if.c revision 1.446
1 /* $NetBSD: if.c,v 1.446 2019/03/01 11:06:57 pgoyette 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 1.446 2019/03/01 11:06:57 pgoyette Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 #include "opt_ipsec.h" 98 #include "opt_atalk.h" 99 #include "opt_wlan.h" 100 #include "opt_net_mpsafe.h" 101 #include "opt_mrouting.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 #include <sys/compat_stub.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 #include <netinet/ip_encap.h> 142 #include <net/bpf.h> 143 144 #ifdef INET6 145 #include <netinet6/in6_var.h> 146 #include <netinet6/nd6.h> 147 #endif 148 149 #include "ether.h" 150 #include "fddi.h" 151 #include "token.h" 152 153 #include "bridge.h" 154 #if NBRIDGE > 0 155 #include <net/if_bridgevar.h> 156 #endif 157 158 #include "carp.h" 159 #if NCARP > 0 160 #include <netinet/ip_carp.h> 161 #endif 162 163 #include <compat/sys/sockio.h> 164 #include <compat/sys/socket.h> 165 166 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 167 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 168 169 /* 170 * Global list of interfaces. 171 */ 172 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 173 struct ifnet_head ifnet_list; 174 175 struct pslist_head ifnet_pslist; 176 static ifnet_t ** ifindex2ifnet = NULL; 177 static u_int if_index = 1; 178 static size_t if_indexlim = 0; 179 static uint64_t index_gen; 180 /* Mutex to protect the above objects. */ 181 kmutex_t ifnet_mtx __cacheline_aligned; 182 static struct psref_class *ifnet_psref_class __read_mostly; 183 static pserialize_t ifnet_psz; 184 185 static kmutex_t if_clone_mtx; 186 187 struct ifnet *lo0ifp; 188 int ifqmaxlen = IFQ_MAXLEN; 189 190 struct psref_class *ifa_psref_class __read_mostly; 191 192 static int if_delroute_matcher(struct rtentry *, void *); 193 194 static bool if_is_unit(const char *); 195 static struct if_clone *if_clone_lookup(const char *, int *); 196 197 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 198 static int if_cloners_count; 199 200 /* Packet filtering hook for interfaces. */ 201 pfil_head_t * if_pfil __read_mostly; 202 203 static kauth_listener_t if_listener; 204 205 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 206 static void if_detach_queues(struct ifnet *, struct ifqueue *); 207 static void sysctl_sndq_setup(struct sysctllog **, const char *, 208 struct ifaltq *); 209 static void if_slowtimo(void *); 210 static void if_attachdomain1(struct ifnet *); 211 static int ifconf(u_long, void *); 212 static int if_transmit(struct ifnet *, struct mbuf *); 213 static int if_clone_create(const char *); 214 static int if_clone_destroy(const char *); 215 static void if_link_state_change_si(void *); 216 static void if_up_locked(struct ifnet *); 217 static void _if_down(struct ifnet *); 218 static void if_down_deactivated(struct ifnet *); 219 220 struct if_percpuq { 221 struct ifnet *ipq_ifp; 222 void *ipq_si; 223 struct percpu *ipq_ifqs; /* struct ifqueue */ 224 }; 225 226 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 227 228 static void if_percpuq_drops(void *, void *, struct cpu_info *); 229 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 230 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 231 struct if_percpuq *); 232 233 struct if_deferred_start { 234 struct ifnet *ids_ifp; 235 void (*ids_if_start)(struct ifnet *); 236 void *ids_si; 237 }; 238 239 static void if_deferred_start_softint(void *); 240 static void if_deferred_start_common(struct ifnet *); 241 static void if_deferred_start_destroy(struct ifnet *); 242 243 #if defined(INET) || defined(INET6) 244 static void sysctl_net_pktq_setup(struct sysctllog **, int); 245 #endif 246 247 static void if_sysctl_setup(struct sysctllog **); 248 249 static int 250 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 251 void *arg0, void *arg1, void *arg2, void *arg3) 252 { 253 int result; 254 enum kauth_network_req req; 255 256 result = KAUTH_RESULT_DEFER; 257 req = (enum kauth_network_req)arg1; 258 259 if (action != KAUTH_NETWORK_INTERFACE) 260 return result; 261 262 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 263 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 264 result = KAUTH_RESULT_ALLOW; 265 266 return result; 267 } 268 269 /* 270 * Network interface utility routines. 271 * 272 * Routines with ifa_ifwith* names take sockaddr *'s as 273 * parameters. 274 */ 275 void 276 ifinit(void) 277 { 278 279 #if (defined(INET) || defined(INET6)) 280 encapinit(); 281 #endif 282 283 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 284 if_listener_cb, NULL); 285 286 /* interfaces are available, inform socket code */ 287 ifioctl = doifioctl; 288 } 289 290 /* 291 * XXX Initialization before configure(). 292 * XXX hack to get pfil_add_hook working in autoconf. 293 */ 294 void 295 ifinit1(void) 296 { 297 298 #ifdef NET_MPSAFE 299 printf("NET_MPSAFE enabled\n"); 300 #endif 301 302 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 303 304 TAILQ_INIT(&ifnet_list); 305 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 306 ifnet_psz = pserialize_create(); 307 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 308 ifa_psref_class = psref_class_create("ifa", IPL_SOFTNET); 309 PSLIST_INIT(&ifnet_pslist); 310 311 if_indexlim = 8; 312 313 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 314 KASSERT(if_pfil != NULL); 315 316 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 317 etherinit(); 318 #endif 319 } 320 321 /* XXX must be after domaininit() */ 322 void 323 ifinit_post(void) 324 { 325 326 if_sysctl_setup(NULL); 327 } 328 329 ifnet_t * 330 if_alloc(u_char type) 331 { 332 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 333 } 334 335 void 336 if_free(ifnet_t *ifp) 337 { 338 kmem_free(ifp, sizeof(ifnet_t)); 339 } 340 341 void 342 if_initname(struct ifnet *ifp, const char *name, int unit) 343 { 344 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 345 "%s%d", name, unit); 346 } 347 348 /* 349 * Null routines used while an interface is going away. These routines 350 * just return an error. 351 */ 352 353 int 354 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 355 const struct sockaddr *so, const struct rtentry *rt) 356 { 357 358 return ENXIO; 359 } 360 361 void 362 if_nullinput(struct ifnet *ifp, struct mbuf *m) 363 { 364 365 /* Nothing. */ 366 } 367 368 void 369 if_nullstart(struct ifnet *ifp) 370 { 371 372 /* Nothing. */ 373 } 374 375 int 376 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 377 { 378 379 m_freem(m); 380 return ENXIO; 381 } 382 383 int 384 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 385 { 386 387 return ENXIO; 388 } 389 390 int 391 if_nullinit(struct ifnet *ifp) 392 { 393 394 return ENXIO; 395 } 396 397 void 398 if_nullstop(struct ifnet *ifp, int disable) 399 { 400 401 /* Nothing. */ 402 } 403 404 void 405 if_nullslowtimo(struct ifnet *ifp) 406 { 407 408 /* Nothing. */ 409 } 410 411 void 412 if_nulldrain(struct ifnet *ifp) 413 { 414 415 /* Nothing. */ 416 } 417 418 void 419 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 420 { 421 struct ifaddr *ifa; 422 struct sockaddr_dl *sdl; 423 424 ifp->if_addrlen = addrlen; 425 if_alloc_sadl(ifp); 426 ifa = ifp->if_dl; 427 sdl = satosdl(ifa->ifa_addr); 428 429 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 430 if (factory) { 431 KASSERT(ifp->if_hwdl == NULL); 432 ifp->if_hwdl = ifp->if_dl; 433 ifaref(ifp->if_hwdl); 434 } 435 /* TBD routing socket */ 436 } 437 438 struct ifaddr * 439 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 440 { 441 unsigned socksize, ifasize; 442 int addrlen, namelen; 443 struct sockaddr_dl *mask, *sdl; 444 struct ifaddr *ifa; 445 446 namelen = strlen(ifp->if_xname); 447 addrlen = ifp->if_addrlen; 448 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 449 ifasize = sizeof(*ifa) + 2 * socksize; 450 ifa = malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 451 452 sdl = (struct sockaddr_dl *)(ifa + 1); 453 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 454 455 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 456 ifp->if_xname, namelen, NULL, addrlen); 457 mask->sdl_family = AF_LINK; 458 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 459 memset(&mask->sdl_data[0], 0xff, namelen); 460 ifa->ifa_rtrequest = link_rtrequest; 461 ifa->ifa_addr = (struct sockaddr *)sdl; 462 ifa->ifa_netmask = (struct sockaddr *)mask; 463 ifa_psref_init(ifa); 464 465 *sdlp = sdl; 466 467 return ifa; 468 } 469 470 static void 471 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 472 { 473 const struct sockaddr_dl *sdl; 474 475 ifp->if_dl = ifa; 476 ifaref(ifa); 477 sdl = satosdl(ifa->ifa_addr); 478 ifp->if_sadl = sdl; 479 } 480 481 /* 482 * Allocate the link level name for the specified interface. This 483 * is an attachment helper. It must be called after ifp->if_addrlen 484 * is initialized, which may not be the case when if_attach() is 485 * called. 486 */ 487 void 488 if_alloc_sadl(struct ifnet *ifp) 489 { 490 struct ifaddr *ifa; 491 const struct sockaddr_dl *sdl; 492 493 /* 494 * If the interface already has a link name, release it 495 * now. This is useful for interfaces that can change 496 * link types, and thus switch link names often. 497 */ 498 if (ifp->if_sadl != NULL) 499 if_free_sadl(ifp, 0); 500 501 ifa = if_dl_create(ifp, &sdl); 502 503 ifa_insert(ifp, ifa); 504 if_sadl_setrefs(ifp, ifa); 505 } 506 507 static void 508 if_deactivate_sadl(struct ifnet *ifp) 509 { 510 struct ifaddr *ifa; 511 512 KASSERT(ifp->if_dl != NULL); 513 514 ifa = ifp->if_dl; 515 516 ifp->if_sadl = NULL; 517 518 ifp->if_dl = NULL; 519 ifafree(ifa); 520 } 521 522 static void 523 if_replace_sadl(struct ifnet *ifp, struct ifaddr *ifa) 524 { 525 struct ifaddr *old; 526 527 KASSERT(ifp->if_dl != NULL); 528 529 old = ifp->if_dl; 530 531 ifaref(ifa); 532 /* XXX Update if_dl and if_sadl atomically */ 533 ifp->if_dl = ifa; 534 ifp->if_sadl = satosdl(ifa->ifa_addr); 535 536 ifafree(old); 537 } 538 539 void 540 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa0, 541 const struct sockaddr_dl *sdl) 542 { 543 int s, ss; 544 struct ifaddr *ifa; 545 int bound = curlwp_bind(); 546 547 KASSERT(ifa_held(ifa0)); 548 549 s = splsoftnet(); 550 551 if_replace_sadl(ifp, ifa0); 552 553 ss = pserialize_read_enter(); 554 IFADDR_READER_FOREACH(ifa, ifp) { 555 struct psref psref; 556 ifa_acquire(ifa, &psref); 557 pserialize_read_exit(ss); 558 559 rtinit(ifa, RTM_LLINFO_UPD, 0); 560 561 ss = pserialize_read_enter(); 562 ifa_release(ifa, &psref); 563 } 564 pserialize_read_exit(ss); 565 566 splx(s); 567 curlwp_bindx(bound); 568 } 569 570 /* 571 * Free the link level name for the specified interface. This is 572 * a detach helper. This is called from if_detach(). 573 */ 574 void 575 if_free_sadl(struct ifnet *ifp, int factory) 576 { 577 struct ifaddr *ifa; 578 int s; 579 580 if (factory && ifp->if_hwdl != NULL) { 581 ifa = ifp->if_hwdl; 582 ifp->if_hwdl = NULL; 583 ifafree(ifa); 584 } 585 586 ifa = ifp->if_dl; 587 if (ifa == NULL) { 588 KASSERT(ifp->if_sadl == NULL); 589 return; 590 } 591 592 KASSERT(ifp->if_sadl != NULL); 593 594 s = splsoftnet(); 595 KASSERT(ifa->ifa_addr->sa_family == AF_LINK); 596 ifa_remove(ifp, ifa); 597 if_deactivate_sadl(ifp); 598 splx(s); 599 } 600 601 static void 602 if_getindex(ifnet_t *ifp) 603 { 604 bool hitlimit = false; 605 606 ifp->if_index_gen = index_gen++; 607 608 ifp->if_index = if_index; 609 if (ifindex2ifnet == NULL) { 610 if_index++; 611 goto skip; 612 } 613 while (if_byindex(ifp->if_index)) { 614 /* 615 * If we hit USHRT_MAX, we skip back to 0 since 616 * there are a number of places where the value 617 * of if_index or if_index itself is compared 618 * to or stored in an unsigned short. By 619 * jumping back, we won't botch those assignments 620 * or comparisons. 621 */ 622 if (++if_index == 0) { 623 if_index = 1; 624 } else if (if_index == USHRT_MAX) { 625 /* 626 * However, if we have to jump back to 627 * zero *twice* without finding an empty 628 * slot in ifindex2ifnet[], then there 629 * there are too many (>65535) interfaces. 630 */ 631 if (hitlimit) { 632 panic("too many interfaces"); 633 } 634 hitlimit = true; 635 if_index = 1; 636 } 637 ifp->if_index = if_index; 638 } 639 skip: 640 /* 641 * ifindex2ifnet is indexed by if_index. Since if_index will 642 * grow dynamically, it should grow too. 643 */ 644 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 645 size_t m, n, oldlim; 646 void *q; 647 648 oldlim = if_indexlim; 649 while (ifp->if_index >= if_indexlim) 650 if_indexlim <<= 1; 651 652 /* grow ifindex2ifnet */ 653 m = oldlim * sizeof(struct ifnet *); 654 n = if_indexlim * sizeof(struct ifnet *); 655 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 656 if (ifindex2ifnet != NULL) { 657 memcpy(q, ifindex2ifnet, m); 658 free(ifindex2ifnet, M_IFADDR); 659 } 660 ifindex2ifnet = (struct ifnet **)q; 661 } 662 ifindex2ifnet[ifp->if_index] = ifp; 663 } 664 665 /* 666 * Initialize an interface and assign an index for it. 667 * 668 * It must be called prior to a device specific attach routine 669 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 670 * and be followed by if_register: 671 * 672 * if_initialize(ifp); 673 * ether_ifattach(ifp, enaddr); 674 * if_register(ifp); 675 */ 676 int 677 if_initialize(ifnet_t *ifp) 678 { 679 int rv = 0; 680 681 KASSERT(if_indexlim > 0); 682 TAILQ_INIT(&ifp->if_addrlist); 683 684 /* 685 * Link level name is allocated later by a separate call to 686 * if_alloc_sadl(). 687 */ 688 689 if (ifp->if_snd.ifq_maxlen == 0) 690 ifp->if_snd.ifq_maxlen = ifqmaxlen; 691 692 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 693 694 ifp->if_link_state = LINK_STATE_UNKNOWN; 695 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 696 697 ifp->if_capenable = 0; 698 ifp->if_csum_flags_tx = 0; 699 ifp->if_csum_flags_rx = 0; 700 701 #ifdef ALTQ 702 ifp->if_snd.altq_type = 0; 703 ifp->if_snd.altq_disc = NULL; 704 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 705 ifp->if_snd.altq_tbr = NULL; 706 ifp->if_snd.altq_ifp = ifp; 707 #endif 708 709 IFQ_LOCK_INIT(&ifp->if_snd); 710 711 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 712 pfil_run_ifhooks(if_pfil, PFIL_IFNET_ATTACH, ifp); 713 714 IF_AFDATA_LOCK_INIT(ifp); 715 716 if (if_is_link_state_changeable(ifp)) { 717 u_int flags = SOFTINT_NET; 718 flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0; 719 ifp->if_link_si = softint_establish(flags, 720 if_link_state_change_si, ifp); 721 if (ifp->if_link_si == NULL) { 722 rv = ENOMEM; 723 goto fail; 724 } 725 } 726 727 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 728 PSLIST_INIT(&ifp->if_addr_pslist); 729 psref_target_init(&ifp->if_psref, ifnet_psref_class); 730 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 731 LIST_INIT(&ifp->if_multiaddrs); 732 733 IFNET_GLOBAL_LOCK(); 734 if_getindex(ifp); 735 IFNET_GLOBAL_UNLOCK(); 736 737 return 0; 738 739 fail: 740 IF_AFDATA_LOCK_DESTROY(ifp); 741 742 pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp); 743 (void)pfil_head_destroy(ifp->if_pfil); 744 745 IFQ_LOCK_DESTROY(&ifp->if_snd); 746 747 return rv; 748 } 749 750 /* 751 * Register an interface to the list of "active" interfaces. 752 */ 753 void 754 if_register(ifnet_t *ifp) 755 { 756 /* 757 * If the driver has not supplied its own if_ioctl, then 758 * supply the default. 759 */ 760 if (ifp->if_ioctl == NULL) 761 ifp->if_ioctl = ifioctl_common; 762 763 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 764 765 if (!STAILQ_EMPTY(&domains)) 766 if_attachdomain1(ifp); 767 768 /* Announce the interface. */ 769 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 770 771 if (ifp->if_slowtimo != NULL) { 772 ifp->if_slowtimo_ch = 773 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 774 callout_init(ifp->if_slowtimo_ch, 0); 775 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 776 if_slowtimo(ifp); 777 } 778 779 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 780 ifp->if_transmit = if_transmit; 781 782 IFNET_GLOBAL_LOCK(); 783 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 784 IFNET_WRITER_INSERT_TAIL(ifp); 785 IFNET_GLOBAL_UNLOCK(); 786 } 787 788 /* 789 * The if_percpuq framework 790 * 791 * It allows network device drivers to execute the network stack 792 * in softint (so called softint-based if_input). It utilizes 793 * softint and percpu ifqueue. It doesn't distribute any packets 794 * between CPUs, unlike pktqueue(9). 795 * 796 * Currently we support two options for device drivers to apply the framework: 797 * - Use it implicitly with less changes 798 * - If you use if_attach in driver's _attach function and if_input in 799 * driver's Rx interrupt handler, a packet is queued and a softint handles 800 * the packet implicitly 801 * - Use it explicitly in each driver (recommended) 802 * - You can use if_percpuq_* directly in your driver 803 * - In this case, you need to allocate struct if_percpuq in driver's softc 804 * - See wm(4) as a reference implementation 805 */ 806 807 static void 808 if_percpuq_softint(void *arg) 809 { 810 struct if_percpuq *ipq = arg; 811 struct ifnet *ifp = ipq->ipq_ifp; 812 struct mbuf *m; 813 814 while ((m = if_percpuq_dequeue(ipq)) != NULL) { 815 ifp->if_ipackets++; 816 bpf_mtap(ifp, m, BPF_D_IN); 817 818 ifp->_if_input(ifp, m); 819 } 820 } 821 822 static void 823 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 824 { 825 struct ifqueue *const ifq = p; 826 827 memset(ifq, 0, sizeof(*ifq)); 828 ifq->ifq_maxlen = IFQ_MAXLEN; 829 } 830 831 struct if_percpuq * 832 if_percpuq_create(struct ifnet *ifp) 833 { 834 struct if_percpuq *ipq; 835 u_int flags = SOFTINT_NET; 836 837 flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0; 838 839 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 840 ipq->ipq_ifp = ifp; 841 ipq->ipq_si = softint_establish(flags, if_percpuq_softint, ipq); 842 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 843 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 844 845 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 846 847 return ipq; 848 } 849 850 static struct mbuf * 851 if_percpuq_dequeue(struct if_percpuq *ipq) 852 { 853 struct mbuf *m; 854 struct ifqueue *ifq; 855 int s; 856 857 s = splnet(); 858 ifq = percpu_getref(ipq->ipq_ifqs); 859 IF_DEQUEUE(ifq, m); 860 percpu_putref(ipq->ipq_ifqs); 861 splx(s); 862 863 return m; 864 } 865 866 static void 867 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 868 { 869 struct ifqueue *const ifq = p; 870 871 IF_PURGE(ifq); 872 } 873 874 void 875 if_percpuq_destroy(struct if_percpuq *ipq) 876 { 877 878 /* if_detach may already destroy it */ 879 if (ipq == NULL) 880 return; 881 882 softint_disestablish(ipq->ipq_si); 883 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 884 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 885 kmem_free(ipq, sizeof(*ipq)); 886 } 887 888 void 889 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 890 { 891 struct ifqueue *ifq; 892 int s; 893 894 KASSERT(ipq != NULL); 895 896 s = splnet(); 897 ifq = percpu_getref(ipq->ipq_ifqs); 898 if (IF_QFULL(ifq)) { 899 IF_DROP(ifq); 900 percpu_putref(ipq->ipq_ifqs); 901 m_freem(m); 902 goto out; 903 } 904 IF_ENQUEUE(ifq, m); 905 percpu_putref(ipq->ipq_ifqs); 906 907 softint_schedule(ipq->ipq_si); 908 out: 909 splx(s); 910 } 911 912 static void 913 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 914 { 915 struct ifqueue *const ifq = p; 916 int *sum = arg; 917 918 *sum += ifq->ifq_drops; 919 } 920 921 static int 922 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 923 { 924 struct sysctlnode node; 925 struct if_percpuq *ipq; 926 int sum = 0; 927 int error; 928 929 node = *rnode; 930 ipq = node.sysctl_data; 931 932 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 933 934 node.sysctl_data = ∑ 935 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 936 if (error != 0 || newp == NULL) 937 return error; 938 939 return 0; 940 } 941 942 static void 943 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 944 struct if_percpuq *ipq) 945 { 946 const struct sysctlnode *cnode, *rnode; 947 948 if (sysctl_createv(clog, 0, NULL, &rnode, 949 CTLFLAG_PERMANENT, 950 CTLTYPE_NODE, "interfaces", 951 SYSCTL_DESCR("Per-interface controls"), 952 NULL, 0, NULL, 0, 953 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 954 goto bad; 955 956 if (sysctl_createv(clog, 0, &rnode, &rnode, 957 CTLFLAG_PERMANENT, 958 CTLTYPE_NODE, ifname, 959 SYSCTL_DESCR("Interface controls"), 960 NULL, 0, NULL, 0, 961 CTL_CREATE, CTL_EOL) != 0) 962 goto bad; 963 964 if (sysctl_createv(clog, 0, &rnode, &rnode, 965 CTLFLAG_PERMANENT, 966 CTLTYPE_NODE, "rcvq", 967 SYSCTL_DESCR("Interface input queue controls"), 968 NULL, 0, NULL, 0, 969 CTL_CREATE, CTL_EOL) != 0) 970 goto bad; 971 972 #ifdef NOTYET 973 /* XXX Should show each per-CPU queue length? */ 974 if (sysctl_createv(clog, 0, &rnode, &rnode, 975 CTLFLAG_PERMANENT, 976 CTLTYPE_INT, "len", 977 SYSCTL_DESCR("Current input queue length"), 978 sysctl_percpuq_len, 0, NULL, 0, 979 CTL_CREATE, CTL_EOL) != 0) 980 goto bad; 981 982 if (sysctl_createv(clog, 0, &rnode, &cnode, 983 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 984 CTLTYPE_INT, "maxlen", 985 SYSCTL_DESCR("Maximum allowed input queue length"), 986 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 987 CTL_CREATE, CTL_EOL) != 0) 988 goto bad; 989 #endif 990 991 if (sysctl_createv(clog, 0, &rnode, &cnode, 992 CTLFLAG_PERMANENT, 993 CTLTYPE_INT, "drops", 994 SYSCTL_DESCR("Total packets dropped due to full input queue"), 995 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 996 CTL_CREATE, CTL_EOL) != 0) 997 goto bad; 998 999 return; 1000 bad: 1001 printf("%s: could not attach sysctl nodes\n", ifname); 1002 return; 1003 } 1004 1005 /* 1006 * The deferred if_start framework 1007 * 1008 * The common APIs to defer if_start to softint when if_start is requested 1009 * from a device driver running in hardware interrupt context. 1010 */ 1011 /* 1012 * Call ifp->if_start (or equivalent) in a dedicated softint for 1013 * deferred if_start. 1014 */ 1015 static void 1016 if_deferred_start_softint(void *arg) 1017 { 1018 struct if_deferred_start *ids = arg; 1019 struct ifnet *ifp = ids->ids_ifp; 1020 1021 ids->ids_if_start(ifp); 1022 } 1023 1024 /* 1025 * The default callback function for deferred if_start. 1026 */ 1027 static void 1028 if_deferred_start_common(struct ifnet *ifp) 1029 { 1030 int s; 1031 1032 s = splnet(); 1033 if_start_lock(ifp); 1034 splx(s); 1035 } 1036 1037 static inline bool 1038 if_snd_is_used(struct ifnet *ifp) 1039 { 1040 1041 return ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit || 1042 ifp->if_transmit == if_transmit || 1043 ALTQ_IS_ENABLED(&ifp->if_snd); 1044 } 1045 1046 /* 1047 * Schedule deferred if_start. 1048 */ 1049 void 1050 if_schedule_deferred_start(struct ifnet *ifp) 1051 { 1052 1053 KASSERT(ifp->if_deferred_start != NULL); 1054 1055 if (if_snd_is_used(ifp) && IFQ_IS_EMPTY(&ifp->if_snd)) 1056 return; 1057 1058 softint_schedule(ifp->if_deferred_start->ids_si); 1059 } 1060 1061 /* 1062 * Create an instance of deferred if_start. A driver should call the function 1063 * only if the driver needs deferred if_start. Drivers can setup their own 1064 * deferred if_start function via 2nd argument. 1065 */ 1066 void 1067 if_deferred_start_init(struct ifnet *ifp, void (*func)(struct ifnet *)) 1068 { 1069 struct if_deferred_start *ids; 1070 u_int flags = SOFTINT_NET; 1071 1072 flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0; 1073 1074 ids = kmem_zalloc(sizeof(*ids), KM_SLEEP); 1075 ids->ids_ifp = ifp; 1076 ids->ids_si = softint_establish(flags, if_deferred_start_softint, ids); 1077 if (func != NULL) 1078 ids->ids_if_start = func; 1079 else 1080 ids->ids_if_start = if_deferred_start_common; 1081 1082 ifp->if_deferred_start = ids; 1083 } 1084 1085 static void 1086 if_deferred_start_destroy(struct ifnet *ifp) 1087 { 1088 1089 if (ifp->if_deferred_start == NULL) 1090 return; 1091 1092 softint_disestablish(ifp->if_deferred_start->ids_si); 1093 kmem_free(ifp->if_deferred_start, sizeof(*ifp->if_deferred_start)); 1094 ifp->if_deferred_start = NULL; 1095 } 1096 1097 /* 1098 * The common interface input routine that is called by device drivers, 1099 * which should be used only when the driver's rx handler already runs 1100 * in softint. 1101 */ 1102 void 1103 if_input(struct ifnet *ifp, struct mbuf *m) 1104 { 1105 1106 KASSERT(ifp->if_percpuq == NULL); 1107 KASSERT(!cpu_intr_p()); 1108 1109 ifp->if_ipackets++; 1110 bpf_mtap(ifp, m, BPF_D_IN); 1111 1112 ifp->_if_input(ifp, m); 1113 } 1114 1115 /* 1116 * DEPRECATED. Use if_initialize and if_register instead. 1117 * See the above comment of if_initialize. 1118 * 1119 * Note that it implicitly enables if_percpuq to make drivers easy to 1120 * migrate softint-based if_input without much changes. If you don't 1121 * want to enable it, use if_initialize instead. 1122 */ 1123 int 1124 if_attach(ifnet_t *ifp) 1125 { 1126 int rv; 1127 1128 rv = if_initialize(ifp); 1129 if (rv != 0) 1130 return rv; 1131 1132 ifp->if_percpuq = if_percpuq_create(ifp); 1133 if_register(ifp); 1134 1135 return 0; 1136 } 1137 1138 void 1139 if_attachdomain(void) 1140 { 1141 struct ifnet *ifp; 1142 int s; 1143 int bound = curlwp_bind(); 1144 1145 s = pserialize_read_enter(); 1146 IFNET_READER_FOREACH(ifp) { 1147 struct psref psref; 1148 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 1149 pserialize_read_exit(s); 1150 if_attachdomain1(ifp); 1151 s = pserialize_read_enter(); 1152 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 1153 } 1154 pserialize_read_exit(s); 1155 curlwp_bindx(bound); 1156 } 1157 1158 static void 1159 if_attachdomain1(struct ifnet *ifp) 1160 { 1161 struct domain *dp; 1162 int s; 1163 1164 s = splsoftnet(); 1165 1166 /* address family dependent data region */ 1167 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 1168 DOMAIN_FOREACH(dp) { 1169 if (dp->dom_ifattach != NULL) 1170 ifp->if_afdata[dp->dom_family] = 1171 (*dp->dom_ifattach)(ifp); 1172 } 1173 1174 splx(s); 1175 } 1176 1177 /* 1178 * Deactivate an interface. This points all of the procedure 1179 * handles at error stubs. May be called from interrupt context. 1180 */ 1181 void 1182 if_deactivate(struct ifnet *ifp) 1183 { 1184 int s; 1185 1186 s = splsoftnet(); 1187 1188 ifp->if_output = if_nulloutput; 1189 ifp->_if_input = if_nullinput; 1190 ifp->if_start = if_nullstart; 1191 ifp->if_transmit = if_nulltransmit; 1192 ifp->if_ioctl = if_nullioctl; 1193 ifp->if_init = if_nullinit; 1194 ifp->if_stop = if_nullstop; 1195 ifp->if_slowtimo = if_nullslowtimo; 1196 ifp->if_drain = if_nulldrain; 1197 1198 /* No more packets may be enqueued. */ 1199 ifp->if_snd.ifq_maxlen = 0; 1200 1201 splx(s); 1202 } 1203 1204 bool 1205 if_is_deactivated(const struct ifnet *ifp) 1206 { 1207 1208 return ifp->if_output == if_nulloutput; 1209 } 1210 1211 void 1212 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1213 { 1214 struct ifaddr *ifa, *nifa; 1215 int s; 1216 1217 s = pserialize_read_enter(); 1218 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) { 1219 nifa = IFADDR_READER_NEXT(ifa); 1220 if (ifa->ifa_addr->sa_family != family) 1221 continue; 1222 pserialize_read_exit(s); 1223 1224 (*purgeaddr)(ifa); 1225 1226 s = pserialize_read_enter(); 1227 } 1228 pserialize_read_exit(s); 1229 } 1230 1231 #ifdef IFAREF_DEBUG 1232 static struct ifaddr **ifa_list; 1233 static int ifa_list_size; 1234 1235 /* Depends on only one if_attach runs at once */ 1236 static void 1237 if_build_ifa_list(struct ifnet *ifp) 1238 { 1239 struct ifaddr *ifa; 1240 int i; 1241 1242 KASSERT(ifa_list == NULL); 1243 KASSERT(ifa_list_size == 0); 1244 1245 IFADDR_READER_FOREACH(ifa, ifp) 1246 ifa_list_size++; 1247 1248 ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP); 1249 i = 0; 1250 IFADDR_READER_FOREACH(ifa, ifp) { 1251 ifa_list[i++] = ifa; 1252 ifaref(ifa); 1253 } 1254 } 1255 1256 static void 1257 if_check_and_free_ifa_list(struct ifnet *ifp) 1258 { 1259 int i; 1260 struct ifaddr *ifa; 1261 1262 if (ifa_list == NULL) 1263 return; 1264 1265 for (i = 0; i < ifa_list_size; i++) { 1266 char buf[64]; 1267 1268 ifa = ifa_list[i]; 1269 sockaddr_format(ifa->ifa_addr, buf, sizeof(buf)); 1270 if (ifa->ifa_refcnt > 1) { 1271 log(LOG_WARNING, 1272 "ifa(%s) still referenced (refcnt=%d)\n", 1273 buf, ifa->ifa_refcnt - 1); 1274 } else 1275 log(LOG_DEBUG, 1276 "ifa(%s) not referenced (refcnt=%d)\n", 1277 buf, ifa->ifa_refcnt - 1); 1278 ifafree(ifa); 1279 } 1280 1281 kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size); 1282 ifa_list = NULL; 1283 ifa_list_size = 0; 1284 } 1285 #endif 1286 1287 /* 1288 * Detach an interface from the list of "active" interfaces, 1289 * freeing any resources as we go along. 1290 * 1291 * NOTE: This routine must be called with a valid thread context, 1292 * as it may block. 1293 */ 1294 void 1295 if_detach(struct ifnet *ifp) 1296 { 1297 struct socket so; 1298 struct ifaddr *ifa; 1299 #ifdef IFAREF_DEBUG 1300 struct ifaddr *last_ifa = NULL; 1301 #endif 1302 struct domain *dp; 1303 const struct protosw *pr; 1304 int s, i, family, purged; 1305 uint64_t xc; 1306 1307 #ifdef IFAREF_DEBUG 1308 if_build_ifa_list(ifp); 1309 #endif 1310 /* 1311 * XXX It's kind of lame that we have to have the 1312 * XXX socket structure... 1313 */ 1314 memset(&so, 0, sizeof(so)); 1315 1316 s = splnet(); 1317 1318 sysctl_teardown(&ifp->if_sysctl_log); 1319 IFNET_LOCK(ifp); 1320 if_deactivate(ifp); 1321 IFNET_UNLOCK(ifp); 1322 1323 /* 1324 * Unlink from the list and wait for all readers to leave 1325 * from pserialize read sections. Note that we can't do 1326 * psref_target_destroy here. See below. 1327 */ 1328 IFNET_GLOBAL_LOCK(); 1329 ifindex2ifnet[ifp->if_index] = NULL; 1330 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1331 IFNET_WRITER_REMOVE(ifp); 1332 pserialize_perform(ifnet_psz); 1333 IFNET_GLOBAL_UNLOCK(); 1334 1335 if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) { 1336 ifp->if_slowtimo = NULL; 1337 callout_halt(ifp->if_slowtimo_ch, NULL); 1338 callout_destroy(ifp->if_slowtimo_ch); 1339 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1340 } 1341 if_deferred_start_destroy(ifp); 1342 1343 /* 1344 * Do an if_down() to give protocols a chance to do something. 1345 */ 1346 if_down_deactivated(ifp); 1347 1348 #ifdef ALTQ 1349 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1350 altq_disable(&ifp->if_snd); 1351 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1352 altq_detach(&ifp->if_snd); 1353 #endif 1354 1355 #if NCARP > 0 1356 /* Remove the interface from any carp group it is a part of. */ 1357 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1358 carp_ifdetach(ifp); 1359 #endif 1360 1361 /* 1362 * Rip all the addresses off the interface. This should make 1363 * all of the routes go away. 1364 * 1365 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1366 * from the list, including our "cursor", ifa. For safety, 1367 * and to honor the TAILQ abstraction, I just restart the 1368 * loop after each removal. Note that the loop will exit 1369 * when all of the remaining ifaddrs belong to the AF_LINK 1370 * family. I am counting on the historical fact that at 1371 * least one pr_usrreq in each address domain removes at 1372 * least one ifaddr. 1373 */ 1374 again: 1375 /* 1376 * At this point, no other one tries to remove ifa in the list, 1377 * so we don't need to take a lock or psref. Avoid using 1378 * IFADDR_READER_FOREACH to pass over an inspection of contract 1379 * violations of pserialize. 1380 */ 1381 IFADDR_WRITER_FOREACH(ifa, ifp) { 1382 family = ifa->ifa_addr->sa_family; 1383 #ifdef IFAREF_DEBUG 1384 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1385 ifa, family, ifa->ifa_refcnt); 1386 if (last_ifa != NULL && ifa == last_ifa) 1387 panic("if_detach: loop detected"); 1388 last_ifa = ifa; 1389 #endif 1390 if (family == AF_LINK) 1391 continue; 1392 dp = pffinddomain(family); 1393 KASSERTMSG(dp != NULL, "no domain for AF %d", family); 1394 /* 1395 * XXX These PURGEIF calls are redundant with the 1396 * purge-all-families calls below, but are left in for 1397 * now both to make a smaller change, and to avoid 1398 * unplanned interactions with clearing of 1399 * ifp->if_addrlist. 1400 */ 1401 purged = 0; 1402 for (pr = dp->dom_protosw; 1403 pr < dp->dom_protoswNPROTOSW; pr++) { 1404 so.so_proto = pr; 1405 if (pr->pr_usrreqs) { 1406 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1407 purged = 1; 1408 } 1409 } 1410 if (purged == 0) { 1411 /* 1412 * XXX What's really the best thing to do 1413 * XXX here? --thorpej (at) NetBSD.org 1414 */ 1415 printf("if_detach: WARNING: AF %d not purged\n", 1416 family); 1417 ifa_remove(ifp, ifa); 1418 } 1419 goto again; 1420 } 1421 1422 if_free_sadl(ifp, 1); 1423 1424 restart: 1425 IFADDR_WRITER_FOREACH(ifa, ifp) { 1426 family = ifa->ifa_addr->sa_family; 1427 KASSERT(family == AF_LINK); 1428 ifa_remove(ifp, ifa); 1429 goto restart; 1430 } 1431 1432 /* Delete stray routes from the routing table. */ 1433 for (i = 0; i <= AF_MAX; i++) 1434 rt_delete_matched_entries(i, if_delroute_matcher, ifp); 1435 1436 DOMAIN_FOREACH(dp) { 1437 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1438 { 1439 void *p = ifp->if_afdata[dp->dom_family]; 1440 if (p) { 1441 ifp->if_afdata[dp->dom_family] = NULL; 1442 (*dp->dom_ifdetach)(ifp, p); 1443 } 1444 } 1445 1446 /* 1447 * One would expect multicast memberships (INET and 1448 * INET6) on UDP sockets to be purged by the PURGEIF 1449 * calls above, but if all addresses were removed from 1450 * the interface prior to destruction, the calls will 1451 * not be made (e.g. ppp, for which pppd(8) generally 1452 * removes addresses before destroying the interface). 1453 * Because there is no invariant that multicast 1454 * memberships only exist for interfaces with IPv4 1455 * addresses, we must call PURGEIF regardless of 1456 * addresses. (Protocols which might store ifnet 1457 * pointers are marked with PR_PURGEIF.) 1458 */ 1459 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1460 so.so_proto = pr; 1461 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1462 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1463 } 1464 } 1465 1466 /* 1467 * Must be done after the above pr_purgeif because if_psref may be 1468 * still used in pr_purgeif. 1469 */ 1470 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1471 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1472 1473 pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp); 1474 (void)pfil_head_destroy(ifp->if_pfil); 1475 1476 /* Announce that the interface is gone. */ 1477 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1478 1479 IF_AFDATA_LOCK_DESTROY(ifp); 1480 1481 if (if_is_link_state_changeable(ifp)) { 1482 softint_disestablish(ifp->if_link_si); 1483 ifp->if_link_si = NULL; 1484 } 1485 1486 /* 1487 * remove packets that came from ifp, from software interrupt queues. 1488 */ 1489 DOMAIN_FOREACH(dp) { 1490 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1491 struct ifqueue *iq = dp->dom_ifqueues[i]; 1492 if (iq == NULL) 1493 break; 1494 dp->dom_ifqueues[i] = NULL; 1495 if_detach_queues(ifp, iq); 1496 } 1497 } 1498 1499 /* 1500 * IP queues have to be processed separately: net-queue barrier 1501 * ensures that the packets are dequeued while a cross-call will 1502 * ensure that the interrupts have completed. FIXME: not quite.. 1503 */ 1504 #ifdef INET 1505 pktq_barrier(ip_pktq); 1506 #endif 1507 #ifdef INET6 1508 if (in6_present) 1509 pktq_barrier(ip6_pktq); 1510 #endif 1511 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1512 xc_wait(xc); 1513 1514 if (ifp->if_percpuq != NULL) { 1515 if_percpuq_destroy(ifp->if_percpuq); 1516 ifp->if_percpuq = NULL; 1517 } 1518 1519 mutex_obj_free(ifp->if_ioctl_lock); 1520 ifp->if_ioctl_lock = NULL; 1521 mutex_obj_free(ifp->if_snd.ifq_lock); 1522 1523 splx(s); 1524 1525 #ifdef IFAREF_DEBUG 1526 if_check_and_free_ifa_list(ifp); 1527 #endif 1528 } 1529 1530 static void 1531 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1532 { 1533 struct mbuf *m, *prev, *next; 1534 1535 prev = NULL; 1536 for (m = q->ifq_head; m != NULL; m = next) { 1537 KASSERT((m->m_flags & M_PKTHDR) != 0); 1538 1539 next = m->m_nextpkt; 1540 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1541 prev = m; 1542 continue; 1543 } 1544 1545 if (prev != NULL) 1546 prev->m_nextpkt = m->m_nextpkt; 1547 else 1548 q->ifq_head = m->m_nextpkt; 1549 if (q->ifq_tail == m) 1550 q->ifq_tail = prev; 1551 q->ifq_len--; 1552 1553 m->m_nextpkt = NULL; 1554 m_freem(m); 1555 IF_DROP(q); 1556 } 1557 } 1558 1559 /* 1560 * Callback for a radix tree walk to delete all references to an 1561 * ifnet. 1562 */ 1563 static int 1564 if_delroute_matcher(struct rtentry *rt, void *v) 1565 { 1566 struct ifnet *ifp = (struct ifnet *)v; 1567 1568 if (rt->rt_ifp == ifp) 1569 return 1; 1570 else 1571 return 0; 1572 } 1573 1574 /* 1575 * Create a clone network interface. 1576 */ 1577 static int 1578 if_clone_create(const char *name) 1579 { 1580 struct if_clone *ifc; 1581 int unit; 1582 struct ifnet *ifp; 1583 struct psref psref; 1584 1585 KASSERT(mutex_owned(&if_clone_mtx)); 1586 1587 ifc = if_clone_lookup(name, &unit); 1588 if (ifc == NULL) 1589 return EINVAL; 1590 1591 ifp = if_get(name, &psref); 1592 if (ifp != NULL) { 1593 if_put(ifp, &psref); 1594 return EEXIST; 1595 } 1596 1597 return (*ifc->ifc_create)(ifc, unit); 1598 } 1599 1600 /* 1601 * Destroy a clone network interface. 1602 */ 1603 static int 1604 if_clone_destroy(const char *name) 1605 { 1606 struct if_clone *ifc; 1607 struct ifnet *ifp; 1608 struct psref psref; 1609 1610 KASSERT(mutex_owned(&if_clone_mtx)); 1611 1612 ifc = if_clone_lookup(name, NULL); 1613 if (ifc == NULL) 1614 return EINVAL; 1615 1616 if (ifc->ifc_destroy == NULL) 1617 return EOPNOTSUPP; 1618 1619 ifp = if_get(name, &psref); 1620 if (ifp == NULL) 1621 return ENXIO; 1622 1623 /* We have to disable ioctls here */ 1624 IFNET_LOCK(ifp); 1625 ifp->if_ioctl = if_nullioctl; 1626 IFNET_UNLOCK(ifp); 1627 1628 /* 1629 * We cannot call ifc_destroy with holding ifp. 1630 * Releasing ifp here is safe thanks to if_clone_mtx. 1631 */ 1632 if_put(ifp, &psref); 1633 1634 return (*ifc->ifc_destroy)(ifp); 1635 } 1636 1637 static bool 1638 if_is_unit(const char *name) 1639 { 1640 1641 while(*name != '\0') { 1642 if (*name < '0' || *name > '9') 1643 return false; 1644 name++; 1645 } 1646 1647 return true; 1648 } 1649 1650 /* 1651 * Look up a network interface cloner. 1652 */ 1653 static struct if_clone * 1654 if_clone_lookup(const char *name, int *unitp) 1655 { 1656 struct if_clone *ifc; 1657 const char *cp; 1658 char *dp, ifname[IFNAMSIZ + 3]; 1659 int unit; 1660 1661 KASSERT(mutex_owned(&if_clone_mtx)); 1662 1663 strcpy(ifname, "if_"); 1664 /* separate interface name from unit */ 1665 /* TODO: search unit number from backward */ 1666 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1667 *cp && !if_is_unit(cp);) 1668 *dp++ = *cp++; 1669 1670 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1671 return NULL; /* No name or unit number */ 1672 *dp++ = '\0'; 1673 1674 again: 1675 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1676 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1677 break; 1678 } 1679 1680 if (ifc == NULL) { 1681 int error; 1682 if (*ifname == '\0') 1683 return NULL; 1684 mutex_exit(&if_clone_mtx); 1685 error = module_autoload(ifname, MODULE_CLASS_DRIVER); 1686 mutex_enter(&if_clone_mtx); 1687 if (error) 1688 return NULL; 1689 *ifname = '\0'; 1690 goto again; 1691 } 1692 1693 unit = 0; 1694 while (cp - name < IFNAMSIZ && *cp) { 1695 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1696 /* Bogus unit number. */ 1697 return NULL; 1698 } 1699 unit = (unit * 10) + (*cp++ - '0'); 1700 } 1701 1702 if (unitp != NULL) 1703 *unitp = unit; 1704 return ifc; 1705 } 1706 1707 /* 1708 * Register a network interface cloner. 1709 */ 1710 void 1711 if_clone_attach(struct if_clone *ifc) 1712 { 1713 1714 mutex_enter(&if_clone_mtx); 1715 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1716 if_cloners_count++; 1717 mutex_exit(&if_clone_mtx); 1718 } 1719 1720 /* 1721 * Unregister a network interface cloner. 1722 */ 1723 void 1724 if_clone_detach(struct if_clone *ifc) 1725 { 1726 1727 mutex_enter(&if_clone_mtx); 1728 LIST_REMOVE(ifc, ifc_list); 1729 if_cloners_count--; 1730 mutex_exit(&if_clone_mtx); 1731 } 1732 1733 /* 1734 * Provide list of interface cloners to userspace. 1735 */ 1736 int 1737 if_clone_list(int buf_count, char *buffer, int *total) 1738 { 1739 char outbuf[IFNAMSIZ], *dst; 1740 struct if_clone *ifc; 1741 int count, error = 0; 1742 1743 mutex_enter(&if_clone_mtx); 1744 *total = if_cloners_count; 1745 if ((dst = buffer) == NULL) { 1746 /* Just asking how many there are. */ 1747 goto out; 1748 } 1749 1750 if (buf_count < 0) { 1751 error = EINVAL; 1752 goto out; 1753 } 1754 1755 count = (if_cloners_count < buf_count) ? 1756 if_cloners_count : buf_count; 1757 1758 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1759 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1760 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1761 if (outbuf[sizeof(outbuf) - 1] != '\0') { 1762 error = ENAMETOOLONG; 1763 goto out; 1764 } 1765 error = copyout(outbuf, dst, sizeof(outbuf)); 1766 if (error != 0) 1767 break; 1768 } 1769 1770 out: 1771 mutex_exit(&if_clone_mtx); 1772 return error; 1773 } 1774 1775 void 1776 ifa_psref_init(struct ifaddr *ifa) 1777 { 1778 1779 psref_target_init(&ifa->ifa_psref, ifa_psref_class); 1780 } 1781 1782 void 1783 ifaref(struct ifaddr *ifa) 1784 { 1785 1786 atomic_inc_uint(&ifa->ifa_refcnt); 1787 } 1788 1789 void 1790 ifafree(struct ifaddr *ifa) 1791 { 1792 KASSERT(ifa != NULL); 1793 KASSERT(ifa->ifa_refcnt > 0); 1794 1795 if (atomic_dec_uint_nv(&ifa->ifa_refcnt) == 0) { 1796 free(ifa, M_IFADDR); 1797 } 1798 } 1799 1800 bool 1801 ifa_is_destroying(struct ifaddr *ifa) 1802 { 1803 1804 return ISSET(ifa->ifa_flags, IFA_DESTROYING); 1805 } 1806 1807 void 1808 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1809 { 1810 1811 ifa->ifa_ifp = ifp; 1812 1813 /* 1814 * Check MP-safety for IFEF_MPSAFE drivers. 1815 * Check !IFF_RUNNING for initialization routines that normally don't 1816 * take IFNET_LOCK but it's safe because there is no competitor. 1817 * XXX there are false positive cases because IFF_RUNNING can be off on 1818 * if_stop. 1819 */ 1820 KASSERT(!if_is_mpsafe(ifp) || !ISSET(ifp->if_flags, IFF_RUNNING) || 1821 IFNET_LOCKED(ifp)); 1822 1823 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1824 IFADDR_ENTRY_INIT(ifa); 1825 IFADDR_WRITER_INSERT_TAIL(ifp, ifa); 1826 1827 ifaref(ifa); 1828 } 1829 1830 void 1831 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1832 { 1833 1834 KASSERT(ifa->ifa_ifp == ifp); 1835 /* 1836 * Check MP-safety for IFEF_MPSAFE drivers. 1837 * if_is_deactivated indicates ifa_remove is called form if_detach 1838 * where is safe even if IFNET_LOCK isn't held. 1839 */ 1840 KASSERT(!if_is_mpsafe(ifp) || if_is_deactivated(ifp) || IFNET_LOCKED(ifp)); 1841 1842 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1843 IFADDR_WRITER_REMOVE(ifa); 1844 #ifdef NET_MPSAFE 1845 IFNET_GLOBAL_LOCK(); 1846 pserialize_perform(ifnet_psz); 1847 IFNET_GLOBAL_UNLOCK(); 1848 #endif 1849 1850 #ifdef NET_MPSAFE 1851 psref_target_destroy(&ifa->ifa_psref, ifa_psref_class); 1852 #endif 1853 IFADDR_ENTRY_DESTROY(ifa); 1854 ifafree(ifa); 1855 } 1856 1857 void 1858 ifa_acquire(struct ifaddr *ifa, struct psref *psref) 1859 { 1860 1861 psref_acquire(psref, &ifa->ifa_psref, ifa_psref_class); 1862 } 1863 1864 void 1865 ifa_release(struct ifaddr *ifa, struct psref *psref) 1866 { 1867 1868 if (ifa == NULL) 1869 return; 1870 1871 psref_release(psref, &ifa->ifa_psref, ifa_psref_class); 1872 } 1873 1874 bool 1875 ifa_held(struct ifaddr *ifa) 1876 { 1877 1878 return psref_held(&ifa->ifa_psref, ifa_psref_class); 1879 } 1880 1881 static inline int 1882 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1883 { 1884 return sockaddr_cmp(sa1, sa2) == 0; 1885 } 1886 1887 /* 1888 * Locate an interface based on a complete address. 1889 */ 1890 /*ARGSUSED*/ 1891 struct ifaddr * 1892 ifa_ifwithaddr(const struct sockaddr *addr) 1893 { 1894 struct ifnet *ifp; 1895 struct ifaddr *ifa; 1896 1897 IFNET_READER_FOREACH(ifp) { 1898 if (if_is_deactivated(ifp)) 1899 continue; 1900 IFADDR_READER_FOREACH(ifa, ifp) { 1901 if (ifa->ifa_addr->sa_family != addr->sa_family) 1902 continue; 1903 if (equal(addr, ifa->ifa_addr)) 1904 return ifa; 1905 if ((ifp->if_flags & IFF_BROADCAST) && 1906 ifa->ifa_broadaddr && 1907 /* IP6 doesn't have broadcast */ 1908 ifa->ifa_broadaddr->sa_len != 0 && 1909 equal(ifa->ifa_broadaddr, addr)) 1910 return ifa; 1911 } 1912 } 1913 return NULL; 1914 } 1915 1916 struct ifaddr * 1917 ifa_ifwithaddr_psref(const struct sockaddr *addr, struct psref *psref) 1918 { 1919 struct ifaddr *ifa; 1920 int s = pserialize_read_enter(); 1921 1922 ifa = ifa_ifwithaddr(addr); 1923 if (ifa != NULL) 1924 ifa_acquire(ifa, psref); 1925 pserialize_read_exit(s); 1926 1927 return ifa; 1928 } 1929 1930 /* 1931 * Locate the point to point interface with a given destination address. 1932 */ 1933 /*ARGSUSED*/ 1934 struct ifaddr * 1935 ifa_ifwithdstaddr(const struct sockaddr *addr) 1936 { 1937 struct ifnet *ifp; 1938 struct ifaddr *ifa; 1939 1940 IFNET_READER_FOREACH(ifp) { 1941 if (if_is_deactivated(ifp)) 1942 continue; 1943 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1944 continue; 1945 IFADDR_READER_FOREACH(ifa, ifp) { 1946 if (ifa->ifa_addr->sa_family != addr->sa_family || 1947 ifa->ifa_dstaddr == NULL) 1948 continue; 1949 if (equal(addr, ifa->ifa_dstaddr)) 1950 return ifa; 1951 } 1952 } 1953 1954 return NULL; 1955 } 1956 1957 struct ifaddr * 1958 ifa_ifwithdstaddr_psref(const struct sockaddr *addr, struct psref *psref) 1959 { 1960 struct ifaddr *ifa; 1961 int s; 1962 1963 s = pserialize_read_enter(); 1964 ifa = ifa_ifwithdstaddr(addr); 1965 if (ifa != NULL) 1966 ifa_acquire(ifa, psref); 1967 pserialize_read_exit(s); 1968 1969 return ifa; 1970 } 1971 1972 /* 1973 * Find an interface on a specific network. If many, choice 1974 * is most specific found. 1975 */ 1976 struct ifaddr * 1977 ifa_ifwithnet(const struct sockaddr *addr) 1978 { 1979 struct ifnet *ifp; 1980 struct ifaddr *ifa, *ifa_maybe = NULL; 1981 const struct sockaddr_dl *sdl; 1982 u_int af = addr->sa_family; 1983 const char *addr_data = addr->sa_data, *cplim; 1984 1985 if (af == AF_LINK) { 1986 sdl = satocsdl(addr); 1987 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1988 ifindex2ifnet[sdl->sdl_index] && 1989 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1990 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1991 } 1992 } 1993 #ifdef NETATALK 1994 if (af == AF_APPLETALK) { 1995 const struct sockaddr_at *sat, *sat2; 1996 sat = (const struct sockaddr_at *)addr; 1997 IFNET_READER_FOREACH(ifp) { 1998 if (if_is_deactivated(ifp)) 1999 continue; 2000 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 2001 if (ifa == NULL) 2002 continue; 2003 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 2004 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 2005 return ifa; /* exact match */ 2006 if (ifa_maybe == NULL) { 2007 /* else keep the if with the right range */ 2008 ifa_maybe = ifa; 2009 } 2010 } 2011 return ifa_maybe; 2012 } 2013 #endif 2014 IFNET_READER_FOREACH(ifp) { 2015 if (if_is_deactivated(ifp)) 2016 continue; 2017 IFADDR_READER_FOREACH(ifa, ifp) { 2018 const char *cp, *cp2, *cp3; 2019 2020 if (ifa->ifa_addr->sa_family != af || 2021 ifa->ifa_netmask == NULL) 2022 next: continue; 2023 cp = addr_data; 2024 cp2 = ifa->ifa_addr->sa_data; 2025 cp3 = ifa->ifa_netmask->sa_data; 2026 cplim = (const char *)ifa->ifa_netmask + 2027 ifa->ifa_netmask->sa_len; 2028 while (cp3 < cplim) { 2029 if ((*cp++ ^ *cp2++) & *cp3++) { 2030 /* want to continue for() loop */ 2031 goto next; 2032 } 2033 } 2034 if (ifa_maybe == NULL || 2035 rt_refines(ifa->ifa_netmask, 2036 ifa_maybe->ifa_netmask)) 2037 ifa_maybe = ifa; 2038 } 2039 } 2040 return ifa_maybe; 2041 } 2042 2043 struct ifaddr * 2044 ifa_ifwithnet_psref(const struct sockaddr *addr, struct psref *psref) 2045 { 2046 struct ifaddr *ifa; 2047 int s; 2048 2049 s = pserialize_read_enter(); 2050 ifa = ifa_ifwithnet(addr); 2051 if (ifa != NULL) 2052 ifa_acquire(ifa, psref); 2053 pserialize_read_exit(s); 2054 2055 return ifa; 2056 } 2057 2058 /* 2059 * Find the interface of the addresss. 2060 */ 2061 struct ifaddr * 2062 ifa_ifwithladdr(const struct sockaddr *addr) 2063 { 2064 struct ifaddr *ia; 2065 2066 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 2067 (ia = ifa_ifwithnet(addr))) 2068 return ia; 2069 return NULL; 2070 } 2071 2072 struct ifaddr * 2073 ifa_ifwithladdr_psref(const struct sockaddr *addr, struct psref *psref) 2074 { 2075 struct ifaddr *ifa; 2076 int s; 2077 2078 s = pserialize_read_enter(); 2079 ifa = ifa_ifwithladdr(addr); 2080 if (ifa != NULL) 2081 ifa_acquire(ifa, psref); 2082 pserialize_read_exit(s); 2083 2084 return ifa; 2085 } 2086 2087 /* 2088 * Find an interface using a specific address family 2089 */ 2090 struct ifaddr * 2091 ifa_ifwithaf(int af) 2092 { 2093 struct ifnet *ifp; 2094 struct ifaddr *ifa = NULL; 2095 int s; 2096 2097 s = pserialize_read_enter(); 2098 IFNET_READER_FOREACH(ifp) { 2099 if (if_is_deactivated(ifp)) 2100 continue; 2101 IFADDR_READER_FOREACH(ifa, ifp) { 2102 if (ifa->ifa_addr->sa_family == af) 2103 goto out; 2104 } 2105 } 2106 out: 2107 pserialize_read_exit(s); 2108 return ifa; 2109 } 2110 2111 /* 2112 * Find an interface address specific to an interface best matching 2113 * a given address. 2114 */ 2115 struct ifaddr * 2116 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2117 { 2118 struct ifaddr *ifa; 2119 const char *cp, *cp2, *cp3; 2120 const char *cplim; 2121 struct ifaddr *ifa_maybe = 0; 2122 u_int af = addr->sa_family; 2123 2124 if (if_is_deactivated(ifp)) 2125 return NULL; 2126 2127 if (af >= AF_MAX) 2128 return NULL; 2129 2130 IFADDR_READER_FOREACH(ifa, ifp) { 2131 if (ifa->ifa_addr->sa_family != af) 2132 continue; 2133 ifa_maybe = ifa; 2134 if (ifa->ifa_netmask == NULL) { 2135 if (equal(addr, ifa->ifa_addr) || 2136 (ifa->ifa_dstaddr && 2137 equal(addr, ifa->ifa_dstaddr))) 2138 return ifa; 2139 continue; 2140 } 2141 cp = addr->sa_data; 2142 cp2 = ifa->ifa_addr->sa_data; 2143 cp3 = ifa->ifa_netmask->sa_data; 2144 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2145 for (; cp3 < cplim; cp3++) { 2146 if ((*cp++ ^ *cp2++) & *cp3) 2147 break; 2148 } 2149 if (cp3 == cplim) 2150 return ifa; 2151 } 2152 return ifa_maybe; 2153 } 2154 2155 struct ifaddr * 2156 ifaof_ifpforaddr_psref(const struct sockaddr *addr, struct ifnet *ifp, 2157 struct psref *psref) 2158 { 2159 struct ifaddr *ifa; 2160 int s; 2161 2162 s = pserialize_read_enter(); 2163 ifa = ifaof_ifpforaddr(addr, ifp); 2164 if (ifa != NULL) 2165 ifa_acquire(ifa, psref); 2166 pserialize_read_exit(s); 2167 2168 return ifa; 2169 } 2170 2171 /* 2172 * Default action when installing a route with a Link Level gateway. 2173 * Lookup an appropriate real ifa to point to. 2174 * This should be moved to /sys/net/link.c eventually. 2175 */ 2176 void 2177 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 2178 { 2179 struct ifaddr *ifa; 2180 const struct sockaddr *dst; 2181 struct ifnet *ifp; 2182 struct psref psref; 2183 2184 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 2185 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL || 2186 ISSET(info->rti_flags, RTF_DONTCHANGEIFA)) 2187 return; 2188 if ((ifa = ifaof_ifpforaddr_psref(dst, ifp, &psref)) != NULL) { 2189 rt_replace_ifa(rt, ifa); 2190 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2191 ifa->ifa_rtrequest(cmd, rt, info); 2192 ifa_release(ifa, &psref); 2193 } 2194 } 2195 2196 /* 2197 * bitmask macros to manage a densely packed link_state change queue. 2198 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 2199 * LINK_STATE_UP(2) we need 2 bits for each state change. 2200 * As a state change to store is 0, treat all bits set as an unset item. 2201 */ 2202 #define LQ_ITEM_BITS 2 2203 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 2204 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 2205 #define LINK_STATE_UNSET LQ_ITEM_MASK 2206 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 2207 #define LQ_STORE(q, i, v) \ 2208 do { \ 2209 (q) &= ~LQ_MASK((i)); \ 2210 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 2211 } while (0 /* CONSTCOND */) 2212 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 2213 #define LQ_POP(q, v) \ 2214 do { \ 2215 (v) = LQ_ITEM((q), 0); \ 2216 (q) >>= LQ_ITEM_BITS; \ 2217 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2218 } while (0 /* CONSTCOND */) 2219 #define LQ_PUSH(q, v) \ 2220 do { \ 2221 (q) >>= LQ_ITEM_BITS; \ 2222 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2223 } while (0 /* CONSTCOND */) 2224 #define LQ_FIND_UNSET(q, i) \ 2225 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 2226 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 2227 break; \ 2228 } 2229 2230 /* 2231 * XXX reusing (ifp)->if_snd->ifq_lock rather than having another spin mutex 2232 * for each ifnet. It doesn't matter because: 2233 * - if IFEF_MPSAFE is enabled, if_snd isn't used and lock contentions on 2234 * ifq_lock don't happen 2235 * - if IFEF_MPSAFE is disabled, there is no lock contention on ifq_lock 2236 * because if_snd, if_link_state_change and if_link_state_change_softint 2237 * are all called with KERNEL_LOCK 2238 */ 2239 #define IF_LINK_STATE_CHANGE_LOCK(ifp) \ 2240 mutex_enter((ifp)->if_snd.ifq_lock) 2241 #define IF_LINK_STATE_CHANGE_UNLOCK(ifp) \ 2242 mutex_exit((ifp)->if_snd.ifq_lock) 2243 2244 /* 2245 * Handle a change in the interface link state and 2246 * queue notifications. 2247 */ 2248 void 2249 if_link_state_change(struct ifnet *ifp, int link_state) 2250 { 2251 int idx; 2252 2253 KASSERTMSG(if_is_link_state_changeable(ifp), 2254 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 2255 ifp->if_xname, ifp->if_extflags); 2256 2257 /* Ensure change is to a valid state */ 2258 switch (link_state) { 2259 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 2260 case LINK_STATE_DOWN: /* FALLTHROUGH */ 2261 case LINK_STATE_UP: 2262 break; 2263 default: 2264 #ifdef DEBUG 2265 printf("%s: invalid link state %d\n", 2266 ifp->if_xname, link_state); 2267 #endif 2268 return; 2269 } 2270 2271 IF_LINK_STATE_CHANGE_LOCK(ifp); 2272 2273 /* Find the last unset event in the queue. */ 2274 LQ_FIND_UNSET(ifp->if_link_queue, idx); 2275 2276 /* 2277 * Ensure link_state doesn't match the last event in the queue. 2278 * ifp->if_link_state is not checked and set here because 2279 * that would present an inconsistent picture to the system. 2280 */ 2281 if (idx != 0 && 2282 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 2283 goto out; 2284 2285 /* Handle queue overflow. */ 2286 if (idx == LQ_MAX(ifp->if_link_queue)) { 2287 uint8_t lost; 2288 2289 /* 2290 * The DOWN state must be protected from being pushed off 2291 * the queue to ensure that userland will always be 2292 * in a sane state. 2293 * Because DOWN is protected, there is no need to protect 2294 * UNKNOWN. 2295 * It should be invalid to change from any other state to 2296 * UNKNOWN anyway ... 2297 */ 2298 lost = LQ_ITEM(ifp->if_link_queue, 0); 2299 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 2300 if (lost == LINK_STATE_DOWN) { 2301 lost = LQ_ITEM(ifp->if_link_queue, 0); 2302 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 2303 } 2304 printf("%s: lost link state change %s\n", 2305 ifp->if_xname, 2306 lost == LINK_STATE_UP ? "UP" : 2307 lost == LINK_STATE_DOWN ? "DOWN" : 2308 "UNKNOWN"); 2309 } else 2310 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 2311 2312 softint_schedule(ifp->if_link_si); 2313 2314 out: 2315 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2316 } 2317 2318 /* 2319 * Handle interface link state change notifications. 2320 */ 2321 void 2322 if_link_state_change_softint(struct ifnet *ifp, int link_state) 2323 { 2324 struct domain *dp; 2325 int s = splnet(); 2326 bool notify; 2327 2328 KASSERT(!cpu_intr_p()); 2329 2330 IF_LINK_STATE_CHANGE_LOCK(ifp); 2331 2332 /* Ensure the change is still valid. */ 2333 if (ifp->if_link_state == link_state) { 2334 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2335 splx(s); 2336 return; 2337 } 2338 2339 #ifdef DEBUG 2340 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 2341 link_state == LINK_STATE_UP ? "UP" : 2342 link_state == LINK_STATE_DOWN ? "DOWN" : 2343 "UNKNOWN", 2344 ifp->if_link_state == LINK_STATE_UP ? "UP" : 2345 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 2346 "UNKNOWN"); 2347 #endif 2348 2349 /* 2350 * When going from UNKNOWN to UP, we need to mark existing 2351 * addresses as tentative and restart DAD as we may have 2352 * erroneously not found a duplicate. 2353 * 2354 * This needs to happen before rt_ifmsg to avoid a race where 2355 * listeners would have an address and expect it to work right 2356 * away. 2357 */ 2358 notify = (link_state == LINK_STATE_UP && 2359 ifp->if_link_state == LINK_STATE_UNKNOWN); 2360 ifp->if_link_state = link_state; 2361 /* The following routines may sleep so release the spin mutex */ 2362 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2363 2364 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 2365 if (notify) { 2366 DOMAIN_FOREACH(dp) { 2367 if (dp->dom_if_link_state_change != NULL) 2368 dp->dom_if_link_state_change(ifp, 2369 LINK_STATE_DOWN); 2370 } 2371 } 2372 2373 /* Notify that the link state has changed. */ 2374 rt_ifmsg(ifp); 2375 2376 #if NCARP > 0 2377 if (ifp->if_carp) 2378 carp_carpdev_state(ifp); 2379 #endif 2380 2381 DOMAIN_FOREACH(dp) { 2382 if (dp->dom_if_link_state_change != NULL) 2383 dp->dom_if_link_state_change(ifp, link_state); 2384 } 2385 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 2386 splx(s); 2387 } 2388 2389 /* 2390 * Process the interface link state change queue. 2391 */ 2392 static void 2393 if_link_state_change_si(void *arg) 2394 { 2395 struct ifnet *ifp = arg; 2396 int s; 2397 uint8_t state; 2398 bool schedule; 2399 2400 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 2401 s = splnet(); 2402 2403 /* Pop a link state change from the queue and process it. */ 2404 IF_LINK_STATE_CHANGE_LOCK(ifp); 2405 LQ_POP(ifp->if_link_queue, state); 2406 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2407 2408 if_link_state_change_softint(ifp, state); 2409 2410 /* If there is a link state change to come, schedule it. */ 2411 IF_LINK_STATE_CHANGE_LOCK(ifp); 2412 schedule = (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET); 2413 IF_LINK_STATE_CHANGE_UNLOCK(ifp); 2414 if (schedule) 2415 softint_schedule(ifp->if_link_si); 2416 2417 splx(s); 2418 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 2419 } 2420 2421 /* 2422 * Default action when installing a local route on a point-to-point 2423 * interface. 2424 */ 2425 void 2426 p2p_rtrequest(int req, struct rtentry *rt, 2427 __unused const struct rt_addrinfo *info) 2428 { 2429 struct ifnet *ifp = rt->rt_ifp; 2430 struct ifaddr *ifa, *lo0ifa; 2431 int s = pserialize_read_enter(); 2432 2433 switch (req) { 2434 case RTM_ADD: 2435 if ((rt->rt_flags & RTF_LOCAL) == 0) 2436 break; 2437 2438 rt->rt_ifp = lo0ifp; 2439 2440 if (ISSET(info->rti_flags, RTF_DONTCHANGEIFA)) 2441 break; 2442 2443 IFADDR_READER_FOREACH(ifa, ifp) { 2444 if (equal(rt_getkey(rt), ifa->ifa_addr)) 2445 break; 2446 } 2447 if (ifa == NULL) 2448 break; 2449 2450 /* 2451 * Ensure lo0 has an address of the same family. 2452 */ 2453 IFADDR_READER_FOREACH(lo0ifa, lo0ifp) { 2454 if (lo0ifa->ifa_addr->sa_family == 2455 ifa->ifa_addr->sa_family) 2456 break; 2457 } 2458 if (lo0ifa == NULL) 2459 break; 2460 2461 /* 2462 * Make sure to set rt->rt_ifa to the interface 2463 * address we are using, otherwise we will have trouble 2464 * with source address selection. 2465 */ 2466 if (ifa != rt->rt_ifa) 2467 rt_replace_ifa(rt, ifa); 2468 break; 2469 case RTM_DELETE: 2470 default: 2471 break; 2472 } 2473 pserialize_read_exit(s); 2474 } 2475 2476 static void 2477 _if_down(struct ifnet *ifp) 2478 { 2479 struct ifaddr *ifa; 2480 struct domain *dp; 2481 int s, bound; 2482 struct psref psref; 2483 2484 ifp->if_flags &= ~IFF_UP; 2485 nanotime(&ifp->if_lastchange); 2486 2487 bound = curlwp_bind(); 2488 s = pserialize_read_enter(); 2489 IFADDR_READER_FOREACH(ifa, ifp) { 2490 ifa_acquire(ifa, &psref); 2491 pserialize_read_exit(s); 2492 2493 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2494 2495 s = pserialize_read_enter(); 2496 ifa_release(ifa, &psref); 2497 } 2498 pserialize_read_exit(s); 2499 curlwp_bindx(bound); 2500 2501 IFQ_PURGE(&ifp->if_snd); 2502 #if NCARP > 0 2503 if (ifp->if_carp) 2504 carp_carpdev_state(ifp); 2505 #endif 2506 rt_ifmsg(ifp); 2507 DOMAIN_FOREACH(dp) { 2508 if (dp->dom_if_down) 2509 dp->dom_if_down(ifp); 2510 } 2511 } 2512 2513 static void 2514 if_down_deactivated(struct ifnet *ifp) 2515 { 2516 2517 KASSERT(if_is_deactivated(ifp)); 2518 _if_down(ifp); 2519 } 2520 2521 void 2522 if_down_locked(struct ifnet *ifp) 2523 { 2524 2525 KASSERT(IFNET_LOCKED(ifp)); 2526 _if_down(ifp); 2527 } 2528 2529 /* 2530 * Mark an interface down and notify protocols of 2531 * the transition. 2532 * NOTE: must be called at splsoftnet or equivalent. 2533 */ 2534 void 2535 if_down(struct ifnet *ifp) 2536 { 2537 2538 IFNET_LOCK(ifp); 2539 if_down_locked(ifp); 2540 IFNET_UNLOCK(ifp); 2541 } 2542 2543 /* 2544 * Must be called with holding if_ioctl_lock. 2545 */ 2546 static void 2547 if_up_locked(struct ifnet *ifp) 2548 { 2549 #ifdef notyet 2550 struct ifaddr *ifa; 2551 #endif 2552 struct domain *dp; 2553 2554 KASSERT(IFNET_LOCKED(ifp)); 2555 2556 KASSERT(!if_is_deactivated(ifp)); 2557 ifp->if_flags |= IFF_UP; 2558 nanotime(&ifp->if_lastchange); 2559 #ifdef notyet 2560 /* this has no effect on IP, and will kill all ISO connections XXX */ 2561 IFADDR_READER_FOREACH(ifa, ifp) 2562 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2563 #endif 2564 #if NCARP > 0 2565 if (ifp->if_carp) 2566 carp_carpdev_state(ifp); 2567 #endif 2568 rt_ifmsg(ifp); 2569 DOMAIN_FOREACH(dp) { 2570 if (dp->dom_if_up) 2571 dp->dom_if_up(ifp); 2572 } 2573 } 2574 2575 /* 2576 * Handle interface slowtimo timer routine. Called 2577 * from softclock, we decrement timer (if set) and 2578 * call the appropriate interface routine on expiration. 2579 */ 2580 static void 2581 if_slowtimo(void *arg) 2582 { 2583 void (*slowtimo)(struct ifnet *); 2584 struct ifnet *ifp = arg; 2585 int s; 2586 2587 slowtimo = ifp->if_slowtimo; 2588 if (__predict_false(slowtimo == NULL)) 2589 return; 2590 2591 s = splnet(); 2592 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2593 (*slowtimo)(ifp); 2594 2595 splx(s); 2596 2597 if (__predict_true(ifp->if_slowtimo != NULL)) 2598 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2599 } 2600 2601 /* 2602 * Mark an interface up and notify protocols of 2603 * the transition. 2604 * NOTE: must be called at splsoftnet or equivalent. 2605 */ 2606 void 2607 if_up(struct ifnet *ifp) 2608 { 2609 2610 IFNET_LOCK(ifp); 2611 if_up_locked(ifp); 2612 IFNET_UNLOCK(ifp); 2613 } 2614 2615 /* 2616 * Set/clear promiscuous mode on interface ifp based on the truth value 2617 * of pswitch. The calls are reference counted so that only the first 2618 * "on" request actually has an effect, as does the final "off" request. 2619 * Results are undefined if the "off" and "on" requests are not matched. 2620 */ 2621 int 2622 ifpromisc_locked(struct ifnet *ifp, int pswitch) 2623 { 2624 int pcount, ret = 0; 2625 short nflags; 2626 2627 KASSERT(IFNET_LOCKED(ifp)); 2628 2629 pcount = ifp->if_pcount; 2630 if (pswitch) { 2631 /* 2632 * Allow the device to be "placed" into promiscuous 2633 * mode even if it is not configured up. It will 2634 * consult IFF_PROMISC when it is brought up. 2635 */ 2636 if (ifp->if_pcount++ != 0) 2637 goto out; 2638 nflags = ifp->if_flags | IFF_PROMISC; 2639 } else { 2640 if (--ifp->if_pcount > 0) 2641 goto out; 2642 nflags = ifp->if_flags & ~IFF_PROMISC; 2643 } 2644 ret = if_flags_set(ifp, nflags); 2645 /* Restore interface state if not successful. */ 2646 if (ret != 0) { 2647 ifp->if_pcount = pcount; 2648 } 2649 out: 2650 return ret; 2651 } 2652 2653 int 2654 ifpromisc(struct ifnet *ifp, int pswitch) 2655 { 2656 int e; 2657 2658 IFNET_LOCK(ifp); 2659 e = ifpromisc_locked(ifp, pswitch); 2660 IFNET_UNLOCK(ifp); 2661 2662 return e; 2663 } 2664 2665 /* 2666 * Map interface name to 2667 * interface structure pointer. 2668 */ 2669 struct ifnet * 2670 ifunit(const char *name) 2671 { 2672 struct ifnet *ifp; 2673 const char *cp = name; 2674 u_int unit = 0; 2675 u_int i; 2676 int s; 2677 2678 /* 2679 * If the entire name is a number, treat it as an ifindex. 2680 */ 2681 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2682 unit = unit * 10 + (*cp - '0'); 2683 } 2684 2685 /* 2686 * If the number took all of the name, then it's a valid ifindex. 2687 */ 2688 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) 2689 return if_byindex(unit); 2690 2691 ifp = NULL; 2692 s = pserialize_read_enter(); 2693 IFNET_READER_FOREACH(ifp) { 2694 if (if_is_deactivated(ifp)) 2695 continue; 2696 if (strcmp(ifp->if_xname, name) == 0) 2697 goto out; 2698 } 2699 out: 2700 pserialize_read_exit(s); 2701 return ifp; 2702 } 2703 2704 /* 2705 * Get a reference of an ifnet object by an interface name. 2706 * The returned reference is protected by psref(9). The caller 2707 * must release a returned reference by if_put after use. 2708 */ 2709 struct ifnet * 2710 if_get(const char *name, struct psref *psref) 2711 { 2712 struct ifnet *ifp; 2713 const char *cp = name; 2714 u_int unit = 0; 2715 u_int i; 2716 int s; 2717 2718 /* 2719 * If the entire name is a number, treat it as an ifindex. 2720 */ 2721 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2722 unit = unit * 10 + (*cp - '0'); 2723 } 2724 2725 /* 2726 * If the number took all of the name, then it's a valid ifindex. 2727 */ 2728 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) 2729 return if_get_byindex(unit, psref); 2730 2731 ifp = NULL; 2732 s = pserialize_read_enter(); 2733 IFNET_READER_FOREACH(ifp) { 2734 if (if_is_deactivated(ifp)) 2735 continue; 2736 if (strcmp(ifp->if_xname, name) == 0) { 2737 psref_acquire(psref, &ifp->if_psref, 2738 ifnet_psref_class); 2739 goto out; 2740 } 2741 } 2742 out: 2743 pserialize_read_exit(s); 2744 return ifp; 2745 } 2746 2747 /* 2748 * Release a reference of an ifnet object given by if_get, if_get_byindex 2749 * or if_get_bylla. 2750 */ 2751 void 2752 if_put(const struct ifnet *ifp, struct psref *psref) 2753 { 2754 2755 if (ifp == NULL) 2756 return; 2757 2758 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2759 } 2760 2761 /* 2762 * Return ifp having idx. Return NULL if not found. Normally if_byindex 2763 * should be used. 2764 */ 2765 ifnet_t * 2766 _if_byindex(u_int idx) 2767 { 2768 2769 return (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2770 } 2771 2772 /* 2773 * Return ifp having idx. Return NULL if not found or the found ifp is 2774 * already deactivated. 2775 */ 2776 ifnet_t * 2777 if_byindex(u_int idx) 2778 { 2779 ifnet_t *ifp; 2780 2781 ifp = _if_byindex(idx); 2782 if (ifp != NULL && if_is_deactivated(ifp)) 2783 ifp = NULL; 2784 return ifp; 2785 } 2786 2787 /* 2788 * Get a reference of an ifnet object by an interface index. 2789 * The returned reference is protected by psref(9). The caller 2790 * must release a returned reference by if_put after use. 2791 */ 2792 ifnet_t * 2793 if_get_byindex(u_int idx, struct psref *psref) 2794 { 2795 ifnet_t *ifp; 2796 int s; 2797 2798 s = pserialize_read_enter(); 2799 ifp = if_byindex(idx); 2800 if (__predict_true(ifp != NULL)) 2801 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2802 pserialize_read_exit(s); 2803 2804 return ifp; 2805 } 2806 2807 ifnet_t * 2808 if_get_bylla(const void *lla, unsigned char lla_len, struct psref *psref) 2809 { 2810 ifnet_t *ifp; 2811 int s; 2812 2813 s = pserialize_read_enter(); 2814 IFNET_READER_FOREACH(ifp) { 2815 if (if_is_deactivated(ifp)) 2816 continue; 2817 if (ifp->if_addrlen != lla_len) 2818 continue; 2819 if (memcmp(lla, CLLADDR(ifp->if_sadl), lla_len) == 0) { 2820 psref_acquire(psref, &ifp->if_psref, 2821 ifnet_psref_class); 2822 break; 2823 } 2824 } 2825 pserialize_read_exit(s); 2826 2827 return ifp; 2828 } 2829 2830 /* 2831 * Note that it's safe only if the passed ifp is guaranteed to not be freed, 2832 * for example using pserialize or the ifp is already held or some other 2833 * object is held which guarantes the ifp to not be freed indirectly. 2834 */ 2835 void 2836 if_acquire(struct ifnet *ifp, struct psref *psref) 2837 { 2838 2839 KASSERT(ifp->if_index != 0); 2840 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2841 } 2842 2843 bool 2844 if_held(struct ifnet *ifp) 2845 { 2846 2847 return psref_held(&ifp->if_psref, ifnet_psref_class); 2848 } 2849 2850 /* 2851 * Some tunnel interfaces can nest, e.g. IPv4 over IPv4 gif(4) tunnel over IPv4. 2852 * Check the tunnel nesting count. 2853 * Return > 0, if tunnel nesting count is more than limit. 2854 * Return 0, if tunnel nesting count is equal or less than limit. 2855 */ 2856 int 2857 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, int limit) 2858 { 2859 struct m_tag *mtag; 2860 int *count; 2861 2862 mtag = m_tag_find(m, PACKET_TAG_TUNNEL_INFO); 2863 if (mtag != NULL) { 2864 count = (int *)(mtag + 1); 2865 if (++(*count) > limit) { 2866 log(LOG_NOTICE, 2867 "%s: recursively called too many times(%d)\n", 2868 ifp->if_xname, *count); 2869 return EIO; 2870 } 2871 } else { 2872 mtag = m_tag_get(PACKET_TAG_TUNNEL_INFO, sizeof(*count), 2873 M_NOWAIT); 2874 if (mtag != NULL) { 2875 m_tag_prepend(m, mtag); 2876 count = (int *)(mtag + 1); 2877 *count = 0; 2878 } else { 2879 log(LOG_DEBUG, 2880 "%s: m_tag_get() failed, recursion calls are not prevented.\n", 2881 ifp->if_xname); 2882 } 2883 } 2884 2885 return 0; 2886 } 2887 2888 /* common */ 2889 int 2890 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2891 { 2892 int s; 2893 struct ifreq *ifr; 2894 struct ifcapreq *ifcr; 2895 struct ifdatareq *ifdr; 2896 2897 switch (cmd) { 2898 case SIOCSIFCAP: 2899 ifcr = data; 2900 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2901 return EINVAL; 2902 2903 if (ifcr->ifcr_capenable == ifp->if_capenable) 2904 return 0; 2905 2906 ifp->if_capenable = ifcr->ifcr_capenable; 2907 2908 /* Pre-compute the checksum flags mask. */ 2909 ifp->if_csum_flags_tx = 0; 2910 ifp->if_csum_flags_rx = 0; 2911 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) 2912 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2913 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) 2914 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2915 2916 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) 2917 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2918 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) 2919 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2920 2921 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) 2922 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2923 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) 2924 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2925 2926 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) 2927 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2928 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) 2929 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2930 2931 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) 2932 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2933 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) 2934 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2935 2936 if (ifp->if_capenable & IFCAP_TSOv4) 2937 ifp->if_csum_flags_tx |= M_CSUM_TSOv4; 2938 if (ifp->if_capenable & IFCAP_TSOv6) 2939 ifp->if_csum_flags_tx |= M_CSUM_TSOv6; 2940 2941 #if NBRIDGE > 0 2942 if (ifp->if_bridge != NULL) 2943 bridge_calc_csum_flags(ifp->if_bridge); 2944 #endif 2945 2946 if (ifp->if_flags & IFF_UP) 2947 return ENETRESET; 2948 return 0; 2949 case SIOCSIFFLAGS: 2950 ifr = data; 2951 /* 2952 * If if_is_mpsafe(ifp), KERNEL_LOCK isn't held here, but if_up 2953 * and if_down aren't MP-safe yet, so we must hold the lock. 2954 */ 2955 KERNEL_LOCK_IF_IFP_MPSAFE(ifp); 2956 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2957 s = splsoftnet(); 2958 if_down_locked(ifp); 2959 splx(s); 2960 } 2961 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2962 s = splsoftnet(); 2963 if_up_locked(ifp); 2964 splx(s); 2965 } 2966 KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp); 2967 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2968 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2969 break; 2970 case SIOCGIFFLAGS: 2971 ifr = data; 2972 ifr->ifr_flags = ifp->if_flags; 2973 break; 2974 2975 case SIOCGIFMETRIC: 2976 ifr = data; 2977 ifr->ifr_metric = ifp->if_metric; 2978 break; 2979 2980 case SIOCGIFMTU: 2981 ifr = data; 2982 ifr->ifr_mtu = ifp->if_mtu; 2983 break; 2984 2985 case SIOCGIFDLT: 2986 ifr = data; 2987 ifr->ifr_dlt = ifp->if_dlt; 2988 break; 2989 2990 case SIOCGIFCAP: 2991 ifcr = data; 2992 ifcr->ifcr_capabilities = ifp->if_capabilities; 2993 ifcr->ifcr_capenable = ifp->if_capenable; 2994 break; 2995 2996 case SIOCSIFMETRIC: 2997 ifr = data; 2998 ifp->if_metric = ifr->ifr_metric; 2999 break; 3000 3001 case SIOCGIFDATA: 3002 ifdr = data; 3003 ifdr->ifdr_data = ifp->if_data; 3004 break; 3005 3006 case SIOCGIFINDEX: 3007 ifr = data; 3008 ifr->ifr_index = ifp->if_index; 3009 break; 3010 3011 case SIOCZIFDATA: 3012 ifdr = data; 3013 ifdr->ifdr_data = ifp->if_data; 3014 /* 3015 * Assumes that the volatile counters that can be 3016 * zero'ed are at the end of if_data. 3017 */ 3018 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 3019 offsetof(struct if_data, ifi_ipackets)); 3020 /* 3021 * The memset() clears to the bottm of if_data. In the area, 3022 * if_lastchange is included. Please be careful if new entry 3023 * will be added into if_data or rewite this. 3024 * 3025 * And also, update if_lastchnage. 3026 */ 3027 getnanotime(&ifp->if_lastchange); 3028 break; 3029 case SIOCSIFMTU: 3030 ifr = data; 3031 if (ifp->if_mtu == ifr->ifr_mtu) 3032 break; 3033 ifp->if_mtu = ifr->ifr_mtu; 3034 /* 3035 * If the link MTU changed, do network layer specific procedure. 3036 */ 3037 #ifdef INET6 3038 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 3039 if (in6_present) 3040 nd6_setmtu(ifp); 3041 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 3042 #endif 3043 return ENETRESET; 3044 default: 3045 return ENOTTY; 3046 } 3047 return 0; 3048 } 3049 3050 int 3051 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 3052 { 3053 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 3054 struct ifaddr *ifa; 3055 const struct sockaddr *any, *sa; 3056 union { 3057 struct sockaddr sa; 3058 struct sockaddr_storage ss; 3059 } u, v; 3060 int s, error = 0; 3061 3062 switch (cmd) { 3063 case SIOCSIFADDRPREF: 3064 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 3065 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 3066 NULL) != 0) 3067 return EPERM; 3068 case SIOCGIFADDRPREF: 3069 break; 3070 default: 3071 return EOPNOTSUPP; 3072 } 3073 3074 /* sanity checks */ 3075 if (data == NULL || ifp == NULL) { 3076 panic("invalid argument to %s", __func__); 3077 /*NOTREACHED*/ 3078 } 3079 3080 /* address must be specified on ADD and DELETE */ 3081 sa = sstocsa(&ifap->ifap_addr); 3082 if (sa->sa_family != sofamily(so)) 3083 return EINVAL; 3084 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 3085 return EINVAL; 3086 3087 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 3088 3089 s = pserialize_read_enter(); 3090 IFADDR_READER_FOREACH(ifa, ifp) { 3091 if (ifa->ifa_addr->sa_family != sa->sa_family) 3092 continue; 3093 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 3094 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 3095 break; 3096 } 3097 if (ifa == NULL) { 3098 error = EADDRNOTAVAIL; 3099 goto out; 3100 } 3101 3102 switch (cmd) { 3103 case SIOCSIFADDRPREF: 3104 ifa->ifa_preference = ifap->ifap_preference; 3105 goto out; 3106 case SIOCGIFADDRPREF: 3107 /* fill in the if_laddrreq structure */ 3108 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 3109 sizeof(ifap->ifap_addr), ifa->ifa_addr); 3110 ifap->ifap_preference = ifa->ifa_preference; 3111 goto out; 3112 default: 3113 error = EOPNOTSUPP; 3114 } 3115 out: 3116 pserialize_read_exit(s); 3117 return error; 3118 } 3119 3120 /* 3121 * Interface ioctls. 3122 */ 3123 static int 3124 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 3125 { 3126 struct ifnet *ifp; 3127 struct ifreq *ifr; 3128 int error = 0, hook; 3129 u_long ocmd = cmd; 3130 short oif_flags; 3131 struct ifreq ifrb; 3132 struct oifreq *oifr = NULL; 3133 int r; 3134 struct psref psref; 3135 int bound; 3136 3137 switch (cmd) { 3138 case SIOCGIFCONF: 3139 return ifconf(cmd, data); 3140 case SIOCINITIFADDR: 3141 return EPERM; 3142 default: 3143 MODULE_HOOK_CALL(uipc_syscalls_40_hook, (cmd, data), enosys(), 3144 error); 3145 if (error != ENOSYS) 3146 return error; 3147 MODULE_HOOK_CALL(uipc_syscalls_50_hook, (l, cmd, data), 3148 enosys(), error); 3149 if (error != ENOSYS) 3150 return error; 3151 error = 0; 3152 break; 3153 } 3154 3155 ifr = data; 3156 MODULE_HOOK_CALL(if_cvtcmd_43_hook, (&cmd, ocmd), enosys(), hook); 3157 if (hook != ENOSYS) { 3158 if (cmd != ocmd) { 3159 oifr = data; 3160 data = ifr = &ifrb; 3161 ifreqo2n(oifr, ifr); 3162 } 3163 } 3164 3165 switch (cmd) { 3166 case SIOCIFCREATE: 3167 case SIOCIFDESTROY: 3168 bound = curlwp_bind(); 3169 if (l != NULL) { 3170 ifp = if_get(ifr->ifr_name, &psref); 3171 error = kauth_authorize_network(l->l_cred, 3172 KAUTH_NETWORK_INTERFACE, 3173 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 3174 (void *)cmd, NULL); 3175 if (ifp != NULL) 3176 if_put(ifp, &psref); 3177 if (error != 0) { 3178 curlwp_bindx(bound); 3179 return error; 3180 } 3181 } 3182 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 3183 mutex_enter(&if_clone_mtx); 3184 r = (cmd == SIOCIFCREATE) ? 3185 if_clone_create(ifr->ifr_name) : 3186 if_clone_destroy(ifr->ifr_name); 3187 mutex_exit(&if_clone_mtx); 3188 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 3189 curlwp_bindx(bound); 3190 return r; 3191 3192 case SIOCIFGCLONERS: 3193 { 3194 struct if_clonereq *req = (struct if_clonereq *)data; 3195 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 3196 &req->ifcr_total); 3197 } 3198 } 3199 3200 bound = curlwp_bind(); 3201 ifp = if_get(ifr->ifr_name, &psref); 3202 if (ifp == NULL) { 3203 curlwp_bindx(bound); 3204 return ENXIO; 3205 } 3206 3207 switch (cmd) { 3208 case SIOCALIFADDR: 3209 case SIOCDLIFADDR: 3210 case SIOCSIFADDRPREF: 3211 case SIOCSIFFLAGS: 3212 case SIOCSIFCAP: 3213 case SIOCSIFMETRIC: 3214 case SIOCZIFDATA: 3215 case SIOCSIFMTU: 3216 case SIOCSIFPHYADDR: 3217 case SIOCDIFPHYADDR: 3218 #ifdef INET6 3219 case SIOCSIFPHYADDR_IN6: 3220 #endif 3221 case SIOCSLIFPHYADDR: 3222 case SIOCADDMULTI: 3223 case SIOCDELMULTI: 3224 case SIOCSETHERCAP: 3225 case SIOCSIFMEDIA: 3226 case SIOCSDRVSPEC: 3227 case SIOCG80211: 3228 case SIOCS80211: 3229 case SIOCS80211NWID: 3230 case SIOCS80211NWKEY: 3231 case SIOCS80211POWER: 3232 case SIOCS80211BSSID: 3233 case SIOCS80211CHANNEL: 3234 case SIOCSLINKSTR: 3235 if (l != NULL) { 3236 error = kauth_authorize_network(l->l_cred, 3237 KAUTH_NETWORK_INTERFACE, 3238 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 3239 (void *)cmd, NULL); 3240 if (error != 0) 3241 goto out; 3242 } 3243 } 3244 3245 oif_flags = ifp->if_flags; 3246 3247 KERNEL_LOCK_UNLESS_IFP_MPSAFE(ifp); 3248 IFNET_LOCK(ifp); 3249 3250 error = (*ifp->if_ioctl)(ifp, cmd, data); 3251 if (error != ENOTTY) 3252 ; 3253 else if (so->so_proto == NULL) 3254 error = EOPNOTSUPP; 3255 else { 3256 KERNEL_LOCK_IF_IFP_MPSAFE(ifp); 3257 MODULE_HOOK_CALL(if_ifioctl_43_hook, 3258 (so, ocmd, cmd, data, l), enosys(), error); 3259 if (error == ENOSYS) 3260 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 3261 cmd, data, ifp); 3262 KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp); 3263 } 3264 3265 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 3266 if ((ifp->if_flags & IFF_UP) != 0) { 3267 int s = splsoftnet(); 3268 if_up_locked(ifp); 3269 splx(s); 3270 } 3271 } 3272 if (cmd != ocmd) 3273 ifreqn2o(oifr, ifr); 3274 3275 IFNET_UNLOCK(ifp); 3276 KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(ifp); 3277 out: 3278 if_put(ifp, &psref); 3279 curlwp_bindx(bound); 3280 return error; 3281 } 3282 3283 /* 3284 * Return interface configuration 3285 * of system. List may be used 3286 * in later ioctl's (above) to get 3287 * other information. 3288 * 3289 * Each record is a struct ifreq. Before the addition of 3290 * sockaddr_storage, the API rule was that sockaddr flavors that did 3291 * not fit would extend beyond the struct ifreq, with the next struct 3292 * ifreq starting sa_len beyond the struct sockaddr. Because the 3293 * union in struct ifreq includes struct sockaddr_storage, every kind 3294 * of sockaddr must fit. Thus, there are no longer any overlength 3295 * records. 3296 * 3297 * Records are added to the user buffer if they fit, and ifc_len is 3298 * adjusted to the length that was written. Thus, the user is only 3299 * assured of getting the complete list if ifc_len on return is at 3300 * least sizeof(struct ifreq) less than it was on entry. 3301 * 3302 * If the user buffer pointer is NULL, this routine copies no data and 3303 * returns the amount of space that would be needed. 3304 * 3305 * Invariants: 3306 * ifrp points to the next part of the user's buffer to be used. If 3307 * ifrp != NULL, space holds the number of bytes remaining that we may 3308 * write at ifrp. Otherwise, space holds the number of bytes that 3309 * would have been written had there been adequate space. 3310 */ 3311 /*ARGSUSED*/ 3312 static int 3313 ifconf(u_long cmd, void *data) 3314 { 3315 struct ifconf *ifc = (struct ifconf *)data; 3316 struct ifnet *ifp; 3317 struct ifaddr *ifa; 3318 struct ifreq ifr, *ifrp = NULL; 3319 int space = 0, error = 0; 3320 const int sz = (int)sizeof(struct ifreq); 3321 const bool docopy = ifc->ifc_req != NULL; 3322 int s; 3323 int bound; 3324 struct psref psref; 3325 3326 if (docopy) { 3327 space = ifc->ifc_len; 3328 ifrp = ifc->ifc_req; 3329 } 3330 3331 bound = curlwp_bind(); 3332 s = pserialize_read_enter(); 3333 IFNET_READER_FOREACH(ifp) { 3334 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 3335 pserialize_read_exit(s); 3336 3337 (void)strncpy(ifr.ifr_name, ifp->if_xname, 3338 sizeof(ifr.ifr_name)); 3339 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 3340 error = ENAMETOOLONG; 3341 goto release_exit; 3342 } 3343 if (IFADDR_READER_EMPTY(ifp)) { 3344 /* Interface with no addresses - send zero sockaddr. */ 3345 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 3346 if (!docopy) { 3347 space += sz; 3348 goto next; 3349 } 3350 if (space >= sz) { 3351 error = copyout(&ifr, ifrp, sz); 3352 if (error != 0) 3353 goto release_exit; 3354 ifrp++; 3355 space -= sz; 3356 } 3357 } 3358 3359 s = pserialize_read_enter(); 3360 IFADDR_READER_FOREACH(ifa, ifp) { 3361 struct sockaddr *sa = ifa->ifa_addr; 3362 /* all sockaddrs must fit in sockaddr_storage */ 3363 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 3364 3365 if (!docopy) { 3366 space += sz; 3367 continue; 3368 } 3369 memcpy(&ifr.ifr_space, sa, sa->sa_len); 3370 pserialize_read_exit(s); 3371 3372 if (space >= sz) { 3373 error = copyout(&ifr, ifrp, sz); 3374 if (error != 0) 3375 goto release_exit; 3376 ifrp++; space -= sz; 3377 } 3378 s = pserialize_read_enter(); 3379 } 3380 pserialize_read_exit(s); 3381 3382 next: 3383 s = pserialize_read_enter(); 3384 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3385 } 3386 pserialize_read_exit(s); 3387 curlwp_bindx(bound); 3388 3389 if (docopy) { 3390 KASSERT(0 <= space && space <= ifc->ifc_len); 3391 ifc->ifc_len -= space; 3392 } else { 3393 KASSERT(space >= 0); 3394 ifc->ifc_len = space; 3395 } 3396 return (0); 3397 3398 release_exit: 3399 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3400 curlwp_bindx(bound); 3401 return error; 3402 } 3403 3404 int 3405 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 3406 { 3407 uint8_t len = sizeof(ifr->ifr_ifru.ifru_space); 3408 struct ifreq ifrb; 3409 struct oifreq *oifr = NULL; 3410 u_long ocmd = cmd; 3411 int hook; 3412 3413 MODULE_HOOK_CALL(if_cvtcmd_43_hook, (&cmd, ocmd), enosys(), hook); 3414 if (hook != ENOSYS) { 3415 if (cmd != ocmd) { 3416 oifr = (struct oifreq *)(void *)ifr; 3417 ifr = &ifrb; 3418 ifreqo2n(oifr, ifr); 3419 len = sizeof(oifr->ifr_addr); 3420 } 3421 } 3422 3423 if (len < sa->sa_len) 3424 return EFBIG; 3425 3426 memset(&ifr->ifr_addr, 0, len); 3427 sockaddr_copy(&ifr->ifr_addr, len, sa); 3428 3429 if (cmd != ocmd) 3430 ifreqn2o(oifr, ifr); 3431 return 0; 3432 } 3433 3434 /* 3435 * wrapper function for the drivers which doesn't have if_transmit(). 3436 */ 3437 static int 3438 if_transmit(struct ifnet *ifp, struct mbuf *m) 3439 { 3440 int s, error; 3441 size_t pktlen = m->m_pkthdr.len; 3442 bool mcast = (m->m_flags & M_MCAST) != 0; 3443 3444 s = splnet(); 3445 3446 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3447 if (error != 0) { 3448 /* mbuf is already freed */ 3449 goto out; 3450 } 3451 3452 ifp->if_obytes += pktlen; 3453 if (mcast) 3454 ifp->if_omcasts++; 3455 3456 if ((ifp->if_flags & IFF_OACTIVE) == 0) 3457 if_start_lock(ifp); 3458 out: 3459 splx(s); 3460 3461 return error; 3462 } 3463 3464 int 3465 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 3466 { 3467 int error; 3468 3469 #ifdef ALTQ 3470 KERNEL_LOCK(1, NULL); 3471 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 3472 error = if_transmit(ifp, m); 3473 KERNEL_UNLOCK_ONE(NULL); 3474 } else { 3475 KERNEL_UNLOCK_ONE(NULL); 3476 error = (*ifp->if_transmit)(ifp, m); 3477 /* mbuf is alredy freed */ 3478 } 3479 #else /* !ALTQ */ 3480 error = (*ifp->if_transmit)(ifp, m); 3481 /* mbuf is alredy freed */ 3482 #endif /* !ALTQ */ 3483 3484 return error; 3485 } 3486 3487 /* 3488 * Queue message on interface, and start output if interface 3489 * not yet active. 3490 */ 3491 int 3492 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 3493 { 3494 3495 return if_transmit_lock(ifp, m); 3496 } 3497 3498 /* 3499 * Queue message on interface, possibly using a second fast queue 3500 */ 3501 int 3502 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 3503 { 3504 int error = 0; 3505 3506 if (ifq != NULL 3507 #ifdef ALTQ 3508 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 3509 #endif 3510 ) { 3511 if (IF_QFULL(ifq)) { 3512 IF_DROP(&ifp->if_snd); 3513 m_freem(m); 3514 if (error == 0) 3515 error = ENOBUFS; 3516 } else 3517 IF_ENQUEUE(ifq, m); 3518 } else 3519 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3520 if (error != 0) { 3521 ++ifp->if_oerrors; 3522 return error; 3523 } 3524 return 0; 3525 } 3526 3527 int 3528 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 3529 { 3530 int rc; 3531 3532 KASSERT(IFNET_LOCKED(ifp)); 3533 if (ifp->if_initaddr != NULL) 3534 rc = (*ifp->if_initaddr)(ifp, ifa, src); 3535 else if (src || 3536 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 3537 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 3538 3539 return rc; 3540 } 3541 3542 int 3543 if_do_dad(struct ifnet *ifp) 3544 { 3545 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 3546 return 0; 3547 3548 switch (ifp->if_type) { 3549 case IFT_FAITH: 3550 /* 3551 * These interfaces do not have the IFF_LOOPBACK flag, 3552 * but loop packets back. We do not have to do DAD on such 3553 * interfaces. We should even omit it, because loop-backed 3554 * responses would confuse the DAD procedure. 3555 */ 3556 return 0; 3557 default: 3558 /* 3559 * Our DAD routine requires the interface up and running. 3560 * However, some interfaces can be up before the RUNNING 3561 * status. Additionaly, users may try to assign addresses 3562 * before the interface becomes up (or running). 3563 * We simply skip DAD in such a case as a work around. 3564 * XXX: we should rather mark "tentative" on such addresses, 3565 * and do DAD after the interface becomes ready. 3566 */ 3567 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 3568 (IFF_UP|IFF_RUNNING)) 3569 return 0; 3570 3571 return 1; 3572 } 3573 } 3574 3575 int 3576 if_flags_set(ifnet_t *ifp, const short flags) 3577 { 3578 int rc; 3579 3580 KASSERT(IFNET_LOCKED(ifp)); 3581 3582 if (ifp->if_setflags != NULL) 3583 rc = (*ifp->if_setflags)(ifp, flags); 3584 else { 3585 short cantflags, chgdflags; 3586 struct ifreq ifr; 3587 3588 chgdflags = ifp->if_flags ^ flags; 3589 cantflags = chgdflags & IFF_CANTCHANGE; 3590 3591 if (cantflags != 0) 3592 ifp->if_flags ^= cantflags; 3593 3594 /* Traditionally, we do not call if_ioctl after 3595 * setting/clearing only IFF_PROMISC if the interface 3596 * isn't IFF_UP. Uphold that tradition. 3597 */ 3598 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 3599 return 0; 3600 3601 memset(&ifr, 0, sizeof(ifr)); 3602 3603 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 3604 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 3605 3606 if (rc != 0 && cantflags != 0) 3607 ifp->if_flags ^= cantflags; 3608 } 3609 3610 return rc; 3611 } 3612 3613 int 3614 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 3615 { 3616 int rc; 3617 struct ifreq ifr; 3618 3619 /* There remain some paths that don't hold IFNET_LOCK yet */ 3620 #ifdef NET_MPSAFE 3621 /* CARP and MROUTING still don't deal with the lock yet */ 3622 #if (!defined(NCARP) || (NCARP == 0)) && !defined(MROUTING) 3623 KASSERT(IFNET_LOCKED(ifp)); 3624 #endif 3625 #endif 3626 if (ifp->if_mcastop != NULL) 3627 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 3628 else { 3629 ifreq_setaddr(cmd, &ifr, sa); 3630 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 3631 } 3632 3633 return rc; 3634 } 3635 3636 static void 3637 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 3638 struct ifaltq *ifq) 3639 { 3640 const struct sysctlnode *cnode, *rnode; 3641 3642 if (sysctl_createv(clog, 0, NULL, &rnode, 3643 CTLFLAG_PERMANENT, 3644 CTLTYPE_NODE, "interfaces", 3645 SYSCTL_DESCR("Per-interface controls"), 3646 NULL, 0, NULL, 0, 3647 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3648 goto bad; 3649 3650 if (sysctl_createv(clog, 0, &rnode, &rnode, 3651 CTLFLAG_PERMANENT, 3652 CTLTYPE_NODE, ifname, 3653 SYSCTL_DESCR("Interface controls"), 3654 NULL, 0, NULL, 0, 3655 CTL_CREATE, CTL_EOL) != 0) 3656 goto bad; 3657 3658 if (sysctl_createv(clog, 0, &rnode, &rnode, 3659 CTLFLAG_PERMANENT, 3660 CTLTYPE_NODE, "sndq", 3661 SYSCTL_DESCR("Interface output queue controls"), 3662 NULL, 0, NULL, 0, 3663 CTL_CREATE, CTL_EOL) != 0) 3664 goto bad; 3665 3666 if (sysctl_createv(clog, 0, &rnode, &cnode, 3667 CTLFLAG_PERMANENT, 3668 CTLTYPE_INT, "len", 3669 SYSCTL_DESCR("Current output queue length"), 3670 NULL, 0, &ifq->ifq_len, 0, 3671 CTL_CREATE, CTL_EOL) != 0) 3672 goto bad; 3673 3674 if (sysctl_createv(clog, 0, &rnode, &cnode, 3675 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3676 CTLTYPE_INT, "maxlen", 3677 SYSCTL_DESCR("Maximum allowed output queue length"), 3678 NULL, 0, &ifq->ifq_maxlen, 0, 3679 CTL_CREATE, CTL_EOL) != 0) 3680 goto bad; 3681 3682 if (sysctl_createv(clog, 0, &rnode, &cnode, 3683 CTLFLAG_PERMANENT, 3684 CTLTYPE_INT, "drops", 3685 SYSCTL_DESCR("Packets dropped due to full output queue"), 3686 NULL, 0, &ifq->ifq_drops, 0, 3687 CTL_CREATE, CTL_EOL) != 0) 3688 goto bad; 3689 3690 return; 3691 bad: 3692 printf("%s: could not attach sysctl nodes\n", ifname); 3693 return; 3694 } 3695 3696 #if defined(INET) || defined(INET6) 3697 3698 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3699 static int \ 3700 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3701 { \ 3702 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3703 } 3704 3705 #if defined(INET) 3706 static int 3707 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3708 { 3709 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3710 } 3711 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3712 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3713 #endif 3714 3715 #if defined(INET6) 3716 static int 3717 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3718 { 3719 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3720 } 3721 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3722 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3723 #endif 3724 3725 static void 3726 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3727 { 3728 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3729 const char *pfname = NULL, *ipname = NULL; 3730 int ipn = 0, qid = 0; 3731 3732 switch (pf) { 3733 #if defined(INET) 3734 case PF_INET: 3735 len_func = sysctl_net_ip_pktq_items; 3736 maxlen_func = sysctl_net_ip_pktq_maxlen; 3737 drops_func = sysctl_net_ip_pktq_drops; 3738 pfname = "inet", ipn = IPPROTO_IP; 3739 ipname = "ip", qid = IPCTL_IFQ; 3740 break; 3741 #endif 3742 #if defined(INET6) 3743 case PF_INET6: 3744 len_func = sysctl_net_ip6_pktq_items; 3745 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3746 drops_func = sysctl_net_ip6_pktq_drops; 3747 pfname = "inet6", ipn = IPPROTO_IPV6; 3748 ipname = "ip6", qid = IPV6CTL_IFQ; 3749 break; 3750 #endif 3751 default: 3752 KASSERT(false); 3753 } 3754 3755 sysctl_createv(clog, 0, NULL, NULL, 3756 CTLFLAG_PERMANENT, 3757 CTLTYPE_NODE, pfname, NULL, 3758 NULL, 0, NULL, 0, 3759 CTL_NET, pf, CTL_EOL); 3760 sysctl_createv(clog, 0, NULL, NULL, 3761 CTLFLAG_PERMANENT, 3762 CTLTYPE_NODE, ipname, NULL, 3763 NULL, 0, NULL, 0, 3764 CTL_NET, pf, ipn, CTL_EOL); 3765 sysctl_createv(clog, 0, NULL, NULL, 3766 CTLFLAG_PERMANENT, 3767 CTLTYPE_NODE, "ifq", 3768 SYSCTL_DESCR("Protocol input queue controls"), 3769 NULL, 0, NULL, 0, 3770 CTL_NET, pf, ipn, qid, CTL_EOL); 3771 3772 sysctl_createv(clog, 0, NULL, NULL, 3773 CTLFLAG_PERMANENT, 3774 CTLTYPE_QUAD, "len", 3775 SYSCTL_DESCR("Current input queue length"), 3776 len_func, 0, NULL, 0, 3777 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3778 sysctl_createv(clog, 0, NULL, NULL, 3779 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3780 CTLTYPE_INT, "maxlen", 3781 SYSCTL_DESCR("Maximum allowed input queue length"), 3782 maxlen_func, 0, NULL, 0, 3783 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3784 sysctl_createv(clog, 0, NULL, NULL, 3785 CTLFLAG_PERMANENT, 3786 CTLTYPE_QUAD, "drops", 3787 SYSCTL_DESCR("Packets dropped due to full input queue"), 3788 drops_func, 0, NULL, 0, 3789 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3790 } 3791 #endif /* INET || INET6 */ 3792 3793 static int 3794 if_sdl_sysctl(SYSCTLFN_ARGS) 3795 { 3796 struct ifnet *ifp; 3797 const struct sockaddr_dl *sdl; 3798 struct psref psref; 3799 int error = 0; 3800 int bound; 3801 3802 if (namelen != 1) 3803 return EINVAL; 3804 3805 bound = curlwp_bind(); 3806 ifp = if_get_byindex(name[0], &psref); 3807 if (ifp == NULL) { 3808 error = ENODEV; 3809 goto out0; 3810 } 3811 3812 sdl = ifp->if_sadl; 3813 if (sdl == NULL) { 3814 *oldlenp = 0; 3815 goto out1; 3816 } 3817 3818 if (oldp == NULL) { 3819 *oldlenp = sdl->sdl_alen; 3820 goto out1; 3821 } 3822 3823 if (*oldlenp >= sdl->sdl_alen) 3824 *oldlenp = sdl->sdl_alen; 3825 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3826 out1: 3827 if_put(ifp, &psref); 3828 out0: 3829 curlwp_bindx(bound); 3830 return error; 3831 } 3832 3833 static void 3834 if_sysctl_setup(struct sysctllog **clog) 3835 { 3836 const struct sysctlnode *rnode = NULL; 3837 3838 sysctl_createv(clog, 0, NULL, &rnode, 3839 CTLFLAG_PERMANENT, 3840 CTLTYPE_NODE, "sdl", 3841 SYSCTL_DESCR("Get active link-layer address"), 3842 if_sdl_sysctl, 0, NULL, 0, 3843 CTL_NET, CTL_CREATE, CTL_EOL); 3844 3845 #if defined(INET) 3846 sysctl_net_pktq_setup(NULL, PF_INET); 3847 #endif 3848 #ifdef INET6 3849 if (in6_present) 3850 sysctl_net_pktq_setup(NULL, PF_INET6); 3851 #endif 3852 } 3853
Indexes created Tue Sep 23 07:09:52 GMT 2025