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/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(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 register u_long i = ntohl(in.s_addr); 100 register 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 register u_long i = ntohl(in.s_addr); 191 register 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 register char *cplim = (char *) &ap->sin_addr; 210 register 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(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(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; 400 401 ia->ia_ifp = ifp; 402 ia->ia_addr = *addr; 403 if (mask->sin_len != 0) { 404 ia->ia_sockmask = *mask; 405 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 406 } else { 407 in_addr_t i = ntohl(addr->sin_addr.s_addr); 408 409 /* 410 * Be compatible with network classes, if netmask isn't 411 * supplied, guess it based on classes. 412 */ 413 if (IN_CLASSA(i)) 414 ia->ia_subnetmask = IN_CLASSA_NET; 415 else if (IN_CLASSB(i)) 416 ia->ia_subnetmask = IN_CLASSB_NET; 417 else 418 ia->ia_subnetmask = IN_CLASSC_NET; 419 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 420 } 421 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask; 422 in_socktrim(&ia->ia_sockmask); 423 424 if (ifp->if_flags & IFF_BROADCAST) { 425 if (broadaddr->sin_len != 0) { 426 ia->ia_broadaddr = *broadaddr; 427 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) { 428 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 429 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 430 ia->ia_broadaddr.sin_family = AF_INET; 431 } else { 432 ia->ia_broadaddr.sin_addr.s_addr = 433 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 434 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 435 ia->ia_broadaddr.sin_family = AF_INET; 436 } 437 } 438 439 if (ifp->if_flags & IFF_POINTOPOINT) 440 ia->ia_dstaddr = *dstaddr; 441 442 /* XXXGL: rtinit() needs this strange assignment. */ 443 if (ifp->if_flags & IFF_LOOPBACK) 444 ia->ia_dstaddr = ia->ia_addr; 445 446 if (vhid != 0) { 447 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 448 if (error) 449 return (error); 450 } 451 452 /* if_addrhead is already referenced by ifa_alloc() */ 453 IF_ADDR_WLOCK(ifp); 454 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 455 IF_ADDR_WUNLOCK(ifp); 456 457 ifa_ref(ifa); /* in_ifaddrhead */ 458 IN_IFADDR_WLOCK(); 459 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 460 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash); 461 IN_IFADDR_WUNLOCK(); 462 463 /* 464 * Give the interface a chance to initialize 465 * if this is its first address, 466 * and to validate the address if necessary. 467 */ 468 if (ifp->if_ioctl != NULL) { 469 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 470 if (error) 471 goto fail1; 472 } 473 474 /* 475 * Add route for the network. 476 */ 477 if (vhid == 0) { 478 int flags = RTF_UP; 479 480 if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT)) 481 flags |= RTF_HOST; 482 483 error = in_addprefix(ia, flags); 484 if (error) 485 goto fail1; 486 } 487 488 /* 489 * Add a loopback route to self. 490 */ 491 if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 && 492 ia->ia_addr.sin_addr.s_addr != INADDR_ANY && 493 !((ifp->if_flags & IFF_POINTOPOINT) && 494 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) { 495 struct in_ifaddr *eia; 496 497 eia = in_localip_more(ia); 498 499 if (eia == NULL) { 500 error = ifa_add_loopback_route((struct ifaddr *)ia, 501 (struct sockaddr *)&ia->ia_addr); 502 if (error) 503 goto fail2; 504 } else 505 ifa_free(&eia->ia_ifa); 506 } 507 508 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 509 struct in_addr allhosts_addr; 510 struct in_ifinfo *ii; 511 512 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 513 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 514 515 error = in_joingroup(ifp, &allhosts_addr, NULL, 516 &ii->ii_allhosts); 517 } 518 519 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 520 521 return (error); 522 523 fail2: 524 if (vhid == 0) 525 (void )in_scrubprefix(ia, LLE_STATIC); 526 527 fail1: 528 if (ia->ia_ifa.ifa_carp) 529 (*carp_detach_p)(&ia->ia_ifa); 530 531 IF_ADDR_WLOCK(ifp); 532 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 533 IF_ADDR_WUNLOCK(ifp); 534 ifa_free(&ia->ia_ifa); /* if_addrhead */ 535 536 IN_IFADDR_WLOCK(); 537 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 538 LIST_REMOVE(ia, ia_hash); 539 IN_IFADDR_WUNLOCK(); 540 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 541 542 return (error); 543 } 544 545 static int 546 in_difaddr_ioctl(caddr_t data, struct ifnet *ifp, struct thread *td) 547 { 548 const struct ifreq *ifr = (struct ifreq *)data; 549 const struct sockaddr_in *addr = (const struct sockaddr_in *) 550 &ifr->ifr_addr; 551 struct ifaddr *ifa; 552 struct in_ifaddr *ia; 553 bool deleteAny, iaIsLast; 554 int error; 555 556 if (td != NULL) { 557 error = priv_check(td, PRIV_NET_DELIFADDR); 558 if (error) 559 return (error); 560 } 561 562 if (addr->sin_len != sizeof(struct sockaddr_in) || 563 addr->sin_family != AF_INET) 564 deleteAny = true; 565 else 566 deleteAny = false; 567 568 iaIsLast = true; 569 ia = NULL; 570 IF_ADDR_WLOCK(ifp); 571 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 572 struct in_ifaddr *it; 573 574 if (ifa->ifa_addr->sa_family != AF_INET) 575 continue; 576 577 it = (struct in_ifaddr *)ifa; 578 if (deleteAny && ia == NULL && (td == NULL || 579 prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0)) 580 ia = it; 581 582 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 583 (td == NULL || prison_check_ip4(td->td_ucred, 584 &addr->sin_addr) == 0)) 585 ia = it; 586 587 if (it != ia) 588 iaIsLast = false; 589 } 590 591 if (ia == NULL) { 592 IF_ADDR_WUNLOCK(ifp); 593 return (EADDRNOTAVAIL); 594 } 595 596 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 597 IF_ADDR_WUNLOCK(ifp); 598 ifa_free(&ia->ia_ifa); /* if_addrhead */ 599 600 IN_IFADDR_WLOCK(); 601 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 602 LIST_REMOVE(ia, ia_hash); 603 IN_IFADDR_WUNLOCK(); 604 605 /* 606 * in_scrubprefix() kills the interface route. 607 */ 608 in_scrubprefix(ia, LLE_STATIC); 609 610 /* 611 * in_ifadown gets rid of all the rest of 612 * the routes. This is not quite the right 613 * thing to do, but at least if we are running 614 * a routing process they will come back. 615 */ 616 in_ifadown(&ia->ia_ifa, 1); 617 618 if (ia->ia_ifa.ifa_carp) 619 (*carp_detach_p)(&ia->ia_ifa); 620 621 /* 622 * If this is the last IPv4 address configured on this 623 * interface, leave the all-hosts group. 624 * No state-change report need be transmitted. 625 */ 626 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 627 struct in_ifinfo *ii; 628 629 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 630 IN_MULTI_LOCK(); 631 if (ii->ii_allhosts) { 632 (void)in_leavegroup_locked(ii->ii_allhosts, NULL); 633 ii->ii_allhosts = NULL; 634 } 635 IN_MULTI_UNLOCK(); 636 } 637 638 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 639 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 640 641 return (0); 642 } 643 644 #define rtinitflags(x) \ 645 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 646 ? RTF_HOST : 0) 647 648 /* 649 * Check if we have a route for the given prefix already or add one accordingly. 650 */ 651 int 652 in_addprefix(struct in_ifaddr *target, int flags) 653 { 654 struct rm_priotracker in_ifa_tracker; 655 struct in_ifaddr *ia; 656 struct in_addr prefix, mask, p, m; 657 int error; 658 659 if ((flags & RTF_HOST) != 0) { 660 prefix = target->ia_dstaddr.sin_addr; 661 mask.s_addr = 0; 662 } else { 663 prefix = target->ia_addr.sin_addr; 664 mask = target->ia_sockmask.sin_addr; 665 prefix.s_addr &= mask.s_addr; 666 } 667 668 IN_IFADDR_RLOCK(&in_ifa_tracker); 669 /* Look for an existing address with the same prefix, mask, and fib */ 670 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 671 if (rtinitflags(ia)) { 672 p = ia->ia_dstaddr.sin_addr; 673 674 if (prefix.s_addr != p.s_addr) 675 continue; 676 } else { 677 p = ia->ia_addr.sin_addr; 678 m = ia->ia_sockmask.sin_addr; 679 p.s_addr &= m.s_addr; 680 681 if (prefix.s_addr != p.s_addr || 682 mask.s_addr != m.s_addr) 683 continue; 684 } 685 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 686 continue; 687 688 /* 689 * If we got a matching prefix route inserted by other 690 * interface address, we are done here. 691 */ 692 if (ia->ia_flags & IFA_ROUTE) { 693 #ifdef RADIX_MPATH 694 if (ia->ia_addr.sin_addr.s_addr == 695 target->ia_addr.sin_addr.s_addr) { 696 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 697 return (EEXIST); 698 } else 699 break; 700 #endif 701 if (V_nosameprefix) { 702 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 703 return (EEXIST); 704 } else { 705 int fibnum; 706 707 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : 708 target->ia_ifp->if_fib; 709 rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum); 710 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 711 return (0); 712 } 713 } 714 } 715 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 716 717 /* 718 * No-one seem to have this prefix route, so we try to insert it. 719 */ 720 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 721 if (!error) 722 target->ia_flags |= IFA_ROUTE; 723 return (error); 724 } 725 726 /* 727 * Removes either all lle entries for given @ia, or lle 728 * corresponding to @ia address. 729 */ 730 static void 731 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags) 732 { 733 struct sockaddr_in addr, mask; 734 struct sockaddr *saddr, *smask; 735 struct ifnet *ifp; 736 737 saddr = (struct sockaddr *)&addr; 738 bzero(&addr, sizeof(addr)); 739 addr.sin_len = sizeof(addr); 740 addr.sin_family = AF_INET; 741 smask = (struct sockaddr *)&mask; 742 bzero(&mask, sizeof(mask)); 743 mask.sin_len = sizeof(mask); 744 mask.sin_family = AF_INET; 745 mask.sin_addr.s_addr = ia->ia_subnetmask; 746 ifp = ia->ia_ifp; 747 748 if (all) { 749 750 /* 751 * Remove all L2 entries matching given prefix. 752 * Convert address to host representation to avoid 753 * doing this on every callback. ia_subnetmask is already 754 * stored in host representation. 755 */ 756 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr); 757 lltable_prefix_free(AF_INET, saddr, smask, flags); 758 } else { 759 /* Remove interface address only */ 760 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr; 761 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr); 762 } 763 } 764 765 /* 766 * If there is no other address in the system that can serve a route to the 767 * same prefix, remove the route. Hand over the route to the new address 768 * otherwise. 769 */ 770 int 771 in_scrubprefix(struct in_ifaddr *target, u_int flags) 772 { 773 struct rm_priotracker in_ifa_tracker; 774 struct in_ifaddr *ia; 775 struct in_addr prefix, mask, p, m; 776 int error = 0; 777 778 /* 779 * Remove the loopback route to the interface address. 780 */ 781 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 782 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 783 (flags & LLE_STATIC)) { 784 struct in_ifaddr *eia; 785 786 /* 787 * XXXME: add fib-aware in_localip. 788 * We definitely don't want to switch between 789 * prefixes in different fibs. 790 */ 791 eia = in_localip_more(target); 792 793 if (eia != NULL) { 794 error = ifa_switch_loopback_route((struct ifaddr *)eia, 795 (struct sockaddr *)&target->ia_addr); 796 ifa_free(&eia->ia_ifa); 797 } else { 798 error = ifa_del_loopback_route((struct ifaddr *)target, 799 (struct sockaddr *)&target->ia_addr); 800 } 801 } 802 803 if (rtinitflags(target)) { 804 prefix = target->ia_dstaddr.sin_addr; 805 mask.s_addr = 0; 806 } else { 807 prefix = target->ia_addr.sin_addr; 808 mask = target->ia_sockmask.sin_addr; 809 prefix.s_addr &= mask.s_addr; 810 } 811 812 if ((target->ia_flags & IFA_ROUTE) == 0) { 813 int fibnum; 814 815 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : 816 target->ia_ifp->if_fib; 817 rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum); 818 819 /* 820 * Removing address from !IFF_UP interface or 821 * prefix which exists on other interface (along with route). 822 * No entries should exist here except target addr. 823 * Given that, delete this entry only. 824 */ 825 in_scrubprefixlle(target, 0, flags); 826 return (0); 827 } 828 829 IN_IFADDR_RLOCK(&in_ifa_tracker); 830 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 831 if (rtinitflags(ia)) { 832 p = ia->ia_dstaddr.sin_addr; 833 834 if (prefix.s_addr != p.s_addr) 835 continue; 836 } else { 837 p = ia->ia_addr.sin_addr; 838 m = ia->ia_sockmask.sin_addr; 839 p.s_addr &= m.s_addr; 840 841 if (prefix.s_addr != p.s_addr || 842 mask.s_addr != m.s_addr) 843 continue; 844 } 845 846 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 847 continue; 848 849 /* 850 * If we got a matching prefix address, move IFA_ROUTE and 851 * the route itself to it. Make sure that routing daemons 852 * get a heads-up. 853 */ 854 if ((ia->ia_flags & IFA_ROUTE) == 0) { 855 ifa_ref(&ia->ia_ifa); 856 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 857 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 858 rtinitflags(target)); 859 if (error == 0) 860 target->ia_flags &= ~IFA_ROUTE; 861 else 862 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 863 error); 864 /* Scrub all entries IFF interface is different */ 865 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp, 866 flags); 867 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 868 rtinitflags(ia) | RTF_UP); 869 if (error == 0) 870 ia->ia_flags |= IFA_ROUTE; 871 else 872 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 873 error); 874 ifa_free(&ia->ia_ifa); 875 return (error); 876 } 877 } 878 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 879 880 /* 881 * remove all L2 entries on the given prefix 882 */ 883 in_scrubprefixlle(target, 1, flags); 884 885 /* 886 * As no-one seem to have this prefix, we can remove the route. 887 */ 888 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 889 if (error == 0) 890 target->ia_flags &= ~IFA_ROUTE; 891 else 892 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 893 return (error); 894 } 895 896 #undef rtinitflags 897 898 void 899 in_ifscrub_all(void) 900 { 901 struct ifnet *ifp; 902 struct ifaddr *ifa, *nifa; 903 struct ifaliasreq ifr; 904 905 IFNET_RLOCK(); 906 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 907 /* Cannot lock here - lock recursion. */ 908 /* IF_ADDR_RLOCK(ifp); */ 909 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 910 if (ifa->ifa_addr->sa_family != AF_INET) 911 continue; 912 913 /* 914 * This is ugly but the only way for legacy IP to 915 * cleanly remove addresses and everything attached. 916 */ 917 bzero(&ifr, sizeof(ifr)); 918 ifr.ifra_addr = *ifa->ifa_addr; 919 if (ifa->ifa_dstaddr) 920 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 921 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, 922 ifp, NULL); 923 } 924 /* IF_ADDR_RUNLOCK(ifp); */ 925 in_purgemaddrs(ifp); 926 igmp_domifdetach(ifp); 927 } 928 IFNET_RUNLOCK(); 929 } 930 931 int 932 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia) 933 { 934 935 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 936 /* 937 * Check for old-style (host 0) broadcast, but 938 * taking into account that RFC 3021 obsoletes it. 939 */ 940 (ia->ia_subnetmask != IN_RFC3021_MASK && 941 ntohl(in.s_addr) == ia->ia_subnet)) && 942 /* 943 * Check for an all one subnetmask. These 944 * only exist when an interface gets a secondary 945 * address. 946 */ 947 ia->ia_subnetmask != (u_long)0xffffffff); 948 } 949 950 /* 951 * Return 1 if the address might be a local broadcast address. 952 */ 953 int 954 in_broadcast(struct in_addr in, struct ifnet *ifp) 955 { 956 register struct ifaddr *ifa; 957 int found; 958 959 if (in.s_addr == INADDR_BROADCAST || 960 in.s_addr == INADDR_ANY) 961 return (1); 962 if ((ifp->if_flags & IFF_BROADCAST) == 0) 963 return (0); 964 found = 0; 965 /* 966 * Look through the list of addresses for a match 967 * with a broadcast address. 968 */ 969 IF_ADDR_RLOCK(ifp); 970 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 971 if (ifa->ifa_addr->sa_family == AF_INET && 972 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) { 973 found = 1; 974 break; 975 } 976 IF_ADDR_RUNLOCK(ifp); 977 return (found); 978 } 979 980 /* 981 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 982 */ 983 void 984 in_ifdetach(struct ifnet *ifp) 985 { 986 987 in_pcbpurgeif0(&V_ripcbinfo, ifp); 988 in_pcbpurgeif0(&V_udbinfo, ifp); 989 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 990 in_purgemaddrs(ifp); 991 } 992 993 /* 994 * Delete all IPv4 multicast address records, and associated link-layer 995 * multicast address records, associated with ifp. 996 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 997 * XXX This should not race with ifma_protospec being set during 998 * a new allocation, if it does, we have bigger problems. 999 */ 1000 static void 1001 in_purgemaddrs(struct ifnet *ifp) 1002 { 1003 LIST_HEAD(,in_multi) purgeinms; 1004 struct in_multi *inm, *tinm; 1005 struct ifmultiaddr *ifma; 1006 1007 LIST_INIT(&purgeinms); 1008 IN_MULTI_LOCK(); 1009 1010 /* 1011 * Extract list of in_multi associated with the detaching ifp 1012 * which the PF_INET layer is about to release. 1013 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1014 * by code further down. 1015 */ 1016 IF_ADDR_RLOCK(ifp); 1017 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1018 if (ifma->ifma_addr->sa_family != AF_INET || 1019 ifma->ifma_protospec == NULL) 1020 continue; 1021 #if 0 1022 KASSERT(ifma->ifma_protospec != NULL, 1023 ("%s: ifma_protospec is NULL", __func__)); 1024 #endif 1025 inm = (struct in_multi *)ifma->ifma_protospec; 1026 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 1027 } 1028 IF_ADDR_RUNLOCK(ifp); 1029 1030 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 1031 LIST_REMOVE(inm, inm_link); 1032 inm_release_locked(inm); 1033 } 1034 igmp_ifdetach(ifp); 1035 1036 IN_MULTI_UNLOCK(); 1037 } 1038 1039 struct in_llentry { 1040 struct llentry base; 1041 }; 1042 1043 #define IN_LLTBL_DEFAULT_HSIZE 32 1044 #define IN_LLTBL_HASH(k, h) \ 1045 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1046 1047 /* 1048 * Do actual deallocation of @lle. 1049 */ 1050 static void 1051 in_lltable_destroy_lle_unlocked(struct llentry *lle) 1052 { 1053 1054 LLE_LOCK_DESTROY(lle); 1055 LLE_REQ_DESTROY(lle); 1056 free(lle, M_LLTABLE); 1057 } 1058 1059 /* 1060 * Called by LLE_FREE_LOCKED when number of references 1061 * drops to zero. 1062 */ 1063 static void 1064 in_lltable_destroy_lle(struct llentry *lle) 1065 { 1066 1067 LLE_WUNLOCK(lle); 1068 in_lltable_destroy_lle_unlocked(lle); 1069 } 1070 1071 static struct llentry * 1072 in_lltable_new(struct in_addr addr4, u_int flags) 1073 { 1074 struct in_llentry *lle; 1075 1076 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1077 if (lle == NULL) /* NB: caller generates msg */ 1078 return NULL; 1079 1080 /* 1081 * For IPv4 this will trigger "arpresolve" to generate 1082 * an ARP request. 1083 */ 1084 lle->base.la_expire = time_uptime; /* mark expired */ 1085 lle->base.r_l3addr.addr4 = addr4; 1086 lle->base.lle_refcnt = 1; 1087 lle->base.lle_free = in_lltable_destroy_lle; 1088 LLE_LOCK_INIT(&lle->base); 1089 LLE_REQ_INIT(&lle->base); 1090 callout_init(&lle->base.lle_timer, 1); 1091 1092 return (&lle->base); 1093 } 1094 1095 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1096 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 ) 1097 1098 static int 1099 in_lltable_match_prefix(const struct sockaddr *saddr, 1100 const struct sockaddr *smask, u_int flags, struct llentry *lle) 1101 { 1102 struct in_addr addr, mask, lle_addr; 1103 1104 addr = ((const struct sockaddr_in *)saddr)->sin_addr; 1105 mask = ((const struct sockaddr_in *)smask)->sin_addr; 1106 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1107 1108 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0) 1109 return (0); 1110 1111 if (lle->la_flags & LLE_IFADDR) { 1112 1113 /* 1114 * Delete LLE_IFADDR records IFF address & flag matches. 1115 * Note that addr is the interface address within prefix 1116 * being matched. 1117 * Note also we should handle 'ifdown' cases without removing 1118 * ifaddr macs. 1119 */ 1120 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0) 1121 return (1); 1122 return (0); 1123 } 1124 1125 /* flags & LLE_STATIC means deleting both dynamic and static entries */ 1126 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)) 1127 return (1); 1128 1129 return (0); 1130 } 1131 1132 static void 1133 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1134 { 1135 struct ifnet *ifp; 1136 size_t pkts_dropped; 1137 1138 LLE_WLOCK_ASSERT(lle); 1139 KASSERT(llt != NULL, ("lltable is NULL")); 1140 1141 /* Unlink entry from table if not already */ 1142 if ((lle->la_flags & LLE_LINKED) != 0) { 1143 ifp = llt->llt_ifp; 1144 IF_AFDATA_WLOCK_ASSERT(ifp); 1145 lltable_unlink_entry(llt, lle); 1146 } 1147 1148 /* cancel timer */ 1149 if (callout_stop(&lle->lle_timer) > 0) 1150 LLE_REMREF(lle); 1151 1152 /* Drop hold queue */ 1153 pkts_dropped = llentry_free(lle); 1154 ARPSTAT_ADD(dropped, pkts_dropped); 1155 } 1156 1157 static int 1158 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1159 { 1160 struct rt_addrinfo info; 1161 struct sockaddr_in rt_key, rt_mask; 1162 struct sockaddr rt_gateway; 1163 int rt_flags; 1164 1165 KASSERT(l3addr->sa_family == AF_INET, 1166 ("sin_family %d", l3addr->sa_family)); 1167 1168 bzero(&rt_key, sizeof(rt_key)); 1169 rt_key.sin_len = sizeof(rt_key); 1170 bzero(&rt_mask, sizeof(rt_mask)); 1171 rt_mask.sin_len = sizeof(rt_mask); 1172 bzero(&rt_gateway, sizeof(rt_gateway)); 1173 rt_gateway.sa_len = sizeof(rt_gateway); 1174 1175 bzero(&info, sizeof(info)); 1176 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; 1177 info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask; 1178 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway; 1179 1180 if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0) 1181 return (EINVAL); 1182 1183 rt_flags = info.rti_flags; 1184 1185 /* 1186 * If the gateway for an existing host route matches the target L3 1187 * address, which is a special route inserted by some implementation 1188 * such as MANET, and the interface is of the correct type, then 1189 * allow for ARP to proceed. 1190 */ 1191 if (rt_flags & RTF_GATEWAY) { 1192 if (!(rt_flags & RTF_HOST) || !info.rti_ifp || 1193 info.rti_ifp->if_type != IFT_ETHER || 1194 (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1195 memcmp(rt_gateway.sa_data, l3addr->sa_data, 1196 sizeof(in_addr_t)) != 0) { 1197 rib_free_info(&info); 1198 return (EINVAL); 1199 } 1200 } 1201 rib_free_info(&info); 1202 1203 /* 1204 * Make sure that at least the destination address is covered 1205 * by the route. This is for handling the case where 2 or more 1206 * interfaces have the same prefix. An incoming packet arrives 1207 * on one interface and the corresponding outgoing packet leaves 1208 * another interface. 1209 */ 1210 if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) { 1211 const char *sa, *mask, *addr, *lim; 1212 int len; 1213 1214 mask = (const char *)&rt_mask; 1215 /* 1216 * Just being extra cautious to avoid some custom 1217 * code getting into trouble. 1218 */ 1219 if ((info.rti_addrs & RTA_NETMASK) == 0) 1220 return (EINVAL); 1221 1222 sa = (const char *)&rt_key; 1223 addr = (const char *)l3addr; 1224 len = ((const struct sockaddr_in *)l3addr)->sin_len; 1225 lim = addr + len; 1226 1227 for ( ; addr < lim; sa++, mask++, addr++) { 1228 if ((*sa ^ *addr) & *mask) { 1229 #ifdef DIAGNOSTIC 1230 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1231 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 1232 #endif 1233 return (EINVAL); 1234 } 1235 } 1236 } 1237 1238 return (0); 1239 } 1240 1241 static inline uint32_t 1242 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 1243 { 1244 1245 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 1246 } 1247 1248 static uint32_t 1249 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 1250 { 1251 1252 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 1253 } 1254 1255 static void 1256 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 1257 { 1258 struct sockaddr_in *sin; 1259 1260 sin = (struct sockaddr_in *)sa; 1261 bzero(sin, sizeof(*sin)); 1262 sin->sin_family = AF_INET; 1263 sin->sin_len = sizeof(*sin); 1264 sin->sin_addr = lle->r_l3addr.addr4; 1265 } 1266 1267 static inline struct llentry * 1268 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 1269 { 1270 struct llentry *lle; 1271 struct llentries *lleh; 1272 u_int hashidx; 1273 1274 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 1275 lleh = &llt->lle_head[hashidx]; 1276 LIST_FOREACH(lle, lleh, lle_next) { 1277 if (lle->la_flags & LLE_DELETED) 1278 continue; 1279 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 1280 break; 1281 } 1282 1283 return (lle); 1284 } 1285 1286 static void 1287 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle) 1288 { 1289 1290 lle->la_flags |= LLE_DELETED; 1291 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1292 #ifdef DIAGNOSTIC 1293 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1294 #endif 1295 llentry_free(lle); 1296 } 1297 1298 static struct llentry * 1299 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1300 { 1301 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1302 struct ifnet *ifp = llt->llt_ifp; 1303 struct llentry *lle; 1304 char linkhdr[LLE_MAX_LINKHDR]; 1305 size_t linkhdrsize; 1306 int lladdr_off; 1307 1308 KASSERT(l3addr->sa_family == AF_INET, 1309 ("sin_family %d", l3addr->sa_family)); 1310 1311 /* 1312 * A route that covers the given address must have 1313 * been installed 1st because we are doing a resolution, 1314 * verify this. 1315 */ 1316 if (!(flags & LLE_IFADDR) && 1317 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1318 return (NULL); 1319 1320 lle = in_lltable_new(sin->sin_addr, flags); 1321 if (lle == NULL) { 1322 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1323 return (NULL); 1324 } 1325 lle->la_flags = flags; 1326 if (flags & LLE_STATIC) 1327 lle->r_flags |= RLLE_VALID; 1328 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 1329 linkhdrsize = LLE_MAX_LINKHDR; 1330 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp), 1331 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1332 in_lltable_destroy_lle_unlocked(lle); 1333 return (NULL); 1334 } 1335 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 1336 lladdr_off); 1337 lle->la_flags |= LLE_STATIC; 1338 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR); 1339 } 1340 1341 return (lle); 1342 } 1343 1344 /* 1345 * Return NULL if not found or marked for deletion. 1346 * If found return lle read locked. 1347 */ 1348 static struct llentry * 1349 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1350 { 1351 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1352 struct llentry *lle; 1353 1354 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 1355 KASSERT(l3addr->sa_family == AF_INET, 1356 ("sin_family %d", l3addr->sa_family)); 1357 lle = in_lltable_find_dst(llt, sin->sin_addr); 1358 1359 if (lle == NULL) 1360 return (NULL); 1361 1362 KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) != 1363 (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X", 1364 flags)); 1365 1366 if (flags & LLE_UNLOCKED) 1367 return (lle); 1368 1369 if (flags & LLE_EXCLUSIVE) 1370 LLE_WLOCK(lle); 1371 else 1372 LLE_RLOCK(lle); 1373 1374 return (lle); 1375 } 1376 1377 static int 1378 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 1379 struct sysctl_req *wr) 1380 { 1381 struct ifnet *ifp = llt->llt_ifp; 1382 /* XXX stack use */ 1383 struct { 1384 struct rt_msghdr rtm; 1385 struct sockaddr_in sin; 1386 struct sockaddr_dl sdl; 1387 } arpc; 1388 struct sockaddr_dl *sdl; 1389 int error; 1390 1391 bzero(&arpc, sizeof(arpc)); 1392 /* skip deleted entries */ 1393 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1394 return (0); 1395 /* Skip if jailed and not a valid IP of the prison. */ 1396 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin); 1397 if (prison_if(wr->td->td_ucred, 1398 (struct sockaddr *)&arpc.sin) != 0) 1399 return (0); 1400 /* 1401 * produce a msg made of: 1402 * struct rt_msghdr; 1403 * struct sockaddr_in; (IPv4) 1404 * struct sockaddr_dl; 1405 */ 1406 arpc.rtm.rtm_msglen = sizeof(arpc); 1407 arpc.rtm.rtm_version = RTM_VERSION; 1408 arpc.rtm.rtm_type = RTM_GET; 1409 arpc.rtm.rtm_flags = RTF_UP; 1410 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1411 1412 /* publish */ 1413 if (lle->la_flags & LLE_PUB) 1414 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1415 1416 sdl = &arpc.sdl; 1417 sdl->sdl_family = AF_LINK; 1418 sdl->sdl_len = sizeof(*sdl); 1419 sdl->sdl_index = ifp->if_index; 1420 sdl->sdl_type = ifp->if_type; 1421 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1422 sdl->sdl_alen = ifp->if_addrlen; 1423 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1424 } else { 1425 sdl->sdl_alen = 0; 1426 bzero(LLADDR(sdl), ifp->if_addrlen); 1427 } 1428 1429 arpc.rtm.rtm_rmx.rmx_expire = 1430 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1431 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1432 if (lle->la_flags & LLE_STATIC) 1433 arpc.rtm.rtm_flags |= RTF_STATIC; 1434 if (lle->la_flags & LLE_IFADDR) 1435 arpc.rtm.rtm_flags |= RTF_PINNED; 1436 arpc.rtm.rtm_index = ifp->if_index; 1437 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1438 1439 return (error); 1440 } 1441 1442 static struct lltable * 1443 in_lltattach(struct ifnet *ifp) 1444 { 1445 struct lltable *llt; 1446 1447 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 1448 llt->llt_af = AF_INET; 1449 llt->llt_ifp = ifp; 1450 1451 llt->llt_lookup = in_lltable_lookup; 1452 llt->llt_alloc_entry = in_lltable_alloc; 1453 llt->llt_delete_entry = in_lltable_delete_entry; 1454 llt->llt_dump_entry = in_lltable_dump_entry; 1455 llt->llt_hash = in_lltable_hash; 1456 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 1457 llt->llt_free_entry = in_lltable_free_entry; 1458 llt->llt_match_prefix = in_lltable_match_prefix; 1459 lltable_link(llt); 1460 1461 return (llt); 1462 } 1463 1464 void * 1465 in_domifattach(struct ifnet *ifp) 1466 { 1467 struct in_ifinfo *ii; 1468 1469 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1470 1471 ii->ii_llt = in_lltattach(ifp); 1472 ii->ii_igmp = igmp_domifattach(ifp); 1473 1474 return (ii); 1475 } 1476 1477 void 1478 in_domifdetach(struct ifnet *ifp, void *aux) 1479 { 1480 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1481 1482 igmp_domifdetach(ifp); 1483 lltable_free(ii->ii_llt); 1484 free(ii, M_IFADDR); 1485 } 1486