if.c revision 1.317
1 /* $NetBSD: if.c,v 1.317 2015/07/17 02:21:08 ozaki-r 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 1.317 2015/07/17 02:21:08 ozaki-r 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 struct ifnet *lo0ifp; 173 int ifqmaxlen = IFQ_MAXLEN; 174 175 static int if_rt_walktree(struct rtentry *, void *); 176 177 static struct if_clone *if_clone_lookup(const char *, int *); 178 179 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 180 static int if_cloners_count; 181 182 /* Packet filtering hook for interfaces. */ 183 pfil_head_t * if_pfil; 184 185 static kauth_listener_t if_listener; 186 187 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 188 static int ifioctl_attach(struct ifnet *); 189 static void ifioctl_detach(struct ifnet *); 190 static void ifnet_lock_enter(struct ifnet_lock *); 191 static void ifnet_lock_exit(struct ifnet_lock *); 192 static void if_detach_queues(struct ifnet *, struct ifqueue *); 193 static void sysctl_sndq_setup(struct sysctllog **, const char *, 194 struct ifaltq *); 195 static void if_slowtimo(void *); 196 static void if_free_sadl(struct ifnet *); 197 static void if_attachdomain1(struct ifnet *); 198 static int ifconf(u_long, void *); 199 static int if_clone_create(const char *); 200 static int if_clone_destroy(const char *); 201 202 #if defined(INET) || defined(INET6) 203 static void sysctl_net_pktq_setup(struct sysctllog **, int); 204 #endif 205 206 static int 207 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 208 void *arg0, void *arg1, void *arg2, void *arg3) 209 { 210 int result; 211 enum kauth_network_req req; 212 213 result = KAUTH_RESULT_DEFER; 214 req = (enum kauth_network_req)arg1; 215 216 if (action != KAUTH_NETWORK_INTERFACE) 217 return result; 218 219 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 220 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 221 result = KAUTH_RESULT_ALLOW; 222 223 return result; 224 } 225 226 /* 227 * Network interface utility routines. 228 * 229 * Routines with ifa_ifwith* names take sockaddr *'s as 230 * parameters. 231 */ 232 void 233 ifinit(void) 234 { 235 #if defined(INET) 236 sysctl_net_pktq_setup(NULL, PF_INET); 237 #endif 238 #ifdef INET6 239 if (in6_present) 240 sysctl_net_pktq_setup(NULL, PF_INET6); 241 #endif 242 243 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 244 if_listener_cb, NULL); 245 246 /* interfaces are available, inform socket code */ 247 ifioctl = doifioctl; 248 } 249 250 /* 251 * XXX Initialization before configure(). 252 * XXX hack to get pfil_add_hook working in autoconf. 253 */ 254 void 255 ifinit1(void) 256 { 257 mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE); 258 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 259 TAILQ_INIT(&ifnet_list); 260 if_indexlim = 8; 261 262 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 263 KASSERT(if_pfil != NULL); 264 265 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 266 etherinit(); 267 #endif 268 } 269 270 ifnet_t * 271 if_alloc(u_char type) 272 { 273 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 274 } 275 276 void 277 if_free(ifnet_t *ifp) 278 { 279 kmem_free(ifp, sizeof(ifnet_t)); 280 } 281 282 void 283 if_initname(struct ifnet *ifp, const char *name, int unit) 284 { 285 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 286 "%s%d", name, unit); 287 } 288 289 /* 290 * Null routines used while an interface is going away. These routines 291 * just return an error. 292 */ 293 294 int 295 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 296 const struct sockaddr *so, struct rtentry *rt) 297 { 298 299 return ENXIO; 300 } 301 302 void 303 if_nullinput(struct ifnet *ifp, struct mbuf *m) 304 { 305 306 /* Nothing. */ 307 } 308 309 void 310 if_nullstart(struct ifnet *ifp) 311 { 312 313 /* Nothing. */ 314 } 315 316 int 317 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 318 { 319 320 /* Wake ifioctl_detach(), who may wait for all threads to 321 * quit the critical section. 322 */ 323 cv_signal(&ifp->if_ioctl_lock->il_emptied); 324 return ENXIO; 325 } 326 327 int 328 if_nullinit(struct ifnet *ifp) 329 { 330 331 return ENXIO; 332 } 333 334 void 335 if_nullstop(struct ifnet *ifp, int disable) 336 { 337 338 /* Nothing. */ 339 } 340 341 void 342 if_nullslowtimo(struct ifnet *ifp) 343 { 344 345 /* Nothing. */ 346 } 347 348 void 349 if_nulldrain(struct ifnet *ifp) 350 { 351 352 /* Nothing. */ 353 } 354 355 void 356 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 357 { 358 struct ifaddr *ifa; 359 struct sockaddr_dl *sdl; 360 361 ifp->if_addrlen = addrlen; 362 if_alloc_sadl(ifp); 363 ifa = ifp->if_dl; 364 sdl = satosdl(ifa->ifa_addr); 365 366 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 367 if (factory) { 368 ifp->if_hwdl = ifp->if_dl; 369 ifaref(ifp->if_hwdl); 370 } 371 /* TBD routing socket */ 372 } 373 374 struct ifaddr * 375 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 376 { 377 unsigned socksize, ifasize; 378 int addrlen, namelen; 379 struct sockaddr_dl *mask, *sdl; 380 struct ifaddr *ifa; 381 382 namelen = strlen(ifp->if_xname); 383 addrlen = ifp->if_addrlen; 384 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 385 ifasize = sizeof(*ifa) + 2 * socksize; 386 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 387 388 sdl = (struct sockaddr_dl *)(ifa + 1); 389 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 390 391 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 392 ifp->if_xname, namelen, NULL, addrlen); 393 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 394 memset(&mask->sdl_data[0], 0xff, namelen); 395 ifa->ifa_rtrequest = link_rtrequest; 396 ifa->ifa_addr = (struct sockaddr *)sdl; 397 ifa->ifa_netmask = (struct sockaddr *)mask; 398 399 *sdlp = sdl; 400 401 return ifa; 402 } 403 404 static void 405 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 406 { 407 const struct sockaddr_dl *sdl; 408 409 ifp->if_dl = ifa; 410 ifaref(ifa); 411 sdl = satosdl(ifa->ifa_addr); 412 ifp->if_sadl = sdl; 413 } 414 415 /* 416 * Allocate the link level name for the specified interface. This 417 * is an attachment helper. It must be called after ifp->if_addrlen 418 * is initialized, which may not be the case when if_attach() is 419 * called. 420 */ 421 void 422 if_alloc_sadl(struct ifnet *ifp) 423 { 424 struct ifaddr *ifa; 425 const struct sockaddr_dl *sdl; 426 427 /* 428 * If the interface already has a link name, release it 429 * now. This is useful for interfaces that can change 430 * link types, and thus switch link names often. 431 */ 432 if (ifp->if_sadl != NULL) 433 if_free_sadl(ifp); 434 435 ifa = if_dl_create(ifp, &sdl); 436 437 ifa_insert(ifp, ifa); 438 if_sadl_setrefs(ifp, ifa); 439 } 440 441 static void 442 if_deactivate_sadl(struct ifnet *ifp) 443 { 444 struct ifaddr *ifa; 445 446 KASSERT(ifp->if_dl != NULL); 447 448 ifa = ifp->if_dl; 449 450 ifp->if_sadl = NULL; 451 452 ifp->if_dl = NULL; 453 ifafree(ifa); 454 } 455 456 void 457 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 458 const struct sockaddr_dl *sdl) 459 { 460 int s; 461 462 s = splnet(); 463 464 if_deactivate_sadl(ifp); 465 466 if_sadl_setrefs(ifp, ifa); 467 IFADDR_FOREACH(ifa, ifp) 468 rtinit(ifa, RTM_LLINFO_UPD, 0); 469 splx(s); 470 } 471 472 /* 473 * Free the link level name for the specified interface. This is 474 * a detach helper. This is called from if_detach(). 475 */ 476 static void 477 if_free_sadl(struct ifnet *ifp) 478 { 479 struct ifaddr *ifa; 480 int s; 481 482 ifa = ifp->if_dl; 483 if (ifa == NULL) { 484 KASSERT(ifp->if_sadl == NULL); 485 return; 486 } 487 488 KASSERT(ifp->if_sadl != NULL); 489 490 s = splnet(); 491 rtinit(ifa, RTM_DELETE, 0); 492 ifa_remove(ifp, ifa); 493 if_deactivate_sadl(ifp); 494 if (ifp->if_hwdl == ifa) { 495 ifafree(ifa); 496 ifp->if_hwdl = NULL; 497 } 498 splx(s); 499 } 500 501 static void 502 if_getindex(ifnet_t *ifp) 503 { 504 bool hitlimit = false; 505 506 mutex_enter(&index_gen_mtx); 507 ifp->if_index_gen = index_gen++; 508 mutex_exit(&index_gen_mtx); 509 510 ifp->if_index = if_index; 511 if (ifindex2ifnet == NULL) { 512 if_index++; 513 goto skip; 514 } 515 while (if_byindex(ifp->if_index)) { 516 /* 517 * If we hit USHRT_MAX, we skip back to 0 since 518 * there are a number of places where the value 519 * of if_index or if_index itself is compared 520 * to or stored in an unsigned short. By 521 * jumping back, we won't botch those assignments 522 * or comparisons. 523 */ 524 if (++if_index == 0) { 525 if_index = 1; 526 } else if (if_index == USHRT_MAX) { 527 /* 528 * However, if we have to jump back to 529 * zero *twice* without finding an empty 530 * slot in ifindex2ifnet[], then there 531 * there are too many (>65535) interfaces. 532 */ 533 if (hitlimit) { 534 panic("too many interfaces"); 535 } 536 hitlimit = true; 537 if_index = 1; 538 } 539 ifp->if_index = if_index; 540 } 541 skip: 542 /* 543 * ifindex2ifnet is indexed by if_index. Since if_index will 544 * grow dynamically, it should grow too. 545 */ 546 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 547 size_t m, n, oldlim; 548 void *q; 549 550 oldlim = if_indexlim; 551 while (ifp->if_index >= if_indexlim) 552 if_indexlim <<= 1; 553 554 /* grow ifindex2ifnet */ 555 m = oldlim * sizeof(struct ifnet *); 556 n = if_indexlim * sizeof(struct ifnet *); 557 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 558 if (ifindex2ifnet != NULL) { 559 memcpy(q, ifindex2ifnet, m); 560 free(ifindex2ifnet, M_IFADDR); 561 } 562 ifindex2ifnet = (struct ifnet **)q; 563 } 564 ifindex2ifnet[ifp->if_index] = ifp; 565 } 566 567 /* 568 * Initialize an interface and assign an index for it. 569 * 570 * It must be called prior to a device specific attach routine 571 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 572 * and be followed by if_register: 573 * 574 * if_initialize(ifp); 575 * ether_ifattach(ifp, enaddr); 576 * if_register(ifp); 577 */ 578 void 579 if_initialize(ifnet_t *ifp) 580 { 581 KASSERT(if_indexlim > 0); 582 TAILQ_INIT(&ifp->if_addrlist); 583 584 /* 585 * Link level name is allocated later by a separate call to 586 * if_alloc_sadl(). 587 */ 588 589 if (ifp->if_snd.ifq_maxlen == 0) 590 ifp->if_snd.ifq_maxlen = ifqmaxlen; 591 592 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 593 594 ifp->if_link_state = LINK_STATE_UNKNOWN; 595 596 ifp->if_capenable = 0; 597 ifp->if_csum_flags_tx = 0; 598 ifp->if_csum_flags_rx = 0; 599 600 #ifdef ALTQ 601 ifp->if_snd.altq_type = 0; 602 ifp->if_snd.altq_disc = NULL; 603 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 604 ifp->if_snd.altq_tbr = NULL; 605 ifp->if_snd.altq_ifp = ifp; 606 #endif 607 608 #ifdef NET_MPSAFE 609 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 610 #else 611 ifp->if_snd.ifq_lock = NULL; 612 #endif 613 614 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 615 (void)pfil_run_hooks(if_pfil, 616 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 617 618 if_getindex(ifp); 619 } 620 621 /* 622 * Register an interface to the list of "active" interfaces. 623 */ 624 void 625 if_register(ifnet_t *ifp) 626 { 627 if (ifioctl_attach(ifp) != 0) 628 panic("%s: ifioctl_attach() failed", __func__); 629 630 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 631 632 if (!STAILQ_EMPTY(&domains)) 633 if_attachdomain1(ifp); 634 635 /* Announce the interface. */ 636 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 637 638 if (ifp->if_slowtimo != NULL) { 639 ifp->if_slowtimo_ch = 640 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 641 callout_init(ifp->if_slowtimo_ch, 0); 642 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 643 if_slowtimo(ifp); 644 } 645 646 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 647 } 648 649 /* 650 * Deprecated. Use if_initialize and if_register instead. 651 * See the above comment of if_initialize. 652 */ 653 void 654 if_attach(ifnet_t *ifp) 655 { 656 if_initialize(ifp); 657 if_register(ifp); 658 } 659 660 void 661 if_attachdomain(void) 662 { 663 struct ifnet *ifp; 664 int s; 665 666 s = splnet(); 667 IFNET_FOREACH(ifp) 668 if_attachdomain1(ifp); 669 splx(s); 670 } 671 672 static void 673 if_attachdomain1(struct ifnet *ifp) 674 { 675 struct domain *dp; 676 int s; 677 678 s = splnet(); 679 680 /* address family dependent data region */ 681 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 682 DOMAIN_FOREACH(dp) { 683 if (dp->dom_ifattach != NULL) 684 ifp->if_afdata[dp->dom_family] = 685 (*dp->dom_ifattach)(ifp); 686 } 687 688 splx(s); 689 } 690 691 /* 692 * Deactivate an interface. This points all of the procedure 693 * handles at error stubs. May be called from interrupt context. 694 */ 695 void 696 if_deactivate(struct ifnet *ifp) 697 { 698 int s; 699 700 s = splnet(); 701 702 ifp->if_output = if_nulloutput; 703 ifp->if_input = if_nullinput; 704 ifp->if_start = if_nullstart; 705 ifp->if_ioctl = if_nullioctl; 706 ifp->if_init = if_nullinit; 707 ifp->if_stop = if_nullstop; 708 ifp->if_slowtimo = if_nullslowtimo; 709 ifp->if_drain = if_nulldrain; 710 711 /* No more packets may be enqueued. */ 712 ifp->if_snd.ifq_maxlen = 0; 713 714 splx(s); 715 } 716 717 void 718 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 719 { 720 struct ifaddr *ifa, *nifa; 721 722 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 723 if (ifa->ifa_addr->sa_family != family) 724 continue; 725 (*purgeaddr)(ifa); 726 } 727 } 728 729 /* 730 * Detach an interface from the list of "active" interfaces, 731 * freeing any resources as we go along. 732 * 733 * NOTE: This routine must be called with a valid thread context, 734 * as it may block. 735 */ 736 void 737 if_detach(struct ifnet *ifp) 738 { 739 struct socket so; 740 struct ifaddr *ifa; 741 #ifdef IFAREF_DEBUG 742 struct ifaddr *last_ifa = NULL; 743 #endif 744 struct domain *dp; 745 const struct protosw *pr; 746 int s, i, family, purged; 747 uint64_t xc; 748 749 /* 750 * XXX It's kind of lame that we have to have the 751 * XXX socket structure... 752 */ 753 memset(&so, 0, sizeof(so)); 754 755 s = splnet(); 756 757 if (ifp->if_slowtimo != NULL) { 758 ifp->if_slowtimo = NULL; 759 callout_halt(ifp->if_slowtimo_ch, NULL); 760 callout_destroy(ifp->if_slowtimo_ch); 761 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 762 } 763 764 /* 765 * Do an if_down() to give protocols a chance to do something. 766 */ 767 if_down(ifp); 768 769 #ifdef ALTQ 770 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 771 altq_disable(&ifp->if_snd); 772 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 773 altq_detach(&ifp->if_snd); 774 #endif 775 776 if (ifp->if_snd.ifq_lock) 777 mutex_obj_free(ifp->if_snd.ifq_lock); 778 779 sysctl_teardown(&ifp->if_sysctl_log); 780 781 #if NCARP > 0 782 /* Remove the interface from any carp group it is a part of. */ 783 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 784 carp_ifdetach(ifp); 785 #endif 786 787 /* 788 * Rip all the addresses off the interface. This should make 789 * all of the routes go away. 790 * 791 * pr_usrreq calls can remove an arbitrary number of ifaddrs 792 * from the list, including our "cursor", ifa. For safety, 793 * and to honor the TAILQ abstraction, I just restart the 794 * loop after each removal. Note that the loop will exit 795 * when all of the remaining ifaddrs belong to the AF_LINK 796 * family. I am counting on the historical fact that at 797 * least one pr_usrreq in each address domain removes at 798 * least one ifaddr. 799 */ 800 again: 801 IFADDR_FOREACH(ifa, ifp) { 802 family = ifa->ifa_addr->sa_family; 803 #ifdef IFAREF_DEBUG 804 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 805 ifa, family, ifa->ifa_refcnt); 806 if (last_ifa != NULL && ifa == last_ifa) 807 panic("if_detach: loop detected"); 808 last_ifa = ifa; 809 #endif 810 if (family == AF_LINK) 811 continue; 812 dp = pffinddomain(family); 813 #ifdef DIAGNOSTIC 814 if (dp == NULL) 815 panic("if_detach: no domain for AF %d", 816 family); 817 #endif 818 /* 819 * XXX These PURGEIF calls are redundant with the 820 * purge-all-families calls below, but are left in for 821 * now both to make a smaller change, and to avoid 822 * unplanned interactions with clearing of 823 * ifp->if_addrlist. 824 */ 825 purged = 0; 826 for (pr = dp->dom_protosw; 827 pr < dp->dom_protoswNPROTOSW; pr++) { 828 so.so_proto = pr; 829 if (pr->pr_usrreqs) { 830 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 831 purged = 1; 832 } 833 } 834 if (purged == 0) { 835 /* 836 * XXX What's really the best thing to do 837 * XXX here? --thorpej (at) NetBSD.org 838 */ 839 printf("if_detach: WARNING: AF %d not purged\n", 840 family); 841 ifa_remove(ifp, ifa); 842 } 843 goto again; 844 } 845 846 if_free_sadl(ifp); 847 848 /* Walk the routing table looking for stragglers. */ 849 for (i = 0; i <= AF_MAX; i++) { 850 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 851 continue; 852 } 853 854 DOMAIN_FOREACH(dp) { 855 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 856 { 857 void *p = ifp->if_afdata[dp->dom_family]; 858 if (p) { 859 ifp->if_afdata[dp->dom_family] = NULL; 860 (*dp->dom_ifdetach)(ifp, p); 861 } 862 } 863 864 /* 865 * One would expect multicast memberships (INET and 866 * INET6) on UDP sockets to be purged by the PURGEIF 867 * calls above, but if all addresses were removed from 868 * the interface prior to destruction, the calls will 869 * not be made (e.g. ppp, for which pppd(8) generally 870 * removes addresses before destroying the interface). 871 * Because there is no invariant that multicast 872 * memberships only exist for interfaces with IPv4 873 * addresses, we must call PURGEIF regardless of 874 * addresses. (Protocols which might store ifnet 875 * pointers are marked with PR_PURGEIF.) 876 */ 877 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 878 so.so_proto = pr; 879 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 880 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 881 } 882 } 883 884 (void)pfil_run_hooks(if_pfil, 885 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 886 (void)pfil_head_destroy(ifp->if_pfil); 887 888 /* Announce that the interface is gone. */ 889 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 890 891 ifindex2ifnet[ifp->if_index] = NULL; 892 893 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 894 895 ifioctl_detach(ifp); 896 897 /* 898 * remove packets that came from ifp, from software interrupt queues. 899 */ 900 DOMAIN_FOREACH(dp) { 901 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 902 struct ifqueue *iq = dp->dom_ifqueues[i]; 903 if (iq == NULL) 904 break; 905 dp->dom_ifqueues[i] = NULL; 906 if_detach_queues(ifp, iq); 907 } 908 } 909 910 /* 911 * IP queues have to be processed separately: net-queue barrier 912 * ensures that the packets are dequeued while a cross-call will 913 * ensure that the interrupts have completed. FIXME: not quite.. 914 */ 915 #ifdef INET 916 pktq_barrier(ip_pktq); 917 #endif 918 #ifdef INET6 919 if (in6_present) 920 pktq_barrier(ip6_pktq); 921 #endif 922 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 923 xc_wait(xc); 924 925 splx(s); 926 } 927 928 static void 929 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 930 { 931 struct mbuf *m, *prev, *next; 932 933 prev = NULL; 934 for (m = q->ifq_head; m != NULL; m = next) { 935 KASSERT((m->m_flags & M_PKTHDR) != 0); 936 937 next = m->m_nextpkt; 938 if (m->m_pkthdr.rcvif != ifp) { 939 prev = m; 940 continue; 941 } 942 943 if (prev != NULL) 944 prev->m_nextpkt = m->m_nextpkt; 945 else 946 q->ifq_head = m->m_nextpkt; 947 if (q->ifq_tail == m) 948 q->ifq_tail = prev; 949 q->ifq_len--; 950 951 m->m_nextpkt = NULL; 952 m_freem(m); 953 IF_DROP(q); 954 } 955 } 956 957 /* 958 * Callback for a radix tree walk to delete all references to an 959 * ifnet. 960 */ 961 static int 962 if_rt_walktree(struct rtentry *rt, void *v) 963 { 964 struct ifnet *ifp = (struct ifnet *)v; 965 int error; 966 struct rtentry *retrt; 967 968 if (rt->rt_ifp != ifp) 969 return 0; 970 971 /* Delete the entry. */ 972 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 973 rt_mask(rt), rt->rt_flags, &retrt); 974 if (error == 0) { 975 KASSERT(retrt == rt); 976 KASSERT((retrt->rt_flags & RTF_UP) == 0); 977 retrt->rt_ifp = NULL; 978 rtfree(retrt); 979 } else { 980 printf("%s: warning: unable to delete rtentry @ %p, " 981 "error = %d\n", ifp->if_xname, rt, error); 982 } 983 return ERESTART; 984 } 985 986 /* 987 * Create a clone network interface. 988 */ 989 static int 990 if_clone_create(const char *name) 991 { 992 struct if_clone *ifc; 993 int unit; 994 995 ifc = if_clone_lookup(name, &unit); 996 if (ifc == NULL) 997 return EINVAL; 998 999 if (ifunit(name) != NULL) 1000 return EEXIST; 1001 1002 return (*ifc->ifc_create)(ifc, unit); 1003 } 1004 1005 /* 1006 * Destroy a clone network interface. 1007 */ 1008 static int 1009 if_clone_destroy(const char *name) 1010 { 1011 struct if_clone *ifc; 1012 struct ifnet *ifp; 1013 1014 ifc = if_clone_lookup(name, NULL); 1015 if (ifc == NULL) 1016 return EINVAL; 1017 1018 ifp = ifunit(name); 1019 if (ifp == NULL) 1020 return ENXIO; 1021 1022 if (ifc->ifc_destroy == NULL) 1023 return EOPNOTSUPP; 1024 1025 return (*ifc->ifc_destroy)(ifp); 1026 } 1027 1028 /* 1029 * Look up a network interface cloner. 1030 */ 1031 static struct if_clone * 1032 if_clone_lookup(const char *name, int *unitp) 1033 { 1034 struct if_clone *ifc; 1035 const char *cp; 1036 char *dp, ifname[IFNAMSIZ + 3]; 1037 int unit; 1038 1039 strcpy(ifname, "if_"); 1040 /* separate interface name from unit */ 1041 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1042 *cp && (*cp < '0' || *cp > '9');) 1043 *dp++ = *cp++; 1044 1045 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1046 return NULL; /* No name or unit number */ 1047 *dp++ = '\0'; 1048 1049 again: 1050 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1051 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1052 break; 1053 } 1054 1055 if (ifc == NULL) { 1056 if (*ifname == '\0' || 1057 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1058 return NULL; 1059 *ifname = '\0'; 1060 goto again; 1061 } 1062 1063 unit = 0; 1064 while (cp - name < IFNAMSIZ && *cp) { 1065 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1066 /* Bogus unit number. */ 1067 return NULL; 1068 } 1069 unit = (unit * 10) + (*cp++ - '0'); 1070 } 1071 1072 if (unitp != NULL) 1073 *unitp = unit; 1074 return ifc; 1075 } 1076 1077 /* 1078 * Register a network interface cloner. 1079 */ 1080 void 1081 if_clone_attach(struct if_clone *ifc) 1082 { 1083 1084 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1085 if_cloners_count++; 1086 } 1087 1088 /* 1089 * Unregister a network interface cloner. 1090 */ 1091 void 1092 if_clone_detach(struct if_clone *ifc) 1093 { 1094 1095 LIST_REMOVE(ifc, ifc_list); 1096 if_cloners_count--; 1097 } 1098 1099 /* 1100 * Provide list of interface cloners to userspace. 1101 */ 1102 int 1103 if_clone_list(int buf_count, char *buffer, int *total) 1104 { 1105 char outbuf[IFNAMSIZ], *dst; 1106 struct if_clone *ifc; 1107 int count, error = 0; 1108 1109 *total = if_cloners_count; 1110 if ((dst = buffer) == NULL) { 1111 /* Just asking how many there are. */ 1112 return 0; 1113 } 1114 1115 if (buf_count < 0) 1116 return EINVAL; 1117 1118 count = (if_cloners_count < buf_count) ? 1119 if_cloners_count : buf_count; 1120 1121 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1122 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1123 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1124 if (outbuf[sizeof(outbuf) - 1] != '\0') 1125 return ENAMETOOLONG; 1126 error = copyout(outbuf, dst, sizeof(outbuf)); 1127 if (error != 0) 1128 break; 1129 } 1130 1131 return error; 1132 } 1133 1134 void 1135 ifaref(struct ifaddr *ifa) 1136 { 1137 ifa->ifa_refcnt++; 1138 } 1139 1140 void 1141 ifafree(struct ifaddr *ifa) 1142 { 1143 KASSERT(ifa != NULL); 1144 KASSERT(ifa->ifa_refcnt > 0); 1145 1146 if (--ifa->ifa_refcnt == 0) { 1147 free(ifa, M_IFADDR); 1148 } 1149 } 1150 1151 void 1152 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1153 { 1154 ifa->ifa_ifp = ifp; 1155 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1156 ifaref(ifa); 1157 } 1158 1159 void 1160 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1161 { 1162 KASSERT(ifa->ifa_ifp == ifp); 1163 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1164 ifafree(ifa); 1165 } 1166 1167 static inline int 1168 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1169 { 1170 return sockaddr_cmp(sa1, sa2) == 0; 1171 } 1172 1173 /* 1174 * Locate an interface based on a complete address. 1175 */ 1176 /*ARGSUSED*/ 1177 struct ifaddr * 1178 ifa_ifwithaddr(const struct sockaddr *addr) 1179 { 1180 struct ifnet *ifp; 1181 struct ifaddr *ifa; 1182 1183 IFNET_FOREACH(ifp) { 1184 if (ifp->if_output == if_nulloutput) 1185 continue; 1186 IFADDR_FOREACH(ifa, ifp) { 1187 if (ifa->ifa_addr->sa_family != addr->sa_family) 1188 continue; 1189 if (equal(addr, ifa->ifa_addr)) 1190 return ifa; 1191 if ((ifp->if_flags & IFF_BROADCAST) && 1192 ifa->ifa_broadaddr && 1193 /* IP6 doesn't have broadcast */ 1194 ifa->ifa_broadaddr->sa_len != 0 && 1195 equal(ifa->ifa_broadaddr, addr)) 1196 return ifa; 1197 } 1198 } 1199 return NULL; 1200 } 1201 1202 /* 1203 * Locate the point to point interface with a given destination address. 1204 */ 1205 /*ARGSUSED*/ 1206 struct ifaddr * 1207 ifa_ifwithdstaddr(const struct sockaddr *addr) 1208 { 1209 struct ifnet *ifp; 1210 struct ifaddr *ifa; 1211 1212 IFNET_FOREACH(ifp) { 1213 if (ifp->if_output == if_nulloutput) 1214 continue; 1215 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1216 continue; 1217 IFADDR_FOREACH(ifa, ifp) { 1218 if (ifa->ifa_addr->sa_family != addr->sa_family || 1219 ifa->ifa_dstaddr == NULL) 1220 continue; 1221 if (equal(addr, ifa->ifa_dstaddr)) 1222 return ifa; 1223 } 1224 } 1225 return NULL; 1226 } 1227 1228 /* 1229 * Find an interface on a specific network. If many, choice 1230 * is most specific found. 1231 */ 1232 struct ifaddr * 1233 ifa_ifwithnet(const struct sockaddr *addr) 1234 { 1235 struct ifnet *ifp; 1236 struct ifaddr *ifa; 1237 const struct sockaddr_dl *sdl; 1238 struct ifaddr *ifa_maybe = 0; 1239 u_int af = addr->sa_family; 1240 const char *addr_data = addr->sa_data, *cplim; 1241 1242 if (af == AF_LINK) { 1243 sdl = satocsdl(addr); 1244 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1245 ifindex2ifnet[sdl->sdl_index] && 1246 ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) { 1247 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1248 } 1249 } 1250 #ifdef NETATALK 1251 if (af == AF_APPLETALK) { 1252 const struct sockaddr_at *sat, *sat2; 1253 sat = (const struct sockaddr_at *)addr; 1254 IFNET_FOREACH(ifp) { 1255 if (ifp->if_output == if_nulloutput) 1256 continue; 1257 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1258 if (ifa == NULL) 1259 continue; 1260 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1261 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1262 return ifa; /* exact match */ 1263 if (ifa_maybe == NULL) { 1264 /* else keep the if with the right range */ 1265 ifa_maybe = ifa; 1266 } 1267 } 1268 return ifa_maybe; 1269 } 1270 #endif 1271 IFNET_FOREACH(ifp) { 1272 if (ifp->if_output == if_nulloutput) 1273 continue; 1274 IFADDR_FOREACH(ifa, ifp) { 1275 const char *cp, *cp2, *cp3; 1276 1277 if (ifa->ifa_addr->sa_family != af || 1278 ifa->ifa_netmask == NULL) 1279 next: continue; 1280 cp = addr_data; 1281 cp2 = ifa->ifa_addr->sa_data; 1282 cp3 = ifa->ifa_netmask->sa_data; 1283 cplim = (const char *)ifa->ifa_netmask + 1284 ifa->ifa_netmask->sa_len; 1285 while (cp3 < cplim) { 1286 if ((*cp++ ^ *cp2++) & *cp3++) { 1287 /* want to continue for() loop */ 1288 goto next; 1289 } 1290 } 1291 if (ifa_maybe == NULL || 1292 rn_refines((void *)ifa->ifa_netmask, 1293 (void *)ifa_maybe->ifa_netmask)) 1294 ifa_maybe = ifa; 1295 } 1296 } 1297 return ifa_maybe; 1298 } 1299 1300 /* 1301 * Find the interface of the addresss. 1302 */ 1303 struct ifaddr * 1304 ifa_ifwithladdr(const struct sockaddr *addr) 1305 { 1306 struct ifaddr *ia; 1307 1308 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1309 (ia = ifa_ifwithnet(addr))) 1310 return ia; 1311 return NULL; 1312 } 1313 1314 /* 1315 * Find an interface using a specific address family 1316 */ 1317 struct ifaddr * 1318 ifa_ifwithaf(int af) 1319 { 1320 struct ifnet *ifp; 1321 struct ifaddr *ifa; 1322 1323 IFNET_FOREACH(ifp) { 1324 if (ifp->if_output == if_nulloutput) 1325 continue; 1326 IFADDR_FOREACH(ifa, ifp) { 1327 if (ifa->ifa_addr->sa_family == af) 1328 return ifa; 1329 } 1330 } 1331 return NULL; 1332 } 1333 1334 /* 1335 * Find an interface address specific to an interface best matching 1336 * a given address. 1337 */ 1338 struct ifaddr * 1339 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1340 { 1341 struct ifaddr *ifa; 1342 const char *cp, *cp2, *cp3; 1343 const char *cplim; 1344 struct ifaddr *ifa_maybe = 0; 1345 u_int af = addr->sa_family; 1346 1347 if (ifp->if_output == if_nulloutput) 1348 return NULL; 1349 1350 if (af >= AF_MAX) 1351 return NULL; 1352 1353 IFADDR_FOREACH(ifa, ifp) { 1354 if (ifa->ifa_addr->sa_family != af) 1355 continue; 1356 ifa_maybe = ifa; 1357 if (ifa->ifa_netmask == NULL) { 1358 if (equal(addr, ifa->ifa_addr) || 1359 (ifa->ifa_dstaddr && 1360 equal(addr, ifa->ifa_dstaddr))) 1361 return ifa; 1362 continue; 1363 } 1364 cp = addr->sa_data; 1365 cp2 = ifa->ifa_addr->sa_data; 1366 cp3 = ifa->ifa_netmask->sa_data; 1367 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1368 for (; cp3 < cplim; cp3++) { 1369 if ((*cp++ ^ *cp2++) & *cp3) 1370 break; 1371 } 1372 if (cp3 == cplim) 1373 return ifa; 1374 } 1375 return ifa_maybe; 1376 } 1377 1378 /* 1379 * Default action when installing a route with a Link Level gateway. 1380 * Lookup an appropriate real ifa to point to. 1381 * This should be moved to /sys/net/link.c eventually. 1382 */ 1383 void 1384 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1385 { 1386 struct ifaddr *ifa; 1387 const struct sockaddr *dst; 1388 struct ifnet *ifp; 1389 1390 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1391 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1392 return; 1393 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1394 rt_replace_ifa(rt, ifa); 1395 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1396 ifa->ifa_rtrequest(cmd, rt, info); 1397 } 1398 } 1399 1400 /* 1401 * Handle a change in the interface link state. 1402 * XXX: We should listen to the routing socket in-kernel rather 1403 * than calling in6_if_link_* functions directly from here. 1404 */ 1405 void 1406 if_link_state_change(struct ifnet *ifp, int link_state) 1407 { 1408 int s; 1409 int old_link_state; 1410 struct domain *dp; 1411 1412 s = splnet(); 1413 if (ifp->if_link_state == link_state) { 1414 splx(s); 1415 return; 1416 } 1417 1418 old_link_state = ifp->if_link_state; 1419 ifp->if_link_state = link_state; 1420 #ifdef DEBUG 1421 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1422 link_state == LINK_STATE_UP ? "UP" : 1423 link_state == LINK_STATE_DOWN ? "DOWN" : 1424 "UNKNOWN", 1425 old_link_state == LINK_STATE_UP ? "UP" : 1426 old_link_state == LINK_STATE_DOWN ? "DOWN" : 1427 "UNKNOWN"); 1428 #endif 1429 1430 /* 1431 * When going from UNKNOWN to UP, we need to mark existing 1432 * addresses as tentative and restart DAD as we may have 1433 * erroneously not found a duplicate. 1434 * 1435 * This needs to happen before rt_ifmsg to avoid a race where 1436 * listeners would have an address and expect it to work right 1437 * away. 1438 */ 1439 if (link_state == LINK_STATE_UP && 1440 old_link_state == LINK_STATE_UNKNOWN) 1441 { 1442 DOMAIN_FOREACH(dp) { 1443 if (dp->dom_if_link_state_change != NULL) 1444 dp->dom_if_link_state_change(ifp, 1445 LINK_STATE_DOWN); 1446 } 1447 } 1448 1449 /* Notify that the link state has changed. */ 1450 rt_ifmsg(ifp); 1451 1452 #if NCARP > 0 1453 if (ifp->if_carp) 1454 carp_carpdev_state(ifp); 1455 #endif 1456 1457 DOMAIN_FOREACH(dp) { 1458 if (dp->dom_if_link_state_change != NULL) 1459 dp->dom_if_link_state_change(ifp, link_state); 1460 } 1461 1462 splx(s); 1463 } 1464 1465 /* 1466 * Default action when installing a local route on a point-to-point 1467 * interface. 1468 */ 1469 void 1470 p2p_rtrequest(int req, struct rtentry *rt, 1471 __unused const struct rt_addrinfo *info) 1472 { 1473 struct ifnet *ifp = rt->rt_ifp; 1474 struct ifaddr *ifa, *lo0ifa; 1475 1476 switch (req) { 1477 case RTM_ADD: 1478 if ((rt->rt_flags & RTF_LOCAL) == 0) 1479 break; 1480 1481 IFADDR_FOREACH(ifa, ifp) { 1482 if (equal(rt_getkey(rt), ifa->ifa_addr)) 1483 break; 1484 } 1485 if (ifa == NULL) 1486 break; 1487 1488 /* 1489 * Ensure lo0 has an address of the same family. 1490 */ 1491 IFADDR_FOREACH(lo0ifa, lo0ifp) { 1492 if (lo0ifa->ifa_addr->sa_family == 1493 ifa->ifa_addr->sa_family) 1494 break; 1495 } 1496 if (lo0ifa == NULL) 1497 break; 1498 1499 rt->rt_ifp = lo0ifp; 1500 rt->rt_flags &= ~RTF_LLINFO; 1501 1502 /* 1503 * Make sure to set rt->rt_ifa to the interface 1504 * address we are using, otherwise we will have trouble 1505 * with source address selection. 1506 */ 1507 if (ifa != rt->rt_ifa) 1508 rt_replace_ifa(rt, ifa); 1509 break; 1510 case RTM_DELETE: 1511 case RTM_RESOLVE: 1512 default: 1513 break; 1514 } 1515 } 1516 1517 /* 1518 * Mark an interface down and notify protocols of 1519 * the transition. 1520 * NOTE: must be called at splsoftnet or equivalent. 1521 */ 1522 void 1523 if_down(struct ifnet *ifp) 1524 { 1525 struct ifaddr *ifa; 1526 struct domain *dp; 1527 1528 ifp->if_flags &= ~IFF_UP; 1529 nanotime(&ifp->if_lastchange); 1530 IFADDR_FOREACH(ifa, ifp) 1531 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1532 IFQ_PURGE(&ifp->if_snd); 1533 #if NCARP > 0 1534 if (ifp->if_carp) 1535 carp_carpdev_state(ifp); 1536 #endif 1537 rt_ifmsg(ifp); 1538 DOMAIN_FOREACH(dp) { 1539 if (dp->dom_if_down) 1540 dp->dom_if_down(ifp); 1541 } 1542 } 1543 1544 /* 1545 * Mark an interface up and notify protocols of 1546 * the transition. 1547 * NOTE: must be called at splsoftnet or equivalent. 1548 */ 1549 void 1550 if_up(struct ifnet *ifp) 1551 { 1552 #ifdef notyet 1553 struct ifaddr *ifa; 1554 #endif 1555 struct domain *dp; 1556 1557 ifp->if_flags |= IFF_UP; 1558 nanotime(&ifp->if_lastchange); 1559 #ifdef notyet 1560 /* this has no effect on IP, and will kill all ISO connections XXX */ 1561 IFADDR_FOREACH(ifa, ifp) 1562 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1563 #endif 1564 #if NCARP > 0 1565 if (ifp->if_carp) 1566 carp_carpdev_state(ifp); 1567 #endif 1568 rt_ifmsg(ifp); 1569 DOMAIN_FOREACH(dp) { 1570 if (dp->dom_if_up) 1571 dp->dom_if_up(ifp); 1572 } 1573 } 1574 1575 /* 1576 * Handle interface slowtimo timer routine. Called 1577 * from softclock, we decrement timer (if set) and 1578 * call the appropriate interface routine on expiration. 1579 */ 1580 static void 1581 if_slowtimo(void *arg) 1582 { 1583 void (*slowtimo)(struct ifnet *); 1584 struct ifnet *ifp = arg; 1585 int s; 1586 1587 slowtimo = ifp->if_slowtimo; 1588 if (__predict_false(slowtimo == NULL)) 1589 return; 1590 1591 s = splnet(); 1592 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 1593 (*slowtimo)(ifp); 1594 1595 splx(s); 1596 1597 if (__predict_true(ifp->if_slowtimo != NULL)) 1598 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 1599 } 1600 1601 /* 1602 * Set/clear promiscuous mode on interface ifp based on the truth value 1603 * of pswitch. The calls are reference counted so that only the first 1604 * "on" request actually has an effect, as does the final "off" request. 1605 * Results are undefined if the "off" and "on" requests are not matched. 1606 */ 1607 int 1608 ifpromisc(struct ifnet *ifp, int pswitch) 1609 { 1610 int pcount, ret; 1611 short nflags; 1612 1613 pcount = ifp->if_pcount; 1614 if (pswitch) { 1615 /* 1616 * Allow the device to be "placed" into promiscuous 1617 * mode even if it is not configured up. It will 1618 * consult IFF_PROMISC when it is brought up. 1619 */ 1620 if (ifp->if_pcount++ != 0) 1621 return 0; 1622 nflags = ifp->if_flags | IFF_PROMISC; 1623 } else { 1624 if (--ifp->if_pcount > 0) 1625 return 0; 1626 nflags = ifp->if_flags & ~IFF_PROMISC; 1627 } 1628 ret = if_flags_set(ifp, nflags); 1629 /* Restore interface state if not successful. */ 1630 if (ret != 0) { 1631 ifp->if_pcount = pcount; 1632 } 1633 return ret; 1634 } 1635 1636 /* 1637 * Map interface name to 1638 * interface structure pointer. 1639 */ 1640 struct ifnet * 1641 ifunit(const char *name) 1642 { 1643 struct ifnet *ifp; 1644 const char *cp = name; 1645 u_int unit = 0; 1646 u_int i; 1647 1648 /* 1649 * If the entire name is a number, treat it as an ifindex. 1650 */ 1651 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 1652 unit = unit * 10 + (*cp - '0'); 1653 } 1654 1655 /* 1656 * If the number took all of the name, then it's a valid ifindex. 1657 */ 1658 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 1659 if (unit >= if_indexlim) 1660 return NULL; 1661 ifp = ifindex2ifnet[unit]; 1662 if (ifp == NULL || ifp->if_output == if_nulloutput) 1663 return NULL; 1664 return ifp; 1665 } 1666 1667 IFNET_FOREACH(ifp) { 1668 if (ifp->if_output == if_nulloutput) 1669 continue; 1670 if (strcmp(ifp->if_xname, name) == 0) 1671 return ifp; 1672 } 1673 return NULL; 1674 } 1675 1676 ifnet_t * 1677 if_byindex(u_int idx) 1678 { 1679 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 1680 } 1681 1682 /* common */ 1683 int 1684 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 1685 { 1686 int s; 1687 struct ifreq *ifr; 1688 struct ifcapreq *ifcr; 1689 struct ifdatareq *ifdr; 1690 1691 switch (cmd) { 1692 case SIOCSIFCAP: 1693 ifcr = data; 1694 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 1695 return EINVAL; 1696 1697 if (ifcr->ifcr_capenable == ifp->if_capenable) 1698 return 0; 1699 1700 ifp->if_capenable = ifcr->ifcr_capenable; 1701 1702 /* Pre-compute the checksum flags mask. */ 1703 ifp->if_csum_flags_tx = 0; 1704 ifp->if_csum_flags_rx = 0; 1705 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 1706 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 1707 } 1708 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 1709 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 1710 } 1711 1712 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 1713 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 1714 } 1715 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 1716 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 1717 } 1718 1719 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 1720 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 1721 } 1722 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 1723 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 1724 } 1725 1726 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 1727 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 1728 } 1729 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 1730 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 1731 } 1732 1733 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 1734 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 1735 } 1736 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 1737 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 1738 } 1739 if (ifp->if_flags & IFF_UP) 1740 return ENETRESET; 1741 return 0; 1742 case SIOCSIFFLAGS: 1743 ifr = data; 1744 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 1745 s = splnet(); 1746 if_down(ifp); 1747 splx(s); 1748 } 1749 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 1750 s = splnet(); 1751 if_up(ifp); 1752 splx(s); 1753 } 1754 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1755 (ifr->ifr_flags &~ IFF_CANTCHANGE); 1756 break; 1757 case SIOCGIFFLAGS: 1758 ifr = data; 1759 ifr->ifr_flags = ifp->if_flags; 1760 break; 1761 1762 case SIOCGIFMETRIC: 1763 ifr = data; 1764 ifr->ifr_metric = ifp->if_metric; 1765 break; 1766 1767 case SIOCGIFMTU: 1768 ifr = data; 1769 ifr->ifr_mtu = ifp->if_mtu; 1770 break; 1771 1772 case SIOCGIFDLT: 1773 ifr = data; 1774 ifr->ifr_dlt = ifp->if_dlt; 1775 break; 1776 1777 case SIOCGIFCAP: 1778 ifcr = data; 1779 ifcr->ifcr_capabilities = ifp->if_capabilities; 1780 ifcr->ifcr_capenable = ifp->if_capenable; 1781 break; 1782 1783 case SIOCSIFMETRIC: 1784 ifr = data; 1785 ifp->if_metric = ifr->ifr_metric; 1786 break; 1787 1788 case SIOCGIFDATA: 1789 ifdr = data; 1790 ifdr->ifdr_data = ifp->if_data; 1791 break; 1792 1793 case SIOCGIFINDEX: 1794 ifr = data; 1795 ifr->ifr_index = ifp->if_index; 1796 break; 1797 1798 case SIOCZIFDATA: 1799 ifdr = data; 1800 ifdr->ifdr_data = ifp->if_data; 1801 /* 1802 * Assumes that the volatile counters that can be 1803 * zero'ed are at the end of if_data. 1804 */ 1805 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 1806 offsetof(struct if_data, ifi_ipackets)); 1807 /* 1808 * The memset() clears to the bottm of if_data. In the area, 1809 * if_lastchange is included. Please be careful if new entry 1810 * will be added into if_data or rewite this. 1811 * 1812 * And also, update if_lastchnage. 1813 */ 1814 getnanotime(&ifp->if_lastchange); 1815 break; 1816 case SIOCSIFMTU: 1817 ifr = data; 1818 if (ifp->if_mtu == ifr->ifr_mtu) 1819 break; 1820 ifp->if_mtu = ifr->ifr_mtu; 1821 /* 1822 * If the link MTU changed, do network layer specific procedure. 1823 */ 1824 #ifdef INET6 1825 if (in6_present) 1826 nd6_setmtu(ifp); 1827 #endif 1828 return ENETRESET; 1829 default: 1830 return ENOTTY; 1831 } 1832 return 0; 1833 } 1834 1835 int 1836 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 1837 { 1838 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 1839 struct ifaddr *ifa; 1840 const struct sockaddr *any, *sa; 1841 union { 1842 struct sockaddr sa; 1843 struct sockaddr_storage ss; 1844 } u, v; 1845 1846 switch (cmd) { 1847 case SIOCSIFADDRPREF: 1848 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 1849 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1850 NULL) != 0) 1851 return EPERM; 1852 case SIOCGIFADDRPREF: 1853 break; 1854 default: 1855 return EOPNOTSUPP; 1856 } 1857 1858 /* sanity checks */ 1859 if (data == NULL || ifp == NULL) { 1860 panic("invalid argument to %s", __func__); 1861 /*NOTREACHED*/ 1862 } 1863 1864 /* address must be specified on ADD and DELETE */ 1865 sa = sstocsa(&ifap->ifap_addr); 1866 if (sa->sa_family != sofamily(so)) 1867 return EINVAL; 1868 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 1869 return EINVAL; 1870 1871 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 1872 1873 IFADDR_FOREACH(ifa, ifp) { 1874 if (ifa->ifa_addr->sa_family != sa->sa_family) 1875 continue; 1876 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 1877 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 1878 break; 1879 } 1880 if (ifa == NULL) 1881 return EADDRNOTAVAIL; 1882 1883 switch (cmd) { 1884 case SIOCSIFADDRPREF: 1885 ifa->ifa_preference = ifap->ifap_preference; 1886 return 0; 1887 case SIOCGIFADDRPREF: 1888 /* fill in the if_laddrreq structure */ 1889 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 1890 sizeof(ifap->ifap_addr), ifa->ifa_addr); 1891 ifap->ifap_preference = ifa->ifa_preference; 1892 return 0; 1893 default: 1894 return EOPNOTSUPP; 1895 } 1896 } 1897 1898 static void 1899 ifnet_lock_enter(struct ifnet_lock *il) 1900 { 1901 uint64_t *nenter; 1902 1903 /* Before trying to acquire the mutex, increase the count of threads 1904 * who have entered or who wait to enter the critical section. 1905 * Avoid one costly locked memory transaction by keeping a count for 1906 * each CPU. 1907 */ 1908 nenter = percpu_getref(il->il_nenter); 1909 (*nenter)++; 1910 percpu_putref(il->il_nenter); 1911 mutex_enter(&il->il_lock); 1912 } 1913 1914 static void 1915 ifnet_lock_exit(struct ifnet_lock *il) 1916 { 1917 /* Increase the count of threads who have exited the critical 1918 * section. Increase while we still hold the lock. 1919 */ 1920 il->il_nexit++; 1921 mutex_exit(&il->il_lock); 1922 } 1923 1924 /* 1925 * Interface ioctls. 1926 */ 1927 static int 1928 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 1929 { 1930 struct ifnet *ifp; 1931 struct ifreq *ifr; 1932 int error = 0; 1933 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 1934 u_long ocmd = cmd; 1935 #endif 1936 short oif_flags; 1937 #ifdef COMPAT_OIFREQ 1938 struct ifreq ifrb; 1939 struct oifreq *oifr = NULL; 1940 #endif 1941 int r; 1942 1943 switch (cmd) { 1944 #ifdef COMPAT_OIFREQ 1945 case OSIOCGIFCONF: 1946 case OOSIOCGIFCONF: 1947 return compat_ifconf(cmd, data); 1948 #endif 1949 #ifdef COMPAT_OIFDATA 1950 case OSIOCGIFDATA: 1951 case OSIOCZIFDATA: 1952 return compat_ifdatareq(l, cmd, data); 1953 #endif 1954 case SIOCGIFCONF: 1955 return ifconf(cmd, data); 1956 case SIOCINITIFADDR: 1957 return EPERM; 1958 } 1959 1960 #ifdef COMPAT_OIFREQ 1961 cmd = compat_cvtcmd(cmd); 1962 if (cmd != ocmd) { 1963 oifr = data; 1964 data = ifr = &ifrb; 1965 ifreqo2n(oifr, ifr); 1966 } else 1967 #endif 1968 ifr = data; 1969 1970 ifp = ifunit(ifr->ifr_name); 1971 1972 switch (cmd) { 1973 case SIOCIFCREATE: 1974 case SIOCIFDESTROY: 1975 if (l != NULL) { 1976 error = kauth_authorize_network(l->l_cred, 1977 KAUTH_NETWORK_INTERFACE, 1978 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 1979 (void *)cmd, NULL); 1980 if (error != 0) 1981 return error; 1982 } 1983 mutex_enter(&if_clone_mtx); 1984 r = (cmd == SIOCIFCREATE) ? 1985 if_clone_create(ifr->ifr_name) : 1986 if_clone_destroy(ifr->ifr_name); 1987 mutex_exit(&if_clone_mtx); 1988 return r; 1989 1990 case SIOCIFGCLONERS: 1991 { 1992 struct if_clonereq *req = (struct if_clonereq *)data; 1993 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 1994 &req->ifcr_total); 1995 } 1996 } 1997 1998 if (ifp == NULL) 1999 return ENXIO; 2000 2001 switch (cmd) { 2002 case SIOCALIFADDR: 2003 case SIOCDLIFADDR: 2004 case SIOCSIFADDRPREF: 2005 case SIOCSIFFLAGS: 2006 case SIOCSIFCAP: 2007 case SIOCSIFMETRIC: 2008 case SIOCZIFDATA: 2009 case SIOCSIFMTU: 2010 case SIOCSIFPHYADDR: 2011 case SIOCDIFPHYADDR: 2012 #ifdef INET6 2013 case SIOCSIFPHYADDR_IN6: 2014 #endif 2015 case SIOCSLIFPHYADDR: 2016 case SIOCADDMULTI: 2017 case SIOCDELMULTI: 2018 case SIOCSIFMEDIA: 2019 case SIOCSDRVSPEC: 2020 case SIOCG80211: 2021 case SIOCS80211: 2022 case SIOCS80211NWID: 2023 case SIOCS80211NWKEY: 2024 case SIOCS80211POWER: 2025 case SIOCS80211BSSID: 2026 case SIOCS80211CHANNEL: 2027 case SIOCSLINKSTR: 2028 if (l != NULL) { 2029 error = kauth_authorize_network(l->l_cred, 2030 KAUTH_NETWORK_INTERFACE, 2031 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2032 (void *)cmd, NULL); 2033 if (error != 0) 2034 return error; 2035 } 2036 } 2037 2038 oif_flags = ifp->if_flags; 2039 2040 ifnet_lock_enter(ifp->if_ioctl_lock); 2041 error = (*ifp->if_ioctl)(ifp, cmd, data); 2042 if (error != ENOTTY) 2043 ; 2044 else if (so->so_proto == NULL) 2045 error = EOPNOTSUPP; 2046 else { 2047 #ifdef COMPAT_OSOCK 2048 error = compat_ifioctl(so, ocmd, cmd, data, l); 2049 #else 2050 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2051 cmd, data, ifp); 2052 #endif 2053 } 2054 2055 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2056 if ((ifp->if_flags & IFF_UP) != 0) { 2057 int s = splnet(); 2058 if_up(ifp); 2059 splx(s); 2060 } 2061 } 2062 #ifdef COMPAT_OIFREQ 2063 if (cmd != ocmd) 2064 ifreqn2o(oifr, ifr); 2065 #endif 2066 2067 ifnet_lock_exit(ifp->if_ioctl_lock); 2068 return error; 2069 } 2070 2071 /* This callback adds to the sum in `arg' the number of 2072 * threads on `ci' who have entered or who wait to enter the 2073 * critical section. 2074 */ 2075 static void 2076 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci) 2077 { 2078 uint64_t *sum = arg, *nenter = p; 2079 2080 *sum += *nenter; 2081 } 2082 2083 /* Return the number of threads who have entered or who wait 2084 * to enter the critical section on all CPUs. 2085 */ 2086 static uint64_t 2087 ifnet_lock_entrances(struct ifnet_lock *il) 2088 { 2089 uint64_t sum = 0; 2090 2091 percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum); 2092 2093 return sum; 2094 } 2095 2096 static int 2097 ifioctl_attach(struct ifnet *ifp) 2098 { 2099 struct ifnet_lock *il; 2100 2101 /* If the driver has not supplied its own if_ioctl, then 2102 * supply the default. 2103 */ 2104 if (ifp->if_ioctl == NULL) 2105 ifp->if_ioctl = ifioctl_common; 2106 2107 /* Create an ifnet_lock for synchronizing ifioctls. */ 2108 if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL) 2109 return ENOMEM; 2110 2111 il->il_nenter = percpu_alloc(sizeof(uint64_t)); 2112 if (il->il_nenter == NULL) { 2113 kmem_free(il, sizeof(*il)); 2114 return ENOMEM; 2115 } 2116 2117 mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE); 2118 cv_init(&il->il_emptied, ifp->if_xname); 2119 2120 ifp->if_ioctl_lock = il; 2121 2122 return 0; 2123 } 2124 2125 /* 2126 * This must not be called until after `ifp' has been withdrawn from the 2127 * ifnet tables so that ifioctl() cannot get a handle on it by calling 2128 * ifunit(). 2129 */ 2130 static void 2131 ifioctl_detach(struct ifnet *ifp) 2132 { 2133 struct ifnet_lock *il; 2134 2135 il = ifp->if_ioctl_lock; 2136 mutex_enter(&il->il_lock); 2137 /* Install if_nullioctl to make sure that any thread that 2138 * subsequently enters the critical section will quit it 2139 * immediately and signal the condition variable that we 2140 * wait on, below. 2141 */ 2142 ifp->if_ioctl = if_nullioctl; 2143 /* Sleep while threads are still in the critical section or 2144 * wait to enter it. 2145 */ 2146 while (ifnet_lock_entrances(il) != il->il_nexit) 2147 cv_wait(&il->il_emptied, &il->il_lock); 2148 /* At this point, we are the only thread still in the critical 2149 * section, and no new thread can get a handle on the ifioctl 2150 * lock, so it is safe to free its memory. 2151 */ 2152 mutex_exit(&il->il_lock); 2153 ifp->if_ioctl_lock = NULL; 2154 percpu_free(il->il_nenter, sizeof(uint64_t)); 2155 il->il_nenter = NULL; 2156 cv_destroy(&il->il_emptied); 2157 mutex_destroy(&il->il_lock); 2158 kmem_free(il, sizeof(*il)); 2159 } 2160 2161 /* 2162 * Return interface configuration 2163 * of system. List may be used 2164 * in later ioctl's (above) to get 2165 * other information. 2166 * 2167 * Each record is a struct ifreq. Before the addition of 2168 * sockaddr_storage, the API rule was that sockaddr flavors that did 2169 * not fit would extend beyond the struct ifreq, with the next struct 2170 * ifreq starting sa_len beyond the struct sockaddr. Because the 2171 * union in struct ifreq includes struct sockaddr_storage, every kind 2172 * of sockaddr must fit. Thus, there are no longer any overlength 2173 * records. 2174 * 2175 * Records are added to the user buffer if they fit, and ifc_len is 2176 * adjusted to the length that was written. Thus, the user is only 2177 * assured of getting the complete list if ifc_len on return is at 2178 * least sizeof(struct ifreq) less than it was on entry. 2179 * 2180 * If the user buffer pointer is NULL, this routine copies no data and 2181 * returns the amount of space that would be needed. 2182 * 2183 * Invariants: 2184 * ifrp points to the next part of the user's buffer to be used. If 2185 * ifrp != NULL, space holds the number of bytes remaining that we may 2186 * write at ifrp. Otherwise, space holds the number of bytes that 2187 * would have been written had there been adequate space. 2188 */ 2189 /*ARGSUSED*/ 2190 static int 2191 ifconf(u_long cmd, void *data) 2192 { 2193 struct ifconf *ifc = (struct ifconf *)data; 2194 struct ifnet *ifp; 2195 struct ifaddr *ifa; 2196 struct ifreq ifr, *ifrp = NULL; 2197 int space = 0, error = 0; 2198 const int sz = (int)sizeof(struct ifreq); 2199 const bool docopy = ifc->ifc_req != NULL; 2200 2201 if (docopy) { 2202 space = ifc->ifc_len; 2203 ifrp = ifc->ifc_req; 2204 } 2205 2206 IFNET_FOREACH(ifp) { 2207 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2208 sizeof(ifr.ifr_name)); 2209 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') 2210 return ENAMETOOLONG; 2211 if (IFADDR_EMPTY(ifp)) { 2212 /* Interface with no addresses - send zero sockaddr. */ 2213 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2214 if (!docopy) { 2215 space += sz; 2216 continue; 2217 } 2218 if (space >= sz) { 2219 error = copyout(&ifr, ifrp, sz); 2220 if (error != 0) 2221 return error; 2222 ifrp++; 2223 space -= sz; 2224 } 2225 } 2226 2227 IFADDR_FOREACH(ifa, ifp) { 2228 struct sockaddr *sa = ifa->ifa_addr; 2229 /* all sockaddrs must fit in sockaddr_storage */ 2230 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2231 2232 if (!docopy) { 2233 space += sz; 2234 continue; 2235 } 2236 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2237 if (space >= sz) { 2238 error = copyout(&ifr, ifrp, sz); 2239 if (error != 0) 2240 return (error); 2241 ifrp++; space -= sz; 2242 } 2243 } 2244 } 2245 if (docopy) { 2246 KASSERT(0 <= space && space <= ifc->ifc_len); 2247 ifc->ifc_len -= space; 2248 } else { 2249 KASSERT(space >= 0); 2250 ifc->ifc_len = space; 2251 } 2252 return (0); 2253 } 2254 2255 int 2256 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2257 { 2258 uint8_t len; 2259 #ifdef COMPAT_OIFREQ 2260 struct ifreq ifrb; 2261 struct oifreq *oifr = NULL; 2262 u_long ocmd = cmd; 2263 cmd = compat_cvtcmd(cmd); 2264 if (cmd != ocmd) { 2265 oifr = (struct oifreq *)(void *)ifr; 2266 ifr = &ifrb; 2267 ifreqo2n(oifr, ifr); 2268 len = sizeof(oifr->ifr_addr); 2269 } else 2270 #endif 2271 len = sizeof(ifr->ifr_ifru.ifru_space); 2272 2273 if (len < sa->sa_len) 2274 return EFBIG; 2275 2276 memset(&ifr->ifr_addr, 0, len); 2277 sockaddr_copy(&ifr->ifr_addr, len, sa); 2278 2279 #ifdef COMPAT_OIFREQ 2280 if (cmd != ocmd) 2281 ifreqn2o(oifr, ifr); 2282 #endif 2283 return 0; 2284 } 2285 2286 /* 2287 * Queue message on interface, and start output if interface 2288 * not yet active. 2289 */ 2290 int 2291 ifq_enqueue(struct ifnet *ifp, struct mbuf *m 2292 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2293 { 2294 int len = m->m_pkthdr.len; 2295 int mflags = m->m_flags; 2296 int s = splnet(); 2297 int error; 2298 2299 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2300 if (error != 0) 2301 goto out; 2302 ifp->if_obytes += len; 2303 if (mflags & M_MCAST) 2304 ifp->if_omcasts++; 2305 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2306 (*ifp->if_start)(ifp); 2307 out: 2308 splx(s); 2309 return error; 2310 } 2311 2312 /* 2313 * Queue message on interface, possibly using a second fast queue 2314 */ 2315 int 2316 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m 2317 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2318 { 2319 int error = 0; 2320 2321 if (ifq != NULL 2322 #ifdef ALTQ 2323 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2324 #endif 2325 ) { 2326 if (IF_QFULL(ifq)) { 2327 IF_DROP(&ifp->if_snd); 2328 m_freem(m); 2329 if (error == 0) 2330 error = ENOBUFS; 2331 } else 2332 IF_ENQUEUE(ifq, m); 2333 } else 2334 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2335 if (error != 0) { 2336 ++ifp->if_oerrors; 2337 return error; 2338 } 2339 return 0; 2340 } 2341 2342 int 2343 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2344 { 2345 int rc; 2346 2347 if (ifp->if_initaddr != NULL) 2348 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2349 else if (src || 2350 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2351 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2352 2353 return rc; 2354 } 2355 2356 int 2357 if_do_dad(struct ifnet *ifp) 2358 { 2359 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2360 return 0; 2361 2362 switch (ifp->if_type) { 2363 case IFT_FAITH: 2364 /* 2365 * These interfaces do not have the IFF_LOOPBACK flag, 2366 * but loop packets back. We do not have to do DAD on such 2367 * interfaces. We should even omit it, because loop-backed 2368 * responses would confuse the DAD procedure. 2369 */ 2370 return 0; 2371 default: 2372 /* 2373 * Our DAD routine requires the interface up and running. 2374 * However, some interfaces can be up before the RUNNING 2375 * status. Additionaly, users may try to assign addresses 2376 * before the interface becomes up (or running). 2377 * We simply skip DAD in such a case as a work around. 2378 * XXX: we should rather mark "tentative" on such addresses, 2379 * and do DAD after the interface becomes ready. 2380 */ 2381 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 2382 (IFF_UP|IFF_RUNNING)) 2383 return 0; 2384 2385 return 1; 2386 } 2387 } 2388 2389 int 2390 if_flags_set(ifnet_t *ifp, const short flags) 2391 { 2392 int rc; 2393 2394 if (ifp->if_setflags != NULL) 2395 rc = (*ifp->if_setflags)(ifp, flags); 2396 else { 2397 short cantflags, chgdflags; 2398 struct ifreq ifr; 2399 2400 chgdflags = ifp->if_flags ^ flags; 2401 cantflags = chgdflags & IFF_CANTCHANGE; 2402 2403 if (cantflags != 0) 2404 ifp->if_flags ^= cantflags; 2405 2406 /* Traditionally, we do not call if_ioctl after 2407 * setting/clearing only IFF_PROMISC if the interface 2408 * isn't IFF_UP. Uphold that tradition. 2409 */ 2410 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2411 return 0; 2412 2413 memset(&ifr, 0, sizeof(ifr)); 2414 2415 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2416 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2417 2418 if (rc != 0 && cantflags != 0) 2419 ifp->if_flags ^= cantflags; 2420 } 2421 2422 return rc; 2423 } 2424 2425 int 2426 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2427 { 2428 int rc; 2429 struct ifreq ifr; 2430 2431 if (ifp->if_mcastop != NULL) 2432 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2433 else { 2434 ifreq_setaddr(cmd, &ifr, sa); 2435 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2436 } 2437 2438 return rc; 2439 } 2440 2441 static void 2442 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2443 struct ifaltq *ifq) 2444 { 2445 const struct sysctlnode *cnode, *rnode; 2446 2447 if (sysctl_createv(clog, 0, NULL, &rnode, 2448 CTLFLAG_PERMANENT, 2449 CTLTYPE_NODE, "interfaces", 2450 SYSCTL_DESCR("Per-interface controls"), 2451 NULL, 0, NULL, 0, 2452 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 2453 goto bad; 2454 2455 if (sysctl_createv(clog, 0, &rnode, &rnode, 2456 CTLFLAG_PERMANENT, 2457 CTLTYPE_NODE, ifname, 2458 SYSCTL_DESCR("Interface controls"), 2459 NULL, 0, NULL, 0, 2460 CTL_CREATE, CTL_EOL) != 0) 2461 goto bad; 2462 2463 if (sysctl_createv(clog, 0, &rnode, &rnode, 2464 CTLFLAG_PERMANENT, 2465 CTLTYPE_NODE, "sndq", 2466 SYSCTL_DESCR("Interface output queue controls"), 2467 NULL, 0, NULL, 0, 2468 CTL_CREATE, CTL_EOL) != 0) 2469 goto bad; 2470 2471 if (sysctl_createv(clog, 0, &rnode, &cnode, 2472 CTLFLAG_PERMANENT, 2473 CTLTYPE_INT, "len", 2474 SYSCTL_DESCR("Current output queue length"), 2475 NULL, 0, &ifq->ifq_len, 0, 2476 CTL_CREATE, CTL_EOL) != 0) 2477 goto bad; 2478 2479 if (sysctl_createv(clog, 0, &rnode, &cnode, 2480 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2481 CTLTYPE_INT, "maxlen", 2482 SYSCTL_DESCR("Maximum allowed output queue length"), 2483 NULL, 0, &ifq->ifq_maxlen, 0, 2484 CTL_CREATE, CTL_EOL) != 0) 2485 goto bad; 2486 2487 if (sysctl_createv(clog, 0, &rnode, &cnode, 2488 CTLFLAG_PERMANENT, 2489 CTLTYPE_INT, "drops", 2490 SYSCTL_DESCR("Packets dropped due to full output queue"), 2491 NULL, 0, &ifq->ifq_drops, 0, 2492 CTL_CREATE, CTL_EOL) != 0) 2493 goto bad; 2494 2495 return; 2496 bad: 2497 printf("%s: could not attach sysctl nodes\n", ifname); 2498 return; 2499 } 2500 2501 #if defined(INET) || defined(INET6) 2502 2503 #define SYSCTL_NET_PKTQ(q, cn, c) \ 2504 static int \ 2505 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 2506 { \ 2507 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 2508 } 2509 2510 #if defined(INET) 2511 static int 2512 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 2513 { 2514 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 2515 } 2516 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 2517 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 2518 #endif 2519 2520 #if defined(INET6) 2521 static int 2522 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 2523 { 2524 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 2525 } 2526 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 2527 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 2528 #endif 2529 2530 static void 2531 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 2532 { 2533 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 2534 const char *pfname = NULL, *ipname = NULL; 2535 int ipn = 0, qid = 0; 2536 2537 switch (pf) { 2538 #if defined(INET) 2539 case PF_INET: 2540 len_func = sysctl_net_ip_pktq_items; 2541 maxlen_func = sysctl_net_ip_pktq_maxlen; 2542 drops_func = sysctl_net_ip_pktq_drops; 2543 pfname = "inet", ipn = IPPROTO_IP; 2544 ipname = "ip", qid = IPCTL_IFQ; 2545 break; 2546 #endif 2547 #if defined(INET6) 2548 case PF_INET6: 2549 len_func = sysctl_net_ip6_pktq_items; 2550 maxlen_func = sysctl_net_ip6_pktq_maxlen; 2551 drops_func = sysctl_net_ip6_pktq_drops; 2552 pfname = "inet6", ipn = IPPROTO_IPV6; 2553 ipname = "ip6", qid = IPV6CTL_IFQ; 2554 break; 2555 #endif 2556 default: 2557 KASSERT(false); 2558 } 2559 2560 sysctl_createv(clog, 0, NULL, NULL, 2561 CTLFLAG_PERMANENT, 2562 CTLTYPE_NODE, pfname, NULL, 2563 NULL, 0, NULL, 0, 2564 CTL_NET, pf, CTL_EOL); 2565 sysctl_createv(clog, 0, NULL, NULL, 2566 CTLFLAG_PERMANENT, 2567 CTLTYPE_NODE, ipname, NULL, 2568 NULL, 0, NULL, 0, 2569 CTL_NET, pf, ipn, CTL_EOL); 2570 sysctl_createv(clog, 0, NULL, NULL, 2571 CTLFLAG_PERMANENT, 2572 CTLTYPE_NODE, "ifq", 2573 SYSCTL_DESCR("Protocol input queue controls"), 2574 NULL, 0, NULL, 0, 2575 CTL_NET, pf, ipn, qid, CTL_EOL); 2576 2577 sysctl_createv(clog, 0, NULL, NULL, 2578 CTLFLAG_PERMANENT, 2579 CTLTYPE_INT, "len", 2580 SYSCTL_DESCR("Current input queue length"), 2581 len_func, 0, NULL, 0, 2582 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 2583 sysctl_createv(clog, 0, NULL, NULL, 2584 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2585 CTLTYPE_INT, "maxlen", 2586 SYSCTL_DESCR("Maximum allowed input queue length"), 2587 maxlen_func, 0, NULL, 0, 2588 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 2589 sysctl_createv(clog, 0, NULL, NULL, 2590 CTLFLAG_PERMANENT, 2591 CTLTYPE_INT, "drops", 2592 SYSCTL_DESCR("Packets dropped due to full input queue"), 2593 drops_func, 0, NULL, 0, 2594 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 2595 } 2596 #endif /* INET || INET6 */ 2597 2598 static int 2599 if_sdl_sysctl(SYSCTLFN_ARGS) 2600 { 2601 struct ifnet *ifp; 2602 const struct sockaddr_dl *sdl; 2603 2604 if (namelen != 1) 2605 return EINVAL; 2606 2607 ifp = if_byindex(name[0]); 2608 if (ifp == NULL) 2609 return ENODEV; 2610 2611 sdl = ifp->if_sadl; 2612 if (sdl == NULL) { 2613 *oldlenp = 0; 2614 return 0; 2615 } 2616 2617 if (oldp == NULL) { 2618 *oldlenp = sdl->sdl_alen; 2619 return 0; 2620 } 2621 2622 if (*oldlenp >= sdl->sdl_alen) 2623 *oldlenp = sdl->sdl_alen; 2624 return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 2625 } 2626 2627 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 2628 { 2629 const struct sysctlnode *rnode = NULL; 2630 2631 sysctl_createv(clog, 0, NULL, &rnode, 2632 CTLFLAG_PERMANENT, 2633 CTLTYPE_NODE, "sdl", 2634 SYSCTL_DESCR("Get active link-layer address"), 2635 if_sdl_sysctl, 0, NULL, 0, 2636 CTL_NET, CTL_CREATE, CTL_EOL); 2637 } 2638
Indexes created Fri Sep 26 20:09:58 GMT 2025