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 and carp(4)-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_carp == NULL || 180 carp_master_p(&ia->ia_ifa)) && 181 ia->ia_ifa.ifa_ifp->if_fib == fib) 182 return (true); 183 184 return (false); 185 } 186 187 /* 188 * Return 1 if an internet address is configured on an interface. 189 */ 190 int 191 in_ifhasaddr(struct ifnet *ifp, struct in_addr in) 192 { 193 struct ifaddr *ifa; 194 struct in_ifaddr *ia; 195 196 NET_EPOCH_ASSERT(); 197 198 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 199 if (ifa->ifa_addr->sa_family != AF_INET) 200 continue; 201 ia = (struct in_ifaddr *)ifa; 202 if (ia->ia_addr.sin_addr.s_addr == in.s_addr) 203 return (1); 204 } 205 206 return (0); 207 } 208 209 /* 210 * Return a reference to the interface address which is different to 211 * the supplied one but with same IP address value. 212 */ 213 static struct in_ifaddr * 214 in_localip_more(struct in_ifaddr *original_ia) 215 { 216 struct epoch_tracker et; 217 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr; 218 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib; 219 struct in_ifaddr *ia; 220 221 NET_EPOCH_ENTER(et); 222 CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) { 223 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr; 224 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib; 225 if (!V_rt_add_addr_allfibs && (original_fib != fib)) 226 continue; 227 if ((original_ia != ia) && (original_addr == addr)) { 228 ifa_ref(&ia->ia_ifa); 229 NET_EPOCH_EXIT(et); 230 return (ia); 231 } 232 } 233 NET_EPOCH_EXIT(et); 234 235 return (NULL); 236 } 237 238 /* 239 * Tries to find first IPv4 address in the provided fib. 240 * Prefers non-loopback addresses and return loopback IFF 241 * @loopback_ok is set. 242 * 243 * Returns ifa or NULL. 244 */ 245 struct in_ifaddr * 246 in_findlocal(uint32_t fibnum, bool loopback_ok) 247 { 248 struct in_ifaddr *ia = NULL, *ia_lo = NULL; 249 250 NET_EPOCH_ASSERT(); 251 252 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 253 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib; 254 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib)) 255 continue; 256 257 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr))) 258 break; 259 if (loopback_ok) 260 ia_lo = ia; 261 } 262 263 if (ia == NULL) 264 ia = ia_lo; 265 266 return (ia); 267 } 268 269 /* 270 * Determine whether an IP address is in a reserved set of addresses 271 * that may not be forwarded, or whether datagrams to that destination 272 * may be forwarded. 273 */ 274 int 275 in_canforward(struct in_addr in) 276 { 277 u_long i = ntohl(in.s_addr); 278 279 if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i)) 280 return (0); 281 if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240) 282 return (0); 283 if (IN_ZERONET(i) && !V_ip_allow_net0) 284 return (0); 285 return (1); 286 } 287 288 /* 289 * Sysctl to manage prefix of reserved loopback network; translate 290 * to/from mask. The mask is always contiguous high-order 1 bits 291 * followed by all 0 bits. 292 */ 293 static int 294 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS) 295 { 296 int error, preflen; 297 298 /* ffs is 1-based; compensate. */ 299 preflen = 33 - ffs(V_in_loopback_mask); 300 error = sysctl_handle_int(oidp, &preflen, 0, req); 301 if (error || !req->newptr) 302 return (error); 303 if (preflen < 8 || preflen > 31) 304 return (EINVAL); 305 V_in_loopback_mask = 0xffffffff << (32 - preflen); 306 return (0); 307 } 308 309 /* 310 * Trim a mask in a sockaddr 311 */ 312 static void 313 in_socktrim(struct sockaddr_in *ap) 314 { 315 char *cplim = (char *) &ap->sin_addr; 316 char *cp = (char *) (&ap->sin_addr + 1); 317 318 ap->sin_len = 0; 319 while (--cp >= cplim) 320 if (*cp) { 321 (ap)->sin_len = cp - (char *) (ap) + 1; 322 break; 323 } 324 } 325 326 /* 327 * Generic internet control operations (ioctl's). 328 */ 329 int 330 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp, 331 struct ucred *cred) 332 { 333 struct ifreq *ifr = (struct ifreq *)data; 334 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr; 335 struct epoch_tracker et; 336 struct ifaddr *ifa; 337 struct in_ifaddr *ia; 338 int error; 339 340 if (ifp == NULL) 341 return (EADDRNOTAVAIL); 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 int 445 in_mask2len(struct in_addr *mask) 446 { 447 int x, y; 448 u_char *p; 449 450 p = (u_char *)mask; 451 for (x = 0; x < sizeof(*mask); x++) { 452 if (p[x] != 0xff) 453 break; 454 } 455 y = 0; 456 if (x < sizeof(*mask)) { 457 for (y = 0; y < 8; y++) { 458 if ((p[x] & (0x80 >> y)) == 0) 459 break; 460 } 461 } 462 return (x * 8 + y); 463 } 464 465 int 466 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp, 467 struct thread *td) 468 { 469 return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL)); 470 } 471 472 static int 473 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 474 { 475 const struct in_aliasreq *ifra = (struct in_aliasreq *)data; 476 const struct sockaddr_in *addr = &ifra->ifra_addr; 477 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr; 478 const struct sockaddr_in *mask = &ifra->ifra_mask; 479 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr; 480 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0; 481 struct epoch_tracker et; 482 struct ifaddr *ifa; 483 struct in_ifaddr *ia; 484 bool iaIsFirst; 485 int error = 0; 486 487 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR); 488 if (error) 489 return (error); 490 491 /* 492 * ifra_addr must be present and be of INET family. 493 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional. 494 */ 495 if (addr->sin_len != sizeof(struct sockaddr_in) || 496 addr->sin_family != AF_INET) 497 return (EINVAL); 498 if (broadaddr->sin_len != 0 && 499 (broadaddr->sin_len != sizeof(struct sockaddr_in) || 500 broadaddr->sin_family != AF_INET)) 501 return (EINVAL); 502 if (mask->sin_len != 0 && 503 (mask->sin_len != sizeof(struct sockaddr_in) || 504 mask->sin_family != AF_INET)) 505 return (EINVAL); 506 if ((ifp->if_flags & IFF_POINTOPOINT) && 507 (dstaddr->sin_len != sizeof(struct sockaddr_in) || 508 dstaddr->sin_addr.s_addr == INADDR_ANY)) 509 return (EDESTADDRREQ); 510 if (vhid != 0 && carp_attach_p == NULL) 511 return (EPROTONOSUPPORT); 512 513 #ifdef MAC 514 /* Check if a MAC policy disallows setting the IPv4 address. */ 515 error = mac_inet_check_add_addr(cred, &addr->sin_addr, ifp); 516 if (error != 0) 517 return (error); 518 #endif 519 520 /* 521 * See whether address already exist. 522 */ 523 iaIsFirst = true; 524 ia = NULL; 525 NET_EPOCH_ENTER(et); 526 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 527 struct in_ifaddr *it; 528 529 if (ifa->ifa_addr->sa_family != AF_INET) 530 continue; 531 532 it = (struct in_ifaddr *)ifa; 533 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 534 prison_check_ip4(cred, &addr->sin_addr) == 0) 535 ia = it; 536 else 537 iaIsFirst = false; 538 } 539 NET_EPOCH_EXIT(et); 540 541 if (ia != NULL) 542 (void )in_difaddr_ioctl(cmd, data, ifp, cred); 543 544 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK); 545 ia = (struct in_ifaddr *)ifa; 546 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 547 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 548 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 549 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock, 550 CALLOUT_RETURNUNLOCKED); 551 552 ia->ia_ifp = ifp; 553 ia->ia_addr = *addr; 554 if (mask->sin_len != 0) { 555 ia->ia_sockmask = *mask; 556 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 557 } else { 558 in_addr_t i = ntohl(addr->sin_addr.s_addr); 559 560 /* 561 * If netmask isn't supplied, use historical default. 562 * This is deprecated for interfaces other than loopback 563 * or point-to-point; warn in other cases. In the future 564 * we should return an error rather than warning. 565 */ 566 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) 567 printf("%s: set address: WARNING: network mask " 568 "should be specified; using historical default\n", 569 ifp->if_xname); 570 if (IN_CLASSA(i)) 571 ia->ia_subnetmask = IN_CLASSA_NET; 572 else if (IN_CLASSB(i)) 573 ia->ia_subnetmask = IN_CLASSB_NET; 574 else 575 ia->ia_subnetmask = IN_CLASSC_NET; 576 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 577 } 578 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask; 579 in_socktrim(&ia->ia_sockmask); 580 581 if (ifp->if_flags & IFF_BROADCAST) { 582 if (broadaddr->sin_len != 0) { 583 ia->ia_broadaddr = *broadaddr; 584 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) { 585 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 586 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 587 ia->ia_broadaddr.sin_family = AF_INET; 588 } else { 589 ia->ia_broadaddr.sin_addr.s_addr = 590 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 591 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 592 ia->ia_broadaddr.sin_family = AF_INET; 593 } 594 } 595 596 if (ifp->if_flags & IFF_POINTOPOINT) 597 ia->ia_dstaddr = *dstaddr; 598 599 if (vhid != 0) { 600 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 601 if (error) 602 return (error); 603 } 604 605 /* if_addrhead is already referenced by ifa_alloc() */ 606 IF_ADDR_WLOCK(ifp); 607 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 608 IF_ADDR_WUNLOCK(ifp); 609 610 ifa_ref(ifa); /* in_ifaddrhead */ 611 sx_assert(&in_control_sx, SA_XLOCKED); 612 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 613 CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, 614 ia_hash); 615 616 /* 617 * Give the interface a chance to initialize 618 * if this is its first address, 619 * and to validate the address if necessary. 620 */ 621 if (ifp->if_ioctl != NULL) { 622 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 623 if (error) 624 goto fail1; 625 } 626 627 /* 628 * Add route for the network. 629 */ 630 if (vhid == 0) { 631 error = in_addprefix(ia); 632 if (error) 633 goto fail1; 634 } 635 636 /* 637 * Add a loopback route to self. 638 */ 639 if (vhid == 0 && ia_need_loopback_route(ia)) { 640 struct in_ifaddr *eia; 641 642 eia = in_localip_more(ia); 643 644 if (eia == NULL) { 645 error = ifa_add_loopback_route((struct ifaddr *)ia, 646 (struct sockaddr *)&ia->ia_addr); 647 if (error) 648 goto fail2; 649 } else 650 ifa_free(&eia->ia_ifa); 651 } 652 653 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 654 struct in_addr allhosts_addr; 655 struct in_ifinfo *ii; 656 657 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 658 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 659 660 error = in_joingroup(ifp, &allhosts_addr, NULL, 661 &ii->ii_allhosts); 662 } 663 664 /* 665 * Note: we don't need extra reference for ifa, since we called 666 * with sx lock held, and ifaddr can not be deleted in concurrent 667 * thread. 668 */ 669 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD); 670 671 return (error); 672 673 fail2: 674 if (vhid == 0) 675 (void )in_scrubprefix(ia, LLE_STATIC); 676 677 fail1: 678 if (ia->ia_ifa.ifa_carp) 679 (*carp_detach_p)(&ia->ia_ifa, false); 680 681 IF_ADDR_WLOCK(ifp); 682 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 683 IF_ADDR_WUNLOCK(ifp); 684 ifa_free(&ia->ia_ifa); /* if_addrhead */ 685 686 sx_assert(&in_control_sx, SA_XLOCKED); 687 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 688 CK_LIST_REMOVE(ia, ia_hash); 689 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 690 691 return (error); 692 } 693 694 static int 695 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 696 { 697 const struct ifreq *ifr = (struct ifreq *)data; 698 const struct sockaddr_in *addr = (const struct sockaddr_in *) 699 &ifr->ifr_addr; 700 struct ifaddr *ifa; 701 struct in_ifaddr *ia; 702 bool deleteAny, iaIsLast; 703 int error; 704 705 if (cred != NULL) { 706 error = priv_check_cred(cred, PRIV_NET_DELIFADDR); 707 if (error) 708 return (error); 709 } 710 711 if (addr->sin_len != sizeof(struct sockaddr_in) || 712 addr->sin_family != AF_INET) 713 deleteAny = true; 714 else 715 deleteAny = false; 716 717 iaIsLast = true; 718 ia = NULL; 719 IF_ADDR_WLOCK(ifp); 720 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 721 struct in_ifaddr *it; 722 723 if (ifa->ifa_addr->sa_family != AF_INET) 724 continue; 725 726 it = (struct in_ifaddr *)ifa; 727 if (deleteAny && ia == NULL && (cred == NULL || 728 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0)) 729 ia = it; 730 731 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 732 (cred == NULL || prison_check_ip4(cred, 733 &addr->sin_addr) == 0)) 734 ia = it; 735 736 if (it != ia) 737 iaIsLast = false; 738 } 739 740 if (ia == NULL) { 741 IF_ADDR_WUNLOCK(ifp); 742 return (EADDRNOTAVAIL); 743 } 744 745 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 746 IF_ADDR_WUNLOCK(ifp); 747 ifa_free(&ia->ia_ifa); /* if_addrhead */ 748 749 sx_assert(&in_control_sx, SA_XLOCKED); 750 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 751 CK_LIST_REMOVE(ia, ia_hash); 752 753 /* 754 * in_scrubprefix() kills the interface route. 755 */ 756 in_scrubprefix(ia, LLE_STATIC); 757 758 /* 759 * in_ifadown gets rid of all the rest of 760 * the routes. This is not quite the right 761 * thing to do, but at least if we are running 762 * a routing process they will come back. 763 */ 764 in_ifadown(&ia->ia_ifa, 1); 765 766 if (ia->ia_ifa.ifa_carp) 767 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR); 768 769 /* 770 * If this is the last IPv4 address configured on this 771 * interface, leave the all-hosts group. 772 * No state-change report need be transmitted. 773 */ 774 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 775 struct in_ifinfo *ii; 776 777 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 778 if (ii->ii_allhosts) { 779 (void)in_leavegroup(ii->ii_allhosts, NULL); 780 ii->ii_allhosts = NULL; 781 } 782 } 783 784 IF_ADDR_WLOCK(ifp); 785 if (callout_stop(&ia->ia_garp_timer) == 1) { 786 ifa_free(&ia->ia_ifa); 787 } 788 IF_ADDR_WUNLOCK(ifp); 789 790 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa, 791 IFADDR_EVENT_DEL); 792 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 793 794 return (0); 795 } 796 797 static int 798 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 799 { 800 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 801 const struct sockaddr_in *addr = &ifra->ifra_addr; 802 struct epoch_tracker et; 803 struct ifaddr *ifa; 804 struct in_ifaddr *ia; 805 806 /* 807 * ifra_addr must be present and be of INET family. 808 */ 809 if (addr->sin_len != sizeof(struct sockaddr_in) || 810 addr->sin_family != AF_INET) 811 return (EINVAL); 812 813 /* 814 * See whether address exist. 815 */ 816 ia = NULL; 817 NET_EPOCH_ENTER(et); 818 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 819 struct in_ifaddr *it; 820 821 if (ifa->ifa_addr->sa_family != AF_INET) 822 continue; 823 824 it = (struct in_ifaddr *)ifa; 825 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 826 prison_check_ip4(cred, &addr->sin_addr) == 0) { 827 ia = it; 828 break; 829 } 830 } 831 if (ia == NULL) { 832 NET_EPOCH_EXIT(et); 833 return (EADDRNOTAVAIL); 834 } 835 836 ifra->ifra_mask = ia->ia_sockmask; 837 if ((ifp->if_flags & IFF_POINTOPOINT) && 838 ia->ia_dstaddr.sin_family == AF_INET) 839 ifra->ifra_dstaddr = ia->ia_dstaddr; 840 else if ((ifp->if_flags & IFF_BROADCAST) && 841 ia->ia_broadaddr.sin_family == AF_INET) 842 ifra->ifra_broadaddr = ia->ia_broadaddr; 843 else 844 memset(&ifra->ifra_broadaddr, 0, 845 sizeof(ifra->ifra_broadaddr)); 846 847 NET_EPOCH_EXIT(et); 848 return (0); 849 } 850 851 static int 852 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg) 853 { 854 855 if (nh->nh_ifa == (struct ifaddr *)arg) 856 return (1); 857 858 return (0); 859 } 860 861 static int 862 in_handle_prefix_route(uint32_t fibnum, int cmd, 863 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa, 864 struct ifnet *ifp) 865 { 866 867 NET_EPOCH_ASSERT(); 868 869 /* Prepare gateway */ 870 struct sockaddr_dl_short sdl = { 871 .sdl_family = AF_LINK, 872 .sdl_len = sizeof(struct sockaddr_dl_short), 873 .sdl_type = ifa->ifa_ifp->if_type, 874 .sdl_index = ifa->ifa_ifp->if_index, 875 }; 876 877 struct rt_addrinfo info = { 878 .rti_ifa = ifa, 879 .rti_ifp = ifp, 880 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST), 881 .rti_info = { 882 [RTAX_DST] = (struct sockaddr *)dst, 883 [RTAX_NETMASK] = (struct sockaddr *)netmask, 884 [RTAX_GATEWAY] = (struct sockaddr *)&sdl, 885 }, 886 /* Ensure we delete the prefix IFF prefix ifa matches */ 887 .rti_filter = in_match_ifaddr, 888 .rti_filterdata = ifa, 889 }; 890 891 return (rib_handle_ifaddr_info(fibnum, cmd, &info)); 892 } 893 894 /* 895 * Routing table interaction with interface addresses. 896 * 897 * In general, two types of routes needs to be installed: 898 * a) "interface" or "prefix" route, telling user that the addresses 899 * behind the ifa prefix are reached directly. 900 * b) "loopback" route installed for the ifa address, telling user that 901 * the address belongs to local system. 902 * 903 * Handling for (a) and (b) differs in multi-fib aspects, hence they 904 * are implemented in different functions below. 905 * 906 * The cases above may intersect - /32 interface aliases results in 907 * the same prefix produced by (a) and (b). This blurs the definition 908 * of the "loopback" route and complicate interactions. The interaction 909 * table is defined below. The case numbers are used in the multiple 910 * functions below to refer to the particular test case. 911 * 912 * There can be multiple options: 913 * 1) Adding address with prefix on non-p2p/non-loopback interface. 914 * Example: 192.0.2.1/24. Action: 915 * * add "prefix" route towards 192.0.2.0/24 via @ia interface, 916 * using @ia as an address source. 917 * * add "loopback" route towards 192.0.2.1 via V_loif, saving 918 * @ia ifp in the gateway and using @ia as an address source. 919 * 920 * 2) Adding address with /32 mask to non-p2p/non-loopback interface. 921 * Example: 192.0.2.2/32. Action: 922 * * add "prefix" host route via V_loif, using @ia as an address source. 923 * 924 * 3) Adding address with or without prefix to p2p interface. 925 * Example: 10.0.0.1/24->10.0.0.2. Action: 926 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia 927 * as an address source. Note: no sense in installing full /24 as the interface 928 * is point-to-point. 929 * * add "loopback" route towards 10.0.9.1 via V_loif, saving 930 * @ia ifp in the gateway and using @ia as an address source. 931 * 932 * 4) Adding address with or without prefix to loopback interface. 933 * Example: 192.0.2.1/24. Action: 934 * * add "prefix" host route via @ia interface, using @ia as an address source. 935 * Note: Skip installing /24 prefix as it would introduce TTL loop 936 * for the traffic destined to these addresses. 937 */ 938 939 /* 940 * Checks if @ia needs to install loopback route to @ia address via 941 * ifa_maintain_loopback_route(). 942 * 943 * Return true on success. 944 */ 945 static bool 946 ia_need_loopback_route(const struct in_ifaddr *ia) 947 { 948 struct ifnet *ifp = ia->ia_ifp; 949 950 /* Case 4: Skip loopback interfaces */ 951 if ((ifp->if_flags & IFF_LOOPBACK) || 952 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)) 953 return (false); 954 955 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */ 956 if ((ifp->if_flags & IFF_POINTOPOINT) && 957 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr) 958 return (false); 959 960 /* Case 2: skip /32 prefixes */ 961 if (!(ifp->if_flags & IFF_POINTOPOINT) && 962 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST)) 963 return (false); 964 965 return (true); 966 } 967 968 /* 969 * Calculate "prefix" route corresponding to @ia. 970 */ 971 static void 972 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask) 973 { 974 975 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) { 976 /* Case 3: return host route for dstaddr */ 977 *prefix = ia->ia_dstaddr.sin_addr; 978 mask->s_addr = INADDR_BROADCAST; 979 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) { 980 /* Case 4: return host route for ifaddr */ 981 *prefix = ia->ia_addr.sin_addr; 982 mask->s_addr = INADDR_BROADCAST; 983 } else { 984 /* Cases 1,2: return actual ia prefix */ 985 *prefix = ia->ia_addr.sin_addr; 986 *mask = ia->ia_sockmask.sin_addr; 987 prefix->s_addr &= mask->s_addr; 988 } 989 } 990 991 /* 992 * Adds or delete interface "prefix" route corresponding to @ifa. 993 * Returns 0 on success or errno. 994 */ 995 static int 996 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia) 997 { 998 struct ifaddr *ifa = &ia->ia_ifa; 999 struct in_addr daddr, maddr; 1000 struct sockaddr_in *pmask; 1001 struct epoch_tracker et; 1002 int error; 1003 1004 ia_getrtprefix(ia, &daddr, &maddr); 1005 1006 struct sockaddr_in mask = { 1007 .sin_family = AF_INET, 1008 .sin_len = sizeof(struct sockaddr_in), 1009 .sin_addr = maddr, 1010 }; 1011 1012 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL; 1013 1014 struct sockaddr_in dst = { 1015 .sin_family = AF_INET, 1016 .sin_len = sizeof(struct sockaddr_in), 1017 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr, 1018 }; 1019 1020 struct ifnet *ifp = ia->ia_ifp; 1021 1022 if ((maddr.s_addr == INADDR_BROADCAST) && 1023 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) { 1024 /* Case 2: host route on broadcast interface */ 1025 ifp = V_loif; 1026 } 1027 1028 uint32_t fibnum = ifa->ifa_ifp->if_fib; 1029 NET_EPOCH_ENTER(et); 1030 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp); 1031 NET_EPOCH_EXIT(et); 1032 1033 return (error); 1034 } 1035 1036 /* 1037 * Check if we have a route for the given prefix already. 1038 */ 1039 static bool 1040 in_hasrtprefix(struct in_ifaddr *target) 1041 { 1042 struct epoch_tracker et; 1043 struct in_ifaddr *ia; 1044 struct in_addr prefix, mask, p, m; 1045 bool result = false; 1046 1047 ia_getrtprefix(target, &prefix, &mask); 1048 1049 /* Look for an existing address with the same prefix, mask, and fib */ 1050 NET_EPOCH_ENTER(et); 1051 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1052 ia_getrtprefix(ia, &p, &m); 1053 1054 if (prefix.s_addr != p.s_addr || 1055 mask.s_addr != m.s_addr) 1056 continue; 1057 1058 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 1059 continue; 1060 1061 /* 1062 * If we got a matching prefix route inserted by other 1063 * interface address, we are done here. 1064 */ 1065 if (ia->ia_flags & IFA_ROUTE) { 1066 result = true; 1067 break; 1068 } 1069 } 1070 NET_EPOCH_EXIT(et); 1071 1072 return (result); 1073 } 1074 1075 int 1076 in_addprefix(struct in_ifaddr *target) 1077 { 1078 int error; 1079 1080 if (in_hasrtprefix(target)) { 1081 if (V_nosameprefix) 1082 return (EEXIST); 1083 else { 1084 rt_addrmsg(RTM_ADD, &target->ia_ifa, 1085 target->ia_ifp->if_fib); 1086 return (0); 1087 } 1088 } 1089 1090 /* 1091 * No-one seem to have this prefix route, so we try to insert it. 1092 */ 1093 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib); 1094 error = in_handle_ifaddr_route(RTM_ADD, target); 1095 if (!error) 1096 target->ia_flags |= IFA_ROUTE; 1097 return (error); 1098 } 1099 1100 /* 1101 * Removes either all lle entries for given @ia, or lle 1102 * corresponding to @ia address. 1103 */ 1104 static void 1105 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags) 1106 { 1107 struct sockaddr_in addr, mask; 1108 struct sockaddr *saddr, *smask; 1109 struct ifnet *ifp; 1110 1111 saddr = (struct sockaddr *)&addr; 1112 bzero(&addr, sizeof(addr)); 1113 addr.sin_len = sizeof(addr); 1114 addr.sin_family = AF_INET; 1115 smask = (struct sockaddr *)&mask; 1116 bzero(&mask, sizeof(mask)); 1117 mask.sin_len = sizeof(mask); 1118 mask.sin_family = AF_INET; 1119 mask.sin_addr.s_addr = ia->ia_subnetmask; 1120 ifp = ia->ia_ifp; 1121 1122 if (all) { 1123 /* 1124 * Remove all L2 entries matching given prefix. 1125 * Convert address to host representation to avoid 1126 * doing this on every callback. ia_subnetmask is already 1127 * stored in host representation. 1128 */ 1129 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr); 1130 lltable_prefix_free(AF_INET, saddr, smask, flags); 1131 } else { 1132 /* Remove interface address only */ 1133 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr; 1134 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr); 1135 } 1136 } 1137 1138 /* 1139 * If there is no other address in the system that can serve a route to the 1140 * same prefix, remove the route. Hand over the route to the new address 1141 * otherwise. 1142 */ 1143 int 1144 in_scrubprefix(struct in_ifaddr *target, u_int flags) 1145 { 1146 struct epoch_tracker et; 1147 struct in_ifaddr *ia; 1148 struct in_addr prefix, mask, p, m; 1149 int error = 0; 1150 1151 /* 1152 * Remove the loopback route to the interface address. 1153 */ 1154 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) { 1155 struct in_ifaddr *eia; 1156 1157 eia = in_localip_more(target); 1158 1159 if (eia != NULL) { 1160 error = ifa_switch_loopback_route((struct ifaddr *)eia, 1161 (struct sockaddr *)&target->ia_addr); 1162 ifa_free(&eia->ia_ifa); 1163 } else { 1164 error = ifa_del_loopback_route((struct ifaddr *)target, 1165 (struct sockaddr *)&target->ia_addr); 1166 } 1167 } 1168 1169 ia_getrtprefix(target, &prefix, &mask); 1170 1171 if ((target->ia_flags & IFA_ROUTE) == 0) { 1172 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib); 1173 1174 /* 1175 * Removing address from !IFF_UP interface or 1176 * prefix which exists on other interface (along with route). 1177 * No entries should exist here except target addr. 1178 * Given that, delete this entry only. 1179 */ 1180 in_scrubprefixlle(target, 0, flags); 1181 return (0); 1182 } 1183 1184 NET_EPOCH_ENTER(et); 1185 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1186 ia_getrtprefix(ia, &p, &m); 1187 1188 if (prefix.s_addr != p.s_addr || 1189 mask.s_addr != m.s_addr) 1190 continue; 1191 1192 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 1193 continue; 1194 1195 /* 1196 * If we got a matching prefix address, move IFA_ROUTE and 1197 * the route itself to it. Make sure that routing daemons 1198 * get a heads-up. 1199 */ 1200 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1201 ifa_ref(&ia->ia_ifa); 1202 NET_EPOCH_EXIT(et); 1203 error = in_handle_ifaddr_route(RTM_DELETE, target); 1204 if (error == 0) 1205 target->ia_flags &= ~IFA_ROUTE; 1206 else 1207 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 1208 error); 1209 /* Scrub all entries IFF interface is different */ 1210 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp, 1211 flags); 1212 error = in_handle_ifaddr_route(RTM_ADD, ia); 1213 if (error == 0) 1214 ia->ia_flags |= IFA_ROUTE; 1215 else 1216 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 1217 error); 1218 ifa_free(&ia->ia_ifa); 1219 return (error); 1220 } 1221 } 1222 NET_EPOCH_EXIT(et); 1223 1224 /* 1225 * remove all L2 entries on the given prefix 1226 */ 1227 in_scrubprefixlle(target, 1, flags); 1228 1229 /* 1230 * As no-one seem to have this prefix, we can remove the route. 1231 */ 1232 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib); 1233 error = in_handle_ifaddr_route(RTM_DELETE, target); 1234 if (error == 0) 1235 target->ia_flags &= ~IFA_ROUTE; 1236 else 1237 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 1238 return (error); 1239 } 1240 1241 void 1242 in_ifscrub_all(void) 1243 { 1244 struct ifnet *ifp; 1245 struct ifaddr *ifa, *nifa; 1246 struct ifreq ifr; 1247 1248 IFNET_RLOCK(); 1249 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1250 /* Cannot lock here - lock recursion. */ 1251 /* NET_EPOCH_ENTER(et); */ 1252 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 1253 if (ifa->ifa_addr->sa_family != AF_INET) 1254 continue; 1255 1256 /* 1257 * This is ugly but the only way for legacy IP to 1258 * cleanly remove addresses and everything attached. 1259 */ 1260 bzero(&ifr, sizeof(ifr)); 1261 ifr.ifr_addr = *ifa->ifa_addr; 1262 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, 1263 ifp, NULL); 1264 } 1265 /* NET_EPOCH_EXIT(et); */ 1266 in_purgemaddrs(ifp); 1267 igmp_domifdetach(ifp); 1268 } 1269 IFNET_RUNLOCK(); 1270 } 1271 1272 int 1273 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia) 1274 { 1275 1276 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 1277 /* 1278 * Optionally check for old-style (host 0) broadcast, but 1279 * taking into account that RFC 3021 obsoletes it. 1280 */ 1281 (V_broadcast_lowest && ia->ia_subnetmask != IN_RFC3021_MASK && 1282 ntohl(in.s_addr) == ia->ia_subnet)) && 1283 /* 1284 * Check for an all one subnetmask. These 1285 * only exist when an interface gets a secondary 1286 * address. 1287 */ 1288 ia->ia_subnetmask != (u_long)0xffffffff); 1289 } 1290 1291 /* 1292 * Return 1 if the address might be a local broadcast address. 1293 */ 1294 int 1295 in_broadcast(struct in_addr in, struct ifnet *ifp) 1296 { 1297 struct ifaddr *ifa; 1298 int found; 1299 1300 NET_EPOCH_ASSERT(); 1301 1302 if (in.s_addr == INADDR_BROADCAST || 1303 in.s_addr == INADDR_ANY) 1304 return (1); 1305 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1306 return (0); 1307 found = 0; 1308 /* 1309 * Look through the list of addresses for a match 1310 * with a broadcast address. 1311 */ 1312 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1313 if (ifa->ifa_addr->sa_family == AF_INET && 1314 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) { 1315 found = 1; 1316 break; 1317 } 1318 return (found); 1319 } 1320 1321 /* 1322 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 1323 */ 1324 void 1325 in_ifdetach(struct ifnet *ifp) 1326 { 1327 IN_MULTI_LOCK(); 1328 in_pcbpurgeif0(&V_ripcbinfo, ifp); 1329 in_pcbpurgeif0(&V_udbinfo, ifp); 1330 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 1331 in_purgemaddrs(ifp); 1332 IN_MULTI_UNLOCK(); 1333 1334 /* 1335 * Make sure all multicast deletions invoking if_ioctl() are 1336 * completed before returning. Else we risk accessing a freed 1337 * ifnet structure pointer. 1338 */ 1339 inm_release_wait(NULL); 1340 } 1341 1342 static void 1343 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event) 1344 { 1345 struct epoch_tracker et; 1346 struct ifaddr *ifa; 1347 struct in_ifaddr *ia; 1348 int error; 1349 1350 NET_EPOCH_ENTER(et); 1351 switch (event) { 1352 case IFNET_EVENT_DOWN: 1353 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1354 if (ifa->ifa_addr->sa_family != AF_INET) 1355 continue; 1356 ia = (struct in_ifaddr *)ifa; 1357 if ((ia->ia_flags & IFA_ROUTE) == 0) 1358 continue; 1359 ifa_ref(ifa); 1360 /* 1361 * in_scrubprefix() kills the interface route. 1362 */ 1363 in_scrubprefix(ia, 0); 1364 /* 1365 * in_ifadown gets rid of all the rest of the 1366 * routes. This is not quite the right thing 1367 * to do, but at least if we are running a 1368 * routing process they will come back. 1369 */ 1370 in_ifadown(ifa, 0); 1371 ifa_free(ifa); 1372 } 1373 break; 1374 1375 case IFNET_EVENT_UP: 1376 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1377 if (ifa->ifa_addr->sa_family != AF_INET) 1378 continue; 1379 ia = (struct in_ifaddr *)ifa; 1380 if (ia->ia_flags & IFA_ROUTE) 1381 continue; 1382 ifa_ref(ifa); 1383 error = ifa_del_loopback_route(ifa, ifa->ifa_addr); 1384 rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib); 1385 error = in_handle_ifaddr_route(RTM_ADD, ia); 1386 if (error == 0) 1387 ia->ia_flags |= IFA_ROUTE; 1388 error = ifa_add_loopback_route(ifa, ifa->ifa_addr); 1389 ifa_free(ifa); 1390 } 1391 break; 1392 } 1393 NET_EPOCH_EXIT(et); 1394 } 1395 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY); 1396 1397 /* 1398 * Delete all IPv4 multicast address records, and associated link-layer 1399 * multicast address records, associated with ifp. 1400 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1401 * XXX This should not race with ifma_protospec being set during 1402 * a new allocation, if it does, we have bigger problems. 1403 */ 1404 static void 1405 in_purgemaddrs(struct ifnet *ifp) 1406 { 1407 struct epoch_tracker et; 1408 struct in_multi_head purgeinms; 1409 struct in_multi *inm; 1410 struct ifmultiaddr *ifma; 1411 1412 SLIST_INIT(&purgeinms); 1413 IN_MULTI_LIST_LOCK(); 1414 1415 /* 1416 * Extract list of in_multi associated with the detaching ifp 1417 * which the PF_INET layer is about to release. 1418 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1419 * by code further down. 1420 */ 1421 IF_ADDR_WLOCK(ifp); 1422 NET_EPOCH_ENTER(et); 1423 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1424 inm = inm_ifmultiaddr_get_inm(ifma); 1425 if (inm == NULL) 1426 continue; 1427 inm_rele_locked(&purgeinms, inm); 1428 } 1429 NET_EPOCH_EXIT(et); 1430 IF_ADDR_WUNLOCK(ifp); 1431 1432 inm_release_list_deferred(&purgeinms); 1433 igmp_ifdetach(ifp); 1434 IN_MULTI_LIST_UNLOCK(); 1435 } 1436 1437 struct in_llentry { 1438 struct llentry base; 1439 }; 1440 1441 #define IN_LLTBL_DEFAULT_HSIZE 32 1442 #define IN_LLTBL_HASH(k, h) \ 1443 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1444 1445 /* 1446 * Do actual deallocation of @lle. 1447 */ 1448 static void 1449 in_lltable_destroy_lle_unlocked(epoch_context_t ctx) 1450 { 1451 struct llentry *lle; 1452 1453 lle = __containerof(ctx, struct llentry, lle_epoch_ctx); 1454 LLE_LOCK_DESTROY(lle); 1455 LLE_REQ_DESTROY(lle); 1456 free(lle, M_LLTABLE); 1457 } 1458 1459 /* 1460 * Called by LLE_FREE_LOCKED when number of references 1461 * drops to zero. 1462 */ 1463 static void 1464 in_lltable_destroy_lle(struct llentry *lle) 1465 { 1466 1467 LLE_WUNLOCK(lle); 1468 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx); 1469 } 1470 1471 static struct llentry * 1472 in_lltable_new(struct in_addr addr4, u_int flags) 1473 { 1474 struct in_llentry *lle; 1475 1476 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1477 if (lle == NULL) /* NB: caller generates msg */ 1478 return NULL; 1479 1480 /* 1481 * For IPv4 this will trigger "arpresolve" to generate 1482 * an ARP request. 1483 */ 1484 lle->base.la_expire = time_uptime; /* mark expired */ 1485 lle->base.r_l3addr.addr4 = addr4; 1486 lle->base.lle_refcnt = 1; 1487 lle->base.lle_free = in_lltable_destroy_lle; 1488 LLE_LOCK_INIT(&lle->base); 1489 LLE_REQ_INIT(&lle->base); 1490 callout_init(&lle->base.lle_timer, 1); 1491 1492 return (&lle->base); 1493 } 1494 1495 static int 1496 in_lltable_match_prefix(const struct sockaddr *saddr, 1497 const struct sockaddr *smask, u_int flags, struct llentry *lle) 1498 { 1499 struct in_addr addr, mask, lle_addr; 1500 1501 addr = ((const struct sockaddr_in *)saddr)->sin_addr; 1502 mask = ((const struct sockaddr_in *)smask)->sin_addr; 1503 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1504 1505 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0) 1506 return (0); 1507 1508 if (lle->la_flags & LLE_IFADDR) { 1509 /* 1510 * Delete LLE_IFADDR records IFF address & flag matches. 1511 * Note that addr is the interface address within prefix 1512 * being matched. 1513 * Note also we should handle 'ifdown' cases without removing 1514 * ifaddr macs. 1515 */ 1516 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0) 1517 return (1); 1518 return (0); 1519 } 1520 1521 /* flags & LLE_STATIC means deleting both dynamic and static entries */ 1522 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)) 1523 return (1); 1524 1525 return (0); 1526 } 1527 1528 static void 1529 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1530 { 1531 size_t pkts_dropped; 1532 1533 LLE_WLOCK_ASSERT(lle); 1534 KASSERT(llt != NULL, ("lltable is NULL")); 1535 1536 /* Unlink entry from table if not already */ 1537 if ((lle->la_flags & LLE_LINKED) != 0) { 1538 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 1539 lltable_unlink_entry(llt, lle); 1540 } 1541 1542 /* Drop hold queue */ 1543 pkts_dropped = llentry_free(lle); 1544 ARPSTAT_ADD(dropped, pkts_dropped); 1545 } 1546 1547 static int 1548 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1549 { 1550 struct nhop_object *nh; 1551 struct in_addr addr; 1552 1553 KASSERT(l3addr->sa_family == AF_INET, 1554 ("sin_family %d", l3addr->sa_family)); 1555 1556 addr = ((const struct sockaddr_in *)l3addr)->sin_addr; 1557 1558 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0); 1559 if (nh == NULL) 1560 return (EINVAL); 1561 1562 /* 1563 * If the gateway for an existing host route matches the target L3 1564 * address, which is a special route inserted by some implementation 1565 * such as MANET, and the interface is of the correct type, then 1566 * allow for ARP to proceed. 1567 */ 1568 if (nh->nh_flags & NHF_GATEWAY) { 1569 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER || 1570 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1571 memcmp(nh->gw_sa.sa_data, l3addr->sa_data, 1572 sizeof(in_addr_t)) != 0) { 1573 return (EINVAL); 1574 } 1575 } 1576 1577 /* 1578 * Make sure that at least the destination address is covered 1579 * by the route. This is for handling the case where 2 or more 1580 * interfaces have the same prefix. An incoming packet arrives 1581 * on one interface and the corresponding outgoing packet leaves 1582 * another interface. 1583 */ 1584 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) { 1585 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp); 1586 struct in_addr dst_addr, mask_addr; 1587 1588 if (ia == NULL) 1589 return (EINVAL); 1590 1591 /* 1592 * ifaof_ifpforaddr() returns _best matching_ IFA. 1593 * It is possible that ifa prefix does not cover our address. 1594 * Explicitly verify and fail if that's the case. 1595 */ 1596 dst_addr = IA_SIN(ia)->sin_addr; 1597 mask_addr.s_addr = htonl(ia->ia_subnetmask); 1598 1599 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr)) 1600 return (EINVAL); 1601 } 1602 1603 return (0); 1604 } 1605 1606 static inline uint32_t 1607 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 1608 { 1609 1610 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 1611 } 1612 1613 static uint32_t 1614 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 1615 { 1616 1617 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 1618 } 1619 1620 static void 1621 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 1622 { 1623 struct sockaddr_in *sin; 1624 1625 sin = (struct sockaddr_in *)sa; 1626 bzero(sin, sizeof(*sin)); 1627 sin->sin_family = AF_INET; 1628 sin->sin_len = sizeof(*sin); 1629 sin->sin_addr = lle->r_l3addr.addr4; 1630 } 1631 1632 static inline struct llentry * 1633 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 1634 { 1635 struct llentry *lle; 1636 struct llentries *lleh; 1637 u_int hashidx; 1638 1639 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 1640 lleh = &llt->lle_head[hashidx]; 1641 CK_LIST_FOREACH(lle, lleh, lle_next) { 1642 if (lle->la_flags & LLE_DELETED) 1643 continue; 1644 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 1645 break; 1646 } 1647 1648 return (lle); 1649 } 1650 1651 static void 1652 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle) 1653 { 1654 1655 lle->la_flags |= LLE_DELETED; 1656 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1657 #ifdef DIAGNOSTIC 1658 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1659 #endif 1660 llentry_free(lle); 1661 } 1662 1663 static struct llentry * 1664 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1665 { 1666 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1667 struct ifnet *ifp = llt->llt_ifp; 1668 struct llentry *lle; 1669 char linkhdr[LLE_MAX_LINKHDR]; 1670 size_t linkhdrsize; 1671 int lladdr_off; 1672 1673 KASSERT(l3addr->sa_family == AF_INET, 1674 ("sin_family %d", l3addr->sa_family)); 1675 1676 /* 1677 * A route that covers the given address must have 1678 * been installed 1st because we are doing a resolution, 1679 * verify this. 1680 */ 1681 if (!(flags & LLE_IFADDR) && 1682 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1683 return (NULL); 1684 1685 lle = in_lltable_new(sin->sin_addr, flags); 1686 if (lle == NULL) { 1687 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1688 return (NULL); 1689 } 1690 lle->la_flags = flags; 1691 if (flags & LLE_STATIC) 1692 lle->r_flags |= RLLE_VALID; 1693 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 1694 linkhdrsize = LLE_MAX_LINKHDR; 1695 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp), 1696 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1697 in_lltable_free_entry(llt, lle); 1698 return (NULL); 1699 } 1700 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 1701 lladdr_off); 1702 lle->la_flags |= LLE_STATIC; 1703 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR); 1704 lle->la_expire = 0; 1705 } 1706 1707 return (lle); 1708 } 1709 1710 /* 1711 * Return NULL if not found or marked for deletion. 1712 * If found return lle read locked. 1713 */ 1714 static struct llentry * 1715 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1716 { 1717 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1718 struct llentry *lle; 1719 1720 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 1721 KASSERT(l3addr->sa_family == AF_INET, 1722 ("sin_family %d", l3addr->sa_family)); 1723 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) != 1724 (LLE_UNLOCKED | LLE_EXCLUSIVE), 1725 ("wrong lle request flags: %#x", flags)); 1726 1727 lle = in_lltable_find_dst(llt, sin->sin_addr); 1728 if (lle == NULL) 1729 return (NULL); 1730 if (flags & LLE_UNLOCKED) 1731 return (lle); 1732 1733 if (flags & LLE_EXCLUSIVE) 1734 LLE_WLOCK(lle); 1735 else 1736 LLE_RLOCK(lle); 1737 1738 /* 1739 * If the afdata lock is not held, the LLE may have been unlinked while 1740 * we were blocked on the LLE lock. Check for this case. 1741 */ 1742 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) { 1743 if (flags & LLE_EXCLUSIVE) 1744 LLE_WUNLOCK(lle); 1745 else 1746 LLE_RUNLOCK(lle); 1747 return (NULL); 1748 } 1749 return (lle); 1750 } 1751 1752 static int 1753 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 1754 struct sysctl_req *wr) 1755 { 1756 struct ifnet *ifp = llt->llt_ifp; 1757 /* XXX stack use */ 1758 struct { 1759 struct rt_msghdr rtm; 1760 struct sockaddr_in sin; 1761 struct sockaddr_dl sdl; 1762 } arpc; 1763 struct sockaddr_dl *sdl; 1764 int error; 1765 1766 bzero(&arpc, sizeof(arpc)); 1767 /* skip deleted entries */ 1768 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1769 return (0); 1770 /* Skip if jailed and not a valid IP of the prison. */ 1771 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin); 1772 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0) 1773 return (0); 1774 /* 1775 * produce a msg made of: 1776 * struct rt_msghdr; 1777 * struct sockaddr_in; (IPv4) 1778 * struct sockaddr_dl; 1779 */ 1780 arpc.rtm.rtm_msglen = sizeof(arpc); 1781 arpc.rtm.rtm_version = RTM_VERSION; 1782 arpc.rtm.rtm_type = RTM_GET; 1783 arpc.rtm.rtm_flags = RTF_UP; 1784 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1785 1786 /* publish */ 1787 if (lle->la_flags & LLE_PUB) 1788 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1789 1790 sdl = &arpc.sdl; 1791 sdl->sdl_family = AF_LINK; 1792 sdl->sdl_len = sizeof(*sdl); 1793 sdl->sdl_index = ifp->if_index; 1794 sdl->sdl_type = ifp->if_type; 1795 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1796 sdl->sdl_alen = ifp->if_addrlen; 1797 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1798 } else { 1799 sdl->sdl_alen = 0; 1800 bzero(LLADDR(sdl), ifp->if_addrlen); 1801 } 1802 1803 arpc.rtm.rtm_rmx.rmx_expire = 1804 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1805 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1806 if (lle->la_flags & LLE_STATIC) 1807 arpc.rtm.rtm_flags |= RTF_STATIC; 1808 if (lle->la_flags & LLE_IFADDR) 1809 arpc.rtm.rtm_flags |= RTF_PINNED; 1810 arpc.rtm.rtm_index = ifp->if_index; 1811 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1812 1813 return (error); 1814 } 1815 1816 static void 1817 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle) 1818 { 1819 struct ifnet *ifp = llt->llt_ifp; 1820 1821 /* gratuitous ARP */ 1822 if ((lle->la_flags & LLE_PUB) != 0) 1823 arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4, 1824 lle->ll_addr); 1825 } 1826 1827 static struct lltable * 1828 in_lltattach(struct ifnet *ifp) 1829 { 1830 struct lltable *llt; 1831 1832 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 1833 llt->llt_af = AF_INET; 1834 llt->llt_ifp = ifp; 1835 1836 llt->llt_lookup = in_lltable_lookup; 1837 llt->llt_alloc_entry = in_lltable_alloc; 1838 llt->llt_delete_entry = in_lltable_delete_entry; 1839 llt->llt_dump_entry = in_lltable_dump_entry; 1840 llt->llt_hash = in_lltable_hash; 1841 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 1842 llt->llt_free_entry = in_lltable_free_entry; 1843 llt->llt_match_prefix = in_lltable_match_prefix; 1844 llt->llt_mark_used = llentry_mark_used; 1845 llt->llt_post_resolved = in_lltable_post_resolved; 1846 lltable_link(llt); 1847 1848 return (llt); 1849 } 1850 1851 struct lltable * 1852 in_lltable_get(struct ifnet *ifp) 1853 { 1854 struct lltable *llt = NULL; 1855 1856 void *afdata_ptr = ifp->if_afdata[AF_INET]; 1857 if (afdata_ptr != NULL) 1858 llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt; 1859 return (llt); 1860 } 1861 1862 void * 1863 in_domifattach(struct ifnet *ifp) 1864 { 1865 struct in_ifinfo *ii; 1866 1867 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1868 1869 ii->ii_llt = in_lltattach(ifp); 1870 ii->ii_igmp = igmp_domifattach(ifp); 1871 1872 return (ii); 1873 } 1874 1875 void 1876 in_domifdetach(struct ifnet *ifp, void *aux) 1877 { 1878 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1879 1880 igmp_domifdetach(ifp); 1881 lltable_free(ii->ii_llt); 1882 free(ii, M_IFADDR); 1883 } 1884