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 * 3. 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/lock.h> 48 #include <sys/proc.h> 49 #include <sys/rmlock.h> 50 #include <sys/sysctl.h> 51 #include <sys/syslog.h> 52 #include <sys/sx.h> 53 54 #include <net/if.h> 55 #include <net/if_var.h> 56 #include <net/if_arp.h> 57 #include <net/if_dl.h> 58 #include <net/if_llatbl.h> 59 #include <net/if_types.h> 60 #include <net/route.h> 61 #include <net/vnet.h> 62 63 #include <netinet/if_ether.h> 64 #include <netinet/in.h> 65 #include <netinet/in_var.h> 66 #include <netinet/in_pcb.h> 67 #include <netinet/ip_var.h> 68 #include <netinet/ip_carp.h> 69 #include <netinet/igmp_var.h> 70 #include <netinet/udp.h> 71 #include <netinet/udp_var.h> 72 73 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *); 74 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *); 75 76 static void in_socktrim(struct sockaddr_in *); 77 static void in_purgemaddrs(struct ifnet *); 78 79 static VNET_DEFINE(int, nosameprefix); 80 #define V_nosameprefix VNET(nosameprefix) 81 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW, 82 &VNET_NAME(nosameprefix), 0, 83 "Refuse to create same prefixes on different interfaces"); 84 85 VNET_DECLARE(struct inpcbinfo, ripcbinfo); 86 #define V_ripcbinfo VNET(ripcbinfo) 87 88 static struct sx in_control_sx; 89 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control"); 90 91 /* 92 * Return 1 if an internet address is for a ``local'' host 93 * (one to which we have a connection). 94 */ 95 int 96 in_localaddr(struct in_addr in) 97 { 98 struct rm_priotracker in_ifa_tracker; 99 u_long i = ntohl(in.s_addr); 100 struct in_ifaddr *ia; 101 102 IN_IFADDR_RLOCK(&in_ifa_tracker); 103 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 104 if ((i & ia->ia_subnetmask) == ia->ia_subnet) { 105 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 106 return (1); 107 } 108 } 109 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 110 return (0); 111 } 112 113 /* 114 * Return 1 if an internet address is for the local host and configured 115 * on one of its interfaces. 116 */ 117 int 118 in_localip(struct in_addr in) 119 { 120 struct rm_priotracker in_ifa_tracker; 121 struct in_ifaddr *ia; 122 123 IN_IFADDR_RLOCK(&in_ifa_tracker); 124 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) { 125 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) { 126 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 127 return (1); 128 } 129 } 130 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 131 return (0); 132 } 133 134 /* 135 * Return 1 if an internet address is configured on an interface. 136 */ 137 int 138 in_ifhasaddr(struct ifnet *ifp, struct in_addr in) 139 { 140 struct ifaddr *ifa; 141 struct in_ifaddr *ia; 142 143 IF_ADDR_RLOCK(ifp); 144 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 145 if (ifa->ifa_addr->sa_family != AF_INET) 146 continue; 147 ia = (struct in_ifaddr *)ifa; 148 if (ia->ia_addr.sin_addr.s_addr == in.s_addr) { 149 IF_ADDR_RUNLOCK(ifp); 150 return (1); 151 } 152 } 153 IF_ADDR_RUNLOCK(ifp); 154 155 return (0); 156 } 157 158 /* 159 * Return a reference to the interface address which is different to 160 * the supplied one but with same IP address value. 161 */ 162 static struct in_ifaddr * 163 in_localip_more(struct in_ifaddr *ia) 164 { 165 struct rm_priotracker in_ifa_tracker; 166 in_addr_t in = IA_SIN(ia)->sin_addr.s_addr; 167 struct in_ifaddr *it; 168 169 IN_IFADDR_RLOCK(&in_ifa_tracker); 170 LIST_FOREACH(it, INADDR_HASH(in), ia_hash) { 171 if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) { 172 ifa_ref(&it->ia_ifa); 173 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 174 return (it); 175 } 176 } 177 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 178 179 return (NULL); 180 } 181 182 /* 183 * Determine whether an IP address is in a reserved set of addresses 184 * that may not be forwarded, or whether datagrams to that destination 185 * may be forwarded. 186 */ 187 int 188 in_canforward(struct in_addr in) 189 { 190 u_long i = ntohl(in.s_addr); 191 u_long net; 192 193 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i)) 194 return (0); 195 if (IN_CLASSA(i)) { 196 net = i & IN_CLASSA_NET; 197 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 198 return (0); 199 } 200 return (1); 201 } 202 203 /* 204 * Trim a mask in a sockaddr 205 */ 206 static void 207 in_socktrim(struct sockaddr_in *ap) 208 { 209 char *cplim = (char *) &ap->sin_addr; 210 char *cp = (char *) (&ap->sin_addr + 1); 211 212 ap->sin_len = 0; 213 while (--cp >= cplim) 214 if (*cp) { 215 (ap)->sin_len = cp - (char *) (ap) + 1; 216 break; 217 } 218 } 219 220 /* 221 * Generic internet control operations (ioctl's). 222 */ 223 int 224 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, 225 struct thread *td) 226 { 227 struct ifreq *ifr = (struct ifreq *)data; 228 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr; 229 struct ifaddr *ifa; 230 struct in_ifaddr *ia; 231 int error; 232 233 if (ifp == NULL) 234 return (EADDRNOTAVAIL); 235 236 /* 237 * Filter out 4 ioctls we implement directly. Forward the rest 238 * to specific functions and ifp->if_ioctl(). 239 */ 240 switch (cmd) { 241 case SIOCGIFADDR: 242 case SIOCGIFBRDADDR: 243 case SIOCGIFDSTADDR: 244 case SIOCGIFNETMASK: 245 break; 246 case SIOCDIFADDR: 247 sx_xlock(&in_control_sx); 248 error = in_difaddr_ioctl(cmd, data, ifp, td); 249 sx_xunlock(&in_control_sx); 250 return (error); 251 case OSIOCAIFADDR: /* 9.x compat */ 252 case SIOCAIFADDR: 253 sx_xlock(&in_control_sx); 254 error = in_aifaddr_ioctl(cmd, data, ifp, td); 255 sx_xunlock(&in_control_sx); 256 return (error); 257 case SIOCSIFADDR: 258 case SIOCSIFBRDADDR: 259 case SIOCSIFDSTADDR: 260 case SIOCSIFNETMASK: 261 /* We no longer support that old commands. */ 262 return (EINVAL); 263 default: 264 if (ifp->if_ioctl == NULL) 265 return (EOPNOTSUPP); 266 return ((*ifp->if_ioctl)(ifp, cmd, data)); 267 } 268 269 if (addr->sin_addr.s_addr != INADDR_ANY && 270 prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0) 271 return (EADDRNOTAVAIL); 272 273 /* 274 * Find address for this interface, if it exists. If an 275 * address was specified, find that one instead of the 276 * first one on the interface, if possible. 277 */ 278 IF_ADDR_RLOCK(ifp); 279 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 280 if (ifa->ifa_addr->sa_family != AF_INET) 281 continue; 282 ia = (struct in_ifaddr *)ifa; 283 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr) 284 break; 285 } 286 if (ifa == NULL) 287 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 288 if (ifa->ifa_addr->sa_family == AF_INET) { 289 ia = (struct in_ifaddr *)ifa; 290 if (prison_check_ip4(td->td_ucred, 291 &ia->ia_addr.sin_addr) == 0) 292 break; 293 } 294 295 if (ifa == NULL) { 296 IF_ADDR_RUNLOCK(ifp); 297 return (EADDRNOTAVAIL); 298 } 299 300 error = 0; 301 switch (cmd) { 302 case SIOCGIFADDR: 303 *addr = ia->ia_addr; 304 break; 305 306 case SIOCGIFBRDADDR: 307 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 308 error = EINVAL; 309 break; 310 } 311 *addr = ia->ia_broadaddr; 312 break; 313 314 case SIOCGIFDSTADDR: 315 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 316 error = EINVAL; 317 break; 318 } 319 *addr = ia->ia_dstaddr; 320 break; 321 322 case SIOCGIFNETMASK: 323 *addr = ia->ia_sockmask; 324 break; 325 } 326 327 IF_ADDR_RUNLOCK(ifp); 328 329 return (error); 330 } 331 332 static int 333 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td) 334 { 335 const struct in_aliasreq *ifra = (struct in_aliasreq *)data; 336 const struct sockaddr_in *addr = &ifra->ifra_addr; 337 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr; 338 const struct sockaddr_in *mask = &ifra->ifra_mask; 339 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr; 340 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0; 341 struct ifaddr *ifa; 342 struct in_ifaddr *ia; 343 bool iaIsFirst; 344 int error = 0; 345 346 error = priv_check(td, PRIV_NET_ADDIFADDR); 347 if (error) 348 return (error); 349 350 /* 351 * ifra_addr must be present and be of INET family. 352 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional. 353 */ 354 if (addr->sin_len != sizeof(struct sockaddr_in) || 355 addr->sin_family != AF_INET) 356 return (EINVAL); 357 if (broadaddr->sin_len != 0 && 358 (broadaddr->sin_len != sizeof(struct sockaddr_in) || 359 broadaddr->sin_family != AF_INET)) 360 return (EINVAL); 361 if (mask->sin_len != 0 && 362 (mask->sin_len != sizeof(struct sockaddr_in) || 363 mask->sin_family != AF_INET)) 364 return (EINVAL); 365 if ((ifp->if_flags & IFF_POINTOPOINT) && 366 (dstaddr->sin_len != sizeof(struct sockaddr_in) || 367 dstaddr->sin_addr.s_addr == INADDR_ANY)) 368 return (EDESTADDRREQ); 369 if (vhid > 0 && carp_attach_p == NULL) 370 return (EPROTONOSUPPORT); 371 372 /* 373 * See whether address already exist. 374 */ 375 iaIsFirst = true; 376 ia = NULL; 377 IF_ADDR_RLOCK(ifp); 378 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 379 struct in_ifaddr *it; 380 381 if (ifa->ifa_addr->sa_family != AF_INET) 382 continue; 383 384 it = (struct in_ifaddr *)ifa; 385 iaIsFirst = false; 386 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 387 prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0) 388 ia = it; 389 } 390 IF_ADDR_RUNLOCK(ifp); 391 392 if (ia != NULL) 393 (void )in_difaddr_ioctl(cmd, data, ifp, td); 394 395 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK); 396 ia = (struct in_ifaddr *)ifa; 397 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 398 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 399 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 400 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock, 401 CALLOUT_RETURNUNLOCKED); 402 403 ia->ia_ifp = ifp; 404 ia->ia_addr = *addr; 405 if (mask->sin_len != 0) { 406 ia->ia_sockmask = *mask; 407 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 408 } else { 409 in_addr_t i = ntohl(addr->sin_addr.s_addr); 410 411 /* 412 * Be compatible with network classes, if netmask isn't 413 * supplied, guess it based on classes. 414 */ 415 if (IN_CLASSA(i)) 416 ia->ia_subnetmask = IN_CLASSA_NET; 417 else if (IN_CLASSB(i)) 418 ia->ia_subnetmask = IN_CLASSB_NET; 419 else 420 ia->ia_subnetmask = IN_CLASSC_NET; 421 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 422 } 423 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask; 424 in_socktrim(&ia->ia_sockmask); 425 426 if (ifp->if_flags & IFF_BROADCAST) { 427 if (broadaddr->sin_len != 0) { 428 ia->ia_broadaddr = *broadaddr; 429 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) { 430 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 431 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 432 ia->ia_broadaddr.sin_family = AF_INET; 433 } else { 434 ia->ia_broadaddr.sin_addr.s_addr = 435 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 436 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 437 ia->ia_broadaddr.sin_family = AF_INET; 438 } 439 } 440 441 if (ifp->if_flags & IFF_POINTOPOINT) 442 ia->ia_dstaddr = *dstaddr; 443 444 /* XXXGL: rtinit() needs this strange assignment. */ 445 if (ifp->if_flags & IFF_LOOPBACK) 446 ia->ia_dstaddr = ia->ia_addr; 447 448 if (vhid != 0) { 449 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 450 if (error) 451 return (error); 452 } 453 454 /* if_addrhead is already referenced by ifa_alloc() */ 455 IF_ADDR_WLOCK(ifp); 456 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 457 IF_ADDR_WUNLOCK(ifp); 458 459 ifa_ref(ifa); /* in_ifaddrhead */ 460 IN_IFADDR_WLOCK(); 461 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 462 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash); 463 IN_IFADDR_WUNLOCK(); 464 465 /* 466 * Give the interface a chance to initialize 467 * if this is its first address, 468 * and to validate the address if necessary. 469 */ 470 if (ifp->if_ioctl != NULL) { 471 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 472 if (error) 473 goto fail1; 474 } 475 476 /* 477 * Add route for the network. 478 */ 479 if (vhid == 0) { 480 int flags = RTF_UP; 481 482 if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT)) 483 flags |= RTF_HOST; 484 485 error = in_addprefix(ia, flags); 486 if (error) 487 goto fail1; 488 } 489 490 /* 491 * Add a loopback route to self. 492 */ 493 if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 && 494 ia->ia_addr.sin_addr.s_addr != INADDR_ANY && 495 !((ifp->if_flags & IFF_POINTOPOINT) && 496 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) { 497 struct in_ifaddr *eia; 498 499 eia = in_localip_more(ia); 500 501 if (eia == NULL) { 502 error = ifa_add_loopback_route((struct ifaddr *)ia, 503 (struct sockaddr *)&ia->ia_addr); 504 if (error) 505 goto fail2; 506 } else 507 ifa_free(&eia->ia_ifa); 508 } 509 510 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 511 struct in_addr allhosts_addr; 512 struct in_ifinfo *ii; 513 514 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 515 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 516 517 error = in_joingroup(ifp, &allhosts_addr, NULL, 518 &ii->ii_allhosts); 519 } 520 521 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 522 523 return (error); 524 525 fail2: 526 if (vhid == 0) 527 (void )in_scrubprefix(ia, LLE_STATIC); 528 529 fail1: 530 if (ia->ia_ifa.ifa_carp) 531 (*carp_detach_p)(&ia->ia_ifa, false); 532 533 IF_ADDR_WLOCK(ifp); 534 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 535 IF_ADDR_WUNLOCK(ifp); 536 ifa_free(&ia->ia_ifa); /* if_addrhead */ 537 538 IN_IFADDR_WLOCK(); 539 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 540 LIST_REMOVE(ia, ia_hash); 541 IN_IFADDR_WUNLOCK(); 542 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 543 544 return (error); 545 } 546 547 static int 548 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td) 549 { 550 const struct ifreq *ifr = (struct ifreq *)data; 551 const struct sockaddr_in *addr = (const struct sockaddr_in *) 552 &ifr->ifr_addr; 553 struct ifaddr *ifa; 554 struct in_ifaddr *ia; 555 bool deleteAny, iaIsLast; 556 int error; 557 558 if (td != NULL) { 559 error = priv_check(td, PRIV_NET_DELIFADDR); 560 if (error) 561 return (error); 562 } 563 564 if (addr->sin_len != sizeof(struct sockaddr_in) || 565 addr->sin_family != AF_INET) 566 deleteAny = true; 567 else 568 deleteAny = false; 569 570 iaIsLast = true; 571 ia = NULL; 572 IF_ADDR_WLOCK(ifp); 573 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 574 struct in_ifaddr *it; 575 576 if (ifa->ifa_addr->sa_family != AF_INET) 577 continue; 578 579 it = (struct in_ifaddr *)ifa; 580 if (deleteAny && ia == NULL && (td == NULL || 581 prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0)) 582 ia = it; 583 584 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 585 (td == NULL || prison_check_ip4(td->td_ucred, 586 &addr->sin_addr) == 0)) 587 ia = it; 588 589 if (it != ia) 590 iaIsLast = false; 591 } 592 593 if (ia == NULL) { 594 IF_ADDR_WUNLOCK(ifp); 595 return (EADDRNOTAVAIL); 596 } 597 598 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 599 IF_ADDR_WUNLOCK(ifp); 600 ifa_free(&ia->ia_ifa); /* if_addrhead */ 601 602 IN_IFADDR_WLOCK(); 603 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 604 LIST_REMOVE(ia, ia_hash); 605 IN_IFADDR_WUNLOCK(); 606 607 /* 608 * in_scrubprefix() kills the interface route. 609 */ 610 in_scrubprefix(ia, LLE_STATIC); 611 612 /* 613 * in_ifadown gets rid of all the rest of 614 * the routes. This is not quite the right 615 * thing to do, but at least if we are running 616 * a routing process they will come back. 617 */ 618 in_ifadown(&ia->ia_ifa, 1); 619 620 if (ia->ia_ifa.ifa_carp) 621 (*carp_detach_p)(&ia->ia_ifa, 622 (cmd == SIOCDIFADDR) ? false : true); 623 624 /* 625 * If this is the last IPv4 address configured on this 626 * interface, leave the all-hosts group. 627 * No state-change report need be transmitted. 628 */ 629 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 630 struct in_ifinfo *ii; 631 632 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 633 IN_MULTI_LOCK(); 634 if (ii->ii_allhosts) { 635 (void)in_leavegroup_locked(ii->ii_allhosts, NULL); 636 ii->ii_allhosts = NULL; 637 } 638 IN_MULTI_UNLOCK(); 639 } 640 641 IF_ADDR_WLOCK(ifp); 642 if (callout_stop(&ia->ia_garp_timer) == 1) { 643 ifa_free(&ia->ia_ifa); 644 } 645 IF_ADDR_WUNLOCK(ifp); 646 647 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 648 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 649 650 return (0); 651 } 652 653 #define rtinitflags(x) \ 654 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 655 ? RTF_HOST : 0) 656 657 /* 658 * Check if we have a route for the given prefix already or add one accordingly. 659 */ 660 int 661 in_addprefix(struct in_ifaddr *target, int flags) 662 { 663 struct rm_priotracker in_ifa_tracker; 664 struct in_ifaddr *ia; 665 struct in_addr prefix, mask, p, m; 666 int error; 667 668 if ((flags & RTF_HOST) != 0) { 669 prefix = target->ia_dstaddr.sin_addr; 670 mask.s_addr = 0; 671 } else { 672 prefix = target->ia_addr.sin_addr; 673 mask = target->ia_sockmask.sin_addr; 674 prefix.s_addr &= mask.s_addr; 675 } 676 677 IN_IFADDR_RLOCK(&in_ifa_tracker); 678 /* Look for an existing address with the same prefix, mask, and fib */ 679 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 680 if (rtinitflags(ia)) { 681 p = ia->ia_dstaddr.sin_addr; 682 683 if (prefix.s_addr != p.s_addr) 684 continue; 685 } else { 686 p = ia->ia_addr.sin_addr; 687 m = ia->ia_sockmask.sin_addr; 688 p.s_addr &= m.s_addr; 689 690 if (prefix.s_addr != p.s_addr || 691 mask.s_addr != m.s_addr) 692 continue; 693 } 694 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 695 continue; 696 697 /* 698 * If we got a matching prefix route inserted by other 699 * interface address, we are done here. 700 */ 701 if (ia->ia_flags & IFA_ROUTE) { 702 #ifdef RADIX_MPATH 703 if (ia->ia_addr.sin_addr.s_addr == 704 target->ia_addr.sin_addr.s_addr) { 705 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 706 return (EEXIST); 707 } else 708 break; 709 #endif 710 if (V_nosameprefix) { 711 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 712 return (EEXIST); 713 } else { 714 int fibnum; 715 716 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : 717 target->ia_ifp->if_fib; 718 rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum); 719 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 720 return (0); 721 } 722 } 723 } 724 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 725 726 /* 727 * No-one seem to have this prefix route, so we try to insert it. 728 */ 729 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 730 if (!error) 731 target->ia_flags |= IFA_ROUTE; 732 return (error); 733 } 734 735 /* 736 * Removes either all lle entries for given @ia, or lle 737 * corresponding to @ia address. 738 */ 739 static void 740 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags) 741 { 742 struct sockaddr_in addr, mask; 743 struct sockaddr *saddr, *smask; 744 struct ifnet *ifp; 745 746 saddr = (struct sockaddr *)&addr; 747 bzero(&addr, sizeof(addr)); 748 addr.sin_len = sizeof(addr); 749 addr.sin_family = AF_INET; 750 smask = (struct sockaddr *)&mask; 751 bzero(&mask, sizeof(mask)); 752 mask.sin_len = sizeof(mask); 753 mask.sin_family = AF_INET; 754 mask.sin_addr.s_addr = ia->ia_subnetmask; 755 ifp = ia->ia_ifp; 756 757 if (all) { 758 759 /* 760 * Remove all L2 entries matching given prefix. 761 * Convert address to host representation to avoid 762 * doing this on every callback. ia_subnetmask is already 763 * stored in host representation. 764 */ 765 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr); 766 lltable_prefix_free(AF_INET, saddr, smask, flags); 767 } else { 768 /* Remove interface address only */ 769 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr; 770 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr); 771 } 772 } 773 774 /* 775 * If there is no other address in the system that can serve a route to the 776 * same prefix, remove the route. Hand over the route to the new address 777 * otherwise. 778 */ 779 int 780 in_scrubprefix(struct in_ifaddr *target, u_int flags) 781 { 782 struct rm_priotracker in_ifa_tracker; 783 struct in_ifaddr *ia; 784 struct in_addr prefix, mask, p, m; 785 int error = 0; 786 787 /* 788 * Remove the loopback route to the interface address. 789 */ 790 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 791 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 792 (flags & LLE_STATIC)) { 793 struct in_ifaddr *eia; 794 795 /* 796 * XXXME: add fib-aware in_localip. 797 * We definitely don't want to switch between 798 * prefixes in different fibs. 799 */ 800 eia = in_localip_more(target); 801 802 if (eia != NULL) { 803 error = ifa_switch_loopback_route((struct ifaddr *)eia, 804 (struct sockaddr *)&target->ia_addr); 805 ifa_free(&eia->ia_ifa); 806 } else { 807 error = ifa_del_loopback_route((struct ifaddr *)target, 808 (struct sockaddr *)&target->ia_addr); 809 } 810 } 811 812 if (rtinitflags(target)) { 813 prefix = target->ia_dstaddr.sin_addr; 814 mask.s_addr = 0; 815 } else { 816 prefix = target->ia_addr.sin_addr; 817 mask = target->ia_sockmask.sin_addr; 818 prefix.s_addr &= mask.s_addr; 819 } 820 821 if ((target->ia_flags & IFA_ROUTE) == 0) { 822 int fibnum; 823 824 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : 825 target->ia_ifp->if_fib; 826 rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum); 827 828 /* 829 * Removing address from !IFF_UP interface or 830 * prefix which exists on other interface (along with route). 831 * No entries should exist here except target addr. 832 * Given that, delete this entry only. 833 */ 834 in_scrubprefixlle(target, 0, flags); 835 return (0); 836 } 837 838 IN_IFADDR_RLOCK(&in_ifa_tracker); 839 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 840 if (rtinitflags(ia)) { 841 p = ia->ia_dstaddr.sin_addr; 842 843 if (prefix.s_addr != p.s_addr) 844 continue; 845 } else { 846 p = ia->ia_addr.sin_addr; 847 m = ia->ia_sockmask.sin_addr; 848 p.s_addr &= m.s_addr; 849 850 if (prefix.s_addr != p.s_addr || 851 mask.s_addr != m.s_addr) 852 continue; 853 } 854 855 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 856 continue; 857 858 /* 859 * If we got a matching prefix address, move IFA_ROUTE and 860 * the route itself to it. Make sure that routing daemons 861 * get a heads-up. 862 */ 863 if ((ia->ia_flags & IFA_ROUTE) == 0) { 864 ifa_ref(&ia->ia_ifa); 865 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 866 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 867 rtinitflags(target)); 868 if (error == 0) 869 target->ia_flags &= ~IFA_ROUTE; 870 else 871 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 872 error); 873 /* Scrub all entries IFF interface is different */ 874 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp, 875 flags); 876 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 877 rtinitflags(ia) | RTF_UP); 878 if (error == 0) 879 ia->ia_flags |= IFA_ROUTE; 880 else 881 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 882 error); 883 ifa_free(&ia->ia_ifa); 884 return (error); 885 } 886 } 887 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 888 889 /* 890 * remove all L2 entries on the given prefix 891 */ 892 in_scrubprefixlle(target, 1, flags); 893 894 /* 895 * As no-one seem to have this prefix, we can remove the route. 896 */ 897 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 898 if (error == 0) 899 target->ia_flags &= ~IFA_ROUTE; 900 else 901 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 902 return (error); 903 } 904 905 #undef rtinitflags 906 907 void 908 in_ifscrub_all(void) 909 { 910 struct ifnet *ifp; 911 struct ifaddr *ifa, *nifa; 912 struct ifaliasreq ifr; 913 914 IFNET_RLOCK(); 915 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 916 /* Cannot lock here - lock recursion. */ 917 /* IF_ADDR_RLOCK(ifp); */ 918 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 919 if (ifa->ifa_addr->sa_family != AF_INET) 920 continue; 921 922 /* 923 * This is ugly but the only way for legacy IP to 924 * cleanly remove addresses and everything attached. 925 */ 926 bzero(&ifr, sizeof(ifr)); 927 ifr.ifra_addr = *ifa->ifa_addr; 928 if (ifa->ifa_dstaddr) 929 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 930 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, 931 ifp, NULL); 932 } 933 /* IF_ADDR_RUNLOCK(ifp); */ 934 in_purgemaddrs(ifp); 935 igmp_domifdetach(ifp); 936 } 937 IFNET_RUNLOCK(); 938 } 939 940 int 941 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia) 942 { 943 944 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 945 /* 946 * Check for old-style (host 0) broadcast, but 947 * taking into account that RFC 3021 obsoletes it. 948 */ 949 (ia->ia_subnetmask != IN_RFC3021_MASK && 950 ntohl(in.s_addr) == ia->ia_subnet)) && 951 /* 952 * Check for an all one subnetmask. These 953 * only exist when an interface gets a secondary 954 * address. 955 */ 956 ia->ia_subnetmask != (u_long)0xffffffff); 957 } 958 959 /* 960 * Return 1 if the address might be a local broadcast address. 961 */ 962 int 963 in_broadcast(struct in_addr in, struct ifnet *ifp) 964 { 965 struct ifaddr *ifa; 966 int found; 967 968 if (in.s_addr == INADDR_BROADCAST || 969 in.s_addr == INADDR_ANY) 970 return (1); 971 if ((ifp->if_flags & IFF_BROADCAST) == 0) 972 return (0); 973 found = 0; 974 /* 975 * Look through the list of addresses for a match 976 * with a broadcast address. 977 */ 978 IF_ADDR_RLOCK(ifp); 979 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 980 if (ifa->ifa_addr->sa_family == AF_INET && 981 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) { 982 found = 1; 983 break; 984 } 985 IF_ADDR_RUNLOCK(ifp); 986 return (found); 987 } 988 989 /* 990 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 991 */ 992 void 993 in_ifdetach(struct ifnet *ifp) 994 { 995 996 in_pcbpurgeif0(&V_ripcbinfo, ifp); 997 in_pcbpurgeif0(&V_udbinfo, ifp); 998 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 999 in_purgemaddrs(ifp); 1000 } 1001 1002 /* 1003 * Delete all IPv4 multicast address records, and associated link-layer 1004 * multicast address records, associated with ifp. 1005 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1006 * XXX This should not race with ifma_protospec being set during 1007 * a new allocation, if it does, we have bigger problems. 1008 */ 1009 static void 1010 in_purgemaddrs(struct ifnet *ifp) 1011 { 1012 LIST_HEAD(,in_multi) purgeinms; 1013 struct in_multi *inm, *tinm; 1014 struct ifmultiaddr *ifma; 1015 1016 LIST_INIT(&purgeinms); 1017 IN_MULTI_LOCK(); 1018 1019 /* 1020 * Extract list of in_multi associated with the detaching ifp 1021 * which the PF_INET layer is about to release. 1022 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1023 * by code further down. 1024 */ 1025 IF_ADDR_RLOCK(ifp); 1026 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1027 if (ifma->ifma_addr->sa_family != AF_INET || 1028 ifma->ifma_protospec == NULL) 1029 continue; 1030 #if 0 1031 KASSERT(ifma->ifma_protospec != NULL, 1032 ("%s: ifma_protospec is NULL", __func__)); 1033 #endif 1034 inm = (struct in_multi *)ifma->ifma_protospec; 1035 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 1036 } 1037 IF_ADDR_RUNLOCK(ifp); 1038 1039 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 1040 LIST_REMOVE(inm, inm_link); 1041 inm_release_locked(inm); 1042 } 1043 igmp_ifdetach(ifp); 1044 1045 IN_MULTI_UNLOCK(); 1046 } 1047 1048 struct in_llentry { 1049 struct llentry base; 1050 }; 1051 1052 #define IN_LLTBL_DEFAULT_HSIZE 32 1053 #define IN_LLTBL_HASH(k, h) \ 1054 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1055 1056 /* 1057 * Do actual deallocation of @lle. 1058 */ 1059 static void 1060 in_lltable_destroy_lle_unlocked(struct llentry *lle) 1061 { 1062 1063 LLE_LOCK_DESTROY(lle); 1064 LLE_REQ_DESTROY(lle); 1065 free(lle, M_LLTABLE); 1066 } 1067 1068 /* 1069 * Called by LLE_FREE_LOCKED when number of references 1070 * drops to zero. 1071 */ 1072 static void 1073 in_lltable_destroy_lle(struct llentry *lle) 1074 { 1075 1076 LLE_WUNLOCK(lle); 1077 in_lltable_destroy_lle_unlocked(lle); 1078 } 1079 1080 static struct llentry * 1081 in_lltable_new(struct in_addr addr4, u_int flags) 1082 { 1083 struct in_llentry *lle; 1084 1085 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1086 if (lle == NULL) /* NB: caller generates msg */ 1087 return NULL; 1088 1089 /* 1090 * For IPv4 this will trigger "arpresolve" to generate 1091 * an ARP request. 1092 */ 1093 lle->base.la_expire = time_uptime; /* mark expired */ 1094 lle->base.r_l3addr.addr4 = addr4; 1095 lle->base.lle_refcnt = 1; 1096 lle->base.lle_free = in_lltable_destroy_lle; 1097 LLE_LOCK_INIT(&lle->base); 1098 LLE_REQ_INIT(&lle->base); 1099 callout_init(&lle->base.lle_timer, 1); 1100 1101 return (&lle->base); 1102 } 1103 1104 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1105 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 ) 1106 1107 static int 1108 in_lltable_match_prefix(const struct sockaddr *saddr, 1109 const struct sockaddr *smask, u_int flags, struct llentry *lle) 1110 { 1111 struct in_addr addr, mask, lle_addr; 1112 1113 addr = ((const struct sockaddr_in *)saddr)->sin_addr; 1114 mask = ((const struct sockaddr_in *)smask)->sin_addr; 1115 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1116 1117 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0) 1118 return (0); 1119 1120 if (lle->la_flags & LLE_IFADDR) { 1121 1122 /* 1123 * Delete LLE_IFADDR records IFF address & flag matches. 1124 * Note that addr is the interface address within prefix 1125 * being matched. 1126 * Note also we should handle 'ifdown' cases without removing 1127 * ifaddr macs. 1128 */ 1129 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0) 1130 return (1); 1131 return (0); 1132 } 1133 1134 /* flags & LLE_STATIC means deleting both dynamic and static entries */ 1135 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)) 1136 return (1); 1137 1138 return (0); 1139 } 1140 1141 static void 1142 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1143 { 1144 struct ifnet *ifp; 1145 size_t pkts_dropped; 1146 1147 LLE_WLOCK_ASSERT(lle); 1148 KASSERT(llt != NULL, ("lltable is NULL")); 1149 1150 /* Unlink entry from table if not already */ 1151 if ((lle->la_flags & LLE_LINKED) != 0) { 1152 ifp = llt->llt_ifp; 1153 IF_AFDATA_WLOCK_ASSERT(ifp); 1154 lltable_unlink_entry(llt, lle); 1155 } 1156 1157 /* cancel timer */ 1158 if (callout_stop(&lle->lle_timer) > 0) 1159 LLE_REMREF(lle); 1160 1161 /* Drop hold queue */ 1162 pkts_dropped = llentry_free(lle); 1163 ARPSTAT_ADD(dropped, pkts_dropped); 1164 } 1165 1166 static int 1167 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1168 { 1169 struct rt_addrinfo info; 1170 struct sockaddr_in rt_key, rt_mask; 1171 struct sockaddr rt_gateway; 1172 int rt_flags; 1173 1174 KASSERT(l3addr->sa_family == AF_INET, 1175 ("sin_family %d", l3addr->sa_family)); 1176 1177 bzero(&rt_key, sizeof(rt_key)); 1178 rt_key.sin_len = sizeof(rt_key); 1179 bzero(&rt_mask, sizeof(rt_mask)); 1180 rt_mask.sin_len = sizeof(rt_mask); 1181 bzero(&rt_gateway, sizeof(rt_gateway)); 1182 rt_gateway.sa_len = sizeof(rt_gateway); 1183 1184 bzero(&info, sizeof(info)); 1185 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; 1186 info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask; 1187 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway; 1188 1189 if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0) 1190 return (EINVAL); 1191 1192 rt_flags = info.rti_flags; 1193 1194 /* 1195 * If the gateway for an existing host route matches the target L3 1196 * address, which is a special route inserted by some implementation 1197 * such as MANET, and the interface is of the correct type, then 1198 * allow for ARP to proceed. 1199 */ 1200 if (rt_flags & RTF_GATEWAY) { 1201 if (!(rt_flags & RTF_HOST) || !info.rti_ifp || 1202 info.rti_ifp->if_type != IFT_ETHER || 1203 (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1204 memcmp(rt_gateway.sa_data, l3addr->sa_data, 1205 sizeof(in_addr_t)) != 0) { 1206 rib_free_info(&info); 1207 return (EINVAL); 1208 } 1209 } 1210 rib_free_info(&info); 1211 1212 /* 1213 * Make sure that at least the destination address is covered 1214 * by the route. This is for handling the case where 2 or more 1215 * interfaces have the same prefix. An incoming packet arrives 1216 * on one interface and the corresponding outgoing packet leaves 1217 * another interface. 1218 */ 1219 if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) { 1220 const char *sa, *mask, *addr, *lim; 1221 const struct sockaddr_in *l3sin; 1222 1223 mask = (const char *)&rt_mask; 1224 /* 1225 * Just being extra cautious to avoid some custom 1226 * code getting into trouble. 1227 */ 1228 if ((info.rti_addrs & RTA_NETMASK) == 0) 1229 return (EINVAL); 1230 1231 sa = (const char *)&rt_key; 1232 addr = (const char *)l3addr; 1233 l3sin = (const struct sockaddr_in *)l3addr; 1234 lim = addr + l3sin->sin_len; 1235 1236 for ( ; addr < lim; sa++, mask++, addr++) { 1237 if ((*sa ^ *addr) & *mask) { 1238 #ifdef DIAGNOSTIC 1239 char addrbuf[INET_ADDRSTRLEN]; 1240 1241 log(LOG_INFO, "IPv4 address: \"%s\" " 1242 "is not on the network\n", 1243 inet_ntoa_r(l3sin->sin_addr, addrbuf)); 1244 #endif 1245 return (EINVAL); 1246 } 1247 } 1248 } 1249 1250 return (0); 1251 } 1252 1253 static inline uint32_t 1254 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 1255 { 1256 1257 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 1258 } 1259 1260 static uint32_t 1261 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 1262 { 1263 1264 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 1265 } 1266 1267 static void 1268 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 1269 { 1270 struct sockaddr_in *sin; 1271 1272 sin = (struct sockaddr_in *)sa; 1273 bzero(sin, sizeof(*sin)); 1274 sin->sin_family = AF_INET; 1275 sin->sin_len = sizeof(*sin); 1276 sin->sin_addr = lle->r_l3addr.addr4; 1277 } 1278 1279 static inline struct llentry * 1280 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 1281 { 1282 struct llentry *lle; 1283 struct llentries *lleh; 1284 u_int hashidx; 1285 1286 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 1287 lleh = &llt->lle_head[hashidx]; 1288 LIST_FOREACH(lle, lleh, lle_next) { 1289 if (lle->la_flags & LLE_DELETED) 1290 continue; 1291 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 1292 break; 1293 } 1294 1295 return (lle); 1296 } 1297 1298 static void 1299 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle) 1300 { 1301 1302 lle->la_flags |= LLE_DELETED; 1303 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1304 #ifdef DIAGNOSTIC 1305 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1306 #endif 1307 llentry_free(lle); 1308 } 1309 1310 static struct llentry * 1311 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1312 { 1313 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1314 struct ifnet *ifp = llt->llt_ifp; 1315 struct llentry *lle; 1316 char linkhdr[LLE_MAX_LINKHDR]; 1317 size_t linkhdrsize; 1318 int lladdr_off; 1319 1320 KASSERT(l3addr->sa_family == AF_INET, 1321 ("sin_family %d", l3addr->sa_family)); 1322 1323 /* 1324 * A route that covers the given address must have 1325 * been installed 1st because we are doing a resolution, 1326 * verify this. 1327 */ 1328 if (!(flags & LLE_IFADDR) && 1329 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1330 return (NULL); 1331 1332 lle = in_lltable_new(sin->sin_addr, flags); 1333 if (lle == NULL) { 1334 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1335 return (NULL); 1336 } 1337 lle->la_flags = flags; 1338 if (flags & LLE_STATIC) 1339 lle->r_flags |= RLLE_VALID; 1340 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 1341 linkhdrsize = LLE_MAX_LINKHDR; 1342 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp), 1343 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1344 in_lltable_destroy_lle_unlocked(lle); 1345 return (NULL); 1346 } 1347 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 1348 lladdr_off); 1349 lle->la_flags |= LLE_STATIC; 1350 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR); 1351 } 1352 1353 return (lle); 1354 } 1355 1356 /* 1357 * Return NULL if not found or marked for deletion. 1358 * If found return lle read locked. 1359 */ 1360 static struct llentry * 1361 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1362 { 1363 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1364 struct llentry *lle; 1365 1366 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 1367 KASSERT(l3addr->sa_family == AF_INET, 1368 ("sin_family %d", l3addr->sa_family)); 1369 lle = in_lltable_find_dst(llt, sin->sin_addr); 1370 1371 if (lle == NULL) 1372 return (NULL); 1373 1374 KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) != 1375 (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X", 1376 flags)); 1377 1378 if (flags & LLE_UNLOCKED) 1379 return (lle); 1380 1381 if (flags & LLE_EXCLUSIVE) 1382 LLE_WLOCK(lle); 1383 else 1384 LLE_RLOCK(lle); 1385 1386 return (lle); 1387 } 1388 1389 static int 1390 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 1391 struct sysctl_req *wr) 1392 { 1393 struct ifnet *ifp = llt->llt_ifp; 1394 /* XXX stack use */ 1395 struct { 1396 struct rt_msghdr rtm; 1397 struct sockaddr_in sin; 1398 struct sockaddr_dl sdl; 1399 } arpc; 1400 struct sockaddr_dl *sdl; 1401 int error; 1402 1403 bzero(&arpc, sizeof(arpc)); 1404 /* skip deleted entries */ 1405 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1406 return (0); 1407 /* Skip if jailed and not a valid IP of the prison. */ 1408 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin); 1409 if (prison_if(wr->td->td_ucred, 1410 (struct sockaddr *)&arpc.sin) != 0) 1411 return (0); 1412 /* 1413 * produce a msg made of: 1414 * struct rt_msghdr; 1415 * struct sockaddr_in; (IPv4) 1416 * struct sockaddr_dl; 1417 */ 1418 arpc.rtm.rtm_msglen = sizeof(arpc); 1419 arpc.rtm.rtm_version = RTM_VERSION; 1420 arpc.rtm.rtm_type = RTM_GET; 1421 arpc.rtm.rtm_flags = RTF_UP; 1422 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1423 1424 /* publish */ 1425 if (lle->la_flags & LLE_PUB) 1426 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1427 1428 sdl = &arpc.sdl; 1429 sdl->sdl_family = AF_LINK; 1430 sdl->sdl_len = sizeof(*sdl); 1431 sdl->sdl_index = ifp->if_index; 1432 sdl->sdl_type = ifp->if_type; 1433 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1434 sdl->sdl_alen = ifp->if_addrlen; 1435 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1436 } else { 1437 sdl->sdl_alen = 0; 1438 bzero(LLADDR(sdl), ifp->if_addrlen); 1439 } 1440 1441 arpc.rtm.rtm_rmx.rmx_expire = 1442 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1443 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1444 if (lle->la_flags & LLE_STATIC) 1445 arpc.rtm.rtm_flags |= RTF_STATIC; 1446 if (lle->la_flags & LLE_IFADDR) 1447 arpc.rtm.rtm_flags |= RTF_PINNED; 1448 arpc.rtm.rtm_index = ifp->if_index; 1449 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1450 1451 return (error); 1452 } 1453 1454 static struct lltable * 1455 in_lltattach(struct ifnet *ifp) 1456 { 1457 struct lltable *llt; 1458 1459 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 1460 llt->llt_af = AF_INET; 1461 llt->llt_ifp = ifp; 1462 1463 llt->llt_lookup = in_lltable_lookup; 1464 llt->llt_alloc_entry = in_lltable_alloc; 1465 llt->llt_delete_entry = in_lltable_delete_entry; 1466 llt->llt_dump_entry = in_lltable_dump_entry; 1467 llt->llt_hash = in_lltable_hash; 1468 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 1469 llt->llt_free_entry = in_lltable_free_entry; 1470 llt->llt_match_prefix = in_lltable_match_prefix; 1471 lltable_link(llt); 1472 1473 return (llt); 1474 } 1475 1476 void * 1477 in_domifattach(struct ifnet *ifp) 1478 { 1479 struct in_ifinfo *ii; 1480 1481 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1482 1483 ii->ii_llt = in_lltattach(ifp); 1484 ii->ii_igmp = igmp_domifattach(ifp); 1485 1486 return (ii); 1487 } 1488 1489 void 1490 in_domifdetach(struct ifnet *ifp, void *aux) 1491 { 1492 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1493 1494 igmp_domifdetach(ifp); 1495 lltable_free(ii->ii_llt); 1496 free(ii, M_IFADDR); 1497 } 1498