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