if.c revision 274043
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: stable/10/sys/net/if.c 274043 2014-11-03 12:38:29Z hselasky $ 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 98struct ifindex_entry { 99 struct ifnet *ife_ifnet; 100}; 101 102SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 103SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 104 105TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen); 106SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 107 &ifqmaxlen, 0, "max send queue size"); 108 109/* Log link state change events */ 110static int log_link_state_change = 1; 111 112SYSCTL_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 */ 117static unsigned int ifdescr_maxlen = 1024; 118SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 119 &ifdescr_maxlen, 0, 120 "administrative maximum length for interface description"); 121 122static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 123 124/* global sx for non-critical path ifdescr */ 125static struct sx ifdescr_sx; 126SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 127 128void (*bridge_linkstate_p)(struct ifnet *ifp); 129void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 130void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 131/* These are external hooks for CARP. */ 132void (*carp_linkstate_p)(struct ifnet *ifp); 133void (*carp_demote_adj_p)(int, char *); 134int (*carp_master_p)(struct ifaddr *); 135#if defined(INET) || defined(INET6) 136int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 137int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 138 const struct sockaddr *sa); 139int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 140int (*carp_attach_p)(struct ifaddr *, int); 141void (*carp_detach_p)(struct ifaddr *); 142#endif 143#ifdef INET 144int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 145#endif 146#ifdef INET6 147struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 148caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 149 const struct in6_addr *taddr); 150#endif 151 152struct 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 */ 159static void if_attachdomain(void *); 160static void if_attachdomain1(struct ifnet *); 161static int ifconf(u_long, caddr_t); 162static void if_freemulti(struct ifmultiaddr *); 163static void if_init(void *); 164static void if_grow(void); 165static void if_route(struct ifnet *, int flag, int fam); 166static int if_setflag(struct ifnet *, int, int, int *, int); 167static int if_transmit(struct ifnet *ifp, struct mbuf *m); 168static void if_unroute(struct ifnet *, int flag, int fam); 169static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 170static int if_rtdel(struct radix_node *, void *); 171static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 172static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 173static void do_link_state_change(void *, int); 174static int if_getgroup(struct ifgroupreq *, struct ifnet *); 175static int if_getgroupmembers(struct ifgroupreq *); 176static void if_delgroups(struct ifnet *); 177static void if_attach_internal(struct ifnet *, int); 178static 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 */ 185extern void nd6_setmtu(struct ifnet *); 186#endif 187 188VNET_DEFINE(int, if_index); 189int ifqmaxlen = IFQ_MAXLEN; 190VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 191VNET_DEFINE(struct ifgrouphead, ifg_head); 192 193static VNET_DEFINE(int, if_indexlim) = 8; 194 195/* Table of ifnet by index. */ 196VNET_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 */ 210struct rwlock ifnet_rwlock; 211struct 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 220static if_com_alloc_t *if_com_alloc[256]; 221static if_com_free_t *if_com_free[256]; 222 223static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 224MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 225MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 226 227struct ifnet * 228ifnet_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 238struct ifnet * 239ifnet_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 249struct ifnet * 250ifnet_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 */ 269static int 270ifindex_alloc_locked(u_short *idxp) 271{ 272 u_short idx; 273 274 IFNET_WLOCK_ASSERT(); 275 276retry: 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 < 1) 288 return (ENOSPC); 289 if (idx >= V_if_indexlim) { 290 if_grow(); 291 goto retry; 292 } 293 if (idx > V_if_index) 294 V_if_index = idx; 295 *idxp = idx; 296 return (0); 297} 298 299static void 300ifindex_free_locked(u_short idx) 301{ 302 303 IFNET_WLOCK_ASSERT(); 304 305 V_ifindex_table[idx].ife_ifnet = NULL; 306 while (V_if_index > 0 && 307 V_ifindex_table[V_if_index].ife_ifnet == NULL) 308 V_if_index--; 309} 310 311static void 312ifindex_free(u_short idx) 313{ 314 315 IFNET_WLOCK(); 316 ifindex_free_locked(idx); 317 IFNET_WUNLOCK(); 318} 319 320static void 321ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) 322{ 323 324 IFNET_WLOCK_ASSERT(); 325 326 V_ifindex_table[idx].ife_ifnet = ifp; 327} 328 329static void 330ifnet_setbyindex(u_short idx, struct ifnet *ifp) 331{ 332 333 IFNET_WLOCK(); 334 ifnet_setbyindex_locked(idx, ifp); 335 IFNET_WUNLOCK(); 336} 337 338struct ifaddr * 339ifaddr_byindex(u_short idx) 340{ 341 struct ifaddr *ifa; 342 343 IFNET_RLOCK_NOSLEEP(); 344 ifa = ifnet_byindex_locked(idx)->if_addr; 345 if (ifa != NULL) 346 ifa_ref(ifa); 347 IFNET_RUNLOCK_NOSLEEP(); 348 return (ifa); 349} 350 351/* 352 * Network interface utility routines. 353 * 354 * Routines with ifa_ifwith* names take sockaddr *'s as 355 * parameters. 356 */ 357 358static void 359vnet_if_init(const void *unused __unused) 360{ 361 362 TAILQ_INIT(&V_ifnet); 363 TAILQ_INIT(&V_ifg_head); 364 IFNET_WLOCK(); 365 if_grow(); /* create initial table */ 366 IFNET_WUNLOCK(); 367 vnet_if_clone_init(); 368} 369VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 370 NULL); 371 372/* ARGSUSED*/ 373static void 374if_init(void *dummy __unused) 375{ 376 377 IFNET_LOCK_INIT(); 378 if_clone_init(); 379} 380SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL); 381 382 383#ifdef VIMAGE 384static void 385vnet_if_uninit(const void *unused __unused) 386{ 387 388 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 389 "not empty", __func__, __LINE__, &V_ifnet)); 390 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 391 "not empty", __func__, __LINE__, &V_ifg_head)); 392 393 free((caddr_t)V_ifindex_table, M_IFNET); 394} 395VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 396 vnet_if_uninit, NULL); 397#endif 398 399static void 400if_grow(void) 401{ 402 int oldlim; 403 u_int n; 404 struct ifindex_entry *e; 405 406 IFNET_WLOCK_ASSERT(); 407 oldlim = V_if_indexlim; 408 IFNET_WUNLOCK(); 409 n = (oldlim << 1) * sizeof(*e); 410 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 411 IFNET_WLOCK(); 412 if (V_if_indexlim != oldlim) { 413 free(e, M_IFNET); 414 return; 415 } 416 if (V_ifindex_table != NULL) { 417 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 418 free((caddr_t)V_ifindex_table, M_IFNET); 419 } 420 V_if_indexlim <<= 1; 421 V_ifindex_table = e; 422} 423 424/* 425 * Allocate a struct ifnet and an index for an interface. A layer 2 426 * common structure will also be allocated if an allocation routine is 427 * registered for the passed type. 428 */ 429struct ifnet * 430if_alloc(u_char type) 431{ 432 struct ifnet *ifp; 433 u_short idx; 434 435 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 436 IFNET_WLOCK(); 437 if (ifindex_alloc_locked(&idx) != 0) { 438 IFNET_WUNLOCK(); 439 free(ifp, M_IFNET); 440 return (NULL); 441 } 442 ifnet_setbyindex_locked(idx, IFNET_HOLD); 443 IFNET_WUNLOCK(); 444 ifp->if_index = idx; 445 ifp->if_type = type; 446 ifp->if_alloctype = type; 447 if (if_com_alloc[type] != NULL) { 448 ifp->if_l2com = if_com_alloc[type](type, ifp); 449 if (ifp->if_l2com == NULL) { 450 free(ifp, M_IFNET); 451 ifindex_free(idx); 452 return (NULL); 453 } 454 } 455 456 IF_ADDR_LOCK_INIT(ifp); 457 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 458 ifp->if_afdata_initialized = 0; 459 IF_AFDATA_LOCK_INIT(ifp); 460 TAILQ_INIT(&ifp->if_addrhead); 461 TAILQ_INIT(&ifp->if_multiaddrs); 462 TAILQ_INIT(&ifp->if_groups); 463#ifdef MAC 464 mac_ifnet_init(ifp); 465#endif 466 ifq_init(&ifp->if_snd, ifp); 467 468 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 469 ifnet_setbyindex(ifp->if_index, ifp); 470 return (ifp); 471} 472 473/* 474 * Do the actual work of freeing a struct ifnet, and layer 2 common 475 * structure. This call is made when the last reference to an 476 * interface is released. 477 */ 478static void 479if_free_internal(struct ifnet *ifp) 480{ 481 482 KASSERT((ifp->if_flags & IFF_DYING), 483 ("if_free_internal: interface not dying")); 484 485 if (if_com_free[ifp->if_alloctype] != NULL) 486 if_com_free[ifp->if_alloctype](ifp->if_l2com, 487 ifp->if_alloctype); 488 489#ifdef MAC 490 mac_ifnet_destroy(ifp); 491#endif /* MAC */ 492 if (ifp->if_description != NULL) 493 free(ifp->if_description, M_IFDESCR); 494 IF_AFDATA_DESTROY(ifp); 495 IF_ADDR_LOCK_DESTROY(ifp); 496 ifq_delete(&ifp->if_snd); 497 free(ifp, M_IFNET); 498} 499 500/* 501 * Deregister an interface and free the associated storage. 502 */ 503void 504if_free(struct ifnet *ifp) 505{ 506 507 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 508 509 CURVNET_SET_QUIET(ifp->if_vnet); 510 IFNET_WLOCK(); 511 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 512 ("%s: freeing unallocated ifnet", ifp->if_xname)); 513 514 ifindex_free_locked(ifp->if_index); 515 IFNET_WUNLOCK(); 516 517 if (refcount_release(&ifp->if_refcount)) 518 if_free_internal(ifp); 519 CURVNET_RESTORE(); 520} 521 522/* 523 * Interfaces to keep an ifnet type-stable despite the possibility of the 524 * driver calling if_free(). If there are additional references, we defer 525 * freeing the underlying data structure. 526 */ 527void 528if_ref(struct ifnet *ifp) 529{ 530 531 /* We don't assert the ifnet list lock here, but arguably should. */ 532 refcount_acquire(&ifp->if_refcount); 533} 534 535void 536if_rele(struct ifnet *ifp) 537{ 538 539 if (!refcount_release(&ifp->if_refcount)) 540 return; 541 if_free_internal(ifp); 542} 543 544void 545ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 546{ 547 548 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 549 550 if (ifq->ifq_maxlen == 0) 551 ifq->ifq_maxlen = ifqmaxlen; 552 553 ifq->altq_type = 0; 554 ifq->altq_disc = NULL; 555 ifq->altq_flags &= ALTQF_CANTCHANGE; 556 ifq->altq_tbr = NULL; 557 ifq->altq_ifp = ifp; 558} 559 560void 561ifq_delete(struct ifaltq *ifq) 562{ 563 mtx_destroy(&ifq->ifq_mtx); 564} 565 566/* 567 * Perform generic interface initalization tasks and attach the interface 568 * to the list of "active" interfaces. If vmove flag is set on entry 569 * to if_attach_internal(), perform only a limited subset of initialization 570 * tasks, given that we are moving from one vnet to another an ifnet which 571 * has already been fully initialized. 572 * 573 * XXX: 574 * - The decision to return void and thus require this function to 575 * succeed is questionable. 576 * - We should probably do more sanity checking. For instance we don't 577 * do anything to insure if_xname is unique or non-empty. 578 */ 579void 580if_attach(struct ifnet *ifp) 581{ 582 583 if_attach_internal(ifp, 0); 584} 585 586/* 587 * Compute the least common TSO limit. 588 */ 589void 590if_hw_tsomax_common(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax) 591{ 592 /* 593 * 1) If there is no limit currently, take the limit from 594 * the network adapter. 595 * 596 * 2) If the network adapter has a limit below the current 597 * limit, apply it. 598 */ 599 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 600 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 601 pmax->tsomaxbytes = ifp->if_hw_tsomax; 602 } 603 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 604 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 605 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 606 } 607 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 608 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 609 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 610 } 611} 612 613/* 614 * Update TSO limit of a network adapter. 615 * 616 * Returns zero if no change. Else non-zero. 617 */ 618int 619if_hw_tsomax_update(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax) 620{ 621 int retval = 0; 622 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 623 ifp->if_hw_tsomax = pmax->tsomaxbytes; 624 retval++; 625 } 626 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 627 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 628 retval++; 629 } 630 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 631 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 632 retval++; 633 } 634 return (retval); 635} 636 637static void 638if_attach_internal(struct ifnet *ifp, int vmove) 639{ 640 unsigned socksize, ifasize; 641 int namelen, masklen; 642 struct sockaddr_dl *sdl; 643 struct ifaddr *ifa; 644 645 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 646 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 647 ifp->if_xname); 648 649#ifdef VIMAGE 650 ifp->if_vnet = curvnet; 651 if (ifp->if_home_vnet == NULL) 652 ifp->if_home_vnet = curvnet; 653#endif 654 655 if_addgroup(ifp, IFG_ALL); 656 657 getmicrotime(&ifp->if_lastchange); 658 ifp->if_data.ifi_epoch = time_uptime; 659 ifp->if_data.ifi_datalen = sizeof(struct if_data); 660 661 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 662 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 663 ("transmit and qflush must both either be set or both be NULL")); 664 if (ifp->if_transmit == NULL) { 665 ifp->if_transmit = if_transmit; 666 ifp->if_qflush = if_qflush; 667 } 668 669 if (!vmove) { 670#ifdef MAC 671 mac_ifnet_create(ifp); 672#endif 673 674 /* 675 * Create a Link Level name for this device. 676 */ 677 namelen = strlen(ifp->if_xname); 678 /* 679 * Always save enough space for any possiable name so we 680 * can do a rename in place later. 681 */ 682 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 683 socksize = masklen + ifp->if_addrlen; 684 if (socksize < sizeof(*sdl)) 685 socksize = sizeof(*sdl); 686 socksize = roundup2(socksize, sizeof(long)); 687 ifasize = sizeof(*ifa) + 2 * socksize; 688 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 689 ifa_init(ifa); 690 sdl = (struct sockaddr_dl *)(ifa + 1); 691 sdl->sdl_len = socksize; 692 sdl->sdl_family = AF_LINK; 693 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 694 sdl->sdl_nlen = namelen; 695 sdl->sdl_index = ifp->if_index; 696 sdl->sdl_type = ifp->if_type; 697 ifp->if_addr = ifa; 698 ifa->ifa_ifp = ifp; 699 ifa->ifa_rtrequest = link_rtrequest; 700 ifa->ifa_addr = (struct sockaddr *)sdl; 701 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 702 ifa->ifa_netmask = (struct sockaddr *)sdl; 703 sdl->sdl_len = masklen; 704 while (namelen != 0) 705 sdl->sdl_data[--namelen] = 0xff; 706 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 707 /* Reliably crash if used uninitialized. */ 708 ifp->if_broadcastaddr = NULL; 709 710#if defined(INET) || defined(INET6) 711 /* Use defaults for TSO, if nothing is set */ 712 if (ifp->if_hw_tsomax == 0 && 713 ifp->if_hw_tsomaxsegcount == 0 && 714 ifp->if_hw_tsomaxsegsize == 0) { 715 /* 716 * The TSO defaults needs to be such that an 717 * NFS mbuf list of 35 mbufs totalling just 718 * below 64K works and that a chain of mbufs 719 * can be defragged into at most 32 segments: 720 */ 721 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 722 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 723 ifp->if_hw_tsomaxsegcount = 35; 724 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 725 726 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 727 if (ifp->if_capabilities & IFCAP_TSO) { 728 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 729 ifp->if_hw_tsomax, 730 ifp->if_hw_tsomaxsegcount, 731 ifp->if_hw_tsomaxsegsize); 732 } 733 } 734 /* 735 * If the "if_hw_tsomax" limit is set, check if it is 736 * too small: 737 */ 738 KASSERT(ifp->if_hw_tsomax == 0 || 739 ifp->if_hw_tsomax >= (IP_MAXPACKET / 8), 740 ("%s: if_hw_tsomax is outside of range", __func__)); 741#endif 742 } 743#ifdef VIMAGE 744 else { 745 /* 746 * Update the interface index in the link layer address 747 * of the interface. 748 */ 749 for (ifa = ifp->if_addr; ifa != NULL; 750 ifa = TAILQ_NEXT(ifa, ifa_link)) { 751 if (ifa->ifa_addr->sa_family == AF_LINK) { 752 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 753 sdl->sdl_index = ifp->if_index; 754 } 755 } 756 } 757#endif 758 759 IFNET_WLOCK(); 760 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 761#ifdef VIMAGE 762 curvnet->vnet_ifcnt++; 763#endif 764 IFNET_WUNLOCK(); 765 766 if (domain_init_status >= 2) 767 if_attachdomain1(ifp); 768 769 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 770 if (IS_DEFAULT_VNET(curvnet)) 771 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 772 773 /* Announce the interface. */ 774 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 775} 776 777static void 778if_attachdomain(void *dummy) 779{ 780 struct ifnet *ifp; 781 782 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 783 if_attachdomain1(ifp); 784} 785SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 786 if_attachdomain, NULL); 787 788static void 789if_attachdomain1(struct ifnet *ifp) 790{ 791 struct domain *dp; 792 793 /* 794 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 795 * cannot lock ifp->if_afdata initialization, entirely. 796 */ 797 if (IF_AFDATA_TRYLOCK(ifp) == 0) 798 return; 799 if (ifp->if_afdata_initialized >= domain_init_status) { 800 IF_AFDATA_UNLOCK(ifp); 801 log(LOG_WARNING, "%s called more than once on %s\n", 802 __func__, ifp->if_xname); 803 return; 804 } 805 ifp->if_afdata_initialized = domain_init_status; 806 IF_AFDATA_UNLOCK(ifp); 807 808 /* address family dependent data region */ 809 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 810 for (dp = domains; dp; dp = dp->dom_next) { 811 if (dp->dom_ifattach) 812 ifp->if_afdata[dp->dom_family] = 813 (*dp->dom_ifattach)(ifp); 814 } 815} 816 817/* 818 * Remove any unicast or broadcast network addresses from an interface. 819 */ 820void 821if_purgeaddrs(struct ifnet *ifp) 822{ 823 struct ifaddr *ifa, *next; 824 825 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 826 if (ifa->ifa_addr->sa_family == AF_LINK) 827 continue; 828#ifdef INET 829 /* XXX: Ugly!! ad hoc just for INET */ 830 if (ifa->ifa_addr->sa_family == AF_INET) { 831 struct ifaliasreq ifr; 832 833 bzero(&ifr, sizeof(ifr)); 834 ifr.ifra_addr = *ifa->ifa_addr; 835 if (ifa->ifa_dstaddr) 836 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 837 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 838 NULL) == 0) 839 continue; 840 } 841#endif /* INET */ 842#ifdef INET6 843 if (ifa->ifa_addr->sa_family == AF_INET6) { 844 in6_purgeaddr(ifa); 845 /* ifp_addrhead is already updated */ 846 continue; 847 } 848#endif /* INET6 */ 849 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 850 ifa_free(ifa); 851 } 852} 853 854/* 855 * Remove any multicast network addresses from an interface when an ifnet 856 * is going away. 857 */ 858static void 859if_purgemaddrs(struct ifnet *ifp) 860{ 861 struct ifmultiaddr *ifma; 862 struct ifmultiaddr *next; 863 864 IF_ADDR_WLOCK(ifp); 865 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 866 if_delmulti_locked(ifp, ifma, 1); 867 IF_ADDR_WUNLOCK(ifp); 868} 869 870/* 871 * Detach an interface, removing it from the list of "active" interfaces. 872 * If vmove flag is set on entry to if_detach_internal(), perform only a 873 * limited subset of cleanup tasks, given that we are moving an ifnet from 874 * one vnet to another, where it must be fully operational. 875 * 876 * XXXRW: There are some significant questions about event ordering, and 877 * how to prevent things from starting to use the interface during detach. 878 */ 879void 880if_detach(struct ifnet *ifp) 881{ 882 883 CURVNET_SET_QUIET(ifp->if_vnet); 884 if_detach_internal(ifp, 0); 885 CURVNET_RESTORE(); 886} 887 888static void 889if_detach_internal(struct ifnet *ifp, int vmove) 890{ 891 struct ifaddr *ifa; 892 struct radix_node_head *rnh; 893 int i, j; 894 struct domain *dp; 895 struct ifnet *iter; 896 int found = 0; 897 898 IFNET_WLOCK(); 899 TAILQ_FOREACH(iter, &V_ifnet, if_link) 900 if (iter == ifp) { 901 TAILQ_REMOVE(&V_ifnet, ifp, if_link); 902 found = 1; 903 break; 904 } 905#ifdef VIMAGE 906 if (found) 907 curvnet->vnet_ifcnt--; 908#endif 909 IFNET_WUNLOCK(); 910 if (!found) { 911 if (vmove) 912 panic("%s: ifp=%p not on the ifnet tailq %p", 913 __func__, ifp, &V_ifnet); 914 else 915 return; /* XXX this should panic as well? */ 916 } 917 918 /* 919 * Remove/wait for pending events. 920 */ 921 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 922 923 /* 924 * Remove routes and flush queues. 925 */ 926 if_down(ifp); 927#ifdef ALTQ 928 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 929 altq_disable(&ifp->if_snd); 930 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 931 altq_detach(&ifp->if_snd); 932#endif 933 934 if_purgeaddrs(ifp); 935 936#ifdef INET 937 in_ifdetach(ifp); 938#endif 939 940#ifdef INET6 941 /* 942 * Remove all IPv6 kernel structs related to ifp. This should be done 943 * before removing routing entries below, since IPv6 interface direct 944 * routes are expected to be removed by the IPv6-specific kernel API. 945 * Otherwise, the kernel will detect some inconsistency and bark it. 946 */ 947 in6_ifdetach(ifp); 948#endif 949 if_purgemaddrs(ifp); 950 951 if (!vmove) { 952 /* 953 * Prevent further calls into the device driver via ifnet. 954 */ 955 if_dead(ifp); 956 957 /* 958 * Remove link ifaddr pointer and maybe decrement if_index. 959 * Clean up all addresses. 960 */ 961 ifp->if_addr = NULL; 962 963 /* We can now free link ifaddr. */ 964 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 965 ifa = TAILQ_FIRST(&ifp->if_addrhead); 966 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 967 ifa_free(ifa); 968 } 969 } 970 971 /* 972 * Delete all remaining routes using this interface 973 * Unfortuneatly the only way to do this is to slog through 974 * the entire routing table looking for routes which point 975 * to this interface...oh well... 976 */ 977 for (i = 1; i <= AF_MAX; i++) { 978 for (j = 0; j < rt_numfibs; j++) { 979 rnh = rt_tables_get_rnh(j, i); 980 if (rnh == NULL) 981 continue; 982 RADIX_NODE_HEAD_LOCK(rnh); 983 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 984 RADIX_NODE_HEAD_UNLOCK(rnh); 985 } 986 } 987 988 /* Announce that the interface is gone. */ 989 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 990 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 991 if (IS_DEFAULT_VNET(curvnet)) 992 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 993 if_delgroups(ifp); 994 995 /* 996 * We cannot hold the lock over dom_ifdetach calls as they might 997 * sleep, for example trying to drain a callout, thus open up the 998 * theoretical race with re-attaching. 999 */ 1000 IF_AFDATA_LOCK(ifp); 1001 i = ifp->if_afdata_initialized; 1002 ifp->if_afdata_initialized = 0; 1003 IF_AFDATA_UNLOCK(ifp); 1004 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1005 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 1006 (*dp->dom_ifdetach)(ifp, 1007 ifp->if_afdata[dp->dom_family]); 1008 } 1009} 1010 1011#ifdef VIMAGE 1012/* 1013 * if_vmove() performs a limited version of if_detach() in current 1014 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1015 * An attempt is made to shrink if_index in current vnet, find an 1016 * unused if_index in target vnet and calls if_grow() if necessary, 1017 * and finally find an unused if_xname for the target vnet. 1018 */ 1019void 1020if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1021{ 1022 u_short idx; 1023 1024 /* 1025 * Detach from current vnet, but preserve LLADDR info, do not 1026 * mark as dead etc. so that the ifnet can be reattached later. 1027 */ 1028 if_detach_internal(ifp, 1); 1029 1030 /* 1031 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1032 * the if_index for that vnet if possible. 1033 * 1034 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1035 * or we'd lock on one vnet and unlock on another. 1036 */ 1037 IFNET_WLOCK(); 1038 ifindex_free_locked(ifp->if_index); 1039 IFNET_WUNLOCK(); 1040 1041 /* 1042 * Perform interface-specific reassignment tasks, if provided by 1043 * the driver. 1044 */ 1045 if (ifp->if_reassign != NULL) 1046 ifp->if_reassign(ifp, new_vnet, NULL); 1047 1048 /* 1049 * Switch to the context of the target vnet. 1050 */ 1051 CURVNET_SET_QUIET(new_vnet); 1052 1053 IFNET_WLOCK(); 1054 if (ifindex_alloc_locked(&idx) != 0) { 1055 IFNET_WUNLOCK(); 1056 panic("if_index overflow"); 1057 } 1058 ifp->if_index = idx; 1059 ifnet_setbyindex_locked(ifp->if_index, ifp); 1060 IFNET_WUNLOCK(); 1061 1062 if_attach_internal(ifp, 1); 1063 1064 CURVNET_RESTORE(); 1065} 1066 1067/* 1068 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1069 */ 1070static int 1071if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1072{ 1073 struct prison *pr; 1074 struct ifnet *difp; 1075 1076 /* Try to find the prison within our visibility. */ 1077 sx_slock(&allprison_lock); 1078 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1079 sx_sunlock(&allprison_lock); 1080 if (pr == NULL) 1081 return (ENXIO); 1082 prison_hold_locked(pr); 1083 mtx_unlock(&pr->pr_mtx); 1084 1085 /* Do not try to move the iface from and to the same prison. */ 1086 if (pr->pr_vnet == ifp->if_vnet) { 1087 prison_free(pr); 1088 return (EEXIST); 1089 } 1090 1091 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1092 /* XXX Lock interfaces to avoid races. */ 1093 CURVNET_SET_QUIET(pr->pr_vnet); 1094 difp = ifunit(ifname); 1095 CURVNET_RESTORE(); 1096 if (difp != NULL) { 1097 prison_free(pr); 1098 return (EEXIST); 1099 } 1100 1101 /* Move the interface into the child jail/vnet. */ 1102 if_vmove(ifp, pr->pr_vnet); 1103 1104 /* Report the new if_xname back to the userland. */ 1105 sprintf(ifname, "%s", ifp->if_xname); 1106 1107 prison_free(pr); 1108 return (0); 1109} 1110 1111static int 1112if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1113{ 1114 struct prison *pr; 1115 struct vnet *vnet_dst; 1116 struct ifnet *ifp; 1117 1118 /* Try to find the prison within our visibility. */ 1119 sx_slock(&allprison_lock); 1120 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1121 sx_sunlock(&allprison_lock); 1122 if (pr == NULL) 1123 return (ENXIO); 1124 prison_hold_locked(pr); 1125 mtx_unlock(&pr->pr_mtx); 1126 1127 /* Make sure the named iface exists in the source prison/vnet. */ 1128 CURVNET_SET(pr->pr_vnet); 1129 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1130 if (ifp == NULL) { 1131 CURVNET_RESTORE(); 1132 prison_free(pr); 1133 return (ENXIO); 1134 } 1135 1136 /* Do not try to move the iface from and to the same prison. */ 1137 vnet_dst = TD_TO_VNET(td); 1138 if (vnet_dst == ifp->if_vnet) { 1139 CURVNET_RESTORE(); 1140 prison_free(pr); 1141 return (EEXIST); 1142 } 1143 1144 /* Get interface back from child jail/vnet. */ 1145 if_vmove(ifp, vnet_dst); 1146 CURVNET_RESTORE(); 1147 1148 /* Report the new if_xname back to the userland. */ 1149 sprintf(ifname, "%s", ifp->if_xname); 1150 1151 prison_free(pr); 1152 return (0); 1153} 1154#endif /* VIMAGE */ 1155 1156/* 1157 * Add a group to an interface 1158 */ 1159int 1160if_addgroup(struct ifnet *ifp, const char *groupname) 1161{ 1162 struct ifg_list *ifgl; 1163 struct ifg_group *ifg = NULL; 1164 struct ifg_member *ifgm; 1165 int new = 0; 1166 1167 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1168 groupname[strlen(groupname) - 1] <= '9') 1169 return (EINVAL); 1170 1171 IFNET_WLOCK(); 1172 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1173 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1174 IFNET_WUNLOCK(); 1175 return (EEXIST); 1176 } 1177 1178 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1179 M_NOWAIT)) == NULL) { 1180 IFNET_WUNLOCK(); 1181 return (ENOMEM); 1182 } 1183 1184 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1185 M_TEMP, M_NOWAIT)) == NULL) { 1186 free(ifgl, M_TEMP); 1187 IFNET_WUNLOCK(); 1188 return (ENOMEM); 1189 } 1190 1191 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1192 if (!strcmp(ifg->ifg_group, groupname)) 1193 break; 1194 1195 if (ifg == NULL) { 1196 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1197 M_TEMP, M_NOWAIT)) == NULL) { 1198 free(ifgl, M_TEMP); 1199 free(ifgm, M_TEMP); 1200 IFNET_WUNLOCK(); 1201 return (ENOMEM); 1202 } 1203 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1204 ifg->ifg_refcnt = 0; 1205 TAILQ_INIT(&ifg->ifg_members); 1206 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1207 new = 1; 1208 } 1209 1210 ifg->ifg_refcnt++; 1211 ifgl->ifgl_group = ifg; 1212 ifgm->ifgm_ifp = ifp; 1213 1214 IF_ADDR_WLOCK(ifp); 1215 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1216 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1217 IF_ADDR_WUNLOCK(ifp); 1218 1219 IFNET_WUNLOCK(); 1220 1221 if (new) 1222 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1223 EVENTHANDLER_INVOKE(group_change_event, groupname); 1224 1225 return (0); 1226} 1227 1228/* 1229 * Remove a group from an interface 1230 */ 1231int 1232if_delgroup(struct ifnet *ifp, const char *groupname) 1233{ 1234 struct ifg_list *ifgl; 1235 struct ifg_member *ifgm; 1236 1237 IFNET_WLOCK(); 1238 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1239 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1240 break; 1241 if (ifgl == NULL) { 1242 IFNET_WUNLOCK(); 1243 return (ENOENT); 1244 } 1245 1246 IF_ADDR_WLOCK(ifp); 1247 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1248 IF_ADDR_WUNLOCK(ifp); 1249 1250 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1251 if (ifgm->ifgm_ifp == ifp) 1252 break; 1253 1254 if (ifgm != NULL) { 1255 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 1256 free(ifgm, M_TEMP); 1257 } 1258 1259 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1260 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1261 IFNET_WUNLOCK(); 1262 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1263 free(ifgl->ifgl_group, M_TEMP); 1264 } else 1265 IFNET_WUNLOCK(); 1266 1267 free(ifgl, M_TEMP); 1268 1269 EVENTHANDLER_INVOKE(group_change_event, groupname); 1270 1271 return (0); 1272} 1273 1274/* 1275 * Remove an interface from all groups 1276 */ 1277static void 1278if_delgroups(struct ifnet *ifp) 1279{ 1280 struct ifg_list *ifgl; 1281 struct ifg_member *ifgm; 1282 char groupname[IFNAMSIZ]; 1283 1284 IFNET_WLOCK(); 1285 while (!TAILQ_EMPTY(&ifp->if_groups)) { 1286 ifgl = TAILQ_FIRST(&ifp->if_groups); 1287 1288 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1289 1290 IF_ADDR_WLOCK(ifp); 1291 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1292 IF_ADDR_WUNLOCK(ifp); 1293 1294 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1295 if (ifgm->ifgm_ifp == ifp) 1296 break; 1297 1298 if (ifgm != NULL) { 1299 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1300 ifgm_next); 1301 free(ifgm, M_TEMP); 1302 } 1303 1304 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1305 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1306 IFNET_WUNLOCK(); 1307 EVENTHANDLER_INVOKE(group_detach_event, 1308 ifgl->ifgl_group); 1309 free(ifgl->ifgl_group, M_TEMP); 1310 } else 1311 IFNET_WUNLOCK(); 1312 1313 free(ifgl, M_TEMP); 1314 1315 EVENTHANDLER_INVOKE(group_change_event, groupname); 1316 1317 IFNET_WLOCK(); 1318 } 1319 IFNET_WUNLOCK(); 1320} 1321 1322/* 1323 * Stores all groups from an interface in memory pointed 1324 * to by data 1325 */ 1326static int 1327if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 1328{ 1329 int len, error; 1330 struct ifg_list *ifgl; 1331 struct ifg_req ifgrq, *ifgp; 1332 struct ifgroupreq *ifgr = data; 1333 1334 if (ifgr->ifgr_len == 0) { 1335 IF_ADDR_RLOCK(ifp); 1336 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1337 ifgr->ifgr_len += sizeof(struct ifg_req); 1338 IF_ADDR_RUNLOCK(ifp); 1339 return (0); 1340 } 1341 1342 len = ifgr->ifgr_len; 1343 ifgp = ifgr->ifgr_groups; 1344 /* XXX: wire */ 1345 IF_ADDR_RLOCK(ifp); 1346 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1347 if (len < sizeof(ifgrq)) { 1348 IF_ADDR_RUNLOCK(ifp); 1349 return (EINVAL); 1350 } 1351 bzero(&ifgrq, sizeof ifgrq); 1352 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1353 sizeof(ifgrq.ifgrq_group)); 1354 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1355 IF_ADDR_RUNLOCK(ifp); 1356 return (error); 1357 } 1358 len -= sizeof(ifgrq); 1359 ifgp++; 1360 } 1361 IF_ADDR_RUNLOCK(ifp); 1362 1363 return (0); 1364} 1365 1366/* 1367 * Stores all members of a group in memory pointed to by data 1368 */ 1369static int 1370if_getgroupmembers(struct ifgroupreq *data) 1371{ 1372 struct ifgroupreq *ifgr = data; 1373 struct ifg_group *ifg; 1374 struct ifg_member *ifgm; 1375 struct ifg_req ifgrq, *ifgp; 1376 int len, error; 1377 1378 IFNET_RLOCK(); 1379 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1380 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1381 break; 1382 if (ifg == NULL) { 1383 IFNET_RUNLOCK(); 1384 return (ENOENT); 1385 } 1386 1387 if (ifgr->ifgr_len == 0) { 1388 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1389 ifgr->ifgr_len += sizeof(ifgrq); 1390 IFNET_RUNLOCK(); 1391 return (0); 1392 } 1393 1394 len = ifgr->ifgr_len; 1395 ifgp = ifgr->ifgr_groups; 1396 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1397 if (len < sizeof(ifgrq)) { 1398 IFNET_RUNLOCK(); 1399 return (EINVAL); 1400 } 1401 bzero(&ifgrq, sizeof ifgrq); 1402 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1403 sizeof(ifgrq.ifgrq_member)); 1404 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1405 IFNET_RUNLOCK(); 1406 return (error); 1407 } 1408 len -= sizeof(ifgrq); 1409 ifgp++; 1410 } 1411 IFNET_RUNLOCK(); 1412 1413 return (0); 1414} 1415 1416/* 1417 * Delete Routes for a Network Interface 1418 * 1419 * Called for each routing entry via the rnh->rnh_walktree() call above 1420 * to delete all route entries referencing a detaching network interface. 1421 * 1422 * Arguments: 1423 * rn pointer to node in the routing table 1424 * arg argument passed to rnh->rnh_walktree() - detaching interface 1425 * 1426 * Returns: 1427 * 0 successful 1428 * errno failed - reason indicated 1429 * 1430 */ 1431static int 1432if_rtdel(struct radix_node *rn, void *arg) 1433{ 1434 struct rtentry *rt = (struct rtentry *)rn; 1435 struct ifnet *ifp = arg; 1436 int err; 1437 1438 if (rt->rt_ifp == ifp) { 1439 1440 /* 1441 * Protect (sorta) against walktree recursion problems 1442 * with cloned routes 1443 */ 1444 if ((rt->rt_flags & RTF_UP) == 0) 1445 return (0); 1446 1447 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway, 1448 rt_mask(rt), 1449 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED, 1450 (struct rtentry **) NULL, rt->rt_fibnum); 1451 if (err) { 1452 log(LOG_WARNING, "if_rtdel: error %d\n", err); 1453 } 1454 } 1455 1456 return (0); 1457} 1458 1459/* 1460 * Wrapper functions for struct ifnet address list locking macros. These are 1461 * used by kernel modules to avoid encoding programming interface or binary 1462 * interface assumptions that may be violated when kernel-internal locking 1463 * approaches change. 1464 */ 1465void 1466if_addr_rlock(struct ifnet *ifp) 1467{ 1468 1469 IF_ADDR_RLOCK(ifp); 1470} 1471 1472void 1473if_addr_runlock(struct ifnet *ifp) 1474{ 1475 1476 IF_ADDR_RUNLOCK(ifp); 1477} 1478 1479void 1480if_maddr_rlock(struct ifnet *ifp) 1481{ 1482 1483 IF_ADDR_RLOCK(ifp); 1484} 1485 1486void 1487if_maddr_runlock(struct ifnet *ifp) 1488{ 1489 1490 IF_ADDR_RUNLOCK(ifp); 1491} 1492 1493/* 1494 * Initialization, destruction and refcounting functions for ifaddrs. 1495 */ 1496void 1497ifa_init(struct ifaddr *ifa) 1498{ 1499 1500 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF); 1501 refcount_init(&ifa->ifa_refcnt, 1); 1502 ifa->if_data.ifi_datalen = sizeof(ifa->if_data); 1503} 1504 1505void 1506ifa_ref(struct ifaddr *ifa) 1507{ 1508 1509 refcount_acquire(&ifa->ifa_refcnt); 1510} 1511 1512void 1513ifa_free(struct ifaddr *ifa) 1514{ 1515 1516 if (refcount_release(&ifa->ifa_refcnt)) { 1517 mtx_destroy(&ifa->ifa_mtx); 1518 free(ifa, M_IFADDR); 1519 } 1520} 1521 1522int 1523ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1524{ 1525 int error = 0; 1526 struct rtentry *rt = NULL; 1527 struct rt_addrinfo info; 1528 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1529 1530 bzero(&info, sizeof(info)); 1531 info.rti_ifp = V_loif; 1532 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; 1533 info.rti_info[RTAX_DST] = ia; 1534 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1535 error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib); 1536 1537 if (error == 0 && rt != NULL) { 1538 RT_LOCK(rt); 1539 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = 1540 ifa->ifa_ifp->if_type; 1541 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = 1542 ifa->ifa_ifp->if_index; 1543 RT_REMREF(rt); 1544 RT_UNLOCK(rt); 1545 } else if (error != 0) 1546 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error); 1547 1548 return (error); 1549} 1550 1551int 1552ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1553{ 1554 int error = 0; 1555 struct rt_addrinfo info; 1556 struct sockaddr_dl null_sdl; 1557 1558 bzero(&null_sdl, sizeof(null_sdl)); 1559 null_sdl.sdl_len = sizeof(null_sdl); 1560 null_sdl.sdl_family = AF_LINK; 1561 null_sdl.sdl_type = ifa->ifa_ifp->if_type; 1562 null_sdl.sdl_index = ifa->ifa_ifp->if_index; 1563 bzero(&info, sizeof(info)); 1564 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; 1565 info.rti_info[RTAX_DST] = ia; 1566 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1567 error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib); 1568 1569 if (error != 0) 1570 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error); 1571 1572 return (error); 1573} 1574 1575/* 1576 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1577 * structs used to represent other address families, it is necessary 1578 * to perform a different comparison. 1579 */ 1580 1581#define sa_equal(a1, a2) \ 1582 (bcmp((a1), (a2), ((a1))->sa_len) == 0) 1583 1584#define sa_dl_equal(a1, a2) \ 1585 ((((struct sockaddr_dl *)(a1))->sdl_len == \ 1586 ((struct sockaddr_dl *)(a2))->sdl_len) && \ 1587 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \ 1588 LLADDR((struct sockaddr_dl *)(a2)), \ 1589 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1590 1591/* 1592 * Locate an interface based on a complete address. 1593 */ 1594/*ARGSUSED*/ 1595static struct ifaddr * 1596ifa_ifwithaddr_internal(struct sockaddr *addr, int getref) 1597{ 1598 struct ifnet *ifp; 1599 struct ifaddr *ifa; 1600 1601 IFNET_RLOCK_NOSLEEP(); 1602 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1603 IF_ADDR_RLOCK(ifp); 1604 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1605 if (ifa->ifa_addr->sa_family != addr->sa_family) 1606 continue; 1607 if (sa_equal(addr, ifa->ifa_addr)) { 1608 if (getref) 1609 ifa_ref(ifa); 1610 IF_ADDR_RUNLOCK(ifp); 1611 goto done; 1612 } 1613 /* IP6 doesn't have broadcast */ 1614 if ((ifp->if_flags & IFF_BROADCAST) && 1615 ifa->ifa_broadaddr && 1616 ifa->ifa_broadaddr->sa_len != 0 && 1617 sa_equal(ifa->ifa_broadaddr, addr)) { 1618 if (getref) 1619 ifa_ref(ifa); 1620 IF_ADDR_RUNLOCK(ifp); 1621 goto done; 1622 } 1623 } 1624 IF_ADDR_RUNLOCK(ifp); 1625 } 1626 ifa = NULL; 1627done: 1628 IFNET_RUNLOCK_NOSLEEP(); 1629 return (ifa); 1630} 1631 1632struct ifaddr * 1633ifa_ifwithaddr(struct sockaddr *addr) 1634{ 1635 1636 return (ifa_ifwithaddr_internal(addr, 1)); 1637} 1638 1639int 1640ifa_ifwithaddr_check(struct sockaddr *addr) 1641{ 1642 1643 return (ifa_ifwithaddr_internal(addr, 0) != NULL); 1644} 1645 1646/* 1647 * Locate an interface based on the broadcast address. 1648 */ 1649/* ARGSUSED */ 1650struct ifaddr * 1651ifa_ifwithbroadaddr(struct sockaddr *addr) 1652{ 1653 struct ifnet *ifp; 1654 struct ifaddr *ifa; 1655 1656 IFNET_RLOCK_NOSLEEP(); 1657 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1658 IF_ADDR_RLOCK(ifp); 1659 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1660 if (ifa->ifa_addr->sa_family != addr->sa_family) 1661 continue; 1662 if ((ifp->if_flags & IFF_BROADCAST) && 1663 ifa->ifa_broadaddr && 1664 ifa->ifa_broadaddr->sa_len != 0 && 1665 sa_equal(ifa->ifa_broadaddr, addr)) { 1666 ifa_ref(ifa); 1667 IF_ADDR_RUNLOCK(ifp); 1668 goto done; 1669 } 1670 } 1671 IF_ADDR_RUNLOCK(ifp); 1672 } 1673 ifa = NULL; 1674done: 1675 IFNET_RUNLOCK_NOSLEEP(); 1676 return (ifa); 1677} 1678 1679/* 1680 * Locate the point to point interface with a given destination address. 1681 */ 1682/*ARGSUSED*/ 1683struct ifaddr * 1684ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum) 1685{ 1686 struct ifnet *ifp; 1687 struct ifaddr *ifa; 1688 1689 IFNET_RLOCK_NOSLEEP(); 1690 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1691 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1692 continue; 1693 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1694 continue; 1695 IF_ADDR_RLOCK(ifp); 1696 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1697 if (ifa->ifa_addr->sa_family != addr->sa_family) 1698 continue; 1699 if (ifa->ifa_dstaddr != NULL && 1700 sa_equal(addr, ifa->ifa_dstaddr)) { 1701 ifa_ref(ifa); 1702 IF_ADDR_RUNLOCK(ifp); 1703 goto done; 1704 } 1705 } 1706 IF_ADDR_RUNLOCK(ifp); 1707 } 1708 ifa = NULL; 1709done: 1710 IFNET_RUNLOCK_NOSLEEP(); 1711 return (ifa); 1712} 1713 1714struct ifaddr * 1715ifa_ifwithdstaddr(struct sockaddr *addr) 1716{ 1717 1718 return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS)); 1719} 1720 1721/* 1722 * Find an interface on a specific network. If many, choice 1723 * is most specific found. 1724 */ 1725struct ifaddr * 1726ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum) 1727{ 1728 struct ifnet *ifp; 1729 struct ifaddr *ifa; 1730 struct ifaddr *ifa_maybe = NULL; 1731 u_int af = addr->sa_family; 1732 char *addr_data = addr->sa_data, *cplim; 1733 1734 /* 1735 * AF_LINK addresses can be looked up directly by their index number, 1736 * so do that if we can. 1737 */ 1738 if (af == AF_LINK) { 1739 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1740 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1741 return (ifaddr_byindex(sdl->sdl_index)); 1742 } 1743 1744 /* 1745 * Scan though each interface, looking for ones that have addresses 1746 * in this address family and the requested fib. Maintain a reference 1747 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 1748 * kept it stable when we move onto the next interface. 1749 */ 1750 IFNET_RLOCK_NOSLEEP(); 1751 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1752 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1753 continue; 1754 IF_ADDR_RLOCK(ifp); 1755 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1756 char *cp, *cp2, *cp3; 1757 1758 if (ifa->ifa_addr->sa_family != af) 1759next: continue; 1760 if (af == AF_INET && 1761 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1762 /* 1763 * This is a bit broken as it doesn't 1764 * take into account that the remote end may 1765 * be a single node in the network we are 1766 * looking for. 1767 * The trouble is that we don't know the 1768 * netmask for the remote end. 1769 */ 1770 if (ifa->ifa_dstaddr != NULL && 1771 sa_equal(addr, ifa->ifa_dstaddr)) { 1772 ifa_ref(ifa); 1773 IF_ADDR_RUNLOCK(ifp); 1774 goto done; 1775 } 1776 } else { 1777 /* 1778 * if we have a special address handler, 1779 * then use it instead of the generic one. 1780 */ 1781 if (ifa->ifa_claim_addr) { 1782 if ((*ifa->ifa_claim_addr)(ifa, addr)) { 1783 ifa_ref(ifa); 1784 IF_ADDR_RUNLOCK(ifp); 1785 goto done; 1786 } 1787 continue; 1788 } 1789 1790 /* 1791 * Scan all the bits in the ifa's address. 1792 * If a bit dissagrees with what we are 1793 * looking for, mask it with the netmask 1794 * to see if it really matters. 1795 * (A byte at a time) 1796 */ 1797 if (ifa->ifa_netmask == 0) 1798 continue; 1799 cp = addr_data; 1800 cp2 = ifa->ifa_addr->sa_data; 1801 cp3 = ifa->ifa_netmask->sa_data; 1802 cplim = ifa->ifa_netmask->sa_len 1803 + (char *)ifa->ifa_netmask; 1804 while (cp3 < cplim) 1805 if ((*cp++ ^ *cp2++) & *cp3++) 1806 goto next; /* next address! */ 1807 /* 1808 * If the netmask of what we just found 1809 * is more specific than what we had before 1810 * (if we had one), or if the virtual status 1811 * of new prefix is better than of the old one, 1812 * then remember the new one before continuing 1813 * to search for an even better one. 1814 */ 1815 if (ifa_maybe == NULL || 1816 ifa_preferred(ifa_maybe, ifa) || 1817 rn_refines((caddr_t)ifa->ifa_netmask, 1818 (caddr_t)ifa_maybe->ifa_netmask)) { 1819 if (ifa_maybe != NULL) 1820 ifa_free(ifa_maybe); 1821 ifa_maybe = ifa; 1822 ifa_ref(ifa_maybe); 1823 } 1824 } 1825 } 1826 IF_ADDR_RUNLOCK(ifp); 1827 } 1828 ifa = ifa_maybe; 1829 ifa_maybe = NULL; 1830done: 1831 IFNET_RUNLOCK_NOSLEEP(); 1832 if (ifa_maybe != NULL) 1833 ifa_free(ifa_maybe); 1834 return (ifa); 1835} 1836 1837struct ifaddr * 1838ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp) 1839{ 1840 1841 return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS)); 1842} 1843 1844/* 1845 * Find an interface address specific to an interface best matching 1846 * a given address. 1847 */ 1848struct ifaddr * 1849ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1850{ 1851 struct ifaddr *ifa; 1852 char *cp, *cp2, *cp3; 1853 char *cplim; 1854 struct ifaddr *ifa_maybe = NULL; 1855 u_int af = addr->sa_family; 1856 1857 if (af >= AF_MAX) 1858 return (NULL); 1859 IF_ADDR_RLOCK(ifp); 1860 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1861 if (ifa->ifa_addr->sa_family != af) 1862 continue; 1863 if (ifa_maybe == NULL) 1864 ifa_maybe = ifa; 1865 if (ifa->ifa_netmask == 0) { 1866 if (sa_equal(addr, ifa->ifa_addr) || 1867 (ifa->ifa_dstaddr && 1868 sa_equal(addr, ifa->ifa_dstaddr))) 1869 goto done; 1870 continue; 1871 } 1872 if (ifp->if_flags & IFF_POINTOPOINT) { 1873 if (sa_equal(addr, ifa->ifa_dstaddr)) 1874 goto done; 1875 } else { 1876 cp = addr->sa_data; 1877 cp2 = ifa->ifa_addr->sa_data; 1878 cp3 = ifa->ifa_netmask->sa_data; 1879 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1880 for (; cp3 < cplim; cp3++) 1881 if ((*cp++ ^ *cp2++) & *cp3) 1882 break; 1883 if (cp3 == cplim) 1884 goto done; 1885 } 1886 } 1887 ifa = ifa_maybe; 1888done: 1889 if (ifa != NULL) 1890 ifa_ref(ifa); 1891 IF_ADDR_RUNLOCK(ifp); 1892 return (ifa); 1893} 1894 1895/* 1896 * See whether new ifa is better than current one: 1897 * 1) A non-virtual one is preferred over virtual. 1898 * 2) A virtual in master state preferred over any other state. 1899 * 1900 * Used in several address selecting functions. 1901 */ 1902int 1903ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 1904{ 1905 1906 return (cur->ifa_carp && (!next->ifa_carp || 1907 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 1908} 1909 1910#include <net/if_llatbl.h> 1911 1912/* 1913 * Default action when installing a route with a Link Level gateway. 1914 * Lookup an appropriate real ifa to point to. 1915 * This should be moved to /sys/net/link.c eventually. 1916 */ 1917static void 1918link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 1919{ 1920 struct ifaddr *ifa, *oifa; 1921 struct sockaddr *dst; 1922 struct ifnet *ifp; 1923 1924 RT_LOCK_ASSERT(rt); 1925 1926 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 1927 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 1928 return; 1929 ifa = ifaof_ifpforaddr(dst, ifp); 1930 if (ifa) { 1931 oifa = rt->rt_ifa; 1932 rt->rt_ifa = ifa; 1933 ifa_free(oifa); 1934 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1935 ifa->ifa_rtrequest(cmd, rt, info); 1936 } 1937} 1938 1939/* 1940 * Mark an interface down and notify protocols of 1941 * the transition. 1942 */ 1943static void 1944if_unroute(struct ifnet *ifp, int flag, int fam) 1945{ 1946 struct ifaddr *ifa; 1947 1948 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 1949 1950 ifp->if_flags &= ~flag; 1951 getmicrotime(&ifp->if_lastchange); 1952 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1953 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1954 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1955 ifp->if_qflush(ifp); 1956 1957 if (ifp->if_carp) 1958 (*carp_linkstate_p)(ifp); 1959 rt_ifmsg(ifp); 1960} 1961 1962/* 1963 * Mark an interface up and notify protocols of 1964 * the transition. 1965 */ 1966static void 1967if_route(struct ifnet *ifp, int flag, int fam) 1968{ 1969 struct ifaddr *ifa; 1970 1971 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 1972 1973 ifp->if_flags |= flag; 1974 getmicrotime(&ifp->if_lastchange); 1975 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1976 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 1977 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1978 if (ifp->if_carp) 1979 (*carp_linkstate_p)(ifp); 1980 rt_ifmsg(ifp); 1981#ifdef INET6 1982 in6_if_up(ifp); 1983#endif 1984} 1985 1986void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 1987void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1988struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 1989struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 1990int (*vlan_tag_p)(struct ifnet *, uint16_t *); 1991int (*vlan_setcookie_p)(struct ifnet *, void *); 1992void *(*vlan_cookie_p)(struct ifnet *); 1993 1994/* 1995 * Handle a change in the interface link state. To avoid LORs 1996 * between driver lock and upper layer locks, as well as possible 1997 * recursions, we post event to taskqueue, and all job 1998 * is done in static do_link_state_change(). 1999 */ 2000void 2001if_link_state_change(struct ifnet *ifp, int link_state) 2002{ 2003 /* Return if state hasn't changed. */ 2004 if (ifp->if_link_state == link_state) 2005 return; 2006 2007 ifp->if_link_state = link_state; 2008 2009 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2010} 2011 2012static void 2013do_link_state_change(void *arg, int pending) 2014{ 2015 struct ifnet *ifp = (struct ifnet *)arg; 2016 int link_state = ifp->if_link_state; 2017 CURVNET_SET(ifp->if_vnet); 2018 2019 /* Notify that the link state has changed. */ 2020 rt_ifmsg(ifp); 2021 if (ifp->if_vlantrunk != NULL) 2022 (*vlan_link_state_p)(ifp); 2023 2024 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2025 IFP2AC(ifp)->ac_netgraph != NULL) 2026 (*ng_ether_link_state_p)(ifp, link_state); 2027 if (ifp->if_carp) 2028 (*carp_linkstate_p)(ifp); 2029 if (ifp->if_bridge) 2030 (*bridge_linkstate_p)(ifp); 2031 if (ifp->if_lagg) 2032 (*lagg_linkstate_p)(ifp, link_state); 2033 2034 if (IS_DEFAULT_VNET(curvnet)) 2035 devctl_notify("IFNET", ifp->if_xname, 2036 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2037 NULL); 2038 if (pending > 1) 2039 if_printf(ifp, "%d link states coalesced\n", pending); 2040 if (log_link_state_change) 2041 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 2042 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2043 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state); 2044 CURVNET_RESTORE(); 2045} 2046 2047/* 2048 * Mark an interface down and notify protocols of 2049 * the transition. 2050 */ 2051void 2052if_down(struct ifnet *ifp) 2053{ 2054 2055 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2056} 2057 2058/* 2059 * Mark an interface up and notify protocols of 2060 * the transition. 2061 */ 2062void 2063if_up(struct ifnet *ifp) 2064{ 2065 2066 if_route(ifp, IFF_UP, AF_UNSPEC); 2067} 2068 2069/* 2070 * Flush an interface queue. 2071 */ 2072void 2073if_qflush(struct ifnet *ifp) 2074{ 2075 struct mbuf *m, *n; 2076 struct ifaltq *ifq; 2077 2078 ifq = &ifp->if_snd; 2079 IFQ_LOCK(ifq); 2080#ifdef ALTQ 2081 if (ALTQ_IS_ENABLED(ifq)) 2082 ALTQ_PURGE(ifq); 2083#endif 2084 n = ifq->ifq_head; 2085 while ((m = n) != 0) { 2086 n = m->m_nextpkt; 2087 m_freem(m); 2088 } 2089 ifq->ifq_head = 0; 2090 ifq->ifq_tail = 0; 2091 ifq->ifq_len = 0; 2092 IFQ_UNLOCK(ifq); 2093} 2094 2095/* 2096 * Map interface name to interface structure pointer, with or without 2097 * returning a reference. 2098 */ 2099struct ifnet * 2100ifunit_ref(const char *name) 2101{ 2102 struct ifnet *ifp; 2103 2104 IFNET_RLOCK_NOSLEEP(); 2105 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2106 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2107 !(ifp->if_flags & IFF_DYING)) 2108 break; 2109 } 2110 if (ifp != NULL) 2111 if_ref(ifp); 2112 IFNET_RUNLOCK_NOSLEEP(); 2113 return (ifp); 2114} 2115 2116struct ifnet * 2117ifunit(const char *name) 2118{ 2119 struct ifnet *ifp; 2120 2121 IFNET_RLOCK_NOSLEEP(); 2122 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2123 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2124 break; 2125 } 2126 IFNET_RUNLOCK_NOSLEEP(); 2127 return (ifp); 2128} 2129 2130/* 2131 * Hardware specific interface ioctls. 2132 */ 2133static int 2134ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2135{ 2136 struct ifreq *ifr; 2137 struct ifstat *ifs; 2138 int error = 0; 2139 int new_flags, temp_flags; 2140 size_t namelen, onamelen; 2141 size_t descrlen; 2142 char *descrbuf, *odescrbuf; 2143 char new_name[IFNAMSIZ]; 2144 struct ifaddr *ifa; 2145 struct sockaddr_dl *sdl; 2146 2147 ifr = (struct ifreq *)data; 2148 switch (cmd) { 2149 case SIOCGIFINDEX: 2150 ifr->ifr_index = ifp->if_index; 2151 break; 2152 2153 case SIOCGIFFLAGS: 2154 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2155 ifr->ifr_flags = temp_flags & 0xffff; 2156 ifr->ifr_flagshigh = temp_flags >> 16; 2157 break; 2158 2159 case SIOCGIFCAP: 2160 ifr->ifr_reqcap = ifp->if_capabilities; 2161 ifr->ifr_curcap = ifp->if_capenable; 2162 break; 2163 2164#ifdef MAC 2165 case SIOCGIFMAC: 2166 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2167 break; 2168#endif 2169 2170 case SIOCGIFMETRIC: 2171 ifr->ifr_metric = ifp->if_metric; 2172 break; 2173 2174 case SIOCGIFMTU: 2175 ifr->ifr_mtu = ifp->if_mtu; 2176 break; 2177 2178 case SIOCGIFPHYS: 2179 ifr->ifr_phys = ifp->if_physical; 2180 break; 2181 2182 case SIOCGIFDESCR: 2183 error = 0; 2184 sx_slock(&ifdescr_sx); 2185 if (ifp->if_description == NULL) 2186 error = ENOMSG; 2187 else { 2188 /* space for terminating nul */ 2189 descrlen = strlen(ifp->if_description) + 1; 2190 if (ifr->ifr_buffer.length < descrlen) 2191 ifr->ifr_buffer.buffer = NULL; 2192 else 2193 error = copyout(ifp->if_description, 2194 ifr->ifr_buffer.buffer, descrlen); 2195 ifr->ifr_buffer.length = descrlen; 2196 } 2197 sx_sunlock(&ifdescr_sx); 2198 break; 2199 2200 case SIOCSIFDESCR: 2201 error = priv_check(td, PRIV_NET_SETIFDESCR); 2202 if (error) 2203 return (error); 2204 2205 /* 2206 * Copy only (length-1) bytes to make sure that 2207 * if_description is always nul terminated. The 2208 * length parameter is supposed to count the 2209 * terminating nul in. 2210 */ 2211 if (ifr->ifr_buffer.length > ifdescr_maxlen) 2212 return (ENAMETOOLONG); 2213 else if (ifr->ifr_buffer.length == 0) 2214 descrbuf = NULL; 2215 else { 2216 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR, 2217 M_WAITOK | M_ZERO); 2218 error = copyin(ifr->ifr_buffer.buffer, descrbuf, 2219 ifr->ifr_buffer.length - 1); 2220 if (error) { 2221 free(descrbuf, M_IFDESCR); 2222 break; 2223 } 2224 } 2225 2226 sx_xlock(&ifdescr_sx); 2227 odescrbuf = ifp->if_description; 2228 ifp->if_description = descrbuf; 2229 sx_xunlock(&ifdescr_sx); 2230 2231 getmicrotime(&ifp->if_lastchange); 2232 free(odescrbuf, M_IFDESCR); 2233 break; 2234 2235 case SIOCGIFFIB: 2236 ifr->ifr_fib = ifp->if_fib; 2237 break; 2238 2239 case SIOCSIFFIB: 2240 error = priv_check(td, PRIV_NET_SETIFFIB); 2241 if (error) 2242 return (error); 2243 if (ifr->ifr_fib >= rt_numfibs) 2244 return (EINVAL); 2245 2246 ifp->if_fib = ifr->ifr_fib; 2247 break; 2248 2249 case SIOCSIFFLAGS: 2250 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2251 if (error) 2252 return (error); 2253 /* 2254 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2255 * check, so we don't need special handling here yet. 2256 */ 2257 new_flags = (ifr->ifr_flags & 0xffff) | 2258 (ifr->ifr_flagshigh << 16); 2259 if (ifp->if_flags & IFF_SMART) { 2260 /* Smart drivers twiddle their own routes */ 2261 } else if (ifp->if_flags & IFF_UP && 2262 (new_flags & IFF_UP) == 0) { 2263 if_down(ifp); 2264 } else if (new_flags & IFF_UP && 2265 (ifp->if_flags & IFF_UP) == 0) { 2266 if_up(ifp); 2267 } 2268 /* See if permanently promiscuous mode bit is about to flip */ 2269 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2270 if (new_flags & IFF_PPROMISC) 2271 ifp->if_flags |= IFF_PROMISC; 2272 else if (ifp->if_pcount == 0) 2273 ifp->if_flags &= ~IFF_PROMISC; 2274 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2275 ifp->if_xname, 2276 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); 2277 } 2278 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2279 (new_flags &~ IFF_CANTCHANGE); 2280 if (ifp->if_ioctl) { 2281 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2282 } 2283 getmicrotime(&ifp->if_lastchange); 2284 break; 2285 2286 case SIOCSIFCAP: 2287 error = priv_check(td, PRIV_NET_SETIFCAP); 2288 if (error) 2289 return (error); 2290 if (ifp->if_ioctl == NULL) 2291 return (EOPNOTSUPP); 2292 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2293 return (EINVAL); 2294 error = (*ifp->if_ioctl)(ifp, cmd, data); 2295 if (error == 0) 2296 getmicrotime(&ifp->if_lastchange); 2297 break; 2298 2299#ifdef MAC 2300 case SIOCSIFMAC: 2301 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2302 break; 2303#endif 2304 2305 case SIOCSIFNAME: 2306 error = priv_check(td, PRIV_NET_SETIFNAME); 2307 if (error) 2308 return (error); 2309 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 2310 if (error != 0) 2311 return (error); 2312 if (new_name[0] == '\0') 2313 return (EINVAL); 2314 if (ifunit(new_name) != NULL) 2315 return (EEXIST); 2316 2317 /* 2318 * XXX: Locking. Nothing else seems to lock if_flags, 2319 * and there are numerous other races with the 2320 * ifunit() checks not being atomic with namespace 2321 * changes (renames, vmoves, if_attach, etc). 2322 */ 2323 ifp->if_flags |= IFF_RENAMING; 2324 2325 /* Announce the departure of the interface. */ 2326 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2327 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2328 2329 log(LOG_INFO, "%s: changing name to '%s'\n", 2330 ifp->if_xname, new_name); 2331 2332 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2333 ifa = ifp->if_addr; 2334 IFA_LOCK(ifa); 2335 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2336 namelen = strlen(new_name); 2337 onamelen = sdl->sdl_nlen; 2338 /* 2339 * Move the address if needed. This is safe because we 2340 * allocate space for a name of length IFNAMSIZ when we 2341 * create this in if_attach(). 2342 */ 2343 if (namelen != onamelen) { 2344 bcopy(sdl->sdl_data + onamelen, 2345 sdl->sdl_data + namelen, sdl->sdl_alen); 2346 } 2347 bcopy(new_name, sdl->sdl_data, namelen); 2348 sdl->sdl_nlen = namelen; 2349 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2350 bzero(sdl->sdl_data, onamelen); 2351 while (namelen != 0) 2352 sdl->sdl_data[--namelen] = 0xff; 2353 IFA_UNLOCK(ifa); 2354 2355 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2356 /* Announce the return of the interface. */ 2357 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2358 2359 ifp->if_flags &= ~IFF_RENAMING; 2360 break; 2361 2362#ifdef VIMAGE 2363 case SIOCSIFVNET: 2364 error = priv_check(td, PRIV_NET_SETIFVNET); 2365 if (error) 2366 return (error); 2367 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2368 break; 2369#endif 2370 2371 case SIOCSIFMETRIC: 2372 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2373 if (error) 2374 return (error); 2375 ifp->if_metric = ifr->ifr_metric; 2376 getmicrotime(&ifp->if_lastchange); 2377 break; 2378 2379 case SIOCSIFPHYS: 2380 error = priv_check(td, PRIV_NET_SETIFPHYS); 2381 if (error) 2382 return (error); 2383 if (ifp->if_ioctl == NULL) 2384 return (EOPNOTSUPP); 2385 error = (*ifp->if_ioctl)(ifp, cmd, data); 2386 if (error == 0) 2387 getmicrotime(&ifp->if_lastchange); 2388 break; 2389 2390 case SIOCSIFMTU: 2391 { 2392 u_long oldmtu = ifp->if_mtu; 2393 2394 error = priv_check(td, PRIV_NET_SETIFMTU); 2395 if (error) 2396 return (error); 2397 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2398 return (EINVAL); 2399 if (ifp->if_ioctl == NULL) 2400 return (EOPNOTSUPP); 2401 error = (*ifp->if_ioctl)(ifp, cmd, data); 2402 if (error == 0) { 2403 getmicrotime(&ifp->if_lastchange); 2404 rt_ifmsg(ifp); 2405 } 2406 /* 2407 * If the link MTU changed, do network layer specific procedure. 2408 */ 2409 if (ifp->if_mtu != oldmtu) { 2410#ifdef INET6 2411 nd6_setmtu(ifp); 2412#endif 2413 } 2414 break; 2415 } 2416 2417 case SIOCADDMULTI: 2418 case SIOCDELMULTI: 2419 if (cmd == SIOCADDMULTI) 2420 error = priv_check(td, PRIV_NET_ADDMULTI); 2421 else 2422 error = priv_check(td, PRIV_NET_DELMULTI); 2423 if (error) 2424 return (error); 2425 2426 /* Don't allow group membership on non-multicast interfaces. */ 2427 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2428 return (EOPNOTSUPP); 2429 2430 /* Don't let users screw up protocols' entries. */ 2431 if (ifr->ifr_addr.sa_family != AF_LINK) 2432 return (EINVAL); 2433 2434 if (cmd == SIOCADDMULTI) { 2435 struct ifmultiaddr *ifma; 2436 2437 /* 2438 * Userland is only permitted to join groups once 2439 * via the if_addmulti() KPI, because it cannot hold 2440 * struct ifmultiaddr * between calls. It may also 2441 * lose a race while we check if the membership 2442 * already exists. 2443 */ 2444 IF_ADDR_RLOCK(ifp); 2445 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2446 IF_ADDR_RUNLOCK(ifp); 2447 if (ifma != NULL) 2448 error = EADDRINUSE; 2449 else 2450 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2451 } else { 2452 error = if_delmulti(ifp, &ifr->ifr_addr); 2453 } 2454 if (error == 0) 2455 getmicrotime(&ifp->if_lastchange); 2456 break; 2457 2458 case SIOCSIFPHYADDR: 2459 case SIOCDIFPHYADDR: 2460#ifdef INET6 2461 case SIOCSIFPHYADDR_IN6: 2462#endif 2463 case SIOCSLIFPHYADDR: 2464 case SIOCSIFMEDIA: 2465 case SIOCSIFGENERIC: 2466 error = priv_check(td, PRIV_NET_HWIOCTL); 2467 if (error) 2468 return (error); 2469 if (ifp->if_ioctl == NULL) 2470 return (EOPNOTSUPP); 2471 error = (*ifp->if_ioctl)(ifp, cmd, data); 2472 if (error == 0) 2473 getmicrotime(&ifp->if_lastchange); 2474 break; 2475 2476 case SIOCGIFSTATUS: 2477 ifs = (struct ifstat *)data; 2478 ifs->ascii[0] = '\0'; 2479 2480 case SIOCGIFPSRCADDR: 2481 case SIOCGIFPDSTADDR: 2482 case SIOCGLIFPHYADDR: 2483 case SIOCGIFMEDIA: 2484 case SIOCGIFGENERIC: 2485 if (ifp->if_ioctl == NULL) 2486 return (EOPNOTSUPP); 2487 error = (*ifp->if_ioctl)(ifp, cmd, data); 2488 break; 2489 2490 case SIOCSIFLLADDR: 2491 error = priv_check(td, PRIV_NET_SETLLADDR); 2492 if (error) 2493 return (error); 2494 error = if_setlladdr(ifp, 2495 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2496 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 2497 break; 2498 2499 case SIOCAIFGROUP: 2500 { 2501 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2502 2503 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2504 if (error) 2505 return (error); 2506 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2507 return (error); 2508 break; 2509 } 2510 2511 case SIOCGIFGROUP: 2512 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2513 return (error); 2514 break; 2515 2516 case SIOCDIFGROUP: 2517 { 2518 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2519 2520 error = priv_check(td, PRIV_NET_DELIFGROUP); 2521 if (error) 2522 return (error); 2523 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2524 return (error); 2525 break; 2526 } 2527 2528 default: 2529 error = ENOIOCTL; 2530 break; 2531 } 2532 return (error); 2533} 2534 2535#ifdef COMPAT_FREEBSD32 2536struct ifconf32 { 2537 int32_t ifc_len; 2538 union { 2539 uint32_t ifcu_buf; 2540 uint32_t ifcu_req; 2541 } ifc_ifcu; 2542}; 2543#define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2544#endif 2545 2546/* 2547 * Interface ioctls. 2548 */ 2549int 2550ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2551{ 2552 struct ifnet *ifp; 2553 struct ifreq *ifr; 2554 int error; 2555 int oif_flags; 2556 2557 CURVNET_SET(so->so_vnet); 2558 switch (cmd) { 2559 case SIOCGIFCONF: 2560 case OSIOCGIFCONF: 2561 error = ifconf(cmd, data); 2562 CURVNET_RESTORE(); 2563 return (error); 2564 2565#ifdef COMPAT_FREEBSD32 2566 case SIOCGIFCONF32: 2567 { 2568 struct ifconf32 *ifc32; 2569 struct ifconf ifc; 2570 2571 ifc32 = (struct ifconf32 *)data; 2572 ifc.ifc_len = ifc32->ifc_len; 2573 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2574 2575 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2576 CURVNET_RESTORE(); 2577 if (error == 0) 2578 ifc32->ifc_len = ifc.ifc_len; 2579 return (error); 2580 } 2581#endif 2582 } 2583 ifr = (struct ifreq *)data; 2584 2585 switch (cmd) { 2586#ifdef VIMAGE 2587 case SIOCSIFRVNET: 2588 error = priv_check(td, PRIV_NET_SETIFVNET); 2589 if (error == 0) 2590 error = if_vmove_reclaim(td, ifr->ifr_name, 2591 ifr->ifr_jid); 2592 CURVNET_RESTORE(); 2593 return (error); 2594#endif 2595 case SIOCIFCREATE: 2596 case SIOCIFCREATE2: 2597 error = priv_check(td, PRIV_NET_IFCREATE); 2598 if (error == 0) 2599 error = if_clone_create(ifr->ifr_name, 2600 sizeof(ifr->ifr_name), 2601 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL); 2602 CURVNET_RESTORE(); 2603 return (error); 2604 case SIOCIFDESTROY: 2605 error = priv_check(td, PRIV_NET_IFDESTROY); 2606 if (error == 0) 2607 error = if_clone_destroy(ifr->ifr_name); 2608 CURVNET_RESTORE(); 2609 return (error); 2610 2611 case SIOCIFGCLONERS: 2612 error = if_clone_list((struct if_clonereq *)data); 2613 CURVNET_RESTORE(); 2614 return (error); 2615 case SIOCGIFGMEMB: 2616 error = if_getgroupmembers((struct ifgroupreq *)data); 2617 CURVNET_RESTORE(); 2618 return (error); 2619#if defined(INET) || defined(INET6) 2620 case SIOCSVH: 2621 case SIOCGVH: 2622 if (carp_ioctl_p == NULL) 2623 error = EPROTONOSUPPORT; 2624 else 2625 error = (*carp_ioctl_p)(ifr, cmd, td); 2626 CURVNET_RESTORE(); 2627 return (error); 2628#endif 2629 } 2630 2631 ifp = ifunit_ref(ifr->ifr_name); 2632 if (ifp == NULL) { 2633 CURVNET_RESTORE(); 2634 return (ENXIO); 2635 } 2636 2637 error = ifhwioctl(cmd, ifp, data, td); 2638 if (error != ENOIOCTL) { 2639 if_rele(ifp); 2640 CURVNET_RESTORE(); 2641 return (error); 2642 } 2643 2644 oif_flags = ifp->if_flags; 2645 if (so->so_proto == NULL) { 2646 if_rele(ifp); 2647 CURVNET_RESTORE(); 2648 return (EOPNOTSUPP); 2649 } 2650 2651 /* 2652 * Pass the request on to the socket control method, and if the 2653 * latter returns EOPNOTSUPP, directly to the interface. 2654 * 2655 * Make an exception for the legacy SIOCSIF* requests. Drivers 2656 * trust SIOCSIFADDR et al to come from an already privileged 2657 * layer, and do not perform any credentials checks or input 2658 * validation. 2659 */ 2660#ifndef COMPAT_43 2661 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 2662 data, 2663 ifp, td)); 2664 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2665 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2666 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2667 error = (*ifp->if_ioctl)(ifp, cmd, data); 2668#else 2669 { 2670 u_long ocmd = cmd; 2671 2672 switch (cmd) { 2673 2674 case SIOCSIFDSTADDR: 2675 case SIOCSIFADDR: 2676 case SIOCSIFBRDADDR: 2677 case SIOCSIFNETMASK: 2678#if BYTE_ORDER != BIG_ENDIAN 2679 if (ifr->ifr_addr.sa_family == 0 && 2680 ifr->ifr_addr.sa_len < 16) { 2681 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 2682 ifr->ifr_addr.sa_len = 16; 2683 } 2684#else 2685 if (ifr->ifr_addr.sa_len == 0) 2686 ifr->ifr_addr.sa_len = 16; 2687#endif 2688 break; 2689 2690 case OSIOCGIFADDR: 2691 cmd = SIOCGIFADDR; 2692 break; 2693 2694 case OSIOCGIFDSTADDR: 2695 cmd = SIOCGIFDSTADDR; 2696 break; 2697 2698 case OSIOCGIFBRDADDR: 2699 cmd = SIOCGIFBRDADDR; 2700 break; 2701 2702 case OSIOCGIFNETMASK: 2703 cmd = SIOCGIFNETMASK; 2704 } 2705 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 2706 cmd, 2707 data, 2708 ifp, td)); 2709 if (error == EOPNOTSUPP && ifp != NULL && 2710 ifp->if_ioctl != NULL && 2711 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2712 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2713 error = (*ifp->if_ioctl)(ifp, cmd, data); 2714 switch (ocmd) { 2715 2716 case OSIOCGIFADDR: 2717 case OSIOCGIFDSTADDR: 2718 case OSIOCGIFBRDADDR: 2719 case OSIOCGIFNETMASK: 2720 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 2721 2722 } 2723 } 2724#endif /* COMPAT_43 */ 2725 2726 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2727#ifdef INET6 2728 if (ifp->if_flags & IFF_UP) 2729 in6_if_up(ifp); 2730#endif 2731 } 2732 if_rele(ifp); 2733 CURVNET_RESTORE(); 2734 return (error); 2735} 2736 2737/* 2738 * The code common to handling reference counted flags, 2739 * e.g., in ifpromisc() and if_allmulti(). 2740 * The "pflag" argument can specify a permanent mode flag to check, 2741 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2742 * 2743 * Only to be used on stack-owned flags, not driver-owned flags. 2744 */ 2745static int 2746if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 2747{ 2748 struct ifreq ifr; 2749 int error; 2750 int oldflags, oldcount; 2751 2752 /* Sanity checks to catch programming errors */ 2753 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 2754 ("%s: setting driver-owned flag %d", __func__, flag)); 2755 2756 if (onswitch) 2757 KASSERT(*refcount >= 0, 2758 ("%s: increment negative refcount %d for flag %d", 2759 __func__, *refcount, flag)); 2760 else 2761 KASSERT(*refcount > 0, 2762 ("%s: decrement non-positive refcount %d for flag %d", 2763 __func__, *refcount, flag)); 2764 2765 /* In case this mode is permanent, just touch refcount */ 2766 if (ifp->if_flags & pflag) { 2767 *refcount += onswitch ? 1 : -1; 2768 return (0); 2769 } 2770 2771 /* Save ifnet parameters for if_ioctl() may fail */ 2772 oldcount = *refcount; 2773 oldflags = ifp->if_flags; 2774 2775 /* 2776 * See if we aren't the only and touching refcount is enough. 2777 * Actually toggle interface flag if we are the first or last. 2778 */ 2779 if (onswitch) { 2780 if ((*refcount)++) 2781 return (0); 2782 ifp->if_flags |= flag; 2783 } else { 2784 if (--(*refcount)) 2785 return (0); 2786 ifp->if_flags &= ~flag; 2787 } 2788 2789 /* Call down the driver since we've changed interface flags */ 2790 if (ifp->if_ioctl == NULL) { 2791 error = EOPNOTSUPP; 2792 goto recover; 2793 } 2794 ifr.ifr_flags = ifp->if_flags & 0xffff; 2795 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2796 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2797 if (error) 2798 goto recover; 2799 /* Notify userland that interface flags have changed */ 2800 rt_ifmsg(ifp); 2801 return (0); 2802 2803recover: 2804 /* Recover after driver error */ 2805 *refcount = oldcount; 2806 ifp->if_flags = oldflags; 2807 return (error); 2808} 2809 2810/* 2811 * Set/clear promiscuous mode on interface ifp based on the truth value 2812 * of pswitch. The calls are reference counted so that only the first 2813 * "on" request actually has an effect, as does the final "off" request. 2814 * Results are undefined if the "off" and "on" requests are not matched. 2815 */ 2816int 2817ifpromisc(struct ifnet *ifp, int pswitch) 2818{ 2819 int error; 2820 int oldflags = ifp->if_flags; 2821 2822 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 2823 &ifp->if_pcount, pswitch); 2824 /* If promiscuous mode status has changed, log a message */ 2825 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) 2826 log(LOG_INFO, "%s: promiscuous mode %s\n", 2827 ifp->if_xname, 2828 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 2829 return (error); 2830} 2831 2832/* 2833 * Return interface configuration 2834 * of system. List may be used 2835 * in later ioctl's (above) to get 2836 * other information. 2837 */ 2838/*ARGSUSED*/ 2839static int 2840ifconf(u_long cmd, caddr_t data) 2841{ 2842 struct ifconf *ifc = (struct ifconf *)data; 2843 struct ifnet *ifp; 2844 struct ifaddr *ifa; 2845 struct ifreq ifr; 2846 struct sbuf *sb; 2847 int error, full = 0, valid_len, max_len; 2848 2849 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2850 max_len = MAXPHYS - 1; 2851 2852 /* Prevent hostile input from being able to crash the system */ 2853 if (ifc->ifc_len <= 0) 2854 return (EINVAL); 2855 2856again: 2857 if (ifc->ifc_len <= max_len) { 2858 max_len = ifc->ifc_len; 2859 full = 1; 2860 } 2861 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2862 max_len = 0; 2863 valid_len = 0; 2864 2865 IFNET_RLOCK(); 2866 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2867 int addrs; 2868 2869 /* 2870 * Zero the ifr_name buffer to make sure we don't 2871 * disclose the contents of the stack. 2872 */ 2873 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 2874 2875 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 2876 >= sizeof(ifr.ifr_name)) { 2877 sbuf_delete(sb); 2878 IFNET_RUNLOCK(); 2879 return (ENAMETOOLONG); 2880 } 2881 2882 addrs = 0; 2883 IF_ADDR_RLOCK(ifp); 2884 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2885 struct sockaddr *sa = ifa->ifa_addr; 2886 2887 if (prison_if(curthread->td_ucred, sa) != 0) 2888 continue; 2889 addrs++; 2890#ifdef COMPAT_43 2891 if (cmd == OSIOCGIFCONF) { 2892 struct osockaddr *osa = 2893 (struct osockaddr *)&ifr.ifr_addr; 2894 ifr.ifr_addr = *sa; 2895 osa->sa_family = sa->sa_family; 2896 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2897 max_len += sizeof(ifr); 2898 } else 2899#endif 2900 if (sa->sa_len <= sizeof(*sa)) { 2901 ifr.ifr_addr = *sa; 2902 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2903 max_len += sizeof(ifr); 2904 } else { 2905 sbuf_bcat(sb, &ifr, 2906 offsetof(struct ifreq, ifr_addr)); 2907 max_len += offsetof(struct ifreq, ifr_addr); 2908 sbuf_bcat(sb, sa, sa->sa_len); 2909 max_len += sa->sa_len; 2910 } 2911 2912 if (sbuf_error(sb) == 0) 2913 valid_len = sbuf_len(sb); 2914 } 2915 IF_ADDR_RUNLOCK(ifp); 2916 if (addrs == 0) { 2917 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 2918 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2919 max_len += sizeof(ifr); 2920 2921 if (sbuf_error(sb) == 0) 2922 valid_len = sbuf_len(sb); 2923 } 2924 } 2925 IFNET_RUNLOCK(); 2926 2927 /* 2928 * If we didn't allocate enough space (uncommon), try again. If 2929 * we have already allocated as much space as we are allowed, 2930 * return what we've got. 2931 */ 2932 if (valid_len != max_len && !full) { 2933 sbuf_delete(sb); 2934 goto again; 2935 } 2936 2937 ifc->ifc_len = valid_len; 2938 sbuf_finish(sb); 2939 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 2940 sbuf_delete(sb); 2941 return (error); 2942} 2943 2944/* 2945 * Just like ifpromisc(), but for all-multicast-reception mode. 2946 */ 2947int 2948if_allmulti(struct ifnet *ifp, int onswitch) 2949{ 2950 2951 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 2952} 2953 2954struct ifmultiaddr * 2955if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 2956{ 2957 struct ifmultiaddr *ifma; 2958 2959 IF_ADDR_LOCK_ASSERT(ifp); 2960 2961 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2962 if (sa->sa_family == AF_LINK) { 2963 if (sa_dl_equal(ifma->ifma_addr, sa)) 2964 break; 2965 } else { 2966 if (sa_equal(ifma->ifma_addr, sa)) 2967 break; 2968 } 2969 } 2970 2971 return ifma; 2972} 2973 2974/* 2975 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 2976 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 2977 * the ifnet multicast address list here, so the caller must do that and 2978 * other setup work (such as notifying the device driver). The reference 2979 * count is initialized to 1. 2980 */ 2981static struct ifmultiaddr * 2982if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 2983 int mflags) 2984{ 2985 struct ifmultiaddr *ifma; 2986 struct sockaddr *dupsa; 2987 2988 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 2989 M_ZERO); 2990 if (ifma == NULL) 2991 return (NULL); 2992 2993 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 2994 if (dupsa == NULL) { 2995 free(ifma, M_IFMADDR); 2996 return (NULL); 2997 } 2998 bcopy(sa, dupsa, sa->sa_len); 2999 ifma->ifma_addr = dupsa; 3000 3001 ifma->ifma_ifp = ifp; 3002 ifma->ifma_refcount = 1; 3003 ifma->ifma_protospec = NULL; 3004 3005 if (llsa == NULL) { 3006 ifma->ifma_lladdr = NULL; 3007 return (ifma); 3008 } 3009 3010 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3011 if (dupsa == NULL) { 3012 free(ifma->ifma_addr, M_IFMADDR); 3013 free(ifma, M_IFMADDR); 3014 return (NULL); 3015 } 3016 bcopy(llsa, dupsa, llsa->sa_len); 3017 ifma->ifma_lladdr = dupsa; 3018 3019 return (ifma); 3020} 3021 3022/* 3023 * if_freemulti: free ifmultiaddr structure and possibly attached related 3024 * addresses. The caller is responsible for implementing reference 3025 * counting, notifying the driver, handling routing messages, and releasing 3026 * any dependent link layer state. 3027 */ 3028static void 3029if_freemulti(struct ifmultiaddr *ifma) 3030{ 3031 3032 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3033 ifma->ifma_refcount)); 3034 KASSERT(ifma->ifma_protospec == NULL, 3035 ("if_freemulti: protospec not NULL")); 3036 3037 if (ifma->ifma_lladdr != NULL) 3038 free(ifma->ifma_lladdr, M_IFMADDR); 3039 free(ifma->ifma_addr, M_IFMADDR); 3040 free(ifma, M_IFMADDR); 3041} 3042 3043/* 3044 * Register an additional multicast address with a network interface. 3045 * 3046 * - If the address is already present, bump the reference count on the 3047 * address and return. 3048 * - If the address is not link-layer, look up a link layer address. 3049 * - Allocate address structures for one or both addresses, and attach to the 3050 * multicast address list on the interface. If automatically adding a link 3051 * layer address, the protocol address will own a reference to the link 3052 * layer address, to be freed when it is freed. 3053 * - Notify the network device driver of an addition to the multicast address 3054 * list. 3055 * 3056 * 'sa' points to caller-owned memory with the desired multicast address. 3057 * 3058 * 'retifma' will be used to return a pointer to the resulting multicast 3059 * address reference, if desired. 3060 */ 3061int 3062if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3063 struct ifmultiaddr **retifma) 3064{ 3065 struct ifmultiaddr *ifma, *ll_ifma; 3066 struct sockaddr *llsa; 3067 int error; 3068 3069 /* 3070 * If the address is already present, return a new reference to it; 3071 * otherwise, allocate storage and set up a new address. 3072 */ 3073 IF_ADDR_WLOCK(ifp); 3074 ifma = if_findmulti(ifp, sa); 3075 if (ifma != NULL) { 3076 ifma->ifma_refcount++; 3077 if (retifma != NULL) 3078 *retifma = ifma; 3079 IF_ADDR_WUNLOCK(ifp); 3080 return (0); 3081 } 3082 3083 /* 3084 * The address isn't already present; resolve the protocol address 3085 * into a link layer address, and then look that up, bump its 3086 * refcount or allocate an ifma for that also. If 'llsa' was 3087 * returned, we will need to free it later. 3088 */ 3089 llsa = NULL; 3090 ll_ifma = NULL; 3091 if (ifp->if_resolvemulti != NULL) { 3092 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3093 if (error) 3094 goto unlock_out; 3095 } 3096 3097 /* 3098 * Allocate the new address. Don't hook it up yet, as we may also 3099 * need to allocate a link layer multicast address. 3100 */ 3101 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3102 if (ifma == NULL) { 3103 error = ENOMEM; 3104 goto free_llsa_out; 3105 } 3106 3107 /* 3108 * If a link layer address is found, we'll need to see if it's 3109 * already present in the address list, or allocate is as well. 3110 * When this block finishes, the link layer address will be on the 3111 * list. 3112 */ 3113 if (llsa != NULL) { 3114 ll_ifma = if_findmulti(ifp, llsa); 3115 if (ll_ifma == NULL) { 3116 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3117 if (ll_ifma == NULL) { 3118 --ifma->ifma_refcount; 3119 if_freemulti(ifma); 3120 error = ENOMEM; 3121 goto free_llsa_out; 3122 } 3123 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3124 ifma_link); 3125 } else 3126 ll_ifma->ifma_refcount++; 3127 ifma->ifma_llifma = ll_ifma; 3128 } 3129 3130 /* 3131 * We now have a new multicast address, ifma, and possibly a new or 3132 * referenced link layer address. Add the primary address to the 3133 * ifnet address list. 3134 */ 3135 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3136 3137 if (retifma != NULL) 3138 *retifma = ifma; 3139 3140 /* 3141 * Must generate the message while holding the lock so that 'ifma' 3142 * pointer is still valid. 3143 */ 3144 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3145 IF_ADDR_WUNLOCK(ifp); 3146 3147 /* 3148 * We are certain we have added something, so call down to the 3149 * interface to let them know about it. 3150 */ 3151 if (ifp->if_ioctl != NULL) { 3152 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3153 } 3154 3155 if (llsa != NULL) 3156 free(llsa, M_IFMADDR); 3157 3158 return (0); 3159 3160free_llsa_out: 3161 if (llsa != NULL) 3162 free(llsa, M_IFMADDR); 3163 3164unlock_out: 3165 IF_ADDR_WUNLOCK(ifp); 3166 return (error); 3167} 3168 3169/* 3170 * Delete a multicast group membership by network-layer group address. 3171 * 3172 * Returns ENOENT if the entry could not be found. If ifp no longer 3173 * exists, results are undefined. This entry point should only be used 3174 * from subsystems which do appropriate locking to hold ifp for the 3175 * duration of the call. 3176 * Network-layer protocol domains must use if_delmulti_ifma(). 3177 */ 3178int 3179if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3180{ 3181 struct ifmultiaddr *ifma; 3182 int lastref; 3183#ifdef INVARIANTS 3184 struct ifnet *oifp; 3185 3186 IFNET_RLOCK_NOSLEEP(); 3187 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3188 if (ifp == oifp) 3189 break; 3190 if (ifp != oifp) 3191 ifp = NULL; 3192 IFNET_RUNLOCK_NOSLEEP(); 3193 3194 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3195#endif 3196 if (ifp == NULL) 3197 return (ENOENT); 3198 3199 IF_ADDR_WLOCK(ifp); 3200 lastref = 0; 3201 ifma = if_findmulti(ifp, sa); 3202 if (ifma != NULL) 3203 lastref = if_delmulti_locked(ifp, ifma, 0); 3204 IF_ADDR_WUNLOCK(ifp); 3205 3206 if (ifma == NULL) 3207 return (ENOENT); 3208 3209 if (lastref && ifp->if_ioctl != NULL) { 3210 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3211 } 3212 3213 return (0); 3214} 3215 3216/* 3217 * Delete all multicast group membership for an interface. 3218 * Should be used to quickly flush all multicast filters. 3219 */ 3220void 3221if_delallmulti(struct ifnet *ifp) 3222{ 3223 struct ifmultiaddr *ifma; 3224 struct ifmultiaddr *next; 3225 3226 IF_ADDR_WLOCK(ifp); 3227 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3228 if_delmulti_locked(ifp, ifma, 0); 3229 IF_ADDR_WUNLOCK(ifp); 3230} 3231 3232/* 3233 * Delete a multicast group membership by group membership pointer. 3234 * Network-layer protocol domains must use this routine. 3235 * 3236 * It is safe to call this routine if the ifp disappeared. 3237 */ 3238void 3239if_delmulti_ifma(struct ifmultiaddr *ifma) 3240{ 3241 struct ifnet *ifp; 3242 int lastref; 3243 3244 ifp = ifma->ifma_ifp; 3245#ifdef DIAGNOSTIC 3246 if (ifp == NULL) { 3247 printf("%s: ifma_ifp seems to be detached\n", __func__); 3248 } else { 3249 struct ifnet *oifp; 3250 3251 IFNET_RLOCK_NOSLEEP(); 3252 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3253 if (ifp == oifp) 3254 break; 3255 if (ifp != oifp) { 3256 printf("%s: ifnet %p disappeared\n", __func__, ifp); 3257 ifp = NULL; 3258 } 3259 IFNET_RUNLOCK_NOSLEEP(); 3260 } 3261#endif 3262 /* 3263 * If and only if the ifnet instance exists: Acquire the address lock. 3264 */ 3265 if (ifp != NULL) 3266 IF_ADDR_WLOCK(ifp); 3267 3268 lastref = if_delmulti_locked(ifp, ifma, 0); 3269 3270 if (ifp != NULL) { 3271 /* 3272 * If and only if the ifnet instance exists: 3273 * Release the address lock. 3274 * If the group was left: update the hardware hash filter. 3275 */ 3276 IF_ADDR_WUNLOCK(ifp); 3277 if (lastref && ifp->if_ioctl != NULL) { 3278 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3279 } 3280 } 3281} 3282 3283/* 3284 * Perform deletion of network-layer and/or link-layer multicast address. 3285 * 3286 * Return 0 if the reference count was decremented. 3287 * Return 1 if the final reference was released, indicating that the 3288 * hardware hash filter should be reprogrammed. 3289 */ 3290static int 3291if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3292{ 3293 struct ifmultiaddr *ll_ifma; 3294 3295 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3296 KASSERT(ifma->ifma_ifp == ifp, 3297 ("%s: inconsistent ifp %p", __func__, ifp)); 3298 IF_ADDR_WLOCK_ASSERT(ifp); 3299 } 3300 3301 ifp = ifma->ifma_ifp; 3302 3303 /* 3304 * If the ifnet is detaching, null out references to ifnet, 3305 * so that upper protocol layers will notice, and not attempt 3306 * to obtain locks for an ifnet which no longer exists. The 3307 * routing socket announcement must happen before the ifnet 3308 * instance is detached from the system. 3309 */ 3310 if (detaching) { 3311#ifdef DIAGNOSTIC 3312 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3313#endif 3314 /* 3315 * ifp may already be nulled out if we are being reentered 3316 * to delete the ll_ifma. 3317 */ 3318 if (ifp != NULL) { 3319 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3320 ifma->ifma_ifp = NULL; 3321 } 3322 } 3323 3324 if (--ifma->ifma_refcount > 0) 3325 return 0; 3326 3327 /* 3328 * If this ifma is a network-layer ifma, a link-layer ifma may 3329 * have been associated with it. Release it first if so. 3330 */ 3331 ll_ifma = ifma->ifma_llifma; 3332 if (ll_ifma != NULL) { 3333 KASSERT(ifma->ifma_lladdr != NULL, 3334 ("%s: llifma w/o lladdr", __func__)); 3335 if (detaching) 3336 ll_ifma->ifma_ifp = NULL; /* XXX */ 3337 if (--ll_ifma->ifma_refcount == 0) { 3338 if (ifp != NULL) { 3339 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3340 ifma_link); 3341 } 3342 if_freemulti(ll_ifma); 3343 } 3344 } 3345 3346 if (ifp != NULL) 3347 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3348 3349 if_freemulti(ifma); 3350 3351 /* 3352 * The last reference to this instance of struct ifmultiaddr 3353 * was released; the hardware should be notified of this change. 3354 */ 3355 return 1; 3356} 3357 3358/* 3359 * Set the link layer address on an interface. 3360 * 3361 * At this time we only support certain types of interfaces, 3362 * and we don't allow the length of the address to change. 3363 */ 3364int 3365if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3366{ 3367 struct sockaddr_dl *sdl; 3368 struct ifaddr *ifa; 3369 struct ifreq ifr; 3370 3371 IF_ADDR_RLOCK(ifp); 3372 ifa = ifp->if_addr; 3373 if (ifa == NULL) { 3374 IF_ADDR_RUNLOCK(ifp); 3375 return (EINVAL); 3376 } 3377 ifa_ref(ifa); 3378 IF_ADDR_RUNLOCK(ifp); 3379 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3380 if (sdl == NULL) { 3381 ifa_free(ifa); 3382 return (EINVAL); 3383 } 3384 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3385 ifa_free(ifa); 3386 return (EINVAL); 3387 } 3388 switch (ifp->if_type) { 3389 case IFT_ETHER: 3390 case IFT_FDDI: 3391 case IFT_XETHER: 3392 case IFT_ISO88025: 3393 case IFT_L2VLAN: 3394 case IFT_BRIDGE: 3395 case IFT_ARCNET: 3396 case IFT_IEEE8023ADLAG: 3397 case IFT_IEEE80211: 3398 bcopy(lladdr, LLADDR(sdl), len); 3399 ifa_free(ifa); 3400 break; 3401 default: 3402 ifa_free(ifa); 3403 return (ENODEV); 3404 } 3405 3406 /* 3407 * If the interface is already up, we need 3408 * to re-init it in order to reprogram its 3409 * address filter. 3410 */ 3411 if ((ifp->if_flags & IFF_UP) != 0) { 3412 if (ifp->if_ioctl) { 3413 ifp->if_flags &= ~IFF_UP; 3414 ifr.ifr_flags = ifp->if_flags & 0xffff; 3415 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3416 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3417 ifp->if_flags |= IFF_UP; 3418 ifr.ifr_flags = ifp->if_flags & 0xffff; 3419 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3420 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3421 } 3422#ifdef INET 3423 /* 3424 * Also send gratuitous ARPs to notify other nodes about 3425 * the address change. 3426 */ 3427 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3428 if (ifa->ifa_addr->sa_family == AF_INET) 3429 arp_ifinit(ifp, ifa); 3430 } 3431#endif 3432 } 3433 return (0); 3434} 3435 3436/* 3437 * The name argument must be a pointer to storage which will last as 3438 * long as the interface does. For physical devices, the result of 3439 * device_get_name(dev) is a good choice and for pseudo-devices a 3440 * static string works well. 3441 */ 3442void 3443if_initname(struct ifnet *ifp, const char *name, int unit) 3444{ 3445 ifp->if_dname = name; 3446 ifp->if_dunit = unit; 3447 if (unit != IF_DUNIT_NONE) 3448 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3449 else 3450 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3451} 3452 3453int 3454if_printf(struct ifnet *ifp, const char * fmt, ...) 3455{ 3456 va_list ap; 3457 int retval; 3458 3459 retval = printf("%s: ", ifp->if_xname); 3460 va_start(ap, fmt); 3461 retval += vprintf(fmt, ap); 3462 va_end(ap); 3463 return (retval); 3464} 3465 3466void 3467if_start(struct ifnet *ifp) 3468{ 3469 3470 (*(ifp)->if_start)(ifp); 3471} 3472 3473/* 3474 * Backwards compatibility interface for drivers 3475 * that have not implemented it 3476 */ 3477static int 3478if_transmit(struct ifnet *ifp, struct mbuf *m) 3479{ 3480 int error; 3481 3482 IFQ_HANDOFF(ifp, m, error); 3483 return (error); 3484} 3485 3486int 3487if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3488{ 3489 int active = 0; 3490 3491 IF_LOCK(ifq); 3492 if (_IF_QFULL(ifq)) { 3493 _IF_DROP(ifq); 3494 IF_UNLOCK(ifq); 3495 m_freem(m); 3496 return (0); 3497 } 3498 if (ifp != NULL) { 3499 ifp->if_obytes += m->m_pkthdr.len + adjust; 3500 if (m->m_flags & (M_BCAST|M_MCAST)) 3501 ifp->if_omcasts++; 3502 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3503 } 3504 _IF_ENQUEUE(ifq, m); 3505 IF_UNLOCK(ifq); 3506 if (ifp != NULL && !active) 3507 (*(ifp)->if_start)(ifp); 3508 return (1); 3509} 3510 3511void 3512if_register_com_alloc(u_char type, 3513 if_com_alloc_t *a, if_com_free_t *f) 3514{ 3515 3516 KASSERT(if_com_alloc[type] == NULL, 3517 ("if_register_com_alloc: %d already registered", type)); 3518 KASSERT(if_com_free[type] == NULL, 3519 ("if_register_com_alloc: %d free already registered", type)); 3520 3521 if_com_alloc[type] = a; 3522 if_com_free[type] = f; 3523} 3524 3525void 3526if_deregister_com_alloc(u_char type) 3527{ 3528 3529 KASSERT(if_com_alloc[type] != NULL, 3530 ("if_deregister_com_alloc: %d not registered", type)); 3531 KASSERT(if_com_free[type] != NULL, 3532 ("if_deregister_com_alloc: %d free not registered", type)); 3533 if_com_alloc[type] = NULL; 3534 if_com_free[type] = NULL; 3535} 3536