if.c revision 194251
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: head/sys/net/if.c 194251 2009-06-15 18:59:29Z jamie $ 31 */ 32 33#include "opt_compat.h" 34#include "opt_inet6.h" 35#include "opt_inet.h" 36#include "opt_carp.h" 37 38#include <sys/param.h> 39#include <sys/types.h> 40#include <sys/conf.h> 41#include <sys/malloc.h> 42#include <sys/sbuf.h> 43#include <sys/bus.h> 44#include <sys/mbuf.h> 45#include <sys/systm.h> 46#include <sys/priv.h> 47#include <sys/proc.h> 48#include <sys/socket.h> 49#include <sys/socketvar.h> 50#include <sys/protosw.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/refcount.h> 54#include <sys/module.h> 55#include <sys/rwlock.h> 56#include <sys/sockio.h> 57#include <sys/syslog.h> 58#include <sys/sysctl.h> 59#include <sys/taskqueue.h> 60#include <sys/domain.h> 61#include <sys/jail.h> 62#include <sys/vimage.h> 63#include <machine/stdarg.h> 64#include <vm/uma.h> 65 66#include <net/if.h> 67#include <net/if_arp.h> 68#include <net/if_clone.h> 69#include <net/if_dl.h> 70#include <net/if_types.h> 71#include <net/if_var.h> 72#include <net/radix.h> 73#include <net/route.h> 74#include <net/vnet.h> 75 76#if defined(INET) || defined(INET6) 77/*XXX*/ 78#include <netinet/in.h> 79#include <netinet/in_var.h> 80#ifdef INET6 81#include <netinet6/in6_var.h> 82#include <netinet6/in6_ifattach.h> 83#endif 84#endif 85#ifdef INET 86#include <netinet/if_ether.h> 87#include <netinet/vinet.h> 88#endif 89#if defined(INET) || defined(INET6) 90#ifdef DEV_CARP 91#include <netinet/ip_carp.h> 92#endif 93#endif 94 95#include <security/mac/mac_framework.h> 96 97#ifndef VIMAGE 98#ifndef VIMAGE_GLOBALS 99struct vnet_net vnet_net_0; 100#endif 101#endif 102 103static int slowtimo_started; 104 105SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 106SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 107 108/* Log link state change events */ 109static int log_link_state_change = 1; 110 111SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 112 &log_link_state_change, 0, 113 "log interface link state change events"); 114 115void (*bstp_linkstate_p)(struct ifnet *ifp, int state); 116void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 117void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 118 119struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 120 121/* 122 * XXX: Style; these should be sorted alphabetically, and unprototyped 123 * static functions should be prototyped. Currently they are sorted by 124 * declaration order. 125 */ 126static void if_attachdomain(void *); 127static void if_attachdomain1(struct ifnet *); 128static int ifconf(u_long, caddr_t); 129static void if_freemulti(struct ifmultiaddr *); 130static void if_init(void *); 131static void if_check(void *); 132static void if_route(struct ifnet *, int flag, int fam); 133static int if_setflag(struct ifnet *, int, int, int *, int); 134static void if_slowtimo(void *); 135static int if_transmit(struct ifnet *ifp, struct mbuf *m); 136static void if_unroute(struct ifnet *, int flag, int fam); 137static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 138static int if_rtdel(struct radix_node *, void *); 139static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 140static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 141static void do_link_state_change(void *, int); 142static int if_getgroup(struct ifgroupreq *, struct ifnet *); 143static int if_getgroupmembers(struct ifgroupreq *); 144static void if_delgroups(struct ifnet *); 145static void if_attach_internal(struct ifnet *, int); 146static void if_detach_internal(struct ifnet *, int); 147 148#ifdef INET6 149/* 150 * XXX: declare here to avoid to include many inet6 related files.. 151 * should be more generalized? 152 */ 153extern void nd6_setmtu(struct ifnet *); 154#endif 155 156static int vnet_net_iattach(const void *); 157#ifdef VIMAGE 158static int vnet_net_idetach(const void *); 159#endif 160 161#ifdef VIMAGE_GLOBALS 162struct ifnethead ifnet; /* depend on static init XXX */ 163struct ifgrouphead ifg_head; 164int if_index; 165static int if_indexlim; 166/* Table of ifnet/cdev by index. Locked with ifnet_lock. */ 167static struct ifindex_entry *ifindex_table; 168static struct knlist ifklist; 169#endif 170 171int ifqmaxlen = IFQ_MAXLEN; 172struct rwlock ifnet_lock; 173static if_com_alloc_t *if_com_alloc[256]; 174static if_com_free_t *if_com_free[256]; 175 176static void filt_netdetach(struct knote *kn); 177static int filt_netdev(struct knote *kn, long hint); 178 179static struct filterops netdev_filtops = 180 { 1, NULL, filt_netdetach, filt_netdev }; 181 182#ifndef VIMAGE_GLOBALS 183static struct vnet_symmap vnet_net_symmap[] = { 184 VNET_SYMMAP(net, ifnet), 185 VNET_SYMMAP(net, rt_tables), 186 VNET_SYMMAP(net, rtstat), 187 VNET_SYMMAP(net, rttrash), 188 VNET_SYMMAP_END 189}; 190 191static const vnet_modinfo_t vnet_net_modinfo = { 192 .vmi_id = VNET_MOD_NET, 193 .vmi_name = "net", 194 .vmi_size = sizeof(struct vnet_net), 195 .vmi_symmap = vnet_net_symmap, 196 .vmi_iattach = vnet_net_iattach, 197#ifdef VIMAGE 198 .vmi_idetach = vnet_net_idetach 199#endif 200}; 201#endif /* !VIMAGE_GLOBALS */ 202 203/* 204 * System initialization 205 */ 206SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL); 207SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL); 208 209MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 210MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 211MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 212 213struct ifnet * 214ifnet_byindex_locked(u_short idx) 215{ 216 INIT_VNET_NET(curvnet); 217 218 if (idx > V_if_index) 219 return (NULL); 220 return (V_ifindex_table[idx].ife_ifnet); 221} 222 223struct ifnet * 224ifnet_byindex(u_short idx) 225{ 226 struct ifnet *ifp; 227 228 IFNET_RLOCK(); 229 ifp = ifnet_byindex_locked(idx); 230 IFNET_RUNLOCK(); 231 return (ifp); 232} 233 234struct ifnet * 235ifnet_byindex_ref(u_short idx) 236{ 237 struct ifnet *ifp; 238 239 IFNET_RLOCK(); 240 ifp = ifnet_byindex_locked(idx); 241 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 242 IFNET_RUNLOCK(); 243 return (NULL); 244 } 245 if_ref(ifp); 246 IFNET_RUNLOCK(); 247 return (ifp); 248} 249 250static void 251ifnet_setbyindex(u_short idx, struct ifnet *ifp) 252{ 253 INIT_VNET_NET(curvnet); 254 255 IFNET_WLOCK_ASSERT(); 256 257 V_ifindex_table[idx].ife_ifnet = ifp; 258} 259 260struct ifaddr * 261ifaddr_byindex(u_short idx) 262{ 263 struct ifaddr *ifa; 264 265 IFNET_RLOCK(); 266 ifa = ifnet_byindex_locked(idx)->if_addr; 267 IFNET_RUNLOCK(); 268 return (ifa); 269} 270 271struct cdev * 272ifdev_byindex(u_short idx) 273{ 274 INIT_VNET_NET(curvnet); 275 struct cdev *cdev; 276 277 IFNET_RLOCK(); 278 cdev = V_ifindex_table[idx].ife_dev; 279 IFNET_RUNLOCK(); 280 return (cdev); 281} 282 283static void 284ifdev_setbyindex(u_short idx, struct cdev *cdev) 285{ 286 INIT_VNET_NET(curvnet); 287 288 IFNET_WLOCK(); 289 V_ifindex_table[idx].ife_dev = cdev; 290 IFNET_WUNLOCK(); 291} 292 293static d_open_t netopen; 294static d_close_t netclose; 295static d_ioctl_t netioctl; 296static d_kqfilter_t netkqfilter; 297 298static struct cdevsw net_cdevsw = { 299 .d_version = D_VERSION, 300 .d_flags = D_NEEDGIANT, 301 .d_open = netopen, 302 .d_close = netclose, 303 .d_ioctl = netioctl, 304 .d_name = "net", 305 .d_kqfilter = netkqfilter, 306}; 307 308static int 309netopen(struct cdev *dev, int flag, int mode, struct thread *td) 310{ 311 return (0); 312} 313 314static int 315netclose(struct cdev *dev, int flags, int fmt, struct thread *td) 316{ 317 return (0); 318} 319 320static int 321netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 322{ 323 struct ifnet *ifp; 324 int error, idx; 325 326 /* only support interface specific ioctls */ 327 if (IOCGROUP(cmd) != 'i') 328 return (EOPNOTSUPP); 329 idx = dev2unit(dev); 330 if (idx == 0) { 331 /* 332 * special network device, not interface. 333 */ 334 if (cmd == SIOCGIFCONF) 335 return (ifconf(cmd, data)); /* XXX remove cmd */ 336#ifdef __amd64__ 337 if (cmd == SIOCGIFCONF32) 338 return (ifconf(cmd, data)); /* XXX remove cmd */ 339#endif 340 return (EOPNOTSUPP); 341 } 342 343 ifp = ifnet_byindex(idx); 344 if (ifp == NULL) 345 return (ENXIO); 346 347 error = ifhwioctl(cmd, ifp, data, td); 348 if (error == ENOIOCTL) 349 error = EOPNOTSUPP; 350 return (error); 351} 352 353static int 354netkqfilter(struct cdev *dev, struct knote *kn) 355{ 356 INIT_VNET_NET(curvnet); 357 struct knlist *klist; 358 struct ifnet *ifp; 359 int idx; 360 361 switch (kn->kn_filter) { 362 case EVFILT_NETDEV: 363 kn->kn_fop = &netdev_filtops; 364 break; 365 default: 366 return (EINVAL); 367 } 368 369 idx = dev2unit(dev); 370 if (idx == 0) { 371 klist = &V_ifklist; 372 } else { 373 ifp = ifnet_byindex(idx); 374 if (ifp == NULL) 375 return (1); 376 klist = &ifp->if_klist; 377 } 378 379 kn->kn_hook = (caddr_t)klist; 380 381 knlist_add(klist, kn, 0); 382 383 return (0); 384} 385 386static void 387filt_netdetach(struct knote *kn) 388{ 389 struct knlist *klist = (struct knlist *)kn->kn_hook; 390 391 knlist_remove(klist, kn, 0); 392} 393 394static int 395filt_netdev(struct knote *kn, long hint) 396{ 397 struct knlist *klist = (struct knlist *)kn->kn_hook; 398 399 /* 400 * Currently NOTE_EXIT is abused to indicate device detach. 401 */ 402 if (hint == NOTE_EXIT) { 403 kn->kn_data = NOTE_LINKINV; 404 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 405 knlist_remove_inevent(klist, kn); 406 return (1); 407 } 408 if (hint != 0) 409 kn->kn_data = hint; /* current status */ 410 if (kn->kn_sfflags & hint) 411 kn->kn_fflags |= hint; 412 return (kn->kn_fflags != 0); 413} 414 415/* 416 * Network interface utility routines. 417 * 418 * Routines with ifa_ifwith* names take sockaddr *'s as 419 * parameters. 420 */ 421 422/* ARGSUSED*/ 423static void 424if_init(void *dummy __unused) 425{ 426 427#ifndef VIMAGE_GLOBALS 428 vnet_mod_register(&vnet_net_modinfo); 429#else 430 vnet_net_iattach(NULL); 431#endif 432 433 IFNET_LOCK_INIT(); 434 ifdev_setbyindex(0, make_dev(&net_cdevsw, 0, UID_ROOT, GID_WHEEL, 435 0600, "network")); 436 if_clone_init(); 437} 438 439static int 440vnet_net_iattach(const void *unused __unused) 441{ 442 INIT_VNET_NET(curvnet); 443 444 V_if_index = 0; 445 V_ifindex_table = NULL; 446 V_if_indexlim = 8; 447 448 TAILQ_INIT(&V_ifnet); 449 TAILQ_INIT(&V_ifg_head); 450 knlist_init_mtx(&V_ifklist, NULL); 451 if_grow(); /* create initial table */ 452 453 return (0); 454} 455 456#ifdef VIMAGE 457static int 458vnet_net_idetach(const void *unused __unused) 459{ 460 INIT_VNET_NET(curvnet); 461 462 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet)); 463 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head)); 464 VNET_ASSERT(SLIST_EMPTY(&V_ifklist.kl_list)); 465 466 free((caddr_t)V_ifindex_table, M_IFNET); 467 468 return (0); 469} 470#endif 471 472void 473if_grow(void) 474{ 475 INIT_VNET_NET(curvnet); 476 u_int n; 477 struct ifindex_entry *e; 478 479 V_if_indexlim <<= 1; 480 n = V_if_indexlim * sizeof(*e); 481 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 482 if (V_ifindex_table != NULL) { 483 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 484 free((caddr_t)V_ifindex_table, M_IFNET); 485 } 486 V_ifindex_table = e; 487} 488 489static void 490if_check(void *dummy __unused) 491{ 492 493 /* 494 * If at least one interface added during boot uses 495 * if_watchdog then start the timer. 496 */ 497 if (slowtimo_started) 498 if_slowtimo(0); 499} 500 501/* 502 * Allocate a struct ifnet and an index for an interface. A layer 2 503 * common structure will also be allocated if an allocation routine is 504 * registered for the passed type. 505 */ 506struct ifnet * 507if_alloc(u_char type) 508{ 509 INIT_VNET_NET(curvnet); 510 struct ifnet *ifp; 511 512 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 513 514 /* 515 * Try to find an empty slot below if_index. If we fail, take 516 * the next slot. 517 * 518 * XXX: should be locked! 519 */ 520 for (ifp->if_index = 1; ifp->if_index <= V_if_index; ifp->if_index++) { 521 if (ifnet_byindex(ifp->if_index) == NULL) 522 break; 523 } 524 /* Catch if_index overflow. */ 525 if (ifp->if_index < 1) { 526 free(ifp, M_IFNET); 527 return (NULL); 528 } 529 if (ifp->if_index > V_if_index) 530 V_if_index = ifp->if_index; 531 if (V_if_index >= V_if_indexlim) 532 if_grow(); 533 534 ifp->if_type = type; 535 ifp->if_alloctype = type; 536 537 if (if_com_alloc[type] != NULL) { 538 ifp->if_l2com = if_com_alloc[type](type, ifp); 539 if (ifp->if_l2com == NULL) { 540 free(ifp, M_IFNET); 541 return (NULL); 542 } 543 } 544 545 IF_ADDR_LOCK_INIT(ifp); 546 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 547 ifp->if_afdata_initialized = 0; 548 IF_AFDATA_LOCK_INIT(ifp); 549 TAILQ_INIT(&ifp->if_addrhead); 550 TAILQ_INIT(&ifp->if_prefixhead); 551 TAILQ_INIT(&ifp->if_multiaddrs); 552 TAILQ_INIT(&ifp->if_groups); 553 knlist_init_mtx(&ifp->if_klist, NULL); 554#ifdef MAC 555 mac_ifnet_init(ifp); 556#endif 557 558 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 559 IFNET_WLOCK(); 560 ifnet_setbyindex(ifp->if_index, ifp); 561 IFNET_WUNLOCK(); 562 return (ifp); 563} 564 565/* 566 * Do the actual work of freeing a struct ifnet, associated index, and layer 567 * 2 common structure. This call is made when the last reference to an 568 * interface is released. 569 */ 570static void 571if_free_internal(struct ifnet *ifp) 572{ 573 INIT_VNET_NET(curvnet); /* ifp->if_vnet is already NULL here */ 574 575 KASSERT((ifp->if_flags & IFF_DYING), 576 ("if_free_internal: interface not dying")); 577 578 IFNET_WLOCK(); 579 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 580 ("%s: freeing unallocated ifnet", ifp->if_xname)); 581 582 ifnet_setbyindex(ifp->if_index, NULL); 583 while (V_if_index > 0 && ifnet_byindex_locked(V_if_index) == NULL) 584 V_if_index--; 585 IFNET_WUNLOCK(); 586 587 if (if_com_free[ifp->if_alloctype] != NULL) 588 if_com_free[ifp->if_alloctype](ifp->if_l2com, 589 ifp->if_alloctype); 590 591#ifdef MAC 592 mac_ifnet_destroy(ifp); 593#endif /* MAC */ 594 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT); 595 knlist_clear(&ifp->if_klist, 0); 596 knlist_destroy(&ifp->if_klist); 597 IF_AFDATA_DESTROY(ifp); 598 IF_ADDR_LOCK_DESTROY(ifp); 599 ifq_detach(&ifp->if_snd); 600 free(ifp, M_IFNET); 601} 602 603/* 604 * This version should only be called by intefaces that switch their type 605 * after calling if_alloc(). if_free_type() will go away again now that we 606 * have if_alloctype to cache the original allocation type. For now, assert 607 * that they match, since we require that in practice. 608 */ 609void 610if_free_type(struct ifnet *ifp, u_char type) 611{ 612 613 KASSERT(ifp->if_alloctype == type, 614 ("if_free_type: type (%d) != alloctype (%d)", type, 615 ifp->if_alloctype)); 616 617 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 618 if (!refcount_release(&ifp->if_refcount)) 619 return; 620 if_free_internal(ifp); 621} 622 623/* 624 * This is the normal version of if_free(), used by device drivers to free a 625 * detached network interface. The contents of if_free_type() will move into 626 * here when if_free_type() goes away. 627 */ 628void 629if_free(struct ifnet *ifp) 630{ 631 632 if_free_type(ifp, ifp->if_alloctype); 633} 634 635/* 636 * Interfaces to keep an ifnet type-stable despite the possibility of the 637 * driver calling if_free(). If there are additional references, we defer 638 * freeing the underlying data structure. 639 */ 640void 641if_ref(struct ifnet *ifp) 642{ 643 644 /* We don't assert the ifnet list lock here, but arguably should. */ 645 refcount_acquire(&ifp->if_refcount); 646} 647 648void 649if_rele(struct ifnet *ifp) 650{ 651 652 if (!refcount_release(&ifp->if_refcount)) 653 return; 654 if_free_internal(ifp); 655} 656 657void 658ifq_attach(struct ifaltq *ifq, struct ifnet *ifp) 659{ 660 661 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 662 663 if (ifq->ifq_maxlen == 0) 664 ifq->ifq_maxlen = ifqmaxlen; 665 666 ifq->altq_type = 0; 667 ifq->altq_disc = NULL; 668 ifq->altq_flags &= ALTQF_CANTCHANGE; 669 ifq->altq_tbr = NULL; 670 ifq->altq_ifp = ifp; 671} 672 673void 674ifq_detach(struct ifaltq *ifq) 675{ 676 mtx_destroy(&ifq->ifq_mtx); 677} 678 679/* 680 * Perform generic interface initalization tasks and attach the interface 681 * to the list of "active" interfaces. If vmove flag is set on entry 682 * to if_attach_internal(), perform only a limited subset of initialization 683 * tasks, given that we are moving from one vnet to another an ifnet which 684 * has already been fully initialized. 685 * 686 * XXX: 687 * - The decision to return void and thus require this function to 688 * succeed is questionable. 689 * - We should probably do more sanity checking. For instance we don't 690 * do anything to insure if_xname is unique or non-empty. 691 */ 692void 693if_attach(struct ifnet *ifp) 694{ 695 696 if_attach_internal(ifp, 0); 697} 698 699static void 700if_attach_internal(struct ifnet *ifp, int vmove) 701{ 702 INIT_VNET_NET(curvnet); 703 unsigned socksize, ifasize; 704 int namelen, masklen; 705 struct sockaddr_dl *sdl; 706 struct ifaddr *ifa; 707 708 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 709 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 710 ifp->if_xname); 711 712#ifdef VIMAGE 713 ifp->if_vnet = curvnet; 714 if (ifp->if_home_vnet == NULL) 715 ifp->if_home_vnet = curvnet; 716#endif 717 718 if_addgroup(ifp, IFG_ALL); 719 720 getmicrotime(&ifp->if_lastchange); 721 ifp->if_data.ifi_epoch = time_uptime; 722 ifp->if_data.ifi_datalen = sizeof(struct if_data); 723 724 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 725 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 726 ("transmit and qflush must both either be set or both be NULL")); 727 if (ifp->if_transmit == NULL) { 728 ifp->if_transmit = if_transmit; 729 ifp->if_qflush = if_qflush; 730 } 731 732 if (!vmove) { 733#ifdef MAC 734 mac_ifnet_create(ifp); 735#endif 736 737 if (IS_DEFAULT_VNET(curvnet)) { 738 ifdev_setbyindex(ifp->if_index, make_dev(&net_cdevsw, 739 ifp->if_index, UID_ROOT, GID_WHEEL, 0600, "%s/%s", 740 net_cdevsw.d_name, ifp->if_xname)); 741 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d", 742 net_cdevsw.d_name, ifp->if_index); 743 } 744 745 ifq_attach(&ifp->if_snd, ifp); 746 747 /* 748 * Create a Link Level name for this device. 749 */ 750 namelen = strlen(ifp->if_xname); 751 /* 752 * Always save enough space for any possiable name so we 753 * can do a rename in place later. 754 */ 755 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 756 socksize = masklen + ifp->if_addrlen; 757 if (socksize < sizeof(*sdl)) 758 socksize = sizeof(*sdl); 759 socksize = roundup2(socksize, sizeof(long)); 760 ifasize = sizeof(*ifa) + 2 * socksize; 761 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 762 IFA_LOCK_INIT(ifa); 763 sdl = (struct sockaddr_dl *)(ifa + 1); 764 sdl->sdl_len = socksize; 765 sdl->sdl_family = AF_LINK; 766 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 767 sdl->sdl_nlen = namelen; 768 sdl->sdl_index = ifp->if_index; 769 sdl->sdl_type = ifp->if_type; 770 ifp->if_addr = ifa; 771 ifa->ifa_ifp = ifp; 772 ifa->ifa_rtrequest = link_rtrequest; 773 ifa->ifa_addr = (struct sockaddr *)sdl; 774 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 775 ifa->ifa_netmask = (struct sockaddr *)sdl; 776 sdl->sdl_len = masklen; 777 while (namelen != 0) 778 sdl->sdl_data[--namelen] = 0xff; 779 ifa->ifa_refcnt = 1; 780 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 781 /* Reliably crash if used uninitialized. */ 782 ifp->if_broadcastaddr = NULL; 783 } 784 785 IFNET_WLOCK(); 786 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 787#ifdef VIMAGE 788 curvnet->ifcnt++; 789#endif 790 IFNET_WUNLOCK(); 791 792 if (domain_init_status >= 2) 793 if_attachdomain1(ifp); 794 795 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 796 if (IS_DEFAULT_VNET(curvnet)) 797 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 798 799 /* Announce the interface. */ 800 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 801 802 if (!vmove && ifp->if_watchdog != NULL) { 803 if_printf(ifp, 804 "WARNING: using obsoleted if_watchdog interface\n"); 805 806 /* 807 * Note that we need if_slowtimo(). If this happens after 808 * boot, then call if_slowtimo() directly. 809 */ 810 if (atomic_cmpset_int(&slowtimo_started, 0, 1) && !cold) 811 if_slowtimo(0); 812 } 813} 814 815static void 816if_attachdomain(void *dummy) 817{ 818 INIT_VNET_NET(curvnet); 819 struct ifnet *ifp; 820 int s; 821 822 s = splnet(); 823 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 824 if_attachdomain1(ifp); 825 splx(s); 826} 827SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 828 if_attachdomain, NULL); 829 830static void 831if_attachdomain1(struct ifnet *ifp) 832{ 833 struct domain *dp; 834 int s; 835 836 s = splnet(); 837 838 /* 839 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 840 * cannot lock ifp->if_afdata initialization, entirely. 841 */ 842 if (IF_AFDATA_TRYLOCK(ifp) == 0) { 843 splx(s); 844 return; 845 } 846 if (ifp->if_afdata_initialized >= domain_init_status) { 847 IF_AFDATA_UNLOCK(ifp); 848 splx(s); 849 printf("if_attachdomain called more than once on %s\n", 850 ifp->if_xname); 851 return; 852 } 853 ifp->if_afdata_initialized = domain_init_status; 854 IF_AFDATA_UNLOCK(ifp); 855 856 /* address family dependent data region */ 857 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 858 for (dp = domains; dp; dp = dp->dom_next) { 859 if (dp->dom_ifattach) 860 ifp->if_afdata[dp->dom_family] = 861 (*dp->dom_ifattach)(ifp); 862 } 863 864 splx(s); 865} 866 867/* 868 * Remove any unicast or broadcast network addresses from an interface. 869 */ 870void 871if_purgeaddrs(struct ifnet *ifp) 872{ 873 struct ifaddr *ifa, *next; 874 875 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 876 if (ifa->ifa_addr->sa_family == AF_LINK) 877 continue; 878#ifdef INET 879 /* XXX: Ugly!! ad hoc just for INET */ 880 if (ifa->ifa_addr->sa_family == AF_INET) { 881 struct ifaliasreq ifr; 882 883 bzero(&ifr, sizeof(ifr)); 884 ifr.ifra_addr = *ifa->ifa_addr; 885 if (ifa->ifa_dstaddr) 886 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 887 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 888 NULL) == 0) 889 continue; 890 } 891#endif /* INET */ 892#ifdef INET6 893 if (ifa->ifa_addr->sa_family == AF_INET6) { 894 in6_purgeaddr(ifa); 895 /* ifp_addrhead is already updated */ 896 continue; 897 } 898#endif /* INET6 */ 899 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 900 IFAFREE(ifa); 901 } 902} 903 904/* 905 * Remove any multicast network addresses from an interface. 906 */ 907void 908if_purgemaddrs(struct ifnet *ifp) 909{ 910 struct ifmultiaddr *ifma; 911 struct ifmultiaddr *next; 912 913 IF_ADDR_LOCK(ifp); 914 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 915 if_delmulti_locked(ifp, ifma, 1); 916 IF_ADDR_UNLOCK(ifp); 917} 918 919/* 920 * Detach an interface, removing it from the list of "active" interfaces. 921 * If vmove flag is set on entry to if_detach_internal(), perform only a 922 * limited subset of cleanup tasks, given that we are moving an ifnet from 923 * one vnet to another, where it must be fully operational. 924 * 925 * XXXRW: There are some significant questions about event ordering, and 926 * how to prevent things from starting to use the interface during detach. 927 */ 928void 929if_detach(struct ifnet *ifp) 930{ 931 932 if_detach_internal(ifp, 0); 933} 934 935static void 936if_detach_internal(struct ifnet *ifp, int vmove) 937{ 938 INIT_VNET_NET(ifp->if_vnet); 939 struct ifaddr *ifa; 940 struct radix_node_head *rnh; 941 int i, j; 942 struct domain *dp; 943 struct ifnet *iter; 944 int found = 0; 945 946 IFNET_WLOCK(); 947 TAILQ_FOREACH(iter, &V_ifnet, if_link) 948 if (iter == ifp) { 949 TAILQ_REMOVE(&V_ifnet, ifp, if_link); 950 found = 1; 951 break; 952 } 953#ifdef VIMAGE 954 if (found) 955 curvnet->ifcnt--; 956#endif 957 IFNET_WUNLOCK(); 958 if (!found) { 959 if (vmove) 960 panic("interface not in it's own ifnet list"); 961 else 962 return; /* XXX this should panic as well? */ 963 } 964 965 /* 966 * Remove/wait for pending events. 967 */ 968 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 969 970 /* 971 * Remove routes and flush queues. 972 */ 973 if_down(ifp); 974#ifdef ALTQ 975 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 976 altq_disable(&ifp->if_snd); 977 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 978 altq_detach(&ifp->if_snd); 979#endif 980 981 if_purgeaddrs(ifp); 982 983#ifdef INET 984 in_ifdetach(ifp); 985#endif 986 987#ifdef INET6 988 /* 989 * Remove all IPv6 kernel structs related to ifp. This should be done 990 * before removing routing entries below, since IPv6 interface direct 991 * routes are expected to be removed by the IPv6-specific kernel API. 992 * Otherwise, the kernel will detect some inconsistency and bark it. 993 */ 994 in6_ifdetach(ifp); 995#endif 996 if_purgemaddrs(ifp); 997 998 if (!vmove) { 999 /* 1000 * Prevent further calls into the device driver via ifnet. 1001 */ 1002 if_dead(ifp); 1003 1004 /* 1005 * Remove link ifaddr pointer and maybe decrement if_index. 1006 * Clean up all addresses. 1007 */ 1008 ifp->if_addr = NULL; 1009 if (IS_DEFAULT_VNET(curvnet)) 1010 destroy_dev(ifdev_byindex(ifp->if_index)); 1011 ifdev_setbyindex(ifp->if_index, NULL); 1012 1013 /* We can now free link ifaddr. */ 1014 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 1015 ifa = TAILQ_FIRST(&ifp->if_addrhead); 1016 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 1017 IFAFREE(ifa); 1018 } 1019 } 1020 1021 /* 1022 * Delete all remaining routes using this interface 1023 * Unfortuneatly the only way to do this is to slog through 1024 * the entire routing table looking for routes which point 1025 * to this interface...oh well... 1026 */ 1027 for (i = 1; i <= AF_MAX; i++) { 1028 for (j = 0; j < rt_numfibs; j++) { 1029 rnh = rt_tables_get_rnh(j, i); 1030 if (rnh == NULL) 1031 continue; 1032 RADIX_NODE_HEAD_LOCK(rnh); 1033 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 1034 RADIX_NODE_HEAD_UNLOCK(rnh); 1035 } 1036 } 1037 1038 /* Announce that the interface is gone. */ 1039 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1040 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1041 if (IS_DEFAULT_VNET(curvnet)) 1042 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1043 if_delgroups(ifp); 1044 1045 IF_AFDATA_LOCK(ifp); 1046 for (dp = domains; dp; dp = dp->dom_next) { 1047 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 1048 (*dp->dom_ifdetach)(ifp, 1049 ifp->if_afdata[dp->dom_family]); 1050 } 1051 ifp->if_afdata_initialized = 0; 1052 IF_AFDATA_UNLOCK(ifp); 1053} 1054 1055#ifdef VIMAGE 1056/* 1057 * if_vmove() performs a limited version of if_detach() in current 1058 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1059 * An attempt is made to shrink if_index in current vnet, find an 1060 * unused if_index in target vnet and calls if_grow() if necessary, 1061 * and finally find an unused if_xname for the target vnet. 1062 */ 1063void 1064if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1065{ 1066 1067 /* 1068 * Detach from current vnet, but preserve LLADDR info, do not 1069 * mark as dead etc. so that the ifnet can be reattached later. 1070 */ 1071 if_detach_internal(ifp, 1); 1072 1073 /* 1074 * Unlink the ifnet from ifindex_table[] in current vnet, 1075 * and shrink the if_index for that vnet if possible. 1076 * do / while construct below is needed to confine the scope 1077 * of INIT_VNET_NET(). 1078 */ 1079 { 1080 INIT_VNET_NET(curvnet); 1081 1082 IFNET_WLOCK(); 1083 ifnet_setbyindex(ifp->if_index, NULL); 1084 while (V_if_index > 0 && \ 1085 ifnet_byindex_locked(V_if_index) == NULL) 1086 V_if_index--; 1087 IFNET_WUNLOCK(); 1088 }; 1089 1090 /* 1091 * Switch to the context of the target vnet. 1092 */ 1093 CURVNET_SET_QUIET(new_vnet); 1094 INIT_VNET_NET(new_vnet); 1095 1096 /* 1097 * Try to find an empty slot below if_index. If we fail, take 1098 * the next slot. 1099 */ 1100 IFNET_WLOCK(); 1101 for (ifp->if_index = 1; ifp->if_index <= V_if_index; ifp->if_index++) { 1102 if (ifnet_byindex_locked(ifp->if_index) == NULL) 1103 break; 1104 } 1105 /* Catch if_index overflow. */ 1106 if (ifp->if_index < 1) 1107 panic("if_index overflow"); 1108 1109 if (ifp->if_index > V_if_index) 1110 V_if_index = ifp->if_index; 1111 if (V_if_index >= V_if_indexlim) 1112 if_grow(); 1113 ifnet_setbyindex(ifp->if_index, ifp); 1114 IFNET_WUNLOCK(); 1115 1116 if_attach_internal(ifp, 1); 1117 1118 CURVNET_RESTORE(); 1119} 1120#endif /* VIMAGE */ 1121 1122/* 1123 * Add a group to an interface 1124 */ 1125int 1126if_addgroup(struct ifnet *ifp, const char *groupname) 1127{ 1128 INIT_VNET_NET(ifp->if_vnet); 1129 struct ifg_list *ifgl; 1130 struct ifg_group *ifg = NULL; 1131 struct ifg_member *ifgm; 1132 1133 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1134 groupname[strlen(groupname) - 1] <= '9') 1135 return (EINVAL); 1136 1137 IFNET_WLOCK(); 1138 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1139 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1140 IFNET_WUNLOCK(); 1141 return (EEXIST); 1142 } 1143 1144 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1145 M_NOWAIT)) == NULL) { 1146 IFNET_WUNLOCK(); 1147 return (ENOMEM); 1148 } 1149 1150 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1151 M_TEMP, M_NOWAIT)) == NULL) { 1152 free(ifgl, M_TEMP); 1153 IFNET_WUNLOCK(); 1154 return (ENOMEM); 1155 } 1156 1157 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1158 if (!strcmp(ifg->ifg_group, groupname)) 1159 break; 1160 1161 if (ifg == NULL) { 1162 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1163 M_TEMP, M_NOWAIT)) == NULL) { 1164 free(ifgl, M_TEMP); 1165 free(ifgm, M_TEMP); 1166 IFNET_WUNLOCK(); 1167 return (ENOMEM); 1168 } 1169 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1170 ifg->ifg_refcnt = 0; 1171 TAILQ_INIT(&ifg->ifg_members); 1172 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1173 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1174 } 1175 1176 ifg->ifg_refcnt++; 1177 ifgl->ifgl_group = ifg; 1178 ifgm->ifgm_ifp = ifp; 1179 1180 IF_ADDR_LOCK(ifp); 1181 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1182 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1183 IF_ADDR_UNLOCK(ifp); 1184 1185 IFNET_WUNLOCK(); 1186 1187 EVENTHANDLER_INVOKE(group_change_event, groupname); 1188 1189 return (0); 1190} 1191 1192/* 1193 * Remove a group from an interface 1194 */ 1195int 1196if_delgroup(struct ifnet *ifp, const char *groupname) 1197{ 1198 INIT_VNET_NET(ifp->if_vnet); 1199 struct ifg_list *ifgl; 1200 struct ifg_member *ifgm; 1201 1202 IFNET_WLOCK(); 1203 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1204 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1205 break; 1206 if (ifgl == NULL) { 1207 IFNET_WUNLOCK(); 1208 return (ENOENT); 1209 } 1210 1211 IF_ADDR_LOCK(ifp); 1212 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1213 IF_ADDR_UNLOCK(ifp); 1214 1215 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1216 if (ifgm->ifgm_ifp == ifp) 1217 break; 1218 1219 if (ifgm != NULL) { 1220 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 1221 free(ifgm, M_TEMP); 1222 } 1223 1224 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1225 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1226 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1227 free(ifgl->ifgl_group, M_TEMP); 1228 } 1229 IFNET_WUNLOCK(); 1230 1231 free(ifgl, M_TEMP); 1232 1233 EVENTHANDLER_INVOKE(group_change_event, groupname); 1234 1235 return (0); 1236} 1237 1238/* 1239 * Remove an interface from all groups 1240 */ 1241static void 1242if_delgroups(struct ifnet *ifp) 1243{ 1244 INIT_VNET_NET(ifp->if_vnet); 1245 struct ifg_list *ifgl; 1246 struct ifg_member *ifgm; 1247 char groupname[IFNAMSIZ]; 1248 1249 IFNET_WLOCK(); 1250 while (!TAILQ_EMPTY(&ifp->if_groups)) { 1251 ifgl = TAILQ_FIRST(&ifp->if_groups); 1252 1253 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1254 1255 IF_ADDR_LOCK(ifp); 1256 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1257 IF_ADDR_UNLOCK(ifp); 1258 1259 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1260 if (ifgm->ifgm_ifp == ifp) 1261 break; 1262 1263 if (ifgm != NULL) { 1264 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1265 ifgm_next); 1266 free(ifgm, M_TEMP); 1267 } 1268 1269 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1270 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1271 EVENTHANDLER_INVOKE(group_detach_event, 1272 ifgl->ifgl_group); 1273 free(ifgl->ifgl_group, M_TEMP); 1274 } 1275 IFNET_WUNLOCK(); 1276 1277 free(ifgl, M_TEMP); 1278 1279 EVENTHANDLER_INVOKE(group_change_event, groupname); 1280 1281 IFNET_WLOCK(); 1282 } 1283 IFNET_WUNLOCK(); 1284} 1285 1286/* 1287 * Stores all groups from an interface in memory pointed 1288 * to by data 1289 */ 1290static int 1291if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 1292{ 1293 int len, error; 1294 struct ifg_list *ifgl; 1295 struct ifg_req ifgrq, *ifgp; 1296 struct ifgroupreq *ifgr = data; 1297 1298 if (ifgr->ifgr_len == 0) { 1299 IF_ADDR_LOCK(ifp); 1300 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1301 ifgr->ifgr_len += sizeof(struct ifg_req); 1302 IF_ADDR_UNLOCK(ifp); 1303 return (0); 1304 } 1305 1306 len = ifgr->ifgr_len; 1307 ifgp = ifgr->ifgr_groups; 1308 /* XXX: wire */ 1309 IF_ADDR_LOCK(ifp); 1310 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1311 if (len < sizeof(ifgrq)) { 1312 IF_ADDR_UNLOCK(ifp); 1313 return (EINVAL); 1314 } 1315 bzero(&ifgrq, sizeof ifgrq); 1316 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1317 sizeof(ifgrq.ifgrq_group)); 1318 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1319 IF_ADDR_UNLOCK(ifp); 1320 return (error); 1321 } 1322 len -= sizeof(ifgrq); 1323 ifgp++; 1324 } 1325 IF_ADDR_UNLOCK(ifp); 1326 1327 return (0); 1328} 1329 1330/* 1331 * Stores all members of a group in memory pointed to by data 1332 */ 1333static int 1334if_getgroupmembers(struct ifgroupreq *data) 1335{ 1336 INIT_VNET_NET(curvnet); 1337 struct ifgroupreq *ifgr = data; 1338 struct ifg_group *ifg; 1339 struct ifg_member *ifgm; 1340 struct ifg_req ifgrq, *ifgp; 1341 int len, error; 1342 1343 IFNET_RLOCK(); 1344 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1345 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1346 break; 1347 if (ifg == NULL) { 1348 IFNET_RUNLOCK(); 1349 return (ENOENT); 1350 } 1351 1352 if (ifgr->ifgr_len == 0) { 1353 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1354 ifgr->ifgr_len += sizeof(ifgrq); 1355 IFNET_RUNLOCK(); 1356 return (0); 1357 } 1358 1359 len = ifgr->ifgr_len; 1360 ifgp = ifgr->ifgr_groups; 1361 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1362 if (len < sizeof(ifgrq)) { 1363 IFNET_RUNLOCK(); 1364 return (EINVAL); 1365 } 1366 bzero(&ifgrq, sizeof ifgrq); 1367 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1368 sizeof(ifgrq.ifgrq_member)); 1369 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1370 IFNET_RUNLOCK(); 1371 return (error); 1372 } 1373 len -= sizeof(ifgrq); 1374 ifgp++; 1375 } 1376 IFNET_RUNLOCK(); 1377 1378 return (0); 1379} 1380 1381/* 1382 * Delete Routes for a Network Interface 1383 * 1384 * Called for each routing entry via the rnh->rnh_walktree() call above 1385 * to delete all route entries referencing a detaching network interface. 1386 * 1387 * Arguments: 1388 * rn pointer to node in the routing table 1389 * arg argument passed to rnh->rnh_walktree() - detaching interface 1390 * 1391 * Returns: 1392 * 0 successful 1393 * errno failed - reason indicated 1394 * 1395 */ 1396static int 1397if_rtdel(struct radix_node *rn, void *arg) 1398{ 1399 struct rtentry *rt = (struct rtentry *)rn; 1400 struct ifnet *ifp = arg; 1401 int err; 1402 1403 if (rt->rt_ifp == ifp) { 1404 1405 /* 1406 * Protect (sorta) against walktree recursion problems 1407 * with cloned routes 1408 */ 1409 if ((rt->rt_flags & RTF_UP) == 0) 1410 return (0); 1411 1412 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway, 1413 rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED, 1414 (struct rtentry **) NULL, rt->rt_fibnum); 1415 if (err) { 1416 log(LOG_WARNING, "if_rtdel: error %d\n", err); 1417 } 1418 } 1419 1420 return (0); 1421} 1422 1423/* 1424 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1425 * structs used to represent other address families, it is necessary 1426 * to perform a different comparison. 1427 */ 1428 1429#define sa_equal(a1, a2) \ 1430 (bcmp((a1), (a2), ((a1))->sa_len) == 0) 1431 1432#define sa_dl_equal(a1, a2) \ 1433 ((((struct sockaddr_dl *)(a1))->sdl_len == \ 1434 ((struct sockaddr_dl *)(a2))->sdl_len) && \ 1435 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \ 1436 LLADDR((struct sockaddr_dl *)(a2)), \ 1437 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1438 1439/* 1440 * Locate an interface based on a complete address. 1441 */ 1442/*ARGSUSED*/ 1443struct ifaddr * 1444ifa_ifwithaddr(struct sockaddr *addr) 1445{ 1446 INIT_VNET_NET(curvnet); 1447 struct ifnet *ifp; 1448 struct ifaddr *ifa; 1449 1450 IFNET_RLOCK(); 1451 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1452 IF_ADDR_LOCK(ifp); 1453 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1454 if (ifa->ifa_addr->sa_family != addr->sa_family) 1455 continue; 1456 if (sa_equal(addr, ifa->ifa_addr)) { 1457 IF_ADDR_UNLOCK(ifp); 1458 goto done; 1459 } 1460 /* IP6 doesn't have broadcast */ 1461 if ((ifp->if_flags & IFF_BROADCAST) && 1462 ifa->ifa_broadaddr && 1463 ifa->ifa_broadaddr->sa_len != 0 && 1464 sa_equal(ifa->ifa_broadaddr, addr)) { 1465 IF_ADDR_UNLOCK(ifp); 1466 goto done; 1467 } 1468 } 1469 IF_ADDR_UNLOCK(ifp); 1470 } 1471 ifa = NULL; 1472done: 1473 IFNET_RUNLOCK(); 1474 return (ifa); 1475} 1476 1477/* 1478 * Locate an interface based on the broadcast address. 1479 */ 1480/* ARGSUSED */ 1481struct ifaddr * 1482ifa_ifwithbroadaddr(struct sockaddr *addr) 1483{ 1484 INIT_VNET_NET(curvnet); 1485 struct ifnet *ifp; 1486 struct ifaddr *ifa; 1487 1488 IFNET_RLOCK(); 1489 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1490 IF_ADDR_LOCK(ifp); 1491 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1492 if (ifa->ifa_addr->sa_family != addr->sa_family) 1493 continue; 1494 if ((ifp->if_flags & IFF_BROADCAST) && 1495 ifa->ifa_broadaddr && 1496 ifa->ifa_broadaddr->sa_len != 0 && 1497 sa_equal(ifa->ifa_broadaddr, addr)) { 1498 IF_ADDR_UNLOCK(ifp); 1499 goto done; 1500 } 1501 } 1502 IF_ADDR_UNLOCK(ifp); 1503 } 1504 ifa = NULL; 1505done: 1506 IFNET_RUNLOCK(); 1507 return (ifa); 1508} 1509 1510/* 1511 * Locate the point to point interface with a given destination address. 1512 */ 1513/*ARGSUSED*/ 1514struct ifaddr * 1515ifa_ifwithdstaddr(struct sockaddr *addr) 1516{ 1517 INIT_VNET_NET(curvnet); 1518 struct ifnet *ifp; 1519 struct ifaddr *ifa; 1520 1521 IFNET_RLOCK(); 1522 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1523 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1524 continue; 1525 IF_ADDR_LOCK(ifp); 1526 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1527 if (ifa->ifa_addr->sa_family != addr->sa_family) 1528 continue; 1529 if (ifa->ifa_dstaddr != NULL && 1530 sa_equal(addr, ifa->ifa_dstaddr)) { 1531 IF_ADDR_UNLOCK(ifp); 1532 goto done; 1533 } 1534 } 1535 IF_ADDR_UNLOCK(ifp); 1536 } 1537 ifa = NULL; 1538done: 1539 IFNET_RUNLOCK(); 1540 return (ifa); 1541} 1542 1543/* 1544 * Find an interface on a specific network. If many, choice 1545 * is most specific found. 1546 */ 1547struct ifaddr * 1548ifa_ifwithnet(struct sockaddr *addr) 1549{ 1550 INIT_VNET_NET(curvnet); 1551 struct ifnet *ifp; 1552 struct ifaddr *ifa; 1553 struct ifaddr *ifa_maybe = (struct ifaddr *) 0; 1554 u_int af = addr->sa_family; 1555 char *addr_data = addr->sa_data, *cplim; 1556 1557 /* 1558 * AF_LINK addresses can be looked up directly by their index number, 1559 * so do that if we can. 1560 */ 1561 if (af == AF_LINK) { 1562 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1563 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1564 return (ifaddr_byindex(sdl->sdl_index)); 1565 } 1566 1567 /* 1568 * Scan though each interface, looking for ones that have 1569 * addresses in this address family. 1570 */ 1571 IFNET_RLOCK(); 1572 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1573 IF_ADDR_LOCK(ifp); 1574 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1575 char *cp, *cp2, *cp3; 1576 1577 if (ifa->ifa_addr->sa_family != af) 1578next: continue; 1579 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) { 1580 /* 1581 * This is a bit broken as it doesn't 1582 * take into account that the remote end may 1583 * be a single node in the network we are 1584 * looking for. 1585 * The trouble is that we don't know the 1586 * netmask for the remote end. 1587 */ 1588 if (ifa->ifa_dstaddr != NULL && 1589 sa_equal(addr, ifa->ifa_dstaddr)) { 1590 IF_ADDR_UNLOCK(ifp); 1591 goto done; 1592 } 1593 } else { 1594 /* 1595 * if we have a special address handler, 1596 * then use it instead of the generic one. 1597 */ 1598 if (ifa->ifa_claim_addr) { 1599 if ((*ifa->ifa_claim_addr)(ifa, addr)) { 1600 IF_ADDR_UNLOCK(ifp); 1601 goto done; 1602 } 1603 continue; 1604 } 1605 1606 /* 1607 * Scan all the bits in the ifa's address. 1608 * If a bit dissagrees with what we are 1609 * looking for, mask it with the netmask 1610 * to see if it really matters. 1611 * (A byte at a time) 1612 */ 1613 if (ifa->ifa_netmask == 0) 1614 continue; 1615 cp = addr_data; 1616 cp2 = ifa->ifa_addr->sa_data; 1617 cp3 = ifa->ifa_netmask->sa_data; 1618 cplim = ifa->ifa_netmask->sa_len 1619 + (char *)ifa->ifa_netmask; 1620 while (cp3 < cplim) 1621 if ((*cp++ ^ *cp2++) & *cp3++) 1622 goto next; /* next address! */ 1623 /* 1624 * If the netmask of what we just found 1625 * is more specific than what we had before 1626 * (if we had one) then remember the new one 1627 * before continuing to search 1628 * for an even better one. 1629 */ 1630 if (ifa_maybe == 0 || 1631 rn_refines((caddr_t)ifa->ifa_netmask, 1632 (caddr_t)ifa_maybe->ifa_netmask)) 1633 ifa_maybe = ifa; 1634 } 1635 } 1636 IF_ADDR_UNLOCK(ifp); 1637 } 1638 ifa = ifa_maybe; 1639done: 1640 IFNET_RUNLOCK(); 1641 return (ifa); 1642} 1643 1644/* 1645 * Find an interface address specific to an interface best matching 1646 * a given address. 1647 */ 1648struct ifaddr * 1649ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1650{ 1651 struct ifaddr *ifa; 1652 char *cp, *cp2, *cp3; 1653 char *cplim; 1654 struct ifaddr *ifa_maybe = 0; 1655 u_int af = addr->sa_family; 1656 1657 if (af >= AF_MAX) 1658 return (0); 1659 IF_ADDR_LOCK(ifp); 1660 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1661 if (ifa->ifa_addr->sa_family != af) 1662 continue; 1663 if (ifa_maybe == 0) 1664 ifa_maybe = ifa; 1665 if (ifa->ifa_netmask == 0) { 1666 if (sa_equal(addr, ifa->ifa_addr) || 1667 (ifa->ifa_dstaddr && 1668 sa_equal(addr, ifa->ifa_dstaddr))) 1669 goto done; 1670 continue; 1671 } 1672 if (ifp->if_flags & IFF_POINTOPOINT) { 1673 if (sa_equal(addr, ifa->ifa_dstaddr)) 1674 goto done; 1675 } else { 1676 cp = addr->sa_data; 1677 cp2 = ifa->ifa_addr->sa_data; 1678 cp3 = ifa->ifa_netmask->sa_data; 1679 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1680 for (; cp3 < cplim; cp3++) 1681 if ((*cp++ ^ *cp2++) & *cp3) 1682 break; 1683 if (cp3 == cplim) 1684 goto done; 1685 } 1686 } 1687 ifa = ifa_maybe; 1688done: 1689 IF_ADDR_UNLOCK(ifp); 1690 return (ifa); 1691} 1692 1693#include <net/route.h> 1694#include <net/if_llatbl.h> 1695 1696/* 1697 * Default action when installing a route with a Link Level gateway. 1698 * Lookup an appropriate real ifa to point to. 1699 * This should be moved to /sys/net/link.c eventually. 1700 */ 1701static void 1702link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 1703{ 1704 struct ifaddr *ifa, *oifa; 1705 struct sockaddr *dst; 1706 struct ifnet *ifp; 1707 1708 RT_LOCK_ASSERT(rt); 1709 1710 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 1711 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 1712 return; 1713 ifa = ifaof_ifpforaddr(dst, ifp); 1714 if (ifa) { 1715 IFAREF(ifa); /* XXX */ 1716 oifa = rt->rt_ifa; 1717 rt->rt_ifa = ifa; 1718 IFAFREE(oifa); 1719 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1720 ifa->ifa_rtrequest(cmd, rt, info); 1721 } 1722} 1723 1724/* 1725 * Mark an interface down and notify protocols of 1726 * the transition. 1727 * NOTE: must be called at splnet or eqivalent. 1728 */ 1729static void 1730if_unroute(struct ifnet *ifp, int flag, int fam) 1731{ 1732 struct ifaddr *ifa; 1733 1734 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 1735 1736 ifp->if_flags &= ~flag; 1737 getmicrotime(&ifp->if_lastchange); 1738 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1739 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1740 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1741 ifp->if_qflush(ifp); 1742 1743#if defined(INET) || defined(INET6) 1744#ifdef DEV_CARP 1745 if (ifp->if_carp) 1746 carp_carpdev_state(ifp->if_carp); 1747#endif 1748#endif 1749 rt_ifmsg(ifp); 1750} 1751 1752/* 1753 * Mark an interface up and notify protocols of 1754 * the transition. 1755 * NOTE: must be called at splnet or eqivalent. 1756 */ 1757static void 1758if_route(struct ifnet *ifp, int flag, int fam) 1759{ 1760 struct ifaddr *ifa; 1761 1762 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 1763 1764 ifp->if_flags |= flag; 1765 getmicrotime(&ifp->if_lastchange); 1766 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1767 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1768 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1769#if defined(INET) || defined(INET6) 1770#ifdef DEV_CARP 1771 if (ifp->if_carp) 1772 carp_carpdev_state(ifp->if_carp); 1773#endif 1774#endif 1775 rt_ifmsg(ifp); 1776#ifdef INET6 1777 in6_if_up(ifp); 1778#endif 1779} 1780 1781void (*vlan_link_state_p)(struct ifnet *, int); /* XXX: private from if_vlan */ 1782void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1783 1784/* 1785 * Handle a change in the interface link state. To avoid LORs 1786 * between driver lock and upper layer locks, as well as possible 1787 * recursions, we post event to taskqueue, and all job 1788 * is done in static do_link_state_change(). 1789 */ 1790void 1791if_link_state_change(struct ifnet *ifp, int link_state) 1792{ 1793 /* Return if state hasn't changed. */ 1794 if (ifp->if_link_state == link_state) 1795 return; 1796 1797 ifp->if_link_state = link_state; 1798 1799 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 1800} 1801 1802static void 1803do_link_state_change(void *arg, int pending) 1804{ 1805 struct ifnet *ifp = (struct ifnet *)arg; 1806 int link_state = ifp->if_link_state; 1807 int link; 1808 CURVNET_SET(ifp->if_vnet); 1809 1810 /* Notify that the link state has changed. */ 1811 rt_ifmsg(ifp); 1812 if (link_state == LINK_STATE_UP) 1813 link = NOTE_LINKUP; 1814 else if (link_state == LINK_STATE_DOWN) 1815 link = NOTE_LINKDOWN; 1816 else 1817 link = NOTE_LINKINV; 1818 KNOTE_UNLOCKED(&ifp->if_klist, link); 1819 if (ifp->if_vlantrunk != NULL) 1820 (*vlan_link_state_p)(ifp, link); 1821 1822 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 1823 IFP2AC(ifp)->ac_netgraph != NULL) 1824 (*ng_ether_link_state_p)(ifp, link_state); 1825#if defined(INET) || defined(INET6) 1826#ifdef DEV_CARP 1827 if (ifp->if_carp) 1828 carp_carpdev_state(ifp->if_carp); 1829#endif 1830#endif 1831 if (ifp->if_bridge) { 1832 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!")); 1833 (*bstp_linkstate_p)(ifp, link_state); 1834 } 1835 if (ifp->if_lagg) { 1836 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!")); 1837 (*lagg_linkstate_p)(ifp, link_state); 1838 } 1839 1840 if (IS_DEFAULT_VNET(curvnet)) 1841 devctl_notify("IFNET", ifp->if_xname, 1842 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 1843 NULL); 1844 if (pending > 1) 1845 if_printf(ifp, "%d link states coalesced\n", pending); 1846 if (log_link_state_change) 1847 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 1848 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 1849 CURVNET_RESTORE(); 1850} 1851 1852/* 1853 * Mark an interface down and notify protocols of 1854 * the transition. 1855 * NOTE: must be called at splnet or eqivalent. 1856 */ 1857void 1858if_down(struct ifnet *ifp) 1859{ 1860 1861 if_unroute(ifp, IFF_UP, AF_UNSPEC); 1862} 1863 1864/* 1865 * Mark an interface up and notify protocols of 1866 * the transition. 1867 * NOTE: must be called at splnet or eqivalent. 1868 */ 1869void 1870if_up(struct ifnet *ifp) 1871{ 1872 1873 if_route(ifp, IFF_UP, AF_UNSPEC); 1874} 1875 1876/* 1877 * Flush an interface queue. 1878 */ 1879void 1880if_qflush(struct ifnet *ifp) 1881{ 1882 struct mbuf *m, *n; 1883 struct ifaltq *ifq; 1884 1885 ifq = &ifp->if_snd; 1886 IFQ_LOCK(ifq); 1887#ifdef ALTQ 1888 if (ALTQ_IS_ENABLED(ifq)) 1889 ALTQ_PURGE(ifq); 1890#endif 1891 n = ifq->ifq_head; 1892 while ((m = n) != 0) { 1893 n = m->m_act; 1894 m_freem(m); 1895 } 1896 ifq->ifq_head = 0; 1897 ifq->ifq_tail = 0; 1898 ifq->ifq_len = 0; 1899 IFQ_UNLOCK(ifq); 1900} 1901 1902/* 1903 * Handle interface watchdog timer routines. Called 1904 * from softclock, we decrement timers (if set) and 1905 * call the appropriate interface routine on expiration. 1906 * 1907 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called 1908 * holding Giant. 1909 */ 1910static void 1911if_slowtimo(void *arg) 1912{ 1913 VNET_ITERATOR_DECL(vnet_iter); 1914 struct ifnet *ifp; 1915 int s = splimp(); 1916 1917 IFNET_RLOCK(); 1918 VNET_LIST_RLOCK(); 1919 VNET_FOREACH(vnet_iter) { 1920 CURVNET_SET(vnet_iter); 1921 INIT_VNET_NET(vnet_iter); 1922 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1923 if (ifp->if_timer == 0 || --ifp->if_timer) 1924 continue; 1925 if (ifp->if_watchdog) 1926 (*ifp->if_watchdog)(ifp); 1927 } 1928 CURVNET_RESTORE(); 1929 } 1930 VNET_LIST_RUNLOCK(); 1931 IFNET_RUNLOCK(); 1932 splx(s); 1933 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ); 1934} 1935 1936/* 1937 * Map interface name to interface structure pointer, with or without 1938 * returning a reference. 1939 */ 1940struct ifnet * 1941ifunit_ref(const char *name) 1942{ 1943 INIT_VNET_NET(curvnet); 1944 struct ifnet *ifp; 1945 1946 IFNET_RLOCK(); 1947 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1948 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 1949 !(ifp->if_flags & IFF_DYING)) 1950 break; 1951 } 1952 if (ifp != NULL) 1953 if_ref(ifp); 1954 IFNET_RUNLOCK(); 1955 return (ifp); 1956} 1957 1958struct ifnet * 1959ifunit(const char *name) 1960{ 1961 INIT_VNET_NET(curvnet); 1962 struct ifnet *ifp; 1963 1964 IFNET_RLOCK(); 1965 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1966 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 1967 break; 1968 } 1969 IFNET_RUNLOCK(); 1970 return (ifp); 1971} 1972 1973/* 1974 * Hardware specific interface ioctls. 1975 */ 1976static int 1977ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 1978{ 1979 struct ifreq *ifr; 1980 struct ifstat *ifs; 1981 int error = 0; 1982 int new_flags, temp_flags; 1983 size_t namelen, onamelen; 1984 char new_name[IFNAMSIZ]; 1985 struct ifaddr *ifa; 1986 struct sockaddr_dl *sdl; 1987 1988 ifr = (struct ifreq *)data; 1989 switch (cmd) { 1990 case SIOCGIFINDEX: 1991 ifr->ifr_index = ifp->if_index; 1992 break; 1993 1994 case SIOCGIFFLAGS: 1995 temp_flags = ifp->if_flags | ifp->if_drv_flags; 1996 ifr->ifr_flags = temp_flags & 0xffff; 1997 ifr->ifr_flagshigh = temp_flags >> 16; 1998 break; 1999 2000 case SIOCGIFCAP: 2001 ifr->ifr_reqcap = ifp->if_capabilities; 2002 ifr->ifr_curcap = ifp->if_capenable; 2003 break; 2004 2005#ifdef MAC 2006 case SIOCGIFMAC: 2007 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2008 break; 2009#endif 2010 2011 case SIOCGIFMETRIC: 2012 ifr->ifr_metric = ifp->if_metric; 2013 break; 2014 2015 case SIOCGIFMTU: 2016 ifr->ifr_mtu = ifp->if_mtu; 2017 break; 2018 2019 case SIOCGIFPHYS: 2020 ifr->ifr_phys = ifp->if_physical; 2021 break; 2022 2023 case SIOCSIFFLAGS: 2024 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2025 if (error) 2026 return (error); 2027 /* 2028 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2029 * check, so we don't need special handling here yet. 2030 */ 2031 new_flags = (ifr->ifr_flags & 0xffff) | 2032 (ifr->ifr_flagshigh << 16); 2033 if (ifp->if_flags & IFF_SMART) { 2034 /* Smart drivers twiddle their own routes */ 2035 } else if (ifp->if_flags & IFF_UP && 2036 (new_flags & IFF_UP) == 0) { 2037 int s = splimp(); 2038 if_down(ifp); 2039 splx(s); 2040 } else if (new_flags & IFF_UP && 2041 (ifp->if_flags & IFF_UP) == 0) { 2042 int s = splimp(); 2043 if_up(ifp); 2044 splx(s); 2045 } 2046 /* See if permanently promiscuous mode bit is about to flip */ 2047 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2048 if (new_flags & IFF_PPROMISC) 2049 ifp->if_flags |= IFF_PROMISC; 2050 else if (ifp->if_pcount == 0) 2051 ifp->if_flags &= ~IFF_PROMISC; 2052 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2053 ifp->if_xname, 2054 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); 2055 } 2056 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2057 (new_flags &~ IFF_CANTCHANGE); 2058 if (ifp->if_ioctl) { 2059 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2060 } 2061 getmicrotime(&ifp->if_lastchange); 2062 break; 2063 2064 case SIOCSIFCAP: 2065 error = priv_check(td, PRIV_NET_SETIFCAP); 2066 if (error) 2067 return (error); 2068 if (ifp->if_ioctl == NULL) 2069 return (EOPNOTSUPP); 2070 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2071 return (EINVAL); 2072 error = (*ifp->if_ioctl)(ifp, cmd, data); 2073 if (error == 0) 2074 getmicrotime(&ifp->if_lastchange); 2075 break; 2076 2077#ifdef MAC 2078 case SIOCSIFMAC: 2079 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2080 break; 2081#endif 2082 2083 case SIOCSIFNAME: 2084 error = priv_check(td, PRIV_NET_SETIFNAME); 2085 if (error) 2086 return (error); 2087 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 2088 if (error != 0) 2089 return (error); 2090 if (new_name[0] == '\0') 2091 return (EINVAL); 2092 if (ifunit(new_name) != NULL) 2093 return (EEXIST); 2094 2095 /* Announce the departure of the interface. */ 2096 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2097 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2098 2099 log(LOG_INFO, "%s: changing name to '%s'\n", 2100 ifp->if_xname, new_name); 2101 2102 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2103 ifa = ifp->if_addr; 2104 IFA_LOCK(ifa); 2105 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2106 namelen = strlen(new_name); 2107 onamelen = sdl->sdl_nlen; 2108 /* 2109 * Move the address if needed. This is safe because we 2110 * allocate space for a name of length IFNAMSIZ when we 2111 * create this in if_attach(). 2112 */ 2113 if (namelen != onamelen) { 2114 bcopy(sdl->sdl_data + onamelen, 2115 sdl->sdl_data + namelen, sdl->sdl_alen); 2116 } 2117 bcopy(new_name, sdl->sdl_data, namelen); 2118 sdl->sdl_nlen = namelen; 2119 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2120 bzero(sdl->sdl_data, onamelen); 2121 while (namelen != 0) 2122 sdl->sdl_data[--namelen] = 0xff; 2123 IFA_UNLOCK(ifa); 2124 2125 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2126 /* Announce the return of the interface. */ 2127 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2128 break; 2129 2130#ifdef VIMAGE 2131 case SIOCSIFVNET: 2132 error = priv_check(td, PRIV_NET_SETIFVNET); 2133 if (error) 2134 return (error); 2135 error = vi_if_move(td, ifp, ifr->ifr_name, ifr->ifr_jid, NULL); 2136 break; 2137#endif 2138 2139 case SIOCSIFMETRIC: 2140 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2141 if (error) 2142 return (error); 2143 ifp->if_metric = ifr->ifr_metric; 2144 getmicrotime(&ifp->if_lastchange); 2145 break; 2146 2147 case SIOCSIFPHYS: 2148 error = priv_check(td, PRIV_NET_SETIFPHYS); 2149 if (error) 2150 return (error); 2151 if (ifp->if_ioctl == NULL) 2152 return (EOPNOTSUPP); 2153 error = (*ifp->if_ioctl)(ifp, cmd, data); 2154 if (error == 0) 2155 getmicrotime(&ifp->if_lastchange); 2156 break; 2157 2158 case SIOCSIFMTU: 2159 { 2160 u_long oldmtu = ifp->if_mtu; 2161 2162 error = priv_check(td, PRIV_NET_SETIFMTU); 2163 if (error) 2164 return (error); 2165 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2166 return (EINVAL); 2167 if (ifp->if_ioctl == NULL) 2168 return (EOPNOTSUPP); 2169 error = (*ifp->if_ioctl)(ifp, cmd, data); 2170 if (error == 0) { 2171 getmicrotime(&ifp->if_lastchange); 2172 rt_ifmsg(ifp); 2173 } 2174 /* 2175 * If the link MTU changed, do network layer specific procedure. 2176 */ 2177 if (ifp->if_mtu != oldmtu) { 2178#ifdef INET6 2179 nd6_setmtu(ifp); 2180#endif 2181 } 2182 break; 2183 } 2184 2185 case SIOCADDMULTI: 2186 case SIOCDELMULTI: 2187 if (cmd == SIOCADDMULTI) 2188 error = priv_check(td, PRIV_NET_ADDMULTI); 2189 else 2190 error = priv_check(td, PRIV_NET_DELMULTI); 2191 if (error) 2192 return (error); 2193 2194 /* Don't allow group membership on non-multicast interfaces. */ 2195 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2196 return (EOPNOTSUPP); 2197 2198 /* Don't let users screw up protocols' entries. */ 2199 if (ifr->ifr_addr.sa_family != AF_LINK) 2200 return (EINVAL); 2201 2202 if (cmd == SIOCADDMULTI) { 2203 struct ifmultiaddr *ifma; 2204 2205 /* 2206 * Userland is only permitted to join groups once 2207 * via the if_addmulti() KPI, because it cannot hold 2208 * struct ifmultiaddr * between calls. It may also 2209 * lose a race while we check if the membership 2210 * already exists. 2211 */ 2212 IF_ADDR_LOCK(ifp); 2213 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2214 IF_ADDR_UNLOCK(ifp); 2215 if (ifma != NULL) 2216 error = EADDRINUSE; 2217 else 2218 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2219 } else { 2220 error = if_delmulti(ifp, &ifr->ifr_addr); 2221 } 2222 if (error == 0) 2223 getmicrotime(&ifp->if_lastchange); 2224 break; 2225 2226 case SIOCSIFPHYADDR: 2227 case SIOCDIFPHYADDR: 2228#ifdef INET6 2229 case SIOCSIFPHYADDR_IN6: 2230#endif 2231 case SIOCSLIFPHYADDR: 2232 case SIOCSIFMEDIA: 2233 case SIOCSIFGENERIC: 2234 error = priv_check(td, PRIV_NET_HWIOCTL); 2235 if (error) 2236 return (error); 2237 if (ifp->if_ioctl == NULL) 2238 return (EOPNOTSUPP); 2239 error = (*ifp->if_ioctl)(ifp, cmd, data); 2240 if (error == 0) 2241 getmicrotime(&ifp->if_lastchange); 2242 break; 2243 2244 case SIOCGIFSTATUS: 2245 ifs = (struct ifstat *)data; 2246 ifs->ascii[0] = '\0'; 2247 2248 case SIOCGIFPSRCADDR: 2249 case SIOCGIFPDSTADDR: 2250 case SIOCGLIFPHYADDR: 2251 case SIOCGIFMEDIA: 2252 case SIOCGIFGENERIC: 2253 if (ifp->if_ioctl == NULL) 2254 return (EOPNOTSUPP); 2255 error = (*ifp->if_ioctl)(ifp, cmd, data); 2256 break; 2257 2258 case SIOCSIFLLADDR: 2259 error = priv_check(td, PRIV_NET_SETLLADDR); 2260 if (error) 2261 return (error); 2262 error = if_setlladdr(ifp, 2263 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2264 break; 2265 2266 case SIOCAIFGROUP: 2267 { 2268 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2269 2270 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2271 if (error) 2272 return (error); 2273 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2274 return (error); 2275 break; 2276 } 2277 2278 case SIOCGIFGROUP: 2279 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2280 return (error); 2281 break; 2282 2283 case SIOCDIFGROUP: 2284 { 2285 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2286 2287 error = priv_check(td, PRIV_NET_DELIFGROUP); 2288 if (error) 2289 return (error); 2290 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2291 return (error); 2292 break; 2293 } 2294 2295 default: 2296 error = ENOIOCTL; 2297 break; 2298 } 2299 return (error); 2300} 2301 2302/* 2303 * Interface ioctls. 2304 */ 2305int 2306ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2307{ 2308 struct ifnet *ifp; 2309 struct ifreq *ifr; 2310 int error; 2311 int oif_flags; 2312 2313 switch (cmd) { 2314 case SIOCGIFCONF: 2315 case OSIOCGIFCONF: 2316#ifdef __amd64__ 2317 case SIOCGIFCONF32: 2318#endif 2319 return (ifconf(cmd, data)); 2320 } 2321 ifr = (struct ifreq *)data; 2322 2323 switch (cmd) { 2324#ifdef VIMAGE 2325 case SIOCSIFRVNET: 2326 error = priv_check(td, PRIV_NET_SETIFVNET); 2327 if (error) 2328 return (error); 2329 return (vi_if_move(td, NULL, ifr->ifr_name, ifr->ifr_jid, 2330 NULL)); 2331 /* 2332 * XXX vnet creation will be implemented through the new jail 2333 * framework - this is just a temporary hack for testing the 2334 * vnet create / destroy mechanisms. 2335 */ 2336 case SIOCSIFVIMAGE: 2337 error = vi_if_move(td, NULL, NULL, 0, (struct vi_req *) data); 2338 return (error); 2339 case SIOCSPVIMAGE: 2340 case SIOCGPVIMAGE: 2341 error = vi_td_ioctl(cmd, (struct vi_req *) data, td); 2342 return (error); 2343#endif 2344 case SIOCIFCREATE: 2345 case SIOCIFCREATE2: 2346 error = priv_check(td, PRIV_NET_IFCREATE); 2347 if (error) 2348 return (error); 2349 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name), 2350 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL)); 2351 case SIOCIFDESTROY: 2352 error = priv_check(td, PRIV_NET_IFDESTROY); 2353 if (error) 2354 return (error); 2355 return if_clone_destroy(ifr->ifr_name); 2356 2357 case SIOCIFGCLONERS: 2358 return (if_clone_list((struct if_clonereq *)data)); 2359 case SIOCGIFGMEMB: 2360 return (if_getgroupmembers((struct ifgroupreq *)data)); 2361 } 2362 2363 ifp = ifunit_ref(ifr->ifr_name); 2364 if (ifp == NULL) 2365 return (ENXIO); 2366 2367 error = ifhwioctl(cmd, ifp, data, td); 2368 if (error != ENOIOCTL) { 2369 if_rele(ifp); 2370 return (error); 2371 } 2372 2373 oif_flags = ifp->if_flags; 2374 if (so->so_proto == NULL) { 2375 if_rele(ifp); 2376 return (EOPNOTSUPP); 2377 } 2378#ifndef COMPAT_43 2379 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 2380 data, 2381 ifp, td)); 2382 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL) 2383 error = (*ifp->if_ioctl)(ifp, cmd, data); 2384#else 2385 { 2386 int ocmd = cmd; 2387 2388 switch (cmd) { 2389 2390 case SIOCSIFDSTADDR: 2391 case SIOCSIFADDR: 2392 case SIOCSIFBRDADDR: 2393 case SIOCSIFNETMASK: 2394#if BYTE_ORDER != BIG_ENDIAN 2395 if (ifr->ifr_addr.sa_family == 0 && 2396 ifr->ifr_addr.sa_len < 16) { 2397 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 2398 ifr->ifr_addr.sa_len = 16; 2399 } 2400#else 2401 if (ifr->ifr_addr.sa_len == 0) 2402 ifr->ifr_addr.sa_len = 16; 2403#endif 2404 break; 2405 2406 case OSIOCGIFADDR: 2407 cmd = SIOCGIFADDR; 2408 break; 2409 2410 case OSIOCGIFDSTADDR: 2411 cmd = SIOCGIFDSTADDR; 2412 break; 2413 2414 case OSIOCGIFBRDADDR: 2415 cmd = SIOCGIFBRDADDR; 2416 break; 2417 2418 case OSIOCGIFNETMASK: 2419 cmd = SIOCGIFNETMASK; 2420 } 2421 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 2422 cmd, 2423 data, 2424 ifp, td)); 2425 if (error == EOPNOTSUPP && ifp != NULL && 2426 ifp->if_ioctl != NULL) 2427 error = (*ifp->if_ioctl)(ifp, cmd, data); 2428 switch (ocmd) { 2429 2430 case OSIOCGIFADDR: 2431 case OSIOCGIFDSTADDR: 2432 case OSIOCGIFBRDADDR: 2433 case OSIOCGIFNETMASK: 2434 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 2435 2436 } 2437 } 2438#endif /* COMPAT_43 */ 2439 2440 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2441#ifdef INET6 2442 DELAY(100);/* XXX: temporary workaround for fxp issue*/ 2443 if (ifp->if_flags & IFF_UP) { 2444 int s = splimp(); 2445 in6_if_up(ifp); 2446 splx(s); 2447 } 2448#endif 2449 } 2450 if_rele(ifp); 2451 return (error); 2452} 2453 2454/* 2455 * The code common to handling reference counted flags, 2456 * e.g., in ifpromisc() and if_allmulti(). 2457 * The "pflag" argument can specify a permanent mode flag to check, 2458 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2459 * 2460 * Only to be used on stack-owned flags, not driver-owned flags. 2461 */ 2462static int 2463if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 2464{ 2465 struct ifreq ifr; 2466 int error; 2467 int oldflags, oldcount; 2468 2469 /* Sanity checks to catch programming errors */ 2470 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 2471 ("%s: setting driver-owned flag %d", __func__, flag)); 2472 2473 if (onswitch) 2474 KASSERT(*refcount >= 0, 2475 ("%s: increment negative refcount %d for flag %d", 2476 __func__, *refcount, flag)); 2477 else 2478 KASSERT(*refcount > 0, 2479 ("%s: decrement non-positive refcount %d for flag %d", 2480 __func__, *refcount, flag)); 2481 2482 /* In case this mode is permanent, just touch refcount */ 2483 if (ifp->if_flags & pflag) { 2484 *refcount += onswitch ? 1 : -1; 2485 return (0); 2486 } 2487 2488 /* Save ifnet parameters for if_ioctl() may fail */ 2489 oldcount = *refcount; 2490 oldflags = ifp->if_flags; 2491 2492 /* 2493 * See if we aren't the only and touching refcount is enough. 2494 * Actually toggle interface flag if we are the first or last. 2495 */ 2496 if (onswitch) { 2497 if ((*refcount)++) 2498 return (0); 2499 ifp->if_flags |= flag; 2500 } else { 2501 if (--(*refcount)) 2502 return (0); 2503 ifp->if_flags &= ~flag; 2504 } 2505 2506 /* Call down the driver since we've changed interface flags */ 2507 if (ifp->if_ioctl == NULL) { 2508 error = EOPNOTSUPP; 2509 goto recover; 2510 } 2511 ifr.ifr_flags = ifp->if_flags & 0xffff; 2512 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2513 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2514 if (error) 2515 goto recover; 2516 /* Notify userland that interface flags have changed */ 2517 rt_ifmsg(ifp); 2518 return (0); 2519 2520recover: 2521 /* Recover after driver error */ 2522 *refcount = oldcount; 2523 ifp->if_flags = oldflags; 2524 return (error); 2525} 2526 2527/* 2528 * Set/clear promiscuous mode on interface ifp based on the truth value 2529 * of pswitch. The calls are reference counted so that only the first 2530 * "on" request actually has an effect, as does the final "off" request. 2531 * Results are undefined if the "off" and "on" requests are not matched. 2532 */ 2533int 2534ifpromisc(struct ifnet *ifp, int pswitch) 2535{ 2536 int error; 2537 int oldflags = ifp->if_flags; 2538 2539 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 2540 &ifp->if_pcount, pswitch); 2541 /* If promiscuous mode status has changed, log a message */ 2542 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) 2543 log(LOG_INFO, "%s: promiscuous mode %s\n", 2544 ifp->if_xname, 2545 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 2546 return (error); 2547} 2548 2549/* 2550 * Return interface configuration 2551 * of system. List may be used 2552 * in later ioctl's (above) to get 2553 * other information. 2554 */ 2555/*ARGSUSED*/ 2556static int 2557ifconf(u_long cmd, caddr_t data) 2558{ 2559 INIT_VNET_NET(curvnet); 2560 struct ifconf *ifc = (struct ifconf *)data; 2561#ifdef __amd64__ 2562 struct ifconf32 *ifc32 = (struct ifconf32 *)data; 2563 struct ifconf ifc_swab; 2564#endif 2565 struct ifnet *ifp; 2566 struct ifaddr *ifa; 2567 struct ifreq ifr; 2568 struct sbuf *sb; 2569 int error, full = 0, valid_len, max_len; 2570 2571#ifdef __amd64__ 2572 if (cmd == SIOCGIFCONF32) { 2573 ifc_swab.ifc_len = ifc32->ifc_len; 2574 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf; 2575 ifc = &ifc_swab; 2576 } 2577#endif 2578 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2579 max_len = MAXPHYS - 1; 2580 2581 /* Prevent hostile input from being able to crash the system */ 2582 if (ifc->ifc_len <= 0) 2583 return (EINVAL); 2584 2585again: 2586 if (ifc->ifc_len <= max_len) { 2587 max_len = ifc->ifc_len; 2588 full = 1; 2589 } 2590 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2591 max_len = 0; 2592 valid_len = 0; 2593 2594 IFNET_RLOCK(); /* could sleep XXX */ 2595 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2596 int addrs; 2597 2598 /* 2599 * Zero the ifr_name buffer to make sure we don't 2600 * disclose the contents of the stack. 2601 */ 2602 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 2603 2604 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 2605 >= sizeof(ifr.ifr_name)) { 2606 sbuf_delete(sb); 2607 IFNET_RUNLOCK(); 2608 return (ENAMETOOLONG); 2609 } 2610 2611 addrs = 0; 2612 IF_ADDR_LOCK(ifp); 2613 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2614 struct sockaddr *sa = ifa->ifa_addr; 2615 2616 if (prison_if(curthread->td_ucred, sa) != 0) 2617 continue; 2618 addrs++; 2619#ifdef COMPAT_43 2620 if (cmd == OSIOCGIFCONF) { 2621 struct osockaddr *osa = 2622 (struct osockaddr *)&ifr.ifr_addr; 2623 ifr.ifr_addr = *sa; 2624 osa->sa_family = sa->sa_family; 2625 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2626 max_len += sizeof(ifr); 2627 } else 2628#endif 2629 if (sa->sa_len <= sizeof(*sa)) { 2630 ifr.ifr_addr = *sa; 2631 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2632 max_len += sizeof(ifr); 2633 } else { 2634 sbuf_bcat(sb, &ifr, 2635 offsetof(struct ifreq, ifr_addr)); 2636 max_len += offsetof(struct ifreq, ifr_addr); 2637 sbuf_bcat(sb, sa, sa->sa_len); 2638 max_len += sa->sa_len; 2639 } 2640 2641 if (!sbuf_overflowed(sb)) 2642 valid_len = sbuf_len(sb); 2643 } 2644 IF_ADDR_UNLOCK(ifp); 2645 if (addrs == 0) { 2646 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 2647 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2648 max_len += sizeof(ifr); 2649 2650 if (!sbuf_overflowed(sb)) 2651 valid_len = sbuf_len(sb); 2652 } 2653 } 2654 IFNET_RUNLOCK(); 2655 2656 /* 2657 * If we didn't allocate enough space (uncommon), try again. If 2658 * we have already allocated as much space as we are allowed, 2659 * return what we've got. 2660 */ 2661 if (valid_len != max_len && !full) { 2662 sbuf_delete(sb); 2663 goto again; 2664 } 2665 2666 ifc->ifc_len = valid_len; 2667#ifdef __amd64__ 2668 if (cmd == SIOCGIFCONF32) 2669 ifc32->ifc_len = valid_len; 2670#endif 2671 sbuf_finish(sb); 2672 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 2673 sbuf_delete(sb); 2674 return (error); 2675} 2676 2677/* 2678 * Just like ifpromisc(), but for all-multicast-reception mode. 2679 */ 2680int 2681if_allmulti(struct ifnet *ifp, int onswitch) 2682{ 2683 2684 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 2685} 2686 2687struct ifmultiaddr * 2688if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 2689{ 2690 struct ifmultiaddr *ifma; 2691 2692 IF_ADDR_LOCK_ASSERT(ifp); 2693 2694 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2695 if (sa->sa_family == AF_LINK) { 2696 if (sa_dl_equal(ifma->ifma_addr, sa)) 2697 break; 2698 } else { 2699 if (sa_equal(ifma->ifma_addr, sa)) 2700 break; 2701 } 2702 } 2703 2704 return ifma; 2705} 2706 2707/* 2708 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 2709 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 2710 * the ifnet multicast address list here, so the caller must do that and 2711 * other setup work (such as notifying the device driver). The reference 2712 * count is initialized to 1. 2713 */ 2714static struct ifmultiaddr * 2715if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 2716 int mflags) 2717{ 2718 struct ifmultiaddr *ifma; 2719 struct sockaddr *dupsa; 2720 2721 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 2722 M_ZERO); 2723 if (ifma == NULL) 2724 return (NULL); 2725 2726 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 2727 if (dupsa == NULL) { 2728 free(ifma, M_IFMADDR); 2729 return (NULL); 2730 } 2731 bcopy(sa, dupsa, sa->sa_len); 2732 ifma->ifma_addr = dupsa; 2733 2734 ifma->ifma_ifp = ifp; 2735 ifma->ifma_refcount = 1; 2736 ifma->ifma_protospec = NULL; 2737 2738 if (llsa == NULL) { 2739 ifma->ifma_lladdr = NULL; 2740 return (ifma); 2741 } 2742 2743 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 2744 if (dupsa == NULL) { 2745 free(ifma->ifma_addr, M_IFMADDR); 2746 free(ifma, M_IFMADDR); 2747 return (NULL); 2748 } 2749 bcopy(llsa, dupsa, llsa->sa_len); 2750 ifma->ifma_lladdr = dupsa; 2751 2752 return (ifma); 2753} 2754 2755/* 2756 * if_freemulti: free ifmultiaddr structure and possibly attached related 2757 * addresses. The caller is responsible for implementing reference 2758 * counting, notifying the driver, handling routing messages, and releasing 2759 * any dependent link layer state. 2760 */ 2761static void 2762if_freemulti(struct ifmultiaddr *ifma) 2763{ 2764 2765 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 2766 ifma->ifma_refcount)); 2767 KASSERT(ifma->ifma_protospec == NULL, 2768 ("if_freemulti: protospec not NULL")); 2769 2770 if (ifma->ifma_lladdr != NULL) 2771 free(ifma->ifma_lladdr, M_IFMADDR); 2772 free(ifma->ifma_addr, M_IFMADDR); 2773 free(ifma, M_IFMADDR); 2774} 2775 2776/* 2777 * Register an additional multicast address with a network interface. 2778 * 2779 * - If the address is already present, bump the reference count on the 2780 * address and return. 2781 * - If the address is not link-layer, look up a link layer address. 2782 * - Allocate address structures for one or both addresses, and attach to the 2783 * multicast address list on the interface. If automatically adding a link 2784 * layer address, the protocol address will own a reference to the link 2785 * layer address, to be freed when it is freed. 2786 * - Notify the network device driver of an addition to the multicast address 2787 * list. 2788 * 2789 * 'sa' points to caller-owned memory with the desired multicast address. 2790 * 2791 * 'retifma' will be used to return a pointer to the resulting multicast 2792 * address reference, if desired. 2793 */ 2794int 2795if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 2796 struct ifmultiaddr **retifma) 2797{ 2798 struct ifmultiaddr *ifma, *ll_ifma; 2799 struct sockaddr *llsa; 2800 int error; 2801 2802 /* 2803 * If the address is already present, return a new reference to it; 2804 * otherwise, allocate storage and set up a new address. 2805 */ 2806 IF_ADDR_LOCK(ifp); 2807 ifma = if_findmulti(ifp, sa); 2808 if (ifma != NULL) { 2809 ifma->ifma_refcount++; 2810 if (retifma != NULL) 2811 *retifma = ifma; 2812 IF_ADDR_UNLOCK(ifp); 2813 return (0); 2814 } 2815 2816 /* 2817 * The address isn't already present; resolve the protocol address 2818 * into a link layer address, and then look that up, bump its 2819 * refcount or allocate an ifma for that also. If 'llsa' was 2820 * returned, we will need to free it later. 2821 */ 2822 llsa = NULL; 2823 ll_ifma = NULL; 2824 if (ifp->if_resolvemulti != NULL) { 2825 error = ifp->if_resolvemulti(ifp, &llsa, sa); 2826 if (error) 2827 goto unlock_out; 2828 } 2829 2830 /* 2831 * Allocate the new address. Don't hook it up yet, as we may also 2832 * need to allocate a link layer multicast address. 2833 */ 2834 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 2835 if (ifma == NULL) { 2836 error = ENOMEM; 2837 goto free_llsa_out; 2838 } 2839 2840 /* 2841 * If a link layer address is found, we'll need to see if it's 2842 * already present in the address list, or allocate is as well. 2843 * When this block finishes, the link layer address will be on the 2844 * list. 2845 */ 2846 if (llsa != NULL) { 2847 ll_ifma = if_findmulti(ifp, llsa); 2848 if (ll_ifma == NULL) { 2849 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 2850 if (ll_ifma == NULL) { 2851 --ifma->ifma_refcount; 2852 if_freemulti(ifma); 2853 error = ENOMEM; 2854 goto free_llsa_out; 2855 } 2856 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 2857 ifma_link); 2858 } else 2859 ll_ifma->ifma_refcount++; 2860 ifma->ifma_llifma = ll_ifma; 2861 } 2862 2863 /* 2864 * We now have a new multicast address, ifma, and possibly a new or 2865 * referenced link layer address. Add the primary address to the 2866 * ifnet address list. 2867 */ 2868 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 2869 2870 if (retifma != NULL) 2871 *retifma = ifma; 2872 2873 /* 2874 * Must generate the message while holding the lock so that 'ifma' 2875 * pointer is still valid. 2876 */ 2877 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 2878 IF_ADDR_UNLOCK(ifp); 2879 2880 /* 2881 * We are certain we have added something, so call down to the 2882 * interface to let them know about it. 2883 */ 2884 if (ifp->if_ioctl != NULL) { 2885 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 2886 } 2887 2888 if (llsa != NULL) 2889 free(llsa, M_IFMADDR); 2890 2891 return (0); 2892 2893free_llsa_out: 2894 if (llsa != NULL) 2895 free(llsa, M_IFMADDR); 2896 2897unlock_out: 2898 IF_ADDR_UNLOCK(ifp); 2899 return (error); 2900} 2901 2902/* 2903 * Delete a multicast group membership by network-layer group address. 2904 * 2905 * Returns ENOENT if the entry could not be found. If ifp no longer 2906 * exists, results are undefined. This entry point should only be used 2907 * from subsystems which do appropriate locking to hold ifp for the 2908 * duration of the call. 2909 * Network-layer protocol domains must use if_delmulti_ifma(). 2910 */ 2911int 2912if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 2913{ 2914 struct ifmultiaddr *ifma; 2915 int lastref; 2916#ifdef INVARIANTS 2917 struct ifnet *oifp; 2918 INIT_VNET_NET(ifp->if_vnet); 2919 2920 IFNET_RLOCK(); 2921 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 2922 if (ifp == oifp) 2923 break; 2924 if (ifp != oifp) 2925 ifp = NULL; 2926 IFNET_RUNLOCK(); 2927 2928 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 2929#endif 2930 if (ifp == NULL) 2931 return (ENOENT); 2932 2933 IF_ADDR_LOCK(ifp); 2934 lastref = 0; 2935 ifma = if_findmulti(ifp, sa); 2936 if (ifma != NULL) 2937 lastref = if_delmulti_locked(ifp, ifma, 0); 2938 IF_ADDR_UNLOCK(ifp); 2939 2940 if (ifma == NULL) 2941 return (ENOENT); 2942 2943 if (lastref && ifp->if_ioctl != NULL) { 2944 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 2945 } 2946 2947 return (0); 2948} 2949 2950/* 2951 * Delete a multicast group membership by group membership pointer. 2952 * Network-layer protocol domains must use this routine. 2953 * 2954 * It is safe to call this routine if the ifp disappeared. 2955 */ 2956void 2957if_delmulti_ifma(struct ifmultiaddr *ifma) 2958{ 2959#ifdef DIAGNOSTIC 2960 INIT_VNET_NET(curvnet); 2961#endif 2962 struct ifnet *ifp; 2963 int lastref; 2964 2965 ifp = ifma->ifma_ifp; 2966#ifdef DIAGNOSTIC 2967 if (ifp == NULL) { 2968 printf("%s: ifma_ifp seems to be detached\n", __func__); 2969 } else { 2970 struct ifnet *oifp; 2971 2972 IFNET_RLOCK(); 2973 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 2974 if (ifp == oifp) 2975 break; 2976 if (ifp != oifp) { 2977 printf("%s: ifnet %p disappeared\n", __func__, ifp); 2978 ifp = NULL; 2979 } 2980 IFNET_RUNLOCK(); 2981 } 2982#endif 2983 /* 2984 * If and only if the ifnet instance exists: Acquire the address lock. 2985 */ 2986 if (ifp != NULL) 2987 IF_ADDR_LOCK(ifp); 2988 2989 lastref = if_delmulti_locked(ifp, ifma, 0); 2990 2991 if (ifp != NULL) { 2992 /* 2993 * If and only if the ifnet instance exists: 2994 * Release the address lock. 2995 * If the group was left: update the hardware hash filter. 2996 */ 2997 IF_ADDR_UNLOCK(ifp); 2998 if (lastref && ifp->if_ioctl != NULL) { 2999 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3000 } 3001 } 3002} 3003 3004/* 3005 * Perform deletion of network-layer and/or link-layer multicast address. 3006 * 3007 * Return 0 if the reference count was decremented. 3008 * Return 1 if the final reference was released, indicating that the 3009 * hardware hash filter should be reprogrammed. 3010 */ 3011static int 3012if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3013{ 3014 struct ifmultiaddr *ll_ifma; 3015 3016 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3017 KASSERT(ifma->ifma_ifp == ifp, 3018 ("%s: inconsistent ifp %p", __func__, ifp)); 3019 IF_ADDR_LOCK_ASSERT(ifp); 3020 } 3021 3022 ifp = ifma->ifma_ifp; 3023 3024 /* 3025 * If the ifnet is detaching, null out references to ifnet, 3026 * so that upper protocol layers will notice, and not attempt 3027 * to obtain locks for an ifnet which no longer exists. The 3028 * routing socket announcement must happen before the ifnet 3029 * instance is detached from the system. 3030 */ 3031 if (detaching) { 3032#ifdef DIAGNOSTIC 3033 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3034#endif 3035 /* 3036 * ifp may already be nulled out if we are being reentered 3037 * to delete the ll_ifma. 3038 */ 3039 if (ifp != NULL) { 3040 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3041 ifma->ifma_ifp = NULL; 3042 } 3043 } 3044 3045 if (--ifma->ifma_refcount > 0) 3046 return 0; 3047 3048 /* 3049 * If this ifma is a network-layer ifma, a link-layer ifma may 3050 * have been associated with it. Release it first if so. 3051 */ 3052 ll_ifma = ifma->ifma_llifma; 3053 if (ll_ifma != NULL) { 3054 KASSERT(ifma->ifma_lladdr != NULL, 3055 ("%s: llifma w/o lladdr", __func__)); 3056 if (detaching) 3057 ll_ifma->ifma_ifp = NULL; /* XXX */ 3058 if (--ll_ifma->ifma_refcount == 0) { 3059 if (ifp != NULL) { 3060 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3061 ifma_link); 3062 } 3063 if_freemulti(ll_ifma); 3064 } 3065 } 3066 3067 if (ifp != NULL) 3068 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3069 3070 if_freemulti(ifma); 3071 3072 /* 3073 * The last reference to this instance of struct ifmultiaddr 3074 * was released; the hardware should be notified of this change. 3075 */ 3076 return 1; 3077} 3078 3079/* 3080 * Set the link layer address on an interface. 3081 * 3082 * At this time we only support certain types of interfaces, 3083 * and we don't allow the length of the address to change. 3084 */ 3085int 3086if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3087{ 3088 struct sockaddr_dl *sdl; 3089 struct ifaddr *ifa; 3090 struct ifreq ifr; 3091 3092 ifa = ifp->if_addr; 3093 if (ifa == NULL) 3094 return (EINVAL); 3095 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3096 if (sdl == NULL) 3097 return (EINVAL); 3098 if (len != sdl->sdl_alen) /* don't allow length to change */ 3099 return (EINVAL); 3100 switch (ifp->if_type) { 3101 case IFT_ETHER: 3102 case IFT_FDDI: 3103 case IFT_XETHER: 3104 case IFT_ISO88025: 3105 case IFT_L2VLAN: 3106 case IFT_BRIDGE: 3107 case IFT_ARCNET: 3108 case IFT_IEEE8023ADLAG: 3109 case IFT_IEEE80211: 3110 bcopy(lladdr, LLADDR(sdl), len); 3111 break; 3112 default: 3113 return (ENODEV); 3114 } 3115 /* 3116 * If the interface is already up, we need 3117 * to re-init it in order to reprogram its 3118 * address filter. 3119 */ 3120 if ((ifp->if_flags & IFF_UP) != 0) { 3121 if (ifp->if_ioctl) { 3122 ifp->if_flags &= ~IFF_UP; 3123 ifr.ifr_flags = ifp->if_flags & 0xffff; 3124 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3125 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3126 ifp->if_flags |= IFF_UP; 3127 ifr.ifr_flags = ifp->if_flags & 0xffff; 3128 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3129 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3130 } 3131#ifdef INET 3132 /* 3133 * Also send gratuitous ARPs to notify other nodes about 3134 * the address change. 3135 */ 3136 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3137 if (ifa->ifa_addr->sa_family == AF_INET) 3138 arp_ifinit(ifp, ifa); 3139 } 3140#endif 3141 } 3142 return (0); 3143} 3144 3145/* 3146 * The name argument must be a pointer to storage which will last as 3147 * long as the interface does. For physical devices, the result of 3148 * device_get_name(dev) is a good choice and for pseudo-devices a 3149 * static string works well. 3150 */ 3151void 3152if_initname(struct ifnet *ifp, const char *name, int unit) 3153{ 3154 ifp->if_dname = name; 3155 ifp->if_dunit = unit; 3156 if (unit != IF_DUNIT_NONE) 3157 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3158 else 3159 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3160} 3161 3162int 3163if_printf(struct ifnet *ifp, const char * fmt, ...) 3164{ 3165 va_list ap; 3166 int retval; 3167 3168 retval = printf("%s: ", ifp->if_xname); 3169 va_start(ap, fmt); 3170 retval += vprintf(fmt, ap); 3171 va_end(ap); 3172 return (retval); 3173} 3174 3175void 3176if_start(struct ifnet *ifp) 3177{ 3178 3179 (*(ifp)->if_start)(ifp); 3180} 3181 3182/* 3183 * Backwards compatibility interface for drivers 3184 * that have not implemented it 3185 */ 3186static int 3187if_transmit(struct ifnet *ifp, struct mbuf *m) 3188{ 3189 int error; 3190 3191 IFQ_HANDOFF(ifp, m, error); 3192 return (error); 3193} 3194 3195int 3196if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3197{ 3198 int active = 0; 3199 3200 IF_LOCK(ifq); 3201 if (_IF_QFULL(ifq)) { 3202 _IF_DROP(ifq); 3203 IF_UNLOCK(ifq); 3204 m_freem(m); 3205 return (0); 3206 } 3207 if (ifp != NULL) { 3208 ifp->if_obytes += m->m_pkthdr.len + adjust; 3209 if (m->m_flags & (M_BCAST|M_MCAST)) 3210 ifp->if_omcasts++; 3211 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3212 } 3213 _IF_ENQUEUE(ifq, m); 3214 IF_UNLOCK(ifq); 3215 if (ifp != NULL && !active) 3216 (*(ifp)->if_start)(ifp); 3217 return (1); 3218} 3219 3220void 3221if_register_com_alloc(u_char type, 3222 if_com_alloc_t *a, if_com_free_t *f) 3223{ 3224 3225 KASSERT(if_com_alloc[type] == NULL, 3226 ("if_register_com_alloc: %d already registered", type)); 3227 KASSERT(if_com_free[type] == NULL, 3228 ("if_register_com_alloc: %d free already registered", type)); 3229 3230 if_com_alloc[type] = a; 3231 if_com_free[type] = f; 3232} 3233 3234void 3235if_deregister_com_alloc(u_char type) 3236{ 3237 3238 KASSERT(if_com_alloc[type] != NULL, 3239 ("if_deregister_com_alloc: %d not registered", type)); 3240 KASSERT(if_com_free[type] != NULL, 3241 ("if_deregister_com_alloc: %d free not registered", type)); 3242 if_com_alloc[type] = NULL; 3243 if_com_free[type] = NULL; 3244} 3245