if.c revision 1.378
1 /* $NetBSD: if.c,v 1.378 2017/02/15 01:48:44 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 1.378 2017/02/15 01:48:44 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 #ifndef IPSEC 142 #include <netinet/ip_encap.h> 143 #endif 144 #include <net/bpf.h> 145 146 #ifdef INET6 147 #include <netinet6/in6_var.h> 148 #include <netinet6/nd6.h> 149 #endif 150 151 #include "ether.h" 152 #include "fddi.h" 153 #include "token.h" 154 155 #include "carp.h" 156 #if NCARP > 0 157 #include <netinet/ip_carp.h> 158 #endif 159 160 #include <compat/sys/sockio.h> 161 #include <compat/sys/socket.h> 162 163 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 164 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 165 166 /* 167 * Global list of interfaces. 168 */ 169 /* DEPRECATED. Remove it once kvm(3) users disappeared */ 170 struct ifnet_head ifnet_list; 171 172 struct pslist_head ifnet_pslist; 173 static ifnet_t ** ifindex2ifnet = NULL; 174 static u_int if_index = 1; 175 static size_t if_indexlim = 0; 176 static uint64_t index_gen; 177 /* Mutex to protect the above objects. */ 178 kmutex_t ifnet_mtx __cacheline_aligned; 179 struct psref_class *ifnet_psref_class __read_mostly; 180 static pserialize_t ifnet_psz; 181 182 static kmutex_t if_clone_mtx; 183 184 struct ifnet *lo0ifp; 185 int ifqmaxlen = IFQ_MAXLEN; 186 187 struct psref_class *ifa_psref_class __read_mostly; 188 189 static int if_delroute_matcher(struct rtentry *, void *); 190 191 static struct if_clone *if_clone_lookup(const char *, int *); 192 193 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners); 194 static int if_cloners_count; 195 196 /* Packet filtering hook for interfaces. */ 197 pfil_head_t * if_pfil __read_mostly; 198 199 static kauth_listener_t if_listener; 200 201 static int doifioctl(struct socket *, u_long, void *, struct lwp *); 202 static void if_detach_queues(struct ifnet *, struct ifqueue *); 203 static void sysctl_sndq_setup(struct sysctllog **, const char *, 204 struct ifaltq *); 205 static void if_slowtimo(void *); 206 static void if_free_sadl(struct ifnet *); 207 static void if_attachdomain1(struct ifnet *); 208 static int ifconf(u_long, void *); 209 static int if_transmit(struct ifnet *, struct mbuf *); 210 static int if_clone_create(const char *); 211 static int if_clone_destroy(const char *); 212 static void if_link_state_change_si(void *); 213 214 struct if_percpuq { 215 struct ifnet *ipq_ifp; 216 void *ipq_si; 217 struct percpu *ipq_ifqs; /* struct ifqueue */ 218 }; 219 220 static struct mbuf *if_percpuq_dequeue(struct if_percpuq *); 221 222 static void if_percpuq_drops(void *, void *, struct cpu_info *); 223 static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO); 224 static void sysctl_percpuq_setup(struct sysctllog **, const char *, 225 struct if_percpuq *); 226 227 struct if_deferred_start { 228 struct ifnet *ids_ifp; 229 void (*ids_if_start)(struct ifnet *); 230 void *ids_si; 231 }; 232 233 static void if_deferred_start_softint(void *); 234 static void if_deferred_start_common(struct ifnet *); 235 static void if_deferred_start_destroy(struct ifnet *); 236 237 #if defined(INET) || defined(INET6) 238 static void sysctl_net_pktq_setup(struct sysctllog **, int); 239 #endif 240 241 static void if_sysctl_setup(struct sysctllog **); 242 243 /* 244 * Pointer to stub or real compat_cvtcmd() depending on presence of 245 * the compat module 246 */ 247 u_long stub_compat_cvtcmd(u_long); 248 u_long (*vec_compat_cvtcmd)(u_long) = stub_compat_cvtcmd; 249 250 /* Similarly, pointer to compat_ifioctl() if it is present */ 251 252 int (*vec_compat_ifioctl)(struct socket *, u_long, u_long, void *, 253 struct lwp *) = NULL; 254 255 /* The stub version of compat_cvtcmd() */ 256 u_long stub_compat_cvtcmd(u_long cmd) 257 { 258 259 return cmd; 260 } 261 262 static int 263 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 264 void *arg0, void *arg1, void *arg2, void *arg3) 265 { 266 int result; 267 enum kauth_network_req req; 268 269 result = KAUTH_RESULT_DEFER; 270 req = (enum kauth_network_req)arg1; 271 272 if (action != KAUTH_NETWORK_INTERFACE) 273 return result; 274 275 if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) || 276 (req == KAUTH_REQ_NETWORK_INTERFACE_SET)) 277 result = KAUTH_RESULT_ALLOW; 278 279 return result; 280 } 281 282 /* 283 * Network interface utility routines. 284 * 285 * Routines with ifa_ifwith* names take sockaddr *'s as 286 * parameters. 287 */ 288 void 289 ifinit(void) 290 { 291 292 if_sysctl_setup(NULL); 293 294 #if (defined(INET) || defined(INET6)) && !defined(IPSEC) 295 encapinit(); 296 #endif 297 298 if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, 299 if_listener_cb, NULL); 300 301 /* interfaces are available, inform socket code */ 302 ifioctl = doifioctl; 303 } 304 305 /* 306 * XXX Initialization before configure(). 307 * XXX hack to get pfil_add_hook working in autoconf. 308 */ 309 void 310 ifinit1(void) 311 { 312 mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE); 313 314 TAILQ_INIT(&ifnet_list); 315 mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE); 316 ifnet_psz = pserialize_create(); 317 ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET); 318 ifa_psref_class = psref_class_create("ifa", IPL_SOFTNET); 319 PSLIST_INIT(&ifnet_pslist); 320 321 if_indexlim = 8; 322 323 if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL); 324 KASSERT(if_pfil != NULL); 325 326 #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN) 327 etherinit(); 328 #endif 329 } 330 331 ifnet_t * 332 if_alloc(u_char type) 333 { 334 return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP); 335 } 336 337 void 338 if_free(ifnet_t *ifp) 339 { 340 kmem_free(ifp, sizeof(ifnet_t)); 341 } 342 343 void 344 if_initname(struct ifnet *ifp, const char *name, int unit) 345 { 346 (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname), 347 "%s%d", name, unit); 348 } 349 350 /* 351 * Null routines used while an interface is going away. These routines 352 * just return an error. 353 */ 354 355 int 356 if_nulloutput(struct ifnet *ifp, struct mbuf *m, 357 const struct sockaddr *so, const struct rtentry *rt) 358 { 359 360 return ENXIO; 361 } 362 363 void 364 if_nullinput(struct ifnet *ifp, struct mbuf *m) 365 { 366 367 /* Nothing. */ 368 } 369 370 void 371 if_nullstart(struct ifnet *ifp) 372 { 373 374 /* Nothing. */ 375 } 376 377 int 378 if_nulltransmit(struct ifnet *ifp, struct mbuf *m) 379 { 380 381 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); 722 723 IFNET_LOCK(); 724 if_getindex(ifp); 725 IFNET_UNLOCK(); 726 } 727 728 /* 729 * Register an interface to the list of "active" interfaces. 730 */ 731 void 732 if_register(ifnet_t *ifp) 733 { 734 /* 735 * If the driver has not supplied its own if_ioctl, then 736 * supply the default. 737 */ 738 if (ifp->if_ioctl == NULL) 739 ifp->if_ioctl = ifioctl_common; 740 741 sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd); 742 743 if (!STAILQ_EMPTY(&domains)) 744 if_attachdomain1(ifp); 745 746 /* Announce the interface. */ 747 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 748 749 if (ifp->if_slowtimo != NULL) { 750 ifp->if_slowtimo_ch = 751 kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP); 752 callout_init(ifp->if_slowtimo_ch, 0); 753 callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp); 754 if_slowtimo(ifp); 755 } 756 757 if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit) 758 ifp->if_transmit = if_transmit; 759 760 IFNET_LOCK(); 761 TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list); 762 IFNET_WRITER_INSERT_TAIL(ifp); 763 IFNET_UNLOCK(); 764 } 765 766 /* 767 * The if_percpuq framework 768 * 769 * It allows network device drivers to execute the network stack 770 * in softint (so called softint-based if_input). It utilizes 771 * softint and percpu ifqueue. It doesn't distribute any packets 772 * between CPUs, unlike pktqueue(9). 773 * 774 * Currently we support two options for device drivers to apply the framework: 775 * - Use it implicitly with less changes 776 * - If you use if_attach in driver's _attach function and if_input in 777 * driver's Rx interrupt handler, a packet is queued and a softint handles 778 * the packet implicitly 779 * - Use it explicitly in each driver (recommended) 780 * - You can use if_percpuq_* directly in your driver 781 * - In this case, you need to allocate struct if_percpuq in driver's softc 782 * - See wm(4) as a reference implementation 783 */ 784 785 static void 786 if_percpuq_softint(void *arg) 787 { 788 struct if_percpuq *ipq = arg; 789 struct ifnet *ifp = ipq->ipq_ifp; 790 struct mbuf *m; 791 792 while ((m = if_percpuq_dequeue(ipq)) != NULL) { 793 ifp->if_ipackets++; 794 bpf_mtap(ifp, m); 795 796 ifp->_if_input(ifp, m); 797 } 798 } 799 800 static void 801 if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 802 { 803 struct ifqueue *const ifq = p; 804 805 memset(ifq, 0, sizeof(*ifq)); 806 ifq->ifq_maxlen = IFQ_MAXLEN; 807 } 808 809 struct if_percpuq * 810 if_percpuq_create(struct ifnet *ifp) 811 { 812 struct if_percpuq *ipq; 813 814 ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP); 815 if (ipq == NULL) 816 panic("kmem_zalloc failed"); 817 818 ipq->ipq_ifp = ifp; 819 ipq->ipq_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 820 if_percpuq_softint, ipq); 821 ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue)); 822 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL); 823 824 sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq); 825 826 return ipq; 827 } 828 829 static struct mbuf * 830 if_percpuq_dequeue(struct if_percpuq *ipq) 831 { 832 struct mbuf *m; 833 struct ifqueue *ifq; 834 int s; 835 836 s = splnet(); 837 ifq = percpu_getref(ipq->ipq_ifqs); 838 IF_DEQUEUE(ifq, m); 839 percpu_putref(ipq->ipq_ifqs); 840 splx(s); 841 842 return m; 843 } 844 845 static void 846 if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused) 847 { 848 struct ifqueue *const ifq = p; 849 850 IF_PURGE(ifq); 851 } 852 853 void 854 if_percpuq_destroy(struct if_percpuq *ipq) 855 { 856 857 /* if_detach may already destroy it */ 858 if (ipq == NULL) 859 return; 860 861 softint_disestablish(ipq->ipq_si); 862 percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL); 863 percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue)); 864 kmem_free(ipq, sizeof(*ipq)); 865 } 866 867 void 868 if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m) 869 { 870 struct ifqueue *ifq; 871 int s; 872 873 KASSERT(ipq != NULL); 874 875 s = splnet(); 876 ifq = percpu_getref(ipq->ipq_ifqs); 877 if (IF_QFULL(ifq)) { 878 IF_DROP(ifq); 879 percpu_putref(ipq->ipq_ifqs); 880 m_freem(m); 881 goto out; 882 } 883 IF_ENQUEUE(ifq, m); 884 percpu_putref(ipq->ipq_ifqs); 885 886 softint_schedule(ipq->ipq_si); 887 out: 888 splx(s); 889 } 890 891 static void 892 if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused) 893 { 894 struct ifqueue *const ifq = p; 895 int *sum = arg; 896 897 *sum += ifq->ifq_drops; 898 } 899 900 static int 901 sysctl_percpuq_drops_handler(SYSCTLFN_ARGS) 902 { 903 struct sysctlnode node; 904 struct if_percpuq *ipq; 905 int sum = 0; 906 int error; 907 908 node = *rnode; 909 ipq = node.sysctl_data; 910 911 percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum); 912 913 node.sysctl_data = ∑ 914 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 915 if (error != 0 || newp == NULL) 916 return error; 917 918 return 0; 919 } 920 921 static void 922 sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname, 923 struct if_percpuq *ipq) 924 { 925 const struct sysctlnode *cnode, *rnode; 926 927 if (sysctl_createv(clog, 0, NULL, &rnode, 928 CTLFLAG_PERMANENT, 929 CTLTYPE_NODE, "interfaces", 930 SYSCTL_DESCR("Per-interface controls"), 931 NULL, 0, NULL, 0, 932 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 933 goto bad; 934 935 if (sysctl_createv(clog, 0, &rnode, &rnode, 936 CTLFLAG_PERMANENT, 937 CTLTYPE_NODE, ifname, 938 SYSCTL_DESCR("Interface controls"), 939 NULL, 0, NULL, 0, 940 CTL_CREATE, CTL_EOL) != 0) 941 goto bad; 942 943 if (sysctl_createv(clog, 0, &rnode, &rnode, 944 CTLFLAG_PERMANENT, 945 CTLTYPE_NODE, "rcvq", 946 SYSCTL_DESCR("Interface input queue controls"), 947 NULL, 0, NULL, 0, 948 CTL_CREATE, CTL_EOL) != 0) 949 goto bad; 950 951 #ifdef NOTYET 952 /* XXX Should show each per-CPU queue length? */ 953 if (sysctl_createv(clog, 0, &rnode, &rnode, 954 CTLFLAG_PERMANENT, 955 CTLTYPE_INT, "len", 956 SYSCTL_DESCR("Current input queue length"), 957 sysctl_percpuq_len, 0, NULL, 0, 958 CTL_CREATE, CTL_EOL) != 0) 959 goto bad; 960 961 if (sysctl_createv(clog, 0, &rnode, &cnode, 962 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 963 CTLTYPE_INT, "maxlen", 964 SYSCTL_DESCR("Maximum allowed input queue length"), 965 sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0, 966 CTL_CREATE, CTL_EOL) != 0) 967 goto bad; 968 #endif 969 970 if (sysctl_createv(clog, 0, &rnode, &cnode, 971 CTLFLAG_PERMANENT, 972 CTLTYPE_INT, "drops", 973 SYSCTL_DESCR("Total packets dropped due to full input queue"), 974 sysctl_percpuq_drops_handler, 0, (void *)ipq, 0, 975 CTL_CREATE, CTL_EOL) != 0) 976 goto bad; 977 978 return; 979 bad: 980 printf("%s: could not attach sysctl nodes\n", ifname); 981 return; 982 } 983 984 /* 985 * The deferred if_start framework 986 * 987 * The common APIs to defer if_start to softint when if_start is requested 988 * from a device driver running in hardware interrupt context. 989 */ 990 /* 991 * Call ifp->if_start (or equivalent) in a dedicated softint for 992 * deferred if_start. 993 */ 994 static void 995 if_deferred_start_softint(void *arg) 996 { 997 struct if_deferred_start *ids = arg; 998 struct ifnet *ifp = ids->ids_ifp; 999 1000 ids->ids_if_start(ifp); 1001 } 1002 1003 /* 1004 * The default callback function for deferred if_start. 1005 */ 1006 static void 1007 if_deferred_start_common(struct ifnet *ifp) 1008 { 1009 1010 if_start_lock(ifp); 1011 } 1012 1013 static inline bool 1014 if_snd_is_used(struct ifnet *ifp) 1015 { 1016 1017 return ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit || 1018 ALTQ_IS_ENABLED(&ifp->if_snd); 1019 } 1020 1021 /* 1022 * Schedule deferred if_start. 1023 */ 1024 void 1025 if_schedule_deferred_start(struct ifnet *ifp) 1026 { 1027 1028 KASSERT(ifp->if_deferred_start != NULL); 1029 1030 if (if_snd_is_used(ifp) && IFQ_IS_EMPTY(&ifp->if_snd)) 1031 return; 1032 1033 softint_schedule(ifp->if_deferred_start->ids_si); 1034 } 1035 1036 /* 1037 * Create an instance of deferred if_start. A driver should call the function 1038 * only if the driver needs deferred if_start. Drivers can setup their own 1039 * deferred if_start function via 2nd argument. 1040 */ 1041 void 1042 if_deferred_start_init(struct ifnet *ifp, void (*func)(struct ifnet *)) 1043 { 1044 struct if_deferred_start *ids; 1045 1046 ids = kmem_zalloc(sizeof(*ids), KM_SLEEP); 1047 if (ids == NULL) 1048 panic("kmem_zalloc failed"); 1049 1050 ids->ids_ifp = ifp; 1051 ids->ids_si = softint_establish(SOFTINT_NET|SOFTINT_MPSAFE, 1052 if_deferred_start_softint, ids); 1053 if (func != NULL) 1054 ids->ids_if_start = func; 1055 else 1056 ids->ids_if_start = if_deferred_start_common; 1057 1058 ifp->if_deferred_start = ids; 1059 } 1060 1061 static void 1062 if_deferred_start_destroy(struct ifnet *ifp) 1063 { 1064 1065 if (ifp->if_deferred_start == NULL) 1066 return; 1067 1068 softint_disestablish(ifp->if_deferred_start->ids_si); 1069 kmem_free(ifp->if_deferred_start, sizeof(*ifp->if_deferred_start)); 1070 ifp->if_deferred_start = NULL; 1071 } 1072 1073 /* 1074 * The common interface input routine that is called by device drivers, 1075 * which should be used only when the driver's rx handler already runs 1076 * in softint. 1077 */ 1078 void 1079 if_input(struct ifnet *ifp, struct mbuf *m) 1080 { 1081 1082 KASSERT(ifp->if_percpuq == NULL); 1083 KASSERT(!cpu_intr_p()); 1084 1085 ifp->if_ipackets++; 1086 bpf_mtap(ifp, m); 1087 1088 ifp->_if_input(ifp, m); 1089 } 1090 1091 /* 1092 * DEPRECATED. Use if_initialize and if_register instead. 1093 * See the above comment of if_initialize. 1094 * 1095 * Note that it implicitly enables if_percpuq to make drivers easy to 1096 * migrate softint-based if_input without much changes. If you don't 1097 * want to enable it, use if_initialize instead. 1098 */ 1099 void 1100 if_attach(ifnet_t *ifp) 1101 { 1102 1103 if_initialize(ifp); 1104 ifp->if_percpuq = if_percpuq_create(ifp); 1105 if_register(ifp); 1106 } 1107 1108 void 1109 if_attachdomain(void) 1110 { 1111 struct ifnet *ifp; 1112 int s; 1113 int bound = curlwp_bind(); 1114 1115 s = pserialize_read_enter(); 1116 IFNET_READER_FOREACH(ifp) { 1117 struct psref psref; 1118 psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class); 1119 pserialize_read_exit(s); 1120 if_attachdomain1(ifp); 1121 s = pserialize_read_enter(); 1122 psref_release(&psref, &ifp->if_psref, ifnet_psref_class); 1123 } 1124 pserialize_read_exit(s); 1125 curlwp_bindx(bound); 1126 } 1127 1128 static void 1129 if_attachdomain1(struct ifnet *ifp) 1130 { 1131 struct domain *dp; 1132 int s; 1133 1134 s = splsoftnet(); 1135 1136 /* address family dependent data region */ 1137 memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata)); 1138 DOMAIN_FOREACH(dp) { 1139 if (dp->dom_ifattach != NULL) 1140 ifp->if_afdata[dp->dom_family] = 1141 (*dp->dom_ifattach)(ifp); 1142 } 1143 1144 splx(s); 1145 } 1146 1147 /* 1148 * Deactivate an interface. This points all of the procedure 1149 * handles at error stubs. May be called from interrupt context. 1150 */ 1151 void 1152 if_deactivate(struct ifnet *ifp) 1153 { 1154 int s; 1155 1156 s = splsoftnet(); 1157 1158 ifp->if_output = if_nulloutput; 1159 ifp->_if_input = if_nullinput; 1160 ifp->if_start = if_nullstart; 1161 ifp->if_transmit = if_nulltransmit; 1162 ifp->if_ioctl = if_nullioctl; 1163 ifp->if_init = if_nullinit; 1164 ifp->if_stop = if_nullstop; 1165 ifp->if_slowtimo = if_nullslowtimo; 1166 ifp->if_drain = if_nulldrain; 1167 1168 /* No more packets may be enqueued. */ 1169 ifp->if_snd.ifq_maxlen = 0; 1170 1171 splx(s); 1172 } 1173 1174 bool 1175 if_is_deactivated(const struct ifnet *ifp) 1176 { 1177 1178 return ifp->if_output == if_nulloutput; 1179 } 1180 1181 void 1182 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *)) 1183 { 1184 struct ifaddr *ifa, *nifa; 1185 int s; 1186 1187 s = pserialize_read_enter(); 1188 for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) { 1189 nifa = IFADDR_READER_NEXT(ifa); 1190 if (ifa->ifa_addr->sa_family != family) 1191 continue; 1192 pserialize_read_exit(s); 1193 1194 (*purgeaddr)(ifa); 1195 1196 s = pserialize_read_enter(); 1197 } 1198 pserialize_read_exit(s); 1199 } 1200 1201 #ifdef IFAREF_DEBUG 1202 static struct ifaddr **ifa_list; 1203 static int ifa_list_size; 1204 1205 /* Depends on only one if_attach runs at once */ 1206 static void 1207 if_build_ifa_list(struct ifnet *ifp) 1208 { 1209 struct ifaddr *ifa; 1210 int i; 1211 1212 KASSERT(ifa_list == NULL); 1213 KASSERT(ifa_list_size == 0); 1214 1215 IFADDR_READER_FOREACH(ifa, ifp) 1216 ifa_list_size++; 1217 1218 ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP); 1219 if (ifa_list == NULL) 1220 return; 1221 1222 i = 0; 1223 IFADDR_READER_FOREACH(ifa, ifp) { 1224 ifa_list[i++] = ifa; 1225 ifaref(ifa); 1226 } 1227 } 1228 1229 static void 1230 if_check_and_free_ifa_list(struct ifnet *ifp) 1231 { 1232 int i; 1233 struct ifaddr *ifa; 1234 1235 if (ifa_list == NULL) 1236 return; 1237 1238 for (i = 0; i < ifa_list_size; i++) { 1239 char buf[64]; 1240 1241 ifa = ifa_list[i]; 1242 sockaddr_format(ifa->ifa_addr, buf, sizeof(buf)); 1243 if (ifa->ifa_refcnt > 1) { 1244 log(LOG_WARNING, 1245 "ifa(%s) still referenced (refcnt=%d)\n", 1246 buf, ifa->ifa_refcnt - 1); 1247 } else 1248 log(LOG_DEBUG, 1249 "ifa(%s) not referenced (refcnt=%d)\n", 1250 buf, ifa->ifa_refcnt - 1); 1251 ifafree(ifa); 1252 } 1253 1254 kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size); 1255 ifa_list = NULL; 1256 ifa_list_size = 0; 1257 } 1258 #endif 1259 1260 /* 1261 * Detach an interface from the list of "active" interfaces, 1262 * freeing any resources as we go along. 1263 * 1264 * NOTE: This routine must be called with a valid thread context, 1265 * as it may block. 1266 */ 1267 void 1268 if_detach(struct ifnet *ifp) 1269 { 1270 struct socket so; 1271 struct ifaddr *ifa; 1272 #ifdef IFAREF_DEBUG 1273 struct ifaddr *last_ifa = NULL; 1274 #endif 1275 struct domain *dp; 1276 const struct protosw *pr; 1277 int s, i, family, purged; 1278 uint64_t xc; 1279 1280 #ifdef IFAREF_DEBUG 1281 if_build_ifa_list(ifp); 1282 #endif 1283 /* 1284 * XXX It's kind of lame that we have to have the 1285 * XXX socket structure... 1286 */ 1287 memset(&so, 0, sizeof(so)); 1288 1289 s = splnet(); 1290 1291 sysctl_teardown(&ifp->if_sysctl_log); 1292 mutex_enter(ifp->if_ioctl_lock); 1293 if_deactivate(ifp); 1294 mutex_exit(ifp->if_ioctl_lock); 1295 1296 IFNET_LOCK(); 1297 ifindex2ifnet[ifp->if_index] = NULL; 1298 TAILQ_REMOVE(&ifnet_list, ifp, if_list); 1299 IFNET_WRITER_REMOVE(ifp); 1300 pserialize_perform(ifnet_psz); 1301 IFNET_UNLOCK(); 1302 1303 /* Wait for all readers to drain before freeing. */ 1304 psref_target_destroy(&ifp->if_psref, ifnet_psref_class); 1305 PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry); 1306 1307 mutex_obj_free(ifp->if_ioctl_lock); 1308 ifp->if_ioctl_lock = NULL; 1309 1310 if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) { 1311 ifp->if_slowtimo = NULL; 1312 callout_halt(ifp->if_slowtimo_ch, NULL); 1313 callout_destroy(ifp->if_slowtimo_ch); 1314 kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch)); 1315 } 1316 if_deferred_start_destroy(ifp); 1317 1318 /* 1319 * Do an if_down() to give protocols a chance to do something. 1320 */ 1321 if_down(ifp); 1322 1323 #ifdef ALTQ 1324 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1325 altq_disable(&ifp->if_snd); 1326 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1327 altq_detach(&ifp->if_snd); 1328 #endif 1329 1330 mutex_obj_free(ifp->if_snd.ifq_lock); 1331 1332 #if NCARP > 0 1333 /* Remove the interface from any carp group it is a part of. */ 1334 if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP) 1335 carp_ifdetach(ifp); 1336 #endif 1337 1338 /* 1339 * Rip all the addresses off the interface. This should make 1340 * all of the routes go away. 1341 * 1342 * pr_usrreq calls can remove an arbitrary number of ifaddrs 1343 * from the list, including our "cursor", ifa. For safety, 1344 * and to honor the TAILQ abstraction, I just restart the 1345 * loop after each removal. Note that the loop will exit 1346 * when all of the remaining ifaddrs belong to the AF_LINK 1347 * family. I am counting on the historical fact that at 1348 * least one pr_usrreq in each address domain removes at 1349 * least one ifaddr. 1350 */ 1351 again: 1352 /* 1353 * At this point, no other one tries to remove ifa in the list, 1354 * so we don't need to take a lock or psref. 1355 */ 1356 IFADDR_READER_FOREACH(ifa, ifp) { 1357 family = ifa->ifa_addr->sa_family; 1358 #ifdef IFAREF_DEBUG 1359 printf("if_detach: ifaddr %p, family %d, refcnt %d\n", 1360 ifa, family, ifa->ifa_refcnt); 1361 if (last_ifa != NULL && ifa == last_ifa) 1362 panic("if_detach: loop detected"); 1363 last_ifa = ifa; 1364 #endif 1365 if (family == AF_LINK) 1366 continue; 1367 dp = pffinddomain(family); 1368 #ifdef DIAGNOSTIC 1369 if (dp == NULL) 1370 panic("if_detach: no domain for AF %d", 1371 family); 1372 #endif 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 /* 1598 * Look up a network interface cloner. 1599 */ 1600 static struct if_clone * 1601 if_clone_lookup(const char *name, int *unitp) 1602 { 1603 struct if_clone *ifc; 1604 const char *cp; 1605 char *dp, ifname[IFNAMSIZ + 3]; 1606 int unit; 1607 1608 KASSERT(mutex_owned(&if_clone_mtx)); 1609 1610 strcpy(ifname, "if_"); 1611 /* separate interface name from unit */ 1612 for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ && 1613 *cp && (*cp < '0' || *cp > '9');) 1614 *dp++ = *cp++; 1615 1616 if (cp == name || cp - name == IFNAMSIZ || !*cp) 1617 return NULL; /* No name or unit number */ 1618 *dp++ = '\0'; 1619 1620 again: 1621 LIST_FOREACH(ifc, &if_cloners, ifc_list) { 1622 if (strcmp(ifname + 3, ifc->ifc_name) == 0) 1623 break; 1624 } 1625 1626 if (ifc == NULL) { 1627 int error; 1628 if (*ifname == '\0') 1629 return NULL; 1630 mutex_exit(&if_clone_mtx); 1631 error = module_autoload(ifname, MODULE_CLASS_DRIVER); 1632 mutex_enter(&if_clone_mtx); 1633 if (error) 1634 return NULL; 1635 *ifname = '\0'; 1636 goto again; 1637 } 1638 1639 unit = 0; 1640 while (cp - name < IFNAMSIZ && *cp) { 1641 if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) { 1642 /* Bogus unit number. */ 1643 return NULL; 1644 } 1645 unit = (unit * 10) + (*cp++ - '0'); 1646 } 1647 1648 if (unitp != NULL) 1649 *unitp = unit; 1650 return ifc; 1651 } 1652 1653 /* 1654 * Register a network interface cloner. 1655 */ 1656 void 1657 if_clone_attach(struct if_clone *ifc) 1658 { 1659 1660 mutex_enter(&if_clone_mtx); 1661 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list); 1662 if_cloners_count++; 1663 mutex_exit(&if_clone_mtx); 1664 } 1665 1666 /* 1667 * Unregister a network interface cloner. 1668 */ 1669 void 1670 if_clone_detach(struct if_clone *ifc) 1671 { 1672 1673 mutex_enter(&if_clone_mtx); 1674 LIST_REMOVE(ifc, ifc_list); 1675 if_cloners_count--; 1676 mutex_exit(&if_clone_mtx); 1677 } 1678 1679 /* 1680 * Provide list of interface cloners to userspace. 1681 */ 1682 int 1683 if_clone_list(int buf_count, char *buffer, int *total) 1684 { 1685 char outbuf[IFNAMSIZ], *dst; 1686 struct if_clone *ifc; 1687 int count, error = 0; 1688 1689 mutex_enter(&if_clone_mtx); 1690 *total = if_cloners_count; 1691 if ((dst = buffer) == NULL) { 1692 /* Just asking how many there are. */ 1693 goto out; 1694 } 1695 1696 if (buf_count < 0) { 1697 error = EINVAL; 1698 goto out; 1699 } 1700 1701 count = (if_cloners_count < buf_count) ? 1702 if_cloners_count : buf_count; 1703 1704 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0; 1705 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) { 1706 (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf)); 1707 if (outbuf[sizeof(outbuf) - 1] != '\0') { 1708 error = ENAMETOOLONG; 1709 goto out; 1710 } 1711 error = copyout(outbuf, dst, sizeof(outbuf)); 1712 if (error != 0) 1713 break; 1714 } 1715 1716 out: 1717 mutex_exit(&if_clone_mtx); 1718 return error; 1719 } 1720 1721 void 1722 ifa_psref_init(struct ifaddr *ifa) 1723 { 1724 1725 psref_target_init(&ifa->ifa_psref, ifa_psref_class); 1726 } 1727 1728 void 1729 ifaref(struct ifaddr *ifa) 1730 { 1731 KASSERT(!ISSET(ifa->ifa_flags, IFA_DESTROYING)); 1732 ifa->ifa_refcnt++; 1733 } 1734 1735 void 1736 ifafree(struct ifaddr *ifa) 1737 { 1738 KASSERT(ifa != NULL); 1739 KASSERT(ifa->ifa_refcnt > 0); 1740 1741 if (--ifa->ifa_refcnt == 0) { 1742 free(ifa, M_IFADDR); 1743 } 1744 } 1745 1746 bool 1747 ifa_is_destroying(struct ifaddr *ifa) 1748 { 1749 1750 return ISSET(ifa->ifa_flags, IFA_DESTROYING); 1751 } 1752 1753 void 1754 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa) 1755 { 1756 1757 ifa->ifa_ifp = ifp; 1758 1759 IFNET_LOCK(); 1760 TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list); 1761 IFADDR_ENTRY_INIT(ifa); 1762 IFADDR_WRITER_INSERT_TAIL(ifp, ifa); 1763 IFNET_UNLOCK(); 1764 1765 ifaref(ifa); 1766 } 1767 1768 void 1769 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa) 1770 { 1771 1772 KASSERT(ifa->ifa_ifp == ifp); 1773 1774 IFNET_LOCK(); 1775 TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list); 1776 IFADDR_WRITER_REMOVE(ifa); 1777 #ifdef NET_MPSAFE 1778 pserialize_perform(ifnet_psz); 1779 #endif 1780 IFNET_UNLOCK(); 1781 1782 #ifdef NET_MPSAFE 1783 psref_target_destroy(&ifa->ifa_psref, ifa_psref_class); 1784 #endif 1785 IFADDR_ENTRY_DESTROY(ifa); 1786 ifafree(ifa); 1787 } 1788 1789 void 1790 ifa_acquire(struct ifaddr *ifa, struct psref *psref) 1791 { 1792 1793 psref_acquire(psref, &ifa->ifa_psref, ifa_psref_class); 1794 } 1795 1796 void 1797 ifa_release(struct ifaddr *ifa, struct psref *psref) 1798 { 1799 1800 if (ifa == NULL) 1801 return; 1802 1803 psref_release(psref, &ifa->ifa_psref, ifa_psref_class); 1804 } 1805 1806 bool 1807 ifa_held(struct ifaddr *ifa) 1808 { 1809 1810 return psref_held(&ifa->ifa_psref, ifa_psref_class); 1811 } 1812 1813 static inline int 1814 equal(const struct sockaddr *sa1, const struct sockaddr *sa2) 1815 { 1816 return sockaddr_cmp(sa1, sa2) == 0; 1817 } 1818 1819 /* 1820 * Locate an interface based on a complete address. 1821 */ 1822 /*ARGSUSED*/ 1823 struct ifaddr * 1824 ifa_ifwithaddr(const struct sockaddr *addr) 1825 { 1826 struct ifnet *ifp; 1827 struct ifaddr *ifa; 1828 1829 IFNET_READER_FOREACH(ifp) { 1830 if (if_is_deactivated(ifp)) 1831 continue; 1832 IFADDR_READER_FOREACH(ifa, ifp) { 1833 if (ifa->ifa_addr->sa_family != addr->sa_family) 1834 continue; 1835 if (equal(addr, ifa->ifa_addr)) 1836 return ifa; 1837 if ((ifp->if_flags & IFF_BROADCAST) && 1838 ifa->ifa_broadaddr && 1839 /* IP6 doesn't have broadcast */ 1840 ifa->ifa_broadaddr->sa_len != 0 && 1841 equal(ifa->ifa_broadaddr, addr)) 1842 return ifa; 1843 } 1844 } 1845 return NULL; 1846 } 1847 1848 struct ifaddr * 1849 ifa_ifwithaddr_psref(const struct sockaddr *addr, struct psref *psref) 1850 { 1851 struct ifaddr *ifa; 1852 int s = pserialize_read_enter(); 1853 1854 ifa = ifa_ifwithaddr(addr); 1855 if (ifa != NULL) 1856 ifa_acquire(ifa, psref); 1857 pserialize_read_exit(s); 1858 1859 return ifa; 1860 } 1861 1862 /* 1863 * Locate the point to point interface with a given destination address. 1864 */ 1865 /*ARGSUSED*/ 1866 struct ifaddr * 1867 ifa_ifwithdstaddr(const struct sockaddr *addr) 1868 { 1869 struct ifnet *ifp; 1870 struct ifaddr *ifa; 1871 1872 IFNET_READER_FOREACH(ifp) { 1873 if (if_is_deactivated(ifp)) 1874 continue; 1875 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1876 continue; 1877 IFADDR_READER_FOREACH(ifa, ifp) { 1878 if (ifa->ifa_addr->sa_family != addr->sa_family || 1879 ifa->ifa_dstaddr == NULL) 1880 continue; 1881 if (equal(addr, ifa->ifa_dstaddr)) 1882 return ifa; 1883 } 1884 } 1885 1886 return NULL; 1887 } 1888 1889 struct ifaddr * 1890 ifa_ifwithdstaddr_psref(const struct sockaddr *addr, struct psref *psref) 1891 { 1892 struct ifaddr *ifa; 1893 int s; 1894 1895 s = pserialize_read_enter(); 1896 ifa = ifa_ifwithdstaddr(addr); 1897 if (ifa != NULL) 1898 ifa_acquire(ifa, psref); 1899 pserialize_read_exit(s); 1900 1901 return ifa; 1902 } 1903 1904 /* 1905 * Find an interface on a specific network. If many, choice 1906 * is most specific found. 1907 */ 1908 struct ifaddr * 1909 ifa_ifwithnet(const struct sockaddr *addr) 1910 { 1911 struct ifnet *ifp; 1912 struct ifaddr *ifa, *ifa_maybe = NULL; 1913 const struct sockaddr_dl *sdl; 1914 u_int af = addr->sa_family; 1915 const char *addr_data = addr->sa_data, *cplim; 1916 1917 if (af == AF_LINK) { 1918 sdl = satocsdl(addr); 1919 if (sdl->sdl_index && sdl->sdl_index < if_indexlim && 1920 ifindex2ifnet[sdl->sdl_index] && 1921 !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) { 1922 return ifindex2ifnet[sdl->sdl_index]->if_dl; 1923 } 1924 } 1925 #ifdef NETATALK 1926 if (af == AF_APPLETALK) { 1927 const struct sockaddr_at *sat, *sat2; 1928 sat = (const struct sockaddr_at *)addr; 1929 IFNET_READER_FOREACH(ifp) { 1930 if (if_is_deactivated(ifp)) 1931 continue; 1932 ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp); 1933 if (ifa == NULL) 1934 continue; 1935 sat2 = (struct sockaddr_at *)ifa->ifa_addr; 1936 if (sat2->sat_addr.s_net == sat->sat_addr.s_net) 1937 return ifa; /* exact match */ 1938 if (ifa_maybe == NULL) { 1939 /* else keep the if with the right range */ 1940 ifa_maybe = ifa; 1941 } 1942 } 1943 return ifa_maybe; 1944 } 1945 #endif 1946 IFNET_READER_FOREACH(ifp) { 1947 if (if_is_deactivated(ifp)) 1948 continue; 1949 IFADDR_READER_FOREACH(ifa, ifp) { 1950 const char *cp, *cp2, *cp3; 1951 1952 if (ifa->ifa_addr->sa_family != af || 1953 ifa->ifa_netmask == NULL) 1954 next: continue; 1955 cp = addr_data; 1956 cp2 = ifa->ifa_addr->sa_data; 1957 cp3 = ifa->ifa_netmask->sa_data; 1958 cplim = (const char *)ifa->ifa_netmask + 1959 ifa->ifa_netmask->sa_len; 1960 while (cp3 < cplim) { 1961 if ((*cp++ ^ *cp2++) & *cp3++) { 1962 /* want to continue for() loop */ 1963 goto next; 1964 } 1965 } 1966 if (ifa_maybe == NULL || 1967 rt_refines(ifa->ifa_netmask, 1968 ifa_maybe->ifa_netmask)) 1969 ifa_maybe = ifa; 1970 } 1971 } 1972 return ifa_maybe; 1973 } 1974 1975 struct ifaddr * 1976 ifa_ifwithnet_psref(const struct sockaddr *addr, struct psref *psref) 1977 { 1978 struct ifaddr *ifa; 1979 int s; 1980 1981 s = pserialize_read_enter(); 1982 ifa = ifa_ifwithnet(addr); 1983 if (ifa != NULL) 1984 ifa_acquire(ifa, psref); 1985 pserialize_read_exit(s); 1986 1987 return ifa; 1988 } 1989 1990 /* 1991 * Find the interface of the addresss. 1992 */ 1993 struct ifaddr * 1994 ifa_ifwithladdr(const struct sockaddr *addr) 1995 { 1996 struct ifaddr *ia; 1997 1998 if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) || 1999 (ia = ifa_ifwithnet(addr))) 2000 return ia; 2001 return NULL; 2002 } 2003 2004 struct ifaddr * 2005 ifa_ifwithladdr_psref(const struct sockaddr *addr, struct psref *psref) 2006 { 2007 struct ifaddr *ifa; 2008 int s; 2009 2010 s = pserialize_read_enter(); 2011 ifa = ifa_ifwithladdr(addr); 2012 if (ifa != NULL) 2013 ifa_acquire(ifa, psref); 2014 pserialize_read_exit(s); 2015 2016 return ifa; 2017 } 2018 2019 /* 2020 * Find an interface using a specific address family 2021 */ 2022 struct ifaddr * 2023 ifa_ifwithaf(int af) 2024 { 2025 struct ifnet *ifp; 2026 struct ifaddr *ifa = NULL; 2027 int s; 2028 2029 s = pserialize_read_enter(); 2030 IFNET_READER_FOREACH(ifp) { 2031 if (if_is_deactivated(ifp)) 2032 continue; 2033 IFADDR_READER_FOREACH(ifa, ifp) { 2034 if (ifa->ifa_addr->sa_family == af) 2035 goto out; 2036 } 2037 } 2038 out: 2039 pserialize_read_exit(s); 2040 return ifa; 2041 } 2042 2043 /* 2044 * Find an interface address specific to an interface best matching 2045 * a given address. 2046 */ 2047 struct ifaddr * 2048 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2049 { 2050 struct ifaddr *ifa; 2051 const char *cp, *cp2, *cp3; 2052 const char *cplim; 2053 struct ifaddr *ifa_maybe = 0; 2054 u_int af = addr->sa_family; 2055 2056 if (if_is_deactivated(ifp)) 2057 return NULL; 2058 2059 if (af >= AF_MAX) 2060 return NULL; 2061 2062 IFADDR_READER_FOREACH(ifa, ifp) { 2063 if (ifa->ifa_addr->sa_family != af) 2064 continue; 2065 ifa_maybe = ifa; 2066 if (ifa->ifa_netmask == NULL) { 2067 if (equal(addr, ifa->ifa_addr) || 2068 (ifa->ifa_dstaddr && 2069 equal(addr, ifa->ifa_dstaddr))) 2070 return ifa; 2071 continue; 2072 } 2073 cp = addr->sa_data; 2074 cp2 = ifa->ifa_addr->sa_data; 2075 cp3 = ifa->ifa_netmask->sa_data; 2076 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2077 for (; cp3 < cplim; cp3++) { 2078 if ((*cp++ ^ *cp2++) & *cp3) 2079 break; 2080 } 2081 if (cp3 == cplim) 2082 return ifa; 2083 } 2084 return ifa_maybe; 2085 } 2086 2087 struct ifaddr * 2088 ifaof_ifpforaddr_psref(const struct sockaddr *addr, struct ifnet *ifp, 2089 struct psref *psref) 2090 { 2091 struct ifaddr *ifa; 2092 int s; 2093 2094 s = pserialize_read_enter(); 2095 ifa = ifaof_ifpforaddr(addr, ifp); 2096 if (ifa != NULL) 2097 ifa_acquire(ifa, psref); 2098 pserialize_read_exit(s); 2099 2100 return ifa; 2101 } 2102 2103 /* 2104 * Default action when installing a route with a Link Level gateway. 2105 * Lookup an appropriate real ifa to point to. 2106 * This should be moved to /sys/net/link.c eventually. 2107 */ 2108 void 2109 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info) 2110 { 2111 struct ifaddr *ifa; 2112 const struct sockaddr *dst; 2113 struct ifnet *ifp; 2114 struct psref psref; 2115 2116 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL || 2117 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL) 2118 return; 2119 if ((ifa = ifaof_ifpforaddr_psref(dst, ifp, &psref)) != NULL) { 2120 rt_replace_ifa(rt, ifa); 2121 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2122 ifa->ifa_rtrequest(cmd, rt, info); 2123 ifa_release(ifa, &psref); 2124 } 2125 } 2126 2127 /* 2128 * bitmask macros to manage a densely packed link_state change queue. 2129 * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and 2130 * LINK_STATE_UP(2) we need 2 bits for each state change. 2131 * As a state change to store is 0, treat all bits set as an unset item. 2132 */ 2133 #define LQ_ITEM_BITS 2 2134 #define LQ_ITEM_MASK ((1 << LQ_ITEM_BITS) - 1) 2135 #define LQ_MASK(i) (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS) 2136 #define LINK_STATE_UNSET LQ_ITEM_MASK 2137 #define LQ_ITEM(q, i) (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS) 2138 #define LQ_STORE(q, i, v) \ 2139 do { \ 2140 (q) &= ~LQ_MASK((i)); \ 2141 (q) |= (v) << (i) * LQ_ITEM_BITS; \ 2142 } while (0 /* CONSTCOND */) 2143 #define LQ_MAX(q) ((sizeof((q)) * NBBY) / LQ_ITEM_BITS) 2144 #define LQ_POP(q, v) \ 2145 do { \ 2146 (v) = LQ_ITEM((q), 0); \ 2147 (q) >>= LQ_ITEM_BITS; \ 2148 (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2149 } while (0 /* CONSTCOND */) 2150 #define LQ_PUSH(q, v) \ 2151 do { \ 2152 (q) >>= LQ_ITEM_BITS; \ 2153 (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS; \ 2154 } while (0 /* CONSTCOND */) 2155 #define LQ_FIND_UNSET(q, i) \ 2156 for ((i) = 0; i < LQ_MAX((q)); (i)++) { \ 2157 if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET) \ 2158 break; \ 2159 } 2160 /* 2161 * Handle a change in the interface link state and 2162 * queue notifications. 2163 */ 2164 void 2165 if_link_state_change(struct ifnet *ifp, int link_state) 2166 { 2167 int s, idx; 2168 2169 KASSERTMSG(if_is_link_state_changeable(ifp), 2170 "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x", 2171 ifp->if_xname, ifp->if_extflags); 2172 2173 /* Ensure change is to a valid state */ 2174 switch (link_state) { 2175 case LINK_STATE_UNKNOWN: /* FALLTHROUGH */ 2176 case LINK_STATE_DOWN: /* FALLTHROUGH */ 2177 case LINK_STATE_UP: 2178 break; 2179 default: 2180 #ifdef DEBUG 2181 printf("%s: invalid link state %d\n", 2182 ifp->if_xname, link_state); 2183 #endif 2184 return; 2185 } 2186 2187 s = splnet(); 2188 2189 /* Find the last unset event in the queue. */ 2190 LQ_FIND_UNSET(ifp->if_link_queue, idx); 2191 2192 /* 2193 * Ensure link_state doesn't match the last event in the queue. 2194 * ifp->if_link_state is not checked and set here because 2195 * that would present an inconsistent picture to the system. 2196 */ 2197 if (idx != 0 && 2198 LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state) 2199 goto out; 2200 2201 /* Handle queue overflow. */ 2202 if (idx == LQ_MAX(ifp->if_link_queue)) { 2203 uint8_t lost; 2204 2205 /* 2206 * The DOWN state must be protected from being pushed off 2207 * the queue to ensure that userland will always be 2208 * in a sane state. 2209 * Because DOWN is protected, there is no need to protect 2210 * UNKNOWN. 2211 * It should be invalid to change from any other state to 2212 * UNKNOWN anyway ... 2213 */ 2214 lost = LQ_ITEM(ifp->if_link_queue, 0); 2215 LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state); 2216 if (lost == LINK_STATE_DOWN) { 2217 lost = LQ_ITEM(ifp->if_link_queue, 0); 2218 LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN); 2219 } 2220 printf("%s: lost link state change %s\n", 2221 ifp->if_xname, 2222 lost == LINK_STATE_UP ? "UP" : 2223 lost == LINK_STATE_DOWN ? "DOWN" : 2224 "UNKNOWN"); 2225 } else 2226 LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state); 2227 2228 softint_schedule(ifp->if_link_si); 2229 2230 out: 2231 splx(s); 2232 } 2233 2234 /* 2235 * Handle interface link state change notifications. 2236 * Must be called at splnet(). 2237 */ 2238 static void 2239 if_link_state_change0(struct ifnet *ifp, int link_state) 2240 { 2241 struct domain *dp; 2242 2243 /* Ensure the change is still valid. */ 2244 if (ifp->if_link_state == link_state) 2245 return; 2246 2247 #ifdef DEBUG 2248 log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname, 2249 link_state == LINK_STATE_UP ? "UP" : 2250 link_state == LINK_STATE_DOWN ? "DOWN" : 2251 "UNKNOWN", 2252 ifp->if_link_state == LINK_STATE_UP ? "UP" : 2253 ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" : 2254 "UNKNOWN"); 2255 #endif 2256 2257 /* 2258 * When going from UNKNOWN to UP, we need to mark existing 2259 * addresses as tentative and restart DAD as we may have 2260 * erroneously not found a duplicate. 2261 * 2262 * This needs to happen before rt_ifmsg to avoid a race where 2263 * listeners would have an address and expect it to work right 2264 * away. 2265 */ 2266 if (link_state == LINK_STATE_UP && 2267 ifp->if_link_state == LINK_STATE_UNKNOWN) 2268 { 2269 DOMAIN_FOREACH(dp) { 2270 if (dp->dom_if_link_state_change != NULL) 2271 dp->dom_if_link_state_change(ifp, 2272 LINK_STATE_DOWN); 2273 } 2274 } 2275 2276 ifp->if_link_state = link_state; 2277 2278 /* Notify that the link state has changed. */ 2279 rt_ifmsg(ifp); 2280 2281 #if NCARP > 0 2282 if (ifp->if_carp) 2283 carp_carpdev_state(ifp); 2284 #endif 2285 2286 DOMAIN_FOREACH(dp) { 2287 if (dp->dom_if_link_state_change != NULL) 2288 dp->dom_if_link_state_change(ifp, link_state); 2289 } 2290 } 2291 2292 /* 2293 * Process the interface link state change queue. 2294 */ 2295 static void 2296 if_link_state_change_si(void *arg) 2297 { 2298 struct ifnet *ifp = arg; 2299 int s; 2300 uint8_t state; 2301 2302 #ifndef NET_MPSAFE 2303 mutex_enter(softnet_lock); 2304 KERNEL_LOCK(1, NULL); 2305 #endif 2306 s = splnet(); 2307 2308 /* Pop a link state change from the queue and process it. */ 2309 LQ_POP(ifp->if_link_queue, state); 2310 if_link_state_change0(ifp, state); 2311 2312 /* If there is a link state change to come, schedule it. */ 2313 if (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET) 2314 softint_schedule(ifp->if_link_si); 2315 2316 splx(s); 2317 #ifndef NET_MPSAFE 2318 KERNEL_UNLOCK_ONE(NULL); 2319 mutex_exit(softnet_lock); 2320 #endif 2321 } 2322 2323 /* 2324 * Default action when installing a local route on a point-to-point 2325 * interface. 2326 */ 2327 void 2328 p2p_rtrequest(int req, struct rtentry *rt, 2329 __unused const struct rt_addrinfo *info) 2330 { 2331 struct ifnet *ifp = rt->rt_ifp; 2332 struct ifaddr *ifa, *lo0ifa; 2333 int s = pserialize_read_enter(); 2334 2335 switch (req) { 2336 case RTM_ADD: 2337 if ((rt->rt_flags & RTF_LOCAL) == 0) 2338 break; 2339 2340 rt->rt_ifp = lo0ifp; 2341 2342 IFADDR_READER_FOREACH(ifa, ifp) { 2343 if (equal(rt_getkey(rt), ifa->ifa_addr)) 2344 break; 2345 } 2346 if (ifa == NULL) 2347 break; 2348 2349 /* 2350 * Ensure lo0 has an address of the same family. 2351 */ 2352 IFADDR_READER_FOREACH(lo0ifa, lo0ifp) { 2353 if (lo0ifa->ifa_addr->sa_family == 2354 ifa->ifa_addr->sa_family) 2355 break; 2356 } 2357 if (lo0ifa == NULL) 2358 break; 2359 2360 /* 2361 * Make sure to set rt->rt_ifa to the interface 2362 * address we are using, otherwise we will have trouble 2363 * with source address selection. 2364 */ 2365 if (ifa != rt->rt_ifa) 2366 rt_replace_ifa(rt, ifa); 2367 break; 2368 case RTM_DELETE: 2369 default: 2370 break; 2371 } 2372 pserialize_read_exit(s); 2373 } 2374 2375 /* 2376 * Mark an interface down and notify protocols of 2377 * the transition. 2378 * NOTE: must be called at splsoftnet or equivalent. 2379 */ 2380 void 2381 if_down(struct ifnet *ifp) 2382 { 2383 struct ifaddr *ifa; 2384 struct domain *dp; 2385 int s, bound; 2386 struct psref psref; 2387 2388 ifp->if_flags &= ~IFF_UP; 2389 nanotime(&ifp->if_lastchange); 2390 2391 bound = curlwp_bind(); 2392 s = pserialize_read_enter(); 2393 IFADDR_READER_FOREACH(ifa, ifp) { 2394 ifa_acquire(ifa, &psref); 2395 pserialize_read_exit(s); 2396 2397 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2398 2399 s = pserialize_read_enter(); 2400 ifa_release(ifa, &psref); 2401 } 2402 pserialize_read_exit(s); 2403 curlwp_bindx(bound); 2404 2405 IFQ_PURGE(&ifp->if_snd); 2406 #if NCARP > 0 2407 if (ifp->if_carp) 2408 carp_carpdev_state(ifp); 2409 #endif 2410 rt_ifmsg(ifp); 2411 DOMAIN_FOREACH(dp) { 2412 if (dp->dom_if_down) 2413 dp->dom_if_down(ifp); 2414 } 2415 } 2416 2417 /* 2418 * Mark an interface up and notify protocols of 2419 * the transition. 2420 * NOTE: must be called at splsoftnet or equivalent. 2421 */ 2422 void 2423 if_up(struct ifnet *ifp) 2424 { 2425 #ifdef notyet 2426 struct ifaddr *ifa; 2427 #endif 2428 struct domain *dp; 2429 2430 ifp->if_flags |= IFF_UP; 2431 nanotime(&ifp->if_lastchange); 2432 #ifdef notyet 2433 /* this has no effect on IP, and will kill all ISO connections XXX */ 2434 IFADDR_READER_FOREACH(ifa, ifp) 2435 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2436 #endif 2437 #if NCARP > 0 2438 if (ifp->if_carp) 2439 carp_carpdev_state(ifp); 2440 #endif 2441 rt_ifmsg(ifp); 2442 DOMAIN_FOREACH(dp) { 2443 if (dp->dom_if_up) 2444 dp->dom_if_up(ifp); 2445 } 2446 } 2447 2448 /* 2449 * Handle interface slowtimo timer routine. Called 2450 * from softclock, we decrement timer (if set) and 2451 * call the appropriate interface routine on expiration. 2452 */ 2453 static void 2454 if_slowtimo(void *arg) 2455 { 2456 void (*slowtimo)(struct ifnet *); 2457 struct ifnet *ifp = arg; 2458 int s; 2459 2460 slowtimo = ifp->if_slowtimo; 2461 if (__predict_false(slowtimo == NULL)) 2462 return; 2463 2464 s = splnet(); 2465 if (ifp->if_timer != 0 && --ifp->if_timer == 0) 2466 (*slowtimo)(ifp); 2467 2468 splx(s); 2469 2470 if (__predict_true(ifp->if_slowtimo != NULL)) 2471 callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ); 2472 } 2473 2474 /* 2475 * Set/clear promiscuous mode on interface ifp based on the truth value 2476 * of pswitch. The calls are reference counted so that only the first 2477 * "on" request actually has an effect, as does the final "off" request. 2478 * Results are undefined if the "off" and "on" requests are not matched. 2479 */ 2480 int 2481 ifpromisc(struct ifnet *ifp, int pswitch) 2482 { 2483 int pcount, ret; 2484 short nflags; 2485 2486 pcount = ifp->if_pcount; 2487 if (pswitch) { 2488 /* 2489 * Allow the device to be "placed" into promiscuous 2490 * mode even if it is not configured up. It will 2491 * consult IFF_PROMISC when it is brought up. 2492 */ 2493 if (ifp->if_pcount++ != 0) 2494 return 0; 2495 nflags = ifp->if_flags | IFF_PROMISC; 2496 } else { 2497 if (--ifp->if_pcount > 0) 2498 return 0; 2499 nflags = ifp->if_flags & ~IFF_PROMISC; 2500 } 2501 ret = if_flags_set(ifp, nflags); 2502 /* Restore interface state if not successful. */ 2503 if (ret != 0) { 2504 ifp->if_pcount = pcount; 2505 } 2506 return ret; 2507 } 2508 2509 /* 2510 * Map interface name to 2511 * interface structure pointer. 2512 */ 2513 struct ifnet * 2514 ifunit(const char *name) 2515 { 2516 struct ifnet *ifp; 2517 const char *cp = name; 2518 u_int unit = 0; 2519 u_int i; 2520 int s; 2521 2522 /* 2523 * If the entire name is a number, treat it as an ifindex. 2524 */ 2525 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2526 unit = unit * 10 + (*cp - '0'); 2527 } 2528 2529 /* 2530 * If the number took all of the name, then it's a valid ifindex. 2531 */ 2532 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2533 if (unit >= if_indexlim) 2534 return NULL; 2535 ifp = ifindex2ifnet[unit]; 2536 if (ifp == NULL || if_is_deactivated(ifp)) 2537 return NULL; 2538 return ifp; 2539 } 2540 2541 ifp = NULL; 2542 s = pserialize_read_enter(); 2543 IFNET_READER_FOREACH(ifp) { 2544 if (if_is_deactivated(ifp)) 2545 continue; 2546 if (strcmp(ifp->if_xname, name) == 0) 2547 goto out; 2548 } 2549 out: 2550 pserialize_read_exit(s); 2551 return ifp; 2552 } 2553 2554 /* 2555 * Get a reference of an ifnet object by an interface name. 2556 * The returned reference is protected by psref(9). The caller 2557 * must release a returned reference by if_put after use. 2558 */ 2559 struct ifnet * 2560 if_get(const char *name, struct psref *psref) 2561 { 2562 struct ifnet *ifp; 2563 const char *cp = name; 2564 u_int unit = 0; 2565 u_int i; 2566 int s; 2567 2568 /* 2569 * If the entire name is a number, treat it as an ifindex. 2570 */ 2571 for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) { 2572 unit = unit * 10 + (*cp - '0'); 2573 } 2574 2575 /* 2576 * If the number took all of the name, then it's a valid ifindex. 2577 */ 2578 if (i == IFNAMSIZ || (cp != name && *cp == '\0')) { 2579 if (unit >= if_indexlim) 2580 return NULL; 2581 ifp = ifindex2ifnet[unit]; 2582 if (ifp == NULL || if_is_deactivated(ifp)) 2583 return NULL; 2584 return ifp; 2585 } 2586 2587 ifp = NULL; 2588 s = pserialize_read_enter(); 2589 IFNET_READER_FOREACH(ifp) { 2590 if (if_is_deactivated(ifp)) 2591 continue; 2592 if (strcmp(ifp->if_xname, name) == 0) { 2593 psref_acquire(psref, &ifp->if_psref, 2594 ifnet_psref_class); 2595 goto out; 2596 } 2597 } 2598 out: 2599 pserialize_read_exit(s); 2600 return ifp; 2601 } 2602 2603 /* 2604 * Release a reference of an ifnet object given by if_get or 2605 * if_get_byindex. 2606 */ 2607 void 2608 if_put(const struct ifnet *ifp, struct psref *psref) 2609 { 2610 2611 if (ifp == NULL) 2612 return; 2613 2614 psref_release(psref, &ifp->if_psref, ifnet_psref_class); 2615 } 2616 2617 ifnet_t * 2618 if_byindex(u_int idx) 2619 { 2620 ifnet_t *ifp; 2621 2622 ifp = (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL; 2623 if (ifp != NULL && if_is_deactivated(ifp)) 2624 ifp = NULL; 2625 return ifp; 2626 } 2627 2628 /* 2629 * Get a reference of an ifnet object by an interface index. 2630 * The returned reference is protected by psref(9). The caller 2631 * must release a returned reference by if_put after use. 2632 */ 2633 ifnet_t * 2634 if_get_byindex(u_int idx, struct psref *psref) 2635 { 2636 ifnet_t *ifp; 2637 int s; 2638 2639 s = pserialize_read_enter(); 2640 ifp = (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL; 2641 if (ifp != NULL && if_is_deactivated(ifp)) 2642 ifp = NULL; 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 * XXX it's safe only if the passed ifp is guaranteed to not be freed, 2652 * for example the ifp is already held or some other object is held which 2653 * guarantes the ifp to not be freed indirectly. 2654 */ 2655 void 2656 if_acquire_NOMPSAFE(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 3211 s = splnet(); 3212 3213 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3214 if (error != 0) { 3215 /* mbuf is already freed */ 3216 goto out; 3217 } 3218 3219 ifp->if_obytes += m->m_pkthdr.len;; 3220 if (m->m_flags & M_MCAST) 3221 ifp->if_omcasts++; 3222 3223 if ((ifp->if_flags & IFF_OACTIVE) == 0) 3224 if_start_lock(ifp); 3225 out: 3226 splx(s); 3227 3228 return error; 3229 } 3230 3231 int 3232 if_transmit_lock(struct ifnet *ifp, struct mbuf *m) 3233 { 3234 int error; 3235 3236 #ifdef ALTQ 3237 KERNEL_LOCK(1, NULL); 3238 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 3239 error = if_transmit(ifp, m); 3240 KERNEL_UNLOCK_ONE(NULL); 3241 } else { 3242 KERNEL_UNLOCK_ONE(NULL); 3243 error = (*ifp->if_transmit)(ifp, m); 3244 } 3245 #else /* !ALTQ */ 3246 error = (*ifp->if_transmit)(ifp, m); 3247 #endif /* !ALTQ */ 3248 3249 return error; 3250 } 3251 3252 /* 3253 * Queue message on interface, and start output if interface 3254 * not yet active. 3255 */ 3256 int 3257 ifq_enqueue(struct ifnet *ifp, struct mbuf *m) 3258 { 3259 3260 return if_transmit_lock(ifp, m); 3261 } 3262 3263 /* 3264 * Queue message on interface, possibly using a second fast queue 3265 */ 3266 int 3267 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m) 3268 { 3269 int error = 0; 3270 3271 if (ifq != NULL 3272 #ifdef ALTQ 3273 && ALTQ_IS_ENABLED(&ifp->if_snd) == 0 3274 #endif 3275 ) { 3276 if (IF_QFULL(ifq)) { 3277 IF_DROP(&ifp->if_snd); 3278 m_freem(m); 3279 if (error == 0) 3280 error = ENOBUFS; 3281 } else 3282 IF_ENQUEUE(ifq, m); 3283 } else 3284 IFQ_ENQUEUE(&ifp->if_snd, m, error); 3285 if (error != 0) { 3286 ++ifp->if_oerrors; 3287 return error; 3288 } 3289 return 0; 3290 } 3291 3292 int 3293 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src) 3294 { 3295 int rc; 3296 3297 if (ifp->if_initaddr != NULL) 3298 rc = (*ifp->if_initaddr)(ifp, ifa, src); 3299 else if (src || 3300 /* FIXME: may not hold if_ioctl_lock */ 3301 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY) 3302 rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa); 3303 3304 return rc; 3305 } 3306 3307 int 3308 if_do_dad(struct ifnet *ifp) 3309 { 3310 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 3311 return 0; 3312 3313 switch (ifp->if_type) { 3314 case IFT_FAITH: 3315 /* 3316 * These interfaces do not have the IFF_LOOPBACK flag, 3317 * but loop packets back. We do not have to do DAD on such 3318 * interfaces. We should even omit it, because loop-backed 3319 * responses would confuse the DAD procedure. 3320 */ 3321 return 0; 3322 default: 3323 /* 3324 * Our DAD routine requires the interface up and running. 3325 * However, some interfaces can be up before the RUNNING 3326 * status. Additionaly, users may try to assign addresses 3327 * before the interface becomes up (or running). 3328 * We simply skip DAD in such a case as a work around. 3329 * XXX: we should rather mark "tentative" on such addresses, 3330 * and do DAD after the interface becomes ready. 3331 */ 3332 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != 3333 (IFF_UP|IFF_RUNNING)) 3334 return 0; 3335 3336 return 1; 3337 } 3338 } 3339 3340 int 3341 if_flags_set(ifnet_t *ifp, const short flags) 3342 { 3343 int rc; 3344 3345 if (ifp->if_setflags != NULL) 3346 rc = (*ifp->if_setflags)(ifp, flags); 3347 else { 3348 short cantflags, chgdflags; 3349 struct ifreq ifr; 3350 3351 chgdflags = ifp->if_flags ^ flags; 3352 cantflags = chgdflags & IFF_CANTCHANGE; 3353 3354 if (cantflags != 0) 3355 ifp->if_flags ^= cantflags; 3356 3357 /* Traditionally, we do not call if_ioctl after 3358 * setting/clearing only IFF_PROMISC if the interface 3359 * isn't IFF_UP. Uphold that tradition. 3360 */ 3361 if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0) 3362 return 0; 3363 3364 memset(&ifr, 0, sizeof(ifr)); 3365 3366 ifr.ifr_flags = flags & ~IFF_CANTCHANGE; 3367 /* FIXME: may not hold if_ioctl_lock */ 3368 rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr); 3369 3370 if (rc != 0 && cantflags != 0) 3371 ifp->if_flags ^= cantflags; 3372 } 3373 3374 return rc; 3375 } 3376 3377 int 3378 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa) 3379 { 3380 int rc; 3381 struct ifreq ifr; 3382 3383 if (ifp->if_mcastop != NULL) 3384 rc = (*ifp->if_mcastop)(ifp, cmd, sa); 3385 else { 3386 ifreq_setaddr(cmd, &ifr, sa); 3387 rc = (*ifp->if_ioctl)(ifp, cmd, &ifr); 3388 } 3389 3390 return rc; 3391 } 3392 3393 static void 3394 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname, 3395 struct ifaltq *ifq) 3396 { 3397 const struct sysctlnode *cnode, *rnode; 3398 3399 if (sysctl_createv(clog, 0, NULL, &rnode, 3400 CTLFLAG_PERMANENT, 3401 CTLTYPE_NODE, "interfaces", 3402 SYSCTL_DESCR("Per-interface controls"), 3403 NULL, 0, NULL, 0, 3404 CTL_NET, CTL_CREATE, CTL_EOL) != 0) 3405 goto bad; 3406 3407 if (sysctl_createv(clog, 0, &rnode, &rnode, 3408 CTLFLAG_PERMANENT, 3409 CTLTYPE_NODE, ifname, 3410 SYSCTL_DESCR("Interface controls"), 3411 NULL, 0, NULL, 0, 3412 CTL_CREATE, CTL_EOL) != 0) 3413 goto bad; 3414 3415 if (sysctl_createv(clog, 0, &rnode, &rnode, 3416 CTLFLAG_PERMANENT, 3417 CTLTYPE_NODE, "sndq", 3418 SYSCTL_DESCR("Interface output queue controls"), 3419 NULL, 0, NULL, 0, 3420 CTL_CREATE, CTL_EOL) != 0) 3421 goto bad; 3422 3423 if (sysctl_createv(clog, 0, &rnode, &cnode, 3424 CTLFLAG_PERMANENT, 3425 CTLTYPE_INT, "len", 3426 SYSCTL_DESCR("Current output queue length"), 3427 NULL, 0, &ifq->ifq_len, 0, 3428 CTL_CREATE, CTL_EOL) != 0) 3429 goto bad; 3430 3431 if (sysctl_createv(clog, 0, &rnode, &cnode, 3432 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3433 CTLTYPE_INT, "maxlen", 3434 SYSCTL_DESCR("Maximum allowed output queue length"), 3435 NULL, 0, &ifq->ifq_maxlen, 0, 3436 CTL_CREATE, CTL_EOL) != 0) 3437 goto bad; 3438 3439 if (sysctl_createv(clog, 0, &rnode, &cnode, 3440 CTLFLAG_PERMANENT, 3441 CTLTYPE_INT, "drops", 3442 SYSCTL_DESCR("Packets dropped due to full output queue"), 3443 NULL, 0, &ifq->ifq_drops, 0, 3444 CTL_CREATE, CTL_EOL) != 0) 3445 goto bad; 3446 3447 return; 3448 bad: 3449 printf("%s: could not attach sysctl nodes\n", ifname); 3450 return; 3451 } 3452 3453 #if defined(INET) || defined(INET6) 3454 3455 #define SYSCTL_NET_PKTQ(q, cn, c) \ 3456 static int \ 3457 sysctl_net_##q##_##cn(SYSCTLFN_ARGS) \ 3458 { \ 3459 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c); \ 3460 } 3461 3462 #if defined(INET) 3463 static int 3464 sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS) 3465 { 3466 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq); 3467 } 3468 SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS) 3469 SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS) 3470 #endif 3471 3472 #if defined(INET6) 3473 static int 3474 sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS) 3475 { 3476 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq); 3477 } 3478 SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS) 3479 SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS) 3480 #endif 3481 3482 static void 3483 sysctl_net_pktq_setup(struct sysctllog **clog, int pf) 3484 { 3485 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL; 3486 const char *pfname = NULL, *ipname = NULL; 3487 int ipn = 0, qid = 0; 3488 3489 switch (pf) { 3490 #if defined(INET) 3491 case PF_INET: 3492 len_func = sysctl_net_ip_pktq_items; 3493 maxlen_func = sysctl_net_ip_pktq_maxlen; 3494 drops_func = sysctl_net_ip_pktq_drops; 3495 pfname = "inet", ipn = IPPROTO_IP; 3496 ipname = "ip", qid = IPCTL_IFQ; 3497 break; 3498 #endif 3499 #if defined(INET6) 3500 case PF_INET6: 3501 len_func = sysctl_net_ip6_pktq_items; 3502 maxlen_func = sysctl_net_ip6_pktq_maxlen; 3503 drops_func = sysctl_net_ip6_pktq_drops; 3504 pfname = "inet6", ipn = IPPROTO_IPV6; 3505 ipname = "ip6", qid = IPV6CTL_IFQ; 3506 break; 3507 #endif 3508 default: 3509 KASSERT(false); 3510 } 3511 3512 sysctl_createv(clog, 0, NULL, NULL, 3513 CTLFLAG_PERMANENT, 3514 CTLTYPE_NODE, pfname, NULL, 3515 NULL, 0, NULL, 0, 3516 CTL_NET, pf, CTL_EOL); 3517 sysctl_createv(clog, 0, NULL, NULL, 3518 CTLFLAG_PERMANENT, 3519 CTLTYPE_NODE, ipname, NULL, 3520 NULL, 0, NULL, 0, 3521 CTL_NET, pf, ipn, CTL_EOL); 3522 sysctl_createv(clog, 0, NULL, NULL, 3523 CTLFLAG_PERMANENT, 3524 CTLTYPE_NODE, "ifq", 3525 SYSCTL_DESCR("Protocol input queue controls"), 3526 NULL, 0, NULL, 0, 3527 CTL_NET, pf, ipn, qid, CTL_EOL); 3528 3529 sysctl_createv(clog, 0, NULL, NULL, 3530 CTLFLAG_PERMANENT, 3531 CTLTYPE_INT, "len", 3532 SYSCTL_DESCR("Current input queue length"), 3533 len_func, 0, NULL, 0, 3534 CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL); 3535 sysctl_createv(clog, 0, NULL, NULL, 3536 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 3537 CTLTYPE_INT, "maxlen", 3538 SYSCTL_DESCR("Maximum allowed input queue length"), 3539 maxlen_func, 0, NULL, 0, 3540 CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL); 3541 sysctl_createv(clog, 0, NULL, NULL, 3542 CTLFLAG_PERMANENT, 3543 CTLTYPE_INT, "drops", 3544 SYSCTL_DESCR("Packets dropped due to full input queue"), 3545 drops_func, 0, NULL, 0, 3546 CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL); 3547 } 3548 #endif /* INET || INET6 */ 3549 3550 static int 3551 if_sdl_sysctl(SYSCTLFN_ARGS) 3552 { 3553 struct ifnet *ifp; 3554 const struct sockaddr_dl *sdl; 3555 struct psref psref; 3556 int error = 0; 3557 int bound; 3558 3559 if (namelen != 1) 3560 return EINVAL; 3561 3562 bound = curlwp_bind(); 3563 ifp = if_get_byindex(name[0], &psref); 3564 if (ifp == NULL) { 3565 error = ENODEV; 3566 goto out0; 3567 } 3568 3569 sdl = ifp->if_sadl; 3570 if (sdl == NULL) { 3571 *oldlenp = 0; 3572 goto out1; 3573 } 3574 3575 if (oldp == NULL) { 3576 *oldlenp = sdl->sdl_alen; 3577 goto out1; 3578 } 3579 3580 if (*oldlenp >= sdl->sdl_alen) 3581 *oldlenp = sdl->sdl_alen; 3582 error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp); 3583 out1: 3584 if_put(ifp, &psref); 3585 out0: 3586 curlwp_bindx(bound); 3587 return error; 3588 } 3589 3590 static void 3591 if_sysctl_setup(struct sysctllog **clog) 3592 { 3593 const struct sysctlnode *rnode = NULL; 3594 3595 sysctl_createv(clog, 0, NULL, &rnode, 3596 CTLFLAG_PERMANENT, 3597 CTLTYPE_NODE, "sdl", 3598 SYSCTL_DESCR("Get active link-layer address"), 3599 if_sdl_sysctl, 0, NULL, 0, 3600 CTL_NET, CTL_CREATE, CTL_EOL); 3601 3602 #if defined(INET) 3603 sysctl_net_pktq_setup(NULL, PF_INET); 3604 #endif 3605 #ifdef INET6 3606 if (in6_present) 3607 sysctl_net_pktq_setup(NULL, PF_INET6); 3608 #endif 3609 } 3610
Indexes created Tue Sep 23 07:09:52 GMT 2025