in.c revision 267193
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: stable/10/sys/netinet/in.c 267193 2014-06-06 21:45:14Z asomers $"); 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); 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 88/* 89 * Return 1 if an internet address is for a ``local'' host 90 * (one to which we have a connection). 91 */ 92int 93in_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 */ 113int 114in_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 */ 134int 135in_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 */ 153static void 154in_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 167static int 168in_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 189static void 190in_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 */ 209int 210in_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 */ 598out: 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 */ 620static int 621in_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 */ 783void 784in_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 */ 794static int 795in_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 * Check if we have a route for the given prefix already or add one accordingly. 905 */ 906int 907in_addprefix(struct in_ifaddr *target, int flags) 908{ 909 struct in_ifaddr *ia; 910 struct in_addr prefix, mask, p, m; 911 int error; 912 913 if ((flags & RTF_HOST) != 0) { 914 prefix = target->ia_dstaddr.sin_addr; 915 mask.s_addr = 0; 916 } else { 917 prefix = target->ia_addr.sin_addr; 918 mask = target->ia_sockmask.sin_addr; 919 prefix.s_addr &= mask.s_addr; 920 } 921 922 IN_IFADDR_RLOCK(); 923 /* Look for an existing address with the same prefix, mask, and fib */ 924 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 925 if (rtinitflags(ia)) { 926 p = ia->ia_dstaddr.sin_addr; 927 928 if (prefix.s_addr != p.s_addr) 929 continue; 930 } else { 931 p = ia->ia_addr.sin_addr; 932 m = ia->ia_sockmask.sin_addr; 933 p.s_addr &= m.s_addr; 934 935 if (prefix.s_addr != p.s_addr || 936 mask.s_addr != m.s_addr) 937 continue; 938 } 939 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 940 continue; 941 942 /* 943 * If we got a matching prefix route inserted by other 944 * interface address, we are done here. 945 */ 946 if (ia->ia_flags & IFA_ROUTE) { 947#ifdef RADIX_MPATH 948 if (ia->ia_addr.sin_addr.s_addr == 949 target->ia_addr.sin_addr.s_addr) { 950 IN_IFADDR_RUNLOCK(); 951 return (EEXIST); 952 } else 953 break; 954#endif 955 if (V_nosameprefix) { 956 IN_IFADDR_RUNLOCK(); 957 return (EEXIST); 958 } else { 959 int fibnum; 960 961 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS : 962 target->ia_ifp->if_fib; 963 rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum); 964 IN_IFADDR_RUNLOCK(); 965 return (0); 966 } 967 } 968 } 969 IN_IFADDR_RUNLOCK(); 970 971 /* 972 * No-one seem to have this prefix route, so we try to insert it. 973 */ 974 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags); 975 if (!error) 976 target->ia_flags |= IFA_ROUTE; 977 return (error); 978} 979 980/* 981 * If there is no other address in the system that can serve a route to the 982 * same prefix, remove the route. Hand over the route to the new address 983 * otherwise. 984 */ 985int 986in_scrubprefix(struct in_ifaddr *target, u_int flags) 987{ 988 struct in_ifaddr *ia; 989 struct in_addr prefix, mask, p, m; 990 int error = 0, fibnum; 991 struct sockaddr_in prefix0, mask0; 992 993 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS : target->ia_ifp->if_fib; 994 995 /* 996 * Remove the loopback route to the interface address. 997 * The "useloopback" setting is not consulted because if the 998 * user configures an interface address, turns off this 999 * setting, and then tries to delete that interface address, 1000 * checking the current setting of "useloopback" would leave 1001 * that interface address loopback route untouched, which 1002 * would be wrong. Therefore the interface address loopback route 1003 * deletion is unconditional. 1004 */ 1005 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) && 1006 !(target->ia_ifp->if_flags & IFF_LOOPBACK) && 1007 (target->ia_flags & IFA_RTSELF)) { 1008 struct route ia_ro; 1009 int freeit = 0; 1010 int fib; 1011 1012 bzero(&ia_ro, sizeof(ia_ro)); 1013 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr; 1014 fib = target->ia_ifa.ifa_ifp->if_fib; 1015 rtalloc_ign_fib(&ia_ro, 0, fib); 1016 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) && 1017 (ia_ro.ro_rt->rt_ifp == V_loif)) { 1018 RT_LOCK(ia_ro.ro_rt); 1019 if (ia_ro.ro_rt->rt_refcnt <= 1) 1020 freeit = 1; 1021 else if (flags & LLE_STATIC) { 1022 RT_REMREF(ia_ro.ro_rt); 1023 target->ia_flags &= ~IFA_RTSELF; 1024 } 1025 RTFREE_LOCKED(ia_ro.ro_rt); 1026 } 1027 if (freeit && (flags & LLE_STATIC)) { 1028 error = ifa_del_loopback_route((struct ifaddr *)target, 1029 (struct sockaddr *)&target->ia_addr); 1030 if (error == 0) 1031 target->ia_flags &= ~IFA_RTSELF; 1032 } 1033 if ((flags & LLE_STATIC) && 1034 !(target->ia_ifp->if_flags & IFF_NOARP)) 1035 /* remove arp cache */ 1036 arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr); 1037 } 1038 1039 if (rtinitflags(target)) { 1040 prefix = target->ia_dstaddr.sin_addr; 1041 mask.s_addr = 0; 1042 } else { 1043 prefix = target->ia_addr.sin_addr; 1044 mask = target->ia_sockmask.sin_addr; 1045 prefix.s_addr &= mask.s_addr; 1046 } 1047 1048 if ((target->ia_flags & IFA_ROUTE) == 0) { 1049 rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum); 1050 return (0); 1051 } 1052 1053 IN_IFADDR_RLOCK(); 1054 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1055 if (rtinitflags(ia)) { 1056 p = ia->ia_dstaddr.sin_addr; 1057 1058 if (prefix.s_addr != p.s_addr) 1059 continue; 1060 } else { 1061 p = ia->ia_addr.sin_addr; 1062 m = ia->ia_sockmask.sin_addr; 1063 p.s_addr &= m.s_addr; 1064 1065 if (prefix.s_addr != p.s_addr || 1066 mask.s_addr != m.s_addr) 1067 continue; 1068 } 1069 1070 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 1071 continue; 1072 1073 /* 1074 * If we got a matching prefix address, move IFA_ROUTE and 1075 * the route itself to it. Make sure that routing daemons 1076 * get a heads-up. 1077 */ 1078 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1079 ifa_ref(&ia->ia_ifa); 1080 IN_IFADDR_RUNLOCK(); 1081 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, 1082 rtinitflags(target)); 1083 if (error == 0) 1084 target->ia_flags &= ~IFA_ROUTE; 1085 else 1086 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 1087 error); 1088 error = rtinit(&ia->ia_ifa, (int)RTM_ADD, 1089 rtinitflags(ia) | RTF_UP); 1090 if (error == 0) 1091 ia->ia_flags |= IFA_ROUTE; 1092 else 1093 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 1094 error); 1095 ifa_free(&ia->ia_ifa); 1096 return (error); 1097 } 1098 } 1099 IN_IFADDR_RUNLOCK(); 1100 1101 /* 1102 * remove all L2 entries on the given prefix 1103 */ 1104 bzero(&prefix0, sizeof(prefix0)); 1105 prefix0.sin_len = sizeof(prefix0); 1106 prefix0.sin_family = AF_INET; 1107 prefix0.sin_addr.s_addr = target->ia_subnet; 1108 bzero(&mask0, sizeof(mask0)); 1109 mask0.sin_len = sizeof(mask0); 1110 mask0.sin_family = AF_INET; 1111 mask0.sin_addr.s_addr = target->ia_subnetmask; 1112 lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0, 1113 (struct sockaddr *)&mask0, flags); 1114 1115 /* 1116 * As no-one seem to have this prefix, we can remove the route. 1117 */ 1118 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target)); 1119 if (error == 0) 1120 target->ia_flags &= ~IFA_ROUTE; 1121 else 1122 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 1123 return (error); 1124} 1125 1126#undef rtinitflags 1127 1128/* 1129 * Return 1 if the address might be a local broadcast address. 1130 */ 1131int 1132in_broadcast(struct in_addr in, struct ifnet *ifp) 1133{ 1134 register struct ifaddr *ifa; 1135 u_long t; 1136 1137 if (in.s_addr == INADDR_BROADCAST || 1138 in.s_addr == INADDR_ANY) 1139 return (1); 1140 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1141 return (0); 1142 t = ntohl(in.s_addr); 1143 /* 1144 * Look through the list of addresses for a match 1145 * with a broadcast address. 1146 */ 1147#define ia ((struct in_ifaddr *)ifa) 1148 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1149 if (ifa->ifa_addr->sa_family == AF_INET && 1150 (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 1151 /* 1152 * Check for old-style (host 0) broadcast, but 1153 * taking into account that RFC 3021 obsoletes it. 1154 */ 1155 (ia->ia_subnetmask != IN_RFC3021_MASK && 1156 t == ia->ia_subnet)) && 1157 /* 1158 * Check for an all one subnetmask. These 1159 * only exist when an interface gets a secondary 1160 * address. 1161 */ 1162 ia->ia_subnetmask != (u_long)0xffffffff) 1163 return (1); 1164 return (0); 1165#undef ia 1166} 1167 1168/* 1169 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 1170 */ 1171void 1172in_ifdetach(struct ifnet *ifp) 1173{ 1174 1175 in_pcbpurgeif0(&V_ripcbinfo, ifp); 1176 in_pcbpurgeif0(&V_udbinfo, ifp); 1177 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 1178 in_purgemaddrs(ifp); 1179} 1180 1181/* 1182 * Delete all IPv4 multicast address records, and associated link-layer 1183 * multicast address records, associated with ifp. 1184 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1185 * XXX This should not race with ifma_protospec being set during 1186 * a new allocation, if it does, we have bigger problems. 1187 */ 1188static void 1189in_purgemaddrs(struct ifnet *ifp) 1190{ 1191 LIST_HEAD(,in_multi) purgeinms; 1192 struct in_multi *inm, *tinm; 1193 struct ifmultiaddr *ifma; 1194 1195 LIST_INIT(&purgeinms); 1196 IN_MULTI_LOCK(); 1197 1198 /* 1199 * Extract list of in_multi associated with the detaching ifp 1200 * which the PF_INET layer is about to release. 1201 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1202 * by code further down. 1203 */ 1204 IF_ADDR_RLOCK(ifp); 1205 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1206 if (ifma->ifma_addr->sa_family != AF_INET || 1207 ifma->ifma_protospec == NULL) 1208 continue; 1209#if 0 1210 KASSERT(ifma->ifma_protospec != NULL, 1211 ("%s: ifma_protospec is NULL", __func__)); 1212#endif 1213 inm = (struct in_multi *)ifma->ifma_protospec; 1214 LIST_INSERT_HEAD(&purgeinms, inm, inm_link); 1215 } 1216 IF_ADDR_RUNLOCK(ifp); 1217 1218 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) { 1219 LIST_REMOVE(inm, inm_link); 1220 inm_release_locked(inm); 1221 } 1222 igmp_ifdetach(ifp); 1223 1224 IN_MULTI_UNLOCK(); 1225} 1226 1227struct in_llentry { 1228 struct llentry base; 1229 struct sockaddr_in l3_addr4; 1230}; 1231 1232/* 1233 * Deletes an address from the address table. 1234 * This function is called by the timer functions 1235 * such as arptimer() and nd6_llinfo_timer(), and 1236 * the caller does the locking. 1237 */ 1238static void 1239in_lltable_free(struct lltable *llt, struct llentry *lle) 1240{ 1241 LLE_WUNLOCK(lle); 1242 LLE_LOCK_DESTROY(lle); 1243 free(lle, M_LLTABLE); 1244} 1245 1246static struct llentry * 1247in_lltable_new(const struct sockaddr *l3addr, u_int flags) 1248{ 1249 struct in_llentry *lle; 1250 1251 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1252 if (lle == NULL) /* NB: caller generates msg */ 1253 return NULL; 1254 1255 /* 1256 * For IPv4 this will trigger "arpresolve" to generate 1257 * an ARP request. 1258 */ 1259 lle->base.la_expire = time_uptime; /* mark expired */ 1260 lle->l3_addr4 = *(const struct sockaddr_in *)l3addr; 1261 lle->base.lle_refcnt = 1; 1262 lle->base.lle_free = in_lltable_free; 1263 LLE_LOCK_INIT(&lle->base); 1264 callout_init_rw(&lle->base.la_timer, &lle->base.lle_lock, 1265 CALLOUT_RETURNUNLOCKED); 1266 1267 return (&lle->base); 1268} 1269 1270#define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1271 (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 1272 1273static void 1274in_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix, 1275 const struct sockaddr *mask, u_int flags) 1276{ 1277 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 1278 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 1279 struct llentry *lle, *next; 1280 int i; 1281 size_t pkts_dropped; 1282 1283 IF_AFDATA_WLOCK(llt->llt_ifp); 1284 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1285 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 1286 /* 1287 * (flags & LLE_STATIC) means deleting all entries 1288 * including static ARP entries. 1289 */ 1290 if (IN_ARE_MASKED_ADDR_EQUAL(satosin(L3_ADDR(lle)), 1291 pfx, msk) && ((flags & LLE_STATIC) || 1292 !(lle->la_flags & LLE_STATIC))) { 1293 LLE_WLOCK(lle); 1294 if (callout_stop(&lle->la_timer)) 1295 LLE_REMREF(lle); 1296 pkts_dropped = llentry_free(lle); 1297 ARPSTAT_ADD(dropped, pkts_dropped); 1298 } 1299 } 1300 } 1301 IF_AFDATA_WUNLOCK(llt->llt_ifp); 1302} 1303 1304 1305static int 1306in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1307{ 1308 struct rtentry *rt; 1309 1310 KASSERT(l3addr->sa_family == AF_INET, 1311 ("sin_family %d", l3addr->sa_family)); 1312 1313 /* XXX rtalloc1_fib should take a const param */ 1314 rt = rtalloc1_fib(__DECONST(struct sockaddr *, l3addr), 0, 0, 1315 ifp->if_fib); 1316 1317 if (rt == NULL) 1318 return (EINVAL); 1319 1320 /* 1321 * If the gateway for an existing host route matches the target L3 1322 * address, which is a special route inserted by some implementation 1323 * such as MANET, and the interface is of the correct type, then 1324 * allow for ARP to proceed. 1325 */ 1326 if (rt->rt_flags & RTF_GATEWAY) { 1327 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 1328 rt->rt_ifp->if_type != IFT_ETHER || 1329 (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1330 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 1331 sizeof(in_addr_t)) != 0) { 1332 RTFREE_LOCKED(rt); 1333 return (EINVAL); 1334 } 1335 } 1336 1337 /* 1338 * Make sure that at least the destination address is covered 1339 * by the route. This is for handling the case where 2 or more 1340 * interfaces have the same prefix. An incoming packet arrives 1341 * on one interface and the corresponding outgoing packet leaves 1342 * another interface. 1343 */ 1344 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 1345 const char *sa, *mask, *addr, *lim; 1346 int len; 1347 1348 mask = (const char *)rt_mask(rt); 1349 /* 1350 * Just being extra cautious to avoid some custom 1351 * code getting into trouble. 1352 */ 1353 if (mask == NULL) { 1354 RTFREE_LOCKED(rt); 1355 return (EINVAL); 1356 } 1357 1358 sa = (const char *)rt_key(rt); 1359 addr = (const char *)l3addr; 1360 len = ((const struct sockaddr_in *)l3addr)->sin_len; 1361 lim = addr + len; 1362 1363 for ( ; addr < lim; sa++, mask++, addr++) { 1364 if ((*sa ^ *addr) & *mask) { 1365#ifdef DIAGNOSTIC 1366 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1367 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 1368#endif 1369 RTFREE_LOCKED(rt); 1370 return (EINVAL); 1371 } 1372 } 1373 } 1374 1375 RTFREE_LOCKED(rt); 1376 return (0); 1377} 1378 1379/* 1380 * Return NULL if not found or marked for deletion. 1381 * If found return lle read locked. 1382 */ 1383static struct llentry * 1384in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1385{ 1386 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1387 struct ifnet *ifp = llt->llt_ifp; 1388 struct llentry *lle; 1389 struct llentries *lleh; 1390 u_int hashkey; 1391 1392 IF_AFDATA_LOCK_ASSERT(ifp); 1393 KASSERT(l3addr->sa_family == AF_INET, 1394 ("sin_family %d", l3addr->sa_family)); 1395 1396 hashkey = sin->sin_addr.s_addr; 1397 lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; 1398 LIST_FOREACH(lle, lleh, lle_next) { 1399 struct sockaddr_in *sa2 = satosin(L3_ADDR(lle)); 1400 if (lle->la_flags & LLE_DELETED) 1401 continue; 1402 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr) 1403 break; 1404 } 1405 if (lle == NULL) { 1406#ifdef DIAGNOSTIC 1407 if (flags & LLE_DELETE) 1408 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); 1409#endif 1410 if (!(flags & LLE_CREATE)) 1411 return (NULL); 1412 IF_AFDATA_WLOCK_ASSERT(ifp); 1413 /* 1414 * A route that covers the given address must have 1415 * been installed 1st because we are doing a resolution, 1416 * verify this. 1417 */ 1418 if (!(flags & LLE_IFADDR) && 1419 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1420 goto done; 1421 1422 lle = in_lltable_new(l3addr, flags); 1423 if (lle == NULL) { 1424 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1425 goto done; 1426 } 1427 lle->la_flags = flags & ~LLE_CREATE; 1428 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { 1429 bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); 1430 lle->la_flags |= (LLE_VALID | LLE_STATIC); 1431 } 1432 1433 lle->lle_tbl = llt; 1434 lle->lle_head = lleh; 1435 lle->la_flags |= LLE_LINKED; 1436 LIST_INSERT_HEAD(lleh, lle, lle_next); 1437 } else if (flags & LLE_DELETE) { 1438 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) { 1439 LLE_WLOCK(lle); 1440 lle->la_flags |= LLE_DELETED; 1441 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1442#ifdef DIAGNOSTIC 1443 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1444#endif 1445 if ((lle->la_flags & 1446 (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC) 1447 llentry_free(lle); 1448 else 1449 LLE_WUNLOCK(lle); 1450 } 1451 lle = (void *)-1; 1452 1453 } 1454 if (LLE_IS_VALID(lle)) { 1455 if (flags & LLE_EXCLUSIVE) 1456 LLE_WLOCK(lle); 1457 else 1458 LLE_RLOCK(lle); 1459 } 1460done: 1461 return (lle); 1462} 1463 1464static int 1465in_lltable_dump(struct lltable *llt, struct sysctl_req *wr) 1466{ 1467#define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle)) 1468 struct ifnet *ifp = llt->llt_ifp; 1469 struct llentry *lle; 1470 /* XXX stack use */ 1471 struct { 1472 struct rt_msghdr rtm; 1473 struct sockaddr_in sin; 1474 struct sockaddr_dl sdl; 1475 } arpc; 1476 int error, i; 1477 1478 LLTABLE_LOCK_ASSERT(); 1479 1480 error = 0; 1481 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { 1482 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1483 struct sockaddr_dl *sdl; 1484 1485 /* skip deleted entries */ 1486 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1487 continue; 1488 /* Skip if jailed and not a valid IP of the prison. */ 1489 if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0) 1490 continue; 1491 /* 1492 * produce a msg made of: 1493 * struct rt_msghdr; 1494 * struct sockaddr_in; (IPv4) 1495 * struct sockaddr_dl; 1496 */ 1497 bzero(&arpc, sizeof(arpc)); 1498 arpc.rtm.rtm_msglen = sizeof(arpc); 1499 arpc.rtm.rtm_version = RTM_VERSION; 1500 arpc.rtm.rtm_type = RTM_GET; 1501 arpc.rtm.rtm_flags = RTF_UP; 1502 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1503 arpc.sin.sin_family = AF_INET; 1504 arpc.sin.sin_len = sizeof(arpc.sin); 1505 arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr; 1506 1507 /* publish */ 1508 if (lle->la_flags & LLE_PUB) 1509 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1510 1511 sdl = &arpc.sdl; 1512 sdl->sdl_family = AF_LINK; 1513 sdl->sdl_len = sizeof(*sdl); 1514 sdl->sdl_index = ifp->if_index; 1515 sdl->sdl_type = ifp->if_type; 1516 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1517 sdl->sdl_alen = ifp->if_addrlen; 1518 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1519 } else { 1520 sdl->sdl_alen = 0; 1521 bzero(LLADDR(sdl), ifp->if_addrlen); 1522 } 1523 1524 arpc.rtm.rtm_rmx.rmx_expire = 1525 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1526 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1527 if (lle->la_flags & LLE_STATIC) 1528 arpc.rtm.rtm_flags |= RTF_STATIC; 1529 arpc.rtm.rtm_index = ifp->if_index; 1530 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1531 if (error) 1532 break; 1533 } 1534 } 1535 return error; 1536#undef SIN 1537} 1538 1539void * 1540in_domifattach(struct ifnet *ifp) 1541{ 1542 struct in_ifinfo *ii; 1543 struct lltable *llt; 1544 1545 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1546 1547 llt = lltable_init(ifp, AF_INET); 1548 if (llt != NULL) { 1549 llt->llt_prefix_free = in_lltable_prefix_free; 1550 llt->llt_lookup = in_lltable_lookup; 1551 llt->llt_dump = in_lltable_dump; 1552 } 1553 ii->ii_llt = llt; 1554 1555 ii->ii_igmp = igmp_domifattach(ifp); 1556 1557 return ii; 1558} 1559 1560void 1561in_domifdetach(struct ifnet *ifp, void *aux) 1562{ 1563 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1564 1565 igmp_domifdetach(ifp); 1566 lltable_free(ii->ii_llt); 1567 free(ii, M_IFADDR); 1568} 1569