1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * Copyright (C) 2001 WIDE Project. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)in.c 8.4 (Berkeley) 1/9/95 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 #include "opt_mpath.h" 39 40 #include <sys/param.h> 41 #include <sys/eventhandler.h> 42 #include <sys/systm.h> 43 #include <sys/sockio.h> 44 #include <sys/malloc.h> 45 #include <sys/priv.h> 46 #include <sys/socket.h> 47 #include <sys/jail.h> 48 #include <sys/kernel.h> 49 #include <sys/lock.h> 50 #include <sys/proc.h> 51 #include <sys/rmlock.h> 52 #include <sys/sysctl.h> 53 #include <sys/syslog.h> 54 #include <sys/sx.h> 55 56 #include <net/if.h> 57 #include <net/if_var.h> 58 #include <net/if_arp.h> 59 #include <net/if_dl.h> 60 #include <net/if_llatbl.h> 61 #include <net/if_types.h> 62 #include <net/route.h> 63 #include <net/vnet.h> 64 65 #include <netinet/if_ether.h> 66 #include <netinet/in.h> 67 #include <netinet/in_var.h> 68 #include <netinet/in_pcb.h> 69 #include <netinet/ip_var.h> 70 #include <netinet/ip_carp.h> 71 #include <netinet/igmp_var.h> 72 #include <netinet/udp.h> 73 #include <netinet/udp_var.h> 74 75 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *); 76 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *); 77 78 static void in_socktrim(struct sockaddr_in *); 79 static void in_purgemaddrs(struct ifnet *); 80 81 VNET_DEFINE_STATIC(int, nosameprefix); 82 #define V_nosameprefix VNET(nosameprefix) 83 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW, 84 &VNET_NAME(nosameprefix), 0, 85 "Refuse to create same prefixes on different interfaces"); 86 87 VNET_DECLARE(struct inpcbinfo, ripcbinfo); 88 #define V_ripcbinfo VNET(ripcbinfo) 89 90 static struct sx in_control_sx; 91 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control"); 92 93 /* 94 * Return 1 if an internet address is for a ``local'' host 95 * (one to which we have a connection). 96 */ 97 int 98 in_localaddr(struct in_addr in) 99 { 100 struct rm_priotracker in_ifa_tracker; 101 u_long i = ntohl(in.s_addr); 102 struct in_ifaddr *ia; 103 104 IN_IFADDR_RLOCK(&in_ifa_tracker); 105 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 106 if ((i & ia->ia_subnetmask) == ia->ia_subnet) { 107 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 108 return (1); 109 } 110 } 111 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 112 return (0); 113 } 114 115 /* 116 * Return 1 if an internet address is for the local host and configured 117 * on one of its interfaces. 118 */ 119 int 120 in_localip(struct in_addr in) 121 { 122 struct rm_priotracker in_ifa_tracker; 123 struct in_ifaddr *ia; 124 125 IN_IFADDR_RLOCK(&in_ifa_tracker); 126 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) { 127 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) { 128 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 129 return (1); 130 } 131 } 132 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 133 return (0); 134 } 135 136 /* 137 * Return 1 if an internet address is configured on an interface. 138 */ 139 int 140 in_ifhasaddr(struct ifnet *ifp, struct in_addr in) 141 { 142 struct ifaddr *ifa; 143 struct in_ifaddr *ia; 144 145 IF_ADDR_RLOCK(ifp); 146 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 147 if (ifa->ifa_addr->sa_family != AF_INET) 148 continue; 149 ia = (struct in_ifaddr *)ifa; 150 if (ia->ia_addr.sin_addr.s_addr == in.s_addr) { 151 IF_ADDR_RUNLOCK(ifp); 152 return (1); 153 } 154 } 155 IF_ADDR_RUNLOCK(ifp); 156 157 return (0); 158 } 159 160 /* 161 * Return a reference to the interface address which is different to 162 * the supplied one but with same IP address value. 163 */ 164 static struct in_ifaddr * 165 in_localip_more(struct in_ifaddr *ia) 166 { 167 struct rm_priotracker in_ifa_tracker; 168 in_addr_t in = IA_SIN(ia)->sin_addr.s_addr; 169 struct in_ifaddr *it; 170 171 IN_IFADDR_RLOCK(&in_ifa_tracker); 172 LIST_FOREACH(it, INADDR_HASH(in), ia_hash) { 173 if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) { 174 ifa_ref(&it->ia_ifa); 175 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 176 return (it); 177 } 178 } 179 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 180 181 return (NULL); 182 } 183 184 /* 185 * Determine whether an IP address is in a reserved set of addresses 186 * that may not be forwarded, or whether datagrams to that destination 187 * may be forwarded. 188 */ 189 int 190 in_canforward(struct in_addr in) 191 { 192 u_long i = ntohl(in.s_addr); 193 u_long net; 194 195 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i)) 196 return (0); 197 if (IN_CLASSA(i)) { 198 net = i & IN_CLASSA_NET; 199 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 200 return (0); 201 } 202 return (1); 203 } 204 205 /* 206 * Trim a mask in a sockaddr 207 */ 208 static void 209 in_socktrim(struct sockaddr_in *ap) 210 { 211 char *cplim = (char *) &ap->sin_addr; 212 char *cp = (char *) (&ap->sin_addr + 1); 213 214 ap->sin_len = 0; 215 while (--cp >= cplim) 216 if (*cp) { 217 (ap)->sin_len = cp - (char *) (ap) + 1; 218 break; 219 } 220 } 221 222 /* 223 * Generic internet control operations (ioctl's). 224 */ 225 int 226 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, 227 struct thread *td) 228 { 229 struct ifreq *ifr = (struct ifreq *)data; 230 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr; 231 struct ifaddr *ifa; 232 struct in_ifaddr *ia; 233 int error; 234 235 if (ifp == NULL) 236 return (EADDRNOTAVAIL); 237 238 /* 239 * Filter out 4 ioctls we implement directly. Forward the rest 240 * to specific functions and ifp->if_ioctl(). 241 */ 242 switch (cmd) { 243 case SIOCGIFADDR: 244 case SIOCGIFBRDADDR: 245 case SIOCGIFDSTADDR: 246 case SIOCGIFNETMASK: 247 break; 248 case SIOCDIFADDR: 249 sx_xlock(&in_control_sx); 250 error = in_difaddr_ioctl(cmd, data, ifp, td); 251 sx_xunlock(&in_control_sx); 252 return (error); 253 case OSIOCAIFADDR: /* 9.x compat */ 254 case SIOCAIFADDR: 255 sx_xlock(&in_control_sx); 256 error = in_aifaddr_ioctl(cmd, data, ifp, td); 257 sx_xunlock(&in_control_sx); 258 return (error); 259 case SIOCSIFADDR: 260 case SIOCSIFBRDADDR: 261 case SIOCSIFDSTADDR: 262 case SIOCSIFNETMASK: 263 /* We no longer support that old commands. */ 264 return (EINVAL); 265 default: 266 if (ifp->if_ioctl == NULL) 267 return (EOPNOTSUPP); 268 return ((*ifp->if_ioctl)(ifp, cmd, data)); 269 } 270 271 if (addr->sin_addr.s_addr != INADDR_ANY && 272 prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0) 273 return (EADDRNOTAVAIL); 274 275 /* 276 * Find address for this interface, if it exists. If an 277 * address was specified, find that one instead of the 278 * first one on the interface, if possible. 279 */ 280 IF_ADDR_RLOCK(ifp); 281 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 282 if (ifa->ifa_addr->sa_family != AF_INET) 283 continue; 284 ia = (struct in_ifaddr *)ifa; 285 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr) 286 break; 287 } 288 if (ifa == NULL) 289 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 290 if (ifa->ifa_addr->sa_family == AF_INET) { 291 ia = (struct in_ifaddr *)ifa; 292 if (prison_check_ip4(td->td_ucred, 293 &ia->ia_addr.sin_addr) == 0) 294 break; 295 } 296 297 if (ifa == NULL) { 298 IF_ADDR_RUNLOCK(ifp); 299 return (EADDRNOTAVAIL); 300 } 301 302 error = 0; 303 switch (cmd) { 304 case SIOCGIFADDR: 305 *addr = ia->ia_addr; 306 break; 307 308 case SIOCGIFBRDADDR: 309 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 310 error = EINVAL; 311 break; 312 } 313 *addr = ia->ia_broadaddr; 314 break; 315 316 case SIOCGIFDSTADDR: 317 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 318 error = EINVAL; 319 break; 320 } 321 *addr = ia->ia_dstaddr; 322 break; 323 324 case SIOCGIFNETMASK: 325 *addr = ia->ia_sockmask; 326 break; 327 } 328 329 IF_ADDR_RUNLOCK(ifp); 330 331 return (error); 332 } 333 334 static int 335 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td) 336 { 337 const struct in_aliasreq *ifra = (struct in_aliasreq *)data; 338 const struct sockaddr_in *addr = &ifra->ifra_addr; 339 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr; 340 const struct sockaddr_in *mask = &ifra->ifra_mask; 341 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr; 342 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0; 343 struct ifaddr *ifa; 344 struct in_ifaddr *ia; 345 bool iaIsFirst; 346 int error = 0; 347 348 error = priv_check(td, PRIV_NET_ADDIFADDR); 349 if (error) 350 return (error); 351 352 /* 353 * ifra_addr must be present and be of INET family. 354 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional. 355 */ 356 if (addr->sin_len != sizeof(struct sockaddr_in) || 357 addr->sin_family != AF_INET) 358 return (EINVAL); 359 if (broadaddr->sin_len != 0 && 360 (broadaddr->sin_len != sizeof(struct sockaddr_in) || 361 broadaddr->sin_family != AF_INET)) 362 return (EINVAL); 363 if (mask->sin_len != 0 && 364 (mask->sin_len != sizeof(struct sockaddr_in) || 365 mask->sin_family != AF_INET)) 366 return (EINVAL); 367 if ((ifp->if_flags & IFF_POINTOPOINT) && 368 (dstaddr->sin_len != sizeof(struct sockaddr_in) || 369 dstaddr->sin_addr.s_addr == INADDR_ANY)) 370 return (EDESTADDRREQ); 371 if (vhid > 0 && carp_attach_p == NULL) 372 return (EPROTONOSUPPORT); 373 374 /* 375 * See whether address already exist. 376 */ 377 iaIsFirst = true; 378 ia = NULL; 379 IF_ADDR_RLOCK(ifp); 380 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 381 struct in_ifaddr *it; 382 383 if (ifa->ifa_addr->sa_family != AF_INET) 384 continue; 385 386 it = (struct in_ifaddr *)ifa; 387 iaIsFirst = false; 388 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 389 prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0) 390 ia = it; 391 } 392 IF_ADDR_RUNLOCK(ifp); 393 394 if (ia != NULL) 395 (void )in_difaddr_ioctl(cmd, data, ifp, td); 396 397 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK); 398 ia = (struct in_ifaddr *)ifa; 399 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 400 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 401 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 402 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock, 403 CALLOUT_RETURNUNLOCKED); 404 405 ia->ia_ifp = ifp; 406 ia->ia_addr = *addr; 407 if (mask->sin_len != 0) { 408 ia->ia_sockmask = *mask; 409 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 410 } else { 411 in_addr_t i = ntohl(addr->sin_addr.s_addr); 412 413 /* 414 * Be compatible with network classes, if netmask isn't 415 * supplied, guess it based on classes. 416 */ 417 if (IN_CLASSA(i)) 418 ia->ia_subnetmask = IN_CLASSA_NET; 419 else if (IN_CLASSB(i)) 420 ia->ia_subnetmask = IN_CLASSB_NET; 421 else 422 ia->ia_subnetmask = IN_CLASSC_NET; 423 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 424 } 425 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask; 426 in_socktrim(&ia->ia_sockmask); 427 428 if (ifp->if_flags & IFF_BROADCAST) { 429 if (broadaddr->sin_len != 0) { 430 ia->ia_broadaddr = *broadaddr; 431 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) { 432 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 433 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 434 ia->ia_broadaddr.sin_family = AF_INET; 435 } else { 436 ia->ia_broadaddr.sin_addr.s_addr = 437 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 438 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 439 ia->ia_broadaddr.sin_family = AF_INET; 440 } 441 } 442 443 if (ifp->if_flags & IFF_POINTOPOINT) 444 ia->ia_dstaddr = *dstaddr; 445 446 /* XXXGL: rtinit() needs this strange assignment. */ 447 if (ifp->if_flags & IFF_LOOPBACK) 448 ia->ia_dstaddr = ia->ia_addr; 449 450 if (vhid != 0) { 451 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 452 if (error) 453 return (error); 454 } 455 456 /* if_addrhead is already referenced by ifa_alloc() */ 457 IF_ADDR_WLOCK(ifp); 458 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 459 IF_ADDR_WUNLOCK(ifp); 460 461 ifa_ref(ifa); /* in_ifaddrhead */ 462 IN_IFADDR_WLOCK(); 463 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 464 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash); 465 IN_IFADDR_WUNLOCK(); 466 467 /* 468 * Give the interface a chance to initialize 469 * if this is its first address, 470 * and to validate the address if necessary. 471 */ 472 if (ifp->if_ioctl != NULL) { 473 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 474 if (error) 475 goto fail1; 476 } 477 478 /* 479 * Add route for the network. 480 */ 481 if (vhid == 0) { 482 int flags = RTF_UP; 483 484 if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT)) 485 flags |= RTF_HOST; 486 487 error = in_addprefix(ia, flags); 488 if (error) 489 goto fail1; 490 } 491 492 /* 493 * Add a loopback route to self. 494 */ 495 if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 && 496 ia->ia_addr.sin_addr.s_addr != INADDR_ANY && 497 !((ifp->if_flags & IFF_POINTOPOINT) && 498 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) { 499 struct in_ifaddr *eia; 500 501 eia = in_localip_more(ia); 502 503 if (eia == NULL) { 504 error = ifa_add_loopback_route((struct ifaddr *)ia, 505 (struct sockaddr *)&ia->ia_addr); 506 if (error) 507 goto fail2; 508 } else 509 ifa_free(&eia->ia_ifa); 510 } 511 512 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 513 struct in_addr allhosts_addr; 514 struct in_ifinfo *ii; 515 516 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 517 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 518 519 error = in_joingroup(ifp, &allhosts_addr, NULL, 520 &ii->ii_allhosts); 521 } 522 523 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 524 525 return (error); 526 527 fail2: 528 if (vhid == 0) 529 (void )in_scrubprefix(ia, LLE_STATIC); 530 531 fail1: 532 if (ia->ia_ifa.ifa_carp) 533 (*carp_detach_p)(&ia->ia_ifa, false); 534 535 IF_ADDR_WLOCK(ifp); 536 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 537 IF_ADDR_WUNLOCK(ifp); 538 ifa_free(&ia->ia_ifa); /* if_addrhead */ 539 540 IN_IFADDR_WLOCK(); 541 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 542 LIST_REMOVE(ia, ia_hash); 543 IN_IFADDR_WUNLOCK(); 544 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 545 546 return (error); 547 } 548 549 static int 550 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td) 551 { 552 const struct ifreq *ifr = (struct ifreq *)data; 553 const struct sockaddr_in *addr = (const struct sockaddr_in *) 554 &ifr->ifr_addr; 555 struct ifaddr *ifa; 556 struct in_ifaddr *ia; 557 bool deleteAny, iaIsLast; 558 int error; 559 560 if (td != NULL) { 561 error = priv_check(td, PRIV_NET_DELIFADDR); 562 if (error) 563 return (error); 564 } 565 566 if (addr->sin_len != sizeof(struct sockaddr_in) || 567 addr->sin_family != AF_INET) 568 deleteAny = true; 569 else 570 deleteAny = false; 571 572 iaIsLast = true; 573 ia = NULL; 574 IF_ADDR_WLOCK(ifp); 575 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 576 struct in_ifaddr *it; 577 578 if (ifa->ifa_addr->sa_family != AF_INET) 579 continue; 580 581 it = (struct in_ifaddr *)ifa; 582 if (deleteAny && ia == NULL && (td == NULL || 583 prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0)) 584 ia = it; 585 586 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 587 (td == NULL || prison_check_ip4(td->td_ucred, 588 &addr->sin_addr) == 0)) 589 ia = it; 590 591 if (it != ia) 592 iaIsLast = false; 593 } 594 595 if (ia == NULL) { 596 IF_ADDR_WUNLOCK(ifp); 597 return (EADDRNOTAVAIL); 598 } 599 600 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 601 IF_ADDR_WUNLOCK(ifp); 602 ifa_free(&ia->ia_ifa); /* if_addrhead */ 603 604 IN_IFADDR_WLOCK(); 605 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 606 LIST_REMOVE(ia, ia_hash); 607 IN_IFADDR_WUNLOCK(); 608 609 /* 610 * in_scrubprefix() kills the interface route. 611 */ 612 in_scrubprefix(ia, LLE_STATIC); 613 614 /* 615 * in_ifadown gets rid of all the rest of 616 * the routes. This is not quite the right 617 * thing to do, but at least if we are running 618 * a routing process they will come back. 619 */ 620 in_ifadown(&ia->ia_ifa, 1); 621 622 if (ia->ia_ifa.ifa_carp) 623 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR); 624 625 /* 626 * If this is the last IPv4 address configured on this 627 * interface, leave the all-hosts group. 628 * No state-change report need be transmitted. 629 */ 630 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 631 struct in_ifinfo *ii; 632 633 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 634 if (ii->ii_allhosts) { 635 (void)in_leavegroup(ii->ii_allhosts, NULL); 636 ii->ii_allhosts = NULL; 637 } 638 } 639 640 IF_ADDR_WLOCK(ifp); 641 if (callout_stop(&ia->ia_garp_timer) == 1) { 642 ifa_free(&ia->ia_ifa); 643 } 644 IF_ADDR_WUNLOCK(ifp); 645 646 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 647 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 648 649 return (0); 650 } 651 652 #define rtinitflags(x) \ 653 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 654 ? RTF_HOST : 0) 655 656 /* 657 * Check if we have a route for the given prefix already or add one accordingly. 658 */ 659 int 660 in_addprefix(struct in_ifaddr *target, int flags) 661 { 662 struct rm_priotracker in_ifa_tracker; 663 struct in_ifaddr *ia; 664 struct in_addr prefix, mask, p, m; 665 int error; 666 667 if ((flags & RTF_HOST) != 0) { 668 prefix = target->ia_dstaddr.sin_addr; 669 mask.s_addr = 0; 670 } else { 671 prefix = target->ia_addr.sin_addr; 672 mask = target->ia_sockmask.sin_addr; 673 prefix.s_addr &= mask.s_addr; 674 } 675 676 IN_IFADDR_RLOCK(&in_ifa_tracker); 677 /* Look for an existing address with the same prefix, mask, and fib */ 678 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 679 if (rtinitflags(ia)) { 680 p = ia->ia_dstaddr.sin_addr; 681 682 if (prefix.s_addr != p.s_addr) 683 continue; 684 } else { 685 p = ia->ia_addr.sin_addr; 686 m = ia->ia_sockmask.sin_addr; 687 p.s_addr &= m.s_addr; 688 689 if (prefix.s_addr != p.s_addr || 690 mask.s_addr != m.s_addr) 691 continue; 692 } 693 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 694 continue; 695 696 /* 697 * If we got a matching prefix route inserted by other 698 * interface address, we are done here. 699 */ 700 if (ia->ia_flags & IFA_ROUTE) { 701 #ifdef RADIX_MPATH 702 if (ia->ia_addr.sin_addr.s_addr == 703 target->ia_addr.sin_addr.s_addr) { 704 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 705 return (EEXIST); 706 } else 707 break; 708 #endif 709 if (V_nosameprefix) { 710 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 711 return (EEXIST); 712 } else { 713 int fibnum; 714 715 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : 716 target->ia_ifp->if_fib; 717 rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum); 718 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 719 return (0); 720 } 721 } 722 } 723 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 724 725 /* 726 * No-one seem to have this prefix route, so we try to insert it. 727 */ 728 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 729 if (!error) 730 target->ia_flags |= IFA_ROUTE; 731 return (error); 732 } 733 734 /* 735 * Removes either all lle entries for given @ia, or lle 736 * corresponding to @ia address. 737 */ 738 static void 739 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags) 740 { 741 struct sockaddr_in addr, mask; 742 struct sockaddr *saddr, *smask; 743 struct ifnet *ifp; 744 745 saddr = (struct sockaddr *)&addr; 746 bzero(&addr, sizeof(addr)); 747 addr.sin_len = sizeof(addr); 748 addr.sin_family = AF_INET; 749 smask = (struct sockaddr *)&mask; 750 bzero(&mask, sizeof(mask)); 751 mask.sin_len = sizeof(mask); 752 mask.sin_family = AF_INET; 753 mask.sin_addr.s_addr = ia->ia_subnetmask; 754 ifp = ia->ia_ifp; 755 756 if (all) { 757 758 /* 759 * Remove all L2 entries matching given prefix. 760 * Convert address to host representation to avoid 761 * doing this on every callback. ia_subnetmask is already 762 * stored in host representation. 763 */ 764 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr); 765 lltable_prefix_free(AF_INET, saddr, smask, flags); 766 } else { 767 /* Remove interface address only */ 768 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr; 769 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr); 770 } 771 } 772 773 /* 774 * If there is no other address in the system that can serve a route to the 775 * same prefix, remove the route. Hand over the route to the new address 776 * otherwise. 777 */ 778 int 779 in_scrubprefix(struct in_ifaddr *target, u_int flags) 780 { 781 struct rm_priotracker in_ifa_tracker; 782 struct in_ifaddr *ia; 783 struct in_addr prefix, mask, p, m; 784 int error = 0; 785 786 /* 787 * Remove the loopback route to the interface address. 788 */ 789 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 790 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 791 (flags & LLE_STATIC)) { 792 struct in_ifaddr *eia; 793 794 /* 795 * XXXME: add fib-aware in_localip. 796 * We definitely don't want to switch between 797 * prefixes in different fibs. 798 */ 799 eia = in_localip_more(target); 800 801 if (eia != NULL) { 802 error = ifa_switch_loopback_route((struct ifaddr *)eia, 803 (struct sockaddr *)&target->ia_addr); 804 ifa_free(&eia->ia_ifa); 805 } else { 806 error = ifa_del_loopback_route((struct ifaddr *)target, 807 (struct sockaddr *)&target->ia_addr); 808 } 809 } 810 811 if (rtinitflags(target)) { 812 prefix = target->ia_dstaddr.sin_addr; 813 mask.s_addr = 0; 814 } else { 815 prefix = target->ia_addr.sin_addr; 816 mask = target->ia_sockmask.sin_addr; 817 prefix.s_addr &= mask.s_addr; 818 } 819 820 if ((target->ia_flags & IFA_ROUTE) == 0) { 821 int fibnum; 822 823 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : 824 target->ia_ifp->if_fib; 825 rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum); 826 827 /* 828 * Removing address from !IFF_UP interface or 829 * prefix which exists on other interface (along with route). 830 * No entries should exist here except target addr. 831 * Given that, delete this entry only. 832 */ 833 in_scrubprefixlle(target, 0, flags); 834 return (0); 835 } 836 837 IN_IFADDR_RLOCK(&in_ifa_tracker); 838 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 839 if (rtinitflags(ia)) { 840 p = ia->ia_dstaddr.sin_addr; 841 842 if (prefix.s_addr != p.s_addr) 843 continue; 844 } else { 845 p = ia->ia_addr.sin_addr; 846 m = ia->ia_sockmask.sin_addr; 847 p.s_addr &= m.s_addr; 848 849 if (prefix.s_addr != p.s_addr || 850 mask.s_addr != m.s_addr) 851 continue; 852 } 853 854 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 855 continue; 856 857 /* 858 * If we got a matching prefix address, move IFA_ROUTE and 859 * the route itself to it. Make sure that routing daemons 860 * get a heads-up. 861 */ 862 if ((ia->ia_flags & IFA_ROUTE) == 0) { 863 ifa_ref(&ia->ia_ifa); 864 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 865 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 866 rtinitflags(target)); 867 if (error == 0) 868 target->ia_flags &= ~IFA_ROUTE; 869 else 870 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 871 error); 872 /* Scrub all entries IFF interface is different */ 873 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp, 874 flags); 875 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 876 rtinitflags(ia) | RTF_UP); 877 if (error == 0) 878 ia->ia_flags |= IFA_ROUTE; 879 else 880 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 881 error); 882 ifa_free(&ia->ia_ifa); 883 return (error); 884 } 885 } 886 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 887 888 /* 889 * remove all L2 entries on the given prefix 890 */ 891 in_scrubprefixlle(target, 1, flags); 892 893 /* 894 * As no-one seem to have this prefix, we can remove the route. 895 */ 896 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 897 if (error == 0) 898 target->ia_flags &= ~IFA_ROUTE; 899 else 900 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 901 return (error); 902 } 903 904 #undef rtinitflags 905 906 void 907 in_ifscrub_all(void) 908 { 909 struct ifnet *ifp; 910 struct ifaddr *ifa, *nifa; 911 struct ifaliasreq ifr; 912 913 IFNET_RLOCK(); 914 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 915 /* Cannot lock here - lock recursion. */ 916 /* IF_ADDR_RLOCK(ifp); */ 917 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 918 if (ifa->ifa_addr->sa_family != AF_INET) 919 continue; 920 921 /* 922 * This is ugly but the only way for legacy IP to 923 * cleanly remove addresses and everything attached. 924 */ 925 bzero(&ifr, sizeof(ifr)); 926 ifr.ifra_addr = *ifa->ifa_addr; 927 if (ifa->ifa_dstaddr) 928 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 929 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, 930 ifp, NULL); 931 } 932 /* IF_ADDR_RUNLOCK(ifp); */ 933 in_purgemaddrs(ifp); 934 igmp_domifdetach(ifp); 935 } 936 IFNET_RUNLOCK(); 937 } 938 939 int 940 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia) 941 { 942 943 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 944 /* 945 * Check for old-style (host 0) broadcast, but 946 * taking into account that RFC 3021 obsoletes it. 947 */ 948 (ia->ia_subnetmask != IN_RFC3021_MASK && 949 ntohl(in.s_addr) == ia->ia_subnet)) && 950 /* 951 * Check for an all one subnetmask. These 952 * only exist when an interface gets a secondary 953 * address. 954 */ 955 ia->ia_subnetmask != (u_long)0xffffffff); 956 } 957 958 /* 959 * Return 1 if the address might be a local broadcast address. 960 */ 961 int 962 in_broadcast(struct in_addr in, struct ifnet *ifp) 963 { 964 struct ifaddr *ifa; 965 int found; 966 967 if (in.s_addr == INADDR_BROADCAST || 968 in.s_addr == INADDR_ANY) 969 return (1); 970 if ((ifp->if_flags & IFF_BROADCAST) == 0) 971 return (0); 972 found = 0; 973 /* 974 * Look through the list of addresses for a match 975 * with a broadcast address. 976 */ 977 IF_ADDR_RLOCK(ifp); 978 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 979 if (ifa->ifa_addr->sa_family == AF_INET && 980 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) { 981 found = 1; 982 break; 983 } 984 IF_ADDR_RUNLOCK(ifp); 985 return (found); 986 } 987 988 /* 989 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 990 */ 991 void 992 in_ifdetach(struct ifnet *ifp) 993 { 994 IN_MULTI_LOCK(); 995 in_pcbpurgeif0(&V_ripcbinfo, ifp); 996 in_pcbpurgeif0(&V_udbinfo, ifp); 997 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 998 in_purgemaddrs(ifp); 999 IN_MULTI_UNLOCK(); 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 struct in_multi_head purgeinms; 1013 struct in_multi *inm; 1014 struct ifmultiaddr *ifma, *next; 1015 1016 SLIST_INIT(&purgeinms); 1017 IN_MULTI_LIST_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_WLOCK(ifp); 1026 restart: 1027 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) { 1028 if (ifma->ifma_addr->sa_family != AF_INET || 1029 ifma->ifma_protospec == NULL) 1030 continue; 1031 inm = (struct in_multi *)ifma->ifma_protospec; 1032 inm_rele_locked(&purgeinms, inm); 1033 if (__predict_false(ifma_restart)) { 1034 ifma_restart = true; 1035 goto restart; 1036 } 1037 } 1038 IF_ADDR_WUNLOCK(ifp); 1039 1040 inm_release_list_deferred(&purgeinms); 1041 igmp_ifdetach(ifp); 1042 IN_MULTI_LIST_UNLOCK(); 1043 } 1044 1045 struct in_llentry { 1046 struct llentry base; 1047 }; 1048 1049 #define IN_LLTBL_DEFAULT_HSIZE 32 1050 #define IN_LLTBL_HASH(k, h) \ 1051 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1052 1053 /* 1054 * Do actual deallocation of @lle. 1055 */ 1056 static void 1057 in_lltable_destroy_lle_unlocked(epoch_context_t ctx) 1058 { 1059 struct llentry *lle; 1060 1061 lle = __containerof(ctx, struct llentry, lle_epoch_ctx); 1062 LLE_LOCK_DESTROY(lle); 1063 LLE_REQ_DESTROY(lle); 1064 free(lle, M_LLTABLE); 1065 } 1066 1067 /* 1068 * Called by the datapath to indicate that 1069 * the entry was used. 1070 */ 1071 static void 1072 in_lltable_mark_used(struct llentry *lle) 1073 { 1074 1075 LLE_REQ_LOCK(lle); 1076 lle->r_skip_req = 0; 1077 LLE_REQ_UNLOCK(lle); 1078 } 1079 1080 /* 1081 * Called by LLE_FREE_LOCKED when number of references 1082 * drops to zero. 1083 */ 1084 static void 1085 in_lltable_destroy_lle(struct llentry *lle) 1086 { 1087 1088 LLE_WUNLOCK(lle); 1089 epoch_call(net_epoch_preempt, &lle->lle_epoch_ctx, in_lltable_destroy_lle_unlocked); 1090 } 1091 1092 static struct llentry * 1093 in_lltable_new(struct in_addr addr4, u_int flags) 1094 { 1095 struct in_llentry *lle; 1096 1097 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1098 if (lle == NULL) /* NB: caller generates msg */ 1099 return NULL; 1100 1101 /* 1102 * For IPv4 this will trigger "arpresolve" to generate 1103 * an ARP request. 1104 */ 1105 lle->base.la_expire = time_uptime; /* mark expired */ 1106 lle->base.r_l3addr.addr4 = addr4; 1107 lle->base.lle_refcnt = 1; 1108 lle->base.lle_free = in_lltable_destroy_lle; 1109 LLE_LOCK_INIT(&lle->base); 1110 LLE_REQ_INIT(&lle->base); 1111 callout_init(&lle->base.lle_timer, 1); 1112 1113 return (&lle->base); 1114 } 1115 1116 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1117 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 ) 1118 1119 static int 1120 in_lltable_match_prefix(const struct sockaddr *saddr, 1121 const struct sockaddr *smask, u_int flags, struct llentry *lle) 1122 { 1123 struct in_addr addr, mask, lle_addr; 1124 1125 addr = ((const struct sockaddr_in *)saddr)->sin_addr; 1126 mask = ((const struct sockaddr_in *)smask)->sin_addr; 1127 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1128 1129 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0) 1130 return (0); 1131 1132 if (lle->la_flags & LLE_IFADDR) { 1133 1134 /* 1135 * Delete LLE_IFADDR records IFF address & flag matches. 1136 * Note that addr is the interface address within prefix 1137 * being matched. 1138 * Note also we should handle 'ifdown' cases without removing 1139 * ifaddr macs. 1140 */ 1141 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0) 1142 return (1); 1143 return (0); 1144 } 1145 1146 /* flags & LLE_STATIC means deleting both dynamic and static entries */ 1147 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)) 1148 return (1); 1149 1150 return (0); 1151 } 1152 1153 static void 1154 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1155 { 1156 size_t pkts_dropped; 1157 1158 LLE_WLOCK_ASSERT(lle); 1159 KASSERT(llt != NULL, ("lltable is NULL")); 1160 1161 /* Unlink entry from table if not already */ 1162 if ((lle->la_flags & LLE_LINKED) != 0) { 1163 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 1164 lltable_unlink_entry(llt, lle); 1165 } 1166 1167 /* Drop hold queue */ 1168 pkts_dropped = llentry_free(lle); 1169 ARPSTAT_ADD(dropped, pkts_dropped); 1170 } 1171 1172 static int 1173 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1174 { 1175 struct rt_addrinfo info; 1176 struct sockaddr_in rt_key, rt_mask; 1177 struct sockaddr rt_gateway; 1178 int rt_flags; 1179 1180 KASSERT(l3addr->sa_family == AF_INET, 1181 ("sin_family %d", l3addr->sa_family)); 1182 1183 bzero(&rt_key, sizeof(rt_key)); 1184 rt_key.sin_len = sizeof(rt_key); 1185 bzero(&rt_mask, sizeof(rt_mask)); 1186 rt_mask.sin_len = sizeof(rt_mask); 1187 bzero(&rt_gateway, sizeof(rt_gateway)); 1188 rt_gateway.sa_len = sizeof(rt_gateway); 1189 1190 bzero(&info, sizeof(info)); 1191 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; 1192 info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask; 1193 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway; 1194 1195 if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0) 1196 return (EINVAL); 1197 1198 rt_flags = info.rti_flags; 1199 1200 /* 1201 * If the gateway for an existing host route matches the target L3 1202 * address, which is a special route inserted by some implementation 1203 * such as MANET, and the interface is of the correct type, then 1204 * allow for ARP to proceed. 1205 */ 1206 if (rt_flags & RTF_GATEWAY) { 1207 if (!(rt_flags & RTF_HOST) || !info.rti_ifp || 1208 info.rti_ifp->if_type != IFT_ETHER || 1209 (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1210 memcmp(rt_gateway.sa_data, l3addr->sa_data, 1211 sizeof(in_addr_t)) != 0) { 1212 rib_free_info(&info); 1213 return (EINVAL); 1214 } 1215 } 1216 rib_free_info(&info); 1217 1218 /* 1219 * Make sure that at least the destination address is covered 1220 * by the route. This is for handling the case where 2 or more 1221 * interfaces have the same prefix. An incoming packet arrives 1222 * on one interface and the corresponding outgoing packet leaves 1223 * another interface. 1224 */ 1225 if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) { 1226 const char *sa, *mask, *addr, *lim; 1227 const struct sockaddr_in *l3sin; 1228 1229 mask = (const char *)&rt_mask; 1230 /* 1231 * Just being extra cautious to avoid some custom 1232 * code getting into trouble. 1233 */ 1234 if ((info.rti_addrs & RTA_NETMASK) == 0) 1235 return (EINVAL); 1236 1237 sa = (const char *)&rt_key; 1238 addr = (const char *)l3addr; 1239 l3sin = (const struct sockaddr_in *)l3addr; 1240 lim = addr + l3sin->sin_len; 1241 1242 for ( ; addr < lim; sa++, mask++, addr++) { 1243 if ((*sa ^ *addr) & *mask) { 1244 #ifdef DIAGNOSTIC 1245 char addrbuf[INET_ADDRSTRLEN]; 1246 1247 log(LOG_INFO, "IPv4 address: \"%s\" " 1248 "is not on the network\n", 1249 inet_ntoa_r(l3sin->sin_addr, addrbuf)); 1250 #endif 1251 return (EINVAL); 1252 } 1253 } 1254 } 1255 1256 return (0); 1257 } 1258 1259 static inline uint32_t 1260 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 1261 { 1262 1263 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 1264 } 1265 1266 static uint32_t 1267 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 1268 { 1269 1270 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 1271 } 1272 1273 static void 1274 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 1275 { 1276 struct sockaddr_in *sin; 1277 1278 sin = (struct sockaddr_in *)sa; 1279 bzero(sin, sizeof(*sin)); 1280 sin->sin_family = AF_INET; 1281 sin->sin_len = sizeof(*sin); 1282 sin->sin_addr = lle->r_l3addr.addr4; 1283 } 1284 1285 static inline struct llentry * 1286 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 1287 { 1288 struct llentry *lle; 1289 struct llentries *lleh; 1290 u_int hashidx; 1291 1292 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 1293 lleh = &llt->lle_head[hashidx]; 1294 CK_LIST_FOREACH(lle, lleh, lle_next) { 1295 if (lle->la_flags & LLE_DELETED) 1296 continue; 1297 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 1298 break; 1299 } 1300 1301 return (lle); 1302 } 1303 1304 static void 1305 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle) 1306 { 1307 1308 lle->la_flags |= LLE_DELETED; 1309 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1310 #ifdef DIAGNOSTIC 1311 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1312 #endif 1313 llentry_free(lle); 1314 } 1315 1316 static struct llentry * 1317 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1318 { 1319 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1320 struct ifnet *ifp = llt->llt_ifp; 1321 struct llentry *lle; 1322 char linkhdr[LLE_MAX_LINKHDR]; 1323 size_t linkhdrsize; 1324 int lladdr_off; 1325 1326 KASSERT(l3addr->sa_family == AF_INET, 1327 ("sin_family %d", l3addr->sa_family)); 1328 1329 /* 1330 * A route that covers the given address must have 1331 * been installed 1st because we are doing a resolution, 1332 * verify this. 1333 */ 1334 if (!(flags & LLE_IFADDR) && 1335 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1336 return (NULL); 1337 1338 lle = in_lltable_new(sin->sin_addr, flags); 1339 if (lle == NULL) { 1340 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1341 return (NULL); 1342 } 1343 lle->la_flags = flags; 1344 if (flags & LLE_STATIC) 1345 lle->r_flags |= RLLE_VALID; 1346 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 1347 linkhdrsize = LLE_MAX_LINKHDR; 1348 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp), 1349 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1350 epoch_call(net_epoch_preempt, &lle->lle_epoch_ctx, in_lltable_destroy_lle_unlocked); 1351 return (NULL); 1352 } 1353 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 1354 lladdr_off); 1355 lle->la_flags |= LLE_STATIC; 1356 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR); 1357 } 1358 1359 return (lle); 1360 } 1361 1362 /* 1363 * Return NULL if not found or marked for deletion. 1364 * If found return lle read locked. 1365 */ 1366 static struct llentry * 1367 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1368 { 1369 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1370 struct llentry *lle; 1371 1372 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 1373 KASSERT(l3addr->sa_family == AF_INET, 1374 ("sin_family %d", l3addr->sa_family)); 1375 lle = in_lltable_find_dst(llt, sin->sin_addr); 1376 1377 if (lle == NULL) 1378 return (NULL); 1379 1380 KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) != 1381 (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X", 1382 flags)); 1383 1384 if (flags & LLE_UNLOCKED) 1385 return (lle); 1386 1387 if (flags & LLE_EXCLUSIVE) 1388 LLE_WLOCK(lle); 1389 else 1390 LLE_RLOCK(lle); 1391 1392 return (lle); 1393 } 1394 1395 static int 1396 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 1397 struct sysctl_req *wr) 1398 { 1399 struct ifnet *ifp = llt->llt_ifp; 1400 /* XXX stack use */ 1401 struct { 1402 struct rt_msghdr rtm; 1403 struct sockaddr_in sin; 1404 struct sockaddr_dl sdl; 1405 } arpc; 1406 struct sockaddr_dl *sdl; 1407 int error; 1408 1409 bzero(&arpc, sizeof(arpc)); 1410 /* skip deleted entries */ 1411 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1412 return (0); 1413 /* Skip if jailed and not a valid IP of the prison. */ 1414 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin); 1415 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0) 1416 return (0); 1417 /* 1418 * produce a msg made of: 1419 * struct rt_msghdr; 1420 * struct sockaddr_in; (IPv4) 1421 * struct sockaddr_dl; 1422 */ 1423 arpc.rtm.rtm_msglen = sizeof(arpc); 1424 arpc.rtm.rtm_version = RTM_VERSION; 1425 arpc.rtm.rtm_type = RTM_GET; 1426 arpc.rtm.rtm_flags = RTF_UP; 1427 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1428 1429 /* publish */ 1430 if (lle->la_flags & LLE_PUB) 1431 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1432 1433 sdl = &arpc.sdl; 1434 sdl->sdl_family = AF_LINK; 1435 sdl->sdl_len = sizeof(*sdl); 1436 sdl->sdl_index = ifp->if_index; 1437 sdl->sdl_type = ifp->if_type; 1438 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1439 sdl->sdl_alen = ifp->if_addrlen; 1440 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1441 } else { 1442 sdl->sdl_alen = 0; 1443 bzero(LLADDR(sdl), ifp->if_addrlen); 1444 } 1445 1446 arpc.rtm.rtm_rmx.rmx_expire = 1447 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1448 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1449 if (lle->la_flags & LLE_STATIC) 1450 arpc.rtm.rtm_flags |= RTF_STATIC; 1451 if (lle->la_flags & LLE_IFADDR) 1452 arpc.rtm.rtm_flags |= RTF_PINNED; 1453 arpc.rtm.rtm_index = ifp->if_index; 1454 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1455 1456 return (error); 1457 } 1458 1459 static struct lltable * 1460 in_lltattach(struct ifnet *ifp) 1461 { 1462 struct lltable *llt; 1463 1464 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 1465 llt->llt_af = AF_INET; 1466 llt->llt_ifp = ifp; 1467 1468 llt->llt_lookup = in_lltable_lookup; 1469 llt->llt_alloc_entry = in_lltable_alloc; 1470 llt->llt_delete_entry = in_lltable_delete_entry; 1471 llt->llt_dump_entry = in_lltable_dump_entry; 1472 llt->llt_hash = in_lltable_hash; 1473 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 1474 llt->llt_free_entry = in_lltable_free_entry; 1475 llt->llt_match_prefix = in_lltable_match_prefix; 1476 llt->llt_mark_used = in_lltable_mark_used; 1477 lltable_link(llt); 1478 1479 return (llt); 1480 } 1481 1482 void * 1483 in_domifattach(struct ifnet *ifp) 1484 { 1485 struct in_ifinfo *ii; 1486 1487 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1488 1489 ii->ii_llt = in_lltattach(ifp); 1490 ii->ii_igmp = igmp_domifattach(ifp); 1491 1492 return (ii); 1493 } 1494 1495 void 1496 in_domifdetach(struct ifnet *ifp, void *aux) 1497 { 1498 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1499 1500 igmp_domifdetach(ifp); 1501 lltable_free(ii->ii_llt); 1502 free(ii, M_IFADDR); 1503 } 1504