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 /* Interface description */ 116 static unsigned int ifdescr_maxlen = 1024; 117 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 118 &ifdescr_maxlen, 0, 119 "administrative maximum length for interface description"); 120 121 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 122 123 /* global sx for non-critical path ifdescr */ 124 static struct sx ifdescr_sx; 125 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 126 127 void (*bridge_linkstate_p)(struct ifnet *ifp); 128 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 129 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 130 /* These are external hooks for CARP. */ 131 void (*carp_linkstate_p)(struct ifnet *ifp); 132 void (*carp_demote_adj_p)(int, char *); 133 int (*carp_master_p)(struct ifaddr *); 134 #if defined(INET) || defined(INET6) 135 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 136 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 137 const struct sockaddr *sa); 138 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 139 int (*carp_attach_p)(struct ifaddr *, int); 140 void (*carp_detach_p)(struct ifaddr *); 141 #endif 142 #ifdef INET 143 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 144 #endif 145 #ifdef INET6 146 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 147 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 148 const struct in6_addr *taddr); 149 #endif 150 151 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 152 153 /* 154 * XXX: Style; these should be sorted alphabetically, and unprototyped 155 * static functions should be prototyped. Currently they are sorted by 156 * declaration order. 157 */ 158 static void if_attachdomain(void *); 159 static void if_attachdomain1(struct ifnet *); 160 static int ifconf(u_long, caddr_t); 161 static void if_freemulti(struct ifmultiaddr *); 162 static void if_grow(void); 163 static void if_input_default(struct ifnet *, struct mbuf *); 164 static void if_route(struct ifnet *, int flag, int fam); 165 static int if_setflag(struct ifnet *, int, int, int *, int); 166 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 167 static void if_unroute(struct ifnet *, int flag, int fam); 168 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 169 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 170 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 171 static void do_link_state_change(void *, int); 172 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 173 static int if_getgroupmembers(struct ifgroupreq *); 174 static void if_delgroups(struct ifnet *); 175 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 176 static void if_detach_internal(struct ifnet *, int, struct if_clone **); 177 178 #ifdef INET6 179 /* 180 * XXX: declare here to avoid to include many inet6 related files.. 181 * should be more generalized? 182 */ 183 extern void nd6_setmtu(struct ifnet *); 184 #endif 185 186 VNET_DEFINE(int, if_index); 187 int ifqmaxlen = IFQ_MAXLEN; 188 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 189 VNET_DEFINE(struct ifgrouphead, ifg_head); 190 191 static VNET_DEFINE(int, if_indexlim) = 8; 192 193 /* Table of ifnet by index. */ 194 VNET_DEFINE(struct ifnet **, ifindex_table); 195 196 #define V_if_indexlim VNET(if_indexlim) 197 #define V_ifindex_table VNET(ifindex_table) 198 199 /* 200 * The global network interface list (V_ifnet) and related state (such as 201 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 202 * an rwlock. Either may be acquired shared to stablize the list, but both 203 * must be acquired writable to modify the list. This model allows us to 204 * both stablize the interface list during interrupt thread processing, but 205 * also to stablize it over long-running ioctls, without introducing priority 206 * inversions and deadlocks. 207 */ 208 struct rwlock ifnet_rwlock; 209 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE); 210 struct sx ifnet_sxlock; 211 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 212 213 /* 214 * The allocation of network interfaces is a rather non-atomic affair; we 215 * need to select an index before we are ready to expose the interface for 216 * use, so will use this pointer value to indicate reservation. 217 */ 218 #define IFNET_HOLD (void *)(uintptr_t)(-1) 219 220 static if_com_alloc_t *if_com_alloc[256]; 221 static if_com_free_t *if_com_free[256]; 222 223 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 224 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 225 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 226 227 struct ifnet * 228 ifnet_byindex_locked(u_short idx) 229 { 230 231 if (idx > V_if_index) 232 return (NULL); 233 if (V_ifindex_table[idx] == IFNET_HOLD) 234 return (NULL); 235 return (V_ifindex_table[idx]); 236 } 237 238 struct ifnet * 239 ifnet_byindex(u_short idx) 240 { 241 struct ifnet *ifp; 242 243 IFNET_RLOCK_NOSLEEP(); 244 ifp = ifnet_byindex_locked(idx); 245 IFNET_RUNLOCK_NOSLEEP(); 246 return (ifp); 247 } 248 249 struct ifnet * 250 ifnet_byindex_ref(u_short idx) 251 { 252 struct ifnet *ifp; 253 254 IFNET_RLOCK_NOSLEEP(); 255 ifp = ifnet_byindex_locked(idx); 256 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 257 IFNET_RUNLOCK_NOSLEEP(); 258 return (NULL); 259 } 260 if_ref(ifp); 261 IFNET_RUNLOCK_NOSLEEP(); 262 return (ifp); 263 } 264 265 /* 266 * Allocate an ifindex array entry; return 0 on success or an error on 267 * failure. 268 */ 269 static u_short 270 ifindex_alloc(void) 271 { 272 u_short idx; 273 274 IFNET_WLOCK_ASSERT(); 275 retry: 276 /* 277 * Try to find an empty slot below V_if_index. If we fail, take the 278 * next slot. 279 */ 280 for (idx = 1; idx <= V_if_index; idx++) { 281 if (V_ifindex_table[idx] == NULL) 282 break; 283 } 284 285 /* Catch if_index overflow. */ 286 if (idx >= V_if_indexlim) { 287 if_grow(); 288 goto retry; 289 } 290 if (idx > V_if_index) 291 V_if_index = idx; 292 return (idx); 293 } 294 295 static void 296 ifindex_free_locked(u_short idx) 297 { 298 299 IFNET_WLOCK_ASSERT(); 300 301 V_ifindex_table[idx] = NULL; 302 while (V_if_index > 0 && 303 V_ifindex_table[V_if_index] == NULL) 304 V_if_index--; 305 } 306 307 static void 308 ifindex_free(u_short idx) 309 { 310 311 IFNET_WLOCK(); 312 ifindex_free_locked(idx); 313 IFNET_WUNLOCK(); 314 } 315 316 static void 317 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) 318 { 319 320 IFNET_WLOCK_ASSERT(); 321 322 V_ifindex_table[idx] = ifp; 323 } 324 325 static void 326 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 327 { 328 329 IFNET_WLOCK(); 330 ifnet_setbyindex_locked(idx, ifp); 331 IFNET_WUNLOCK(); 332 } 333 334 struct ifaddr * 335 ifaddr_byindex(u_short idx) 336 { 337 struct ifnet *ifp; 338 struct ifaddr *ifa = NULL; 339 340 IFNET_RLOCK_NOSLEEP(); 341 ifp = ifnet_byindex_locked(idx); 342 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 343 ifa_ref(ifa); 344 IFNET_RUNLOCK_NOSLEEP(); 345 return (ifa); 346 } 347 348 /* 349 * Network interface utility routines. 350 * 351 * Routines with ifa_ifwith* names take sockaddr *'s as 352 * parameters. 353 */ 354 355 static void 356 vnet_if_init(const void *unused __unused) 357 { 358 359 TAILQ_INIT(&V_ifnet); 360 TAILQ_INIT(&V_ifg_head); 361 IFNET_WLOCK(); 362 if_grow(); /* create initial table */ 363 IFNET_WUNLOCK(); 364 vnet_if_clone_init(); 365 } 366 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 367 NULL); 368 369 #ifdef VIMAGE 370 static void 371 vnet_if_uninit(const void *unused __unused) 372 { 373 374 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 375 "not empty", __func__, __LINE__, &V_ifnet)); 376 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 377 "not empty", __func__, __LINE__, &V_ifg_head)); 378 379 free((caddr_t)V_ifindex_table, M_IFNET); 380 } 381 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 382 vnet_if_uninit, NULL); 383 #endif 384 385 static void 386 if_grow(void) 387 { 388 int oldlim; 389 u_int n; 390 struct ifnet **e; 391 392 IFNET_WLOCK_ASSERT(); 393 oldlim = V_if_indexlim; 394 IFNET_WUNLOCK(); 395 n = (oldlim << 1) * sizeof(*e); 396 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 397 IFNET_WLOCK(); 398 if (V_if_indexlim != oldlim) { 399 free(e, M_IFNET); 400 return; 401 } 402 if (V_ifindex_table != NULL) { 403 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 404 free((caddr_t)V_ifindex_table, M_IFNET); 405 } 406 V_if_indexlim <<= 1; 407 V_ifindex_table = e; 408 } 409 410 /* 411 * Allocate a struct ifnet and an index for an interface. A layer 2 412 * common structure will also be allocated if an allocation routine is 413 * registered for the passed type. 414 */ 415 struct ifnet * 416 if_alloc(u_char type) 417 { 418 struct ifnet *ifp; 419 u_short idx; 420 421 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 422 IFNET_WLOCK(); 423 idx = ifindex_alloc(); 424 ifnet_setbyindex_locked(idx, IFNET_HOLD); 425 IFNET_WUNLOCK(); 426 ifp->if_index = idx; 427 ifp->if_type = type; 428 ifp->if_alloctype = type; 429 if (if_com_alloc[type] != NULL) { 430 ifp->if_l2com = if_com_alloc[type](type, ifp); 431 if (ifp->if_l2com == NULL) { 432 free(ifp, M_IFNET); 433 ifindex_free(idx); 434 return (NULL); 435 } 436 } 437 438 IF_ADDR_LOCK_INIT(ifp); 439 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 440 ifp->if_afdata_initialized = 0; 441 IF_AFDATA_LOCK_INIT(ifp); 442 TAILQ_INIT(&ifp->if_addrhead); 443 TAILQ_INIT(&ifp->if_multiaddrs); 444 TAILQ_INIT(&ifp->if_groups); 445 #ifdef MAC 446 mac_ifnet_init(ifp); 447 #endif 448 ifq_init(&ifp->if_snd, ifp); 449 450 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 451 for (int i = 0; i < IFCOUNTERS; i++) 452 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 453 ifp->if_get_counter = if_get_counter_default; 454 ifnet_setbyindex(ifp->if_index, ifp); 455 return (ifp); 456 } 457 458 /* 459 * Do the actual work of freeing a struct ifnet, and layer 2 common 460 * structure. This call is made when the last reference to an 461 * interface is released. 462 */ 463 static void 464 if_free_internal(struct ifnet *ifp) 465 { 466 467 KASSERT((ifp->if_flags & IFF_DYING), 468 ("if_free_internal: interface not dying")); 469 470 if (if_com_free[ifp->if_alloctype] != NULL) 471 if_com_free[ifp->if_alloctype](ifp->if_l2com, 472 ifp->if_alloctype); 473 474 #ifdef MAC 475 mac_ifnet_destroy(ifp); 476 #endif /* MAC */ 477 if (ifp->if_description != NULL) 478 free(ifp->if_description, M_IFDESCR); 479 IF_AFDATA_DESTROY(ifp); 480 IF_ADDR_LOCK_DESTROY(ifp); 481 ifq_delete(&ifp->if_snd); 482 483 for (int i = 0; i < IFCOUNTERS; i++) 484 counter_u64_free(ifp->if_counters[i]); 485 486 free(ifp, M_IFNET); 487 } 488 489 /* 490 * Deregister an interface and free the associated storage. 491 */ 492 void 493 if_free(struct ifnet *ifp) 494 { 495 496 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 497 498 CURVNET_SET_QUIET(ifp->if_vnet); 499 IFNET_WLOCK(); 500 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 501 ("%s: freeing unallocated ifnet", ifp->if_xname)); 502 503 ifindex_free_locked(ifp->if_index); 504 IFNET_WUNLOCK(); 505 506 if (refcount_release(&ifp->if_refcount)) 507 if_free_internal(ifp); 508 CURVNET_RESTORE(); 509 } 510 511 /* 512 * Interfaces to keep an ifnet type-stable despite the possibility of the 513 * driver calling if_free(). If there are additional references, we defer 514 * freeing the underlying data structure. 515 */ 516 void 517 if_ref(struct ifnet *ifp) 518 { 519 520 /* We don't assert the ifnet list lock here, but arguably should. */ 521 refcount_acquire(&ifp->if_refcount); 522 } 523 524 void 525 if_rele(struct ifnet *ifp) 526 { 527 528 if (!refcount_release(&ifp->if_refcount)) 529 return; 530 if_free_internal(ifp); 531 } 532 533 void 534 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 535 { 536 537 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 538 539 if (ifq->ifq_maxlen == 0) 540 ifq->ifq_maxlen = ifqmaxlen; 541 542 ifq->altq_type = 0; 543 ifq->altq_disc = NULL; 544 ifq->altq_flags &= ALTQF_CANTCHANGE; 545 ifq->altq_tbr = NULL; 546 ifq->altq_ifp = ifp; 547 } 548 549 void 550 ifq_delete(struct ifaltq *ifq) 551 { 552 mtx_destroy(&ifq->ifq_mtx); 553 } 554 555 /* 556 * Perform generic interface initalization tasks and attach the interface 557 * to the list of "active" interfaces. If vmove flag is set on entry 558 * to if_attach_internal(), perform only a limited subset of initialization 559 * tasks, given that we are moving from one vnet to another an ifnet which 560 * has already been fully initialized. 561 * 562 * Note that if_detach_internal() removes group membership unconditionally 563 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 564 * Thus, when if_vmove() is applied to a cloned interface, group membership 565 * is lost while a cloned one always joins a group whose name is 566 * ifc->ifc_name. To recover this after if_detach_internal() and 567 * if_attach_internal(), the cloner should be specified to 568 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 569 * attempts to join a group whose name is ifc->ifc_name. 570 * 571 * XXX: 572 * - The decision to return void and thus require this function to 573 * succeed is questionable. 574 * - We should probably do more sanity checking. For instance we don't 575 * do anything to insure if_xname is unique or non-empty. 576 */ 577 void 578 if_attach(struct ifnet *ifp) 579 { 580 581 if_attach_internal(ifp, 0, NULL); 582 } 583 584 /* 585 * Compute the least common TSO limit. 586 */ 587 void 588 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 589 { 590 /* 591 * 1) If there is no limit currently, take the limit from 592 * the network adapter. 593 * 594 * 2) If the network adapter has a limit below the current 595 * limit, apply it. 596 */ 597 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 598 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 599 pmax->tsomaxbytes = ifp->if_hw_tsomax; 600 } 601 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 602 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 603 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 604 } 605 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 606 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 607 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 608 } 609 } 610 611 /* 612 * Update TSO limit of a network adapter. 613 * 614 * Returns zero if no change. Else non-zero. 615 */ 616 int 617 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 618 { 619 int retval = 0; 620 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 621 ifp->if_hw_tsomax = pmax->tsomaxbytes; 622 retval++; 623 } 624 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 625 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 626 retval++; 627 } 628 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 629 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 630 retval++; 631 } 632 return (retval); 633 } 634 635 static void 636 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 637 { 638 unsigned socksize, ifasize; 639 int namelen, masklen; 640 struct sockaddr_dl *sdl; 641 struct ifaddr *ifa; 642 643 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 644 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 645 ifp->if_xname); 646 647 #ifdef VIMAGE 648 ifp->if_vnet = curvnet; 649 if (ifp->if_home_vnet == NULL) 650 ifp->if_home_vnet = curvnet; 651 #endif 652 653 if_addgroup(ifp, IFG_ALL); 654 655 /* Restore group membership for cloned interfaces. */ 656 if (vmove && ifc != NULL) 657 if_clone_addgroup(ifp, ifc); 658 659 getmicrotime(&ifp->if_lastchange); 660 ifp->if_epoch = time_uptime; 661 662 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 663 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 664 ("transmit and qflush must both either be set or both be NULL")); 665 if (ifp->if_transmit == NULL) { 666 ifp->if_transmit = if_transmit; 667 ifp->if_qflush = if_qflush; 668 } 669 if (ifp->if_input == NULL) 670 ifp->if_input = if_input_default; 671 672 if (!vmove) { 673 #ifdef MAC 674 mac_ifnet_create(ifp); 675 #endif 676 677 /* 678 * Create a Link Level name for this device. 679 */ 680 namelen = strlen(ifp->if_xname); 681 /* 682 * Always save enough space for any possiable name so we 683 * can do a rename in place later. 684 */ 685 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 686 socksize = masklen + ifp->if_addrlen; 687 if (socksize < sizeof(*sdl)) 688 socksize = sizeof(*sdl); 689 socksize = roundup2(socksize, sizeof(long)); 690 ifasize = sizeof(*ifa) + 2 * socksize; 691 ifa = ifa_alloc(ifasize, M_WAITOK); 692 sdl = (struct sockaddr_dl *)(ifa + 1); 693 sdl->sdl_len = socksize; 694 sdl->sdl_family = AF_LINK; 695 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 696 sdl->sdl_nlen = namelen; 697 sdl->sdl_index = ifp->if_index; 698 sdl->sdl_type = ifp->if_type; 699 ifp->if_addr = ifa; 700 ifa->ifa_ifp = ifp; 701 ifa->ifa_rtrequest = link_rtrequest; 702 ifa->ifa_addr = (struct sockaddr *)sdl; 703 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 704 ifa->ifa_netmask = (struct sockaddr *)sdl; 705 sdl->sdl_len = masklen; 706 while (namelen != 0) 707 sdl->sdl_data[--namelen] = 0xff; 708 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 709 /* Reliably crash if used uninitialized. */ 710 ifp->if_broadcastaddr = NULL; 711 712 #if defined(INET) || defined(INET6) 713 /* Use defaults for TSO, if nothing is set */ 714 if (ifp->if_hw_tsomax == 0 && 715 ifp->if_hw_tsomaxsegcount == 0 && 716 ifp->if_hw_tsomaxsegsize == 0) { 717 /* 718 * The TSO defaults needs to be such that an 719 * NFS mbuf list of 35 mbufs totalling just 720 * below 64K works and that a chain of mbufs 721 * can be defragged into at most 32 segments: 722 */ 723 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 724 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 725 ifp->if_hw_tsomaxsegcount = 35; 726 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 727 728 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 729 if (ifp->if_capabilities & IFCAP_TSO) { 730 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 731 ifp->if_hw_tsomax, 732 ifp->if_hw_tsomaxsegcount, 733 ifp->if_hw_tsomaxsegsize); 734 } 735 } 736 #endif 737 } 738 #ifdef VIMAGE 739 else { 740 /* 741 * Update the interface index in the link layer address 742 * of the interface. 743 */ 744 for (ifa = ifp->if_addr; ifa != NULL; 745 ifa = TAILQ_NEXT(ifa, ifa_link)) { 746 if (ifa->ifa_addr->sa_family == AF_LINK) { 747 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 748 sdl->sdl_index = ifp->if_index; 749 } 750 } 751 } 752 #endif 753 754 IFNET_WLOCK(); 755 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 756 #ifdef VIMAGE 757 curvnet->vnet_ifcnt++; 758 #endif 759 IFNET_WUNLOCK(); 760 761 if (domain_init_status >= 2) 762 if_attachdomain1(ifp); 763 764 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 765 if (IS_DEFAULT_VNET(curvnet)) 766 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 767 768 /* Announce the interface. */ 769 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 770 } 771 772 static void 773 if_attachdomain(void *dummy) 774 { 775 struct ifnet *ifp; 776 777 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 778 if_attachdomain1(ifp); 779 } 780 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 781 if_attachdomain, NULL); 782 783 static void 784 if_attachdomain1(struct ifnet *ifp) 785 { 786 struct domain *dp; 787 788 /* 789 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 790 * cannot lock ifp->if_afdata initialization, entirely. 791 */ 792 if (IF_AFDATA_TRYLOCK(ifp) == 0) 793 return; 794 if (ifp->if_afdata_initialized >= domain_init_status) { 795 IF_AFDATA_UNLOCK(ifp); 796 log(LOG_WARNING, "%s called more than once on %s\n", 797 __func__, ifp->if_xname); 798 return; 799 } 800 ifp->if_afdata_initialized = domain_init_status; 801 IF_AFDATA_UNLOCK(ifp); 802 803 /* address family dependent data region */ 804 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 805 for (dp = domains; dp; dp = dp->dom_next) { 806 if (dp->dom_ifattach) 807 ifp->if_afdata[dp->dom_family] = 808 (*dp->dom_ifattach)(ifp); 809 } 810 } 811 812 /* 813 * Remove any unicast or broadcast network addresses from an interface. 814 */ 815 void 816 if_purgeaddrs(struct ifnet *ifp) 817 { 818 struct ifaddr *ifa, *next; 819 820 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 821 if (ifa->ifa_addr->sa_family == AF_LINK) 822 continue; 823 #ifdef INET 824 /* XXX: Ugly!! ad hoc just for INET */ 825 if (ifa->ifa_addr->sa_family == AF_INET) { 826 struct ifaliasreq ifr; 827 828 bzero(&ifr, sizeof(ifr)); 829 ifr.ifra_addr = *ifa->ifa_addr; 830 if (ifa->ifa_dstaddr) 831 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 832 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 833 NULL) == 0) 834 continue; 835 } 836 #endif /* INET */ 837 #ifdef INET6 838 if (ifa->ifa_addr->sa_family == AF_INET6) { 839 in6_purgeaddr(ifa); 840 /* ifp_addrhead is already updated */ 841 continue; 842 } 843 #endif /* INET6 */ 844 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 845 ifa_free(ifa); 846 } 847 } 848 849 /* 850 * Remove any multicast network addresses from an interface when an ifnet 851 * is going away. 852 */ 853 static void 854 if_purgemaddrs(struct ifnet *ifp) 855 { 856 struct ifmultiaddr *ifma; 857 struct ifmultiaddr *next; 858 859 IF_ADDR_WLOCK(ifp); 860 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 861 if_delmulti_locked(ifp, ifma, 1); 862 IF_ADDR_WUNLOCK(ifp); 863 } 864 865 /* 866 * Detach an interface, removing it from the list of "active" interfaces. 867 * If vmove flag is set on entry to if_detach_internal(), perform only a 868 * limited subset of cleanup tasks, given that we are moving an ifnet from 869 * one vnet to another, where it must be fully operational. 870 * 871 * XXXRW: There are some significant questions about event ordering, and 872 * how to prevent things from starting to use the interface during detach. 873 */ 874 void 875 if_detach(struct ifnet *ifp) 876 { 877 878 CURVNET_SET_QUIET(ifp->if_vnet); 879 if_detach_internal(ifp, 0, NULL); 880 CURVNET_RESTORE(); 881 } 882 883 static void 884 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 885 { 886 struct ifaddr *ifa; 887 int i; 888 struct domain *dp; 889 struct ifnet *iter; 890 int found = 0; 891 892 IFNET_WLOCK(); 893 TAILQ_FOREACH(iter, &V_ifnet, if_link) 894 if (iter == ifp) { 895 TAILQ_REMOVE(&V_ifnet, ifp, if_link); 896 found = 1; 897 break; 898 } 899 #ifdef VIMAGE 900 if (found) 901 curvnet->vnet_ifcnt--; 902 #endif 903 IFNET_WUNLOCK(); 904 if (!found) { 905 if (vmove) 906 panic("%s: ifp=%p not on the ifnet tailq %p", 907 __func__, ifp, &V_ifnet); 908 else 909 return; /* XXX this should panic as well? */ 910 } 911 912 /* Check if this is a cloned interface or not. */ 913 if (vmove && ifcp != NULL) 914 *ifcp = if_clone_findifc(ifp); 915 916 /* 917 * Remove/wait for pending events. 918 */ 919 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 920 921 /* 922 * Remove routes and flush queues. 923 */ 924 if_down(ifp); 925 #ifdef ALTQ 926 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 927 altq_disable(&ifp->if_snd); 928 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 929 altq_detach(&ifp->if_snd); 930 #endif 931 932 if_purgeaddrs(ifp); 933 934 #ifdef INET 935 in_ifdetach(ifp); 936 #endif 937 938 #ifdef INET6 939 /* 940 * Remove all IPv6 kernel structs related to ifp. This should be done 941 * before removing routing entries below, since IPv6 interface direct 942 * routes are expected to be removed by the IPv6-specific kernel API. 943 * Otherwise, the kernel will detect some inconsistency and bark it. 944 */ 945 in6_ifdetach(ifp); 946 #endif 947 if_purgemaddrs(ifp); 948 949 /* Announce that the interface is gone. */ 950 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 951 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 952 if (IS_DEFAULT_VNET(curvnet)) 953 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 954 955 if (!vmove) { 956 /* 957 * Prevent further calls into the device driver via ifnet. 958 */ 959 if_dead(ifp); 960 961 /* 962 * Remove link ifaddr pointer and maybe decrement if_index. 963 * Clean up all addresses. 964 */ 965 ifp->if_addr = NULL; 966 967 /* We can now free link ifaddr. */ 968 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 969 ifa = TAILQ_FIRST(&ifp->if_addrhead); 970 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 971 ifa_free(ifa); 972 } 973 } 974 975 rt_flushifroutes(ifp); 976 if_delgroups(ifp); 977 978 /* 979 * We cannot hold the lock over dom_ifdetach calls as they might 980 * sleep, for example trying to drain a callout, thus open up the 981 * theoretical race with re-attaching. 982 */ 983 IF_AFDATA_LOCK(ifp); 984 i = ifp->if_afdata_initialized; 985 ifp->if_afdata_initialized = 0; 986 IF_AFDATA_UNLOCK(ifp); 987 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 988 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 989 (*dp->dom_ifdetach)(ifp, 990 ifp->if_afdata[dp->dom_family]); 991 } 992 } 993 994 #ifdef VIMAGE 995 /* 996 * if_vmove() performs a limited version of if_detach() in current 997 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 998 * An attempt is made to shrink if_index in current vnet, find an 999 * unused if_index in target vnet and calls if_grow() if necessary, 1000 * and finally find an unused if_xname for the target vnet. 1001 */ 1002 void 1003 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1004 { 1005 struct if_clone *ifc; 1006 1007 /* 1008 * Detach from current vnet, but preserve LLADDR info, do not 1009 * mark as dead etc. so that the ifnet can be reattached later. 1010 */ 1011 if_detach_internal(ifp, 1, &ifc); 1012 1013 /* 1014 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1015 * the if_index for that vnet if possible. 1016 * 1017 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1018 * or we'd lock on one vnet and unlock on another. 1019 */ 1020 IFNET_WLOCK(); 1021 ifindex_free_locked(ifp->if_index); 1022 IFNET_WUNLOCK(); 1023 1024 /* 1025 * Perform interface-specific reassignment tasks, if provided by 1026 * the driver. 1027 */ 1028 if (ifp->if_reassign != NULL) 1029 ifp->if_reassign(ifp, new_vnet, NULL); 1030 1031 /* 1032 * Switch to the context of the target vnet. 1033 */ 1034 CURVNET_SET_QUIET(new_vnet); 1035 1036 IFNET_WLOCK(); 1037 ifp->if_index = ifindex_alloc(); 1038 ifnet_setbyindex_locked(ifp->if_index, ifp); 1039 IFNET_WUNLOCK(); 1040 1041 if_attach_internal(ifp, 1, ifc); 1042 1043 CURVNET_RESTORE(); 1044 } 1045 1046 /* 1047 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1048 */ 1049 static int 1050 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1051 { 1052 struct prison *pr; 1053 struct ifnet *difp; 1054 1055 /* Try to find the prison within our visibility. */ 1056 sx_slock(&allprison_lock); 1057 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1058 sx_sunlock(&allprison_lock); 1059 if (pr == NULL) 1060 return (ENXIO); 1061 prison_hold_locked(pr); 1062 mtx_unlock(&pr->pr_mtx); 1063 1064 /* Do not try to move the iface from and to the same prison. */ 1065 if (pr->pr_vnet == ifp->if_vnet) { 1066 prison_free(pr); 1067 return (EEXIST); 1068 } 1069 1070 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1071 /* XXX Lock interfaces to avoid races. */ 1072 CURVNET_SET_QUIET(pr->pr_vnet); 1073 difp = ifunit(ifname); 1074 CURVNET_RESTORE(); 1075 if (difp != NULL) { 1076 prison_free(pr); 1077 return (EEXIST); 1078 } 1079 1080 /* Move the interface into the child jail/vnet. */ 1081 if_vmove(ifp, pr->pr_vnet); 1082 1083 /* Report the new if_xname back to the userland. */ 1084 sprintf(ifname, "%s", ifp->if_xname); 1085 1086 prison_free(pr); 1087 return (0); 1088 } 1089 1090 static int 1091 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1092 { 1093 struct prison *pr; 1094 struct vnet *vnet_dst; 1095 struct ifnet *ifp; 1096 1097 /* Try to find the prison within our visibility. */ 1098 sx_slock(&allprison_lock); 1099 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1100 sx_sunlock(&allprison_lock); 1101 if (pr == NULL) 1102 return (ENXIO); 1103 prison_hold_locked(pr); 1104 mtx_unlock(&pr->pr_mtx); 1105 1106 /* Make sure the named iface exists in the source prison/vnet. */ 1107 CURVNET_SET(pr->pr_vnet); 1108 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1109 if (ifp == NULL) { 1110 CURVNET_RESTORE(); 1111 prison_free(pr); 1112 return (ENXIO); 1113 } 1114 1115 /* Do not try to move the iface from and to the same prison. */ 1116 vnet_dst = TD_TO_VNET(td); 1117 if (vnet_dst == ifp->if_vnet) { 1118 CURVNET_RESTORE(); 1119 prison_free(pr); 1120 return (EEXIST); 1121 } 1122 1123 /* Get interface back from child jail/vnet. */ 1124 if_vmove(ifp, vnet_dst); 1125 CURVNET_RESTORE(); 1126 1127 /* Report the new if_xname back to the userland. */ 1128 sprintf(ifname, "%s", ifp->if_xname); 1129 1130 prison_free(pr); 1131 return (0); 1132 } 1133 #endif /* VIMAGE */ 1134 1135 /* 1136 * Add a group to an interface 1137 */ 1138 int 1139 if_addgroup(struct ifnet *ifp, const char *groupname) 1140 { 1141 struct ifg_list *ifgl; 1142 struct ifg_group *ifg = NULL; 1143 struct ifg_member *ifgm; 1144 int new = 0; 1145 1146 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1147 groupname[strlen(groupname) - 1] <= '9') 1148 return (EINVAL); 1149 1150 IFNET_WLOCK(); 1151 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1152 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1153 IFNET_WUNLOCK(); 1154 return (EEXIST); 1155 } 1156 1157 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1158 M_NOWAIT)) == NULL) { 1159 IFNET_WUNLOCK(); 1160 return (ENOMEM); 1161 } 1162 1163 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1164 M_TEMP, M_NOWAIT)) == NULL) { 1165 free(ifgl, M_TEMP); 1166 IFNET_WUNLOCK(); 1167 return (ENOMEM); 1168 } 1169 1170 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1171 if (!strcmp(ifg->ifg_group, groupname)) 1172 break; 1173 1174 if (ifg == NULL) { 1175 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1176 M_TEMP, M_NOWAIT)) == NULL) { 1177 free(ifgl, M_TEMP); 1178 free(ifgm, M_TEMP); 1179 IFNET_WUNLOCK(); 1180 return (ENOMEM); 1181 } 1182 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1183 ifg->ifg_refcnt = 0; 1184 TAILQ_INIT(&ifg->ifg_members); 1185 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1186 new = 1; 1187 } 1188 1189 ifg->ifg_refcnt++; 1190 ifgl->ifgl_group = ifg; 1191 ifgm->ifgm_ifp = ifp; 1192 1193 IF_ADDR_WLOCK(ifp); 1194 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1195 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1196 IF_ADDR_WUNLOCK(ifp); 1197 1198 IFNET_WUNLOCK(); 1199 1200 if (new) 1201 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1202 EVENTHANDLER_INVOKE(group_change_event, groupname); 1203 1204 return (0); 1205 } 1206 1207 /* 1208 * Remove a group from an interface 1209 */ 1210 int 1211 if_delgroup(struct ifnet *ifp, const char *groupname) 1212 { 1213 struct ifg_list *ifgl; 1214 struct ifg_member *ifgm; 1215 1216 IFNET_WLOCK(); 1217 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1218 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1219 break; 1220 if (ifgl == NULL) { 1221 IFNET_WUNLOCK(); 1222 return (ENOENT); 1223 } 1224 1225 IF_ADDR_WLOCK(ifp); 1226 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1227 IF_ADDR_WUNLOCK(ifp); 1228 1229 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1230 if (ifgm->ifgm_ifp == ifp) 1231 break; 1232 1233 if (ifgm != NULL) { 1234 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); 1235 free(ifgm, M_TEMP); 1236 } 1237 1238 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1239 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1240 IFNET_WUNLOCK(); 1241 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1242 free(ifgl->ifgl_group, M_TEMP); 1243 } else 1244 IFNET_WUNLOCK(); 1245 1246 free(ifgl, M_TEMP); 1247 1248 EVENTHANDLER_INVOKE(group_change_event, groupname); 1249 1250 return (0); 1251 } 1252 1253 /* 1254 * Remove an interface from all groups 1255 */ 1256 static void 1257 if_delgroups(struct ifnet *ifp) 1258 { 1259 struct ifg_list *ifgl; 1260 struct ifg_member *ifgm; 1261 char groupname[IFNAMSIZ]; 1262 1263 IFNET_WLOCK(); 1264 while (!TAILQ_EMPTY(&ifp->if_groups)) { 1265 ifgl = TAILQ_FIRST(&ifp->if_groups); 1266 1267 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1268 1269 IF_ADDR_WLOCK(ifp); 1270 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); 1271 IF_ADDR_WUNLOCK(ifp); 1272 1273 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1274 if (ifgm->ifgm_ifp == ifp) 1275 break; 1276 1277 if (ifgm != NULL) { 1278 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1279 ifgm_next); 1280 free(ifgm, M_TEMP); 1281 } 1282 1283 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1284 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); 1285 IFNET_WUNLOCK(); 1286 EVENTHANDLER_INVOKE(group_detach_event, 1287 ifgl->ifgl_group); 1288 free(ifgl->ifgl_group, M_TEMP); 1289 } else 1290 IFNET_WUNLOCK(); 1291 1292 free(ifgl, M_TEMP); 1293 1294 EVENTHANDLER_INVOKE(group_change_event, groupname); 1295 1296 IFNET_WLOCK(); 1297 } 1298 IFNET_WUNLOCK(); 1299 } 1300 1301 /* 1302 * Stores all groups from an interface in memory pointed 1303 * to by data 1304 */ 1305 static int 1306 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) 1307 { 1308 int len, error; 1309 struct ifg_list *ifgl; 1310 struct ifg_req ifgrq, *ifgp; 1311 struct ifgroupreq *ifgr = data; 1312 1313 if (ifgr->ifgr_len == 0) { 1314 IF_ADDR_RLOCK(ifp); 1315 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1316 ifgr->ifgr_len += sizeof(struct ifg_req); 1317 IF_ADDR_RUNLOCK(ifp); 1318 return (0); 1319 } 1320 1321 len = ifgr->ifgr_len; 1322 ifgp = ifgr->ifgr_groups; 1323 /* XXX: wire */ 1324 IF_ADDR_RLOCK(ifp); 1325 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1326 if (len < sizeof(ifgrq)) { 1327 IF_ADDR_RUNLOCK(ifp); 1328 return (EINVAL); 1329 } 1330 bzero(&ifgrq, sizeof ifgrq); 1331 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1332 sizeof(ifgrq.ifgrq_group)); 1333 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1334 IF_ADDR_RUNLOCK(ifp); 1335 return (error); 1336 } 1337 len -= sizeof(ifgrq); 1338 ifgp++; 1339 } 1340 IF_ADDR_RUNLOCK(ifp); 1341 1342 return (0); 1343 } 1344 1345 /* 1346 * Stores all members of a group in memory pointed to by data 1347 */ 1348 static int 1349 if_getgroupmembers(struct ifgroupreq *data) 1350 { 1351 struct ifgroupreq *ifgr = data; 1352 struct ifg_group *ifg; 1353 struct ifg_member *ifgm; 1354 struct ifg_req ifgrq, *ifgp; 1355 int len, error; 1356 1357 IFNET_RLOCK(); 1358 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1359 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1360 break; 1361 if (ifg == NULL) { 1362 IFNET_RUNLOCK(); 1363 return (ENOENT); 1364 } 1365 1366 if (ifgr->ifgr_len == 0) { 1367 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1368 ifgr->ifgr_len += sizeof(ifgrq); 1369 IFNET_RUNLOCK(); 1370 return (0); 1371 } 1372 1373 len = ifgr->ifgr_len; 1374 ifgp = ifgr->ifgr_groups; 1375 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1376 if (len < sizeof(ifgrq)) { 1377 IFNET_RUNLOCK(); 1378 return (EINVAL); 1379 } 1380 bzero(&ifgrq, sizeof ifgrq); 1381 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1382 sizeof(ifgrq.ifgrq_member)); 1383 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1384 IFNET_RUNLOCK(); 1385 return (error); 1386 } 1387 len -= sizeof(ifgrq); 1388 ifgp++; 1389 } 1390 IFNET_RUNLOCK(); 1391 1392 return (0); 1393 } 1394 1395 /* 1396 * Return counter values from counter(9)s stored in ifnet. 1397 */ 1398 uint64_t 1399 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1400 { 1401 1402 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1403 1404 return (counter_u64_fetch(ifp->if_counters[cnt])); 1405 } 1406 1407 /* 1408 * Increase an ifnet counter. Usually used for counters shared 1409 * between the stack and a driver, but function supports them all. 1410 */ 1411 void 1412 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1413 { 1414 1415 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1416 1417 counter_u64_add(ifp->if_counters[cnt], inc); 1418 } 1419 1420 /* 1421 * Copy data from ifnet to userland API structure if_data. 1422 */ 1423 void 1424 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1425 { 1426 1427 ifd->ifi_type = ifp->if_type; 1428 ifd->ifi_physical = 0; 1429 ifd->ifi_addrlen = ifp->if_addrlen; 1430 ifd->ifi_hdrlen = ifp->if_hdrlen; 1431 ifd->ifi_link_state = ifp->if_link_state; 1432 ifd->ifi_vhid = 0; 1433 ifd->ifi_datalen = sizeof(struct if_data); 1434 ifd->ifi_mtu = ifp->if_mtu; 1435 ifd->ifi_metric = ifp->if_metric; 1436 ifd->ifi_baudrate = ifp->if_baudrate; 1437 ifd->ifi_hwassist = ifp->if_hwassist; 1438 ifd->ifi_epoch = ifp->if_epoch; 1439 ifd->ifi_lastchange = ifp->if_lastchange; 1440 1441 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1442 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1443 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1444 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1445 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1446 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1447 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1448 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1449 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1450 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1451 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1452 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1453 } 1454 1455 /* 1456 * Wrapper functions for struct ifnet address list locking macros. These are 1457 * used by kernel modules to avoid encoding programming interface or binary 1458 * interface assumptions that may be violated when kernel-internal locking 1459 * approaches change. 1460 */ 1461 void 1462 if_addr_rlock(struct ifnet *ifp) 1463 { 1464 1465 IF_ADDR_RLOCK(ifp); 1466 } 1467 1468 void 1469 if_addr_runlock(struct ifnet *ifp) 1470 { 1471 1472 IF_ADDR_RUNLOCK(ifp); 1473 } 1474 1475 void 1476 if_maddr_rlock(if_t ifp) 1477 { 1478 1479 IF_ADDR_RLOCK((struct ifnet *)ifp); 1480 } 1481 1482 void 1483 if_maddr_runlock(if_t ifp) 1484 { 1485 1486 IF_ADDR_RUNLOCK((struct ifnet *)ifp); 1487 } 1488 1489 /* 1490 * Initialization, destruction and refcounting functions for ifaddrs. 1491 */ 1492 struct ifaddr * 1493 ifa_alloc(size_t size, int flags) 1494 { 1495 struct ifaddr *ifa; 1496 1497 KASSERT(size >= sizeof(struct ifaddr), 1498 ("%s: invalid size %zu", __func__, size)); 1499 1500 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1501 if (ifa == NULL) 1502 return (NULL); 1503 1504 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1505 goto fail; 1506 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1507 goto fail; 1508 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1509 goto fail; 1510 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1511 goto fail; 1512 1513 refcount_init(&ifa->ifa_refcnt, 1); 1514 1515 return (ifa); 1516 1517 fail: 1518 /* free(NULL) is okay */ 1519 counter_u64_free(ifa->ifa_opackets); 1520 counter_u64_free(ifa->ifa_ipackets); 1521 counter_u64_free(ifa->ifa_obytes); 1522 counter_u64_free(ifa->ifa_ibytes); 1523 free(ifa, M_IFADDR); 1524 1525 return (NULL); 1526 } 1527 1528 void 1529 ifa_ref(struct ifaddr *ifa) 1530 { 1531 1532 refcount_acquire(&ifa->ifa_refcnt); 1533 } 1534 1535 void 1536 ifa_free(struct ifaddr *ifa) 1537 { 1538 1539 if (refcount_release(&ifa->ifa_refcnt)) { 1540 counter_u64_free(ifa->ifa_opackets); 1541 counter_u64_free(ifa->ifa_ipackets); 1542 counter_u64_free(ifa->ifa_obytes); 1543 counter_u64_free(ifa->ifa_ibytes); 1544 free(ifa, M_IFADDR); 1545 } 1546 } 1547 1548 int 1549 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1550 { 1551 int error = 0; 1552 struct rtentry *rt = NULL; 1553 struct rt_addrinfo info; 1554 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 1555 1556 bzero(&info, sizeof(info)); 1557 info.rti_ifp = V_loif; 1558 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; 1559 info.rti_info[RTAX_DST] = ia; 1560 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1561 error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib); 1562 1563 if (error == 0 && rt != NULL) { 1564 RT_LOCK(rt); 1565 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = 1566 ifa->ifa_ifp->if_type; 1567 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = 1568 ifa->ifa_ifp->if_index; 1569 RT_REMREF(rt); 1570 RT_UNLOCK(rt); 1571 } else if (error != 0) 1572 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error); 1573 1574 return (error); 1575 } 1576 1577 int 1578 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1579 { 1580 int error = 0; 1581 struct rt_addrinfo info; 1582 struct sockaddr_dl null_sdl; 1583 1584 bzero(&null_sdl, sizeof(null_sdl)); 1585 null_sdl.sdl_len = sizeof(null_sdl); 1586 null_sdl.sdl_family = AF_LINK; 1587 null_sdl.sdl_type = ifa->ifa_ifp->if_type; 1588 null_sdl.sdl_index = ifa->ifa_ifp->if_index; 1589 bzero(&info, sizeof(info)); 1590 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; 1591 info.rti_info[RTAX_DST] = ia; 1592 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1593 error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib); 1594 1595 if (error != 0) 1596 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error); 1597 1598 return (error); 1599 } 1600 1601 int 1602 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa, int fib) 1603 { 1604 struct rtentry *rt; 1605 1606 rt = rtalloc1_fib(sa, 0, 0, fib); 1607 if (rt == NULL) { 1608 log(LOG_DEBUG, "%s: fail", __func__); 1609 return (EHOSTUNREACH); 1610 } 1611 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = 1612 ifa->ifa_ifp->if_type; 1613 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = 1614 ifa->ifa_ifp->if_index; 1615 RTFREE_LOCKED(rt); 1616 1617 return (0); 1618 } 1619 1620 /* 1621 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1622 * structs used to represent other address families, it is necessary 1623 * to perform a different comparison. 1624 */ 1625 1626 #define sa_dl_equal(a1, a2) \ 1627 ((((struct sockaddr_dl *)(a1))->sdl_len == \ 1628 ((struct sockaddr_dl *)(a2))->sdl_len) && \ 1629 (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \ 1630 LLADDR((struct sockaddr_dl *)(a2)), \ 1631 ((struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1632 1633 /* 1634 * Locate an interface based on a complete address. 1635 */ 1636 /*ARGSUSED*/ 1637 static struct ifaddr * 1638 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref) 1639 { 1640 struct ifnet *ifp; 1641 struct ifaddr *ifa; 1642 1643 IFNET_RLOCK_NOSLEEP(); 1644 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1645 IF_ADDR_RLOCK(ifp); 1646 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1647 if (ifa->ifa_addr->sa_family != addr->sa_family) 1648 continue; 1649 if (sa_equal(addr, ifa->ifa_addr)) { 1650 if (getref) 1651 ifa_ref(ifa); 1652 IF_ADDR_RUNLOCK(ifp); 1653 goto done; 1654 } 1655 /* IP6 doesn't have broadcast */ 1656 if ((ifp->if_flags & IFF_BROADCAST) && 1657 ifa->ifa_broadaddr && 1658 ifa->ifa_broadaddr->sa_len != 0 && 1659 sa_equal(ifa->ifa_broadaddr, addr)) { 1660 if (getref) 1661 ifa_ref(ifa); 1662 IF_ADDR_RUNLOCK(ifp); 1663 goto done; 1664 } 1665 } 1666 IF_ADDR_RUNLOCK(ifp); 1667 } 1668 ifa = NULL; 1669 done: 1670 IFNET_RUNLOCK_NOSLEEP(); 1671 return (ifa); 1672 } 1673 1674 struct ifaddr * 1675 ifa_ifwithaddr(struct sockaddr *addr) 1676 { 1677 1678 return (ifa_ifwithaddr_internal(addr, 1)); 1679 } 1680 1681 int 1682 ifa_ifwithaddr_check(struct sockaddr *addr) 1683 { 1684 1685 return (ifa_ifwithaddr_internal(addr, 0) != NULL); 1686 } 1687 1688 /* 1689 * Locate an interface based on the broadcast address. 1690 */ 1691 /* ARGSUSED */ 1692 struct ifaddr * 1693 ifa_ifwithbroadaddr(struct sockaddr *addr, int fibnum) 1694 { 1695 struct ifnet *ifp; 1696 struct ifaddr *ifa; 1697 1698 IFNET_RLOCK_NOSLEEP(); 1699 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1700 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1701 continue; 1702 IF_ADDR_RLOCK(ifp); 1703 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1704 if (ifa->ifa_addr->sa_family != addr->sa_family) 1705 continue; 1706 if ((ifp->if_flags & IFF_BROADCAST) && 1707 ifa->ifa_broadaddr && 1708 ifa->ifa_broadaddr->sa_len != 0 && 1709 sa_equal(ifa->ifa_broadaddr, addr)) { 1710 ifa_ref(ifa); 1711 IF_ADDR_RUNLOCK(ifp); 1712 goto done; 1713 } 1714 } 1715 IF_ADDR_RUNLOCK(ifp); 1716 } 1717 ifa = NULL; 1718 done: 1719 IFNET_RUNLOCK_NOSLEEP(); 1720 return (ifa); 1721 } 1722 1723 /* 1724 * Locate the point to point interface with a given destination address. 1725 */ 1726 /*ARGSUSED*/ 1727 struct ifaddr * 1728 ifa_ifwithdstaddr(struct sockaddr *addr, int fibnum) 1729 { 1730 struct ifnet *ifp; 1731 struct ifaddr *ifa; 1732 1733 IFNET_RLOCK_NOSLEEP(); 1734 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1735 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1736 continue; 1737 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1738 continue; 1739 IF_ADDR_RLOCK(ifp); 1740 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1741 if (ifa->ifa_addr->sa_family != addr->sa_family) 1742 continue; 1743 if (ifa->ifa_dstaddr != NULL && 1744 sa_equal(addr, ifa->ifa_dstaddr)) { 1745 ifa_ref(ifa); 1746 IF_ADDR_RUNLOCK(ifp); 1747 goto done; 1748 } 1749 } 1750 IF_ADDR_RUNLOCK(ifp); 1751 } 1752 ifa = NULL; 1753 done: 1754 IFNET_RUNLOCK_NOSLEEP(); 1755 return (ifa); 1756 } 1757 1758 /* 1759 * Find an interface on a specific network. If many, choice 1760 * is most specific found. 1761 */ 1762 struct ifaddr * 1763 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp, int fibnum) 1764 { 1765 struct ifnet *ifp; 1766 struct ifaddr *ifa; 1767 struct ifaddr *ifa_maybe = NULL; 1768 u_int af = addr->sa_family; 1769 char *addr_data = addr->sa_data, *cplim; 1770 1771 /* 1772 * AF_LINK addresses can be looked up directly by their index number, 1773 * so do that if we can. 1774 */ 1775 if (af == AF_LINK) { 1776 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 1777 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 1778 return (ifaddr_byindex(sdl->sdl_index)); 1779 } 1780 1781 /* 1782 * Scan though each interface, looking for ones that have addresses 1783 * in this address family and the requested fib. Maintain a reference 1784 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 1785 * kept it stable when we move onto the next interface. 1786 */ 1787 IFNET_RLOCK_NOSLEEP(); 1788 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1789 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1790 continue; 1791 IF_ADDR_RLOCK(ifp); 1792 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1793 char *cp, *cp2, *cp3; 1794 1795 if (ifa->ifa_addr->sa_family != af) 1796 next: continue; 1797 if (af == AF_INET && 1798 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1799 /* 1800 * This is a bit broken as it doesn't 1801 * take into account that the remote end may 1802 * be a single node in the network we are 1803 * looking for. 1804 * The trouble is that we don't know the 1805 * netmask for the remote end. 1806 */ 1807 if (ifa->ifa_dstaddr != NULL && 1808 sa_equal(addr, ifa->ifa_dstaddr)) { 1809 ifa_ref(ifa); 1810 IF_ADDR_RUNLOCK(ifp); 1811 goto done; 1812 } 1813 } else { 1814 /* 1815 * Scan all the bits in the ifa's address. 1816 * If a bit dissagrees with what we are 1817 * looking for, mask it with the netmask 1818 * to see if it really matters. 1819 * (A byte at a time) 1820 */ 1821 if (ifa->ifa_netmask == 0) 1822 continue; 1823 cp = addr_data; 1824 cp2 = ifa->ifa_addr->sa_data; 1825 cp3 = ifa->ifa_netmask->sa_data; 1826 cplim = ifa->ifa_netmask->sa_len 1827 + (char *)ifa->ifa_netmask; 1828 while (cp3 < cplim) 1829 if ((*cp++ ^ *cp2++) & *cp3++) 1830 goto next; /* next address! */ 1831 /* 1832 * If the netmask of what we just found 1833 * is more specific than what we had before 1834 * (if we had one), or if the virtual status 1835 * of new prefix is better than of the old one, 1836 * then remember the new one before continuing 1837 * to search for an even better one. 1838 */ 1839 if (ifa_maybe == NULL || 1840 ifa_preferred(ifa_maybe, ifa) || 1841 rn_refines((caddr_t)ifa->ifa_netmask, 1842 (caddr_t)ifa_maybe->ifa_netmask)) { 1843 if (ifa_maybe != NULL) 1844 ifa_free(ifa_maybe); 1845 ifa_maybe = ifa; 1846 ifa_ref(ifa_maybe); 1847 } 1848 } 1849 } 1850 IF_ADDR_RUNLOCK(ifp); 1851 } 1852 ifa = ifa_maybe; 1853 ifa_maybe = NULL; 1854 done: 1855 IFNET_RUNLOCK_NOSLEEP(); 1856 if (ifa_maybe != NULL) 1857 ifa_free(ifa_maybe); 1858 return (ifa); 1859 } 1860 1861 /* 1862 * Find an interface address specific to an interface best matching 1863 * a given address. 1864 */ 1865 struct ifaddr * 1866 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 1867 { 1868 struct ifaddr *ifa; 1869 char *cp, *cp2, *cp3; 1870 char *cplim; 1871 struct ifaddr *ifa_maybe = NULL; 1872 u_int af = addr->sa_family; 1873 1874 if (af >= AF_MAX) 1875 return (NULL); 1876 IF_ADDR_RLOCK(ifp); 1877 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1878 if (ifa->ifa_addr->sa_family != af) 1879 continue; 1880 if (ifa_maybe == NULL) 1881 ifa_maybe = ifa; 1882 if (ifa->ifa_netmask == 0) { 1883 if (sa_equal(addr, ifa->ifa_addr) || 1884 (ifa->ifa_dstaddr && 1885 sa_equal(addr, ifa->ifa_dstaddr))) 1886 goto done; 1887 continue; 1888 } 1889 if (ifp->if_flags & IFF_POINTOPOINT) { 1890 if (sa_equal(addr, ifa->ifa_dstaddr)) 1891 goto done; 1892 } else { 1893 cp = addr->sa_data; 1894 cp2 = ifa->ifa_addr->sa_data; 1895 cp3 = ifa->ifa_netmask->sa_data; 1896 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1897 for (; cp3 < cplim; cp3++) 1898 if ((*cp++ ^ *cp2++) & *cp3) 1899 break; 1900 if (cp3 == cplim) 1901 goto done; 1902 } 1903 } 1904 ifa = ifa_maybe; 1905 done: 1906 if (ifa != NULL) 1907 ifa_ref(ifa); 1908 IF_ADDR_RUNLOCK(ifp); 1909 return (ifa); 1910 } 1911 1912 /* 1913 * See whether new ifa is better than current one: 1914 * 1) A non-virtual one is preferred over virtual. 1915 * 2) A virtual in master state preferred over any other state. 1916 * 1917 * Used in several address selecting functions. 1918 */ 1919 int 1920 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 1921 { 1922 1923 return (cur->ifa_carp && (!next->ifa_carp || 1924 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 1925 } 1926 1927 #include <net/if_llatbl.h> 1928 1929 /* 1930 * Default action when installing a route with a Link Level gateway. 1931 * Lookup an appropriate real ifa to point to. 1932 * This should be moved to /sys/net/link.c eventually. 1933 */ 1934 static void 1935 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 1936 { 1937 struct ifaddr *ifa, *oifa; 1938 struct sockaddr *dst; 1939 struct ifnet *ifp; 1940 1941 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 1942 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 1943 return; 1944 ifa = ifaof_ifpforaddr(dst, ifp); 1945 if (ifa) { 1946 oifa = rt->rt_ifa; 1947 rt->rt_ifa = ifa; 1948 ifa_free(oifa); 1949 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1950 ifa->ifa_rtrequest(cmd, rt, info); 1951 } 1952 } 1953 1954 struct sockaddr_dl * 1955 link_alloc_sdl(size_t size, int flags) 1956 { 1957 1958 return (malloc(size, M_TEMP, flags)); 1959 } 1960 1961 void 1962 link_free_sdl(struct sockaddr *sa) 1963 { 1964 free(sa, M_TEMP); 1965 } 1966 1967 /* 1968 * Fills in given sdl with interface basic info. 1969 * Returns pointer to filled sdl. 1970 */ 1971 struct sockaddr_dl * 1972 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 1973 { 1974 struct sockaddr_dl *sdl; 1975 1976 sdl = (struct sockaddr_dl *)paddr; 1977 memset(sdl, 0, sizeof(struct sockaddr_dl)); 1978 sdl->sdl_len = sizeof(struct sockaddr_dl); 1979 sdl->sdl_family = AF_LINK; 1980 sdl->sdl_index = ifp->if_index; 1981 sdl->sdl_type = iftype; 1982 1983 return (sdl); 1984 } 1985 1986 /* 1987 * Mark an interface down and notify protocols of 1988 * the transition. 1989 */ 1990 static void 1991 if_unroute(struct ifnet *ifp, int flag, int fam) 1992 { 1993 struct ifaddr *ifa; 1994 1995 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 1996 1997 ifp->if_flags &= ~flag; 1998 getmicrotime(&ifp->if_lastchange); 1999 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2000 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2001 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2002 ifp->if_qflush(ifp); 2003 2004 if (ifp->if_carp) 2005 (*carp_linkstate_p)(ifp); 2006 rt_ifmsg(ifp); 2007 } 2008 2009 /* 2010 * Mark an interface up and notify protocols of 2011 * the transition. 2012 */ 2013 static void 2014 if_route(struct ifnet *ifp, int flag, int fam) 2015 { 2016 struct ifaddr *ifa; 2017 2018 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2019 2020 ifp->if_flags |= flag; 2021 getmicrotime(&ifp->if_lastchange); 2022 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2023 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2024 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2025 if (ifp->if_carp) 2026 (*carp_linkstate_p)(ifp); 2027 rt_ifmsg(ifp); 2028 #ifdef INET6 2029 in6_if_up(ifp); 2030 #endif 2031 } 2032 2033 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2034 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2035 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2036 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2037 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2038 int (*vlan_setcookie_p)(struct ifnet *, void *); 2039 void *(*vlan_cookie_p)(struct ifnet *); 2040 2041 /* 2042 * Handle a change in the interface link state. To avoid LORs 2043 * between driver lock and upper layer locks, as well as possible 2044 * recursions, we post event to taskqueue, and all job 2045 * is done in static do_link_state_change(). 2046 */ 2047 void 2048 if_link_state_change(struct ifnet *ifp, int link_state) 2049 { 2050 /* Return if state hasn't changed. */ 2051 if (ifp->if_link_state == link_state) 2052 return; 2053 2054 ifp->if_link_state = link_state; 2055 2056 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2057 } 2058 2059 static void 2060 do_link_state_change(void *arg, int pending) 2061 { 2062 struct ifnet *ifp = (struct ifnet *)arg; 2063 int link_state = ifp->if_link_state; 2064 CURVNET_SET(ifp->if_vnet); 2065 2066 /* Notify that the link state has changed. */ 2067 rt_ifmsg(ifp); 2068 if (ifp->if_vlantrunk != NULL) 2069 (*vlan_link_state_p)(ifp); 2070 2071 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2072 ifp->if_l2com != NULL) 2073 (*ng_ether_link_state_p)(ifp, link_state); 2074 if (ifp->if_carp) 2075 (*carp_linkstate_p)(ifp); 2076 if (ifp->if_bridge) 2077 (*bridge_linkstate_p)(ifp); 2078 if (ifp->if_lagg) 2079 (*lagg_linkstate_p)(ifp, link_state); 2080 2081 if (IS_DEFAULT_VNET(curvnet)) 2082 devctl_notify("IFNET", ifp->if_xname, 2083 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2084 NULL); 2085 if (pending > 1) 2086 if_printf(ifp, "%d link states coalesced\n", pending); 2087 if (log_link_state_change) 2088 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, 2089 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2090 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state); 2091 CURVNET_RESTORE(); 2092 } 2093 2094 /* 2095 * Mark an interface down and notify protocols of 2096 * the transition. 2097 */ 2098 void 2099 if_down(struct ifnet *ifp) 2100 { 2101 2102 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2103 } 2104 2105 /* 2106 * Mark an interface up and notify protocols of 2107 * the transition. 2108 */ 2109 void 2110 if_up(struct ifnet *ifp) 2111 { 2112 2113 if_route(ifp, IFF_UP, AF_UNSPEC); 2114 } 2115 2116 /* 2117 * Flush an interface queue. 2118 */ 2119 void 2120 if_qflush(struct ifnet *ifp) 2121 { 2122 struct mbuf *m, *n; 2123 struct ifaltq *ifq; 2124 2125 ifq = &ifp->if_snd; 2126 IFQ_LOCK(ifq); 2127 #ifdef ALTQ 2128 if (ALTQ_IS_ENABLED(ifq)) 2129 ALTQ_PURGE(ifq); 2130 #endif 2131 n = ifq->ifq_head; 2132 while ((m = n) != 0) { 2133 n = m->m_nextpkt; 2134 m_freem(m); 2135 } 2136 ifq->ifq_head = 0; 2137 ifq->ifq_tail = 0; 2138 ifq->ifq_len = 0; 2139 IFQ_UNLOCK(ifq); 2140 } 2141 2142 /* 2143 * Map interface name to interface structure pointer, with or without 2144 * returning a reference. 2145 */ 2146 struct ifnet * 2147 ifunit_ref(const char *name) 2148 { 2149 struct ifnet *ifp; 2150 2151 IFNET_RLOCK_NOSLEEP(); 2152 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2153 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2154 !(ifp->if_flags & IFF_DYING)) 2155 break; 2156 } 2157 if (ifp != NULL) 2158 if_ref(ifp); 2159 IFNET_RUNLOCK_NOSLEEP(); 2160 return (ifp); 2161 } 2162 2163 struct ifnet * 2164 ifunit(const char *name) 2165 { 2166 struct ifnet *ifp; 2167 2168 IFNET_RLOCK_NOSLEEP(); 2169 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2170 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2171 break; 2172 } 2173 IFNET_RUNLOCK_NOSLEEP(); 2174 return (ifp); 2175 } 2176 2177 /* 2178 * Hardware specific interface ioctls. 2179 */ 2180 static int 2181 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2182 { 2183 struct ifreq *ifr; 2184 int error = 0; 2185 int new_flags, temp_flags; 2186 size_t namelen, onamelen; 2187 size_t descrlen; 2188 char *descrbuf, *odescrbuf; 2189 char new_name[IFNAMSIZ]; 2190 struct ifaddr *ifa; 2191 struct sockaddr_dl *sdl; 2192 2193 ifr = (struct ifreq *)data; 2194 switch (cmd) { 2195 case SIOCGIFINDEX: 2196 ifr->ifr_index = ifp->if_index; 2197 break; 2198 2199 case SIOCGIFFLAGS: 2200 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2201 ifr->ifr_flags = temp_flags & 0xffff; 2202 ifr->ifr_flagshigh = temp_flags >> 16; 2203 break; 2204 2205 case SIOCGIFCAP: 2206 ifr->ifr_reqcap = ifp->if_capabilities; 2207 ifr->ifr_curcap = ifp->if_capenable; 2208 break; 2209 2210 #ifdef MAC 2211 case SIOCGIFMAC: 2212 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2213 break; 2214 #endif 2215 2216 case SIOCGIFMETRIC: 2217 ifr->ifr_metric = ifp->if_metric; 2218 break; 2219 2220 case SIOCGIFMTU: 2221 ifr->ifr_mtu = ifp->if_mtu; 2222 break; 2223 2224 case SIOCGIFPHYS: 2225 /* XXXGL: did this ever worked? */ 2226 ifr->ifr_phys = 0; 2227 break; 2228 2229 case SIOCGIFDESCR: 2230 error = 0; 2231 sx_slock(&ifdescr_sx); 2232 if (ifp->if_description == NULL) 2233 error = ENOMSG; 2234 else { 2235 /* space for terminating nul */ 2236 descrlen = strlen(ifp->if_description) + 1; 2237 if (ifr->ifr_buffer.length < descrlen) 2238 ifr->ifr_buffer.buffer = NULL; 2239 else 2240 error = copyout(ifp->if_description, 2241 ifr->ifr_buffer.buffer, descrlen); 2242 ifr->ifr_buffer.length = descrlen; 2243 } 2244 sx_sunlock(&ifdescr_sx); 2245 break; 2246 2247 case SIOCSIFDESCR: 2248 error = priv_check(td, PRIV_NET_SETIFDESCR); 2249 if (error) 2250 return (error); 2251 2252 /* 2253 * Copy only (length-1) bytes to make sure that 2254 * if_description is always nul terminated. The 2255 * length parameter is supposed to count the 2256 * terminating nul in. 2257 */ 2258 if (ifr->ifr_buffer.length > ifdescr_maxlen) 2259 return (ENAMETOOLONG); 2260 else if (ifr->ifr_buffer.length == 0) 2261 descrbuf = NULL; 2262 else { 2263 descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR, 2264 M_WAITOK | M_ZERO); 2265 error = copyin(ifr->ifr_buffer.buffer, descrbuf, 2266 ifr->ifr_buffer.length - 1); 2267 if (error) { 2268 free(descrbuf, M_IFDESCR); 2269 break; 2270 } 2271 } 2272 2273 sx_xlock(&ifdescr_sx); 2274 odescrbuf = ifp->if_description; 2275 ifp->if_description = descrbuf; 2276 sx_xunlock(&ifdescr_sx); 2277 2278 getmicrotime(&ifp->if_lastchange); 2279 free(odescrbuf, M_IFDESCR); 2280 break; 2281 2282 case SIOCGIFFIB: 2283 ifr->ifr_fib = ifp->if_fib; 2284 break; 2285 2286 case SIOCSIFFIB: 2287 error = priv_check(td, PRIV_NET_SETIFFIB); 2288 if (error) 2289 return (error); 2290 if (ifr->ifr_fib >= rt_numfibs) 2291 return (EINVAL); 2292 2293 ifp->if_fib = ifr->ifr_fib; 2294 break; 2295 2296 case SIOCSIFFLAGS: 2297 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2298 if (error) 2299 return (error); 2300 /* 2301 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2302 * check, so we don't need special handling here yet. 2303 */ 2304 new_flags = (ifr->ifr_flags & 0xffff) | 2305 (ifr->ifr_flagshigh << 16); 2306 if (ifp->if_flags & IFF_UP && 2307 (new_flags & IFF_UP) == 0) { 2308 if_down(ifp); 2309 } else if (new_flags & IFF_UP && 2310 (ifp->if_flags & IFF_UP) == 0) { 2311 if_up(ifp); 2312 } 2313 /* See if permanently promiscuous mode bit is about to flip */ 2314 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2315 if (new_flags & IFF_PPROMISC) 2316 ifp->if_flags |= IFF_PROMISC; 2317 else if (ifp->if_pcount == 0) 2318 ifp->if_flags &= ~IFF_PROMISC; 2319 log(LOG_INFO, "%s: permanently promiscuous mode %s\n", 2320 ifp->if_xname, 2321 (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); 2322 } 2323 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2324 (new_flags &~ IFF_CANTCHANGE); 2325 if (ifp->if_ioctl) { 2326 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2327 } 2328 getmicrotime(&ifp->if_lastchange); 2329 break; 2330 2331 case SIOCSIFCAP: 2332 error = priv_check(td, PRIV_NET_SETIFCAP); 2333 if (error) 2334 return (error); 2335 if (ifp->if_ioctl == NULL) 2336 return (EOPNOTSUPP); 2337 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2338 return (EINVAL); 2339 error = (*ifp->if_ioctl)(ifp, cmd, data); 2340 if (error == 0) 2341 getmicrotime(&ifp->if_lastchange); 2342 break; 2343 2344 #ifdef MAC 2345 case SIOCSIFMAC: 2346 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2347 break; 2348 #endif 2349 2350 case SIOCSIFNAME: 2351 error = priv_check(td, PRIV_NET_SETIFNAME); 2352 if (error) 2353 return (error); 2354 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 2355 if (error != 0) 2356 return (error); 2357 if (new_name[0] == '\0') 2358 return (EINVAL); 2359 if (ifunit(new_name) != NULL) 2360 return (EEXIST); 2361 2362 /* 2363 * XXX: Locking. Nothing else seems to lock if_flags, 2364 * and there are numerous other races with the 2365 * ifunit() checks not being atomic with namespace 2366 * changes (renames, vmoves, if_attach, etc). 2367 */ 2368 ifp->if_flags |= IFF_RENAMING; 2369 2370 /* Announce the departure of the interface. */ 2371 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2372 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2373 2374 log(LOG_INFO, "%s: changing name to '%s'\n", 2375 ifp->if_xname, new_name); 2376 2377 IF_ADDR_WLOCK(ifp); 2378 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2379 ifa = ifp->if_addr; 2380 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2381 namelen = strlen(new_name); 2382 onamelen = sdl->sdl_nlen; 2383 /* 2384 * Move the address if needed. This is safe because we 2385 * allocate space for a name of length IFNAMSIZ when we 2386 * create this in if_attach(). 2387 */ 2388 if (namelen != onamelen) { 2389 bcopy(sdl->sdl_data + onamelen, 2390 sdl->sdl_data + namelen, sdl->sdl_alen); 2391 } 2392 bcopy(new_name, sdl->sdl_data, namelen); 2393 sdl->sdl_nlen = namelen; 2394 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2395 bzero(sdl->sdl_data, onamelen); 2396 while (namelen != 0) 2397 sdl->sdl_data[--namelen] = 0xff; 2398 IF_ADDR_WUNLOCK(ifp); 2399 2400 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2401 /* Announce the return of the interface. */ 2402 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2403 2404 ifp->if_flags &= ~IFF_RENAMING; 2405 break; 2406 2407 #ifdef VIMAGE 2408 case SIOCSIFVNET: 2409 error = priv_check(td, PRIV_NET_SETIFVNET); 2410 if (error) 2411 return (error); 2412 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2413 break; 2414 #endif 2415 2416 case SIOCSIFMETRIC: 2417 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2418 if (error) 2419 return (error); 2420 ifp->if_metric = ifr->ifr_metric; 2421 getmicrotime(&ifp->if_lastchange); 2422 break; 2423 2424 case SIOCSIFPHYS: 2425 error = priv_check(td, PRIV_NET_SETIFPHYS); 2426 if (error) 2427 return (error); 2428 if (ifp->if_ioctl == NULL) 2429 return (EOPNOTSUPP); 2430 error = (*ifp->if_ioctl)(ifp, cmd, data); 2431 if (error == 0) 2432 getmicrotime(&ifp->if_lastchange); 2433 break; 2434 2435 case SIOCSIFMTU: 2436 { 2437 u_long oldmtu = ifp->if_mtu; 2438 2439 error = priv_check(td, PRIV_NET_SETIFMTU); 2440 if (error) 2441 return (error); 2442 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2443 return (EINVAL); 2444 if (ifp->if_ioctl == NULL) 2445 return (EOPNOTSUPP); 2446 error = (*ifp->if_ioctl)(ifp, cmd, data); 2447 if (error == 0) { 2448 getmicrotime(&ifp->if_lastchange); 2449 rt_ifmsg(ifp); 2450 } 2451 /* 2452 * If the link MTU changed, do network layer specific procedure. 2453 */ 2454 if (ifp->if_mtu != oldmtu) { 2455 #ifdef INET6 2456 nd6_setmtu(ifp); 2457 #endif 2458 rt_updatemtu(ifp); 2459 } 2460 break; 2461 } 2462 2463 case SIOCADDMULTI: 2464 case SIOCDELMULTI: 2465 if (cmd == SIOCADDMULTI) 2466 error = priv_check(td, PRIV_NET_ADDMULTI); 2467 else 2468 error = priv_check(td, PRIV_NET_DELMULTI); 2469 if (error) 2470 return (error); 2471 2472 /* Don't allow group membership on non-multicast interfaces. */ 2473 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2474 return (EOPNOTSUPP); 2475 2476 /* Don't let users screw up protocols' entries. */ 2477 if (ifr->ifr_addr.sa_family != AF_LINK) 2478 return (EINVAL); 2479 2480 if (cmd == SIOCADDMULTI) { 2481 struct ifmultiaddr *ifma; 2482 2483 /* 2484 * Userland is only permitted to join groups once 2485 * via the if_addmulti() KPI, because it cannot hold 2486 * struct ifmultiaddr * between calls. It may also 2487 * lose a race while we check if the membership 2488 * already exists. 2489 */ 2490 IF_ADDR_RLOCK(ifp); 2491 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2492 IF_ADDR_RUNLOCK(ifp); 2493 if (ifma != NULL) 2494 error = EADDRINUSE; 2495 else 2496 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2497 } else { 2498 error = if_delmulti(ifp, &ifr->ifr_addr); 2499 } 2500 if (error == 0) 2501 getmicrotime(&ifp->if_lastchange); 2502 break; 2503 2504 case SIOCSIFPHYADDR: 2505 case SIOCDIFPHYADDR: 2506 #ifdef INET6 2507 case SIOCSIFPHYADDR_IN6: 2508 #endif 2509 case SIOCSIFMEDIA: 2510 case SIOCSIFGENERIC: 2511 error = priv_check(td, PRIV_NET_HWIOCTL); 2512 if (error) 2513 return (error); 2514 if (ifp->if_ioctl == NULL) 2515 return (EOPNOTSUPP); 2516 error = (*ifp->if_ioctl)(ifp, cmd, data); 2517 if (error == 0) 2518 getmicrotime(&ifp->if_lastchange); 2519 break; 2520 2521 case SIOCGIFSTATUS: 2522 case SIOCGIFPSRCADDR: 2523 case SIOCGIFPDSTADDR: 2524 case SIOCGIFMEDIA: 2525 case SIOCGIFXMEDIA: 2526 case SIOCGIFGENERIC: 2527 if (ifp->if_ioctl == NULL) 2528 return (EOPNOTSUPP); 2529 error = (*ifp->if_ioctl)(ifp, cmd, data); 2530 break; 2531 2532 case SIOCSIFLLADDR: 2533 error = priv_check(td, PRIV_NET_SETLLADDR); 2534 if (error) 2535 return (error); 2536 error = if_setlladdr(ifp, 2537 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2538 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 2539 break; 2540 2541 case SIOCAIFGROUP: 2542 { 2543 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2544 2545 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2546 if (error) 2547 return (error); 2548 if ((error = if_addgroup(ifp, ifgr->ifgr_group))) 2549 return (error); 2550 break; 2551 } 2552 2553 case SIOCGIFGROUP: 2554 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) 2555 return (error); 2556 break; 2557 2558 case SIOCDIFGROUP: 2559 { 2560 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; 2561 2562 error = priv_check(td, PRIV_NET_DELIFGROUP); 2563 if (error) 2564 return (error); 2565 if ((error = if_delgroup(ifp, ifgr->ifgr_group))) 2566 return (error); 2567 break; 2568 } 2569 2570 default: 2571 error = ENOIOCTL; 2572 break; 2573 } 2574 return (error); 2575 } 2576 2577 #ifdef COMPAT_FREEBSD32 2578 struct ifconf32 { 2579 int32_t ifc_len; 2580 union { 2581 uint32_t ifcu_buf; 2582 uint32_t ifcu_req; 2583 } ifc_ifcu; 2584 }; 2585 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2586 #endif 2587 2588 /* 2589 * Interface ioctls. 2590 */ 2591 int 2592 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2593 { 2594 struct ifnet *ifp; 2595 struct ifreq *ifr; 2596 int error; 2597 int oif_flags; 2598 2599 CURVNET_SET(so->so_vnet); 2600 switch (cmd) { 2601 case SIOCGIFCONF: 2602 error = ifconf(cmd, data); 2603 CURVNET_RESTORE(); 2604 return (error); 2605 2606 #ifdef COMPAT_FREEBSD32 2607 case SIOCGIFCONF32: 2608 { 2609 struct ifconf32 *ifc32; 2610 struct ifconf ifc; 2611 2612 ifc32 = (struct ifconf32 *)data; 2613 ifc.ifc_len = ifc32->ifc_len; 2614 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2615 2616 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2617 CURVNET_RESTORE(); 2618 if (error == 0) 2619 ifc32->ifc_len = ifc.ifc_len; 2620 return (error); 2621 } 2622 #endif 2623 } 2624 ifr = (struct ifreq *)data; 2625 2626 switch (cmd) { 2627 #ifdef VIMAGE 2628 case SIOCSIFRVNET: 2629 error = priv_check(td, PRIV_NET_SETIFVNET); 2630 if (error == 0) 2631 error = if_vmove_reclaim(td, ifr->ifr_name, 2632 ifr->ifr_jid); 2633 CURVNET_RESTORE(); 2634 return (error); 2635 #endif 2636 case SIOCIFCREATE: 2637 case SIOCIFCREATE2: 2638 error = priv_check(td, PRIV_NET_IFCREATE); 2639 if (error == 0) 2640 error = if_clone_create(ifr->ifr_name, 2641 sizeof(ifr->ifr_name), 2642 cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL); 2643 CURVNET_RESTORE(); 2644 return (error); 2645 case SIOCIFDESTROY: 2646 error = priv_check(td, PRIV_NET_IFDESTROY); 2647 if (error == 0) 2648 error = if_clone_destroy(ifr->ifr_name); 2649 CURVNET_RESTORE(); 2650 return (error); 2651 2652 case SIOCIFGCLONERS: 2653 error = if_clone_list((struct if_clonereq *)data); 2654 CURVNET_RESTORE(); 2655 return (error); 2656 case SIOCGIFGMEMB: 2657 error = if_getgroupmembers((struct ifgroupreq *)data); 2658 CURVNET_RESTORE(); 2659 return (error); 2660 #if defined(INET) || defined(INET6) 2661 case SIOCSVH: 2662 case SIOCGVH: 2663 if (carp_ioctl_p == NULL) 2664 error = EPROTONOSUPPORT; 2665 else 2666 error = (*carp_ioctl_p)(ifr, cmd, td); 2667 CURVNET_RESTORE(); 2668 return (error); 2669 #endif 2670 } 2671 2672 ifp = ifunit_ref(ifr->ifr_name); 2673 if (ifp == NULL) { 2674 CURVNET_RESTORE(); 2675 return (ENXIO); 2676 } 2677 2678 error = ifhwioctl(cmd, ifp, data, td); 2679 if (error != ENOIOCTL) { 2680 if_rele(ifp); 2681 CURVNET_RESTORE(); 2682 return (error); 2683 } 2684 2685 oif_flags = ifp->if_flags; 2686 if (so->so_proto == NULL) { 2687 if_rele(ifp); 2688 CURVNET_RESTORE(); 2689 return (EOPNOTSUPP); 2690 } 2691 2692 /* 2693 * Pass the request on to the socket control method, and if the 2694 * latter returns EOPNOTSUPP, directly to the interface. 2695 * 2696 * Make an exception for the legacy SIOCSIF* requests. Drivers 2697 * trust SIOCSIFADDR et al to come from an already privileged 2698 * layer, and do not perform any credentials checks or input 2699 * validation. 2700 */ 2701 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 2702 ifp, td)); 2703 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2704 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2705 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2706 error = (*ifp->if_ioctl)(ifp, cmd, data); 2707 2708 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 2709 #ifdef INET6 2710 if (ifp->if_flags & IFF_UP) 2711 in6_if_up(ifp); 2712 #endif 2713 } 2714 if_rele(ifp); 2715 CURVNET_RESTORE(); 2716 return (error); 2717 } 2718 2719 /* 2720 * The code common to handling reference counted flags, 2721 * e.g., in ifpromisc() and if_allmulti(). 2722 * The "pflag" argument can specify a permanent mode flag to check, 2723 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 2724 * 2725 * Only to be used on stack-owned flags, not driver-owned flags. 2726 */ 2727 static int 2728 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 2729 { 2730 struct ifreq ifr; 2731 int error; 2732 int oldflags, oldcount; 2733 2734 /* Sanity checks to catch programming errors */ 2735 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 2736 ("%s: setting driver-owned flag %d", __func__, flag)); 2737 2738 if (onswitch) 2739 KASSERT(*refcount >= 0, 2740 ("%s: increment negative refcount %d for flag %d", 2741 __func__, *refcount, flag)); 2742 else 2743 KASSERT(*refcount > 0, 2744 ("%s: decrement non-positive refcount %d for flag %d", 2745 __func__, *refcount, flag)); 2746 2747 /* In case this mode is permanent, just touch refcount */ 2748 if (ifp->if_flags & pflag) { 2749 *refcount += onswitch ? 1 : -1; 2750 return (0); 2751 } 2752 2753 /* Save ifnet parameters for if_ioctl() may fail */ 2754 oldcount = *refcount; 2755 oldflags = ifp->if_flags; 2756 2757 /* 2758 * See if we aren't the only and touching refcount is enough. 2759 * Actually toggle interface flag if we are the first or last. 2760 */ 2761 if (onswitch) { 2762 if ((*refcount)++) 2763 return (0); 2764 ifp->if_flags |= flag; 2765 } else { 2766 if (--(*refcount)) 2767 return (0); 2768 ifp->if_flags &= ~flag; 2769 } 2770 2771 /* Call down the driver since we've changed interface flags */ 2772 if (ifp->if_ioctl == NULL) { 2773 error = EOPNOTSUPP; 2774 goto recover; 2775 } 2776 ifr.ifr_flags = ifp->if_flags & 0xffff; 2777 ifr.ifr_flagshigh = ifp->if_flags >> 16; 2778 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 2779 if (error) 2780 goto recover; 2781 /* Notify userland that interface flags have changed */ 2782 rt_ifmsg(ifp); 2783 return (0); 2784 2785 recover: 2786 /* Recover after driver error */ 2787 *refcount = oldcount; 2788 ifp->if_flags = oldflags; 2789 return (error); 2790 } 2791 2792 /* 2793 * Set/clear promiscuous mode on interface ifp based on the truth value 2794 * of pswitch. The calls are reference counted so that only the first 2795 * "on" request actually has an effect, as does the final "off" request. 2796 * Results are undefined if the "off" and "on" requests are not matched. 2797 */ 2798 int 2799 ifpromisc(struct ifnet *ifp, int pswitch) 2800 { 2801 int error; 2802 int oldflags = ifp->if_flags; 2803 2804 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 2805 &ifp->if_pcount, pswitch); 2806 /* If promiscuous mode status has changed, log a message */ 2807 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) 2808 log(LOG_INFO, "%s: promiscuous mode %s\n", 2809 ifp->if_xname, 2810 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 2811 return (error); 2812 } 2813 2814 /* 2815 * Return interface configuration 2816 * of system. List may be used 2817 * in later ioctl's (above) to get 2818 * other information. 2819 */ 2820 /*ARGSUSED*/ 2821 static int 2822 ifconf(u_long cmd, caddr_t data) 2823 { 2824 struct ifconf *ifc = (struct ifconf *)data; 2825 struct ifnet *ifp; 2826 struct ifaddr *ifa; 2827 struct ifreq ifr; 2828 struct sbuf *sb; 2829 int error, full = 0, valid_len, max_len; 2830 2831 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 2832 max_len = MAXPHYS - 1; 2833 2834 /* Prevent hostile input from being able to crash the system */ 2835 if (ifc->ifc_len <= 0) 2836 return (EINVAL); 2837 2838 again: 2839 if (ifc->ifc_len <= max_len) { 2840 max_len = ifc->ifc_len; 2841 full = 1; 2842 } 2843 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 2844 max_len = 0; 2845 valid_len = 0; 2846 2847 IFNET_RLOCK(); 2848 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2849 int addrs; 2850 2851 /* 2852 * Zero the ifr_name buffer to make sure we don't 2853 * disclose the contents of the stack. 2854 */ 2855 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); 2856 2857 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 2858 >= sizeof(ifr.ifr_name)) { 2859 sbuf_delete(sb); 2860 IFNET_RUNLOCK(); 2861 return (ENAMETOOLONG); 2862 } 2863 2864 addrs = 0; 2865 IF_ADDR_RLOCK(ifp); 2866 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2867 struct sockaddr *sa = ifa->ifa_addr; 2868 2869 if (prison_if(curthread->td_ucred, sa) != 0) 2870 continue; 2871 addrs++; 2872 if (sa->sa_len <= sizeof(*sa)) { 2873 ifr.ifr_addr = *sa; 2874 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2875 max_len += sizeof(ifr); 2876 } else { 2877 sbuf_bcat(sb, &ifr, 2878 offsetof(struct ifreq, ifr_addr)); 2879 max_len += offsetof(struct ifreq, ifr_addr); 2880 sbuf_bcat(sb, sa, sa->sa_len); 2881 max_len += sa->sa_len; 2882 } 2883 2884 if (sbuf_error(sb) == 0) 2885 valid_len = sbuf_len(sb); 2886 } 2887 IF_ADDR_RUNLOCK(ifp); 2888 if (addrs == 0) { 2889 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 2890 sbuf_bcat(sb, &ifr, sizeof(ifr)); 2891 max_len += sizeof(ifr); 2892 2893 if (sbuf_error(sb) == 0) 2894 valid_len = sbuf_len(sb); 2895 } 2896 } 2897 IFNET_RUNLOCK(); 2898 2899 /* 2900 * If we didn't allocate enough space (uncommon), try again. If 2901 * we have already allocated as much space as we are allowed, 2902 * return what we've got. 2903 */ 2904 if (valid_len != max_len && !full) { 2905 sbuf_delete(sb); 2906 goto again; 2907 } 2908 2909 ifc->ifc_len = valid_len; 2910 sbuf_finish(sb); 2911 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 2912 sbuf_delete(sb); 2913 return (error); 2914 } 2915 2916 /* 2917 * Just like ifpromisc(), but for all-multicast-reception mode. 2918 */ 2919 int 2920 if_allmulti(struct ifnet *ifp, int onswitch) 2921 { 2922 2923 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 2924 } 2925 2926 struct ifmultiaddr * 2927 if_findmulti(struct ifnet *ifp, struct sockaddr *sa) 2928 { 2929 struct ifmultiaddr *ifma; 2930 2931 IF_ADDR_LOCK_ASSERT(ifp); 2932 2933 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2934 if (sa->sa_family == AF_LINK) { 2935 if (sa_dl_equal(ifma->ifma_addr, sa)) 2936 break; 2937 } else { 2938 if (sa_equal(ifma->ifma_addr, sa)) 2939 break; 2940 } 2941 } 2942 2943 return ifma; 2944 } 2945 2946 /* 2947 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 2948 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 2949 * the ifnet multicast address list here, so the caller must do that and 2950 * other setup work (such as notifying the device driver). The reference 2951 * count is initialized to 1. 2952 */ 2953 static struct ifmultiaddr * 2954 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 2955 int mflags) 2956 { 2957 struct ifmultiaddr *ifma; 2958 struct sockaddr *dupsa; 2959 2960 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 2961 M_ZERO); 2962 if (ifma == NULL) 2963 return (NULL); 2964 2965 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 2966 if (dupsa == NULL) { 2967 free(ifma, M_IFMADDR); 2968 return (NULL); 2969 } 2970 bcopy(sa, dupsa, sa->sa_len); 2971 ifma->ifma_addr = dupsa; 2972 2973 ifma->ifma_ifp = ifp; 2974 ifma->ifma_refcount = 1; 2975 ifma->ifma_protospec = NULL; 2976 2977 if (llsa == NULL) { 2978 ifma->ifma_lladdr = NULL; 2979 return (ifma); 2980 } 2981 2982 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 2983 if (dupsa == NULL) { 2984 free(ifma->ifma_addr, M_IFMADDR); 2985 free(ifma, M_IFMADDR); 2986 return (NULL); 2987 } 2988 bcopy(llsa, dupsa, llsa->sa_len); 2989 ifma->ifma_lladdr = dupsa; 2990 2991 return (ifma); 2992 } 2993 2994 /* 2995 * if_freemulti: free ifmultiaddr structure and possibly attached related 2996 * addresses. The caller is responsible for implementing reference 2997 * counting, notifying the driver, handling routing messages, and releasing 2998 * any dependent link layer state. 2999 */ 3000 static void 3001 if_freemulti(struct ifmultiaddr *ifma) 3002 { 3003 3004 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3005 ifma->ifma_refcount)); 3006 3007 if (ifma->ifma_lladdr != NULL) 3008 free(ifma->ifma_lladdr, M_IFMADDR); 3009 free(ifma->ifma_addr, M_IFMADDR); 3010 free(ifma, M_IFMADDR); 3011 } 3012 3013 /* 3014 * Register an additional multicast address with a network interface. 3015 * 3016 * - If the address is already present, bump the reference count on the 3017 * address and return. 3018 * - If the address is not link-layer, look up a link layer address. 3019 * - Allocate address structures for one or both addresses, and attach to the 3020 * multicast address list on the interface. If automatically adding a link 3021 * layer address, the protocol address will own a reference to the link 3022 * layer address, to be freed when it is freed. 3023 * - Notify the network device driver of an addition to the multicast address 3024 * list. 3025 * 3026 * 'sa' points to caller-owned memory with the desired multicast address. 3027 * 3028 * 'retifma' will be used to return a pointer to the resulting multicast 3029 * address reference, if desired. 3030 */ 3031 int 3032 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3033 struct ifmultiaddr **retifma) 3034 { 3035 struct ifmultiaddr *ifma, *ll_ifma; 3036 struct sockaddr *llsa; 3037 struct sockaddr_dl sdl; 3038 int error; 3039 3040 /* 3041 * If the address is already present, return a new reference to it; 3042 * otherwise, allocate storage and set up a new address. 3043 */ 3044 IF_ADDR_WLOCK(ifp); 3045 ifma = if_findmulti(ifp, sa); 3046 if (ifma != NULL) { 3047 ifma->ifma_refcount++; 3048 if (retifma != NULL) 3049 *retifma = ifma; 3050 IF_ADDR_WUNLOCK(ifp); 3051 return (0); 3052 } 3053 3054 /* 3055 * The address isn't already present; resolve the protocol address 3056 * into a link layer address, and then look that up, bump its 3057 * refcount or allocate an ifma for that also. 3058 * Most link layer resolving functions returns address data which 3059 * fits inside default sockaddr_dl structure. However callback 3060 * can allocate another sockaddr structure, in that case we need to 3061 * free it later. 3062 */ 3063 llsa = NULL; 3064 ll_ifma = NULL; 3065 if (ifp->if_resolvemulti != NULL) { 3066 /* Provide called function with buffer size information */ 3067 sdl.sdl_len = sizeof(sdl); 3068 llsa = (struct sockaddr *)&sdl; 3069 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3070 if (error) 3071 goto unlock_out; 3072 } 3073 3074 /* 3075 * Allocate the new address. Don't hook it up yet, as we may also 3076 * need to allocate a link layer multicast address. 3077 */ 3078 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3079 if (ifma == NULL) { 3080 error = ENOMEM; 3081 goto free_llsa_out; 3082 } 3083 3084 /* 3085 * If a link layer address is found, we'll need to see if it's 3086 * already present in the address list, or allocate is as well. 3087 * When this block finishes, the link layer address will be on the 3088 * list. 3089 */ 3090 if (llsa != NULL) { 3091 ll_ifma = if_findmulti(ifp, llsa); 3092 if (ll_ifma == NULL) { 3093 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3094 if (ll_ifma == NULL) { 3095 --ifma->ifma_refcount; 3096 if_freemulti(ifma); 3097 error = ENOMEM; 3098 goto free_llsa_out; 3099 } 3100 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3101 ifma_link); 3102 } else 3103 ll_ifma->ifma_refcount++; 3104 ifma->ifma_llifma = ll_ifma; 3105 } 3106 3107 /* 3108 * We now have a new multicast address, ifma, and possibly a new or 3109 * referenced link layer address. Add the primary address to the 3110 * ifnet address list. 3111 */ 3112 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3113 3114 if (retifma != NULL) 3115 *retifma = ifma; 3116 3117 /* 3118 * Must generate the message while holding the lock so that 'ifma' 3119 * pointer is still valid. 3120 */ 3121 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3122 IF_ADDR_WUNLOCK(ifp); 3123 3124 /* 3125 * We are certain we have added something, so call down to the 3126 * interface to let them know about it. 3127 */ 3128 if (ifp->if_ioctl != NULL) { 3129 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3130 } 3131 3132 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3133 link_free_sdl(llsa); 3134 3135 return (0); 3136 3137 free_llsa_out: 3138 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3139 link_free_sdl(llsa); 3140 3141 unlock_out: 3142 IF_ADDR_WUNLOCK(ifp); 3143 return (error); 3144 } 3145 3146 /* 3147 * Delete a multicast group membership by network-layer group address. 3148 * 3149 * Returns ENOENT if the entry could not be found. If ifp no longer 3150 * exists, results are undefined. This entry point should only be used 3151 * from subsystems which do appropriate locking to hold ifp for the 3152 * duration of the call. 3153 * Network-layer protocol domains must use if_delmulti_ifma(). 3154 */ 3155 int 3156 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3157 { 3158 struct ifmultiaddr *ifma; 3159 int lastref; 3160 #ifdef INVARIANTS 3161 struct ifnet *oifp; 3162 3163 IFNET_RLOCK_NOSLEEP(); 3164 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3165 if (ifp == oifp) 3166 break; 3167 if (ifp != oifp) 3168 ifp = NULL; 3169 IFNET_RUNLOCK_NOSLEEP(); 3170 3171 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3172 #endif 3173 if (ifp == NULL) 3174 return (ENOENT); 3175 3176 IF_ADDR_WLOCK(ifp); 3177 lastref = 0; 3178 ifma = if_findmulti(ifp, sa); 3179 if (ifma != NULL) 3180 lastref = if_delmulti_locked(ifp, ifma, 0); 3181 IF_ADDR_WUNLOCK(ifp); 3182 3183 if (ifma == NULL) 3184 return (ENOENT); 3185 3186 if (lastref && ifp->if_ioctl != NULL) { 3187 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3188 } 3189 3190 return (0); 3191 } 3192 3193 /* 3194 * Delete all multicast group membership for an interface. 3195 * Should be used to quickly flush all multicast filters. 3196 */ 3197 void 3198 if_delallmulti(struct ifnet *ifp) 3199 { 3200 struct ifmultiaddr *ifma; 3201 struct ifmultiaddr *next; 3202 3203 IF_ADDR_WLOCK(ifp); 3204 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3205 if_delmulti_locked(ifp, ifma, 0); 3206 IF_ADDR_WUNLOCK(ifp); 3207 } 3208 3209 /* 3210 * Delete a multicast group membership by group membership pointer. 3211 * Network-layer protocol domains must use this routine. 3212 * 3213 * It is safe to call this routine if the ifp disappeared. 3214 */ 3215 void 3216 if_delmulti_ifma(struct ifmultiaddr *ifma) 3217 { 3218 struct ifnet *ifp; 3219 int lastref; 3220 3221 ifp = ifma->ifma_ifp; 3222 #ifdef DIAGNOSTIC 3223 if (ifp == NULL) { 3224 printf("%s: ifma_ifp seems to be detached\n", __func__); 3225 } else { 3226 struct ifnet *oifp; 3227 3228 IFNET_RLOCK_NOSLEEP(); 3229 TAILQ_FOREACH(oifp, &V_ifnet, if_link) 3230 if (ifp == oifp) 3231 break; 3232 if (ifp != oifp) { 3233 printf("%s: ifnet %p disappeared\n", __func__, ifp); 3234 ifp = NULL; 3235 } 3236 IFNET_RUNLOCK_NOSLEEP(); 3237 } 3238 #endif 3239 /* 3240 * If and only if the ifnet instance exists: Acquire the address lock. 3241 */ 3242 if (ifp != NULL) 3243 IF_ADDR_WLOCK(ifp); 3244 3245 lastref = if_delmulti_locked(ifp, ifma, 0); 3246 3247 if (ifp != NULL) { 3248 /* 3249 * If and only if the ifnet instance exists: 3250 * Release the address lock. 3251 * If the group was left: update the hardware hash filter. 3252 */ 3253 IF_ADDR_WUNLOCK(ifp); 3254 if (lastref && ifp->if_ioctl != NULL) { 3255 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3256 } 3257 } 3258 } 3259 3260 /* 3261 * Perform deletion of network-layer and/or link-layer multicast address. 3262 * 3263 * Return 0 if the reference count was decremented. 3264 * Return 1 if the final reference was released, indicating that the 3265 * hardware hash filter should be reprogrammed. 3266 */ 3267 static int 3268 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3269 { 3270 struct ifmultiaddr *ll_ifma; 3271 3272 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3273 KASSERT(ifma->ifma_ifp == ifp, 3274 ("%s: inconsistent ifp %p", __func__, ifp)); 3275 IF_ADDR_WLOCK_ASSERT(ifp); 3276 } 3277 3278 ifp = ifma->ifma_ifp; 3279 3280 /* 3281 * If the ifnet is detaching, null out references to ifnet, 3282 * so that upper protocol layers will notice, and not attempt 3283 * to obtain locks for an ifnet which no longer exists. The 3284 * routing socket announcement must happen before the ifnet 3285 * instance is detached from the system. 3286 */ 3287 if (detaching) { 3288 #ifdef DIAGNOSTIC 3289 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3290 #endif 3291 /* 3292 * ifp may already be nulled out if we are being reentered 3293 * to delete the ll_ifma. 3294 */ 3295 if (ifp != NULL) { 3296 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3297 ifma->ifma_ifp = NULL; 3298 } 3299 } 3300 3301 if (--ifma->ifma_refcount > 0) 3302 return 0; 3303 3304 /* 3305 * If this ifma is a network-layer ifma, a link-layer ifma may 3306 * have been associated with it. Release it first if so. 3307 */ 3308 ll_ifma = ifma->ifma_llifma; 3309 if (ll_ifma != NULL) { 3310 KASSERT(ifma->ifma_lladdr != NULL, 3311 ("%s: llifma w/o lladdr", __func__)); 3312 if (detaching) 3313 ll_ifma->ifma_ifp = NULL; /* XXX */ 3314 if (--ll_ifma->ifma_refcount == 0) { 3315 if (ifp != NULL) { 3316 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, 3317 ifma_link); 3318 } 3319 if_freemulti(ll_ifma); 3320 } 3321 } 3322 3323 if (ifp != NULL) 3324 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 3325 3326 if_freemulti(ifma); 3327 3328 /* 3329 * The last reference to this instance of struct ifmultiaddr 3330 * was released; the hardware should be notified of this change. 3331 */ 3332 return 1; 3333 } 3334 3335 /* 3336 * Set the link layer address on an interface. 3337 * 3338 * At this time we only support certain types of interfaces, 3339 * and we don't allow the length of the address to change. 3340 */ 3341 int 3342 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3343 { 3344 struct sockaddr_dl *sdl; 3345 struct ifaddr *ifa; 3346 struct ifreq ifr; 3347 3348 IF_ADDR_RLOCK(ifp); 3349 ifa = ifp->if_addr; 3350 if (ifa == NULL) { 3351 IF_ADDR_RUNLOCK(ifp); 3352 return (EINVAL); 3353 } 3354 ifa_ref(ifa); 3355 IF_ADDR_RUNLOCK(ifp); 3356 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3357 if (sdl == NULL) { 3358 ifa_free(ifa); 3359 return (EINVAL); 3360 } 3361 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3362 ifa_free(ifa); 3363 return (EINVAL); 3364 } 3365 switch (ifp->if_type) { 3366 case IFT_ETHER: 3367 case IFT_FDDI: 3368 case IFT_XETHER: 3369 case IFT_ISO88025: 3370 case IFT_L2VLAN: 3371 case IFT_BRIDGE: 3372 case IFT_ARCNET: 3373 case IFT_IEEE8023ADLAG: 3374 case IFT_IEEE80211: 3375 bcopy(lladdr, LLADDR(sdl), len); 3376 ifa_free(ifa); 3377 break; 3378 default: 3379 ifa_free(ifa); 3380 return (ENODEV); 3381 } 3382 3383 /* 3384 * If the interface is already up, we need 3385 * to re-init it in order to reprogram its 3386 * address filter. 3387 */ 3388 if ((ifp->if_flags & IFF_UP) != 0) { 3389 if (ifp->if_ioctl) { 3390 ifp->if_flags &= ~IFF_UP; 3391 ifr.ifr_flags = ifp->if_flags & 0xffff; 3392 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3393 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3394 ifp->if_flags |= IFF_UP; 3395 ifr.ifr_flags = ifp->if_flags & 0xffff; 3396 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3397 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3398 } 3399 #ifdef INET 3400 /* 3401 * Also send gratuitous ARPs to notify other nodes about 3402 * the address change. 3403 */ 3404 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3405 if (ifa->ifa_addr->sa_family == AF_INET) 3406 arp_ifinit(ifp, ifa); 3407 } 3408 #endif 3409 } 3410 return (0); 3411 } 3412 3413 /* 3414 * The name argument must be a pointer to storage which will last as 3415 * long as the interface does. For physical devices, the result of 3416 * device_get_name(dev) is a good choice and for pseudo-devices a 3417 * static string works well. 3418 */ 3419 void 3420 if_initname(struct ifnet *ifp, const char *name, int unit) 3421 { 3422 ifp->if_dname = name; 3423 ifp->if_dunit = unit; 3424 if (unit != IF_DUNIT_NONE) 3425 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3426 else 3427 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3428 } 3429 3430 int 3431 if_printf(struct ifnet *ifp, const char * fmt, ...) 3432 { 3433 va_list ap; 3434 int retval; 3435 3436 retval = printf("%s: ", ifp->if_xname); 3437 va_start(ap, fmt); 3438 retval += vprintf(fmt, ap); 3439 va_end(ap); 3440 return (retval); 3441 } 3442 3443 void 3444 if_start(struct ifnet *ifp) 3445 { 3446 3447 (*(ifp)->if_start)(ifp); 3448 } 3449 3450 /* 3451 * Backwards compatibility interface for drivers 3452 * that have not implemented it 3453 */ 3454 static int 3455 if_transmit(struct ifnet *ifp, struct mbuf *m) 3456 { 3457 int error; 3458 3459 IFQ_HANDOFF(ifp, m, error); 3460 return (error); 3461 } 3462 3463 static void 3464 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3465 { 3466 3467 m_freem(m); 3468 } 3469 3470 int 3471 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3472 { 3473 int active = 0; 3474 3475 IF_LOCK(ifq); 3476 if (_IF_QFULL(ifq)) { 3477 IF_UNLOCK(ifq); 3478 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 3479 m_freem(m); 3480 return (0); 3481 } 3482 if (ifp != NULL) { 3483 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 3484 if (m->m_flags & (M_BCAST|M_MCAST)) 3485 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 3486 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3487 } 3488 _IF_ENQUEUE(ifq, m); 3489 IF_UNLOCK(ifq); 3490 if (ifp != NULL && !active) 3491 (*(ifp)->if_start)(ifp); 3492 return (1); 3493 } 3494 3495 void 3496 if_register_com_alloc(u_char type, 3497 if_com_alloc_t *a, if_com_free_t *f) 3498 { 3499 3500 KASSERT(if_com_alloc[type] == NULL, 3501 ("if_register_com_alloc: %d already registered", type)); 3502 KASSERT(if_com_free[type] == NULL, 3503 ("if_register_com_alloc: %d free already registered", type)); 3504 3505 if_com_alloc[type] = a; 3506 if_com_free[type] = f; 3507 } 3508 3509 void 3510 if_deregister_com_alloc(u_char type) 3511 { 3512 3513 KASSERT(if_com_alloc[type] != NULL, 3514 ("if_deregister_com_alloc: %d not registered", type)); 3515 KASSERT(if_com_free[type] != NULL, 3516 ("if_deregister_com_alloc: %d free not registered", type)); 3517 if_com_alloc[type] = NULL; 3518 if_com_free[type] = NULL; 3519 } 3520 3521 /* API for driver access to network stack owned ifnet.*/ 3522 uint64_t 3523 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 3524 { 3525 uint64_t oldbrate; 3526 3527 oldbrate = ifp->if_baudrate; 3528 ifp->if_baudrate = baudrate; 3529 return (oldbrate); 3530 } 3531 3532 uint64_t 3533 if_getbaudrate(if_t ifp) 3534 { 3535 3536 return (((struct ifnet *)ifp)->if_baudrate); 3537 } 3538 3539 int 3540 if_setcapabilities(if_t ifp, int capabilities) 3541 { 3542 ((struct ifnet *)ifp)->if_capabilities = capabilities; 3543 return (0); 3544 } 3545 3546 int 3547 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 3548 { 3549 ((struct ifnet *)ifp)->if_capabilities |= setbit; 3550 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 3551 3552 return (0); 3553 } 3554 3555 int 3556 if_getcapabilities(if_t ifp) 3557 { 3558 return ((struct ifnet *)ifp)->if_capabilities; 3559 } 3560 3561 int 3562 if_setcapenable(if_t ifp, int capabilities) 3563 { 3564 ((struct ifnet *)ifp)->if_capenable = capabilities; 3565 return (0); 3566 } 3567 3568 int 3569 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 3570 { 3571 if(setcap) 3572 ((struct ifnet *)ifp)->if_capenable |= setcap; 3573 if(clearcap) 3574 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 3575 3576 return (0); 3577 } 3578 3579 const char * 3580 if_getdname(if_t ifp) 3581 { 3582 return ((struct ifnet *)ifp)->if_dname; 3583 } 3584 3585 int 3586 if_togglecapenable(if_t ifp, int togglecap) 3587 { 3588 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 3589 return (0); 3590 } 3591 3592 int 3593 if_getcapenable(if_t ifp) 3594 { 3595 return ((struct ifnet *)ifp)->if_capenable; 3596 } 3597 3598 /* 3599 * This is largely undesirable because it ties ifnet to a device, but does 3600 * provide flexiblity for an embedded product vendor. Should be used with 3601 * the understanding that it violates the interface boundaries, and should be 3602 * a last resort only. 3603 */ 3604 int 3605 if_setdev(if_t ifp, void *dev) 3606 { 3607 return (0); 3608 } 3609 3610 int 3611 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 3612 { 3613 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 3614 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 3615 3616 return (0); 3617 } 3618 3619 int 3620 if_getdrvflags(if_t ifp) 3621 { 3622 return ((struct ifnet *)ifp)->if_drv_flags; 3623 } 3624 3625 int 3626 if_setdrvflags(if_t ifp, int flags) 3627 { 3628 ((struct ifnet *)ifp)->if_drv_flags = flags; 3629 return (0); 3630 } 3631 3632 3633 int 3634 if_setflags(if_t ifp, int flags) 3635 { 3636 ((struct ifnet *)ifp)->if_flags = flags; 3637 return (0); 3638 } 3639 3640 int 3641 if_setflagbits(if_t ifp, int set, int clear) 3642 { 3643 ((struct ifnet *)ifp)->if_flags |= set; 3644 ((struct ifnet *)ifp)->if_flags &= ~clear; 3645 3646 return (0); 3647 } 3648 3649 int 3650 if_getflags(if_t ifp) 3651 { 3652 return ((struct ifnet *)ifp)->if_flags; 3653 } 3654 3655 int 3656 if_clearhwassist(if_t ifp) 3657 { 3658 ((struct ifnet *)ifp)->if_hwassist = 0; 3659 return (0); 3660 } 3661 3662 int 3663 if_sethwassistbits(if_t ifp, int toset, int toclear) 3664 { 3665 ((struct ifnet *)ifp)->if_hwassist |= toset; 3666 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 3667 3668 return (0); 3669 } 3670 3671 int 3672 if_sethwassist(if_t ifp, int hwassist_bit) 3673 { 3674 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 3675 return (0); 3676 } 3677 3678 int 3679 if_gethwassist(if_t ifp) 3680 { 3681 return ((struct ifnet *)ifp)->if_hwassist; 3682 } 3683 3684 int 3685 if_setmtu(if_t ifp, int mtu) 3686 { 3687 ((struct ifnet *)ifp)->if_mtu = mtu; 3688 return (0); 3689 } 3690 3691 int 3692 if_getmtu(if_t ifp) 3693 { 3694 return ((struct ifnet *)ifp)->if_mtu; 3695 } 3696 3697 int 3698 if_getmtu_family(if_t ifp, int family) 3699 { 3700 struct domain *dp; 3701 3702 for (dp = domains; dp; dp = dp->dom_next) { 3703 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 3704 return (dp->dom_ifmtu((struct ifnet *)ifp)); 3705 } 3706 3707 return (((struct ifnet *)ifp)->if_mtu); 3708 } 3709 3710 int 3711 if_setsoftc(if_t ifp, void *softc) 3712 { 3713 ((struct ifnet *)ifp)->if_softc = softc; 3714 return (0); 3715 } 3716 3717 void * 3718 if_getsoftc(if_t ifp) 3719 { 3720 return ((struct ifnet *)ifp)->if_softc; 3721 } 3722 3723 void 3724 if_setrcvif(struct mbuf *m, if_t ifp) 3725 { 3726 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 3727 } 3728 3729 void 3730 if_setvtag(struct mbuf *m, uint16_t tag) 3731 { 3732 m->m_pkthdr.ether_vtag = tag; 3733 } 3734 3735 uint16_t 3736 if_getvtag(struct mbuf *m) 3737 { 3738 3739 return (m->m_pkthdr.ether_vtag); 3740 } 3741 3742 int 3743 if_sendq_empty(if_t ifp) 3744 { 3745 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 3746 } 3747 3748 struct ifaddr * 3749 if_getifaddr(if_t ifp) 3750 { 3751 return ((struct ifnet *)ifp)->if_addr; 3752 } 3753 3754 int 3755 if_getamcount(if_t ifp) 3756 { 3757 return ((struct ifnet *)ifp)->if_amcount; 3758 } 3759 3760 3761 int 3762 if_setsendqready(if_t ifp) 3763 { 3764 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 3765 return (0); 3766 } 3767 3768 int 3769 if_setsendqlen(if_t ifp, int tx_desc_count) 3770 { 3771 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 3772 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 3773 3774 return (0); 3775 } 3776 3777 int 3778 if_vlantrunkinuse(if_t ifp) 3779 { 3780 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 3781 } 3782 3783 int 3784 if_input(if_t ifp, struct mbuf* sendmp) 3785 { 3786 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 3787 return (0); 3788 3789 } 3790 3791 /* XXX */ 3792 #ifndef ETH_ADDR_LEN 3793 #define ETH_ADDR_LEN 6 3794 #endif 3795 3796 int 3797 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max) 3798 { 3799 struct ifmultiaddr *ifma; 3800 uint8_t *lmta = (uint8_t *)mta; 3801 int mcnt = 0; 3802 3803 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 3804 if (ifma->ifma_addr->sa_family != AF_LINK) 3805 continue; 3806 3807 if (mcnt == max) 3808 break; 3809 3810 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 3811 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 3812 mcnt++; 3813 } 3814 *cnt = mcnt; 3815 3816 return (0); 3817 } 3818 3819 int 3820 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max) 3821 { 3822 int error; 3823 3824 if_maddr_rlock(ifp); 3825 error = if_setupmultiaddr(ifp, mta, cnt, max); 3826 if_maddr_runlock(ifp); 3827 return (error); 3828 } 3829 3830 int 3831 if_multiaddr_count(if_t ifp, int max) 3832 { 3833 struct ifmultiaddr *ifma; 3834 int count; 3835 3836 count = 0; 3837 if_maddr_rlock(ifp); 3838 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 3839 if (ifma->ifma_addr->sa_family != AF_LINK) 3840 continue; 3841 count++; 3842 if (count == max) 3843 break; 3844 } 3845 if_maddr_runlock(ifp); 3846 return (count); 3847 } 3848 3849 struct mbuf * 3850 if_dequeue(if_t ifp) 3851 { 3852 struct mbuf *m; 3853 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 3854 3855 return (m); 3856 } 3857 3858 int 3859 if_sendq_prepend(if_t ifp, struct mbuf *m) 3860 { 3861 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 3862 return (0); 3863 } 3864 3865 int 3866 if_setifheaderlen(if_t ifp, int len) 3867 { 3868 ((struct ifnet *)ifp)->if_hdrlen = len; 3869 return (0); 3870 } 3871 3872 caddr_t 3873 if_getlladdr(if_t ifp) 3874 { 3875 return (IF_LLADDR((struct ifnet *)ifp)); 3876 } 3877 3878 void * 3879 if_gethandle(u_char type) 3880 { 3881 return (if_alloc(type)); 3882 } 3883 3884 void 3885 if_bpfmtap(if_t ifh, struct mbuf *m) 3886 { 3887 struct ifnet *ifp = (struct ifnet *)ifh; 3888 3889 BPF_MTAP(ifp, m); 3890 } 3891 3892 void 3893 if_etherbpfmtap(if_t ifh, struct mbuf *m) 3894 { 3895 struct ifnet *ifp = (struct ifnet *)ifh; 3896 3897 ETHER_BPF_MTAP(ifp, m); 3898 } 3899 3900 void 3901 if_vlancap(if_t ifh) 3902 { 3903 struct ifnet *ifp = (struct ifnet *)ifh; 3904 VLAN_CAPABILITIES(ifp); 3905 } 3906 3907 void 3908 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 3909 { 3910 ((struct ifnet *)ifp)->if_init = init_fn; 3911 } 3912 3913 void 3914 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 3915 { 3916 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 3917 } 3918 3919 void 3920 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 3921 { 3922 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 3923 } 3924 3925 void 3926 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 3927 { 3928 ((struct ifnet *)ifp)->if_transmit = start_fn; 3929 } 3930 3931 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 3932 { 3933 ((struct ifnet *)ifp)->if_qflush = flush_fn; 3934 3935 } 3936 3937 void 3938 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 3939 { 3940 3941 ifp->if_get_counter = fn; 3942 } 3943 3944 /* Revisit these - These are inline functions originally. */ 3945 int 3946 drbr_inuse_drv(if_t ifh, struct buf_ring *br) 3947 { 3948 return drbr_inuse(ifh, br); 3949 } 3950 3951 struct mbuf* 3952 drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 3953 { 3954 return drbr_dequeue(ifh, br); 3955 } 3956 3957 int 3958 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 3959 { 3960 return drbr_needs_enqueue(ifh, br); 3961 } 3962 3963 int 3964 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 3965 { 3966 return drbr_enqueue(ifh, br, m); 3967 3968 } 3969