if.c revision 1.293
1 /* $NetBSD: if.c,v 1.293 2014/11/17 13:58:53 pooka 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 1.293 2014/11/17 13:58:53 pooka Exp $"); 94 95 #include "opt_inet.h" 96 97 #include "opt_atalk.h" 98 #include "opt_natm.h" 99 #include "opt_wlan.h" 100 101 #include <sys/param.h> 102 #include <sys/mbuf.h> 103 #include <sys/systm.h> 104 #include <sys/callout.h> 105 #include <sys/proc.h> 106 #include <sys/socket.h> 107 #include <sys/socketvar.h> 108 #include <sys/domain.h> 109 #include <sys/protosw.h> 110 #include <sys/kernel.h> 111 #include <sys/ioctl.h> 112 #include <sys/sysctl.h> 113 #include <sys/syslog.h> 114 #include <sys/kauth.h> 115 #include <sys/kmem.h> 116 #include <sys/xcall.h> 117 118 #include <net/if.h> 119 #include <net/if_dl.h> 120 #include <net/if_ether.h> 121 #include <net/if_media.h> 122 #include <net80211/ieee80211.h> 123 #include <net80211/ieee80211_ioctl.h> 124 #include <net/if_types.h> 125 #include <net/radix.h> 126 #include <net/route.h> 127 #include <net/netisr.h> 128 #include <sys/module.h> 129 #ifdef NETATALK 130 #include <netatalk/at_extern.h> 131 #include <netatalk/at.h> 132 #endif 133 #include <net/pfil.h> 134 #include <netinet/in.h> 135 #include <netinet/in_var.h> 136 137 #ifdef INET6 138 #include <netinet6/in6_var.h> 139 #include <netinet6/nd6.h> 140 #endif 141 142 #include "ether.h" 143 #include "fddi.h" 144 #include "token.h" 145 146 #include "carp.h" 147 #if NCARP > 0 148 #include <netinet/ip_carp.h> 149 #endif 150 151 #include <compat/sys/sockio.h> 152 #include <compat/sys/socket.h> 153 154 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 155 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 156 157 /* 158 * Global list of interfaces. 159 */ 160 struct ifnet_head ifnet_list; 161 static ifnet_t ** ifindex2ifnet = NULL; 162 163 static u_int if_index = 1; 164 static size_t if_indexlim = 0; 165 static uint64_t index_gen; 166 static kmutex_t index_gen_mtx; 167 static kmutex_t if_clone_mtx; 168 169 static struct ifaddr ** ifnet_addrs = NULL; 170 171 static callout_t if_slowtimo_ch; 172 173 struct ifnet *lo0ifp; 174 int ifqmaxlen = IFQ_MAXLEN; 175 176 static int if_rt_walktree(struct rtentry *, void *); 177 178 static struct if_clone *if_clone_lookup(const char *, int *); 179 static int if_clone_list(struct if_clonereq *); 180 181 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 182 static int if_cloners_count; 183 184 /* Packet filtering hook for interfaces. */ 185 pfil_head_t * if_pfil; 186 187 static kauth_listener_t if_listener; 188 189 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 190 static int ifioctl_attach(struct ifnet *); 191 static void ifioctl_detach(struct ifnet *); 192 static void ifnet_lock_enter(struct ifnet_lock *); 193 static void ifnet_lock_exit(struct ifnet_lock *); 194 static void if_detach_queues(struct ifnet *, struct ifqueue *); 195 static void sysctl_sndq_setup(struct sysctllog **, const char *, 196 struct ifaltq *); 197 198 #if defined(INET) || defined(INET6) 199 static void sysctl_net_pktq_setup(struct sysctllog **, int); 200 #endif 201 202 static int 203 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 204 void *arg0, void *arg1, void *arg2, void *arg3) 205 { 206 int result; 207 enum kauth_network_req req; 208 209 result = KAUTH_RESULT_DEFER; 210 req = (enum kauth_network_req)arg1; 211 212 if (action != KAUTH_NETWORK_INTERFACE) 213 return result; 214 215 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 216 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 217 result = KAUTH_RESULT_ALLOW; 218 219 return result; 220 } 221 222 /* 223 * Network interface utility routines. 224 * 225 * Routines with ifa_ifwith* names take sockaddr *'s as 226 * parameters. 227 */ 228 void 229 ifinit(void) 230 { 231 #if defined(INET) 232 sysctl_net_pktq_setup(NULL, PF_INET); 233 #endif 234 #ifdef INET6 235 sysctl_net_pktq_setup(NULL, PF_INET6); 236 #endif 237 238 callout_init(&if_slowtimo_ch, 0); 239 if_slowtimo(NULL); 240 241 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 242 if_listener_cb, NULL); 243 244 /* interfaces are available, inform socket code */ 245 ifioctl = doifioctl; 246 } 247 248 /* 249 * XXX Initialization before configure(). 250 * XXX hack to get pfil_add_hook working in autoconf. 251 */ 252 void 253 ifinit1(void) 254 { 255 mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE); 256 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 257 TAILQ_INIT(&ifnet_list); 258 if_indexlim = 8; 259 260 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 261 KASSERT(if_pfil != NULL); 262 263 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 264 etherinit(); 265 #endif 266 } 267 268 ifnet_t * 269 if_alloc(u_char type) 270 { 271 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 272 } 273 274 void 275 if_free(ifnet_t *ifp) 276 { 277 kmem_free(ifp, sizeof(ifnet_t)); 278 } 279 280 void 281 if_initname(struct ifnet *ifp, const char *name, int unit) 282 { 283 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 284 "%s%d", name, unit); 285 } 286 287 /* 288 * Null routines used while an interface is going away. These routines 289 * just return an error. 290 */ 291 292 int 293 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 294 const struct sockaddr *so, struct rtentry *rt) 295 { 296 297 return ENXIO; 298 } 299 300 void 301 if_nullinput(struct ifnet *ifp, struct mbuf *m) 302 { 303 304 /* Nothing. */ 305 } 306 307 void 308 if_nullstart(struct ifnet *ifp) 309 { 310 311 /* Nothing. */ 312 } 313 314 int 315 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 316 { 317 318 /* Wake ifioctl_detach(), who may wait for all threads to 319 * quit the critical section. 320 */ 321 cv_signal(&ifp->if_ioctl_lock->il_emptied); 322 return ENXIO; 323 } 324 325 int 326 if_nullinit(struct ifnet *ifp) 327 { 328 329 return ENXIO; 330 } 331 332 void 333 if_nullstop(struct ifnet *ifp, int disable) 334 { 335 336 /* Nothing. */ 337 } 338 339 void 340 if_nullwatchdog(struct ifnet *ifp) 341 { 342 343 /* Nothing. */ 344 } 345 346 void 347 if_nulldrain(struct ifnet *ifp) 348 { 349 350 /* Nothing. */ 351 } 352 353 void 354 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 355 { 356 struct ifaddr *ifa; 357 struct sockaddr_dl *sdl; 358 359 ifp->if_addrlen = addrlen; 360 if_alloc_sadl(ifp); 361 ifa = ifp->if_dl; 362 sdl = satosdl(ifa->ifa_addr); 363 364 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 365 if (factory) { 366 ifp->if_hwdl = ifp->if_dl; 367 ifaref(ifp->if_hwdl); 368 } 369 /* TBD routing socket */ 370 } 371 372 struct ifaddr * 373 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 374 { 375 unsigned socksize, ifasize; 376 int addrlen, namelen; 377 struct sockaddr_dl *mask, *sdl; 378 struct ifaddr *ifa; 379 380 namelen = strlen(ifp->if_xname); 381 addrlen = ifp->if_addrlen; 382 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 383 ifasize = sizeof(*ifa) + 2 * socksize; 384 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 385 386 sdl = (struct sockaddr_dl *)(ifa + 1); 387 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 388 389 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 390 ifp->if_xname, namelen, NULL, addrlen); 391 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 392 memset(&mask->sdl_data[0], 0xff, namelen); 393 ifa->ifa_rtrequest = link_rtrequest; 394 ifa->ifa_addr = (struct sockaddr *)sdl; 395 ifa->ifa_netmask = (struct sockaddr *)mask; 396 397 *sdlp = sdl; 398 399 return ifa; 400 } 401 402 static void 403 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 404 { 405 const struct sockaddr_dl *sdl; 406 ifnet_addrs[ifp->if_index] = ifa; 407 ifaref(ifa); 408 ifp->if_dl = ifa; 409 ifaref(ifa); 410 sdl = satosdl(ifa->ifa_addr); 411 ifp->if_sadl = sdl; 412 } 413 414 /* 415 * Allocate the link level name for the specified interface. This 416 * is an attachment helper. It must be called after ifp->if_addrlen 417 * is initialized, which may not be the case when if_attach() is 418 * called. 419 */ 420 void 421 if_alloc_sadl(struct ifnet *ifp) 422 { 423 struct ifaddr *ifa; 424 const struct sockaddr_dl *sdl; 425 426 /* 427 * If the interface already has a link name, release it 428 * now. This is useful for interfaces that can change 429 * link types, and thus switch link names often. 430 */ 431 if (ifp->if_sadl != NULL) 432 if_free_sadl(ifp); 433 434 ifa = if_dl_create(ifp, &sdl); 435 436 ifa_insert(ifp, ifa); 437 if_sadl_setrefs(ifp, ifa); 438 } 439 440 static void 441 if_deactivate_sadl(struct ifnet *ifp) 442 { 443 struct ifaddr *ifa; 444 445 KASSERT(ifp->if_dl != NULL); 446 447 ifa = ifp->if_dl; 448 449 ifp->if_sadl = NULL; 450 451 ifnet_addrs[ifp->if_index] = NULL; 452 ifafree(ifa); 453 ifp->if_dl = NULL; 454 ifafree(ifa); 455 } 456 457 void 458 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 459 const struct sockaddr_dl *sdl) 460 { 461 int s; 462 463 s = splnet(); 464 465 if_deactivate_sadl(ifp); 466 467 if_sadl_setrefs(ifp, ifa); 468 IFADDR_FOREACH(ifa, ifp) 469 rtinit(ifa, RTM_LLINFO_UPD, 0); 470 splx(s); 471 } 472 473 /* 474 * Free the link level name for the specified interface. This is 475 * a detach helper. This is called from if_detach() or from 476 * link layer type specific detach functions. 477 */ 478 void 479 if_free_sadl(struct ifnet *ifp) 480 { 481 struct ifaddr *ifa; 482 int s; 483 484 ifa = ifnet_addrs[ifp->if_index]; 485 if (ifa == NULL) { 486 KASSERT(ifp->if_sadl == NULL); 487 KASSERT(ifp->if_dl == NULL); 488 return; 489 } 490 491 KASSERT(ifp->if_sadl != NULL); 492 KASSERT(ifp->if_dl != NULL); 493 494 s = splnet(); 495 rtinit(ifa, RTM_DELETE, 0); 496 ifa_remove(ifp, ifa); 497 if_deactivate_sadl(ifp); 498 if (ifp->if_hwdl == ifa) { 499 ifafree(ifa); 500 ifp->if_hwdl = NULL; 501 } 502 splx(s); 503 } 504 505 static void 506 if_getindex(ifnet_t *ifp) 507 { 508 bool hitlimit = false; 509 510 mutex_enter(&index_gen_mtx); 511 ifp->if_index_gen = index_gen++; 512 mutex_exit(&index_gen_mtx); 513 514 ifp->if_index = if_index; 515 if (ifindex2ifnet == NULL) { 516 if_index++; 517 goto skip; 518 } 519 while (if_byindex(ifp->if_index)) { 520 /* 521 * If we hit USHRT_MAX, we skip back to 0 since 522 * there are a number of places where the value 523 * of if_index or if_index itself is compared 524 * to or stored in an unsigned short. By 525 * jumping back, we won't botch those assignments 526 * or comparisons. 527 */ 528 if (++if_index == 0) { 529 if_index = 1; 530 } else if (if_index == USHRT_MAX) { 531 /* 532 * However, if we have to jump back to 533 * zero *twice* without finding an empty 534 * slot in ifindex2ifnet[], then there 535 * there are too many (>65535) interfaces. 536 */ 537 if (hitlimit) { 538 panic("too many interfaces"); 539 } 540 hitlimit = true; 541 if_index = 1; 542 } 543 ifp->if_index = if_index; 544 } 545 skip: 546 /* 547 * We have some arrays that should be indexed by if_index. 548 * since if_index will grow dynamically, they should grow too. 549 * struct ifadd **ifnet_addrs 550 * struct ifnet **ifindex2ifnet 551 */ 552 if (ifnet_addrs == NULL || ifindex2ifnet == NULL || 553 ifp->if_index >= if_indexlim) { 554 size_t m, n, oldlim; 555 void *q; 556 557 oldlim = if_indexlim; 558 while (ifp->if_index >= if_indexlim) 559 if_indexlim <<= 1; 560 561 /* grow ifnet_addrs */ 562 m = oldlim * sizeof(struct ifaddr *); 563 n = if_indexlim * sizeof(struct ifaddr *); 564 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 565 if (ifnet_addrs != NULL) { 566 memcpy(q, ifnet_addrs, m); 567 free(ifnet_addrs, M_IFADDR); 568 } 569 ifnet_addrs = (struct ifaddr **)q; 570 571 /* grow ifindex2ifnet */ 572 m = oldlim * sizeof(struct ifnet *); 573 n = if_indexlim * sizeof(struct ifnet *); 574 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 575 if (ifindex2ifnet != NULL) { 576 memcpy(q, ifindex2ifnet, m); 577 free(ifindex2ifnet, M_IFADDR); 578 } 579 ifindex2ifnet = (struct ifnet **)q; 580 } 581 ifindex2ifnet[ifp->if_index] = ifp; 582 } 583 584 /* 585 * Attach an interface to the list of "active" interfaces. 586 */ 587 void 588 if_attach(ifnet_t *ifp) 589 { 590 KASSERT(if_indexlim > 0); 591 TAILQ_INIT(&ifp->if_addrlist); 592 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 593 594 if (ifioctl_attach(ifp) != 0) 595 panic("%s: ifioctl_attach() failed", __func__); 596 597 if_getindex(ifp); 598 599 /* 600 * Link level name is allocated later by a separate call to 601 * if_alloc_sadl(). 602 */ 603 604 if (ifp->if_snd.ifq_maxlen == 0) 605 ifp->if_snd.ifq_maxlen = ifqmaxlen; 606 607 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 608 609 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 610 611 ifp->if_link_state = LINK_STATE_UNKNOWN; 612 613 ifp->if_capenable = 0; 614 ifp->if_csum_flags_tx = 0; 615 ifp->if_csum_flags_rx = 0; 616 617 #ifdef ALTQ 618 ifp->if_snd.altq_type = 0; 619 ifp->if_snd.altq_disc = NULL; 620 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 621 ifp->if_snd.altq_tbr = NULL; 622 ifp->if_snd.altq_ifp = ifp; 623 #endif 624 625 #ifdef NET_MPSAFE 626 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 627 #else 628 ifp->if_snd.ifq_lock = NULL; 629 #endif 630 631 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 632 (void)pfil_run_hooks(if_pfil, 633 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 634 635 if (!STAILQ_EMPTY(&domains)) 636 if_attachdomain1(ifp); 637 638 /* Announce the interface. */ 639 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 640 } 641 642 void 643 if_attachdomain(void) 644 { 645 struct ifnet *ifp; 646 int s; 647 648 s = splnet(); 649 IFNET_FOREACH(ifp) 650 if_attachdomain1(ifp); 651 splx(s); 652 } 653 654 void 655 if_attachdomain1(struct ifnet *ifp) 656 { 657 struct domain *dp; 658 int s; 659 660 s = splnet(); 661 662 /* address family dependent data region */ 663 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 664 DOMAIN_FOREACH(dp) { 665 if (dp->dom_ifattach != NULL) 666 ifp->if_afdata[dp->dom_family] = 667 (*dp->dom_ifattach)(ifp); 668 } 669 670 splx(s); 671 } 672 673 /* 674 * Deactivate an interface. This points all of the procedure 675 * handles at error stubs. May be called from interrupt context. 676 */ 677 void 678 if_deactivate(struct ifnet *ifp) 679 { 680 int s; 681 682 s = splnet(); 683 684 ifp->if_output = if_nulloutput; 685 ifp->if_input = if_nullinput; 686 ifp->if_start = if_nullstart; 687 ifp->if_ioctl = if_nullioctl; 688 ifp->if_init = if_nullinit; 689 ifp->if_stop = if_nullstop; 690 ifp->if_watchdog = if_nullwatchdog; 691 ifp->if_drain = if_nulldrain; 692 693 /* No more packets may be enqueued. */ 694 ifp->if_snd.ifq_maxlen = 0; 695 696 splx(s); 697 } 698 699 void 700 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 701 { 702 struct ifaddr *ifa, *nifa; 703 704 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 705 if (ifa->ifa_addr->sa_family != family) 706 continue; 707 (*purgeaddr)(ifa); 708 } 709 } 710 711 /* 712 * Detach an interface from the list of "active" interfaces, 713 * freeing any resources as we go along. 714 * 715 * NOTE: This routine must be called with a valid thread context, 716 * as it may block. 717 */ 718 void 719 if_detach(struct ifnet *ifp) 720 { 721 struct socket so; 722 struct ifaddr *ifa; 723 #ifdef IFAREF_DEBUG 724 struct ifaddr *last_ifa = NULL; 725 #endif 726 struct domain *dp; 727 const struct protosw *pr; 728 int s, i, family, purged; 729 uint64_t xc; 730 731 /* 732 * XXX It's kind of lame that we have to have the 733 * XXX socket structure... 734 */ 735 memset(&so, 0, sizeof(so)); 736 737 s = splnet(); 738 739 /* 740 * Do an if_down() to give protocols a chance to do something. 741 */ 742 if_down(ifp); 743 744 #ifdef ALTQ 745 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 746 altq_disable(&ifp->if_snd); 747 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 748 altq_detach(&ifp->if_snd); 749 #endif 750 751 if (ifp->if_snd.ifq_lock) 752 mutex_obj_free(ifp->if_snd.ifq_lock); 753 754 sysctl_teardown(&ifp->if_sysctl_log); 755 756 #if NCARP > 0 757 /* Remove the interface from any carp group it is a part of. */ 758 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 759 carp_ifdetach(ifp); 760 #endif 761 762 /* 763 * Rip all the addresses off the interface. This should make 764 * all of the routes go away. 765 * 766 * pr_usrreq calls can remove an arbitrary number of ifaddrs 767 * from the list, including our "cursor", ifa. For safety, 768 * and to honor the TAILQ abstraction, I just restart the 769 * loop after each removal. Note that the loop will exit 770 * when all of the remaining ifaddrs belong to the AF_LINK 771 * family. I am counting on the historical fact that at 772 * least one pr_usrreq in each address domain removes at 773 * least one ifaddr. 774 */ 775 again: 776 IFADDR_FOREACH(ifa, ifp) { 777 family = ifa->ifa_addr->sa_family; 778 #ifdef IFAREF_DEBUG 779 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 780 ifa, family, ifa->ifa_refcnt); 781 if (last_ifa != NULL && ifa == last_ifa) 782 panic("if_detach: loop detected"); 783 last_ifa = ifa; 784 #endif 785 if (family == AF_LINK) 786 continue; 787 dp = pffinddomain(family); 788 #ifdef DIAGNOSTIC 789 if (dp == NULL) 790 panic("if_detach: no domain for AF %d", 791 family); 792 #endif 793 /* 794 * XXX These PURGEIF calls are redundant with the 795 * purge-all-families calls below, but are left in for 796 * now both to make a smaller change, and to avoid 797 * unplanned interactions with clearing of 798 * ifp->if_addrlist. 799 */ 800 purged = 0; 801 for (pr = dp->dom_protosw; 802 pr < dp->dom_protoswNPROTOSW; pr++) { 803 so.so_proto = pr; 804 if (pr->pr_usrreqs) { 805 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 806 purged = 1; 807 } 808 } 809 if (purged == 0) { 810 /* 811 * XXX What's really the best thing to do 812 * XXX here? --thorpej (at) NetBSD.org 813 */ 814 printf("if_detach: WARNING: AF %d not purged\n", 815 family); 816 ifa_remove(ifp, ifa); 817 } 818 goto again; 819 } 820 821 if_free_sadl(ifp); 822 823 /* Walk the routing table looking for stragglers. */ 824 for (i = 0; i <= AF_MAX; i++) { 825 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 826 continue; 827 } 828 829 DOMAIN_FOREACH(dp) { 830 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 831 { 832 void *p = ifp->if_afdata[dp->dom_family]; 833 if (p) { 834 ifp->if_afdata[dp->dom_family] = NULL; 835 (*dp->dom_ifdetach)(ifp, p); 836 } 837 } 838 839 /* 840 * One would expect multicast memberships (INET and 841 * INET6) on UDP sockets to be purged by the PURGEIF 842 * calls above, but if all addresses were removed from 843 * the interface prior to destruction, the calls will 844 * not be made (e.g. ppp, for which pppd(8) generally 845 * removes addresses before destroying the interface). 846 * Because there is no invariant that multicast 847 * memberships only exist for interfaces with IPv4 848 * addresses, we must call PURGEIF regardless of 849 * addresses. (Protocols which might store ifnet 850 * pointers are marked with PR_PURGEIF.) 851 */ 852 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 853 so.so_proto = pr; 854 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 855 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 856 } 857 } 858 859 (void)pfil_run_hooks(if_pfil, 860 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 861 (void)pfil_head_destroy(ifp->if_pfil); 862 863 /* Announce that the interface is gone. */ 864 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 865 866 ifindex2ifnet[ifp->if_index] = NULL; 867 868 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 869 870 ifioctl_detach(ifp); 871 872 /* 873 * remove packets that came from ifp, from software interrupt queues. 874 */ 875 DOMAIN_FOREACH(dp) { 876 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 877 struct ifqueue *iq = dp->dom_ifqueues[i]; 878 if (iq == NULL) 879 break; 880 dp->dom_ifqueues[i] = NULL; 881 if_detach_queues(ifp, iq); 882 } 883 } 884 885 /* 886 * IP queues have to be processed separately: net-queue barrier 887 * ensures that the packets are dequeued while a cross-call will 888 * ensure that the interrupts have completed. FIXME: not quite.. 889 */ 890 #ifdef INET 891 pktq_barrier(ip_pktq); 892 #endif 893 #ifdef INET6 894 if (in6_present) 895 pktq_barrier(ip6_pktq); 896 #endif 897 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 898 xc_wait(xc); 899 900 splx(s); 901 } 902 903 static void 904 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 905 { 906 struct mbuf *m, *prev, *next; 907 908 prev = NULL; 909 for (m = q->ifq_head; m != NULL; m = next) { 910 KASSERT((m->m_flags & M_PKTHDR) != 0); 911 912 next = m->m_nextpkt; 913 if (m->m_pkthdr.rcvif != ifp) { 914 prev = m; 915 continue; 916 } 917 918 if (prev != NULL) 919 prev->m_nextpkt = m->m_nextpkt; 920 else 921 q->ifq_head = m->m_nextpkt; 922 if (q->ifq_tail == m) 923 q->ifq_tail = prev; 924 q->ifq_len--; 925 926 m->m_nextpkt = NULL; 927 m_freem(m); 928 IF_DROP(q); 929 } 930 } 931 932 /* 933 * Callback for a radix tree walk to delete all references to an 934 * ifnet. 935 */ 936 static int 937 if_rt_walktree(struct rtentry *rt, void *v) 938 { 939 struct ifnet *ifp = (struct ifnet *)v; 940 int error; 941 942 if (rt->rt_ifp != ifp) 943 return 0; 944 945 /* Delete the entry. */ 946 ++rt->rt_refcnt; 947 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 948 rt_mask(rt), rt->rt_flags, NULL); 949 KASSERT((rt->rt_flags & RTF_UP) == 0); 950 rt->rt_ifp = NULL; 951 rtfree(rt); 952 if (error != 0) 953 printf("%s: warning: unable to delete rtentry @ %p, " 954 "error = %d\n", ifp->if_xname, rt, error); 955 return ERESTART; 956 } 957 958 /* 959 * Create a clone network interface. 960 */ 961 int 962 if_clone_create(const char *name) 963 { 964 struct if_clone *ifc; 965 int unit; 966 967 ifc = if_clone_lookup(name, &unit); 968 if (ifc == NULL) 969 return EINVAL; 970 971 if (ifunit(name) != NULL) 972 return EEXIST; 973 974 return (*ifc->ifc_create)(ifc, unit); 975 } 976 977 /* 978 * Destroy a clone network interface. 979 */ 980 int 981 if_clone_destroy(const char *name) 982 { 983 struct if_clone *ifc; 984 struct ifnet *ifp; 985 986 ifc = if_clone_lookup(name, NULL); 987 if (ifc == NULL) 988 return EINVAL; 989 990 ifp = ifunit(name); 991 if (ifp == NULL) 992 return ENXIO; 993 994 if (ifc->ifc_destroy == NULL) 995 return EOPNOTSUPP; 996 997 return (*ifc->ifc_destroy)(ifp); 998 } 999 1000 /* 1001 * Look up a network interface cloner. 1002 */ 1003 static struct if_clone * 1004 if_clone_lookup(const char *name, int *unitp) 1005 { 1006 struct if_clone *ifc; 1007 const char *cp; 1008 char *dp, ifname[IFNAMSIZ + 3]; 1009 int unit; 1010 1011 strcpy(ifname, "if_"); 1012 /* separate interface name from unit */ 1013 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1014 *cp && (*cp < '0' || *cp > '9');) 1015 *dp++ = *cp++; 1016 1017 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1018 return NULL; /* No name or unit number */ 1019 *dp++ = '\0'; 1020 1021 again: 1022 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1023 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1024 break; 1025 } 1026 1027 if (ifc == NULL) { 1028 if (*ifname == '\0' || 1029 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1030 return NULL; 1031 *ifname = '\0'; 1032 goto again; 1033 } 1034 1035 unit = 0; 1036 while (cp - name < IFNAMSIZ && *cp) { 1037 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1038 /* Bogus unit number. */ 1039 return NULL; 1040 } 1041 unit = (unit * 10) + (*cp++ - '0'); 1042 } 1043 1044 if (unitp != NULL) 1045 *unitp = unit; 1046 return ifc; 1047 } 1048 1049 /* 1050 * Register a network interface cloner. 1051 */ 1052 void 1053 if_clone_attach(struct if_clone *ifc) 1054 { 1055 1056 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1057 if_cloners_count++; 1058 } 1059 1060 /* 1061 * Unregister a network interface cloner. 1062 */ 1063 void 1064 if_clone_detach(struct if_clone *ifc) 1065 { 1066 1067 LIST_REMOVE(ifc, ifc_list); 1068 if_cloners_count--; 1069 } 1070 1071 /* 1072 * Provide list of interface cloners to userspace. 1073 */ 1074 static int 1075 if_clone_list(struct if_clonereq *ifcr) 1076 { 1077 char outbuf[IFNAMSIZ], *dst; 1078 struct if_clone *ifc; 1079 int count, error = 0; 1080 1081 ifcr->ifcr_total = if_cloners_count; 1082 if ((dst = ifcr->ifcr_buffer) == NULL) { 1083 /* Just asking how many there are. */ 1084 return 0; 1085 } 1086 1087 if (ifcr->ifcr_count < 0) 1088 return EINVAL; 1089 1090 count = (if_cloners_count < ifcr->ifcr_count) ? 1091 if_cloners_count : ifcr->ifcr_count; 1092 1093 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1094 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1095 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1096 if (outbuf[sizeof(outbuf) - 1] != '\0') 1097 return ENAMETOOLONG; 1098 error = copyout(outbuf, dst, sizeof(outbuf)); 1099 if (error != 0) 1100 break; 1101 } 1102 1103 return error; 1104 } 1105 1106 void 1107 ifaref(struct ifaddr *ifa) 1108 { 1109 ifa->ifa_refcnt++; 1110 } 1111 1112 void 1113 ifafree(struct ifaddr *ifa) 1114 { 1115 KASSERT(ifa != NULL); 1116 KASSERT(ifa->ifa_refcnt > 0); 1117 1118 if (--ifa->ifa_refcnt == 0) { 1119 free(ifa, M_IFADDR); 1120 } 1121 } 1122 1123 void 1124 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1125 { 1126 ifa->ifa_ifp = ifp; 1127 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1128 ifaref(ifa); 1129 } 1130 1131 void 1132 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1133 { 1134 KASSERT(ifa->ifa_ifp == ifp); 1135 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1136 ifafree(ifa); 1137 } 1138 1139 static inline int 1140 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1141 { 1142 return sockaddr_cmp(sa1, sa2) == 0; 1143 } 1144 1145 /* 1146 * Locate an interface based on a complete address. 1147 */ 1148 /*ARGSUSED*/ 1149 struct ifaddr * 1150 ifa_ifwithaddr(const struct sockaddr *addr) 1151 { 1152 struct ifnet *ifp; 1153 struct ifaddr *ifa; 1154 1155 IFNET_FOREACH(ifp) { 1156 if (ifp->if_output == if_nulloutput) 1157 continue; 1158 IFADDR_FOREACH(ifa, ifp) { 1159 if (ifa->ifa_addr->sa_family != addr->sa_family) 1160 continue; 1161 if (equal(addr, ifa->ifa_addr)) 1162 return ifa; 1163 if ((ifp->if_flags & IFF_BROADCAST) && 1164 ifa->ifa_broadaddr && 1165 /* IP6 doesn't have broadcast */ 1166 ifa->ifa_broadaddr->sa_len != 0 && 1167 equal(ifa->ifa_broadaddr, addr)) 1168 return ifa; 1169 } 1170 } 1171 return NULL; 1172 } 1173 1174 /* 1175 * Locate the point to point interface with a given destination address. 1176 */ 1177 /*ARGSUSED*/ 1178 struct ifaddr * 1179 ifa_ifwithdstaddr(const struct sockaddr *addr) 1180 { 1181 struct ifnet *ifp; 1182 struct ifaddr *ifa; 1183 1184 IFNET_FOREACH(ifp) { 1185 if (ifp->if_output == if_nulloutput) 1186 continue; 1187 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1188 continue; 1189 IFADDR_FOREACH(ifa, ifp) { 1190 if (ifa->ifa_addr->sa_family != addr->sa_family || 1191 ifa->ifa_dstaddr == NULL) 1192 continue; 1193 if (equal(addr, ifa->ifa_dstaddr)) 1194 return ifa; 1195 } 1196 } 1197 return NULL; 1198 } 1199 1200 /* 1201 * Find an interface on a specific network. If many, choice 1202 * is most specific found. 1203 */ 1204 struct ifaddr * 1205 ifa_ifwithnet(const struct sockaddr *addr) 1206 { 1207 struct ifnet *ifp; 1208 struct ifaddr *ifa; 1209 const struct sockaddr_dl *sdl; 1210 struct ifaddr *ifa_maybe = 0; 1211 u_int af = addr->sa_family; 1212 const char *addr_data = addr->sa_data, *cplim; 1213 1214 if (af == AF_LINK) { 1215 sdl = satocsdl(addr); 1216 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1217 ifindex2ifnet[sdl->sdl_index] && 1218 ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) 1219 return ifnet_addrs[sdl->sdl_index]; 1220 } 1221 #ifdef NETATALK 1222 if (af == AF_APPLETALK) { 1223 const struct sockaddr_at *sat, *sat2; 1224 sat = (const struct sockaddr_at *)addr; 1225 IFNET_FOREACH(ifp) { 1226 if (ifp->if_output == if_nulloutput) 1227 continue; 1228 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1229 if (ifa == NULL) 1230 continue; 1231 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1232 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1233 return ifa; /* exact match */ 1234 if (ifa_maybe == NULL) { 1235 /* else keep the if with the right range */ 1236 ifa_maybe = ifa; 1237 } 1238 } 1239 return ifa_maybe; 1240 } 1241 #endif 1242 IFNET_FOREACH(ifp) { 1243 if (ifp->if_output == if_nulloutput) 1244 continue; 1245 IFADDR_FOREACH(ifa, ifp) { 1246 const char *cp, *cp2, *cp3; 1247 1248 if (ifa->ifa_addr->sa_family != af || 1249 ifa->ifa_netmask == NULL) 1250 next: continue; 1251 cp = addr_data; 1252 cp2 = ifa->ifa_addr->sa_data; 1253 cp3 = ifa->ifa_netmask->sa_data; 1254 cplim = (const char *)ifa->ifa_netmask + 1255 ifa->ifa_netmask->sa_len; 1256 while (cp3 < cplim) { 1257 if ((*cp++ ^ *cp2++) & *cp3++) { 1258 /* want to continue for() loop */ 1259 goto next; 1260 } 1261 } 1262 if (ifa_maybe == NULL || 1263 rn_refines((void *)ifa->ifa_netmask, 1264 (void *)ifa_maybe->ifa_netmask)) 1265 ifa_maybe = ifa; 1266 } 1267 } 1268 return ifa_maybe; 1269 } 1270 1271 /* 1272 * Find the interface of the addresss. 1273 */ 1274 struct ifaddr * 1275 ifa_ifwithladdr(const struct sockaddr *addr) 1276 { 1277 struct ifaddr *ia; 1278 1279 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1280 (ia = ifa_ifwithnet(addr))) 1281 return ia; 1282 return NULL; 1283 } 1284 1285 /* 1286 * Find an interface using a specific address family 1287 */ 1288 struct ifaddr * 1289 ifa_ifwithaf(int af) 1290 { 1291 struct ifnet *ifp; 1292 struct ifaddr *ifa; 1293 1294 IFNET_FOREACH(ifp) { 1295 if (ifp->if_output == if_nulloutput) 1296 continue; 1297 IFADDR_FOREACH(ifa, ifp) { 1298 if (ifa->ifa_addr->sa_family == af) 1299 return ifa; 1300 } 1301 } 1302 return NULL; 1303 } 1304 1305 /* 1306 * Find an interface address specific to an interface best matching 1307 * a given address. 1308 */ 1309 struct ifaddr * 1310 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1311 { 1312 struct ifaddr *ifa; 1313 const char *cp, *cp2, *cp3; 1314 const char *cplim; 1315 struct ifaddr *ifa_maybe = 0; 1316 u_int af = addr->sa_family; 1317 1318 if (ifp->if_output == if_nulloutput) 1319 return NULL; 1320 1321 if (af >= AF_MAX) 1322 return NULL; 1323 1324 IFADDR_FOREACH(ifa, ifp) { 1325 if (ifa->ifa_addr->sa_family != af) 1326 continue; 1327 ifa_maybe = ifa; 1328 if (ifa->ifa_netmask == NULL) { 1329 if (equal(addr, ifa->ifa_addr) || 1330 (ifa->ifa_dstaddr && 1331 equal(addr, ifa->ifa_dstaddr))) 1332 return ifa; 1333 continue; 1334 } 1335 cp = addr->sa_data; 1336 cp2 = ifa->ifa_addr->sa_data; 1337 cp3 = ifa->ifa_netmask->sa_data; 1338 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1339 for (; cp3 < cplim; cp3++) { 1340 if ((*cp++ ^ *cp2++) & *cp3) 1341 break; 1342 } 1343 if (cp3 == cplim) 1344 return ifa; 1345 } 1346 return ifa_maybe; 1347 } 1348 1349 /* 1350 * Default action when installing a route with a Link Level gateway. 1351 * Lookup an appropriate real ifa to point to. 1352 * This should be moved to /sys/net/link.c eventually. 1353 */ 1354 void 1355 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1356 { 1357 struct ifaddr *ifa; 1358 const struct sockaddr *dst; 1359 struct ifnet *ifp; 1360 1361 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1362 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1363 return; 1364 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1365 rt_replace_ifa(rt, ifa); 1366 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1367 ifa->ifa_rtrequest(cmd, rt, info); 1368 } 1369 } 1370 1371 /* 1372 * Handle a change in the interface link state. 1373 * XXX: We should listen to the routing socket in-kernel rather 1374 * than calling in6_if_link_* functions directly from here. 1375 */ 1376 void 1377 if_link_state_change(struct ifnet *ifp, int link_state) 1378 { 1379 int s; 1380 #if defined(DEBUG) || defined(INET6) 1381 int old_link_state; 1382 #endif 1383 1384 s = splnet(); 1385 if (ifp->if_link_state == link_state) { 1386 splx(s); 1387 return; 1388 } 1389 1390 #if defined(DEBUG) || defined(INET6) 1391 old_link_state = ifp->if_link_state; 1392 #endif 1393 ifp->if_link_state = link_state; 1394 #ifdef DEBUG 1395 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1396 link_state == LINK_STATE_UP ? "UP" : 1397 link_state == LINK_STATE_DOWN ? "DOWN" : 1398 "UNKNOWN", 1399 old_link_state == LINK_STATE_UP ? "UP" : 1400 old_link_state == LINK_STATE_DOWN ? "DOWN" : 1401 "UNKNOWN"); 1402 #endif 1403 1404 #ifdef INET6 1405 /* 1406 * When going from UNKNOWN to UP, we need to mark existing 1407 * IPv6 addresses as tentative and restart DAD as we may have 1408 * erroneously not found a duplicate. 1409 * 1410 * This needs to happen before rt_ifmsg to avoid a race where 1411 * listeners would have an address and expect it to work right 1412 * away. 1413 */ 1414 if (in6_present && link_state == LINK_STATE_UP && 1415 old_link_state == LINK_STATE_UNKNOWN) 1416 in6_if_link_down(ifp); 1417 #endif 1418 1419 /* Notify that the link state has changed. */ 1420 rt_ifmsg(ifp); 1421 1422 #if NCARP > 0 1423 if (ifp->if_carp) 1424 carp_carpdev_state(ifp); 1425 #endif 1426 1427 #ifdef INET6 1428 if (in6_present) { 1429 if (link_state == LINK_STATE_DOWN) 1430 in6_if_link_down(ifp); 1431 else if (link_state == LINK_STATE_UP) 1432 in6_if_link_up(ifp); 1433 } 1434 #endif 1435 1436 splx(s); 1437 } 1438 1439 /* 1440 * Mark an interface down and notify protocols of 1441 * the transition. 1442 * NOTE: must be called at splsoftnet or equivalent. 1443 */ 1444 void 1445 if_down(struct ifnet *ifp) 1446 { 1447 struct ifaddr *ifa; 1448 1449 ifp->if_flags &= ~IFF_UP; 1450 nanotime(&ifp->if_lastchange); 1451 IFADDR_FOREACH(ifa, ifp) 1452 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1453 IFQ_PURGE(&ifp->if_snd); 1454 #if NCARP > 0 1455 if (ifp->if_carp) 1456 carp_carpdev_state(ifp); 1457 #endif 1458 rt_ifmsg(ifp); 1459 #ifdef INET6 1460 if (in6_present) 1461 in6_if_down(ifp); 1462 #endif 1463 } 1464 1465 /* 1466 * Mark an interface up and notify protocols of 1467 * the transition. 1468 * NOTE: must be called at splsoftnet or equivalent. 1469 */ 1470 void 1471 if_up(struct ifnet *ifp) 1472 { 1473 #ifdef notyet 1474 struct ifaddr *ifa; 1475 #endif 1476 1477 ifp->if_flags |= IFF_UP; 1478 nanotime(&ifp->if_lastchange); 1479 #ifdef notyet 1480 /* this has no effect on IP, and will kill all ISO connections XXX */ 1481 IFADDR_FOREACH(ifa, ifp) 1482 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1483 #endif 1484 #if NCARP > 0 1485 if (ifp->if_carp) 1486 carp_carpdev_state(ifp); 1487 #endif 1488 rt_ifmsg(ifp); 1489 #ifdef INET6 1490 if (in6_present) 1491 in6_if_up(ifp); 1492 #endif 1493 } 1494 1495 /* 1496 * Handle interface watchdog timer routines. Called 1497 * from softclock, we decrement timers (if set) and 1498 * call the appropriate interface routine on expiration. 1499 */ 1500 void 1501 if_slowtimo(void *arg) 1502 { 1503 struct ifnet *ifp; 1504 int s = splnet(); 1505 1506 IFNET_FOREACH(ifp) { 1507 if (ifp->if_timer == 0 || --ifp->if_timer) 1508 continue; 1509 if (ifp->if_watchdog != NULL) 1510 (*ifp->if_watchdog)(ifp); 1511 } 1512 splx(s); 1513 callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL); 1514 } 1515 1516 /* 1517 * Set/clear promiscuous mode on interface ifp based on the truth value 1518 * of pswitch. The calls are reference counted so that only the first 1519 * "on" request actually has an effect, as does the final "off" request. 1520 * Results are undefined if the "off" and "on" requests are not matched. 1521 */ 1522 int 1523 ifpromisc(struct ifnet *ifp, int pswitch) 1524 { 1525 int pcount, ret; 1526 short nflags; 1527 1528 pcount = ifp->if_pcount; 1529 if (pswitch) { 1530 /* 1531 * Allow the device to be "placed" into promiscuous 1532 * mode even if it is not configured up. It will 1533 * consult IFF_PROMISC when it is brought up. 1534 */ 1535 if (ifp->if_pcount++ != 0) 1536 return 0; 1537 nflags = ifp->if_flags | IFF_PROMISC; 1538 } else { 1539 if (--ifp->if_pcount > 0) 1540 return 0; 1541 nflags = ifp->if_flags & ~IFF_PROMISC; 1542 } 1543 ret = if_flags_set(ifp, nflags); 1544 /* Restore interface state if not successful. */ 1545 if (ret != 0) { 1546 ifp->if_pcount = pcount; 1547 } 1548 return ret; 1549 } 1550 1551 /* 1552 * Map interface name to 1553 * interface structure pointer. 1554 */ 1555 struct ifnet * 1556 ifunit(const char *name) 1557 { 1558 struct ifnet *ifp; 1559 const char *cp = name; 1560 u_int unit = 0; 1561 u_int i; 1562 1563 /* 1564 * If the entire name is a number, treat it as an ifindex. 1565 */ 1566 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 1567 unit = unit * 10 + (*cp - '0'); 1568 } 1569 1570 /* 1571 * If the number took all of the name, then it's a valid ifindex. 1572 */ 1573 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 1574 if (unit >= if_indexlim) 1575 return NULL; 1576 ifp = ifindex2ifnet[unit]; 1577 if (ifp == NULL || ifp->if_output == if_nulloutput) 1578 return NULL; 1579 return ifp; 1580 } 1581 1582 IFNET_FOREACH(ifp) { 1583 if (ifp->if_output == if_nulloutput) 1584 continue; 1585 if (strcmp(ifp->if_xname, name) == 0) 1586 return ifp; 1587 } 1588 return NULL; 1589 } 1590 1591 ifnet_t * 1592 if_byindex(u_int idx) 1593 { 1594 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 1595 } 1596 1597 /* common */ 1598 int 1599 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 1600 { 1601 int s; 1602 struct ifreq *ifr; 1603 struct ifcapreq *ifcr; 1604 struct ifdatareq *ifdr; 1605 1606 switch (cmd) { 1607 case SIOCSIFCAP: 1608 ifcr = data; 1609 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 1610 return EINVAL; 1611 1612 if (ifcr->ifcr_capenable == ifp->if_capenable) 1613 return 0; 1614 1615 ifp->if_capenable = ifcr->ifcr_capenable; 1616 1617 /* Pre-compute the checksum flags mask. */ 1618 ifp->if_csum_flags_tx = 0; 1619 ifp->if_csum_flags_rx = 0; 1620 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 1621 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 1622 } 1623 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 1624 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 1625 } 1626 1627 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 1628 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 1629 } 1630 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 1631 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 1632 } 1633 1634 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 1635 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 1636 } 1637 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 1638 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 1639 } 1640 1641 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 1642 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 1643 } 1644 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 1645 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 1646 } 1647 1648 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 1649 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 1650 } 1651 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 1652 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 1653 } 1654 if (ifp->if_flags & IFF_UP) 1655 return ENETRESET; 1656 return 0; 1657 case SIOCSIFFLAGS: 1658 ifr = data; 1659 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 1660 s = splnet(); 1661 if_down(ifp); 1662 splx(s); 1663 } 1664 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 1665 s = splnet(); 1666 if_up(ifp); 1667 splx(s); 1668 } 1669 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1670 (ifr->ifr_flags &~ IFF_CANTCHANGE); 1671 break; 1672 case SIOCGIFFLAGS: 1673 ifr = data; 1674 ifr->ifr_flags = ifp->if_flags; 1675 break; 1676 1677 case SIOCGIFMETRIC: 1678 ifr = data; 1679 ifr->ifr_metric = ifp->if_metric; 1680 break; 1681 1682 case SIOCGIFMTU: 1683 ifr = data; 1684 ifr->ifr_mtu = ifp->if_mtu; 1685 break; 1686 1687 case SIOCGIFDLT: 1688 ifr = data; 1689 ifr->ifr_dlt = ifp->if_dlt; 1690 break; 1691 1692 case SIOCGIFCAP: 1693 ifcr = data; 1694 ifcr->ifcr_capabilities = ifp->if_capabilities; 1695 ifcr->ifcr_capenable = ifp->if_capenable; 1696 break; 1697 1698 case SIOCSIFMETRIC: 1699 ifr = data; 1700 ifp->if_metric = ifr->ifr_metric; 1701 break; 1702 1703 case SIOCGIFDATA: 1704 ifdr = data; 1705 ifdr->ifdr_data = ifp->if_data; 1706 break; 1707 1708 case SIOCGIFINDEX: 1709 ifr = data; 1710 ifr->ifr_index = ifp->if_index; 1711 break; 1712 1713 case SIOCZIFDATA: 1714 ifdr = data; 1715 ifdr->ifdr_data = ifp->if_data; 1716 /* 1717 * Assumes that the volatile counters that can be 1718 * zero'ed are at the end of if_data. 1719 */ 1720 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 1721 offsetof(struct if_data, ifi_ipackets)); 1722 /* 1723 * The memset() clears to the bottm of if_data. In the area, 1724 * if_lastchange is included. Please be careful if new entry 1725 * will be added into if_data or rewite this. 1726 * 1727 * And also, update if_lastchnage. 1728 */ 1729 getnanotime(&ifp->if_lastchange); 1730 break; 1731 case SIOCSIFMTU: 1732 ifr = data; 1733 if (ifp->if_mtu == ifr->ifr_mtu) 1734 break; 1735 ifp->if_mtu = ifr->ifr_mtu; 1736 /* 1737 * If the link MTU changed, do network layer specific procedure. 1738 */ 1739 #ifdef INET6 1740 if (in6_present) 1741 nd6_setmtu(ifp); 1742 #endif 1743 return ENETRESET; 1744 default: 1745 return ENOTTY; 1746 } 1747 return 0; 1748 } 1749 1750 int 1751 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 1752 { 1753 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 1754 struct ifaddr *ifa; 1755 const struct sockaddr *any, *sa; 1756 union { 1757 struct sockaddr sa; 1758 struct sockaddr_storage ss; 1759 } u, v; 1760 1761 switch (cmd) { 1762 case SIOCSIFADDRPREF: 1763 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 1764 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1765 NULL) != 0) 1766 return EPERM; 1767 case SIOCGIFADDRPREF: 1768 break; 1769 default: 1770 return EOPNOTSUPP; 1771 } 1772 1773 /* sanity checks */ 1774 if (data == NULL || ifp == NULL) { 1775 panic("invalid argument to %s", __func__); 1776 /*NOTREACHED*/ 1777 } 1778 1779 /* address must be specified on ADD and DELETE */ 1780 sa = sstocsa(&ifap->ifap_addr); 1781 if (sa->sa_family != sofamily(so)) 1782 return EINVAL; 1783 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 1784 return EINVAL; 1785 1786 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 1787 1788 IFADDR_FOREACH(ifa, ifp) { 1789 if (ifa->ifa_addr->sa_family != sa->sa_family) 1790 continue; 1791 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 1792 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 1793 break; 1794 } 1795 if (ifa == NULL) 1796 return EADDRNOTAVAIL; 1797 1798 switch (cmd) { 1799 case SIOCSIFADDRPREF: 1800 ifa->ifa_preference = ifap->ifap_preference; 1801 return 0; 1802 case SIOCGIFADDRPREF: 1803 /* fill in the if_laddrreq structure */ 1804 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 1805 sizeof(ifap->ifap_addr), ifa->ifa_addr); 1806 ifap->ifap_preference = ifa->ifa_preference; 1807 return 0; 1808 default: 1809 return EOPNOTSUPP; 1810 } 1811 } 1812 1813 static void 1814 ifnet_lock_enter(struct ifnet_lock *il) 1815 { 1816 uint64_t *nenter; 1817 1818 /* Before trying to acquire the mutex, increase the count of threads 1819 * who have entered or who wait to enter the critical section. 1820 * Avoid one costly locked memory transaction by keeping a count for 1821 * each CPU. 1822 */ 1823 nenter = percpu_getref(il->il_nenter); 1824 (*nenter)++; 1825 percpu_putref(il->il_nenter); 1826 mutex_enter(&il->il_lock); 1827 } 1828 1829 static void 1830 ifnet_lock_exit(struct ifnet_lock *il) 1831 { 1832 /* Increase the count of threads who have exited the critical 1833 * section. Increase while we still hold the lock. 1834 */ 1835 il->il_nexit++; 1836 mutex_exit(&il->il_lock); 1837 } 1838 1839 /* 1840 * Interface ioctls. 1841 */ 1842 static int 1843 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 1844 { 1845 struct ifnet *ifp; 1846 struct ifreq *ifr; 1847 int error = 0; 1848 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 1849 u_long ocmd = cmd; 1850 #endif 1851 short oif_flags; 1852 #ifdef COMPAT_OIFREQ 1853 struct ifreq ifrb; 1854 struct oifreq *oifr = NULL; 1855 #endif 1856 int r; 1857 1858 switch (cmd) { 1859 #ifdef COMPAT_OIFREQ 1860 case OSIOCGIFCONF: 1861 case OOSIOCGIFCONF: 1862 return compat_ifconf(cmd, data); 1863 #endif 1864 #ifdef COMPAT_OIFDATA 1865 case OSIOCGIFDATA: 1866 case OSIOCZIFDATA: 1867 return compat_ifdatareq(l, cmd, data); 1868 #endif 1869 case SIOCGIFCONF: 1870 return ifconf(cmd, data); 1871 case SIOCINITIFADDR: 1872 return EPERM; 1873 } 1874 1875 #ifdef COMPAT_OIFREQ 1876 cmd = compat_cvtcmd(cmd); 1877 if (cmd != ocmd) { 1878 oifr = data; 1879 data = ifr = &ifrb; 1880 ifreqo2n(oifr, ifr); 1881 } else 1882 #endif 1883 ifr = data; 1884 1885 ifp = ifunit(ifr->ifr_name); 1886 1887 switch (cmd) { 1888 case SIOCIFCREATE: 1889 case SIOCIFDESTROY: 1890 if (l != NULL) { 1891 error = kauth_authorize_network(l->l_cred, 1892 KAUTH_NETWORK_INTERFACE, 1893 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 1894 (void *)cmd, NULL); 1895 if (error != 0) 1896 return error; 1897 } 1898 mutex_enter(&if_clone_mtx); 1899 r = (cmd == SIOCIFCREATE) ? 1900 if_clone_create(ifr->ifr_name) : 1901 if_clone_destroy(ifr->ifr_name); 1902 mutex_exit(&if_clone_mtx); 1903 return r; 1904 1905 case SIOCIFGCLONERS: 1906 return if_clone_list((struct if_clonereq *)data); 1907 } 1908 1909 if (ifp == NULL) 1910 return ENXIO; 1911 1912 switch (cmd) { 1913 case SIOCALIFADDR: 1914 case SIOCDLIFADDR: 1915 case SIOCSIFADDRPREF: 1916 case SIOCSIFFLAGS: 1917 case SIOCSIFCAP: 1918 case SIOCSIFMETRIC: 1919 case SIOCZIFDATA: 1920 case SIOCSIFMTU: 1921 case SIOCSIFPHYADDR: 1922 case SIOCDIFPHYADDR: 1923 #ifdef INET6 1924 case SIOCSIFPHYADDR_IN6: 1925 #endif 1926 case SIOCSLIFPHYADDR: 1927 case SIOCADDMULTI: 1928 case SIOCDELMULTI: 1929 case SIOCSIFMEDIA: 1930 case SIOCSDRVSPEC: 1931 case SIOCG80211: 1932 case SIOCS80211: 1933 case SIOCS80211NWID: 1934 case SIOCS80211NWKEY: 1935 case SIOCS80211POWER: 1936 case SIOCS80211BSSID: 1937 case SIOCS80211CHANNEL: 1938 case SIOCSLINKSTR: 1939 if (l != NULL) { 1940 error = kauth_authorize_network(l->l_cred, 1941 KAUTH_NETWORK_INTERFACE, 1942 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 1943 (void *)cmd, NULL); 1944 if (error != 0) 1945 return error; 1946 } 1947 } 1948 1949 oif_flags = ifp->if_flags; 1950 1951 ifnet_lock_enter(ifp->if_ioctl_lock); 1952 error = (*ifp->if_ioctl)(ifp, cmd, data); 1953 if (error != ENOTTY) 1954 ; 1955 else if (so->so_proto == NULL) 1956 error = EOPNOTSUPP; 1957 else { 1958 #ifdef COMPAT_OSOCK 1959 error = compat_ifioctl(so, ocmd, cmd, data, l); 1960 #else 1961 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 1962 cmd, data, ifp); 1963 #endif 1964 } 1965 1966 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 1967 #ifdef INET6 1968 if (in6_present && (ifp->if_flags & IFF_UP) != 0) { 1969 int s = splnet(); 1970 in6_if_up(ifp); 1971 splx(s); 1972 } 1973 #endif 1974 } 1975 #ifdef COMPAT_OIFREQ 1976 if (cmd != ocmd) 1977 ifreqn2o(oifr, ifr); 1978 #endif 1979 1980 ifnet_lock_exit(ifp->if_ioctl_lock); 1981 return error; 1982 } 1983 1984 /* This callback adds to the sum in `arg' the number of 1985 * threads on `ci' who have entered or who wait to enter the 1986 * critical section. 1987 */ 1988 static void 1989 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci) 1990 { 1991 uint64_t *sum = arg, *nenter = p; 1992 1993 *sum += *nenter; 1994 } 1995 1996 /* Return the number of threads who have entered or who wait 1997 * to enter the critical section on all CPUs. 1998 */ 1999 static uint64_t 2000 ifnet_lock_entrances(struct ifnet_lock *il) 2001 { 2002 uint64_t sum = 0; 2003 2004 percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum); 2005 2006 return sum; 2007 } 2008 2009 static int 2010 ifioctl_attach(struct ifnet *ifp) 2011 { 2012 struct ifnet_lock *il; 2013 2014 /* If the driver has not supplied its own if_ioctl, then 2015 * supply the default. 2016 */ 2017 if (ifp->if_ioctl == NULL) 2018 ifp->if_ioctl = ifioctl_common; 2019 2020 /* Create an ifnet_lock for synchronizing ifioctls. */ 2021 if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL) 2022 return ENOMEM; 2023 2024 il->il_nenter = percpu_alloc(sizeof(uint64_t)); 2025 if (il->il_nenter == NULL) { 2026 kmem_free(il, sizeof(*il)); 2027 return ENOMEM; 2028 } 2029 2030 mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE); 2031 cv_init(&il->il_emptied, ifp->if_xname); 2032 2033 ifp->if_ioctl_lock = il; 2034 2035 return 0; 2036 } 2037 2038 /* 2039 * This must not be called until after `ifp' has been withdrawn from the 2040 * ifnet tables so that ifioctl() cannot get a handle on it by calling 2041 * ifunit(). 2042 */ 2043 static void 2044 ifioctl_detach(struct ifnet *ifp) 2045 { 2046 struct ifnet_lock *il; 2047 2048 il = ifp->if_ioctl_lock; 2049 mutex_enter(&il->il_lock); 2050 /* Install if_nullioctl to make sure that any thread that 2051 * subsequently enters the critical section will quit it 2052 * immediately and signal the condition variable that we 2053 * wait on, below. 2054 */ 2055 ifp->if_ioctl = if_nullioctl; 2056 /* Sleep while threads are still in the critical section or 2057 * wait to enter it. 2058 */ 2059 while (ifnet_lock_entrances(il) != il->il_nexit) 2060 cv_wait(&il->il_emptied, &il->il_lock); 2061 /* At this point, we are the only thread still in the critical 2062 * section, and no new thread can get a handle on the ifioctl 2063 * lock, so it is safe to free its memory. 2064 */ 2065 mutex_exit(&il->il_lock); 2066 ifp->if_ioctl_lock = NULL; 2067 percpu_free(il->il_nenter, sizeof(uint64_t)); 2068 il->il_nenter = NULL; 2069 cv_destroy(&il->il_emptied); 2070 mutex_destroy(&il->il_lock); 2071 kmem_free(il, sizeof(*il)); 2072 } 2073 2074 /* 2075 * Return interface configuration 2076 * of system. List may be used 2077 * in later ioctl's (above) to get 2078 * other information. 2079 * 2080 * Each record is a struct ifreq. Before the addition of 2081 * sockaddr_storage, the API rule was that sockaddr flavors that did 2082 * not fit would extend beyond the struct ifreq, with the next struct 2083 * ifreq starting sa_len beyond the struct sockaddr. Because the 2084 * union in struct ifreq includes struct sockaddr_storage, every kind 2085 * of sockaddr must fit. Thus, there are no longer any overlength 2086 * records. 2087 * 2088 * Records are added to the user buffer if they fit, and ifc_len is 2089 * adjusted to the length that was written. Thus, the user is only 2090 * assured of getting the complete list if ifc_len on return is at 2091 * least sizeof(struct ifreq) less than it was on entry. 2092 * 2093 * If the user buffer pointer is NULL, this routine copies no data and 2094 * returns the amount of space that would be needed. 2095 * 2096 * Invariants: 2097 * ifrp points to the next part of the user's buffer to be used. If 2098 * ifrp != NULL, space holds the number of bytes remaining that we may 2099 * write at ifrp. Otherwise, space holds the number of bytes that 2100 * would have been written had there been adequate space. 2101 */ 2102 /*ARGSUSED*/ 2103 int 2104 ifconf(u_long cmd, void *data) 2105 { 2106 struct ifconf *ifc = (struct ifconf *)data; 2107 struct ifnet *ifp; 2108 struct ifaddr *ifa; 2109 struct ifreq ifr, *ifrp; 2110 int space, error = 0; 2111 const int sz = (int)sizeof(struct ifreq); 2112 2113 if ((ifrp = ifc->ifc_req) == NULL) 2114 space = 0; 2115 else 2116 space = ifc->ifc_len; 2117 IFNET_FOREACH(ifp) { 2118 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2119 sizeof(ifr.ifr_name)); 2120 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') 2121 return ENAMETOOLONG; 2122 if (IFADDR_EMPTY(ifp)) { 2123 /* Interface with no addresses - send zero sockaddr. */ 2124 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2125 if (ifrp == NULL) { 2126 space += sz; 2127 continue; 2128 } 2129 if (space >= sz) { 2130 error = copyout(&ifr, ifrp, sz); 2131 if (error != 0) 2132 return error; 2133 ifrp++; 2134 space -= sz; 2135 } 2136 } 2137 2138 IFADDR_FOREACH(ifa, ifp) { 2139 struct sockaddr *sa = ifa->ifa_addr; 2140 /* all sockaddrs must fit in sockaddr_storage */ 2141 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2142 2143 if (ifrp == NULL) { 2144 space += sz; 2145 continue; 2146 } 2147 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2148 if (space >= sz) { 2149 error = copyout(&ifr, ifrp, sz); 2150 if (error != 0) 2151 return (error); 2152 ifrp++; space -= sz; 2153 } 2154 } 2155 } 2156 if (ifrp != NULL) { 2157 KASSERT(0 <= space && space <= ifc->ifc_len); 2158 ifc->ifc_len -= space; 2159 } else { 2160 KASSERT(space >= 0); 2161 ifc->ifc_len = space; 2162 } 2163 return (0); 2164 } 2165 2166 int 2167 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2168 { 2169 uint8_t len; 2170 #ifdef COMPAT_OIFREQ 2171 struct ifreq ifrb; 2172 struct oifreq *oifr = NULL; 2173 u_long ocmd = cmd; 2174 cmd = compat_cvtcmd(cmd); 2175 if (cmd != ocmd) { 2176 oifr = (struct oifreq *)(void *)ifr; 2177 ifr = &ifrb; 2178 ifreqo2n(oifr, ifr); 2179 len = sizeof(oifr->ifr_addr); 2180 } else 2181 #endif 2182 len = sizeof(ifr->ifr_ifru.ifru_space); 2183 2184 if (len < sa->sa_len) 2185 return EFBIG; 2186 2187 memset(&ifr->ifr_addr, 0, len); 2188 sockaddr_copy(&ifr->ifr_addr, len, sa); 2189 2190 #ifdef COMPAT_OIFREQ 2191 if (cmd != ocmd) 2192 ifreqn2o(oifr, ifr); 2193 #endif 2194 return 0; 2195 } 2196 2197 /* 2198 * Queue message on interface, and start output if interface 2199 * not yet active. 2200 */ 2201 int 2202 ifq_enqueue(struct ifnet *ifp, struct mbuf *m 2203 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2204 { 2205 int len = m->m_pkthdr.len; 2206 int mflags = m->m_flags; 2207 int s = splnet(); 2208 int error; 2209 2210 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2211 if (error != 0) 2212 goto out; 2213 ifp->if_obytes += len; 2214 if (mflags & M_MCAST) 2215 ifp->if_omcasts++; 2216 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2217 (*ifp->if_start)(ifp); 2218 out: 2219 splx(s); 2220 return error; 2221 } 2222 2223 /* 2224 * Queue message on interface, possibly using a second fast queue 2225 */ 2226 int 2227 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m 2228 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2229 { 2230 int error = 0; 2231 2232 if (ifq != NULL 2233 #ifdef ALTQ 2234 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2235 #endif 2236 ) { 2237 if (IF_QFULL(ifq)) { 2238 IF_DROP(&ifp->if_snd); 2239 m_freem(m); 2240 if (error == 0) 2241 error = ENOBUFS; 2242 } else 2243 IF_ENQUEUE(ifq, m); 2244 } else 2245 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2246 if (error != 0) { 2247 ++ifp->if_oerrors; 2248 return error; 2249 } 2250 return 0; 2251 } 2252 2253 int 2254 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2255 { 2256 int rc; 2257 2258 if (ifp->if_initaddr != NULL) 2259 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2260 else if (src || 2261 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2262 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2263 2264 return rc; 2265 } 2266 2267 int 2268 if_flags_set(ifnet_t *ifp, const short flags) 2269 { 2270 int rc; 2271 2272 if (ifp->if_setflags != NULL) 2273 rc = (*ifp->if_setflags)(ifp, flags); 2274 else { 2275 short cantflags, chgdflags; 2276 struct ifreq ifr; 2277 2278 chgdflags = ifp->if_flags ^ flags; 2279 cantflags = chgdflags & IFF_CANTCHANGE; 2280 2281 if (cantflags != 0) 2282 ifp->if_flags ^= cantflags; 2283 2284 /* Traditionally, we do not call if_ioctl after 2285 * setting/clearing only IFF_PROMISC if the interface 2286 * isn't IFF_UP. Uphold that tradition. 2287 */ 2288 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2289 return 0; 2290 2291 memset(&ifr, 0, sizeof(ifr)); 2292 2293 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2294 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2295 2296 if (rc != 0 && cantflags != 0) 2297 ifp->if_flags ^= cantflags; 2298 } 2299 2300 return rc; 2301 } 2302 2303 int 2304 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2305 { 2306 int rc; 2307 struct ifreq ifr; 2308 2309 if (ifp->if_mcastop != NULL) 2310 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2311 else { 2312 ifreq_setaddr(cmd, &ifr, sa); 2313 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2314 } 2315 2316 return rc; 2317 } 2318 2319 static void 2320 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2321 struct ifaltq *ifq) 2322 { 2323 const struct sysctlnode *cnode, *rnode; 2324 2325 if (sysctl_createv(clog, 0, NULL, &rnode, 2326 CTLFLAG_PERMANENT, 2327 CTLTYPE_NODE, "interfaces", 2328 SYSCTL_DESCR("Per-interface controls"), 2329 NULL, 0, NULL, 0, 2330 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 2331 goto bad; 2332 2333 if (sysctl_createv(clog, 0, &rnode, &rnode, 2334 CTLFLAG_PERMANENT, 2335 CTLTYPE_NODE, ifname, 2336 SYSCTL_DESCR("Interface controls"), 2337 NULL, 0, NULL, 0, 2338 CTL_CREATE, CTL_EOL) != 0) 2339 goto bad; 2340 2341 if (sysctl_createv(clog, 0, &rnode, &rnode, 2342 CTLFLAG_PERMANENT, 2343 CTLTYPE_NODE, "sndq", 2344 SYSCTL_DESCR("Interface output queue controls"), 2345 NULL, 0, NULL, 0, 2346 CTL_CREATE, CTL_EOL) != 0) 2347 goto bad; 2348 2349 if (sysctl_createv(clog, 0, &rnode, &cnode, 2350 CTLFLAG_PERMANENT, 2351 CTLTYPE_INT, "len", 2352 SYSCTL_DESCR("Current output queue length"), 2353 NULL, 0, &ifq->ifq_len, 0, 2354 CTL_CREATE, CTL_EOL) != 0) 2355 goto bad; 2356 2357 if (sysctl_createv(clog, 0, &rnode, &cnode, 2358 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2359 CTLTYPE_INT, "maxlen", 2360 SYSCTL_DESCR("Maximum allowed output queue length"), 2361 NULL, 0, &ifq->ifq_maxlen, 0, 2362 CTL_CREATE, CTL_EOL) != 0) 2363 goto bad; 2364 2365 if (sysctl_createv(clog, 0, &rnode, &cnode, 2366 CTLFLAG_PERMANENT, 2367 CTLTYPE_INT, "drops", 2368 SYSCTL_DESCR("Packets dropped due to full output queue"), 2369 NULL, 0, &ifq->ifq_drops, 0, 2370 CTL_CREATE, CTL_EOL) != 0) 2371 goto bad; 2372 2373 return; 2374 bad: 2375 printf("%s: could not attach sysctl nodes\n", ifname); 2376 return; 2377 } 2378 2379 #if defined(INET) || defined(INET6) 2380 2381 #define SYSCTL_NET_PKTQ(q, cn, c) \ 2382 static int \ 2383 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 2384 { \ 2385 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 2386 } 2387 2388 #if defined(INET) 2389 static int 2390 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 2391 { 2392 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 2393 } 2394 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 2395 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 2396 #endif 2397 2398 #if defined(INET6) 2399 static int 2400 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 2401 { 2402 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 2403 } 2404 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 2405 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 2406 #endif 2407 2408 static void 2409 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 2410 { 2411 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 2412 const char *pfname = NULL, *ipname = NULL; 2413 int ipn = 0, qid = 0; 2414 2415 switch (pf) { 2416 #if defined(INET) 2417 case PF_INET: 2418 len_func = sysctl_net_ip_pktq_items; 2419 maxlen_func = sysctl_net_ip_pktq_maxlen; 2420 drops_func = sysctl_net_ip_pktq_drops; 2421 pfname = "inet", ipn = IPPROTO_IP; 2422 ipname = "ip", qid = IPCTL_IFQ; 2423 break; 2424 #endif 2425 #if defined(INET6) 2426 case PF_INET6: 2427 len_func = sysctl_net_ip6_pktq_items; 2428 maxlen_func = sysctl_net_ip6_pktq_maxlen; 2429 drops_func = sysctl_net_ip6_pktq_drops; 2430 pfname = "inet6", ipn = IPPROTO_IPV6; 2431 ipname = "ip6", qid = IPV6CTL_IFQ; 2432 break; 2433 #endif 2434 default: 2435 KASSERT(false); 2436 } 2437 2438 sysctl_createv(clog, 0, NULL, NULL, 2439 CTLFLAG_PERMANENT, 2440 CTLTYPE_NODE, pfname, NULL, 2441 NULL, 0, NULL, 0, 2442 CTL_NET, pf, CTL_EOL); 2443 sysctl_createv(clog, 0, NULL, NULL, 2444 CTLFLAG_PERMANENT, 2445 CTLTYPE_NODE, ipname, NULL, 2446 NULL, 0, NULL, 0, 2447 CTL_NET, pf, ipn, CTL_EOL); 2448 sysctl_createv(clog, 0, NULL, NULL, 2449 CTLFLAG_PERMANENT, 2450 CTLTYPE_NODE, "ifq", 2451 SYSCTL_DESCR("Protocol input queue controls"), 2452 NULL, 0, NULL, 0, 2453 CTL_NET, pf, ipn, qid, CTL_EOL); 2454 2455 sysctl_createv(clog, 0, NULL, NULL, 2456 CTLFLAG_PERMANENT, 2457 CTLTYPE_INT, "len", 2458 SYSCTL_DESCR("Current input queue length"), 2459 len_func, 0, NULL, 0, 2460 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 2461 sysctl_createv(clog, 0, NULL, NULL, 2462 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2463 CTLTYPE_INT, "maxlen", 2464 SYSCTL_DESCR("Maximum allowed input queue length"), 2465 maxlen_func, 0, NULL, 0, 2466 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 2467 sysctl_createv(clog, 0, NULL, NULL, 2468 CTLFLAG_PERMANENT, 2469 CTLTYPE_INT, "drops", 2470 SYSCTL_DESCR("Packets dropped due to full input queue"), 2471 drops_func, 0, NULL, 0, 2472 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 2473 } 2474 #endif /* INET || INET6 */ 2475 2476 static int 2477 if_sdl_sysctl(SYSCTLFN_ARGS) 2478 { 2479 struct ifnet *ifp; 2480 const struct sockaddr_dl *sdl; 2481 2482 if (namelen != 1) 2483 return EINVAL; 2484 2485 ifp = if_byindex(name[0]); 2486 if (ifp == NULL) 2487 return ENODEV; 2488 2489 sdl = ifp->if_sadl; 2490 if (sdl == NULL) { 2491 *oldlenp = 0; 2492 return 0; 2493 } 2494 2495 if (oldp == NULL) { 2496 *oldlenp = sdl->sdl_alen; 2497 return 0; 2498 } 2499 2500 if (*oldlenp >= sdl->sdl_alen) 2501 *oldlenp = sdl->sdl_alen; 2502 return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 2503 } 2504 2505 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 2506 { 2507 const struct sysctlnode *rnode = NULL; 2508 2509 sysctl_createv(clog, 0, NULL, &rnode, 2510 CTLFLAG_PERMANENT, 2511 CTLTYPE_NODE, "sdl", 2512 SYSCTL_DESCR("Get active link-layer address"), 2513 if_sdl_sysctl, 0, NULL, 0, 2514 CTL_NET, CTL_CREATE, CTL_EOL); 2515 } 2516
Indexes created Sun Sep 21 20:09:37 GMT 2025