in.c revision 229621
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: head/sys/netinet/in.c 229621 2012-01-05 19:00:36Z jhb $"); 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 69static int in_mask2len(struct in_addr *); 70static void in_len2mask(struct in_addr *, int); 71static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t, 72 struct ifnet *, struct thread *); 73 74static void in_socktrim(struct sockaddr_in *); 75static int in_ifinit(struct ifnet *, struct in_ifaddr *, 76 struct sockaddr_in *, int, int, int); 77static void in_purgemaddrs(struct ifnet *); 78 79static VNET_DEFINE(int, nosameprefix); 80#define V_nosameprefix VNET(nosameprefix) 81SYSCTL_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 85VNET_DECLARE(struct inpcbinfo, ripcbinfo); 86#define V_ripcbinfo VNET(ripcbinfo) 87 88VNET_DECLARE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */ 89#define V_arpstat VNET(arpstat) 90 91/* 92 * Return 1 if an internet address is for a ``local'' host 93 * (one to which we have a connection). 94 */ 95int 96in_localaddr(struct in_addr in) 97{ 98 register u_long i = ntohl(in.s_addr); 99 register struct in_ifaddr *ia; 100 101 IN_IFADDR_RLOCK(); 102 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 103 if ((i & ia->ia_subnetmask) == ia->ia_subnet) { 104 IN_IFADDR_RUNLOCK(); 105 return (1); 106 } 107 } 108 IN_IFADDR_RUNLOCK(); 109 return (0); 110} 111 112/* 113 * Return 1 if an internet address is for the local host and configured 114 * on one of its interfaces. 115 */ 116int 117in_localip(struct in_addr in) 118{ 119 struct in_ifaddr *ia; 120 121 IN_IFADDR_RLOCK(); 122 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) { 123 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) { 124 IN_IFADDR_RUNLOCK(); 125 return (1); 126 } 127 } 128 IN_IFADDR_RUNLOCK(); 129 return (0); 130} 131 132/* 133 * Determine whether an IP address is in a reserved set of addresses 134 * that may not be forwarded, or whether datagrams to that destination 135 * may be forwarded. 136 */ 137int 138in_canforward(struct in_addr in) 139{ 140 register u_long i = ntohl(in.s_addr); 141 register u_long net; 142 143 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i)) 144 return (0); 145 if (IN_CLASSA(i)) { 146 net = i & IN_CLASSA_NET; 147 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 148 return (0); 149 } 150 return (1); 151} 152 153/* 154 * Trim a mask in a sockaddr 155 */ 156static void 157in_socktrim(struct sockaddr_in *ap) 158{ 159 register char *cplim = (char *) &ap->sin_addr; 160 register char *cp = (char *) (&ap->sin_addr + 1); 161 162 ap->sin_len = 0; 163 while (--cp >= cplim) 164 if (*cp) { 165 (ap)->sin_len = cp - (char *) (ap) + 1; 166 break; 167 } 168} 169 170static int 171in_mask2len(mask) 172 struct in_addr *mask; 173{ 174 int x, y; 175 u_char *p; 176 177 p = (u_char *)mask; 178 for (x = 0; x < sizeof(*mask); x++) { 179 if (p[x] != 0xff) 180 break; 181 } 182 y = 0; 183 if (x < sizeof(*mask)) { 184 for (y = 0; y < 8; y++) { 185 if ((p[x] & (0x80 >> y)) == 0) 186 break; 187 } 188 } 189 return (x * 8 + y); 190} 191 192static void 193in_len2mask(struct in_addr *mask, int len) 194{ 195 int i; 196 u_char *p; 197 198 p = (u_char *)mask; 199 bzero(mask, sizeof(*mask)); 200 for (i = 0; i < len / 8; i++) 201 p[i] = 0xff; 202 if (len % 8) 203 p[i] = (0xff00 >> (len % 8)) & 0xff; 204} 205 206/* 207 * Generic internet control operations (ioctl's). 208 * 209 * ifp is NULL if not an interface-specific ioctl. 210 */ 211/* ARGSUSED */ 212int 213in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, 214 struct thread *td) 215{ 216 register struct ifreq *ifr = (struct ifreq *)data; 217 register struct in_ifaddr *ia, *iap; 218 register struct ifaddr *ifa; 219 struct in_addr allhosts_addr; 220 struct in_addr dst; 221 struct in_ifinfo *ii; 222 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 223 struct sockaddr_in oldaddr; 224 int error, hostIsNew, iaIsNew, maskIsNew; 225 int iaIsFirst; 226 u_long ocmd = cmd; 227 228 /* 229 * Pre-10.x compat: OSIOCAIFADDR passes a shorter 230 * struct in_aliasreq, without ifra_vhid. 231 */ 232 if (cmd == OSIOCAIFADDR) 233 cmd = SIOCAIFADDR; 234 235 ia = NULL; 236 iaIsFirst = 0; 237 iaIsNew = 0; 238 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 239 240 /* 241 * Filter out ioctls we implement directly; forward the rest on to 242 * in_lifaddr_ioctl() and ifp->if_ioctl(). 243 */ 244 switch (cmd) { 245 case SIOCGIFADDR: 246 case SIOCGIFBRDADDR: 247 case SIOCGIFDSTADDR: 248 case SIOCGIFNETMASK: 249 case SIOCDIFADDR: 250 break; 251 case SIOCAIFADDR: 252 /* 253 * ifra_addr must be present and be of INET family. 254 * ifra_broadaddr and ifra_mask are optional. 255 */ 256 if (ifra->ifra_addr.sin_len != sizeof(struct sockaddr_in) || 257 ifra->ifra_addr.sin_family != AF_INET) 258 return (EINVAL); 259 if (ifra->ifra_broadaddr.sin_len != 0 && 260 (ifra->ifra_broadaddr.sin_len != 261 sizeof(struct sockaddr_in) || 262 ifra->ifra_broadaddr.sin_family != AF_INET)) 263 return (EINVAL); 264#if 0 265 /* 266 * ifconfig(8) in pre-10.x doesn't set sin_family for the 267 * mask. The code is disabled for the 10.x timeline, to 268 * make SIOCAIFADDR compatible with 9.x ifconfig(8). 269 * The code should be enabled in 11.x 270 */ 271 if (ifra->ifra_mask.sin_len != 0 && 272 (ifra->ifra_mask.sin_len != sizeof(struct sockaddr_in) || 273 ifra->ifra_mask.sin_family != AF_INET)) 274 return (EINVAL); 275#endif 276 break; 277 case SIOCSIFADDR: 278 case SIOCSIFBRDADDR: 279 case SIOCSIFDSTADDR: 280 case SIOCSIFNETMASK: 281 if (ifr->ifr_addr.sa_family != AF_INET || 282 ifr->ifr_addr.sa_len != sizeof(struct sockaddr_in)) 283 return (EINVAL); 284 break; 285 286 case SIOCALIFADDR: 287 if (td != NULL) { 288 error = priv_check(td, PRIV_NET_ADDIFADDR); 289 if (error) 290 return (error); 291 } 292 if (ifp == NULL) 293 return (EINVAL); 294 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 295 296 case SIOCDLIFADDR: 297 if (td != NULL) { 298 error = priv_check(td, PRIV_NET_DELIFADDR); 299 if (error) 300 return (error); 301 } 302 if (ifp == NULL) 303 return (EINVAL); 304 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 305 306 case SIOCGLIFADDR: 307 if (ifp == NULL) 308 return (EINVAL); 309 return in_lifaddr_ioctl(so, cmd, data, ifp, td); 310 311 default: 312 if (ifp == NULL || ifp->if_ioctl == NULL) 313 return (EOPNOTSUPP); 314 return ((*ifp->if_ioctl)(ifp, cmd, data)); 315 } 316 317 if (ifp == NULL) 318 return (EADDRNOTAVAIL); 319 320 /* 321 * Security checks before we get involved in any work. 322 */ 323 switch (cmd) { 324 case SIOCAIFADDR: 325 case SIOCSIFADDR: 326 case SIOCSIFBRDADDR: 327 case SIOCSIFNETMASK: 328 case SIOCSIFDSTADDR: 329 if (td != NULL) { 330 error = priv_check(td, PRIV_NET_ADDIFADDR); 331 if (error) 332 return (error); 333 } 334 break; 335 336 case SIOCDIFADDR: 337 if (td != NULL) { 338 error = priv_check(td, PRIV_NET_DELIFADDR); 339 if (error) 340 return (error); 341 } 342 break; 343 } 344 345 /* 346 * Find address for this interface, if it exists. 347 * 348 * If an alias address was specified, find that one instead of the 349 * first one on the interface, if possible. 350 */ 351 dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr; 352 IN_IFADDR_RLOCK(); 353 LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) { 354 if (iap->ia_ifp == ifp && 355 iap->ia_addr.sin_addr.s_addr == dst.s_addr) { 356 if (td == NULL || prison_check_ip4(td->td_ucred, 357 &dst) == 0) 358 ia = iap; 359 break; 360 } 361 } 362 if (ia != NULL) 363 ifa_ref(&ia->ia_ifa); 364 IN_IFADDR_RUNLOCK(); 365 if (ia == NULL) { 366 IF_ADDR_RLOCK(ifp); 367 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 368 iap = ifatoia(ifa); 369 if (iap->ia_addr.sin_family == AF_INET) { 370 if (td != NULL && 371 prison_check_ip4(td->td_ucred, 372 &iap->ia_addr.sin_addr) != 0) 373 continue; 374 ia = iap; 375 break; 376 } 377 } 378 if (ia != NULL) 379 ifa_ref(&ia->ia_ifa); 380 IF_ADDR_RUNLOCK(ifp); 381 } 382 if (ia == NULL) 383 iaIsFirst = 1; 384 385 error = 0; 386 switch (cmd) { 387 case SIOCAIFADDR: 388 case SIOCDIFADDR: 389 if (ifra->ifra_addr.sin_family == AF_INET) { 390 struct in_ifaddr *oia; 391 392 IN_IFADDR_RLOCK(); 393 for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) { 394 if (ia->ia_ifp == ifp && 395 ia->ia_addr.sin_addr.s_addr == 396 ifra->ifra_addr.sin_addr.s_addr) 397 break; 398 } 399 if (ia != NULL && ia != oia) 400 ifa_ref(&ia->ia_ifa); 401 if (oia != NULL && ia != oia) 402 ifa_free(&oia->ia_ifa); 403 IN_IFADDR_RUNLOCK(); 404 if ((ifp->if_flags & IFF_POINTOPOINT) 405 && (cmd == SIOCAIFADDR) 406 && (ifra->ifra_dstaddr.sin_addr.s_addr 407 == INADDR_ANY)) { 408 error = EDESTADDRREQ; 409 goto out; 410 } 411 } 412 if (cmd == SIOCDIFADDR && ia == NULL) { 413 error = EADDRNOTAVAIL; 414 goto out; 415 } 416 /* FALLTHROUGH */ 417 case SIOCSIFADDR: 418 case SIOCSIFNETMASK: 419 case SIOCSIFDSTADDR: 420 if (ia == NULL) { 421 ia = (struct in_ifaddr *) 422 malloc(sizeof *ia, M_IFADDR, M_NOWAIT | 423 M_ZERO); 424 if (ia == NULL) { 425 error = ENOBUFS; 426 goto out; 427 } 428 429 ifa = &ia->ia_ifa; 430 ifa_init(ifa); 431 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 432 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 433 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 434 435 ia->ia_sockmask.sin_len = 8; 436 ia->ia_sockmask.sin_family = AF_INET; 437 if (ifp->if_flags & IFF_BROADCAST) { 438 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr); 439 ia->ia_broadaddr.sin_family = AF_INET; 440 } 441 ia->ia_ifp = ifp; 442 443 ifa_ref(ifa); /* if_addrhead */ 444 IF_ADDR_WLOCK(ifp); 445 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 446 IF_ADDR_WUNLOCK(ifp); 447 ifa_ref(ifa); /* in_ifaddrhead */ 448 IN_IFADDR_WLOCK(); 449 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 450 IN_IFADDR_WUNLOCK(); 451 iaIsNew = 1; 452 } 453 break; 454 455 case SIOCSIFBRDADDR: 456 case SIOCGIFADDR: 457 case SIOCGIFNETMASK: 458 case SIOCGIFDSTADDR: 459 case SIOCGIFBRDADDR: 460 if (ia == NULL) { 461 error = EADDRNOTAVAIL; 462 goto out; 463 } 464 break; 465 } 466 467 /* 468 * Most paths in this switch return directly or via out. Only paths 469 * that remove the address break in order to hit common removal code. 470 */ 471 switch (cmd) { 472 case SIOCGIFADDR: 473 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr; 474 goto out; 475 476 case SIOCGIFBRDADDR: 477 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 478 error = EINVAL; 479 goto out; 480 } 481 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr; 482 goto out; 483 484 case SIOCGIFDSTADDR: 485 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 486 error = EINVAL; 487 goto out; 488 } 489 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr; 490 goto out; 491 492 case SIOCGIFNETMASK: 493 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask; 494 goto out; 495 496 case SIOCSIFDSTADDR: 497 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 498 error = EINVAL; 499 goto out; 500 } 501 oldaddr = ia->ia_dstaddr; 502 ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr; 503 if (ifp->if_ioctl != NULL) { 504 error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, 505 (caddr_t)ia); 506 if (error) { 507 ia->ia_dstaddr = oldaddr; 508 goto out; 509 } 510 } 511 if (ia->ia_flags & IFA_ROUTE) { 512 ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr; 513 rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST); 514 ia->ia_ifa.ifa_dstaddr = 515 (struct sockaddr *)&ia->ia_dstaddr; 516 rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP); 517 } 518 goto out; 519 520 case SIOCSIFBRDADDR: 521 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 522 error = EINVAL; 523 goto out; 524 } 525 ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr; 526 goto out; 527 528 case SIOCSIFADDR: 529 error = in_ifinit(ifp, ia, 530 (struct sockaddr_in *) &ifr->ifr_addr, 1, 0, 0); 531 if (error != 0 && iaIsNew) 532 break; 533 if (error == 0) { 534 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 535 if (iaIsFirst && 536 (ifp->if_flags & IFF_MULTICAST) != 0) { 537 error = in_joingroup(ifp, &allhosts_addr, 538 NULL, &ii->ii_allhosts); 539 } 540 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 541 } 542 error = 0; 543 goto out; 544 545 case SIOCSIFNETMASK: 546 ia->ia_sockmask.sin_addr = ((struct sockaddr_in *) 547 &ifr->ifr_addr)->sin_addr; 548 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 549 goto out; 550 551 case SIOCAIFADDR: 552 maskIsNew = 0; 553 hostIsNew = 1; 554 error = 0; 555 if (ifra->ifra_addr.sin_addr.s_addr == 556 ia->ia_addr.sin_addr.s_addr) 557 hostIsNew = 0; 558 if (ifra->ifra_mask.sin_len) { 559 /* 560 * QL: XXX 561 * Need to scrub the prefix here in case 562 * the issued command is SIOCAIFADDR with 563 * the same address, but with a different 564 * prefix length. And if the prefix length 565 * is the same as before, then the call is 566 * un-necessarily executed here. 567 */ 568 in_ifscrub(ifp, ia, LLE_STATIC); 569 ia->ia_sockmask = ifra->ifra_mask; 570 ia->ia_sockmask.sin_family = AF_INET; 571 ia->ia_subnetmask = 572 ntohl(ia->ia_sockmask.sin_addr.s_addr); 573 maskIsNew = 1; 574 } 575 if ((ifp->if_flags & IFF_POINTOPOINT) && 576 (ifra->ifra_dstaddr.sin_family == AF_INET)) { 577 in_ifscrub(ifp, ia, LLE_STATIC); 578 ia->ia_dstaddr = ifra->ifra_dstaddr; 579 maskIsNew = 1; /* We lie; but the effect's the same */ 580 } 581 if (hostIsNew || maskIsNew) 582 error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0, 583 maskIsNew, (ocmd == cmd ? ifra->ifra_vhid : 0)); 584 if (error != 0 && iaIsNew) 585 break; 586 587 if ((ifp->if_flags & IFF_BROADCAST) && 588 ifra->ifra_broadaddr.sin_len) 589 ia->ia_broadaddr = ifra->ifra_broadaddr; 590 if (error == 0) { 591 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 592 if (iaIsFirst && 593 (ifp->if_flags & IFF_MULTICAST) != 0) { 594 error = in_joingroup(ifp, &allhosts_addr, 595 NULL, &ii->ii_allhosts); 596 } 597 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 598 } 599 goto out; 600 601 case SIOCDIFADDR: 602 /* 603 * in_ifscrub kills the interface route. 604 */ 605 in_ifscrub(ifp, ia, LLE_STATIC); 606 607 /* 608 * in_ifadown gets rid of all the rest of 609 * the routes. This is not quite the right 610 * thing to do, but at least if we are running 611 * a routing process they will come back. 612 */ 613 in_ifadown(&ia->ia_ifa, 1); 614 EVENTHANDLER_INVOKE(ifaddr_event, ifp); 615 error = 0; 616 break; 617 618 default: 619 panic("in_control: unsupported ioctl"); 620 } 621 622 if (ia->ia_ifa.ifa_carp) 623 (*carp_detach_p)(&ia->ia_ifa); 624 625 IF_ADDR_WLOCK(ifp); 626 /* Re-check that ia is still part of the list. */ 627 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 628 if (ifa == &ia->ia_ifa) 629 break; 630 } 631 if (ifa == NULL) { 632 /* 633 * If we lost the race with another thread, there is no need to 634 * try it again for the next loop as there is no other exit 635 * path between here and out. 636 */ 637 IF_ADDR_WUNLOCK(ifp); 638 error = EADDRNOTAVAIL; 639 goto out; 640 } 641 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); 642 IF_ADDR_WUNLOCK(ifp); 643 ifa_free(&ia->ia_ifa); /* if_addrhead */ 644 645 IN_IFADDR_WLOCK(); 646 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link); 647 648 LIST_REMOVE(ia, ia_hash); 649 IN_IFADDR_WUNLOCK(); 650 /* 651 * If this is the last IPv4 address configured on this 652 * interface, leave the all-hosts group. 653 * No state-change report need be transmitted. 654 */ 655 IFP_TO_IA(ifp, iap); 656 if (iap == NULL) { 657 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 658 IN_MULTI_LOCK(); 659 if (ii->ii_allhosts) { 660 (void)in_leavegroup_locked(ii->ii_allhosts, NULL); 661 ii->ii_allhosts = NULL; 662 } 663 IN_MULTI_UNLOCK(); 664 } else 665 ifa_free(&iap->ia_ifa); 666 667 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 668out: 669 if (ia != NULL) 670 ifa_free(&ia->ia_ifa); 671 return (error); 672} 673 674/* 675 * SIOC[GAD]LIFADDR. 676 * SIOCGLIFADDR: get first address. (?!?) 677 * SIOCGLIFADDR with IFLR_PREFIX: 678 * get first address that matches the specified prefix. 679 * SIOCALIFADDR: add the specified address. 680 * SIOCALIFADDR with IFLR_PREFIX: 681 * EINVAL since we can't deduce hostid part of the address. 682 * SIOCDLIFADDR: delete the specified address. 683 * SIOCDLIFADDR with IFLR_PREFIX: 684 * delete the first address that matches the specified prefix. 685 * return values: 686 * EINVAL on invalid parameters 687 * EADDRNOTAVAIL on prefix match failed/specified address not found 688 * other values may be returned from in_ioctl() 689 */ 690static int 691in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data, 692 struct ifnet *ifp, struct thread *td) 693{ 694 struct if_laddrreq *iflr = (struct if_laddrreq *)data; 695 struct ifaddr *ifa; 696 697 /* sanity checks */ 698 if (data == NULL || ifp == NULL) { 699 panic("invalid argument to in_lifaddr_ioctl"); 700 /*NOTRECHED*/ 701 } 702 703 switch (cmd) { 704 case SIOCGLIFADDR: 705 /* address must be specified on GET with IFLR_PREFIX */ 706 if ((iflr->flags & IFLR_PREFIX) == 0) 707 break; 708 /*FALLTHROUGH*/ 709 case SIOCALIFADDR: 710 case SIOCDLIFADDR: 711 /* address must be specified on ADD and DELETE */ 712 if (iflr->addr.ss_family != AF_INET) 713 return (EINVAL); 714 if (iflr->addr.ss_len != sizeof(struct sockaddr_in)) 715 return (EINVAL); 716 /* XXX need improvement */ 717 if (iflr->dstaddr.ss_family 718 && iflr->dstaddr.ss_family != AF_INET) 719 return (EINVAL); 720 if (iflr->dstaddr.ss_family 721 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in)) 722 return (EINVAL); 723 break; 724 default: /*shouldn't happen*/ 725 return (EOPNOTSUPP); 726 } 727 if (sizeof(struct in_addr) * 8 < iflr->prefixlen) 728 return (EINVAL); 729 730 switch (cmd) { 731 case SIOCALIFADDR: 732 { 733 struct in_aliasreq ifra; 734 735 if (iflr->flags & IFLR_PREFIX) 736 return (EINVAL); 737 738 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */ 739 bzero(&ifra, sizeof(ifra)); 740 bcopy(iflr->iflr_name, ifra.ifra_name, 741 sizeof(ifra.ifra_name)); 742 743 bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len); 744 745 if (iflr->dstaddr.ss_family) { /*XXX*/ 746 bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr, 747 iflr->dstaddr.ss_len); 748 } 749 750 ifra.ifra_mask.sin_family = AF_INET; 751 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in); 752 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen); 753 754 return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td)); 755 } 756 case SIOCGLIFADDR: 757 case SIOCDLIFADDR: 758 { 759 struct in_ifaddr *ia; 760 struct in_addr mask, candidate, match; 761 struct sockaddr_in *sin; 762 763 bzero(&mask, sizeof(mask)); 764 bzero(&match, sizeof(match)); 765 if (iflr->flags & IFLR_PREFIX) { 766 /* lookup a prefix rather than address. */ 767 in_len2mask(&mask, iflr->prefixlen); 768 769 sin = (struct sockaddr_in *)&iflr->addr; 770 match.s_addr = sin->sin_addr.s_addr; 771 match.s_addr &= mask.s_addr; 772 773 /* if you set extra bits, that's wrong */ 774 if (match.s_addr != sin->sin_addr.s_addr) 775 return (EINVAL); 776 777 } else { 778 /* on getting an address, take the 1st match */ 779 /* on deleting an address, do exact match */ 780 if (cmd != SIOCGLIFADDR) { 781 in_len2mask(&mask, 32); 782 sin = (struct sockaddr_in *)&iflr->addr; 783 match.s_addr = sin->sin_addr.s_addr; 784 } 785 } 786 787 IF_ADDR_RLOCK(ifp); 788 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 789 if (ifa->ifa_addr->sa_family != AF_INET) 790 continue; 791 if (match.s_addr == 0) 792 break; 793 sin = (struct sockaddr_in *)&ifa->ifa_addr; 794 candidate.s_addr = sin->sin_addr.s_addr; 795 candidate.s_addr &= mask.s_addr; 796 if (candidate.s_addr == match.s_addr) 797 break; 798 } 799 if (ifa != NULL) 800 ifa_ref(ifa); 801 IF_ADDR_RUNLOCK(ifp); 802 if (ifa == NULL) 803 return (EADDRNOTAVAIL); 804 ia = (struct in_ifaddr *)ifa; 805 806 if (cmd == SIOCGLIFADDR) { 807 /* fill in the if_laddrreq structure */ 808 bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len); 809 810 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 811 bcopy(&ia->ia_dstaddr, &iflr->dstaddr, 812 ia->ia_dstaddr.sin_len); 813 } else 814 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr)); 815 816 iflr->prefixlen = 817 in_mask2len(&ia->ia_sockmask.sin_addr); 818 819 iflr->flags = 0; /*XXX*/ 820 ifa_free(ifa); 821 822 return (0); 823 } else { 824 struct in_aliasreq ifra; 825 826 /* fill in_aliasreq and do ioctl(SIOCDIFADDR) */ 827 bzero(&ifra, sizeof(ifra)); 828 bcopy(iflr->iflr_name, ifra.ifra_name, 829 sizeof(ifra.ifra_name)); 830 831 bcopy(&ia->ia_addr, &ifra.ifra_addr, 832 ia->ia_addr.sin_len); 833 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 834 bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr, 835 ia->ia_dstaddr.sin_len); 836 } 837 bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr, 838 ia->ia_sockmask.sin_len); 839 ifa_free(ifa); 840 841 return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra, 842 ifp, td)); 843 } 844 } 845 } 846 847 return (EOPNOTSUPP); /*just for safety*/ 848} 849 850/* 851 * Delete any existing route for an interface. 852 */ 853void 854in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags) 855{ 856 857 in_scrubprefix(ia, flags); 858} 859 860/* 861 * Initialize an interface's internet address 862 * and routing table entry. 863 */ 864static int 865in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin, 866 int scrub, int masksupplied, int vhid) 867{ 868 register u_long i = ntohl(sin->sin_addr.s_addr); 869 int flags = RTF_UP, error = 0; 870 871 if (scrub) 872 in_scrubprefix(ia, LLE_STATIC); 873 874 IN_IFADDR_WLOCK(); 875 if (ia->ia_addr.sin_family == AF_INET) 876 LIST_REMOVE(ia, ia_hash); 877 ia->ia_addr = *sin; 878 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), 879 ia, ia_hash); 880 IN_IFADDR_WUNLOCK(); 881 882 if (vhid > 0) { 883 if (carp_attach_p != NULL) 884 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 885 else 886 error = EPROTONOSUPPORT; 887 } 888 if (error) 889 return (error); 890 891 /* 892 * Give the interface a chance to initialize 893 * if this is its first address, 894 * and to validate the address if necessary. 895 */ 896 if (ifp->if_ioctl != NULL && 897 (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia)) != 0) 898 /* LIST_REMOVE(ia, ia_hash) is done in in_control */ 899 return (error); 900 901 /* 902 * Be compatible with network classes, if netmask isn't supplied, 903 * guess it based on classes. 904 */ 905 if (!masksupplied) { 906 if (IN_CLASSA(i)) 907 ia->ia_subnetmask = IN_CLASSA_NET; 908 else if (IN_CLASSB(i)) 909 ia->ia_subnetmask = IN_CLASSB_NET; 910 else 911 ia->ia_subnetmask = IN_CLASSC_NET; 912 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 913 } 914 ia->ia_subnet = i & ia->ia_subnetmask; 915 in_socktrim(&ia->ia_sockmask); 916 /* 917 * Add route for the network. 918 */ 919 ia->ia_ifa.ifa_metric = ifp->if_metric; 920 if (ifp->if_flags & IFF_BROADCAST) { 921 if (ia->ia_subnetmask == IN_RFC3021_MASK) 922 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 923 else 924 ia->ia_broadaddr.sin_addr.s_addr = 925 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 926 } else if (ifp->if_flags & IFF_LOOPBACK) { 927 ia->ia_dstaddr = ia->ia_addr; 928 flags |= RTF_HOST; 929 } else if (ifp->if_flags & IFF_POINTOPOINT) { 930 if (ia->ia_dstaddr.sin_family != AF_INET) 931 return (0); 932 flags |= RTF_HOST; 933 } 934 if (!vhid && (error = in_addprefix(ia, flags)) != 0) 935 return (error); 936 937 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY) 938 return (0); 939 940 if (ifp->if_flags & IFF_POINTOPOINT && 941 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr) 942 return (0); 943 944 /* 945 * add a loopback route to self 946 */ 947 if (V_useloopback && !vhid && !(ifp->if_flags & IFF_LOOPBACK)) { 948 struct route ia_ro; 949 950 bzero(&ia_ro, sizeof(ia_ro)); 951 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr; 952 rtalloc_ign_fib(&ia_ro, 0, 0); 953 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) && 954 (ia_ro.ro_rt->rt_ifp == V_loif)) { 955 RT_LOCK(ia_ro.ro_rt); 956 RT_ADDREF(ia_ro.ro_rt); 957 RTFREE_LOCKED(ia_ro.ro_rt); 958 } else 959 error = ifa_add_loopback_route((struct ifaddr *)ia, 960 (struct sockaddr *)&ia->ia_addr); 961 if (error == 0) 962 ia->ia_flags |= IFA_RTSELF; 963 if (ia_ro.ro_rt != NULL) 964 RTFREE(ia_ro.ro_rt); 965 } 966 967 return (error); 968} 969 970#define rtinitflags(x) \ 971 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 972 ? RTF_HOST : 0) 973 974/* 975 * Generate a routing message when inserting or deleting 976 * an interface address alias. 977 */ 978static void in_addralias_rtmsg(int cmd, struct in_addr *prefix, 979 struct in_ifaddr *target) 980{ 981 struct route pfx_ro; 982 struct sockaddr_in *pfx_addr; 983 struct rtentry msg_rt; 984 985 /* QL: XXX 986 * This is a bit questionable because there is no 987 * additional route entry added/deleted for an address 988 * alias. Therefore this route report is inaccurate. 989 */ 990 bzero(&pfx_ro, sizeof(pfx_ro)); 991 pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst); 992 pfx_addr->sin_len = sizeof(*pfx_addr); 993 pfx_addr->sin_family = AF_INET; 994 pfx_addr->sin_addr = *prefix; 995 rtalloc_ign_fib(&pfx_ro, 0, 0); 996 if (pfx_ro.ro_rt != NULL) { 997 msg_rt = *pfx_ro.ro_rt; 998 999 /* QL: XXX 1000 * Point the gateway to the new interface 1001 * address as if a new prefix route entry has 1002 * been added through the new address alias. 1003 * All other parts of the rtentry is accurate, 1004 * e.g., rt_key, rt_mask, rt_ifp etc. 1005 */ 1006 msg_rt.rt_gateway = 1007 (struct sockaddr *)&target->ia_addr; 1008 rt_newaddrmsg(cmd, 1009 (struct ifaddr *)target, 1010 0, &msg_rt); 1011 RTFREE(pfx_ro.ro_rt); 1012 } 1013 return; 1014} 1015 1016/* 1017 * Check if we have a route for the given prefix already or add one accordingly. 1018 */ 1019int 1020in_addprefix(struct in_ifaddr *target, int flags) 1021{ 1022 struct in_ifaddr *ia; 1023 struct in_addr prefix, mask, p, m; 1024 int error; 1025 1026 if ((flags & RTF_HOST) != 0) { 1027 prefix = target->ia_dstaddr.sin_addr; 1028 mask.s_addr = 0; 1029 } else { 1030 prefix = target->ia_addr.sin_addr; 1031 mask = target->ia_sockmask.sin_addr; 1032 prefix.s_addr &= mask.s_addr; 1033 } 1034 1035 IN_IFADDR_RLOCK(); 1036 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1037 if (rtinitflags(ia)) { 1038 p = ia->ia_dstaddr.sin_addr; 1039 1040 if (prefix.s_addr != p.s_addr) 1041 continue; 1042 } else { 1043 p = ia->ia_addr.sin_addr; 1044 m = ia->ia_sockmask.sin_addr; 1045 p.s_addr &= m.s_addr; 1046 1047 if (prefix.s_addr != p.s_addr || 1048 mask.s_addr != m.s_addr) 1049 continue; 1050 } 1051 1052 /* 1053 * If we got a matching prefix route inserted by other 1054 * interface address, we are done here. 1055 */ 1056 if (ia->ia_flags & IFA_ROUTE) { 1057#ifdef RADIX_MPATH 1058 if (ia->ia_addr.sin_addr.s_addr == 1059 target->ia_addr.sin_addr.s_addr) { 1060 IN_IFADDR_RUNLOCK(); 1061 return (EEXIST); 1062 } else 1063 break; 1064#endif 1065 if (V_nosameprefix) { 1066 IN_IFADDR_RUNLOCK(); 1067 return (EEXIST); 1068 } else { 1069 in_addralias_rtmsg(RTM_ADD, &prefix, target); 1070 IN_IFADDR_RUNLOCK(); 1071 return (0); 1072 } 1073 } 1074 } 1075 IN_IFADDR_RUNLOCK(); 1076 1077 /* 1078 * No-one seem to have this prefix route, so we try to insert it. 1079 */ 1080 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 1081 if (!error) 1082 target->ia_flags |= IFA_ROUTE; 1083 return (error); 1084} 1085 1086/* 1087 * If there is no other address in the system that can serve a route to the 1088 * same prefix, remove the route. Hand over the route to the new address 1089 * otherwise. 1090 */ 1091int 1092in_scrubprefix(struct in_ifaddr *target, u_int flags) 1093{ 1094 struct in_ifaddr *ia; 1095 struct in_addr prefix, mask, p, m; 1096 int error = 0; 1097 struct sockaddr_in prefix0, mask0; 1098 1099 /* 1100 * Remove the loopback route to the interface address. 1101 * The "useloopback" setting is not consulted because if the 1102 * user configures an interface address, turns off this 1103 * setting, and then tries to delete that interface address, 1104 * checking the current setting of "useloopback" would leave 1105 * that interface address loopback route untouched, which 1106 * would be wrong. Therefore the interface address loopback route 1107 * deletion is unconditional. 1108 */ 1109 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 1110 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 1111 (target->ia_flags & IFA_RTSELF)) { 1112 struct route ia_ro; 1113 int freeit = 0; 1114 1115 bzero(&ia_ro, sizeof(ia_ro)); 1116 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr; 1117 rtalloc_ign_fib(&ia_ro, 0, 0); 1118 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) && 1119 (ia_ro.ro_rt->rt_ifp == V_loif)) { 1120 RT_LOCK(ia_ro.ro_rt); 1121 if (ia_ro.ro_rt->rt_refcnt <= 1) 1122 freeit = 1; 1123 else if (flags & LLE_STATIC) { 1124 RT_REMREF(ia_ro.ro_rt); 1125 target->ia_flags &= ~IFA_RTSELF; 1126 } 1127 RTFREE_LOCKED(ia_ro.ro_rt); 1128 } 1129 if (freeit && (flags & LLE_STATIC)) { 1130 error = ifa_del_loopback_route((struct ifaddr *)target, 1131 (struct sockaddr *)&target->ia_addr); 1132 if (error == 0) 1133 target->ia_flags &= ~IFA_RTSELF; 1134 } 1135 if ((flags & LLE_STATIC) && 1136 !(target->ia_ifp->if_flags & IFF_NOARP)) 1137 /* remove arp cache */ 1138 arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr); 1139 } 1140 1141 if (rtinitflags(target)) { 1142 prefix = target->ia_dstaddr.sin_addr; 1143 mask.s_addr = 0; 1144 } else { 1145 prefix = target->ia_addr.sin_addr; 1146 mask = target->ia_sockmask.sin_addr; 1147 prefix.s_addr &= mask.s_addr; 1148 } 1149 1150 if ((target->ia_flags & IFA_ROUTE) == 0) { 1151 in_addralias_rtmsg(RTM_DELETE, &prefix, target); 1152 return (0); 1153 } 1154 1155 IN_IFADDR_RLOCK(); 1156 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1157 if (rtinitflags(ia)) { 1158 p = ia->ia_dstaddr.sin_addr; 1159 1160 if (prefix.s_addr != p.s_addr) 1161 continue; 1162 } else { 1163 p = ia->ia_addr.sin_addr; 1164 m = ia->ia_sockmask.sin_addr; 1165 p.s_addr &= m.s_addr; 1166 1167 if (prefix.s_addr != p.s_addr || 1168 mask.s_addr != m.s_addr) 1169 continue; 1170 } 1171 1172 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 1173 continue; 1174 1175 /* 1176 * If we got a matching prefix address, move IFA_ROUTE and 1177 * the route itself to it. Make sure that routing daemons 1178 * get a heads-up. 1179 */ 1180 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1181 ifa_ref(&ia->ia_ifa); 1182 IN_IFADDR_RUNLOCK(); 1183 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 1184 rtinitflags(target)); 1185 if (error == 0) 1186 target->ia_flags &= ~IFA_ROUTE; 1187 else 1188 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 1189 error); 1190 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 1191 rtinitflags(ia) | RTF_UP); 1192 if (error == 0) 1193 ia->ia_flags |= IFA_ROUTE; 1194 else 1195 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 1196 error); 1197 ifa_free(&ia->ia_ifa); 1198 return (error); 1199 } 1200 } 1201 IN_IFADDR_RUNLOCK(); 1202 1203 /* 1204 * remove all L2 entries on the given prefix 1205 */ 1206 bzero(&prefix0, sizeof(prefix0)); 1207 prefix0.sin_len = sizeof(prefix0); 1208 prefix0.sin_family = AF_INET; 1209 prefix0.sin_addr.s_addr = target->ia_subnet; 1210 bzero(&mask0, sizeof(mask0)); 1211 mask0.sin_len = sizeof(mask0); 1212 mask0.sin_family = AF_INET; 1213 mask0.sin_addr.s_addr = target->ia_subnetmask; 1214 lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0, 1215 (struct sockaddr *)&mask0, flags); 1216 1217 /* 1218 * As no-one seem to have this prefix, we can remove the route. 1219 */ 1220 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 1221 if (error == 0) 1222 target->ia_flags &= ~IFA_ROUTE; 1223 else 1224 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 1225 return (error); 1226} 1227 1228#undef rtinitflags 1229 1230/* 1231 * Return 1 if the address might be a local broadcast address. 1232 */ 1233int 1234in_broadcast(struct in_addr in, struct ifnet *ifp) 1235{ 1236 register struct ifaddr *ifa; 1237 u_long t; 1238 1239 if (in.s_addr == INADDR_BROADCAST || 1240 in.s_addr == INADDR_ANY) 1241 return (1); 1242 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1243 return (0); 1244 t = ntohl(in.s_addr); 1245 /* 1246 * Look through the list of addresses for a match 1247 * with a broadcast address. 1248 */ 1249#define ia ((struct in_ifaddr *)ifa) 1250 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1251 if (ifa->ifa_addr->sa_family == AF_INET && 1252 (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 1253 /* 1254 * Check for old-style (host 0) broadcast, but 1255 * taking into account that RFC 3021 obsoletes it. 1256 */ 1257 (ia->ia_subnetmask != IN_RFC3021_MASK && 1258 t == ia->ia_subnet)) && 1259 /* 1260 * Check for an all one subnetmask. These 1261 * only exist when an interface gets a secondary 1262 * address. 1263 */ 1264 ia->ia_subnetmask != (u_long)0xffffffff) 1265 return (1); 1266 return (0); 1267#undef ia 1268} 1269 1270/* 1271 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 1272 */ 1273void 1274in_ifdetach(struct ifnet *ifp) 1275{ 1276 1277 in_pcbpurgeif0(&V_ripcbinfo, ifp); 1278 in_pcbpurgeif0(&V_udbinfo, ifp); 1279 in_purgemaddrs(ifp); 1280} 1281 1282/* 1283 * Delete all IPv4 multicast address records, and associated link-layer 1284 * multicast address records, associated with ifp. 1285 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1286 * XXX This should not race with ifma_protospec being set during 1287 * a new allocation, if it does, we have bigger problems. 1288 */ 1289static void 1290in_purgemaddrs(struct ifnet *ifp) 1291{ 1292 LIST_HEAD(,in_multi) purgeinms; 1293 struct in_multi *inm, *tinm; 1294 struct ifmultiaddr *ifma; 1295 1296 LIST_INIT(&purgeinms); 1297 IN_MULTI_LOCK(); 1298 1299 /* 1300 * Extract list of in_multi associated with the detaching ifp 1301 * which the PF_INET layer is about to release. 1302 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1303 * by code further down. 1304 */ 1305 IF_ADDR_RLOCK(ifp); 1306 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1307 if (ifma->ifma_addr->sa_family != AF_INET || 1308 ifma->ifma_protospec == NULL) 1309 continue; 1310#if 0 1311 KASSERT(ifma->ifma_protospec != NULL, 1312 ("%s: ifma_protospec is NULL", __func__)); 1313#endif 1314 inm = (struct in_multi *)ifma->ifma_protospec; 1315 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 1316 } 1317 IF_ADDR_RUNLOCK(ifp); 1318 1319 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 1320 LIST_REMOVE(inm, inm_link); 1321 inm_release_locked(inm); 1322 } 1323 igmp_ifdetach(ifp); 1324 1325 IN_MULTI_UNLOCK(); 1326} 1327 1328struct in_llentry { 1329 struct llentry base; 1330 struct sockaddr_in l3_addr4; 1331}; 1332 1333static struct llentry * 1334in_lltable_new(const struct sockaddr *l3addr, u_int flags) 1335{ 1336 struct in_llentry *lle; 1337 1338 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO); 1339 if (lle == NULL) /* NB: caller generates msg */ 1340 return NULL; 1341 1342 callout_init(&lle->base.la_timer, CALLOUT_MPSAFE); 1343 /* 1344 * For IPv4 this will trigger "arpresolve" to generate 1345 * an ARP request. 1346 */ 1347 lle->base.la_expire = time_uptime; /* mark expired */ 1348 lle->l3_addr4 = *(const struct sockaddr_in *)l3addr; 1349 lle->base.lle_refcnt = 1; 1350 LLE_LOCK_INIT(&lle->base); 1351 return &lle->base; 1352} 1353 1354/* 1355 * Deletes an address from the address table. 1356 * This function is called by the timer functions 1357 * such as arptimer() and nd6_llinfo_timer(), and 1358 * the caller does the locking. 1359 */ 1360static void 1361in_lltable_free(struct lltable *llt, struct llentry *lle) 1362{ 1363 LLE_WUNLOCK(lle); 1364 LLE_LOCK_DESTROY(lle); 1365 free(lle, M_LLTABLE); 1366} 1367 1368 1369#define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1370 (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 1371 1372static void 1373in_lltable_prefix_free(struct lltable *llt, 1374 const struct sockaddr *prefix, 1375 const struct sockaddr *mask, 1376 u_int flags) 1377{ 1378 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 1379 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 1380 struct llentry *lle, *next; 1381 register int i; 1382 size_t pkts_dropped; 1383 1384 for (i=0; i < LLTBL_HASHTBL_SIZE; i++) { 1385 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 1386 1387 /* 1388 * (flags & LLE_STATIC) means deleting all entries 1389 * including static ARP entries 1390 */ 1391 if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle), 1392 pfx, msk) && 1393 ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) { 1394 int canceled; 1395 1396 canceled = callout_drain(&lle->la_timer); 1397 LLE_WLOCK(lle); 1398 if (canceled) 1399 LLE_REMREF(lle); 1400 pkts_dropped = llentry_free(lle); 1401 ARPSTAT_ADD(dropped, pkts_dropped); 1402 } 1403 } 1404 } 1405} 1406 1407 1408static int 1409in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1410{ 1411 struct rtentry *rt; 1412 1413 KASSERT(l3addr->sa_family == AF_INET, 1414 ("sin_family %d", l3addr->sa_family)); 1415 1416 /* XXX rtalloc1 should take a const param */ 1417 rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0); 1418 1419 if (rt == NULL) 1420 return (EINVAL); 1421 1422 /* 1423 * If the gateway for an existing host route matches the target L3 1424 * address, which is a special route inserted by some implementation 1425 * such as MANET, and the interface is of the correct type, then 1426 * allow for ARP to proceed. 1427 */ 1428 if (rt->rt_flags & RTF_GATEWAY) { 1429 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 1430 rt->rt_ifp->if_type != IFT_ETHER || 1431 (rt->rt_ifp->if_flags & 1432 (IFF_NOARP | IFF_STATICARP)) != 0 || 1433 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 1434 sizeof(in_addr_t)) != 0) { 1435 RTFREE_LOCKED(rt); 1436 return (EINVAL); 1437 } 1438 } 1439 1440 /* 1441 * Make sure that at least the destination address is covered 1442 * by the route. This is for handling the case where 2 or more 1443 * interfaces have the same prefix. An incoming packet arrives 1444 * on one interface and the corresponding outgoing packet leaves 1445 * another interface. 1446 */ 1447 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 1448 const char *sa, *mask, *addr, *lim; 1449 int len; 1450 1451 mask = (const char *)rt_mask(rt); 1452 /* 1453 * Just being extra cautious to avoid some custom 1454 * code getting into trouble. 1455 */ 1456 if (mask == NULL) { 1457 RTFREE_LOCKED(rt); 1458 return (EINVAL); 1459 } 1460 1461 sa = (const char *)rt_key(rt); 1462 addr = (const char *)l3addr; 1463 len = ((const struct sockaddr_in *)l3addr)->sin_len; 1464 lim = addr + len; 1465 1466 for ( ; addr < lim; sa++, mask++, addr++) { 1467 if ((*sa ^ *addr) & *mask) { 1468#ifdef DIAGNOSTIC 1469 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1470 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 1471#endif 1472 RTFREE_LOCKED(rt); 1473 return (EINVAL); 1474 } 1475 } 1476 } 1477 1478 RTFREE_LOCKED(rt); 1479 return (0); 1480} 1481 1482/* 1483 * Return NULL if not found or marked for deletion. 1484 * If found return lle read locked. 1485 */ 1486static struct llentry * 1487in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1488{ 1489 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1490 struct ifnet *ifp = llt->llt_ifp; 1491 struct llentry *lle; 1492 struct llentries *lleh; 1493 u_int hashkey; 1494 1495 IF_AFDATA_LOCK_ASSERT(ifp); 1496 KASSERT(l3addr->sa_family == AF_INET, 1497 ("sin_family %d", l3addr->sa_family)); 1498 1499 hashkey = sin->sin_addr.s_addr; 1500 lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; 1501 LIST_FOREACH(lle, lleh, lle_next) { 1502 struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle); 1503 if (lle->la_flags & LLE_DELETED) 1504 continue; 1505 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr) 1506 break; 1507 } 1508 if (lle == NULL) { 1509#ifdef DIAGNOSTIC 1510 if (flags & LLE_DELETE) 1511 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); 1512#endif 1513 if (!(flags & LLE_CREATE)) 1514 return (NULL); 1515 /* 1516 * A route that covers the given address must have 1517 * been installed 1st because we are doing a resolution, 1518 * verify this. 1519 */ 1520 if (!(flags & LLE_IFADDR) && 1521 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1522 goto done; 1523 1524 lle = in_lltable_new(l3addr, flags); 1525 if (lle == NULL) { 1526 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1527 goto done; 1528 } 1529 lle->la_flags = flags & ~LLE_CREATE; 1530 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { 1531 bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); 1532 lle->la_flags |= (LLE_VALID | LLE_STATIC); 1533 } 1534 1535 lle->lle_tbl = llt; 1536 lle->lle_head = lleh; 1537 LIST_INSERT_HEAD(lleh, lle, lle_next); 1538 } else if (flags & LLE_DELETE) { 1539 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) { 1540 LLE_WLOCK(lle); 1541 lle->la_flags = LLE_DELETED; 1542 EVENTHANDLER_INVOKE(arp_update_event, lle); 1543 LLE_WUNLOCK(lle); 1544#ifdef DIAGNOSTIC 1545 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1546#endif 1547 } 1548 lle = (void *)-1; 1549 1550 } 1551 if (LLE_IS_VALID(lle)) { 1552 if (flags & LLE_EXCLUSIVE) 1553 LLE_WLOCK(lle); 1554 else 1555 LLE_RLOCK(lle); 1556 } 1557done: 1558 return (lle); 1559} 1560 1561static int 1562in_lltable_dump(struct lltable *llt, struct sysctl_req *wr) 1563{ 1564#define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle)) 1565 struct ifnet *ifp = llt->llt_ifp; 1566 struct llentry *lle; 1567 /* XXX stack use */ 1568 struct { 1569 struct rt_msghdr rtm; 1570 struct sockaddr_inarp sin; 1571 struct sockaddr_dl sdl; 1572 } arpc; 1573 int error, i; 1574 1575 LLTABLE_LOCK_ASSERT(); 1576 1577 error = 0; 1578 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1579 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1580 struct sockaddr_dl *sdl; 1581 1582 /* skip deleted entries */ 1583 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1584 continue; 1585 /* Skip if jailed and not a valid IP of the prison. */ 1586 if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0) 1587 continue; 1588 /* 1589 * produce a msg made of: 1590 * struct rt_msghdr; 1591 * struct sockaddr_inarp; (IPv4) 1592 * struct sockaddr_dl; 1593 */ 1594 bzero(&arpc, sizeof(arpc)); 1595 arpc.rtm.rtm_msglen = sizeof(arpc); 1596 arpc.rtm.rtm_version = RTM_VERSION; 1597 arpc.rtm.rtm_type = RTM_GET; 1598 arpc.rtm.rtm_flags = RTF_UP; 1599 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1600 arpc.sin.sin_family = AF_INET; 1601 arpc.sin.sin_len = sizeof(arpc.sin); 1602 arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr; 1603 1604 /* publish */ 1605 if (lle->la_flags & LLE_PUB) { 1606 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1607 /* proxy only */ 1608 if (lle->la_flags & LLE_PROXY) 1609 arpc.sin.sin_other = SIN_PROXY; 1610 } 1611 1612 sdl = &arpc.sdl; 1613 sdl->sdl_family = AF_LINK; 1614 sdl->sdl_len = sizeof(*sdl); 1615 sdl->sdl_index = ifp->if_index; 1616 sdl->sdl_type = ifp->if_type; 1617 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1618 sdl->sdl_alen = ifp->if_addrlen; 1619 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1620 } else { 1621 sdl->sdl_alen = 0; 1622 bzero(LLADDR(sdl), ifp->if_addrlen); 1623 } 1624 1625 arpc.rtm.rtm_rmx.rmx_expire = 1626 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1627 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1628 if (lle->la_flags & LLE_STATIC) 1629 arpc.rtm.rtm_flags |= RTF_STATIC; 1630 arpc.rtm.rtm_index = ifp->if_index; 1631 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1632 if (error) 1633 break; 1634 } 1635 } 1636 return error; 1637#undef SIN 1638} 1639 1640void * 1641in_domifattach(struct ifnet *ifp) 1642{ 1643 struct in_ifinfo *ii; 1644 struct lltable *llt; 1645 1646 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1647 1648 llt = lltable_init(ifp, AF_INET); 1649 if (llt != NULL) { 1650 llt->llt_free = in_lltable_free; 1651 llt->llt_prefix_free = in_lltable_prefix_free; 1652 llt->llt_lookup = in_lltable_lookup; 1653 llt->llt_dump = in_lltable_dump; 1654 } 1655 ii->ii_llt = llt; 1656 1657 ii->ii_igmp = igmp_domifattach(ifp); 1658 1659 return ii; 1660} 1661 1662void 1663in_domifdetach(struct ifnet *ifp, void *aux) 1664{ 1665 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1666 1667 igmp_domifdetach(ifp); 1668 lltable_free(ii->ii_llt); 1669 free(ii, M_IFADDR); 1670} 1671