if.c revision 332288
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/11/sys/net/if.c 332288 2018-04-08 16:54:07Z brooks $ 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/sysent.h> 59#include <sys/taskqueue.h> 60#include <sys/domain.h> 61#include <sys/jail.h> 62#include <sys/priv.h> 63 64#include <machine/stdarg.h> 65#include <vm/uma.h> 66 67#include <net/bpf.h> 68#include <net/ethernet.h> 69#include <net/if.h> 70#include <net/if_arp.h> 71#include <net/if_clone.h> 72#include <net/if_dl.h> 73#include <net/if_types.h> 74#include <net/if_var.h> 75#include <net/if_media.h> 76#include <net/if_vlan_var.h> 77#include <net/radix.h> 78#include <net/route.h> 79#include <net/vnet.h> 80 81#if defined(INET) || defined(INET6) 82#include <net/ethernet.h> 83#include <netinet/in.h> 84#include <netinet/in_var.h> 85#include <netinet/ip.h> 86#include <netinet/ip_carp.h> 87#ifdef INET 88#include <netinet/if_ether.h> 89#endif /* INET */ 90#ifdef INET6 91#include <netinet6/in6_var.h> 92#include <netinet6/in6_ifattach.h> 93#endif /* INET6 */ 94#endif /* INET || INET6 */ 95 96#include <security/mac/mac_framework.h> 97 98#ifdef COMPAT_FREEBSD32 99#include <sys/mount.h> 100#include <compat/freebsd32/freebsd32.h> 101 102struct ifreq_buffer32 { 103 uint32_t length; /* (size_t) */ 104 uint32_t buffer; /* (void *) */ 105}; 106 107/* 108 * Interface request structure used for socket 109 * ioctl's. All interface ioctl's must have parameter 110 * definitions which begin with ifr_name. The 111 * remainder may be interface specific. 112 */ 113struct ifreq32 { 114 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 115 union { 116 struct sockaddr ifru_addr; 117 struct sockaddr ifru_dstaddr; 118 struct sockaddr ifru_broadaddr; 119 struct ifreq_buffer32 ifru_buffer; 120 short ifru_flags[2]; 121 short ifru_index; 122 int ifru_jid; 123 int ifru_metric; 124 int ifru_mtu; 125 int ifru_phys; 126 int ifru_media; 127 uint32_t ifru_data; 128 int ifru_cap[2]; 129 u_int ifru_fib; 130 u_char ifru_vlan_pcp; 131 } ifr_ifru; 132}; 133CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 134CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 135 __offsetof(struct ifreq32, ifr_ifru)); 136#endif 137 138union ifreq_union { 139 struct ifreq ifr; 140#ifdef COMPAT_FREEBSD32 141 struct ifreq32 ifr32; 142#endif 143}; 144 145SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 146SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 147 148SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 149 &ifqmaxlen, 0, "max send queue size"); 150 151/* Log link state change events */ 152static int log_link_state_change = 1; 153 154SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 155 &log_link_state_change, 0, 156 "log interface link state change events"); 157 158/* Log promiscuous mode change events */ 159static int log_promisc_mode_change = 1; 160 161SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 162 &log_promisc_mode_change, 1, 163 "log promiscuous mode change events"); 164 165/* Interface description */ 166static unsigned int ifdescr_maxlen = 1024; 167SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 168 &ifdescr_maxlen, 0, 169 "administrative maximum length for interface description"); 170 171static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 172 173/* global sx for non-critical path ifdescr */ 174static struct sx ifdescr_sx; 175SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 176 177void (*bridge_linkstate_p)(struct ifnet *ifp); 178void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 179void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 180/* These are external hooks for CARP. */ 181void (*carp_linkstate_p)(struct ifnet *ifp); 182void (*carp_demote_adj_p)(int, char *); 183int (*carp_master_p)(struct ifaddr *); 184#if defined(INET) || defined(INET6) 185int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 186int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 187 const struct sockaddr *sa); 188int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 189int (*carp_attach_p)(struct ifaddr *, int); 190void (*carp_detach_p)(struct ifaddr *); 191#endif 192#ifdef INET 193int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 194#endif 195#ifdef INET6 196struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 197caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 198 const struct in6_addr *taddr); 199#endif 200 201struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 202 203/* 204 * XXX: Style; these should be sorted alphabetically, and unprototyped 205 * static functions should be prototyped. Currently they are sorted by 206 * declaration order. 207 */ 208static void if_attachdomain(void *); 209static void if_attachdomain1(struct ifnet *); 210static int ifconf(u_long, caddr_t); 211static void if_freemulti(struct ifmultiaddr *); 212static void if_grow(void); 213static void if_input_default(struct ifnet *, struct mbuf *); 214static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 215static void if_route(struct ifnet *, int flag, int fam); 216static int if_setflag(struct ifnet *, int, int, int *, int); 217static int if_transmit(struct ifnet *ifp, struct mbuf *m); 218static void if_unroute(struct ifnet *, int flag, int fam); 219static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 220static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 221static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 222static void do_link_state_change(void *, int); 223static int if_getgroup(struct ifgroupreq *, struct ifnet *); 224static int if_getgroupmembers(struct ifgroupreq *); 225static void if_delgroups(struct ifnet *); 226static void if_attach_internal(struct ifnet *, int, struct if_clone *); 227static int if_detach_internal(struct ifnet *, int, struct if_clone **); 228#ifdef VIMAGE 229static void if_vmove(struct ifnet *, struct vnet *); 230#endif 231 232#ifdef INET6 233/* 234 * XXX: declare here to avoid to include many inet6 related files.. 235 * should be more generalized? 236 */ 237extern void nd6_setmtu(struct ifnet *); 238#endif 239 240/* ipsec helper hooks */ 241VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 242VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 243 244VNET_DEFINE(int, if_index); 245int ifqmaxlen = IFQ_MAXLEN; 246VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 247VNET_DEFINE(struct ifgrouphead, ifg_head); 248 249static VNET_DEFINE(int, if_indexlim) = 8; 250 251/* Table of ifnet by index. */ 252VNET_DEFINE(struct ifnet **, ifindex_table); 253 254#define V_if_indexlim VNET(if_indexlim) 255#define V_ifindex_table VNET(ifindex_table) 256 257/* 258 * The global network interface list (V_ifnet) and related state (such as 259 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 260 * an rwlock. Either may be acquired shared to stablize the list, but both 261 * must be acquired writable to modify the list. This model allows us to 262 * both stablize the interface list during interrupt thread processing, but 263 * also to stablize it over long-running ioctls, without introducing priority 264 * inversions and deadlocks. 265 */ 266struct rwlock ifnet_rwlock; 267RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE); 268struct sx ifnet_sxlock; 269SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 270 271/* 272 * The allocation of network interfaces is a rather non-atomic affair; we 273 * need to select an index before we are ready to expose the interface for 274 * use, so will use this pointer value to indicate reservation. 275 */ 276#define IFNET_HOLD (void *)(uintptr_t)(-1) 277 278static if_com_alloc_t *if_com_alloc[256]; 279static if_com_free_t *if_com_free[256]; 280 281static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 282MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 283MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 284 285struct ifnet * 286ifnet_byindex_locked(u_short idx) 287{ 288 289 if (idx > V_if_index) 290 return (NULL); 291 if (V_ifindex_table[idx] == IFNET_HOLD) 292 return (NULL); 293 return (V_ifindex_table[idx]); 294} 295 296struct ifnet * 297ifnet_byindex(u_short idx) 298{ 299 struct ifnet *ifp; 300 301 IFNET_RLOCK_NOSLEEP(); 302 ifp = ifnet_byindex_locked(idx); 303 IFNET_RUNLOCK_NOSLEEP(); 304 return (ifp); 305} 306 307struct ifnet * 308ifnet_byindex_ref(u_short idx) 309{ 310 struct ifnet *ifp; 311 312 IFNET_RLOCK_NOSLEEP(); 313 ifp = ifnet_byindex_locked(idx); 314 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 315 IFNET_RUNLOCK_NOSLEEP(); 316 return (NULL); 317 } 318 if_ref(ifp); 319 IFNET_RUNLOCK_NOSLEEP(); 320 return (ifp); 321} 322 323/* 324 * Allocate an ifindex array entry; return 0 on success or an error on 325 * failure. 326 */ 327static u_short 328ifindex_alloc(void) 329{ 330 u_short idx; 331 332 IFNET_WLOCK_ASSERT(); 333retry: 334 /* 335 * Try to find an empty slot below V_if_index. If we fail, take the 336 * next slot. 337 */ 338 for (idx = 1; idx <= V_if_index; idx++) { 339 if (V_ifindex_table[idx] == NULL) 340 break; 341 } 342 343 /* Catch if_index overflow. */ 344 if (idx >= V_if_indexlim) { 345 if_grow(); 346 goto retry; 347 } 348 if (idx > V_if_index) 349 V_if_index = idx; 350 return (idx); 351} 352 353static void 354ifindex_free_locked(u_short idx) 355{ 356 357 IFNET_WLOCK_ASSERT(); 358 359 V_ifindex_table[idx] = NULL; 360 while (V_if_index > 0 && 361 V_ifindex_table[V_if_index] == NULL) 362 V_if_index--; 363} 364 365static void 366ifindex_free(u_short idx) 367{ 368 369 IFNET_WLOCK(); 370 ifindex_free_locked(idx); 371 IFNET_WUNLOCK(); 372} 373 374static void 375ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) 376{ 377 378 IFNET_WLOCK_ASSERT(); 379 380 V_ifindex_table[idx] = ifp; 381} 382 383static void 384ifnet_setbyindex(u_short idx, struct ifnet *ifp) 385{ 386 387 IFNET_WLOCK(); 388 ifnet_setbyindex_locked(idx, ifp); 389 IFNET_WUNLOCK(); 390} 391 392struct ifaddr * 393ifaddr_byindex(u_short idx) 394{ 395 struct ifnet *ifp; 396 struct ifaddr *ifa = NULL; 397 398 IFNET_RLOCK_NOSLEEP(); 399 ifp = ifnet_byindex_locked(idx); 400 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 401 ifa_ref(ifa); 402 IFNET_RUNLOCK_NOSLEEP(); 403 return (ifa); 404} 405 406/* 407 * Network interface utility routines. 408 * 409 * Routines with ifa_ifwith* names take sockaddr *'s as 410 * parameters. 411 */ 412 413static void 414vnet_if_init(const void *unused __unused) 415{ 416 417 TAILQ_INIT(&V_ifnet); 418 TAILQ_INIT(&V_ifg_head); 419 IFNET_WLOCK(); 420 if_grow(); /* create initial table */ 421 IFNET_WUNLOCK(); 422 vnet_if_clone_init(); 423} 424VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 425 NULL); 426 427#ifdef VIMAGE 428static void 429vnet_if_uninit(const void *unused __unused) 430{ 431 432 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 433 "not empty", __func__, __LINE__, &V_ifnet)); 434 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 435 "not empty", __func__, __LINE__, &V_ifg_head)); 436 437 free((caddr_t)V_ifindex_table, M_IFNET); 438} 439VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 440 vnet_if_uninit, NULL); 441 442static void 443vnet_if_return(const void *unused __unused) 444{ 445 struct ifnet *ifp, *nifp; 446 447 /* Return all inherited interfaces to their parent vnets. */ 448 TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 449 if (ifp->if_home_vnet != ifp->if_vnet) 450 if_vmove(ifp, ifp->if_home_vnet); 451 } 452} 453VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 454 vnet_if_return, NULL); 455#endif 456 457static void 458if_grow(void) 459{ 460 int oldlim; 461 u_int n; 462 struct ifnet **e; 463 464 IFNET_WLOCK_ASSERT(); 465 oldlim = V_if_indexlim; 466 IFNET_WUNLOCK(); 467 n = (oldlim << 1) * sizeof(*e); 468 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 469 IFNET_WLOCK(); 470 if (V_if_indexlim != oldlim) { 471 free(e, M_IFNET); 472 return; 473 } 474 if (V_ifindex_table != NULL) { 475 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 476 free((caddr_t)V_ifindex_table, M_IFNET); 477 } 478 V_if_indexlim <<= 1; 479 V_ifindex_table = e; 480} 481 482/* 483 * Allocate a struct ifnet and an index for an interface. A layer 2 484 * common structure will also be allocated if an allocation routine is 485 * registered for the passed type. 486 */ 487struct ifnet * 488if_alloc(u_char type) 489{ 490 struct ifnet *ifp; 491 u_short idx; 492 493 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 494 IFNET_WLOCK(); 495 idx = ifindex_alloc(); 496 ifnet_setbyindex_locked(idx, IFNET_HOLD); 497 IFNET_WUNLOCK(); 498 ifp->if_index = idx; 499 ifp->if_type = type; 500 ifp->if_alloctype = type; 501#ifdef VIMAGE 502 ifp->if_vnet = curvnet; 503#endif 504 if (if_com_alloc[type] != NULL) { 505 ifp->if_l2com = if_com_alloc[type](type, ifp); 506 if (ifp->if_l2com == NULL) { 507 free(ifp, M_IFNET); 508 ifindex_free(idx); 509 return (NULL); 510 } 511 } 512 513 IF_ADDR_LOCK_INIT(ifp); 514 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 515 ifp->if_afdata_initialized = 0; 516 IF_AFDATA_LOCK_INIT(ifp); 517 TAILQ_INIT(&ifp->if_addrhead); 518 TAILQ_INIT(&ifp->if_multiaddrs); 519 TAILQ_INIT(&ifp->if_groups); 520#ifdef MAC 521 mac_ifnet_init(ifp); 522#endif 523 ifq_init(&ifp->if_snd, ifp); 524 525 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 526 for (int i = 0; i < IFCOUNTERS; i++) 527 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 528 ifp->if_get_counter = if_get_counter_default; 529 ifnet_setbyindex(ifp->if_index, ifp); 530 return (ifp); 531} 532 533/* 534 * Do the actual work of freeing a struct ifnet, and layer 2 common 535 * structure. This call is made when the last reference to an 536 * interface is released. 537 */ 538static void 539if_free_internal(struct ifnet *ifp) 540{ 541 542 KASSERT((ifp->if_flags & IFF_DYING), 543 ("if_free_internal: interface not dying")); 544 545 if (if_com_free[ifp->if_alloctype] != NULL) 546 if_com_free[ifp->if_alloctype](ifp->if_l2com, 547 ifp->if_alloctype); 548 549#ifdef MAC 550 mac_ifnet_destroy(ifp); 551#endif /* MAC */ 552 if (ifp->if_description != NULL) 553 free(ifp->if_description, M_IFDESCR); 554 IF_AFDATA_DESTROY(ifp); 555 IF_ADDR_LOCK_DESTROY(ifp); 556 ifq_delete(&ifp->if_snd); 557 558 for (int i = 0; i < IFCOUNTERS; i++) 559 counter_u64_free(ifp->if_counters[i]); 560 561 free(ifp, M_IFNET); 562} 563 564/* 565 * Deregister an interface and free the associated storage. 566 */ 567void 568if_free(struct ifnet *ifp) 569{ 570 571 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 572 573 CURVNET_SET_QUIET(ifp->if_vnet); 574 IFNET_WLOCK(); 575 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 576 ("%s: freeing unallocated ifnet", ifp->if_xname)); 577 578 ifindex_free_locked(ifp->if_index); 579 IFNET_WUNLOCK(); 580 581 if (refcount_release(&ifp->if_refcount)) 582 if_free_internal(ifp); 583 CURVNET_RESTORE(); 584} 585 586/* 587 * Interfaces to keep an ifnet type-stable despite the possibility of the 588 * driver calling if_free(). If there are additional references, we defer 589 * freeing the underlying data structure. 590 */ 591void 592if_ref(struct ifnet *ifp) 593{ 594 595 /* We don't assert the ifnet list lock here, but arguably should. */ 596 refcount_acquire(&ifp->if_refcount); 597} 598 599void 600if_rele(struct ifnet *ifp) 601{ 602 603 if (!refcount_release(&ifp->if_refcount)) 604 return; 605 if_free_internal(ifp); 606} 607 608void 609ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 610{ 611 612 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 613 614 if (ifq->ifq_maxlen == 0) 615 ifq->ifq_maxlen = ifqmaxlen; 616 617 ifq->altq_type = 0; 618 ifq->altq_disc = NULL; 619 ifq->altq_flags &= ALTQF_CANTCHANGE; 620 ifq->altq_tbr = NULL; 621 ifq->altq_ifp = ifp; 622} 623 624void 625ifq_delete(struct ifaltq *ifq) 626{ 627 mtx_destroy(&ifq->ifq_mtx); 628} 629 630/* 631 * Perform generic interface initialization tasks and attach the interface 632 * to the list of "active" interfaces. If vmove flag is set on entry 633 * to if_attach_internal(), perform only a limited subset of initialization 634 * tasks, given that we are moving from one vnet to another an ifnet which 635 * has already been fully initialized. 636 * 637 * Note that if_detach_internal() removes group membership unconditionally 638 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 639 * Thus, when if_vmove() is applied to a cloned interface, group membership 640 * is lost while a cloned one always joins a group whose name is 641 * ifc->ifc_name. To recover this after if_detach_internal() and 642 * if_attach_internal(), the cloner should be specified to 643 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 644 * attempts to join a group whose name is ifc->ifc_name. 645 * 646 * XXX: 647 * - The decision to return void and thus require this function to 648 * succeed is questionable. 649 * - We should probably do more sanity checking. For instance we don't 650 * do anything to insure if_xname is unique or non-empty. 651 */ 652void 653if_attach(struct ifnet *ifp) 654{ 655 656 if_attach_internal(ifp, 0, NULL); 657} 658 659/* 660 * Compute the least common TSO limit. 661 */ 662void 663if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 664{ 665 /* 666 * 1) If there is no limit currently, take the limit from 667 * the network adapter. 668 * 669 * 2) If the network adapter has a limit below the current 670 * limit, apply it. 671 */ 672 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 673 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 674 pmax->tsomaxbytes = ifp->if_hw_tsomax; 675 } 676 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 677 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 678 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 679 } 680 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 681 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 682 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 683 } 684} 685 686/* 687 * Update TSO limit of a network adapter. 688 * 689 * Returns zero if no change. Else non-zero. 690 */ 691int 692if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 693{ 694 int retval = 0; 695 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 696 ifp->if_hw_tsomax = pmax->tsomaxbytes; 697 retval++; 698 } 699 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 700 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 701 retval++; 702 } 703 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 704 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 705 retval++; 706 } 707 return (retval); 708} 709 710static void 711if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 712{ 713 unsigned socksize, ifasize; 714 int namelen, masklen; 715 struct sockaddr_dl *sdl; 716 struct ifaddr *ifa; 717 718 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 719 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 720 ifp->if_xname); 721 722#ifdef VIMAGE 723 ifp->if_vnet = curvnet; 724 if (ifp->if_home_vnet == NULL) 725 ifp->if_home_vnet = curvnet; 726#endif 727 728 if_addgroup(ifp, IFG_ALL); 729 730 /* Restore group membership for cloned interfaces. */ 731 if (vmove && ifc != NULL) 732 if_clone_addgroup(ifp, ifc); 733 734 getmicrotime(&ifp->if_lastchange); 735 ifp->if_epoch = time_uptime; 736 737 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 738 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 739 ("transmit and qflush must both either be set or both be NULL")); 740 if (ifp->if_transmit == NULL) { 741 ifp->if_transmit = if_transmit; 742 ifp->if_qflush = if_qflush; 743 } 744 if (ifp->if_input == NULL) 745 ifp->if_input = if_input_default; 746 747 if (ifp->if_requestencap == NULL) 748 ifp->if_requestencap = if_requestencap_default; 749 750 if (!vmove) { 751#ifdef MAC 752 mac_ifnet_create(ifp); 753#endif 754 755 /* 756 * Create a Link Level name for this device. 757 */ 758 namelen = strlen(ifp->if_xname); 759 /* 760 * Always save enough space for any possiable name so we 761 * can do a rename in place later. 762 */ 763 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 764 socksize = masklen + ifp->if_addrlen; 765 if (socksize < sizeof(*sdl)) 766 socksize = sizeof(*sdl); 767 socksize = roundup2(socksize, sizeof(long)); 768 ifasize = sizeof(*ifa) + 2 * socksize; 769 ifa = ifa_alloc(ifasize, M_WAITOK); 770 sdl = (struct sockaddr_dl *)(ifa + 1); 771 sdl->sdl_len = socksize; 772 sdl->sdl_family = AF_LINK; 773 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 774 sdl->sdl_nlen = namelen; 775 sdl->sdl_index = ifp->if_index; 776 sdl->sdl_type = ifp->if_type; 777 ifp->if_addr = ifa; 778 ifa->ifa_ifp = ifp; 779 ifa->ifa_rtrequest = link_rtrequest; 780 ifa->ifa_addr = (struct sockaddr *)sdl; 781 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 782 ifa->ifa_netmask = (struct sockaddr *)sdl; 783 sdl->sdl_len = masklen; 784 while (namelen != 0) 785 sdl->sdl_data[--namelen] = 0xff; 786 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 787 /* Reliably crash if used uninitialized. */ 788 ifp->if_broadcastaddr = NULL; 789 790 if (ifp->if_type == IFT_ETHER) { 791 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 792 M_WAITOK | M_ZERO); 793 } 794 795#if defined(INET) || defined(INET6) 796 /* Use defaults for TSO, if nothing is set */ 797 if (ifp->if_hw_tsomax == 0 && 798 ifp->if_hw_tsomaxsegcount == 0 && 799 ifp->if_hw_tsomaxsegsize == 0) { 800 /* 801 * The TSO defaults needs to be such that an 802 * NFS mbuf list of 35 mbufs totalling just 803 * below 64K works and that a chain of mbufs 804 * can be defragged into at most 32 segments: 805 */ 806 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 807 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 808 ifp->if_hw_tsomaxsegcount = 35; 809 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 810 811 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 812 if (ifp->if_capabilities & IFCAP_TSO) { 813 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 814 ifp->if_hw_tsomax, 815 ifp->if_hw_tsomaxsegcount, 816 ifp->if_hw_tsomaxsegsize); 817 } 818 } 819#endif 820 } 821#ifdef VIMAGE 822 else { 823 /* 824 * Update the interface index in the link layer address 825 * of the interface. 826 */ 827 for (ifa = ifp->if_addr; ifa != NULL; 828 ifa = TAILQ_NEXT(ifa, ifa_link)) { 829 if (ifa->ifa_addr->sa_family == AF_LINK) { 830 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 831 sdl->sdl_index = ifp->if_index; 832 } 833 } 834 } 835#endif 836 837 IFNET_WLOCK(); 838 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 839#ifdef VIMAGE 840 curvnet->vnet_ifcnt++; 841#endif 842 IFNET_WUNLOCK(); 843 844 if (domain_init_status >= 2) 845 if_attachdomain1(ifp); 846 847 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 848 if (IS_DEFAULT_VNET(curvnet)) 849 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 850 851 /* Announce the interface. */ 852 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 853} 854 855static void 856if_attachdomain(void *dummy) 857{ 858 struct ifnet *ifp; 859 860 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 861 if_attachdomain1(ifp); 862} 863SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 864 if_attachdomain, NULL); 865 866static void 867if_attachdomain1(struct ifnet *ifp) 868{ 869 struct domain *dp; 870 871 /* 872 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 873 * cannot lock ifp->if_afdata initialization, entirely. 874 */ 875 IF_AFDATA_LOCK(ifp); 876 if (ifp->if_afdata_initialized >= domain_init_status) { 877 IF_AFDATA_UNLOCK(ifp); 878 log(LOG_WARNING, "%s called more than once on %s\n", 879 __func__, ifp->if_xname); 880 return; 881 } 882 ifp->if_afdata_initialized = domain_init_status; 883 IF_AFDATA_UNLOCK(ifp); 884 885 /* address family dependent data region */ 886 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 887 for (dp = domains; dp; dp = dp->dom_next) { 888 if (dp->dom_ifattach) 889 ifp->if_afdata[dp->dom_family] = 890 (*dp->dom_ifattach)(ifp); 891 } 892} 893 894/* 895 * Remove any unicast or broadcast network addresses from an interface. 896 */ 897void 898if_purgeaddrs(struct ifnet *ifp) 899{ 900 struct ifaddr *ifa, *next; 901 902 /* XXX cannot hold IF_ADDR_WLOCK over called functions. */ 903 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 904 if (ifa->ifa_addr->sa_family == AF_LINK) 905 continue; 906#ifdef INET 907 /* XXX: Ugly!! ad hoc just for INET */ 908 if (ifa->ifa_addr->sa_family == AF_INET) { 909 struct ifaliasreq ifr; 910 911 bzero(&ifr, sizeof(ifr)); 912 ifr.ifra_addr = *ifa->ifa_addr; 913 if (ifa->ifa_dstaddr) 914 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 915 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 916 NULL) == 0) 917 continue; 918 } 919#endif /* INET */ 920#ifdef INET6 921 if (ifa->ifa_addr->sa_family == AF_INET6) { 922 in6_purgeaddr(ifa); 923 /* ifp_addrhead is already updated */ 924 continue; 925 } 926#endif /* INET6 */ 927 IF_ADDR_WLOCK(ifp); 928 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 929 IF_ADDR_WUNLOCK(ifp); 930 ifa_free(ifa); 931 } 932} 933 934/* 935 * Remove any multicast network addresses from an interface when an ifnet 936 * is going away. 937 */ 938static void 939if_purgemaddrs(struct ifnet *ifp) 940{ 941 struct ifmultiaddr *ifma; 942 struct ifmultiaddr *next; 943 944 IF_ADDR_WLOCK(ifp); 945 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 946 if_delmulti_locked(ifp, ifma, 1); 947 IF_ADDR_WUNLOCK(ifp); 948} 949 950/* 951 * Detach an interface, removing it from the list of "active" interfaces. 952 * If vmove flag is set on entry to if_detach_internal(), perform only a 953 * limited subset of cleanup tasks, given that we are moving an ifnet from 954 * one vnet to another, where it must be fully operational. 955 * 956 * XXXRW: There are some significant questions about event ordering, and 957 * how to prevent things from starting to use the interface during detach. 958 */ 959void 960if_detach(struct ifnet *ifp) 961{ 962 963 CURVNET_SET_QUIET(ifp->if_vnet); 964 if_detach_internal(ifp, 0, NULL); 965 CURVNET_RESTORE(); 966} 967 968/* 969 * The vmove flag, if set, indicates that we are called from a callpath 970 * that is moving an interface to a different vnet instance. 971 * 972 * The shutdown flag, if set, indicates that we are called in the 973 * process of shutting down a vnet instance. Currently only the 974 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 975 * on a vnet instance shutdown without this flag being set, e.g., when 976 * the cloned interfaces are destoyed as first thing of teardown. 977 */ 978static int 979if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 980{ 981 struct ifaddr *ifa; 982 int i; 983 struct domain *dp; 984 struct ifnet *iter; 985 int found = 0; 986#ifdef VIMAGE 987 int shutdown; 988 989 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 990 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 991#endif 992 IFNET_WLOCK(); 993 TAILQ_FOREACH(iter, &V_ifnet, if_link) 994 if (iter == ifp) { 995 TAILQ_REMOVE(&V_ifnet, ifp, if_link); 996 found = 1; 997 break; 998 } 999 IFNET_WUNLOCK(); 1000 if (!found) { 1001 /* 1002 * While we would want to panic here, we cannot 1003 * guarantee that the interface is indeed still on 1004 * the list given we don't hold locks all the way. 1005 */ 1006 return (ENOENT); 1007#if 0 1008 if (vmove) 1009 panic("%s: ifp=%p not on the ifnet tailq %p", 1010 __func__, ifp, &V_ifnet); 1011 else 1012 return; /* XXX this should panic as well? */ 1013#endif 1014 } 1015 1016 /* 1017 * At this point we know the interface still was on the ifnet list 1018 * and we removed it so we are in a stable state. 1019 */ 1020#ifdef VIMAGE 1021 curvnet->vnet_ifcnt--; 1022#endif 1023 1024 /* 1025 * In any case (destroy or vmove) detach us from the groups 1026 * and remove/wait for pending events on the taskq. 1027 * XXX-BZ in theory an interface could still enqueue a taskq change? 1028 */ 1029 if_delgroups(ifp); 1030 1031 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1032 1033 /* 1034 * Check if this is a cloned interface or not. Must do even if 1035 * shutting down as a if_vmove_reclaim() would move the ifp and 1036 * the if_clone_addgroup() will have a corrupted string overwise 1037 * from a gibberish pointer. 1038 */ 1039 if (vmove && ifcp != NULL) 1040 *ifcp = if_clone_findifc(ifp); 1041 1042 if_down(ifp); 1043 1044#ifdef VIMAGE 1045 /* 1046 * On VNET shutdown abort here as the stack teardown will do all 1047 * the work top-down for us. 1048 */ 1049 if (shutdown) { 1050 /* 1051 * In case of a vmove we are done here without error. 1052 * If we would signal an error it would lead to the same 1053 * abort as if we did not find the ifnet anymore. 1054 * if_detach() calls us in void context and does not care 1055 * about an early abort notification, so life is splendid :) 1056 */ 1057 goto finish_vnet_shutdown; 1058 } 1059#endif 1060 1061 /* 1062 * At this point we are not tearing down a VNET and are either 1063 * going to destroy or vmove the interface and have to cleanup 1064 * accordingly. 1065 */ 1066 1067 /* 1068 * Remove routes and flush queues. 1069 */ 1070#ifdef ALTQ 1071 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1072 altq_disable(&ifp->if_snd); 1073 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1074 altq_detach(&ifp->if_snd); 1075#endif 1076 1077 if_purgeaddrs(ifp); 1078 1079#ifdef INET 1080 in_ifdetach(ifp); 1081#endif 1082 1083#ifdef INET6 1084 /* 1085 * Remove all IPv6 kernel structs related to ifp. This should be done 1086 * before removing routing entries below, since IPv6 interface direct 1087 * routes are expected to be removed by the IPv6-specific kernel API. 1088 * Otherwise, the kernel will detect some inconsistency and bark it. 1089 */ 1090 in6_ifdetach(ifp); 1091#endif 1092 if_purgemaddrs(ifp); 1093 1094 /* Announce that the interface is gone. */ 1095 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1096 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1097 if (IS_DEFAULT_VNET(curvnet)) 1098 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1099 1100 if (!vmove) { 1101 /* 1102 * Prevent further calls into the device driver via ifnet. 1103 */ 1104 if_dead(ifp); 1105 1106 /* 1107 * Remove link ifaddr pointer and maybe decrement if_index. 1108 * Clean up all addresses. 1109 */ 1110 free(ifp->if_hw_addr, M_IFADDR); 1111 ifp->if_hw_addr = NULL; 1112 ifp->if_addr = NULL; 1113 1114 /* We can now free link ifaddr. */ 1115 IF_ADDR_WLOCK(ifp); 1116 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 1117 ifa = TAILQ_FIRST(&ifp->if_addrhead); 1118 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 1119 IF_ADDR_WUNLOCK(ifp); 1120 ifa_free(ifa); 1121 } else 1122 IF_ADDR_WUNLOCK(ifp); 1123 } 1124 1125 rt_flushifroutes(ifp); 1126 1127#ifdef VIMAGE 1128finish_vnet_shutdown: 1129#endif 1130 /* 1131 * We cannot hold the lock over dom_ifdetach calls as they might 1132 * sleep, for example trying to drain a callout, thus open up the 1133 * theoretical race with re-attaching. 1134 */ 1135 IF_AFDATA_LOCK(ifp); 1136 i = ifp->if_afdata_initialized; 1137 ifp->if_afdata_initialized = 0; 1138 IF_AFDATA_UNLOCK(ifp); 1139 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1140 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1141 (*dp->dom_ifdetach)(ifp, 1142 ifp->if_afdata[dp->dom_family]); 1143 ifp->if_afdata[dp->dom_family] = NULL; 1144 } 1145 } 1146 1147 return (0); 1148} 1149 1150#ifdef VIMAGE 1151/* 1152 * if_vmove() performs a limited version of if_detach() in current 1153 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1154 * An attempt is made to shrink if_index in current vnet, find an 1155 * unused if_index in target vnet and calls if_grow() if necessary, 1156 * and finally find an unused if_xname for the target vnet. 1157 */ 1158static void 1159if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1160{ 1161 struct if_clone *ifc; 1162 u_int bif_dlt, bif_hdrlen; 1163 int rc; 1164 1165 /* 1166 * if_detach_internal() will call the eventhandler to notify 1167 * interface departure. That will detach if_bpf. We need to 1168 * safe the dlt and hdrlen so we can re-attach it later. 1169 */ 1170 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1171 1172 /* 1173 * Detach from current vnet, but preserve LLADDR info, do not 1174 * mark as dead etc. so that the ifnet can be reattached later. 1175 * If we cannot find it, we lost the race to someone else. 1176 */ 1177 rc = if_detach_internal(ifp, 1, &ifc); 1178 if (rc != 0) 1179 return; 1180 1181 /* 1182 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1183 * the if_index for that vnet if possible. 1184 * 1185 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1186 * or we'd lock on one vnet and unlock on another. 1187 */ 1188 IFNET_WLOCK(); 1189 ifindex_free_locked(ifp->if_index); 1190 IFNET_WUNLOCK(); 1191 1192 /* 1193 * Perform interface-specific reassignment tasks, if provided by 1194 * the driver. 1195 */ 1196 if (ifp->if_reassign != NULL) 1197 ifp->if_reassign(ifp, new_vnet, NULL); 1198 1199 /* 1200 * Switch to the context of the target vnet. 1201 */ 1202 CURVNET_SET_QUIET(new_vnet); 1203 1204 IFNET_WLOCK(); 1205 ifp->if_index = ifindex_alloc(); 1206 ifnet_setbyindex_locked(ifp->if_index, ifp); 1207 IFNET_WUNLOCK(); 1208 1209 if_attach_internal(ifp, 1, ifc); 1210 1211 if (ifp->if_bpf == NULL) 1212 bpfattach(ifp, bif_dlt, bif_hdrlen); 1213 1214 CURVNET_RESTORE(); 1215} 1216 1217/* 1218 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1219 */ 1220static int 1221if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1222{ 1223 struct prison *pr; 1224 struct ifnet *difp; 1225 int shutdown; 1226 1227 /* Try to find the prison within our visibility. */ 1228 sx_slock(&allprison_lock); 1229 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1230 sx_sunlock(&allprison_lock); 1231 if (pr == NULL) 1232 return (ENXIO); 1233 prison_hold_locked(pr); 1234 mtx_unlock(&pr->pr_mtx); 1235 1236 /* Do not try to move the iface from and to the same prison. */ 1237 if (pr->pr_vnet == ifp->if_vnet) { 1238 prison_free(pr); 1239 return (EEXIST); 1240 } 1241 1242 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1243 /* XXX Lock interfaces to avoid races. */ 1244 CURVNET_SET_QUIET(pr->pr_vnet); 1245 difp = ifunit(ifname); 1246 if (difp != NULL) { 1247 CURVNET_RESTORE(); 1248 prison_free(pr); 1249 return (EEXIST); 1250 } 1251 1252 /* Make sure the VNET is stable. */ 1253 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1254 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1255 if (shutdown) { 1256 CURVNET_RESTORE(); 1257 prison_free(pr); 1258 return (EBUSY); 1259 } 1260 CURVNET_RESTORE(); 1261 1262 /* Move the interface into the child jail/vnet. */ 1263 if_vmove(ifp, pr->pr_vnet); 1264 1265 /* Report the new if_xname back to the userland. */ 1266 sprintf(ifname, "%s", ifp->if_xname); 1267 1268 prison_free(pr); 1269 return (0); 1270} 1271 1272static int 1273if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1274{ 1275 struct prison *pr; 1276 struct vnet *vnet_dst; 1277 struct ifnet *ifp; 1278 int shutdown; 1279 1280 /* Try to find the prison within our visibility. */ 1281 sx_slock(&allprison_lock); 1282 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1283 sx_sunlock(&allprison_lock); 1284 if (pr == NULL) 1285 return (ENXIO); 1286 prison_hold_locked(pr); 1287 mtx_unlock(&pr->pr_mtx); 1288 1289 /* Make sure the named iface exists in the source prison/vnet. */ 1290 CURVNET_SET(pr->pr_vnet); 1291 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1292 if (ifp == NULL) { 1293 CURVNET_RESTORE(); 1294 prison_free(pr); 1295 return (ENXIO); 1296 } 1297 1298 /* Do not try to move the iface from and to the same prison. */ 1299 vnet_dst = TD_TO_VNET(td); 1300 if (vnet_dst == ifp->if_vnet) { 1301 CURVNET_RESTORE(); 1302 prison_free(pr); 1303 return (EEXIST); 1304 } 1305 1306 /* Make sure the VNET is stable. */ 1307 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1308 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1309 if (shutdown) { 1310 CURVNET_RESTORE(); 1311 prison_free(pr); 1312 return (EBUSY); 1313 } 1314 1315 /* Get interface back from child jail/vnet. */ 1316 if_vmove(ifp, vnet_dst); 1317 CURVNET_RESTORE(); 1318 1319 /* Report the new if_xname back to the userland. */ 1320 sprintf(ifname, "%s", ifp->if_xname); 1321 1322 prison_free(pr); 1323 return (0); 1324} 1325#endif /* VIMAGE */ 1326 1327/* 1328 * Add a group to an interface 1329 */ 1330int 1331if_addgroup(struct ifnet *ifp, const char *groupname) 1332{ 1333 struct ifg_list *ifgl; 1334 struct ifg_group *ifg = NULL; 1335 struct ifg_member *ifgm; 1336 int new = 0; 1337 1338 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1339 groupname[strlen(groupname) - 1] <= '9') 1340 return (EINVAL); 1341 1342 IFNET_WLOCK(); 1343 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1344 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1345 IFNET_WUNLOCK(); 1346 return (EEXIST); 1347 } 1348 1349 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1350 M_NOWAIT)) == NULL) { 1351 IFNET_WUNLOCK(); 1352 return (ENOMEM); 1353 } 1354 1355 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1356 M_TEMP, M_NOWAIT)) == NULL) { 1357 free(ifgl, M_TEMP); 1358 IFNET_WUNLOCK(); 1359 return (ENOMEM); 1360 } 1361 1362 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1363 if (!strcmp(ifg->ifg_group, groupname)) 1364 break; 1365 1366 if (ifg == NULL) { 1367 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1368 M_TEMP, M_NOWAIT)) == NULL) { 1369 free(ifgl, M_TEMP); 1370 free(ifgm, M_TEMP); 1371 IFNET_WUNLOCK(); 1372 return (ENOMEM); 1373 } 1374 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1375 ifg->ifg_refcnt = 0; 1376 TAILQ_INIT(&ifg->ifg_members); 1377 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1378 new = 1; 1379 } 1380 1381 ifg->ifg_refcnt++; 1382 ifgl->ifgl_group = ifg; 1383 ifgm->ifgm_ifp = ifp; 1384 1385 IF_ADDR_WLOCK(ifp); 1386 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1387 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1388 IF_ADDR_WUNLOCK(ifp); 1389 1390 IFNET_WUNLOCK(); 1391 1392 if (new) 1393 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1394 EVENTHANDLER_INVOKE(group_change_event, groupname); 1395 1396 return (0); 1397} 1398 1399/* 1400 * Remove a group from an interface 1401 */ 1402int 1403if_delgroup(struct ifnet *ifp, const char *groupname) 1404{ 1405 struct ifg_list *ifgl; 1406 struct ifg_member *ifgm; 1407 1408 IFNET_WLOCK(); 1409 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1410 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1411 break; 1412 if (ifgl == NULL) { 1413 IFNET_WUNLOCK(); 1414 return (ENOENT); 1415 } 1416 1417 IF_ADDR_WLOCK(ifp); 1418 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1419 IF_ADDR_WUNLOCK(ifp); 1420 1421 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1422 if (ifgm->ifgm_ifp == ifp) 1423 break; 1424 1425 if (ifgm != NULL) { 1426 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 1427 free(ifgm, M_TEMP); 1428 } 1429 1430 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1431 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1432 IFNET_WUNLOCK(); 1433 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1434 free(ifgl->ifgl_group, M_TEMP); 1435 } else 1436 IFNET_WUNLOCK(); 1437 1438 free(ifgl, M_TEMP); 1439 1440 EVENTHANDLER_INVOKE(group_change_event, groupname); 1441 1442 return (0); 1443} 1444 1445/* 1446 * Remove an interface from all groups 1447 */ 1448static void 1449if_delgroups(struct ifnet *ifp) 1450{ 1451 struct ifg_list *ifgl; 1452 struct ifg_member *ifgm; 1453 char groupname[IFNAMSIZ]; 1454 1455 IFNET_WLOCK(); 1456 while (!TAILQ_EMPTY(&ifp->if_groups)) { 1457 ifgl = TAILQ_FIRST(&ifp->if_groups); 1458 1459 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1460 1461 IF_ADDR_WLOCK(ifp); 1462 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1463 IF_ADDR_WUNLOCK(ifp); 1464 1465 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1466 if (ifgm->ifgm_ifp == ifp) 1467 break; 1468 1469 if (ifgm != NULL) { 1470 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1471 ifgm_next); 1472 free(ifgm, M_TEMP); 1473 } 1474 1475 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1476 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1477 IFNET_WUNLOCK(); 1478 EVENTHANDLER_INVOKE(group_detach_event, 1479 ifgl->ifgl_group); 1480 free(ifgl->ifgl_group, M_TEMP); 1481 } else 1482 IFNET_WUNLOCK(); 1483 1484 free(ifgl, M_TEMP); 1485 1486 EVENTHANDLER_INVOKE(group_change_event, groupname); 1487 1488 IFNET_WLOCK(); 1489 } 1490 IFNET_WUNLOCK(); 1491} 1492 1493/* 1494 * Stores all groups from an interface in memory pointed 1495 * to by data 1496 */ 1497static int 1498if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 1499{ 1500 int len, error; 1501 struct ifg_list *ifgl; 1502 struct ifg_req ifgrq, *ifgp; 1503 struct ifgroupreq *ifgr = data; 1504 1505 if (ifgr->ifgr_len == 0) { 1506 IF_ADDR_RLOCK(ifp); 1507 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1508 ifgr->ifgr_len += sizeof(struct ifg_req); 1509 IF_ADDR_RUNLOCK(ifp); 1510 return (0); 1511 } 1512 1513 len = ifgr->ifgr_len; 1514 ifgp = ifgr->ifgr_groups; 1515 /* XXX: wire */ 1516 IF_ADDR_RLOCK(ifp); 1517 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1518 if (len < sizeof(ifgrq)) { 1519 IF_ADDR_RUNLOCK(ifp); 1520 return (EINVAL); 1521 } 1522 bzero(&ifgrq, sizeof ifgrq); 1523 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1524 sizeof(ifgrq.ifgrq_group)); 1525 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1526 IF_ADDR_RUNLOCK(ifp); 1527 return (error); 1528 } 1529 len -= sizeof(ifgrq); 1530 ifgp++; 1531 } 1532 IF_ADDR_RUNLOCK(ifp); 1533 1534 return (0); 1535} 1536 1537/* 1538 * Stores all members of a group in memory pointed to by data 1539 */ 1540static int 1541if_getgroupmembers(struct ifgroupreq *data) 1542{ 1543 struct ifgroupreq *ifgr = data; 1544 struct ifg_group *ifg; 1545 struct ifg_member *ifgm; 1546 struct ifg_req ifgrq, *ifgp; 1547 int len, error; 1548 1549 IFNET_RLOCK(); 1550 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1551 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1552 break; 1553 if (ifg == NULL) { 1554 IFNET_RUNLOCK(); 1555 return (ENOENT); 1556 } 1557 1558 if (ifgr->ifgr_len == 0) { 1559 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1560 ifgr->ifgr_len += sizeof(ifgrq); 1561 IFNET_RUNLOCK(); 1562 return (0); 1563 } 1564 1565 len = ifgr->ifgr_len; 1566 ifgp = ifgr->ifgr_groups; 1567 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1568 if (len < sizeof(ifgrq)) { 1569 IFNET_RUNLOCK(); 1570 return (EINVAL); 1571 } 1572 bzero(&ifgrq, sizeof ifgrq); 1573 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1574 sizeof(ifgrq.ifgrq_member)); 1575 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1576 IFNET_RUNLOCK(); 1577 return (error); 1578 } 1579 len -= sizeof(ifgrq); 1580 ifgp++; 1581 } 1582 IFNET_RUNLOCK(); 1583 1584 return (0); 1585} 1586 1587/* 1588 * Return counter values from counter(9)s stored in ifnet. 1589 */ 1590uint64_t 1591if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1592{ 1593 1594 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1595 1596 return (counter_u64_fetch(ifp->if_counters[cnt])); 1597} 1598 1599/* 1600 * Increase an ifnet counter. Usually used for counters shared 1601 * between the stack and a driver, but function supports them all. 1602 */ 1603void 1604if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1605{ 1606 1607 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1608 1609 counter_u64_add(ifp->if_counters[cnt], inc); 1610} 1611 1612/* 1613 * Copy data from ifnet to userland API structure if_data. 1614 */ 1615void 1616if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1617{ 1618 1619 ifd->ifi_type = ifp->if_type; 1620 ifd->ifi_physical = 0; 1621 ifd->ifi_addrlen = ifp->if_addrlen; 1622 ifd->ifi_hdrlen = ifp->if_hdrlen; 1623 ifd->ifi_link_state = ifp->if_link_state; 1624 ifd->ifi_vhid = 0; 1625 ifd->ifi_datalen = sizeof(struct if_data); 1626 ifd->ifi_mtu = ifp->if_mtu; 1627 ifd->ifi_metric = ifp->if_metric; 1628 ifd->ifi_baudrate = ifp->if_baudrate; 1629 ifd->ifi_hwassist = ifp->if_hwassist; 1630 ifd->ifi_epoch = ifp->if_epoch; 1631 ifd->ifi_lastchange = ifp->if_lastchange; 1632 1633 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1634 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1635 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1636 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1637 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1638 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1639 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1640 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1641 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1642 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1643 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1644 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1645} 1646 1647/* 1648 * Wrapper functions for struct ifnet address list locking macros. These are 1649 * used by kernel modules to avoid encoding programming interface or binary 1650 * interface assumptions that may be violated when kernel-internal locking 1651 * approaches change. 1652 */ 1653void 1654if_addr_rlock(struct ifnet *ifp) 1655{ 1656 1657 IF_ADDR_RLOCK(ifp); 1658} 1659 1660void 1661if_addr_runlock(struct ifnet *ifp) 1662{ 1663 1664 IF_ADDR_RUNLOCK(ifp); 1665} 1666 1667void 1668if_maddr_rlock(if_t ifp) 1669{ 1670 1671 IF_ADDR_RLOCK((struct ifnet *)ifp); 1672} 1673 1674void 1675if_maddr_runlock(if_t ifp) 1676{ 1677 1678 IF_ADDR_RUNLOCK((struct ifnet *)ifp); 1679} 1680 1681/* 1682 * Initialization, destruction and refcounting functions for ifaddrs. 1683 */ 1684struct ifaddr * 1685ifa_alloc(size_t size, int flags) 1686{ 1687 struct ifaddr *ifa; 1688 1689 KASSERT(size >= sizeof(struct ifaddr), 1690 ("%s: invalid size %zu", __func__, size)); 1691 1692 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1693 if (ifa == NULL) 1694 return (NULL); 1695 1696 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1697 goto fail; 1698 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1699 goto fail; 1700 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1701 goto fail; 1702 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1703 goto fail; 1704 1705 refcount_init(&ifa->ifa_refcnt, 1); 1706 1707 return (ifa); 1708 1709fail: 1710 /* free(NULL) is okay */ 1711 counter_u64_free(ifa->ifa_opackets); 1712 counter_u64_free(ifa->ifa_ipackets); 1713 counter_u64_free(ifa->ifa_obytes); 1714 counter_u64_free(ifa->ifa_ibytes); 1715 free(ifa, M_IFADDR); 1716 1717 return (NULL); 1718} 1719 1720void 1721ifa_ref(struct ifaddr *ifa) 1722{ 1723 1724 refcount_acquire(&ifa->ifa_refcnt); 1725} 1726 1727void 1728ifa_free(struct ifaddr *ifa) 1729{ 1730 1731 if (refcount_release(&ifa->ifa_refcnt)) { 1732 counter_u64_free(ifa->ifa_opackets); 1733 counter_u64_free(ifa->ifa_ipackets); 1734 counter_u64_free(ifa->ifa_obytes); 1735 counter_u64_free(ifa->ifa_ibytes); 1736 free(ifa, M_IFADDR); 1737 } 1738} 1739 1740static int 1741ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa, 1742 struct sockaddr *ia) 1743{ 1744 int error; 1745 struct rt_addrinfo info; 1746 struct sockaddr_dl null_sdl; 1747 struct ifnet *ifp; 1748 1749 ifp = ifa->ifa_ifp; 1750 1751 bzero(&info, sizeof(info)); 1752 if (cmd != RTM_DELETE) 1753 info.rti_ifp = V_loif; 1754 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED; 1755 info.rti_info[RTAX_DST] = ia; 1756 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1757 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type); 1758 1759 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib); 1760 1761 if (error != 0) 1762 log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n", 1763 __func__, otype, if_name(ifp), error); 1764 1765 return (error); 1766} 1767 1768int 1769ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1770{ 1771 1772 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia)); 1773} 1774 1775int 1776ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1777{ 1778 1779 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia)); 1780} 1781 1782int 1783ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1784{ 1785 1786 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia)); 1787} 1788 1789/* 1790 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1791 * structs used to represent other address families, it is necessary 1792 * to perform a different comparison. 1793 */ 1794 1795#define sa_dl_equal(a1, a2) \ 1796 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1797 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1798 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1799 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1800 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1801 1802/* 1803 * Locate an interface based on a complete address. 1804 */ 1805/*ARGSUSED*/ 1806static struct ifaddr * 1807ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref) 1808{ 1809 struct ifnet *ifp; 1810 struct ifaddr *ifa; 1811 1812 IFNET_RLOCK_NOSLEEP(); 1813 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1814 IF_ADDR_RLOCK(ifp); 1815 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1816 if (ifa->ifa_addr->sa_family != addr->sa_family) 1817 continue; 1818 if (sa_equal(addr, ifa->ifa_addr)) { 1819 if (getref) 1820 ifa_ref(ifa); 1821 IF_ADDR_RUNLOCK(ifp); 1822 goto done; 1823 } 1824 /* IP6 doesn't have broadcast */ 1825 if ((ifp->if_flags & IFF_BROADCAST) && 1826 ifa->ifa_broadaddr && 1827 ifa->ifa_broadaddr->sa_len != 0 && 1828 sa_equal(ifa->ifa_broadaddr, addr)) { 1829 if (getref) 1830 ifa_ref(ifa); 1831 IF_ADDR_RUNLOCK(ifp); 1832 goto done; 1833 } 1834 } 1835 IF_ADDR_RUNLOCK(ifp); 1836 } 1837 ifa = NULL; 1838done: 1839 IFNET_RUNLOCK_NOSLEEP(); 1840 return (ifa); 1841} 1842 1843struct ifaddr * 1844ifa_ifwithaddr(const struct sockaddr *addr) 1845{ 1846 1847 return (ifa_ifwithaddr_internal(addr, 1)); 1848} 1849 1850int 1851ifa_ifwithaddr_check(const struct sockaddr *addr) 1852{ 1853 1854 return (ifa_ifwithaddr_internal(addr, 0) != NULL); 1855} 1856 1857/* 1858 * Locate an interface based on the broadcast address. 1859 */ 1860/* ARGSUSED */ 1861struct ifaddr * 1862ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1863{ 1864 struct ifnet *ifp; 1865 struct ifaddr *ifa; 1866 1867 IFNET_RLOCK_NOSLEEP(); 1868 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1869 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1870 continue; 1871 IF_ADDR_RLOCK(ifp); 1872 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1873 if (ifa->ifa_addr->sa_family != addr->sa_family) 1874 continue; 1875 if ((ifp->if_flags & IFF_BROADCAST) && 1876 ifa->ifa_broadaddr && 1877 ifa->ifa_broadaddr->sa_len != 0 && 1878 sa_equal(ifa->ifa_broadaddr, addr)) { 1879 ifa_ref(ifa); 1880 IF_ADDR_RUNLOCK(ifp); 1881 goto done; 1882 } 1883 } 1884 IF_ADDR_RUNLOCK(ifp); 1885 } 1886 ifa = NULL; 1887done: 1888 IFNET_RUNLOCK_NOSLEEP(); 1889 return (ifa); 1890} 1891 1892/* 1893 * Locate the point to point interface with a given destination address. 1894 */ 1895/*ARGSUSED*/ 1896struct ifaddr * 1897ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1898{ 1899 struct ifnet *ifp; 1900 struct ifaddr *ifa; 1901 1902 IFNET_RLOCK_NOSLEEP(); 1903 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1904 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1905 continue; 1906 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1907 continue; 1908 IF_ADDR_RLOCK(ifp); 1909 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1910 if (ifa->ifa_addr->sa_family != addr->sa_family) 1911 continue; 1912 if (ifa->ifa_dstaddr != NULL && 1913 sa_equal(addr, ifa->ifa_dstaddr)) { 1914 ifa_ref(ifa); 1915 IF_ADDR_RUNLOCK(ifp); 1916 goto done; 1917 } 1918 } 1919 IF_ADDR_RUNLOCK(ifp); 1920 } 1921 ifa = NULL; 1922done: 1923 IFNET_RUNLOCK_NOSLEEP(); 1924 return (ifa); 1925} 1926 1927/* 1928 * Find an interface on a specific network. If many, choice 1929 * is most specific found. 1930 */ 1931struct ifaddr * 1932ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1933{ 1934 struct ifnet *ifp; 1935 struct ifaddr *ifa; 1936 struct ifaddr *ifa_maybe = NULL; 1937 u_int af = addr->sa_family; 1938 const char *addr_data = addr->sa_data, *cplim; 1939 1940 /* 1941 * AF_LINK addresses can be looked up directly by their index number, 1942 * so do that if we can. 1943 */ 1944 if (af == AF_LINK) { 1945 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 1946 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1947 return (ifaddr_byindex(sdl->sdl_index)); 1948 } 1949 1950 /* 1951 * Scan though each interface, looking for ones that have addresses 1952 * in this address family and the requested fib. Maintain a reference 1953 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 1954 * kept it stable when we move onto the next interface. 1955 */ 1956 IFNET_RLOCK_NOSLEEP(); 1957 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1958 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1959 continue; 1960 IF_ADDR_RLOCK(ifp); 1961 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1962 const char *cp, *cp2, *cp3; 1963 1964 if (ifa->ifa_addr->sa_family != af) 1965next: continue; 1966 if (af == AF_INET && 1967 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1968 /* 1969 * This is a bit broken as it doesn't 1970 * take into account that the remote end may 1971 * be a single node in the network we are 1972 * looking for. 1973 * The trouble is that we don't know the 1974 * netmask for the remote end. 1975 */ 1976 if (ifa->ifa_dstaddr != NULL && 1977 sa_equal(addr, ifa->ifa_dstaddr)) { 1978 ifa_ref(ifa); 1979 IF_ADDR_RUNLOCK(ifp); 1980 goto done; 1981 } 1982 } else { 1983 /* 1984 * Scan all the bits in the ifa's address. 1985 * If a bit dissagrees with what we are 1986 * looking for, mask it with the netmask 1987 * to see if it really matters. 1988 * (A byte at a time) 1989 */ 1990 if (ifa->ifa_netmask == 0) 1991 continue; 1992 cp = addr_data; 1993 cp2 = ifa->ifa_addr->sa_data; 1994 cp3 = ifa->ifa_netmask->sa_data; 1995 cplim = ifa->ifa_netmask->sa_len 1996 + (char *)ifa->ifa_netmask; 1997 while (cp3 < cplim) 1998 if ((*cp++ ^ *cp2++) & *cp3++) 1999 goto next; /* next address! */ 2000 /* 2001 * If the netmask of what we just found 2002 * is more specific than what we had before 2003 * (if we had one), or if the virtual status 2004 * of new prefix is better than of the old one, 2005 * then remember the new one before continuing 2006 * to search for an even better one. 2007 */ 2008 if (ifa_maybe == NULL || 2009 ifa_preferred(ifa_maybe, ifa) || 2010 rn_refines((caddr_t)ifa->ifa_netmask, 2011 (caddr_t)ifa_maybe->ifa_netmask)) { 2012 if (ifa_maybe != NULL) 2013 ifa_free(ifa_maybe); 2014 ifa_maybe = ifa; 2015 ifa_ref(ifa_maybe); 2016 } 2017 } 2018 } 2019 IF_ADDR_RUNLOCK(ifp); 2020 } 2021 ifa = ifa_maybe; 2022 ifa_maybe = NULL; 2023done: 2024 IFNET_RUNLOCK_NOSLEEP(); 2025 if (ifa_maybe != NULL) 2026 ifa_free(ifa_maybe); 2027 return (ifa); 2028} 2029 2030/* 2031 * Find an interface address specific to an interface best matching 2032 * a given address. 2033 */ 2034struct ifaddr * 2035ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2036{ 2037 struct ifaddr *ifa; 2038 const char *cp, *cp2, *cp3; 2039 char *cplim; 2040 struct ifaddr *ifa_maybe = NULL; 2041 u_int af = addr->sa_family; 2042 2043 if (af >= AF_MAX) 2044 return (NULL); 2045 IF_ADDR_RLOCK(ifp); 2046 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2047 if (ifa->ifa_addr->sa_family != af) 2048 continue; 2049 if (ifa_maybe == NULL) 2050 ifa_maybe = ifa; 2051 if (ifa->ifa_netmask == 0) { 2052 if (sa_equal(addr, ifa->ifa_addr) || 2053 (ifa->ifa_dstaddr && 2054 sa_equal(addr, ifa->ifa_dstaddr))) 2055 goto done; 2056 continue; 2057 } 2058 if (ifp->if_flags & IFF_POINTOPOINT) { 2059 if (sa_equal(addr, ifa->ifa_dstaddr)) 2060 goto done; 2061 } else { 2062 cp = addr->sa_data; 2063 cp2 = ifa->ifa_addr->sa_data; 2064 cp3 = ifa->ifa_netmask->sa_data; 2065 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2066 for (; cp3 < cplim; cp3++) 2067 if ((*cp++ ^ *cp2++) & *cp3) 2068 break; 2069 if (cp3 == cplim) 2070 goto done; 2071 } 2072 } 2073 ifa = ifa_maybe; 2074done: 2075 if (ifa != NULL) 2076 ifa_ref(ifa); 2077 IF_ADDR_RUNLOCK(ifp); 2078 return (ifa); 2079} 2080 2081/* 2082 * See whether new ifa is better than current one: 2083 * 1) A non-virtual one is preferred over virtual. 2084 * 2) A virtual in master state preferred over any other state. 2085 * 2086 * Used in several address selecting functions. 2087 */ 2088int 2089ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2090{ 2091 2092 return (cur->ifa_carp && (!next->ifa_carp || 2093 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2094} 2095 2096#include <net/if_llatbl.h> 2097 2098/* 2099 * Default action when installing a route with a Link Level gateway. 2100 * Lookup an appropriate real ifa to point to. 2101 * This should be moved to /sys/net/link.c eventually. 2102 */ 2103static void 2104link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 2105{ 2106 struct ifaddr *ifa, *oifa; 2107 struct sockaddr *dst; 2108 struct ifnet *ifp; 2109 2110 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) || 2111 ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL)) 2112 return; 2113 ifa = ifaof_ifpforaddr(dst, ifp); 2114 if (ifa) { 2115 oifa = rt->rt_ifa; 2116 rt->rt_ifa = ifa; 2117 ifa_free(oifa); 2118 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2119 ifa->ifa_rtrequest(cmd, rt, info); 2120 } 2121} 2122 2123struct sockaddr_dl * 2124link_alloc_sdl(size_t size, int flags) 2125{ 2126 2127 return (malloc(size, M_TEMP, flags)); 2128} 2129 2130void 2131link_free_sdl(struct sockaddr *sa) 2132{ 2133 free(sa, M_TEMP); 2134} 2135 2136/* 2137 * Fills in given sdl with interface basic info. 2138 * Returns pointer to filled sdl. 2139 */ 2140struct sockaddr_dl * 2141link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2142{ 2143 struct sockaddr_dl *sdl; 2144 2145 sdl = (struct sockaddr_dl *)paddr; 2146 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2147 sdl->sdl_len = sizeof(struct sockaddr_dl); 2148 sdl->sdl_family = AF_LINK; 2149 sdl->sdl_index = ifp->if_index; 2150 sdl->sdl_type = iftype; 2151 2152 return (sdl); 2153} 2154 2155/* 2156 * Mark an interface down and notify protocols of 2157 * the transition. 2158 */ 2159static void 2160if_unroute(struct ifnet *ifp, int flag, int fam) 2161{ 2162 struct ifaddr *ifa; 2163 2164 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2165 2166 ifp->if_flags &= ~flag; 2167 getmicrotime(&ifp->if_lastchange); 2168 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2169 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2170 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2171 ifp->if_qflush(ifp); 2172 2173 if (ifp->if_carp) 2174 (*carp_linkstate_p)(ifp); 2175 rt_ifmsg(ifp); 2176} 2177 2178/* 2179 * Mark an interface up and notify protocols of 2180 * the transition. 2181 */ 2182static void 2183if_route(struct ifnet *ifp, int flag, int fam) 2184{ 2185 struct ifaddr *ifa; 2186 2187 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2188 2189 ifp->if_flags |= flag; 2190 getmicrotime(&ifp->if_lastchange); 2191 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2192 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2193 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2194 if (ifp->if_carp) 2195 (*carp_linkstate_p)(ifp); 2196 rt_ifmsg(ifp); 2197#ifdef INET6 2198 in6_if_up(ifp); 2199#endif 2200} 2201 2202void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2203void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2204struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2205struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2206int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2207int (*vlan_setcookie_p)(struct ifnet *, void *); 2208void *(*vlan_cookie_p)(struct ifnet *); 2209 2210/* 2211 * Handle a change in the interface link state. To avoid LORs 2212 * between driver lock and upper layer locks, as well as possible 2213 * recursions, we post event to taskqueue, and all job 2214 * is done in static do_link_state_change(). 2215 */ 2216void 2217if_link_state_change(struct ifnet *ifp, int link_state) 2218{ 2219 /* Return if state hasn't changed. */ 2220 if (ifp->if_link_state == link_state) 2221 return; 2222 2223 ifp->if_link_state = link_state; 2224 2225 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2226} 2227 2228static void 2229do_link_state_change(void *arg, int pending) 2230{ 2231 struct ifnet *ifp = (struct ifnet *)arg; 2232 int link_state = ifp->if_link_state; 2233 CURVNET_SET(ifp->if_vnet); 2234 2235 /* Notify that the link state has changed. */ 2236 rt_ifmsg(ifp); 2237 if (ifp->if_vlantrunk != NULL) 2238 (*vlan_link_state_p)(ifp); 2239 2240 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2241 ifp->if_l2com != NULL) 2242 (*ng_ether_link_state_p)(ifp, link_state); 2243 if (ifp->if_carp) 2244 (*carp_linkstate_p)(ifp); 2245 if (ifp->if_bridge) 2246 (*bridge_linkstate_p)(ifp); 2247 if (ifp->if_lagg) 2248 (*lagg_linkstate_p)(ifp, link_state); 2249 2250 if (IS_DEFAULT_VNET(curvnet)) 2251 devctl_notify("IFNET", ifp->if_xname, 2252 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2253 NULL); 2254 if (pending > 1) 2255 if_printf(ifp, "%d link states coalesced\n", pending); 2256 if (log_link_state_change) 2257 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 2258 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2259 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2260 CURVNET_RESTORE(); 2261} 2262 2263/* 2264 * Mark an interface down and notify protocols of 2265 * the transition. 2266 */ 2267void 2268if_down(struct ifnet *ifp) 2269{ 2270 2271 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2272 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2273} 2274 2275/* 2276 * Mark an interface up and notify protocols of 2277 * the transition. 2278 */ 2279void 2280if_up(struct ifnet *ifp) 2281{ 2282 2283 if_route(ifp, IFF_UP, AF_UNSPEC); 2284 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2285} 2286 2287/* 2288 * Flush an interface queue. 2289 */ 2290void 2291if_qflush(struct ifnet *ifp) 2292{ 2293 struct mbuf *m, *n; 2294 struct ifaltq *ifq; 2295 2296 ifq = &ifp->if_snd; 2297 IFQ_LOCK(ifq); 2298#ifdef ALTQ 2299 if (ALTQ_IS_ENABLED(ifq)) 2300 ALTQ_PURGE(ifq); 2301#endif 2302 n = ifq->ifq_head; 2303 while ((m = n) != NULL) { 2304 n = m->m_nextpkt; 2305 m_freem(m); 2306 } 2307 ifq->ifq_head = 0; 2308 ifq->ifq_tail = 0; 2309 ifq->ifq_len = 0; 2310 IFQ_UNLOCK(ifq); 2311} 2312 2313/* 2314 * Map interface name to interface structure pointer, with or without 2315 * returning a reference. 2316 */ 2317struct ifnet * 2318ifunit_ref(const char *name) 2319{ 2320 struct ifnet *ifp; 2321 2322 IFNET_RLOCK_NOSLEEP(); 2323 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2324 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2325 !(ifp->if_flags & IFF_DYING)) 2326 break; 2327 } 2328 if (ifp != NULL) 2329 if_ref(ifp); 2330 IFNET_RUNLOCK_NOSLEEP(); 2331 return (ifp); 2332} 2333 2334struct ifnet * 2335ifunit(const char *name) 2336{ 2337 struct ifnet *ifp; 2338 2339 IFNET_RLOCK_NOSLEEP(); 2340 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2341 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2342 break; 2343 } 2344 IFNET_RUNLOCK_NOSLEEP(); 2345 return (ifp); 2346} 2347 2348static void * 2349ifr_buffer_get_buffer(void *data) 2350{ 2351 union ifreq_union *ifrup; 2352 2353 ifrup = data; 2354#ifdef COMPAT_FREEBSD32 2355 if (SV_CURPROC_FLAG(SV_ILP32)) 2356 return ((void *)(uintptr_t) 2357 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2358#endif 2359 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2360} 2361 2362static void 2363ifr_buffer_set_buffer_null(void *data) 2364{ 2365 union ifreq_union *ifrup; 2366 2367 ifrup = data; 2368#ifdef COMPAT_FREEBSD32 2369 if (SV_CURPROC_FLAG(SV_ILP32)) 2370 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2371 else 2372#endif 2373 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2374} 2375 2376static size_t 2377ifr_buffer_get_length(void *data) 2378{ 2379 union ifreq_union *ifrup; 2380 2381 ifrup = data; 2382#ifdef COMPAT_FREEBSD32 2383 if (SV_CURPROC_FLAG(SV_ILP32)) 2384 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2385#endif 2386 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2387} 2388 2389static void 2390ifr_buffer_set_length(void *data, size_t len) 2391{ 2392 union ifreq_union *ifrup; 2393 2394 ifrup = data; 2395#ifdef COMPAT_FREEBSD32 2396 if (SV_CURPROC_FLAG(SV_ILP32)) 2397 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2398 else 2399#endif 2400 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2401} 2402 2403void * 2404ifr_data_get_ptr(void *ifrp) 2405{ 2406 union ifreq_union *ifrup; 2407 2408 ifrup = ifrp; 2409#ifdef COMPAT_FREEBSD32 2410 if (SV_CURPROC_FLAG(SV_ILP32)) 2411 return ((void *)(uintptr_t) 2412 ifrup->ifr32.ifr_ifru.ifru_data); 2413#endif 2414 return (ifrup->ifr.ifr_ifru.ifru_data); 2415} 2416 2417/* 2418 * Hardware specific interface ioctls. 2419 */ 2420static int 2421ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2422{ 2423 struct ifreq *ifr; 2424 int error = 0, do_ifup = 0; 2425 int new_flags, temp_flags; 2426 size_t namelen, onamelen; 2427 size_t descrlen; 2428 char *descrbuf, *odescrbuf; 2429 char new_name[IFNAMSIZ]; 2430 struct ifaddr *ifa; 2431 struct sockaddr_dl *sdl; 2432 2433 ifr = (struct ifreq *)data; 2434 switch (cmd) { 2435 case SIOCGIFINDEX: 2436 ifr->ifr_index = ifp->if_index; 2437 break; 2438 2439 case SIOCGIFFLAGS: 2440 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2441 ifr->ifr_flags = temp_flags & 0xffff; 2442 ifr->ifr_flagshigh = temp_flags >> 16; 2443 break; 2444 2445 case SIOCGIFCAP: 2446 ifr->ifr_reqcap = ifp->if_capabilities; 2447 ifr->ifr_curcap = ifp->if_capenable; 2448 break; 2449 2450#ifdef MAC 2451 case SIOCGIFMAC: 2452 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2453 break; 2454#endif 2455 2456 case SIOCGIFMETRIC: 2457 ifr->ifr_metric = ifp->if_metric; 2458 break; 2459 2460 case SIOCGIFMTU: 2461 ifr->ifr_mtu = ifp->if_mtu; 2462 break; 2463 2464 case SIOCGIFPHYS: 2465 /* XXXGL: did this ever worked? */ 2466 ifr->ifr_phys = 0; 2467 break; 2468 2469 case SIOCGIFDESCR: 2470 error = 0; 2471 sx_slock(&ifdescr_sx); 2472 if (ifp->if_description == NULL) 2473 error = ENOMSG; 2474 else { 2475 /* space for terminating nul */ 2476 descrlen = strlen(ifp->if_description) + 1; 2477 if (ifr_buffer_get_length(ifr) < descrlen) 2478 ifr_buffer_set_buffer_null(ifr); 2479 else 2480 error = copyout(ifp->if_description, 2481 ifr_buffer_get_buffer(ifr), descrlen); 2482 ifr_buffer_set_length(ifr, descrlen); 2483 } 2484 sx_sunlock(&ifdescr_sx); 2485 break; 2486 2487 case SIOCSIFDESCR: 2488 error = priv_check(td, PRIV_NET_SETIFDESCR); 2489 if (error) 2490 return (error); 2491 2492 /* 2493 * Copy only (length-1) bytes to make sure that 2494 * if_description is always nul terminated. The 2495 * length parameter is supposed to count the 2496 * terminating nul in. 2497 */ 2498 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2499 return (ENAMETOOLONG); 2500 else if (ifr_buffer_get_length(ifr) == 0) 2501 descrbuf = NULL; 2502 else { 2503 descrbuf = malloc(ifr_buffer_get_length(ifr), 2504 M_IFDESCR, M_WAITOK | M_ZERO); 2505 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2506 ifr_buffer_get_length(ifr) - 1); 2507 if (error) { 2508 free(descrbuf, M_IFDESCR); 2509 break; 2510 } 2511 } 2512 2513 sx_xlock(&ifdescr_sx); 2514 odescrbuf = ifp->if_description; 2515 ifp->if_description = descrbuf; 2516 sx_xunlock(&ifdescr_sx); 2517 2518 getmicrotime(&ifp->if_lastchange); 2519 free(odescrbuf, M_IFDESCR); 2520 break; 2521 2522 case SIOCGIFFIB: 2523 ifr->ifr_fib = ifp->if_fib; 2524 break; 2525 2526 case SIOCSIFFIB: 2527 error = priv_check(td, PRIV_NET_SETIFFIB); 2528 if (error) 2529 return (error); 2530 if (ifr->ifr_fib >= rt_numfibs) 2531 return (EINVAL); 2532 2533 ifp->if_fib = ifr->ifr_fib; 2534 break; 2535 2536 case SIOCSIFFLAGS: 2537 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2538 if (error) 2539 return (error); 2540 /* 2541 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2542 * check, so we don't need special handling here yet. 2543 */ 2544 new_flags = (ifr->ifr_flags & 0xffff) | 2545 (ifr->ifr_flagshigh << 16); 2546 if (ifp->if_flags & IFF_UP && 2547 (new_flags & IFF_UP) == 0) { 2548 if_down(ifp); 2549 } else if (new_flags & IFF_UP && 2550 (ifp->if_flags & IFF_UP) == 0) { 2551 do_ifup = 1; 2552 } 2553 /* See if permanently promiscuous mode bit is about to flip */ 2554 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2555 if (new_flags & IFF_PPROMISC) 2556 ifp->if_flags |= IFF_PROMISC; 2557 else if (ifp->if_pcount == 0) 2558 ifp->if_flags &= ~IFF_PROMISC; 2559 if (log_promisc_mode_change) 2560 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2561 ifp->if_xname, 2562 ((new_flags & IFF_PPROMISC) ? 2563 "enabled" : "disabled")); 2564 } 2565 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2566 (new_flags &~ IFF_CANTCHANGE); 2567 if (ifp->if_ioctl) { 2568 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2569 } 2570 if (do_ifup) 2571 if_up(ifp); 2572 getmicrotime(&ifp->if_lastchange); 2573 break; 2574 2575 case SIOCSIFCAP: 2576 error = priv_check(td, PRIV_NET_SETIFCAP); 2577 if (error) 2578 return (error); 2579 if (ifp->if_ioctl == NULL) 2580 return (EOPNOTSUPP); 2581 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2582 return (EINVAL); 2583 error = (*ifp->if_ioctl)(ifp, cmd, data); 2584 if (error == 0) 2585 getmicrotime(&ifp->if_lastchange); 2586 break; 2587 2588#ifdef MAC 2589 case SIOCSIFMAC: 2590 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2591 break; 2592#endif 2593 2594 case SIOCSIFNAME: 2595 error = priv_check(td, PRIV_NET_SETIFNAME); 2596 if (error) 2597 return (error); 2598 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2599 NULL); 2600 if (error != 0) 2601 return (error); 2602 if (new_name[0] == '\0') 2603 return (EINVAL); 2604 if (new_name[IFNAMSIZ-1] != '\0') { 2605 new_name[IFNAMSIZ-1] = '\0'; 2606 if (strlen(new_name) == IFNAMSIZ-1) 2607 return (EINVAL); 2608 } 2609 if (ifunit(new_name) != NULL) 2610 return (EEXIST); 2611 2612 /* 2613 * XXX: Locking. Nothing else seems to lock if_flags, 2614 * and there are numerous other races with the 2615 * ifunit() checks not being atomic with namespace 2616 * changes (renames, vmoves, if_attach, etc). 2617 */ 2618 ifp->if_flags |= IFF_RENAMING; 2619 2620 /* Announce the departure of the interface. */ 2621 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2622 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2623 2624 log(LOG_INFO, "%s: changing name to '%s'\n", 2625 ifp->if_xname, new_name); 2626 2627 IF_ADDR_WLOCK(ifp); 2628 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2629 ifa = ifp->if_addr; 2630 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2631 namelen = strlen(new_name); 2632 onamelen = sdl->sdl_nlen; 2633 /* 2634 * Move the address if needed. This is safe because we 2635 * allocate space for a name of length IFNAMSIZ when we 2636 * create this in if_attach(). 2637 */ 2638 if (namelen != onamelen) { 2639 bcopy(sdl->sdl_data + onamelen, 2640 sdl->sdl_data + namelen, sdl->sdl_alen); 2641 } 2642 bcopy(new_name, sdl->sdl_data, namelen); 2643 sdl->sdl_nlen = namelen; 2644 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2645 bzero(sdl->sdl_data, onamelen); 2646 while (namelen != 0) 2647 sdl->sdl_data[--namelen] = 0xff; 2648 IF_ADDR_WUNLOCK(ifp); 2649 2650 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2651 /* Announce the return of the interface. */ 2652 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2653 2654 ifp->if_flags &= ~IFF_RENAMING; 2655 break; 2656 2657#ifdef VIMAGE 2658 case SIOCSIFVNET: 2659 error = priv_check(td, PRIV_NET_SETIFVNET); 2660 if (error) 2661 return (error); 2662 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2663 break; 2664#endif 2665 2666 case SIOCSIFMETRIC: 2667 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2668 if (error) 2669 return (error); 2670 ifp->if_metric = ifr->ifr_metric; 2671 getmicrotime(&ifp->if_lastchange); 2672 break; 2673 2674 case SIOCSIFPHYS: 2675 error = priv_check(td, PRIV_NET_SETIFPHYS); 2676 if (error) 2677 return (error); 2678 if (ifp->if_ioctl == NULL) 2679 return (EOPNOTSUPP); 2680 error = (*ifp->if_ioctl)(ifp, cmd, data); 2681 if (error == 0) 2682 getmicrotime(&ifp->if_lastchange); 2683 break; 2684 2685 case SIOCSIFMTU: 2686 { 2687 u_long oldmtu = ifp->if_mtu; 2688 2689 error = priv_check(td, PRIV_NET_SETIFMTU); 2690 if (error) 2691 return (error); 2692 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2693 return (EINVAL); 2694 if (ifp->if_ioctl == NULL) 2695 return (EOPNOTSUPP); 2696 error = (*ifp->if_ioctl)(ifp, cmd, data); 2697 if (error == 0) { 2698 getmicrotime(&ifp->if_lastchange); 2699 rt_ifmsg(ifp); 2700 } 2701 /* 2702 * If the link MTU changed, do network layer specific procedure. 2703 */ 2704 if (ifp->if_mtu != oldmtu) { 2705#ifdef INET6 2706 nd6_setmtu(ifp); 2707#endif 2708 rt_updatemtu(ifp); 2709 } 2710 break; 2711 } 2712 2713 case SIOCADDMULTI: 2714 case SIOCDELMULTI: 2715 if (cmd == SIOCADDMULTI) 2716 error = priv_check(td, PRIV_NET_ADDMULTI); 2717 else 2718 error = priv_check(td, PRIV_NET_DELMULTI); 2719 if (error) 2720 return (error); 2721 2722 /* Don't allow group membership on non-multicast interfaces. */ 2723 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2724 return (EOPNOTSUPP); 2725 2726 /* Don't let users screw up protocols' entries. */ 2727 if (ifr->ifr_addr.sa_family != AF_LINK) 2728 return (EINVAL); 2729 2730 if (cmd == SIOCADDMULTI) { 2731 struct ifmultiaddr *ifma; 2732 2733 /* 2734 * Userland is only permitted to join groups once 2735 * via the if_addmulti() KPI, because it cannot hold 2736 * struct ifmultiaddr * between calls. It may also 2737 * lose a race while we check if the membership 2738 * already exists. 2739 */ 2740 IF_ADDR_RLOCK(ifp); 2741 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2742 IF_ADDR_RUNLOCK(ifp); 2743 if (ifma != NULL) 2744 error = EADDRINUSE; 2745 else 2746 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2747 } else { 2748 error = if_delmulti(ifp, &ifr->ifr_addr); 2749 } 2750 if (error == 0) 2751 getmicrotime(&ifp->if_lastchange); 2752 break; 2753 2754 case SIOCSIFPHYADDR: 2755 case SIOCDIFPHYADDR: 2756#ifdef INET6 2757 case SIOCSIFPHYADDR_IN6: 2758#endif 2759 case SIOCSIFMEDIA: 2760 case SIOCSIFGENERIC: 2761 error = priv_check(td, PRIV_NET_HWIOCTL); 2762 if (error) 2763 return (error); 2764 if (ifp->if_ioctl == NULL) 2765 return (EOPNOTSUPP); 2766 error = (*ifp->if_ioctl)(ifp, cmd, data); 2767 if (error == 0) 2768 getmicrotime(&ifp->if_lastchange); 2769 break; 2770 2771 case SIOCGIFSTATUS: 2772 case SIOCGIFPSRCADDR: 2773 case SIOCGIFPDSTADDR: 2774 case SIOCGIFMEDIA: 2775 case SIOCGIFXMEDIA: 2776 case SIOCGIFGENERIC: 2777 case SIOCGIFRSSKEY: 2778 case SIOCGIFRSSHASH: 2779 if (ifp->if_ioctl == NULL) 2780 return (EOPNOTSUPP); 2781 error = (*ifp->if_ioctl)(ifp, cmd, data); 2782 break; 2783 2784 case SIOCSIFLLADDR: 2785 error = priv_check(td, PRIV_NET_SETLLADDR); 2786 if (error) 2787 return (error); 2788 error = if_setlladdr(ifp, 2789 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2790 break; 2791 2792 case SIOCGHWADDR: 2793 error = if_gethwaddr(ifp, ifr); 2794 break; 2795 2796 case SIOCAIFGROUP: 2797 { 2798 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2799 2800 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2801 if (error) 2802 return (error); 2803 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2804 return (error); 2805 break; 2806 } 2807 2808 case SIOCGIFGROUP: 2809 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2810 return (error); 2811 break; 2812 2813 case SIOCDIFGROUP: 2814 { 2815 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2816 2817 error = priv_check(td, PRIV_NET_DELIFGROUP); 2818 if (error) 2819 return (error); 2820 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2821 return (error); 2822 break; 2823 } 2824 2825 default: 2826 error = ENOIOCTL; 2827 break; 2828 } 2829 return (error); 2830} 2831 2832/* COMPAT_SVR4 */ 2833#define OSIOCGIFCONF _IOWR('i', 20, struct ifconf) 2834 2835#ifdef COMPAT_FREEBSD32 2836struct ifconf32 { 2837 int32_t ifc_len; 2838 union { 2839 uint32_t ifcu_buf; 2840 uint32_t ifcu_req; 2841 } ifc_ifcu; 2842}; 2843#define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2844#endif 2845 2846/* 2847 * Interface ioctls. 2848 */ 2849int 2850ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2851{ 2852 struct ifnet *ifp; 2853 struct ifreq *ifr; 2854 int error; 2855 int oif_flags; 2856#ifdef VIMAGE 2857 int shutdown; 2858#endif 2859 2860 CURVNET_SET(so->so_vnet); 2861#ifdef VIMAGE 2862 /* Make sure the VNET is stable. */ 2863 shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET && 2864 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 2865 if (shutdown) { 2866 CURVNET_RESTORE(); 2867 return (EBUSY); 2868 } 2869#endif 2870 2871 2872 switch (cmd) { 2873 case SIOCGIFCONF: 2874 case OSIOCGIFCONF: /* COMPAT_SVR4 */ 2875 error = ifconf(cmd, data); 2876 CURVNET_RESTORE(); 2877 return (error); 2878 2879#ifdef COMPAT_FREEBSD32 2880 case SIOCGIFCONF32: 2881 { 2882 struct ifconf32 *ifc32; 2883 struct ifconf ifc; 2884 2885 ifc32 = (struct ifconf32 *)data; 2886 ifc.ifc_len = ifc32->ifc_len; 2887 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2888 2889 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2890 CURVNET_RESTORE(); 2891 if (error == 0) 2892 ifc32->ifc_len = ifc.ifc_len; 2893 return (error); 2894 } 2895#endif 2896 } 2897 ifr = (struct ifreq *)data; 2898 2899 switch (cmd) { 2900#ifdef VIMAGE 2901 case SIOCSIFRVNET: 2902 error = priv_check(td, PRIV_NET_SETIFVNET); 2903 if (error == 0) 2904 error = if_vmove_reclaim(td, ifr->ifr_name, 2905 ifr->ifr_jid); 2906 CURVNET_RESTORE(); 2907 return (error); 2908#endif 2909 case SIOCIFCREATE: 2910 case SIOCIFCREATE2: 2911 error = priv_check(td, PRIV_NET_IFCREATE); 2912 if (error == 0) 2913 error = if_clone_create(ifr->ifr_name, 2914 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 2915 ifr_data_get_ptr(ifr) : NULL); 2916 CURVNET_RESTORE(); 2917 return (error); 2918 case SIOCIFDESTROY: 2919 error = priv_check(td, PRIV_NET_IFDESTROY); 2920 if (error == 0) 2921 error = if_clone_destroy(ifr->ifr_name); 2922 CURVNET_RESTORE(); 2923 return (error); 2924 2925 case SIOCIFGCLONERS: 2926 error = if_clone_list((struct if_clonereq *)data); 2927 CURVNET_RESTORE(); 2928 return (error); 2929 case SIOCGIFGMEMB: 2930 error = if_getgroupmembers((struct ifgroupreq *)data); 2931 CURVNET_RESTORE(); 2932 return (error); 2933#if defined(INET) || defined(INET6) 2934 case SIOCSVH: 2935 case SIOCGVH: 2936 if (carp_ioctl_p == NULL) 2937 error = EPROTONOSUPPORT; 2938 else 2939 error = (*carp_ioctl_p)(ifr, cmd, td); 2940 CURVNET_RESTORE(); 2941 return (error); 2942#endif 2943 } 2944 2945 ifp = ifunit_ref(ifr->ifr_name); 2946 if (ifp == NULL) { 2947 CURVNET_RESTORE(); 2948 return (ENXIO); 2949 } 2950 2951 error = ifhwioctl(cmd, ifp, data, td); 2952 if (error != ENOIOCTL) { 2953 if_rele(ifp); 2954 CURVNET_RESTORE(); 2955 return (error); 2956 } 2957 2958 oif_flags = ifp->if_flags; 2959 if (so->so_proto == NULL) { 2960 if_rele(ifp); 2961 CURVNET_RESTORE(); 2962 return (EOPNOTSUPP); 2963 } 2964 2965 /* 2966 * Pass the request on to the socket control method, and if the 2967 * latter returns EOPNOTSUPP, directly to the interface. 2968 * 2969 * Make an exception for the legacy SIOCSIF* requests. Drivers 2970 * trust SIOCSIFADDR et al to come from an already privileged 2971 * layer, and do not perform any credentials checks or input 2972 * validation. 2973 */ 2974 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 2975 ifp, td)); 2976 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2977 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2978 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2979 error = (*ifp->if_ioctl)(ifp, cmd, data); 2980 2981 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2982#ifdef INET6 2983 if (ifp->if_flags & IFF_UP) 2984 in6_if_up(ifp); 2985#endif 2986 } 2987 if_rele(ifp); 2988 CURVNET_RESTORE(); 2989 return (error); 2990} 2991 2992/* 2993 * The code common to handling reference counted flags, 2994 * e.g., in ifpromisc() and if_allmulti(). 2995 * The "pflag" argument can specify a permanent mode flag to check, 2996 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2997 * 2998 * Only to be used on stack-owned flags, not driver-owned flags. 2999 */ 3000static int 3001if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3002{ 3003 struct ifreq ifr; 3004 int error; 3005 int oldflags, oldcount; 3006 3007 /* Sanity checks to catch programming errors */ 3008 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3009 ("%s: setting driver-owned flag %d", __func__, flag)); 3010 3011 if (onswitch) 3012 KASSERT(*refcount >= 0, 3013 ("%s: increment negative refcount %d for flag %d", 3014 __func__, *refcount, flag)); 3015 else 3016 KASSERT(*refcount > 0, 3017 ("%s: decrement non-positive refcount %d for flag %d", 3018 __func__, *refcount, flag)); 3019 3020 /* In case this mode is permanent, just touch refcount */ 3021 if (ifp->if_flags & pflag) { 3022 *refcount += onswitch ? 1 : -1; 3023 return (0); 3024 } 3025 3026 /* Save ifnet parameters for if_ioctl() may fail */ 3027 oldcount = *refcount; 3028 oldflags = ifp->if_flags; 3029 3030 /* 3031 * See if we aren't the only and touching refcount is enough. 3032 * Actually toggle interface flag if we are the first or last. 3033 */ 3034 if (onswitch) { 3035 if ((*refcount)++) 3036 return (0); 3037 ifp->if_flags |= flag; 3038 } else { 3039 if (--(*refcount)) 3040 return (0); 3041 ifp->if_flags &= ~flag; 3042 } 3043 3044 /* Call down the driver since we've changed interface flags */ 3045 if (ifp->if_ioctl == NULL) { 3046 error = EOPNOTSUPP; 3047 goto recover; 3048 } 3049 ifr.ifr_flags = ifp->if_flags & 0xffff; 3050 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3051 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3052 if (error) 3053 goto recover; 3054 /* Notify userland that interface flags have changed */ 3055 rt_ifmsg(ifp); 3056 return (0); 3057 3058recover: 3059 /* Recover after driver error */ 3060 *refcount = oldcount; 3061 ifp->if_flags = oldflags; 3062 return (error); 3063} 3064 3065/* 3066 * Set/clear promiscuous mode on interface ifp based on the truth value 3067 * of pswitch. The calls are reference counted so that only the first 3068 * "on" request actually has an effect, as does the final "off" request. 3069 * Results are undefined if the "off" and "on" requests are not matched. 3070 */ 3071int 3072ifpromisc(struct ifnet *ifp, int pswitch) 3073{ 3074 int error; 3075 int oldflags = ifp->if_flags; 3076 3077 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3078 &ifp->if_pcount, pswitch); 3079 /* If promiscuous mode status has changed, log a message */ 3080 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3081 log_promisc_mode_change) 3082 log(LOG_INFO, "%s: promiscuous mode %s\n", 3083 ifp->if_xname, 3084 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3085 return (error); 3086} 3087 3088/* 3089 * Return interface configuration 3090 * of system. List may be used 3091 * in later ioctl's (above) to get 3092 * other information. 3093 */ 3094/*ARGSUSED*/ 3095static int 3096ifconf(u_long cmd, caddr_t data) 3097{ 3098 struct ifconf *ifc = (struct ifconf *)data; 3099 struct ifnet *ifp; 3100 struct ifaddr *ifa; 3101 struct ifreq ifr; 3102 struct sbuf *sb; 3103 int error, full = 0, valid_len, max_len; 3104 3105 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 3106 max_len = MAXPHYS - 1; 3107 3108 /* Prevent hostile input from being able to crash the system */ 3109 if (ifc->ifc_len <= 0) 3110 return (EINVAL); 3111 3112again: 3113 if (ifc->ifc_len <= max_len) { 3114 max_len = ifc->ifc_len; 3115 full = 1; 3116 } 3117 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3118 max_len = 0; 3119 valid_len = 0; 3120 3121 IFNET_RLOCK(); 3122 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3123 int addrs; 3124 3125 /* 3126 * Zero the ifr_name buffer to make sure we don't 3127 * disclose the contents of the stack. 3128 */ 3129 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 3130 3131 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3132 >= sizeof(ifr.ifr_name)) { 3133 sbuf_delete(sb); 3134 IFNET_RUNLOCK(); 3135 return (ENAMETOOLONG); 3136 } 3137 3138 addrs = 0; 3139 IF_ADDR_RLOCK(ifp); 3140 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3141 struct sockaddr *sa = ifa->ifa_addr; 3142 3143 if (prison_if(curthread->td_ucred, sa) != 0) 3144 continue; 3145 addrs++; 3146 /* COMPAT_SVR4 */ 3147 if (cmd == OSIOCGIFCONF) { 3148 struct osockaddr *osa = 3149 (struct osockaddr *)&ifr.ifr_addr; 3150 ifr.ifr_addr = *sa; 3151 osa->sa_family = sa->sa_family; 3152 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3153 max_len += sizeof(ifr); 3154 } else 3155 if (sa->sa_len <= sizeof(*sa)) { 3156 ifr.ifr_addr = *sa; 3157 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3158 max_len += sizeof(ifr); 3159 } else { 3160 sbuf_bcat(sb, &ifr, 3161 offsetof(struct ifreq, ifr_addr)); 3162 max_len += offsetof(struct ifreq, ifr_addr); 3163 sbuf_bcat(sb, sa, sa->sa_len); 3164 max_len += sa->sa_len; 3165 } 3166 3167 if (sbuf_error(sb) == 0) 3168 valid_len = sbuf_len(sb); 3169 } 3170 IF_ADDR_RUNLOCK(ifp); 3171 if (addrs == 0) { 3172 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 3173 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3174 max_len += sizeof(ifr); 3175 3176 if (sbuf_error(sb) == 0) 3177 valid_len = sbuf_len(sb); 3178 } 3179 } 3180 IFNET_RUNLOCK(); 3181 3182 /* 3183 * If we didn't allocate enough space (uncommon), try again. If 3184 * we have already allocated as much space as we are allowed, 3185 * return what we've got. 3186 */ 3187 if (valid_len != max_len && !full) { 3188 sbuf_delete(sb); 3189 goto again; 3190 } 3191 3192 ifc->ifc_len = valid_len; 3193 sbuf_finish(sb); 3194 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3195 sbuf_delete(sb); 3196 return (error); 3197} 3198 3199/* 3200 * Just like ifpromisc(), but for all-multicast-reception mode. 3201 */ 3202int 3203if_allmulti(struct ifnet *ifp, int onswitch) 3204{ 3205 3206 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3207} 3208 3209struct ifmultiaddr * 3210if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3211{ 3212 struct ifmultiaddr *ifma; 3213 3214 IF_ADDR_LOCK_ASSERT(ifp); 3215 3216 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3217 if (sa->sa_family == AF_LINK) { 3218 if (sa_dl_equal(ifma->ifma_addr, sa)) 3219 break; 3220 } else { 3221 if (sa_equal(ifma->ifma_addr, sa)) 3222 break; 3223 } 3224 } 3225 3226 return ifma; 3227} 3228 3229/* 3230 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3231 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3232 * the ifnet multicast address list here, so the caller must do that and 3233 * other setup work (such as notifying the device driver). The reference 3234 * count is initialized to 1. 3235 */ 3236static struct ifmultiaddr * 3237if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3238 int mflags) 3239{ 3240 struct ifmultiaddr *ifma; 3241 struct sockaddr *dupsa; 3242 3243 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3244 M_ZERO); 3245 if (ifma == NULL) 3246 return (NULL); 3247 3248 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3249 if (dupsa == NULL) { 3250 free(ifma, M_IFMADDR); 3251 return (NULL); 3252 } 3253 bcopy(sa, dupsa, sa->sa_len); 3254 ifma->ifma_addr = dupsa; 3255 3256 ifma->ifma_ifp = ifp; 3257 ifma->ifma_refcount = 1; 3258 ifma->ifma_protospec = NULL; 3259 3260 if (llsa == NULL) { 3261 ifma->ifma_lladdr = NULL; 3262 return (ifma); 3263 } 3264 3265 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3266 if (dupsa == NULL) { 3267 free(ifma->ifma_addr, M_IFMADDR); 3268 free(ifma, M_IFMADDR); 3269 return (NULL); 3270 } 3271 bcopy(llsa, dupsa, llsa->sa_len); 3272 ifma->ifma_lladdr = dupsa; 3273 3274 return (ifma); 3275} 3276 3277/* 3278 * if_freemulti: free ifmultiaddr structure and possibly attached related 3279 * addresses. The caller is responsible for implementing reference 3280 * counting, notifying the driver, handling routing messages, and releasing 3281 * any dependent link layer state. 3282 */ 3283static void 3284if_freemulti(struct ifmultiaddr *ifma) 3285{ 3286 3287 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3288 ifma->ifma_refcount)); 3289 3290 if (ifma->ifma_lladdr != NULL) 3291 free(ifma->ifma_lladdr, M_IFMADDR); 3292 free(ifma->ifma_addr, M_IFMADDR); 3293 free(ifma, M_IFMADDR); 3294} 3295 3296/* 3297 * Register an additional multicast address with a network interface. 3298 * 3299 * - If the address is already present, bump the reference count on the 3300 * address and return. 3301 * - If the address is not link-layer, look up a link layer address. 3302 * - Allocate address structures for one or both addresses, and attach to the 3303 * multicast address list on the interface. If automatically adding a link 3304 * layer address, the protocol address will own a reference to the link 3305 * layer address, to be freed when it is freed. 3306 * - Notify the network device driver of an addition to the multicast address 3307 * list. 3308 * 3309 * 'sa' points to caller-owned memory with the desired multicast address. 3310 * 3311 * 'retifma' will be used to return a pointer to the resulting multicast 3312 * address reference, if desired. 3313 */ 3314int 3315if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3316 struct ifmultiaddr **retifma) 3317{ 3318 struct ifmultiaddr *ifma, *ll_ifma; 3319 struct sockaddr *llsa; 3320 struct sockaddr_dl sdl; 3321 int error; 3322 3323 /* 3324 * If the address is already present, return a new reference to it; 3325 * otherwise, allocate storage and set up a new address. 3326 */ 3327 IF_ADDR_WLOCK(ifp); 3328 ifma = if_findmulti(ifp, sa); 3329 if (ifma != NULL) { 3330 ifma->ifma_refcount++; 3331 if (retifma != NULL) 3332 *retifma = ifma; 3333 IF_ADDR_WUNLOCK(ifp); 3334 return (0); 3335 } 3336 3337 /* 3338 * The address isn't already present; resolve the protocol address 3339 * into a link layer address, and then look that up, bump its 3340 * refcount or allocate an ifma for that also. 3341 * Most link layer resolving functions returns address data which 3342 * fits inside default sockaddr_dl structure. However callback 3343 * can allocate another sockaddr structure, in that case we need to 3344 * free it later. 3345 */ 3346 llsa = NULL; 3347 ll_ifma = NULL; 3348 if (ifp->if_resolvemulti != NULL) { 3349 /* Provide called function with buffer size information */ 3350 sdl.sdl_len = sizeof(sdl); 3351 llsa = (struct sockaddr *)&sdl; 3352 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3353 if (error) 3354 goto unlock_out; 3355 } 3356 3357 /* 3358 * Allocate the new address. Don't hook it up yet, as we may also 3359 * need to allocate a link layer multicast address. 3360 */ 3361 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3362 if (ifma == NULL) { 3363 error = ENOMEM; 3364 goto free_llsa_out; 3365 } 3366 3367 /* 3368 * If a link layer address is found, we'll need to see if it's 3369 * already present in the address list, or allocate is as well. 3370 * When this block finishes, the link layer address will be on the 3371 * list. 3372 */ 3373 if (llsa != NULL) { 3374 ll_ifma = if_findmulti(ifp, llsa); 3375 if (ll_ifma == NULL) { 3376 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3377 if (ll_ifma == NULL) { 3378 --ifma->ifma_refcount; 3379 if_freemulti(ifma); 3380 error = ENOMEM; 3381 goto free_llsa_out; 3382 } 3383 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3384 ifma_link); 3385 } else 3386 ll_ifma->ifma_refcount++; 3387 ifma->ifma_llifma = ll_ifma; 3388 } 3389 3390 /* 3391 * We now have a new multicast address, ifma, and possibly a new or 3392 * referenced link layer address. Add the primary address to the 3393 * ifnet address list. 3394 */ 3395 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3396 3397 if (retifma != NULL) 3398 *retifma = ifma; 3399 3400 /* 3401 * Must generate the message while holding the lock so that 'ifma' 3402 * pointer is still valid. 3403 */ 3404 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3405 IF_ADDR_WUNLOCK(ifp); 3406 3407 /* 3408 * We are certain we have added something, so call down to the 3409 * interface to let them know about it. 3410 */ 3411 if (ifp->if_ioctl != NULL) { 3412 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3413 } 3414 3415 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3416 link_free_sdl(llsa); 3417 3418 return (0); 3419 3420free_llsa_out: 3421 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3422 link_free_sdl(llsa); 3423 3424unlock_out: 3425 IF_ADDR_WUNLOCK(ifp); 3426 return (error); 3427} 3428 3429/* 3430 * Delete a multicast group membership by network-layer group address. 3431 * 3432 * Returns ENOENT if the entry could not be found. If ifp no longer 3433 * exists, results are undefined. This entry point should only be used 3434 * from subsystems which do appropriate locking to hold ifp for the 3435 * duration of the call. 3436 * Network-layer protocol domains must use if_delmulti_ifma(). 3437 */ 3438int 3439if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3440{ 3441 struct ifmultiaddr *ifma; 3442 int lastref; 3443#ifdef INVARIANTS 3444 struct ifnet *oifp; 3445 3446 IFNET_RLOCK_NOSLEEP(); 3447 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3448 if (ifp == oifp) 3449 break; 3450 if (ifp != oifp) 3451 ifp = NULL; 3452 IFNET_RUNLOCK_NOSLEEP(); 3453 3454 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3455#endif 3456 if (ifp == NULL) 3457 return (ENOENT); 3458 3459 IF_ADDR_WLOCK(ifp); 3460 lastref = 0; 3461 ifma = if_findmulti(ifp, sa); 3462 if (ifma != NULL) 3463 lastref = if_delmulti_locked(ifp, ifma, 0); 3464 IF_ADDR_WUNLOCK(ifp); 3465 3466 if (ifma == NULL) 3467 return (ENOENT); 3468 3469 if (lastref && ifp->if_ioctl != NULL) { 3470 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3471 } 3472 3473 return (0); 3474} 3475 3476/* 3477 * Delete all multicast group membership for an interface. 3478 * Should be used to quickly flush all multicast filters. 3479 */ 3480void 3481if_delallmulti(struct ifnet *ifp) 3482{ 3483 struct ifmultiaddr *ifma; 3484 struct ifmultiaddr *next; 3485 3486 IF_ADDR_WLOCK(ifp); 3487 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3488 if_delmulti_locked(ifp, ifma, 0); 3489 IF_ADDR_WUNLOCK(ifp); 3490} 3491 3492/* 3493 * Delete a multicast group membership by group membership pointer. 3494 * Network-layer protocol domains must use this routine. 3495 * 3496 * It is safe to call this routine if the ifp disappeared. 3497 */ 3498void 3499if_delmulti_ifma(struct ifmultiaddr *ifma) 3500{ 3501 struct ifnet *ifp; 3502 int lastref; 3503 3504 ifp = ifma->ifma_ifp; 3505#ifdef DIAGNOSTIC 3506 if (ifp == NULL) { 3507 printf("%s: ifma_ifp seems to be detached\n", __func__); 3508 } else { 3509 struct ifnet *oifp; 3510 3511 IFNET_RLOCK_NOSLEEP(); 3512 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3513 if (ifp == oifp) 3514 break; 3515 if (ifp != oifp) { 3516 printf("%s: ifnet %p disappeared\n", __func__, ifp); 3517 ifp = NULL; 3518 } 3519 IFNET_RUNLOCK_NOSLEEP(); 3520 } 3521#endif 3522 /* 3523 * If and only if the ifnet instance exists: Acquire the address lock. 3524 */ 3525 if (ifp != NULL) 3526 IF_ADDR_WLOCK(ifp); 3527 3528 lastref = if_delmulti_locked(ifp, ifma, 0); 3529 3530 if (ifp != NULL) { 3531 /* 3532 * If and only if the ifnet instance exists: 3533 * Release the address lock. 3534 * If the group was left: update the hardware hash filter. 3535 */ 3536 IF_ADDR_WUNLOCK(ifp); 3537 if (lastref && ifp->if_ioctl != NULL) { 3538 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3539 } 3540 } 3541} 3542 3543/* 3544 * Perform deletion of network-layer and/or link-layer multicast address. 3545 * 3546 * Return 0 if the reference count was decremented. 3547 * Return 1 if the final reference was released, indicating that the 3548 * hardware hash filter should be reprogrammed. 3549 */ 3550static int 3551if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3552{ 3553 struct ifmultiaddr *ll_ifma; 3554 3555 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3556 KASSERT(ifma->ifma_ifp == ifp, 3557 ("%s: inconsistent ifp %p", __func__, ifp)); 3558 IF_ADDR_WLOCK_ASSERT(ifp); 3559 } 3560 3561 ifp = ifma->ifma_ifp; 3562 3563 /* 3564 * If the ifnet is detaching, null out references to ifnet, 3565 * so that upper protocol layers will notice, and not attempt 3566 * to obtain locks for an ifnet which no longer exists. The 3567 * routing socket announcement must happen before the ifnet 3568 * instance is detached from the system. 3569 */ 3570 if (detaching) { 3571#ifdef DIAGNOSTIC 3572 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3573#endif 3574 /* 3575 * ifp may already be nulled out if we are being reentered 3576 * to delete the ll_ifma. 3577 */ 3578 if (ifp != NULL) { 3579 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3580 ifma->ifma_ifp = NULL; 3581 } 3582 } 3583 3584 if (--ifma->ifma_refcount > 0) 3585 return 0; 3586 3587 /* 3588 * If this ifma is a network-layer ifma, a link-layer ifma may 3589 * have been associated with it. Release it first if so. 3590 */ 3591 ll_ifma = ifma->ifma_llifma; 3592 if (ll_ifma != NULL) { 3593 KASSERT(ifma->ifma_lladdr != NULL, 3594 ("%s: llifma w/o lladdr", __func__)); 3595 if (detaching) 3596 ll_ifma->ifma_ifp = NULL; /* XXX */ 3597 if (--ll_ifma->ifma_refcount == 0) { 3598 if (ifp != NULL) { 3599 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3600 ifma_link); 3601 } 3602 if_freemulti(ll_ifma); 3603 } 3604 } 3605 3606 if (ifp != NULL) 3607 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3608 3609 if_freemulti(ifma); 3610 3611 /* 3612 * The last reference to this instance of struct ifmultiaddr 3613 * was released; the hardware should be notified of this change. 3614 */ 3615 return 1; 3616} 3617 3618/* 3619 * Set the link layer address on an interface. 3620 * 3621 * At this time we only support certain types of interfaces, 3622 * and we don't allow the length of the address to change. 3623 * 3624 * Set noinline to be dtrace-friendly 3625 */ 3626__noinline int 3627if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3628{ 3629 struct sockaddr_dl *sdl; 3630 struct ifaddr *ifa; 3631 struct ifreq ifr; 3632 3633 IF_ADDR_RLOCK(ifp); 3634 ifa = ifp->if_addr; 3635 if (ifa == NULL) { 3636 IF_ADDR_RUNLOCK(ifp); 3637 return (EINVAL); 3638 } 3639 ifa_ref(ifa); 3640 IF_ADDR_RUNLOCK(ifp); 3641 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3642 if (sdl == NULL) { 3643 ifa_free(ifa); 3644 return (EINVAL); 3645 } 3646 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3647 ifa_free(ifa); 3648 return (EINVAL); 3649 } 3650 switch (ifp->if_type) { 3651 case IFT_ETHER: 3652 case IFT_FDDI: 3653 case IFT_XETHER: 3654 case IFT_ISO88025: 3655 case IFT_L2VLAN: 3656 case IFT_BRIDGE: 3657 case IFT_ARCNET: 3658 case IFT_IEEE8023ADLAG: 3659 case IFT_IEEE80211: 3660 bcopy(lladdr, LLADDR(sdl), len); 3661 ifa_free(ifa); 3662 break; 3663 default: 3664 ifa_free(ifa); 3665 return (ENODEV); 3666 } 3667 3668 /* 3669 * If the interface is already up, we need 3670 * to re-init it in order to reprogram its 3671 * address filter. 3672 */ 3673 if ((ifp->if_flags & IFF_UP) != 0) { 3674 if (ifp->if_ioctl) { 3675 ifp->if_flags &= ~IFF_UP; 3676 ifr.ifr_flags = ifp->if_flags & 0xffff; 3677 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3678 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3679 ifp->if_flags |= IFF_UP; 3680 ifr.ifr_flags = ifp->if_flags & 0xffff; 3681 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3682 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3683 } 3684 } 3685 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3686 return (0); 3687} 3688 3689/* 3690 * Compat function for handling basic encapsulation requests. 3691 * Not converted stacks (FDDI, IB, ..) supports traditional 3692 * output model: ARP (and other similar L2 protocols) are handled 3693 * inside output routine, arpresolve/nd6_resolve() returns MAC 3694 * address instead of full prepend. 3695 * 3696 * This function creates calculated header==MAC for IPv4/IPv6 and 3697 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3698 * address families. 3699 */ 3700static int 3701if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3702{ 3703 3704 if (req->rtype != IFENCAP_LL) 3705 return (EOPNOTSUPP); 3706 3707 if (req->bufsize < req->lladdr_len) 3708 return (ENOMEM); 3709 3710 switch (req->family) { 3711 case AF_INET: 3712 case AF_INET6: 3713 break; 3714 default: 3715 return (EAFNOSUPPORT); 3716 } 3717 3718 /* Copy lladdr to storage as is */ 3719 memmove(req->buf, req->lladdr, req->lladdr_len); 3720 req->bufsize = req->lladdr_len; 3721 req->lladdr_off = 0; 3722 3723 return (0); 3724} 3725 3726/* 3727 * Get the link layer address that was read from the hardware at attach. 3728 * 3729 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3730 * their component interfaces as IFT_IEEE8023ADLAG. 3731 */ 3732int 3733if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3734{ 3735 3736 if (ifp->if_hw_addr == NULL) 3737 return (ENODEV); 3738 3739 switch (ifp->if_type) { 3740 case IFT_ETHER: 3741 case IFT_IEEE8023ADLAG: 3742 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3743 return (0); 3744 default: 3745 return (ENODEV); 3746 } 3747} 3748 3749/* 3750 * The name argument must be a pointer to storage which will last as 3751 * long as the interface does. For physical devices, the result of 3752 * device_get_name(dev) is a good choice and for pseudo-devices a 3753 * static string works well. 3754 */ 3755void 3756if_initname(struct ifnet *ifp, const char *name, int unit) 3757{ 3758 ifp->if_dname = name; 3759 ifp->if_dunit = unit; 3760 if (unit != IF_DUNIT_NONE) 3761 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3762 else 3763 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3764} 3765 3766int 3767if_printf(struct ifnet *ifp, const char * fmt, ...) 3768{ 3769 va_list ap; 3770 int retval; 3771 3772 retval = printf("%s: ", ifp->if_xname); 3773 va_start(ap, fmt); 3774 retval += vprintf(fmt, ap); 3775 va_end(ap); 3776 return (retval); 3777} 3778 3779void 3780if_start(struct ifnet *ifp) 3781{ 3782 3783 (*(ifp)->if_start)(ifp); 3784} 3785 3786/* 3787 * Backwards compatibility interface for drivers 3788 * that have not implemented it 3789 */ 3790static int 3791if_transmit(struct ifnet *ifp, struct mbuf *m) 3792{ 3793 int error; 3794 3795 IFQ_HANDOFF(ifp, m, error); 3796 return (error); 3797} 3798 3799static void 3800if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3801{ 3802 3803 m_freem(m); 3804} 3805 3806int 3807if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3808{ 3809 int active = 0; 3810 3811 IF_LOCK(ifq); 3812 if (_IF_QFULL(ifq)) { 3813 IF_UNLOCK(ifq); 3814 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 3815 m_freem(m); 3816 return (0); 3817 } 3818 if (ifp != NULL) { 3819 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 3820 if (m->m_flags & (M_BCAST|M_MCAST)) 3821 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 3822 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3823 } 3824 _IF_ENQUEUE(ifq, m); 3825 IF_UNLOCK(ifq); 3826 if (ifp != NULL && !active) 3827 (*(ifp)->if_start)(ifp); 3828 return (1); 3829} 3830 3831void 3832if_register_com_alloc(u_char type, 3833 if_com_alloc_t *a, if_com_free_t *f) 3834{ 3835 3836 KASSERT(if_com_alloc[type] == NULL, 3837 ("if_register_com_alloc: %d already registered", type)); 3838 KASSERT(if_com_free[type] == NULL, 3839 ("if_register_com_alloc: %d free already registered", type)); 3840 3841 if_com_alloc[type] = a; 3842 if_com_free[type] = f; 3843} 3844 3845void 3846if_deregister_com_alloc(u_char type) 3847{ 3848 3849 KASSERT(if_com_alloc[type] != NULL, 3850 ("if_deregister_com_alloc: %d not registered", type)); 3851 KASSERT(if_com_free[type] != NULL, 3852 ("if_deregister_com_alloc: %d free not registered", type)); 3853 if_com_alloc[type] = NULL; 3854 if_com_free[type] = NULL; 3855} 3856 3857/* API for driver access to network stack owned ifnet.*/ 3858uint64_t 3859if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 3860{ 3861 uint64_t oldbrate; 3862 3863 oldbrate = ifp->if_baudrate; 3864 ifp->if_baudrate = baudrate; 3865 return (oldbrate); 3866} 3867 3868uint64_t 3869if_getbaudrate(if_t ifp) 3870{ 3871 3872 return (((struct ifnet *)ifp)->if_baudrate); 3873} 3874 3875int 3876if_setcapabilities(if_t ifp, int capabilities) 3877{ 3878 ((struct ifnet *)ifp)->if_capabilities = capabilities; 3879 return (0); 3880} 3881 3882int 3883if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 3884{ 3885 ((struct ifnet *)ifp)->if_capabilities |= setbit; 3886 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 3887 3888 return (0); 3889} 3890 3891int 3892if_getcapabilities(if_t ifp) 3893{ 3894 return ((struct ifnet *)ifp)->if_capabilities; 3895} 3896 3897int 3898if_setcapenable(if_t ifp, int capabilities) 3899{ 3900 ((struct ifnet *)ifp)->if_capenable = capabilities; 3901 return (0); 3902} 3903 3904int 3905if_setcapenablebit(if_t ifp, int setcap, int clearcap) 3906{ 3907 if(setcap) 3908 ((struct ifnet *)ifp)->if_capenable |= setcap; 3909 if(clearcap) 3910 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 3911 3912 return (0); 3913} 3914 3915const char * 3916if_getdname(if_t ifp) 3917{ 3918 return ((struct ifnet *)ifp)->if_dname; 3919} 3920 3921int 3922if_togglecapenable(if_t ifp, int togglecap) 3923{ 3924 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 3925 return (0); 3926} 3927 3928int 3929if_getcapenable(if_t ifp) 3930{ 3931 return ((struct ifnet *)ifp)->if_capenable; 3932} 3933 3934/* 3935 * This is largely undesirable because it ties ifnet to a device, but does 3936 * provide flexiblity for an embedded product vendor. Should be used with 3937 * the understanding that it violates the interface boundaries, and should be 3938 * a last resort only. 3939 */ 3940int 3941if_setdev(if_t ifp, void *dev) 3942{ 3943 return (0); 3944} 3945 3946int 3947if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 3948{ 3949 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 3950 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 3951 3952 return (0); 3953} 3954 3955int 3956if_getdrvflags(if_t ifp) 3957{ 3958 return ((struct ifnet *)ifp)->if_drv_flags; 3959} 3960 3961int 3962if_setdrvflags(if_t ifp, int flags) 3963{ 3964 ((struct ifnet *)ifp)->if_drv_flags = flags; 3965 return (0); 3966} 3967 3968 3969int 3970if_setflags(if_t ifp, int flags) 3971{ 3972 ((struct ifnet *)ifp)->if_flags = flags; 3973 return (0); 3974} 3975 3976int 3977if_setflagbits(if_t ifp, int set, int clear) 3978{ 3979 ((struct ifnet *)ifp)->if_flags |= set; 3980 ((struct ifnet *)ifp)->if_flags &= ~clear; 3981 3982 return (0); 3983} 3984 3985int 3986if_getflags(if_t ifp) 3987{ 3988 return ((struct ifnet *)ifp)->if_flags; 3989} 3990 3991int 3992if_clearhwassist(if_t ifp) 3993{ 3994 ((struct ifnet *)ifp)->if_hwassist = 0; 3995 return (0); 3996} 3997 3998int 3999if_sethwassistbits(if_t ifp, int toset, int toclear) 4000{ 4001 ((struct ifnet *)ifp)->if_hwassist |= toset; 4002 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4003 4004 return (0); 4005} 4006 4007int 4008if_sethwassist(if_t ifp, int hwassist_bit) 4009{ 4010 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4011 return (0); 4012} 4013 4014int 4015if_gethwassist(if_t ifp) 4016{ 4017 return ((struct ifnet *)ifp)->if_hwassist; 4018} 4019 4020int 4021if_setmtu(if_t ifp, int mtu) 4022{ 4023 ((struct ifnet *)ifp)->if_mtu = mtu; 4024 return (0); 4025} 4026 4027int 4028if_getmtu(if_t ifp) 4029{ 4030 return ((struct ifnet *)ifp)->if_mtu; 4031} 4032 4033int 4034if_getmtu_family(if_t ifp, int family) 4035{ 4036 struct domain *dp; 4037 4038 for (dp = domains; dp; dp = dp->dom_next) { 4039 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4040 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4041 } 4042 4043 return (((struct ifnet *)ifp)->if_mtu); 4044} 4045 4046int 4047if_setsoftc(if_t ifp, void *softc) 4048{ 4049 ((struct ifnet *)ifp)->if_softc = softc; 4050 return (0); 4051} 4052 4053void * 4054if_getsoftc(if_t ifp) 4055{ 4056 return ((struct ifnet *)ifp)->if_softc; 4057} 4058 4059void 4060if_setrcvif(struct mbuf *m, if_t ifp) 4061{ 4062 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4063} 4064 4065void 4066if_setvtag(struct mbuf *m, uint16_t tag) 4067{ 4068 m->m_pkthdr.ether_vtag = tag; 4069} 4070 4071uint16_t 4072if_getvtag(struct mbuf *m) 4073{ 4074 4075 return (m->m_pkthdr.ether_vtag); 4076} 4077 4078int 4079if_sendq_empty(if_t ifp) 4080{ 4081 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4082} 4083 4084struct ifaddr * 4085if_getifaddr(if_t ifp) 4086{ 4087 return ((struct ifnet *)ifp)->if_addr; 4088} 4089 4090int 4091if_getamcount(if_t ifp) 4092{ 4093 return ((struct ifnet *)ifp)->if_amcount; 4094} 4095 4096 4097int 4098if_setsendqready(if_t ifp) 4099{ 4100 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4101 return (0); 4102} 4103 4104int 4105if_setsendqlen(if_t ifp, int tx_desc_count) 4106{ 4107 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4108 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4109 4110 return (0); 4111} 4112 4113int 4114if_vlantrunkinuse(if_t ifp) 4115{ 4116 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4117} 4118 4119int 4120if_input(if_t ifp, struct mbuf* sendmp) 4121{ 4122 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4123 return (0); 4124 4125} 4126 4127/* XXX */ 4128#ifndef ETH_ADDR_LEN 4129#define ETH_ADDR_LEN 6 4130#endif 4131 4132int 4133if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max) 4134{ 4135 struct ifmultiaddr *ifma; 4136 uint8_t *lmta = (uint8_t *)mta; 4137 int mcnt = 0; 4138 4139 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4140 if (ifma->ifma_addr->sa_family != AF_LINK) 4141 continue; 4142 4143 if (mcnt == max) 4144 break; 4145 4146 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 4147 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 4148 mcnt++; 4149 } 4150 *cnt = mcnt; 4151 4152 return (0); 4153} 4154 4155int 4156if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max) 4157{ 4158 int error; 4159 4160 if_maddr_rlock(ifp); 4161 error = if_setupmultiaddr(ifp, mta, cnt, max); 4162 if_maddr_runlock(ifp); 4163 return (error); 4164} 4165 4166int 4167if_multiaddr_count(if_t ifp, int max) 4168{ 4169 struct ifmultiaddr *ifma; 4170 int count; 4171 4172 count = 0; 4173 if_maddr_rlock(ifp); 4174 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4175 if (ifma->ifma_addr->sa_family != AF_LINK) 4176 continue; 4177 count++; 4178 if (count == max) 4179 break; 4180 } 4181 if_maddr_runlock(ifp); 4182 return (count); 4183} 4184 4185int 4186if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg) 4187{ 4188 struct ifmultiaddr *ifma; 4189 int cnt = 0; 4190 4191 if_maddr_rlock(ifp); 4192 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4193 cnt += filter(arg, ifma, cnt); 4194 if_maddr_runlock(ifp); 4195 return (cnt); 4196} 4197 4198struct mbuf * 4199if_dequeue(if_t ifp) 4200{ 4201 struct mbuf *m; 4202 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4203 4204 return (m); 4205} 4206 4207int 4208if_sendq_prepend(if_t ifp, struct mbuf *m) 4209{ 4210 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4211 return (0); 4212} 4213 4214int 4215if_setifheaderlen(if_t ifp, int len) 4216{ 4217 ((struct ifnet *)ifp)->if_hdrlen = len; 4218 return (0); 4219} 4220 4221caddr_t 4222if_getlladdr(if_t ifp) 4223{ 4224 return (IF_LLADDR((struct ifnet *)ifp)); 4225} 4226 4227void * 4228if_gethandle(u_char type) 4229{ 4230 return (if_alloc(type)); 4231} 4232 4233void 4234if_bpfmtap(if_t ifh, struct mbuf *m) 4235{ 4236 struct ifnet *ifp = (struct ifnet *)ifh; 4237 4238 BPF_MTAP(ifp, m); 4239} 4240 4241void 4242if_etherbpfmtap(if_t ifh, struct mbuf *m) 4243{ 4244 struct ifnet *ifp = (struct ifnet *)ifh; 4245 4246 ETHER_BPF_MTAP(ifp, m); 4247} 4248 4249void 4250if_vlancap(if_t ifh) 4251{ 4252 struct ifnet *ifp = (struct ifnet *)ifh; 4253 VLAN_CAPABILITIES(ifp); 4254} 4255 4256void 4257if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4258{ 4259 ((struct ifnet *)ifp)->if_init = init_fn; 4260} 4261 4262void 4263if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4264{ 4265 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4266} 4267 4268void 4269if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4270{ 4271 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4272} 4273 4274void 4275if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4276{ 4277 ((struct ifnet *)ifp)->if_transmit = start_fn; 4278} 4279 4280void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4281{ 4282 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4283 4284} 4285 4286void 4287if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4288{ 4289 4290 ifp->if_get_counter = fn; 4291} 4292 4293/* Revisit these - These are inline functions originally. */ 4294int 4295drbr_inuse_drv(if_t ifh, struct buf_ring *br) 4296{ 4297 return drbr_inuse(ifh, br); 4298} 4299 4300struct mbuf* 4301drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 4302{ 4303 return drbr_dequeue(ifh, br); 4304} 4305 4306int 4307drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 4308{ 4309 return drbr_needs_enqueue(ifh, br); 4310} 4311 4312int 4313drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 4314{ 4315 return drbr_enqueue(ifh, br, m); 4316 4317} 4318