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 37 #include <sys/param.h> 38 #include <sys/types.h> 39 #include <sys/conf.h> 40 #include <sys/malloc.h> 41 #include <sys/sbuf.h> 42 #include <sys/bus.h> 43 #include <sys/mbuf.h> 44 #include <sys/systm.h> 45 #include <sys/priv.h> 46 #include <sys/proc.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/protosw.h> 50 #include <sys/kernel.h> 51 #include <sys/lock.h> 52 #include <sys/refcount.h> 53 #include <sys/module.h> 54 #include <sys/rwlock.h> 55 #include <sys/sockio.h> 56 #include <sys/syslog.h> 57 #include <sys/sysctl.h> 58 #include <sys/taskqueue.h> 59 #include <sys/domain.h> 60 #include <sys/jail.h> 61 #include <sys/priv.h> 62 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 #include <netinet/in.h> 78 #include <netinet/in_var.h> 79 #include <netinet/ip.h> 80 #include <netinet/ip_carp.h> 81 #ifdef INET 82 #include <netinet/if_ether.h> 83 #endif /* INET */ 84 #ifdef INET6 85 #include <netinet6/in6_var.h> 86 #include <netinet6/in6_ifattach.h> 87 #endif /* INET6 */ 88 #endif /* INET || INET6 */ 89 90 #include <security/mac/mac_framework.h> 91 92 #ifdef COMPAT_FREEBSD32 93 #include <sys/mount.h> 94 #include <compat/freebsd32/freebsd32.h> 95 #endif 96 97 struct ifindex_entry { 98 struct ifnet *ife_ifnet; 99 }; 100 101 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 102 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 103 104 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen); 105 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 106 &ifqmaxlen, 0, "max send queue size"); 107 108 /* Log link state change events */ 109 static int log_link_state_change = 1; 110 111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 112 &log_link_state_change, 0, 113 "log interface link state change events"); 114 115 /* Interface description */ 116 static unsigned int ifdescr_maxlen = 1024; 117 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 118 &ifdescr_maxlen, 0, 119 "administrative maximum length for interface description"); 120 121 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 122 123 /* global sx for non-critical path ifdescr */ 124 static struct sx ifdescr_sx; 125 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 126 127 void (*bridge_linkstate_p)(struct ifnet *ifp); 128 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 129 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 130 /* These are external hooks for CARP. */ 131 void (*carp_linkstate_p)(struct ifnet *ifp); 132 void (*carp_demote_adj_p)(int, char *); 133 int (*carp_master_p)(struct ifaddr *); 134 #if defined(INET) || defined(INET6) 135 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 136 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 137 const struct sockaddr *sa); 138 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 139 int (*carp_attach_p)(struct ifaddr *, int); 140 void (*carp_detach_p)(struct ifaddr *); 141 #endif 142 #ifdef INET 143 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 144 #endif 145 #ifdef INET6 146 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 147 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 148 const struct in6_addr *taddr); 149 #endif 150 151 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 152 153 /* 154 * XXX: Style; these should be sorted alphabetically, and unprototyped 155 * static functions should be prototyped. Currently they are sorted by 156 * declaration order. 157 */ 158 static void if_attachdomain(void *); 159 static void if_attachdomain1(struct ifnet *); 160 static int ifconf(u_long, caddr_t); 161 static void if_freemulti(struct ifmultiaddr *); 162 static void if_init(void *); 163 static void if_grow(void); 164 static void if_route(struct ifnet *, int flag, int fam); 165 static int if_setflag(struct ifnet *, int, int, int *, int); 166 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 167 static void if_unroute(struct ifnet *, int flag, int fam); 168 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 169 static int if_rtdel(struct radix_node *, void *); 170 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 171 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 172 static void do_link_state_change(void *, int); 173 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 174 static int if_getgroupmembers(struct ifgroupreq *); 175 static void if_delgroups(struct ifnet *); 176 static void if_attach_internal(struct ifnet *, int); 177 static void if_detach_internal(struct ifnet *, int); 178 179 #ifdef INET6 180 /* 181 * XXX: declare here to avoid to include many inet6 related files.. 182 * should be more generalized? 183 */ 184 extern void nd6_setmtu(struct ifnet *); 185 #endif 186 187 VNET_DEFINE(int, if_index); 188 int ifqmaxlen = IFQ_MAXLEN; 189 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 190 VNET_DEFINE(struct ifgrouphead, ifg_head); 191 192 static VNET_DEFINE(int, if_indexlim) = 8; 193 194 /* Table of ifnet by index. */ 195 VNET_DEFINE(struct ifindex_entry *, ifindex_table); 196 197 #define V_if_indexlim VNET(if_indexlim) 198 #define V_ifindex_table VNET(ifindex_table) 199 200 /* 201 * The global network interface list (V_ifnet) and related state (such as 202 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 203 * an rwlock. Either may be acquired shared to stablize the list, but both 204 * must be acquired writable to modify the list. This model allows us to 205 * both stablize the interface list during interrupt thread processing, but 206 * also to stablize it over long-running ioctls, without introducing priority 207 * inversions and deadlocks. 208 */ 209 struct rwlock ifnet_rwlock; 210 struct sx ifnet_sxlock; 211 212 /* 213 * The allocation of network interfaces is a rather non-atomic affair; we 214 * need to select an index before we are ready to expose the interface for 215 * use, so will use this pointer value to indicate reservation. 216 */ 217 #define IFNET_HOLD (void *)(uintptr_t)(-1) 218 219 static if_com_alloc_t *if_com_alloc[256]; 220 static if_com_free_t *if_com_free[256]; 221 222 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 223 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 224 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 225 226 struct ifnet * 227 ifnet_byindex_locked(u_short idx) 228 { 229 230 if (idx > V_if_index) 231 return (NULL); 232 if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD) 233 return (NULL); 234 return (V_ifindex_table[idx].ife_ifnet); 235 } 236 237 struct ifnet * 238 ifnet_byindex(u_short idx) 239 { 240 struct ifnet *ifp; 241 242 IFNET_RLOCK_NOSLEEP(); 243 ifp = ifnet_byindex_locked(idx); 244 IFNET_RUNLOCK_NOSLEEP(); 245 return (ifp); 246 } 247 248 struct ifnet * 249 ifnet_byindex_ref(u_short idx) 250 { 251 struct ifnet *ifp; 252 253 IFNET_RLOCK_NOSLEEP(); 254 ifp = ifnet_byindex_locked(idx); 255 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 256 IFNET_RUNLOCK_NOSLEEP(); 257 return (NULL); 258 } 259 if_ref(ifp); 260 IFNET_RUNLOCK_NOSLEEP(); 261 return (ifp); 262 } 263 264 /* 265 * Allocate an ifindex array entry; return 0 on success or an error on 266 * failure. 267 */ 268 static int 269 ifindex_alloc_locked(u_short *idxp) 270 { 271 u_short idx; 272 273 IFNET_WLOCK_ASSERT(); 274 275 retry: 276 /* 277 * Try to find an empty slot below V_if_index. If we fail, take the 278 * next slot. 279 */ 280 for (idx = 1; idx <= V_if_index; idx++) { 281 if (V_ifindex_table[idx].ife_ifnet == NULL) 282 break; 283 } 284 285 /* Catch if_index overflow. */ 286 if (idx < 1) 287 return (ENOSPC); 288 if (idx >= V_if_indexlim) { 289 if_grow(); 290 goto retry; 291 } 292 if (idx > V_if_index) 293 V_if_index = idx; 294 *idxp = idx; 295 return (0); 296 } 297 298 static void 299 ifindex_free_locked(u_short idx) 300 { 301 302 IFNET_WLOCK_ASSERT(); 303 304 V_ifindex_table[idx].ife_ifnet = NULL; 305 while (V_if_index > 0 && 306 V_ifindex_table[V_if_index].ife_ifnet == NULL) 307 V_if_index--; 308 } 309 310 static void 311 ifindex_free(u_short idx) 312 { 313 314 IFNET_WLOCK(); 315 ifindex_free_locked(idx); 316 IFNET_WUNLOCK(); 317 } 318 319 static void 320 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) 321 { 322 323 IFNET_WLOCK_ASSERT(); 324 325 V_ifindex_table[idx].ife_ifnet = ifp; 326 } 327 328 static void 329 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 330 { 331 332 IFNET_WLOCK(); 333 ifnet_setbyindex_locked(idx, ifp); 334 IFNET_WUNLOCK(); 335 } 336 337 struct ifaddr * 338 ifaddr_byindex(u_short idx) 339 { 340 struct ifaddr *ifa; 341 342 IFNET_RLOCK_NOSLEEP(); 343 ifa = ifnet_byindex_locked(idx)->if_addr; 344 if (ifa != NULL) 345 ifa_ref(ifa); 346 IFNET_RUNLOCK_NOSLEEP(); 347 return (ifa); 348 } 349 350 /* 351 * Network interface utility routines. 352 * 353 * Routines with ifa_ifwith* names take sockaddr *'s as 354 * parameters. 355 */ 356 357 static void 358 vnet_if_init(const void *unused __unused) 359 { 360 361 TAILQ_INIT(&V_ifnet); 362 TAILQ_INIT(&V_ifg_head); 363 IFNET_WLOCK(); 364 if_grow(); /* create initial table */ 365 IFNET_WUNLOCK(); 366 vnet_if_clone_init(); 367 } 368 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 369 NULL); 370 371 /* ARGSUSED*/ 372 static void 373 if_init(void *dummy __unused) 374 { 375 376 IFNET_LOCK_INIT(); 377 if_clone_init(); 378 } 379 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL); 380 381 382 #ifdef VIMAGE 383 static void 384 vnet_if_uninit(const void *unused __unused) 385 { 386 387 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 388 "not empty", __func__, __LINE__, &V_ifnet)); 389 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 390 "not empty", __func__, __LINE__, &V_ifg_head)); 391 392 free((caddr_t)V_ifindex_table, M_IFNET); 393 } 394 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 395 vnet_if_uninit, NULL); 396 #endif 397 398 static void 399 if_grow(void) 400 { 401 int oldlim; 402 u_int n; 403 struct ifindex_entry *e; 404 405 IFNET_WLOCK_ASSERT(); 406 oldlim = V_if_indexlim; 407 IFNET_WUNLOCK(); 408 n = (oldlim << 1) * sizeof(*e); 409 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 410 IFNET_WLOCK(); 411 if (V_if_indexlim != oldlim) { 412 free(e, M_IFNET); 413 return; 414 } 415 if (V_ifindex_table != NULL) { 416 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 417 free((caddr_t)V_ifindex_table, M_IFNET); 418 } 419 V_if_indexlim <<= 1; 420 V_ifindex_table = e; 421 } 422 423 /* 424 * Allocate a struct ifnet and an index for an interface. A layer 2 425 * common structure will also be allocated if an allocation routine is 426 * registered for the passed type. 427 */ 428 struct ifnet * 429 if_alloc(u_char type) 430 { 431 struct ifnet *ifp; 432 u_short idx; 433 434 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 435 IFNET_WLOCK(); 436 if (ifindex_alloc_locked(&idx) != 0) { 437 IFNET_WUNLOCK(); 438 free(ifp, M_IFNET); 439 return (NULL); 440 } 441 ifnet_setbyindex_locked(idx, IFNET_HOLD); 442 IFNET_WUNLOCK(); 443 ifp->if_index = idx; 444 ifp->if_type = type; 445 ifp->if_alloctype = type; 446 if (if_com_alloc[type] != NULL) { 447 ifp->if_l2com = if_com_alloc[type](type, ifp); 448 if (ifp->if_l2com == NULL) { 449 free(ifp, M_IFNET); 450 ifindex_free(idx); 451 return (NULL); 452 } 453 } 454 455 IF_ADDR_LOCK_INIT(ifp); 456 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 457 ifp->if_afdata_initialized = 0; 458 IF_AFDATA_LOCK_INIT(ifp); 459 TAILQ_INIT(&ifp->if_addrhead); 460 TAILQ_INIT(&ifp->if_multiaddrs); 461 TAILQ_INIT(&ifp->if_groups); 462 #ifdef MAC 463 mac_ifnet_init(ifp); 464 #endif 465 ifq_init(&ifp->if_snd, ifp); 466 467 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 468 ifnet_setbyindex(ifp->if_index, ifp); 469 return (ifp); 470 } 471 472 /* 473 * Do the actual work of freeing a struct ifnet, and layer 2 common 474 * structure. This call is made when the last reference to an 475 * interface is released. 476 */ 477 static void 478 if_free_internal(struct ifnet *ifp) 479 { 480 481 KASSERT((ifp->if_flags & IFF_DYING), 482 ("if_free_internal: interface not dying")); 483 484 if (if_com_free[ifp->if_alloctype] != NULL) 485 if_com_free[ifp->if_alloctype](ifp->if_l2com, 486 ifp->if_alloctype); 487 488 #ifdef MAC 489 mac_ifnet_destroy(ifp); 490 #endif /* MAC */ 491 if (ifp->if_description != NULL) 492 free(ifp->if_description, M_IFDESCR); 493 IF_AFDATA_DESTROY(ifp); 494 IF_ADDR_LOCK_DESTROY(ifp); 495 ifq_delete(&ifp->if_snd); 496 free(ifp, M_IFNET); 497 } 498 499 /* 500 * Deregister an interface and free the associated storage. 501 */ 502 void 503 if_free(struct ifnet *ifp) 504 { 505 506 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 507 508 CURVNET_SET_QUIET(ifp->if_vnet); 509 IFNET_WLOCK(); 510 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 511 ("%s: freeing unallocated ifnet", ifp->if_xname)); 512 513 ifindex_free_locked(ifp->if_index); 514 IFNET_WUNLOCK(); 515 516 if (refcount_release(&ifp->if_refcount)) 517 if_free_internal(ifp); 518 CURVNET_RESTORE(); 519 } 520 521 /* 522 * Interfaces to keep an ifnet type-stable despite the possibility of the 523 * driver calling if_free(). If there are additional references, we defer 524 * freeing the underlying data structure. 525 */ 526 void 527 if_ref(struct ifnet *ifp) 528 { 529 530 /* We don't assert the ifnet list lock here, but arguably should. */ 531 refcount_acquire(&ifp->if_refcount); 532 } 533 534 void 535 if_rele(struct ifnet *ifp) 536 { 537 538 if (!refcount_release(&ifp->if_refcount)) 539 return; 540 if_free_internal(ifp); 541 } 542 543 void 544 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 545 { 546 547 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 548 549 if (ifq->ifq_maxlen == 0) 550 ifq->ifq_maxlen = ifqmaxlen; 551 552 ifq->altq_type = 0; 553 ifq->altq_disc = NULL; 554 ifq->altq_flags &= ALTQF_CANTCHANGE; 555 ifq->altq_tbr = NULL; 556 ifq->altq_ifp = ifp; 557 } 558 559 void 560 ifq_delete(struct ifaltq *ifq) 561 { 562 mtx_destroy(&ifq->ifq_mtx); 563 } 564 565 /* 566 * Perform generic interface initalization tasks and attach the interface 567 * to the list of "active" interfaces. If vmove flag is set on entry 568 * to if_attach_internal(), perform only a limited subset of initialization 569 * tasks, given that we are moving from one vnet to another an ifnet which 570 * has already been fully initialized. 571 * 572 * XXX: 573 * - The decision to return void and thus require this function to 574 * succeed is questionable. 575 * - We should probably do more sanity checking. For instance we don't 576 * do anything to insure if_xname is unique or non-empty. 577 */ 578 void 579 if_attach(struct ifnet *ifp) 580 { 581 582 if_attach_internal(ifp, 0); 583 } 584 585 static void 586 if_attach_internal(struct ifnet *ifp, int vmove) 587 { 588 unsigned socksize, ifasize; 589 int namelen, masklen; 590 struct sockaddr_dl *sdl; 591 struct ifaddr *ifa; 592 593 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 594 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 595 ifp->if_xname); 596 597 #ifdef VIMAGE 598 ifp->if_vnet = curvnet; 599 if (ifp->if_home_vnet == NULL) 600 ifp->if_home_vnet = curvnet; 601 #endif 602 603 if_addgroup(ifp, IFG_ALL); 604 605 getmicrotime(&ifp->if_lastchange); 606 ifp->if_data.ifi_epoch = time_uptime; 607 ifp->if_data.ifi_datalen = sizeof(struct if_data); 608 609 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 610 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 611 ("transmit and qflush must both either be set or both be NULL")); 612 if (ifp->if_transmit == NULL) { 613 ifp->if_transmit = if_transmit; 614 ifp->if_qflush = if_qflush; 615 } 616 617 if (!vmove) { 618 #ifdef MAC 619 mac_ifnet_create(ifp); 620 #endif 621 622 /* 623 * Create a Link Level name for this device. 624 */ 625 namelen = strlen(ifp->if_xname); 626 /* 627 * Always save enough space for any possiable name so we 628 * can do a rename in place later. 629 */ 630 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 631 socksize = masklen + ifp->if_addrlen; 632 if (socksize < sizeof(*sdl)) 633 socksize = sizeof(*sdl); 634 socksize = roundup2(socksize, sizeof(long)); 635 ifasize = sizeof(*ifa) + 2 * socksize; 636 ifa = ifa_alloc(ifasize, M_WAITOK); 637 sdl = (struct sockaddr_dl *)(ifa + 1); 638 sdl->sdl_len = socksize; 639 sdl->sdl_family = AF_LINK; 640 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 641 sdl->sdl_nlen = namelen; 642 sdl->sdl_index = ifp->if_index; 643 sdl->sdl_type = ifp->if_type; 644 ifp->if_addr = ifa; 645 ifa->ifa_ifp = ifp; 646 ifa->ifa_rtrequest = link_rtrequest; 647 ifa->ifa_addr = (struct sockaddr *)sdl; 648 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 649 ifa->ifa_netmask = (struct sockaddr *)sdl; 650 sdl->sdl_len = masklen; 651 while (namelen != 0) 652 sdl->sdl_data[--namelen] = 0xff; 653 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 654 /* Reliably crash if used uninitialized. */ 655 ifp->if_broadcastaddr = NULL; 656 657 #if defined(INET) || defined(INET6) 658 /* Initialize to max value. */ 659 if (ifp->if_hw_tsomax == 0) 660 ifp->if_hw_tsomax = IP_MAXPACKET; 661 KASSERT(ifp->if_hw_tsomax <= IP_MAXPACKET && 662 ifp->if_hw_tsomax >= IP_MAXPACKET / 8, 663 ("%s: tsomax outside of range", __func__)); 664 #endif 665 } 666 #ifdef VIMAGE 667 else { 668 /* 669 * Update the interface index in the link layer address 670 * of the interface. 671 */ 672 for (ifa = ifp->if_addr; ifa != NULL; 673 ifa = TAILQ_NEXT(ifa, ifa_link)) { 674 if (ifa->ifa_addr->sa_family == AF_LINK) { 675 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 676 sdl->sdl_index = ifp->if_index; 677 } 678 } 679 } 680 #endif 681 682 IFNET_WLOCK(); 683 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 684 #ifdef VIMAGE 685 curvnet->vnet_ifcnt++; 686 #endif 687 IFNET_WUNLOCK(); 688 689 if (domain_init_status >= 2) 690 if_attachdomain1(ifp); 691 692 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 693 if (IS_DEFAULT_VNET(curvnet)) 694 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 695 696 /* Announce the interface. */ 697 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 698 } 699 700 static void 701 if_attachdomain(void *dummy) 702 { 703 struct ifnet *ifp; 704 705 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 706 if_attachdomain1(ifp); 707 } 708 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 709 if_attachdomain, NULL); 710 711 static void 712 if_attachdomain1(struct ifnet *ifp) 713 { 714 struct domain *dp; 715 716 /* 717 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 718 * cannot lock ifp->if_afdata initialization, entirely. 719 */ 720 if (IF_AFDATA_TRYLOCK(ifp) == 0) 721 return; 722 if (ifp->if_afdata_initialized >= domain_init_status) { 723 IF_AFDATA_UNLOCK(ifp); 724 log(LOG_WARNING, "%s called more than once on %s\n", 725 __func__, ifp->if_xname); 726 return; 727 } 728 ifp->if_afdata_initialized = domain_init_status; 729 IF_AFDATA_UNLOCK(ifp); 730 731 /* address family dependent data region */ 732 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 733 for (dp = domains; dp; dp = dp->dom_next) { 734 if (dp->dom_ifattach) 735 ifp->if_afdata[dp->dom_family] = 736 (*dp->dom_ifattach)(ifp); 737 } 738 } 739 740 /* 741 * Remove any unicast or broadcast network addresses from an interface. 742 */ 743 void 744 if_purgeaddrs(struct ifnet *ifp) 745 { 746 struct ifaddr *ifa, *next; 747 748 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 749 if (ifa->ifa_addr->sa_family == AF_LINK) 750 continue; 751 #ifdef INET 752 /* XXX: Ugly!! ad hoc just for INET */ 753 if (ifa->ifa_addr->sa_family == AF_INET) { 754 struct ifaliasreq ifr; 755 756 bzero(&ifr, sizeof(ifr)); 757 ifr.ifra_addr = *ifa->ifa_addr; 758 if (ifa->ifa_dstaddr) 759 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 760 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 761 NULL) == 0) 762 continue; 763 } 764 #endif /* INET */ 765 #ifdef INET6 766 if (ifa->ifa_addr->sa_family == AF_INET6) { 767 in6_purgeaddr(ifa); 768 /* ifp_addrhead is already updated */ 769 continue; 770 } 771 #endif /* INET6 */ 772 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 773 ifa_free(ifa); 774 } 775 } 776 777 /* 778 * Remove any multicast network addresses from an interface when an ifnet 779 * is going away. 780 */ 781 static void 782 if_purgemaddrs(struct ifnet *ifp) 783 { 784 struct ifmultiaddr *ifma; 785 struct ifmultiaddr *next; 786 787 IF_ADDR_WLOCK(ifp); 788 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 789 if_delmulti_locked(ifp, ifma, 1); 790 IF_ADDR_WUNLOCK(ifp); 791 } 792 793 /* 794 * Detach an interface, removing it from the list of "active" interfaces. 795 * If vmove flag is set on entry to if_detach_internal(), perform only a 796 * limited subset of cleanup tasks, given that we are moving an ifnet from 797 * one vnet to another, where it must be fully operational. 798 * 799 * XXXRW: There are some significant questions about event ordering, and 800 * how to prevent things from starting to use the interface during detach. 801 */ 802 void 803 if_detach(struct ifnet *ifp) 804 { 805 806 CURVNET_SET_QUIET(ifp->if_vnet); 807 if_detach_internal(ifp, 0); 808 CURVNET_RESTORE(); 809 } 810 811 static void 812 if_detach_internal(struct ifnet *ifp, int vmove) 813 { 814 struct ifaddr *ifa; 815 struct radix_node_head *rnh; 816 int i, j; 817 struct domain *dp; 818 struct ifnet *iter; 819 int found = 0; 820 821 IFNET_WLOCK(); 822 TAILQ_FOREACH(iter, &V_ifnet, if_link) 823 if (iter == ifp) { 824 TAILQ_REMOVE(&V_ifnet, ifp, if_link); 825 found = 1; 826 break; 827 } 828 #ifdef VIMAGE 829 if (found) 830 curvnet->vnet_ifcnt--; 831 #endif 832 IFNET_WUNLOCK(); 833 if (!found) { 834 if (vmove) 835 panic("%s: ifp=%p not on the ifnet tailq %p", 836 __func__, ifp, &V_ifnet); 837 else 838 return; /* XXX this should panic as well? */ 839 } 840 841 /* 842 * Remove/wait for pending events. 843 */ 844 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 845 846 /* 847 * Remove routes and flush queues. 848 */ 849 if_down(ifp); 850 #ifdef ALTQ 851 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 852 altq_disable(&ifp->if_snd); 853 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 854 altq_detach(&ifp->if_snd); 855 #endif 856 857 if_purgeaddrs(ifp); 858 859 #ifdef INET 860 in_ifdetach(ifp); 861 #endif 862 863 #ifdef INET6 864 /* 865 * Remove all IPv6 kernel structs related to ifp. This should be done 866 * before removing routing entries below, since IPv6 interface direct 867 * routes are expected to be removed by the IPv6-specific kernel API. 868 * Otherwise, the kernel will detect some inconsistency and bark it. 869 */ 870 in6_ifdetach(ifp); 871 #endif 872 if_purgemaddrs(ifp); 873 874 if (!vmove) { 875 /* 876 * Prevent further calls into the device driver via ifnet. 877 */ 878 if_dead(ifp); 879 880 /* 881 * Remove link ifaddr pointer and maybe decrement if_index. 882 * Clean up all addresses. 883 */ 884 ifp->if_addr = NULL; 885 886 /* We can now free link ifaddr. */ 887 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 888 ifa = TAILQ_FIRST(&ifp->if_addrhead); 889 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 890 ifa_free(ifa); 891 } 892 } 893 894 /* 895 * Delete all remaining routes using this interface 896 * Unfortuneatly the only way to do this is to slog through 897 * the entire routing table looking for routes which point 898 * to this interface...oh well... 899 */ 900 for (i = 1; i <= AF_MAX; i++) { 901 for (j = 0; j < rt_numfibs; j++) { 902 rnh = rt_tables_get_rnh(j, i); 903 if (rnh == NULL) 904 continue; 905 RADIX_NODE_HEAD_LOCK(rnh); 906 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 907 RADIX_NODE_HEAD_UNLOCK(rnh); 908 } 909 } 910 911 /* Announce that the interface is gone. */ 912 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 913 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 914 if (IS_DEFAULT_VNET(curvnet)) 915 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 916 if_delgroups(ifp); 917 918 /* 919 * We cannot hold the lock over dom_ifdetach calls as they might 920 * sleep, for example trying to drain a callout, thus open up the 921 * theoretical race with re-attaching. 922 */ 923 IF_AFDATA_LOCK(ifp); 924 i = ifp->if_afdata_initialized; 925 ifp->if_afdata_initialized = 0; 926 IF_AFDATA_UNLOCK(ifp); 927 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 928 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 929 (*dp->dom_ifdetach)(ifp, 930 ifp->if_afdata[dp->dom_family]); 931 } 932 } 933 934 #ifdef VIMAGE 935 /* 936 * if_vmove() performs a limited version of if_detach() in current 937 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 938 * An attempt is made to shrink if_index in current vnet, find an 939 * unused if_index in target vnet and calls if_grow() if necessary, 940 * and finally find an unused if_xname for the target vnet. 941 */ 942 void 943 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 944 { 945 u_short idx; 946 947 /* 948 * Detach from current vnet, but preserve LLADDR info, do not 949 * mark as dead etc. so that the ifnet can be reattached later. 950 */ 951 if_detach_internal(ifp, 1); 952 953 /* 954 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 955 * the if_index for that vnet if possible. 956 * 957 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 958 * or we'd lock on one vnet and unlock on another. 959 */ 960 IFNET_WLOCK(); 961 ifindex_free_locked(ifp->if_index); 962 IFNET_WUNLOCK(); 963 964 /* 965 * Perform interface-specific reassignment tasks, if provided by 966 * the driver. 967 */ 968 if (ifp->if_reassign != NULL) 969 ifp->if_reassign(ifp, new_vnet, NULL); 970 971 /* 972 * Switch to the context of the target vnet. 973 */ 974 CURVNET_SET_QUIET(new_vnet); 975 976 IFNET_WLOCK(); 977 if (ifindex_alloc_locked(&idx) != 0) { 978 IFNET_WUNLOCK(); 979 panic("if_index overflow"); 980 } 981 ifp->if_index = idx; 982 ifnet_setbyindex_locked(ifp->if_index, ifp); 983 IFNET_WUNLOCK(); 984 985 if_attach_internal(ifp, 1); 986 987 CURVNET_RESTORE(); 988 } 989 990 /* 991 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 992 */ 993 static int 994 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 995 { 996 struct prison *pr; 997 struct ifnet *difp; 998 999 /* Try to find the prison within our visibility. */ 1000 sx_slock(&allprison_lock); 1001 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1002 sx_sunlock(&allprison_lock); 1003 if (pr == NULL) 1004 return (ENXIO); 1005 prison_hold_locked(pr); 1006 mtx_unlock(&pr->pr_mtx); 1007 1008 /* Do not try to move the iface from and to the same prison. */ 1009 if (pr->pr_vnet == ifp->if_vnet) { 1010 prison_free(pr); 1011 return (EEXIST); 1012 } 1013 1014 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1015 /* XXX Lock interfaces to avoid races. */ 1016 CURVNET_SET_QUIET(pr->pr_vnet); 1017 difp = ifunit(ifname); 1018 CURVNET_RESTORE(); 1019 if (difp != NULL) { 1020 prison_free(pr); 1021 return (EEXIST); 1022 } 1023 1024 /* Move the interface into the child jail/vnet. */ 1025 if_vmove(ifp, pr->pr_vnet); 1026 1027 /* Report the new if_xname back to the userland. */ 1028 sprintf(ifname, "%s", ifp->if_xname); 1029 1030 prison_free(pr); 1031 return (0); 1032 } 1033 1034 static int 1035 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1036 { 1037 struct prison *pr; 1038 struct vnet *vnet_dst; 1039 struct ifnet *ifp; 1040 1041 /* Try to find the prison within our visibility. */ 1042 sx_slock(&allprison_lock); 1043 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1044 sx_sunlock(&allprison_lock); 1045 if (pr == NULL) 1046 return (ENXIO); 1047 prison_hold_locked(pr); 1048 mtx_unlock(&pr->pr_mtx); 1049 1050 /* Make sure the named iface exists in the source prison/vnet. */ 1051 CURVNET_SET(pr->pr_vnet); 1052 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1053 if (ifp == NULL) { 1054 CURVNET_RESTORE(); 1055 prison_free(pr); 1056 return (ENXIO); 1057 } 1058 1059 /* Do not try to move the iface from and to the same prison. */ 1060 vnet_dst = TD_TO_VNET(td); 1061 if (vnet_dst == ifp->if_vnet) { 1062 CURVNET_RESTORE(); 1063 prison_free(pr); 1064 return (EEXIST); 1065 } 1066 1067 /* Get interface back from child jail/vnet. */ 1068 if_vmove(ifp, vnet_dst); 1069 CURVNET_RESTORE(); 1070 1071 /* Report the new if_xname back to the userland. */ 1072 sprintf(ifname, "%s", ifp->if_xname); 1073 1074 prison_free(pr); 1075 return (0); 1076 } 1077 #endif /* VIMAGE */ 1078 1079 /* 1080 * Add a group to an interface 1081 */ 1082 int 1083 if_addgroup(struct ifnet *ifp, const char *groupname) 1084 { 1085 struct ifg_list *ifgl; 1086 struct ifg_group *ifg = NULL; 1087 struct ifg_member *ifgm; 1088 int new = 0; 1089 1090 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1091 groupname[strlen(groupname) - 1] <= '9') 1092 return (EINVAL); 1093 1094 IFNET_WLOCK(); 1095 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1096 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1097 IFNET_WUNLOCK(); 1098 return (EEXIST); 1099 } 1100 1101 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1102 M_NOWAIT)) == NULL) { 1103 IFNET_WUNLOCK(); 1104 return (ENOMEM); 1105 } 1106 1107 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1108 M_TEMP, M_NOWAIT)) == NULL) { 1109 free(ifgl, M_TEMP); 1110 IFNET_WUNLOCK(); 1111 return (ENOMEM); 1112 } 1113 1114 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1115 if (!strcmp(ifg->ifg_group, groupname)) 1116 break; 1117 1118 if (ifg == NULL) { 1119 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1120 M_TEMP, M_NOWAIT)) == NULL) { 1121 free(ifgl, M_TEMP); 1122 free(ifgm, M_TEMP); 1123 IFNET_WUNLOCK(); 1124 return (ENOMEM); 1125 } 1126 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1127 ifg->ifg_refcnt = 0; 1128 TAILQ_INIT(&ifg->ifg_members); 1129 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1130 new = 1; 1131 } 1132 1133 ifg->ifg_refcnt++; 1134 ifgl->ifgl_group = ifg; 1135 ifgm->ifgm_ifp = ifp; 1136 1137 IF_ADDR_WLOCK(ifp); 1138 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1139 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1140 IF_ADDR_WUNLOCK(ifp); 1141 1142 IFNET_WUNLOCK(); 1143 1144 if (new) 1145 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1146 EVENTHANDLER_INVOKE(group_change_event, groupname); 1147 1148 return (0); 1149 } 1150 1151 /* 1152 * Remove a group from an interface 1153 */ 1154 int 1155 if_delgroup(struct ifnet *ifp, const char *groupname) 1156 { 1157 struct ifg_list *ifgl; 1158 struct ifg_member *ifgm; 1159 1160 IFNET_WLOCK(); 1161 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1162 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1163 break; 1164 if (ifgl == NULL) { 1165 IFNET_WUNLOCK(); 1166 return (ENOENT); 1167 } 1168 1169 IF_ADDR_WLOCK(ifp); 1170 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1171 IF_ADDR_WUNLOCK(ifp); 1172 1173 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1174 if (ifgm->ifgm_ifp == ifp) 1175 break; 1176 1177 if (ifgm != NULL) { 1178 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 1179 free(ifgm, M_TEMP); 1180 } 1181 1182 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1183 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1184 IFNET_WUNLOCK(); 1185 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1186 free(ifgl->ifgl_group, M_TEMP); 1187 } else 1188 IFNET_WUNLOCK(); 1189 1190 free(ifgl, M_TEMP); 1191 1192 EVENTHANDLER_INVOKE(group_change_event, groupname); 1193 1194 return (0); 1195 } 1196 1197 /* 1198 * Remove an interface from all groups 1199 */ 1200 static void 1201 if_delgroups(struct ifnet *ifp) 1202 { 1203 struct ifg_list *ifgl; 1204 struct ifg_member *ifgm; 1205 char groupname[IFNAMSIZ]; 1206 1207 IFNET_WLOCK(); 1208 while (!TAILQ_EMPTY(&ifp->if_groups)) { 1209 ifgl = TAILQ_FIRST(&ifp->if_groups); 1210 1211 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1212 1213 IF_ADDR_WLOCK(ifp); 1214 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1215 IF_ADDR_WUNLOCK(ifp); 1216 1217 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1218 if (ifgm->ifgm_ifp == ifp) 1219 break; 1220 1221 if (ifgm != NULL) { 1222 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1223 ifgm_next); 1224 free(ifgm, M_TEMP); 1225 } 1226 1227 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1228 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1229 IFNET_WUNLOCK(); 1230 EVENTHANDLER_INVOKE(group_detach_event, 1231 ifgl->ifgl_group); 1232 free(ifgl->ifgl_group, M_TEMP); 1233 } else 1234 IFNET_WUNLOCK(); 1235 1236 free(ifgl, M_TEMP); 1237 1238 EVENTHANDLER_INVOKE(group_change_event, groupname); 1239 1240 IFNET_WLOCK(); 1241 } 1242 IFNET_WUNLOCK(); 1243 } 1244 1245 /* 1246 * Stores all groups from an interface in memory pointed 1247 * to by data 1248 */ 1249 static int 1250 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 1251 { 1252 int len, error; 1253 struct ifg_list *ifgl; 1254 struct ifg_req ifgrq, *ifgp; 1255 struct ifgroupreq *ifgr = data; 1256 1257 if (ifgr->ifgr_len == 0) { 1258 IF_ADDR_RLOCK(ifp); 1259 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1260 ifgr->ifgr_len += sizeof(struct ifg_req); 1261 IF_ADDR_RUNLOCK(ifp); 1262 return (0); 1263 } 1264 1265 len = ifgr->ifgr_len; 1266 ifgp = ifgr->ifgr_groups; 1267 /* XXX: wire */ 1268 IF_ADDR_RLOCK(ifp); 1269 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1270 if (len < sizeof(ifgrq)) { 1271 IF_ADDR_RUNLOCK(ifp); 1272 return (EINVAL); 1273 } 1274 bzero(&ifgrq, sizeof ifgrq); 1275 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1276 sizeof(ifgrq.ifgrq_group)); 1277 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1278 IF_ADDR_RUNLOCK(ifp); 1279 return (error); 1280 } 1281 len -= sizeof(ifgrq); 1282 ifgp++; 1283 } 1284 IF_ADDR_RUNLOCK(ifp); 1285 1286 return (0); 1287 } 1288 1289 /* 1290 * Stores all members of a group in memory pointed to by data 1291 */ 1292 static int 1293 if_getgroupmembers(struct ifgroupreq *data) 1294 { 1295 struct ifgroupreq *ifgr = data; 1296 struct ifg_group *ifg; 1297 struct ifg_member *ifgm; 1298 struct ifg_req ifgrq, *ifgp; 1299 int len, error; 1300 1301 IFNET_RLOCK(); 1302 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1303 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1304 break; 1305 if (ifg == NULL) { 1306 IFNET_RUNLOCK(); 1307 return (ENOENT); 1308 } 1309 1310 if (ifgr->ifgr_len == 0) { 1311 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1312 ifgr->ifgr_len += sizeof(ifgrq); 1313 IFNET_RUNLOCK(); 1314 return (0); 1315 } 1316 1317 len = ifgr->ifgr_len; 1318 ifgp = ifgr->ifgr_groups; 1319 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1320 if (len < sizeof(ifgrq)) { 1321 IFNET_RUNLOCK(); 1322 return (EINVAL); 1323 } 1324 bzero(&ifgrq, sizeof ifgrq); 1325 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1326 sizeof(ifgrq.ifgrq_member)); 1327 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1328 IFNET_RUNLOCK(); 1329 return (error); 1330 } 1331 len -= sizeof(ifgrq); 1332 ifgp++; 1333 } 1334 IFNET_RUNLOCK(); 1335 1336 return (0); 1337 } 1338 1339 /* 1340 * Delete Routes for a Network Interface 1341 * 1342 * Called for each routing entry via the rnh->rnh_walktree() call above 1343 * to delete all route entries referencing a detaching network interface. 1344 * 1345 * Arguments: 1346 * rn pointer to node in the routing table 1347 * arg argument passed to rnh->rnh_walktree() - detaching interface 1348 * 1349 * Returns: 1350 * 0 successful 1351 * errno failed - reason indicated 1352 * 1353 */ 1354 static int 1355 if_rtdel(struct radix_node *rn, void *arg) 1356 { 1357 struct rtentry *rt = (struct rtentry *)rn; 1358 struct ifnet *ifp = arg; 1359 int err; 1360 1361 if (rt->rt_ifp == ifp) { 1362 1363 /* 1364 * Protect (sorta) against walktree recursion problems 1365 * with cloned routes 1366 */ 1367 if ((rt->rt_flags & RTF_UP) == 0) 1368 return (0); 1369 1370 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway, 1371 rt_mask(rt), 1372 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED, 1373 (struct rtentry **) NULL, rt->rt_fibnum); 1374 if (err) { 1375 log(LOG_WARNING, "if_rtdel: error %d\n", err); 1376 } 1377 } 1378 1379 return (0); 1380 } 1381 1382 /* 1383 * Wrapper functions for struct ifnet address list locking macros. These are 1384 * used by kernel modules to avoid encoding programming interface or binary 1385 * interface assumptions that may be violated when kernel-internal locking 1386 * approaches change. 1387 */ 1388 void 1389 if_addr_rlock(struct ifnet *ifp) 1390 { 1391 1392 IF_ADDR_RLOCK(ifp); 1393 } 1394 1395 void 1396 if_addr_runlock(struct ifnet *ifp) 1397 { 1398 1399 IF_ADDR_RUNLOCK(ifp); 1400 } 1401 1402 void 1403 if_maddr_rlock(struct ifnet *ifp) 1404 { 1405 1406 IF_ADDR_RLOCK(ifp); 1407 } 1408 1409 void 1410 if_maddr_runlock(struct ifnet *ifp) 1411 { 1412 1413 IF_ADDR_RUNLOCK(ifp); 1414 } 1415 1416 /* 1417 * Initialization, destruction and refcounting functions for ifaddrs. 1418 */ 1419 struct ifaddr * 1420 ifa_alloc(size_t size, int flags) 1421 { 1422 struct ifaddr *ifa; 1423 1424 KASSERT(size >= sizeof(struct ifaddr), 1425 ("%s: invalid size %zu", __func__, size)); 1426 1427 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1428 if (ifa == NULL) 1429 return (NULL); 1430 1431 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1432 goto fail; 1433 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1434 goto fail; 1435 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1436 goto fail; 1437 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1438 goto fail; 1439 1440 refcount_init(&ifa->ifa_refcnt, 1); 1441 1442 return (ifa); 1443 1444 fail: 1445 /* free(NULL) is okay */ 1446 counter_u64_free(ifa->ifa_opackets); 1447 counter_u64_free(ifa->ifa_ipackets); 1448 counter_u64_free(ifa->ifa_obytes); 1449 counter_u64_free(ifa->ifa_ibytes); 1450 free(ifa, M_IFADDR); 1451 1452 return (NULL); 1453 } 1454 1455 void 1456 ifa_ref(struct ifaddr *ifa) 1457 { 1458 1459 refcount_acquire(&ifa->ifa_refcnt); 1460 } 1461 1462 void 1463 ifa_free(struct ifaddr *ifa) 1464 { 1465 1466 if (refcount_release(&ifa->ifa_refcnt)) { 1467 counter_u64_free(ifa->ifa_opackets); 1468 counter_u64_free(ifa->ifa_ipackets); 1469 counter_u64_free(ifa->ifa_obytes); 1470 counter_u64_free(ifa->ifa_ibytes); 1471 free(ifa, M_IFADDR); 1472 } 1473 } 1474 1475 int 1476 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1477 { 1478 int error = 0; 1479 struct rtentry *rt = NULL; 1480 struct rt_addrinfo info; 1481 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1482 1483 bzero(&info, sizeof(info)); 1484 info.rti_ifp = V_loif; 1485 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; 1486 info.rti_info[RTAX_DST] = ia; 1487 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1488 error = rtrequest1_fib(RTM_ADD, &info, &rt, 0); 1489 1490 if (error == 0 && rt != NULL) { 1491 RT_LOCK(rt); 1492 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = 1493 ifa->ifa_ifp->if_type; 1494 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = 1495 ifa->ifa_ifp->if_index; 1496 RT_REMREF(rt); 1497 RT_UNLOCK(rt); 1498 } else if (error != 0) 1499 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error); 1500 1501 return (error); 1502 } 1503 1504 int 1505 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1506 { 1507 int error = 0; 1508 struct rt_addrinfo info; 1509 struct sockaddr_dl null_sdl; 1510 1511 bzero(&null_sdl, sizeof(null_sdl)); 1512 null_sdl.sdl_len = sizeof(null_sdl); 1513 null_sdl.sdl_family = AF_LINK; 1514 null_sdl.sdl_type = ifa->ifa_ifp->if_type; 1515 null_sdl.sdl_index = ifa->ifa_ifp->if_index; 1516 bzero(&info, sizeof(info)); 1517 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; 1518 info.rti_info[RTAX_DST] = ia; 1519 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1520 error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0); 1521 1522 if (error != 0) 1523 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error); 1524 1525 return (error); 1526 } 1527 1528 /* 1529 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1530 * structs used to represent other address families, it is necessary 1531 * to perform a different comparison. 1532 */ 1533 1534 #define sa_equal(a1, a2) \ 1535 (bcmp((a1), (a2), ((a1))->sa_len) == 0) 1536 1537 #define sa_dl_equal(a1, a2) \ 1538 ((((struct sockaddr_dl *)(a1))->sdl_len == \ 1539 ((struct sockaddr_dl *)(a2))->sdl_len) && \ 1540 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \ 1541 LLADDR((struct sockaddr_dl *)(a2)), \ 1542 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1543 1544 /* 1545 * Locate an interface based on a complete address. 1546 */ 1547 /*ARGSUSED*/ 1548 static struct ifaddr * 1549 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref) 1550 { 1551 struct ifnet *ifp; 1552 struct ifaddr *ifa; 1553 1554 IFNET_RLOCK_NOSLEEP(); 1555 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1556 IF_ADDR_RLOCK(ifp); 1557 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1558 if (ifa->ifa_addr->sa_family != addr->sa_family) 1559 continue; 1560 if (sa_equal(addr, ifa->ifa_addr)) { 1561 if (getref) 1562 ifa_ref(ifa); 1563 IF_ADDR_RUNLOCK(ifp); 1564 goto done; 1565 } 1566 /* IP6 doesn't have broadcast */ 1567 if ((ifp->if_flags & IFF_BROADCAST) && 1568 ifa->ifa_broadaddr && 1569 ifa->ifa_broadaddr->sa_len != 0 && 1570 sa_equal(ifa->ifa_broadaddr, addr)) { 1571 if (getref) 1572 ifa_ref(ifa); 1573 IF_ADDR_RUNLOCK(ifp); 1574 goto done; 1575 } 1576 } 1577 IF_ADDR_RUNLOCK(ifp); 1578 } 1579 ifa = NULL; 1580 done: 1581 IFNET_RUNLOCK_NOSLEEP(); 1582 return (ifa); 1583 } 1584 1585 struct ifaddr * 1586 ifa_ifwithaddr(struct sockaddr *addr) 1587 { 1588 1589 return (ifa_ifwithaddr_internal(addr, 1)); 1590 } 1591 1592 int 1593 ifa_ifwithaddr_check(struct sockaddr *addr) 1594 { 1595 1596 return (ifa_ifwithaddr_internal(addr, 0) != NULL); 1597 } 1598 1599 /* 1600 * Locate an interface based on the broadcast address. 1601 */ 1602 /* ARGSUSED */ 1603 struct ifaddr * 1604 ifa_ifwithbroadaddr(struct sockaddr *addr) 1605 { 1606 struct ifnet *ifp; 1607 struct ifaddr *ifa; 1608 1609 IFNET_RLOCK_NOSLEEP(); 1610 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1611 IF_ADDR_RLOCK(ifp); 1612 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1613 if (ifa->ifa_addr->sa_family != addr->sa_family) 1614 continue; 1615 if ((ifp->if_flags & IFF_BROADCAST) && 1616 ifa->ifa_broadaddr && 1617 ifa->ifa_broadaddr->sa_len != 0 && 1618 sa_equal(ifa->ifa_broadaddr, addr)) { 1619 ifa_ref(ifa); 1620 IF_ADDR_RUNLOCK(ifp); 1621 goto done; 1622 } 1623 } 1624 IF_ADDR_RUNLOCK(ifp); 1625 } 1626 ifa = NULL; 1627 done: 1628 IFNET_RUNLOCK_NOSLEEP(); 1629 return (ifa); 1630 } 1631 1632 /* 1633 * Locate the point to point interface with a given destination address. 1634 */ 1635 /*ARGSUSED*/ 1636 struct ifaddr * 1637 ifa_ifwithdstaddr(struct sockaddr *addr) 1638 { 1639 struct ifnet *ifp; 1640 struct ifaddr *ifa; 1641 1642 IFNET_RLOCK_NOSLEEP(); 1643 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1644 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1645 continue; 1646 IF_ADDR_RLOCK(ifp); 1647 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1648 if (ifa->ifa_addr->sa_family != addr->sa_family) 1649 continue; 1650 if (ifa->ifa_dstaddr != NULL && 1651 sa_equal(addr, ifa->ifa_dstaddr)) { 1652 ifa_ref(ifa); 1653 IF_ADDR_RUNLOCK(ifp); 1654 goto done; 1655 } 1656 } 1657 IF_ADDR_RUNLOCK(ifp); 1658 } 1659 ifa = NULL; 1660 done: 1661 IFNET_RUNLOCK_NOSLEEP(); 1662 return (ifa); 1663 } 1664 1665 /* 1666 * Find an interface on a specific network. If many, choice 1667 * is most specific found. 1668 */ 1669 struct ifaddr * 1670 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp) 1671 { 1672 struct ifnet *ifp; 1673 struct ifaddr *ifa; 1674 struct ifaddr *ifa_maybe = NULL; 1675 u_int af = addr->sa_family; 1676 char *addr_data = addr->sa_data, *cplim; 1677 1678 /* 1679 * AF_LINK addresses can be looked up directly by their index number, 1680 * so do that if we can. 1681 */ 1682 if (af == AF_LINK) { 1683 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1684 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1685 return (ifaddr_byindex(sdl->sdl_index)); 1686 } 1687 1688 /* 1689 * Scan though each interface, looking for ones that have addresses 1690 * in this address family. Maintain a reference on ifa_maybe once 1691 * we find one, as we release the IF_ADDR_RLOCK() that kept it stable 1692 * when we move onto the next interface. 1693 */ 1694 IFNET_RLOCK_NOSLEEP(); 1695 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1696 IF_ADDR_RLOCK(ifp); 1697 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1698 char *cp, *cp2, *cp3; 1699 1700 if (ifa->ifa_addr->sa_family != af) 1701 next: continue; 1702 if (af == AF_INET && 1703 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1704 /* 1705 * This is a bit broken as it doesn't 1706 * take into account that the remote end may 1707 * be a single node in the network we are 1708 * looking for. 1709 * The trouble is that we don't know the 1710 * netmask for the remote end. 1711 */ 1712 if (ifa->ifa_dstaddr != NULL && 1713 sa_equal(addr, ifa->ifa_dstaddr)) { 1714 ifa_ref(ifa); 1715 IF_ADDR_RUNLOCK(ifp); 1716 goto done; 1717 } 1718 } else { 1719 /* 1720 * if we have a special address handler, 1721 * then use it instead of the generic one. 1722 */ 1723 if (ifa->ifa_claim_addr) { 1724 if ((*ifa->ifa_claim_addr)(ifa, addr)) { 1725 ifa_ref(ifa); 1726 IF_ADDR_RUNLOCK(ifp); 1727 goto done; 1728 } 1729 continue; 1730 } 1731 1732 /* 1733 * Scan all the bits in the ifa's address. 1734 * If a bit dissagrees with what we are 1735 * looking for, mask it with the netmask 1736 * to see if it really matters. 1737 * (A byte at a time) 1738 */ 1739 if (ifa->ifa_netmask == 0) 1740 continue; 1741 cp = addr_data; 1742 cp2 = ifa->ifa_addr->sa_data; 1743 cp3 = ifa->ifa_netmask->sa_data; 1744 cplim = ifa->ifa_netmask->sa_len 1745 + (char *)ifa->ifa_netmask; 1746 while (cp3 < cplim) 1747 if ((*cp++ ^ *cp2++) & *cp3++) 1748 goto next; /* next address! */ 1749 /* 1750 * If the netmask of what we just found 1751 * is more specific than what we had before 1752 * (if we had one), or if the virtual status 1753 * of new prefix is better than of the old one, 1754 * then remember the new one before continuing 1755 * to search for an even better one. 1756 */ 1757 if (ifa_maybe == NULL || 1758 ifa_preferred(ifa_maybe, ifa) || 1759 rn_refines((caddr_t)ifa->ifa_netmask, 1760 (caddr_t)ifa_maybe->ifa_netmask)) { 1761 if (ifa_maybe != NULL) 1762 ifa_free(ifa_maybe); 1763 ifa_maybe = ifa; 1764 ifa_ref(ifa_maybe); 1765 } 1766 } 1767 } 1768 IF_ADDR_RUNLOCK(ifp); 1769 } 1770 ifa = ifa_maybe; 1771 ifa_maybe = NULL; 1772 done: 1773 IFNET_RUNLOCK_NOSLEEP(); 1774 if (ifa_maybe != NULL) 1775 ifa_free(ifa_maybe); 1776 return (ifa); 1777 } 1778 1779 /* 1780 * Find an interface address specific to an interface best matching 1781 * a given address. 1782 */ 1783 struct ifaddr * 1784 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1785 { 1786 struct ifaddr *ifa; 1787 char *cp, *cp2, *cp3; 1788 char *cplim; 1789 struct ifaddr *ifa_maybe = NULL; 1790 u_int af = addr->sa_family; 1791 1792 if (af >= AF_MAX) 1793 return (NULL); 1794 IF_ADDR_RLOCK(ifp); 1795 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1796 if (ifa->ifa_addr->sa_family != af) 1797 continue; 1798 if (ifa_maybe == NULL) 1799 ifa_maybe = ifa; 1800 if (ifa->ifa_netmask == 0) { 1801 if (sa_equal(addr, ifa->ifa_addr) || 1802 (ifa->ifa_dstaddr && 1803 sa_equal(addr, ifa->ifa_dstaddr))) 1804 goto done; 1805 continue; 1806 } 1807 if (ifp->if_flags & IFF_POINTOPOINT) { 1808 if (sa_equal(addr, ifa->ifa_dstaddr)) 1809 goto done; 1810 } else { 1811 cp = addr->sa_data; 1812 cp2 = ifa->ifa_addr->sa_data; 1813 cp3 = ifa->ifa_netmask->sa_data; 1814 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1815 for (; cp3 < cplim; cp3++) 1816 if ((*cp++ ^ *cp2++) & *cp3) 1817 break; 1818 if (cp3 == cplim) 1819 goto done; 1820 } 1821 } 1822 ifa = ifa_maybe; 1823 done: 1824 if (ifa != NULL) 1825 ifa_ref(ifa); 1826 IF_ADDR_RUNLOCK(ifp); 1827 return (ifa); 1828 } 1829 1830 /* 1831 * See whether new ifa is better than current one: 1832 * 1) A non-virtual one is preferred over virtual. 1833 * 2) A virtual in master state preferred over any other state. 1834 * 1835 * Used in several address selecting functions. 1836 */ 1837 int 1838 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 1839 { 1840 1841 return (cur->ifa_carp && (!next->ifa_carp || 1842 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 1843 } 1844 1845 #include <net/if_llatbl.h> 1846 1847 /* 1848 * Default action when installing a route with a Link Level gateway. 1849 * Lookup an appropriate real ifa to point to. 1850 * This should be moved to /sys/net/link.c eventually. 1851 */ 1852 static void 1853 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 1854 { 1855 struct ifaddr *ifa, *oifa; 1856 struct sockaddr *dst; 1857 struct ifnet *ifp; 1858 1859 RT_LOCK_ASSERT(rt); 1860 1861 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 1862 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 1863 return; 1864 ifa = ifaof_ifpforaddr(dst, ifp); 1865 if (ifa) { 1866 oifa = rt->rt_ifa; 1867 rt->rt_ifa = ifa; 1868 ifa_free(oifa); 1869 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1870 ifa->ifa_rtrequest(cmd, rt, info); 1871 } 1872 } 1873 1874 /* 1875 * Mark an interface down and notify protocols of 1876 * the transition. 1877 */ 1878 static void 1879 if_unroute(struct ifnet *ifp, int flag, int fam) 1880 { 1881 struct ifaddr *ifa; 1882 1883 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 1884 1885 ifp->if_flags &= ~flag; 1886 getmicrotime(&ifp->if_lastchange); 1887 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1888 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1889 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1890 ifp->if_qflush(ifp); 1891 1892 if (ifp->if_carp) 1893 (*carp_linkstate_p)(ifp); 1894 rt_ifmsg(ifp); 1895 } 1896 1897 /* 1898 * Mark an interface up and notify protocols of 1899 * the transition. 1900 */ 1901 static void 1902 if_route(struct ifnet *ifp, int flag, int fam) 1903 { 1904 struct ifaddr *ifa; 1905 1906 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 1907 1908 ifp->if_flags |= flag; 1909 getmicrotime(&ifp->if_lastchange); 1910 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1911 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1912 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1913 if (ifp->if_carp) 1914 (*carp_linkstate_p)(ifp); 1915 rt_ifmsg(ifp); 1916 #ifdef INET6 1917 in6_if_up(ifp); 1918 #endif 1919 } 1920 1921 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 1922 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1923 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 1924 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 1925 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 1926 int (*vlan_setcookie_p)(struct ifnet *, void *); 1927 void *(*vlan_cookie_p)(struct ifnet *); 1928 1929 /* 1930 * Handle a change in the interface link state. To avoid LORs 1931 * between driver lock and upper layer locks, as well as possible 1932 * recursions, we post event to taskqueue, and all job 1933 * is done in static do_link_state_change(). 1934 */ 1935 void 1936 if_link_state_change(struct ifnet *ifp, int link_state) 1937 { 1938 /* Return if state hasn't changed. */ 1939 if (ifp->if_link_state == link_state) 1940 return; 1941 1942 ifp->if_link_state = link_state; 1943 1944 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 1945 } 1946 1947 static void 1948 do_link_state_change(void *arg, int pending) 1949 { 1950 struct ifnet *ifp = (struct ifnet *)arg; 1951 int link_state = ifp->if_link_state; 1952 CURVNET_SET(ifp->if_vnet); 1953 1954 /* Notify that the link state has changed. */ 1955 rt_ifmsg(ifp); 1956 if (ifp->if_vlantrunk != NULL) 1957 (*vlan_link_state_p)(ifp); 1958 1959 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 1960 IFP2AC(ifp)->ac_netgraph != NULL) 1961 (*ng_ether_link_state_p)(ifp, link_state); 1962 if (ifp->if_carp) 1963 (*carp_linkstate_p)(ifp); 1964 if (ifp->if_bridge) 1965 (*bridge_linkstate_p)(ifp); 1966 if (ifp->if_lagg) 1967 (*lagg_linkstate_p)(ifp, link_state); 1968 1969 if (IS_DEFAULT_VNET(curvnet)) 1970 devctl_notify("IFNET", ifp->if_xname, 1971 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 1972 NULL); 1973 if (pending > 1) 1974 if_printf(ifp, "%d link states coalesced\n", pending); 1975 if (log_link_state_change) 1976 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 1977 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 1978 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state); 1979 CURVNET_RESTORE(); 1980 } 1981 1982 /* 1983 * Mark an interface down and notify protocols of 1984 * the transition. 1985 */ 1986 void 1987 if_down(struct ifnet *ifp) 1988 { 1989 1990 if_unroute(ifp, IFF_UP, AF_UNSPEC); 1991 } 1992 1993 /* 1994 * Mark an interface up and notify protocols of 1995 * the transition. 1996 */ 1997 void 1998 if_up(struct ifnet *ifp) 1999 { 2000 2001 if_route(ifp, IFF_UP, AF_UNSPEC); 2002 } 2003 2004 /* 2005 * Flush an interface queue. 2006 */ 2007 void 2008 if_qflush(struct ifnet *ifp) 2009 { 2010 struct mbuf *m, *n; 2011 struct ifaltq *ifq; 2012 2013 ifq = &ifp->if_snd; 2014 IFQ_LOCK(ifq); 2015 #ifdef ALTQ 2016 if (ALTQ_IS_ENABLED(ifq)) 2017 ALTQ_PURGE(ifq); 2018 #endif 2019 n = ifq->ifq_head; 2020 while ((m = n) != 0) { 2021 n = m->m_act; 2022 m_freem(m); 2023 } 2024 ifq->ifq_head = 0; 2025 ifq->ifq_tail = 0; 2026 ifq->ifq_len = 0; 2027 IFQ_UNLOCK(ifq); 2028 } 2029 2030 /* 2031 * Map interface name to interface structure pointer, with or without 2032 * returning a reference. 2033 */ 2034 struct ifnet * 2035 ifunit_ref(const char *name) 2036 { 2037 struct ifnet *ifp; 2038 2039 IFNET_RLOCK_NOSLEEP(); 2040 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2041 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2042 !(ifp->if_flags & IFF_DYING)) 2043 break; 2044 } 2045 if (ifp != NULL) 2046 if_ref(ifp); 2047 IFNET_RUNLOCK_NOSLEEP(); 2048 return (ifp); 2049 } 2050 2051 struct ifnet * 2052 ifunit(const char *name) 2053 { 2054 struct ifnet *ifp; 2055 2056 IFNET_RLOCK_NOSLEEP(); 2057 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2058 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2059 break; 2060 } 2061 IFNET_RUNLOCK_NOSLEEP(); 2062 return (ifp); 2063 } 2064 2065 /* 2066 * Hardware specific interface ioctls. 2067 */ 2068 static int 2069 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2070 { 2071 struct ifreq *ifr; 2072 struct ifstat *ifs; 2073 int error = 0; 2074 int new_flags, temp_flags; 2075 size_t namelen, onamelen; 2076 size_t descrlen; 2077 char *descrbuf, *odescrbuf; 2078 char new_name[IFNAMSIZ]; 2079 struct ifaddr *ifa; 2080 struct sockaddr_dl *sdl; 2081 2082 ifr = (struct ifreq *)data; 2083 switch (cmd) { 2084 case SIOCGIFINDEX: 2085 ifr->ifr_index = ifp->if_index; 2086 break; 2087 2088 case SIOCGIFFLAGS: 2089 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2090 ifr->ifr_flags = temp_flags & 0xffff; 2091 ifr->ifr_flagshigh = temp_flags >> 16; 2092 break; 2093 2094 case SIOCGIFCAP: 2095 ifr->ifr_reqcap = ifp->if_capabilities; 2096 ifr->ifr_curcap = ifp->if_capenable; 2097 break; 2098 2099 #ifdef MAC 2100 case SIOCGIFMAC: 2101 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2102 break; 2103 #endif 2104 2105 case SIOCGIFMETRIC: 2106 ifr->ifr_metric = ifp->if_metric; 2107 break; 2108 2109 case SIOCGIFMTU: 2110 ifr->ifr_mtu = ifp->if_mtu; 2111 break; 2112 2113 case SIOCGIFPHYS: 2114 ifr->ifr_phys = ifp->if_physical; 2115 break; 2116 2117 case SIOCGIFDESCR: 2118 error = 0; 2119 sx_slock(&ifdescr_sx); 2120 if (ifp->if_description == NULL) 2121 error = ENOMSG; 2122 else { 2123 /* space for terminating nul */ 2124 descrlen = strlen(ifp->if_description) + 1; 2125 if (ifr->ifr_buffer.length < descrlen) 2126 ifr->ifr_buffer.buffer = NULL; 2127 else 2128 error = copyout(ifp->if_description, 2129 ifr->ifr_buffer.buffer, descrlen); 2130 ifr->ifr_buffer.length = descrlen; 2131 } 2132 sx_sunlock(&ifdescr_sx); 2133 break; 2134 2135 case SIOCSIFDESCR: 2136 error = priv_check(td, PRIV_NET_SETIFDESCR); 2137 if (error) 2138 return (error); 2139 2140 /* 2141 * Copy only (length-1) bytes to make sure that 2142 * if_description is always nul terminated. The 2143 * length parameter is supposed to count the 2144 * terminating nul in. 2145 */ 2146 if (ifr->ifr_buffer.length > ifdescr_maxlen) 2147 return (ENAMETOOLONG); 2148 else if (ifr->ifr_buffer.length == 0) 2149 descrbuf = NULL; 2150 else { 2151 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR, 2152 M_WAITOK | M_ZERO); 2153 error = copyin(ifr->ifr_buffer.buffer, descrbuf, 2154 ifr->ifr_buffer.length - 1); 2155 if (error) { 2156 free(descrbuf, M_IFDESCR); 2157 break; 2158 } 2159 } 2160 2161 sx_xlock(&ifdescr_sx); 2162 odescrbuf = ifp->if_description; 2163 ifp->if_description = descrbuf; 2164 sx_xunlock(&ifdescr_sx); 2165 2166 getmicrotime(&ifp->if_lastchange); 2167 free(odescrbuf, M_IFDESCR); 2168 break; 2169 2170 case SIOCGIFFIB: 2171 ifr->ifr_fib = ifp->if_fib; 2172 break; 2173 2174 case SIOCSIFFIB: 2175 error = priv_check(td, PRIV_NET_SETIFFIB); 2176 if (error) 2177 return (error); 2178 if (ifr->ifr_fib >= rt_numfibs) 2179 return (EINVAL); 2180 2181 ifp->if_fib = ifr->ifr_fib; 2182 break; 2183 2184 case SIOCSIFFLAGS: 2185 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2186 if (error) 2187 return (error); 2188 /* 2189 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2190 * check, so we don't need special handling here yet. 2191 */ 2192 new_flags = (ifr->ifr_flags & 0xffff) | 2193 (ifr->ifr_flagshigh << 16); 2194 if (ifp->if_flags & IFF_SMART) { 2195 /* Smart drivers twiddle their own routes */ 2196 } else if (ifp->if_flags & IFF_UP && 2197 (new_flags & IFF_UP) == 0) { 2198 if_down(ifp); 2199 } else if (new_flags & IFF_UP && 2200 (ifp->if_flags & IFF_UP) == 0) { 2201 if_up(ifp); 2202 } 2203 /* See if permanently promiscuous mode bit is about to flip */ 2204 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2205 if (new_flags & IFF_PPROMISC) 2206 ifp->if_flags |= IFF_PROMISC; 2207 else if (ifp->if_pcount == 0) 2208 ifp->if_flags &= ~IFF_PROMISC; 2209 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2210 ifp->if_xname, 2211 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); 2212 } 2213 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2214 (new_flags &~ IFF_CANTCHANGE); 2215 if (ifp->if_ioctl) { 2216 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2217 } 2218 getmicrotime(&ifp->if_lastchange); 2219 break; 2220 2221 case SIOCSIFCAP: 2222 error = priv_check(td, PRIV_NET_SETIFCAP); 2223 if (error) 2224 return (error); 2225 if (ifp->if_ioctl == NULL) 2226 return (EOPNOTSUPP); 2227 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2228 return (EINVAL); 2229 error = (*ifp->if_ioctl)(ifp, cmd, data); 2230 if (error == 0) 2231 getmicrotime(&ifp->if_lastchange); 2232 break; 2233 2234 #ifdef MAC 2235 case SIOCSIFMAC: 2236 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2237 break; 2238 #endif 2239 2240 case SIOCSIFNAME: 2241 error = priv_check(td, PRIV_NET_SETIFNAME); 2242 if (error) 2243 return (error); 2244 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 2245 if (error != 0) 2246 return (error); 2247 if (new_name[0] == '\0') 2248 return (EINVAL); 2249 if (ifunit(new_name) != NULL) 2250 return (EEXIST); 2251 2252 /* 2253 * XXX: Locking. Nothing else seems to lock if_flags, 2254 * and there are numerous other races with the 2255 * ifunit() checks not being atomic with namespace 2256 * changes (renames, vmoves, if_attach, etc). 2257 */ 2258 ifp->if_flags |= IFF_RENAMING; 2259 2260 /* Announce the departure of the interface. */ 2261 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2262 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2263 2264 log(LOG_INFO, "%s: changing name to '%s'\n", 2265 ifp->if_xname, new_name); 2266 2267 IF_ADDR_WLOCK(ifp); 2268 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2269 ifa = ifp->if_addr; 2270 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2271 namelen = strlen(new_name); 2272 onamelen = sdl->sdl_nlen; 2273 /* 2274 * Move the address if needed. This is safe because we 2275 * allocate space for a name of length IFNAMSIZ when we 2276 * create this in if_attach(). 2277 */ 2278 if (namelen != onamelen) { 2279 bcopy(sdl->sdl_data + onamelen, 2280 sdl->sdl_data + namelen, sdl->sdl_alen); 2281 } 2282 bcopy(new_name, sdl->sdl_data, namelen); 2283 sdl->sdl_nlen = namelen; 2284 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2285 bzero(sdl->sdl_data, onamelen); 2286 while (namelen != 0) 2287 sdl->sdl_data[--namelen] = 0xff; 2288 IF_ADDR_WUNLOCK(ifp); 2289 2290 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2291 /* Announce the return of the interface. */ 2292 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2293 2294 ifp->if_flags &= ~IFF_RENAMING; 2295 break; 2296 2297 #ifdef VIMAGE 2298 case SIOCSIFVNET: 2299 error = priv_check(td, PRIV_NET_SETIFVNET); 2300 if (error) 2301 return (error); 2302 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2303 break; 2304 #endif 2305 2306 case SIOCSIFMETRIC: 2307 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2308 if (error) 2309 return (error); 2310 ifp->if_metric = ifr->ifr_metric; 2311 getmicrotime(&ifp->if_lastchange); 2312 break; 2313 2314 case SIOCSIFPHYS: 2315 error = priv_check(td, PRIV_NET_SETIFPHYS); 2316 if (error) 2317 return (error); 2318 if (ifp->if_ioctl == NULL) 2319 return (EOPNOTSUPP); 2320 error = (*ifp->if_ioctl)(ifp, cmd, data); 2321 if (error == 0) 2322 getmicrotime(&ifp->if_lastchange); 2323 break; 2324 2325 case SIOCSIFMTU: 2326 { 2327 u_long oldmtu = ifp->if_mtu; 2328 2329 error = priv_check(td, PRIV_NET_SETIFMTU); 2330 if (error) 2331 return (error); 2332 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2333 return (EINVAL); 2334 if (ifp->if_ioctl == NULL) 2335 return (EOPNOTSUPP); 2336 error = (*ifp->if_ioctl)(ifp, cmd, data); 2337 if (error == 0) { 2338 getmicrotime(&ifp->if_lastchange); 2339 rt_ifmsg(ifp); 2340 } 2341 /* 2342 * If the link MTU changed, do network layer specific procedure. 2343 */ 2344 if (ifp->if_mtu != oldmtu) { 2345 #ifdef INET6 2346 nd6_setmtu(ifp); 2347 #endif 2348 } 2349 break; 2350 } 2351 2352 case SIOCADDMULTI: 2353 case SIOCDELMULTI: 2354 if (cmd == SIOCADDMULTI) 2355 error = priv_check(td, PRIV_NET_ADDMULTI); 2356 else 2357 error = priv_check(td, PRIV_NET_DELMULTI); 2358 if (error) 2359 return (error); 2360 2361 /* Don't allow group membership on non-multicast interfaces. */ 2362 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2363 return (EOPNOTSUPP); 2364 2365 /* Don't let users screw up protocols' entries. */ 2366 if (ifr->ifr_addr.sa_family != AF_LINK) 2367 return (EINVAL); 2368 2369 if (cmd == SIOCADDMULTI) { 2370 struct ifmultiaddr *ifma; 2371 2372 /* 2373 * Userland is only permitted to join groups once 2374 * via the if_addmulti() KPI, because it cannot hold 2375 * struct ifmultiaddr * between calls. It may also 2376 * lose a race while we check if the membership 2377 * already exists. 2378 */ 2379 IF_ADDR_RLOCK(ifp); 2380 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2381 IF_ADDR_RUNLOCK(ifp); 2382 if (ifma != NULL) 2383 error = EADDRINUSE; 2384 else 2385 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2386 } else { 2387 error = if_delmulti(ifp, &ifr->ifr_addr); 2388 } 2389 if (error == 0) 2390 getmicrotime(&ifp->if_lastchange); 2391 break; 2392 2393 case SIOCSIFPHYADDR: 2394 case SIOCDIFPHYADDR: 2395 #ifdef INET6 2396 case SIOCSIFPHYADDR_IN6: 2397 #endif 2398 case SIOCSLIFPHYADDR: 2399 case SIOCSIFMEDIA: 2400 case SIOCSIFGENERIC: 2401 error = priv_check(td, PRIV_NET_HWIOCTL); 2402 if (error) 2403 return (error); 2404 if (ifp->if_ioctl == NULL) 2405 return (EOPNOTSUPP); 2406 error = (*ifp->if_ioctl)(ifp, cmd, data); 2407 if (error == 0) 2408 getmicrotime(&ifp->if_lastchange); 2409 break; 2410 2411 case SIOCGIFSTATUS: 2412 ifs = (struct ifstat *)data; 2413 ifs->ascii[0] = '\0'; 2414 2415 case SIOCGIFPSRCADDR: 2416 case SIOCGIFPDSTADDR: 2417 case SIOCGLIFPHYADDR: 2418 case SIOCGIFMEDIA: 2419 case SIOCGIFGENERIC: 2420 if (ifp->if_ioctl == NULL) 2421 return (EOPNOTSUPP); 2422 error = (*ifp->if_ioctl)(ifp, cmd, data); 2423 break; 2424 2425 case SIOCSIFLLADDR: 2426 error = priv_check(td, PRIV_NET_SETLLADDR); 2427 if (error) 2428 return (error); 2429 error = if_setlladdr(ifp, 2430 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2431 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 2432 break; 2433 2434 case SIOCAIFGROUP: 2435 { 2436 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2437 2438 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2439 if (error) 2440 return (error); 2441 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2442 return (error); 2443 break; 2444 } 2445 2446 case SIOCGIFGROUP: 2447 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2448 return (error); 2449 break; 2450 2451 case SIOCDIFGROUP: 2452 { 2453 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2454 2455 error = priv_check(td, PRIV_NET_DELIFGROUP); 2456 if (error) 2457 return (error); 2458 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2459 return (error); 2460 break; 2461 } 2462 2463 default: 2464 error = ENOIOCTL; 2465 break; 2466 } 2467 return (error); 2468 } 2469 2470 #ifdef COMPAT_FREEBSD32 2471 struct ifconf32 { 2472 int32_t ifc_len; 2473 union { 2474 uint32_t ifcu_buf; 2475 uint32_t ifcu_req; 2476 } ifc_ifcu; 2477 }; 2478 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2479 #endif 2480 2481 /* 2482 * Interface ioctls. 2483 */ 2484 int 2485 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2486 { 2487 struct ifnet *ifp; 2488 struct ifreq *ifr; 2489 int error; 2490 int oif_flags; 2491 2492 CURVNET_SET(so->so_vnet); 2493 switch (cmd) { 2494 case SIOCGIFCONF: 2495 case OSIOCGIFCONF: 2496 error = ifconf(cmd, data); 2497 CURVNET_RESTORE(); 2498 return (error); 2499 2500 #ifdef COMPAT_FREEBSD32 2501 case SIOCGIFCONF32: 2502 { 2503 struct ifconf32 *ifc32; 2504 struct ifconf ifc; 2505 2506 ifc32 = (struct ifconf32 *)data; 2507 ifc.ifc_len = ifc32->ifc_len; 2508 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2509 2510 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2511 CURVNET_RESTORE(); 2512 if (error == 0) 2513 ifc32->ifc_len = ifc.ifc_len; 2514 return (error); 2515 } 2516 #endif 2517 } 2518 ifr = (struct ifreq *)data; 2519 2520 switch (cmd) { 2521 #ifdef VIMAGE 2522 case SIOCSIFRVNET: 2523 error = priv_check(td, PRIV_NET_SETIFVNET); 2524 if (error == 0) 2525 error = if_vmove_reclaim(td, ifr->ifr_name, 2526 ifr->ifr_jid); 2527 CURVNET_RESTORE(); 2528 return (error); 2529 #endif 2530 case SIOCIFCREATE: 2531 case SIOCIFCREATE2: 2532 error = priv_check(td, PRIV_NET_IFCREATE); 2533 if (error == 0) 2534 error = if_clone_create(ifr->ifr_name, 2535 sizeof(ifr->ifr_name), 2536 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL); 2537 CURVNET_RESTORE(); 2538 return (error); 2539 case SIOCIFDESTROY: 2540 error = priv_check(td, PRIV_NET_IFDESTROY); 2541 if (error == 0) 2542 error = if_clone_destroy(ifr->ifr_name); 2543 CURVNET_RESTORE(); 2544 return (error); 2545 2546 case SIOCIFGCLONERS: 2547 error = if_clone_list((struct if_clonereq *)data); 2548 CURVNET_RESTORE(); 2549 return (error); 2550 case SIOCGIFGMEMB: 2551 error = if_getgroupmembers((struct ifgroupreq *)data); 2552 CURVNET_RESTORE(); 2553 return (error); 2554 #if defined(INET) || defined(INET6) 2555 case SIOCSVH: 2556 case SIOCGVH: 2557 if (carp_ioctl_p == NULL) 2558 error = EPROTONOSUPPORT; 2559 else 2560 error = (*carp_ioctl_p)(ifr, cmd, td); 2561 CURVNET_RESTORE(); 2562 return (error); 2563 #endif 2564 } 2565 2566 ifp = ifunit_ref(ifr->ifr_name); 2567 if (ifp == NULL) { 2568 CURVNET_RESTORE(); 2569 return (ENXIO); 2570 } 2571 2572 error = ifhwioctl(cmd, ifp, data, td); 2573 if (error != ENOIOCTL) { 2574 if_rele(ifp); 2575 CURVNET_RESTORE(); 2576 return (error); 2577 } 2578 2579 oif_flags = ifp->if_flags; 2580 if (so->so_proto == NULL) { 2581 if_rele(ifp); 2582 CURVNET_RESTORE(); 2583 return (EOPNOTSUPP); 2584 } 2585 2586 /* 2587 * Pass the request on to the socket control method, and if the 2588 * latter returns EOPNOTSUPP, directly to the interface. 2589 * 2590 * Make an exception for the legacy SIOCSIF* requests. Drivers 2591 * trust SIOCSIFADDR et al to come from an already privileged 2592 * layer, and do not perform any credentials checks or input 2593 * validation. 2594 */ 2595 #ifndef COMPAT_43 2596 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 2597 data, 2598 ifp, td)); 2599 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2600 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2601 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2602 error = (*ifp->if_ioctl)(ifp, cmd, data); 2603 #else 2604 { 2605 u_long ocmd = cmd; 2606 2607 switch (cmd) { 2608 2609 case SIOCSIFDSTADDR: 2610 case SIOCSIFADDR: 2611 case SIOCSIFBRDADDR: 2612 case SIOCSIFNETMASK: 2613 #if BYTE_ORDER != BIG_ENDIAN 2614 if (ifr->ifr_addr.sa_family == 0 && 2615 ifr->ifr_addr.sa_len < 16) { 2616 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 2617 ifr->ifr_addr.sa_len = 16; 2618 } 2619 #else 2620 if (ifr->ifr_addr.sa_len == 0) 2621 ifr->ifr_addr.sa_len = 16; 2622 #endif 2623 break; 2624 2625 case OSIOCGIFADDR: 2626 cmd = SIOCGIFADDR; 2627 break; 2628 2629 case OSIOCGIFDSTADDR: 2630 cmd = SIOCGIFDSTADDR; 2631 break; 2632 2633 case OSIOCGIFBRDADDR: 2634 cmd = SIOCGIFBRDADDR; 2635 break; 2636 2637 case OSIOCGIFNETMASK: 2638 cmd = SIOCGIFNETMASK; 2639 } 2640 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 2641 cmd, 2642 data, 2643 ifp, td)); 2644 if (error == EOPNOTSUPP && ifp != NULL && 2645 ifp->if_ioctl != NULL && 2646 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2647 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2648 error = (*ifp->if_ioctl)(ifp, cmd, data); 2649 switch (ocmd) { 2650 2651 case OSIOCGIFADDR: 2652 case OSIOCGIFDSTADDR: 2653 case OSIOCGIFBRDADDR: 2654 case OSIOCGIFNETMASK: 2655 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 2656 2657 } 2658 } 2659 #endif /* COMPAT_43 */ 2660 2661 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2662 #ifdef INET6 2663 if (ifp->if_flags & IFF_UP) 2664 in6_if_up(ifp); 2665 #endif 2666 } 2667 if_rele(ifp); 2668 CURVNET_RESTORE(); 2669 return (error); 2670 } 2671 2672 /* 2673 * The code common to handling reference counted flags, 2674 * e.g., in ifpromisc() and if_allmulti(). 2675 * The "pflag" argument can specify a permanent mode flag to check, 2676 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2677 * 2678 * Only to be used on stack-owned flags, not driver-owned flags. 2679 */ 2680 static int 2681 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 2682 { 2683 struct ifreq ifr; 2684 int error; 2685 int oldflags, oldcount; 2686 2687 /* Sanity checks to catch programming errors */ 2688 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 2689 ("%s: setting driver-owned flag %d", __func__, flag)); 2690 2691 if (onswitch) 2692 KASSERT(*refcount >= 0, 2693 ("%s: increment negative refcount %d for flag %d", 2694 __func__, *refcount, flag)); 2695 else 2696 KASSERT(*refcount > 0, 2697 ("%s: decrement non-positive refcount %d for flag %d", 2698 __func__, *refcount, flag)); 2699 2700 /* In case this mode is permanent, just touch refcount */ 2701 if (ifp->if_flags & pflag) { 2702 *refcount += onswitch ? 1 : -1; 2703 return (0); 2704 } 2705 2706 /* Save ifnet parameters for if_ioctl() may fail */ 2707 oldcount = *refcount; 2708 oldflags = ifp->if_flags; 2709 2710 /* 2711 * See if we aren't the only and touching refcount is enough. 2712 * Actually toggle interface flag if we are the first or last. 2713 */ 2714 if (onswitch) { 2715 if ((*refcount)++) 2716 return (0); 2717 ifp->if_flags |= flag; 2718 } else { 2719 if (--(*refcount)) 2720 return (0); 2721 ifp->if_flags &= ~flag; 2722 } 2723 2724 /* Call down the driver since we've changed interface flags */ 2725 if (ifp->if_ioctl == NULL) { 2726 error = EOPNOTSUPP; 2727 goto recover; 2728 } 2729 ifr.ifr_flags = ifp->if_flags & 0xffff; 2730 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2731 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2732 if (error) 2733 goto recover; 2734 /* Notify userland that interface flags have changed */ 2735 rt_ifmsg(ifp); 2736 return (0); 2737 2738 recover: 2739 /* Recover after driver error */ 2740 *refcount = oldcount; 2741 ifp->if_flags = oldflags; 2742 return (error); 2743 } 2744 2745 /* 2746 * Set/clear promiscuous mode on interface ifp based on the truth value 2747 * of pswitch. The calls are reference counted so that only the first 2748 * "on" request actually has an effect, as does the final "off" request. 2749 * Results are undefined if the "off" and "on" requests are not matched. 2750 */ 2751 int 2752 ifpromisc(struct ifnet *ifp, int pswitch) 2753 { 2754 int error; 2755 int oldflags = ifp->if_flags; 2756 2757 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 2758 &ifp->if_pcount, pswitch); 2759 /* If promiscuous mode status has changed, log a message */ 2760 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) 2761 log(LOG_INFO, "%s: promiscuous mode %s\n", 2762 ifp->if_xname, 2763 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 2764 return (error); 2765 } 2766 2767 /* 2768 * Return interface configuration 2769 * of system. List may be used 2770 * in later ioctl's (above) to get 2771 * other information. 2772 */ 2773 /*ARGSUSED*/ 2774 static int 2775 ifconf(u_long cmd, caddr_t data) 2776 { 2777 struct ifconf *ifc = (struct ifconf *)data; 2778 struct ifnet *ifp; 2779 struct ifaddr *ifa; 2780 struct ifreq ifr; 2781 struct sbuf *sb; 2782 int error, full = 0, valid_len, max_len; 2783 2784 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2785 max_len = MAXPHYS - 1; 2786 2787 /* Prevent hostile input from being able to crash the system */ 2788 if (ifc->ifc_len <= 0) 2789 return (EINVAL); 2790 2791 again: 2792 if (ifc->ifc_len <= max_len) { 2793 max_len = ifc->ifc_len; 2794 full = 1; 2795 } 2796 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2797 max_len = 0; 2798 valid_len = 0; 2799 2800 IFNET_RLOCK(); 2801 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2802 int addrs; 2803 2804 /* 2805 * Zero the ifr_name buffer to make sure we don't 2806 * disclose the contents of the stack. 2807 */ 2808 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 2809 2810 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 2811 >= sizeof(ifr.ifr_name)) { 2812 sbuf_delete(sb); 2813 IFNET_RUNLOCK(); 2814 return (ENAMETOOLONG); 2815 } 2816 2817 addrs = 0; 2818 IF_ADDR_RLOCK(ifp); 2819 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2820 struct sockaddr *sa = ifa->ifa_addr; 2821 2822 if (prison_if(curthread->td_ucred, sa) != 0) 2823 continue; 2824 addrs++; 2825 #ifdef COMPAT_43 2826 if (cmd == OSIOCGIFCONF) { 2827 struct osockaddr *osa = 2828 (struct osockaddr *)&ifr.ifr_addr; 2829 ifr.ifr_addr = *sa; 2830 osa->sa_family = sa->sa_family; 2831 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2832 max_len += sizeof(ifr); 2833 } else 2834 #endif 2835 if (sa->sa_len <= sizeof(*sa)) { 2836 ifr.ifr_addr = *sa; 2837 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2838 max_len += sizeof(ifr); 2839 } else { 2840 sbuf_bcat(sb, &ifr, 2841 offsetof(struct ifreq, ifr_addr)); 2842 max_len += offsetof(struct ifreq, ifr_addr); 2843 sbuf_bcat(sb, sa, sa->sa_len); 2844 max_len += sa->sa_len; 2845 } 2846 2847 if (sbuf_error(sb) == 0) 2848 valid_len = sbuf_len(sb); 2849 } 2850 IF_ADDR_RUNLOCK(ifp); 2851 if (addrs == 0) { 2852 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 2853 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2854 max_len += sizeof(ifr); 2855 2856 if (sbuf_error(sb) == 0) 2857 valid_len = sbuf_len(sb); 2858 } 2859 } 2860 IFNET_RUNLOCK(); 2861 2862 /* 2863 * If we didn't allocate enough space (uncommon), try again. If 2864 * we have already allocated as much space as we are allowed, 2865 * return what we've got. 2866 */ 2867 if (valid_len != max_len && !full) { 2868 sbuf_delete(sb); 2869 goto again; 2870 } 2871 2872 ifc->ifc_len = valid_len; 2873 sbuf_finish(sb); 2874 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 2875 sbuf_delete(sb); 2876 return (error); 2877 } 2878 2879 /* 2880 * Just like ifpromisc(), but for all-multicast-reception mode. 2881 */ 2882 int 2883 if_allmulti(struct ifnet *ifp, int onswitch) 2884 { 2885 2886 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 2887 } 2888 2889 struct ifmultiaddr * 2890 if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 2891 { 2892 struct ifmultiaddr *ifma; 2893 2894 IF_ADDR_LOCK_ASSERT(ifp); 2895 2896 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2897 if (sa->sa_family == AF_LINK) { 2898 if (sa_dl_equal(ifma->ifma_addr, sa)) 2899 break; 2900 } else { 2901 if (sa_equal(ifma->ifma_addr, sa)) 2902 break; 2903 } 2904 } 2905 2906 return ifma; 2907 } 2908 2909 /* 2910 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 2911 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 2912 * the ifnet multicast address list here, so the caller must do that and 2913 * other setup work (such as notifying the device driver). The reference 2914 * count is initialized to 1. 2915 */ 2916 static struct ifmultiaddr * 2917 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 2918 int mflags) 2919 { 2920 struct ifmultiaddr *ifma; 2921 struct sockaddr *dupsa; 2922 2923 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 2924 M_ZERO); 2925 if (ifma == NULL) 2926 return (NULL); 2927 2928 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 2929 if (dupsa == NULL) { 2930 free(ifma, M_IFMADDR); 2931 return (NULL); 2932 } 2933 bcopy(sa, dupsa, sa->sa_len); 2934 ifma->ifma_addr = dupsa; 2935 2936 ifma->ifma_ifp = ifp; 2937 ifma->ifma_refcount = 1; 2938 ifma->ifma_protospec = NULL; 2939 2940 if (llsa == NULL) { 2941 ifma->ifma_lladdr = NULL; 2942 return (ifma); 2943 } 2944 2945 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 2946 if (dupsa == NULL) { 2947 free(ifma->ifma_addr, M_IFMADDR); 2948 free(ifma, M_IFMADDR); 2949 return (NULL); 2950 } 2951 bcopy(llsa, dupsa, llsa->sa_len); 2952 ifma->ifma_lladdr = dupsa; 2953 2954 return (ifma); 2955 } 2956 2957 /* 2958 * if_freemulti: free ifmultiaddr structure and possibly attached related 2959 * addresses. The caller is responsible for implementing reference 2960 * counting, notifying the driver, handling routing messages, and releasing 2961 * any dependent link layer state. 2962 */ 2963 static void 2964 if_freemulti(struct ifmultiaddr *ifma) 2965 { 2966 2967 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 2968 ifma->ifma_refcount)); 2969 KASSERT(ifma->ifma_protospec == NULL, 2970 ("if_freemulti: protospec not NULL")); 2971 2972 if (ifma->ifma_lladdr != NULL) 2973 free(ifma->ifma_lladdr, M_IFMADDR); 2974 free(ifma->ifma_addr, M_IFMADDR); 2975 free(ifma, M_IFMADDR); 2976 } 2977 2978 /* 2979 * Register an additional multicast address with a network interface. 2980 * 2981 * - If the address is already present, bump the reference count on the 2982 * address and return. 2983 * - If the address is not link-layer, look up a link layer address. 2984 * - Allocate address structures for one or both addresses, and attach to the 2985 * multicast address list on the interface. If automatically adding a link 2986 * layer address, the protocol address will own a reference to the link 2987 * layer address, to be freed when it is freed. 2988 * - Notify the network device driver of an addition to the multicast address 2989 * list. 2990 * 2991 * 'sa' points to caller-owned memory with the desired multicast address. 2992 * 2993 * 'retifma' will be used to return a pointer to the resulting multicast 2994 * address reference, if desired. 2995 */ 2996 int 2997 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 2998 struct ifmultiaddr **retifma) 2999 { 3000 struct ifmultiaddr *ifma, *ll_ifma; 3001 struct sockaddr *llsa; 3002 int error; 3003 3004 /* 3005 * If the address is already present, return a new reference to it; 3006 * otherwise, allocate storage and set up a new address. 3007 */ 3008 IF_ADDR_WLOCK(ifp); 3009 ifma = if_findmulti(ifp, sa); 3010 if (ifma != NULL) { 3011 ifma->ifma_refcount++; 3012 if (retifma != NULL) 3013 *retifma = ifma; 3014 IF_ADDR_WUNLOCK(ifp); 3015 return (0); 3016 } 3017 3018 /* 3019 * The address isn't already present; resolve the protocol address 3020 * into a link layer address, and then look that up, bump its 3021 * refcount or allocate an ifma for that also. If 'llsa' was 3022 * returned, we will need to free it later. 3023 */ 3024 llsa = NULL; 3025 ll_ifma = NULL; 3026 if (ifp->if_resolvemulti != NULL) { 3027 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3028 if (error) 3029 goto unlock_out; 3030 } 3031 3032 /* 3033 * Allocate the new address. Don't hook it up yet, as we may also 3034 * need to allocate a link layer multicast address. 3035 */ 3036 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3037 if (ifma == NULL) { 3038 error = ENOMEM; 3039 goto free_llsa_out; 3040 } 3041 3042 /* 3043 * If a link layer address is found, we'll need to see if it's 3044 * already present in the address list, or allocate is as well. 3045 * When this block finishes, the link layer address will be on the 3046 * list. 3047 */ 3048 if (llsa != NULL) { 3049 ll_ifma = if_findmulti(ifp, llsa); 3050 if (ll_ifma == NULL) { 3051 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3052 if (ll_ifma == NULL) { 3053 --ifma->ifma_refcount; 3054 if_freemulti(ifma); 3055 error = ENOMEM; 3056 goto free_llsa_out; 3057 } 3058 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3059 ifma_link); 3060 } else 3061 ll_ifma->ifma_refcount++; 3062 ifma->ifma_llifma = ll_ifma; 3063 } 3064 3065 /* 3066 * We now have a new multicast address, ifma, and possibly a new or 3067 * referenced link layer address. Add the primary address to the 3068 * ifnet address list. 3069 */ 3070 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3071 3072 if (retifma != NULL) 3073 *retifma = ifma; 3074 3075 /* 3076 * Must generate the message while holding the lock so that 'ifma' 3077 * pointer is still valid. 3078 */ 3079 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3080 IF_ADDR_WUNLOCK(ifp); 3081 3082 /* 3083 * We are certain we have added something, so call down to the 3084 * interface to let them know about it. 3085 */ 3086 if (ifp->if_ioctl != NULL) { 3087 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3088 } 3089 3090 if (llsa != NULL) 3091 free(llsa, M_IFMADDR); 3092 3093 return (0); 3094 3095 free_llsa_out: 3096 if (llsa != NULL) 3097 free(llsa, M_IFMADDR); 3098 3099 unlock_out: 3100 IF_ADDR_WUNLOCK(ifp); 3101 return (error); 3102 } 3103 3104 /* 3105 * Delete a multicast group membership by network-layer group address. 3106 * 3107 * Returns ENOENT if the entry could not be found. If ifp no longer 3108 * exists, results are undefined. This entry point should only be used 3109 * from subsystems which do appropriate locking to hold ifp for the 3110 * duration of the call. 3111 * Network-layer protocol domains must use if_delmulti_ifma(). 3112 */ 3113 int 3114 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3115 { 3116 struct ifmultiaddr *ifma; 3117 int lastref; 3118 #ifdef INVARIANTS 3119 struct ifnet *oifp; 3120 3121 IFNET_RLOCK_NOSLEEP(); 3122 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3123 if (ifp == oifp) 3124 break; 3125 if (ifp != oifp) 3126 ifp = NULL; 3127 IFNET_RUNLOCK_NOSLEEP(); 3128 3129 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3130 #endif 3131 if (ifp == NULL) 3132 return (ENOENT); 3133 3134 IF_ADDR_WLOCK(ifp); 3135 lastref = 0; 3136 ifma = if_findmulti(ifp, sa); 3137 if (ifma != NULL) 3138 lastref = if_delmulti_locked(ifp, ifma, 0); 3139 IF_ADDR_WUNLOCK(ifp); 3140 3141 if (ifma == NULL) 3142 return (ENOENT); 3143 3144 if (lastref && ifp->if_ioctl != NULL) { 3145 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3146 } 3147 3148 return (0); 3149 } 3150 3151 /* 3152 * Delete all multicast group membership for an interface. 3153 * Should be used to quickly flush all multicast filters. 3154 */ 3155 void 3156 if_delallmulti(struct ifnet *ifp) 3157 { 3158 struct ifmultiaddr *ifma; 3159 struct ifmultiaddr *next; 3160 3161 IF_ADDR_WLOCK(ifp); 3162 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3163 if_delmulti_locked(ifp, ifma, 0); 3164 IF_ADDR_WUNLOCK(ifp); 3165 } 3166 3167 /* 3168 * Delete a multicast group membership by group membership pointer. 3169 * Network-layer protocol domains must use this routine. 3170 * 3171 * It is safe to call this routine if the ifp disappeared. 3172 */ 3173 void 3174 if_delmulti_ifma(struct ifmultiaddr *ifma) 3175 { 3176 struct ifnet *ifp; 3177 int lastref; 3178 3179 ifp = ifma->ifma_ifp; 3180 #ifdef DIAGNOSTIC 3181 if (ifp == NULL) { 3182 printf("%s: ifma_ifp seems to be detached\n", __func__); 3183 } else { 3184 struct ifnet *oifp; 3185 3186 IFNET_RLOCK_NOSLEEP(); 3187 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3188 if (ifp == oifp) 3189 break; 3190 if (ifp != oifp) { 3191 printf("%s: ifnet %p disappeared\n", __func__, ifp); 3192 ifp = NULL; 3193 } 3194 IFNET_RUNLOCK_NOSLEEP(); 3195 } 3196 #endif 3197 /* 3198 * If and only if the ifnet instance exists: Acquire the address lock. 3199 */ 3200 if (ifp != NULL) 3201 IF_ADDR_WLOCK(ifp); 3202 3203 lastref = if_delmulti_locked(ifp, ifma, 0); 3204 3205 if (ifp != NULL) { 3206 /* 3207 * If and only if the ifnet instance exists: 3208 * Release the address lock. 3209 * If the group was left: update the hardware hash filter. 3210 */ 3211 IF_ADDR_WUNLOCK(ifp); 3212 if (lastref && ifp->if_ioctl != NULL) { 3213 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3214 } 3215 } 3216 } 3217 3218 /* 3219 * Perform deletion of network-layer and/or link-layer multicast address. 3220 * 3221 * Return 0 if the reference count was decremented. 3222 * Return 1 if the final reference was released, indicating that the 3223 * hardware hash filter should be reprogrammed. 3224 */ 3225 static int 3226 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3227 { 3228 struct ifmultiaddr *ll_ifma; 3229 3230 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3231 KASSERT(ifma->ifma_ifp == ifp, 3232 ("%s: inconsistent ifp %p", __func__, ifp)); 3233 IF_ADDR_WLOCK_ASSERT(ifp); 3234 } 3235 3236 ifp = ifma->ifma_ifp; 3237 3238 /* 3239 * If the ifnet is detaching, null out references to ifnet, 3240 * so that upper protocol layers will notice, and not attempt 3241 * to obtain locks for an ifnet which no longer exists. The 3242 * routing socket announcement must happen before the ifnet 3243 * instance is detached from the system. 3244 */ 3245 if (detaching) { 3246 #ifdef DIAGNOSTIC 3247 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3248 #endif 3249 /* 3250 * ifp may already be nulled out if we are being reentered 3251 * to delete the ll_ifma. 3252 */ 3253 if (ifp != NULL) { 3254 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3255 ifma->ifma_ifp = NULL; 3256 } 3257 } 3258 3259 if (--ifma->ifma_refcount > 0) 3260 return 0; 3261 3262 /* 3263 * If this ifma is a network-layer ifma, a link-layer ifma may 3264 * have been associated with it. Release it first if so. 3265 */ 3266 ll_ifma = ifma->ifma_llifma; 3267 if (ll_ifma != NULL) { 3268 KASSERT(ifma->ifma_lladdr != NULL, 3269 ("%s: llifma w/o lladdr", __func__)); 3270 if (detaching) 3271 ll_ifma->ifma_ifp = NULL; /* XXX */ 3272 if (--ll_ifma->ifma_refcount == 0) { 3273 if (ifp != NULL) { 3274 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3275 ifma_link); 3276 } 3277 if_freemulti(ll_ifma); 3278 } 3279 } 3280 3281 if (ifp != NULL) 3282 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3283 3284 if_freemulti(ifma); 3285 3286 /* 3287 * The last reference to this instance of struct ifmultiaddr 3288 * was released; the hardware should be notified of this change. 3289 */ 3290 return 1; 3291 } 3292 3293 /* 3294 * Set the link layer address on an interface. 3295 * 3296 * At this time we only support certain types of interfaces, 3297 * and we don't allow the length of the address to change. 3298 */ 3299 int 3300 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3301 { 3302 struct sockaddr_dl *sdl; 3303 struct ifaddr *ifa; 3304 struct ifreq ifr; 3305 3306 IF_ADDR_RLOCK(ifp); 3307 ifa = ifp->if_addr; 3308 if (ifa == NULL) { 3309 IF_ADDR_RUNLOCK(ifp); 3310 return (EINVAL); 3311 } 3312 ifa_ref(ifa); 3313 IF_ADDR_RUNLOCK(ifp); 3314 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3315 if (sdl == NULL) { 3316 ifa_free(ifa); 3317 return (EINVAL); 3318 } 3319 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3320 ifa_free(ifa); 3321 return (EINVAL); 3322 } 3323 switch (ifp->if_type) { 3324 case IFT_ETHER: 3325 case IFT_FDDI: 3326 case IFT_XETHER: 3327 case IFT_ISO88025: 3328 case IFT_L2VLAN: 3329 case IFT_BRIDGE: 3330 case IFT_ARCNET: 3331 case IFT_IEEE8023ADLAG: 3332 case IFT_IEEE80211: 3333 bcopy(lladdr, LLADDR(sdl), len); 3334 ifa_free(ifa); 3335 break; 3336 default: 3337 ifa_free(ifa); 3338 return (ENODEV); 3339 } 3340 3341 /* 3342 * If the interface is already up, we need 3343 * to re-init it in order to reprogram its 3344 * address filter. 3345 */ 3346 if ((ifp->if_flags & IFF_UP) != 0) { 3347 if (ifp->if_ioctl) { 3348 ifp->if_flags &= ~IFF_UP; 3349 ifr.ifr_flags = ifp->if_flags & 0xffff; 3350 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3351 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3352 ifp->if_flags |= IFF_UP; 3353 ifr.ifr_flags = ifp->if_flags & 0xffff; 3354 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3355 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3356 } 3357 #ifdef INET 3358 /* 3359 * Also send gratuitous ARPs to notify other nodes about 3360 * the address change. 3361 */ 3362 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3363 if (ifa->ifa_addr->sa_family == AF_INET) 3364 arp_ifinit(ifp, ifa); 3365 } 3366 #endif 3367 } 3368 return (0); 3369 } 3370 3371 /* 3372 * The name argument must be a pointer to storage which will last as 3373 * long as the interface does. For physical devices, the result of 3374 * device_get_name(dev) is a good choice and for pseudo-devices a 3375 * static string works well. 3376 */ 3377 void 3378 if_initname(struct ifnet *ifp, const char *name, int unit) 3379 { 3380 ifp->if_dname = name; 3381 ifp->if_dunit = unit; 3382 if (unit != IF_DUNIT_NONE) 3383 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3384 else 3385 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3386 } 3387 3388 int 3389 if_printf(struct ifnet *ifp, const char * fmt, ...) 3390 { 3391 va_list ap; 3392 int retval; 3393 3394 retval = printf("%s: ", ifp->if_xname); 3395 va_start(ap, fmt); 3396 retval += vprintf(fmt, ap); 3397 va_end(ap); 3398 return (retval); 3399 } 3400 3401 void 3402 if_start(struct ifnet *ifp) 3403 { 3404 3405 (*(ifp)->if_start)(ifp); 3406 } 3407 3408 /* 3409 * Backwards compatibility interface for drivers 3410 * that have not implemented it 3411 */ 3412 static int 3413 if_transmit(struct ifnet *ifp, struct mbuf *m) 3414 { 3415 int error; 3416 3417 IFQ_HANDOFF(ifp, m, error); 3418 return (error); 3419 } 3420 3421 int 3422 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3423 { 3424 int active = 0; 3425 3426 IF_LOCK(ifq); 3427 if (_IF_QFULL(ifq)) { 3428 _IF_DROP(ifq); 3429 IF_UNLOCK(ifq); 3430 m_freem(m); 3431 return (0); 3432 } 3433 if (ifp != NULL) { 3434 ifp->if_obytes += m->m_pkthdr.len + adjust; 3435 if (m->m_flags & (M_BCAST|M_MCAST)) 3436 ifp->if_omcasts++; 3437 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3438 } 3439 _IF_ENQUEUE(ifq, m); 3440 IF_UNLOCK(ifq); 3441 if (ifp != NULL && !active) 3442 (*(ifp)->if_start)(ifp); 3443 return (1); 3444 } 3445 3446 void 3447 if_register_com_alloc(u_char type, 3448 if_com_alloc_t *a, if_com_free_t *f) 3449 { 3450 3451 KASSERT(if_com_alloc[type] == NULL, 3452 ("if_register_com_alloc: %d already registered", type)); 3453 KASSERT(if_com_free[type] == NULL, 3454 ("if_register_com_alloc: %d free already registered", type)); 3455 3456 if_com_alloc[type] = a; 3457 if_com_free[type] = f; 3458 } 3459 3460 void 3461 if_deregister_com_alloc(u_char type) 3462 { 3463 3464 KASSERT(if_com_alloc[type] != NULL, 3465 ("if_deregister_com_alloc: %d not registered", type)); 3466 KASSERT(if_com_free[type] != NULL, 3467 ("if_deregister_com_alloc: %d free not registered", type)); 3468 if_com_alloc[type] = NULL; 3469 if_com_free[type] = NULL; 3470 } 3471