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 (vhid != 0) { 411 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 412 if (error) 413 return (error); 414 } 415 416 /* if_addrhead is already referenced by ifa_alloc() */ 417 IF_ADDR_WLOCK(ifp); 418 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 419 IF_ADDR_WUNLOCK(ifp); 420 421 ifa_ref(ifa); /* in_ifaddrhead */ 422 IN_IFADDR_WLOCK(); 423 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 424 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash); 425 IN_IFADDR_WUNLOCK(); 426 427 /* 428 * Give the interface a chance to initialize 429 * if this is its first address, 430 * and to validate the address if necessary. 431 */ 432 if (ifp->if_ioctl != NULL) { 433 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 434 if (error) 435 goto fail1; 436 } 437 438 /* 439 * Add route for the network. 440 */ 441 if (vhid == 0) { 442 int flags = RTF_UP; 443 444 if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT)) 445 flags |= RTF_HOST; 446 447 error = in_addprefix(ia, flags); 448 if (error) 449 goto fail1; 450 } 451 452 /* 453 * Add a loopback route to self. 454 */ 455 if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 && 456 ia->ia_addr.sin_addr.s_addr != INADDR_ANY && 457 !((ifp->if_flags & IFF_POINTOPOINT) && 458 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) { 459 struct in_ifaddr *eia; 460 461 eia = in_localip_more(ia); 462 463 if (eia == NULL) { 464 error = ifa_add_loopback_route((struct ifaddr *)ia, 465 (struct sockaddr *)&ia->ia_addr); 466 if (error) 467 goto fail2; 468 } else 469 ifa_free(&eia->ia_ifa); 470 } 471 472 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 473 struct in_addr allhosts_addr; 474 struct in_ifinfo *ii; 475 476 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 477 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 478 479 error = in_joingroup(ifp, &allhosts_addr, NULL, 480 &ii->ii_allhosts); 481 } 482 483 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 484 485 return (error); 486 487 fail2: 488 if (vhid == 0) 489 (void )in_scrubprefix(ia, LLE_STATIC); 490 491 fail1: 492 if (ia->ia_ifa.ifa_carp) 493 (*carp_detach_p)(&ia->ia_ifa); 494 495 IF_ADDR_WLOCK(ifp); 496 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 497 IF_ADDR_WUNLOCK(ifp); 498 ifa_free(&ia->ia_ifa); /* if_addrhead */ 499 500 IN_IFADDR_WLOCK(); 501 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 502 LIST_REMOVE(ia, ia_hash); 503 IN_IFADDR_WUNLOCK(); 504 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 505 506 return (error); 507 } 508 509 static int 510 in_difaddr_ioctl(caddr_t data, struct ifnet *ifp, struct thread *td) 511 { 512 const struct ifreq *ifr = (struct ifreq *)data; 513 const struct sockaddr_in *addr = (const struct sockaddr_in *) 514 &ifr->ifr_addr; 515 struct ifaddr *ifa; 516 struct in_ifaddr *ia; 517 bool deleteAny, iaIsLast; 518 int error; 519 520 if (td != NULL) { 521 error = priv_check(td, PRIV_NET_DELIFADDR); 522 if (error) 523 return (error); 524 } 525 526 if (addr->sin_len != sizeof(struct sockaddr_in) || 527 addr->sin_family != AF_INET) 528 deleteAny = true; 529 else 530 deleteAny = false; 531 532 iaIsLast = true; 533 ia = NULL; 534 IF_ADDR_WLOCK(ifp); 535 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 536 struct in_ifaddr *it; 537 538 if (ifa->ifa_addr->sa_family != AF_INET) 539 continue; 540 541 it = (struct in_ifaddr *)ifa; 542 if (deleteAny && ia == NULL && (td == NULL || 543 prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0)) 544 ia = it; 545 546 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 547 (td == NULL || prison_check_ip4(td->td_ucred, 548 &addr->sin_addr) == 0)) 549 ia = it; 550 551 if (it != ia) 552 iaIsLast = false; 553 } 554 555 if (ia == NULL) { 556 IF_ADDR_WUNLOCK(ifp); 557 return (EADDRNOTAVAIL); 558 } 559 560 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 561 IF_ADDR_WUNLOCK(ifp); 562 ifa_free(&ia->ia_ifa); /* if_addrhead */ 563 564 IN_IFADDR_WLOCK(); 565 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 566 LIST_REMOVE(ia, ia_hash); 567 IN_IFADDR_WUNLOCK(); 568 569 /* 570 * in_scrubprefix() kills the interface route. 571 */ 572 in_scrubprefix(ia, LLE_STATIC); 573 574 /* 575 * in_ifadown gets rid of all the rest of 576 * the routes. This is not quite the right 577 * thing to do, but at least if we are running 578 * a routing process they will come back. 579 */ 580 in_ifadown(&ia->ia_ifa, 1); 581 582 if (ia->ia_ifa.ifa_carp) 583 (*carp_detach_p)(&ia->ia_ifa); 584 585 /* 586 * If this is the last IPv4 address configured on this 587 * interface, leave the all-hosts group. 588 * No state-change report need be transmitted. 589 */ 590 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 591 struct in_ifinfo *ii; 592 593 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 594 IN_MULTI_LOCK(); 595 if (ii->ii_allhosts) { 596 (void)in_leavegroup_locked(ii->ii_allhosts, NULL); 597 ii->ii_allhosts = NULL; 598 } 599 IN_MULTI_UNLOCK(); 600 } 601 602 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 603 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 604 605 return (0); 606 } 607 608 #define rtinitflags(x) \ 609 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 610 ? RTF_HOST : 0) 611 612 /* 613 * Check if we have a route for the given prefix already or add one accordingly. 614 */ 615 int 616 in_addprefix(struct in_ifaddr *target, int flags) 617 { 618 struct in_ifaddr *ia; 619 struct in_addr prefix, mask, p, m; 620 int error; 621 622 if ((flags & RTF_HOST) != 0) { 623 prefix = target->ia_dstaddr.sin_addr; 624 mask.s_addr = 0; 625 } else { 626 prefix = target->ia_addr.sin_addr; 627 mask = target->ia_sockmask.sin_addr; 628 prefix.s_addr &= mask.s_addr; 629 } 630 631 IN_IFADDR_RLOCK(); 632 /* Look for an existing address with the same prefix, mask, and fib */ 633 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 634 if (rtinitflags(ia)) { 635 p = ia->ia_dstaddr.sin_addr; 636 637 if (prefix.s_addr != p.s_addr) 638 continue; 639 } else { 640 p = ia->ia_addr.sin_addr; 641 m = ia->ia_sockmask.sin_addr; 642 p.s_addr &= m.s_addr; 643 644 if (prefix.s_addr != p.s_addr || 645 mask.s_addr != m.s_addr) 646 continue; 647 } 648 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 649 continue; 650 651 /* 652 * If we got a matching prefix route inserted by other 653 * interface address, we are done here. 654 */ 655 if (ia->ia_flags & IFA_ROUTE) { 656 #ifdef RADIX_MPATH 657 if (ia->ia_addr.sin_addr.s_addr == 658 target->ia_addr.sin_addr.s_addr) { 659 IN_IFADDR_RUNLOCK(); 660 return (EEXIST); 661 } else 662 break; 663 #endif 664 if (V_nosameprefix) { 665 IN_IFADDR_RUNLOCK(); 666 return (EEXIST); 667 } else { 668 int fibnum; 669 670 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS : 671 target->ia_ifp->if_fib; 672 rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum); 673 IN_IFADDR_RUNLOCK(); 674 return (0); 675 } 676 } 677 } 678 IN_IFADDR_RUNLOCK(); 679 680 /* 681 * No-one seem to have this prefix route, so we try to insert it. 682 */ 683 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 684 if (!error) 685 target->ia_flags |= IFA_ROUTE; 686 return (error); 687 } 688 689 /* 690 * If there is no other address in the system that can serve a route to the 691 * same prefix, remove the route. Hand over the route to the new address 692 * otherwise. 693 */ 694 int 695 in_scrubprefix(struct in_ifaddr *target, u_int flags) 696 { 697 struct in_ifaddr *ia; 698 struct in_addr prefix, mask, p, m; 699 int error = 0; 700 struct sockaddr_in prefix0, mask0; 701 702 /* 703 * Remove the loopback route to the interface address. 704 */ 705 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 706 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 707 (flags & LLE_STATIC)) { 708 struct in_ifaddr *eia; 709 710 eia = in_localip_more(target); 711 712 if (eia != NULL) { 713 int fibnum = target->ia_ifp->if_fib; 714 715 error = ifa_switch_loopback_route((struct ifaddr *)eia, 716 (struct sockaddr *)&target->ia_addr, fibnum); 717 ifa_free(&eia->ia_ifa); 718 } else { 719 error = ifa_del_loopback_route((struct ifaddr *)target, 720 (struct sockaddr *)&target->ia_addr); 721 } 722 723 if (!(target->ia_ifp->if_flags & IFF_NOARP)) 724 /* remove arp cache */ 725 arp_ifscrub(target->ia_ifp, 726 IA_SIN(target)->sin_addr.s_addr); 727 } 728 729 if (rtinitflags(target)) { 730 prefix = target->ia_dstaddr.sin_addr; 731 mask.s_addr = 0; 732 } else { 733 prefix = target->ia_addr.sin_addr; 734 mask = target->ia_sockmask.sin_addr; 735 prefix.s_addr &= mask.s_addr; 736 } 737 738 if ((target->ia_flags & IFA_ROUTE) == 0) { 739 int fibnum; 740 741 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS : 742 target->ia_ifp->if_fib; 743 rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum); 744 return (0); 745 } 746 747 IN_IFADDR_RLOCK(); 748 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 749 if (rtinitflags(ia)) { 750 p = ia->ia_dstaddr.sin_addr; 751 752 if (prefix.s_addr != p.s_addr) 753 continue; 754 } else { 755 p = ia->ia_addr.sin_addr; 756 m = ia->ia_sockmask.sin_addr; 757 p.s_addr &= m.s_addr; 758 759 if (prefix.s_addr != p.s_addr || 760 mask.s_addr != m.s_addr) 761 continue; 762 } 763 764 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 765 continue; 766 767 /* 768 * If we got a matching prefix address, move IFA_ROUTE and 769 * the route itself to it. Make sure that routing daemons 770 * get a heads-up. 771 */ 772 if ((ia->ia_flags & IFA_ROUTE) == 0) { 773 ifa_ref(&ia->ia_ifa); 774 IN_IFADDR_RUNLOCK(); 775 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 776 rtinitflags(target)); 777 if (error == 0) 778 target->ia_flags &= ~IFA_ROUTE; 779 else 780 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 781 error); 782 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 783 rtinitflags(ia) | RTF_UP); 784 if (error == 0) 785 ia->ia_flags |= IFA_ROUTE; 786 else 787 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 788 error); 789 ifa_free(&ia->ia_ifa); 790 return (error); 791 } 792 } 793 IN_IFADDR_RUNLOCK(); 794 795 /* 796 * remove all L2 entries on the given prefix 797 */ 798 bzero(&prefix0, sizeof(prefix0)); 799 prefix0.sin_len = sizeof(prefix0); 800 prefix0.sin_family = AF_INET; 801 prefix0.sin_addr.s_addr = target->ia_subnet; 802 bzero(&mask0, sizeof(mask0)); 803 mask0.sin_len = sizeof(mask0); 804 mask0.sin_family = AF_INET; 805 mask0.sin_addr.s_addr = target->ia_subnetmask; 806 lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0, 807 (struct sockaddr *)&mask0, flags); 808 809 /* 810 * As no-one seem to have this prefix, we can remove the route. 811 */ 812 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 813 if (error == 0) 814 target->ia_flags &= ~IFA_ROUTE; 815 else 816 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 817 return (error); 818 } 819 820 #undef rtinitflags 821 822 /* 823 * Return 1 if the address might be a local broadcast address. 824 */ 825 int 826 in_broadcast(struct in_addr in, struct ifnet *ifp) 827 { 828 register struct ifaddr *ifa; 829 u_long t; 830 831 if (in.s_addr == INADDR_BROADCAST || 832 in.s_addr == INADDR_ANY) 833 return (1); 834 if ((ifp->if_flags & IFF_BROADCAST) == 0) 835 return (0); 836 t = ntohl(in.s_addr); 837 /* 838 * Look through the list of addresses for a match 839 * with a broadcast address. 840 */ 841 #define ia ((struct in_ifaddr *)ifa) 842 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 843 if (ifa->ifa_addr->sa_family == AF_INET && 844 (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 845 /* 846 * Check for old-style (host 0) broadcast, but 847 * taking into account that RFC 3021 obsoletes it. 848 */ 849 (ia->ia_subnetmask != IN_RFC3021_MASK && 850 t == ia->ia_subnet)) && 851 /* 852 * Check for an all one subnetmask. These 853 * only exist when an interface gets a secondary 854 * address. 855 */ 856 ia->ia_subnetmask != (u_long)0xffffffff) 857 return (1); 858 return (0); 859 #undef ia 860 } 861 862 /* 863 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 864 */ 865 void 866 in_ifdetach(struct ifnet *ifp) 867 { 868 869 in_pcbpurgeif0(&V_ripcbinfo, ifp); 870 in_pcbpurgeif0(&V_udbinfo, ifp); 871 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 872 in_purgemaddrs(ifp); 873 } 874 875 /* 876 * Delete all IPv4 multicast address records, and associated link-layer 877 * multicast address records, associated with ifp. 878 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 879 * XXX This should not race with ifma_protospec being set during 880 * a new allocation, if it does, we have bigger problems. 881 */ 882 static void 883 in_purgemaddrs(struct ifnet *ifp) 884 { 885 LIST_HEAD(,in_multi) purgeinms; 886 struct in_multi *inm, *tinm; 887 struct ifmultiaddr *ifma; 888 889 LIST_INIT(&purgeinms); 890 IN_MULTI_LOCK(); 891 892 /* 893 * Extract list of in_multi associated with the detaching ifp 894 * which the PF_INET layer is about to release. 895 * We need to do this as IF_ADDR_LOCK() may be re-acquired 896 * by code further down. 897 */ 898 IF_ADDR_RLOCK(ifp); 899 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 900 if (ifma->ifma_addr->sa_family != AF_INET || 901 ifma->ifma_protospec == NULL) 902 continue; 903 #if 0 904 KASSERT(ifma->ifma_protospec != NULL, 905 ("%s: ifma_protospec is NULL", __func__)); 906 #endif 907 inm = (struct in_multi *)ifma->ifma_protospec; 908 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 909 } 910 IF_ADDR_RUNLOCK(ifp); 911 912 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 913 LIST_REMOVE(inm, inm_link); 914 inm_release_locked(inm); 915 } 916 igmp_ifdetach(ifp); 917 918 IN_MULTI_UNLOCK(); 919 } 920 921 struct in_llentry { 922 struct llentry base; 923 struct sockaddr_in l3_addr4; 924 }; 925 926 /* 927 * Deletes an address from the address table. 928 * This function is called by the timer functions 929 * such as arptimer() and nd6_llinfo_timer(), and 930 * the caller does the locking. 931 */ 932 static void 933 in_lltable_free(struct lltable *llt, struct llentry *lle) 934 { 935 LLE_WUNLOCK(lle); 936 LLE_LOCK_DESTROY(lle); 937 free(lle, M_LLTABLE); 938 } 939 940 static struct llentry * 941 in_lltable_new(const struct sockaddr *l3addr, u_int flags) 942 { 943 struct in_llentry *lle; 944 945 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 946 if (lle == NULL) /* NB: caller generates msg */ 947 return NULL; 948 949 /* 950 * For IPv4 this will trigger "arpresolve" to generate 951 * an ARP request. 952 */ 953 lle->base.la_expire = time_uptime; /* mark expired */ 954 lle->l3_addr4 = *(const struct sockaddr_in *)l3addr; 955 lle->base.lle_refcnt = 1; 956 lle->base.lle_free = in_lltable_free; 957 LLE_LOCK_INIT(&lle->base); 958 callout_init_rw(&lle->base.la_timer, &lle->base.lle_lock, 959 CALLOUT_RETURNUNLOCKED); 960 961 return (&lle->base); 962 } 963 964 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 965 (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 966 967 static void 968 in_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix, 969 const struct sockaddr *mask, u_int flags) 970 { 971 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 972 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 973 struct llentry *lle, *next; 974 int i; 975 size_t pkts_dropped; 976 977 IF_AFDATA_WLOCK(llt->llt_ifp); 978 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 979 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 980 /* 981 * (flags & LLE_STATIC) means deleting all entries 982 * including static ARP entries. 983 */ 984 if (IN_ARE_MASKED_ADDR_EQUAL(satosin(L3_ADDR(lle)), 985 pfx, msk) && ((flags & LLE_STATIC) || 986 !(lle->la_flags & LLE_STATIC))) { 987 LLE_WLOCK(lle); 988 if (callout_stop(&lle->la_timer)) 989 LLE_REMREF(lle); 990 pkts_dropped = llentry_free(lle); 991 ARPSTAT_ADD(dropped, pkts_dropped); 992 } 993 } 994 } 995 IF_AFDATA_WUNLOCK(llt->llt_ifp); 996 } 997 998 999 static int 1000 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1001 { 1002 struct rtentry *rt; 1003 1004 KASSERT(l3addr->sa_family == AF_INET, 1005 ("sin_family %d", l3addr->sa_family)); 1006 1007 /* XXX rtalloc1_fib should take a const param */ 1008 rt = rtalloc1_fib(__DECONST(struct sockaddr *, l3addr), 0, 0, 1009 ifp->if_fib); 1010 1011 if (rt == NULL) 1012 return (EINVAL); 1013 1014 /* 1015 * If the gateway for an existing host route matches the target L3 1016 * address, which is a special route inserted by some implementation 1017 * such as MANET, and the interface is of the correct type, then 1018 * allow for ARP to proceed. 1019 */ 1020 if (rt->rt_flags & RTF_GATEWAY) { 1021 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 1022 rt->rt_ifp->if_type != IFT_ETHER || 1023 (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1024 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 1025 sizeof(in_addr_t)) != 0) { 1026 RTFREE_LOCKED(rt); 1027 return (EINVAL); 1028 } 1029 } 1030 1031 /* 1032 * Make sure that at least the destination address is covered 1033 * by the route. This is for handling the case where 2 or more 1034 * interfaces have the same prefix. An incoming packet arrives 1035 * on one interface and the corresponding outgoing packet leaves 1036 * another interface. 1037 */ 1038 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 1039 const char *sa, *mask, *addr, *lim; 1040 int len; 1041 1042 mask = (const char *)rt_mask(rt); 1043 /* 1044 * Just being extra cautious to avoid some custom 1045 * code getting into trouble. 1046 */ 1047 if (mask == NULL) { 1048 RTFREE_LOCKED(rt); 1049 return (EINVAL); 1050 } 1051 1052 sa = (const char *)rt_key(rt); 1053 addr = (const char *)l3addr; 1054 len = ((const struct sockaddr_in *)l3addr)->sin_len; 1055 lim = addr + len; 1056 1057 for ( ; addr < lim; sa++, mask++, addr++) { 1058 if ((*sa ^ *addr) & *mask) { 1059 #ifdef DIAGNOSTIC 1060 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1061 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 1062 #endif 1063 RTFREE_LOCKED(rt); 1064 return (EINVAL); 1065 } 1066 } 1067 } 1068 1069 RTFREE_LOCKED(rt); 1070 return (0); 1071 } 1072 1073 /* 1074 * Return NULL if not found or marked for deletion. 1075 * If found return lle read locked. 1076 */ 1077 static struct llentry * 1078 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1079 { 1080 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1081 struct ifnet *ifp = llt->llt_ifp; 1082 struct llentry *lle; 1083 struct llentries *lleh; 1084 u_int hashkey; 1085 1086 IF_AFDATA_LOCK_ASSERT(ifp); 1087 KASSERT(l3addr->sa_family == AF_INET, 1088 ("sin_family %d", l3addr->sa_family)); 1089 1090 hashkey = sin->sin_addr.s_addr; 1091 lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; 1092 LIST_FOREACH(lle, lleh, lle_next) { 1093 struct sockaddr_in *sa2 = satosin(L3_ADDR(lle)); 1094 if (lle->la_flags & LLE_DELETED) 1095 continue; 1096 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr) 1097 break; 1098 } 1099 if (lle == NULL) { 1100 #ifdef DIAGNOSTIC 1101 if (flags & LLE_DELETE) 1102 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); 1103 #endif 1104 if (!(flags & LLE_CREATE)) 1105 return (NULL); 1106 IF_AFDATA_WLOCK_ASSERT(ifp); 1107 /* 1108 * A route that covers the given address must have 1109 * been installed 1st because we are doing a resolution, 1110 * verify this. 1111 */ 1112 if (!(flags & LLE_IFADDR) && 1113 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1114 goto done; 1115 1116 lle = in_lltable_new(l3addr, flags); 1117 if (lle == NULL) { 1118 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1119 goto done; 1120 } 1121 lle->la_flags = flags & ~LLE_CREATE; 1122 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { 1123 bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); 1124 lle->la_flags |= (LLE_VALID | LLE_STATIC); 1125 } 1126 1127 lle->lle_tbl = llt; 1128 lle->lle_head = lleh; 1129 lle->la_flags |= LLE_LINKED; 1130 LIST_INSERT_HEAD(lleh, lle, lle_next); 1131 } else if (flags & LLE_DELETE) { 1132 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) { 1133 LLE_WLOCK(lle); 1134 lle->la_flags |= LLE_DELETED; 1135 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1136 #ifdef DIAGNOSTIC 1137 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1138 #endif 1139 if ((lle->la_flags & 1140 (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC) 1141 llentry_free(lle); 1142 else 1143 LLE_WUNLOCK(lle); 1144 } 1145 lle = (void *)-1; 1146 1147 } 1148 if (LLE_IS_VALID(lle)) { 1149 if (flags & LLE_EXCLUSIVE) 1150 LLE_WLOCK(lle); 1151 else 1152 LLE_RLOCK(lle); 1153 } 1154 done: 1155 return (lle); 1156 } 1157 1158 static int 1159 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr) 1160 { 1161 #define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle)) 1162 struct ifnet *ifp = llt->llt_ifp; 1163 struct llentry *lle; 1164 /* XXX stack use */ 1165 struct { 1166 struct rt_msghdr rtm; 1167 struct sockaddr_in sin; 1168 struct sockaddr_dl sdl; 1169 } arpc; 1170 int error, i; 1171 1172 LLTABLE_LOCK_ASSERT(); 1173 1174 error = 0; 1175 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1176 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1177 struct sockaddr_dl *sdl; 1178 1179 /* skip deleted entries */ 1180 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1181 continue; 1182 /* Skip if jailed and not a valid IP of the prison. */ 1183 if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0) 1184 continue; 1185 /* 1186 * produce a msg made of: 1187 * struct rt_msghdr; 1188 * struct sockaddr_in; (IPv4) 1189 * struct sockaddr_dl; 1190 */ 1191 bzero(&arpc, sizeof(arpc)); 1192 arpc.rtm.rtm_msglen = sizeof(arpc); 1193 arpc.rtm.rtm_version = RTM_VERSION; 1194 arpc.rtm.rtm_type = RTM_GET; 1195 arpc.rtm.rtm_flags = RTF_UP; 1196 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1197 arpc.sin.sin_family = AF_INET; 1198 arpc.sin.sin_len = sizeof(arpc.sin); 1199 arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr; 1200 1201 /* publish */ 1202 if (lle->la_flags & LLE_PUB) 1203 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1204 1205 sdl = &arpc.sdl; 1206 sdl->sdl_family = AF_LINK; 1207 sdl->sdl_len = sizeof(*sdl); 1208 sdl->sdl_index = ifp->if_index; 1209 sdl->sdl_type = ifp->if_type; 1210 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1211 sdl->sdl_alen = ifp->if_addrlen; 1212 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1213 } else { 1214 sdl->sdl_alen = 0; 1215 bzero(LLADDR(sdl), ifp->if_addrlen); 1216 } 1217 1218 arpc.rtm.rtm_rmx.rmx_expire = 1219 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1220 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1221 if (lle->la_flags & LLE_STATIC) 1222 arpc.rtm.rtm_flags |= RTF_STATIC; 1223 arpc.rtm.rtm_index = ifp->if_index; 1224 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1225 if (error) 1226 break; 1227 } 1228 } 1229 return error; 1230 #undef SIN 1231 } 1232 1233 void * 1234 in_domifattach(struct ifnet *ifp) 1235 { 1236 struct in_ifinfo *ii; 1237 struct lltable *llt; 1238 1239 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1240 1241 llt = lltable_init(ifp, AF_INET); 1242 if (llt != NULL) { 1243 llt->llt_prefix_free = in_lltable_prefix_free; 1244 llt->llt_lookup = in_lltable_lookup; 1245 llt->llt_dump = in_lltable_dump; 1246 } 1247 ii->ii_llt = llt; 1248 1249 ii->ii_igmp = igmp_domifattach(ifp); 1250 1251 return ii; 1252 } 1253 1254 void 1255 in_domifdetach(struct ifnet *ifp, void *aux) 1256 { 1257 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1258 1259 igmp_domifdetach(ifp); 1260 lltable_free(ii->ii_llt); 1261 free(ii, M_IFADDR); 1262 } 1263