if.c revision 1.355
1 /* $NetBSD: if.c,v 1.355 2016/07/22 07:09:40 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 1.355 2016/07/22 07:09:40 knakahara Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 #include "opt_ipsec.h" 98 99 #include "opt_atalk.h" 100 #include "opt_natm.h" 101 #include "opt_wlan.h" 102 #include "opt_net_mpsafe.h" 103 #endif 104 105 #include <sys/param.h> 106 #include <sys/mbuf.h> 107 #include <sys/systm.h> 108 #include <sys/callout.h> 109 #include <sys/proc.h> 110 #include <sys/socket.h> 111 #include <sys/socketvar.h> 112 #include <sys/domain.h> 113 #include <sys/protosw.h> 114 #include <sys/kernel.h> 115 #include <sys/ioctl.h> 116 #include <sys/sysctl.h> 117 #include <sys/syslog.h> 118 #include <sys/kauth.h> 119 #include <sys/kmem.h> 120 #include <sys/xcall.h> 121 #include <sys/cpu.h> 122 #include <sys/intr.h> 123 124 #include <net/if.h> 125 #include <net/if_dl.h> 126 #include <net/if_ether.h> 127 #include <net/if_media.h> 128 #include <net80211/ieee80211.h> 129 #include <net80211/ieee80211_ioctl.h> 130 #include <net/if_types.h> 131 #include <net/route.h> 132 #include <net/netisr.h> 133 #include <sys/module.h> 134 #ifdef NETATALK 135 #include <netatalk/at_extern.h> 136 #include <netatalk/at.h> 137 #endif 138 #include <net/pfil.h> 139 #include <netinet/in.h> 140 #include <netinet/in_var.h> 141 #ifndef IPSEC 142 #include <netinet/ip_encap.h> 143 #endif 144 145 #ifdef INET6 146 #include <netinet6/in6_var.h> 147 #include <netinet6/nd6.h> 148 #endif 149 150 #include "ether.h" 151 #include "fddi.h" 152 #include "token.h" 153 154 #include "carp.h" 155 #if NCARP > 0 156 #include <netinet/ip_carp.h> 157 #endif 158 159 #include <compat/sys/sockio.h> 160 #include <compat/sys/socket.h> 161 162 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 163 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 164 165 /* 166 * Global list of interfaces. 167 */ 168 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 169 struct ifnet_head ifnet_list; 170 171 struct pslist_head ifnet_pslist; 172 static ifnet_t ** ifindex2ifnet = NULL; 173 static u_int if_index = 1; 174 static size_t if_indexlim = 0; 175 static uint64_t index_gen; 176 /* Mutex to protect the above objects. */ 177 kmutex_t ifnet_mtx __cacheline_aligned; 178 struct psref_class *ifnet_psref_class __read_mostly; 179 static pserialize_t ifnet_psz; 180 181 static kmutex_t if_clone_mtx; 182 183 struct ifnet *lo0ifp; 184 int ifqmaxlen = IFQ_MAXLEN; 185 186 static int if_rt_walktree(struct rtentry *, void *); 187 188 static struct if_clone *if_clone_lookup(const char *, int *); 189 190 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 191 static int if_cloners_count; 192 193 /* Packet filtering hook for interfaces. */ 194 pfil_head_t * if_pfil; 195 196 static kauth_listener_t if_listener; 197 198 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 199 static void if_detach_queues(struct ifnet *, struct ifqueue *); 200 static void sysctl_sndq_setup(struct sysctllog **, const char *, 201 struct ifaltq *); 202 static void if_slowtimo(void *); 203 static void if_free_sadl(struct ifnet *); 204 static void if_attachdomain1(struct ifnet *); 205 static int ifconf(u_long, void *); 206 static int if_transmit(struct ifnet *, struct mbuf *); 207 static int if_clone_create(const char *); 208 static int if_clone_destroy(const char *); 209 static void if_link_state_change_si(void *); 210 211 struct if_percpuq { 212 struct ifnet *ipq_ifp; 213 void *ipq_si; 214 struct percpu *ipq_ifqs; /* struct ifqueue */ 215 }; 216 217 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 218 219 static void if_percpuq_drops(void *, void *, struct cpu_info *); 220 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 221 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 222 struct if_percpuq *); 223 224 #if defined(INET) || defined(INET6) 225 static void sysctl_net_pktq_setup(struct sysctllog **, int); 226 #endif 227 228 static int 229 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 230 void *arg0, void *arg1, void *arg2, void *arg3) 231 { 232 int result; 233 enum kauth_network_req req; 234 235 result = KAUTH_RESULT_DEFER; 236 req = (enum kauth_network_req)arg1; 237 238 if (action != KAUTH_NETWORK_INTERFACE) 239 return result; 240 241 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 242 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 243 result = KAUTH_RESULT_ALLOW; 244 245 return result; 246 } 247 248 /* 249 * Network interface utility routines. 250 * 251 * Routines with ifa_ifwith* names take sockaddr *'s as 252 * parameters. 253 */ 254 void 255 ifinit(void) 256 { 257 #if defined(INET) 258 sysctl_net_pktq_setup(NULL, PF_INET); 259 #endif 260 #ifdef INET6 261 if (in6_present) 262 sysctl_net_pktq_setup(NULL, PF_INET6); 263 #endif 264 265 #if (defined(INET) || defined(INET6)) && !defined(IPSEC) 266 encapinit(); 267 #endif 268 269 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 270 if_listener_cb, NULL); 271 272 /* interfaces are available, inform socket code */ 273 ifioctl = doifioctl; 274 } 275 276 /* 277 * XXX Initialization before configure(). 278 * XXX hack to get pfil_add_hook working in autoconf. 279 */ 280 void 281 ifinit1(void) 282 { 283 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 284 TAILQ_INIT(&ifnet_list); 285 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 286 ifnet_psz = pserialize_create(); 287 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 288 PSLIST_INIT(&ifnet_pslist); 289 if_indexlim = 8; 290 291 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 292 KASSERT(if_pfil != NULL); 293 294 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 295 etherinit(); 296 #endif 297 } 298 299 ifnet_t * 300 if_alloc(u_char type) 301 { 302 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 303 } 304 305 void 306 if_free(ifnet_t *ifp) 307 { 308 kmem_free(ifp, sizeof(ifnet_t)); 309 } 310 311 void 312 if_initname(struct ifnet *ifp, const char *name, int unit) 313 { 314 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 315 "%s%d", name, unit); 316 } 317 318 /* 319 * Null routines used while an interface is going away. These routines 320 * just return an error. 321 */ 322 323 int 324 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 325 const struct sockaddr *so, const struct rtentry *rt) 326 { 327 328 return ENXIO; 329 } 330 331 void 332 if_nullinput(struct ifnet *ifp, struct mbuf *m) 333 { 334 335 /* Nothing. */ 336 } 337 338 void 339 if_nullstart(struct ifnet *ifp) 340 { 341 342 /* Nothing. */ 343 } 344 345 int 346 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 347 { 348 349 return ENXIO; 350 } 351 352 int 353 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 354 { 355 356 return ENXIO; 357 } 358 359 int 360 if_nullinit(struct ifnet *ifp) 361 { 362 363 return ENXIO; 364 } 365 366 void 367 if_nullstop(struct ifnet *ifp, int disable) 368 { 369 370 /* Nothing. */ 371 } 372 373 void 374 if_nullslowtimo(struct ifnet *ifp) 375 { 376 377 /* Nothing. */ 378 } 379 380 void 381 if_nulldrain(struct ifnet *ifp) 382 { 383 384 /* Nothing. */ 385 } 386 387 void 388 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 389 { 390 struct ifaddr *ifa; 391 struct sockaddr_dl *sdl; 392 393 ifp->if_addrlen = addrlen; 394 if_alloc_sadl(ifp); 395 ifa = ifp->if_dl; 396 sdl = satosdl(ifa->ifa_addr); 397 398 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 399 if (factory) { 400 ifp->if_hwdl = ifp->if_dl; 401 ifaref(ifp->if_hwdl); 402 } 403 /* TBD routing socket */ 404 } 405 406 struct ifaddr * 407 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 408 { 409 unsigned socksize, ifasize; 410 int addrlen, namelen; 411 struct sockaddr_dl *mask, *sdl; 412 struct ifaddr *ifa; 413 414 namelen = strlen(ifp->if_xname); 415 addrlen = ifp->if_addrlen; 416 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 417 ifasize = sizeof(*ifa) + 2 * socksize; 418 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 419 420 sdl = (struct sockaddr_dl *)(ifa + 1); 421 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 422 423 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 424 ifp->if_xname, namelen, NULL, addrlen); 425 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 426 memset(&mask->sdl_data[0], 0xff, namelen); 427 ifa->ifa_rtrequest = link_rtrequest; 428 ifa->ifa_addr = (struct sockaddr *)sdl; 429 ifa->ifa_netmask = (struct sockaddr *)mask; 430 431 *sdlp = sdl; 432 433 return ifa; 434 } 435 436 static void 437 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 438 { 439 const struct sockaddr_dl *sdl; 440 441 ifp->if_dl = ifa; 442 ifaref(ifa); 443 sdl = satosdl(ifa->ifa_addr); 444 ifp->if_sadl = sdl; 445 } 446 447 /* 448 * Allocate the link level name for the specified interface. This 449 * is an attachment helper. It must be called after ifp->if_addrlen 450 * is initialized, which may not be the case when if_attach() is 451 * called. 452 */ 453 void 454 if_alloc_sadl(struct ifnet *ifp) 455 { 456 struct ifaddr *ifa; 457 const struct sockaddr_dl *sdl; 458 459 /* 460 * If the interface already has a link name, release it 461 * now. This is useful for interfaces that can change 462 * link types, and thus switch link names often. 463 */ 464 if (ifp->if_sadl != NULL) 465 if_free_sadl(ifp); 466 467 ifa = if_dl_create(ifp, &sdl); 468 469 ifa_insert(ifp, ifa); 470 if_sadl_setrefs(ifp, ifa); 471 } 472 473 static void 474 if_deactivate_sadl(struct ifnet *ifp) 475 { 476 struct ifaddr *ifa; 477 478 KASSERT(ifp->if_dl != NULL); 479 480 ifa = ifp->if_dl; 481 482 ifp->if_sadl = NULL; 483 484 ifp->if_dl = NULL; 485 ifafree(ifa); 486 } 487 488 void 489 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 490 const struct sockaddr_dl *sdl) 491 { 492 int s; 493 494 s = splnet(); 495 496 if_deactivate_sadl(ifp); 497 498 if_sadl_setrefs(ifp, ifa); 499 IFADDR_READER_FOREACH(ifa, ifp) 500 rtinit(ifa, RTM_LLINFO_UPD, 0); 501 502 splx(s); 503 } 504 505 /* 506 * Free the link level name for the specified interface. This is 507 * a detach helper. This is called from if_detach(). 508 */ 509 static void 510 if_free_sadl(struct ifnet *ifp) 511 { 512 struct ifaddr *ifa; 513 int s; 514 515 ifa = ifp->if_dl; 516 if (ifa == NULL) { 517 KASSERT(ifp->if_sadl == NULL); 518 return; 519 } 520 521 KASSERT(ifp->if_sadl != NULL); 522 523 s = splnet(); 524 rtinit(ifa, RTM_DELETE, 0); 525 ifa_remove(ifp, ifa); 526 if_deactivate_sadl(ifp); 527 if (ifp->if_hwdl == ifa) { 528 ifafree(ifa); 529 ifp->if_hwdl = NULL; 530 } 531 splx(s); 532 } 533 534 static void 535 if_getindex(ifnet_t *ifp) 536 { 537 bool hitlimit = false; 538 539 ifp->if_index_gen = index_gen++; 540 541 ifp->if_index = if_index; 542 if (ifindex2ifnet == NULL) { 543 if_index++; 544 goto skip; 545 } 546 while (if_byindex(ifp->if_index)) { 547 /* 548 * If we hit USHRT_MAX, we skip back to 0 since 549 * there are a number of places where the value 550 * of if_index or if_index itself is compared 551 * to or stored in an unsigned short. By 552 * jumping back, we won't botch those assignments 553 * or comparisons. 554 */ 555 if (++if_index == 0) { 556 if_index = 1; 557 } else if (if_index == USHRT_MAX) { 558 /* 559 * However, if we have to jump back to 560 * zero *twice* without finding an empty 561 * slot in ifindex2ifnet[], then there 562 * there are too many (>65535) interfaces. 563 */ 564 if (hitlimit) { 565 panic("too many interfaces"); 566 } 567 hitlimit = true; 568 if_index = 1; 569 } 570 ifp->if_index = if_index; 571 } 572 skip: 573 /* 574 * ifindex2ifnet is indexed by if_index. Since if_index will 575 * grow dynamically, it should grow too. 576 */ 577 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 578 size_t m, n, oldlim; 579 void *q; 580 581 oldlim = if_indexlim; 582 while (ifp->if_index >= if_indexlim) 583 if_indexlim <<= 1; 584 585 /* grow ifindex2ifnet */ 586 m = oldlim * sizeof(struct ifnet *); 587 n = if_indexlim * sizeof(struct ifnet *); 588 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 589 if (ifindex2ifnet != NULL) { 590 memcpy(q, ifindex2ifnet, m); 591 free(ifindex2ifnet, M_IFADDR); 592 } 593 ifindex2ifnet = (struct ifnet **)q; 594 } 595 ifindex2ifnet[ifp->if_index] = ifp; 596 } 597 598 /* 599 * Initialize an interface and assign an index for it. 600 * 601 * It must be called prior to a device specific attach routine 602 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 603 * and be followed by if_register: 604 * 605 * if_initialize(ifp); 606 * ether_ifattach(ifp, enaddr); 607 * if_register(ifp); 608 */ 609 void 610 if_initialize(ifnet_t *ifp) 611 { 612 KASSERT(if_indexlim > 0); 613 TAILQ_INIT(&ifp->if_addrlist); 614 615 /* 616 * Link level name is allocated later by a separate call to 617 * if_alloc_sadl(). 618 */ 619 620 if (ifp->if_snd.ifq_maxlen == 0) 621 ifp->if_snd.ifq_maxlen = ifqmaxlen; 622 623 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 624 625 ifp->if_link_state = LINK_STATE_UNKNOWN; 626 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 627 628 ifp->if_capenable = 0; 629 ifp->if_csum_flags_tx = 0; 630 ifp->if_csum_flags_rx = 0; 631 632 #ifdef ALTQ 633 ifp->if_snd.altq_type = 0; 634 ifp->if_snd.altq_disc = NULL; 635 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 636 ifp->if_snd.altq_tbr = NULL; 637 ifp->if_snd.altq_ifp = ifp; 638 #endif 639 640 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 641 642 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 643 (void)pfil_run_hooks(if_pfil, 644 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 645 646 IF_AFDATA_LOCK_INIT(ifp); 647 648 if (if_is_link_state_changeable(ifp)) { 649 ifp->if_link_si = softint_establish(SOFTINT_NET, 650 if_link_state_change_si, ifp); 651 if (ifp->if_link_si == NULL) 652 panic("%s: softint_establish() failed", __func__); 653 } 654 655 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 656 PSLIST_INIT(&ifp->if_addr_pslist); 657 psref_target_init(&ifp->if_psref, ifnet_psref_class); 658 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 659 660 IFNET_LOCK(); 661 if_getindex(ifp); 662 IFNET_UNLOCK(); 663 } 664 665 /* 666 * Register an interface to the list of "active" interfaces. 667 */ 668 void 669 if_register(ifnet_t *ifp) 670 { 671 /* 672 * If the driver has not supplied its own if_ioctl, then 673 * supply the default. 674 */ 675 if (ifp->if_ioctl == NULL) 676 ifp->if_ioctl = ifioctl_common; 677 678 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 679 680 if (!STAILQ_EMPTY(&domains)) 681 if_attachdomain1(ifp); 682 683 /* Announce the interface. */ 684 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 685 686 if (ifp->if_slowtimo != NULL) { 687 ifp->if_slowtimo_ch = 688 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 689 callout_init(ifp->if_slowtimo_ch, 0); 690 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 691 if_slowtimo(ifp); 692 } 693 694 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 695 ifp->if_transmit = if_transmit; 696 697 IFNET_LOCK(); 698 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 699 IFNET_WRITER_INSERT_TAIL(ifp); 700 IFNET_UNLOCK(); 701 } 702 703 /* 704 * The if_percpuq framework 705 * 706 * It allows network device drivers to execute the network stack 707 * in softint (so called softint-based if_input). It utilizes 708 * softint and percpu ifqueue. It doesn't distribute any packets 709 * between CPUs, unlike pktqueue(9). 710 * 711 * Currently we support two options for device drivers to apply the framework: 712 * - Use it implicitly with less changes 713 * - If you use if_attach in driver's _attach function and if_input in 714 * driver's Rx interrupt handler, a packet is queued and a softint handles 715 * the packet implicitly 716 * - Use it explicitly in each driver (recommended) 717 * - You can use if_percpuq_* directly in your driver 718 * - In this case, you need to allocate struct if_percpuq in driver's softc 719 * - See wm(4) as a reference implementation 720 */ 721 722 static void 723 if_percpuq_softint(void *arg) 724 { 725 struct if_percpuq *ipq = arg; 726 struct ifnet *ifp = ipq->ipq_ifp; 727 struct mbuf *m; 728 729 while ((m = if_percpuq_dequeue(ipq)) != NULL) 730 ifp->_if_input(ifp, m); 731 } 732 733 static void 734 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 735 { 736 struct ifqueue *const ifq = p; 737 738 memset(ifq, 0, sizeof(*ifq)); 739 ifq->ifq_maxlen = IFQ_MAXLEN; 740 } 741 742 struct if_percpuq * 743 if_percpuq_create(struct ifnet *ifp) 744 { 745 struct if_percpuq *ipq; 746 747 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 748 if (ipq == NULL) 749 panic("kmem_zalloc failed"); 750 751 ipq->ipq_ifp = ifp; 752 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 753 if_percpuq_softint, ipq); 754 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 755 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 756 757 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 758 759 return ipq; 760 } 761 762 static struct mbuf * 763 if_percpuq_dequeue(struct if_percpuq *ipq) 764 { 765 struct mbuf *m; 766 struct ifqueue *ifq; 767 int s; 768 769 s = splnet(); 770 ifq = percpu_getref(ipq->ipq_ifqs); 771 IF_DEQUEUE(ifq, m); 772 percpu_putref(ipq->ipq_ifqs); 773 splx(s); 774 775 return m; 776 } 777 778 static void 779 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 780 { 781 struct ifqueue *const ifq = p; 782 783 IF_PURGE(ifq); 784 } 785 786 void 787 if_percpuq_destroy(struct if_percpuq *ipq) 788 { 789 790 /* if_detach may already destroy it */ 791 if (ipq == NULL) 792 return; 793 794 softint_disestablish(ipq->ipq_si); 795 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 796 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 797 } 798 799 void 800 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 801 { 802 struct ifqueue *ifq; 803 int s; 804 805 KASSERT(ipq != NULL); 806 807 s = splnet(); 808 ifq = percpu_getref(ipq->ipq_ifqs); 809 if (IF_QFULL(ifq)) { 810 IF_DROP(ifq); 811 percpu_putref(ipq->ipq_ifqs); 812 m_freem(m); 813 goto out; 814 } 815 IF_ENQUEUE(ifq, m); 816 percpu_putref(ipq->ipq_ifqs); 817 818 softint_schedule(ipq->ipq_si); 819 out: 820 splx(s); 821 } 822 823 static void 824 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 825 { 826 struct ifqueue *const ifq = p; 827 int *sum = arg; 828 829 *sum += ifq->ifq_drops; 830 } 831 832 static int 833 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 834 { 835 struct sysctlnode node; 836 struct if_percpuq *ipq; 837 int sum = 0; 838 int error; 839 840 node = *rnode; 841 ipq = node.sysctl_data; 842 843 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 844 845 node.sysctl_data = ∑ 846 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 847 if (error != 0 || newp == NULL) 848 return error; 849 850 return 0; 851 } 852 853 static void 854 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 855 struct if_percpuq *ipq) 856 { 857 const struct sysctlnode *cnode, *rnode; 858 859 if (sysctl_createv(clog, 0, NULL, &rnode, 860 CTLFLAG_PERMANENT, 861 CTLTYPE_NODE, "interfaces", 862 SYSCTL_DESCR("Per-interface controls"), 863 NULL, 0, NULL, 0, 864 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 865 goto bad; 866 867 if (sysctl_createv(clog, 0, &rnode, &rnode, 868 CTLFLAG_PERMANENT, 869 CTLTYPE_NODE, ifname, 870 SYSCTL_DESCR("Interface controls"), 871 NULL, 0, NULL, 0, 872 CTL_CREATE, CTL_EOL) != 0) 873 goto bad; 874 875 if (sysctl_createv(clog, 0, &rnode, &rnode, 876 CTLFLAG_PERMANENT, 877 CTLTYPE_NODE, "rcvq", 878 SYSCTL_DESCR("Interface input queue controls"), 879 NULL, 0, NULL, 0, 880 CTL_CREATE, CTL_EOL) != 0) 881 goto bad; 882 883 #ifdef NOTYET 884 /* XXX Should show each per-CPU queue length? */ 885 if (sysctl_createv(clog, 0, &rnode, &rnode, 886 CTLFLAG_PERMANENT, 887 CTLTYPE_INT, "len", 888 SYSCTL_DESCR("Current input queue length"), 889 sysctl_percpuq_len, 0, NULL, 0, 890 CTL_CREATE, CTL_EOL) != 0) 891 goto bad; 892 893 if (sysctl_createv(clog, 0, &rnode, &cnode, 894 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 895 CTLTYPE_INT, "maxlen", 896 SYSCTL_DESCR("Maximum allowed input queue length"), 897 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 898 CTL_CREATE, CTL_EOL) != 0) 899 goto bad; 900 #endif 901 902 if (sysctl_createv(clog, 0, &rnode, &cnode, 903 CTLFLAG_PERMANENT, 904 CTLTYPE_INT, "drops", 905 SYSCTL_DESCR("Total packets dropped due to full input queue"), 906 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 907 CTL_CREATE, CTL_EOL) != 0) 908 goto bad; 909 910 return; 911 bad: 912 printf("%s: could not attach sysctl nodes\n", ifname); 913 return; 914 } 915 916 917 /* 918 * The common interface input routine that is called by device drivers, 919 * which should be used only when the driver's rx handler already runs 920 * in softint. 921 */ 922 void 923 if_input(struct ifnet *ifp, struct mbuf *m) 924 { 925 926 KASSERT(ifp->if_percpuq == NULL); 927 KASSERT(!cpu_intr_p()); 928 929 ifp->_if_input(ifp, m); 930 } 931 932 /* 933 * DEPRECATED. Use if_initialize and if_register instead. 934 * See the above comment of if_initialize. 935 * 936 * Note that it implicitly enables if_percpuq to make drivers easy to 937 * migrate softint-based if_input without much changes. If you don't 938 * want to enable it, use if_initialize instead. 939 */ 940 void 941 if_attach(ifnet_t *ifp) 942 { 943 944 if_initialize(ifp); 945 ifp->if_percpuq = if_percpuq_create(ifp); 946 if_register(ifp); 947 } 948 949 void 950 if_attachdomain(void) 951 { 952 struct ifnet *ifp; 953 int s; 954 int bound = curlwp_bind(); 955 956 s = pserialize_read_enter(); 957 IFNET_READER_FOREACH(ifp) { 958 struct psref psref; 959 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 960 pserialize_read_exit(s); 961 if_attachdomain1(ifp); 962 s = pserialize_read_enter(); 963 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 964 } 965 pserialize_read_exit(s); 966 curlwp_bindx(bound); 967 } 968 969 static void 970 if_attachdomain1(struct ifnet *ifp) 971 { 972 struct domain *dp; 973 int s; 974 975 s = splnet(); 976 977 /* address family dependent data region */ 978 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 979 DOMAIN_FOREACH(dp) { 980 if (dp->dom_ifattach != NULL) 981 ifp->if_afdata[dp->dom_family] = 982 (*dp->dom_ifattach)(ifp); 983 } 984 985 splx(s); 986 } 987 988 /* 989 * Deactivate an interface. This points all of the procedure 990 * handles at error stubs. May be called from interrupt context. 991 */ 992 void 993 if_deactivate(struct ifnet *ifp) 994 { 995 int s; 996 997 s = splnet(); 998 999 ifp->if_output = if_nulloutput; 1000 ifp->_if_input = if_nullinput; 1001 ifp->if_start = if_nullstart; 1002 ifp->if_transmit = if_nulltransmit; 1003 ifp->if_ioctl = if_nullioctl; 1004 ifp->if_init = if_nullinit; 1005 ifp->if_stop = if_nullstop; 1006 ifp->if_slowtimo = if_nullslowtimo; 1007 ifp->if_drain = if_nulldrain; 1008 1009 /* No more packets may be enqueued. */ 1010 ifp->if_snd.ifq_maxlen = 0; 1011 1012 splx(s); 1013 } 1014 1015 bool 1016 if_is_deactivated(struct ifnet *ifp) 1017 { 1018 1019 return ifp->if_output == if_nulloutput; 1020 } 1021 1022 void 1023 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1024 { 1025 struct ifaddr *ifa, *nifa; 1026 1027 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) { 1028 nifa = IFADDR_READER_NEXT(ifa); 1029 if (ifa->ifa_addr->sa_family != family) 1030 continue; 1031 (*purgeaddr)(ifa); 1032 } 1033 } 1034 1035 #ifdef IFAREF_DEBUG 1036 static struct ifaddr **ifa_list; 1037 static int ifa_list_size; 1038 1039 /* Depends on only one if_attach runs at once */ 1040 static void 1041 if_build_ifa_list(struct ifnet *ifp) 1042 { 1043 struct ifaddr *ifa; 1044 int i; 1045 1046 KASSERT(ifa_list == NULL); 1047 KASSERT(ifa_list_size == 0); 1048 1049 IFADDR_READER_FOREACH(ifa, ifp) 1050 ifa_list_size++; 1051 1052 ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP); 1053 if (ifa_list == NULL) 1054 return; 1055 1056 i = 0; 1057 IFADDR_READER_FOREACH(ifa, ifp) { 1058 ifa_list[i++] = ifa; 1059 ifaref(ifa); 1060 } 1061 } 1062 1063 static void 1064 if_check_and_free_ifa_list(struct ifnet *ifp) 1065 { 1066 int i; 1067 struct ifaddr *ifa; 1068 1069 if (ifa_list == NULL) 1070 return; 1071 1072 for (i = 0; i < ifa_list_size; i++) { 1073 char buf[64]; 1074 1075 ifa = ifa_list[i]; 1076 sockaddr_format(ifa->ifa_addr, buf, sizeof(buf)); 1077 if (ifa->ifa_refcnt > 1) { 1078 log(LOG_WARNING, 1079 "ifa(%s) still referenced (refcnt=%d)\n", 1080 buf, ifa->ifa_refcnt - 1); 1081 } else 1082 log(LOG_DEBUG, 1083 "ifa(%s) not referenced (refcnt=%d)\n", 1084 buf, ifa->ifa_refcnt - 1); 1085 ifafree(ifa); 1086 } 1087 1088 kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size); 1089 ifa_list = NULL; 1090 ifa_list_size = 0; 1091 } 1092 #endif 1093 1094 /* 1095 * Detach an interface from the list of "active" interfaces, 1096 * freeing any resources as we go along. 1097 * 1098 * NOTE: This routine must be called with a valid thread context, 1099 * as it may block. 1100 */ 1101 void 1102 if_detach(struct ifnet *ifp) 1103 { 1104 struct socket so; 1105 struct ifaddr *ifa; 1106 #ifdef IFAREF_DEBUG 1107 struct ifaddr *last_ifa = NULL; 1108 #endif 1109 struct domain *dp; 1110 const struct protosw *pr; 1111 int s, i, family, purged; 1112 uint64_t xc; 1113 1114 #ifdef IFAREF_DEBUG 1115 if_build_ifa_list(ifp); 1116 #endif 1117 /* 1118 * XXX It's kind of lame that we have to have the 1119 * XXX socket structure... 1120 */ 1121 memset(&so, 0, sizeof(so)); 1122 1123 s = splnet(); 1124 1125 sysctl_teardown(&ifp->if_sysctl_log); 1126 mutex_enter(ifp->if_ioctl_lock); 1127 if_deactivate(ifp); 1128 mutex_exit(ifp->if_ioctl_lock); 1129 1130 IFNET_LOCK(); 1131 ifindex2ifnet[ifp->if_index] = NULL; 1132 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1133 IFNET_WRITER_REMOVE(ifp); 1134 pserialize_perform(ifnet_psz); 1135 IFNET_UNLOCK(); 1136 1137 mutex_obj_free(ifp->if_ioctl_lock); 1138 ifp->if_ioctl_lock = NULL; 1139 1140 if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) { 1141 ifp->if_slowtimo = NULL; 1142 callout_halt(ifp->if_slowtimo_ch, NULL); 1143 callout_destroy(ifp->if_slowtimo_ch); 1144 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1145 } 1146 1147 /* 1148 * Do an if_down() to give protocols a chance to do something. 1149 */ 1150 if_down(ifp); 1151 1152 #ifdef ALTQ 1153 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1154 altq_disable(&ifp->if_snd); 1155 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1156 altq_detach(&ifp->if_snd); 1157 #endif 1158 1159 mutex_obj_free(ifp->if_snd.ifq_lock); 1160 1161 #if NCARP > 0 1162 /* Remove the interface from any carp group it is a part of. */ 1163 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1164 carp_ifdetach(ifp); 1165 #endif 1166 1167 /* 1168 * Rip all the addresses off the interface. This should make 1169 * all of the routes go away. 1170 * 1171 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1172 * from the list, including our "cursor", ifa. For safety, 1173 * and to honor the TAILQ abstraction, I just restart the 1174 * loop after each removal. Note that the loop will exit 1175 * when all of the remaining ifaddrs belong to the AF_LINK 1176 * family. I am counting on the historical fact that at 1177 * least one pr_usrreq in each address domain removes at 1178 * least one ifaddr. 1179 */ 1180 again: 1181 IFADDR_READER_FOREACH(ifa, ifp) { 1182 family = ifa->ifa_addr->sa_family; 1183 #ifdef IFAREF_DEBUG 1184 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1185 ifa, family, ifa->ifa_refcnt); 1186 if (last_ifa != NULL && ifa == last_ifa) 1187 panic("if_detach: loop detected"); 1188 last_ifa = ifa; 1189 #endif 1190 if (family == AF_LINK) 1191 continue; 1192 dp = pffinddomain(family); 1193 #ifdef DIAGNOSTIC 1194 if (dp == NULL) 1195 panic("if_detach: no domain for AF %d", 1196 family); 1197 #endif 1198 /* 1199 * XXX These PURGEIF calls are redundant with the 1200 * purge-all-families calls below, but are left in for 1201 * now both to make a smaller change, and to avoid 1202 * unplanned interactions with clearing of 1203 * ifp->if_addrlist. 1204 */ 1205 purged = 0; 1206 for (pr = dp->dom_protosw; 1207 pr < dp->dom_protoswNPROTOSW; pr++) { 1208 so.so_proto = pr; 1209 if (pr->pr_usrreqs) { 1210 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1211 purged = 1; 1212 } 1213 } 1214 if (purged == 0) { 1215 /* 1216 * XXX What's really the best thing to do 1217 * XXX here? --thorpej (at) NetBSD.org 1218 */ 1219 printf("if_detach: WARNING: AF %d not purged\n", 1220 family); 1221 ifa_remove(ifp, ifa); 1222 } 1223 goto again; 1224 } 1225 1226 if_free_sadl(ifp); 1227 1228 /* Walk the routing table looking for stragglers. */ 1229 for (i = 0; i <= AF_MAX; i++) { 1230 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 1231 continue; 1232 } 1233 1234 DOMAIN_FOREACH(dp) { 1235 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1236 { 1237 void *p = ifp->if_afdata[dp->dom_family]; 1238 if (p) { 1239 ifp->if_afdata[dp->dom_family] = NULL; 1240 (*dp->dom_ifdetach)(ifp, p); 1241 } 1242 } 1243 1244 /* 1245 * One would expect multicast memberships (INET and 1246 * INET6) on UDP sockets to be purged by the PURGEIF 1247 * calls above, but if all addresses were removed from 1248 * the interface prior to destruction, the calls will 1249 * not be made (e.g. ppp, for which pppd(8) generally 1250 * removes addresses before destroying the interface). 1251 * Because there is no invariant that multicast 1252 * memberships only exist for interfaces with IPv4 1253 * addresses, we must call PURGEIF regardless of 1254 * addresses. (Protocols which might store ifnet 1255 * pointers are marked with PR_PURGEIF.) 1256 */ 1257 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1258 so.so_proto = pr; 1259 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1260 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1261 } 1262 } 1263 1264 (void)pfil_run_hooks(if_pfil, 1265 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 1266 (void)pfil_head_destroy(ifp->if_pfil); 1267 1268 /* Announce that the interface is gone. */ 1269 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1270 1271 IF_AFDATA_LOCK_DESTROY(ifp); 1272 1273 if (if_is_link_state_changeable(ifp)) { 1274 softint_disestablish(ifp->if_link_si); 1275 ifp->if_link_si = NULL; 1276 } 1277 1278 /* 1279 * remove packets that came from ifp, from software interrupt queues. 1280 */ 1281 DOMAIN_FOREACH(dp) { 1282 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1283 struct ifqueue *iq = dp->dom_ifqueues[i]; 1284 if (iq == NULL) 1285 break; 1286 dp->dom_ifqueues[i] = NULL; 1287 if_detach_queues(ifp, iq); 1288 } 1289 } 1290 1291 /* 1292 * IP queues have to be processed separately: net-queue barrier 1293 * ensures that the packets are dequeued while a cross-call will 1294 * ensure that the interrupts have completed. FIXME: not quite.. 1295 */ 1296 #ifdef INET 1297 pktq_barrier(ip_pktq); 1298 #endif 1299 #ifdef INET6 1300 if (in6_present) 1301 pktq_barrier(ip6_pktq); 1302 #endif 1303 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1304 xc_wait(xc); 1305 1306 /* Wait for all readers to drain before freeing. */ 1307 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1308 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1309 1310 if (ifp->if_percpuq != NULL) { 1311 if_percpuq_destroy(ifp->if_percpuq); 1312 ifp->if_percpuq = NULL; 1313 } 1314 1315 splx(s); 1316 1317 #ifdef IFAREF_DEBUG 1318 if_check_and_free_ifa_list(ifp); 1319 #endif 1320 } 1321 1322 static void 1323 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1324 { 1325 struct mbuf *m, *prev, *next; 1326 1327 prev = NULL; 1328 for (m = q->ifq_head; m != NULL; m = next) { 1329 KASSERT((m->m_flags & M_PKTHDR) != 0); 1330 1331 next = m->m_nextpkt; 1332 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1333 prev = m; 1334 continue; 1335 } 1336 1337 if (prev != NULL) 1338 prev->m_nextpkt = m->m_nextpkt; 1339 else 1340 q->ifq_head = m->m_nextpkt; 1341 if (q->ifq_tail == m) 1342 q->ifq_tail = prev; 1343 q->ifq_len--; 1344 1345 m->m_nextpkt = NULL; 1346 m_freem(m); 1347 IF_DROP(q); 1348 } 1349 } 1350 1351 /* 1352 * Callback for a radix tree walk to delete all references to an 1353 * ifnet. 1354 */ 1355 static int 1356 if_rt_walktree(struct rtentry *rt, void *v) 1357 { 1358 struct ifnet *ifp = (struct ifnet *)v; 1359 int error; 1360 struct rtentry *retrt; 1361 1362 if (rt->rt_ifp != ifp) 1363 return 0; 1364 1365 /* Delete the entry. */ 1366 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 1367 rt_mask(rt), rt->rt_flags, &retrt); 1368 if (error == 0) { 1369 KASSERT(retrt == rt); 1370 KASSERT((retrt->rt_flags & RTF_UP) == 0); 1371 retrt->rt_ifp = NULL; 1372 rtfree(retrt); 1373 } else { 1374 printf("%s: warning: unable to delete rtentry @ %p, " 1375 "error = %d\n", ifp->if_xname, rt, error); 1376 } 1377 return ERESTART; 1378 } 1379 1380 /* 1381 * Create a clone network interface. 1382 */ 1383 static int 1384 if_clone_create(const char *name) 1385 { 1386 struct if_clone *ifc; 1387 int unit; 1388 struct ifnet *ifp; 1389 struct psref psref; 1390 1391 ifc = if_clone_lookup(name, &unit); 1392 if (ifc == NULL) 1393 return EINVAL; 1394 1395 ifp = if_get(name, &psref); 1396 if (ifp != NULL) { 1397 if_put(ifp, &psref); 1398 return EEXIST; 1399 } 1400 1401 return (*ifc->ifc_create)(ifc, unit); 1402 } 1403 1404 /* 1405 * Destroy a clone network interface. 1406 */ 1407 static int 1408 if_clone_destroy(const char *name) 1409 { 1410 struct if_clone *ifc; 1411 struct ifnet *ifp; 1412 struct psref psref; 1413 1414 ifc = if_clone_lookup(name, NULL); 1415 if (ifc == NULL) 1416 return EINVAL; 1417 1418 if (ifc->ifc_destroy == NULL) 1419 return EOPNOTSUPP; 1420 1421 ifp = if_get(name, &psref); 1422 if (ifp == NULL) 1423 return ENXIO; 1424 1425 /* We have to disable ioctls here */ 1426 mutex_enter(ifp->if_ioctl_lock); 1427 ifp->if_ioctl = if_nullioctl; 1428 mutex_exit(ifp->if_ioctl_lock); 1429 1430 /* 1431 * We cannot call ifc_destroy with holding ifp. 1432 * Releasing ifp here is safe thanks to if_clone_mtx. 1433 */ 1434 if_put(ifp, &psref); 1435 1436 return (*ifc->ifc_destroy)(ifp); 1437 } 1438 1439 /* 1440 * Look up a network interface cloner. 1441 */ 1442 static struct if_clone * 1443 if_clone_lookup(const char *name, int *unitp) 1444 { 1445 struct if_clone *ifc; 1446 const char *cp; 1447 char *dp, ifname[IFNAMSIZ + 3]; 1448 int unit; 1449 1450 strcpy(ifname, "if_"); 1451 /* separate interface name from unit */ 1452 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1453 *cp && (*cp < '0' || *cp > '9');) 1454 *dp++ = *cp++; 1455 1456 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1457 return NULL; /* No name or unit number */ 1458 *dp++ = '\0'; 1459 1460 again: 1461 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1462 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1463 break; 1464 } 1465 1466 if (ifc == NULL) { 1467 if (*ifname == '\0' || 1468 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1469 return NULL; 1470 *ifname = '\0'; 1471 goto again; 1472 } 1473 1474 unit = 0; 1475 while (cp - name < IFNAMSIZ && *cp) { 1476 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1477 /* Bogus unit number. */ 1478 return NULL; 1479 } 1480 unit = (unit * 10) + (*cp++ - '0'); 1481 } 1482 1483 if (unitp != NULL) 1484 *unitp = unit; 1485 return ifc; 1486 } 1487 1488 /* 1489 * Register a network interface cloner. 1490 */ 1491 void 1492 if_clone_attach(struct if_clone *ifc) 1493 { 1494 1495 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1496 if_cloners_count++; 1497 } 1498 1499 /* 1500 * Unregister a network interface cloner. 1501 */ 1502 void 1503 if_clone_detach(struct if_clone *ifc) 1504 { 1505 1506 LIST_REMOVE(ifc, ifc_list); 1507 if_cloners_count--; 1508 } 1509 1510 /* 1511 * Provide list of interface cloners to userspace. 1512 */ 1513 int 1514 if_clone_list(int buf_count, char *buffer, int *total) 1515 { 1516 char outbuf[IFNAMSIZ], *dst; 1517 struct if_clone *ifc; 1518 int count, error = 0; 1519 1520 *total = if_cloners_count; 1521 if ((dst = buffer) == NULL) { 1522 /* Just asking how many there are. */ 1523 return 0; 1524 } 1525 1526 if (buf_count < 0) 1527 return EINVAL; 1528 1529 count = (if_cloners_count < buf_count) ? 1530 if_cloners_count : buf_count; 1531 1532 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1533 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1534 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1535 if (outbuf[sizeof(outbuf) - 1] != '\0') 1536 return ENAMETOOLONG; 1537 error = copyout(outbuf, dst, sizeof(outbuf)); 1538 if (error != 0) 1539 break; 1540 } 1541 1542 return error; 1543 } 1544 1545 void 1546 ifaref(struct ifaddr *ifa) 1547 { 1548 ifa->ifa_refcnt++; 1549 } 1550 1551 void 1552 ifafree(struct ifaddr *ifa) 1553 { 1554 KASSERT(ifa != NULL); 1555 KASSERT(ifa->ifa_refcnt > 0); 1556 1557 if (--ifa->ifa_refcnt == 0) { 1558 free(ifa, M_IFADDR); 1559 } 1560 } 1561 1562 void 1563 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1564 { 1565 ifa->ifa_ifp = ifp; 1566 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1567 IFADDR_ENTRY_INIT(ifa); 1568 IFADDR_WRITER_INSERT_TAIL(ifp, ifa); 1569 ifaref(ifa); 1570 } 1571 1572 void 1573 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1574 { 1575 KASSERT(ifa->ifa_ifp == ifp); 1576 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1577 IFADDR_WRITER_REMOVE(ifa); 1578 /* TODO psref_target_destroy */ 1579 IFADDR_ENTRY_DESTROY(ifa); 1580 ifafree(ifa); 1581 } 1582 1583 static inline int 1584 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1585 { 1586 return sockaddr_cmp(sa1, sa2) == 0; 1587 } 1588 1589 /* 1590 * Locate an interface based on a complete address. 1591 */ 1592 /*ARGSUSED*/ 1593 struct ifaddr * 1594 ifa_ifwithaddr(const struct sockaddr *addr) 1595 { 1596 struct ifnet *ifp; 1597 struct ifaddr *ifa; 1598 int s; 1599 1600 s = pserialize_read_enter(); 1601 IFNET_READER_FOREACH(ifp) { 1602 if (if_is_deactivated(ifp)) 1603 continue; 1604 IFADDR_READER_FOREACH(ifa, ifp) { 1605 if (ifa->ifa_addr->sa_family != addr->sa_family) 1606 continue; 1607 if (equal(addr, ifa->ifa_addr)) 1608 return ifa; 1609 if ((ifp->if_flags & IFF_BROADCAST) && 1610 ifa->ifa_broadaddr && 1611 /* IP6 doesn't have broadcast */ 1612 ifa->ifa_broadaddr->sa_len != 0 && 1613 equal(ifa->ifa_broadaddr, addr)) 1614 return ifa; 1615 } 1616 } 1617 pserialize_read_exit(s); 1618 return NULL; 1619 } 1620 1621 /* 1622 * Locate the point to point interface with a given destination address. 1623 */ 1624 /*ARGSUSED*/ 1625 struct ifaddr * 1626 ifa_ifwithdstaddr(const struct sockaddr *addr) 1627 { 1628 struct ifnet *ifp; 1629 struct ifaddr *ifa; 1630 int s; 1631 1632 s = pserialize_read_enter(); 1633 IFNET_READER_FOREACH(ifp) { 1634 if (if_is_deactivated(ifp)) 1635 continue; 1636 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1637 continue; 1638 IFADDR_READER_FOREACH(ifa, ifp) { 1639 if (ifa->ifa_addr->sa_family != addr->sa_family || 1640 ifa->ifa_dstaddr == NULL) 1641 continue; 1642 if (equal(addr, ifa->ifa_dstaddr)) 1643 return ifa; 1644 } 1645 } 1646 pserialize_read_exit(s); 1647 return NULL; 1648 } 1649 1650 /* 1651 * Find an interface on a specific network. If many, choice 1652 * is most specific found. 1653 */ 1654 struct ifaddr * 1655 ifa_ifwithnet(const struct sockaddr *addr) 1656 { 1657 struct ifnet *ifp; 1658 struct ifaddr *ifa; 1659 const struct sockaddr_dl *sdl; 1660 struct ifaddr *ifa_maybe = 0; 1661 u_int af = addr->sa_family; 1662 const char *addr_data = addr->sa_data, *cplim; 1663 int s; 1664 1665 if (af == AF_LINK) { 1666 sdl = satocsdl(addr); 1667 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1668 ifindex2ifnet[sdl->sdl_index] && 1669 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1670 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1671 } 1672 } 1673 #ifdef NETATALK 1674 if (af == AF_APPLETALK) { 1675 const struct sockaddr_at *sat, *sat2; 1676 sat = (const struct sockaddr_at *)addr; 1677 s = pserialize_read_enter(); 1678 IFNET_READER_FOREACH(ifp) { 1679 if (if_is_deactivated(ifp)) 1680 continue; 1681 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1682 if (ifa == NULL) 1683 continue; 1684 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1685 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1686 return ifa; /* exact match */ 1687 if (ifa_maybe == NULL) { 1688 /* else keep the if with the right range */ 1689 ifa_maybe = ifa; 1690 } 1691 } 1692 pserialize_read_exit(s); 1693 return ifa_maybe; 1694 } 1695 #endif 1696 s = pserialize_read_enter(); 1697 IFNET_READER_FOREACH(ifp) { 1698 if (if_is_deactivated(ifp)) 1699 continue; 1700 IFADDR_READER_FOREACH(ifa, ifp) { 1701 const char *cp, *cp2, *cp3; 1702 1703 if (ifa->ifa_addr->sa_family != af || 1704 ifa->ifa_netmask == NULL) 1705 next: continue; 1706 cp = addr_data; 1707 cp2 = ifa->ifa_addr->sa_data; 1708 cp3 = ifa->ifa_netmask->sa_data; 1709 cplim = (const char *)ifa->ifa_netmask + 1710 ifa->ifa_netmask->sa_len; 1711 while (cp3 < cplim) { 1712 if ((*cp++ ^ *cp2++) & *cp3++) { 1713 /* want to continue for() loop */ 1714 goto next; 1715 } 1716 } 1717 if (ifa_maybe == NULL || 1718 rt_refines(ifa->ifa_netmask, 1719 ifa_maybe->ifa_netmask)) 1720 ifa_maybe = ifa; 1721 } 1722 } 1723 pserialize_read_exit(s); 1724 return ifa_maybe; 1725 } 1726 1727 /* 1728 * Find the interface of the addresss. 1729 */ 1730 struct ifaddr * 1731 ifa_ifwithladdr(const struct sockaddr *addr) 1732 { 1733 struct ifaddr *ia; 1734 1735 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1736 (ia = ifa_ifwithnet(addr))) 1737 return ia; 1738 return NULL; 1739 } 1740 1741 /* 1742 * Find an interface using a specific address family 1743 */ 1744 struct ifaddr * 1745 ifa_ifwithaf(int af) 1746 { 1747 struct ifnet *ifp; 1748 struct ifaddr *ifa = NULL; 1749 int s; 1750 1751 s = pserialize_read_enter(); 1752 IFNET_READER_FOREACH(ifp) { 1753 if (if_is_deactivated(ifp)) 1754 continue; 1755 IFADDR_READER_FOREACH(ifa, ifp) { 1756 if (ifa->ifa_addr->sa_family == af) 1757 goto out; 1758 } 1759 } 1760 out: 1761 pserialize_read_exit(s); 1762 return ifa; 1763 } 1764 1765 /* 1766 * Find an interface address specific to an interface best matching 1767 * a given address. 1768 */ 1769 struct ifaddr * 1770 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1771 { 1772 struct ifaddr *ifa; 1773 const char *cp, *cp2, *cp3; 1774 const char *cplim; 1775 struct ifaddr *ifa_maybe = 0; 1776 u_int af = addr->sa_family; 1777 1778 if (if_is_deactivated(ifp)) 1779 return NULL; 1780 1781 if (af >= AF_MAX) 1782 return NULL; 1783 1784 IFADDR_READER_FOREACH(ifa, ifp) { 1785 if (ifa->ifa_addr->sa_family != af) 1786 continue; 1787 ifa_maybe = ifa; 1788 if (ifa->ifa_netmask == NULL) { 1789 if (equal(addr, ifa->ifa_addr) || 1790 (ifa->ifa_dstaddr && 1791 equal(addr, ifa->ifa_dstaddr))) 1792 return ifa; 1793 continue; 1794 } 1795 cp = addr->sa_data; 1796 cp2 = ifa->ifa_addr->sa_data; 1797 cp3 = ifa->ifa_netmask->sa_data; 1798 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1799 for (; cp3 < cplim; cp3++) { 1800 if ((*cp++ ^ *cp2++) & *cp3) 1801 break; 1802 } 1803 if (cp3 == cplim) 1804 return ifa; 1805 } 1806 return ifa_maybe; 1807 } 1808 1809 /* 1810 * Default action when installing a route with a Link Level gateway. 1811 * Lookup an appropriate real ifa to point to. 1812 * This should be moved to /sys/net/link.c eventually. 1813 */ 1814 void 1815 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1816 { 1817 struct ifaddr *ifa; 1818 const struct sockaddr *dst; 1819 struct ifnet *ifp; 1820 1821 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1822 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1823 return; 1824 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1825 rt_replace_ifa(rt, ifa); 1826 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1827 ifa->ifa_rtrequest(cmd, rt, info); 1828 } 1829 } 1830 1831 /* 1832 * bitmask macros to manage a densely packed link_state change queue. 1833 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 1834 * LINK_STATE_UP(2) we need 2 bits for each state change. 1835 * As a state change to store is 0, treat all bits set as an unset item. 1836 */ 1837 #define LQ_ITEM_BITS 2 1838 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 1839 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 1840 #define LINK_STATE_UNSET LQ_ITEM_MASK 1841 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 1842 #define LQ_STORE(q, i, v) \ 1843 do { \ 1844 (q) &= ~LQ_MASK((i)); \ 1845 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 1846 } while (0 /* CONSTCOND */) 1847 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 1848 #define LQ_POP(q, v) \ 1849 do { \ 1850 (v) = LQ_ITEM((q), 0); \ 1851 (q) >>= LQ_ITEM_BITS; \ 1852 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1853 } while (0 /* CONSTCOND */) 1854 #define LQ_PUSH(q, v) \ 1855 do { \ 1856 (q) >>= LQ_ITEM_BITS; \ 1857 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1858 } while (0 /* CONSTCOND */) 1859 #define LQ_FIND_UNSET(q, i) \ 1860 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 1861 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 1862 break; \ 1863 } 1864 /* 1865 * Handle a change in the interface link state and 1866 * queue notifications. 1867 */ 1868 void 1869 if_link_state_change(struct ifnet *ifp, int link_state) 1870 { 1871 int s, idx; 1872 1873 KASSERTMSG(if_is_link_state_changeable(ifp), 1874 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 1875 ifp->if_xname, ifp->if_extflags); 1876 1877 /* Ensure change is to a valid state */ 1878 switch (link_state) { 1879 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 1880 case LINK_STATE_DOWN: /* FALLTHROUGH */ 1881 case LINK_STATE_UP: 1882 break; 1883 default: 1884 #ifdef DEBUG 1885 printf("%s: invalid link state %d\n", 1886 ifp->if_xname, link_state); 1887 #endif 1888 return; 1889 } 1890 1891 s = splnet(); 1892 1893 /* Find the last unset event in the queue. */ 1894 LQ_FIND_UNSET(ifp->if_link_queue, idx); 1895 1896 /* 1897 * Ensure link_state doesn't match the last event in the queue. 1898 * ifp->if_link_state is not checked and set here because 1899 * that would present an inconsistent picture to the system. 1900 */ 1901 if (idx != 0 && 1902 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 1903 goto out; 1904 1905 /* Handle queue overflow. */ 1906 if (idx == LQ_MAX(ifp->if_link_queue)) { 1907 uint8_t lost; 1908 1909 /* 1910 * The DOWN state must be protected from being pushed off 1911 * the queue to ensure that userland will always be 1912 * in a sane state. 1913 * Because DOWN is protected, there is no need to protect 1914 * UNKNOWN. 1915 * It should be invalid to change from any other state to 1916 * UNKNOWN anyway ... 1917 */ 1918 lost = LQ_ITEM(ifp->if_link_queue, 0); 1919 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 1920 if (lost == LINK_STATE_DOWN) { 1921 lost = LQ_ITEM(ifp->if_link_queue, 0); 1922 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 1923 } 1924 printf("%s: lost link state change %s\n", 1925 ifp->if_xname, 1926 lost == LINK_STATE_UP ? "UP" : 1927 lost == LINK_STATE_DOWN ? "DOWN" : 1928 "UNKNOWN"); 1929 } else 1930 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 1931 1932 softint_schedule(ifp->if_link_si); 1933 1934 out: 1935 splx(s); 1936 } 1937 1938 /* 1939 * Handle interface link state change notifications. 1940 * Must be called at splnet(). 1941 */ 1942 static void 1943 if_link_state_change0(struct ifnet *ifp, int link_state) 1944 { 1945 struct domain *dp; 1946 1947 /* Ensure the change is still valid. */ 1948 if (ifp->if_link_state == link_state) 1949 return; 1950 1951 #ifdef DEBUG 1952 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1953 link_state == LINK_STATE_UP ? "UP" : 1954 link_state == LINK_STATE_DOWN ? "DOWN" : 1955 "UNKNOWN", 1956 ifp->if_link_state == LINK_STATE_UP ? "UP" : 1957 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 1958 "UNKNOWN"); 1959 #endif 1960 1961 /* 1962 * When going from UNKNOWN to UP, we need to mark existing 1963 * addresses as tentative and restart DAD as we may have 1964 * erroneously not found a duplicate. 1965 * 1966 * This needs to happen before rt_ifmsg to avoid a race where 1967 * listeners would have an address and expect it to work right 1968 * away. 1969 */ 1970 if (link_state == LINK_STATE_UP && 1971 ifp->if_link_state == LINK_STATE_UNKNOWN) 1972 { 1973 DOMAIN_FOREACH(dp) { 1974 if (dp->dom_if_link_state_change != NULL) 1975 dp->dom_if_link_state_change(ifp, 1976 LINK_STATE_DOWN); 1977 } 1978 } 1979 1980 ifp->if_link_state = link_state; 1981 1982 /* Notify that the link state has changed. */ 1983 rt_ifmsg(ifp); 1984 1985 #if NCARP > 0 1986 if (ifp->if_carp) 1987 carp_carpdev_state(ifp); 1988 #endif 1989 1990 DOMAIN_FOREACH(dp) { 1991 if (dp->dom_if_link_state_change != NULL) 1992 dp->dom_if_link_state_change(ifp, link_state); 1993 } 1994 } 1995 1996 /* 1997 * Process the interface link state change queue. 1998 */ 1999 static void 2000 if_link_state_change_si(void *arg) 2001 { 2002 struct ifnet *ifp = arg; 2003 int s; 2004 uint8_t state; 2005 2006 s = splnet(); 2007 2008 /* Pop a link state change from the queue and process it. */ 2009 LQ_POP(ifp->if_link_queue, state); 2010 if_link_state_change0(ifp, state); 2011 2012 /* If there is a link state change to come, schedule it. */ 2013 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 2014 softint_schedule(ifp->if_link_si); 2015 2016 splx(s); 2017 } 2018 2019 /* 2020 * Default action when installing a local route on a point-to-point 2021 * interface. 2022 */ 2023 void 2024 p2p_rtrequest(int req, struct rtentry *rt, 2025 __unused const struct rt_addrinfo *info) 2026 { 2027 struct ifnet *ifp = rt->rt_ifp; 2028 struct ifaddr *ifa, *lo0ifa; 2029 2030 switch (req) { 2031 case RTM_ADD: 2032 if ((rt->rt_flags & RTF_LOCAL) == 0) 2033 break; 2034 2035 rt->rt_ifp = lo0ifp; 2036 2037 IFADDR_READER_FOREACH(ifa, ifp) { 2038 if (equal(rt_getkey(rt), ifa->ifa_addr)) 2039 break; 2040 } 2041 if (ifa == NULL) 2042 break; 2043 2044 /* 2045 * Ensure lo0 has an address of the same family. 2046 */ 2047 IFADDR_READER_FOREACH(lo0ifa, lo0ifp) { 2048 if (lo0ifa->ifa_addr->sa_family == 2049 ifa->ifa_addr->sa_family) 2050 break; 2051 } 2052 if (lo0ifa == NULL) 2053 break; 2054 2055 /* 2056 * Make sure to set rt->rt_ifa to the interface 2057 * address we are using, otherwise we will have trouble 2058 * with source address selection. 2059 */ 2060 if (ifa != rt->rt_ifa) 2061 rt_replace_ifa(rt, ifa); 2062 break; 2063 case RTM_DELETE: 2064 default: 2065 break; 2066 } 2067 } 2068 2069 /* 2070 * Mark an interface down and notify protocols of 2071 * the transition. 2072 * NOTE: must be called at splsoftnet or equivalent. 2073 */ 2074 void 2075 if_down(struct ifnet *ifp) 2076 { 2077 struct ifaddr *ifa; 2078 struct domain *dp; 2079 2080 ifp->if_flags &= ~IFF_UP; 2081 nanotime(&ifp->if_lastchange); 2082 IFADDR_READER_FOREACH(ifa, ifp) 2083 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2084 IFQ_PURGE(&ifp->if_snd); 2085 #if NCARP > 0 2086 if (ifp->if_carp) 2087 carp_carpdev_state(ifp); 2088 #endif 2089 rt_ifmsg(ifp); 2090 DOMAIN_FOREACH(dp) { 2091 if (dp->dom_if_down) 2092 dp->dom_if_down(ifp); 2093 } 2094 } 2095 2096 /* 2097 * Mark an interface up and notify protocols of 2098 * the transition. 2099 * NOTE: must be called at splsoftnet or equivalent. 2100 */ 2101 void 2102 if_up(struct ifnet *ifp) 2103 { 2104 #ifdef notyet 2105 struct ifaddr *ifa; 2106 #endif 2107 struct domain *dp; 2108 2109 ifp->if_flags |= IFF_UP; 2110 nanotime(&ifp->if_lastchange); 2111 #ifdef notyet 2112 /* this has no effect on IP, and will kill all ISO connections XXX */ 2113 IFADDR_READER_FOREACH(ifa, ifp) 2114 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2115 #endif 2116 #if NCARP > 0 2117 if (ifp->if_carp) 2118 carp_carpdev_state(ifp); 2119 #endif 2120 rt_ifmsg(ifp); 2121 DOMAIN_FOREACH(dp) { 2122 if (dp->dom_if_up) 2123 dp->dom_if_up(ifp); 2124 } 2125 } 2126 2127 /* 2128 * Handle interface slowtimo timer routine. Called 2129 * from softclock, we decrement timer (if set) and 2130 * call the appropriate interface routine on expiration. 2131 */ 2132 static void 2133 if_slowtimo(void *arg) 2134 { 2135 void (*slowtimo)(struct ifnet *); 2136 struct ifnet *ifp = arg; 2137 int s; 2138 2139 slowtimo = ifp->if_slowtimo; 2140 if (__predict_false(slowtimo == NULL)) 2141 return; 2142 2143 s = splnet(); 2144 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2145 (*slowtimo)(ifp); 2146 2147 splx(s); 2148 2149 if (__predict_true(ifp->if_slowtimo != NULL)) 2150 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2151 } 2152 2153 /* 2154 * Set/clear promiscuous mode on interface ifp based on the truth value 2155 * of pswitch. The calls are reference counted so that only the first 2156 * "on" request actually has an effect, as does the final "off" request. 2157 * Results are undefined if the "off" and "on" requests are not matched. 2158 */ 2159 int 2160 ifpromisc(struct ifnet *ifp, int pswitch) 2161 { 2162 int pcount, ret; 2163 short nflags; 2164 2165 pcount = ifp->if_pcount; 2166 if (pswitch) { 2167 /* 2168 * Allow the device to be "placed" into promiscuous 2169 * mode even if it is not configured up. It will 2170 * consult IFF_PROMISC when it is brought up. 2171 */ 2172 if (ifp->if_pcount++ != 0) 2173 return 0; 2174 nflags = ifp->if_flags | IFF_PROMISC; 2175 } else { 2176 if (--ifp->if_pcount > 0) 2177 return 0; 2178 nflags = ifp->if_flags & ~IFF_PROMISC; 2179 } 2180 ret = if_flags_set(ifp, nflags); 2181 /* Restore interface state if not successful. */ 2182 if (ret != 0) { 2183 ifp->if_pcount = pcount; 2184 } 2185 return ret; 2186 } 2187 2188 /* 2189 * Map interface name to 2190 * interface structure pointer. 2191 */ 2192 struct ifnet * 2193 ifunit(const char *name) 2194 { 2195 struct ifnet *ifp; 2196 const char *cp = name; 2197 u_int unit = 0; 2198 u_int i; 2199 int s; 2200 2201 /* 2202 * If the entire name is a number, treat it as an ifindex. 2203 */ 2204 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2205 unit = unit * 10 + (*cp - '0'); 2206 } 2207 2208 /* 2209 * If the number took all of the name, then it's a valid ifindex. 2210 */ 2211 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2212 if (unit >= if_indexlim) 2213 return NULL; 2214 ifp = ifindex2ifnet[unit]; 2215 if (ifp == NULL || if_is_deactivated(ifp)) 2216 return NULL; 2217 return ifp; 2218 } 2219 2220 ifp = NULL; 2221 s = pserialize_read_enter(); 2222 IFNET_READER_FOREACH(ifp) { 2223 if (if_is_deactivated(ifp)) 2224 continue; 2225 if (strcmp(ifp->if_xname, name) == 0) 2226 goto out; 2227 } 2228 out: 2229 pserialize_read_exit(s); 2230 return ifp; 2231 } 2232 2233 /* 2234 * Get a reference of an ifnet object by an interface name. 2235 * The returned reference is protected by psref(9). The caller 2236 * must release a returned reference by if_put after use. 2237 */ 2238 struct ifnet * 2239 if_get(const char *name, struct psref *psref) 2240 { 2241 struct ifnet *ifp; 2242 const char *cp = name; 2243 u_int unit = 0; 2244 u_int i; 2245 int s; 2246 2247 /* 2248 * If the entire name is a number, treat it as an ifindex. 2249 */ 2250 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2251 unit = unit * 10 + (*cp - '0'); 2252 } 2253 2254 /* 2255 * If the number took all of the name, then it's a valid ifindex. 2256 */ 2257 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2258 if (unit >= if_indexlim) 2259 return NULL; 2260 ifp = ifindex2ifnet[unit]; 2261 if (ifp == NULL || if_is_deactivated(ifp)) 2262 return NULL; 2263 return ifp; 2264 } 2265 2266 ifp = NULL; 2267 s = pserialize_read_enter(); 2268 IFNET_READER_FOREACH(ifp) { 2269 if (if_is_deactivated(ifp)) 2270 continue; 2271 if (strcmp(ifp->if_xname, name) == 0) { 2272 psref_acquire(psref, &ifp->if_psref, 2273 ifnet_psref_class); 2274 goto out; 2275 } 2276 } 2277 out: 2278 pserialize_read_exit(s); 2279 return ifp; 2280 } 2281 2282 /* 2283 * Release a reference of an ifnet object given by if_get or 2284 * if_get_byindex. 2285 */ 2286 void 2287 if_put(const struct ifnet *ifp, struct psref *psref) 2288 { 2289 2290 if (ifp == NULL) 2291 return; 2292 2293 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2294 } 2295 2296 ifnet_t * 2297 if_byindex(u_int idx) 2298 { 2299 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2300 } 2301 2302 /* 2303 * Get a reference of an ifnet object by an interface index. 2304 * The returned reference is protected by psref(9). The caller 2305 * must release a returned reference by if_put after use. 2306 */ 2307 ifnet_t * 2308 if_get_byindex(u_int idx, struct psref *psref) 2309 { 2310 ifnet_t *ifp; 2311 int s; 2312 2313 s = pserialize_read_enter(); 2314 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2315 if (__predict_true(ifp != NULL)) 2316 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2317 pserialize_read_exit(s); 2318 2319 return ifp; 2320 } 2321 2322 /* 2323 * XXX it's safe only if the passed ifp is guaranteed to not be freed, 2324 * for example the ifp is already held or some other object is held which 2325 * guarantes the ifp to not be freed indirectly. 2326 */ 2327 void 2328 if_acquire_NOMPSAFE(struct ifnet *ifp, struct psref *psref) 2329 { 2330 2331 KASSERT(ifp->if_index != 0); 2332 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2333 } 2334 2335 bool 2336 if_held(struct ifnet *ifp) 2337 { 2338 2339 return psref_held(&ifp->if_psref, ifnet_psref_class); 2340 } 2341 2342 2343 /* common */ 2344 int 2345 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2346 { 2347 int s; 2348 struct ifreq *ifr; 2349 struct ifcapreq *ifcr; 2350 struct ifdatareq *ifdr; 2351 2352 switch (cmd) { 2353 case SIOCSIFCAP: 2354 ifcr = data; 2355 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2356 return EINVAL; 2357 2358 if (ifcr->ifcr_capenable == ifp->if_capenable) 2359 return 0; 2360 2361 ifp->if_capenable = ifcr->ifcr_capenable; 2362 2363 /* Pre-compute the checksum flags mask. */ 2364 ifp->if_csum_flags_tx = 0; 2365 ifp->if_csum_flags_rx = 0; 2366 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2367 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2368 } 2369 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2370 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2371 } 2372 2373 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2374 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2375 } 2376 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2377 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2378 } 2379 2380 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2381 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2382 } 2383 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2384 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2385 } 2386 2387 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2388 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2389 } 2390 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2391 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2392 } 2393 2394 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2395 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2396 } 2397 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2398 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2399 } 2400 if (ifp->if_flags & IFF_UP) 2401 return ENETRESET; 2402 return 0; 2403 case SIOCSIFFLAGS: 2404 ifr = data; 2405 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2406 s = splnet(); 2407 if_down(ifp); 2408 splx(s); 2409 } 2410 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2411 s = splnet(); 2412 if_up(ifp); 2413 splx(s); 2414 } 2415 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2416 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2417 break; 2418 case SIOCGIFFLAGS: 2419 ifr = data; 2420 ifr->ifr_flags = ifp->if_flags; 2421 break; 2422 2423 case SIOCGIFMETRIC: 2424 ifr = data; 2425 ifr->ifr_metric = ifp->if_metric; 2426 break; 2427 2428 case SIOCGIFMTU: 2429 ifr = data; 2430 ifr->ifr_mtu = ifp->if_mtu; 2431 break; 2432 2433 case SIOCGIFDLT: 2434 ifr = data; 2435 ifr->ifr_dlt = ifp->if_dlt; 2436 break; 2437 2438 case SIOCGIFCAP: 2439 ifcr = data; 2440 ifcr->ifcr_capabilities = ifp->if_capabilities; 2441 ifcr->ifcr_capenable = ifp->if_capenable; 2442 break; 2443 2444 case SIOCSIFMETRIC: 2445 ifr = data; 2446 ifp->if_metric = ifr->ifr_metric; 2447 break; 2448 2449 case SIOCGIFDATA: 2450 ifdr = data; 2451 ifdr->ifdr_data = ifp->if_data; 2452 break; 2453 2454 case SIOCGIFINDEX: 2455 ifr = data; 2456 ifr->ifr_index = ifp->if_index; 2457 break; 2458 2459 case SIOCZIFDATA: 2460 ifdr = data; 2461 ifdr->ifdr_data = ifp->if_data; 2462 /* 2463 * Assumes that the volatile counters that can be 2464 * zero'ed are at the end of if_data. 2465 */ 2466 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2467 offsetof(struct if_data, ifi_ipackets)); 2468 /* 2469 * The memset() clears to the bottm of if_data. In the area, 2470 * if_lastchange is included. Please be careful if new entry 2471 * will be added into if_data or rewite this. 2472 * 2473 * And also, update if_lastchnage. 2474 */ 2475 getnanotime(&ifp->if_lastchange); 2476 break; 2477 case SIOCSIFMTU: 2478 ifr = data; 2479 if (ifp->if_mtu == ifr->ifr_mtu) 2480 break; 2481 ifp->if_mtu = ifr->ifr_mtu; 2482 /* 2483 * If the link MTU changed, do network layer specific procedure. 2484 */ 2485 #ifdef INET6 2486 if (in6_present) 2487 nd6_setmtu(ifp); 2488 #endif 2489 return ENETRESET; 2490 default: 2491 return ENOTTY; 2492 } 2493 return 0; 2494 } 2495 2496 int 2497 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2498 { 2499 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2500 struct ifaddr *ifa; 2501 const struct sockaddr *any, *sa; 2502 union { 2503 struct sockaddr sa; 2504 struct sockaddr_storage ss; 2505 } u, v; 2506 2507 switch (cmd) { 2508 case SIOCSIFADDRPREF: 2509 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2510 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2511 NULL) != 0) 2512 return EPERM; 2513 case SIOCGIFADDRPREF: 2514 break; 2515 default: 2516 return EOPNOTSUPP; 2517 } 2518 2519 /* sanity checks */ 2520 if (data == NULL || ifp == NULL) { 2521 panic("invalid argument to %s", __func__); 2522 /*NOTREACHED*/ 2523 } 2524 2525 /* address must be specified on ADD and DELETE */ 2526 sa = sstocsa(&ifap->ifap_addr); 2527 if (sa->sa_family != sofamily(so)) 2528 return EINVAL; 2529 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2530 return EINVAL; 2531 2532 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2533 2534 IFADDR_READER_FOREACH(ifa, ifp) { 2535 if (ifa->ifa_addr->sa_family != sa->sa_family) 2536 continue; 2537 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2538 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2539 break; 2540 } 2541 if (ifa == NULL) 2542 return EADDRNOTAVAIL; 2543 2544 switch (cmd) { 2545 case SIOCSIFADDRPREF: 2546 ifa->ifa_preference = ifap->ifap_preference; 2547 return 0; 2548 case SIOCGIFADDRPREF: 2549 /* fill in the if_laddrreq structure */ 2550 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2551 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2552 ifap->ifap_preference = ifa->ifa_preference; 2553 return 0; 2554 default: 2555 return EOPNOTSUPP; 2556 } 2557 } 2558 2559 /* 2560 * Interface ioctls. 2561 */ 2562 static int 2563 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2564 { 2565 struct ifnet *ifp; 2566 struct ifreq *ifr; 2567 int error = 0; 2568 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2569 u_long ocmd = cmd; 2570 #endif 2571 short oif_flags; 2572 #ifdef COMPAT_OIFREQ 2573 struct ifreq ifrb; 2574 struct oifreq *oifr = NULL; 2575 #endif 2576 int r; 2577 struct psref psref; 2578 int bound; 2579 2580 switch (cmd) { 2581 #ifdef COMPAT_OIFREQ 2582 case OSIOCGIFCONF: 2583 case OOSIOCGIFCONF: 2584 return compat_ifconf(cmd, data); 2585 #endif 2586 #ifdef COMPAT_OIFDATA 2587 case OSIOCGIFDATA: 2588 case OSIOCZIFDATA: 2589 return compat_ifdatareq(l, cmd, data); 2590 #endif 2591 case SIOCGIFCONF: 2592 return ifconf(cmd, data); 2593 case SIOCINITIFADDR: 2594 return EPERM; 2595 } 2596 2597 #ifdef COMPAT_OIFREQ 2598 cmd = compat_cvtcmd(cmd); 2599 if (cmd != ocmd) { 2600 oifr = data; 2601 data = ifr = &ifrb; 2602 ifreqo2n(oifr, ifr); 2603 } else 2604 #endif 2605 ifr = data; 2606 2607 switch (cmd) { 2608 case SIOCIFCREATE: 2609 case SIOCIFDESTROY: 2610 bound = curlwp_bind(); 2611 if (l != NULL) { 2612 ifp = if_get(ifr->ifr_name, &psref); 2613 error = kauth_authorize_network(l->l_cred, 2614 KAUTH_NETWORK_INTERFACE, 2615 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2616 (void *)cmd, NULL); 2617 if (ifp != NULL) 2618 if_put(ifp, &psref); 2619 if (error != 0) { 2620 curlwp_bindx(bound); 2621 return error; 2622 } 2623 } 2624 mutex_enter(&if_clone_mtx); 2625 r = (cmd == SIOCIFCREATE) ? 2626 if_clone_create(ifr->ifr_name) : 2627 if_clone_destroy(ifr->ifr_name); 2628 mutex_exit(&if_clone_mtx); 2629 curlwp_bindx(bound); 2630 return r; 2631 2632 case SIOCIFGCLONERS: 2633 { 2634 struct if_clonereq *req = (struct if_clonereq *)data; 2635 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2636 &req->ifcr_total); 2637 } 2638 } 2639 2640 bound = curlwp_bind(); 2641 ifp = if_get(ifr->ifr_name, &psref); 2642 if (ifp == NULL) { 2643 curlwp_bindx(bound); 2644 return ENXIO; 2645 } 2646 2647 switch (cmd) { 2648 case SIOCALIFADDR: 2649 case SIOCDLIFADDR: 2650 case SIOCSIFADDRPREF: 2651 case SIOCSIFFLAGS: 2652 case SIOCSIFCAP: 2653 case SIOCSIFMETRIC: 2654 case SIOCZIFDATA: 2655 case SIOCSIFMTU: 2656 case SIOCSIFPHYADDR: 2657 case SIOCDIFPHYADDR: 2658 #ifdef INET6 2659 case SIOCSIFPHYADDR_IN6: 2660 #endif 2661 case SIOCSLIFPHYADDR: 2662 case SIOCADDMULTI: 2663 case SIOCDELMULTI: 2664 case SIOCSIFMEDIA: 2665 case SIOCSDRVSPEC: 2666 case SIOCG80211: 2667 case SIOCS80211: 2668 case SIOCS80211NWID: 2669 case SIOCS80211NWKEY: 2670 case SIOCS80211POWER: 2671 case SIOCS80211BSSID: 2672 case SIOCS80211CHANNEL: 2673 case SIOCSLINKSTR: 2674 if (l != NULL) { 2675 error = kauth_authorize_network(l->l_cred, 2676 KAUTH_NETWORK_INTERFACE, 2677 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2678 (void *)cmd, NULL); 2679 if (error != 0) 2680 goto out; 2681 } 2682 } 2683 2684 oif_flags = ifp->if_flags; 2685 2686 mutex_enter(ifp->if_ioctl_lock); 2687 2688 error = (*ifp->if_ioctl)(ifp, cmd, data); 2689 if (error != ENOTTY) 2690 ; 2691 else if (so->so_proto == NULL) 2692 error = EOPNOTSUPP; 2693 else { 2694 #ifdef COMPAT_OSOCK 2695 error = compat_ifioctl(so, ocmd, cmd, data, l); 2696 #else 2697 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2698 cmd, data, ifp); 2699 #endif 2700 } 2701 2702 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2703 if ((ifp->if_flags & IFF_UP) != 0) { 2704 int s = splnet(); 2705 if_up(ifp); 2706 splx(s); 2707 } 2708 } 2709 #ifdef COMPAT_OIFREQ 2710 if (cmd != ocmd) 2711 ifreqn2o(oifr, ifr); 2712 #endif 2713 2714 mutex_exit(ifp->if_ioctl_lock); 2715 out: 2716 if_put(ifp, &psref); 2717 curlwp_bindx(bound); 2718 return error; 2719 } 2720 2721 /* 2722 * Return interface configuration 2723 * of system. List may be used 2724 * in later ioctl's (above) to get 2725 * other information. 2726 * 2727 * Each record is a struct ifreq. Before the addition of 2728 * sockaddr_storage, the API rule was that sockaddr flavors that did 2729 * not fit would extend beyond the struct ifreq, with the next struct 2730 * ifreq starting sa_len beyond the struct sockaddr. Because the 2731 * union in struct ifreq includes struct sockaddr_storage, every kind 2732 * of sockaddr must fit. Thus, there are no longer any overlength 2733 * records. 2734 * 2735 * Records are added to the user buffer if they fit, and ifc_len is 2736 * adjusted to the length that was written. Thus, the user is only 2737 * assured of getting the complete list if ifc_len on return is at 2738 * least sizeof(struct ifreq) less than it was on entry. 2739 * 2740 * If the user buffer pointer is NULL, this routine copies no data and 2741 * returns the amount of space that would be needed. 2742 * 2743 * Invariants: 2744 * ifrp points to the next part of the user's buffer to be used. If 2745 * ifrp != NULL, space holds the number of bytes remaining that we may 2746 * write at ifrp. Otherwise, space holds the number of bytes that 2747 * would have been written had there been adequate space. 2748 */ 2749 /*ARGSUSED*/ 2750 static int 2751 ifconf(u_long cmd, void *data) 2752 { 2753 struct ifconf *ifc = (struct ifconf *)data; 2754 struct ifnet *ifp; 2755 struct ifaddr *ifa; 2756 struct ifreq ifr, *ifrp = NULL; 2757 int space = 0, error = 0; 2758 const int sz = (int)sizeof(struct ifreq); 2759 const bool docopy = ifc->ifc_req != NULL; 2760 int s; 2761 int bound; 2762 struct psref psref; 2763 2764 if (docopy) { 2765 space = ifc->ifc_len; 2766 ifrp = ifc->ifc_req; 2767 } 2768 2769 bound = curlwp_bind(); 2770 s = pserialize_read_enter(); 2771 IFNET_READER_FOREACH(ifp) { 2772 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 2773 pserialize_read_exit(s); 2774 2775 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2776 sizeof(ifr.ifr_name)); 2777 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 2778 error = ENAMETOOLONG; 2779 goto release_exit; 2780 } 2781 if (IFADDR_READER_EMPTY(ifp)) { 2782 /* Interface with no addresses - send zero sockaddr. */ 2783 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2784 if (!docopy) { 2785 space += sz; 2786 continue; 2787 } 2788 if (space >= sz) { 2789 error = copyout(&ifr, ifrp, sz); 2790 if (error != 0) 2791 goto release_exit; 2792 ifrp++; 2793 space -= sz; 2794 } 2795 } 2796 2797 IFADDR_READER_FOREACH(ifa, ifp) { 2798 struct sockaddr *sa = ifa->ifa_addr; 2799 /* all sockaddrs must fit in sockaddr_storage */ 2800 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2801 2802 if (!docopy) { 2803 space += sz; 2804 continue; 2805 } 2806 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2807 if (space >= sz) { 2808 error = copyout(&ifr, ifrp, sz); 2809 if (error != 0) 2810 goto release_exit; 2811 ifrp++; space -= sz; 2812 } 2813 } 2814 2815 s = pserialize_read_enter(); 2816 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2817 } 2818 pserialize_read_exit(s); 2819 curlwp_bindx(bound); 2820 2821 if (docopy) { 2822 KASSERT(0 <= space && space <= ifc->ifc_len); 2823 ifc->ifc_len -= space; 2824 } else { 2825 KASSERT(space >= 0); 2826 ifc->ifc_len = space; 2827 } 2828 return (0); 2829 2830 release_exit: 2831 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2832 curlwp_bindx(bound); 2833 return error; 2834 } 2835 2836 int 2837 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2838 { 2839 uint8_t len; 2840 #ifdef COMPAT_OIFREQ 2841 struct ifreq ifrb; 2842 struct oifreq *oifr = NULL; 2843 u_long ocmd = cmd; 2844 cmd = compat_cvtcmd(cmd); 2845 if (cmd != ocmd) { 2846 oifr = (struct oifreq *)(void *)ifr; 2847 ifr = &ifrb; 2848 ifreqo2n(oifr, ifr); 2849 len = sizeof(oifr->ifr_addr); 2850 } else 2851 #endif 2852 len = sizeof(ifr->ifr_ifru.ifru_space); 2853 2854 if (len < sa->sa_len) 2855 return EFBIG; 2856 2857 memset(&ifr->ifr_addr, 0, len); 2858 sockaddr_copy(&ifr->ifr_addr, len, sa); 2859 2860 #ifdef COMPAT_OIFREQ 2861 if (cmd != ocmd) 2862 ifreqn2o(oifr, ifr); 2863 #endif 2864 return 0; 2865 } 2866 2867 /* 2868 * wrapper function for the drivers which doesn't have if_transmit(). 2869 */ 2870 static int 2871 if_transmit(struct ifnet *ifp, struct mbuf *m) 2872 { 2873 int s, error; 2874 2875 s = splnet(); 2876 2877 /* 2878 * If NET_MPSAFE is not defined , IFQ_LOCK() is nop. 2879 * use KERNEL_LOCK instead of ifq_lock. 2880 */ 2881 #ifndef NET_MPSAFE 2882 KERNEL_LOCK(1, NULL); 2883 #endif 2884 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2885 #ifndef NET_MPSAFE 2886 KERNEL_UNLOCK_ONE(NULL); 2887 #endif 2888 if (error != 0) { 2889 /* mbuf is already freed */ 2890 goto out; 2891 } 2892 2893 ifp->if_obytes += m->m_pkthdr.len;; 2894 if (m->m_flags & M_MCAST) 2895 ifp->if_omcasts++; 2896 2897 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2898 if_start_lock(ifp); 2899 out: 2900 splx(s); 2901 2902 return error; 2903 } 2904 2905 int 2906 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 2907 { 2908 int error; 2909 2910 #ifdef ALTQ 2911 KERNEL_LOCK(1, NULL); 2912 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 2913 error = if_transmit(ifp, m); 2914 KERNEL_UNLOCK_ONE(NULL); 2915 } else { 2916 KERNEL_UNLOCK_ONE(NULL); 2917 error = (*ifp->if_transmit)(ifp, m); 2918 } 2919 #else /* !ALTQ */ 2920 error = (*ifp->if_transmit)(ifp, m); 2921 #endif /* !ALTQ */ 2922 2923 return error; 2924 } 2925 2926 /* 2927 * Queue message on interface, and start output if interface 2928 * not yet active. 2929 */ 2930 int 2931 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 2932 { 2933 2934 return if_transmit_lock(ifp, m); 2935 } 2936 2937 /* 2938 * Queue message on interface, possibly using a second fast queue 2939 */ 2940 int 2941 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 2942 { 2943 int error = 0; 2944 2945 if (ifq != NULL 2946 #ifdef ALTQ 2947 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2948 #endif 2949 ) { 2950 if (IF_QFULL(ifq)) { 2951 IF_DROP(&ifp->if_snd); 2952 m_freem(m); 2953 if (error == 0) 2954 error = ENOBUFS; 2955 } else 2956 IF_ENQUEUE(ifq, m); 2957 } else 2958 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2959 if (error != 0) { 2960 ++ifp->if_oerrors; 2961 return error; 2962 } 2963 return 0; 2964 } 2965 2966 int 2967 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2968 { 2969 int rc; 2970 2971 if (ifp->if_initaddr != NULL) 2972 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2973 else if (src || 2974 /* FIXME: may not hold if_ioctl_lock */ 2975 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2976 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2977 2978 return rc; 2979 } 2980 2981 int 2982 if_do_dad(struct ifnet *ifp) 2983 { 2984 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2985 return 0; 2986 2987 switch (ifp->if_type) { 2988 case IFT_FAITH: 2989 /* 2990 * These interfaces do not have the IFF_LOOPBACK flag, 2991 * but loop packets back. We do not have to do DAD on such 2992 * interfaces. We should even omit it, because loop-backed 2993 * responses would confuse the DAD procedure. 2994 */ 2995 return 0; 2996 default: 2997 /* 2998 * Our DAD routine requires the interface up and running. 2999 * However, some interfaces can be up before the RUNNING 3000 * status. Additionaly, users may try to assign addresses 3001 * before the interface becomes up (or running). 3002 * We simply skip DAD in such a case as a work around. 3003 * XXX: we should rather mark "tentative" on such addresses, 3004 * and do DAD after the interface becomes ready. 3005 */ 3006 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 3007 (IFF_UP|IFF_RUNNING)) 3008 return 0; 3009 3010 return 1; 3011 } 3012 } 3013 3014 int 3015 if_flags_set(ifnet_t *ifp, const short flags) 3016 { 3017 int rc; 3018 3019 if (ifp->if_setflags != NULL) 3020 rc = (*ifp->if_setflags)(ifp, flags); 3021 else { 3022 short cantflags, chgdflags; 3023 struct ifreq ifr; 3024 3025 chgdflags = ifp->if_flags ^ flags; 3026 cantflags = chgdflags & IFF_CANTCHANGE; 3027 3028 if (cantflags != 0) 3029 ifp->if_flags ^= cantflags; 3030 3031 /* Traditionally, we do not call if_ioctl after 3032 * setting/clearing only IFF_PROMISC if the interface 3033 * isn't IFF_UP. Uphold that tradition. 3034 */ 3035 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 3036 return 0; 3037 3038 memset(&ifr, 0, sizeof(ifr)); 3039 3040 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 3041 /* FIXME: may not hold if_ioctl_lock */ 3042 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 3043 3044 if (rc != 0 && cantflags != 0) 3045 ifp->if_flags ^= cantflags; 3046 } 3047 3048 return rc; 3049 } 3050 3051 int 3052 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 3053 { 3054 int rc; 3055 struct ifreq ifr; 3056 3057 if (ifp->if_mcastop != NULL) 3058 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 3059 else { 3060 ifreq_setaddr(cmd, &ifr, sa); 3061 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 3062 } 3063 3064 return rc; 3065 } 3066 3067 static void 3068 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 3069 struct ifaltq *ifq) 3070 { 3071 const struct sysctlnode *cnode, *rnode; 3072 3073 if (sysctl_createv(clog, 0, NULL, &rnode, 3074 CTLFLAG_PERMANENT, 3075 CTLTYPE_NODE, "interfaces", 3076 SYSCTL_DESCR("Per-interface controls"), 3077 NULL, 0, NULL, 0, 3078 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3079 goto bad; 3080 3081 if (sysctl_createv(clog, 0, &rnode, &rnode, 3082 CTLFLAG_PERMANENT, 3083 CTLTYPE_NODE, ifname, 3084 SYSCTL_DESCR("Interface controls"), 3085 NULL, 0, NULL, 0, 3086 CTL_CREATE, CTL_EOL) != 0) 3087 goto bad; 3088 3089 if (sysctl_createv(clog, 0, &rnode, &rnode, 3090 CTLFLAG_PERMANENT, 3091 CTLTYPE_NODE, "sndq", 3092 SYSCTL_DESCR("Interface output queue controls"), 3093 NULL, 0, NULL, 0, 3094 CTL_CREATE, CTL_EOL) != 0) 3095 goto bad; 3096 3097 if (sysctl_createv(clog, 0, &rnode, &cnode, 3098 CTLFLAG_PERMANENT, 3099 CTLTYPE_INT, "len", 3100 SYSCTL_DESCR("Current output queue length"), 3101 NULL, 0, &ifq->ifq_len, 0, 3102 CTL_CREATE, CTL_EOL) != 0) 3103 goto bad; 3104 3105 if (sysctl_createv(clog, 0, &rnode, &cnode, 3106 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3107 CTLTYPE_INT, "maxlen", 3108 SYSCTL_DESCR("Maximum allowed output queue length"), 3109 NULL, 0, &ifq->ifq_maxlen, 0, 3110 CTL_CREATE, CTL_EOL) != 0) 3111 goto bad; 3112 3113 if (sysctl_createv(clog, 0, &rnode, &cnode, 3114 CTLFLAG_PERMANENT, 3115 CTLTYPE_INT, "drops", 3116 SYSCTL_DESCR("Packets dropped due to full output queue"), 3117 NULL, 0, &ifq->ifq_drops, 0, 3118 CTL_CREATE, CTL_EOL) != 0) 3119 goto bad; 3120 3121 return; 3122 bad: 3123 printf("%s: could not attach sysctl nodes\n", ifname); 3124 return; 3125 } 3126 3127 #if defined(INET) || defined(INET6) 3128 3129 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3130 static int \ 3131 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3132 { \ 3133 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3134 } 3135 3136 #if defined(INET) 3137 static int 3138 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3139 { 3140 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3141 } 3142 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3143 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3144 #endif 3145 3146 #if defined(INET6) 3147 static int 3148 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3149 { 3150 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3151 } 3152 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3153 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3154 #endif 3155 3156 static void 3157 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3158 { 3159 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3160 const char *pfname = NULL, *ipname = NULL; 3161 int ipn = 0, qid = 0; 3162 3163 switch (pf) { 3164 #if defined(INET) 3165 case PF_INET: 3166 len_func = sysctl_net_ip_pktq_items; 3167 maxlen_func = sysctl_net_ip_pktq_maxlen; 3168 drops_func = sysctl_net_ip_pktq_drops; 3169 pfname = "inet", ipn = IPPROTO_IP; 3170 ipname = "ip", qid = IPCTL_IFQ; 3171 break; 3172 #endif 3173 #if defined(INET6) 3174 case PF_INET6: 3175 len_func = sysctl_net_ip6_pktq_items; 3176 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3177 drops_func = sysctl_net_ip6_pktq_drops; 3178 pfname = "inet6", ipn = IPPROTO_IPV6; 3179 ipname = "ip6", qid = IPV6CTL_IFQ; 3180 break; 3181 #endif 3182 default: 3183 KASSERT(false); 3184 } 3185 3186 sysctl_createv(clog, 0, NULL, NULL, 3187 CTLFLAG_PERMANENT, 3188 CTLTYPE_NODE, pfname, NULL, 3189 NULL, 0, NULL, 0, 3190 CTL_NET, pf, CTL_EOL); 3191 sysctl_createv(clog, 0, NULL, NULL, 3192 CTLFLAG_PERMANENT, 3193 CTLTYPE_NODE, ipname, NULL, 3194 NULL, 0, NULL, 0, 3195 CTL_NET, pf, ipn, CTL_EOL); 3196 sysctl_createv(clog, 0, NULL, NULL, 3197 CTLFLAG_PERMANENT, 3198 CTLTYPE_NODE, "ifq", 3199 SYSCTL_DESCR("Protocol input queue controls"), 3200 NULL, 0, NULL, 0, 3201 CTL_NET, pf, ipn, qid, CTL_EOL); 3202 3203 sysctl_createv(clog, 0, NULL, NULL, 3204 CTLFLAG_PERMANENT, 3205 CTLTYPE_INT, "len", 3206 SYSCTL_DESCR("Current input queue length"), 3207 len_func, 0, NULL, 0, 3208 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3209 sysctl_createv(clog, 0, NULL, NULL, 3210 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3211 CTLTYPE_INT, "maxlen", 3212 SYSCTL_DESCR("Maximum allowed input queue length"), 3213 maxlen_func, 0, NULL, 0, 3214 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3215 sysctl_createv(clog, 0, NULL, NULL, 3216 CTLFLAG_PERMANENT, 3217 CTLTYPE_INT, "drops", 3218 SYSCTL_DESCR("Packets dropped due to full input queue"), 3219 drops_func, 0, NULL, 0, 3220 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3221 } 3222 #endif /* INET || INET6 */ 3223 3224 static int 3225 if_sdl_sysctl(SYSCTLFN_ARGS) 3226 { 3227 struct ifnet *ifp; 3228 const struct sockaddr_dl *sdl; 3229 struct psref psref; 3230 int error = 0; 3231 int bound; 3232 3233 if (namelen != 1) 3234 return EINVAL; 3235 3236 bound = curlwp_bind(); 3237 ifp = if_get_byindex(name[0], &psref); 3238 if (ifp == NULL) { 3239 error = ENODEV; 3240 goto out0; 3241 } 3242 3243 sdl = ifp->if_sadl; 3244 if (sdl == NULL) { 3245 *oldlenp = 0; 3246 goto out1; 3247 } 3248 3249 if (oldp == NULL) { 3250 *oldlenp = sdl->sdl_alen; 3251 goto out1; 3252 } 3253 3254 if (*oldlenp >= sdl->sdl_alen) 3255 *oldlenp = sdl->sdl_alen; 3256 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3257 out1: 3258 if_put(ifp, &psref); 3259 out0: 3260 curlwp_bindx(bound); 3261 return error; 3262 } 3263 3264 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 3265 { 3266 const struct sysctlnode *rnode = NULL; 3267 3268 sysctl_createv(clog, 0, NULL, &rnode, 3269 CTLFLAG_PERMANENT, 3270 CTLTYPE_NODE, "sdl", 3271 SYSCTL_DESCR("Get active link-layer address"), 3272 if_sdl_sysctl, 0, NULL, 0, 3273 CTL_NET, CTL_CREATE, CTL_EOL); 3274 } 3275
Indexes created Mon Sep 22 13:09:51 GMT 2025