1 /*- 2 * Copyright (c) 1980, 1986, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)if.c 8.5 (Berkeley) 1/9/95 30 * $FreeBSD$ 31 */ 32 33 #include "opt_compat.h" 34 #include "opt_inet6.h" 35 #include "opt_inet.h" 36 37 #include <sys/param.h> 38 #include <sys/types.h> 39 #include <sys/conf.h> 40 #include <sys/malloc.h> 41 #include <sys/sbuf.h> 42 #include <sys/bus.h> 43 #include <sys/mbuf.h> 44 #include <sys/systm.h> 45 #include <sys/priv.h> 46 #include <sys/proc.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/protosw.h> 50 #include <sys/kernel.h> 51 #include <sys/lock.h> 52 #include <sys/refcount.h> 53 #include <sys/module.h> 54 #include <sys/rwlock.h> 55 #include <sys/sockio.h> 56 #include <sys/syslog.h> 57 #include <sys/sysctl.h> 58 #include <sys/taskqueue.h> 59 #include <sys/domain.h> 60 #include <sys/jail.h> 61 #include <sys/priv.h> 62 63 #include <machine/stdarg.h> 64 #include <vm/uma.h> 65 66 #include <net/bpf.h> 67 #include <net/ethernet.h> 68 #include <net/if.h> 69 #include <net/if_arp.h> 70 #include <net/if_clone.h> 71 #include <net/if_dl.h> 72 #include <net/if_types.h> 73 #include <net/if_var.h> 74 #include <net/if_media.h> 75 #include <net/if_vlan_var.h> 76 #include <net/radix.h> 77 #include <net/route.h> 78 #include <net/vnet.h> 79 80 #if defined(INET) || defined(INET6) 81 #include <net/ethernet.h> 82 #include <netinet/in.h> 83 #include <netinet/in_var.h> 84 #include <netinet/ip.h> 85 #include <netinet/ip_carp.h> 86 #ifdef INET 87 #include <netinet/if_ether.h> 88 #endif /* INET */ 89 #ifdef INET6 90 #include <netinet6/in6_var.h> 91 #include <netinet6/in6_ifattach.h> 92 #endif /* INET6 */ 93 #endif /* INET || INET6 */ 94 95 #include <security/mac/mac_framework.h> 96 97 #ifdef COMPAT_FREEBSD32 98 #include <sys/mount.h> 99 #include <compat/freebsd32/freebsd32.h> 100 #endif 101 102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 104 105 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 106 &ifqmaxlen, 0, "max send queue size"); 107 108 /* Log link state change events */ 109 static int log_link_state_change = 1; 110 111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 112 &log_link_state_change, 0, 113 "log interface link state change events"); 114 115 /* Log promiscuous mode change events */ 116 static int log_promisc_mode_change = 1; 117 118 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 119 &log_promisc_mode_change, 1, 120 "log promiscuous mode change events"); 121 122 /* Interface description */ 123 static unsigned int ifdescr_maxlen = 1024; 124 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 125 &ifdescr_maxlen, 0, 126 "administrative maximum length for interface description"); 127 128 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 129 130 /* global sx for non-critical path ifdescr */ 131 static struct sx ifdescr_sx; 132 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 133 134 void (*bridge_linkstate_p)(struct ifnet *ifp); 135 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 136 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 137 /* These are external hooks for CARP. */ 138 void (*carp_linkstate_p)(struct ifnet *ifp); 139 void (*carp_demote_adj_p)(int, char *); 140 int (*carp_master_p)(struct ifaddr *); 141 #if defined(INET) || defined(INET6) 142 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 143 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 144 const struct sockaddr *sa); 145 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 146 int (*carp_attach_p)(struct ifaddr *, int); 147 void (*carp_detach_p)(struct ifaddr *); 148 #endif 149 #ifdef INET 150 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 151 #endif 152 #ifdef INET6 153 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 154 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 155 const struct in6_addr *taddr); 156 #endif 157 158 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 159 160 /* 161 * XXX: Style; these should be sorted alphabetically, and unprototyped 162 * static functions should be prototyped. Currently they are sorted by 163 * declaration order. 164 */ 165 static void if_attachdomain(void *); 166 static void if_attachdomain1(struct ifnet *); 167 static int ifconf(u_long, caddr_t); 168 static void if_freemulti(struct ifmultiaddr *); 169 static void if_grow(void); 170 static void if_input_default(struct ifnet *, struct mbuf *); 171 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 172 static void if_route(struct ifnet *, int flag, int fam); 173 static int if_setflag(struct ifnet *, int, int, int *, int); 174 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 175 static void if_unroute(struct ifnet *, int flag, int fam); 176 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 177 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 178 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 179 static void do_link_state_change(void *, int); 180 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 181 static int if_getgroupmembers(struct ifgroupreq *); 182 static void if_delgroups(struct ifnet *); 183 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 184 static int if_detach_internal(struct ifnet *, int, struct if_clone **); 185 #ifdef VIMAGE 186 static void if_vmove(struct ifnet *, struct vnet *); 187 #endif 188 189 #ifdef INET6 190 /* 191 * XXX: declare here to avoid to include many inet6 related files.. 192 * should be more generalized? 193 */ 194 extern void nd6_setmtu(struct ifnet *); 195 #endif 196 197 /* ipsec helper hooks */ 198 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 199 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 200 201 VNET_DEFINE(int, if_index); 202 int ifqmaxlen = IFQ_MAXLEN; 203 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 204 VNET_DEFINE(struct ifgrouphead, ifg_head); 205 206 static VNET_DEFINE(int, if_indexlim) = 8; 207 208 /* Table of ifnet by index. */ 209 VNET_DEFINE(struct ifnet **, ifindex_table); 210 211 #define V_if_indexlim VNET(if_indexlim) 212 #define V_ifindex_table VNET(ifindex_table) 213 214 /* 215 * The global network interface list (V_ifnet) and related state (such as 216 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 217 * an rwlock. Either may be acquired shared to stablize the list, but both 218 * must be acquired writable to modify the list. This model allows us to 219 * both stablize the interface list during interrupt thread processing, but 220 * also to stablize it over long-running ioctls, without introducing priority 221 * inversions and deadlocks. 222 */ 223 struct rwlock ifnet_rwlock; 224 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE); 225 struct sx ifnet_sxlock; 226 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 227 228 /* 229 * The allocation of network interfaces is a rather non-atomic affair; we 230 * need to select an index before we are ready to expose the interface for 231 * use, so will use this pointer value to indicate reservation. 232 */ 233 #define IFNET_HOLD (void *)(uintptr_t)(-1) 234 235 static if_com_alloc_t *if_com_alloc[256]; 236 static if_com_free_t *if_com_free[256]; 237 238 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 239 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 240 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 241 242 struct ifnet * 243 ifnet_byindex_locked(u_short idx) 244 { 245 246 if (idx > V_if_index) 247 return (NULL); 248 if (V_ifindex_table[idx] == IFNET_HOLD) 249 return (NULL); 250 return (V_ifindex_table[idx]); 251 } 252 253 struct ifnet * 254 ifnet_byindex(u_short idx) 255 { 256 struct ifnet *ifp; 257 258 IFNET_RLOCK_NOSLEEP(); 259 ifp = ifnet_byindex_locked(idx); 260 IFNET_RUNLOCK_NOSLEEP(); 261 return (ifp); 262 } 263 264 struct ifnet * 265 ifnet_byindex_ref(u_short idx) 266 { 267 struct ifnet *ifp; 268 269 IFNET_RLOCK_NOSLEEP(); 270 ifp = ifnet_byindex_locked(idx); 271 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 272 IFNET_RUNLOCK_NOSLEEP(); 273 return (NULL); 274 } 275 if_ref(ifp); 276 IFNET_RUNLOCK_NOSLEEP(); 277 return (ifp); 278 } 279 280 /* 281 * Allocate an ifindex array entry; return 0 on success or an error on 282 * failure. 283 */ 284 static u_short 285 ifindex_alloc(void) 286 { 287 u_short idx; 288 289 IFNET_WLOCK_ASSERT(); 290 retry: 291 /* 292 * Try to find an empty slot below V_if_index. If we fail, take the 293 * next slot. 294 */ 295 for (idx = 1; idx <= V_if_index; idx++) { 296 if (V_ifindex_table[idx] == NULL) 297 break; 298 } 299 300 /* Catch if_index overflow. */ 301 if (idx >= V_if_indexlim) { 302 if_grow(); 303 goto retry; 304 } 305 if (idx > V_if_index) 306 V_if_index = idx; 307 return (idx); 308 } 309 310 static void 311 ifindex_free_locked(u_short idx) 312 { 313 314 IFNET_WLOCK_ASSERT(); 315 316 V_ifindex_table[idx] = NULL; 317 while (V_if_index > 0 && 318 V_ifindex_table[V_if_index] == NULL) 319 V_if_index--; 320 } 321 322 static void 323 ifindex_free(u_short idx) 324 { 325 326 IFNET_WLOCK(); 327 ifindex_free_locked(idx); 328 IFNET_WUNLOCK(); 329 } 330 331 static void 332 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) 333 { 334 335 IFNET_WLOCK_ASSERT(); 336 337 V_ifindex_table[idx] = ifp; 338 } 339 340 static void 341 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 342 { 343 344 IFNET_WLOCK(); 345 ifnet_setbyindex_locked(idx, ifp); 346 IFNET_WUNLOCK(); 347 } 348 349 struct ifaddr * 350 ifaddr_byindex(u_short idx) 351 { 352 struct ifnet *ifp; 353 struct ifaddr *ifa = NULL; 354 355 IFNET_RLOCK_NOSLEEP(); 356 ifp = ifnet_byindex_locked(idx); 357 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 358 ifa_ref(ifa); 359 IFNET_RUNLOCK_NOSLEEP(); 360 return (ifa); 361 } 362 363 /* 364 * Network interface utility routines. 365 * 366 * Routines with ifa_ifwith* names take sockaddr *'s as 367 * parameters. 368 */ 369 370 static void 371 vnet_if_init(const void *unused __unused) 372 { 373 374 TAILQ_INIT(&V_ifnet); 375 TAILQ_INIT(&V_ifg_head); 376 IFNET_WLOCK(); 377 if_grow(); /* create initial table */ 378 IFNET_WUNLOCK(); 379 vnet_if_clone_init(); 380 } 381 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 382 NULL); 383 384 #ifdef VIMAGE 385 static void 386 vnet_if_uninit(const void *unused __unused) 387 { 388 389 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 390 "not empty", __func__, __LINE__, &V_ifnet)); 391 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 392 "not empty", __func__, __LINE__, &V_ifg_head)); 393 394 free((caddr_t)V_ifindex_table, M_IFNET); 395 } 396 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 397 vnet_if_uninit, NULL); 398 399 static void 400 vnet_if_return(const void *unused __unused) 401 { 402 struct ifnet *ifp, *nifp; 403 404 /* Return all inherited interfaces to their parent vnets. */ 405 TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 406 if (ifp->if_home_vnet != ifp->if_vnet) 407 if_vmove(ifp, ifp->if_home_vnet); 408 } 409 } 410 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 411 vnet_if_return, NULL); 412 #endif 413 414 static void 415 if_grow(void) 416 { 417 int oldlim; 418 u_int n; 419 struct ifnet **e; 420 421 IFNET_WLOCK_ASSERT(); 422 oldlim = V_if_indexlim; 423 IFNET_WUNLOCK(); 424 n = (oldlim << 1) * sizeof(*e); 425 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 426 IFNET_WLOCK(); 427 if (V_if_indexlim != oldlim) { 428 free(e, M_IFNET); 429 return; 430 } 431 if (V_ifindex_table != NULL) { 432 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 433 free((caddr_t)V_ifindex_table, M_IFNET); 434 } 435 V_if_indexlim <<= 1; 436 V_ifindex_table = e; 437 } 438 439 /* 440 * Allocate a struct ifnet and an index for an interface. A layer 2 441 * common structure will also be allocated if an allocation routine is 442 * registered for the passed type. 443 */ 444 struct ifnet * 445 if_alloc(u_char type) 446 { 447 struct ifnet *ifp; 448 u_short idx; 449 450 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 451 IFNET_WLOCK(); 452 idx = ifindex_alloc(); 453 ifnet_setbyindex_locked(idx, IFNET_HOLD); 454 IFNET_WUNLOCK(); 455 ifp->if_index = idx; 456 ifp->if_type = type; 457 ifp->if_alloctype = type; 458 #ifdef VIMAGE 459 ifp->if_vnet = curvnet; 460 #endif 461 if (if_com_alloc[type] != NULL) { 462 ifp->if_l2com = if_com_alloc[type](type, ifp); 463 if (ifp->if_l2com == NULL) { 464 free(ifp, M_IFNET); 465 ifindex_free(idx); 466 return (NULL); 467 } 468 } 469 470 IF_ADDR_LOCK_INIT(ifp); 471 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 472 ifp->if_afdata_initialized = 0; 473 IF_AFDATA_LOCK_INIT(ifp); 474 TAILQ_INIT(&ifp->if_addrhead); 475 TAILQ_INIT(&ifp->if_multiaddrs); 476 TAILQ_INIT(&ifp->if_groups); 477 #ifdef MAC 478 mac_ifnet_init(ifp); 479 #endif 480 ifq_init(&ifp->if_snd, ifp); 481 482 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 483 for (int i = 0; i < IFCOUNTERS; i++) 484 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 485 ifp->if_get_counter = if_get_counter_default; 486 ifnet_setbyindex(ifp->if_index, ifp); 487 return (ifp); 488 } 489 490 /* 491 * Do the actual work of freeing a struct ifnet, and layer 2 common 492 * structure. This call is made when the last reference to an 493 * interface is released. 494 */ 495 static void 496 if_free_internal(struct ifnet *ifp) 497 { 498 499 KASSERT((ifp->if_flags & IFF_DYING), 500 ("if_free_internal: interface not dying")); 501 502 if (if_com_free[ifp->if_alloctype] != NULL) 503 if_com_free[ifp->if_alloctype](ifp->if_l2com, 504 ifp->if_alloctype); 505 506 #ifdef MAC 507 mac_ifnet_destroy(ifp); 508 #endif /* MAC */ 509 if (ifp->if_description != NULL) 510 free(ifp->if_description, M_IFDESCR); 511 IF_AFDATA_DESTROY(ifp); 512 IF_ADDR_LOCK_DESTROY(ifp); 513 ifq_delete(&ifp->if_snd); 514 515 for (int i = 0; i < IFCOUNTERS; i++) 516 counter_u64_free(ifp->if_counters[i]); 517 518 free(ifp, M_IFNET); 519 } 520 521 /* 522 * Deregister an interface and free the associated storage. 523 */ 524 void 525 if_free(struct ifnet *ifp) 526 { 527 528 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 529 530 CURVNET_SET_QUIET(ifp->if_vnet); 531 IFNET_WLOCK(); 532 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 533 ("%s: freeing unallocated ifnet", ifp->if_xname)); 534 535 ifindex_free_locked(ifp->if_index); 536 IFNET_WUNLOCK(); 537 538 if (refcount_release(&ifp->if_refcount)) 539 if_free_internal(ifp); 540 CURVNET_RESTORE(); 541 } 542 543 /* 544 * Interfaces to keep an ifnet type-stable despite the possibility of the 545 * driver calling if_free(). If there are additional references, we defer 546 * freeing the underlying data structure. 547 */ 548 void 549 if_ref(struct ifnet *ifp) 550 { 551 552 /* We don't assert the ifnet list lock here, but arguably should. */ 553 refcount_acquire(&ifp->if_refcount); 554 } 555 556 void 557 if_rele(struct ifnet *ifp) 558 { 559 560 if (!refcount_release(&ifp->if_refcount)) 561 return; 562 if_free_internal(ifp); 563 } 564 565 void 566 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 567 { 568 569 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 570 571 if (ifq->ifq_maxlen == 0) 572 ifq->ifq_maxlen = ifqmaxlen; 573 574 ifq->altq_type = 0; 575 ifq->altq_disc = NULL; 576 ifq->altq_flags &= ALTQF_CANTCHANGE; 577 ifq->altq_tbr = NULL; 578 ifq->altq_ifp = ifp; 579 } 580 581 void 582 ifq_delete(struct ifaltq *ifq) 583 { 584 mtx_destroy(&ifq->ifq_mtx); 585 } 586 587 /* 588 * Perform generic interface initialization tasks and attach the interface 589 * to the list of "active" interfaces. If vmove flag is set on entry 590 * to if_attach_internal(), perform only a limited subset of initialization 591 * tasks, given that we are moving from one vnet to another an ifnet which 592 * has already been fully initialized. 593 * 594 * Note that if_detach_internal() removes group membership unconditionally 595 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 596 * Thus, when if_vmove() is applied to a cloned interface, group membership 597 * is lost while a cloned one always joins a group whose name is 598 * ifc->ifc_name. To recover this after if_detach_internal() and 599 * if_attach_internal(), the cloner should be specified to 600 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 601 * attempts to join a group whose name is ifc->ifc_name. 602 * 603 * XXX: 604 * - The decision to return void and thus require this function to 605 * succeed is questionable. 606 * - We should probably do more sanity checking. For instance we don't 607 * do anything to insure if_xname is unique or non-empty. 608 */ 609 void 610 if_attach(struct ifnet *ifp) 611 { 612 613 if_attach_internal(ifp, 0, NULL); 614 } 615 616 /* 617 * Compute the least common TSO limit. 618 */ 619 void 620 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 621 { 622 /* 623 * 1) If there is no limit currently, take the limit from 624 * the network adapter. 625 * 626 * 2) If the network adapter has a limit below the current 627 * limit, apply it. 628 */ 629 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 630 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 631 pmax->tsomaxbytes = ifp->if_hw_tsomax; 632 } 633 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 634 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 635 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 636 } 637 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 638 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 639 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 640 } 641 } 642 643 /* 644 * Update TSO limit of a network adapter. 645 * 646 * Returns zero if no change. Else non-zero. 647 */ 648 int 649 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 650 { 651 int retval = 0; 652 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 653 ifp->if_hw_tsomax = pmax->tsomaxbytes; 654 retval++; 655 } 656 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 657 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 658 retval++; 659 } 660 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 661 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 662 retval++; 663 } 664 return (retval); 665 } 666 667 static void 668 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 669 { 670 unsigned socksize, ifasize; 671 int namelen, masklen; 672 struct sockaddr_dl *sdl; 673 struct ifaddr *ifa; 674 675 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 676 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 677 ifp->if_xname); 678 679 #ifdef VIMAGE 680 ifp->if_vnet = curvnet; 681 if (ifp->if_home_vnet == NULL) 682 ifp->if_home_vnet = curvnet; 683 #endif 684 685 if_addgroup(ifp, IFG_ALL); 686 687 /* Restore group membership for cloned interfaces. */ 688 if (vmove && ifc != NULL) 689 if_clone_addgroup(ifp, ifc); 690 691 getmicrotime(&ifp->if_lastchange); 692 ifp->if_epoch = time_uptime; 693 694 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 695 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 696 ("transmit and qflush must both either be set or both be NULL")); 697 if (ifp->if_transmit == NULL) { 698 ifp->if_transmit = if_transmit; 699 ifp->if_qflush = if_qflush; 700 } 701 if (ifp->if_input == NULL) 702 ifp->if_input = if_input_default; 703 704 if (ifp->if_requestencap == NULL) 705 ifp->if_requestencap = if_requestencap_default; 706 707 if (!vmove) { 708 #ifdef MAC 709 mac_ifnet_create(ifp); 710 #endif 711 712 /* 713 * Create a Link Level name for this device. 714 */ 715 namelen = strlen(ifp->if_xname); 716 /* 717 * Always save enough space for any possiable name so we 718 * can do a rename in place later. 719 */ 720 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 721 socksize = masklen + ifp->if_addrlen; 722 if (socksize < sizeof(*sdl)) 723 socksize = sizeof(*sdl); 724 socksize = roundup2(socksize, sizeof(long)); 725 ifasize = sizeof(*ifa) + 2 * socksize; 726 ifa = ifa_alloc(ifasize, M_WAITOK); 727 sdl = (struct sockaddr_dl *)(ifa + 1); 728 sdl->sdl_len = socksize; 729 sdl->sdl_family = AF_LINK; 730 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 731 sdl->sdl_nlen = namelen; 732 sdl->sdl_index = ifp->if_index; 733 sdl->sdl_type = ifp->if_type; 734 ifp->if_addr = ifa; 735 ifa->ifa_ifp = ifp; 736 ifa->ifa_rtrequest = link_rtrequest; 737 ifa->ifa_addr = (struct sockaddr *)sdl; 738 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 739 ifa->ifa_netmask = (struct sockaddr *)sdl; 740 sdl->sdl_len = masklen; 741 while (namelen != 0) 742 sdl->sdl_data[--namelen] = 0xff; 743 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 744 /* Reliably crash if used uninitialized. */ 745 ifp->if_broadcastaddr = NULL; 746 747 #if defined(INET) || defined(INET6) 748 /* Use defaults for TSO, if nothing is set */ 749 if (ifp->if_hw_tsomax == 0 && 750 ifp->if_hw_tsomaxsegcount == 0 && 751 ifp->if_hw_tsomaxsegsize == 0) { 752 /* 753 * The TSO defaults needs to be such that an 754 * NFS mbuf list of 35 mbufs totalling just 755 * below 64K works and that a chain of mbufs 756 * can be defragged into at most 32 segments: 757 */ 758 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 759 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 760 ifp->if_hw_tsomaxsegcount = 35; 761 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 762 763 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 764 if (ifp->if_capabilities & IFCAP_TSO) { 765 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 766 ifp->if_hw_tsomax, 767 ifp->if_hw_tsomaxsegcount, 768 ifp->if_hw_tsomaxsegsize); 769 } 770 } 771 #endif 772 } 773 #ifdef VIMAGE 774 else { 775 /* 776 * Update the interface index in the link layer address 777 * of the interface. 778 */ 779 for (ifa = ifp->if_addr; ifa != NULL; 780 ifa = TAILQ_NEXT(ifa, ifa_link)) { 781 if (ifa->ifa_addr->sa_family == AF_LINK) { 782 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 783 sdl->sdl_index = ifp->if_index; 784 } 785 } 786 } 787 #endif 788 789 IFNET_WLOCK(); 790 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 791 #ifdef VIMAGE 792 curvnet->vnet_ifcnt++; 793 #endif 794 IFNET_WUNLOCK(); 795 796 if (domain_init_status >= 2) 797 if_attachdomain1(ifp); 798 799 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 800 if (IS_DEFAULT_VNET(curvnet)) 801 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 802 803 /* Announce the interface. */ 804 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 805 } 806 807 static void 808 if_attachdomain(void *dummy) 809 { 810 struct ifnet *ifp; 811 812 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 813 if_attachdomain1(ifp); 814 } 815 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 816 if_attachdomain, NULL); 817 818 static void 819 if_attachdomain1(struct ifnet *ifp) 820 { 821 struct domain *dp; 822 823 /* 824 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 825 * cannot lock ifp->if_afdata initialization, entirely. 826 */ 827 IF_AFDATA_LOCK(ifp); 828 if (ifp->if_afdata_initialized >= domain_init_status) { 829 IF_AFDATA_UNLOCK(ifp); 830 log(LOG_WARNING, "%s called more than once on %s\n", 831 __func__, ifp->if_xname); 832 return; 833 } 834 ifp->if_afdata_initialized = domain_init_status; 835 IF_AFDATA_UNLOCK(ifp); 836 837 /* address family dependent data region */ 838 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 839 for (dp = domains; dp; dp = dp->dom_next) { 840 if (dp->dom_ifattach) 841 ifp->if_afdata[dp->dom_family] = 842 (*dp->dom_ifattach)(ifp); 843 } 844 } 845 846 /* 847 * Remove any unicast or broadcast network addresses from an interface. 848 */ 849 void 850 if_purgeaddrs(struct ifnet *ifp) 851 { 852 struct ifaddr *ifa, *next; 853 854 /* XXX cannot hold IF_ADDR_WLOCK over called functions. */ 855 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 856 if (ifa->ifa_addr->sa_family == AF_LINK) 857 continue; 858 #ifdef INET 859 /* XXX: Ugly!! ad hoc just for INET */ 860 if (ifa->ifa_addr->sa_family == AF_INET) { 861 struct ifaliasreq ifr; 862 863 bzero(&ifr, sizeof(ifr)); 864 ifr.ifra_addr = *ifa->ifa_addr; 865 if (ifa->ifa_dstaddr) 866 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 867 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 868 NULL) == 0) 869 continue; 870 } 871 #endif /* INET */ 872 #ifdef INET6 873 if (ifa->ifa_addr->sa_family == AF_INET6) { 874 in6_purgeaddr(ifa); 875 /* ifp_addrhead is already updated */ 876 continue; 877 } 878 #endif /* INET6 */ 879 IF_ADDR_WLOCK(ifp); 880 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 881 IF_ADDR_WUNLOCK(ifp); 882 ifa_free(ifa); 883 } 884 } 885 886 /* 887 * Remove any multicast network addresses from an interface when an ifnet 888 * is going away. 889 */ 890 static void 891 if_purgemaddrs(struct ifnet *ifp) 892 { 893 struct ifmultiaddr *ifma; 894 struct ifmultiaddr *next; 895 896 IF_ADDR_WLOCK(ifp); 897 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 898 if_delmulti_locked(ifp, ifma, 1); 899 IF_ADDR_WUNLOCK(ifp); 900 } 901 902 /* 903 * Detach an interface, removing it from the list of "active" interfaces. 904 * If vmove flag is set on entry to if_detach_internal(), perform only a 905 * limited subset of cleanup tasks, given that we are moving an ifnet from 906 * one vnet to another, where it must be fully operational. 907 * 908 * XXXRW: There are some significant questions about event ordering, and 909 * how to prevent things from starting to use the interface during detach. 910 */ 911 void 912 if_detach(struct ifnet *ifp) 913 { 914 915 CURVNET_SET_QUIET(ifp->if_vnet); 916 if_detach_internal(ifp, 0, NULL); 917 CURVNET_RESTORE(); 918 } 919 920 /* 921 * The vmove flag, if set, indicates that we are called from a callpath 922 * that is moving an interface to a different vnet instance. 923 * 924 * The shutdown flag, if set, indicates that we are called in the 925 * process of shutting down a vnet instance. Currently only the 926 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 927 * on a vnet instance shutdown without this flag being set, e.g., when 928 * the cloned interfaces are destoyed as first thing of teardown. 929 */ 930 static int 931 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 932 { 933 struct ifaddr *ifa; 934 int i; 935 struct domain *dp; 936 struct ifnet *iter; 937 int found = 0; 938 #ifdef VIMAGE 939 int shutdown; 940 941 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 942 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 943 #endif 944 IFNET_WLOCK(); 945 TAILQ_FOREACH(iter, &V_ifnet, if_link) 946 if (iter == ifp) { 947 TAILQ_REMOVE(&V_ifnet, ifp, if_link); 948 found = 1; 949 break; 950 } 951 IFNET_WUNLOCK(); 952 if (!found) { 953 /* 954 * While we would want to panic here, we cannot 955 * guarantee that the interface is indeed still on 956 * the list given we don't hold locks all the way. 957 */ 958 return (ENOENT); 959 #if 0 960 if (vmove) 961 panic("%s: ifp=%p not on the ifnet tailq %p", 962 __func__, ifp, &V_ifnet); 963 else 964 return; /* XXX this should panic as well? */ 965 #endif 966 } 967 968 /* 969 * At this point we know the interface still was on the ifnet list 970 * and we removed it so we are in a stable state. 971 */ 972 #ifdef VIMAGE 973 curvnet->vnet_ifcnt--; 974 #endif 975 976 /* 977 * In any case (destroy or vmove) detach us from the groups 978 * and remove/wait for pending events on the taskq. 979 * XXX-BZ in theory an interface could still enqueue a taskq change? 980 */ 981 if_delgroups(ifp); 982 983 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 984 985 /* 986 * Check if this is a cloned interface or not. Must do even if 987 * shutting down as a if_vmove_reclaim() would move the ifp and 988 * the if_clone_addgroup() will have a corrupted string overwise 989 * from a gibberish pointer. 990 */ 991 if (vmove && ifcp != NULL) 992 *ifcp = if_clone_findifc(ifp); 993 994 if_down(ifp); 995 996 #ifdef VIMAGE 997 /* 998 * On VNET shutdown abort here as the stack teardown will do all 999 * the work top-down for us. 1000 */ 1001 if (shutdown) { 1002 /* 1003 * In case of a vmove we are done here without error. 1004 * If we would signal an error it would lead to the same 1005 * abort as if we did not find the ifnet anymore. 1006 * if_detach() calls us in void context and does not care 1007 * about an early abort notification, so life is splendid :) 1008 */ 1009 goto finish_vnet_shutdown; 1010 } 1011 #endif 1012 1013 /* 1014 * At this point we are not tearing down a VNET and are either 1015 * going to destroy or vmove the interface and have to cleanup 1016 * accordingly. 1017 */ 1018 1019 /* 1020 * Remove routes and flush queues. 1021 */ 1022 #ifdef ALTQ 1023 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1024 altq_disable(&ifp->if_snd); 1025 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1026 altq_detach(&ifp->if_snd); 1027 #endif 1028 1029 if_purgeaddrs(ifp); 1030 1031 #ifdef INET 1032 in_ifdetach(ifp); 1033 #endif 1034 1035 #ifdef INET6 1036 /* 1037 * Remove all IPv6 kernel structs related to ifp. This should be done 1038 * before removing routing entries below, since IPv6 interface direct 1039 * routes are expected to be removed by the IPv6-specific kernel API. 1040 * Otherwise, the kernel will detect some inconsistency and bark it. 1041 */ 1042 in6_ifdetach(ifp); 1043 #endif 1044 if_purgemaddrs(ifp); 1045 1046 /* Announce that the interface is gone. */ 1047 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1048 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1049 if (IS_DEFAULT_VNET(curvnet)) 1050 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1051 1052 if (!vmove) { 1053 /* 1054 * Prevent further calls into the device driver via ifnet. 1055 */ 1056 if_dead(ifp); 1057 1058 /* 1059 * Remove link ifaddr pointer and maybe decrement if_index. 1060 * Clean up all addresses. 1061 */ 1062 ifp->if_addr = NULL; 1063 1064 /* We can now free link ifaddr. */ 1065 IF_ADDR_WLOCK(ifp); 1066 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 1067 ifa = TAILQ_FIRST(&ifp->if_addrhead); 1068 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 1069 IF_ADDR_WUNLOCK(ifp); 1070 ifa_free(ifa); 1071 } else 1072 IF_ADDR_WUNLOCK(ifp); 1073 } 1074 1075 rt_flushifroutes(ifp); 1076 1077 #ifdef VIMAGE 1078 finish_vnet_shutdown: 1079 #endif 1080 /* 1081 * We cannot hold the lock over dom_ifdetach calls as they might 1082 * sleep, for example trying to drain a callout, thus open up the 1083 * theoretical race with re-attaching. 1084 */ 1085 IF_AFDATA_LOCK(ifp); 1086 i = ifp->if_afdata_initialized; 1087 ifp->if_afdata_initialized = 0; 1088 IF_AFDATA_UNLOCK(ifp); 1089 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1090 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1091 (*dp->dom_ifdetach)(ifp, 1092 ifp->if_afdata[dp->dom_family]); 1093 ifp->if_afdata[dp->dom_family] = NULL; 1094 } 1095 } 1096 1097 return (0); 1098 } 1099 1100 #ifdef VIMAGE 1101 /* 1102 * if_vmove() performs a limited version of if_detach() in current 1103 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1104 * An attempt is made to shrink if_index in current vnet, find an 1105 * unused if_index in target vnet and calls if_grow() if necessary, 1106 * and finally find an unused if_xname for the target vnet. 1107 */ 1108 static void 1109 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1110 { 1111 struct if_clone *ifc; 1112 u_int bif_dlt, bif_hdrlen; 1113 int rc; 1114 1115 /* 1116 * if_detach_internal() will call the eventhandler to notify 1117 * interface departure. That will detach if_bpf. We need to 1118 * safe the dlt and hdrlen so we can re-attach it later. 1119 */ 1120 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1121 1122 /* 1123 * Detach from current vnet, but preserve LLADDR info, do not 1124 * mark as dead etc. so that the ifnet can be reattached later. 1125 * If we cannot find it, we lost the race to someone else. 1126 */ 1127 rc = if_detach_internal(ifp, 1, &ifc); 1128 if (rc != 0) 1129 return; 1130 1131 /* 1132 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1133 * the if_index for that vnet if possible. 1134 * 1135 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1136 * or we'd lock on one vnet and unlock on another. 1137 */ 1138 IFNET_WLOCK(); 1139 ifindex_free_locked(ifp->if_index); 1140 IFNET_WUNLOCK(); 1141 1142 /* 1143 * Perform interface-specific reassignment tasks, if provided by 1144 * the driver. 1145 */ 1146 if (ifp->if_reassign != NULL) 1147 ifp->if_reassign(ifp, new_vnet, NULL); 1148 1149 /* 1150 * Switch to the context of the target vnet. 1151 */ 1152 CURVNET_SET_QUIET(new_vnet); 1153 1154 IFNET_WLOCK(); 1155 ifp->if_index = ifindex_alloc(); 1156 ifnet_setbyindex_locked(ifp->if_index, ifp); 1157 IFNET_WUNLOCK(); 1158 1159 if_attach_internal(ifp, 1, ifc); 1160 1161 if (ifp->if_bpf == NULL) 1162 bpfattach(ifp, bif_dlt, bif_hdrlen); 1163 1164 CURVNET_RESTORE(); 1165 } 1166 1167 /* 1168 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1169 */ 1170 static int 1171 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1172 { 1173 struct prison *pr; 1174 struct ifnet *difp; 1175 int shutdown; 1176 1177 /* Try to find the prison within our visibility. */ 1178 sx_slock(&allprison_lock); 1179 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1180 sx_sunlock(&allprison_lock); 1181 if (pr == NULL) 1182 return (ENXIO); 1183 prison_hold_locked(pr); 1184 mtx_unlock(&pr->pr_mtx); 1185 1186 /* Do not try to move the iface from and to the same prison. */ 1187 if (pr->pr_vnet == ifp->if_vnet) { 1188 prison_free(pr); 1189 return (EEXIST); 1190 } 1191 1192 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1193 /* XXX Lock interfaces to avoid races. */ 1194 CURVNET_SET_QUIET(pr->pr_vnet); 1195 difp = ifunit(ifname); 1196 if (difp != NULL) { 1197 CURVNET_RESTORE(); 1198 prison_free(pr); 1199 return (EEXIST); 1200 } 1201 1202 /* Make sure the VNET is stable. */ 1203 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1204 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1205 if (shutdown) { 1206 CURVNET_RESTORE(); 1207 prison_free(pr); 1208 return (EBUSY); 1209 } 1210 CURVNET_RESTORE(); 1211 1212 /* Move the interface into the child jail/vnet. */ 1213 if_vmove(ifp, pr->pr_vnet); 1214 1215 /* Report the new if_xname back to the userland. */ 1216 sprintf(ifname, "%s", ifp->if_xname); 1217 1218 prison_free(pr); 1219 return (0); 1220 } 1221 1222 static int 1223 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1224 { 1225 struct prison *pr; 1226 struct vnet *vnet_dst; 1227 struct ifnet *ifp; 1228 int shutdown; 1229 1230 /* Try to find the prison within our visibility. */ 1231 sx_slock(&allprison_lock); 1232 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1233 sx_sunlock(&allprison_lock); 1234 if (pr == NULL) 1235 return (ENXIO); 1236 prison_hold_locked(pr); 1237 mtx_unlock(&pr->pr_mtx); 1238 1239 /* Make sure the named iface exists in the source prison/vnet. */ 1240 CURVNET_SET(pr->pr_vnet); 1241 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1242 if (ifp == NULL) { 1243 CURVNET_RESTORE(); 1244 prison_free(pr); 1245 return (ENXIO); 1246 } 1247 1248 /* Do not try to move the iface from and to the same prison. */ 1249 vnet_dst = TD_TO_VNET(td); 1250 if (vnet_dst == ifp->if_vnet) { 1251 CURVNET_RESTORE(); 1252 prison_free(pr); 1253 return (EEXIST); 1254 } 1255 1256 /* Make sure the VNET is stable. */ 1257 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1258 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1259 if (shutdown) { 1260 CURVNET_RESTORE(); 1261 prison_free(pr); 1262 return (EBUSY); 1263 } 1264 1265 /* Get interface back from child jail/vnet. */ 1266 if_vmove(ifp, vnet_dst); 1267 CURVNET_RESTORE(); 1268 1269 /* Report the new if_xname back to the userland. */ 1270 sprintf(ifname, "%s", ifp->if_xname); 1271 1272 prison_free(pr); 1273 return (0); 1274 } 1275 #endif /* VIMAGE */ 1276 1277 /* 1278 * Add a group to an interface 1279 */ 1280 int 1281 if_addgroup(struct ifnet *ifp, const char *groupname) 1282 { 1283 struct ifg_list *ifgl; 1284 struct ifg_group *ifg = NULL; 1285 struct ifg_member *ifgm; 1286 int new = 0; 1287 1288 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1289 groupname[strlen(groupname) - 1] <= '9') 1290 return (EINVAL); 1291 1292 IFNET_WLOCK(); 1293 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1294 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1295 IFNET_WUNLOCK(); 1296 return (EEXIST); 1297 } 1298 1299 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1300 M_NOWAIT)) == NULL) { 1301 IFNET_WUNLOCK(); 1302 return (ENOMEM); 1303 } 1304 1305 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1306 M_TEMP, M_NOWAIT)) == NULL) { 1307 free(ifgl, M_TEMP); 1308 IFNET_WUNLOCK(); 1309 return (ENOMEM); 1310 } 1311 1312 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1313 if (!strcmp(ifg->ifg_group, groupname)) 1314 break; 1315 1316 if (ifg == NULL) { 1317 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1318 M_TEMP, M_NOWAIT)) == NULL) { 1319 free(ifgl, M_TEMP); 1320 free(ifgm, M_TEMP); 1321 IFNET_WUNLOCK(); 1322 return (ENOMEM); 1323 } 1324 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1325 ifg->ifg_refcnt = 0; 1326 TAILQ_INIT(&ifg->ifg_members); 1327 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1328 new = 1; 1329 } 1330 1331 ifg->ifg_refcnt++; 1332 ifgl->ifgl_group = ifg; 1333 ifgm->ifgm_ifp = ifp; 1334 1335 IF_ADDR_WLOCK(ifp); 1336 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1337 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1338 IF_ADDR_WUNLOCK(ifp); 1339 1340 IFNET_WUNLOCK(); 1341 1342 if (new) 1343 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1344 EVENTHANDLER_INVOKE(group_change_event, groupname); 1345 1346 return (0); 1347 } 1348 1349 /* 1350 * Remove a group from an interface 1351 */ 1352 int 1353 if_delgroup(struct ifnet *ifp, const char *groupname) 1354 { 1355 struct ifg_list *ifgl; 1356 struct ifg_member *ifgm; 1357 1358 IFNET_WLOCK(); 1359 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1360 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1361 break; 1362 if (ifgl == NULL) { 1363 IFNET_WUNLOCK(); 1364 return (ENOENT); 1365 } 1366 1367 IF_ADDR_WLOCK(ifp); 1368 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1369 IF_ADDR_WUNLOCK(ifp); 1370 1371 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1372 if (ifgm->ifgm_ifp == ifp) 1373 break; 1374 1375 if (ifgm != NULL) { 1376 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 1377 free(ifgm, M_TEMP); 1378 } 1379 1380 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1381 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1382 IFNET_WUNLOCK(); 1383 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1384 free(ifgl->ifgl_group, M_TEMP); 1385 } else 1386 IFNET_WUNLOCK(); 1387 1388 free(ifgl, M_TEMP); 1389 1390 EVENTHANDLER_INVOKE(group_change_event, groupname); 1391 1392 return (0); 1393 } 1394 1395 /* 1396 * Remove an interface from all groups 1397 */ 1398 static void 1399 if_delgroups(struct ifnet *ifp) 1400 { 1401 struct ifg_list *ifgl; 1402 struct ifg_member *ifgm; 1403 char groupname[IFNAMSIZ]; 1404 1405 IFNET_WLOCK(); 1406 while (!TAILQ_EMPTY(&ifp->if_groups)) { 1407 ifgl = TAILQ_FIRST(&ifp->if_groups); 1408 1409 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1410 1411 IF_ADDR_WLOCK(ifp); 1412 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1413 IF_ADDR_WUNLOCK(ifp); 1414 1415 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1416 if (ifgm->ifgm_ifp == ifp) 1417 break; 1418 1419 if (ifgm != NULL) { 1420 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1421 ifgm_next); 1422 free(ifgm, M_TEMP); 1423 } 1424 1425 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1426 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1427 IFNET_WUNLOCK(); 1428 EVENTHANDLER_INVOKE(group_detach_event, 1429 ifgl->ifgl_group); 1430 free(ifgl->ifgl_group, M_TEMP); 1431 } else 1432 IFNET_WUNLOCK(); 1433 1434 free(ifgl, M_TEMP); 1435 1436 EVENTHANDLER_INVOKE(group_change_event, groupname); 1437 1438 IFNET_WLOCK(); 1439 } 1440 IFNET_WUNLOCK(); 1441 } 1442 1443 /* 1444 * Stores all groups from an interface in memory pointed 1445 * to by data 1446 */ 1447 static int 1448 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 1449 { 1450 int len, error; 1451 struct ifg_list *ifgl; 1452 struct ifg_req ifgrq, *ifgp; 1453 struct ifgroupreq *ifgr = data; 1454 1455 if (ifgr->ifgr_len == 0) { 1456 IF_ADDR_RLOCK(ifp); 1457 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1458 ifgr->ifgr_len += sizeof(struct ifg_req); 1459 IF_ADDR_RUNLOCK(ifp); 1460 return (0); 1461 } 1462 1463 len = ifgr->ifgr_len; 1464 ifgp = ifgr->ifgr_groups; 1465 /* XXX: wire */ 1466 IF_ADDR_RLOCK(ifp); 1467 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1468 if (len < sizeof(ifgrq)) { 1469 IF_ADDR_RUNLOCK(ifp); 1470 return (EINVAL); 1471 } 1472 bzero(&ifgrq, sizeof ifgrq); 1473 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1474 sizeof(ifgrq.ifgrq_group)); 1475 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1476 IF_ADDR_RUNLOCK(ifp); 1477 return (error); 1478 } 1479 len -= sizeof(ifgrq); 1480 ifgp++; 1481 } 1482 IF_ADDR_RUNLOCK(ifp); 1483 1484 return (0); 1485 } 1486 1487 /* 1488 * Stores all members of a group in memory pointed to by data 1489 */ 1490 static int 1491 if_getgroupmembers(struct ifgroupreq *data) 1492 { 1493 struct ifgroupreq *ifgr = data; 1494 struct ifg_group *ifg; 1495 struct ifg_member *ifgm; 1496 struct ifg_req ifgrq, *ifgp; 1497 int len, error; 1498 1499 IFNET_RLOCK(); 1500 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1501 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1502 break; 1503 if (ifg == NULL) { 1504 IFNET_RUNLOCK(); 1505 return (ENOENT); 1506 } 1507 1508 if (ifgr->ifgr_len == 0) { 1509 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1510 ifgr->ifgr_len += sizeof(ifgrq); 1511 IFNET_RUNLOCK(); 1512 return (0); 1513 } 1514 1515 len = ifgr->ifgr_len; 1516 ifgp = ifgr->ifgr_groups; 1517 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1518 if (len < sizeof(ifgrq)) { 1519 IFNET_RUNLOCK(); 1520 return (EINVAL); 1521 } 1522 bzero(&ifgrq, sizeof ifgrq); 1523 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1524 sizeof(ifgrq.ifgrq_member)); 1525 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1526 IFNET_RUNLOCK(); 1527 return (error); 1528 } 1529 len -= sizeof(ifgrq); 1530 ifgp++; 1531 } 1532 IFNET_RUNLOCK(); 1533 1534 return (0); 1535 } 1536 1537 /* 1538 * Return counter values from counter(9)s stored in ifnet. 1539 */ 1540 uint64_t 1541 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1542 { 1543 1544 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1545 1546 return (counter_u64_fetch(ifp->if_counters[cnt])); 1547 } 1548 1549 /* 1550 * Increase an ifnet counter. Usually used for counters shared 1551 * between the stack and a driver, but function supports them all. 1552 */ 1553 void 1554 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1555 { 1556 1557 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1558 1559 counter_u64_add(ifp->if_counters[cnt], inc); 1560 } 1561 1562 /* 1563 * Copy data from ifnet to userland API structure if_data. 1564 */ 1565 void 1566 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1567 { 1568 1569 ifd->ifi_type = ifp->if_type; 1570 ifd->ifi_physical = 0; 1571 ifd->ifi_addrlen = ifp->if_addrlen; 1572 ifd->ifi_hdrlen = ifp->if_hdrlen; 1573 ifd->ifi_link_state = ifp->if_link_state; 1574 ifd->ifi_vhid = 0; 1575 ifd->ifi_datalen = sizeof(struct if_data); 1576 ifd->ifi_mtu = ifp->if_mtu; 1577 ifd->ifi_metric = ifp->if_metric; 1578 ifd->ifi_baudrate = ifp->if_baudrate; 1579 ifd->ifi_hwassist = ifp->if_hwassist; 1580 ifd->ifi_epoch = ifp->if_epoch; 1581 ifd->ifi_lastchange = ifp->if_lastchange; 1582 1583 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1584 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1585 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1586 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1587 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1588 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1589 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1590 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1591 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1592 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1593 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1594 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1595 } 1596 1597 /* 1598 * Wrapper functions for struct ifnet address list locking macros. These are 1599 * used by kernel modules to avoid encoding programming interface or binary 1600 * interface assumptions that may be violated when kernel-internal locking 1601 * approaches change. 1602 */ 1603 void 1604 if_addr_rlock(struct ifnet *ifp) 1605 { 1606 1607 IF_ADDR_RLOCK(ifp); 1608 } 1609 1610 void 1611 if_addr_runlock(struct ifnet *ifp) 1612 { 1613 1614 IF_ADDR_RUNLOCK(ifp); 1615 } 1616 1617 void 1618 if_maddr_rlock(if_t ifp) 1619 { 1620 1621 IF_ADDR_RLOCK((struct ifnet *)ifp); 1622 } 1623 1624 void 1625 if_maddr_runlock(if_t ifp) 1626 { 1627 1628 IF_ADDR_RUNLOCK((struct ifnet *)ifp); 1629 } 1630 1631 /* 1632 * Initialization, destruction and refcounting functions for ifaddrs. 1633 */ 1634 struct ifaddr * 1635 ifa_alloc(size_t size, int flags) 1636 { 1637 struct ifaddr *ifa; 1638 1639 KASSERT(size >= sizeof(struct ifaddr), 1640 ("%s: invalid size %zu", __func__, size)); 1641 1642 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1643 if (ifa == NULL) 1644 return (NULL); 1645 1646 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1647 goto fail; 1648 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1649 goto fail; 1650 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1651 goto fail; 1652 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1653 goto fail; 1654 1655 refcount_init(&ifa->ifa_refcnt, 1); 1656 1657 return (ifa); 1658 1659 fail: 1660 /* free(NULL) is okay */ 1661 counter_u64_free(ifa->ifa_opackets); 1662 counter_u64_free(ifa->ifa_ipackets); 1663 counter_u64_free(ifa->ifa_obytes); 1664 counter_u64_free(ifa->ifa_ibytes); 1665 free(ifa, M_IFADDR); 1666 1667 return (NULL); 1668 } 1669 1670 void 1671 ifa_ref(struct ifaddr *ifa) 1672 { 1673 1674 refcount_acquire(&ifa->ifa_refcnt); 1675 } 1676 1677 void 1678 ifa_free(struct ifaddr *ifa) 1679 { 1680 1681 if (refcount_release(&ifa->ifa_refcnt)) { 1682 counter_u64_free(ifa->ifa_opackets); 1683 counter_u64_free(ifa->ifa_ipackets); 1684 counter_u64_free(ifa->ifa_obytes); 1685 counter_u64_free(ifa->ifa_ibytes); 1686 free(ifa, M_IFADDR); 1687 } 1688 } 1689 1690 static int 1691 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa, 1692 struct sockaddr *ia) 1693 { 1694 int error; 1695 struct rt_addrinfo info; 1696 struct sockaddr_dl null_sdl; 1697 struct ifnet *ifp; 1698 1699 ifp = ifa->ifa_ifp; 1700 1701 bzero(&info, sizeof(info)); 1702 if (cmd != RTM_DELETE) 1703 info.rti_ifp = V_loif; 1704 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; 1705 info.rti_info[RTAX_DST] = ia; 1706 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1707 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type); 1708 1709 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib); 1710 1711 if (error != 0) 1712 log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n", 1713 __func__, otype, if_name(ifp), error); 1714 1715 return (error); 1716 } 1717 1718 int 1719 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1720 { 1721 1722 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia)); 1723 } 1724 1725 int 1726 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1727 { 1728 1729 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia)); 1730 } 1731 1732 int 1733 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1734 { 1735 1736 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia)); 1737 } 1738 1739 /* 1740 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1741 * structs used to represent other address families, it is necessary 1742 * to perform a different comparison. 1743 */ 1744 1745 #define sa_dl_equal(a1, a2) \ 1746 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1747 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1748 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1749 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1750 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1751 1752 /* 1753 * Locate an interface based on a complete address. 1754 */ 1755 /*ARGSUSED*/ 1756 static struct ifaddr * 1757 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref) 1758 { 1759 struct ifnet *ifp; 1760 struct ifaddr *ifa; 1761 1762 IFNET_RLOCK_NOSLEEP(); 1763 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1764 IF_ADDR_RLOCK(ifp); 1765 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1766 if (ifa->ifa_addr->sa_family != addr->sa_family) 1767 continue; 1768 if (sa_equal(addr, ifa->ifa_addr)) { 1769 if (getref) 1770 ifa_ref(ifa); 1771 IF_ADDR_RUNLOCK(ifp); 1772 goto done; 1773 } 1774 /* IP6 doesn't have broadcast */ 1775 if ((ifp->if_flags & IFF_BROADCAST) && 1776 ifa->ifa_broadaddr && 1777 ifa->ifa_broadaddr->sa_len != 0 && 1778 sa_equal(ifa->ifa_broadaddr, addr)) { 1779 if (getref) 1780 ifa_ref(ifa); 1781 IF_ADDR_RUNLOCK(ifp); 1782 goto done; 1783 } 1784 } 1785 IF_ADDR_RUNLOCK(ifp); 1786 } 1787 ifa = NULL; 1788 done: 1789 IFNET_RUNLOCK_NOSLEEP(); 1790 return (ifa); 1791 } 1792 1793 struct ifaddr * 1794 ifa_ifwithaddr(const struct sockaddr *addr) 1795 { 1796 1797 return (ifa_ifwithaddr_internal(addr, 1)); 1798 } 1799 1800 int 1801 ifa_ifwithaddr_check(const struct sockaddr *addr) 1802 { 1803 1804 return (ifa_ifwithaddr_internal(addr, 0) != NULL); 1805 } 1806 1807 /* 1808 * Locate an interface based on the broadcast address. 1809 */ 1810 /* ARGSUSED */ 1811 struct ifaddr * 1812 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1813 { 1814 struct ifnet *ifp; 1815 struct ifaddr *ifa; 1816 1817 IFNET_RLOCK_NOSLEEP(); 1818 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1819 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1820 continue; 1821 IF_ADDR_RLOCK(ifp); 1822 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1823 if (ifa->ifa_addr->sa_family != addr->sa_family) 1824 continue; 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 ifa_ref(ifa); 1830 IF_ADDR_RUNLOCK(ifp); 1831 goto done; 1832 } 1833 } 1834 IF_ADDR_RUNLOCK(ifp); 1835 } 1836 ifa = NULL; 1837 done: 1838 IFNET_RUNLOCK_NOSLEEP(); 1839 return (ifa); 1840 } 1841 1842 /* 1843 * Locate the point to point interface with a given destination address. 1844 */ 1845 /*ARGSUSED*/ 1846 struct ifaddr * 1847 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1848 { 1849 struct ifnet *ifp; 1850 struct ifaddr *ifa; 1851 1852 IFNET_RLOCK_NOSLEEP(); 1853 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1854 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1855 continue; 1856 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1857 continue; 1858 IF_ADDR_RLOCK(ifp); 1859 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1860 if (ifa->ifa_addr->sa_family != addr->sa_family) 1861 continue; 1862 if (ifa->ifa_dstaddr != NULL && 1863 sa_equal(addr, ifa->ifa_dstaddr)) { 1864 ifa_ref(ifa); 1865 IF_ADDR_RUNLOCK(ifp); 1866 goto done; 1867 } 1868 } 1869 IF_ADDR_RUNLOCK(ifp); 1870 } 1871 ifa = NULL; 1872 done: 1873 IFNET_RUNLOCK_NOSLEEP(); 1874 return (ifa); 1875 } 1876 1877 /* 1878 * Find an interface on a specific network. If many, choice 1879 * is most specific found. 1880 */ 1881 struct ifaddr * 1882 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1883 { 1884 struct ifnet *ifp; 1885 struct ifaddr *ifa; 1886 struct ifaddr *ifa_maybe = NULL; 1887 u_int af = addr->sa_family; 1888 const char *addr_data = addr->sa_data, *cplim; 1889 1890 /* 1891 * AF_LINK addresses can be looked up directly by their index number, 1892 * so do that if we can. 1893 */ 1894 if (af == AF_LINK) { 1895 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 1896 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1897 return (ifaddr_byindex(sdl->sdl_index)); 1898 } 1899 1900 /* 1901 * Scan though each interface, looking for ones that have addresses 1902 * in this address family and the requested fib. Maintain a reference 1903 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 1904 * kept it stable when we move onto the next interface. 1905 */ 1906 IFNET_RLOCK_NOSLEEP(); 1907 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1908 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1909 continue; 1910 IF_ADDR_RLOCK(ifp); 1911 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1912 const char *cp, *cp2, *cp3; 1913 1914 if (ifa->ifa_addr->sa_family != af) 1915 next: continue; 1916 if (af == AF_INET && 1917 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1918 /* 1919 * This is a bit broken as it doesn't 1920 * take into account that the remote end may 1921 * be a single node in the network we are 1922 * looking for. 1923 * The trouble is that we don't know the 1924 * netmask for the remote end. 1925 */ 1926 if (ifa->ifa_dstaddr != NULL && 1927 sa_equal(addr, ifa->ifa_dstaddr)) { 1928 ifa_ref(ifa); 1929 IF_ADDR_RUNLOCK(ifp); 1930 goto done; 1931 } 1932 } else { 1933 /* 1934 * Scan all the bits in the ifa's address. 1935 * If a bit dissagrees with what we are 1936 * looking for, mask it with the netmask 1937 * to see if it really matters. 1938 * (A byte at a time) 1939 */ 1940 if (ifa->ifa_netmask == 0) 1941 continue; 1942 cp = addr_data; 1943 cp2 = ifa->ifa_addr->sa_data; 1944 cp3 = ifa->ifa_netmask->sa_data; 1945 cplim = ifa->ifa_netmask->sa_len 1946 + (char *)ifa->ifa_netmask; 1947 while (cp3 < cplim) 1948 if ((*cp++ ^ *cp2++) & *cp3++) 1949 goto next; /* next address! */ 1950 /* 1951 * If the netmask of what we just found 1952 * is more specific than what we had before 1953 * (if we had one), or if the virtual status 1954 * of new prefix is better than of the old one, 1955 * then remember the new one before continuing 1956 * to search for an even better one. 1957 */ 1958 if (ifa_maybe == NULL || 1959 ifa_preferred(ifa_maybe, ifa) || 1960 rn_refines((caddr_t)ifa->ifa_netmask, 1961 (caddr_t)ifa_maybe->ifa_netmask)) { 1962 if (ifa_maybe != NULL) 1963 ifa_free(ifa_maybe); 1964 ifa_maybe = ifa; 1965 ifa_ref(ifa_maybe); 1966 } 1967 } 1968 } 1969 IF_ADDR_RUNLOCK(ifp); 1970 } 1971 ifa = ifa_maybe; 1972 ifa_maybe = NULL; 1973 done: 1974 IFNET_RUNLOCK_NOSLEEP(); 1975 if (ifa_maybe != NULL) 1976 ifa_free(ifa_maybe); 1977 return (ifa); 1978 } 1979 1980 /* 1981 * Find an interface address specific to an interface best matching 1982 * a given address. 1983 */ 1984 struct ifaddr * 1985 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1986 { 1987 struct ifaddr *ifa; 1988 const char *cp, *cp2, *cp3; 1989 char *cplim; 1990 struct ifaddr *ifa_maybe = NULL; 1991 u_int af = addr->sa_family; 1992 1993 if (af >= AF_MAX) 1994 return (NULL); 1995 IF_ADDR_RLOCK(ifp); 1996 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1997 if (ifa->ifa_addr->sa_family != af) 1998 continue; 1999 if (ifa_maybe == NULL) 2000 ifa_maybe = ifa; 2001 if (ifa->ifa_netmask == 0) { 2002 if (sa_equal(addr, ifa->ifa_addr) || 2003 (ifa->ifa_dstaddr && 2004 sa_equal(addr, ifa->ifa_dstaddr))) 2005 goto done; 2006 continue; 2007 } 2008 if (ifp->if_flags & IFF_POINTOPOINT) { 2009 if (sa_equal(addr, ifa->ifa_dstaddr)) 2010 goto done; 2011 } else { 2012 cp = addr->sa_data; 2013 cp2 = ifa->ifa_addr->sa_data; 2014 cp3 = ifa->ifa_netmask->sa_data; 2015 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2016 for (; cp3 < cplim; cp3++) 2017 if ((*cp++ ^ *cp2++) & *cp3) 2018 break; 2019 if (cp3 == cplim) 2020 goto done; 2021 } 2022 } 2023 ifa = ifa_maybe; 2024 done: 2025 if (ifa != NULL) 2026 ifa_ref(ifa); 2027 IF_ADDR_RUNLOCK(ifp); 2028 return (ifa); 2029 } 2030 2031 /* 2032 * See whether new ifa is better than current one: 2033 * 1) A non-virtual one is preferred over virtual. 2034 * 2) A virtual in master state preferred over any other state. 2035 * 2036 * Used in several address selecting functions. 2037 */ 2038 int 2039 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2040 { 2041 2042 return (cur->ifa_carp && (!next->ifa_carp || 2043 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2044 } 2045 2046 #include <net/if_llatbl.h> 2047 2048 /* 2049 * Default action when installing a route with a Link Level gateway. 2050 * Lookup an appropriate real ifa to point to. 2051 * This should be moved to /sys/net/link.c eventually. 2052 */ 2053 static void 2054 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 2055 { 2056 struct ifaddr *ifa, *oifa; 2057 struct sockaddr *dst; 2058 struct ifnet *ifp; 2059 2060 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) || 2061 ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL)) 2062 return; 2063 ifa = ifaof_ifpforaddr(dst, ifp); 2064 if (ifa) { 2065 oifa = rt->rt_ifa; 2066 rt->rt_ifa = ifa; 2067 ifa_free(oifa); 2068 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2069 ifa->ifa_rtrequest(cmd, rt, info); 2070 } 2071 } 2072 2073 struct sockaddr_dl * 2074 link_alloc_sdl(size_t size, int flags) 2075 { 2076 2077 return (malloc(size, M_TEMP, flags)); 2078 } 2079 2080 void 2081 link_free_sdl(struct sockaddr *sa) 2082 { 2083 free(sa, M_TEMP); 2084 } 2085 2086 /* 2087 * Fills in given sdl with interface basic info. 2088 * Returns pointer to filled sdl. 2089 */ 2090 struct sockaddr_dl * 2091 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2092 { 2093 struct sockaddr_dl *sdl; 2094 2095 sdl = (struct sockaddr_dl *)paddr; 2096 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2097 sdl->sdl_len = sizeof(struct sockaddr_dl); 2098 sdl->sdl_family = AF_LINK; 2099 sdl->sdl_index = ifp->if_index; 2100 sdl->sdl_type = iftype; 2101 2102 return (sdl); 2103 } 2104 2105 /* 2106 * Mark an interface down and notify protocols of 2107 * the transition. 2108 */ 2109 static void 2110 if_unroute(struct ifnet *ifp, int flag, int fam) 2111 { 2112 struct ifaddr *ifa; 2113 2114 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2115 2116 ifp->if_flags &= ~flag; 2117 getmicrotime(&ifp->if_lastchange); 2118 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2119 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2120 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2121 ifp->if_qflush(ifp); 2122 2123 if (ifp->if_carp) 2124 (*carp_linkstate_p)(ifp); 2125 rt_ifmsg(ifp); 2126 } 2127 2128 /* 2129 * Mark an interface up and notify protocols of 2130 * the transition. 2131 */ 2132 static void 2133 if_route(struct ifnet *ifp, int flag, int fam) 2134 { 2135 struct ifaddr *ifa; 2136 2137 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2138 2139 ifp->if_flags |= flag; 2140 getmicrotime(&ifp->if_lastchange); 2141 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2142 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2143 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2144 if (ifp->if_carp) 2145 (*carp_linkstate_p)(ifp); 2146 rt_ifmsg(ifp); 2147 #ifdef INET6 2148 in6_if_up(ifp); 2149 #endif 2150 } 2151 2152 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2153 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2154 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2155 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2156 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2157 int (*vlan_setcookie_p)(struct ifnet *, void *); 2158 void *(*vlan_cookie_p)(struct ifnet *); 2159 2160 /* 2161 * Handle a change in the interface link state. To avoid LORs 2162 * between driver lock and upper layer locks, as well as possible 2163 * recursions, we post event to taskqueue, and all job 2164 * is done in static do_link_state_change(). 2165 */ 2166 void 2167 if_link_state_change(struct ifnet *ifp, int link_state) 2168 { 2169 /* Return if state hasn't changed. */ 2170 if (ifp->if_link_state == link_state) 2171 return; 2172 2173 ifp->if_link_state = link_state; 2174 2175 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2176 } 2177 2178 static void 2179 do_link_state_change(void *arg, int pending) 2180 { 2181 struct ifnet *ifp = (struct ifnet *)arg; 2182 int link_state = ifp->if_link_state; 2183 CURVNET_SET(ifp->if_vnet); 2184 2185 /* Notify that the link state has changed. */ 2186 rt_ifmsg(ifp); 2187 if (ifp->if_vlantrunk != NULL) 2188 (*vlan_link_state_p)(ifp); 2189 2190 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2191 ifp->if_l2com != NULL) 2192 (*ng_ether_link_state_p)(ifp, link_state); 2193 if (ifp->if_carp) 2194 (*carp_linkstate_p)(ifp); 2195 if (ifp->if_bridge) 2196 (*bridge_linkstate_p)(ifp); 2197 if (ifp->if_lagg) 2198 (*lagg_linkstate_p)(ifp, link_state); 2199 2200 if (IS_DEFAULT_VNET(curvnet)) 2201 devctl_notify("IFNET", ifp->if_xname, 2202 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2203 NULL); 2204 if (pending > 1) 2205 if_printf(ifp, "%d link states coalesced\n", pending); 2206 if (log_link_state_change) 2207 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 2208 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2209 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2210 CURVNET_RESTORE(); 2211 } 2212 2213 /* 2214 * Mark an interface down and notify protocols of 2215 * the transition. 2216 */ 2217 void 2218 if_down(struct ifnet *ifp) 2219 { 2220 2221 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2222 } 2223 2224 /* 2225 * Mark an interface up and notify protocols of 2226 * the transition. 2227 */ 2228 void 2229 if_up(struct ifnet *ifp) 2230 { 2231 2232 if_route(ifp, IFF_UP, AF_UNSPEC); 2233 } 2234 2235 /* 2236 * Flush an interface queue. 2237 */ 2238 void 2239 if_qflush(struct ifnet *ifp) 2240 { 2241 struct mbuf *m, *n; 2242 struct ifaltq *ifq; 2243 2244 ifq = &ifp->if_snd; 2245 IFQ_LOCK(ifq); 2246 #ifdef ALTQ 2247 if (ALTQ_IS_ENABLED(ifq)) 2248 ALTQ_PURGE(ifq); 2249 #endif 2250 n = ifq->ifq_head; 2251 while ((m = n) != NULL) { 2252 n = m->m_nextpkt; 2253 m_freem(m); 2254 } 2255 ifq->ifq_head = 0; 2256 ifq->ifq_tail = 0; 2257 ifq->ifq_len = 0; 2258 IFQ_UNLOCK(ifq); 2259 } 2260 2261 /* 2262 * Map interface name to interface structure pointer, with or without 2263 * returning a reference. 2264 */ 2265 struct ifnet * 2266 ifunit_ref(const char *name) 2267 { 2268 struct ifnet *ifp; 2269 2270 IFNET_RLOCK_NOSLEEP(); 2271 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2272 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2273 !(ifp->if_flags & IFF_DYING)) 2274 break; 2275 } 2276 if (ifp != NULL) 2277 if_ref(ifp); 2278 IFNET_RUNLOCK_NOSLEEP(); 2279 return (ifp); 2280 } 2281 2282 struct ifnet * 2283 ifunit(const char *name) 2284 { 2285 struct ifnet *ifp; 2286 2287 IFNET_RLOCK_NOSLEEP(); 2288 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2289 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2290 break; 2291 } 2292 IFNET_RUNLOCK_NOSLEEP(); 2293 return (ifp); 2294 } 2295 2296 /* 2297 * Hardware specific interface ioctls. 2298 */ 2299 static int 2300 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2301 { 2302 struct ifreq *ifr; 2303 int error = 0, do_ifup = 0; 2304 int new_flags, temp_flags; 2305 size_t namelen, onamelen; 2306 size_t descrlen; 2307 char *descrbuf, *odescrbuf; 2308 char new_name[IFNAMSIZ]; 2309 struct ifaddr *ifa; 2310 struct sockaddr_dl *sdl; 2311 2312 ifr = (struct ifreq *)data; 2313 switch (cmd) { 2314 case SIOCGIFINDEX: 2315 ifr->ifr_index = ifp->if_index; 2316 break; 2317 2318 case SIOCGIFFLAGS: 2319 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2320 ifr->ifr_flags = temp_flags & 0xffff; 2321 ifr->ifr_flagshigh = temp_flags >> 16; 2322 break; 2323 2324 case SIOCGIFCAP: 2325 ifr->ifr_reqcap = ifp->if_capabilities; 2326 ifr->ifr_curcap = ifp->if_capenable; 2327 break; 2328 2329 #ifdef MAC 2330 case SIOCGIFMAC: 2331 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2332 break; 2333 #endif 2334 2335 case SIOCGIFMETRIC: 2336 ifr->ifr_metric = ifp->if_metric; 2337 break; 2338 2339 case SIOCGIFMTU: 2340 ifr->ifr_mtu = ifp->if_mtu; 2341 break; 2342 2343 case SIOCGIFPHYS: 2344 /* XXXGL: did this ever worked? */ 2345 ifr->ifr_phys = 0; 2346 break; 2347 2348 case SIOCGIFDESCR: 2349 error = 0; 2350 sx_slock(&ifdescr_sx); 2351 if (ifp->if_description == NULL) 2352 error = ENOMSG; 2353 else { 2354 /* space for terminating nul */ 2355 descrlen = strlen(ifp->if_description) + 1; 2356 if (ifr->ifr_buffer.length < descrlen) 2357 ifr->ifr_buffer.buffer = NULL; 2358 else 2359 error = copyout(ifp->if_description, 2360 ifr->ifr_buffer.buffer, descrlen); 2361 ifr->ifr_buffer.length = descrlen; 2362 } 2363 sx_sunlock(&ifdescr_sx); 2364 break; 2365 2366 case SIOCSIFDESCR: 2367 error = priv_check(td, PRIV_NET_SETIFDESCR); 2368 if (error) 2369 return (error); 2370 2371 /* 2372 * Copy only (length-1) bytes to make sure that 2373 * if_description is always nul terminated. The 2374 * length parameter is supposed to count the 2375 * terminating nul in. 2376 */ 2377 if (ifr->ifr_buffer.length > ifdescr_maxlen) 2378 return (ENAMETOOLONG); 2379 else if (ifr->ifr_buffer.length == 0) 2380 descrbuf = NULL; 2381 else { 2382 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR, 2383 M_WAITOK | M_ZERO); 2384 error = copyin(ifr->ifr_buffer.buffer, descrbuf, 2385 ifr->ifr_buffer.length - 1); 2386 if (error) { 2387 free(descrbuf, M_IFDESCR); 2388 break; 2389 } 2390 } 2391 2392 sx_xlock(&ifdescr_sx); 2393 odescrbuf = ifp->if_description; 2394 ifp->if_description = descrbuf; 2395 sx_xunlock(&ifdescr_sx); 2396 2397 getmicrotime(&ifp->if_lastchange); 2398 free(odescrbuf, M_IFDESCR); 2399 break; 2400 2401 case SIOCGIFFIB: 2402 ifr->ifr_fib = ifp->if_fib; 2403 break; 2404 2405 case SIOCSIFFIB: 2406 error = priv_check(td, PRIV_NET_SETIFFIB); 2407 if (error) 2408 return (error); 2409 if (ifr->ifr_fib >= rt_numfibs) 2410 return (EINVAL); 2411 2412 ifp->if_fib = ifr->ifr_fib; 2413 break; 2414 2415 case SIOCSIFFLAGS: 2416 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2417 if (error) 2418 return (error); 2419 /* 2420 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2421 * check, so we don't need special handling here yet. 2422 */ 2423 new_flags = (ifr->ifr_flags & 0xffff) | 2424 (ifr->ifr_flagshigh << 16); 2425 if (ifp->if_flags & IFF_UP && 2426 (new_flags & IFF_UP) == 0) { 2427 if_down(ifp); 2428 } else if (new_flags & IFF_UP && 2429 (ifp->if_flags & IFF_UP) == 0) { 2430 do_ifup = 1; 2431 } 2432 /* See if permanently promiscuous mode bit is about to flip */ 2433 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2434 if (new_flags & IFF_PPROMISC) 2435 ifp->if_flags |= IFF_PROMISC; 2436 else if (ifp->if_pcount == 0) 2437 ifp->if_flags &= ~IFF_PROMISC; 2438 if (log_promisc_mode_change) 2439 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2440 ifp->if_xname, 2441 ((new_flags & IFF_PPROMISC) ? 2442 "enabled" : "disabled")); 2443 } 2444 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2445 (new_flags &~ IFF_CANTCHANGE); 2446 if (ifp->if_ioctl) { 2447 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2448 } 2449 if (do_ifup) 2450 if_up(ifp); 2451 getmicrotime(&ifp->if_lastchange); 2452 break; 2453 2454 case SIOCSIFCAP: 2455 error = priv_check(td, PRIV_NET_SETIFCAP); 2456 if (error) 2457 return (error); 2458 if (ifp->if_ioctl == NULL) 2459 return (EOPNOTSUPP); 2460 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2461 return (EINVAL); 2462 error = (*ifp->if_ioctl)(ifp, cmd, data); 2463 if (error == 0) 2464 getmicrotime(&ifp->if_lastchange); 2465 break; 2466 2467 #ifdef MAC 2468 case SIOCSIFMAC: 2469 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2470 break; 2471 #endif 2472 2473 case SIOCSIFNAME: 2474 error = priv_check(td, PRIV_NET_SETIFNAME); 2475 if (error) 2476 return (error); 2477 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 2478 if (error != 0) 2479 return (error); 2480 if (new_name[0] == '\0') 2481 return (EINVAL); 2482 if (new_name[IFNAMSIZ-1] != '\0') { 2483 new_name[IFNAMSIZ-1] = '\0'; 2484 if (strlen(new_name) == IFNAMSIZ-1) 2485 return (EINVAL); 2486 } 2487 if (ifunit(new_name) != NULL) 2488 return (EEXIST); 2489 2490 /* 2491 * XXX: Locking. Nothing else seems to lock if_flags, 2492 * and there are numerous other races with the 2493 * ifunit() checks not being atomic with namespace 2494 * changes (renames, vmoves, if_attach, etc). 2495 */ 2496 ifp->if_flags |= IFF_RENAMING; 2497 2498 /* Announce the departure of the interface. */ 2499 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2500 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2501 2502 log(LOG_INFO, "%s: changing name to '%s'\n", 2503 ifp->if_xname, new_name); 2504 2505 IF_ADDR_WLOCK(ifp); 2506 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2507 ifa = ifp->if_addr; 2508 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2509 namelen = strlen(new_name); 2510 onamelen = sdl->sdl_nlen; 2511 /* 2512 * Move the address if needed. This is safe because we 2513 * allocate space for a name of length IFNAMSIZ when we 2514 * create this in if_attach(). 2515 */ 2516 if (namelen != onamelen) { 2517 bcopy(sdl->sdl_data + onamelen, 2518 sdl->sdl_data + namelen, sdl->sdl_alen); 2519 } 2520 bcopy(new_name, sdl->sdl_data, namelen); 2521 sdl->sdl_nlen = namelen; 2522 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2523 bzero(sdl->sdl_data, onamelen); 2524 while (namelen != 0) 2525 sdl->sdl_data[--namelen] = 0xff; 2526 IF_ADDR_WUNLOCK(ifp); 2527 2528 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2529 /* Announce the return of the interface. */ 2530 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2531 2532 ifp->if_flags &= ~IFF_RENAMING; 2533 break; 2534 2535 #ifdef VIMAGE 2536 case SIOCSIFVNET: 2537 error = priv_check(td, PRIV_NET_SETIFVNET); 2538 if (error) 2539 return (error); 2540 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2541 break; 2542 #endif 2543 2544 case SIOCSIFMETRIC: 2545 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2546 if (error) 2547 return (error); 2548 ifp->if_metric = ifr->ifr_metric; 2549 getmicrotime(&ifp->if_lastchange); 2550 break; 2551 2552 case SIOCSIFPHYS: 2553 error = priv_check(td, PRIV_NET_SETIFPHYS); 2554 if (error) 2555 return (error); 2556 if (ifp->if_ioctl == NULL) 2557 return (EOPNOTSUPP); 2558 error = (*ifp->if_ioctl)(ifp, cmd, data); 2559 if (error == 0) 2560 getmicrotime(&ifp->if_lastchange); 2561 break; 2562 2563 case SIOCSIFMTU: 2564 { 2565 u_long oldmtu = ifp->if_mtu; 2566 2567 error = priv_check(td, PRIV_NET_SETIFMTU); 2568 if (error) 2569 return (error); 2570 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2571 return (EINVAL); 2572 if (ifp->if_ioctl == NULL) 2573 return (EOPNOTSUPP); 2574 error = (*ifp->if_ioctl)(ifp, cmd, data); 2575 if (error == 0) { 2576 getmicrotime(&ifp->if_lastchange); 2577 rt_ifmsg(ifp); 2578 } 2579 /* 2580 * If the link MTU changed, do network layer specific procedure. 2581 */ 2582 if (ifp->if_mtu != oldmtu) { 2583 #ifdef INET6 2584 nd6_setmtu(ifp); 2585 #endif 2586 rt_updatemtu(ifp); 2587 } 2588 break; 2589 } 2590 2591 case SIOCADDMULTI: 2592 case SIOCDELMULTI: 2593 if (cmd == SIOCADDMULTI) 2594 error = priv_check(td, PRIV_NET_ADDMULTI); 2595 else 2596 error = priv_check(td, PRIV_NET_DELMULTI); 2597 if (error) 2598 return (error); 2599 2600 /* Don't allow group membership on non-multicast interfaces. */ 2601 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2602 return (EOPNOTSUPP); 2603 2604 /* Don't let users screw up protocols' entries. */ 2605 if (ifr->ifr_addr.sa_family != AF_LINK) 2606 return (EINVAL); 2607 2608 if (cmd == SIOCADDMULTI) { 2609 struct ifmultiaddr *ifma; 2610 2611 /* 2612 * Userland is only permitted to join groups once 2613 * via the if_addmulti() KPI, because it cannot hold 2614 * struct ifmultiaddr * between calls. It may also 2615 * lose a race while we check if the membership 2616 * already exists. 2617 */ 2618 IF_ADDR_RLOCK(ifp); 2619 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2620 IF_ADDR_RUNLOCK(ifp); 2621 if (ifma != NULL) 2622 error = EADDRINUSE; 2623 else 2624 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2625 } else { 2626 error = if_delmulti(ifp, &ifr->ifr_addr); 2627 } 2628 if (error == 0) 2629 getmicrotime(&ifp->if_lastchange); 2630 break; 2631 2632 case SIOCSIFPHYADDR: 2633 case SIOCDIFPHYADDR: 2634 #ifdef INET6 2635 case SIOCSIFPHYADDR_IN6: 2636 #endif 2637 case SIOCSIFMEDIA: 2638 case SIOCSIFGENERIC: 2639 error = priv_check(td, PRIV_NET_HWIOCTL); 2640 if (error) 2641 return (error); 2642 if (ifp->if_ioctl == NULL) 2643 return (EOPNOTSUPP); 2644 error = (*ifp->if_ioctl)(ifp, cmd, data); 2645 if (error == 0) 2646 getmicrotime(&ifp->if_lastchange); 2647 break; 2648 2649 case SIOCGIFSTATUS: 2650 case SIOCGIFPSRCADDR: 2651 case SIOCGIFPDSTADDR: 2652 case SIOCGIFMEDIA: 2653 case SIOCGIFXMEDIA: 2654 case SIOCGIFGENERIC: 2655 if (ifp->if_ioctl == NULL) 2656 return (EOPNOTSUPP); 2657 error = (*ifp->if_ioctl)(ifp, cmd, data); 2658 break; 2659 2660 case SIOCSIFLLADDR: 2661 error = priv_check(td, PRIV_NET_SETLLADDR); 2662 if (error) 2663 return (error); 2664 error = if_setlladdr(ifp, 2665 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2666 break; 2667 2668 case SIOCAIFGROUP: 2669 { 2670 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2671 2672 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2673 if (error) 2674 return (error); 2675 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2676 return (error); 2677 break; 2678 } 2679 2680 case SIOCGIFGROUP: 2681 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2682 return (error); 2683 break; 2684 2685 case SIOCDIFGROUP: 2686 { 2687 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2688 2689 error = priv_check(td, PRIV_NET_DELIFGROUP); 2690 if (error) 2691 return (error); 2692 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2693 return (error); 2694 break; 2695 } 2696 2697 default: 2698 error = ENOIOCTL; 2699 break; 2700 } 2701 return (error); 2702 } 2703 2704 /* COMPAT_SVR4 */ 2705 #define OSIOCGIFCONF _IOWR('i', 20, struct ifconf) 2706 2707 #ifdef COMPAT_FREEBSD32 2708 struct ifconf32 { 2709 int32_t ifc_len; 2710 union { 2711 uint32_t ifcu_buf; 2712 uint32_t ifcu_req; 2713 } ifc_ifcu; 2714 }; 2715 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2716 #endif 2717 2718 /* 2719 * Interface ioctls. 2720 */ 2721 int 2722 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2723 { 2724 struct ifnet *ifp; 2725 struct ifreq *ifr; 2726 int error; 2727 int oif_flags; 2728 #ifdef VIMAGE 2729 int shutdown; 2730 #endif 2731 2732 CURVNET_SET(so->so_vnet); 2733 #ifdef VIMAGE 2734 /* Make sure the VNET is stable. */ 2735 shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET && 2736 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 2737 if (shutdown) { 2738 CURVNET_RESTORE(); 2739 return (EBUSY); 2740 } 2741 #endif 2742 2743 2744 switch (cmd) { 2745 case SIOCGIFCONF: 2746 case OSIOCGIFCONF: /* COMPAT_SVR4 */ 2747 error = ifconf(cmd, data); 2748 CURVNET_RESTORE(); 2749 return (error); 2750 2751 #ifdef COMPAT_FREEBSD32 2752 case SIOCGIFCONF32: 2753 { 2754 struct ifconf32 *ifc32; 2755 struct ifconf ifc; 2756 2757 ifc32 = (struct ifconf32 *)data; 2758 ifc.ifc_len = ifc32->ifc_len; 2759 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2760 2761 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2762 CURVNET_RESTORE(); 2763 if (error == 0) 2764 ifc32->ifc_len = ifc.ifc_len; 2765 return (error); 2766 } 2767 #endif 2768 } 2769 ifr = (struct ifreq *)data; 2770 2771 switch (cmd) { 2772 #ifdef VIMAGE 2773 case SIOCSIFRVNET: 2774 error = priv_check(td, PRIV_NET_SETIFVNET); 2775 if (error == 0) 2776 error = if_vmove_reclaim(td, ifr->ifr_name, 2777 ifr->ifr_jid); 2778 CURVNET_RESTORE(); 2779 return (error); 2780 #endif 2781 case SIOCIFCREATE: 2782 case SIOCIFCREATE2: 2783 error = priv_check(td, PRIV_NET_IFCREATE); 2784 if (error == 0) 2785 error = if_clone_create(ifr->ifr_name, 2786 sizeof(ifr->ifr_name), 2787 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL); 2788 CURVNET_RESTORE(); 2789 return (error); 2790 case SIOCIFDESTROY: 2791 error = priv_check(td, PRIV_NET_IFDESTROY); 2792 if (error == 0) 2793 error = if_clone_destroy(ifr->ifr_name); 2794 CURVNET_RESTORE(); 2795 return (error); 2796 2797 case SIOCIFGCLONERS: 2798 error = if_clone_list((struct if_clonereq *)data); 2799 CURVNET_RESTORE(); 2800 return (error); 2801 case SIOCGIFGMEMB: 2802 error = if_getgroupmembers((struct ifgroupreq *)data); 2803 CURVNET_RESTORE(); 2804 return (error); 2805 #if defined(INET) || defined(INET6) 2806 case SIOCSVH: 2807 case SIOCGVH: 2808 if (carp_ioctl_p == NULL) 2809 error = EPROTONOSUPPORT; 2810 else 2811 error = (*carp_ioctl_p)(ifr, cmd, td); 2812 CURVNET_RESTORE(); 2813 return (error); 2814 #endif 2815 } 2816 2817 ifp = ifunit_ref(ifr->ifr_name); 2818 if (ifp == NULL) { 2819 CURVNET_RESTORE(); 2820 return (ENXIO); 2821 } 2822 2823 error = ifhwioctl(cmd, ifp, data, td); 2824 if (error != ENOIOCTL) { 2825 if_rele(ifp); 2826 CURVNET_RESTORE(); 2827 return (error); 2828 } 2829 2830 oif_flags = ifp->if_flags; 2831 if (so->so_proto == NULL) { 2832 if_rele(ifp); 2833 CURVNET_RESTORE(); 2834 return (EOPNOTSUPP); 2835 } 2836 2837 /* 2838 * Pass the request on to the socket control method, and if the 2839 * latter returns EOPNOTSUPP, directly to the interface. 2840 * 2841 * Make an exception for the legacy SIOCSIF* requests. Drivers 2842 * trust SIOCSIFADDR et al to come from an already privileged 2843 * layer, and do not perform any credentials checks or input 2844 * validation. 2845 */ 2846 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 2847 ifp, td)); 2848 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2849 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2850 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2851 error = (*ifp->if_ioctl)(ifp, cmd, data); 2852 2853 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2854 #ifdef INET6 2855 if (ifp->if_flags & IFF_UP) 2856 in6_if_up(ifp); 2857 #endif 2858 } 2859 if_rele(ifp); 2860 CURVNET_RESTORE(); 2861 return (error); 2862 } 2863 2864 /* 2865 * The code common to handling reference counted flags, 2866 * e.g., in ifpromisc() and if_allmulti(). 2867 * The "pflag" argument can specify a permanent mode flag to check, 2868 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2869 * 2870 * Only to be used on stack-owned flags, not driver-owned flags. 2871 */ 2872 static int 2873 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 2874 { 2875 struct ifreq ifr; 2876 int error; 2877 int oldflags, oldcount; 2878 2879 /* Sanity checks to catch programming errors */ 2880 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 2881 ("%s: setting driver-owned flag %d", __func__, flag)); 2882 2883 if (onswitch) 2884 KASSERT(*refcount >= 0, 2885 ("%s: increment negative refcount %d for flag %d", 2886 __func__, *refcount, flag)); 2887 else 2888 KASSERT(*refcount > 0, 2889 ("%s: decrement non-positive refcount %d for flag %d", 2890 __func__, *refcount, flag)); 2891 2892 /* In case this mode is permanent, just touch refcount */ 2893 if (ifp->if_flags & pflag) { 2894 *refcount += onswitch ? 1 : -1; 2895 return (0); 2896 } 2897 2898 /* Save ifnet parameters for if_ioctl() may fail */ 2899 oldcount = *refcount; 2900 oldflags = ifp->if_flags; 2901 2902 /* 2903 * See if we aren't the only and touching refcount is enough. 2904 * Actually toggle interface flag if we are the first or last. 2905 */ 2906 if (onswitch) { 2907 if ((*refcount)++) 2908 return (0); 2909 ifp->if_flags |= flag; 2910 } else { 2911 if (--(*refcount)) 2912 return (0); 2913 ifp->if_flags &= ~flag; 2914 } 2915 2916 /* Call down the driver since we've changed interface flags */ 2917 if (ifp->if_ioctl == NULL) { 2918 error = EOPNOTSUPP; 2919 goto recover; 2920 } 2921 ifr.ifr_flags = ifp->if_flags & 0xffff; 2922 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2923 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2924 if (error) 2925 goto recover; 2926 /* Notify userland that interface flags have changed */ 2927 rt_ifmsg(ifp); 2928 return (0); 2929 2930 recover: 2931 /* Recover after driver error */ 2932 *refcount = oldcount; 2933 ifp->if_flags = oldflags; 2934 return (error); 2935 } 2936 2937 /* 2938 * Set/clear promiscuous mode on interface ifp based on the truth value 2939 * of pswitch. The calls are reference counted so that only the first 2940 * "on" request actually has an effect, as does the final "off" request. 2941 * Results are undefined if the "off" and "on" requests are not matched. 2942 */ 2943 int 2944 ifpromisc(struct ifnet *ifp, int pswitch) 2945 { 2946 int error; 2947 int oldflags = ifp->if_flags; 2948 2949 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 2950 &ifp->if_pcount, pswitch); 2951 /* If promiscuous mode status has changed, log a message */ 2952 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 2953 log_promisc_mode_change) 2954 log(LOG_INFO, "%s: promiscuous mode %s\n", 2955 ifp->if_xname, 2956 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 2957 return (error); 2958 } 2959 2960 /* 2961 * Return interface configuration 2962 * of system. List may be used 2963 * in later ioctl's (above) to get 2964 * other information. 2965 */ 2966 /*ARGSUSED*/ 2967 static int 2968 ifconf(u_long cmd, caddr_t data) 2969 { 2970 struct ifconf *ifc = (struct ifconf *)data; 2971 struct ifnet *ifp; 2972 struct ifaddr *ifa; 2973 struct ifreq ifr; 2974 struct sbuf *sb; 2975 int error, full = 0, valid_len, max_len; 2976 2977 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2978 max_len = MAXPHYS - 1; 2979 2980 /* Prevent hostile input from being able to crash the system */ 2981 if (ifc->ifc_len <= 0) 2982 return (EINVAL); 2983 2984 again: 2985 if (ifc->ifc_len <= max_len) { 2986 max_len = ifc->ifc_len; 2987 full = 1; 2988 } 2989 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2990 max_len = 0; 2991 valid_len = 0; 2992 2993 IFNET_RLOCK(); 2994 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2995 int addrs; 2996 2997 /* 2998 * Zero the ifr_name buffer to make sure we don't 2999 * disclose the contents of the stack. 3000 */ 3001 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 3002 3003 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3004 >= sizeof(ifr.ifr_name)) { 3005 sbuf_delete(sb); 3006 IFNET_RUNLOCK(); 3007 return (ENAMETOOLONG); 3008 } 3009 3010 addrs = 0; 3011 IF_ADDR_RLOCK(ifp); 3012 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3013 struct sockaddr *sa = ifa->ifa_addr; 3014 3015 if (prison_if(curthread->td_ucred, sa) != 0) 3016 continue; 3017 addrs++; 3018 /* COMPAT_SVR4 */ 3019 if (cmd == OSIOCGIFCONF) { 3020 struct osockaddr *osa = 3021 (struct osockaddr *)&ifr.ifr_addr; 3022 ifr.ifr_addr = *sa; 3023 osa->sa_family = sa->sa_family; 3024 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3025 max_len += sizeof(ifr); 3026 } else 3027 if (sa->sa_len <= sizeof(*sa)) { 3028 ifr.ifr_addr = *sa; 3029 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3030 max_len += sizeof(ifr); 3031 } else { 3032 sbuf_bcat(sb, &ifr, 3033 offsetof(struct ifreq, ifr_addr)); 3034 max_len += offsetof(struct ifreq, ifr_addr); 3035 sbuf_bcat(sb, sa, sa->sa_len); 3036 max_len += sa->sa_len; 3037 } 3038 3039 if (sbuf_error(sb) == 0) 3040 valid_len = sbuf_len(sb); 3041 } 3042 IF_ADDR_RUNLOCK(ifp); 3043 if (addrs == 0) { 3044 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 3045 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3046 max_len += sizeof(ifr); 3047 3048 if (sbuf_error(sb) == 0) 3049 valid_len = sbuf_len(sb); 3050 } 3051 } 3052 IFNET_RUNLOCK(); 3053 3054 /* 3055 * If we didn't allocate enough space (uncommon), try again. If 3056 * we have already allocated as much space as we are allowed, 3057 * return what we've got. 3058 */ 3059 if (valid_len != max_len && !full) { 3060 sbuf_delete(sb); 3061 goto again; 3062 } 3063 3064 ifc->ifc_len = valid_len; 3065 sbuf_finish(sb); 3066 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3067 sbuf_delete(sb); 3068 return (error); 3069 } 3070 3071 /* 3072 * Just like ifpromisc(), but for all-multicast-reception mode. 3073 */ 3074 int 3075 if_allmulti(struct ifnet *ifp, int onswitch) 3076 { 3077 3078 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3079 } 3080 3081 struct ifmultiaddr * 3082 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3083 { 3084 struct ifmultiaddr *ifma; 3085 3086 IF_ADDR_LOCK_ASSERT(ifp); 3087 3088 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3089 if (sa->sa_family == AF_LINK) { 3090 if (sa_dl_equal(ifma->ifma_addr, sa)) 3091 break; 3092 } else { 3093 if (sa_equal(ifma->ifma_addr, sa)) 3094 break; 3095 } 3096 } 3097 3098 return ifma; 3099 } 3100 3101 /* 3102 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3103 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3104 * the ifnet multicast address list here, so the caller must do that and 3105 * other setup work (such as notifying the device driver). The reference 3106 * count is initialized to 1. 3107 */ 3108 static struct ifmultiaddr * 3109 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3110 int mflags) 3111 { 3112 struct ifmultiaddr *ifma; 3113 struct sockaddr *dupsa; 3114 3115 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3116 M_ZERO); 3117 if (ifma == NULL) 3118 return (NULL); 3119 3120 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3121 if (dupsa == NULL) { 3122 free(ifma, M_IFMADDR); 3123 return (NULL); 3124 } 3125 bcopy(sa, dupsa, sa->sa_len); 3126 ifma->ifma_addr = dupsa; 3127 3128 ifma->ifma_ifp = ifp; 3129 ifma->ifma_refcount = 1; 3130 ifma->ifma_protospec = NULL; 3131 3132 if (llsa == NULL) { 3133 ifma->ifma_lladdr = NULL; 3134 return (ifma); 3135 } 3136 3137 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3138 if (dupsa == NULL) { 3139 free(ifma->ifma_addr, M_IFMADDR); 3140 free(ifma, M_IFMADDR); 3141 return (NULL); 3142 } 3143 bcopy(llsa, dupsa, llsa->sa_len); 3144 ifma->ifma_lladdr = dupsa; 3145 3146 return (ifma); 3147 } 3148 3149 /* 3150 * if_freemulti: free ifmultiaddr structure and possibly attached related 3151 * addresses. The caller is responsible for implementing reference 3152 * counting, notifying the driver, handling routing messages, and releasing 3153 * any dependent link layer state. 3154 */ 3155 static void 3156 if_freemulti(struct ifmultiaddr *ifma) 3157 { 3158 3159 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3160 ifma->ifma_refcount)); 3161 3162 if (ifma->ifma_lladdr != NULL) 3163 free(ifma->ifma_lladdr, M_IFMADDR); 3164 free(ifma->ifma_addr, M_IFMADDR); 3165 free(ifma, M_IFMADDR); 3166 } 3167 3168 /* 3169 * Register an additional multicast address with a network interface. 3170 * 3171 * - If the address is already present, bump the reference count on the 3172 * address and return. 3173 * - If the address is not link-layer, look up a link layer address. 3174 * - Allocate address structures for one or both addresses, and attach to the 3175 * multicast address list on the interface. If automatically adding a link 3176 * layer address, the protocol address will own a reference to the link 3177 * layer address, to be freed when it is freed. 3178 * - Notify the network device driver of an addition to the multicast address 3179 * list. 3180 * 3181 * 'sa' points to caller-owned memory with the desired multicast address. 3182 * 3183 * 'retifma' will be used to return a pointer to the resulting multicast 3184 * address reference, if desired. 3185 */ 3186 int 3187 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3188 struct ifmultiaddr **retifma) 3189 { 3190 struct ifmultiaddr *ifma, *ll_ifma; 3191 struct sockaddr *llsa; 3192 struct sockaddr_dl sdl; 3193 int error; 3194 3195 /* 3196 * If the address is already present, return a new reference to it; 3197 * otherwise, allocate storage and set up a new address. 3198 */ 3199 IF_ADDR_WLOCK(ifp); 3200 ifma = if_findmulti(ifp, sa); 3201 if (ifma != NULL) { 3202 ifma->ifma_refcount++; 3203 if (retifma != NULL) 3204 *retifma = ifma; 3205 IF_ADDR_WUNLOCK(ifp); 3206 return (0); 3207 } 3208 3209 /* 3210 * The address isn't already present; resolve the protocol address 3211 * into a link layer address, and then look that up, bump its 3212 * refcount or allocate an ifma for that also. 3213 * Most link layer resolving functions returns address data which 3214 * fits inside default sockaddr_dl structure. However callback 3215 * can allocate another sockaddr structure, in that case we need to 3216 * free it later. 3217 */ 3218 llsa = NULL; 3219 ll_ifma = NULL; 3220 if (ifp->if_resolvemulti != NULL) { 3221 /* Provide called function with buffer size information */ 3222 sdl.sdl_len = sizeof(sdl); 3223 llsa = (struct sockaddr *)&sdl; 3224 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3225 if (error) 3226 goto unlock_out; 3227 } 3228 3229 /* 3230 * Allocate the new address. Don't hook it up yet, as we may also 3231 * need to allocate a link layer multicast address. 3232 */ 3233 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3234 if (ifma == NULL) { 3235 error = ENOMEM; 3236 goto free_llsa_out; 3237 } 3238 3239 /* 3240 * If a link layer address is found, we'll need to see if it's 3241 * already present in the address list, or allocate is as well. 3242 * When this block finishes, the link layer address will be on the 3243 * list. 3244 */ 3245 if (llsa != NULL) { 3246 ll_ifma = if_findmulti(ifp, llsa); 3247 if (ll_ifma == NULL) { 3248 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3249 if (ll_ifma == NULL) { 3250 --ifma->ifma_refcount; 3251 if_freemulti(ifma); 3252 error = ENOMEM; 3253 goto free_llsa_out; 3254 } 3255 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3256 ifma_link); 3257 } else 3258 ll_ifma->ifma_refcount++; 3259 ifma->ifma_llifma = ll_ifma; 3260 } 3261 3262 /* 3263 * We now have a new multicast address, ifma, and possibly a new or 3264 * referenced link layer address. Add the primary address to the 3265 * ifnet address list. 3266 */ 3267 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3268 3269 if (retifma != NULL) 3270 *retifma = ifma; 3271 3272 /* 3273 * Must generate the message while holding the lock so that 'ifma' 3274 * pointer is still valid. 3275 */ 3276 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3277 IF_ADDR_WUNLOCK(ifp); 3278 3279 /* 3280 * We are certain we have added something, so call down to the 3281 * interface to let them know about it. 3282 */ 3283 if (ifp->if_ioctl != NULL) { 3284 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3285 } 3286 3287 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3288 link_free_sdl(llsa); 3289 3290 return (0); 3291 3292 free_llsa_out: 3293 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3294 link_free_sdl(llsa); 3295 3296 unlock_out: 3297 IF_ADDR_WUNLOCK(ifp); 3298 return (error); 3299 } 3300 3301 /* 3302 * Delete a multicast group membership by network-layer group address. 3303 * 3304 * Returns ENOENT if the entry could not be found. If ifp no longer 3305 * exists, results are undefined. This entry point should only be used 3306 * from subsystems which do appropriate locking to hold ifp for the 3307 * duration of the call. 3308 * Network-layer protocol domains must use if_delmulti_ifma(). 3309 */ 3310 int 3311 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3312 { 3313 struct ifmultiaddr *ifma; 3314 int lastref; 3315 #ifdef INVARIANTS 3316 struct ifnet *oifp; 3317 3318 IFNET_RLOCK_NOSLEEP(); 3319 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3320 if (ifp == oifp) 3321 break; 3322 if (ifp != oifp) 3323 ifp = NULL; 3324 IFNET_RUNLOCK_NOSLEEP(); 3325 3326 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3327 #endif 3328 if (ifp == NULL) 3329 return (ENOENT); 3330 3331 IF_ADDR_WLOCK(ifp); 3332 lastref = 0; 3333 ifma = if_findmulti(ifp, sa); 3334 if (ifma != NULL) 3335 lastref = if_delmulti_locked(ifp, ifma, 0); 3336 IF_ADDR_WUNLOCK(ifp); 3337 3338 if (ifma == NULL) 3339 return (ENOENT); 3340 3341 if (lastref && ifp->if_ioctl != NULL) { 3342 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3343 } 3344 3345 return (0); 3346 } 3347 3348 /* 3349 * Delete all multicast group membership for an interface. 3350 * Should be used to quickly flush all multicast filters. 3351 */ 3352 void 3353 if_delallmulti(struct ifnet *ifp) 3354 { 3355 struct ifmultiaddr *ifma; 3356 struct ifmultiaddr *next; 3357 3358 IF_ADDR_WLOCK(ifp); 3359 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3360 if_delmulti_locked(ifp, ifma, 0); 3361 IF_ADDR_WUNLOCK(ifp); 3362 } 3363 3364 /* 3365 * Delete a multicast group membership by group membership pointer. 3366 * Network-layer protocol domains must use this routine. 3367 * 3368 * It is safe to call this routine if the ifp disappeared. 3369 */ 3370 void 3371 if_delmulti_ifma(struct ifmultiaddr *ifma) 3372 { 3373 struct ifnet *ifp; 3374 int lastref; 3375 3376 ifp = ifma->ifma_ifp; 3377 #ifdef DIAGNOSTIC 3378 if (ifp == NULL) { 3379 printf("%s: ifma_ifp seems to be detached\n", __func__); 3380 } else { 3381 struct ifnet *oifp; 3382 3383 IFNET_RLOCK_NOSLEEP(); 3384 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3385 if (ifp == oifp) 3386 break; 3387 if (ifp != oifp) { 3388 printf("%s: ifnet %p disappeared\n", __func__, ifp); 3389 ifp = NULL; 3390 } 3391 IFNET_RUNLOCK_NOSLEEP(); 3392 } 3393 #endif 3394 /* 3395 * If and only if the ifnet instance exists: Acquire the address lock. 3396 */ 3397 if (ifp != NULL) 3398 IF_ADDR_WLOCK(ifp); 3399 3400 lastref = if_delmulti_locked(ifp, ifma, 0); 3401 3402 if (ifp != NULL) { 3403 /* 3404 * If and only if the ifnet instance exists: 3405 * Release the address lock. 3406 * If the group was left: update the hardware hash filter. 3407 */ 3408 IF_ADDR_WUNLOCK(ifp); 3409 if (lastref && ifp->if_ioctl != NULL) { 3410 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3411 } 3412 } 3413 } 3414 3415 /* 3416 * Perform deletion of network-layer and/or link-layer multicast address. 3417 * 3418 * Return 0 if the reference count was decremented. 3419 * Return 1 if the final reference was released, indicating that the 3420 * hardware hash filter should be reprogrammed. 3421 */ 3422 static int 3423 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3424 { 3425 struct ifmultiaddr *ll_ifma; 3426 3427 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3428 KASSERT(ifma->ifma_ifp == ifp, 3429 ("%s: inconsistent ifp %p", __func__, ifp)); 3430 IF_ADDR_WLOCK_ASSERT(ifp); 3431 } 3432 3433 ifp = ifma->ifma_ifp; 3434 3435 /* 3436 * If the ifnet is detaching, null out references to ifnet, 3437 * so that upper protocol layers will notice, and not attempt 3438 * to obtain locks for an ifnet which no longer exists. The 3439 * routing socket announcement must happen before the ifnet 3440 * instance is detached from the system. 3441 */ 3442 if (detaching) { 3443 #ifdef DIAGNOSTIC 3444 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3445 #endif 3446 /* 3447 * ifp may already be nulled out if we are being reentered 3448 * to delete the ll_ifma. 3449 */ 3450 if (ifp != NULL) { 3451 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3452 ifma->ifma_ifp = NULL; 3453 } 3454 } 3455 3456 if (--ifma->ifma_refcount > 0) 3457 return 0; 3458 3459 /* 3460 * If this ifma is a network-layer ifma, a link-layer ifma may 3461 * have been associated with it. Release it first if so. 3462 */ 3463 ll_ifma = ifma->ifma_llifma; 3464 if (ll_ifma != NULL) { 3465 KASSERT(ifma->ifma_lladdr != NULL, 3466 ("%s: llifma w/o lladdr", __func__)); 3467 if (detaching) 3468 ll_ifma->ifma_ifp = NULL; /* XXX */ 3469 if (--ll_ifma->ifma_refcount == 0) { 3470 if (ifp != NULL) { 3471 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3472 ifma_link); 3473 } 3474 if_freemulti(ll_ifma); 3475 } 3476 } 3477 3478 if (ifp != NULL) 3479 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3480 3481 if_freemulti(ifma); 3482 3483 /* 3484 * The last reference to this instance of struct ifmultiaddr 3485 * was released; the hardware should be notified of this change. 3486 */ 3487 return 1; 3488 } 3489 3490 /* 3491 * Set the link layer address on an interface. 3492 * 3493 * At this time we only support certain types of interfaces, 3494 * and we don't allow the length of the address to change. 3495 * 3496 * Set noinline to be dtrace-friendly 3497 */ 3498 __noinline int 3499 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3500 { 3501 struct sockaddr_dl *sdl; 3502 struct ifaddr *ifa; 3503 struct ifreq ifr; 3504 3505 IF_ADDR_RLOCK(ifp); 3506 ifa = ifp->if_addr; 3507 if (ifa == NULL) { 3508 IF_ADDR_RUNLOCK(ifp); 3509 return (EINVAL); 3510 } 3511 ifa_ref(ifa); 3512 IF_ADDR_RUNLOCK(ifp); 3513 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3514 if (sdl == NULL) { 3515 ifa_free(ifa); 3516 return (EINVAL); 3517 } 3518 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3519 ifa_free(ifa); 3520 return (EINVAL); 3521 } 3522 switch (ifp->if_type) { 3523 case IFT_ETHER: 3524 case IFT_FDDI: 3525 case IFT_XETHER: 3526 case IFT_ISO88025: 3527 case IFT_L2VLAN: 3528 case IFT_BRIDGE: 3529 case IFT_ARCNET: 3530 case IFT_IEEE8023ADLAG: 3531 case IFT_IEEE80211: 3532 bcopy(lladdr, LLADDR(sdl), len); 3533 ifa_free(ifa); 3534 break; 3535 default: 3536 ifa_free(ifa); 3537 return (ENODEV); 3538 } 3539 3540 /* 3541 * If the interface is already up, we need 3542 * to re-init it in order to reprogram its 3543 * address filter. 3544 */ 3545 if ((ifp->if_flags & IFF_UP) != 0) { 3546 if (ifp->if_ioctl) { 3547 ifp->if_flags &= ~IFF_UP; 3548 ifr.ifr_flags = ifp->if_flags & 0xffff; 3549 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3550 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3551 ifp->if_flags |= IFF_UP; 3552 ifr.ifr_flags = ifp->if_flags & 0xffff; 3553 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3554 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3555 } 3556 } 3557 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3558 return (0); 3559 } 3560 3561 /* 3562 * Compat function for handling basic encapsulation requests. 3563 * Not converted stacks (FDDI, IB, ..) supports traditional 3564 * output model: ARP (and other similar L2 protocols) are handled 3565 * inside output routine, arpresolve/nd6_resolve() returns MAC 3566 * address instead of full prepend. 3567 * 3568 * This function creates calculated header==MAC for IPv4/IPv6 and 3569 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3570 * address families. 3571 */ 3572 static int 3573 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3574 { 3575 3576 if (req->rtype != IFENCAP_LL) 3577 return (EOPNOTSUPP); 3578 3579 if (req->bufsize < req->lladdr_len) 3580 return (ENOMEM); 3581 3582 switch (req->family) { 3583 case AF_INET: 3584 case AF_INET6: 3585 break; 3586 default: 3587 return (EAFNOSUPPORT); 3588 } 3589 3590 /* Copy lladdr to storage as is */ 3591 memmove(req->buf, req->lladdr, req->lladdr_len); 3592 req->bufsize = req->lladdr_len; 3593 req->lladdr_off = 0; 3594 3595 return (0); 3596 } 3597 3598 /* 3599 * The name argument must be a pointer to storage which will last as 3600 * long as the interface does. For physical devices, the result of 3601 * device_get_name(dev) is a good choice and for pseudo-devices a 3602 * static string works well. 3603 */ 3604 void 3605 if_initname(struct ifnet *ifp, const char *name, int unit) 3606 { 3607 ifp->if_dname = name; 3608 ifp->if_dunit = unit; 3609 if (unit != IF_DUNIT_NONE) 3610 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3611 else 3612 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3613 } 3614 3615 int 3616 if_printf(struct ifnet *ifp, const char * fmt, ...) 3617 { 3618 va_list ap; 3619 int retval; 3620 3621 retval = printf("%s: ", ifp->if_xname); 3622 va_start(ap, fmt); 3623 retval += vprintf(fmt, ap); 3624 va_end(ap); 3625 return (retval); 3626 } 3627 3628 void 3629 if_start(struct ifnet *ifp) 3630 { 3631 3632 (*(ifp)->if_start)(ifp); 3633 } 3634 3635 /* 3636 * Backwards compatibility interface for drivers 3637 * that have not implemented it 3638 */ 3639 static int 3640 if_transmit(struct ifnet *ifp, struct mbuf *m) 3641 { 3642 int error; 3643 3644 IFQ_HANDOFF(ifp, m, error); 3645 return (error); 3646 } 3647 3648 static void 3649 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3650 { 3651 3652 m_freem(m); 3653 } 3654 3655 int 3656 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3657 { 3658 int active = 0; 3659 3660 IF_LOCK(ifq); 3661 if (_IF_QFULL(ifq)) { 3662 IF_UNLOCK(ifq); 3663 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 3664 m_freem(m); 3665 return (0); 3666 } 3667 if (ifp != NULL) { 3668 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 3669 if (m->m_flags & (M_BCAST|M_MCAST)) 3670 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 3671 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3672 } 3673 _IF_ENQUEUE(ifq, m); 3674 IF_UNLOCK(ifq); 3675 if (ifp != NULL && !active) 3676 (*(ifp)->if_start)(ifp); 3677 return (1); 3678 } 3679 3680 void 3681 if_register_com_alloc(u_char type, 3682 if_com_alloc_t *a, if_com_free_t *f) 3683 { 3684 3685 KASSERT(if_com_alloc[type] == NULL, 3686 ("if_register_com_alloc: %d already registered", type)); 3687 KASSERT(if_com_free[type] == NULL, 3688 ("if_register_com_alloc: %d free already registered", type)); 3689 3690 if_com_alloc[type] = a; 3691 if_com_free[type] = f; 3692 } 3693 3694 void 3695 if_deregister_com_alloc(u_char type) 3696 { 3697 3698 KASSERT(if_com_alloc[type] != NULL, 3699 ("if_deregister_com_alloc: %d not registered", type)); 3700 KASSERT(if_com_free[type] != NULL, 3701 ("if_deregister_com_alloc: %d free not registered", type)); 3702 if_com_alloc[type] = NULL; 3703 if_com_free[type] = NULL; 3704 } 3705 3706 /* API for driver access to network stack owned ifnet.*/ 3707 uint64_t 3708 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 3709 { 3710 uint64_t oldbrate; 3711 3712 oldbrate = ifp->if_baudrate; 3713 ifp->if_baudrate = baudrate; 3714 return (oldbrate); 3715 } 3716 3717 uint64_t 3718 if_getbaudrate(if_t ifp) 3719 { 3720 3721 return (((struct ifnet *)ifp)->if_baudrate); 3722 } 3723 3724 int 3725 if_setcapabilities(if_t ifp, int capabilities) 3726 { 3727 ((struct ifnet *)ifp)->if_capabilities = capabilities; 3728 return (0); 3729 } 3730 3731 int 3732 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 3733 { 3734 ((struct ifnet *)ifp)->if_capabilities |= setbit; 3735 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 3736 3737 return (0); 3738 } 3739 3740 int 3741 if_getcapabilities(if_t ifp) 3742 { 3743 return ((struct ifnet *)ifp)->if_capabilities; 3744 } 3745 3746 int 3747 if_setcapenable(if_t ifp, int capabilities) 3748 { 3749 ((struct ifnet *)ifp)->if_capenable = capabilities; 3750 return (0); 3751 } 3752 3753 int 3754 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 3755 { 3756 if(setcap) 3757 ((struct ifnet *)ifp)->if_capenable |= setcap; 3758 if(clearcap) 3759 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 3760 3761 return (0); 3762 } 3763 3764 const char * 3765 if_getdname(if_t ifp) 3766 { 3767 return ((struct ifnet *)ifp)->if_dname; 3768 } 3769 3770 int 3771 if_togglecapenable(if_t ifp, int togglecap) 3772 { 3773 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 3774 return (0); 3775 } 3776 3777 int 3778 if_getcapenable(if_t ifp) 3779 { 3780 return ((struct ifnet *)ifp)->if_capenable; 3781 } 3782 3783 /* 3784 * This is largely undesirable because it ties ifnet to a device, but does 3785 * provide flexiblity for an embedded product vendor. Should be used with 3786 * the understanding that it violates the interface boundaries, and should be 3787 * a last resort only. 3788 */ 3789 int 3790 if_setdev(if_t ifp, void *dev) 3791 { 3792 return (0); 3793 } 3794 3795 int 3796 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 3797 { 3798 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 3799 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 3800 3801 return (0); 3802 } 3803 3804 int 3805 if_getdrvflags(if_t ifp) 3806 { 3807 return ((struct ifnet *)ifp)->if_drv_flags; 3808 } 3809 3810 int 3811 if_setdrvflags(if_t ifp, int flags) 3812 { 3813 ((struct ifnet *)ifp)->if_drv_flags = flags; 3814 return (0); 3815 } 3816 3817 3818 int 3819 if_setflags(if_t ifp, int flags) 3820 { 3821 ((struct ifnet *)ifp)->if_flags = flags; 3822 return (0); 3823 } 3824 3825 int 3826 if_setflagbits(if_t ifp, int set, int clear) 3827 { 3828 ((struct ifnet *)ifp)->if_flags |= set; 3829 ((struct ifnet *)ifp)->if_flags &= ~clear; 3830 3831 return (0); 3832 } 3833 3834 int 3835 if_getflags(if_t ifp) 3836 { 3837 return ((struct ifnet *)ifp)->if_flags; 3838 } 3839 3840 int 3841 if_clearhwassist(if_t ifp) 3842 { 3843 ((struct ifnet *)ifp)->if_hwassist = 0; 3844 return (0); 3845 } 3846 3847 int 3848 if_sethwassistbits(if_t ifp, int toset, int toclear) 3849 { 3850 ((struct ifnet *)ifp)->if_hwassist |= toset; 3851 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 3852 3853 return (0); 3854 } 3855 3856 int 3857 if_sethwassist(if_t ifp, int hwassist_bit) 3858 { 3859 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 3860 return (0); 3861 } 3862 3863 int 3864 if_gethwassist(if_t ifp) 3865 { 3866 return ((struct ifnet *)ifp)->if_hwassist; 3867 } 3868 3869 int 3870 if_setmtu(if_t ifp, int mtu) 3871 { 3872 ((struct ifnet *)ifp)->if_mtu = mtu; 3873 return (0); 3874 } 3875 3876 int 3877 if_getmtu(if_t ifp) 3878 { 3879 return ((struct ifnet *)ifp)->if_mtu; 3880 } 3881 3882 int 3883 if_getmtu_family(if_t ifp, int family) 3884 { 3885 struct domain *dp; 3886 3887 for (dp = domains; dp; dp = dp->dom_next) { 3888 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 3889 return (dp->dom_ifmtu((struct ifnet *)ifp)); 3890 } 3891 3892 return (((struct ifnet *)ifp)->if_mtu); 3893 } 3894 3895 int 3896 if_setsoftc(if_t ifp, void *softc) 3897 { 3898 ((struct ifnet *)ifp)->if_softc = softc; 3899 return (0); 3900 } 3901 3902 void * 3903 if_getsoftc(if_t ifp) 3904 { 3905 return ((struct ifnet *)ifp)->if_softc; 3906 } 3907 3908 void 3909 if_setrcvif(struct mbuf *m, if_t ifp) 3910 { 3911 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 3912 } 3913 3914 void 3915 if_setvtag(struct mbuf *m, uint16_t tag) 3916 { 3917 m->m_pkthdr.ether_vtag = tag; 3918 } 3919 3920 uint16_t 3921 if_getvtag(struct mbuf *m) 3922 { 3923 3924 return (m->m_pkthdr.ether_vtag); 3925 } 3926 3927 int 3928 if_sendq_empty(if_t ifp) 3929 { 3930 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 3931 } 3932 3933 struct ifaddr * 3934 if_getifaddr(if_t ifp) 3935 { 3936 return ((struct ifnet *)ifp)->if_addr; 3937 } 3938 3939 int 3940 if_getamcount(if_t ifp) 3941 { 3942 return ((struct ifnet *)ifp)->if_amcount; 3943 } 3944 3945 3946 int 3947 if_setsendqready(if_t ifp) 3948 { 3949 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 3950 return (0); 3951 } 3952 3953 int 3954 if_setsendqlen(if_t ifp, int tx_desc_count) 3955 { 3956 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 3957 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 3958 3959 return (0); 3960 } 3961 3962 int 3963 if_vlantrunkinuse(if_t ifp) 3964 { 3965 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 3966 } 3967 3968 int 3969 if_input(if_t ifp, struct mbuf* sendmp) 3970 { 3971 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 3972 return (0); 3973 3974 } 3975 3976 /* XXX */ 3977 #ifndef ETH_ADDR_LEN 3978 #define ETH_ADDR_LEN 6 3979 #endif 3980 3981 int 3982 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max) 3983 { 3984 struct ifmultiaddr *ifma; 3985 uint8_t *lmta = (uint8_t *)mta; 3986 int mcnt = 0; 3987 3988 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 3989 if (ifma->ifma_addr->sa_family != AF_LINK) 3990 continue; 3991 3992 if (mcnt == max) 3993 break; 3994 3995 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 3996 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 3997 mcnt++; 3998 } 3999 *cnt = mcnt; 4000 4001 return (0); 4002 } 4003 4004 int 4005 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max) 4006 { 4007 int error; 4008 4009 if_maddr_rlock(ifp); 4010 error = if_setupmultiaddr(ifp, mta, cnt, max); 4011 if_maddr_runlock(ifp); 4012 return (error); 4013 } 4014 4015 int 4016 if_multiaddr_count(if_t ifp, int max) 4017 { 4018 struct ifmultiaddr *ifma; 4019 int count; 4020 4021 count = 0; 4022 if_maddr_rlock(ifp); 4023 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4024 if (ifma->ifma_addr->sa_family != AF_LINK) 4025 continue; 4026 count++; 4027 if (count == max) 4028 break; 4029 } 4030 if_maddr_runlock(ifp); 4031 return (count); 4032 } 4033 4034 int 4035 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg) 4036 { 4037 struct ifmultiaddr *ifma; 4038 int cnt = 0; 4039 4040 if_maddr_rlock(ifp); 4041 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4042 cnt += filter(arg, ifma, cnt); 4043 if_maddr_runlock(ifp); 4044 return (cnt); 4045 } 4046 4047 struct mbuf * 4048 if_dequeue(if_t ifp) 4049 { 4050 struct mbuf *m; 4051 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4052 4053 return (m); 4054 } 4055 4056 int 4057 if_sendq_prepend(if_t ifp, struct mbuf *m) 4058 { 4059 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4060 return (0); 4061 } 4062 4063 int 4064 if_setifheaderlen(if_t ifp, int len) 4065 { 4066 ((struct ifnet *)ifp)->if_hdrlen = len; 4067 return (0); 4068 } 4069 4070 caddr_t 4071 if_getlladdr(if_t ifp) 4072 { 4073 return (IF_LLADDR((struct ifnet *)ifp)); 4074 } 4075 4076 void * 4077 if_gethandle(u_char type) 4078 { 4079 return (if_alloc(type)); 4080 } 4081 4082 void 4083 if_bpfmtap(if_t ifh, struct mbuf *m) 4084 { 4085 struct ifnet *ifp = (struct ifnet *)ifh; 4086 4087 BPF_MTAP(ifp, m); 4088 } 4089 4090 void 4091 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4092 { 4093 struct ifnet *ifp = (struct ifnet *)ifh; 4094 4095 ETHER_BPF_MTAP(ifp, m); 4096 } 4097 4098 void 4099 if_vlancap(if_t ifh) 4100 { 4101 struct ifnet *ifp = (struct ifnet *)ifh; 4102 VLAN_CAPABILITIES(ifp); 4103 } 4104 4105 void 4106 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4107 { 4108 ((struct ifnet *)ifp)->if_init = init_fn; 4109 } 4110 4111 void 4112 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4113 { 4114 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4115 } 4116 4117 void 4118 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4119 { 4120 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4121 } 4122 4123 void 4124 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4125 { 4126 ((struct ifnet *)ifp)->if_transmit = start_fn; 4127 } 4128 4129 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4130 { 4131 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4132 4133 } 4134 4135 void 4136 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4137 { 4138 4139 ifp->if_get_counter = fn; 4140 } 4141 4142 /* Revisit these - These are inline functions originally. */ 4143 int 4144 drbr_inuse_drv(if_t ifh, struct buf_ring *br) 4145 { 4146 return drbr_inuse(ifh, br); 4147 } 4148 4149 struct mbuf* 4150 drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 4151 { 4152 return drbr_dequeue(ifh, br); 4153 } 4154 4155 int 4156 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 4157 { 4158 return drbr_needs_enqueue(ifh, br); 4159 } 4160 4161 int 4162 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 4163 { 4164 return drbr_enqueue(ifh, br, m); 4165 4166 } 4167