if.c revision 1.345
1 /* $NetBSD: if.c,v 1.345 2016/06/22 10:44:32 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 1.345 2016/06/22 10:44:32 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 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 165 struct ifnet_head ifnet_list; 166 167 struct pslist_head ifnet_pslist; 168 static ifnet_t ** ifindex2ifnet = NULL; 169 static u_int if_index = 1; 170 static size_t if_indexlim = 0; 171 static uint64_t index_gen; 172 /* Mutex to protect the above objects. */ 173 kmutex_t ifnet_mtx __cacheline_aligned; 174 struct psref_class *ifnet_psref_class __read_mostly; 175 static pserialize_t ifnet_psz; 176 177 static kmutex_t if_clone_mtx; 178 179 struct ifnet *lo0ifp; 180 int ifqmaxlen = IFQ_MAXLEN; 181 182 static int if_rt_walktree(struct rtentry *, void *); 183 184 static struct if_clone *if_clone_lookup(const char *, int *); 185 186 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 187 static int if_cloners_count; 188 189 /* Packet filtering hook for interfaces. */ 190 pfil_head_t * if_pfil; 191 192 static kauth_listener_t if_listener; 193 194 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 195 static void if_detach_queues(struct ifnet *, struct ifqueue *); 196 static void sysctl_sndq_setup(struct sysctllog **, const char *, 197 struct ifaltq *); 198 static void if_slowtimo(void *); 199 static void if_free_sadl(struct ifnet *); 200 static void if_attachdomain1(struct ifnet *); 201 static int ifconf(u_long, void *); 202 static int if_transmit(struct ifnet *, struct mbuf *); 203 static int if_clone_create(const char *); 204 static int if_clone_destroy(const char *); 205 static void if_link_state_change_si(void *); 206 207 struct if_percpuq { 208 struct ifnet *ipq_ifp; 209 void *ipq_si; 210 struct percpu *ipq_ifqs; /* struct ifqueue */ 211 }; 212 213 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 214 215 static void if_percpuq_drops(void *, void *, struct cpu_info *); 216 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 217 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 218 struct if_percpuq *); 219 220 #if defined(INET) || defined(INET6) 221 static void sysctl_net_pktq_setup(struct sysctllog **, int); 222 #endif 223 224 static int 225 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 226 void *arg0, void *arg1, void *arg2, void *arg3) 227 { 228 int result; 229 enum kauth_network_req req; 230 231 result = KAUTH_RESULT_DEFER; 232 req = (enum kauth_network_req)arg1; 233 234 if (action != KAUTH_NETWORK_INTERFACE) 235 return result; 236 237 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 238 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 239 result = KAUTH_RESULT_ALLOW; 240 241 return result; 242 } 243 244 /* 245 * Network interface utility routines. 246 * 247 * Routines with ifa_ifwith* names take sockaddr *'s as 248 * parameters. 249 */ 250 void 251 ifinit(void) 252 { 253 #if defined(INET) 254 sysctl_net_pktq_setup(NULL, PF_INET); 255 #endif 256 #ifdef INET6 257 if (in6_present) 258 sysctl_net_pktq_setup(NULL, PF_INET6); 259 #endif 260 261 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 262 if_listener_cb, NULL); 263 264 /* interfaces are available, inform socket code */ 265 ifioctl = doifioctl; 266 } 267 268 /* 269 * XXX Initialization before configure(). 270 * XXX hack to get pfil_add_hook working in autoconf. 271 */ 272 void 273 ifinit1(void) 274 { 275 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 276 TAILQ_INIT(&ifnet_list); 277 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 278 ifnet_psz = pserialize_create(); 279 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 280 PSLIST_INIT(&ifnet_pslist); 281 if_indexlim = 8; 282 283 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 284 KASSERT(if_pfil != NULL); 285 286 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 287 etherinit(); 288 #endif 289 } 290 291 ifnet_t * 292 if_alloc(u_char type) 293 { 294 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 295 } 296 297 void 298 if_free(ifnet_t *ifp) 299 { 300 kmem_free(ifp, sizeof(ifnet_t)); 301 } 302 303 void 304 if_initname(struct ifnet *ifp, const char *name, int unit) 305 { 306 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 307 "%s%d", name, unit); 308 } 309 310 /* 311 * Null routines used while an interface is going away. These routines 312 * just return an error. 313 */ 314 315 int 316 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 317 const struct sockaddr *so, const struct rtentry *rt) 318 { 319 320 return ENXIO; 321 } 322 323 void 324 if_nullinput(struct ifnet *ifp, struct mbuf *m) 325 { 326 327 /* Nothing. */ 328 } 329 330 void 331 if_nullstart(struct ifnet *ifp) 332 { 333 334 /* Nothing. */ 335 } 336 337 int 338 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 339 { 340 341 return ENXIO; 342 } 343 344 int 345 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 346 { 347 348 return ENXIO; 349 } 350 351 int 352 if_nullinit(struct ifnet *ifp) 353 { 354 355 return ENXIO; 356 } 357 358 void 359 if_nullstop(struct ifnet *ifp, int disable) 360 { 361 362 /* Nothing. */ 363 } 364 365 void 366 if_nullslowtimo(struct ifnet *ifp) 367 { 368 369 /* Nothing. */ 370 } 371 372 void 373 if_nulldrain(struct ifnet *ifp) 374 { 375 376 /* Nothing. */ 377 } 378 379 void 380 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 381 { 382 struct ifaddr *ifa; 383 struct sockaddr_dl *sdl; 384 385 ifp->if_addrlen = addrlen; 386 if_alloc_sadl(ifp); 387 ifa = ifp->if_dl; 388 sdl = satosdl(ifa->ifa_addr); 389 390 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 391 if (factory) { 392 ifp->if_hwdl = ifp->if_dl; 393 ifaref(ifp->if_hwdl); 394 } 395 /* TBD routing socket */ 396 } 397 398 struct ifaddr * 399 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 400 { 401 unsigned socksize, ifasize; 402 int addrlen, namelen; 403 struct sockaddr_dl *mask, *sdl; 404 struct ifaddr *ifa; 405 406 namelen = strlen(ifp->if_xname); 407 addrlen = ifp->if_addrlen; 408 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 409 ifasize = sizeof(*ifa) + 2 * socksize; 410 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 411 412 sdl = (struct sockaddr_dl *)(ifa + 1); 413 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 414 415 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 416 ifp->if_xname, namelen, NULL, addrlen); 417 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 418 memset(&mask->sdl_data[0], 0xff, namelen); 419 ifa->ifa_rtrequest = link_rtrequest; 420 ifa->ifa_addr = (struct sockaddr *)sdl; 421 ifa->ifa_netmask = (struct sockaddr *)mask; 422 423 *sdlp = sdl; 424 425 return ifa; 426 } 427 428 static void 429 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 430 { 431 const struct sockaddr_dl *sdl; 432 433 ifp->if_dl = ifa; 434 ifaref(ifa); 435 sdl = satosdl(ifa->ifa_addr); 436 ifp->if_sadl = sdl; 437 } 438 439 /* 440 * Allocate the link level name for the specified interface. This 441 * is an attachment helper. It must be called after ifp->if_addrlen 442 * is initialized, which may not be the case when if_attach() is 443 * called. 444 */ 445 void 446 if_alloc_sadl(struct ifnet *ifp) 447 { 448 struct ifaddr *ifa; 449 const struct sockaddr_dl *sdl; 450 451 /* 452 * If the interface already has a link name, release it 453 * now. This is useful for interfaces that can change 454 * link types, and thus switch link names often. 455 */ 456 if (ifp->if_sadl != NULL) 457 if_free_sadl(ifp); 458 459 ifa = if_dl_create(ifp, &sdl); 460 461 ifa_insert(ifp, ifa); 462 if_sadl_setrefs(ifp, ifa); 463 } 464 465 static void 466 if_deactivate_sadl(struct ifnet *ifp) 467 { 468 struct ifaddr *ifa; 469 470 KASSERT(ifp->if_dl != NULL); 471 472 ifa = ifp->if_dl; 473 474 ifp->if_sadl = NULL; 475 476 ifp->if_dl = NULL; 477 ifafree(ifa); 478 } 479 480 void 481 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 482 const struct sockaddr_dl *sdl) 483 { 484 int s; 485 486 s = splnet(); 487 488 if_deactivate_sadl(ifp); 489 490 if_sadl_setrefs(ifp, ifa); 491 IFADDR_FOREACH(ifa, ifp) 492 rtinit(ifa, RTM_LLINFO_UPD, 0); 493 splx(s); 494 } 495 496 /* 497 * Free the link level name for the specified interface. This is 498 * a detach helper. This is called from if_detach(). 499 */ 500 static void 501 if_free_sadl(struct ifnet *ifp) 502 { 503 struct ifaddr *ifa; 504 int s; 505 506 ifa = ifp->if_dl; 507 if (ifa == NULL) { 508 KASSERT(ifp->if_sadl == NULL); 509 return; 510 } 511 512 KASSERT(ifp->if_sadl != NULL); 513 514 s = splnet(); 515 rtinit(ifa, RTM_DELETE, 0); 516 ifa_remove(ifp, ifa); 517 if_deactivate_sadl(ifp); 518 if (ifp->if_hwdl == ifa) { 519 ifafree(ifa); 520 ifp->if_hwdl = NULL; 521 } 522 splx(s); 523 } 524 525 static void 526 if_getindex(ifnet_t *ifp) 527 { 528 bool hitlimit = false; 529 530 ifp->if_index_gen = index_gen++; 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 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 648 psref_target_init(&ifp->if_psref, ifnet_psref_class); 649 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 650 651 IFNET_LOCK(); 652 if_getindex(ifp); 653 IFNET_UNLOCK(); 654 } 655 656 /* 657 * Register an interface to the list of "active" interfaces. 658 */ 659 void 660 if_register(ifnet_t *ifp) 661 { 662 /* 663 * If the driver has not supplied its own if_ioctl, then 664 * supply the default. 665 */ 666 if (ifp->if_ioctl == NULL) 667 ifp->if_ioctl = ifioctl_common; 668 669 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 670 671 if (!STAILQ_EMPTY(&domains)) 672 if_attachdomain1(ifp); 673 674 /* Announce the interface. */ 675 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 676 677 if (ifp->if_slowtimo != NULL) { 678 ifp->if_slowtimo_ch = 679 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 680 callout_init(ifp->if_slowtimo_ch, 0); 681 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 682 if_slowtimo(ifp); 683 } 684 685 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 686 ifp->if_transmit = if_transmit; 687 688 IFNET_LOCK(); 689 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 690 IFNET_WRITER_INSERT_TAIL(ifp); 691 IFNET_UNLOCK(); 692 } 693 694 /* 695 * The if_percpuq framework 696 * 697 * It allows network device drivers to execute the network stack 698 * in softint (so called softint-based if_input). It utilizes 699 * softint and percpu ifqueue. It doesn't distribute any packets 700 * between CPUs, unlike pktqueue(9). 701 * 702 * Currently we support two options for device drivers to apply the framework: 703 * - Use it implicitly with less changes 704 * - If you use if_attach in driver's _attach function and if_input in 705 * driver's Rx interrupt handler, a packet is queued and a softint handles 706 * the packet implicitly 707 * - Use it explicitly in each driver (recommended) 708 * - You can use if_percpuq_* directly in your driver 709 * - In this case, you need to allocate struct if_percpuq in driver's softc 710 * - See wm(4) as a reference implementation 711 */ 712 713 static void 714 if_percpuq_softint(void *arg) 715 { 716 struct if_percpuq *ipq = arg; 717 struct ifnet *ifp = ipq->ipq_ifp; 718 struct mbuf *m; 719 720 while ((m = if_percpuq_dequeue(ipq)) != NULL) 721 ifp->_if_input(ifp, m); 722 } 723 724 static void 725 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 726 { 727 struct ifqueue *const ifq = p; 728 729 memset(ifq, 0, sizeof(*ifq)); 730 ifq->ifq_maxlen = IFQ_MAXLEN; 731 } 732 733 struct if_percpuq * 734 if_percpuq_create(struct ifnet *ifp) 735 { 736 struct if_percpuq *ipq; 737 738 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 739 if (ipq == NULL) 740 panic("kmem_zalloc failed"); 741 742 ipq->ipq_ifp = ifp; 743 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 744 if_percpuq_softint, ipq); 745 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 746 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 747 748 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 749 750 return ipq; 751 } 752 753 static struct mbuf * 754 if_percpuq_dequeue(struct if_percpuq *ipq) 755 { 756 struct mbuf *m; 757 struct ifqueue *ifq; 758 int s; 759 760 s = splnet(); 761 ifq = percpu_getref(ipq->ipq_ifqs); 762 IF_DEQUEUE(ifq, m); 763 percpu_putref(ipq->ipq_ifqs); 764 splx(s); 765 766 return m; 767 } 768 769 static void 770 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 771 { 772 struct ifqueue *const ifq = p; 773 774 IF_PURGE(ifq); 775 } 776 777 void 778 if_percpuq_destroy(struct if_percpuq *ipq) 779 { 780 781 /* if_detach may already destroy it */ 782 if (ipq == NULL) 783 return; 784 785 softint_disestablish(ipq->ipq_si); 786 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 787 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 788 } 789 790 void 791 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 792 { 793 struct ifqueue *ifq; 794 int s; 795 796 KASSERT(ipq != NULL); 797 798 s = splnet(); 799 ifq = percpu_getref(ipq->ipq_ifqs); 800 if (IF_QFULL(ifq)) { 801 IF_DROP(ifq); 802 percpu_putref(ipq->ipq_ifqs); 803 m_freem(m); 804 goto out; 805 } 806 IF_ENQUEUE(ifq, m); 807 percpu_putref(ipq->ipq_ifqs); 808 809 softint_schedule(ipq->ipq_si); 810 out: 811 splx(s); 812 } 813 814 static void 815 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 816 { 817 struct ifqueue *const ifq = p; 818 int *sum = arg; 819 820 *sum += ifq->ifq_drops; 821 } 822 823 static int 824 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 825 { 826 struct sysctlnode node; 827 struct if_percpuq *ipq; 828 int sum = 0; 829 int error; 830 831 node = *rnode; 832 ipq = node.sysctl_data; 833 834 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 835 836 node.sysctl_data = ∑ 837 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 838 if (error != 0 || newp == NULL) 839 return error; 840 841 return 0; 842 } 843 844 static void 845 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 846 struct if_percpuq *ipq) 847 { 848 const struct sysctlnode *cnode, *rnode; 849 850 if (sysctl_createv(clog, 0, NULL, &rnode, 851 CTLFLAG_PERMANENT, 852 CTLTYPE_NODE, "interfaces", 853 SYSCTL_DESCR("Per-interface controls"), 854 NULL, 0, NULL, 0, 855 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 856 goto bad; 857 858 if (sysctl_createv(clog, 0, &rnode, &rnode, 859 CTLFLAG_PERMANENT, 860 CTLTYPE_NODE, ifname, 861 SYSCTL_DESCR("Interface controls"), 862 NULL, 0, NULL, 0, 863 CTL_CREATE, CTL_EOL) != 0) 864 goto bad; 865 866 if (sysctl_createv(clog, 0, &rnode, &rnode, 867 CTLFLAG_PERMANENT, 868 CTLTYPE_NODE, "rcvq", 869 SYSCTL_DESCR("Interface input queue controls"), 870 NULL, 0, NULL, 0, 871 CTL_CREATE, CTL_EOL) != 0) 872 goto bad; 873 874 #ifdef NOTYET 875 /* XXX Should show each per-CPU queue length? */ 876 if (sysctl_createv(clog, 0, &rnode, &rnode, 877 CTLFLAG_PERMANENT, 878 CTLTYPE_INT, "len", 879 SYSCTL_DESCR("Current input queue length"), 880 sysctl_percpuq_len, 0, NULL, 0, 881 CTL_CREATE, CTL_EOL) != 0) 882 goto bad; 883 884 if (sysctl_createv(clog, 0, &rnode, &cnode, 885 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 886 CTLTYPE_INT, "maxlen", 887 SYSCTL_DESCR("Maximum allowed input queue length"), 888 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 889 CTL_CREATE, CTL_EOL) != 0) 890 goto bad; 891 #endif 892 893 if (sysctl_createv(clog, 0, &rnode, &cnode, 894 CTLFLAG_PERMANENT, 895 CTLTYPE_INT, "drops", 896 SYSCTL_DESCR("Total packets dropped due to full input queue"), 897 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 898 CTL_CREATE, CTL_EOL) != 0) 899 goto bad; 900 901 return; 902 bad: 903 printf("%s: could not attach sysctl nodes\n", ifname); 904 return; 905 } 906 907 908 /* 909 * The common interface input routine that is called by device drivers, 910 * which should be used only when the driver's rx handler already runs 911 * in softint. 912 */ 913 void 914 if_input(struct ifnet *ifp, struct mbuf *m) 915 { 916 917 KASSERT(ifp->if_percpuq == NULL); 918 KASSERT(!cpu_intr_p()); 919 920 ifp->_if_input(ifp, m); 921 } 922 923 /* 924 * DEPRECATED. Use if_initialize and if_register instead. 925 * See the above comment of if_initialize. 926 * 927 * Note that it implicitly enables if_percpuq to make drivers easy to 928 * migrate softint-based if_input without much changes. If you don't 929 * want to enable it, use if_initialize instead. 930 */ 931 void 932 if_attach(ifnet_t *ifp) 933 { 934 935 if_initialize(ifp); 936 ifp->if_percpuq = if_percpuq_create(ifp); 937 if_register(ifp); 938 } 939 940 void 941 if_attachdomain(void) 942 { 943 struct ifnet *ifp; 944 int s; 945 int bound = curlwp_bind(); 946 947 s = pserialize_read_enter(); 948 IFNET_READER_FOREACH(ifp) { 949 struct psref psref; 950 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 951 pserialize_read_exit(s); 952 if_attachdomain1(ifp); 953 s = pserialize_read_enter(); 954 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 955 } 956 pserialize_read_exit(s); 957 curlwp_bindx(bound); 958 } 959 960 static void 961 if_attachdomain1(struct ifnet *ifp) 962 { 963 struct domain *dp; 964 int s; 965 966 s = splnet(); 967 968 /* address family dependent data region */ 969 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 970 DOMAIN_FOREACH(dp) { 971 if (dp->dom_ifattach != NULL) 972 ifp->if_afdata[dp->dom_family] = 973 (*dp->dom_ifattach)(ifp); 974 } 975 976 splx(s); 977 } 978 979 /* 980 * Deactivate an interface. This points all of the procedure 981 * handles at error stubs. May be called from interrupt context. 982 */ 983 void 984 if_deactivate(struct ifnet *ifp) 985 { 986 int s; 987 988 s = splnet(); 989 990 ifp->if_output = if_nulloutput; 991 ifp->_if_input = if_nullinput; 992 ifp->if_start = if_nullstart; 993 ifp->if_transmit = if_nulltransmit; 994 ifp->if_ioctl = if_nullioctl; 995 ifp->if_init = if_nullinit; 996 ifp->if_stop = if_nullstop; 997 ifp->if_slowtimo = if_nullslowtimo; 998 ifp->if_drain = if_nulldrain; 999 1000 /* No more packets may be enqueued. */ 1001 ifp->if_snd.ifq_maxlen = 0; 1002 1003 splx(s); 1004 } 1005 1006 void 1007 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1008 { 1009 struct ifaddr *ifa, *nifa; 1010 1011 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 1012 if (ifa->ifa_addr->sa_family != family) 1013 continue; 1014 (*purgeaddr)(ifa); 1015 } 1016 } 1017 1018 /* 1019 * Detach an interface from the list of "active" interfaces, 1020 * freeing any resources as we go along. 1021 * 1022 * NOTE: This routine must be called with a valid thread context, 1023 * as it may block. 1024 */ 1025 void 1026 if_detach(struct ifnet *ifp) 1027 { 1028 struct socket so; 1029 struct ifaddr *ifa; 1030 #ifdef IFAREF_DEBUG 1031 struct ifaddr *last_ifa = NULL; 1032 #endif 1033 struct domain *dp; 1034 const struct protosw *pr; 1035 int s, i, family, purged; 1036 uint64_t xc; 1037 1038 /* 1039 * XXX It's kind of lame that we have to have the 1040 * XXX socket structure... 1041 */ 1042 memset(&so, 0, sizeof(so)); 1043 1044 s = splnet(); 1045 1046 sysctl_teardown(&ifp->if_sysctl_log); 1047 mutex_enter(ifp->if_ioctl_lock); 1048 ifp->if_ioctl = if_nullioctl; 1049 mutex_exit(ifp->if_ioctl_lock); 1050 1051 IFNET_LOCK(); 1052 ifindex2ifnet[ifp->if_index] = NULL; 1053 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1054 IFNET_WRITER_REMOVE(ifp); 1055 pserialize_perform(ifnet_psz); 1056 IFNET_UNLOCK(); 1057 1058 mutex_obj_free(ifp->if_ioctl_lock); 1059 ifp->if_ioctl_lock = NULL; 1060 1061 if (ifp->if_slowtimo != NULL) { 1062 ifp->if_slowtimo = NULL; 1063 callout_halt(ifp->if_slowtimo_ch, NULL); 1064 callout_destroy(ifp->if_slowtimo_ch); 1065 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1066 } 1067 1068 /* 1069 * Do an if_down() to give protocols a chance to do something. 1070 */ 1071 if_down(ifp); 1072 1073 #ifdef ALTQ 1074 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1075 altq_disable(&ifp->if_snd); 1076 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1077 altq_detach(&ifp->if_snd); 1078 #endif 1079 1080 if (ifp->if_snd.ifq_lock) 1081 mutex_obj_free(ifp->if_snd.ifq_lock); 1082 1083 #if NCARP > 0 1084 /* Remove the interface from any carp group it is a part of. */ 1085 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1086 carp_ifdetach(ifp); 1087 #endif 1088 1089 /* 1090 * Rip all the addresses off the interface. This should make 1091 * all of the routes go away. 1092 * 1093 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1094 * from the list, including our "cursor", ifa. For safety, 1095 * and to honor the TAILQ abstraction, I just restart the 1096 * loop after each removal. Note that the loop will exit 1097 * when all of the remaining ifaddrs belong to the AF_LINK 1098 * family. I am counting on the historical fact that at 1099 * least one pr_usrreq in each address domain removes at 1100 * least one ifaddr. 1101 */ 1102 again: 1103 IFADDR_FOREACH(ifa, ifp) { 1104 family = ifa->ifa_addr->sa_family; 1105 #ifdef IFAREF_DEBUG 1106 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1107 ifa, family, ifa->ifa_refcnt); 1108 if (last_ifa != NULL && ifa == last_ifa) 1109 panic("if_detach: loop detected"); 1110 last_ifa = ifa; 1111 #endif 1112 if (family == AF_LINK) 1113 continue; 1114 dp = pffinddomain(family); 1115 #ifdef DIAGNOSTIC 1116 if (dp == NULL) 1117 panic("if_detach: no domain for AF %d", 1118 family); 1119 #endif 1120 /* 1121 * XXX These PURGEIF calls are redundant with the 1122 * purge-all-families calls below, but are left in for 1123 * now both to make a smaller change, and to avoid 1124 * unplanned interactions with clearing of 1125 * ifp->if_addrlist. 1126 */ 1127 purged = 0; 1128 for (pr = dp->dom_protosw; 1129 pr < dp->dom_protoswNPROTOSW; pr++) { 1130 so.so_proto = pr; 1131 if (pr->pr_usrreqs) { 1132 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1133 purged = 1; 1134 } 1135 } 1136 if (purged == 0) { 1137 /* 1138 * XXX What's really the best thing to do 1139 * XXX here? --thorpej (at) NetBSD.org 1140 */ 1141 printf("if_detach: WARNING: AF %d not purged\n", 1142 family); 1143 ifa_remove(ifp, ifa); 1144 } 1145 goto again; 1146 } 1147 1148 if_free_sadl(ifp); 1149 1150 /* Walk the routing table looking for stragglers. */ 1151 for (i = 0; i <= AF_MAX; i++) { 1152 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 1153 continue; 1154 } 1155 1156 DOMAIN_FOREACH(dp) { 1157 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1158 { 1159 void *p = ifp->if_afdata[dp->dom_family]; 1160 if (p) { 1161 ifp->if_afdata[dp->dom_family] = NULL; 1162 (*dp->dom_ifdetach)(ifp, p); 1163 } 1164 } 1165 1166 /* 1167 * One would expect multicast memberships (INET and 1168 * INET6) on UDP sockets to be purged by the PURGEIF 1169 * calls above, but if all addresses were removed from 1170 * the interface prior to destruction, the calls will 1171 * not be made (e.g. ppp, for which pppd(8) generally 1172 * removes addresses before destroying the interface). 1173 * Because there is no invariant that multicast 1174 * memberships only exist for interfaces with IPv4 1175 * addresses, we must call PURGEIF regardless of 1176 * addresses. (Protocols which might store ifnet 1177 * pointers are marked with PR_PURGEIF.) 1178 */ 1179 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1180 so.so_proto = pr; 1181 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1182 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1183 } 1184 } 1185 1186 (void)pfil_run_hooks(if_pfil, 1187 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 1188 (void)pfil_head_destroy(ifp->if_pfil); 1189 1190 /* Announce that the interface is gone. */ 1191 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1192 1193 IF_AFDATA_LOCK_DESTROY(ifp); 1194 1195 softint_disestablish(ifp->if_link_si); 1196 ifp->if_link_si = NULL; 1197 1198 /* 1199 * remove packets that came from ifp, from software interrupt queues. 1200 */ 1201 DOMAIN_FOREACH(dp) { 1202 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1203 struct ifqueue *iq = dp->dom_ifqueues[i]; 1204 if (iq == NULL) 1205 break; 1206 dp->dom_ifqueues[i] = NULL; 1207 if_detach_queues(ifp, iq); 1208 } 1209 } 1210 1211 /* 1212 * IP queues have to be processed separately: net-queue barrier 1213 * ensures that the packets are dequeued while a cross-call will 1214 * ensure that the interrupts have completed. FIXME: not quite.. 1215 */ 1216 #ifdef INET 1217 pktq_barrier(ip_pktq); 1218 #endif 1219 #ifdef INET6 1220 if (in6_present) 1221 pktq_barrier(ip6_pktq); 1222 #endif 1223 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1224 xc_wait(xc); 1225 1226 /* Wait for all readers to drain before freeing. */ 1227 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1228 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1229 1230 if (ifp->if_percpuq != NULL) { 1231 if_percpuq_destroy(ifp->if_percpuq); 1232 ifp->if_percpuq = NULL; 1233 } 1234 1235 splx(s); 1236 } 1237 1238 static void 1239 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1240 { 1241 struct mbuf *m, *prev, *next; 1242 1243 prev = NULL; 1244 for (m = q->ifq_head; m != NULL; m = next) { 1245 KASSERT((m->m_flags & M_PKTHDR) != 0); 1246 1247 next = m->m_nextpkt; 1248 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1249 prev = m; 1250 continue; 1251 } 1252 1253 if (prev != NULL) 1254 prev->m_nextpkt = m->m_nextpkt; 1255 else 1256 q->ifq_head = m->m_nextpkt; 1257 if (q->ifq_tail == m) 1258 q->ifq_tail = prev; 1259 q->ifq_len--; 1260 1261 m->m_nextpkt = NULL; 1262 m_freem(m); 1263 IF_DROP(q); 1264 } 1265 } 1266 1267 /* 1268 * Callback for a radix tree walk to delete all references to an 1269 * ifnet. 1270 */ 1271 static int 1272 if_rt_walktree(struct rtentry *rt, void *v) 1273 { 1274 struct ifnet *ifp = (struct ifnet *)v; 1275 int error; 1276 struct rtentry *retrt; 1277 1278 if (rt->rt_ifp != ifp) 1279 return 0; 1280 1281 /* Delete the entry. */ 1282 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 1283 rt_mask(rt), rt->rt_flags, &retrt); 1284 if (error == 0) { 1285 KASSERT(retrt == rt); 1286 KASSERT((retrt->rt_flags & RTF_UP) == 0); 1287 retrt->rt_ifp = NULL; 1288 rtfree(retrt); 1289 } else { 1290 printf("%s: warning: unable to delete rtentry @ %p, " 1291 "error = %d\n", ifp->if_xname, rt, error); 1292 } 1293 return ERESTART; 1294 } 1295 1296 /* 1297 * Create a clone network interface. 1298 */ 1299 static int 1300 if_clone_create(const char *name) 1301 { 1302 struct if_clone *ifc; 1303 int unit; 1304 struct ifnet *ifp; 1305 struct psref psref; 1306 1307 ifc = if_clone_lookup(name, &unit); 1308 if (ifc == NULL) 1309 return EINVAL; 1310 1311 ifp = if_get(name, &psref); 1312 if (ifp != NULL) { 1313 if_put(ifp, &psref); 1314 return EEXIST; 1315 } 1316 1317 return (*ifc->ifc_create)(ifc, unit); 1318 } 1319 1320 /* 1321 * Destroy a clone network interface. 1322 */ 1323 static int 1324 if_clone_destroy(const char *name) 1325 { 1326 struct if_clone *ifc; 1327 struct ifnet *ifp; 1328 struct psref psref; 1329 1330 ifc = if_clone_lookup(name, NULL); 1331 if (ifc == NULL) 1332 return EINVAL; 1333 1334 if (ifc->ifc_destroy == NULL) 1335 return EOPNOTSUPP; 1336 1337 ifp = if_get(name, &psref); 1338 if (ifp == NULL) 1339 return ENXIO; 1340 1341 /* We have to disable ioctls here */ 1342 mutex_enter(ifp->if_ioctl_lock); 1343 ifp->if_ioctl = if_nullioctl; 1344 mutex_exit(ifp->if_ioctl_lock); 1345 1346 /* 1347 * We cannot call ifc_destroy with holding ifp. 1348 * Releasing ifp here is safe thanks to if_clone_mtx. 1349 */ 1350 if_put(ifp, &psref); 1351 1352 return (*ifc->ifc_destroy)(ifp); 1353 } 1354 1355 /* 1356 * Look up a network interface cloner. 1357 */ 1358 static struct if_clone * 1359 if_clone_lookup(const char *name, int *unitp) 1360 { 1361 struct if_clone *ifc; 1362 const char *cp; 1363 char *dp, ifname[IFNAMSIZ + 3]; 1364 int unit; 1365 1366 strcpy(ifname, "if_"); 1367 /* separate interface name from unit */ 1368 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1369 *cp && (*cp < '0' || *cp > '9');) 1370 *dp++ = *cp++; 1371 1372 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1373 return NULL; /* No name or unit number */ 1374 *dp++ = '\0'; 1375 1376 again: 1377 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1378 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1379 break; 1380 } 1381 1382 if (ifc == NULL) { 1383 if (*ifname == '\0' || 1384 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1385 return NULL; 1386 *ifname = '\0'; 1387 goto again; 1388 } 1389 1390 unit = 0; 1391 while (cp - name < IFNAMSIZ && *cp) { 1392 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1393 /* Bogus unit number. */ 1394 return NULL; 1395 } 1396 unit = (unit * 10) + (*cp++ - '0'); 1397 } 1398 1399 if (unitp != NULL) 1400 *unitp = unit; 1401 return ifc; 1402 } 1403 1404 /* 1405 * Register a network interface cloner. 1406 */ 1407 void 1408 if_clone_attach(struct if_clone *ifc) 1409 { 1410 1411 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1412 if_cloners_count++; 1413 } 1414 1415 /* 1416 * Unregister a network interface cloner. 1417 */ 1418 void 1419 if_clone_detach(struct if_clone *ifc) 1420 { 1421 1422 LIST_REMOVE(ifc, ifc_list); 1423 if_cloners_count--; 1424 } 1425 1426 /* 1427 * Provide list of interface cloners to userspace. 1428 */ 1429 int 1430 if_clone_list(int buf_count, char *buffer, int *total) 1431 { 1432 char outbuf[IFNAMSIZ], *dst; 1433 struct if_clone *ifc; 1434 int count, error = 0; 1435 1436 *total = if_cloners_count; 1437 if ((dst = buffer) == NULL) { 1438 /* Just asking how many there are. */ 1439 return 0; 1440 } 1441 1442 if (buf_count < 0) 1443 return EINVAL; 1444 1445 count = (if_cloners_count < buf_count) ? 1446 if_cloners_count : buf_count; 1447 1448 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1449 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1450 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1451 if (outbuf[sizeof(outbuf) - 1] != '\0') 1452 return ENAMETOOLONG; 1453 error = copyout(outbuf, dst, sizeof(outbuf)); 1454 if (error != 0) 1455 break; 1456 } 1457 1458 return error; 1459 } 1460 1461 void 1462 ifaref(struct ifaddr *ifa) 1463 { 1464 ifa->ifa_refcnt++; 1465 } 1466 1467 void 1468 ifafree(struct ifaddr *ifa) 1469 { 1470 KASSERT(ifa != NULL); 1471 KASSERT(ifa->ifa_refcnt > 0); 1472 1473 if (--ifa->ifa_refcnt == 0) { 1474 free(ifa, M_IFADDR); 1475 } 1476 } 1477 1478 void 1479 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1480 { 1481 ifa->ifa_ifp = ifp; 1482 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1483 ifaref(ifa); 1484 } 1485 1486 void 1487 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1488 { 1489 KASSERT(ifa->ifa_ifp == ifp); 1490 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1491 ifafree(ifa); 1492 } 1493 1494 static inline int 1495 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1496 { 1497 return sockaddr_cmp(sa1, sa2) == 0; 1498 } 1499 1500 /* 1501 * Locate an interface based on a complete address. 1502 */ 1503 /*ARGSUSED*/ 1504 struct ifaddr * 1505 ifa_ifwithaddr(const struct sockaddr *addr) 1506 { 1507 struct ifnet *ifp; 1508 struct ifaddr *ifa; 1509 int s; 1510 1511 s = pserialize_read_enter(); 1512 IFNET_READER_FOREACH(ifp) { 1513 if (ifp->if_output == if_nulloutput) 1514 continue; 1515 IFADDR_FOREACH(ifa, ifp) { 1516 if (ifa->ifa_addr->sa_family != addr->sa_family) 1517 continue; 1518 if (equal(addr, ifa->ifa_addr)) 1519 return ifa; 1520 if ((ifp->if_flags & IFF_BROADCAST) && 1521 ifa->ifa_broadaddr && 1522 /* IP6 doesn't have broadcast */ 1523 ifa->ifa_broadaddr->sa_len != 0 && 1524 equal(ifa->ifa_broadaddr, addr)) 1525 return ifa; 1526 } 1527 } 1528 pserialize_read_exit(s); 1529 return NULL; 1530 } 1531 1532 /* 1533 * Locate the point to point interface with a given destination address. 1534 */ 1535 /*ARGSUSED*/ 1536 struct ifaddr * 1537 ifa_ifwithdstaddr(const struct sockaddr *addr) 1538 { 1539 struct ifnet *ifp; 1540 struct ifaddr *ifa; 1541 int s; 1542 1543 s = pserialize_read_enter(); 1544 IFNET_READER_FOREACH(ifp) { 1545 if (ifp->if_output == if_nulloutput) 1546 continue; 1547 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1548 continue; 1549 IFADDR_FOREACH(ifa, ifp) { 1550 if (ifa->ifa_addr->sa_family != addr->sa_family || 1551 ifa->ifa_dstaddr == NULL) 1552 continue; 1553 if (equal(addr, ifa->ifa_dstaddr)) 1554 return ifa; 1555 } 1556 } 1557 pserialize_read_exit(s); 1558 return NULL; 1559 } 1560 1561 /* 1562 * Find an interface on a specific network. If many, choice 1563 * is most specific found. 1564 */ 1565 struct ifaddr * 1566 ifa_ifwithnet(const struct sockaddr *addr) 1567 { 1568 struct ifnet *ifp; 1569 struct ifaddr *ifa; 1570 const struct sockaddr_dl *sdl; 1571 struct ifaddr *ifa_maybe = 0; 1572 u_int af = addr->sa_family; 1573 const char *addr_data = addr->sa_data, *cplim; 1574 int s; 1575 1576 if (af == AF_LINK) { 1577 sdl = satocsdl(addr); 1578 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1579 ifindex2ifnet[sdl->sdl_index] && 1580 ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) { 1581 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1582 } 1583 } 1584 #ifdef NETATALK 1585 if (af == AF_APPLETALK) { 1586 const struct sockaddr_at *sat, *sat2; 1587 sat = (const struct sockaddr_at *)addr; 1588 s = pserialize_read_enter(); 1589 IFNET_READER_FOREACH(ifp) { 1590 if (ifp->if_output == if_nulloutput) 1591 continue; 1592 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1593 if (ifa == NULL) 1594 continue; 1595 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1596 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1597 return ifa; /* exact match */ 1598 if (ifa_maybe == NULL) { 1599 /* else keep the if with the right range */ 1600 ifa_maybe = ifa; 1601 } 1602 } 1603 pserialize_read_exit(s); 1604 return ifa_maybe; 1605 } 1606 #endif 1607 s = pserialize_read_enter(); 1608 IFNET_READER_FOREACH(ifp) { 1609 if (ifp->if_output == if_nulloutput) 1610 continue; 1611 IFADDR_FOREACH(ifa, ifp) { 1612 const char *cp, *cp2, *cp3; 1613 1614 if (ifa->ifa_addr->sa_family != af || 1615 ifa->ifa_netmask == NULL) 1616 next: continue; 1617 cp = addr_data; 1618 cp2 = ifa->ifa_addr->sa_data; 1619 cp3 = ifa->ifa_netmask->sa_data; 1620 cplim = (const char *)ifa->ifa_netmask + 1621 ifa->ifa_netmask->sa_len; 1622 while (cp3 < cplim) { 1623 if ((*cp++ ^ *cp2++) & *cp3++) { 1624 /* want to continue for() loop */ 1625 goto next; 1626 } 1627 } 1628 if (ifa_maybe == NULL || 1629 rt_refines(ifa->ifa_netmask, 1630 ifa_maybe->ifa_netmask)) 1631 ifa_maybe = ifa; 1632 } 1633 } 1634 pserialize_read_exit(s); 1635 return ifa_maybe; 1636 } 1637 1638 /* 1639 * Find the interface of the addresss. 1640 */ 1641 struct ifaddr * 1642 ifa_ifwithladdr(const struct sockaddr *addr) 1643 { 1644 struct ifaddr *ia; 1645 1646 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1647 (ia = ifa_ifwithnet(addr))) 1648 return ia; 1649 return NULL; 1650 } 1651 1652 /* 1653 * Find an interface using a specific address family 1654 */ 1655 struct ifaddr * 1656 ifa_ifwithaf(int af) 1657 { 1658 struct ifnet *ifp; 1659 struct ifaddr *ifa = NULL; 1660 int s; 1661 1662 s = pserialize_read_enter(); 1663 IFNET_READER_FOREACH(ifp) { 1664 if (ifp->if_output == if_nulloutput) 1665 continue; 1666 IFADDR_FOREACH(ifa, ifp) { 1667 if (ifa->ifa_addr->sa_family == af) 1668 goto out; 1669 } 1670 } 1671 out: 1672 pserialize_read_exit(s); 1673 return ifa; 1674 } 1675 1676 /* 1677 * Find an interface address specific to an interface best matching 1678 * a given address. 1679 */ 1680 struct ifaddr * 1681 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1682 { 1683 struct ifaddr *ifa; 1684 const char *cp, *cp2, *cp3; 1685 const char *cplim; 1686 struct ifaddr *ifa_maybe = 0; 1687 u_int af = addr->sa_family; 1688 1689 if (ifp->if_output == if_nulloutput) 1690 return NULL; 1691 1692 if (af >= AF_MAX) 1693 return NULL; 1694 1695 IFADDR_FOREACH(ifa, ifp) { 1696 if (ifa->ifa_addr->sa_family != af) 1697 continue; 1698 ifa_maybe = ifa; 1699 if (ifa->ifa_netmask == NULL) { 1700 if (equal(addr, ifa->ifa_addr) || 1701 (ifa->ifa_dstaddr && 1702 equal(addr, ifa->ifa_dstaddr))) 1703 return ifa; 1704 continue; 1705 } 1706 cp = addr->sa_data; 1707 cp2 = ifa->ifa_addr->sa_data; 1708 cp3 = ifa->ifa_netmask->sa_data; 1709 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1710 for (; cp3 < cplim; cp3++) { 1711 if ((*cp++ ^ *cp2++) & *cp3) 1712 break; 1713 } 1714 if (cp3 == cplim) 1715 return ifa; 1716 } 1717 return ifa_maybe; 1718 } 1719 1720 /* 1721 * Default action when installing a route with a Link Level gateway. 1722 * Lookup an appropriate real ifa to point to. 1723 * This should be moved to /sys/net/link.c eventually. 1724 */ 1725 void 1726 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1727 { 1728 struct ifaddr *ifa; 1729 const struct sockaddr *dst; 1730 struct ifnet *ifp; 1731 1732 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1733 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1734 return; 1735 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1736 rt_replace_ifa(rt, ifa); 1737 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1738 ifa->ifa_rtrequest(cmd, rt, info); 1739 } 1740 } 1741 1742 /* 1743 * bitmask macros to manage a densely packed link_state change queue. 1744 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 1745 * LINK_STATE_UP(2) we need 2 bits for each state change. 1746 * As a state change to store is 0, treat all bits set as an unset item. 1747 */ 1748 #define LQ_ITEM_BITS 2 1749 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 1750 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 1751 #define LINK_STATE_UNSET LQ_ITEM_MASK 1752 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 1753 #define LQ_STORE(q, i, v) \ 1754 do { \ 1755 (q) &= ~LQ_MASK((i)); \ 1756 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 1757 } while (0 /* CONSTCOND */) 1758 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 1759 #define LQ_POP(q, v) \ 1760 do { \ 1761 (v) = LQ_ITEM((q), 0); \ 1762 (q) >>= LQ_ITEM_BITS; \ 1763 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1764 } while (0 /* CONSTCOND */) 1765 #define LQ_PUSH(q, v) \ 1766 do { \ 1767 (q) >>= LQ_ITEM_BITS; \ 1768 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1769 } while (0 /* CONSTCOND */) 1770 #define LQ_FIND_UNSET(q, i) \ 1771 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 1772 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 1773 break; \ 1774 } 1775 /* 1776 * Handle a change in the interface link state and 1777 * queue notifications. 1778 */ 1779 void 1780 if_link_state_change(struct ifnet *ifp, int link_state) 1781 { 1782 int s, idx; 1783 1784 /* Ensure change is to a valid state */ 1785 switch (link_state) { 1786 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 1787 case LINK_STATE_DOWN: /* FALLTHROUGH */ 1788 case LINK_STATE_UP: 1789 break; 1790 default: 1791 #ifdef DEBUG 1792 printf("%s: invalid link state %d\n", 1793 ifp->if_xname, link_state); 1794 #endif 1795 return; 1796 } 1797 1798 s = splnet(); 1799 1800 /* Find the last unset event in the queue. */ 1801 LQ_FIND_UNSET(ifp->if_link_queue, idx); 1802 1803 /* 1804 * Ensure link_state doesn't match the last event in the queue. 1805 * ifp->if_link_state is not checked and set here because 1806 * that would present an inconsistent picture to the system. 1807 */ 1808 if (idx != 0 && 1809 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 1810 goto out; 1811 1812 /* Handle queue overflow. */ 1813 if (idx == LQ_MAX(ifp->if_link_queue)) { 1814 uint8_t lost; 1815 1816 /* 1817 * The DOWN state must be protected from being pushed off 1818 * the queue to ensure that userland will always be 1819 * in a sane state. 1820 * Because DOWN is protected, there is no need to protect 1821 * UNKNOWN. 1822 * It should be invalid to change from any other state to 1823 * UNKNOWN anyway ... 1824 */ 1825 lost = LQ_ITEM(ifp->if_link_queue, 0); 1826 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 1827 if (lost == LINK_STATE_DOWN) { 1828 lost = LQ_ITEM(ifp->if_link_queue, 0); 1829 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 1830 } 1831 printf("%s: lost link state change %s\n", 1832 ifp->if_xname, 1833 lost == LINK_STATE_UP ? "UP" : 1834 lost == LINK_STATE_DOWN ? "DOWN" : 1835 "UNKNOWN"); 1836 } else 1837 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 1838 1839 softint_schedule(ifp->if_link_si); 1840 1841 out: 1842 splx(s); 1843 } 1844 1845 /* 1846 * Handle interface link state change notifications. 1847 * Must be called at splnet(). 1848 */ 1849 static void 1850 if_link_state_change0(struct ifnet *ifp, int link_state) 1851 { 1852 struct domain *dp; 1853 1854 /* Ensure the change is still valid. */ 1855 if (ifp->if_link_state == link_state) 1856 return; 1857 1858 #ifdef DEBUG 1859 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1860 link_state == LINK_STATE_UP ? "UP" : 1861 link_state == LINK_STATE_DOWN ? "DOWN" : 1862 "UNKNOWN", 1863 ifp->if_link_state == LINK_STATE_UP ? "UP" : 1864 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 1865 "UNKNOWN"); 1866 #endif 1867 1868 /* 1869 * When going from UNKNOWN to UP, we need to mark existing 1870 * addresses as tentative and restart DAD as we may have 1871 * erroneously not found a duplicate. 1872 * 1873 * This needs to happen before rt_ifmsg to avoid a race where 1874 * listeners would have an address and expect it to work right 1875 * away. 1876 */ 1877 if (link_state == LINK_STATE_UP && 1878 ifp->if_link_state == LINK_STATE_UNKNOWN) 1879 { 1880 DOMAIN_FOREACH(dp) { 1881 if (dp->dom_if_link_state_change != NULL) 1882 dp->dom_if_link_state_change(ifp, 1883 LINK_STATE_DOWN); 1884 } 1885 } 1886 1887 ifp->if_link_state = link_state; 1888 1889 /* Notify that the link state has changed. */ 1890 rt_ifmsg(ifp); 1891 1892 #if NCARP > 0 1893 if (ifp->if_carp) 1894 carp_carpdev_state(ifp); 1895 #endif 1896 1897 DOMAIN_FOREACH(dp) { 1898 if (dp->dom_if_link_state_change != NULL) 1899 dp->dom_if_link_state_change(ifp, link_state); 1900 } 1901 } 1902 1903 /* 1904 * Process the interface link state change queue. 1905 */ 1906 static void 1907 if_link_state_change_si(void *arg) 1908 { 1909 struct ifnet *ifp = arg; 1910 int s; 1911 uint8_t state; 1912 1913 s = splnet(); 1914 1915 /* Pop a link state change from the queue and process it. */ 1916 LQ_POP(ifp->if_link_queue, state); 1917 if_link_state_change0(ifp, state); 1918 1919 /* If there is a link state change to come, schedule it. */ 1920 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 1921 softint_schedule(ifp->if_link_si); 1922 1923 splx(s); 1924 } 1925 1926 /* 1927 * Default action when installing a local route on a point-to-point 1928 * interface. 1929 */ 1930 void 1931 p2p_rtrequest(int req, struct rtentry *rt, 1932 __unused const struct rt_addrinfo *info) 1933 { 1934 struct ifnet *ifp = rt->rt_ifp; 1935 struct ifaddr *ifa, *lo0ifa; 1936 1937 switch (req) { 1938 case RTM_ADD: 1939 if ((rt->rt_flags & RTF_LOCAL) == 0) 1940 break; 1941 1942 IFADDR_FOREACH(ifa, ifp) { 1943 if (equal(rt_getkey(rt), ifa->ifa_addr)) 1944 break; 1945 } 1946 if (ifa == NULL) 1947 break; 1948 1949 /* 1950 * Ensure lo0 has an address of the same family. 1951 */ 1952 IFADDR_FOREACH(lo0ifa, lo0ifp) { 1953 if (lo0ifa->ifa_addr->sa_family == 1954 ifa->ifa_addr->sa_family) 1955 break; 1956 } 1957 if (lo0ifa == NULL) 1958 break; 1959 1960 rt->rt_ifp = lo0ifp; 1961 1962 /* 1963 * Make sure to set rt->rt_ifa to the interface 1964 * address we are using, otherwise we will have trouble 1965 * with source address selection. 1966 */ 1967 if (ifa != rt->rt_ifa) 1968 rt_replace_ifa(rt, ifa); 1969 break; 1970 case RTM_DELETE: 1971 default: 1972 break; 1973 } 1974 } 1975 1976 /* 1977 * Mark an interface down and notify protocols of 1978 * the transition. 1979 * NOTE: must be called at splsoftnet or equivalent. 1980 */ 1981 void 1982 if_down(struct ifnet *ifp) 1983 { 1984 struct ifaddr *ifa; 1985 struct domain *dp; 1986 1987 ifp->if_flags &= ~IFF_UP; 1988 nanotime(&ifp->if_lastchange); 1989 IFADDR_FOREACH(ifa, ifp) 1990 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1991 IFQ_PURGE(&ifp->if_snd); 1992 #if NCARP > 0 1993 if (ifp->if_carp) 1994 carp_carpdev_state(ifp); 1995 #endif 1996 rt_ifmsg(ifp); 1997 DOMAIN_FOREACH(dp) { 1998 if (dp->dom_if_down) 1999 dp->dom_if_down(ifp); 2000 } 2001 } 2002 2003 /* 2004 * Mark an interface up and notify protocols of 2005 * the transition. 2006 * NOTE: must be called at splsoftnet or equivalent. 2007 */ 2008 void 2009 if_up(struct ifnet *ifp) 2010 { 2011 #ifdef notyet 2012 struct ifaddr *ifa; 2013 #endif 2014 struct domain *dp; 2015 2016 ifp->if_flags |= IFF_UP; 2017 nanotime(&ifp->if_lastchange); 2018 #ifdef notyet 2019 /* this has no effect on IP, and will kill all ISO connections XXX */ 2020 IFADDR_FOREACH(ifa, ifp) 2021 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2022 #endif 2023 #if NCARP > 0 2024 if (ifp->if_carp) 2025 carp_carpdev_state(ifp); 2026 #endif 2027 rt_ifmsg(ifp); 2028 DOMAIN_FOREACH(dp) { 2029 if (dp->dom_if_up) 2030 dp->dom_if_up(ifp); 2031 } 2032 } 2033 2034 /* 2035 * Handle interface slowtimo timer routine. Called 2036 * from softclock, we decrement timer (if set) and 2037 * call the appropriate interface routine on expiration. 2038 */ 2039 static void 2040 if_slowtimo(void *arg) 2041 { 2042 void (*slowtimo)(struct ifnet *); 2043 struct ifnet *ifp = arg; 2044 int s; 2045 2046 slowtimo = ifp->if_slowtimo; 2047 if (__predict_false(slowtimo == NULL)) 2048 return; 2049 2050 s = splnet(); 2051 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2052 (*slowtimo)(ifp); 2053 2054 splx(s); 2055 2056 if (__predict_true(ifp->if_slowtimo != NULL)) 2057 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2058 } 2059 2060 /* 2061 * Set/clear promiscuous mode on interface ifp based on the truth value 2062 * of pswitch. The calls are reference counted so that only the first 2063 * "on" request actually has an effect, as does the final "off" request. 2064 * Results are undefined if the "off" and "on" requests are not matched. 2065 */ 2066 int 2067 ifpromisc(struct ifnet *ifp, int pswitch) 2068 { 2069 int pcount, ret; 2070 short nflags; 2071 2072 pcount = ifp->if_pcount; 2073 if (pswitch) { 2074 /* 2075 * Allow the device to be "placed" into promiscuous 2076 * mode even if it is not configured up. It will 2077 * consult IFF_PROMISC when it is brought up. 2078 */ 2079 if (ifp->if_pcount++ != 0) 2080 return 0; 2081 nflags = ifp->if_flags | IFF_PROMISC; 2082 } else { 2083 if (--ifp->if_pcount > 0) 2084 return 0; 2085 nflags = ifp->if_flags & ~IFF_PROMISC; 2086 } 2087 ret = if_flags_set(ifp, nflags); 2088 /* Restore interface state if not successful. */ 2089 if (ret != 0) { 2090 ifp->if_pcount = pcount; 2091 } 2092 return ret; 2093 } 2094 2095 /* 2096 * Map interface name to 2097 * interface structure pointer. 2098 */ 2099 struct ifnet * 2100 ifunit(const char *name) 2101 { 2102 struct ifnet *ifp; 2103 const char *cp = name; 2104 u_int unit = 0; 2105 u_int i; 2106 int s; 2107 2108 /* 2109 * If the entire name is a number, treat it as an ifindex. 2110 */ 2111 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2112 unit = unit * 10 + (*cp - '0'); 2113 } 2114 2115 /* 2116 * If the number took all of the name, then it's a valid ifindex. 2117 */ 2118 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2119 if (unit >= if_indexlim) 2120 return NULL; 2121 ifp = ifindex2ifnet[unit]; 2122 if (ifp == NULL || ifp->if_output == if_nulloutput) 2123 return NULL; 2124 return ifp; 2125 } 2126 2127 ifp = NULL; 2128 s = pserialize_read_enter(); 2129 IFNET_READER_FOREACH(ifp) { 2130 if (ifp->if_output == if_nulloutput) 2131 continue; 2132 if (strcmp(ifp->if_xname, name) == 0) 2133 goto out; 2134 } 2135 out: 2136 pserialize_read_exit(s); 2137 return ifp; 2138 } 2139 2140 /* 2141 * Get a reference of an ifnet object by an interface name. 2142 * The returned reference is protected by psref(9). The caller 2143 * must release a returned reference by if_put after use. 2144 */ 2145 struct ifnet * 2146 if_get(const char *name, struct psref *psref) 2147 { 2148 struct ifnet *ifp; 2149 const char *cp = name; 2150 u_int unit = 0; 2151 u_int i; 2152 int s; 2153 2154 /* 2155 * If the entire name is a number, treat it as an ifindex. 2156 */ 2157 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2158 unit = unit * 10 + (*cp - '0'); 2159 } 2160 2161 /* 2162 * If the number took all of the name, then it's a valid ifindex. 2163 */ 2164 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2165 if (unit >= if_indexlim) 2166 return NULL; 2167 ifp = ifindex2ifnet[unit]; 2168 if (ifp == NULL || ifp->if_output == if_nulloutput) 2169 return NULL; 2170 return ifp; 2171 } 2172 2173 ifp = NULL; 2174 s = pserialize_read_enter(); 2175 IFNET_READER_FOREACH(ifp) { 2176 if (ifp->if_output == if_nulloutput) 2177 continue; 2178 if (strcmp(ifp->if_xname, name) == 0) { 2179 psref_acquire(psref, &ifp->if_psref, 2180 ifnet_psref_class); 2181 goto out; 2182 } 2183 } 2184 out: 2185 pserialize_read_exit(s); 2186 return ifp; 2187 } 2188 2189 /* 2190 * Release a reference of an ifnet object given by if_get or 2191 * if_get_byindex. 2192 */ 2193 void 2194 if_put(const struct ifnet *ifp, struct psref *psref) 2195 { 2196 2197 if (ifp == NULL) 2198 return; 2199 2200 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2201 } 2202 2203 ifnet_t * 2204 if_byindex(u_int idx) 2205 { 2206 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2207 } 2208 2209 /* 2210 * Get a reference of an ifnet object by an interface index. 2211 * The returned reference is protected by psref(9). The caller 2212 * must release a returned reference by if_put after use. 2213 */ 2214 ifnet_t * 2215 if_get_byindex(u_int idx, struct psref *psref) 2216 { 2217 ifnet_t *ifp; 2218 int s; 2219 2220 s = pserialize_read_enter(); 2221 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2222 if (__predict_true(ifp != NULL)) 2223 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2224 pserialize_read_exit(s); 2225 2226 return ifp; 2227 } 2228 2229 /* 2230 * XXX it's safe only if the passed ifp is guaranteed to not be freed, 2231 * for example the ifp is already held or some other object is held which 2232 * guarantes the ifp to not be freed indirectly. 2233 */ 2234 void 2235 if_acquire_NOMPSAFE(struct ifnet *ifp, struct psref *psref) 2236 { 2237 2238 KASSERT(ifp->if_index != 0); 2239 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2240 } 2241 2242 bool 2243 if_held(struct ifnet *ifp) 2244 { 2245 2246 return psref_held(&ifp->if_psref, ifnet_psref_class); 2247 } 2248 2249 2250 /* common */ 2251 int 2252 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2253 { 2254 int s; 2255 struct ifreq *ifr; 2256 struct ifcapreq *ifcr; 2257 struct ifdatareq *ifdr; 2258 2259 switch (cmd) { 2260 case SIOCSIFCAP: 2261 ifcr = data; 2262 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2263 return EINVAL; 2264 2265 if (ifcr->ifcr_capenable == ifp->if_capenable) 2266 return 0; 2267 2268 ifp->if_capenable = ifcr->ifcr_capenable; 2269 2270 /* Pre-compute the checksum flags mask. */ 2271 ifp->if_csum_flags_tx = 0; 2272 ifp->if_csum_flags_rx = 0; 2273 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2274 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2275 } 2276 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2277 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2278 } 2279 2280 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2281 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2282 } 2283 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2284 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2285 } 2286 2287 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2288 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2289 } 2290 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2291 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2292 } 2293 2294 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2295 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2296 } 2297 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2298 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2299 } 2300 2301 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2302 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2303 } 2304 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2305 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2306 } 2307 if (ifp->if_flags & IFF_UP) 2308 return ENETRESET; 2309 return 0; 2310 case SIOCSIFFLAGS: 2311 ifr = data; 2312 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2313 s = splnet(); 2314 if_down(ifp); 2315 splx(s); 2316 } 2317 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2318 s = splnet(); 2319 if_up(ifp); 2320 splx(s); 2321 } 2322 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2323 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2324 break; 2325 case SIOCGIFFLAGS: 2326 ifr = data; 2327 ifr->ifr_flags = ifp->if_flags; 2328 break; 2329 2330 case SIOCGIFMETRIC: 2331 ifr = data; 2332 ifr->ifr_metric = ifp->if_metric; 2333 break; 2334 2335 case SIOCGIFMTU: 2336 ifr = data; 2337 ifr->ifr_mtu = ifp->if_mtu; 2338 break; 2339 2340 case SIOCGIFDLT: 2341 ifr = data; 2342 ifr->ifr_dlt = ifp->if_dlt; 2343 break; 2344 2345 case SIOCGIFCAP: 2346 ifcr = data; 2347 ifcr->ifcr_capabilities = ifp->if_capabilities; 2348 ifcr->ifcr_capenable = ifp->if_capenable; 2349 break; 2350 2351 case SIOCSIFMETRIC: 2352 ifr = data; 2353 ifp->if_metric = ifr->ifr_metric; 2354 break; 2355 2356 case SIOCGIFDATA: 2357 ifdr = data; 2358 ifdr->ifdr_data = ifp->if_data; 2359 break; 2360 2361 case SIOCGIFINDEX: 2362 ifr = data; 2363 ifr->ifr_index = ifp->if_index; 2364 break; 2365 2366 case SIOCZIFDATA: 2367 ifdr = data; 2368 ifdr->ifdr_data = ifp->if_data; 2369 /* 2370 * Assumes that the volatile counters that can be 2371 * zero'ed are at the end of if_data. 2372 */ 2373 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2374 offsetof(struct if_data, ifi_ipackets)); 2375 /* 2376 * The memset() clears to the bottm of if_data. In the area, 2377 * if_lastchange is included. Please be careful if new entry 2378 * will be added into if_data or rewite this. 2379 * 2380 * And also, update if_lastchnage. 2381 */ 2382 getnanotime(&ifp->if_lastchange); 2383 break; 2384 case SIOCSIFMTU: 2385 ifr = data; 2386 if (ifp->if_mtu == ifr->ifr_mtu) 2387 break; 2388 ifp->if_mtu = ifr->ifr_mtu; 2389 /* 2390 * If the link MTU changed, do network layer specific procedure. 2391 */ 2392 #ifdef INET6 2393 if (in6_present) 2394 nd6_setmtu(ifp); 2395 #endif 2396 return ENETRESET; 2397 default: 2398 return ENOTTY; 2399 } 2400 return 0; 2401 } 2402 2403 int 2404 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2405 { 2406 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2407 struct ifaddr *ifa; 2408 const struct sockaddr *any, *sa; 2409 union { 2410 struct sockaddr sa; 2411 struct sockaddr_storage ss; 2412 } u, v; 2413 2414 switch (cmd) { 2415 case SIOCSIFADDRPREF: 2416 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2417 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2418 NULL) != 0) 2419 return EPERM; 2420 case SIOCGIFADDRPREF: 2421 break; 2422 default: 2423 return EOPNOTSUPP; 2424 } 2425 2426 /* sanity checks */ 2427 if (data == NULL || ifp == NULL) { 2428 panic("invalid argument to %s", __func__); 2429 /*NOTREACHED*/ 2430 } 2431 2432 /* address must be specified on ADD and DELETE */ 2433 sa = sstocsa(&ifap->ifap_addr); 2434 if (sa->sa_family != sofamily(so)) 2435 return EINVAL; 2436 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2437 return EINVAL; 2438 2439 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2440 2441 IFADDR_FOREACH(ifa, ifp) { 2442 if (ifa->ifa_addr->sa_family != sa->sa_family) 2443 continue; 2444 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2445 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2446 break; 2447 } 2448 if (ifa == NULL) 2449 return EADDRNOTAVAIL; 2450 2451 switch (cmd) { 2452 case SIOCSIFADDRPREF: 2453 ifa->ifa_preference = ifap->ifap_preference; 2454 return 0; 2455 case SIOCGIFADDRPREF: 2456 /* fill in the if_laddrreq structure */ 2457 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2458 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2459 ifap->ifap_preference = ifa->ifa_preference; 2460 return 0; 2461 default: 2462 return EOPNOTSUPP; 2463 } 2464 } 2465 2466 /* 2467 * Interface ioctls. 2468 */ 2469 static int 2470 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2471 { 2472 struct ifnet *ifp; 2473 struct ifreq *ifr; 2474 int error = 0; 2475 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2476 u_long ocmd = cmd; 2477 #endif 2478 short oif_flags; 2479 #ifdef COMPAT_OIFREQ 2480 struct ifreq ifrb; 2481 struct oifreq *oifr = NULL; 2482 #endif 2483 int r; 2484 struct psref psref; 2485 int bound; 2486 2487 switch (cmd) { 2488 #ifdef COMPAT_OIFREQ 2489 case OSIOCGIFCONF: 2490 case OOSIOCGIFCONF: 2491 return compat_ifconf(cmd, data); 2492 #endif 2493 #ifdef COMPAT_OIFDATA 2494 case OSIOCGIFDATA: 2495 case OSIOCZIFDATA: 2496 return compat_ifdatareq(l, cmd, data); 2497 #endif 2498 case SIOCGIFCONF: 2499 return ifconf(cmd, data); 2500 case SIOCINITIFADDR: 2501 return EPERM; 2502 } 2503 2504 #ifdef COMPAT_OIFREQ 2505 cmd = compat_cvtcmd(cmd); 2506 if (cmd != ocmd) { 2507 oifr = data; 2508 data = ifr = &ifrb; 2509 ifreqo2n(oifr, ifr); 2510 } else 2511 #endif 2512 ifr = data; 2513 2514 switch (cmd) { 2515 case SIOCIFCREATE: 2516 case SIOCIFDESTROY: 2517 bound = curlwp_bind(); 2518 if (l != NULL) { 2519 ifp = if_get(ifr->ifr_name, &psref); 2520 error = kauth_authorize_network(l->l_cred, 2521 KAUTH_NETWORK_INTERFACE, 2522 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2523 (void *)cmd, NULL); 2524 if (ifp != NULL) 2525 if_put(ifp, &psref); 2526 if (error != 0) { 2527 curlwp_bindx(bound); 2528 return error; 2529 } 2530 } 2531 mutex_enter(&if_clone_mtx); 2532 r = (cmd == SIOCIFCREATE) ? 2533 if_clone_create(ifr->ifr_name) : 2534 if_clone_destroy(ifr->ifr_name); 2535 mutex_exit(&if_clone_mtx); 2536 curlwp_bindx(bound); 2537 return r; 2538 2539 case SIOCIFGCLONERS: 2540 { 2541 struct if_clonereq *req = (struct if_clonereq *)data; 2542 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2543 &req->ifcr_total); 2544 } 2545 } 2546 2547 bound = curlwp_bind(); 2548 ifp = if_get(ifr->ifr_name, &psref); 2549 if (ifp == NULL) { 2550 curlwp_bindx(bound); 2551 return ENXIO; 2552 } 2553 2554 switch (cmd) { 2555 case SIOCALIFADDR: 2556 case SIOCDLIFADDR: 2557 case SIOCSIFADDRPREF: 2558 case SIOCSIFFLAGS: 2559 case SIOCSIFCAP: 2560 case SIOCSIFMETRIC: 2561 case SIOCZIFDATA: 2562 case SIOCSIFMTU: 2563 case SIOCSIFPHYADDR: 2564 case SIOCDIFPHYADDR: 2565 #ifdef INET6 2566 case SIOCSIFPHYADDR_IN6: 2567 #endif 2568 case SIOCSLIFPHYADDR: 2569 case SIOCADDMULTI: 2570 case SIOCDELMULTI: 2571 case SIOCSIFMEDIA: 2572 case SIOCSDRVSPEC: 2573 case SIOCG80211: 2574 case SIOCS80211: 2575 case SIOCS80211NWID: 2576 case SIOCS80211NWKEY: 2577 case SIOCS80211POWER: 2578 case SIOCS80211BSSID: 2579 case SIOCS80211CHANNEL: 2580 case SIOCSLINKSTR: 2581 if (l != NULL) { 2582 error = kauth_authorize_network(l->l_cred, 2583 KAUTH_NETWORK_INTERFACE, 2584 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2585 (void *)cmd, NULL); 2586 if (error != 0) 2587 goto out; 2588 } 2589 } 2590 2591 oif_flags = ifp->if_flags; 2592 2593 mutex_enter(ifp->if_ioctl_lock); 2594 2595 error = (*ifp->if_ioctl)(ifp, cmd, data); 2596 if (error != ENOTTY) 2597 ; 2598 else if (so->so_proto == NULL) 2599 error = EOPNOTSUPP; 2600 else { 2601 #ifdef COMPAT_OSOCK 2602 error = compat_ifioctl(so, ocmd, cmd, data, l); 2603 #else 2604 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2605 cmd, data, ifp); 2606 #endif 2607 } 2608 2609 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2610 if ((ifp->if_flags & IFF_UP) != 0) { 2611 int s = splnet(); 2612 if_up(ifp); 2613 splx(s); 2614 } 2615 } 2616 #ifdef COMPAT_OIFREQ 2617 if (cmd != ocmd) 2618 ifreqn2o(oifr, ifr); 2619 #endif 2620 2621 mutex_exit(ifp->if_ioctl_lock); 2622 out: 2623 if_put(ifp, &psref); 2624 curlwp_bindx(bound); 2625 return error; 2626 } 2627 2628 /* 2629 * Return interface configuration 2630 * of system. List may be used 2631 * in later ioctl's (above) to get 2632 * other information. 2633 * 2634 * Each record is a struct ifreq. Before the addition of 2635 * sockaddr_storage, the API rule was that sockaddr flavors that did 2636 * not fit would extend beyond the struct ifreq, with the next struct 2637 * ifreq starting sa_len beyond the struct sockaddr. Because the 2638 * union in struct ifreq includes struct sockaddr_storage, every kind 2639 * of sockaddr must fit. Thus, there are no longer any overlength 2640 * records. 2641 * 2642 * Records are added to the user buffer if they fit, and ifc_len is 2643 * adjusted to the length that was written. Thus, the user is only 2644 * assured of getting the complete list if ifc_len on return is at 2645 * least sizeof(struct ifreq) less than it was on entry. 2646 * 2647 * If the user buffer pointer is NULL, this routine copies no data and 2648 * returns the amount of space that would be needed. 2649 * 2650 * Invariants: 2651 * ifrp points to the next part of the user's buffer to be used. If 2652 * ifrp != NULL, space holds the number of bytes remaining that we may 2653 * write at ifrp. Otherwise, space holds the number of bytes that 2654 * would have been written had there been adequate space. 2655 */ 2656 /*ARGSUSED*/ 2657 static int 2658 ifconf(u_long cmd, void *data) 2659 { 2660 struct ifconf *ifc = (struct ifconf *)data; 2661 struct ifnet *ifp; 2662 struct ifaddr *ifa; 2663 struct ifreq ifr, *ifrp = NULL; 2664 int space = 0, error = 0; 2665 const int sz = (int)sizeof(struct ifreq); 2666 const bool docopy = ifc->ifc_req != NULL; 2667 int s; 2668 int bound; 2669 struct psref psref; 2670 2671 if (docopy) { 2672 space = ifc->ifc_len; 2673 ifrp = ifc->ifc_req; 2674 } 2675 2676 bound = curlwp_bind(); 2677 s = pserialize_read_enter(); 2678 IFNET_READER_FOREACH(ifp) { 2679 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 2680 pserialize_read_exit(s); 2681 2682 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2683 sizeof(ifr.ifr_name)); 2684 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 2685 error = ENAMETOOLONG; 2686 goto release_exit; 2687 } 2688 if (IFADDR_EMPTY(ifp)) { 2689 /* Interface with no addresses - send zero sockaddr. */ 2690 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2691 if (!docopy) { 2692 space += sz; 2693 continue; 2694 } 2695 if (space >= sz) { 2696 error = copyout(&ifr, ifrp, sz); 2697 if (error != 0) 2698 goto release_exit; 2699 ifrp++; 2700 space -= sz; 2701 } 2702 } 2703 2704 IFADDR_FOREACH(ifa, ifp) { 2705 struct sockaddr *sa = ifa->ifa_addr; 2706 /* all sockaddrs must fit in sockaddr_storage */ 2707 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2708 2709 if (!docopy) { 2710 space += sz; 2711 continue; 2712 } 2713 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2714 if (space >= sz) { 2715 error = copyout(&ifr, ifrp, sz); 2716 if (error != 0) 2717 goto release_exit; 2718 ifrp++; space -= sz; 2719 } 2720 } 2721 2722 s = pserialize_read_enter(); 2723 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2724 } 2725 pserialize_read_exit(s); 2726 curlwp_bindx(bound); 2727 2728 if (docopy) { 2729 KASSERT(0 <= space && space <= ifc->ifc_len); 2730 ifc->ifc_len -= space; 2731 } else { 2732 KASSERT(space >= 0); 2733 ifc->ifc_len = space; 2734 } 2735 return (0); 2736 2737 release_exit: 2738 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2739 curlwp_bindx(bound); 2740 return error; 2741 } 2742 2743 int 2744 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2745 { 2746 uint8_t len; 2747 #ifdef COMPAT_OIFREQ 2748 struct ifreq ifrb; 2749 struct oifreq *oifr = NULL; 2750 u_long ocmd = cmd; 2751 cmd = compat_cvtcmd(cmd); 2752 if (cmd != ocmd) { 2753 oifr = (struct oifreq *)(void *)ifr; 2754 ifr = &ifrb; 2755 ifreqo2n(oifr, ifr); 2756 len = sizeof(oifr->ifr_addr); 2757 } else 2758 #endif 2759 len = sizeof(ifr->ifr_ifru.ifru_space); 2760 2761 if (len < sa->sa_len) 2762 return EFBIG; 2763 2764 memset(&ifr->ifr_addr, 0, len); 2765 sockaddr_copy(&ifr->ifr_addr, len, sa); 2766 2767 #ifdef COMPAT_OIFREQ 2768 if (cmd != ocmd) 2769 ifreqn2o(oifr, ifr); 2770 #endif 2771 return 0; 2772 } 2773 2774 /* 2775 * wrapper function for the drivers which doesn't have if_transmit(). 2776 */ 2777 static int 2778 if_transmit(struct ifnet *ifp, struct mbuf *m) 2779 { 2780 int s, error; 2781 2782 s = splnet(); 2783 2784 /* 2785 * If NET_MPSAFE is not defined , IFQ_LOCK() is nop. 2786 * use KERNEL_LOCK instead of ifq_lock. 2787 */ 2788 #ifndef NET_MPSAFE 2789 KERNEL_LOCK(1, NULL); 2790 #endif 2791 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2792 #ifndef NET_MPSAFE 2793 KERNEL_UNLOCK_ONE(NULL); 2794 #endif 2795 if (error != 0) { 2796 /* mbuf is already freed */ 2797 goto out; 2798 } 2799 2800 ifp->if_obytes += m->m_pkthdr.len;; 2801 if (m->m_flags & M_MCAST) 2802 ifp->if_omcasts++; 2803 2804 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2805 if_start_lock(ifp); 2806 out: 2807 splx(s); 2808 2809 return error; 2810 } 2811 2812 int 2813 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 2814 { 2815 int error; 2816 2817 #ifdef ALTQ 2818 KERNEL_LOCK(1, NULL); 2819 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 2820 error = if_transmit(ifp, m); 2821 KERNEL_UNLOCK_ONE(NULL); 2822 } else { 2823 KERNEL_UNLOCK_ONE(NULL); 2824 error = (*ifp->if_transmit)(ifp, m); 2825 } 2826 #else /* !ALTQ */ 2827 error = (*ifp->if_transmit)(ifp, m); 2828 #endif /* !ALTQ */ 2829 2830 return error; 2831 } 2832 2833 /* 2834 * Queue message on interface, and start output if interface 2835 * not yet active. 2836 */ 2837 int 2838 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 2839 { 2840 2841 return if_transmit_lock(ifp, m); 2842 } 2843 2844 /* 2845 * Queue message on interface, possibly using a second fast queue 2846 */ 2847 int 2848 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 2849 { 2850 int error = 0; 2851 2852 if (ifq != NULL 2853 #ifdef ALTQ 2854 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2855 #endif 2856 ) { 2857 if (IF_QFULL(ifq)) { 2858 IF_DROP(&ifp->if_snd); 2859 m_freem(m); 2860 if (error == 0) 2861 error = ENOBUFS; 2862 } else 2863 IF_ENQUEUE(ifq, m); 2864 } else 2865 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2866 if (error != 0) { 2867 ++ifp->if_oerrors; 2868 return error; 2869 } 2870 return 0; 2871 } 2872 2873 int 2874 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2875 { 2876 int rc; 2877 2878 if (ifp->if_initaddr != NULL) 2879 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2880 else if (src || 2881 /* FIXME: may not hold if_ioctl_lock */ 2882 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2883 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2884 2885 return rc; 2886 } 2887 2888 int 2889 if_do_dad(struct ifnet *ifp) 2890 { 2891 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2892 return 0; 2893 2894 switch (ifp->if_type) { 2895 case IFT_FAITH: 2896 /* 2897 * These interfaces do not have the IFF_LOOPBACK flag, 2898 * but loop packets back. We do not have to do DAD on such 2899 * interfaces. We should even omit it, because loop-backed 2900 * responses would confuse the DAD procedure. 2901 */ 2902 return 0; 2903 default: 2904 /* 2905 * Our DAD routine requires the interface up and running. 2906 * However, some interfaces can be up before the RUNNING 2907 * status. Additionaly, users may try to assign addresses 2908 * before the interface becomes up (or running). 2909 * We simply skip DAD in such a case as a work around. 2910 * XXX: we should rather mark "tentative" on such addresses, 2911 * and do DAD after the interface becomes ready. 2912 */ 2913 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 2914 (IFF_UP|IFF_RUNNING)) 2915 return 0; 2916 2917 return 1; 2918 } 2919 } 2920 2921 int 2922 if_flags_set(ifnet_t *ifp, const short flags) 2923 { 2924 int rc; 2925 2926 if (ifp->if_setflags != NULL) 2927 rc = (*ifp->if_setflags)(ifp, flags); 2928 else { 2929 short cantflags, chgdflags; 2930 struct ifreq ifr; 2931 2932 chgdflags = ifp->if_flags ^ flags; 2933 cantflags = chgdflags & IFF_CANTCHANGE; 2934 2935 if (cantflags != 0) 2936 ifp->if_flags ^= cantflags; 2937 2938 /* Traditionally, we do not call if_ioctl after 2939 * setting/clearing only IFF_PROMISC if the interface 2940 * isn't IFF_UP. Uphold that tradition. 2941 */ 2942 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2943 return 0; 2944 2945 memset(&ifr, 0, sizeof(ifr)); 2946 2947 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2948 /* FIXME: may not hold if_ioctl_lock */ 2949 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2950 2951 if (rc != 0 && cantflags != 0) 2952 ifp->if_flags ^= cantflags; 2953 } 2954 2955 return rc; 2956 } 2957 2958 int 2959 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2960 { 2961 int rc; 2962 struct ifreq ifr; 2963 2964 if (ifp->if_mcastop != NULL) 2965 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2966 else { 2967 ifreq_setaddr(cmd, &ifr, sa); 2968 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2969 } 2970 2971 return rc; 2972 } 2973 2974 static void 2975 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2976 struct ifaltq *ifq) 2977 { 2978 const struct sysctlnode *cnode, *rnode; 2979 2980 if (sysctl_createv(clog, 0, NULL, &rnode, 2981 CTLFLAG_PERMANENT, 2982 CTLTYPE_NODE, "interfaces", 2983 SYSCTL_DESCR("Per-interface controls"), 2984 NULL, 0, NULL, 0, 2985 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 2986 goto bad; 2987 2988 if (sysctl_createv(clog, 0, &rnode, &rnode, 2989 CTLFLAG_PERMANENT, 2990 CTLTYPE_NODE, ifname, 2991 SYSCTL_DESCR("Interface controls"), 2992 NULL, 0, NULL, 0, 2993 CTL_CREATE, CTL_EOL) != 0) 2994 goto bad; 2995 2996 if (sysctl_createv(clog, 0, &rnode, &rnode, 2997 CTLFLAG_PERMANENT, 2998 CTLTYPE_NODE, "sndq", 2999 SYSCTL_DESCR("Interface output queue controls"), 3000 NULL, 0, NULL, 0, 3001 CTL_CREATE, CTL_EOL) != 0) 3002 goto bad; 3003 3004 if (sysctl_createv(clog, 0, &rnode, &cnode, 3005 CTLFLAG_PERMANENT, 3006 CTLTYPE_INT, "len", 3007 SYSCTL_DESCR("Current output queue length"), 3008 NULL, 0, &ifq->ifq_len, 0, 3009 CTL_CREATE, CTL_EOL) != 0) 3010 goto bad; 3011 3012 if (sysctl_createv(clog, 0, &rnode, &cnode, 3013 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3014 CTLTYPE_INT, "maxlen", 3015 SYSCTL_DESCR("Maximum allowed output queue length"), 3016 NULL, 0, &ifq->ifq_maxlen, 0, 3017 CTL_CREATE, CTL_EOL) != 0) 3018 goto bad; 3019 3020 if (sysctl_createv(clog, 0, &rnode, &cnode, 3021 CTLFLAG_PERMANENT, 3022 CTLTYPE_INT, "drops", 3023 SYSCTL_DESCR("Packets dropped due to full output queue"), 3024 NULL, 0, &ifq->ifq_drops, 0, 3025 CTL_CREATE, CTL_EOL) != 0) 3026 goto bad; 3027 3028 return; 3029 bad: 3030 printf("%s: could not attach sysctl nodes\n", ifname); 3031 return; 3032 } 3033 3034 #if defined(INET) || defined(INET6) 3035 3036 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3037 static int \ 3038 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3039 { \ 3040 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3041 } 3042 3043 #if defined(INET) 3044 static int 3045 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3046 { 3047 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3048 } 3049 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3050 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3051 #endif 3052 3053 #if defined(INET6) 3054 static int 3055 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3056 { 3057 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3058 } 3059 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3060 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3061 #endif 3062 3063 static void 3064 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3065 { 3066 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3067 const char *pfname = NULL, *ipname = NULL; 3068 int ipn = 0, qid = 0; 3069 3070 switch (pf) { 3071 #if defined(INET) 3072 case PF_INET: 3073 len_func = sysctl_net_ip_pktq_items; 3074 maxlen_func = sysctl_net_ip_pktq_maxlen; 3075 drops_func = sysctl_net_ip_pktq_drops; 3076 pfname = "inet", ipn = IPPROTO_IP; 3077 ipname = "ip", qid = IPCTL_IFQ; 3078 break; 3079 #endif 3080 #if defined(INET6) 3081 case PF_INET6: 3082 len_func = sysctl_net_ip6_pktq_items; 3083 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3084 drops_func = sysctl_net_ip6_pktq_drops; 3085 pfname = "inet6", ipn = IPPROTO_IPV6; 3086 ipname = "ip6", qid = IPV6CTL_IFQ; 3087 break; 3088 #endif 3089 default: 3090 KASSERT(false); 3091 } 3092 3093 sysctl_createv(clog, 0, NULL, NULL, 3094 CTLFLAG_PERMANENT, 3095 CTLTYPE_NODE, pfname, NULL, 3096 NULL, 0, NULL, 0, 3097 CTL_NET, pf, CTL_EOL); 3098 sysctl_createv(clog, 0, NULL, NULL, 3099 CTLFLAG_PERMANENT, 3100 CTLTYPE_NODE, ipname, NULL, 3101 NULL, 0, NULL, 0, 3102 CTL_NET, pf, ipn, CTL_EOL); 3103 sysctl_createv(clog, 0, NULL, NULL, 3104 CTLFLAG_PERMANENT, 3105 CTLTYPE_NODE, "ifq", 3106 SYSCTL_DESCR("Protocol input queue controls"), 3107 NULL, 0, NULL, 0, 3108 CTL_NET, pf, ipn, qid, CTL_EOL); 3109 3110 sysctl_createv(clog, 0, NULL, NULL, 3111 CTLFLAG_PERMANENT, 3112 CTLTYPE_INT, "len", 3113 SYSCTL_DESCR("Current input queue length"), 3114 len_func, 0, NULL, 0, 3115 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3116 sysctl_createv(clog, 0, NULL, NULL, 3117 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3118 CTLTYPE_INT, "maxlen", 3119 SYSCTL_DESCR("Maximum allowed input queue length"), 3120 maxlen_func, 0, NULL, 0, 3121 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3122 sysctl_createv(clog, 0, NULL, NULL, 3123 CTLFLAG_PERMANENT, 3124 CTLTYPE_INT, "drops", 3125 SYSCTL_DESCR("Packets dropped due to full input queue"), 3126 drops_func, 0, NULL, 0, 3127 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3128 } 3129 #endif /* INET || INET6 */ 3130 3131 static int 3132 if_sdl_sysctl(SYSCTLFN_ARGS) 3133 { 3134 struct ifnet *ifp; 3135 const struct sockaddr_dl *sdl; 3136 struct psref psref; 3137 int error = 0; 3138 int bound; 3139 3140 if (namelen != 1) 3141 return EINVAL; 3142 3143 bound = curlwp_bind(); 3144 ifp = if_get_byindex(name[0], &psref); 3145 if (ifp == NULL) { 3146 error = ENODEV; 3147 goto out0; 3148 } 3149 3150 sdl = ifp->if_sadl; 3151 if (sdl == NULL) { 3152 *oldlenp = 0; 3153 goto out1; 3154 } 3155 3156 if (oldp == NULL) { 3157 *oldlenp = sdl->sdl_alen; 3158 goto out1; 3159 } 3160 3161 if (*oldlenp >= sdl->sdl_alen) 3162 *oldlenp = sdl->sdl_alen; 3163 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3164 out1: 3165 if_put(ifp, &psref); 3166 out0: 3167 curlwp_bindx(bound); 3168 return error; 3169 } 3170 3171 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 3172 { 3173 const struct sysctlnode *rnode = NULL; 3174 3175 sysctl_createv(clog, 0, NULL, &rnode, 3176 CTLFLAG_PERMANENT, 3177 CTLTYPE_NODE, "sdl", 3178 SYSCTL_DESCR("Get active link-layer address"), 3179 if_sdl_sysctl, 0, NULL, 0, 3180 CTL_NET, CTL_CREATE, CTL_EOL); 3181 } 3182
Indexes created Wed Sep 24 05:09:52 GMT 2025