1 /* $NetBSD: if.c,v 1.473 2020/02/21 00:26:23 joerg Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by William Studenmund and Jason R. Thorpe. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 34 * All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the project nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 */ 60 61 /* 62 * Copyright (c) 1980, 1986, 1993 63 * The Regents of the University of California. All rights reserved. 64 * 65 * Redistribution and use in source and binary forms, with or without 66 * modification, are permitted provided that the following conditions 67 * are met: 68 * 1. Redistributions of source code must retain the above copyright 69 * notice, this list of conditions and the following disclaimer. 70 * 2. Redistributions in binary form must reproduce the above copyright 71 * notice, this list of conditions and the following disclaimer in the 72 * documentation and/or other materials provided with the distribution. 73 * 3. Neither the name of the University nor the names of its contributors 74 * may be used to endorse or promote products derived from this software 75 * without specific prior written permission. 76 * 77 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 78 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 79 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 80 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 81 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 82 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 83 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 84 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 85 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 86 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 87 * SUCH DAMAGE. 88 * 89 * @(#)if.c 8.5 (Berkeley) 1/9/95 90 */ 91 92 #include <sys/cdefs.h> 93 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.473 2020/02/21 00:26:23 joerg Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 #include "opt_ipsec.h" 98 #include "opt_atalk.h" 99 #include "opt_wlan.h" 100 #include "opt_net_mpsafe.h" 101 #include "opt_mrouting.h" 102 #endif 103 104 #include <sys/param.h> 105 #include <sys/mbuf.h> 106 #include <sys/systm.h> 107 #include <sys/callout.h> 108 #include <sys/proc.h> 109 #include <sys/socket.h> 110 #include <sys/socketvar.h> 111 #include <sys/domain.h> 112 #include <sys/protosw.h> 113 #include <sys/kernel.h> 114 #include <sys/ioctl.h> 115 #include <sys/sysctl.h> 116 #include <sys/syslog.h> 117 #include <sys/kauth.h> 118 #include <sys/kmem.h> 119 #include <sys/xcall.h> 120 #include <sys/cpu.h> 121 #include <sys/intr.h> 122 #include <sys/module_hook.h> 123 #include <sys/compat_stub.h> 124 #include <sys/msan.h> 125 126 #include <net/if.h> 127 #include <net/if_dl.h> 128 #include <net/if_ether.h> 129 #include <net/if_media.h> 130 #include <net80211/ieee80211.h> 131 #include <net80211/ieee80211_ioctl.h> 132 #include <net/if_types.h> 133 #include <net/route.h> 134 #include <net/netisr.h> 135 #include <sys/module.h> 136 #ifdef NETATALK 137 #include <netatalk/at_extern.h> 138 #include <netatalk/at.h> 139 #endif 140 #include <net/pfil.h> 141 #include <netinet/in.h> 142 #include <netinet/in_var.h> 143 #include <netinet/ip_encap.h> 144 #include <net/bpf.h> 145 146 #ifdef INET6 147 #include <netinet6/in6_var.h> 148 #include <netinet6/nd6.h> 149 #endif 150 151 #include "ether.h" 152 153 #include "bridge.h" 154 #if NBRIDGE > 0 155 #include <net/if_bridgevar.h> 156 #endif 157 158 #include "carp.h" 159 #if NCARP > 0 160 #include <netinet/ip_carp.h> 161 #endif 162 163 #include <compat/sys/sockio.h> 164 165 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 166 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 167 168 /* 169 * Global list of interfaces. 170 */ 171 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 172 struct ifnet_head ifnet_list; 173 174 struct pslist_head ifnet_pslist; 175 static ifnet_t ** ifindex2ifnet = NULL; 176 static u_int if_index = 1; 177 static size_t if_indexlim = 0; 178 static uint64_t index_gen; 179 /* Mutex to protect the above objects. */ 180 kmutex_t ifnet_mtx __cacheline_aligned; 181 static struct psref_class *ifnet_psref_class __read_mostly; 182 static pserialize_t ifnet_psz; 183 static struct workqueue *ifnet_link_state_wq __read_mostly; 184 185 static kmutex_t if_clone_mtx; 186 187 struct ifnet *lo0ifp; 188 int ifqmaxlen = IFQ_MAXLEN; 189 190 struct psref_class *ifa_psref_class __read_mostly; 191 192 static int if_delroute_matcher(struct rtentry *, void *); 193 194 static bool if_is_unit(const char *); 195 static struct if_clone *if_clone_lookup(const char *, int *); 196 197 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 198 static int if_cloners_count; 199 200 /* Packet filtering hook for interfaces. */ 201 pfil_head_t * if_pfil __read_mostly; 202 203 static kauth_listener_t if_listener; 204 205 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 206 static void if_detach_queues(struct ifnet *, struct ifqueue *); 207 static void sysctl_sndq_setup(struct sysctllog **, const char *, 208 struct ifaltq *); 209 static void if_slowtimo(void *); 210 static void if_attachdomain1(struct ifnet *); 211 static int ifconf(u_long, void *); 212 static int if_transmit(struct ifnet *, struct mbuf *); 213 static int if_clone_create(const char *); 214 static int if_clone_destroy(const char *); 215 static void if_link_state_change_work(struct work *, void *); 216 static void if_up_locked(struct ifnet *); 217 static void _if_down(struct ifnet *); 218 static void if_down_deactivated(struct ifnet *); 219 220 struct if_percpuq { 221 struct ifnet *ipq_ifp; 222 void *ipq_si; 223 struct percpu *ipq_ifqs; /* struct ifqueue */ 224 }; 225 226 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 227 228 static void if_percpuq_drops(void *, void *, struct cpu_info *); 229 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 230 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 231 struct if_percpuq *); 232 233 struct if_deferred_start { 234 struct ifnet *ids_ifp; 235 void (*ids_if_start)(struct ifnet *); 236 void *ids_si; 237 }; 238 239 static void if_deferred_start_softint(void *); 240 static void if_deferred_start_common(struct ifnet *); 241 static void if_deferred_start_destroy(struct ifnet *); 242 243 #if defined(INET) || defined(INET6) 244 static void sysctl_net_pktq_setup(struct sysctllog **, int); 245 #endif 246 247 /* 248 * Hook for if_vlan - needed by if_agr 249 */ 250 struct if_vlan_vlan_input_hook_t if_vlan_vlan_input_hook; 251 252 static void if_sysctl_setup(struct sysctllog **); 253 254 static int 255 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 256 void *arg0, void *arg1, void *arg2, void *arg3) 257 { 258 int result; 259 enum kauth_network_req req; 260 261 result = KAUTH_RESULT_DEFER; 262 req = (enum kauth_network_req)(uintptr_t)arg1; 263 264 if (action != KAUTH_NETWORK_INTERFACE) 265 return result; 266 267 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 268 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 269 result = KAUTH_RESULT_ALLOW; 270 271 return result; 272 } 273 274 /* 275 * Network interface utility routines. 276 * 277 * Routines with ifa_ifwith* names take sockaddr *'s as 278 * parameters. 279 */ 280 void 281 ifinit(void) 282 { 283 284 #if (defined(INET) || defined(INET6)) 285 encapinit(); 286 #endif 287 288 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 289 if_listener_cb, NULL); 290 291 /* interfaces are available, inform socket code */ 292 ifioctl = doifioctl; 293 } 294 295 /* 296 * XXX Initialization before configure(). 297 * XXX hack to get pfil_add_hook working in autoconf. 298 */ 299 void 300 ifinit1(void) 301 { 302 int error __diagused; 303 304 #ifdef NET_MPSAFE 305 printf("NET_MPSAFE enabled\n"); 306 #endif 307 308 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 309 310 TAILQ_INIT(&ifnet_list); 311 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 312 ifnet_psz = pserialize_create(); 313 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 314 ifa_psref_class = psref_class_create("ifa", IPL_SOFTNET); 315 error = workqueue_create(&ifnet_link_state_wq, "iflnkst", 316 if_link_state_change_work, NULL, PRI_SOFTNET, IPL_NET, 317 WQ_MPSAFE); 318 KASSERT(error == 0); 319 PSLIST_INIT(&ifnet_pslist); 320 321 if_indexlim = 8; 322 323 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 324 KASSERT(if_pfil != NULL); 325 326 #if NETHER > 0 || defined(NETATALK) || defined(WLAN) 327 etherinit(); 328 #endif 329 } 330 331 /* XXX must be after domaininit() */ 332 void 333 ifinit_post(void) 334 { 335 336 if_sysctl_setup(NULL); 337 } 338 339 ifnet_t * 340 if_alloc(u_char type) 341 { 342 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 343 } 344 345 void 346 if_free(ifnet_t *ifp) 347 { 348 kmem_free(ifp, sizeof(ifnet_t)); 349 } 350 351 void 352 if_initname(struct ifnet *ifp, const char *name, int unit) 353 { 354 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 355 "%s%d", name, unit); 356 } 357 358 /* 359 * Null routines used while an interface is going away. These routines 360 * just return an error. 361 */ 362 363 int 364 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 365 const struct sockaddr *so, const struct rtentry *rt) 366 { 367 368 return ENXIO; 369 } 370 371 void 372 if_nullinput(struct ifnet *ifp, struct mbuf *m) 373 { 374 375 /* Nothing. */ 376 } 377 378 void 379 if_nullstart(struct ifnet *ifp) 380 { 381 382 /* Nothing. */ 383 } 384 385 int 386 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 387 { 388 389 m_freem(m); 390 return ENXIO; 391 } 392 393 int 394 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 395 { 396 397 return ENXIO; 398 } 399 400 int 401 if_nullinit(struct ifnet *ifp) 402 { 403 404 return ENXIO; 405 } 406 407 void 408 if_nullstop(struct ifnet *ifp, int disable) 409 { 410 411 /* Nothing. */ 412 } 413 414 void 415 if_nullslowtimo(struct ifnet *ifp) 416 { 417 418 /* Nothing. */ 419 } 420 421 void 422 if_nulldrain(struct ifnet *ifp) 423 { 424 425 /* Nothing. */ 426 } 427 428 void 429 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 430 { 431 struct ifaddr *ifa; 432 struct sockaddr_dl *sdl; 433 434 ifp->if_addrlen = addrlen; 435 if_alloc_sadl(ifp); 436 ifa = ifp->if_dl; 437 sdl = satosdl(ifa->ifa_addr); 438 439 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 440 if (factory) { 441 KASSERT(ifp->if_hwdl == NULL); 442 ifp->if_hwdl = ifp->if_dl; 443 ifaref(ifp->if_hwdl); 444 } 445 /* TBD routing socket */ 446 } 447 448 struct ifaddr * 449 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 450 { 451 unsigned socksize, ifasize; 452 int addrlen, namelen; 453 struct sockaddr_dl *mask, *sdl; 454 struct ifaddr *ifa; 455 456 namelen = strlen(ifp->if_xname); 457 addrlen = ifp->if_addrlen; 458 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 459 ifasize = sizeof(*ifa) + 2 * socksize; 460 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 461 462 sdl = (struct sockaddr_dl *)(ifa + 1); 463 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 464 465 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 466 ifp->if_xname, namelen, NULL, addrlen); 467 mask->sdl_family = AF_LINK; 468 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 469 memset(&mask->sdl_data[0], 0xff, namelen); 470 ifa->ifa_rtrequest = link_rtrequest; 471 ifa->ifa_addr = (struct sockaddr *)sdl; 472 ifa->ifa_netmask = (struct sockaddr *)mask; 473 ifa_psref_init(ifa); 474 475 *sdlp = sdl; 476 477 return ifa; 478 } 479 480 static void 481 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 482 { 483 const struct sockaddr_dl *sdl; 484 485 ifp->if_dl = ifa; 486 ifaref(ifa); 487 sdl = satosdl(ifa->ifa_addr); 488 ifp->if_sadl = sdl; 489 } 490 491 /* 492 * Allocate the link level name for the specified interface. This 493 * is an attachment helper. It must be called after ifp->if_addrlen 494 * is initialized, which may not be the case when if_attach() is 495 * called. 496 */ 497 void 498 if_alloc_sadl(struct ifnet *ifp) 499 { 500 struct ifaddr *ifa; 501 const struct sockaddr_dl *sdl; 502 503 /* 504 * If the interface already has a link name, release it 505 * now. This is useful for interfaces that can change 506 * link types, and thus switch link names often. 507 */ 508 if (ifp->if_sadl != NULL) 509 if_free_sadl(ifp, 0); 510 511 ifa = if_dl_create(ifp, &sdl); 512 513 ifa_insert(ifp, ifa); 514 if_sadl_setrefs(ifp, ifa); 515 } 516 517 static void 518 if_deactivate_sadl(struct ifnet *ifp) 519 { 520 struct ifaddr *ifa; 521 522 KASSERT(ifp->if_dl != NULL); 523 524 ifa = ifp->if_dl; 525 526 ifp->if_sadl = NULL; 527 528 ifp->if_dl = NULL; 529 ifafree(ifa); 530 } 531 532 static void 533 if_replace_sadl(struct ifnet *ifp, struct ifaddr *ifa) 534 { 535 struct ifaddr *old; 536 537 KASSERT(ifp->if_dl != NULL); 538 539 old = ifp->if_dl; 540 541 ifaref(ifa); 542 /* XXX Update if_dl and if_sadl atomically */ 543 ifp->if_dl = ifa; 544 ifp->if_sadl = satosdl(ifa->ifa_addr); 545 546 ifafree(old); 547 } 548 549 void 550 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa0, 551 const struct sockaddr_dl *sdl) 552 { 553 int s, ss; 554 struct ifaddr *ifa; 555 int bound = curlwp_bind(); 556 557 KASSERT(ifa_held(ifa0)); 558 559 s = splsoftnet(); 560 561 if_replace_sadl(ifp, ifa0); 562 563 ss = pserialize_read_enter(); 564 IFADDR_READER_FOREACH(ifa, ifp) { 565 struct psref psref; 566 ifa_acquire(ifa, &psref); 567 pserialize_read_exit(ss); 568 569 rtinit(ifa, RTM_LLINFO_UPD, 0); 570 571 ss = pserialize_read_enter(); 572 ifa_release(ifa, &psref); 573 } 574 pserialize_read_exit(ss); 575 576 splx(s); 577 curlwp_bindx(bound); 578 } 579 580 /* 581 * Free the link level name for the specified interface. This is 582 * a detach helper. This is called from if_detach(). 583 */ 584 void 585 if_free_sadl(struct ifnet *ifp, int factory) 586 { 587 struct ifaddr *ifa; 588 int s; 589 590 if (factory && ifp->if_hwdl != NULL) { 591 ifa = ifp->if_hwdl; 592 ifp->if_hwdl = NULL; 593 ifafree(ifa); 594 } 595 596 ifa = ifp->if_dl; 597 if (ifa == NULL) { 598 KASSERT(ifp->if_sadl == NULL); 599 return; 600 } 601 602 KASSERT(ifp->if_sadl != NULL); 603 604 s = splsoftnet(); 605 KASSERT(ifa->ifa_addr->sa_family == AF_LINK); 606 ifa_remove(ifp, ifa); 607 if_deactivate_sadl(ifp); 608 splx(s); 609 } 610 611 static void 612 if_getindex(ifnet_t *ifp) 613 { 614 bool hitlimit = false; 615 616 ifp->if_index_gen = index_gen++; 617 618 ifp->if_index = if_index; 619 if (ifindex2ifnet == NULL) { 620 if_index++; 621 goto skip; 622 } 623 while (if_byindex(ifp->if_index)) { 624 /* 625 * If we hit USHRT_MAX, we skip back to 0 since 626 * there are a number of places where the value 627 * of if_index or if_index itself is compared 628 * to or stored in an unsigned short. By 629 * jumping back, we won't botch those assignments 630 * or comparisons. 631 */ 632 if (++if_index == 0) { 633 if_index = 1; 634 } else if (if_index == USHRT_MAX) { 635 /* 636 * However, if we have to jump back to 637 * zero *twice* without finding an empty 638 * slot in ifindex2ifnet[], then there 639 * there are too many (>65535) interfaces. 640 */ 641 if (hitlimit) { 642 panic("too many interfaces"); 643 } 644 hitlimit = true; 645 if_index = 1; 646 } 647 ifp->if_index = if_index; 648 } 649 skip: 650 /* 651 * ifindex2ifnet is indexed by if_index. Since if_index will 652 * grow dynamically, it should grow too. 653 */ 654 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 655 size_t m, n, oldlim; 656 void *q; 657 658 oldlim = if_indexlim; 659 while (ifp->if_index >= if_indexlim) 660 if_indexlim <<= 1; 661 662 /* grow ifindex2ifnet */ 663 m = oldlim * sizeof(struct ifnet *); 664 n = if_indexlim * sizeof(struct ifnet *); 665 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO); 666 if (ifindex2ifnet != NULL) { 667 memcpy(q, ifindex2ifnet, m); 668 free(ifindex2ifnet, M_IFADDR); 669 } 670 ifindex2ifnet = (struct ifnet **)q; 671 } 672 ifindex2ifnet[ifp->if_index] = ifp; 673 } 674 675 /* 676 * Initialize an interface and assign an index for it. 677 * 678 * It must be called prior to a device specific attach routine 679 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 680 * and be followed by if_register: 681 * 682 * if_initialize(ifp); 683 * ether_ifattach(ifp, enaddr); 684 * if_register(ifp); 685 */ 686 int 687 if_initialize(ifnet_t *ifp) 688 { 689 int rv = 0; 690 691 KASSERT(if_indexlim > 0); 692 TAILQ_INIT(&ifp->if_addrlist); 693 694 /* 695 * Link level name is allocated later by a separate call to 696 * if_alloc_sadl(). 697 */ 698 699 if (ifp->if_snd.ifq_maxlen == 0) 700 ifp->if_snd.ifq_maxlen = ifqmaxlen; 701 702 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 703 704 ifp->if_link_state = LINK_STATE_UNKNOWN; 705 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 706 707 ifp->if_capenable = 0; 708 ifp->if_csum_flags_tx = 0; 709 ifp->if_csum_flags_rx = 0; 710 711 #ifdef ALTQ 712 ifp->if_snd.altq_type = 0; 713 ifp->if_snd.altq_disc = NULL; 714 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 715 ifp->if_snd.altq_tbr = NULL; 716 ifp->if_snd.altq_ifp = ifp; 717 #endif 718 719 IFQ_LOCK_INIT(&ifp->if_snd); 720 721 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 722 pfil_run_ifhooks(if_pfil, PFIL_IFNET_ATTACH, ifp); 723 724 IF_AFDATA_LOCK_INIT(ifp); 725 726 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 727 PSLIST_INIT(&ifp->if_addr_pslist); 728 psref_target_init(&ifp->if_psref, ifnet_psref_class); 729 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 730 LIST_INIT(&ifp->if_multiaddrs); 731 if ((rv = if_stats_init(ifp)) != 0) { 732 goto fail; 733 } 734 735 IFNET_GLOBAL_LOCK(); 736 if_getindex(ifp); 737 IFNET_GLOBAL_UNLOCK(); 738 739 return 0; 740 741 fail: 742 IF_AFDATA_LOCK_DESTROY(ifp); 743 744 pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp); 745 (void)pfil_head_destroy(ifp->if_pfil); 746 747 IFQ_LOCK_DESTROY(&ifp->if_snd); 748 749 return rv; 750 } 751 752 /* 753 * Register an interface to the list of "active" interfaces. 754 */ 755 void 756 if_register(ifnet_t *ifp) 757 { 758 /* 759 * If the driver has not supplied its own if_ioctl, then 760 * supply the default. 761 */ 762 if (ifp->if_ioctl == NULL) 763 ifp->if_ioctl = ifioctl_common; 764 765 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 766 767 if (!STAILQ_EMPTY(&domains)) 768 if_attachdomain1(ifp); 769 770 /* Announce the interface. */ 771 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 772 773 if (ifp->if_slowtimo != NULL) { 774 ifp->if_slowtimo_ch = 775 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 776 callout_init(ifp->if_slowtimo_ch, 0); 777 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 778 if_slowtimo(ifp); 779 } 780 781 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 782 ifp->if_transmit = if_transmit; 783 784 IFNET_GLOBAL_LOCK(); 785 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 786 IFNET_WRITER_INSERT_TAIL(ifp); 787 IFNET_GLOBAL_UNLOCK(); 788 } 789 790 /* 791 * The if_percpuq framework 792 * 793 * It allows network device drivers to execute the network stack 794 * in softint (so called softint-based if_input). It utilizes 795 * softint and percpu ifqueue. It doesn't distribute any packets 796 * between CPUs, unlike pktqueue(9). 797 * 798 * Currently we support two options for device drivers to apply the framework: 799 * - Use it implicitly with less changes 800 * - If you use if_attach in driver's _attach function and if_input in 801 * driver's Rx interrupt handler, a packet is queued and a softint handles 802 * the packet implicitly 803 * - Use it explicitly in each driver (recommended) 804 * - You can use if_percpuq_* directly in your driver 805 * - In this case, you need to allocate struct if_percpuq in driver's softc 806 * - See wm(4) as a reference implementation 807 */ 808 809 static void 810 if_percpuq_softint(void *arg) 811 { 812 struct if_percpuq *ipq = arg; 813 struct ifnet *ifp = ipq->ipq_ifp; 814 struct mbuf *m; 815 816 while ((m = if_percpuq_dequeue(ipq)) != NULL) { 817 if_statinc(ifp, if_ipackets); 818 bpf_mtap(ifp, m, BPF_D_IN); 819 820 ifp->_if_input(ifp, m); 821 } 822 } 823 824 static void 825 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 826 { 827 struct ifqueue *const ifq = p; 828 829 memset(ifq, 0, sizeof(*ifq)); 830 ifq->ifq_maxlen = IFQ_MAXLEN; 831 } 832 833 struct if_percpuq * 834 if_percpuq_create(struct ifnet *ifp) 835 { 836 struct if_percpuq *ipq; 837 u_int flags = SOFTINT_NET; 838 839 flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0; 840 841 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 842 ipq->ipq_ifp = ifp; 843 ipq->ipq_si = softint_establish(flags, if_percpuq_softint, ipq); 844 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 845 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 846 847 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 848 849 return ipq; 850 } 851 852 static struct mbuf * 853 if_percpuq_dequeue(struct if_percpuq *ipq) 854 { 855 struct mbuf *m; 856 struct ifqueue *ifq; 857 int s; 858 859 s = splnet(); 860 ifq = percpu_getref(ipq->ipq_ifqs); 861 IF_DEQUEUE(ifq, m); 862 percpu_putref(ipq->ipq_ifqs); 863 splx(s); 864 865 return m; 866 } 867 868 static void 869 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 870 { 871 struct ifqueue *const ifq = p; 872 873 IF_PURGE(ifq); 874 } 875 876 void 877 if_percpuq_destroy(struct if_percpuq *ipq) 878 { 879 880 /* if_detach may already destroy it */ 881 if (ipq == NULL) 882 return; 883 884 softint_disestablish(ipq->ipq_si); 885 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 886 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 887 kmem_free(ipq, sizeof(*ipq)); 888 } 889 890 void 891 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 892 { 893 struct ifqueue *ifq; 894 int s; 895 896 KASSERT(ipq != NULL); 897 898 s = splnet(); 899 ifq = percpu_getref(ipq->ipq_ifqs); 900 if (IF_QFULL(ifq)) { 901 IF_DROP(ifq); 902 percpu_putref(ipq->ipq_ifqs); 903 m_freem(m); 904 goto out; 905 } 906 IF_ENQUEUE(ifq, m); 907 percpu_putref(ipq->ipq_ifqs); 908 909 softint_schedule(ipq->ipq_si); 910 out: 911 splx(s); 912 } 913 914 static void 915 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 916 { 917 struct ifqueue *const ifq = p; 918 int *sum = arg; 919 920 *sum += ifq->ifq_drops; 921 } 922 923 static int 924 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 925 { 926 struct sysctlnode node; 927 struct if_percpuq *ipq; 928 int sum = 0; 929 int error; 930 931 node = *rnode; 932 ipq = node.sysctl_data; 933 934 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 935 936 node.sysctl_data = ∑ 937 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 938 if (error != 0 || newp == NULL) 939 return error; 940 941 return 0; 942 } 943 944 static void 945 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 946 struct if_percpuq *ipq) 947 { 948 const struct sysctlnode *cnode, *rnode; 949 950 if (sysctl_createv(clog, 0, NULL, &rnode, 951 CTLFLAG_PERMANENT, 952 CTLTYPE_NODE, "interfaces", 953 SYSCTL_DESCR("Per-interface controls"), 954 NULL, 0, NULL, 0, 955 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 956 goto bad; 957 958 if (sysctl_createv(clog, 0, &rnode, &rnode, 959 CTLFLAG_PERMANENT, 960 CTLTYPE_NODE, ifname, 961 SYSCTL_DESCR("Interface controls"), 962 NULL, 0, NULL, 0, 963 CTL_CREATE, CTL_EOL) != 0) 964 goto bad; 965 966 if (sysctl_createv(clog, 0, &rnode, &rnode, 967 CTLFLAG_PERMANENT, 968 CTLTYPE_NODE, "rcvq", 969 SYSCTL_DESCR("Interface input queue controls"), 970 NULL, 0, NULL, 0, 971 CTL_CREATE, CTL_EOL) != 0) 972 goto bad; 973 974 #ifdef NOTYET 975 /* XXX Should show each per-CPU queue length? */ 976 if (sysctl_createv(clog, 0, &rnode, &rnode, 977 CTLFLAG_PERMANENT, 978 CTLTYPE_INT, "len", 979 SYSCTL_DESCR("Current input queue length"), 980 sysctl_percpuq_len, 0, NULL, 0, 981 CTL_CREATE, CTL_EOL) != 0) 982 goto bad; 983 984 if (sysctl_createv(clog, 0, &rnode, &cnode, 985 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 986 CTLTYPE_INT, "maxlen", 987 SYSCTL_DESCR("Maximum allowed input queue length"), 988 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 989 CTL_CREATE, CTL_EOL) != 0) 990 goto bad; 991 #endif 992 993 if (sysctl_createv(clog, 0, &rnode, &cnode, 994 CTLFLAG_PERMANENT, 995 CTLTYPE_INT, "drops", 996 SYSCTL_DESCR("Total packets dropped due to full input queue"), 997 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 998 CTL_CREATE, CTL_EOL) != 0) 999 goto bad; 1000 1001 return; 1002 bad: 1003 printf("%s: could not attach sysctl nodes\n", ifname); 1004 return; 1005 } 1006 1007 /* 1008 * The deferred if_start framework 1009 * 1010 * The common APIs to defer if_start to softint when if_start is requested 1011 * from a device driver running in hardware interrupt context. 1012 */ 1013 /* 1014 * Call ifp->if_start (or equivalent) in a dedicated softint for 1015 * deferred if_start. 1016 */ 1017 static void 1018 if_deferred_start_softint(void *arg) 1019 { 1020 struct if_deferred_start *ids = arg; 1021 struct ifnet *ifp = ids->ids_ifp; 1022 1023 ids->ids_if_start(ifp); 1024 } 1025 1026 /* 1027 * The default callback function for deferred if_start. 1028 */ 1029 static void 1030 if_deferred_start_common(struct ifnet *ifp) 1031 { 1032 int s; 1033 1034 s = splnet(); 1035 if_start_lock(ifp); 1036 splx(s); 1037 } 1038 1039 static inline bool 1040 if_snd_is_used(struct ifnet *ifp) 1041 { 1042 1043 return ALTQ_IS_ENABLED(&ifp->if_snd) || 1044 ifp->if_transmit == if_transmit || 1045 ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit; 1046 } 1047 1048 /* 1049 * Schedule deferred if_start. 1050 */ 1051 void 1052 if_schedule_deferred_start(struct ifnet *ifp) 1053 { 1054 1055 KASSERT(ifp->if_deferred_start != NULL); 1056 1057 if (if_snd_is_used(ifp) && IFQ_IS_EMPTY(&ifp->if_snd)) 1058 return; 1059 1060 softint_schedule(ifp->if_deferred_start->ids_si); 1061 } 1062 1063 /* 1064 * Create an instance of deferred if_start. A driver should call the function 1065 * only if the driver needs deferred if_start. Drivers can setup their own 1066 * deferred if_start function via 2nd argument. 1067 */ 1068 void 1069 if_deferred_start_init(struct ifnet *ifp, void (*func)(struct ifnet *)) 1070 { 1071 struct if_deferred_start *ids; 1072 u_int flags = SOFTINT_NET; 1073 1074 flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0; 1075 1076 ids = kmem_zalloc(sizeof(*ids), KM_SLEEP); 1077 ids->ids_ifp = ifp; 1078 ids->ids_si = softint_establish(flags, if_deferred_start_softint, ids); 1079 if (func != NULL) 1080 ids->ids_if_start = func; 1081 else 1082 ids->ids_if_start = if_deferred_start_common; 1083 1084 ifp->if_deferred_start = ids; 1085 } 1086 1087 static void 1088 if_deferred_start_destroy(struct ifnet *ifp) 1089 { 1090 1091 if (ifp->if_deferred_start == NULL) 1092 return; 1093 1094 softint_disestablish(ifp->if_deferred_start->ids_si); 1095 kmem_free(ifp->if_deferred_start, sizeof(*ifp->if_deferred_start)); 1096 ifp->if_deferred_start = NULL; 1097 } 1098 1099 /* 1100 * The common interface input routine that is called by device drivers, 1101 * which should be used only when the driver's rx handler already runs 1102 * in softint. 1103 */ 1104 void 1105 if_input(struct ifnet *ifp, struct mbuf *m) 1106 { 1107 1108 KASSERT(ifp->if_percpuq == NULL); 1109 KASSERT(!cpu_intr_p()); 1110 1111 if_statinc(ifp, if_ipackets); 1112 bpf_mtap(ifp, m, BPF_D_IN); 1113 1114 ifp->_if_input(ifp, m); 1115 } 1116 1117 /* 1118 * DEPRECATED. Use if_initialize and if_register instead. 1119 * See the above comment of if_initialize. 1120 * 1121 * Note that it implicitly enables if_percpuq to make drivers easy to 1122 * migrate softint-based if_input without much changes. If you don't 1123 * want to enable it, use if_initialize instead. 1124 */ 1125 int 1126 if_attach(ifnet_t *ifp) 1127 { 1128 int rv; 1129 1130 rv = if_initialize(ifp); 1131 if (rv != 0) 1132 return rv; 1133 1134 ifp->if_percpuq = if_percpuq_create(ifp); 1135 if_register(ifp); 1136 1137 return 0; 1138 } 1139 1140 void 1141 if_attachdomain(void) 1142 { 1143 struct ifnet *ifp; 1144 int s; 1145 int bound = curlwp_bind(); 1146 1147 s = pserialize_read_enter(); 1148 IFNET_READER_FOREACH(ifp) { 1149 struct psref psref; 1150 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 1151 pserialize_read_exit(s); 1152 if_attachdomain1(ifp); 1153 s = pserialize_read_enter(); 1154 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 1155 } 1156 pserialize_read_exit(s); 1157 curlwp_bindx(bound); 1158 } 1159 1160 static void 1161 if_attachdomain1(struct ifnet *ifp) 1162 { 1163 struct domain *dp; 1164 int s; 1165 1166 s = splsoftnet(); 1167 1168 /* address family dependent data region */ 1169 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 1170 DOMAIN_FOREACH(dp) { 1171 if (dp->dom_ifattach != NULL) 1172 ifp->if_afdata[dp->dom_family] = 1173 (*dp->dom_ifattach)(ifp); 1174 } 1175 1176 splx(s); 1177 } 1178 1179 /* 1180 * Deactivate an interface. This points all of the procedure 1181 * handles at error stubs. May be called from interrupt context. 1182 */ 1183 void 1184 if_deactivate(struct ifnet *ifp) 1185 { 1186 int s; 1187 1188 s = splsoftnet(); 1189 1190 ifp->if_output = if_nulloutput; 1191 ifp->_if_input = if_nullinput; 1192 ifp->if_start = if_nullstart; 1193 ifp->if_transmit = if_nulltransmit; 1194 ifp->if_ioctl = if_nullioctl; 1195 ifp->if_init = if_nullinit; 1196 ifp->if_stop = if_nullstop; 1197 ifp->if_slowtimo = if_nullslowtimo; 1198 ifp->if_drain = if_nulldrain; 1199 1200 /* No more packets may be enqueued. */ 1201 ifp->if_snd.ifq_maxlen = 0; 1202 1203 splx(s); 1204 } 1205 1206 bool 1207 if_is_deactivated(const struct ifnet *ifp) 1208 { 1209 1210 return ifp->if_output == if_nulloutput; 1211 } 1212 1213 void 1214 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1215 { 1216 struct ifaddr *ifa, *nifa; 1217 int s; 1218 1219 s = pserialize_read_enter(); 1220 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) { 1221 nifa = IFADDR_READER_NEXT(ifa); 1222 if (ifa->ifa_addr->sa_family != family) 1223 continue; 1224 pserialize_read_exit(s); 1225 1226 (*purgeaddr)(ifa); 1227 1228 s = pserialize_read_enter(); 1229 } 1230 pserialize_read_exit(s); 1231 } 1232 1233 #ifdef IFAREF_DEBUG 1234 static struct ifaddr **ifa_list; 1235 static int ifa_list_size; 1236 1237 /* Depends on only one if_attach runs at once */ 1238 static void 1239 if_build_ifa_list(struct ifnet *ifp) 1240 { 1241 struct ifaddr *ifa; 1242 int i; 1243 1244 KASSERT(ifa_list == NULL); 1245 KASSERT(ifa_list_size == 0); 1246 1247 IFADDR_READER_FOREACH(ifa, ifp) 1248 ifa_list_size++; 1249 1250 ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP); 1251 i = 0; 1252 IFADDR_READER_FOREACH(ifa, ifp) { 1253 ifa_list[i++] = ifa; 1254 ifaref(ifa); 1255 } 1256 } 1257 1258 static void 1259 if_check_and_free_ifa_list(struct ifnet *ifp) 1260 { 1261 int i; 1262 struct ifaddr *ifa; 1263 1264 if (ifa_list == NULL) 1265 return; 1266 1267 for (i = 0; i < ifa_list_size; i++) { 1268 char buf[64]; 1269 1270 ifa = ifa_list[i]; 1271 sockaddr_format(ifa->ifa_addr, buf, sizeof(buf)); 1272 if (ifa->ifa_refcnt > 1) { 1273 log(LOG_WARNING, 1274 "ifa(%s) still referenced (refcnt=%d)\n", 1275 buf, ifa->ifa_refcnt - 1); 1276 } else 1277 log(LOG_DEBUG, 1278 "ifa(%s) not referenced (refcnt=%d)\n", 1279 buf, ifa->ifa_refcnt - 1); 1280 ifafree(ifa); 1281 } 1282 1283 kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size); 1284 ifa_list = NULL; 1285 ifa_list_size = 0; 1286 } 1287 #endif 1288 1289 /* 1290 * Detach an interface from the list of "active" interfaces, 1291 * freeing any resources as we go along. 1292 * 1293 * NOTE: This routine must be called with a valid thread context, 1294 * as it may block. 1295 */ 1296 void 1297 if_detach(struct ifnet *ifp) 1298 { 1299 struct socket so; 1300 struct ifaddr *ifa; 1301 #ifdef IFAREF_DEBUG 1302 struct ifaddr *last_ifa = NULL; 1303 #endif 1304 struct domain *dp; 1305 const struct protosw *pr; 1306 int s, i, family, purged; 1307 1308 #ifdef IFAREF_DEBUG 1309 if_build_ifa_list(ifp); 1310 #endif 1311 /* 1312 * XXX It's kind of lame that we have to have the 1313 * XXX socket structure... 1314 */ 1315 memset(&so, 0, sizeof(so)); 1316 1317 s = splnet(); 1318 1319 sysctl_teardown(&ifp->if_sysctl_log); 1320 IFNET_LOCK(ifp); 1321 if_deactivate(ifp); 1322 IFNET_UNLOCK(ifp); 1323 1324 /* 1325 * Unlink from the list and wait for all readers to leave 1326 * from pserialize read sections. Note that we can't do 1327 * psref_target_destroy here. See below. 1328 */ 1329 IFNET_GLOBAL_LOCK(); 1330 ifindex2ifnet[ifp->if_index] = NULL; 1331 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1332 IFNET_WRITER_REMOVE(ifp); 1333 pserialize_perform(ifnet_psz); 1334 IFNET_GLOBAL_UNLOCK(); 1335 1336 if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) { 1337 ifp->if_slowtimo = NULL; 1338 callout_halt(ifp->if_slowtimo_ch, NULL); 1339 callout_destroy(ifp->if_slowtimo_ch); 1340 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1341 } 1342 if_deferred_start_destroy(ifp); 1343 1344 /* 1345 * Do an if_down() to give protocols a chance to do something. 1346 */ 1347 if_down_deactivated(ifp); 1348 1349 #ifdef ALTQ 1350 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1351 altq_disable(&ifp->if_snd); 1352 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1353 altq_detach(&ifp->if_snd); 1354 #endif 1355 1356 #if NCARP > 0 1357 /* Remove the interface from any carp group it is a part of. */ 1358 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1359 carp_ifdetach(ifp); 1360 #endif 1361 1362 /* 1363 * Rip all the addresses off the interface. This should make 1364 * all of the routes go away. 1365 * 1366 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1367 * from the list, including our "cursor", ifa. For safety, 1368 * and to honor the TAILQ abstraction, I just restart the 1369 * loop after each removal. Note that the loop will exit 1370 * when all of the remaining ifaddrs belong to the AF_LINK 1371 * family. I am counting on the historical fact that at 1372 * least one pr_usrreq in each address domain removes at 1373 * least one ifaddr. 1374 */ 1375 again: 1376 /* 1377 * At this point, no other one tries to remove ifa in the list, 1378 * so we don't need to take a lock or psref. Avoid using 1379 * IFADDR_READER_FOREACH to pass over an inspection of contract 1380 * violations of pserialize. 1381 */ 1382 IFADDR_WRITER_FOREACH(ifa, ifp) { 1383 family = ifa->ifa_addr->sa_family; 1384 #ifdef IFAREF_DEBUG 1385 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1386 ifa, family, ifa->ifa_refcnt); 1387 if (last_ifa != NULL && ifa == last_ifa) 1388 panic("if_detach: loop detected"); 1389 last_ifa = ifa; 1390 #endif 1391 if (family == AF_LINK) 1392 continue; 1393 dp = pffinddomain(family); 1394 KASSERTMSG(dp != NULL, "no domain for AF %d", family); 1395 /* 1396 * XXX These PURGEIF calls are redundant with the 1397 * purge-all-families calls below, but are left in for 1398 * now both to make a smaller change, and to avoid 1399 * unplanned interactions with clearing of 1400 * ifp->if_addrlist. 1401 */ 1402 purged = 0; 1403 for (pr = dp->dom_protosw; 1404 pr < dp->dom_protoswNPROTOSW; pr++) { 1405 so.so_proto = pr; 1406 if (pr->pr_usrreqs) { 1407 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1408 purged = 1; 1409 } 1410 } 1411 if (purged == 0) { 1412 /* 1413 * XXX What's really the best thing to do 1414 * XXX here? --thorpej@NetBSD.org 1415 */ 1416 printf("if_detach: WARNING: AF %d not purged\n", 1417 family); 1418 ifa_remove(ifp, ifa); 1419 } 1420 goto again; 1421 } 1422 1423 if_free_sadl(ifp, 1); 1424 1425 restart: 1426 IFADDR_WRITER_FOREACH(ifa, ifp) { 1427 family = ifa->ifa_addr->sa_family; 1428 KASSERT(family == AF_LINK); 1429 ifa_remove(ifp, ifa); 1430 goto restart; 1431 } 1432 1433 /* Delete stray routes from the routing table. */ 1434 for (i = 0; i <= AF_MAX; i++) 1435 rt_delete_matched_entries(i, if_delroute_matcher, ifp); 1436 1437 DOMAIN_FOREACH(dp) { 1438 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1439 { 1440 void *p = ifp->if_afdata[dp->dom_family]; 1441 if (p) { 1442 ifp->if_afdata[dp->dom_family] = NULL; 1443 (*dp->dom_ifdetach)(ifp, p); 1444 } 1445 } 1446 1447 /* 1448 * One would expect multicast memberships (INET and 1449 * INET6) on UDP sockets to be purged by the PURGEIF 1450 * calls above, but if all addresses were removed from 1451 * the interface prior to destruction, the calls will 1452 * not be made (e.g. ppp, for which pppd(8) generally 1453 * removes addresses before destroying the interface). 1454 * Because there is no invariant that multicast 1455 * memberships only exist for interfaces with IPv4 1456 * addresses, we must call PURGEIF regardless of 1457 * addresses. (Protocols which might store ifnet 1458 * pointers are marked with PR_PURGEIF.) 1459 */ 1460 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1461 so.so_proto = pr; 1462 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1463 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1464 } 1465 } 1466 1467 /* 1468 * Must be done after the above pr_purgeif because if_psref may be 1469 * still used in pr_purgeif. 1470 */ 1471 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1472 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1473 1474 pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp); 1475 (void)pfil_head_destroy(ifp->if_pfil); 1476 1477 /* Announce that the interface is gone. */ 1478 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1479 1480 IF_AFDATA_LOCK_DESTROY(ifp); 1481 1482 /* 1483 * remove packets that came from ifp, from software interrupt queues. 1484 */ 1485 DOMAIN_FOREACH(dp) { 1486 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1487 struct ifqueue *iq = dp->dom_ifqueues[i]; 1488 if (iq == NULL) 1489 break; 1490 dp->dom_ifqueues[i] = NULL; 1491 if_detach_queues(ifp, iq); 1492 } 1493 } 1494 1495 /* 1496 * IP queues have to be processed separately: net-queue barrier 1497 * ensures that the packets are dequeued while a cross-call will 1498 * ensure that the interrupts have completed. FIXME: not quite.. 1499 */ 1500 #ifdef INET 1501 pktq_barrier(ip_pktq); 1502 #endif 1503 #ifdef INET6 1504 if (in6_present) 1505 pktq_barrier(ip6_pktq); 1506 #endif 1507 xc_barrier(0); 1508 1509 if (ifp->if_percpuq != NULL) { 1510 if_percpuq_destroy(ifp->if_percpuq); 1511 ifp->if_percpuq = NULL; 1512 } 1513 1514 mutex_obj_free(ifp->if_ioctl_lock); 1515 ifp->if_ioctl_lock = NULL; 1516 mutex_obj_free(ifp->if_snd.ifq_lock); 1517 if_stats_fini(ifp); 1518 1519 splx(s); 1520 1521 #ifdef IFAREF_DEBUG 1522 if_check_and_free_ifa_list(ifp); 1523 #endif 1524 } 1525 1526 static void 1527 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1528 { 1529 struct mbuf *m, *prev, *next; 1530 1531 prev = NULL; 1532 for (m = q->ifq_head; m != NULL; m = next) { 1533 KASSERT((m->m_flags & M_PKTHDR) != 0); 1534 1535 next = m->m_nextpkt; 1536 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1537 prev = m; 1538 continue; 1539 } 1540 1541 if (prev != NULL) 1542 prev->m_nextpkt = m->m_nextpkt; 1543 else 1544 q->ifq_head = m->m_nextpkt; 1545 if (q->ifq_tail == m) 1546 q->ifq_tail = prev; 1547 q->ifq_len--; 1548 1549 m->m_nextpkt = NULL; 1550 m_freem(m); 1551 IF_DROP(q); 1552 } 1553 } 1554 1555 /* 1556 * Callback for a radix tree walk to delete all references to an 1557 * ifnet. 1558 */ 1559 static int 1560 if_delroute_matcher(struct rtentry *rt, void *v) 1561 { 1562 struct ifnet *ifp = (struct ifnet *)v; 1563 1564 if (rt->rt_ifp == ifp) 1565 return 1; 1566 else 1567 return 0; 1568 } 1569 1570 /* 1571 * Create a clone network interface. 1572 */ 1573 static int 1574 if_clone_create(const char *name) 1575 { 1576 struct if_clone *ifc; 1577 int unit; 1578 struct ifnet *ifp; 1579 struct psref psref; 1580 1581 KASSERT(mutex_owned(&if_clone_mtx)); 1582 1583 ifc = if_clone_lookup(name, &unit); 1584 if (ifc == NULL) 1585 return EINVAL; 1586 1587 ifp = if_get(name, &psref); 1588 if (ifp != NULL) { 1589 if_put(ifp, &psref); 1590 return EEXIST; 1591 } 1592 1593 return (*ifc->ifc_create)(ifc, unit); 1594 } 1595 1596 /* 1597 * Destroy a clone network interface. 1598 */ 1599 static int 1600 if_clone_destroy(const char *name) 1601 { 1602 struct if_clone *ifc; 1603 struct ifnet *ifp; 1604 struct psref psref; 1605 int error; 1606 int (*if_ioctl)(struct ifnet *, u_long, void *); 1607 1608 KASSERT(mutex_owned(&if_clone_mtx)); 1609 1610 ifc = if_clone_lookup(name, NULL); 1611 if (ifc == NULL) 1612 return EINVAL; 1613 1614 if (ifc->ifc_destroy == NULL) 1615 return EOPNOTSUPP; 1616 1617 ifp = if_get(name, &psref); 1618 if (ifp == NULL) 1619 return ENXIO; 1620 1621 /* We have to disable ioctls here */ 1622 IFNET_LOCK(ifp); 1623 if_ioctl = ifp->if_ioctl; 1624 ifp->if_ioctl = if_nullioctl; 1625 IFNET_UNLOCK(ifp); 1626 1627 /* 1628 * We cannot call ifc_destroy with holding ifp. 1629 * Releasing ifp here is safe thanks to if_clone_mtx. 1630 */ 1631 if_put(ifp, &psref); 1632 1633 error = (*ifc->ifc_destroy)(ifp); 1634 1635 if (error != 0) { 1636 /* We have to restore if_ioctl on error */ 1637 IFNET_LOCK(ifp); 1638 ifp->if_ioctl = if_ioctl; 1639 IFNET_UNLOCK(ifp); 1640 } 1641 1642 return error; 1643 } 1644 1645 static bool 1646 if_is_unit(const char *name) 1647 { 1648 1649 while (*name != '\0') { 1650 if (*name < '0' || *name > '9') 1651 return false; 1652 name++; 1653 } 1654 1655 return true; 1656 } 1657 1658 /* 1659 * Look up a network interface cloner. 1660 */ 1661 static struct if_clone * 1662 if_clone_lookup(const char *name, int *unitp) 1663 { 1664 struct if_clone *ifc; 1665 const char *cp; 1666 char *dp, ifname[IFNAMSIZ + 3]; 1667 int unit; 1668 1669 KASSERT(mutex_owned(&if_clone_mtx)); 1670 1671 strcpy(ifname, "if_"); 1672 /* separate interface name from unit */ 1673 /* TODO: search unit number from backward */ 1674 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1675 *cp && !if_is_unit(cp);) 1676 *dp++ = *cp++; 1677 1678 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1679 return NULL; /* No name or unit number */ 1680 *dp++ = '\0'; 1681 1682 again: 1683 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1684 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1685 break; 1686 } 1687 1688 if (ifc == NULL) { 1689 int error; 1690 if (*ifname == '\0') 1691 return NULL; 1692 mutex_exit(&if_clone_mtx); 1693 error = module_autoload(ifname, MODULE_CLASS_DRIVER); 1694 mutex_enter(&if_clone_mtx); 1695 if (error) 1696 return NULL; 1697 *ifname = '\0'; 1698 goto again; 1699 } 1700 1701 unit = 0; 1702 while (cp - name < IFNAMSIZ && *cp) { 1703 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1704 /* Bogus unit number. */ 1705 return NULL; 1706 } 1707 unit = (unit * 10) + (*cp++ - '0'); 1708 } 1709 1710 if (unitp != NULL) 1711 *unitp = unit; 1712 return ifc; 1713 } 1714 1715 /* 1716 * Register a network interface cloner. 1717 */ 1718 void 1719 if_clone_attach(struct if_clone *ifc) 1720 { 1721 1722 mutex_enter(&if_clone_mtx); 1723 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1724 if_cloners_count++; 1725 mutex_exit(&if_clone_mtx); 1726 } 1727 1728 /* 1729 * Unregister a network interface cloner. 1730 */ 1731 void 1732 if_clone_detach(struct if_clone *ifc) 1733 { 1734 1735 mutex_enter(&if_clone_mtx); 1736 LIST_REMOVE(ifc, ifc_list); 1737 if_cloners_count--; 1738 mutex_exit(&if_clone_mtx); 1739 } 1740 1741 /* 1742 * Provide list of interface cloners to userspace. 1743 */ 1744 int 1745 if_clone_list(int buf_count, char *buffer, int *total) 1746 { 1747 char outbuf[IFNAMSIZ], *dst; 1748 struct if_clone *ifc; 1749 int count, error = 0; 1750 1751 mutex_enter(&if_clone_mtx); 1752 *total = if_cloners_count; 1753 if ((dst = buffer) == NULL) { 1754 /* Just asking how many there are. */ 1755 goto out; 1756 } 1757 1758 if (buf_count < 0) { 1759 error = EINVAL; 1760 goto out; 1761 } 1762 1763 count = (if_cloners_count < buf_count) ? 1764 if_cloners_count : buf_count; 1765 1766 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1767 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1768 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1769 if (outbuf[sizeof(outbuf) - 1] != '\0') { 1770 error = ENAMETOOLONG; 1771 goto out; 1772 } 1773 error = copyout(outbuf, dst, sizeof(outbuf)); 1774 if (error != 0) 1775 break; 1776 } 1777 1778 out: 1779 mutex_exit(&if_clone_mtx); 1780 return error; 1781 } 1782 1783 void 1784 ifa_psref_init(struct ifaddr *ifa) 1785 { 1786 1787 psref_target_init(&ifa->ifa_psref, ifa_psref_class); 1788 } 1789 1790 void 1791 ifaref(struct ifaddr *ifa) 1792 { 1793 1794 atomic_inc_uint(&ifa->ifa_refcnt); 1795 } 1796 1797 void 1798 ifafree(struct ifaddr *ifa) 1799 { 1800 KASSERT(ifa != NULL); 1801 KASSERTMSG(ifa->ifa_refcnt > 0, "ifa_refcnt=%d", ifa->ifa_refcnt); 1802 1803 if (atomic_dec_uint_nv(&ifa->ifa_refcnt) == 0) { 1804 free(ifa, M_IFADDR); 1805 } 1806 } 1807 1808 bool 1809 ifa_is_destroying(struct ifaddr *ifa) 1810 { 1811 1812 return ISSET(ifa->ifa_flags, IFA_DESTROYING); 1813 } 1814 1815 void 1816 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1817 { 1818 1819 ifa->ifa_ifp = ifp; 1820 1821 /* 1822 * Check MP-safety for IFEF_MPSAFE drivers. 1823 * Check !IFF_RUNNING for initialization routines that normally don't 1824 * take IFNET_LOCK but it's safe because there is no competitor. 1825 * XXX there are false positive cases because IFF_RUNNING can be off on 1826 * if_stop. 1827 */ 1828 KASSERT(!if_is_mpsafe(ifp) || !ISSET(ifp->if_flags, IFF_RUNNING) || 1829 IFNET_LOCKED(ifp)); 1830 1831 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1832 IFADDR_ENTRY_INIT(ifa); 1833 IFADDR_WRITER_INSERT_TAIL(ifp, ifa); 1834 1835 ifaref(ifa); 1836 } 1837 1838 void 1839 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1840 { 1841 1842 KASSERT(ifa->ifa_ifp == ifp); 1843 /* 1844 * Check MP-safety for IFEF_MPSAFE drivers. 1845 * if_is_deactivated indicates ifa_remove is called form if_detach 1846 * where is safe even if IFNET_LOCK isn't held. 1847 */ 1848 KASSERT(!if_is_mpsafe(ifp) || if_is_deactivated(ifp) || IFNET_LOCKED(ifp)); 1849 1850 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1851 IFADDR_WRITER_REMOVE(ifa); 1852 #ifdef NET_MPSAFE 1853 IFNET_GLOBAL_LOCK(); 1854 pserialize_perform(ifnet_psz); 1855 IFNET_GLOBAL_UNLOCK(); 1856 #endif 1857 1858 #ifdef NET_MPSAFE 1859 psref_target_destroy(&ifa->ifa_psref, ifa_psref_class); 1860 #endif 1861 IFADDR_ENTRY_DESTROY(ifa); 1862 ifafree(ifa); 1863 } 1864 1865 void 1866 ifa_acquire(struct ifaddr *ifa, struct psref *psref) 1867 { 1868 1869 PSREF_DEBUG_FILL_RETURN_ADDRESS(psref); 1870 psref_acquire(psref, &ifa->ifa_psref, ifa_psref_class); 1871 } 1872 1873 void 1874 ifa_release(struct ifaddr *ifa, struct psref *psref) 1875 { 1876 1877 if (ifa == NULL) 1878 return; 1879 1880 psref_release(psref, &ifa->ifa_psref, ifa_psref_class); 1881 } 1882 1883 bool 1884 ifa_held(struct ifaddr *ifa) 1885 { 1886 1887 return psref_held(&ifa->ifa_psref, ifa_psref_class); 1888 } 1889 1890 static inline int 1891 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1892 { 1893 return sockaddr_cmp(sa1, sa2) == 0; 1894 } 1895 1896 /* 1897 * Locate an interface based on a complete address. 1898 */ 1899 /*ARGSUSED*/ 1900 struct ifaddr * 1901 ifa_ifwithaddr(const struct sockaddr *addr) 1902 { 1903 struct ifnet *ifp; 1904 struct ifaddr *ifa; 1905 1906 IFNET_READER_FOREACH(ifp) { 1907 if (if_is_deactivated(ifp)) 1908 continue; 1909 IFADDR_READER_FOREACH(ifa, ifp) { 1910 if (ifa->ifa_addr->sa_family != addr->sa_family) 1911 continue; 1912 if (equal(addr, ifa->ifa_addr)) 1913 return ifa; 1914 if ((ifp->if_flags & IFF_BROADCAST) && 1915 ifa->ifa_broadaddr && 1916 /* IP6 doesn't have broadcast */ 1917 ifa->ifa_broadaddr->sa_len != 0 && 1918 equal(ifa->ifa_broadaddr, addr)) 1919 return ifa; 1920 } 1921 } 1922 return NULL; 1923 } 1924 1925 struct ifaddr * 1926 ifa_ifwithaddr_psref(const struct sockaddr *addr, struct psref *psref) 1927 { 1928 struct ifaddr *ifa; 1929 int s = pserialize_read_enter(); 1930 1931 ifa = ifa_ifwithaddr(addr); 1932 if (ifa != NULL) 1933 ifa_acquire(ifa, psref); 1934 pserialize_read_exit(s); 1935 1936 return ifa; 1937 } 1938 1939 /* 1940 * Locate the point to point interface with a given destination address. 1941 */ 1942 /*ARGSUSED*/ 1943 struct ifaddr * 1944 ifa_ifwithdstaddr(const struct sockaddr *addr) 1945 { 1946 struct ifnet *ifp; 1947 struct ifaddr *ifa; 1948 1949 IFNET_READER_FOREACH(ifp) { 1950 if (if_is_deactivated(ifp)) 1951 continue; 1952 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1953 continue; 1954 IFADDR_READER_FOREACH(ifa, ifp) { 1955 if (ifa->ifa_addr->sa_family != addr->sa_family || 1956 ifa->ifa_dstaddr == NULL) 1957 continue; 1958 if (equal(addr, ifa->ifa_dstaddr)) 1959 return ifa; 1960 } 1961 } 1962 1963 return NULL; 1964 } 1965 1966 struct ifaddr * 1967 ifa_ifwithdstaddr_psref(const struct sockaddr *addr, struct psref *psref) 1968 { 1969 struct ifaddr *ifa; 1970 int s; 1971 1972 s = pserialize_read_enter(); 1973 ifa = ifa_ifwithdstaddr(addr); 1974 if (ifa != NULL) 1975 ifa_acquire(ifa, psref); 1976 pserialize_read_exit(s); 1977 1978 return ifa; 1979 } 1980 1981 /* 1982 * Find an interface on a specific network. If many, choice 1983 * is most specific found. 1984 */ 1985 struct ifaddr * 1986 ifa_ifwithnet(const struct sockaddr *addr) 1987 { 1988 struct ifnet *ifp; 1989 struct ifaddr *ifa, *ifa_maybe = NULL; 1990 const struct sockaddr_dl *sdl; 1991 u_int af = addr->sa_family; 1992 const char *addr_data = addr->sa_data, *cplim; 1993 1994 if (af == AF_LINK) { 1995 sdl = satocsdl(addr); 1996 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1997 ifindex2ifnet[sdl->sdl_index] && 1998 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1999 return ifindex2ifnet[sdl->sdl_index]->if_dl; 2000 } 2001 } 2002 #ifdef NETATALK 2003 if (af == AF_APPLETALK) { 2004 const struct sockaddr_at *sat, *sat2; 2005 sat = (const struct sockaddr_at *)addr; 2006 IFNET_READER_FOREACH(ifp) { 2007 if (if_is_deactivated(ifp)) 2008 continue; 2009 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 2010 if (ifa == NULL) 2011 continue; 2012 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 2013 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 2014 return ifa; /* exact match */ 2015 if (ifa_maybe == NULL) { 2016 /* else keep the if with the right range */ 2017 ifa_maybe = ifa; 2018 } 2019 } 2020 return ifa_maybe; 2021 } 2022 #endif 2023 IFNET_READER_FOREACH(ifp) { 2024 if (if_is_deactivated(ifp)) 2025 continue; 2026 IFADDR_READER_FOREACH(ifa, ifp) { 2027 const char *cp, *cp2, *cp3; 2028 2029 if (ifa->ifa_addr->sa_family != af || 2030 ifa->ifa_netmask == NULL) 2031 next: continue; 2032 cp = addr_data; 2033 cp2 = ifa->ifa_addr->sa_data; 2034 cp3 = ifa->ifa_netmask->sa_data; 2035 cplim = (const char *)ifa->ifa_netmask + 2036 ifa->ifa_netmask->sa_len; 2037 while (cp3 < cplim) { 2038 if ((*cp++ ^ *cp2++) & *cp3++) { 2039 /* want to continue for() loop */ 2040 goto next; 2041 } 2042 } 2043 if (ifa_maybe == NULL || 2044 rt_refines(ifa->ifa_netmask, 2045 ifa_maybe->ifa_netmask)) 2046 ifa_maybe = ifa; 2047 } 2048 } 2049 return ifa_maybe; 2050 } 2051 2052 struct ifaddr * 2053 ifa_ifwithnet_psref(const struct sockaddr *addr, struct psref *psref) 2054 { 2055 struct ifaddr *ifa; 2056 int s; 2057 2058 s = pserialize_read_enter(); 2059 ifa = ifa_ifwithnet(addr); 2060 if (ifa != NULL) 2061 ifa_acquire(ifa, psref); 2062 pserialize_read_exit(s); 2063 2064 return ifa; 2065 } 2066 2067 /* 2068 * Find the interface of the addresss. 2069 */ 2070 struct ifaddr * 2071 ifa_ifwithladdr(const struct sockaddr *addr) 2072 { 2073 struct ifaddr *ia; 2074 2075 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 2076 (ia = ifa_ifwithnet(addr))) 2077 return ia; 2078 return NULL; 2079 } 2080 2081 struct ifaddr * 2082 ifa_ifwithladdr_psref(const struct sockaddr *addr, struct psref *psref) 2083 { 2084 struct ifaddr *ifa; 2085 int s; 2086 2087 s = pserialize_read_enter(); 2088 ifa = ifa_ifwithladdr(addr); 2089 if (ifa != NULL) 2090 ifa_acquire(ifa, psref); 2091 pserialize_read_exit(s); 2092 2093 return ifa; 2094 } 2095 2096 /* 2097 * Find an interface using a specific address family 2098 */ 2099 struct ifaddr * 2100 ifa_ifwithaf(int af) 2101 { 2102 struct ifnet *ifp; 2103 struct ifaddr *ifa = NULL; 2104 int s; 2105 2106 s = pserialize_read_enter(); 2107 IFNET_READER_FOREACH(ifp) { 2108 if (if_is_deactivated(ifp)) 2109 continue; 2110 IFADDR_READER_FOREACH(ifa, ifp) { 2111 if (ifa->ifa_addr->sa_family == af) 2112 goto out; 2113 } 2114 } 2115 out: 2116 pserialize_read_exit(s); 2117 return ifa; 2118 } 2119 2120 /* 2121 * Find an interface address specific to an interface best matching 2122 * a given address. 2123 */ 2124 struct ifaddr * 2125 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2126 { 2127 struct ifaddr *ifa; 2128 const char *cp, *cp2, *cp3; 2129 const char *cplim; 2130 struct ifaddr *ifa_maybe = 0; 2131 u_int af = addr->sa_family; 2132 2133 if (if_is_deactivated(ifp)) 2134 return NULL; 2135 2136 if (af >= AF_MAX) 2137 return NULL; 2138 2139 IFADDR_READER_FOREACH(ifa, ifp) { 2140 if (ifa->ifa_addr->sa_family != af) 2141 continue; 2142 ifa_maybe = ifa; 2143 if (ifa->ifa_netmask == NULL) { 2144 if (equal(addr, ifa->ifa_addr) || 2145 (ifa->ifa_dstaddr && 2146 equal(addr, ifa->ifa_dstaddr))) 2147 return ifa; 2148 continue; 2149 } 2150 cp = addr->sa_data; 2151 cp2 = ifa->ifa_addr->sa_data; 2152 cp3 = ifa->ifa_netmask->sa_data; 2153 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2154 for (; cp3 < cplim; cp3++) { 2155 if ((*cp++ ^ *cp2++) & *cp3) 2156 break; 2157 } 2158 if (cp3 == cplim) 2159 return ifa; 2160 } 2161 return ifa_maybe; 2162 } 2163 2164 struct ifaddr * 2165 ifaof_ifpforaddr_psref(const struct sockaddr *addr, struct ifnet *ifp, 2166 struct psref *psref) 2167 { 2168 struct ifaddr *ifa; 2169 int s; 2170 2171 s = pserialize_read_enter(); 2172 ifa = ifaof_ifpforaddr(addr, ifp); 2173 if (ifa != NULL) 2174 ifa_acquire(ifa, psref); 2175 pserialize_read_exit(s); 2176 2177 return ifa; 2178 } 2179 2180 /* 2181 * Default action when installing a route with a Link Level gateway. 2182 * Lookup an appropriate real ifa to point to. 2183 * This should be moved to /sys/net/link.c eventually. 2184 */ 2185 void 2186 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 2187 { 2188 struct ifaddr *ifa; 2189 const struct sockaddr *dst; 2190 struct ifnet *ifp; 2191 struct psref psref; 2192 2193 if (cmd != RTM_ADD || ISSET(info->rti_flags, RTF_DONTCHANGEIFA)) 2194 return; 2195 ifp = rt->rt_ifa->ifa_ifp; 2196 dst = rt_getkey(rt); 2197 if ((ifa = ifaof_ifpforaddr_psref(dst, ifp, &psref)) != NULL) { 2198 rt_replace_ifa(rt, ifa); 2199 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2200 ifa->ifa_rtrequest(cmd, rt, info); 2201 ifa_release(ifa, &psref); 2202 } 2203 } 2204 2205 /* 2206 * bitmask macros to manage a densely packed link_state change queue. 2207 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 2208 * LINK_STATE_UP(2) we need 2 bits for each state change. 2209 * As a state change to store is 0, treat all bits set as an unset item. 2210 */ 2211 #define LQ_ITEM_BITS 2 2212 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 2213 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 2214 #define LINK_STATE_UNSET LQ_ITEM_MASK 2215 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 2216 #define LQ_STORE(q, i, v) \ 2217 do { \ 2218 (q) &= ~LQ_MASK((i)); \ 2219 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 2220 } while (0 /* CONSTCOND */) 2221 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 2222 #define LQ_POP(q, v) \ 2223 do { \ 2224 (v) = LQ_ITEM((q), 0); \ 2225 (q) >>= LQ_ITEM_BITS; \ 2226 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2227 } while (0 /* CONSTCOND */) 2228 #define LQ_PUSH(q, v) \ 2229 do { \ 2230 (q) >>= LQ_ITEM_BITS; \ 2231 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2232 } while (0 /* CONSTCOND */) 2233 #define LQ_FIND_UNSET(q, i) \ 2234 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 2235 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 2236 break; \ 2237 } 2238 2239 /* 2240 * XXX reusing (ifp)->if_snd->ifq_lock rather than having another spin mutex 2241 * for each ifnet. It doesn't matter because: 2242 * - if IFEF_MPSAFE is enabled, if_snd isn't used and lock contentions on 2243 * ifq_lock don't happen 2244 * - if IFEF_MPSAFE is disabled, there is no lock contention on ifq_lock 2245 * because if_snd, if_link_state_change and if_link_state_change_process 2246 * are all called with KERNEL_LOCK 2247 */ 2248 #define IF_LINK_STATE_CHANGE_LOCK(ifp) \ 2249 mutex_enter((ifp)->if_snd.ifq_lock) 2250 #define IF_LINK_STATE_CHANGE_UNLOCK(ifp) \ 2251 mutex_exit((ifp)->if_snd.ifq_lock) 2252 2253 static void 2254 if_link_state_change_work_schedule(struct ifnet *ifp) 2255 { 2256 if (ifp->if_link_cansched && !ifp->if_link_scheduled) { 2257 ifp->if_link_scheduled = true; 2258 workqueue_enqueue(ifnet_link_state_wq, &ifp->if_link_work, 2259 NULL); 2260 } 2261 } 2262 2263 /* 2264 * Handle a change in the interface link state and 2265 * queue notifications. 2266 */ 2267 void 2268 if_link_state_change(struct ifnet *ifp, int link_state) 2269 { 2270 int idx; 2271 2272 KASSERTMSG(if_is_link_state_changeable(ifp), 2273 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 2274 ifp->if_xname, ifp->if_extflags); 2275 2276 /* Ensure change is to a valid state */ 2277 switch (link_state) { 2278 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 2279 case LINK_STATE_DOWN: /* FALLTHROUGH */ 2280 case LINK_STATE_UP: 2281 break; 2282 default: 2283 #ifdef DEBUG 2284 printf("%s: invalid link state %d\n", 2285 ifp->if_xname, link_state); 2286 #endif 2287 return; 2288 } 2289 2290 IF_LINK_STATE_CHANGE_LOCK(ifp); 2291 2292 /* Find the last unset event in the queue. */ 2293 LQ_FIND_UNSET(ifp->if_link_queue, idx); 2294 2295 /* 2296 * Ensure link_state doesn't match the last event in the queue. 2297 * ifp->if_link_state is not checked and set here because 2298 * that would present an inconsistent picture to the system. 2299 */ 2300 if (idx != 0 && 2301 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 2302 goto out; 2303 2304 /* Handle queue overflow. */ 2305 if (idx == LQ_MAX(ifp->if_link_queue)) { 2306 uint8_t lost; 2307 2308 /* 2309 * The DOWN state must be protected from being pushed off 2310 * the queue to ensure that userland will always be 2311 * in a sane state. 2312 * Because DOWN is protected, there is no need to protect 2313 * UNKNOWN. 2314 * It should be invalid to change from any other state to 2315 * UNKNOWN anyway ... 2316 */ 2317 lost = LQ_ITEM(ifp->if_link_queue, 0); 2318 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 2319 if (lost == LINK_STATE_DOWN) { 2320 lost = LQ_ITEM(ifp->if_link_queue, 0); 2321 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 2322 } 2323 printf("%s: lost link state change %s\n", 2324 ifp->if_xname, 2325 lost == LINK_STATE_UP ? "UP" : 2326 lost == LINK_STATE_DOWN ? "DOWN" : 2327 "UNKNOWN"); 2328 } else 2329 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 2330 2331 if_link_state_change_work_schedule(ifp); 2332 2333 out: 2334 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2335 } 2336 2337 /* 2338 * Handle interface link state change notifications. 2339 */ 2340 static void 2341 if_link_state_change_process(struct ifnet *ifp, int link_state) 2342 { 2343 struct domain *dp; 2344 int s = splnet(); 2345 bool notify; 2346 2347 KASSERT(!cpu_intr_p()); 2348 2349 IF_LINK_STATE_CHANGE_LOCK(ifp); 2350 2351 /* Ensure the change is still valid. */ 2352 if (ifp->if_link_state == link_state) { 2353 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2354 splx(s); 2355 return; 2356 } 2357 2358 #ifdef DEBUG 2359 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 2360 link_state == LINK_STATE_UP ? "UP" : 2361 link_state == LINK_STATE_DOWN ? "DOWN" : 2362 "UNKNOWN", 2363 ifp->if_link_state == LINK_STATE_UP ? "UP" : 2364 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 2365 "UNKNOWN"); 2366 #endif 2367 2368 /* 2369 * When going from UNKNOWN to UP, we need to mark existing 2370 * addresses as tentative and restart DAD as we may have 2371 * erroneously not found a duplicate. 2372 * 2373 * This needs to happen before rt_ifmsg to avoid a race where 2374 * listeners would have an address and expect it to work right 2375 * away. 2376 */ 2377 notify = (link_state == LINK_STATE_UP && 2378 ifp->if_link_state == LINK_STATE_UNKNOWN); 2379 ifp->if_link_state = link_state; 2380 /* The following routines may sleep so release the spin mutex */ 2381 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2382 2383 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 2384 if (notify) { 2385 DOMAIN_FOREACH(dp) { 2386 if (dp->dom_if_link_state_change != NULL) 2387 dp->dom_if_link_state_change(ifp, 2388 LINK_STATE_DOWN); 2389 } 2390 } 2391 2392 /* Notify that the link state has changed. */ 2393 rt_ifmsg(ifp); 2394 2395 #if NCARP > 0 2396 if (ifp->if_carp) 2397 carp_carpdev_state(ifp); 2398 #endif 2399 2400 DOMAIN_FOREACH(dp) { 2401 if (dp->dom_if_link_state_change != NULL) 2402 dp->dom_if_link_state_change(ifp, link_state); 2403 } 2404 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 2405 splx(s); 2406 } 2407 2408 /* 2409 * Process the interface link state change queue. 2410 */ 2411 static void 2412 if_link_state_change_work(struct work *work, void *arg) 2413 { 2414 struct ifnet *ifp = container_of(work, struct ifnet, if_link_work); 2415 int s; 2416 uint8_t state; 2417 bool schedule; 2418 2419 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 2420 s = splnet(); 2421 2422 /* Pop a link state change from the queue and process it. */ 2423 IF_LINK_STATE_CHANGE_LOCK(ifp); 2424 ifp->if_link_scheduled = false; 2425 LQ_POP(ifp->if_link_queue, state); 2426 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2427 2428 if_link_state_change_process(ifp, state); 2429 2430 /* If there is a link state change to come, schedule it. */ 2431 IF_LINK_STATE_CHANGE_LOCK(ifp); 2432 schedule = (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET); 2433 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2434 2435 if (schedule) 2436 if_link_state_change_work_schedule(ifp); 2437 2438 splx(s); 2439 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 2440 } 2441 2442 /* 2443 * Default action when installing a local route on a point-to-point 2444 * interface. 2445 */ 2446 void 2447 p2p_rtrequest(int req, struct rtentry *rt, 2448 __unused const struct rt_addrinfo *info) 2449 { 2450 struct ifnet *ifp = rt->rt_ifp; 2451 struct ifaddr *ifa, *lo0ifa; 2452 int s = pserialize_read_enter(); 2453 2454 switch (req) { 2455 case RTM_ADD: 2456 if ((rt->rt_flags & RTF_LOCAL) == 0) 2457 break; 2458 2459 rt->rt_ifp = lo0ifp; 2460 2461 if (ISSET(info->rti_flags, RTF_DONTCHANGEIFA)) 2462 break; 2463 2464 IFADDR_READER_FOREACH(ifa, ifp) { 2465 if (equal(rt_getkey(rt), ifa->ifa_addr)) 2466 break; 2467 } 2468 if (ifa == NULL) 2469 break; 2470 2471 /* 2472 * Ensure lo0 has an address of the same family. 2473 */ 2474 IFADDR_READER_FOREACH(lo0ifa, lo0ifp) { 2475 if (lo0ifa->ifa_addr->sa_family == 2476 ifa->ifa_addr->sa_family) 2477 break; 2478 } 2479 if (lo0ifa == NULL) 2480 break; 2481 2482 /* 2483 * Make sure to set rt->rt_ifa to the interface 2484 * address we are using, otherwise we will have trouble 2485 * with source address selection. 2486 */ 2487 if (ifa != rt->rt_ifa) 2488 rt_replace_ifa(rt, ifa); 2489 break; 2490 case RTM_DELETE: 2491 default: 2492 break; 2493 } 2494 pserialize_read_exit(s); 2495 } 2496 2497 static void 2498 _if_down(struct ifnet *ifp) 2499 { 2500 struct ifaddr *ifa; 2501 struct domain *dp; 2502 int s, bound; 2503 struct psref psref; 2504 2505 ifp->if_flags &= ~IFF_UP; 2506 nanotime(&ifp->if_lastchange); 2507 2508 bound = curlwp_bind(); 2509 s = pserialize_read_enter(); 2510 IFADDR_READER_FOREACH(ifa, ifp) { 2511 ifa_acquire(ifa, &psref); 2512 pserialize_read_exit(s); 2513 2514 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2515 2516 s = pserialize_read_enter(); 2517 ifa_release(ifa, &psref); 2518 } 2519 pserialize_read_exit(s); 2520 curlwp_bindx(bound); 2521 2522 IF_LINK_STATE_CHANGE_LOCK(ifp); 2523 ifp->if_link_cansched = false; 2524 workqueue_wait(ifnet_link_state_wq, &ifp->if_link_work); 2525 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2526 2527 IFQ_PURGE(&ifp->if_snd); 2528 #if NCARP > 0 2529 if (ifp->if_carp) 2530 carp_carpdev_state(ifp); 2531 #endif 2532 rt_ifmsg(ifp); 2533 DOMAIN_FOREACH(dp) { 2534 if (dp->dom_if_down) 2535 dp->dom_if_down(ifp); 2536 } 2537 } 2538 2539 static void 2540 if_down_deactivated(struct ifnet *ifp) 2541 { 2542 2543 KASSERT(if_is_deactivated(ifp)); 2544 _if_down(ifp); 2545 } 2546 2547 void 2548 if_down_locked(struct ifnet *ifp) 2549 { 2550 2551 KASSERT(IFNET_LOCKED(ifp)); 2552 _if_down(ifp); 2553 } 2554 2555 /* 2556 * Mark an interface down and notify protocols of 2557 * the transition. 2558 * NOTE: must be called at splsoftnet or equivalent. 2559 */ 2560 void 2561 if_down(struct ifnet *ifp) 2562 { 2563 2564 IFNET_LOCK(ifp); 2565 if_down_locked(ifp); 2566 IFNET_UNLOCK(ifp); 2567 } 2568 2569 /* 2570 * Must be called with holding if_ioctl_lock. 2571 */ 2572 static void 2573 if_up_locked(struct ifnet *ifp) 2574 { 2575 #ifdef notyet 2576 struct ifaddr *ifa; 2577 #endif 2578 struct domain *dp; 2579 2580 KASSERT(IFNET_LOCKED(ifp)); 2581 2582 KASSERT(!if_is_deactivated(ifp)); 2583 ifp->if_flags |= IFF_UP; 2584 nanotime(&ifp->if_lastchange); 2585 #ifdef notyet 2586 /* this has no effect on IP, and will kill all ISO connections XXX */ 2587 IFADDR_READER_FOREACH(ifa, ifp) 2588 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2589 #endif 2590 #if NCARP > 0 2591 if (ifp->if_carp) 2592 carp_carpdev_state(ifp); 2593 #endif 2594 rt_ifmsg(ifp); 2595 DOMAIN_FOREACH(dp) { 2596 if (dp->dom_if_up) 2597 dp->dom_if_up(ifp); 2598 } 2599 2600 IF_LINK_STATE_CHANGE_LOCK(ifp); 2601 ifp->if_link_cansched = true; 2602 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2603 } 2604 2605 /* 2606 * Handle interface slowtimo timer routine. Called 2607 * from softclock, we decrement timer (if set) and 2608 * call the appropriate interface routine on expiration. 2609 */ 2610 static void 2611 if_slowtimo(void *arg) 2612 { 2613 void (*slowtimo)(struct ifnet *); 2614 struct ifnet *ifp = arg; 2615 int s; 2616 2617 slowtimo = ifp->if_slowtimo; 2618 if (__predict_false(slowtimo == NULL)) 2619 return; 2620 2621 s = splnet(); 2622 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2623 (*slowtimo)(ifp); 2624 2625 splx(s); 2626 2627 if (__predict_true(ifp->if_slowtimo != NULL)) 2628 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2629 } 2630 2631 /* 2632 * Mark an interface up and notify protocols of 2633 * the transition. 2634 * NOTE: must be called at splsoftnet or equivalent. 2635 */ 2636 void 2637 if_up(struct ifnet *ifp) 2638 { 2639 2640 IFNET_LOCK(ifp); 2641 if_up_locked(ifp); 2642 IFNET_UNLOCK(ifp); 2643 } 2644 2645 /* 2646 * Set/clear promiscuous mode on interface ifp based on the truth value 2647 * of pswitch. The calls are reference counted so that only the first 2648 * "on" request actually has an effect, as does the final "off" request. 2649 * Results are undefined if the "off" and "on" requests are not matched. 2650 */ 2651 int 2652 ifpromisc_locked(struct ifnet *ifp, int pswitch) 2653 { 2654 int pcount, ret = 0; 2655 u_short nflags; 2656 2657 KASSERT(IFNET_LOCKED(ifp)); 2658 2659 pcount = ifp->if_pcount; 2660 if (pswitch) { 2661 /* 2662 * Allow the device to be "placed" into promiscuous 2663 * mode even if it is not configured up. It will 2664 * consult IFF_PROMISC when it is brought up. 2665 */ 2666 if (ifp->if_pcount++ != 0) 2667 goto out; 2668 nflags = ifp->if_flags | IFF_PROMISC; 2669 } else { 2670 if (--ifp->if_pcount > 0) 2671 goto out; 2672 nflags = ifp->if_flags & ~IFF_PROMISC; 2673 } 2674 ret = if_flags_set(ifp, nflags); 2675 /* Restore interface state if not successful. */ 2676 if (ret != 0) { 2677 ifp->if_pcount = pcount; 2678 } 2679 out: 2680 return ret; 2681 } 2682 2683 int 2684 ifpromisc(struct ifnet *ifp, int pswitch) 2685 { 2686 int e; 2687 2688 IFNET_LOCK(ifp); 2689 e = ifpromisc_locked(ifp, pswitch); 2690 IFNET_UNLOCK(ifp); 2691 2692 return e; 2693 } 2694 2695 /* 2696 * Map interface name to 2697 * interface structure pointer. 2698 */ 2699 struct ifnet * 2700 ifunit(const char *name) 2701 { 2702 struct ifnet *ifp; 2703 const char *cp = name; 2704 u_int unit = 0; 2705 u_int i; 2706 int s; 2707 2708 /* 2709 * If the entire name is a number, treat it as an ifindex. 2710 */ 2711 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2712 unit = unit * 10 + (*cp - '0'); 2713 } 2714 2715 /* 2716 * If the number took all of the name, then it's a valid ifindex. 2717 */ 2718 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) 2719 return if_byindex(unit); 2720 2721 ifp = NULL; 2722 s = pserialize_read_enter(); 2723 IFNET_READER_FOREACH(ifp) { 2724 if (if_is_deactivated(ifp)) 2725 continue; 2726 if (strcmp(ifp->if_xname, name) == 0) 2727 goto out; 2728 } 2729 out: 2730 pserialize_read_exit(s); 2731 return ifp; 2732 } 2733 2734 /* 2735 * Get a reference of an ifnet object by an interface name. 2736 * The returned reference is protected by psref(9). The caller 2737 * must release a returned reference by if_put after use. 2738 */ 2739 struct ifnet * 2740 if_get(const char *name, struct psref *psref) 2741 { 2742 struct ifnet *ifp; 2743 const char *cp = name; 2744 u_int unit = 0; 2745 u_int i; 2746 int s; 2747 2748 /* 2749 * If the entire name is a number, treat it as an ifindex. 2750 */ 2751 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2752 unit = unit * 10 + (*cp - '0'); 2753 } 2754 2755 /* 2756 * If the number took all of the name, then it's a valid ifindex. 2757 */ 2758 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) 2759 return if_get_byindex(unit, psref); 2760 2761 ifp = NULL; 2762 s = pserialize_read_enter(); 2763 IFNET_READER_FOREACH(ifp) { 2764 if (if_is_deactivated(ifp)) 2765 continue; 2766 if (strcmp(ifp->if_xname, name) == 0) { 2767 PSREF_DEBUG_FILL_RETURN_ADDRESS(psref); 2768 psref_acquire(psref, &ifp->if_psref, 2769 ifnet_psref_class); 2770 goto out; 2771 } 2772 } 2773 out: 2774 pserialize_read_exit(s); 2775 return ifp; 2776 } 2777 2778 /* 2779 * Release a reference of an ifnet object given by if_get, if_get_byindex 2780 * or if_get_bylla. 2781 */ 2782 void 2783 if_put(const struct ifnet *ifp, struct psref *psref) 2784 { 2785 2786 if (ifp == NULL) 2787 return; 2788 2789 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2790 } 2791 2792 /* 2793 * Return ifp having idx. Return NULL if not found. Normally if_byindex 2794 * should be used. 2795 */ 2796 ifnet_t * 2797 _if_byindex(u_int idx) 2798 { 2799 2800 return (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2801 } 2802 2803 /* 2804 * Return ifp having idx. Return NULL if not found or the found ifp is 2805 * already deactivated. 2806 */ 2807 ifnet_t * 2808 if_byindex(u_int idx) 2809 { 2810 ifnet_t *ifp; 2811 2812 ifp = _if_byindex(idx); 2813 if (ifp != NULL && if_is_deactivated(ifp)) 2814 ifp = NULL; 2815 return ifp; 2816 } 2817 2818 /* 2819 * Get a reference of an ifnet object by an interface index. 2820 * The returned reference is protected by psref(9). The caller 2821 * must release a returned reference by if_put after use. 2822 */ 2823 ifnet_t * 2824 if_get_byindex(u_int idx, struct psref *psref) 2825 { 2826 ifnet_t *ifp; 2827 int s; 2828 2829 s = pserialize_read_enter(); 2830 ifp = if_byindex(idx); 2831 if (__predict_true(ifp != NULL)) { 2832 PSREF_DEBUG_FILL_RETURN_ADDRESS(psref); 2833 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2834 } 2835 pserialize_read_exit(s); 2836 2837 return ifp; 2838 } 2839 2840 ifnet_t * 2841 if_get_bylla(const void *lla, unsigned char lla_len, struct psref *psref) 2842 { 2843 ifnet_t *ifp; 2844 int s; 2845 2846 s = pserialize_read_enter(); 2847 IFNET_READER_FOREACH(ifp) { 2848 if (if_is_deactivated(ifp)) 2849 continue; 2850 if (ifp->if_addrlen != lla_len) 2851 continue; 2852 if (memcmp(lla, CLLADDR(ifp->if_sadl), lla_len) == 0) { 2853 psref_acquire(psref, &ifp->if_psref, 2854 ifnet_psref_class); 2855 break; 2856 } 2857 } 2858 pserialize_read_exit(s); 2859 2860 return ifp; 2861 } 2862 2863 /* 2864 * Note that it's safe only if the passed ifp is guaranteed to not be freed, 2865 * for example using pserialize or the ifp is already held or some other 2866 * object is held which guarantes the ifp to not be freed indirectly. 2867 */ 2868 void 2869 if_acquire(struct ifnet *ifp, struct psref *psref) 2870 { 2871 2872 KASSERT(ifp->if_index != 0); 2873 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2874 } 2875 2876 bool 2877 if_held(struct ifnet *ifp) 2878 { 2879 2880 return psref_held(&ifp->if_psref, ifnet_psref_class); 2881 } 2882 2883 /* 2884 * Some tunnel interfaces can nest, e.g. IPv4 over IPv4 gif(4) tunnel over IPv4. 2885 * Check the tunnel nesting count. 2886 * Return > 0, if tunnel nesting count is more than limit. 2887 * Return 0, if tunnel nesting count is equal or less than limit. 2888 */ 2889 int 2890 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, int limit) 2891 { 2892 struct m_tag *mtag; 2893 int *count; 2894 2895 mtag = m_tag_find(m, PACKET_TAG_TUNNEL_INFO); 2896 if (mtag != NULL) { 2897 count = (int *)(mtag + 1); 2898 if (++(*count) > limit) { 2899 log(LOG_NOTICE, 2900 "%s: recursively called too many times(%d)\n", 2901 ifp->if_xname, *count); 2902 return EIO; 2903 } 2904 } else { 2905 mtag = m_tag_get(PACKET_TAG_TUNNEL_INFO, sizeof(*count), 2906 M_NOWAIT); 2907 if (mtag != NULL) { 2908 m_tag_prepend(m, mtag); 2909 count = (int *)(mtag + 1); 2910 *count = 0; 2911 } else { 2912 log(LOG_DEBUG, 2913 "%s: m_tag_get() failed, recursion calls are not prevented.\n", 2914 ifp->if_xname); 2915 } 2916 } 2917 2918 return 0; 2919 } 2920 2921 static void 2922 if_tunnel_ro_init_pc(void *p, void *arg __unused, struct cpu_info *ci __unused) 2923 { 2924 struct tunnel_ro *tro = p; 2925 2926 tro->tr_ro = kmem_zalloc(sizeof(*tro->tr_ro), KM_SLEEP); 2927 tro->tr_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 2928 } 2929 2930 static void 2931 if_tunnel_ro_fini_pc(void *p, void *arg __unused, struct cpu_info *ci __unused) 2932 { 2933 struct tunnel_ro *tro = p; 2934 2935 rtcache_free(tro->tr_ro); 2936 kmem_free(tro->tr_ro, sizeof(*tro->tr_ro)); 2937 2938 mutex_obj_free(tro->tr_lock); 2939 } 2940 2941 percpu_t * 2942 if_tunnel_alloc_ro_percpu(void) 2943 { 2944 2945 return percpu_create(sizeof(struct tunnel_ro), 2946 if_tunnel_ro_init_pc, if_tunnel_ro_fini_pc, NULL); 2947 } 2948 2949 void 2950 if_tunnel_free_ro_percpu(percpu_t *ro_percpu) 2951 { 2952 2953 percpu_free(ro_percpu, sizeof(struct tunnel_ro)); 2954 } 2955 2956 2957 static void 2958 if_tunnel_rtcache_free_pc(void *p, void *arg __unused, struct cpu_info *ci __unused) 2959 { 2960 struct tunnel_ro *tro = p; 2961 2962 mutex_enter(tro->tr_lock); 2963 rtcache_free(tro->tr_ro); 2964 mutex_exit(tro->tr_lock); 2965 } 2966 2967 void if_tunnel_ro_percpu_rtcache_free(percpu_t *ro_percpu) 2968 { 2969 2970 percpu_foreach(ro_percpu, if_tunnel_rtcache_free_pc, NULL); 2971 } 2972 2973 void 2974 if_export_if_data(ifnet_t * const ifp, struct if_data *ifi, bool zero_stats) 2975 { 2976 2977 /* Collet the volatile stats first; this zeros *ifi. */ 2978 if_stats_to_if_data(ifp, ifi, zero_stats); 2979 2980 ifi->ifi_type = ifp->if_type; 2981 ifi->ifi_addrlen = ifp->if_addrlen; 2982 ifi->ifi_hdrlen = ifp->if_hdrlen; 2983 ifi->ifi_link_state = ifp->if_link_state; 2984 ifi->ifi_mtu = ifp->if_mtu; 2985 ifi->ifi_metric = ifp->if_metric; 2986 ifi->ifi_baudrate = ifp->if_baudrate; 2987 ifi->ifi_lastchange = ifp->if_lastchange; 2988 } 2989 2990 /* common */ 2991 int 2992 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2993 { 2994 int s; 2995 struct ifreq *ifr; 2996 struct ifcapreq *ifcr; 2997 struct ifdatareq *ifdr; 2998 unsigned short flags; 2999 char *descr; 3000 int error; 3001 3002 switch (cmd) { 3003 case SIOCSIFCAP: 3004 ifcr = data; 3005 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 3006 return EINVAL; 3007 3008 if (ifcr->ifcr_capenable == ifp->if_capenable) 3009 return 0; 3010 3011 ifp->if_capenable = ifcr->ifcr_capenable; 3012 3013 /* Pre-compute the checksum flags mask. */ 3014 ifp->if_csum_flags_tx = 0; 3015 ifp->if_csum_flags_rx = 0; 3016 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) 3017 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 3018 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) 3019 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 3020 3021 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) 3022 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 3023 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) 3024 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 3025 3026 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) 3027 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 3028 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) 3029 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 3030 3031 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) 3032 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 3033 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) 3034 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 3035 3036 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) 3037 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 3038 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) 3039 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 3040 3041 if (ifp->if_capenable & IFCAP_TSOv4) 3042 ifp->if_csum_flags_tx |= M_CSUM_TSOv4; 3043 if (ifp->if_capenable & IFCAP_TSOv6) 3044 ifp->if_csum_flags_tx |= M_CSUM_TSOv6; 3045 3046 #if NBRIDGE > 0 3047 if (ifp->if_bridge != NULL) 3048 bridge_calc_csum_flags(ifp->if_bridge); 3049 #endif 3050 3051 if (ifp->if_flags & IFF_UP) 3052 return ENETRESET; 3053 return 0; 3054 case SIOCSIFFLAGS: 3055 ifr = data; 3056 /* 3057 * If if_is_mpsafe(ifp), KERNEL_LOCK isn't held here, but if_up 3058 * and if_down aren't MP-safe yet, so we must hold the lock. 3059 */ 3060 KERNEL_LOCK_IF_IFP_MPSAFE(ifp); 3061 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 3062 s = splsoftnet(); 3063 if_down_locked(ifp); 3064 splx(s); 3065 } 3066 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 3067 s = splsoftnet(); 3068 if_up_locked(ifp); 3069 splx(s); 3070 } 3071 KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp); 3072 flags = (ifp->if_flags & IFF_CANTCHANGE) | 3073 (ifr->ifr_flags &~ IFF_CANTCHANGE); 3074 if (ifp->if_flags != flags) { 3075 ifp->if_flags = flags; 3076 /* Notify that the flags have changed. */ 3077 rt_ifmsg(ifp); 3078 } 3079 break; 3080 case SIOCGIFFLAGS: 3081 ifr = data; 3082 ifr->ifr_flags = ifp->if_flags; 3083 break; 3084 3085 case SIOCGIFMETRIC: 3086 ifr = data; 3087 ifr->ifr_metric = ifp->if_metric; 3088 break; 3089 3090 case SIOCGIFMTU: 3091 ifr = data; 3092 ifr->ifr_mtu = ifp->if_mtu; 3093 break; 3094 3095 case SIOCGIFDLT: 3096 ifr = data; 3097 ifr->ifr_dlt = ifp->if_dlt; 3098 break; 3099 3100 case SIOCGIFCAP: 3101 ifcr = data; 3102 ifcr->ifcr_capabilities = ifp->if_capabilities; 3103 ifcr->ifcr_capenable = ifp->if_capenable; 3104 break; 3105 3106 case SIOCSIFMETRIC: 3107 ifr = data; 3108 ifp->if_metric = ifr->ifr_metric; 3109 break; 3110 3111 case SIOCGIFDATA: 3112 ifdr = data; 3113 if_export_if_data(ifp, &ifdr->ifdr_data, false); 3114 break; 3115 3116 case SIOCGIFINDEX: 3117 ifr = data; 3118 ifr->ifr_index = ifp->if_index; 3119 break; 3120 3121 case SIOCZIFDATA: 3122 ifdr = data; 3123 if_export_if_data(ifp, &ifdr->ifdr_data, true); 3124 getnanotime(&ifp->if_lastchange); 3125 break; 3126 case SIOCSIFMTU: 3127 ifr = data; 3128 if (ifp->if_mtu == ifr->ifr_mtu) 3129 break; 3130 ifp->if_mtu = ifr->ifr_mtu; 3131 /* 3132 * If the link MTU changed, do network layer specific procedure. 3133 */ 3134 #ifdef INET6 3135 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 3136 if (in6_present) 3137 nd6_setmtu(ifp); 3138 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 3139 #endif 3140 return ENETRESET; 3141 case SIOCSIFDESCR: 3142 error = kauth_authorize_network(curlwp->l_cred, 3143 KAUTH_NETWORK_INTERFACE, 3144 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, KAUTH_ARG(cmd), 3145 NULL); 3146 if (error) 3147 return error; 3148 3149 ifr = data; 3150 3151 if (ifr->ifr_buflen > IFDESCRSIZE) 3152 return ENAMETOOLONG; 3153 3154 if (ifr->ifr_buf == NULL || ifr->ifr_buflen == 0) { 3155 /* unset description */ 3156 descr = NULL; 3157 } else { 3158 descr = kmem_zalloc(IFDESCRSIZE, KM_SLEEP); 3159 /* 3160 * copy (IFDESCRSIZE - 1) bytes to ensure 3161 * terminating nul 3162 */ 3163 error = copyin(ifr->ifr_buf, descr, IFDESCRSIZE - 1); 3164 if (error) { 3165 kmem_free(descr, IFDESCRSIZE); 3166 return error; 3167 } 3168 } 3169 3170 if (ifp->if_description != NULL) 3171 kmem_free(ifp->if_description, IFDESCRSIZE); 3172 3173 ifp->if_description = descr; 3174 break; 3175 3176 case SIOCGIFDESCR: 3177 ifr = data; 3178 descr = ifp->if_description; 3179 3180 if (descr == NULL) 3181 return ENOMSG; 3182 3183 if (ifr->ifr_buflen < IFDESCRSIZE) 3184 return EINVAL; 3185 3186 error = copyout(descr, ifr->ifr_buf, IFDESCRSIZE); 3187 if (error) 3188 return error; 3189 break; 3190 3191 default: 3192 return ENOTTY; 3193 } 3194 return 0; 3195 } 3196 3197 int 3198 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 3199 { 3200 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 3201 struct ifaddr *ifa; 3202 const struct sockaddr *any, *sa; 3203 union { 3204 struct sockaddr sa; 3205 struct sockaddr_storage ss; 3206 } u, v; 3207 int s, error = 0; 3208 3209 switch (cmd) { 3210 case SIOCSIFADDRPREF: 3211 error = kauth_authorize_network(curlwp->l_cred, 3212 KAUTH_NETWORK_INTERFACE, 3213 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, KAUTH_ARG(cmd), 3214 NULL); 3215 if (error) 3216 return error; 3217 break; 3218 case SIOCGIFADDRPREF: 3219 break; 3220 default: 3221 return EOPNOTSUPP; 3222 } 3223 3224 /* sanity checks */ 3225 if (data == NULL || ifp == NULL) { 3226 panic("invalid argument to %s", __func__); 3227 /*NOTREACHED*/ 3228 } 3229 3230 /* address must be specified on ADD and DELETE */ 3231 sa = sstocsa(&ifap->ifap_addr); 3232 if (sa->sa_family != sofamily(so)) 3233 return EINVAL; 3234 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 3235 return EINVAL; 3236 3237 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 3238 3239 s = pserialize_read_enter(); 3240 IFADDR_READER_FOREACH(ifa, ifp) { 3241 if (ifa->ifa_addr->sa_family != sa->sa_family) 3242 continue; 3243 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 3244 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 3245 break; 3246 } 3247 if (ifa == NULL) { 3248 error = EADDRNOTAVAIL; 3249 goto out; 3250 } 3251 3252 switch (cmd) { 3253 case SIOCSIFADDRPREF: 3254 ifa->ifa_preference = ifap->ifap_preference; 3255 goto out; 3256 case SIOCGIFADDRPREF: 3257 /* fill in the if_laddrreq structure */ 3258 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 3259 sizeof(ifap->ifap_addr), ifa->ifa_addr); 3260 ifap->ifap_preference = ifa->ifa_preference; 3261 goto out; 3262 default: 3263 error = EOPNOTSUPP; 3264 } 3265 out: 3266 pserialize_read_exit(s); 3267 return error; 3268 } 3269 3270 /* 3271 * Interface ioctls. 3272 */ 3273 static int 3274 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 3275 { 3276 struct ifnet *ifp; 3277 struct ifreq *ifr; 3278 int error = 0; 3279 u_long ocmd = cmd; 3280 u_short oif_flags; 3281 struct ifreq ifrb; 3282 struct oifreq *oifr = NULL; 3283 int r; 3284 struct psref psref; 3285 int bound; 3286 bool do_if43_post = false; 3287 bool do_ifm80_post = false; 3288 3289 switch (cmd) { 3290 case SIOCGIFCONF: 3291 return ifconf(cmd, data); 3292 case SIOCINITIFADDR: 3293 return EPERM; 3294 default: 3295 MODULE_HOOK_CALL(uipc_syscalls_40_hook, (cmd, data), enosys(), 3296 error); 3297 if (error != ENOSYS) 3298 return error; 3299 MODULE_HOOK_CALL(uipc_syscalls_50_hook, (l, cmd, data), 3300 enosys(), error); 3301 if (error != ENOSYS) 3302 return error; 3303 error = 0; 3304 break; 3305 } 3306 3307 ifr = data; 3308 /* Pre-conversion */ 3309 MODULE_HOOK_CALL(if_cvtcmd_43_hook, (&cmd, ocmd), enosys(), error); 3310 if (cmd != ocmd) { 3311 oifr = data; 3312 data = ifr = &ifrb; 3313 IFREQO2N_43(oifr, ifr); 3314 do_if43_post = true; 3315 } 3316 MODULE_HOOK_CALL(ifmedia_80_pre_hook, (ifr, &cmd, &do_ifm80_post), 3317 enosys(), error); 3318 3319 switch (cmd) { 3320 case SIOCIFCREATE: 3321 case SIOCIFDESTROY: 3322 bound = curlwp_bind(); 3323 if (l != NULL) { 3324 ifp = if_get(ifr->ifr_name, &psref); 3325 error = kauth_authorize_network(l->l_cred, 3326 KAUTH_NETWORK_INTERFACE, 3327 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 3328 KAUTH_ARG(cmd), NULL); 3329 if (ifp != NULL) 3330 if_put(ifp, &psref); 3331 if (error != 0) { 3332 curlwp_bindx(bound); 3333 return error; 3334 } 3335 } 3336 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 3337 mutex_enter(&if_clone_mtx); 3338 r = (cmd == SIOCIFCREATE) ? 3339 if_clone_create(ifr->ifr_name) : 3340 if_clone_destroy(ifr->ifr_name); 3341 mutex_exit(&if_clone_mtx); 3342 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 3343 curlwp_bindx(bound); 3344 return r; 3345 3346 case SIOCIFGCLONERS: 3347 { 3348 struct if_clonereq *req = (struct if_clonereq *)data; 3349 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 3350 &req->ifcr_total); 3351 } 3352 } 3353 3354 bound = curlwp_bind(); 3355 ifp = if_get(ifr->ifr_name, &psref); 3356 if (ifp == NULL) { 3357 curlwp_bindx(bound); 3358 return ENXIO; 3359 } 3360 3361 switch (cmd) { 3362 case SIOCALIFADDR: 3363 case SIOCDLIFADDR: 3364 case SIOCSIFADDRPREF: 3365 case SIOCSIFFLAGS: 3366 case SIOCSIFCAP: 3367 case SIOCSIFMETRIC: 3368 case SIOCZIFDATA: 3369 case SIOCSIFMTU: 3370 case SIOCSIFPHYADDR: 3371 case SIOCDIFPHYADDR: 3372 #ifdef INET6 3373 case SIOCSIFPHYADDR_IN6: 3374 #endif 3375 case SIOCSLIFPHYADDR: 3376 case SIOCADDMULTI: 3377 case SIOCDELMULTI: 3378 case SIOCSETHERCAP: 3379 case SIOCSIFMEDIA: 3380 case SIOCSDRVSPEC: 3381 case SIOCG80211: 3382 case SIOCS80211: 3383 case SIOCS80211NWID: 3384 case SIOCS80211NWKEY: 3385 case SIOCS80211POWER: 3386 case SIOCS80211BSSID: 3387 case SIOCS80211CHANNEL: 3388 case SIOCSLINKSTR: 3389 if (l != NULL) { 3390 error = kauth_authorize_network(l->l_cred, 3391 KAUTH_NETWORK_INTERFACE, 3392 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 3393 KAUTH_ARG(cmd), NULL); 3394 if (error != 0) 3395 goto out; 3396 } 3397 } 3398 3399 oif_flags = ifp->if_flags; 3400 3401 KERNEL_LOCK_UNLESS_IFP_MPSAFE(ifp); 3402 IFNET_LOCK(ifp); 3403 3404 error = (*ifp->if_ioctl)(ifp, cmd, data); 3405 if (error != ENOTTY) 3406 ; 3407 else if (so->so_proto == NULL) 3408 error = EOPNOTSUPP; 3409 else { 3410 KERNEL_LOCK_IF_IFP_MPSAFE(ifp); 3411 MODULE_HOOK_CALL(if_ifioctl_43_hook, 3412 (so, ocmd, cmd, data, l), enosys(), error); 3413 if (error == ENOSYS) 3414 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 3415 cmd, data, ifp); 3416 KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp); 3417 } 3418 3419 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 3420 if ((ifp->if_flags & IFF_UP) != 0) { 3421 int s = splsoftnet(); 3422 if_up_locked(ifp); 3423 splx(s); 3424 } 3425 } 3426 3427 /* Post-conversion */ 3428 if (do_ifm80_post && (error == 0)) 3429 MODULE_HOOK_CALL(ifmedia_80_post_hook, (ifr, cmd), 3430 enosys(), error); 3431 if (do_if43_post) 3432 IFREQN2O_43(oifr, ifr); 3433 3434 IFNET_UNLOCK(ifp); 3435 KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(ifp); 3436 out: 3437 if_put(ifp, &psref); 3438 curlwp_bindx(bound); 3439 return error; 3440 } 3441 3442 /* 3443 * Return interface configuration 3444 * of system. List may be used 3445 * in later ioctl's (above) to get 3446 * other information. 3447 * 3448 * Each record is a struct ifreq. Before the addition of 3449 * sockaddr_storage, the API rule was that sockaddr flavors that did 3450 * not fit would extend beyond the struct ifreq, with the next struct 3451 * ifreq starting sa_len beyond the struct sockaddr. Because the 3452 * union in struct ifreq includes struct sockaddr_storage, every kind 3453 * of sockaddr must fit. Thus, there are no longer any overlength 3454 * records. 3455 * 3456 * Records are added to the user buffer if they fit, and ifc_len is 3457 * adjusted to the length that was written. Thus, the user is only 3458 * assured of getting the complete list if ifc_len on return is at 3459 * least sizeof(struct ifreq) less than it was on entry. 3460 * 3461 * If the user buffer pointer is NULL, this routine copies no data and 3462 * returns the amount of space that would be needed. 3463 * 3464 * Invariants: 3465 * ifrp points to the next part of the user's buffer to be used. If 3466 * ifrp != NULL, space holds the number of bytes remaining that we may 3467 * write at ifrp. Otherwise, space holds the number of bytes that 3468 * would have been written had there been adequate space. 3469 */ 3470 /*ARGSUSED*/ 3471 static int 3472 ifconf(u_long cmd, void *data) 3473 { 3474 struct ifconf *ifc = (struct ifconf *)data; 3475 struct ifnet *ifp; 3476 struct ifaddr *ifa; 3477 struct ifreq ifr, *ifrp = NULL; 3478 int space = 0, error = 0; 3479 const int sz = (int)sizeof(struct ifreq); 3480 const bool docopy = ifc->ifc_req != NULL; 3481 int s; 3482 int bound; 3483 struct psref psref; 3484 3485 memset(&ifr, 0, sizeof(ifr)); 3486 if (docopy) { 3487 space = ifc->ifc_len; 3488 ifrp = ifc->ifc_req; 3489 } 3490 3491 bound = curlwp_bind(); 3492 s = pserialize_read_enter(); 3493 IFNET_READER_FOREACH(ifp) { 3494 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 3495 pserialize_read_exit(s); 3496 3497 (void)strncpy(ifr.ifr_name, ifp->if_xname, 3498 sizeof(ifr.ifr_name)); 3499 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 3500 error = ENAMETOOLONG; 3501 goto release_exit; 3502 } 3503 if (IFADDR_READER_EMPTY(ifp)) { 3504 /* Interface with no addresses - send zero sockaddr. */ 3505 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 3506 if (!docopy) { 3507 space += sz; 3508 goto next; 3509 } 3510 if (space >= sz) { 3511 error = copyout(&ifr, ifrp, sz); 3512 if (error != 0) 3513 goto release_exit; 3514 ifrp++; 3515 space -= sz; 3516 } 3517 } 3518 3519 s = pserialize_read_enter(); 3520 IFADDR_READER_FOREACH(ifa, ifp) { 3521 struct sockaddr *sa = ifa->ifa_addr; 3522 /* all sockaddrs must fit in sockaddr_storage */ 3523 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 3524 3525 if (!docopy) { 3526 space += sz; 3527 continue; 3528 } 3529 memcpy(&ifr.ifr_space, sa, sa->sa_len); 3530 pserialize_read_exit(s); 3531 3532 if (space >= sz) { 3533 error = copyout(&ifr, ifrp, sz); 3534 if (error != 0) 3535 goto release_exit; 3536 ifrp++; space -= sz; 3537 } 3538 s = pserialize_read_enter(); 3539 } 3540 pserialize_read_exit(s); 3541 3542 next: 3543 s = pserialize_read_enter(); 3544 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3545 } 3546 pserialize_read_exit(s); 3547 curlwp_bindx(bound); 3548 3549 if (docopy) { 3550 KASSERT(0 <= space && space <= ifc->ifc_len); 3551 ifc->ifc_len -= space; 3552 } else { 3553 KASSERT(space >= 0); 3554 ifc->ifc_len = space; 3555 } 3556 return (0); 3557 3558 release_exit: 3559 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3560 curlwp_bindx(bound); 3561 return error; 3562 } 3563 3564 int 3565 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 3566 { 3567 uint8_t len = sizeof(ifr->ifr_ifru.ifru_space); 3568 struct ifreq ifrb; 3569 struct oifreq *oifr = NULL; 3570 u_long ocmd = cmd; 3571 int hook; 3572 3573 MODULE_HOOK_CALL(if_cvtcmd_43_hook, (&cmd, ocmd), enosys(), hook); 3574 if (hook != ENOSYS) { 3575 if (cmd != ocmd) { 3576 oifr = (struct oifreq *)(void *)ifr; 3577 ifr = &ifrb; 3578 IFREQO2N_43(oifr, ifr); 3579 len = sizeof(oifr->ifr_addr); 3580 } 3581 } 3582 3583 if (len < sa->sa_len) 3584 return EFBIG; 3585 3586 memset(&ifr->ifr_addr, 0, len); 3587 sockaddr_copy(&ifr->ifr_addr, len, sa); 3588 3589 if (cmd != ocmd) 3590 IFREQN2O_43(oifr, ifr); 3591 return 0; 3592 } 3593 3594 /* 3595 * wrapper function for the drivers which doesn't have if_transmit(). 3596 */ 3597 static int 3598 if_transmit(struct ifnet *ifp, struct mbuf *m) 3599 { 3600 int s, error; 3601 size_t pktlen = m->m_pkthdr.len; 3602 bool mcast = (m->m_flags & M_MCAST) != 0; 3603 3604 s = splnet(); 3605 3606 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3607 if (error != 0) { 3608 /* mbuf is already freed */ 3609 goto out; 3610 } 3611 3612 net_stat_ref_t nsr = IF_STAT_GETREF(ifp); 3613 if_statadd_ref(nsr, if_obytes, pktlen); 3614 if (mcast) 3615 if_statinc_ref(nsr, if_omcasts); 3616 IF_STAT_PUTREF(ifp); 3617 3618 if ((ifp->if_flags & IFF_OACTIVE) == 0) 3619 if_start_lock(ifp); 3620 out: 3621 splx(s); 3622 3623 return error; 3624 } 3625 3626 int 3627 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 3628 { 3629 int error; 3630 3631 kmsan_check_mbuf(m); 3632 3633 #ifdef ALTQ 3634 KERNEL_LOCK(1, NULL); 3635 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 3636 error = if_transmit(ifp, m); 3637 KERNEL_UNLOCK_ONE(NULL); 3638 } else { 3639 KERNEL_UNLOCK_ONE(NULL); 3640 error = (*ifp->if_transmit)(ifp, m); 3641 /* mbuf is alredy freed */ 3642 } 3643 #else /* !ALTQ */ 3644 error = (*ifp->if_transmit)(ifp, m); 3645 /* mbuf is alredy freed */ 3646 #endif /* !ALTQ */ 3647 3648 return error; 3649 } 3650 3651 /* 3652 * Queue message on interface, and start output if interface 3653 * not yet active. 3654 */ 3655 int 3656 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 3657 { 3658 3659 return if_transmit_lock(ifp, m); 3660 } 3661 3662 /* 3663 * Queue message on interface, possibly using a second fast queue 3664 */ 3665 int 3666 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 3667 { 3668 int error = 0; 3669 3670 if (ifq != NULL 3671 #ifdef ALTQ 3672 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 3673 #endif 3674 ) { 3675 if (IF_QFULL(ifq)) { 3676 IF_DROP(&ifp->if_snd); 3677 m_freem(m); 3678 if (error == 0) 3679 error = ENOBUFS; 3680 } else 3681 IF_ENQUEUE(ifq, m); 3682 } else 3683 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3684 if (error != 0) { 3685 if_statinc(ifp, if_oerrors); 3686 return error; 3687 } 3688 return 0; 3689 } 3690 3691 int 3692 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 3693 { 3694 int rc; 3695 3696 KASSERT(IFNET_LOCKED(ifp)); 3697 if (ifp->if_initaddr != NULL) 3698 rc = (*ifp->if_initaddr)(ifp, ifa, src); 3699 else if (src || 3700 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 3701 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 3702 3703 return rc; 3704 } 3705 3706 int 3707 if_do_dad(struct ifnet *ifp) 3708 { 3709 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 3710 return 0; 3711 3712 switch (ifp->if_type) { 3713 case IFT_FAITH: 3714 /* 3715 * These interfaces do not have the IFF_LOOPBACK flag, 3716 * but loop packets back. We do not have to do DAD on such 3717 * interfaces. We should even omit it, because loop-backed 3718 * responses would confuse the DAD procedure. 3719 */ 3720 return 0; 3721 default: 3722 /* 3723 * Our DAD routine requires the interface up and running. 3724 * However, some interfaces can be up before the RUNNING 3725 * status. Additionaly, users may try to assign addresses 3726 * before the interface becomes up (or running). 3727 * We simply skip DAD in such a case as a work around. 3728 * XXX: we should rather mark "tentative" on such addresses, 3729 * and do DAD after the interface becomes ready. 3730 */ 3731 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 3732 (IFF_UP | IFF_RUNNING)) 3733 return 0; 3734 3735 return 1; 3736 } 3737 } 3738 3739 int 3740 if_flags_set(ifnet_t *ifp, const u_short flags) 3741 { 3742 int rc; 3743 3744 KASSERT(IFNET_LOCKED(ifp)); 3745 3746 if (ifp->if_setflags != NULL) 3747 rc = (*ifp->if_setflags)(ifp, flags); 3748 else { 3749 u_short cantflags, chgdflags; 3750 struct ifreq ifr; 3751 3752 chgdflags = ifp->if_flags ^ flags; 3753 cantflags = chgdflags & IFF_CANTCHANGE; 3754 3755 if (cantflags != 0) 3756 ifp->if_flags ^= cantflags; 3757 3758 /* Traditionally, we do not call if_ioctl after 3759 * setting/clearing only IFF_PROMISC if the interface 3760 * isn't IFF_UP. Uphold that tradition. 3761 */ 3762 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 3763 return 0; 3764 3765 memset(&ifr, 0, sizeof(ifr)); 3766 3767 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 3768 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 3769 3770 if (rc != 0 && cantflags != 0) 3771 ifp->if_flags ^= cantflags; 3772 } 3773 3774 return rc; 3775 } 3776 3777 int 3778 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 3779 { 3780 int rc; 3781 struct ifreq ifr; 3782 3783 if (ifp->if_mcastop != NULL) 3784 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 3785 else { 3786 ifreq_setaddr(cmd, &ifr, sa); 3787 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 3788 } 3789 3790 return rc; 3791 } 3792 3793 static void 3794 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 3795 struct ifaltq *ifq) 3796 { 3797 const struct sysctlnode *cnode, *rnode; 3798 3799 if (sysctl_createv(clog, 0, NULL, &rnode, 3800 CTLFLAG_PERMANENT, 3801 CTLTYPE_NODE, "interfaces", 3802 SYSCTL_DESCR("Per-interface controls"), 3803 NULL, 0, NULL, 0, 3804 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3805 goto bad; 3806 3807 if (sysctl_createv(clog, 0, &rnode, &rnode, 3808 CTLFLAG_PERMANENT, 3809 CTLTYPE_NODE, ifname, 3810 SYSCTL_DESCR("Interface controls"), 3811 NULL, 0, NULL, 0, 3812 CTL_CREATE, CTL_EOL) != 0) 3813 goto bad; 3814 3815 if (sysctl_createv(clog, 0, &rnode, &rnode, 3816 CTLFLAG_PERMANENT, 3817 CTLTYPE_NODE, "sndq", 3818 SYSCTL_DESCR("Interface output queue controls"), 3819 NULL, 0, NULL, 0, 3820 CTL_CREATE, CTL_EOL) != 0) 3821 goto bad; 3822 3823 if (sysctl_createv(clog, 0, &rnode, &cnode, 3824 CTLFLAG_PERMANENT, 3825 CTLTYPE_INT, "len", 3826 SYSCTL_DESCR("Current output queue length"), 3827 NULL, 0, &ifq->ifq_len, 0, 3828 CTL_CREATE, CTL_EOL) != 0) 3829 goto bad; 3830 3831 if (sysctl_createv(clog, 0, &rnode, &cnode, 3832 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 3833 CTLTYPE_INT, "maxlen", 3834 SYSCTL_DESCR("Maximum allowed output queue length"), 3835 NULL, 0, &ifq->ifq_maxlen, 0, 3836 CTL_CREATE, CTL_EOL) != 0) 3837 goto bad; 3838 3839 if (sysctl_createv(clog, 0, &rnode, &cnode, 3840 CTLFLAG_PERMANENT, 3841 CTLTYPE_INT, "drops", 3842 SYSCTL_DESCR("Packets dropped due to full output queue"), 3843 NULL, 0, &ifq->ifq_drops, 0, 3844 CTL_CREATE, CTL_EOL) != 0) 3845 goto bad; 3846 3847 return; 3848 bad: 3849 printf("%s: could not attach sysctl nodes\n", ifname); 3850 return; 3851 } 3852 3853 #if defined(INET) || defined(INET6) 3854 3855 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3856 static int \ 3857 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3858 { \ 3859 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3860 } 3861 3862 #if defined(INET) 3863 static int 3864 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3865 { 3866 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3867 } 3868 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3869 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3870 #endif 3871 3872 #if defined(INET6) 3873 static int 3874 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3875 { 3876 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3877 } 3878 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3879 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3880 #endif 3881 3882 static void 3883 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3884 { 3885 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3886 const char *pfname = NULL, *ipname = NULL; 3887 int ipn = 0, qid = 0; 3888 3889 switch (pf) { 3890 #if defined(INET) 3891 case PF_INET: 3892 len_func = sysctl_net_ip_pktq_items; 3893 maxlen_func = sysctl_net_ip_pktq_maxlen; 3894 drops_func = sysctl_net_ip_pktq_drops; 3895 pfname = "inet", ipn = IPPROTO_IP; 3896 ipname = "ip", qid = IPCTL_IFQ; 3897 break; 3898 #endif 3899 #if defined(INET6) 3900 case PF_INET6: 3901 len_func = sysctl_net_ip6_pktq_items; 3902 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3903 drops_func = sysctl_net_ip6_pktq_drops; 3904 pfname = "inet6", ipn = IPPROTO_IPV6; 3905 ipname = "ip6", qid = IPV6CTL_IFQ; 3906 break; 3907 #endif 3908 default: 3909 KASSERT(false); 3910 } 3911 3912 sysctl_createv(clog, 0, NULL, NULL, 3913 CTLFLAG_PERMANENT, 3914 CTLTYPE_NODE, pfname, NULL, 3915 NULL, 0, NULL, 0, 3916 CTL_NET, pf, CTL_EOL); 3917 sysctl_createv(clog, 0, NULL, NULL, 3918 CTLFLAG_PERMANENT, 3919 CTLTYPE_NODE, ipname, NULL, 3920 NULL, 0, NULL, 0, 3921 CTL_NET, pf, ipn, CTL_EOL); 3922 sysctl_createv(clog, 0, NULL, NULL, 3923 CTLFLAG_PERMANENT, 3924 CTLTYPE_NODE, "ifq", 3925 SYSCTL_DESCR("Protocol input queue controls"), 3926 NULL, 0, NULL, 0, 3927 CTL_NET, pf, ipn, qid, CTL_EOL); 3928 3929 sysctl_createv(clog, 0, NULL, NULL, 3930 CTLFLAG_PERMANENT, 3931 CTLTYPE_QUAD, "len", 3932 SYSCTL_DESCR("Current input queue length"), 3933 len_func, 0, NULL, 0, 3934 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3935 sysctl_createv(clog, 0, NULL, NULL, 3936 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 3937 CTLTYPE_INT, "maxlen", 3938 SYSCTL_DESCR("Maximum allowed input queue length"), 3939 maxlen_func, 0, NULL, 0, 3940 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3941 sysctl_createv(clog, 0, NULL, NULL, 3942 CTLFLAG_PERMANENT, 3943 CTLTYPE_QUAD, "drops", 3944 SYSCTL_DESCR("Packets dropped due to full input queue"), 3945 drops_func, 0, NULL, 0, 3946 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3947 } 3948 #endif /* INET || INET6 */ 3949 3950 static int 3951 if_sdl_sysctl(SYSCTLFN_ARGS) 3952 { 3953 struct ifnet *ifp; 3954 const struct sockaddr_dl *sdl; 3955 struct psref psref; 3956 int error = 0; 3957 int bound; 3958 3959 if (namelen != 1) 3960 return EINVAL; 3961 3962 bound = curlwp_bind(); 3963 ifp = if_get_byindex(name[0], &psref); 3964 if (ifp == NULL) { 3965 error = ENODEV; 3966 goto out0; 3967 } 3968 3969 sdl = ifp->if_sadl; 3970 if (sdl == NULL) { 3971 *oldlenp = 0; 3972 goto out1; 3973 } 3974 3975 if (oldp == NULL) { 3976 *oldlenp = sdl->sdl_alen; 3977 goto out1; 3978 } 3979 3980 if (*oldlenp >= sdl->sdl_alen) 3981 *oldlenp = sdl->sdl_alen; 3982 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3983 out1: 3984 if_put(ifp, &psref); 3985 out0: 3986 curlwp_bindx(bound); 3987 return error; 3988 } 3989 3990 static void 3991 if_sysctl_setup(struct sysctllog **clog) 3992 { 3993 const struct sysctlnode *rnode = NULL; 3994 3995 sysctl_createv(clog, 0, NULL, &rnode, 3996 CTLFLAG_PERMANENT, 3997 CTLTYPE_NODE, "sdl", 3998 SYSCTL_DESCR("Get active link-layer address"), 3999 if_sdl_sysctl, 0, NULL, 0, 4000 CTL_NET, CTL_CREATE, CTL_EOL); 4001 4002 #if defined(INET) 4003 sysctl_net_pktq_setup(NULL, PF_INET); 4004 #endif 4005 #ifdef INET6 4006 if (in6_present) 4007 sysctl_net_pktq_setup(NULL, PF_INET6); 4008 #endif 4009 } 4010