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 <net/ethernet.h> 78 #include <netinet/in.h> 79 #include <netinet/in_var.h> 80 #include <netinet/ip.h> 81 #include <netinet/ip_carp.h> 82 #ifdef INET 83 #include <netinet/if_ether.h> 84 #endif /* INET */ 85 #ifdef INET6 86 #include <netinet6/in6_var.h> 87 #include <netinet6/in6_ifattach.h> 88 #endif /* INET6 */ 89 #endif /* INET || INET6 */ 90 91 #include <security/mac/mac_framework.h> 92 93 #ifdef COMPAT_FREEBSD32 94 #include <sys/mount.h> 95 #include <compat/freebsd32/freebsd32.h> 96 #endif 97 98 struct ifindex_entry { 99 struct ifnet *ife_ifnet; 100 }; 101 102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 104 105 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen); 106 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 107 &ifqmaxlen, 0, "max send queue size"); 108 109 /* Log link state change events */ 110 static int log_link_state_change = 1; 111 112 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 113 &log_link_state_change, 0, 114 "log interface link state change events"); 115 116 /* Interface description */ 117 static unsigned int ifdescr_maxlen = 1024; 118 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 119 &ifdescr_maxlen, 0, 120 "administrative maximum length for interface description"); 121 122 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 123 124 /* global sx for non-critical path ifdescr */ 125 static struct sx ifdescr_sx; 126 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 127 128 void (*bridge_linkstate_p)(struct ifnet *ifp); 129 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 130 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 131 /* These are external hooks for CARP. */ 132 void (*carp_linkstate_p)(struct ifnet *ifp); 133 void (*carp_demote_adj_p)(int, char *); 134 int (*carp_master_p)(struct ifaddr *); 135 #if defined(INET) || defined(INET6) 136 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 137 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 138 const struct sockaddr *sa); 139 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 140 int (*carp_attach_p)(struct ifaddr *, int); 141 void (*carp_detach_p)(struct ifaddr *); 142 #endif 143 #ifdef INET 144 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 145 #endif 146 #ifdef INET6 147 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 148 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 149 const struct in6_addr *taddr); 150 #endif 151 152 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 153 154 /* 155 * XXX: Style; these should be sorted alphabetically, and unprototyped 156 * static functions should be prototyped. Currently they are sorted by 157 * declaration order. 158 */ 159 static void if_attachdomain(void *); 160 static void if_attachdomain1(struct ifnet *); 161 static int ifconf(u_long, caddr_t); 162 static void if_freemulti(struct ifmultiaddr *); 163 static void if_init(void *); 164 static void if_grow(void); 165 static void if_route(struct ifnet *, int flag, int fam); 166 static int if_setflag(struct ifnet *, int, int, int *, int); 167 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 168 static void if_unroute(struct ifnet *, int flag, int fam); 169 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 170 static int if_rtdel(struct radix_node *, void *); 171 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 172 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 173 static void do_link_state_change(void *, int); 174 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 175 static int if_getgroupmembers(struct ifgroupreq *); 176 static void if_delgroups(struct ifnet *); 177 static void if_attach_internal(struct ifnet *, int); 178 static void if_detach_internal(struct ifnet *, int); 179 180 #ifdef INET6 181 /* 182 * XXX: declare here to avoid to include many inet6 related files.. 183 * should be more generalized? 184 */ 185 extern void nd6_setmtu(struct ifnet *); 186 #endif 187 188 VNET_DEFINE(int, if_index); 189 int ifqmaxlen = IFQ_MAXLEN; 190 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 191 VNET_DEFINE(struct ifgrouphead, ifg_head); 192 193 static VNET_DEFINE(int, if_indexlim) = 8; 194 195 /* Table of ifnet by index. */ 196 VNET_DEFINE(struct ifindex_entry *, ifindex_table); 197 198 #define V_if_indexlim VNET(if_indexlim) 199 #define V_ifindex_table VNET(ifindex_table) 200 201 /* 202 * The global network interface list (V_ifnet) and related state (such as 203 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 204 * an rwlock. Either may be acquired shared to stablize the list, but both 205 * must be acquired writable to modify the list. This model allows us to 206 * both stablize the interface list during interrupt thread processing, but 207 * also to stablize it over long-running ioctls, without introducing priority 208 * inversions and deadlocks. 209 */ 210 struct rwlock ifnet_rwlock; 211 struct sx ifnet_sxlock; 212 213 /* 214 * The allocation of network interfaces is a rather non-atomic affair; we 215 * need to select an index before we are ready to expose the interface for 216 * use, so will use this pointer value to indicate reservation. 217 */ 218 #define IFNET_HOLD (void *)(uintptr_t)(-1) 219 220 static if_com_alloc_t *if_com_alloc[256]; 221 static if_com_free_t *if_com_free[256]; 222 223 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 224 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 225 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 226 227 struct ifnet * 228 ifnet_byindex_locked(u_short idx) 229 { 230 231 if (idx > V_if_index) 232 return (NULL); 233 if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD) 234 return (NULL); 235 return (V_ifindex_table[idx].ife_ifnet); 236 } 237 238 struct ifnet * 239 ifnet_byindex(u_short idx) 240 { 241 struct ifnet *ifp; 242 243 IFNET_RLOCK_NOSLEEP(); 244 ifp = ifnet_byindex_locked(idx); 245 IFNET_RUNLOCK_NOSLEEP(); 246 return (ifp); 247 } 248 249 struct ifnet * 250 ifnet_byindex_ref(u_short idx) 251 { 252 struct ifnet *ifp; 253 254 IFNET_RLOCK_NOSLEEP(); 255 ifp = ifnet_byindex_locked(idx); 256 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 257 IFNET_RUNLOCK_NOSLEEP(); 258 return (NULL); 259 } 260 if_ref(ifp); 261 IFNET_RUNLOCK_NOSLEEP(); 262 return (ifp); 263 } 264 265 /* 266 * Allocate an ifindex array entry; return 0 on success or an error on 267 * failure. 268 */ 269 static int 270 ifindex_alloc_locked(u_short *idxp) 271 { 272 u_short idx; 273 274 IFNET_WLOCK_ASSERT(); 275 276 retry: 277 /* 278 * Try to find an empty slot below V_if_index. If we fail, take the 279 * next slot. 280 */ 281 for (idx = 1; idx <= V_if_index; idx++) { 282 if (V_ifindex_table[idx].ife_ifnet == NULL) 283 break; 284 } 285 286 /* Catch if_index overflow. */ 287 if (idx >= V_if_indexlim) { 288 if_grow(); 289 goto retry; 290 } 291 if (idx > V_if_index) 292 V_if_index = idx; 293 *idxp = idx; 294 return (0); 295 } 296 297 static void 298 ifindex_free_locked(u_short idx) 299 { 300 301 IFNET_WLOCK_ASSERT(); 302 303 V_ifindex_table[idx].ife_ifnet = NULL; 304 while (V_if_index > 0 && 305 V_ifindex_table[V_if_index].ife_ifnet == NULL) 306 V_if_index--; 307 } 308 309 static void 310 ifindex_free(u_short idx) 311 { 312 313 IFNET_WLOCK(); 314 ifindex_free_locked(idx); 315 IFNET_WUNLOCK(); 316 } 317 318 static void 319 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) 320 { 321 322 IFNET_WLOCK_ASSERT(); 323 324 V_ifindex_table[idx].ife_ifnet = ifp; 325 } 326 327 static void 328 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 329 { 330 331 IFNET_WLOCK(); 332 ifnet_setbyindex_locked(idx, ifp); 333 IFNET_WUNLOCK(); 334 } 335 336 struct ifaddr * 337 ifaddr_byindex(u_short idx) 338 { 339 struct ifaddr *ifa; 340 341 IFNET_RLOCK_NOSLEEP(); 342 ifa = ifnet_byindex_locked(idx)->if_addr; 343 if (ifa != NULL) 344 ifa_ref(ifa); 345 IFNET_RUNLOCK_NOSLEEP(); 346 return (ifa); 347 } 348 349 /* 350 * Network interface utility routines. 351 * 352 * Routines with ifa_ifwith* names take sockaddr *'s as 353 * parameters. 354 */ 355 356 static void 357 vnet_if_init(const void *unused __unused) 358 { 359 360 TAILQ_INIT(&V_ifnet); 361 TAILQ_INIT(&V_ifg_head); 362 IFNET_WLOCK(); 363 if_grow(); /* create initial table */ 364 IFNET_WUNLOCK(); 365 vnet_if_clone_init(); 366 } 367 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 368 NULL); 369 370 /* ARGSUSED*/ 371 static void 372 if_init(void *dummy __unused) 373 { 374 375 IFNET_LOCK_INIT(); 376 if_clone_init(); 377 } 378 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL); 379 380 381 #ifdef VIMAGE 382 static void 383 vnet_if_uninit(const void *unused __unused) 384 { 385 386 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 387 "not empty", __func__, __LINE__, &V_ifnet)); 388 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 389 "not empty", __func__, __LINE__, &V_ifg_head)); 390 391 free((caddr_t)V_ifindex_table, M_IFNET); 392 } 393 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 394 vnet_if_uninit, NULL); 395 #endif 396 397 static void 398 if_grow(void) 399 { 400 int oldlim; 401 u_int n; 402 struct ifindex_entry *e; 403 404 IFNET_WLOCK_ASSERT(); 405 oldlim = V_if_indexlim; 406 IFNET_WUNLOCK(); 407 n = (oldlim << 1) * sizeof(*e); 408 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 409 IFNET_WLOCK(); 410 if (V_if_indexlim != oldlim) { 411 free(e, M_IFNET); 412 return; 413 } 414 if (V_ifindex_table != NULL) { 415 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 416 free((caddr_t)V_ifindex_table, M_IFNET); 417 } 418 V_if_indexlim <<= 1; 419 V_ifindex_table = e; 420 } 421 422 /* 423 * Allocate a struct ifnet and an index for an interface. A layer 2 424 * common structure will also be allocated if an allocation routine is 425 * registered for the passed type. 426 */ 427 struct ifnet * 428 if_alloc(u_char type) 429 { 430 struct ifnet *ifp; 431 u_short idx; 432 433 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 434 IFNET_WLOCK(); 435 if (ifindex_alloc_locked(&idx) != 0) { 436 IFNET_WUNLOCK(); 437 free(ifp, M_IFNET); 438 return (NULL); 439 } 440 ifnet_setbyindex_locked(idx, IFNET_HOLD); 441 IFNET_WUNLOCK(); 442 ifp->if_index = idx; 443 ifp->if_type = type; 444 ifp->if_alloctype = type; 445 if (if_com_alloc[type] != NULL) { 446 ifp->if_l2com = if_com_alloc[type](type, ifp); 447 if (ifp->if_l2com == NULL) { 448 free(ifp, M_IFNET); 449 ifindex_free(idx); 450 return (NULL); 451 } 452 } 453 454 IF_ADDR_LOCK_INIT(ifp); 455 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 456 ifp->if_afdata_initialized = 0; 457 IF_AFDATA_LOCK_INIT(ifp); 458 TAILQ_INIT(&ifp->if_addrhead); 459 TAILQ_INIT(&ifp->if_multiaddrs); 460 TAILQ_INIT(&ifp->if_groups); 461 #ifdef MAC 462 mac_ifnet_init(ifp); 463 #endif 464 ifq_init(&ifp->if_snd, ifp); 465 466 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 467 ifnet_setbyindex(ifp->if_index, ifp); 468 return (ifp); 469 } 470 471 /* 472 * Do the actual work of freeing a struct ifnet, and layer 2 common 473 * structure. This call is made when the last reference to an 474 * interface is released. 475 */ 476 static void 477 if_free_internal(struct ifnet *ifp) 478 { 479 480 KASSERT((ifp->if_flags & IFF_DYING), 481 ("if_free_internal: interface not dying")); 482 483 if (if_com_free[ifp->if_alloctype] != NULL) 484 if_com_free[ifp->if_alloctype](ifp->if_l2com, 485 ifp->if_alloctype); 486 487 #ifdef MAC 488 mac_ifnet_destroy(ifp); 489 #endif /* MAC */ 490 if (ifp->if_description != NULL) 491 free(ifp->if_description, M_IFDESCR); 492 IF_AFDATA_DESTROY(ifp); 493 IF_ADDR_LOCK_DESTROY(ifp); 494 ifq_delete(&ifp->if_snd); 495 free(ifp, M_IFNET); 496 } 497 498 /* 499 * Deregister an interface and free the associated storage. 500 */ 501 void 502 if_free(struct ifnet *ifp) 503 { 504 505 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 506 507 CURVNET_SET_QUIET(ifp->if_vnet); 508 IFNET_WLOCK(); 509 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 510 ("%s: freeing unallocated ifnet", ifp->if_xname)); 511 512 ifindex_free_locked(ifp->if_index); 513 IFNET_WUNLOCK(); 514 515 if (refcount_release(&ifp->if_refcount)) 516 if_free_internal(ifp); 517 CURVNET_RESTORE(); 518 } 519 520 /* 521 * Interfaces to keep an ifnet type-stable despite the possibility of the 522 * driver calling if_free(). If there are additional references, we defer 523 * freeing the underlying data structure. 524 */ 525 void 526 if_ref(struct ifnet *ifp) 527 { 528 529 /* We don't assert the ifnet list lock here, but arguably should. */ 530 refcount_acquire(&ifp->if_refcount); 531 } 532 533 void 534 if_rele(struct ifnet *ifp) 535 { 536 537 if (!refcount_release(&ifp->if_refcount)) 538 return; 539 if_free_internal(ifp); 540 } 541 542 void 543 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 544 { 545 546 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 547 548 if (ifq->ifq_maxlen == 0) 549 ifq->ifq_maxlen = ifqmaxlen; 550 551 ifq->altq_type = 0; 552 ifq->altq_disc = NULL; 553 ifq->altq_flags &= ALTQF_CANTCHANGE; 554 ifq->altq_tbr = NULL; 555 ifq->altq_ifp = ifp; 556 } 557 558 void 559 ifq_delete(struct ifaltq *ifq) 560 { 561 mtx_destroy(&ifq->ifq_mtx); 562 } 563 564 /* 565 * Perform generic interface initalization tasks and attach the interface 566 * to the list of "active" interfaces. If vmove flag is set on entry 567 * to if_attach_internal(), perform only a limited subset of initialization 568 * tasks, given that we are moving from one vnet to another an ifnet which 569 * has already been fully initialized. 570 * 571 * XXX: 572 * - The decision to return void and thus require this function to 573 * succeed is questionable. 574 * - We should probably do more sanity checking. For instance we don't 575 * do anything to insure if_xname is unique or non-empty. 576 */ 577 void 578 if_attach(struct ifnet *ifp) 579 { 580 581 if_attach_internal(ifp, 0); 582 } 583 584 static void 585 if_attach_internal(struct ifnet *ifp, int vmove) 586 { 587 unsigned socksize, ifasize; 588 int namelen, masklen; 589 struct sockaddr_dl *sdl; 590 struct ifaddr *ifa; 591 592 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 593 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 594 ifp->if_xname); 595 596 #ifdef VIMAGE 597 ifp->if_vnet = curvnet; 598 if (ifp->if_home_vnet == NULL) 599 ifp->if_home_vnet = curvnet; 600 #endif 601 602 if_addgroup(ifp, IFG_ALL); 603 604 getmicrotime(&ifp->if_lastchange); 605 ifp->if_data.ifi_epoch = time_uptime; 606 ifp->if_data.ifi_datalen = sizeof(struct if_data); 607 608 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 609 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 610 ("transmit and qflush must both either be set or both be NULL")); 611 if (ifp->if_transmit == NULL) { 612 ifp->if_transmit = if_transmit; 613 ifp->if_qflush = if_qflush; 614 } 615 616 if (!vmove) { 617 #ifdef MAC 618 mac_ifnet_create(ifp); 619 #endif 620 621 /* 622 * Create a Link Level name for this device. 623 */ 624 namelen = strlen(ifp->if_xname); 625 /* 626 * Always save enough space for any possiable name so we 627 * can do a rename in place later. 628 */ 629 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 630 socksize = masklen + ifp->if_addrlen; 631 if (socksize < sizeof(*sdl)) 632 socksize = sizeof(*sdl); 633 socksize = roundup2(socksize, sizeof(long)); 634 ifasize = sizeof(*ifa) + 2 * socksize; 635 ifa = ifa_alloc(ifasize, M_WAITOK); 636 sdl = (struct sockaddr_dl *)(ifa + 1); 637 sdl->sdl_len = socksize; 638 sdl->sdl_family = AF_LINK; 639 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 640 sdl->sdl_nlen = namelen; 641 sdl->sdl_index = ifp->if_index; 642 sdl->sdl_type = ifp->if_type; 643 ifp->if_addr = ifa; 644 ifa->ifa_ifp = ifp; 645 ifa->ifa_rtrequest = link_rtrequest; 646 ifa->ifa_addr = (struct sockaddr *)sdl; 647 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 648 ifa->ifa_netmask = (struct sockaddr *)sdl; 649 sdl->sdl_len = masklen; 650 while (namelen != 0) 651 sdl->sdl_data[--namelen] = 0xff; 652 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 653 /* Reliably crash if used uninitialized. */ 654 ifp->if_broadcastaddr = NULL; 655 656 #if defined(INET) || defined(INET6) 657 /* Initialize to max value. */ 658 if (ifp->if_hw_tsomax == 0) 659 ifp->if_hw_tsomax = min(IP_MAXPACKET, 32 * MCLBYTES - 660 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 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 int 1529 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa) 1530 { 1531 struct rtentry *rt; 1532 1533 rt = rtalloc1_fib(sa, 0, 0, 0); 1534 if (rt == NULL) { 1535 log(LOG_DEBUG, "%s: fail", __func__); 1536 return (EHOSTUNREACH); 1537 } 1538 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = 1539 ifa->ifa_ifp->if_type; 1540 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = 1541 ifa->ifa_ifp->if_index; 1542 RTFREE_LOCKED(rt); 1543 1544 return (0); 1545 } 1546 1547 /* 1548 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1549 * structs used to represent other address families, it is necessary 1550 * to perform a different comparison. 1551 */ 1552 1553 #define sa_equal(a1, a2) \ 1554 (bcmp((a1), (a2), ((a1))->sa_len) == 0) 1555 1556 #define sa_dl_equal(a1, a2) \ 1557 ((((struct sockaddr_dl *)(a1))->sdl_len == \ 1558 ((struct sockaddr_dl *)(a2))->sdl_len) && \ 1559 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \ 1560 LLADDR((struct sockaddr_dl *)(a2)), \ 1561 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1562 1563 /* 1564 * Locate an interface based on a complete address. 1565 */ 1566 /*ARGSUSED*/ 1567 static struct ifaddr * 1568 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref) 1569 { 1570 struct ifnet *ifp; 1571 struct ifaddr *ifa; 1572 1573 IFNET_RLOCK_NOSLEEP(); 1574 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1575 IF_ADDR_RLOCK(ifp); 1576 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1577 if (ifa->ifa_addr->sa_family != addr->sa_family) 1578 continue; 1579 if (sa_equal(addr, ifa->ifa_addr)) { 1580 if (getref) 1581 ifa_ref(ifa); 1582 IF_ADDR_RUNLOCK(ifp); 1583 goto done; 1584 } 1585 /* IP6 doesn't have broadcast */ 1586 if ((ifp->if_flags & IFF_BROADCAST) && 1587 ifa->ifa_broadaddr && 1588 ifa->ifa_broadaddr->sa_len != 0 && 1589 sa_equal(ifa->ifa_broadaddr, addr)) { 1590 if (getref) 1591 ifa_ref(ifa); 1592 IF_ADDR_RUNLOCK(ifp); 1593 goto done; 1594 } 1595 } 1596 IF_ADDR_RUNLOCK(ifp); 1597 } 1598 ifa = NULL; 1599 done: 1600 IFNET_RUNLOCK_NOSLEEP(); 1601 return (ifa); 1602 } 1603 1604 struct ifaddr * 1605 ifa_ifwithaddr(struct sockaddr *addr) 1606 { 1607 1608 return (ifa_ifwithaddr_internal(addr, 1)); 1609 } 1610 1611 int 1612 ifa_ifwithaddr_check(struct sockaddr *addr) 1613 { 1614 1615 return (ifa_ifwithaddr_internal(addr, 0) != NULL); 1616 } 1617 1618 /* 1619 * Locate an interface based on the broadcast address. 1620 */ 1621 /* ARGSUSED */ 1622 struct ifaddr * 1623 ifa_ifwithbroadaddr(struct sockaddr *addr) 1624 { 1625 struct ifnet *ifp; 1626 struct ifaddr *ifa; 1627 1628 IFNET_RLOCK_NOSLEEP(); 1629 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1630 IF_ADDR_RLOCK(ifp); 1631 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1632 if (ifa->ifa_addr->sa_family != addr->sa_family) 1633 continue; 1634 if ((ifp->if_flags & IFF_BROADCAST) && 1635 ifa->ifa_broadaddr && 1636 ifa->ifa_broadaddr->sa_len != 0 && 1637 sa_equal(ifa->ifa_broadaddr, addr)) { 1638 ifa_ref(ifa); 1639 IF_ADDR_RUNLOCK(ifp); 1640 goto done; 1641 } 1642 } 1643 IF_ADDR_RUNLOCK(ifp); 1644 } 1645 ifa = NULL; 1646 done: 1647 IFNET_RUNLOCK_NOSLEEP(); 1648 return (ifa); 1649 } 1650 1651 /* 1652 * Locate the point to point interface with a given destination address. 1653 */ 1654 /*ARGSUSED*/ 1655 struct ifaddr * 1656 ifa_ifwithdstaddr(struct sockaddr *addr) 1657 { 1658 struct ifnet *ifp; 1659 struct ifaddr *ifa; 1660 1661 IFNET_RLOCK_NOSLEEP(); 1662 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1663 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1664 continue; 1665 IF_ADDR_RLOCK(ifp); 1666 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1667 if (ifa->ifa_addr->sa_family != addr->sa_family) 1668 continue; 1669 if (ifa->ifa_dstaddr != NULL && 1670 sa_equal(addr, ifa->ifa_dstaddr)) { 1671 ifa_ref(ifa); 1672 IF_ADDR_RUNLOCK(ifp); 1673 goto done; 1674 } 1675 } 1676 IF_ADDR_RUNLOCK(ifp); 1677 } 1678 ifa = NULL; 1679 done: 1680 IFNET_RUNLOCK_NOSLEEP(); 1681 return (ifa); 1682 } 1683 1684 /* 1685 * Find an interface on a specific network. If many, choice 1686 * is most specific found. 1687 */ 1688 struct ifaddr * 1689 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp) 1690 { 1691 struct ifnet *ifp; 1692 struct ifaddr *ifa; 1693 struct ifaddr *ifa_maybe = NULL; 1694 u_int af = addr->sa_family; 1695 char *addr_data = addr->sa_data, *cplim; 1696 1697 /* 1698 * AF_LINK addresses can be looked up directly by their index number, 1699 * so do that if we can. 1700 */ 1701 if (af == AF_LINK) { 1702 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1703 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1704 return (ifaddr_byindex(sdl->sdl_index)); 1705 } 1706 1707 /* 1708 * Scan though each interface, looking for ones that have addresses 1709 * in this address family. Maintain a reference on ifa_maybe once 1710 * we find one, as we release the IF_ADDR_RLOCK() that kept it stable 1711 * when we move onto the next interface. 1712 */ 1713 IFNET_RLOCK_NOSLEEP(); 1714 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1715 IF_ADDR_RLOCK(ifp); 1716 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1717 char *cp, *cp2, *cp3; 1718 1719 if (ifa->ifa_addr->sa_family != af) 1720 next: continue; 1721 if (af == AF_INET && 1722 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1723 /* 1724 * This is a bit broken as it doesn't 1725 * take into account that the remote end may 1726 * be a single node in the network we are 1727 * looking for. 1728 * The trouble is that we don't know the 1729 * netmask for the remote end. 1730 */ 1731 if (ifa->ifa_dstaddr != NULL && 1732 sa_equal(addr, ifa->ifa_dstaddr)) { 1733 ifa_ref(ifa); 1734 IF_ADDR_RUNLOCK(ifp); 1735 goto done; 1736 } 1737 } else { 1738 /* 1739 * if we have a special address handler, 1740 * then use it instead of the generic one. 1741 */ 1742 if (ifa->ifa_claim_addr) { 1743 if ((*ifa->ifa_claim_addr)(ifa, addr)) { 1744 ifa_ref(ifa); 1745 IF_ADDR_RUNLOCK(ifp); 1746 goto done; 1747 } 1748 continue; 1749 } 1750 1751 /* 1752 * Scan all the bits in the ifa's address. 1753 * If a bit dissagrees with what we are 1754 * looking for, mask it with the netmask 1755 * to see if it really matters. 1756 * (A byte at a time) 1757 */ 1758 if (ifa->ifa_netmask == 0) 1759 continue; 1760 cp = addr_data; 1761 cp2 = ifa->ifa_addr->sa_data; 1762 cp3 = ifa->ifa_netmask->sa_data; 1763 cplim = ifa->ifa_netmask->sa_len 1764 + (char *)ifa->ifa_netmask; 1765 while (cp3 < cplim) 1766 if ((*cp++ ^ *cp2++) & *cp3++) 1767 goto next; /* next address! */ 1768 /* 1769 * If the netmask of what we just found 1770 * is more specific than what we had before 1771 * (if we had one), or if the virtual status 1772 * of new prefix is better than of the old one, 1773 * then remember the new one before continuing 1774 * to search for an even better one. 1775 */ 1776 if (ifa_maybe == NULL || 1777 ifa_preferred(ifa_maybe, ifa) || 1778 rn_refines((caddr_t)ifa->ifa_netmask, 1779 (caddr_t)ifa_maybe->ifa_netmask)) { 1780 if (ifa_maybe != NULL) 1781 ifa_free(ifa_maybe); 1782 ifa_maybe = ifa; 1783 ifa_ref(ifa_maybe); 1784 } 1785 } 1786 } 1787 IF_ADDR_RUNLOCK(ifp); 1788 } 1789 ifa = ifa_maybe; 1790 ifa_maybe = NULL; 1791 done: 1792 IFNET_RUNLOCK_NOSLEEP(); 1793 if (ifa_maybe != NULL) 1794 ifa_free(ifa_maybe); 1795 return (ifa); 1796 } 1797 1798 /* 1799 * Find an interface address specific to an interface best matching 1800 * a given address. 1801 */ 1802 struct ifaddr * 1803 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1804 { 1805 struct ifaddr *ifa; 1806 char *cp, *cp2, *cp3; 1807 char *cplim; 1808 struct ifaddr *ifa_maybe = NULL; 1809 u_int af = addr->sa_family; 1810 1811 if (af >= AF_MAX) 1812 return (NULL); 1813 IF_ADDR_RLOCK(ifp); 1814 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1815 if (ifa->ifa_addr->sa_family != af) 1816 continue; 1817 if (ifa_maybe == NULL) 1818 ifa_maybe = ifa; 1819 if (ifa->ifa_netmask == 0) { 1820 if (sa_equal(addr, ifa->ifa_addr) || 1821 (ifa->ifa_dstaddr && 1822 sa_equal(addr, ifa->ifa_dstaddr))) 1823 goto done; 1824 continue; 1825 } 1826 if (ifp->if_flags & IFF_POINTOPOINT) { 1827 if (sa_equal(addr, ifa->ifa_dstaddr)) 1828 goto done; 1829 } else { 1830 cp = addr->sa_data; 1831 cp2 = ifa->ifa_addr->sa_data; 1832 cp3 = ifa->ifa_netmask->sa_data; 1833 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1834 for (; cp3 < cplim; cp3++) 1835 if ((*cp++ ^ *cp2++) & *cp3) 1836 break; 1837 if (cp3 == cplim) 1838 goto done; 1839 } 1840 } 1841 ifa = ifa_maybe; 1842 done: 1843 if (ifa != NULL) 1844 ifa_ref(ifa); 1845 IF_ADDR_RUNLOCK(ifp); 1846 return (ifa); 1847 } 1848 1849 /* 1850 * See whether new ifa is better than current one: 1851 * 1) A non-virtual one is preferred over virtual. 1852 * 2) A virtual in master state preferred over any other state. 1853 * 1854 * Used in several address selecting functions. 1855 */ 1856 int 1857 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 1858 { 1859 1860 return (cur->ifa_carp && (!next->ifa_carp || 1861 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 1862 } 1863 1864 #include <net/if_llatbl.h> 1865 1866 /* 1867 * Default action when installing a route with a Link Level gateway. 1868 * Lookup an appropriate real ifa to point to. 1869 * This should be moved to /sys/net/link.c eventually. 1870 */ 1871 static void 1872 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 1873 { 1874 struct ifaddr *ifa, *oifa; 1875 struct sockaddr *dst; 1876 struct ifnet *ifp; 1877 1878 RT_LOCK_ASSERT(rt); 1879 1880 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 1881 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 1882 return; 1883 ifa = ifaof_ifpforaddr(dst, ifp); 1884 if (ifa) { 1885 oifa = rt->rt_ifa; 1886 rt->rt_ifa = ifa; 1887 ifa_free(oifa); 1888 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1889 ifa->ifa_rtrequest(cmd, rt, info); 1890 } 1891 } 1892 1893 struct sockaddr_dl * 1894 link_alloc_sdl(size_t size, int flags) 1895 { 1896 1897 return (malloc(size, M_TEMP, flags)); 1898 } 1899 1900 void 1901 link_free_sdl(struct sockaddr *sa) 1902 { 1903 free(sa, M_TEMP); 1904 } 1905 1906 /* 1907 * Fills in given sdl with interface basic info. 1908 * Returns pointer to filled sdl. 1909 */ 1910 struct sockaddr_dl * 1911 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 1912 { 1913 struct sockaddr_dl *sdl; 1914 1915 sdl = (struct sockaddr_dl *)paddr; 1916 memset(sdl, 0, sizeof(struct sockaddr_dl)); 1917 sdl->sdl_len = sizeof(struct sockaddr_dl); 1918 sdl->sdl_family = AF_LINK; 1919 sdl->sdl_index = ifp->if_index; 1920 sdl->sdl_type = iftype; 1921 1922 return (sdl); 1923 } 1924 1925 /* 1926 * Mark an interface down and notify protocols of 1927 * the transition. 1928 */ 1929 static void 1930 if_unroute(struct ifnet *ifp, int flag, int fam) 1931 { 1932 struct ifaddr *ifa; 1933 1934 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 1935 1936 ifp->if_flags &= ~flag; 1937 getmicrotime(&ifp->if_lastchange); 1938 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1939 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1940 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1941 ifp->if_qflush(ifp); 1942 1943 if (ifp->if_carp) 1944 (*carp_linkstate_p)(ifp); 1945 rt_ifmsg(ifp); 1946 } 1947 1948 /* 1949 * Mark an interface up and notify protocols of 1950 * the transition. 1951 */ 1952 static void 1953 if_route(struct ifnet *ifp, int flag, int fam) 1954 { 1955 struct ifaddr *ifa; 1956 1957 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 1958 1959 ifp->if_flags |= flag; 1960 getmicrotime(&ifp->if_lastchange); 1961 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1962 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1963 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1964 if (ifp->if_carp) 1965 (*carp_linkstate_p)(ifp); 1966 rt_ifmsg(ifp); 1967 #ifdef INET6 1968 in6_if_up(ifp); 1969 #endif 1970 } 1971 1972 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 1973 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1974 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 1975 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 1976 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 1977 int (*vlan_setcookie_p)(struct ifnet *, void *); 1978 void *(*vlan_cookie_p)(struct ifnet *); 1979 1980 /* 1981 * Handle a change in the interface link state. To avoid LORs 1982 * between driver lock and upper layer locks, as well as possible 1983 * recursions, we post event to taskqueue, and all job 1984 * is done in static do_link_state_change(). 1985 */ 1986 void 1987 if_link_state_change(struct ifnet *ifp, int link_state) 1988 { 1989 /* Return if state hasn't changed. */ 1990 if (ifp->if_link_state == link_state) 1991 return; 1992 1993 ifp->if_link_state = link_state; 1994 1995 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 1996 } 1997 1998 static void 1999 do_link_state_change(void *arg, int pending) 2000 { 2001 struct ifnet *ifp = (struct ifnet *)arg; 2002 int link_state = ifp->if_link_state; 2003 CURVNET_SET(ifp->if_vnet); 2004 2005 /* Notify that the link state has changed. */ 2006 rt_ifmsg(ifp); 2007 if (ifp->if_vlantrunk != NULL) 2008 (*vlan_link_state_p)(ifp); 2009 2010 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2011 IFP2AC(ifp)->ac_netgraph != NULL) 2012 (*ng_ether_link_state_p)(ifp, link_state); 2013 if (ifp->if_carp) 2014 (*carp_linkstate_p)(ifp); 2015 if (ifp->if_bridge) 2016 (*bridge_linkstate_p)(ifp); 2017 if (ifp->if_lagg) 2018 (*lagg_linkstate_p)(ifp, link_state); 2019 2020 if (IS_DEFAULT_VNET(curvnet)) 2021 devctl_notify("IFNET", ifp->if_xname, 2022 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2023 NULL); 2024 if (pending > 1) 2025 if_printf(ifp, "%d link states coalesced\n", pending); 2026 if (log_link_state_change) 2027 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 2028 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2029 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state); 2030 CURVNET_RESTORE(); 2031 } 2032 2033 /* 2034 * Mark an interface down and notify protocols of 2035 * the transition. 2036 */ 2037 void 2038 if_down(struct ifnet *ifp) 2039 { 2040 2041 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2042 } 2043 2044 /* 2045 * Mark an interface up and notify protocols of 2046 * the transition. 2047 */ 2048 void 2049 if_up(struct ifnet *ifp) 2050 { 2051 2052 if_route(ifp, IFF_UP, AF_UNSPEC); 2053 } 2054 2055 /* 2056 * Flush an interface queue. 2057 */ 2058 void 2059 if_qflush(struct ifnet *ifp) 2060 { 2061 struct mbuf *m, *n; 2062 struct ifaltq *ifq; 2063 2064 ifq = &ifp->if_snd; 2065 IFQ_LOCK(ifq); 2066 #ifdef ALTQ 2067 if (ALTQ_IS_ENABLED(ifq)) 2068 ALTQ_PURGE(ifq); 2069 #endif 2070 n = ifq->ifq_head; 2071 while ((m = n) != 0) { 2072 n = m->m_act; 2073 m_freem(m); 2074 } 2075 ifq->ifq_head = 0; 2076 ifq->ifq_tail = 0; 2077 ifq->ifq_len = 0; 2078 IFQ_UNLOCK(ifq); 2079 } 2080 2081 /* 2082 * Map interface name to interface structure pointer, with or without 2083 * returning a reference. 2084 */ 2085 struct ifnet * 2086 ifunit_ref(const char *name) 2087 { 2088 struct ifnet *ifp; 2089 2090 IFNET_RLOCK_NOSLEEP(); 2091 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2092 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2093 !(ifp->if_flags & IFF_DYING)) 2094 break; 2095 } 2096 if (ifp != NULL) 2097 if_ref(ifp); 2098 IFNET_RUNLOCK_NOSLEEP(); 2099 return (ifp); 2100 } 2101 2102 struct ifnet * 2103 ifunit(const char *name) 2104 { 2105 struct ifnet *ifp; 2106 2107 IFNET_RLOCK_NOSLEEP(); 2108 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2109 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2110 break; 2111 } 2112 IFNET_RUNLOCK_NOSLEEP(); 2113 return (ifp); 2114 } 2115 2116 /* 2117 * Hardware specific interface ioctls. 2118 */ 2119 static int 2120 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2121 { 2122 struct ifreq *ifr; 2123 int error = 0; 2124 int new_flags, temp_flags; 2125 size_t namelen, onamelen; 2126 size_t descrlen; 2127 char *descrbuf, *odescrbuf; 2128 char new_name[IFNAMSIZ]; 2129 struct ifaddr *ifa; 2130 struct sockaddr_dl *sdl; 2131 2132 ifr = (struct ifreq *)data; 2133 switch (cmd) { 2134 case SIOCGIFINDEX: 2135 ifr->ifr_index = ifp->if_index; 2136 break; 2137 2138 case SIOCGIFFLAGS: 2139 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2140 ifr->ifr_flags = temp_flags & 0xffff; 2141 ifr->ifr_flagshigh = temp_flags >> 16; 2142 break; 2143 2144 case SIOCGIFCAP: 2145 ifr->ifr_reqcap = ifp->if_capabilities; 2146 ifr->ifr_curcap = ifp->if_capenable; 2147 break; 2148 2149 #ifdef MAC 2150 case SIOCGIFMAC: 2151 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2152 break; 2153 #endif 2154 2155 case SIOCGIFMETRIC: 2156 ifr->ifr_metric = ifp->if_metric; 2157 break; 2158 2159 case SIOCGIFMTU: 2160 ifr->ifr_mtu = ifp->if_mtu; 2161 break; 2162 2163 case SIOCGIFPHYS: 2164 ifr->ifr_phys = ifp->if_physical; 2165 break; 2166 2167 case SIOCGIFDESCR: 2168 error = 0; 2169 sx_slock(&ifdescr_sx); 2170 if (ifp->if_description == NULL) 2171 error = ENOMSG; 2172 else { 2173 /* space for terminating nul */ 2174 descrlen = strlen(ifp->if_description) + 1; 2175 if (ifr->ifr_buffer.length < descrlen) 2176 ifr->ifr_buffer.buffer = NULL; 2177 else 2178 error = copyout(ifp->if_description, 2179 ifr->ifr_buffer.buffer, descrlen); 2180 ifr->ifr_buffer.length = descrlen; 2181 } 2182 sx_sunlock(&ifdescr_sx); 2183 break; 2184 2185 case SIOCSIFDESCR: 2186 error = priv_check(td, PRIV_NET_SETIFDESCR); 2187 if (error) 2188 return (error); 2189 2190 /* 2191 * Copy only (length-1) bytes to make sure that 2192 * if_description is always nul terminated. The 2193 * length parameter is supposed to count the 2194 * terminating nul in. 2195 */ 2196 if (ifr->ifr_buffer.length > ifdescr_maxlen) 2197 return (ENAMETOOLONG); 2198 else if (ifr->ifr_buffer.length == 0) 2199 descrbuf = NULL; 2200 else { 2201 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR, 2202 M_WAITOK | M_ZERO); 2203 error = copyin(ifr->ifr_buffer.buffer, descrbuf, 2204 ifr->ifr_buffer.length - 1); 2205 if (error) { 2206 free(descrbuf, M_IFDESCR); 2207 break; 2208 } 2209 } 2210 2211 sx_xlock(&ifdescr_sx); 2212 odescrbuf = ifp->if_description; 2213 ifp->if_description = descrbuf; 2214 sx_xunlock(&ifdescr_sx); 2215 2216 getmicrotime(&ifp->if_lastchange); 2217 free(odescrbuf, M_IFDESCR); 2218 break; 2219 2220 case SIOCGIFFIB: 2221 ifr->ifr_fib = ifp->if_fib; 2222 break; 2223 2224 case SIOCSIFFIB: 2225 error = priv_check(td, PRIV_NET_SETIFFIB); 2226 if (error) 2227 return (error); 2228 if (ifr->ifr_fib >= rt_numfibs) 2229 return (EINVAL); 2230 2231 ifp->if_fib = ifr->ifr_fib; 2232 break; 2233 2234 case SIOCSIFFLAGS: 2235 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2236 if (error) 2237 return (error); 2238 /* 2239 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2240 * check, so we don't need special handling here yet. 2241 */ 2242 new_flags = (ifr->ifr_flags & 0xffff) | 2243 (ifr->ifr_flagshigh << 16); 2244 if (ifp->if_flags & IFF_UP && 2245 (new_flags & IFF_UP) == 0) { 2246 if_down(ifp); 2247 } else if (new_flags & IFF_UP && 2248 (ifp->if_flags & IFF_UP) == 0) { 2249 if_up(ifp); 2250 } 2251 /* See if permanently promiscuous mode bit is about to flip */ 2252 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2253 if (new_flags & IFF_PPROMISC) 2254 ifp->if_flags |= IFF_PROMISC; 2255 else if (ifp->if_pcount == 0) 2256 ifp->if_flags &= ~IFF_PROMISC; 2257 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2258 ifp->if_xname, 2259 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); 2260 } 2261 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2262 (new_flags &~ IFF_CANTCHANGE); 2263 if (ifp->if_ioctl) { 2264 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2265 } 2266 getmicrotime(&ifp->if_lastchange); 2267 break; 2268 2269 case SIOCSIFCAP: 2270 error = priv_check(td, PRIV_NET_SETIFCAP); 2271 if (error) 2272 return (error); 2273 if (ifp->if_ioctl == NULL) 2274 return (EOPNOTSUPP); 2275 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2276 return (EINVAL); 2277 error = (*ifp->if_ioctl)(ifp, cmd, data); 2278 if (error == 0) 2279 getmicrotime(&ifp->if_lastchange); 2280 break; 2281 2282 #ifdef MAC 2283 case SIOCSIFMAC: 2284 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2285 break; 2286 #endif 2287 2288 case SIOCSIFNAME: 2289 error = priv_check(td, PRIV_NET_SETIFNAME); 2290 if (error) 2291 return (error); 2292 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 2293 if (error != 0) 2294 return (error); 2295 if (new_name[0] == '\0') 2296 return (EINVAL); 2297 if (ifunit(new_name) != NULL) 2298 return (EEXIST); 2299 2300 /* 2301 * XXX: Locking. Nothing else seems to lock if_flags, 2302 * and there are numerous other races with the 2303 * ifunit() checks not being atomic with namespace 2304 * changes (renames, vmoves, if_attach, etc). 2305 */ 2306 ifp->if_flags |= IFF_RENAMING; 2307 2308 /* Announce the departure of the interface. */ 2309 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2310 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2311 2312 log(LOG_INFO, "%s: changing name to '%s'\n", 2313 ifp->if_xname, new_name); 2314 2315 IF_ADDR_WLOCK(ifp); 2316 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2317 ifa = ifp->if_addr; 2318 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2319 namelen = strlen(new_name); 2320 onamelen = sdl->sdl_nlen; 2321 /* 2322 * Move the address if needed. This is safe because we 2323 * allocate space for a name of length IFNAMSIZ when we 2324 * create this in if_attach(). 2325 */ 2326 if (namelen != onamelen) { 2327 bcopy(sdl->sdl_data + onamelen, 2328 sdl->sdl_data + namelen, sdl->sdl_alen); 2329 } 2330 bcopy(new_name, sdl->sdl_data, namelen); 2331 sdl->sdl_nlen = namelen; 2332 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2333 bzero(sdl->sdl_data, onamelen); 2334 while (namelen != 0) 2335 sdl->sdl_data[--namelen] = 0xff; 2336 IF_ADDR_WUNLOCK(ifp); 2337 2338 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2339 /* Announce the return of the interface. */ 2340 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2341 2342 ifp->if_flags &= ~IFF_RENAMING; 2343 break; 2344 2345 #ifdef VIMAGE 2346 case SIOCSIFVNET: 2347 error = priv_check(td, PRIV_NET_SETIFVNET); 2348 if (error) 2349 return (error); 2350 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2351 break; 2352 #endif 2353 2354 case SIOCSIFMETRIC: 2355 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2356 if (error) 2357 return (error); 2358 ifp->if_metric = ifr->ifr_metric; 2359 getmicrotime(&ifp->if_lastchange); 2360 break; 2361 2362 case SIOCSIFPHYS: 2363 error = priv_check(td, PRIV_NET_SETIFPHYS); 2364 if (error) 2365 return (error); 2366 if (ifp->if_ioctl == NULL) 2367 return (EOPNOTSUPP); 2368 error = (*ifp->if_ioctl)(ifp, cmd, data); 2369 if (error == 0) 2370 getmicrotime(&ifp->if_lastchange); 2371 break; 2372 2373 case SIOCSIFMTU: 2374 { 2375 u_long oldmtu = ifp->if_mtu; 2376 2377 error = priv_check(td, PRIV_NET_SETIFMTU); 2378 if (error) 2379 return (error); 2380 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2381 return (EINVAL); 2382 if (ifp->if_ioctl == NULL) 2383 return (EOPNOTSUPP); 2384 error = (*ifp->if_ioctl)(ifp, cmd, data); 2385 if (error == 0) { 2386 getmicrotime(&ifp->if_lastchange); 2387 rt_ifmsg(ifp); 2388 } 2389 /* 2390 * If the link MTU changed, do network layer specific procedure. 2391 */ 2392 if (ifp->if_mtu != oldmtu) { 2393 #ifdef INET6 2394 nd6_setmtu(ifp); 2395 #endif 2396 } 2397 break; 2398 } 2399 2400 case SIOCADDMULTI: 2401 case SIOCDELMULTI: 2402 if (cmd == SIOCADDMULTI) 2403 error = priv_check(td, PRIV_NET_ADDMULTI); 2404 else 2405 error = priv_check(td, PRIV_NET_DELMULTI); 2406 if (error) 2407 return (error); 2408 2409 /* Don't allow group membership on non-multicast interfaces. */ 2410 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2411 return (EOPNOTSUPP); 2412 2413 /* Don't let users screw up protocols' entries. */ 2414 if (ifr->ifr_addr.sa_family != AF_LINK) 2415 return (EINVAL); 2416 2417 if (cmd == SIOCADDMULTI) { 2418 struct ifmultiaddr *ifma; 2419 2420 /* 2421 * Userland is only permitted to join groups once 2422 * via the if_addmulti() KPI, because it cannot hold 2423 * struct ifmultiaddr * between calls. It may also 2424 * lose a race while we check if the membership 2425 * already exists. 2426 */ 2427 IF_ADDR_RLOCK(ifp); 2428 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2429 IF_ADDR_RUNLOCK(ifp); 2430 if (ifma != NULL) 2431 error = EADDRINUSE; 2432 else 2433 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2434 } else { 2435 error = if_delmulti(ifp, &ifr->ifr_addr); 2436 } 2437 if (error == 0) 2438 getmicrotime(&ifp->if_lastchange); 2439 break; 2440 2441 case SIOCSIFPHYADDR: 2442 case SIOCDIFPHYADDR: 2443 #ifdef INET6 2444 case SIOCSIFPHYADDR_IN6: 2445 #endif 2446 case SIOCSIFMEDIA: 2447 case SIOCSIFGENERIC: 2448 error = priv_check(td, PRIV_NET_HWIOCTL); 2449 if (error) 2450 return (error); 2451 if (ifp->if_ioctl == NULL) 2452 return (EOPNOTSUPP); 2453 error = (*ifp->if_ioctl)(ifp, cmd, data); 2454 if (error == 0) 2455 getmicrotime(&ifp->if_lastchange); 2456 break; 2457 2458 case SIOCGIFSTATUS: 2459 case SIOCGIFPSRCADDR: 2460 case SIOCGIFPDSTADDR: 2461 case SIOCGIFMEDIA: 2462 case SIOCGIFGENERIC: 2463 if (ifp->if_ioctl == NULL) 2464 return (EOPNOTSUPP); 2465 error = (*ifp->if_ioctl)(ifp, cmd, data); 2466 break; 2467 2468 case SIOCSIFLLADDR: 2469 error = priv_check(td, PRIV_NET_SETLLADDR); 2470 if (error) 2471 return (error); 2472 error = if_setlladdr(ifp, 2473 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2474 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 2475 break; 2476 2477 case SIOCAIFGROUP: 2478 { 2479 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2480 2481 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2482 if (error) 2483 return (error); 2484 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2485 return (error); 2486 break; 2487 } 2488 2489 case SIOCGIFGROUP: 2490 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2491 return (error); 2492 break; 2493 2494 case SIOCDIFGROUP: 2495 { 2496 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2497 2498 error = priv_check(td, PRIV_NET_DELIFGROUP); 2499 if (error) 2500 return (error); 2501 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2502 return (error); 2503 break; 2504 } 2505 2506 default: 2507 error = ENOIOCTL; 2508 break; 2509 } 2510 return (error); 2511 } 2512 2513 #ifdef COMPAT_FREEBSD32 2514 struct ifconf32 { 2515 int32_t ifc_len; 2516 union { 2517 uint32_t ifcu_buf; 2518 uint32_t ifcu_req; 2519 } ifc_ifcu; 2520 }; 2521 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2522 #endif 2523 2524 /* 2525 * Interface ioctls. 2526 */ 2527 int 2528 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2529 { 2530 struct ifnet *ifp; 2531 struct ifreq *ifr; 2532 int error; 2533 int oif_flags; 2534 2535 CURVNET_SET(so->so_vnet); 2536 switch (cmd) { 2537 case SIOCGIFCONF: 2538 error = ifconf(cmd, data); 2539 CURVNET_RESTORE(); 2540 return (error); 2541 2542 #ifdef COMPAT_FREEBSD32 2543 case SIOCGIFCONF32: 2544 { 2545 struct ifconf32 *ifc32; 2546 struct ifconf ifc; 2547 2548 ifc32 = (struct ifconf32 *)data; 2549 ifc.ifc_len = ifc32->ifc_len; 2550 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2551 2552 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2553 CURVNET_RESTORE(); 2554 if (error == 0) 2555 ifc32->ifc_len = ifc.ifc_len; 2556 return (error); 2557 } 2558 #endif 2559 } 2560 ifr = (struct ifreq *)data; 2561 2562 switch (cmd) { 2563 #ifdef VIMAGE 2564 case SIOCSIFRVNET: 2565 error = priv_check(td, PRIV_NET_SETIFVNET); 2566 if (error == 0) 2567 error = if_vmove_reclaim(td, ifr->ifr_name, 2568 ifr->ifr_jid); 2569 CURVNET_RESTORE(); 2570 return (error); 2571 #endif 2572 case SIOCIFCREATE: 2573 case SIOCIFCREATE2: 2574 error = priv_check(td, PRIV_NET_IFCREATE); 2575 if (error == 0) 2576 error = if_clone_create(ifr->ifr_name, 2577 sizeof(ifr->ifr_name), 2578 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL); 2579 CURVNET_RESTORE(); 2580 return (error); 2581 case SIOCIFDESTROY: 2582 error = priv_check(td, PRIV_NET_IFDESTROY); 2583 if (error == 0) 2584 error = if_clone_destroy(ifr->ifr_name); 2585 CURVNET_RESTORE(); 2586 return (error); 2587 2588 case SIOCIFGCLONERS: 2589 error = if_clone_list((struct if_clonereq *)data); 2590 CURVNET_RESTORE(); 2591 return (error); 2592 case SIOCGIFGMEMB: 2593 error = if_getgroupmembers((struct ifgroupreq *)data); 2594 CURVNET_RESTORE(); 2595 return (error); 2596 #if defined(INET) || defined(INET6) 2597 case SIOCSVH: 2598 case SIOCGVH: 2599 if (carp_ioctl_p == NULL) 2600 error = EPROTONOSUPPORT; 2601 else 2602 error = (*carp_ioctl_p)(ifr, cmd, td); 2603 CURVNET_RESTORE(); 2604 return (error); 2605 #endif 2606 } 2607 2608 ifp = ifunit_ref(ifr->ifr_name); 2609 if (ifp == NULL) { 2610 CURVNET_RESTORE(); 2611 return (ENXIO); 2612 } 2613 2614 error = ifhwioctl(cmd, ifp, data, td); 2615 if (error != ENOIOCTL) { 2616 if_rele(ifp); 2617 CURVNET_RESTORE(); 2618 return (error); 2619 } 2620 2621 oif_flags = ifp->if_flags; 2622 if (so->so_proto == NULL) { 2623 if_rele(ifp); 2624 CURVNET_RESTORE(); 2625 return (EOPNOTSUPP); 2626 } 2627 2628 /* 2629 * Pass the request on to the socket control method, and if the 2630 * latter returns EOPNOTSUPP, directly to the interface. 2631 * 2632 * Make an exception for the legacy SIOCSIF* requests. Drivers 2633 * trust SIOCSIFADDR et al to come from an already privileged 2634 * layer, and do not perform any credentials checks or input 2635 * validation. 2636 */ 2637 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 2638 ifp, td)); 2639 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2640 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2641 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2642 error = (*ifp->if_ioctl)(ifp, cmd, data); 2643 2644 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2645 #ifdef INET6 2646 if (ifp->if_flags & IFF_UP) 2647 in6_if_up(ifp); 2648 #endif 2649 } 2650 if_rele(ifp); 2651 CURVNET_RESTORE(); 2652 return (error); 2653 } 2654 2655 /* 2656 * The code common to handling reference counted flags, 2657 * e.g., in ifpromisc() and if_allmulti(). 2658 * The "pflag" argument can specify a permanent mode flag to check, 2659 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2660 * 2661 * Only to be used on stack-owned flags, not driver-owned flags. 2662 */ 2663 static int 2664 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 2665 { 2666 struct ifreq ifr; 2667 int error; 2668 int oldflags, oldcount; 2669 2670 /* Sanity checks to catch programming errors */ 2671 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 2672 ("%s: setting driver-owned flag %d", __func__, flag)); 2673 2674 if (onswitch) 2675 KASSERT(*refcount >= 0, 2676 ("%s: increment negative refcount %d for flag %d", 2677 __func__, *refcount, flag)); 2678 else 2679 KASSERT(*refcount > 0, 2680 ("%s: decrement non-positive refcount %d for flag %d", 2681 __func__, *refcount, flag)); 2682 2683 /* In case this mode is permanent, just touch refcount */ 2684 if (ifp->if_flags & pflag) { 2685 *refcount += onswitch ? 1 : -1; 2686 return (0); 2687 } 2688 2689 /* Save ifnet parameters for if_ioctl() may fail */ 2690 oldcount = *refcount; 2691 oldflags = ifp->if_flags; 2692 2693 /* 2694 * See if we aren't the only and touching refcount is enough. 2695 * Actually toggle interface flag if we are the first or last. 2696 */ 2697 if (onswitch) { 2698 if ((*refcount)++) 2699 return (0); 2700 ifp->if_flags |= flag; 2701 } else { 2702 if (--(*refcount)) 2703 return (0); 2704 ifp->if_flags &= ~flag; 2705 } 2706 2707 /* Call down the driver since we've changed interface flags */ 2708 if (ifp->if_ioctl == NULL) { 2709 error = EOPNOTSUPP; 2710 goto recover; 2711 } 2712 ifr.ifr_flags = ifp->if_flags & 0xffff; 2713 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2714 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2715 if (error) 2716 goto recover; 2717 /* Notify userland that interface flags have changed */ 2718 rt_ifmsg(ifp); 2719 return (0); 2720 2721 recover: 2722 /* Recover after driver error */ 2723 *refcount = oldcount; 2724 ifp->if_flags = oldflags; 2725 return (error); 2726 } 2727 2728 /* 2729 * Set/clear promiscuous mode on interface ifp based on the truth value 2730 * of pswitch. The calls are reference counted so that only the first 2731 * "on" request actually has an effect, as does the final "off" request. 2732 * Results are undefined if the "off" and "on" requests are not matched. 2733 */ 2734 int 2735 ifpromisc(struct ifnet *ifp, int pswitch) 2736 { 2737 int error; 2738 int oldflags = ifp->if_flags; 2739 2740 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 2741 &ifp->if_pcount, pswitch); 2742 /* If promiscuous mode status has changed, log a message */ 2743 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) 2744 log(LOG_INFO, "%s: promiscuous mode %s\n", 2745 ifp->if_xname, 2746 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 2747 return (error); 2748 } 2749 2750 /* 2751 * Return interface configuration 2752 * of system. List may be used 2753 * in later ioctl's (above) to get 2754 * other information. 2755 */ 2756 /*ARGSUSED*/ 2757 static int 2758 ifconf(u_long cmd, caddr_t data) 2759 { 2760 struct ifconf *ifc = (struct ifconf *)data; 2761 struct ifnet *ifp; 2762 struct ifaddr *ifa; 2763 struct ifreq ifr; 2764 struct sbuf *sb; 2765 int error, full = 0, valid_len, max_len; 2766 2767 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2768 max_len = MAXPHYS - 1; 2769 2770 /* Prevent hostile input from being able to crash the system */ 2771 if (ifc->ifc_len <= 0) 2772 return (EINVAL); 2773 2774 again: 2775 if (ifc->ifc_len <= max_len) { 2776 max_len = ifc->ifc_len; 2777 full = 1; 2778 } 2779 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2780 max_len = 0; 2781 valid_len = 0; 2782 2783 IFNET_RLOCK(); 2784 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2785 int addrs; 2786 2787 /* 2788 * Zero the ifr_name buffer to make sure we don't 2789 * disclose the contents of the stack. 2790 */ 2791 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 2792 2793 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 2794 >= sizeof(ifr.ifr_name)) { 2795 sbuf_delete(sb); 2796 IFNET_RUNLOCK(); 2797 return (ENAMETOOLONG); 2798 } 2799 2800 addrs = 0; 2801 IF_ADDR_RLOCK(ifp); 2802 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2803 struct sockaddr *sa = ifa->ifa_addr; 2804 2805 if (prison_if(curthread->td_ucred, sa) != 0) 2806 continue; 2807 addrs++; 2808 if (sa->sa_len <= sizeof(*sa)) { 2809 ifr.ifr_addr = *sa; 2810 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2811 max_len += sizeof(ifr); 2812 } else { 2813 sbuf_bcat(sb, &ifr, 2814 offsetof(struct ifreq, ifr_addr)); 2815 max_len += offsetof(struct ifreq, ifr_addr); 2816 sbuf_bcat(sb, sa, sa->sa_len); 2817 max_len += sa->sa_len; 2818 } 2819 2820 if (sbuf_error(sb) == 0) 2821 valid_len = sbuf_len(sb); 2822 } 2823 IF_ADDR_RUNLOCK(ifp); 2824 if (addrs == 0) { 2825 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 2826 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2827 max_len += sizeof(ifr); 2828 2829 if (sbuf_error(sb) == 0) 2830 valid_len = sbuf_len(sb); 2831 } 2832 } 2833 IFNET_RUNLOCK(); 2834 2835 /* 2836 * If we didn't allocate enough space (uncommon), try again. If 2837 * we have already allocated as much space as we are allowed, 2838 * return what we've got. 2839 */ 2840 if (valid_len != max_len && !full) { 2841 sbuf_delete(sb); 2842 goto again; 2843 } 2844 2845 ifc->ifc_len = valid_len; 2846 sbuf_finish(sb); 2847 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 2848 sbuf_delete(sb); 2849 return (error); 2850 } 2851 2852 /* 2853 * Just like ifpromisc(), but for all-multicast-reception mode. 2854 */ 2855 int 2856 if_allmulti(struct ifnet *ifp, int onswitch) 2857 { 2858 2859 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 2860 } 2861 2862 struct ifmultiaddr * 2863 if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 2864 { 2865 struct ifmultiaddr *ifma; 2866 2867 IF_ADDR_LOCK_ASSERT(ifp); 2868 2869 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2870 if (sa->sa_family == AF_LINK) { 2871 if (sa_dl_equal(ifma->ifma_addr, sa)) 2872 break; 2873 } else { 2874 if (sa_equal(ifma->ifma_addr, sa)) 2875 break; 2876 } 2877 } 2878 2879 return ifma; 2880 } 2881 2882 /* 2883 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 2884 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 2885 * the ifnet multicast address list here, so the caller must do that and 2886 * other setup work (such as notifying the device driver). The reference 2887 * count is initialized to 1. 2888 */ 2889 static struct ifmultiaddr * 2890 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 2891 int mflags) 2892 { 2893 struct ifmultiaddr *ifma; 2894 struct sockaddr *dupsa; 2895 2896 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 2897 M_ZERO); 2898 if (ifma == NULL) 2899 return (NULL); 2900 2901 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 2902 if (dupsa == NULL) { 2903 free(ifma, M_IFMADDR); 2904 return (NULL); 2905 } 2906 bcopy(sa, dupsa, sa->sa_len); 2907 ifma->ifma_addr = dupsa; 2908 2909 ifma->ifma_ifp = ifp; 2910 ifma->ifma_refcount = 1; 2911 ifma->ifma_protospec = NULL; 2912 2913 if (llsa == NULL) { 2914 ifma->ifma_lladdr = NULL; 2915 return (ifma); 2916 } 2917 2918 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 2919 if (dupsa == NULL) { 2920 free(ifma->ifma_addr, M_IFMADDR); 2921 free(ifma, M_IFMADDR); 2922 return (NULL); 2923 } 2924 bcopy(llsa, dupsa, llsa->sa_len); 2925 ifma->ifma_lladdr = dupsa; 2926 2927 return (ifma); 2928 } 2929 2930 /* 2931 * if_freemulti: free ifmultiaddr structure and possibly attached related 2932 * addresses. The caller is responsible for implementing reference 2933 * counting, notifying the driver, handling routing messages, and releasing 2934 * any dependent link layer state. 2935 */ 2936 static void 2937 if_freemulti(struct ifmultiaddr *ifma) 2938 { 2939 2940 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 2941 ifma->ifma_refcount)); 2942 2943 if (ifma->ifma_lladdr != NULL) 2944 free(ifma->ifma_lladdr, M_IFMADDR); 2945 free(ifma->ifma_addr, M_IFMADDR); 2946 free(ifma, M_IFMADDR); 2947 } 2948 2949 /* 2950 * Register an additional multicast address with a network interface. 2951 * 2952 * - If the address is already present, bump the reference count on the 2953 * address and return. 2954 * - If the address is not link-layer, look up a link layer address. 2955 * - Allocate address structures for one or both addresses, and attach to the 2956 * multicast address list on the interface. If automatically adding a link 2957 * layer address, the protocol address will own a reference to the link 2958 * layer address, to be freed when it is freed. 2959 * - Notify the network device driver of an addition to the multicast address 2960 * list. 2961 * 2962 * 'sa' points to caller-owned memory with the desired multicast address. 2963 * 2964 * 'retifma' will be used to return a pointer to the resulting multicast 2965 * address reference, if desired. 2966 */ 2967 int 2968 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 2969 struct ifmultiaddr **retifma) 2970 { 2971 struct ifmultiaddr *ifma, *ll_ifma; 2972 struct sockaddr *llsa; 2973 struct sockaddr_dl sdl; 2974 int error; 2975 2976 /* 2977 * If the address is already present, return a new reference to it; 2978 * otherwise, allocate storage and set up a new address. 2979 */ 2980 IF_ADDR_WLOCK(ifp); 2981 ifma = if_findmulti(ifp, sa); 2982 if (ifma != NULL) { 2983 ifma->ifma_refcount++; 2984 if (retifma != NULL) 2985 *retifma = ifma; 2986 IF_ADDR_WUNLOCK(ifp); 2987 return (0); 2988 } 2989 2990 /* 2991 * The address isn't already present; resolve the protocol address 2992 * into a link layer address, and then look that up, bump its 2993 * refcount or allocate an ifma for that also. 2994 * Most link layer resolving functions returns address data which 2995 * fits inside default sockaddr_dl structure. However callback 2996 * can allocate another sockaddr structure, in that case we need to 2997 * free it later. 2998 */ 2999 llsa = NULL; 3000 ll_ifma = NULL; 3001 if (ifp->if_resolvemulti != NULL) { 3002 /* Provide called function with buffer size information */ 3003 sdl.sdl_len = sizeof(sdl); 3004 llsa = (struct sockaddr *)&sdl; 3005 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3006 if (error) 3007 goto unlock_out; 3008 } 3009 3010 /* 3011 * Allocate the new address. Don't hook it up yet, as we may also 3012 * need to allocate a link layer multicast address. 3013 */ 3014 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3015 if (ifma == NULL) { 3016 error = ENOMEM; 3017 goto free_llsa_out; 3018 } 3019 3020 /* 3021 * If a link layer address is found, we'll need to see if it's 3022 * already present in the address list, or allocate is as well. 3023 * When this block finishes, the link layer address will be on the 3024 * list. 3025 */ 3026 if (llsa != NULL) { 3027 ll_ifma = if_findmulti(ifp, llsa); 3028 if (ll_ifma == NULL) { 3029 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3030 if (ll_ifma == NULL) { 3031 --ifma->ifma_refcount; 3032 if_freemulti(ifma); 3033 error = ENOMEM; 3034 goto free_llsa_out; 3035 } 3036 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3037 ifma_link); 3038 } else 3039 ll_ifma->ifma_refcount++; 3040 ifma->ifma_llifma = ll_ifma; 3041 } 3042 3043 /* 3044 * We now have a new multicast address, ifma, and possibly a new or 3045 * referenced link layer address. Add the primary address to the 3046 * ifnet address list. 3047 */ 3048 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3049 3050 if (retifma != NULL) 3051 *retifma = ifma; 3052 3053 /* 3054 * Must generate the message while holding the lock so that 'ifma' 3055 * pointer is still valid. 3056 */ 3057 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3058 IF_ADDR_WUNLOCK(ifp); 3059 3060 /* 3061 * We are certain we have added something, so call down to the 3062 * interface to let them know about it. 3063 */ 3064 if (ifp->if_ioctl != NULL) { 3065 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3066 } 3067 3068 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3069 link_free_sdl(llsa); 3070 3071 return (0); 3072 3073 free_llsa_out: 3074 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3075 link_free_sdl(llsa); 3076 3077 unlock_out: 3078 IF_ADDR_WUNLOCK(ifp); 3079 return (error); 3080 } 3081 3082 /* 3083 * Delete a multicast group membership by network-layer group address. 3084 * 3085 * Returns ENOENT if the entry could not be found. If ifp no longer 3086 * exists, results are undefined. This entry point should only be used 3087 * from subsystems which do appropriate locking to hold ifp for the 3088 * duration of the call. 3089 * Network-layer protocol domains must use if_delmulti_ifma(). 3090 */ 3091 int 3092 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3093 { 3094 struct ifmultiaddr *ifma; 3095 int lastref; 3096 #ifdef INVARIANTS 3097 struct ifnet *oifp; 3098 3099 IFNET_RLOCK_NOSLEEP(); 3100 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3101 if (ifp == oifp) 3102 break; 3103 if (ifp != oifp) 3104 ifp = NULL; 3105 IFNET_RUNLOCK_NOSLEEP(); 3106 3107 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3108 #endif 3109 if (ifp == NULL) 3110 return (ENOENT); 3111 3112 IF_ADDR_WLOCK(ifp); 3113 lastref = 0; 3114 ifma = if_findmulti(ifp, sa); 3115 if (ifma != NULL) 3116 lastref = if_delmulti_locked(ifp, ifma, 0); 3117 IF_ADDR_WUNLOCK(ifp); 3118 3119 if (ifma == NULL) 3120 return (ENOENT); 3121 3122 if (lastref && ifp->if_ioctl != NULL) { 3123 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3124 } 3125 3126 return (0); 3127 } 3128 3129 /* 3130 * Delete all multicast group membership for an interface. 3131 * Should be used to quickly flush all multicast filters. 3132 */ 3133 void 3134 if_delallmulti(struct ifnet *ifp) 3135 { 3136 struct ifmultiaddr *ifma; 3137 struct ifmultiaddr *next; 3138 3139 IF_ADDR_WLOCK(ifp); 3140 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3141 if_delmulti_locked(ifp, ifma, 0); 3142 IF_ADDR_WUNLOCK(ifp); 3143 } 3144 3145 /* 3146 * Delete a multicast group membership by group membership pointer. 3147 * Network-layer protocol domains must use this routine. 3148 * 3149 * It is safe to call this routine if the ifp disappeared. 3150 */ 3151 void 3152 if_delmulti_ifma(struct ifmultiaddr *ifma) 3153 { 3154 struct ifnet *ifp; 3155 int lastref; 3156 3157 ifp = ifma->ifma_ifp; 3158 #ifdef DIAGNOSTIC 3159 if (ifp == NULL) { 3160 printf("%s: ifma_ifp seems to be detached\n", __func__); 3161 } else { 3162 struct ifnet *oifp; 3163 3164 IFNET_RLOCK_NOSLEEP(); 3165 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3166 if (ifp == oifp) 3167 break; 3168 if (ifp != oifp) { 3169 printf("%s: ifnet %p disappeared\n", __func__, ifp); 3170 ifp = NULL; 3171 } 3172 IFNET_RUNLOCK_NOSLEEP(); 3173 } 3174 #endif 3175 /* 3176 * If and only if the ifnet instance exists: Acquire the address lock. 3177 */ 3178 if (ifp != NULL) 3179 IF_ADDR_WLOCK(ifp); 3180 3181 lastref = if_delmulti_locked(ifp, ifma, 0); 3182 3183 if (ifp != NULL) { 3184 /* 3185 * If and only if the ifnet instance exists: 3186 * Release the address lock. 3187 * If the group was left: update the hardware hash filter. 3188 */ 3189 IF_ADDR_WUNLOCK(ifp); 3190 if (lastref && ifp->if_ioctl != NULL) { 3191 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3192 } 3193 } 3194 } 3195 3196 /* 3197 * Perform deletion of network-layer and/or link-layer multicast address. 3198 * 3199 * Return 0 if the reference count was decremented. 3200 * Return 1 if the final reference was released, indicating that the 3201 * hardware hash filter should be reprogrammed. 3202 */ 3203 static int 3204 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3205 { 3206 struct ifmultiaddr *ll_ifma; 3207 3208 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3209 KASSERT(ifma->ifma_ifp == ifp, 3210 ("%s: inconsistent ifp %p", __func__, ifp)); 3211 IF_ADDR_WLOCK_ASSERT(ifp); 3212 } 3213 3214 ifp = ifma->ifma_ifp; 3215 3216 /* 3217 * If the ifnet is detaching, null out references to ifnet, 3218 * so that upper protocol layers will notice, and not attempt 3219 * to obtain locks for an ifnet which no longer exists. The 3220 * routing socket announcement must happen before the ifnet 3221 * instance is detached from the system. 3222 */ 3223 if (detaching) { 3224 #ifdef DIAGNOSTIC 3225 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3226 #endif 3227 /* 3228 * ifp may already be nulled out if we are being reentered 3229 * to delete the ll_ifma. 3230 */ 3231 if (ifp != NULL) { 3232 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3233 ifma->ifma_ifp = NULL; 3234 } 3235 } 3236 3237 if (--ifma->ifma_refcount > 0) 3238 return 0; 3239 3240 /* 3241 * If this ifma is a network-layer ifma, a link-layer ifma may 3242 * have been associated with it. Release it first if so. 3243 */ 3244 ll_ifma = ifma->ifma_llifma; 3245 if (ll_ifma != NULL) { 3246 KASSERT(ifma->ifma_lladdr != NULL, 3247 ("%s: llifma w/o lladdr", __func__)); 3248 if (detaching) 3249 ll_ifma->ifma_ifp = NULL; /* XXX */ 3250 if (--ll_ifma->ifma_refcount == 0) { 3251 if (ifp != NULL) { 3252 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3253 ifma_link); 3254 } 3255 if_freemulti(ll_ifma); 3256 } 3257 } 3258 3259 if (ifp != NULL) 3260 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3261 3262 if_freemulti(ifma); 3263 3264 /* 3265 * The last reference to this instance of struct ifmultiaddr 3266 * was released; the hardware should be notified of this change. 3267 */ 3268 return 1; 3269 } 3270 3271 /* 3272 * Set the link layer address on an interface. 3273 * 3274 * At this time we only support certain types of interfaces, 3275 * and we don't allow the length of the address to change. 3276 */ 3277 int 3278 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3279 { 3280 struct sockaddr_dl *sdl; 3281 struct ifaddr *ifa; 3282 struct ifreq ifr; 3283 3284 IF_ADDR_RLOCK(ifp); 3285 ifa = ifp->if_addr; 3286 if (ifa == NULL) { 3287 IF_ADDR_RUNLOCK(ifp); 3288 return (EINVAL); 3289 } 3290 ifa_ref(ifa); 3291 IF_ADDR_RUNLOCK(ifp); 3292 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3293 if (sdl == NULL) { 3294 ifa_free(ifa); 3295 return (EINVAL); 3296 } 3297 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3298 ifa_free(ifa); 3299 return (EINVAL); 3300 } 3301 switch (ifp->if_type) { 3302 case IFT_ETHER: 3303 case IFT_FDDI: 3304 case IFT_XETHER: 3305 case IFT_ISO88025: 3306 case IFT_L2VLAN: 3307 case IFT_BRIDGE: 3308 case IFT_ARCNET: 3309 case IFT_IEEE8023ADLAG: 3310 case IFT_IEEE80211: 3311 bcopy(lladdr, LLADDR(sdl), len); 3312 ifa_free(ifa); 3313 break; 3314 default: 3315 ifa_free(ifa); 3316 return (ENODEV); 3317 } 3318 3319 /* 3320 * If the interface is already up, we need 3321 * to re-init it in order to reprogram its 3322 * address filter. 3323 */ 3324 if ((ifp->if_flags & IFF_UP) != 0) { 3325 if (ifp->if_ioctl) { 3326 ifp->if_flags &= ~IFF_UP; 3327 ifr.ifr_flags = ifp->if_flags & 0xffff; 3328 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3329 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3330 ifp->if_flags |= IFF_UP; 3331 ifr.ifr_flags = ifp->if_flags & 0xffff; 3332 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3333 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3334 } 3335 #ifdef INET 3336 /* 3337 * Also send gratuitous ARPs to notify other nodes about 3338 * the address change. 3339 */ 3340 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3341 if (ifa->ifa_addr->sa_family == AF_INET) 3342 arp_ifinit(ifp, ifa); 3343 } 3344 #endif 3345 } 3346 return (0); 3347 } 3348 3349 /* 3350 * The name argument must be a pointer to storage which will last as 3351 * long as the interface does. For physical devices, the result of 3352 * device_get_name(dev) is a good choice and for pseudo-devices a 3353 * static string works well. 3354 */ 3355 void 3356 if_initname(struct ifnet *ifp, const char *name, int unit) 3357 { 3358 ifp->if_dname = name; 3359 ifp->if_dunit = unit; 3360 if (unit != IF_DUNIT_NONE) 3361 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3362 else 3363 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3364 } 3365 3366 int 3367 if_printf(struct ifnet *ifp, const char * fmt, ...) 3368 { 3369 va_list ap; 3370 int retval; 3371 3372 retval = printf("%s: ", ifp->if_xname); 3373 va_start(ap, fmt); 3374 retval += vprintf(fmt, ap); 3375 va_end(ap); 3376 return (retval); 3377 } 3378 3379 void 3380 if_start(struct ifnet *ifp) 3381 { 3382 3383 (*(ifp)->if_start)(ifp); 3384 } 3385 3386 /* 3387 * Backwards compatibility interface for drivers 3388 * that have not implemented it 3389 */ 3390 static int 3391 if_transmit(struct ifnet *ifp, struct mbuf *m) 3392 { 3393 int error; 3394 3395 IFQ_HANDOFF(ifp, m, error); 3396 return (error); 3397 } 3398 3399 int 3400 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3401 { 3402 int active = 0; 3403 3404 IF_LOCK(ifq); 3405 if (_IF_QFULL(ifq)) { 3406 _IF_DROP(ifq); 3407 IF_UNLOCK(ifq); 3408 m_freem(m); 3409 return (0); 3410 } 3411 if (ifp != NULL) { 3412 ifp->if_obytes += m->m_pkthdr.len + adjust; 3413 if (m->m_flags & (M_BCAST|M_MCAST)) 3414 ifp->if_omcasts++; 3415 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3416 } 3417 _IF_ENQUEUE(ifq, m); 3418 IF_UNLOCK(ifq); 3419 if (ifp != NULL && !active) 3420 (*(ifp)->if_start)(ifp); 3421 return (1); 3422 } 3423 3424 void 3425 if_register_com_alloc(u_char type, 3426 if_com_alloc_t *a, if_com_free_t *f) 3427 { 3428 3429 KASSERT(if_com_alloc[type] == NULL, 3430 ("if_register_com_alloc: %d already registered", type)); 3431 KASSERT(if_com_free[type] == NULL, 3432 ("if_register_com_alloc: %d free already registered", type)); 3433 3434 if_com_alloc[type] = a; 3435 if_com_free[type] = f; 3436 } 3437 3438 void 3439 if_deregister_com_alloc(u_char type) 3440 { 3441 3442 KASSERT(if_com_alloc[type] != NULL, 3443 ("if_deregister_com_alloc: %d not registered", type)); 3444 KASSERT(if_com_free[type] != NULL, 3445 ("if_deregister_com_alloc: %d free not registered", type)); 3446 if_com_alloc[type] = NULL; 3447 if_com_free[type] = NULL; 3448 } 3449