1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org> 5 * Copyright (c) 1980, 1986, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)if.c 8.5 (Berkeley) 1/9/95 33 * $FreeBSD$ 34 */ 35 36 #include "opt_bpf.h" 37 #include "opt_inet6.h" 38 #include "opt_inet.h" 39 #include "opt_ddb.h" 40 41 #include <sys/param.h> 42 #include <sys/capsicum.h> 43 #include <sys/conf.h> 44 #include <sys/eventhandler.h> 45 #include <sys/malloc.h> 46 #include <sys/domainset.h> 47 #include <sys/sbuf.h> 48 #include <sys/bus.h> 49 #include <sys/epoch.h> 50 #include <sys/mbuf.h> 51 #include <sys/systm.h> 52 #include <sys/priv.h> 53 #include <sys/proc.h> 54 #include <sys/socket.h> 55 #include <sys/socketvar.h> 56 #include <sys/protosw.h> 57 #include <sys/kernel.h> 58 #include <sys/lock.h> 59 #include <sys/refcount.h> 60 #include <sys/module.h> 61 #include <sys/nv.h> 62 #include <sys/rwlock.h> 63 #include <sys/sockio.h> 64 #include <sys/syslog.h> 65 #include <sys/sysctl.h> 66 #include <sys/sysent.h> 67 #include <sys/taskqueue.h> 68 #include <sys/domain.h> 69 #include <sys/jail.h> 70 #include <sys/priv.h> 71 72 #ifdef DDB 73 #include <ddb/ddb.h> 74 #endif 75 76 #include <machine/stdarg.h> 77 #include <vm/uma.h> 78 79 #include <net/bpf.h> 80 #include <net/ethernet.h> 81 #include <net/if.h> 82 #include <net/if_arp.h> 83 #include <net/if_clone.h> 84 #include <net/if_dl.h> 85 #include <net/if_types.h> 86 #include <net/if_var.h> 87 #include <net/if_media.h> 88 #include <net/if_mib.h> 89 #include <net/if_private.h> 90 #include <net/if_vlan_var.h> 91 #include <net/radix.h> 92 #include <net/route.h> 93 #include <net/route/route_ctl.h> 94 #include <net/vnet.h> 95 96 #if defined(INET) || defined(INET6) 97 #include <net/ethernet.h> 98 #include <netinet/in.h> 99 #include <netinet/in_var.h> 100 #include <netinet/ip.h> 101 #include <netinet/ip_carp.h> 102 #ifdef INET 103 #include <net/debugnet.h> 104 #include <netinet/if_ether.h> 105 #endif /* INET */ 106 #ifdef INET6 107 #include <netinet6/in6_var.h> 108 #include <netinet6/in6_ifattach.h> 109 #endif /* INET6 */ 110 #endif /* INET || INET6 */ 111 112 #include <security/mac/mac_framework.h> 113 114 /* 115 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 116 * and ifr_ifru when it is used in SIOCGIFCONF. 117 */ 118 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 119 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 120 121 __read_mostly epoch_t net_epoch_preempt; 122 #ifdef COMPAT_FREEBSD32 123 #include <sys/mount.h> 124 #include <compat/freebsd32/freebsd32.h> 125 126 struct ifreq_buffer32 { 127 uint32_t length; /* (size_t) */ 128 uint32_t buffer; /* (void *) */ 129 }; 130 131 /* 132 * Interface request structure used for socket 133 * ioctl's. All interface ioctl's must have parameter 134 * definitions which begin with ifr_name. The 135 * remainder may be interface specific. 136 */ 137 struct ifreq32 { 138 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 139 union { 140 struct sockaddr ifru_addr; 141 struct sockaddr ifru_dstaddr; 142 struct sockaddr ifru_broadaddr; 143 struct ifreq_buffer32 ifru_buffer; 144 short ifru_flags[2]; 145 short ifru_index; 146 int ifru_jid; 147 int ifru_metric; 148 int ifru_mtu; 149 int ifru_phys; 150 int ifru_media; 151 uint32_t ifru_data; 152 int ifru_cap[2]; 153 u_int ifru_fib; 154 u_char ifru_vlan_pcp; 155 } ifr_ifru; 156 }; 157 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 158 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 159 __offsetof(struct ifreq32, ifr_ifru)); 160 161 struct ifconf32 { 162 int32_t ifc_len; 163 union { 164 uint32_t ifcu_buf; 165 uint32_t ifcu_req; 166 } ifc_ifcu; 167 }; 168 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 169 170 struct ifdrv32 { 171 char ifd_name[IFNAMSIZ]; 172 uint32_t ifd_cmd; 173 uint32_t ifd_len; 174 uint32_t ifd_data; 175 }; 176 #define SIOCSDRVSPEC32 _IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32) 177 #define SIOCGDRVSPEC32 _IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32) 178 179 struct ifgroupreq32 { 180 char ifgr_name[IFNAMSIZ]; 181 u_int ifgr_len; 182 union { 183 char ifgru_group[IFNAMSIZ]; 184 uint32_t ifgru_groups; 185 } ifgr_ifgru; 186 }; 187 #define SIOCAIFGROUP32 _IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32) 188 #define SIOCGIFGROUP32 _IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32) 189 #define SIOCDIFGROUP32 _IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32) 190 #define SIOCGIFGMEMB32 _IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32) 191 192 struct ifmediareq32 { 193 char ifm_name[IFNAMSIZ]; 194 int ifm_current; 195 int ifm_mask; 196 int ifm_status; 197 int ifm_active; 198 int ifm_count; 199 uint32_t ifm_ulist; /* (int *) */ 200 }; 201 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 202 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 203 #endif /* COMPAT_FREEBSD32 */ 204 205 union ifreq_union { 206 struct ifreq ifr; 207 #ifdef COMPAT_FREEBSD32 208 struct ifreq32 ifr32; 209 #endif 210 }; 211 212 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 213 "Link layers"); 214 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 215 "Generic link-management"); 216 217 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 218 &ifqmaxlen, 0, "max send queue size"); 219 220 /* Log link state change events */ 221 static int log_link_state_change = 1; 222 223 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 224 &log_link_state_change, 0, 225 "log interface link state change events"); 226 227 /* Log promiscuous mode change events */ 228 static int log_promisc_mode_change = 1; 229 230 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 231 &log_promisc_mode_change, 1, 232 "log promiscuous mode change events"); 233 234 /* Interface description */ 235 static unsigned int ifdescr_maxlen = 1024; 236 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 237 &ifdescr_maxlen, 0, 238 "administrative maximum length for interface description"); 239 240 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 241 242 /* global sx for non-critical path ifdescr */ 243 static struct sx ifdescr_sx; 244 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 245 246 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 247 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 248 /* These are external hooks for CARP. */ 249 void (*carp_linkstate_p)(struct ifnet *ifp); 250 void (*carp_demote_adj_p)(int, char *); 251 int (*carp_master_p)(struct ifaddr *); 252 #if defined(INET) || defined(INET6) 253 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 254 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 255 const struct sockaddr *sa); 256 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 257 int (*carp_attach_p)(struct ifaddr *, int); 258 void (*carp_detach_p)(struct ifaddr *, bool); 259 #endif 260 #ifdef INET 261 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 262 #endif 263 #ifdef INET6 264 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 265 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 266 const struct in6_addr *taddr); 267 #endif 268 269 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 270 271 /* 272 * XXX: Style; these should be sorted alphabetically, and unprototyped 273 * static functions should be prototyped. Currently they are sorted by 274 * declaration order. 275 */ 276 static void if_attachdomain(void *); 277 static void if_attachdomain1(struct ifnet *); 278 static int ifconf(u_long, caddr_t); 279 static void if_input_default(struct ifnet *, struct mbuf *); 280 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 281 static void if_route(struct ifnet *, int flag, int fam); 282 static int if_setflag(struct ifnet *, int, int, int *, int); 283 static int if_transmit_default(struct ifnet *ifp, struct mbuf *m); 284 static void if_unroute(struct ifnet *, int flag, int fam); 285 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 286 static void do_link_state_change(void *, int); 287 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 288 static int if_getgroupmembers(struct ifgroupreq *); 289 static void if_delgroups(struct ifnet *); 290 static void if_attach_internal(struct ifnet *, bool); 291 static int if_detach_internal(struct ifnet *, bool); 292 static void if_siocaddmulti(void *, int); 293 static void if_link_ifnet(struct ifnet *); 294 static bool if_unlink_ifnet(struct ifnet *, bool); 295 #ifdef VIMAGE 296 static int if_vmove(struct ifnet *, struct vnet *); 297 #endif 298 299 #ifdef INET6 300 /* 301 * XXX: declare here to avoid to include many inet6 related files.. 302 * should be more generalized? 303 */ 304 extern void nd6_setmtu(struct ifnet *); 305 #endif 306 307 /* ipsec helper hooks */ 308 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 309 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 310 311 int ifqmaxlen = IFQ_MAXLEN; 312 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 313 VNET_DEFINE(struct ifgrouphead, ifg_head); 314 315 /* Table of ifnet by index. */ 316 static int if_index; 317 static int if_indexlim = 8; 318 static struct ifindex_entry { 319 struct ifnet *ife_ifnet; 320 uint16_t ife_gencnt; 321 } *ifindex_table; 322 323 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system, 324 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 325 "Variables global to all interfaces"); 326 static int 327 sysctl_ifcount(SYSCTL_HANDLER_ARGS) 328 { 329 int rv = 0; 330 331 IFNET_RLOCK(); 332 for (int i = 1; i <= if_index; i++) 333 if (ifindex_table[i].ife_ifnet != NULL && 334 ifindex_table[i].ife_ifnet->if_vnet == curvnet) 335 rv = i; 336 IFNET_RUNLOCK(); 337 338 return (sysctl_handle_int(oidp, &rv, 0, req)); 339 } 340 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount, 341 CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I", 342 "Maximum known interface index"); 343 344 /* 345 * The global network interface list (V_ifnet) and related state (such as 346 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 347 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 348 */ 349 struct sx ifnet_sxlock; 350 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 351 352 struct sx ifnet_detach_sxlock; 353 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 354 SX_RECURSE); 355 356 #ifdef VIMAGE 357 #define VNET_IS_SHUTTING_DOWN(_vnet) \ 358 ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE) 359 #endif 360 361 static if_com_alloc_t *if_com_alloc[256]; 362 static if_com_free_t *if_com_free[256]; 363 364 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 365 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 366 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 367 368 struct ifnet * 369 ifnet_byindex(u_int idx) 370 { 371 struct ifnet *ifp; 372 373 NET_EPOCH_ASSERT(); 374 375 if (__predict_false(idx > if_index)) 376 return (NULL); 377 378 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 379 380 if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet) 381 ifp = NULL; 382 383 return (ifp); 384 } 385 386 struct ifnet * 387 ifnet_byindex_ref(u_int idx) 388 { 389 struct ifnet *ifp; 390 391 ifp = ifnet_byindex(idx); 392 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 393 return (NULL); 394 if (!if_try_ref(ifp)) 395 return (NULL); 396 return (ifp); 397 } 398 399 struct ifnet * 400 ifnet_byindexgen(uint16_t idx, uint16_t gen) 401 { 402 struct ifnet *ifp; 403 404 NET_EPOCH_ASSERT(); 405 406 if (__predict_false(idx > if_index)) 407 return (NULL); 408 409 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 410 411 if (ifindex_table[idx].ife_gencnt == gen) 412 return (ifp); 413 else 414 return (NULL); 415 } 416 417 /* 418 * Network interface utility routines. 419 * 420 * Routines with ifa_ifwith* names take sockaddr *'s as 421 * parameters. 422 */ 423 424 static void 425 if_init_idxtable(void *arg __unused) 426 { 427 428 ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table), 429 M_IFNET, M_WAITOK | M_ZERO); 430 } 431 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL); 432 433 static void 434 vnet_if_init(const void *unused __unused) 435 { 436 437 CK_STAILQ_INIT(&V_ifnet); 438 CK_STAILQ_INIT(&V_ifg_head); 439 vnet_if_clone_init(); 440 } 441 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 442 NULL); 443 444 static void 445 if_link_ifnet(struct ifnet *ifp) 446 { 447 448 IFNET_WLOCK(); 449 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 450 #ifdef VIMAGE 451 curvnet->vnet_ifcnt++; 452 #endif 453 IFNET_WUNLOCK(); 454 } 455 456 static bool 457 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 458 { 459 struct ifnet *iter; 460 int found = 0; 461 462 IFNET_WLOCK(); 463 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 464 if (iter == ifp) { 465 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 466 if (!vmove) 467 ifp->if_flags |= IFF_DYING; 468 found = 1; 469 break; 470 } 471 #ifdef VIMAGE 472 curvnet->vnet_ifcnt--; 473 #endif 474 IFNET_WUNLOCK(); 475 476 return (found); 477 } 478 479 #ifdef VIMAGE 480 static void 481 vnet_if_return(const void *unused __unused) 482 { 483 struct ifnet *ifp, *nifp; 484 struct ifnet **pending; 485 int found __diagused; 486 int i; 487 488 i = 0; 489 490 /* 491 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 492 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 493 * if_detach_internal(), which waits for NET_EPOCH callbacks to 494 * complete. We can't do that from within NET_EPOCH. 495 * 496 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 497 * read/write lock. We cannot hold the lock as we call if_vmove() 498 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 499 * ctx lock. 500 */ 501 IFNET_WLOCK(); 502 503 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 504 M_IFNET, M_WAITOK | M_ZERO); 505 506 /* Return all inherited interfaces to their parent vnets. */ 507 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 508 if (ifp->if_home_vnet != ifp->if_vnet) { 509 found = if_unlink_ifnet(ifp, true); 510 MPASS(found); 511 512 pending[i++] = ifp; 513 } 514 } 515 IFNET_WUNLOCK(); 516 517 for (int j = 0; j < i; j++) { 518 sx_xlock(&ifnet_detach_sxlock); 519 if_vmove(pending[j], pending[j]->if_home_vnet); 520 sx_xunlock(&ifnet_detach_sxlock); 521 } 522 523 free(pending, M_IFNET); 524 } 525 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 526 vnet_if_return, NULL); 527 #endif 528 529 /* 530 * Allocate a struct ifnet and an index for an interface. A layer 2 531 * common structure will also be allocated if an allocation routine is 532 * registered for the passed type. 533 */ 534 static struct ifnet * 535 if_alloc_domain(u_char type, int numa_domain) 536 { 537 struct ifnet *ifp; 538 u_short idx; 539 540 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 541 if (numa_domain == IF_NODOM) 542 ifp = malloc(sizeof(struct ifnet), M_IFNET, 543 M_WAITOK | M_ZERO); 544 else 545 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 546 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 547 ifp->if_type = type; 548 ifp->if_alloctype = type; 549 ifp->if_numa_domain = numa_domain; 550 #ifdef VIMAGE 551 ifp->if_vnet = curvnet; 552 #endif 553 if (if_com_alloc[type] != NULL) { 554 ifp->if_l2com = if_com_alloc[type](type, ifp); 555 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 556 type)); 557 } 558 559 IF_ADDR_LOCK_INIT(ifp); 560 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 561 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 562 ifp->if_afdata_initialized = 0; 563 IF_AFDATA_LOCK_INIT(ifp); 564 CK_STAILQ_INIT(&ifp->if_addrhead); 565 CK_STAILQ_INIT(&ifp->if_multiaddrs); 566 CK_STAILQ_INIT(&ifp->if_groups); 567 #ifdef MAC 568 mac_ifnet_init(ifp); 569 #endif 570 ifq_init(&ifp->if_snd, ifp); 571 572 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 573 for (int i = 0; i < IFCOUNTERS; i++) 574 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 575 ifp->if_get_counter = if_get_counter_default; 576 ifp->if_pcp = IFNET_PCP_NONE; 577 578 /* Allocate an ifindex array entry. */ 579 IFNET_WLOCK(); 580 /* 581 * Try to find an empty slot below if_index. If we fail, take the 582 * next slot. 583 */ 584 for (idx = 1; idx <= if_index; idx++) { 585 if (ifindex_table[idx].ife_ifnet == NULL) 586 break; 587 } 588 589 /* Catch if_index overflow. */ 590 if (idx >= if_indexlim) { 591 struct ifindex_entry *new, *old; 592 int newlim; 593 594 newlim = if_indexlim * 2; 595 new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO); 596 memcpy(new, ifindex_table, if_indexlim * sizeof(*new)); 597 old = ifindex_table; 598 ck_pr_store_ptr(&ifindex_table, new); 599 if_indexlim = newlim; 600 epoch_wait_preempt(net_epoch_preempt); 601 free(old, M_IFNET); 602 } 603 if (idx > if_index) 604 if_index = idx; 605 606 ifp->if_index = idx; 607 ifp->if_idxgen = ifindex_table[idx].ife_gencnt; 608 ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp); 609 IFNET_WUNLOCK(); 610 611 return (ifp); 612 } 613 614 struct ifnet * 615 if_alloc_dev(u_char type, device_t dev) 616 { 617 int numa_domain; 618 619 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 620 return (if_alloc_domain(type, IF_NODOM)); 621 return (if_alloc_domain(type, numa_domain)); 622 } 623 624 struct ifnet * 625 if_alloc(u_char type) 626 { 627 628 return (if_alloc_domain(type, IF_NODOM)); 629 } 630 /* 631 * Do the actual work of freeing a struct ifnet, and layer 2 common 632 * structure. This call is made when the network epoch guarantees 633 * us that nobody holds a pointer to the interface. 634 */ 635 static void 636 if_free_deferred(epoch_context_t ctx) 637 { 638 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 639 640 KASSERT((ifp->if_flags & IFF_DYING), 641 ("%s: interface not dying", __func__)); 642 643 if (if_com_free[ifp->if_alloctype] != NULL) 644 if_com_free[ifp->if_alloctype](ifp->if_l2com, 645 ifp->if_alloctype); 646 647 #ifdef MAC 648 mac_ifnet_destroy(ifp); 649 #endif /* MAC */ 650 IF_AFDATA_DESTROY(ifp); 651 IF_ADDR_LOCK_DESTROY(ifp); 652 ifq_delete(&ifp->if_snd); 653 654 for (int i = 0; i < IFCOUNTERS; i++) 655 counter_u64_free(ifp->if_counters[i]); 656 657 if_freedescr(ifp->if_description); 658 free(ifp->if_hw_addr, M_IFADDR); 659 free(ifp, M_IFNET); 660 } 661 662 /* 663 * Deregister an interface and free the associated storage. 664 */ 665 void 666 if_free(struct ifnet *ifp) 667 { 668 669 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 670 671 /* 672 * XXXGL: An interface index is really an alias to ifp pointer. 673 * Why would we clear the alias now, and not in the deferred 674 * context? Indeed there is nothing wrong with some network 675 * thread obtaining ifp via ifnet_byindex() inside the network 676 * epoch and then dereferencing ifp while we perform if_free(), 677 * and after if_free() finished, too. 678 * 679 * This early index freeing was important back when ifindex was 680 * virtualized and interface would outlive the vnet. 681 */ 682 IFNET_WLOCK(); 683 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 684 ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL); 685 ifindex_table[ifp->if_index].ife_gencnt++; 686 while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL) 687 if_index--; 688 IFNET_WUNLOCK(); 689 690 if (refcount_release(&ifp->if_refcount)) 691 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 692 } 693 694 /* 695 * Interfaces to keep an ifnet type-stable despite the possibility of the 696 * driver calling if_free(). If there are additional references, we defer 697 * freeing the underlying data structure. 698 */ 699 void 700 if_ref(struct ifnet *ifp) 701 { 702 u_int old __diagused; 703 704 /* We don't assert the ifnet list lock here, but arguably should. */ 705 old = refcount_acquire(&ifp->if_refcount); 706 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 707 } 708 709 bool 710 if_try_ref(struct ifnet *ifp) 711 { 712 NET_EPOCH_ASSERT(); 713 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 714 } 715 716 void 717 if_rele(struct ifnet *ifp) 718 { 719 720 if (!refcount_release(&ifp->if_refcount)) 721 return; 722 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 723 } 724 725 void 726 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 727 { 728 729 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 730 731 if (ifq->ifq_maxlen == 0) 732 ifq->ifq_maxlen = ifqmaxlen; 733 734 ifq->altq_type = 0; 735 ifq->altq_disc = NULL; 736 ifq->altq_flags &= ALTQF_CANTCHANGE; 737 ifq->altq_tbr = NULL; 738 ifq->altq_ifp = ifp; 739 } 740 741 void 742 ifq_delete(struct ifaltq *ifq) 743 { 744 mtx_destroy(&ifq->ifq_mtx); 745 } 746 747 /* 748 * Perform generic interface initialization tasks and attach the interface 749 * to the list of "active" interfaces. If vmove flag is set on entry 750 * to if_attach_internal(), perform only a limited subset of initialization 751 * tasks, given that we are moving from one vnet to another an ifnet which 752 * has already been fully initialized. 753 * 754 * Note that if_detach_internal() removes group membership unconditionally 755 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 756 * Thus, when if_vmove() is applied to a cloned interface, group membership 757 * is lost while a cloned one always joins a group whose name is 758 * ifc->ifc_name. To recover this after if_detach_internal() and 759 * if_attach_internal(), the cloner should be specified to 760 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 761 * attempts to join a group whose name is ifc->ifc_name. 762 * 763 * XXX: 764 * - The decision to return void and thus require this function to 765 * succeed is questionable. 766 * - We should probably do more sanity checking. For instance we don't 767 * do anything to insure if_xname is unique or non-empty. 768 */ 769 void 770 if_attach(struct ifnet *ifp) 771 { 772 773 if_attach_internal(ifp, false); 774 } 775 776 /* 777 * Compute the least common TSO limit. 778 */ 779 void 780 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 781 { 782 /* 783 * 1) If there is no limit currently, take the limit from 784 * the network adapter. 785 * 786 * 2) If the network adapter has a limit below the current 787 * limit, apply it. 788 */ 789 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 790 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 791 pmax->tsomaxbytes = ifp->if_hw_tsomax; 792 } 793 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 794 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 795 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 796 } 797 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 798 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 799 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 800 } 801 } 802 803 /* 804 * Update TSO limit of a network adapter. 805 * 806 * Returns zero if no change. Else non-zero. 807 */ 808 int 809 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 810 { 811 int retval = 0; 812 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 813 ifp->if_hw_tsomax = pmax->tsomaxbytes; 814 retval++; 815 } 816 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 817 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 818 retval++; 819 } 820 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 821 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 822 retval++; 823 } 824 return (retval); 825 } 826 827 static void 828 if_attach_internal(struct ifnet *ifp, bool vmove) 829 { 830 unsigned socksize, ifasize; 831 int namelen, masklen; 832 struct sockaddr_dl *sdl; 833 struct ifaddr *ifa; 834 835 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 836 837 #ifdef VIMAGE 838 ifp->if_vnet = curvnet; 839 if (ifp->if_home_vnet == NULL) 840 ifp->if_home_vnet = curvnet; 841 #endif 842 843 if_addgroup(ifp, IFG_ALL); 844 845 #ifdef VIMAGE 846 /* Restore group membership for cloned interface. */ 847 if (vmove) 848 if_clone_restoregroup(ifp); 849 #endif 850 851 getmicrotime(&ifp->if_lastchange); 852 ifp->if_epoch = time_uptime; 853 854 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 855 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 856 ("transmit and qflush must both either be set or both be NULL")); 857 if (ifp->if_transmit == NULL) { 858 ifp->if_transmit = if_transmit_default; 859 ifp->if_qflush = if_qflush; 860 } 861 if (ifp->if_input == NULL) 862 ifp->if_input = if_input_default; 863 864 if (ifp->if_requestencap == NULL) 865 ifp->if_requestencap = if_requestencap_default; 866 867 if (!vmove) { 868 #ifdef MAC 869 mac_ifnet_create(ifp); 870 #endif 871 872 /* 873 * Create a Link Level name for this device. 874 */ 875 namelen = strlen(ifp->if_xname); 876 /* 877 * Always save enough space for any possiable name so we 878 * can do a rename in place later. 879 */ 880 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 881 socksize = masklen + ifp->if_addrlen; 882 if (socksize < sizeof(*sdl)) 883 socksize = sizeof(*sdl); 884 socksize = roundup2(socksize, sizeof(long)); 885 ifasize = sizeof(*ifa) + 2 * socksize; 886 ifa = ifa_alloc(ifasize, M_WAITOK); 887 sdl = (struct sockaddr_dl *)(ifa + 1); 888 sdl->sdl_len = socksize; 889 sdl->sdl_family = AF_LINK; 890 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 891 sdl->sdl_nlen = namelen; 892 sdl->sdl_index = ifp->if_index; 893 sdl->sdl_type = ifp->if_type; 894 ifp->if_addr = ifa; 895 ifa->ifa_ifp = ifp; 896 ifa->ifa_addr = (struct sockaddr *)sdl; 897 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 898 ifa->ifa_netmask = (struct sockaddr *)sdl; 899 sdl->sdl_len = masklen; 900 while (namelen != 0) 901 sdl->sdl_data[--namelen] = 0xff; 902 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 903 /* Reliably crash if used uninitialized. */ 904 ifp->if_broadcastaddr = NULL; 905 906 if (ifp->if_type == IFT_ETHER) { 907 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 908 M_WAITOK | M_ZERO); 909 } 910 911 #if defined(INET) || defined(INET6) 912 /* Use defaults for TSO, if nothing is set */ 913 if (ifp->if_hw_tsomax == 0 && 914 ifp->if_hw_tsomaxsegcount == 0 && 915 ifp->if_hw_tsomaxsegsize == 0) { 916 /* 917 * The TSO defaults needs to be such that an 918 * NFS mbuf list of 35 mbufs totalling just 919 * below 64K works and that a chain of mbufs 920 * can be defragged into at most 32 segments: 921 */ 922 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 923 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 924 ifp->if_hw_tsomaxsegcount = 35; 925 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 926 927 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 928 if (ifp->if_capabilities & IFCAP_TSO) { 929 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 930 ifp->if_hw_tsomax, 931 ifp->if_hw_tsomaxsegcount, 932 ifp->if_hw_tsomaxsegsize); 933 } 934 } 935 #endif 936 } 937 #ifdef VIMAGE 938 else { 939 /* 940 * Update the interface index in the link layer address 941 * of the interface. 942 */ 943 for (ifa = ifp->if_addr; ifa != NULL; 944 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 945 if (ifa->ifa_addr->sa_family == AF_LINK) { 946 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 947 sdl->sdl_index = ifp->if_index; 948 } 949 } 950 } 951 #endif 952 953 if_link_ifnet(ifp); 954 955 if (domain_init_status >= 2) 956 if_attachdomain1(ifp); 957 958 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 959 if (IS_DEFAULT_VNET(curvnet)) 960 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 961 } 962 963 static void 964 if_epochalloc(void *dummy __unused) 965 { 966 967 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 968 } 969 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 970 971 static void 972 if_attachdomain(void *dummy) 973 { 974 struct ifnet *ifp; 975 976 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 977 if_attachdomain1(ifp); 978 } 979 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 980 if_attachdomain, NULL); 981 982 static void 983 if_attachdomain1(struct ifnet *ifp) 984 { 985 struct domain *dp; 986 987 /* 988 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 989 * cannot lock ifp->if_afdata initialization, entirely. 990 */ 991 IF_AFDATA_LOCK(ifp); 992 if (ifp->if_afdata_initialized >= domain_init_status) { 993 IF_AFDATA_UNLOCK(ifp); 994 log(LOG_WARNING, "%s called more than once on %s\n", 995 __func__, ifp->if_xname); 996 return; 997 } 998 ifp->if_afdata_initialized = domain_init_status; 999 IF_AFDATA_UNLOCK(ifp); 1000 1001 /* address family dependent data region */ 1002 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 1003 SLIST_FOREACH(dp, &domains, dom_next) { 1004 if (dp->dom_ifattach) 1005 ifp->if_afdata[dp->dom_family] = 1006 (*dp->dom_ifattach)(ifp); 1007 } 1008 } 1009 1010 /* 1011 * Remove any unicast or broadcast network addresses from an interface. 1012 */ 1013 void 1014 if_purgeaddrs(struct ifnet *ifp) 1015 { 1016 struct ifaddr *ifa; 1017 1018 #ifdef INET6 1019 /* 1020 * Need to leave multicast addresses of proxy NDP llentries 1021 * before in6_purgeifaddr() because the llentries are keys 1022 * for in6_multi objects of proxy NDP entries. 1023 * in6_purgeifaddr()s clean up llentries including proxy NDPs 1024 * then we would lose the keys if they are called earlier. 1025 */ 1026 in6_purge_proxy_ndp(ifp); 1027 #endif 1028 while (1) { 1029 struct epoch_tracker et; 1030 1031 NET_EPOCH_ENTER(et); 1032 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1033 if (ifa->ifa_addr->sa_family != AF_LINK) 1034 break; 1035 } 1036 NET_EPOCH_EXIT(et); 1037 1038 if (ifa == NULL) 1039 break; 1040 #ifdef INET 1041 /* XXX: Ugly!! ad hoc just for INET */ 1042 if (ifa->ifa_addr->sa_family == AF_INET) { 1043 struct ifaliasreq ifr; 1044 1045 bzero(&ifr, sizeof(ifr)); 1046 ifr.ifra_addr = *ifa->ifa_addr; 1047 if (ifa->ifa_dstaddr) 1048 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1049 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1050 NULL) == 0) 1051 continue; 1052 } 1053 #endif /* INET */ 1054 #ifdef INET6 1055 if (ifa->ifa_addr->sa_family == AF_INET6) { 1056 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1057 /* ifp_addrhead is already updated */ 1058 continue; 1059 } 1060 #endif /* INET6 */ 1061 IF_ADDR_WLOCK(ifp); 1062 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1063 IF_ADDR_WUNLOCK(ifp); 1064 ifa_free(ifa); 1065 } 1066 } 1067 1068 /* 1069 * Remove any multicast network addresses from an interface when an ifnet 1070 * is going away. 1071 */ 1072 static void 1073 if_purgemaddrs(struct ifnet *ifp) 1074 { 1075 struct ifmultiaddr *ifma; 1076 1077 IF_ADDR_WLOCK(ifp); 1078 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1079 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1080 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1081 if_delmulti_locked(ifp, ifma, 1); 1082 } 1083 IF_ADDR_WUNLOCK(ifp); 1084 } 1085 1086 /* 1087 * Detach an interface, removing it from the list of "active" interfaces. 1088 * If vmove flag is set on entry to if_detach_internal(), perform only a 1089 * limited subset of cleanup tasks, given that we are moving an ifnet from 1090 * one vnet to another, where it must be fully operational. 1091 * 1092 * XXXRW: There are some significant questions about event ordering, and 1093 * how to prevent things from starting to use the interface during detach. 1094 */ 1095 void 1096 if_detach(struct ifnet *ifp) 1097 { 1098 bool found; 1099 1100 CURVNET_SET_QUIET(ifp->if_vnet); 1101 found = if_unlink_ifnet(ifp, false); 1102 if (found) { 1103 sx_xlock(&ifnet_detach_sxlock); 1104 if_detach_internal(ifp, false); 1105 sx_xunlock(&ifnet_detach_sxlock); 1106 } 1107 CURVNET_RESTORE(); 1108 } 1109 1110 /* 1111 * The vmove flag, if set, indicates that we are called from a callpath 1112 * that is moving an interface to a different vnet instance. 1113 * 1114 * The shutdown flag, if set, indicates that we are called in the 1115 * process of shutting down a vnet instance. Currently only the 1116 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1117 * on a vnet instance shutdown without this flag being set, e.g., when 1118 * the cloned interfaces are destoyed as first thing of teardown. 1119 */ 1120 static int 1121 if_detach_internal(struct ifnet *ifp, bool vmove) 1122 { 1123 struct ifaddr *ifa; 1124 int i; 1125 struct domain *dp; 1126 #ifdef VIMAGE 1127 bool shutdown; 1128 1129 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1130 #endif 1131 1132 /* 1133 * At this point we know the interface still was on the ifnet list 1134 * and we removed it so we are in a stable state. 1135 */ 1136 epoch_wait_preempt(net_epoch_preempt); 1137 1138 /* 1139 * Ensure all pending EPOCH(9) callbacks have been executed. This 1140 * fixes issues about late destruction of multicast options 1141 * which lead to leave group calls, which in turn access the 1142 * belonging ifnet structure: 1143 */ 1144 NET_EPOCH_DRAIN_CALLBACKS(); 1145 1146 /* 1147 * In any case (destroy or vmove) detach us from the groups 1148 * and remove/wait for pending events on the taskq. 1149 * XXX-BZ in theory an interface could still enqueue a taskq change? 1150 */ 1151 if_delgroups(ifp); 1152 1153 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1154 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1155 1156 if_down(ifp); 1157 1158 #ifdef VIMAGE 1159 /* 1160 * On VNET shutdown abort here as the stack teardown will do all 1161 * the work top-down for us. 1162 */ 1163 if (shutdown) { 1164 /* Give interface users the chance to clean up. */ 1165 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1166 1167 /* 1168 * In case of a vmove we are done here without error. 1169 * If we would signal an error it would lead to the same 1170 * abort as if we did not find the ifnet anymore. 1171 * if_detach() calls us in void context and does not care 1172 * about an early abort notification, so life is splendid :) 1173 */ 1174 goto finish_vnet_shutdown; 1175 } 1176 #endif 1177 1178 /* 1179 * At this point we are not tearing down a VNET and are either 1180 * going to destroy or vmove the interface and have to cleanup 1181 * accordingly. 1182 */ 1183 1184 /* 1185 * Remove routes and flush queues. 1186 */ 1187 #ifdef ALTQ 1188 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1189 altq_disable(&ifp->if_snd); 1190 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1191 altq_detach(&ifp->if_snd); 1192 #endif 1193 1194 if_purgeaddrs(ifp); 1195 1196 #ifdef INET 1197 in_ifdetach(ifp); 1198 #endif 1199 1200 #ifdef INET6 1201 /* 1202 * Remove all IPv6 kernel structs related to ifp. This should be done 1203 * before removing routing entries below, since IPv6 interface direct 1204 * routes are expected to be removed by the IPv6-specific kernel API. 1205 * Otherwise, the kernel will detect some inconsistency and bark it. 1206 */ 1207 in6_ifdetach(ifp); 1208 #endif 1209 if_purgemaddrs(ifp); 1210 1211 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1212 if (IS_DEFAULT_VNET(curvnet)) 1213 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1214 1215 if (!vmove) { 1216 /* 1217 * Prevent further calls into the device driver via ifnet. 1218 */ 1219 if_dead(ifp); 1220 1221 /* 1222 * Clean up all addresses. 1223 */ 1224 IF_ADDR_WLOCK(ifp); 1225 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1226 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1227 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1228 IF_ADDR_WUNLOCK(ifp); 1229 ifa_free(ifa); 1230 } else 1231 IF_ADDR_WUNLOCK(ifp); 1232 } 1233 1234 rt_flushifroutes(ifp); 1235 1236 #ifdef VIMAGE 1237 finish_vnet_shutdown: 1238 #endif 1239 /* 1240 * We cannot hold the lock over dom_ifdetach calls as they might 1241 * sleep, for example trying to drain a callout, thus open up the 1242 * theoretical race with re-attaching. 1243 */ 1244 IF_AFDATA_LOCK(ifp); 1245 i = ifp->if_afdata_initialized; 1246 ifp->if_afdata_initialized = 0; 1247 IF_AFDATA_UNLOCK(ifp); 1248 if (i == 0) 1249 return (0); 1250 SLIST_FOREACH(dp, &domains, dom_next) { 1251 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1252 (*dp->dom_ifdetach)(ifp, 1253 ifp->if_afdata[dp->dom_family]); 1254 ifp->if_afdata[dp->dom_family] = NULL; 1255 } 1256 } 1257 1258 return (0); 1259 } 1260 1261 #ifdef VIMAGE 1262 /* 1263 * if_vmove() performs a limited version of if_detach() in current 1264 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1265 */ 1266 static int 1267 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1268 { 1269 #ifdef DEV_BPF 1270 u_int bif_dlt, bif_hdrlen; 1271 #endif 1272 int rc; 1273 1274 #ifdef DEV_BPF 1275 /* 1276 * if_detach_internal() will call the eventhandler to notify 1277 * interface departure. That will detach if_bpf. We need to 1278 * safe the dlt and hdrlen so we can re-attach it later. 1279 */ 1280 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1281 #endif 1282 1283 /* 1284 * Detach from current vnet, but preserve LLADDR info, do not 1285 * mark as dead etc. so that the ifnet can be reattached later. 1286 * If we cannot find it, we lost the race to someone else. 1287 */ 1288 rc = if_detach_internal(ifp, true); 1289 if (rc != 0) 1290 return (rc); 1291 1292 /* 1293 * Perform interface-specific reassignment tasks, if provided by 1294 * the driver. 1295 */ 1296 if (ifp->if_reassign != NULL) 1297 ifp->if_reassign(ifp, new_vnet, NULL); 1298 1299 /* 1300 * Switch to the context of the target vnet. 1301 */ 1302 CURVNET_SET_QUIET(new_vnet); 1303 if_attach_internal(ifp, true); 1304 1305 #ifdef DEV_BPF 1306 if (ifp->if_bpf == NULL) 1307 bpfattach(ifp, bif_dlt, bif_hdrlen); 1308 #endif 1309 1310 CURVNET_RESTORE(); 1311 return (0); 1312 } 1313 1314 /* 1315 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1316 */ 1317 static int 1318 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1319 { 1320 struct prison *pr; 1321 struct ifnet *difp; 1322 int error; 1323 bool found __diagused; 1324 bool shutdown; 1325 1326 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 1327 1328 /* Try to find the prison within our visibility. */ 1329 sx_slock(&allprison_lock); 1330 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1331 sx_sunlock(&allprison_lock); 1332 if (pr == NULL) 1333 return (ENXIO); 1334 prison_hold_locked(pr); 1335 mtx_unlock(&pr->pr_mtx); 1336 1337 /* Do not try to move the iface from and to the same prison. */ 1338 if (pr->pr_vnet == ifp->if_vnet) { 1339 prison_free(pr); 1340 return (EEXIST); 1341 } 1342 1343 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1344 /* XXX Lock interfaces to avoid races. */ 1345 CURVNET_SET_QUIET(pr->pr_vnet); 1346 difp = ifunit(ifname); 1347 if (difp != NULL) { 1348 CURVNET_RESTORE(); 1349 prison_free(pr); 1350 return (EEXIST); 1351 } 1352 sx_xlock(&ifnet_detach_sxlock); 1353 1354 /* Make sure the VNET is stable. */ 1355 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1356 if (shutdown) { 1357 sx_xunlock(&ifnet_detach_sxlock); 1358 CURVNET_RESTORE(); 1359 prison_free(pr); 1360 return (EBUSY); 1361 } 1362 CURVNET_RESTORE(); 1363 1364 found = if_unlink_ifnet(ifp, true); 1365 if (! found) { 1366 sx_xunlock(&ifnet_detach_sxlock); 1367 CURVNET_RESTORE(); 1368 prison_free(pr); 1369 return (ENODEV); 1370 } 1371 1372 /* Move the interface into the child jail/vnet. */ 1373 error = if_vmove(ifp, pr->pr_vnet); 1374 1375 /* Report the new if_xname back to the userland on success. */ 1376 if (error == 0) 1377 sprintf(ifname, "%s", ifp->if_xname); 1378 1379 sx_xunlock(&ifnet_detach_sxlock); 1380 1381 prison_free(pr); 1382 return (error); 1383 } 1384 1385 static int 1386 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1387 { 1388 struct prison *pr; 1389 struct vnet *vnet_dst; 1390 struct ifnet *ifp; 1391 int error, found __diagused; 1392 bool shutdown; 1393 1394 /* Try to find the prison within our visibility. */ 1395 sx_slock(&allprison_lock); 1396 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1397 sx_sunlock(&allprison_lock); 1398 if (pr == NULL) 1399 return (ENXIO); 1400 prison_hold_locked(pr); 1401 mtx_unlock(&pr->pr_mtx); 1402 1403 /* Make sure the named iface exists in the source prison/vnet. */ 1404 CURVNET_SET(pr->pr_vnet); 1405 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1406 if (ifp == NULL) { 1407 CURVNET_RESTORE(); 1408 prison_free(pr); 1409 return (ENXIO); 1410 } 1411 1412 /* Do not try to move the iface from and to the same prison. */ 1413 vnet_dst = TD_TO_VNET(td); 1414 if (vnet_dst == ifp->if_vnet) { 1415 CURVNET_RESTORE(); 1416 prison_free(pr); 1417 return (EEXIST); 1418 } 1419 1420 /* Make sure the VNET is stable. */ 1421 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1422 if (shutdown) { 1423 CURVNET_RESTORE(); 1424 prison_free(pr); 1425 return (EBUSY); 1426 } 1427 1428 /* Get interface back from child jail/vnet. */ 1429 found = if_unlink_ifnet(ifp, true); 1430 MPASS(found); 1431 sx_xlock(&ifnet_detach_sxlock); 1432 error = if_vmove(ifp, vnet_dst); 1433 sx_xunlock(&ifnet_detach_sxlock); 1434 CURVNET_RESTORE(); 1435 1436 /* Report the new if_xname back to the userland on success. */ 1437 if (error == 0) 1438 sprintf(ifname, "%s", ifp->if_xname); 1439 1440 prison_free(pr); 1441 return (error); 1442 } 1443 #endif /* VIMAGE */ 1444 1445 /* 1446 * Add a group to an interface 1447 */ 1448 int 1449 if_addgroup(struct ifnet *ifp, const char *groupname) 1450 { 1451 struct ifg_list *ifgl; 1452 struct ifg_group *ifg = NULL; 1453 struct ifg_member *ifgm; 1454 int new = 0; 1455 1456 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1457 groupname[strlen(groupname) - 1] <= '9') 1458 return (EINVAL); 1459 1460 IFNET_WLOCK(); 1461 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1462 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1463 IFNET_WUNLOCK(); 1464 return (EEXIST); 1465 } 1466 1467 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1468 IFNET_WUNLOCK(); 1469 return (ENOMEM); 1470 } 1471 1472 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1473 free(ifgl, M_TEMP); 1474 IFNET_WUNLOCK(); 1475 return (ENOMEM); 1476 } 1477 1478 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1479 if (!strcmp(ifg->ifg_group, groupname)) 1480 break; 1481 1482 if (ifg == NULL) { 1483 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1484 free(ifgl, M_TEMP); 1485 free(ifgm, M_TEMP); 1486 IFNET_WUNLOCK(); 1487 return (ENOMEM); 1488 } 1489 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1490 ifg->ifg_refcnt = 0; 1491 CK_STAILQ_INIT(&ifg->ifg_members); 1492 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1493 new = 1; 1494 } 1495 1496 ifg->ifg_refcnt++; 1497 ifgl->ifgl_group = ifg; 1498 ifgm->ifgm_ifp = ifp; 1499 1500 IF_ADDR_WLOCK(ifp); 1501 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1502 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1503 IF_ADDR_WUNLOCK(ifp); 1504 1505 IFNET_WUNLOCK(); 1506 1507 if (new) 1508 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1509 EVENTHANDLER_INVOKE(group_change_event, groupname); 1510 1511 return (0); 1512 } 1513 1514 /* 1515 * Helper function to remove a group out of an interface. Expects the global 1516 * ifnet lock to be write-locked, and drops it before returning. 1517 */ 1518 static void 1519 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1520 const char *groupname) 1521 { 1522 struct ifg_member *ifgm; 1523 bool freeifgl; 1524 1525 IFNET_WLOCK_ASSERT(); 1526 1527 IF_ADDR_WLOCK(ifp); 1528 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1529 IF_ADDR_WUNLOCK(ifp); 1530 1531 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1532 if (ifgm->ifgm_ifp == ifp) { 1533 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1534 ifg_member, ifgm_next); 1535 break; 1536 } 1537 } 1538 1539 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1540 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1541 ifg_next); 1542 freeifgl = true; 1543 } else { 1544 freeifgl = false; 1545 } 1546 IFNET_WUNLOCK(); 1547 1548 epoch_wait_preempt(net_epoch_preempt); 1549 EVENTHANDLER_INVOKE(group_change_event, groupname); 1550 if (freeifgl) { 1551 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1552 free(ifgl->ifgl_group, M_TEMP); 1553 } 1554 free(ifgm, M_TEMP); 1555 free(ifgl, M_TEMP); 1556 } 1557 1558 /* 1559 * Remove a group from an interface 1560 */ 1561 int 1562 if_delgroup(struct ifnet *ifp, const char *groupname) 1563 { 1564 struct ifg_list *ifgl; 1565 1566 IFNET_WLOCK(); 1567 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1568 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1569 break; 1570 if (ifgl == NULL) { 1571 IFNET_WUNLOCK(); 1572 return (ENOENT); 1573 } 1574 1575 _if_delgroup_locked(ifp, ifgl, groupname); 1576 1577 return (0); 1578 } 1579 1580 /* 1581 * Remove an interface from all groups 1582 */ 1583 static void 1584 if_delgroups(struct ifnet *ifp) 1585 { 1586 struct ifg_list *ifgl; 1587 char groupname[IFNAMSIZ]; 1588 1589 IFNET_WLOCK(); 1590 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1591 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1592 _if_delgroup_locked(ifp, ifgl, groupname); 1593 IFNET_WLOCK(); 1594 } 1595 IFNET_WUNLOCK(); 1596 } 1597 1598 /* 1599 * Stores all groups from an interface in memory pointed to by ifgr. 1600 */ 1601 static int 1602 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1603 { 1604 int len, error; 1605 struct ifg_list *ifgl; 1606 struct ifg_req ifgrq, *ifgp; 1607 1608 NET_EPOCH_ASSERT(); 1609 1610 if (ifgr->ifgr_len == 0) { 1611 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1612 ifgr->ifgr_len += sizeof(struct ifg_req); 1613 return (0); 1614 } 1615 1616 len = ifgr->ifgr_len; 1617 ifgp = ifgr->ifgr_groups; 1618 /* XXX: wire */ 1619 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1620 if (len < sizeof(ifgrq)) 1621 return (EINVAL); 1622 bzero(&ifgrq, sizeof ifgrq); 1623 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1624 sizeof(ifgrq.ifgrq_group)); 1625 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1626 return (error); 1627 len -= sizeof(ifgrq); 1628 ifgp++; 1629 } 1630 1631 return (0); 1632 } 1633 1634 /* 1635 * Stores all members of a group in memory pointed to by igfr 1636 */ 1637 static int 1638 if_getgroupmembers(struct ifgroupreq *ifgr) 1639 { 1640 struct ifg_group *ifg; 1641 struct ifg_member *ifgm; 1642 struct ifg_req ifgrq, *ifgp; 1643 int len, error; 1644 1645 IFNET_RLOCK(); 1646 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1647 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1648 break; 1649 if (ifg == NULL) { 1650 IFNET_RUNLOCK(); 1651 return (ENOENT); 1652 } 1653 1654 if (ifgr->ifgr_len == 0) { 1655 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1656 ifgr->ifgr_len += sizeof(ifgrq); 1657 IFNET_RUNLOCK(); 1658 return (0); 1659 } 1660 1661 len = ifgr->ifgr_len; 1662 ifgp = ifgr->ifgr_groups; 1663 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1664 if (len < sizeof(ifgrq)) { 1665 IFNET_RUNLOCK(); 1666 return (EINVAL); 1667 } 1668 bzero(&ifgrq, sizeof ifgrq); 1669 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1670 sizeof(ifgrq.ifgrq_member)); 1671 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1672 IFNET_RUNLOCK(); 1673 return (error); 1674 } 1675 len -= sizeof(ifgrq); 1676 ifgp++; 1677 } 1678 IFNET_RUNLOCK(); 1679 1680 return (0); 1681 } 1682 1683 /* 1684 * Return counter values from counter(9)s stored in ifnet. 1685 */ 1686 uint64_t 1687 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1688 { 1689 1690 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1691 1692 return (counter_u64_fetch(ifp->if_counters[cnt])); 1693 } 1694 1695 /* 1696 * Increase an ifnet counter. Usually used for counters shared 1697 * between the stack and a driver, but function supports them all. 1698 */ 1699 void 1700 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1701 { 1702 1703 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1704 1705 counter_u64_add(ifp->if_counters[cnt], inc); 1706 } 1707 1708 /* 1709 * Copy data from ifnet to userland API structure if_data. 1710 */ 1711 void 1712 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1713 { 1714 1715 ifd->ifi_type = ifp->if_type; 1716 ifd->ifi_physical = 0; 1717 ifd->ifi_addrlen = ifp->if_addrlen; 1718 ifd->ifi_hdrlen = ifp->if_hdrlen; 1719 ifd->ifi_link_state = ifp->if_link_state; 1720 ifd->ifi_vhid = 0; 1721 ifd->ifi_datalen = sizeof(struct if_data); 1722 ifd->ifi_mtu = ifp->if_mtu; 1723 ifd->ifi_metric = ifp->if_metric; 1724 ifd->ifi_baudrate = ifp->if_baudrate; 1725 ifd->ifi_hwassist = ifp->if_hwassist; 1726 ifd->ifi_epoch = ifp->if_epoch; 1727 ifd->ifi_lastchange = ifp->if_lastchange; 1728 1729 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1730 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1731 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1732 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1733 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1734 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1735 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1736 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1737 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1738 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1739 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1740 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1741 } 1742 1743 /* 1744 * Initialization, destruction and refcounting functions for ifaddrs. 1745 */ 1746 struct ifaddr * 1747 ifa_alloc(size_t size, int flags) 1748 { 1749 struct ifaddr *ifa; 1750 1751 KASSERT(size >= sizeof(struct ifaddr), 1752 ("%s: invalid size %zu", __func__, size)); 1753 1754 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1755 if (ifa == NULL) 1756 return (NULL); 1757 1758 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1759 goto fail; 1760 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1761 goto fail; 1762 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1763 goto fail; 1764 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1765 goto fail; 1766 1767 refcount_init(&ifa->ifa_refcnt, 1); 1768 1769 return (ifa); 1770 1771 fail: 1772 /* free(NULL) is okay */ 1773 counter_u64_free(ifa->ifa_opackets); 1774 counter_u64_free(ifa->ifa_ipackets); 1775 counter_u64_free(ifa->ifa_obytes); 1776 counter_u64_free(ifa->ifa_ibytes); 1777 free(ifa, M_IFADDR); 1778 1779 return (NULL); 1780 } 1781 1782 void 1783 ifa_ref(struct ifaddr *ifa) 1784 { 1785 u_int old __diagused; 1786 1787 old = refcount_acquire(&ifa->ifa_refcnt); 1788 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1789 } 1790 1791 int 1792 ifa_try_ref(struct ifaddr *ifa) 1793 { 1794 1795 NET_EPOCH_ASSERT(); 1796 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1797 } 1798 1799 static void 1800 ifa_destroy(epoch_context_t ctx) 1801 { 1802 struct ifaddr *ifa; 1803 1804 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1805 counter_u64_free(ifa->ifa_opackets); 1806 counter_u64_free(ifa->ifa_ipackets); 1807 counter_u64_free(ifa->ifa_obytes); 1808 counter_u64_free(ifa->ifa_ibytes); 1809 free(ifa, M_IFADDR); 1810 } 1811 1812 void 1813 ifa_free(struct ifaddr *ifa) 1814 { 1815 1816 if (refcount_release(&ifa->ifa_refcnt)) 1817 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1818 } 1819 1820 /* 1821 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1822 * structs used to represent other address families, it is necessary 1823 * to perform a different comparison. 1824 */ 1825 1826 #define sa_dl_equal(a1, a2) \ 1827 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1828 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1829 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1830 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1831 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1832 1833 /* 1834 * Locate an interface based on a complete address. 1835 */ 1836 /*ARGSUSED*/ 1837 struct ifaddr * 1838 ifa_ifwithaddr(const struct sockaddr *addr) 1839 { 1840 struct ifnet *ifp; 1841 struct ifaddr *ifa; 1842 1843 NET_EPOCH_ASSERT(); 1844 1845 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1846 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1847 if (ifa->ifa_addr->sa_family != addr->sa_family) 1848 continue; 1849 if (sa_equal(addr, ifa->ifa_addr)) { 1850 goto done; 1851 } 1852 /* IP6 doesn't have broadcast */ 1853 if ((ifp->if_flags & IFF_BROADCAST) && 1854 ifa->ifa_broadaddr && 1855 ifa->ifa_broadaddr->sa_len != 0 && 1856 sa_equal(ifa->ifa_broadaddr, addr)) { 1857 goto done; 1858 } 1859 } 1860 } 1861 ifa = NULL; 1862 done: 1863 return (ifa); 1864 } 1865 1866 int 1867 ifa_ifwithaddr_check(const struct sockaddr *addr) 1868 { 1869 struct epoch_tracker et; 1870 int rc; 1871 1872 NET_EPOCH_ENTER(et); 1873 rc = (ifa_ifwithaddr(addr) != NULL); 1874 NET_EPOCH_EXIT(et); 1875 return (rc); 1876 } 1877 1878 /* 1879 * Locate an interface based on the broadcast address. 1880 */ 1881 /* ARGSUSED */ 1882 struct ifaddr * 1883 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1884 { 1885 struct ifnet *ifp; 1886 struct ifaddr *ifa; 1887 1888 NET_EPOCH_ASSERT(); 1889 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1890 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1891 continue; 1892 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1893 if (ifa->ifa_addr->sa_family != addr->sa_family) 1894 continue; 1895 if ((ifp->if_flags & IFF_BROADCAST) && 1896 ifa->ifa_broadaddr && 1897 ifa->ifa_broadaddr->sa_len != 0 && 1898 sa_equal(ifa->ifa_broadaddr, addr)) { 1899 goto done; 1900 } 1901 } 1902 } 1903 ifa = NULL; 1904 done: 1905 return (ifa); 1906 } 1907 1908 /* 1909 * Locate the point to point interface with a given destination address. 1910 */ 1911 /*ARGSUSED*/ 1912 struct ifaddr * 1913 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1914 { 1915 struct ifnet *ifp; 1916 struct ifaddr *ifa; 1917 1918 NET_EPOCH_ASSERT(); 1919 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1920 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1921 continue; 1922 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1923 continue; 1924 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1925 if (ifa->ifa_addr->sa_family != addr->sa_family) 1926 continue; 1927 if (ifa->ifa_dstaddr != NULL && 1928 sa_equal(addr, ifa->ifa_dstaddr)) { 1929 goto done; 1930 } 1931 } 1932 } 1933 ifa = NULL; 1934 done: 1935 return (ifa); 1936 } 1937 1938 /* 1939 * Find an interface on a specific network. If many, choice 1940 * is most specific found. 1941 */ 1942 struct ifaddr * 1943 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1944 { 1945 struct ifnet *ifp; 1946 struct ifaddr *ifa; 1947 struct ifaddr *ifa_maybe = NULL; 1948 u_int af = addr->sa_family; 1949 const char *addr_data = addr->sa_data, *cplim; 1950 1951 NET_EPOCH_ASSERT(); 1952 /* 1953 * AF_LINK addresses can be looked up directly by their index number, 1954 * so do that if we can. 1955 */ 1956 if (af == AF_LINK) { 1957 ifp = ifnet_byindex( 1958 ((const struct sockaddr_dl *)addr)->sdl_index); 1959 return (ifp ? ifp->if_addr : NULL); 1960 } 1961 1962 /* 1963 * Scan though each interface, looking for ones that have addresses 1964 * in this address family and the requested fib. 1965 */ 1966 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1967 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1968 continue; 1969 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1970 const char *cp, *cp2, *cp3; 1971 1972 if (ifa->ifa_addr->sa_family != af) 1973 next: continue; 1974 if (af == AF_INET && 1975 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1976 /* 1977 * This is a bit broken as it doesn't 1978 * take into account that the remote end may 1979 * be a single node in the network we are 1980 * looking for. 1981 * The trouble is that we don't know the 1982 * netmask for the remote end. 1983 */ 1984 if (ifa->ifa_dstaddr != NULL && 1985 sa_equal(addr, ifa->ifa_dstaddr)) { 1986 goto done; 1987 } 1988 } else { 1989 /* 1990 * Scan all the bits in the ifa's address. 1991 * If a bit dissagrees with what we are 1992 * looking for, mask it with the netmask 1993 * to see if it really matters. 1994 * (A byte at a time) 1995 */ 1996 if (ifa->ifa_netmask == 0) 1997 continue; 1998 cp = addr_data; 1999 cp2 = ifa->ifa_addr->sa_data; 2000 cp3 = ifa->ifa_netmask->sa_data; 2001 cplim = ifa->ifa_netmask->sa_len 2002 + (char *)ifa->ifa_netmask; 2003 while (cp3 < cplim) 2004 if ((*cp++ ^ *cp2++) & *cp3++) 2005 goto next; /* next address! */ 2006 /* 2007 * If the netmask of what we just found 2008 * is more specific than what we had before 2009 * (if we had one), or if the virtual status 2010 * of new prefix is better than of the old one, 2011 * then remember the new one before continuing 2012 * to search for an even better one. 2013 */ 2014 if (ifa_maybe == NULL || 2015 ifa_preferred(ifa_maybe, ifa) || 2016 rn_refines((caddr_t)ifa->ifa_netmask, 2017 (caddr_t)ifa_maybe->ifa_netmask)) { 2018 ifa_maybe = ifa; 2019 } 2020 } 2021 } 2022 } 2023 ifa = ifa_maybe; 2024 ifa_maybe = NULL; 2025 done: 2026 return (ifa); 2027 } 2028 2029 /* 2030 * Find an interface address specific to an interface best matching 2031 * a given address. 2032 */ 2033 struct ifaddr * 2034 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2035 { 2036 struct ifaddr *ifa; 2037 const char *cp, *cp2, *cp3; 2038 char *cplim; 2039 struct ifaddr *ifa_maybe = NULL; 2040 u_int af = addr->sa_family; 2041 2042 if (af >= AF_MAX) 2043 return (NULL); 2044 2045 NET_EPOCH_ASSERT(); 2046 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2047 if (ifa->ifa_addr->sa_family != af) 2048 continue; 2049 if (ifa_maybe == NULL) 2050 ifa_maybe = ifa; 2051 if (ifa->ifa_netmask == 0) { 2052 if (sa_equal(addr, ifa->ifa_addr) || 2053 (ifa->ifa_dstaddr && 2054 sa_equal(addr, ifa->ifa_dstaddr))) 2055 goto done; 2056 continue; 2057 } 2058 if (ifp->if_flags & IFF_POINTOPOINT) { 2059 if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr)) 2060 goto done; 2061 } else { 2062 cp = addr->sa_data; 2063 cp2 = ifa->ifa_addr->sa_data; 2064 cp3 = ifa->ifa_netmask->sa_data; 2065 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2066 for (; cp3 < cplim; cp3++) 2067 if ((*cp++ ^ *cp2++) & *cp3) 2068 break; 2069 if (cp3 == cplim) 2070 goto done; 2071 } 2072 } 2073 ifa = ifa_maybe; 2074 done: 2075 return (ifa); 2076 } 2077 2078 /* 2079 * See whether new ifa is better than current one: 2080 * 1) A non-virtual one is preferred over virtual. 2081 * 2) A virtual in master state preferred over any other state. 2082 * 2083 * Used in several address selecting functions. 2084 */ 2085 int 2086 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2087 { 2088 2089 return (cur->ifa_carp && (!next->ifa_carp || 2090 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2091 } 2092 2093 struct sockaddr_dl * 2094 link_alloc_sdl(size_t size, int flags) 2095 { 2096 2097 return (malloc(size, M_TEMP, flags)); 2098 } 2099 2100 void 2101 link_free_sdl(struct sockaddr *sa) 2102 { 2103 free(sa, M_TEMP); 2104 } 2105 2106 /* 2107 * Fills in given sdl with interface basic info. 2108 * Returns pointer to filled sdl. 2109 */ 2110 struct sockaddr_dl * 2111 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2112 { 2113 struct sockaddr_dl *sdl; 2114 2115 sdl = (struct sockaddr_dl *)paddr; 2116 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2117 sdl->sdl_len = sizeof(struct sockaddr_dl); 2118 sdl->sdl_family = AF_LINK; 2119 sdl->sdl_index = ifp->if_index; 2120 sdl->sdl_type = iftype; 2121 2122 return (sdl); 2123 } 2124 2125 /* 2126 * Mark an interface down and notify protocols of 2127 * the transition. 2128 */ 2129 static void 2130 if_unroute(struct ifnet *ifp, int flag, int fam) 2131 { 2132 2133 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2134 2135 ifp->if_flags &= ~flag; 2136 getmicrotime(&ifp->if_lastchange); 2137 ifp->if_qflush(ifp); 2138 2139 if (ifp->if_carp) 2140 (*carp_linkstate_p)(ifp); 2141 rt_ifmsg(ifp, IFF_UP); 2142 } 2143 2144 /* 2145 * Mark an interface up and notify protocols of 2146 * the transition. 2147 */ 2148 static void 2149 if_route(struct ifnet *ifp, int flag, int fam) 2150 { 2151 2152 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2153 2154 ifp->if_flags |= flag; 2155 getmicrotime(&ifp->if_lastchange); 2156 if (ifp->if_carp) 2157 (*carp_linkstate_p)(ifp); 2158 rt_ifmsg(ifp, IFF_UP); 2159 #ifdef INET6 2160 in6_if_up(ifp); 2161 #endif 2162 } 2163 2164 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2165 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2166 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2167 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2168 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2169 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2170 int (*vlan_setcookie_p)(struct ifnet *, void *); 2171 void *(*vlan_cookie_p)(struct ifnet *); 2172 2173 /* 2174 * Handle a change in the interface link state. To avoid LORs 2175 * between driver lock and upper layer locks, as well as possible 2176 * recursions, we post event to taskqueue, and all job 2177 * is done in static do_link_state_change(). 2178 */ 2179 void 2180 if_link_state_change(struct ifnet *ifp, int link_state) 2181 { 2182 /* Return if state hasn't changed. */ 2183 if (ifp->if_link_state == link_state) 2184 return; 2185 2186 ifp->if_link_state = link_state; 2187 2188 /* XXXGL: reference ifp? */ 2189 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2190 } 2191 2192 static void 2193 do_link_state_change(void *arg, int pending) 2194 { 2195 struct ifnet *ifp; 2196 int link_state; 2197 2198 ifp = arg; 2199 link_state = ifp->if_link_state; 2200 2201 CURVNET_SET(ifp->if_vnet); 2202 rt_ifmsg(ifp, 0); 2203 if (ifp->if_vlantrunk != NULL) 2204 (*vlan_link_state_p)(ifp); 2205 2206 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2207 ifp->if_l2com != NULL) 2208 (*ng_ether_link_state_p)(ifp, link_state); 2209 if (ifp->if_carp) 2210 (*carp_linkstate_p)(ifp); 2211 if (ifp->if_bridge) 2212 ifp->if_bridge_linkstate(ifp); 2213 if (ifp->if_lagg) 2214 (*lagg_linkstate_p)(ifp, link_state); 2215 2216 if (IS_DEFAULT_VNET(curvnet)) 2217 devctl_notify("IFNET", ifp->if_xname, 2218 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2219 NULL); 2220 if (pending > 1) 2221 if_printf(ifp, "%d link states coalesced\n", pending); 2222 if (log_link_state_change) 2223 if_printf(ifp, "link state changed to %s\n", 2224 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2225 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2226 CURVNET_RESTORE(); 2227 } 2228 2229 /* 2230 * Mark an interface down and notify protocols of 2231 * the transition. 2232 */ 2233 void 2234 if_down(struct ifnet *ifp) 2235 { 2236 2237 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2238 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2239 } 2240 2241 /* 2242 * Mark an interface up and notify protocols of 2243 * the transition. 2244 */ 2245 void 2246 if_up(struct ifnet *ifp) 2247 { 2248 2249 if_route(ifp, IFF_UP, AF_UNSPEC); 2250 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2251 } 2252 2253 /* 2254 * Flush an interface queue. 2255 */ 2256 void 2257 if_qflush(struct ifnet *ifp) 2258 { 2259 struct mbuf *m, *n; 2260 struct ifaltq *ifq; 2261 2262 ifq = &ifp->if_snd; 2263 IFQ_LOCK(ifq); 2264 #ifdef ALTQ 2265 if (ALTQ_IS_ENABLED(ifq)) 2266 ALTQ_PURGE(ifq); 2267 #endif 2268 n = ifq->ifq_head; 2269 while ((m = n) != NULL) { 2270 n = m->m_nextpkt; 2271 m_freem(m); 2272 } 2273 ifq->ifq_head = 0; 2274 ifq->ifq_tail = 0; 2275 ifq->ifq_len = 0; 2276 IFQ_UNLOCK(ifq); 2277 } 2278 2279 /* 2280 * Map interface name to interface structure pointer, with or without 2281 * returning a reference. 2282 */ 2283 struct ifnet * 2284 ifunit_ref(const char *name) 2285 { 2286 struct epoch_tracker et; 2287 struct ifnet *ifp; 2288 2289 NET_EPOCH_ENTER(et); 2290 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2291 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2292 !(ifp->if_flags & IFF_DYING)) 2293 break; 2294 } 2295 if (ifp != NULL) { 2296 if_ref(ifp); 2297 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2298 } 2299 2300 NET_EPOCH_EXIT(et); 2301 return (ifp); 2302 } 2303 2304 struct ifnet * 2305 ifunit(const char *name) 2306 { 2307 struct epoch_tracker et; 2308 struct ifnet *ifp; 2309 2310 NET_EPOCH_ENTER(et); 2311 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2312 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2313 break; 2314 } 2315 NET_EPOCH_EXIT(et); 2316 return (ifp); 2317 } 2318 2319 void * 2320 ifr_buffer_get_buffer(void *data) 2321 { 2322 union ifreq_union *ifrup; 2323 2324 ifrup = data; 2325 #ifdef COMPAT_FREEBSD32 2326 if (SV_CURPROC_FLAG(SV_ILP32)) 2327 return ((void *)(uintptr_t) 2328 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2329 #endif 2330 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2331 } 2332 2333 static void 2334 ifr_buffer_set_buffer_null(void *data) 2335 { 2336 union ifreq_union *ifrup; 2337 2338 ifrup = data; 2339 #ifdef COMPAT_FREEBSD32 2340 if (SV_CURPROC_FLAG(SV_ILP32)) 2341 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2342 else 2343 #endif 2344 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2345 } 2346 2347 size_t 2348 ifr_buffer_get_length(void *data) 2349 { 2350 union ifreq_union *ifrup; 2351 2352 ifrup = data; 2353 #ifdef COMPAT_FREEBSD32 2354 if (SV_CURPROC_FLAG(SV_ILP32)) 2355 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2356 #endif 2357 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2358 } 2359 2360 static void 2361 ifr_buffer_set_length(void *data, size_t len) 2362 { 2363 union ifreq_union *ifrup; 2364 2365 ifrup = data; 2366 #ifdef COMPAT_FREEBSD32 2367 if (SV_CURPROC_FLAG(SV_ILP32)) 2368 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2369 else 2370 #endif 2371 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2372 } 2373 2374 void * 2375 ifr_data_get_ptr(void *ifrp) 2376 { 2377 union ifreq_union *ifrup; 2378 2379 ifrup = ifrp; 2380 #ifdef COMPAT_FREEBSD32 2381 if (SV_CURPROC_FLAG(SV_ILP32)) 2382 return ((void *)(uintptr_t) 2383 ifrup->ifr32.ifr_ifru.ifru_data); 2384 #endif 2385 return (ifrup->ifr.ifr_ifru.ifru_data); 2386 } 2387 2388 struct ifcap_nv_bit_name { 2389 uint64_t cap_bit; 2390 const char *cap_name; 2391 }; 2392 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2393 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2394 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2395 CAPNV(RXCSUM), 2396 CAPNV(TXCSUM), 2397 CAPNV(NETCONS), 2398 CAPNV(VLAN_MTU), 2399 CAPNV(VLAN_HWTAGGING), 2400 CAPNV(JUMBO_MTU), 2401 CAPNV(POLLING), 2402 CAPNV(VLAN_HWCSUM), 2403 CAPNV(TSO4), 2404 CAPNV(TSO6), 2405 CAPNV(LRO), 2406 CAPNV(WOL_UCAST), 2407 CAPNV(WOL_MCAST), 2408 CAPNV(WOL_MAGIC), 2409 CAPNV(TOE4), 2410 CAPNV(TOE6), 2411 CAPNV(VLAN_HWFILTER), 2412 CAPNV(VLAN_HWTSO), 2413 CAPNV(LINKSTATE), 2414 CAPNV(NETMAP), 2415 CAPNV(RXCSUM_IPV6), 2416 CAPNV(TXCSUM_IPV6), 2417 CAPNV(HWSTATS), 2418 CAPNV(TXRTLMT), 2419 CAPNV(HWRXTSTMP), 2420 CAPNV(MEXTPG), 2421 CAPNV(TXTLS4), 2422 CAPNV(TXTLS6), 2423 CAPNV(VXLAN_HWCSUM), 2424 CAPNV(VXLAN_HWTSO), 2425 CAPNV(TXTLS_RTLMT), 2426 {0, NULL} 2427 }; 2428 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \ 2429 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2430 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2431 CAP2NV(RXTLS4), 2432 CAP2NV(RXTLS6), 2433 {0, NULL} 2434 }; 2435 #undef CAPNV 2436 #undef CAP2NV 2437 2438 int 2439 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2440 const struct ifcap_nv_bit_name *nn, bool all) 2441 { 2442 int i, res; 2443 2444 res = 0; 2445 for (i = 0; nn[i].cap_name != NULL; i++) { 2446 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2447 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2448 res |= nn[i].cap_bit; 2449 } else { 2450 res |= *old_cap & nn[i].cap_bit; 2451 } 2452 } 2453 return (res); 2454 } 2455 2456 void 2457 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2458 int ifr_cap, int ifr_req) 2459 { 2460 int i; 2461 2462 for (i = 0; nn[i].cap_name != NULL; i++) { 2463 if ((nn[i].cap_bit & ifr_cap) != 0) { 2464 nvlist_add_bool(nv, nn[i].cap_name, 2465 (nn[i].cap_bit & ifr_req) != 0); 2466 } 2467 } 2468 } 2469 2470 /* 2471 * Hardware specific interface ioctls. 2472 */ 2473 int 2474 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2475 { 2476 struct ifreq *ifr; 2477 int error = 0, do_ifup = 0; 2478 int new_flags, temp_flags; 2479 size_t namelen, onamelen; 2480 size_t descrlen, nvbuflen; 2481 char *descrbuf; 2482 char new_name[IFNAMSIZ]; 2483 char old_name[IFNAMSIZ], strbuf[IFNAMSIZ + 8]; 2484 struct ifaddr *ifa; 2485 struct sockaddr_dl *sdl; 2486 void *buf; 2487 nvlist_t *nvcap; 2488 struct siocsifcapnv_driver_data drv_ioctl_data; 2489 2490 ifr = (struct ifreq *)data; 2491 switch (cmd) { 2492 case SIOCGIFINDEX: 2493 ifr->ifr_index = ifp->if_index; 2494 break; 2495 2496 case SIOCGIFFLAGS: 2497 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2498 ifr->ifr_flags = temp_flags & 0xffff; 2499 ifr->ifr_flagshigh = temp_flags >> 16; 2500 break; 2501 2502 case SIOCGIFCAP: 2503 ifr->ifr_reqcap = ifp->if_capabilities; 2504 ifr->ifr_curcap = ifp->if_capenable; 2505 break; 2506 2507 case SIOCGIFCAPNV: 2508 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2509 error = EINVAL; 2510 break; 2511 } 2512 buf = NULL; 2513 nvcap = nvlist_create(0); 2514 for (;;) { 2515 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2516 ifp->if_capabilities, ifp->if_capenable); 2517 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2518 ifp->if_capabilities2, ifp->if_capenable2); 2519 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2520 __DECONST(caddr_t, nvcap)); 2521 if (error != 0) { 2522 if_printf(ifp, 2523 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2524 error); 2525 break; 2526 } 2527 buf = nvlist_pack(nvcap, &nvbuflen); 2528 if (buf == NULL) { 2529 error = nvlist_error(nvcap); 2530 if (error == 0) 2531 error = EDOOFUS; 2532 break; 2533 } 2534 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2535 ifr->ifr_cap_nv.length = nvbuflen; 2536 ifr->ifr_cap_nv.buffer = NULL; 2537 error = EFBIG; 2538 break; 2539 } 2540 ifr->ifr_cap_nv.length = nvbuflen; 2541 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2542 break; 2543 } 2544 free(buf, M_NVLIST); 2545 nvlist_destroy(nvcap); 2546 break; 2547 2548 case SIOCGIFDATA: 2549 { 2550 struct if_data ifd; 2551 2552 /* Ensure uninitialised padding is not leaked. */ 2553 memset(&ifd, 0, sizeof(ifd)); 2554 2555 if_data_copy(ifp, &ifd); 2556 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2557 break; 2558 } 2559 2560 #ifdef MAC 2561 case SIOCGIFMAC: 2562 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2563 break; 2564 #endif 2565 2566 case SIOCGIFMETRIC: 2567 ifr->ifr_metric = ifp->if_metric; 2568 break; 2569 2570 case SIOCGIFMTU: 2571 ifr->ifr_mtu = ifp->if_mtu; 2572 break; 2573 2574 case SIOCGIFPHYS: 2575 /* XXXGL: did this ever worked? */ 2576 ifr->ifr_phys = 0; 2577 break; 2578 2579 case SIOCGIFDESCR: 2580 error = 0; 2581 sx_slock(&ifdescr_sx); 2582 if (ifp->if_description == NULL) 2583 error = ENOMSG; 2584 else { 2585 /* space for terminating nul */ 2586 descrlen = strlen(ifp->if_description) + 1; 2587 if (ifr_buffer_get_length(ifr) < descrlen) 2588 ifr_buffer_set_buffer_null(ifr); 2589 else 2590 error = copyout(ifp->if_description, 2591 ifr_buffer_get_buffer(ifr), descrlen); 2592 ifr_buffer_set_length(ifr, descrlen); 2593 } 2594 sx_sunlock(&ifdescr_sx); 2595 break; 2596 2597 case SIOCSIFDESCR: 2598 error = priv_check(td, PRIV_NET_SETIFDESCR); 2599 if (error) 2600 return (error); 2601 2602 /* 2603 * Copy only (length-1) bytes to make sure that 2604 * if_description is always nul terminated. The 2605 * length parameter is supposed to count the 2606 * terminating nul in. 2607 */ 2608 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2609 return (ENAMETOOLONG); 2610 else if (ifr_buffer_get_length(ifr) == 0) 2611 descrbuf = NULL; 2612 else { 2613 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2614 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2615 ifr_buffer_get_length(ifr) - 1); 2616 if (error) { 2617 if_freedescr(descrbuf); 2618 break; 2619 } 2620 } 2621 2622 if_setdescr(ifp, descrbuf); 2623 getmicrotime(&ifp->if_lastchange); 2624 break; 2625 2626 case SIOCGIFFIB: 2627 ifr->ifr_fib = ifp->if_fib; 2628 break; 2629 2630 case SIOCSIFFIB: 2631 error = priv_check(td, PRIV_NET_SETIFFIB); 2632 if (error) 2633 return (error); 2634 if (ifr->ifr_fib >= rt_numfibs) 2635 return (EINVAL); 2636 2637 ifp->if_fib = ifr->ifr_fib; 2638 break; 2639 2640 case SIOCSIFFLAGS: 2641 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2642 if (error) 2643 return (error); 2644 /* 2645 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2646 * check, so we don't need special handling here yet. 2647 */ 2648 new_flags = (ifr->ifr_flags & 0xffff) | 2649 (ifr->ifr_flagshigh << 16); 2650 if (ifp->if_flags & IFF_UP && 2651 (new_flags & IFF_UP) == 0) { 2652 if_down(ifp); 2653 } else if (new_flags & IFF_UP && 2654 (ifp->if_flags & IFF_UP) == 0) { 2655 do_ifup = 1; 2656 } 2657 /* See if permanently promiscuous mode bit is about to flip */ 2658 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2659 if (new_flags & IFF_PPROMISC) 2660 ifp->if_flags |= IFF_PROMISC; 2661 else if (ifp->if_pcount == 0) 2662 ifp->if_flags &= ~IFF_PROMISC; 2663 if (log_promisc_mode_change) 2664 if_printf(ifp, "permanently promiscuous mode %s\n", 2665 ((new_flags & IFF_PPROMISC) ? 2666 "enabled" : "disabled")); 2667 } 2668 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2669 (new_flags &~ IFF_CANTCHANGE); 2670 if (ifp->if_ioctl) { 2671 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2672 } 2673 if (do_ifup) 2674 if_up(ifp); 2675 getmicrotime(&ifp->if_lastchange); 2676 break; 2677 2678 case SIOCSIFCAP: 2679 error = priv_check(td, PRIV_NET_SETIFCAP); 2680 if (error != 0) 2681 return (error); 2682 if (ifp->if_ioctl == NULL) 2683 return (EOPNOTSUPP); 2684 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2685 return (EINVAL); 2686 error = (*ifp->if_ioctl)(ifp, cmd, data); 2687 if (error == 0) 2688 getmicrotime(&ifp->if_lastchange); 2689 break; 2690 2691 case SIOCSIFCAPNV: 2692 error = priv_check(td, PRIV_NET_SETIFCAP); 2693 if (error != 0) 2694 return (error); 2695 if (ifp->if_ioctl == NULL) 2696 return (EOPNOTSUPP); 2697 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2698 return (EINVAL); 2699 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2700 return (EINVAL); 2701 nvcap = NULL; 2702 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2703 for (;;) { 2704 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2705 ifr->ifr_cap_nv.length); 2706 if (error != 0) 2707 break; 2708 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2709 if (nvcap == NULL) { 2710 error = EINVAL; 2711 break; 2712 } 2713 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2714 &ifp->if_capenable, ifcap_nv_bit_names, false); 2715 if ((drv_ioctl_data.reqcap & 2716 ~ifp->if_capabilities) != 0) { 2717 error = EINVAL; 2718 break; 2719 } 2720 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2721 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2722 if ((drv_ioctl_data.reqcap2 & 2723 ~ifp->if_capabilities2) != 0) { 2724 error = EINVAL; 2725 break; 2726 } 2727 drv_ioctl_data.nvcap = nvcap; 2728 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2729 (caddr_t)&drv_ioctl_data); 2730 break; 2731 } 2732 nvlist_destroy(nvcap); 2733 free(buf, M_TEMP); 2734 if (error == 0) 2735 getmicrotime(&ifp->if_lastchange); 2736 break; 2737 2738 #ifdef MAC 2739 case SIOCSIFMAC: 2740 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2741 break; 2742 #endif 2743 2744 case SIOCSIFNAME: 2745 error = priv_check(td, PRIV_NET_SETIFNAME); 2746 if (error) 2747 return (error); 2748 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2749 NULL); 2750 if (error != 0) 2751 return (error); 2752 if (new_name[0] == '\0') 2753 return (EINVAL); 2754 if (strcmp(new_name, ifp->if_xname) == 0) 2755 break; 2756 if (ifunit(new_name) != NULL) 2757 return (EEXIST); 2758 2759 /* 2760 * XXX: Locking. Nothing else seems to lock if_flags, 2761 * and there are numerous other races with the 2762 * ifunit() checks not being atomic with namespace 2763 * changes (renames, vmoves, if_attach, etc). 2764 */ 2765 ifp->if_flags |= IFF_RENAMING; 2766 2767 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2768 2769 if_printf(ifp, "changing name to '%s'\n", new_name); 2770 2771 IF_ADDR_WLOCK(ifp); 2772 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 2773 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2774 ifa = ifp->if_addr; 2775 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2776 namelen = strlen(new_name); 2777 onamelen = sdl->sdl_nlen; 2778 /* 2779 * Move the address if needed. This is safe because we 2780 * allocate space for a name of length IFNAMSIZ when we 2781 * create this in if_attach(). 2782 */ 2783 if (namelen != onamelen) { 2784 bcopy(sdl->sdl_data + onamelen, 2785 sdl->sdl_data + namelen, sdl->sdl_alen); 2786 } 2787 bcopy(new_name, sdl->sdl_data, namelen); 2788 sdl->sdl_nlen = namelen; 2789 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2790 bzero(sdl->sdl_data, onamelen); 2791 while (namelen != 0) 2792 sdl->sdl_data[--namelen] = 0xff; 2793 IF_ADDR_WUNLOCK(ifp); 2794 2795 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2796 2797 ifp->if_flags &= ~IFF_RENAMING; 2798 2799 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 2800 devctl_notify("IFNET", old_name, "RENAME", strbuf); 2801 break; 2802 2803 #ifdef VIMAGE 2804 case SIOCSIFVNET: 2805 error = priv_check(td, PRIV_NET_SETIFVNET); 2806 if (error) 2807 return (error); 2808 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2809 break; 2810 #endif 2811 2812 case SIOCSIFMETRIC: 2813 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2814 if (error) 2815 return (error); 2816 ifp->if_metric = ifr->ifr_metric; 2817 getmicrotime(&ifp->if_lastchange); 2818 break; 2819 2820 case SIOCSIFPHYS: 2821 error = priv_check(td, PRIV_NET_SETIFPHYS); 2822 if (error) 2823 return (error); 2824 if (ifp->if_ioctl == NULL) 2825 return (EOPNOTSUPP); 2826 error = (*ifp->if_ioctl)(ifp, cmd, data); 2827 if (error == 0) 2828 getmicrotime(&ifp->if_lastchange); 2829 break; 2830 2831 case SIOCSIFMTU: 2832 { 2833 u_long oldmtu = ifp->if_mtu; 2834 2835 error = priv_check(td, PRIV_NET_SETIFMTU); 2836 if (error) 2837 return (error); 2838 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2839 return (EINVAL); 2840 if (ifp->if_ioctl == NULL) 2841 return (EOPNOTSUPP); 2842 /* Disallow MTU changes on bridge member interfaces. */ 2843 if (ifp->if_bridge) 2844 return (EOPNOTSUPP); 2845 error = (*ifp->if_ioctl)(ifp, cmd, data); 2846 if (error == 0) { 2847 getmicrotime(&ifp->if_lastchange); 2848 rt_ifmsg(ifp, 0); 2849 #ifdef INET 2850 DEBUGNET_NOTIFY_MTU(ifp); 2851 #endif 2852 } 2853 /* 2854 * If the link MTU changed, do network layer specific procedure. 2855 */ 2856 if (ifp->if_mtu != oldmtu) 2857 if_notifymtu(ifp); 2858 break; 2859 } 2860 2861 case SIOCADDMULTI: 2862 case SIOCDELMULTI: 2863 if (cmd == SIOCADDMULTI) 2864 error = priv_check(td, PRIV_NET_ADDMULTI); 2865 else 2866 error = priv_check(td, PRIV_NET_DELMULTI); 2867 if (error) 2868 return (error); 2869 2870 /* Don't allow group membership on non-multicast interfaces. */ 2871 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2872 return (EOPNOTSUPP); 2873 2874 /* Don't let users screw up protocols' entries. */ 2875 if (ifr->ifr_addr.sa_family != AF_LINK) 2876 return (EINVAL); 2877 2878 if (cmd == SIOCADDMULTI) { 2879 struct epoch_tracker et; 2880 struct ifmultiaddr *ifma; 2881 2882 /* 2883 * Userland is only permitted to join groups once 2884 * via the if_addmulti() KPI, because it cannot hold 2885 * struct ifmultiaddr * between calls. It may also 2886 * lose a race while we check if the membership 2887 * already exists. 2888 */ 2889 NET_EPOCH_ENTER(et); 2890 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2891 NET_EPOCH_EXIT(et); 2892 if (ifma != NULL) 2893 error = EADDRINUSE; 2894 else 2895 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2896 } else { 2897 error = if_delmulti(ifp, &ifr->ifr_addr); 2898 } 2899 if (error == 0) 2900 getmicrotime(&ifp->if_lastchange); 2901 break; 2902 2903 case SIOCSIFPHYADDR: 2904 case SIOCDIFPHYADDR: 2905 #ifdef INET6 2906 case SIOCSIFPHYADDR_IN6: 2907 #endif 2908 case SIOCSIFMEDIA: 2909 case SIOCSIFGENERIC: 2910 error = priv_check(td, PRIV_NET_HWIOCTL); 2911 if (error) 2912 return (error); 2913 if (ifp->if_ioctl == NULL) 2914 return (EOPNOTSUPP); 2915 error = (*ifp->if_ioctl)(ifp, cmd, data); 2916 if (error == 0) 2917 getmicrotime(&ifp->if_lastchange); 2918 break; 2919 2920 case SIOCGIFSTATUS: 2921 case SIOCGIFPSRCADDR: 2922 case SIOCGIFPDSTADDR: 2923 case SIOCGIFMEDIA: 2924 case SIOCGIFXMEDIA: 2925 case SIOCGIFGENERIC: 2926 case SIOCGIFRSSKEY: 2927 case SIOCGIFRSSHASH: 2928 case SIOCGIFDOWNREASON: 2929 if (ifp->if_ioctl == NULL) 2930 return (EOPNOTSUPP); 2931 error = (*ifp->if_ioctl)(ifp, cmd, data); 2932 break; 2933 2934 case SIOCSIFLLADDR: 2935 error = priv_check(td, PRIV_NET_SETLLADDR); 2936 if (error) 2937 return (error); 2938 error = if_setlladdr(ifp, 2939 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2940 break; 2941 2942 case SIOCGHWADDR: 2943 error = if_gethwaddr(ifp, ifr); 2944 break; 2945 2946 case SIOCAIFGROUP: 2947 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2948 if (error) 2949 return (error); 2950 error = if_addgroup(ifp, 2951 ((struct ifgroupreq *)data)->ifgr_group); 2952 if (error != 0) 2953 return (error); 2954 break; 2955 2956 case SIOCGIFGROUP: 2957 { 2958 struct epoch_tracker et; 2959 2960 NET_EPOCH_ENTER(et); 2961 error = if_getgroup((struct ifgroupreq *)data, ifp); 2962 NET_EPOCH_EXIT(et); 2963 break; 2964 } 2965 2966 case SIOCDIFGROUP: 2967 error = priv_check(td, PRIV_NET_DELIFGROUP); 2968 if (error) 2969 return (error); 2970 error = if_delgroup(ifp, 2971 ((struct ifgroupreq *)data)->ifgr_group); 2972 if (error != 0) 2973 return (error); 2974 break; 2975 2976 default: 2977 error = ENOIOCTL; 2978 break; 2979 } 2980 return (error); 2981 } 2982 2983 /* 2984 * Interface ioctls. 2985 */ 2986 int 2987 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2988 { 2989 #ifdef COMPAT_FREEBSD32 2990 union { 2991 struct ifconf ifc; 2992 struct ifdrv ifd; 2993 struct ifgroupreq ifgr; 2994 struct ifmediareq ifmr; 2995 } thunk; 2996 u_long saved_cmd; 2997 struct ifconf32 *ifc32; 2998 struct ifdrv32 *ifd32; 2999 struct ifgroupreq32 *ifgr32; 3000 struct ifmediareq32 *ifmr32; 3001 #endif 3002 struct ifnet *ifp; 3003 struct ifreq *ifr; 3004 int error; 3005 int oif_flags; 3006 #ifdef VIMAGE 3007 bool shutdown; 3008 #endif 3009 3010 CURVNET_SET(so->so_vnet); 3011 #ifdef VIMAGE 3012 /* Make sure the VNET is stable. */ 3013 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 3014 if (shutdown) { 3015 CURVNET_RESTORE(); 3016 return (EBUSY); 3017 } 3018 #endif 3019 3020 #ifdef COMPAT_FREEBSD32 3021 saved_cmd = cmd; 3022 switch (cmd) { 3023 case SIOCGIFCONF32: 3024 ifc32 = (struct ifconf32 *)data; 3025 thunk.ifc.ifc_len = ifc32->ifc_len; 3026 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 3027 data = (caddr_t)&thunk.ifc; 3028 cmd = SIOCGIFCONF; 3029 break; 3030 case SIOCGDRVSPEC32: 3031 case SIOCSDRVSPEC32: 3032 ifd32 = (struct ifdrv32 *)data; 3033 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 3034 sizeof(thunk.ifd.ifd_name)); 3035 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 3036 thunk.ifd.ifd_len = ifd32->ifd_len; 3037 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 3038 data = (caddr_t)&thunk.ifd; 3039 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 3040 break; 3041 case SIOCAIFGROUP32: 3042 case SIOCGIFGROUP32: 3043 case SIOCDIFGROUP32: 3044 case SIOCGIFGMEMB32: 3045 ifgr32 = (struct ifgroupreq32 *)data; 3046 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 3047 sizeof(thunk.ifgr.ifgr_name)); 3048 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 3049 switch (cmd) { 3050 case SIOCAIFGROUP32: 3051 case SIOCDIFGROUP32: 3052 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 3053 sizeof(thunk.ifgr.ifgr_group)); 3054 break; 3055 case SIOCGIFGROUP32: 3056 case SIOCGIFGMEMB32: 3057 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 3058 break; 3059 } 3060 data = (caddr_t)&thunk.ifgr; 3061 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 3062 break; 3063 case SIOCGIFMEDIA32: 3064 case SIOCGIFXMEDIA32: 3065 ifmr32 = (struct ifmediareq32 *)data; 3066 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 3067 sizeof(thunk.ifmr.ifm_name)); 3068 thunk.ifmr.ifm_current = ifmr32->ifm_current; 3069 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 3070 thunk.ifmr.ifm_status = ifmr32->ifm_status; 3071 thunk.ifmr.ifm_active = ifmr32->ifm_active; 3072 thunk.ifmr.ifm_count = ifmr32->ifm_count; 3073 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 3074 data = (caddr_t)&thunk.ifmr; 3075 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 3076 break; 3077 } 3078 #endif 3079 3080 switch (cmd) { 3081 case SIOCGIFCONF: 3082 error = ifconf(cmd, data); 3083 goto out_noref; 3084 } 3085 3086 ifr = (struct ifreq *)data; 3087 switch (cmd) { 3088 #ifdef VIMAGE 3089 case SIOCSIFRVNET: 3090 error = priv_check(td, PRIV_NET_SETIFVNET); 3091 if (error == 0) 3092 error = if_vmove_reclaim(td, ifr->ifr_name, 3093 ifr->ifr_jid); 3094 goto out_noref; 3095 #endif 3096 case SIOCIFCREATE: 3097 case SIOCIFCREATE2: 3098 error = priv_check(td, PRIV_NET_IFCREATE); 3099 if (error == 0) 3100 error = if_clone_create(ifr->ifr_name, 3101 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3102 ifr_data_get_ptr(ifr) : NULL); 3103 goto out_noref; 3104 case SIOCIFDESTROY: 3105 error = priv_check(td, PRIV_NET_IFDESTROY); 3106 3107 if (error == 0) { 3108 sx_xlock(&ifnet_detach_sxlock); 3109 error = if_clone_destroy(ifr->ifr_name); 3110 sx_xunlock(&ifnet_detach_sxlock); 3111 } 3112 goto out_noref; 3113 3114 case SIOCIFGCLONERS: 3115 error = if_clone_list((struct if_clonereq *)data); 3116 goto out_noref; 3117 3118 case SIOCGIFGMEMB: 3119 error = if_getgroupmembers((struct ifgroupreq *)data); 3120 goto out_noref; 3121 3122 #if defined(INET) || defined(INET6) 3123 case SIOCSVH: 3124 case SIOCGVH: 3125 if (carp_ioctl_p == NULL) 3126 error = EPROTONOSUPPORT; 3127 else 3128 error = (*carp_ioctl_p)(ifr, cmd, td); 3129 goto out_noref; 3130 #endif 3131 } 3132 3133 ifp = ifunit_ref(ifr->ifr_name); 3134 if (ifp == NULL) { 3135 error = ENXIO; 3136 goto out_noref; 3137 } 3138 3139 error = ifhwioctl(cmd, ifp, data, td); 3140 if (error != ENOIOCTL) 3141 goto out_ref; 3142 3143 oif_flags = ifp->if_flags; 3144 if (so->so_proto == NULL) { 3145 error = EOPNOTSUPP; 3146 goto out_ref; 3147 } 3148 3149 /* 3150 * Pass the request on to the socket control method, and if the 3151 * latter returns EOPNOTSUPP, directly to the interface. 3152 * 3153 * Make an exception for the legacy SIOCSIF* requests. Drivers 3154 * trust SIOCSIFADDR et al to come from an already privileged 3155 * layer, and do not perform any credentials checks or input 3156 * validation. 3157 */ 3158 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 3159 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3160 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3161 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3162 error = (*ifp->if_ioctl)(ifp, cmd, data); 3163 3164 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3165 #ifdef INET6 3166 if (ifp->if_flags & IFF_UP) 3167 in6_if_up(ifp); 3168 #endif 3169 } 3170 3171 out_ref: 3172 if_rele(ifp); 3173 out_noref: 3174 CURVNET_RESTORE(); 3175 #ifdef COMPAT_FREEBSD32 3176 if (error != 0) 3177 return (error); 3178 switch (saved_cmd) { 3179 case SIOCGIFCONF32: 3180 ifc32->ifc_len = thunk.ifc.ifc_len; 3181 break; 3182 case SIOCGDRVSPEC32: 3183 /* 3184 * SIOCGDRVSPEC is IOWR, but nothing actually touches 3185 * the struct so just assert that ifd_len (the only 3186 * field it might make sense to update) hasn't 3187 * changed. 3188 */ 3189 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 3190 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 3191 thunk.ifd.ifd_len)); 3192 break; 3193 case SIOCGIFGROUP32: 3194 case SIOCGIFGMEMB32: 3195 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 3196 break; 3197 case SIOCGIFMEDIA32: 3198 case SIOCGIFXMEDIA32: 3199 ifmr32->ifm_current = thunk.ifmr.ifm_current; 3200 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 3201 ifmr32->ifm_status = thunk.ifmr.ifm_status; 3202 ifmr32->ifm_active = thunk.ifmr.ifm_active; 3203 ifmr32->ifm_count = thunk.ifmr.ifm_count; 3204 break; 3205 } 3206 #endif 3207 return (error); 3208 } 3209 3210 /* 3211 * The code common to handling reference counted flags, 3212 * e.g., in ifpromisc() and if_allmulti(). 3213 * The "pflag" argument can specify a permanent mode flag to check, 3214 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3215 * 3216 * Only to be used on stack-owned flags, not driver-owned flags. 3217 */ 3218 static int 3219 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3220 { 3221 struct ifreq ifr; 3222 int error; 3223 int oldflags, oldcount; 3224 3225 /* Sanity checks to catch programming errors */ 3226 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3227 ("%s: setting driver-owned flag %d", __func__, flag)); 3228 3229 if (onswitch) 3230 KASSERT(*refcount >= 0, 3231 ("%s: increment negative refcount %d for flag %d", 3232 __func__, *refcount, flag)); 3233 else 3234 KASSERT(*refcount > 0, 3235 ("%s: decrement non-positive refcount %d for flag %d", 3236 __func__, *refcount, flag)); 3237 3238 /* In case this mode is permanent, just touch refcount */ 3239 if (ifp->if_flags & pflag) { 3240 *refcount += onswitch ? 1 : -1; 3241 return (0); 3242 } 3243 3244 /* Save ifnet parameters for if_ioctl() may fail */ 3245 oldcount = *refcount; 3246 oldflags = ifp->if_flags; 3247 3248 /* 3249 * See if we aren't the only and touching refcount is enough. 3250 * Actually toggle interface flag if we are the first or last. 3251 */ 3252 if (onswitch) { 3253 if ((*refcount)++) 3254 return (0); 3255 ifp->if_flags |= flag; 3256 } else { 3257 if (--(*refcount)) 3258 return (0); 3259 ifp->if_flags &= ~flag; 3260 } 3261 3262 /* Call down the driver since we've changed interface flags */ 3263 if (ifp->if_ioctl == NULL) { 3264 error = EOPNOTSUPP; 3265 goto recover; 3266 } 3267 ifr.ifr_flags = ifp->if_flags & 0xffff; 3268 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3269 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3270 if (error) 3271 goto recover; 3272 /* Notify userland that interface flags have changed */ 3273 rt_ifmsg(ifp, flag); 3274 return (0); 3275 3276 recover: 3277 /* Recover after driver error */ 3278 *refcount = oldcount; 3279 ifp->if_flags = oldflags; 3280 return (error); 3281 } 3282 3283 /* 3284 * Set/clear promiscuous mode on interface ifp based on the truth value 3285 * of pswitch. The calls are reference counted so that only the first 3286 * "on" request actually has an effect, as does the final "off" request. 3287 * Results are undefined if the "off" and "on" requests are not matched. 3288 */ 3289 int 3290 ifpromisc(struct ifnet *ifp, int pswitch) 3291 { 3292 int error; 3293 int oldflags = ifp->if_flags; 3294 3295 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3296 &ifp->if_pcount, pswitch); 3297 /* If promiscuous mode status has changed, log a message */ 3298 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3299 log_promisc_mode_change) 3300 if_printf(ifp, "promiscuous mode %s\n", 3301 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3302 return (error); 3303 } 3304 3305 /* 3306 * Return interface configuration 3307 * of system. List may be used 3308 * in later ioctl's (above) to get 3309 * other information. 3310 */ 3311 /*ARGSUSED*/ 3312 static int 3313 ifconf(u_long cmd, caddr_t data) 3314 { 3315 struct ifconf *ifc = (struct ifconf *)data; 3316 struct ifnet *ifp; 3317 struct ifaddr *ifa; 3318 struct ifreq ifr; 3319 struct sbuf *sb; 3320 int error, full = 0, valid_len, max_len; 3321 3322 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3323 max_len = maxphys - 1; 3324 3325 /* Prevent hostile input from being able to crash the system */ 3326 if (ifc->ifc_len <= 0) 3327 return (EINVAL); 3328 3329 again: 3330 if (ifc->ifc_len <= max_len) { 3331 max_len = ifc->ifc_len; 3332 full = 1; 3333 } 3334 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3335 max_len = 0; 3336 valid_len = 0; 3337 3338 IFNET_RLOCK(); 3339 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3340 struct epoch_tracker et; 3341 int addrs; 3342 3343 /* 3344 * Zero the ifr to make sure we don't disclose the contents 3345 * of the stack. 3346 */ 3347 memset(&ifr, 0, sizeof(ifr)); 3348 3349 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3350 >= sizeof(ifr.ifr_name)) { 3351 sbuf_delete(sb); 3352 IFNET_RUNLOCK(); 3353 return (ENAMETOOLONG); 3354 } 3355 3356 addrs = 0; 3357 NET_EPOCH_ENTER(et); 3358 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3359 struct sockaddr *sa = ifa->ifa_addr; 3360 3361 if (prison_if(curthread->td_ucred, sa) != 0) 3362 continue; 3363 addrs++; 3364 if (sa->sa_len <= sizeof(*sa)) { 3365 if (sa->sa_len < sizeof(*sa)) { 3366 memset(&ifr.ifr_ifru.ifru_addr, 0, 3367 sizeof(ifr.ifr_ifru.ifru_addr)); 3368 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3369 sa->sa_len); 3370 } else 3371 ifr.ifr_ifru.ifru_addr = *sa; 3372 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3373 max_len += sizeof(ifr); 3374 } else { 3375 sbuf_bcat(sb, &ifr, 3376 offsetof(struct ifreq, ifr_addr)); 3377 max_len += offsetof(struct ifreq, ifr_addr); 3378 sbuf_bcat(sb, sa, sa->sa_len); 3379 max_len += sa->sa_len; 3380 } 3381 3382 if (sbuf_error(sb) == 0) 3383 valid_len = sbuf_len(sb); 3384 } 3385 NET_EPOCH_EXIT(et); 3386 if (addrs == 0) { 3387 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3388 max_len += sizeof(ifr); 3389 3390 if (sbuf_error(sb) == 0) 3391 valid_len = sbuf_len(sb); 3392 } 3393 } 3394 IFNET_RUNLOCK(); 3395 3396 /* 3397 * If we didn't allocate enough space (uncommon), try again. If 3398 * we have already allocated as much space as we are allowed, 3399 * return what we've got. 3400 */ 3401 if (valid_len != max_len && !full) { 3402 sbuf_delete(sb); 3403 goto again; 3404 } 3405 3406 ifc->ifc_len = valid_len; 3407 sbuf_finish(sb); 3408 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3409 sbuf_delete(sb); 3410 return (error); 3411 } 3412 3413 /* 3414 * Just like ifpromisc(), but for all-multicast-reception mode. 3415 */ 3416 int 3417 if_allmulti(struct ifnet *ifp, int onswitch) 3418 { 3419 3420 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3421 } 3422 3423 struct ifmultiaddr * 3424 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3425 { 3426 struct ifmultiaddr *ifma; 3427 3428 IF_ADDR_LOCK_ASSERT(ifp); 3429 3430 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3431 if (sa->sa_family == AF_LINK) { 3432 if (sa_dl_equal(ifma->ifma_addr, sa)) 3433 break; 3434 } else { 3435 if (sa_equal(ifma->ifma_addr, sa)) 3436 break; 3437 } 3438 } 3439 3440 return ifma; 3441 } 3442 3443 /* 3444 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3445 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3446 * the ifnet multicast address list here, so the caller must do that and 3447 * other setup work (such as notifying the device driver). The reference 3448 * count is initialized to 1. 3449 */ 3450 static struct ifmultiaddr * 3451 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3452 int mflags) 3453 { 3454 struct ifmultiaddr *ifma; 3455 struct sockaddr *dupsa; 3456 3457 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3458 M_ZERO); 3459 if (ifma == NULL) 3460 return (NULL); 3461 3462 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3463 if (dupsa == NULL) { 3464 free(ifma, M_IFMADDR); 3465 return (NULL); 3466 } 3467 bcopy(sa, dupsa, sa->sa_len); 3468 ifma->ifma_addr = dupsa; 3469 3470 ifma->ifma_ifp = ifp; 3471 ifma->ifma_refcount = 1; 3472 ifma->ifma_protospec = NULL; 3473 3474 if (llsa == NULL) { 3475 ifma->ifma_lladdr = NULL; 3476 return (ifma); 3477 } 3478 3479 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3480 if (dupsa == NULL) { 3481 free(ifma->ifma_addr, M_IFMADDR); 3482 free(ifma, M_IFMADDR); 3483 return (NULL); 3484 } 3485 bcopy(llsa, dupsa, llsa->sa_len); 3486 ifma->ifma_lladdr = dupsa; 3487 3488 return (ifma); 3489 } 3490 3491 /* 3492 * if_freemulti: free ifmultiaddr structure and possibly attached related 3493 * addresses. The caller is responsible for implementing reference 3494 * counting, notifying the driver, handling routing messages, and releasing 3495 * any dependent link layer state. 3496 */ 3497 #ifdef MCAST_VERBOSE 3498 extern void kdb_backtrace(void); 3499 #endif 3500 static void 3501 if_freemulti_internal(struct ifmultiaddr *ifma) 3502 { 3503 3504 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3505 ifma->ifma_refcount)); 3506 3507 if (ifma->ifma_lladdr != NULL) 3508 free(ifma->ifma_lladdr, M_IFMADDR); 3509 #ifdef MCAST_VERBOSE 3510 kdb_backtrace(); 3511 printf("%s freeing ifma: %p\n", __func__, ifma); 3512 #endif 3513 free(ifma->ifma_addr, M_IFMADDR); 3514 free(ifma, M_IFMADDR); 3515 } 3516 3517 static void 3518 if_destroymulti(epoch_context_t ctx) 3519 { 3520 struct ifmultiaddr *ifma; 3521 3522 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3523 if_freemulti_internal(ifma); 3524 } 3525 3526 void 3527 if_freemulti(struct ifmultiaddr *ifma) 3528 { 3529 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3530 ifma->ifma_refcount)); 3531 3532 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3533 } 3534 3535 /* 3536 * Register an additional multicast address with a network interface. 3537 * 3538 * - If the address is already present, bump the reference count on the 3539 * address and return. 3540 * - If the address is not link-layer, look up a link layer address. 3541 * - Allocate address structures for one or both addresses, and attach to the 3542 * multicast address list on the interface. If automatically adding a link 3543 * layer address, the protocol address will own a reference to the link 3544 * layer address, to be freed when it is freed. 3545 * - Notify the network device driver of an addition to the multicast address 3546 * list. 3547 * 3548 * 'sa' points to caller-owned memory with the desired multicast address. 3549 * 3550 * 'retifma' will be used to return a pointer to the resulting multicast 3551 * address reference, if desired. 3552 */ 3553 int 3554 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3555 struct ifmultiaddr **retifma) 3556 { 3557 struct ifmultiaddr *ifma, *ll_ifma; 3558 struct sockaddr *llsa; 3559 struct sockaddr_dl sdl; 3560 int error; 3561 3562 #ifdef INET 3563 IN_MULTI_LIST_UNLOCK_ASSERT(); 3564 #endif 3565 #ifdef INET6 3566 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3567 #endif 3568 /* 3569 * If the address is already present, return a new reference to it; 3570 * otherwise, allocate storage and set up a new address. 3571 */ 3572 IF_ADDR_WLOCK(ifp); 3573 ifma = if_findmulti(ifp, sa); 3574 if (ifma != NULL) { 3575 ifma->ifma_refcount++; 3576 if (retifma != NULL) 3577 *retifma = ifma; 3578 IF_ADDR_WUNLOCK(ifp); 3579 return (0); 3580 } 3581 3582 /* 3583 * The address isn't already present; resolve the protocol address 3584 * into a link layer address, and then look that up, bump its 3585 * refcount or allocate an ifma for that also. 3586 * Most link layer resolving functions returns address data which 3587 * fits inside default sockaddr_dl structure. However callback 3588 * can allocate another sockaddr structure, in that case we need to 3589 * free it later. 3590 */ 3591 llsa = NULL; 3592 ll_ifma = NULL; 3593 if (ifp->if_resolvemulti != NULL) { 3594 /* Provide called function with buffer size information */ 3595 sdl.sdl_len = sizeof(sdl); 3596 llsa = (struct sockaddr *)&sdl; 3597 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3598 if (error) 3599 goto unlock_out; 3600 } 3601 3602 /* 3603 * Allocate the new address. Don't hook it up yet, as we may also 3604 * need to allocate a link layer multicast address. 3605 */ 3606 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3607 if (ifma == NULL) { 3608 error = ENOMEM; 3609 goto free_llsa_out; 3610 } 3611 3612 /* 3613 * If a link layer address is found, we'll need to see if it's 3614 * already present in the address list, or allocate is as well. 3615 * When this block finishes, the link layer address will be on the 3616 * list. 3617 */ 3618 if (llsa != NULL) { 3619 ll_ifma = if_findmulti(ifp, llsa); 3620 if (ll_ifma == NULL) { 3621 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3622 if (ll_ifma == NULL) { 3623 --ifma->ifma_refcount; 3624 if_freemulti(ifma); 3625 error = ENOMEM; 3626 goto free_llsa_out; 3627 } 3628 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3629 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3630 ifma_link); 3631 } else 3632 ll_ifma->ifma_refcount++; 3633 ifma->ifma_llifma = ll_ifma; 3634 } 3635 3636 /* 3637 * We now have a new multicast address, ifma, and possibly a new or 3638 * referenced link layer address. Add the primary address to the 3639 * ifnet address list. 3640 */ 3641 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3642 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3643 3644 if (retifma != NULL) 3645 *retifma = ifma; 3646 3647 /* 3648 * Must generate the message while holding the lock so that 'ifma' 3649 * pointer is still valid. 3650 */ 3651 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3652 IF_ADDR_WUNLOCK(ifp); 3653 3654 /* 3655 * We are certain we have added something, so call down to the 3656 * interface to let them know about it. 3657 */ 3658 if (ifp->if_ioctl != NULL) { 3659 if (THREAD_CAN_SLEEP()) 3660 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3661 else 3662 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3663 } 3664 3665 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3666 link_free_sdl(llsa); 3667 3668 return (0); 3669 3670 free_llsa_out: 3671 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3672 link_free_sdl(llsa); 3673 3674 unlock_out: 3675 IF_ADDR_WUNLOCK(ifp); 3676 return (error); 3677 } 3678 3679 static void 3680 if_siocaddmulti(void *arg, int pending) 3681 { 3682 struct ifnet *ifp; 3683 3684 ifp = arg; 3685 #ifdef DIAGNOSTIC 3686 if (pending > 1) 3687 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3688 #endif 3689 CURVNET_SET(ifp->if_vnet); 3690 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3691 CURVNET_RESTORE(); 3692 } 3693 3694 /* 3695 * Delete a multicast group membership by network-layer group address. 3696 * 3697 * Returns ENOENT if the entry could not be found. If ifp no longer 3698 * exists, results are undefined. This entry point should only be used 3699 * from subsystems which do appropriate locking to hold ifp for the 3700 * duration of the call. 3701 * Network-layer protocol domains must use if_delmulti_ifma(). 3702 */ 3703 int 3704 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3705 { 3706 struct ifmultiaddr *ifma; 3707 int lastref; 3708 3709 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3710 3711 IF_ADDR_WLOCK(ifp); 3712 lastref = 0; 3713 ifma = if_findmulti(ifp, sa); 3714 if (ifma != NULL) 3715 lastref = if_delmulti_locked(ifp, ifma, 0); 3716 IF_ADDR_WUNLOCK(ifp); 3717 3718 if (ifma == NULL) 3719 return (ENOENT); 3720 3721 if (lastref && ifp->if_ioctl != NULL) { 3722 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3723 } 3724 3725 return (0); 3726 } 3727 3728 /* 3729 * Delete all multicast group membership for an interface. 3730 * Should be used to quickly flush all multicast filters. 3731 */ 3732 void 3733 if_delallmulti(struct ifnet *ifp) 3734 { 3735 struct ifmultiaddr *ifma; 3736 struct ifmultiaddr *next; 3737 3738 IF_ADDR_WLOCK(ifp); 3739 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3740 if_delmulti_locked(ifp, ifma, 0); 3741 IF_ADDR_WUNLOCK(ifp); 3742 } 3743 3744 void 3745 if_delmulti_ifma(struct ifmultiaddr *ifma) 3746 { 3747 if_delmulti_ifma_flags(ifma, 0); 3748 } 3749 3750 /* 3751 * Delete a multicast group membership by group membership pointer. 3752 * Network-layer protocol domains must use this routine. 3753 * 3754 * It is safe to call this routine if the ifp disappeared. 3755 */ 3756 void 3757 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3758 { 3759 struct ifnet *ifp; 3760 int lastref; 3761 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3762 #ifdef INET 3763 IN_MULTI_LIST_UNLOCK_ASSERT(); 3764 #endif 3765 ifp = ifma->ifma_ifp; 3766 #ifdef DIAGNOSTIC 3767 if (ifp == NULL) { 3768 printf("%s: ifma_ifp seems to be detached\n", __func__); 3769 } else { 3770 struct epoch_tracker et; 3771 struct ifnet *oifp; 3772 3773 NET_EPOCH_ENTER(et); 3774 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3775 if (ifp == oifp) 3776 break; 3777 NET_EPOCH_EXIT(et); 3778 if (ifp != oifp) 3779 ifp = NULL; 3780 } 3781 #endif 3782 /* 3783 * If and only if the ifnet instance exists: Acquire the address lock. 3784 */ 3785 if (ifp != NULL) 3786 IF_ADDR_WLOCK(ifp); 3787 3788 lastref = if_delmulti_locked(ifp, ifma, flags); 3789 3790 if (ifp != NULL) { 3791 /* 3792 * If and only if the ifnet instance exists: 3793 * Release the address lock. 3794 * If the group was left: update the hardware hash filter. 3795 */ 3796 IF_ADDR_WUNLOCK(ifp); 3797 if (lastref && ifp->if_ioctl != NULL) { 3798 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3799 } 3800 } 3801 } 3802 3803 /* 3804 * Perform deletion of network-layer and/or link-layer multicast address. 3805 * 3806 * Return 0 if the reference count was decremented. 3807 * Return 1 if the final reference was released, indicating that the 3808 * hardware hash filter should be reprogrammed. 3809 */ 3810 static int 3811 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3812 { 3813 struct ifmultiaddr *ll_ifma; 3814 3815 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3816 KASSERT(ifma->ifma_ifp == ifp, 3817 ("%s: inconsistent ifp %p", __func__, ifp)); 3818 IF_ADDR_WLOCK_ASSERT(ifp); 3819 } 3820 3821 ifp = ifma->ifma_ifp; 3822 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3823 3824 /* 3825 * If the ifnet is detaching, null out references to ifnet, 3826 * so that upper protocol layers will notice, and not attempt 3827 * to obtain locks for an ifnet which no longer exists. The 3828 * routing socket announcement must happen before the ifnet 3829 * instance is detached from the system. 3830 */ 3831 if (detaching) { 3832 #ifdef DIAGNOSTIC 3833 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3834 #endif 3835 /* 3836 * ifp may already be nulled out if we are being reentered 3837 * to delete the ll_ifma. 3838 */ 3839 if (ifp != NULL) { 3840 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3841 ifma->ifma_ifp = NULL; 3842 } 3843 } 3844 3845 if (--ifma->ifma_refcount > 0) 3846 return 0; 3847 3848 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3849 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3850 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3851 } 3852 /* 3853 * If this ifma is a network-layer ifma, a link-layer ifma may 3854 * have been associated with it. Release it first if so. 3855 */ 3856 ll_ifma = ifma->ifma_llifma; 3857 if (ll_ifma != NULL) { 3858 KASSERT(ifma->ifma_lladdr != NULL, 3859 ("%s: llifma w/o lladdr", __func__)); 3860 if (detaching) 3861 ll_ifma->ifma_ifp = NULL; /* XXX */ 3862 if (--ll_ifma->ifma_refcount == 0) { 3863 if (ifp != NULL) { 3864 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3865 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3866 ifma_link); 3867 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3868 } 3869 } 3870 if_freemulti(ll_ifma); 3871 } 3872 } 3873 #ifdef INVARIANTS 3874 if (ifp) { 3875 struct ifmultiaddr *ifmatmp; 3876 3877 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3878 MPASS(ifma != ifmatmp); 3879 } 3880 #endif 3881 if_freemulti(ifma); 3882 /* 3883 * The last reference to this instance of struct ifmultiaddr 3884 * was released; the hardware should be notified of this change. 3885 */ 3886 return 1; 3887 } 3888 3889 /* 3890 * Set the link layer address on an interface. 3891 * 3892 * At this time we only support certain types of interfaces, 3893 * and we don't allow the length of the address to change. 3894 * 3895 * Set noinline to be dtrace-friendly 3896 */ 3897 __noinline int 3898 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3899 { 3900 struct sockaddr_dl *sdl; 3901 struct ifaddr *ifa; 3902 struct ifreq ifr; 3903 3904 ifa = ifp->if_addr; 3905 if (ifa == NULL) 3906 return (EINVAL); 3907 3908 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3909 if (sdl == NULL) 3910 return (EINVAL); 3911 3912 if (len != sdl->sdl_alen) /* don't allow length to change */ 3913 return (EINVAL); 3914 3915 switch (ifp->if_type) { 3916 case IFT_ETHER: 3917 case IFT_XETHER: 3918 case IFT_L2VLAN: 3919 case IFT_BRIDGE: 3920 case IFT_IEEE8023ADLAG: 3921 bcopy(lladdr, LLADDR(sdl), len); 3922 break; 3923 default: 3924 return (ENODEV); 3925 } 3926 3927 /* 3928 * If the interface is already up, we need 3929 * to re-init it in order to reprogram its 3930 * address filter. 3931 */ 3932 if ((ifp->if_flags & IFF_UP) != 0) { 3933 if (ifp->if_ioctl) { 3934 ifp->if_flags &= ~IFF_UP; 3935 ifr.ifr_flags = ifp->if_flags & 0xffff; 3936 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3937 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3938 ifp->if_flags |= IFF_UP; 3939 ifr.ifr_flags = ifp->if_flags & 0xffff; 3940 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3941 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3942 } 3943 } 3944 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3945 3946 return (0); 3947 } 3948 3949 /* 3950 * Compat function for handling basic encapsulation requests. 3951 * Not converted stacks (FDDI, IB, ..) supports traditional 3952 * output model: ARP (and other similar L2 protocols) are handled 3953 * inside output routine, arpresolve/nd6_resolve() returns MAC 3954 * address instead of full prepend. 3955 * 3956 * This function creates calculated header==MAC for IPv4/IPv6 and 3957 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3958 * address families. 3959 */ 3960 static int 3961 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3962 { 3963 if (req->rtype != IFENCAP_LL) 3964 return (EOPNOTSUPP); 3965 3966 if (req->bufsize < req->lladdr_len) 3967 return (ENOMEM); 3968 3969 switch (req->family) { 3970 case AF_INET: 3971 case AF_INET6: 3972 break; 3973 default: 3974 return (EAFNOSUPPORT); 3975 } 3976 3977 /* Copy lladdr to storage as is */ 3978 memmove(req->buf, req->lladdr, req->lladdr_len); 3979 req->bufsize = req->lladdr_len; 3980 req->lladdr_off = 0; 3981 3982 return (0); 3983 } 3984 3985 /* 3986 * Tunnel interfaces can nest, also they may cause infinite recursion 3987 * calls when misconfigured. We'll prevent this by detecting loops. 3988 * High nesting level may cause stack exhaustion. We'll prevent this 3989 * by introducing upper limit. 3990 * 3991 * Return 0, if tunnel nesting count is equal or less than limit. 3992 */ 3993 int 3994 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3995 int limit) 3996 { 3997 struct m_tag *mtag; 3998 int count; 3999 4000 count = 1; 4001 mtag = NULL; 4002 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 4003 if (*(struct ifnet **)(mtag + 1) == ifp) { 4004 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 4005 return (EIO); 4006 } 4007 count++; 4008 } 4009 if (count > limit) { 4010 log(LOG_NOTICE, 4011 "%s: if_output recursively called too many times(%d)\n", 4012 if_name(ifp), count); 4013 return (EIO); 4014 } 4015 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 4016 if (mtag == NULL) 4017 return (ENOMEM); 4018 *(struct ifnet **)(mtag + 1) = ifp; 4019 m_tag_prepend(m, mtag); 4020 return (0); 4021 } 4022 4023 /* 4024 * Get the link layer address that was read from the hardware at attach. 4025 * 4026 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 4027 * their component interfaces as IFT_IEEE8023ADLAG. 4028 */ 4029 int 4030 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 4031 { 4032 if (ifp->if_hw_addr == NULL) 4033 return (ENODEV); 4034 4035 switch (ifp->if_type) { 4036 case IFT_ETHER: 4037 case IFT_IEEE8023ADLAG: 4038 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 4039 return (0); 4040 default: 4041 return (ENODEV); 4042 } 4043 } 4044 4045 /* 4046 * The name argument must be a pointer to storage which will last as 4047 * long as the interface does. For physical devices, the result of 4048 * device_get_name(dev) is a good choice and for pseudo-devices a 4049 * static string works well. 4050 */ 4051 void 4052 if_initname(struct ifnet *ifp, const char *name, int unit) 4053 { 4054 ifp->if_dname = name; 4055 ifp->if_dunit = unit; 4056 if (unit != IF_DUNIT_NONE) 4057 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 4058 else 4059 strlcpy(ifp->if_xname, name, IFNAMSIZ); 4060 } 4061 4062 static int 4063 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 4064 { 4065 char if_fmt[256]; 4066 4067 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 4068 vlog(pri, if_fmt, ap); 4069 return (0); 4070 } 4071 4072 4073 int 4074 if_printf(struct ifnet *ifp, const char *fmt, ...) 4075 { 4076 va_list ap; 4077 4078 va_start(ap, fmt); 4079 if_vlog(ifp, LOG_INFO, fmt, ap); 4080 va_end(ap); 4081 return (0); 4082 } 4083 4084 int 4085 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 4086 { 4087 va_list ap; 4088 4089 va_start(ap, fmt); 4090 if_vlog(ifp, pri, fmt, ap); 4091 va_end(ap); 4092 return (0); 4093 } 4094 4095 void 4096 if_start(struct ifnet *ifp) 4097 { 4098 4099 (*(ifp)->if_start)(ifp); 4100 } 4101 4102 /* 4103 * Backwards compatibility interface for drivers 4104 * that have not implemented it 4105 */ 4106 static int 4107 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 4108 { 4109 int error; 4110 4111 IFQ_HANDOFF(ifp, m, error); 4112 return (error); 4113 } 4114 4115 static void 4116 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4117 { 4118 m_freem(m); 4119 } 4120 4121 int 4122 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4123 { 4124 int active = 0; 4125 4126 IF_LOCK(ifq); 4127 if (_IF_QFULL(ifq)) { 4128 IF_UNLOCK(ifq); 4129 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4130 m_freem(m); 4131 return (0); 4132 } 4133 if (ifp != NULL) { 4134 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4135 if (m->m_flags & (M_BCAST|M_MCAST)) 4136 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4137 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4138 } 4139 _IF_ENQUEUE(ifq, m); 4140 IF_UNLOCK(ifq); 4141 if (ifp != NULL && !active) 4142 (*(ifp)->if_start)(ifp); 4143 return (1); 4144 } 4145 4146 void 4147 if_register_com_alloc(u_char type, 4148 if_com_alloc_t *a, if_com_free_t *f) 4149 { 4150 4151 KASSERT(if_com_alloc[type] == NULL, 4152 ("if_register_com_alloc: %d already registered", type)); 4153 KASSERT(if_com_free[type] == NULL, 4154 ("if_register_com_alloc: %d free already registered", type)); 4155 4156 if_com_alloc[type] = a; 4157 if_com_free[type] = f; 4158 } 4159 4160 void 4161 if_deregister_com_alloc(u_char type) 4162 { 4163 4164 KASSERT(if_com_alloc[type] != NULL, 4165 ("if_deregister_com_alloc: %d not registered", type)); 4166 KASSERT(if_com_free[type] != NULL, 4167 ("if_deregister_com_alloc: %d free not registered", type)); 4168 4169 /* 4170 * Ensure all pending EPOCH(9) callbacks have been executed. This 4171 * fixes issues about late invocation of if_destroy(), which leads 4172 * to memory leak from if_com_alloc[type] allocated if_l2com. 4173 */ 4174 NET_EPOCH_DRAIN_CALLBACKS(); 4175 4176 if_com_alloc[type] = NULL; 4177 if_com_free[type] = NULL; 4178 } 4179 4180 /* API for driver access to network stack owned ifnet.*/ 4181 uint64_t 4182 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4183 { 4184 uint64_t oldbrate; 4185 4186 oldbrate = ifp->if_baudrate; 4187 ifp->if_baudrate = baudrate; 4188 return (oldbrate); 4189 } 4190 4191 uint64_t 4192 if_getbaudrate(const if_t ifp) 4193 { 4194 return (ifp->if_baudrate); 4195 } 4196 4197 int 4198 if_setcapabilities(if_t ifp, int capabilities) 4199 { 4200 ifp->if_capabilities = capabilities; 4201 return (0); 4202 } 4203 4204 int 4205 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4206 { 4207 ifp->if_capabilities &= ~clearbit; 4208 ifp->if_capabilities |= setbit; 4209 return (0); 4210 } 4211 4212 int 4213 if_getcapabilities(const if_t ifp) 4214 { 4215 return (ifp->if_capabilities); 4216 } 4217 4218 int 4219 if_setcapenable(if_t ifp, int capabilities) 4220 { 4221 ifp->if_capenable = capabilities; 4222 return (0); 4223 } 4224 4225 int 4226 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4227 { 4228 ifp->if_capenable &= ~clearcap; 4229 ifp->if_capenable |= setcap; 4230 return (0); 4231 } 4232 4233 int 4234 if_setcapabilities2(if_t ifp, int capabilities) 4235 { 4236 ifp->if_capabilities2 = capabilities; 4237 return (0); 4238 } 4239 4240 int 4241 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit) 4242 { 4243 ifp->if_capabilities2 &= ~clearbit; 4244 ifp->if_capabilities2 |= setbit; 4245 return (0); 4246 } 4247 4248 int 4249 if_getcapabilities2(const if_t ifp) 4250 { 4251 return (ifp->if_capabilities2); 4252 } 4253 4254 int 4255 if_setcapenable2(if_t ifp, int capabilities2) 4256 { 4257 ifp->if_capenable2 = capabilities2; 4258 return (0); 4259 } 4260 4261 int 4262 if_setcapenable2bit(if_t ifp, int setcap, int clearcap) 4263 { 4264 ifp->if_capenable2 &= ~clearcap; 4265 ifp->if_capenable2 |= setcap; 4266 return (0); 4267 } 4268 4269 const char * 4270 if_getdname(const if_t ifp) 4271 { 4272 return (ifp->if_dname); 4273 } 4274 4275 void 4276 if_setdname(if_t ifp, const char *dname) 4277 { 4278 ifp->if_dname = dname; 4279 } 4280 4281 const char * 4282 if_name(if_t ifp) 4283 { 4284 return (ifp->if_xname); 4285 } 4286 4287 int 4288 if_setname(if_t ifp, const char *name) 4289 { 4290 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4291 return (ENAMETOOLONG); 4292 strcpy(ifp->if_xname, name); 4293 4294 return (0); 4295 } 4296 4297 int 4298 if_togglecapenable(if_t ifp, int togglecap) 4299 { 4300 ifp->if_capenable ^= togglecap; 4301 return (0); 4302 } 4303 4304 int 4305 if_getcapenable(const if_t ifp) 4306 { 4307 return (ifp->if_capenable); 4308 } 4309 4310 int 4311 if_togglecapenable2(if_t ifp, int togglecap) 4312 { 4313 ifp->if_capenable2 ^= togglecap; 4314 return (0); 4315 } 4316 4317 int 4318 if_getcapenable2(const if_t ifp) 4319 { 4320 return (ifp->if_capenable2); 4321 } 4322 4323 int 4324 if_getdunit(const if_t ifp) 4325 { 4326 return (ifp->if_dunit); 4327 } 4328 4329 int 4330 if_getindex(const if_t ifp) 4331 { 4332 return (ifp->if_index); 4333 } 4334 4335 int 4336 if_getidxgen(const if_t ifp) 4337 { 4338 return (ifp->if_idxgen); 4339 } 4340 4341 void 4342 if_setdescr(if_t ifp, char *descrbuf) 4343 { 4344 sx_xlock(&ifdescr_sx); 4345 char *odescrbuf = ifp->if_description; 4346 ifp->if_description = descrbuf; 4347 sx_xunlock(&ifdescr_sx); 4348 4349 if_freedescr(odescrbuf); 4350 } 4351 4352 char * 4353 if_allocdescr(size_t sz, int malloc_flag) 4354 { 4355 malloc_flag &= (M_WAITOK | M_NOWAIT); 4356 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4357 } 4358 4359 void 4360 if_freedescr(char *descrbuf) 4361 { 4362 free(descrbuf, M_IFDESCR); 4363 } 4364 4365 int 4366 if_getalloctype(const if_t ifp) 4367 { 4368 return (ifp->if_alloctype); 4369 } 4370 4371 /* 4372 * This is largely undesirable because it ties ifnet to a device, but does 4373 * provide flexiblity for an embedded product vendor. Should be used with 4374 * the understanding that it violates the interface boundaries, and should be 4375 * a last resort only. 4376 */ 4377 int 4378 if_setdev(if_t ifp, void *dev) 4379 { 4380 return (0); 4381 } 4382 4383 int 4384 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4385 { 4386 ifp->if_drv_flags &= ~clear_flags; 4387 ifp->if_drv_flags |= set_flags; 4388 4389 return (0); 4390 } 4391 4392 int 4393 if_getdrvflags(const if_t ifp) 4394 { 4395 return (ifp->if_drv_flags); 4396 } 4397 4398 int 4399 if_setdrvflags(if_t ifp, int flags) 4400 { 4401 ifp->if_drv_flags = flags; 4402 return (0); 4403 } 4404 4405 int 4406 if_setflags(if_t ifp, int flags) 4407 { 4408 ifp->if_flags = flags; 4409 return (0); 4410 } 4411 4412 int 4413 if_setflagbits(if_t ifp, int set, int clear) 4414 { 4415 ifp->if_flags &= ~clear; 4416 ifp->if_flags |= set; 4417 return (0); 4418 } 4419 4420 int 4421 if_getflags(const if_t ifp) 4422 { 4423 return (ifp->if_flags); 4424 } 4425 4426 int 4427 if_clearhwassist(if_t ifp) 4428 { 4429 ifp->if_hwassist = 0; 4430 return (0); 4431 } 4432 4433 int 4434 if_sethwassistbits(if_t ifp, int toset, int toclear) 4435 { 4436 ifp->if_hwassist &= ~toclear; 4437 ifp->if_hwassist |= toset; 4438 4439 return (0); 4440 } 4441 4442 int 4443 if_sethwassist(if_t ifp, int hwassist_bit) 4444 { 4445 ifp->if_hwassist = hwassist_bit; 4446 return (0); 4447 } 4448 4449 int 4450 if_gethwassist(const if_t ifp) 4451 { 4452 return (ifp->if_hwassist); 4453 } 4454 4455 int 4456 if_togglehwassist(if_t ifp, int toggle_bits) 4457 { 4458 ifp->if_hwassist ^= toggle_bits; 4459 return (0); 4460 } 4461 4462 int 4463 if_setmtu(if_t ifp, int mtu) 4464 { 4465 ifp->if_mtu = mtu; 4466 return (0); 4467 } 4468 4469 void 4470 if_notifymtu(if_t ifp) 4471 { 4472 #ifdef INET6 4473 nd6_setmtu(ifp); 4474 #endif 4475 rt_updatemtu(ifp); 4476 } 4477 4478 int 4479 if_getmtu(const if_t ifp) 4480 { 4481 return (ifp->if_mtu); 4482 } 4483 4484 int 4485 if_getmtu_family(const if_t ifp, int family) 4486 { 4487 struct domain *dp; 4488 4489 SLIST_FOREACH(dp, &domains, dom_next) { 4490 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4491 return (dp->dom_ifmtu(ifp)); 4492 } 4493 4494 return (ifp->if_mtu); 4495 } 4496 4497 /* 4498 * Methods for drivers to access interface unicast and multicast 4499 * link level addresses. Driver shall not know 'struct ifaddr' neither 4500 * 'struct ifmultiaddr'. 4501 */ 4502 u_int 4503 if_lladdr_count(if_t ifp) 4504 { 4505 struct epoch_tracker et; 4506 struct ifaddr *ifa; 4507 u_int count; 4508 4509 count = 0; 4510 NET_EPOCH_ENTER(et); 4511 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4512 if (ifa->ifa_addr->sa_family == AF_LINK) 4513 count++; 4514 NET_EPOCH_EXIT(et); 4515 4516 return (count); 4517 } 4518 4519 int 4520 if_foreach(if_foreach_cb_t cb, void *cb_arg) 4521 { 4522 if_t ifp; 4523 int error; 4524 4525 NET_EPOCH_ASSERT(); 4526 MPASS(cb); 4527 4528 error = 0; 4529 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4530 error = cb(ifp, cb_arg); 4531 if (error != 0) 4532 break; 4533 } 4534 4535 return (error); 4536 } 4537 4538 /* 4539 * Iterates over the list of interfaces, permitting callback function @cb to sleep. 4540 * Stops iteration if @cb returns non-zero error code. 4541 * Returns the last error code from @cb. 4542 * @match_cb: optional match callback limiting the iteration to only matched interfaces 4543 * @match_arg: argument to pass to @match_cb 4544 * @cb: iteration callback 4545 * @cb_arg: argument to pass to @cb 4546 */ 4547 int 4548 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb, 4549 void *cb_arg) 4550 { 4551 int match_count = 0, array_size = 16; /* 128 bytes for malloc */ 4552 struct ifnet **match_array = NULL; 4553 int error = 0; 4554 4555 MPASS(cb); 4556 4557 while (true) { 4558 struct ifnet **new_array; 4559 int new_size = array_size; 4560 struct epoch_tracker et; 4561 struct ifnet *ifp; 4562 4563 while (new_size < match_count) 4564 new_size *= 2; 4565 new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK); 4566 if (match_array != NULL) 4567 memcpy(new_array, match_array, array_size * sizeof(void *)); 4568 free(match_array, M_TEMP); 4569 match_array = new_array; 4570 array_size = new_size; 4571 4572 match_count = 0; 4573 NET_EPOCH_ENTER(et); 4574 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4575 if (match_cb != NULL && !match_cb(ifp, match_arg)) 4576 continue; 4577 if (match_count < array_size) { 4578 if (if_try_ref(ifp)) 4579 match_array[match_count++] = ifp; 4580 } else 4581 match_count++; 4582 } 4583 NET_EPOCH_EXIT(et); 4584 4585 if (match_count > array_size) { 4586 for (int i = 0; i < array_size; i++) 4587 if_rele(match_array[i]); 4588 continue; 4589 } else { 4590 for (int i = 0; i < match_count; i++) { 4591 if (error == 0) 4592 error = cb(match_array[i], cb_arg); 4593 if_rele(match_array[i]); 4594 } 4595 free(match_array, M_TEMP); 4596 break; 4597 } 4598 } 4599 4600 return (error); 4601 } 4602 4603 4604 /* 4605 * Uses just 1 pointer of the 4 available in the public struct. 4606 */ 4607 if_t 4608 if_iter_start(struct if_iter *iter) 4609 { 4610 if_t ifp; 4611 4612 NET_EPOCH_ASSERT(); 4613 4614 bzero(iter, sizeof(*iter)); 4615 ifp = CK_STAILQ_FIRST(&V_ifnet); 4616 if (ifp != NULL) 4617 iter->context[0] = CK_STAILQ_NEXT(ifp, if_link); 4618 else 4619 iter->context[0] = NULL; 4620 return (ifp); 4621 } 4622 4623 if_t 4624 if_iter_next(struct if_iter *iter) 4625 { 4626 if_t cur_ifp = iter->context[0]; 4627 4628 if (cur_ifp != NULL) 4629 iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link); 4630 return (cur_ifp); 4631 } 4632 4633 void 4634 if_iter_finish(struct if_iter *iter) 4635 { 4636 /* Nothing to do here for now. */ 4637 } 4638 4639 u_int 4640 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4641 { 4642 struct epoch_tracker et; 4643 struct ifaddr *ifa; 4644 u_int count; 4645 4646 MPASS(cb); 4647 4648 count = 0; 4649 NET_EPOCH_ENTER(et); 4650 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4651 if (ifa->ifa_addr->sa_family != AF_LINK) 4652 continue; 4653 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4654 count); 4655 } 4656 NET_EPOCH_EXIT(et); 4657 4658 return (count); 4659 } 4660 4661 u_int 4662 if_llmaddr_count(if_t ifp) 4663 { 4664 struct epoch_tracker et; 4665 struct ifmultiaddr *ifma; 4666 int count; 4667 4668 count = 0; 4669 NET_EPOCH_ENTER(et); 4670 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4671 if (ifma->ifma_addr->sa_family == AF_LINK) 4672 count++; 4673 NET_EPOCH_EXIT(et); 4674 4675 return (count); 4676 } 4677 4678 u_int 4679 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4680 { 4681 struct epoch_tracker et; 4682 struct ifmultiaddr *ifma; 4683 u_int count; 4684 4685 MPASS(cb); 4686 4687 count = 0; 4688 NET_EPOCH_ENTER(et); 4689 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4690 if (ifma->ifma_addr->sa_family != AF_LINK) 4691 continue; 4692 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4693 count); 4694 } 4695 NET_EPOCH_EXIT(et); 4696 4697 return (count); 4698 } 4699 4700 u_int 4701 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4702 { 4703 struct epoch_tracker et; 4704 struct ifaddr *ifa; 4705 u_int count; 4706 4707 MPASS(cb); 4708 4709 count = 0; 4710 NET_EPOCH_ENTER(et); 4711 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4712 if (ifa->ifa_addr->sa_family != type) 4713 continue; 4714 count += (*cb)(cb_arg, ifa, count); 4715 } 4716 NET_EPOCH_EXIT(et); 4717 4718 return (count); 4719 } 4720 4721 int 4722 if_setsoftc(if_t ifp, void *softc) 4723 { 4724 ifp->if_softc = softc; 4725 return (0); 4726 } 4727 4728 void * 4729 if_getsoftc(const if_t ifp) 4730 { 4731 return (ifp->if_softc); 4732 } 4733 4734 void 4735 if_setrcvif(struct mbuf *m, if_t ifp) 4736 { 4737 4738 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4739 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4740 } 4741 4742 void 4743 if_setvtag(struct mbuf *m, uint16_t tag) 4744 { 4745 m->m_pkthdr.ether_vtag = tag; 4746 } 4747 4748 uint16_t 4749 if_getvtag(struct mbuf *m) 4750 { 4751 return (m->m_pkthdr.ether_vtag); 4752 } 4753 4754 int 4755 if_sendq_empty(if_t ifp) 4756 { 4757 return (IFQ_DRV_IS_EMPTY(&ifp->if_snd)); 4758 } 4759 4760 struct ifaddr * 4761 if_getifaddr(const if_t ifp) 4762 { 4763 return (ifp->if_addr); 4764 } 4765 4766 int 4767 if_getamcount(const if_t ifp) 4768 { 4769 return (ifp->if_amcount); 4770 } 4771 4772 int 4773 if_setsendqready(if_t ifp) 4774 { 4775 IFQ_SET_READY(&ifp->if_snd); 4776 return (0); 4777 } 4778 4779 int 4780 if_setsendqlen(if_t ifp, int tx_desc_count) 4781 { 4782 IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count); 4783 ifp->if_snd.ifq_drv_maxlen = tx_desc_count; 4784 return (0); 4785 } 4786 4787 void 4788 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na) 4789 { 4790 ifp->if_netmap = na; 4791 } 4792 4793 struct netmap_adapter * 4794 if_getnetmapadapter(if_t ifp) 4795 { 4796 return (ifp->if_netmap); 4797 } 4798 4799 int 4800 if_vlantrunkinuse(if_t ifp) 4801 { 4802 return (ifp->if_vlantrunk != NULL); 4803 } 4804 4805 void 4806 if_init(if_t ifp, void *ctx) 4807 { 4808 (*ifp->if_init)(ctx); 4809 } 4810 4811 void 4812 if_input(if_t ifp, struct mbuf* sendmp) 4813 { 4814 (*ifp->if_input)(ifp, sendmp); 4815 } 4816 4817 int 4818 if_transmit(if_t ifp, struct mbuf *m) 4819 { 4820 return ((*ifp->if_transmit)(ifp, m)); 4821 } 4822 4823 int 4824 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst) 4825 { 4826 if (ifp->if_resolvemulti == NULL) 4827 return (EOPNOTSUPP); 4828 4829 return (ifp->if_resolvemulti(ifp, srcs, dst)); 4830 } 4831 4832 struct mbuf * 4833 if_dequeue(if_t ifp) 4834 { 4835 struct mbuf *m; 4836 4837 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 4838 return (m); 4839 } 4840 4841 int 4842 if_sendq_prepend(if_t ifp, struct mbuf *m) 4843 { 4844 IFQ_DRV_PREPEND(&ifp->if_snd, m); 4845 return (0); 4846 } 4847 4848 int 4849 if_setifheaderlen(if_t ifp, int len) 4850 { 4851 ifp->if_hdrlen = len; 4852 return (0); 4853 } 4854 4855 caddr_t 4856 if_getlladdr(const if_t ifp) 4857 { 4858 return (IF_LLADDR(ifp)); 4859 } 4860 4861 void * 4862 if_gethandle(u_char type) 4863 { 4864 return (if_alloc(type)); 4865 } 4866 4867 void 4868 if_bpfmtap(if_t ifp, struct mbuf *m) 4869 { 4870 BPF_MTAP(ifp, m); 4871 } 4872 4873 void 4874 if_etherbpfmtap(if_t ifp, struct mbuf *m) 4875 { 4876 ETHER_BPF_MTAP(ifp, m); 4877 } 4878 4879 void 4880 if_vlancap(if_t ifp) 4881 { 4882 VLAN_CAPABILITIES(ifp); 4883 } 4884 4885 int 4886 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4887 { 4888 ifp->if_hw_tsomax = if_hw_tsomax; 4889 return (0); 4890 } 4891 4892 int 4893 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4894 { 4895 ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4896 return (0); 4897 } 4898 4899 int 4900 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4901 { 4902 ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4903 return (0); 4904 } 4905 4906 u_int 4907 if_gethwtsomax(const if_t ifp) 4908 { 4909 return (ifp->if_hw_tsomax); 4910 } 4911 4912 u_int 4913 if_gethwtsomaxsegcount(const if_t ifp) 4914 { 4915 return (ifp->if_hw_tsomaxsegcount); 4916 } 4917 4918 u_int 4919 if_gethwtsomaxsegsize(const if_t ifp) 4920 { 4921 return (ifp->if_hw_tsomaxsegsize); 4922 } 4923 4924 void 4925 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4926 { 4927 ifp->if_init = init_fn; 4928 } 4929 4930 void 4931 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4932 { 4933 ifp->if_input = input_fn; 4934 } 4935 4936 if_input_fn_t 4937 if_getinputfn(if_t ifp) 4938 { 4939 return (ifp->if_input); 4940 } 4941 4942 void 4943 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4944 { 4945 ifp->if_ioctl = ioctl_fn; 4946 } 4947 4948 void 4949 if_setoutputfn(if_t ifp, if_output_fn_t output_fn) 4950 { 4951 ifp->if_output = output_fn; 4952 } 4953 4954 void 4955 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4956 { 4957 ifp->if_start = start_fn; 4958 } 4959 4960 if_start_fn_t 4961 if_getstartfn(if_t ifp) 4962 { 4963 return (ifp->if_start); 4964 } 4965 4966 void 4967 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4968 { 4969 ifp->if_transmit = start_fn; 4970 } 4971 4972 if_transmit_fn_t 4973 if_gettransmitfn(if_t ifp) 4974 { 4975 return (ifp->if_transmit); 4976 } 4977 4978 void 4979 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4980 { 4981 ifp->if_qflush = flush_fn; 4982 } 4983 4984 void 4985 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 4986 { 4987 ifp->if_snd_tag_alloc = alloc_fn; 4988 } 4989 4990 int 4991 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params, 4992 struct m_snd_tag **mstp) 4993 { 4994 if (ifp->if_snd_tag_alloc == NULL) 4995 return (EOPNOTSUPP); 4996 return (ifp->if_snd_tag_alloc(ifp, params, mstp)); 4997 } 4998 4999 void 5000 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 5001 { 5002 ifp->if_get_counter = fn; 5003 } 5004 5005 void 5006 if_setreassignfn(if_t ifp, if_reassign_fn_t fn) 5007 { 5008 ifp->if_reassign = fn; 5009 } 5010 5011 void 5012 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn) 5013 { 5014 ifp->if_ratelimit_query = fn; 5015 } 5016 5017 void 5018 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m) 5019 { 5020 ifp->if_debugnet_methods = m; 5021 } 5022 5023 struct label * 5024 if_getmaclabel(if_t ifp) 5025 { 5026 return (ifp->if_label); 5027 } 5028 5029 void 5030 if_setmaclabel(if_t ifp, struct label *label) 5031 { 5032 ifp->if_label = label; 5033 } 5034 5035 int 5036 if_gettype(if_t ifp) 5037 { 5038 return (ifp->if_type); 5039 } 5040 5041 void * 5042 if_getllsoftc(if_t ifp) 5043 { 5044 return (ifp->if_llsoftc); 5045 } 5046 5047 void 5048 if_setllsoftc(if_t ifp, void *llsoftc) 5049 { 5050 ifp->if_llsoftc = llsoftc; 5051 }; 5052 5053 int 5054 if_getlinkstate(if_t ifp) 5055 { 5056 return (ifp->if_link_state); 5057 } 5058 5059 const uint8_t * 5060 if_getbroadcastaddr(if_t ifp) 5061 { 5062 return (ifp->if_broadcastaddr); 5063 } 5064 5065 void 5066 if_setbroadcastaddr(if_t ifp, const uint8_t *addr) 5067 { 5068 ifp->if_broadcastaddr = addr; 5069 } 5070 5071 int 5072 if_getnumadomain(if_t ifp) 5073 { 5074 return (ifp->if_numa_domain); 5075 } 5076 5077 uint64_t 5078 if_getcounter(if_t ifp, ift_counter counter) 5079 { 5080 return (ifp->if_get_counter(ifp, counter)); 5081 } 5082 5083 bool 5084 if_altq_is_enabled(if_t ifp) 5085 { 5086 return (ALTQ_IS_ENABLED(&ifp->if_snd)); 5087 } 5088 5089 struct vnet * 5090 if_getvnet(if_t ifp) 5091 { 5092 return (ifp->if_vnet); 5093 } 5094 5095 void * 5096 if_getafdata(if_t ifp, int af) 5097 { 5098 return (ifp->if_afdata[af]); 5099 } 5100 5101 u_int 5102 if_getfib(if_t ifp) 5103 { 5104 return (ifp->if_fib); 5105 } 5106 5107 uint8_t 5108 if_getaddrlen(if_t ifp) 5109 { 5110 return (ifp->if_addrlen); 5111 } 5112 5113 struct bpf_if * 5114 if_getbpf(if_t ifp) 5115 { 5116 return (ifp->if_bpf); 5117 } 5118 5119 struct ifvlantrunk * 5120 if_getvlantrunk(if_t ifp) 5121 { 5122 return (ifp->if_vlantrunk); 5123 } 5124 5125 uint8_t 5126 if_getpcp(if_t ifp) 5127 { 5128 return (ifp->if_pcp); 5129 } 5130 5131 void * 5132 if_getl2com(if_t ifp) 5133 { 5134 return (ifp->if_l2com); 5135 } 5136 5137 #ifdef DDB 5138 static void 5139 if_show_ifnet(struct ifnet *ifp) 5140 { 5141 if (ifp == NULL) 5142 return; 5143 db_printf("%s:\n", ifp->if_xname); 5144 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 5145 IF_DB_PRINTF("%s", if_dname); 5146 IF_DB_PRINTF("%d", if_dunit); 5147 IF_DB_PRINTF("%s", if_description); 5148 IF_DB_PRINTF("%u", if_index); 5149 IF_DB_PRINTF("%d", if_idxgen); 5150 IF_DB_PRINTF("%u", if_refcount); 5151 IF_DB_PRINTF("%p", if_softc); 5152 IF_DB_PRINTF("%p", if_l2com); 5153 IF_DB_PRINTF("%p", if_llsoftc); 5154 IF_DB_PRINTF("%d", if_amcount); 5155 IF_DB_PRINTF("%p", if_addr); 5156 IF_DB_PRINTF("%p", if_broadcastaddr); 5157 IF_DB_PRINTF("%p", if_afdata); 5158 IF_DB_PRINTF("%d", if_afdata_initialized); 5159 IF_DB_PRINTF("%u", if_fib); 5160 IF_DB_PRINTF("%p", if_vnet); 5161 IF_DB_PRINTF("%p", if_home_vnet); 5162 IF_DB_PRINTF("%p", if_vlantrunk); 5163 IF_DB_PRINTF("%p", if_bpf); 5164 IF_DB_PRINTF("%u", if_pcount); 5165 IF_DB_PRINTF("%p", if_bridge); 5166 IF_DB_PRINTF("%p", if_lagg); 5167 IF_DB_PRINTF("%p", if_pf_kif); 5168 IF_DB_PRINTF("%p", if_carp); 5169 IF_DB_PRINTF("%p", if_label); 5170 IF_DB_PRINTF("%p", if_netmap); 5171 IF_DB_PRINTF("0x%08x", if_flags); 5172 IF_DB_PRINTF("0x%08x", if_drv_flags); 5173 IF_DB_PRINTF("0x%08x", if_capabilities); 5174 IF_DB_PRINTF("0x%08x", if_capenable); 5175 IF_DB_PRINTF("%p", if_snd.ifq_head); 5176 IF_DB_PRINTF("%p", if_snd.ifq_tail); 5177 IF_DB_PRINTF("%d", if_snd.ifq_len); 5178 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 5179 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 5180 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 5181 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 5182 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 5183 IF_DB_PRINTF("%d", if_snd.altq_type); 5184 IF_DB_PRINTF("%x", if_snd.altq_flags); 5185 #undef IF_DB_PRINTF 5186 } 5187 5188 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 5189 { 5190 if (!have_addr) { 5191 db_printf("usage: show ifnet <struct ifnet *>\n"); 5192 return; 5193 } 5194 5195 if_show_ifnet((struct ifnet *)addr); 5196 } 5197 5198 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 5199 { 5200 struct ifnet *ifp; 5201 u_short idx; 5202 5203 for (idx = 1; idx <= if_index; idx++) { 5204 ifp = ifindex_table[idx].ife_ifnet; 5205 if (ifp == NULL) 5206 continue; 5207 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 5208 if (db_pager_quit) 5209 break; 5210 } 5211 } 5212 #endif /* DDB */ 5213