if.c revision 1.361
1 /* $NetBSD: if.c,v 1.361 2016/11/05 23:30:22 pgoyette 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 1.361 2016/11/05 23:30:22 pgoyette Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 #include "opt_ipsec.h" 98 99 #include "opt_atalk.h" 100 #include "opt_natm.h" 101 #include "opt_wlan.h" 102 #include "opt_net_mpsafe.h" 103 #endif 104 105 #include <sys/param.h> 106 #include <sys/mbuf.h> 107 #include <sys/systm.h> 108 #include <sys/callout.h> 109 #include <sys/proc.h> 110 #include <sys/socket.h> 111 #include <sys/socketvar.h> 112 #include <sys/domain.h> 113 #include <sys/protosw.h> 114 #include <sys/kernel.h> 115 #include <sys/ioctl.h> 116 #include <sys/sysctl.h> 117 #include <sys/syslog.h> 118 #include <sys/kauth.h> 119 #include <sys/kmem.h> 120 #include <sys/xcall.h> 121 #include <sys/cpu.h> 122 #include <sys/intr.h> 123 124 #include <net/if.h> 125 #include <net/if_dl.h> 126 #include <net/if_ether.h> 127 #include <net/if_media.h> 128 #include <net80211/ieee80211.h> 129 #include <net80211/ieee80211_ioctl.h> 130 #include <net/if_types.h> 131 #include <net/route.h> 132 #include <net/netisr.h> 133 #include <sys/module.h> 134 #ifdef NETATALK 135 #include <netatalk/at_extern.h> 136 #include <netatalk/at.h> 137 #endif 138 #include <net/pfil.h> 139 #include <netinet/in.h> 140 #include <netinet/in_var.h> 141 #ifndef IPSEC 142 #include <netinet/ip_encap.h> 143 #endif 144 145 #ifdef INET6 146 #include <netinet6/in6_var.h> 147 #include <netinet6/nd6.h> 148 #endif 149 150 #include "ether.h" 151 #include "fddi.h" 152 #include "token.h" 153 154 #include "carp.h" 155 #if NCARP > 0 156 #include <netinet/ip_carp.h> 157 #endif 158 159 #include <compat/sys/sockio.h> 160 #include <compat/sys/socket.h> 161 162 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 163 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 164 165 /* 166 * Global list of interfaces. 167 */ 168 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 169 struct ifnet_head ifnet_list; 170 171 struct pslist_head ifnet_pslist; 172 static ifnet_t ** ifindex2ifnet = NULL; 173 static u_int if_index = 1; 174 static size_t if_indexlim = 0; 175 static uint64_t index_gen; 176 /* Mutex to protect the above objects. */ 177 kmutex_t ifnet_mtx __cacheline_aligned; 178 struct psref_class *ifnet_psref_class __read_mostly; 179 static pserialize_t ifnet_psz; 180 181 static kmutex_t if_clone_mtx; 182 183 struct ifnet *lo0ifp; 184 int ifqmaxlen = IFQ_MAXLEN; 185 186 struct psref_class *ifa_psref_class __read_mostly; 187 188 static int if_rt_walktree(struct rtentry *, void *); 189 190 static struct if_clone *if_clone_lookup(const char *, int *); 191 192 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 193 static int if_cloners_count; 194 195 /* Packet filtering hook for interfaces. */ 196 pfil_head_t * if_pfil; 197 198 static kauth_listener_t if_listener; 199 200 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 201 static void if_detach_queues(struct ifnet *, struct ifqueue *); 202 static void sysctl_sndq_setup(struct sysctllog **, const char *, 203 struct ifaltq *); 204 static void if_slowtimo(void *); 205 static void if_free_sadl(struct ifnet *); 206 static void if_attachdomain1(struct ifnet *); 207 static int ifconf(u_long, void *); 208 static int if_transmit(struct ifnet *, struct mbuf *); 209 static int if_clone_create(const char *); 210 static int if_clone_destroy(const char *); 211 static void if_link_state_change_si(void *); 212 213 struct if_percpuq { 214 struct ifnet *ipq_ifp; 215 void *ipq_si; 216 struct percpu *ipq_ifqs; /* struct ifqueue */ 217 }; 218 219 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 220 221 static void if_percpuq_drops(void *, void *, struct cpu_info *); 222 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 223 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 224 struct if_percpuq *); 225 226 #if defined(INET) || defined(INET6) 227 static void sysctl_net_pktq_setup(struct sysctllog **, int); 228 #endif 229 230 /* 231 * Pointer to stub or real compat_cvtcmd() depending on presence of 232 * the compat module 233 */ 234 u_long stub_compat_cvtcmd(u_long); 235 u_long (*vec_compat_cvtcmd)(u_long) = stub_compat_cvtcmd; 236 237 /* Similarly, pointer to compat_ifioctl() if it is present */ 238 239 int (*vec_compat_ifioctl)(struct socket *, u_long, u_long, void *, 240 struct lwp *) = NULL; 241 242 /* The stub version of compat_cvtcmd() */ 243 u_long stub_compat_cvtcmd(u_long cmd) 244 { 245 246 return cmd; 247 } 248 249 static int 250 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 251 void *arg0, void *arg1, void *arg2, void *arg3) 252 { 253 int result; 254 enum kauth_network_req req; 255 256 result = KAUTH_RESULT_DEFER; 257 req = (enum kauth_network_req)arg1; 258 259 if (action != KAUTH_NETWORK_INTERFACE) 260 return result; 261 262 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 263 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 264 result = KAUTH_RESULT_ALLOW; 265 266 return result; 267 } 268 269 /* 270 * Network interface utility routines. 271 * 272 * Routines with ifa_ifwith* names take sockaddr *'s as 273 * parameters. 274 */ 275 void 276 ifinit(void) 277 { 278 #if defined(INET) 279 sysctl_net_pktq_setup(NULL, PF_INET); 280 #endif 281 #ifdef INET6 282 if (in6_present) 283 sysctl_net_pktq_setup(NULL, PF_INET6); 284 #endif 285 286 #if (defined(INET) || defined(INET6)) && !defined(IPSEC) 287 encapinit(); 288 #endif 289 290 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 291 if_listener_cb, NULL); 292 293 /* interfaces are available, inform socket code */ 294 ifioctl = doifioctl; 295 } 296 297 /* 298 * XXX Initialization before configure(). 299 * XXX hack to get pfil_add_hook working in autoconf. 300 */ 301 void 302 ifinit1(void) 303 { 304 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 305 306 TAILQ_INIT(&ifnet_list); 307 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 308 ifnet_psz = pserialize_create(); 309 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 310 ifa_psref_class = psref_class_create("ifa", IPL_SOFTNET); 311 PSLIST_INIT(&ifnet_pslist); 312 313 if_indexlim = 8; 314 315 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 316 KASSERT(if_pfil != NULL); 317 318 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 319 etherinit(); 320 #endif 321 } 322 323 ifnet_t * 324 if_alloc(u_char type) 325 { 326 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 327 } 328 329 void 330 if_free(ifnet_t *ifp) 331 { 332 kmem_free(ifp, sizeof(ifnet_t)); 333 } 334 335 void 336 if_initname(struct ifnet *ifp, const char *name, int unit) 337 { 338 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 339 "%s%d", name, unit); 340 } 341 342 /* 343 * Null routines used while an interface is going away. These routines 344 * just return an error. 345 */ 346 347 int 348 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 349 const struct sockaddr *so, const struct rtentry *rt) 350 { 351 352 return ENXIO; 353 } 354 355 void 356 if_nullinput(struct ifnet *ifp, struct mbuf *m) 357 { 358 359 /* Nothing. */ 360 } 361 362 void 363 if_nullstart(struct ifnet *ifp) 364 { 365 366 /* Nothing. */ 367 } 368 369 int 370 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 371 { 372 373 return ENXIO; 374 } 375 376 int 377 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 378 { 379 380 return ENXIO; 381 } 382 383 int 384 if_nullinit(struct ifnet *ifp) 385 { 386 387 return ENXIO; 388 } 389 390 void 391 if_nullstop(struct ifnet *ifp, int disable) 392 { 393 394 /* Nothing. */ 395 } 396 397 void 398 if_nullslowtimo(struct ifnet *ifp) 399 { 400 401 /* Nothing. */ 402 } 403 404 void 405 if_nulldrain(struct ifnet *ifp) 406 { 407 408 /* Nothing. */ 409 } 410 411 void 412 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 413 { 414 struct ifaddr *ifa; 415 struct sockaddr_dl *sdl; 416 417 ifp->if_addrlen = addrlen; 418 if_alloc_sadl(ifp); 419 ifa = ifp->if_dl; 420 sdl = satosdl(ifa->ifa_addr); 421 422 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 423 if (factory) { 424 ifp->if_hwdl = ifp->if_dl; 425 ifaref(ifp->if_hwdl); 426 } 427 /* TBD routing socket */ 428 } 429 430 struct ifaddr * 431 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 432 { 433 unsigned socksize, ifasize; 434 int addrlen, namelen; 435 struct sockaddr_dl *mask, *sdl; 436 struct ifaddr *ifa; 437 438 namelen = strlen(ifp->if_xname); 439 addrlen = ifp->if_addrlen; 440 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 441 ifasize = sizeof(*ifa) + 2 * socksize; 442 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 443 444 sdl = (struct sockaddr_dl *)(ifa + 1); 445 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 446 447 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 448 ifp->if_xname, namelen, NULL, addrlen); 449 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 450 memset(&mask->sdl_data[0], 0xff, namelen); 451 ifa->ifa_rtrequest = link_rtrequest; 452 ifa->ifa_addr = (struct sockaddr *)sdl; 453 ifa->ifa_netmask = (struct sockaddr *)mask; 454 ifa_psref_init(ifa); 455 456 *sdlp = sdl; 457 458 return ifa; 459 } 460 461 static void 462 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 463 { 464 const struct sockaddr_dl *sdl; 465 466 ifp->if_dl = ifa; 467 ifaref(ifa); 468 sdl = satosdl(ifa->ifa_addr); 469 ifp->if_sadl = sdl; 470 } 471 472 /* 473 * Allocate the link level name for the specified interface. This 474 * is an attachment helper. It must be called after ifp->if_addrlen 475 * is initialized, which may not be the case when if_attach() is 476 * called. 477 */ 478 void 479 if_alloc_sadl(struct ifnet *ifp) 480 { 481 struct ifaddr *ifa; 482 const struct sockaddr_dl *sdl; 483 484 /* 485 * If the interface already has a link name, release it 486 * now. This is useful for interfaces that can change 487 * link types, and thus switch link names often. 488 */ 489 if (ifp->if_sadl != NULL) 490 if_free_sadl(ifp); 491 492 ifa = if_dl_create(ifp, &sdl); 493 494 ifa_insert(ifp, ifa); 495 if_sadl_setrefs(ifp, ifa); 496 } 497 498 static void 499 if_deactivate_sadl(struct ifnet *ifp) 500 { 501 struct ifaddr *ifa; 502 503 KASSERT(ifp->if_dl != NULL); 504 505 ifa = ifp->if_dl; 506 507 ifp->if_sadl = NULL; 508 509 ifp->if_dl = NULL; 510 ifafree(ifa); 511 } 512 513 void 514 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa0, 515 const struct sockaddr_dl *sdl) 516 { 517 int s, ss; 518 struct ifaddr *ifa; 519 int bound = curlwp_bind(); 520 521 s = splnet(); 522 523 if_deactivate_sadl(ifp); 524 525 if_sadl_setrefs(ifp, ifa0); 526 527 ss = pserialize_read_enter(); 528 IFADDR_READER_FOREACH(ifa, ifp) { 529 struct psref psref; 530 ifa_acquire(ifa, &psref); 531 pserialize_read_exit(ss); 532 533 rtinit(ifa, RTM_LLINFO_UPD, 0); 534 535 ss = pserialize_read_enter(); 536 ifa_release(ifa, &psref); 537 } 538 pserialize_read_exit(ss); 539 540 splx(s); 541 curlwp_bindx(bound); 542 } 543 544 /* 545 * Free the link level name for the specified interface. This is 546 * a detach helper. This is called from if_detach(). 547 */ 548 static void 549 if_free_sadl(struct ifnet *ifp) 550 { 551 struct ifaddr *ifa; 552 int s; 553 554 ifa = ifp->if_dl; 555 if (ifa == NULL) { 556 KASSERT(ifp->if_sadl == NULL); 557 return; 558 } 559 560 KASSERT(ifp->if_sadl != NULL); 561 562 s = splnet(); 563 rtinit(ifa, RTM_DELETE, 0); 564 ifa_remove(ifp, ifa); 565 if_deactivate_sadl(ifp); 566 if (ifp->if_hwdl == ifa) { 567 ifafree(ifa); 568 ifp->if_hwdl = NULL; 569 } 570 splx(s); 571 } 572 573 static void 574 if_getindex(ifnet_t *ifp) 575 { 576 bool hitlimit = false; 577 578 ifp->if_index_gen = index_gen++; 579 580 ifp->if_index = if_index; 581 if (ifindex2ifnet == NULL) { 582 if_index++; 583 goto skip; 584 } 585 while (if_byindex(ifp->if_index)) { 586 /* 587 * If we hit USHRT_MAX, we skip back to 0 since 588 * there are a number of places where the value 589 * of if_index or if_index itself is compared 590 * to or stored in an unsigned short. By 591 * jumping back, we won't botch those assignments 592 * or comparisons. 593 */ 594 if (++if_index == 0) { 595 if_index = 1; 596 } else if (if_index == USHRT_MAX) { 597 /* 598 * However, if we have to jump back to 599 * zero *twice* without finding an empty 600 * slot in ifindex2ifnet[], then there 601 * there are too many (>65535) interfaces. 602 */ 603 if (hitlimit) { 604 panic("too many interfaces"); 605 } 606 hitlimit = true; 607 if_index = 1; 608 } 609 ifp->if_index = if_index; 610 } 611 skip: 612 /* 613 * ifindex2ifnet is indexed by if_index. Since if_index will 614 * grow dynamically, it should grow too. 615 */ 616 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 617 size_t m, n, oldlim; 618 void *q; 619 620 oldlim = if_indexlim; 621 while (ifp->if_index >= if_indexlim) 622 if_indexlim <<= 1; 623 624 /* grow ifindex2ifnet */ 625 m = oldlim * sizeof(struct ifnet *); 626 n = if_indexlim * sizeof(struct ifnet *); 627 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 628 if (ifindex2ifnet != NULL) { 629 memcpy(q, ifindex2ifnet, m); 630 free(ifindex2ifnet, M_IFADDR); 631 } 632 ifindex2ifnet = (struct ifnet **)q; 633 } 634 ifindex2ifnet[ifp->if_index] = ifp; 635 } 636 637 /* 638 * Initialize an interface and assign an index for it. 639 * 640 * It must be called prior to a device specific attach routine 641 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 642 * and be followed by if_register: 643 * 644 * if_initialize(ifp); 645 * ether_ifattach(ifp, enaddr); 646 * if_register(ifp); 647 */ 648 void 649 if_initialize(ifnet_t *ifp) 650 { 651 KASSERT(if_indexlim > 0); 652 TAILQ_INIT(&ifp->if_addrlist); 653 654 /* 655 * Link level name is allocated later by a separate call to 656 * if_alloc_sadl(). 657 */ 658 659 if (ifp->if_snd.ifq_maxlen == 0) 660 ifp->if_snd.ifq_maxlen = ifqmaxlen; 661 662 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 663 664 ifp->if_link_state = LINK_STATE_UNKNOWN; 665 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 666 667 ifp->if_capenable = 0; 668 ifp->if_csum_flags_tx = 0; 669 ifp->if_csum_flags_rx = 0; 670 671 #ifdef ALTQ 672 ifp->if_snd.altq_type = 0; 673 ifp->if_snd.altq_disc = NULL; 674 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 675 ifp->if_snd.altq_tbr = NULL; 676 ifp->if_snd.altq_ifp = ifp; 677 #endif 678 679 IFQ_LOCK_INIT(&ifp->if_snd); 680 681 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 682 (void)pfil_run_hooks(if_pfil, 683 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 684 685 IF_AFDATA_LOCK_INIT(ifp); 686 687 if (if_is_link_state_changeable(ifp)) { 688 ifp->if_link_si = softint_establish(SOFTINT_NET, 689 if_link_state_change_si, ifp); 690 if (ifp->if_link_si == NULL) 691 panic("%s: softint_establish() failed", __func__); 692 } 693 694 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 695 PSLIST_INIT(&ifp->if_addr_pslist); 696 psref_target_init(&ifp->if_psref, ifnet_psref_class); 697 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 698 699 IFNET_LOCK(); 700 if_getindex(ifp); 701 IFNET_UNLOCK(); 702 } 703 704 /* 705 * Register an interface to the list of "active" interfaces. 706 */ 707 void 708 if_register(ifnet_t *ifp) 709 { 710 /* 711 * If the driver has not supplied its own if_ioctl, then 712 * supply the default. 713 */ 714 if (ifp->if_ioctl == NULL) 715 ifp->if_ioctl = ifioctl_common; 716 717 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 718 719 if (!STAILQ_EMPTY(&domains)) 720 if_attachdomain1(ifp); 721 722 /* Announce the interface. */ 723 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 724 725 if (ifp->if_slowtimo != NULL) { 726 ifp->if_slowtimo_ch = 727 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 728 callout_init(ifp->if_slowtimo_ch, 0); 729 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 730 if_slowtimo(ifp); 731 } 732 733 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 734 ifp->if_transmit = if_transmit; 735 736 IFNET_LOCK(); 737 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 738 IFNET_WRITER_INSERT_TAIL(ifp); 739 IFNET_UNLOCK(); 740 } 741 742 /* 743 * The if_percpuq framework 744 * 745 * It allows network device drivers to execute the network stack 746 * in softint (so called softint-based if_input). It utilizes 747 * softint and percpu ifqueue. It doesn't distribute any packets 748 * between CPUs, unlike pktqueue(9). 749 * 750 * Currently we support two options for device drivers to apply the framework: 751 * - Use it implicitly with less changes 752 * - If you use if_attach in driver's _attach function and if_input in 753 * driver's Rx interrupt handler, a packet is queued and a softint handles 754 * the packet implicitly 755 * - Use it explicitly in each driver (recommended) 756 * - You can use if_percpuq_* directly in your driver 757 * - In this case, you need to allocate struct if_percpuq in driver's softc 758 * - See wm(4) as a reference implementation 759 */ 760 761 static void 762 if_percpuq_softint(void *arg) 763 { 764 struct if_percpuq *ipq = arg; 765 struct ifnet *ifp = ipq->ipq_ifp; 766 struct mbuf *m; 767 768 while ((m = if_percpuq_dequeue(ipq)) != NULL) 769 ifp->_if_input(ifp, m); 770 } 771 772 static void 773 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 774 { 775 struct ifqueue *const ifq = p; 776 777 memset(ifq, 0, sizeof(*ifq)); 778 ifq->ifq_maxlen = IFQ_MAXLEN; 779 } 780 781 struct if_percpuq * 782 if_percpuq_create(struct ifnet *ifp) 783 { 784 struct if_percpuq *ipq; 785 786 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 787 if (ipq == NULL) 788 panic("kmem_zalloc failed"); 789 790 ipq->ipq_ifp = ifp; 791 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 792 if_percpuq_softint, ipq); 793 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 794 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 795 796 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 797 798 return ipq; 799 } 800 801 static struct mbuf * 802 if_percpuq_dequeue(struct if_percpuq *ipq) 803 { 804 struct mbuf *m; 805 struct ifqueue *ifq; 806 int s; 807 808 s = splnet(); 809 ifq = percpu_getref(ipq->ipq_ifqs); 810 IF_DEQUEUE(ifq, m); 811 percpu_putref(ipq->ipq_ifqs); 812 splx(s); 813 814 return m; 815 } 816 817 static void 818 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 819 { 820 struct ifqueue *const ifq = p; 821 822 IF_PURGE(ifq); 823 } 824 825 void 826 if_percpuq_destroy(struct if_percpuq *ipq) 827 { 828 829 /* if_detach may already destroy it */ 830 if (ipq == NULL) 831 return; 832 833 softint_disestablish(ipq->ipq_si); 834 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 835 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 836 } 837 838 void 839 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 840 { 841 struct ifqueue *ifq; 842 int s; 843 844 KASSERT(ipq != NULL); 845 846 s = splnet(); 847 ifq = percpu_getref(ipq->ipq_ifqs); 848 if (IF_QFULL(ifq)) { 849 IF_DROP(ifq); 850 percpu_putref(ipq->ipq_ifqs); 851 m_freem(m); 852 goto out; 853 } 854 IF_ENQUEUE(ifq, m); 855 percpu_putref(ipq->ipq_ifqs); 856 857 softint_schedule(ipq->ipq_si); 858 out: 859 splx(s); 860 } 861 862 static void 863 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 864 { 865 struct ifqueue *const ifq = p; 866 int *sum = arg; 867 868 *sum += ifq->ifq_drops; 869 } 870 871 static int 872 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 873 { 874 struct sysctlnode node; 875 struct if_percpuq *ipq; 876 int sum = 0; 877 int error; 878 879 node = *rnode; 880 ipq = node.sysctl_data; 881 882 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 883 884 node.sysctl_data = ∑ 885 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 886 if (error != 0 || newp == NULL) 887 return error; 888 889 return 0; 890 } 891 892 static void 893 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 894 struct if_percpuq *ipq) 895 { 896 const struct sysctlnode *cnode, *rnode; 897 898 if (sysctl_createv(clog, 0, NULL, &rnode, 899 CTLFLAG_PERMANENT, 900 CTLTYPE_NODE, "interfaces", 901 SYSCTL_DESCR("Per-interface controls"), 902 NULL, 0, NULL, 0, 903 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 904 goto bad; 905 906 if (sysctl_createv(clog, 0, &rnode, &rnode, 907 CTLFLAG_PERMANENT, 908 CTLTYPE_NODE, ifname, 909 SYSCTL_DESCR("Interface controls"), 910 NULL, 0, NULL, 0, 911 CTL_CREATE, CTL_EOL) != 0) 912 goto bad; 913 914 if (sysctl_createv(clog, 0, &rnode, &rnode, 915 CTLFLAG_PERMANENT, 916 CTLTYPE_NODE, "rcvq", 917 SYSCTL_DESCR("Interface input queue controls"), 918 NULL, 0, NULL, 0, 919 CTL_CREATE, CTL_EOL) != 0) 920 goto bad; 921 922 #ifdef NOTYET 923 /* XXX Should show each per-CPU queue length? */ 924 if (sysctl_createv(clog, 0, &rnode, &rnode, 925 CTLFLAG_PERMANENT, 926 CTLTYPE_INT, "len", 927 SYSCTL_DESCR("Current input queue length"), 928 sysctl_percpuq_len, 0, NULL, 0, 929 CTL_CREATE, CTL_EOL) != 0) 930 goto bad; 931 932 if (sysctl_createv(clog, 0, &rnode, &cnode, 933 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 934 CTLTYPE_INT, "maxlen", 935 SYSCTL_DESCR("Maximum allowed input queue length"), 936 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 937 CTL_CREATE, CTL_EOL) != 0) 938 goto bad; 939 #endif 940 941 if (sysctl_createv(clog, 0, &rnode, &cnode, 942 CTLFLAG_PERMANENT, 943 CTLTYPE_INT, "drops", 944 SYSCTL_DESCR("Total packets dropped due to full input queue"), 945 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 946 CTL_CREATE, CTL_EOL) != 0) 947 goto bad; 948 949 return; 950 bad: 951 printf("%s: could not attach sysctl nodes\n", ifname); 952 return; 953 } 954 955 956 /* 957 * The common interface input routine that is called by device drivers, 958 * which should be used only when the driver's rx handler already runs 959 * in softint. 960 */ 961 void 962 if_input(struct ifnet *ifp, struct mbuf *m) 963 { 964 965 KASSERT(ifp->if_percpuq == NULL); 966 KASSERT(!cpu_intr_p()); 967 968 ifp->_if_input(ifp, m); 969 } 970 971 /* 972 * DEPRECATED. Use if_initialize and if_register instead. 973 * See the above comment of if_initialize. 974 * 975 * Note that it implicitly enables if_percpuq to make drivers easy to 976 * migrate softint-based if_input without much changes. If you don't 977 * want to enable it, use if_initialize instead. 978 */ 979 void 980 if_attach(ifnet_t *ifp) 981 { 982 983 if_initialize(ifp); 984 ifp->if_percpuq = if_percpuq_create(ifp); 985 if_register(ifp); 986 } 987 988 void 989 if_attachdomain(void) 990 { 991 struct ifnet *ifp; 992 int s; 993 int bound = curlwp_bind(); 994 995 s = pserialize_read_enter(); 996 IFNET_READER_FOREACH(ifp) { 997 struct psref psref; 998 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 999 pserialize_read_exit(s); 1000 if_attachdomain1(ifp); 1001 s = pserialize_read_enter(); 1002 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 1003 } 1004 pserialize_read_exit(s); 1005 curlwp_bindx(bound); 1006 } 1007 1008 static void 1009 if_attachdomain1(struct ifnet *ifp) 1010 { 1011 struct domain *dp; 1012 int s; 1013 1014 s = splnet(); 1015 1016 /* address family dependent data region */ 1017 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 1018 DOMAIN_FOREACH(dp) { 1019 if (dp->dom_ifattach != NULL) 1020 ifp->if_afdata[dp->dom_family] = 1021 (*dp->dom_ifattach)(ifp); 1022 } 1023 1024 splx(s); 1025 } 1026 1027 /* 1028 * Deactivate an interface. This points all of the procedure 1029 * handles at error stubs. May be called from interrupt context. 1030 */ 1031 void 1032 if_deactivate(struct ifnet *ifp) 1033 { 1034 int s; 1035 1036 s = splnet(); 1037 1038 ifp->if_output = if_nulloutput; 1039 ifp->_if_input = if_nullinput; 1040 ifp->if_start = if_nullstart; 1041 ifp->if_transmit = if_nulltransmit; 1042 ifp->if_ioctl = if_nullioctl; 1043 ifp->if_init = if_nullinit; 1044 ifp->if_stop = if_nullstop; 1045 ifp->if_slowtimo = if_nullslowtimo; 1046 ifp->if_drain = if_nulldrain; 1047 1048 /* No more packets may be enqueued. */ 1049 ifp->if_snd.ifq_maxlen = 0; 1050 1051 splx(s); 1052 } 1053 1054 bool 1055 if_is_deactivated(struct ifnet *ifp) 1056 { 1057 1058 return ifp->if_output == if_nulloutput; 1059 } 1060 1061 void 1062 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1063 { 1064 struct ifaddr *ifa, *nifa; 1065 int s; 1066 1067 s = pserialize_read_enter(); 1068 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) { 1069 nifa = IFADDR_READER_NEXT(ifa); 1070 if (ifa->ifa_addr->sa_family != family) 1071 continue; 1072 pserialize_read_exit(s); 1073 1074 (*purgeaddr)(ifa); 1075 1076 s = pserialize_read_enter(); 1077 } 1078 pserialize_read_exit(s); 1079 } 1080 1081 #ifdef IFAREF_DEBUG 1082 static struct ifaddr **ifa_list; 1083 static int ifa_list_size; 1084 1085 /* Depends on only one if_attach runs at once */ 1086 static void 1087 if_build_ifa_list(struct ifnet *ifp) 1088 { 1089 struct ifaddr *ifa; 1090 int i; 1091 1092 KASSERT(ifa_list == NULL); 1093 KASSERT(ifa_list_size == 0); 1094 1095 IFADDR_READER_FOREACH(ifa, ifp) 1096 ifa_list_size++; 1097 1098 ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP); 1099 if (ifa_list == NULL) 1100 return; 1101 1102 i = 0; 1103 IFADDR_READER_FOREACH(ifa, ifp) { 1104 ifa_list[i++] = ifa; 1105 ifaref(ifa); 1106 } 1107 } 1108 1109 static void 1110 if_check_and_free_ifa_list(struct ifnet *ifp) 1111 { 1112 int i; 1113 struct ifaddr *ifa; 1114 1115 if (ifa_list == NULL) 1116 return; 1117 1118 for (i = 0; i < ifa_list_size; i++) { 1119 char buf[64]; 1120 1121 ifa = ifa_list[i]; 1122 sockaddr_format(ifa->ifa_addr, buf, sizeof(buf)); 1123 if (ifa->ifa_refcnt > 1) { 1124 log(LOG_WARNING, 1125 "ifa(%s) still referenced (refcnt=%d)\n", 1126 buf, ifa->ifa_refcnt - 1); 1127 } else 1128 log(LOG_DEBUG, 1129 "ifa(%s) not referenced (refcnt=%d)\n", 1130 buf, ifa->ifa_refcnt - 1); 1131 ifafree(ifa); 1132 } 1133 1134 kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size); 1135 ifa_list = NULL; 1136 ifa_list_size = 0; 1137 } 1138 #endif 1139 1140 /* 1141 * Detach an interface from the list of "active" interfaces, 1142 * freeing any resources as we go along. 1143 * 1144 * NOTE: This routine must be called with a valid thread context, 1145 * as it may block. 1146 */ 1147 void 1148 if_detach(struct ifnet *ifp) 1149 { 1150 struct socket so; 1151 struct ifaddr *ifa; 1152 #ifdef IFAREF_DEBUG 1153 struct ifaddr *last_ifa = NULL; 1154 #endif 1155 struct domain *dp; 1156 const struct protosw *pr; 1157 int s, i, family, purged; 1158 uint64_t xc; 1159 1160 #ifdef IFAREF_DEBUG 1161 if_build_ifa_list(ifp); 1162 #endif 1163 /* 1164 * XXX It's kind of lame that we have to have the 1165 * XXX socket structure... 1166 */ 1167 memset(&so, 0, sizeof(so)); 1168 1169 s = splnet(); 1170 1171 sysctl_teardown(&ifp->if_sysctl_log); 1172 mutex_enter(ifp->if_ioctl_lock); 1173 if_deactivate(ifp); 1174 mutex_exit(ifp->if_ioctl_lock); 1175 1176 IFNET_LOCK(); 1177 ifindex2ifnet[ifp->if_index] = NULL; 1178 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1179 IFNET_WRITER_REMOVE(ifp); 1180 pserialize_perform(ifnet_psz); 1181 IFNET_UNLOCK(); 1182 1183 /* Wait for all readers to drain before freeing. */ 1184 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1185 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1186 1187 mutex_obj_free(ifp->if_ioctl_lock); 1188 ifp->if_ioctl_lock = NULL; 1189 1190 if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) { 1191 ifp->if_slowtimo = NULL; 1192 callout_halt(ifp->if_slowtimo_ch, NULL); 1193 callout_destroy(ifp->if_slowtimo_ch); 1194 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1195 } 1196 1197 /* 1198 * Do an if_down() to give protocols a chance to do something. 1199 */ 1200 if_down(ifp); 1201 1202 #ifdef ALTQ 1203 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1204 altq_disable(&ifp->if_snd); 1205 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1206 altq_detach(&ifp->if_snd); 1207 #endif 1208 1209 mutex_obj_free(ifp->if_snd.ifq_lock); 1210 1211 #if NCARP > 0 1212 /* Remove the interface from any carp group it is a part of. */ 1213 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1214 carp_ifdetach(ifp); 1215 #endif 1216 1217 /* 1218 * Rip all the addresses off the interface. This should make 1219 * all of the routes go away. 1220 * 1221 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1222 * from the list, including our "cursor", ifa. For safety, 1223 * and to honor the TAILQ abstraction, I just restart the 1224 * loop after each removal. Note that the loop will exit 1225 * when all of the remaining ifaddrs belong to the AF_LINK 1226 * family. I am counting on the historical fact that at 1227 * least one pr_usrreq in each address domain removes at 1228 * least one ifaddr. 1229 */ 1230 again: 1231 /* 1232 * At this point, no other one tries to remove ifa in the list, 1233 * so we don't need to take a lock or psref. 1234 */ 1235 IFADDR_READER_FOREACH(ifa, ifp) { 1236 family = ifa->ifa_addr->sa_family; 1237 #ifdef IFAREF_DEBUG 1238 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1239 ifa, family, ifa->ifa_refcnt); 1240 if (last_ifa != NULL && ifa == last_ifa) 1241 panic("if_detach: loop detected"); 1242 last_ifa = ifa; 1243 #endif 1244 if (family == AF_LINK) 1245 continue; 1246 dp = pffinddomain(family); 1247 #ifdef DIAGNOSTIC 1248 if (dp == NULL) 1249 panic("if_detach: no domain for AF %d", 1250 family); 1251 #endif 1252 /* 1253 * XXX These PURGEIF calls are redundant with the 1254 * purge-all-families calls below, but are left in for 1255 * now both to make a smaller change, and to avoid 1256 * unplanned interactions with clearing of 1257 * ifp->if_addrlist. 1258 */ 1259 purged = 0; 1260 for (pr = dp->dom_protosw; 1261 pr < dp->dom_protoswNPROTOSW; pr++) { 1262 so.so_proto = pr; 1263 if (pr->pr_usrreqs) { 1264 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1265 purged = 1; 1266 } 1267 } 1268 if (purged == 0) { 1269 /* 1270 * XXX What's really the best thing to do 1271 * XXX here? --thorpej (at) NetBSD.org 1272 */ 1273 printf("if_detach: WARNING: AF %d not purged\n", 1274 family); 1275 ifa_remove(ifp, ifa); 1276 } 1277 goto again; 1278 } 1279 1280 if_free_sadl(ifp); 1281 1282 /* Walk the routing table looking for stragglers. */ 1283 for (i = 0; i <= AF_MAX; i++) { 1284 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 1285 continue; 1286 } 1287 1288 DOMAIN_FOREACH(dp) { 1289 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1290 { 1291 void *p = ifp->if_afdata[dp->dom_family]; 1292 if (p) { 1293 ifp->if_afdata[dp->dom_family] = NULL; 1294 (*dp->dom_ifdetach)(ifp, p); 1295 } 1296 } 1297 1298 /* 1299 * One would expect multicast memberships (INET and 1300 * INET6) on UDP sockets to be purged by the PURGEIF 1301 * calls above, but if all addresses were removed from 1302 * the interface prior to destruction, the calls will 1303 * not be made (e.g. ppp, for which pppd(8) generally 1304 * removes addresses before destroying the interface). 1305 * Because there is no invariant that multicast 1306 * memberships only exist for interfaces with IPv4 1307 * addresses, we must call PURGEIF regardless of 1308 * addresses. (Protocols which might store ifnet 1309 * pointers are marked with PR_PURGEIF.) 1310 */ 1311 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1312 so.so_proto = pr; 1313 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1314 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1315 } 1316 } 1317 1318 (void)pfil_run_hooks(if_pfil, 1319 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 1320 (void)pfil_head_destroy(ifp->if_pfil); 1321 1322 /* Announce that the interface is gone. */ 1323 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1324 1325 IF_AFDATA_LOCK_DESTROY(ifp); 1326 1327 if (if_is_link_state_changeable(ifp)) { 1328 softint_disestablish(ifp->if_link_si); 1329 ifp->if_link_si = NULL; 1330 } 1331 1332 /* 1333 * remove packets that came from ifp, from software interrupt queues. 1334 */ 1335 DOMAIN_FOREACH(dp) { 1336 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1337 struct ifqueue *iq = dp->dom_ifqueues[i]; 1338 if (iq == NULL) 1339 break; 1340 dp->dom_ifqueues[i] = NULL; 1341 if_detach_queues(ifp, iq); 1342 } 1343 } 1344 1345 /* 1346 * IP queues have to be processed separately: net-queue barrier 1347 * ensures that the packets are dequeued while a cross-call will 1348 * ensure that the interrupts have completed. FIXME: not quite.. 1349 */ 1350 #ifdef INET 1351 pktq_barrier(ip_pktq); 1352 #endif 1353 #ifdef INET6 1354 if (in6_present) 1355 pktq_barrier(ip6_pktq); 1356 #endif 1357 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1358 xc_wait(xc); 1359 1360 if (ifp->if_percpuq != NULL) { 1361 if_percpuq_destroy(ifp->if_percpuq); 1362 ifp->if_percpuq = NULL; 1363 } 1364 1365 splx(s); 1366 1367 #ifdef IFAREF_DEBUG 1368 if_check_and_free_ifa_list(ifp); 1369 #endif 1370 } 1371 1372 static void 1373 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1374 { 1375 struct mbuf *m, *prev, *next; 1376 1377 prev = NULL; 1378 for (m = q->ifq_head; m != NULL; m = next) { 1379 KASSERT((m->m_flags & M_PKTHDR) != 0); 1380 1381 next = m->m_nextpkt; 1382 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1383 prev = m; 1384 continue; 1385 } 1386 1387 if (prev != NULL) 1388 prev->m_nextpkt = m->m_nextpkt; 1389 else 1390 q->ifq_head = m->m_nextpkt; 1391 if (q->ifq_tail == m) 1392 q->ifq_tail = prev; 1393 q->ifq_len--; 1394 1395 m->m_nextpkt = NULL; 1396 m_freem(m); 1397 IF_DROP(q); 1398 } 1399 } 1400 1401 /* 1402 * Callback for a radix tree walk to delete all references to an 1403 * ifnet. 1404 */ 1405 static int 1406 if_rt_walktree(struct rtentry *rt, void *v) 1407 { 1408 struct ifnet *ifp = (struct ifnet *)v; 1409 int error; 1410 struct rtentry *retrt; 1411 1412 if (rt->rt_ifp != ifp) 1413 return 0; 1414 1415 /* Delete the entry. */ 1416 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 1417 rt_mask(rt), rt->rt_flags, &retrt); 1418 if (error == 0) { 1419 KASSERT(retrt == rt); 1420 KASSERT((retrt->rt_flags & RTF_UP) == 0); 1421 retrt->rt_ifp = NULL; 1422 rtfree(retrt); 1423 } else { 1424 printf("%s: warning: unable to delete rtentry @ %p, " 1425 "error = %d\n", ifp->if_xname, rt, error); 1426 } 1427 return ERESTART; 1428 } 1429 1430 /* 1431 * Create a clone network interface. 1432 */ 1433 static int 1434 if_clone_create(const char *name) 1435 { 1436 struct if_clone *ifc; 1437 int unit; 1438 struct ifnet *ifp; 1439 struct psref psref; 1440 1441 ifc = if_clone_lookup(name, &unit); 1442 if (ifc == NULL) 1443 return EINVAL; 1444 1445 ifp = if_get(name, &psref); 1446 if (ifp != NULL) { 1447 if_put(ifp, &psref); 1448 return EEXIST; 1449 } 1450 1451 return (*ifc->ifc_create)(ifc, unit); 1452 } 1453 1454 /* 1455 * Destroy a clone network interface. 1456 */ 1457 static int 1458 if_clone_destroy(const char *name) 1459 { 1460 struct if_clone *ifc; 1461 struct ifnet *ifp; 1462 struct psref psref; 1463 1464 ifc = if_clone_lookup(name, NULL); 1465 if (ifc == NULL) 1466 return EINVAL; 1467 1468 if (ifc->ifc_destroy == NULL) 1469 return EOPNOTSUPP; 1470 1471 ifp = if_get(name, &psref); 1472 if (ifp == NULL) 1473 return ENXIO; 1474 1475 /* We have to disable ioctls here */ 1476 mutex_enter(ifp->if_ioctl_lock); 1477 ifp->if_ioctl = if_nullioctl; 1478 mutex_exit(ifp->if_ioctl_lock); 1479 1480 /* 1481 * We cannot call ifc_destroy with holding ifp. 1482 * Releasing ifp here is safe thanks to if_clone_mtx. 1483 */ 1484 if_put(ifp, &psref); 1485 1486 return (*ifc->ifc_destroy)(ifp); 1487 } 1488 1489 /* 1490 * Look up a network interface cloner. 1491 */ 1492 static struct if_clone * 1493 if_clone_lookup(const char *name, int *unitp) 1494 { 1495 struct if_clone *ifc; 1496 const char *cp; 1497 char *dp, ifname[IFNAMSIZ + 3]; 1498 int unit; 1499 1500 strcpy(ifname, "if_"); 1501 /* separate interface name from unit */ 1502 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1503 *cp && (*cp < '0' || *cp > '9');) 1504 *dp++ = *cp++; 1505 1506 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1507 return NULL; /* No name or unit number */ 1508 *dp++ = '\0'; 1509 1510 again: 1511 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1512 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1513 break; 1514 } 1515 1516 if (ifc == NULL) { 1517 if (*ifname == '\0' || 1518 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1519 return NULL; 1520 *ifname = '\0'; 1521 goto again; 1522 } 1523 1524 unit = 0; 1525 while (cp - name < IFNAMSIZ && *cp) { 1526 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1527 /* Bogus unit number. */ 1528 return NULL; 1529 } 1530 unit = (unit * 10) + (*cp++ - '0'); 1531 } 1532 1533 if (unitp != NULL) 1534 *unitp = unit; 1535 return ifc; 1536 } 1537 1538 /* 1539 * Register a network interface cloner. 1540 */ 1541 void 1542 if_clone_attach(struct if_clone *ifc) 1543 { 1544 1545 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1546 if_cloners_count++; 1547 } 1548 1549 /* 1550 * Unregister a network interface cloner. 1551 */ 1552 void 1553 if_clone_detach(struct if_clone *ifc) 1554 { 1555 1556 LIST_REMOVE(ifc, ifc_list); 1557 if_cloners_count--; 1558 } 1559 1560 /* 1561 * Provide list of interface cloners to userspace. 1562 */ 1563 int 1564 if_clone_list(int buf_count, char *buffer, int *total) 1565 { 1566 char outbuf[IFNAMSIZ], *dst; 1567 struct if_clone *ifc; 1568 int count, error = 0; 1569 1570 *total = if_cloners_count; 1571 if ((dst = buffer) == NULL) { 1572 /* Just asking how many there are. */ 1573 return 0; 1574 } 1575 1576 if (buf_count < 0) 1577 return EINVAL; 1578 1579 count = (if_cloners_count < buf_count) ? 1580 if_cloners_count : buf_count; 1581 1582 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1583 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1584 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1585 if (outbuf[sizeof(outbuf) - 1] != '\0') 1586 return ENAMETOOLONG; 1587 error = copyout(outbuf, dst, sizeof(outbuf)); 1588 if (error != 0) 1589 break; 1590 } 1591 1592 return error; 1593 } 1594 1595 void 1596 ifa_psref_init(struct ifaddr *ifa) 1597 { 1598 1599 psref_target_init(&ifa->ifa_psref, ifa_psref_class); 1600 } 1601 1602 void 1603 ifaref(struct ifaddr *ifa) 1604 { 1605 ifa->ifa_refcnt++; 1606 } 1607 1608 void 1609 ifafree(struct ifaddr *ifa) 1610 { 1611 KASSERT(ifa != NULL); 1612 KASSERT(ifa->ifa_refcnt > 0); 1613 1614 if (--ifa->ifa_refcnt == 0) { 1615 free(ifa, M_IFADDR); 1616 } 1617 } 1618 1619 void 1620 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1621 { 1622 1623 ifa->ifa_ifp = ifp; 1624 1625 IFNET_LOCK(); 1626 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1627 IFADDR_ENTRY_INIT(ifa); 1628 IFADDR_WRITER_INSERT_TAIL(ifp, ifa); 1629 IFNET_UNLOCK(); 1630 1631 ifaref(ifa); 1632 } 1633 1634 void 1635 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1636 { 1637 1638 KASSERT(ifa->ifa_ifp == ifp); 1639 1640 IFNET_LOCK(); 1641 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1642 IFADDR_WRITER_REMOVE(ifa); 1643 #ifdef NET_MPSAFE 1644 pserialize_perform(ifnet_psz); 1645 #endif 1646 IFNET_UNLOCK(); 1647 1648 #ifdef NET_MPSAFE 1649 psref_target_destroy(&ifa->ifa_psref, ifa_psref_class); 1650 #endif 1651 IFADDR_ENTRY_DESTROY(ifa); 1652 ifafree(ifa); 1653 } 1654 1655 void 1656 ifa_acquire(struct ifaddr *ifa, struct psref *psref) 1657 { 1658 1659 psref_acquire(psref, &ifa->ifa_psref, ifa_psref_class); 1660 } 1661 1662 void 1663 ifa_release(struct ifaddr *ifa, struct psref *psref) 1664 { 1665 1666 if (ifa == NULL) 1667 return; 1668 1669 psref_release(psref, &ifa->ifa_psref, ifa_psref_class); 1670 } 1671 1672 bool 1673 ifa_held(struct ifaddr *ifa) 1674 { 1675 1676 return psref_held(&ifa->ifa_psref, ifa_psref_class); 1677 } 1678 1679 static inline int 1680 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1681 { 1682 return sockaddr_cmp(sa1, sa2) == 0; 1683 } 1684 1685 /* 1686 * Locate an interface based on a complete address. 1687 */ 1688 /*ARGSUSED*/ 1689 struct ifaddr * 1690 ifa_ifwithaddr(const struct sockaddr *addr) 1691 { 1692 struct ifnet *ifp; 1693 struct ifaddr *ifa; 1694 1695 IFNET_READER_FOREACH(ifp) { 1696 if (if_is_deactivated(ifp)) 1697 continue; 1698 IFADDR_READER_FOREACH(ifa, ifp) { 1699 if (ifa->ifa_addr->sa_family != addr->sa_family) 1700 continue; 1701 if (equal(addr, ifa->ifa_addr)) 1702 return ifa; 1703 if ((ifp->if_flags & IFF_BROADCAST) && 1704 ifa->ifa_broadaddr && 1705 /* IP6 doesn't have broadcast */ 1706 ifa->ifa_broadaddr->sa_len != 0 && 1707 equal(ifa->ifa_broadaddr, addr)) 1708 return ifa; 1709 } 1710 } 1711 return NULL; 1712 } 1713 1714 struct ifaddr * 1715 ifa_ifwithaddr_psref(const struct sockaddr *addr, struct psref *psref) 1716 { 1717 struct ifaddr *ifa; 1718 int s = pserialize_read_enter(); 1719 1720 ifa = ifa_ifwithaddr(addr); 1721 if (ifa != NULL) 1722 ifa_acquire(ifa, psref); 1723 pserialize_read_exit(s); 1724 1725 return ifa; 1726 } 1727 1728 /* 1729 * Locate the point to point interface with a given destination address. 1730 */ 1731 /*ARGSUSED*/ 1732 struct ifaddr * 1733 ifa_ifwithdstaddr(const struct sockaddr *addr) 1734 { 1735 struct ifnet *ifp; 1736 struct ifaddr *ifa; 1737 1738 IFNET_READER_FOREACH(ifp) { 1739 if (if_is_deactivated(ifp)) 1740 continue; 1741 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1742 continue; 1743 IFADDR_READER_FOREACH(ifa, ifp) { 1744 if (ifa->ifa_addr->sa_family != addr->sa_family || 1745 ifa->ifa_dstaddr == NULL) 1746 continue; 1747 if (equal(addr, ifa->ifa_dstaddr)) 1748 return ifa; 1749 } 1750 } 1751 1752 return NULL; 1753 } 1754 1755 struct ifaddr * 1756 ifa_ifwithdstaddr_psref(const struct sockaddr *addr, struct psref *psref) 1757 { 1758 struct ifaddr *ifa; 1759 int s; 1760 1761 s = pserialize_read_enter(); 1762 ifa = ifa_ifwithdstaddr(addr); 1763 if (ifa != NULL) 1764 ifa_acquire(ifa, psref); 1765 pserialize_read_exit(s); 1766 1767 return ifa; 1768 } 1769 1770 /* 1771 * Find an interface on a specific network. If many, choice 1772 * is most specific found. 1773 */ 1774 struct ifaddr * 1775 ifa_ifwithnet(const struct sockaddr *addr) 1776 { 1777 struct ifnet *ifp; 1778 struct ifaddr *ifa, *ifa_maybe = NULL; 1779 const struct sockaddr_dl *sdl; 1780 u_int af = addr->sa_family; 1781 const char *addr_data = addr->sa_data, *cplim; 1782 1783 if (af == AF_LINK) { 1784 sdl = satocsdl(addr); 1785 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1786 ifindex2ifnet[sdl->sdl_index] && 1787 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1788 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1789 } 1790 } 1791 #ifdef NETATALK 1792 if (af == AF_APPLETALK) { 1793 const struct sockaddr_at *sat, *sat2; 1794 sat = (const struct sockaddr_at *)addr; 1795 IFNET_READER_FOREACH(ifp) { 1796 if (if_is_deactivated(ifp)) 1797 continue; 1798 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1799 if (ifa == NULL) 1800 continue; 1801 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1802 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1803 return ifa; /* exact match */ 1804 if (ifa_maybe == NULL) { 1805 /* else keep the if with the right range */ 1806 ifa_maybe = ifa; 1807 } 1808 } 1809 return ifa_maybe; 1810 } 1811 #endif 1812 IFNET_READER_FOREACH(ifp) { 1813 if (if_is_deactivated(ifp)) 1814 continue; 1815 IFADDR_READER_FOREACH(ifa, ifp) { 1816 const char *cp, *cp2, *cp3; 1817 1818 if (ifa->ifa_addr->sa_family != af || 1819 ifa->ifa_netmask == NULL) 1820 next: continue; 1821 cp = addr_data; 1822 cp2 = ifa->ifa_addr->sa_data; 1823 cp3 = ifa->ifa_netmask->sa_data; 1824 cplim = (const char *)ifa->ifa_netmask + 1825 ifa->ifa_netmask->sa_len; 1826 while (cp3 < cplim) { 1827 if ((*cp++ ^ *cp2++) & *cp3++) { 1828 /* want to continue for() loop */ 1829 goto next; 1830 } 1831 } 1832 if (ifa_maybe == NULL || 1833 rt_refines(ifa->ifa_netmask, 1834 ifa_maybe->ifa_netmask)) 1835 ifa_maybe = ifa; 1836 } 1837 } 1838 return ifa_maybe; 1839 } 1840 1841 struct ifaddr * 1842 ifa_ifwithnet_psref(const struct sockaddr *addr, struct psref *psref) 1843 { 1844 struct ifaddr *ifa; 1845 int s; 1846 1847 s = pserialize_read_enter(); 1848 ifa = ifa_ifwithnet(addr); 1849 if (ifa != NULL) 1850 ifa_acquire(ifa, psref); 1851 pserialize_read_exit(s); 1852 1853 return ifa; 1854 } 1855 1856 /* 1857 * Find the interface of the addresss. 1858 */ 1859 struct ifaddr * 1860 ifa_ifwithladdr(const struct sockaddr *addr) 1861 { 1862 struct ifaddr *ia; 1863 1864 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1865 (ia = ifa_ifwithnet(addr))) 1866 return ia; 1867 return NULL; 1868 } 1869 1870 struct ifaddr * 1871 ifa_ifwithladdr_psref(const struct sockaddr *addr, struct psref *psref) 1872 { 1873 struct ifaddr *ifa; 1874 int s; 1875 1876 s = pserialize_read_enter(); 1877 ifa = ifa_ifwithladdr(addr); 1878 if (ifa != NULL) 1879 ifa_acquire(ifa, psref); 1880 pserialize_read_exit(s); 1881 1882 return ifa; 1883 } 1884 1885 /* 1886 * Find an interface using a specific address family 1887 */ 1888 struct ifaddr * 1889 ifa_ifwithaf(int af) 1890 { 1891 struct ifnet *ifp; 1892 struct ifaddr *ifa = NULL; 1893 int s; 1894 1895 s = pserialize_read_enter(); 1896 IFNET_READER_FOREACH(ifp) { 1897 if (if_is_deactivated(ifp)) 1898 continue; 1899 IFADDR_READER_FOREACH(ifa, ifp) { 1900 if (ifa->ifa_addr->sa_family == af) 1901 goto out; 1902 } 1903 } 1904 out: 1905 pserialize_read_exit(s); 1906 return ifa; 1907 } 1908 1909 /* 1910 * Find an interface address specific to an interface best matching 1911 * a given address. 1912 */ 1913 struct ifaddr * 1914 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1915 { 1916 struct ifaddr *ifa; 1917 const char *cp, *cp2, *cp3; 1918 const char *cplim; 1919 struct ifaddr *ifa_maybe = 0; 1920 u_int af = addr->sa_family; 1921 1922 if (if_is_deactivated(ifp)) 1923 return NULL; 1924 1925 if (af >= AF_MAX) 1926 return NULL; 1927 1928 IFADDR_READER_FOREACH(ifa, ifp) { 1929 if (ifa->ifa_addr->sa_family != af) 1930 continue; 1931 ifa_maybe = ifa; 1932 if (ifa->ifa_netmask == NULL) { 1933 if (equal(addr, ifa->ifa_addr) || 1934 (ifa->ifa_dstaddr && 1935 equal(addr, ifa->ifa_dstaddr))) 1936 return ifa; 1937 continue; 1938 } 1939 cp = addr->sa_data; 1940 cp2 = ifa->ifa_addr->sa_data; 1941 cp3 = ifa->ifa_netmask->sa_data; 1942 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1943 for (; cp3 < cplim; cp3++) { 1944 if ((*cp++ ^ *cp2++) & *cp3) 1945 break; 1946 } 1947 if (cp3 == cplim) 1948 return ifa; 1949 } 1950 return ifa_maybe; 1951 } 1952 1953 struct ifaddr * 1954 ifaof_ifpforaddr_psref(const struct sockaddr *addr, struct ifnet *ifp, 1955 struct psref *psref) 1956 { 1957 struct ifaddr *ifa; 1958 int s; 1959 1960 s = pserialize_read_enter(); 1961 ifa = ifaof_ifpforaddr(addr, ifp); 1962 if (ifa != NULL) 1963 ifa_acquire(ifa, psref); 1964 pserialize_read_exit(s); 1965 1966 return ifa; 1967 } 1968 1969 /* 1970 * Default action when installing a route with a Link Level gateway. 1971 * Lookup an appropriate real ifa to point to. 1972 * This should be moved to /sys/net/link.c eventually. 1973 */ 1974 void 1975 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1976 { 1977 struct ifaddr *ifa; 1978 const struct sockaddr *dst; 1979 struct ifnet *ifp; 1980 struct psref psref; 1981 1982 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1983 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1984 return; 1985 if ((ifa = ifaof_ifpforaddr_psref(dst, ifp, &psref)) != NULL) { 1986 rt_replace_ifa(rt, ifa); 1987 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1988 ifa->ifa_rtrequest(cmd, rt, info); 1989 ifa_release(ifa, &psref); 1990 } 1991 } 1992 1993 /* 1994 * bitmask macros to manage a densely packed link_state change queue. 1995 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 1996 * LINK_STATE_UP(2) we need 2 bits for each state change. 1997 * As a state change to store is 0, treat all bits set as an unset item. 1998 */ 1999 #define LQ_ITEM_BITS 2 2000 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 2001 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 2002 #define LINK_STATE_UNSET LQ_ITEM_MASK 2003 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 2004 #define LQ_STORE(q, i, v) \ 2005 do { \ 2006 (q) &= ~LQ_MASK((i)); \ 2007 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 2008 } while (0 /* CONSTCOND */) 2009 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 2010 #define LQ_POP(q, v) \ 2011 do { \ 2012 (v) = LQ_ITEM((q), 0); \ 2013 (q) >>= LQ_ITEM_BITS; \ 2014 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2015 } while (0 /* CONSTCOND */) 2016 #define LQ_PUSH(q, v) \ 2017 do { \ 2018 (q) >>= LQ_ITEM_BITS; \ 2019 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2020 } while (0 /* CONSTCOND */) 2021 #define LQ_FIND_UNSET(q, i) \ 2022 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 2023 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 2024 break; \ 2025 } 2026 /* 2027 * Handle a change in the interface link state and 2028 * queue notifications. 2029 */ 2030 void 2031 if_link_state_change(struct ifnet *ifp, int link_state) 2032 { 2033 int s, idx; 2034 2035 KASSERTMSG(if_is_link_state_changeable(ifp), 2036 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 2037 ifp->if_xname, ifp->if_extflags); 2038 2039 /* Ensure change is to a valid state */ 2040 switch (link_state) { 2041 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 2042 case LINK_STATE_DOWN: /* FALLTHROUGH */ 2043 case LINK_STATE_UP: 2044 break; 2045 default: 2046 #ifdef DEBUG 2047 printf("%s: invalid link state %d\n", 2048 ifp->if_xname, link_state); 2049 #endif 2050 return; 2051 } 2052 2053 s = splnet(); 2054 2055 /* Find the last unset event in the queue. */ 2056 LQ_FIND_UNSET(ifp->if_link_queue, idx); 2057 2058 /* 2059 * Ensure link_state doesn't match the last event in the queue. 2060 * ifp->if_link_state is not checked and set here because 2061 * that would present an inconsistent picture to the system. 2062 */ 2063 if (idx != 0 && 2064 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 2065 goto out; 2066 2067 /* Handle queue overflow. */ 2068 if (idx == LQ_MAX(ifp->if_link_queue)) { 2069 uint8_t lost; 2070 2071 /* 2072 * The DOWN state must be protected from being pushed off 2073 * the queue to ensure that userland will always be 2074 * in a sane state. 2075 * Because DOWN is protected, there is no need to protect 2076 * UNKNOWN. 2077 * It should be invalid to change from any other state to 2078 * UNKNOWN anyway ... 2079 */ 2080 lost = LQ_ITEM(ifp->if_link_queue, 0); 2081 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 2082 if (lost == LINK_STATE_DOWN) { 2083 lost = LQ_ITEM(ifp->if_link_queue, 0); 2084 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 2085 } 2086 printf("%s: lost link state change %s\n", 2087 ifp->if_xname, 2088 lost == LINK_STATE_UP ? "UP" : 2089 lost == LINK_STATE_DOWN ? "DOWN" : 2090 "UNKNOWN"); 2091 } else 2092 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 2093 2094 softint_schedule(ifp->if_link_si); 2095 2096 out: 2097 splx(s); 2098 } 2099 2100 /* 2101 * Handle interface link state change notifications. 2102 * Must be called at splnet(). 2103 */ 2104 static void 2105 if_link_state_change0(struct ifnet *ifp, int link_state) 2106 { 2107 struct domain *dp; 2108 2109 /* Ensure the change is still valid. */ 2110 if (ifp->if_link_state == link_state) 2111 return; 2112 2113 #ifdef DEBUG 2114 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 2115 link_state == LINK_STATE_UP ? "UP" : 2116 link_state == LINK_STATE_DOWN ? "DOWN" : 2117 "UNKNOWN", 2118 ifp->if_link_state == LINK_STATE_UP ? "UP" : 2119 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 2120 "UNKNOWN"); 2121 #endif 2122 2123 /* 2124 * When going from UNKNOWN to UP, we need to mark existing 2125 * addresses as tentative and restart DAD as we may have 2126 * erroneously not found a duplicate. 2127 * 2128 * This needs to happen before rt_ifmsg to avoid a race where 2129 * listeners would have an address and expect it to work right 2130 * away. 2131 */ 2132 if (link_state == LINK_STATE_UP && 2133 ifp->if_link_state == LINK_STATE_UNKNOWN) 2134 { 2135 DOMAIN_FOREACH(dp) { 2136 if (dp->dom_if_link_state_change != NULL) 2137 dp->dom_if_link_state_change(ifp, 2138 LINK_STATE_DOWN); 2139 } 2140 } 2141 2142 ifp->if_link_state = link_state; 2143 2144 /* Notify that the link state has changed. */ 2145 rt_ifmsg(ifp); 2146 2147 #if NCARP > 0 2148 if (ifp->if_carp) 2149 carp_carpdev_state(ifp); 2150 #endif 2151 2152 DOMAIN_FOREACH(dp) { 2153 if (dp->dom_if_link_state_change != NULL) 2154 dp->dom_if_link_state_change(ifp, link_state); 2155 } 2156 } 2157 2158 /* 2159 * Process the interface link state change queue. 2160 */ 2161 static void 2162 if_link_state_change_si(void *arg) 2163 { 2164 struct ifnet *ifp = arg; 2165 int s; 2166 uint8_t state; 2167 2168 s = splnet(); 2169 2170 /* Pop a link state change from the queue and process it. */ 2171 LQ_POP(ifp->if_link_queue, state); 2172 if_link_state_change0(ifp, state); 2173 2174 /* If there is a link state change to come, schedule it. */ 2175 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 2176 softint_schedule(ifp->if_link_si); 2177 2178 splx(s); 2179 } 2180 2181 /* 2182 * Default action when installing a local route on a point-to-point 2183 * interface. 2184 */ 2185 void 2186 p2p_rtrequest(int req, struct rtentry *rt, 2187 __unused const struct rt_addrinfo *info) 2188 { 2189 struct ifnet *ifp = rt->rt_ifp; 2190 struct ifaddr *ifa, *lo0ifa; 2191 int s = pserialize_read_enter(); 2192 2193 switch (req) { 2194 case RTM_ADD: 2195 if ((rt->rt_flags & RTF_LOCAL) == 0) 2196 break; 2197 2198 rt->rt_ifp = lo0ifp; 2199 2200 IFADDR_READER_FOREACH(ifa, ifp) { 2201 if (equal(rt_getkey(rt), ifa->ifa_addr)) 2202 break; 2203 } 2204 if (ifa == NULL) 2205 break; 2206 2207 /* 2208 * Ensure lo0 has an address of the same family. 2209 */ 2210 IFADDR_READER_FOREACH(lo0ifa, lo0ifp) { 2211 if (lo0ifa->ifa_addr->sa_family == 2212 ifa->ifa_addr->sa_family) 2213 break; 2214 } 2215 if (lo0ifa == NULL) 2216 break; 2217 2218 /* 2219 * Make sure to set rt->rt_ifa to the interface 2220 * address we are using, otherwise we will have trouble 2221 * with source address selection. 2222 */ 2223 if (ifa != rt->rt_ifa) 2224 rt_replace_ifa(rt, ifa); 2225 break; 2226 case RTM_DELETE: 2227 default: 2228 break; 2229 } 2230 pserialize_read_exit(s); 2231 } 2232 2233 /* 2234 * Mark an interface down and notify protocols of 2235 * the transition. 2236 * NOTE: must be called at splsoftnet or equivalent. 2237 */ 2238 void 2239 if_down(struct ifnet *ifp) 2240 { 2241 struct ifaddr *ifa; 2242 struct domain *dp; 2243 int s, bound; 2244 struct psref psref; 2245 2246 ifp->if_flags &= ~IFF_UP; 2247 nanotime(&ifp->if_lastchange); 2248 2249 bound = curlwp_bind(); 2250 s = pserialize_read_enter(); 2251 IFADDR_READER_FOREACH(ifa, ifp) { 2252 ifa_acquire(ifa, &psref); 2253 pserialize_read_exit(s); 2254 2255 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2256 2257 s = pserialize_read_enter(); 2258 ifa_release(ifa, &psref); 2259 } 2260 pserialize_read_exit(s); 2261 curlwp_bindx(bound); 2262 2263 IFQ_PURGE(&ifp->if_snd); 2264 #if NCARP > 0 2265 if (ifp->if_carp) 2266 carp_carpdev_state(ifp); 2267 #endif 2268 rt_ifmsg(ifp); 2269 DOMAIN_FOREACH(dp) { 2270 if (dp->dom_if_down) 2271 dp->dom_if_down(ifp); 2272 } 2273 } 2274 2275 /* 2276 * Mark an interface up and notify protocols of 2277 * the transition. 2278 * NOTE: must be called at splsoftnet or equivalent. 2279 */ 2280 void 2281 if_up(struct ifnet *ifp) 2282 { 2283 #ifdef notyet 2284 struct ifaddr *ifa; 2285 #endif 2286 struct domain *dp; 2287 2288 ifp->if_flags |= IFF_UP; 2289 nanotime(&ifp->if_lastchange); 2290 #ifdef notyet 2291 /* this has no effect on IP, and will kill all ISO connections XXX */ 2292 IFADDR_READER_FOREACH(ifa, ifp) 2293 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2294 #endif 2295 #if NCARP > 0 2296 if (ifp->if_carp) 2297 carp_carpdev_state(ifp); 2298 #endif 2299 rt_ifmsg(ifp); 2300 DOMAIN_FOREACH(dp) { 2301 if (dp->dom_if_up) 2302 dp->dom_if_up(ifp); 2303 } 2304 } 2305 2306 /* 2307 * Handle interface slowtimo timer routine. Called 2308 * from softclock, we decrement timer (if set) and 2309 * call the appropriate interface routine on expiration. 2310 */ 2311 static void 2312 if_slowtimo(void *arg) 2313 { 2314 void (*slowtimo)(struct ifnet *); 2315 struct ifnet *ifp = arg; 2316 int s; 2317 2318 slowtimo = ifp->if_slowtimo; 2319 if (__predict_false(slowtimo == NULL)) 2320 return; 2321 2322 s = splnet(); 2323 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2324 (*slowtimo)(ifp); 2325 2326 splx(s); 2327 2328 if (__predict_true(ifp->if_slowtimo != NULL)) 2329 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2330 } 2331 2332 /* 2333 * Set/clear promiscuous mode on interface ifp based on the truth value 2334 * of pswitch. The calls are reference counted so that only the first 2335 * "on" request actually has an effect, as does the final "off" request. 2336 * Results are undefined if the "off" and "on" requests are not matched. 2337 */ 2338 int 2339 ifpromisc(struct ifnet *ifp, int pswitch) 2340 { 2341 int pcount, ret; 2342 short nflags; 2343 2344 pcount = ifp->if_pcount; 2345 if (pswitch) { 2346 /* 2347 * Allow the device to be "placed" into promiscuous 2348 * mode even if it is not configured up. It will 2349 * consult IFF_PROMISC when it is brought up. 2350 */ 2351 if (ifp->if_pcount++ != 0) 2352 return 0; 2353 nflags = ifp->if_flags | IFF_PROMISC; 2354 } else { 2355 if (--ifp->if_pcount > 0) 2356 return 0; 2357 nflags = ifp->if_flags & ~IFF_PROMISC; 2358 } 2359 ret = if_flags_set(ifp, nflags); 2360 /* Restore interface state if not successful. */ 2361 if (ret != 0) { 2362 ifp->if_pcount = pcount; 2363 } 2364 return ret; 2365 } 2366 2367 /* 2368 * Map interface name to 2369 * interface structure pointer. 2370 */ 2371 struct ifnet * 2372 ifunit(const char *name) 2373 { 2374 struct ifnet *ifp; 2375 const char *cp = name; 2376 u_int unit = 0; 2377 u_int i; 2378 int s; 2379 2380 /* 2381 * If the entire name is a number, treat it as an ifindex. 2382 */ 2383 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2384 unit = unit * 10 + (*cp - '0'); 2385 } 2386 2387 /* 2388 * If the number took all of the name, then it's a valid ifindex. 2389 */ 2390 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2391 if (unit >= if_indexlim) 2392 return NULL; 2393 ifp = ifindex2ifnet[unit]; 2394 if (ifp == NULL || if_is_deactivated(ifp)) 2395 return NULL; 2396 return ifp; 2397 } 2398 2399 ifp = NULL; 2400 s = pserialize_read_enter(); 2401 IFNET_READER_FOREACH(ifp) { 2402 if (if_is_deactivated(ifp)) 2403 continue; 2404 if (strcmp(ifp->if_xname, name) == 0) 2405 goto out; 2406 } 2407 out: 2408 pserialize_read_exit(s); 2409 return ifp; 2410 } 2411 2412 /* 2413 * Get a reference of an ifnet object by an interface name. 2414 * The returned reference is protected by psref(9). The caller 2415 * must release a returned reference by if_put after use. 2416 */ 2417 struct ifnet * 2418 if_get(const char *name, struct psref *psref) 2419 { 2420 struct ifnet *ifp; 2421 const char *cp = name; 2422 u_int unit = 0; 2423 u_int i; 2424 int s; 2425 2426 /* 2427 * If the entire name is a number, treat it as an ifindex. 2428 */ 2429 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2430 unit = unit * 10 + (*cp - '0'); 2431 } 2432 2433 /* 2434 * If the number took all of the name, then it's a valid ifindex. 2435 */ 2436 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2437 if (unit >= if_indexlim) 2438 return NULL; 2439 ifp = ifindex2ifnet[unit]; 2440 if (ifp == NULL || if_is_deactivated(ifp)) 2441 return NULL; 2442 return ifp; 2443 } 2444 2445 ifp = NULL; 2446 s = pserialize_read_enter(); 2447 IFNET_READER_FOREACH(ifp) { 2448 if (if_is_deactivated(ifp)) 2449 continue; 2450 if (strcmp(ifp->if_xname, name) == 0) { 2451 psref_acquire(psref, &ifp->if_psref, 2452 ifnet_psref_class); 2453 goto out; 2454 } 2455 } 2456 out: 2457 pserialize_read_exit(s); 2458 return ifp; 2459 } 2460 2461 /* 2462 * Release a reference of an ifnet object given by if_get or 2463 * if_get_byindex. 2464 */ 2465 void 2466 if_put(const struct ifnet *ifp, struct psref *psref) 2467 { 2468 2469 if (ifp == NULL) 2470 return; 2471 2472 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2473 } 2474 2475 ifnet_t * 2476 if_byindex(u_int idx) 2477 { 2478 ifnet_t *ifp; 2479 2480 ifp = (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2481 if (ifp != NULL && if_is_deactivated(ifp)) 2482 ifp = NULL; 2483 return ifp; 2484 } 2485 2486 /* 2487 * Get a reference of an ifnet object by an interface index. 2488 * The returned reference is protected by psref(9). The caller 2489 * must release a returned reference by if_put after use. 2490 */ 2491 ifnet_t * 2492 if_get_byindex(u_int idx, struct psref *psref) 2493 { 2494 ifnet_t *ifp; 2495 int s; 2496 2497 s = pserialize_read_enter(); 2498 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2499 if (ifp != NULL && if_is_deactivated(ifp)) 2500 ifp = NULL; 2501 if (__predict_true(ifp != NULL)) 2502 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2503 pserialize_read_exit(s); 2504 2505 return ifp; 2506 } 2507 2508 /* 2509 * XXX it's safe only if the passed ifp is guaranteed to not be freed, 2510 * for example the ifp is already held or some other object is held which 2511 * guarantes the ifp to not be freed indirectly. 2512 */ 2513 void 2514 if_acquire_NOMPSAFE(struct ifnet *ifp, struct psref *psref) 2515 { 2516 2517 KASSERT(ifp->if_index != 0); 2518 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2519 } 2520 2521 bool 2522 if_held(struct ifnet *ifp) 2523 { 2524 2525 return psref_held(&ifp->if_psref, ifnet_psref_class); 2526 } 2527 2528 2529 /* common */ 2530 int 2531 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2532 { 2533 int s; 2534 struct ifreq *ifr; 2535 struct ifcapreq *ifcr; 2536 struct ifdatareq *ifdr; 2537 2538 switch (cmd) { 2539 case SIOCSIFCAP: 2540 ifcr = data; 2541 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2542 return EINVAL; 2543 2544 if (ifcr->ifcr_capenable == ifp->if_capenable) 2545 return 0; 2546 2547 ifp->if_capenable = ifcr->ifcr_capenable; 2548 2549 /* Pre-compute the checksum flags mask. */ 2550 ifp->if_csum_flags_tx = 0; 2551 ifp->if_csum_flags_rx = 0; 2552 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2553 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2554 } 2555 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2556 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2557 } 2558 2559 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2560 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2561 } 2562 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2563 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2564 } 2565 2566 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2567 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2568 } 2569 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2570 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2571 } 2572 2573 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2574 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2575 } 2576 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2577 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2578 } 2579 2580 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2581 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2582 } 2583 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2584 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2585 } 2586 if (ifp->if_flags & IFF_UP) 2587 return ENETRESET; 2588 return 0; 2589 case SIOCSIFFLAGS: 2590 ifr = data; 2591 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2592 s = splnet(); 2593 if_down(ifp); 2594 splx(s); 2595 } 2596 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2597 s = splnet(); 2598 if_up(ifp); 2599 splx(s); 2600 } 2601 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2602 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2603 break; 2604 case SIOCGIFFLAGS: 2605 ifr = data; 2606 ifr->ifr_flags = ifp->if_flags; 2607 break; 2608 2609 case SIOCGIFMETRIC: 2610 ifr = data; 2611 ifr->ifr_metric = ifp->if_metric; 2612 break; 2613 2614 case SIOCGIFMTU: 2615 ifr = data; 2616 ifr->ifr_mtu = ifp->if_mtu; 2617 break; 2618 2619 case SIOCGIFDLT: 2620 ifr = data; 2621 ifr->ifr_dlt = ifp->if_dlt; 2622 break; 2623 2624 case SIOCGIFCAP: 2625 ifcr = data; 2626 ifcr->ifcr_capabilities = ifp->if_capabilities; 2627 ifcr->ifcr_capenable = ifp->if_capenable; 2628 break; 2629 2630 case SIOCSIFMETRIC: 2631 ifr = data; 2632 ifp->if_metric = ifr->ifr_metric; 2633 break; 2634 2635 case SIOCGIFDATA: 2636 ifdr = data; 2637 ifdr->ifdr_data = ifp->if_data; 2638 break; 2639 2640 case SIOCGIFINDEX: 2641 ifr = data; 2642 ifr->ifr_index = ifp->if_index; 2643 break; 2644 2645 case SIOCZIFDATA: 2646 ifdr = data; 2647 ifdr->ifdr_data = ifp->if_data; 2648 /* 2649 * Assumes that the volatile counters that can be 2650 * zero'ed are at the end of if_data. 2651 */ 2652 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2653 offsetof(struct if_data, ifi_ipackets)); 2654 /* 2655 * The memset() clears to the bottm of if_data. In the area, 2656 * if_lastchange is included. Please be careful if new entry 2657 * will be added into if_data or rewite this. 2658 * 2659 * And also, update if_lastchnage. 2660 */ 2661 getnanotime(&ifp->if_lastchange); 2662 break; 2663 case SIOCSIFMTU: 2664 ifr = data; 2665 if (ifp->if_mtu == ifr->ifr_mtu) 2666 break; 2667 ifp->if_mtu = ifr->ifr_mtu; 2668 /* 2669 * If the link MTU changed, do network layer specific procedure. 2670 */ 2671 #ifdef INET6 2672 if (in6_present) 2673 nd6_setmtu(ifp); 2674 #endif 2675 return ENETRESET; 2676 default: 2677 return ENOTTY; 2678 } 2679 return 0; 2680 } 2681 2682 int 2683 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2684 { 2685 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2686 struct ifaddr *ifa; 2687 const struct sockaddr *any, *sa; 2688 union { 2689 struct sockaddr sa; 2690 struct sockaddr_storage ss; 2691 } u, v; 2692 int s, error = 0; 2693 2694 switch (cmd) { 2695 case SIOCSIFADDRPREF: 2696 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2697 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2698 NULL) != 0) 2699 return EPERM; 2700 case SIOCGIFADDRPREF: 2701 break; 2702 default: 2703 return EOPNOTSUPP; 2704 } 2705 2706 /* sanity checks */ 2707 if (data == NULL || ifp == NULL) { 2708 panic("invalid argument to %s", __func__); 2709 /*NOTREACHED*/ 2710 } 2711 2712 /* address must be specified on ADD and DELETE */ 2713 sa = sstocsa(&ifap->ifap_addr); 2714 if (sa->sa_family != sofamily(so)) 2715 return EINVAL; 2716 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2717 return EINVAL; 2718 2719 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2720 2721 s = pserialize_read_enter(); 2722 IFADDR_READER_FOREACH(ifa, ifp) { 2723 if (ifa->ifa_addr->sa_family != sa->sa_family) 2724 continue; 2725 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2726 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2727 break; 2728 } 2729 if (ifa == NULL) { 2730 error = EADDRNOTAVAIL; 2731 goto out; 2732 } 2733 2734 switch (cmd) { 2735 case SIOCSIFADDRPREF: 2736 ifa->ifa_preference = ifap->ifap_preference; 2737 goto out; 2738 case SIOCGIFADDRPREF: 2739 /* fill in the if_laddrreq structure */ 2740 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2741 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2742 ifap->ifap_preference = ifa->ifa_preference; 2743 goto out; 2744 default: 2745 error = EOPNOTSUPP; 2746 } 2747 out: 2748 pserialize_read_exit(s); 2749 return error; 2750 } 2751 2752 /* 2753 * Interface ioctls. 2754 */ 2755 static int 2756 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2757 { 2758 struct ifnet *ifp; 2759 struct ifreq *ifr; 2760 int error = 0; 2761 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2762 u_long ocmd = cmd; 2763 #endif 2764 short oif_flags; 2765 #ifdef COMPAT_OIFREQ 2766 struct ifreq ifrb; 2767 struct oifreq *oifr = NULL; 2768 #endif 2769 int r; 2770 struct psref psref; 2771 int bound; 2772 2773 switch (cmd) { 2774 #ifdef COMPAT_OIFREQ 2775 case OSIOCGIFCONF: 2776 case OOSIOCGIFCONF: 2777 return compat_ifconf(cmd, data); 2778 #endif 2779 #ifdef COMPAT_OIFDATA 2780 case OSIOCGIFDATA: 2781 case OSIOCZIFDATA: 2782 return compat_ifdatareq(l, cmd, data); 2783 #endif 2784 case SIOCGIFCONF: 2785 return ifconf(cmd, data); 2786 case SIOCINITIFADDR: 2787 return EPERM; 2788 } 2789 2790 #ifdef COMPAT_OIFREQ 2791 cmd = (*vec_compat_cvtcmd)(cmd); 2792 if (cmd != ocmd) { 2793 oifr = data; 2794 data = ifr = &ifrb; 2795 ifreqo2n(oifr, ifr); 2796 } else 2797 #endif 2798 ifr = data; 2799 2800 switch (cmd) { 2801 case SIOCIFCREATE: 2802 case SIOCIFDESTROY: 2803 bound = curlwp_bind(); 2804 if (l != NULL) { 2805 ifp = if_get(ifr->ifr_name, &psref); 2806 error = kauth_authorize_network(l->l_cred, 2807 KAUTH_NETWORK_INTERFACE, 2808 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2809 (void *)cmd, NULL); 2810 if (ifp != NULL) 2811 if_put(ifp, &psref); 2812 if (error != 0) { 2813 curlwp_bindx(bound); 2814 return error; 2815 } 2816 } 2817 mutex_enter(&if_clone_mtx); 2818 r = (cmd == SIOCIFCREATE) ? 2819 if_clone_create(ifr->ifr_name) : 2820 if_clone_destroy(ifr->ifr_name); 2821 mutex_exit(&if_clone_mtx); 2822 curlwp_bindx(bound); 2823 return r; 2824 2825 case SIOCIFGCLONERS: 2826 { 2827 struct if_clonereq *req = (struct if_clonereq *)data; 2828 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2829 &req->ifcr_total); 2830 } 2831 } 2832 2833 bound = curlwp_bind(); 2834 ifp = if_get(ifr->ifr_name, &psref); 2835 if (ifp == NULL) { 2836 curlwp_bindx(bound); 2837 return ENXIO; 2838 } 2839 2840 switch (cmd) { 2841 case SIOCALIFADDR: 2842 case SIOCDLIFADDR: 2843 case SIOCSIFADDRPREF: 2844 case SIOCSIFFLAGS: 2845 case SIOCSIFCAP: 2846 case SIOCSIFMETRIC: 2847 case SIOCZIFDATA: 2848 case SIOCSIFMTU: 2849 case SIOCSIFPHYADDR: 2850 case SIOCDIFPHYADDR: 2851 #ifdef INET6 2852 case SIOCSIFPHYADDR_IN6: 2853 #endif 2854 case SIOCSLIFPHYADDR: 2855 case SIOCADDMULTI: 2856 case SIOCDELMULTI: 2857 case SIOCSIFMEDIA: 2858 case SIOCSDRVSPEC: 2859 case SIOCG80211: 2860 case SIOCS80211: 2861 case SIOCS80211NWID: 2862 case SIOCS80211NWKEY: 2863 case SIOCS80211POWER: 2864 case SIOCS80211BSSID: 2865 case SIOCS80211CHANNEL: 2866 case SIOCSLINKSTR: 2867 if (l != NULL) { 2868 error = kauth_authorize_network(l->l_cred, 2869 KAUTH_NETWORK_INTERFACE, 2870 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2871 (void *)cmd, NULL); 2872 if (error != 0) 2873 goto out; 2874 } 2875 } 2876 2877 oif_flags = ifp->if_flags; 2878 2879 mutex_enter(ifp->if_ioctl_lock); 2880 2881 error = (*ifp->if_ioctl)(ifp, cmd, data); 2882 if (error != ENOTTY) 2883 ; 2884 else if (so->so_proto == NULL) 2885 error = EOPNOTSUPP; 2886 else { 2887 #ifdef COMPAT_OSOCK 2888 if (vec_compat_ifioctl != NULL) 2889 error = (*vec_compat_ifioctl)(so, ocmd, cmd, data, l); 2890 else 2891 #endif 2892 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2893 cmd, data, ifp); 2894 } 2895 2896 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2897 if ((ifp->if_flags & IFF_UP) != 0) { 2898 int s = splnet(); 2899 if_up(ifp); 2900 splx(s); 2901 } 2902 } 2903 #ifdef COMPAT_OIFREQ 2904 if (cmd != ocmd) 2905 ifreqn2o(oifr, ifr); 2906 #endif 2907 2908 mutex_exit(ifp->if_ioctl_lock); 2909 out: 2910 if_put(ifp, &psref); 2911 curlwp_bindx(bound); 2912 return error; 2913 } 2914 2915 /* 2916 * Return interface configuration 2917 * of system. List may be used 2918 * in later ioctl's (above) to get 2919 * other information. 2920 * 2921 * Each record is a struct ifreq. Before the addition of 2922 * sockaddr_storage, the API rule was that sockaddr flavors that did 2923 * not fit would extend beyond the struct ifreq, with the next struct 2924 * ifreq starting sa_len beyond the struct sockaddr. Because the 2925 * union in struct ifreq includes struct sockaddr_storage, every kind 2926 * of sockaddr must fit. Thus, there are no longer any overlength 2927 * records. 2928 * 2929 * Records are added to the user buffer if they fit, and ifc_len is 2930 * adjusted to the length that was written. Thus, the user is only 2931 * assured of getting the complete list if ifc_len on return is at 2932 * least sizeof(struct ifreq) less than it was on entry. 2933 * 2934 * If the user buffer pointer is NULL, this routine copies no data and 2935 * returns the amount of space that would be needed. 2936 * 2937 * Invariants: 2938 * ifrp points to the next part of the user's buffer to be used. If 2939 * ifrp != NULL, space holds the number of bytes remaining that we may 2940 * write at ifrp. Otherwise, space holds the number of bytes that 2941 * would have been written had there been adequate space. 2942 */ 2943 /*ARGSUSED*/ 2944 static int 2945 ifconf(u_long cmd, void *data) 2946 { 2947 struct ifconf *ifc = (struct ifconf *)data; 2948 struct ifnet *ifp; 2949 struct ifaddr *ifa; 2950 struct ifreq ifr, *ifrp = NULL; 2951 int space = 0, error = 0; 2952 const int sz = (int)sizeof(struct ifreq); 2953 const bool docopy = ifc->ifc_req != NULL; 2954 int s; 2955 int bound; 2956 struct psref psref; 2957 2958 if (docopy) { 2959 space = ifc->ifc_len; 2960 ifrp = ifc->ifc_req; 2961 } 2962 2963 bound = curlwp_bind(); 2964 s = pserialize_read_enter(); 2965 IFNET_READER_FOREACH(ifp) { 2966 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 2967 pserialize_read_exit(s); 2968 2969 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2970 sizeof(ifr.ifr_name)); 2971 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 2972 error = ENAMETOOLONG; 2973 goto release_exit; 2974 } 2975 if (IFADDR_READER_EMPTY(ifp)) { 2976 /* Interface with no addresses - send zero sockaddr. */ 2977 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2978 if (!docopy) { 2979 space += sz; 2980 continue; 2981 } 2982 if (space >= sz) { 2983 error = copyout(&ifr, ifrp, sz); 2984 if (error != 0) 2985 goto release_exit; 2986 ifrp++; 2987 space -= sz; 2988 } 2989 } 2990 2991 IFADDR_READER_FOREACH(ifa, ifp) { 2992 struct sockaddr *sa = ifa->ifa_addr; 2993 /* all sockaddrs must fit in sockaddr_storage */ 2994 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2995 2996 if (!docopy) { 2997 space += sz; 2998 continue; 2999 } 3000 memcpy(&ifr.ifr_space, sa, sa->sa_len); 3001 if (space >= sz) { 3002 error = copyout(&ifr, ifrp, sz); 3003 if (error != 0) 3004 goto release_exit; 3005 ifrp++; space -= sz; 3006 } 3007 } 3008 3009 s = pserialize_read_enter(); 3010 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3011 } 3012 pserialize_read_exit(s); 3013 curlwp_bindx(bound); 3014 3015 if (docopy) { 3016 KASSERT(0 <= space && space <= ifc->ifc_len); 3017 ifc->ifc_len -= space; 3018 } else { 3019 KASSERT(space >= 0); 3020 ifc->ifc_len = space; 3021 } 3022 return (0); 3023 3024 release_exit: 3025 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3026 curlwp_bindx(bound); 3027 return error; 3028 } 3029 3030 int 3031 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 3032 { 3033 uint8_t len; 3034 #ifdef COMPAT_OIFREQ 3035 struct ifreq ifrb; 3036 struct oifreq *oifr = NULL; 3037 u_long ocmd = cmd; 3038 cmd = (*vec_compat_cvtcmd)(cmd); 3039 if (cmd != ocmd) { 3040 oifr = (struct oifreq *)(void *)ifr; 3041 ifr = &ifrb; 3042 ifreqo2n(oifr, ifr); 3043 len = sizeof(oifr->ifr_addr); 3044 } else 3045 #endif 3046 len = sizeof(ifr->ifr_ifru.ifru_space); 3047 3048 if (len < sa->sa_len) 3049 return EFBIG; 3050 3051 memset(&ifr->ifr_addr, 0, len); 3052 sockaddr_copy(&ifr->ifr_addr, len, sa); 3053 3054 #ifdef COMPAT_OIFREQ 3055 if (cmd != ocmd) 3056 ifreqn2o(oifr, ifr); 3057 #endif 3058 return 0; 3059 } 3060 3061 /* 3062 * wrapper function for the drivers which doesn't have if_transmit(). 3063 */ 3064 static int 3065 if_transmit(struct ifnet *ifp, struct mbuf *m) 3066 { 3067 int s, error; 3068 3069 s = splnet(); 3070 3071 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3072 if (error != 0) { 3073 /* mbuf is already freed */ 3074 goto out; 3075 } 3076 3077 ifp->if_obytes += m->m_pkthdr.len;; 3078 if (m->m_flags & M_MCAST) 3079 ifp->if_omcasts++; 3080 3081 if ((ifp->if_flags & IFF_OACTIVE) == 0) 3082 if_start_lock(ifp); 3083 out: 3084 splx(s); 3085 3086 return error; 3087 } 3088 3089 int 3090 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 3091 { 3092 int error; 3093 3094 #ifdef ALTQ 3095 KERNEL_LOCK(1, NULL); 3096 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 3097 error = if_transmit(ifp, m); 3098 KERNEL_UNLOCK_ONE(NULL); 3099 } else { 3100 KERNEL_UNLOCK_ONE(NULL); 3101 error = (*ifp->if_transmit)(ifp, m); 3102 } 3103 #else /* !ALTQ */ 3104 error = (*ifp->if_transmit)(ifp, m); 3105 #endif /* !ALTQ */ 3106 3107 return error; 3108 } 3109 3110 /* 3111 * Queue message on interface, and start output if interface 3112 * not yet active. 3113 */ 3114 int 3115 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 3116 { 3117 3118 return if_transmit_lock(ifp, m); 3119 } 3120 3121 /* 3122 * Queue message on interface, possibly using a second fast queue 3123 */ 3124 int 3125 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 3126 { 3127 int error = 0; 3128 3129 if (ifq != NULL 3130 #ifdef ALTQ 3131 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 3132 #endif 3133 ) { 3134 if (IF_QFULL(ifq)) { 3135 IF_DROP(&ifp->if_snd); 3136 m_freem(m); 3137 if (error == 0) 3138 error = ENOBUFS; 3139 } else 3140 IF_ENQUEUE(ifq, m); 3141 } else 3142 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3143 if (error != 0) { 3144 ++ifp->if_oerrors; 3145 return error; 3146 } 3147 return 0; 3148 } 3149 3150 int 3151 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 3152 { 3153 int rc; 3154 3155 if (ifp->if_initaddr != NULL) 3156 rc = (*ifp->if_initaddr)(ifp, ifa, src); 3157 else if (src || 3158 /* FIXME: may not hold if_ioctl_lock */ 3159 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 3160 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 3161 3162 return rc; 3163 } 3164 3165 int 3166 if_do_dad(struct ifnet *ifp) 3167 { 3168 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 3169 return 0; 3170 3171 switch (ifp->if_type) { 3172 case IFT_FAITH: 3173 /* 3174 * These interfaces do not have the IFF_LOOPBACK flag, 3175 * but loop packets back. We do not have to do DAD on such 3176 * interfaces. We should even omit it, because loop-backed 3177 * responses would confuse the DAD procedure. 3178 */ 3179 return 0; 3180 default: 3181 /* 3182 * Our DAD routine requires the interface up and running. 3183 * However, some interfaces can be up before the RUNNING 3184 * status. Additionaly, users may try to assign addresses 3185 * before the interface becomes up (or running). 3186 * We simply skip DAD in such a case as a work around. 3187 * XXX: we should rather mark "tentative" on such addresses, 3188 * and do DAD after the interface becomes ready. 3189 */ 3190 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 3191 (IFF_UP|IFF_RUNNING)) 3192 return 0; 3193 3194 return 1; 3195 } 3196 } 3197 3198 int 3199 if_flags_set(ifnet_t *ifp, const short flags) 3200 { 3201 int rc; 3202 3203 if (ifp->if_setflags != NULL) 3204 rc = (*ifp->if_setflags)(ifp, flags); 3205 else { 3206 short cantflags, chgdflags; 3207 struct ifreq ifr; 3208 3209 chgdflags = ifp->if_flags ^ flags; 3210 cantflags = chgdflags & IFF_CANTCHANGE; 3211 3212 if (cantflags != 0) 3213 ifp->if_flags ^= cantflags; 3214 3215 /* Traditionally, we do not call if_ioctl after 3216 * setting/clearing only IFF_PROMISC if the interface 3217 * isn't IFF_UP. Uphold that tradition. 3218 */ 3219 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 3220 return 0; 3221 3222 memset(&ifr, 0, sizeof(ifr)); 3223 3224 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 3225 /* FIXME: may not hold if_ioctl_lock */ 3226 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 3227 3228 if (rc != 0 && cantflags != 0) 3229 ifp->if_flags ^= cantflags; 3230 } 3231 3232 return rc; 3233 } 3234 3235 int 3236 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 3237 { 3238 int rc; 3239 struct ifreq ifr; 3240 3241 if (ifp->if_mcastop != NULL) 3242 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 3243 else { 3244 ifreq_setaddr(cmd, &ifr, sa); 3245 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 3246 } 3247 3248 return rc; 3249 } 3250 3251 static void 3252 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 3253 struct ifaltq *ifq) 3254 { 3255 const struct sysctlnode *cnode, *rnode; 3256 3257 if (sysctl_createv(clog, 0, NULL, &rnode, 3258 CTLFLAG_PERMANENT, 3259 CTLTYPE_NODE, "interfaces", 3260 SYSCTL_DESCR("Per-interface controls"), 3261 NULL, 0, NULL, 0, 3262 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3263 goto bad; 3264 3265 if (sysctl_createv(clog, 0, &rnode, &rnode, 3266 CTLFLAG_PERMANENT, 3267 CTLTYPE_NODE, ifname, 3268 SYSCTL_DESCR("Interface controls"), 3269 NULL, 0, NULL, 0, 3270 CTL_CREATE, CTL_EOL) != 0) 3271 goto bad; 3272 3273 if (sysctl_createv(clog, 0, &rnode, &rnode, 3274 CTLFLAG_PERMANENT, 3275 CTLTYPE_NODE, "sndq", 3276 SYSCTL_DESCR("Interface output queue controls"), 3277 NULL, 0, NULL, 0, 3278 CTL_CREATE, CTL_EOL) != 0) 3279 goto bad; 3280 3281 if (sysctl_createv(clog, 0, &rnode, &cnode, 3282 CTLFLAG_PERMANENT, 3283 CTLTYPE_INT, "len", 3284 SYSCTL_DESCR("Current output queue length"), 3285 NULL, 0, &ifq->ifq_len, 0, 3286 CTL_CREATE, CTL_EOL) != 0) 3287 goto bad; 3288 3289 if (sysctl_createv(clog, 0, &rnode, &cnode, 3290 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3291 CTLTYPE_INT, "maxlen", 3292 SYSCTL_DESCR("Maximum allowed output queue length"), 3293 NULL, 0, &ifq->ifq_maxlen, 0, 3294 CTL_CREATE, CTL_EOL) != 0) 3295 goto bad; 3296 3297 if (sysctl_createv(clog, 0, &rnode, &cnode, 3298 CTLFLAG_PERMANENT, 3299 CTLTYPE_INT, "drops", 3300 SYSCTL_DESCR("Packets dropped due to full output queue"), 3301 NULL, 0, &ifq->ifq_drops, 0, 3302 CTL_CREATE, CTL_EOL) != 0) 3303 goto bad; 3304 3305 return; 3306 bad: 3307 printf("%s: could not attach sysctl nodes\n", ifname); 3308 return; 3309 } 3310 3311 #if defined(INET) || defined(INET6) 3312 3313 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3314 static int \ 3315 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3316 { \ 3317 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3318 } 3319 3320 #if defined(INET) 3321 static int 3322 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3323 { 3324 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3325 } 3326 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3327 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3328 #endif 3329 3330 #if defined(INET6) 3331 static int 3332 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3333 { 3334 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3335 } 3336 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3337 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3338 #endif 3339 3340 static void 3341 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3342 { 3343 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3344 const char *pfname = NULL, *ipname = NULL; 3345 int ipn = 0, qid = 0; 3346 3347 switch (pf) { 3348 #if defined(INET) 3349 case PF_INET: 3350 len_func = sysctl_net_ip_pktq_items; 3351 maxlen_func = sysctl_net_ip_pktq_maxlen; 3352 drops_func = sysctl_net_ip_pktq_drops; 3353 pfname = "inet", ipn = IPPROTO_IP; 3354 ipname = "ip", qid = IPCTL_IFQ; 3355 break; 3356 #endif 3357 #if defined(INET6) 3358 case PF_INET6: 3359 len_func = sysctl_net_ip6_pktq_items; 3360 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3361 drops_func = sysctl_net_ip6_pktq_drops; 3362 pfname = "inet6", ipn = IPPROTO_IPV6; 3363 ipname = "ip6", qid = IPV6CTL_IFQ; 3364 break; 3365 #endif 3366 default: 3367 KASSERT(false); 3368 } 3369 3370 sysctl_createv(clog, 0, NULL, NULL, 3371 CTLFLAG_PERMANENT, 3372 CTLTYPE_NODE, pfname, NULL, 3373 NULL, 0, NULL, 0, 3374 CTL_NET, pf, CTL_EOL); 3375 sysctl_createv(clog, 0, NULL, NULL, 3376 CTLFLAG_PERMANENT, 3377 CTLTYPE_NODE, ipname, NULL, 3378 NULL, 0, NULL, 0, 3379 CTL_NET, pf, ipn, CTL_EOL); 3380 sysctl_createv(clog, 0, NULL, NULL, 3381 CTLFLAG_PERMANENT, 3382 CTLTYPE_NODE, "ifq", 3383 SYSCTL_DESCR("Protocol input queue controls"), 3384 NULL, 0, NULL, 0, 3385 CTL_NET, pf, ipn, qid, CTL_EOL); 3386 3387 sysctl_createv(clog, 0, NULL, NULL, 3388 CTLFLAG_PERMANENT, 3389 CTLTYPE_INT, "len", 3390 SYSCTL_DESCR("Current input queue length"), 3391 len_func, 0, NULL, 0, 3392 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3393 sysctl_createv(clog, 0, NULL, NULL, 3394 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3395 CTLTYPE_INT, "maxlen", 3396 SYSCTL_DESCR("Maximum allowed input queue length"), 3397 maxlen_func, 0, NULL, 0, 3398 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3399 sysctl_createv(clog, 0, NULL, NULL, 3400 CTLFLAG_PERMANENT, 3401 CTLTYPE_INT, "drops", 3402 SYSCTL_DESCR("Packets dropped due to full input queue"), 3403 drops_func, 0, NULL, 0, 3404 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3405 } 3406 #endif /* INET || INET6 */ 3407 3408 static int 3409 if_sdl_sysctl(SYSCTLFN_ARGS) 3410 { 3411 struct ifnet *ifp; 3412 const struct sockaddr_dl *sdl; 3413 struct psref psref; 3414 int error = 0; 3415 int bound; 3416 3417 if (namelen != 1) 3418 return EINVAL; 3419 3420 bound = curlwp_bind(); 3421 ifp = if_get_byindex(name[0], &psref); 3422 if (ifp == NULL) { 3423 error = ENODEV; 3424 goto out0; 3425 } 3426 3427 sdl = ifp->if_sadl; 3428 if (sdl == NULL) { 3429 *oldlenp = 0; 3430 goto out1; 3431 } 3432 3433 if (oldp == NULL) { 3434 *oldlenp = sdl->sdl_alen; 3435 goto out1; 3436 } 3437 3438 if (*oldlenp >= sdl->sdl_alen) 3439 *oldlenp = sdl->sdl_alen; 3440 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3441 out1: 3442 if_put(ifp, &psref); 3443 out0: 3444 curlwp_bindx(bound); 3445 return error; 3446 } 3447 3448 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 3449 { 3450 const struct sysctlnode *rnode = NULL; 3451 3452 sysctl_createv(clog, 0, NULL, &rnode, 3453 CTLFLAG_PERMANENT, 3454 CTLTYPE_NODE, "sdl", 3455 SYSCTL_DESCR("Get active link-layer address"), 3456 if_sdl_sysctl, 0, NULL, 0, 3457 CTL_NET, CTL_CREATE, CTL_EOL); 3458 } 3459
Indexes created Mon Sep 22 10:09:38 GMT 2025