if.c revision 1.289
1 /* $NetBSD: if.c,v 1.289 2014/07/31 06:35:47 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 1.289 2014/07/31 06:35:47 ozaki-r 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; 167 168 static struct ifaddr ** ifnet_addrs = NULL; 169 170 static callout_t if_slowtimo_ch; 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 static int if_clone_list(struct if_clonereq *); 179 180 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 181 static int if_cloners_count; 182 183 /* Packet filtering hook for interfaces. */ 184 pfil_head_t * if_pfil; 185 186 static kauth_listener_t if_listener; 187 188 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 189 static int ifioctl_attach(struct ifnet *); 190 static void ifioctl_detach(struct ifnet *); 191 static void ifnet_lock_enter(struct ifnet_lock *); 192 static void ifnet_lock_exit(struct ifnet_lock *); 193 static void if_detach_queues(struct ifnet *, struct ifqueue *); 194 static void sysctl_sndq_setup(struct sysctllog **, const char *, 195 struct ifaltq *); 196 197 #if defined(INET) || defined(INET6) 198 static void sysctl_net_pktq_setup(struct sysctllog **, int); 199 #endif 200 201 static int 202 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 203 void *arg0, void *arg1, void *arg2, void *arg3) 204 { 205 int result; 206 enum kauth_network_req req; 207 208 result = KAUTH_RESULT_DEFER; 209 req = (enum kauth_network_req)arg1; 210 211 if (action != KAUTH_NETWORK_INTERFACE) 212 return result; 213 214 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 215 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 216 result = KAUTH_RESULT_ALLOW; 217 218 return result; 219 } 220 221 /* 222 * Network interface utility routines. 223 * 224 * Routines with ifa_ifwith* names take sockaddr *'s as 225 * parameters. 226 */ 227 void 228 ifinit(void) 229 { 230 #if defined(INET) 231 sysctl_net_pktq_setup(NULL, PF_INET); 232 #endif 233 #ifdef INET6 234 sysctl_net_pktq_setup(NULL, PF_INET6); 235 #endif 236 237 callout_init(&if_slowtimo_ch, 0); 238 if_slowtimo(NULL); 239 240 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 241 if_listener_cb, NULL); 242 243 /* interfaces are available, inform socket code */ 244 ifioctl = doifioctl; 245 } 246 247 /* 248 * XXX Initialization before configure(). 249 * XXX hack to get pfil_add_hook working in autoconf. 250 */ 251 void 252 ifinit1(void) 253 { 254 mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE); 255 TAILQ_INIT(&ifnet_list); 256 if_indexlim = 8; 257 258 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 259 KASSERT(if_pfil != NULL); 260 261 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 262 etherinit(); 263 #endif 264 } 265 266 ifnet_t * 267 if_alloc(u_char type) 268 { 269 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 270 } 271 272 void 273 if_free(ifnet_t *ifp) 274 { 275 kmem_free(ifp, sizeof(ifnet_t)); 276 } 277 278 void 279 if_initname(struct ifnet *ifp, const char *name, int unit) 280 { 281 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 282 "%s%d", name, unit); 283 } 284 285 /* 286 * Null routines used while an interface is going away. These routines 287 * just return an error. 288 */ 289 290 int 291 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 292 const struct sockaddr *so, struct rtentry *rt) 293 { 294 295 return ENXIO; 296 } 297 298 void 299 if_nullinput(struct ifnet *ifp, struct mbuf *m) 300 { 301 302 /* Nothing. */ 303 } 304 305 void 306 if_nullstart(struct ifnet *ifp) 307 { 308 309 /* Nothing. */ 310 } 311 312 int 313 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 314 { 315 316 /* Wake ifioctl_detach(), who may wait for all threads to 317 * quit the critical section. 318 */ 319 cv_signal(&ifp->if_ioctl_lock->il_emptied); 320 return ENXIO; 321 } 322 323 int 324 if_nullinit(struct ifnet *ifp) 325 { 326 327 return ENXIO; 328 } 329 330 void 331 if_nullstop(struct ifnet *ifp, int disable) 332 { 333 334 /* Nothing. */ 335 } 336 337 void 338 if_nullwatchdog(struct ifnet *ifp) 339 { 340 341 /* Nothing. */ 342 } 343 344 void 345 if_nulldrain(struct ifnet *ifp) 346 { 347 348 /* Nothing. */ 349 } 350 351 void 352 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 353 { 354 struct ifaddr *ifa; 355 struct sockaddr_dl *sdl; 356 357 ifp->if_addrlen = addrlen; 358 if_alloc_sadl(ifp); 359 ifa = ifp->if_dl; 360 sdl = satosdl(ifa->ifa_addr); 361 362 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 363 if (factory) { 364 ifp->if_hwdl = ifp->if_dl; 365 IFAREF(ifp->if_hwdl); 366 } 367 /* TBD routing socket */ 368 } 369 370 struct ifaddr * 371 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 372 { 373 unsigned socksize, ifasize; 374 int addrlen, namelen; 375 struct sockaddr_dl *mask, *sdl; 376 struct ifaddr *ifa; 377 378 namelen = strlen(ifp->if_xname); 379 addrlen = ifp->if_addrlen; 380 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 381 ifasize = sizeof(*ifa) + 2 * socksize; 382 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 383 384 sdl = (struct sockaddr_dl *)(ifa + 1); 385 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 386 387 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 388 ifp->if_xname, namelen, NULL, addrlen); 389 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 390 memset(&mask->sdl_data[0], 0xff, namelen); 391 ifa->ifa_rtrequest = link_rtrequest; 392 ifa->ifa_addr = (struct sockaddr *)sdl; 393 ifa->ifa_netmask = (struct sockaddr *)mask; 394 395 *sdlp = sdl; 396 397 return ifa; 398 } 399 400 static void 401 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 402 { 403 const struct sockaddr_dl *sdl; 404 ifnet_addrs[ifp->if_index] = ifa; 405 IFAREF(ifa); 406 ifp->if_dl = ifa; 407 IFAREF(ifa); 408 sdl = satosdl(ifa->ifa_addr); 409 ifp->if_sadl = sdl; 410 } 411 412 /* 413 * Allocate the link level name for the specified interface. This 414 * is an attachment helper. It must be called after ifp->if_addrlen 415 * is initialized, which may not be the case when if_attach() is 416 * called. 417 */ 418 void 419 if_alloc_sadl(struct ifnet *ifp) 420 { 421 struct ifaddr *ifa; 422 const struct sockaddr_dl *sdl; 423 424 /* 425 * If the interface already has a link name, release it 426 * now. This is useful for interfaces that can change 427 * link types, and thus switch link names often. 428 */ 429 if (ifp->if_sadl != NULL) 430 if_free_sadl(ifp); 431 432 ifa = if_dl_create(ifp, &sdl); 433 434 ifa_insert(ifp, ifa); 435 if_sadl_setrefs(ifp, ifa); 436 } 437 438 static void 439 if_deactivate_sadl(struct ifnet *ifp) 440 { 441 struct ifaddr *ifa; 442 443 KASSERT(ifp->if_dl != NULL); 444 445 ifa = ifp->if_dl; 446 447 ifp->if_sadl = NULL; 448 449 ifnet_addrs[ifp->if_index] = NULL; 450 IFAFREE(ifa); 451 ifp->if_dl = NULL; 452 IFAFREE(ifa); 453 } 454 455 void 456 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 457 const struct sockaddr_dl *sdl) 458 { 459 int s; 460 461 s = splnet(); 462 463 if_deactivate_sadl(ifp); 464 465 if_sadl_setrefs(ifp, ifa); 466 IFADDR_FOREACH(ifa, ifp) 467 rtinit(ifa, RTM_LLINFO_UPD, 0); 468 splx(s); 469 } 470 471 /* 472 * Free the link level name for the specified interface. This is 473 * a detach helper. This is called from if_detach() or from 474 * link layer type specific detach functions. 475 */ 476 void 477 if_free_sadl(struct ifnet *ifp) 478 { 479 struct ifaddr *ifa; 480 int s; 481 482 ifa = ifnet_addrs[ifp->if_index]; 483 if (ifa == NULL) { 484 KASSERT(ifp->if_sadl == NULL); 485 KASSERT(ifp->if_dl == NULL); 486 return; 487 } 488 489 KASSERT(ifp->if_sadl != NULL); 490 KASSERT(ifp->if_dl != NULL); 491 492 s = splnet(); 493 rtinit(ifa, RTM_DELETE, 0); 494 ifa_remove(ifp, ifa); 495 if_deactivate_sadl(ifp); 496 if (ifp->if_hwdl == ifa) { 497 IFAFREE(ifa); 498 ifp->if_hwdl = NULL; 499 } 500 splx(s); 501 } 502 503 static void 504 if_getindex(ifnet_t *ifp) 505 { 506 bool hitlimit = false; 507 508 mutex_enter(&index_gen_mtx); 509 ifp->if_index_gen = index_gen++; 510 mutex_exit(&index_gen_mtx); 511 512 ifp->if_index = if_index; 513 if (ifindex2ifnet == NULL) { 514 if_index++; 515 goto skip; 516 } 517 while (if_byindex(ifp->if_index)) { 518 /* 519 * If we hit USHRT_MAX, we skip back to 0 since 520 * there are a number of places where the value 521 * of if_index or if_index itself is compared 522 * to or stored in an unsigned short. By 523 * jumping back, we won't botch those assignments 524 * or comparisons. 525 */ 526 if (++if_index == 0) { 527 if_index = 1; 528 } else if (if_index == USHRT_MAX) { 529 /* 530 * However, if we have to jump back to 531 * zero *twice* without finding an empty 532 * slot in ifindex2ifnet[], then there 533 * there are too many (>65535) interfaces. 534 */ 535 if (hitlimit) { 536 panic("too many interfaces"); 537 } 538 hitlimit = true; 539 if_index = 1; 540 } 541 ifp->if_index = if_index; 542 } 543 skip: 544 /* 545 * We have some arrays that should be indexed by if_index. 546 * since if_index will grow dynamically, they should grow too. 547 * struct ifadd **ifnet_addrs 548 * struct ifnet **ifindex2ifnet 549 */ 550 if (ifnet_addrs == NULL || ifindex2ifnet == NULL || 551 ifp->if_index >= if_indexlim) { 552 size_t m, n, oldlim; 553 void *q; 554 555 oldlim = if_indexlim; 556 while (ifp->if_index >= if_indexlim) 557 if_indexlim <<= 1; 558 559 /* grow ifnet_addrs */ 560 m = oldlim * sizeof(struct ifaddr *); 561 n = if_indexlim * sizeof(struct ifaddr *); 562 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 563 if (ifnet_addrs != NULL) { 564 memcpy(q, ifnet_addrs, m); 565 free(ifnet_addrs, M_IFADDR); 566 } 567 ifnet_addrs = (struct ifaddr **)q; 568 569 /* grow ifindex2ifnet */ 570 m = oldlim * sizeof(struct ifnet *); 571 n = if_indexlim * sizeof(struct ifnet *); 572 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 573 if (ifindex2ifnet != NULL) { 574 memcpy(q, ifindex2ifnet, m); 575 free(ifindex2ifnet, M_IFADDR); 576 } 577 ifindex2ifnet = (struct ifnet **)q; 578 } 579 ifindex2ifnet[ifp->if_index] = ifp; 580 } 581 582 /* 583 * Attach an interface to the list of "active" interfaces. 584 */ 585 void 586 if_attach(ifnet_t *ifp) 587 { 588 KASSERT(if_indexlim > 0); 589 TAILQ_INIT(&ifp->if_addrlist); 590 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 591 592 if (ifioctl_attach(ifp) != 0) 593 panic("%s: ifioctl_attach() failed", __func__); 594 595 if_getindex(ifp); 596 597 /* 598 * Link level name is allocated later by a separate call to 599 * if_alloc_sadl(). 600 */ 601 602 if (ifp->if_snd.ifq_maxlen == 0) 603 ifp->if_snd.ifq_maxlen = ifqmaxlen; 604 605 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 606 607 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 608 609 ifp->if_link_state = LINK_STATE_UNKNOWN; 610 611 ifp->if_capenable = 0; 612 ifp->if_csum_flags_tx = 0; 613 ifp->if_csum_flags_rx = 0; 614 615 #ifdef ALTQ 616 ifp->if_snd.altq_type = 0; 617 ifp->if_snd.altq_disc = NULL; 618 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 619 ifp->if_snd.altq_tbr = NULL; 620 ifp->if_snd.altq_ifp = ifp; 621 #endif 622 623 #ifdef NET_MPSAFE 624 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 625 #else 626 ifp->if_snd.ifq_lock = NULL; 627 #endif 628 629 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 630 (void)pfil_run_hooks(if_pfil, 631 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 632 633 if (!STAILQ_EMPTY(&domains)) 634 if_attachdomain1(ifp); 635 636 /* Announce the interface. */ 637 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 638 } 639 640 void 641 if_attachdomain(void) 642 { 643 struct ifnet *ifp; 644 int s; 645 646 s = splnet(); 647 IFNET_FOREACH(ifp) 648 if_attachdomain1(ifp); 649 splx(s); 650 } 651 652 void 653 if_attachdomain1(struct ifnet *ifp) 654 { 655 struct domain *dp; 656 int s; 657 658 s = splnet(); 659 660 /* address family dependent data region */ 661 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 662 DOMAIN_FOREACH(dp) { 663 if (dp->dom_ifattach != NULL) 664 ifp->if_afdata[dp->dom_family] = 665 (*dp->dom_ifattach)(ifp); 666 } 667 668 splx(s); 669 } 670 671 /* 672 * Deactivate an interface. This points all of the procedure 673 * handles at error stubs. May be called from interrupt context. 674 */ 675 void 676 if_deactivate(struct ifnet *ifp) 677 { 678 int s; 679 680 s = splnet(); 681 682 ifp->if_output = if_nulloutput; 683 ifp->if_input = if_nullinput; 684 ifp->if_start = if_nullstart; 685 ifp->if_ioctl = if_nullioctl; 686 ifp->if_init = if_nullinit; 687 ifp->if_stop = if_nullstop; 688 ifp->if_watchdog = if_nullwatchdog; 689 ifp->if_drain = if_nulldrain; 690 691 /* No more packets may be enqueued. */ 692 ifp->if_snd.ifq_maxlen = 0; 693 694 splx(s); 695 } 696 697 void 698 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 699 { 700 struct ifaddr *ifa, *nifa; 701 702 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 703 if (ifa->ifa_addr->sa_family != family) 704 continue; 705 (*purgeaddr)(ifa); 706 } 707 } 708 709 /* 710 * Detach an interface from the list of "active" interfaces, 711 * freeing any resources as we go along. 712 * 713 * NOTE: This routine must be called with a valid thread context, 714 * as it may block. 715 */ 716 void 717 if_detach(struct ifnet *ifp) 718 { 719 struct socket so; 720 struct ifaddr *ifa; 721 #ifdef IFAREF_DEBUG 722 struct ifaddr *last_ifa = NULL; 723 #endif 724 struct domain *dp; 725 const struct protosw *pr; 726 int s, i, family, purged; 727 uint64_t xc; 728 729 /* 730 * XXX It's kind of lame that we have to have the 731 * XXX socket structure... 732 */ 733 memset(&so, 0, sizeof(so)); 734 735 s = splnet(); 736 737 /* 738 * Do an if_down() to give protocols a chance to do something. 739 */ 740 if_down(ifp); 741 742 #ifdef ALTQ 743 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 744 altq_disable(&ifp->if_snd); 745 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 746 altq_detach(&ifp->if_snd); 747 #endif 748 749 if (ifp->if_snd.ifq_lock) 750 mutex_obj_free(ifp->if_snd.ifq_lock); 751 752 sysctl_teardown(&ifp->if_sysctl_log); 753 754 #if NCARP > 0 755 /* Remove the interface from any carp group it is a part of. */ 756 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 757 carp_ifdetach(ifp); 758 #endif 759 760 /* 761 * Rip all the addresses off the interface. This should make 762 * all of the routes go away. 763 * 764 * pr_usrreq calls can remove an arbitrary number of ifaddrs 765 * from the list, including our "cursor", ifa. For safety, 766 * and to honor the TAILQ abstraction, I just restart the 767 * loop after each removal. Note that the loop will exit 768 * when all of the remaining ifaddrs belong to the AF_LINK 769 * family. I am counting on the historical fact that at 770 * least one pr_usrreq in each address domain removes at 771 * least one ifaddr. 772 */ 773 again: 774 IFADDR_FOREACH(ifa, ifp) { 775 family = ifa->ifa_addr->sa_family; 776 #ifdef IFAREF_DEBUG 777 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 778 ifa, family, ifa->ifa_refcnt); 779 if (last_ifa != NULL && ifa == last_ifa) 780 panic("if_detach: loop detected"); 781 last_ifa = ifa; 782 #endif 783 if (family == AF_LINK) 784 continue; 785 dp = pffinddomain(family); 786 #ifdef DIAGNOSTIC 787 if (dp == NULL) 788 panic("if_detach: no domain for AF %d", 789 family); 790 #endif 791 /* 792 * XXX These PURGEIF calls are redundant with the 793 * purge-all-families calls below, but are left in for 794 * now both to make a smaller change, and to avoid 795 * unplanned interactions with clearing of 796 * ifp->if_addrlist. 797 */ 798 purged = 0; 799 for (pr = dp->dom_protosw; 800 pr < dp->dom_protoswNPROTOSW; pr++) { 801 so.so_proto = pr; 802 if (pr->pr_usrreqs) { 803 (void) (*pr->pr_usrreqs->pr_generic)(&so, 804 PRU_PURGEIF, NULL, NULL, 805 (struct mbuf *) ifp, curlwp); 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_generic)(&so, 856 PRU_PURGEIF, NULL, NULL, 857 (struct mbuf *)ifp, curlwp); 858 } 859 } 860 861 (void)pfil_run_hooks(if_pfil, 862 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 863 (void)pfil_head_destroy(ifp->if_pfil); 864 865 /* Announce that the interface is gone. */ 866 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 867 868 ifindex2ifnet[ifp->if_index] = NULL; 869 870 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 871 872 ifioctl_detach(ifp); 873 874 /* 875 * remove packets that came from ifp, from software interrupt queues. 876 */ 877 DOMAIN_FOREACH(dp) { 878 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 879 struct ifqueue *iq = dp->dom_ifqueues[i]; 880 if (iq == NULL) 881 break; 882 dp->dom_ifqueues[i] = NULL; 883 if_detach_queues(ifp, iq); 884 } 885 } 886 887 /* 888 * IP queues have to be processed separately: net-queue barrier 889 * ensures that the packets are dequeued while a cross-call will 890 * ensure that the interrupts have completed. FIXME: not quite.. 891 */ 892 #ifdef INET 893 pktq_barrier(ip_pktq); 894 #endif 895 #ifdef INET6 896 pktq_barrier(ip6_pktq); 897 #endif 898 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 899 xc_wait(xc); 900 901 splx(s); 902 } 903 904 static void 905 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 906 { 907 struct mbuf *m, *prev, *next; 908 909 prev = NULL; 910 for (m = q->ifq_head; m != NULL; m = next) { 911 KASSERT((m->m_flags & M_PKTHDR) != 0); 912 913 next = m->m_nextpkt; 914 if (m->m_pkthdr.rcvif != ifp) { 915 prev = m; 916 continue; 917 } 918 919 if (prev != NULL) 920 prev->m_nextpkt = m->m_nextpkt; 921 else 922 q->ifq_head = m->m_nextpkt; 923 if (q->ifq_tail == m) 924 q->ifq_tail = prev; 925 q->ifq_len--; 926 927 m->m_nextpkt = NULL; 928 m_freem(m); 929 IF_DROP(q); 930 } 931 } 932 933 /* 934 * Callback for a radix tree walk to delete all references to an 935 * ifnet. 936 */ 937 static int 938 if_rt_walktree(struct rtentry *rt, void *v) 939 { 940 struct ifnet *ifp = (struct ifnet *)v; 941 int error; 942 943 if (rt->rt_ifp != ifp) 944 return 0; 945 946 /* Delete the entry. */ 947 ++rt->rt_refcnt; 948 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 949 rt_mask(rt), rt->rt_flags, NULL); 950 KASSERT((rt->rt_flags & RTF_UP) == 0); 951 rt->rt_ifp = NULL; 952 rtfree(rt); 953 if (error != 0) 954 printf("%s: warning: unable to delete rtentry @ %p, " 955 "error = %d\n", ifp->if_xname, rt, error); 956 return ERESTART; 957 } 958 959 /* 960 * Create a clone network interface. 961 */ 962 int 963 if_clone_create(const char *name) 964 { 965 struct if_clone *ifc; 966 int unit; 967 968 ifc = if_clone_lookup(name, &unit); 969 if (ifc == NULL) 970 return EINVAL; 971 972 if (ifunit(name) != NULL) 973 return EEXIST; 974 975 return (*ifc->ifc_create)(ifc, unit); 976 } 977 978 /* 979 * Destroy a clone network interface. 980 */ 981 int 982 if_clone_destroy(const char *name) 983 { 984 struct if_clone *ifc; 985 struct ifnet *ifp; 986 987 ifc = if_clone_lookup(name, NULL); 988 if (ifc == NULL) 989 return EINVAL; 990 991 ifp = ifunit(name); 992 if (ifp == NULL) 993 return ENXIO; 994 995 if (ifc->ifc_destroy == NULL) 996 return EOPNOTSUPP; 997 998 return (*ifc->ifc_destroy)(ifp); 999 } 1000 1001 /* 1002 * Look up a network interface cloner. 1003 */ 1004 static struct if_clone * 1005 if_clone_lookup(const char *name, int *unitp) 1006 { 1007 struct if_clone *ifc; 1008 const char *cp; 1009 char *dp, ifname[IFNAMSIZ + 3]; 1010 int unit; 1011 1012 strcpy(ifname, "if_"); 1013 /* separate interface name from unit */ 1014 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1015 *cp && (*cp < '0' || *cp > '9');) 1016 *dp++ = *cp++; 1017 1018 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1019 return NULL; /* No name or unit number */ 1020 *dp++ = '\0'; 1021 1022 again: 1023 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1024 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1025 break; 1026 } 1027 1028 if (ifc == NULL) { 1029 if (*ifname == '\0' || 1030 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1031 return NULL; 1032 *ifname = '\0'; 1033 goto again; 1034 } 1035 1036 unit = 0; 1037 while (cp - name < IFNAMSIZ && *cp) { 1038 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1039 /* Bogus unit number. */ 1040 return NULL; 1041 } 1042 unit = (unit * 10) + (*cp++ - '0'); 1043 } 1044 1045 if (unitp != NULL) 1046 *unitp = unit; 1047 return ifc; 1048 } 1049 1050 /* 1051 * Register a network interface cloner. 1052 */ 1053 void 1054 if_clone_attach(struct if_clone *ifc) 1055 { 1056 1057 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1058 if_cloners_count++; 1059 } 1060 1061 /* 1062 * Unregister a network interface cloner. 1063 */ 1064 void 1065 if_clone_detach(struct if_clone *ifc) 1066 { 1067 1068 LIST_REMOVE(ifc, ifc_list); 1069 if_cloners_count--; 1070 } 1071 1072 /* 1073 * Provide list of interface cloners to userspace. 1074 */ 1075 static int 1076 if_clone_list(struct if_clonereq *ifcr) 1077 { 1078 char outbuf[IFNAMSIZ], *dst; 1079 struct if_clone *ifc; 1080 int count, error = 0; 1081 1082 ifcr->ifcr_total = if_cloners_count; 1083 if ((dst = ifcr->ifcr_buffer) == NULL) { 1084 /* Just asking how many there are. */ 1085 return 0; 1086 } 1087 1088 if (ifcr->ifcr_count < 0) 1089 return EINVAL; 1090 1091 count = (if_cloners_count < ifcr->ifcr_count) ? 1092 if_cloners_count : ifcr->ifcr_count; 1093 1094 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1095 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1096 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1097 if (outbuf[sizeof(outbuf) - 1] != '\0') 1098 return ENAMETOOLONG; 1099 error = copyout(outbuf, dst, sizeof(outbuf)); 1100 if (error != 0) 1101 break; 1102 } 1103 1104 return error; 1105 } 1106 1107 void 1108 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1109 { 1110 ifa->ifa_ifp = ifp; 1111 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1112 IFAREF(ifa); 1113 } 1114 1115 void 1116 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1117 { 1118 KASSERT(ifa->ifa_ifp == ifp); 1119 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1120 IFAFREE(ifa); 1121 } 1122 1123 static inline int 1124 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1125 { 1126 return sockaddr_cmp(sa1, sa2) == 0; 1127 } 1128 1129 /* 1130 * Locate an interface based on a complete address. 1131 */ 1132 /*ARGSUSED*/ 1133 struct ifaddr * 1134 ifa_ifwithaddr(const struct sockaddr *addr) 1135 { 1136 struct ifnet *ifp; 1137 struct ifaddr *ifa; 1138 1139 IFNET_FOREACH(ifp) { 1140 if (ifp->if_output == if_nulloutput) 1141 continue; 1142 IFADDR_FOREACH(ifa, ifp) { 1143 if (ifa->ifa_addr->sa_family != addr->sa_family) 1144 continue; 1145 if (equal(addr, ifa->ifa_addr)) 1146 return ifa; 1147 if ((ifp->if_flags & IFF_BROADCAST) && 1148 ifa->ifa_broadaddr && 1149 /* IP6 doesn't have broadcast */ 1150 ifa->ifa_broadaddr->sa_len != 0 && 1151 equal(ifa->ifa_broadaddr, addr)) 1152 return ifa; 1153 } 1154 } 1155 return NULL; 1156 } 1157 1158 /* 1159 * Locate the point to point interface with a given destination address. 1160 */ 1161 /*ARGSUSED*/ 1162 struct ifaddr * 1163 ifa_ifwithdstaddr(const struct sockaddr *addr) 1164 { 1165 struct ifnet *ifp; 1166 struct ifaddr *ifa; 1167 1168 IFNET_FOREACH(ifp) { 1169 if (ifp->if_output == if_nulloutput) 1170 continue; 1171 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1172 continue; 1173 IFADDR_FOREACH(ifa, ifp) { 1174 if (ifa->ifa_addr->sa_family != addr->sa_family || 1175 ifa->ifa_dstaddr == NULL) 1176 continue; 1177 if (equal(addr, ifa->ifa_dstaddr)) 1178 return ifa; 1179 } 1180 } 1181 return NULL; 1182 } 1183 1184 /* 1185 * Find an interface on a specific network. If many, choice 1186 * is most specific found. 1187 */ 1188 struct ifaddr * 1189 ifa_ifwithnet(const struct sockaddr *addr) 1190 { 1191 struct ifnet *ifp; 1192 struct ifaddr *ifa; 1193 const struct sockaddr_dl *sdl; 1194 struct ifaddr *ifa_maybe = 0; 1195 u_int af = addr->sa_family; 1196 const char *addr_data = addr->sa_data, *cplim; 1197 1198 if (af == AF_LINK) { 1199 sdl = satocsdl(addr); 1200 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1201 ifindex2ifnet[sdl->sdl_index] && 1202 ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) 1203 return ifnet_addrs[sdl->sdl_index]; 1204 } 1205 #ifdef NETATALK 1206 if (af == AF_APPLETALK) { 1207 const struct sockaddr_at *sat, *sat2; 1208 sat = (const struct sockaddr_at *)addr; 1209 IFNET_FOREACH(ifp) { 1210 if (ifp->if_output == if_nulloutput) 1211 continue; 1212 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1213 if (ifa == NULL) 1214 continue; 1215 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1216 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1217 return ifa; /* exact match */ 1218 if (ifa_maybe == NULL) { 1219 /* else keep the if with the right range */ 1220 ifa_maybe = ifa; 1221 } 1222 } 1223 return ifa_maybe; 1224 } 1225 #endif 1226 IFNET_FOREACH(ifp) { 1227 if (ifp->if_output == if_nulloutput) 1228 continue; 1229 IFADDR_FOREACH(ifa, ifp) { 1230 const char *cp, *cp2, *cp3; 1231 1232 if (ifa->ifa_addr->sa_family != af || 1233 ifa->ifa_netmask == NULL) 1234 next: continue; 1235 cp = addr_data; 1236 cp2 = ifa->ifa_addr->sa_data; 1237 cp3 = ifa->ifa_netmask->sa_data; 1238 cplim = (const char *)ifa->ifa_netmask + 1239 ifa->ifa_netmask->sa_len; 1240 while (cp3 < cplim) { 1241 if ((*cp++ ^ *cp2++) & *cp3++) { 1242 /* want to continue for() loop */ 1243 goto next; 1244 } 1245 } 1246 if (ifa_maybe == NULL || 1247 rn_refines((void *)ifa->ifa_netmask, 1248 (void *)ifa_maybe->ifa_netmask)) 1249 ifa_maybe = ifa; 1250 } 1251 } 1252 return ifa_maybe; 1253 } 1254 1255 /* 1256 * Find the interface of the addresss. 1257 */ 1258 struct ifaddr * 1259 ifa_ifwithladdr(const struct sockaddr *addr) 1260 { 1261 struct ifaddr *ia; 1262 1263 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1264 (ia = ifa_ifwithnet(addr))) 1265 return ia; 1266 return NULL; 1267 } 1268 1269 /* 1270 * Find an interface using a specific address family 1271 */ 1272 struct ifaddr * 1273 ifa_ifwithaf(int af) 1274 { 1275 struct ifnet *ifp; 1276 struct ifaddr *ifa; 1277 1278 IFNET_FOREACH(ifp) { 1279 if (ifp->if_output == if_nulloutput) 1280 continue; 1281 IFADDR_FOREACH(ifa, ifp) { 1282 if (ifa->ifa_addr->sa_family == af) 1283 return ifa; 1284 } 1285 } 1286 return NULL; 1287 } 1288 1289 /* 1290 * Find an interface address specific to an interface best matching 1291 * a given address. 1292 */ 1293 struct ifaddr * 1294 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1295 { 1296 struct ifaddr *ifa; 1297 const char *cp, *cp2, *cp3; 1298 const char *cplim; 1299 struct ifaddr *ifa_maybe = 0; 1300 u_int af = addr->sa_family; 1301 1302 if (ifp->if_output == if_nulloutput) 1303 return NULL; 1304 1305 if (af >= AF_MAX) 1306 return NULL; 1307 1308 IFADDR_FOREACH(ifa, ifp) { 1309 if (ifa->ifa_addr->sa_family != af) 1310 continue; 1311 ifa_maybe = ifa; 1312 if (ifa->ifa_netmask == NULL) { 1313 if (equal(addr, ifa->ifa_addr) || 1314 (ifa->ifa_dstaddr && 1315 equal(addr, ifa->ifa_dstaddr))) 1316 return ifa; 1317 continue; 1318 } 1319 cp = addr->sa_data; 1320 cp2 = ifa->ifa_addr->sa_data; 1321 cp3 = ifa->ifa_netmask->sa_data; 1322 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1323 for (; cp3 < cplim; cp3++) { 1324 if ((*cp++ ^ *cp2++) & *cp3) 1325 break; 1326 } 1327 if (cp3 == cplim) 1328 return ifa; 1329 } 1330 return ifa_maybe; 1331 } 1332 1333 /* 1334 * Default action when installing a route with a Link Level gateway. 1335 * Lookup an appropriate real ifa to point to. 1336 * This should be moved to /sys/net/link.c eventually. 1337 */ 1338 void 1339 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1340 { 1341 struct ifaddr *ifa; 1342 const struct sockaddr *dst; 1343 struct ifnet *ifp; 1344 1345 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1346 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1347 return; 1348 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1349 rt_replace_ifa(rt, ifa); 1350 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1351 ifa->ifa_rtrequest(cmd, rt, info); 1352 } 1353 } 1354 1355 /* 1356 * Handle a change in the interface link state. 1357 * XXX: We should listen to the routing socket in-kernel rather 1358 * than calling in6_if_link_* functions directly from here. 1359 */ 1360 void 1361 if_link_state_change(struct ifnet *ifp, int link_state) 1362 { 1363 int s; 1364 #if defined(DEBUG) || defined(INET6) 1365 int old_link_state; 1366 #endif 1367 1368 s = splnet(); 1369 if (ifp->if_link_state == link_state) { 1370 splx(s); 1371 return; 1372 } 1373 1374 #if defined(DEBUG) || defined(INET6) 1375 old_link_state = ifp->if_link_state; 1376 #endif 1377 ifp->if_link_state = link_state; 1378 #ifdef DEBUG 1379 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1380 link_state == LINK_STATE_UP ? "UP" : 1381 link_state == LINK_STATE_DOWN ? "DOWN" : 1382 "UNKNOWN", 1383 old_link_state == LINK_STATE_UP ? "UP" : 1384 old_link_state == LINK_STATE_DOWN ? "DOWN" : 1385 "UNKNOWN"); 1386 #endif 1387 1388 #ifdef INET6 1389 /* 1390 * When going from UNKNOWN to UP, we need to mark existing 1391 * IPv6 addresses as tentative and restart DAD as we may have 1392 * erroneously not found a duplicate. 1393 * 1394 * This needs to happen before rt_ifmsg to avoid a race where 1395 * listeners would have an address and expect it to work right 1396 * away. 1397 */ 1398 if (in6_present && link_state == LINK_STATE_UP && 1399 old_link_state == LINK_STATE_UNKNOWN) 1400 in6_if_link_down(ifp); 1401 #endif 1402 1403 /* Notify that the link state has changed. */ 1404 rt_ifmsg(ifp); 1405 1406 #if NCARP > 0 1407 if (ifp->if_carp) 1408 carp_carpdev_state(ifp); 1409 #endif 1410 1411 #ifdef INET6 1412 if (in6_present) { 1413 if (link_state == LINK_STATE_DOWN) 1414 in6_if_link_down(ifp); 1415 else if (link_state == LINK_STATE_UP) 1416 in6_if_link_up(ifp); 1417 } 1418 #endif 1419 1420 splx(s); 1421 } 1422 1423 /* 1424 * Mark an interface down and notify protocols of 1425 * the transition. 1426 * NOTE: must be called at splsoftnet or equivalent. 1427 */ 1428 void 1429 if_down(struct ifnet *ifp) 1430 { 1431 struct ifaddr *ifa; 1432 1433 ifp->if_flags &= ~IFF_UP; 1434 nanotime(&ifp->if_lastchange); 1435 IFADDR_FOREACH(ifa, ifp) 1436 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1437 IFQ_PURGE(&ifp->if_snd); 1438 #if NCARP > 0 1439 if (ifp->if_carp) 1440 carp_carpdev_state(ifp); 1441 #endif 1442 rt_ifmsg(ifp); 1443 #ifdef INET6 1444 if (in6_present) 1445 in6_if_down(ifp); 1446 #endif 1447 } 1448 1449 /* 1450 * Mark an interface up and notify protocols of 1451 * the transition. 1452 * NOTE: must be called at splsoftnet or equivalent. 1453 */ 1454 void 1455 if_up(struct ifnet *ifp) 1456 { 1457 #ifdef notyet 1458 struct ifaddr *ifa; 1459 #endif 1460 1461 ifp->if_flags |= IFF_UP; 1462 nanotime(&ifp->if_lastchange); 1463 #ifdef notyet 1464 /* this has no effect on IP, and will kill all ISO connections XXX */ 1465 IFADDR_FOREACH(ifa, ifp) 1466 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1467 #endif 1468 #if NCARP > 0 1469 if (ifp->if_carp) 1470 carp_carpdev_state(ifp); 1471 #endif 1472 rt_ifmsg(ifp); 1473 #ifdef INET6 1474 if (in6_present) 1475 in6_if_up(ifp); 1476 #endif 1477 } 1478 1479 /* 1480 * Handle interface watchdog timer routines. Called 1481 * from softclock, we decrement timers (if set) and 1482 * call the appropriate interface routine on expiration. 1483 */ 1484 void 1485 if_slowtimo(void *arg) 1486 { 1487 struct ifnet *ifp; 1488 int s = splnet(); 1489 1490 IFNET_FOREACH(ifp) { 1491 if (ifp->if_timer == 0 || --ifp->if_timer) 1492 continue; 1493 if (ifp->if_watchdog != NULL) 1494 (*ifp->if_watchdog)(ifp); 1495 } 1496 splx(s); 1497 callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL); 1498 } 1499 1500 /* 1501 * Set/clear promiscuous mode on interface ifp based on the truth value 1502 * of pswitch. The calls are reference counted so that only the first 1503 * "on" request actually has an effect, as does the final "off" request. 1504 * Results are undefined if the "off" and "on" requests are not matched. 1505 */ 1506 int 1507 ifpromisc(struct ifnet *ifp, int pswitch) 1508 { 1509 int pcount, ret; 1510 short nflags; 1511 1512 pcount = ifp->if_pcount; 1513 if (pswitch) { 1514 /* 1515 * Allow the device to be "placed" into promiscuous 1516 * mode even if it is not configured up. It will 1517 * consult IFF_PROMISC when it is brought up. 1518 */ 1519 if (ifp->if_pcount++ != 0) 1520 return 0; 1521 nflags = ifp->if_flags | IFF_PROMISC; 1522 } else { 1523 if (--ifp->if_pcount > 0) 1524 return 0; 1525 nflags = ifp->if_flags & ~IFF_PROMISC; 1526 } 1527 ret = if_flags_set(ifp, nflags); 1528 /* Restore interface state if not successful. */ 1529 if (ret != 0) { 1530 ifp->if_pcount = pcount; 1531 } 1532 return ret; 1533 } 1534 1535 /* 1536 * Map interface name to 1537 * interface structure pointer. 1538 */ 1539 struct ifnet * 1540 ifunit(const char *name) 1541 { 1542 struct ifnet *ifp; 1543 const char *cp = name; 1544 u_int unit = 0; 1545 u_int i; 1546 1547 /* 1548 * If the entire name is a number, treat it as an ifindex. 1549 */ 1550 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 1551 unit = unit * 10 + (*cp - '0'); 1552 } 1553 1554 /* 1555 * If the number took all of the name, then it's a valid ifindex. 1556 */ 1557 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 1558 if (unit >= if_indexlim) 1559 return NULL; 1560 ifp = ifindex2ifnet[unit]; 1561 if (ifp == NULL || ifp->if_output == if_nulloutput) 1562 return NULL; 1563 return ifp; 1564 } 1565 1566 IFNET_FOREACH(ifp) { 1567 if (ifp->if_output == if_nulloutput) 1568 continue; 1569 if (strcmp(ifp->if_xname, name) == 0) 1570 return ifp; 1571 } 1572 return NULL; 1573 } 1574 1575 ifnet_t * 1576 if_byindex(u_int idx) 1577 { 1578 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 1579 } 1580 1581 /* common */ 1582 int 1583 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 1584 { 1585 int s; 1586 struct ifreq *ifr; 1587 struct ifcapreq *ifcr; 1588 struct ifdatareq *ifdr; 1589 1590 switch (cmd) { 1591 case SIOCSIFCAP: 1592 ifcr = data; 1593 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 1594 return EINVAL; 1595 1596 if (ifcr->ifcr_capenable == ifp->if_capenable) 1597 return 0; 1598 1599 ifp->if_capenable = ifcr->ifcr_capenable; 1600 1601 /* Pre-compute the checksum flags mask. */ 1602 ifp->if_csum_flags_tx = 0; 1603 ifp->if_csum_flags_rx = 0; 1604 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 1605 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 1606 } 1607 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 1608 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 1609 } 1610 1611 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 1612 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 1613 } 1614 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 1615 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 1616 } 1617 1618 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 1619 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 1620 } 1621 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 1622 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 1623 } 1624 1625 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 1626 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 1627 } 1628 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 1629 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 1630 } 1631 1632 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 1633 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 1634 } 1635 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 1636 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 1637 } 1638 if (ifp->if_flags & IFF_UP) 1639 return ENETRESET; 1640 return 0; 1641 case SIOCSIFFLAGS: 1642 ifr = data; 1643 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 1644 s = splnet(); 1645 if_down(ifp); 1646 splx(s); 1647 } 1648 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 1649 s = splnet(); 1650 if_up(ifp); 1651 splx(s); 1652 } 1653 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1654 (ifr->ifr_flags &~ IFF_CANTCHANGE); 1655 break; 1656 case SIOCGIFFLAGS: 1657 ifr = data; 1658 ifr->ifr_flags = ifp->if_flags; 1659 break; 1660 1661 case SIOCGIFMETRIC: 1662 ifr = data; 1663 ifr->ifr_metric = ifp->if_metric; 1664 break; 1665 1666 case SIOCGIFMTU: 1667 ifr = data; 1668 ifr->ifr_mtu = ifp->if_mtu; 1669 break; 1670 1671 case SIOCGIFDLT: 1672 ifr = data; 1673 ifr->ifr_dlt = ifp->if_dlt; 1674 break; 1675 1676 case SIOCGIFCAP: 1677 ifcr = data; 1678 ifcr->ifcr_capabilities = ifp->if_capabilities; 1679 ifcr->ifcr_capenable = ifp->if_capenable; 1680 break; 1681 1682 case SIOCSIFMETRIC: 1683 ifr = data; 1684 ifp->if_metric = ifr->ifr_metric; 1685 break; 1686 1687 case SIOCGIFDATA: 1688 ifdr = data; 1689 ifdr->ifdr_data = ifp->if_data; 1690 break; 1691 1692 case SIOCGIFINDEX: 1693 ifr = data; 1694 ifr->ifr_index = ifp->if_index; 1695 break; 1696 1697 case SIOCZIFDATA: 1698 ifdr = data; 1699 ifdr->ifdr_data = ifp->if_data; 1700 /* 1701 * Assumes that the volatile counters that can be 1702 * zero'ed are at the end of if_data. 1703 */ 1704 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 1705 offsetof(struct if_data, ifi_ipackets)); 1706 /* 1707 * The memset() clears to the bottm of if_data. In the area, 1708 * if_lastchange is included. Please be careful if new entry 1709 * will be added into if_data or rewite this. 1710 * 1711 * And also, update if_lastchnage. 1712 */ 1713 getnanotime(&ifp->if_lastchange); 1714 break; 1715 case SIOCSIFMTU: 1716 ifr = data; 1717 if (ifp->if_mtu == ifr->ifr_mtu) 1718 break; 1719 ifp->if_mtu = ifr->ifr_mtu; 1720 /* 1721 * If the link MTU changed, do network layer specific procedure. 1722 */ 1723 #ifdef INET6 1724 if (in6_present) 1725 nd6_setmtu(ifp); 1726 #endif 1727 return ENETRESET; 1728 default: 1729 return ENOTTY; 1730 } 1731 return 0; 1732 } 1733 1734 int 1735 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 1736 { 1737 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 1738 struct ifaddr *ifa; 1739 const struct sockaddr *any, *sa; 1740 union { 1741 struct sockaddr sa; 1742 struct sockaddr_storage ss; 1743 } u, v; 1744 1745 switch (cmd) { 1746 case SIOCSIFADDRPREF: 1747 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 1748 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1749 NULL) != 0) 1750 return EPERM; 1751 case SIOCGIFADDRPREF: 1752 break; 1753 default: 1754 return EOPNOTSUPP; 1755 } 1756 1757 /* sanity checks */ 1758 if (data == NULL || ifp == NULL) { 1759 panic("invalid argument to %s", __func__); 1760 /*NOTREACHED*/ 1761 } 1762 1763 /* address must be specified on ADD and DELETE */ 1764 sa = sstocsa(&ifap->ifap_addr); 1765 if (sa->sa_family != sofamily(so)) 1766 return EINVAL; 1767 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 1768 return EINVAL; 1769 1770 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 1771 1772 IFADDR_FOREACH(ifa, ifp) { 1773 if (ifa->ifa_addr->sa_family != sa->sa_family) 1774 continue; 1775 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 1776 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 1777 break; 1778 } 1779 if (ifa == NULL) 1780 return EADDRNOTAVAIL; 1781 1782 switch (cmd) { 1783 case SIOCSIFADDRPREF: 1784 ifa->ifa_preference = ifap->ifap_preference; 1785 return 0; 1786 case SIOCGIFADDRPREF: 1787 /* fill in the if_laddrreq structure */ 1788 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 1789 sizeof(ifap->ifap_addr), ifa->ifa_addr); 1790 ifap->ifap_preference = ifa->ifa_preference; 1791 return 0; 1792 default: 1793 return EOPNOTSUPP; 1794 } 1795 } 1796 1797 static void 1798 ifnet_lock_enter(struct ifnet_lock *il) 1799 { 1800 uint64_t *nenter; 1801 1802 /* Before trying to acquire the mutex, increase the count of threads 1803 * who have entered or who wait to enter the critical section. 1804 * Avoid one costly locked memory transaction by keeping a count for 1805 * each CPU. 1806 */ 1807 nenter = percpu_getref(il->il_nenter); 1808 (*nenter)++; 1809 percpu_putref(il->il_nenter); 1810 mutex_enter(&il->il_lock); 1811 } 1812 1813 static void 1814 ifnet_lock_exit(struct ifnet_lock *il) 1815 { 1816 /* Increase the count of threads who have exited the critical 1817 * section. Increase while we still hold the lock. 1818 */ 1819 il->il_nexit++; 1820 mutex_exit(&il->il_lock); 1821 } 1822 1823 /* 1824 * Interface ioctls. 1825 */ 1826 static int 1827 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 1828 { 1829 struct ifnet *ifp; 1830 struct ifreq *ifr; 1831 int error = 0; 1832 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 1833 u_long ocmd = cmd; 1834 #endif 1835 short oif_flags; 1836 #ifdef COMPAT_OIFREQ 1837 struct ifreq ifrb; 1838 struct oifreq *oifr = NULL; 1839 #endif 1840 1841 switch (cmd) { 1842 #ifdef COMPAT_OIFREQ 1843 case OSIOCGIFCONF: 1844 case OOSIOCGIFCONF: 1845 return compat_ifconf(cmd, data); 1846 #endif 1847 #ifdef COMPAT_OIFDATA 1848 case OSIOCGIFDATA: 1849 case OSIOCZIFDATA: 1850 return compat_ifdatareq(l, cmd, data); 1851 #endif 1852 case SIOCGIFCONF: 1853 return ifconf(cmd, data); 1854 case SIOCINITIFADDR: 1855 return EPERM; 1856 } 1857 1858 #ifdef COMPAT_OIFREQ 1859 cmd = compat_cvtcmd(cmd); 1860 if (cmd != ocmd) { 1861 oifr = data; 1862 data = ifr = &ifrb; 1863 ifreqo2n(oifr, ifr); 1864 } else 1865 #endif 1866 ifr = data; 1867 1868 ifp = ifunit(ifr->ifr_name); 1869 1870 switch (cmd) { 1871 case SIOCIFCREATE: 1872 case SIOCIFDESTROY: 1873 if (l != NULL) { 1874 error = kauth_authorize_network(l->l_cred, 1875 KAUTH_NETWORK_INTERFACE, 1876 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 1877 (void *)cmd, NULL); 1878 if (error != 0) 1879 return error; 1880 } 1881 return (cmd == SIOCIFCREATE) ? 1882 if_clone_create(ifr->ifr_name) : 1883 if_clone_destroy(ifr->ifr_name); 1884 1885 case SIOCIFGCLONERS: 1886 return if_clone_list((struct if_clonereq *)data); 1887 } 1888 1889 if (ifp == NULL) 1890 return ENXIO; 1891 1892 switch (cmd) { 1893 case SIOCALIFADDR: 1894 case SIOCDLIFADDR: 1895 case SIOCSIFADDRPREF: 1896 case SIOCSIFFLAGS: 1897 case SIOCSIFCAP: 1898 case SIOCSIFMETRIC: 1899 case SIOCZIFDATA: 1900 case SIOCSIFMTU: 1901 case SIOCSIFPHYADDR: 1902 case SIOCDIFPHYADDR: 1903 #ifdef INET6 1904 case SIOCSIFPHYADDR_IN6: 1905 #endif 1906 case SIOCSLIFPHYADDR: 1907 case SIOCADDMULTI: 1908 case SIOCDELMULTI: 1909 case SIOCSIFMEDIA: 1910 case SIOCSDRVSPEC: 1911 case SIOCG80211: 1912 case SIOCS80211: 1913 case SIOCS80211NWID: 1914 case SIOCS80211NWKEY: 1915 case SIOCS80211POWER: 1916 case SIOCS80211BSSID: 1917 case SIOCS80211CHANNEL: 1918 case SIOCSLINKSTR: 1919 if (l != NULL) { 1920 error = kauth_authorize_network(l->l_cred, 1921 KAUTH_NETWORK_INTERFACE, 1922 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 1923 (void *)cmd, NULL); 1924 if (error != 0) 1925 return error; 1926 } 1927 } 1928 1929 oif_flags = ifp->if_flags; 1930 1931 ifnet_lock_enter(ifp->if_ioctl_lock); 1932 error = (*ifp->if_ioctl)(ifp, cmd, data); 1933 if (error != ENOTTY) 1934 ; 1935 else if (so->so_proto == NULL) 1936 error = EOPNOTSUPP; 1937 else { 1938 #ifdef COMPAT_OSOCK 1939 error = compat_ifioctl(so, ocmd, cmd, data, l); 1940 #else 1941 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 1942 cmd, data, ifp); 1943 #endif 1944 } 1945 1946 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 1947 #ifdef INET6 1948 if (in6_present && (ifp->if_flags & IFF_UP) != 0) { 1949 int s = splnet(); 1950 in6_if_up(ifp); 1951 splx(s); 1952 } 1953 #endif 1954 } 1955 #ifdef COMPAT_OIFREQ 1956 if (cmd != ocmd) 1957 ifreqn2o(oifr, ifr); 1958 #endif 1959 1960 ifnet_lock_exit(ifp->if_ioctl_lock); 1961 return error; 1962 } 1963 1964 /* This callback adds to the sum in `arg' the number of 1965 * threads on `ci' who have entered or who wait to enter the 1966 * critical section. 1967 */ 1968 static void 1969 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci) 1970 { 1971 uint64_t *sum = arg, *nenter = p; 1972 1973 *sum += *nenter; 1974 } 1975 1976 /* Return the number of threads who have entered or who wait 1977 * to enter the critical section on all CPUs. 1978 */ 1979 static uint64_t 1980 ifnet_lock_entrances(struct ifnet_lock *il) 1981 { 1982 uint64_t sum = 0; 1983 1984 percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum); 1985 1986 return sum; 1987 } 1988 1989 static int 1990 ifioctl_attach(struct ifnet *ifp) 1991 { 1992 struct ifnet_lock *il; 1993 1994 /* If the driver has not supplied its own if_ioctl, then 1995 * supply the default. 1996 */ 1997 if (ifp->if_ioctl == NULL) 1998 ifp->if_ioctl = ifioctl_common; 1999 2000 /* Create an ifnet_lock for synchronizing ifioctls. */ 2001 if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL) 2002 return ENOMEM; 2003 2004 il->il_nenter = percpu_alloc(sizeof(uint64_t)); 2005 if (il->il_nenter == NULL) { 2006 kmem_free(il, sizeof(*il)); 2007 return ENOMEM; 2008 } 2009 2010 mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE); 2011 cv_init(&il->il_emptied, ifp->if_xname); 2012 2013 ifp->if_ioctl_lock = il; 2014 2015 return 0; 2016 } 2017 2018 /* 2019 * This must not be called until after `ifp' has been withdrawn from the 2020 * ifnet tables so that ifioctl() cannot get a handle on it by calling 2021 * ifunit(). 2022 */ 2023 static void 2024 ifioctl_detach(struct ifnet *ifp) 2025 { 2026 struct ifnet_lock *il; 2027 2028 il = ifp->if_ioctl_lock; 2029 mutex_enter(&il->il_lock); 2030 /* Install if_nullioctl to make sure that any thread that 2031 * subsequently enters the critical section will quit it 2032 * immediately and signal the condition variable that we 2033 * wait on, below. 2034 */ 2035 ifp->if_ioctl = if_nullioctl; 2036 /* Sleep while threads are still in the critical section or 2037 * wait to enter it. 2038 */ 2039 while (ifnet_lock_entrances(il) != il->il_nexit) 2040 cv_wait(&il->il_emptied, &il->il_lock); 2041 /* At this point, we are the only thread still in the critical 2042 * section, and no new thread can get a handle on the ifioctl 2043 * lock, so it is safe to free its memory. 2044 */ 2045 mutex_exit(&il->il_lock); 2046 ifp->if_ioctl_lock = NULL; 2047 percpu_free(il->il_nenter, sizeof(uint64_t)); 2048 il->il_nenter = NULL; 2049 cv_destroy(&il->il_emptied); 2050 mutex_destroy(&il->il_lock); 2051 kmem_free(il, sizeof(*il)); 2052 } 2053 2054 /* 2055 * Return interface configuration 2056 * of system. List may be used 2057 * in later ioctl's (above) to get 2058 * other information. 2059 * 2060 * Each record is a struct ifreq. Before the addition of 2061 * sockaddr_storage, the API rule was that sockaddr flavors that did 2062 * not fit would extend beyond the struct ifreq, with the next struct 2063 * ifreq starting sa_len beyond the struct sockaddr. Because the 2064 * union in struct ifreq includes struct sockaddr_storage, every kind 2065 * of sockaddr must fit. Thus, there are no longer any overlength 2066 * records. 2067 * 2068 * Records are added to the user buffer if they fit, and ifc_len is 2069 * adjusted to the length that was written. Thus, the user is only 2070 * assured of getting the complete list if ifc_len on return is at 2071 * least sizeof(struct ifreq) less than it was on entry. 2072 * 2073 * If the user buffer pointer is NULL, this routine copies no data and 2074 * returns the amount of space that would be needed. 2075 * 2076 * Invariants: 2077 * ifrp points to the next part of the user's buffer to be used. If 2078 * ifrp != NULL, space holds the number of bytes remaining that we may 2079 * write at ifrp. Otherwise, space holds the number of bytes that 2080 * would have been written had there been adequate space. 2081 */ 2082 /*ARGSUSED*/ 2083 int 2084 ifconf(u_long cmd, void *data) 2085 { 2086 struct ifconf *ifc = (struct ifconf *)data; 2087 struct ifnet *ifp; 2088 struct ifaddr *ifa; 2089 struct ifreq ifr, *ifrp; 2090 int space, error = 0; 2091 const int sz = (int)sizeof(struct ifreq); 2092 2093 if ((ifrp = ifc->ifc_req) == NULL) 2094 space = 0; 2095 else 2096 space = ifc->ifc_len; 2097 IFNET_FOREACH(ifp) { 2098 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2099 sizeof(ifr.ifr_name)); 2100 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') 2101 return ENAMETOOLONG; 2102 if (IFADDR_EMPTY(ifp)) { 2103 /* Interface with no addresses - send zero sockaddr. */ 2104 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2105 if (ifrp == NULL) { 2106 space += sz; 2107 continue; 2108 } 2109 if (space >= sz) { 2110 error = copyout(&ifr, ifrp, sz); 2111 if (error != 0) 2112 return error; 2113 ifrp++; 2114 space -= sz; 2115 } 2116 } 2117 2118 IFADDR_FOREACH(ifa, ifp) { 2119 struct sockaddr *sa = ifa->ifa_addr; 2120 /* all sockaddrs must fit in sockaddr_storage */ 2121 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2122 2123 if (ifrp == NULL) { 2124 space += sz; 2125 continue; 2126 } 2127 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2128 if (space >= sz) { 2129 error = copyout(&ifr, ifrp, sz); 2130 if (error != 0) 2131 return (error); 2132 ifrp++; space -= sz; 2133 } 2134 } 2135 } 2136 if (ifrp != NULL) { 2137 KASSERT(0 <= space && space <= ifc->ifc_len); 2138 ifc->ifc_len -= space; 2139 } else { 2140 KASSERT(space >= 0); 2141 ifc->ifc_len = space; 2142 } 2143 return (0); 2144 } 2145 2146 int 2147 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2148 { 2149 uint8_t len; 2150 #ifdef COMPAT_OIFREQ 2151 struct ifreq ifrb; 2152 struct oifreq *oifr = NULL; 2153 u_long ocmd = cmd; 2154 cmd = compat_cvtcmd(cmd); 2155 if (cmd != ocmd) { 2156 oifr = (struct oifreq *)(void *)ifr; 2157 ifr = &ifrb; 2158 ifreqo2n(oifr, ifr); 2159 len = sizeof(oifr->ifr_addr); 2160 } else 2161 #endif 2162 len = sizeof(ifr->ifr_ifru.ifru_space); 2163 2164 if (len < sa->sa_len) 2165 return EFBIG; 2166 2167 memset(&ifr->ifr_addr, 0, len); 2168 sockaddr_copy(&ifr->ifr_addr, len, sa); 2169 2170 #ifdef COMPAT_OIFREQ 2171 if (cmd != ocmd) 2172 ifreqn2o(oifr, ifr); 2173 #endif 2174 return 0; 2175 } 2176 2177 /* 2178 * Queue message on interface, and start output if interface 2179 * not yet active. 2180 */ 2181 int 2182 ifq_enqueue(struct ifnet *ifp, struct mbuf *m 2183 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2184 { 2185 int len = m->m_pkthdr.len; 2186 int mflags = m->m_flags; 2187 int s = splnet(); 2188 int error; 2189 2190 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2191 if (error != 0) 2192 goto out; 2193 ifp->if_obytes += len; 2194 if (mflags & M_MCAST) 2195 ifp->if_omcasts++; 2196 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2197 (*ifp->if_start)(ifp); 2198 out: 2199 splx(s); 2200 return error; 2201 } 2202 2203 /* 2204 * Queue message on interface, possibly using a second fast queue 2205 */ 2206 int 2207 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m 2208 ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr)) 2209 { 2210 int error = 0; 2211 2212 if (ifq != NULL 2213 #ifdef ALTQ 2214 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2215 #endif 2216 ) { 2217 if (IF_QFULL(ifq)) { 2218 IF_DROP(&ifp->if_snd); 2219 m_freem(m); 2220 if (error == 0) 2221 error = ENOBUFS; 2222 } else 2223 IF_ENQUEUE(ifq, m); 2224 } else 2225 IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error); 2226 if (error != 0) { 2227 ++ifp->if_oerrors; 2228 return error; 2229 } 2230 return 0; 2231 } 2232 2233 int 2234 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2235 { 2236 int rc; 2237 2238 if (ifp->if_initaddr != NULL) 2239 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2240 else if (src || 2241 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2242 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2243 2244 return rc; 2245 } 2246 2247 int 2248 if_flags_set(ifnet_t *ifp, const short flags) 2249 { 2250 int rc; 2251 2252 if (ifp->if_setflags != NULL) 2253 rc = (*ifp->if_setflags)(ifp, flags); 2254 else { 2255 short cantflags, chgdflags; 2256 struct ifreq ifr; 2257 2258 chgdflags = ifp->if_flags ^ flags; 2259 cantflags = chgdflags & IFF_CANTCHANGE; 2260 2261 if (cantflags != 0) 2262 ifp->if_flags ^= cantflags; 2263 2264 /* Traditionally, we do not call if_ioctl after 2265 * setting/clearing only IFF_PROMISC if the interface 2266 * isn't IFF_UP. Uphold that tradition. 2267 */ 2268 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2269 return 0; 2270 2271 memset(&ifr, 0, sizeof(ifr)); 2272 2273 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2274 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2275 2276 if (rc != 0 && cantflags != 0) 2277 ifp->if_flags ^= cantflags; 2278 } 2279 2280 return rc; 2281 } 2282 2283 int 2284 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2285 { 2286 int rc; 2287 struct ifreq ifr; 2288 2289 if (ifp->if_mcastop != NULL) 2290 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2291 else { 2292 ifreq_setaddr(cmd, &ifr, sa); 2293 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2294 } 2295 2296 return rc; 2297 } 2298 2299 static void 2300 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2301 struct ifaltq *ifq) 2302 { 2303 const struct sysctlnode *cnode, *rnode; 2304 2305 if (sysctl_createv(clog, 0, NULL, &rnode, 2306 CTLFLAG_PERMANENT, 2307 CTLTYPE_NODE, "interfaces", 2308 SYSCTL_DESCR("Per-interface controls"), 2309 NULL, 0, NULL, 0, 2310 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 2311 goto bad; 2312 2313 if (sysctl_createv(clog, 0, &rnode, &rnode, 2314 CTLFLAG_PERMANENT, 2315 CTLTYPE_NODE, ifname, 2316 SYSCTL_DESCR("Interface controls"), 2317 NULL, 0, NULL, 0, 2318 CTL_CREATE, CTL_EOL) != 0) 2319 goto bad; 2320 2321 if (sysctl_createv(clog, 0, &rnode, &rnode, 2322 CTLFLAG_PERMANENT, 2323 CTLTYPE_NODE, "sndq", 2324 SYSCTL_DESCR("Interface output queue controls"), 2325 NULL, 0, NULL, 0, 2326 CTL_CREATE, CTL_EOL) != 0) 2327 goto bad; 2328 2329 if (sysctl_createv(clog, 0, &rnode, &cnode, 2330 CTLFLAG_PERMANENT, 2331 CTLTYPE_INT, "len", 2332 SYSCTL_DESCR("Current output queue length"), 2333 NULL, 0, &ifq->ifq_len, 0, 2334 CTL_CREATE, CTL_EOL) != 0) 2335 goto bad; 2336 2337 if (sysctl_createv(clog, 0, &rnode, &cnode, 2338 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2339 CTLTYPE_INT, "maxlen", 2340 SYSCTL_DESCR("Maximum allowed output queue length"), 2341 NULL, 0, &ifq->ifq_maxlen, 0, 2342 CTL_CREATE, CTL_EOL) != 0) 2343 goto bad; 2344 2345 if (sysctl_createv(clog, 0, &rnode, &cnode, 2346 CTLFLAG_PERMANENT, 2347 CTLTYPE_INT, "drops", 2348 SYSCTL_DESCR("Packets dropped due to full output queue"), 2349 NULL, 0, &ifq->ifq_drops, 0, 2350 CTL_CREATE, CTL_EOL) != 0) 2351 goto bad; 2352 2353 return; 2354 bad: 2355 printf("%s: could not attach sysctl nodes\n", ifname); 2356 return; 2357 } 2358 2359 #if defined(INET) || defined(INET6) 2360 2361 #define SYSCTL_NET_PKTQ(q, cn, c) \ 2362 static int \ 2363 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 2364 { \ 2365 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 2366 } 2367 2368 #if defined(INET) 2369 static int 2370 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 2371 { 2372 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 2373 } 2374 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 2375 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 2376 #endif 2377 2378 #if defined(INET6) 2379 static int 2380 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 2381 { 2382 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 2383 } 2384 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 2385 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 2386 #endif 2387 2388 static void 2389 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 2390 { 2391 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 2392 const char *pfname = NULL, *ipname = NULL; 2393 int ipn = 0, qid = 0; 2394 2395 switch (pf) { 2396 #if defined(INET) 2397 case PF_INET: 2398 len_func = sysctl_net_ip_pktq_items; 2399 maxlen_func = sysctl_net_ip_pktq_maxlen; 2400 drops_func = sysctl_net_ip_pktq_drops; 2401 pfname = "inet", ipn = IPPROTO_IP; 2402 ipname = "ip", qid = IPCTL_IFQ; 2403 break; 2404 #endif 2405 #if defined(INET6) 2406 case PF_INET6: 2407 len_func = sysctl_net_ip6_pktq_items; 2408 maxlen_func = sysctl_net_ip6_pktq_maxlen; 2409 drops_func = sysctl_net_ip6_pktq_drops; 2410 pfname = "inet6", ipn = IPPROTO_IPV6; 2411 ipname = "ip6", qid = IPV6CTL_IFQ; 2412 break; 2413 #endif 2414 default: 2415 KASSERT(false); 2416 } 2417 2418 sysctl_createv(clog, 0, NULL, NULL, 2419 CTLFLAG_PERMANENT, 2420 CTLTYPE_NODE, pfname, NULL, 2421 NULL, 0, NULL, 0, 2422 CTL_NET, pf, CTL_EOL); 2423 sysctl_createv(clog, 0, NULL, NULL, 2424 CTLFLAG_PERMANENT, 2425 CTLTYPE_NODE, ipname, NULL, 2426 NULL, 0, NULL, 0, 2427 CTL_NET, pf, ipn, CTL_EOL); 2428 sysctl_createv(clog, 0, NULL, NULL, 2429 CTLFLAG_PERMANENT, 2430 CTLTYPE_NODE, "ifq", 2431 SYSCTL_DESCR("Protocol input queue controls"), 2432 NULL, 0, NULL, 0, 2433 CTL_NET, pf, ipn, qid, CTL_EOL); 2434 2435 sysctl_createv(clog, 0, NULL, NULL, 2436 CTLFLAG_PERMANENT, 2437 CTLTYPE_INT, "len", 2438 SYSCTL_DESCR("Current input queue length"), 2439 len_func, 0, NULL, 0, 2440 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 2441 sysctl_createv(clog, 0, NULL, NULL, 2442 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2443 CTLTYPE_INT, "maxlen", 2444 SYSCTL_DESCR("Maximum allowed input queue length"), 2445 maxlen_func, 0, NULL, 0, 2446 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 2447 sysctl_createv(clog, 0, NULL, NULL, 2448 CTLFLAG_PERMANENT, 2449 CTLTYPE_INT, "drops", 2450 SYSCTL_DESCR("Packets dropped due to full input queue"), 2451 drops_func, 0, NULL, 0, 2452 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 2453 } 2454 #endif /* INET || INET6 */ 2455 2456 static int 2457 if_sdl_sysctl(SYSCTLFN_ARGS) 2458 { 2459 struct ifnet *ifp; 2460 const struct sockaddr_dl *sdl; 2461 2462 if (namelen != 1) 2463 return EINVAL; 2464 2465 ifp = if_byindex(name[0]); 2466 if (ifp == NULL) 2467 return ENODEV; 2468 2469 sdl = ifp->if_sadl; 2470 if (sdl == NULL) { 2471 *oldlenp = 0; 2472 return 0; 2473 } 2474 2475 if (oldp == NULL) { 2476 *oldlenp = sdl->sdl_alen; 2477 return 0; 2478 } 2479 2480 if (*oldlenp >= sdl->sdl_alen) 2481 *oldlenp = sdl->sdl_alen; 2482 return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 2483 } 2484 2485 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 2486 { 2487 const struct sysctlnode *rnode = NULL; 2488 2489 sysctl_createv(clog, 0, NULL, &rnode, 2490 CTLFLAG_PERMANENT, 2491 CTLTYPE_NODE, "sdl", 2492 SYSCTL_DESCR("Get active link-layer address"), 2493 if_sdl_sysctl, 0, NULL, 0, 2494 CTL_NET, CTL_CREATE, CTL_EOL); 2495 } 2496
Indexes created Thu Sep 25 02:09:54 GMT 2025