if.c revision 1.342
1 /* $NetBSD: if.c,v 1.342 2016/06/20 06:41:30 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 1.342 2016/06/20 06:41:30 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_clone_create(const char *); 203 static int if_clone_destroy(const char *); 204 static void if_link_state_change_si(void *); 205 206 struct if_percpuq { 207 struct ifnet *ipq_ifp; 208 void *ipq_si; 209 struct percpu *ipq_ifqs; /* struct ifqueue */ 210 }; 211 212 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 213 214 static void if_percpuq_drops(void *, void *, struct cpu_info *); 215 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 216 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 217 struct if_percpuq *); 218 219 #if defined(INET) || defined(INET6) 220 static void sysctl_net_pktq_setup(struct sysctllog **, int); 221 #endif 222 223 static int 224 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 225 void *arg0, void *arg1, void *arg2, void *arg3) 226 { 227 int result; 228 enum kauth_network_req req; 229 230 result = KAUTH_RESULT_DEFER; 231 req = (enum kauth_network_req)arg1; 232 233 if (action != KAUTH_NETWORK_INTERFACE) 234 return result; 235 236 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 237 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 238 result = KAUTH_RESULT_ALLOW; 239 240 return result; 241 } 242 243 /* 244 * Network interface utility routines. 245 * 246 * Routines with ifa_ifwith* names take sockaddr *'s as 247 * parameters. 248 */ 249 void 250 ifinit(void) 251 { 252 #if defined(INET) 253 sysctl_net_pktq_setup(NULL, PF_INET); 254 #endif 255 #ifdef INET6 256 if (in6_present) 257 sysctl_net_pktq_setup(NULL, PF_INET6); 258 #endif 259 260 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 261 if_listener_cb, NULL); 262 263 /* interfaces are available, inform socket code */ 264 ifioctl = doifioctl; 265 } 266 267 /* 268 * XXX Initialization before configure(). 269 * XXX hack to get pfil_add_hook working in autoconf. 270 */ 271 void 272 ifinit1(void) 273 { 274 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 275 TAILQ_INIT(&ifnet_list); 276 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 277 ifnet_psz = pserialize_create(); 278 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 279 PSLIST_INIT(&ifnet_pslist); 280 if_indexlim = 8; 281 282 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 283 KASSERT(if_pfil != NULL); 284 285 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 286 etherinit(); 287 #endif 288 } 289 290 ifnet_t * 291 if_alloc(u_char type) 292 { 293 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 294 } 295 296 void 297 if_free(ifnet_t *ifp) 298 { 299 kmem_free(ifp, sizeof(ifnet_t)); 300 } 301 302 void 303 if_initname(struct ifnet *ifp, const char *name, int unit) 304 { 305 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 306 "%s%d", name, unit); 307 } 308 309 /* 310 * Null routines used while an interface is going away. These routines 311 * just return an error. 312 */ 313 314 int 315 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 316 const struct sockaddr *so, const struct rtentry *rt) 317 { 318 319 return ENXIO; 320 } 321 322 void 323 if_nullinput(struct ifnet *ifp, struct mbuf *m) 324 { 325 326 /* Nothing. */ 327 } 328 329 void 330 if_nullstart(struct ifnet *ifp) 331 { 332 333 /* Nothing. */ 334 } 335 336 int 337 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 338 { 339 340 return ENXIO; 341 } 342 343 int 344 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 345 { 346 347 return ENXIO; 348 } 349 350 int 351 if_nullinit(struct ifnet *ifp) 352 { 353 354 return ENXIO; 355 } 356 357 void 358 if_nullstop(struct ifnet *ifp, int disable) 359 { 360 361 /* Nothing. */ 362 } 363 364 void 365 if_nullslowtimo(struct ifnet *ifp) 366 { 367 368 /* Nothing. */ 369 } 370 371 void 372 if_nulldrain(struct ifnet *ifp) 373 { 374 375 /* Nothing. */ 376 } 377 378 void 379 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 380 { 381 struct ifaddr *ifa; 382 struct sockaddr_dl *sdl; 383 384 ifp->if_addrlen = addrlen; 385 if_alloc_sadl(ifp); 386 ifa = ifp->if_dl; 387 sdl = satosdl(ifa->ifa_addr); 388 389 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 390 if (factory) { 391 ifp->if_hwdl = ifp->if_dl; 392 ifaref(ifp->if_hwdl); 393 } 394 /* TBD routing socket */ 395 } 396 397 struct ifaddr * 398 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 399 { 400 unsigned socksize, ifasize; 401 int addrlen, namelen; 402 struct sockaddr_dl *mask, *sdl; 403 struct ifaddr *ifa; 404 405 namelen = strlen(ifp->if_xname); 406 addrlen = ifp->if_addrlen; 407 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 408 ifasize = sizeof(*ifa) + 2 * socksize; 409 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 410 411 sdl = (struct sockaddr_dl *)(ifa + 1); 412 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 413 414 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 415 ifp->if_xname, namelen, NULL, addrlen); 416 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 417 memset(&mask->sdl_data[0], 0xff, namelen); 418 ifa->ifa_rtrequest = link_rtrequest; 419 ifa->ifa_addr = (struct sockaddr *)sdl; 420 ifa->ifa_netmask = (struct sockaddr *)mask; 421 422 *sdlp = sdl; 423 424 return ifa; 425 } 426 427 static void 428 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 429 { 430 const struct sockaddr_dl *sdl; 431 432 ifp->if_dl = ifa; 433 ifaref(ifa); 434 sdl = satosdl(ifa->ifa_addr); 435 ifp->if_sadl = sdl; 436 } 437 438 /* 439 * Allocate the link level name for the specified interface. This 440 * is an attachment helper. It must be called after ifp->if_addrlen 441 * is initialized, which may not be the case when if_attach() is 442 * called. 443 */ 444 void 445 if_alloc_sadl(struct ifnet *ifp) 446 { 447 struct ifaddr *ifa; 448 const struct sockaddr_dl *sdl; 449 450 /* 451 * If the interface already has a link name, release it 452 * now. This is useful for interfaces that can change 453 * link types, and thus switch link names often. 454 */ 455 if (ifp->if_sadl != NULL) 456 if_free_sadl(ifp); 457 458 ifa = if_dl_create(ifp, &sdl); 459 460 ifa_insert(ifp, ifa); 461 if_sadl_setrefs(ifp, ifa); 462 } 463 464 static void 465 if_deactivate_sadl(struct ifnet *ifp) 466 { 467 struct ifaddr *ifa; 468 469 KASSERT(ifp->if_dl != NULL); 470 471 ifa = ifp->if_dl; 472 473 ifp->if_sadl = NULL; 474 475 ifp->if_dl = NULL; 476 ifafree(ifa); 477 } 478 479 void 480 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa, 481 const struct sockaddr_dl *sdl) 482 { 483 int s; 484 485 s = splnet(); 486 487 if_deactivate_sadl(ifp); 488 489 if_sadl_setrefs(ifp, ifa); 490 IFADDR_FOREACH(ifa, ifp) 491 rtinit(ifa, RTM_LLINFO_UPD, 0); 492 splx(s); 493 } 494 495 /* 496 * Free the link level name for the specified interface. This is 497 * a detach helper. This is called from if_detach(). 498 */ 499 static void 500 if_free_sadl(struct ifnet *ifp) 501 { 502 struct ifaddr *ifa; 503 int s; 504 505 ifa = ifp->if_dl; 506 if (ifa == NULL) { 507 KASSERT(ifp->if_sadl == NULL); 508 return; 509 } 510 511 KASSERT(ifp->if_sadl != NULL); 512 513 s = splnet(); 514 rtinit(ifa, RTM_DELETE, 0); 515 ifa_remove(ifp, ifa); 516 if_deactivate_sadl(ifp); 517 if (ifp->if_hwdl == ifa) { 518 ifafree(ifa); 519 ifp->if_hwdl = NULL; 520 } 521 splx(s); 522 } 523 524 static void 525 if_getindex(ifnet_t *ifp) 526 { 527 bool hitlimit = false; 528 529 ifp->if_index_gen = index_gen++; 530 531 ifp->if_index = if_index; 532 if (ifindex2ifnet == NULL) { 533 if_index++; 534 goto skip; 535 } 536 while (if_byindex(ifp->if_index)) { 537 /* 538 * If we hit USHRT_MAX, we skip back to 0 since 539 * there are a number of places where the value 540 * of if_index or if_index itself is compared 541 * to or stored in an unsigned short. By 542 * jumping back, we won't botch those assignments 543 * or comparisons. 544 */ 545 if (++if_index == 0) { 546 if_index = 1; 547 } else if (if_index == USHRT_MAX) { 548 /* 549 * However, if we have to jump back to 550 * zero *twice* without finding an empty 551 * slot in ifindex2ifnet[], then there 552 * there are too many (>65535) interfaces. 553 */ 554 if (hitlimit) { 555 panic("too many interfaces"); 556 } 557 hitlimit = true; 558 if_index = 1; 559 } 560 ifp->if_index = if_index; 561 } 562 skip: 563 /* 564 * ifindex2ifnet is indexed by if_index. Since if_index will 565 * grow dynamically, it should grow too. 566 */ 567 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 568 size_t m, n, oldlim; 569 void *q; 570 571 oldlim = if_indexlim; 572 while (ifp->if_index >= if_indexlim) 573 if_indexlim <<= 1; 574 575 /* grow ifindex2ifnet */ 576 m = oldlim * sizeof(struct ifnet *); 577 n = if_indexlim * sizeof(struct ifnet *); 578 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 579 if (ifindex2ifnet != NULL) { 580 memcpy(q, ifindex2ifnet, m); 581 free(ifindex2ifnet, M_IFADDR); 582 } 583 ifindex2ifnet = (struct ifnet **)q; 584 } 585 ifindex2ifnet[ifp->if_index] = ifp; 586 } 587 588 /* 589 * Initialize an interface and assign an index for it. 590 * 591 * It must be called prior to a device specific attach routine 592 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 593 * and be followed by if_register: 594 * 595 * if_initialize(ifp); 596 * ether_ifattach(ifp, enaddr); 597 * if_register(ifp); 598 */ 599 void 600 if_initialize(ifnet_t *ifp) 601 { 602 KASSERT(if_indexlim > 0); 603 TAILQ_INIT(&ifp->if_addrlist); 604 605 /* 606 * Link level name is allocated later by a separate call to 607 * if_alloc_sadl(). 608 */ 609 610 if (ifp->if_snd.ifq_maxlen == 0) 611 ifp->if_snd.ifq_maxlen = ifqmaxlen; 612 613 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 614 615 ifp->if_link_state = LINK_STATE_UNKNOWN; 616 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 617 618 ifp->if_capenable = 0; 619 ifp->if_csum_flags_tx = 0; 620 ifp->if_csum_flags_rx = 0; 621 622 #ifdef ALTQ 623 ifp->if_snd.altq_type = 0; 624 ifp->if_snd.altq_disc = NULL; 625 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 626 ifp->if_snd.altq_tbr = NULL; 627 ifp->if_snd.altq_ifp = ifp; 628 #endif 629 630 #ifdef NET_MPSAFE 631 ifp->if_snd.ifq_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 632 #else 633 ifp->if_snd.ifq_lock = NULL; 634 #endif 635 636 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 637 (void)pfil_run_hooks(if_pfil, 638 (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET); 639 640 IF_AFDATA_LOCK_INIT(ifp); 641 642 ifp->if_link_si = softint_establish(SOFTINT_NET, if_link_state_change_si, ifp); 643 if (ifp->if_link_si == NULL) 644 panic("%s: softint_establish() failed", __func__); 645 646 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 647 psref_target_init(&ifp->if_psref, ifnet_psref_class); 648 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 649 650 IFNET_LOCK(); 651 if_getindex(ifp); 652 IFNET_UNLOCK(); 653 } 654 655 /* 656 * Register an interface to the list of "active" interfaces. 657 */ 658 void 659 if_register(ifnet_t *ifp) 660 { 661 /* 662 * If the driver has not supplied its own if_ioctl, then 663 * supply the default. 664 */ 665 if (ifp->if_ioctl == NULL) 666 ifp->if_ioctl = ifioctl_common; 667 668 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 669 670 if (!STAILQ_EMPTY(&domains)) 671 if_attachdomain1(ifp); 672 673 /* Announce the interface. */ 674 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 675 676 if (ifp->if_slowtimo != NULL) { 677 ifp->if_slowtimo_ch = 678 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 679 callout_init(ifp->if_slowtimo_ch, 0); 680 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 681 if_slowtimo(ifp); 682 } 683 684 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 685 ifp->if_transmit = if_transmit; 686 687 IFNET_LOCK(); 688 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 689 IFNET_WRITER_INSERT_TAIL(ifp); 690 IFNET_UNLOCK(); 691 } 692 693 /* 694 * The if_percpuq framework 695 * 696 * It allows network device drivers to execute the network stack 697 * in softint (so called softint-based if_input). It utilizes 698 * softint and percpu ifqueue. It doesn't distribute any packets 699 * between CPUs, unlike pktqueue(9). 700 * 701 * Currently we support two options for device drivers to apply the framework: 702 * - Use it implicitly with less changes 703 * - If you use if_attach in driver's _attach function and if_input in 704 * driver's Rx interrupt handler, a packet is queued and a softint handles 705 * the packet implicitly 706 * - Use it explicitly in each driver (recommended) 707 * - You can use if_percpuq_* directly in your driver 708 * - In this case, you need to allocate struct if_percpuq in driver's softc 709 * - See wm(4) as a reference implementation 710 */ 711 712 static void 713 if_percpuq_softint(void *arg) 714 { 715 struct if_percpuq *ipq = arg; 716 struct ifnet *ifp = ipq->ipq_ifp; 717 struct mbuf *m; 718 719 while ((m = if_percpuq_dequeue(ipq)) != NULL) 720 ifp->_if_input(ifp, m); 721 } 722 723 static void 724 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 725 { 726 struct ifqueue *const ifq = p; 727 728 memset(ifq, 0, sizeof(*ifq)); 729 ifq->ifq_maxlen = IFQ_MAXLEN; 730 } 731 732 struct if_percpuq * 733 if_percpuq_create(struct ifnet *ifp) 734 { 735 struct if_percpuq *ipq; 736 737 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 738 if (ipq == NULL) 739 panic("kmem_zalloc failed"); 740 741 ipq->ipq_ifp = ifp; 742 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 743 if_percpuq_softint, ipq); 744 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 745 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 746 747 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 748 749 return ipq; 750 } 751 752 static struct mbuf * 753 if_percpuq_dequeue(struct if_percpuq *ipq) 754 { 755 struct mbuf *m; 756 struct ifqueue *ifq; 757 int s; 758 759 s = splnet(); 760 ifq = percpu_getref(ipq->ipq_ifqs); 761 IF_DEQUEUE(ifq, m); 762 percpu_putref(ipq->ipq_ifqs); 763 splx(s); 764 765 return m; 766 } 767 768 static void 769 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 770 { 771 struct ifqueue *const ifq = p; 772 773 IF_PURGE(ifq); 774 } 775 776 void 777 if_percpuq_destroy(struct if_percpuq *ipq) 778 { 779 780 /* if_detach may already destroy it */ 781 if (ipq == NULL) 782 return; 783 784 softint_disestablish(ipq->ipq_si); 785 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 786 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 787 } 788 789 void 790 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 791 { 792 struct ifqueue *ifq; 793 int s; 794 795 KASSERT(ipq != NULL); 796 797 s = splnet(); 798 ifq = percpu_getref(ipq->ipq_ifqs); 799 if (IF_QFULL(ifq)) { 800 IF_DROP(ifq); 801 percpu_putref(ipq->ipq_ifqs); 802 m_freem(m); 803 goto out; 804 } 805 IF_ENQUEUE(ifq, m); 806 percpu_putref(ipq->ipq_ifqs); 807 808 softint_schedule(ipq->ipq_si); 809 out: 810 splx(s); 811 } 812 813 static void 814 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 815 { 816 struct ifqueue *const ifq = p; 817 int *sum = arg; 818 819 *sum += ifq->ifq_drops; 820 } 821 822 static int 823 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 824 { 825 struct sysctlnode node; 826 struct if_percpuq *ipq; 827 int sum = 0; 828 int error; 829 830 node = *rnode; 831 ipq = node.sysctl_data; 832 833 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 834 835 node.sysctl_data = ∑ 836 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 837 if (error != 0 || newp == NULL) 838 return error; 839 840 return 0; 841 } 842 843 static void 844 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 845 struct if_percpuq *ipq) 846 { 847 const struct sysctlnode *cnode, *rnode; 848 849 if (sysctl_createv(clog, 0, NULL, &rnode, 850 CTLFLAG_PERMANENT, 851 CTLTYPE_NODE, "interfaces", 852 SYSCTL_DESCR("Per-interface controls"), 853 NULL, 0, NULL, 0, 854 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 855 goto bad; 856 857 if (sysctl_createv(clog, 0, &rnode, &rnode, 858 CTLFLAG_PERMANENT, 859 CTLTYPE_NODE, ifname, 860 SYSCTL_DESCR("Interface controls"), 861 NULL, 0, NULL, 0, 862 CTL_CREATE, CTL_EOL) != 0) 863 goto bad; 864 865 if (sysctl_createv(clog, 0, &rnode, &rnode, 866 CTLFLAG_PERMANENT, 867 CTLTYPE_NODE, "rcvq", 868 SYSCTL_DESCR("Interface input queue controls"), 869 NULL, 0, NULL, 0, 870 CTL_CREATE, CTL_EOL) != 0) 871 goto bad; 872 873 #ifdef NOTYET 874 /* XXX Should show each per-CPU queue length? */ 875 if (sysctl_createv(clog, 0, &rnode, &rnode, 876 CTLFLAG_PERMANENT, 877 CTLTYPE_INT, "len", 878 SYSCTL_DESCR("Current input queue length"), 879 sysctl_percpuq_len, 0, NULL, 0, 880 CTL_CREATE, CTL_EOL) != 0) 881 goto bad; 882 883 if (sysctl_createv(clog, 0, &rnode, &cnode, 884 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 885 CTLTYPE_INT, "maxlen", 886 SYSCTL_DESCR("Maximum allowed input queue length"), 887 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 888 CTL_CREATE, CTL_EOL) != 0) 889 goto bad; 890 #endif 891 892 if (sysctl_createv(clog, 0, &rnode, &cnode, 893 CTLFLAG_PERMANENT, 894 CTLTYPE_INT, "drops", 895 SYSCTL_DESCR("Total packets dropped due to full input queue"), 896 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 897 CTL_CREATE, CTL_EOL) != 0) 898 goto bad; 899 900 return; 901 bad: 902 printf("%s: could not attach sysctl nodes\n", ifname); 903 return; 904 } 905 906 907 /* 908 * The common interface input routine that is called by device drivers, 909 * which should be used only when the driver's rx handler already runs 910 * in softint. 911 */ 912 void 913 if_input(struct ifnet *ifp, struct mbuf *m) 914 { 915 916 KASSERT(ifp->if_percpuq == NULL); 917 KASSERT(!cpu_intr_p()); 918 919 ifp->_if_input(ifp, m); 920 } 921 922 /* 923 * DEPRECATED. Use if_initialize and if_register instead. 924 * See the above comment of if_initialize. 925 * 926 * Note that it implicitly enables if_percpuq to make drivers easy to 927 * migrate softint-based if_input without much changes. If you don't 928 * want to enable it, use if_initialize instead. 929 */ 930 void 931 if_attach(ifnet_t *ifp) 932 { 933 934 if_initialize(ifp); 935 ifp->if_percpuq = if_percpuq_create(ifp); 936 if_register(ifp); 937 } 938 939 void 940 if_attachdomain(void) 941 { 942 struct ifnet *ifp; 943 int s; 944 int bound = curlwp_bind(); 945 946 s = pserialize_read_enter(); 947 IFNET_READER_FOREACH(ifp) { 948 struct psref psref; 949 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 950 pserialize_read_exit(s); 951 if_attachdomain1(ifp); 952 s = pserialize_read_enter(); 953 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 954 } 955 pserialize_read_exit(s); 956 curlwp_bindx(bound); 957 } 958 959 static void 960 if_attachdomain1(struct ifnet *ifp) 961 { 962 struct domain *dp; 963 int s; 964 965 s = splnet(); 966 967 /* address family dependent data region */ 968 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 969 DOMAIN_FOREACH(dp) { 970 if (dp->dom_ifattach != NULL) 971 ifp->if_afdata[dp->dom_family] = 972 (*dp->dom_ifattach)(ifp); 973 } 974 975 splx(s); 976 } 977 978 /* 979 * Deactivate an interface. This points all of the procedure 980 * handles at error stubs. May be called from interrupt context. 981 */ 982 void 983 if_deactivate(struct ifnet *ifp) 984 { 985 int s; 986 987 s = splnet(); 988 989 ifp->if_output = if_nulloutput; 990 ifp->_if_input = if_nullinput; 991 ifp->if_start = if_nullstart; 992 ifp->if_transmit = if_nulltransmit; 993 ifp->if_ioctl = if_nullioctl; 994 ifp->if_init = if_nullinit; 995 ifp->if_stop = if_nullstop; 996 ifp->if_slowtimo = if_nullslowtimo; 997 ifp->if_drain = if_nulldrain; 998 999 /* No more packets may be enqueued. */ 1000 ifp->if_snd.ifq_maxlen = 0; 1001 1002 splx(s); 1003 } 1004 1005 void 1006 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1007 { 1008 struct ifaddr *ifa, *nifa; 1009 1010 IFADDR_FOREACH_SAFE(ifa, ifp, nifa) { 1011 if (ifa->ifa_addr->sa_family != family) 1012 continue; 1013 (*purgeaddr)(ifa); 1014 } 1015 } 1016 1017 /* 1018 * Detach an interface from the list of "active" interfaces, 1019 * freeing any resources as we go along. 1020 * 1021 * NOTE: This routine must be called with a valid thread context, 1022 * as it may block. 1023 */ 1024 void 1025 if_detach(struct ifnet *ifp) 1026 { 1027 struct socket so; 1028 struct ifaddr *ifa; 1029 #ifdef IFAREF_DEBUG 1030 struct ifaddr *last_ifa = NULL; 1031 #endif 1032 struct domain *dp; 1033 const struct protosw *pr; 1034 int s, i, family, purged; 1035 uint64_t xc; 1036 1037 /* 1038 * XXX It's kind of lame that we have to have the 1039 * XXX socket structure... 1040 */ 1041 memset(&so, 0, sizeof(so)); 1042 1043 s = splnet(); 1044 1045 sysctl_teardown(&ifp->if_sysctl_log); 1046 mutex_enter(ifp->if_ioctl_lock); 1047 ifp->if_ioctl = if_nullioctl; 1048 mutex_exit(ifp->if_ioctl_lock); 1049 1050 IFNET_LOCK(); 1051 ifindex2ifnet[ifp->if_index] = NULL; 1052 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1053 IFNET_WRITER_REMOVE(ifp); 1054 pserialize_perform(ifnet_psz); 1055 IFNET_UNLOCK(); 1056 1057 mutex_obj_free(ifp->if_ioctl_lock); 1058 ifp->if_ioctl_lock = NULL; 1059 1060 if (ifp->if_slowtimo != NULL) { 1061 ifp->if_slowtimo = NULL; 1062 callout_halt(ifp->if_slowtimo_ch, NULL); 1063 callout_destroy(ifp->if_slowtimo_ch); 1064 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1065 } 1066 1067 /* 1068 * Do an if_down() to give protocols a chance to do something. 1069 */ 1070 if_down(ifp); 1071 1072 #ifdef ALTQ 1073 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1074 altq_disable(&ifp->if_snd); 1075 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1076 altq_detach(&ifp->if_snd); 1077 #endif 1078 1079 if (ifp->if_snd.ifq_lock) 1080 mutex_obj_free(ifp->if_snd.ifq_lock); 1081 1082 #if NCARP > 0 1083 /* Remove the interface from any carp group it is a part of. */ 1084 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1085 carp_ifdetach(ifp); 1086 #endif 1087 1088 /* 1089 * Rip all the addresses off the interface. This should make 1090 * all of the routes go away. 1091 * 1092 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1093 * from the list, including our "cursor", ifa. For safety, 1094 * and to honor the TAILQ abstraction, I just restart the 1095 * loop after each removal. Note that the loop will exit 1096 * when all of the remaining ifaddrs belong to the AF_LINK 1097 * family. I am counting on the historical fact that at 1098 * least one pr_usrreq in each address domain removes at 1099 * least one ifaddr. 1100 */ 1101 again: 1102 IFADDR_FOREACH(ifa, ifp) { 1103 family = ifa->ifa_addr->sa_family; 1104 #ifdef IFAREF_DEBUG 1105 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1106 ifa, family, ifa->ifa_refcnt); 1107 if (last_ifa != NULL && ifa == last_ifa) 1108 panic("if_detach: loop detected"); 1109 last_ifa = ifa; 1110 #endif 1111 if (family == AF_LINK) 1112 continue; 1113 dp = pffinddomain(family); 1114 #ifdef DIAGNOSTIC 1115 if (dp == NULL) 1116 panic("if_detach: no domain for AF %d", 1117 family); 1118 #endif 1119 /* 1120 * XXX These PURGEIF calls are redundant with the 1121 * purge-all-families calls below, but are left in for 1122 * now both to make a smaller change, and to avoid 1123 * unplanned interactions with clearing of 1124 * ifp->if_addrlist. 1125 */ 1126 purged = 0; 1127 for (pr = dp->dom_protosw; 1128 pr < dp->dom_protoswNPROTOSW; pr++) { 1129 so.so_proto = pr; 1130 if (pr->pr_usrreqs) { 1131 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1132 purged = 1; 1133 } 1134 } 1135 if (purged == 0) { 1136 /* 1137 * XXX What's really the best thing to do 1138 * XXX here? --thorpej (at) NetBSD.org 1139 */ 1140 printf("if_detach: WARNING: AF %d not purged\n", 1141 family); 1142 ifa_remove(ifp, ifa); 1143 } 1144 goto again; 1145 } 1146 1147 if_free_sadl(ifp); 1148 1149 /* Walk the routing table looking for stragglers. */ 1150 for (i = 0; i <= AF_MAX; i++) { 1151 while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART) 1152 continue; 1153 } 1154 1155 DOMAIN_FOREACH(dp) { 1156 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1157 { 1158 void *p = ifp->if_afdata[dp->dom_family]; 1159 if (p) { 1160 ifp->if_afdata[dp->dom_family] = NULL; 1161 (*dp->dom_ifdetach)(ifp, p); 1162 } 1163 } 1164 1165 /* 1166 * One would expect multicast memberships (INET and 1167 * INET6) on UDP sockets to be purged by the PURGEIF 1168 * calls above, but if all addresses were removed from 1169 * the interface prior to destruction, the calls will 1170 * not be made (e.g. ppp, for which pppd(8) generally 1171 * removes addresses before destroying the interface). 1172 * Because there is no invariant that multicast 1173 * memberships only exist for interfaces with IPv4 1174 * addresses, we must call PURGEIF regardless of 1175 * addresses. (Protocols which might store ifnet 1176 * pointers are marked with PR_PURGEIF.) 1177 */ 1178 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1179 so.so_proto = pr; 1180 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1181 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1182 } 1183 } 1184 1185 (void)pfil_run_hooks(if_pfil, 1186 (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET); 1187 (void)pfil_head_destroy(ifp->if_pfil); 1188 1189 /* Announce that the interface is gone. */ 1190 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1191 1192 IF_AFDATA_LOCK_DESTROY(ifp); 1193 1194 softint_disestablish(ifp->if_link_si); 1195 ifp->if_link_si = NULL; 1196 1197 /* 1198 * remove packets that came from ifp, from software interrupt queues. 1199 */ 1200 DOMAIN_FOREACH(dp) { 1201 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1202 struct ifqueue *iq = dp->dom_ifqueues[i]; 1203 if (iq == NULL) 1204 break; 1205 dp->dom_ifqueues[i] = NULL; 1206 if_detach_queues(ifp, iq); 1207 } 1208 } 1209 1210 /* 1211 * IP queues have to be processed separately: net-queue barrier 1212 * ensures that the packets are dequeued while a cross-call will 1213 * ensure that the interrupts have completed. FIXME: not quite.. 1214 */ 1215 #ifdef INET 1216 pktq_barrier(ip_pktq); 1217 #endif 1218 #ifdef INET6 1219 if (in6_present) 1220 pktq_barrier(ip6_pktq); 1221 #endif 1222 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1223 xc_wait(xc); 1224 1225 /* Wait for all readers to drain before freeing. */ 1226 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1227 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1228 1229 if (ifp->if_percpuq != NULL) { 1230 if_percpuq_destroy(ifp->if_percpuq); 1231 ifp->if_percpuq = NULL; 1232 } 1233 1234 splx(s); 1235 } 1236 1237 static void 1238 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1239 { 1240 struct mbuf *m, *prev, *next; 1241 1242 prev = NULL; 1243 for (m = q->ifq_head; m != NULL; m = next) { 1244 KASSERT((m->m_flags & M_PKTHDR) != 0); 1245 1246 next = m->m_nextpkt; 1247 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1248 prev = m; 1249 continue; 1250 } 1251 1252 if (prev != NULL) 1253 prev->m_nextpkt = m->m_nextpkt; 1254 else 1255 q->ifq_head = m->m_nextpkt; 1256 if (q->ifq_tail == m) 1257 q->ifq_tail = prev; 1258 q->ifq_len--; 1259 1260 m->m_nextpkt = NULL; 1261 m_freem(m); 1262 IF_DROP(q); 1263 } 1264 } 1265 1266 /* 1267 * Callback for a radix tree walk to delete all references to an 1268 * ifnet. 1269 */ 1270 static int 1271 if_rt_walktree(struct rtentry *rt, void *v) 1272 { 1273 struct ifnet *ifp = (struct ifnet *)v; 1274 int error; 1275 struct rtentry *retrt; 1276 1277 if (rt->rt_ifp != ifp) 1278 return 0; 1279 1280 /* Delete the entry. */ 1281 error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway, 1282 rt_mask(rt), rt->rt_flags, &retrt); 1283 if (error == 0) { 1284 KASSERT(retrt == rt); 1285 KASSERT((retrt->rt_flags & RTF_UP) == 0); 1286 retrt->rt_ifp = NULL; 1287 rtfree(retrt); 1288 } else { 1289 printf("%s: warning: unable to delete rtentry @ %p, " 1290 "error = %d\n", ifp->if_xname, rt, error); 1291 } 1292 return ERESTART; 1293 } 1294 1295 /* 1296 * Create a clone network interface. 1297 */ 1298 static int 1299 if_clone_create(const char *name) 1300 { 1301 struct if_clone *ifc; 1302 int unit; 1303 struct ifnet *ifp; 1304 struct psref psref; 1305 1306 ifc = if_clone_lookup(name, &unit); 1307 if (ifc == NULL) 1308 return EINVAL; 1309 1310 ifp = if_get(name, &psref); 1311 if (ifp != NULL) { 1312 if_put(ifp, &psref); 1313 return EEXIST; 1314 } 1315 1316 return (*ifc->ifc_create)(ifc, unit); 1317 } 1318 1319 /* 1320 * Destroy a clone network interface. 1321 */ 1322 static int 1323 if_clone_destroy(const char *name) 1324 { 1325 struct if_clone *ifc; 1326 struct ifnet *ifp; 1327 struct psref psref; 1328 1329 ifc = if_clone_lookup(name, NULL); 1330 if (ifc == NULL) 1331 return EINVAL; 1332 1333 if (ifc->ifc_destroy == NULL) 1334 return EOPNOTSUPP; 1335 1336 ifp = if_get(name, &psref); 1337 if (ifp == NULL) 1338 return ENXIO; 1339 1340 /* We have to disable ioctls here */ 1341 mutex_enter(ifp->if_ioctl_lock); 1342 ifp->if_ioctl = if_nullioctl; 1343 mutex_exit(ifp->if_ioctl_lock); 1344 1345 /* 1346 * We cannot call ifc_destroy with holding ifp. 1347 * Releasing ifp here is safe thanks to if_clone_mtx. 1348 */ 1349 if_put(ifp, &psref); 1350 1351 return (*ifc->ifc_destroy)(ifp); 1352 } 1353 1354 /* 1355 * Look up a network interface cloner. 1356 */ 1357 static struct if_clone * 1358 if_clone_lookup(const char *name, int *unitp) 1359 { 1360 struct if_clone *ifc; 1361 const char *cp; 1362 char *dp, ifname[IFNAMSIZ + 3]; 1363 int unit; 1364 1365 strcpy(ifname, "if_"); 1366 /* separate interface name from unit */ 1367 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1368 *cp && (*cp < '0' || *cp > '9');) 1369 *dp++ = *cp++; 1370 1371 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1372 return NULL; /* No name or unit number */ 1373 *dp++ = '\0'; 1374 1375 again: 1376 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1377 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1378 break; 1379 } 1380 1381 if (ifc == NULL) { 1382 if (*ifname == '\0' || 1383 module_autoload(ifname, MODULE_CLASS_DRIVER)) 1384 return NULL; 1385 *ifname = '\0'; 1386 goto again; 1387 } 1388 1389 unit = 0; 1390 while (cp - name < IFNAMSIZ && *cp) { 1391 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1392 /* Bogus unit number. */ 1393 return NULL; 1394 } 1395 unit = (unit * 10) + (*cp++ - '0'); 1396 } 1397 1398 if (unitp != NULL) 1399 *unitp = unit; 1400 return ifc; 1401 } 1402 1403 /* 1404 * Register a network interface cloner. 1405 */ 1406 void 1407 if_clone_attach(struct if_clone *ifc) 1408 { 1409 1410 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1411 if_cloners_count++; 1412 } 1413 1414 /* 1415 * Unregister a network interface cloner. 1416 */ 1417 void 1418 if_clone_detach(struct if_clone *ifc) 1419 { 1420 1421 LIST_REMOVE(ifc, ifc_list); 1422 if_cloners_count--; 1423 } 1424 1425 /* 1426 * Provide list of interface cloners to userspace. 1427 */ 1428 int 1429 if_clone_list(int buf_count, char *buffer, int *total) 1430 { 1431 char outbuf[IFNAMSIZ], *dst; 1432 struct if_clone *ifc; 1433 int count, error = 0; 1434 1435 *total = if_cloners_count; 1436 if ((dst = buffer) == NULL) { 1437 /* Just asking how many there are. */ 1438 return 0; 1439 } 1440 1441 if (buf_count < 0) 1442 return EINVAL; 1443 1444 count = (if_cloners_count < buf_count) ? 1445 if_cloners_count : buf_count; 1446 1447 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1448 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1449 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1450 if (outbuf[sizeof(outbuf) - 1] != '\0') 1451 return ENAMETOOLONG; 1452 error = copyout(outbuf, dst, sizeof(outbuf)); 1453 if (error != 0) 1454 break; 1455 } 1456 1457 return error; 1458 } 1459 1460 void 1461 ifaref(struct ifaddr *ifa) 1462 { 1463 ifa->ifa_refcnt++; 1464 } 1465 1466 void 1467 ifafree(struct ifaddr *ifa) 1468 { 1469 KASSERT(ifa != NULL); 1470 KASSERT(ifa->ifa_refcnt > 0); 1471 1472 if (--ifa->ifa_refcnt == 0) { 1473 free(ifa, M_IFADDR); 1474 } 1475 } 1476 1477 void 1478 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1479 { 1480 ifa->ifa_ifp = ifp; 1481 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1482 ifaref(ifa); 1483 } 1484 1485 void 1486 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1487 { 1488 KASSERT(ifa->ifa_ifp == ifp); 1489 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1490 ifafree(ifa); 1491 } 1492 1493 static inline int 1494 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1495 { 1496 return sockaddr_cmp(sa1, sa2) == 0; 1497 } 1498 1499 /* 1500 * Locate an interface based on a complete address. 1501 */ 1502 /*ARGSUSED*/ 1503 struct ifaddr * 1504 ifa_ifwithaddr(const struct sockaddr *addr) 1505 { 1506 struct ifnet *ifp; 1507 struct ifaddr *ifa; 1508 int s; 1509 1510 s = pserialize_read_enter(); 1511 IFNET_READER_FOREACH(ifp) { 1512 if (ifp->if_output == if_nulloutput) 1513 continue; 1514 IFADDR_FOREACH(ifa, ifp) { 1515 if (ifa->ifa_addr->sa_family != addr->sa_family) 1516 continue; 1517 if (equal(addr, ifa->ifa_addr)) 1518 return ifa; 1519 if ((ifp->if_flags & IFF_BROADCAST) && 1520 ifa->ifa_broadaddr && 1521 /* IP6 doesn't have broadcast */ 1522 ifa->ifa_broadaddr->sa_len != 0 && 1523 equal(ifa->ifa_broadaddr, addr)) 1524 return ifa; 1525 } 1526 } 1527 pserialize_read_exit(s); 1528 return NULL; 1529 } 1530 1531 /* 1532 * Locate the point to point interface with a given destination address. 1533 */ 1534 /*ARGSUSED*/ 1535 struct ifaddr * 1536 ifa_ifwithdstaddr(const struct sockaddr *addr) 1537 { 1538 struct ifnet *ifp; 1539 struct ifaddr *ifa; 1540 int s; 1541 1542 s = pserialize_read_enter(); 1543 IFNET_READER_FOREACH(ifp) { 1544 if (ifp->if_output == if_nulloutput) 1545 continue; 1546 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1547 continue; 1548 IFADDR_FOREACH(ifa, ifp) { 1549 if (ifa->ifa_addr->sa_family != addr->sa_family || 1550 ifa->ifa_dstaddr == NULL) 1551 continue; 1552 if (equal(addr, ifa->ifa_dstaddr)) 1553 return ifa; 1554 } 1555 } 1556 pserialize_read_exit(s); 1557 return NULL; 1558 } 1559 1560 /* 1561 * Find an interface on a specific network. If many, choice 1562 * is most specific found. 1563 */ 1564 struct ifaddr * 1565 ifa_ifwithnet(const struct sockaddr *addr) 1566 { 1567 struct ifnet *ifp; 1568 struct ifaddr *ifa; 1569 const struct sockaddr_dl *sdl; 1570 struct ifaddr *ifa_maybe = 0; 1571 u_int af = addr->sa_family; 1572 const char *addr_data = addr->sa_data, *cplim; 1573 int s; 1574 1575 if (af == AF_LINK) { 1576 sdl = satocsdl(addr); 1577 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1578 ifindex2ifnet[sdl->sdl_index] && 1579 ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput) { 1580 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1581 } 1582 } 1583 #ifdef NETATALK 1584 if (af == AF_APPLETALK) { 1585 const struct sockaddr_at *sat, *sat2; 1586 sat = (const struct sockaddr_at *)addr; 1587 s = pserialize_read_enter(); 1588 IFNET_READER_FOREACH(ifp) { 1589 if (ifp->if_output == if_nulloutput) 1590 continue; 1591 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1592 if (ifa == NULL) 1593 continue; 1594 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1595 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1596 return ifa; /* exact match */ 1597 if (ifa_maybe == NULL) { 1598 /* else keep the if with the right range */ 1599 ifa_maybe = ifa; 1600 } 1601 } 1602 pserialize_read_exit(s); 1603 return ifa_maybe; 1604 } 1605 #endif 1606 s = pserialize_read_enter(); 1607 IFNET_READER_FOREACH(ifp) { 1608 if (ifp->if_output == if_nulloutput) 1609 continue; 1610 IFADDR_FOREACH(ifa, ifp) { 1611 const char *cp, *cp2, *cp3; 1612 1613 if (ifa->ifa_addr->sa_family != af || 1614 ifa->ifa_netmask == NULL) 1615 next: continue; 1616 cp = addr_data; 1617 cp2 = ifa->ifa_addr->sa_data; 1618 cp3 = ifa->ifa_netmask->sa_data; 1619 cplim = (const char *)ifa->ifa_netmask + 1620 ifa->ifa_netmask->sa_len; 1621 while (cp3 < cplim) { 1622 if ((*cp++ ^ *cp2++) & *cp3++) { 1623 /* want to continue for() loop */ 1624 goto next; 1625 } 1626 } 1627 if (ifa_maybe == NULL || 1628 rt_refines(ifa->ifa_netmask, 1629 ifa_maybe->ifa_netmask)) 1630 ifa_maybe = ifa; 1631 } 1632 } 1633 pserialize_read_exit(s); 1634 return ifa_maybe; 1635 } 1636 1637 /* 1638 * Find the interface of the addresss. 1639 */ 1640 struct ifaddr * 1641 ifa_ifwithladdr(const struct sockaddr *addr) 1642 { 1643 struct ifaddr *ia; 1644 1645 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1646 (ia = ifa_ifwithnet(addr))) 1647 return ia; 1648 return NULL; 1649 } 1650 1651 /* 1652 * Find an interface using a specific address family 1653 */ 1654 struct ifaddr * 1655 ifa_ifwithaf(int af) 1656 { 1657 struct ifnet *ifp; 1658 struct ifaddr *ifa = NULL; 1659 int s; 1660 1661 s = pserialize_read_enter(); 1662 IFNET_READER_FOREACH(ifp) { 1663 if (ifp->if_output == if_nulloutput) 1664 continue; 1665 IFADDR_FOREACH(ifa, ifp) { 1666 if (ifa->ifa_addr->sa_family == af) 1667 goto out; 1668 } 1669 } 1670 out: 1671 pserialize_read_exit(s); 1672 return ifa; 1673 } 1674 1675 /* 1676 * Find an interface address specific to an interface best matching 1677 * a given address. 1678 */ 1679 struct ifaddr * 1680 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1681 { 1682 struct ifaddr *ifa; 1683 const char *cp, *cp2, *cp3; 1684 const char *cplim; 1685 struct ifaddr *ifa_maybe = 0; 1686 u_int af = addr->sa_family; 1687 1688 if (ifp->if_output == if_nulloutput) 1689 return NULL; 1690 1691 if (af >= AF_MAX) 1692 return NULL; 1693 1694 IFADDR_FOREACH(ifa, ifp) { 1695 if (ifa->ifa_addr->sa_family != af) 1696 continue; 1697 ifa_maybe = ifa; 1698 if (ifa->ifa_netmask == NULL) { 1699 if (equal(addr, ifa->ifa_addr) || 1700 (ifa->ifa_dstaddr && 1701 equal(addr, ifa->ifa_dstaddr))) 1702 return ifa; 1703 continue; 1704 } 1705 cp = addr->sa_data; 1706 cp2 = ifa->ifa_addr->sa_data; 1707 cp3 = ifa->ifa_netmask->sa_data; 1708 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1709 for (; cp3 < cplim; cp3++) { 1710 if ((*cp++ ^ *cp2++) & *cp3) 1711 break; 1712 } 1713 if (cp3 == cplim) 1714 return ifa; 1715 } 1716 return ifa_maybe; 1717 } 1718 1719 /* 1720 * Default action when installing a route with a Link Level gateway. 1721 * Lookup an appropriate real ifa to point to. 1722 * This should be moved to /sys/net/link.c eventually. 1723 */ 1724 void 1725 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 1726 { 1727 struct ifaddr *ifa; 1728 const struct sockaddr *dst; 1729 struct ifnet *ifp; 1730 1731 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 1732 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 1733 return; 1734 if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) { 1735 rt_replace_ifa(rt, ifa); 1736 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 1737 ifa->ifa_rtrequest(cmd, rt, info); 1738 } 1739 } 1740 1741 /* 1742 * bitmask macros to manage a densely packed link_state change queue. 1743 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 1744 * LINK_STATE_UP(2) we need 2 bits for each state change. 1745 * As a state change to store is 0, treat all bits set as an unset item. 1746 */ 1747 #define LQ_ITEM_BITS 2 1748 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 1749 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 1750 #define LINK_STATE_UNSET LQ_ITEM_MASK 1751 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 1752 #define LQ_STORE(q, i, v) \ 1753 do { \ 1754 (q) &= ~LQ_MASK((i)); \ 1755 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 1756 } while (0 /* CONSTCOND */) 1757 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 1758 #define LQ_POP(q, v) \ 1759 do { \ 1760 (v) = LQ_ITEM((q), 0); \ 1761 (q) >>= LQ_ITEM_BITS; \ 1762 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1763 } while (0 /* CONSTCOND */) 1764 #define LQ_PUSH(q, v) \ 1765 do { \ 1766 (q) >>= LQ_ITEM_BITS; \ 1767 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 1768 } while (0 /* CONSTCOND */) 1769 #define LQ_FIND_UNSET(q, i) \ 1770 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 1771 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 1772 break; \ 1773 } 1774 /* 1775 * Handle a change in the interface link state and 1776 * queue notifications. 1777 */ 1778 void 1779 if_link_state_change(struct ifnet *ifp, int link_state) 1780 { 1781 int s, idx; 1782 1783 /* Ensure change is to a valid state */ 1784 switch (link_state) { 1785 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 1786 case LINK_STATE_DOWN: /* FALLTHROUGH */ 1787 case LINK_STATE_UP: 1788 break; 1789 default: 1790 #ifdef DEBUG 1791 printf("%s: invalid link state %d\n", 1792 ifp->if_xname, link_state); 1793 #endif 1794 return; 1795 } 1796 1797 s = splnet(); 1798 1799 /* Find the last unset event in the queue. */ 1800 LQ_FIND_UNSET(ifp->if_link_queue, idx); 1801 1802 /* 1803 * Ensure link_state doesn't match the last event in the queue. 1804 * ifp->if_link_state is not checked and set here because 1805 * that would present an inconsistent picture to the system. 1806 */ 1807 if (idx != 0 && 1808 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 1809 goto out; 1810 1811 /* Handle queue overflow. */ 1812 if (idx == LQ_MAX(ifp->if_link_queue)) { 1813 uint8_t lost; 1814 1815 /* 1816 * The DOWN state must be protected from being pushed off 1817 * the queue to ensure that userland will always be 1818 * in a sane state. 1819 * Because DOWN is protected, there is no need to protect 1820 * UNKNOWN. 1821 * It should be invalid to change from any other state to 1822 * UNKNOWN anyway ... 1823 */ 1824 lost = LQ_ITEM(ifp->if_link_queue, 0); 1825 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 1826 if (lost == LINK_STATE_DOWN) { 1827 lost = LQ_ITEM(ifp->if_link_queue, 0); 1828 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 1829 } 1830 printf("%s: lost link state change %s\n", 1831 ifp->if_xname, 1832 lost == LINK_STATE_UP ? "UP" : 1833 lost == LINK_STATE_DOWN ? "DOWN" : 1834 "UNKNOWN"); 1835 } else 1836 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 1837 1838 softint_schedule(ifp->if_link_si); 1839 1840 out: 1841 splx(s); 1842 } 1843 1844 /* 1845 * Handle interface link state change notifications. 1846 * Must be called at splnet(). 1847 */ 1848 static void 1849 if_link_state_change0(struct ifnet *ifp, int link_state) 1850 { 1851 struct domain *dp; 1852 1853 /* Ensure the change is still valid. */ 1854 if (ifp->if_link_state == link_state) 1855 return; 1856 1857 #ifdef DEBUG 1858 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 1859 link_state == LINK_STATE_UP ? "UP" : 1860 link_state == LINK_STATE_DOWN ? "DOWN" : 1861 "UNKNOWN", 1862 ifp->if_link_state == LINK_STATE_UP ? "UP" : 1863 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 1864 "UNKNOWN"); 1865 #endif 1866 1867 /* 1868 * When going from UNKNOWN to UP, we need to mark existing 1869 * addresses as tentative and restart DAD as we may have 1870 * erroneously not found a duplicate. 1871 * 1872 * This needs to happen before rt_ifmsg to avoid a race where 1873 * listeners would have an address and expect it to work right 1874 * away. 1875 */ 1876 if (link_state == LINK_STATE_UP && 1877 ifp->if_link_state == LINK_STATE_UNKNOWN) 1878 { 1879 DOMAIN_FOREACH(dp) { 1880 if (dp->dom_if_link_state_change != NULL) 1881 dp->dom_if_link_state_change(ifp, 1882 LINK_STATE_DOWN); 1883 } 1884 } 1885 1886 ifp->if_link_state = link_state; 1887 1888 /* Notify that the link state has changed. */ 1889 rt_ifmsg(ifp); 1890 1891 #if NCARP > 0 1892 if (ifp->if_carp) 1893 carp_carpdev_state(ifp); 1894 #endif 1895 1896 DOMAIN_FOREACH(dp) { 1897 if (dp->dom_if_link_state_change != NULL) 1898 dp->dom_if_link_state_change(ifp, link_state); 1899 } 1900 } 1901 1902 /* 1903 * Process the interface link state change queue. 1904 */ 1905 static void 1906 if_link_state_change_si(void *arg) 1907 { 1908 struct ifnet *ifp = arg; 1909 int s; 1910 uint8_t state; 1911 1912 s = splnet(); 1913 1914 /* Pop a link state change from the queue and process it. */ 1915 LQ_POP(ifp->if_link_queue, state); 1916 if_link_state_change0(ifp, state); 1917 1918 /* If there is a link state change to come, schedule it. */ 1919 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 1920 softint_schedule(ifp->if_link_si); 1921 1922 splx(s); 1923 } 1924 1925 /* 1926 * Default action when installing a local route on a point-to-point 1927 * interface. 1928 */ 1929 void 1930 p2p_rtrequest(int req, struct rtentry *rt, 1931 __unused const struct rt_addrinfo *info) 1932 { 1933 struct ifnet *ifp = rt->rt_ifp; 1934 struct ifaddr *ifa, *lo0ifa; 1935 1936 switch (req) { 1937 case RTM_ADD: 1938 if ((rt->rt_flags & RTF_LOCAL) == 0) 1939 break; 1940 1941 IFADDR_FOREACH(ifa, ifp) { 1942 if (equal(rt_getkey(rt), ifa->ifa_addr)) 1943 break; 1944 } 1945 if (ifa == NULL) 1946 break; 1947 1948 /* 1949 * Ensure lo0 has an address of the same family. 1950 */ 1951 IFADDR_FOREACH(lo0ifa, lo0ifp) { 1952 if (lo0ifa->ifa_addr->sa_family == 1953 ifa->ifa_addr->sa_family) 1954 break; 1955 } 1956 if (lo0ifa == NULL) 1957 break; 1958 1959 rt->rt_ifp = lo0ifp; 1960 1961 /* 1962 * Make sure to set rt->rt_ifa to the interface 1963 * address we are using, otherwise we will have trouble 1964 * with source address selection. 1965 */ 1966 if (ifa != rt->rt_ifa) 1967 rt_replace_ifa(rt, ifa); 1968 break; 1969 case RTM_DELETE: 1970 default: 1971 break; 1972 } 1973 } 1974 1975 /* 1976 * Mark an interface down and notify protocols of 1977 * the transition. 1978 * NOTE: must be called at splsoftnet or equivalent. 1979 */ 1980 void 1981 if_down(struct ifnet *ifp) 1982 { 1983 struct ifaddr *ifa; 1984 struct domain *dp; 1985 1986 ifp->if_flags &= ~IFF_UP; 1987 nanotime(&ifp->if_lastchange); 1988 IFADDR_FOREACH(ifa, ifp) 1989 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 1990 IFQ_PURGE(&ifp->if_snd); 1991 #if NCARP > 0 1992 if (ifp->if_carp) 1993 carp_carpdev_state(ifp); 1994 #endif 1995 rt_ifmsg(ifp); 1996 DOMAIN_FOREACH(dp) { 1997 if (dp->dom_if_down) 1998 dp->dom_if_down(ifp); 1999 } 2000 } 2001 2002 /* 2003 * Mark an interface up and notify protocols of 2004 * the transition. 2005 * NOTE: must be called at splsoftnet or equivalent. 2006 */ 2007 void 2008 if_up(struct ifnet *ifp) 2009 { 2010 #ifdef notyet 2011 struct ifaddr *ifa; 2012 #endif 2013 struct domain *dp; 2014 2015 ifp->if_flags |= IFF_UP; 2016 nanotime(&ifp->if_lastchange); 2017 #ifdef notyet 2018 /* this has no effect on IP, and will kill all ISO connections XXX */ 2019 IFADDR_FOREACH(ifa, ifp) 2020 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2021 #endif 2022 #if NCARP > 0 2023 if (ifp->if_carp) 2024 carp_carpdev_state(ifp); 2025 #endif 2026 rt_ifmsg(ifp); 2027 DOMAIN_FOREACH(dp) { 2028 if (dp->dom_if_up) 2029 dp->dom_if_up(ifp); 2030 } 2031 } 2032 2033 /* 2034 * Handle interface slowtimo timer routine. Called 2035 * from softclock, we decrement timer (if set) and 2036 * call the appropriate interface routine on expiration. 2037 */ 2038 static void 2039 if_slowtimo(void *arg) 2040 { 2041 void (*slowtimo)(struct ifnet *); 2042 struct ifnet *ifp = arg; 2043 int s; 2044 2045 slowtimo = ifp->if_slowtimo; 2046 if (__predict_false(slowtimo == NULL)) 2047 return; 2048 2049 s = splnet(); 2050 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2051 (*slowtimo)(ifp); 2052 2053 splx(s); 2054 2055 if (__predict_true(ifp->if_slowtimo != NULL)) 2056 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2057 } 2058 2059 /* 2060 * Set/clear promiscuous mode on interface ifp based on the truth value 2061 * of pswitch. The calls are reference counted so that only the first 2062 * "on" request actually has an effect, as does the final "off" request. 2063 * Results are undefined if the "off" and "on" requests are not matched. 2064 */ 2065 int 2066 ifpromisc(struct ifnet *ifp, int pswitch) 2067 { 2068 int pcount, ret; 2069 short nflags; 2070 2071 pcount = ifp->if_pcount; 2072 if (pswitch) { 2073 /* 2074 * Allow the device to be "placed" into promiscuous 2075 * mode even if it is not configured up. It will 2076 * consult IFF_PROMISC when it is brought up. 2077 */ 2078 if (ifp->if_pcount++ != 0) 2079 return 0; 2080 nflags = ifp->if_flags | IFF_PROMISC; 2081 } else { 2082 if (--ifp->if_pcount > 0) 2083 return 0; 2084 nflags = ifp->if_flags & ~IFF_PROMISC; 2085 } 2086 ret = if_flags_set(ifp, nflags); 2087 /* Restore interface state if not successful. */ 2088 if (ret != 0) { 2089 ifp->if_pcount = pcount; 2090 } 2091 return ret; 2092 } 2093 2094 /* 2095 * Map interface name to 2096 * interface structure pointer. 2097 */ 2098 struct ifnet * 2099 ifunit(const char *name) 2100 { 2101 struct ifnet *ifp; 2102 const char *cp = name; 2103 u_int unit = 0; 2104 u_int i; 2105 int s; 2106 2107 /* 2108 * If the entire name is a number, treat it as an ifindex. 2109 */ 2110 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2111 unit = unit * 10 + (*cp - '0'); 2112 } 2113 2114 /* 2115 * If the number took all of the name, then it's a valid ifindex. 2116 */ 2117 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2118 if (unit >= if_indexlim) 2119 return NULL; 2120 ifp = ifindex2ifnet[unit]; 2121 if (ifp == NULL || ifp->if_output == if_nulloutput) 2122 return NULL; 2123 return ifp; 2124 } 2125 2126 ifp = NULL; 2127 s = pserialize_read_enter(); 2128 IFNET_READER_FOREACH(ifp) { 2129 if (ifp->if_output == if_nulloutput) 2130 continue; 2131 if (strcmp(ifp->if_xname, name) == 0) 2132 goto out; 2133 } 2134 out: 2135 pserialize_read_exit(s); 2136 return ifp; 2137 } 2138 2139 /* 2140 * Get a reference of an ifnet object by an interface name. 2141 * The returned reference is protected by psref(9). The caller 2142 * must release a returned reference by if_put after use. 2143 */ 2144 struct ifnet * 2145 if_get(const char *name, struct psref *psref) 2146 { 2147 struct ifnet *ifp; 2148 const char *cp = name; 2149 u_int unit = 0; 2150 u_int i; 2151 int s; 2152 2153 /* 2154 * If the entire name is a number, treat it as an ifindex. 2155 */ 2156 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2157 unit = unit * 10 + (*cp - '0'); 2158 } 2159 2160 /* 2161 * If the number took all of the name, then it's a valid ifindex. 2162 */ 2163 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2164 if (unit >= if_indexlim) 2165 return NULL; 2166 ifp = ifindex2ifnet[unit]; 2167 if (ifp == NULL || ifp->if_output == if_nulloutput) 2168 return NULL; 2169 return ifp; 2170 } 2171 2172 ifp = NULL; 2173 s = pserialize_read_enter(); 2174 IFNET_READER_FOREACH(ifp) { 2175 if (ifp->if_output == if_nulloutput) 2176 continue; 2177 if (strcmp(ifp->if_xname, name) == 0) { 2178 psref_acquire(psref, &ifp->if_psref, 2179 ifnet_psref_class); 2180 goto out; 2181 } 2182 } 2183 out: 2184 pserialize_read_exit(s); 2185 return ifp; 2186 } 2187 2188 /* 2189 * Release a reference of an ifnet object given by if_get or 2190 * if_get_byindex. 2191 */ 2192 void 2193 if_put(const struct ifnet *ifp, struct psref *psref) 2194 { 2195 2196 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2197 } 2198 2199 ifnet_t * 2200 if_byindex(u_int idx) 2201 { 2202 return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2203 } 2204 2205 /* 2206 * Get a reference of an ifnet object by an interface index. 2207 * The returned reference is protected by psref(9). The caller 2208 * must release a returned reference by if_put after use. 2209 */ 2210 ifnet_t * 2211 if_get_byindex(u_int idx, struct psref *psref) 2212 { 2213 ifnet_t *ifp; 2214 int s; 2215 2216 s = pserialize_read_enter(); 2217 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2218 if (__predict_true(ifp != NULL)) 2219 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2220 pserialize_read_exit(s); 2221 2222 return ifp; 2223 } 2224 2225 /* 2226 * XXX unsafe 2227 */ 2228 void 2229 if_acquire_unsafe(struct ifnet *ifp, struct psref *psref) 2230 { 2231 2232 KASSERT(ifp->if_index != 0); 2233 KASSERT(if_byindex(ifp->if_index) != NULL); 2234 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2235 } 2236 2237 bool 2238 if_held(struct ifnet *ifp) 2239 { 2240 2241 return psref_held(&ifp->if_psref, ifnet_psref_class); 2242 } 2243 2244 2245 /* common */ 2246 int 2247 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2248 { 2249 int s; 2250 struct ifreq *ifr; 2251 struct ifcapreq *ifcr; 2252 struct ifdatareq *ifdr; 2253 2254 switch (cmd) { 2255 case SIOCSIFCAP: 2256 ifcr = data; 2257 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2258 return EINVAL; 2259 2260 if (ifcr->ifcr_capenable == ifp->if_capenable) 2261 return 0; 2262 2263 ifp->if_capenable = ifcr->ifcr_capenable; 2264 2265 /* Pre-compute the checksum flags mask. */ 2266 ifp->if_csum_flags_tx = 0; 2267 ifp->if_csum_flags_rx = 0; 2268 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2269 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2270 } 2271 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2272 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2273 } 2274 2275 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2276 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2277 } 2278 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2279 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2280 } 2281 2282 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2283 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2284 } 2285 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2286 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2287 } 2288 2289 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2290 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2291 } 2292 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2293 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2294 } 2295 2296 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2297 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2298 } 2299 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2300 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2301 } 2302 if (ifp->if_flags & IFF_UP) 2303 return ENETRESET; 2304 return 0; 2305 case SIOCSIFFLAGS: 2306 ifr = data; 2307 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2308 s = splnet(); 2309 if_down(ifp); 2310 splx(s); 2311 } 2312 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2313 s = splnet(); 2314 if_up(ifp); 2315 splx(s); 2316 } 2317 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2318 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2319 break; 2320 case SIOCGIFFLAGS: 2321 ifr = data; 2322 ifr->ifr_flags = ifp->if_flags; 2323 break; 2324 2325 case SIOCGIFMETRIC: 2326 ifr = data; 2327 ifr->ifr_metric = ifp->if_metric; 2328 break; 2329 2330 case SIOCGIFMTU: 2331 ifr = data; 2332 ifr->ifr_mtu = ifp->if_mtu; 2333 break; 2334 2335 case SIOCGIFDLT: 2336 ifr = data; 2337 ifr->ifr_dlt = ifp->if_dlt; 2338 break; 2339 2340 case SIOCGIFCAP: 2341 ifcr = data; 2342 ifcr->ifcr_capabilities = ifp->if_capabilities; 2343 ifcr->ifcr_capenable = ifp->if_capenable; 2344 break; 2345 2346 case SIOCSIFMETRIC: 2347 ifr = data; 2348 ifp->if_metric = ifr->ifr_metric; 2349 break; 2350 2351 case SIOCGIFDATA: 2352 ifdr = data; 2353 ifdr->ifdr_data = ifp->if_data; 2354 break; 2355 2356 case SIOCGIFINDEX: 2357 ifr = data; 2358 ifr->ifr_index = ifp->if_index; 2359 break; 2360 2361 case SIOCZIFDATA: 2362 ifdr = data; 2363 ifdr->ifdr_data = ifp->if_data; 2364 /* 2365 * Assumes that the volatile counters that can be 2366 * zero'ed are at the end of if_data. 2367 */ 2368 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2369 offsetof(struct if_data, ifi_ipackets)); 2370 /* 2371 * The memset() clears to the bottm of if_data. In the area, 2372 * if_lastchange is included. Please be careful if new entry 2373 * will be added into if_data or rewite this. 2374 * 2375 * And also, update if_lastchnage. 2376 */ 2377 getnanotime(&ifp->if_lastchange); 2378 break; 2379 case SIOCSIFMTU: 2380 ifr = data; 2381 if (ifp->if_mtu == ifr->ifr_mtu) 2382 break; 2383 ifp->if_mtu = ifr->ifr_mtu; 2384 /* 2385 * If the link MTU changed, do network layer specific procedure. 2386 */ 2387 #ifdef INET6 2388 if (in6_present) 2389 nd6_setmtu(ifp); 2390 #endif 2391 return ENETRESET; 2392 default: 2393 return ENOTTY; 2394 } 2395 return 0; 2396 } 2397 2398 int 2399 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2400 { 2401 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2402 struct ifaddr *ifa; 2403 const struct sockaddr *any, *sa; 2404 union { 2405 struct sockaddr sa; 2406 struct sockaddr_storage ss; 2407 } u, v; 2408 2409 switch (cmd) { 2410 case SIOCSIFADDRPREF: 2411 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2412 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2413 NULL) != 0) 2414 return EPERM; 2415 case SIOCGIFADDRPREF: 2416 break; 2417 default: 2418 return EOPNOTSUPP; 2419 } 2420 2421 /* sanity checks */ 2422 if (data == NULL || ifp == NULL) { 2423 panic("invalid argument to %s", __func__); 2424 /*NOTREACHED*/ 2425 } 2426 2427 /* address must be specified on ADD and DELETE */ 2428 sa = sstocsa(&ifap->ifap_addr); 2429 if (sa->sa_family != sofamily(so)) 2430 return EINVAL; 2431 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2432 return EINVAL; 2433 2434 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2435 2436 IFADDR_FOREACH(ifa, ifp) { 2437 if (ifa->ifa_addr->sa_family != sa->sa_family) 2438 continue; 2439 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2440 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2441 break; 2442 } 2443 if (ifa == NULL) 2444 return EADDRNOTAVAIL; 2445 2446 switch (cmd) { 2447 case SIOCSIFADDRPREF: 2448 ifa->ifa_preference = ifap->ifap_preference; 2449 return 0; 2450 case SIOCGIFADDRPREF: 2451 /* fill in the if_laddrreq structure */ 2452 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2453 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2454 ifap->ifap_preference = ifa->ifa_preference; 2455 return 0; 2456 default: 2457 return EOPNOTSUPP; 2458 } 2459 } 2460 2461 /* 2462 * Interface ioctls. 2463 */ 2464 static int 2465 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2466 { 2467 struct ifnet *ifp; 2468 struct ifreq *ifr; 2469 int error = 0; 2470 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2471 u_long ocmd = cmd; 2472 #endif 2473 short oif_flags; 2474 #ifdef COMPAT_OIFREQ 2475 struct ifreq ifrb; 2476 struct oifreq *oifr = NULL; 2477 #endif 2478 int r; 2479 struct psref psref; 2480 int bound; 2481 2482 switch (cmd) { 2483 #ifdef COMPAT_OIFREQ 2484 case OSIOCGIFCONF: 2485 case OOSIOCGIFCONF: 2486 return compat_ifconf(cmd, data); 2487 #endif 2488 #ifdef COMPAT_OIFDATA 2489 case OSIOCGIFDATA: 2490 case OSIOCZIFDATA: 2491 return compat_ifdatareq(l, cmd, data); 2492 #endif 2493 case SIOCGIFCONF: 2494 return ifconf(cmd, data); 2495 case SIOCINITIFADDR: 2496 return EPERM; 2497 } 2498 2499 #ifdef COMPAT_OIFREQ 2500 cmd = compat_cvtcmd(cmd); 2501 if (cmd != ocmd) { 2502 oifr = data; 2503 data = ifr = &ifrb; 2504 ifreqo2n(oifr, ifr); 2505 } else 2506 #endif 2507 ifr = data; 2508 2509 switch (cmd) { 2510 case SIOCIFCREATE: 2511 case SIOCIFDESTROY: 2512 bound = curlwp_bind(); 2513 if (l != NULL) { 2514 ifp = if_get(ifr->ifr_name, &psref); 2515 error = kauth_authorize_network(l->l_cred, 2516 KAUTH_NETWORK_INTERFACE, 2517 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2518 (void *)cmd, NULL); 2519 if (ifp != NULL) 2520 if_put(ifp, &psref); 2521 if (error != 0) { 2522 curlwp_bindx(bound); 2523 return error; 2524 } 2525 } 2526 mutex_enter(&if_clone_mtx); 2527 r = (cmd == SIOCIFCREATE) ? 2528 if_clone_create(ifr->ifr_name) : 2529 if_clone_destroy(ifr->ifr_name); 2530 mutex_exit(&if_clone_mtx); 2531 curlwp_bindx(bound); 2532 return r; 2533 2534 case SIOCIFGCLONERS: 2535 { 2536 struct if_clonereq *req = (struct if_clonereq *)data; 2537 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2538 &req->ifcr_total); 2539 } 2540 } 2541 2542 bound = curlwp_bind(); 2543 ifp = if_get(ifr->ifr_name, &psref); 2544 if (ifp == NULL) { 2545 curlwp_bindx(bound); 2546 return ENXIO; 2547 } 2548 2549 switch (cmd) { 2550 case SIOCALIFADDR: 2551 case SIOCDLIFADDR: 2552 case SIOCSIFADDRPREF: 2553 case SIOCSIFFLAGS: 2554 case SIOCSIFCAP: 2555 case SIOCSIFMETRIC: 2556 case SIOCZIFDATA: 2557 case SIOCSIFMTU: 2558 case SIOCSIFPHYADDR: 2559 case SIOCDIFPHYADDR: 2560 #ifdef INET6 2561 case SIOCSIFPHYADDR_IN6: 2562 #endif 2563 case SIOCSLIFPHYADDR: 2564 case SIOCADDMULTI: 2565 case SIOCDELMULTI: 2566 case SIOCSIFMEDIA: 2567 case SIOCSDRVSPEC: 2568 case SIOCG80211: 2569 case SIOCS80211: 2570 case SIOCS80211NWID: 2571 case SIOCS80211NWKEY: 2572 case SIOCS80211POWER: 2573 case SIOCS80211BSSID: 2574 case SIOCS80211CHANNEL: 2575 case SIOCSLINKSTR: 2576 if (l != NULL) { 2577 error = kauth_authorize_network(l->l_cred, 2578 KAUTH_NETWORK_INTERFACE, 2579 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2580 (void *)cmd, NULL); 2581 if (error != 0) 2582 goto out; 2583 } 2584 } 2585 2586 oif_flags = ifp->if_flags; 2587 2588 mutex_enter(ifp->if_ioctl_lock); 2589 2590 error = (*ifp->if_ioctl)(ifp, cmd, data); 2591 if (error != ENOTTY) 2592 ; 2593 else if (so->so_proto == NULL) 2594 error = EOPNOTSUPP; 2595 else { 2596 #ifdef COMPAT_OSOCK 2597 error = compat_ifioctl(so, ocmd, cmd, data, l); 2598 #else 2599 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 2600 cmd, data, ifp); 2601 #endif 2602 } 2603 2604 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 2605 if ((ifp->if_flags & IFF_UP) != 0) { 2606 int s = splnet(); 2607 if_up(ifp); 2608 splx(s); 2609 } 2610 } 2611 #ifdef COMPAT_OIFREQ 2612 if (cmd != ocmd) 2613 ifreqn2o(oifr, ifr); 2614 #endif 2615 2616 mutex_exit(ifp->if_ioctl_lock); 2617 out: 2618 if_put(ifp, &psref); 2619 curlwp_bindx(bound); 2620 return error; 2621 } 2622 2623 /* 2624 * Return interface configuration 2625 * of system. List may be used 2626 * in later ioctl's (above) to get 2627 * other information. 2628 * 2629 * Each record is a struct ifreq. Before the addition of 2630 * sockaddr_storage, the API rule was that sockaddr flavors that did 2631 * not fit would extend beyond the struct ifreq, with the next struct 2632 * ifreq starting sa_len beyond the struct sockaddr. Because the 2633 * union in struct ifreq includes struct sockaddr_storage, every kind 2634 * of sockaddr must fit. Thus, there are no longer any overlength 2635 * records. 2636 * 2637 * Records are added to the user buffer if they fit, and ifc_len is 2638 * adjusted to the length that was written. Thus, the user is only 2639 * assured of getting the complete list if ifc_len on return is at 2640 * least sizeof(struct ifreq) less than it was on entry. 2641 * 2642 * If the user buffer pointer is NULL, this routine copies no data and 2643 * returns the amount of space that would be needed. 2644 * 2645 * Invariants: 2646 * ifrp points to the next part of the user's buffer to be used. If 2647 * ifrp != NULL, space holds the number of bytes remaining that we may 2648 * write at ifrp. Otherwise, space holds the number of bytes that 2649 * would have been written had there been adequate space. 2650 */ 2651 /*ARGSUSED*/ 2652 static int 2653 ifconf(u_long cmd, void *data) 2654 { 2655 struct ifconf *ifc = (struct ifconf *)data; 2656 struct ifnet *ifp; 2657 struct ifaddr *ifa; 2658 struct ifreq ifr, *ifrp = NULL; 2659 int space = 0, error = 0; 2660 const int sz = (int)sizeof(struct ifreq); 2661 const bool docopy = ifc->ifc_req != NULL; 2662 int s; 2663 int bound; 2664 struct psref psref; 2665 2666 if (docopy) { 2667 space = ifc->ifc_len; 2668 ifrp = ifc->ifc_req; 2669 } 2670 2671 bound = curlwp_bind(); 2672 s = pserialize_read_enter(); 2673 IFNET_READER_FOREACH(ifp) { 2674 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 2675 pserialize_read_exit(s); 2676 2677 (void)strncpy(ifr.ifr_name, ifp->if_xname, 2678 sizeof(ifr.ifr_name)); 2679 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 2680 error = ENAMETOOLONG; 2681 goto release_exit; 2682 } 2683 if (IFADDR_EMPTY(ifp)) { 2684 /* Interface with no addresses - send zero sockaddr. */ 2685 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 2686 if (!docopy) { 2687 space += sz; 2688 continue; 2689 } 2690 if (space >= sz) { 2691 error = copyout(&ifr, ifrp, sz); 2692 if (error != 0) 2693 goto release_exit; 2694 ifrp++; 2695 space -= sz; 2696 } 2697 } 2698 2699 IFADDR_FOREACH(ifa, ifp) { 2700 struct sockaddr *sa = ifa->ifa_addr; 2701 /* all sockaddrs must fit in sockaddr_storage */ 2702 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 2703 2704 if (!docopy) { 2705 space += sz; 2706 continue; 2707 } 2708 memcpy(&ifr.ifr_space, sa, sa->sa_len); 2709 if (space >= sz) { 2710 error = copyout(&ifr, ifrp, sz); 2711 if (error != 0) 2712 goto release_exit; 2713 ifrp++; space -= sz; 2714 } 2715 } 2716 2717 s = pserialize_read_enter(); 2718 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2719 } 2720 pserialize_read_exit(s); 2721 curlwp_bindx(bound); 2722 2723 if (docopy) { 2724 KASSERT(0 <= space && space <= ifc->ifc_len); 2725 ifc->ifc_len -= space; 2726 } else { 2727 KASSERT(space >= 0); 2728 ifc->ifc_len = space; 2729 } 2730 return (0); 2731 2732 release_exit: 2733 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 2734 curlwp_bindx(bound); 2735 return error; 2736 } 2737 2738 int 2739 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 2740 { 2741 uint8_t len; 2742 #ifdef COMPAT_OIFREQ 2743 struct ifreq ifrb; 2744 struct oifreq *oifr = NULL; 2745 u_long ocmd = cmd; 2746 cmd = compat_cvtcmd(cmd); 2747 if (cmd != ocmd) { 2748 oifr = (struct oifreq *)(void *)ifr; 2749 ifr = &ifrb; 2750 ifreqo2n(oifr, ifr); 2751 len = sizeof(oifr->ifr_addr); 2752 } else 2753 #endif 2754 len = sizeof(ifr->ifr_ifru.ifru_space); 2755 2756 if (len < sa->sa_len) 2757 return EFBIG; 2758 2759 memset(&ifr->ifr_addr, 0, len); 2760 sockaddr_copy(&ifr->ifr_addr, len, sa); 2761 2762 #ifdef COMPAT_OIFREQ 2763 if (cmd != ocmd) 2764 ifreqn2o(oifr, ifr); 2765 #endif 2766 return 0; 2767 } 2768 2769 /* 2770 * wrapper function for the drivers which doesn't have if_transmit(). 2771 */ 2772 int 2773 if_transmit(struct ifnet *ifp, struct mbuf *m) 2774 { 2775 int s, error; 2776 2777 s = splnet(); 2778 2779 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2780 if (error != 0) { 2781 /* mbuf is already freed */ 2782 goto out; 2783 } 2784 2785 ifp->if_obytes += m->m_pkthdr.len;; 2786 if (m->m_flags & M_MCAST) 2787 ifp->if_omcasts++; 2788 2789 if ((ifp->if_flags & IFF_OACTIVE) == 0) 2790 (*ifp->if_start)(ifp); 2791 out: 2792 splx(s); 2793 2794 return error; 2795 } 2796 2797 /* 2798 * Queue message on interface, and start output if interface 2799 * not yet active. 2800 */ 2801 int 2802 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 2803 { 2804 2805 return (*ifp->if_transmit)(ifp, m); 2806 } 2807 2808 /* 2809 * Queue message on interface, possibly using a second fast queue 2810 */ 2811 int 2812 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 2813 { 2814 int error = 0; 2815 2816 if (ifq != NULL 2817 #ifdef ALTQ 2818 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 2819 #endif 2820 ) { 2821 if (IF_QFULL(ifq)) { 2822 IF_DROP(&ifp->if_snd); 2823 m_freem(m); 2824 if (error == 0) 2825 error = ENOBUFS; 2826 } else 2827 IF_ENQUEUE(ifq, m); 2828 } else 2829 IFQ_ENQUEUE(&ifp->if_snd, m, error); 2830 if (error != 0) { 2831 ++ifp->if_oerrors; 2832 return error; 2833 } 2834 return 0; 2835 } 2836 2837 int 2838 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 2839 { 2840 int rc; 2841 2842 if (ifp->if_initaddr != NULL) 2843 rc = (*ifp->if_initaddr)(ifp, ifa, src); 2844 else if (src || 2845 /* FIXME: may not hold if_ioctl_lock */ 2846 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 2847 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 2848 2849 return rc; 2850 } 2851 2852 int 2853 if_do_dad(struct ifnet *ifp) 2854 { 2855 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2856 return 0; 2857 2858 switch (ifp->if_type) { 2859 case IFT_FAITH: 2860 /* 2861 * These interfaces do not have the IFF_LOOPBACK flag, 2862 * but loop packets back. We do not have to do DAD on such 2863 * interfaces. We should even omit it, because loop-backed 2864 * responses would confuse the DAD procedure. 2865 */ 2866 return 0; 2867 default: 2868 /* 2869 * Our DAD routine requires the interface up and running. 2870 * However, some interfaces can be up before the RUNNING 2871 * status. Additionaly, users may try to assign addresses 2872 * before the interface becomes up (or running). 2873 * We simply skip DAD in such a case as a work around. 2874 * XXX: we should rather mark "tentative" on such addresses, 2875 * and do DAD after the interface becomes ready. 2876 */ 2877 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 2878 (IFF_UP|IFF_RUNNING)) 2879 return 0; 2880 2881 return 1; 2882 } 2883 } 2884 2885 int 2886 if_flags_set(ifnet_t *ifp, const short flags) 2887 { 2888 int rc; 2889 2890 if (ifp->if_setflags != NULL) 2891 rc = (*ifp->if_setflags)(ifp, flags); 2892 else { 2893 short cantflags, chgdflags; 2894 struct ifreq ifr; 2895 2896 chgdflags = ifp->if_flags ^ flags; 2897 cantflags = chgdflags & IFF_CANTCHANGE; 2898 2899 if (cantflags != 0) 2900 ifp->if_flags ^= cantflags; 2901 2902 /* Traditionally, we do not call if_ioctl after 2903 * setting/clearing only IFF_PROMISC if the interface 2904 * isn't IFF_UP. Uphold that tradition. 2905 */ 2906 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 2907 return 0; 2908 2909 memset(&ifr, 0, sizeof(ifr)); 2910 2911 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 2912 /* FIXME: may not hold if_ioctl_lock */ 2913 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 2914 2915 if (rc != 0 && cantflags != 0) 2916 ifp->if_flags ^= cantflags; 2917 } 2918 2919 return rc; 2920 } 2921 2922 int 2923 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 2924 { 2925 int rc; 2926 struct ifreq ifr; 2927 2928 if (ifp->if_mcastop != NULL) 2929 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 2930 else { 2931 ifreq_setaddr(cmd, &ifr, sa); 2932 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 2933 } 2934 2935 return rc; 2936 } 2937 2938 static void 2939 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 2940 struct ifaltq *ifq) 2941 { 2942 const struct sysctlnode *cnode, *rnode; 2943 2944 if (sysctl_createv(clog, 0, NULL, &rnode, 2945 CTLFLAG_PERMANENT, 2946 CTLTYPE_NODE, "interfaces", 2947 SYSCTL_DESCR("Per-interface controls"), 2948 NULL, 0, NULL, 0, 2949 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 2950 goto bad; 2951 2952 if (sysctl_createv(clog, 0, &rnode, &rnode, 2953 CTLFLAG_PERMANENT, 2954 CTLTYPE_NODE, ifname, 2955 SYSCTL_DESCR("Interface controls"), 2956 NULL, 0, NULL, 0, 2957 CTL_CREATE, CTL_EOL) != 0) 2958 goto bad; 2959 2960 if (sysctl_createv(clog, 0, &rnode, &rnode, 2961 CTLFLAG_PERMANENT, 2962 CTLTYPE_NODE, "sndq", 2963 SYSCTL_DESCR("Interface output queue controls"), 2964 NULL, 0, NULL, 0, 2965 CTL_CREATE, CTL_EOL) != 0) 2966 goto bad; 2967 2968 if (sysctl_createv(clog, 0, &rnode, &cnode, 2969 CTLFLAG_PERMANENT, 2970 CTLTYPE_INT, "len", 2971 SYSCTL_DESCR("Current output queue length"), 2972 NULL, 0, &ifq->ifq_len, 0, 2973 CTL_CREATE, CTL_EOL) != 0) 2974 goto bad; 2975 2976 if (sysctl_createv(clog, 0, &rnode, &cnode, 2977 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2978 CTLTYPE_INT, "maxlen", 2979 SYSCTL_DESCR("Maximum allowed output queue length"), 2980 NULL, 0, &ifq->ifq_maxlen, 0, 2981 CTL_CREATE, CTL_EOL) != 0) 2982 goto bad; 2983 2984 if (sysctl_createv(clog, 0, &rnode, &cnode, 2985 CTLFLAG_PERMANENT, 2986 CTLTYPE_INT, "drops", 2987 SYSCTL_DESCR("Packets dropped due to full output queue"), 2988 NULL, 0, &ifq->ifq_drops, 0, 2989 CTL_CREATE, CTL_EOL) != 0) 2990 goto bad; 2991 2992 return; 2993 bad: 2994 printf("%s: could not attach sysctl nodes\n", ifname); 2995 return; 2996 } 2997 2998 #if defined(INET) || defined(INET6) 2999 3000 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3001 static int \ 3002 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3003 { \ 3004 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3005 } 3006 3007 #if defined(INET) 3008 static int 3009 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3010 { 3011 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3012 } 3013 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3014 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3015 #endif 3016 3017 #if defined(INET6) 3018 static int 3019 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3020 { 3021 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3022 } 3023 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3024 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3025 #endif 3026 3027 static void 3028 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3029 { 3030 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3031 const char *pfname = NULL, *ipname = NULL; 3032 int ipn = 0, qid = 0; 3033 3034 switch (pf) { 3035 #if defined(INET) 3036 case PF_INET: 3037 len_func = sysctl_net_ip_pktq_items; 3038 maxlen_func = sysctl_net_ip_pktq_maxlen; 3039 drops_func = sysctl_net_ip_pktq_drops; 3040 pfname = "inet", ipn = IPPROTO_IP; 3041 ipname = "ip", qid = IPCTL_IFQ; 3042 break; 3043 #endif 3044 #if defined(INET6) 3045 case PF_INET6: 3046 len_func = sysctl_net_ip6_pktq_items; 3047 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3048 drops_func = sysctl_net_ip6_pktq_drops; 3049 pfname = "inet6", ipn = IPPROTO_IPV6; 3050 ipname = "ip6", qid = IPV6CTL_IFQ; 3051 break; 3052 #endif 3053 default: 3054 KASSERT(false); 3055 } 3056 3057 sysctl_createv(clog, 0, NULL, NULL, 3058 CTLFLAG_PERMANENT, 3059 CTLTYPE_NODE, pfname, NULL, 3060 NULL, 0, NULL, 0, 3061 CTL_NET, pf, CTL_EOL); 3062 sysctl_createv(clog, 0, NULL, NULL, 3063 CTLFLAG_PERMANENT, 3064 CTLTYPE_NODE, ipname, NULL, 3065 NULL, 0, NULL, 0, 3066 CTL_NET, pf, ipn, CTL_EOL); 3067 sysctl_createv(clog, 0, NULL, NULL, 3068 CTLFLAG_PERMANENT, 3069 CTLTYPE_NODE, "ifq", 3070 SYSCTL_DESCR("Protocol input queue controls"), 3071 NULL, 0, NULL, 0, 3072 CTL_NET, pf, ipn, qid, CTL_EOL); 3073 3074 sysctl_createv(clog, 0, NULL, NULL, 3075 CTLFLAG_PERMANENT, 3076 CTLTYPE_INT, "len", 3077 SYSCTL_DESCR("Current input queue length"), 3078 len_func, 0, NULL, 0, 3079 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3080 sysctl_createv(clog, 0, NULL, NULL, 3081 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3082 CTLTYPE_INT, "maxlen", 3083 SYSCTL_DESCR("Maximum allowed input queue length"), 3084 maxlen_func, 0, NULL, 0, 3085 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3086 sysctl_createv(clog, 0, NULL, NULL, 3087 CTLFLAG_PERMANENT, 3088 CTLTYPE_INT, "drops", 3089 SYSCTL_DESCR("Packets dropped due to full input queue"), 3090 drops_func, 0, NULL, 0, 3091 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3092 } 3093 #endif /* INET || INET6 */ 3094 3095 static int 3096 if_sdl_sysctl(SYSCTLFN_ARGS) 3097 { 3098 struct ifnet *ifp; 3099 const struct sockaddr_dl *sdl; 3100 struct psref psref; 3101 int error = 0; 3102 int bound; 3103 3104 if (namelen != 1) 3105 return EINVAL; 3106 3107 bound = curlwp_bind(); 3108 ifp = if_get_byindex(name[0], &psref); 3109 if (ifp == NULL) { 3110 error = ENODEV; 3111 goto out0; 3112 } 3113 3114 sdl = ifp->if_sadl; 3115 if (sdl == NULL) { 3116 *oldlenp = 0; 3117 goto out1; 3118 } 3119 3120 if (oldp == NULL) { 3121 *oldlenp = sdl->sdl_alen; 3122 goto out1; 3123 } 3124 3125 if (*oldlenp >= sdl->sdl_alen) 3126 *oldlenp = sdl->sdl_alen; 3127 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3128 out1: 3129 if_put(ifp, &psref); 3130 out0: 3131 curlwp_bindx(bound); 3132 return error; 3133 } 3134 3135 SYSCTL_SETUP(sysctl_net_sdl_setup, "sysctl net.sdl subtree setup") 3136 { 3137 const struct sysctlnode *rnode = NULL; 3138 3139 sysctl_createv(clog, 0, NULL, &rnode, 3140 CTLFLAG_PERMANENT, 3141 CTLTYPE_NODE, "sdl", 3142 SYSCTL_DESCR("Get active link-layer address"), 3143 if_sdl_sysctl, 0, NULL, 0, 3144 CTL_NET, CTL_CREATE, CTL_EOL); 3145 } 3146
Indexes created Sun Sep 21 20:09:37 GMT 2025