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_carp.h" 37 #include "opt_mpath.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/sockio.h> 42 #include <sys/malloc.h> 43 #include <sys/priv.h> 44 #include <sys/socket.h> 45 #include <sys/jail.h> 46 #include <sys/kernel.h> 47 #include <sys/proc.h> 48 #include <sys/sysctl.h> 49 #include <sys/syslog.h> 50 51 #include <net/if.h> 52 #include <net/if_var.h> 53 #include <net/if_dl.h> 54 #include <net/if_llatbl.h> 55 #include <net/if_types.h> 56 #include <net/route.h> 57 #include <net/vnet.h> 58 59 #include <netinet/in.h> 60 #include <netinet/in_var.h> 61 #include <netinet/in_pcb.h> 62 #include <netinet/ip_var.h> 63 #include <netinet/igmp_var.h> 64 #include <netinet/udp.h> 65 #include <netinet/udp_var.h> 66 67 static int in_mask2len(struct in_addr *); 68 static void in_len2mask(struct in_addr *, int); 69 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t, 70 struct ifnet *, struct thread *); 71 72 static int in_addprefix(struct in_ifaddr *, int); 73 static int in_scrubprefix(struct in_ifaddr *); 74 static void in_socktrim(struct sockaddr_in *); 75 static int in_ifinit(struct ifnet *, 76 struct in_ifaddr *, struct sockaddr_in *, int); 77 static void in_purgemaddrs(struct ifnet *); 78 79 static VNET_DEFINE(int, subnetsarelocal); 80 #define V_subnetsarelocal VNET(subnetsarelocal) 81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW, 82 &VNET_NAME(subnetsarelocal), 0, 83 "Treat all subnets as directly connected"); 84 static VNET_DEFINE(int, sameprefixcarponly); 85 #define V_sameprefixcarponly VNET(sameprefixcarponly) 86 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW, 87 &VNET_NAME(sameprefixcarponly), 0, 88 "Refuse to create same prefixes on different interfaces"); 89 90 VNET_DECLARE(struct inpcbinfo, ripcbinfo); 91 #define V_ripcbinfo VNET(ripcbinfo) 92 93 /* 94 * Return 1 if an internet address is for a ``local'' host 95 * (one to which we have a connection). If subnetsarelocal 96 * is true, this includes other subnets of the local net. 97 * Otherwise, it includes only the directly-connected (sub)nets. 98 */ 99 int 100 in_localaddr(struct in_addr in) 101 { 102 register u_long i = ntohl(in.s_addr); 103 register struct in_ifaddr *ia; 104 105 IN_IFADDR_RLOCK(); 106 if (V_subnetsarelocal) { 107 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 108 if ((i & ia->ia_netmask) == ia->ia_net) { 109 IN_IFADDR_RUNLOCK(); 110 return (1); 111 } 112 } 113 } else { 114 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 115 if ((i & ia->ia_subnetmask) == ia->ia_subnet) { 116 IN_IFADDR_RUNLOCK(); 117 return (1); 118 } 119 } 120 } 121 IN_IFADDR_RUNLOCK(); 122 return (0); 123 } 124 125 /* 126 * Return 1 if an internet address is for the local host and configured 127 * on one of its interfaces. 128 */ 129 int 130 in_localip(struct in_addr in) 131 { 132 struct in_ifaddr *ia; 133 134 IN_IFADDR_RLOCK(); 135 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) { 136 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) { 137 IN_IFADDR_RUNLOCK(); 138 return (1); 139 } 140 } 141 IN_IFADDR_RUNLOCK(); 142 return (0); 143 } 144 145 /* 146 * Determine whether an IP address is in a reserved set of addresses 147 * that may not be forwarded, or whether datagrams to that destination 148 * may be forwarded. 149 */ 150 int 151 in_canforward(struct in_addr in) 152 { 153 register u_long i = ntohl(in.s_addr); 154 register u_long net; 155 156 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i)) 157 return (0); 158 if (IN_CLASSA(i)) { 159 net = i & IN_CLASSA_NET; 160 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 161 return (0); 162 } 163 return (1); 164 } 165 166 /* 167 * Trim a mask in a sockaddr 168 */ 169 static void 170 in_socktrim(struct sockaddr_in *ap) 171 { 172 register char *cplim = (char *) &ap->sin_addr; 173 register char *cp = (char *) (&ap->sin_addr + 1); 174 175 ap->sin_len = 0; 176 while (--cp >= cplim) 177 if (*cp) { 178 (ap)->sin_len = cp - (char *) (ap) + 1; 179 break; 180 } 181 } 182 183 static int 184 in_mask2len(mask) 185 struct in_addr *mask; 186 { 187 int x, y; 188 u_char *p; 189 190 p = (u_char *)mask; 191 for (x = 0; x < sizeof(*mask); x++) { 192 if (p[x] != 0xff) 193 break; 194 } 195 y = 0; 196 if (x < sizeof(*mask)) { 197 for (y = 0; y < 8; y++) { 198 if ((p[x] & (0x80 >> y)) == 0) 199 break; 200 } 201 } 202 return (x * 8 + y); 203 } 204 205 static void 206 in_len2mask(struct in_addr *mask, int len) 207 { 208 int i; 209 u_char *p; 210 211 p = (u_char *)mask; 212 bzero(mask, sizeof(*mask)); 213 for (i = 0; i < len / 8; i++) 214 p[i] = 0xff; 215 if (len % 8) 216 p[i] = (0xff00 >> (len % 8)) & 0xff; 217 } 218 219 /* 220 * Generic internet control operations (ioctl's). 221 * 222 * ifp is NULL if not an interface-specific ioctl. 223 */ 224 /* ARGSUSED */ 225 int 226 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, 227 struct thread *td) 228 { 229 register struct ifreq *ifr = (struct ifreq *)data; 230 register struct in_ifaddr *ia, *iap; 231 register struct ifaddr *ifa; 232 struct in_addr allhosts_addr; 233 struct in_addr dst; 234 struct in_ifinfo *ii; 235 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 236 struct sockaddr_in oldaddr; 237 int error, hostIsNew, iaIsNew, maskIsNew; 238 int iaIsFirst; 239 240 ia = NULL; 241 iaIsFirst = 0; 242 iaIsNew = 0; 243 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 244 245 /* 246 * Filter out ioctls we implement directly; forward the rest on to 247 * in_lifaddr_ioctl() and ifp->if_ioctl(). 248 */ 249 switch (cmd) { 250 case SIOCAIFADDR: 251 case SIOCDIFADDR: 252 case SIOCGIFADDR: 253 case SIOCGIFBRDADDR: 254 case SIOCGIFDSTADDR: 255 case SIOCGIFNETMASK: 256 case SIOCSIFADDR: 257 case SIOCSIFBRDADDR: 258 case SIOCSIFDSTADDR: 259 case SIOCSIFNETMASK: 260 break; 261 262 case SIOCALIFADDR: 263 if (td != NULL) { 264 error = priv_check(td, PRIV_NET_ADDIFADDR); 265 if (error) 266 return (error); 267 } 268 if (ifp == NULL) 269 return (EINVAL); 270 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 271 272 case SIOCDLIFADDR: 273 if (td != NULL) { 274 error = priv_check(td, PRIV_NET_DELIFADDR); 275 if (error) 276 return (error); 277 } 278 if (ifp == NULL) 279 return (EINVAL); 280 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 281 282 case SIOCGLIFADDR: 283 if (ifp == NULL) 284 return (EINVAL); 285 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 286 287 default: 288 if (ifp == NULL || ifp->if_ioctl == NULL) 289 return (EOPNOTSUPP); 290 return ((*ifp->if_ioctl)(ifp, cmd, data)); 291 } 292 293 if (ifp == NULL) 294 return (EADDRNOTAVAIL); 295 296 /* 297 * Security checks before we get involved in any work. 298 */ 299 switch (cmd) { 300 case SIOCAIFADDR: 301 case SIOCSIFADDR: 302 case SIOCSIFBRDADDR: 303 case SIOCSIFNETMASK: 304 case SIOCSIFDSTADDR: 305 if (td != NULL) { 306 error = priv_check(td, PRIV_NET_ADDIFADDR); 307 if (error) 308 return (error); 309 } 310 break; 311 312 case SIOCDIFADDR: 313 if (td != NULL) { 314 error = priv_check(td, PRIV_NET_DELIFADDR); 315 if (error) 316 return (error); 317 } 318 break; 319 } 320 321 /* 322 * Find address for this interface, if it exists. 323 * 324 * If an alias address was specified, find that one instead of the 325 * first one on the interface, if possible. 326 */ 327 dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr; 328 IN_IFADDR_RLOCK(); 329 LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) { 330 if (iap->ia_ifp == ifp && 331 iap->ia_addr.sin_addr.s_addr == dst.s_addr) { 332 if (td == NULL || prison_check_ip4(td->td_ucred, 333 &dst) == 0) 334 ia = iap; 335 break; 336 } 337 } 338 if (ia != NULL) 339 ifa_ref(&ia->ia_ifa); 340 IN_IFADDR_RUNLOCK(); 341 if (ia == NULL) { 342 IF_ADDR_LOCK(ifp); 343 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 344 iap = ifatoia(ifa); 345 if (iap->ia_addr.sin_family == AF_INET) { 346 if (td != NULL && 347 prison_check_ip4(td->td_ucred, 348 &iap->ia_addr.sin_addr) != 0) 349 continue; 350 ia = iap; 351 break; 352 } 353 } 354 if (ia != NULL) 355 ifa_ref(&ia->ia_ifa); 356 IF_ADDR_UNLOCK(ifp); 357 } 358 if (ia == NULL) 359 iaIsFirst = 1; 360 361 error = 0; 362 switch (cmd) { 363 case SIOCAIFADDR: 364 case SIOCDIFADDR: 365 if (ifra->ifra_addr.sin_family == AF_INET) { 366 struct in_ifaddr *oia; 367 368 IN_IFADDR_RLOCK(); 369 for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) { 370 if (ia->ia_ifp == ifp && 371 ia->ia_addr.sin_addr.s_addr == 372 ifra->ifra_addr.sin_addr.s_addr) 373 break; 374 } 375 if (ia != NULL && ia != oia) 376 ifa_ref(&ia->ia_ifa); 377 if (oia != NULL && ia != oia) 378 ifa_free(&oia->ia_ifa); 379 IN_IFADDR_RUNLOCK(); 380 if ((ifp->if_flags & IFF_POINTOPOINT) 381 && (cmd == SIOCAIFADDR) 382 && (ifra->ifra_dstaddr.sin_addr.s_addr 383 == INADDR_ANY)) { 384 error = EDESTADDRREQ; 385 goto out; 386 } 387 } 388 if (cmd == SIOCDIFADDR && ia == NULL) { 389 error = EADDRNOTAVAIL; 390 goto out; 391 } 392 /* FALLTHROUGH */ 393 case SIOCSIFADDR: 394 case SIOCSIFNETMASK: 395 case SIOCSIFDSTADDR: 396 if (ia == NULL) { 397 ia = (struct in_ifaddr *) 398 malloc(sizeof *ia, M_IFADDR, M_NOWAIT | 399 M_ZERO); 400 if (ia == NULL) { 401 error = ENOBUFS; 402 goto out; 403 } 404 405 ifa = &ia->ia_ifa; 406 ifa_init(ifa); 407 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 408 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 409 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 410 411 ia->ia_sockmask.sin_len = 8; 412 ia->ia_sockmask.sin_family = AF_INET; 413 if (ifp->if_flags & IFF_BROADCAST) { 414 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr); 415 ia->ia_broadaddr.sin_family = AF_INET; 416 } 417 ia->ia_ifp = ifp; 418 419 ifa_ref(ifa); /* if_addrhead */ 420 IF_ADDR_LOCK(ifp); 421 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 422 IF_ADDR_UNLOCK(ifp); 423 ifa_ref(ifa); /* in_ifaddrhead */ 424 IN_IFADDR_WLOCK(); 425 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 426 IN_IFADDR_WUNLOCK(); 427 iaIsNew = 1; 428 } 429 break; 430 431 case SIOCSIFBRDADDR: 432 case SIOCGIFADDR: 433 case SIOCGIFNETMASK: 434 case SIOCGIFDSTADDR: 435 case SIOCGIFBRDADDR: 436 if (ia == NULL) { 437 error = EADDRNOTAVAIL; 438 goto out; 439 } 440 break; 441 } 442 443 /* 444 * Most paths in this switch return directly or via out. Only paths 445 * that remove the address break in order to hit common removal code. 446 */ 447 switch (cmd) { 448 case SIOCGIFADDR: 449 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr; 450 goto out; 451 452 case SIOCGIFBRDADDR: 453 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 454 error = EINVAL; 455 goto out; 456 } 457 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr; 458 goto out; 459 460 case SIOCGIFDSTADDR: 461 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 462 error = EINVAL; 463 goto out; 464 } 465 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr; 466 goto out; 467 468 case SIOCGIFNETMASK: 469 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask; 470 goto out; 471 472 case SIOCSIFDSTADDR: 473 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 474 error = EINVAL; 475 goto out; 476 } 477 oldaddr = ia->ia_dstaddr; 478 ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr; 479 if (ifp->if_ioctl != NULL) { 480 error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, 481 (caddr_t)ia); 482 if (error) { 483 ia->ia_dstaddr = oldaddr; 484 goto out; 485 } 486 } 487 if (ia->ia_flags & IFA_ROUTE) { 488 ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr; 489 rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST); 490 ia->ia_ifa.ifa_dstaddr = 491 (struct sockaddr *)&ia->ia_dstaddr; 492 rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP); 493 } 494 goto out; 495 496 case SIOCSIFBRDADDR: 497 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 498 error = EINVAL; 499 goto out; 500 } 501 ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr; 502 goto out; 503 504 case SIOCSIFADDR: 505 error = in_ifinit(ifp, ia, 506 (struct sockaddr_in *) &ifr->ifr_addr, 1); 507 if (error != 0 && iaIsNew) 508 break; 509 if (error == 0) { 510 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 511 if (iaIsFirst && 512 (ifp->if_flags & IFF_MULTICAST) != 0) { 513 error = in_joingroup(ifp, &allhosts_addr, 514 NULL, &ii->ii_allhosts); 515 } 516 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 517 } 518 error = 0; 519 goto out; 520 521 case SIOCSIFNETMASK: 522 ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr; 523 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 524 goto out; 525 526 case SIOCAIFADDR: 527 maskIsNew = 0; 528 hostIsNew = 1; 529 error = 0; 530 if (ia->ia_addr.sin_family == AF_INET) { 531 if (ifra->ifra_addr.sin_len == 0) { 532 ifra->ifra_addr = ia->ia_addr; 533 hostIsNew = 0; 534 } else if (ifra->ifra_addr.sin_addr.s_addr == 535 ia->ia_addr.sin_addr.s_addr) 536 hostIsNew = 0; 537 } 538 if (ifra->ifra_mask.sin_len) { 539 /* 540 * QL: XXX 541 * Need to scrub the prefix here in case 542 * the issued command is SIOCAIFADDR with 543 * the same address, but with a different 544 * prefix length. And if the prefix length 545 * is the same as before, then the call is 546 * un-necessarily executed here. 547 */ 548 in_ifscrub(ifp, ia); 549 ia->ia_sockmask = ifra->ifra_mask; 550 ia->ia_sockmask.sin_family = AF_INET; 551 ia->ia_subnetmask = 552 ntohl(ia->ia_sockmask.sin_addr.s_addr); 553 maskIsNew = 1; 554 } 555 if ((ifp->if_flags & IFF_POINTOPOINT) && 556 (ifra->ifra_dstaddr.sin_family == AF_INET)) { 557 in_ifscrub(ifp, ia); 558 ia->ia_dstaddr = ifra->ifra_dstaddr; 559 maskIsNew = 1; /* We lie; but the effect's the same */ 560 } 561 if (ifra->ifra_addr.sin_family == AF_INET && 562 (hostIsNew || maskIsNew)) 563 error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0); 564 if (error != 0 && iaIsNew) 565 break; 566 567 if ((ifp->if_flags & IFF_BROADCAST) && 568 (ifra->ifra_broadaddr.sin_family == AF_INET)) 569 ia->ia_broadaddr = ifra->ifra_broadaddr; 570 if (error == 0) { 571 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 572 if (iaIsFirst && 573 (ifp->if_flags & IFF_MULTICAST) != 0) { 574 error = in_joingroup(ifp, &allhosts_addr, 575 NULL, &ii->ii_allhosts); 576 } 577 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 578 } 579 goto out; 580 581 case SIOCDIFADDR: 582 /* 583 * in_ifscrub kills the interface route. 584 */ 585 in_ifscrub(ifp, ia); 586 587 /* 588 * in_ifadown gets rid of all the rest of 589 * the routes. This is not quite the right 590 * thing to do, but at least if we are running 591 * a routing process they will come back. 592 */ 593 in_ifadown(&ia->ia_ifa, 1); 594 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 595 error = 0; 596 break; 597 598 default: 599 panic("in_control: unsupported ioctl"); 600 } 601 602 IF_ADDR_LOCK(ifp); 603 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 604 IF_ADDR_UNLOCK(ifp); 605 ifa_free(&ia->ia_ifa); /* if_addrhead */ 606 607 IN_IFADDR_WLOCK(); 608 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 609 if (ia->ia_addr.sin_family == AF_INET) { 610 struct in_ifaddr *if_ia; 611 612 LIST_REMOVE(ia, ia_hash); 613 IN_IFADDR_WUNLOCK(); 614 /* 615 * If this is the last IPv4 address configured on this 616 * interface, leave the all-hosts group. 617 * No state-change report need be transmitted. 618 */ 619 if_ia = NULL; 620 IFP_TO_IA(ifp, if_ia); 621 if (if_ia == NULL) { 622 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 623 IN_MULTI_LOCK(); 624 if (ii->ii_allhosts) { 625 (void)in_leavegroup_locked(ii->ii_allhosts, 626 NULL); 627 ii->ii_allhosts = NULL; 628 } 629 IN_MULTI_UNLOCK(); 630 } else 631 ifa_free(&if_ia->ia_ifa); 632 } else 633 IN_IFADDR_WUNLOCK(); 634 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 635 out: 636 if (ia != NULL) 637 ifa_free(&ia->ia_ifa); 638 return (error); 639 } 640 641 /* 642 * SIOC[GAD]LIFADDR. 643 * SIOCGLIFADDR: get first address. (?!?) 644 * SIOCGLIFADDR with IFLR_PREFIX: 645 * get first address that matches the specified prefix. 646 * SIOCALIFADDR: add the specified address. 647 * SIOCALIFADDR with IFLR_PREFIX: 648 * EINVAL since we can't deduce hostid part of the address. 649 * SIOCDLIFADDR: delete the specified address. 650 * SIOCDLIFADDR with IFLR_PREFIX: 651 * delete the first address that matches the specified prefix. 652 * return values: 653 * EINVAL on invalid parameters 654 * EADDRNOTAVAIL on prefix match failed/specified address not found 655 * other values may be returned from in_ioctl() 656 */ 657 static int 658 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data, 659 struct ifnet *ifp, struct thread *td) 660 { 661 struct if_laddrreq *iflr = (struct if_laddrreq *)data; 662 struct ifaddr *ifa; 663 664 /* sanity checks */ 665 if (data == NULL || ifp == NULL) { 666 panic("invalid argument to in_lifaddr_ioctl"); 667 /*NOTRECHED*/ 668 } 669 670 switch (cmd) { 671 case SIOCGLIFADDR: 672 /* address must be specified on GET with IFLR_PREFIX */ 673 if ((iflr->flags & IFLR_PREFIX) == 0) 674 break; 675 /*FALLTHROUGH*/ 676 case SIOCALIFADDR: 677 case SIOCDLIFADDR: 678 /* address must be specified on ADD and DELETE */ 679 if (iflr->addr.ss_family != AF_INET) 680 return (EINVAL); 681 if (iflr->addr.ss_len != sizeof(struct sockaddr_in)) 682 return (EINVAL); 683 /* XXX need improvement */ 684 if (iflr->dstaddr.ss_family 685 && iflr->dstaddr.ss_family != AF_INET) 686 return (EINVAL); 687 if (iflr->dstaddr.ss_family 688 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in)) 689 return (EINVAL); 690 break; 691 default: /*shouldn't happen*/ 692 return (EOPNOTSUPP); 693 } 694 if (sizeof(struct in_addr) * 8 < iflr->prefixlen) 695 return (EINVAL); 696 697 switch (cmd) { 698 case SIOCALIFADDR: 699 { 700 struct in_aliasreq ifra; 701 702 if (iflr->flags & IFLR_PREFIX) 703 return (EINVAL); 704 705 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */ 706 bzero(&ifra, sizeof(ifra)); 707 bcopy(iflr->iflr_name, ifra.ifra_name, 708 sizeof(ifra.ifra_name)); 709 710 bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len); 711 712 if (iflr->dstaddr.ss_family) { /*XXX*/ 713 bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr, 714 iflr->dstaddr.ss_len); 715 } 716 717 ifra.ifra_mask.sin_family = AF_INET; 718 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in); 719 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen); 720 721 return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td)); 722 } 723 case SIOCGLIFADDR: 724 case SIOCDLIFADDR: 725 { 726 struct in_ifaddr *ia; 727 struct in_addr mask, candidate, match; 728 struct sockaddr_in *sin; 729 730 bzero(&mask, sizeof(mask)); 731 bzero(&match, sizeof(match)); 732 if (iflr->flags & IFLR_PREFIX) { 733 /* lookup a prefix rather than address. */ 734 in_len2mask(&mask, iflr->prefixlen); 735 736 sin = (struct sockaddr_in *)&iflr->addr; 737 match.s_addr = sin->sin_addr.s_addr; 738 match.s_addr &= mask.s_addr; 739 740 /* if you set extra bits, that's wrong */ 741 if (match.s_addr != sin->sin_addr.s_addr) 742 return (EINVAL); 743 744 } else { 745 /* on getting an address, take the 1st match */ 746 /* on deleting an address, do exact match */ 747 if (cmd != SIOCGLIFADDR) { 748 in_len2mask(&mask, 32); 749 sin = (struct sockaddr_in *)&iflr->addr; 750 match.s_addr = sin->sin_addr.s_addr; 751 } 752 } 753 754 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 755 if (ifa->ifa_addr->sa_family != AF_INET6) 756 continue; 757 if (match.s_addr == 0) 758 break; 759 candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr; 760 candidate.s_addr &= mask.s_addr; 761 if (candidate.s_addr == match.s_addr) 762 break; 763 } 764 if (ifa == NULL) 765 return (EADDRNOTAVAIL); 766 ia = (struct in_ifaddr *)ifa; 767 768 if (cmd == SIOCGLIFADDR) { 769 /* fill in the if_laddrreq structure */ 770 bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len); 771 772 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 773 bcopy(&ia->ia_dstaddr, &iflr->dstaddr, 774 ia->ia_dstaddr.sin_len); 775 } else 776 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr)); 777 778 iflr->prefixlen = 779 in_mask2len(&ia->ia_sockmask.sin_addr); 780 781 iflr->flags = 0; /*XXX*/ 782 783 return (0); 784 } else { 785 struct in_aliasreq ifra; 786 787 /* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */ 788 bzero(&ifra, sizeof(ifra)); 789 bcopy(iflr->iflr_name, ifra.ifra_name, 790 sizeof(ifra.ifra_name)); 791 792 bcopy(&ia->ia_addr, &ifra.ifra_addr, 793 ia->ia_addr.sin_len); 794 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 795 bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr, 796 ia->ia_dstaddr.sin_len); 797 } 798 bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr, 799 ia->ia_sockmask.sin_len); 800 801 return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra, 802 ifp, td)); 803 } 804 } 805 } 806 807 return (EOPNOTSUPP); /*just for safety*/ 808 } 809 810 /* 811 * Delete any existing route for an interface. 812 */ 813 void 814 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia) 815 { 816 817 in_scrubprefix(ia); 818 } 819 820 /* 821 * Initialize an interface's internet address 822 * and routing table entry. 823 */ 824 static int 825 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin, 826 int scrub) 827 { 828 register u_long i = ntohl(sin->sin_addr.s_addr); 829 struct sockaddr_in oldaddr; 830 int s = splimp(), flags = RTF_UP, error = 0; 831 832 oldaddr = ia->ia_addr; 833 if (oldaddr.sin_family == AF_INET) 834 LIST_REMOVE(ia, ia_hash); 835 ia->ia_addr = *sin; 836 if (ia->ia_addr.sin_family == AF_INET) { 837 IN_IFADDR_WLOCK(); 838 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), 839 ia, ia_hash); 840 IN_IFADDR_WUNLOCK(); 841 } 842 /* 843 * Give the interface a chance to initialize 844 * if this is its first address, 845 * and to validate the address if necessary. 846 */ 847 if (ifp->if_ioctl != NULL) { 848 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 849 if (error) { 850 splx(s); 851 /* LIST_REMOVE(ia, ia_hash) is done in in_control */ 852 ia->ia_addr = oldaddr; 853 IN_IFADDR_WLOCK(); 854 if (ia->ia_addr.sin_family == AF_INET) 855 LIST_INSERT_HEAD(INADDR_HASH( 856 ia->ia_addr.sin_addr.s_addr), ia, ia_hash); 857 else 858 /* 859 * If oldaddr family is not AF_INET (e.g. 860 * interface has been just created) in_control 861 * does not call LIST_REMOVE, and we end up 862 * with bogus ia entries in hash 863 */ 864 LIST_REMOVE(ia, ia_hash); 865 IN_IFADDR_WUNLOCK(); 866 return (error); 867 } 868 } 869 splx(s); 870 if (scrub) { 871 ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr; 872 in_ifscrub(ifp, ia); 873 ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr; 874 } 875 if (IN_CLASSA(i)) 876 ia->ia_netmask = IN_CLASSA_NET; 877 else if (IN_CLASSB(i)) 878 ia->ia_netmask = IN_CLASSB_NET; 879 else 880 ia->ia_netmask = IN_CLASSC_NET; 881 /* 882 * The subnet mask usually includes at least the standard network part, 883 * but may may be smaller in the case of supernetting. 884 * If it is set, we believe it. 885 */ 886 if (ia->ia_subnetmask == 0) { 887 ia->ia_subnetmask = ia->ia_netmask; 888 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 889 } else 890 ia->ia_netmask &= ia->ia_subnetmask; 891 ia->ia_net = i & ia->ia_netmask; 892 ia->ia_subnet = i & ia->ia_subnetmask; 893 in_socktrim(&ia->ia_sockmask); 894 #ifdef DEV_CARP 895 /* 896 * XXX: carp(4) does not have interface route 897 */ 898 if (ifp->if_type == IFT_CARP) 899 return (0); 900 #endif 901 /* 902 * Add route for the network. 903 */ 904 ia->ia_ifa.ifa_metric = ifp->if_metric; 905 if (ifp->if_flags & IFF_BROADCAST) { 906 ia->ia_broadaddr.sin_addr.s_addr = 907 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 908 ia->ia_netbroadcast.s_addr = 909 htonl(ia->ia_net | ~ ia->ia_netmask); 910 } else if (ifp->if_flags & IFF_LOOPBACK) { 911 ia->ia_dstaddr = ia->ia_addr; 912 flags |= RTF_HOST; 913 } else if (ifp->if_flags & IFF_POINTOPOINT) { 914 if (ia->ia_dstaddr.sin_family != AF_INET) 915 return (0); 916 flags |= RTF_HOST; 917 } 918 if ((error = in_addprefix(ia, flags)) != 0) 919 return (error); 920 921 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY) 922 return (0); 923 924 if (ifp->if_flags & IFF_POINTOPOINT) { 925 if (ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr) 926 return (0); 927 } 928 929 930 /* 931 * add a loopback route to self 932 */ 933 if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) { 934 struct route ia_ro; 935 936 bzero(&ia_ro, sizeof(ia_ro)); 937 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr; 938 rtalloc_ign_fib(&ia_ro, 0, 0); 939 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) && 940 (ia_ro.ro_rt->rt_ifp == V_loif)) { 941 RT_LOCK(ia_ro.ro_rt); 942 RT_ADDREF(ia_ro.ro_rt); 943 RTFREE_LOCKED(ia_ro.ro_rt); 944 } else 945 error = ifa_add_loopback_route((struct ifaddr *)ia, 946 (struct sockaddr *)&ia->ia_addr); 947 if (error == 0) 948 ia->ia_flags |= IFA_RTSELF; 949 if (ia_ro.ro_rt != NULL) 950 RTFREE(ia_ro.ro_rt); 951 } 952 953 return (error); 954 } 955 956 #define rtinitflags(x) \ 957 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 958 ? RTF_HOST : 0) 959 960 /* 961 * Generate a routing message when inserting or deleting 962 * an interface address alias. 963 */ 964 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix, 965 struct in_ifaddr *target) 966 { 967 struct route pfx_ro; 968 struct sockaddr_in *pfx_addr; 969 struct rtentry msg_rt; 970 971 /* QL: XXX 972 * This is a bit questionable because there is no 973 * additional route entry added/deleted for an address 974 * alias. Therefore this route report is inaccurate. 975 */ 976 bzero(&pfx_ro, sizeof(pfx_ro)); 977 pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst); 978 pfx_addr->sin_len = sizeof(*pfx_addr); 979 pfx_addr->sin_family = AF_INET; 980 pfx_addr->sin_addr = *prefix; 981 rtalloc_ign_fib(&pfx_ro, 0, 0); 982 if (pfx_ro.ro_rt != NULL) { 983 msg_rt = *pfx_ro.ro_rt; 984 985 /* QL: XXX 986 * Point the gateway to the new interface 987 * address as if a new prefix route entry has 988 * been added through the new address alias. 989 * All other parts of the rtentry is accurate, 990 * e.g., rt_key, rt_mask, rt_ifp etc. 991 */ 992 msg_rt.rt_gateway = 993 (struct sockaddr *)&target->ia_addr; 994 rt_newaddrmsg(cmd, 995 (struct ifaddr *)target, 996 0, &msg_rt); 997 RTFREE(pfx_ro.ro_rt); 998 } 999 return; 1000 } 1001 1002 /* 1003 * Check if we have a route for the given prefix already or add one accordingly. 1004 */ 1005 static int 1006 in_addprefix(struct in_ifaddr *target, int flags) 1007 { 1008 struct in_ifaddr *ia; 1009 struct in_addr prefix, mask, p, m; 1010 int error; 1011 1012 if ((flags & RTF_HOST) != 0) { 1013 prefix = target->ia_dstaddr.sin_addr; 1014 mask.s_addr = 0; 1015 } else { 1016 prefix = target->ia_addr.sin_addr; 1017 mask = target->ia_sockmask.sin_addr; 1018 prefix.s_addr &= mask.s_addr; 1019 } 1020 1021 IN_IFADDR_RLOCK(); 1022 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1023 if (rtinitflags(ia)) { 1024 p = ia->ia_addr.sin_addr; 1025 1026 if (prefix.s_addr != p.s_addr) 1027 continue; 1028 } else { 1029 p = ia->ia_addr.sin_addr; 1030 m = ia->ia_sockmask.sin_addr; 1031 p.s_addr &= m.s_addr; 1032 1033 if (prefix.s_addr != p.s_addr || 1034 mask.s_addr != m.s_addr) 1035 continue; 1036 } 1037 1038 /* 1039 * If we got a matching prefix route inserted by other 1040 * interface address, we are done here. 1041 */ 1042 if (ia->ia_flags & IFA_ROUTE) { 1043 #ifdef RADIX_MPATH 1044 if (ia->ia_addr.sin_addr.s_addr == 1045 target->ia_addr.sin_addr.s_addr) 1046 return (EEXIST); 1047 else 1048 break; 1049 #endif 1050 if (V_sameprefixcarponly && 1051 target->ia_ifp->if_type != IFT_CARP && 1052 ia->ia_ifp->if_type != IFT_CARP) { 1053 IN_IFADDR_RUNLOCK(); 1054 return (EEXIST); 1055 } else { 1056 in_addralias_rtmsg(RTM_ADD, &prefix, target); 1057 IN_IFADDR_RUNLOCK(); 1058 return (0); 1059 } 1060 } 1061 } 1062 IN_IFADDR_RUNLOCK(); 1063 1064 /* 1065 * No-one seem to have this prefix route, so we try to insert it. 1066 */ 1067 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 1068 if (!error) 1069 target->ia_flags |= IFA_ROUTE; 1070 return (error); 1071 } 1072 1073 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr); 1074 1075 /* 1076 * If there is no other address in the system that can serve a route to the 1077 * same prefix, remove the route. Hand over the route to the new address 1078 * otherwise. 1079 */ 1080 static int 1081 in_scrubprefix(struct in_ifaddr *target) 1082 { 1083 struct in_ifaddr *ia; 1084 struct in_addr prefix, mask, p; 1085 int error = 0; 1086 struct sockaddr_in prefix0, mask0; 1087 1088 /* 1089 * Remove the loopback route to the interface address. 1090 * The "useloopback" setting is not consulted because if the 1091 * user configures an interface address, turns off this 1092 * setting, and then tries to delete that interface address, 1093 * checking the current setting of "useloopback" would leave 1094 * that interface address loopback route untouched, which 1095 * would be wrong. Therefore the interface address loopback route 1096 * deletion is unconditional. 1097 */ 1098 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 1099 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 1100 (target->ia_flags & IFA_RTSELF)) { 1101 struct route ia_ro; 1102 int freeit = 0; 1103 1104 bzero(&ia_ro, sizeof(ia_ro)); 1105 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr; 1106 rtalloc_ign_fib(&ia_ro, 0, 0); 1107 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) && 1108 (ia_ro.ro_rt->rt_ifp == V_loif)) { 1109 RT_LOCK(ia_ro.ro_rt); 1110 if (ia_ro.ro_rt->rt_refcnt <= 1) 1111 freeit = 1; 1112 else 1113 RT_REMREF(ia_ro.ro_rt); 1114 RTFREE_LOCKED(ia_ro.ro_rt); 1115 } 1116 if (freeit) 1117 error = ifa_del_loopback_route((struct ifaddr *)target, 1118 (struct sockaddr *)&target->ia_addr); 1119 if (error == 0) 1120 target->ia_flags &= ~IFA_RTSELF; 1121 /* remove arp cache */ 1122 arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr); 1123 } 1124 1125 if (rtinitflags(target)) 1126 prefix = target->ia_dstaddr.sin_addr; 1127 else { 1128 prefix = target->ia_addr.sin_addr; 1129 mask = target->ia_sockmask.sin_addr; 1130 prefix.s_addr &= mask.s_addr; 1131 } 1132 1133 if ((target->ia_flags & IFA_ROUTE) == 0) { 1134 in_addralias_rtmsg(RTM_DELETE, &prefix, target); 1135 return (0); 1136 } 1137 1138 IN_IFADDR_RLOCK(); 1139 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1140 if (rtinitflags(ia)) 1141 p = ia->ia_dstaddr.sin_addr; 1142 else { 1143 p = ia->ia_addr.sin_addr; 1144 p.s_addr &= ia->ia_sockmask.sin_addr.s_addr; 1145 } 1146 1147 if (prefix.s_addr != p.s_addr) 1148 continue; 1149 1150 /* 1151 * If we got a matching prefix address, move IFA_ROUTE and 1152 * the route itself to it. Make sure that routing daemons 1153 * get a heads-up. 1154 * 1155 * XXX: a special case for carp(4) interface 1156 */ 1157 if ((ia->ia_flags & IFA_ROUTE) == 0 1158 #ifdef DEV_CARP 1159 && (ia->ia_ifp->if_type != IFT_CARP) 1160 #endif 1161 ) { 1162 IN_IFADDR_RUNLOCK(); 1163 rtinit(&(target->ia_ifa), (int)RTM_DELETE, 1164 rtinitflags(target)); 1165 target->ia_flags &= ~IFA_ROUTE; 1166 1167 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 1168 rtinitflags(ia) | RTF_UP); 1169 if (error == 0) 1170 ia->ia_flags |= IFA_ROUTE; 1171 return (error); 1172 } 1173 } 1174 IN_IFADDR_RUNLOCK(); 1175 1176 /* 1177 * remove all L2 entries on the given prefix 1178 */ 1179 bzero(&prefix0, sizeof(prefix0)); 1180 prefix0.sin_len = sizeof(prefix0); 1181 prefix0.sin_family = AF_INET; 1182 prefix0.sin_addr.s_addr = target->ia_subnet; 1183 bzero(&mask0, sizeof(mask0)); 1184 mask0.sin_len = sizeof(mask0); 1185 mask0.sin_family = AF_INET; 1186 mask0.sin_addr.s_addr = target->ia_subnetmask; 1187 lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0, 1188 (struct sockaddr *)&mask0); 1189 1190 /* 1191 * As no-one seem to have this prefix, we can remove the route. 1192 */ 1193 rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 1194 target->ia_flags &= ~IFA_ROUTE; 1195 return (0); 1196 } 1197 1198 #undef rtinitflags 1199 1200 /* 1201 * Return 1 if the address might be a local broadcast address. 1202 */ 1203 int 1204 in_broadcast(struct in_addr in, struct ifnet *ifp) 1205 { 1206 register struct ifaddr *ifa; 1207 u_long t; 1208 1209 if (in.s_addr == INADDR_BROADCAST || 1210 in.s_addr == INADDR_ANY) 1211 return (1); 1212 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1213 return (0); 1214 t = ntohl(in.s_addr); 1215 /* 1216 * Look through the list of addresses for a match 1217 * with a broadcast address. 1218 */ 1219 #define ia ((struct in_ifaddr *)ifa) 1220 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1221 if (ifa->ifa_addr->sa_family == AF_INET && 1222 (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 1223 in.s_addr == ia->ia_netbroadcast.s_addr || 1224 /* 1225 * Check for old-style (host 0) broadcast. 1226 */ 1227 t == ia->ia_subnet || t == ia->ia_net) && 1228 /* 1229 * Check for an all one subnetmask. These 1230 * only exist when an interface gets a secondary 1231 * address. 1232 */ 1233 ia->ia_subnetmask != (u_long)0xffffffff) 1234 return (1); 1235 return (0); 1236 #undef ia 1237 } 1238 1239 /* 1240 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 1241 */ 1242 void 1243 in_ifdetach(struct ifnet *ifp) 1244 { 1245 1246 in_pcbpurgeif0(&V_ripcbinfo, ifp); 1247 in_pcbpurgeif0(&V_udbinfo, ifp); 1248 in_purgemaddrs(ifp); 1249 } 1250 1251 /* 1252 * Delete all IPv4 multicast address records, and associated link-layer 1253 * multicast address records, associated with ifp. 1254 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1255 * XXX This should not race with ifma_protospec being set during 1256 * a new allocation, if it does, we have bigger problems. 1257 */ 1258 static void 1259 in_purgemaddrs(struct ifnet *ifp) 1260 { 1261 LIST_HEAD(,in_multi) purgeinms; 1262 struct in_multi *inm, *tinm; 1263 struct ifmultiaddr *ifma; 1264 1265 LIST_INIT(&purgeinms); 1266 IN_MULTI_LOCK(); 1267 1268 /* 1269 * Extract list of in_multi associated with the detaching ifp 1270 * which the PF_INET layer is about to release. 1271 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1272 * by code further down. 1273 */ 1274 IF_ADDR_LOCK(ifp); 1275 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1276 if (ifma->ifma_addr->sa_family != AF_INET || 1277 ifma->ifma_protospec == NULL) 1278 continue; 1279 #if 0 1280 KASSERT(ifma->ifma_protospec != NULL, 1281 ("%s: ifma_protospec is NULL", __func__)); 1282 #endif 1283 inm = (struct in_multi *)ifma->ifma_protospec; 1284 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 1285 } 1286 IF_ADDR_UNLOCK(ifp); 1287 1288 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 1289 LIST_REMOVE(inm, inm_link); 1290 inm_release_locked(inm); 1291 } 1292 igmp_ifdetach(ifp); 1293 1294 IN_MULTI_UNLOCK(); 1295 } 1296 1297 #include <net/if_dl.h> 1298 #include <netinet/if_ether.h> 1299 1300 struct in_llentry { 1301 struct llentry base; 1302 struct sockaddr_in l3_addr4; 1303 }; 1304 1305 static struct llentry * 1306 in_lltable_new(const struct sockaddr *l3addr, u_int flags) 1307 { 1308 struct in_llentry *lle; 1309 1310 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO); 1311 if (lle == NULL) /* NB: caller generates msg */ 1312 return NULL; 1313 1314 callout_init(&lle->base.la_timer, CALLOUT_MPSAFE); 1315 /* 1316 * For IPv4 this will trigger "arpresolve" to generate 1317 * an ARP request. 1318 */ 1319 lle->base.la_expire = time_second; /* mark expired */ 1320 lle->l3_addr4 = *(const struct sockaddr_in *)l3addr; 1321 lle->base.lle_refcnt = 1; 1322 LLE_LOCK_INIT(&lle->base); 1323 return &lle->base; 1324 } 1325 1326 /* 1327 * Deletes an address from the address table. 1328 * This function is called by the timer functions 1329 * such as arptimer() and nd6_llinfo_timer(), and 1330 * the caller does the locking. 1331 */ 1332 static void 1333 in_lltable_free(struct lltable *llt, struct llentry *lle) 1334 { 1335 LLE_WUNLOCK(lle); 1336 LLE_LOCK_DESTROY(lle); 1337 free(lle, M_LLTABLE); 1338 } 1339 1340 1341 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1342 (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 1343 1344 static void 1345 in_lltable_prefix_free(struct lltable *llt, 1346 const struct sockaddr *prefix, 1347 const struct sockaddr *mask) 1348 { 1349 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 1350 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 1351 struct llentry *lle, *next; 1352 register int i; 1353 1354 for (i=0; i < LLTBL_HASHTBL_SIZE; i++) { 1355 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 1356 1357 if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle), 1358 pfx, msk)) { 1359 int canceled; 1360 1361 canceled = callout_drain(&lle->la_timer); 1362 LLE_WLOCK(lle); 1363 if (canceled) 1364 LLE_REMREF(lle); 1365 llentry_free(lle); 1366 } 1367 } 1368 } 1369 } 1370 1371 1372 static int 1373 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1374 { 1375 struct rtentry *rt; 1376 1377 KASSERT(l3addr->sa_family == AF_INET, 1378 ("sin_family %d", l3addr->sa_family)); 1379 1380 /* XXX rtalloc1 should take a const param */ 1381 rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0); 1382 if (rt == NULL || (!(flags & LLE_PUB) && 1383 ((rt->rt_flags & RTF_GATEWAY) || 1384 (rt->rt_ifp != ifp)))) { 1385 #ifdef DIAGNOSTIC 1386 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1387 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 1388 #endif 1389 if (rt != NULL) 1390 RTFREE_LOCKED(rt); 1391 return (EINVAL); 1392 } 1393 RTFREE_LOCKED(rt); 1394 return 0; 1395 } 1396 1397 /* 1398 * Return NULL if not found or marked for deletion. 1399 * If found return lle read locked. 1400 */ 1401 static struct llentry * 1402 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1403 { 1404 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1405 struct ifnet *ifp = llt->llt_ifp; 1406 struct llentry *lle; 1407 struct llentries *lleh; 1408 u_int hashkey; 1409 1410 IF_AFDATA_LOCK_ASSERT(ifp); 1411 KASSERT(l3addr->sa_family == AF_INET, 1412 ("sin_family %d", l3addr->sa_family)); 1413 1414 hashkey = sin->sin_addr.s_addr; 1415 lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; 1416 LIST_FOREACH(lle, lleh, lle_next) { 1417 struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle); 1418 if (lle->la_flags & LLE_DELETED) 1419 continue; 1420 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr) 1421 break; 1422 } 1423 if (lle == NULL) { 1424 #ifdef DIAGNOSTIC 1425 if (flags & LLE_DELETE) 1426 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); 1427 #endif 1428 if (!(flags & LLE_CREATE)) 1429 return (NULL); 1430 /* 1431 * A route that covers the given address must have 1432 * been installed 1st because we are doing a resolution, 1433 * verify this. 1434 */ 1435 if (!(flags & LLE_IFADDR) && 1436 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1437 goto done; 1438 1439 lle = in_lltable_new(l3addr, flags); 1440 if (lle == NULL) { 1441 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1442 goto done; 1443 } 1444 lle->la_flags = flags & ~LLE_CREATE; 1445 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { 1446 bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); 1447 lle->la_flags |= (LLE_VALID | LLE_STATIC); 1448 } 1449 1450 lle->lle_tbl = llt; 1451 lle->lle_head = lleh; 1452 LIST_INSERT_HEAD(lleh, lle, lle_next); 1453 } else if (flags & LLE_DELETE) { 1454 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) { 1455 LLE_WLOCK(lle); 1456 lle->la_flags = LLE_DELETED; 1457 EVENTHANDLER_INVOKE(arp_update_event, lle); 1458 LLE_WUNLOCK(lle); 1459 #ifdef DIAGNOSTIC 1460 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1461 #endif 1462 } 1463 lle = (void *)-1; 1464 1465 } 1466 if (LLE_IS_VALID(lle)) { 1467 if (flags & LLE_EXCLUSIVE) 1468 LLE_WLOCK(lle); 1469 else 1470 LLE_RLOCK(lle); 1471 } 1472 done: 1473 return (lle); 1474 } 1475 1476 static int 1477 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr) 1478 { 1479 #define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle)) 1480 struct ifnet *ifp = llt->llt_ifp; 1481 struct llentry *lle; 1482 /* XXX stack use */ 1483 struct { 1484 struct rt_msghdr rtm; 1485 struct sockaddr_inarp sin; 1486 struct sockaddr_dl sdl; 1487 } arpc; 1488 int error, i; 1489 1490 LLTABLE_LOCK_ASSERT(); 1491 1492 error = 0; 1493 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1494 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1495 struct sockaddr_dl *sdl; 1496 1497 /* skip deleted entries */ 1498 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1499 continue; 1500 /* Skip if jailed and not a valid IP of the prison. */ 1501 if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0) 1502 continue; 1503 /* 1504 * produce a msg made of: 1505 * struct rt_msghdr; 1506 * struct sockaddr_inarp; (IPv4) 1507 * struct sockaddr_dl; 1508 */ 1509 bzero(&arpc, sizeof(arpc)); 1510 arpc.rtm.rtm_msglen = sizeof(arpc); 1511 arpc.rtm.rtm_version = RTM_VERSION; 1512 arpc.rtm.rtm_type = RTM_GET; 1513 arpc.rtm.rtm_flags = RTF_UP; 1514 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1515 arpc.sin.sin_family = AF_INET; 1516 arpc.sin.sin_len = sizeof(arpc.sin); 1517 arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr; 1518 1519 /* publish */ 1520 if (lle->la_flags & LLE_PUB) { 1521 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1522 /* proxy only */ 1523 if (lle->la_flags & LLE_PROXY) 1524 arpc.sin.sin_other = SIN_PROXY; 1525 } 1526 1527 sdl = &arpc.sdl; 1528 sdl->sdl_family = AF_LINK; 1529 sdl->sdl_len = sizeof(*sdl); 1530 sdl->sdl_index = ifp->if_index; 1531 sdl->sdl_type = ifp->if_type; 1532 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1533 sdl->sdl_alen = ifp->if_addrlen; 1534 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1535 } else { 1536 sdl->sdl_alen = 0; 1537 bzero(LLADDR(sdl), ifp->if_addrlen); 1538 } 1539 1540 arpc.rtm.rtm_rmx.rmx_expire = 1541 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1542 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1543 if (lle->la_flags & LLE_STATIC) 1544 arpc.rtm.rtm_flags |= RTF_STATIC; 1545 arpc.rtm.rtm_index = ifp->if_index; 1546 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1547 if (error) 1548 break; 1549 } 1550 } 1551 return error; 1552 #undef SIN 1553 } 1554 1555 void * 1556 in_domifattach(struct ifnet *ifp) 1557 { 1558 struct in_ifinfo *ii; 1559 struct lltable *llt; 1560 1561 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1562 1563 llt = lltable_init(ifp, AF_INET); 1564 if (llt != NULL) { 1565 llt->llt_new = in_lltable_new; 1566 llt->llt_free = in_lltable_free; 1567 llt->llt_prefix_free = in_lltable_prefix_free; 1568 llt->llt_rtcheck = in_lltable_rtcheck; 1569 llt->llt_lookup = in_lltable_lookup; 1570 llt->llt_dump = in_lltable_dump; 1571 } 1572 ii->ii_llt = llt; 1573 1574 ii->ii_igmp = igmp_domifattach(ifp); 1575 1576 return ii; 1577 } 1578 1579 void 1580 in_domifdetach(struct ifnet *ifp, void *aux) 1581 { 1582 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1583 1584 igmp_domifdetach(ifp); 1585 lltable_free(ii->ii_llt); 1586 free(ii, M_IFADDR); 1587 } 1588