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/malloc.h> 43 #include <sys/sbuf.h> 44 #include <sys/bus.h> 45 #include <sys/mbuf.h> 46 #include <sys/systm.h> 47 #include <sys/priv.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 #include <security/mac/mac_framework.h> 86 87 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 88 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 89 90 /* Log link state change events */ 91 static int log_link_state_change = 1; 92 93 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 94 &log_link_state_change, 0, 95 "log interface link state change events"); 96 97 void (*bstp_linkstate_p)(struct ifnet *ifp, int state); 98 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 99 100 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 101 102 static void if_attachdomain(void *); 103 static void if_attachdomain1(struct ifnet *); 104 static int ifconf(u_long, caddr_t); 105 static void if_grow(void); 106 static void if_init(void *); 107 static void if_check(void *); 108 static void if_qflush(struct ifaltq *); 109 static void if_route(struct ifnet *, int flag, int fam); 110 static int if_setflag(struct ifnet *, int, int, int *, int); 111 static void if_slowtimo(void *); 112 static void if_unroute(struct ifnet *, int flag, int fam); 113 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 114 static int if_rtdel(struct radix_node *, void *); 115 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 116 static void if_start_deferred(void *context, int pending); 117 static void do_link_state_change(void *, int); 118 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 119 static int if_getgroupmembers(struct ifgroupreq *); 120 #ifdef INET6 121 /* 122 * XXX: declare here to avoid to include many inet6 related files.. 123 * should be more generalized? 124 */ 125 extern void nd6_setmtu(struct ifnet *); 126 #endif 127 128 int if_index = 0; 129 struct ifindex_entry *ifindex_table = NULL; 130 int ifqmaxlen = IFQ_MAXLEN; 131 struct ifnethead ifnet; /* depend on static init XXX */ 132 struct ifgrouphead ifg_head; 133 struct mtx ifnet_lock; 134 static if_com_alloc_t *if_com_alloc[256]; 135 static if_com_free_t *if_com_free[256]; 136 137 static int if_indexlim = 8; 138 static struct knlist ifklist; 139 140 static void filt_netdetach(struct knote *kn); 141 static int filt_netdev(struct knote *kn, long hint); 142 143 static struct filterops netdev_filtops = 144 { 1, NULL, filt_netdetach, filt_netdev }; 145 146 /* 147 * System initialization 148 */ 149 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL) 150 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL) 151 152 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 153 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 154 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 155 156 static d_open_t netopen; 157 static d_close_t netclose; 158 static d_ioctl_t netioctl; 159 static d_kqfilter_t netkqfilter; 160 161 static struct cdevsw net_cdevsw = { 162 .d_version = D_VERSION, 163 .d_flags = D_NEEDGIANT, 164 .d_open = netopen, 165 .d_close = netclose, 166 .d_ioctl = netioctl, 167 .d_name = "net", 168 .d_kqfilter = netkqfilter, 169 }; 170 171 static int 172 netopen(struct cdev *dev, int flag, int mode, struct thread *td) 173 { 174 return (0); 175 } 176 177 static int 178 netclose(struct cdev *dev, int flags, int fmt, struct thread *td) 179 { 180 return (0); 181 } 182 183 static int 184 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 185 { 186 struct ifnet *ifp; 187 int error, idx; 188 189 /* only support interface specific ioctls */ 190 if (IOCGROUP(cmd) != 'i') 191 return (EOPNOTSUPP); 192 idx = minor(dev); 193 if (idx == 0) { 194 /* 195 * special network device, not interface. 196 */ 197 if (cmd == SIOCGIFCONF) 198 return (ifconf(cmd, data)); /* XXX remove cmd */ 199 #ifdef __amd64__ 200 if (cmd == SIOCGIFCONF32) 201 return (ifconf(cmd, data)); /* XXX remove cmd */ 202 #endif 203 return (EOPNOTSUPP); 204 } 205 206 ifp = ifnet_byindex(idx); 207 if (ifp == NULL) 208 return (ENXIO); 209 210 error = ifhwioctl(cmd, ifp, data, td); 211 if (error == ENOIOCTL) 212 error = EOPNOTSUPP; 213 return (error); 214 } 215 216 static int 217 netkqfilter(struct cdev *dev, struct knote *kn) 218 { 219 struct knlist *klist; 220 struct ifnet *ifp; 221 int idx; 222 223 switch (kn->kn_filter) { 224 case EVFILT_NETDEV: 225 kn->kn_fop = &netdev_filtops; 226 break; 227 default: 228 return (EINVAL); 229 } 230 231 idx = minor(dev); 232 if (idx == 0) { 233 klist = &ifklist; 234 } else { 235 ifp = ifnet_byindex(idx); 236 if (ifp == NULL) 237 return (1); 238 klist = &ifp->if_klist; 239 } 240 241 kn->kn_hook = (caddr_t)klist; 242 243 knlist_add(klist, kn, 0); 244 245 return (0); 246 } 247 248 static void 249 filt_netdetach(struct knote *kn) 250 { 251 struct knlist *klist = (struct knlist *)kn->kn_hook; 252 253 knlist_remove(klist, kn, 0); 254 } 255 256 static int 257 filt_netdev(struct knote *kn, long hint) 258 { 259 struct knlist *klist = (struct knlist *)kn->kn_hook; 260 261 /* 262 * Currently NOTE_EXIT is abused to indicate device detach. 263 */ 264 if (hint == NOTE_EXIT) { 265 kn->kn_data = NOTE_LINKINV; 266 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 267 knlist_remove_inevent(klist, kn); 268 return (1); 269 } 270 if (hint != 0) 271 kn->kn_data = hint; /* current status */ 272 if (kn->kn_sfflags & hint) 273 kn->kn_fflags |= hint; 274 return (kn->kn_fflags != 0); 275 } 276 277 /* 278 * Network interface utility routines. 279 * 280 * Routines with ifa_ifwith* names take sockaddr *'s as 281 * parameters. 282 */ 283 /* ARGSUSED*/ 284 static void 285 if_init(void *dummy __unused) 286 { 287 288 IFNET_LOCK_INIT(); 289 TAILQ_INIT(&ifnet); 290 TAILQ_INIT(&ifg_head); 291 knlist_init(&ifklist, NULL, NULL, NULL, NULL); 292 if_grow(); /* create initial table */ 293 ifdev_byindex(0) = make_dev(&net_cdevsw, 0, 294 UID_ROOT, GID_WHEEL, 0600, "network"); 295 if_clone_init(); 296 } 297 298 static void 299 if_grow(void) 300 { 301 u_int n; 302 struct ifindex_entry *e; 303 304 if_indexlim <<= 1; 305 n = if_indexlim * sizeof(*e); 306 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 307 if (ifindex_table != NULL) { 308 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2); 309 free((caddr_t)ifindex_table, M_IFNET); 310 } 311 ifindex_table = e; 312 } 313 314 /* ARGSUSED*/ 315 static void 316 if_check(void *dummy __unused) 317 { 318 struct ifnet *ifp; 319 int s; 320 321 s = splimp(); 322 IFNET_RLOCK(); /* could sleep on rare error; mostly okay XXX */ 323 TAILQ_FOREACH(ifp, &ifnet, if_link) { 324 if (ifp->if_snd.ifq_maxlen == 0) { 325 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n"); 326 ifp->if_snd.ifq_maxlen = ifqmaxlen; 327 } 328 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) { 329 if_printf(ifp, 330 "XXX: driver didn't initialize queue mtx\n"); 331 mtx_init(&ifp->if_snd.ifq_mtx, "unknown", 332 MTX_NETWORK_LOCK, MTX_DEF); 333 } 334 } 335 IFNET_RUNLOCK(); 336 splx(s); 337 if_slowtimo(0); 338 } 339 340 /* 341 * Allocate a struct ifnet and in index for an interface. 342 */ 343 struct ifnet* 344 if_alloc(u_char type) 345 { 346 struct ifnet *ifp; 347 348 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 349 350 /* 351 * Try to find an empty slot below if_index. If we fail, take 352 * the next slot. 353 * 354 * XXX: should be locked! 355 */ 356 for (ifp->if_index = 1; ifp->if_index <= if_index; ifp->if_index++) { 357 if (ifnet_byindex(ifp->if_index) == NULL) 358 break; 359 } 360 /* Catch if_index overflow. */ 361 if (ifp->if_index < 1) { 362 free(ifp, M_IFNET); 363 return (NULL); 364 } 365 if (ifp->if_index > if_index) 366 if_index = ifp->if_index; 367 if (if_index >= if_indexlim) 368 if_grow(); 369 ifnet_byindex(ifp->if_index) = ifp; 370 371 ifp->if_type = type; 372 373 if (if_com_alloc[type] != NULL) { 374 ifp->if_l2com = if_com_alloc[type](type, ifp); 375 if (ifp->if_l2com == NULL) { 376 free(ifp, M_IFNET); 377 return (NULL); 378 } 379 } 380 IF_ADDR_LOCK_INIT(ifp); 381 382 return (ifp); 383 } 384 385 void 386 if_free(struct ifnet *ifp) 387 { 388 389 /* Do not add code to this function! Add it to if_free_type(). */ 390 if_free_type(ifp, ifp->if_type); 391 } 392 393 void 394 if_free_type(struct ifnet *ifp, u_char type) 395 { 396 397 if (ifp != ifnet_byindex(ifp->if_index)) { 398 if_printf(ifp, "%s: value was not if_alloced, skipping\n", 399 __func__); 400 return; 401 } 402 403 IF_ADDR_LOCK_DESTROY(ifp); 404 405 ifnet_byindex(ifp->if_index) = NULL; 406 407 /* XXX: should be locked with if_findindex() */ 408 while (if_index > 0 && ifnet_byindex(if_index) == NULL) 409 if_index--; 410 411 if (if_com_free[type] != NULL) 412 if_com_free[type](ifp->if_l2com, type); 413 414 free(ifp, M_IFNET); 415 }; 416 417 /* 418 * Attach an interface to the 419 * list of "active" interfaces. 420 */ 421 void 422 if_attach(struct ifnet *ifp) 423 { 424 unsigned socksize, ifasize; 425 int namelen, masklen; 426 struct sockaddr_dl *sdl; 427 struct ifaddr *ifa; 428 429 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 430 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 431 ifp->if_xname); 432 433 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp); 434 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 435 IF_AFDATA_LOCK_INIT(ifp); 436 ifp->if_afdata_initialized = 0; 437 /* 438 * XXX - 439 * The old code would work if the interface passed a pre-existing 440 * chain of ifaddrs to this code. We don't trust our callers to 441 * properly initialize the tailq, however, so we no longer allow 442 * this unlikely case. 443 */ 444 TAILQ_INIT(&ifp->if_addrhead); 445 TAILQ_INIT(&ifp->if_prefixhead); 446 TAILQ_INIT(&ifp->if_multiaddrs); 447 TAILQ_INIT(&ifp->if_groups); 448 449 if_addgroup(ifp, IFG_ALL); 450 451 knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL); 452 getmicrotime(&ifp->if_lastchange); 453 ifp->if_data.ifi_epoch = time_uptime; 454 ifp->if_data.ifi_datalen = sizeof(struct if_data); 455 456 #ifdef MAC 457 mac_init_ifnet(ifp); 458 mac_create_ifnet(ifp); 459 #endif 460 461 ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw, 462 unit2minor(ifp->if_index), 463 UID_ROOT, GID_WHEEL, 0600, "%s/%s", 464 net_cdevsw.d_name, ifp->if_xname); 465 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d", 466 net_cdevsw.d_name, ifp->if_index); 467 468 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 469 470 /* 471 * create a Link Level name for this device 472 */ 473 namelen = strlen(ifp->if_xname); 474 /* 475 * Always save enough space for any possiable name so we can do 476 * a rename in place later. 477 */ 478 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 479 socksize = masklen + ifp->if_addrlen; 480 if (socksize < sizeof(*sdl)) 481 socksize = sizeof(*sdl); 482 socksize = roundup2(socksize, sizeof(long)); 483 ifasize = sizeof(*ifa) + 2 * socksize; 484 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 485 IFA_LOCK_INIT(ifa); 486 sdl = (struct sockaddr_dl *)(ifa + 1); 487 sdl->sdl_len = socksize; 488 sdl->sdl_family = AF_LINK; 489 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 490 sdl->sdl_nlen = namelen; 491 sdl->sdl_index = ifp->if_index; 492 sdl->sdl_type = ifp->if_type; 493 ifp->if_addr = ifa; 494 ifa->ifa_ifp = ifp; 495 ifa->ifa_rtrequest = link_rtrequest; 496 ifa->ifa_addr = (struct sockaddr *)sdl; 497 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 498 ifa->ifa_netmask = (struct sockaddr *)sdl; 499 sdl->sdl_len = masklen; 500 while (namelen != 0) 501 sdl->sdl_data[--namelen] = 0xff; 502 ifa->ifa_refcnt = 1; 503 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 504 ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */ 505 ifp->if_snd.altq_type = 0; 506 ifp->if_snd.altq_disc = NULL; 507 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 508 ifp->if_snd.altq_tbr = NULL; 509 ifp->if_snd.altq_ifp = ifp; 510 511 IFNET_WLOCK(); 512 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link); 513 IFNET_WUNLOCK(); 514 515 if (domain_init_status >= 2) 516 if_attachdomain1(ifp); 517 518 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 519 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 520 521 /* Announce the interface. */ 522 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 523 524 if (ifp->if_watchdog != NULL) 525 if_printf(ifp, "using obsoleted if_watchdog interface\n"); 526 } 527 528 static void 529 if_attachdomain(void *dummy) 530 { 531 struct ifnet *ifp; 532 int s; 533 534 s = splnet(); 535 TAILQ_FOREACH(ifp, &ifnet, if_link) 536 if_attachdomain1(ifp); 537 splx(s); 538 } 539 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 540 if_attachdomain, NULL); 541 542 static void 543 if_attachdomain1(struct ifnet *ifp) 544 { 545 struct domain *dp; 546 int s; 547 548 s = splnet(); 549 550 /* 551 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 552 * cannot lock ifp->if_afdata initialization, entirely. 553 */ 554 if (IF_AFDATA_TRYLOCK(ifp) == 0) { 555 splx(s); 556 return; 557 } 558 if (ifp->if_afdata_initialized >= domain_init_status) { 559 IF_AFDATA_UNLOCK(ifp); 560 splx(s); 561 printf("if_attachdomain called more than once on %s\n", 562 ifp->if_xname); 563 return; 564 } 565 ifp->if_afdata_initialized = domain_init_status; 566 IF_AFDATA_UNLOCK(ifp); 567 568 /* address family dependent data region */ 569 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 570 for (dp = domains; dp; dp = dp->dom_next) { 571 if (dp->dom_ifattach) 572 ifp->if_afdata[dp->dom_family] = 573 (*dp->dom_ifattach)(ifp); 574 } 575 576 splx(s); 577 } 578 579 /* 580 * Remove any network addresses from an interface. 581 */ 582 583 void 584 if_purgeaddrs(struct ifnet *ifp) 585 { 586 struct ifaddr *ifa, *next; 587 588 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 589 if (ifa->ifa_addr->sa_family == AF_LINK) 590 continue; 591 #ifdef INET 592 /* XXX: Ugly!! ad hoc just for INET */ 593 if (ifa->ifa_addr->sa_family == AF_INET) { 594 struct ifaliasreq ifr; 595 596 bzero(&ifr, sizeof(ifr)); 597 ifr.ifra_addr = *ifa->ifa_addr; 598 if (ifa->ifa_dstaddr) 599 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 600 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 601 NULL) == 0) 602 continue; 603 } 604 #endif /* INET */ 605 #ifdef INET6 606 if (ifa->ifa_addr->sa_family == AF_INET6) { 607 in6_purgeaddr(ifa); 608 /* ifp_addrhead is already updated */ 609 continue; 610 } 611 #endif /* INET6 */ 612 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 613 IFAFREE(ifa); 614 } 615 } 616 617 /* 618 * Detach an interface, removing it from the 619 * list of "active" interfaces. 620 * 621 * XXXRW: There are some significant questions about event ordering, and 622 * how to prevent things from starting to use the interface during detach. 623 */ 624 void 625 if_detach(struct ifnet *ifp) 626 { 627 struct ifaddr *ifa; 628 struct radix_node_head *rnh; 629 int s; 630 int i; 631 struct domain *dp; 632 struct ifnet *iter; 633 int found = 0; 634 635 IFNET_WLOCK(); 636 TAILQ_FOREACH(iter, &ifnet, if_link) 637 if (iter == ifp) { 638 TAILQ_REMOVE(&ifnet, ifp, if_link); 639 found = 1; 640 break; 641 } 642 IFNET_WUNLOCK(); 643 if (!found) 644 return; 645 646 /* 647 * Remove/wait for pending events. 648 */ 649 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 650 651 /* 652 * Remove routes and flush queues. 653 */ 654 s = splnet(); 655 if_down(ifp); 656 #ifdef ALTQ 657 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 658 altq_disable(&ifp->if_snd); 659 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 660 altq_detach(&ifp->if_snd); 661 #endif 662 663 if_purgeaddrs(ifp); 664 665 #ifdef INET 666 in_ifdetach(ifp); 667 #endif 668 669 #ifdef INET6 670 /* 671 * Remove all IPv6 kernel structs related to ifp. This should be done 672 * before removing routing entries below, since IPv6 interface direct 673 * routes are expected to be removed by the IPv6-specific kernel API. 674 * Otherwise, the kernel will detect some inconsistency and bark it. 675 */ 676 in6_ifdetach(ifp); 677 #endif 678 /* 679 * Remove link ifaddr pointer and maybe decrement if_index. 680 * Clean up all addresses. 681 */ 682 ifp->if_addr = NULL; 683 destroy_dev(ifdev_byindex(ifp->if_index)); 684 ifdev_byindex(ifp->if_index) = NULL; 685 686 /* We can now free link ifaddr. */ 687 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 688 ifa = TAILQ_FIRST(&ifp->if_addrhead); 689 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 690 IFAFREE(ifa); 691 } 692 693 /* 694 * Delete all remaining routes using this interface 695 * Unfortuneatly the only way to do this is to slog through 696 * the entire routing table looking for routes which point 697 * to this interface...oh well... 698 */ 699 for (i = 1; i <= AF_MAX; i++) { 700 if ((rnh = rt_tables[i]) == NULL) 701 continue; 702 RADIX_NODE_HEAD_LOCK(rnh); 703 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 704 RADIX_NODE_HEAD_UNLOCK(rnh); 705 } 706 707 /* Announce that the interface is gone. */ 708 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 709 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 710 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 711 712 IF_AFDATA_LOCK(ifp); 713 for (dp = domains; dp; dp = dp->dom_next) { 714 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 715 (*dp->dom_ifdetach)(ifp, 716 ifp->if_afdata[dp->dom_family]); 717 } 718 IF_AFDATA_UNLOCK(ifp); 719 720 #ifdef MAC 721 mac_destroy_ifnet(ifp); 722 #endif /* MAC */ 723 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT); 724 knlist_clear(&ifp->if_klist, 0); 725 knlist_destroy(&ifp->if_klist); 726 mtx_destroy(&ifp->if_snd.ifq_mtx); 727 IF_AFDATA_DESTROY(ifp); 728 splx(s); 729 } 730 731 /* 732 * Add a group to an interface 733 */ 734 int 735 if_addgroup(struct ifnet *ifp, const char *groupname) 736 { 737 struct ifg_list *ifgl; 738 struct ifg_group *ifg = NULL; 739 struct ifg_member *ifgm; 740 741 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 742 groupname[strlen(groupname) - 1] <= '9') 743 return (EINVAL); 744 745 IFNET_WLOCK(); 746 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 747 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 748 IFNET_WUNLOCK(); 749 return (EEXIST); 750 } 751 752 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 753 M_NOWAIT)) == NULL) { 754 IFNET_WUNLOCK(); 755 return (ENOMEM); 756 } 757 758 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 759 M_TEMP, M_NOWAIT)) == NULL) { 760 free(ifgl, M_TEMP); 761 IFNET_WUNLOCK(); 762 return (ENOMEM); 763 } 764 765 TAILQ_FOREACH(ifg, &ifg_head, ifg_next) 766 if (!strcmp(ifg->ifg_group, groupname)) 767 break; 768 769 if (ifg == NULL) { 770 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 771 M_TEMP, M_NOWAIT)) == NULL) { 772 free(ifgl, M_TEMP); 773 free(ifgm, M_TEMP); 774 IFNET_WUNLOCK(); 775 return (ENOMEM); 776 } 777 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 778 ifg->ifg_refcnt = 0; 779 TAILQ_INIT(&ifg->ifg_members); 780 EVENTHANDLER_INVOKE(group_attach_event, ifg); 781 TAILQ_INSERT_TAIL(&ifg_head, ifg, ifg_next); 782 } 783 784 ifg->ifg_refcnt++; 785 ifgl->ifgl_group = ifg; 786 ifgm->ifgm_ifp = ifp; 787 788 IF_ADDR_LOCK(ifp); 789 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 790 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 791 IF_ADDR_UNLOCK(ifp); 792 793 IFNET_WUNLOCK(); 794 795 EVENTHANDLER_INVOKE(group_change_event, groupname); 796 797 return (0); 798 } 799 800 /* 801 * Remove a group from an interface 802 */ 803 int 804 if_delgroup(struct ifnet *ifp, const char *groupname) 805 { 806 struct ifg_list *ifgl; 807 struct ifg_member *ifgm; 808 809 IFNET_WLOCK(); 810 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 811 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 812 break; 813 if (ifgl == NULL) { 814 IFNET_WUNLOCK(); 815 return (ENOENT); 816 } 817 818 IF_ADDR_LOCK(ifp); 819 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 820 IF_ADDR_UNLOCK(ifp); 821 822 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 823 if (ifgm->ifgm_ifp == ifp) 824 break; 825 826 if (ifgm != NULL) { 827 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 828 free(ifgm, M_TEMP); 829 } 830 831 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 832 TAILQ_REMOVE(&ifg_head, ifgl->ifgl_group, ifg_next); 833 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 834 free(ifgl->ifgl_group, M_TEMP); 835 } 836 IFNET_WUNLOCK(); 837 838 free(ifgl, M_TEMP); 839 840 EVENTHANDLER_INVOKE(group_change_event, groupname); 841 842 return (0); 843 } 844 845 /* 846 * Stores all groups from an interface in memory pointed 847 * to by data 848 */ 849 static int 850 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 851 { 852 int len, error; 853 struct ifg_list *ifgl; 854 struct ifg_req ifgrq, *ifgp; 855 struct ifgroupreq *ifgr = data; 856 857 if (ifgr->ifgr_len == 0) { 858 IF_ADDR_LOCK(ifp); 859 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 860 ifgr->ifgr_len += sizeof(struct ifg_req); 861 IF_ADDR_UNLOCK(ifp); 862 return (0); 863 } 864 865 len = ifgr->ifgr_len; 866 ifgp = ifgr->ifgr_groups; 867 /* XXX: wire */ 868 IF_ADDR_LOCK(ifp); 869 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 870 if (len < sizeof(ifgrq)) { 871 IF_ADDR_UNLOCK(ifp); 872 return (EINVAL); 873 } 874 bzero(&ifgrq, sizeof ifgrq); 875 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 876 sizeof(ifgrq.ifgrq_group)); 877 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 878 IF_ADDR_UNLOCK(ifp); 879 return (error); 880 } 881 len -= sizeof(ifgrq); 882 ifgp++; 883 } 884 IF_ADDR_UNLOCK(ifp); 885 886 return (0); 887 } 888 889 /* 890 * Stores all members of a group in memory pointed to by data 891 */ 892 static int 893 if_getgroupmembers(struct ifgroupreq *data) 894 { 895 struct ifgroupreq *ifgr = data; 896 struct ifg_group *ifg; 897 struct ifg_member *ifgm; 898 struct ifg_req ifgrq, *ifgp; 899 int len, error; 900 901 IFNET_RLOCK(); 902 TAILQ_FOREACH(ifg, &ifg_head, ifg_next) 903 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 904 break; 905 if (ifg == NULL) { 906 IFNET_RUNLOCK(); 907 return (ENOENT); 908 } 909 910 if (ifgr->ifgr_len == 0) { 911 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 912 ifgr->ifgr_len += sizeof(ifgrq); 913 IFNET_RUNLOCK(); 914 return (0); 915 } 916 917 len = ifgr->ifgr_len; 918 ifgp = ifgr->ifgr_groups; 919 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 920 if (len < sizeof(ifgrq)) { 921 IFNET_RUNLOCK(); 922 return (EINVAL); 923 } 924 bzero(&ifgrq, sizeof ifgrq); 925 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 926 sizeof(ifgrq.ifgrq_member)); 927 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 928 IFNET_RUNLOCK(); 929 return (error); 930 } 931 len -= sizeof(ifgrq); 932 ifgp++; 933 } 934 IFNET_RUNLOCK(); 935 936 return (0); 937 } 938 939 /* 940 * Delete Routes for a Network Interface 941 * 942 * Called for each routing entry via the rnh->rnh_walktree() call above 943 * to delete all route entries referencing a detaching network interface. 944 * 945 * Arguments: 946 * rn pointer to node in the routing table 947 * arg argument passed to rnh->rnh_walktree() - detaching interface 948 * 949 * Returns: 950 * 0 successful 951 * errno failed - reason indicated 952 * 953 */ 954 static int 955 if_rtdel(struct radix_node *rn, void *arg) 956 { 957 struct rtentry *rt = (struct rtentry *)rn; 958 struct ifnet *ifp = arg; 959 int err; 960 961 if (rt->rt_ifp == ifp) { 962 963 /* 964 * Protect (sorta) against walktree recursion problems 965 * with cloned routes 966 */ 967 if ((rt->rt_flags & RTF_UP) == 0) 968 return (0); 969 970 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 971 rt_mask(rt), rt->rt_flags, 972 (struct rtentry **) NULL); 973 if (err) { 974 log(LOG_WARNING, "if_rtdel: error %d\n", err); 975 } 976 } 977 978 return (0); 979 } 980 981 #define sa_equal(a1, a2) (bcmp((a1), (a2), ((a1))->sa_len) == 0) 982 983 /* 984 * Locate an interface based on a complete address. 985 */ 986 /*ARGSUSED*/ 987 struct ifaddr * 988 ifa_ifwithaddr(struct sockaddr *addr) 989 { 990 struct ifnet *ifp; 991 struct ifaddr *ifa; 992 993 IFNET_RLOCK(); 994 TAILQ_FOREACH(ifp, &ifnet, if_link) 995 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 996 if (ifa->ifa_addr->sa_family != addr->sa_family) 997 continue; 998 if (sa_equal(addr, ifa->ifa_addr)) 999 goto done; 1000 /* IP6 doesn't have broadcast */ 1001 if ((ifp->if_flags & IFF_BROADCAST) && 1002 ifa->ifa_broadaddr && 1003 ifa->ifa_broadaddr->sa_len != 0 && 1004 sa_equal(ifa->ifa_broadaddr, addr)) 1005 goto done; 1006 } 1007 ifa = NULL; 1008 done: 1009 IFNET_RUNLOCK(); 1010 return (ifa); 1011 } 1012 1013 /* 1014 * Locate an interface based on the broadcast address. 1015 */ 1016 /* ARGSUSED */ 1017 struct ifaddr * 1018 ifa_ifwithbroadaddr(struct sockaddr *addr) 1019 { 1020 struct ifnet *ifp; 1021 struct ifaddr *ifa; 1022 1023 IFNET_RLOCK(); 1024 TAILQ_FOREACH(ifp, &ifnet, if_link) 1025 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1026 if (ifa->ifa_addr->sa_family != addr->sa_family) 1027 continue; 1028 if ((ifp->if_flags & IFF_BROADCAST) && 1029 ifa->ifa_broadaddr && 1030 ifa->ifa_broadaddr->sa_len != 0 && 1031 sa_equal(ifa->ifa_broadaddr, addr)) 1032 goto done; 1033 } 1034 ifa = NULL; 1035 done: 1036 IFNET_RUNLOCK(); 1037 return (ifa); 1038 } 1039 1040 /* 1041 * Locate the point to point interface with a given destination address. 1042 */ 1043 /*ARGSUSED*/ 1044 struct ifaddr * 1045 ifa_ifwithdstaddr(struct sockaddr *addr) 1046 { 1047 struct ifnet *ifp; 1048 struct ifaddr *ifa; 1049 1050 IFNET_RLOCK(); 1051 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1052 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1053 continue; 1054 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1055 if (ifa->ifa_addr->sa_family != addr->sa_family) 1056 continue; 1057 if (ifa->ifa_dstaddr && 1058 sa_equal(addr, ifa->ifa_dstaddr)) 1059 goto done; 1060 } 1061 } 1062 ifa = NULL; 1063 done: 1064 IFNET_RUNLOCK(); 1065 return (ifa); 1066 } 1067 1068 /* 1069 * Find an interface on a specific network. If many, choice 1070 * is most specific found. 1071 */ 1072 struct ifaddr * 1073 ifa_ifwithnet(struct sockaddr *addr) 1074 { 1075 struct ifnet *ifp; 1076 struct ifaddr *ifa; 1077 struct ifaddr *ifa_maybe = (struct ifaddr *) 0; 1078 u_int af = addr->sa_family; 1079 char *addr_data = addr->sa_data, *cplim; 1080 1081 /* 1082 * AF_LINK addresses can be looked up directly by their index number, 1083 * so do that if we can. 1084 */ 1085 if (af == AF_LINK) { 1086 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1087 if (sdl->sdl_index && sdl->sdl_index <= if_index) 1088 return (ifaddr_byindex(sdl->sdl_index)); 1089 } 1090 1091 /* 1092 * Scan though each interface, looking for ones that have 1093 * addresses in this address family. 1094 */ 1095 IFNET_RLOCK(); 1096 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1097 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1098 char *cp, *cp2, *cp3; 1099 1100 if (ifa->ifa_addr->sa_family != af) 1101 next: continue; 1102 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) { 1103 /* 1104 * This is a bit broken as it doesn't 1105 * take into account that the remote end may 1106 * be a single node in the network we are 1107 * looking for. 1108 * The trouble is that we don't know the 1109 * netmask for the remote end. 1110 */ 1111 if (ifa->ifa_dstaddr != 0 && 1112 sa_equal(addr, ifa->ifa_dstaddr)) 1113 goto done; 1114 } else { 1115 /* 1116 * if we have a special address handler, 1117 * then use it instead of the generic one. 1118 */ 1119 if (ifa->ifa_claim_addr) { 1120 if ((*ifa->ifa_claim_addr)(ifa, addr)) 1121 goto done; 1122 continue; 1123 } 1124 1125 /* 1126 * Scan all the bits in the ifa's address. 1127 * If a bit dissagrees with what we are 1128 * looking for, mask it with the netmask 1129 * to see if it really matters. 1130 * (A byte at a time) 1131 */ 1132 if (ifa->ifa_netmask == 0) 1133 continue; 1134 cp = addr_data; 1135 cp2 = ifa->ifa_addr->sa_data; 1136 cp3 = ifa->ifa_netmask->sa_data; 1137 cplim = ifa->ifa_netmask->sa_len 1138 + (char *)ifa->ifa_netmask; 1139 while (cp3 < cplim) 1140 if ((*cp++ ^ *cp2++) & *cp3++) 1141 goto next; /* next address! */ 1142 /* 1143 * If the netmask of what we just found 1144 * is more specific than what we had before 1145 * (if we had one) then remember the new one 1146 * before continuing to search 1147 * for an even better one. 1148 */ 1149 if (ifa_maybe == 0 || 1150 rn_refines((caddr_t)ifa->ifa_netmask, 1151 (caddr_t)ifa_maybe->ifa_netmask)) 1152 ifa_maybe = ifa; 1153 } 1154 } 1155 } 1156 ifa = ifa_maybe; 1157 done: 1158 IFNET_RUNLOCK(); 1159 return (ifa); 1160 } 1161 1162 /* 1163 * Find an interface address specific to an interface best matching 1164 * a given address. 1165 */ 1166 struct ifaddr * 1167 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1168 { 1169 struct ifaddr *ifa; 1170 char *cp, *cp2, *cp3; 1171 char *cplim; 1172 struct ifaddr *ifa_maybe = 0; 1173 u_int af = addr->sa_family; 1174 1175 if (af >= AF_MAX) 1176 return (0); 1177 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1178 if (ifa->ifa_addr->sa_family != af) 1179 continue; 1180 if (ifa_maybe == 0) 1181 ifa_maybe = ifa; 1182 if (ifa->ifa_netmask == 0) { 1183 if (sa_equal(addr, ifa->ifa_addr) || 1184 (ifa->ifa_dstaddr && 1185 sa_equal(addr, ifa->ifa_dstaddr))) 1186 goto done; 1187 continue; 1188 } 1189 if (ifp->if_flags & IFF_POINTOPOINT) { 1190 if (sa_equal(addr, ifa->ifa_dstaddr)) 1191 goto done; 1192 } else { 1193 cp = addr->sa_data; 1194 cp2 = ifa->ifa_addr->sa_data; 1195 cp3 = ifa->ifa_netmask->sa_data; 1196 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1197 for (; cp3 < cplim; cp3++) 1198 if ((*cp++ ^ *cp2++) & *cp3) 1199 break; 1200 if (cp3 == cplim) 1201 goto done; 1202 } 1203 } 1204 ifa = ifa_maybe; 1205 done: 1206 return (ifa); 1207 } 1208 1209 #include <net/route.h> 1210 1211 /* 1212 * Default action when installing a route with a Link Level gateway. 1213 * Lookup an appropriate real ifa to point to. 1214 * This should be moved to /sys/net/link.c eventually. 1215 */ 1216 static void 1217 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 1218 { 1219 struct ifaddr *ifa, *oifa; 1220 struct sockaddr *dst; 1221 struct ifnet *ifp; 1222 1223 RT_LOCK_ASSERT(rt); 1224 1225 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 1226 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 1227 return; 1228 ifa = ifaof_ifpforaddr(dst, ifp); 1229 if (ifa) { 1230 IFAREF(ifa); /* XXX */ 1231 oifa = rt->rt_ifa; 1232 rt->rt_ifa = ifa; 1233 IFAFREE(oifa); 1234 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1235 ifa->ifa_rtrequest(cmd, rt, info); 1236 } 1237 } 1238 1239 /* 1240 * Mark an interface down and notify protocols of 1241 * the transition. 1242 * NOTE: must be called at splnet or eqivalent. 1243 */ 1244 static void 1245 if_unroute(struct ifnet *ifp, int flag, int fam) 1246 { 1247 struct ifaddr *ifa; 1248 1249 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 1250 1251 ifp->if_flags &= ~flag; 1252 getmicrotime(&ifp->if_lastchange); 1253 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1254 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1255 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1256 if_qflush(&ifp->if_snd); 1257 #ifdef DEV_CARP 1258 if (ifp->if_carp) 1259 carp_carpdev_state(ifp->if_carp); 1260 #endif 1261 rt_ifmsg(ifp); 1262 } 1263 1264 /* 1265 * Mark an interface up and notify protocols of 1266 * the transition. 1267 * NOTE: must be called at splnet or eqivalent. 1268 */ 1269 static void 1270 if_route(struct ifnet *ifp, int flag, int fam) 1271 { 1272 struct ifaddr *ifa; 1273 1274 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 1275 1276 ifp->if_flags |= flag; 1277 getmicrotime(&ifp->if_lastchange); 1278 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1279 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1280 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1281 #ifdef DEV_CARP 1282 if (ifp->if_carp) 1283 carp_carpdev_state(ifp->if_carp); 1284 #endif 1285 rt_ifmsg(ifp); 1286 #ifdef INET6 1287 in6_if_up(ifp); 1288 #endif 1289 } 1290 1291 void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */ 1292 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1293 1294 /* 1295 * Handle a change in the interface link state. To avoid LORs 1296 * between driver lock and upper layer locks, as well as possible 1297 * recursions, we post event to taskqueue, and all job 1298 * is done in static do_link_state_change(). 1299 */ 1300 void 1301 if_link_state_change(struct ifnet *ifp, int link_state) 1302 { 1303 /* Return if state hasn't changed. */ 1304 if (ifp->if_link_state == link_state) 1305 return; 1306 1307 ifp->if_link_state = link_state; 1308 1309 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 1310 } 1311 1312 static void 1313 do_link_state_change(void *arg, int pending) 1314 { 1315 struct ifnet *ifp = (struct ifnet *)arg; 1316 int link_state = ifp->if_link_state; 1317 int link; 1318 1319 /* Notify that the link state has changed. */ 1320 rt_ifmsg(ifp); 1321 if (link_state == LINK_STATE_UP) 1322 link = NOTE_LINKUP; 1323 else if (link_state == LINK_STATE_DOWN) 1324 link = NOTE_LINKDOWN; 1325 else 1326 link = NOTE_LINKINV; 1327 KNOTE_UNLOCKED(&ifp->if_klist, link); 1328 if (ifp->if_vlantrunk != NULL) 1329 (*vlan_link_state_p)(ifp, link); 1330 1331 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 1332 IFP2AC(ifp)->ac_netgraph != NULL) 1333 (*ng_ether_link_state_p)(ifp, link_state); 1334 #ifdef DEV_CARP 1335 if (ifp->if_carp) 1336 carp_carpdev_state(ifp->if_carp); 1337 #endif 1338 if (ifp->if_bridge) { 1339 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!")); 1340 (*bstp_linkstate_p)(ifp, link_state); 1341 } 1342 1343 devctl_notify("IFNET", ifp->if_xname, 1344 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL); 1345 if (pending > 1) 1346 if_printf(ifp, "%d link states coalesced\n", pending); 1347 if (log_link_state_change) 1348 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 1349 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 1350 } 1351 1352 /* 1353 * Mark an interface down and notify protocols of 1354 * the transition. 1355 * NOTE: must be called at splnet or eqivalent. 1356 */ 1357 void 1358 if_down(struct ifnet *ifp) 1359 { 1360 1361 if_unroute(ifp, IFF_UP, AF_UNSPEC); 1362 } 1363 1364 /* 1365 * Mark an interface up and notify protocols of 1366 * the transition. 1367 * NOTE: must be called at splnet or eqivalent. 1368 */ 1369 void 1370 if_up(struct ifnet *ifp) 1371 { 1372 1373 if_route(ifp, IFF_UP, AF_UNSPEC); 1374 } 1375 1376 /* 1377 * Flush an interface queue. 1378 */ 1379 static void 1380 if_qflush(struct ifaltq *ifq) 1381 { 1382 struct mbuf *m, *n; 1383 1384 IFQ_LOCK(ifq); 1385 #ifdef ALTQ 1386 if (ALTQ_IS_ENABLED(ifq)) 1387 ALTQ_PURGE(ifq); 1388 #endif 1389 n = ifq->ifq_head; 1390 while ((m = n) != 0) { 1391 n = m->m_act; 1392 m_freem(m); 1393 } 1394 ifq->ifq_head = 0; 1395 ifq->ifq_tail = 0; 1396 ifq->ifq_len = 0; 1397 IFQ_UNLOCK(ifq); 1398 } 1399 1400 /* 1401 * Handle interface watchdog timer routines. Called 1402 * from softclock, we decrement timers (if set) and 1403 * call the appropriate interface routine on expiration. 1404 * 1405 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called 1406 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab 1407 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface. 1408 */ 1409 static void 1410 if_slowtimo(void *arg) 1411 { 1412 struct ifnet *ifp; 1413 int s = splimp(); 1414 1415 IFNET_RLOCK(); 1416 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1417 if (ifp->if_timer == 0 || --ifp->if_timer) 1418 continue; 1419 if (ifp->if_watchdog) 1420 (*ifp->if_watchdog)(ifp); 1421 } 1422 IFNET_RUNLOCK(); 1423 splx(s); 1424 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ); 1425 } 1426 1427 /* 1428 * Map interface name to 1429 * interface structure pointer. 1430 */ 1431 struct ifnet * 1432 ifunit(const char *name) 1433 { 1434 struct ifnet *ifp; 1435 1436 IFNET_RLOCK(); 1437 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1438 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 1439 break; 1440 } 1441 IFNET_RUNLOCK(); 1442 return (ifp); 1443 } 1444 1445 /* 1446 * Hardware specific interface ioctls. 1447 */ 1448 static int 1449 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 1450 { 1451 struct ifreq *ifr; 1452 struct ifstat *ifs; 1453 int error = 0; 1454 int new_flags, temp_flags; 1455 size_t namelen, onamelen; 1456 char new_name[IFNAMSIZ]; 1457 struct ifaddr *ifa; 1458 struct sockaddr_dl *sdl; 1459 1460 ifr = (struct ifreq *)data; 1461 switch (cmd) { 1462 case SIOCGIFINDEX: 1463 ifr->ifr_index = ifp->if_index; 1464 break; 1465 1466 case SIOCGIFFLAGS: 1467 temp_flags = ifp->if_flags | ifp->if_drv_flags; 1468 ifr->ifr_flags = temp_flags & 0xffff; 1469 ifr->ifr_flagshigh = temp_flags >> 16; 1470 break; 1471 1472 case SIOCGIFCAP: 1473 ifr->ifr_reqcap = ifp->if_capabilities; 1474 ifr->ifr_curcap = ifp->if_capenable; 1475 break; 1476 1477 #ifdef MAC 1478 case SIOCGIFMAC: 1479 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp); 1480 break; 1481 #endif 1482 1483 case SIOCGIFMETRIC: 1484 ifr->ifr_metric = ifp->if_metric; 1485 break; 1486 1487 case SIOCGIFMTU: 1488 ifr->ifr_mtu = ifp->if_mtu; 1489 break; 1490 1491 case SIOCGIFPHYS: 1492 ifr->ifr_phys = ifp->if_physical; 1493 break; 1494 1495 case SIOCSIFFLAGS: 1496 error = priv_check(td, PRIV_NET_SETIFFLAGS); 1497 if (error) 1498 return (error); 1499 /* 1500 * Currently, no driver owned flags pass the IFF_CANTCHANGE 1501 * check, so we don't need special handling here yet. 1502 */ 1503 new_flags = (ifr->ifr_flags & 0xffff) | 1504 (ifr->ifr_flagshigh << 16); 1505 if (ifp->if_flags & IFF_SMART) { 1506 /* Smart drivers twiddle their own routes */ 1507 } else if (ifp->if_flags & IFF_UP && 1508 (new_flags & IFF_UP) == 0) { 1509 int s = splimp(); 1510 if_down(ifp); 1511 splx(s); 1512 } else if (new_flags & IFF_UP && 1513 (ifp->if_flags & IFF_UP) == 0) { 1514 int s = splimp(); 1515 if_up(ifp); 1516 splx(s); 1517 } 1518 /* See if permanently promiscuous mode bit is about to flip */ 1519 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 1520 if (new_flags & IFF_PPROMISC) 1521 ifp->if_flags |= IFF_PROMISC; 1522 else if (ifp->if_pcount == 0) 1523 ifp->if_flags &= ~IFF_PROMISC; 1524 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 1525 ifp->if_xname, 1526 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); 1527 } 1528 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1529 (new_flags &~ IFF_CANTCHANGE); 1530 if (ifp->if_ioctl) { 1531 IFF_LOCKGIANT(ifp); 1532 (void) (*ifp->if_ioctl)(ifp, cmd, data); 1533 IFF_UNLOCKGIANT(ifp); 1534 } 1535 getmicrotime(&ifp->if_lastchange); 1536 break; 1537 1538 case SIOCSIFCAP: 1539 error = priv_check(td, PRIV_NET_SETIFCAP); 1540 if (error) 1541 return (error); 1542 if (ifp->if_ioctl == NULL) 1543 return (EOPNOTSUPP); 1544 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 1545 return (EINVAL); 1546 IFF_LOCKGIANT(ifp); 1547 error = (*ifp->if_ioctl)(ifp, cmd, data); 1548 IFF_UNLOCKGIANT(ifp); 1549 if (error == 0) 1550 getmicrotime(&ifp->if_lastchange); 1551 break; 1552 1553 #ifdef MAC 1554 case SIOCSIFMAC: 1555 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp); 1556 break; 1557 #endif 1558 1559 case SIOCSIFNAME: 1560 error = priv_check(td, PRIV_NET_SETIFNAME); 1561 if (error) 1562 return (error); 1563 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 1564 if (error != 0) 1565 return (error); 1566 if (new_name[0] == '\0') 1567 return (EINVAL); 1568 if (ifunit(new_name) != NULL) 1569 return (EEXIST); 1570 1571 /* Announce the departure of the interface. */ 1572 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1573 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1574 1575 log(LOG_INFO, "%s: changing name to '%s'\n", 1576 ifp->if_xname, new_name); 1577 1578 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 1579 ifa = ifp->if_addr; 1580 IFA_LOCK(ifa); 1581 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1582 namelen = strlen(new_name); 1583 onamelen = sdl->sdl_nlen; 1584 /* 1585 * Move the address if needed. This is safe because we 1586 * allocate space for a name of length IFNAMSIZ when we 1587 * create this in if_attach(). 1588 */ 1589 if (namelen != onamelen) { 1590 bcopy(sdl->sdl_data + onamelen, 1591 sdl->sdl_data + namelen, sdl->sdl_alen); 1592 } 1593 bcopy(new_name, sdl->sdl_data, namelen); 1594 sdl->sdl_nlen = namelen; 1595 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 1596 bzero(sdl->sdl_data, onamelen); 1597 while (namelen != 0) 1598 sdl->sdl_data[--namelen] = 0xff; 1599 IFA_UNLOCK(ifa); 1600 1601 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 1602 /* Announce the return of the interface. */ 1603 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 1604 break; 1605 1606 case SIOCSIFMETRIC: 1607 error = priv_check(td, PRIV_NET_SETIFMETRIC); 1608 if (error) 1609 return (error); 1610 ifp->if_metric = ifr->ifr_metric; 1611 getmicrotime(&ifp->if_lastchange); 1612 break; 1613 1614 case SIOCSIFPHYS: 1615 error = priv_check(td, PRIV_NET_SETIFPHYS); 1616 if (error) 1617 return (error); 1618 if (ifp->if_ioctl == NULL) 1619 return (EOPNOTSUPP); 1620 IFF_LOCKGIANT(ifp); 1621 error = (*ifp->if_ioctl)(ifp, cmd, data); 1622 IFF_UNLOCKGIANT(ifp); 1623 if (error == 0) 1624 getmicrotime(&ifp->if_lastchange); 1625 break; 1626 1627 case SIOCSIFMTU: 1628 { 1629 u_long oldmtu = ifp->if_mtu; 1630 1631 error = priv_check(td, PRIV_NET_SETIFMTU); 1632 if (error) 1633 return (error); 1634 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 1635 return (EINVAL); 1636 if (ifp->if_ioctl == NULL) 1637 return (EOPNOTSUPP); 1638 IFF_LOCKGIANT(ifp); 1639 error = (*ifp->if_ioctl)(ifp, cmd, data); 1640 IFF_UNLOCKGIANT(ifp); 1641 if (error == 0) { 1642 getmicrotime(&ifp->if_lastchange); 1643 rt_ifmsg(ifp); 1644 } 1645 /* 1646 * If the link MTU changed, do network layer specific procedure. 1647 */ 1648 if (ifp->if_mtu != oldmtu) { 1649 #ifdef INET6 1650 nd6_setmtu(ifp); 1651 #endif 1652 } 1653 break; 1654 } 1655 1656 case SIOCADDMULTI: 1657 case SIOCDELMULTI: 1658 if (cmd == SIOCADDMULTI) 1659 error = priv_check(td, PRIV_NET_ADDMULTI); 1660 else 1661 error = priv_check(td, PRIV_NET_DELMULTI); 1662 if (error) 1663 return (error); 1664 1665 /* Don't allow group membership on non-multicast interfaces. */ 1666 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1667 return (EOPNOTSUPP); 1668 1669 /* Don't let users screw up protocols' entries. */ 1670 if (ifr->ifr_addr.sa_family != AF_LINK) 1671 return (EINVAL); 1672 1673 if (cmd == SIOCADDMULTI) { 1674 struct ifmultiaddr *ifma; 1675 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 1676 } else { 1677 error = if_delmulti(ifp, &ifr->ifr_addr); 1678 } 1679 if (error == 0) 1680 getmicrotime(&ifp->if_lastchange); 1681 break; 1682 1683 case SIOCSIFPHYADDR: 1684 case SIOCDIFPHYADDR: 1685 #ifdef INET6 1686 case SIOCSIFPHYADDR_IN6: 1687 #endif 1688 case SIOCSLIFPHYADDR: 1689 case SIOCSIFMEDIA: 1690 case SIOCSIFGENERIC: 1691 error = priv_check(td, PRIV_NET_HWIOCTL); 1692 if (error) 1693 return (error); 1694 if (ifp->if_ioctl == NULL) 1695 return (EOPNOTSUPP); 1696 IFF_LOCKGIANT(ifp); 1697 error = (*ifp->if_ioctl)(ifp, cmd, data); 1698 IFF_UNLOCKGIANT(ifp); 1699 if (error == 0) 1700 getmicrotime(&ifp->if_lastchange); 1701 break; 1702 1703 case SIOCGIFSTATUS: 1704 ifs = (struct ifstat *)data; 1705 ifs->ascii[0] = '\0'; 1706 1707 case SIOCGIFPSRCADDR: 1708 case SIOCGIFPDSTADDR: 1709 case SIOCGLIFPHYADDR: 1710 case SIOCGIFMEDIA: 1711 case SIOCGIFGENERIC: 1712 if (ifp->if_ioctl == NULL) 1713 return (EOPNOTSUPP); 1714 IFF_LOCKGIANT(ifp); 1715 error = (*ifp->if_ioctl)(ifp, cmd, data); 1716 IFF_UNLOCKGIANT(ifp); 1717 break; 1718 1719 case SIOCSIFLLADDR: 1720 error = priv_check(td, PRIV_NET_SETLLADDR); 1721 if (error) 1722 return (error); 1723 error = if_setlladdr(ifp, 1724 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 1725 break; 1726 1727 case SIOCAIFGROUP: 1728 { 1729 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 1730 1731 error = priv_check(td, PRIV_NET_ADDIFGROUP); 1732 if (error) 1733 return (error); 1734 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 1735 return (error); 1736 break; 1737 } 1738 1739 case SIOCGIFGROUP: 1740 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 1741 return (error); 1742 break; 1743 1744 case SIOCDIFGROUP: 1745 { 1746 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 1747 1748 error = priv_check(td, PRIV_NET_DELIFGROUP); 1749 if (error) 1750 return (error); 1751 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 1752 return (error); 1753 break; 1754 } 1755 1756 default: 1757 error = ENOIOCTL; 1758 break; 1759 } 1760 return (error); 1761 } 1762 1763 /* 1764 * Interface ioctls. 1765 */ 1766 int 1767 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 1768 { 1769 struct ifnet *ifp; 1770 struct ifreq *ifr; 1771 int error; 1772 int oif_flags; 1773 1774 switch (cmd) { 1775 case SIOCGIFCONF: 1776 case OSIOCGIFCONF: 1777 #ifdef __amd64__ 1778 case SIOCGIFCONF32: 1779 #endif 1780 return (ifconf(cmd, data)); 1781 } 1782 ifr = (struct ifreq *)data; 1783 1784 switch (cmd) { 1785 case SIOCIFCREATE: 1786 case SIOCIFCREATE2: 1787 error = priv_check(td, PRIV_NET_IFCREATE); 1788 if (error) 1789 return (error); 1790 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name), 1791 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL)); 1792 case SIOCIFDESTROY: 1793 error = priv_check(td, PRIV_NET_IFDESTROY); 1794 if (error) 1795 return (error); 1796 return if_clone_destroy(ifr->ifr_name); 1797 1798 case SIOCIFGCLONERS: 1799 return (if_clone_list((struct if_clonereq *)data)); 1800 case SIOCGIFGMEMB: 1801 return (if_getgroupmembers((struct ifgroupreq *)data)); 1802 } 1803 1804 ifp = ifunit(ifr->ifr_name); 1805 if (ifp == 0) 1806 return (ENXIO); 1807 1808 error = ifhwioctl(cmd, ifp, data, td); 1809 if (error != ENOIOCTL) 1810 return (error); 1811 1812 oif_flags = ifp->if_flags; 1813 if (so->so_proto == 0) 1814 return (EOPNOTSUPP); 1815 #ifndef COMPAT_43 1816 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 1817 data, 1818 ifp, td)); 1819 #else 1820 { 1821 int ocmd = cmd; 1822 1823 switch (cmd) { 1824 1825 case SIOCSIFDSTADDR: 1826 case SIOCSIFADDR: 1827 case SIOCSIFBRDADDR: 1828 case SIOCSIFNETMASK: 1829 #if BYTE_ORDER != BIG_ENDIAN 1830 if (ifr->ifr_addr.sa_family == 0 && 1831 ifr->ifr_addr.sa_len < 16) { 1832 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 1833 ifr->ifr_addr.sa_len = 16; 1834 } 1835 #else 1836 if (ifr->ifr_addr.sa_len == 0) 1837 ifr->ifr_addr.sa_len = 16; 1838 #endif 1839 break; 1840 1841 case OSIOCGIFADDR: 1842 cmd = SIOCGIFADDR; 1843 break; 1844 1845 case OSIOCGIFDSTADDR: 1846 cmd = SIOCGIFDSTADDR; 1847 break; 1848 1849 case OSIOCGIFBRDADDR: 1850 cmd = SIOCGIFBRDADDR; 1851 break; 1852 1853 case OSIOCGIFNETMASK: 1854 cmd = SIOCGIFNETMASK; 1855 } 1856 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 1857 cmd, 1858 data, 1859 ifp, td)); 1860 switch (ocmd) { 1861 1862 case OSIOCGIFADDR: 1863 case OSIOCGIFDSTADDR: 1864 case OSIOCGIFBRDADDR: 1865 case OSIOCGIFNETMASK: 1866 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 1867 1868 } 1869 } 1870 #endif /* COMPAT_43 */ 1871 1872 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 1873 #ifdef INET6 1874 DELAY(100);/* XXX: temporary workaround for fxp issue*/ 1875 if (ifp->if_flags & IFF_UP) { 1876 int s = splimp(); 1877 in6_if_up(ifp); 1878 splx(s); 1879 } 1880 #endif 1881 } 1882 return (error); 1883 } 1884 1885 /* 1886 * The code common to handling reference counted flags, 1887 * e.g., in ifpromisc() and if_allmulti(). 1888 * The "pflag" argument can specify a permanent mode flag to check, 1889 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 1890 * 1891 * Only to be used on stack-owned flags, not driver-owned flags. 1892 */ 1893 static int 1894 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 1895 { 1896 struct ifreq ifr; 1897 int error; 1898 int oldflags, oldcount; 1899 1900 /* Sanity checks to catch programming errors */ 1901 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 1902 ("%s: setting driver-owned flag %d", __func__, flag)); 1903 1904 if (onswitch) 1905 KASSERT(*refcount >= 0, 1906 ("%s: increment negative refcount %d for flag %d", 1907 __func__, *refcount, flag)); 1908 else 1909 KASSERT(*refcount > 0, 1910 ("%s: decrement non-positive refcount %d for flag %d", 1911 __func__, *refcount, flag)); 1912 1913 /* In case this mode is permanent, just touch refcount */ 1914 if (ifp->if_flags & pflag) { 1915 *refcount += onswitch ? 1 : -1; 1916 return (0); 1917 } 1918 1919 /* Save ifnet parameters for if_ioctl() may fail */ 1920 oldcount = *refcount; 1921 oldflags = ifp->if_flags; 1922 1923 /* 1924 * See if we aren't the only and touching refcount is enough. 1925 * Actually toggle interface flag if we are the first or last. 1926 */ 1927 if (onswitch) { 1928 if ((*refcount)++) 1929 return (0); 1930 ifp->if_flags |= flag; 1931 } else { 1932 if (--(*refcount)) 1933 return (0); 1934 ifp->if_flags &= ~flag; 1935 } 1936 1937 /* Call down the driver since we've changed interface flags */ 1938 if (ifp->if_ioctl == NULL) { 1939 error = EOPNOTSUPP; 1940 goto recover; 1941 } 1942 ifr.ifr_flags = ifp->if_flags & 0xffff; 1943 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1944 IFF_LOCKGIANT(ifp); 1945 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1946 IFF_UNLOCKGIANT(ifp); 1947 if (error) 1948 goto recover; 1949 /* Notify userland that interface flags have changed */ 1950 rt_ifmsg(ifp); 1951 return (0); 1952 1953 recover: 1954 /* Recover after driver error */ 1955 *refcount = oldcount; 1956 ifp->if_flags = oldflags; 1957 return (error); 1958 } 1959 1960 /* 1961 * Set/clear promiscuous mode on interface ifp based on the truth value 1962 * of pswitch. The calls are reference counted so that only the first 1963 * "on" request actually has an effect, as does the final "off" request. 1964 * Results are undefined if the "off" and "on" requests are not matched. 1965 */ 1966 int 1967 ifpromisc(struct ifnet *ifp, int pswitch) 1968 { 1969 int error; 1970 int oldflags = ifp->if_flags; 1971 1972 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 1973 &ifp->if_pcount, pswitch); 1974 /* If promiscuous mode status has changed, log a message */ 1975 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) 1976 log(LOG_INFO, "%s: promiscuous mode %s\n", 1977 ifp->if_xname, 1978 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 1979 return (error); 1980 } 1981 1982 /* 1983 * Return interface configuration 1984 * of system. List may be used 1985 * in later ioctl's (above) to get 1986 * other information. 1987 */ 1988 /*ARGSUSED*/ 1989 static int 1990 ifconf(u_long cmd, caddr_t data) 1991 { 1992 struct ifconf *ifc = (struct ifconf *)data; 1993 #ifdef __amd64__ 1994 struct ifconf32 *ifc32 = (struct ifconf32 *)data; 1995 struct ifconf ifc_swab; 1996 #endif 1997 struct ifnet *ifp; 1998 struct ifaddr *ifa; 1999 struct ifreq ifr; 2000 struct sbuf *sb; 2001 int error, full = 0, valid_len, max_len; 2002 2003 #ifdef __amd64__ 2004 if (cmd == SIOCGIFCONF32) { 2005 ifc_swab.ifc_len = ifc32->ifc_len; 2006 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf; 2007 ifc = &ifc_swab; 2008 } 2009 #endif 2010 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2011 max_len = MAXPHYS - 1; 2012 2013 /* Prevent hostile input from being able to crash the system */ 2014 if (ifc->ifc_len <= 0) 2015 return (EINVAL); 2016 2017 again: 2018 if (ifc->ifc_len <= max_len) { 2019 max_len = ifc->ifc_len; 2020 full = 1; 2021 } 2022 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2023 max_len = 0; 2024 valid_len = 0; 2025 2026 IFNET_RLOCK(); /* could sleep XXX */ 2027 TAILQ_FOREACH(ifp, &ifnet, if_link) { 2028 int addrs; 2029 2030 /* 2031 * Zero the ifr_name buffer to make sure we don't 2032 * disclose the contents of the stack. 2033 */ 2034 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 2035 2036 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 2037 >= sizeof(ifr.ifr_name)) { 2038 sbuf_delete(sb); 2039 IFNET_RUNLOCK(); 2040 return (ENAMETOOLONG); 2041 } 2042 2043 addrs = 0; 2044 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2045 struct sockaddr *sa = ifa->ifa_addr; 2046 2047 if (jailed(curthread->td_ucred) && 2048 prison_if(curthread->td_ucred, sa)) 2049 continue; 2050 addrs++; 2051 #ifdef COMPAT_43 2052 if (cmd == OSIOCGIFCONF) { 2053 struct osockaddr *osa = 2054 (struct osockaddr *)&ifr.ifr_addr; 2055 ifr.ifr_addr = *sa; 2056 osa->sa_family = sa->sa_family; 2057 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2058 max_len += sizeof(ifr); 2059 } else 2060 #endif 2061 if (sa->sa_len <= sizeof(*sa)) { 2062 ifr.ifr_addr = *sa; 2063 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2064 max_len += sizeof(ifr); 2065 } else { 2066 sbuf_bcat(sb, &ifr, 2067 offsetof(struct ifreq, ifr_addr)); 2068 max_len += offsetof(struct ifreq, ifr_addr); 2069 sbuf_bcat(sb, sa, sa->sa_len); 2070 max_len += sa->sa_len; 2071 } 2072 2073 if (!sbuf_overflowed(sb)) 2074 valid_len = sbuf_len(sb); 2075 } 2076 if (addrs == 0) { 2077 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 2078 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2079 max_len += sizeof(ifr); 2080 2081 if (!sbuf_overflowed(sb)) 2082 valid_len = sbuf_len(sb); 2083 } 2084 } 2085 IFNET_RUNLOCK(); 2086 2087 /* 2088 * If we didn't allocate enough space (uncommon), try again. If 2089 * we have already allocated as much space as we are allowed, 2090 * return what we've got. 2091 */ 2092 if (valid_len != max_len && !full) { 2093 sbuf_delete(sb); 2094 goto again; 2095 } 2096 2097 ifc->ifc_len = valid_len; 2098 #ifdef __amd64__ 2099 if (cmd == SIOCGIFCONF32) 2100 ifc32->ifc_len = valid_len; 2101 #endif 2102 sbuf_finish(sb); 2103 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 2104 sbuf_delete(sb); 2105 return (error); 2106 } 2107 2108 /* 2109 * Just like ifpromisc(), but for all-multicast-reception mode. 2110 */ 2111 int 2112 if_allmulti(struct ifnet *ifp, int onswitch) 2113 { 2114 2115 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 2116 } 2117 2118 static struct ifmultiaddr * 2119 if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 2120 { 2121 struct ifmultiaddr *ifma; 2122 2123 IF_ADDR_LOCK_ASSERT(ifp); 2124 2125 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2126 if (sa_equal(ifma->ifma_addr, sa)) 2127 break; 2128 } 2129 2130 return ifma; 2131 } 2132 2133 /* 2134 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 2135 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 2136 * the ifnet multicast address list here, so the caller must do that and 2137 * other setup work (such as notifying the device driver). The reference 2138 * count is initialized to 1. 2139 */ 2140 static struct ifmultiaddr * 2141 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 2142 int mflags) 2143 { 2144 struct ifmultiaddr *ifma; 2145 struct sockaddr *dupsa; 2146 2147 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags | 2148 M_ZERO); 2149 if (ifma == NULL) 2150 return (NULL); 2151 2152 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags); 2153 if (dupsa == NULL) { 2154 FREE(ifma, M_IFMADDR); 2155 return (NULL); 2156 } 2157 bcopy(sa, dupsa, sa->sa_len); 2158 ifma->ifma_addr = dupsa; 2159 2160 ifma->ifma_ifp = ifp; 2161 ifma->ifma_refcount = 1; 2162 ifma->ifma_protospec = NULL; 2163 2164 if (llsa == NULL) { 2165 ifma->ifma_lladdr = NULL; 2166 return (ifma); 2167 } 2168 2169 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags); 2170 if (dupsa == NULL) { 2171 FREE(ifma->ifma_addr, M_IFMADDR); 2172 FREE(ifma, M_IFMADDR); 2173 return (NULL); 2174 } 2175 bcopy(llsa, dupsa, llsa->sa_len); 2176 ifma->ifma_lladdr = dupsa; 2177 2178 return (ifma); 2179 } 2180 2181 /* 2182 * if_freemulti: free ifmultiaddr structure and possibly attached related 2183 * addresses. The caller is responsible for implementing reference 2184 * counting, notifying the driver, handling routing messages, and releasing 2185 * any dependent link layer state. 2186 */ 2187 static void 2188 if_freemulti(struct ifmultiaddr *ifma) 2189 { 2190 2191 KASSERT(ifma->ifma_refcount == 1, ("if_freemulti: refcount %d", 2192 ifma->ifma_refcount)); 2193 KASSERT(ifma->ifma_protospec == NULL, 2194 ("if_freemulti: protospec not NULL")); 2195 2196 if (ifma->ifma_lladdr != NULL) 2197 FREE(ifma->ifma_lladdr, M_IFMADDR); 2198 FREE(ifma->ifma_addr, M_IFMADDR); 2199 FREE(ifma, M_IFMADDR); 2200 } 2201 2202 /* 2203 * Register an additional multicast address with a network interface. 2204 * 2205 * - If the address is already present, bump the reference count on the 2206 * address and return. 2207 * - If the address is not link-layer, look up a link layer address. 2208 * - Allocate address structures for one or both addresses, and attach to the 2209 * multicast address list on the interface. If automatically adding a link 2210 * layer address, the protocol address will own a reference to the link 2211 * layer address, to be freed when it is freed. 2212 * - Notify the network device driver of an addition to the multicast address 2213 * list. 2214 * 2215 * 'sa' points to caller-owned memory with the desired multicast address. 2216 * 2217 * 'retifma' will be used to return a pointer to the resulting multicast 2218 * address reference, if desired. 2219 */ 2220 int 2221 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 2222 struct ifmultiaddr **retifma) 2223 { 2224 struct ifmultiaddr *ifma, *ll_ifma; 2225 struct sockaddr *llsa; 2226 int error; 2227 2228 /* 2229 * If the address is already present, return a new reference to it; 2230 * otherwise, allocate storage and set up a new address. 2231 */ 2232 IF_ADDR_LOCK(ifp); 2233 ifma = if_findmulti(ifp, sa); 2234 if (ifma != NULL) { 2235 ifma->ifma_refcount++; 2236 if (retifma != NULL) 2237 *retifma = ifma; 2238 IF_ADDR_UNLOCK(ifp); 2239 return (0); 2240 } 2241 2242 /* 2243 * The address isn't already present; resolve the protocol address 2244 * into a link layer address, and then look that up, bump its 2245 * refcount or allocate an ifma for that also. If 'llsa' was 2246 * returned, we will need to free it later. 2247 */ 2248 llsa = NULL; 2249 ll_ifma = NULL; 2250 if (ifp->if_resolvemulti != NULL) { 2251 error = ifp->if_resolvemulti(ifp, &llsa, sa); 2252 if (error) 2253 goto unlock_out; 2254 } 2255 2256 /* 2257 * Allocate the new address. Don't hook it up yet, as we may also 2258 * need to allocate a link layer multicast address. 2259 */ 2260 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 2261 if (ifma == NULL) { 2262 error = ENOMEM; 2263 goto free_llsa_out; 2264 } 2265 2266 /* 2267 * If a link layer address is found, we'll need to see if it's 2268 * already present in the address list, or allocate is as well. 2269 * When this block finishes, the link layer address will be on the 2270 * list. 2271 */ 2272 if (llsa != NULL) { 2273 ll_ifma = if_findmulti(ifp, llsa); 2274 if (ll_ifma == NULL) { 2275 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 2276 if (ll_ifma == NULL) { 2277 if_freemulti(ifma); 2278 error = ENOMEM; 2279 goto free_llsa_out; 2280 } 2281 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 2282 ifma_link); 2283 } else 2284 ll_ifma->ifma_refcount++; 2285 } 2286 2287 /* 2288 * We now have a new multicast address, ifma, and possibly a new or 2289 * referenced link layer address. Add the primary address to the 2290 * ifnet address list. 2291 */ 2292 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 2293 2294 if (retifma != NULL) 2295 *retifma = ifma; 2296 2297 /* 2298 * Must generate the message while holding the lock so that 'ifma' 2299 * pointer is still valid. 2300 * 2301 * XXXRW: How come we don't announce ll_ifma? 2302 */ 2303 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 2304 IF_ADDR_UNLOCK(ifp); 2305 2306 /* 2307 * We are certain we have added something, so call down to the 2308 * interface to let them know about it. 2309 */ 2310 if (ifp->if_ioctl != NULL) { 2311 IFF_LOCKGIANT(ifp); 2312 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 2313 IFF_UNLOCKGIANT(ifp); 2314 } 2315 2316 if (llsa != NULL) 2317 FREE(llsa, M_IFMADDR); 2318 2319 return (0); 2320 2321 free_llsa_out: 2322 if (llsa != NULL) 2323 FREE(llsa, M_IFMADDR); 2324 2325 unlock_out: 2326 IF_ADDR_UNLOCK(ifp); 2327 return (error); 2328 } 2329 2330 /* 2331 * Remove a reference to a multicast address on this interface. Yell 2332 * if the request does not match an existing membership. 2333 */ 2334 int 2335 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 2336 { 2337 struct ifmultiaddr *ifma, *ll_ifma; 2338 2339 IF_ADDR_LOCK(ifp); 2340 ifma = if_findmulti(ifp, sa); 2341 if (ifma == NULL) { 2342 IF_ADDR_UNLOCK(ifp); 2343 return ENOENT; 2344 } 2345 2346 if (ifma->ifma_refcount > 1) { 2347 ifma->ifma_refcount--; 2348 IF_ADDR_UNLOCK(ifp); 2349 return 0; 2350 } 2351 2352 sa = ifma->ifma_lladdr; 2353 if (sa != NULL) 2354 ll_ifma = if_findmulti(ifp, sa); 2355 else 2356 ll_ifma = NULL; 2357 2358 /* 2359 * XXXRW: How come we don't announce ll_ifma? 2360 */ 2361 rt_newmaddrmsg(RTM_DELMADDR, ifma); 2362 2363 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 2364 if_freemulti(ifma); 2365 2366 if (ll_ifma != NULL) { 2367 if (ll_ifma->ifma_refcount == 1) { 2368 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifma_link); 2369 if_freemulti(ll_ifma); 2370 } else 2371 ll_ifma->ifma_refcount--; 2372 } 2373 IF_ADDR_UNLOCK(ifp); 2374 2375 /* 2376 * Make sure the interface driver is notified 2377 * in the case of a link layer mcast group being left. 2378 */ 2379 if (ifp->if_ioctl) { 2380 IFF_LOCKGIANT(ifp); 2381 (void) (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 2382 IFF_UNLOCKGIANT(ifp); 2383 } 2384 2385 return 0; 2386 } 2387 2388 /* 2389 * Set the link layer address on an interface. 2390 * 2391 * At this time we only support certain types of interfaces, 2392 * and we don't allow the length of the address to change. 2393 */ 2394 int 2395 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 2396 { 2397 struct sockaddr_dl *sdl; 2398 struct ifaddr *ifa; 2399 struct ifreq ifr; 2400 2401 ifa = ifp->if_addr; 2402 if (ifa == NULL) 2403 return (EINVAL); 2404 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2405 if (sdl == NULL) 2406 return (EINVAL); 2407 if (len != sdl->sdl_alen) /* don't allow length to change */ 2408 return (EINVAL); 2409 switch (ifp->if_type) { 2410 case IFT_ETHER: 2411 case IFT_FDDI: 2412 case IFT_XETHER: 2413 case IFT_ISO88025: 2414 case IFT_L2VLAN: 2415 case IFT_BRIDGE: 2416 case IFT_ARCNET: 2417 bcopy(lladdr, LLADDR(sdl), len); 2418 break; 2419 default: 2420 return (ENODEV); 2421 } 2422 /* 2423 * If the interface is already up, we need 2424 * to re-init it in order to reprogram its 2425 * address filter. 2426 */ 2427 if ((ifp->if_flags & IFF_UP) != 0) { 2428 if (ifp->if_ioctl) { 2429 IFF_LOCKGIANT(ifp); 2430 ifp->if_flags &= ~IFF_UP; 2431 ifr.ifr_flags = ifp->if_flags & 0xffff; 2432 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2433 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2434 ifp->if_flags |= IFF_UP; 2435 ifr.ifr_flags = ifp->if_flags & 0xffff; 2436 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2437 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2438 IFF_UNLOCKGIANT(ifp); 2439 } 2440 #ifdef INET 2441 /* 2442 * Also send gratuitous ARPs to notify other nodes about 2443 * the address change. 2444 */ 2445 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2446 if (ifa->ifa_addr->sa_family == AF_INET) 2447 arp_ifinit(ifp, ifa); 2448 } 2449 #endif 2450 } 2451 return (0); 2452 } 2453 2454 /* 2455 * The name argument must be a pointer to storage which will last as 2456 * long as the interface does. For physical devices, the result of 2457 * device_get_name(dev) is a good choice and for pseudo-devices a 2458 * static string works well. 2459 */ 2460 void 2461 if_initname(struct ifnet *ifp, const char *name, int unit) 2462 { 2463 ifp->if_dname = name; 2464 ifp->if_dunit = unit; 2465 if (unit != IF_DUNIT_NONE) 2466 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 2467 else 2468 strlcpy(ifp->if_xname, name, IFNAMSIZ); 2469 } 2470 2471 int 2472 if_printf(struct ifnet *ifp, const char * fmt, ...) 2473 { 2474 va_list ap; 2475 int retval; 2476 2477 retval = printf("%s: ", ifp->if_xname); 2478 va_start(ap, fmt); 2479 retval += vprintf(fmt, ap); 2480 va_end(ap); 2481 return (retval); 2482 } 2483 2484 /* 2485 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot 2486 * be called without Giant. However, we often can't acquire the Giant lock 2487 * at those points; instead, we run it via a task queue that holds Giant via 2488 * if_start_deferred. 2489 * 2490 * XXXRW: We need to make sure that the ifnet isn't fully detached until any 2491 * outstanding if_start_deferred() tasks that will run after the free. This 2492 * probably means waiting in if_detach(). 2493 */ 2494 void 2495 if_start(struct ifnet *ifp) 2496 { 2497 2498 NET_ASSERT_GIANT(); 2499 2500 if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) { 2501 if (mtx_owned(&Giant)) 2502 (*(ifp)->if_start)(ifp); 2503 else 2504 taskqueue_enqueue(taskqueue_swi_giant, 2505 &ifp->if_starttask); 2506 } else 2507 (*(ifp)->if_start)(ifp); 2508 } 2509 2510 static void 2511 if_start_deferred(void *context, int pending) 2512 { 2513 struct ifnet *ifp; 2514 2515 /* 2516 * This code must be entered with Giant, and should never run if 2517 * we're not running with debug.mpsafenet. 2518 */ 2519 KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet")); 2520 GIANT_REQUIRED; 2521 2522 ifp = context; 2523 (ifp->if_start)(ifp); 2524 } 2525 2526 int 2527 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 2528 { 2529 int active = 0; 2530 2531 IF_LOCK(ifq); 2532 if (_IF_QFULL(ifq)) { 2533 _IF_DROP(ifq); 2534 IF_UNLOCK(ifq); 2535 m_freem(m); 2536 return (0); 2537 } 2538 if (ifp != NULL) { 2539 ifp->if_obytes += m->m_pkthdr.len + adjust; 2540 if (m->m_flags & (M_BCAST|M_MCAST)) 2541 ifp->if_omcasts++; 2542 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 2543 } 2544 _IF_ENQUEUE(ifq, m); 2545 IF_UNLOCK(ifq); 2546 if (ifp != NULL && !active) 2547 if_start(ifp); 2548 return (1); 2549 } 2550 2551 void 2552 if_register_com_alloc(u_char type, 2553 if_com_alloc_t *a, if_com_free_t *f) 2554 { 2555 2556 KASSERT(if_com_alloc[type] == NULL, 2557 ("if_register_com_alloc: %d already registered", type)); 2558 KASSERT(if_com_free[type] == NULL, 2559 ("if_register_com_alloc: %d free already registered", type)); 2560 2561 if_com_alloc[type] = a; 2562 if_com_free[type] = f; 2563 } 2564 2565 void 2566 if_deregister_com_alloc(u_char type) 2567 { 2568 2569 KASSERT(if_com_alloc[type] != NULL, 2570 ("if_deregister_com_alloc: %d not registered", type)); 2571 KASSERT(if_com_free[type] != NULL, 2572 ("if_deregister_com_alloc: %d free not registered", type)); 2573 if_com_alloc[type] = NULL; 2574 if_com_free[type] = NULL; 2575 } 2576