if.c revision 1.379
1 /* $NetBSD: if.c,v 1.379 2017/02/16 08:13:43 knakahara Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by William Studenmund and Jason R. Thorpe. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 34 * All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the project nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 */ 60 61 /* 62 * Copyright (c) 1980, 1986, 1993 63 * The Regents of the University of California. All rights reserved. 64 * 65 * Redistribution and use in source and binary forms, with or without 66 * modification, are permitted provided that the following conditions 67 * are met: 68 * 1. Redistributions of source code must retain the above copyright 69 * notice, this list of conditions and the following disclaimer. 70 * 2. Redistributions in binary form must reproduce the above copyright 71 * notice, this list of conditions and the following disclaimer in the 72 * documentation and/or other materials provided with the distribution. 73 * 3. Neither the name of the University nor the names of its contributors 74 * may be used to endorse or promote products derived from this software 75 * without specific prior written permission. 76 * 77 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 78 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 79 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 80 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 81 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 82 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 83 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 84 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 85 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 86 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 87 * SUCH DAMAGE. 88 * 89 * @(#)if.c 8.5 (Berkeley) 1/9/95 90 */ 91 92 #include <sys/cdefs.h> 93 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.379 2017/02/16 08:13:43 knakahara Exp $"); 94 95 #if defined(_KERNEL_OPT) 96 #include "opt_inet.h" 97 #include "opt_ipsec.h" 98 99 #include "opt_atalk.h" 100 #include "opt_natm.h" 101 #include "opt_wlan.h" 102 #include "opt_net_mpsafe.h" 103 #endif 104 105 #include <sys/param.h> 106 #include <sys/mbuf.h> 107 #include <sys/systm.h> 108 #include <sys/callout.h> 109 #include <sys/proc.h> 110 #include <sys/socket.h> 111 #include <sys/socketvar.h> 112 #include <sys/domain.h> 113 #include <sys/protosw.h> 114 #include <sys/kernel.h> 115 #include <sys/ioctl.h> 116 #include <sys/sysctl.h> 117 #include <sys/syslog.h> 118 #include <sys/kauth.h> 119 #include <sys/kmem.h> 120 #include <sys/xcall.h> 121 #include <sys/cpu.h> 122 #include <sys/intr.h> 123 124 #include <net/if.h> 125 #include <net/if_dl.h> 126 #include <net/if_ether.h> 127 #include <net/if_media.h> 128 #include <net80211/ieee80211.h> 129 #include <net80211/ieee80211_ioctl.h> 130 #include <net/if_types.h> 131 #include <net/route.h> 132 #include <net/netisr.h> 133 #include <sys/module.h> 134 #ifdef NETATALK 135 #include <netatalk/at_extern.h> 136 #include <netatalk/at.h> 137 #endif 138 #include <net/pfil.h> 139 #include <netinet/in.h> 140 #include <netinet/in_var.h> 141 #ifndef IPSEC 142 #include <netinet/ip_encap.h> 143 #endif 144 #include <net/bpf.h> 145 146 #ifdef INET6 147 #include <netinet6/in6_var.h> 148 #include <netinet6/nd6.h> 149 #endif 150 151 #include "ether.h" 152 #include "fddi.h" 153 #include "token.h" 154 155 #include "carp.h" 156 #if NCARP > 0 157 #include <netinet/ip_carp.h> 158 #endif 159 160 #include <compat/sys/sockio.h> 161 #include <compat/sys/socket.h> 162 163 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 164 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 165 166 /* 167 * Global list of interfaces. 168 */ 169 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 170 struct ifnet_head ifnet_list; 171 172 struct pslist_head ifnet_pslist; 173 static ifnet_t ** ifindex2ifnet = NULL; 174 static u_int if_index = 1; 175 static size_t if_indexlim = 0; 176 static uint64_t index_gen; 177 /* Mutex to protect the above objects. */ 178 kmutex_t ifnet_mtx __cacheline_aligned; 179 struct psref_class *ifnet_psref_class __read_mostly; 180 static pserialize_t ifnet_psz; 181 182 static kmutex_t if_clone_mtx; 183 184 struct ifnet *lo0ifp; 185 int ifqmaxlen = IFQ_MAXLEN; 186 187 struct psref_class *ifa_psref_class __read_mostly; 188 189 static int if_delroute_matcher(struct rtentry *, void *); 190 191 static bool if_is_unit(const char *); 192 static struct if_clone *if_clone_lookup(const char *, int *); 193 194 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 195 static int if_cloners_count; 196 197 /* Packet filtering hook for interfaces. */ 198 pfil_head_t * if_pfil __read_mostly; 199 200 static kauth_listener_t if_listener; 201 202 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 203 static void if_detach_queues(struct ifnet *, struct ifqueue *); 204 static void sysctl_sndq_setup(struct sysctllog **, const char *, 205 struct ifaltq *); 206 static void if_slowtimo(void *); 207 static void if_free_sadl(struct ifnet *); 208 static void if_attachdomain1(struct ifnet *); 209 static int ifconf(u_long, void *); 210 static int if_transmit(struct ifnet *, struct mbuf *); 211 static int if_clone_create(const char *); 212 static int if_clone_destroy(const char *); 213 static void if_link_state_change_si(void *); 214 215 struct if_percpuq { 216 struct ifnet *ipq_ifp; 217 void *ipq_si; 218 struct percpu *ipq_ifqs; /* struct ifqueue */ 219 }; 220 221 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 222 223 static void if_percpuq_drops(void *, void *, struct cpu_info *); 224 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 225 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 226 struct if_percpuq *); 227 228 struct if_deferred_start { 229 struct ifnet *ids_ifp; 230 void (*ids_if_start)(struct ifnet *); 231 void *ids_si; 232 }; 233 234 static void if_deferred_start_softint(void *); 235 static void if_deferred_start_common(struct ifnet *); 236 static void if_deferred_start_destroy(struct ifnet *); 237 238 #if defined(INET) || defined(INET6) 239 static void sysctl_net_pktq_setup(struct sysctllog **, int); 240 #endif 241 242 static void if_sysctl_setup(struct sysctllog **); 243 244 /* 245 * Pointer to stub or real compat_cvtcmd() depending on presence of 246 * the compat module 247 */ 248 u_long stub_compat_cvtcmd(u_long); 249 u_long (*vec_compat_cvtcmd)(u_long) = stub_compat_cvtcmd; 250 251 /* Similarly, pointer to compat_ifioctl() if it is present */ 252 253 int (*vec_compat_ifioctl)(struct socket *, u_long, u_long, void *, 254 struct lwp *) = NULL; 255 256 /* The stub version of compat_cvtcmd() */ 257 u_long stub_compat_cvtcmd(u_long cmd) 258 { 259 260 return cmd; 261 } 262 263 static int 264 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 265 void *arg0, void *arg1, void *arg2, void *arg3) 266 { 267 int result; 268 enum kauth_network_req req; 269 270 result = KAUTH_RESULT_DEFER; 271 req = (enum kauth_network_req)arg1; 272 273 if (action != KAUTH_NETWORK_INTERFACE) 274 return result; 275 276 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 277 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 278 result = KAUTH_RESULT_ALLOW; 279 280 return result; 281 } 282 283 /* 284 * Network interface utility routines. 285 * 286 * Routines with ifa_ifwith* names take sockaddr *'s as 287 * parameters. 288 */ 289 void 290 ifinit(void) 291 { 292 293 if_sysctl_setup(NULL); 294 295 #if (defined(INET) || defined(INET6)) && !defined(IPSEC) 296 encapinit(); 297 #endif 298 299 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 300 if_listener_cb, NULL); 301 302 /* interfaces are available, inform socket code */ 303 ifioctl = doifioctl; 304 } 305 306 /* 307 * XXX Initialization before configure(). 308 * XXX hack to get pfil_add_hook working in autoconf. 309 */ 310 void 311 ifinit1(void) 312 { 313 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 314 315 TAILQ_INIT(&ifnet_list); 316 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 317 ifnet_psz = pserialize_create(); 318 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 319 ifa_psref_class = psref_class_create("ifa", IPL_SOFTNET); 320 PSLIST_INIT(&ifnet_pslist); 321 322 if_indexlim = 8; 323 324 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 325 KASSERT(if_pfil != NULL); 326 327 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 328 etherinit(); 329 #endif 330 } 331 332 ifnet_t * 333 if_alloc(u_char type) 334 { 335 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 336 } 337 338 void 339 if_free(ifnet_t *ifp) 340 { 341 kmem_free(ifp, sizeof(ifnet_t)); 342 } 343 344 void 345 if_initname(struct ifnet *ifp, const char *name, int unit) 346 { 347 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 348 "%s%d", name, unit); 349 } 350 351 /* 352 * Null routines used while an interface is going away. These routines 353 * just return an error. 354 */ 355 356 int 357 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 358 const struct sockaddr *so, const struct rtentry *rt) 359 { 360 361 return ENXIO; 362 } 363 364 void 365 if_nullinput(struct ifnet *ifp, struct mbuf *m) 366 { 367 368 /* Nothing. */ 369 } 370 371 void 372 if_nullstart(struct ifnet *ifp) 373 { 374 375 /* Nothing. */ 376 } 377 378 int 379 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 380 { 381 382 return ENXIO; 383 } 384 385 int 386 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 387 { 388 389 return ENXIO; 390 } 391 392 int 393 if_nullinit(struct ifnet *ifp) 394 { 395 396 return ENXIO; 397 } 398 399 void 400 if_nullstop(struct ifnet *ifp, int disable) 401 { 402 403 /* Nothing. */ 404 } 405 406 void 407 if_nullslowtimo(struct ifnet *ifp) 408 { 409 410 /* Nothing. */ 411 } 412 413 void 414 if_nulldrain(struct ifnet *ifp) 415 { 416 417 /* Nothing. */ 418 } 419 420 void 421 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 422 { 423 struct ifaddr *ifa; 424 struct sockaddr_dl *sdl; 425 426 ifp->if_addrlen = addrlen; 427 if_alloc_sadl(ifp); 428 ifa = ifp->if_dl; 429 sdl = satosdl(ifa->ifa_addr); 430 431 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 432 if (factory) { 433 ifp->if_hwdl = ifp->if_dl; 434 ifaref(ifp->if_hwdl); 435 } 436 /* TBD routing socket */ 437 } 438 439 struct ifaddr * 440 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 441 { 442 unsigned socksize, ifasize; 443 int addrlen, namelen; 444 struct sockaddr_dl *mask, *sdl; 445 struct ifaddr *ifa; 446 447 namelen = strlen(ifp->if_xname); 448 addrlen = ifp->if_addrlen; 449 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 450 ifasize = sizeof(*ifa) + 2 * socksize; 451 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 452 453 sdl = (struct sockaddr_dl *)(ifa + 1); 454 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 455 456 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 457 ifp->if_xname, namelen, NULL, addrlen); 458 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 459 memset(&mask->sdl_data[0], 0xff, namelen); 460 ifa->ifa_rtrequest = link_rtrequest; 461 ifa->ifa_addr = (struct sockaddr *)sdl; 462 ifa->ifa_netmask = (struct sockaddr *)mask; 463 ifa_psref_init(ifa); 464 465 *sdlp = sdl; 466 467 return ifa; 468 } 469 470 static void 471 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 472 { 473 const struct sockaddr_dl *sdl; 474 475 ifp->if_dl = ifa; 476 ifaref(ifa); 477 sdl = satosdl(ifa->ifa_addr); 478 ifp->if_sadl = sdl; 479 } 480 481 /* 482 * Allocate the link level name for the specified interface. This 483 * is an attachment helper. It must be called after ifp->if_addrlen 484 * is initialized, which may not be the case when if_attach() is 485 * called. 486 */ 487 void 488 if_alloc_sadl(struct ifnet *ifp) 489 { 490 struct ifaddr *ifa; 491 const struct sockaddr_dl *sdl; 492 493 /* 494 * If the interface already has a link name, release it 495 * now. This is useful for interfaces that can change 496 * link types, and thus switch link names often. 497 */ 498 if (ifp->if_sadl != NULL) 499 if_free_sadl(ifp); 500 501 ifa = if_dl_create(ifp, &sdl); 502 503 ifa_insert(ifp, ifa); 504 if_sadl_setrefs(ifp, ifa); 505 } 506 507 static void 508 if_deactivate_sadl(struct ifnet *ifp) 509 { 510 struct ifaddr *ifa; 511 512 KASSERT(ifp->if_dl != NULL); 513 514 ifa = ifp->if_dl; 515 516 ifp->if_sadl = NULL; 517 518 ifp->if_dl = NULL; 519 ifafree(ifa); 520 } 521 522 static void 523 if_replace_sadl(struct ifnet *ifp, struct ifaddr *ifa) 524 { 525 struct ifaddr *old; 526 527 KASSERT(ifp->if_dl != NULL); 528 529 old = ifp->if_dl; 530 531 ifaref(ifa); 532 /* XXX Update if_dl and if_sadl atomically */ 533 ifp->if_dl = ifa; 534 ifp->if_sadl = satosdl(ifa->ifa_addr); 535 536 ifafree(old); 537 } 538 539 void 540 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa0, 541 const struct sockaddr_dl *sdl) 542 { 543 int s, ss; 544 struct ifaddr *ifa; 545 int bound = curlwp_bind(); 546 547 KASSERT(ifa_held(ifa0)); 548 549 s = splsoftnet(); 550 551 if_replace_sadl(ifp, ifa0); 552 553 ss = pserialize_read_enter(); 554 IFADDR_READER_FOREACH(ifa, ifp) { 555 struct psref psref; 556 ifa_acquire(ifa, &psref); 557 pserialize_read_exit(ss); 558 559 rtinit(ifa, RTM_LLINFO_UPD, 0); 560 561 ss = pserialize_read_enter(); 562 ifa_release(ifa, &psref); 563 } 564 pserialize_read_exit(ss); 565 566 splx(s); 567 curlwp_bindx(bound); 568 } 569 570 /* 571 * Free the link level name for the specified interface. This is 572 * a detach helper. This is called from if_detach(). 573 */ 574 static void 575 if_free_sadl(struct ifnet *ifp) 576 { 577 struct ifaddr *ifa; 578 int s; 579 580 ifa = ifp->if_dl; 581 if (ifa == NULL) { 582 KASSERT(ifp->if_sadl == NULL); 583 return; 584 } 585 586 KASSERT(ifp->if_sadl != NULL); 587 588 s = splsoftnet(); 589 rtinit(ifa, RTM_DELETE, 0); 590 ifa_remove(ifp, ifa); 591 if_deactivate_sadl(ifp); 592 if (ifp->if_hwdl == ifa) { 593 ifafree(ifa); 594 ifp->if_hwdl = NULL; 595 } 596 splx(s); 597 } 598 599 static void 600 if_getindex(ifnet_t *ifp) 601 { 602 bool hitlimit = false; 603 604 ifp->if_index_gen = index_gen++; 605 606 ifp->if_index = if_index; 607 if (ifindex2ifnet == NULL) { 608 if_index++; 609 goto skip; 610 } 611 while (if_byindex(ifp->if_index)) { 612 /* 613 * If we hit USHRT_MAX, we skip back to 0 since 614 * there are a number of places where the value 615 * of if_index or if_index itself is compared 616 * to or stored in an unsigned short. By 617 * jumping back, we won't botch those assignments 618 * or comparisons. 619 */ 620 if (++if_index == 0) { 621 if_index = 1; 622 } else if (if_index == USHRT_MAX) { 623 /* 624 * However, if we have to jump back to 625 * zero *twice* without finding an empty 626 * slot in ifindex2ifnet[], then there 627 * there are too many (>65535) interfaces. 628 */ 629 if (hitlimit) { 630 panic("too many interfaces"); 631 } 632 hitlimit = true; 633 if_index = 1; 634 } 635 ifp->if_index = if_index; 636 } 637 skip: 638 /* 639 * ifindex2ifnet is indexed by if_index. Since if_index will 640 * grow dynamically, it should grow too. 641 */ 642 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 643 size_t m, n, oldlim; 644 void *q; 645 646 oldlim = if_indexlim; 647 while (ifp->if_index >= if_indexlim) 648 if_indexlim <<= 1; 649 650 /* grow ifindex2ifnet */ 651 m = oldlim * sizeof(struct ifnet *); 652 n = if_indexlim * sizeof(struct ifnet *); 653 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 654 if (ifindex2ifnet != NULL) { 655 memcpy(q, ifindex2ifnet, m); 656 free(ifindex2ifnet, M_IFADDR); 657 } 658 ifindex2ifnet = (struct ifnet **)q; 659 } 660 ifindex2ifnet[ifp->if_index] = ifp; 661 } 662 663 /* 664 * Initialize an interface and assign an index for it. 665 * 666 * It must be called prior to a device specific attach routine 667 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 668 * and be followed by if_register: 669 * 670 * if_initialize(ifp); 671 * ether_ifattach(ifp, enaddr); 672 * if_register(ifp); 673 */ 674 void 675 if_initialize(ifnet_t *ifp) 676 { 677 KASSERT(if_indexlim > 0); 678 TAILQ_INIT(&ifp->if_addrlist); 679 680 /* 681 * Link level name is allocated later by a separate call to 682 * if_alloc_sadl(). 683 */ 684 685 if (ifp->if_snd.ifq_maxlen == 0) 686 ifp->if_snd.ifq_maxlen = ifqmaxlen; 687 688 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 689 690 ifp->if_link_state = LINK_STATE_UNKNOWN; 691 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 692 693 ifp->if_capenable = 0; 694 ifp->if_csum_flags_tx = 0; 695 ifp->if_csum_flags_rx = 0; 696 697 #ifdef ALTQ 698 ifp->if_snd.altq_type = 0; 699 ifp->if_snd.altq_disc = NULL; 700 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 701 ifp->if_snd.altq_tbr = NULL; 702 ifp->if_snd.altq_ifp = ifp; 703 #endif 704 705 IFQ_LOCK_INIT(&ifp->if_snd); 706 707 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 708 pfil_run_ifhooks(if_pfil, PFIL_IFNET_ATTACH, ifp); 709 710 IF_AFDATA_LOCK_INIT(ifp); 711 712 if (if_is_link_state_changeable(ifp)) { 713 ifp->if_link_si = softint_establish(SOFTINT_NET, 714 if_link_state_change_si, ifp); 715 if (ifp->if_link_si == NULL) 716 panic("%s: softint_establish() failed", __func__); 717 } 718 719 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 720 PSLIST_INIT(&ifp->if_addr_pslist); 721 psref_target_init(&ifp->if_psref, ifnet_psref_class); 722 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 723 724 IFNET_LOCK(); 725 if_getindex(ifp); 726 IFNET_UNLOCK(); 727 } 728 729 /* 730 * Register an interface to the list of "active" interfaces. 731 */ 732 void 733 if_register(ifnet_t *ifp) 734 { 735 /* 736 * If the driver has not supplied its own if_ioctl, then 737 * supply the default. 738 */ 739 if (ifp->if_ioctl == NULL) 740 ifp->if_ioctl = ifioctl_common; 741 742 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 743 744 if (!STAILQ_EMPTY(&domains)) 745 if_attachdomain1(ifp); 746 747 /* Announce the interface. */ 748 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 749 750 if (ifp->if_slowtimo != NULL) { 751 ifp->if_slowtimo_ch = 752 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 753 callout_init(ifp->if_slowtimo_ch, 0); 754 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 755 if_slowtimo(ifp); 756 } 757 758 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 759 ifp->if_transmit = if_transmit; 760 761 IFNET_LOCK(); 762 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 763 IFNET_WRITER_INSERT_TAIL(ifp); 764 IFNET_UNLOCK(); 765 } 766 767 /* 768 * The if_percpuq framework 769 * 770 * It allows network device drivers to execute the network stack 771 * in softint (so called softint-based if_input). It utilizes 772 * softint and percpu ifqueue. It doesn't distribute any packets 773 * between CPUs, unlike pktqueue(9). 774 * 775 * Currently we support two options for device drivers to apply the framework: 776 * - Use it implicitly with less changes 777 * - If you use if_attach in driver's _attach function and if_input in 778 * driver's Rx interrupt handler, a packet is queued and a softint handles 779 * the packet implicitly 780 * - Use it explicitly in each driver (recommended) 781 * - You can use if_percpuq_* directly in your driver 782 * - In this case, you need to allocate struct if_percpuq in driver's softc 783 * - See wm(4) as a reference implementation 784 */ 785 786 static void 787 if_percpuq_softint(void *arg) 788 { 789 struct if_percpuq *ipq = arg; 790 struct ifnet *ifp = ipq->ipq_ifp; 791 struct mbuf *m; 792 793 while ((m = if_percpuq_dequeue(ipq)) != NULL) { 794 ifp->if_ipackets++; 795 bpf_mtap(ifp, m); 796 797 ifp->_if_input(ifp, m); 798 } 799 } 800 801 static void 802 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 803 { 804 struct ifqueue *const ifq = p; 805 806 memset(ifq, 0, sizeof(*ifq)); 807 ifq->ifq_maxlen = IFQ_MAXLEN; 808 } 809 810 struct if_percpuq * 811 if_percpuq_create(struct ifnet *ifp) 812 { 813 struct if_percpuq *ipq; 814 815 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 816 if (ipq == NULL) 817 panic("kmem_zalloc failed"); 818 819 ipq->ipq_ifp = ifp; 820 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 821 if_percpuq_softint, ipq); 822 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 823 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 824 825 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 826 827 return ipq; 828 } 829 830 static struct mbuf * 831 if_percpuq_dequeue(struct if_percpuq *ipq) 832 { 833 struct mbuf *m; 834 struct ifqueue *ifq; 835 int s; 836 837 s = splnet(); 838 ifq = percpu_getref(ipq->ipq_ifqs); 839 IF_DEQUEUE(ifq, m); 840 percpu_putref(ipq->ipq_ifqs); 841 splx(s); 842 843 return m; 844 } 845 846 static void 847 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 848 { 849 struct ifqueue *const ifq = p; 850 851 IF_PURGE(ifq); 852 } 853 854 void 855 if_percpuq_destroy(struct if_percpuq *ipq) 856 { 857 858 /* if_detach may already destroy it */ 859 if (ipq == NULL) 860 return; 861 862 softint_disestablish(ipq->ipq_si); 863 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 864 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 865 kmem_free(ipq, sizeof(*ipq)); 866 } 867 868 void 869 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 870 { 871 struct ifqueue *ifq; 872 int s; 873 874 KASSERT(ipq != NULL); 875 876 s = splnet(); 877 ifq = percpu_getref(ipq->ipq_ifqs); 878 if (IF_QFULL(ifq)) { 879 IF_DROP(ifq); 880 percpu_putref(ipq->ipq_ifqs); 881 m_freem(m); 882 goto out; 883 } 884 IF_ENQUEUE(ifq, m); 885 percpu_putref(ipq->ipq_ifqs); 886 887 softint_schedule(ipq->ipq_si); 888 out: 889 splx(s); 890 } 891 892 static void 893 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 894 { 895 struct ifqueue *const ifq = p; 896 int *sum = arg; 897 898 *sum += ifq->ifq_drops; 899 } 900 901 static int 902 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 903 { 904 struct sysctlnode node; 905 struct if_percpuq *ipq; 906 int sum = 0; 907 int error; 908 909 node = *rnode; 910 ipq = node.sysctl_data; 911 912 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 913 914 node.sysctl_data = ∑ 915 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 916 if (error != 0 || newp == NULL) 917 return error; 918 919 return 0; 920 } 921 922 static void 923 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 924 struct if_percpuq *ipq) 925 { 926 const struct sysctlnode *cnode, *rnode; 927 928 if (sysctl_createv(clog, 0, NULL, &rnode, 929 CTLFLAG_PERMANENT, 930 CTLTYPE_NODE, "interfaces", 931 SYSCTL_DESCR("Per-interface controls"), 932 NULL, 0, NULL, 0, 933 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 934 goto bad; 935 936 if (sysctl_createv(clog, 0, &rnode, &rnode, 937 CTLFLAG_PERMANENT, 938 CTLTYPE_NODE, ifname, 939 SYSCTL_DESCR("Interface controls"), 940 NULL, 0, NULL, 0, 941 CTL_CREATE, CTL_EOL) != 0) 942 goto bad; 943 944 if (sysctl_createv(clog, 0, &rnode, &rnode, 945 CTLFLAG_PERMANENT, 946 CTLTYPE_NODE, "rcvq", 947 SYSCTL_DESCR("Interface input queue controls"), 948 NULL, 0, NULL, 0, 949 CTL_CREATE, CTL_EOL) != 0) 950 goto bad; 951 952 #ifdef NOTYET 953 /* XXX Should show each per-CPU queue length? */ 954 if (sysctl_createv(clog, 0, &rnode, &rnode, 955 CTLFLAG_PERMANENT, 956 CTLTYPE_INT, "len", 957 SYSCTL_DESCR("Current input queue length"), 958 sysctl_percpuq_len, 0, NULL, 0, 959 CTL_CREATE, CTL_EOL) != 0) 960 goto bad; 961 962 if (sysctl_createv(clog, 0, &rnode, &cnode, 963 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 964 CTLTYPE_INT, "maxlen", 965 SYSCTL_DESCR("Maximum allowed input queue length"), 966 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 967 CTL_CREATE, CTL_EOL) != 0) 968 goto bad; 969 #endif 970 971 if (sysctl_createv(clog, 0, &rnode, &cnode, 972 CTLFLAG_PERMANENT, 973 CTLTYPE_INT, "drops", 974 SYSCTL_DESCR("Total packets dropped due to full input queue"), 975 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 976 CTL_CREATE, CTL_EOL) != 0) 977 goto bad; 978 979 return; 980 bad: 981 printf("%s: could not attach sysctl nodes\n", ifname); 982 return; 983 } 984 985 /* 986 * The deferred if_start framework 987 * 988 * The common APIs to defer if_start to softint when if_start is requested 989 * from a device driver running in hardware interrupt context. 990 */ 991 /* 992 * Call ifp->if_start (or equivalent) in a dedicated softint for 993 * deferred if_start. 994 */ 995 static void 996 if_deferred_start_softint(void *arg) 997 { 998 struct if_deferred_start *ids = arg; 999 struct ifnet *ifp = ids->ids_ifp; 1000 1001 ids->ids_if_start(ifp); 1002 } 1003 1004 /* 1005 * The default callback function for deferred if_start. 1006 */ 1007 static void 1008 if_deferred_start_common(struct ifnet *ifp) 1009 { 1010 1011 if_start_lock(ifp); 1012 } 1013 1014 static inline bool 1015 if_snd_is_used(struct ifnet *ifp) 1016 { 1017 1018 return ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit || 1019 ALTQ_IS_ENABLED(&ifp->if_snd); 1020 } 1021 1022 /* 1023 * Schedule deferred if_start. 1024 */ 1025 void 1026 if_schedule_deferred_start(struct ifnet *ifp) 1027 { 1028 1029 KASSERT(ifp->if_deferred_start != NULL); 1030 1031 if (if_snd_is_used(ifp) && IFQ_IS_EMPTY(&ifp->if_snd)) 1032 return; 1033 1034 softint_schedule(ifp->if_deferred_start->ids_si); 1035 } 1036 1037 /* 1038 * Create an instance of deferred if_start. A driver should call the function 1039 * only if the driver needs deferred if_start. Drivers can setup their own 1040 * deferred if_start function via 2nd argument. 1041 */ 1042 void 1043 if_deferred_start_init(struct ifnet *ifp, void (*func)(struct ifnet *)) 1044 { 1045 struct if_deferred_start *ids; 1046 1047 ids = kmem_zalloc(sizeof(*ids), KM_SLEEP); 1048 if (ids == NULL) 1049 panic("kmem_zalloc failed"); 1050 1051 ids->ids_ifp = ifp; 1052 ids->ids_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 1053 if_deferred_start_softint, ids); 1054 if (func != NULL) 1055 ids->ids_if_start = func; 1056 else 1057 ids->ids_if_start = if_deferred_start_common; 1058 1059 ifp->if_deferred_start = ids; 1060 } 1061 1062 static void 1063 if_deferred_start_destroy(struct ifnet *ifp) 1064 { 1065 1066 if (ifp->if_deferred_start == NULL) 1067 return; 1068 1069 softint_disestablish(ifp->if_deferred_start->ids_si); 1070 kmem_free(ifp->if_deferred_start, sizeof(*ifp->if_deferred_start)); 1071 ifp->if_deferred_start = NULL; 1072 } 1073 1074 /* 1075 * The common interface input routine that is called by device drivers, 1076 * which should be used only when the driver's rx handler already runs 1077 * in softint. 1078 */ 1079 void 1080 if_input(struct ifnet *ifp, struct mbuf *m) 1081 { 1082 1083 KASSERT(ifp->if_percpuq == NULL); 1084 KASSERT(!cpu_intr_p()); 1085 1086 ifp->if_ipackets++; 1087 bpf_mtap(ifp, m); 1088 1089 ifp->_if_input(ifp, m); 1090 } 1091 1092 /* 1093 * DEPRECATED. Use if_initialize and if_register instead. 1094 * See the above comment of if_initialize. 1095 * 1096 * Note that it implicitly enables if_percpuq to make drivers easy to 1097 * migrate softint-based if_input without much changes. If you don't 1098 * want to enable it, use if_initialize instead. 1099 */ 1100 void 1101 if_attach(ifnet_t *ifp) 1102 { 1103 1104 if_initialize(ifp); 1105 ifp->if_percpuq = if_percpuq_create(ifp); 1106 if_register(ifp); 1107 } 1108 1109 void 1110 if_attachdomain(void) 1111 { 1112 struct ifnet *ifp; 1113 int s; 1114 int bound = curlwp_bind(); 1115 1116 s = pserialize_read_enter(); 1117 IFNET_READER_FOREACH(ifp) { 1118 struct psref psref; 1119 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 1120 pserialize_read_exit(s); 1121 if_attachdomain1(ifp); 1122 s = pserialize_read_enter(); 1123 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 1124 } 1125 pserialize_read_exit(s); 1126 curlwp_bindx(bound); 1127 } 1128 1129 static void 1130 if_attachdomain1(struct ifnet *ifp) 1131 { 1132 struct domain *dp; 1133 int s; 1134 1135 s = splsoftnet(); 1136 1137 /* address family dependent data region */ 1138 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 1139 DOMAIN_FOREACH(dp) { 1140 if (dp->dom_ifattach != NULL) 1141 ifp->if_afdata[dp->dom_family] = 1142 (*dp->dom_ifattach)(ifp); 1143 } 1144 1145 splx(s); 1146 } 1147 1148 /* 1149 * Deactivate an interface. This points all of the procedure 1150 * handles at error stubs. May be called from interrupt context. 1151 */ 1152 void 1153 if_deactivate(struct ifnet *ifp) 1154 { 1155 int s; 1156 1157 s = splsoftnet(); 1158 1159 ifp->if_output = if_nulloutput; 1160 ifp->_if_input = if_nullinput; 1161 ifp->if_start = if_nullstart; 1162 ifp->if_transmit = if_nulltransmit; 1163 ifp->if_ioctl = if_nullioctl; 1164 ifp->if_init = if_nullinit; 1165 ifp->if_stop = if_nullstop; 1166 ifp->if_slowtimo = if_nullslowtimo; 1167 ifp->if_drain = if_nulldrain; 1168 1169 /* No more packets may be enqueued. */ 1170 ifp->if_snd.ifq_maxlen = 0; 1171 1172 splx(s); 1173 } 1174 1175 bool 1176 if_is_deactivated(const struct ifnet *ifp) 1177 { 1178 1179 return ifp->if_output == if_nulloutput; 1180 } 1181 1182 void 1183 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1184 { 1185 struct ifaddr *ifa, *nifa; 1186 int s; 1187 1188 s = pserialize_read_enter(); 1189 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) { 1190 nifa = IFADDR_READER_NEXT(ifa); 1191 if (ifa->ifa_addr->sa_family != family) 1192 continue; 1193 pserialize_read_exit(s); 1194 1195 (*purgeaddr)(ifa); 1196 1197 s = pserialize_read_enter(); 1198 } 1199 pserialize_read_exit(s); 1200 } 1201 1202 #ifdef IFAREF_DEBUG 1203 static struct ifaddr **ifa_list; 1204 static int ifa_list_size; 1205 1206 /* Depends on only one if_attach runs at once */ 1207 static void 1208 if_build_ifa_list(struct ifnet *ifp) 1209 { 1210 struct ifaddr *ifa; 1211 int i; 1212 1213 KASSERT(ifa_list == NULL); 1214 KASSERT(ifa_list_size == 0); 1215 1216 IFADDR_READER_FOREACH(ifa, ifp) 1217 ifa_list_size++; 1218 1219 ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP); 1220 if (ifa_list == NULL) 1221 return; 1222 1223 i = 0; 1224 IFADDR_READER_FOREACH(ifa, ifp) { 1225 ifa_list[i++] = ifa; 1226 ifaref(ifa); 1227 } 1228 } 1229 1230 static void 1231 if_check_and_free_ifa_list(struct ifnet *ifp) 1232 { 1233 int i; 1234 struct ifaddr *ifa; 1235 1236 if (ifa_list == NULL) 1237 return; 1238 1239 for (i = 0; i < ifa_list_size; i++) { 1240 char buf[64]; 1241 1242 ifa = ifa_list[i]; 1243 sockaddr_format(ifa->ifa_addr, buf, sizeof(buf)); 1244 if (ifa->ifa_refcnt > 1) { 1245 log(LOG_WARNING, 1246 "ifa(%s) still referenced (refcnt=%d)\n", 1247 buf, ifa->ifa_refcnt - 1); 1248 } else 1249 log(LOG_DEBUG, 1250 "ifa(%s) not referenced (refcnt=%d)\n", 1251 buf, ifa->ifa_refcnt - 1); 1252 ifafree(ifa); 1253 } 1254 1255 kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size); 1256 ifa_list = NULL; 1257 ifa_list_size = 0; 1258 } 1259 #endif 1260 1261 /* 1262 * Detach an interface from the list of "active" interfaces, 1263 * freeing any resources as we go along. 1264 * 1265 * NOTE: This routine must be called with a valid thread context, 1266 * as it may block. 1267 */ 1268 void 1269 if_detach(struct ifnet *ifp) 1270 { 1271 struct socket so; 1272 struct ifaddr *ifa; 1273 #ifdef IFAREF_DEBUG 1274 struct ifaddr *last_ifa = NULL; 1275 #endif 1276 struct domain *dp; 1277 const struct protosw *pr; 1278 int s, i, family, purged; 1279 uint64_t xc; 1280 1281 #ifdef IFAREF_DEBUG 1282 if_build_ifa_list(ifp); 1283 #endif 1284 /* 1285 * XXX It's kind of lame that we have to have the 1286 * XXX socket structure... 1287 */ 1288 memset(&so, 0, sizeof(so)); 1289 1290 s = splnet(); 1291 1292 sysctl_teardown(&ifp->if_sysctl_log); 1293 mutex_enter(ifp->if_ioctl_lock); 1294 if_deactivate(ifp); 1295 mutex_exit(ifp->if_ioctl_lock); 1296 1297 IFNET_LOCK(); 1298 ifindex2ifnet[ifp->if_index] = NULL; 1299 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1300 IFNET_WRITER_REMOVE(ifp); 1301 pserialize_perform(ifnet_psz); 1302 IFNET_UNLOCK(); 1303 1304 /* Wait for all readers to drain before freeing. */ 1305 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1306 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1307 1308 mutex_obj_free(ifp->if_ioctl_lock); 1309 ifp->if_ioctl_lock = NULL; 1310 1311 if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) { 1312 ifp->if_slowtimo = NULL; 1313 callout_halt(ifp->if_slowtimo_ch, NULL); 1314 callout_destroy(ifp->if_slowtimo_ch); 1315 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1316 } 1317 if_deferred_start_destroy(ifp); 1318 1319 /* 1320 * Do an if_down() to give protocols a chance to do something. 1321 */ 1322 if_down(ifp); 1323 1324 #ifdef ALTQ 1325 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1326 altq_disable(&ifp->if_snd); 1327 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1328 altq_detach(&ifp->if_snd); 1329 #endif 1330 1331 mutex_obj_free(ifp->if_snd.ifq_lock); 1332 1333 #if NCARP > 0 1334 /* Remove the interface from any carp group it is a part of. */ 1335 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1336 carp_ifdetach(ifp); 1337 #endif 1338 1339 /* 1340 * Rip all the addresses off the interface. This should make 1341 * all of the routes go away. 1342 * 1343 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1344 * from the list, including our "cursor", ifa. For safety, 1345 * and to honor the TAILQ abstraction, I just restart the 1346 * loop after each removal. Note that the loop will exit 1347 * when all of the remaining ifaddrs belong to the AF_LINK 1348 * family. I am counting on the historical fact that at 1349 * least one pr_usrreq in each address domain removes at 1350 * least one ifaddr. 1351 */ 1352 again: 1353 /* 1354 * At this point, no other one tries to remove ifa in the list, 1355 * so we don't need to take a lock or psref. 1356 */ 1357 IFADDR_READER_FOREACH(ifa, ifp) { 1358 family = ifa->ifa_addr->sa_family; 1359 #ifdef IFAREF_DEBUG 1360 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1361 ifa, family, ifa->ifa_refcnt); 1362 if (last_ifa != NULL && ifa == last_ifa) 1363 panic("if_detach: loop detected"); 1364 last_ifa = ifa; 1365 #endif 1366 if (family == AF_LINK) 1367 continue; 1368 dp = pffinddomain(family); 1369 #ifdef DIAGNOSTIC 1370 if (dp == NULL) 1371 panic("if_detach: no domain for AF %d", 1372 family); 1373 #endif 1374 /* 1375 * XXX These PURGEIF calls are redundant with the 1376 * purge-all-families calls below, but are left in for 1377 * now both to make a smaller change, and to avoid 1378 * unplanned interactions with clearing of 1379 * ifp->if_addrlist. 1380 */ 1381 purged = 0; 1382 for (pr = dp->dom_protosw; 1383 pr < dp->dom_protoswNPROTOSW; pr++) { 1384 so.so_proto = pr; 1385 if (pr->pr_usrreqs) { 1386 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1387 purged = 1; 1388 } 1389 } 1390 if (purged == 0) { 1391 /* 1392 * XXX What's really the best thing to do 1393 * XXX here? --thorpej (at) NetBSD.org 1394 */ 1395 printf("if_detach: WARNING: AF %d not purged\n", 1396 family); 1397 ifa_remove(ifp, ifa); 1398 } 1399 goto again; 1400 } 1401 1402 if_free_sadl(ifp); 1403 1404 /* Delete stray routes from the routing table. */ 1405 for (i = 0; i <= AF_MAX; i++) 1406 rt_delete_matched_entries(i, if_delroute_matcher, ifp); 1407 1408 DOMAIN_FOREACH(dp) { 1409 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1410 { 1411 void *p = ifp->if_afdata[dp->dom_family]; 1412 if (p) { 1413 ifp->if_afdata[dp->dom_family] = NULL; 1414 (*dp->dom_ifdetach)(ifp, p); 1415 } 1416 } 1417 1418 /* 1419 * One would expect multicast memberships (INET and 1420 * INET6) on UDP sockets to be purged by the PURGEIF 1421 * calls above, but if all addresses were removed from 1422 * the interface prior to destruction, the calls will 1423 * not be made (e.g. ppp, for which pppd(8) generally 1424 * removes addresses before destroying the interface). 1425 * Because there is no invariant that multicast 1426 * memberships only exist for interfaces with IPv4 1427 * addresses, we must call PURGEIF regardless of 1428 * addresses. (Protocols which might store ifnet 1429 * pointers are marked with PR_PURGEIF.) 1430 */ 1431 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1432 so.so_proto = pr; 1433 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1434 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1435 } 1436 } 1437 1438 pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp); 1439 (void)pfil_head_destroy(ifp->if_pfil); 1440 1441 /* Announce that the interface is gone. */ 1442 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1443 1444 IF_AFDATA_LOCK_DESTROY(ifp); 1445 1446 if (if_is_link_state_changeable(ifp)) { 1447 softint_disestablish(ifp->if_link_si); 1448 ifp->if_link_si = NULL; 1449 } 1450 1451 /* 1452 * remove packets that came from ifp, from software interrupt queues. 1453 */ 1454 DOMAIN_FOREACH(dp) { 1455 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1456 struct ifqueue *iq = dp->dom_ifqueues[i]; 1457 if (iq == NULL) 1458 break; 1459 dp->dom_ifqueues[i] = NULL; 1460 if_detach_queues(ifp, iq); 1461 } 1462 } 1463 1464 /* 1465 * IP queues have to be processed separately: net-queue barrier 1466 * ensures that the packets are dequeued while a cross-call will 1467 * ensure that the interrupts have completed. FIXME: not quite.. 1468 */ 1469 #ifdef INET 1470 pktq_barrier(ip_pktq); 1471 #endif 1472 #ifdef INET6 1473 if (in6_present) 1474 pktq_barrier(ip6_pktq); 1475 #endif 1476 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1477 xc_wait(xc); 1478 1479 if (ifp->if_percpuq != NULL) { 1480 if_percpuq_destroy(ifp->if_percpuq); 1481 ifp->if_percpuq = NULL; 1482 } 1483 1484 splx(s); 1485 1486 #ifdef IFAREF_DEBUG 1487 if_check_and_free_ifa_list(ifp); 1488 #endif 1489 } 1490 1491 static void 1492 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1493 { 1494 struct mbuf *m, *prev, *next; 1495 1496 prev = NULL; 1497 for (m = q->ifq_head; m != NULL; m = next) { 1498 KASSERT((m->m_flags & M_PKTHDR) != 0); 1499 1500 next = m->m_nextpkt; 1501 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1502 prev = m; 1503 continue; 1504 } 1505 1506 if (prev != NULL) 1507 prev->m_nextpkt = m->m_nextpkt; 1508 else 1509 q->ifq_head = m->m_nextpkt; 1510 if (q->ifq_tail == m) 1511 q->ifq_tail = prev; 1512 q->ifq_len--; 1513 1514 m->m_nextpkt = NULL; 1515 m_freem(m); 1516 IF_DROP(q); 1517 } 1518 } 1519 1520 /* 1521 * Callback for a radix tree walk to delete all references to an 1522 * ifnet. 1523 */ 1524 static int 1525 if_delroute_matcher(struct rtentry *rt, void *v) 1526 { 1527 struct ifnet *ifp = (struct ifnet *)v; 1528 1529 if (rt->rt_ifp == ifp) 1530 return 1; 1531 else 1532 return 0; 1533 } 1534 1535 /* 1536 * Create a clone network interface. 1537 */ 1538 static int 1539 if_clone_create(const char *name) 1540 { 1541 struct if_clone *ifc; 1542 int unit; 1543 struct ifnet *ifp; 1544 struct psref psref; 1545 1546 KASSERT(mutex_owned(&if_clone_mtx)); 1547 1548 ifc = if_clone_lookup(name, &unit); 1549 if (ifc == NULL) 1550 return EINVAL; 1551 1552 ifp = if_get(name, &psref); 1553 if (ifp != NULL) { 1554 if_put(ifp, &psref); 1555 return EEXIST; 1556 } 1557 1558 return (*ifc->ifc_create)(ifc, unit); 1559 } 1560 1561 /* 1562 * Destroy a clone network interface. 1563 */ 1564 static int 1565 if_clone_destroy(const char *name) 1566 { 1567 struct if_clone *ifc; 1568 struct ifnet *ifp; 1569 struct psref psref; 1570 1571 KASSERT(mutex_owned(&if_clone_mtx)); 1572 1573 ifc = if_clone_lookup(name, NULL); 1574 if (ifc == NULL) 1575 return EINVAL; 1576 1577 if (ifc->ifc_destroy == NULL) 1578 return EOPNOTSUPP; 1579 1580 ifp = if_get(name, &psref); 1581 if (ifp == NULL) 1582 return ENXIO; 1583 1584 /* We have to disable ioctls here */ 1585 mutex_enter(ifp->if_ioctl_lock); 1586 ifp->if_ioctl = if_nullioctl; 1587 mutex_exit(ifp->if_ioctl_lock); 1588 1589 /* 1590 * We cannot call ifc_destroy with holding ifp. 1591 * Releasing ifp here is safe thanks to if_clone_mtx. 1592 */ 1593 if_put(ifp, &psref); 1594 1595 return (*ifc->ifc_destroy)(ifp); 1596 } 1597 1598 static bool 1599 if_is_unit(const char *name) 1600 { 1601 1602 while(*name != '\0') { 1603 if (*name < '0' || *name > '9') 1604 return false; 1605 name++; 1606 } 1607 1608 return true; 1609 } 1610 1611 /* 1612 * Look up a network interface cloner. 1613 */ 1614 static struct if_clone * 1615 if_clone_lookup(const char *name, int *unitp) 1616 { 1617 struct if_clone *ifc; 1618 const char *cp; 1619 char *dp, ifname[IFNAMSIZ + 3]; 1620 int unit; 1621 1622 KASSERT(mutex_owned(&if_clone_mtx)); 1623 1624 strcpy(ifname, "if_"); 1625 /* separate interface name from unit */ 1626 /* TODO: search unit number from backward */ 1627 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1628 *cp && !if_is_unit(cp);) 1629 *dp++ = *cp++; 1630 1631 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1632 return NULL; /* No name or unit number */ 1633 *dp++ = '\0'; 1634 1635 again: 1636 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1637 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1638 break; 1639 } 1640 1641 if (ifc == NULL) { 1642 int error; 1643 if (*ifname == '\0') 1644 return NULL; 1645 mutex_exit(&if_clone_mtx); 1646 error = module_autoload(ifname, MODULE_CLASS_DRIVER); 1647 mutex_enter(&if_clone_mtx); 1648 if (error) 1649 return NULL; 1650 *ifname = '\0'; 1651 goto again; 1652 } 1653 1654 unit = 0; 1655 while (cp - name < IFNAMSIZ && *cp) { 1656 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1657 /* Bogus unit number. */ 1658 return NULL; 1659 } 1660 unit = (unit * 10) + (*cp++ - '0'); 1661 } 1662 1663 if (unitp != NULL) 1664 *unitp = unit; 1665 return ifc; 1666 } 1667 1668 /* 1669 * Register a network interface cloner. 1670 */ 1671 void 1672 if_clone_attach(struct if_clone *ifc) 1673 { 1674 1675 mutex_enter(&if_clone_mtx); 1676 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1677 if_cloners_count++; 1678 mutex_exit(&if_clone_mtx); 1679 } 1680 1681 /* 1682 * Unregister a network interface cloner. 1683 */ 1684 void 1685 if_clone_detach(struct if_clone *ifc) 1686 { 1687 1688 mutex_enter(&if_clone_mtx); 1689 LIST_REMOVE(ifc, ifc_list); 1690 if_cloners_count--; 1691 mutex_exit(&if_clone_mtx); 1692 } 1693 1694 /* 1695 * Provide list of interface cloners to userspace. 1696 */ 1697 int 1698 if_clone_list(int buf_count, char *buffer, int *total) 1699 { 1700 char outbuf[IFNAMSIZ], *dst; 1701 struct if_clone *ifc; 1702 int count, error = 0; 1703 1704 mutex_enter(&if_clone_mtx); 1705 *total = if_cloners_count; 1706 if ((dst = buffer) == NULL) { 1707 /* Just asking how many there are. */ 1708 goto out; 1709 } 1710 1711 if (buf_count < 0) { 1712 error = EINVAL; 1713 goto out; 1714 } 1715 1716 count = (if_cloners_count < buf_count) ? 1717 if_cloners_count : buf_count; 1718 1719 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1720 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1721 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1722 if (outbuf[sizeof(outbuf) - 1] != '\0') { 1723 error = ENAMETOOLONG; 1724 goto out; 1725 } 1726 error = copyout(outbuf, dst, sizeof(outbuf)); 1727 if (error != 0) 1728 break; 1729 } 1730 1731 out: 1732 mutex_exit(&if_clone_mtx); 1733 return error; 1734 } 1735 1736 void 1737 ifa_psref_init(struct ifaddr *ifa) 1738 { 1739 1740 psref_target_init(&ifa->ifa_psref, ifa_psref_class); 1741 } 1742 1743 void 1744 ifaref(struct ifaddr *ifa) 1745 { 1746 KASSERT(!ISSET(ifa->ifa_flags, IFA_DESTROYING)); 1747 ifa->ifa_refcnt++; 1748 } 1749 1750 void 1751 ifafree(struct ifaddr *ifa) 1752 { 1753 KASSERT(ifa != NULL); 1754 KASSERT(ifa->ifa_refcnt > 0); 1755 1756 if (--ifa->ifa_refcnt == 0) { 1757 free(ifa, M_IFADDR); 1758 } 1759 } 1760 1761 bool 1762 ifa_is_destroying(struct ifaddr *ifa) 1763 { 1764 1765 return ISSET(ifa->ifa_flags, IFA_DESTROYING); 1766 } 1767 1768 void 1769 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1770 { 1771 1772 ifa->ifa_ifp = ifp; 1773 1774 IFNET_LOCK(); 1775 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1776 IFADDR_ENTRY_INIT(ifa); 1777 IFADDR_WRITER_INSERT_TAIL(ifp, ifa); 1778 IFNET_UNLOCK(); 1779 1780 ifaref(ifa); 1781 } 1782 1783 void 1784 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1785 { 1786 1787 KASSERT(ifa->ifa_ifp == ifp); 1788 1789 IFNET_LOCK(); 1790 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1791 IFADDR_WRITER_REMOVE(ifa); 1792 #ifdef NET_MPSAFE 1793 pserialize_perform(ifnet_psz); 1794 #endif 1795 IFNET_UNLOCK(); 1796 1797 #ifdef NET_MPSAFE 1798 psref_target_destroy(&ifa->ifa_psref, ifa_psref_class); 1799 #endif 1800 IFADDR_ENTRY_DESTROY(ifa); 1801 ifafree(ifa); 1802 } 1803 1804 void 1805 ifa_acquire(struct ifaddr *ifa, struct psref *psref) 1806 { 1807 1808 psref_acquire(psref, &ifa->ifa_psref, ifa_psref_class); 1809 } 1810 1811 void 1812 ifa_release(struct ifaddr *ifa, struct psref *psref) 1813 { 1814 1815 if (ifa == NULL) 1816 return; 1817 1818 psref_release(psref, &ifa->ifa_psref, ifa_psref_class); 1819 } 1820 1821 bool 1822 ifa_held(struct ifaddr *ifa) 1823 { 1824 1825 return psref_held(&ifa->ifa_psref, ifa_psref_class); 1826 } 1827 1828 static inline int 1829 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1830 { 1831 return sockaddr_cmp(sa1, sa2) == 0; 1832 } 1833 1834 /* 1835 * Locate an interface based on a complete address. 1836 */ 1837 /*ARGSUSED*/ 1838 struct ifaddr * 1839 ifa_ifwithaddr(const struct sockaddr *addr) 1840 { 1841 struct ifnet *ifp; 1842 struct ifaddr *ifa; 1843 1844 IFNET_READER_FOREACH(ifp) { 1845 if (if_is_deactivated(ifp)) 1846 continue; 1847 IFADDR_READER_FOREACH(ifa, ifp) { 1848 if (ifa->ifa_addr->sa_family != addr->sa_family) 1849 continue; 1850 if (equal(addr, ifa->ifa_addr)) 1851 return ifa; 1852 if ((ifp->if_flags & IFF_BROADCAST) && 1853 ifa->ifa_broadaddr && 1854 /* IP6 doesn't have broadcast */ 1855 ifa->ifa_broadaddr->sa_len != 0 && 1856 equal(ifa->ifa_broadaddr, addr)) 1857 return ifa; 1858 } 1859 } 1860 return NULL; 1861 } 1862 1863 struct ifaddr * 1864 ifa_ifwithaddr_psref(const struct sockaddr *addr, struct psref *psref) 1865 { 1866 struct ifaddr *ifa; 1867 int s = pserialize_read_enter(); 1868 1869 ifa = ifa_ifwithaddr(addr); 1870 if (ifa != NULL) 1871 ifa_acquire(ifa, psref); 1872 pserialize_read_exit(s); 1873 1874 return ifa; 1875 } 1876 1877 /* 1878 * Locate the point to point interface with a given destination address. 1879 */ 1880 /*ARGSUSED*/ 1881 struct ifaddr * 1882 ifa_ifwithdstaddr(const struct sockaddr *addr) 1883 { 1884 struct ifnet *ifp; 1885 struct ifaddr *ifa; 1886 1887 IFNET_READER_FOREACH(ifp) { 1888 if (if_is_deactivated(ifp)) 1889 continue; 1890 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1891 continue; 1892 IFADDR_READER_FOREACH(ifa, ifp) { 1893 if (ifa->ifa_addr->sa_family != addr->sa_family || 1894 ifa->ifa_dstaddr == NULL) 1895 continue; 1896 if (equal(addr, ifa->ifa_dstaddr)) 1897 return ifa; 1898 } 1899 } 1900 1901 return NULL; 1902 } 1903 1904 struct ifaddr * 1905 ifa_ifwithdstaddr_psref(const struct sockaddr *addr, struct psref *psref) 1906 { 1907 struct ifaddr *ifa; 1908 int s; 1909 1910 s = pserialize_read_enter(); 1911 ifa = ifa_ifwithdstaddr(addr); 1912 if (ifa != NULL) 1913 ifa_acquire(ifa, psref); 1914 pserialize_read_exit(s); 1915 1916 return ifa; 1917 } 1918 1919 /* 1920 * Find an interface on a specific network. If many, choice 1921 * is most specific found. 1922 */ 1923 struct ifaddr * 1924 ifa_ifwithnet(const struct sockaddr *addr) 1925 { 1926 struct ifnet *ifp; 1927 struct ifaddr *ifa, *ifa_maybe = NULL; 1928 const struct sockaddr_dl *sdl; 1929 u_int af = addr->sa_family; 1930 const char *addr_data = addr->sa_data, *cplim; 1931 1932 if (af == AF_LINK) { 1933 sdl = satocsdl(addr); 1934 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1935 ifindex2ifnet[sdl->sdl_index] && 1936 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1937 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1938 } 1939 } 1940 #ifdef NETATALK 1941 if (af == AF_APPLETALK) { 1942 const struct sockaddr_at *sat, *sat2; 1943 sat = (const struct sockaddr_at *)addr; 1944 IFNET_READER_FOREACH(ifp) { 1945 if (if_is_deactivated(ifp)) 1946 continue; 1947 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1948 if (ifa == NULL) 1949 continue; 1950 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1951 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1952 return ifa; /* exact match */ 1953 if (ifa_maybe == NULL) { 1954 /* else keep the if with the right range */ 1955 ifa_maybe = ifa; 1956 } 1957 } 1958 return ifa_maybe; 1959 } 1960 #endif 1961 IFNET_READER_FOREACH(ifp) { 1962 if (if_is_deactivated(ifp)) 1963 continue; 1964 IFADDR_READER_FOREACH(ifa, ifp) { 1965 const char *cp, *cp2, *cp3; 1966 1967 if (ifa->ifa_addr->sa_family != af || 1968 ifa->ifa_netmask == NULL) 1969 next: continue; 1970 cp = addr_data; 1971 cp2 = ifa->ifa_addr->sa_data; 1972 cp3 = ifa->ifa_netmask->sa_data; 1973 cplim = (const char *)ifa->ifa_netmask + 1974 ifa->ifa_netmask->sa_len; 1975 while (cp3 < cplim) { 1976 if ((*cp++ ^ *cp2++) & *cp3++) { 1977 /* want to continue for() loop */ 1978 goto next; 1979 } 1980 } 1981 if (ifa_maybe == NULL || 1982 rt_refines(ifa->ifa_netmask, 1983 ifa_maybe->ifa_netmask)) 1984 ifa_maybe = ifa; 1985 } 1986 } 1987 return ifa_maybe; 1988 } 1989 1990 struct ifaddr * 1991 ifa_ifwithnet_psref(const struct sockaddr *addr, struct psref *psref) 1992 { 1993 struct ifaddr *ifa; 1994 int s; 1995 1996 s = pserialize_read_enter(); 1997 ifa = ifa_ifwithnet(addr); 1998 if (ifa != NULL) 1999 ifa_acquire(ifa, psref); 2000 pserialize_read_exit(s); 2001 2002 return ifa; 2003 } 2004 2005 /* 2006 * Find the interface of the addresss. 2007 */ 2008 struct ifaddr * 2009 ifa_ifwithladdr(const struct sockaddr *addr) 2010 { 2011 struct ifaddr *ia; 2012 2013 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 2014 (ia = ifa_ifwithnet(addr))) 2015 return ia; 2016 return NULL; 2017 } 2018 2019 struct ifaddr * 2020 ifa_ifwithladdr_psref(const struct sockaddr *addr, struct psref *psref) 2021 { 2022 struct ifaddr *ifa; 2023 int s; 2024 2025 s = pserialize_read_enter(); 2026 ifa = ifa_ifwithladdr(addr); 2027 if (ifa != NULL) 2028 ifa_acquire(ifa, psref); 2029 pserialize_read_exit(s); 2030 2031 return ifa; 2032 } 2033 2034 /* 2035 * Find an interface using a specific address family 2036 */ 2037 struct ifaddr * 2038 ifa_ifwithaf(int af) 2039 { 2040 struct ifnet *ifp; 2041 struct ifaddr *ifa = NULL; 2042 int s; 2043 2044 s = pserialize_read_enter(); 2045 IFNET_READER_FOREACH(ifp) { 2046 if (if_is_deactivated(ifp)) 2047 continue; 2048 IFADDR_READER_FOREACH(ifa, ifp) { 2049 if (ifa->ifa_addr->sa_family == af) 2050 goto out; 2051 } 2052 } 2053 out: 2054 pserialize_read_exit(s); 2055 return ifa; 2056 } 2057 2058 /* 2059 * Find an interface address specific to an interface best matching 2060 * a given address. 2061 */ 2062 struct ifaddr * 2063 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2064 { 2065 struct ifaddr *ifa; 2066 const char *cp, *cp2, *cp3; 2067 const char *cplim; 2068 struct ifaddr *ifa_maybe = 0; 2069 u_int af = addr->sa_family; 2070 2071 if (if_is_deactivated(ifp)) 2072 return NULL; 2073 2074 if (af >= AF_MAX) 2075 return NULL; 2076 2077 IFADDR_READER_FOREACH(ifa, ifp) { 2078 if (ifa->ifa_addr->sa_family != af) 2079 continue; 2080 ifa_maybe = ifa; 2081 if (ifa->ifa_netmask == NULL) { 2082 if (equal(addr, ifa->ifa_addr) || 2083 (ifa->ifa_dstaddr && 2084 equal(addr, ifa->ifa_dstaddr))) 2085 return ifa; 2086 continue; 2087 } 2088 cp = addr->sa_data; 2089 cp2 = ifa->ifa_addr->sa_data; 2090 cp3 = ifa->ifa_netmask->sa_data; 2091 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2092 for (; cp3 < cplim; cp3++) { 2093 if ((*cp++ ^ *cp2++) & *cp3) 2094 break; 2095 } 2096 if (cp3 == cplim) 2097 return ifa; 2098 } 2099 return ifa_maybe; 2100 } 2101 2102 struct ifaddr * 2103 ifaof_ifpforaddr_psref(const struct sockaddr *addr, struct ifnet *ifp, 2104 struct psref *psref) 2105 { 2106 struct ifaddr *ifa; 2107 int s; 2108 2109 s = pserialize_read_enter(); 2110 ifa = ifaof_ifpforaddr(addr, ifp); 2111 if (ifa != NULL) 2112 ifa_acquire(ifa, psref); 2113 pserialize_read_exit(s); 2114 2115 return ifa; 2116 } 2117 2118 /* 2119 * Default action when installing a route with a Link Level gateway. 2120 * Lookup an appropriate real ifa to point to. 2121 * This should be moved to /sys/net/link.c eventually. 2122 */ 2123 void 2124 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 2125 { 2126 struct ifaddr *ifa; 2127 const struct sockaddr *dst; 2128 struct ifnet *ifp; 2129 struct psref psref; 2130 2131 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 2132 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 2133 return; 2134 if ((ifa = ifaof_ifpforaddr_psref(dst, ifp, &psref)) != NULL) { 2135 rt_replace_ifa(rt, ifa); 2136 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2137 ifa->ifa_rtrequest(cmd, rt, info); 2138 ifa_release(ifa, &psref); 2139 } 2140 } 2141 2142 /* 2143 * bitmask macros to manage a densely packed link_state change queue. 2144 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 2145 * LINK_STATE_UP(2) we need 2 bits for each state change. 2146 * As a state change to store is 0, treat all bits set as an unset item. 2147 */ 2148 #define LQ_ITEM_BITS 2 2149 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 2150 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 2151 #define LINK_STATE_UNSET LQ_ITEM_MASK 2152 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 2153 #define LQ_STORE(q, i, v) \ 2154 do { \ 2155 (q) &= ~LQ_MASK((i)); \ 2156 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 2157 } while (0 /* CONSTCOND */) 2158 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 2159 #define LQ_POP(q, v) \ 2160 do { \ 2161 (v) = LQ_ITEM((q), 0); \ 2162 (q) >>= LQ_ITEM_BITS; \ 2163 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2164 } while (0 /* CONSTCOND */) 2165 #define LQ_PUSH(q, v) \ 2166 do { \ 2167 (q) >>= LQ_ITEM_BITS; \ 2168 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2169 } while (0 /* CONSTCOND */) 2170 #define LQ_FIND_UNSET(q, i) \ 2171 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 2172 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 2173 break; \ 2174 } 2175 /* 2176 * Handle a change in the interface link state and 2177 * queue notifications. 2178 */ 2179 void 2180 if_link_state_change(struct ifnet *ifp, int link_state) 2181 { 2182 int s, idx; 2183 2184 KASSERTMSG(if_is_link_state_changeable(ifp), 2185 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 2186 ifp->if_xname, ifp->if_extflags); 2187 2188 /* Ensure change is to a valid state */ 2189 switch (link_state) { 2190 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 2191 case LINK_STATE_DOWN: /* FALLTHROUGH */ 2192 case LINK_STATE_UP: 2193 break; 2194 default: 2195 #ifdef DEBUG 2196 printf("%s: invalid link state %d\n", 2197 ifp->if_xname, link_state); 2198 #endif 2199 return; 2200 } 2201 2202 s = splnet(); 2203 2204 /* Find the last unset event in the queue. */ 2205 LQ_FIND_UNSET(ifp->if_link_queue, idx); 2206 2207 /* 2208 * Ensure link_state doesn't match the last event in the queue. 2209 * ifp->if_link_state is not checked and set here because 2210 * that would present an inconsistent picture to the system. 2211 */ 2212 if (idx != 0 && 2213 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 2214 goto out; 2215 2216 /* Handle queue overflow. */ 2217 if (idx == LQ_MAX(ifp->if_link_queue)) { 2218 uint8_t lost; 2219 2220 /* 2221 * The DOWN state must be protected from being pushed off 2222 * the queue to ensure that userland will always be 2223 * in a sane state. 2224 * Because DOWN is protected, there is no need to protect 2225 * UNKNOWN. 2226 * It should be invalid to change from any other state to 2227 * UNKNOWN anyway ... 2228 */ 2229 lost = LQ_ITEM(ifp->if_link_queue, 0); 2230 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 2231 if (lost == LINK_STATE_DOWN) { 2232 lost = LQ_ITEM(ifp->if_link_queue, 0); 2233 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 2234 } 2235 printf("%s: lost link state change %s\n", 2236 ifp->if_xname, 2237 lost == LINK_STATE_UP ? "UP" : 2238 lost == LINK_STATE_DOWN ? "DOWN" : 2239 "UNKNOWN"); 2240 } else 2241 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 2242 2243 softint_schedule(ifp->if_link_si); 2244 2245 out: 2246 splx(s); 2247 } 2248 2249 /* 2250 * Handle interface link state change notifications. 2251 * Must be called at splnet(). 2252 */ 2253 static void 2254 if_link_state_change0(struct ifnet *ifp, int link_state) 2255 { 2256 struct domain *dp; 2257 2258 /* Ensure the change is still valid. */ 2259 if (ifp->if_link_state == link_state) 2260 return; 2261 2262 #ifdef DEBUG 2263 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 2264 link_state == LINK_STATE_UP ? "UP" : 2265 link_state == LINK_STATE_DOWN ? "DOWN" : 2266 "UNKNOWN", 2267 ifp->if_link_state == LINK_STATE_UP ? "UP" : 2268 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 2269 "UNKNOWN"); 2270 #endif 2271 2272 /* 2273 * When going from UNKNOWN to UP, we need to mark existing 2274 * addresses as tentative and restart DAD as we may have 2275 * erroneously not found a duplicate. 2276 * 2277 * This needs to happen before rt_ifmsg to avoid a race where 2278 * listeners would have an address and expect it to work right 2279 * away. 2280 */ 2281 if (link_state == LINK_STATE_UP && 2282 ifp->if_link_state == LINK_STATE_UNKNOWN) 2283 { 2284 DOMAIN_FOREACH(dp) { 2285 if (dp->dom_if_link_state_change != NULL) 2286 dp->dom_if_link_state_change(ifp, 2287 LINK_STATE_DOWN); 2288 } 2289 } 2290 2291 ifp->if_link_state = link_state; 2292 2293 /* Notify that the link state has changed. */ 2294 rt_ifmsg(ifp); 2295 2296 #if NCARP > 0 2297 if (ifp->if_carp) 2298 carp_carpdev_state(ifp); 2299 #endif 2300 2301 DOMAIN_FOREACH(dp) { 2302 if (dp->dom_if_link_state_change != NULL) 2303 dp->dom_if_link_state_change(ifp, link_state); 2304 } 2305 } 2306 2307 /* 2308 * Process the interface link state change queue. 2309 */ 2310 static void 2311 if_link_state_change_si(void *arg) 2312 { 2313 struct ifnet *ifp = arg; 2314 int s; 2315 uint8_t state; 2316 2317 #ifndef NET_MPSAFE 2318 mutex_enter(softnet_lock); 2319 KERNEL_LOCK(1, NULL); 2320 #endif 2321 s = splnet(); 2322 2323 /* Pop a link state change from the queue and process it. */ 2324 LQ_POP(ifp->if_link_queue, state); 2325 if_link_state_change0(ifp, state); 2326 2327 /* If there is a link state change to come, schedule it. */ 2328 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 2329 softint_schedule(ifp->if_link_si); 2330 2331 splx(s); 2332 #ifndef NET_MPSAFE 2333 KERNEL_UNLOCK_ONE(NULL); 2334 mutex_exit(softnet_lock); 2335 #endif 2336 } 2337 2338 /* 2339 * Default action when installing a local route on a point-to-point 2340 * interface. 2341 */ 2342 void 2343 p2p_rtrequest(int req, struct rtentry *rt, 2344 __unused const struct rt_addrinfo *info) 2345 { 2346 struct ifnet *ifp = rt->rt_ifp; 2347 struct ifaddr *ifa, *lo0ifa; 2348 int s = pserialize_read_enter(); 2349 2350 switch (req) { 2351 case RTM_ADD: 2352 if ((rt->rt_flags & RTF_LOCAL) == 0) 2353 break; 2354 2355 rt->rt_ifp = lo0ifp; 2356 2357 IFADDR_READER_FOREACH(ifa, ifp) { 2358 if (equal(rt_getkey(rt), ifa->ifa_addr)) 2359 break; 2360 } 2361 if (ifa == NULL) 2362 break; 2363 2364 /* 2365 * Ensure lo0 has an address of the same family. 2366 */ 2367 IFADDR_READER_FOREACH(lo0ifa, lo0ifp) { 2368 if (lo0ifa->ifa_addr->sa_family == 2369 ifa->ifa_addr->sa_family) 2370 break; 2371 } 2372 if (lo0ifa == NULL) 2373 break; 2374 2375 /* 2376 * Make sure to set rt->rt_ifa to the interface 2377 * address we are using, otherwise we will have trouble 2378 * with source address selection. 2379 */ 2380 if (ifa != rt->rt_ifa) 2381 rt_replace_ifa(rt, ifa); 2382 break; 2383 case RTM_DELETE: 2384 default: 2385 break; 2386 } 2387 pserialize_read_exit(s); 2388 } 2389 2390 /* 2391 * Mark an interface down and notify protocols of 2392 * the transition. 2393 * NOTE: must be called at splsoftnet or equivalent. 2394 */ 2395 void 2396 if_down(struct ifnet *ifp) 2397 { 2398 struct ifaddr *ifa; 2399 struct domain *dp; 2400 int s, bound; 2401 struct psref psref; 2402 2403 ifp->if_flags &= ~IFF_UP; 2404 nanotime(&ifp->if_lastchange); 2405 2406 bound = curlwp_bind(); 2407 s = pserialize_read_enter(); 2408 IFADDR_READER_FOREACH(ifa, ifp) { 2409 ifa_acquire(ifa, &psref); 2410 pserialize_read_exit(s); 2411 2412 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2413 2414 s = pserialize_read_enter(); 2415 ifa_release(ifa, &psref); 2416 } 2417 pserialize_read_exit(s); 2418 curlwp_bindx(bound); 2419 2420 IFQ_PURGE(&ifp->if_snd); 2421 #if NCARP > 0 2422 if (ifp->if_carp) 2423 carp_carpdev_state(ifp); 2424 #endif 2425 rt_ifmsg(ifp); 2426 DOMAIN_FOREACH(dp) { 2427 if (dp->dom_if_down) 2428 dp->dom_if_down(ifp); 2429 } 2430 } 2431 2432 /* 2433 * Mark an interface up and notify protocols of 2434 * the transition. 2435 * NOTE: must be called at splsoftnet or equivalent. 2436 */ 2437 void 2438 if_up(struct ifnet *ifp) 2439 { 2440 #ifdef notyet 2441 struct ifaddr *ifa; 2442 #endif 2443 struct domain *dp; 2444 2445 ifp->if_flags |= IFF_UP; 2446 nanotime(&ifp->if_lastchange); 2447 #ifdef notyet 2448 /* this has no effect on IP, and will kill all ISO connections XXX */ 2449 IFADDR_READER_FOREACH(ifa, ifp) 2450 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2451 #endif 2452 #if NCARP > 0 2453 if (ifp->if_carp) 2454 carp_carpdev_state(ifp); 2455 #endif 2456 rt_ifmsg(ifp); 2457 DOMAIN_FOREACH(dp) { 2458 if (dp->dom_if_up) 2459 dp->dom_if_up(ifp); 2460 } 2461 } 2462 2463 /* 2464 * Handle interface slowtimo timer routine. Called 2465 * from softclock, we decrement timer (if set) and 2466 * call the appropriate interface routine on expiration. 2467 */ 2468 static void 2469 if_slowtimo(void *arg) 2470 { 2471 void (*slowtimo)(struct ifnet *); 2472 struct ifnet *ifp = arg; 2473 int s; 2474 2475 slowtimo = ifp->if_slowtimo; 2476 if (__predict_false(slowtimo == NULL)) 2477 return; 2478 2479 s = splnet(); 2480 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2481 (*slowtimo)(ifp); 2482 2483 splx(s); 2484 2485 if (__predict_true(ifp->if_slowtimo != NULL)) 2486 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2487 } 2488 2489 /* 2490 * Set/clear promiscuous mode on interface ifp based on the truth value 2491 * of pswitch. The calls are reference counted so that only the first 2492 * "on" request actually has an effect, as does the final "off" request. 2493 * Results are undefined if the "off" and "on" requests are not matched. 2494 */ 2495 int 2496 ifpromisc(struct ifnet *ifp, int pswitch) 2497 { 2498 int pcount, ret; 2499 short nflags; 2500 2501 pcount = ifp->if_pcount; 2502 if (pswitch) { 2503 /* 2504 * Allow the device to be "placed" into promiscuous 2505 * mode even if it is not configured up. It will 2506 * consult IFF_PROMISC when it is brought up. 2507 */ 2508 if (ifp->if_pcount++ != 0) 2509 return 0; 2510 nflags = ifp->if_flags | IFF_PROMISC; 2511 } else { 2512 if (--ifp->if_pcount > 0) 2513 return 0; 2514 nflags = ifp->if_flags & ~IFF_PROMISC; 2515 } 2516 ret = if_flags_set(ifp, nflags); 2517 /* Restore interface state if not successful. */ 2518 if (ret != 0) { 2519 ifp->if_pcount = pcount; 2520 } 2521 return ret; 2522 } 2523 2524 /* 2525 * Map interface name to 2526 * interface structure pointer. 2527 */ 2528 struct ifnet * 2529 ifunit(const char *name) 2530 { 2531 struct ifnet *ifp; 2532 const char *cp = name; 2533 u_int unit = 0; 2534 u_int i; 2535 int s; 2536 2537 /* 2538 * If the entire name is a number, treat it as an ifindex. 2539 */ 2540 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2541 unit = unit * 10 + (*cp - '0'); 2542 } 2543 2544 /* 2545 * If the number took all of the name, then it's a valid ifindex. 2546 */ 2547 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2548 if (unit >= if_indexlim) 2549 return NULL; 2550 ifp = ifindex2ifnet[unit]; 2551 if (ifp == NULL || if_is_deactivated(ifp)) 2552 return NULL; 2553 return ifp; 2554 } 2555 2556 ifp = NULL; 2557 s = pserialize_read_enter(); 2558 IFNET_READER_FOREACH(ifp) { 2559 if (if_is_deactivated(ifp)) 2560 continue; 2561 if (strcmp(ifp->if_xname, name) == 0) 2562 goto out; 2563 } 2564 out: 2565 pserialize_read_exit(s); 2566 return ifp; 2567 } 2568 2569 /* 2570 * Get a reference of an ifnet object by an interface name. 2571 * The returned reference is protected by psref(9). The caller 2572 * must release a returned reference by if_put after use. 2573 */ 2574 struct ifnet * 2575 if_get(const char *name, struct psref *psref) 2576 { 2577 struct ifnet *ifp; 2578 const char *cp = name; 2579 u_int unit = 0; 2580 u_int i; 2581 int s; 2582 2583 /* 2584 * If the entire name is a number, treat it as an ifindex. 2585 */ 2586 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2587 unit = unit * 10 + (*cp - '0'); 2588 } 2589 2590 /* 2591 * If the number took all of the name, then it's a valid ifindex. 2592 */ 2593 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2594 if (unit >= if_indexlim) 2595 return NULL; 2596 ifp = ifindex2ifnet[unit]; 2597 if (ifp == NULL || if_is_deactivated(ifp)) 2598 return NULL; 2599 return ifp; 2600 } 2601 2602 ifp = NULL; 2603 s = pserialize_read_enter(); 2604 IFNET_READER_FOREACH(ifp) { 2605 if (if_is_deactivated(ifp)) 2606 continue; 2607 if (strcmp(ifp->if_xname, name) == 0) { 2608 psref_acquire(psref, &ifp->if_psref, 2609 ifnet_psref_class); 2610 goto out; 2611 } 2612 } 2613 out: 2614 pserialize_read_exit(s); 2615 return ifp; 2616 } 2617 2618 /* 2619 * Release a reference of an ifnet object given by if_get or 2620 * if_get_byindex. 2621 */ 2622 void 2623 if_put(const struct ifnet *ifp, struct psref *psref) 2624 { 2625 2626 if (ifp == NULL) 2627 return; 2628 2629 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2630 } 2631 2632 ifnet_t * 2633 if_byindex(u_int idx) 2634 { 2635 ifnet_t *ifp; 2636 2637 ifp = (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2638 if (ifp != NULL && if_is_deactivated(ifp)) 2639 ifp = NULL; 2640 return ifp; 2641 } 2642 2643 /* 2644 * Get a reference of an ifnet object by an interface index. 2645 * The returned reference is protected by psref(9). The caller 2646 * must release a returned reference by if_put after use. 2647 */ 2648 ifnet_t * 2649 if_get_byindex(u_int idx, struct psref *psref) 2650 { 2651 ifnet_t *ifp; 2652 int s; 2653 2654 s = pserialize_read_enter(); 2655 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2656 if (ifp != NULL && if_is_deactivated(ifp)) 2657 ifp = NULL; 2658 if (__predict_true(ifp != NULL)) 2659 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2660 pserialize_read_exit(s); 2661 2662 return ifp; 2663 } 2664 2665 /* 2666 * XXX it's safe only if the passed ifp is guaranteed to not be freed, 2667 * for example the ifp is already held or some other object is held which 2668 * guarantes the ifp to not be freed indirectly. 2669 */ 2670 void 2671 if_acquire_NOMPSAFE(struct ifnet *ifp, struct psref *psref) 2672 { 2673 2674 KASSERT(ifp->if_index != 0); 2675 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2676 } 2677 2678 bool 2679 if_held(struct ifnet *ifp) 2680 { 2681 2682 return psref_held(&ifp->if_psref, ifnet_psref_class); 2683 } 2684 2685 2686 /* common */ 2687 int 2688 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2689 { 2690 int s; 2691 struct ifreq *ifr; 2692 struct ifcapreq *ifcr; 2693 struct ifdatareq *ifdr; 2694 2695 switch (cmd) { 2696 case SIOCSIFCAP: 2697 ifcr = data; 2698 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2699 return EINVAL; 2700 2701 if (ifcr->ifcr_capenable == ifp->if_capenable) 2702 return 0; 2703 2704 ifp->if_capenable = ifcr->ifcr_capenable; 2705 2706 /* Pre-compute the checksum flags mask. */ 2707 ifp->if_csum_flags_tx = 0; 2708 ifp->if_csum_flags_rx = 0; 2709 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2710 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2711 } 2712 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2713 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2714 } 2715 2716 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2717 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2718 } 2719 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2720 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2721 } 2722 2723 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2724 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2725 } 2726 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2727 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2728 } 2729 2730 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2731 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2732 } 2733 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2734 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2735 } 2736 2737 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2738 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2739 } 2740 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2741 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2742 } 2743 if (ifp->if_flags & IFF_UP) 2744 return ENETRESET; 2745 return 0; 2746 case SIOCSIFFLAGS: 2747 ifr = data; 2748 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2749 s = splsoftnet(); 2750 if_down(ifp); 2751 splx(s); 2752 } 2753 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2754 s = splsoftnet(); 2755 if_up(ifp); 2756 splx(s); 2757 } 2758 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2759 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2760 break; 2761 case SIOCGIFFLAGS: 2762 ifr = data; 2763 ifr->ifr_flags = ifp->if_flags; 2764 break; 2765 2766 case SIOCGIFMETRIC: 2767 ifr = data; 2768 ifr->ifr_metric = ifp->if_metric; 2769 break; 2770 2771 case SIOCGIFMTU: 2772 ifr = data; 2773 ifr->ifr_mtu = ifp->if_mtu; 2774 break; 2775 2776 case SIOCGIFDLT: 2777 ifr = data; 2778 ifr->ifr_dlt = ifp->if_dlt; 2779 break; 2780 2781 case SIOCGIFCAP: 2782 ifcr = data; 2783 ifcr->ifcr_capabilities = ifp->if_capabilities; 2784 ifcr->ifcr_capenable = ifp->if_capenable; 2785 break; 2786 2787 case SIOCSIFMETRIC: 2788 ifr = data; 2789 ifp->if_metric = ifr->ifr_metric; 2790 break; 2791 2792 case SIOCGIFDATA: 2793 ifdr = data; 2794 ifdr->ifdr_data = ifp->if_data; 2795 break; 2796 2797 case SIOCGIFINDEX: 2798 ifr = data; 2799 ifr->ifr_index = ifp->if_index; 2800 break; 2801 2802 case SIOCZIFDATA: 2803 ifdr = data; 2804 ifdr->ifdr_data = ifp->if_data; 2805 /* 2806 * Assumes that the volatile counters that can be 2807 * zero'ed are at the end of if_data. 2808 */ 2809 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2810 offsetof(struct if_data, ifi_ipackets)); 2811 /* 2812 * The memset() clears to the bottm of if_data. In the area, 2813 * if_lastchange is included. Please be careful if new entry 2814 * will be added into if_data or rewite this. 2815 * 2816 * And also, update if_lastchnage. 2817 */ 2818 getnanotime(&ifp->if_lastchange); 2819 break; 2820 case SIOCSIFMTU: 2821 ifr = data; 2822 if (ifp->if_mtu == ifr->ifr_mtu) 2823 break; 2824 ifp->if_mtu = ifr->ifr_mtu; 2825 /* 2826 * If the link MTU changed, do network layer specific procedure. 2827 */ 2828 #ifdef INET6 2829 if (in6_present) 2830 nd6_setmtu(ifp); 2831 #endif 2832 return ENETRESET; 2833 default: 2834 return ENOTTY; 2835 } 2836 return 0; 2837 } 2838 2839 int 2840 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2841 { 2842 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2843 struct ifaddr *ifa; 2844 const struct sockaddr *any, *sa; 2845 union { 2846 struct sockaddr sa; 2847 struct sockaddr_storage ss; 2848 } u, v; 2849 int s, error = 0; 2850 2851 switch (cmd) { 2852 case SIOCSIFADDRPREF: 2853 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2854 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2855 NULL) != 0) 2856 return EPERM; 2857 case SIOCGIFADDRPREF: 2858 break; 2859 default: 2860 return EOPNOTSUPP; 2861 } 2862 2863 /* sanity checks */ 2864 if (data == NULL || ifp == NULL) { 2865 panic("invalid argument to %s", __func__); 2866 /*NOTREACHED*/ 2867 } 2868 2869 /* address must be specified on ADD and DELETE */ 2870 sa = sstocsa(&ifap->ifap_addr); 2871 if (sa->sa_family != sofamily(so)) 2872 return EINVAL; 2873 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2874 return EINVAL; 2875 2876 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2877 2878 s = pserialize_read_enter(); 2879 IFADDR_READER_FOREACH(ifa, ifp) { 2880 if (ifa->ifa_addr->sa_family != sa->sa_family) 2881 continue; 2882 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2883 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2884 break; 2885 } 2886 if (ifa == NULL) { 2887 error = EADDRNOTAVAIL; 2888 goto out; 2889 } 2890 2891 switch (cmd) { 2892 case SIOCSIFADDRPREF: 2893 ifa->ifa_preference = ifap->ifap_preference; 2894 goto out; 2895 case SIOCGIFADDRPREF: 2896 /* fill in the if_laddrreq structure */ 2897 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2898 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2899 ifap->ifap_preference = ifa->ifa_preference; 2900 goto out; 2901 default: 2902 error = EOPNOTSUPP; 2903 } 2904 out: 2905 pserialize_read_exit(s); 2906 return error; 2907 } 2908 2909 /* 2910 * Interface ioctls. 2911 */ 2912 static int 2913 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2914 { 2915 struct ifnet *ifp; 2916 struct ifreq *ifr; 2917 int error = 0; 2918 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2919 u_long ocmd = cmd; 2920 #endif 2921 short oif_flags; 2922 #ifdef COMPAT_OIFREQ 2923 struct ifreq ifrb; 2924 struct oifreq *oifr = NULL; 2925 #endif 2926 int r; 2927 struct psref psref; 2928 int bound; 2929 2930 switch (cmd) { 2931 #ifdef COMPAT_OIFREQ 2932 case OSIOCGIFCONF: 2933 case OOSIOCGIFCONF: 2934 return compat_ifconf(cmd, data); 2935 #endif 2936 #ifdef COMPAT_OIFDATA 2937 case OSIOCGIFDATA: 2938 case OSIOCZIFDATA: 2939 return compat_ifdatareq(l, cmd, data); 2940 #endif 2941 case SIOCGIFCONF: 2942 return ifconf(cmd, data); 2943 case SIOCINITIFADDR: 2944 return EPERM; 2945 } 2946 2947 #ifdef COMPAT_OIFREQ 2948 cmd = (*vec_compat_cvtcmd)(cmd); 2949 if (cmd != ocmd) { 2950 oifr = data; 2951 data = ifr = &ifrb; 2952 ifreqo2n(oifr, ifr); 2953 } else 2954 #endif 2955 ifr = data; 2956 2957 switch (cmd) { 2958 case SIOCIFCREATE: 2959 case SIOCIFDESTROY: 2960 bound = curlwp_bind(); 2961 if (l != NULL) { 2962 ifp = if_get(ifr->ifr_name, &psref); 2963 error = kauth_authorize_network(l->l_cred, 2964 KAUTH_NETWORK_INTERFACE, 2965 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2966 (void *)cmd, NULL); 2967 if (ifp != NULL) 2968 if_put(ifp, &psref); 2969 if (error != 0) { 2970 curlwp_bindx(bound); 2971 return error; 2972 } 2973 } 2974 mutex_enter(&if_clone_mtx); 2975 r = (cmd == SIOCIFCREATE) ? 2976 if_clone_create(ifr->ifr_name) : 2977 if_clone_destroy(ifr->ifr_name); 2978 mutex_exit(&if_clone_mtx); 2979 curlwp_bindx(bound); 2980 return r; 2981 2982 case SIOCIFGCLONERS: 2983 { 2984 struct if_clonereq *req = (struct if_clonereq *)data; 2985 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2986 &req->ifcr_total); 2987 } 2988 } 2989 2990 bound = curlwp_bind(); 2991 ifp = if_get(ifr->ifr_name, &psref); 2992 if (ifp == NULL) { 2993 curlwp_bindx(bound); 2994 return ENXIO; 2995 } 2996 2997 switch (cmd) { 2998 case SIOCALIFADDR: 2999 case SIOCDLIFADDR: 3000 case SIOCSIFADDRPREF: 3001 case SIOCSIFFLAGS: 3002 case SIOCSIFCAP: 3003 case SIOCSIFMETRIC: 3004 case SIOCZIFDATA: 3005 case SIOCSIFMTU: 3006 case SIOCSIFPHYADDR: 3007 case SIOCDIFPHYADDR: 3008 #ifdef INET6 3009 case SIOCSIFPHYADDR_IN6: 3010 #endif 3011 case SIOCSLIFPHYADDR: 3012 case SIOCADDMULTI: 3013 case SIOCDELMULTI: 3014 case SIOCSIFMEDIA: 3015 case SIOCSDRVSPEC: 3016 case SIOCG80211: 3017 case SIOCS80211: 3018 case SIOCS80211NWID: 3019 case SIOCS80211NWKEY: 3020 case SIOCS80211POWER: 3021 case SIOCS80211BSSID: 3022 case SIOCS80211CHANNEL: 3023 case SIOCSLINKSTR: 3024 if (l != NULL) { 3025 error = kauth_authorize_network(l->l_cred, 3026 KAUTH_NETWORK_INTERFACE, 3027 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 3028 (void *)cmd, NULL); 3029 if (error != 0) 3030 goto out; 3031 } 3032 } 3033 3034 oif_flags = ifp->if_flags; 3035 3036 mutex_enter(ifp->if_ioctl_lock); 3037 3038 error = (*ifp->if_ioctl)(ifp, cmd, data); 3039 if (error != ENOTTY) 3040 ; 3041 else if (so->so_proto == NULL) 3042 error = EOPNOTSUPP; 3043 else { 3044 #ifdef COMPAT_OSOCK 3045 if (vec_compat_ifioctl != NULL) 3046 error = (*vec_compat_ifioctl)(so, ocmd, cmd, data, l); 3047 else 3048 #endif 3049 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 3050 cmd, data, ifp); 3051 } 3052 3053 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 3054 if ((ifp->if_flags & IFF_UP) != 0) { 3055 int s = splsoftnet(); 3056 if_up(ifp); 3057 splx(s); 3058 } 3059 } 3060 #ifdef COMPAT_OIFREQ 3061 if (cmd != ocmd) 3062 ifreqn2o(oifr, ifr); 3063 #endif 3064 3065 mutex_exit(ifp->if_ioctl_lock); 3066 out: 3067 if_put(ifp, &psref); 3068 curlwp_bindx(bound); 3069 return error; 3070 } 3071 3072 /* 3073 * Return interface configuration 3074 * of system. List may be used 3075 * in later ioctl's (above) to get 3076 * other information. 3077 * 3078 * Each record is a struct ifreq. Before the addition of 3079 * sockaddr_storage, the API rule was that sockaddr flavors that did 3080 * not fit would extend beyond the struct ifreq, with the next struct 3081 * ifreq starting sa_len beyond the struct sockaddr. Because the 3082 * union in struct ifreq includes struct sockaddr_storage, every kind 3083 * of sockaddr must fit. Thus, there are no longer any overlength 3084 * records. 3085 * 3086 * Records are added to the user buffer if they fit, and ifc_len is 3087 * adjusted to the length that was written. Thus, the user is only 3088 * assured of getting the complete list if ifc_len on return is at 3089 * least sizeof(struct ifreq) less than it was on entry. 3090 * 3091 * If the user buffer pointer is NULL, this routine copies no data and 3092 * returns the amount of space that would be needed. 3093 * 3094 * Invariants: 3095 * ifrp points to the next part of the user's buffer to be used. If 3096 * ifrp != NULL, space holds the number of bytes remaining that we may 3097 * write at ifrp. Otherwise, space holds the number of bytes that 3098 * would have been written had there been adequate space. 3099 */ 3100 /*ARGSUSED*/ 3101 static int 3102 ifconf(u_long cmd, void *data) 3103 { 3104 struct ifconf *ifc = (struct ifconf *)data; 3105 struct ifnet *ifp; 3106 struct ifaddr *ifa; 3107 struct ifreq ifr, *ifrp = NULL; 3108 int space = 0, error = 0; 3109 const int sz = (int)sizeof(struct ifreq); 3110 const bool docopy = ifc->ifc_req != NULL; 3111 int s; 3112 int bound; 3113 struct psref psref; 3114 3115 if (docopy) { 3116 space = ifc->ifc_len; 3117 ifrp = ifc->ifc_req; 3118 } 3119 3120 bound = curlwp_bind(); 3121 s = pserialize_read_enter(); 3122 IFNET_READER_FOREACH(ifp) { 3123 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 3124 pserialize_read_exit(s); 3125 3126 (void)strncpy(ifr.ifr_name, ifp->if_xname, 3127 sizeof(ifr.ifr_name)); 3128 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 3129 error = ENAMETOOLONG; 3130 goto release_exit; 3131 } 3132 if (IFADDR_READER_EMPTY(ifp)) { 3133 /* Interface with no addresses - send zero sockaddr. */ 3134 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 3135 if (!docopy) { 3136 space += sz; 3137 continue; 3138 } 3139 if (space >= sz) { 3140 error = copyout(&ifr, ifrp, sz); 3141 if (error != 0) 3142 goto release_exit; 3143 ifrp++; 3144 space -= sz; 3145 } 3146 } 3147 3148 IFADDR_READER_FOREACH(ifa, ifp) { 3149 struct sockaddr *sa = ifa->ifa_addr; 3150 /* all sockaddrs must fit in sockaddr_storage */ 3151 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 3152 3153 if (!docopy) { 3154 space += sz; 3155 continue; 3156 } 3157 memcpy(&ifr.ifr_space, sa, sa->sa_len); 3158 if (space >= sz) { 3159 error = copyout(&ifr, ifrp, sz); 3160 if (error != 0) 3161 goto release_exit; 3162 ifrp++; space -= sz; 3163 } 3164 } 3165 3166 s = pserialize_read_enter(); 3167 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3168 } 3169 pserialize_read_exit(s); 3170 curlwp_bindx(bound); 3171 3172 if (docopy) { 3173 KASSERT(0 <= space && space <= ifc->ifc_len); 3174 ifc->ifc_len -= space; 3175 } else { 3176 KASSERT(space >= 0); 3177 ifc->ifc_len = space; 3178 } 3179 return (0); 3180 3181 release_exit: 3182 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3183 curlwp_bindx(bound); 3184 return error; 3185 } 3186 3187 int 3188 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 3189 { 3190 uint8_t len; 3191 #ifdef COMPAT_OIFREQ 3192 struct ifreq ifrb; 3193 struct oifreq *oifr = NULL; 3194 u_long ocmd = cmd; 3195 cmd = (*vec_compat_cvtcmd)(cmd); 3196 if (cmd != ocmd) { 3197 oifr = (struct oifreq *)(void *)ifr; 3198 ifr = &ifrb; 3199 ifreqo2n(oifr, ifr); 3200 len = sizeof(oifr->ifr_addr); 3201 } else 3202 #endif 3203 len = sizeof(ifr->ifr_ifru.ifru_space); 3204 3205 if (len < sa->sa_len) 3206 return EFBIG; 3207 3208 memset(&ifr->ifr_addr, 0, len); 3209 sockaddr_copy(&ifr->ifr_addr, len, sa); 3210 3211 #ifdef COMPAT_OIFREQ 3212 if (cmd != ocmd) 3213 ifreqn2o(oifr, ifr); 3214 #endif 3215 return 0; 3216 } 3217 3218 /* 3219 * wrapper function for the drivers which doesn't have if_transmit(). 3220 */ 3221 static int 3222 if_transmit(struct ifnet *ifp, struct mbuf *m) 3223 { 3224 int s, error; 3225 3226 s = splnet(); 3227 3228 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3229 if (error != 0) { 3230 /* mbuf is already freed */ 3231 goto out; 3232 } 3233 3234 ifp->if_obytes += m->m_pkthdr.len;; 3235 if (m->m_flags & M_MCAST) 3236 ifp->if_omcasts++; 3237 3238 if ((ifp->if_flags & IFF_OACTIVE) == 0) 3239 if_start_lock(ifp); 3240 out: 3241 splx(s); 3242 3243 return error; 3244 } 3245 3246 int 3247 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 3248 { 3249 int error; 3250 3251 #ifdef ALTQ 3252 KERNEL_LOCK(1, NULL); 3253 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 3254 error = if_transmit(ifp, m); 3255 KERNEL_UNLOCK_ONE(NULL); 3256 } else { 3257 KERNEL_UNLOCK_ONE(NULL); 3258 error = (*ifp->if_transmit)(ifp, m); 3259 } 3260 #else /* !ALTQ */ 3261 error = (*ifp->if_transmit)(ifp, m); 3262 #endif /* !ALTQ */ 3263 3264 return error; 3265 } 3266 3267 /* 3268 * Queue message on interface, and start output if interface 3269 * not yet active. 3270 */ 3271 int 3272 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 3273 { 3274 3275 return if_transmit_lock(ifp, m); 3276 } 3277 3278 /* 3279 * Queue message on interface, possibly using a second fast queue 3280 */ 3281 int 3282 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 3283 { 3284 int error = 0; 3285 3286 if (ifq != NULL 3287 #ifdef ALTQ 3288 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 3289 #endif 3290 ) { 3291 if (IF_QFULL(ifq)) { 3292 IF_DROP(&ifp->if_snd); 3293 m_freem(m); 3294 if (error == 0) 3295 error = ENOBUFS; 3296 } else 3297 IF_ENQUEUE(ifq, m); 3298 } else 3299 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3300 if (error != 0) { 3301 ++ifp->if_oerrors; 3302 return error; 3303 } 3304 return 0; 3305 } 3306 3307 int 3308 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 3309 { 3310 int rc; 3311 3312 if (ifp->if_initaddr != NULL) 3313 rc = (*ifp->if_initaddr)(ifp, ifa, src); 3314 else if (src || 3315 /* FIXME: may not hold if_ioctl_lock */ 3316 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 3317 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 3318 3319 return rc; 3320 } 3321 3322 int 3323 if_do_dad(struct ifnet *ifp) 3324 { 3325 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 3326 return 0; 3327 3328 switch (ifp->if_type) { 3329 case IFT_FAITH: 3330 /* 3331 * These interfaces do not have the IFF_LOOPBACK flag, 3332 * but loop packets back. We do not have to do DAD on such 3333 * interfaces. We should even omit it, because loop-backed 3334 * responses would confuse the DAD procedure. 3335 */ 3336 return 0; 3337 default: 3338 /* 3339 * Our DAD routine requires the interface up and running. 3340 * However, some interfaces can be up before the RUNNING 3341 * status. Additionaly, users may try to assign addresses 3342 * before the interface becomes up (or running). 3343 * We simply skip DAD in such a case as a work around. 3344 * XXX: we should rather mark "tentative" on such addresses, 3345 * and do DAD after the interface becomes ready. 3346 */ 3347 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 3348 (IFF_UP|IFF_RUNNING)) 3349 return 0; 3350 3351 return 1; 3352 } 3353 } 3354 3355 int 3356 if_flags_set(ifnet_t *ifp, const short flags) 3357 { 3358 int rc; 3359 3360 if (ifp->if_setflags != NULL) 3361 rc = (*ifp->if_setflags)(ifp, flags); 3362 else { 3363 short cantflags, chgdflags; 3364 struct ifreq ifr; 3365 3366 chgdflags = ifp->if_flags ^ flags; 3367 cantflags = chgdflags & IFF_CANTCHANGE; 3368 3369 if (cantflags != 0) 3370 ifp->if_flags ^= cantflags; 3371 3372 /* Traditionally, we do not call if_ioctl after 3373 * setting/clearing only IFF_PROMISC if the interface 3374 * isn't IFF_UP. Uphold that tradition. 3375 */ 3376 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 3377 return 0; 3378 3379 memset(&ifr, 0, sizeof(ifr)); 3380 3381 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 3382 /* FIXME: may not hold if_ioctl_lock */ 3383 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 3384 3385 if (rc != 0 && cantflags != 0) 3386 ifp->if_flags ^= cantflags; 3387 } 3388 3389 return rc; 3390 } 3391 3392 int 3393 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 3394 { 3395 int rc; 3396 struct ifreq ifr; 3397 3398 if (ifp->if_mcastop != NULL) 3399 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 3400 else { 3401 ifreq_setaddr(cmd, &ifr, sa); 3402 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 3403 } 3404 3405 return rc; 3406 } 3407 3408 static void 3409 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 3410 struct ifaltq *ifq) 3411 { 3412 const struct sysctlnode *cnode, *rnode; 3413 3414 if (sysctl_createv(clog, 0, NULL, &rnode, 3415 CTLFLAG_PERMANENT, 3416 CTLTYPE_NODE, "interfaces", 3417 SYSCTL_DESCR("Per-interface controls"), 3418 NULL, 0, NULL, 0, 3419 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3420 goto bad; 3421 3422 if (sysctl_createv(clog, 0, &rnode, &rnode, 3423 CTLFLAG_PERMANENT, 3424 CTLTYPE_NODE, ifname, 3425 SYSCTL_DESCR("Interface controls"), 3426 NULL, 0, NULL, 0, 3427 CTL_CREATE, CTL_EOL) != 0) 3428 goto bad; 3429 3430 if (sysctl_createv(clog, 0, &rnode, &rnode, 3431 CTLFLAG_PERMANENT, 3432 CTLTYPE_NODE, "sndq", 3433 SYSCTL_DESCR("Interface output queue controls"), 3434 NULL, 0, NULL, 0, 3435 CTL_CREATE, CTL_EOL) != 0) 3436 goto bad; 3437 3438 if (sysctl_createv(clog, 0, &rnode, &cnode, 3439 CTLFLAG_PERMANENT, 3440 CTLTYPE_INT, "len", 3441 SYSCTL_DESCR("Current output queue length"), 3442 NULL, 0, &ifq->ifq_len, 0, 3443 CTL_CREATE, CTL_EOL) != 0) 3444 goto bad; 3445 3446 if (sysctl_createv(clog, 0, &rnode, &cnode, 3447 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3448 CTLTYPE_INT, "maxlen", 3449 SYSCTL_DESCR("Maximum allowed output queue length"), 3450 NULL, 0, &ifq->ifq_maxlen, 0, 3451 CTL_CREATE, CTL_EOL) != 0) 3452 goto bad; 3453 3454 if (sysctl_createv(clog, 0, &rnode, &cnode, 3455 CTLFLAG_PERMANENT, 3456 CTLTYPE_INT, "drops", 3457 SYSCTL_DESCR("Packets dropped due to full output queue"), 3458 NULL, 0, &ifq->ifq_drops, 0, 3459 CTL_CREATE, CTL_EOL) != 0) 3460 goto bad; 3461 3462 return; 3463 bad: 3464 printf("%s: could not attach sysctl nodes\n", ifname); 3465 return; 3466 } 3467 3468 #if defined(INET) || defined(INET6) 3469 3470 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3471 static int \ 3472 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3473 { \ 3474 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3475 } 3476 3477 #if defined(INET) 3478 static int 3479 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3480 { 3481 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3482 } 3483 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3484 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3485 #endif 3486 3487 #if defined(INET6) 3488 static int 3489 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3490 { 3491 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3492 } 3493 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3494 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3495 #endif 3496 3497 static void 3498 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3499 { 3500 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3501 const char *pfname = NULL, *ipname = NULL; 3502 int ipn = 0, qid = 0; 3503 3504 switch (pf) { 3505 #if defined(INET) 3506 case PF_INET: 3507 len_func = sysctl_net_ip_pktq_items; 3508 maxlen_func = sysctl_net_ip_pktq_maxlen; 3509 drops_func = sysctl_net_ip_pktq_drops; 3510 pfname = "inet", ipn = IPPROTO_IP; 3511 ipname = "ip", qid = IPCTL_IFQ; 3512 break; 3513 #endif 3514 #if defined(INET6) 3515 case PF_INET6: 3516 len_func = sysctl_net_ip6_pktq_items; 3517 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3518 drops_func = sysctl_net_ip6_pktq_drops; 3519 pfname = "inet6", ipn = IPPROTO_IPV6; 3520 ipname = "ip6", qid = IPV6CTL_IFQ; 3521 break; 3522 #endif 3523 default: 3524 KASSERT(false); 3525 } 3526 3527 sysctl_createv(clog, 0, NULL, NULL, 3528 CTLFLAG_PERMANENT, 3529 CTLTYPE_NODE, pfname, NULL, 3530 NULL, 0, NULL, 0, 3531 CTL_NET, pf, CTL_EOL); 3532 sysctl_createv(clog, 0, NULL, NULL, 3533 CTLFLAG_PERMANENT, 3534 CTLTYPE_NODE, ipname, NULL, 3535 NULL, 0, NULL, 0, 3536 CTL_NET, pf, ipn, CTL_EOL); 3537 sysctl_createv(clog, 0, NULL, NULL, 3538 CTLFLAG_PERMANENT, 3539 CTLTYPE_NODE, "ifq", 3540 SYSCTL_DESCR("Protocol input queue controls"), 3541 NULL, 0, NULL, 0, 3542 CTL_NET, pf, ipn, qid, CTL_EOL); 3543 3544 sysctl_createv(clog, 0, NULL, NULL, 3545 CTLFLAG_PERMANENT, 3546 CTLTYPE_INT, "len", 3547 SYSCTL_DESCR("Current input queue length"), 3548 len_func, 0, NULL, 0, 3549 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3550 sysctl_createv(clog, 0, NULL, NULL, 3551 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3552 CTLTYPE_INT, "maxlen", 3553 SYSCTL_DESCR("Maximum allowed input queue length"), 3554 maxlen_func, 0, NULL, 0, 3555 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3556 sysctl_createv(clog, 0, NULL, NULL, 3557 CTLFLAG_PERMANENT, 3558 CTLTYPE_INT, "drops", 3559 SYSCTL_DESCR("Packets dropped due to full input queue"), 3560 drops_func, 0, NULL, 0, 3561 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3562 } 3563 #endif /* INET || INET6 */ 3564 3565 static int 3566 if_sdl_sysctl(SYSCTLFN_ARGS) 3567 { 3568 struct ifnet *ifp; 3569 const struct sockaddr_dl *sdl; 3570 struct psref psref; 3571 int error = 0; 3572 int bound; 3573 3574 if (namelen != 1) 3575 return EINVAL; 3576 3577 bound = curlwp_bind(); 3578 ifp = if_get_byindex(name[0], &psref); 3579 if (ifp == NULL) { 3580 error = ENODEV; 3581 goto out0; 3582 } 3583 3584 sdl = ifp->if_sadl; 3585 if (sdl == NULL) { 3586 *oldlenp = 0; 3587 goto out1; 3588 } 3589 3590 if (oldp == NULL) { 3591 *oldlenp = sdl->sdl_alen; 3592 goto out1; 3593 } 3594 3595 if (*oldlenp >= sdl->sdl_alen) 3596 *oldlenp = sdl->sdl_alen; 3597 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3598 out1: 3599 if_put(ifp, &psref); 3600 out0: 3601 curlwp_bindx(bound); 3602 return error; 3603 } 3604 3605 static void 3606 if_sysctl_setup(struct sysctllog **clog) 3607 { 3608 const struct sysctlnode *rnode = NULL; 3609 3610 sysctl_createv(clog, 0, NULL, &rnode, 3611 CTLFLAG_PERMANENT, 3612 CTLTYPE_NODE, "sdl", 3613 SYSCTL_DESCR("Get active link-layer address"), 3614 if_sdl_sysctl, 0, NULL, 0, 3615 CTL_NET, CTL_CREATE, CTL_EOL); 3616 3617 #if defined(INET) 3618 sysctl_net_pktq_setup(NULL, PF_INET); 3619 #endif 3620 #ifdef INET6 3621 if (in6_present) 3622 sysctl_net_pktq_setup(NULL, PF_INET6); 3623 #endif 3624 } 3625
Indexes created Tue Sep 23 07:09:52 GMT 2025