if.c revision 329913
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 329913 2018-02-24 16:50:57Z eugen $ 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/* Log promiscuous mode change events */ 117static int log_promisc_mode_change = 1; 118 119TUNABLE_INT("net.link.log_promisc_mode_change", &log_promisc_mode_change); 120SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 121 &log_promisc_mode_change, 1, 122 "log promiscuous mode change events"); 123 124/* Interface description */ 125static unsigned int ifdescr_maxlen = 1024; 126SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 127 &ifdescr_maxlen, 0, 128 "administrative maximum length for interface description"); 129 130static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 131 132/* global sx for non-critical path ifdescr */ 133static struct sx ifdescr_sx; 134SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 135 136void (*bridge_linkstate_p)(struct ifnet *ifp); 137void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 138void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 139/* These are external hooks for CARP. */ 140void (*carp_linkstate_p)(struct ifnet *ifp); 141void (*carp_demote_adj_p)(int, char *); 142int (*carp_master_p)(struct ifaddr *); 143#if defined(INET) || defined(INET6) 144int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 145int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 146 const struct sockaddr *sa); 147int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 148int (*carp_attach_p)(struct ifaddr *, int); 149void (*carp_detach_p)(struct ifaddr *); 150#endif 151#ifdef INET 152int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 153#endif 154#ifdef INET6 155struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 156caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 157 const struct in6_addr *taddr); 158#endif 159 160struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 161 162/* 163 * XXX: Style; these should be sorted alphabetically, and unprototyped 164 * static functions should be prototyped. Currently they are sorted by 165 * declaration order. 166 */ 167static void if_attachdomain(void *); 168static void if_attachdomain1(struct ifnet *); 169static int ifconf(u_long, caddr_t); 170static void if_freemulti(struct ifmultiaddr *); 171static void if_init(void *); 172static void if_grow(void); 173static void if_input_default(struct ifnet *, struct mbuf *); 174static void if_route(struct ifnet *, int flag, int fam); 175static int if_setflag(struct ifnet *, int, int, int *, int); 176static int if_transmit(struct ifnet *ifp, struct mbuf *m); 177static void if_unroute(struct ifnet *, int flag, int fam); 178static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 179static int if_rtdel(struct radix_node *, void *); 180static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 181static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 182static void do_link_state_change(void *, int); 183static int if_getgroup(struct ifgroupreq *, struct ifnet *); 184static int if_getgroupmembers(struct ifgroupreq *); 185static void if_delgroups(struct ifnet *); 186static void if_attach_internal(struct ifnet *, int, struct if_clone *); 187static int if_detach_internal(struct ifnet *, int, struct if_clone **); 188 189#ifdef INET6 190/* 191 * XXX: declare here to avoid to include many inet6 related files.. 192 * should be more generalized? 193 */ 194extern void nd6_setmtu(struct ifnet *); 195#endif 196 197VNET_DEFINE(int, if_index); 198int ifqmaxlen = IFQ_MAXLEN; 199VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 200VNET_DEFINE(struct ifgrouphead, ifg_head); 201 202static VNET_DEFINE(int, if_indexlim) = 8; 203 204/* Table of ifnet by index. */ 205VNET_DEFINE(struct ifindex_entry *, ifindex_table); 206 207#define V_if_indexlim VNET(if_indexlim) 208#define V_ifindex_table VNET(ifindex_table) 209 210/* 211 * The global network interface list (V_ifnet) and related state (such as 212 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 213 * an rwlock. Either may be acquired shared to stablize the list, but both 214 * must be acquired writable to modify the list. This model allows us to 215 * both stablize the interface list during interrupt thread processing, but 216 * also to stablize it over long-running ioctls, without introducing priority 217 * inversions and deadlocks. 218 */ 219struct rwlock ifnet_rwlock; 220struct sx ifnet_sxlock; 221 222/* 223 * The allocation of network interfaces is a rather non-atomic affair; we 224 * need to select an index before we are ready to expose the interface for 225 * use, so will use this pointer value to indicate reservation. 226 */ 227#define IFNET_HOLD (void *)(uintptr_t)(-1) 228 229static if_com_alloc_t *if_com_alloc[256]; 230static if_com_free_t *if_com_free[256]; 231 232static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 233MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 234MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 235 236struct ifnet * 237ifnet_byindex_locked(u_short idx) 238{ 239 240 if (idx > V_if_index) 241 return (NULL); 242 if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD) 243 return (NULL); 244 return (V_ifindex_table[idx].ife_ifnet); 245} 246 247struct ifnet * 248ifnet_byindex(u_short idx) 249{ 250 struct ifnet *ifp; 251 252 IFNET_RLOCK_NOSLEEP(); 253 ifp = ifnet_byindex_locked(idx); 254 IFNET_RUNLOCK_NOSLEEP(); 255 return (ifp); 256} 257 258struct ifnet * 259ifnet_byindex_ref(u_short idx) 260{ 261 struct ifnet *ifp; 262 263 IFNET_RLOCK_NOSLEEP(); 264 ifp = ifnet_byindex_locked(idx); 265 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 266 IFNET_RUNLOCK_NOSLEEP(); 267 return (NULL); 268 } 269 if_ref(ifp); 270 IFNET_RUNLOCK_NOSLEEP(); 271 return (ifp); 272} 273 274/* 275 * Allocate an ifindex array entry; return 0 on success or an error on 276 * failure. 277 */ 278static int 279ifindex_alloc_locked(u_short *idxp) 280{ 281 u_short idx; 282 283 IFNET_WLOCK_ASSERT(); 284 285retry: 286 /* 287 * Try to find an empty slot below V_if_index. If we fail, take the 288 * next slot. 289 */ 290 for (idx = 1; idx <= V_if_index; idx++) { 291 if (V_ifindex_table[idx].ife_ifnet == NULL) 292 break; 293 } 294 295 /* Catch if_index overflow. */ 296 if (idx < 1) 297 return (ENOSPC); 298 if (idx >= V_if_indexlim) { 299 if_grow(); 300 goto retry; 301 } 302 if (idx > V_if_index) 303 V_if_index = idx; 304 *idxp = idx; 305 return (0); 306} 307 308static void 309ifindex_free_locked(u_short idx) 310{ 311 312 IFNET_WLOCK_ASSERT(); 313 314 V_ifindex_table[idx].ife_ifnet = NULL; 315 while (V_if_index > 0 && 316 V_ifindex_table[V_if_index].ife_ifnet == NULL) 317 V_if_index--; 318} 319 320static void 321ifindex_free(u_short idx) 322{ 323 324 IFNET_WLOCK(); 325 ifindex_free_locked(idx); 326 IFNET_WUNLOCK(); 327} 328 329static void 330ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) 331{ 332 333 IFNET_WLOCK_ASSERT(); 334 335 V_ifindex_table[idx].ife_ifnet = ifp; 336} 337 338static void 339ifnet_setbyindex(u_short idx, struct ifnet *ifp) 340{ 341 342 IFNET_WLOCK(); 343 ifnet_setbyindex_locked(idx, ifp); 344 IFNET_WUNLOCK(); 345} 346 347struct ifaddr * 348ifaddr_byindex(u_short idx) 349{ 350 struct ifnet *ifp; 351 struct ifaddr *ifa = NULL; 352 353 IFNET_RLOCK_NOSLEEP(); 354 ifp = ifnet_byindex_locked(idx); 355 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 356 ifa_ref(ifa); 357 IFNET_RUNLOCK_NOSLEEP(); 358 return (ifa); 359} 360 361/* 362 * Network interface utility routines. 363 * 364 * Routines with ifa_ifwith* names take sockaddr *'s as 365 * parameters. 366 */ 367 368static void 369vnet_if_init(const void *unused __unused) 370{ 371 372 TAILQ_INIT(&V_ifnet); 373 TAILQ_INIT(&V_ifg_head); 374 IFNET_WLOCK(); 375 if_grow(); /* create initial table */ 376 IFNET_WUNLOCK(); 377 vnet_if_clone_init(); 378} 379VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 380 NULL); 381 382/* ARGSUSED*/ 383static void 384if_init(void *dummy __unused) 385{ 386 387 IFNET_LOCK_INIT(); 388 if_clone_init(); 389} 390SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL); 391 392 393#ifdef VIMAGE 394static void 395vnet_if_uninit(const void *unused __unused) 396{ 397 398 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 399 "not empty", __func__, __LINE__, &V_ifnet)); 400 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 401 "not empty", __func__, __LINE__, &V_ifg_head)); 402 403 free((caddr_t)V_ifindex_table, M_IFNET); 404} 405VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 406 vnet_if_uninit, NULL); 407#endif 408 409static void 410if_grow(void) 411{ 412 int oldlim; 413 u_int n; 414 struct ifindex_entry *e; 415 416 IFNET_WLOCK_ASSERT(); 417 oldlim = V_if_indexlim; 418 IFNET_WUNLOCK(); 419 n = (oldlim << 1) * sizeof(*e); 420 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 421 IFNET_WLOCK(); 422 if (V_if_indexlim != oldlim) { 423 free(e, M_IFNET); 424 return; 425 } 426 if (V_ifindex_table != NULL) { 427 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 428 free((caddr_t)V_ifindex_table, M_IFNET); 429 } 430 V_if_indexlim <<= 1; 431 V_ifindex_table = e; 432} 433 434/* 435 * Allocate a struct ifnet and an index for an interface. A layer 2 436 * common structure will also be allocated if an allocation routine is 437 * registered for the passed type. 438 */ 439struct ifnet * 440if_alloc(u_char type) 441{ 442 struct ifnet *ifp; 443 u_short idx; 444 445 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 446 IFNET_WLOCK(); 447 if (ifindex_alloc_locked(&idx) != 0) { 448 IFNET_WUNLOCK(); 449 free(ifp, M_IFNET); 450 return (NULL); 451 } 452 ifnet_setbyindex_locked(idx, IFNET_HOLD); 453 IFNET_WUNLOCK(); 454 ifp->if_index = idx; 455 ifp->if_type = type; 456 ifp->if_alloctype = type; 457 if (if_com_alloc[type] != NULL) { 458 ifp->if_l2com = if_com_alloc[type](type, ifp); 459 if (ifp->if_l2com == NULL) { 460 free(ifp, M_IFNET); 461 ifindex_free(idx); 462 return (NULL); 463 } 464 } 465 466 IF_ADDR_LOCK_INIT(ifp); 467 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 468 ifp->if_afdata_initialized = 0; 469 IF_AFDATA_LOCK_INIT(ifp); 470 TAILQ_INIT(&ifp->if_addrhead); 471 TAILQ_INIT(&ifp->if_multiaddrs); 472 TAILQ_INIT(&ifp->if_groups); 473#ifdef MAC 474 mac_ifnet_init(ifp); 475#endif 476 ifq_init(&ifp->if_snd, ifp); 477 478 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 479 ifnet_setbyindex(ifp->if_index, ifp); 480 return (ifp); 481} 482 483/* 484 * Do the actual work of freeing a struct ifnet, and layer 2 common 485 * structure. This call is made when the last reference to an 486 * interface is released. 487 */ 488static void 489if_free_internal(struct ifnet *ifp) 490{ 491 492 KASSERT((ifp->if_flags & IFF_DYING), 493 ("if_free_internal: interface not dying")); 494 495 if (if_com_free[ifp->if_alloctype] != NULL) 496 if_com_free[ifp->if_alloctype](ifp->if_l2com, 497 ifp->if_alloctype); 498 499#ifdef MAC 500 mac_ifnet_destroy(ifp); 501#endif /* MAC */ 502 if (ifp->if_description != NULL) 503 free(ifp->if_description, M_IFDESCR); 504 IF_AFDATA_DESTROY(ifp); 505 IF_ADDR_LOCK_DESTROY(ifp); 506 ifq_delete(&ifp->if_snd); 507 free(ifp, M_IFNET); 508} 509 510/* 511 * Deregister an interface and free the associated storage. 512 */ 513void 514if_free(struct ifnet *ifp) 515{ 516 517 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 518 519 CURVNET_SET_QUIET(ifp->if_vnet); 520 IFNET_WLOCK(); 521 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 522 ("%s: freeing unallocated ifnet", ifp->if_xname)); 523 524 ifindex_free_locked(ifp->if_index); 525 IFNET_WUNLOCK(); 526 527 if (refcount_release(&ifp->if_refcount)) 528 if_free_internal(ifp); 529 CURVNET_RESTORE(); 530} 531 532/* 533 * Interfaces to keep an ifnet type-stable despite the possibility of the 534 * driver calling if_free(). If there are additional references, we defer 535 * freeing the underlying data structure. 536 */ 537void 538if_ref(struct ifnet *ifp) 539{ 540 541 /* We don't assert the ifnet list lock here, but arguably should. */ 542 refcount_acquire(&ifp->if_refcount); 543} 544 545void 546if_rele(struct ifnet *ifp) 547{ 548 549 if (!refcount_release(&ifp->if_refcount)) 550 return; 551 if_free_internal(ifp); 552} 553 554void 555ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 556{ 557 558 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 559 560 if (ifq->ifq_maxlen == 0) 561 ifq->ifq_maxlen = ifqmaxlen; 562 563 ifq->altq_type = 0; 564 ifq->altq_disc = NULL; 565 ifq->altq_flags &= ALTQF_CANTCHANGE; 566 ifq->altq_tbr = NULL; 567 ifq->altq_ifp = ifp; 568} 569 570void 571ifq_delete(struct ifaltq *ifq) 572{ 573 mtx_destroy(&ifq->ifq_mtx); 574} 575 576/* 577 * Perform generic interface initalization tasks and attach the interface 578 * to the list of "active" interfaces. If vmove flag is set on entry 579 * to if_attach_internal(), perform only a limited subset of initialization 580 * tasks, given that we are moving from one vnet to another an ifnet which 581 * has already been fully initialized. 582 * 583 * Note that if_detach_internal() removes group membership unconditionally 584 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 585 * Thus, when if_vmove() is applied to a cloned interface, group membership 586 * is lost while a cloned one always joins a group whose name is 587 * ifc->ifc_name. To recover this after if_detach_internal() and 588 * if_attach_internal(), the cloner should be specified to 589 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 590 * attempts to join a group whose name is ifc->ifc_name. 591 * 592 * XXX: 593 * - The decision to return void and thus require this function to 594 * succeed is questionable. 595 * - We should probably do more sanity checking. For instance we don't 596 * do anything to insure if_xname is unique or non-empty. 597 */ 598void 599if_attach(struct ifnet *ifp) 600{ 601 602 if_attach_internal(ifp, 0, NULL); 603} 604 605/* 606 * Compute the least common TSO limit. 607 */ 608void 609if_hw_tsomax_common(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax) 610{ 611 /* 612 * 1) If there is no limit currently, take the limit from 613 * the network adapter. 614 * 615 * 2) If the network adapter has a limit below the current 616 * limit, apply it. 617 */ 618 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 619 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 620 pmax->tsomaxbytes = ifp->if_hw_tsomax; 621 } 622 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 623 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 624 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 625 } 626 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 627 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 628 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 629 } 630} 631 632/* 633 * Update TSO limit of a network adapter. 634 * 635 * Returns zero if no change. Else non-zero. 636 */ 637int 638if_hw_tsomax_update(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax) 639{ 640 int retval = 0; 641 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 642 ifp->if_hw_tsomax = pmax->tsomaxbytes; 643 retval++; 644 } 645 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 646 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 647 retval++; 648 } 649 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 650 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 651 retval++; 652 } 653 return (retval); 654} 655 656static void 657if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 658{ 659 unsigned socksize, ifasize; 660 int namelen, masklen; 661 struct sockaddr_dl *sdl; 662 struct ifaddr *ifa; 663 664 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 665 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 666 ifp->if_xname); 667 668#ifdef VIMAGE 669 ifp->if_vnet = curvnet; 670 if (ifp->if_home_vnet == NULL) 671 ifp->if_home_vnet = curvnet; 672#endif 673 674 if_addgroup(ifp, IFG_ALL); 675 676 /* Restore group membership for cloned interfaces. */ 677 if (vmove && ifc != NULL) 678 if_clone_addgroup(ifp, ifc); 679 680 getmicrotime(&ifp->if_lastchange); 681 ifp->if_data.ifi_epoch = time_uptime; 682 ifp->if_data.ifi_datalen = sizeof(struct if_data); 683 684 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 685 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 686 ("transmit and qflush must both either be set or both be NULL")); 687 if (ifp->if_transmit == NULL) { 688 ifp->if_transmit = if_transmit; 689 ifp->if_qflush = if_qflush; 690 } 691 if (ifp->if_input == NULL) 692 ifp->if_input = if_input_default; 693 694 if (!vmove) { 695#ifdef MAC 696 mac_ifnet_create(ifp); 697#endif 698 699 /* 700 * Create a Link Level name for this device. 701 */ 702 namelen = strlen(ifp->if_xname); 703 /* 704 * Always save enough space for any possiable name so we 705 * can do a rename in place later. 706 */ 707 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 708 socksize = masklen + ifp->if_addrlen; 709 if (socksize < sizeof(*sdl)) 710 socksize = sizeof(*sdl); 711 socksize = roundup2(socksize, sizeof(long)); 712 ifasize = sizeof(*ifa) + 2 * socksize; 713 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 714 ifa_init(ifa); 715 sdl = (struct sockaddr_dl *)(ifa + 1); 716 sdl->sdl_len = socksize; 717 sdl->sdl_family = AF_LINK; 718 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 719 sdl->sdl_nlen = namelen; 720 sdl->sdl_index = ifp->if_index; 721 sdl->sdl_type = ifp->if_type; 722 ifp->if_addr = ifa; 723 ifa->ifa_ifp = ifp; 724 ifa->ifa_rtrequest = link_rtrequest; 725 ifa->ifa_addr = (struct sockaddr *)sdl; 726 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 727 ifa->ifa_netmask = (struct sockaddr *)sdl; 728 sdl->sdl_len = masklen; 729 while (namelen != 0) 730 sdl->sdl_data[--namelen] = 0xff; 731 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 732 /* Reliably crash if used uninitialized. */ 733 ifp->if_broadcastaddr = NULL; 734 735 if (ifp->if_type == IFT_ETHER) { 736 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 737 M_WAITOK | M_ZERO); 738 } 739 740#if defined(INET) || defined(INET6) 741 /* Use defaults for TSO, if nothing is set */ 742 if (ifp->if_hw_tsomax == 0 && 743 ifp->if_hw_tsomaxsegcount == 0 && 744 ifp->if_hw_tsomaxsegsize == 0) { 745 /* 746 * The TSO defaults needs to be such that an 747 * NFS mbuf list of 35 mbufs totalling just 748 * below 64K works and that a chain of mbufs 749 * can be defragged into at most 32 segments: 750 */ 751 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 752 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 753 ifp->if_hw_tsomaxsegcount = 35; 754 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 755 756 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 757 if (ifp->if_capabilities & IFCAP_TSO) { 758 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 759 ifp->if_hw_tsomax, 760 ifp->if_hw_tsomaxsegcount, 761 ifp->if_hw_tsomaxsegsize); 762 } 763 } 764#endif 765 } 766#ifdef VIMAGE 767 else { 768 /* 769 * Update the interface index in the link layer address 770 * of the interface. 771 */ 772 for (ifa = ifp->if_addr; ifa != NULL; 773 ifa = TAILQ_NEXT(ifa, ifa_link)) { 774 if (ifa->ifa_addr->sa_family == AF_LINK) { 775 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 776 sdl->sdl_index = ifp->if_index; 777 } 778 } 779 } 780#endif 781 782 IFNET_WLOCK(); 783 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 784#ifdef VIMAGE 785 curvnet->vnet_ifcnt++; 786#endif 787 IFNET_WUNLOCK(); 788 789 if (domain_init_status >= 2) 790 if_attachdomain1(ifp); 791 792 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 793 if (IS_DEFAULT_VNET(curvnet)) 794 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 795 796 /* Announce the interface. */ 797 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 798} 799 800static void 801if_attachdomain(void *dummy) 802{ 803 struct ifnet *ifp; 804 805 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 806 if_attachdomain1(ifp); 807} 808SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 809 if_attachdomain, NULL); 810 811static void 812if_attachdomain1(struct ifnet *ifp) 813{ 814 struct domain *dp; 815 816 /* 817 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 818 * cannot lock ifp->if_afdata initialization, entirely. 819 */ 820 if (IF_AFDATA_TRYLOCK(ifp) == 0) 821 return; 822 if (ifp->if_afdata_initialized >= domain_init_status) { 823 IF_AFDATA_UNLOCK(ifp); 824 log(LOG_WARNING, "%s called more than once on %s\n", 825 __func__, ifp->if_xname); 826 return; 827 } 828 ifp->if_afdata_initialized = domain_init_status; 829 IF_AFDATA_UNLOCK(ifp); 830 831 /* address family dependent data region */ 832 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 833 for (dp = domains; dp; dp = dp->dom_next) { 834 if (dp->dom_ifattach) 835 ifp->if_afdata[dp->dom_family] = 836 (*dp->dom_ifattach)(ifp); 837 } 838} 839 840/* 841 * Remove any unicast or broadcast network addresses from an interface. 842 */ 843void 844if_purgeaddrs(struct ifnet *ifp) 845{ 846 struct ifaddr *ifa, *next; 847 848 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 849 if (ifa->ifa_addr->sa_family == AF_LINK) 850 continue; 851#ifdef INET 852 /* XXX: Ugly!! ad hoc just for INET */ 853 if (ifa->ifa_addr->sa_family == AF_INET) { 854 struct ifaliasreq ifr; 855 856 bzero(&ifr, sizeof(ifr)); 857 ifr.ifra_addr = *ifa->ifa_addr; 858 if (ifa->ifa_dstaddr) 859 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 860 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 861 NULL) == 0) 862 continue; 863 } 864#endif /* INET */ 865#ifdef INET6 866 if (ifa->ifa_addr->sa_family == AF_INET6) { 867 in6_purgeaddr(ifa); 868 /* ifp_addrhead is already updated */ 869 continue; 870 } 871#endif /* INET6 */ 872 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 873 ifa_free(ifa); 874 } 875} 876 877/* 878 * Remove any multicast network addresses from an interface when an ifnet 879 * is going away. 880 */ 881static void 882if_purgemaddrs(struct ifnet *ifp) 883{ 884 struct ifmultiaddr *ifma; 885 struct ifmultiaddr *next; 886 887 IF_ADDR_WLOCK(ifp); 888 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 889 if_delmulti_locked(ifp, ifma, 1); 890 IF_ADDR_WUNLOCK(ifp); 891} 892 893/* 894 * Detach an interface, removing it from the list of "active" interfaces. 895 * If vmove flag is set on entry to if_detach_internal(), perform only a 896 * limited subset of cleanup tasks, given that we are moving an ifnet from 897 * one vnet to another, where it must be fully operational. 898 * 899 * XXXRW: There are some significant questions about event ordering, and 900 * how to prevent things from starting to use the interface during detach. 901 */ 902void 903if_detach(struct ifnet *ifp) 904{ 905 906 CURVNET_SET_QUIET(ifp->if_vnet); 907 if_detach_internal(ifp, 0, NULL); 908 CURVNET_RESTORE(); 909} 910 911static int 912if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 913{ 914 struct ifaddr *ifa; 915 struct radix_node_head *rnh; 916 int i, j; 917 struct domain *dp; 918 struct ifnet *iter; 919 int found = 0; 920 921 IFNET_WLOCK(); 922 TAILQ_FOREACH(iter, &V_ifnet, if_link) 923 if (iter == ifp) { 924 TAILQ_REMOVE(&V_ifnet, ifp, if_link); 925 found = 1; 926 break; 927 } 928#ifdef VIMAGE 929 if (found) 930 curvnet->vnet_ifcnt--; 931#endif 932 IFNET_WUNLOCK(); 933 if (!found) { 934 /* 935 * While we would want to panic here, we cannot 936 * guarantee that the interface is indeed still on 937 * the list given we don't hold locks all the way. 938 */ 939 return (ENOENT); 940#if 0 941 if (vmove) 942 panic("%s: ifp=%p not on the ifnet tailq %p", 943 __func__, ifp, &V_ifnet); 944 else 945 return; /* XXX this should panic as well? */ 946#endif 947 } 948 949 /* Check if this is a cloned interface or not. */ 950 if (vmove && ifcp != NULL) 951 *ifcp = if_clone_findifc(ifp); 952 953 /* 954 * Remove/wait for pending events. 955 */ 956 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 957 958 /* 959 * Remove routes and flush queues. 960 */ 961 if_down(ifp); 962#ifdef ALTQ 963 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 964 altq_disable(&ifp->if_snd); 965 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 966 altq_detach(&ifp->if_snd); 967#endif 968 969 if_purgeaddrs(ifp); 970 971#ifdef INET 972 in_ifdetach(ifp); 973#endif 974 975#ifdef INET6 976 /* 977 * Remove all IPv6 kernel structs related to ifp. This should be done 978 * before removing routing entries below, since IPv6 interface direct 979 * routes are expected to be removed by the IPv6-specific kernel API. 980 * Otherwise, the kernel will detect some inconsistency and bark it. 981 */ 982 in6_ifdetach(ifp); 983#endif 984 if_purgemaddrs(ifp); 985 986 if (!vmove) { 987 /* 988 * Prevent further calls into the device driver via ifnet. 989 */ 990 if_dead(ifp); 991 992 /* 993 * Remove link ifaddr pointer and maybe decrement if_index. 994 * Clean up all addresses. 995 */ 996 free(ifp->if_hw_addr, M_IFADDR); 997 ifp->if_hw_addr = NULL; 998 ifp->if_addr = NULL; 999 1000 /* We can now free link ifaddr. */ 1001 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 1002 ifa = TAILQ_FIRST(&ifp->if_addrhead); 1003 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 1004 ifa_free(ifa); 1005 } 1006 } 1007 1008 /* 1009 * Delete all remaining routes using this interface 1010 * Unfortuneatly the only way to do this is to slog through 1011 * the entire routing table looking for routes which point 1012 * to this interface...oh well... 1013 */ 1014 for (i = 1; i <= AF_MAX; i++) { 1015 for (j = 0; j < rt_numfibs; j++) { 1016 rnh = rt_tables_get_rnh(j, i); 1017 if (rnh == NULL) 1018 continue; 1019 RADIX_NODE_HEAD_LOCK(rnh); 1020 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 1021 RADIX_NODE_HEAD_UNLOCK(rnh); 1022 } 1023 } 1024 1025 /* Announce that the interface is gone. */ 1026 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1027 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1028 if (IS_DEFAULT_VNET(curvnet)) 1029 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1030 if_delgroups(ifp); 1031 1032 /* 1033 * We cannot hold the lock over dom_ifdetach calls as they might 1034 * sleep, for example trying to drain a callout, thus open up the 1035 * theoretical race with re-attaching. 1036 */ 1037 IF_AFDATA_LOCK(ifp); 1038 i = ifp->if_afdata_initialized; 1039 ifp->if_afdata_initialized = 0; 1040 IF_AFDATA_UNLOCK(ifp); 1041 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1042 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 1043 (*dp->dom_ifdetach)(ifp, 1044 ifp->if_afdata[dp->dom_family]); 1045 } 1046 1047 return (0); 1048} 1049 1050#ifdef VIMAGE 1051/* 1052 * if_vmove() performs a limited version of if_detach() in current 1053 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1054 * An attempt is made to shrink if_index in current vnet, find an 1055 * unused if_index in target vnet and calls if_grow() if necessary, 1056 * and finally find an unused if_xname for the target vnet. 1057 */ 1058void 1059if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1060{ 1061 u_short idx; 1062 struct if_clone *ifc; 1063 int rc; 1064 1065 /* 1066 * Detach from current vnet, but preserve LLADDR info, do not 1067 * mark as dead etc. so that the ifnet can be reattached later. 1068 * If we cannot find it, we lost the race to someone else. 1069 */ 1070 rc = if_detach_internal(ifp, 1, &ifc); 1071 if (rc != 0) 1072 return; 1073 1074 /* 1075 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1076 * the if_index for that vnet if possible. 1077 * 1078 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1079 * or we'd lock on one vnet and unlock on another. 1080 */ 1081 IFNET_WLOCK(); 1082 ifindex_free_locked(ifp->if_index); 1083 IFNET_WUNLOCK(); 1084 1085 /* 1086 * Perform interface-specific reassignment tasks, if provided by 1087 * the driver. 1088 */ 1089 if (ifp->if_reassign != NULL) 1090 ifp->if_reassign(ifp, new_vnet, NULL); 1091 1092 /* 1093 * Switch to the context of the target vnet. 1094 */ 1095 CURVNET_SET_QUIET(new_vnet); 1096 1097 IFNET_WLOCK(); 1098 if (ifindex_alloc_locked(&idx) != 0) { 1099 IFNET_WUNLOCK(); 1100 panic("if_index overflow"); 1101 } 1102 ifp->if_index = idx; 1103 ifnet_setbyindex_locked(ifp->if_index, ifp); 1104 IFNET_WUNLOCK(); 1105 1106 if_attach_internal(ifp, 1, ifc); 1107 1108 CURVNET_RESTORE(); 1109} 1110 1111/* 1112 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1113 */ 1114static int 1115if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1116{ 1117 struct prison *pr; 1118 struct ifnet *difp; 1119 1120 /* Try to find the prison within our visibility. */ 1121 sx_slock(&allprison_lock); 1122 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1123 sx_sunlock(&allprison_lock); 1124 if (pr == NULL) 1125 return (ENXIO); 1126 prison_hold_locked(pr); 1127 mtx_unlock(&pr->pr_mtx); 1128 1129 /* Do not try to move the iface from and to the same prison. */ 1130 if (pr->pr_vnet == ifp->if_vnet) { 1131 prison_free(pr); 1132 return (EEXIST); 1133 } 1134 1135 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1136 /* XXX Lock interfaces to avoid races. */ 1137 CURVNET_SET_QUIET(pr->pr_vnet); 1138 difp = ifunit(ifname); 1139 CURVNET_RESTORE(); 1140 if (difp != NULL) { 1141 prison_free(pr); 1142 return (EEXIST); 1143 } 1144 1145 /* Move the interface into the child jail/vnet. */ 1146 if_vmove(ifp, pr->pr_vnet); 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 1155static int 1156if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1157{ 1158 struct prison *pr; 1159 struct vnet *vnet_dst; 1160 struct ifnet *ifp; 1161 1162 /* Try to find the prison within our visibility. */ 1163 sx_slock(&allprison_lock); 1164 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1165 sx_sunlock(&allprison_lock); 1166 if (pr == NULL) 1167 return (ENXIO); 1168 prison_hold_locked(pr); 1169 mtx_unlock(&pr->pr_mtx); 1170 1171 /* Make sure the named iface exists in the source prison/vnet. */ 1172 CURVNET_SET(pr->pr_vnet); 1173 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1174 if (ifp == NULL) { 1175 CURVNET_RESTORE(); 1176 prison_free(pr); 1177 return (ENXIO); 1178 } 1179 1180 /* Do not try to move the iface from and to the same prison. */ 1181 vnet_dst = TD_TO_VNET(td); 1182 if (vnet_dst == ifp->if_vnet) { 1183 CURVNET_RESTORE(); 1184 prison_free(pr); 1185 return (EEXIST); 1186 } 1187 1188 /* Get interface back from child jail/vnet. */ 1189 if_vmove(ifp, vnet_dst); 1190 CURVNET_RESTORE(); 1191 1192 /* Report the new if_xname back to the userland. */ 1193 sprintf(ifname, "%s", ifp->if_xname); 1194 1195 prison_free(pr); 1196 return (0); 1197} 1198#endif /* VIMAGE */ 1199 1200/* 1201 * Add a group to an interface 1202 */ 1203int 1204if_addgroup(struct ifnet *ifp, const char *groupname) 1205{ 1206 struct ifg_list *ifgl; 1207 struct ifg_group *ifg = NULL; 1208 struct ifg_member *ifgm; 1209 int new = 0; 1210 1211 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1212 groupname[strlen(groupname) - 1] <= '9') 1213 return (EINVAL); 1214 1215 IFNET_WLOCK(); 1216 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1217 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1218 IFNET_WUNLOCK(); 1219 return (EEXIST); 1220 } 1221 1222 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1223 M_NOWAIT)) == NULL) { 1224 IFNET_WUNLOCK(); 1225 return (ENOMEM); 1226 } 1227 1228 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1229 M_TEMP, M_NOWAIT)) == NULL) { 1230 free(ifgl, M_TEMP); 1231 IFNET_WUNLOCK(); 1232 return (ENOMEM); 1233 } 1234 1235 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1236 if (!strcmp(ifg->ifg_group, groupname)) 1237 break; 1238 1239 if (ifg == NULL) { 1240 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1241 M_TEMP, M_NOWAIT)) == NULL) { 1242 free(ifgl, M_TEMP); 1243 free(ifgm, M_TEMP); 1244 IFNET_WUNLOCK(); 1245 return (ENOMEM); 1246 } 1247 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1248 ifg->ifg_refcnt = 0; 1249 TAILQ_INIT(&ifg->ifg_members); 1250 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1251 new = 1; 1252 } 1253 1254 ifg->ifg_refcnt++; 1255 ifgl->ifgl_group = ifg; 1256 ifgm->ifgm_ifp = ifp; 1257 1258 IF_ADDR_WLOCK(ifp); 1259 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1260 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1261 IF_ADDR_WUNLOCK(ifp); 1262 1263 IFNET_WUNLOCK(); 1264 1265 if (new) 1266 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1267 EVENTHANDLER_INVOKE(group_change_event, groupname); 1268 1269 return (0); 1270} 1271 1272/* 1273 * Remove a group from an interface 1274 */ 1275int 1276if_delgroup(struct ifnet *ifp, const char *groupname) 1277{ 1278 struct ifg_list *ifgl; 1279 struct ifg_member *ifgm; 1280 1281 IFNET_WLOCK(); 1282 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1283 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1284 break; 1285 if (ifgl == NULL) { 1286 IFNET_WUNLOCK(); 1287 return (ENOENT); 1288 } 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, ifgm_next); 1300 free(ifgm, M_TEMP); 1301 } 1302 1303 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1304 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1305 IFNET_WUNLOCK(); 1306 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1307 free(ifgl->ifgl_group, M_TEMP); 1308 } else 1309 IFNET_WUNLOCK(); 1310 1311 free(ifgl, M_TEMP); 1312 1313 EVENTHANDLER_INVOKE(group_change_event, groupname); 1314 1315 return (0); 1316} 1317 1318/* 1319 * Remove an interface from all groups 1320 */ 1321static void 1322if_delgroups(struct ifnet *ifp) 1323{ 1324 struct ifg_list *ifgl; 1325 struct ifg_member *ifgm; 1326 char groupname[IFNAMSIZ]; 1327 1328 IFNET_WLOCK(); 1329 while (!TAILQ_EMPTY(&ifp->if_groups)) { 1330 ifgl = TAILQ_FIRST(&ifp->if_groups); 1331 1332 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1333 1334 IF_ADDR_WLOCK(ifp); 1335 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1336 IF_ADDR_WUNLOCK(ifp); 1337 1338 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1339 if (ifgm->ifgm_ifp == ifp) 1340 break; 1341 1342 if (ifgm != NULL) { 1343 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1344 ifgm_next); 1345 free(ifgm, M_TEMP); 1346 } 1347 1348 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1349 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1350 IFNET_WUNLOCK(); 1351 EVENTHANDLER_INVOKE(group_detach_event, 1352 ifgl->ifgl_group); 1353 free(ifgl->ifgl_group, M_TEMP); 1354 } else 1355 IFNET_WUNLOCK(); 1356 1357 free(ifgl, M_TEMP); 1358 1359 EVENTHANDLER_INVOKE(group_change_event, groupname); 1360 1361 IFNET_WLOCK(); 1362 } 1363 IFNET_WUNLOCK(); 1364} 1365 1366/* 1367 * Stores all groups from an interface in memory pointed 1368 * to by data 1369 */ 1370static int 1371if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 1372{ 1373 int len, error; 1374 struct ifg_list *ifgl; 1375 struct ifg_req ifgrq, *ifgp; 1376 struct ifgroupreq *ifgr = data; 1377 1378 if (ifgr->ifgr_len == 0) { 1379 IF_ADDR_RLOCK(ifp); 1380 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1381 ifgr->ifgr_len += sizeof(struct ifg_req); 1382 IF_ADDR_RUNLOCK(ifp); 1383 return (0); 1384 } 1385 1386 len = ifgr->ifgr_len; 1387 ifgp = ifgr->ifgr_groups; 1388 /* XXX: wire */ 1389 IF_ADDR_RLOCK(ifp); 1390 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1391 if (len < sizeof(ifgrq)) { 1392 IF_ADDR_RUNLOCK(ifp); 1393 return (EINVAL); 1394 } 1395 bzero(&ifgrq, sizeof ifgrq); 1396 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1397 sizeof(ifgrq.ifgrq_group)); 1398 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1399 IF_ADDR_RUNLOCK(ifp); 1400 return (error); 1401 } 1402 len -= sizeof(ifgrq); 1403 ifgp++; 1404 } 1405 IF_ADDR_RUNLOCK(ifp); 1406 1407 return (0); 1408} 1409 1410/* 1411 * Stores all members of a group in memory pointed to by data 1412 */ 1413static int 1414if_getgroupmembers(struct ifgroupreq *data) 1415{ 1416 struct ifgroupreq *ifgr = data; 1417 struct ifg_group *ifg; 1418 struct ifg_member *ifgm; 1419 struct ifg_req ifgrq, *ifgp; 1420 int len, error; 1421 1422 IFNET_RLOCK(); 1423 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1424 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1425 break; 1426 if (ifg == NULL) { 1427 IFNET_RUNLOCK(); 1428 return (ENOENT); 1429 } 1430 1431 if (ifgr->ifgr_len == 0) { 1432 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1433 ifgr->ifgr_len += sizeof(ifgrq); 1434 IFNET_RUNLOCK(); 1435 return (0); 1436 } 1437 1438 len = ifgr->ifgr_len; 1439 ifgp = ifgr->ifgr_groups; 1440 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1441 if (len < sizeof(ifgrq)) { 1442 IFNET_RUNLOCK(); 1443 return (EINVAL); 1444 } 1445 bzero(&ifgrq, sizeof ifgrq); 1446 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1447 sizeof(ifgrq.ifgrq_member)); 1448 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1449 IFNET_RUNLOCK(); 1450 return (error); 1451 } 1452 len -= sizeof(ifgrq); 1453 ifgp++; 1454 } 1455 IFNET_RUNLOCK(); 1456 1457 return (0); 1458} 1459 1460/* 1461 * Delete Routes for a Network Interface 1462 * 1463 * Called for each routing entry via the rnh->rnh_walktree() call above 1464 * to delete all route entries referencing a detaching network interface. 1465 * 1466 * Arguments: 1467 * rn pointer to node in the routing table 1468 * arg argument passed to rnh->rnh_walktree() - detaching interface 1469 * 1470 * Returns: 1471 * 0 successful 1472 * errno failed - reason indicated 1473 * 1474 */ 1475static int 1476if_rtdel(struct radix_node *rn, void *arg) 1477{ 1478 struct rtentry *rt = (struct rtentry *)rn; 1479 struct ifnet *ifp = arg; 1480 int err; 1481 1482 if (rt->rt_ifp == ifp) { 1483 1484 /* 1485 * Protect (sorta) against walktree recursion problems 1486 * with cloned routes 1487 */ 1488 if ((rt->rt_flags & RTF_UP) == 0) 1489 return (0); 1490 1491 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway, 1492 rt_mask(rt), 1493 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED, 1494 (struct rtentry **) NULL, rt->rt_fibnum); 1495 if (err) { 1496 log(LOG_WARNING, "if_rtdel: error %d\n", err); 1497 } 1498 } 1499 1500 return (0); 1501} 1502 1503/* 1504 * A compatibility function returns ifnet counter values. 1505 */ 1506uint64_t 1507if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1508{ 1509 1510 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1511 switch (cnt) { 1512 case IFCOUNTER_IPACKETS: 1513 return (ifp->if_ipackets); 1514 case IFCOUNTER_IERRORS: 1515 return (ifp->if_ierrors); 1516 case IFCOUNTER_OPACKETS: 1517 return (ifp->if_opackets); 1518 case IFCOUNTER_OERRORS: 1519 return (ifp->if_oerrors); 1520 case IFCOUNTER_COLLISIONS: 1521 return (ifp->if_collisions); 1522 case IFCOUNTER_IBYTES: 1523 return (ifp->if_ibytes); 1524 case IFCOUNTER_OBYTES: 1525 return (ifp->if_obytes); 1526 case IFCOUNTER_IMCASTS: 1527 return (ifp->if_imcasts); 1528 case IFCOUNTER_OMCASTS: 1529 return (ifp->if_omcasts); 1530 case IFCOUNTER_IQDROPS: 1531 return (ifp->if_iqdrops); 1532#ifdef _IFI_OQDROPS 1533 case IFCOUNTER_OQDROPS: 1534 return (ifp->if_oqdrops); 1535#endif 1536 case IFCOUNTER_NOPROTO: 1537 return (ifp->if_noproto); 1538 default: 1539 break; 1540 }; 1541 return (0); 1542} 1543 1544/* 1545 * Increase an ifnet counter. Usually used for counters shared 1546 * between the stack and a driver, but function supports them all. 1547 */ 1548void 1549if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1550{ 1551 1552 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1553 switch (cnt) { 1554 case IFCOUNTER_IPACKETS: 1555 ifp->if_ipackets += inc; 1556 break; 1557 case IFCOUNTER_IERRORS: 1558 ifp->if_ierrors += inc; 1559 break; 1560 case IFCOUNTER_OPACKETS: 1561 ifp->if_opackets += inc; 1562 break; 1563 case IFCOUNTER_OERRORS: 1564 ifp->if_oerrors += inc; 1565 break; 1566 case IFCOUNTER_COLLISIONS: 1567 ifp->if_collisions += inc; 1568 break; 1569 case IFCOUNTER_IBYTES: 1570 ifp->if_ibytes += inc; 1571 break; 1572 case IFCOUNTER_OBYTES: 1573 ifp->if_obytes += inc; 1574 break; 1575 case IFCOUNTER_IMCASTS: 1576 ifp->if_imcasts += inc; 1577 break; 1578 case IFCOUNTER_OMCASTS: 1579 ifp->if_omcasts += inc; 1580 break; 1581 case IFCOUNTER_IQDROPS: 1582 ifp->if_iqdrops += inc; 1583 break; 1584#ifdef _IFI_OQDROPS 1585 case IFCOUNTER_OQDROPS: 1586 ifp->if_oqdrops += inc; 1587 break; 1588#endif 1589 case IFCOUNTER_NOPROTO: 1590 ifp->if_noproto += inc; 1591 break; 1592 default: 1593 break; 1594 }; 1595} 1596 1597/* 1598 * Wrapper functions for struct ifnet address list locking macros. These are 1599 * used by kernel modules to avoid encoding programming interface or binary 1600 * interface assumptions that may be violated when kernel-internal locking 1601 * approaches change. 1602 */ 1603void 1604if_addr_rlock(struct ifnet *ifp) 1605{ 1606 1607 IF_ADDR_RLOCK(ifp); 1608} 1609 1610void 1611if_addr_runlock(struct ifnet *ifp) 1612{ 1613 1614 IF_ADDR_RUNLOCK(ifp); 1615} 1616 1617void 1618if_maddr_rlock(struct ifnet *ifp) 1619{ 1620 1621 IF_ADDR_RLOCK(ifp); 1622} 1623 1624void 1625if_maddr_runlock(struct ifnet *ifp) 1626{ 1627 1628 IF_ADDR_RUNLOCK(ifp); 1629} 1630 1631/* 1632 * Initialization, destruction and refcounting functions for ifaddrs. 1633 */ 1634void 1635ifa_init(struct ifaddr *ifa) 1636{ 1637 1638 mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF); 1639 refcount_init(&ifa->ifa_refcnt, 1); 1640 ifa->if_data.ifi_datalen = sizeof(ifa->if_data); 1641} 1642 1643void 1644ifa_ref(struct ifaddr *ifa) 1645{ 1646 1647 refcount_acquire(&ifa->ifa_refcnt); 1648} 1649 1650void 1651ifa_free(struct ifaddr *ifa) 1652{ 1653 1654 if (refcount_release(&ifa->ifa_refcnt)) { 1655 mtx_destroy(&ifa->ifa_mtx); 1656 free(ifa, M_IFADDR); 1657 } 1658} 1659 1660int 1661ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1662{ 1663 int error = 0; 1664 struct rtentry *rt = NULL; 1665 struct rt_addrinfo info; 1666 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1667 1668 bzero(&info, sizeof(info)); 1669 info.rti_ifp = V_loif; 1670 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED; 1671 info.rti_info[RTAX_DST] = ia; 1672 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1673 error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib); 1674 1675 if (error == 0 && rt != NULL) { 1676 RT_LOCK(rt); 1677 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = 1678 ifa->ifa_ifp->if_type; 1679 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = 1680 ifa->ifa_ifp->if_index; 1681 RT_REMREF(rt); 1682 RT_UNLOCK(rt); 1683 } else if (error != 0) 1684 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error); 1685 1686 return (error); 1687} 1688 1689int 1690ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1691{ 1692 int error = 0; 1693 struct rt_addrinfo info; 1694 struct sockaddr_dl null_sdl; 1695 1696 bzero(&null_sdl, sizeof(null_sdl)); 1697 null_sdl.sdl_len = sizeof(null_sdl); 1698 null_sdl.sdl_family = AF_LINK; 1699 null_sdl.sdl_type = ifa->ifa_ifp->if_type; 1700 null_sdl.sdl_index = ifa->ifa_ifp->if_index; 1701 bzero(&info, sizeof(info)); 1702 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED; 1703 info.rti_info[RTAX_DST] = ia; 1704 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1705 error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib); 1706 1707 if (error != 0) 1708 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error); 1709 1710 return (error); 1711} 1712 1713/* 1714 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1715 * structs used to represent other address families, it is necessary 1716 * to perform a different comparison. 1717 */ 1718 1719#define sa_equal(a1, a2) \ 1720 (bcmp((a1), (a2), ((a1))->sa_len) == 0) 1721 1722#define sa_dl_equal(a1, a2) \ 1723 ((((struct sockaddr_dl *)(a1))->sdl_len == \ 1724 ((struct sockaddr_dl *)(a2))->sdl_len) && \ 1725 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \ 1726 LLADDR((struct sockaddr_dl *)(a2)), \ 1727 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1728 1729/* 1730 * Locate an interface based on a complete address. 1731 */ 1732/*ARGSUSED*/ 1733static struct ifaddr * 1734ifa_ifwithaddr_internal(struct sockaddr *addr, int getref) 1735{ 1736 struct ifnet *ifp; 1737 struct ifaddr *ifa; 1738 1739 IFNET_RLOCK_NOSLEEP(); 1740 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1741 IF_ADDR_RLOCK(ifp); 1742 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1743 if (ifa->ifa_addr->sa_family != addr->sa_family) 1744 continue; 1745 if (sa_equal(addr, ifa->ifa_addr)) { 1746 if (getref) 1747 ifa_ref(ifa); 1748 IF_ADDR_RUNLOCK(ifp); 1749 goto done; 1750 } 1751 /* IP6 doesn't have broadcast */ 1752 if ((ifp->if_flags & IFF_BROADCAST) && 1753 ifa->ifa_broadaddr && 1754 ifa->ifa_broadaddr->sa_len != 0 && 1755 sa_equal(ifa->ifa_broadaddr, addr)) { 1756 if (getref) 1757 ifa_ref(ifa); 1758 IF_ADDR_RUNLOCK(ifp); 1759 goto done; 1760 } 1761 } 1762 IF_ADDR_RUNLOCK(ifp); 1763 } 1764 ifa = NULL; 1765done: 1766 IFNET_RUNLOCK_NOSLEEP(); 1767 return (ifa); 1768} 1769 1770struct ifaddr * 1771ifa_ifwithaddr(struct sockaddr *addr) 1772{ 1773 1774 return (ifa_ifwithaddr_internal(addr, 1)); 1775} 1776 1777int 1778ifa_ifwithaddr_check(struct sockaddr *addr) 1779{ 1780 1781 return (ifa_ifwithaddr_internal(addr, 0) != NULL); 1782} 1783 1784/* 1785 * Locate an interface based on the broadcast address. 1786 */ 1787/* ARGSUSED */ 1788struct ifaddr * 1789ifa_ifwithbroadaddr(struct sockaddr *addr) 1790{ 1791 struct ifnet *ifp; 1792 struct ifaddr *ifa; 1793 1794 IFNET_RLOCK_NOSLEEP(); 1795 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1796 IF_ADDR_RLOCK(ifp); 1797 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1798 if (ifa->ifa_addr->sa_family != addr->sa_family) 1799 continue; 1800 if ((ifp->if_flags & IFF_BROADCAST) && 1801 ifa->ifa_broadaddr && 1802 ifa->ifa_broadaddr->sa_len != 0 && 1803 sa_equal(ifa->ifa_broadaddr, addr)) { 1804 ifa_ref(ifa); 1805 IF_ADDR_RUNLOCK(ifp); 1806 goto done; 1807 } 1808 } 1809 IF_ADDR_RUNLOCK(ifp); 1810 } 1811 ifa = NULL; 1812done: 1813 IFNET_RUNLOCK_NOSLEEP(); 1814 return (ifa); 1815} 1816 1817/* 1818 * Locate the point to point interface with a given destination address. 1819 */ 1820/*ARGSUSED*/ 1821struct ifaddr * 1822ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum) 1823{ 1824 struct ifnet *ifp; 1825 struct ifaddr *ifa; 1826 1827 IFNET_RLOCK_NOSLEEP(); 1828 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1829 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1830 continue; 1831 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1832 continue; 1833 IF_ADDR_RLOCK(ifp); 1834 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1835 if (ifa->ifa_addr->sa_family != addr->sa_family) 1836 continue; 1837 if (ifa->ifa_dstaddr != NULL && 1838 sa_equal(addr, ifa->ifa_dstaddr)) { 1839 ifa_ref(ifa); 1840 IF_ADDR_RUNLOCK(ifp); 1841 goto done; 1842 } 1843 } 1844 IF_ADDR_RUNLOCK(ifp); 1845 } 1846 ifa = NULL; 1847done: 1848 IFNET_RUNLOCK_NOSLEEP(); 1849 return (ifa); 1850} 1851 1852struct ifaddr * 1853ifa_ifwithdstaddr(struct sockaddr *addr) 1854{ 1855 1856 return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS)); 1857} 1858 1859/* 1860 * Find an interface on a specific network. If many, choice 1861 * is most specific found. 1862 */ 1863struct ifaddr * 1864ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum) 1865{ 1866 struct ifnet *ifp; 1867 struct ifaddr *ifa; 1868 struct ifaddr *ifa_maybe = NULL; 1869 u_int af = addr->sa_family; 1870 char *addr_data = addr->sa_data, *cplim; 1871 1872 /* 1873 * AF_LINK addresses can be looked up directly by their index number, 1874 * so do that if we can. 1875 */ 1876 if (af == AF_LINK) { 1877 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1878 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1879 return (ifaddr_byindex(sdl->sdl_index)); 1880 } 1881 1882 /* 1883 * Scan though each interface, looking for ones that have addresses 1884 * in this address family and the requested fib. Maintain a reference 1885 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 1886 * kept it stable when we move onto the next interface. 1887 */ 1888 IFNET_RLOCK_NOSLEEP(); 1889 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1890 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1891 continue; 1892 IF_ADDR_RLOCK(ifp); 1893 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1894 char *cp, *cp2, *cp3; 1895 1896 if (ifa->ifa_addr->sa_family != af) 1897next: continue; 1898 if (af == AF_INET && 1899 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1900 /* 1901 * This is a bit broken as it doesn't 1902 * take into account that the remote end may 1903 * be a single node in the network we are 1904 * looking for. 1905 * The trouble is that we don't know the 1906 * netmask for the remote end. 1907 */ 1908 if (ifa->ifa_dstaddr != NULL && 1909 sa_equal(addr, ifa->ifa_dstaddr)) { 1910 ifa_ref(ifa); 1911 IF_ADDR_RUNLOCK(ifp); 1912 goto done; 1913 } 1914 } else { 1915 /* 1916 * if we have a special address handler, 1917 * then use it instead of the generic one. 1918 */ 1919 if (ifa->ifa_claim_addr) { 1920 if ((*ifa->ifa_claim_addr)(ifa, addr)) { 1921 ifa_ref(ifa); 1922 IF_ADDR_RUNLOCK(ifp); 1923 goto done; 1924 } 1925 continue; 1926 } 1927 1928 /* 1929 * Scan all the bits in the ifa's address. 1930 * If a bit dissagrees with what we are 1931 * looking for, mask it with the netmask 1932 * to see if it really matters. 1933 * (A byte at a time) 1934 */ 1935 if (ifa->ifa_netmask == 0) 1936 continue; 1937 cp = addr_data; 1938 cp2 = ifa->ifa_addr->sa_data; 1939 cp3 = ifa->ifa_netmask->sa_data; 1940 cplim = ifa->ifa_netmask->sa_len 1941 + (char *)ifa->ifa_netmask; 1942 while (cp3 < cplim) 1943 if ((*cp++ ^ *cp2++) & *cp3++) 1944 goto next; /* next address! */ 1945 /* 1946 * If the netmask of what we just found 1947 * is more specific than what we had before 1948 * (if we had one), or if the virtual status 1949 * of new prefix is better than of the old one, 1950 * then remember the new one before continuing 1951 * to search for an even better one. 1952 */ 1953 if (ifa_maybe == NULL || 1954 ifa_preferred(ifa_maybe, ifa) || 1955 rn_refines((caddr_t)ifa->ifa_netmask, 1956 (caddr_t)ifa_maybe->ifa_netmask)) { 1957 if (ifa_maybe != NULL) 1958 ifa_free(ifa_maybe); 1959 ifa_maybe = ifa; 1960 ifa_ref(ifa_maybe); 1961 } 1962 } 1963 } 1964 IF_ADDR_RUNLOCK(ifp); 1965 } 1966 ifa = ifa_maybe; 1967 ifa_maybe = NULL; 1968done: 1969 IFNET_RUNLOCK_NOSLEEP(); 1970 if (ifa_maybe != NULL) 1971 ifa_free(ifa_maybe); 1972 return (ifa); 1973} 1974 1975struct ifaddr * 1976ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp) 1977{ 1978 1979 return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS)); 1980} 1981 1982/* 1983 * Find an interface address specific to an interface best matching 1984 * a given address. 1985 */ 1986struct ifaddr * 1987ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1988{ 1989 struct ifaddr *ifa; 1990 char *cp, *cp2, *cp3; 1991 char *cplim; 1992 struct ifaddr *ifa_maybe = NULL; 1993 u_int af = addr->sa_family; 1994 1995 if (af >= AF_MAX) 1996 return (NULL); 1997 IF_ADDR_RLOCK(ifp); 1998 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1999 if (ifa->ifa_addr->sa_family != af) 2000 continue; 2001 if (ifa_maybe == NULL) 2002 ifa_maybe = ifa; 2003 if (ifa->ifa_netmask == 0) { 2004 if (sa_equal(addr, ifa->ifa_addr) || 2005 (ifa->ifa_dstaddr && 2006 sa_equal(addr, ifa->ifa_dstaddr))) 2007 goto done; 2008 continue; 2009 } 2010 if (ifp->if_flags & IFF_POINTOPOINT) { 2011 if (sa_equal(addr, ifa->ifa_dstaddr)) 2012 goto done; 2013 } else { 2014 cp = addr->sa_data; 2015 cp2 = ifa->ifa_addr->sa_data; 2016 cp3 = ifa->ifa_netmask->sa_data; 2017 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2018 for (; cp3 < cplim; cp3++) 2019 if ((*cp++ ^ *cp2++) & *cp3) 2020 break; 2021 if (cp3 == cplim) 2022 goto done; 2023 } 2024 } 2025 ifa = ifa_maybe; 2026done: 2027 if (ifa != NULL) 2028 ifa_ref(ifa); 2029 IF_ADDR_RUNLOCK(ifp); 2030 return (ifa); 2031} 2032 2033/* 2034 * See whether new ifa is better than current one: 2035 * 1) A non-virtual one is preferred over virtual. 2036 * 2) A virtual in master state preferred over any other state. 2037 * 2038 * Used in several address selecting functions. 2039 */ 2040int 2041ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2042{ 2043 2044 return (cur->ifa_carp && (!next->ifa_carp || 2045 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2046} 2047 2048#include <net/if_llatbl.h> 2049 2050/* 2051 * Default action when installing a route with a Link Level gateway. 2052 * Lookup an appropriate real ifa to point to. 2053 * This should be moved to /sys/net/link.c eventually. 2054 */ 2055static void 2056link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 2057{ 2058 struct ifaddr *ifa, *oifa; 2059 struct sockaddr *dst; 2060 struct ifnet *ifp; 2061 2062 RT_LOCK_ASSERT(rt); 2063 2064 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 2065 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 2066 return; 2067 ifa = ifaof_ifpforaddr(dst, ifp); 2068 if (ifa) { 2069 oifa = rt->rt_ifa; 2070 rt->rt_ifa = ifa; 2071 ifa_free(oifa); 2072 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2073 ifa->ifa_rtrequest(cmd, rt, info); 2074 } 2075} 2076 2077/* 2078 * Mark an interface down and notify protocols of 2079 * the transition. 2080 */ 2081static void 2082if_unroute(struct ifnet *ifp, int flag, int fam) 2083{ 2084 struct ifaddr *ifa; 2085 2086 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2087 2088 ifp->if_flags &= ~flag; 2089 getmicrotime(&ifp->if_lastchange); 2090 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2091 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2092 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2093 ifp->if_qflush(ifp); 2094 2095 if (ifp->if_carp) 2096 (*carp_linkstate_p)(ifp); 2097 rt_ifmsg(ifp); 2098} 2099 2100/* 2101 * Mark an interface up and notify protocols of 2102 * the transition. 2103 */ 2104static void 2105if_route(struct ifnet *ifp, int flag, int fam) 2106{ 2107 struct ifaddr *ifa; 2108 2109 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2110 2111 ifp->if_flags |= flag; 2112 getmicrotime(&ifp->if_lastchange); 2113 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2114 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2115 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2116 if (ifp->if_carp) 2117 (*carp_linkstate_p)(ifp); 2118 rt_ifmsg(ifp); 2119#ifdef INET6 2120 in6_if_up(ifp); 2121#endif 2122} 2123 2124void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2125void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2126struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2127struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2128int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2129int (*vlan_setcookie_p)(struct ifnet *, void *); 2130void *(*vlan_cookie_p)(struct ifnet *); 2131 2132/* 2133 * Handle a change in the interface link state. To avoid LORs 2134 * between driver lock and upper layer locks, as well as possible 2135 * recursions, we post event to taskqueue, and all job 2136 * is done in static do_link_state_change(). 2137 */ 2138void 2139if_link_state_change(struct ifnet *ifp, int link_state) 2140{ 2141 /* Return if state hasn't changed. */ 2142 if (ifp->if_link_state == link_state) 2143 return; 2144 2145 ifp->if_link_state = link_state; 2146 2147 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2148} 2149 2150static void 2151do_link_state_change(void *arg, int pending) 2152{ 2153 struct ifnet *ifp = (struct ifnet *)arg; 2154 int link_state = ifp->if_link_state; 2155 CURVNET_SET(ifp->if_vnet); 2156 2157 /* Notify that the link state has changed. */ 2158 rt_ifmsg(ifp); 2159 if (ifp->if_vlantrunk != NULL) 2160 (*vlan_link_state_p)(ifp); 2161 2162 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2163 IFP2AC(ifp)->ac_netgraph != NULL) 2164 (*ng_ether_link_state_p)(ifp, link_state); 2165 if (ifp->if_carp) 2166 (*carp_linkstate_p)(ifp); 2167 if (ifp->if_bridge) 2168 (*bridge_linkstate_p)(ifp); 2169 if (ifp->if_lagg) 2170 (*lagg_linkstate_p)(ifp, link_state); 2171 2172 if (IS_DEFAULT_VNET(curvnet)) 2173 devctl_notify("IFNET", ifp->if_xname, 2174 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2175 NULL); 2176 if (pending > 1) 2177 if_printf(ifp, "%d link states coalesced\n", pending); 2178 if (log_link_state_change) 2179 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 2180 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2181 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2182 CURVNET_RESTORE(); 2183} 2184 2185/* 2186 * Mark an interface down and notify protocols of 2187 * the transition. 2188 */ 2189void 2190if_down(struct ifnet *ifp) 2191{ 2192 2193 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2194 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2195} 2196 2197/* 2198 * Mark an interface up and notify protocols of 2199 * the transition. 2200 */ 2201void 2202if_up(struct ifnet *ifp) 2203{ 2204 2205 if_route(ifp, IFF_UP, AF_UNSPEC); 2206 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2207} 2208 2209/* 2210 * Flush an interface queue. 2211 */ 2212void 2213if_qflush(struct ifnet *ifp) 2214{ 2215 struct mbuf *m, *n; 2216 struct ifaltq *ifq; 2217 2218 ifq = &ifp->if_snd; 2219 IFQ_LOCK(ifq); 2220#ifdef ALTQ 2221 if (ALTQ_IS_ENABLED(ifq)) 2222 ALTQ_PURGE(ifq); 2223#endif 2224 n = ifq->ifq_head; 2225 while ((m = n) != 0) { 2226 n = m->m_nextpkt; 2227 m_freem(m); 2228 } 2229 ifq->ifq_head = 0; 2230 ifq->ifq_tail = 0; 2231 ifq->ifq_len = 0; 2232 IFQ_UNLOCK(ifq); 2233} 2234 2235/* 2236 * Map interface name to interface structure pointer, with or without 2237 * returning a reference. 2238 */ 2239struct ifnet * 2240ifunit_ref(const char *name) 2241{ 2242 struct ifnet *ifp; 2243 2244 IFNET_RLOCK_NOSLEEP(); 2245 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2246 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2247 !(ifp->if_flags & IFF_DYING)) 2248 break; 2249 } 2250 if (ifp != NULL) 2251 if_ref(ifp); 2252 IFNET_RUNLOCK_NOSLEEP(); 2253 return (ifp); 2254} 2255 2256struct ifnet * 2257ifunit(const char *name) 2258{ 2259 struct ifnet *ifp; 2260 2261 IFNET_RLOCK_NOSLEEP(); 2262 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2263 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2264 break; 2265 } 2266 IFNET_RUNLOCK_NOSLEEP(); 2267 return (ifp); 2268} 2269 2270/* 2271 * Hardware specific interface ioctls. 2272 */ 2273static int 2274ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2275{ 2276 struct ifreq *ifr; 2277 struct ifstat *ifs; 2278 int error = 0, do_ifup = 0; 2279 int new_flags, temp_flags; 2280 size_t namelen, onamelen; 2281 size_t descrlen; 2282 char *descrbuf, *odescrbuf; 2283 char new_name[IFNAMSIZ]; 2284 struct ifaddr *ifa; 2285 struct sockaddr_dl *sdl; 2286 2287 ifr = (struct ifreq *)data; 2288 switch (cmd) { 2289 case SIOCGIFINDEX: 2290 ifr->ifr_index = ifp->if_index; 2291 break; 2292 2293 case SIOCGIFFLAGS: 2294 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2295 ifr->ifr_flags = temp_flags & 0xffff; 2296 ifr->ifr_flagshigh = temp_flags >> 16; 2297 break; 2298 2299 case SIOCGIFCAP: 2300 ifr->ifr_reqcap = ifp->if_capabilities; 2301 ifr->ifr_curcap = ifp->if_capenable; 2302 break; 2303 2304#ifdef MAC 2305 case SIOCGIFMAC: 2306 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2307 break; 2308#endif 2309 2310 case SIOCGIFMETRIC: 2311 ifr->ifr_metric = ifp->if_metric; 2312 break; 2313 2314 case SIOCGIFMTU: 2315 ifr->ifr_mtu = ifp->if_mtu; 2316 break; 2317 2318 case SIOCGIFPHYS: 2319 ifr->ifr_phys = ifp->if_physical; 2320 break; 2321 2322 case SIOCGIFDESCR: 2323 error = 0; 2324 sx_slock(&ifdescr_sx); 2325 if (ifp->if_description == NULL) 2326 error = ENOMSG; 2327 else { 2328 /* space for terminating nul */ 2329 descrlen = strlen(ifp->if_description) + 1; 2330 if (ifr->ifr_buffer.length < descrlen) 2331 ifr->ifr_buffer.buffer = NULL; 2332 else 2333 error = copyout(ifp->if_description, 2334 ifr->ifr_buffer.buffer, descrlen); 2335 ifr->ifr_buffer.length = descrlen; 2336 } 2337 sx_sunlock(&ifdescr_sx); 2338 break; 2339 2340 case SIOCSIFDESCR: 2341 error = priv_check(td, PRIV_NET_SETIFDESCR); 2342 if (error) 2343 return (error); 2344 2345 /* 2346 * Copy only (length-1) bytes to make sure that 2347 * if_description is always nul terminated. The 2348 * length parameter is supposed to count the 2349 * terminating nul in. 2350 */ 2351 if (ifr->ifr_buffer.length > ifdescr_maxlen) 2352 return (ENAMETOOLONG); 2353 else if (ifr->ifr_buffer.length == 0) 2354 descrbuf = NULL; 2355 else { 2356 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR, 2357 M_WAITOK | M_ZERO); 2358 error = copyin(ifr->ifr_buffer.buffer, descrbuf, 2359 ifr->ifr_buffer.length - 1); 2360 if (error) { 2361 free(descrbuf, M_IFDESCR); 2362 break; 2363 } 2364 } 2365 2366 sx_xlock(&ifdescr_sx); 2367 odescrbuf = ifp->if_description; 2368 ifp->if_description = descrbuf; 2369 sx_xunlock(&ifdescr_sx); 2370 2371 getmicrotime(&ifp->if_lastchange); 2372 free(odescrbuf, M_IFDESCR); 2373 break; 2374 2375 case SIOCGIFFIB: 2376 ifr->ifr_fib = ifp->if_fib; 2377 break; 2378 2379 case SIOCSIFFIB: 2380 error = priv_check(td, PRIV_NET_SETIFFIB); 2381 if (error) 2382 return (error); 2383 if (ifr->ifr_fib >= rt_numfibs) 2384 return (EINVAL); 2385 2386 ifp->if_fib = ifr->ifr_fib; 2387 break; 2388 2389 case SIOCSIFFLAGS: 2390 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2391 if (error) 2392 return (error); 2393 /* 2394 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2395 * check, so we don't need special handling here yet. 2396 */ 2397 new_flags = (ifr->ifr_flags & 0xffff) | 2398 (ifr->ifr_flagshigh << 16); 2399 if (ifp->if_flags & IFF_SMART) { 2400 /* Smart drivers twiddle their own routes */ 2401 } else if (ifp->if_flags & IFF_UP && 2402 (new_flags & IFF_UP) == 0) { 2403 if_down(ifp); 2404 } else if (new_flags & IFF_UP && 2405 (ifp->if_flags & IFF_UP) == 0) { 2406 do_ifup = 1; 2407 } 2408 /* See if permanently promiscuous mode bit is about to flip */ 2409 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2410 if (new_flags & IFF_PPROMISC) 2411 ifp->if_flags |= IFF_PROMISC; 2412 else if (ifp->if_pcount == 0) 2413 ifp->if_flags &= ~IFF_PROMISC; 2414 if (log_promisc_mode_change) 2415 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2416 ifp->if_xname, 2417 ((new_flags & IFF_PPROMISC) ? 2418 "enabled" : "disabled")); 2419 } 2420 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2421 (new_flags &~ IFF_CANTCHANGE); 2422 if (ifp->if_ioctl) { 2423 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2424 } 2425 if (do_ifup) 2426 if_up(ifp); 2427 getmicrotime(&ifp->if_lastchange); 2428 break; 2429 2430 case SIOCSIFCAP: 2431 error = priv_check(td, PRIV_NET_SETIFCAP); 2432 if (error) 2433 return (error); 2434 if (ifp->if_ioctl == NULL) 2435 return (EOPNOTSUPP); 2436 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2437 return (EINVAL); 2438 error = (*ifp->if_ioctl)(ifp, cmd, data); 2439 if (error == 0) 2440 getmicrotime(&ifp->if_lastchange); 2441 break; 2442 2443#ifdef MAC 2444 case SIOCSIFMAC: 2445 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2446 break; 2447#endif 2448 2449 case SIOCSIFNAME: 2450 error = priv_check(td, PRIV_NET_SETIFNAME); 2451 if (error) 2452 return (error); 2453 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 2454 if (error != 0) 2455 return (error); 2456 if (new_name[0] == '\0') 2457 return (EINVAL); 2458 if (new_name[IFNAMSIZ-1] != '\0') { 2459 new_name[IFNAMSIZ-1] = '\0'; 2460 if (strlen(new_name) == IFNAMSIZ-1) 2461 return (EINVAL); 2462 } 2463 if (ifunit(new_name) != NULL) 2464 return (EEXIST); 2465 2466 /* 2467 * XXX: Locking. Nothing else seems to lock if_flags, 2468 * and there are numerous other races with the 2469 * ifunit() checks not being atomic with namespace 2470 * changes (renames, vmoves, if_attach, etc). 2471 */ 2472 ifp->if_flags |= IFF_RENAMING; 2473 2474 /* Announce the departure of the interface. */ 2475 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2476 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2477 2478 log(LOG_INFO, "%s: changing name to '%s'\n", 2479 ifp->if_xname, new_name); 2480 2481 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2482 ifa = ifp->if_addr; 2483 IFA_LOCK(ifa); 2484 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2485 namelen = strlen(new_name); 2486 onamelen = sdl->sdl_nlen; 2487 /* 2488 * Move the address if needed. This is safe because we 2489 * allocate space for a name of length IFNAMSIZ when we 2490 * create this in if_attach(). 2491 */ 2492 if (namelen != onamelen) { 2493 bcopy(sdl->sdl_data + onamelen, 2494 sdl->sdl_data + namelen, sdl->sdl_alen); 2495 } 2496 bcopy(new_name, sdl->sdl_data, namelen); 2497 sdl->sdl_nlen = namelen; 2498 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2499 bzero(sdl->sdl_data, onamelen); 2500 while (namelen != 0) 2501 sdl->sdl_data[--namelen] = 0xff; 2502 IFA_UNLOCK(ifa); 2503 2504 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2505 /* Announce the return of the interface. */ 2506 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2507 2508 ifp->if_flags &= ~IFF_RENAMING; 2509 break; 2510 2511#ifdef VIMAGE 2512 case SIOCSIFVNET: 2513 error = priv_check(td, PRIV_NET_SETIFVNET); 2514 if (error) 2515 return (error); 2516 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2517 break; 2518#endif 2519 2520 case SIOCSIFMETRIC: 2521 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2522 if (error) 2523 return (error); 2524 ifp->if_metric = ifr->ifr_metric; 2525 getmicrotime(&ifp->if_lastchange); 2526 break; 2527 2528 case SIOCSIFPHYS: 2529 error = priv_check(td, PRIV_NET_SETIFPHYS); 2530 if (error) 2531 return (error); 2532 if (ifp->if_ioctl == NULL) 2533 return (EOPNOTSUPP); 2534 error = (*ifp->if_ioctl)(ifp, cmd, data); 2535 if (error == 0) 2536 getmicrotime(&ifp->if_lastchange); 2537 break; 2538 2539 case SIOCSIFMTU: 2540 { 2541 u_long oldmtu = ifp->if_mtu; 2542 2543 error = priv_check(td, PRIV_NET_SETIFMTU); 2544 if (error) 2545 return (error); 2546 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2547 return (EINVAL); 2548 if (ifp->if_ioctl == NULL) 2549 return (EOPNOTSUPP); 2550 error = (*ifp->if_ioctl)(ifp, cmd, data); 2551 if (error == 0) { 2552 getmicrotime(&ifp->if_lastchange); 2553 rt_ifmsg(ifp); 2554 } 2555 /* 2556 * If the link MTU changed, do network layer specific procedure. 2557 */ 2558 if (ifp->if_mtu != oldmtu) { 2559#ifdef INET6 2560 nd6_setmtu(ifp); 2561#endif 2562 } 2563 break; 2564 } 2565 2566 case SIOCADDMULTI: 2567 case SIOCDELMULTI: 2568 if (cmd == SIOCADDMULTI) 2569 error = priv_check(td, PRIV_NET_ADDMULTI); 2570 else 2571 error = priv_check(td, PRIV_NET_DELMULTI); 2572 if (error) 2573 return (error); 2574 2575 /* Don't allow group membership on non-multicast interfaces. */ 2576 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2577 return (EOPNOTSUPP); 2578 2579 /* Don't let users screw up protocols' entries. */ 2580 if (ifr->ifr_addr.sa_family != AF_LINK) 2581 return (EINVAL); 2582 2583 if (cmd == SIOCADDMULTI) { 2584 struct ifmultiaddr *ifma; 2585 2586 /* 2587 * Userland is only permitted to join groups once 2588 * via the if_addmulti() KPI, because it cannot hold 2589 * struct ifmultiaddr * between calls. It may also 2590 * lose a race while we check if the membership 2591 * already exists. 2592 */ 2593 IF_ADDR_RLOCK(ifp); 2594 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2595 IF_ADDR_RUNLOCK(ifp); 2596 if (ifma != NULL) 2597 error = EADDRINUSE; 2598 else 2599 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2600 } else { 2601 error = if_delmulti(ifp, &ifr->ifr_addr); 2602 } 2603 if (error == 0) 2604 getmicrotime(&ifp->if_lastchange); 2605 break; 2606 2607 case SIOCSIFPHYADDR: 2608 case SIOCDIFPHYADDR: 2609#ifdef INET6 2610 case SIOCSIFPHYADDR_IN6: 2611#endif 2612 case SIOCSLIFPHYADDR: 2613 case SIOCSIFMEDIA: 2614 case SIOCSIFGENERIC: 2615 error = priv_check(td, PRIV_NET_HWIOCTL); 2616 if (error) 2617 return (error); 2618 if (ifp->if_ioctl == NULL) 2619 return (EOPNOTSUPP); 2620 error = (*ifp->if_ioctl)(ifp, cmd, data); 2621 if (error == 0) 2622 getmicrotime(&ifp->if_lastchange); 2623 break; 2624 2625 case SIOCGIFSTATUS: 2626 ifs = (struct ifstat *)data; 2627 ifs->ascii[0] = '\0'; 2628 2629 case SIOCGIFPSRCADDR: 2630 case SIOCGIFPDSTADDR: 2631 case SIOCGLIFPHYADDR: 2632 case SIOCGIFMEDIA: 2633 case SIOCGIFXMEDIA: 2634 case SIOCGIFGENERIC: 2635 case SIOCGIFRSSKEY: 2636 case SIOCGIFRSSHASH: 2637 if (ifp->if_ioctl == NULL) 2638 return (EOPNOTSUPP); 2639 error = (*ifp->if_ioctl)(ifp, cmd, data); 2640 break; 2641 2642 case SIOCSIFLLADDR: 2643 error = priv_check(td, PRIV_NET_SETLLADDR); 2644 if (error) 2645 return (error); 2646 error = if_setlladdr(ifp, 2647 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2648 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 2649 break; 2650 2651 case SIOCGHWADDR: 2652 error = if_gethwaddr(ifp, ifr); 2653 break; 2654 2655 case SIOCAIFGROUP: 2656 { 2657 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2658 2659 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2660 if (error) 2661 return (error); 2662 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2663 return (error); 2664 break; 2665 } 2666 2667 case SIOCGIFGROUP: 2668 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2669 return (error); 2670 break; 2671 2672 case SIOCDIFGROUP: 2673 { 2674 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2675 2676 error = priv_check(td, PRIV_NET_DELIFGROUP); 2677 if (error) 2678 return (error); 2679 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2680 return (error); 2681 break; 2682 } 2683 2684 default: 2685 error = ENOIOCTL; 2686 break; 2687 } 2688 return (error); 2689} 2690 2691#ifdef COMPAT_FREEBSD32 2692struct ifconf32 { 2693 int32_t ifc_len; 2694 union { 2695 uint32_t ifcu_buf; 2696 uint32_t ifcu_req; 2697 } ifc_ifcu; 2698}; 2699#define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2700#endif 2701 2702/* 2703 * Interface ioctls. 2704 */ 2705int 2706ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2707{ 2708 struct ifnet *ifp; 2709 struct ifreq *ifr; 2710 int error; 2711 int oif_flags; 2712 2713 CURVNET_SET(so->so_vnet); 2714 switch (cmd) { 2715 case SIOCGIFCONF: 2716 case OSIOCGIFCONF: 2717 error = ifconf(cmd, data); 2718 CURVNET_RESTORE(); 2719 return (error); 2720 2721#ifdef COMPAT_FREEBSD32 2722 case SIOCGIFCONF32: 2723 { 2724 struct ifconf32 *ifc32; 2725 struct ifconf ifc; 2726 2727 ifc32 = (struct ifconf32 *)data; 2728 ifc.ifc_len = ifc32->ifc_len; 2729 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2730 2731 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2732 CURVNET_RESTORE(); 2733 if (error == 0) 2734 ifc32->ifc_len = ifc.ifc_len; 2735 return (error); 2736 } 2737#endif 2738 } 2739 ifr = (struct ifreq *)data; 2740 2741 switch (cmd) { 2742#ifdef VIMAGE 2743 case SIOCSIFRVNET: 2744 error = priv_check(td, PRIV_NET_SETIFVNET); 2745 if (error == 0) 2746 error = if_vmove_reclaim(td, ifr->ifr_name, 2747 ifr->ifr_jid); 2748 CURVNET_RESTORE(); 2749 return (error); 2750#endif 2751 case SIOCIFCREATE: 2752 case SIOCIFCREATE2: 2753 error = priv_check(td, PRIV_NET_IFCREATE); 2754 if (error == 0) 2755 error = if_clone_create(ifr->ifr_name, 2756 sizeof(ifr->ifr_name), 2757 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL); 2758 CURVNET_RESTORE(); 2759 return (error); 2760 case SIOCIFDESTROY: 2761 error = priv_check(td, PRIV_NET_IFDESTROY); 2762 if (error == 0) 2763 error = if_clone_destroy(ifr->ifr_name); 2764 CURVNET_RESTORE(); 2765 return (error); 2766 2767 case SIOCIFGCLONERS: 2768 error = if_clone_list((struct if_clonereq *)data); 2769 CURVNET_RESTORE(); 2770 return (error); 2771 case SIOCGIFGMEMB: 2772 error = if_getgroupmembers((struct ifgroupreq *)data); 2773 CURVNET_RESTORE(); 2774 return (error); 2775#if defined(INET) || defined(INET6) 2776 case SIOCSVH: 2777 case SIOCGVH: 2778 if (carp_ioctl_p == NULL) 2779 error = EPROTONOSUPPORT; 2780 else 2781 error = (*carp_ioctl_p)(ifr, cmd, td); 2782 CURVNET_RESTORE(); 2783 return (error); 2784#endif 2785 } 2786 2787 ifp = ifunit_ref(ifr->ifr_name); 2788 if (ifp == NULL) { 2789 CURVNET_RESTORE(); 2790 return (ENXIO); 2791 } 2792 2793 error = ifhwioctl(cmd, ifp, data, td); 2794 if (error != ENOIOCTL) { 2795 if_rele(ifp); 2796 CURVNET_RESTORE(); 2797 return (error); 2798 } 2799 2800 oif_flags = ifp->if_flags; 2801 if (so->so_proto == NULL) { 2802 if_rele(ifp); 2803 CURVNET_RESTORE(); 2804 return (EOPNOTSUPP); 2805 } 2806 2807 /* 2808 * Pass the request on to the socket control method, and if the 2809 * latter returns EOPNOTSUPP, directly to the interface. 2810 * 2811 * Make an exception for the legacy SIOCSIF* requests. Drivers 2812 * trust SIOCSIFADDR et al to come from an already privileged 2813 * layer, and do not perform any credentials checks or input 2814 * validation. 2815 */ 2816#ifndef COMPAT_43 2817 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 2818 data, 2819 ifp, td)); 2820 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2821 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2822 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2823 error = (*ifp->if_ioctl)(ifp, cmd, data); 2824#else 2825 { 2826 u_long ocmd = cmd; 2827 2828 switch (cmd) { 2829 2830 case SIOCSIFDSTADDR: 2831 case SIOCSIFADDR: 2832 case SIOCSIFBRDADDR: 2833 case SIOCSIFNETMASK: 2834#if BYTE_ORDER != BIG_ENDIAN 2835 if (ifr->ifr_addr.sa_family == 0 && 2836 ifr->ifr_addr.sa_len < 16) { 2837 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 2838 ifr->ifr_addr.sa_len = 16; 2839 } 2840#else 2841 if (ifr->ifr_addr.sa_len == 0) 2842 ifr->ifr_addr.sa_len = 16; 2843#endif 2844 break; 2845 2846 case OSIOCGIFADDR: 2847 cmd = SIOCGIFADDR; 2848 break; 2849 2850 case OSIOCGIFDSTADDR: 2851 cmd = SIOCGIFDSTADDR; 2852 break; 2853 2854 case OSIOCGIFBRDADDR: 2855 cmd = SIOCGIFBRDADDR; 2856 break; 2857 2858 case OSIOCGIFNETMASK: 2859 cmd = SIOCGIFNETMASK; 2860 } 2861 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 2862 cmd, 2863 data, 2864 ifp, td)); 2865 if (error == EOPNOTSUPP && ifp != NULL && 2866 ifp->if_ioctl != NULL && 2867 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2868 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2869 error = (*ifp->if_ioctl)(ifp, cmd, data); 2870 switch (ocmd) { 2871 2872 case OSIOCGIFADDR: 2873 case OSIOCGIFDSTADDR: 2874 case OSIOCGIFBRDADDR: 2875 case OSIOCGIFNETMASK: 2876 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 2877 2878 } 2879 } 2880#endif /* COMPAT_43 */ 2881 2882 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2883#ifdef INET6 2884 if (ifp->if_flags & IFF_UP) 2885 in6_if_up(ifp); 2886#endif 2887 } 2888 if_rele(ifp); 2889 CURVNET_RESTORE(); 2890 return (error); 2891} 2892 2893/* 2894 * The code common to handling reference counted flags, 2895 * e.g., in ifpromisc() and if_allmulti(). 2896 * The "pflag" argument can specify a permanent mode flag to check, 2897 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2898 * 2899 * Only to be used on stack-owned flags, not driver-owned flags. 2900 */ 2901static int 2902if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 2903{ 2904 struct ifreq ifr; 2905 int error; 2906 int oldflags, oldcount; 2907 2908 /* Sanity checks to catch programming errors */ 2909 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 2910 ("%s: setting driver-owned flag %d", __func__, flag)); 2911 2912 if (onswitch) 2913 KASSERT(*refcount >= 0, 2914 ("%s: increment negative refcount %d for flag %d", 2915 __func__, *refcount, flag)); 2916 else 2917 KASSERT(*refcount > 0, 2918 ("%s: decrement non-positive refcount %d for flag %d", 2919 __func__, *refcount, flag)); 2920 2921 /* In case this mode is permanent, just touch refcount */ 2922 if (ifp->if_flags & pflag) { 2923 *refcount += onswitch ? 1 : -1; 2924 return (0); 2925 } 2926 2927 /* Save ifnet parameters for if_ioctl() may fail */ 2928 oldcount = *refcount; 2929 oldflags = ifp->if_flags; 2930 2931 /* 2932 * See if we aren't the only and touching refcount is enough. 2933 * Actually toggle interface flag if we are the first or last. 2934 */ 2935 if (onswitch) { 2936 if ((*refcount)++) 2937 return (0); 2938 ifp->if_flags |= flag; 2939 } else { 2940 if (--(*refcount)) 2941 return (0); 2942 ifp->if_flags &= ~flag; 2943 } 2944 2945 /* Call down the driver since we've changed interface flags */ 2946 if (ifp->if_ioctl == NULL) { 2947 error = EOPNOTSUPP; 2948 goto recover; 2949 } 2950 ifr.ifr_flags = ifp->if_flags & 0xffff; 2951 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2952 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2953 if (error) 2954 goto recover; 2955 /* Notify userland that interface flags have changed */ 2956 rt_ifmsg(ifp); 2957 return (0); 2958 2959recover: 2960 /* Recover after driver error */ 2961 *refcount = oldcount; 2962 ifp->if_flags = oldflags; 2963 return (error); 2964} 2965 2966/* 2967 * Set/clear promiscuous mode on interface ifp based on the truth value 2968 * of pswitch. The calls are reference counted so that only the first 2969 * "on" request actually has an effect, as does the final "off" request. 2970 * Results are undefined if the "off" and "on" requests are not matched. 2971 */ 2972int 2973ifpromisc(struct ifnet *ifp, int pswitch) 2974{ 2975 int error; 2976 int oldflags = ifp->if_flags; 2977 2978 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 2979 &ifp->if_pcount, pswitch); 2980 /* If promiscuous mode status has changed, log a message */ 2981 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 2982 log_promisc_mode_change) 2983 log(LOG_INFO, "%s: promiscuous mode %s\n", 2984 ifp->if_xname, 2985 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 2986 return (error); 2987} 2988 2989/* 2990 * Return interface configuration 2991 * of system. List may be used 2992 * in later ioctl's (above) to get 2993 * other information. 2994 */ 2995/*ARGSUSED*/ 2996static int 2997ifconf(u_long cmd, caddr_t data) 2998{ 2999 struct ifconf *ifc = (struct ifconf *)data; 3000 struct ifnet *ifp; 3001 struct ifaddr *ifa; 3002 struct ifreq ifr; 3003 struct sbuf *sb; 3004 int error, full = 0, valid_len, max_len; 3005 3006 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 3007 max_len = MAXPHYS - 1; 3008 3009 /* Prevent hostile input from being able to crash the system */ 3010 if (ifc->ifc_len <= 0) 3011 return (EINVAL); 3012 3013again: 3014 if (ifc->ifc_len <= max_len) { 3015 max_len = ifc->ifc_len; 3016 full = 1; 3017 } 3018 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3019 max_len = 0; 3020 valid_len = 0; 3021 3022 IFNET_RLOCK(); 3023 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3024 int addrs; 3025 3026 /* 3027 * Zero the ifr_name buffer to make sure we don't 3028 * disclose the contents of the stack. 3029 */ 3030 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 3031 3032 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3033 >= sizeof(ifr.ifr_name)) { 3034 sbuf_delete(sb); 3035 IFNET_RUNLOCK(); 3036 return (ENAMETOOLONG); 3037 } 3038 3039 addrs = 0; 3040 IF_ADDR_RLOCK(ifp); 3041 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3042 struct sockaddr *sa = ifa->ifa_addr; 3043 3044 if (prison_if(curthread->td_ucred, sa) != 0) 3045 continue; 3046 addrs++; 3047#ifdef COMPAT_43 3048 if (cmd == OSIOCGIFCONF) { 3049 struct osockaddr *osa = 3050 (struct osockaddr *)&ifr.ifr_addr; 3051 ifr.ifr_addr = *sa; 3052 osa->sa_family = sa->sa_family; 3053 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3054 max_len += sizeof(ifr); 3055 } else 3056#endif 3057 if (sa->sa_len <= sizeof(*sa)) { 3058 ifr.ifr_addr = *sa; 3059 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3060 max_len += sizeof(ifr); 3061 } else { 3062 sbuf_bcat(sb, &ifr, 3063 offsetof(struct ifreq, ifr_addr)); 3064 max_len += offsetof(struct ifreq, ifr_addr); 3065 sbuf_bcat(sb, sa, sa->sa_len); 3066 max_len += sa->sa_len; 3067 } 3068 3069 if (sbuf_error(sb) == 0) 3070 valid_len = sbuf_len(sb); 3071 } 3072 IF_ADDR_RUNLOCK(ifp); 3073 if (addrs == 0) { 3074 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 3075 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3076 max_len += sizeof(ifr); 3077 3078 if (sbuf_error(sb) == 0) 3079 valid_len = sbuf_len(sb); 3080 } 3081 } 3082 IFNET_RUNLOCK(); 3083 3084 /* 3085 * If we didn't allocate enough space (uncommon), try again. If 3086 * we have already allocated as much space as we are allowed, 3087 * return what we've got. 3088 */ 3089 if (valid_len != max_len && !full) { 3090 sbuf_delete(sb); 3091 goto again; 3092 } 3093 3094 ifc->ifc_len = valid_len; 3095 sbuf_finish(sb); 3096 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3097 sbuf_delete(sb); 3098 return (error); 3099} 3100 3101/* 3102 * Just like ifpromisc(), but for all-multicast-reception mode. 3103 */ 3104int 3105if_allmulti(struct ifnet *ifp, int onswitch) 3106{ 3107 3108 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3109} 3110 3111struct ifmultiaddr * 3112if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 3113{ 3114 struct ifmultiaddr *ifma; 3115 3116 IF_ADDR_LOCK_ASSERT(ifp); 3117 3118 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3119 if (sa->sa_family == AF_LINK) { 3120 if (sa_dl_equal(ifma->ifma_addr, sa)) 3121 break; 3122 } else { 3123 if (sa_equal(ifma->ifma_addr, sa)) 3124 break; 3125 } 3126 } 3127 3128 return ifma; 3129} 3130 3131/* 3132 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3133 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3134 * the ifnet multicast address list here, so the caller must do that and 3135 * other setup work (such as notifying the device driver). The reference 3136 * count is initialized to 1. 3137 */ 3138static struct ifmultiaddr * 3139if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3140 int mflags) 3141{ 3142 struct ifmultiaddr *ifma; 3143 struct sockaddr *dupsa; 3144 3145 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3146 M_ZERO); 3147 if (ifma == NULL) 3148 return (NULL); 3149 3150 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3151 if (dupsa == NULL) { 3152 free(ifma, M_IFMADDR); 3153 return (NULL); 3154 } 3155 bcopy(sa, dupsa, sa->sa_len); 3156 ifma->ifma_addr = dupsa; 3157 3158 ifma->ifma_ifp = ifp; 3159 ifma->ifma_refcount = 1; 3160 ifma->ifma_protospec = NULL; 3161 3162 if (llsa == NULL) { 3163 ifma->ifma_lladdr = NULL; 3164 return (ifma); 3165 } 3166 3167 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3168 if (dupsa == NULL) { 3169 free(ifma->ifma_addr, M_IFMADDR); 3170 free(ifma, M_IFMADDR); 3171 return (NULL); 3172 } 3173 bcopy(llsa, dupsa, llsa->sa_len); 3174 ifma->ifma_lladdr = dupsa; 3175 3176 return (ifma); 3177} 3178 3179/* 3180 * if_freemulti: free ifmultiaddr structure and possibly attached related 3181 * addresses. The caller is responsible for implementing reference 3182 * counting, notifying the driver, handling routing messages, and releasing 3183 * any dependent link layer state. 3184 */ 3185static void 3186if_freemulti(struct ifmultiaddr *ifma) 3187{ 3188 3189 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3190 ifma->ifma_refcount)); 3191 KASSERT(ifma->ifma_protospec == NULL, 3192 ("if_freemulti: protospec not NULL")); 3193 3194 if (ifma->ifma_lladdr != NULL) 3195 free(ifma->ifma_lladdr, M_IFMADDR); 3196 free(ifma->ifma_addr, M_IFMADDR); 3197 free(ifma, M_IFMADDR); 3198} 3199 3200/* 3201 * Register an additional multicast address with a network interface. 3202 * 3203 * - If the address is already present, bump the reference count on the 3204 * address and return. 3205 * - If the address is not link-layer, look up a link layer address. 3206 * - Allocate address structures for one or both addresses, and attach to the 3207 * multicast address list on the interface. If automatically adding a link 3208 * layer address, the protocol address will own a reference to the link 3209 * layer address, to be freed when it is freed. 3210 * - Notify the network device driver of an addition to the multicast address 3211 * list. 3212 * 3213 * 'sa' points to caller-owned memory with the desired multicast address. 3214 * 3215 * 'retifma' will be used to return a pointer to the resulting multicast 3216 * address reference, if desired. 3217 */ 3218int 3219if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3220 struct ifmultiaddr **retifma) 3221{ 3222 struct ifmultiaddr *ifma, *ll_ifma; 3223 struct sockaddr *llsa; 3224 int error; 3225 3226 /* 3227 * If the address is already present, return a new reference to it; 3228 * otherwise, allocate storage and set up a new address. 3229 */ 3230 IF_ADDR_WLOCK(ifp); 3231 ifma = if_findmulti(ifp, sa); 3232 if (ifma != NULL) { 3233 ifma->ifma_refcount++; 3234 if (retifma != NULL) 3235 *retifma = ifma; 3236 IF_ADDR_WUNLOCK(ifp); 3237 return (0); 3238 } 3239 3240 /* 3241 * The address isn't already present; resolve the protocol address 3242 * into a link layer address, and then look that up, bump its 3243 * refcount or allocate an ifma for that also. If 'llsa' was 3244 * returned, we will need to free it later. 3245 */ 3246 llsa = NULL; 3247 ll_ifma = NULL; 3248 if (ifp->if_resolvemulti != NULL) { 3249 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3250 if (error) 3251 goto unlock_out; 3252 } 3253 3254 /* 3255 * Allocate the new address. Don't hook it up yet, as we may also 3256 * need to allocate a link layer multicast address. 3257 */ 3258 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3259 if (ifma == NULL) { 3260 error = ENOMEM; 3261 goto free_llsa_out; 3262 } 3263 3264 /* 3265 * If a link layer address is found, we'll need to see if it's 3266 * already present in the address list, or allocate is as well. 3267 * When this block finishes, the link layer address will be on the 3268 * list. 3269 */ 3270 if (llsa != NULL) { 3271 ll_ifma = if_findmulti(ifp, llsa); 3272 if (ll_ifma == NULL) { 3273 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3274 if (ll_ifma == NULL) { 3275 --ifma->ifma_refcount; 3276 if_freemulti(ifma); 3277 error = ENOMEM; 3278 goto free_llsa_out; 3279 } 3280 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3281 ifma_link); 3282 } else 3283 ll_ifma->ifma_refcount++; 3284 ifma->ifma_llifma = ll_ifma; 3285 } 3286 3287 /* 3288 * We now have a new multicast address, ifma, and possibly a new or 3289 * referenced link layer address. Add the primary address to the 3290 * ifnet address list. 3291 */ 3292 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3293 3294 if (retifma != NULL) 3295 *retifma = ifma; 3296 3297 /* 3298 * Must generate the message while holding the lock so that 'ifma' 3299 * pointer is still valid. 3300 */ 3301 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3302 IF_ADDR_WUNLOCK(ifp); 3303 3304 /* 3305 * We are certain we have added something, so call down to the 3306 * interface to let them know about it. 3307 */ 3308 if (ifp->if_ioctl != NULL) { 3309 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3310 } 3311 3312 if (llsa != NULL) 3313 free(llsa, M_IFMADDR); 3314 3315 return (0); 3316 3317free_llsa_out: 3318 if (llsa != NULL) 3319 free(llsa, M_IFMADDR); 3320 3321unlock_out: 3322 IF_ADDR_WUNLOCK(ifp); 3323 return (error); 3324} 3325 3326/* 3327 * Delete a multicast group membership by network-layer group address. 3328 * 3329 * Returns ENOENT if the entry could not be found. If ifp no longer 3330 * exists, results are undefined. This entry point should only be used 3331 * from subsystems which do appropriate locking to hold ifp for the 3332 * duration of the call. 3333 * Network-layer protocol domains must use if_delmulti_ifma(). 3334 */ 3335int 3336if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3337{ 3338 struct ifmultiaddr *ifma; 3339 int lastref; 3340#ifdef INVARIANTS 3341 struct ifnet *oifp; 3342 3343 IFNET_RLOCK_NOSLEEP(); 3344 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3345 if (ifp == oifp) 3346 break; 3347 if (ifp != oifp) 3348 ifp = NULL; 3349 IFNET_RUNLOCK_NOSLEEP(); 3350 3351 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3352#endif 3353 if (ifp == NULL) 3354 return (ENOENT); 3355 3356 IF_ADDR_WLOCK(ifp); 3357 lastref = 0; 3358 ifma = if_findmulti(ifp, sa); 3359 if (ifma != NULL) 3360 lastref = if_delmulti_locked(ifp, ifma, 0); 3361 IF_ADDR_WUNLOCK(ifp); 3362 3363 if (ifma == NULL) 3364 return (ENOENT); 3365 3366 if (lastref && ifp->if_ioctl != NULL) { 3367 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3368 } 3369 3370 return (0); 3371} 3372 3373/* 3374 * Delete all multicast group membership for an interface. 3375 * Should be used to quickly flush all multicast filters. 3376 */ 3377void 3378if_delallmulti(struct ifnet *ifp) 3379{ 3380 struct ifmultiaddr *ifma; 3381 struct ifmultiaddr *next; 3382 3383 IF_ADDR_WLOCK(ifp); 3384 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3385 if_delmulti_locked(ifp, ifma, 0); 3386 IF_ADDR_WUNLOCK(ifp); 3387} 3388 3389/* 3390 * Delete a multicast group membership by group membership pointer. 3391 * Network-layer protocol domains must use this routine. 3392 * 3393 * It is safe to call this routine if the ifp disappeared. 3394 */ 3395void 3396if_delmulti_ifma(struct ifmultiaddr *ifma) 3397{ 3398 struct ifnet *ifp; 3399 int lastref; 3400 3401 ifp = ifma->ifma_ifp; 3402#ifdef DIAGNOSTIC 3403 if (ifp == NULL) { 3404 printf("%s: ifma_ifp seems to be detached\n", __func__); 3405 } else { 3406 struct ifnet *oifp; 3407 3408 IFNET_RLOCK_NOSLEEP(); 3409 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3410 if (ifp == oifp) 3411 break; 3412 if (ifp != oifp) { 3413 printf("%s: ifnet %p disappeared\n", __func__, ifp); 3414 ifp = NULL; 3415 } 3416 IFNET_RUNLOCK_NOSLEEP(); 3417 } 3418#endif 3419 /* 3420 * If and only if the ifnet instance exists: Acquire the address lock. 3421 */ 3422 if (ifp != NULL) 3423 IF_ADDR_WLOCK(ifp); 3424 3425 lastref = if_delmulti_locked(ifp, ifma, 0); 3426 3427 if (ifp != NULL) { 3428 /* 3429 * If and only if the ifnet instance exists: 3430 * Release the address lock. 3431 * If the group was left: update the hardware hash filter. 3432 */ 3433 IF_ADDR_WUNLOCK(ifp); 3434 if (lastref && ifp->if_ioctl != NULL) { 3435 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3436 } 3437 } 3438} 3439 3440/* 3441 * Perform deletion of network-layer and/or link-layer multicast address. 3442 * 3443 * Return 0 if the reference count was decremented. 3444 * Return 1 if the final reference was released, indicating that the 3445 * hardware hash filter should be reprogrammed. 3446 */ 3447static int 3448if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3449{ 3450 struct ifmultiaddr *ll_ifma; 3451 3452 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3453 KASSERT(ifma->ifma_ifp == ifp, 3454 ("%s: inconsistent ifp %p", __func__, ifp)); 3455 IF_ADDR_WLOCK_ASSERT(ifp); 3456 } 3457 3458 ifp = ifma->ifma_ifp; 3459 3460 /* 3461 * If the ifnet is detaching, null out references to ifnet, 3462 * so that upper protocol layers will notice, and not attempt 3463 * to obtain locks for an ifnet which no longer exists. The 3464 * routing socket announcement must happen before the ifnet 3465 * instance is detached from the system. 3466 */ 3467 if (detaching) { 3468#ifdef DIAGNOSTIC 3469 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3470#endif 3471 /* 3472 * ifp may already be nulled out if we are being reentered 3473 * to delete the ll_ifma. 3474 */ 3475 if (ifp != NULL) { 3476 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3477 ifma->ifma_ifp = NULL; 3478 } 3479 } 3480 3481 if (--ifma->ifma_refcount > 0) 3482 return 0; 3483 3484 /* 3485 * If this ifma is a network-layer ifma, a link-layer ifma may 3486 * have been associated with it. Release it first if so. 3487 */ 3488 ll_ifma = ifma->ifma_llifma; 3489 if (ll_ifma != NULL) { 3490 KASSERT(ifma->ifma_lladdr != NULL, 3491 ("%s: llifma w/o lladdr", __func__)); 3492 if (detaching) 3493 ll_ifma->ifma_ifp = NULL; /* XXX */ 3494 if (--ll_ifma->ifma_refcount == 0) { 3495 if (ifp != NULL) { 3496 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3497 ifma_link); 3498 } 3499 if_freemulti(ll_ifma); 3500 } 3501 } 3502 3503 if (ifp != NULL) 3504 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3505 3506 if_freemulti(ifma); 3507 3508 /* 3509 * The last reference to this instance of struct ifmultiaddr 3510 * was released; the hardware should be notified of this change. 3511 */ 3512 return 1; 3513} 3514 3515/* 3516 * Set the link layer address on an interface. 3517 * 3518 * At this time we only support certain types of interfaces, 3519 * and we don't allow the length of the address to change. 3520 */ 3521int 3522if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3523{ 3524 struct sockaddr_dl *sdl; 3525 struct ifaddr *ifa; 3526 struct ifreq ifr; 3527 3528 IF_ADDR_RLOCK(ifp); 3529 ifa = ifp->if_addr; 3530 if (ifa == NULL) { 3531 IF_ADDR_RUNLOCK(ifp); 3532 return (EINVAL); 3533 } 3534 ifa_ref(ifa); 3535 IF_ADDR_RUNLOCK(ifp); 3536 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3537 if (sdl == NULL) { 3538 ifa_free(ifa); 3539 return (EINVAL); 3540 } 3541 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3542 ifa_free(ifa); 3543 return (EINVAL); 3544 } 3545 switch (ifp->if_type) { 3546 case IFT_ETHER: 3547 case IFT_FDDI: 3548 case IFT_XETHER: 3549 case IFT_ISO88025: 3550 case IFT_L2VLAN: 3551 case IFT_BRIDGE: 3552 case IFT_ARCNET: 3553 case IFT_IEEE8023ADLAG: 3554 case IFT_IEEE80211: 3555 bcopy(lladdr, LLADDR(sdl), len); 3556 ifa_free(ifa); 3557 break; 3558 default: 3559 ifa_free(ifa); 3560 return (ENODEV); 3561 } 3562 3563 /* 3564 * If the interface is already up, we need 3565 * to re-init it in order to reprogram its 3566 * address filter. 3567 */ 3568 if ((ifp->if_flags & IFF_UP) != 0) { 3569 if (ifp->if_ioctl) { 3570 ifp->if_flags &= ~IFF_UP; 3571 ifr.ifr_flags = ifp->if_flags & 0xffff; 3572 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3573 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3574 ifp->if_flags |= IFF_UP; 3575 ifr.ifr_flags = ifp->if_flags & 0xffff; 3576 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3577 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3578 } 3579#ifdef INET 3580 /* 3581 * Also send gratuitous ARPs to notify other nodes about 3582 * the address change. 3583 */ 3584 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3585 if (ifa->ifa_addr->sa_family == AF_INET) 3586 arp_ifinit(ifp, ifa); 3587 } 3588#endif 3589 } 3590 return (0); 3591} 3592 3593/* 3594 * Get the link layer address that was read from the hardware at attach. 3595 * 3596 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3597 * their component interfaces as IFT_IEEE8023ADLAG. 3598 */ 3599int 3600if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3601{ 3602 3603 if (ifp->if_hw_addr == NULL) 3604 return (ENODEV); 3605 3606 switch (ifp->if_type) { 3607 case IFT_ETHER: 3608 case IFT_IEEE8023ADLAG: 3609 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3610 return (0); 3611 default: 3612 return (ENODEV); 3613 } 3614} 3615 3616/* 3617 * The name argument must be a pointer to storage which will last as 3618 * long as the interface does. For physical devices, the result of 3619 * device_get_name(dev) is a good choice and for pseudo-devices a 3620 * static string works well. 3621 */ 3622void 3623if_initname(struct ifnet *ifp, const char *name, int unit) 3624{ 3625 ifp->if_dname = name; 3626 ifp->if_dunit = unit; 3627 if (unit != IF_DUNIT_NONE) 3628 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3629 else 3630 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3631} 3632 3633int 3634if_printf(struct ifnet *ifp, const char * fmt, ...) 3635{ 3636 va_list ap; 3637 int retval; 3638 3639 retval = printf("%s: ", ifp->if_xname); 3640 va_start(ap, fmt); 3641 retval += vprintf(fmt, ap); 3642 va_end(ap); 3643 return (retval); 3644} 3645 3646void 3647if_start(struct ifnet *ifp) 3648{ 3649 3650 (*(ifp)->if_start)(ifp); 3651} 3652 3653/* 3654 * Backwards compatibility interface for drivers 3655 * that have not implemented it 3656 */ 3657static int 3658if_transmit(struct ifnet *ifp, struct mbuf *m) 3659{ 3660 int error; 3661 3662 IFQ_HANDOFF(ifp, m, error); 3663 return (error); 3664} 3665 3666static void 3667if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3668{ 3669 3670 m_freem(m); 3671} 3672 3673int 3674if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3675{ 3676 int active = 0; 3677 3678 IF_LOCK(ifq); 3679 if (_IF_QFULL(ifq)) { 3680 _IF_DROP(ifq); 3681 IF_UNLOCK(ifq); 3682 m_freem(m); 3683 return (0); 3684 } 3685 if (ifp != NULL) { 3686 ifp->if_obytes += m->m_pkthdr.len + adjust; 3687 if (m->m_flags & (M_BCAST|M_MCAST)) 3688 ifp->if_omcasts++; 3689 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3690 } 3691 _IF_ENQUEUE(ifq, m); 3692 IF_UNLOCK(ifq); 3693 if (ifp != NULL && !active) 3694 (*(ifp)->if_start)(ifp); 3695 return (1); 3696} 3697 3698void 3699if_register_com_alloc(u_char type, 3700 if_com_alloc_t *a, if_com_free_t *f) 3701{ 3702 3703 KASSERT(if_com_alloc[type] == NULL, 3704 ("if_register_com_alloc: %d already registered", type)); 3705 KASSERT(if_com_free[type] == NULL, 3706 ("if_register_com_alloc: %d free already registered", type)); 3707 3708 if_com_alloc[type] = a; 3709 if_com_free[type] = f; 3710} 3711 3712void 3713if_deregister_com_alloc(u_char type) 3714{ 3715 3716 KASSERT(if_com_alloc[type] != NULL, 3717 ("if_deregister_com_alloc: %d not registered", type)); 3718 KASSERT(if_com_free[type] != NULL, 3719 ("if_deregister_com_alloc: %d free not registered", type)); 3720 if_com_alloc[type] = NULL; 3721 if_com_free[type] = NULL; 3722} 3723