if.c revision 1.348
1 /* $NetBSD: if.c,v 1.348 2016/06/28 02:36:54 ozaki-r Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by William Studenmund and Jason R. Thorpe. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 34 * All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the project nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 */ 60 61 /* 62 * Copyright (c) 1980, 1986, 1993 63 * The Regents of the University of California. All rights reserved. 64 * 65 * Redistribution and use in source and binary forms, with or without 66 * modification, are permitted provided that the following conditions 67 * are met: 68 * 1. Redistributions of source code must retain the above copyright 69 * notice, this list of conditions and the following disclaimer. 70 * 2. Redistributions in binary form must reproduce the above copyright 71 * notice, this list of conditions and the following disclaimer in the 72 * documentation and/or other materials provided with the distribution. 73 * 3. Neither the name of the University nor the names of its contributors 74 * may be used to endorse or promote products derived from this software 75 * without specific prior written permission. 76 * 77 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 78 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 79 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 80 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 81 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 82 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 83 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 84 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 85 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 86 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 87 * SUCH DAMAGE. 88 * 89 * @(#)if.c 8.5 (Berkeley) 1/9/95 90 */ 91 92 #include <sys/cdefs.h> 93 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.348 2016/06/28 02:36:54 ozaki-r Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 98 #include "opt_atalk.h" 99 #include "opt_natm.h" 100 #include "opt_wlan.h" 101 #include "opt_net_mpsafe.h" 102 #endif 103 104 #include <sys/param.h> 105 #include <sys/mbuf.h> 106 #include <sys/systm.h> 107 #include <sys/callout.h> 108 #include <sys/proc.h> 109 #include <sys/socket.h> 110 #include <sys/socketvar.h> 111 #include <sys/domain.h> 112 #include <sys/protosw.h> 113 #include <sys/kernel.h> 114 #include <sys/ioctl.h> 115 #include <sys/sysctl.h> 116 #include <sys/syslog.h> 117 #include <sys/kauth.h> 118 #include <sys/kmem.h> 119 #include <sys/xcall.h> 120 #include <sys/cpu.h> 121 #include <sys/intr.h> 122 123 #include <net/if.h> 124 #include <net/if_dl.h> 125 #include <net/if_ether.h> 126 #include <net/if_media.h> 127 #include <net80211/ieee80211.h> 128 #include <net80211/ieee80211_ioctl.h> 129 #include <net/if_types.h> 130 #include <net/route.h> 131 #include <net/netisr.h> 132 #include <sys/module.h> 133 #ifdef NETATALK 134 #include <netatalk/at_extern.h> 135 #include <netatalk/at.h> 136 #endif 137 #include <net/pfil.h> 138 #include <netinet/in.h> 139 #include <netinet/in_var.h> 140 141 #ifdef INET6 142 #include <netinet6/in6_var.h> 143 #include <netinet6/nd6.h> 144 #endif 145 146 #include "ether.h" 147 #include "fddi.h" 148 #include "token.h" 149 150 #include "carp.h" 151 #if NCARP > 0 152 #include <netinet/ip_carp.h> 153 #endif 154 155 #include <compat/sys/sockio.h> 156 #include <compat/sys/socket.h> 157 158 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 159 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 160 161 /* 162 * Global list of interfaces. 163 */ 164 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 165 struct ifnet_head ifnet_list; 166 167 struct pslist_head ifnet_pslist; 168 static ifnet_t ** ifindex2ifnet = NULL; 169 static u_int if_index = 1; 170 static size_t if_indexlim = 0; 171 static uint64_t index_gen; 172 /* Mutex to protect the above objects. */ 173 kmutex_t ifnet_mtx __cacheline_aligned; 174 struct psref_class *ifnet_psref_class __read_mostly; 175 static pserialize_t ifnet_psz; 176 177 static kmutex_t if_clone_mtx; 178 179 struct ifnet *lo0ifp; 180 int ifqmaxlen = IFQ_MAXLEN; 181 182 static int if_rt_walktree(struct rtentry *, void *); 183 184 static struct if_clone *if_clone_lookup(const char *, int *); 185 186 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 187 static int if_cloners_count; 188 189 /* Packet filtering hook for interfaces. */ 190 pfil_head_t * if_pfil; 191 192 static kauth_listener_t if_listener; 193 194 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 195 static void if_detach_queues(struct ifnet *, struct ifqueue *); 196 static void sysctl_sndq_setup(struct sysctllog **, const char *, 197 struct ifaltq *); 198 static void if_slowtimo(void *); 199 static void if_free_sadl(struct ifnet *); 200 static void if_attachdomain1(struct ifnet *); 201 static int ifconf(u_long, void *); 202 static int if_transmit(struct ifnet *, struct mbuf *); 203 static int if_clone_create(const char *); 204 static int if_clone_destroy(const char *); 205 static void if_link_state_change_si(void *); 206 207 struct if_percpuq { 208 struct ifnet *ipq_ifp; 209 void *ipq_si; 210 struct percpu *ipq_ifqs; /* struct ifqueue */ 211 }; 212 213 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 214 215 static void if_percpuq_drops(void *, void *, struct cpu_info *); 216 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 217 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 218 struct if_percpuq *); 219 220 #if defined(INET) || defined(INET6) 221 static void sysctl_net_pktq_setup(struct sysctllog **, int); 222 #endif 223 224 static int 225 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 226 void *arg0, void *arg1, void *arg2, void *arg3) 227 { 228 int result; 229 enum kauth_network_req req; 230 231 result = KAUTH_RESULT_DEFER; 232 req = (enum kauth_network_req)arg1; 233 234 if (action != KAUTH_NETWORK_INTERFACE) 235 return result; 236 237 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 238 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 239 result = KAUTH_RESULT_ALLOW; 240 241 return result; 242 } 243 244 /* 245 * Network interface utility routines. 246 * 247 * Routines with ifa_ifwith* names take sockaddr *'s as 248 * parameters. 249 */ 250 void 251 ifinit(void) 252 { 253 #if defined(INET) 254 sysctl_net_pktq_setup(NULL, PF_INET); 255 #endif 256 #ifdef INET6 257 if (in6_present) 258 sysctl_net_pktq_setup(NULL, PF_INET6); 259 #endif 260 261 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 262 if_listener_cb, NULL); 263 264 /* interfaces are available, inform socket code */ 265 ifioctl = doifioctl; 266 } 267 268 /* 269 * XXX Initialization before configure(). 270 * XXX hack to get pfil_add_hook working in autoconf. 271 */ 272 void 273 ifinit1(void) 274 { 275 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 276 TAILQ_INIT(&ifnet_list); 277 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 278 ifnet_psz = pserialize_create(); 279 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 280 PSLIST_INIT(&ifnet_pslist); 281 if_indexlim = 8; 282 283 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 284 KASSERT(if_pfil != NULL); 285 286 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 287 etherinit(); 288 #endif 289 } 290 291 ifnet_t * 292 if_alloc(u_char type) 293 { 294 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 295 } 296 297 void 298 if_free(ifnet_t *ifp) 299 { 300 kmem_free(ifp, sizeof(ifnet_t)); 301 } 302 303 void 304 if_initname(struct ifnet *ifp, const char *name, int unit) 305 { 306 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 307 "%s%d", name, unit); 308 } 309 310 /* 311 * Null routines used while an interface is going away. These routines 312 * just return an error. 313 */ 314 315 int 316 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 317 const struct sockaddr *so, const struct rtentry *rt) 318 { 319 320 return ENXIO; 321 } 322 323 void 324 if_nullinput(struct ifnet *ifp, struct mbuf *m) 325 { 326 327 /* Nothing. */ 328 } 329 330 void 331 if_nullstart(struct ifnet *ifp) 332 { 333 334 /* Nothing. */ 335 } 336 337 int 338 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 339 { 340 341 return ENXIO; 342 } 343 344 int 345 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 346 { 347 348 return ENXIO; 349 } 350 351 int 352 if_nullinit(struct ifnet *ifp) 353 { 354 355 return ENXIO; 356 } 357 358 void 359 if_nullstop(struct ifnet *ifp, int disable) 360 { 361 362 /* Nothing. */ 363 } 364 365 void 366 if_nullslowtimo(struct ifnet *ifp) 367 { 368 369 /* Nothing. */ 370 } 371 372 void 373 if_nulldrain(struct ifnet *ifp) 374 { 375 376 /* Nothing. */ 377 } 378 379 void 380 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 381 { 382 struct ifaddr *ifa; 383 struct sockaddr_dl *sdl; 384 385 ifp->if_addrlen = addrlen; 386 if_alloc_sadl(ifp); 387 ifa = ifp->if_dl; 388 sdl = satosdl(ifa->ifa_addr); 389 390 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 391 if (factory) { 392 ifp->if_hwdl = ifp->if_dl; 393 ifaref(ifp->if_hwdl); 394 } 395 /* TBD routing socket */ 396 } 397 398 struct ifaddr * 399 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 400 { 401 unsigned socksize, ifasize; 402 int addrlen, namelen; 403 struct sockaddr_dl *mask, *sdl; 404 struct ifaddr *ifa; 405 406 namelen = strlen(ifp->if_xname); 407 addrlen = ifp->if_addrlen; 408 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 409 ifasize = sizeof(*ifa) + 2 * socksize; 410 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 411 412 sdl = (struct sockaddr_dl *)(ifa + 1); 413 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 414 415 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 416 ifp->if_xname, namelen, NULL, addrlen); 417 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 418 memset(&mask->sdl_data[0], 0xff, namelen); 419 ifa->ifa_rtrequest = link_rtrequest; 420 ifa->ifa_addr = (struct sockaddr *)sdl; 421 ifa->ifa_netmask = (struct sockaddr *)mask; 422 423 *sdlp = sdl; 424 425 return ifa; 426 } 427 428 static void 429 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 430 { 431 const struct sockaddr_dl *sdl; 432 433 ifp->if_dl = ifa; 434 ifaref(ifa); 435 sdl = satosdl(ifa->ifa_addr); 436 ifp->if_sadl = sdl; 437 } 438 439 /* 440 * Allocate the link level name for the specified interface. This 441 * is an attachment helper. It must be called after ifp->if_addrlen 442 * is initialized, which may not be the case when if_attach() is 443 * called. 444 */ 445 void 446 if_alloc_sadl(struct ifnet *ifp) 447 { 448 struct ifaddr *ifa; 449 const struct sockaddr_dl *sdl; 450 451 /* 452 * If the interface already has a link name, release it 453 * now. This is useful for interfaces that can change 454 * link types, and thus switch link names often. 455 */ 456 if (ifp->if_sadl != NULL) 457 if_free_sadl(ifp); 458 459 ifa = if_dl_create(ifp, &sdl); 460 461 ifa_insert(ifp, ifa); 462 if_sadl_setrefs(ifp, ifa); 463 } 464 465 static void 466 if_deactivate_sadl(struct ifnet *ifp) 467 { 468 struct ifaddr *ifa; 469 470 KASSERT(ifp->if_dl != NULL); 471 472 ifa = ifp->if_dl; 473 474 ifp->if_sadl = NULL; 475 476 ifp->if_dl = NULL; 477 ifafree(ifa); 478 } 479 480 void 481 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 482 const struct sockaddr_dl *sdl) 483 { 484 int s; 485 486 s = splnet(); 487 488 if_deactivate_sadl(ifp); 489 490 if_sadl_setrefs(ifp, ifa); 491 IFADDR_FOREACH(ifa, ifp) 492 rtinit(ifa, RTM_LLINFO_UPD, 0); 493 splx(s); 494 } 495 496 /* 497 * Free the link level name for the specified interface. This is 498 * a detach helper. This is called from if_detach(). 499 */ 500 static void 501 if_free_sadl(struct ifnet *ifp) 502 { 503 struct ifaddr *ifa; 504 int s; 505 506 ifa = ifp->if_dl; 507 if (ifa == NULL) { 508 KASSERT(ifp->if_sadl == NULL); 509 return; 510 } 511 512 KASSERT(ifp->if_sadl != NULL); 513 514 s = splnet(); 515 rtinit(ifa, RTM_DELETE, 0); 516 ifa_remove(ifp, ifa); 517 if_deactivate_sadl(ifp); 518 if (ifp->if_hwdl == ifa) { 519 ifafree(ifa); 520 ifp->if_hwdl = NULL; 521 } 522 splx(s); 523 } 524 525 static void 526 if_getindex(ifnet_t *ifp) 527 { 528 bool hitlimit = false; 529 530 ifp->if_index_gen = index_gen++; 531 532 ifp->if_index = if_index; 533 if (ifindex2ifnet == NULL) { 534 if_index++; 535 goto skip; 536 } 537 while (if_byindex(ifp->if_index)) { 538 /* 539 * If we hit USHRT_MAX, we skip back to 0 since 540 * there are a number of places where the value 541 * of if_index or if_index itself is compared 542 * to or stored in an unsigned short. By 543 * jumping back, we won't botch those assignments 544 * or comparisons. 545 */ 546 if (++if_index == 0) { 547 if_index = 1; 548 } else if (if_index == USHRT_MAX) { 549 /* 550 * However, if we have to jump back to 551 * zero *twice* without finding an empty 552 * slot in ifindex2ifnet[], then there 553 * there are too many (>65535) interfaces. 554 */ 555 if (hitlimit) { 556 panic("too many interfaces"); 557 } 558 hitlimit = true; 559 if_index = 1; 560 } 561 ifp->if_index = if_index; 562 } 563 skip: 564 /* 565 * ifindex2ifnet is indexed by if_index. Since if_index will 566 * grow dynamically, it should grow too. 567 */ 568 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 569 size_t m, n, oldlim; 570 void *q; 571 572 oldlim = if_indexlim; 573 while (ifp->if_index >= if_indexlim) 574 if_indexlim <<= 1; 575 576 /* grow ifindex2ifnet */ 577 m = oldlim * sizeof(struct ifnet *); 578 n = if_indexlim * sizeof(struct ifnet *); 579 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 580 if (ifindex2ifnet != NULL) { 581 memcpy(q, ifindex2ifnet, m); 582 free(ifindex2ifnet, M_IFADDR); 583 } 584 ifindex2ifnet = (struct ifnet **)q; 585 } 586 ifindex2ifnet[ifp->if_index] = ifp; 587 } 588 589 /* 590 * Initialize an interface and assign an index for it. 591 * 592 * It must be called prior to a device specific attach routine 593 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 594 * and be followed by if_register: 595 * 596 * if_initialize(ifp); 597 * ether_ifattach(ifp, enaddr); 598 * if_register(ifp); 599 */ 600 void 601 if_initialize(ifnet_t *ifp) 602 { 603 KASSERT(if_indexlim > 0); 604 TAILQ_INIT(&ifp->if_addrlist); 605 606 /* 607 * Link level name is allocated later by a separate call to 608 * if_alloc_sadl(). 609 */ 610 611 if (ifp->if_snd.ifq_maxlen == 0) 612 ifp->if_snd.ifq_maxlen = ifqmaxlen; 613 614 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 615 616 ifp->if_link_state = LINK_STATE_UNKNOWN; 617 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 618 619 ifp->if_capenable = 0; 620 ifp->if_csum_flags_tx = 0; 621 ifp->if_csum_flags_rx = 0; 622 623 #ifdef ALTQ 624 ifp->if_snd.altq_type = 0; 625 ifp->if_snd.altq_disc = NULL; 626 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 627 ifp->if_snd.altq_tbr = NULL; 628 ifp->if_snd.altq_ifp = ifp; 629 #endif 630 631 #ifdef NET_MPSAFE 632 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 633 #else 634 ifp->if_snd.ifq_lock = NULL; 635 #endif 636 637 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 638 (void)pfil_run_hooks(if_pfil, 639 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 640 641 IF_AFDATA_LOCK_INIT(ifp); 642 643 if (if_is_link_state_changeable(ifp)) { 644 ifp->if_link_si = softint_establish(SOFTINT_NET, 645 if_link_state_change_si, ifp); 646 if (ifp->if_link_si == NULL) 647 panic("%s: softint_establish() failed", __func__); 648 } 649 650 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 651 psref_target_init(&ifp->if_psref, ifnet_psref_class); 652 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 653 654 IFNET_LOCK(); 655 if_getindex(ifp); 656 IFNET_UNLOCK(); 657 } 658 659 /* 660 * Register an interface to the list of "active" interfaces. 661 */ 662 void 663 if_register(ifnet_t *ifp) 664 { 665 /* 666 * If the driver has not supplied its own if_ioctl, then 667 * supply the default. 668 */ 669 if (ifp->if_ioctl == NULL) 670 ifp->if_ioctl = ifioctl_common; 671 672 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 673 674 if (!STAILQ_EMPTY(&domains)) 675 if_attachdomain1(ifp); 676 677 /* Announce the interface. */ 678 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 679 680 if (ifp->if_slowtimo != NULL) { 681 ifp->if_slowtimo_ch = 682 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 683 callout_init(ifp->if_slowtimo_ch, 0); 684 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 685 if_slowtimo(ifp); 686 } 687 688 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 689 ifp->if_transmit = if_transmit; 690 691 IFNET_LOCK(); 692 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 693 IFNET_WRITER_INSERT_TAIL(ifp); 694 IFNET_UNLOCK(); 695 } 696 697 /* 698 * The if_percpuq framework 699 * 700 * It allows network device drivers to execute the network stack 701 * in softint (so called softint-based if_input). It utilizes 702 * softint and percpu ifqueue. It doesn't distribute any packets 703 * between CPUs, unlike pktqueue(9). 704 * 705 * Currently we support two options for device drivers to apply the framework: 706 * - Use it implicitly with less changes 707 * - If you use if_attach in driver's _attach function and if_input in 708 * driver's Rx interrupt handler, a packet is queued and a softint handles 709 * the packet implicitly 710 * - Use it explicitly in each driver (recommended) 711 * - You can use if_percpuq_* directly in your driver 712 * - In this case, you need to allocate struct if_percpuq in driver's softc 713 * - See wm(4) as a reference implementation 714 */ 715 716 static void 717 if_percpuq_softint(void *arg) 718 { 719 struct if_percpuq *ipq = arg; 720 struct ifnet *ifp = ipq->ipq_ifp; 721 struct mbuf *m; 722 723 while ((m = if_percpuq_dequeue(ipq)) != NULL) 724 ifp->_if_input(ifp, m); 725 } 726 727 static void 728 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 729 { 730 struct ifqueue *const ifq = p; 731 732 memset(ifq, 0, sizeof(*ifq)); 733 ifq->ifq_maxlen = IFQ_MAXLEN; 734 } 735 736 struct if_percpuq * 737 if_percpuq_create(struct ifnet *ifp) 738 { 739 struct if_percpuq *ipq; 740 741 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 742 if (ipq == NULL) 743 panic("kmem_zalloc failed"); 744 745 ipq->ipq_ifp = ifp; 746 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 747 if_percpuq_softint, ipq); 748 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 749 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 750 751 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 752 753 return ipq; 754 } 755 756 static struct mbuf * 757 if_percpuq_dequeue(struct if_percpuq *ipq) 758 { 759 struct mbuf *m; 760 struct ifqueue *ifq; 761 int s; 762 763 s = splnet(); 764 ifq = percpu_getref(ipq->ipq_ifqs); 765 IF_DEQUEUE(ifq, m); 766 percpu_putref(ipq->ipq_ifqs); 767 splx(s); 768 769 return m; 770 } 771 772 static void 773 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 774 { 775 struct ifqueue *const ifq = p; 776 777 IF_PURGE(ifq); 778 } 779 780 void 781 if_percpuq_destroy(struct if_percpuq *ipq) 782 { 783 784 /* if_detach may already destroy it */ 785 if (ipq == NULL) 786 return; 787 788 softint_disestablish(ipq->ipq_si); 789 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 790 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 791 } 792 793 void 794 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 795 { 796 struct ifqueue *ifq; 797 int s; 798 799 KASSERT(ipq != NULL); 800 801 s = splnet(); 802 ifq = percpu_getref(ipq->ipq_ifqs); 803 if (IF_QFULL(ifq)) { 804 IF_DROP(ifq); 805 percpu_putref(ipq->ipq_ifqs); 806 m_freem(m); 807 goto out; 808 } 809 IF_ENQUEUE(ifq, m); 810 percpu_putref(ipq->ipq_ifqs); 811 812 softint_schedule(ipq->ipq_si); 813 out: 814 splx(s); 815 } 816 817 static void 818 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 819 { 820 struct ifqueue *const ifq = p; 821 int *sum = arg; 822 823 *sum += ifq->ifq_drops; 824 } 825 826 static int 827 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 828 { 829 struct sysctlnode node; 830 struct if_percpuq *ipq; 831 int sum = 0; 832 int error; 833 834 node = *rnode; 835 ipq = node.sysctl_data; 836 837 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 838 839 node.sysctl_data = ∑ 840 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 841 if (error != 0 || newp == NULL) 842 return error; 843 844 return 0; 845 } 846 847 static void 848 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 849 struct if_percpuq *ipq) 850 { 851 const struct sysctlnode *cnode, *rnode; 852 853 if (sysctl_createv(clog, 0, NULL, &rnode, 854 CTLFLAG_PERMANENT, 855 CTLTYPE_NODE, "interfaces", 856 SYSCTL_DESCR("Per-interface controls"), 857 NULL, 0, NULL, 0, 858 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 859 goto bad; 860 861 if (sysctl_createv(clog, 0, &rnode, &rnode, 862 CTLFLAG_PERMANENT, 863 CTLTYPE_NODE, ifname, 864 SYSCTL_DESCR("Interface controls"), 865 NULL, 0, NULL, 0, 866 CTL_CREATE, CTL_EOL) != 0) 867 goto bad; 868 869 if (sysctl_createv(clog, 0, &rnode, &rnode, 870 CTLFLAG_PERMANENT, 871 CTLTYPE_NODE, "rcvq", 872 SYSCTL_DESCR("Interface input queue controls"), 873 NULL, 0, NULL, 0, 874 CTL_CREATE, CTL_EOL) != 0) 875 goto bad; 876 877 #ifdef NOTYET 878 /* XXX Should show each per-CPU queue length? */ 879 if (sysctl_createv(clog, 0, &rnode, &rnode, 880 CTLFLAG_PERMANENT, 881 CTLTYPE_INT, "len", 882 SYSCTL_DESCR("Current input queue length"), 883 sysctl_percpuq_len, 0, NULL, 0, 884 CTL_CREATE, CTL_EOL) != 0) 885 goto bad; 886 887 if (sysctl_createv(clog, 0, &rnode, &cnode, 888 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 889 CTLTYPE_INT, "maxlen", 890 SYSCTL_DESCR("Maximum allowed input queue length"), 891 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 892 CTL_CREATE, CTL_EOL) != 0) 893 goto bad; 894 #endif 895 896 if (sysctl_createv(clog, 0, &rnode, &cnode, 897 CTLFLAG_PERMANENT, 898 CTLTYPE_INT, "drops", 899 SYSCTL_DESCR("Total packets dropped due to full input queue"), 900 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 901 CTL_CREATE, CTL_EOL) != 0) 902 goto bad; 903 904 return; 905 bad: 906 printf("%s: could not attach sysctl nodes\n", ifname); 907 return; 908 } 909 910 911 /* 912 * The common interface input routine that is called by device drivers, 913 * which should be used only when the driver's rx handler already runs 914 * in softint. 915 */ 916 void 917 if_input(struct ifnet *ifp, struct mbuf *m) 918 { 919 920 KASSERT(ifp->if_percpuq == NULL); 921 KASSERT(!cpu_intr_p()); 922 923 ifp->_if_input(ifp, m); 924 } 925 926 /* 927 * DEPRECATED. Use if_initialize and if_register instead. 928 * See the above comment of if_initialize. 929 * 930 * Note that it implicitly enables if_percpuq to make drivers easy to 931 * migrate softint-based if_input without much changes. If you don't 932 * want to enable it, use if_initialize instead. 933 */ 934 void 935 if_attach(ifnet_t *ifp) 936 { 937 938 if_initialize(ifp); 939 ifp->if_percpuq = if_percpuq_create(ifp); 940 if_register(ifp); 941 } 942 943 void 944 if_attachdomain(void) 945 { 946 struct ifnet *ifp; 947 int s; 948 int bound = curlwp_bind(); 949 950 s = pserialize_read_enter(); 951 IFNET_READER_FOREACH(ifp) { 952 struct psref psref; 953 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 954 pserialize_read_exit(s); 955 if_attachdomain1(ifp); 956 s = pserialize_read_enter(); 957 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 958 } 959 pserialize_read_exit(s); 960 curlwp_bindx(bound); 961 } 962 963 static void 964 if_attachdomain1(struct ifnet *ifp) 965 { 966 struct domain *dp; 967 int s; 968 969 s = splnet(); 970 971 /* address family dependent data region */ 972 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 973 DOMAIN_FOREACH(dp) { 974 if (dp->dom_ifattach != NULL) 975 ifp->if_afdata[dp->dom_family] = 976 (*dp->dom_ifattach)(ifp); 977 } 978 979 splx(s); 980 } 981 982 /* 983 * Deactivate an interface. This points all of the procedure 984 * handles at error stubs. May be called from interrupt context. 985 */ 986 void 987 if_deactivate(struct ifnet *ifp) 988 { 989 int s; 990 991 s = splnet(); 992 993 ifp->if_output = if_nulloutput; 994 ifp->_if_input = if_nullinput; 995 ifp->if_start = if_nullstart; 996 ifp->if_transmit = if_nulltransmit; 997 ifp->if_ioctl = if_nullioctl; 998 ifp->if_init = if_nullinit; 999 ifp->if_stop = if_nullstop; 1000 ifp->if_slowtimo = if_nullslowtimo; 1001 ifp->if_drain = if_nulldrain; 1002 1003 /* No more packets may be enqueued. */ 1004 ifp->if_snd.ifq_maxlen = 0; 1005 1006 splx(s); 1007 } 1008 1009 bool 1010 if_is_deactivated(struct ifnet *ifp) 1011 { 1012 1013 return ifp->if_output == if_nulloutput; 1014 } 1015 1016 void 1017 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1018 { 1019 struct ifaddr *ifa, *nifa; 1020 1021 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 1022 if (ifa->ifa_addr->sa_family != family) 1023 continue; 1024 (*purgeaddr)(ifa); 1025 } 1026 } 1027 1028 /* 1029 * Detach an interface from the list of "active" interfaces, 1030 * freeing any resources as we go along. 1031 * 1032 * NOTE: This routine must be called with a valid thread context, 1033 * as it may block. 1034 */ 1035 void 1036 if_detach(struct ifnet *ifp) 1037 { 1038 struct socket so; 1039 struct ifaddr *ifa; 1040 #ifdef IFAREF_DEBUG 1041 struct ifaddr *last_ifa = NULL; 1042 #endif 1043 struct domain *dp; 1044 const struct protosw *pr; 1045 int s, i, family, purged; 1046 uint64_t xc; 1047 1048 /* 1049 * XXX It's kind of lame that we have to have the 1050 * XXX socket structure... 1051 */ 1052 memset(&so, 0, sizeof(so)); 1053 1054 s = splnet(); 1055 1056 sysctl_teardown(&ifp->if_sysctl_log); 1057 mutex_enter(ifp->if_ioctl_lock); 1058 ifp->if_ioctl = if_nullioctl; 1059 mutex_exit(ifp->if_ioctl_lock); 1060 1061 IFNET_LOCK(); 1062 ifindex2ifnet[ifp->if_index] = NULL; 1063 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1064 IFNET_WRITER_REMOVE(ifp); 1065 pserialize_perform(ifnet_psz); 1066 IFNET_UNLOCK(); 1067 1068 mutex_obj_free(ifp->if_ioctl_lock); 1069 ifp->if_ioctl_lock = NULL; 1070 1071 if (ifp->if_slowtimo != NULL) { 1072 ifp->if_slowtimo = NULL; 1073 callout_halt(ifp->if_slowtimo_ch, NULL); 1074 callout_destroy(ifp->if_slowtimo_ch); 1075 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1076 } 1077 1078 /* 1079 * Do an if_down() to give protocols a chance to do something. 1080 */ 1081 if_down(ifp); 1082 1083 #ifdef ALTQ 1084 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1085 altq_disable(&ifp->if_snd); 1086 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1087 altq_detach(&ifp->if_snd); 1088 #endif 1089 1090 if (ifp->if_snd.ifq_lock) 1091 mutex_obj_free(ifp->if_snd.ifq_lock); 1092 1093 #if NCARP > 0 1094 /* Remove the interface from any carp group it is a part of. */ 1095 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1096 carp_ifdetach(ifp); 1097 #endif 1098 1099 /* 1100 * Rip all the addresses off the interface. This should make 1101 * all of the routes go away. 1102 * 1103 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1104 * from the list, including our "cursor", ifa. For safety, 1105 * and to honor the TAILQ abstraction, I just restart the 1106 * loop after each removal. Note that the loop will exit 1107 * when all of the remaining ifaddrs belong to the AF_LINK 1108 * family. I am counting on the historical fact that at 1109 * least one pr_usrreq in each address domain removes at 1110 * least one ifaddr. 1111 */ 1112 again: 1113 IFADDR_FOREACH(ifa, ifp) { 1114 family = ifa->ifa_addr->sa_family; 1115 #ifdef IFAREF_DEBUG 1116 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1117 ifa, family, ifa->ifa_refcnt); 1118 if (last_ifa != NULL && ifa == last_ifa) 1119 panic("if_detach: loop detected"); 1120 last_ifa = ifa; 1121 #endif 1122 if (family == AF_LINK) 1123 continue; 1124 dp = pffinddomain(family); 1125 #ifdef DIAGNOSTIC 1126 if (dp == NULL) 1127 panic("if_detach: no domain for AF %d", 1128 family); 1129 #endif 1130 /* 1131 * XXX These PURGEIF calls are redundant with the 1132 * purge-all-families calls below, but are left in for 1133 * now both to make a smaller change, and to avoid 1134 * unplanned interactions with clearing of 1135 * ifp->if_addrlist. 1136 */ 1137 purged = 0; 1138 for (pr = dp->dom_protosw; 1139 pr < dp->dom_protoswNPROTOSW; pr++) { 1140 so.so_proto = pr; 1141 if (pr->pr_usrreqs) { 1142 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1143 purged = 1; 1144 } 1145 } 1146 if (purged == 0) { 1147 /* 1148 * XXX What's really the best thing to do 1149 * XXX here? --thorpej (at) NetBSD.org 1150 */ 1151 printf("if_detach: WARNING: AF %d not purged\n", 1152 family); 1153 ifa_remove(ifp, ifa); 1154 } 1155 goto again; 1156 } 1157 1158 if_free_sadl(ifp); 1159 1160 /* Walk the routing table looking for stragglers. */ 1161 for (i = 0; i <= AF_MAX; i++) { 1162 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 1163 continue; 1164 } 1165 1166 DOMAIN_FOREACH(dp) { 1167 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1168 { 1169 void *p = ifp->if_afdata[dp->dom_family]; 1170 if (p) { 1171 ifp->if_afdata[dp->dom_family] = NULL; 1172 (*dp->dom_ifdetach)(ifp, p); 1173 } 1174 } 1175 1176 /* 1177 * One would expect multicast memberships (INET and 1178 * INET6) on UDP sockets to be purged by the PURGEIF 1179 * calls above, but if all addresses were removed from 1180 * the interface prior to destruction, the calls will 1181 * not be made (e.g. ppp, for which pppd(8) generally 1182 * removes addresses before destroying the interface). 1183 * Because there is no invariant that multicast 1184 * memberships only exist for interfaces with IPv4 1185 * addresses, we must call PURGEIF regardless of 1186 * addresses. (Protocols which might store ifnet 1187 * pointers are marked with PR_PURGEIF.) 1188 */ 1189 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1190 so.so_proto = pr; 1191 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1192 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1193 } 1194 } 1195 1196 (void)pfil_run_hooks(if_pfil, 1197 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 1198 (void)pfil_head_destroy(ifp->if_pfil); 1199 1200 /* Announce that the interface is gone. */ 1201 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1202 1203 IF_AFDATA_LOCK_DESTROY(ifp); 1204 1205 if (if_is_link_state_changeable(ifp)) { 1206 softint_disestablish(ifp->if_link_si); 1207 ifp->if_link_si = NULL; 1208 } 1209 1210 /* 1211 * remove packets that came from ifp, from software interrupt queues. 1212 */ 1213 DOMAIN_FOREACH(dp) { 1214 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1215 struct ifqueue *iq = dp->dom_ifqueues[i]; 1216 if (iq == NULL) 1217 break; 1218 dp->dom_ifqueues[i] = NULL; 1219 if_detach_queues(ifp, iq); 1220 } 1221 } 1222 1223 /* 1224 * IP queues have to be processed separately: net-queue barrier 1225 * ensures that the packets are dequeued while a cross-call will 1226 * ensure that the interrupts have completed. FIXME: not quite.. 1227 */ 1228 #ifdef INET 1229 pktq_barrier(ip_pktq); 1230 #endif 1231 #ifdef INET6 1232 if (in6_present) 1233 pktq_barrier(ip6_pktq); 1234 #endif 1235 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1236 xc_wait(xc); 1237 1238 /* Wait for all readers to drain before freeing. */ 1239 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1240 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1241 1242 if (ifp->if_percpuq != NULL) { 1243 if_percpuq_destroy(ifp->if_percpuq); 1244 ifp->if_percpuq = NULL; 1245 } 1246 1247 splx(s); 1248 } 1249 1250 static void 1251 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1252 { 1253 struct mbuf *m, *prev, *next; 1254 1255 prev = NULL; 1256 for (m = q->ifq_head; m != NULL; m = next) { 1257 KASSERT((m->m_flags & M_PKTHDR) != 0); 1258 1259 next = m->m_nextpkt; 1260 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1261 prev = m; 1262 continue; 1263 } 1264 1265 if (prev != NULL) 1266 prev->m_nextpkt = m->m_nextpkt; 1267 else 1268 q->ifq_head = m->m_nextpkt; 1269 if (q->ifq_tail == m) 1270 q->ifq_tail = prev; 1271 q->ifq_len--; 1272 1273 m->m_nextpkt = NULL; 1274 m_freem(m); 1275 IF_DROP(q); 1276 } 1277 } 1278 1279 /* 1280 * Callback for a radix tree walk to delete all references to an 1281 * ifnet. 1282 */ 1283 static int 1284 if_rt_walktree(struct rtentry *rt, void *v) 1285 { 1286 struct ifnet *ifp = (struct ifnet *)v; 1287 int error; 1288 struct rtentry *retrt; 1289 1290 if (rt->rt_ifp != ifp) 1291 return 0; 1292 1293 /* Delete the entry. */ 1294 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 1295 rt_mask(rt), rt->rt_flags, &retrt); 1296 if (error == 0) { 1297 KASSERT(retrt == rt); 1298 KASSERT((retrt->rt_flags & RTF_UP) == 0); 1299 retrt->rt_ifp = NULL; 1300 rtfree(retrt); 1301 } else { 1302 printf("%s: warning: unable to delete rtentry @ %p, " 1303 "error = %d\n", ifp->if_xname, rt, error); 1304 } 1305 return ERESTART; 1306 } 1307 1308 /* 1309 * Create a clone network interface. 1310 */ 1311 static int 1312 if_clone_create(const char *name) 1313 { 1314 struct if_clone *ifc; 1315 int unit; 1316 struct ifnet *ifp; 1317 struct psref psref; 1318 1319 ifc = if_clone_lookup(name, &unit); 1320 if (ifc == NULL) 1321 return EINVAL; 1322 1323 ifp = if_get(name, &psref); 1324 if (ifp != NULL) { 1325 if_put(ifp, &psref); 1326 return EEXIST; 1327 } 1328 1329 return (*ifc->ifc_create)(ifc, unit); 1330 } 1331 1332 /* 1333 * Destroy a clone network interface. 1334 */ 1335 static int 1336 if_clone_destroy(const char *name) 1337 { 1338 struct if_clone *ifc; 1339 struct ifnet *ifp; 1340 struct psref psref; 1341 1342 ifc = if_clone_lookup(name, NULL); 1343 if (ifc == NULL) 1344 return EINVAL; 1345 1346 if (ifc->ifc_destroy == NULL) 1347 return EOPNOTSUPP; 1348 1349 ifp = if_get(name, &psref); 1350 if (ifp == NULL) 1351 return ENXIO; 1352 1353 /* We have to disable ioctls here */ 1354 mutex_enter(ifp->if_ioctl_lock); 1355 ifp->if_ioctl = if_nullioctl; 1356 mutex_exit(ifp->if_ioctl_lock); 1357 1358 /* 1359 * We cannot call ifc_destroy with holding ifp. 1360 * Releasing ifp here is safe thanks to if_clone_mtx. 1361 */ 1362 if_put(ifp, &psref); 1363 1364 return (*ifc->ifc_destroy)(ifp); 1365 } 1366 1367 /* 1368 * Look up a network interface cloner. 1369 */ 1370 static struct if_clone * 1371 if_clone_lookup(const char *name, int *unitp) 1372 { 1373 struct if_clone *ifc; 1374 const char *cp; 1375 char *dp, ifname[IFNAMSIZ + 3]; 1376 int unit; 1377 1378 strcpy(ifname, "if_"); 1379 /* separate interface name from unit */ 1380 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1381 *cp && (*cp < '0' || *cp > '9');) 1382 *dp++ = *cp++; 1383 1384 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1385 return NULL; /* No name or unit number */ 1386 *dp++ = '\0'; 1387 1388 again: 1389 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1390 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1391 break; 1392 } 1393 1394 if (ifc == NULL) { 1395 if (*ifname == '\0' || 1396 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1397 return NULL; 1398 *ifname = '\0'; 1399 goto again; 1400 } 1401 1402 unit = 0; 1403 while (cp - name < IFNAMSIZ && *cp) { 1404 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1405 /* Bogus unit number. */ 1406 return NULL; 1407 } 1408 unit = (unit * 10) + (*cp++ - '0'); 1409 } 1410 1411 if (unitp != NULL) 1412 *unitp = unit; 1413 return ifc; 1414 } 1415 1416 /* 1417 * Register a network interface cloner. 1418 */ 1419 void 1420 if_clone_attach(struct if_clone *ifc) 1421 { 1422 1423 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1424 if_cloners_count++; 1425 } 1426 1427 /* 1428 * Unregister a network interface cloner. 1429 */ 1430 void 1431 if_clone_detach(struct if_clone *ifc) 1432 { 1433 1434 LIST_REMOVE(ifc, ifc_list); 1435 if_cloners_count--; 1436 } 1437 1438 /* 1439 * Provide list of interface cloners to userspace. 1440 */ 1441 int 1442 if_clone_list(int buf_count, char *buffer, int *total) 1443 { 1444 char outbuf[IFNAMSIZ], *dst; 1445 struct if_clone *ifc; 1446 int count, error = 0; 1447 1448 *total = if_cloners_count; 1449 if ((dst = buffer) == NULL) { 1450 /* Just asking how many there are. */ 1451 return 0; 1452 } 1453 1454 if (buf_count < 0) 1455 return EINVAL; 1456 1457 count = (if_cloners_count < buf_count) ? 1458 if_cloners_count : buf_count; 1459 1460 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1461 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1462 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1463 if (outbuf[sizeof(outbuf) - 1] != '\0') 1464 return ENAMETOOLONG; 1465 error = copyout(outbuf, dst, sizeof(outbuf)); 1466 if (error != 0) 1467 break; 1468 } 1469 1470 return error; 1471 } 1472 1473 void 1474 ifaref(struct ifaddr *ifa) 1475 { 1476 ifa->ifa_refcnt++; 1477 } 1478 1479 void 1480 ifafree(struct ifaddr *ifa) 1481 { 1482 KASSERT(ifa != NULL); 1483 KASSERT(ifa->ifa_refcnt > 0); 1484 1485 if (--ifa->ifa_refcnt == 0) { 1486 free(ifa, M_IFADDR); 1487 } 1488 } 1489 1490 void 1491 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1492 { 1493 ifa->ifa_ifp = ifp; 1494 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1495 ifaref(ifa); 1496 } 1497 1498 void 1499 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1500 { 1501 KASSERT(ifa->ifa_ifp == ifp); 1502 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1503 ifafree(ifa); 1504 } 1505 1506 static inline int 1507 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1508 { 1509 return sockaddr_cmp(sa1, sa2) == 0; 1510 } 1511 1512 /* 1513 * Locate an interface based on a complete address. 1514 */ 1515 /*ARGSUSED*/ 1516 struct ifaddr * 1517 ifa_ifwithaddr(const struct sockaddr *addr) 1518 { 1519 struct ifnet *ifp; 1520 struct ifaddr *ifa; 1521 int s; 1522 1523 s = pserialize_read_enter(); 1524 IFNET_READER_FOREACH(ifp) { 1525 if (if_is_deactivated(ifp)) 1526 continue; 1527 IFADDR_FOREACH(ifa, ifp) { 1528 if (ifa->ifa_addr->sa_family != addr->sa_family) 1529 continue; 1530 if (equal(addr, ifa->ifa_addr)) 1531 return ifa; 1532 if ((ifp->if_flags & IFF_BROADCAST) && 1533 ifa->ifa_broadaddr && 1534 /* IP6 doesn't have broadcast */ 1535 ifa->ifa_broadaddr->sa_len != 0 && 1536 equal(ifa->ifa_broadaddr, addr)) 1537 return ifa; 1538 } 1539 } 1540 pserialize_read_exit(s); 1541 return NULL; 1542 } 1543 1544 /* 1545 * Locate the point to point interface with a given destination address. 1546 */ 1547 /*ARGSUSED*/ 1548 struct ifaddr * 1549 ifa_ifwithdstaddr(const struct sockaddr *addr) 1550 { 1551 struct ifnet *ifp; 1552 struct ifaddr *ifa; 1553 int s; 1554 1555 s = pserialize_read_enter(); 1556 IFNET_READER_FOREACH(ifp) { 1557 if (if_is_deactivated(ifp)) 1558 continue; 1559 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1560 continue; 1561 IFADDR_FOREACH(ifa, ifp) { 1562 if (ifa->ifa_addr->sa_family != addr->sa_family || 1563 ifa->ifa_dstaddr == NULL) 1564 continue; 1565 if (equal(addr, ifa->ifa_dstaddr)) 1566 return ifa; 1567 } 1568 } 1569 pserialize_read_exit(s); 1570 return NULL; 1571 } 1572 1573 /* 1574 * Find an interface on a specific network. If many, choice 1575 * is most specific found. 1576 */ 1577 struct ifaddr * 1578 ifa_ifwithnet(const struct sockaddr *addr) 1579 { 1580 struct ifnet *ifp; 1581 struct ifaddr *ifa; 1582 const struct sockaddr_dl *sdl; 1583 struct ifaddr *ifa_maybe = 0; 1584 u_int af = addr->sa_family; 1585 const char *addr_data = addr->sa_data, *cplim; 1586 int s; 1587 1588 if (af == AF_LINK) { 1589 sdl = satocsdl(addr); 1590 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1591 ifindex2ifnet[sdl->sdl_index] && 1592 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1593 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1594 } 1595 } 1596 #ifdef NETATALK 1597 if (af == AF_APPLETALK) { 1598 const struct sockaddr_at *sat, *sat2; 1599 sat = (const struct sockaddr_at *)addr; 1600 s = pserialize_read_enter(); 1601 IFNET_READER_FOREACH(ifp) { 1602 if (if_is_deactivated(ifp)) 1603 continue; 1604 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1605 if (ifa == NULL) 1606 continue; 1607 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1608 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1609 return ifa; /* exact match */ 1610 if (ifa_maybe == NULL) { 1611 /* else keep the if with the right range */ 1612 ifa_maybe = ifa; 1613 } 1614 } 1615 pserialize_read_exit(s); 1616 return ifa_maybe; 1617 } 1618 #endif 1619 s = pserialize_read_enter(); 1620 IFNET_READER_FOREACH(ifp) { 1621 if (if_is_deactivated(ifp)) 1622 continue; 1623 IFADDR_FOREACH(ifa, ifp) { 1624 const char *cp, *cp2, *cp3; 1625 1626 if (ifa->ifa_addr->sa_family != af || 1627 ifa->ifa_netmask == NULL) 1628 next: continue; 1629 cp = addr_data; 1630 cp2 = ifa->ifa_addr->sa_data; 1631 cp3 = ifa->ifa_netmask->sa_data; 1632 cplim = (const char *)ifa->ifa_netmask + 1633 ifa->ifa_netmask->sa_len; 1634 while (cp3 < cplim) { 1635 if ((*cp++ ^ *cp2++) & *cp3++) { 1636 /* want to continue for() loop */ 1637 goto next; 1638 } 1639 } 1640 if (ifa_maybe == NULL || 1641 rt_refines(ifa->ifa_netmask, 1642 ifa_maybe->ifa_netmask)) 1643 ifa_maybe = ifa; 1644 } 1645 } 1646 pserialize_read_exit(s); 1647 return ifa_maybe; 1648 } 1649 1650 /* 1651 * Find the interface of the addresss. 1652 */ 1653 struct ifaddr * 1654 ifa_ifwithladdr(const struct sockaddr *addr) 1655 { 1656 struct ifaddr *ia; 1657 1658 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1659 (ia = ifa_ifwithnet(addr))) 1660 return ia; 1661 return NULL; 1662 } 1663 1664 /* 1665 * Find an interface using a specific address family 1666 */ 1667 struct ifaddr * 1668 ifa_ifwithaf(int af) 1669 { 1670 struct ifnet *ifp; 1671 struct ifaddr *ifa = NULL; 1672 int s; 1673 1674 s = pserialize_read_enter(); 1675 IFNET_READER_FOREACH(ifp) { 1676 if (if_is_deactivated(ifp)) 1677 continue; 1678 IFADDR_FOREACH(ifa, ifp) { 1679 if (ifa->ifa_addr->sa_family == af) 1680 goto out; 1681 } 1682 } 1683 out: 1684 pserialize_read_exit(s); 1685 return ifa; 1686 } 1687 1688 /* 1689 * Find an interface address specific to an interface best matching 1690 * a given address. 1691 */ 1692 struct ifaddr * 1693 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1694 { 1695 struct ifaddr *ifa; 1696 const char *cp, *cp2, *cp3; 1697 const char *cplim; 1698 struct ifaddr *ifa_maybe = 0; 1699 u_int af = addr->sa_family; 1700 1701 if (if_is_deactivated(ifp)) 1702 return NULL; 1703 1704 if (af >= AF_MAX) 1705 return NULL; 1706 1707 IFADDR_FOREACH(ifa, ifp) { 1708 if (ifa->ifa_addr->sa_family != af) 1709 continue; 1710 ifa_maybe = ifa; 1711 if (ifa->ifa_netmask == NULL) { 1712 if (equal(addr, ifa->ifa_addr) || 1713 (ifa->ifa_dstaddr && 1714 equal(addr, ifa->ifa_dstaddr))) 1715 return ifa; 1716 continue; 1717 } 1718 cp = addr->sa_data; 1719 cp2 = ifa->ifa_addr->sa_data; 1720 cp3 = ifa->ifa_netmask->sa_data; 1721 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1722 for (; cp3 < cplim; cp3++) { 1723 if ((*cp++ ^ *cp2++) & *cp3) 1724 break; 1725 } 1726 if (cp3 == cplim) 1727 return ifa; 1728 } 1729 return ifa_maybe; 1730 } 1731 1732 /* 1733 * Default action when installing a route with a Link Level gateway. 1734 * Lookup an appropriate real ifa to point to. 1735 * This should be moved to /sys/net/link.c eventually. 1736 */ 1737 void 1738 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1739 { 1740 struct ifaddr *ifa; 1741 const struct sockaddr *dst; 1742 struct ifnet *ifp; 1743 1744 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1745 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1746 return; 1747 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1748 rt_replace_ifa(rt, ifa); 1749 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1750 ifa->ifa_rtrequest(cmd, rt, info); 1751 } 1752 } 1753 1754 /* 1755 * bitmask macros to manage a densely packed link_state change queue. 1756 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 1757 * LINK_STATE_UP(2) we need 2 bits for each state change. 1758 * As a state change to store is 0, treat all bits set as an unset item. 1759 */ 1760 #define LQ_ITEM_BITS 2 1761 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 1762 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 1763 #define LINK_STATE_UNSET LQ_ITEM_MASK 1764 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 1765 #define LQ_STORE(q, i, v) \ 1766 do { \ 1767 (q) &= ~LQ_MASK((i)); \ 1768 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 1769 } while (0 /* CONSTCOND */) 1770 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 1771 #define LQ_POP(q, v) \ 1772 do { \ 1773 (v) = LQ_ITEM((q), 0); \ 1774 (q) >>= LQ_ITEM_BITS; \ 1775 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1776 } while (0 /* CONSTCOND */) 1777 #define LQ_PUSH(q, v) \ 1778 do { \ 1779 (q) >>= LQ_ITEM_BITS; \ 1780 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1781 } while (0 /* CONSTCOND */) 1782 #define LQ_FIND_UNSET(q, i) \ 1783 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 1784 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 1785 break; \ 1786 } 1787 /* 1788 * Handle a change in the interface link state and 1789 * queue notifications. 1790 */ 1791 void 1792 if_link_state_change(struct ifnet *ifp, int link_state) 1793 { 1794 int s, idx; 1795 1796 KASSERTMSG(if_is_link_state_changeable(ifp), 1797 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 1798 ifp->if_xname, ifp->if_extflags); 1799 1800 /* Ensure change is to a valid state */ 1801 switch (link_state) { 1802 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 1803 case LINK_STATE_DOWN: /* FALLTHROUGH */ 1804 case LINK_STATE_UP: 1805 break; 1806 default: 1807 #ifdef DEBUG 1808 printf("%s: invalid link state %d\n", 1809 ifp->if_xname, link_state); 1810 #endif 1811 return; 1812 } 1813 1814 s = splnet(); 1815 1816 /* Find the last unset event in the queue. */ 1817 LQ_FIND_UNSET(ifp->if_link_queue, idx); 1818 1819 /* 1820 * Ensure link_state doesn't match the last event in the queue. 1821 * ifp->if_link_state is not checked and set here because 1822 * that would present an inconsistent picture to the system. 1823 */ 1824 if (idx != 0 && 1825 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 1826 goto out; 1827 1828 /* Handle queue overflow. */ 1829 if (idx == LQ_MAX(ifp->if_link_queue)) { 1830 uint8_t lost; 1831 1832 /* 1833 * The DOWN state must be protected from being pushed off 1834 * the queue to ensure that userland will always be 1835 * in a sane state. 1836 * Because DOWN is protected, there is no need to protect 1837 * UNKNOWN. 1838 * It should be invalid to change from any other state to 1839 * UNKNOWN anyway ... 1840 */ 1841 lost = LQ_ITEM(ifp->if_link_queue, 0); 1842 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 1843 if (lost == LINK_STATE_DOWN) { 1844 lost = LQ_ITEM(ifp->if_link_queue, 0); 1845 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 1846 } 1847 printf("%s: lost link state change %s\n", 1848 ifp->if_xname, 1849 lost == LINK_STATE_UP ? "UP" : 1850 lost == LINK_STATE_DOWN ? "DOWN" : 1851 "UNKNOWN"); 1852 } else 1853 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 1854 1855 softint_schedule(ifp->if_link_si); 1856 1857 out: 1858 splx(s); 1859 } 1860 1861 /* 1862 * Handle interface link state change notifications. 1863 * Must be called at splnet(). 1864 */ 1865 static void 1866 if_link_state_change0(struct ifnet *ifp, int link_state) 1867 { 1868 struct domain *dp; 1869 1870 /* Ensure the change is still valid. */ 1871 if (ifp->if_link_state == link_state) 1872 return; 1873 1874 #ifdef DEBUG 1875 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1876 link_state == LINK_STATE_UP ? "UP" : 1877 link_state == LINK_STATE_DOWN ? "DOWN" : 1878 "UNKNOWN", 1879 ifp->if_link_state == LINK_STATE_UP ? "UP" : 1880 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 1881 "UNKNOWN"); 1882 #endif 1883 1884 /* 1885 * When going from UNKNOWN to UP, we need to mark existing 1886 * addresses as tentative and restart DAD as we may have 1887 * erroneously not found a duplicate. 1888 * 1889 * This needs to happen before rt_ifmsg to avoid a race where 1890 * listeners would have an address and expect it to work right 1891 * away. 1892 */ 1893 if (link_state == LINK_STATE_UP && 1894 ifp->if_link_state == LINK_STATE_UNKNOWN) 1895 { 1896 DOMAIN_FOREACH(dp) { 1897 if (dp->dom_if_link_state_change != NULL) 1898 dp->dom_if_link_state_change(ifp, 1899 LINK_STATE_DOWN); 1900 } 1901 } 1902 1903 ifp->if_link_state = link_state; 1904 1905 /* Notify that the link state has changed. */ 1906 rt_ifmsg(ifp); 1907 1908 #if NCARP > 0 1909 if (ifp->if_carp) 1910 carp_carpdev_state(ifp); 1911 #endif 1912 1913 DOMAIN_FOREACH(dp) { 1914 if (dp->dom_if_link_state_change != NULL) 1915 dp->dom_if_link_state_change(ifp, link_state); 1916 } 1917 } 1918 1919 /* 1920 * Process the interface link state change queue. 1921 */ 1922 static void 1923 if_link_state_change_si(void *arg) 1924 { 1925 struct ifnet *ifp = arg; 1926 int s; 1927 uint8_t state; 1928 1929 s = splnet(); 1930 1931 /* Pop a link state change from the queue and process it. */ 1932 LQ_POP(ifp->if_link_queue, state); 1933 if_link_state_change0(ifp, state); 1934 1935 /* If there is a link state change to come, schedule it. */ 1936 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 1937 softint_schedule(ifp->if_link_si); 1938 1939 splx(s); 1940 } 1941 1942 /* 1943 * Default action when installing a local route on a point-to-point 1944 * interface. 1945 */ 1946 void 1947 p2p_rtrequest(int req, struct rtentry *rt, 1948 __unused const struct rt_addrinfo *info) 1949 { 1950 struct ifnet *ifp = rt->rt_ifp; 1951 struct ifaddr *ifa, *lo0ifa; 1952 1953 switch (req) { 1954 case RTM_ADD: 1955 if ((rt->rt_flags & RTF_LOCAL) == 0) 1956 break; 1957 1958 IFADDR_FOREACH(ifa, ifp) { 1959 if (equal(rt_getkey(rt), ifa->ifa_addr)) 1960 break; 1961 } 1962 if (ifa == NULL) 1963 break; 1964 1965 /* 1966 * Ensure lo0 has an address of the same family. 1967 */ 1968 IFADDR_FOREACH(lo0ifa, lo0ifp) { 1969 if (lo0ifa->ifa_addr->sa_family == 1970 ifa->ifa_addr->sa_family) 1971 break; 1972 } 1973 if (lo0ifa == NULL) 1974 break; 1975 1976 rt->rt_ifp = lo0ifp; 1977 1978 /* 1979 * Make sure to set rt->rt_ifa to the interface 1980 * address we are using, otherwise we will have trouble 1981 * with source address selection. 1982 */ 1983 if (ifa != rt->rt_ifa) 1984 rt_replace_ifa(rt, ifa); 1985 break; 1986 case RTM_DELETE: 1987 default: 1988 break; 1989 } 1990 } 1991 1992 /* 1993 * Mark an interface down and notify protocols of 1994 * the transition. 1995 * NOTE: must be called at splsoftnet or equivalent. 1996 */ 1997 void 1998 if_down(struct ifnet *ifp) 1999 { 2000 struct ifaddr *ifa; 2001 struct domain *dp; 2002 2003 ifp->if_flags &= ~IFF_UP; 2004 nanotime(&ifp->if_lastchange); 2005 IFADDR_FOREACH(ifa, ifp) 2006 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2007 IFQ_PURGE(&ifp->if_snd); 2008 #if NCARP > 0 2009 if (ifp->if_carp) 2010 carp_carpdev_state(ifp); 2011 #endif 2012 rt_ifmsg(ifp); 2013 DOMAIN_FOREACH(dp) { 2014 if (dp->dom_if_down) 2015 dp->dom_if_down(ifp); 2016 } 2017 } 2018 2019 /* 2020 * Mark an interface up and notify protocols of 2021 * the transition. 2022 * NOTE: must be called at splsoftnet or equivalent. 2023 */ 2024 void 2025 if_up(struct ifnet *ifp) 2026 { 2027 #ifdef notyet 2028 struct ifaddr *ifa; 2029 #endif 2030 struct domain *dp; 2031 2032 ifp->if_flags |= IFF_UP; 2033 nanotime(&ifp->if_lastchange); 2034 #ifdef notyet 2035 /* this has no effect on IP, and will kill all ISO connections XXX */ 2036 IFADDR_FOREACH(ifa, ifp) 2037 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2038 #endif 2039 #if NCARP > 0 2040 if (ifp->if_carp) 2041 carp_carpdev_state(ifp); 2042 #endif 2043 rt_ifmsg(ifp); 2044 DOMAIN_FOREACH(dp) { 2045 if (dp->dom_if_up) 2046 dp->dom_if_up(ifp); 2047 } 2048 } 2049 2050 /* 2051 * Handle interface slowtimo timer routine. Called 2052 * from softclock, we decrement timer (if set) and 2053 * call the appropriate interface routine on expiration. 2054 */ 2055 static void 2056 if_slowtimo(void *arg) 2057 { 2058 void (*slowtimo)(struct ifnet *); 2059 struct ifnet *ifp = arg; 2060 int s; 2061 2062 slowtimo = ifp->if_slowtimo; 2063 if (__predict_false(slowtimo == NULL)) 2064 return; 2065 2066 s = splnet(); 2067 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2068 (*slowtimo)(ifp); 2069 2070 splx(s); 2071 2072 if (__predict_true(ifp->if_slowtimo != NULL)) 2073 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2074 } 2075 2076 /* 2077 * Set/clear promiscuous mode on interface ifp based on the truth value 2078 * of pswitch. The calls are reference counted so that only the first 2079 * "on" request actually has an effect, as does the final "off" request. 2080 * Results are undefined if the "off" and "on" requests are not matched. 2081 */ 2082 int 2083 ifpromisc(struct ifnet *ifp, int pswitch) 2084 { 2085 int pcount, ret; 2086 short nflags; 2087 2088 pcount = ifp->if_pcount; 2089 if (pswitch) { 2090 /* 2091 * Allow the device to be "placed" into promiscuous 2092 * mode even if it is not configured up. It will 2093 * consult IFF_PROMISC when it is brought up. 2094 */ 2095 if (ifp->if_pcount++ != 0) 2096 return 0; 2097 nflags = ifp->if_flags | IFF_PROMISC; 2098 } else { 2099 if (--ifp->if_pcount > 0) 2100 return 0; 2101 nflags = ifp->if_flags & ~IFF_PROMISC; 2102 } 2103 ret = if_flags_set(ifp, nflags); 2104 /* Restore interface state if not successful. */ 2105 if (ret != 0) { 2106 ifp->if_pcount = pcount; 2107 } 2108 return ret; 2109 } 2110 2111 /* 2112 * Map interface name to 2113 * interface structure pointer. 2114 */ 2115 struct ifnet * 2116 ifunit(const char *name) 2117 { 2118 struct ifnet *ifp; 2119 const char *cp = name; 2120 u_int unit = 0; 2121 u_int i; 2122 int s; 2123 2124 /* 2125 * If the entire name is a number, treat it as an ifindex. 2126 */ 2127 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2128 unit = unit * 10 + (*cp - '0'); 2129 } 2130 2131 /* 2132 * If the number took all of the name, then it's a valid ifindex. 2133 */ 2134 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2135 if (unit >= if_indexlim) 2136 return NULL; 2137 ifp = ifindex2ifnet[unit]; 2138 if (ifp == NULL || if_is_deactivated(ifp)) 2139 return NULL; 2140 return ifp; 2141 } 2142 2143 ifp = NULL; 2144 s = pserialize_read_enter(); 2145 IFNET_READER_FOREACH(ifp) { 2146 if (if_is_deactivated(ifp)) 2147 continue; 2148 if (strcmp(ifp->if_xname, name) == 0) 2149 goto out; 2150 } 2151 out: 2152 pserialize_read_exit(s); 2153 return ifp; 2154 } 2155 2156 /* 2157 * Get a reference of an ifnet object by an interface name. 2158 * The returned reference is protected by psref(9). The caller 2159 * must release a returned reference by if_put after use. 2160 */ 2161 struct ifnet * 2162 if_get(const char *name, struct psref *psref) 2163 { 2164 struct ifnet *ifp; 2165 const char *cp = name; 2166 u_int unit = 0; 2167 u_int i; 2168 int s; 2169 2170 /* 2171 * If the entire name is a number, treat it as an ifindex. 2172 */ 2173 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2174 unit = unit * 10 + (*cp - '0'); 2175 } 2176 2177 /* 2178 * If the number took all of the name, then it's a valid ifindex. 2179 */ 2180 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2181 if (unit >= if_indexlim) 2182 return NULL; 2183 ifp = ifindex2ifnet[unit]; 2184 if (ifp == NULL || if_is_deactivated(ifp)) 2185 return NULL; 2186 return ifp; 2187 } 2188 2189 ifp = NULL; 2190 s = pserialize_read_enter(); 2191 IFNET_READER_FOREACH(ifp) { 2192 if (if_is_deactivated(ifp)) 2193 continue; 2194 if (strcmp(ifp->if_xname, name) == 0) { 2195 psref_acquire(psref, &ifp->if_psref, 2196 ifnet_psref_class); 2197 goto out; 2198 } 2199 } 2200 out: 2201 pserialize_read_exit(s); 2202 return ifp; 2203 } 2204 2205 /* 2206 * Release a reference of an ifnet object given by if_get or 2207 * if_get_byindex. 2208 */ 2209 void 2210 if_put(const struct ifnet *ifp, struct psref *psref) 2211 { 2212 2213 if (ifp == NULL) 2214 return; 2215 2216 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2217 } 2218 2219 ifnet_t * 2220 if_byindex(u_int idx) 2221 { 2222 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2223 } 2224 2225 /* 2226 * Get a reference of an ifnet object by an interface index. 2227 * The returned reference is protected by psref(9). The caller 2228 * must release a returned reference by if_put after use. 2229 */ 2230 ifnet_t * 2231 if_get_byindex(u_int idx, struct psref *psref) 2232 { 2233 ifnet_t *ifp; 2234 int s; 2235 2236 s = pserialize_read_enter(); 2237 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2238 if (__predict_true(ifp != NULL)) 2239 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2240 pserialize_read_exit(s); 2241 2242 return ifp; 2243 } 2244 2245 /* 2246 * XXX it's safe only if the passed ifp is guaranteed to not be freed, 2247 * for example the ifp is already held or some other object is held which 2248 * guarantes the ifp to not be freed indirectly. 2249 */ 2250 void 2251 if_acquire_NOMPSAFE(struct ifnet *ifp, struct psref *psref) 2252 { 2253 2254 KASSERT(ifp->if_index != 0); 2255 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2256 } 2257 2258 bool 2259 if_held(struct ifnet *ifp) 2260 { 2261 2262 return psref_held(&ifp->if_psref, ifnet_psref_class); 2263 } 2264 2265 2266 /* common */ 2267 int 2268 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2269 { 2270 int s; 2271 struct ifreq *ifr; 2272 struct ifcapreq *ifcr; 2273 struct ifdatareq *ifdr; 2274 2275 switch (cmd) { 2276 case SIOCSIFCAP: 2277 ifcr = data; 2278 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2279 return EINVAL; 2280 2281 if (ifcr->ifcr_capenable == ifp->if_capenable) 2282 return 0; 2283 2284 ifp->if_capenable = ifcr->ifcr_capenable; 2285 2286 /* Pre-compute the checksum flags mask. */ 2287 ifp->if_csum_flags_tx = 0; 2288 ifp->if_csum_flags_rx = 0; 2289 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2290 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2291 } 2292 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2293 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2294 } 2295 2296 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2297 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2298 } 2299 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2300 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2301 } 2302 2303 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2304 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2305 } 2306 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2307 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2308 } 2309 2310 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2311 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2312 } 2313 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2314 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2315 } 2316 2317 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2318 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2319 } 2320 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2321 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2322 } 2323 if (ifp->if_flags & IFF_UP) 2324 return ENETRESET; 2325 return 0; 2326 case SIOCSIFFLAGS: 2327 ifr = data; 2328 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2329 s = splnet(); 2330 if_down(ifp); 2331 splx(s); 2332 } 2333 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2334 s = splnet(); 2335 if_up(ifp); 2336 splx(s); 2337 } 2338 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2339 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2340 break; 2341 case SIOCGIFFLAGS: 2342 ifr = data; 2343 ifr->ifr_flags = ifp->if_flags; 2344 break; 2345 2346 case SIOCGIFMETRIC: 2347 ifr = data; 2348 ifr->ifr_metric = ifp->if_metric; 2349 break; 2350 2351 case SIOCGIFMTU: 2352 ifr = data; 2353 ifr->ifr_mtu = ifp->if_mtu; 2354 break; 2355 2356 case SIOCGIFDLT: 2357 ifr = data; 2358 ifr->ifr_dlt = ifp->if_dlt; 2359 break; 2360 2361 case SIOCGIFCAP: 2362 ifcr = data; 2363 ifcr->ifcr_capabilities = ifp->if_capabilities; 2364 ifcr->ifcr_capenable = ifp->if_capenable; 2365 break; 2366 2367 case SIOCSIFMETRIC: 2368 ifr = data; 2369 ifp->if_metric = ifr->ifr_metric; 2370 break; 2371 2372 case SIOCGIFDATA: 2373 ifdr = data; 2374 ifdr->ifdr_data = ifp->if_data; 2375 break; 2376 2377 case SIOCGIFINDEX: 2378 ifr = data; 2379 ifr->ifr_index = ifp->if_index; 2380 break; 2381 2382 case SIOCZIFDATA: 2383 ifdr = data; 2384 ifdr->ifdr_data = ifp->if_data; 2385 /* 2386 * Assumes that the volatile counters that can be 2387 * zero'ed are at the end of if_data. 2388 */ 2389 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2390 offsetof(struct if_data, ifi_ipackets)); 2391 /* 2392 * The memset() clears to the bottm of if_data. In the area, 2393 * if_lastchange is included. Please be careful if new entry 2394 * will be added into if_data or rewite this. 2395 * 2396 * And also, update if_lastchnage. 2397 */ 2398 getnanotime(&ifp->if_lastchange); 2399 break; 2400 case SIOCSIFMTU: 2401 ifr = data; 2402 if (ifp->if_mtu == ifr->ifr_mtu) 2403 break; 2404 ifp->if_mtu = ifr->ifr_mtu; 2405 /* 2406 * If the link MTU changed, do network layer specific procedure. 2407 */ 2408 #ifdef INET6 2409 if (in6_present) 2410 nd6_setmtu(ifp); 2411 #endif 2412 return ENETRESET; 2413 default: 2414 return ENOTTY; 2415 } 2416 return 0; 2417 } 2418 2419 int 2420 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2421 { 2422 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2423 struct ifaddr *ifa; 2424 const struct sockaddr *any, *sa; 2425 union { 2426 struct sockaddr sa; 2427 struct sockaddr_storage ss; 2428 } u, v; 2429 2430 switch (cmd) { 2431 case SIOCSIFADDRPREF: 2432 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2433 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2434 NULL) != 0) 2435 return EPERM; 2436 case SIOCGIFADDRPREF: 2437 break; 2438 default: 2439 return EOPNOTSUPP; 2440 } 2441 2442 /* sanity checks */ 2443 if (data == NULL || ifp == NULL) { 2444 panic("invalid argument to %s", __func__); 2445 /*NOTREACHED*/ 2446 } 2447 2448 /* address must be specified on ADD and DELETE */ 2449 sa = sstocsa(&ifap->ifap_addr); 2450 if (sa->sa_family != sofamily(so)) 2451 return EINVAL; 2452 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2453 return EINVAL; 2454 2455 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2456 2457 IFADDR_FOREACH(ifa, ifp) { 2458 if (ifa->ifa_addr->sa_family != sa->sa_family) 2459 continue; 2460 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2461 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2462 break; 2463 } 2464 if (ifa == NULL) 2465 return EADDRNOTAVAIL; 2466 2467 switch (cmd) { 2468 case SIOCSIFADDRPREF: 2469 ifa->ifa_preference = ifap->ifap_preference; 2470 return 0; 2471 case SIOCGIFADDRPREF: 2472 /* fill in the if_laddrreq structure */ 2473 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2474 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2475 ifap->ifap_preference = ifa->ifa_preference; 2476 return 0; 2477 default: 2478 return EOPNOTSUPP; 2479 } 2480 } 2481 2482 /* 2483 * Interface ioctls. 2484 */ 2485 static int 2486 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2487 { 2488 struct ifnet *ifp; 2489 struct ifreq *ifr; 2490 int error = 0; 2491 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2492 u_long ocmd = cmd; 2493 #endif 2494 short oif_flags; 2495 #ifdef COMPAT_OIFREQ 2496 struct ifreq ifrb; 2497 struct oifreq *oifr = NULL; 2498 #endif 2499 int r; 2500 struct psref psref; 2501 int bound; 2502 2503 switch (cmd) { 2504 #ifdef COMPAT_OIFREQ 2505 case OSIOCGIFCONF: 2506 case OOSIOCGIFCONF: 2507 return compat_ifconf(cmd, data); 2508 #endif 2509 #ifdef COMPAT_OIFDATA 2510 case OSIOCGIFDATA: 2511 case OSIOCZIFDATA: 2512 return compat_ifdatareq(l, cmd, data); 2513 #endif 2514 case SIOCGIFCONF: 2515 return ifconf(cmd, data); 2516 case SIOCINITIFADDR: 2517 return EPERM; 2518 } 2519 2520 #ifdef COMPAT_OIFREQ 2521 cmd = compat_cvtcmd(cmd); 2522 if (cmd != ocmd) { 2523 oifr = data; 2524 data = ifr = &ifrb; 2525 ifreqo2n(oifr, ifr); 2526 } else 2527 #endif 2528 ifr = data; 2529 2530 switch (cmd) { 2531 case SIOCIFCREATE: 2532 case SIOCIFDESTROY: 2533 bound = curlwp_bind(); 2534 if (l != NULL) { 2535 ifp = if_get(ifr->ifr_name, &psref); 2536 error = kauth_authorize_network(l->l_cred, 2537 KAUTH_NETWORK_INTERFACE, 2538 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2539 (void *)cmd, NULL); 2540 if (ifp != NULL) 2541 if_put(ifp, &psref); 2542 if (error != 0) { 2543 curlwp_bindx(bound); 2544 return error; 2545 } 2546 } 2547 mutex_enter(&if_clone_mtx); 2548 r = (cmd == SIOCIFCREATE) ? 2549 if_clone_create(ifr->ifr_name) : 2550 if_clone_destroy(ifr->ifr_name); 2551 mutex_exit(&if_clone_mtx); 2552 curlwp_bindx(bound); 2553 return r; 2554 2555 case SIOCIFGCLONERS: 2556 { 2557 struct if_clonereq *req = (struct if_clonereq *)data; 2558 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2559 &req->ifcr_total); 2560 } 2561 } 2562 2563 bound = curlwp_bind(); 2564 ifp = if_get(ifr->ifr_name, &psref); 2565 if (ifp == NULL) { 2566 curlwp_bindx(bound); 2567 return ENXIO; 2568 } 2569 2570 switch (cmd) { 2571 case SIOCALIFADDR: 2572 case SIOCDLIFADDR: 2573 case SIOCSIFADDRPREF: 2574 case SIOCSIFFLAGS: 2575 case SIOCSIFCAP: 2576 case SIOCSIFMETRIC: 2577 case SIOCZIFDATA: 2578 case SIOCSIFMTU: 2579 case SIOCSIFPHYADDR: 2580 case SIOCDIFPHYADDR: 2581 #ifdef INET6 2582 case SIOCSIFPHYADDR_IN6: 2583 #endif 2584 case SIOCSLIFPHYADDR: 2585 case SIOCADDMULTI: 2586 case SIOCDELMULTI: 2587 case SIOCSIFMEDIA: 2588 case SIOCSDRVSPEC: 2589 case SIOCG80211: 2590 case SIOCS80211: 2591 case SIOCS80211NWID: 2592 case SIOCS80211NWKEY: 2593 case SIOCS80211POWER: 2594 case SIOCS80211BSSID: 2595 case SIOCS80211CHANNEL: 2596 case SIOCSLINKSTR: 2597 if (l != NULL) { 2598 error = kauth_authorize_network(l->l_cred, 2599 KAUTH_NETWORK_INTERFACE, 2600 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2601 (void *)cmd, NULL); 2602 if (error != 0) 2603 goto out; 2604 } 2605 } 2606 2607 oif_flags = ifp->if_flags; 2608 2609 mutex_enter(ifp->if_ioctl_lock); 2610 2611 error = (*ifp->if_ioctl)(ifp, cmd, data); 2612 if (error != ENOTTY) 2613 ; 2614 else if (so->so_proto == NULL) 2615 error = EOPNOTSUPP; 2616 else { 2617 #ifdef COMPAT_OSOCK 2618 error = compat_ifioctl(so, ocmd, cmd, data, l); 2619 #else 2620 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2621 cmd, data, ifp); 2622 #endif 2623 } 2624 2625 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2626 if ((ifp->if_flags & IFF_UP) != 0) { 2627 int s = splnet(); 2628 if_up(ifp); 2629 splx(s); 2630 } 2631 } 2632 #ifdef COMPAT_OIFREQ 2633 if (cmd != ocmd) 2634 ifreqn2o(oifr, ifr); 2635 #endif 2636 2637 mutex_exit(ifp->if_ioctl_lock); 2638 out: 2639 if_put(ifp, &psref); 2640 curlwp_bindx(bound); 2641 return error; 2642 } 2643 2644 /* 2645 * Return interface configuration 2646 * of system. List may be used 2647 * in later ioctl's (above) to get 2648 * other information. 2649 * 2650 * Each record is a struct ifreq. Before the addition of 2651 * sockaddr_storage, the API rule was that sockaddr flavors that did 2652 * not fit would extend beyond the struct ifreq, with the next struct 2653 * ifreq starting sa_len beyond the struct sockaddr. Because the 2654 * union in struct ifreq includes struct sockaddr_storage, every kind 2655 * of sockaddr must fit. Thus, there are no longer any overlength 2656 * records. 2657 * 2658 * Records are added to the user buffer if they fit, and ifc_len is 2659 * adjusted to the length that was written. Thus, the user is only 2660 * assured of getting the complete list if ifc_len on return is at 2661 * least sizeof(struct ifreq) less than it was on entry. 2662 * 2663 * If the user buffer pointer is NULL, this routine copies no data and 2664 * returns the amount of space that would be needed. 2665 * 2666 * Invariants: 2667 * ifrp points to the next part of the user's buffer to be used. If 2668 * ifrp != NULL, space holds the number of bytes remaining that we may 2669 * write at ifrp. Otherwise, space holds the number of bytes that 2670 * would have been written had there been adequate space. 2671 */ 2672 /*ARGSUSED*/ 2673 static int 2674 ifconf(u_long cmd, void *data) 2675 { 2676 struct ifconf *ifc = (struct ifconf *)data; 2677 struct ifnet *ifp; 2678 struct ifaddr *ifa; 2679 struct ifreq ifr, *ifrp = NULL; 2680 int space = 0, error = 0; 2681 const int sz = (int)sizeof(struct ifreq); 2682 const bool docopy = ifc->ifc_req != NULL; 2683 int s; 2684 int bound; 2685 struct psref psref; 2686 2687 if (docopy) { 2688 space = ifc->ifc_len; 2689 ifrp = ifc->ifc_req; 2690 } 2691 2692 bound = curlwp_bind(); 2693 s = pserialize_read_enter(); 2694 IFNET_READER_FOREACH(ifp) { 2695 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 2696 pserialize_read_exit(s); 2697 2698 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2699 sizeof(ifr.ifr_name)); 2700 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 2701 error = ENAMETOOLONG; 2702 goto release_exit; 2703 } 2704 if (IFADDR_EMPTY(ifp)) { 2705 /* Interface with no addresses - send zero sockaddr. */ 2706 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2707 if (!docopy) { 2708 space += sz; 2709 continue; 2710 } 2711 if (space >= sz) { 2712 error = copyout(&ifr, ifrp, sz); 2713 if (error != 0) 2714 goto release_exit; 2715 ifrp++; 2716 space -= sz; 2717 } 2718 } 2719 2720 IFADDR_FOREACH(ifa, ifp) { 2721 struct sockaddr *sa = ifa->ifa_addr; 2722 /* all sockaddrs must fit in sockaddr_storage */ 2723 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2724 2725 if (!docopy) { 2726 space += sz; 2727 continue; 2728 } 2729 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2730 if (space >= sz) { 2731 error = copyout(&ifr, ifrp, sz); 2732 if (error != 0) 2733 goto release_exit; 2734 ifrp++; space -= sz; 2735 } 2736 } 2737 2738 s = pserialize_read_enter(); 2739 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2740 } 2741 pserialize_read_exit(s); 2742 curlwp_bindx(bound); 2743 2744 if (docopy) { 2745 KASSERT(0 <= space && space <= ifc->ifc_len); 2746 ifc->ifc_len -= space; 2747 } else { 2748 KASSERT(space >= 0); 2749 ifc->ifc_len = space; 2750 } 2751 return (0); 2752 2753 release_exit: 2754 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2755 curlwp_bindx(bound); 2756 return error; 2757 } 2758 2759 int 2760 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2761 { 2762 uint8_t len; 2763 #ifdef COMPAT_OIFREQ 2764 struct ifreq ifrb; 2765 struct oifreq *oifr = NULL; 2766 u_long ocmd = cmd; 2767 cmd = compat_cvtcmd(cmd); 2768 if (cmd != ocmd) { 2769 oifr = (struct oifreq *)(void *)ifr; 2770 ifr = &ifrb; 2771 ifreqo2n(oifr, ifr); 2772 len = sizeof(oifr->ifr_addr); 2773 } else 2774 #endif 2775 len = sizeof(ifr->ifr_ifru.ifru_space); 2776 2777 if (len < sa->sa_len) 2778 return EFBIG; 2779 2780 memset(&ifr->ifr_addr, 0, len); 2781 sockaddr_copy(&ifr->ifr_addr, len, sa); 2782 2783 #ifdef COMPAT_OIFREQ 2784 if (cmd != ocmd) 2785 ifreqn2o(oifr, ifr); 2786 #endif 2787 return 0; 2788 } 2789 2790 /* 2791 * wrapper function for the drivers which doesn't have if_transmit(). 2792 */ 2793 static int 2794 if_transmit(struct ifnet *ifp, struct mbuf *m) 2795 { 2796 int s, error; 2797 2798 s = splnet(); 2799 2800 /* 2801 * If NET_MPSAFE is not defined , IFQ_LOCK() is nop. 2802 * use KERNEL_LOCK instead of ifq_lock. 2803 */ 2804 #ifndef NET_MPSAFE 2805 KERNEL_LOCK(1, NULL); 2806 #endif 2807 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2808 #ifndef NET_MPSAFE 2809 KERNEL_UNLOCK_ONE(NULL); 2810 #endif 2811 if (error != 0) { 2812 /* mbuf is already freed */ 2813 goto out; 2814 } 2815 2816 ifp->if_obytes += m->m_pkthdr.len;; 2817 if (m->m_flags & M_MCAST) 2818 ifp->if_omcasts++; 2819 2820 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2821 if_start_lock(ifp); 2822 out: 2823 splx(s); 2824 2825 return error; 2826 } 2827 2828 int 2829 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 2830 { 2831 int error; 2832 2833 #ifdef ALTQ 2834 KERNEL_LOCK(1, NULL); 2835 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 2836 error = if_transmit(ifp, m); 2837 KERNEL_UNLOCK_ONE(NULL); 2838 } else { 2839 KERNEL_UNLOCK_ONE(NULL); 2840 error = (*ifp->if_transmit)(ifp, m); 2841 } 2842 #else /* !ALTQ */ 2843 error = (*ifp->if_transmit)(ifp, m); 2844 #endif /* !ALTQ */ 2845 2846 return error; 2847 } 2848 2849 /* 2850 * Queue message on interface, and start output if interface 2851 * not yet active. 2852 */ 2853 int 2854 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 2855 { 2856 2857 return if_transmit_lock(ifp, m); 2858 } 2859 2860 /* 2861 * Queue message on interface, possibly using a second fast queue 2862 */ 2863 int 2864 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 2865 { 2866 int error = 0; 2867 2868 if (ifq != NULL 2869 #ifdef ALTQ 2870 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2871 #endif 2872 ) { 2873 if (IF_QFULL(ifq)) { 2874 IF_DROP(&ifp->if_snd); 2875 m_freem(m); 2876 if (error == 0) 2877 error = ENOBUFS; 2878 } else 2879 IF_ENQUEUE(ifq, m); 2880 } else 2881 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2882 if (error != 0) { 2883 ++ifp->if_oerrors; 2884 return error; 2885 } 2886 return 0; 2887 } 2888 2889 int 2890 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2891 { 2892 int rc; 2893 2894 if (ifp->if_initaddr != NULL) 2895 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2896 else if (src || 2897 /* FIXME: may not hold if_ioctl_lock */ 2898 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2899 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2900 2901 return rc; 2902 } 2903 2904 int 2905 if_do_dad(struct ifnet *ifp) 2906 { 2907 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2908 return 0; 2909 2910 switch (ifp->if_type) { 2911 case IFT_FAITH: 2912 /* 2913 * These interfaces do not have the IFF_LOOPBACK flag, 2914 * but loop packets back. We do not have to do DAD on such 2915 * interfaces. We should even omit it, because loop-backed 2916 * responses would confuse the DAD procedure. 2917 */ 2918 return 0; 2919 default: 2920 /* 2921 * Our DAD routine requires the interface up and running. 2922 * However, some interfaces can be up before the RUNNING 2923 * status. Additionaly, users may try to assign addresses 2924 * before the interface becomes up (or running). 2925 * We simply skip DAD in such a case as a work around. 2926 * XXX: we should rather mark "tentative" on such addresses, 2927 * and do DAD after the interface becomes ready. 2928 */ 2929 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 2930 (IFF_UP|IFF_RUNNING)) 2931 return 0; 2932 2933 return 1; 2934 } 2935 } 2936 2937 int 2938 if_flags_set(ifnet_t *ifp, const short flags) 2939 { 2940 int rc; 2941 2942 if (ifp->if_setflags != NULL) 2943 rc = (*ifp->if_setflags)(ifp, flags); 2944 else { 2945 short cantflags, chgdflags; 2946 struct ifreq ifr; 2947 2948 chgdflags = ifp->if_flags ^ flags; 2949 cantflags = chgdflags & IFF_CANTCHANGE; 2950 2951 if (cantflags != 0) 2952 ifp->if_flags ^= cantflags; 2953 2954 /* Traditionally, we do not call if_ioctl after 2955 * setting/clearing only IFF_PROMISC if the interface 2956 * isn't IFF_UP. Uphold that tradition. 2957 */ 2958 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2959 return 0; 2960 2961 memset(&ifr, 0, sizeof(ifr)); 2962 2963 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2964 /* FIXME: may not hold if_ioctl_lock */ 2965 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2966 2967 if (rc != 0 && cantflags != 0) 2968 ifp->if_flags ^= cantflags; 2969 } 2970 2971 return rc; 2972 } 2973 2974 int 2975 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2976 { 2977 int rc; 2978 struct ifreq ifr; 2979 2980 if (ifp->if_mcastop != NULL) 2981 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2982 else { 2983 ifreq_setaddr(cmd, &ifr, sa); 2984 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2985 } 2986 2987 return rc; 2988 } 2989 2990 static void 2991 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2992 struct ifaltq *ifq) 2993 { 2994 const struct sysctlnode *cnode, *rnode; 2995 2996 if (sysctl_createv(clog, 0, NULL, &rnode, 2997 CTLFLAG_PERMANENT, 2998 CTLTYPE_NODE, "interfaces", 2999 SYSCTL_DESCR("Per-interface controls"), 3000 NULL, 0, NULL, 0, 3001 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3002 goto bad; 3003 3004 if (sysctl_createv(clog, 0, &rnode, &rnode, 3005 CTLFLAG_PERMANENT, 3006 CTLTYPE_NODE, ifname, 3007 SYSCTL_DESCR("Interface controls"), 3008 NULL, 0, NULL, 0, 3009 CTL_CREATE, CTL_EOL) != 0) 3010 goto bad; 3011 3012 if (sysctl_createv(clog, 0, &rnode, &rnode, 3013 CTLFLAG_PERMANENT, 3014 CTLTYPE_NODE, "sndq", 3015 SYSCTL_DESCR("Interface output queue controls"), 3016 NULL, 0, NULL, 0, 3017 CTL_CREATE, CTL_EOL) != 0) 3018 goto bad; 3019 3020 if (sysctl_createv(clog, 0, &rnode, &cnode, 3021 CTLFLAG_PERMANENT, 3022 CTLTYPE_INT, "len", 3023 SYSCTL_DESCR("Current output queue length"), 3024 NULL, 0, &ifq->ifq_len, 0, 3025 CTL_CREATE, CTL_EOL) != 0) 3026 goto bad; 3027 3028 if (sysctl_createv(clog, 0, &rnode, &cnode, 3029 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3030 CTLTYPE_INT, "maxlen", 3031 SYSCTL_DESCR("Maximum allowed output queue length"), 3032 NULL, 0, &ifq->ifq_maxlen, 0, 3033 CTL_CREATE, CTL_EOL) != 0) 3034 goto bad; 3035 3036 if (sysctl_createv(clog, 0, &rnode, &cnode, 3037 CTLFLAG_PERMANENT, 3038 CTLTYPE_INT, "drops", 3039 SYSCTL_DESCR("Packets dropped due to full output queue"), 3040 NULL, 0, &ifq->ifq_drops, 0, 3041 CTL_CREATE, CTL_EOL) != 0) 3042 goto bad; 3043 3044 return; 3045 bad: 3046 printf("%s: could not attach sysctl nodes\n", ifname); 3047 return; 3048 } 3049 3050 #if defined(INET) || defined(INET6) 3051 3052 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3053 static int \ 3054 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3055 { \ 3056 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3057 } 3058 3059 #if defined(INET) 3060 static int 3061 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3062 { 3063 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3064 } 3065 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3066 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3067 #endif 3068 3069 #if defined(INET6) 3070 static int 3071 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3072 { 3073 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3074 } 3075 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3076 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3077 #endif 3078 3079 static void 3080 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3081 { 3082 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3083 const char *pfname = NULL, *ipname = NULL; 3084 int ipn = 0, qid = 0; 3085 3086 switch (pf) { 3087 #if defined(INET) 3088 case PF_INET: 3089 len_func = sysctl_net_ip_pktq_items; 3090 maxlen_func = sysctl_net_ip_pktq_maxlen; 3091 drops_func = sysctl_net_ip_pktq_drops; 3092 pfname = "inet", ipn = IPPROTO_IP; 3093 ipname = "ip", qid = IPCTL_IFQ; 3094 break; 3095 #endif 3096 #if defined(INET6) 3097 case PF_INET6: 3098 len_func = sysctl_net_ip6_pktq_items; 3099 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3100 drops_func = sysctl_net_ip6_pktq_drops; 3101 pfname = "inet6", ipn = IPPROTO_IPV6; 3102 ipname = "ip6", qid = IPV6CTL_IFQ; 3103 break; 3104 #endif 3105 default: 3106 KASSERT(false); 3107 } 3108 3109 sysctl_createv(clog, 0, NULL, NULL, 3110 CTLFLAG_PERMANENT, 3111 CTLTYPE_NODE, pfname, NULL, 3112 NULL, 0, NULL, 0, 3113 CTL_NET, pf, CTL_EOL); 3114 sysctl_createv(clog, 0, NULL, NULL, 3115 CTLFLAG_PERMANENT, 3116 CTLTYPE_NODE, ipname, NULL, 3117 NULL, 0, NULL, 0, 3118 CTL_NET, pf, ipn, CTL_EOL); 3119 sysctl_createv(clog, 0, NULL, NULL, 3120 CTLFLAG_PERMANENT, 3121 CTLTYPE_NODE, "ifq", 3122 SYSCTL_DESCR("Protocol input queue controls"), 3123 NULL, 0, NULL, 0, 3124 CTL_NET, pf, ipn, qid, CTL_EOL); 3125 3126 sysctl_createv(clog, 0, NULL, NULL, 3127 CTLFLAG_PERMANENT, 3128 CTLTYPE_INT, "len", 3129 SYSCTL_DESCR("Current input queue length"), 3130 len_func, 0, NULL, 0, 3131 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3132 sysctl_createv(clog, 0, NULL, NULL, 3133 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3134 CTLTYPE_INT, "maxlen", 3135 SYSCTL_DESCR("Maximum allowed input queue length"), 3136 maxlen_func, 0, NULL, 0, 3137 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3138 sysctl_createv(clog, 0, NULL, NULL, 3139 CTLFLAG_PERMANENT, 3140 CTLTYPE_INT, "drops", 3141 SYSCTL_DESCR("Packets dropped due to full input queue"), 3142 drops_func, 0, NULL, 0, 3143 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3144 } 3145 #endif /* INET || INET6 */ 3146 3147 static int 3148 if_sdl_sysctl(SYSCTLFN_ARGS) 3149 { 3150 struct ifnet *ifp; 3151 const struct sockaddr_dl *sdl; 3152 struct psref psref; 3153 int error = 0; 3154 int bound; 3155 3156 if (namelen != 1) 3157 return EINVAL; 3158 3159 bound = curlwp_bind(); 3160 ifp = if_get_byindex(name[0], &psref); 3161 if (ifp == NULL) { 3162 error = ENODEV; 3163 goto out0; 3164 } 3165 3166 sdl = ifp->if_sadl; 3167 if (sdl == NULL) { 3168 *oldlenp = 0; 3169 goto out1; 3170 } 3171 3172 if (oldp == NULL) { 3173 *oldlenp = sdl->sdl_alen; 3174 goto out1; 3175 } 3176 3177 if (*oldlenp >= sdl->sdl_alen) 3178 *oldlenp = sdl->sdl_alen; 3179 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3180 out1: 3181 if_put(ifp, &psref); 3182 out0: 3183 curlwp_bindx(bound); 3184 return error; 3185 } 3186 3187 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 3188 { 3189 const struct sysctlnode *rnode = NULL; 3190 3191 sysctl_createv(clog, 0, NULL, &rnode, 3192 CTLFLAG_PERMANENT, 3193 CTLTYPE_NODE, "sdl", 3194 SYSCTL_DESCR("Get active link-layer address"), 3195 if_sdl_sysctl, 0, NULL, 0, 3196 CTL_NET, CTL_CREATE, CTL_EOL); 3197 } 3198
Indexes created Wed Sep 24 05:09:52 GMT 2025