1 /*- 2 * Copyright (c) 1980, 1986, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)if.c 8.5 (Berkeley) 1/9/95 30 * $FreeBSD$ 31 */ 32 33 #include "opt_compat.h" 34 #include "opt_inet6.h" 35 #include "opt_inet.h" 36 #include "opt_mac.h" 37 #include "opt_carp.h" 38 39 #include <sys/param.h> 40 #include <sys/types.h> 41 #include <sys/conf.h> 42 #include <sys/mac.h> 43 #include <sys/malloc.h> 44 #include <sys/sbuf.h> 45 #include <sys/bus.h> 46 #include <sys/mbuf.h> 47 #include <sys/systm.h> 48 #include <sys/proc.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/protosw.h> 52 #include <sys/kernel.h> 53 #include <sys/sockio.h> 54 #include <sys/syslog.h> 55 #include <sys/sysctl.h> 56 #include <sys/taskqueue.h> 57 #include <sys/domain.h> 58 #include <sys/jail.h> 59 #include <machine/stdarg.h> 60 61 #include <net/if.h> 62 #include <net/if_clone.h> 63 #include <net/if_dl.h> 64 #include <net/if_types.h> 65 #include <net/if_var.h> 66 #include <net/radix.h> 67 #include <net/route.h> 68 69 #if defined(INET) || defined(INET6) 70 /*XXX*/ 71 #include <netinet/in.h> 72 #include <netinet/in_var.h> 73 #ifdef INET6 74 #include <netinet6/in6_var.h> 75 #include <netinet6/in6_ifattach.h> 76 #endif 77 #endif 78 #ifdef INET 79 #include <netinet/if_ether.h> 80 #endif 81 #ifdef DEV_CARP 82 #include <netinet/ip_carp.h> 83 #endif 84 85 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 86 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 87 88 /* Log link state change events */ 89 static int log_link_state_change = 1; 90 91 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 92 &log_link_state_change, 0, 93 "log interface link state change events"); 94 95 void (*bstp_linkstate_p)(struct ifnet *ifp, int state); 96 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 97 98 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 99 100 static void if_attachdomain(void *); 101 static void if_attachdomain1(struct ifnet *); 102 static int ifconf(u_long, caddr_t); 103 static void if_grow(void); 104 static void if_init(void *); 105 static void if_check(void *); 106 static void if_qflush(struct ifaltq *); 107 static void if_route(struct ifnet *, int flag, int fam); 108 static int if_setflag(struct ifnet *, int, int, int *, int); 109 static void if_slowtimo(void *); 110 static void if_unroute(struct ifnet *, int flag, int fam); 111 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 112 static int if_rtdel(struct radix_node *, void *); 113 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 114 static void if_start_deferred(void *context, int pending); 115 static void do_link_state_change(void *, int); 116 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 117 static int if_getgroupmembers(struct ifgroupreq *); 118 #ifdef INET6 119 /* 120 * XXX: declare here to avoid to include many inet6 related files.. 121 * should be more generalized? 122 */ 123 extern void nd6_setmtu(struct ifnet *); 124 #endif 125 126 int if_index = 0; 127 struct ifindex_entry *ifindex_table = NULL; 128 int ifqmaxlen = IFQ_MAXLEN; 129 struct ifnethead ifnet; /* depend on static init XXX */ 130 struct ifgrouphead ifg_head; 131 struct mtx ifnet_lock; 132 static if_com_alloc_t *if_com_alloc[256]; 133 static if_com_free_t *if_com_free[256]; 134 135 static int if_indexlim = 8; 136 static struct knlist ifklist; 137 138 static void filt_netdetach(struct knote *kn); 139 static int filt_netdev(struct knote *kn, long hint); 140 141 static struct filterops netdev_filtops = 142 { 1, NULL, filt_netdetach, filt_netdev }; 143 144 /* 145 * System initialization 146 */ 147 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL) 148 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL) 149 150 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 151 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 152 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 153 154 static d_open_t netopen; 155 static d_close_t netclose; 156 static d_ioctl_t netioctl; 157 static d_kqfilter_t netkqfilter; 158 159 static struct cdevsw net_cdevsw = { 160 .d_version = D_VERSION, 161 .d_flags = D_NEEDGIANT, 162 .d_open = netopen, 163 .d_close = netclose, 164 .d_ioctl = netioctl, 165 .d_name = "net", 166 .d_kqfilter = netkqfilter, 167 }; 168 169 static int 170 netopen(struct cdev *dev, int flag, int mode, struct thread *td) 171 { 172 return (0); 173 } 174 175 static int 176 netclose(struct cdev *dev, int flags, int fmt, struct thread *td) 177 { 178 return (0); 179 } 180 181 static int 182 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 183 { 184 struct ifnet *ifp; 185 int error, idx; 186 187 /* only support interface specific ioctls */ 188 if (IOCGROUP(cmd) != 'i') 189 return (EOPNOTSUPP); 190 idx = minor(dev); 191 if (idx == 0) { 192 /* 193 * special network device, not interface. 194 */ 195 if (cmd == SIOCGIFCONF) 196 return (ifconf(cmd, data)); /* XXX remove cmd */ 197 #ifdef __amd64__ 198 if (cmd == SIOCGIFCONF32) 199 return (ifconf(cmd, data)); /* XXX remove cmd */ 200 #endif 201 return (EOPNOTSUPP); 202 } 203 204 ifp = ifnet_byindex(idx); 205 if (ifp == NULL) 206 return (ENXIO); 207 208 error = ifhwioctl(cmd, ifp, data, td); 209 if (error == ENOIOCTL) 210 error = EOPNOTSUPP; 211 return (error); 212 } 213 214 static int 215 netkqfilter(struct cdev *dev, struct knote *kn) 216 { 217 struct knlist *klist; 218 struct ifnet *ifp; 219 int idx; 220 221 switch (kn->kn_filter) { 222 case EVFILT_NETDEV: 223 kn->kn_fop = &netdev_filtops; 224 break; 225 default: 226 return (EINVAL); 227 } 228 229 idx = minor(dev); 230 if (idx == 0) { 231 klist = &ifklist; 232 } else { 233 ifp = ifnet_byindex(idx); 234 if (ifp == NULL) 235 return (1); 236 klist = &ifp->if_klist; 237 } 238 239 kn->kn_hook = (caddr_t)klist; 240 241 knlist_add(klist, kn, 0); 242 243 return (0); 244 } 245 246 static void 247 filt_netdetach(struct knote *kn) 248 { 249 struct knlist *klist = (struct knlist *)kn->kn_hook; 250 251 knlist_remove(klist, kn, 0); 252 } 253 254 static int 255 filt_netdev(struct knote *kn, long hint) 256 { 257 struct knlist *klist = (struct knlist *)kn->kn_hook; 258 259 /* 260 * Currently NOTE_EXIT is abused to indicate device detach. 261 */ 262 if (hint == NOTE_EXIT) { 263 kn->kn_data = NOTE_LINKINV; 264 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 265 knlist_remove_inevent(klist, kn); 266 return (1); 267 } 268 if (hint != 0) 269 kn->kn_data = hint; /* current status */ 270 if (kn->kn_sfflags & hint) 271 kn->kn_fflags |= hint; 272 return (kn->kn_fflags != 0); 273 } 274 275 /* 276 * Network interface utility routines. 277 * 278 * Routines with ifa_ifwith* names take sockaddr *'s as 279 * parameters. 280 */ 281 /* ARGSUSED*/ 282 static void 283 if_init(void *dummy __unused) 284 { 285 286 IFNET_LOCK_INIT(); 287 TAILQ_INIT(&ifnet); 288 TAILQ_INIT(&ifg_head); 289 knlist_init(&ifklist, NULL, NULL, NULL, NULL); 290 if_grow(); /* create initial table */ 291 ifdev_byindex(0) = make_dev(&net_cdevsw, 0, 292 UID_ROOT, GID_WHEEL, 0600, "network"); 293 if_clone_init(); 294 } 295 296 static void 297 if_grow(void) 298 { 299 u_int n; 300 struct ifindex_entry *e; 301 302 if_indexlim <<= 1; 303 n = if_indexlim * sizeof(*e); 304 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 305 if (ifindex_table != NULL) { 306 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2); 307 free((caddr_t)ifindex_table, M_IFNET); 308 } 309 ifindex_table = e; 310 } 311 312 /* ARGSUSED*/ 313 static void 314 if_check(void *dummy __unused) 315 { 316 struct ifnet *ifp; 317 int s; 318 319 s = splimp(); 320 IFNET_RLOCK(); /* could sleep on rare error; mostly okay XXX */ 321 TAILQ_FOREACH(ifp, &ifnet, if_link) { 322 if (ifp->if_snd.ifq_maxlen == 0) { 323 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n"); 324 ifp->if_snd.ifq_maxlen = ifqmaxlen; 325 } 326 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) { 327 if_printf(ifp, 328 "XXX: driver didn't initialize queue mtx\n"); 329 mtx_init(&ifp->if_snd.ifq_mtx, "unknown", 330 MTX_NETWORK_LOCK, MTX_DEF); 331 } 332 } 333 IFNET_RUNLOCK(); 334 splx(s); 335 if_slowtimo(0); 336 } 337 338 /* 339 * Allocate a struct ifnet and in index for an interface. 340 */ 341 struct ifnet* 342 if_alloc(u_char type) 343 { 344 struct ifnet *ifp; 345 346 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 347 348 /* 349 * Try to find an empty slot below if_index. If we fail, take 350 * the next slot. 351 * 352 * XXX: should be locked! 353 */ 354 for (ifp->if_index = 1; ifp->if_index <= if_index; ifp->if_index++) { 355 if (ifnet_byindex(ifp->if_index) == NULL) 356 break; 357 } 358 /* Catch if_index overflow. */ 359 if (ifp->if_index < 1) { 360 free(ifp, M_IFNET); 361 return (NULL); 362 } 363 if (ifp->if_index > if_index) 364 if_index = ifp->if_index; 365 if (if_index >= if_indexlim) 366 if_grow(); 367 ifnet_byindex(ifp->if_index) = ifp; 368 369 ifp->if_type = type; 370 371 if (if_com_alloc[type] != NULL) { 372 ifp->if_l2com = if_com_alloc[type](type, ifp); 373 if (ifp->if_l2com == NULL) { 374 free(ifp, M_IFNET); 375 return (NULL); 376 } 377 } 378 IF_ADDR_LOCK_INIT(ifp); 379 380 return (ifp); 381 } 382 383 void 384 if_free(struct ifnet *ifp) 385 { 386 387 /* Do not add code to this function! Add it to if_free_type(). */ 388 if_free_type(ifp, ifp->if_type); 389 } 390 391 void 392 if_free_type(struct ifnet *ifp, u_char type) 393 { 394 395 if (ifp != ifnet_byindex(ifp->if_index)) { 396 if_printf(ifp, "%s: value was not if_alloced, skipping\n", 397 __func__); 398 return; 399 } 400 401 IF_ADDR_LOCK_DESTROY(ifp); 402 403 ifnet_byindex(ifp->if_index) = NULL; 404 405 /* XXX: should be locked with if_findindex() */ 406 while (if_index > 0 && ifnet_byindex(if_index) == NULL) 407 if_index--; 408 409 if (if_com_free[type] != NULL) 410 if_com_free[type](ifp->if_l2com, type); 411 412 free(ifp, M_IFNET); 413 }; 414 415 /* 416 * Attach an interface to the 417 * list of "active" interfaces. 418 */ 419 void 420 if_attach(struct ifnet *ifp) 421 { 422 unsigned socksize, ifasize; 423 int namelen, masklen; 424 struct sockaddr_dl *sdl; 425 struct ifaddr *ifa; 426 427 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 428 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 429 ifp->if_xname); 430 431 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp); 432 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 433 IF_AFDATA_LOCK_INIT(ifp); 434 ifp->if_afdata_initialized = 0; 435 /* 436 * XXX - 437 * The old code would work if the interface passed a pre-existing 438 * chain of ifaddrs to this code. We don't trust our callers to 439 * properly initialize the tailq, however, so we no longer allow 440 * this unlikely case. 441 */ 442 TAILQ_INIT(&ifp->if_addrhead); 443 TAILQ_INIT(&ifp->if_prefixhead); 444 TAILQ_INIT(&ifp->if_multiaddrs); 445 TAILQ_INIT(&ifp->if_groups); 446 447 if_addgroup(ifp, IFG_ALL); 448 449 knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL); 450 getmicrotime(&ifp->if_lastchange); 451 ifp->if_data.ifi_epoch = time_uptime; 452 ifp->if_data.ifi_datalen = sizeof(struct if_data); 453 454 #ifdef MAC 455 mac_init_ifnet(ifp); 456 mac_create_ifnet(ifp); 457 #endif 458 459 ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw, 460 unit2minor(ifp->if_index), 461 UID_ROOT, GID_WHEEL, 0600, "%s/%s", 462 net_cdevsw.d_name, ifp->if_xname); 463 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d", 464 net_cdevsw.d_name, ifp->if_index); 465 466 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 467 468 /* 469 * create a Link Level name for this device 470 */ 471 namelen = strlen(ifp->if_xname); 472 /* 473 * Always save enough space for any possiable name so we can do 474 * a rename in place later. 475 */ 476 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 477 socksize = masklen + ifp->if_addrlen; 478 if (socksize < sizeof(*sdl)) 479 socksize = sizeof(*sdl); 480 socksize = roundup2(socksize, sizeof(long)); 481 ifasize = sizeof(*ifa) + 2 * socksize; 482 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 483 IFA_LOCK_INIT(ifa); 484 sdl = (struct sockaddr_dl *)(ifa + 1); 485 sdl->sdl_len = socksize; 486 sdl->sdl_family = AF_LINK; 487 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 488 sdl->sdl_nlen = namelen; 489 sdl->sdl_index = ifp->if_index; 490 sdl->sdl_type = ifp->if_type; 491 ifp->if_addr = ifa; 492 ifa->ifa_ifp = ifp; 493 ifa->ifa_rtrequest = link_rtrequest; 494 ifa->ifa_addr = (struct sockaddr *)sdl; 495 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 496 ifa->ifa_netmask = (struct sockaddr *)sdl; 497 sdl->sdl_len = masklen; 498 while (namelen != 0) 499 sdl->sdl_data[--namelen] = 0xff; 500 ifa->ifa_refcnt = 1; 501 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 502 ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */ 503 ifp->if_snd.altq_type = 0; 504 ifp->if_snd.altq_disc = NULL; 505 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 506 ifp->if_snd.altq_tbr = NULL; 507 ifp->if_snd.altq_ifp = ifp; 508 509 IFNET_WLOCK(); 510 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link); 511 IFNET_WUNLOCK(); 512 513 if (domain_init_status >= 2) 514 if_attachdomain1(ifp); 515 516 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 517 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 518 519 /* Announce the interface. */ 520 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 521 } 522 523 static void 524 if_attachdomain(void *dummy) 525 { 526 struct ifnet *ifp; 527 int s; 528 529 s = splnet(); 530 TAILQ_FOREACH(ifp, &ifnet, if_link) 531 if_attachdomain1(ifp); 532 splx(s); 533 } 534 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 535 if_attachdomain, NULL); 536 537 static void 538 if_attachdomain1(struct ifnet *ifp) 539 { 540 struct domain *dp; 541 int s; 542 543 s = splnet(); 544 545 /* 546 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 547 * cannot lock ifp->if_afdata initialization, entirely. 548 */ 549 if (IF_AFDATA_TRYLOCK(ifp) == 0) { 550 splx(s); 551 return; 552 } 553 if (ifp->if_afdata_initialized >= domain_init_status) { 554 IF_AFDATA_UNLOCK(ifp); 555 splx(s); 556 printf("if_attachdomain called more than once on %s\n", 557 ifp->if_xname); 558 return; 559 } 560 ifp->if_afdata_initialized = domain_init_status; 561 IF_AFDATA_UNLOCK(ifp); 562 563 /* address family dependent data region */ 564 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 565 for (dp = domains; dp; dp = dp->dom_next) { 566 if (dp->dom_ifattach) 567 ifp->if_afdata[dp->dom_family] = 568 (*dp->dom_ifattach)(ifp); 569 } 570 571 splx(s); 572 } 573 574 /* 575 * Remove any network addresses from an interface. 576 */ 577 578 void 579 if_purgeaddrs(struct ifnet *ifp) 580 { 581 struct ifaddr *ifa, *next; 582 583 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 584 if (ifa->ifa_addr->sa_family == AF_LINK) 585 continue; 586 #ifdef INET 587 /* XXX: Ugly!! ad hoc just for INET */ 588 if (ifa->ifa_addr->sa_family == AF_INET) { 589 struct ifaliasreq ifr; 590 591 bzero(&ifr, sizeof(ifr)); 592 ifr.ifra_addr = *ifa->ifa_addr; 593 if (ifa->ifa_dstaddr) 594 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 595 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 596 NULL) == 0) 597 continue; 598 } 599 #endif /* INET */ 600 #ifdef INET6 601 if (ifa->ifa_addr->sa_family == AF_INET6) { 602 in6_purgeaddr(ifa); 603 /* ifp_addrhead is already updated */ 604 continue; 605 } 606 #endif /* INET6 */ 607 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 608 IFAFREE(ifa); 609 } 610 } 611 612 /* 613 * Detach an interface, removing it from the 614 * list of "active" interfaces. 615 * 616 * XXXRW: There are some significant questions about event ordering, and 617 * how to prevent things from starting to use the interface during detach. 618 */ 619 void 620 if_detach(struct ifnet *ifp) 621 { 622 struct ifaddr *ifa; 623 struct radix_node_head *rnh; 624 int s; 625 int i; 626 struct domain *dp; 627 struct ifnet *iter; 628 int found = 0; 629 630 IFNET_WLOCK(); 631 TAILQ_FOREACH(iter, &ifnet, if_link) 632 if (iter == ifp) { 633 TAILQ_REMOVE(&ifnet, ifp, if_link); 634 found = 1; 635 break; 636 } 637 IFNET_WUNLOCK(); 638 if (!found) 639 return; 640 641 /* 642 * Remove/wait for pending events. 643 */ 644 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 645 646 /* 647 * Remove routes and flush queues. 648 */ 649 s = splnet(); 650 if_down(ifp); 651 #ifdef ALTQ 652 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 653 altq_disable(&ifp->if_snd); 654 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 655 altq_detach(&ifp->if_snd); 656 #endif 657 658 if_purgeaddrs(ifp); 659 660 #ifdef INET 661 in_ifdetach(ifp); 662 #endif 663 664 #ifdef INET6 665 /* 666 * Remove all IPv6 kernel structs related to ifp. This should be done 667 * before removing routing entries below, since IPv6 interface direct 668 * routes are expected to be removed by the IPv6-specific kernel API. 669 * Otherwise, the kernel will detect some inconsistency and bark it. 670 */ 671 in6_ifdetach(ifp); 672 #endif 673 /* 674 * Remove link ifaddr pointer and maybe decrement if_index. 675 * Clean up all addresses. 676 */ 677 ifp->if_addr = NULL; 678 destroy_dev(ifdev_byindex(ifp->if_index)); 679 ifdev_byindex(ifp->if_index) = NULL; 680 681 /* We can now free link ifaddr. */ 682 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 683 ifa = TAILQ_FIRST(&ifp->if_addrhead); 684 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 685 IFAFREE(ifa); 686 } 687 688 /* 689 * Delete all remaining routes using this interface 690 * Unfortuneatly the only way to do this is to slog through 691 * the entire routing table looking for routes which point 692 * to this interface...oh well... 693 */ 694 for (i = 1; i <= AF_MAX; i++) { 695 if ((rnh = rt_tables[i]) == NULL) 696 continue; 697 RADIX_NODE_HEAD_LOCK(rnh); 698 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 699 RADIX_NODE_HEAD_UNLOCK(rnh); 700 } 701 702 /* Announce that the interface is gone. */ 703 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 704 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 705 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 706 707 IF_AFDATA_LOCK(ifp); 708 for (dp = domains; dp; dp = dp->dom_next) { 709 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 710 (*dp->dom_ifdetach)(ifp, 711 ifp->if_afdata[dp->dom_family]); 712 } 713 IF_AFDATA_UNLOCK(ifp); 714 715 #ifdef MAC 716 mac_destroy_ifnet(ifp); 717 #endif /* MAC */ 718 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT); 719 knlist_clear(&ifp->if_klist, 0); 720 knlist_destroy(&ifp->if_klist); 721 mtx_destroy(&ifp->if_snd.ifq_mtx); 722 IF_AFDATA_DESTROY(ifp); 723 splx(s); 724 } 725 726 /* 727 * Add a group to an interface 728 */ 729 int 730 if_addgroup(struct ifnet *ifp, const char *groupname) 731 { 732 struct ifg_list *ifgl; 733 struct ifg_group *ifg = NULL; 734 struct ifg_member *ifgm; 735 736 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 737 groupname[strlen(groupname) - 1] <= '9') 738 return (EINVAL); 739 740 IFNET_WLOCK(); 741 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 742 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 743 IFNET_WUNLOCK(); 744 return (EEXIST); 745 } 746 747 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 748 M_NOWAIT)) == NULL) { 749 IFNET_WUNLOCK(); 750 return (ENOMEM); 751 } 752 753 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 754 M_TEMP, M_NOWAIT)) == NULL) { 755 free(ifgl, M_TEMP); 756 IFNET_WUNLOCK(); 757 return (ENOMEM); 758 } 759 760 TAILQ_FOREACH(ifg, &ifg_head, ifg_next) 761 if (!strcmp(ifg->ifg_group, groupname)) 762 break; 763 764 if (ifg == NULL) { 765 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 766 M_TEMP, M_NOWAIT)) == NULL) { 767 free(ifgl, M_TEMP); 768 free(ifgm, M_TEMP); 769 IFNET_WUNLOCK(); 770 return (ENOMEM); 771 } 772 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 773 ifg->ifg_refcnt = 0; 774 TAILQ_INIT(&ifg->ifg_members); 775 EVENTHANDLER_INVOKE(group_attach_event, ifg); 776 TAILQ_INSERT_TAIL(&ifg_head, ifg, ifg_next); 777 } 778 779 ifg->ifg_refcnt++; 780 ifgl->ifgl_group = ifg; 781 ifgm->ifgm_ifp = ifp; 782 783 IF_ADDR_LOCK(ifp); 784 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 785 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 786 IF_ADDR_UNLOCK(ifp); 787 788 IFNET_WUNLOCK(); 789 790 EVENTHANDLER_INVOKE(group_change_event, groupname); 791 792 return (0); 793 } 794 795 /* 796 * Remove a group from an interface 797 */ 798 int 799 if_delgroup(struct ifnet *ifp, const char *groupname) 800 { 801 struct ifg_list *ifgl; 802 struct ifg_member *ifgm; 803 804 IFNET_WLOCK(); 805 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 806 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 807 break; 808 if (ifgl == NULL) { 809 IFNET_WUNLOCK(); 810 return (ENOENT); 811 } 812 813 IF_ADDR_LOCK(ifp); 814 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 815 IF_ADDR_UNLOCK(ifp); 816 817 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 818 if (ifgm->ifgm_ifp == ifp) 819 break; 820 821 if (ifgm != NULL) { 822 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 823 free(ifgm, M_TEMP); 824 } 825 826 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 827 TAILQ_REMOVE(&ifg_head, ifgl->ifgl_group, ifg_next); 828 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 829 free(ifgl->ifgl_group, M_TEMP); 830 } 831 IFNET_WUNLOCK(); 832 833 free(ifgl, M_TEMP); 834 835 EVENTHANDLER_INVOKE(group_change_event, groupname); 836 837 return (0); 838 } 839 840 /* 841 * Stores all groups from an interface in memory pointed 842 * to by data 843 */ 844 static int 845 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 846 { 847 int len, error; 848 struct ifg_list *ifgl; 849 struct ifg_req ifgrq, *ifgp; 850 struct ifgroupreq *ifgr = data; 851 852 if (ifgr->ifgr_len == 0) { 853 IF_ADDR_LOCK(ifp); 854 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 855 ifgr->ifgr_len += sizeof(struct ifg_req); 856 IF_ADDR_UNLOCK(ifp); 857 return (0); 858 } 859 860 len = ifgr->ifgr_len; 861 ifgp = ifgr->ifgr_groups; 862 /* XXX: wire */ 863 IF_ADDR_LOCK(ifp); 864 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 865 if (len < sizeof(ifgrq)) { 866 IF_ADDR_UNLOCK(ifp); 867 return (EINVAL); 868 } 869 bzero(&ifgrq, sizeof ifgrq); 870 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 871 sizeof(ifgrq.ifgrq_group)); 872 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 873 IF_ADDR_UNLOCK(ifp); 874 return (error); 875 } 876 len -= sizeof(ifgrq); 877 ifgp++; 878 } 879 IF_ADDR_UNLOCK(ifp); 880 881 return (0); 882 } 883 884 /* 885 * Stores all members of a group in memory pointed to by data 886 */ 887 static int 888 if_getgroupmembers(struct ifgroupreq *data) 889 { 890 struct ifgroupreq *ifgr = data; 891 struct ifg_group *ifg; 892 struct ifg_member *ifgm; 893 struct ifg_req ifgrq, *ifgp; 894 int len, error; 895 896 IFNET_RLOCK(); 897 TAILQ_FOREACH(ifg, &ifg_head, ifg_next) 898 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 899 break; 900 if (ifg == NULL) { 901 IFNET_RUNLOCK(); 902 return (ENOENT); 903 } 904 905 if (ifgr->ifgr_len == 0) { 906 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 907 ifgr->ifgr_len += sizeof(ifgrq); 908 IFNET_RUNLOCK(); 909 return (0); 910 } 911 912 len = ifgr->ifgr_len; 913 ifgp = ifgr->ifgr_groups; 914 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 915 if (len < sizeof(ifgrq)) { 916 IFNET_RUNLOCK(); 917 return (EINVAL); 918 } 919 bzero(&ifgrq, sizeof ifgrq); 920 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 921 sizeof(ifgrq.ifgrq_member)); 922 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 923 IFNET_RUNLOCK(); 924 return (error); 925 } 926 len -= sizeof(ifgrq); 927 ifgp++; 928 } 929 IFNET_RUNLOCK(); 930 931 return (0); 932 } 933 934 /* 935 * Delete Routes for a Network Interface 936 * 937 * Called for each routing entry via the rnh->rnh_walktree() call above 938 * to delete all route entries referencing a detaching network interface. 939 * 940 * Arguments: 941 * rn pointer to node in the routing table 942 * arg argument passed to rnh->rnh_walktree() - detaching interface 943 * 944 * Returns: 945 * 0 successful 946 * errno failed - reason indicated 947 * 948 */ 949 static int 950 if_rtdel(struct radix_node *rn, void *arg) 951 { 952 struct rtentry *rt = (struct rtentry *)rn; 953 struct ifnet *ifp = arg; 954 int err; 955 956 if (rt->rt_ifp == ifp) { 957 958 /* 959 * Protect (sorta) against walktree recursion problems 960 * with cloned routes 961 */ 962 if ((rt->rt_flags & RTF_UP) == 0) 963 return (0); 964 965 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 966 rt_mask(rt), rt->rt_flags, 967 (struct rtentry **) NULL); 968 if (err) { 969 log(LOG_WARNING, "if_rtdel: error %d\n", err); 970 } 971 } 972 973 return (0); 974 } 975 976 #define sa_equal(a1, a2) (bcmp((a1), (a2), ((a1))->sa_len) == 0) 977 978 /* 979 * Locate an interface based on a complete address. 980 */ 981 /*ARGSUSED*/ 982 struct ifaddr * 983 ifa_ifwithaddr(struct sockaddr *addr) 984 { 985 struct ifnet *ifp; 986 struct ifaddr *ifa; 987 988 IFNET_RLOCK(); 989 TAILQ_FOREACH(ifp, &ifnet, if_link) 990 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 991 if (ifa->ifa_addr->sa_family != addr->sa_family) 992 continue; 993 if (sa_equal(addr, ifa->ifa_addr)) 994 goto done; 995 /* IP6 doesn't have broadcast */ 996 if ((ifp->if_flags & IFF_BROADCAST) && 997 ifa->ifa_broadaddr && 998 ifa->ifa_broadaddr->sa_len != 0 && 999 sa_equal(ifa->ifa_broadaddr, addr)) 1000 goto done; 1001 } 1002 ifa = NULL; 1003 done: 1004 IFNET_RUNLOCK(); 1005 return (ifa); 1006 } 1007 1008 /* 1009 * Locate an interface based on the broadcast address. 1010 */ 1011 /* ARGSUSED */ 1012 struct ifaddr * 1013 ifa_ifwithbroadaddr(struct sockaddr *addr) 1014 { 1015 struct ifnet *ifp; 1016 struct ifaddr *ifa; 1017 1018 IFNET_RLOCK(); 1019 TAILQ_FOREACH(ifp, &ifnet, if_link) 1020 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1021 if (ifa->ifa_addr->sa_family != addr->sa_family) 1022 continue; 1023 if ((ifp->if_flags & IFF_BROADCAST) && 1024 ifa->ifa_broadaddr && 1025 ifa->ifa_broadaddr->sa_len != 0 && 1026 sa_equal(ifa->ifa_broadaddr, addr)) 1027 goto done; 1028 } 1029 ifa = NULL; 1030 done: 1031 IFNET_RUNLOCK(); 1032 return (ifa); 1033 } 1034 1035 /* 1036 * Locate the point to point interface with a given destination address. 1037 */ 1038 /*ARGSUSED*/ 1039 struct ifaddr * 1040 ifa_ifwithdstaddr(struct sockaddr *addr) 1041 { 1042 struct ifnet *ifp; 1043 struct ifaddr *ifa; 1044 1045 IFNET_RLOCK(); 1046 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1047 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1048 continue; 1049 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1050 if (ifa->ifa_addr->sa_family != addr->sa_family) 1051 continue; 1052 if (ifa->ifa_dstaddr && 1053 sa_equal(addr, ifa->ifa_dstaddr)) 1054 goto done; 1055 } 1056 } 1057 ifa = NULL; 1058 done: 1059 IFNET_RUNLOCK(); 1060 return (ifa); 1061 } 1062 1063 /* 1064 * Find an interface on a specific network. If many, choice 1065 * is most specific found. 1066 */ 1067 struct ifaddr * 1068 ifa_ifwithnet(struct sockaddr *addr) 1069 { 1070 struct ifnet *ifp; 1071 struct ifaddr *ifa; 1072 struct ifaddr *ifa_maybe = (struct ifaddr *) 0; 1073 u_int af = addr->sa_family; 1074 char *addr_data = addr->sa_data, *cplim; 1075 1076 /* 1077 * AF_LINK addresses can be looked up directly by their index number, 1078 * so do that if we can. 1079 */ 1080 if (af == AF_LINK) { 1081 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1082 if (sdl->sdl_index && sdl->sdl_index <= if_index) 1083 return (ifaddr_byindex(sdl->sdl_index)); 1084 } 1085 1086 /* 1087 * Scan though each interface, looking for ones that have 1088 * addresses in this address family. 1089 */ 1090 IFNET_RLOCK(); 1091 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1092 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1093 char *cp, *cp2, *cp3; 1094 1095 if (ifa->ifa_addr->sa_family != af) 1096 next: continue; 1097 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) { 1098 /* 1099 * This is a bit broken as it doesn't 1100 * take into account that the remote end may 1101 * be a single node in the network we are 1102 * looking for. 1103 * The trouble is that we don't know the 1104 * netmask for the remote end. 1105 */ 1106 if (ifa->ifa_dstaddr != 0 && 1107 sa_equal(addr, ifa->ifa_dstaddr)) 1108 goto done; 1109 } else { 1110 /* 1111 * if we have a special address handler, 1112 * then use it instead of the generic one. 1113 */ 1114 if (ifa->ifa_claim_addr) { 1115 if ((*ifa->ifa_claim_addr)(ifa, addr)) 1116 goto done; 1117 continue; 1118 } 1119 1120 /* 1121 * Scan all the bits in the ifa's address. 1122 * If a bit dissagrees with what we are 1123 * looking for, mask it with the netmask 1124 * to see if it really matters. 1125 * (A byte at a time) 1126 */ 1127 if (ifa->ifa_netmask == 0) 1128 continue; 1129 cp = addr_data; 1130 cp2 = ifa->ifa_addr->sa_data; 1131 cp3 = ifa->ifa_netmask->sa_data; 1132 cplim = ifa->ifa_netmask->sa_len 1133 + (char *)ifa->ifa_netmask; 1134 while (cp3 < cplim) 1135 if ((*cp++ ^ *cp2++) & *cp3++) 1136 goto next; /* next address! */ 1137 /* 1138 * If the netmask of what we just found 1139 * is more specific than what we had before 1140 * (if we had one) then remember the new one 1141 * before continuing to search 1142 * for an even better one. 1143 */ 1144 if (ifa_maybe == 0 || 1145 rn_refines((caddr_t)ifa->ifa_netmask, 1146 (caddr_t)ifa_maybe->ifa_netmask)) 1147 ifa_maybe = ifa; 1148 } 1149 } 1150 } 1151 ifa = ifa_maybe; 1152 done: 1153 IFNET_RUNLOCK(); 1154 return (ifa); 1155 } 1156 1157 /* 1158 * Find an interface address specific to an interface best matching 1159 * a given address. 1160 */ 1161 struct ifaddr * 1162 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1163 { 1164 struct ifaddr *ifa; 1165 char *cp, *cp2, *cp3; 1166 char *cplim; 1167 struct ifaddr *ifa_maybe = 0; 1168 u_int af = addr->sa_family; 1169 1170 if (af >= AF_MAX) 1171 return (0); 1172 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1173 if (ifa->ifa_addr->sa_family != af) 1174 continue; 1175 if (ifa_maybe == 0) 1176 ifa_maybe = ifa; 1177 if (ifa->ifa_netmask == 0) { 1178 if (sa_equal(addr, ifa->ifa_addr) || 1179 (ifa->ifa_dstaddr && 1180 sa_equal(addr, ifa->ifa_dstaddr))) 1181 goto done; 1182 continue; 1183 } 1184 if (ifp->if_flags & IFF_POINTOPOINT) { 1185 if (sa_equal(addr, ifa->ifa_dstaddr)) 1186 goto done; 1187 } else { 1188 cp = addr->sa_data; 1189 cp2 = ifa->ifa_addr->sa_data; 1190 cp3 = ifa->ifa_netmask->sa_data; 1191 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1192 for (; cp3 < cplim; cp3++) 1193 if ((*cp++ ^ *cp2++) & *cp3) 1194 break; 1195 if (cp3 == cplim) 1196 goto done; 1197 } 1198 } 1199 ifa = ifa_maybe; 1200 done: 1201 return (ifa); 1202 } 1203 1204 #include <net/route.h> 1205 1206 /* 1207 * Default action when installing a route with a Link Level gateway. 1208 * Lookup an appropriate real ifa to point to. 1209 * This should be moved to /sys/net/link.c eventually. 1210 */ 1211 static void 1212 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 1213 { 1214 struct ifaddr *ifa, *oifa; 1215 struct sockaddr *dst; 1216 struct ifnet *ifp; 1217 1218 RT_LOCK_ASSERT(rt); 1219 1220 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 1221 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 1222 return; 1223 ifa = ifaof_ifpforaddr(dst, ifp); 1224 if (ifa) { 1225 IFAREF(ifa); /* XXX */ 1226 oifa = rt->rt_ifa; 1227 rt->rt_ifa = ifa; 1228 IFAFREE(oifa); 1229 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1230 ifa->ifa_rtrequest(cmd, rt, info); 1231 } 1232 } 1233 1234 /* 1235 * Mark an interface down and notify protocols of 1236 * the transition. 1237 * NOTE: must be called at splnet or eqivalent. 1238 */ 1239 static void 1240 if_unroute(struct ifnet *ifp, int flag, int fam) 1241 { 1242 struct ifaddr *ifa; 1243 1244 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 1245 1246 ifp->if_flags &= ~flag; 1247 getmicrotime(&ifp->if_lastchange); 1248 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1249 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1250 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1251 if_qflush(&ifp->if_snd); 1252 #ifdef DEV_CARP 1253 if (ifp->if_carp) 1254 carp_carpdev_state(ifp->if_carp); 1255 #endif 1256 rt_ifmsg(ifp); 1257 } 1258 1259 /* 1260 * Mark an interface up and notify protocols of 1261 * the transition. 1262 * NOTE: must be called at splnet or eqivalent. 1263 */ 1264 static void 1265 if_route(struct ifnet *ifp, int flag, int fam) 1266 { 1267 struct ifaddr *ifa; 1268 1269 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 1270 1271 ifp->if_flags |= flag; 1272 getmicrotime(&ifp->if_lastchange); 1273 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1274 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1275 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1276 #ifdef DEV_CARP 1277 if (ifp->if_carp) 1278 carp_carpdev_state(ifp->if_carp); 1279 #endif 1280 rt_ifmsg(ifp); 1281 #ifdef INET6 1282 in6_if_up(ifp); 1283 #endif 1284 } 1285 1286 void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */ 1287 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1288 1289 /* 1290 * Handle a change in the interface link state. To avoid LORs 1291 * between driver lock and upper layer locks, as well as possible 1292 * recursions, we post event to taskqueue, and all job 1293 * is done in static do_link_state_change(). 1294 */ 1295 void 1296 if_link_state_change(struct ifnet *ifp, int link_state) 1297 { 1298 /* Return if state hasn't changed. */ 1299 if (ifp->if_link_state == link_state) 1300 return; 1301 1302 ifp->if_link_state = link_state; 1303 1304 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 1305 } 1306 1307 static void 1308 do_link_state_change(void *arg, int pending) 1309 { 1310 struct ifnet *ifp = (struct ifnet *)arg; 1311 int link_state = ifp->if_link_state; 1312 int link; 1313 1314 /* Notify that the link state has changed. */ 1315 rt_ifmsg(ifp); 1316 if (link_state == LINK_STATE_UP) 1317 link = NOTE_LINKUP; 1318 else if (link_state == LINK_STATE_DOWN) 1319 link = NOTE_LINKDOWN; 1320 else 1321 link = NOTE_LINKINV; 1322 KNOTE_UNLOCKED(&ifp->if_klist, link); 1323 if (ifp->if_vlantrunk != NULL) 1324 (*vlan_link_state_p)(ifp, link); 1325 1326 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 1327 IFP2AC(ifp)->ac_netgraph != NULL) 1328 (*ng_ether_link_state_p)(ifp, link_state); 1329 #ifdef DEV_CARP 1330 if (ifp->if_carp) 1331 carp_carpdev_state(ifp->if_carp); 1332 #endif 1333 if (ifp->if_bridge) { 1334 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!")); 1335 (*bstp_linkstate_p)(ifp, link_state); 1336 } 1337 1338 devctl_notify("IFNET", ifp->if_xname, 1339 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL); 1340 if (pending > 1) 1341 if_printf(ifp, "%d link states coalesced\n", pending); 1342 if (log_link_state_change) 1343 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 1344 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 1345 } 1346 1347 /* 1348 * Mark an interface down and notify protocols of 1349 * the transition. 1350 * NOTE: must be called at splnet or eqivalent. 1351 */ 1352 void 1353 if_down(struct ifnet *ifp) 1354 { 1355 1356 if_unroute(ifp, IFF_UP, AF_UNSPEC); 1357 } 1358 1359 /* 1360 * Mark an interface up and notify protocols of 1361 * the transition. 1362 * NOTE: must be called at splnet or eqivalent. 1363 */ 1364 void 1365 if_up(struct ifnet *ifp) 1366 { 1367 1368 if_route(ifp, IFF_UP, AF_UNSPEC); 1369 } 1370 1371 /* 1372 * Flush an interface queue. 1373 */ 1374 static void 1375 if_qflush(struct ifaltq *ifq) 1376 { 1377 struct mbuf *m, *n; 1378 1379 IFQ_LOCK(ifq); 1380 #ifdef ALTQ 1381 if (ALTQ_IS_ENABLED(ifq)) 1382 ALTQ_PURGE(ifq); 1383 #endif 1384 n = ifq->ifq_head; 1385 while ((m = n) != 0) { 1386 n = m->m_act; 1387 m_freem(m); 1388 } 1389 ifq->ifq_head = 0; 1390 ifq->ifq_tail = 0; 1391 ifq->ifq_len = 0; 1392 IFQ_UNLOCK(ifq); 1393 } 1394 1395 /* 1396 * Handle interface watchdog timer routines. Called 1397 * from softclock, we decrement timers (if set) and 1398 * call the appropriate interface routine on expiration. 1399 * 1400 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called 1401 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab 1402 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface. 1403 */ 1404 static void 1405 if_slowtimo(void *arg) 1406 { 1407 struct ifnet *ifp; 1408 int s = splimp(); 1409 1410 IFNET_RLOCK(); 1411 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1412 if (ifp->if_timer == 0 || --ifp->if_timer) 1413 continue; 1414 if (ifp->if_watchdog) 1415 (*ifp->if_watchdog)(ifp); 1416 } 1417 IFNET_RUNLOCK(); 1418 splx(s); 1419 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ); 1420 } 1421 1422 /* 1423 * Map interface name to 1424 * interface structure pointer. 1425 */ 1426 struct ifnet * 1427 ifunit(const char *name) 1428 { 1429 struct ifnet *ifp; 1430 1431 IFNET_RLOCK(); 1432 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1433 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 1434 break; 1435 } 1436 IFNET_RUNLOCK(); 1437 return (ifp); 1438 } 1439 1440 /* 1441 * Hardware specific interface ioctls. 1442 */ 1443 static int 1444 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 1445 { 1446 struct ifreq *ifr; 1447 struct ifstat *ifs; 1448 int error = 0; 1449 int new_flags, temp_flags; 1450 size_t namelen, onamelen; 1451 char new_name[IFNAMSIZ]; 1452 struct ifaddr *ifa; 1453 struct sockaddr_dl *sdl; 1454 1455 ifr = (struct ifreq *)data; 1456 switch (cmd) { 1457 case SIOCGIFINDEX: 1458 ifr->ifr_index = ifp->if_index; 1459 break; 1460 1461 case SIOCGIFFLAGS: 1462 temp_flags = ifp->if_flags | ifp->if_drv_flags; 1463 ifr->ifr_flags = temp_flags & 0xffff; 1464 ifr->ifr_flagshigh = temp_flags >> 16; 1465 break; 1466 1467 case SIOCGIFCAP: 1468 ifr->ifr_reqcap = ifp->if_capabilities; 1469 ifr->ifr_curcap = ifp->if_capenable; 1470 break; 1471 1472 #ifdef MAC 1473 case SIOCGIFMAC: 1474 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp); 1475 break; 1476 #endif 1477 1478 case SIOCGIFMETRIC: 1479 ifr->ifr_metric = ifp->if_metric; 1480 break; 1481 1482 case SIOCGIFMTU: 1483 ifr->ifr_mtu = ifp->if_mtu; 1484 break; 1485 1486 case SIOCGIFPHYS: 1487 ifr->ifr_phys = ifp->if_physical; 1488 break; 1489 1490 case SIOCSIFFLAGS: 1491 error = suser(td); 1492 if (error) 1493 return (error); 1494 /* 1495 * Currently, no driver owned flags pass the IFF_CANTCHANGE 1496 * check, so we don't need special handling here yet. 1497 */ 1498 new_flags = (ifr->ifr_flags & 0xffff) | 1499 (ifr->ifr_flagshigh << 16); 1500 if (ifp->if_flags & IFF_SMART) { 1501 /* Smart drivers twiddle their own routes */ 1502 } else if (ifp->if_flags & IFF_UP && 1503 (new_flags & IFF_UP) == 0) { 1504 int s = splimp(); 1505 if_down(ifp); 1506 splx(s); 1507 } else if (new_flags & IFF_UP && 1508 (ifp->if_flags & IFF_UP) == 0) { 1509 int s = splimp(); 1510 if_up(ifp); 1511 splx(s); 1512 } 1513 /* See if permanently promiscuous mode bit is about to flip */ 1514 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 1515 if (new_flags & IFF_PPROMISC) 1516 ifp->if_flags |= IFF_PROMISC; 1517 else if (ifp->if_pcount == 0) 1518 ifp->if_flags &= ~IFF_PROMISC; 1519 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 1520 ifp->if_xname, 1521 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); 1522 } 1523 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1524 (new_flags &~ IFF_CANTCHANGE); 1525 if (ifp->if_ioctl) { 1526 IFF_LOCKGIANT(ifp); 1527 (void) (*ifp->if_ioctl)(ifp, cmd, data); 1528 IFF_UNLOCKGIANT(ifp); 1529 } 1530 getmicrotime(&ifp->if_lastchange); 1531 break; 1532 1533 case SIOCSIFCAP: 1534 error = suser(td); 1535 if (error) 1536 return (error); 1537 if (ifp->if_ioctl == NULL) 1538 return (EOPNOTSUPP); 1539 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 1540 return (EINVAL); 1541 IFF_LOCKGIANT(ifp); 1542 error = (*ifp->if_ioctl)(ifp, cmd, data); 1543 IFF_UNLOCKGIANT(ifp); 1544 if (error == 0) 1545 getmicrotime(&ifp->if_lastchange); 1546 break; 1547 1548 #ifdef MAC 1549 case SIOCSIFMAC: 1550 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp); 1551 break; 1552 #endif 1553 1554 case SIOCSIFNAME: 1555 error = suser(td); 1556 if (error != 0) 1557 return (error); 1558 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 1559 if (error != 0) 1560 return (error); 1561 if (new_name[0] == '\0') 1562 return (EINVAL); 1563 if (ifunit(new_name) != NULL) 1564 return (EEXIST); 1565 1566 /* Announce the departure of the interface. */ 1567 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1568 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1569 1570 log(LOG_INFO, "%s: changing name to '%s'\n", 1571 ifp->if_xname, new_name); 1572 1573 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 1574 ifa = ifp->if_addr; 1575 IFA_LOCK(ifa); 1576 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1577 namelen = strlen(new_name); 1578 onamelen = sdl->sdl_nlen; 1579 /* 1580 * Move the address if needed. This is safe because we 1581 * allocate space for a name of length IFNAMSIZ when we 1582 * create this in if_attach(). 1583 */ 1584 if (namelen != onamelen) { 1585 bcopy(sdl->sdl_data + onamelen, 1586 sdl->sdl_data + namelen, sdl->sdl_alen); 1587 } 1588 bcopy(new_name, sdl->sdl_data, namelen); 1589 sdl->sdl_nlen = namelen; 1590 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 1591 bzero(sdl->sdl_data, onamelen); 1592 while (namelen != 0) 1593 sdl->sdl_data[--namelen] = 0xff; 1594 IFA_UNLOCK(ifa); 1595 1596 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 1597 /* Announce the return of the interface. */ 1598 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 1599 break; 1600 1601 case SIOCSIFMETRIC: 1602 error = suser(td); 1603 if (error) 1604 return (error); 1605 ifp->if_metric = ifr->ifr_metric; 1606 getmicrotime(&ifp->if_lastchange); 1607 break; 1608 1609 case SIOCSIFPHYS: 1610 error = suser(td); 1611 if (error) 1612 return (error); 1613 if (ifp->if_ioctl == NULL) 1614 return (EOPNOTSUPP); 1615 IFF_LOCKGIANT(ifp); 1616 error = (*ifp->if_ioctl)(ifp, cmd, data); 1617 IFF_UNLOCKGIANT(ifp); 1618 if (error == 0) 1619 getmicrotime(&ifp->if_lastchange); 1620 break; 1621 1622 case SIOCSIFMTU: 1623 { 1624 u_long oldmtu = ifp->if_mtu; 1625 1626 error = suser(td); 1627 if (error) 1628 return (error); 1629 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 1630 return (EINVAL); 1631 if (ifp->if_ioctl == NULL) 1632 return (EOPNOTSUPP); 1633 IFF_LOCKGIANT(ifp); 1634 error = (*ifp->if_ioctl)(ifp, cmd, data); 1635 IFF_UNLOCKGIANT(ifp); 1636 if (error == 0) { 1637 getmicrotime(&ifp->if_lastchange); 1638 rt_ifmsg(ifp); 1639 } 1640 /* 1641 * If the link MTU changed, do network layer specific procedure. 1642 */ 1643 if (ifp->if_mtu != oldmtu) { 1644 #ifdef INET6 1645 nd6_setmtu(ifp); 1646 #endif 1647 } 1648 break; 1649 } 1650 1651 case SIOCADDMULTI: 1652 case SIOCDELMULTI: 1653 error = suser(td); 1654 if (error) 1655 return (error); 1656 1657 /* Don't allow group membership on non-multicast interfaces. */ 1658 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1659 return (EOPNOTSUPP); 1660 1661 /* Don't let users screw up protocols' entries. */ 1662 if (ifr->ifr_addr.sa_family != AF_LINK) 1663 return (EINVAL); 1664 1665 if (cmd == SIOCADDMULTI) { 1666 struct ifmultiaddr *ifma; 1667 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 1668 } else { 1669 error = if_delmulti(ifp, &ifr->ifr_addr); 1670 } 1671 if (error == 0) 1672 getmicrotime(&ifp->if_lastchange); 1673 break; 1674 1675 case SIOCSIFPHYADDR: 1676 case SIOCDIFPHYADDR: 1677 #ifdef INET6 1678 case SIOCSIFPHYADDR_IN6: 1679 #endif 1680 case SIOCSLIFPHYADDR: 1681 case SIOCSIFMEDIA: 1682 case SIOCSIFGENERIC: 1683 error = suser(td); 1684 if (error) 1685 return (error); 1686 if (ifp->if_ioctl == NULL) 1687 return (EOPNOTSUPP); 1688 IFF_LOCKGIANT(ifp); 1689 error = (*ifp->if_ioctl)(ifp, cmd, data); 1690 IFF_UNLOCKGIANT(ifp); 1691 if (error == 0) 1692 getmicrotime(&ifp->if_lastchange); 1693 break; 1694 1695 case SIOCGIFSTATUS: 1696 ifs = (struct ifstat *)data; 1697 ifs->ascii[0] = '\0'; 1698 1699 case SIOCGIFPSRCADDR: 1700 case SIOCGIFPDSTADDR: 1701 case SIOCGLIFPHYADDR: 1702 case SIOCGIFMEDIA: 1703 case SIOCGIFGENERIC: 1704 if (ifp->if_ioctl == NULL) 1705 return (EOPNOTSUPP); 1706 IFF_LOCKGIANT(ifp); 1707 error = (*ifp->if_ioctl)(ifp, cmd, data); 1708 IFF_UNLOCKGIANT(ifp); 1709 break; 1710 1711 case SIOCSIFLLADDR: 1712 error = suser(td); 1713 if (error) 1714 return (error); 1715 error = if_setlladdr(ifp, 1716 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 1717 break; 1718 1719 case SIOCAIFGROUP: 1720 { 1721 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 1722 1723 error = suser(td); 1724 if (error) 1725 return (error); 1726 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 1727 return (error); 1728 break; 1729 } 1730 1731 case SIOCGIFGROUP: 1732 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 1733 return (error); 1734 break; 1735 1736 case SIOCDIFGROUP: 1737 { 1738 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 1739 1740 error = suser(td); 1741 if (error) 1742 return (error); 1743 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 1744 return (error); 1745 break; 1746 } 1747 1748 default: 1749 error = ENOIOCTL; 1750 break; 1751 } 1752 return (error); 1753 } 1754 1755 /* 1756 * Interface ioctls. 1757 */ 1758 int 1759 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 1760 { 1761 struct ifnet *ifp; 1762 struct ifreq *ifr; 1763 int error; 1764 int oif_flags; 1765 1766 switch (cmd) { 1767 case SIOCGIFCONF: 1768 case OSIOCGIFCONF: 1769 #ifdef __amd64__ 1770 case SIOCGIFCONF32: 1771 #endif 1772 return (ifconf(cmd, data)); 1773 } 1774 ifr = (struct ifreq *)data; 1775 1776 switch (cmd) { 1777 case SIOCIFCREATE: 1778 case SIOCIFCREATE2: 1779 if ((error = suser(td)) != 0) 1780 return (error); 1781 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name), 1782 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL)); 1783 case SIOCIFDESTROY: 1784 if ((error = suser(td)) != 0) 1785 return (error); 1786 return if_clone_destroy(ifr->ifr_name); 1787 1788 case SIOCIFGCLONERS: 1789 return (if_clone_list((struct if_clonereq *)data)); 1790 case SIOCGIFGMEMB: 1791 return (if_getgroupmembers((struct ifgroupreq *)data)); 1792 } 1793 1794 ifp = ifunit(ifr->ifr_name); 1795 if (ifp == 0) 1796 return (ENXIO); 1797 1798 error = ifhwioctl(cmd, ifp, data, td); 1799 if (error != ENOIOCTL) 1800 return (error); 1801 1802 oif_flags = ifp->if_flags; 1803 if (so->so_proto == 0) 1804 return (EOPNOTSUPP); 1805 #ifndef COMPAT_43 1806 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 1807 data, 1808 ifp, td)); 1809 #else 1810 { 1811 int ocmd = cmd; 1812 1813 switch (cmd) { 1814 1815 case SIOCSIFDSTADDR: 1816 case SIOCSIFADDR: 1817 case SIOCSIFBRDADDR: 1818 case SIOCSIFNETMASK: 1819 #if BYTE_ORDER != BIG_ENDIAN 1820 if (ifr->ifr_addr.sa_family == 0 && 1821 ifr->ifr_addr.sa_len < 16) { 1822 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 1823 ifr->ifr_addr.sa_len = 16; 1824 } 1825 #else 1826 if (ifr->ifr_addr.sa_len == 0) 1827 ifr->ifr_addr.sa_len = 16; 1828 #endif 1829 break; 1830 1831 case OSIOCGIFADDR: 1832 cmd = SIOCGIFADDR; 1833 break; 1834 1835 case OSIOCGIFDSTADDR: 1836 cmd = SIOCGIFDSTADDR; 1837 break; 1838 1839 case OSIOCGIFBRDADDR: 1840 cmd = SIOCGIFBRDADDR; 1841 break; 1842 1843 case OSIOCGIFNETMASK: 1844 cmd = SIOCGIFNETMASK; 1845 } 1846 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 1847 cmd, 1848 data, 1849 ifp, td)); 1850 switch (ocmd) { 1851 1852 case OSIOCGIFADDR: 1853 case OSIOCGIFDSTADDR: 1854 case OSIOCGIFBRDADDR: 1855 case OSIOCGIFNETMASK: 1856 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 1857 1858 } 1859 } 1860 #endif /* COMPAT_43 */ 1861 1862 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 1863 #ifdef INET6 1864 DELAY(100);/* XXX: temporary workaround for fxp issue*/ 1865 if (ifp->if_flags & IFF_UP) { 1866 int s = splimp(); 1867 in6_if_up(ifp); 1868 splx(s); 1869 } 1870 #endif 1871 } 1872 return (error); 1873 } 1874 1875 /* 1876 * The code common to handling reference counted flags, 1877 * e.g., in ifpromisc() and if_allmulti(). 1878 * The "pflag" argument can specify a permanent mode flag to check, 1879 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 1880 * 1881 * Only to be used on stack-owned flags, not driver-owned flags. 1882 */ 1883 static int 1884 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 1885 { 1886 struct ifreq ifr; 1887 int error; 1888 int oldflags, oldcount; 1889 1890 /* Sanity checks to catch programming errors */ 1891 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 1892 ("%s: setting driver-owned flag %d", __func__, flag)); 1893 1894 if (onswitch) 1895 KASSERT(*refcount >= 0, 1896 ("%s: increment negative refcount %d for flag %d", 1897 __func__, *refcount, flag)); 1898 else 1899 KASSERT(*refcount > 0, 1900 ("%s: decrement non-positive refcount %d for flag %d", 1901 __func__, *refcount, flag)); 1902 1903 /* In case this mode is permanent, just touch refcount */ 1904 if (ifp->if_flags & pflag) { 1905 *refcount += onswitch ? 1 : -1; 1906 return (0); 1907 } 1908 1909 /* Save ifnet parameters for if_ioctl() may fail */ 1910 oldcount = *refcount; 1911 oldflags = ifp->if_flags; 1912 1913 /* 1914 * See if we aren't the only and touching refcount is enough. 1915 * Actually toggle interface flag if we are the first or last. 1916 */ 1917 if (onswitch) { 1918 if ((*refcount)++) 1919 return (0); 1920 ifp->if_flags |= flag; 1921 } else { 1922 if (--(*refcount)) 1923 return (0); 1924 ifp->if_flags &= ~flag; 1925 } 1926 1927 /* Call down the driver since we've changed interface flags */ 1928 if (ifp->if_ioctl == NULL) { 1929 error = EOPNOTSUPP; 1930 goto recover; 1931 } 1932 ifr.ifr_flags = ifp->if_flags & 0xffff; 1933 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1934 IFF_LOCKGIANT(ifp); 1935 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1936 IFF_UNLOCKGIANT(ifp); 1937 if (error) 1938 goto recover; 1939 /* Notify userland that interface flags have changed */ 1940 rt_ifmsg(ifp); 1941 return (0); 1942 1943 recover: 1944 /* Recover after driver error */ 1945 *refcount = oldcount; 1946 ifp->if_flags = oldflags; 1947 return (error); 1948 } 1949 1950 /* 1951 * Set/clear promiscuous mode on interface ifp based on the truth value 1952 * of pswitch. The calls are reference counted so that only the first 1953 * "on" request actually has an effect, as does the final "off" request. 1954 * Results are undefined if the "off" and "on" requests are not matched. 1955 */ 1956 int 1957 ifpromisc(struct ifnet *ifp, int pswitch) 1958 { 1959 int error; 1960 int oldflags = ifp->if_flags; 1961 1962 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 1963 &ifp->if_pcount, pswitch); 1964 /* If promiscuous mode status has changed, log a message */ 1965 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) 1966 log(LOG_INFO, "%s: promiscuous mode %s\n", 1967 ifp->if_xname, 1968 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 1969 return (error); 1970 } 1971 1972 /* 1973 * Return interface configuration 1974 * of system. List may be used 1975 * in later ioctl's (above) to get 1976 * other information. 1977 */ 1978 /*ARGSUSED*/ 1979 static int 1980 ifconf(u_long cmd, caddr_t data) 1981 { 1982 struct ifconf *ifc = (struct ifconf *)data; 1983 #ifdef __amd64__ 1984 struct ifconf32 *ifc32 = (struct ifconf32 *)data; 1985 struct ifconf ifc_swab; 1986 #endif 1987 struct ifnet *ifp; 1988 struct ifaddr *ifa; 1989 struct ifreq ifr; 1990 struct sbuf *sb; 1991 int error, full = 0, valid_len, max_len; 1992 1993 #ifdef __amd64__ 1994 if (cmd == SIOCGIFCONF32) { 1995 ifc_swab.ifc_len = ifc32->ifc_len; 1996 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf; 1997 ifc = &ifc_swab; 1998 } 1999 #endif 2000 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2001 max_len = MAXPHYS - 1; 2002 2003 /* Prevent hostile input from being able to crash the system */ 2004 if (ifc->ifc_len <= 0) 2005 return (EINVAL); 2006 2007 again: 2008 if (ifc->ifc_len <= max_len) { 2009 max_len = ifc->ifc_len; 2010 full = 1; 2011 } 2012 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2013 max_len = 0; 2014 valid_len = 0; 2015 2016 IFNET_RLOCK(); /* could sleep XXX */ 2017 TAILQ_FOREACH(ifp, &ifnet, if_link) { 2018 int addrs; 2019 2020 /* 2021 * Zero the ifr_name buffer to make sure we don't 2022 * disclose the contents of the stack. 2023 */ 2024 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 2025 2026 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 2027 >= sizeof(ifr.ifr_name)) { 2028 sbuf_delete(sb); 2029 IFNET_RUNLOCK(); 2030 return (ENAMETOOLONG); 2031 } 2032 2033 addrs = 0; 2034 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2035 struct sockaddr *sa = ifa->ifa_addr; 2036 2037 if (jailed(curthread->td_ucred) && 2038 prison_if(curthread->td_ucred, sa)) 2039 continue; 2040 addrs++; 2041 #ifdef COMPAT_43 2042 if (cmd == OSIOCGIFCONF) { 2043 struct osockaddr *osa = 2044 (struct osockaddr *)&ifr.ifr_addr; 2045 ifr.ifr_addr = *sa; 2046 osa->sa_family = sa->sa_family; 2047 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2048 max_len += sizeof(ifr); 2049 } else 2050 #endif 2051 if (sa->sa_len <= sizeof(*sa)) { 2052 ifr.ifr_addr = *sa; 2053 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2054 max_len += sizeof(ifr); 2055 } else { 2056 sbuf_bcat(sb, &ifr, 2057 offsetof(struct ifreq, ifr_addr)); 2058 max_len += offsetof(struct ifreq, ifr_addr); 2059 sbuf_bcat(sb, sa, sa->sa_len); 2060 max_len += sa->sa_len; 2061 } 2062 2063 if (!sbuf_overflowed(sb)) 2064 valid_len = sbuf_len(sb); 2065 } 2066 if (addrs == 0) { 2067 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 2068 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2069 max_len += sizeof(ifr); 2070 2071 if (!sbuf_overflowed(sb)) 2072 valid_len = sbuf_len(sb); 2073 } 2074 } 2075 IFNET_RUNLOCK(); 2076 2077 /* 2078 * If we didn't allocate enough space (uncommon), try again. If 2079 * we have already allocated as much space as we are allowed, 2080 * return what we've got. 2081 */ 2082 if (valid_len != max_len && !full) { 2083 sbuf_delete(sb); 2084 goto again; 2085 } 2086 2087 ifc->ifc_len = valid_len; 2088 #ifdef __amd64__ 2089 if (cmd == SIOCGIFCONF32) 2090 ifc32->ifc_len = valid_len; 2091 #endif 2092 sbuf_finish(sb); 2093 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 2094 sbuf_delete(sb); 2095 return (error); 2096 } 2097 2098 /* 2099 * Just like ifpromisc(), but for all-multicast-reception mode. 2100 */ 2101 int 2102 if_allmulti(struct ifnet *ifp, int onswitch) 2103 { 2104 2105 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 2106 } 2107 2108 static struct ifmultiaddr * 2109 if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 2110 { 2111 struct ifmultiaddr *ifma; 2112 2113 IF_ADDR_LOCK_ASSERT(ifp); 2114 2115 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2116 if (sa_equal(ifma->ifma_addr, sa)) 2117 break; 2118 } 2119 2120 return ifma; 2121 } 2122 2123 /* 2124 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 2125 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 2126 * the ifnet multicast address list here, so the caller must do that and 2127 * other setup work (such as notifying the device driver). The reference 2128 * count is initialized to 1. 2129 */ 2130 static struct ifmultiaddr * 2131 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 2132 int mflags) 2133 { 2134 struct ifmultiaddr *ifma; 2135 struct sockaddr *dupsa; 2136 2137 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags | 2138 M_ZERO); 2139 if (ifma == NULL) 2140 return (NULL); 2141 2142 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags); 2143 if (dupsa == NULL) { 2144 FREE(ifma, M_IFMADDR); 2145 return (NULL); 2146 } 2147 bcopy(sa, dupsa, sa->sa_len); 2148 ifma->ifma_addr = dupsa; 2149 2150 ifma->ifma_ifp = ifp; 2151 ifma->ifma_refcount = 1; 2152 ifma->ifma_protospec = NULL; 2153 2154 if (llsa == NULL) { 2155 ifma->ifma_lladdr = NULL; 2156 return (ifma); 2157 } 2158 2159 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags); 2160 if (dupsa == NULL) { 2161 FREE(ifma->ifma_addr, M_IFMADDR); 2162 FREE(ifma, M_IFMADDR); 2163 return (NULL); 2164 } 2165 bcopy(llsa, dupsa, llsa->sa_len); 2166 ifma->ifma_lladdr = dupsa; 2167 2168 return (ifma); 2169 } 2170 2171 /* 2172 * if_freemulti: free ifmultiaddr structure and possibly attached related 2173 * addresses. The caller is responsible for implementing reference 2174 * counting, notifying the driver, handling routing messages, and releasing 2175 * any dependent link layer state. 2176 */ 2177 static void 2178 if_freemulti(struct ifmultiaddr *ifma) 2179 { 2180 2181 KASSERT(ifma->ifma_refcount == 1, ("if_freemulti: refcount %d", 2182 ifma->ifma_refcount)); 2183 KASSERT(ifma->ifma_protospec == NULL, 2184 ("if_freemulti: protospec not NULL")); 2185 2186 if (ifma->ifma_lladdr != NULL) 2187 FREE(ifma->ifma_lladdr, M_IFMADDR); 2188 FREE(ifma->ifma_addr, M_IFMADDR); 2189 FREE(ifma, M_IFMADDR); 2190 } 2191 2192 /* 2193 * Register an additional multicast address with a network interface. 2194 * 2195 * - If the address is already present, bump the reference count on the 2196 * address and return. 2197 * - If the address is not link-layer, look up a link layer address. 2198 * - Allocate address structures for one or both addresses, and attach to the 2199 * multicast address list on the interface. If automatically adding a link 2200 * layer address, the protocol address will own a reference to the link 2201 * layer address, to be freed when it is freed. 2202 * - Notify the network device driver of an addition to the multicast address 2203 * list. 2204 * 2205 * 'sa' points to caller-owned memory with the desired multicast address. 2206 * 2207 * 'retifma' will be used to return a pointer to the resulting multicast 2208 * address reference, if desired. 2209 */ 2210 int 2211 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 2212 struct ifmultiaddr **retifma) 2213 { 2214 struct ifmultiaddr *ifma, *ll_ifma; 2215 struct sockaddr *llsa; 2216 int error; 2217 2218 /* 2219 * If the address is already present, return a new reference to it; 2220 * otherwise, allocate storage and set up a new address. 2221 */ 2222 IF_ADDR_LOCK(ifp); 2223 ifma = if_findmulti(ifp, sa); 2224 if (ifma != NULL) { 2225 ifma->ifma_refcount++; 2226 if (retifma != NULL) 2227 *retifma = ifma; 2228 IF_ADDR_UNLOCK(ifp); 2229 return (0); 2230 } 2231 2232 /* 2233 * The address isn't already present; resolve the protocol address 2234 * into a link layer address, and then look that up, bump its 2235 * refcount or allocate an ifma for that also. If 'llsa' was 2236 * returned, we will need to free it later. 2237 */ 2238 llsa = NULL; 2239 ll_ifma = NULL; 2240 if (ifp->if_resolvemulti != NULL) { 2241 error = ifp->if_resolvemulti(ifp, &llsa, sa); 2242 if (error) 2243 goto unlock_out; 2244 } 2245 2246 /* 2247 * Allocate the new address. Don't hook it up yet, as we may also 2248 * need to allocate a link layer multicast address. 2249 */ 2250 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 2251 if (ifma == NULL) { 2252 error = ENOMEM; 2253 goto free_llsa_out; 2254 } 2255 2256 /* 2257 * If a link layer address is found, we'll need to see if it's 2258 * already present in the address list, or allocate is as well. 2259 * When this block finishes, the link layer address will be on the 2260 * list. 2261 */ 2262 if (llsa != NULL) { 2263 ll_ifma = if_findmulti(ifp, llsa); 2264 if (ll_ifma == NULL) { 2265 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 2266 if (ll_ifma == NULL) { 2267 if_freemulti(ifma); 2268 error = ENOMEM; 2269 goto free_llsa_out; 2270 } 2271 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 2272 ifma_link); 2273 } else 2274 ll_ifma->ifma_refcount++; 2275 } 2276 2277 /* 2278 * We now have a new multicast address, ifma, and possibly a new or 2279 * referenced link layer address. Add the primary address to the 2280 * ifnet address list. 2281 */ 2282 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 2283 2284 if (retifma != NULL) 2285 *retifma = ifma; 2286 2287 /* 2288 * Must generate the message while holding the lock so that 'ifma' 2289 * pointer is still valid. 2290 * 2291 * XXXRW: How come we don't announce ll_ifma? 2292 */ 2293 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 2294 IF_ADDR_UNLOCK(ifp); 2295 2296 /* 2297 * We are certain we have added something, so call down to the 2298 * interface to let them know about it. 2299 */ 2300 if (ifp->if_ioctl != NULL) { 2301 IFF_LOCKGIANT(ifp); 2302 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 2303 IFF_UNLOCKGIANT(ifp); 2304 } 2305 2306 if (llsa != NULL) 2307 FREE(llsa, M_IFMADDR); 2308 2309 return (0); 2310 2311 free_llsa_out: 2312 if (llsa != NULL) 2313 FREE(llsa, M_IFMADDR); 2314 2315 unlock_out: 2316 IF_ADDR_UNLOCK(ifp); 2317 return (error); 2318 } 2319 2320 /* 2321 * Remove a reference to a multicast address on this interface. Yell 2322 * if the request does not match an existing membership. 2323 */ 2324 int 2325 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 2326 { 2327 struct ifmultiaddr *ifma, *ll_ifma; 2328 2329 IF_ADDR_LOCK(ifp); 2330 ifma = if_findmulti(ifp, sa); 2331 if (ifma == NULL) { 2332 IF_ADDR_UNLOCK(ifp); 2333 return ENOENT; 2334 } 2335 2336 if (ifma->ifma_refcount > 1) { 2337 ifma->ifma_refcount--; 2338 IF_ADDR_UNLOCK(ifp); 2339 return 0; 2340 } 2341 2342 sa = ifma->ifma_lladdr; 2343 if (sa != NULL) 2344 ll_ifma = if_findmulti(ifp, sa); 2345 else 2346 ll_ifma = NULL; 2347 2348 /* 2349 * XXXRW: How come we don't announce ll_ifma? 2350 */ 2351 rt_newmaddrmsg(RTM_DELMADDR, ifma); 2352 2353 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 2354 if_freemulti(ifma); 2355 2356 if (ll_ifma != NULL) { 2357 if (ll_ifma->ifma_refcount == 1) { 2358 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifma_link); 2359 if_freemulti(ll_ifma); 2360 } else 2361 ll_ifma->ifma_refcount--; 2362 } 2363 IF_ADDR_UNLOCK(ifp); 2364 2365 /* 2366 * Make sure the interface driver is notified 2367 * in the case of a link layer mcast group being left. 2368 */ 2369 if (ifp->if_ioctl) { 2370 IFF_LOCKGIANT(ifp); 2371 (void) (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 2372 IFF_UNLOCKGIANT(ifp); 2373 } 2374 2375 return 0; 2376 } 2377 2378 /* 2379 * Set the link layer address on an interface. 2380 * 2381 * At this time we only support certain types of interfaces, 2382 * and we don't allow the length of the address to change. 2383 */ 2384 int 2385 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 2386 { 2387 struct sockaddr_dl *sdl; 2388 struct ifaddr *ifa; 2389 struct ifreq ifr; 2390 2391 ifa = ifp->if_addr; 2392 if (ifa == NULL) 2393 return (EINVAL); 2394 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2395 if (sdl == NULL) 2396 return (EINVAL); 2397 if (len != sdl->sdl_alen) /* don't allow length to change */ 2398 return (EINVAL); 2399 switch (ifp->if_type) { 2400 case IFT_ETHER: 2401 case IFT_FDDI: 2402 case IFT_XETHER: 2403 case IFT_ISO88025: 2404 case IFT_L2VLAN: 2405 case IFT_BRIDGE: 2406 case IFT_ARCNET: 2407 bcopy(lladdr, LLADDR(sdl), len); 2408 break; 2409 default: 2410 return (ENODEV); 2411 } 2412 /* 2413 * If the interface is already up, we need 2414 * to re-init it in order to reprogram its 2415 * address filter. 2416 */ 2417 if ((ifp->if_flags & IFF_UP) != 0) { 2418 if (ifp->if_ioctl) { 2419 IFF_LOCKGIANT(ifp); 2420 ifp->if_flags &= ~IFF_UP; 2421 ifr.ifr_flags = ifp->if_flags & 0xffff; 2422 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2423 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2424 ifp->if_flags |= IFF_UP; 2425 ifr.ifr_flags = ifp->if_flags & 0xffff; 2426 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2427 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2428 IFF_UNLOCKGIANT(ifp); 2429 } 2430 #ifdef INET 2431 /* 2432 * Also send gratuitous ARPs to notify other nodes about 2433 * the address change. 2434 */ 2435 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2436 if (ifa->ifa_addr->sa_family == AF_INET) 2437 arp_ifinit(ifp, ifa); 2438 } 2439 #endif 2440 } 2441 return (0); 2442 } 2443 2444 /* 2445 * The name argument must be a pointer to storage which will last as 2446 * long as the interface does. For physical devices, the result of 2447 * device_get_name(dev) is a good choice and for pseudo-devices a 2448 * static string works well. 2449 */ 2450 void 2451 if_initname(struct ifnet *ifp, const char *name, int unit) 2452 { 2453 ifp->if_dname = name; 2454 ifp->if_dunit = unit; 2455 if (unit != IF_DUNIT_NONE) 2456 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 2457 else 2458 strlcpy(ifp->if_xname, name, IFNAMSIZ); 2459 } 2460 2461 int 2462 if_printf(struct ifnet *ifp, const char * fmt, ...) 2463 { 2464 va_list ap; 2465 int retval; 2466 2467 retval = printf("%s: ", ifp->if_xname); 2468 va_start(ap, fmt); 2469 retval += vprintf(fmt, ap); 2470 va_end(ap); 2471 return (retval); 2472 } 2473 2474 /* 2475 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot 2476 * be called without Giant. However, we often can't acquire the Giant lock 2477 * at those points; instead, we run it via a task queue that holds Giant via 2478 * if_start_deferred. 2479 * 2480 * XXXRW: We need to make sure that the ifnet isn't fully detached until any 2481 * outstanding if_start_deferred() tasks that will run after the free. This 2482 * probably means waiting in if_detach(). 2483 */ 2484 void 2485 if_start(struct ifnet *ifp) 2486 { 2487 2488 NET_ASSERT_GIANT(); 2489 2490 if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) { 2491 if (mtx_owned(&Giant)) 2492 (*(ifp)->if_start)(ifp); 2493 else 2494 taskqueue_enqueue(taskqueue_swi_giant, 2495 &ifp->if_starttask); 2496 } else 2497 (*(ifp)->if_start)(ifp); 2498 } 2499 2500 static void 2501 if_start_deferred(void *context, int pending) 2502 { 2503 struct ifnet *ifp; 2504 2505 /* 2506 * This code must be entered with Giant, and should never run if 2507 * we're not running with debug.mpsafenet. 2508 */ 2509 KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet")); 2510 GIANT_REQUIRED; 2511 2512 ifp = context; 2513 (ifp->if_start)(ifp); 2514 } 2515 2516 int 2517 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 2518 { 2519 int active = 0; 2520 2521 IF_LOCK(ifq); 2522 if (_IF_QFULL(ifq)) { 2523 _IF_DROP(ifq); 2524 IF_UNLOCK(ifq); 2525 m_freem(m); 2526 return (0); 2527 } 2528 if (ifp != NULL) { 2529 ifp->if_obytes += m->m_pkthdr.len + adjust; 2530 if (m->m_flags & (M_BCAST|M_MCAST)) 2531 ifp->if_omcasts++; 2532 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 2533 } 2534 _IF_ENQUEUE(ifq, m); 2535 IF_UNLOCK(ifq); 2536 if (ifp != NULL && !active) 2537 if_start(ifp); 2538 return (1); 2539 } 2540 2541 void 2542 if_register_com_alloc(u_char type, 2543 if_com_alloc_t *a, if_com_free_t *f) 2544 { 2545 2546 KASSERT(if_com_alloc[type] == NULL, 2547 ("if_register_com_alloc: %d already registered", type)); 2548 KASSERT(if_com_free[type] == NULL, 2549 ("if_register_com_alloc: %d free already registered", type)); 2550 2551 if_com_alloc[type] = a; 2552 if_com_free[type] = f; 2553 } 2554 2555 void 2556 if_deregister_com_alloc(u_char type) 2557 { 2558 2559 KASSERT(if_com_alloc[type] != NULL, 2560 ("if_deregister_com_alloc: %d not registered", type)); 2561 KASSERT(if_com_free[type] != NULL, 2562 ("if_deregister_com_alloc: %d free not registered", type)); 2563 if_com_alloc[type] = NULL; 2564 if_com_free[type] = NULL; 2565 } 2566