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/systm.h> 40 #include <sys/sockio.h> 41 #include <sys/malloc.h> 42 #include <sys/priv.h> 43 #include <sys/socket.h> 44 #include <sys/jail.h> 45 #include <sys/kernel.h> 46 #include <sys/proc.h> 47 #include <sys/sysctl.h> 48 #include <sys/syslog.h> 49 50 #include <net/if.h> 51 #include <net/if_var.h> 52 #include <net/if_arp.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/if_ether.h> 60 #include <netinet/in.h> 61 #include <netinet/in_var.h> 62 #include <netinet/in_pcb.h> 63 #include <netinet/ip_var.h> 64 #include <netinet/ip_carp.h> 65 #include <netinet/igmp_var.h> 66 #include <netinet/udp.h> 67 #include <netinet/udp_var.h> 68 69 static int in_mask2len(struct in_addr *); 70 static void in_len2mask(struct in_addr *, int); 71 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t, 72 struct ifnet *, struct thread *); 73 74 static void in_socktrim(struct sockaddr_in *); 75 static int in_ifinit(struct ifnet *, struct in_ifaddr *, 76 struct sockaddr_in *, int, int); 77 static void in_purgemaddrs(struct ifnet *); 78 79 static VNET_DEFINE(int, nosameprefix); 80 #define V_nosameprefix VNET(nosameprefix) 81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, no_same_prefix, 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 /* 89 * Return 1 if an internet address is for a ``local'' host 90 * (one to which we have a connection). 91 */ 92 int 93 in_localaddr(struct in_addr in) 94 { 95 register u_long i = ntohl(in.s_addr); 96 register struct in_ifaddr *ia; 97 98 IN_IFADDR_RLOCK(); 99 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 100 if ((i & ia->ia_subnetmask) == ia->ia_subnet) { 101 IN_IFADDR_RUNLOCK(); 102 return (1); 103 } 104 } 105 IN_IFADDR_RUNLOCK(); 106 return (0); 107 } 108 109 /* 110 * Return 1 if an internet address is for the local host and configured 111 * on one of its interfaces. 112 */ 113 int 114 in_localip(struct in_addr in) 115 { 116 struct in_ifaddr *ia; 117 118 IN_IFADDR_RLOCK(); 119 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) { 120 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) { 121 IN_IFADDR_RUNLOCK(); 122 return (1); 123 } 124 } 125 IN_IFADDR_RUNLOCK(); 126 return (0); 127 } 128 129 /* 130 * Determine whether an IP address is in a reserved set of addresses 131 * that may not be forwarded, or whether datagrams to that destination 132 * may be forwarded. 133 */ 134 int 135 in_canforward(struct in_addr in) 136 { 137 register u_long i = ntohl(in.s_addr); 138 register u_long net; 139 140 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i)) 141 return (0); 142 if (IN_CLASSA(i)) { 143 net = i & IN_CLASSA_NET; 144 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 145 return (0); 146 } 147 return (1); 148 } 149 150 /* 151 * Trim a mask in a sockaddr 152 */ 153 static void 154 in_socktrim(struct sockaddr_in *ap) 155 { 156 register char *cplim = (char *) &ap->sin_addr; 157 register char *cp = (char *) (&ap->sin_addr + 1); 158 159 ap->sin_len = 0; 160 while (--cp >= cplim) 161 if (*cp) { 162 (ap)->sin_len = cp - (char *) (ap) + 1; 163 break; 164 } 165 } 166 167 static int 168 in_mask2len(mask) 169 struct in_addr *mask; 170 { 171 int x, y; 172 u_char *p; 173 174 p = (u_char *)mask; 175 for (x = 0; x < sizeof(*mask); x++) { 176 if (p[x] != 0xff) 177 break; 178 } 179 y = 0; 180 if (x < sizeof(*mask)) { 181 for (y = 0; y < 8; y++) { 182 if ((p[x] & (0x80 >> y)) == 0) 183 break; 184 } 185 } 186 return (x * 8 + y); 187 } 188 189 static void 190 in_len2mask(struct in_addr *mask, int len) 191 { 192 int i; 193 u_char *p; 194 195 p = (u_char *)mask; 196 bzero(mask, sizeof(*mask)); 197 for (i = 0; i < len / 8; i++) 198 p[i] = 0xff; 199 if (len % 8) 200 p[i] = (0xff00 >> (len % 8)) & 0xff; 201 } 202 203 /* 204 * Generic internet control operations (ioctl's). 205 * 206 * ifp is NULL if not an interface-specific ioctl. 207 */ 208 /* ARGSUSED */ 209 int 210 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, 211 struct thread *td) 212 { 213 register struct ifreq *ifr = (struct ifreq *)data; 214 register struct in_ifaddr *ia, *iap; 215 register struct ifaddr *ifa; 216 struct in_addr allhosts_addr; 217 struct in_addr dst; 218 struct in_ifinfo *ii; 219 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 220 int error, hostIsNew, iaIsNew, maskIsNew; 221 int iaIsFirst; 222 u_long ocmd = cmd; 223 224 /* 225 * Pre-10.x compat: OSIOCAIFADDR passes a shorter 226 * struct in_aliasreq, without ifra_vhid. 227 */ 228 if (cmd == OSIOCAIFADDR) 229 cmd = SIOCAIFADDR; 230 231 ia = NULL; 232 iaIsFirst = 0; 233 iaIsNew = 0; 234 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 235 236 /* 237 * Filter out ioctls we implement directly; forward the rest on to 238 * in_lifaddr_ioctl() and ifp->if_ioctl(). 239 */ 240 switch (cmd) { 241 case SIOCGIFADDR: 242 case SIOCGIFBRDADDR: 243 case SIOCGIFDSTADDR: 244 case SIOCGIFNETMASK: 245 case SIOCDIFADDR: 246 break; 247 case SIOCAIFADDR: 248 /* 249 * ifra_addr must be present and be of INET family. 250 * ifra_broadaddr and ifra_mask are optional. 251 */ 252 if (ifra->ifra_addr.sin_len != sizeof(struct sockaddr_in) || 253 ifra->ifra_addr.sin_family != AF_INET) 254 return (EINVAL); 255 if (ifra->ifra_broadaddr.sin_len != 0 && 256 (ifra->ifra_broadaddr.sin_len != 257 sizeof(struct sockaddr_in) || 258 ifra->ifra_broadaddr.sin_family != AF_INET)) 259 return (EINVAL); 260 #if 0 261 /* 262 * ifconfig(8) in pre-10.x doesn't set sin_family for the 263 * mask. The code is disabled for the 10.x timeline, to 264 * make SIOCAIFADDR compatible with 9.x ifconfig(8). 265 * The code should be enabled in 11.x 266 */ 267 if (ifra->ifra_mask.sin_len != 0 && 268 (ifra->ifra_mask.sin_len != sizeof(struct sockaddr_in) || 269 ifra->ifra_mask.sin_family != AF_INET)) 270 return (EINVAL); 271 #endif 272 break; 273 case SIOCSIFADDR: 274 case SIOCSIFBRDADDR: 275 case SIOCSIFDSTADDR: 276 case SIOCSIFNETMASK: 277 /* We no longer support that old commands. */ 278 return (EINVAL); 279 280 case SIOCALIFADDR: 281 if (td != NULL) { 282 error = priv_check(td, PRIV_NET_ADDIFADDR); 283 if (error) 284 return (error); 285 } 286 if (ifp == NULL) 287 return (EINVAL); 288 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 289 290 case SIOCDLIFADDR: 291 if (td != NULL) { 292 error = priv_check(td, PRIV_NET_DELIFADDR); 293 if (error) 294 return (error); 295 } 296 if (ifp == NULL) 297 return (EINVAL); 298 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 299 300 case SIOCGLIFADDR: 301 if (ifp == NULL) 302 return (EINVAL); 303 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 304 305 default: 306 if (ifp == NULL || ifp->if_ioctl == NULL) 307 return (EOPNOTSUPP); 308 return ((*ifp->if_ioctl)(ifp, cmd, data)); 309 } 310 311 if (ifp == NULL) 312 return (EADDRNOTAVAIL); 313 314 /* 315 * Security checks before we get involved in any work. 316 */ 317 switch (cmd) { 318 case SIOCAIFADDR: 319 if (td != NULL) { 320 error = priv_check(td, PRIV_NET_ADDIFADDR); 321 if (error) 322 return (error); 323 } 324 break; 325 326 case SIOCDIFADDR: 327 if (td != NULL) { 328 error = priv_check(td, PRIV_NET_DELIFADDR); 329 if (error) 330 return (error); 331 } 332 break; 333 } 334 335 /* 336 * Find address for this interface, if it exists. 337 * 338 * If an alias address was specified, find that one instead of the 339 * first one on the interface, if possible. 340 */ 341 dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr; 342 IN_IFADDR_RLOCK(); 343 LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) { 344 if (iap->ia_ifp == ifp && 345 iap->ia_addr.sin_addr.s_addr == dst.s_addr) { 346 if (td == NULL || prison_check_ip4(td->td_ucred, 347 &dst) == 0) 348 ia = iap; 349 break; 350 } 351 } 352 if (ia != NULL) 353 ifa_ref(&ia->ia_ifa); 354 IN_IFADDR_RUNLOCK(); 355 if (ia == NULL) { 356 IF_ADDR_RLOCK(ifp); 357 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 358 iap = ifatoia(ifa); 359 if (iap->ia_addr.sin_family == AF_INET) { 360 if (td != NULL && 361 prison_check_ip4(td->td_ucred, 362 &iap->ia_addr.sin_addr) != 0) 363 continue; 364 ia = iap; 365 break; 366 } 367 } 368 if (ia != NULL) 369 ifa_ref(&ia->ia_ifa); 370 IF_ADDR_RUNLOCK(ifp); 371 } 372 if (ia == NULL) 373 iaIsFirst = 1; 374 375 error = 0; 376 switch (cmd) { 377 case SIOCAIFADDR: 378 case SIOCDIFADDR: 379 if (ifra->ifra_addr.sin_family == AF_INET) { 380 struct in_ifaddr *oia; 381 382 IN_IFADDR_RLOCK(); 383 for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) { 384 if (ia->ia_ifp == ifp && 385 ia->ia_addr.sin_addr.s_addr == 386 ifra->ifra_addr.sin_addr.s_addr) 387 break; 388 } 389 if (ia != NULL && ia != oia) 390 ifa_ref(&ia->ia_ifa); 391 if (oia != NULL && ia != oia) 392 ifa_free(&oia->ia_ifa); 393 IN_IFADDR_RUNLOCK(); 394 if ((ifp->if_flags & IFF_POINTOPOINT) 395 && (cmd == SIOCAIFADDR) 396 && (ifra->ifra_dstaddr.sin_addr.s_addr 397 == INADDR_ANY)) { 398 error = EDESTADDRREQ; 399 goto out; 400 } 401 } 402 if (cmd == SIOCDIFADDR && ia == NULL) { 403 error = EADDRNOTAVAIL; 404 goto out; 405 } 406 if (ia == NULL) { 407 ia = (struct in_ifaddr *) 408 malloc(sizeof *ia, M_IFADDR, M_NOWAIT | 409 M_ZERO); 410 if (ia == NULL) { 411 error = ENOBUFS; 412 goto out; 413 } 414 415 ifa = &ia->ia_ifa; 416 ifa_init(ifa); 417 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 418 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 419 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 420 421 ia->ia_sockmask.sin_len = 8; 422 ia->ia_sockmask.sin_family = AF_INET; 423 if (ifp->if_flags & IFF_BROADCAST) { 424 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr); 425 ia->ia_broadaddr.sin_family = AF_INET; 426 } 427 ia->ia_ifp = ifp; 428 429 ifa_ref(ifa); /* if_addrhead */ 430 IF_ADDR_WLOCK(ifp); 431 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 432 IF_ADDR_WUNLOCK(ifp); 433 ifa_ref(ifa); /* in_ifaddrhead */ 434 IN_IFADDR_WLOCK(); 435 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 436 IN_IFADDR_WUNLOCK(); 437 iaIsNew = 1; 438 } 439 break; 440 441 case SIOCGIFADDR: 442 case SIOCGIFNETMASK: 443 case SIOCGIFDSTADDR: 444 case SIOCGIFBRDADDR: 445 if (ia == NULL) { 446 error = EADDRNOTAVAIL; 447 goto out; 448 } 449 break; 450 } 451 452 /* 453 * Most paths in this switch return directly or via out. Only paths 454 * that remove the address break in order to hit common removal code. 455 */ 456 switch (cmd) { 457 case SIOCGIFADDR: 458 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr; 459 goto out; 460 461 case SIOCGIFBRDADDR: 462 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 463 error = EINVAL; 464 goto out; 465 } 466 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr; 467 goto out; 468 469 case SIOCGIFDSTADDR: 470 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 471 error = EINVAL; 472 goto out; 473 } 474 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr; 475 goto out; 476 477 case SIOCGIFNETMASK: 478 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask; 479 goto out; 480 481 case SIOCAIFADDR: 482 maskIsNew = 0; 483 hostIsNew = 1; 484 error = 0; 485 if (ifra->ifra_addr.sin_addr.s_addr == 486 ia->ia_addr.sin_addr.s_addr) 487 hostIsNew = 0; 488 if (ifra->ifra_mask.sin_len) { 489 /* 490 * QL: XXX 491 * Need to scrub the prefix here in case 492 * the issued command is SIOCAIFADDR with 493 * the same address, but with a different 494 * prefix length. And if the prefix length 495 * is the same as before, then the call is 496 * un-necessarily executed here. 497 */ 498 in_ifscrub(ifp, ia, LLE_STATIC); 499 ia->ia_sockmask = ifra->ifra_mask; 500 ia->ia_sockmask.sin_family = AF_INET; 501 ia->ia_subnetmask = 502 ntohl(ia->ia_sockmask.sin_addr.s_addr); 503 maskIsNew = 1; 504 } 505 if ((ifp->if_flags & IFF_POINTOPOINT) && 506 (ifra->ifra_dstaddr.sin_family == AF_INET)) { 507 in_ifscrub(ifp, ia, LLE_STATIC); 508 ia->ia_dstaddr = ifra->ifra_dstaddr; 509 maskIsNew = 1; /* We lie; but the effect's the same */ 510 } 511 if (hostIsNew || maskIsNew) 512 error = in_ifinit(ifp, ia, &ifra->ifra_addr, maskIsNew, 513 (ocmd == cmd ? ifra->ifra_vhid : 0)); 514 if (error != 0 && iaIsNew) 515 break; 516 517 if ((ifp->if_flags & IFF_BROADCAST) && 518 ifra->ifra_broadaddr.sin_len) 519 ia->ia_broadaddr = ifra->ifra_broadaddr; 520 if (error == 0) { 521 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 522 if (iaIsFirst && 523 (ifp->if_flags & IFF_MULTICAST) != 0) { 524 error = in_joingroup(ifp, &allhosts_addr, 525 NULL, &ii->ii_allhosts); 526 } 527 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 528 } 529 goto out; 530 531 case SIOCDIFADDR: 532 /* 533 * in_ifscrub kills the interface route. 534 */ 535 in_ifscrub(ifp, ia, LLE_STATIC); 536 537 /* 538 * in_ifadown gets rid of all the rest of 539 * the routes. This is not quite the right 540 * thing to do, but at least if we are running 541 * a routing process they will come back. 542 */ 543 in_ifadown(&ia->ia_ifa, 1); 544 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 545 error = 0; 546 break; 547 548 default: 549 panic("in_control: unsupported ioctl"); 550 } 551 552 if (ia->ia_ifa.ifa_carp) 553 (*carp_detach_p)(&ia->ia_ifa); 554 555 IF_ADDR_WLOCK(ifp); 556 /* Re-check that ia is still part of the list. */ 557 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 558 if (ifa == &ia->ia_ifa) 559 break; 560 } 561 if (ifa == NULL) { 562 /* 563 * If we lost the race with another thread, there is no need to 564 * try it again for the next loop as there is no other exit 565 * path between here and out. 566 */ 567 IF_ADDR_WUNLOCK(ifp); 568 error = EADDRNOTAVAIL; 569 goto out; 570 } 571 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 572 IF_ADDR_WUNLOCK(ifp); 573 ifa_free(&ia->ia_ifa); /* if_addrhead */ 574 575 IN_IFADDR_WLOCK(); 576 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 577 578 LIST_REMOVE(ia, ia_hash); 579 IN_IFADDR_WUNLOCK(); 580 /* 581 * If this is the last IPv4 address configured on this 582 * interface, leave the all-hosts group. 583 * No state-change report need be transmitted. 584 */ 585 IFP_TO_IA(ifp, iap); 586 if (iap == NULL) { 587 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 588 IN_MULTI_LOCK(); 589 if (ii->ii_allhosts) { 590 (void)in_leavegroup_locked(ii->ii_allhosts, NULL); 591 ii->ii_allhosts = NULL; 592 } 593 IN_MULTI_UNLOCK(); 594 } else 595 ifa_free(&iap->ia_ifa); 596 597 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 598 out: 599 if (ia != NULL) 600 ifa_free(&ia->ia_ifa); 601 return (error); 602 } 603 604 /* 605 * SIOC[GAD]LIFADDR. 606 * SIOCGLIFADDR: get first address. (?!?) 607 * SIOCGLIFADDR with IFLR_PREFIX: 608 * get first address that matches the specified prefix. 609 * SIOCALIFADDR: add the specified address. 610 * SIOCALIFADDR with IFLR_PREFIX: 611 * EINVAL since we can't deduce hostid part of the address. 612 * SIOCDLIFADDR: delete the specified address. 613 * SIOCDLIFADDR with IFLR_PREFIX: 614 * delete the first address that matches the specified prefix. 615 * return values: 616 * EINVAL on invalid parameters 617 * EADDRNOTAVAIL on prefix match failed/specified address not found 618 * other values may be returned from in_ioctl() 619 */ 620 static int 621 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data, 622 struct ifnet *ifp, struct thread *td) 623 { 624 struct if_laddrreq *iflr = (struct if_laddrreq *)data; 625 struct ifaddr *ifa; 626 627 /* sanity checks */ 628 if (data == NULL || ifp == NULL) { 629 panic("invalid argument to in_lifaddr_ioctl"); 630 /*NOTRECHED*/ 631 } 632 633 switch (cmd) { 634 case SIOCGLIFADDR: 635 /* address must be specified on GET with IFLR_PREFIX */ 636 if ((iflr->flags & IFLR_PREFIX) == 0) 637 break; 638 /*FALLTHROUGH*/ 639 case SIOCALIFADDR: 640 case SIOCDLIFADDR: 641 /* address must be specified on ADD and DELETE */ 642 if (iflr->addr.ss_family != AF_INET) 643 return (EINVAL); 644 if (iflr->addr.ss_len != sizeof(struct sockaddr_in)) 645 return (EINVAL); 646 /* XXX need improvement */ 647 if (iflr->dstaddr.ss_family 648 && iflr->dstaddr.ss_family != AF_INET) 649 return (EINVAL); 650 if (iflr->dstaddr.ss_family 651 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in)) 652 return (EINVAL); 653 break; 654 default: /*shouldn't happen*/ 655 return (EOPNOTSUPP); 656 } 657 if (sizeof(struct in_addr) * 8 < iflr->prefixlen) 658 return (EINVAL); 659 660 switch (cmd) { 661 case SIOCALIFADDR: 662 { 663 struct in_aliasreq ifra; 664 665 if (iflr->flags & IFLR_PREFIX) 666 return (EINVAL); 667 668 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */ 669 bzero(&ifra, sizeof(ifra)); 670 bcopy(iflr->iflr_name, ifra.ifra_name, 671 sizeof(ifra.ifra_name)); 672 673 bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len); 674 675 if (iflr->dstaddr.ss_family) { /*XXX*/ 676 bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr, 677 iflr->dstaddr.ss_len); 678 } 679 680 ifra.ifra_mask.sin_family = AF_INET; 681 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in); 682 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen); 683 684 return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td)); 685 } 686 case SIOCGLIFADDR: 687 case SIOCDLIFADDR: 688 { 689 struct in_ifaddr *ia; 690 struct in_addr mask, candidate, match; 691 struct sockaddr_in *sin; 692 693 bzero(&mask, sizeof(mask)); 694 bzero(&match, sizeof(match)); 695 if (iflr->flags & IFLR_PREFIX) { 696 /* lookup a prefix rather than address. */ 697 in_len2mask(&mask, iflr->prefixlen); 698 699 sin = (struct sockaddr_in *)&iflr->addr; 700 match.s_addr = sin->sin_addr.s_addr; 701 match.s_addr &= mask.s_addr; 702 703 /* if you set extra bits, that's wrong */ 704 if (match.s_addr != sin->sin_addr.s_addr) 705 return (EINVAL); 706 707 } else { 708 /* on getting an address, take the 1st match */ 709 /* on deleting an address, do exact match */ 710 if (cmd != SIOCGLIFADDR) { 711 in_len2mask(&mask, 32); 712 sin = (struct sockaddr_in *)&iflr->addr; 713 match.s_addr = sin->sin_addr.s_addr; 714 } 715 } 716 717 IF_ADDR_RLOCK(ifp); 718 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 719 if (ifa->ifa_addr->sa_family != AF_INET) 720 continue; 721 if (match.s_addr == 0) 722 break; 723 sin = (struct sockaddr_in *)&ifa->ifa_addr; 724 candidate.s_addr = sin->sin_addr.s_addr; 725 candidate.s_addr &= mask.s_addr; 726 if (candidate.s_addr == match.s_addr) 727 break; 728 } 729 if (ifa != NULL) 730 ifa_ref(ifa); 731 IF_ADDR_RUNLOCK(ifp); 732 if (ifa == NULL) 733 return (EADDRNOTAVAIL); 734 ia = (struct in_ifaddr *)ifa; 735 736 if (cmd == SIOCGLIFADDR) { 737 /* fill in the if_laddrreq structure */ 738 bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len); 739 740 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 741 bcopy(&ia->ia_dstaddr, &iflr->dstaddr, 742 ia->ia_dstaddr.sin_len); 743 } else 744 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr)); 745 746 iflr->prefixlen = 747 in_mask2len(&ia->ia_sockmask.sin_addr); 748 749 iflr->flags = 0; /*XXX*/ 750 ifa_free(ifa); 751 752 return (0); 753 } else { 754 struct in_aliasreq ifra; 755 756 /* fill in_aliasreq and do ioctl(SIOCDIFADDR) */ 757 bzero(&ifra, sizeof(ifra)); 758 bcopy(iflr->iflr_name, ifra.ifra_name, 759 sizeof(ifra.ifra_name)); 760 761 bcopy(&ia->ia_addr, &ifra.ifra_addr, 762 ia->ia_addr.sin_len); 763 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 764 bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr, 765 ia->ia_dstaddr.sin_len); 766 } 767 bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr, 768 ia->ia_sockmask.sin_len); 769 ifa_free(ifa); 770 771 return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra, 772 ifp, td)); 773 } 774 } 775 } 776 777 return (EOPNOTSUPP); /*just for safety*/ 778 } 779 780 /* 781 * Delete any existing route for an interface. 782 */ 783 void 784 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags) 785 { 786 787 in_scrubprefix(ia, flags); 788 } 789 790 /* 791 * Initialize an interface's internet address 792 * and routing table entry. 793 */ 794 static int 795 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin, 796 int masksupplied, int vhid) 797 { 798 register u_long i = ntohl(sin->sin_addr.s_addr); 799 int flags, error = 0; 800 801 IN_IFADDR_WLOCK(); 802 if (ia->ia_addr.sin_family == AF_INET) 803 LIST_REMOVE(ia, ia_hash); 804 ia->ia_addr = *sin; 805 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), 806 ia, ia_hash); 807 IN_IFADDR_WUNLOCK(); 808 809 if (vhid > 0) { 810 if (carp_attach_p != NULL) 811 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 812 else 813 error = EPROTONOSUPPORT; 814 } 815 if (error) 816 return (error); 817 818 /* 819 * Give the interface a chance to initialize 820 * if this is its first address, 821 * and to validate the address if necessary. 822 */ 823 if (ifp->if_ioctl != NULL && 824 (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia)) != 0) 825 /* LIST_REMOVE(ia, ia_hash) is done in in_control */ 826 return (error); 827 828 /* 829 * Be compatible with network classes, if netmask isn't supplied, 830 * guess it based on classes. 831 */ 832 if (!masksupplied) { 833 if (IN_CLASSA(i)) 834 ia->ia_subnetmask = IN_CLASSA_NET; 835 else if (IN_CLASSB(i)) 836 ia->ia_subnetmask = IN_CLASSB_NET; 837 else 838 ia->ia_subnetmask = IN_CLASSC_NET; 839 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 840 } 841 ia->ia_subnet = i & ia->ia_subnetmask; 842 in_socktrim(&ia->ia_sockmask); 843 844 /* 845 * Add route for the network. 846 */ 847 flags = RTF_UP; 848 ia->ia_ifa.ifa_metric = ifp->if_metric; 849 if (ifp->if_flags & IFF_BROADCAST) { 850 if (ia->ia_subnetmask == IN_RFC3021_MASK) 851 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 852 else 853 ia->ia_broadaddr.sin_addr.s_addr = 854 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 855 } else if (ifp->if_flags & IFF_LOOPBACK) { 856 ia->ia_dstaddr = ia->ia_addr; 857 flags |= RTF_HOST; 858 } else if (ifp->if_flags & IFF_POINTOPOINT) { 859 if (ia->ia_dstaddr.sin_family != AF_INET) 860 return (0); 861 flags |= RTF_HOST; 862 } 863 if (!vhid && (error = in_addprefix(ia, flags)) != 0) 864 return (error); 865 866 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY) 867 return (0); 868 869 if (ifp->if_flags & IFF_POINTOPOINT && 870 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr) 871 return (0); 872 873 /* 874 * add a loopback route to self 875 */ 876 if (V_useloopback && !vhid && !(ifp->if_flags & IFF_LOOPBACK)) { 877 struct route ia_ro; 878 879 bzero(&ia_ro, sizeof(ia_ro)); 880 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr; 881 rtalloc_ign_fib(&ia_ro, 0, RT_DEFAULT_FIB); 882 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) && 883 (ia_ro.ro_rt->rt_ifp == V_loif)) { 884 RT_LOCK(ia_ro.ro_rt); 885 RT_ADDREF(ia_ro.ro_rt); 886 RTFREE_LOCKED(ia_ro.ro_rt); 887 } else 888 error = ifa_add_loopback_route((struct ifaddr *)ia, 889 (struct sockaddr *)&ia->ia_addr); 890 if (error == 0) 891 ia->ia_flags |= IFA_RTSELF; 892 if (ia_ro.ro_rt != NULL) 893 RTFREE(ia_ro.ro_rt); 894 } 895 896 return (error); 897 } 898 899 #define rtinitflags(x) \ 900 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 901 ? RTF_HOST : 0) 902 903 /* 904 * Generate a routing message when inserting or deleting 905 * an interface address alias. 906 */ 907 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix, 908 struct in_ifaddr *target) 909 { 910 struct route pfx_ro; 911 struct sockaddr_in *pfx_addr; 912 struct rtentry msg_rt; 913 914 /* QL: XXX 915 * This is a bit questionable because there is no 916 * additional route entry added/deleted for an address 917 * alias. Therefore this route report is inaccurate. 918 */ 919 bzero(&pfx_ro, sizeof(pfx_ro)); 920 pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst); 921 pfx_addr->sin_len = sizeof(*pfx_addr); 922 pfx_addr->sin_family = AF_INET; 923 pfx_addr->sin_addr = *prefix; 924 rtalloc_ign_fib(&pfx_ro, 0, 0); 925 if (pfx_ro.ro_rt != NULL) { 926 msg_rt = *pfx_ro.ro_rt; 927 928 /* QL: XXX 929 * Point the gateway to the new interface 930 * address as if a new prefix route entry has 931 * been added through the new address alias. 932 * All other parts of the rtentry is accurate, 933 * e.g., rt_key, rt_mask, rt_ifp etc. 934 */ 935 msg_rt.rt_gateway = (struct sockaddr *)&target->ia_addr; 936 rt_newaddrmsg(cmd, (struct ifaddr *)target, 0, &msg_rt); 937 RTFREE(pfx_ro.ro_rt); 938 } 939 return; 940 } 941 942 /* 943 * Check if we have a route for the given prefix already or add one accordingly. 944 */ 945 int 946 in_addprefix(struct in_ifaddr *target, int flags) 947 { 948 struct in_ifaddr *ia; 949 struct in_addr prefix, mask, p, m; 950 int error; 951 952 if ((flags & RTF_HOST) != 0) { 953 prefix = target->ia_dstaddr.sin_addr; 954 mask.s_addr = 0; 955 } else { 956 prefix = target->ia_addr.sin_addr; 957 mask = target->ia_sockmask.sin_addr; 958 prefix.s_addr &= mask.s_addr; 959 } 960 961 IN_IFADDR_RLOCK(); 962 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 963 if (rtinitflags(ia)) { 964 p = ia->ia_dstaddr.sin_addr; 965 966 if (prefix.s_addr != p.s_addr) 967 continue; 968 } else { 969 p = ia->ia_addr.sin_addr; 970 m = ia->ia_sockmask.sin_addr; 971 p.s_addr &= m.s_addr; 972 973 if (prefix.s_addr != p.s_addr || 974 mask.s_addr != m.s_addr) 975 continue; 976 } 977 978 /* 979 * If we got a matching prefix route inserted by other 980 * interface address, we are done here. 981 */ 982 if (ia->ia_flags & IFA_ROUTE) { 983 #ifdef RADIX_MPATH 984 if (ia->ia_addr.sin_addr.s_addr == 985 target->ia_addr.sin_addr.s_addr) { 986 IN_IFADDR_RUNLOCK(); 987 return (EEXIST); 988 } else 989 break; 990 #endif 991 if (V_nosameprefix) { 992 IN_IFADDR_RUNLOCK(); 993 return (EEXIST); 994 } else { 995 in_addralias_rtmsg(RTM_ADD, &prefix, target); 996 IN_IFADDR_RUNLOCK(); 997 return (0); 998 } 999 } 1000 } 1001 IN_IFADDR_RUNLOCK(); 1002 1003 /* 1004 * No-one seem to have this prefix route, so we try to insert it. 1005 */ 1006 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 1007 if (!error) 1008 target->ia_flags |= IFA_ROUTE; 1009 return (error); 1010 } 1011 1012 /* 1013 * If there is no other address in the system that can serve a route to the 1014 * same prefix, remove the route. Hand over the route to the new address 1015 * otherwise. 1016 */ 1017 int 1018 in_scrubprefix(struct in_ifaddr *target, u_int flags) 1019 { 1020 struct in_ifaddr *ia; 1021 struct in_addr prefix, mask, p, m; 1022 int error = 0; 1023 struct sockaddr_in prefix0, mask0; 1024 1025 /* 1026 * Remove the loopback route to the interface address. 1027 * The "useloopback" setting is not consulted because if the 1028 * user configures an interface address, turns off this 1029 * setting, and then tries to delete that interface address, 1030 * checking the current setting of "useloopback" would leave 1031 * that interface address loopback route untouched, which 1032 * would be wrong. Therefore the interface address loopback route 1033 * deletion is unconditional. 1034 */ 1035 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 1036 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 1037 (target->ia_flags & IFA_RTSELF)) { 1038 struct route ia_ro; 1039 int freeit = 0; 1040 1041 bzero(&ia_ro, sizeof(ia_ro)); 1042 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr; 1043 rtalloc_ign_fib(&ia_ro, 0, 0); 1044 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) && 1045 (ia_ro.ro_rt->rt_ifp == V_loif)) { 1046 RT_LOCK(ia_ro.ro_rt); 1047 if (ia_ro.ro_rt->rt_refcnt <= 1) 1048 freeit = 1; 1049 else if (flags & LLE_STATIC) { 1050 RT_REMREF(ia_ro.ro_rt); 1051 target->ia_flags &= ~IFA_RTSELF; 1052 } 1053 RTFREE_LOCKED(ia_ro.ro_rt); 1054 } 1055 if (freeit && (flags & LLE_STATIC)) { 1056 error = ifa_del_loopback_route((struct ifaddr *)target, 1057 (struct sockaddr *)&target->ia_addr); 1058 if (error == 0) 1059 target->ia_flags &= ~IFA_RTSELF; 1060 } 1061 if ((flags & LLE_STATIC) && 1062 !(target->ia_ifp->if_flags & IFF_NOARP)) 1063 /* remove arp cache */ 1064 arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr); 1065 } 1066 1067 if (rtinitflags(target)) { 1068 prefix = target->ia_dstaddr.sin_addr; 1069 mask.s_addr = 0; 1070 } else { 1071 prefix = target->ia_addr.sin_addr; 1072 mask = target->ia_sockmask.sin_addr; 1073 prefix.s_addr &= mask.s_addr; 1074 } 1075 1076 if ((target->ia_flags & IFA_ROUTE) == 0) { 1077 in_addralias_rtmsg(RTM_DELETE, &prefix, target); 1078 return (0); 1079 } 1080 1081 IN_IFADDR_RLOCK(); 1082 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1083 if (rtinitflags(ia)) { 1084 p = ia->ia_dstaddr.sin_addr; 1085 1086 if (prefix.s_addr != p.s_addr) 1087 continue; 1088 } else { 1089 p = ia->ia_addr.sin_addr; 1090 m = ia->ia_sockmask.sin_addr; 1091 p.s_addr &= m.s_addr; 1092 1093 if (prefix.s_addr != p.s_addr || 1094 mask.s_addr != m.s_addr) 1095 continue; 1096 } 1097 1098 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 1099 continue; 1100 1101 /* 1102 * If we got a matching prefix address, move IFA_ROUTE and 1103 * the route itself to it. Make sure that routing daemons 1104 * get a heads-up. 1105 */ 1106 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1107 ifa_ref(&ia->ia_ifa); 1108 IN_IFADDR_RUNLOCK(); 1109 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 1110 rtinitflags(target)); 1111 if (error == 0) 1112 target->ia_flags &= ~IFA_ROUTE; 1113 else 1114 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 1115 error); 1116 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 1117 rtinitflags(ia) | RTF_UP); 1118 if (error == 0) 1119 ia->ia_flags |= IFA_ROUTE; 1120 else 1121 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 1122 error); 1123 ifa_free(&ia->ia_ifa); 1124 return (error); 1125 } 1126 } 1127 IN_IFADDR_RUNLOCK(); 1128 1129 /* 1130 * remove all L2 entries on the given prefix 1131 */ 1132 bzero(&prefix0, sizeof(prefix0)); 1133 prefix0.sin_len = sizeof(prefix0); 1134 prefix0.sin_family = AF_INET; 1135 prefix0.sin_addr.s_addr = target->ia_subnet; 1136 bzero(&mask0, sizeof(mask0)); 1137 mask0.sin_len = sizeof(mask0); 1138 mask0.sin_family = AF_INET; 1139 mask0.sin_addr.s_addr = target->ia_subnetmask; 1140 lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0, 1141 (struct sockaddr *)&mask0, flags); 1142 1143 /* 1144 * As no-one seem to have this prefix, we can remove the route. 1145 */ 1146 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 1147 if (error == 0) 1148 target->ia_flags &= ~IFA_ROUTE; 1149 else 1150 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 1151 return (error); 1152 } 1153 1154 #undef rtinitflags 1155 1156 /* 1157 * Return 1 if the address might be a local broadcast address. 1158 */ 1159 int 1160 in_broadcast(struct in_addr in, struct ifnet *ifp) 1161 { 1162 register struct ifaddr *ifa; 1163 u_long t; 1164 1165 if (in.s_addr == INADDR_BROADCAST || 1166 in.s_addr == INADDR_ANY) 1167 return (1); 1168 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1169 return (0); 1170 t = ntohl(in.s_addr); 1171 /* 1172 * Look through the list of addresses for a match 1173 * with a broadcast address. 1174 */ 1175 #define ia ((struct in_ifaddr *)ifa) 1176 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1177 if (ifa->ifa_addr->sa_family == AF_INET && 1178 (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 1179 /* 1180 * Check for old-style (host 0) broadcast, but 1181 * taking into account that RFC 3021 obsoletes it. 1182 */ 1183 (ia->ia_subnetmask != IN_RFC3021_MASK && 1184 t == ia->ia_subnet)) && 1185 /* 1186 * Check for an all one subnetmask. These 1187 * only exist when an interface gets a secondary 1188 * address. 1189 */ 1190 ia->ia_subnetmask != (u_long)0xffffffff) 1191 return (1); 1192 return (0); 1193 #undef ia 1194 } 1195 1196 /* 1197 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 1198 */ 1199 void 1200 in_ifdetach(struct ifnet *ifp) 1201 { 1202 1203 in_pcbpurgeif0(&V_ripcbinfo, ifp); 1204 in_pcbpurgeif0(&V_udbinfo, ifp); 1205 in_purgemaddrs(ifp); 1206 } 1207 1208 /* 1209 * Delete all IPv4 multicast address records, and associated link-layer 1210 * multicast address records, associated with ifp. 1211 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1212 * XXX This should not race with ifma_protospec being set during 1213 * a new allocation, if it does, we have bigger problems. 1214 */ 1215 static void 1216 in_purgemaddrs(struct ifnet *ifp) 1217 { 1218 LIST_HEAD(,in_multi) purgeinms; 1219 struct in_multi *inm, *tinm; 1220 struct ifmultiaddr *ifma; 1221 1222 LIST_INIT(&purgeinms); 1223 IN_MULTI_LOCK(); 1224 1225 /* 1226 * Extract list of in_multi associated with the detaching ifp 1227 * which the PF_INET layer is about to release. 1228 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1229 * by code further down. 1230 */ 1231 IF_ADDR_RLOCK(ifp); 1232 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1233 if (ifma->ifma_addr->sa_family != AF_INET || 1234 ifma->ifma_protospec == NULL) 1235 continue; 1236 #if 0 1237 KASSERT(ifma->ifma_protospec != NULL, 1238 ("%s: ifma_protospec is NULL", __func__)); 1239 #endif 1240 inm = (struct in_multi *)ifma->ifma_protospec; 1241 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 1242 } 1243 IF_ADDR_RUNLOCK(ifp); 1244 1245 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 1246 LIST_REMOVE(inm, inm_link); 1247 inm_release_locked(inm); 1248 } 1249 igmp_ifdetach(ifp); 1250 1251 IN_MULTI_UNLOCK(); 1252 } 1253 1254 struct in_llentry { 1255 struct llentry base; 1256 struct sockaddr_in l3_addr4; 1257 }; 1258 1259 /* 1260 * Deletes an address from the address table. 1261 * This function is called by the timer functions 1262 * such as arptimer() and nd6_llinfo_timer(), and 1263 * the caller does the locking. 1264 */ 1265 static void 1266 in_lltable_free(struct lltable *llt, struct llentry *lle) 1267 { 1268 LLE_WUNLOCK(lle); 1269 LLE_LOCK_DESTROY(lle); 1270 free(lle, M_LLTABLE); 1271 } 1272 1273 static struct llentry * 1274 in_lltable_new(const struct sockaddr *l3addr, u_int flags) 1275 { 1276 struct in_llentry *lle; 1277 1278 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1279 if (lle == NULL) /* NB: caller generates msg */ 1280 return NULL; 1281 1282 /* 1283 * For IPv4 this will trigger "arpresolve" to generate 1284 * an ARP request. 1285 */ 1286 lle->base.la_expire = time_uptime; /* mark expired */ 1287 lle->l3_addr4 = *(const struct sockaddr_in *)l3addr; 1288 lle->base.lle_refcnt = 1; 1289 lle->base.lle_free = in_lltable_free; 1290 LLE_LOCK_INIT(&lle->base); 1291 callout_init_rw(&lle->base.la_timer, &lle->base.lle_lock, 1292 CALLOUT_RETURNUNLOCKED); 1293 1294 return (&lle->base); 1295 } 1296 1297 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1298 (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 1299 1300 static void 1301 in_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix, 1302 const struct sockaddr *mask, u_int flags) 1303 { 1304 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 1305 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 1306 struct llentry *lle, *next; 1307 int i; 1308 size_t pkts_dropped; 1309 1310 IF_AFDATA_WLOCK(llt->llt_ifp); 1311 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1312 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 1313 /* 1314 * (flags & LLE_STATIC) means deleting all entries 1315 * including static ARP entries. 1316 */ 1317 if (IN_ARE_MASKED_ADDR_EQUAL(satosin(L3_ADDR(lle)), 1318 pfx, msk) && ((flags & LLE_STATIC) || 1319 !(lle->la_flags & LLE_STATIC))) { 1320 LLE_WLOCK(lle); 1321 if (callout_stop(&lle->la_timer)) 1322 LLE_REMREF(lle); 1323 pkts_dropped = llentry_free(lle); 1324 ARPSTAT_ADD(dropped, pkts_dropped); 1325 } 1326 } 1327 } 1328 IF_AFDATA_WUNLOCK(llt->llt_ifp); 1329 } 1330 1331 1332 static int 1333 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1334 { 1335 struct rtentry *rt; 1336 1337 KASSERT(l3addr->sa_family == AF_INET, 1338 ("sin_family %d", l3addr->sa_family)); 1339 1340 /* XXX rtalloc1 should take a const param */ 1341 rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0); 1342 1343 if (rt == NULL) 1344 return (EINVAL); 1345 1346 /* 1347 * If the gateway for an existing host route matches the target L3 1348 * address, which is a special route inserted by some implementation 1349 * such as MANET, and the interface is of the correct type, then 1350 * allow for ARP to proceed. 1351 */ 1352 if (rt->rt_flags & RTF_GATEWAY) { 1353 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 1354 rt->rt_ifp->if_type != IFT_ETHER || 1355 (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1356 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 1357 sizeof(in_addr_t)) != 0) { 1358 RTFREE_LOCKED(rt); 1359 return (EINVAL); 1360 } 1361 } 1362 1363 /* 1364 * Make sure that at least the destination address is covered 1365 * by the route. This is for handling the case where 2 or more 1366 * interfaces have the same prefix. An incoming packet arrives 1367 * on one interface and the corresponding outgoing packet leaves 1368 * another interface. 1369 */ 1370 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 1371 const char *sa, *mask, *addr, *lim; 1372 int len; 1373 1374 mask = (const char *)rt_mask(rt); 1375 /* 1376 * Just being extra cautious to avoid some custom 1377 * code getting into trouble. 1378 */ 1379 if (mask == NULL) { 1380 RTFREE_LOCKED(rt); 1381 return (EINVAL); 1382 } 1383 1384 sa = (const char *)rt_key(rt); 1385 addr = (const char *)l3addr; 1386 len = ((const struct sockaddr_in *)l3addr)->sin_len; 1387 lim = addr + len; 1388 1389 for ( ; addr < lim; sa++, mask++, addr++) { 1390 if ((*sa ^ *addr) & *mask) { 1391 #ifdef DIAGNOSTIC 1392 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1393 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 1394 #endif 1395 RTFREE_LOCKED(rt); 1396 return (EINVAL); 1397 } 1398 } 1399 } 1400 1401 RTFREE_LOCKED(rt); 1402 return (0); 1403 } 1404 1405 /* 1406 * Return NULL if not found or marked for deletion. 1407 * If found return lle read locked. 1408 */ 1409 static struct llentry * 1410 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1411 { 1412 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1413 struct ifnet *ifp = llt->llt_ifp; 1414 struct llentry *lle; 1415 struct llentries *lleh; 1416 u_int hashkey; 1417 1418 IF_AFDATA_LOCK_ASSERT(ifp); 1419 KASSERT(l3addr->sa_family == AF_INET, 1420 ("sin_family %d", l3addr->sa_family)); 1421 1422 hashkey = sin->sin_addr.s_addr; 1423 lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; 1424 LIST_FOREACH(lle, lleh, lle_next) { 1425 struct sockaddr_in *sa2 = satosin(L3_ADDR(lle)); 1426 if (lle->la_flags & LLE_DELETED) 1427 continue; 1428 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr) 1429 break; 1430 } 1431 if (lle == NULL) { 1432 #ifdef DIAGNOSTIC 1433 if (flags & LLE_DELETE) 1434 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); 1435 #endif 1436 if (!(flags & LLE_CREATE)) 1437 return (NULL); 1438 /* 1439 * A route that covers the given address must have 1440 * been installed 1st because we are doing a resolution, 1441 * verify this. 1442 */ 1443 if (!(flags & LLE_IFADDR) && 1444 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1445 goto done; 1446 1447 lle = in_lltable_new(l3addr, flags); 1448 if (lle == NULL) { 1449 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1450 goto done; 1451 } 1452 lle->la_flags = flags & ~LLE_CREATE; 1453 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { 1454 bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); 1455 lle->la_flags |= (LLE_VALID | LLE_STATIC); 1456 } 1457 1458 lle->lle_tbl = llt; 1459 lle->lle_head = lleh; 1460 lle->la_flags |= LLE_LINKED; 1461 LIST_INSERT_HEAD(lleh, lle, lle_next); 1462 } else if (flags & LLE_DELETE) { 1463 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) { 1464 LLE_WLOCK(lle); 1465 lle->la_flags |= LLE_DELETED; 1466 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1467 #ifdef DIAGNOSTIC 1468 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1469 #endif 1470 if ((lle->la_flags & 1471 (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC) 1472 llentry_free(lle); 1473 else 1474 LLE_WUNLOCK(lle); 1475 } 1476 lle = (void *)-1; 1477 1478 } 1479 if (LLE_IS_VALID(lle)) { 1480 if (flags & LLE_EXCLUSIVE) 1481 LLE_WLOCK(lle); 1482 else 1483 LLE_RLOCK(lle); 1484 } 1485 done: 1486 return (lle); 1487 } 1488 1489 static int 1490 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr) 1491 { 1492 #define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle)) 1493 struct ifnet *ifp = llt->llt_ifp; 1494 struct llentry *lle; 1495 /* XXX stack use */ 1496 struct { 1497 struct rt_msghdr rtm; 1498 struct sockaddr_in sin; 1499 struct sockaddr_dl sdl; 1500 } arpc; 1501 int error, i; 1502 1503 LLTABLE_LOCK_ASSERT(); 1504 1505 error = 0; 1506 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1507 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1508 struct sockaddr_dl *sdl; 1509 1510 /* skip deleted entries */ 1511 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1512 continue; 1513 /* Skip if jailed and not a valid IP of the prison. */ 1514 if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0) 1515 continue; 1516 /* 1517 * produce a msg made of: 1518 * struct rt_msghdr; 1519 * struct sockaddr_in; (IPv4) 1520 * struct sockaddr_dl; 1521 */ 1522 bzero(&arpc, sizeof(arpc)); 1523 arpc.rtm.rtm_msglen = sizeof(arpc); 1524 arpc.rtm.rtm_version = RTM_VERSION; 1525 arpc.rtm.rtm_type = RTM_GET; 1526 arpc.rtm.rtm_flags = RTF_UP; 1527 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1528 arpc.sin.sin_family = AF_INET; 1529 arpc.sin.sin_len = sizeof(arpc.sin); 1530 arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr; 1531 1532 /* publish */ 1533 if (lle->la_flags & LLE_PUB) 1534 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1535 1536 sdl = &arpc.sdl; 1537 sdl->sdl_family = AF_LINK; 1538 sdl->sdl_len = sizeof(*sdl); 1539 sdl->sdl_index = ifp->if_index; 1540 sdl->sdl_type = ifp->if_type; 1541 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1542 sdl->sdl_alen = ifp->if_addrlen; 1543 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1544 } else { 1545 sdl->sdl_alen = 0; 1546 bzero(LLADDR(sdl), ifp->if_addrlen); 1547 } 1548 1549 arpc.rtm.rtm_rmx.rmx_expire = 1550 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1551 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1552 if (lle->la_flags & LLE_STATIC) 1553 arpc.rtm.rtm_flags |= RTF_STATIC; 1554 arpc.rtm.rtm_index = ifp->if_index; 1555 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1556 if (error) 1557 break; 1558 } 1559 } 1560 return error; 1561 #undef SIN 1562 } 1563 1564 void * 1565 in_domifattach(struct ifnet *ifp) 1566 { 1567 struct in_ifinfo *ii; 1568 struct lltable *llt; 1569 1570 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1571 1572 llt = lltable_init(ifp, AF_INET); 1573 if (llt != NULL) { 1574 llt->llt_prefix_free = in_lltable_prefix_free; 1575 llt->llt_lookup = in_lltable_lookup; 1576 llt->llt_dump = in_lltable_dump; 1577 } 1578 ii->ii_llt = llt; 1579 1580 ii->ii_igmp = igmp_domifattach(ifp); 1581 1582 return ii; 1583 } 1584 1585 void 1586 in_domifdetach(struct ifnet *ifp, void *aux) 1587 { 1588 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1589 1590 igmp_domifdetach(ifp); 1591 lltable_free(ii->ii_llt); 1592 free(ii, M_IFADDR); 1593 } 1594