if.c revision 1.332
1 /* $NetBSD: if.c,v 1.332 2016/04/28 01:37:17 knakahara 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 1.332 2016/04/28 01:37:17 knakahara Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 98 #include "opt_atalk.h" 99 #include "opt_natm.h" 100 #include "opt_wlan.h" 101 #include "opt_net_mpsafe.h" 102 #endif 103 104 #include <sys/param.h> 105 #include <sys/mbuf.h> 106 #include <sys/systm.h> 107 #include <sys/callout.h> 108 #include <sys/proc.h> 109 #include <sys/socket.h> 110 #include <sys/socketvar.h> 111 #include <sys/domain.h> 112 #include <sys/protosw.h> 113 #include <sys/kernel.h> 114 #include <sys/ioctl.h> 115 #include <sys/sysctl.h> 116 #include <sys/syslog.h> 117 #include <sys/kauth.h> 118 #include <sys/kmem.h> 119 #include <sys/xcall.h> 120 #include <sys/cpu.h> 121 #include <sys/intr.h> 122 123 #include <net/if.h> 124 #include <net/if_dl.h> 125 #include <net/if_ether.h> 126 #include <net/if_media.h> 127 #include <net80211/ieee80211.h> 128 #include <net80211/ieee80211_ioctl.h> 129 #include <net/if_types.h> 130 #include <net/route.h> 131 #include <net/netisr.h> 132 #include <sys/module.h> 133 #ifdef NETATALK 134 #include <netatalk/at_extern.h> 135 #include <netatalk/at.h> 136 #endif 137 #include <net/pfil.h> 138 #include <netinet/in.h> 139 #include <netinet/in_var.h> 140 141 #ifdef INET6 142 #include <netinet6/in6_var.h> 143 #include <netinet6/nd6.h> 144 #endif 145 146 #include "ether.h" 147 #include "fddi.h" 148 #include "token.h" 149 150 #include "carp.h" 151 #if NCARP > 0 152 #include <netinet/ip_carp.h> 153 #endif 154 155 #include <compat/sys/sockio.h> 156 #include <compat/sys/socket.h> 157 158 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 159 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 160 161 /* 162 * Global list of interfaces. 163 */ 164 struct ifnet_head ifnet_list; 165 static ifnet_t ** ifindex2ifnet = NULL; 166 167 static u_int if_index = 1; 168 static size_t if_indexlim = 0; 169 static uint64_t index_gen; 170 static kmutex_t index_gen_mtx; 171 static kmutex_t if_clone_mtx; 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 *); 179 180 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 static void if_slowtimo(void *); 197 static void if_free_sadl(struct ifnet *); 198 static void if_attachdomain1(struct ifnet *); 199 static int ifconf(u_long, void *); 200 static int if_clone_create(const char *); 201 static int if_clone_destroy(const char *); 202 static void if_link_state_change_si(void *); 203 204 struct if_percpuq { 205 struct ifnet *ipq_ifp; 206 void *ipq_si; 207 struct percpu *ipq_ifqs; /* struct ifqueue */ 208 }; 209 210 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 211 212 static void if_percpuq_drops(void *, void *, struct cpu_info *); 213 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 214 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 215 struct if_percpuq *); 216 217 #if defined(INET) || defined(INET6) 218 static void sysctl_net_pktq_setup(struct sysctllog **, int); 219 #endif 220 221 static int 222 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 223 void *arg0, void *arg1, void *arg2, void *arg3) 224 { 225 int result; 226 enum kauth_network_req req; 227 228 result = KAUTH_RESULT_DEFER; 229 req = (enum kauth_network_req)arg1; 230 231 if (action != KAUTH_NETWORK_INTERFACE) 232 return result; 233 234 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 235 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 236 result = KAUTH_RESULT_ALLOW; 237 238 return result; 239 } 240 241 /* 242 * Network interface utility routines. 243 * 244 * Routines with ifa_ifwith* names take sockaddr *'s as 245 * parameters. 246 */ 247 void 248 ifinit(void) 249 { 250 #if defined(INET) 251 sysctl_net_pktq_setup(NULL, PF_INET); 252 #endif 253 #ifdef INET6 254 if (in6_present) 255 sysctl_net_pktq_setup(NULL, PF_INET6); 256 #endif 257 258 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 259 if_listener_cb, NULL); 260 261 /* interfaces are available, inform socket code */ 262 ifioctl = doifioctl; 263 } 264 265 /* 266 * XXX Initialization before configure(). 267 * XXX hack to get pfil_add_hook working in autoconf. 268 */ 269 void 270 ifinit1(void) 271 { 272 mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE); 273 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 274 TAILQ_INIT(&ifnet_list); 275 if_indexlim = 8; 276 277 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 278 KASSERT(if_pfil != NULL); 279 280 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 281 etherinit(); 282 #endif 283 } 284 285 ifnet_t * 286 if_alloc(u_char type) 287 { 288 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 289 } 290 291 void 292 if_free(ifnet_t *ifp) 293 { 294 kmem_free(ifp, sizeof(ifnet_t)); 295 } 296 297 void 298 if_initname(struct ifnet *ifp, const char *name, int unit) 299 { 300 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 301 "%s%d", name, unit); 302 } 303 304 /* 305 * Null routines used while an interface is going away. These routines 306 * just return an error. 307 */ 308 309 int 310 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 311 const struct sockaddr *so, const struct rtentry *rt) 312 { 313 314 return ENXIO; 315 } 316 317 void 318 if_nullinput(struct ifnet *ifp, struct mbuf *m) 319 { 320 321 /* Nothing. */ 322 } 323 324 void 325 if_nullstart(struct ifnet *ifp) 326 { 327 328 /* Nothing. */ 329 } 330 331 int 332 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 333 { 334 335 return ENXIO; 336 } 337 338 int 339 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 340 { 341 342 /* Wake ifioctl_detach(), who may wait for all threads to 343 * quit the critical section. 344 */ 345 cv_signal(&ifp->if_ioctl_lock->il_emptied); 346 return ENXIO; 347 } 348 349 int 350 if_nullinit(struct ifnet *ifp) 351 { 352 353 return ENXIO; 354 } 355 356 void 357 if_nullstop(struct ifnet *ifp, int disable) 358 { 359 360 /* Nothing. */ 361 } 362 363 void 364 if_nullslowtimo(struct ifnet *ifp) 365 { 366 367 /* Nothing. */ 368 } 369 370 void 371 if_nulldrain(struct ifnet *ifp) 372 { 373 374 /* Nothing. */ 375 } 376 377 void 378 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 379 { 380 struct ifaddr *ifa; 381 struct sockaddr_dl *sdl; 382 383 ifp->if_addrlen = addrlen; 384 if_alloc_sadl(ifp); 385 ifa = ifp->if_dl; 386 sdl = satosdl(ifa->ifa_addr); 387 388 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 389 if (factory) { 390 ifp->if_hwdl = ifp->if_dl; 391 ifaref(ifp->if_hwdl); 392 } 393 /* TBD routing socket */ 394 } 395 396 struct ifaddr * 397 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 398 { 399 unsigned socksize, ifasize; 400 int addrlen, namelen; 401 struct sockaddr_dl *mask, *sdl; 402 struct ifaddr *ifa; 403 404 namelen = strlen(ifp->if_xname); 405 addrlen = ifp->if_addrlen; 406 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 407 ifasize = sizeof(*ifa) + 2 * socksize; 408 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 409 410 sdl = (struct sockaddr_dl *)(ifa + 1); 411 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 412 413 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 414 ifp->if_xname, namelen, NULL, addrlen); 415 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 416 memset(&mask->sdl_data[0], 0xff, namelen); 417 ifa->ifa_rtrequest = link_rtrequest; 418 ifa->ifa_addr = (struct sockaddr *)sdl; 419 ifa->ifa_netmask = (struct sockaddr *)mask; 420 421 *sdlp = sdl; 422 423 return ifa; 424 } 425 426 static void 427 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 428 { 429 const struct sockaddr_dl *sdl; 430 431 ifp->if_dl = ifa; 432 ifaref(ifa); 433 sdl = satosdl(ifa->ifa_addr); 434 ifp->if_sadl = sdl; 435 } 436 437 /* 438 * Allocate the link level name for the specified interface. This 439 * is an attachment helper. It must be called after ifp->if_addrlen 440 * is initialized, which may not be the case when if_attach() is 441 * called. 442 */ 443 void 444 if_alloc_sadl(struct ifnet *ifp) 445 { 446 struct ifaddr *ifa; 447 const struct sockaddr_dl *sdl; 448 449 /* 450 * If the interface already has a link name, release it 451 * now. This is useful for interfaces that can change 452 * link types, and thus switch link names often. 453 */ 454 if (ifp->if_sadl != NULL) 455 if_free_sadl(ifp); 456 457 ifa = if_dl_create(ifp, &sdl); 458 459 ifa_insert(ifp, ifa); 460 if_sadl_setrefs(ifp, ifa); 461 } 462 463 static void 464 if_deactivate_sadl(struct ifnet *ifp) 465 { 466 struct ifaddr *ifa; 467 468 KASSERT(ifp->if_dl != NULL); 469 470 ifa = ifp->if_dl; 471 472 ifp->if_sadl = NULL; 473 474 ifp->if_dl = NULL; 475 ifafree(ifa); 476 } 477 478 void 479 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 480 const struct sockaddr_dl *sdl) 481 { 482 int s; 483 484 s = splnet(); 485 486 if_deactivate_sadl(ifp); 487 488 if_sadl_setrefs(ifp, ifa); 489 IFADDR_FOREACH(ifa, ifp) 490 rtinit(ifa, RTM_LLINFO_UPD, 0); 491 splx(s); 492 } 493 494 /* 495 * Free the link level name for the specified interface. This is 496 * a detach helper. This is called from if_detach(). 497 */ 498 static void 499 if_free_sadl(struct ifnet *ifp) 500 { 501 struct ifaddr *ifa; 502 int s; 503 504 ifa = ifp->if_dl; 505 if (ifa == NULL) { 506 KASSERT(ifp->if_sadl == NULL); 507 return; 508 } 509 510 KASSERT(ifp->if_sadl != NULL); 511 512 s = splnet(); 513 rtinit(ifa, RTM_DELETE, 0); 514 ifa_remove(ifp, ifa); 515 if_deactivate_sadl(ifp); 516 if (ifp->if_hwdl == ifa) { 517 ifafree(ifa); 518 ifp->if_hwdl = NULL; 519 } 520 splx(s); 521 } 522 523 static void 524 if_getindex(ifnet_t *ifp) 525 { 526 bool hitlimit = false; 527 528 mutex_enter(&index_gen_mtx); 529 ifp->if_index_gen = index_gen++; 530 mutex_exit(&index_gen_mtx); 531 532 ifp->if_index = if_index; 533 if (ifindex2ifnet == NULL) { 534 if_index++; 535 goto skip; 536 } 537 while (if_byindex(ifp->if_index)) { 538 /* 539 * If we hit USHRT_MAX, we skip back to 0 since 540 * there are a number of places where the value 541 * of if_index or if_index itself is compared 542 * to or stored in an unsigned short. By 543 * jumping back, we won't botch those assignments 544 * or comparisons. 545 */ 546 if (++if_index == 0) { 547 if_index = 1; 548 } else if (if_index == USHRT_MAX) { 549 /* 550 * However, if we have to jump back to 551 * zero *twice* without finding an empty 552 * slot in ifindex2ifnet[], then there 553 * there are too many (>65535) interfaces. 554 */ 555 if (hitlimit) { 556 panic("too many interfaces"); 557 } 558 hitlimit = true; 559 if_index = 1; 560 } 561 ifp->if_index = if_index; 562 } 563 skip: 564 /* 565 * ifindex2ifnet is indexed by if_index. Since if_index will 566 * grow dynamically, it should grow too. 567 */ 568 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 569 size_t m, n, oldlim; 570 void *q; 571 572 oldlim = if_indexlim; 573 while (ifp->if_index >= if_indexlim) 574 if_indexlim <<= 1; 575 576 /* grow ifindex2ifnet */ 577 m = oldlim * sizeof(struct ifnet *); 578 n = if_indexlim * sizeof(struct ifnet *); 579 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 580 if (ifindex2ifnet != NULL) { 581 memcpy(q, ifindex2ifnet, m); 582 free(ifindex2ifnet, M_IFADDR); 583 } 584 ifindex2ifnet = (struct ifnet **)q; 585 } 586 ifindex2ifnet[ifp->if_index] = ifp; 587 } 588 589 /* 590 * Initialize an interface and assign an index for it. 591 * 592 * It must be called prior to a device specific attach routine 593 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 594 * and be followed by if_register: 595 * 596 * if_initialize(ifp); 597 * ether_ifattach(ifp, enaddr); 598 * if_register(ifp); 599 */ 600 void 601 if_initialize(ifnet_t *ifp) 602 { 603 KASSERT(if_indexlim > 0); 604 TAILQ_INIT(&ifp->if_addrlist); 605 606 /* 607 * Link level name is allocated later by a separate call to 608 * if_alloc_sadl(). 609 */ 610 611 if (ifp->if_snd.ifq_maxlen == 0) 612 ifp->if_snd.ifq_maxlen = ifqmaxlen; 613 614 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 615 616 ifp->if_link_state = LINK_STATE_UNKNOWN; 617 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 618 619 ifp->if_capenable = 0; 620 ifp->if_csum_flags_tx = 0; 621 ifp->if_csum_flags_rx = 0; 622 623 #ifdef ALTQ 624 ifp->if_snd.altq_type = 0; 625 ifp->if_snd.altq_disc = NULL; 626 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 627 ifp->if_snd.altq_tbr = NULL; 628 ifp->if_snd.altq_ifp = ifp; 629 #endif 630 631 #ifdef NET_MPSAFE 632 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 633 #else 634 ifp->if_snd.ifq_lock = NULL; 635 #endif 636 637 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 638 (void)pfil_run_hooks(if_pfil, 639 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 640 641 IF_AFDATA_LOCK_INIT(ifp); 642 643 ifp->if_link_si = softint_establish(SOFTINT_NET, if_link_state_change_si, ifp); 644 if (ifp->if_link_si == NULL) 645 panic("%s: softint_establish() failed", __func__); 646 647 if_getindex(ifp); 648 } 649 650 /* 651 * Register an interface to the list of "active" interfaces. 652 */ 653 void 654 if_register(ifnet_t *ifp) 655 { 656 if (ifioctl_attach(ifp) != 0) 657 panic("%s: ifioctl_attach() failed", __func__); 658 659 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 660 661 if (!STAILQ_EMPTY(&domains)) 662 if_attachdomain1(ifp); 663 664 /* Announce the interface. */ 665 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 666 667 if (ifp->if_slowtimo != NULL) { 668 ifp->if_slowtimo_ch = 669 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 670 callout_init(ifp->if_slowtimo_ch, 0); 671 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 672 if_slowtimo(ifp); 673 } 674 675 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 676 ifp->if_transmit = if_transmit; 677 678 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 679 } 680 681 /* 682 * The if_percpuq framework 683 * 684 * It allows network device drivers to execute the network stack 685 * in softint (so called softint-based if_input). It utilizes 686 * softint and percpu ifqueue. It doesn't distribute any packets 687 * between CPUs, unlike pktqueue(9). 688 * 689 * Currently we support two options for device drivers to apply the framework: 690 * - Use it implicitly with less changes 691 * - If you use if_attach in driver's _attach function and if_input in 692 * driver's Rx interrupt handler, a packet is queued and a softint handles 693 * the packet implicitly 694 * - Use it explicitly in each driver (recommended) 695 * - You can use if_percpuq_* directly in your driver 696 * - In this case, you need to allocate struct if_percpuq in driver's softc 697 * - See wm(4) as a reference implementation 698 */ 699 700 static void 701 if_percpuq_softint(void *arg) 702 { 703 struct if_percpuq *ipq = arg; 704 struct ifnet *ifp = ipq->ipq_ifp; 705 struct mbuf *m; 706 707 while ((m = if_percpuq_dequeue(ipq)) != NULL) 708 ifp->_if_input(ifp, m); 709 } 710 711 static void 712 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 713 { 714 struct ifqueue *const ifq = p; 715 716 memset(ifq, 0, sizeof(*ifq)); 717 ifq->ifq_maxlen = IFQ_MAXLEN; 718 } 719 720 struct if_percpuq * 721 if_percpuq_create(struct ifnet *ifp) 722 { 723 struct if_percpuq *ipq; 724 725 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 726 if (ipq == NULL) 727 panic("kmem_zalloc failed"); 728 729 ipq->ipq_ifp = ifp; 730 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 731 if_percpuq_softint, ipq); 732 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 733 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 734 735 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 736 737 return ipq; 738 } 739 740 static struct mbuf * 741 if_percpuq_dequeue(struct if_percpuq *ipq) 742 { 743 struct mbuf *m; 744 struct ifqueue *ifq; 745 int s; 746 747 s = splnet(); 748 ifq = percpu_getref(ipq->ipq_ifqs); 749 IF_DEQUEUE(ifq, m); 750 percpu_putref(ipq->ipq_ifqs); 751 splx(s); 752 753 return m; 754 } 755 756 static void 757 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 758 { 759 struct ifqueue *const ifq = p; 760 761 IF_PURGE(ifq); 762 } 763 764 void 765 if_percpuq_destroy(struct if_percpuq *ipq) 766 { 767 768 /* if_detach may already destroy it */ 769 if (ipq == NULL) 770 return; 771 772 softint_disestablish(ipq->ipq_si); 773 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 774 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 775 } 776 777 void 778 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 779 { 780 struct ifqueue *ifq; 781 int s; 782 783 KASSERT(ipq != NULL); 784 785 s = splnet(); 786 ifq = percpu_getref(ipq->ipq_ifqs); 787 if (IF_QFULL(ifq)) { 788 IF_DROP(ifq); 789 percpu_putref(ipq->ipq_ifqs); 790 m_freem(m); 791 goto out; 792 } 793 IF_ENQUEUE(ifq, m); 794 percpu_putref(ipq->ipq_ifqs); 795 796 softint_schedule(ipq->ipq_si); 797 out: 798 splx(s); 799 } 800 801 static void 802 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 803 { 804 struct ifqueue *const ifq = p; 805 int *sum = arg; 806 807 *sum += ifq->ifq_drops; 808 } 809 810 static int 811 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 812 { 813 struct sysctlnode node; 814 struct if_percpuq *ipq; 815 int sum = 0; 816 int error; 817 818 node = *rnode; 819 ipq = node.sysctl_data; 820 821 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 822 823 node.sysctl_data = ∑ 824 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 825 if (error != 0 || newp == NULL) 826 return error; 827 828 return 0; 829 } 830 831 static void 832 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 833 struct if_percpuq *ipq) 834 { 835 const struct sysctlnode *cnode, *rnode; 836 837 if (sysctl_createv(clog, 0, NULL, &rnode, 838 CTLFLAG_PERMANENT, 839 CTLTYPE_NODE, "interfaces", 840 SYSCTL_DESCR("Per-interface controls"), 841 NULL, 0, NULL, 0, 842 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 843 goto bad; 844 845 if (sysctl_createv(clog, 0, &rnode, &rnode, 846 CTLFLAG_PERMANENT, 847 CTLTYPE_NODE, ifname, 848 SYSCTL_DESCR("Interface controls"), 849 NULL, 0, NULL, 0, 850 CTL_CREATE, CTL_EOL) != 0) 851 goto bad; 852 853 if (sysctl_createv(clog, 0, &rnode, &rnode, 854 CTLFLAG_PERMANENT, 855 CTLTYPE_NODE, "rcvq", 856 SYSCTL_DESCR("Interface input queue controls"), 857 NULL, 0, NULL, 0, 858 CTL_CREATE, CTL_EOL) != 0) 859 goto bad; 860 861 #ifdef NOTYET 862 /* XXX Should show each per-CPU queue length? */ 863 if (sysctl_createv(clog, 0, &rnode, &rnode, 864 CTLFLAG_PERMANENT, 865 CTLTYPE_INT, "len", 866 SYSCTL_DESCR("Current input queue length"), 867 sysctl_percpuq_len, 0, NULL, 0, 868 CTL_CREATE, CTL_EOL) != 0) 869 goto bad; 870 871 if (sysctl_createv(clog, 0, &rnode, &cnode, 872 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 873 CTLTYPE_INT, "maxlen", 874 SYSCTL_DESCR("Maximum allowed input queue length"), 875 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 876 CTL_CREATE, CTL_EOL) != 0) 877 goto bad; 878 #endif 879 880 if (sysctl_createv(clog, 0, &rnode, &cnode, 881 CTLFLAG_PERMANENT, 882 CTLTYPE_INT, "drops", 883 SYSCTL_DESCR("Total packets dropped due to full input queue"), 884 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 885 CTL_CREATE, CTL_EOL) != 0) 886 goto bad; 887 888 return; 889 bad: 890 printf("%s: could not attach sysctl nodes\n", ifname); 891 return; 892 } 893 894 895 /* 896 * The common interface input routine that is called by device drivers, 897 * which should be used only when the driver's rx handler already runs 898 * in softint. 899 */ 900 void 901 if_input(struct ifnet *ifp, struct mbuf *m) 902 { 903 904 KASSERT(ifp->if_percpuq == NULL); 905 KASSERT(!cpu_intr_p()); 906 907 ifp->_if_input(ifp, m); 908 } 909 910 /* 911 * DEPRECATED. Use if_initialize and if_register instead. 912 * See the above comment of if_initialize. 913 * 914 * Note that it implicitly enables if_percpuq to make drivers easy to 915 * migrate softinet-based if_input without much changes. If you don't 916 * want to enable it, use if_initialize instead. 917 */ 918 void 919 if_attach(ifnet_t *ifp) 920 { 921 922 if_initialize(ifp); 923 ifp->if_percpuq = if_percpuq_create(ifp); 924 if_register(ifp); 925 } 926 927 void 928 if_attachdomain(void) 929 { 930 struct ifnet *ifp; 931 int s; 932 933 s = splnet(); 934 IFNET_FOREACH(ifp) 935 if_attachdomain1(ifp); 936 splx(s); 937 } 938 939 static void 940 if_attachdomain1(struct ifnet *ifp) 941 { 942 struct domain *dp; 943 int s; 944 945 s = splnet(); 946 947 /* address family dependent data region */ 948 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 949 DOMAIN_FOREACH(dp) { 950 if (dp->dom_ifattach != NULL) 951 ifp->if_afdata[dp->dom_family] = 952 (*dp->dom_ifattach)(ifp); 953 } 954 955 splx(s); 956 } 957 958 /* 959 * Deactivate an interface. This points all of the procedure 960 * handles at error stubs. May be called from interrupt context. 961 */ 962 void 963 if_deactivate(struct ifnet *ifp) 964 { 965 int s; 966 967 s = splnet(); 968 969 ifp->if_output = if_nulloutput; 970 ifp->_if_input = if_nullinput; 971 ifp->if_start = if_nullstart; 972 ifp->if_transmit = if_nulltransmit; 973 ifp->if_ioctl = if_nullioctl; 974 ifp->if_init = if_nullinit; 975 ifp->if_stop = if_nullstop; 976 ifp->if_slowtimo = if_nullslowtimo; 977 ifp->if_drain = if_nulldrain; 978 979 /* No more packets may be enqueued. */ 980 ifp->if_snd.ifq_maxlen = 0; 981 982 splx(s); 983 } 984 985 void 986 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 987 { 988 struct ifaddr *ifa, *nifa; 989 990 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 991 if (ifa->ifa_addr->sa_family != family) 992 continue; 993 (*purgeaddr)(ifa); 994 } 995 } 996 997 /* 998 * Detach an interface from the list of "active" interfaces, 999 * freeing any resources as we go along. 1000 * 1001 * NOTE: This routine must be called with a valid thread context, 1002 * as it may block. 1003 */ 1004 void 1005 if_detach(struct ifnet *ifp) 1006 { 1007 struct socket so; 1008 struct ifaddr *ifa; 1009 #ifdef IFAREF_DEBUG 1010 struct ifaddr *last_ifa = NULL; 1011 #endif 1012 struct domain *dp; 1013 const struct protosw *pr; 1014 int s, i, family, purged; 1015 uint64_t xc; 1016 1017 /* 1018 * XXX It's kind of lame that we have to have the 1019 * XXX socket structure... 1020 */ 1021 memset(&so, 0, sizeof(so)); 1022 1023 s = splnet(); 1024 1025 ifindex2ifnet[ifp->if_index] = NULL; 1026 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1027 1028 if (ifp->if_slowtimo != NULL) { 1029 ifp->if_slowtimo = NULL; 1030 callout_halt(ifp->if_slowtimo_ch, NULL); 1031 callout_destroy(ifp->if_slowtimo_ch); 1032 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1033 } 1034 1035 /* 1036 * Do an if_down() to give protocols a chance to do something. 1037 */ 1038 if_down(ifp); 1039 1040 #ifdef ALTQ 1041 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1042 altq_disable(&ifp->if_snd); 1043 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1044 altq_detach(&ifp->if_snd); 1045 #endif 1046 1047 if (ifp->if_snd.ifq_lock) 1048 mutex_obj_free(ifp->if_snd.ifq_lock); 1049 1050 sysctl_teardown(&ifp->if_sysctl_log); 1051 1052 #if NCARP > 0 1053 /* Remove the interface from any carp group it is a part of. */ 1054 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1055 carp_ifdetach(ifp); 1056 #endif 1057 1058 /* 1059 * Rip all the addresses off the interface. This should make 1060 * all of the routes go away. 1061 * 1062 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1063 * from the list, including our "cursor", ifa. For safety, 1064 * and to honor the TAILQ abstraction, I just restart the 1065 * loop after each removal. Note that the loop will exit 1066 * when all of the remaining ifaddrs belong to the AF_LINK 1067 * family. I am counting on the historical fact that at 1068 * least one pr_usrreq in each address domain removes at 1069 * least one ifaddr. 1070 */ 1071 again: 1072 IFADDR_FOREACH(ifa, ifp) { 1073 family = ifa->ifa_addr->sa_family; 1074 #ifdef IFAREF_DEBUG 1075 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1076 ifa, family, ifa->ifa_refcnt); 1077 if (last_ifa != NULL && ifa == last_ifa) 1078 panic("if_detach: loop detected"); 1079 last_ifa = ifa; 1080 #endif 1081 if (family == AF_LINK) 1082 continue; 1083 dp = pffinddomain(family); 1084 #ifdef DIAGNOSTIC 1085 if (dp == NULL) 1086 panic("if_detach: no domain for AF %d", 1087 family); 1088 #endif 1089 /* 1090 * XXX These PURGEIF calls are redundant with the 1091 * purge-all-families calls below, but are left in for 1092 * now both to make a smaller change, and to avoid 1093 * unplanned interactions with clearing of 1094 * ifp->if_addrlist. 1095 */ 1096 purged = 0; 1097 for (pr = dp->dom_protosw; 1098 pr < dp->dom_protoswNPROTOSW; pr++) { 1099 so.so_proto = pr; 1100 if (pr->pr_usrreqs) { 1101 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1102 purged = 1; 1103 } 1104 } 1105 if (purged == 0) { 1106 /* 1107 * XXX What's really the best thing to do 1108 * XXX here? --thorpej (at) NetBSD.org 1109 */ 1110 printf("if_detach: WARNING: AF %d not purged\n", 1111 family); 1112 ifa_remove(ifp, ifa); 1113 } 1114 goto again; 1115 } 1116 1117 if_free_sadl(ifp); 1118 1119 /* Walk the routing table looking for stragglers. */ 1120 for (i = 0; i <= AF_MAX; i++) { 1121 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 1122 continue; 1123 } 1124 1125 DOMAIN_FOREACH(dp) { 1126 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1127 { 1128 void *p = ifp->if_afdata[dp->dom_family]; 1129 if (p) { 1130 ifp->if_afdata[dp->dom_family] = NULL; 1131 (*dp->dom_ifdetach)(ifp, p); 1132 } 1133 } 1134 1135 /* 1136 * One would expect multicast memberships (INET and 1137 * INET6) on UDP sockets to be purged by the PURGEIF 1138 * calls above, but if all addresses were removed from 1139 * the interface prior to destruction, the calls will 1140 * not be made (e.g. ppp, for which pppd(8) generally 1141 * removes addresses before destroying the interface). 1142 * Because there is no invariant that multicast 1143 * memberships only exist for interfaces with IPv4 1144 * addresses, we must call PURGEIF regardless of 1145 * addresses. (Protocols which might store ifnet 1146 * pointers are marked with PR_PURGEIF.) 1147 */ 1148 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1149 so.so_proto = pr; 1150 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1151 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1152 } 1153 } 1154 1155 (void)pfil_run_hooks(if_pfil, 1156 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 1157 (void)pfil_head_destroy(ifp->if_pfil); 1158 1159 /* Announce that the interface is gone. */ 1160 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1161 1162 ifioctl_detach(ifp); 1163 1164 IF_AFDATA_LOCK_DESTROY(ifp); 1165 1166 softint_disestablish(ifp->if_link_si); 1167 ifp->if_link_si = NULL; 1168 1169 /* 1170 * remove packets that came from ifp, from software interrupt queues. 1171 */ 1172 DOMAIN_FOREACH(dp) { 1173 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1174 struct ifqueue *iq = dp->dom_ifqueues[i]; 1175 if (iq == NULL) 1176 break; 1177 dp->dom_ifqueues[i] = NULL; 1178 if_detach_queues(ifp, iq); 1179 } 1180 } 1181 1182 /* 1183 * IP queues have to be processed separately: net-queue barrier 1184 * ensures that the packets are dequeued while a cross-call will 1185 * ensure that the interrupts have completed. FIXME: not quite.. 1186 */ 1187 #ifdef INET 1188 pktq_barrier(ip_pktq); 1189 #endif 1190 #ifdef INET6 1191 if (in6_present) 1192 pktq_barrier(ip6_pktq); 1193 #endif 1194 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1195 xc_wait(xc); 1196 1197 if (ifp->if_percpuq != NULL) { 1198 if_percpuq_destroy(ifp->if_percpuq); 1199 ifp->if_percpuq = NULL; 1200 } 1201 1202 splx(s); 1203 } 1204 1205 static void 1206 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1207 { 1208 struct mbuf *m, *prev, *next; 1209 1210 prev = NULL; 1211 for (m = q->ifq_head; m != NULL; m = next) { 1212 KASSERT((m->m_flags & M_PKTHDR) != 0); 1213 1214 next = m->m_nextpkt; 1215 if (m->m_pkthdr.rcvif != ifp) { 1216 prev = m; 1217 continue; 1218 } 1219 1220 if (prev != NULL) 1221 prev->m_nextpkt = m->m_nextpkt; 1222 else 1223 q->ifq_head = m->m_nextpkt; 1224 if (q->ifq_tail == m) 1225 q->ifq_tail = prev; 1226 q->ifq_len--; 1227 1228 m->m_nextpkt = NULL; 1229 m_freem(m); 1230 IF_DROP(q); 1231 } 1232 } 1233 1234 /* 1235 * Callback for a radix tree walk to delete all references to an 1236 * ifnet. 1237 */ 1238 static int 1239 if_rt_walktree(struct rtentry *rt, void *v) 1240 { 1241 struct ifnet *ifp = (struct ifnet *)v; 1242 int error; 1243 struct rtentry *retrt; 1244 1245 if (rt->rt_ifp != ifp) 1246 return 0; 1247 1248 /* Delete the entry. */ 1249 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 1250 rt_mask(rt), rt->rt_flags, &retrt); 1251 if (error == 0) { 1252 KASSERT(retrt == rt); 1253 KASSERT((retrt->rt_flags & RTF_UP) == 0); 1254 retrt->rt_ifp = NULL; 1255 rtfree(retrt); 1256 } else { 1257 printf("%s: warning: unable to delete rtentry @ %p, " 1258 "error = %d\n", ifp->if_xname, rt, error); 1259 } 1260 return ERESTART; 1261 } 1262 1263 /* 1264 * Create a clone network interface. 1265 */ 1266 static int 1267 if_clone_create(const char *name) 1268 { 1269 struct if_clone *ifc; 1270 int unit; 1271 1272 ifc = if_clone_lookup(name, &unit); 1273 if (ifc == NULL) 1274 return EINVAL; 1275 1276 if (ifunit(name) != NULL) 1277 return EEXIST; 1278 1279 return (*ifc->ifc_create)(ifc, unit); 1280 } 1281 1282 /* 1283 * Destroy a clone network interface. 1284 */ 1285 static int 1286 if_clone_destroy(const char *name) 1287 { 1288 struct if_clone *ifc; 1289 struct ifnet *ifp; 1290 1291 ifc = if_clone_lookup(name, NULL); 1292 if (ifc == NULL) 1293 return EINVAL; 1294 1295 ifp = ifunit(name); 1296 if (ifp == NULL) 1297 return ENXIO; 1298 1299 if (ifc->ifc_destroy == NULL) 1300 return EOPNOTSUPP; 1301 1302 return (*ifc->ifc_destroy)(ifp); 1303 } 1304 1305 /* 1306 * Look up a network interface cloner. 1307 */ 1308 static struct if_clone * 1309 if_clone_lookup(const char *name, int *unitp) 1310 { 1311 struct if_clone *ifc; 1312 const char *cp; 1313 char *dp, ifname[IFNAMSIZ + 3]; 1314 int unit; 1315 1316 strcpy(ifname, "if_"); 1317 /* separate interface name from unit */ 1318 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1319 *cp && (*cp < '0' || *cp > '9');) 1320 *dp++ = *cp++; 1321 1322 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1323 return NULL; /* No name or unit number */ 1324 *dp++ = '\0'; 1325 1326 again: 1327 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1328 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1329 break; 1330 } 1331 1332 if (ifc == NULL) { 1333 if (*ifname == '\0' || 1334 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1335 return NULL; 1336 *ifname = '\0'; 1337 goto again; 1338 } 1339 1340 unit = 0; 1341 while (cp - name < IFNAMSIZ && *cp) { 1342 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1343 /* Bogus unit number. */ 1344 return NULL; 1345 } 1346 unit = (unit * 10) + (*cp++ - '0'); 1347 } 1348 1349 if (unitp != NULL) 1350 *unitp = unit; 1351 return ifc; 1352 } 1353 1354 /* 1355 * Register a network interface cloner. 1356 */ 1357 void 1358 if_clone_attach(struct if_clone *ifc) 1359 { 1360 1361 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1362 if_cloners_count++; 1363 } 1364 1365 /* 1366 * Unregister a network interface cloner. 1367 */ 1368 void 1369 if_clone_detach(struct if_clone *ifc) 1370 { 1371 1372 LIST_REMOVE(ifc, ifc_list); 1373 if_cloners_count--; 1374 } 1375 1376 /* 1377 * Provide list of interface cloners to userspace. 1378 */ 1379 int 1380 if_clone_list(int buf_count, char *buffer, int *total) 1381 { 1382 char outbuf[IFNAMSIZ], *dst; 1383 struct if_clone *ifc; 1384 int count, error = 0; 1385 1386 *total = if_cloners_count; 1387 if ((dst = buffer) == NULL) { 1388 /* Just asking how many there are. */ 1389 return 0; 1390 } 1391 1392 if (buf_count < 0) 1393 return EINVAL; 1394 1395 count = (if_cloners_count < buf_count) ? 1396 if_cloners_count : buf_count; 1397 1398 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1399 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1400 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1401 if (outbuf[sizeof(outbuf) - 1] != '\0') 1402 return ENAMETOOLONG; 1403 error = copyout(outbuf, dst, sizeof(outbuf)); 1404 if (error != 0) 1405 break; 1406 } 1407 1408 return error; 1409 } 1410 1411 void 1412 ifaref(struct ifaddr *ifa) 1413 { 1414 ifa->ifa_refcnt++; 1415 } 1416 1417 void 1418 ifafree(struct ifaddr *ifa) 1419 { 1420 KASSERT(ifa != NULL); 1421 KASSERT(ifa->ifa_refcnt > 0); 1422 1423 if (--ifa->ifa_refcnt == 0) { 1424 free(ifa, M_IFADDR); 1425 } 1426 } 1427 1428 void 1429 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1430 { 1431 ifa->ifa_ifp = ifp; 1432 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1433 ifaref(ifa); 1434 } 1435 1436 void 1437 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1438 { 1439 KASSERT(ifa->ifa_ifp == ifp); 1440 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1441 ifafree(ifa); 1442 } 1443 1444 static inline int 1445 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1446 { 1447 return sockaddr_cmp(sa1, sa2) == 0; 1448 } 1449 1450 /* 1451 * Locate an interface based on a complete address. 1452 */ 1453 /*ARGSUSED*/ 1454 struct ifaddr * 1455 ifa_ifwithaddr(const struct sockaddr *addr) 1456 { 1457 struct ifnet *ifp; 1458 struct ifaddr *ifa; 1459 1460 IFNET_FOREACH(ifp) { 1461 if (ifp->if_output == if_nulloutput) 1462 continue; 1463 IFADDR_FOREACH(ifa, ifp) { 1464 if (ifa->ifa_addr->sa_family != addr->sa_family) 1465 continue; 1466 if (equal(addr, ifa->ifa_addr)) 1467 return ifa; 1468 if ((ifp->if_flags & IFF_BROADCAST) && 1469 ifa->ifa_broadaddr && 1470 /* IP6 doesn't have broadcast */ 1471 ifa->ifa_broadaddr->sa_len != 0 && 1472 equal(ifa->ifa_broadaddr, addr)) 1473 return ifa; 1474 } 1475 } 1476 return NULL; 1477 } 1478 1479 /* 1480 * Locate the point to point interface with a given destination address. 1481 */ 1482 /*ARGSUSED*/ 1483 struct ifaddr * 1484 ifa_ifwithdstaddr(const struct sockaddr *addr) 1485 { 1486 struct ifnet *ifp; 1487 struct ifaddr *ifa; 1488 1489 IFNET_FOREACH(ifp) { 1490 if (ifp->if_output == if_nulloutput) 1491 continue; 1492 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1493 continue; 1494 IFADDR_FOREACH(ifa, ifp) { 1495 if (ifa->ifa_addr->sa_family != addr->sa_family || 1496 ifa->ifa_dstaddr == NULL) 1497 continue; 1498 if (equal(addr, ifa->ifa_dstaddr)) 1499 return ifa; 1500 } 1501 } 1502 return NULL; 1503 } 1504 1505 /* 1506 * Find an interface on a specific network. If many, choice 1507 * is most specific found. 1508 */ 1509 struct ifaddr * 1510 ifa_ifwithnet(const struct sockaddr *addr) 1511 { 1512 struct ifnet *ifp; 1513 struct ifaddr *ifa; 1514 const struct sockaddr_dl *sdl; 1515 struct ifaddr *ifa_maybe = 0; 1516 u_int af = addr->sa_family; 1517 const char *addr_data = addr->sa_data, *cplim; 1518 1519 if (af == AF_LINK) { 1520 sdl = satocsdl(addr); 1521 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1522 ifindex2ifnet[sdl->sdl_index] && 1523 ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) { 1524 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1525 } 1526 } 1527 #ifdef NETATALK 1528 if (af == AF_APPLETALK) { 1529 const struct sockaddr_at *sat, *sat2; 1530 sat = (const struct sockaddr_at *)addr; 1531 IFNET_FOREACH(ifp) { 1532 if (ifp->if_output == if_nulloutput) 1533 continue; 1534 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1535 if (ifa == NULL) 1536 continue; 1537 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1538 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1539 return ifa; /* exact match */ 1540 if (ifa_maybe == NULL) { 1541 /* else keep the if with the right range */ 1542 ifa_maybe = ifa; 1543 } 1544 } 1545 return ifa_maybe; 1546 } 1547 #endif 1548 IFNET_FOREACH(ifp) { 1549 if (ifp->if_output == if_nulloutput) 1550 continue; 1551 IFADDR_FOREACH(ifa, ifp) { 1552 const char *cp, *cp2, *cp3; 1553 1554 if (ifa->ifa_addr->sa_family != af || 1555 ifa->ifa_netmask == NULL) 1556 next: continue; 1557 cp = addr_data; 1558 cp2 = ifa->ifa_addr->sa_data; 1559 cp3 = ifa->ifa_netmask->sa_data; 1560 cplim = (const char *)ifa->ifa_netmask + 1561 ifa->ifa_netmask->sa_len; 1562 while (cp3 < cplim) { 1563 if ((*cp++ ^ *cp2++) & *cp3++) { 1564 /* want to continue for() loop */ 1565 goto next; 1566 } 1567 } 1568 if (ifa_maybe == NULL || 1569 rt_refines(ifa->ifa_netmask, 1570 ifa_maybe->ifa_netmask)) 1571 ifa_maybe = ifa; 1572 } 1573 } 1574 return ifa_maybe; 1575 } 1576 1577 /* 1578 * Find the interface of the addresss. 1579 */ 1580 struct ifaddr * 1581 ifa_ifwithladdr(const struct sockaddr *addr) 1582 { 1583 struct ifaddr *ia; 1584 1585 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1586 (ia = ifa_ifwithnet(addr))) 1587 return ia; 1588 return NULL; 1589 } 1590 1591 /* 1592 * Find an interface using a specific address family 1593 */ 1594 struct ifaddr * 1595 ifa_ifwithaf(int af) 1596 { 1597 struct ifnet *ifp; 1598 struct ifaddr *ifa; 1599 1600 IFNET_FOREACH(ifp) { 1601 if (ifp->if_output == if_nulloutput) 1602 continue; 1603 IFADDR_FOREACH(ifa, ifp) { 1604 if (ifa->ifa_addr->sa_family == af) 1605 return ifa; 1606 } 1607 } 1608 return NULL; 1609 } 1610 1611 /* 1612 * Find an interface address specific to an interface best matching 1613 * a given address. 1614 */ 1615 struct ifaddr * 1616 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1617 { 1618 struct ifaddr *ifa; 1619 const char *cp, *cp2, *cp3; 1620 const char *cplim; 1621 struct ifaddr *ifa_maybe = 0; 1622 u_int af = addr->sa_family; 1623 1624 if (ifp->if_output == if_nulloutput) 1625 return NULL; 1626 1627 if (af >= AF_MAX) 1628 return NULL; 1629 1630 IFADDR_FOREACH(ifa, ifp) { 1631 if (ifa->ifa_addr->sa_family != af) 1632 continue; 1633 ifa_maybe = ifa; 1634 if (ifa->ifa_netmask == NULL) { 1635 if (equal(addr, ifa->ifa_addr) || 1636 (ifa->ifa_dstaddr && 1637 equal(addr, ifa->ifa_dstaddr))) 1638 return ifa; 1639 continue; 1640 } 1641 cp = addr->sa_data; 1642 cp2 = ifa->ifa_addr->sa_data; 1643 cp3 = ifa->ifa_netmask->sa_data; 1644 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1645 for (; cp3 < cplim; cp3++) { 1646 if ((*cp++ ^ *cp2++) & *cp3) 1647 break; 1648 } 1649 if (cp3 == cplim) 1650 return ifa; 1651 } 1652 return ifa_maybe; 1653 } 1654 1655 /* 1656 * Default action when installing a route with a Link Level gateway. 1657 * Lookup an appropriate real ifa to point to. 1658 * This should be moved to /sys/net/link.c eventually. 1659 */ 1660 void 1661 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1662 { 1663 struct ifaddr *ifa; 1664 const struct sockaddr *dst; 1665 struct ifnet *ifp; 1666 1667 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1668 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1669 return; 1670 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1671 rt_replace_ifa(rt, ifa); 1672 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1673 ifa->ifa_rtrequest(cmd, rt, info); 1674 } 1675 } 1676 1677 /* 1678 * bitmask macros to manage a densely packed link_state change queue. 1679 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 1680 * LINK_STATE_UP(2) we need 2 bits for each state change. 1681 * As a state change to store is 0, treat all bits set as an unset item. 1682 */ 1683 #define LQ_ITEM_BITS 2 1684 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 1685 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 1686 #define LINK_STATE_UNSET LQ_ITEM_MASK 1687 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 1688 #define LQ_STORE(q, i, v) \ 1689 do { \ 1690 (q) &= ~LQ_MASK((i)); \ 1691 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 1692 } while (0 /* CONSTCOND */) 1693 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 1694 #define LQ_POP(q, v) \ 1695 do { \ 1696 (v) = LQ_ITEM((q), 0); \ 1697 (q) >>= LQ_ITEM_BITS; \ 1698 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1699 } while (0 /* CONSTCOND */) 1700 #define LQ_PUSH(q, v) \ 1701 do { \ 1702 (q) >>= LQ_ITEM_BITS; \ 1703 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1704 } while (0 /* CONSTCOND */) 1705 #define LQ_FIND_UNSET(q, i) \ 1706 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 1707 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 1708 break; \ 1709 } 1710 /* 1711 * Handle a change in the interface link state and 1712 * queue notifications. 1713 */ 1714 void 1715 if_link_state_change(struct ifnet *ifp, int link_state) 1716 { 1717 int s, idx; 1718 1719 /* Ensure change is to a valid state */ 1720 switch (link_state) { 1721 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 1722 case LINK_STATE_DOWN: /* FALLTHROUGH */ 1723 case LINK_STATE_UP: 1724 break; 1725 default: 1726 #ifdef DEBUG 1727 printf("%s: invalid link state %d\n", 1728 ifp->if_xname, link_state); 1729 #endif 1730 return; 1731 } 1732 1733 s = splnet(); 1734 1735 /* Find the last unset event in the queue. */ 1736 LQ_FIND_UNSET(ifp->if_link_queue, idx); 1737 1738 /* 1739 * Ensure link_state doesn't match the last event in the queue. 1740 * ifp->if_link_state is not checked and set here because 1741 * that would present an inconsistent picture to the system. 1742 */ 1743 if (idx != 0 && 1744 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 1745 goto out; 1746 1747 /* Handle queue overflow. */ 1748 if (idx == LQ_MAX(ifp->if_link_queue)) { 1749 uint8_t lost; 1750 1751 /* 1752 * The DOWN state must be protected from being pushed off 1753 * the queue to ensure that userland will always be 1754 * in a sane state. 1755 * Because DOWN is protected, there is no need to protect 1756 * UNKNOWN. 1757 * It should be invalid to change from any other state to 1758 * UNKNOWN anyway ... 1759 */ 1760 lost = LQ_ITEM(ifp->if_link_queue, 0); 1761 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 1762 if (lost == LINK_STATE_DOWN) { 1763 lost = LQ_ITEM(ifp->if_link_queue, 0); 1764 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 1765 } 1766 printf("%s: lost link state change %s\n", 1767 ifp->if_xname, 1768 lost == LINK_STATE_UP ? "UP" : 1769 lost == LINK_STATE_DOWN ? "DOWN" : 1770 "UNKNOWN"); 1771 } else 1772 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 1773 1774 softint_schedule(ifp->if_link_si); 1775 1776 out: 1777 splx(s); 1778 } 1779 1780 /* 1781 * Handle interface link state change notifications. 1782 * Must be called at splnet(). 1783 */ 1784 static void 1785 if_link_state_change0(struct ifnet *ifp, int link_state) 1786 { 1787 struct domain *dp; 1788 1789 /* Ensure the change is still valid. */ 1790 if (ifp->if_link_state == link_state) 1791 return; 1792 1793 #ifdef DEBUG 1794 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1795 link_state == LINK_STATE_UP ? "UP" : 1796 link_state == LINK_STATE_DOWN ? "DOWN" : 1797 "UNKNOWN", 1798 ifp->if_link_state == LINK_STATE_UP ? "UP" : 1799 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 1800 "UNKNOWN"); 1801 #endif 1802 1803 /* 1804 * When going from UNKNOWN to UP, we need to mark existing 1805 * addresses as tentative and restart DAD as we may have 1806 * erroneously not found a duplicate. 1807 * 1808 * This needs to happen before rt_ifmsg to avoid a race where 1809 * listeners would have an address and expect it to work right 1810 * away. 1811 */ 1812 if (link_state == LINK_STATE_UP && 1813 ifp->if_link_state == LINK_STATE_UNKNOWN) 1814 { 1815 DOMAIN_FOREACH(dp) { 1816 if (dp->dom_if_link_state_change != NULL) 1817 dp->dom_if_link_state_change(ifp, 1818 LINK_STATE_DOWN); 1819 } 1820 } 1821 1822 ifp->if_link_state = link_state; 1823 1824 /* Notify that the link state has changed. */ 1825 rt_ifmsg(ifp); 1826 1827 #if NCARP > 0 1828 if (ifp->if_carp) 1829 carp_carpdev_state(ifp); 1830 #endif 1831 1832 DOMAIN_FOREACH(dp) { 1833 if (dp->dom_if_link_state_change != NULL) 1834 dp->dom_if_link_state_change(ifp, link_state); 1835 } 1836 } 1837 1838 /* 1839 * Process the interface link state change queue. 1840 */ 1841 static void 1842 if_link_state_change_si(void *arg) 1843 { 1844 struct ifnet *ifp = arg; 1845 int s; 1846 uint8_t state; 1847 1848 s = splnet(); 1849 1850 /* Pop a link state change from the queue and process it. */ 1851 LQ_POP(ifp->if_link_queue, state); 1852 if_link_state_change0(ifp, state); 1853 1854 /* If there is a link state change to come, schedule it. */ 1855 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 1856 softint_schedule(ifp->if_link_si); 1857 1858 splx(s); 1859 } 1860 1861 /* 1862 * Default action when installing a local route on a point-to-point 1863 * interface. 1864 */ 1865 void 1866 p2p_rtrequest(int req, struct rtentry *rt, 1867 __unused const struct rt_addrinfo *info) 1868 { 1869 struct ifnet *ifp = rt->rt_ifp; 1870 struct ifaddr *ifa, *lo0ifa; 1871 1872 switch (req) { 1873 case RTM_ADD: 1874 if ((rt->rt_flags & RTF_LOCAL) == 0) 1875 break; 1876 1877 IFADDR_FOREACH(ifa, ifp) { 1878 if (equal(rt_getkey(rt), ifa->ifa_addr)) 1879 break; 1880 } 1881 if (ifa == NULL) 1882 break; 1883 1884 /* 1885 * Ensure lo0 has an address of the same family. 1886 */ 1887 IFADDR_FOREACH(lo0ifa, lo0ifp) { 1888 if (lo0ifa->ifa_addr->sa_family == 1889 ifa->ifa_addr->sa_family) 1890 break; 1891 } 1892 if (lo0ifa == NULL) 1893 break; 1894 1895 rt->rt_ifp = lo0ifp; 1896 1897 /* 1898 * Make sure to set rt->rt_ifa to the interface 1899 * address we are using, otherwise we will have trouble 1900 * with source address selection. 1901 */ 1902 if (ifa != rt->rt_ifa) 1903 rt_replace_ifa(rt, ifa); 1904 break; 1905 case RTM_DELETE: 1906 default: 1907 break; 1908 } 1909 } 1910 1911 /* 1912 * Mark an interface down and notify protocols of 1913 * the transition. 1914 * NOTE: must be called at splsoftnet or equivalent. 1915 */ 1916 void 1917 if_down(struct ifnet *ifp) 1918 { 1919 struct ifaddr *ifa; 1920 struct domain *dp; 1921 1922 ifp->if_flags &= ~IFF_UP; 1923 nanotime(&ifp->if_lastchange); 1924 IFADDR_FOREACH(ifa, ifp) 1925 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1926 IFQ_PURGE(&ifp->if_snd); 1927 #if NCARP > 0 1928 if (ifp->if_carp) 1929 carp_carpdev_state(ifp); 1930 #endif 1931 rt_ifmsg(ifp); 1932 DOMAIN_FOREACH(dp) { 1933 if (dp->dom_if_down) 1934 dp->dom_if_down(ifp); 1935 } 1936 } 1937 1938 /* 1939 * Mark an interface up and notify protocols of 1940 * the transition. 1941 * NOTE: must be called at splsoftnet or equivalent. 1942 */ 1943 void 1944 if_up(struct ifnet *ifp) 1945 { 1946 #ifdef notyet 1947 struct ifaddr *ifa; 1948 #endif 1949 struct domain *dp; 1950 1951 ifp->if_flags |= IFF_UP; 1952 nanotime(&ifp->if_lastchange); 1953 #ifdef notyet 1954 /* this has no effect on IP, and will kill all ISO connections XXX */ 1955 IFADDR_FOREACH(ifa, ifp) 1956 pfctlinput(PRC_IFUP, ifa->ifa_addr); 1957 #endif 1958 #if NCARP > 0 1959 if (ifp->if_carp) 1960 carp_carpdev_state(ifp); 1961 #endif 1962 rt_ifmsg(ifp); 1963 DOMAIN_FOREACH(dp) { 1964 if (dp->dom_if_up) 1965 dp->dom_if_up(ifp); 1966 } 1967 } 1968 1969 /* 1970 * Handle interface slowtimo timer routine. Called 1971 * from softclock, we decrement timer (if set) and 1972 * call the appropriate interface routine on expiration. 1973 */ 1974 static void 1975 if_slowtimo(void *arg) 1976 { 1977 void (*slowtimo)(struct ifnet *); 1978 struct ifnet *ifp = arg; 1979 int s; 1980 1981 slowtimo = ifp->if_slowtimo; 1982 if (__predict_false(slowtimo == NULL)) 1983 return; 1984 1985 s = splnet(); 1986 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 1987 (*slowtimo)(ifp); 1988 1989 splx(s); 1990 1991 if (__predict_true(ifp->if_slowtimo != NULL)) 1992 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 1993 } 1994 1995 /* 1996 * Set/clear promiscuous mode on interface ifp based on the truth value 1997 * of pswitch. The calls are reference counted so that only the first 1998 * "on" request actually has an effect, as does the final "off" request. 1999 * Results are undefined if the "off" and "on" requests are not matched. 2000 */ 2001 int 2002 ifpromisc(struct ifnet *ifp, int pswitch) 2003 { 2004 int pcount, ret; 2005 short nflags; 2006 2007 pcount = ifp->if_pcount; 2008 if (pswitch) { 2009 /* 2010 * Allow the device to be "placed" into promiscuous 2011 * mode even if it is not configured up. It will 2012 * consult IFF_PROMISC when it is brought up. 2013 */ 2014 if (ifp->if_pcount++ != 0) 2015 return 0; 2016 nflags = ifp->if_flags | IFF_PROMISC; 2017 } else { 2018 if (--ifp->if_pcount > 0) 2019 return 0; 2020 nflags = ifp->if_flags & ~IFF_PROMISC; 2021 } 2022 ret = if_flags_set(ifp, nflags); 2023 /* Restore interface state if not successful. */ 2024 if (ret != 0) { 2025 ifp->if_pcount = pcount; 2026 } 2027 return ret; 2028 } 2029 2030 /* 2031 * Map interface name to 2032 * interface structure pointer. 2033 */ 2034 struct ifnet * 2035 ifunit(const char *name) 2036 { 2037 struct ifnet *ifp; 2038 const char *cp = name; 2039 u_int unit = 0; 2040 u_int i; 2041 2042 /* 2043 * If the entire name is a number, treat it as an ifindex. 2044 */ 2045 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2046 unit = unit * 10 + (*cp - '0'); 2047 } 2048 2049 /* 2050 * If the number took all of the name, then it's a valid ifindex. 2051 */ 2052 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2053 if (unit >= if_indexlim) 2054 return NULL; 2055 ifp = ifindex2ifnet[unit]; 2056 if (ifp == NULL || ifp->if_output == if_nulloutput) 2057 return NULL; 2058 return ifp; 2059 } 2060 2061 IFNET_FOREACH(ifp) { 2062 if (ifp->if_output == if_nulloutput) 2063 continue; 2064 if (strcmp(ifp->if_xname, name) == 0) 2065 return ifp; 2066 } 2067 return NULL; 2068 } 2069 2070 ifnet_t * 2071 if_byindex(u_int idx) 2072 { 2073 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2074 } 2075 2076 /* common */ 2077 int 2078 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2079 { 2080 int s; 2081 struct ifreq *ifr; 2082 struct ifcapreq *ifcr; 2083 struct ifdatareq *ifdr; 2084 2085 switch (cmd) { 2086 case SIOCSIFCAP: 2087 ifcr = data; 2088 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2089 return EINVAL; 2090 2091 if (ifcr->ifcr_capenable == ifp->if_capenable) 2092 return 0; 2093 2094 ifp->if_capenable = ifcr->ifcr_capenable; 2095 2096 /* Pre-compute the checksum flags mask. */ 2097 ifp->if_csum_flags_tx = 0; 2098 ifp->if_csum_flags_rx = 0; 2099 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2100 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2101 } 2102 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2103 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2104 } 2105 2106 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2107 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2108 } 2109 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2110 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2111 } 2112 2113 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2114 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2115 } 2116 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2117 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2118 } 2119 2120 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2121 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2122 } 2123 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2124 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2125 } 2126 2127 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2128 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2129 } 2130 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2131 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2132 } 2133 if (ifp->if_flags & IFF_UP) 2134 return ENETRESET; 2135 return 0; 2136 case SIOCSIFFLAGS: 2137 ifr = data; 2138 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2139 s = splnet(); 2140 if_down(ifp); 2141 splx(s); 2142 } 2143 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2144 s = splnet(); 2145 if_up(ifp); 2146 splx(s); 2147 } 2148 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2149 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2150 break; 2151 case SIOCGIFFLAGS: 2152 ifr = data; 2153 ifr->ifr_flags = ifp->if_flags; 2154 break; 2155 2156 case SIOCGIFMETRIC: 2157 ifr = data; 2158 ifr->ifr_metric = ifp->if_metric; 2159 break; 2160 2161 case SIOCGIFMTU: 2162 ifr = data; 2163 ifr->ifr_mtu = ifp->if_mtu; 2164 break; 2165 2166 case SIOCGIFDLT: 2167 ifr = data; 2168 ifr->ifr_dlt = ifp->if_dlt; 2169 break; 2170 2171 case SIOCGIFCAP: 2172 ifcr = data; 2173 ifcr->ifcr_capabilities = ifp->if_capabilities; 2174 ifcr->ifcr_capenable = ifp->if_capenable; 2175 break; 2176 2177 case SIOCSIFMETRIC: 2178 ifr = data; 2179 ifp->if_metric = ifr->ifr_metric; 2180 break; 2181 2182 case SIOCGIFDATA: 2183 ifdr = data; 2184 ifdr->ifdr_data = ifp->if_data; 2185 break; 2186 2187 case SIOCGIFINDEX: 2188 ifr = data; 2189 ifr->ifr_index = ifp->if_index; 2190 break; 2191 2192 case SIOCZIFDATA: 2193 ifdr = data; 2194 ifdr->ifdr_data = ifp->if_data; 2195 /* 2196 * Assumes that the volatile counters that can be 2197 * zero'ed are at the end of if_data. 2198 */ 2199 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2200 offsetof(struct if_data, ifi_ipackets)); 2201 /* 2202 * The memset() clears to the bottm of if_data. In the area, 2203 * if_lastchange is included. Please be careful if new entry 2204 * will be added into if_data or rewite this. 2205 * 2206 * And also, update if_lastchnage. 2207 */ 2208 getnanotime(&ifp->if_lastchange); 2209 break; 2210 case SIOCSIFMTU: 2211 ifr = data; 2212 if (ifp->if_mtu == ifr->ifr_mtu) 2213 break; 2214 ifp->if_mtu = ifr->ifr_mtu; 2215 /* 2216 * If the link MTU changed, do network layer specific procedure. 2217 */ 2218 #ifdef INET6 2219 if (in6_present) 2220 nd6_setmtu(ifp); 2221 #endif 2222 return ENETRESET; 2223 default: 2224 return ENOTTY; 2225 } 2226 return 0; 2227 } 2228 2229 int 2230 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2231 { 2232 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2233 struct ifaddr *ifa; 2234 const struct sockaddr *any, *sa; 2235 union { 2236 struct sockaddr sa; 2237 struct sockaddr_storage ss; 2238 } u, v; 2239 2240 switch (cmd) { 2241 case SIOCSIFADDRPREF: 2242 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2243 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2244 NULL) != 0) 2245 return EPERM; 2246 case SIOCGIFADDRPREF: 2247 break; 2248 default: 2249 return EOPNOTSUPP; 2250 } 2251 2252 /* sanity checks */ 2253 if (data == NULL || ifp == NULL) { 2254 panic("invalid argument to %s", __func__); 2255 /*NOTREACHED*/ 2256 } 2257 2258 /* address must be specified on ADD and DELETE */ 2259 sa = sstocsa(&ifap->ifap_addr); 2260 if (sa->sa_family != sofamily(so)) 2261 return EINVAL; 2262 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2263 return EINVAL; 2264 2265 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2266 2267 IFADDR_FOREACH(ifa, ifp) { 2268 if (ifa->ifa_addr->sa_family != sa->sa_family) 2269 continue; 2270 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2271 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2272 break; 2273 } 2274 if (ifa == NULL) 2275 return EADDRNOTAVAIL; 2276 2277 switch (cmd) { 2278 case SIOCSIFADDRPREF: 2279 ifa->ifa_preference = ifap->ifap_preference; 2280 return 0; 2281 case SIOCGIFADDRPREF: 2282 /* fill in the if_laddrreq structure */ 2283 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2284 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2285 ifap->ifap_preference = ifa->ifa_preference; 2286 return 0; 2287 default: 2288 return EOPNOTSUPP; 2289 } 2290 } 2291 2292 static void 2293 ifnet_lock_enter(struct ifnet_lock *il) 2294 { 2295 uint64_t *nenter; 2296 2297 /* Before trying to acquire the mutex, increase the count of threads 2298 * who have entered or who wait to enter the critical section. 2299 * Avoid one costly locked memory transaction by keeping a count for 2300 * each CPU. 2301 */ 2302 nenter = percpu_getref(il->il_nenter); 2303 (*nenter)++; 2304 percpu_putref(il->il_nenter); 2305 mutex_enter(&il->il_lock); 2306 } 2307 2308 static void 2309 ifnet_lock_exit(struct ifnet_lock *il) 2310 { 2311 /* Increase the count of threads who have exited the critical 2312 * section. Increase while we still hold the lock. 2313 */ 2314 il->il_nexit++; 2315 mutex_exit(&il->il_lock); 2316 } 2317 2318 /* 2319 * Interface ioctls. 2320 */ 2321 static int 2322 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2323 { 2324 struct ifnet *ifp; 2325 struct ifreq *ifr; 2326 int error = 0; 2327 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2328 u_long ocmd = cmd; 2329 #endif 2330 short oif_flags; 2331 #ifdef COMPAT_OIFREQ 2332 struct ifreq ifrb; 2333 struct oifreq *oifr = NULL; 2334 #endif 2335 int r; 2336 2337 switch (cmd) { 2338 #ifdef COMPAT_OIFREQ 2339 case OSIOCGIFCONF: 2340 case OOSIOCGIFCONF: 2341 return compat_ifconf(cmd, data); 2342 #endif 2343 #ifdef COMPAT_OIFDATA 2344 case OSIOCGIFDATA: 2345 case OSIOCZIFDATA: 2346 return compat_ifdatareq(l, cmd, data); 2347 #endif 2348 case SIOCGIFCONF: 2349 return ifconf(cmd, data); 2350 case SIOCINITIFADDR: 2351 return EPERM; 2352 } 2353 2354 #ifdef COMPAT_OIFREQ 2355 cmd = compat_cvtcmd(cmd); 2356 if (cmd != ocmd) { 2357 oifr = data; 2358 data = ifr = &ifrb; 2359 ifreqo2n(oifr, ifr); 2360 } else 2361 #endif 2362 ifr = data; 2363 2364 ifp = ifunit(ifr->ifr_name); 2365 2366 switch (cmd) { 2367 case SIOCIFCREATE: 2368 case SIOCIFDESTROY: 2369 if (l != NULL) { 2370 error = kauth_authorize_network(l->l_cred, 2371 KAUTH_NETWORK_INTERFACE, 2372 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2373 (void *)cmd, NULL); 2374 if (error != 0) 2375 return error; 2376 } 2377 mutex_enter(&if_clone_mtx); 2378 r = (cmd == SIOCIFCREATE) ? 2379 if_clone_create(ifr->ifr_name) : 2380 if_clone_destroy(ifr->ifr_name); 2381 mutex_exit(&if_clone_mtx); 2382 return r; 2383 2384 case SIOCIFGCLONERS: 2385 { 2386 struct if_clonereq *req = (struct if_clonereq *)data; 2387 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2388 &req->ifcr_total); 2389 } 2390 } 2391 2392 if (ifp == NULL) 2393 return ENXIO; 2394 2395 switch (cmd) { 2396 case SIOCALIFADDR: 2397 case SIOCDLIFADDR: 2398 case SIOCSIFADDRPREF: 2399 case SIOCSIFFLAGS: 2400 case SIOCSIFCAP: 2401 case SIOCSIFMETRIC: 2402 case SIOCZIFDATA: 2403 case SIOCSIFMTU: 2404 case SIOCSIFPHYADDR: 2405 case SIOCDIFPHYADDR: 2406 #ifdef INET6 2407 case SIOCSIFPHYADDR_IN6: 2408 #endif 2409 case SIOCSLIFPHYADDR: 2410 case SIOCADDMULTI: 2411 case SIOCDELMULTI: 2412 case SIOCSIFMEDIA: 2413 case SIOCSDRVSPEC: 2414 case SIOCG80211: 2415 case SIOCS80211: 2416 case SIOCS80211NWID: 2417 case SIOCS80211NWKEY: 2418 case SIOCS80211POWER: 2419 case SIOCS80211BSSID: 2420 case SIOCS80211CHANNEL: 2421 case SIOCSLINKSTR: 2422 if (l != NULL) { 2423 error = kauth_authorize_network(l->l_cred, 2424 KAUTH_NETWORK_INTERFACE, 2425 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2426 (void *)cmd, NULL); 2427 if (error != 0) 2428 return error; 2429 } 2430 } 2431 2432 oif_flags = ifp->if_flags; 2433 2434 ifnet_lock_enter(ifp->if_ioctl_lock); 2435 error = (*ifp->if_ioctl)(ifp, cmd, data); 2436 if (error != ENOTTY) 2437 ; 2438 else if (so->so_proto == NULL) 2439 error = EOPNOTSUPP; 2440 else { 2441 #ifdef COMPAT_OSOCK 2442 error = compat_ifioctl(so, ocmd, cmd, data, l); 2443 #else 2444 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2445 cmd, data, ifp); 2446 #endif 2447 } 2448 2449 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2450 if ((ifp->if_flags & IFF_UP) != 0) { 2451 int s = splnet(); 2452 if_up(ifp); 2453 splx(s); 2454 } 2455 } 2456 #ifdef COMPAT_OIFREQ 2457 if (cmd != ocmd) 2458 ifreqn2o(oifr, ifr); 2459 #endif 2460 2461 ifnet_lock_exit(ifp->if_ioctl_lock); 2462 return error; 2463 } 2464 2465 /* This callback adds to the sum in `arg' the number of 2466 * threads on `ci' who have entered or who wait to enter the 2467 * critical section. 2468 */ 2469 static void 2470 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci) 2471 { 2472 uint64_t *sum = arg, *nenter = p; 2473 2474 *sum += *nenter; 2475 } 2476 2477 /* Return the number of threads who have entered or who wait 2478 * to enter the critical section on all CPUs. 2479 */ 2480 static uint64_t 2481 ifnet_lock_entrances(struct ifnet_lock *il) 2482 { 2483 uint64_t sum = 0; 2484 2485 percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum); 2486 2487 return sum; 2488 } 2489 2490 static int 2491 ifioctl_attach(struct ifnet *ifp) 2492 { 2493 struct ifnet_lock *il; 2494 2495 /* If the driver has not supplied its own if_ioctl, then 2496 * supply the default. 2497 */ 2498 if (ifp->if_ioctl == NULL) 2499 ifp->if_ioctl = ifioctl_common; 2500 2501 /* Create an ifnet_lock for synchronizing ifioctls. */ 2502 if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL) 2503 return ENOMEM; 2504 2505 il->il_nenter = percpu_alloc(sizeof(uint64_t)); 2506 if (il->il_nenter == NULL) { 2507 kmem_free(il, sizeof(*il)); 2508 return ENOMEM; 2509 } 2510 2511 mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE); 2512 cv_init(&il->il_emptied, ifp->if_xname); 2513 2514 ifp->if_ioctl_lock = il; 2515 2516 return 0; 2517 } 2518 2519 /* 2520 * This must not be called until after `ifp' has been withdrawn from the 2521 * ifnet tables so that ifioctl() cannot get a handle on it by calling 2522 * ifunit(). 2523 */ 2524 static void 2525 ifioctl_detach(struct ifnet *ifp) 2526 { 2527 struct ifnet_lock *il; 2528 2529 il = ifp->if_ioctl_lock; 2530 mutex_enter(&il->il_lock); 2531 /* Install if_nullioctl to make sure that any thread that 2532 * subsequently enters the critical section will quit it 2533 * immediately and signal the condition variable that we 2534 * wait on, below. 2535 */ 2536 ifp->if_ioctl = if_nullioctl; 2537 /* Sleep while threads are still in the critical section or 2538 * wait to enter it. 2539 */ 2540 while (ifnet_lock_entrances(il) != il->il_nexit) 2541 cv_wait(&il->il_emptied, &il->il_lock); 2542 /* At this point, we are the only thread still in the critical 2543 * section, and no new thread can get a handle on the ifioctl 2544 * lock, so it is safe to free its memory. 2545 */ 2546 mutex_exit(&il->il_lock); 2547 ifp->if_ioctl_lock = NULL; 2548 percpu_free(il->il_nenter, sizeof(uint64_t)); 2549 il->il_nenter = NULL; 2550 cv_destroy(&il->il_emptied); 2551 mutex_destroy(&il->il_lock); 2552 kmem_free(il, sizeof(*il)); 2553 } 2554 2555 /* 2556 * Return interface configuration 2557 * of system. List may be used 2558 * in later ioctl's (above) to get 2559 * other information. 2560 * 2561 * Each record is a struct ifreq. Before the addition of 2562 * sockaddr_storage, the API rule was that sockaddr flavors that did 2563 * not fit would extend beyond the struct ifreq, with the next struct 2564 * ifreq starting sa_len beyond the struct sockaddr. Because the 2565 * union in struct ifreq includes struct sockaddr_storage, every kind 2566 * of sockaddr must fit. Thus, there are no longer any overlength 2567 * records. 2568 * 2569 * Records are added to the user buffer if they fit, and ifc_len is 2570 * adjusted to the length that was written. Thus, the user is only 2571 * assured of getting the complete list if ifc_len on return is at 2572 * least sizeof(struct ifreq) less than it was on entry. 2573 * 2574 * If the user buffer pointer is NULL, this routine copies no data and 2575 * returns the amount of space that would be needed. 2576 * 2577 * Invariants: 2578 * ifrp points to the next part of the user's buffer to be used. If 2579 * ifrp != NULL, space holds the number of bytes remaining that we may 2580 * write at ifrp. Otherwise, space holds the number of bytes that 2581 * would have been written had there been adequate space. 2582 */ 2583 /*ARGSUSED*/ 2584 static int 2585 ifconf(u_long cmd, void *data) 2586 { 2587 struct ifconf *ifc = (struct ifconf *)data; 2588 struct ifnet *ifp; 2589 struct ifaddr *ifa; 2590 struct ifreq ifr, *ifrp = NULL; 2591 int space = 0, error = 0; 2592 const int sz = (int)sizeof(struct ifreq); 2593 const bool docopy = ifc->ifc_req != NULL; 2594 2595 if (docopy) { 2596 space = ifc->ifc_len; 2597 ifrp = ifc->ifc_req; 2598 } 2599 2600 IFNET_FOREACH(ifp) { 2601 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2602 sizeof(ifr.ifr_name)); 2603 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') 2604 return ENAMETOOLONG; 2605 if (IFADDR_EMPTY(ifp)) { 2606 /* Interface with no addresses - send zero sockaddr. */ 2607 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2608 if (!docopy) { 2609 space += sz; 2610 continue; 2611 } 2612 if (space >= sz) { 2613 error = copyout(&ifr, ifrp, sz); 2614 if (error != 0) 2615 return error; 2616 ifrp++; 2617 space -= sz; 2618 } 2619 } 2620 2621 IFADDR_FOREACH(ifa, ifp) { 2622 struct sockaddr *sa = ifa->ifa_addr; 2623 /* all sockaddrs must fit in sockaddr_storage */ 2624 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2625 2626 if (!docopy) { 2627 space += sz; 2628 continue; 2629 } 2630 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2631 if (space >= sz) { 2632 error = copyout(&ifr, ifrp, sz); 2633 if (error != 0) 2634 return (error); 2635 ifrp++; space -= sz; 2636 } 2637 } 2638 } 2639 if (docopy) { 2640 KASSERT(0 <= space && space <= ifc->ifc_len); 2641 ifc->ifc_len -= space; 2642 } else { 2643 KASSERT(space >= 0); 2644 ifc->ifc_len = space; 2645 } 2646 return (0); 2647 } 2648 2649 int 2650 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2651 { 2652 uint8_t len; 2653 #ifdef COMPAT_OIFREQ 2654 struct ifreq ifrb; 2655 struct oifreq *oifr = NULL; 2656 u_long ocmd = cmd; 2657 cmd = compat_cvtcmd(cmd); 2658 if (cmd != ocmd) { 2659 oifr = (struct oifreq *)(void *)ifr; 2660 ifr = &ifrb; 2661 ifreqo2n(oifr, ifr); 2662 len = sizeof(oifr->ifr_addr); 2663 } else 2664 #endif 2665 len = sizeof(ifr->ifr_ifru.ifru_space); 2666 2667 if (len < sa->sa_len) 2668 return EFBIG; 2669 2670 memset(&ifr->ifr_addr, 0, len); 2671 sockaddr_copy(&ifr->ifr_addr, len, sa); 2672 2673 #ifdef COMPAT_OIFREQ 2674 if (cmd != ocmd) 2675 ifreqn2o(oifr, ifr); 2676 #endif 2677 return 0; 2678 } 2679 2680 /* 2681 * wrapper function for the drivers which doesn't have if_transmit(). 2682 */ 2683 int 2684 if_transmit(struct ifnet *ifp, struct mbuf *m) 2685 { 2686 int s, error; 2687 2688 s = splnet(); 2689 2690 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2691 if (error != 0) { 2692 /* mbuf is already freed */ 2693 goto out; 2694 } 2695 2696 ifp->if_obytes += m->m_pkthdr.len;; 2697 if (m->m_flags & M_MCAST) 2698 ifp->if_omcasts++; 2699 2700 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2701 (*ifp->if_start)(ifp); 2702 out: 2703 splx(s); 2704 2705 return error; 2706 } 2707 2708 /* 2709 * Queue message on interface, and start output if interface 2710 * not yet active. 2711 */ 2712 int 2713 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 2714 { 2715 2716 return (*ifp->if_transmit)(ifp, m); 2717 } 2718 2719 /* 2720 * Queue message on interface, possibly using a second fast queue 2721 */ 2722 int 2723 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 2724 { 2725 int error = 0; 2726 2727 if (ifq != NULL 2728 #ifdef ALTQ 2729 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2730 #endif 2731 ) { 2732 if (IF_QFULL(ifq)) { 2733 IF_DROP(&ifp->if_snd); 2734 m_freem(m); 2735 if (error == 0) 2736 error = ENOBUFS; 2737 } else 2738 IF_ENQUEUE(ifq, m); 2739 } else 2740 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2741 if (error != 0) { 2742 ++ifp->if_oerrors; 2743 return error; 2744 } 2745 return 0; 2746 } 2747 2748 int 2749 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2750 { 2751 int rc; 2752 2753 if (ifp->if_initaddr != NULL) 2754 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2755 else if (src || 2756 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2757 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2758 2759 return rc; 2760 } 2761 2762 int 2763 if_do_dad(struct ifnet *ifp) 2764 { 2765 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2766 return 0; 2767 2768 switch (ifp->if_type) { 2769 case IFT_FAITH: 2770 /* 2771 * These interfaces do not have the IFF_LOOPBACK flag, 2772 * but loop packets back. We do not have to do DAD on such 2773 * interfaces. We should even omit it, because loop-backed 2774 * responses would confuse the DAD procedure. 2775 */ 2776 return 0; 2777 default: 2778 /* 2779 * Our DAD routine requires the interface up and running. 2780 * However, some interfaces can be up before the RUNNING 2781 * status. Additionaly, users may try to assign addresses 2782 * before the interface becomes up (or running). 2783 * We simply skip DAD in such a case as a work around. 2784 * XXX: we should rather mark "tentative" on such addresses, 2785 * and do DAD after the interface becomes ready. 2786 */ 2787 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 2788 (IFF_UP|IFF_RUNNING)) 2789 return 0; 2790 2791 return 1; 2792 } 2793 } 2794 2795 int 2796 if_flags_set(ifnet_t *ifp, const short flags) 2797 { 2798 int rc; 2799 2800 if (ifp->if_setflags != NULL) 2801 rc = (*ifp->if_setflags)(ifp, flags); 2802 else { 2803 short cantflags, chgdflags; 2804 struct ifreq ifr; 2805 2806 chgdflags = ifp->if_flags ^ flags; 2807 cantflags = chgdflags & IFF_CANTCHANGE; 2808 2809 if (cantflags != 0) 2810 ifp->if_flags ^= cantflags; 2811 2812 /* Traditionally, we do not call if_ioctl after 2813 * setting/clearing only IFF_PROMISC if the interface 2814 * isn't IFF_UP. Uphold that tradition. 2815 */ 2816 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2817 return 0; 2818 2819 memset(&ifr, 0, sizeof(ifr)); 2820 2821 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2822 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2823 2824 if (rc != 0 && cantflags != 0) 2825 ifp->if_flags ^= cantflags; 2826 } 2827 2828 return rc; 2829 } 2830 2831 int 2832 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2833 { 2834 int rc; 2835 struct ifreq ifr; 2836 2837 if (ifp->if_mcastop != NULL) 2838 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2839 else { 2840 ifreq_setaddr(cmd, &ifr, sa); 2841 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2842 } 2843 2844 return rc; 2845 } 2846 2847 static void 2848 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2849 struct ifaltq *ifq) 2850 { 2851 const struct sysctlnode *cnode, *rnode; 2852 2853 if (sysctl_createv(clog, 0, NULL, &rnode, 2854 CTLFLAG_PERMANENT, 2855 CTLTYPE_NODE, "interfaces", 2856 SYSCTL_DESCR("Per-interface controls"), 2857 NULL, 0, NULL, 0, 2858 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 2859 goto bad; 2860 2861 if (sysctl_createv(clog, 0, &rnode, &rnode, 2862 CTLFLAG_PERMANENT, 2863 CTLTYPE_NODE, ifname, 2864 SYSCTL_DESCR("Interface controls"), 2865 NULL, 0, NULL, 0, 2866 CTL_CREATE, CTL_EOL) != 0) 2867 goto bad; 2868 2869 if (sysctl_createv(clog, 0, &rnode, &rnode, 2870 CTLFLAG_PERMANENT, 2871 CTLTYPE_NODE, "sndq", 2872 SYSCTL_DESCR("Interface output queue controls"), 2873 NULL, 0, NULL, 0, 2874 CTL_CREATE, CTL_EOL) != 0) 2875 goto bad; 2876 2877 if (sysctl_createv(clog, 0, &rnode, &cnode, 2878 CTLFLAG_PERMANENT, 2879 CTLTYPE_INT, "len", 2880 SYSCTL_DESCR("Current output queue length"), 2881 NULL, 0, &ifq->ifq_len, 0, 2882 CTL_CREATE, CTL_EOL) != 0) 2883 goto bad; 2884 2885 if (sysctl_createv(clog, 0, &rnode, &cnode, 2886 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2887 CTLTYPE_INT, "maxlen", 2888 SYSCTL_DESCR("Maximum allowed output queue length"), 2889 NULL, 0, &ifq->ifq_maxlen, 0, 2890 CTL_CREATE, CTL_EOL) != 0) 2891 goto bad; 2892 2893 if (sysctl_createv(clog, 0, &rnode, &cnode, 2894 CTLFLAG_PERMANENT, 2895 CTLTYPE_INT, "drops", 2896 SYSCTL_DESCR("Packets dropped due to full output queue"), 2897 NULL, 0, &ifq->ifq_drops, 0, 2898 CTL_CREATE, CTL_EOL) != 0) 2899 goto bad; 2900 2901 return; 2902 bad: 2903 printf("%s: could not attach sysctl nodes\n", ifname); 2904 return; 2905 } 2906 2907 #if defined(INET) || defined(INET6) 2908 2909 #define SYSCTL_NET_PKTQ(q, cn, c) \ 2910 static int \ 2911 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 2912 { \ 2913 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 2914 } 2915 2916 #if defined(INET) 2917 static int 2918 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 2919 { 2920 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 2921 } 2922 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 2923 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 2924 #endif 2925 2926 #if defined(INET6) 2927 static int 2928 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 2929 { 2930 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 2931 } 2932 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 2933 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 2934 #endif 2935 2936 static void 2937 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 2938 { 2939 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 2940 const char *pfname = NULL, *ipname = NULL; 2941 int ipn = 0, qid = 0; 2942 2943 switch (pf) { 2944 #if defined(INET) 2945 case PF_INET: 2946 len_func = sysctl_net_ip_pktq_items; 2947 maxlen_func = sysctl_net_ip_pktq_maxlen; 2948 drops_func = sysctl_net_ip_pktq_drops; 2949 pfname = "inet", ipn = IPPROTO_IP; 2950 ipname = "ip", qid = IPCTL_IFQ; 2951 break; 2952 #endif 2953 #if defined(INET6) 2954 case PF_INET6: 2955 len_func = sysctl_net_ip6_pktq_items; 2956 maxlen_func = sysctl_net_ip6_pktq_maxlen; 2957 drops_func = sysctl_net_ip6_pktq_drops; 2958 pfname = "inet6", ipn = IPPROTO_IPV6; 2959 ipname = "ip6", qid = IPV6CTL_IFQ; 2960 break; 2961 #endif 2962 default: 2963 KASSERT(false); 2964 } 2965 2966 sysctl_createv(clog, 0, NULL, NULL, 2967 CTLFLAG_PERMANENT, 2968 CTLTYPE_NODE, pfname, NULL, 2969 NULL, 0, NULL, 0, 2970 CTL_NET, pf, CTL_EOL); 2971 sysctl_createv(clog, 0, NULL, NULL, 2972 CTLFLAG_PERMANENT, 2973 CTLTYPE_NODE, ipname, NULL, 2974 NULL, 0, NULL, 0, 2975 CTL_NET, pf, ipn, CTL_EOL); 2976 sysctl_createv(clog, 0, NULL, NULL, 2977 CTLFLAG_PERMANENT, 2978 CTLTYPE_NODE, "ifq", 2979 SYSCTL_DESCR("Protocol input queue controls"), 2980 NULL, 0, NULL, 0, 2981 CTL_NET, pf, ipn, qid, CTL_EOL); 2982 2983 sysctl_createv(clog, 0, NULL, NULL, 2984 CTLFLAG_PERMANENT, 2985 CTLTYPE_INT, "len", 2986 SYSCTL_DESCR("Current input queue length"), 2987 len_func, 0, NULL, 0, 2988 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 2989 sysctl_createv(clog, 0, NULL, NULL, 2990 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2991 CTLTYPE_INT, "maxlen", 2992 SYSCTL_DESCR("Maximum allowed input queue length"), 2993 maxlen_func, 0, NULL, 0, 2994 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 2995 sysctl_createv(clog, 0, NULL, NULL, 2996 CTLFLAG_PERMANENT, 2997 CTLTYPE_INT, "drops", 2998 SYSCTL_DESCR("Packets dropped due to full input queue"), 2999 drops_func, 0, NULL, 0, 3000 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3001 } 3002 #endif /* INET || INET6 */ 3003 3004 static int 3005 if_sdl_sysctl(SYSCTLFN_ARGS) 3006 { 3007 struct ifnet *ifp; 3008 const struct sockaddr_dl *sdl; 3009 3010 if (namelen != 1) 3011 return EINVAL; 3012 3013 ifp = if_byindex(name[0]); 3014 if (ifp == NULL) 3015 return ENODEV; 3016 3017 sdl = ifp->if_sadl; 3018 if (sdl == NULL) { 3019 *oldlenp = 0; 3020 return 0; 3021 } 3022 3023 if (oldp == NULL) { 3024 *oldlenp = sdl->sdl_alen; 3025 return 0; 3026 } 3027 3028 if (*oldlenp >= sdl->sdl_alen) 3029 *oldlenp = sdl->sdl_alen; 3030 return sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3031 } 3032 3033 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 3034 { 3035 const struct sysctlnode *rnode = NULL; 3036 3037 sysctl_createv(clog, 0, NULL, &rnode, 3038 CTLFLAG_PERMANENT, 3039 CTLTYPE_NODE, "sdl", 3040 SYSCTL_DESCR("Get active link-layer address"), 3041 if_sdl_sysctl, 0, NULL, 0, 3042 CTL_NET, CTL_CREATE, CTL_EOL); 3043 } 3044
Indexes created Mon Sep 22 13:09:51 GMT 2025