1 /*- 2 * Copyright (c) 1982, 1986, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * Copyright (C) 2001 WIDE Project. All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 4. Neither the name of the University nor the names of its contributors 15 * may be used to endorse or promote products derived from this software 16 * without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 * 30 * @(#)in.c 8.4 (Berkeley) 1/9/95 31 */ 32 33 #include <sys/cdefs.h> 34 __FBSDID("$FreeBSD$"); 35 36 #include "opt_mpath.h" 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/proc.h> 48 #include <sys/sysctl.h> 49 #include <sys/syslog.h> 50 #include <sys/sx.h> 51 52 #include <net/if.h> 53 #include <net/if_var.h> 54 #include <net/if_arp.h> 55 #include <net/if_dl.h> 56 #include <net/if_llatbl.h> 57 #include <net/if_types.h> 58 #include <net/route.h> 59 #include <net/vnet.h> 60 61 #include <netinet/if_ether.h> 62 #include <netinet/in.h> 63 #include <netinet/in_var.h> 64 #include <netinet/in_pcb.h> 65 #include <netinet/ip_var.h> 66 #include <netinet/ip_carp.h> 67 #include <netinet/igmp_var.h> 68 #include <netinet/udp.h> 69 #include <netinet/udp_var.h> 70 71 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *); 72 static int in_difaddr_ioctl(caddr_t, struct ifnet *, struct thread *); 73 74 static void in_socktrim(struct sockaddr_in *); 75 static void in_purgemaddrs(struct ifnet *); 76 77 static VNET_DEFINE(int, nosameprefix); 78 #define V_nosameprefix VNET(nosameprefix) 79 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_RW, 80 &VNET_NAME(nosameprefix), 0, 81 "Refuse to create same prefixes on different interfaces"); 82 83 VNET_DECLARE(struct inpcbinfo, ripcbinfo); 84 #define V_ripcbinfo VNET(ripcbinfo) 85 86 static struct sx in_control_sx; 87 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control"); 88 89 /* 90 * Return 1 if an internet address is for a ``local'' host 91 * (one to which we have a connection). 92 */ 93 int 94 in_localaddr(struct in_addr in) 95 { 96 register u_long i = ntohl(in.s_addr); 97 register struct in_ifaddr *ia; 98 99 IN_IFADDR_RLOCK(); 100 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 101 if ((i & ia->ia_subnetmask) == ia->ia_subnet) { 102 IN_IFADDR_RUNLOCK(); 103 return (1); 104 } 105 } 106 IN_IFADDR_RUNLOCK(); 107 return (0); 108 } 109 110 /* 111 * Return 1 if an internet address is for the local host and configured 112 * on one of its interfaces. 113 */ 114 int 115 in_localip(struct in_addr in) 116 { 117 struct in_ifaddr *ia; 118 119 IN_IFADDR_RLOCK(); 120 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) { 121 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) { 122 IN_IFADDR_RUNLOCK(); 123 return (1); 124 } 125 } 126 IN_IFADDR_RUNLOCK(); 127 return (0); 128 } 129 130 /* 131 * Return a reference to the interface address which is different to 132 * the supplied one but with same IP address value. 133 */ 134 static struct in_ifaddr * 135 in_localip_more(struct in_ifaddr *ia) 136 { 137 in_addr_t in = IA_SIN(ia)->sin_addr.s_addr; 138 struct in_ifaddr *it; 139 140 IN_IFADDR_RLOCK(); 141 LIST_FOREACH(it, INADDR_HASH(in), ia_hash) { 142 if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) { 143 ifa_ref(&it->ia_ifa); 144 IN_IFADDR_RUNLOCK(); 145 return (it); 146 } 147 } 148 IN_IFADDR_RUNLOCK(); 149 150 return (NULL); 151 } 152 153 /* 154 * Determine whether an IP address is in a reserved set of addresses 155 * that may not be forwarded, or whether datagrams to that destination 156 * may be forwarded. 157 */ 158 int 159 in_canforward(struct in_addr in) 160 { 161 register u_long i = ntohl(in.s_addr); 162 register u_long net; 163 164 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i)) 165 return (0); 166 if (IN_CLASSA(i)) { 167 net = i & IN_CLASSA_NET; 168 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 169 return (0); 170 } 171 return (1); 172 } 173 174 /* 175 * Trim a mask in a sockaddr 176 */ 177 static void 178 in_socktrim(struct sockaddr_in *ap) 179 { 180 register char *cplim = (char *) &ap->sin_addr; 181 register char *cp = (char *) (&ap->sin_addr + 1); 182 183 ap->sin_len = 0; 184 while (--cp >= cplim) 185 if (*cp) { 186 (ap)->sin_len = cp - (char *) (ap) + 1; 187 break; 188 } 189 } 190 191 /* 192 * Generic internet control operations (ioctl's). 193 */ 194 int 195 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, 196 struct thread *td) 197 { 198 struct ifreq *ifr = (struct ifreq *)data; 199 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr; 200 struct ifaddr *ifa; 201 struct in_ifaddr *ia; 202 int error; 203 204 if (ifp == NULL) 205 return (EADDRNOTAVAIL); 206 207 /* 208 * Filter out 4 ioctls we implement directly. Forward the rest 209 * to specific functions and ifp->if_ioctl(). 210 */ 211 switch (cmd) { 212 case SIOCGIFADDR: 213 case SIOCGIFBRDADDR: 214 case SIOCGIFDSTADDR: 215 case SIOCGIFNETMASK: 216 break; 217 case SIOCDIFADDR: 218 sx_xlock(&in_control_sx); 219 error = in_difaddr_ioctl(data, ifp, td); 220 sx_xunlock(&in_control_sx); 221 return (error); 222 case OSIOCAIFADDR: /* 9.x compat */ 223 case SIOCAIFADDR: 224 sx_xlock(&in_control_sx); 225 error = in_aifaddr_ioctl(cmd, data, ifp, td); 226 sx_xunlock(&in_control_sx); 227 return (error); 228 case SIOCSIFADDR: 229 case SIOCSIFBRDADDR: 230 case SIOCSIFDSTADDR: 231 case SIOCSIFNETMASK: 232 /* We no longer support that old commands. */ 233 return (EINVAL); 234 default: 235 if (ifp->if_ioctl == NULL) 236 return (EOPNOTSUPP); 237 return ((*ifp->if_ioctl)(ifp, cmd, data)); 238 } 239 240 if (addr->sin_addr.s_addr != INADDR_ANY && 241 prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0) 242 return (EADDRNOTAVAIL); 243 244 /* 245 * For SIOCGIFADDR, pick the first address. For the rest of 246 * ioctls, try to find specified address. 247 */ 248 IF_ADDR_RLOCK(ifp); 249 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 250 if (ifa->ifa_addr->sa_family != AF_INET) 251 continue; 252 ia = (struct in_ifaddr *)ifa; 253 if (cmd == SIOCGIFADDR || addr->sin_addr.s_addr == INADDR_ANY) 254 break; 255 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr) 256 break; 257 } 258 259 if (ifa == NULL) { 260 IF_ADDR_RUNLOCK(ifp); 261 return (EADDRNOTAVAIL); 262 } 263 264 error = 0; 265 switch (cmd) { 266 case SIOCGIFADDR: 267 *addr = ia->ia_addr; 268 break; 269 270 case SIOCGIFBRDADDR: 271 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 272 error = EINVAL; 273 break; 274 } 275 *addr = ia->ia_broadaddr; 276 break; 277 278 case SIOCGIFDSTADDR: 279 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 280 error = EINVAL; 281 break; 282 } 283 *addr = ia->ia_dstaddr; 284 break; 285 286 case SIOCGIFNETMASK: 287 *addr = ia->ia_sockmask; 288 break; 289 } 290 291 IF_ADDR_RUNLOCK(ifp); 292 293 return (error); 294 } 295 296 static int 297 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td) 298 { 299 const struct in_aliasreq *ifra = (struct in_aliasreq *)data; 300 const struct sockaddr_in *addr = &ifra->ifra_addr; 301 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr; 302 const struct sockaddr_in *mask = &ifra->ifra_mask; 303 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr; 304 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0; 305 struct ifaddr *ifa; 306 struct in_ifaddr *ia; 307 bool iaIsFirst; 308 int error = 0; 309 310 error = priv_check(td, PRIV_NET_ADDIFADDR); 311 if (error) 312 return (error); 313 314 /* 315 * ifra_addr must be present and be of INET family. 316 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional. 317 */ 318 if (addr->sin_len != sizeof(struct sockaddr_in) || 319 addr->sin_family != AF_INET) 320 return (EINVAL); 321 if (broadaddr->sin_len != 0 && 322 (broadaddr->sin_len != sizeof(struct sockaddr_in) || 323 broadaddr->sin_family != AF_INET)) 324 return (EINVAL); 325 if (mask->sin_len != 0 && 326 (mask->sin_len != sizeof(struct sockaddr_in) || 327 mask->sin_family != AF_INET)) 328 return (EINVAL); 329 if ((ifp->if_flags & IFF_POINTOPOINT) && 330 (dstaddr->sin_len != sizeof(struct sockaddr_in) || 331 dstaddr->sin_addr.s_addr == INADDR_ANY)) 332 return (EDESTADDRREQ); 333 if (vhid > 0 && carp_attach_p == NULL) 334 return (EPROTONOSUPPORT); 335 336 /* 337 * See whether address already exist. 338 */ 339 iaIsFirst = true; 340 ia = NULL; 341 IF_ADDR_RLOCK(ifp); 342 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 343 struct in_ifaddr *it; 344 345 if (ifa->ifa_addr->sa_family != AF_INET) 346 continue; 347 348 it = (struct in_ifaddr *)ifa; 349 iaIsFirst = false; 350 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 351 prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0) 352 ia = it; 353 } 354 IF_ADDR_RUNLOCK(ifp); 355 356 if (ia != NULL) 357 (void )in_difaddr_ioctl(data, ifp, td); 358 359 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK); 360 ia = (struct in_ifaddr *)ifa; 361 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 362 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 363 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 364 365 ia->ia_ifp = ifp; 366 ia->ia_addr = *addr; 367 if (mask->sin_len != 0) { 368 ia->ia_sockmask = *mask; 369 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 370 } else { 371 in_addr_t i = ntohl(addr->sin_addr.s_addr); 372 373 /* 374 * Be compatible with network classes, if netmask isn't 375 * supplied, guess it based on classes. 376 */ 377 if (IN_CLASSA(i)) 378 ia->ia_subnetmask = IN_CLASSA_NET; 379 else if (IN_CLASSB(i)) 380 ia->ia_subnetmask = IN_CLASSB_NET; 381 else 382 ia->ia_subnetmask = IN_CLASSC_NET; 383 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 384 } 385 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask; 386 in_socktrim(&ia->ia_sockmask); 387 388 if (ifp->if_flags & IFF_BROADCAST) { 389 if (broadaddr->sin_len != 0) { 390 ia->ia_broadaddr = *broadaddr; 391 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) { 392 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 393 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 394 ia->ia_broadaddr.sin_family = AF_INET; 395 } else { 396 ia->ia_broadaddr.sin_addr.s_addr = 397 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 398 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 399 ia->ia_broadaddr.sin_family = AF_INET; 400 } 401 } 402 403 if (ifp->if_flags & IFF_POINTOPOINT) 404 ia->ia_dstaddr = *dstaddr; 405 406 /* XXXGL: rtinit() needs this strange assignment. */ 407 if (ifp->if_flags & IFF_LOOPBACK) 408 ia->ia_dstaddr = ia->ia_addr; 409 410 /* if_addrhead is already referenced by ifa_alloc() */ 411 IF_ADDR_WLOCK(ifp); 412 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 413 IF_ADDR_WUNLOCK(ifp); 414 415 ifa_ref(ifa); /* in_ifaddrhead */ 416 IN_IFADDR_WLOCK(); 417 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 418 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash); 419 IN_IFADDR_WUNLOCK(); 420 421 if (vhid != 0) 422 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 423 if (error) 424 goto fail1; 425 426 /* 427 * Give the interface a chance to initialize 428 * if this is its first address, 429 * and to validate the address if necessary. 430 */ 431 if (ifp->if_ioctl != NULL) 432 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 433 if (error) 434 goto fail2; 435 436 /* 437 * Add route for the network. 438 */ 439 if (vhid == 0) { 440 int flags = RTF_UP; 441 442 if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT)) 443 flags |= RTF_HOST; 444 445 error = in_addprefix(ia, flags); 446 if (error) 447 goto fail2; 448 } 449 450 /* 451 * Add a loopback route to self. 452 */ 453 if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 && 454 ia->ia_addr.sin_addr.s_addr != INADDR_ANY && 455 !((ifp->if_flags & IFF_POINTOPOINT) && 456 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) { 457 struct in_ifaddr *eia; 458 459 eia = in_localip_more(ia); 460 461 if (eia == NULL) { 462 error = ifa_add_loopback_route((struct ifaddr *)ia, 463 (struct sockaddr *)&ia->ia_addr); 464 if (error) 465 goto fail3; 466 } else 467 ifa_free(&eia->ia_ifa); 468 } 469 470 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 471 struct in_addr allhosts_addr; 472 struct in_ifinfo *ii; 473 474 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 475 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 476 477 error = in_joingroup(ifp, &allhosts_addr, NULL, 478 &ii->ii_allhosts); 479 } 480 481 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 482 483 return (error); 484 485 fail3: 486 if (vhid == 0) 487 (void )in_scrubprefix(ia, LLE_STATIC); 488 489 fail2: 490 if (ia->ia_ifa.ifa_carp) 491 (*carp_detach_p)(&ia->ia_ifa); 492 493 fail1: 494 IF_ADDR_WLOCK(ifp); 495 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 496 IF_ADDR_WUNLOCK(ifp); 497 ifa_free(&ia->ia_ifa); /* if_addrhead */ 498 499 IN_IFADDR_WLOCK(); 500 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 501 LIST_REMOVE(ia, ia_hash); 502 IN_IFADDR_WUNLOCK(); 503 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 504 505 return (error); 506 } 507 508 static int 509 in_difaddr_ioctl(caddr_t data, struct ifnet *ifp, struct thread *td) 510 { 511 const struct ifreq *ifr = (struct ifreq *)data; 512 const struct sockaddr_in *addr = (const struct sockaddr_in *) 513 &ifr->ifr_addr; 514 struct ifaddr *ifa; 515 struct in_ifaddr *ia; 516 bool deleteAny, iaIsLast; 517 int error; 518 519 if (td != NULL) { 520 error = priv_check(td, PRIV_NET_DELIFADDR); 521 if (error) 522 return (error); 523 } 524 525 if (addr->sin_len != sizeof(struct sockaddr_in) || 526 addr->sin_family != AF_INET) 527 deleteAny = true; 528 else 529 deleteAny = false; 530 531 iaIsLast = true; 532 ia = NULL; 533 IF_ADDR_WLOCK(ifp); 534 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 535 struct in_ifaddr *it; 536 537 if (ifa->ifa_addr->sa_family != AF_INET) 538 continue; 539 540 it = (struct in_ifaddr *)ifa; 541 if (deleteAny && ia == NULL && (td == NULL || 542 prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0)) 543 ia = it; 544 545 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 546 (td == NULL || prison_check_ip4(td->td_ucred, 547 &addr->sin_addr) == 0)) 548 ia = it; 549 550 if (it != ia) 551 iaIsLast = false; 552 } 553 554 if (ia == NULL) { 555 IF_ADDR_WUNLOCK(ifp); 556 return (EADDRNOTAVAIL); 557 } 558 559 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 560 IF_ADDR_WUNLOCK(ifp); 561 ifa_free(&ia->ia_ifa); /* if_addrhead */ 562 563 IN_IFADDR_WLOCK(); 564 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 565 LIST_REMOVE(ia, ia_hash); 566 IN_IFADDR_WUNLOCK(); 567 568 /* 569 * in_scrubprefix() kills the interface route. 570 */ 571 in_scrubprefix(ia, LLE_STATIC); 572 573 /* 574 * in_ifadown gets rid of all the rest of 575 * the routes. This is not quite the right 576 * thing to do, but at least if we are running 577 * a routing process they will come back. 578 */ 579 in_ifadown(&ia->ia_ifa, 1); 580 581 if (ia->ia_ifa.ifa_carp) 582 (*carp_detach_p)(&ia->ia_ifa); 583 584 /* 585 * If this is the last IPv4 address configured on this 586 * interface, leave the all-hosts group. 587 * No state-change report need be transmitted. 588 */ 589 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 590 struct in_ifinfo *ii; 591 592 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 593 IN_MULTI_LOCK(); 594 if (ii->ii_allhosts) { 595 (void)in_leavegroup_locked(ii->ii_allhosts, NULL); 596 ii->ii_allhosts = NULL; 597 } 598 IN_MULTI_UNLOCK(); 599 } 600 601 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 602 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 603 604 return (0); 605 } 606 607 #define rtinitflags(x) \ 608 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 609 ? RTF_HOST : 0) 610 611 /* 612 * Check if we have a route for the given prefix already or add one accordingly. 613 */ 614 int 615 in_addprefix(struct in_ifaddr *target, int flags) 616 { 617 struct in_ifaddr *ia; 618 struct in_addr prefix, mask, p, m; 619 int error; 620 621 if ((flags & RTF_HOST) != 0) { 622 prefix = target->ia_dstaddr.sin_addr; 623 mask.s_addr = 0; 624 } else { 625 prefix = target->ia_addr.sin_addr; 626 mask = target->ia_sockmask.sin_addr; 627 prefix.s_addr &= mask.s_addr; 628 } 629 630 IN_IFADDR_RLOCK(); 631 /* Look for an existing address with the same prefix, mask, and fib */ 632 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 633 if (rtinitflags(ia)) { 634 p = ia->ia_dstaddr.sin_addr; 635 636 if (prefix.s_addr != p.s_addr) 637 continue; 638 } else { 639 p = ia->ia_addr.sin_addr; 640 m = ia->ia_sockmask.sin_addr; 641 p.s_addr &= m.s_addr; 642 643 if (prefix.s_addr != p.s_addr || 644 mask.s_addr != m.s_addr) 645 continue; 646 } 647 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 648 continue; 649 650 /* 651 * If we got a matching prefix route inserted by other 652 * interface address, we are done here. 653 */ 654 if (ia->ia_flags & IFA_ROUTE) { 655 #ifdef RADIX_MPATH 656 if (ia->ia_addr.sin_addr.s_addr == 657 target->ia_addr.sin_addr.s_addr) { 658 IN_IFADDR_RUNLOCK(); 659 return (EEXIST); 660 } else 661 break; 662 #endif 663 if (V_nosameprefix) { 664 IN_IFADDR_RUNLOCK(); 665 return (EEXIST); 666 } else { 667 int fibnum; 668 669 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS : 670 target->ia_ifp->if_fib; 671 rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum); 672 IN_IFADDR_RUNLOCK(); 673 return (0); 674 } 675 } 676 } 677 IN_IFADDR_RUNLOCK(); 678 679 /* 680 * No-one seem to have this prefix route, so we try to insert it. 681 */ 682 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 683 if (!error) 684 target->ia_flags |= IFA_ROUTE; 685 return (error); 686 } 687 688 /* 689 * If there is no other address in the system that can serve a route to the 690 * same prefix, remove the route. Hand over the route to the new address 691 * otherwise. 692 */ 693 int 694 in_scrubprefix(struct in_ifaddr *target, u_int flags) 695 { 696 struct in_ifaddr *ia; 697 struct in_addr prefix, mask, p, m; 698 int error = 0; 699 struct sockaddr_in prefix0, mask0; 700 701 /* 702 * Remove the loopback route to the interface address. 703 */ 704 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 705 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 706 (flags & LLE_STATIC)) { 707 struct in_ifaddr *eia; 708 709 eia = in_localip_more(target); 710 711 if (eia != NULL) { 712 int fibnum = target->ia_ifp->if_fib; 713 714 error = ifa_switch_loopback_route((struct ifaddr *)eia, 715 (struct sockaddr *)&target->ia_addr, fibnum); 716 ifa_free(&eia->ia_ifa); 717 } else { 718 error = ifa_del_loopback_route((struct ifaddr *)target, 719 (struct sockaddr *)&target->ia_addr); 720 } 721 722 if (!(target->ia_ifp->if_flags & IFF_NOARP)) 723 /* remove arp cache */ 724 arp_ifscrub(target->ia_ifp, 725 IA_SIN(target)->sin_addr.s_addr); 726 } 727 728 if (rtinitflags(target)) { 729 prefix = target->ia_dstaddr.sin_addr; 730 mask.s_addr = 0; 731 } else { 732 prefix = target->ia_addr.sin_addr; 733 mask = target->ia_sockmask.sin_addr; 734 prefix.s_addr &= mask.s_addr; 735 } 736 737 if ((target->ia_flags & IFA_ROUTE) == 0) { 738 int fibnum; 739 740 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS : 741 target->ia_ifp->if_fib; 742 rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum); 743 return (0); 744 } 745 746 IN_IFADDR_RLOCK(); 747 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 748 if (rtinitflags(ia)) { 749 p = ia->ia_dstaddr.sin_addr; 750 751 if (prefix.s_addr != p.s_addr) 752 continue; 753 } else { 754 p = ia->ia_addr.sin_addr; 755 m = ia->ia_sockmask.sin_addr; 756 p.s_addr &= m.s_addr; 757 758 if (prefix.s_addr != p.s_addr || 759 mask.s_addr != m.s_addr) 760 continue; 761 } 762 763 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 764 continue; 765 766 /* 767 * If we got a matching prefix address, move IFA_ROUTE and 768 * the route itself to it. Make sure that routing daemons 769 * get a heads-up. 770 */ 771 if ((ia->ia_flags & IFA_ROUTE) == 0) { 772 ifa_ref(&ia->ia_ifa); 773 IN_IFADDR_RUNLOCK(); 774 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 775 rtinitflags(target)); 776 if (error == 0) 777 target->ia_flags &= ~IFA_ROUTE; 778 else 779 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 780 error); 781 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 782 rtinitflags(ia) | RTF_UP); 783 if (error == 0) 784 ia->ia_flags |= IFA_ROUTE; 785 else 786 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 787 error); 788 ifa_free(&ia->ia_ifa); 789 return (error); 790 } 791 } 792 IN_IFADDR_RUNLOCK(); 793 794 /* 795 * remove all L2 entries on the given prefix 796 */ 797 bzero(&prefix0, sizeof(prefix0)); 798 prefix0.sin_len = sizeof(prefix0); 799 prefix0.sin_family = AF_INET; 800 prefix0.sin_addr.s_addr = target->ia_subnet; 801 bzero(&mask0, sizeof(mask0)); 802 mask0.sin_len = sizeof(mask0); 803 mask0.sin_family = AF_INET; 804 mask0.sin_addr.s_addr = target->ia_subnetmask; 805 lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0, 806 (struct sockaddr *)&mask0, flags); 807 808 /* 809 * As no-one seem to have this prefix, we can remove the route. 810 */ 811 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 812 if (error == 0) 813 target->ia_flags &= ~IFA_ROUTE; 814 else 815 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 816 return (error); 817 } 818 819 #undef rtinitflags 820 821 /* 822 * Return 1 if the address might be a local broadcast address. 823 */ 824 int 825 in_broadcast(struct in_addr in, struct ifnet *ifp) 826 { 827 register struct ifaddr *ifa; 828 u_long t; 829 830 if (in.s_addr == INADDR_BROADCAST || 831 in.s_addr == INADDR_ANY) 832 return (1); 833 if ((ifp->if_flags & IFF_BROADCAST) == 0) 834 return (0); 835 t = ntohl(in.s_addr); 836 /* 837 * Look through the list of addresses for a match 838 * with a broadcast address. 839 */ 840 #define ia ((struct in_ifaddr *)ifa) 841 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 842 if (ifa->ifa_addr->sa_family == AF_INET && 843 (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 844 /* 845 * Check for old-style (host 0) broadcast, but 846 * taking into account that RFC 3021 obsoletes it. 847 */ 848 (ia->ia_subnetmask != IN_RFC3021_MASK && 849 t == ia->ia_subnet)) && 850 /* 851 * Check for an all one subnetmask. These 852 * only exist when an interface gets a secondary 853 * address. 854 */ 855 ia->ia_subnetmask != (u_long)0xffffffff) 856 return (1); 857 return (0); 858 #undef ia 859 } 860 861 /* 862 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 863 */ 864 void 865 in_ifdetach(struct ifnet *ifp) 866 { 867 868 in_pcbpurgeif0(&V_ripcbinfo, ifp); 869 in_pcbpurgeif0(&V_udbinfo, ifp); 870 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 871 in_purgemaddrs(ifp); 872 } 873 874 /* 875 * Delete all IPv4 multicast address records, and associated link-layer 876 * multicast address records, associated with ifp. 877 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 878 * XXX This should not race with ifma_protospec being set during 879 * a new allocation, if it does, we have bigger problems. 880 */ 881 static void 882 in_purgemaddrs(struct ifnet *ifp) 883 { 884 LIST_HEAD(,in_multi) purgeinms; 885 struct in_multi *inm, *tinm; 886 struct ifmultiaddr *ifma; 887 888 LIST_INIT(&purgeinms); 889 IN_MULTI_LOCK(); 890 891 /* 892 * Extract list of in_multi associated with the detaching ifp 893 * which the PF_INET layer is about to release. 894 * We need to do this as IF_ADDR_LOCK() may be re-acquired 895 * by code further down. 896 */ 897 IF_ADDR_RLOCK(ifp); 898 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 899 if (ifma->ifma_addr->sa_family != AF_INET || 900 ifma->ifma_protospec == NULL) 901 continue; 902 #if 0 903 KASSERT(ifma->ifma_protospec != NULL, 904 ("%s: ifma_protospec is NULL", __func__)); 905 #endif 906 inm = (struct in_multi *)ifma->ifma_protospec; 907 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 908 } 909 IF_ADDR_RUNLOCK(ifp); 910 911 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 912 LIST_REMOVE(inm, inm_link); 913 inm_release_locked(inm); 914 } 915 igmp_ifdetach(ifp); 916 917 IN_MULTI_UNLOCK(); 918 } 919 920 struct in_llentry { 921 struct llentry base; 922 struct sockaddr_in l3_addr4; 923 }; 924 925 /* 926 * Deletes an address from the address table. 927 * This function is called by the timer functions 928 * such as arptimer() and nd6_llinfo_timer(), and 929 * the caller does the locking. 930 */ 931 static void 932 in_lltable_free(struct lltable *llt, struct llentry *lle) 933 { 934 LLE_WUNLOCK(lle); 935 LLE_LOCK_DESTROY(lle); 936 free(lle, M_LLTABLE); 937 } 938 939 static struct llentry * 940 in_lltable_new(const struct sockaddr *l3addr, u_int flags) 941 { 942 struct in_llentry *lle; 943 944 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 945 if (lle == NULL) /* NB: caller generates msg */ 946 return NULL; 947 948 /* 949 * For IPv4 this will trigger "arpresolve" to generate 950 * an ARP request. 951 */ 952 lle->base.la_expire = time_uptime; /* mark expired */ 953 lle->l3_addr4 = *(const struct sockaddr_in *)l3addr; 954 lle->base.lle_refcnt = 1; 955 lle->base.lle_free = in_lltable_free; 956 LLE_LOCK_INIT(&lle->base); 957 callout_init_rw(&lle->base.la_timer, &lle->base.lle_lock, 958 CALLOUT_RETURNUNLOCKED); 959 960 return (&lle->base); 961 } 962 963 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 964 (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 965 966 static void 967 in_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix, 968 const struct sockaddr *mask, u_int flags) 969 { 970 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 971 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 972 struct llentry *lle, *next; 973 int i; 974 size_t pkts_dropped; 975 976 IF_AFDATA_WLOCK(llt->llt_ifp); 977 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 978 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 979 /* 980 * (flags & LLE_STATIC) means deleting all entries 981 * including static ARP entries. 982 */ 983 if (IN_ARE_MASKED_ADDR_EQUAL(satosin(L3_ADDR(lle)), 984 pfx, msk) && ((flags & LLE_STATIC) || 985 !(lle->la_flags & LLE_STATIC))) { 986 LLE_WLOCK(lle); 987 if (callout_stop(&lle->la_timer)) 988 LLE_REMREF(lle); 989 pkts_dropped = llentry_free(lle); 990 ARPSTAT_ADD(dropped, pkts_dropped); 991 } 992 } 993 } 994 IF_AFDATA_WUNLOCK(llt->llt_ifp); 995 } 996 997 998 static int 999 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1000 { 1001 struct rtentry *rt; 1002 1003 KASSERT(l3addr->sa_family == AF_INET, 1004 ("sin_family %d", l3addr->sa_family)); 1005 1006 /* XXX rtalloc1_fib should take a const param */ 1007 rt = rtalloc1_fib(__DECONST(struct sockaddr *, l3addr), 0, 0, 1008 ifp->if_fib); 1009 1010 if (rt == NULL) 1011 return (EINVAL); 1012 1013 /* 1014 * If the gateway for an existing host route matches the target L3 1015 * address, which is a special route inserted by some implementation 1016 * such as MANET, and the interface is of the correct type, then 1017 * allow for ARP to proceed. 1018 */ 1019 if (rt->rt_flags & RTF_GATEWAY) { 1020 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 1021 rt->rt_ifp->if_type != IFT_ETHER || 1022 (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1023 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 1024 sizeof(in_addr_t)) != 0) { 1025 RTFREE_LOCKED(rt); 1026 return (EINVAL); 1027 } 1028 } 1029 1030 /* 1031 * Make sure that at least the destination address is covered 1032 * by the route. This is for handling the case where 2 or more 1033 * interfaces have the same prefix. An incoming packet arrives 1034 * on one interface and the corresponding outgoing packet leaves 1035 * another interface. 1036 */ 1037 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 1038 const char *sa, *mask, *addr, *lim; 1039 int len; 1040 1041 mask = (const char *)rt_mask(rt); 1042 /* 1043 * Just being extra cautious to avoid some custom 1044 * code getting into trouble. 1045 */ 1046 if (mask == NULL) { 1047 RTFREE_LOCKED(rt); 1048 return (EINVAL); 1049 } 1050 1051 sa = (const char *)rt_key(rt); 1052 addr = (const char *)l3addr; 1053 len = ((const struct sockaddr_in *)l3addr)->sin_len; 1054 lim = addr + len; 1055 1056 for ( ; addr < lim; sa++, mask++, addr++) { 1057 if ((*sa ^ *addr) & *mask) { 1058 #ifdef DIAGNOSTIC 1059 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1060 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 1061 #endif 1062 RTFREE_LOCKED(rt); 1063 return (EINVAL); 1064 } 1065 } 1066 } 1067 1068 RTFREE_LOCKED(rt); 1069 return (0); 1070 } 1071 1072 /* 1073 * Return NULL if not found or marked for deletion. 1074 * If found return lle read locked. 1075 */ 1076 static struct llentry * 1077 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1078 { 1079 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1080 struct ifnet *ifp = llt->llt_ifp; 1081 struct llentry *lle; 1082 struct llentries *lleh; 1083 u_int hashkey; 1084 1085 IF_AFDATA_LOCK_ASSERT(ifp); 1086 KASSERT(l3addr->sa_family == AF_INET, 1087 ("sin_family %d", l3addr->sa_family)); 1088 1089 hashkey = sin->sin_addr.s_addr; 1090 lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; 1091 LIST_FOREACH(lle, lleh, lle_next) { 1092 struct sockaddr_in *sa2 = satosin(L3_ADDR(lle)); 1093 if (lle->la_flags & LLE_DELETED) 1094 continue; 1095 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr) 1096 break; 1097 } 1098 if (lle == NULL) { 1099 #ifdef DIAGNOSTIC 1100 if (flags & LLE_DELETE) 1101 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); 1102 #endif 1103 if (!(flags & LLE_CREATE)) 1104 return (NULL); 1105 IF_AFDATA_WLOCK_ASSERT(ifp); 1106 /* 1107 * A route that covers the given address must have 1108 * been installed 1st because we are doing a resolution, 1109 * verify this. 1110 */ 1111 if (!(flags & LLE_IFADDR) && 1112 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1113 goto done; 1114 1115 lle = in_lltable_new(l3addr, flags); 1116 if (lle == NULL) { 1117 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1118 goto done; 1119 } 1120 lle->la_flags = flags & ~LLE_CREATE; 1121 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { 1122 bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); 1123 lle->la_flags |= (LLE_VALID | LLE_STATIC); 1124 } 1125 1126 lle->lle_tbl = llt; 1127 lle->lle_head = lleh; 1128 lle->la_flags |= LLE_LINKED; 1129 LIST_INSERT_HEAD(lleh, lle, lle_next); 1130 } else if (flags & LLE_DELETE) { 1131 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) { 1132 LLE_WLOCK(lle); 1133 lle->la_flags |= LLE_DELETED; 1134 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1135 #ifdef DIAGNOSTIC 1136 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1137 #endif 1138 if ((lle->la_flags & 1139 (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC) 1140 llentry_free(lle); 1141 else 1142 LLE_WUNLOCK(lle); 1143 } 1144 lle = (void *)-1; 1145 1146 } 1147 if (LLE_IS_VALID(lle)) { 1148 if (flags & LLE_EXCLUSIVE) 1149 LLE_WLOCK(lle); 1150 else 1151 LLE_RLOCK(lle); 1152 } 1153 done: 1154 return (lle); 1155 } 1156 1157 static int 1158 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr) 1159 { 1160 #define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle)) 1161 struct ifnet *ifp = llt->llt_ifp; 1162 struct llentry *lle; 1163 /* XXX stack use */ 1164 struct { 1165 struct rt_msghdr rtm; 1166 struct sockaddr_in sin; 1167 struct sockaddr_dl sdl; 1168 } arpc; 1169 int error, i; 1170 1171 LLTABLE_LOCK_ASSERT(); 1172 1173 error = 0; 1174 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1175 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1176 struct sockaddr_dl *sdl; 1177 1178 /* skip deleted entries */ 1179 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1180 continue; 1181 /* Skip if jailed and not a valid IP of the prison. */ 1182 if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0) 1183 continue; 1184 /* 1185 * produce a msg made of: 1186 * struct rt_msghdr; 1187 * struct sockaddr_in; (IPv4) 1188 * struct sockaddr_dl; 1189 */ 1190 bzero(&arpc, sizeof(arpc)); 1191 arpc.rtm.rtm_msglen = sizeof(arpc); 1192 arpc.rtm.rtm_version = RTM_VERSION; 1193 arpc.rtm.rtm_type = RTM_GET; 1194 arpc.rtm.rtm_flags = RTF_UP; 1195 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1196 arpc.sin.sin_family = AF_INET; 1197 arpc.sin.sin_len = sizeof(arpc.sin); 1198 arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr; 1199 1200 /* publish */ 1201 if (lle->la_flags & LLE_PUB) 1202 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1203 1204 sdl = &arpc.sdl; 1205 sdl->sdl_family = AF_LINK; 1206 sdl->sdl_len = sizeof(*sdl); 1207 sdl->sdl_index = ifp->if_index; 1208 sdl->sdl_type = ifp->if_type; 1209 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1210 sdl->sdl_alen = ifp->if_addrlen; 1211 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1212 } else { 1213 sdl->sdl_alen = 0; 1214 bzero(LLADDR(sdl), ifp->if_addrlen); 1215 } 1216 1217 arpc.rtm.rtm_rmx.rmx_expire = 1218 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1219 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1220 if (lle->la_flags & LLE_STATIC) 1221 arpc.rtm.rtm_flags |= RTF_STATIC; 1222 arpc.rtm.rtm_index = ifp->if_index; 1223 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1224 if (error) 1225 break; 1226 } 1227 } 1228 return error; 1229 #undef SIN 1230 } 1231 1232 void * 1233 in_domifattach(struct ifnet *ifp) 1234 { 1235 struct in_ifinfo *ii; 1236 struct lltable *llt; 1237 1238 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1239 1240 llt = lltable_init(ifp, AF_INET); 1241 if (llt != NULL) { 1242 llt->llt_prefix_free = in_lltable_prefix_free; 1243 llt->llt_lookup = in_lltable_lookup; 1244 llt->llt_dump = in_lltable_dump; 1245 } 1246 ii->ii_llt = llt; 1247 1248 ii->ii_igmp = igmp_domifattach(ifp); 1249 1250 return ii; 1251 } 1252 1253 void 1254 in_domifdetach(struct ifnet *ifp, void *aux) 1255 { 1256 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1257 1258 igmp_domifdetach(ifp); 1259 lltable_free(ii->ii_llt); 1260 free(ii, M_IFADDR); 1261 } 1262