1 /* $NetBSD: if.c,v 1.309 2015/04/07 23:30:36 roy 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 1.309 2015/04/07 23:30:36 roy Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 98 #include "opt_atalk.h" 99 #include "opt_natm.h" 100 #include "opt_wlan.h" 101 #include "opt_net_mpsafe.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> 120 121 #include <net/if.h> 122 #include <net/if_dl.h> 123 #include <net/if_ether.h> 124 #include <net/if_media.h> 125 #include <net80211/ieee80211.h> 126 #include <net80211/ieee80211_ioctl.h> 127 #include <net/if_types.h> 128 #include <net/radix.h> 129 #include <net/route.h> 130 #include <net/netisr.h> 131 #include <sys/module.h> 132 #ifdef NETATALK 133 #include <netatalk/at_extern.h> 134 #include <netatalk/at.h> 135 #endif 136 #include <net/pfil.h> 137 #include <netinet/in.h> 138 #include <netinet/in_var.h> 139 140 #ifdef INET6 141 #include <netinet6/in6_var.h> 142 #include <netinet6/nd6.h> 143 #endif 144 145 #include "ether.h" 146 #include "fddi.h" 147 #include "token.h" 148 149 #include "carp.h" 150 #if NCARP > 0 151 #include <netinet/ip_carp.h> 152 #endif 153 154 #include <compat/sys/sockio.h> 155 #include <compat/sys/socket.h> 156 157 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 158 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 159 160 /* 161 * Global list of interfaces. 162 */ 163 struct ifnet_head ifnet_list; 164 static ifnet_t ** ifindex2ifnet = NULL; 165 166 static u_int if_index = 1; 167 static size_t if_indexlim = 0; 168 static uint64_t index_gen; 169 static kmutex_t index_gen_mtx; 170 static kmutex_t if_clone_mtx; 171 172 static struct ifaddr ** ifnet_addrs = NULL; 173 174 struct ifnet *lo0ifp; 175 int ifqmaxlen = IFQ_MAXLEN; 176 177 static int if_rt_walktree(struct rtentry *, void *); 178 179 static struct if_clone *if_clone_lookup(const char *, int *); 180 static int if_clone_list(struct if_clonereq *); 181 182 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 183 static int if_cloners_count; 184 185 /* Packet filtering hook for interfaces. */ 186 pfil_head_t * if_pfil; 187 188 static kauth_listener_t if_listener; 189 190 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 191 static int ifioctl_attach(struct ifnet *); 192 static void ifioctl_detach(struct ifnet *); 193 static void ifnet_lock_enter(struct ifnet_lock *); 194 static void ifnet_lock_exit(struct ifnet_lock *); 195 static void if_detach_queues(struct ifnet *, struct ifqueue *); 196 static void sysctl_sndq_setup(struct sysctllog **, const char *, 197 struct ifaltq *); 198 static void if_slowtimo(void *); 199 static void if_free_sadl(struct ifnet *); 200 static void if_attachdomain1(struct ifnet *); 201 static int ifconf(u_long, void *); 202 static int if_clone_create(const char *); 203 static int if_clone_destroy(const char *); 204 205 #if defined(INET) || defined(INET6) 206 static void sysctl_net_pktq_setup(struct sysctllog **, int); 207 #endif 208 209 static int 210 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 211 void *arg0, void *arg1, void *arg2, void *arg3) 212 { 213 int result; 214 enum kauth_network_req req; 215 216 result = KAUTH_RESULT_DEFER; 217 req = (enum kauth_network_req)arg1; 218 219 if (action != KAUTH_NETWORK_INTERFACE) 220 return result; 221 222 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 223 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 224 result = KAUTH_RESULT_ALLOW; 225 226 return result; 227 } 228 229 /* 230 * Network interface utility routines. 231 * 232 * Routines with ifa_ifwith* names take sockaddr *'s as 233 * parameters. 234 */ 235 void 236 ifinit(void) 237 { 238 #if defined(INET) 239 sysctl_net_pktq_setup(NULL, PF_INET); 240 #endif 241 #ifdef INET6 242 sysctl_net_pktq_setup(NULL, PF_INET6); 243 #endif 244 245 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 246 if_listener_cb, NULL); 247 248 /* interfaces are available, inform socket code */ 249 ifioctl = doifioctl; 250 } 251 252 /* 253 * XXX Initialization before configure(). 254 * XXX hack to get pfil_add_hook working in autoconf. 255 */ 256 void 257 ifinit1(void) 258 { 259 mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE); 260 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 261 TAILQ_INIT(&ifnet_list); 262 if_indexlim = 8; 263 264 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 265 KASSERT(if_pfil != NULL); 266 267 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 268 etherinit(); 269 #endif 270 } 271 272 ifnet_t * 273 if_alloc(u_char type) 274 { 275 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 276 } 277 278 void 279 if_free(ifnet_t *ifp) 280 { 281 kmem_free(ifp, sizeof(ifnet_t)); 282 } 283 284 void 285 if_initname(struct ifnet *ifp, const char *name, int unit) 286 { 287 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 288 "%s%d", name, unit); 289 } 290 291 /* 292 * Null routines used while an interface is going away. These routines 293 * just return an error. 294 */ 295 296 int 297 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 298 const struct sockaddr *so, struct rtentry *rt) 299 { 300 301 return ENXIO; 302 } 303 304 void 305 if_nullinput(struct ifnet *ifp, struct mbuf *m) 306 { 307 308 /* Nothing. */ 309 } 310 311 void 312 if_nullstart(struct ifnet *ifp) 313 { 314 315 /* Nothing. */ 316 } 317 318 int 319 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 320 { 321 322 /* Wake ifioctl_detach(), who may wait for all threads to 323 * quit the critical section. 324 */ 325 cv_signal(&ifp->if_ioctl_lock->il_emptied); 326 return ENXIO; 327 } 328 329 int 330 if_nullinit(struct ifnet *ifp) 331 { 332 333 return ENXIO; 334 } 335 336 void 337 if_nullstop(struct ifnet *ifp, int disable) 338 { 339 340 /* Nothing. */ 341 } 342 343 void 344 if_nullslowtimo(struct ifnet *ifp) 345 { 346 347 /* Nothing. */ 348 } 349 350 void 351 if_nulldrain(struct ifnet *ifp) 352 { 353 354 /* Nothing. */ 355 } 356 357 void 358 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 359 { 360 struct ifaddr *ifa; 361 struct sockaddr_dl *sdl; 362 363 ifp->if_addrlen = addrlen; 364 if_alloc_sadl(ifp); 365 ifa = ifp->if_dl; 366 sdl = satosdl(ifa->ifa_addr); 367 368 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 369 if (factory) { 370 ifp->if_hwdl = ifp->if_dl; 371 ifaref(ifp->if_hwdl); 372 } 373 /* TBD routing socket */ 374 } 375 376 struct ifaddr * 377 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 378 { 379 unsigned socksize, ifasize; 380 int addrlen, namelen; 381 struct sockaddr_dl *mask, *sdl; 382 struct ifaddr *ifa; 383 384 namelen = strlen(ifp->if_xname); 385 addrlen = ifp->if_addrlen; 386 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 387 ifasize = sizeof(*ifa) + 2 * socksize; 388 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 389 390 sdl = (struct sockaddr_dl *)(ifa + 1); 391 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 392 393 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 394 ifp->if_xname, namelen, NULL, addrlen); 395 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 396 memset(&mask->sdl_data[0], 0xff, namelen); 397 ifa->ifa_rtrequest = link_rtrequest; 398 ifa->ifa_addr = (struct sockaddr *)sdl; 399 ifa->ifa_netmask = (struct sockaddr *)mask; 400 401 *sdlp = sdl; 402 403 return ifa; 404 } 405 406 static void 407 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 408 { 409 const struct sockaddr_dl *sdl; 410 ifnet_addrs[ifp->if_index] = ifa; 411 ifaref(ifa); 412 ifp->if_dl = ifa; 413 ifaref(ifa); 414 sdl = satosdl(ifa->ifa_addr); 415 ifp->if_sadl = sdl; 416 } 417 418 /* 419 * Allocate the link level name for the specified interface. This 420 * is an attachment helper. It must be called after ifp->if_addrlen 421 * is initialized, which may not be the case when if_attach() is 422 * called. 423 */ 424 void 425 if_alloc_sadl(struct ifnet *ifp) 426 { 427 struct ifaddr *ifa; 428 const struct sockaddr_dl *sdl; 429 430 /* 431 * If the interface already has a link name, release it 432 * now. This is useful for interfaces that can change 433 * link types, and thus switch link names often. 434 */ 435 if (ifp->if_sadl != NULL) 436 if_free_sadl(ifp); 437 438 ifa = if_dl_create(ifp, &sdl); 439 440 ifa_insert(ifp, ifa); 441 if_sadl_setrefs(ifp, ifa); 442 } 443 444 static void 445 if_deactivate_sadl(struct ifnet *ifp) 446 { 447 struct ifaddr *ifa; 448 449 KASSERT(ifp->if_dl != NULL); 450 451 ifa = ifp->if_dl; 452 453 ifp->if_sadl = NULL; 454 455 ifnet_addrs[ifp->if_index] = NULL; 456 ifafree(ifa); 457 ifp->if_dl = NULL; 458 ifafree(ifa); 459 } 460 461 void 462 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 463 const struct sockaddr_dl *sdl) 464 { 465 int s; 466 467 s = splnet(); 468 469 if_deactivate_sadl(ifp); 470 471 if_sadl_setrefs(ifp, ifa); 472 IFADDR_FOREACH(ifa, ifp) 473 rtinit(ifa, RTM_LLINFO_UPD, 0); 474 splx(s); 475 } 476 477 /* 478 * Free the link level name for the specified interface. This is 479 * a detach helper. This is called from if_detach(). 480 */ 481 static void 482 if_free_sadl(struct ifnet *ifp) 483 { 484 struct ifaddr *ifa; 485 int s; 486 487 ifa = ifnet_addrs[ifp->if_index]; 488 if (ifa == NULL) { 489 KASSERT(ifp->if_sadl == NULL); 490 KASSERT(ifp->if_dl == NULL); 491 return; 492 } 493 494 KASSERT(ifp->if_sadl != NULL); 495 KASSERT(ifp->if_dl != NULL); 496 497 s = splnet(); 498 rtinit(ifa, RTM_DELETE, 0); 499 ifa_remove(ifp, ifa); 500 if_deactivate_sadl(ifp); 501 if (ifp->if_hwdl == ifa) { 502 ifafree(ifa); 503 ifp->if_hwdl = NULL; 504 } 505 splx(s); 506 } 507 508 static void 509 if_getindex(ifnet_t *ifp) 510 { 511 bool hitlimit = false; 512 513 mutex_enter(&index_gen_mtx); 514 ifp->if_index_gen = index_gen++; 515 mutex_exit(&index_gen_mtx); 516 517 ifp->if_index = if_index; 518 if (ifindex2ifnet == NULL) { 519 if_index++; 520 goto skip; 521 } 522 while (if_byindex(ifp->if_index)) { 523 /* 524 * If we hit USHRT_MAX, we skip back to 0 since 525 * there are a number of places where the value 526 * of if_index or if_index itself is compared 527 * to or stored in an unsigned short. By 528 * jumping back, we won't botch those assignments 529 * or comparisons. 530 */ 531 if (++if_index == 0) { 532 if_index = 1; 533 } else if (if_index == USHRT_MAX) { 534 /* 535 * However, if we have to jump back to 536 * zero *twice* without finding an empty 537 * slot in ifindex2ifnet[], then there 538 * there are too many (>65535) interfaces. 539 */ 540 if (hitlimit) { 541 panic("too many interfaces"); 542 } 543 hitlimit = true; 544 if_index = 1; 545 } 546 ifp->if_index = if_index; 547 } 548 skip: 549 /* 550 * We have some arrays that should be indexed by if_index. 551 * since if_index will grow dynamically, they should grow too. 552 * struct ifadd **ifnet_addrs 553 * struct ifnet **ifindex2ifnet 554 */ 555 if (ifnet_addrs == NULL || ifindex2ifnet == NULL || 556 ifp->if_index >= if_indexlim) { 557 size_t m, n, oldlim; 558 void *q; 559 560 oldlim = if_indexlim; 561 while (ifp->if_index >= if_indexlim) 562 if_indexlim <<= 1; 563 564 /* grow ifnet_addrs */ 565 m = oldlim * sizeof(struct ifaddr *); 566 n = if_indexlim * sizeof(struct ifaddr *); 567 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 568 if (ifnet_addrs != NULL) { 569 memcpy(q, ifnet_addrs, m); 570 free(ifnet_addrs, M_IFADDR); 571 } 572 ifnet_addrs = (struct ifaddr **)q; 573 574 /* grow ifindex2ifnet */ 575 m = oldlim * sizeof(struct ifnet *); 576 n = if_indexlim * sizeof(struct ifnet *); 577 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 578 if (ifindex2ifnet != NULL) { 579 memcpy(q, ifindex2ifnet, m); 580 free(ifindex2ifnet, M_IFADDR); 581 } 582 ifindex2ifnet = (struct ifnet **)q; 583 } 584 ifindex2ifnet[ifp->if_index] = ifp; 585 } 586 587 /* 588 * Initialize an interface and assign an index for it. 589 * 590 * It must be called prior to a device specific attach routine 591 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 592 * and be followed by if_register: 593 * 594 * if_initialize(ifp); 595 * ether_ifattach(ifp, enaddr); 596 * if_register(ifp); 597 */ 598 void 599 if_initialize(ifnet_t *ifp) 600 { 601 KASSERT(if_indexlim > 0); 602 TAILQ_INIT(&ifp->if_addrlist); 603 604 /* 605 * Link level name is allocated later by a separate call to 606 * if_alloc_sadl(). 607 */ 608 609 if (ifp->if_snd.ifq_maxlen == 0) 610 ifp->if_snd.ifq_maxlen = ifqmaxlen; 611 612 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 613 614 ifp->if_link_state = LINK_STATE_UNKNOWN; 615 616 ifp->if_capenable = 0; 617 ifp->if_csum_flags_tx = 0; 618 ifp->if_csum_flags_rx = 0; 619 620 #ifdef ALTQ 621 ifp->if_snd.altq_type = 0; 622 ifp->if_snd.altq_disc = NULL; 623 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 624 ifp->if_snd.altq_tbr = NULL; 625 ifp->if_snd.altq_ifp = ifp; 626 #endif 627 628 #ifdef NET_MPSAFE 629 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 630 #else 631 ifp->if_snd.ifq_lock = NULL; 632 #endif 633 634 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 635 (void)pfil_run_hooks(if_pfil, 636 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 637 638 if_getindex(ifp); 639 } 640 641 /* 642 * Register an interface to the list of "active" interfaces. 643 */ 644 void 645 if_register(ifnet_t *ifp) 646 { 647 if (ifioctl_attach(ifp) != 0) 648 panic("%s: ifioctl_attach() failed", __func__); 649 650 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 651 652 if (!STAILQ_EMPTY(&domains)) 653 if_attachdomain1(ifp); 654 655 /* Announce the interface. */ 656 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 657 658 if (ifp->if_slowtimo != NULL) { 659 ifp->if_slowtimo_ch = 660 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 661 callout_init(ifp->if_slowtimo_ch, 0); 662 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 663 if_slowtimo(ifp); 664 } 665 666 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 667 } 668 669 /* 670 * Deprecated. Use if_initialize and if_register instead. 671 * See the above comment of if_initialize. 672 */ 673 void 674 if_attach(ifnet_t *ifp) 675 { 676 if_initialize(ifp); 677 if_register(ifp); 678 } 679 680 void 681 if_attachdomain(void) 682 { 683 struct ifnet *ifp; 684 int s; 685 686 s = splnet(); 687 IFNET_FOREACH(ifp) 688 if_attachdomain1(ifp); 689 splx(s); 690 } 691 692 static void 693 if_attachdomain1(struct ifnet *ifp) 694 { 695 struct domain *dp; 696 int s; 697 698 s = splnet(); 699 700 /* address family dependent data region */ 701 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 702 DOMAIN_FOREACH(dp) { 703 if (dp->dom_ifattach != NULL) 704 ifp->if_afdata[dp->dom_family] = 705 (*dp->dom_ifattach)(ifp); 706 } 707 708 splx(s); 709 } 710 711 /* 712 * Deactivate an interface. This points all of the procedure 713 * handles at error stubs. May be called from interrupt context. 714 */ 715 void 716 if_deactivate(struct ifnet *ifp) 717 { 718 int s; 719 720 s = splnet(); 721 722 ifp->if_output = if_nulloutput; 723 ifp->if_input = if_nullinput; 724 ifp->if_start = if_nullstart; 725 ifp->if_ioctl = if_nullioctl; 726 ifp->if_init = if_nullinit; 727 ifp->if_stop = if_nullstop; 728 ifp->if_slowtimo = if_nullslowtimo; 729 ifp->if_drain = if_nulldrain; 730 731 /* No more packets may be enqueued. */ 732 ifp->if_snd.ifq_maxlen = 0; 733 734 splx(s); 735 } 736 737 void 738 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 739 { 740 struct ifaddr *ifa, *nifa; 741 742 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 743 if (ifa->ifa_addr->sa_family != family) 744 continue; 745 (*purgeaddr)(ifa); 746 } 747 } 748 749 /* 750 * Detach an interface from the list of "active" interfaces, 751 * freeing any resources as we go along. 752 * 753 * NOTE: This routine must be called with a valid thread context, 754 * as it may block. 755 */ 756 void 757 if_detach(struct ifnet *ifp) 758 { 759 struct socket so; 760 struct ifaddr *ifa; 761 #ifdef IFAREF_DEBUG 762 struct ifaddr *last_ifa = NULL; 763 #endif 764 struct domain *dp; 765 const struct protosw *pr; 766 int s, i, family, purged; 767 uint64_t xc; 768 769 /* 770 * XXX It's kind of lame that we have to have the 771 * XXX socket structure... 772 */ 773 memset(&so, 0, sizeof(so)); 774 775 s = splnet(); 776 777 if (ifp->if_slowtimo != NULL) { 778 ifp->if_slowtimo = NULL; 779 callout_halt(ifp->if_slowtimo_ch, NULL); 780 callout_destroy(ifp->if_slowtimo_ch); 781 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 782 } 783 784 /* 785 * Do an if_down() to give protocols a chance to do something. 786 */ 787 if_down(ifp); 788 789 #ifdef ALTQ 790 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 791 altq_disable(&ifp->if_snd); 792 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 793 altq_detach(&ifp->if_snd); 794 #endif 795 796 if (ifp->if_snd.ifq_lock) 797 mutex_obj_free(ifp->if_snd.ifq_lock); 798 799 sysctl_teardown(&ifp->if_sysctl_log); 800 801 #if NCARP > 0 802 /* Remove the interface from any carp group it is a part of. */ 803 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 804 carp_ifdetach(ifp); 805 #endif 806 807 /* 808 * Rip all the addresses off the interface. This should make 809 * all of the routes go away. 810 * 811 * pr_usrreq calls can remove an arbitrary number of ifaddrs 812 * from the list, including our "cursor", ifa. For safety, 813 * and to honor the TAILQ abstraction, I just restart the 814 * loop after each removal. Note that the loop will exit 815 * when all of the remaining ifaddrs belong to the AF_LINK 816 * family. I am counting on the historical fact that at 817 * least one pr_usrreq in each address domain removes at 818 * least one ifaddr. 819 */ 820 again: 821 IFADDR_FOREACH(ifa, ifp) { 822 family = ifa->ifa_addr->sa_family; 823 #ifdef IFAREF_DEBUG 824 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 825 ifa, family, ifa->ifa_refcnt); 826 if (last_ifa != NULL && ifa == last_ifa) 827 panic("if_detach: loop detected"); 828 last_ifa = ifa; 829 #endif 830 if (family == AF_LINK) 831 continue; 832 dp = pffinddomain(family); 833 #ifdef DIAGNOSTIC 834 if (dp == NULL) 835 panic("if_detach: no domain for AF %d", 836 family); 837 #endif 838 /* 839 * XXX These PURGEIF calls are redundant with the 840 * purge-all-families calls below, but are left in for 841 * now both to make a smaller change, and to avoid 842 * unplanned interactions with clearing of 843 * ifp->if_addrlist. 844 */ 845 purged = 0; 846 for (pr = dp->dom_protosw; 847 pr < dp->dom_protoswNPROTOSW; pr++) { 848 so.so_proto = pr; 849 if (pr->pr_usrreqs) { 850 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 851 purged = 1; 852 } 853 } 854 if (purged == 0) { 855 /* 856 * XXX What's really the best thing to do 857 * XXX here? --thorpej@NetBSD.org 858 */ 859 printf("if_detach: WARNING: AF %d not purged\n", 860 family); 861 ifa_remove(ifp, ifa); 862 } 863 goto again; 864 } 865 866 if_free_sadl(ifp); 867 868 /* Walk the routing table looking for stragglers. */ 869 for (i = 0; i <= AF_MAX; i++) { 870 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 871 continue; 872 } 873 874 DOMAIN_FOREACH(dp) { 875 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 876 { 877 void *p = ifp->if_afdata[dp->dom_family]; 878 if (p) { 879 ifp->if_afdata[dp->dom_family] = NULL; 880 (*dp->dom_ifdetach)(ifp, p); 881 } 882 } 883 884 /* 885 * One would expect multicast memberships (INET and 886 * INET6) on UDP sockets to be purged by the PURGEIF 887 * calls above, but if all addresses were removed from 888 * the interface prior to destruction, the calls will 889 * not be made (e.g. ppp, for which pppd(8) generally 890 * removes addresses before destroying the interface). 891 * Because there is no invariant that multicast 892 * memberships only exist for interfaces with IPv4 893 * addresses, we must call PURGEIF regardless of 894 * addresses. (Protocols which might store ifnet 895 * pointers are marked with PR_PURGEIF.) 896 */ 897 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 898 so.so_proto = pr; 899 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 900 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 901 } 902 } 903 904 (void)pfil_run_hooks(if_pfil, 905 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 906 (void)pfil_head_destroy(ifp->if_pfil); 907 908 /* Announce that the interface is gone. */ 909 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 910 911 ifindex2ifnet[ifp->if_index] = NULL; 912 913 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 914 915 ifioctl_detach(ifp); 916 917 /* 918 * remove packets that came from ifp, from software interrupt queues. 919 */ 920 DOMAIN_FOREACH(dp) { 921 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 922 struct ifqueue *iq = dp->dom_ifqueues[i]; 923 if (iq == NULL) 924 break; 925 dp->dom_ifqueues[i] = NULL; 926 if_detach_queues(ifp, iq); 927 } 928 } 929 930 /* 931 * IP queues have to be processed separately: net-queue barrier 932 * ensures that the packets are dequeued while a cross-call will 933 * ensure that the interrupts have completed. FIXME: not quite.. 934 */ 935 #ifdef INET 936 pktq_barrier(ip_pktq); 937 #endif 938 #ifdef INET6 939 if (in6_present) 940 pktq_barrier(ip6_pktq); 941 #endif 942 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 943 xc_wait(xc); 944 945 splx(s); 946 } 947 948 static void 949 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 950 { 951 struct mbuf *m, *prev, *next; 952 953 prev = NULL; 954 for (m = q->ifq_head; m != NULL; m = next) { 955 KASSERT((m->m_flags & M_PKTHDR) != 0); 956 957 next = m->m_nextpkt; 958 if (m->m_pkthdr.rcvif != ifp) { 959 prev = m; 960 continue; 961 } 962 963 if (prev != NULL) 964 prev->m_nextpkt = m->m_nextpkt; 965 else 966 q->ifq_head = m->m_nextpkt; 967 if (q->ifq_tail == m) 968 q->ifq_tail = prev; 969 q->ifq_len--; 970 971 m->m_nextpkt = NULL; 972 m_freem(m); 973 IF_DROP(q); 974 } 975 } 976 977 /* 978 * Callback for a radix tree walk to delete all references to an 979 * ifnet. 980 */ 981 static int 982 if_rt_walktree(struct rtentry *rt, void *v) 983 { 984 struct ifnet *ifp = (struct ifnet *)v; 985 int error; 986 987 if (rt->rt_ifp != ifp) 988 return 0; 989 990 /* Delete the entry. */ 991 ++rt->rt_refcnt; 992 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 993 rt_mask(rt), rt->rt_flags, NULL); 994 KASSERT((rt->rt_flags & RTF_UP) == 0); 995 rt->rt_ifp = NULL; 996 rtfree(rt); 997 if (error != 0) 998 printf("%s: warning: unable to delete rtentry @ %p, " 999 "error = %d\n", ifp->if_xname, rt, error); 1000 return ERESTART; 1001 } 1002 1003 /* 1004 * Create a clone network interface. 1005 */ 1006 static int 1007 if_clone_create(const char *name) 1008 { 1009 struct if_clone *ifc; 1010 int unit; 1011 1012 ifc = if_clone_lookup(name, &unit); 1013 if (ifc == NULL) 1014 return EINVAL; 1015 1016 if (ifunit(name) != NULL) 1017 return EEXIST; 1018 1019 return (*ifc->ifc_create)(ifc, unit); 1020 } 1021 1022 /* 1023 * Destroy a clone network interface. 1024 */ 1025 static int 1026 if_clone_destroy(const char *name) 1027 { 1028 struct if_clone *ifc; 1029 struct ifnet *ifp; 1030 1031 ifc = if_clone_lookup(name, NULL); 1032 if (ifc == NULL) 1033 return EINVAL; 1034 1035 ifp = ifunit(name); 1036 if (ifp == NULL) 1037 return ENXIO; 1038 1039 if (ifc->ifc_destroy == NULL) 1040 return EOPNOTSUPP; 1041 1042 return (*ifc->ifc_destroy)(ifp); 1043 } 1044 1045 /* 1046 * Look up a network interface cloner. 1047 */ 1048 static struct if_clone * 1049 if_clone_lookup(const char *name, int *unitp) 1050 { 1051 struct if_clone *ifc; 1052 const char *cp; 1053 char *dp, ifname[IFNAMSIZ + 3]; 1054 int unit; 1055 1056 strcpy(ifname, "if_"); 1057 /* separate interface name from unit */ 1058 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1059 *cp && (*cp < '0' || *cp > '9');) 1060 *dp++ = *cp++; 1061 1062 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1063 return NULL; /* No name or unit number */ 1064 *dp++ = '\0'; 1065 1066 again: 1067 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1068 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1069 break; 1070 } 1071 1072 if (ifc == NULL) { 1073 if (*ifname == '\0' || 1074 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1075 return NULL; 1076 *ifname = '\0'; 1077 goto again; 1078 } 1079 1080 unit = 0; 1081 while (cp - name < IFNAMSIZ && *cp) { 1082 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1083 /* Bogus unit number. */ 1084 return NULL; 1085 } 1086 unit = (unit * 10) + (*cp++ - '0'); 1087 } 1088 1089 if (unitp != NULL) 1090 *unitp = unit; 1091 return ifc; 1092 } 1093 1094 /* 1095 * Register a network interface cloner. 1096 */ 1097 void 1098 if_clone_attach(struct if_clone *ifc) 1099 { 1100 1101 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1102 if_cloners_count++; 1103 } 1104 1105 /* 1106 * Unregister a network interface cloner. 1107 */ 1108 void 1109 if_clone_detach(struct if_clone *ifc) 1110 { 1111 1112 LIST_REMOVE(ifc, ifc_list); 1113 if_cloners_count--; 1114 } 1115 1116 /* 1117 * Provide list of interface cloners to userspace. 1118 */ 1119 static int 1120 if_clone_list(struct if_clonereq *ifcr) 1121 { 1122 char outbuf[IFNAMSIZ], *dst; 1123 struct if_clone *ifc; 1124 int count, error = 0; 1125 1126 ifcr->ifcr_total = if_cloners_count; 1127 if ((dst = ifcr->ifcr_buffer) == NULL) { 1128 /* Just asking how many there are. */ 1129 return 0; 1130 } 1131 1132 if (ifcr->ifcr_count < 0) 1133 return EINVAL; 1134 1135 count = (if_cloners_count < ifcr->ifcr_count) ? 1136 if_cloners_count : ifcr->ifcr_count; 1137 1138 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1139 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1140 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1141 if (outbuf[sizeof(outbuf) - 1] != '\0') 1142 return ENAMETOOLONG; 1143 error = copyout(outbuf, dst, sizeof(outbuf)); 1144 if (error != 0) 1145 break; 1146 } 1147 1148 return error; 1149 } 1150 1151 void 1152 ifaref(struct ifaddr *ifa) 1153 { 1154 ifa->ifa_refcnt++; 1155 } 1156 1157 void 1158 ifafree(struct ifaddr *ifa) 1159 { 1160 KASSERT(ifa != NULL); 1161 KASSERT(ifa->ifa_refcnt > 0); 1162 1163 if (--ifa->ifa_refcnt == 0) { 1164 free(ifa, M_IFADDR); 1165 } 1166 } 1167 1168 void 1169 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1170 { 1171 ifa->ifa_ifp = ifp; 1172 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1173 ifaref(ifa); 1174 } 1175 1176 void 1177 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1178 { 1179 KASSERT(ifa->ifa_ifp == ifp); 1180 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1181 ifafree(ifa); 1182 } 1183 1184 static inline int 1185 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1186 { 1187 return sockaddr_cmp(sa1, sa2) == 0; 1188 } 1189 1190 /* 1191 * Locate an interface based on a complete address. 1192 */ 1193 /*ARGSUSED*/ 1194 struct ifaddr * 1195 ifa_ifwithaddr(const struct sockaddr *addr) 1196 { 1197 struct ifnet *ifp; 1198 struct ifaddr *ifa; 1199 1200 IFNET_FOREACH(ifp) { 1201 if (ifp->if_output == if_nulloutput) 1202 continue; 1203 IFADDR_FOREACH(ifa, ifp) { 1204 if (ifa->ifa_addr->sa_family != addr->sa_family) 1205 continue; 1206 if (equal(addr, ifa->ifa_addr)) 1207 return ifa; 1208 if ((ifp->if_flags & IFF_BROADCAST) && 1209 ifa->ifa_broadaddr && 1210 /* IP6 doesn't have broadcast */ 1211 ifa->ifa_broadaddr->sa_len != 0 && 1212 equal(ifa->ifa_broadaddr, addr)) 1213 return ifa; 1214 } 1215 } 1216 return NULL; 1217 } 1218 1219 /* 1220 * Locate the point to point interface with a given destination address. 1221 */ 1222 /*ARGSUSED*/ 1223 struct ifaddr * 1224 ifa_ifwithdstaddr(const struct sockaddr *addr) 1225 { 1226 struct ifnet *ifp; 1227 struct ifaddr *ifa; 1228 1229 IFNET_FOREACH(ifp) { 1230 if (ifp->if_output == if_nulloutput) 1231 continue; 1232 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1233 continue; 1234 IFADDR_FOREACH(ifa, ifp) { 1235 if (ifa->ifa_addr->sa_family != addr->sa_family || 1236 ifa->ifa_dstaddr == NULL) 1237 continue; 1238 if (equal(addr, ifa->ifa_dstaddr)) 1239 return ifa; 1240 } 1241 } 1242 return NULL; 1243 } 1244 1245 /* 1246 * Find an interface on a specific network. If many, choice 1247 * is most specific found. 1248 */ 1249 struct ifaddr * 1250 ifa_ifwithnet(const struct sockaddr *addr) 1251 { 1252 struct ifnet *ifp; 1253 struct ifaddr *ifa; 1254 const struct sockaddr_dl *sdl; 1255 struct ifaddr *ifa_maybe = 0; 1256 u_int af = addr->sa_family; 1257 const char *addr_data = addr->sa_data, *cplim; 1258 1259 if (af == AF_LINK) { 1260 sdl = satocsdl(addr); 1261 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1262 ifindex2ifnet[sdl->sdl_index] && 1263 ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) 1264 return ifnet_addrs[sdl->sdl_index]; 1265 } 1266 #ifdef NETATALK 1267 if (af == AF_APPLETALK) { 1268 const struct sockaddr_at *sat, *sat2; 1269 sat = (const struct sockaddr_at *)addr; 1270 IFNET_FOREACH(ifp) { 1271 if (ifp->if_output == if_nulloutput) 1272 continue; 1273 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1274 if (ifa == NULL) 1275 continue; 1276 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1277 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1278 return ifa; /* exact match */ 1279 if (ifa_maybe == NULL) { 1280 /* else keep the if with the right range */ 1281 ifa_maybe = ifa; 1282 } 1283 } 1284 return ifa_maybe; 1285 } 1286 #endif 1287 IFNET_FOREACH(ifp) { 1288 if (ifp->if_output == if_nulloutput) 1289 continue; 1290 IFADDR_FOREACH(ifa, ifp) { 1291 const char *cp, *cp2, *cp3; 1292 1293 if (ifa->ifa_addr->sa_family != af || 1294 ifa->ifa_netmask == NULL) 1295 next: continue; 1296 cp = addr_data; 1297 cp2 = ifa->ifa_addr->sa_data; 1298 cp3 = ifa->ifa_netmask->sa_data; 1299 cplim = (const char *)ifa->ifa_netmask + 1300 ifa->ifa_netmask->sa_len; 1301 while (cp3 < cplim) { 1302 if ((*cp++ ^ *cp2++) & *cp3++) { 1303 /* want to continue for() loop */ 1304 goto next; 1305 } 1306 } 1307 if (ifa_maybe == NULL || 1308 rn_refines((void *)ifa->ifa_netmask, 1309 (void *)ifa_maybe->ifa_netmask)) 1310 ifa_maybe = ifa; 1311 } 1312 } 1313 return ifa_maybe; 1314 } 1315 1316 /* 1317 * Find the interface of the addresss. 1318 */ 1319 struct ifaddr * 1320 ifa_ifwithladdr(const struct sockaddr *addr) 1321 { 1322 struct ifaddr *ia; 1323 1324 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1325 (ia = ifa_ifwithnet(addr))) 1326 return ia; 1327 return NULL; 1328 } 1329 1330 /* 1331 * Find an interface using a specific address family 1332 */ 1333 struct ifaddr * 1334 ifa_ifwithaf(int af) 1335 { 1336 struct ifnet *ifp; 1337 struct ifaddr *ifa; 1338 1339 IFNET_FOREACH(ifp) { 1340 if (ifp->if_output == if_nulloutput) 1341 continue; 1342 IFADDR_FOREACH(ifa, ifp) { 1343 if (ifa->ifa_addr->sa_family == af) 1344 return ifa; 1345 } 1346 } 1347 return NULL; 1348 } 1349 1350 /* 1351 * Find an interface address specific to an interface best matching 1352 * a given address. 1353 */ 1354 struct ifaddr * 1355 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1356 { 1357 struct ifaddr *ifa; 1358 const char *cp, *cp2, *cp3; 1359 const char *cplim; 1360 struct ifaddr *ifa_maybe = 0; 1361 u_int af = addr->sa_family; 1362 1363 if (ifp->if_output == if_nulloutput) 1364 return NULL; 1365 1366 if (af >= AF_MAX) 1367 return NULL; 1368 1369 IFADDR_FOREACH(ifa, ifp) { 1370 if (ifa->ifa_addr->sa_family != af) 1371 continue; 1372 ifa_maybe = ifa; 1373 if (ifa->ifa_netmask == NULL) { 1374 if (equal(addr, ifa->ifa_addr) || 1375 (ifa->ifa_dstaddr && 1376 equal(addr, ifa->ifa_dstaddr))) 1377 return ifa; 1378 continue; 1379 } 1380 cp = addr->sa_data; 1381 cp2 = ifa->ifa_addr->sa_data; 1382 cp3 = ifa->ifa_netmask->sa_data; 1383 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1384 for (; cp3 < cplim; cp3++) { 1385 if ((*cp++ ^ *cp2++) & *cp3) 1386 break; 1387 } 1388 if (cp3 == cplim) 1389 return ifa; 1390 } 1391 return ifa_maybe; 1392 } 1393 1394 /* 1395 * Default action when installing a route with a Link Level gateway. 1396 * Lookup an appropriate real ifa to point to. 1397 * This should be moved to /sys/net/link.c eventually. 1398 */ 1399 void 1400 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1401 { 1402 struct ifaddr *ifa; 1403 const struct sockaddr *dst; 1404 struct ifnet *ifp; 1405 1406 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1407 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1408 return; 1409 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1410 rt_replace_ifa(rt, ifa); 1411 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1412 ifa->ifa_rtrequest(cmd, rt, info); 1413 } 1414 } 1415 1416 /* 1417 * Handle a change in the interface link state. 1418 * XXX: We should listen to the routing socket in-kernel rather 1419 * than calling in6_if_link_* functions directly from here. 1420 */ 1421 void 1422 if_link_state_change(struct ifnet *ifp, int link_state) 1423 { 1424 int s; 1425 #if defined(DEBUG) || defined(INET6) 1426 int old_link_state; 1427 #endif 1428 1429 s = splnet(); 1430 if (ifp->if_link_state == link_state) { 1431 splx(s); 1432 return; 1433 } 1434 1435 #if defined(DEBUG) || defined(INET6) 1436 old_link_state = ifp->if_link_state; 1437 #endif 1438 ifp->if_link_state = link_state; 1439 #ifdef DEBUG 1440 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1441 link_state == LINK_STATE_UP ? "UP" : 1442 link_state == LINK_STATE_DOWN ? "DOWN" : 1443 "UNKNOWN", 1444 old_link_state == LINK_STATE_UP ? "UP" : 1445 old_link_state == LINK_STATE_DOWN ? "DOWN" : 1446 "UNKNOWN"); 1447 #endif 1448 1449 #ifdef INET6 1450 /* 1451 * When going from UNKNOWN to UP, we need to mark existing 1452 * IPv6 addresses as tentative and restart DAD as we may have 1453 * erroneously not found a duplicate. 1454 * 1455 * This needs to happen before rt_ifmsg to avoid a race where 1456 * listeners would have an address and expect it to work right 1457 * away. 1458 */ 1459 if (in6_present && link_state == LINK_STATE_UP && 1460 old_link_state == LINK_STATE_UNKNOWN) 1461 in6_if_link_down(ifp); 1462 #endif 1463 1464 /* Notify that the link state has changed. */ 1465 rt_ifmsg(ifp); 1466 1467 #if NCARP > 0 1468 if (ifp->if_carp) 1469 carp_carpdev_state(ifp); 1470 #endif 1471 1472 #ifdef INET6 1473 if (in6_present) { 1474 if (link_state == LINK_STATE_DOWN) 1475 in6_if_link_down(ifp); 1476 else if (link_state == LINK_STATE_UP) 1477 in6_if_link_up(ifp); 1478 } 1479 #endif 1480 1481 splx(s); 1482 } 1483 1484 /* 1485 * Mark an interface down and notify protocols of 1486 * the transition. 1487 * NOTE: must be called at splsoftnet or equivalent. 1488 */ 1489 void 1490 if_down(struct ifnet *ifp) 1491 { 1492 struct ifaddr *ifa; 1493 1494 ifp->if_flags &= ~IFF_UP; 1495 nanotime(&ifp->if_lastchange); 1496 IFADDR_FOREACH(ifa, ifp) 1497 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1498 IFQ_PURGE(&ifp->if_snd); 1499 #if NCARP > 0 1500 if (ifp->if_carp) 1501 carp_carpdev_state(ifp); 1502 #endif 1503 rt_ifmsg(ifp); 1504 #ifdef INET6 1505 if (in6_present) 1506 in6_if_down(ifp); 1507 #endif 1508 } 1509 1510 /* 1511 * Mark an interface up and notify protocols of 1512 * the transition. 1513 * NOTE: must be called at splsoftnet or equivalent. 1514 */ 1515 void 1516 if_up(struct ifnet *ifp) 1517 { 1518 #ifdef notyet 1519 struct ifaddr *ifa; 1520 #endif 1521 1522 ifp->if_flags |= IFF_UP; 1523 nanotime(&ifp->if_lastchange); 1524 #ifdef notyet 1525 /* this has no effect on IP, and will kill all ISO connections XXX */ 1526 IFADDR_FOREACH(ifa, ifp) 1527 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1528 #endif 1529 #if NCARP > 0 1530 if (ifp->if_carp) 1531 carp_carpdev_state(ifp); 1532 #endif 1533 rt_ifmsg(ifp); 1534 #ifdef INET6 1535 if (in6_present) 1536 in6_if_up(ifp); 1537 #endif 1538 } 1539 1540 /* 1541 * Handle interface slowtimo timer routine. Called 1542 * from softclock, we decrement timer (if set) and 1543 * call the appropriate interface routine on expiration. 1544 */ 1545 static void 1546 if_slowtimo(void *arg) 1547 { 1548 void (*slowtimo)(struct ifnet *); 1549 struct ifnet *ifp = arg; 1550 int s; 1551 1552 slowtimo = ifp->if_slowtimo; 1553 if (__predict_false(slowtimo == NULL)) 1554 return; 1555 1556 s = splnet(); 1557 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 1558 (*slowtimo)(ifp); 1559 1560 splx(s); 1561 1562 if (__predict_true(ifp->if_slowtimo != NULL)) 1563 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 1564 } 1565 1566 /* 1567 * Set/clear promiscuous mode on interface ifp based on the truth value 1568 * of pswitch. The calls are reference counted so that only the first 1569 * "on" request actually has an effect, as does the final "off" request. 1570 * Results are undefined if the "off" and "on" requests are not matched. 1571 */ 1572 int 1573 ifpromisc(struct ifnet *ifp, int pswitch) 1574 { 1575 int pcount, ret; 1576 short nflags; 1577 1578 pcount = ifp->if_pcount; 1579 if (pswitch) { 1580 /* 1581 * Allow the device to be "placed" into promiscuous 1582 * mode even if it is not configured up. It will 1583 * consult IFF_PROMISC when it is brought up. 1584 */ 1585 if (ifp->if_pcount++ != 0) 1586 return 0; 1587 nflags = ifp->if_flags | IFF_PROMISC; 1588 } else { 1589 if (--ifp->if_pcount > 0) 1590 return 0; 1591 nflags = ifp->if_flags & ~IFF_PROMISC; 1592 } 1593 ret = if_flags_set(ifp, nflags); 1594 /* Restore interface state if not successful. */ 1595 if (ret != 0) { 1596 ifp->if_pcount = pcount; 1597 } 1598 return ret; 1599 } 1600 1601 /* 1602 * Map interface name to 1603 * interface structure pointer. 1604 */ 1605 struct ifnet * 1606 ifunit(const char *name) 1607 { 1608 struct ifnet *ifp; 1609 const char *cp = name; 1610 u_int unit = 0; 1611 u_int i; 1612 1613 /* 1614 * If the entire name is a number, treat it as an ifindex. 1615 */ 1616 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 1617 unit = unit * 10 + (*cp - '0'); 1618 } 1619 1620 /* 1621 * If the number took all of the name, then it's a valid ifindex. 1622 */ 1623 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 1624 if (unit >= if_indexlim) 1625 return NULL; 1626 ifp = ifindex2ifnet[unit]; 1627 if (ifp == NULL || ifp->if_output == if_nulloutput) 1628 return NULL; 1629 return ifp; 1630 } 1631 1632 IFNET_FOREACH(ifp) { 1633 if (ifp->if_output == if_nulloutput) 1634 continue; 1635 if (strcmp(ifp->if_xname, name) == 0) 1636 return ifp; 1637 } 1638 return NULL; 1639 } 1640 1641 ifnet_t * 1642 if_byindex(u_int idx) 1643 { 1644 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 1645 } 1646 1647 /* common */ 1648 int 1649 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 1650 { 1651 int s; 1652 struct ifreq *ifr; 1653 struct ifcapreq *ifcr; 1654 struct ifdatareq *ifdr; 1655 1656 switch (cmd) { 1657 case SIOCSIFCAP: 1658 ifcr = data; 1659 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 1660 return EINVAL; 1661 1662 if (ifcr->ifcr_capenable == ifp->if_capenable) 1663 return 0; 1664 1665 ifp->if_capenable = ifcr->ifcr_capenable; 1666 1667 /* Pre-compute the checksum flags mask. */ 1668 ifp->if_csum_flags_tx = 0; 1669 ifp->if_csum_flags_rx = 0; 1670 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 1671 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 1672 } 1673 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 1674 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 1675 } 1676 1677 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 1678 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 1679 } 1680 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 1681 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 1682 } 1683 1684 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 1685 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 1686 } 1687 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 1688 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 1689 } 1690 1691 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 1692 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 1693 } 1694 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 1695 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 1696 } 1697 1698 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 1699 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 1700 } 1701 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 1702 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 1703 } 1704 if (ifp->if_flags & IFF_UP) 1705 return ENETRESET; 1706 return 0; 1707 case SIOCSIFFLAGS: 1708 ifr = data; 1709 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 1710 s = splnet(); 1711 if_down(ifp); 1712 splx(s); 1713 } 1714 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 1715 s = splnet(); 1716 if_up(ifp); 1717 splx(s); 1718 } 1719 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1720 (ifr->ifr_flags &~ IFF_CANTCHANGE); 1721 break; 1722 case SIOCGIFFLAGS: 1723 ifr = data; 1724 ifr->ifr_flags = ifp->if_flags; 1725 break; 1726 1727 case SIOCGIFMETRIC: 1728 ifr = data; 1729 ifr->ifr_metric = ifp->if_metric; 1730 break; 1731 1732 case SIOCGIFMTU: 1733 ifr = data; 1734 ifr->ifr_mtu = ifp->if_mtu; 1735 break; 1736 1737 case SIOCGIFDLT: 1738 ifr = data; 1739 ifr->ifr_dlt = ifp->if_dlt; 1740 break; 1741 1742 case SIOCGIFCAP: 1743 ifcr = data; 1744 ifcr->ifcr_capabilities = ifp->if_capabilities; 1745 ifcr->ifcr_capenable = ifp->if_capenable; 1746 break; 1747 1748 case SIOCSIFMETRIC: 1749 ifr = data; 1750 ifp->if_metric = ifr->ifr_metric; 1751 break; 1752 1753 case SIOCGIFDATA: 1754 ifdr = data; 1755 ifdr->ifdr_data = ifp->if_data; 1756 break; 1757 1758 case SIOCGIFINDEX: 1759 ifr = data; 1760 ifr->ifr_index = ifp->if_index; 1761 break; 1762 1763 case SIOCZIFDATA: 1764 ifdr = data; 1765 ifdr->ifdr_data = ifp->if_data; 1766 /* 1767 * Assumes that the volatile counters that can be 1768 * zero'ed are at the end of if_data. 1769 */ 1770 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 1771 offsetof(struct if_data, ifi_ipackets)); 1772 /* 1773 * The memset() clears to the bottm of if_data. In the area, 1774 * if_lastchange is included. Please be careful if new entry 1775 * will be added into if_data or rewite this. 1776 * 1777 * And also, update if_lastchnage. 1778 */ 1779 getnanotime(&ifp->if_lastchange); 1780 break; 1781 case SIOCSIFMTU: 1782 ifr = data; 1783 if (ifp->if_mtu == ifr->ifr_mtu) 1784 break; 1785 ifp->if_mtu = ifr->ifr_mtu; 1786 /* 1787 * If the link MTU changed, do network layer specific procedure. 1788 */ 1789 #ifdef INET6 1790 if (in6_present) 1791 nd6_setmtu(ifp); 1792 #endif 1793 return ENETRESET; 1794 default: 1795 return ENOTTY; 1796 } 1797 return 0; 1798 } 1799 1800 int 1801 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 1802 { 1803 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 1804 struct ifaddr *ifa; 1805 const struct sockaddr *any, *sa; 1806 union { 1807 struct sockaddr sa; 1808 struct sockaddr_storage ss; 1809 } u, v; 1810 1811 switch (cmd) { 1812 case SIOCSIFADDRPREF: 1813 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 1814 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1815 NULL) != 0) 1816 return EPERM; 1817 case SIOCGIFADDRPREF: 1818 break; 1819 default: 1820 return EOPNOTSUPP; 1821 } 1822 1823 /* sanity checks */ 1824 if (data == NULL || ifp == NULL) { 1825 panic("invalid argument to %s", __func__); 1826 /*NOTREACHED*/ 1827 } 1828 1829 /* address must be specified on ADD and DELETE */ 1830 sa = sstocsa(&ifap->ifap_addr); 1831 if (sa->sa_family != sofamily(so)) 1832 return EINVAL; 1833 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 1834 return EINVAL; 1835 1836 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 1837 1838 IFADDR_FOREACH(ifa, ifp) { 1839 if (ifa->ifa_addr->sa_family != sa->sa_family) 1840 continue; 1841 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 1842 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 1843 break; 1844 } 1845 if (ifa == NULL) 1846 return EADDRNOTAVAIL; 1847 1848 switch (cmd) { 1849 case SIOCSIFADDRPREF: 1850 ifa->ifa_preference = ifap->ifap_preference; 1851 return 0; 1852 case SIOCGIFADDRPREF: 1853 /* fill in the if_laddrreq structure */ 1854 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 1855 sizeof(ifap->ifap_addr), ifa->ifa_addr); 1856 ifap->ifap_preference = ifa->ifa_preference; 1857 return 0; 1858 default: 1859 return EOPNOTSUPP; 1860 } 1861 } 1862 1863 static void 1864 ifnet_lock_enter(struct ifnet_lock *il) 1865 { 1866 uint64_t *nenter; 1867 1868 /* Before trying to acquire the mutex, increase the count of threads 1869 * who have entered or who wait to enter the critical section. 1870 * Avoid one costly locked memory transaction by keeping a count for 1871 * each CPU. 1872 */ 1873 nenter = percpu_getref(il->il_nenter); 1874 (*nenter)++; 1875 percpu_putref(il->il_nenter); 1876 mutex_enter(&il->il_lock); 1877 } 1878 1879 static void 1880 ifnet_lock_exit(struct ifnet_lock *il) 1881 { 1882 /* Increase the count of threads who have exited the critical 1883 * section. Increase while we still hold the lock. 1884 */ 1885 il->il_nexit++; 1886 mutex_exit(&il->il_lock); 1887 } 1888 1889 /* 1890 * Interface ioctls. 1891 */ 1892 static int 1893 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 1894 { 1895 struct ifnet *ifp; 1896 struct ifreq *ifr; 1897 int error = 0; 1898 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 1899 u_long ocmd = cmd; 1900 #endif 1901 short oif_flags; 1902 #ifdef COMPAT_OIFREQ 1903 struct ifreq ifrb; 1904 struct oifreq *oifr = NULL; 1905 #endif 1906 int r; 1907 1908 switch (cmd) { 1909 #ifdef COMPAT_OIFREQ 1910 case OSIOCGIFCONF: 1911 case OOSIOCGIFCONF: 1912 return compat_ifconf(cmd, data); 1913 #endif 1914 #ifdef COMPAT_OIFDATA 1915 case OSIOCGIFDATA: 1916 case OSIOCZIFDATA: 1917 return compat_ifdatareq(l, cmd, data); 1918 #endif 1919 case SIOCGIFCONF: 1920 return ifconf(cmd, data); 1921 case SIOCINITIFADDR: 1922 return EPERM; 1923 } 1924 1925 #ifdef COMPAT_OIFREQ 1926 cmd = compat_cvtcmd(cmd); 1927 if (cmd != ocmd) { 1928 oifr = data; 1929 data = ifr = &ifrb; 1930 ifreqo2n(oifr, ifr); 1931 } else 1932 #endif 1933 ifr = data; 1934 1935 ifp = ifunit(ifr->ifr_name); 1936 1937 switch (cmd) { 1938 case SIOCIFCREATE: 1939 case SIOCIFDESTROY: 1940 if (l != NULL) { 1941 error = kauth_authorize_network(l->l_cred, 1942 KAUTH_NETWORK_INTERFACE, 1943 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 1944 (void *)cmd, NULL); 1945 if (error != 0) 1946 return error; 1947 } 1948 mutex_enter(&if_clone_mtx); 1949 r = (cmd == SIOCIFCREATE) ? 1950 if_clone_create(ifr->ifr_name) : 1951 if_clone_destroy(ifr->ifr_name); 1952 mutex_exit(&if_clone_mtx); 1953 return r; 1954 1955 case SIOCIFGCLONERS: 1956 return if_clone_list((struct if_clonereq *)data); 1957 } 1958 1959 if (ifp == NULL) 1960 return ENXIO; 1961 1962 switch (cmd) { 1963 case SIOCALIFADDR: 1964 case SIOCDLIFADDR: 1965 case SIOCSIFADDRPREF: 1966 case SIOCSIFFLAGS: 1967 case SIOCSIFCAP: 1968 case SIOCSIFMETRIC: 1969 case SIOCZIFDATA: 1970 case SIOCSIFMTU: 1971 case SIOCSIFPHYADDR: 1972 case SIOCDIFPHYADDR: 1973 #ifdef INET6 1974 case SIOCSIFPHYADDR_IN6: 1975 #endif 1976 case SIOCSLIFPHYADDR: 1977 case SIOCADDMULTI: 1978 case SIOCDELMULTI: 1979 case SIOCSIFMEDIA: 1980 case SIOCSDRVSPEC: 1981 case SIOCG80211: 1982 case SIOCS80211: 1983 case SIOCS80211NWID: 1984 case SIOCS80211NWKEY: 1985 case SIOCS80211POWER: 1986 case SIOCS80211BSSID: 1987 case SIOCS80211CHANNEL: 1988 case SIOCSLINKSTR: 1989 if (l != NULL) { 1990 error = kauth_authorize_network(l->l_cred, 1991 KAUTH_NETWORK_INTERFACE, 1992 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 1993 (void *)cmd, NULL); 1994 if (error != 0) 1995 return error; 1996 } 1997 } 1998 1999 oif_flags = ifp->if_flags; 2000 2001 ifnet_lock_enter(ifp->if_ioctl_lock); 2002 error = (*ifp->if_ioctl)(ifp, cmd, data); 2003 if (error != ENOTTY) 2004 ; 2005 else if (so->so_proto == NULL) 2006 error = EOPNOTSUPP; 2007 else { 2008 #ifdef COMPAT_OSOCK 2009 error = compat_ifioctl(so, ocmd, cmd, data, l); 2010 #else 2011 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2012 cmd, data, ifp); 2013 #endif 2014 } 2015 2016 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2017 #ifdef INET6 2018 if (in6_present && (ifp->if_flags & IFF_UP) != 0) { 2019 int s = splnet(); 2020 in6_if_up(ifp); 2021 splx(s); 2022 } 2023 #endif 2024 } 2025 #ifdef COMPAT_OIFREQ 2026 if (cmd != ocmd) 2027 ifreqn2o(oifr, ifr); 2028 #endif 2029 2030 ifnet_lock_exit(ifp->if_ioctl_lock); 2031 return error; 2032 } 2033 2034 /* This callback adds to the sum in `arg' the number of 2035 * threads on `ci' who have entered or who wait to enter the 2036 * critical section. 2037 */ 2038 static void 2039 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci) 2040 { 2041 uint64_t *sum = arg, *nenter = p; 2042 2043 *sum += *nenter; 2044 } 2045 2046 /* Return the number of threads who have entered or who wait 2047 * to enter the critical section on all CPUs. 2048 */ 2049 static uint64_t 2050 ifnet_lock_entrances(struct ifnet_lock *il) 2051 { 2052 uint64_t sum = 0; 2053 2054 percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum); 2055 2056 return sum; 2057 } 2058 2059 static int 2060 ifioctl_attach(struct ifnet *ifp) 2061 { 2062 struct ifnet_lock *il; 2063 2064 /* If the driver has not supplied its own if_ioctl, then 2065 * supply the default. 2066 */ 2067 if (ifp->if_ioctl == NULL) 2068 ifp->if_ioctl = ifioctl_common; 2069 2070 /* Create an ifnet_lock for synchronizing ifioctls. */ 2071 if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL) 2072 return ENOMEM; 2073 2074 il->il_nenter = percpu_alloc(sizeof(uint64_t)); 2075 if (il->il_nenter == NULL) { 2076 kmem_free(il, sizeof(*il)); 2077 return ENOMEM; 2078 } 2079 2080 mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE); 2081 cv_init(&il->il_emptied, ifp->if_xname); 2082 2083 ifp->if_ioctl_lock = il; 2084 2085 return 0; 2086 } 2087 2088 /* 2089 * This must not be called until after `ifp' has been withdrawn from the 2090 * ifnet tables so that ifioctl() cannot get a handle on it by calling 2091 * ifunit(). 2092 */ 2093 static void 2094 ifioctl_detach(struct ifnet *ifp) 2095 { 2096 struct ifnet_lock *il; 2097 2098 il = ifp->if_ioctl_lock; 2099 mutex_enter(&il->il_lock); 2100 /* Install if_nullioctl to make sure that any thread that 2101 * subsequently enters the critical section will quit it 2102 * immediately and signal the condition variable that we 2103 * wait on, below. 2104 */ 2105 ifp->if_ioctl = if_nullioctl; 2106 /* Sleep while threads are still in the critical section or 2107 * wait to enter it. 2108 */ 2109 while (ifnet_lock_entrances(il) != il->il_nexit) 2110 cv_wait(&il->il_emptied, &il->il_lock); 2111 /* At this point, we are the only thread still in the critical 2112 * section, and no new thread can get a handle on the ifioctl 2113 * lock, so it is safe to free its memory. 2114 */ 2115 mutex_exit(&il->il_lock); 2116 ifp->if_ioctl_lock = NULL; 2117 percpu_free(il->il_nenter, sizeof(uint64_t)); 2118 il->il_nenter = NULL; 2119 cv_destroy(&il->il_emptied); 2120 mutex_destroy(&il->il_lock); 2121 kmem_free(il, sizeof(*il)); 2122 } 2123 2124 /* 2125 * Return interface configuration 2126 * of system. List may be used 2127 * in later ioctl's (above) to get 2128 * other information. 2129 * 2130 * Each record is a struct ifreq. Before the addition of 2131 * sockaddr_storage, the API rule was that sockaddr flavors that did 2132 * not fit would extend beyond the struct ifreq, with the next struct 2133 * ifreq starting sa_len beyond the struct sockaddr. Because the 2134 * union in struct ifreq includes struct sockaddr_storage, every kind 2135 * of sockaddr must fit. Thus, there are no longer any overlength 2136 * records. 2137 * 2138 * Records are added to the user buffer if they fit, and ifc_len is 2139 * adjusted to the length that was written. Thus, the user is only 2140 * assured of getting the complete list if ifc_len on return is at 2141 * least sizeof(struct ifreq) less than it was on entry. 2142 * 2143 * If the user buffer pointer is NULL, this routine copies no data and 2144 * returns the amount of space that would be needed. 2145 * 2146 * Invariants: 2147 * ifrp points to the next part of the user's buffer to be used. If 2148 * ifrp != NULL, space holds the number of bytes remaining that we may 2149 * write at ifrp. Otherwise, space holds the number of bytes that 2150 * would have been written had there been adequate space. 2151 */ 2152 /*ARGSUSED*/ 2153 static int 2154 ifconf(u_long cmd, void *data) 2155 { 2156 struct ifconf *ifc = (struct ifconf *)data; 2157 struct ifnet *ifp; 2158 struct ifaddr *ifa; 2159 struct ifreq ifr, *ifrp = NULL; 2160 int space = 0, error = 0; 2161 const int sz = (int)sizeof(struct ifreq); 2162 const bool docopy = ifc->ifc_req != NULL; 2163 2164 if (docopy) { 2165 space = ifc->ifc_len; 2166 ifrp = ifc->ifc_req; 2167 } 2168 2169 IFNET_FOREACH(ifp) { 2170 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2171 sizeof(ifr.ifr_name)); 2172 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') 2173 return ENAMETOOLONG; 2174 if (IFADDR_EMPTY(ifp)) { 2175 /* Interface with no addresses - send zero sockaddr. */ 2176 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2177 if (!docopy) { 2178 space += sz; 2179 continue; 2180 } 2181 if (space >= sz) { 2182 error = copyout(&ifr, ifrp, sz); 2183 if (error != 0) 2184 return error; 2185 ifrp++; 2186 space -= sz; 2187 } 2188 } 2189 2190 IFADDR_FOREACH(ifa, ifp) { 2191 struct sockaddr *sa = ifa->ifa_addr; 2192 /* all sockaddrs must fit in sockaddr_storage */ 2193 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2194 2195 if (!docopy) { 2196 space += sz; 2197 continue; 2198 } 2199 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2200 if (space >= sz) { 2201 error = copyout(&ifr, ifrp, sz); 2202 if (error != 0) 2203 return (error); 2204 ifrp++; space -= sz; 2205 } 2206 } 2207 } 2208 if (docopy) { 2209 KASSERT(0 <= space && space <= ifc->ifc_len); 2210 ifc->ifc_len -= space; 2211 } else { 2212 KASSERT(space >= 0); 2213 ifc->ifc_len = space; 2214 } 2215 return (0); 2216 } 2217 2218 int 2219 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2220 { 2221 uint8_t len; 2222 #ifdef COMPAT_OIFREQ 2223 struct ifreq ifrb; 2224 struct oifreq *oifr = NULL; 2225 u_long ocmd = cmd; 2226 cmd = compat_cvtcmd(cmd); 2227 if (cmd != ocmd) { 2228 oifr = (struct oifreq *)(void *)ifr; 2229 ifr = &ifrb; 2230 ifreqo2n(oifr, ifr); 2231 len = sizeof(oifr->ifr_addr); 2232 } else 2233 #endif 2234 len = sizeof(ifr->ifr_ifru.ifru_space); 2235 2236 if (len < sa->sa_len) 2237 return EFBIG; 2238 2239 memset(&ifr->ifr_addr, 0, len); 2240 sockaddr_copy(&ifr->ifr_addr, len, sa); 2241 2242 #ifdef COMPAT_OIFREQ 2243 if (cmd != ocmd) 2244 ifreqn2o(oifr, ifr); 2245 #endif 2246 return 0; 2247 } 2248 2249 /* 2250 * Queue message on interface, and start output if interface 2251 * not yet active. 2252 */ 2253 int 2254 ifq_enqueue(struct ifnet *ifp, struct mbuf *m 2255 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2256 { 2257 int len = m->m_pkthdr.len; 2258 int mflags = m->m_flags; 2259 int s = splnet(); 2260 int error; 2261 2262 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2263 if (error != 0) 2264 goto out; 2265 ifp->if_obytes += len; 2266 if (mflags & M_MCAST) 2267 ifp->if_omcasts++; 2268 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2269 (*ifp->if_start)(ifp); 2270 out: 2271 splx(s); 2272 return error; 2273 } 2274 2275 /* 2276 * Queue message on interface, possibly using a second fast queue 2277 */ 2278 int 2279 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m 2280 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2281 { 2282 int error = 0; 2283 2284 if (ifq != NULL 2285 #ifdef ALTQ 2286 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2287 #endif 2288 ) { 2289 if (IF_QFULL(ifq)) { 2290 IF_DROP(&ifp->if_snd); 2291 m_freem(m); 2292 if (error == 0) 2293 error = ENOBUFS; 2294 } else 2295 IF_ENQUEUE(ifq, m); 2296 } else 2297 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2298 if (error != 0) { 2299 ++ifp->if_oerrors; 2300 return error; 2301 } 2302 return 0; 2303 } 2304 2305 int 2306 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2307 { 2308 int rc; 2309 2310 if (ifp->if_initaddr != NULL) 2311 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2312 else if (src || 2313 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2314 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2315 2316 return rc; 2317 } 2318 2319 int 2320 if_do_dad(struct ifnet *ifp) 2321 { 2322 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2323 return 0; 2324 2325 switch (ifp->if_type) { 2326 case IFT_FAITH: 2327 /* 2328 * These interfaces do not have the IFF_LOOPBACK flag, 2329 * but loop packets back. We do not have to do DAD on such 2330 * interfaces. We should even omit it, because loop-backed 2331 * responses would confuse the DAD procedure. 2332 */ 2333 return 0; 2334 default: 2335 /* 2336 * Our DAD routine requires the interface up and running. 2337 * However, some interfaces can be up before the RUNNING 2338 * status. Additionaly, users may try to assign addresses 2339 * before the interface becomes up (or running). 2340 * We simply skip DAD in such a case as a work around. 2341 * XXX: we should rather mark "tentative" on such addresses, 2342 * and do DAD after the interface becomes ready. 2343 */ 2344 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 2345 (IFF_UP|IFF_RUNNING)) 2346 return 0; 2347 2348 return 1; 2349 } 2350 } 2351 2352 int 2353 if_flags_set(ifnet_t *ifp, const short flags) 2354 { 2355 int rc; 2356 2357 if (ifp->if_setflags != NULL) 2358 rc = (*ifp->if_setflags)(ifp, flags); 2359 else { 2360 short cantflags, chgdflags; 2361 struct ifreq ifr; 2362 2363 chgdflags = ifp->if_flags ^ flags; 2364 cantflags = chgdflags & IFF_CANTCHANGE; 2365 2366 if (cantflags != 0) 2367 ifp->if_flags ^= cantflags; 2368 2369 /* Traditionally, we do not call if_ioctl after 2370 * setting/clearing only IFF_PROMISC if the interface 2371 * isn't IFF_UP. Uphold that tradition. 2372 */ 2373 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2374 return 0; 2375 2376 memset(&ifr, 0, sizeof(ifr)); 2377 2378 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2379 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2380 2381 if (rc != 0 && cantflags != 0) 2382 ifp->if_flags ^= cantflags; 2383 } 2384 2385 return rc; 2386 } 2387 2388 int 2389 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2390 { 2391 int rc; 2392 struct ifreq ifr; 2393 2394 if (ifp->if_mcastop != NULL) 2395 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2396 else { 2397 ifreq_setaddr(cmd, &ifr, sa); 2398 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2399 } 2400 2401 return rc; 2402 } 2403 2404 static void 2405 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2406 struct ifaltq *ifq) 2407 { 2408 const struct sysctlnode *cnode, *rnode; 2409 2410 if (sysctl_createv(clog, 0, NULL, &rnode, 2411 CTLFLAG_PERMANENT, 2412 CTLTYPE_NODE, "interfaces", 2413 SYSCTL_DESCR("Per-interface controls"), 2414 NULL, 0, NULL, 0, 2415 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 2416 goto bad; 2417 2418 if (sysctl_createv(clog, 0, &rnode, &rnode, 2419 CTLFLAG_PERMANENT, 2420 CTLTYPE_NODE, ifname, 2421 SYSCTL_DESCR("Interface controls"), 2422 NULL, 0, NULL, 0, 2423 CTL_CREATE, CTL_EOL) != 0) 2424 goto bad; 2425 2426 if (sysctl_createv(clog, 0, &rnode, &rnode, 2427 CTLFLAG_PERMANENT, 2428 CTLTYPE_NODE, "sndq", 2429 SYSCTL_DESCR("Interface output queue controls"), 2430 NULL, 0, NULL, 0, 2431 CTL_CREATE, CTL_EOL) != 0) 2432 goto bad; 2433 2434 if (sysctl_createv(clog, 0, &rnode, &cnode, 2435 CTLFLAG_PERMANENT, 2436 CTLTYPE_INT, "len", 2437 SYSCTL_DESCR("Current output queue length"), 2438 NULL, 0, &ifq->ifq_len, 0, 2439 CTL_CREATE, CTL_EOL) != 0) 2440 goto bad; 2441 2442 if (sysctl_createv(clog, 0, &rnode, &cnode, 2443 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2444 CTLTYPE_INT, "maxlen", 2445 SYSCTL_DESCR("Maximum allowed output queue length"), 2446 NULL, 0, &ifq->ifq_maxlen, 0, 2447 CTL_CREATE, CTL_EOL) != 0) 2448 goto bad; 2449 2450 if (sysctl_createv(clog, 0, &rnode, &cnode, 2451 CTLFLAG_PERMANENT, 2452 CTLTYPE_INT, "drops", 2453 SYSCTL_DESCR("Packets dropped due to full output queue"), 2454 NULL, 0, &ifq->ifq_drops, 0, 2455 CTL_CREATE, CTL_EOL) != 0) 2456 goto bad; 2457 2458 return; 2459 bad: 2460 printf("%s: could not attach sysctl nodes\n", ifname); 2461 return; 2462 } 2463 2464 #if defined(INET) || defined(INET6) 2465 2466 #define SYSCTL_NET_PKTQ(q, cn, c) \ 2467 static int \ 2468 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 2469 { \ 2470 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 2471 } 2472 2473 #if defined(INET) 2474 static int 2475 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 2476 { 2477 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 2478 } 2479 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 2480 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 2481 #endif 2482 2483 #if defined(INET6) 2484 static int 2485 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 2486 { 2487 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 2488 } 2489 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 2490 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 2491 #endif 2492 2493 static void 2494 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 2495 { 2496 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 2497 const char *pfname = NULL, *ipname = NULL; 2498 int ipn = 0, qid = 0; 2499 2500 switch (pf) { 2501 #if defined(INET) 2502 case PF_INET: 2503 len_func = sysctl_net_ip_pktq_items; 2504 maxlen_func = sysctl_net_ip_pktq_maxlen; 2505 drops_func = sysctl_net_ip_pktq_drops; 2506 pfname = "inet", ipn = IPPROTO_IP; 2507 ipname = "ip", qid = IPCTL_IFQ; 2508 break; 2509 #endif 2510 #if defined(INET6) 2511 case PF_INET6: 2512 len_func = sysctl_net_ip6_pktq_items; 2513 maxlen_func = sysctl_net_ip6_pktq_maxlen; 2514 drops_func = sysctl_net_ip6_pktq_drops; 2515 pfname = "inet6", ipn = IPPROTO_IPV6; 2516 ipname = "ip6", qid = IPV6CTL_IFQ; 2517 break; 2518 #endif 2519 default: 2520 KASSERT(false); 2521 } 2522 2523 sysctl_createv(clog, 0, NULL, NULL, 2524 CTLFLAG_PERMANENT, 2525 CTLTYPE_NODE, pfname, NULL, 2526 NULL, 0, NULL, 0, 2527 CTL_NET, pf, CTL_EOL); 2528 sysctl_createv(clog, 0, NULL, NULL, 2529 CTLFLAG_PERMANENT, 2530 CTLTYPE_NODE, ipname, NULL, 2531 NULL, 0, NULL, 0, 2532 CTL_NET, pf, ipn, CTL_EOL); 2533 sysctl_createv(clog, 0, NULL, NULL, 2534 CTLFLAG_PERMANENT, 2535 CTLTYPE_NODE, "ifq", 2536 SYSCTL_DESCR("Protocol input queue controls"), 2537 NULL, 0, NULL, 0, 2538 CTL_NET, pf, ipn, qid, CTL_EOL); 2539 2540 sysctl_createv(clog, 0, NULL, NULL, 2541 CTLFLAG_PERMANENT, 2542 CTLTYPE_INT, "len", 2543 SYSCTL_DESCR("Current input queue length"), 2544 len_func, 0, NULL, 0, 2545 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 2546 sysctl_createv(clog, 0, NULL, NULL, 2547 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2548 CTLTYPE_INT, "maxlen", 2549 SYSCTL_DESCR("Maximum allowed input queue length"), 2550 maxlen_func, 0, NULL, 0, 2551 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 2552 sysctl_createv(clog, 0, NULL, NULL, 2553 CTLFLAG_PERMANENT, 2554 CTLTYPE_INT, "drops", 2555 SYSCTL_DESCR("Packets dropped due to full input queue"), 2556 drops_func, 0, NULL, 0, 2557 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 2558 } 2559 #endif /* INET || INET6 */ 2560 2561 static int 2562 if_sdl_sysctl(SYSCTLFN_ARGS) 2563 { 2564 struct ifnet *ifp; 2565 const struct sockaddr_dl *sdl; 2566 2567 if (namelen != 1) 2568 return EINVAL; 2569 2570 ifp = if_byindex(name[0]); 2571 if (ifp == NULL) 2572 return ENODEV; 2573 2574 sdl = ifp->if_sadl; 2575 if (sdl == NULL) { 2576 *oldlenp = 0; 2577 return 0; 2578 } 2579 2580 if (oldp == NULL) { 2581 *oldlenp = sdl->sdl_alen; 2582 return 0; 2583 } 2584 2585 if (*oldlenp >= sdl->sdl_alen) 2586 *oldlenp = sdl->sdl_alen; 2587 return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 2588 } 2589 2590 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 2591 { 2592 const struct sysctlnode *rnode = NULL; 2593 2594 sysctl_createv(clog, 0, NULL, &rnode, 2595 CTLFLAG_PERMANENT, 2596 CTLTYPE_NODE, "sdl", 2597 SYSCTL_DESCR("Get active link-layer address"), 2598 if_sdl_sysctl, 0, NULL, 0, 2599 CTL_NET, CTL_CREATE, CTL_EOL); 2600 } 2601