if.c revision 1.392
1 /* $NetBSD: if.c,v 1.392 2017/04/06 09:20:07 ozaki-r Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by William Studenmund and Jason R. Thorpe. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 34 * All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the project nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 */ 60 61 /* 62 * Copyright (c) 1980, 1986, 1993 63 * The Regents of the University of California. All rights reserved. 64 * 65 * Redistribution and use in source and binary forms, with or without 66 * modification, are permitted provided that the following conditions 67 * are met: 68 * 1. Redistributions of source code must retain the above copyright 69 * notice, this list of conditions and the following disclaimer. 70 * 2. Redistributions in binary form must reproduce the above copyright 71 * notice, this list of conditions and the following disclaimer in the 72 * documentation and/or other materials provided with the distribution. 73 * 3. Neither the name of the University nor the names of its contributors 74 * may be used to endorse or promote products derived from this software 75 * without specific prior written permission. 76 * 77 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 78 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 79 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 80 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 81 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 82 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 83 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 84 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 85 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 86 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 87 * SUCH DAMAGE. 88 * 89 * @(#)if.c 8.5 (Berkeley) 1/9/95 90 */ 91 92 #include <sys/cdefs.h> 93 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.392 2017/04/06 09:20:07 ozaki-r 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 #include <netinet/ip_encap.h> 142 #include <net/bpf.h> 143 144 #ifdef INET6 145 #include <netinet6/in6_var.h> 146 #include <netinet6/nd6.h> 147 #endif 148 149 #include "ether.h" 150 #include "fddi.h" 151 #include "token.h" 152 153 #include "carp.h" 154 #if NCARP > 0 155 #include <netinet/ip_carp.h> 156 #endif 157 158 #include <compat/sys/sockio.h> 159 #include <compat/sys/socket.h> 160 161 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 162 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 163 164 /* 165 * Global list of interfaces. 166 */ 167 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 168 struct ifnet_head ifnet_list; 169 170 struct pslist_head ifnet_pslist; 171 static ifnet_t ** ifindex2ifnet = NULL; 172 static u_int if_index = 1; 173 static size_t if_indexlim = 0; 174 static uint64_t index_gen; 175 /* Mutex to protect the above objects. */ 176 kmutex_t ifnet_mtx __cacheline_aligned; 177 static struct psref_class *ifnet_psref_class __read_mostly; 178 static pserialize_t ifnet_psz; 179 180 static kmutex_t if_clone_mtx; 181 182 struct ifnet *lo0ifp; 183 int ifqmaxlen = IFQ_MAXLEN; 184 185 struct psref_class *ifa_psref_class __read_mostly; 186 187 static int if_delroute_matcher(struct rtentry *, void *); 188 189 static bool if_is_unit(const char *); 190 static struct if_clone *if_clone_lookup(const char *, int *); 191 192 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 193 static int if_cloners_count; 194 195 /* Packet filtering hook for interfaces. */ 196 pfil_head_t * if_pfil __read_mostly; 197 198 static kauth_listener_t if_listener; 199 200 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 201 static void if_detach_queues(struct ifnet *, struct ifqueue *); 202 static void sysctl_sndq_setup(struct sysctllog **, const char *, 203 struct ifaltq *); 204 static void if_slowtimo(void *); 205 static void if_free_sadl(struct ifnet *); 206 static void if_attachdomain1(struct ifnet *); 207 static int ifconf(u_long, void *); 208 static int if_transmit(struct ifnet *, struct mbuf *); 209 static int if_clone_create(const char *); 210 static int if_clone_destroy(const char *); 211 static void if_link_state_change_si(void *); 212 213 struct if_percpuq { 214 struct ifnet *ipq_ifp; 215 void *ipq_si; 216 struct percpu *ipq_ifqs; /* struct ifqueue */ 217 }; 218 219 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 220 221 static void if_percpuq_drops(void *, void *, struct cpu_info *); 222 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 223 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 224 struct if_percpuq *); 225 226 struct if_deferred_start { 227 struct ifnet *ids_ifp; 228 void (*ids_if_start)(struct ifnet *); 229 void *ids_si; 230 }; 231 232 static void if_deferred_start_softint(void *); 233 static void if_deferred_start_common(struct ifnet *); 234 static void if_deferred_start_destroy(struct ifnet *); 235 236 #if defined(INET) || defined(INET6) 237 static void sysctl_net_pktq_setup(struct sysctllog **, int); 238 #endif 239 240 static void if_sysctl_setup(struct sysctllog **); 241 242 /* 243 * Pointer to stub or real compat_cvtcmd() depending on presence of 244 * the compat module 245 */ 246 u_long stub_compat_cvtcmd(u_long); 247 u_long (*vec_compat_cvtcmd)(u_long) = stub_compat_cvtcmd; 248 249 /* Similarly, pointer to compat_ifioctl() if it is present */ 250 251 int (*vec_compat_ifioctl)(struct socket *, u_long, u_long, void *, 252 struct lwp *) = NULL; 253 254 /* The stub version of compat_cvtcmd() */ 255 u_long stub_compat_cvtcmd(u_long cmd) 256 { 257 258 return cmd; 259 } 260 261 static int 262 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 263 void *arg0, void *arg1, void *arg2, void *arg3) 264 { 265 int result; 266 enum kauth_network_req req; 267 268 result = KAUTH_RESULT_DEFER; 269 req = (enum kauth_network_req)arg1; 270 271 if (action != KAUTH_NETWORK_INTERFACE) 272 return result; 273 274 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 275 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 276 result = KAUTH_RESULT_ALLOW; 277 278 return result; 279 } 280 281 /* 282 * Network interface utility routines. 283 * 284 * Routines with ifa_ifwith* names take sockaddr *'s as 285 * parameters. 286 */ 287 void 288 ifinit(void) 289 { 290 291 if_sysctl_setup(NULL); 292 293 #if (defined(INET) || defined(INET6)) 294 encapinit(); 295 #endif 296 297 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 298 if_listener_cb, NULL); 299 300 /* interfaces are available, inform socket code */ 301 ifioctl = doifioctl; 302 } 303 304 /* 305 * XXX Initialization before configure(). 306 * XXX hack to get pfil_add_hook working in autoconf. 307 */ 308 void 309 ifinit1(void) 310 { 311 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 312 313 TAILQ_INIT(&ifnet_list); 314 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 315 ifnet_psz = pserialize_create(); 316 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 317 ifa_psref_class = psref_class_create("ifa", IPL_SOFTNET); 318 PSLIST_INIT(&ifnet_pslist); 319 320 if_indexlim = 8; 321 322 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 323 KASSERT(if_pfil != NULL); 324 325 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 326 etherinit(); 327 #endif 328 } 329 330 ifnet_t * 331 if_alloc(u_char type) 332 { 333 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 334 } 335 336 void 337 if_free(ifnet_t *ifp) 338 { 339 kmem_free(ifp, sizeof(ifnet_t)); 340 } 341 342 void 343 if_initname(struct ifnet *ifp, const char *name, int unit) 344 { 345 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 346 "%s%d", name, unit); 347 } 348 349 /* 350 * Null routines used while an interface is going away. These routines 351 * just return an error. 352 */ 353 354 int 355 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 356 const struct sockaddr *so, const struct rtentry *rt) 357 { 358 359 return ENXIO; 360 } 361 362 void 363 if_nullinput(struct ifnet *ifp, struct mbuf *m) 364 { 365 366 /* Nothing. */ 367 } 368 369 void 370 if_nullstart(struct ifnet *ifp) 371 { 372 373 /* Nothing. */ 374 } 375 376 int 377 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 378 { 379 380 m_freem(m); 381 return ENXIO; 382 } 383 384 int 385 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data) 386 { 387 388 return ENXIO; 389 } 390 391 int 392 if_nullinit(struct ifnet *ifp) 393 { 394 395 return ENXIO; 396 } 397 398 void 399 if_nullstop(struct ifnet *ifp, int disable) 400 { 401 402 /* Nothing. */ 403 } 404 405 void 406 if_nullslowtimo(struct ifnet *ifp) 407 { 408 409 /* Nothing. */ 410 } 411 412 void 413 if_nulldrain(struct ifnet *ifp) 414 { 415 416 /* Nothing. */ 417 } 418 419 void 420 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory) 421 { 422 struct ifaddr *ifa; 423 struct sockaddr_dl *sdl; 424 425 ifp->if_addrlen = addrlen; 426 if_alloc_sadl(ifp); 427 ifa = ifp->if_dl; 428 sdl = satosdl(ifa->ifa_addr); 429 430 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen); 431 if (factory) { 432 ifp->if_hwdl = ifp->if_dl; 433 ifaref(ifp->if_hwdl); 434 } 435 /* TBD routing socket */ 436 } 437 438 struct ifaddr * 439 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp) 440 { 441 unsigned socksize, ifasize; 442 int addrlen, namelen; 443 struct sockaddr_dl *mask, *sdl; 444 struct ifaddr *ifa; 445 446 namelen = strlen(ifp->if_xname); 447 addrlen = ifp->if_addrlen; 448 socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long)); 449 ifasize = sizeof(*ifa) + 2 * socksize; 450 ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO); 451 452 sdl = (struct sockaddr_dl *)(ifa + 1); 453 mask = (struct sockaddr_dl *)(socksize + (char *)sdl); 454 455 sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type, 456 ifp->if_xname, namelen, NULL, addrlen); 457 mask->sdl_len = sockaddr_dl_measure(namelen, 0); 458 memset(&mask->sdl_data[0], 0xff, namelen); 459 ifa->ifa_rtrequest = link_rtrequest; 460 ifa->ifa_addr = (struct sockaddr *)sdl; 461 ifa->ifa_netmask = (struct sockaddr *)mask; 462 ifa_psref_init(ifa); 463 464 *sdlp = sdl; 465 466 return ifa; 467 } 468 469 static void 470 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa) 471 { 472 const struct sockaddr_dl *sdl; 473 474 ifp->if_dl = ifa; 475 ifaref(ifa); 476 sdl = satosdl(ifa->ifa_addr); 477 ifp->if_sadl = sdl; 478 } 479 480 /* 481 * Allocate the link level name for the specified interface. This 482 * is an attachment helper. It must be called after ifp->if_addrlen 483 * is initialized, which may not be the case when if_attach() is 484 * called. 485 */ 486 void 487 if_alloc_sadl(struct ifnet *ifp) 488 { 489 struct ifaddr *ifa; 490 const struct sockaddr_dl *sdl; 491 492 /* 493 * If the interface already has a link name, release it 494 * now. This is useful for interfaces that can change 495 * link types, and thus switch link names often. 496 */ 497 if (ifp->if_sadl != NULL) 498 if_free_sadl(ifp); 499 500 ifa = if_dl_create(ifp, &sdl); 501 502 ifa_insert(ifp, ifa); 503 if_sadl_setrefs(ifp, ifa); 504 } 505 506 static void 507 if_deactivate_sadl(struct ifnet *ifp) 508 { 509 struct ifaddr *ifa; 510 511 KASSERT(ifp->if_dl != NULL); 512 513 ifa = ifp->if_dl; 514 515 ifp->if_sadl = NULL; 516 517 ifp->if_dl = NULL; 518 ifafree(ifa); 519 } 520 521 static void 522 if_replace_sadl(struct ifnet *ifp, struct ifaddr *ifa) 523 { 524 struct ifaddr *old; 525 526 KASSERT(ifp->if_dl != NULL); 527 528 old = ifp->if_dl; 529 530 ifaref(ifa); 531 /* XXX Update if_dl and if_sadl atomically */ 532 ifp->if_dl = ifa; 533 ifp->if_sadl = satosdl(ifa->ifa_addr); 534 535 ifafree(old); 536 } 537 538 void 539 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa0, 540 const struct sockaddr_dl *sdl) 541 { 542 int s, ss; 543 struct ifaddr *ifa; 544 int bound = curlwp_bind(); 545 546 KASSERT(ifa_held(ifa0)); 547 548 s = splsoftnet(); 549 550 if_replace_sadl(ifp, ifa0); 551 552 ss = pserialize_read_enter(); 553 IFADDR_READER_FOREACH(ifa, ifp) { 554 struct psref psref; 555 ifa_acquire(ifa, &psref); 556 pserialize_read_exit(ss); 557 558 rtinit(ifa, RTM_LLINFO_UPD, 0); 559 560 ss = pserialize_read_enter(); 561 ifa_release(ifa, &psref); 562 } 563 pserialize_read_exit(ss); 564 565 splx(s); 566 curlwp_bindx(bound); 567 } 568 569 /* 570 * Free the link level name for the specified interface. This is 571 * a detach helper. This is called from if_detach(). 572 */ 573 static void 574 if_free_sadl(struct ifnet *ifp) 575 { 576 struct ifaddr *ifa; 577 int s; 578 579 ifa = ifp->if_dl; 580 if (ifa == NULL) { 581 KASSERT(ifp->if_sadl == NULL); 582 return; 583 } 584 585 KASSERT(ifp->if_sadl != NULL); 586 587 s = splsoftnet(); 588 rtinit(ifa, RTM_DELETE, 0); 589 ifa_remove(ifp, ifa); 590 if_deactivate_sadl(ifp); 591 if (ifp->if_hwdl == ifa) { 592 ifafree(ifa); 593 ifp->if_hwdl = NULL; 594 } 595 splx(s); 596 } 597 598 static void 599 if_getindex(ifnet_t *ifp) 600 { 601 bool hitlimit = false; 602 603 ifp->if_index_gen = index_gen++; 604 605 ifp->if_index = if_index; 606 if (ifindex2ifnet == NULL) { 607 if_index++; 608 goto skip; 609 } 610 while (if_byindex(ifp->if_index)) { 611 /* 612 * If we hit USHRT_MAX, we skip back to 0 since 613 * there are a number of places where the value 614 * of if_index or if_index itself is compared 615 * to or stored in an unsigned short. By 616 * jumping back, we won't botch those assignments 617 * or comparisons. 618 */ 619 if (++if_index == 0) { 620 if_index = 1; 621 } else if (if_index == USHRT_MAX) { 622 /* 623 * However, if we have to jump back to 624 * zero *twice* without finding an empty 625 * slot in ifindex2ifnet[], then there 626 * there are too many (>65535) interfaces. 627 */ 628 if (hitlimit) { 629 panic("too many interfaces"); 630 } 631 hitlimit = true; 632 if_index = 1; 633 } 634 ifp->if_index = if_index; 635 } 636 skip: 637 /* 638 * ifindex2ifnet is indexed by if_index. Since if_index will 639 * grow dynamically, it should grow too. 640 */ 641 if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) { 642 size_t m, n, oldlim; 643 void *q; 644 645 oldlim = if_indexlim; 646 while (ifp->if_index >= if_indexlim) 647 if_indexlim <<= 1; 648 649 /* grow ifindex2ifnet */ 650 m = oldlim * sizeof(struct ifnet *); 651 n = if_indexlim * sizeof(struct ifnet *); 652 q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO); 653 if (ifindex2ifnet != NULL) { 654 memcpy(q, ifindex2ifnet, m); 655 free(ifindex2ifnet, M_IFADDR); 656 } 657 ifindex2ifnet = (struct ifnet **)q; 658 } 659 ifindex2ifnet[ifp->if_index] = ifp; 660 } 661 662 /* 663 * Initialize an interface and assign an index for it. 664 * 665 * It must be called prior to a device specific attach routine 666 * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl, 667 * and be followed by if_register: 668 * 669 * if_initialize(ifp); 670 * ether_ifattach(ifp, enaddr); 671 * if_register(ifp); 672 */ 673 void 674 if_initialize(ifnet_t *ifp) 675 { 676 KASSERT(if_indexlim > 0); 677 TAILQ_INIT(&ifp->if_addrlist); 678 679 /* 680 * Link level name is allocated later by a separate call to 681 * if_alloc_sadl(). 682 */ 683 684 if (ifp->if_snd.ifq_maxlen == 0) 685 ifp->if_snd.ifq_maxlen = ifqmaxlen; 686 687 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 688 689 ifp->if_link_state = LINK_STATE_UNKNOWN; 690 ifp->if_link_queue = -1; /* all bits set, see link_state_change() */ 691 692 ifp->if_capenable = 0; 693 ifp->if_csum_flags_tx = 0; 694 ifp->if_csum_flags_rx = 0; 695 696 #ifdef ALTQ 697 ifp->if_snd.altq_type = 0; 698 ifp->if_snd.altq_disc = NULL; 699 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 700 ifp->if_snd.altq_tbr = NULL; 701 ifp->if_snd.altq_ifp = ifp; 702 #endif 703 704 IFQ_LOCK_INIT(&ifp->if_snd); 705 706 ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp); 707 pfil_run_ifhooks(if_pfil, PFIL_IFNET_ATTACH, ifp); 708 709 IF_AFDATA_LOCK_INIT(ifp); 710 711 if (if_is_link_state_changeable(ifp)) { 712 ifp->if_link_si = softint_establish(SOFTINT_NET, 713 if_link_state_change_si, ifp); 714 if (ifp->if_link_si == NULL) 715 panic("%s: softint_establish() failed", __func__); 716 } 717 718 PSLIST_ENTRY_INIT(ifp, if_pslist_entry); 719 PSLIST_INIT(&ifp->if_addr_pslist); 720 psref_target_init(&ifp->if_psref, ifnet_psref_class); 721 ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 722 LIST_INIT(&ifp->if_multiaddrs); 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 int s; 1011 1012 s = splnet(); 1013 if_start_lock(ifp); 1014 splx(s); 1015 } 1016 1017 static inline bool 1018 if_snd_is_used(struct ifnet *ifp) 1019 { 1020 1021 return ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit || 1022 ALTQ_IS_ENABLED(&ifp->if_snd); 1023 } 1024 1025 /* 1026 * Schedule deferred if_start. 1027 */ 1028 void 1029 if_schedule_deferred_start(struct ifnet *ifp) 1030 { 1031 1032 KASSERT(ifp->if_deferred_start != NULL); 1033 1034 if (if_snd_is_used(ifp) && IFQ_IS_EMPTY(&ifp->if_snd)) 1035 return; 1036 1037 softint_schedule(ifp->if_deferred_start->ids_si); 1038 } 1039 1040 /* 1041 * Create an instance of deferred if_start. A driver should call the function 1042 * only if the driver needs deferred if_start. Drivers can setup their own 1043 * deferred if_start function via 2nd argument. 1044 */ 1045 void 1046 if_deferred_start_init(struct ifnet *ifp, void (*func)(struct ifnet *)) 1047 { 1048 struct if_deferred_start *ids; 1049 1050 ids = kmem_zalloc(sizeof(*ids), KM_SLEEP); 1051 if (ids == NULL) 1052 panic("kmem_zalloc failed"); 1053 1054 ids->ids_ifp = ifp; 1055 ids->ids_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 1056 if_deferred_start_softint, ids); 1057 if (func != NULL) 1058 ids->ids_if_start = func; 1059 else 1060 ids->ids_if_start = if_deferred_start_common; 1061 1062 ifp->if_deferred_start = ids; 1063 } 1064 1065 static void 1066 if_deferred_start_destroy(struct ifnet *ifp) 1067 { 1068 1069 if (ifp->if_deferred_start == NULL) 1070 return; 1071 1072 softint_disestablish(ifp->if_deferred_start->ids_si); 1073 kmem_free(ifp->if_deferred_start, sizeof(*ifp->if_deferred_start)); 1074 ifp->if_deferred_start = NULL; 1075 } 1076 1077 /* 1078 * The common interface input routine that is called by device drivers, 1079 * which should be used only when the driver's rx handler already runs 1080 * in softint. 1081 */ 1082 void 1083 if_input(struct ifnet *ifp, struct mbuf *m) 1084 { 1085 1086 KASSERT(ifp->if_percpuq == NULL); 1087 KASSERT(!cpu_intr_p()); 1088 1089 ifp->if_ipackets++; 1090 bpf_mtap(ifp, m); 1091 1092 ifp->_if_input(ifp, m); 1093 } 1094 1095 /* 1096 * DEPRECATED. Use if_initialize and if_register instead. 1097 * See the above comment of if_initialize. 1098 * 1099 * Note that it implicitly enables if_percpuq to make drivers easy to 1100 * migrate softint-based if_input without much changes. If you don't 1101 * want to enable it, use if_initialize instead. 1102 */ 1103 void 1104 if_attach(ifnet_t *ifp) 1105 { 1106 1107 if_initialize(ifp); 1108 ifp->if_percpuq = if_percpuq_create(ifp); 1109 if_register(ifp); 1110 } 1111 1112 void 1113 if_attachdomain(void) 1114 { 1115 struct ifnet *ifp; 1116 int s; 1117 int bound = curlwp_bind(); 1118 1119 s = pserialize_read_enter(); 1120 IFNET_READER_FOREACH(ifp) { 1121 struct psref psref; 1122 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 1123 pserialize_read_exit(s); 1124 if_attachdomain1(ifp); 1125 s = pserialize_read_enter(); 1126 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 1127 } 1128 pserialize_read_exit(s); 1129 curlwp_bindx(bound); 1130 } 1131 1132 static void 1133 if_attachdomain1(struct ifnet *ifp) 1134 { 1135 struct domain *dp; 1136 int s; 1137 1138 s = splsoftnet(); 1139 1140 /* address family dependent data region */ 1141 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 1142 DOMAIN_FOREACH(dp) { 1143 if (dp->dom_ifattach != NULL) 1144 ifp->if_afdata[dp->dom_family] = 1145 (*dp->dom_ifattach)(ifp); 1146 } 1147 1148 splx(s); 1149 } 1150 1151 /* 1152 * Deactivate an interface. This points all of the procedure 1153 * handles at error stubs. May be called from interrupt context. 1154 */ 1155 void 1156 if_deactivate(struct ifnet *ifp) 1157 { 1158 int s; 1159 1160 s = splsoftnet(); 1161 1162 ifp->if_output = if_nulloutput; 1163 ifp->_if_input = if_nullinput; 1164 ifp->if_start = if_nullstart; 1165 ifp->if_transmit = if_nulltransmit; 1166 ifp->if_ioctl = if_nullioctl; 1167 ifp->if_init = if_nullinit; 1168 ifp->if_stop = if_nullstop; 1169 ifp->if_slowtimo = if_nullslowtimo; 1170 ifp->if_drain = if_nulldrain; 1171 1172 /* No more packets may be enqueued. */ 1173 ifp->if_snd.ifq_maxlen = 0; 1174 1175 splx(s); 1176 } 1177 1178 bool 1179 if_is_deactivated(const struct ifnet *ifp) 1180 { 1181 1182 return ifp->if_output == if_nulloutput; 1183 } 1184 1185 void 1186 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1187 { 1188 struct ifaddr *ifa, *nifa; 1189 int s; 1190 1191 s = pserialize_read_enter(); 1192 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) { 1193 nifa = IFADDR_READER_NEXT(ifa); 1194 if (ifa->ifa_addr->sa_family != family) 1195 continue; 1196 pserialize_read_exit(s); 1197 1198 (*purgeaddr)(ifa); 1199 1200 s = pserialize_read_enter(); 1201 } 1202 pserialize_read_exit(s); 1203 } 1204 1205 #ifdef IFAREF_DEBUG 1206 static struct ifaddr **ifa_list; 1207 static int ifa_list_size; 1208 1209 /* Depends on only one if_attach runs at once */ 1210 static void 1211 if_build_ifa_list(struct ifnet *ifp) 1212 { 1213 struct ifaddr *ifa; 1214 int i; 1215 1216 KASSERT(ifa_list == NULL); 1217 KASSERT(ifa_list_size == 0); 1218 1219 IFADDR_READER_FOREACH(ifa, ifp) 1220 ifa_list_size++; 1221 1222 ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP); 1223 if (ifa_list == NULL) 1224 return; 1225 1226 i = 0; 1227 IFADDR_READER_FOREACH(ifa, ifp) { 1228 ifa_list[i++] = ifa; 1229 ifaref(ifa); 1230 } 1231 } 1232 1233 static void 1234 if_check_and_free_ifa_list(struct ifnet *ifp) 1235 { 1236 int i; 1237 struct ifaddr *ifa; 1238 1239 if (ifa_list == NULL) 1240 return; 1241 1242 for (i = 0; i < ifa_list_size; i++) { 1243 char buf[64]; 1244 1245 ifa = ifa_list[i]; 1246 sockaddr_format(ifa->ifa_addr, buf, sizeof(buf)); 1247 if (ifa->ifa_refcnt > 1) { 1248 log(LOG_WARNING, 1249 "ifa(%s) still referenced (refcnt=%d)\n", 1250 buf, ifa->ifa_refcnt - 1); 1251 } else 1252 log(LOG_DEBUG, 1253 "ifa(%s) not referenced (refcnt=%d)\n", 1254 buf, ifa->ifa_refcnt - 1); 1255 ifafree(ifa); 1256 } 1257 1258 kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size); 1259 ifa_list = NULL; 1260 ifa_list_size = 0; 1261 } 1262 #endif 1263 1264 /* 1265 * Detach an interface from the list of "active" interfaces, 1266 * freeing any resources as we go along. 1267 * 1268 * NOTE: This routine must be called with a valid thread context, 1269 * as it may block. 1270 */ 1271 void 1272 if_detach(struct ifnet *ifp) 1273 { 1274 struct socket so; 1275 struct ifaddr *ifa; 1276 #ifdef IFAREF_DEBUG 1277 struct ifaddr *last_ifa = NULL; 1278 #endif 1279 struct domain *dp; 1280 const struct protosw *pr; 1281 int s, i, family, purged; 1282 uint64_t xc; 1283 1284 #ifdef IFAREF_DEBUG 1285 if_build_ifa_list(ifp); 1286 #endif 1287 /* 1288 * XXX It's kind of lame that we have to have the 1289 * XXX socket structure... 1290 */ 1291 memset(&so, 0, sizeof(so)); 1292 1293 s = splnet(); 1294 1295 sysctl_teardown(&ifp->if_sysctl_log); 1296 mutex_enter(ifp->if_ioctl_lock); 1297 if_deactivate(ifp); 1298 mutex_exit(ifp->if_ioctl_lock); 1299 1300 IFNET_LOCK(); 1301 ifindex2ifnet[ifp->if_index] = NULL; 1302 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1303 IFNET_WRITER_REMOVE(ifp); 1304 pserialize_perform(ifnet_psz); 1305 IFNET_UNLOCK(); 1306 1307 /* Wait for all readers to drain before freeing. */ 1308 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1309 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1310 1311 mutex_obj_free(ifp->if_ioctl_lock); 1312 ifp->if_ioctl_lock = NULL; 1313 1314 if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) { 1315 ifp->if_slowtimo = NULL; 1316 callout_halt(ifp->if_slowtimo_ch, NULL); 1317 callout_destroy(ifp->if_slowtimo_ch); 1318 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1319 } 1320 if_deferred_start_destroy(ifp); 1321 1322 /* 1323 * Do an if_down() to give protocols a chance to do something. 1324 */ 1325 if_down(ifp); 1326 1327 #ifdef ALTQ 1328 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1329 altq_disable(&ifp->if_snd); 1330 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1331 altq_detach(&ifp->if_snd); 1332 #endif 1333 1334 mutex_obj_free(ifp->if_snd.ifq_lock); 1335 1336 #if NCARP > 0 1337 /* Remove the interface from any carp group it is a part of. */ 1338 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1339 carp_ifdetach(ifp); 1340 #endif 1341 1342 /* 1343 * Rip all the addresses off the interface. This should make 1344 * all of the routes go away. 1345 * 1346 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1347 * from the list, including our "cursor", ifa. For safety, 1348 * and to honor the TAILQ abstraction, I just restart the 1349 * loop after each removal. Note that the loop will exit 1350 * when all of the remaining ifaddrs belong to the AF_LINK 1351 * family. I am counting on the historical fact that at 1352 * least one pr_usrreq in each address domain removes at 1353 * least one ifaddr. 1354 */ 1355 again: 1356 /* 1357 * At this point, no other one tries to remove ifa in the list, 1358 * so we don't need to take a lock or psref. 1359 */ 1360 IFADDR_READER_FOREACH(ifa, ifp) { 1361 family = ifa->ifa_addr->sa_family; 1362 #ifdef IFAREF_DEBUG 1363 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1364 ifa, family, ifa->ifa_refcnt); 1365 if (last_ifa != NULL && ifa == last_ifa) 1366 panic("if_detach: loop detected"); 1367 last_ifa = ifa; 1368 #endif 1369 if (family == AF_LINK) 1370 continue; 1371 dp = pffinddomain(family); 1372 KASSERTMSG(dp != NULL, "no domain for AF %d", family); 1373 /* 1374 * XXX These PURGEIF calls are redundant with the 1375 * purge-all-families calls below, but are left in for 1376 * now both to make a smaller change, and to avoid 1377 * unplanned interactions with clearing of 1378 * ifp->if_addrlist. 1379 */ 1380 purged = 0; 1381 for (pr = dp->dom_protosw; 1382 pr < dp->dom_protoswNPROTOSW; pr++) { 1383 so.so_proto = pr; 1384 if (pr->pr_usrreqs) { 1385 (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1386 purged = 1; 1387 } 1388 } 1389 if (purged == 0) { 1390 /* 1391 * XXX What's really the best thing to do 1392 * XXX here? --thorpej (at) NetBSD.org 1393 */ 1394 printf("if_detach: WARNING: AF %d not purged\n", 1395 family); 1396 ifa_remove(ifp, ifa); 1397 } 1398 goto again; 1399 } 1400 1401 if_free_sadl(ifp); 1402 1403 /* Delete stray routes from the routing table. */ 1404 for (i = 0; i <= AF_MAX; i++) 1405 rt_delete_matched_entries(i, if_delroute_matcher, ifp); 1406 1407 DOMAIN_FOREACH(dp) { 1408 if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family]) 1409 { 1410 void *p = ifp->if_afdata[dp->dom_family]; 1411 if (p) { 1412 ifp->if_afdata[dp->dom_family] = NULL; 1413 (*dp->dom_ifdetach)(ifp, p); 1414 } 1415 } 1416 1417 /* 1418 * One would expect multicast memberships (INET and 1419 * INET6) on UDP sockets to be purged by the PURGEIF 1420 * calls above, but if all addresses were removed from 1421 * the interface prior to destruction, the calls will 1422 * not be made (e.g. ppp, for which pppd(8) generally 1423 * removes addresses before destroying the interface). 1424 * Because there is no invariant that multicast 1425 * memberships only exist for interfaces with IPv4 1426 * addresses, we must call PURGEIF regardless of 1427 * addresses. (Protocols which might store ifnet 1428 * pointers are marked with PR_PURGEIF.) 1429 */ 1430 for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { 1431 so.so_proto = pr; 1432 if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF) 1433 (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp); 1434 } 1435 } 1436 1437 pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp); 1438 (void)pfil_head_destroy(ifp->if_pfil); 1439 1440 /* Announce that the interface is gone. */ 1441 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1442 1443 IF_AFDATA_LOCK_DESTROY(ifp); 1444 1445 if (if_is_link_state_changeable(ifp)) { 1446 softint_disestablish(ifp->if_link_si); 1447 ifp->if_link_si = NULL; 1448 } 1449 1450 /* 1451 * remove packets that came from ifp, from software interrupt queues. 1452 */ 1453 DOMAIN_FOREACH(dp) { 1454 for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) { 1455 struct ifqueue *iq = dp->dom_ifqueues[i]; 1456 if (iq == NULL) 1457 break; 1458 dp->dom_ifqueues[i] = NULL; 1459 if_detach_queues(ifp, iq); 1460 } 1461 } 1462 1463 /* 1464 * IP queues have to be processed separately: net-queue barrier 1465 * ensures that the packets are dequeued while a cross-call will 1466 * ensure that the interrupts have completed. FIXME: not quite.. 1467 */ 1468 #ifdef INET 1469 pktq_barrier(ip_pktq); 1470 #endif 1471 #ifdef INET6 1472 if (in6_present) 1473 pktq_barrier(ip6_pktq); 1474 #endif 1475 xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL); 1476 xc_wait(xc); 1477 1478 if (ifp->if_percpuq != NULL) { 1479 if_percpuq_destroy(ifp->if_percpuq); 1480 ifp->if_percpuq = NULL; 1481 } 1482 1483 splx(s); 1484 1485 #ifdef IFAREF_DEBUG 1486 if_check_and_free_ifa_list(ifp); 1487 #endif 1488 } 1489 1490 static void 1491 if_detach_queues(struct ifnet *ifp, struct ifqueue *q) 1492 { 1493 struct mbuf *m, *prev, *next; 1494 1495 prev = NULL; 1496 for (m = q->ifq_head; m != NULL; m = next) { 1497 KASSERT((m->m_flags & M_PKTHDR) != 0); 1498 1499 next = m->m_nextpkt; 1500 if (m->m_pkthdr.rcvif_index != ifp->if_index) { 1501 prev = m; 1502 continue; 1503 } 1504 1505 if (prev != NULL) 1506 prev->m_nextpkt = m->m_nextpkt; 1507 else 1508 q->ifq_head = m->m_nextpkt; 1509 if (q->ifq_tail == m) 1510 q->ifq_tail = prev; 1511 q->ifq_len--; 1512 1513 m->m_nextpkt = NULL; 1514 m_freem(m); 1515 IF_DROP(q); 1516 } 1517 } 1518 1519 /* 1520 * Callback for a radix tree walk to delete all references to an 1521 * ifnet. 1522 */ 1523 static int 1524 if_delroute_matcher(struct rtentry *rt, void *v) 1525 { 1526 struct ifnet *ifp = (struct ifnet *)v; 1527 1528 if (rt->rt_ifp == ifp) 1529 return 1; 1530 else 1531 return 0; 1532 } 1533 1534 /* 1535 * Create a clone network interface. 1536 */ 1537 static int 1538 if_clone_create(const char *name) 1539 { 1540 struct if_clone *ifc; 1541 int unit; 1542 struct ifnet *ifp; 1543 struct psref psref; 1544 1545 KASSERT(mutex_owned(&if_clone_mtx)); 1546 1547 ifc = if_clone_lookup(name, &unit); 1548 if (ifc == NULL) 1549 return EINVAL; 1550 1551 ifp = if_get(name, &psref); 1552 if (ifp != NULL) { 1553 if_put(ifp, &psref); 1554 return EEXIST; 1555 } 1556 1557 return (*ifc->ifc_create)(ifc, unit); 1558 } 1559 1560 /* 1561 * Destroy a clone network interface. 1562 */ 1563 static int 1564 if_clone_destroy(const char *name) 1565 { 1566 struct if_clone *ifc; 1567 struct ifnet *ifp; 1568 struct psref psref; 1569 1570 KASSERT(mutex_owned(&if_clone_mtx)); 1571 1572 ifc = if_clone_lookup(name, NULL); 1573 if (ifc == NULL) 1574 return EINVAL; 1575 1576 if (ifc->ifc_destroy == NULL) 1577 return EOPNOTSUPP; 1578 1579 ifp = if_get(name, &psref); 1580 if (ifp == NULL) 1581 return ENXIO; 1582 1583 /* We have to disable ioctls here */ 1584 mutex_enter(ifp->if_ioctl_lock); 1585 ifp->if_ioctl = if_nullioctl; 1586 mutex_exit(ifp->if_ioctl_lock); 1587 1588 /* 1589 * We cannot call ifc_destroy with holding ifp. 1590 * Releasing ifp here is safe thanks to if_clone_mtx. 1591 */ 1592 if_put(ifp, &psref); 1593 1594 return (*ifc->ifc_destroy)(ifp); 1595 } 1596 1597 static bool 1598 if_is_unit(const char *name) 1599 { 1600 1601 while(*name != '\0') { 1602 if (*name < '0' || *name > '9') 1603 return false; 1604 name++; 1605 } 1606 1607 return true; 1608 } 1609 1610 /* 1611 * Look up a network interface cloner. 1612 */ 1613 static struct if_clone * 1614 if_clone_lookup(const char *name, int *unitp) 1615 { 1616 struct if_clone *ifc; 1617 const char *cp; 1618 char *dp, ifname[IFNAMSIZ + 3]; 1619 int unit; 1620 1621 KASSERT(mutex_owned(&if_clone_mtx)); 1622 1623 strcpy(ifname, "if_"); 1624 /* separate interface name from unit */ 1625 /* TODO: search unit number from backward */ 1626 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1627 *cp && !if_is_unit(cp);) 1628 *dp++ = *cp++; 1629 1630 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1631 return NULL; /* No name or unit number */ 1632 *dp++ = '\0'; 1633 1634 again: 1635 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1636 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1637 break; 1638 } 1639 1640 if (ifc == NULL) { 1641 int error; 1642 if (*ifname == '\0') 1643 return NULL; 1644 mutex_exit(&if_clone_mtx); 1645 error = module_autoload(ifname, MODULE_CLASS_DRIVER); 1646 mutex_enter(&if_clone_mtx); 1647 if (error) 1648 return NULL; 1649 *ifname = '\0'; 1650 goto again; 1651 } 1652 1653 unit = 0; 1654 while (cp - name < IFNAMSIZ && *cp) { 1655 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1656 /* Bogus unit number. */ 1657 return NULL; 1658 } 1659 unit = (unit * 10) + (*cp++ - '0'); 1660 } 1661 1662 if (unitp != NULL) 1663 *unitp = unit; 1664 return ifc; 1665 } 1666 1667 /* 1668 * Register a network interface cloner. 1669 */ 1670 void 1671 if_clone_attach(struct if_clone *ifc) 1672 { 1673 1674 mutex_enter(&if_clone_mtx); 1675 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1676 if_cloners_count++; 1677 mutex_exit(&if_clone_mtx); 1678 } 1679 1680 /* 1681 * Unregister a network interface cloner. 1682 */ 1683 void 1684 if_clone_detach(struct if_clone *ifc) 1685 { 1686 1687 mutex_enter(&if_clone_mtx); 1688 LIST_REMOVE(ifc, ifc_list); 1689 if_cloners_count--; 1690 mutex_exit(&if_clone_mtx); 1691 } 1692 1693 /* 1694 * Provide list of interface cloners to userspace. 1695 */ 1696 int 1697 if_clone_list(int buf_count, char *buffer, int *total) 1698 { 1699 char outbuf[IFNAMSIZ], *dst; 1700 struct if_clone *ifc; 1701 int count, error = 0; 1702 1703 mutex_enter(&if_clone_mtx); 1704 *total = if_cloners_count; 1705 if ((dst = buffer) == NULL) { 1706 /* Just asking how many there are. */ 1707 goto out; 1708 } 1709 1710 if (buf_count < 0) { 1711 error = EINVAL; 1712 goto out; 1713 } 1714 1715 count = (if_cloners_count < buf_count) ? 1716 if_cloners_count : buf_count; 1717 1718 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1719 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1720 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1721 if (outbuf[sizeof(outbuf) - 1] != '\0') { 1722 error = ENAMETOOLONG; 1723 goto out; 1724 } 1725 error = copyout(outbuf, dst, sizeof(outbuf)); 1726 if (error != 0) 1727 break; 1728 } 1729 1730 out: 1731 mutex_exit(&if_clone_mtx); 1732 return error; 1733 } 1734 1735 void 1736 ifa_psref_init(struct ifaddr *ifa) 1737 { 1738 1739 psref_target_init(&ifa->ifa_psref, ifa_psref_class); 1740 } 1741 1742 void 1743 ifaref(struct ifaddr *ifa) 1744 { 1745 KASSERT(!ISSET(ifa->ifa_flags, IFA_DESTROYING)); 1746 ifa->ifa_refcnt++; 1747 } 1748 1749 void 1750 ifafree(struct ifaddr *ifa) 1751 { 1752 KASSERT(ifa != NULL); 1753 KASSERT(ifa->ifa_refcnt > 0); 1754 1755 if (--ifa->ifa_refcnt == 0) { 1756 free(ifa, M_IFADDR); 1757 } 1758 } 1759 1760 bool 1761 ifa_is_destroying(struct ifaddr *ifa) 1762 { 1763 1764 return ISSET(ifa->ifa_flags, IFA_DESTROYING); 1765 } 1766 1767 void 1768 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1769 { 1770 1771 ifa->ifa_ifp = ifp; 1772 1773 IFNET_LOCK(); 1774 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1775 IFADDR_ENTRY_INIT(ifa); 1776 IFADDR_WRITER_INSERT_TAIL(ifp, ifa); 1777 IFNET_UNLOCK(); 1778 1779 ifaref(ifa); 1780 } 1781 1782 void 1783 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1784 { 1785 1786 KASSERT(ifa->ifa_ifp == ifp); 1787 1788 IFNET_LOCK(); 1789 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1790 IFADDR_WRITER_REMOVE(ifa); 1791 #ifdef NET_MPSAFE 1792 pserialize_perform(ifnet_psz); 1793 #endif 1794 IFNET_UNLOCK(); 1795 1796 #ifdef NET_MPSAFE 1797 psref_target_destroy(&ifa->ifa_psref, ifa_psref_class); 1798 #endif 1799 IFADDR_ENTRY_DESTROY(ifa); 1800 ifafree(ifa); 1801 } 1802 1803 void 1804 ifa_acquire(struct ifaddr *ifa, struct psref *psref) 1805 { 1806 1807 psref_acquire(psref, &ifa->ifa_psref, ifa_psref_class); 1808 } 1809 1810 void 1811 ifa_release(struct ifaddr *ifa, struct psref *psref) 1812 { 1813 1814 if (ifa == NULL) 1815 return; 1816 1817 psref_release(psref, &ifa->ifa_psref, ifa_psref_class); 1818 } 1819 1820 bool 1821 ifa_held(struct ifaddr *ifa) 1822 { 1823 1824 return psref_held(&ifa->ifa_psref, ifa_psref_class); 1825 } 1826 1827 static inline int 1828 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1829 { 1830 return sockaddr_cmp(sa1, sa2) == 0; 1831 } 1832 1833 /* 1834 * Locate an interface based on a complete address. 1835 */ 1836 /*ARGSUSED*/ 1837 struct ifaddr * 1838 ifa_ifwithaddr(const struct sockaddr *addr) 1839 { 1840 struct ifnet *ifp; 1841 struct ifaddr *ifa; 1842 1843 IFNET_READER_FOREACH(ifp) { 1844 if (if_is_deactivated(ifp)) 1845 continue; 1846 IFADDR_READER_FOREACH(ifa, ifp) { 1847 if (ifa->ifa_addr->sa_family != addr->sa_family) 1848 continue; 1849 if (equal(addr, ifa->ifa_addr)) 1850 return ifa; 1851 if ((ifp->if_flags & IFF_BROADCAST) && 1852 ifa->ifa_broadaddr && 1853 /* IP6 doesn't have broadcast */ 1854 ifa->ifa_broadaddr->sa_len != 0 && 1855 equal(ifa->ifa_broadaddr, addr)) 1856 return ifa; 1857 } 1858 } 1859 return NULL; 1860 } 1861 1862 struct ifaddr * 1863 ifa_ifwithaddr_psref(const struct sockaddr *addr, struct psref *psref) 1864 { 1865 struct ifaddr *ifa; 1866 int s = pserialize_read_enter(); 1867 1868 ifa = ifa_ifwithaddr(addr); 1869 if (ifa != NULL) 1870 ifa_acquire(ifa, psref); 1871 pserialize_read_exit(s); 1872 1873 return ifa; 1874 } 1875 1876 /* 1877 * Locate the point to point interface with a given destination address. 1878 */ 1879 /*ARGSUSED*/ 1880 struct ifaddr * 1881 ifa_ifwithdstaddr(const struct sockaddr *addr) 1882 { 1883 struct ifnet *ifp; 1884 struct ifaddr *ifa; 1885 1886 IFNET_READER_FOREACH(ifp) { 1887 if (if_is_deactivated(ifp)) 1888 continue; 1889 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1890 continue; 1891 IFADDR_READER_FOREACH(ifa, ifp) { 1892 if (ifa->ifa_addr->sa_family != addr->sa_family || 1893 ifa->ifa_dstaddr == NULL) 1894 continue; 1895 if (equal(addr, ifa->ifa_dstaddr)) 1896 return ifa; 1897 } 1898 } 1899 1900 return NULL; 1901 } 1902 1903 struct ifaddr * 1904 ifa_ifwithdstaddr_psref(const struct sockaddr *addr, struct psref *psref) 1905 { 1906 struct ifaddr *ifa; 1907 int s; 1908 1909 s = pserialize_read_enter(); 1910 ifa = ifa_ifwithdstaddr(addr); 1911 if (ifa != NULL) 1912 ifa_acquire(ifa, psref); 1913 pserialize_read_exit(s); 1914 1915 return ifa; 1916 } 1917 1918 /* 1919 * Find an interface on a specific network. If many, choice 1920 * is most specific found. 1921 */ 1922 struct ifaddr * 1923 ifa_ifwithnet(const struct sockaddr *addr) 1924 { 1925 struct ifnet *ifp; 1926 struct ifaddr *ifa, *ifa_maybe = NULL; 1927 const struct sockaddr_dl *sdl; 1928 u_int af = addr->sa_family; 1929 const char *addr_data = addr->sa_data, *cplim; 1930 1931 if (af == AF_LINK) { 1932 sdl = satocsdl(addr); 1933 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1934 ifindex2ifnet[sdl->sdl_index] && 1935 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1936 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1937 } 1938 } 1939 #ifdef NETATALK 1940 if (af == AF_APPLETALK) { 1941 const struct sockaddr_at *sat, *sat2; 1942 sat = (const struct sockaddr_at *)addr; 1943 IFNET_READER_FOREACH(ifp) { 1944 if (if_is_deactivated(ifp)) 1945 continue; 1946 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1947 if (ifa == NULL) 1948 continue; 1949 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1950 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1951 return ifa; /* exact match */ 1952 if (ifa_maybe == NULL) { 1953 /* else keep the if with the right range */ 1954 ifa_maybe = ifa; 1955 } 1956 } 1957 return ifa_maybe; 1958 } 1959 #endif 1960 IFNET_READER_FOREACH(ifp) { 1961 if (if_is_deactivated(ifp)) 1962 continue; 1963 IFADDR_READER_FOREACH(ifa, ifp) { 1964 const char *cp, *cp2, *cp3; 1965 1966 if (ifa->ifa_addr->sa_family != af || 1967 ifa->ifa_netmask == NULL) 1968 next: continue; 1969 cp = addr_data; 1970 cp2 = ifa->ifa_addr->sa_data; 1971 cp3 = ifa->ifa_netmask->sa_data; 1972 cplim = (const char *)ifa->ifa_netmask + 1973 ifa->ifa_netmask->sa_len; 1974 while (cp3 < cplim) { 1975 if ((*cp++ ^ *cp2++) & *cp3++) { 1976 /* want to continue for() loop */ 1977 goto next; 1978 } 1979 } 1980 if (ifa_maybe == NULL || 1981 rt_refines(ifa->ifa_netmask, 1982 ifa_maybe->ifa_netmask)) 1983 ifa_maybe = ifa; 1984 } 1985 } 1986 return ifa_maybe; 1987 } 1988 1989 struct ifaddr * 1990 ifa_ifwithnet_psref(const struct sockaddr *addr, struct psref *psref) 1991 { 1992 struct ifaddr *ifa; 1993 int s; 1994 1995 s = pserialize_read_enter(); 1996 ifa = ifa_ifwithnet(addr); 1997 if (ifa != NULL) 1998 ifa_acquire(ifa, psref); 1999 pserialize_read_exit(s); 2000 2001 return ifa; 2002 } 2003 2004 /* 2005 * Find the interface of the addresss. 2006 */ 2007 struct ifaddr * 2008 ifa_ifwithladdr(const struct sockaddr *addr) 2009 { 2010 struct ifaddr *ia; 2011 2012 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 2013 (ia = ifa_ifwithnet(addr))) 2014 return ia; 2015 return NULL; 2016 } 2017 2018 struct ifaddr * 2019 ifa_ifwithladdr_psref(const struct sockaddr *addr, struct psref *psref) 2020 { 2021 struct ifaddr *ifa; 2022 int s; 2023 2024 s = pserialize_read_enter(); 2025 ifa = ifa_ifwithladdr(addr); 2026 if (ifa != NULL) 2027 ifa_acquire(ifa, psref); 2028 pserialize_read_exit(s); 2029 2030 return ifa; 2031 } 2032 2033 /* 2034 * Find an interface using a specific address family 2035 */ 2036 struct ifaddr * 2037 ifa_ifwithaf(int af) 2038 { 2039 struct ifnet *ifp; 2040 struct ifaddr *ifa = NULL; 2041 int s; 2042 2043 s = pserialize_read_enter(); 2044 IFNET_READER_FOREACH(ifp) { 2045 if (if_is_deactivated(ifp)) 2046 continue; 2047 IFADDR_READER_FOREACH(ifa, ifp) { 2048 if (ifa->ifa_addr->sa_family == af) 2049 goto out; 2050 } 2051 } 2052 out: 2053 pserialize_read_exit(s); 2054 return ifa; 2055 } 2056 2057 /* 2058 * Find an interface address specific to an interface best matching 2059 * a given address. 2060 */ 2061 struct ifaddr * 2062 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2063 { 2064 struct ifaddr *ifa; 2065 const char *cp, *cp2, *cp3; 2066 const char *cplim; 2067 struct ifaddr *ifa_maybe = 0; 2068 u_int af = addr->sa_family; 2069 2070 if (if_is_deactivated(ifp)) 2071 return NULL; 2072 2073 if (af >= AF_MAX) 2074 return NULL; 2075 2076 IFADDR_READER_FOREACH(ifa, ifp) { 2077 if (ifa->ifa_addr->sa_family != af) 2078 continue; 2079 ifa_maybe = ifa; 2080 if (ifa->ifa_netmask == NULL) { 2081 if (equal(addr, ifa->ifa_addr) || 2082 (ifa->ifa_dstaddr && 2083 equal(addr, ifa->ifa_dstaddr))) 2084 return ifa; 2085 continue; 2086 } 2087 cp = addr->sa_data; 2088 cp2 = ifa->ifa_addr->sa_data; 2089 cp3 = ifa->ifa_netmask->sa_data; 2090 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2091 for (; cp3 < cplim; cp3++) { 2092 if ((*cp++ ^ *cp2++) & *cp3) 2093 break; 2094 } 2095 if (cp3 == cplim) 2096 return ifa; 2097 } 2098 return ifa_maybe; 2099 } 2100 2101 struct ifaddr * 2102 ifaof_ifpforaddr_psref(const struct sockaddr *addr, struct ifnet *ifp, 2103 struct psref *psref) 2104 { 2105 struct ifaddr *ifa; 2106 int s; 2107 2108 s = pserialize_read_enter(); 2109 ifa = ifaof_ifpforaddr(addr, ifp); 2110 if (ifa != NULL) 2111 ifa_acquire(ifa, psref); 2112 pserialize_read_exit(s); 2113 2114 return ifa; 2115 } 2116 2117 /* 2118 * Default action when installing a route with a Link Level gateway. 2119 * Lookup an appropriate real ifa to point to. 2120 * This should be moved to /sys/net/link.c eventually. 2121 */ 2122 void 2123 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 2124 { 2125 struct ifaddr *ifa; 2126 const struct sockaddr *dst; 2127 struct ifnet *ifp; 2128 struct psref psref; 2129 2130 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 2131 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 2132 return; 2133 if ((ifa = ifaof_ifpforaddr_psref(dst, ifp, &psref)) != NULL) { 2134 rt_replace_ifa(rt, ifa); 2135 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2136 ifa->ifa_rtrequest(cmd, rt, info); 2137 ifa_release(ifa, &psref); 2138 } 2139 } 2140 2141 /* 2142 * bitmask macros to manage a densely packed link_state change queue. 2143 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 2144 * LINK_STATE_UP(2) we need 2 bits for each state change. 2145 * As a state change to store is 0, treat all bits set as an unset item. 2146 */ 2147 #define LQ_ITEM_BITS 2 2148 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 2149 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 2150 #define LINK_STATE_UNSET LQ_ITEM_MASK 2151 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 2152 #define LQ_STORE(q, i, v) \ 2153 do { \ 2154 (q) &= ~LQ_MASK((i)); \ 2155 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 2156 } while (0 /* CONSTCOND */) 2157 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 2158 #define LQ_POP(q, v) \ 2159 do { \ 2160 (v) = LQ_ITEM((q), 0); \ 2161 (q) >>= LQ_ITEM_BITS; \ 2162 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2163 } while (0 /* CONSTCOND */) 2164 #define LQ_PUSH(q, v) \ 2165 do { \ 2166 (q) >>= LQ_ITEM_BITS; \ 2167 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2168 } while (0 /* CONSTCOND */) 2169 #define LQ_FIND_UNSET(q, i) \ 2170 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 2171 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 2172 break; \ 2173 } 2174 /* 2175 * Handle a change in the interface link state and 2176 * queue notifications. 2177 */ 2178 void 2179 if_link_state_change(struct ifnet *ifp, int link_state) 2180 { 2181 int s, idx; 2182 2183 KASSERTMSG(if_is_link_state_changeable(ifp), 2184 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 2185 ifp->if_xname, ifp->if_extflags); 2186 2187 /* Ensure change is to a valid state */ 2188 switch (link_state) { 2189 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 2190 case LINK_STATE_DOWN: /* FALLTHROUGH */ 2191 case LINK_STATE_UP: 2192 break; 2193 default: 2194 #ifdef DEBUG 2195 printf("%s: invalid link state %d\n", 2196 ifp->if_xname, link_state); 2197 #endif 2198 return; 2199 } 2200 2201 s = splnet(); 2202 2203 /* Find the last unset event in the queue. */ 2204 LQ_FIND_UNSET(ifp->if_link_queue, idx); 2205 2206 /* 2207 * Ensure link_state doesn't match the last event in the queue. 2208 * ifp->if_link_state is not checked and set here because 2209 * that would present an inconsistent picture to the system. 2210 */ 2211 if (idx != 0 && 2212 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 2213 goto out; 2214 2215 /* Handle queue overflow. */ 2216 if (idx == LQ_MAX(ifp->if_link_queue)) { 2217 uint8_t lost; 2218 2219 /* 2220 * The DOWN state must be protected from being pushed off 2221 * the queue to ensure that userland will always be 2222 * in a sane state. 2223 * Because DOWN is protected, there is no need to protect 2224 * UNKNOWN. 2225 * It should be invalid to change from any other state to 2226 * UNKNOWN anyway ... 2227 */ 2228 lost = LQ_ITEM(ifp->if_link_queue, 0); 2229 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 2230 if (lost == LINK_STATE_DOWN) { 2231 lost = LQ_ITEM(ifp->if_link_queue, 0); 2232 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 2233 } 2234 printf("%s: lost link state change %s\n", 2235 ifp->if_xname, 2236 lost == LINK_STATE_UP ? "UP" : 2237 lost == LINK_STATE_DOWN ? "DOWN" : 2238 "UNKNOWN"); 2239 } else 2240 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 2241 2242 softint_schedule(ifp->if_link_si); 2243 2244 out: 2245 splx(s); 2246 } 2247 2248 /* 2249 * Handle interface link state change notifications. 2250 * Must be called at splnet(). 2251 */ 2252 static void 2253 if_link_state_change0(struct ifnet *ifp, int link_state) 2254 { 2255 struct domain *dp; 2256 2257 /* Ensure the change is still valid. */ 2258 if (ifp->if_link_state == link_state) 2259 return; 2260 2261 #ifdef DEBUG 2262 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 2263 link_state == LINK_STATE_UP ? "UP" : 2264 link_state == LINK_STATE_DOWN ? "DOWN" : 2265 "UNKNOWN", 2266 ifp->if_link_state == LINK_STATE_UP ? "UP" : 2267 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 2268 "UNKNOWN"); 2269 #endif 2270 2271 /* 2272 * When going from UNKNOWN to UP, we need to mark existing 2273 * addresses as tentative and restart DAD as we may have 2274 * erroneously not found a duplicate. 2275 * 2276 * This needs to happen before rt_ifmsg to avoid a race where 2277 * listeners would have an address and expect it to work right 2278 * away. 2279 */ 2280 if (link_state == LINK_STATE_UP && 2281 ifp->if_link_state == LINK_STATE_UNKNOWN) 2282 { 2283 DOMAIN_FOREACH(dp) { 2284 if (dp->dom_if_link_state_change != NULL) 2285 dp->dom_if_link_state_change(ifp, 2286 LINK_STATE_DOWN); 2287 } 2288 } 2289 2290 ifp->if_link_state = link_state; 2291 2292 /* Notify that the link state has changed. */ 2293 rt_ifmsg(ifp); 2294 2295 #if NCARP > 0 2296 if (ifp->if_carp) 2297 carp_carpdev_state(ifp); 2298 #endif 2299 2300 DOMAIN_FOREACH(dp) { 2301 if (dp->dom_if_link_state_change != NULL) 2302 dp->dom_if_link_state_change(ifp, link_state); 2303 } 2304 } 2305 2306 /* 2307 * Process the interface link state change queue. 2308 */ 2309 static void 2310 if_link_state_change_si(void *arg) 2311 { 2312 struct ifnet *ifp = arg; 2313 int s; 2314 uint8_t state; 2315 2316 #ifndef NET_MPSAFE 2317 mutex_enter(softnet_lock); 2318 KERNEL_LOCK(1, NULL); 2319 #endif 2320 s = splnet(); 2321 2322 /* Pop a link state change from the queue and process it. */ 2323 LQ_POP(ifp->if_link_queue, state); 2324 if_link_state_change0(ifp, state); 2325 2326 /* If there is a link state change to come, schedule it. */ 2327 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 2328 softint_schedule(ifp->if_link_si); 2329 2330 splx(s); 2331 #ifndef NET_MPSAFE 2332 KERNEL_UNLOCK_ONE(NULL); 2333 mutex_exit(softnet_lock); 2334 #endif 2335 } 2336 2337 /* 2338 * Default action when installing a local route on a point-to-point 2339 * interface. 2340 */ 2341 void 2342 p2p_rtrequest(int req, struct rtentry *rt, 2343 __unused const struct rt_addrinfo *info) 2344 { 2345 struct ifnet *ifp = rt->rt_ifp; 2346 struct ifaddr *ifa, *lo0ifa; 2347 int s = pserialize_read_enter(); 2348 2349 switch (req) { 2350 case RTM_ADD: 2351 if ((rt->rt_flags & RTF_LOCAL) == 0) 2352 break; 2353 2354 rt->rt_ifp = lo0ifp; 2355 2356 IFADDR_READER_FOREACH(ifa, ifp) { 2357 if (equal(rt_getkey(rt), ifa->ifa_addr)) 2358 break; 2359 } 2360 if (ifa == NULL) 2361 break; 2362 2363 /* 2364 * Ensure lo0 has an address of the same family. 2365 */ 2366 IFADDR_READER_FOREACH(lo0ifa, lo0ifp) { 2367 if (lo0ifa->ifa_addr->sa_family == 2368 ifa->ifa_addr->sa_family) 2369 break; 2370 } 2371 if (lo0ifa == NULL) 2372 break; 2373 2374 /* 2375 * Make sure to set rt->rt_ifa to the interface 2376 * address we are using, otherwise we will have trouble 2377 * with source address selection. 2378 */ 2379 if (ifa != rt->rt_ifa) 2380 rt_replace_ifa(rt, ifa); 2381 break; 2382 case RTM_DELETE: 2383 default: 2384 break; 2385 } 2386 pserialize_read_exit(s); 2387 } 2388 2389 /* 2390 * Mark an interface down and notify protocols of 2391 * the transition. 2392 * NOTE: must be called at splsoftnet or equivalent. 2393 */ 2394 void 2395 if_down(struct ifnet *ifp) 2396 { 2397 struct ifaddr *ifa; 2398 struct domain *dp; 2399 int s, bound; 2400 struct psref psref; 2401 2402 ifp->if_flags &= ~IFF_UP; 2403 nanotime(&ifp->if_lastchange); 2404 2405 bound = curlwp_bind(); 2406 s = pserialize_read_enter(); 2407 IFADDR_READER_FOREACH(ifa, ifp) { 2408 ifa_acquire(ifa, &psref); 2409 pserialize_read_exit(s); 2410 2411 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2412 2413 s = pserialize_read_enter(); 2414 ifa_release(ifa, &psref); 2415 } 2416 pserialize_read_exit(s); 2417 curlwp_bindx(bound); 2418 2419 IFQ_PURGE(&ifp->if_snd); 2420 #if NCARP > 0 2421 if (ifp->if_carp) 2422 carp_carpdev_state(ifp); 2423 #endif 2424 rt_ifmsg(ifp); 2425 DOMAIN_FOREACH(dp) { 2426 if (dp->dom_if_down) 2427 dp->dom_if_down(ifp); 2428 } 2429 } 2430 2431 /* 2432 * Mark an interface up and notify protocols of 2433 * the transition. 2434 * NOTE: must be called at splsoftnet or equivalent. 2435 */ 2436 void 2437 if_up(struct ifnet *ifp) 2438 { 2439 #ifdef notyet 2440 struct ifaddr *ifa; 2441 #endif 2442 struct domain *dp; 2443 2444 ifp->if_flags |= IFF_UP; 2445 nanotime(&ifp->if_lastchange); 2446 #ifdef notyet 2447 /* this has no effect on IP, and will kill all ISO connections XXX */ 2448 IFADDR_READER_FOREACH(ifa, ifp) 2449 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2450 #endif 2451 #if NCARP > 0 2452 if (ifp->if_carp) 2453 carp_carpdev_state(ifp); 2454 #endif 2455 rt_ifmsg(ifp); 2456 DOMAIN_FOREACH(dp) { 2457 if (dp->dom_if_up) 2458 dp->dom_if_up(ifp); 2459 } 2460 } 2461 2462 /* 2463 * Handle interface slowtimo timer routine. Called 2464 * from softclock, we decrement timer (if set) and 2465 * call the appropriate interface routine on expiration. 2466 */ 2467 static void 2468 if_slowtimo(void *arg) 2469 { 2470 void (*slowtimo)(struct ifnet *); 2471 struct ifnet *ifp = arg; 2472 int s; 2473 2474 slowtimo = ifp->if_slowtimo; 2475 if (__predict_false(slowtimo == NULL)) 2476 return; 2477 2478 s = splnet(); 2479 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2480 (*slowtimo)(ifp); 2481 2482 splx(s); 2483 2484 if (__predict_true(ifp->if_slowtimo != NULL)) 2485 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2486 } 2487 2488 /* 2489 * Set/clear promiscuous mode on interface ifp based on the truth value 2490 * of pswitch. The calls are reference counted so that only the first 2491 * "on" request actually has an effect, as does the final "off" request. 2492 * Results are undefined if the "off" and "on" requests are not matched. 2493 */ 2494 int 2495 ifpromisc(struct ifnet *ifp, int pswitch) 2496 { 2497 int pcount, ret; 2498 short nflags; 2499 2500 pcount = ifp->if_pcount; 2501 if (pswitch) { 2502 /* 2503 * Allow the device to be "placed" into promiscuous 2504 * mode even if it is not configured up. It will 2505 * consult IFF_PROMISC when it is brought up. 2506 */ 2507 if (ifp->if_pcount++ != 0) 2508 return 0; 2509 nflags = ifp->if_flags | IFF_PROMISC; 2510 } else { 2511 if (--ifp->if_pcount > 0) 2512 return 0; 2513 nflags = ifp->if_flags & ~IFF_PROMISC; 2514 } 2515 ret = if_flags_set(ifp, nflags); 2516 /* Restore interface state if not successful. */ 2517 if (ret != 0) { 2518 ifp->if_pcount = pcount; 2519 } 2520 return ret; 2521 } 2522 2523 /* 2524 * Map interface name to 2525 * interface structure pointer. 2526 */ 2527 struct ifnet * 2528 ifunit(const char *name) 2529 { 2530 struct ifnet *ifp; 2531 const char *cp = name; 2532 u_int unit = 0; 2533 u_int i; 2534 int s; 2535 2536 /* 2537 * If the entire name is a number, treat it as an ifindex. 2538 */ 2539 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2540 unit = unit * 10 + (*cp - '0'); 2541 } 2542 2543 /* 2544 * If the number took all of the name, then it's a valid ifindex. 2545 */ 2546 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) 2547 return if_byindex(unit); 2548 2549 ifp = NULL; 2550 s = pserialize_read_enter(); 2551 IFNET_READER_FOREACH(ifp) { 2552 if (if_is_deactivated(ifp)) 2553 continue; 2554 if (strcmp(ifp->if_xname, name) == 0) 2555 goto out; 2556 } 2557 out: 2558 pserialize_read_exit(s); 2559 return ifp; 2560 } 2561 2562 /* 2563 * Get a reference of an ifnet object by an interface name. 2564 * The returned reference is protected by psref(9). The caller 2565 * must release a returned reference by if_put after use. 2566 */ 2567 struct ifnet * 2568 if_get(const char *name, struct psref *psref) 2569 { 2570 struct ifnet *ifp; 2571 const char *cp = name; 2572 u_int unit = 0; 2573 u_int i; 2574 int s; 2575 2576 /* 2577 * If the entire name is a number, treat it as an ifindex. 2578 */ 2579 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2580 unit = unit * 10 + (*cp - '0'); 2581 } 2582 2583 /* 2584 * If the number took all of the name, then it's a valid ifindex. 2585 */ 2586 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) 2587 return if_get_byindex(unit, psref); 2588 2589 ifp = NULL; 2590 s = pserialize_read_enter(); 2591 IFNET_READER_FOREACH(ifp) { 2592 if (if_is_deactivated(ifp)) 2593 continue; 2594 if (strcmp(ifp->if_xname, name) == 0) { 2595 psref_acquire(psref, &ifp->if_psref, 2596 ifnet_psref_class); 2597 goto out; 2598 } 2599 } 2600 out: 2601 pserialize_read_exit(s); 2602 return ifp; 2603 } 2604 2605 /* 2606 * Release a reference of an ifnet object given by if_get or 2607 * if_get_byindex. 2608 */ 2609 void 2610 if_put(const struct ifnet *ifp, struct psref *psref) 2611 { 2612 2613 if (ifp == NULL) 2614 return; 2615 2616 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2617 } 2618 2619 ifnet_t * 2620 if_byindex(u_int idx) 2621 { 2622 ifnet_t *ifp; 2623 2624 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2625 if (ifp != NULL && if_is_deactivated(ifp)) 2626 ifp = NULL; 2627 return ifp; 2628 } 2629 2630 /* 2631 * Get a reference of an ifnet object by an interface index. 2632 * The returned reference is protected by psref(9). The caller 2633 * must release a returned reference by if_put after use. 2634 */ 2635 ifnet_t * 2636 if_get_byindex(u_int idx, struct psref *psref) 2637 { 2638 ifnet_t *ifp; 2639 int s; 2640 2641 s = pserialize_read_enter(); 2642 ifp = if_byindex(idx); 2643 if (__predict_true(ifp != NULL)) 2644 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2645 pserialize_read_exit(s); 2646 2647 return ifp; 2648 } 2649 2650 /* 2651 * Note that it's safe only if the passed ifp is guaranteed to not be freed, 2652 * for example using pserialize or the ifp is already held or some other 2653 * object is held which guarantes the ifp to not be freed indirectly. 2654 */ 2655 void 2656 if_acquire(struct ifnet *ifp, struct psref *psref) 2657 { 2658 2659 KASSERT(ifp->if_index != 0); 2660 psref_acquire(psref, &ifp->if_psref, ifnet_psref_class); 2661 } 2662 2663 bool 2664 if_held(struct ifnet *ifp) 2665 { 2666 2667 return psref_held(&ifp->if_psref, ifnet_psref_class); 2668 } 2669 2670 2671 /* common */ 2672 int 2673 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data) 2674 { 2675 int s; 2676 struct ifreq *ifr; 2677 struct ifcapreq *ifcr; 2678 struct ifdatareq *ifdr; 2679 2680 switch (cmd) { 2681 case SIOCSIFCAP: 2682 ifcr = data; 2683 if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0) 2684 return EINVAL; 2685 2686 if (ifcr->ifcr_capenable == ifp->if_capenable) 2687 return 0; 2688 2689 ifp->if_capenable = ifcr->ifcr_capenable; 2690 2691 /* Pre-compute the checksum flags mask. */ 2692 ifp->if_csum_flags_tx = 0; 2693 ifp->if_csum_flags_rx = 0; 2694 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) { 2695 ifp->if_csum_flags_tx |= M_CSUM_IPv4; 2696 } 2697 if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) { 2698 ifp->if_csum_flags_rx |= M_CSUM_IPv4; 2699 } 2700 2701 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) { 2702 ifp->if_csum_flags_tx |= M_CSUM_TCPv4; 2703 } 2704 if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) { 2705 ifp->if_csum_flags_rx |= M_CSUM_TCPv4; 2706 } 2707 2708 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) { 2709 ifp->if_csum_flags_tx |= M_CSUM_UDPv4; 2710 } 2711 if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) { 2712 ifp->if_csum_flags_rx |= M_CSUM_UDPv4; 2713 } 2714 2715 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) { 2716 ifp->if_csum_flags_tx |= M_CSUM_TCPv6; 2717 } 2718 if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) { 2719 ifp->if_csum_flags_rx |= M_CSUM_TCPv6; 2720 } 2721 2722 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) { 2723 ifp->if_csum_flags_tx |= M_CSUM_UDPv6; 2724 } 2725 if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) { 2726 ifp->if_csum_flags_rx |= M_CSUM_UDPv6; 2727 } 2728 if (ifp->if_flags & IFF_UP) 2729 return ENETRESET; 2730 return 0; 2731 case SIOCSIFFLAGS: 2732 ifr = data; 2733 if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) { 2734 s = splsoftnet(); 2735 if_down(ifp); 2736 splx(s); 2737 } 2738 if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { 2739 s = splsoftnet(); 2740 if_up(ifp); 2741 splx(s); 2742 } 2743 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2744 (ifr->ifr_flags &~ IFF_CANTCHANGE); 2745 break; 2746 case SIOCGIFFLAGS: 2747 ifr = data; 2748 ifr->ifr_flags = ifp->if_flags; 2749 break; 2750 2751 case SIOCGIFMETRIC: 2752 ifr = data; 2753 ifr->ifr_metric = ifp->if_metric; 2754 break; 2755 2756 case SIOCGIFMTU: 2757 ifr = data; 2758 ifr->ifr_mtu = ifp->if_mtu; 2759 break; 2760 2761 case SIOCGIFDLT: 2762 ifr = data; 2763 ifr->ifr_dlt = ifp->if_dlt; 2764 break; 2765 2766 case SIOCGIFCAP: 2767 ifcr = data; 2768 ifcr->ifcr_capabilities = ifp->if_capabilities; 2769 ifcr->ifcr_capenable = ifp->if_capenable; 2770 break; 2771 2772 case SIOCSIFMETRIC: 2773 ifr = data; 2774 ifp->if_metric = ifr->ifr_metric; 2775 break; 2776 2777 case SIOCGIFDATA: 2778 ifdr = data; 2779 ifdr->ifdr_data = ifp->if_data; 2780 break; 2781 2782 case SIOCGIFINDEX: 2783 ifr = data; 2784 ifr->ifr_index = ifp->if_index; 2785 break; 2786 2787 case SIOCZIFDATA: 2788 ifdr = data; 2789 ifdr->ifdr_data = ifp->if_data; 2790 /* 2791 * Assumes that the volatile counters that can be 2792 * zero'ed are at the end of if_data. 2793 */ 2794 memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) - 2795 offsetof(struct if_data, ifi_ipackets)); 2796 /* 2797 * The memset() clears to the bottm of if_data. In the area, 2798 * if_lastchange is included. Please be careful if new entry 2799 * will be added into if_data or rewite this. 2800 * 2801 * And also, update if_lastchnage. 2802 */ 2803 getnanotime(&ifp->if_lastchange); 2804 break; 2805 case SIOCSIFMTU: 2806 ifr = data; 2807 if (ifp->if_mtu == ifr->ifr_mtu) 2808 break; 2809 ifp->if_mtu = ifr->ifr_mtu; 2810 /* 2811 * If the link MTU changed, do network layer specific procedure. 2812 */ 2813 #ifdef INET6 2814 if (in6_present) 2815 nd6_setmtu(ifp); 2816 #endif 2817 return ENETRESET; 2818 default: 2819 return ENOTTY; 2820 } 2821 return 0; 2822 } 2823 2824 int 2825 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 2826 { 2827 struct if_addrprefreq *ifap = (struct if_addrprefreq *)data; 2828 struct ifaddr *ifa; 2829 const struct sockaddr *any, *sa; 2830 union { 2831 struct sockaddr sa; 2832 struct sockaddr_storage ss; 2833 } u, v; 2834 int s, error = 0; 2835 2836 switch (cmd) { 2837 case SIOCSIFADDRPREF: 2838 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 2839 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2840 NULL) != 0) 2841 return EPERM; 2842 case SIOCGIFADDRPREF: 2843 break; 2844 default: 2845 return EOPNOTSUPP; 2846 } 2847 2848 /* sanity checks */ 2849 if (data == NULL || ifp == NULL) { 2850 panic("invalid argument to %s", __func__); 2851 /*NOTREACHED*/ 2852 } 2853 2854 /* address must be specified on ADD and DELETE */ 2855 sa = sstocsa(&ifap->ifap_addr); 2856 if (sa->sa_family != sofamily(so)) 2857 return EINVAL; 2858 if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len) 2859 return EINVAL; 2860 2861 sockaddr_externalize(&v.sa, sizeof(v.ss), sa); 2862 2863 s = pserialize_read_enter(); 2864 IFADDR_READER_FOREACH(ifa, ifp) { 2865 if (ifa->ifa_addr->sa_family != sa->sa_family) 2866 continue; 2867 sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr); 2868 if (sockaddr_cmp(&u.sa, &v.sa) == 0) 2869 break; 2870 } 2871 if (ifa == NULL) { 2872 error = EADDRNOTAVAIL; 2873 goto out; 2874 } 2875 2876 switch (cmd) { 2877 case SIOCSIFADDRPREF: 2878 ifa->ifa_preference = ifap->ifap_preference; 2879 goto out; 2880 case SIOCGIFADDRPREF: 2881 /* fill in the if_laddrreq structure */ 2882 (void)sockaddr_copy(sstosa(&ifap->ifap_addr), 2883 sizeof(ifap->ifap_addr), ifa->ifa_addr); 2884 ifap->ifap_preference = ifa->ifa_preference; 2885 goto out; 2886 default: 2887 error = EOPNOTSUPP; 2888 } 2889 out: 2890 pserialize_read_exit(s); 2891 return error; 2892 } 2893 2894 /* 2895 * Interface ioctls. 2896 */ 2897 static int 2898 doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l) 2899 { 2900 struct ifnet *ifp; 2901 struct ifreq *ifr; 2902 int error = 0; 2903 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ) 2904 u_long ocmd = cmd; 2905 #endif 2906 short oif_flags; 2907 #ifdef COMPAT_OIFREQ 2908 struct ifreq ifrb; 2909 struct oifreq *oifr = NULL; 2910 #endif 2911 int r; 2912 struct psref psref; 2913 int bound; 2914 2915 switch (cmd) { 2916 #ifdef COMPAT_OIFREQ 2917 case OSIOCGIFCONF: 2918 case OOSIOCGIFCONF: 2919 return compat_ifconf(cmd, data); 2920 #endif 2921 #ifdef COMPAT_OIFDATA 2922 case OSIOCGIFDATA: 2923 case OSIOCZIFDATA: 2924 return compat_ifdatareq(l, cmd, data); 2925 #endif 2926 case SIOCGIFCONF: 2927 return ifconf(cmd, data); 2928 case SIOCINITIFADDR: 2929 return EPERM; 2930 } 2931 2932 #ifdef COMPAT_OIFREQ 2933 cmd = (*vec_compat_cvtcmd)(cmd); 2934 if (cmd != ocmd) { 2935 oifr = data; 2936 data = ifr = &ifrb; 2937 ifreqo2n(oifr, ifr); 2938 } else 2939 #endif 2940 ifr = data; 2941 2942 switch (cmd) { 2943 case SIOCIFCREATE: 2944 case SIOCIFDESTROY: 2945 bound = curlwp_bind(); 2946 if (l != NULL) { 2947 ifp = if_get(ifr->ifr_name, &psref); 2948 error = kauth_authorize_network(l->l_cred, 2949 KAUTH_NETWORK_INTERFACE, 2950 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 2951 (void *)cmd, NULL); 2952 if (ifp != NULL) 2953 if_put(ifp, &psref); 2954 if (error != 0) { 2955 curlwp_bindx(bound); 2956 return error; 2957 } 2958 } 2959 mutex_enter(&if_clone_mtx); 2960 r = (cmd == SIOCIFCREATE) ? 2961 if_clone_create(ifr->ifr_name) : 2962 if_clone_destroy(ifr->ifr_name); 2963 mutex_exit(&if_clone_mtx); 2964 curlwp_bindx(bound); 2965 return r; 2966 2967 case SIOCIFGCLONERS: 2968 { 2969 struct if_clonereq *req = (struct if_clonereq *)data; 2970 return if_clone_list(req->ifcr_count, req->ifcr_buffer, 2971 &req->ifcr_total); 2972 } 2973 } 2974 2975 bound = curlwp_bind(); 2976 ifp = if_get(ifr->ifr_name, &psref); 2977 if (ifp == NULL) { 2978 curlwp_bindx(bound); 2979 return ENXIO; 2980 } 2981 2982 switch (cmd) { 2983 case SIOCALIFADDR: 2984 case SIOCDLIFADDR: 2985 case SIOCSIFADDRPREF: 2986 case SIOCSIFFLAGS: 2987 case SIOCSIFCAP: 2988 case SIOCSIFMETRIC: 2989 case SIOCZIFDATA: 2990 case SIOCSIFMTU: 2991 case SIOCSIFPHYADDR: 2992 case SIOCDIFPHYADDR: 2993 #ifdef INET6 2994 case SIOCSIFPHYADDR_IN6: 2995 #endif 2996 case SIOCSLIFPHYADDR: 2997 case SIOCADDMULTI: 2998 case SIOCDELMULTI: 2999 case SIOCSIFMEDIA: 3000 case SIOCSDRVSPEC: 3001 case SIOCG80211: 3002 case SIOCS80211: 3003 case SIOCS80211NWID: 3004 case SIOCS80211NWKEY: 3005 case SIOCS80211POWER: 3006 case SIOCS80211BSSID: 3007 case SIOCS80211CHANNEL: 3008 case SIOCSLINKSTR: 3009 if (l != NULL) { 3010 error = kauth_authorize_network(l->l_cred, 3011 KAUTH_NETWORK_INTERFACE, 3012 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, 3013 (void *)cmd, NULL); 3014 if (error != 0) 3015 goto out; 3016 } 3017 } 3018 3019 oif_flags = ifp->if_flags; 3020 3021 mutex_enter(ifp->if_ioctl_lock); 3022 3023 error = (*ifp->if_ioctl)(ifp, cmd, data); 3024 if (error != ENOTTY) 3025 ; 3026 else if (so->so_proto == NULL) 3027 error = EOPNOTSUPP; 3028 else { 3029 #ifdef COMPAT_OSOCK 3030 if (vec_compat_ifioctl != NULL) 3031 error = (*vec_compat_ifioctl)(so, ocmd, cmd, data, l); 3032 else 3033 #endif 3034 error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so, 3035 cmd, data, ifp); 3036 } 3037 3038 if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) { 3039 if ((ifp->if_flags & IFF_UP) != 0) { 3040 int s = splsoftnet(); 3041 if_up(ifp); 3042 splx(s); 3043 } 3044 } 3045 #ifdef COMPAT_OIFREQ 3046 if (cmd != ocmd) 3047 ifreqn2o(oifr, ifr); 3048 #endif 3049 3050 mutex_exit(ifp->if_ioctl_lock); 3051 out: 3052 if_put(ifp, &psref); 3053 curlwp_bindx(bound); 3054 return error; 3055 } 3056 3057 /* 3058 * Return interface configuration 3059 * of system. List may be used 3060 * in later ioctl's (above) to get 3061 * other information. 3062 * 3063 * Each record is a struct ifreq. Before the addition of 3064 * sockaddr_storage, the API rule was that sockaddr flavors that did 3065 * not fit would extend beyond the struct ifreq, with the next struct 3066 * ifreq starting sa_len beyond the struct sockaddr. Because the 3067 * union in struct ifreq includes struct sockaddr_storage, every kind 3068 * of sockaddr must fit. Thus, there are no longer any overlength 3069 * records. 3070 * 3071 * Records are added to the user buffer if they fit, and ifc_len is 3072 * adjusted to the length that was written. Thus, the user is only 3073 * assured of getting the complete list if ifc_len on return is at 3074 * least sizeof(struct ifreq) less than it was on entry. 3075 * 3076 * If the user buffer pointer is NULL, this routine copies no data and 3077 * returns the amount of space that would be needed. 3078 * 3079 * Invariants: 3080 * ifrp points to the next part of the user's buffer to be used. If 3081 * ifrp != NULL, space holds the number of bytes remaining that we may 3082 * write at ifrp. Otherwise, space holds the number of bytes that 3083 * would have been written had there been adequate space. 3084 */ 3085 /*ARGSUSED*/ 3086 static int 3087 ifconf(u_long cmd, void *data) 3088 { 3089 struct ifconf *ifc = (struct ifconf *)data; 3090 struct ifnet *ifp; 3091 struct ifaddr *ifa; 3092 struct ifreq ifr, *ifrp = NULL; 3093 int space = 0, error = 0; 3094 const int sz = (int)sizeof(struct ifreq); 3095 const bool docopy = ifc->ifc_req != NULL; 3096 int s; 3097 int bound; 3098 struct psref psref; 3099 3100 if (docopy) { 3101 space = ifc->ifc_len; 3102 ifrp = ifc->ifc_req; 3103 } 3104 3105 bound = curlwp_bind(); 3106 s = pserialize_read_enter(); 3107 IFNET_READER_FOREACH(ifp) { 3108 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 3109 pserialize_read_exit(s); 3110 3111 (void)strncpy(ifr.ifr_name, ifp->if_xname, 3112 sizeof(ifr.ifr_name)); 3113 if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') { 3114 error = ENAMETOOLONG; 3115 goto release_exit; 3116 } 3117 if (IFADDR_READER_EMPTY(ifp)) { 3118 /* Interface with no addresses - send zero sockaddr. */ 3119 memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr)); 3120 if (!docopy) { 3121 space += sz; 3122 continue; 3123 } 3124 if (space >= sz) { 3125 error = copyout(&ifr, ifrp, sz); 3126 if (error != 0) 3127 goto release_exit; 3128 ifrp++; 3129 space -= sz; 3130 } 3131 } 3132 3133 IFADDR_READER_FOREACH(ifa, ifp) { 3134 struct sockaddr *sa = ifa->ifa_addr; 3135 /* all sockaddrs must fit in sockaddr_storage */ 3136 KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru)); 3137 3138 if (!docopy) { 3139 space += sz; 3140 continue; 3141 } 3142 memcpy(&ifr.ifr_space, sa, sa->sa_len); 3143 if (space >= sz) { 3144 error = copyout(&ifr, ifrp, sz); 3145 if (error != 0) 3146 goto release_exit; 3147 ifrp++; space -= sz; 3148 } 3149 } 3150 3151 s = pserialize_read_enter(); 3152 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3153 } 3154 pserialize_read_exit(s); 3155 curlwp_bindx(bound); 3156 3157 if (docopy) { 3158 KASSERT(0 <= space && space <= ifc->ifc_len); 3159 ifc->ifc_len -= space; 3160 } else { 3161 KASSERT(space >= 0); 3162 ifc->ifc_len = space; 3163 } 3164 return (0); 3165 3166 release_exit: 3167 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 3168 curlwp_bindx(bound); 3169 return error; 3170 } 3171 3172 int 3173 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa) 3174 { 3175 uint8_t len; 3176 #ifdef COMPAT_OIFREQ 3177 struct ifreq ifrb; 3178 struct oifreq *oifr = NULL; 3179 u_long ocmd = cmd; 3180 cmd = (*vec_compat_cvtcmd)(cmd); 3181 if (cmd != ocmd) { 3182 oifr = (struct oifreq *)(void *)ifr; 3183 ifr = &ifrb; 3184 ifreqo2n(oifr, ifr); 3185 len = sizeof(oifr->ifr_addr); 3186 } else 3187 #endif 3188 len = sizeof(ifr->ifr_ifru.ifru_space); 3189 3190 if (len < sa->sa_len) 3191 return EFBIG; 3192 3193 memset(&ifr->ifr_addr, 0, len); 3194 sockaddr_copy(&ifr->ifr_addr, len, sa); 3195 3196 #ifdef COMPAT_OIFREQ 3197 if (cmd != ocmd) 3198 ifreqn2o(oifr, ifr); 3199 #endif 3200 return 0; 3201 } 3202 3203 /* 3204 * wrapper function for the drivers which doesn't have if_transmit(). 3205 */ 3206 static int 3207 if_transmit(struct ifnet *ifp, struct mbuf *m) 3208 { 3209 int s, error; 3210 size_t pktlen = m->m_pkthdr.len; 3211 bool mcast = (m->m_flags & M_MCAST) != 0; 3212 3213 s = splnet(); 3214 3215 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3216 if (error != 0) { 3217 /* mbuf is already freed */ 3218 goto out; 3219 } 3220 3221 ifp->if_obytes += pktlen; 3222 if (mcast) 3223 ifp->if_omcasts++; 3224 3225 if ((ifp->if_flags & IFF_OACTIVE) == 0) 3226 if_start_lock(ifp); 3227 out: 3228 splx(s); 3229 3230 return error; 3231 } 3232 3233 int 3234 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 3235 { 3236 int error; 3237 3238 #ifdef ALTQ 3239 KERNEL_LOCK(1, NULL); 3240 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 3241 error = if_transmit(ifp, m); 3242 KERNEL_UNLOCK_ONE(NULL); 3243 } else { 3244 KERNEL_UNLOCK_ONE(NULL); 3245 error = (*ifp->if_transmit)(ifp, m); 3246 /* mbuf is alredy freed */ 3247 } 3248 #else /* !ALTQ */ 3249 error = (*ifp->if_transmit)(ifp, m); 3250 /* mbuf is alredy freed */ 3251 #endif /* !ALTQ */ 3252 3253 return error; 3254 } 3255 3256 /* 3257 * Queue message on interface, and start output if interface 3258 * not yet active. 3259 */ 3260 int 3261 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 3262 { 3263 3264 return if_transmit_lock(ifp, m); 3265 } 3266 3267 /* 3268 * Queue message on interface, possibly using a second fast queue 3269 */ 3270 int 3271 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 3272 { 3273 int error = 0; 3274 3275 if (ifq != NULL 3276 #ifdef ALTQ 3277 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 3278 #endif 3279 ) { 3280 if (IF_QFULL(ifq)) { 3281 IF_DROP(&ifp->if_snd); 3282 m_freem(m); 3283 if (error == 0) 3284 error = ENOBUFS; 3285 } else 3286 IF_ENQUEUE(ifq, m); 3287 } else 3288 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3289 if (error != 0) { 3290 ++ifp->if_oerrors; 3291 return error; 3292 } 3293 return 0; 3294 } 3295 3296 int 3297 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 3298 { 3299 int rc; 3300 3301 if (ifp->if_initaddr != NULL) 3302 rc = (*ifp->if_initaddr)(ifp, ifa, src); 3303 else if (src || 3304 /* FIXME: may not hold if_ioctl_lock */ 3305 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 3306 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 3307 3308 return rc; 3309 } 3310 3311 int 3312 if_do_dad(struct ifnet *ifp) 3313 { 3314 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 3315 return 0; 3316 3317 switch (ifp->if_type) { 3318 case IFT_FAITH: 3319 /* 3320 * These interfaces do not have the IFF_LOOPBACK flag, 3321 * but loop packets back. We do not have to do DAD on such 3322 * interfaces. We should even omit it, because loop-backed 3323 * responses would confuse the DAD procedure. 3324 */ 3325 return 0; 3326 default: 3327 /* 3328 * Our DAD routine requires the interface up and running. 3329 * However, some interfaces can be up before the RUNNING 3330 * status. Additionaly, users may try to assign addresses 3331 * before the interface becomes up (or running). 3332 * We simply skip DAD in such a case as a work around. 3333 * XXX: we should rather mark "tentative" on such addresses, 3334 * and do DAD after the interface becomes ready. 3335 */ 3336 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 3337 (IFF_UP|IFF_RUNNING)) 3338 return 0; 3339 3340 return 1; 3341 } 3342 } 3343 3344 int 3345 if_flags_set(ifnet_t *ifp, const short flags) 3346 { 3347 int rc; 3348 3349 if (ifp->if_setflags != NULL) 3350 rc = (*ifp->if_setflags)(ifp, flags); 3351 else { 3352 short cantflags, chgdflags; 3353 struct ifreq ifr; 3354 3355 chgdflags = ifp->if_flags ^ flags; 3356 cantflags = chgdflags & IFF_CANTCHANGE; 3357 3358 if (cantflags != 0) 3359 ifp->if_flags ^= cantflags; 3360 3361 /* Traditionally, we do not call if_ioctl after 3362 * setting/clearing only IFF_PROMISC if the interface 3363 * isn't IFF_UP. Uphold that tradition. 3364 */ 3365 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 3366 return 0; 3367 3368 memset(&ifr, 0, sizeof(ifr)); 3369 3370 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 3371 /* FIXME: may not hold if_ioctl_lock */ 3372 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 3373 3374 if (rc != 0 && cantflags != 0) 3375 ifp->if_flags ^= cantflags; 3376 } 3377 3378 return rc; 3379 } 3380 3381 int 3382 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 3383 { 3384 int rc; 3385 struct ifreq ifr; 3386 3387 if (ifp->if_mcastop != NULL) 3388 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 3389 else { 3390 ifreq_setaddr(cmd, &ifr, sa); 3391 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 3392 } 3393 3394 return rc; 3395 } 3396 3397 static void 3398 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 3399 struct ifaltq *ifq) 3400 { 3401 const struct sysctlnode *cnode, *rnode; 3402 3403 if (sysctl_createv(clog, 0, NULL, &rnode, 3404 CTLFLAG_PERMANENT, 3405 CTLTYPE_NODE, "interfaces", 3406 SYSCTL_DESCR("Per-interface controls"), 3407 NULL, 0, NULL, 0, 3408 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3409 goto bad; 3410 3411 if (sysctl_createv(clog, 0, &rnode, &rnode, 3412 CTLFLAG_PERMANENT, 3413 CTLTYPE_NODE, ifname, 3414 SYSCTL_DESCR("Interface controls"), 3415 NULL, 0, NULL, 0, 3416 CTL_CREATE, CTL_EOL) != 0) 3417 goto bad; 3418 3419 if (sysctl_createv(clog, 0, &rnode, &rnode, 3420 CTLFLAG_PERMANENT, 3421 CTLTYPE_NODE, "sndq", 3422 SYSCTL_DESCR("Interface output queue controls"), 3423 NULL, 0, NULL, 0, 3424 CTL_CREATE, CTL_EOL) != 0) 3425 goto bad; 3426 3427 if (sysctl_createv(clog, 0, &rnode, &cnode, 3428 CTLFLAG_PERMANENT, 3429 CTLTYPE_INT, "len", 3430 SYSCTL_DESCR("Current output queue length"), 3431 NULL, 0, &ifq->ifq_len, 0, 3432 CTL_CREATE, CTL_EOL) != 0) 3433 goto bad; 3434 3435 if (sysctl_createv(clog, 0, &rnode, &cnode, 3436 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3437 CTLTYPE_INT, "maxlen", 3438 SYSCTL_DESCR("Maximum allowed output queue length"), 3439 NULL, 0, &ifq->ifq_maxlen, 0, 3440 CTL_CREATE, CTL_EOL) != 0) 3441 goto bad; 3442 3443 if (sysctl_createv(clog, 0, &rnode, &cnode, 3444 CTLFLAG_PERMANENT, 3445 CTLTYPE_INT, "drops", 3446 SYSCTL_DESCR("Packets dropped due to full output queue"), 3447 NULL, 0, &ifq->ifq_drops, 0, 3448 CTL_CREATE, CTL_EOL) != 0) 3449 goto bad; 3450 3451 return; 3452 bad: 3453 printf("%s: could not attach sysctl nodes\n", ifname); 3454 return; 3455 } 3456 3457 #if defined(INET) || defined(INET6) 3458 3459 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3460 static int \ 3461 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3462 { \ 3463 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3464 } 3465 3466 #if defined(INET) 3467 static int 3468 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3469 { 3470 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3471 } 3472 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3473 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3474 #endif 3475 3476 #if defined(INET6) 3477 static int 3478 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3479 { 3480 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3481 } 3482 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3483 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3484 #endif 3485 3486 static void 3487 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3488 { 3489 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3490 const char *pfname = NULL, *ipname = NULL; 3491 int ipn = 0, qid = 0; 3492 3493 switch (pf) { 3494 #if defined(INET) 3495 case PF_INET: 3496 len_func = sysctl_net_ip_pktq_items; 3497 maxlen_func = sysctl_net_ip_pktq_maxlen; 3498 drops_func = sysctl_net_ip_pktq_drops; 3499 pfname = "inet", ipn = IPPROTO_IP; 3500 ipname = "ip", qid = IPCTL_IFQ; 3501 break; 3502 #endif 3503 #if defined(INET6) 3504 case PF_INET6: 3505 len_func = sysctl_net_ip6_pktq_items; 3506 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3507 drops_func = sysctl_net_ip6_pktq_drops; 3508 pfname = "inet6", ipn = IPPROTO_IPV6; 3509 ipname = "ip6", qid = IPV6CTL_IFQ; 3510 break; 3511 #endif 3512 default: 3513 KASSERT(false); 3514 } 3515 3516 sysctl_createv(clog, 0, NULL, NULL, 3517 CTLFLAG_PERMANENT, 3518 CTLTYPE_NODE, pfname, NULL, 3519 NULL, 0, NULL, 0, 3520 CTL_NET, pf, CTL_EOL); 3521 sysctl_createv(clog, 0, NULL, NULL, 3522 CTLFLAG_PERMANENT, 3523 CTLTYPE_NODE, ipname, NULL, 3524 NULL, 0, NULL, 0, 3525 CTL_NET, pf, ipn, CTL_EOL); 3526 sysctl_createv(clog, 0, NULL, NULL, 3527 CTLFLAG_PERMANENT, 3528 CTLTYPE_NODE, "ifq", 3529 SYSCTL_DESCR("Protocol input queue controls"), 3530 NULL, 0, NULL, 0, 3531 CTL_NET, pf, ipn, qid, CTL_EOL); 3532 3533 sysctl_createv(clog, 0, NULL, NULL, 3534 CTLFLAG_PERMANENT, 3535 CTLTYPE_INT, "len", 3536 SYSCTL_DESCR("Current input queue length"), 3537 len_func, 0, NULL, 0, 3538 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3539 sysctl_createv(clog, 0, NULL, NULL, 3540 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3541 CTLTYPE_INT, "maxlen", 3542 SYSCTL_DESCR("Maximum allowed input queue length"), 3543 maxlen_func, 0, NULL, 0, 3544 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3545 sysctl_createv(clog, 0, NULL, NULL, 3546 CTLFLAG_PERMANENT, 3547 CTLTYPE_INT, "drops", 3548 SYSCTL_DESCR("Packets dropped due to full input queue"), 3549 drops_func, 0, NULL, 0, 3550 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3551 } 3552 #endif /* INET || INET6 */ 3553 3554 static int 3555 if_sdl_sysctl(SYSCTLFN_ARGS) 3556 { 3557 struct ifnet *ifp; 3558 const struct sockaddr_dl *sdl; 3559 struct psref psref; 3560 int error = 0; 3561 int bound; 3562 3563 if (namelen != 1) 3564 return EINVAL; 3565 3566 bound = curlwp_bind(); 3567 ifp = if_get_byindex(name[0], &psref); 3568 if (ifp == NULL) { 3569 error = ENODEV; 3570 goto out0; 3571 } 3572 3573 sdl = ifp->if_sadl; 3574 if (sdl == NULL) { 3575 *oldlenp = 0; 3576 goto out1; 3577 } 3578 3579 if (oldp == NULL) { 3580 *oldlenp = sdl->sdl_alen; 3581 goto out1; 3582 } 3583 3584 if (*oldlenp >= sdl->sdl_alen) 3585 *oldlenp = sdl->sdl_alen; 3586 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3587 out1: 3588 if_put(ifp, &psref); 3589 out0: 3590 curlwp_bindx(bound); 3591 return error; 3592 } 3593 3594 static void 3595 if_sysctl_setup(struct sysctllog **clog) 3596 { 3597 const struct sysctlnode *rnode = NULL; 3598 3599 sysctl_createv(clog, 0, NULL, &rnode, 3600 CTLFLAG_PERMANENT, 3601 CTLTYPE_NODE, "sdl", 3602 SYSCTL_DESCR("Get active link-layer address"), 3603 if_sdl_sysctl, 0, NULL, 0, 3604 CTL_NET, CTL_CREATE, CTL_EOL); 3605 3606 #if defined(INET) 3607 sysctl_net_pktq_setup(NULL, PF_INET); 3608 #endif 3609 #ifdef INET6 3610 if (in6_present) 3611 sysctl_net_pktq_setup(NULL, PF_INET6); 3612 #endif 3613 } 3614
Indexes created Wed Sep 24 14:09:57 GMT 2025