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