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