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 33 #include "opt_bpf.h" 34 #include "opt_inet6.h" 35 #include "opt_inet.h" 36 #include "opt_ddb.h" 37 38 #include <sys/param.h> 39 #include <sys/capsicum.h> 40 #include <sys/conf.h> 41 #include <sys/eventhandler.h> 42 #include <sys/malloc.h> 43 #include <sys/domainset.h> 44 #include <sys/sbuf.h> 45 #include <sys/bus.h> 46 #include <sys/epoch.h> 47 #include <sys/mbuf.h> 48 #include <sys/systm.h> 49 #include <sys/priv.h> 50 #include <sys/proc.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/protosw.h> 54 #include <sys/kernel.h> 55 #include <sys/lock.h> 56 #include <sys/refcount.h> 57 #include <sys/module.h> 58 #include <sys/nv.h> 59 #include <sys/rwlock.h> 60 #include <sys/sockio.h> 61 #include <sys/syslog.h> 62 #include <sys/sysctl.h> 63 #include <sys/sysent.h> 64 #include <sys/taskqueue.h> 65 #include <sys/domain.h> 66 #include <sys/jail.h> 67 #include <sys/priv.h> 68 69 #ifdef DDB 70 #include <ddb/ddb.h> 71 #endif 72 73 #include <machine/stdarg.h> 74 #include <vm/uma.h> 75 76 #include <net/bpf.h> 77 #include <net/ethernet.h> 78 #include <net/if.h> 79 #include <net/if_arp.h> 80 #include <net/if_clone.h> 81 #include <net/if_dl.h> 82 #include <net/if_strings.h> 83 #include <net/if_types.h> 84 #include <net/if_var.h> 85 #include <net/if_media.h> 86 #include <net/if_mib.h> 87 #include <net/if_private.h> 88 #include <net/if_vlan_var.h> 89 #include <net/radix.h> 90 #include <net/route.h> 91 #include <net/route/route_ctl.h> 92 #include <net/vnet.h> 93 94 #if defined(INET) || defined(INET6) 95 #include <net/ethernet.h> 96 #include <netinet/in.h> 97 #include <netinet/in_var.h> 98 #include <netinet/ip.h> 99 #include <netinet/ip_carp.h> 100 #ifdef INET 101 #include <net/debugnet.h> 102 #include <netinet/if_ether.h> 103 #endif /* INET */ 104 #ifdef INET6 105 #include <netinet6/in6_var.h> 106 #include <netinet6/in6_ifattach.h> 107 #endif /* INET6 */ 108 #endif /* INET || INET6 */ 109 110 #include <security/mac/mac_framework.h> 111 112 /* 113 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 114 * and ifr_ifru when it is used in SIOCGIFCONF. 115 */ 116 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 117 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 118 119 __read_mostly epoch_t net_epoch_preempt; 120 #ifdef COMPAT_FREEBSD32 121 #include <sys/mount.h> 122 #include <compat/freebsd32/freebsd32.h> 123 124 struct ifreq_buffer32 { 125 uint32_t length; /* (size_t) */ 126 uint32_t buffer; /* (void *) */ 127 }; 128 129 /* 130 * Interface request structure used for socket 131 * ioctl's. All interface ioctl's must have parameter 132 * definitions which begin with ifr_name. The 133 * remainder may be interface specific. 134 */ 135 struct ifreq32 { 136 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 137 union { 138 struct sockaddr ifru_addr; 139 struct sockaddr ifru_dstaddr; 140 struct sockaddr ifru_broadaddr; 141 struct ifreq_buffer32 ifru_buffer; 142 short ifru_flags[2]; 143 short ifru_index; 144 int ifru_jid; 145 int ifru_metric; 146 int ifru_mtu; 147 int ifru_phys; 148 int ifru_media; 149 uint32_t ifru_data; 150 int ifru_cap[2]; 151 u_int ifru_fib; 152 u_char ifru_vlan_pcp; 153 } ifr_ifru; 154 }; 155 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 156 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 157 __offsetof(struct ifreq32, ifr_ifru)); 158 159 struct ifconf32 { 160 int32_t ifc_len; 161 union { 162 uint32_t ifcu_buf; 163 uint32_t ifcu_req; 164 } ifc_ifcu; 165 }; 166 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 167 168 struct ifdrv32 { 169 char ifd_name[IFNAMSIZ]; 170 uint32_t ifd_cmd; 171 uint32_t ifd_len; 172 uint32_t ifd_data; 173 }; 174 #define SIOCSDRVSPEC32 _IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32) 175 #define SIOCGDRVSPEC32 _IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32) 176 177 struct ifgroupreq32 { 178 char ifgr_name[IFNAMSIZ]; 179 u_int ifgr_len; 180 union { 181 char ifgru_group[IFNAMSIZ]; 182 uint32_t ifgru_groups; 183 } ifgr_ifgru; 184 }; 185 #define SIOCAIFGROUP32 _IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32) 186 #define SIOCGIFGROUP32 _IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32) 187 #define SIOCDIFGROUP32 _IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32) 188 #define SIOCGIFGMEMB32 _IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32) 189 190 struct ifmediareq32 { 191 char ifm_name[IFNAMSIZ]; 192 int ifm_current; 193 int ifm_mask; 194 int ifm_status; 195 int ifm_active; 196 int ifm_count; 197 uint32_t ifm_ulist; /* (int *) */ 198 }; 199 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 200 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 201 #endif /* COMPAT_FREEBSD32 */ 202 203 union ifreq_union { 204 struct ifreq ifr; 205 #ifdef COMPAT_FREEBSD32 206 struct ifreq32 ifr32; 207 #endif 208 }; 209 210 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 211 "Link layers"); 212 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 213 "Generic link-management"); 214 215 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 216 &ifqmaxlen, 0, "max send queue size"); 217 218 /* Log link state change events */ 219 static int log_link_state_change = 1; 220 221 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 222 &log_link_state_change, 0, 223 "log interface link state change events"); 224 225 /* Log promiscuous mode change events */ 226 static int log_promisc_mode_change = 1; 227 228 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 229 &log_promisc_mode_change, 1, 230 "log promiscuous mode change events"); 231 232 /* Interface description */ 233 static unsigned int ifdescr_maxlen = 1024; 234 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 235 &ifdescr_maxlen, 0, 236 "administrative maximum length for interface description"); 237 238 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 239 240 /* global sx for non-critical path ifdescr */ 241 static struct sx ifdescr_sx; 242 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 243 244 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 245 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 246 /* These are external hooks for CARP. */ 247 void (*carp_linkstate_p)(struct ifnet *ifp); 248 void (*carp_demote_adj_p)(int, char *); 249 int (*carp_master_p)(struct ifaddr *); 250 #if defined(INET) || defined(INET6) 251 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 252 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 253 const struct sockaddr *sa); 254 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 255 int (*carp_attach_p)(struct ifaddr *, int); 256 void (*carp_detach_p)(struct ifaddr *, bool); 257 #endif 258 #ifdef INET 259 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 260 #endif 261 #ifdef INET6 262 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 263 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 264 const struct in6_addr *taddr); 265 #endif 266 267 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 268 269 /* 270 * XXX: Style; these should be sorted alphabetically, and unprototyped 271 * static functions should be prototyped. Currently they are sorted by 272 * declaration order. 273 */ 274 static void if_attachdomain(void *); 275 static void if_attachdomain1(struct ifnet *); 276 static int ifconf(u_long, caddr_t); 277 static void if_input_default(struct ifnet *, struct mbuf *); 278 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 279 static int if_setflag(struct ifnet *, int, int, int *, int); 280 static int if_transmit_default(struct ifnet *ifp, struct mbuf *m); 281 static void if_unroute(struct ifnet *, int flag, int fam); 282 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 283 static void do_link_state_change(void *, int); 284 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 285 static int if_getgroupmembers(struct ifgroupreq *); 286 static void if_delgroups(struct ifnet *); 287 static void if_attach_internal(struct ifnet *, bool); 288 static int if_detach_internal(struct ifnet *, bool); 289 static void if_siocaddmulti(void *, int); 290 static void if_link_ifnet(struct ifnet *); 291 static bool if_unlink_ifnet(struct ifnet *, bool); 292 #ifdef VIMAGE 293 static int if_vmove(struct ifnet *, struct vnet *); 294 #endif 295 296 #ifdef INET6 297 /* 298 * XXX: declare here to avoid to include many inet6 related files.. 299 * should be more generalized? 300 */ 301 extern void nd6_setmtu(struct ifnet *); 302 #endif 303 304 /* ipsec helper hooks */ 305 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 306 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 307 308 int ifqmaxlen = IFQ_MAXLEN; 309 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 310 VNET_DEFINE(struct ifgrouphead, ifg_head); 311 312 /* Table of ifnet by index. */ 313 static int if_index; 314 static int if_indexlim = 8; 315 static struct ifindex_entry { 316 struct ifnet *ife_ifnet; 317 uint16_t ife_gencnt; 318 } *ifindex_table; 319 320 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system, 321 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 322 "Variables global to all interfaces"); 323 static int 324 sysctl_ifcount(SYSCTL_HANDLER_ARGS) 325 { 326 int rv = 0; 327 328 IFNET_RLOCK(); 329 for (int i = 1; i <= if_index; i++) 330 if (ifindex_table[i].ife_ifnet != NULL && 331 ifindex_table[i].ife_ifnet->if_vnet == curvnet) 332 rv = i; 333 IFNET_RUNLOCK(); 334 335 return (sysctl_handle_int(oidp, &rv, 0, req)); 336 } 337 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount, 338 CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I", 339 "Maximum known interface index"); 340 341 /* 342 * The global network interface list (V_ifnet) and related state (such as 343 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 344 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 345 */ 346 struct sx ifnet_sxlock; 347 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 348 349 struct sx ifnet_detach_sxlock; 350 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 351 SX_RECURSE); 352 353 #ifdef VIMAGE 354 #define VNET_IS_SHUTTING_DOWN(_vnet) \ 355 ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE) 356 #endif 357 358 static if_com_alloc_t *if_com_alloc[256]; 359 static if_com_free_t *if_com_free[256]; 360 361 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 362 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 363 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 364 365 struct ifnet * 366 ifnet_byindex(u_int idx) 367 { 368 struct ifnet *ifp; 369 370 NET_EPOCH_ASSERT(); 371 372 if (__predict_false(idx > if_index)) 373 return (NULL); 374 375 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 376 377 if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet) 378 ifp = NULL; 379 380 return (ifp); 381 } 382 383 struct ifnet * 384 ifnet_byindex_ref(u_int idx) 385 { 386 struct ifnet *ifp; 387 388 ifp = ifnet_byindex(idx); 389 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 390 return (NULL); 391 if (!if_try_ref(ifp)) 392 return (NULL); 393 return (ifp); 394 } 395 396 struct ifnet * 397 ifnet_byindexgen(uint16_t idx, uint16_t gen) 398 { 399 struct ifnet *ifp; 400 401 NET_EPOCH_ASSERT(); 402 403 if (__predict_false(idx > if_index)) 404 return (NULL); 405 406 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 407 408 if (ifindex_table[idx].ife_gencnt == gen) 409 return (ifp); 410 else 411 return (NULL); 412 } 413 414 /* 415 * Network interface utility routines. 416 * 417 * Routines with ifa_ifwith* names take sockaddr *'s as 418 * parameters. 419 */ 420 421 static void 422 if_init_idxtable(void *arg __unused) 423 { 424 425 ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table), 426 M_IFNET, M_WAITOK | M_ZERO); 427 } 428 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL); 429 430 static void 431 vnet_if_init(const void *unused __unused) 432 { 433 434 CK_STAILQ_INIT(&V_ifnet); 435 CK_STAILQ_INIT(&V_ifg_head); 436 vnet_if_clone_init(); 437 } 438 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 439 NULL); 440 441 static void 442 if_link_ifnet(struct ifnet *ifp) 443 { 444 445 IFNET_WLOCK(); 446 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 447 #ifdef VIMAGE 448 curvnet->vnet_ifcnt++; 449 #endif 450 IFNET_WUNLOCK(); 451 } 452 453 static bool 454 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 455 { 456 struct ifnet *iter; 457 int found = 0; 458 459 IFNET_WLOCK(); 460 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 461 if (iter == ifp) { 462 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 463 if (!vmove) 464 ifp->if_flags |= IFF_DYING; 465 found = 1; 466 break; 467 } 468 #ifdef VIMAGE 469 curvnet->vnet_ifcnt--; 470 #endif 471 IFNET_WUNLOCK(); 472 473 return (found); 474 } 475 476 #ifdef VIMAGE 477 static void 478 vnet_if_return(const void *unused __unused) 479 { 480 struct ifnet *ifp, *nifp; 481 struct ifnet **pending; 482 int found __diagused; 483 int i; 484 485 i = 0; 486 487 /* 488 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 489 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 490 * if_detach_internal(), which waits for NET_EPOCH callbacks to 491 * complete. We can't do that from within NET_EPOCH. 492 * 493 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 494 * read/write lock. We cannot hold the lock as we call if_vmove() 495 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 496 * ctx lock. 497 */ 498 IFNET_WLOCK(); 499 500 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 501 M_IFNET, M_WAITOK | M_ZERO); 502 503 /* Return all inherited interfaces to their parent vnets. */ 504 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 505 if (ifp->if_home_vnet != ifp->if_vnet) { 506 found = if_unlink_ifnet(ifp, true); 507 MPASS(found); 508 509 pending[i++] = ifp; 510 } 511 } 512 IFNET_WUNLOCK(); 513 514 for (int j = 0; j < i; j++) { 515 sx_xlock(&ifnet_detach_sxlock); 516 if_vmove(pending[j], pending[j]->if_home_vnet); 517 sx_xunlock(&ifnet_detach_sxlock); 518 } 519 520 free(pending, M_IFNET); 521 } 522 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 523 vnet_if_return, NULL); 524 #endif 525 526 /* 527 * Allocate a struct ifnet and an index for an interface. A layer 2 528 * common structure will also be allocated if an allocation routine is 529 * registered for the passed type. 530 */ 531 static struct ifnet * 532 if_alloc_domain(u_char type, int numa_domain) 533 { 534 struct ifnet *ifp; 535 u_short idx; 536 537 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 538 if (numa_domain == IF_NODOM) 539 ifp = malloc(sizeof(struct ifnet), M_IFNET, 540 M_WAITOK | M_ZERO); 541 else 542 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 543 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 544 ifp->if_type = type; 545 ifp->if_alloctype = type; 546 ifp->if_numa_domain = numa_domain; 547 #ifdef VIMAGE 548 ifp->if_vnet = curvnet; 549 #endif 550 if (if_com_alloc[type] != NULL) { 551 ifp->if_l2com = if_com_alloc[type](type, ifp); 552 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 553 type)); 554 } 555 556 IF_ADDR_LOCK_INIT(ifp); 557 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 558 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 559 ifp->if_afdata_initialized = 0; 560 IF_AFDATA_LOCK_INIT(ifp); 561 CK_STAILQ_INIT(&ifp->if_addrhead); 562 CK_STAILQ_INIT(&ifp->if_multiaddrs); 563 CK_STAILQ_INIT(&ifp->if_groups); 564 #ifdef MAC 565 mac_ifnet_init(ifp); 566 #endif 567 ifq_init(&ifp->if_snd, ifp); 568 569 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 570 for (int i = 0; i < IFCOUNTERS; i++) 571 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 572 ifp->if_get_counter = if_get_counter_default; 573 ifp->if_pcp = IFNET_PCP_NONE; 574 575 /* Allocate an ifindex array entry. */ 576 IFNET_WLOCK(); 577 /* 578 * Try to find an empty slot below if_index. If we fail, take the 579 * next slot. 580 */ 581 for (idx = 1; idx <= if_index; idx++) { 582 if (ifindex_table[idx].ife_ifnet == NULL) 583 break; 584 } 585 586 /* Catch if_index overflow. */ 587 if (idx >= if_indexlim) { 588 struct ifindex_entry *new, *old; 589 int newlim; 590 591 newlim = if_indexlim * 2; 592 new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO); 593 memcpy(new, ifindex_table, if_indexlim * sizeof(*new)); 594 old = ifindex_table; 595 ck_pr_store_ptr(&ifindex_table, new); 596 if_indexlim = newlim; 597 NET_EPOCH_WAIT(); 598 free(old, M_IFNET); 599 } 600 if (idx > if_index) 601 if_index = idx; 602 603 ifp->if_index = idx; 604 ifp->if_idxgen = ifindex_table[idx].ife_gencnt; 605 ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp); 606 IFNET_WUNLOCK(); 607 608 return (ifp); 609 } 610 611 struct ifnet * 612 if_alloc_dev(u_char type, device_t dev) 613 { 614 int numa_domain; 615 616 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 617 return (if_alloc_domain(type, IF_NODOM)); 618 return (if_alloc_domain(type, numa_domain)); 619 } 620 621 struct ifnet * 622 if_alloc(u_char type) 623 { 624 625 return (if_alloc_domain(type, IF_NODOM)); 626 } 627 /* 628 * Do the actual work of freeing a struct ifnet, and layer 2 common 629 * structure. This call is made when the network epoch guarantees 630 * us that nobody holds a pointer to the interface. 631 */ 632 static void 633 if_free_deferred(epoch_context_t ctx) 634 { 635 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 636 637 KASSERT((ifp->if_flags & IFF_DYING), 638 ("%s: interface not dying", __func__)); 639 640 if (if_com_free[ifp->if_alloctype] != NULL) 641 if_com_free[ifp->if_alloctype](ifp->if_l2com, 642 ifp->if_alloctype); 643 644 #ifdef MAC 645 mac_ifnet_destroy(ifp); 646 #endif /* MAC */ 647 IF_AFDATA_DESTROY(ifp); 648 IF_ADDR_LOCK_DESTROY(ifp); 649 ifq_delete(&ifp->if_snd); 650 651 for (int i = 0; i < IFCOUNTERS; i++) 652 counter_u64_free(ifp->if_counters[i]); 653 654 if_freedescr(ifp->if_description); 655 free(ifp->if_hw_addr, M_IFADDR); 656 free(ifp, M_IFNET); 657 } 658 659 /* 660 * Deregister an interface and free the associated storage. 661 */ 662 void 663 if_free(struct ifnet *ifp) 664 { 665 666 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 667 668 /* 669 * XXXGL: An interface index is really an alias to ifp pointer. 670 * Why would we clear the alias now, and not in the deferred 671 * context? Indeed there is nothing wrong with some network 672 * thread obtaining ifp via ifnet_byindex() inside the network 673 * epoch and then dereferencing ifp while we perform if_free(), 674 * and after if_free() finished, too. 675 * 676 * This early index freeing was important back when ifindex was 677 * virtualized and interface would outlive the vnet. 678 */ 679 IFNET_WLOCK(); 680 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 681 ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL); 682 ifindex_table[ifp->if_index].ife_gencnt++; 683 while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL) 684 if_index--; 685 IFNET_WUNLOCK(); 686 687 if (refcount_release(&ifp->if_refcount)) 688 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 689 } 690 691 /* 692 * Interfaces to keep an ifnet type-stable despite the possibility of the 693 * driver calling if_free(). If there are additional references, we defer 694 * freeing the underlying data structure. 695 */ 696 void 697 if_ref(struct ifnet *ifp) 698 { 699 u_int old __diagused; 700 701 /* We don't assert the ifnet list lock here, but arguably should. */ 702 old = refcount_acquire(&ifp->if_refcount); 703 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 704 } 705 706 bool 707 if_try_ref(struct ifnet *ifp) 708 { 709 NET_EPOCH_ASSERT(); 710 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 711 } 712 713 void 714 if_rele(struct ifnet *ifp) 715 { 716 717 if (!refcount_release(&ifp->if_refcount)) 718 return; 719 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 720 } 721 722 void 723 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 724 { 725 726 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 727 728 if (ifq->ifq_maxlen == 0) 729 ifq->ifq_maxlen = ifqmaxlen; 730 731 ifq->altq_type = 0; 732 ifq->altq_disc = NULL; 733 ifq->altq_flags &= ALTQF_CANTCHANGE; 734 ifq->altq_tbr = NULL; 735 ifq->altq_ifp = ifp; 736 } 737 738 void 739 ifq_delete(struct ifaltq *ifq) 740 { 741 mtx_destroy(&ifq->ifq_mtx); 742 } 743 744 /* 745 * Perform generic interface initialization tasks and attach the interface 746 * to the list of "active" interfaces. If vmove flag is set on entry 747 * to if_attach_internal(), perform only a limited subset of initialization 748 * tasks, given that we are moving from one vnet to another an ifnet which 749 * has already been fully initialized. 750 * 751 * Note that if_detach_internal() removes group membership unconditionally 752 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 753 * Thus, when if_vmove() is applied to a cloned interface, group membership 754 * is lost while a cloned one always joins a group whose name is 755 * ifc->ifc_name. To recover this after if_detach_internal() and 756 * if_attach_internal(), the cloner should be specified to 757 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 758 * attempts to join a group whose name is ifc->ifc_name. 759 * 760 * XXX: 761 * - The decision to return void and thus require this function to 762 * succeed is questionable. 763 * - We should probably do more sanity checking. For instance we don't 764 * do anything to insure if_xname is unique or non-empty. 765 */ 766 void 767 if_attach(struct ifnet *ifp) 768 { 769 770 if_attach_internal(ifp, false); 771 } 772 773 /* 774 * Compute the least common TSO limit. 775 */ 776 void 777 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 778 { 779 /* 780 * 1) If there is no limit currently, take the limit from 781 * the network adapter. 782 * 783 * 2) If the network adapter has a limit below the current 784 * limit, apply it. 785 */ 786 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 787 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 788 pmax->tsomaxbytes = ifp->if_hw_tsomax; 789 } 790 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 791 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 792 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 793 } 794 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 795 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 796 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 797 } 798 } 799 800 /* 801 * Update TSO limit of a network adapter. 802 * 803 * Returns zero if no change. Else non-zero. 804 */ 805 int 806 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 807 { 808 int retval = 0; 809 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 810 ifp->if_hw_tsomax = pmax->tsomaxbytes; 811 retval++; 812 } 813 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 814 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 815 retval++; 816 } 817 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 818 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 819 retval++; 820 } 821 return (retval); 822 } 823 824 static void 825 if_attach_internal(struct ifnet *ifp, bool vmove) 826 { 827 unsigned socksize, ifasize; 828 int namelen, masklen; 829 struct sockaddr_dl *sdl; 830 struct ifaddr *ifa; 831 832 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 833 834 #ifdef VIMAGE 835 CURVNET_ASSERT_SET(); 836 ifp->if_vnet = curvnet; 837 if (ifp->if_home_vnet == NULL) 838 ifp->if_home_vnet = curvnet; 839 #endif 840 841 if_addgroup(ifp, IFG_ALL); 842 843 #ifdef VIMAGE 844 /* Restore group membership for cloned interface. */ 845 if (vmove) 846 if_clone_restoregroup(ifp); 847 #endif 848 849 getmicrotime(&ifp->if_lastchange); 850 ifp->if_epoch = time_uptime; 851 852 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 853 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 854 ("transmit and qflush must both either be set or both be NULL")); 855 if (ifp->if_transmit == NULL) { 856 ifp->if_transmit = if_transmit_default; 857 ifp->if_qflush = if_qflush; 858 } 859 if (ifp->if_input == NULL) 860 ifp->if_input = if_input_default; 861 862 if (ifp->if_requestencap == NULL) 863 ifp->if_requestencap = if_requestencap_default; 864 865 if (!vmove) { 866 #ifdef MAC 867 mac_ifnet_create(ifp); 868 #endif 869 870 /* 871 * Create a Link Level name for this device. 872 */ 873 namelen = strlen(ifp->if_xname); 874 /* 875 * Always save enough space for any possiable name so we 876 * can do a rename in place later. 877 */ 878 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 879 socksize = masklen + ifp->if_addrlen; 880 if (socksize < sizeof(*sdl)) 881 socksize = sizeof(*sdl); 882 socksize = roundup2(socksize, sizeof(long)); 883 ifasize = sizeof(*ifa) + 2 * socksize; 884 ifa = ifa_alloc(ifasize, M_WAITOK); 885 sdl = (struct sockaddr_dl *)(ifa + 1); 886 sdl->sdl_len = socksize; 887 sdl->sdl_family = AF_LINK; 888 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 889 sdl->sdl_nlen = namelen; 890 sdl->sdl_index = ifp->if_index; 891 sdl->sdl_type = ifp->if_type; 892 ifp->if_addr = ifa; 893 ifa->ifa_ifp = ifp; 894 ifa->ifa_addr = (struct sockaddr *)sdl; 895 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 896 ifa->ifa_netmask = (struct sockaddr *)sdl; 897 sdl->sdl_len = masklen; 898 while (namelen != 0) 899 sdl->sdl_data[--namelen] = 0xff; 900 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 901 /* Reliably crash if used uninitialized. */ 902 ifp->if_broadcastaddr = NULL; 903 904 if (ifp->if_type == IFT_ETHER) { 905 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 906 M_WAITOK | M_ZERO); 907 } 908 909 #if defined(INET) || defined(INET6) 910 /* Use defaults for TSO, if nothing is set */ 911 if (ifp->if_hw_tsomax == 0 && 912 ifp->if_hw_tsomaxsegcount == 0 && 913 ifp->if_hw_tsomaxsegsize == 0) { 914 /* 915 * The TSO defaults needs to be such that an 916 * NFS mbuf list of 35 mbufs totalling just 917 * below 64K works and that a chain of mbufs 918 * can be defragged into at most 32 segments: 919 */ 920 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 921 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 922 ifp->if_hw_tsomaxsegcount = 35; 923 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 924 925 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 926 if (ifp->if_capabilities & IFCAP_TSO) { 927 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 928 ifp->if_hw_tsomax, 929 ifp->if_hw_tsomaxsegcount, 930 ifp->if_hw_tsomaxsegsize); 931 } 932 } 933 #endif 934 } 935 #ifdef VIMAGE 936 else { 937 /* 938 * Update the interface index in the link layer address 939 * of the interface. 940 */ 941 for (ifa = ifp->if_addr; ifa != NULL; 942 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 943 if (ifa->ifa_addr->sa_family == AF_LINK) { 944 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 945 sdl->sdl_index = ifp->if_index; 946 } 947 } 948 } 949 #endif 950 951 if_link_ifnet(ifp); 952 953 if (domain_init_status >= 2) 954 if_attachdomain1(ifp); 955 956 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 957 if (IS_DEFAULT_VNET(curvnet)) 958 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 959 } 960 961 static void 962 if_epochalloc(void *dummy __unused) 963 { 964 965 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 966 } 967 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 968 969 static void 970 if_attachdomain(void *dummy) 971 { 972 struct ifnet *ifp; 973 974 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 975 if_attachdomain1(ifp); 976 } 977 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 978 if_attachdomain, NULL); 979 980 static void 981 if_attachdomain1(struct ifnet *ifp) 982 { 983 struct domain *dp; 984 985 /* 986 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 987 * cannot lock ifp->if_afdata initialization, entirely. 988 */ 989 IF_AFDATA_LOCK(ifp); 990 if (ifp->if_afdata_initialized >= domain_init_status) { 991 IF_AFDATA_UNLOCK(ifp); 992 log(LOG_WARNING, "%s called more than once on %s\n", 993 __func__, ifp->if_xname); 994 return; 995 } 996 ifp->if_afdata_initialized = domain_init_status; 997 IF_AFDATA_UNLOCK(ifp); 998 999 /* address family dependent data region */ 1000 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 1001 SLIST_FOREACH(dp, &domains, dom_next) { 1002 if (dp->dom_ifattach) 1003 ifp->if_afdata[dp->dom_family] = 1004 (*dp->dom_ifattach)(ifp); 1005 } 1006 } 1007 1008 /* 1009 * Remove any unicast or broadcast network addresses from an interface. 1010 */ 1011 void 1012 if_purgeaddrs(struct ifnet *ifp) 1013 { 1014 struct ifaddr *ifa; 1015 1016 #ifdef INET6 1017 /* 1018 * Need to leave multicast addresses of proxy NDP llentries 1019 * before in6_purgeifaddr() because the llentries are keys 1020 * for in6_multi objects of proxy NDP entries. 1021 * in6_purgeifaddr()s clean up llentries including proxy NDPs 1022 * then we would lose the keys if they are called earlier. 1023 */ 1024 in6_purge_proxy_ndp(ifp); 1025 #endif 1026 while (1) { 1027 struct epoch_tracker et; 1028 1029 NET_EPOCH_ENTER(et); 1030 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1031 if (ifa->ifa_addr->sa_family != AF_LINK) 1032 break; 1033 } 1034 NET_EPOCH_EXIT(et); 1035 1036 if (ifa == NULL) 1037 break; 1038 #ifdef INET 1039 /* XXX: Ugly!! ad hoc just for INET */ 1040 if (ifa->ifa_addr->sa_family == AF_INET) { 1041 struct ifreq ifr; 1042 1043 bzero(&ifr, sizeof(ifr)); 1044 ifr.ifr_addr = *ifa->ifa_addr; 1045 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1046 NULL) == 0) 1047 continue; 1048 } 1049 #endif /* INET */ 1050 #ifdef INET6 1051 if (ifa->ifa_addr->sa_family == AF_INET6) { 1052 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1053 /* ifp_addrhead is already updated */ 1054 continue; 1055 } 1056 #endif /* INET6 */ 1057 IF_ADDR_WLOCK(ifp); 1058 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1059 IF_ADDR_WUNLOCK(ifp); 1060 ifa_free(ifa); 1061 } 1062 } 1063 1064 /* 1065 * Remove any multicast network addresses from an interface when an ifnet 1066 * is going away. 1067 */ 1068 static void 1069 if_purgemaddrs(struct ifnet *ifp) 1070 { 1071 struct ifmultiaddr *ifma; 1072 1073 IF_ADDR_WLOCK(ifp); 1074 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1075 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1076 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1077 if_delmulti_locked(ifp, ifma, 1); 1078 } 1079 IF_ADDR_WUNLOCK(ifp); 1080 } 1081 1082 /* 1083 * Detach an interface, removing it from the list of "active" interfaces. 1084 * If vmove flag is set on entry to if_detach_internal(), perform only a 1085 * limited subset of cleanup tasks, given that we are moving an ifnet from 1086 * one vnet to another, where it must be fully operational. 1087 * 1088 * XXXRW: There are some significant questions about event ordering, and 1089 * how to prevent things from starting to use the interface during detach. 1090 */ 1091 void 1092 if_detach(struct ifnet *ifp) 1093 { 1094 bool found; 1095 1096 CURVNET_SET_QUIET(ifp->if_vnet); 1097 found = if_unlink_ifnet(ifp, false); 1098 if (found) { 1099 sx_xlock(&ifnet_detach_sxlock); 1100 if_detach_internal(ifp, false); 1101 sx_xunlock(&ifnet_detach_sxlock); 1102 } 1103 CURVNET_RESTORE(); 1104 } 1105 1106 /* 1107 * The vmove flag, if set, indicates that we are called from a callpath 1108 * that is moving an interface to a different vnet instance. 1109 * 1110 * The shutdown flag, if set, indicates that we are called in the 1111 * process of shutting down a vnet instance. Currently only the 1112 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1113 * on a vnet instance shutdown without this flag being set, e.g., when 1114 * the cloned interfaces are destoyed as first thing of teardown. 1115 */ 1116 static int 1117 if_detach_internal(struct ifnet *ifp, bool vmove) 1118 { 1119 struct ifaddr *ifa; 1120 int i; 1121 struct domain *dp; 1122 #ifdef VIMAGE 1123 bool shutdown; 1124 1125 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1126 #endif 1127 1128 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 1129 1130 /* 1131 * At this point we know the interface still was on the ifnet list 1132 * and we removed it so we are in a stable state. 1133 */ 1134 NET_EPOCH_WAIT(); 1135 1136 /* 1137 * Ensure all pending EPOCH(9) callbacks have been executed. This 1138 * fixes issues about late destruction of multicast options 1139 * which lead to leave group calls, which in turn access the 1140 * belonging ifnet structure: 1141 */ 1142 NET_EPOCH_DRAIN_CALLBACKS(); 1143 1144 /* 1145 * In any case (destroy or vmove) detach us from the groups 1146 * and remove/wait for pending events on the taskq. 1147 * XXX-BZ in theory an interface could still enqueue a taskq change? 1148 */ 1149 if_delgroups(ifp); 1150 1151 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1152 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1153 1154 if_down(ifp); 1155 1156 #ifdef VIMAGE 1157 /* 1158 * On VNET shutdown abort here as the stack teardown will do all 1159 * the work top-down for us. 1160 */ 1161 if (shutdown) { 1162 /* Give interface users the chance to clean up. */ 1163 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1164 1165 /* 1166 * In case of a vmove we are done here without error. 1167 * If we would signal an error it would lead to the same 1168 * abort as if we did not find the ifnet anymore. 1169 * if_detach() calls us in void context and does not care 1170 * about an early abort notification, so life is splendid :) 1171 */ 1172 goto finish_vnet_shutdown; 1173 } 1174 #endif 1175 1176 /* 1177 * At this point we are not tearing down a VNET and are either 1178 * going to destroy or vmove the interface and have to cleanup 1179 * accordingly. 1180 */ 1181 1182 /* 1183 * Remove routes and flush queues. 1184 */ 1185 #ifdef ALTQ 1186 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1187 altq_disable(&ifp->if_snd); 1188 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1189 altq_detach(&ifp->if_snd); 1190 #endif 1191 1192 if_purgeaddrs(ifp); 1193 1194 #ifdef INET 1195 in_ifdetach(ifp); 1196 #endif 1197 1198 #ifdef INET6 1199 /* 1200 * Remove all IPv6 kernel structs related to ifp. This should be done 1201 * before removing routing entries below, since IPv6 interface direct 1202 * routes are expected to be removed by the IPv6-specific kernel API. 1203 * Otherwise, the kernel will detect some inconsistency and bark it. 1204 */ 1205 in6_ifdetach(ifp); 1206 #endif 1207 if_purgemaddrs(ifp); 1208 1209 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1210 if (IS_DEFAULT_VNET(curvnet)) 1211 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1212 1213 if (!vmove) { 1214 /* 1215 * Prevent further calls into the device driver via ifnet. 1216 */ 1217 if_dead(ifp); 1218 1219 /* 1220 * Clean up all addresses. 1221 */ 1222 IF_ADDR_WLOCK(ifp); 1223 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1224 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1225 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1226 IF_ADDR_WUNLOCK(ifp); 1227 ifa_free(ifa); 1228 } else 1229 IF_ADDR_WUNLOCK(ifp); 1230 } 1231 1232 rt_flushifroutes(ifp); 1233 1234 #ifdef VIMAGE 1235 finish_vnet_shutdown: 1236 #endif 1237 /* 1238 * We cannot hold the lock over dom_ifdetach calls as they might 1239 * sleep, for example trying to drain a callout, thus open up the 1240 * theoretical race with re-attaching. 1241 */ 1242 IF_AFDATA_LOCK(ifp); 1243 i = ifp->if_afdata_initialized; 1244 ifp->if_afdata_initialized = 0; 1245 IF_AFDATA_UNLOCK(ifp); 1246 if (i == 0) 1247 return (0); 1248 SLIST_FOREACH(dp, &domains, dom_next) { 1249 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1250 (*dp->dom_ifdetach)(ifp, 1251 ifp->if_afdata[dp->dom_family]); 1252 ifp->if_afdata[dp->dom_family] = NULL; 1253 } 1254 } 1255 1256 return (0); 1257 } 1258 1259 #ifdef VIMAGE 1260 /* 1261 * if_vmove() performs a limited version of if_detach() in current 1262 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1263 */ 1264 static int 1265 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1266 { 1267 int rc; 1268 1269 /* 1270 * Detach from current vnet, but preserve LLADDR info, do not 1271 * mark as dead etc. so that the ifnet can be reattached later. 1272 * If we cannot find it, we lost the race to someone else. 1273 */ 1274 rc = if_detach_internal(ifp, true); 1275 if (rc != 0) 1276 return (rc); 1277 1278 /* 1279 * Perform interface-specific reassignment tasks, if provided by 1280 * the driver. 1281 */ 1282 if (ifp->if_reassign != NULL) 1283 ifp->if_reassign(ifp, new_vnet, NULL); 1284 1285 /* 1286 * Switch to the context of the target vnet. 1287 */ 1288 CURVNET_SET_QUIET(new_vnet); 1289 if_attach_internal(ifp, true); 1290 CURVNET_RESTORE(); 1291 return (0); 1292 } 1293 1294 /* 1295 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1296 */ 1297 static int 1298 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1299 { 1300 struct prison *pr; 1301 struct ifnet *difp; 1302 int error; 1303 bool found __diagused; 1304 bool shutdown; 1305 1306 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 1307 1308 /* Try to find the prison within our visibility. */ 1309 sx_slock(&allprison_lock); 1310 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1311 sx_sunlock(&allprison_lock); 1312 if (pr == NULL) 1313 return (ENXIO); 1314 prison_hold_locked(pr); 1315 mtx_unlock(&pr->pr_mtx); 1316 1317 /* Do not try to move the iface from and to the same prison. */ 1318 if (pr->pr_vnet == ifp->if_vnet) { 1319 prison_free(pr); 1320 return (EEXIST); 1321 } 1322 1323 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1324 /* XXX Lock interfaces to avoid races. */ 1325 CURVNET_SET_QUIET(pr->pr_vnet); 1326 difp = ifunit(ifname); 1327 CURVNET_RESTORE(); 1328 if (difp != NULL) { 1329 prison_free(pr); 1330 return (EEXIST); 1331 } 1332 sx_xlock(&ifnet_detach_sxlock); 1333 1334 /* Make sure the VNET is stable. */ 1335 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1336 if (shutdown) { 1337 sx_xunlock(&ifnet_detach_sxlock); 1338 prison_free(pr); 1339 return (EBUSY); 1340 } 1341 1342 found = if_unlink_ifnet(ifp, true); 1343 if (! found) { 1344 sx_xunlock(&ifnet_detach_sxlock); 1345 prison_free(pr); 1346 return (ENODEV); 1347 } 1348 1349 /* Move the interface into the child jail/vnet. */ 1350 error = if_vmove(ifp, pr->pr_vnet); 1351 1352 /* Report the new if_xname back to the userland on success. */ 1353 if (error == 0) 1354 sprintf(ifname, "%s", ifp->if_xname); 1355 1356 sx_xunlock(&ifnet_detach_sxlock); 1357 1358 prison_free(pr); 1359 return (error); 1360 } 1361 1362 static int 1363 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1364 { 1365 struct prison *pr; 1366 struct vnet *vnet_dst; 1367 struct ifnet *ifp; 1368 int error, found __diagused; 1369 bool shutdown; 1370 1371 /* Try to find the prison within our visibility. */ 1372 sx_slock(&allprison_lock); 1373 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1374 sx_sunlock(&allprison_lock); 1375 if (pr == NULL) 1376 return (ENXIO); 1377 prison_hold_locked(pr); 1378 mtx_unlock(&pr->pr_mtx); 1379 1380 /* Make sure the named iface exists in the source prison/vnet. */ 1381 CURVNET_SET(pr->pr_vnet); 1382 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1383 if (ifp == NULL) { 1384 CURVNET_RESTORE(); 1385 prison_free(pr); 1386 return (ENXIO); 1387 } 1388 1389 /* Do not try to move the iface from and to the same prison. */ 1390 vnet_dst = TD_TO_VNET(td); 1391 if (vnet_dst == ifp->if_vnet) { 1392 CURVNET_RESTORE(); 1393 prison_free(pr); 1394 return (EEXIST); 1395 } 1396 1397 /* Make sure the VNET is stable. */ 1398 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1399 if (shutdown) { 1400 CURVNET_RESTORE(); 1401 prison_free(pr); 1402 return (EBUSY); 1403 } 1404 1405 /* Get interface back from child jail/vnet. */ 1406 found = if_unlink_ifnet(ifp, true); 1407 MPASS(found); 1408 sx_xlock(&ifnet_detach_sxlock); 1409 error = if_vmove(ifp, vnet_dst); 1410 sx_xunlock(&ifnet_detach_sxlock); 1411 CURVNET_RESTORE(); 1412 1413 /* Report the new if_xname back to the userland on success. */ 1414 if (error == 0) 1415 sprintf(ifname, "%s", ifp->if_xname); 1416 1417 prison_free(pr); 1418 return (error); 1419 } 1420 #endif /* VIMAGE */ 1421 1422 /* 1423 * Add a group to an interface 1424 */ 1425 int 1426 if_addgroup(struct ifnet *ifp, const char *groupname) 1427 { 1428 struct ifg_list *ifgl; 1429 struct ifg_group *ifg = NULL; 1430 struct ifg_member *ifgm; 1431 int new = 0; 1432 1433 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1434 groupname[strlen(groupname) - 1] <= '9') 1435 return (EINVAL); 1436 1437 IFNET_WLOCK(); 1438 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1439 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1440 IFNET_WUNLOCK(); 1441 return (EEXIST); 1442 } 1443 1444 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1445 IFNET_WUNLOCK(); 1446 return (ENOMEM); 1447 } 1448 1449 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1450 free(ifgl, M_TEMP); 1451 IFNET_WUNLOCK(); 1452 return (ENOMEM); 1453 } 1454 1455 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1456 if (!strcmp(ifg->ifg_group, groupname)) 1457 break; 1458 1459 if (ifg == NULL) { 1460 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1461 free(ifgl, M_TEMP); 1462 free(ifgm, M_TEMP); 1463 IFNET_WUNLOCK(); 1464 return (ENOMEM); 1465 } 1466 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1467 ifg->ifg_refcnt = 0; 1468 CK_STAILQ_INIT(&ifg->ifg_members); 1469 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1470 new = 1; 1471 } 1472 1473 ifg->ifg_refcnt++; 1474 ifgl->ifgl_group = ifg; 1475 ifgm->ifgm_ifp = ifp; 1476 1477 IF_ADDR_WLOCK(ifp); 1478 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1479 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1480 IF_ADDR_WUNLOCK(ifp); 1481 1482 IFNET_WUNLOCK(); 1483 1484 if (new) 1485 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1486 EVENTHANDLER_INVOKE(group_change_event, groupname); 1487 1488 return (0); 1489 } 1490 1491 /* 1492 * Helper function to remove a group out of an interface. Expects the global 1493 * ifnet lock to be write-locked, and drops it before returning. 1494 */ 1495 static void 1496 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1497 const char *groupname) 1498 { 1499 struct ifg_member *ifgm; 1500 bool freeifgl; 1501 1502 IFNET_WLOCK_ASSERT(); 1503 1504 IF_ADDR_WLOCK(ifp); 1505 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1506 IF_ADDR_WUNLOCK(ifp); 1507 1508 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1509 if (ifgm->ifgm_ifp == ifp) { 1510 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1511 ifg_member, ifgm_next); 1512 break; 1513 } 1514 } 1515 1516 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1517 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1518 ifg_next); 1519 freeifgl = true; 1520 } else { 1521 freeifgl = false; 1522 } 1523 IFNET_WUNLOCK(); 1524 1525 NET_EPOCH_WAIT(); 1526 EVENTHANDLER_INVOKE(group_change_event, groupname); 1527 if (freeifgl) { 1528 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1529 free(ifgl->ifgl_group, M_TEMP); 1530 } 1531 free(ifgm, M_TEMP); 1532 free(ifgl, M_TEMP); 1533 } 1534 1535 /* 1536 * Remove a group from an interface 1537 */ 1538 int 1539 if_delgroup(struct ifnet *ifp, const char *groupname) 1540 { 1541 struct ifg_list *ifgl; 1542 1543 IFNET_WLOCK(); 1544 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1545 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1546 break; 1547 if (ifgl == NULL) { 1548 IFNET_WUNLOCK(); 1549 return (ENOENT); 1550 } 1551 1552 _if_delgroup_locked(ifp, ifgl, groupname); 1553 1554 return (0); 1555 } 1556 1557 /* 1558 * Remove an interface from all groups 1559 */ 1560 static void 1561 if_delgroups(struct ifnet *ifp) 1562 { 1563 struct ifg_list *ifgl; 1564 char groupname[IFNAMSIZ]; 1565 1566 IFNET_WLOCK(); 1567 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1568 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1569 _if_delgroup_locked(ifp, ifgl, groupname); 1570 IFNET_WLOCK(); 1571 } 1572 IFNET_WUNLOCK(); 1573 } 1574 1575 /* 1576 * Stores all groups from an interface in memory pointed to by ifgr. 1577 */ 1578 static int 1579 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1580 { 1581 int len, error; 1582 struct ifg_list *ifgl; 1583 struct ifg_req ifgrq, *ifgp; 1584 1585 NET_EPOCH_ASSERT(); 1586 1587 if (ifgr->ifgr_len == 0) { 1588 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1589 ifgr->ifgr_len += sizeof(struct ifg_req); 1590 return (0); 1591 } 1592 1593 len = ifgr->ifgr_len; 1594 ifgp = ifgr->ifgr_groups; 1595 /* XXX: wire */ 1596 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1597 if (len < sizeof(ifgrq)) 1598 return (EINVAL); 1599 bzero(&ifgrq, sizeof ifgrq); 1600 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1601 sizeof(ifgrq.ifgrq_group)); 1602 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1603 return (error); 1604 len -= sizeof(ifgrq); 1605 ifgp++; 1606 } 1607 1608 return (0); 1609 } 1610 1611 /* 1612 * Stores all members of a group in memory pointed to by igfr 1613 */ 1614 static int 1615 if_getgroupmembers(struct ifgroupreq *ifgr) 1616 { 1617 struct ifg_group *ifg; 1618 struct ifg_member *ifgm; 1619 struct ifg_req ifgrq, *ifgp; 1620 int len, error; 1621 1622 IFNET_RLOCK(); 1623 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1624 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1625 break; 1626 if (ifg == NULL) { 1627 IFNET_RUNLOCK(); 1628 return (ENOENT); 1629 } 1630 1631 if (ifgr->ifgr_len == 0) { 1632 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1633 ifgr->ifgr_len += sizeof(ifgrq); 1634 IFNET_RUNLOCK(); 1635 return (0); 1636 } 1637 1638 len = ifgr->ifgr_len; 1639 ifgp = ifgr->ifgr_groups; 1640 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1641 if (len < sizeof(ifgrq)) { 1642 IFNET_RUNLOCK(); 1643 return (EINVAL); 1644 } 1645 bzero(&ifgrq, sizeof ifgrq); 1646 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1647 sizeof(ifgrq.ifgrq_member)); 1648 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1649 IFNET_RUNLOCK(); 1650 return (error); 1651 } 1652 len -= sizeof(ifgrq); 1653 ifgp++; 1654 } 1655 IFNET_RUNLOCK(); 1656 1657 return (0); 1658 } 1659 1660 /* 1661 * Return counter values from counter(9)s stored in ifnet. 1662 */ 1663 uint64_t 1664 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1665 { 1666 1667 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1668 1669 return (counter_u64_fetch(ifp->if_counters[cnt])); 1670 } 1671 1672 /* 1673 * Increase an ifnet counter. Usually used for counters shared 1674 * between the stack and a driver, but function supports them all. 1675 */ 1676 void 1677 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1678 { 1679 1680 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1681 1682 counter_u64_add(ifp->if_counters[cnt], inc); 1683 } 1684 1685 /* 1686 * Copy data from ifnet to userland API structure if_data. 1687 */ 1688 void 1689 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1690 { 1691 1692 ifd->ifi_type = ifp->if_type; 1693 ifd->ifi_physical = 0; 1694 ifd->ifi_addrlen = ifp->if_addrlen; 1695 ifd->ifi_hdrlen = ifp->if_hdrlen; 1696 ifd->ifi_link_state = ifp->if_link_state; 1697 ifd->ifi_vhid = 0; 1698 ifd->ifi_datalen = sizeof(struct if_data); 1699 ifd->ifi_mtu = ifp->if_mtu; 1700 ifd->ifi_metric = ifp->if_metric; 1701 ifd->ifi_baudrate = ifp->if_baudrate; 1702 ifd->ifi_hwassist = ifp->if_hwassist; 1703 ifd->ifi_epoch = ifp->if_epoch; 1704 ifd->ifi_lastchange = ifp->if_lastchange; 1705 1706 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1707 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1708 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1709 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1710 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1711 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1712 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1713 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1714 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1715 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1716 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1717 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1718 } 1719 1720 /* 1721 * Initialization, destruction and refcounting functions for ifaddrs. 1722 */ 1723 struct ifaddr * 1724 ifa_alloc(size_t size, int flags) 1725 { 1726 struct ifaddr *ifa; 1727 1728 KASSERT(size >= sizeof(struct ifaddr), 1729 ("%s: invalid size %zu", __func__, size)); 1730 1731 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1732 if (ifa == NULL) 1733 return (NULL); 1734 1735 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1736 goto fail; 1737 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1738 goto fail; 1739 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1740 goto fail; 1741 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1742 goto fail; 1743 1744 refcount_init(&ifa->ifa_refcnt, 1); 1745 1746 return (ifa); 1747 1748 fail: 1749 /* free(NULL) is okay */ 1750 counter_u64_free(ifa->ifa_opackets); 1751 counter_u64_free(ifa->ifa_ipackets); 1752 counter_u64_free(ifa->ifa_obytes); 1753 counter_u64_free(ifa->ifa_ibytes); 1754 free(ifa, M_IFADDR); 1755 1756 return (NULL); 1757 } 1758 1759 void 1760 ifa_ref(struct ifaddr *ifa) 1761 { 1762 u_int old __diagused; 1763 1764 old = refcount_acquire(&ifa->ifa_refcnt); 1765 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1766 } 1767 1768 int 1769 ifa_try_ref(struct ifaddr *ifa) 1770 { 1771 1772 NET_EPOCH_ASSERT(); 1773 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1774 } 1775 1776 static void 1777 ifa_destroy(epoch_context_t ctx) 1778 { 1779 struct ifaddr *ifa; 1780 1781 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1782 counter_u64_free(ifa->ifa_opackets); 1783 counter_u64_free(ifa->ifa_ipackets); 1784 counter_u64_free(ifa->ifa_obytes); 1785 counter_u64_free(ifa->ifa_ibytes); 1786 free(ifa, M_IFADDR); 1787 } 1788 1789 void 1790 ifa_free(struct ifaddr *ifa) 1791 { 1792 1793 if (refcount_release(&ifa->ifa_refcnt)) 1794 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1795 } 1796 1797 /* 1798 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1799 * structs used to represent other address families, it is necessary 1800 * to perform a different comparison. 1801 */ 1802 1803 #define sa_dl_equal(a1, a2) \ 1804 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1805 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1806 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1807 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1808 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1809 1810 /* 1811 * Locate an interface based on a complete address. 1812 */ 1813 /*ARGSUSED*/ 1814 struct ifaddr * 1815 ifa_ifwithaddr(const struct sockaddr *addr) 1816 { 1817 struct ifnet *ifp; 1818 struct ifaddr *ifa; 1819 1820 NET_EPOCH_ASSERT(); 1821 1822 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1823 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1824 if (ifa->ifa_addr->sa_family != addr->sa_family) 1825 continue; 1826 if (sa_equal(addr, ifa->ifa_addr)) { 1827 goto done; 1828 } 1829 /* IP6 doesn't have broadcast */ 1830 if ((ifp->if_flags & IFF_BROADCAST) && 1831 ifa->ifa_broadaddr && 1832 ifa->ifa_broadaddr->sa_len != 0 && 1833 sa_equal(ifa->ifa_broadaddr, addr)) { 1834 goto done; 1835 } 1836 } 1837 } 1838 ifa = NULL; 1839 done: 1840 return (ifa); 1841 } 1842 1843 int 1844 ifa_ifwithaddr_check(const struct sockaddr *addr) 1845 { 1846 struct epoch_tracker et; 1847 int rc; 1848 1849 NET_EPOCH_ENTER(et); 1850 rc = (ifa_ifwithaddr(addr) != NULL); 1851 NET_EPOCH_EXIT(et); 1852 return (rc); 1853 } 1854 1855 /* 1856 * Locate an interface based on the broadcast address. 1857 */ 1858 /* ARGSUSED */ 1859 struct ifaddr * 1860 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1861 { 1862 struct ifnet *ifp; 1863 struct ifaddr *ifa; 1864 1865 NET_EPOCH_ASSERT(); 1866 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1867 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1868 continue; 1869 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1870 if (ifa->ifa_addr->sa_family != addr->sa_family) 1871 continue; 1872 if ((ifp->if_flags & IFF_BROADCAST) && 1873 ifa->ifa_broadaddr && 1874 ifa->ifa_broadaddr->sa_len != 0 && 1875 sa_equal(ifa->ifa_broadaddr, addr)) { 1876 goto done; 1877 } 1878 } 1879 } 1880 ifa = NULL; 1881 done: 1882 return (ifa); 1883 } 1884 1885 /* 1886 * Locate the point to point interface with a given destination address. 1887 */ 1888 /*ARGSUSED*/ 1889 struct ifaddr * 1890 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1891 { 1892 struct ifnet *ifp; 1893 struct ifaddr *ifa; 1894 1895 NET_EPOCH_ASSERT(); 1896 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1897 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1898 continue; 1899 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1900 continue; 1901 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1902 if (ifa->ifa_addr->sa_family != addr->sa_family) 1903 continue; 1904 if (ifa->ifa_dstaddr != NULL && 1905 sa_equal(addr, ifa->ifa_dstaddr)) { 1906 goto done; 1907 } 1908 } 1909 } 1910 ifa = NULL; 1911 done: 1912 return (ifa); 1913 } 1914 1915 /* 1916 * Find an interface on a specific network. If many, choice 1917 * is most specific found. 1918 */ 1919 struct ifaddr * 1920 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1921 { 1922 struct ifnet *ifp; 1923 struct ifaddr *ifa; 1924 struct ifaddr *ifa_maybe = NULL; 1925 u_int af = addr->sa_family; 1926 const char *addr_data = addr->sa_data, *cplim; 1927 1928 NET_EPOCH_ASSERT(); 1929 /* 1930 * AF_LINK addresses can be looked up directly by their index number, 1931 * so do that if we can. 1932 */ 1933 if (af == AF_LINK) { 1934 ifp = ifnet_byindex( 1935 ((const struct sockaddr_dl *)addr)->sdl_index); 1936 return (ifp ? ifp->if_addr : NULL); 1937 } 1938 1939 /* 1940 * Scan though each interface, looking for ones that have addresses 1941 * in this address family and the requested fib. 1942 */ 1943 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1944 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1945 continue; 1946 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1947 const char *cp, *cp2, *cp3; 1948 1949 if (ifa->ifa_addr->sa_family != af) 1950 next: continue; 1951 if (af == AF_INET && 1952 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1953 /* 1954 * This is a bit broken as it doesn't 1955 * take into account that the remote end may 1956 * be a single node in the network we are 1957 * looking for. 1958 * The trouble is that we don't know the 1959 * netmask for the remote end. 1960 */ 1961 if (ifa->ifa_dstaddr != NULL && 1962 sa_equal(addr, ifa->ifa_dstaddr)) { 1963 goto done; 1964 } 1965 } else { 1966 /* 1967 * Scan all the bits in the ifa's address. 1968 * If a bit dissagrees with what we are 1969 * looking for, mask it with the netmask 1970 * to see if it really matters. 1971 * (A byte at a time) 1972 */ 1973 if (ifa->ifa_netmask == 0) 1974 continue; 1975 cp = addr_data; 1976 cp2 = ifa->ifa_addr->sa_data; 1977 cp3 = ifa->ifa_netmask->sa_data; 1978 cplim = ifa->ifa_netmask->sa_len 1979 + (char *)ifa->ifa_netmask; 1980 while (cp3 < cplim) 1981 if ((*cp++ ^ *cp2++) & *cp3++) 1982 goto next; /* next address! */ 1983 /* 1984 * If the netmask of what we just found 1985 * is more specific than what we had before 1986 * (if we had one), or if the virtual status 1987 * of new prefix is better than of the old one, 1988 * then remember the new one before continuing 1989 * to search for an even better one. 1990 */ 1991 if (ifa_maybe == NULL || 1992 ifa_preferred(ifa_maybe, ifa) || 1993 rn_refines((caddr_t)ifa->ifa_netmask, 1994 (caddr_t)ifa_maybe->ifa_netmask)) { 1995 ifa_maybe = ifa; 1996 } 1997 } 1998 } 1999 } 2000 ifa = ifa_maybe; 2001 ifa_maybe = NULL; 2002 done: 2003 return (ifa); 2004 } 2005 2006 /* 2007 * Find an interface address specific to an interface best matching 2008 * a given address. 2009 */ 2010 struct ifaddr * 2011 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2012 { 2013 struct ifaddr *ifa; 2014 const char *cp, *cp2, *cp3; 2015 char *cplim; 2016 struct ifaddr *ifa_maybe = NULL; 2017 u_int af = addr->sa_family; 2018 2019 if (af >= AF_MAX) 2020 return (NULL); 2021 2022 NET_EPOCH_ASSERT(); 2023 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2024 if (ifa->ifa_addr->sa_family != af) 2025 continue; 2026 if (ifa_maybe == NULL) 2027 ifa_maybe = ifa; 2028 if (ifa->ifa_netmask == 0) { 2029 if (sa_equal(addr, ifa->ifa_addr) || 2030 (ifa->ifa_dstaddr && 2031 sa_equal(addr, ifa->ifa_dstaddr))) 2032 goto done; 2033 continue; 2034 } 2035 if (ifp->if_flags & IFF_POINTOPOINT) { 2036 if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr)) 2037 goto done; 2038 } else { 2039 cp = addr->sa_data; 2040 cp2 = ifa->ifa_addr->sa_data; 2041 cp3 = ifa->ifa_netmask->sa_data; 2042 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2043 for (; cp3 < cplim; cp3++) 2044 if ((*cp++ ^ *cp2++) & *cp3) 2045 break; 2046 if (cp3 == cplim) 2047 goto done; 2048 } 2049 } 2050 ifa = ifa_maybe; 2051 done: 2052 return (ifa); 2053 } 2054 2055 /* 2056 * See whether new ifa is better than current one: 2057 * 1) A non-virtual one is preferred over virtual. 2058 * 2) A virtual in master state preferred over any other state. 2059 * 2060 * Used in several address selecting functions. 2061 */ 2062 int 2063 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2064 { 2065 2066 return (cur->ifa_carp && (!next->ifa_carp || 2067 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2068 } 2069 2070 struct sockaddr_dl * 2071 link_alloc_sdl(size_t size, int flags) 2072 { 2073 2074 return (malloc(size, M_TEMP, flags)); 2075 } 2076 2077 void 2078 link_free_sdl(struct sockaddr *sa) 2079 { 2080 free(sa, M_TEMP); 2081 } 2082 2083 /* 2084 * Fills in given sdl with interface basic info. 2085 * Returns pointer to filled sdl. 2086 */ 2087 struct sockaddr_dl * 2088 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2089 { 2090 struct sockaddr_dl *sdl; 2091 2092 sdl = (struct sockaddr_dl *)paddr; 2093 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2094 sdl->sdl_len = sizeof(struct sockaddr_dl); 2095 sdl->sdl_family = AF_LINK; 2096 sdl->sdl_index = ifp->if_index; 2097 sdl->sdl_type = iftype; 2098 2099 return (sdl); 2100 } 2101 2102 /* 2103 * Mark an interface down and notify protocols of 2104 * the transition. 2105 */ 2106 static void 2107 if_unroute(struct ifnet *ifp, int flag, int fam) 2108 { 2109 2110 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2111 2112 ifp->if_flags &= ~flag; 2113 getmicrotime(&ifp->if_lastchange); 2114 ifp->if_qflush(ifp); 2115 2116 if (ifp->if_carp) 2117 (*carp_linkstate_p)(ifp); 2118 rt_ifmsg(ifp, IFF_UP); 2119 } 2120 2121 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2122 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2123 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2124 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2125 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2126 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2127 int (*vlan_setcookie_p)(struct ifnet *, void *); 2128 void *(*vlan_cookie_p)(struct ifnet *); 2129 2130 /* 2131 * Handle a change in the interface link state. To avoid LORs 2132 * between driver lock and upper layer locks, as well as possible 2133 * recursions, we post event to taskqueue, and all job 2134 * is done in static do_link_state_change(). 2135 */ 2136 void 2137 if_link_state_change(struct ifnet *ifp, int link_state) 2138 { 2139 /* Return if state hasn't changed. */ 2140 if (ifp->if_link_state == link_state) 2141 return; 2142 2143 ifp->if_link_state = link_state; 2144 2145 /* XXXGL: reference ifp? */ 2146 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2147 } 2148 2149 static void 2150 do_link_state_change(void *arg, int pending) 2151 { 2152 struct ifnet *ifp; 2153 int link_state; 2154 2155 ifp = arg; 2156 link_state = ifp->if_link_state; 2157 2158 CURVNET_SET(ifp->if_vnet); 2159 rt_ifmsg(ifp, 0); 2160 if (ifp->if_vlantrunk != NULL) 2161 (*vlan_link_state_p)(ifp); 2162 2163 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2164 ifp->if_l2com != NULL) 2165 (*ng_ether_link_state_p)(ifp, link_state); 2166 if (ifp->if_carp) 2167 (*carp_linkstate_p)(ifp); 2168 if (ifp->if_bridge) 2169 ifp->if_bridge_linkstate(ifp); 2170 if (ifp->if_lagg) 2171 (*lagg_linkstate_p)(ifp, link_state); 2172 2173 if (IS_DEFAULT_VNET(curvnet)) 2174 devctl_notify("IFNET", ifp->if_xname, 2175 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2176 NULL); 2177 if (pending > 1) 2178 if_printf(ifp, "%d link states coalesced\n", pending); 2179 if (log_link_state_change) 2180 if_printf(ifp, "link state changed to %s\n", 2181 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2182 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2183 CURVNET_RESTORE(); 2184 } 2185 2186 /* 2187 * Mark an interface down and notify protocols of 2188 * the transition. 2189 */ 2190 void 2191 if_down(struct ifnet *ifp) 2192 { 2193 2194 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2195 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2196 } 2197 2198 /* 2199 * Mark an interface up and notify protocols of 2200 * the transition. 2201 */ 2202 void 2203 if_up(struct ifnet *ifp) 2204 { 2205 2206 ifp->if_flags |= IFF_UP; 2207 getmicrotime(&ifp->if_lastchange); 2208 if (ifp->if_carp) 2209 (*carp_linkstate_p)(ifp); 2210 rt_ifmsg(ifp, IFF_UP); 2211 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2212 } 2213 2214 /* 2215 * Flush an interface queue. 2216 */ 2217 void 2218 if_qflush(struct ifnet *ifp) 2219 { 2220 struct mbuf *m, *n; 2221 struct ifaltq *ifq; 2222 2223 ifq = &ifp->if_snd; 2224 IFQ_LOCK(ifq); 2225 #ifdef ALTQ 2226 if (ALTQ_IS_ENABLED(ifq)) 2227 ALTQ_PURGE(ifq); 2228 #endif 2229 n = ifq->ifq_head; 2230 while ((m = n) != NULL) { 2231 n = m->m_nextpkt; 2232 m_freem(m); 2233 } 2234 ifq->ifq_head = 0; 2235 ifq->ifq_tail = 0; 2236 ifq->ifq_len = 0; 2237 IFQ_UNLOCK(ifq); 2238 } 2239 2240 /* 2241 * Map interface name to interface structure pointer, with or without 2242 * returning a reference. 2243 */ 2244 struct ifnet * 2245 ifunit_ref(const char *name) 2246 { 2247 struct epoch_tracker et; 2248 struct ifnet *ifp; 2249 2250 NET_EPOCH_ENTER(et); 2251 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2252 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2253 !(ifp->if_flags & IFF_DYING)) 2254 break; 2255 } 2256 if (ifp != NULL) { 2257 if_ref(ifp); 2258 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2259 } 2260 2261 NET_EPOCH_EXIT(et); 2262 return (ifp); 2263 } 2264 2265 struct ifnet * 2266 ifunit(const char *name) 2267 { 2268 struct epoch_tracker et; 2269 struct ifnet *ifp; 2270 2271 NET_EPOCH_ENTER(et); 2272 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2273 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2274 break; 2275 } 2276 NET_EPOCH_EXIT(et); 2277 return (ifp); 2278 } 2279 2280 void * 2281 ifr_buffer_get_buffer(void *data) 2282 { 2283 union ifreq_union *ifrup; 2284 2285 ifrup = data; 2286 #ifdef COMPAT_FREEBSD32 2287 if (SV_CURPROC_FLAG(SV_ILP32)) 2288 return ((void *)(uintptr_t) 2289 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2290 #endif 2291 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2292 } 2293 2294 static void 2295 ifr_buffer_set_buffer_null(void *data) 2296 { 2297 union ifreq_union *ifrup; 2298 2299 ifrup = data; 2300 #ifdef COMPAT_FREEBSD32 2301 if (SV_CURPROC_FLAG(SV_ILP32)) 2302 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2303 else 2304 #endif 2305 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2306 } 2307 2308 size_t 2309 ifr_buffer_get_length(void *data) 2310 { 2311 union ifreq_union *ifrup; 2312 2313 ifrup = data; 2314 #ifdef COMPAT_FREEBSD32 2315 if (SV_CURPROC_FLAG(SV_ILP32)) 2316 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2317 #endif 2318 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2319 } 2320 2321 static void 2322 ifr_buffer_set_length(void *data, size_t len) 2323 { 2324 union ifreq_union *ifrup; 2325 2326 ifrup = data; 2327 #ifdef COMPAT_FREEBSD32 2328 if (SV_CURPROC_FLAG(SV_ILP32)) 2329 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2330 else 2331 #endif 2332 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2333 } 2334 2335 void * 2336 ifr_data_get_ptr(void *ifrp) 2337 { 2338 union ifreq_union *ifrup; 2339 2340 ifrup = ifrp; 2341 #ifdef COMPAT_FREEBSD32 2342 if (SV_CURPROC_FLAG(SV_ILP32)) 2343 return ((void *)(uintptr_t) 2344 ifrup->ifr32.ifr_ifru.ifru_data); 2345 #endif 2346 return (ifrup->ifr.ifr_ifru.ifru_data); 2347 } 2348 2349 struct ifcap_nv_bit_name { 2350 uint64_t cap_bit; 2351 const char *cap_name; 2352 }; 2353 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2354 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2355 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2356 CAPNV(RXCSUM), 2357 CAPNV(TXCSUM), 2358 CAPNV(NETCONS), 2359 CAPNV(VLAN_MTU), 2360 CAPNV(VLAN_HWTAGGING), 2361 CAPNV(JUMBO_MTU), 2362 CAPNV(POLLING), 2363 CAPNV(VLAN_HWCSUM), 2364 CAPNV(TSO4), 2365 CAPNV(TSO6), 2366 CAPNV(LRO), 2367 CAPNV(WOL_UCAST), 2368 CAPNV(WOL_MCAST), 2369 CAPNV(WOL_MAGIC), 2370 CAPNV(TOE4), 2371 CAPNV(TOE6), 2372 CAPNV(VLAN_HWFILTER), 2373 CAPNV(VLAN_HWTSO), 2374 CAPNV(LINKSTATE), 2375 CAPNV(NETMAP), 2376 CAPNV(RXCSUM_IPV6), 2377 CAPNV(TXCSUM_IPV6), 2378 CAPNV(HWSTATS), 2379 CAPNV(TXRTLMT), 2380 CAPNV(HWRXTSTMP), 2381 CAPNV(MEXTPG), 2382 CAPNV(TXTLS4), 2383 CAPNV(TXTLS6), 2384 CAPNV(VXLAN_HWCSUM), 2385 CAPNV(VXLAN_HWTSO), 2386 CAPNV(TXTLS_RTLMT), 2387 {0, NULL} 2388 }; 2389 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \ 2390 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2391 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2392 CAP2NV(RXTLS4), 2393 CAP2NV(RXTLS6), 2394 CAP2NV(IPSEC_OFFLOAD), 2395 {0, NULL} 2396 }; 2397 #undef CAPNV 2398 #undef CAP2NV 2399 2400 int 2401 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2402 const struct ifcap_nv_bit_name *nn, bool all) 2403 { 2404 int i, res; 2405 2406 res = 0; 2407 for (i = 0; nn[i].cap_name != NULL; i++) { 2408 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2409 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2410 res |= nn[i].cap_bit; 2411 } else { 2412 res |= *old_cap & nn[i].cap_bit; 2413 } 2414 } 2415 return (res); 2416 } 2417 2418 void 2419 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2420 int ifr_cap, int ifr_req) 2421 { 2422 int i; 2423 2424 for (i = 0; nn[i].cap_name != NULL; i++) { 2425 if ((nn[i].cap_bit & ifr_cap) != 0) { 2426 nvlist_add_bool(nv, nn[i].cap_name, 2427 (nn[i].cap_bit & ifr_req) != 0); 2428 } 2429 } 2430 } 2431 2432 /* 2433 * Hardware specific interface ioctls. 2434 */ 2435 int 2436 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2437 { 2438 struct ifreq *ifr; 2439 int error = 0, do_ifup = 0; 2440 int new_flags, temp_flags; 2441 size_t descrlen, nvbuflen; 2442 char *descrbuf; 2443 char new_name[IFNAMSIZ]; 2444 void *buf; 2445 nvlist_t *nvcap; 2446 struct siocsifcapnv_driver_data drv_ioctl_data; 2447 2448 ifr = (struct ifreq *)data; 2449 switch (cmd) { 2450 case SIOCGIFINDEX: 2451 ifr->ifr_index = ifp->if_index; 2452 break; 2453 2454 case SIOCGIFFLAGS: 2455 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2456 ifr->ifr_flags = temp_flags & 0xffff; 2457 ifr->ifr_flagshigh = temp_flags >> 16; 2458 break; 2459 2460 case SIOCGIFCAP: 2461 ifr->ifr_reqcap = ifp->if_capabilities; 2462 ifr->ifr_curcap = ifp->if_capenable; 2463 break; 2464 2465 case SIOCGIFCAPNV: 2466 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2467 error = EINVAL; 2468 break; 2469 } 2470 buf = NULL; 2471 nvcap = nvlist_create(0); 2472 for (;;) { 2473 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2474 ifp->if_capabilities, ifp->if_capenable); 2475 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2476 ifp->if_capabilities2, ifp->if_capenable2); 2477 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2478 __DECONST(caddr_t, nvcap)); 2479 if (error != 0) { 2480 if_printf(ifp, 2481 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2482 error); 2483 break; 2484 } 2485 buf = nvlist_pack(nvcap, &nvbuflen); 2486 if (buf == NULL) { 2487 error = nvlist_error(nvcap); 2488 if (error == 0) 2489 error = EDOOFUS; 2490 break; 2491 } 2492 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2493 ifr->ifr_cap_nv.length = nvbuflen; 2494 ifr->ifr_cap_nv.buffer = NULL; 2495 error = EFBIG; 2496 break; 2497 } 2498 ifr->ifr_cap_nv.length = nvbuflen; 2499 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2500 break; 2501 } 2502 free(buf, M_NVLIST); 2503 nvlist_destroy(nvcap); 2504 break; 2505 2506 case SIOCGIFDATA: 2507 { 2508 struct if_data ifd; 2509 2510 /* Ensure uninitialised padding is not leaked. */ 2511 memset(&ifd, 0, sizeof(ifd)); 2512 2513 if_data_copy(ifp, &ifd); 2514 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2515 break; 2516 } 2517 2518 #ifdef MAC 2519 case SIOCGIFMAC: 2520 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2521 break; 2522 #endif 2523 2524 case SIOCGIFMETRIC: 2525 ifr->ifr_metric = ifp->if_metric; 2526 break; 2527 2528 case SIOCGIFMTU: 2529 ifr->ifr_mtu = ifp->if_mtu; 2530 break; 2531 2532 case SIOCGIFPHYS: 2533 /* XXXGL: did this ever worked? */ 2534 ifr->ifr_phys = 0; 2535 break; 2536 2537 case SIOCGIFDESCR: 2538 error = 0; 2539 sx_slock(&ifdescr_sx); 2540 if (ifp->if_description == NULL) 2541 error = ENOMSG; 2542 else { 2543 /* space for terminating nul */ 2544 descrlen = strlen(ifp->if_description) + 1; 2545 if (ifr_buffer_get_length(ifr) < descrlen) 2546 ifr_buffer_set_buffer_null(ifr); 2547 else 2548 error = copyout(ifp->if_description, 2549 ifr_buffer_get_buffer(ifr), descrlen); 2550 ifr_buffer_set_length(ifr, descrlen); 2551 } 2552 sx_sunlock(&ifdescr_sx); 2553 break; 2554 2555 case SIOCSIFDESCR: 2556 error = priv_check(td, PRIV_NET_SETIFDESCR); 2557 if (error) 2558 return (error); 2559 2560 /* 2561 * Copy only (length-1) bytes to make sure that 2562 * if_description is always nul terminated. The 2563 * length parameter is supposed to count the 2564 * terminating nul in. 2565 */ 2566 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2567 return (ENAMETOOLONG); 2568 else if (ifr_buffer_get_length(ifr) == 0) 2569 descrbuf = NULL; 2570 else { 2571 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2572 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2573 ifr_buffer_get_length(ifr) - 1); 2574 if (error) { 2575 if_freedescr(descrbuf); 2576 break; 2577 } 2578 } 2579 2580 if_setdescr(ifp, descrbuf); 2581 getmicrotime(&ifp->if_lastchange); 2582 break; 2583 2584 case SIOCGIFFIB: 2585 ifr->ifr_fib = ifp->if_fib; 2586 break; 2587 2588 case SIOCSIFFIB: 2589 error = priv_check(td, PRIV_NET_SETIFFIB); 2590 if (error) 2591 return (error); 2592 if (ifr->ifr_fib >= rt_numfibs) 2593 return (EINVAL); 2594 2595 ifp->if_fib = ifr->ifr_fib; 2596 break; 2597 2598 case SIOCSIFFLAGS: 2599 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2600 if (error) 2601 return (error); 2602 /* 2603 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2604 * check, so we don't need special handling here yet. 2605 */ 2606 new_flags = (ifr->ifr_flags & 0xffff) | 2607 (ifr->ifr_flagshigh << 16); 2608 if (ifp->if_flags & IFF_UP && 2609 (new_flags & IFF_UP) == 0) { 2610 if_down(ifp); 2611 } else if (new_flags & IFF_UP && 2612 (ifp->if_flags & IFF_UP) == 0) { 2613 do_ifup = 1; 2614 } 2615 2616 /* 2617 * See if the promiscuous mode or allmulti bits are about to 2618 * flip. They require special handling because in-kernel 2619 * consumers may indepdently toggle them. 2620 */ 2621 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2622 if (new_flags & IFF_PPROMISC) 2623 ifp->if_flags |= IFF_PROMISC; 2624 else if (ifp->if_pcount == 0) 2625 ifp->if_flags &= ~IFF_PROMISC; 2626 if (log_promisc_mode_change) 2627 if_printf(ifp, "permanently promiscuous mode %s\n", 2628 ((new_flags & IFF_PPROMISC) ? 2629 "enabled" : "disabled")); 2630 } 2631 if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) { 2632 if (new_flags & IFF_PALLMULTI) 2633 ifp->if_flags |= IFF_ALLMULTI; 2634 else if (ifp->if_amcount == 0) 2635 ifp->if_flags &= ~IFF_ALLMULTI; 2636 } 2637 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2638 (new_flags &~ IFF_CANTCHANGE); 2639 if (ifp->if_ioctl) { 2640 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2641 } 2642 if (do_ifup) 2643 if_up(ifp); 2644 getmicrotime(&ifp->if_lastchange); 2645 break; 2646 2647 case SIOCSIFCAP: 2648 error = priv_check(td, PRIV_NET_SETIFCAP); 2649 if (error != 0) 2650 return (error); 2651 if (ifp->if_ioctl == NULL) 2652 return (EOPNOTSUPP); 2653 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2654 return (EINVAL); 2655 error = (*ifp->if_ioctl)(ifp, cmd, data); 2656 if (error == 0) 2657 getmicrotime(&ifp->if_lastchange); 2658 break; 2659 2660 case SIOCSIFCAPNV: 2661 error = priv_check(td, PRIV_NET_SETIFCAP); 2662 if (error != 0) 2663 return (error); 2664 if (ifp->if_ioctl == NULL) 2665 return (EOPNOTSUPP); 2666 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2667 return (EINVAL); 2668 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2669 return (EINVAL); 2670 nvcap = NULL; 2671 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2672 for (;;) { 2673 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2674 ifr->ifr_cap_nv.length); 2675 if (error != 0) 2676 break; 2677 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2678 if (nvcap == NULL) { 2679 error = EINVAL; 2680 break; 2681 } 2682 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2683 &ifp->if_capenable, ifcap_nv_bit_names, false); 2684 if ((drv_ioctl_data.reqcap & 2685 ~ifp->if_capabilities) != 0) { 2686 error = EINVAL; 2687 break; 2688 } 2689 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2690 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2691 if ((drv_ioctl_data.reqcap2 & 2692 ~ifp->if_capabilities2) != 0) { 2693 error = EINVAL; 2694 break; 2695 } 2696 drv_ioctl_data.nvcap = nvcap; 2697 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2698 (caddr_t)&drv_ioctl_data); 2699 break; 2700 } 2701 nvlist_destroy(nvcap); 2702 free(buf, M_TEMP); 2703 if (error == 0) 2704 getmicrotime(&ifp->if_lastchange); 2705 break; 2706 2707 #ifdef MAC 2708 case SIOCSIFMAC: 2709 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2710 break; 2711 #endif 2712 2713 case SIOCSIFNAME: 2714 error = priv_check(td, PRIV_NET_SETIFNAME); 2715 if (error) 2716 return (error); 2717 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2718 NULL); 2719 if (error != 0) 2720 return (error); 2721 error = if_rename(ifp, new_name); 2722 break; 2723 2724 #ifdef VIMAGE 2725 case SIOCSIFVNET: 2726 error = priv_check(td, PRIV_NET_SETIFVNET); 2727 if (error) 2728 return (error); 2729 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2730 break; 2731 #endif 2732 2733 case SIOCSIFMETRIC: 2734 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2735 if (error) 2736 return (error); 2737 ifp->if_metric = ifr->ifr_metric; 2738 getmicrotime(&ifp->if_lastchange); 2739 break; 2740 2741 case SIOCSIFPHYS: 2742 error = priv_check(td, PRIV_NET_SETIFPHYS); 2743 if (error) 2744 return (error); 2745 if (ifp->if_ioctl == NULL) 2746 return (EOPNOTSUPP); 2747 error = (*ifp->if_ioctl)(ifp, cmd, data); 2748 if (error == 0) 2749 getmicrotime(&ifp->if_lastchange); 2750 break; 2751 2752 case SIOCSIFMTU: 2753 { 2754 u_long oldmtu = ifp->if_mtu; 2755 2756 error = priv_check(td, PRIV_NET_SETIFMTU); 2757 if (error) 2758 return (error); 2759 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2760 return (EINVAL); 2761 if (ifp->if_ioctl == NULL) 2762 return (EOPNOTSUPP); 2763 /* Disallow MTU changes on bridge member interfaces. */ 2764 if (ifp->if_bridge) 2765 return (EOPNOTSUPP); 2766 error = (*ifp->if_ioctl)(ifp, cmd, data); 2767 if (error == 0) { 2768 getmicrotime(&ifp->if_lastchange); 2769 rt_ifmsg(ifp, 0); 2770 #ifdef INET 2771 DEBUGNET_NOTIFY_MTU(ifp); 2772 #endif 2773 } 2774 /* 2775 * If the link MTU changed, do network layer specific procedure. 2776 */ 2777 if (ifp->if_mtu != oldmtu) 2778 if_notifymtu(ifp); 2779 break; 2780 } 2781 2782 case SIOCADDMULTI: 2783 case SIOCDELMULTI: 2784 if (cmd == SIOCADDMULTI) 2785 error = priv_check(td, PRIV_NET_ADDMULTI); 2786 else 2787 error = priv_check(td, PRIV_NET_DELMULTI); 2788 if (error) 2789 return (error); 2790 2791 /* Don't allow group membership on non-multicast interfaces. */ 2792 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2793 return (EOPNOTSUPP); 2794 2795 /* Don't let users screw up protocols' entries. */ 2796 if (ifr->ifr_addr.sa_family != AF_LINK) 2797 return (EINVAL); 2798 2799 if (cmd == SIOCADDMULTI) { 2800 struct epoch_tracker et; 2801 struct ifmultiaddr *ifma; 2802 2803 /* 2804 * Userland is only permitted to join groups once 2805 * via the if_addmulti() KPI, because it cannot hold 2806 * struct ifmultiaddr * between calls. It may also 2807 * lose a race while we check if the membership 2808 * already exists. 2809 */ 2810 NET_EPOCH_ENTER(et); 2811 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2812 NET_EPOCH_EXIT(et); 2813 if (ifma != NULL) 2814 error = EADDRINUSE; 2815 else 2816 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2817 } else { 2818 error = if_delmulti(ifp, &ifr->ifr_addr); 2819 } 2820 if (error == 0) 2821 getmicrotime(&ifp->if_lastchange); 2822 break; 2823 2824 case SIOCSIFPHYADDR: 2825 case SIOCDIFPHYADDR: 2826 #ifdef INET6 2827 case SIOCSIFPHYADDR_IN6: 2828 #endif 2829 case SIOCSIFMEDIA: 2830 case SIOCSIFGENERIC: 2831 error = priv_check(td, PRIV_NET_HWIOCTL); 2832 if (error) 2833 return (error); 2834 if (ifp->if_ioctl == NULL) 2835 return (EOPNOTSUPP); 2836 error = (*ifp->if_ioctl)(ifp, cmd, data); 2837 if (error == 0) 2838 getmicrotime(&ifp->if_lastchange); 2839 break; 2840 2841 case SIOCGIFSTATUS: 2842 case SIOCGIFPSRCADDR: 2843 case SIOCGIFPDSTADDR: 2844 case SIOCGIFMEDIA: 2845 case SIOCGIFXMEDIA: 2846 case SIOCGIFGENERIC: 2847 case SIOCGIFRSSKEY: 2848 case SIOCGIFRSSHASH: 2849 case SIOCGIFDOWNREASON: 2850 if (ifp->if_ioctl == NULL) 2851 return (EOPNOTSUPP); 2852 error = (*ifp->if_ioctl)(ifp, cmd, data); 2853 break; 2854 2855 case SIOCSIFLLADDR: 2856 error = priv_check(td, PRIV_NET_SETLLADDR); 2857 if (error) 2858 return (error); 2859 error = if_setlladdr(ifp, 2860 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2861 break; 2862 2863 case SIOCGHWADDR: 2864 error = if_gethwaddr(ifp, ifr); 2865 break; 2866 2867 case SIOCAIFGROUP: 2868 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2869 if (error) 2870 return (error); 2871 error = if_addgroup(ifp, 2872 ((struct ifgroupreq *)data)->ifgr_group); 2873 if (error != 0) 2874 return (error); 2875 break; 2876 2877 case SIOCGIFGROUP: 2878 { 2879 struct epoch_tracker et; 2880 2881 NET_EPOCH_ENTER(et); 2882 error = if_getgroup((struct ifgroupreq *)data, ifp); 2883 NET_EPOCH_EXIT(et); 2884 break; 2885 } 2886 2887 case SIOCDIFGROUP: 2888 error = priv_check(td, PRIV_NET_DELIFGROUP); 2889 if (error) 2890 return (error); 2891 error = if_delgroup(ifp, 2892 ((struct ifgroupreq *)data)->ifgr_group); 2893 if (error != 0) 2894 return (error); 2895 break; 2896 2897 default: 2898 error = ENOIOCTL; 2899 break; 2900 } 2901 return (error); 2902 } 2903 2904 /* 2905 * Interface ioctls. 2906 */ 2907 int 2908 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2909 { 2910 #ifdef COMPAT_FREEBSD32 2911 union { 2912 struct ifconf ifc; 2913 struct ifdrv ifd; 2914 struct ifgroupreq ifgr; 2915 struct ifmediareq ifmr; 2916 } thunk; 2917 u_long saved_cmd; 2918 struct ifconf32 *ifc32; 2919 struct ifdrv32 *ifd32; 2920 struct ifgroupreq32 *ifgr32; 2921 struct ifmediareq32 *ifmr32; 2922 #endif 2923 struct ifnet *ifp; 2924 struct ifreq *ifr; 2925 int error; 2926 int oif_flags; 2927 #ifdef VIMAGE 2928 bool shutdown; 2929 #endif 2930 2931 CURVNET_SET(so->so_vnet); 2932 #ifdef VIMAGE 2933 /* Make sure the VNET is stable. */ 2934 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2935 if (shutdown) { 2936 CURVNET_RESTORE(); 2937 return (EBUSY); 2938 } 2939 #endif 2940 2941 #ifdef COMPAT_FREEBSD32 2942 saved_cmd = cmd; 2943 switch (cmd) { 2944 case SIOCGIFCONF32: 2945 ifc32 = (struct ifconf32 *)data; 2946 thunk.ifc.ifc_len = ifc32->ifc_len; 2947 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2948 data = (caddr_t)&thunk.ifc; 2949 cmd = SIOCGIFCONF; 2950 break; 2951 case SIOCGDRVSPEC32: 2952 case SIOCSDRVSPEC32: 2953 ifd32 = (struct ifdrv32 *)data; 2954 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2955 sizeof(thunk.ifd.ifd_name)); 2956 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2957 thunk.ifd.ifd_len = ifd32->ifd_len; 2958 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2959 data = (caddr_t)&thunk.ifd; 2960 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2961 break; 2962 case SIOCAIFGROUP32: 2963 case SIOCGIFGROUP32: 2964 case SIOCDIFGROUP32: 2965 case SIOCGIFGMEMB32: 2966 ifgr32 = (struct ifgroupreq32 *)data; 2967 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2968 sizeof(thunk.ifgr.ifgr_name)); 2969 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2970 switch (cmd) { 2971 case SIOCAIFGROUP32: 2972 case SIOCDIFGROUP32: 2973 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2974 sizeof(thunk.ifgr.ifgr_group)); 2975 break; 2976 case SIOCGIFGROUP32: 2977 case SIOCGIFGMEMB32: 2978 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2979 break; 2980 } 2981 data = (caddr_t)&thunk.ifgr; 2982 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2983 break; 2984 case SIOCGIFMEDIA32: 2985 case SIOCGIFXMEDIA32: 2986 ifmr32 = (struct ifmediareq32 *)data; 2987 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2988 sizeof(thunk.ifmr.ifm_name)); 2989 thunk.ifmr.ifm_current = ifmr32->ifm_current; 2990 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 2991 thunk.ifmr.ifm_status = ifmr32->ifm_status; 2992 thunk.ifmr.ifm_active = ifmr32->ifm_active; 2993 thunk.ifmr.ifm_count = ifmr32->ifm_count; 2994 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 2995 data = (caddr_t)&thunk.ifmr; 2996 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 2997 break; 2998 } 2999 #endif 3000 3001 switch (cmd) { 3002 case SIOCGIFCONF: 3003 error = ifconf(cmd, data); 3004 goto out_noref; 3005 } 3006 3007 ifr = (struct ifreq *)data; 3008 switch (cmd) { 3009 #ifdef VIMAGE 3010 case SIOCSIFRVNET: 3011 error = priv_check(td, PRIV_NET_SETIFVNET); 3012 if (error == 0) 3013 error = if_vmove_reclaim(td, ifr->ifr_name, 3014 ifr->ifr_jid); 3015 goto out_noref; 3016 #endif 3017 case SIOCIFCREATE: 3018 case SIOCIFCREATE2: 3019 error = priv_check(td, PRIV_NET_IFCREATE); 3020 if (error == 0) 3021 error = if_clone_create(ifr->ifr_name, 3022 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3023 ifr_data_get_ptr(ifr) : NULL); 3024 goto out_noref; 3025 case SIOCIFDESTROY: 3026 error = priv_check(td, PRIV_NET_IFDESTROY); 3027 3028 if (error == 0) { 3029 sx_xlock(&ifnet_detach_sxlock); 3030 error = if_clone_destroy(ifr->ifr_name); 3031 sx_xunlock(&ifnet_detach_sxlock); 3032 } 3033 goto out_noref; 3034 3035 case SIOCIFGCLONERS: 3036 error = if_clone_list((struct if_clonereq *)data); 3037 goto out_noref; 3038 3039 case SIOCGIFGMEMB: 3040 error = if_getgroupmembers((struct ifgroupreq *)data); 3041 goto out_noref; 3042 3043 #if defined(INET) || defined(INET6) 3044 case SIOCSVH: 3045 case SIOCGVH: 3046 if (carp_ioctl_p == NULL) 3047 error = EPROTONOSUPPORT; 3048 else 3049 error = (*carp_ioctl_p)(ifr, cmd, td); 3050 goto out_noref; 3051 #endif 3052 } 3053 3054 ifp = ifunit_ref(ifr->ifr_name); 3055 if (ifp == NULL) { 3056 error = ENXIO; 3057 goto out_noref; 3058 } 3059 3060 error = ifhwioctl(cmd, ifp, data, td); 3061 if (error != ENOIOCTL) 3062 goto out_ref; 3063 3064 oif_flags = ifp->if_flags; 3065 if (so->so_proto == NULL) { 3066 error = EOPNOTSUPP; 3067 goto out_ref; 3068 } 3069 3070 /* 3071 * Pass the request on to the socket control method, and if the 3072 * latter returns EOPNOTSUPP, directly to the interface. 3073 * 3074 * Make an exception for the legacy SIOCSIF* requests. Drivers 3075 * trust SIOCSIFADDR et al to come from an already privileged 3076 * layer, and do not perform any credentials checks or input 3077 * validation. 3078 */ 3079 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 3080 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3081 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3082 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3083 error = (*ifp->if_ioctl)(ifp, cmd, data); 3084 3085 if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP)) 3086 if_up(ifp); 3087 out_ref: 3088 if_rele(ifp); 3089 out_noref: 3090 CURVNET_RESTORE(); 3091 #ifdef COMPAT_FREEBSD32 3092 if (error != 0) 3093 return (error); 3094 switch (saved_cmd) { 3095 case SIOCGIFCONF32: 3096 ifc32->ifc_len = thunk.ifc.ifc_len; 3097 break; 3098 case SIOCGDRVSPEC32: 3099 /* 3100 * SIOCGDRVSPEC is IOWR, but nothing actually touches 3101 * the struct so just assert that ifd_len (the only 3102 * field it might make sense to update) hasn't 3103 * changed. 3104 */ 3105 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 3106 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 3107 thunk.ifd.ifd_len)); 3108 break; 3109 case SIOCGIFGROUP32: 3110 case SIOCGIFGMEMB32: 3111 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 3112 break; 3113 case SIOCGIFMEDIA32: 3114 case SIOCGIFXMEDIA32: 3115 ifmr32->ifm_current = thunk.ifmr.ifm_current; 3116 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 3117 ifmr32->ifm_status = thunk.ifmr.ifm_status; 3118 ifmr32->ifm_active = thunk.ifmr.ifm_active; 3119 ifmr32->ifm_count = thunk.ifmr.ifm_count; 3120 break; 3121 } 3122 #endif 3123 return (error); 3124 } 3125 3126 int 3127 if_rename(struct ifnet *ifp, char *new_name) 3128 { 3129 struct ifaddr *ifa; 3130 struct sockaddr_dl *sdl; 3131 size_t namelen, onamelen; 3132 char old_name[IFNAMSIZ]; 3133 char strbuf[IFNAMSIZ + 8]; 3134 3135 if (new_name[0] == '\0') 3136 return (EINVAL); 3137 if (strcmp(new_name, ifp->if_xname) == 0) 3138 return (0); 3139 if (ifunit(new_name) != NULL) 3140 return (EEXIST); 3141 3142 /* 3143 * XXX: Locking. Nothing else seems to lock if_flags, 3144 * and there are numerous other races with the 3145 * ifunit() checks not being atomic with namespace 3146 * changes (renames, vmoves, if_attach, etc). 3147 */ 3148 ifp->if_flags |= IFF_RENAMING; 3149 3150 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 3151 3152 if_printf(ifp, "changing name to '%s'\n", new_name); 3153 3154 IF_ADDR_WLOCK(ifp); 3155 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 3156 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 3157 ifa = ifp->if_addr; 3158 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3159 namelen = strlen(new_name); 3160 onamelen = sdl->sdl_nlen; 3161 /* 3162 * Move the address if needed. This is safe because we 3163 * allocate space for a name of length IFNAMSIZ when we 3164 * create this in if_attach(). 3165 */ 3166 if (namelen != onamelen) { 3167 bcopy(sdl->sdl_data + onamelen, 3168 sdl->sdl_data + namelen, sdl->sdl_alen); 3169 } 3170 bcopy(new_name, sdl->sdl_data, namelen); 3171 sdl->sdl_nlen = namelen; 3172 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 3173 bzero(sdl->sdl_data, onamelen); 3174 while (namelen != 0) 3175 sdl->sdl_data[--namelen] = 0xff; 3176 IF_ADDR_WUNLOCK(ifp); 3177 3178 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 3179 3180 ifp->if_flags &= ~IFF_RENAMING; 3181 3182 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 3183 devctl_notify("IFNET", old_name, "RENAME", strbuf); 3184 3185 return (0); 3186 } 3187 3188 /* 3189 * The code common to handling reference counted flags, 3190 * e.g., in ifpromisc() and if_allmulti(). 3191 * The "pflag" argument can specify a permanent mode flag to check, 3192 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3193 * 3194 * Only to be used on stack-owned flags, not driver-owned flags. 3195 */ 3196 static int 3197 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3198 { 3199 struct ifreq ifr; 3200 int error; 3201 int oldflags, oldcount; 3202 3203 /* Sanity checks to catch programming errors */ 3204 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3205 ("%s: setting driver-owned flag %d", __func__, flag)); 3206 3207 if (onswitch) 3208 KASSERT(*refcount >= 0, 3209 ("%s: increment negative refcount %d for flag %d", 3210 __func__, *refcount, flag)); 3211 else 3212 KASSERT(*refcount > 0, 3213 ("%s: decrement non-positive refcount %d for flag %d", 3214 __func__, *refcount, flag)); 3215 3216 /* In case this mode is permanent, just touch refcount */ 3217 if (ifp->if_flags & pflag) { 3218 *refcount += onswitch ? 1 : -1; 3219 return (0); 3220 } 3221 3222 /* Save ifnet parameters for if_ioctl() may fail */ 3223 oldcount = *refcount; 3224 oldflags = ifp->if_flags; 3225 3226 /* 3227 * See if we aren't the only and touching refcount is enough. 3228 * Actually toggle interface flag if we are the first or last. 3229 */ 3230 if (onswitch) { 3231 if ((*refcount)++) 3232 return (0); 3233 ifp->if_flags |= flag; 3234 } else { 3235 if (--(*refcount)) 3236 return (0); 3237 ifp->if_flags &= ~flag; 3238 } 3239 3240 /* Call down the driver since we've changed interface flags */ 3241 if (ifp->if_ioctl == NULL) { 3242 error = EOPNOTSUPP; 3243 goto recover; 3244 } 3245 ifr.ifr_flags = ifp->if_flags & 0xffff; 3246 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3247 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3248 if (error) 3249 goto recover; 3250 /* Notify userland that interface flags have changed */ 3251 rt_ifmsg(ifp, flag); 3252 return (0); 3253 3254 recover: 3255 /* Recover after driver error */ 3256 *refcount = oldcount; 3257 ifp->if_flags = oldflags; 3258 return (error); 3259 } 3260 3261 /* 3262 * Set/clear promiscuous mode on interface ifp based on the truth value 3263 * of pswitch. The calls are reference counted so that only the first 3264 * "on" request actually has an effect, as does the final "off" request. 3265 * Results are undefined if the "off" and "on" requests are not matched. 3266 */ 3267 int 3268 ifpromisc(struct ifnet *ifp, int pswitch) 3269 { 3270 int error; 3271 int oldflags = ifp->if_flags; 3272 3273 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3274 &ifp->if_pcount, pswitch); 3275 /* If promiscuous mode status has changed, log a message */ 3276 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3277 log_promisc_mode_change) 3278 if_printf(ifp, "promiscuous mode %s\n", 3279 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3280 return (error); 3281 } 3282 3283 /* 3284 * Return interface configuration 3285 * of system. List may be used 3286 * in later ioctl's (above) to get 3287 * other information. 3288 */ 3289 /*ARGSUSED*/ 3290 static int 3291 ifconf(u_long cmd, caddr_t data) 3292 { 3293 struct ifconf *ifc = (struct ifconf *)data; 3294 struct ifnet *ifp; 3295 struct ifaddr *ifa; 3296 struct ifreq ifr; 3297 struct sbuf *sb; 3298 int error, full = 0, valid_len, max_len; 3299 3300 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3301 max_len = maxphys - 1; 3302 3303 /* Prevent hostile input from being able to crash the system */ 3304 if (ifc->ifc_len <= 0) 3305 return (EINVAL); 3306 3307 again: 3308 if (ifc->ifc_len <= max_len) { 3309 max_len = ifc->ifc_len; 3310 full = 1; 3311 } 3312 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3313 max_len = 0; 3314 valid_len = 0; 3315 3316 IFNET_RLOCK(); 3317 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3318 struct epoch_tracker et; 3319 int addrs; 3320 3321 /* 3322 * Zero the ifr to make sure we don't disclose the contents 3323 * of the stack. 3324 */ 3325 memset(&ifr, 0, sizeof(ifr)); 3326 3327 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3328 >= sizeof(ifr.ifr_name)) { 3329 sbuf_delete(sb); 3330 IFNET_RUNLOCK(); 3331 return (ENAMETOOLONG); 3332 } 3333 3334 addrs = 0; 3335 NET_EPOCH_ENTER(et); 3336 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3337 struct sockaddr *sa = ifa->ifa_addr; 3338 3339 if (prison_if(curthread->td_ucred, sa) != 0) 3340 continue; 3341 addrs++; 3342 if (sa->sa_len <= sizeof(*sa)) { 3343 if (sa->sa_len < sizeof(*sa)) { 3344 memset(&ifr.ifr_ifru.ifru_addr, 0, 3345 sizeof(ifr.ifr_ifru.ifru_addr)); 3346 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3347 sa->sa_len); 3348 } else 3349 ifr.ifr_ifru.ifru_addr = *sa; 3350 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3351 max_len += sizeof(ifr); 3352 } else { 3353 sbuf_bcat(sb, &ifr, 3354 offsetof(struct ifreq, ifr_addr)); 3355 max_len += offsetof(struct ifreq, ifr_addr); 3356 sbuf_bcat(sb, sa, sa->sa_len); 3357 max_len += sa->sa_len; 3358 } 3359 3360 if (sbuf_error(sb) == 0) 3361 valid_len = sbuf_len(sb); 3362 } 3363 NET_EPOCH_EXIT(et); 3364 if (addrs == 0) { 3365 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3366 max_len += sizeof(ifr); 3367 3368 if (sbuf_error(sb) == 0) 3369 valid_len = sbuf_len(sb); 3370 } 3371 } 3372 IFNET_RUNLOCK(); 3373 3374 /* 3375 * If we didn't allocate enough space (uncommon), try again. If 3376 * we have already allocated as much space as we are allowed, 3377 * return what we've got. 3378 */ 3379 if (valid_len != max_len && !full) { 3380 sbuf_delete(sb); 3381 goto again; 3382 } 3383 3384 ifc->ifc_len = valid_len; 3385 sbuf_finish(sb); 3386 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3387 sbuf_delete(sb); 3388 return (error); 3389 } 3390 3391 /* 3392 * Just like ifpromisc(), but for all-multicast-reception mode. 3393 */ 3394 int 3395 if_allmulti(struct ifnet *ifp, int onswitch) 3396 { 3397 3398 return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount, 3399 onswitch)); 3400 } 3401 3402 struct ifmultiaddr * 3403 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3404 { 3405 struct ifmultiaddr *ifma; 3406 3407 IF_ADDR_LOCK_ASSERT(ifp); 3408 3409 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3410 if (sa->sa_family == AF_LINK) { 3411 if (sa_dl_equal(ifma->ifma_addr, sa)) 3412 break; 3413 } else { 3414 if (sa_equal(ifma->ifma_addr, sa)) 3415 break; 3416 } 3417 } 3418 3419 return ifma; 3420 } 3421 3422 /* 3423 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3424 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3425 * the ifnet multicast address list here, so the caller must do that and 3426 * other setup work (such as notifying the device driver). The reference 3427 * count is initialized to 1. 3428 */ 3429 static struct ifmultiaddr * 3430 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3431 int mflags) 3432 { 3433 struct ifmultiaddr *ifma; 3434 struct sockaddr *dupsa; 3435 3436 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3437 M_ZERO); 3438 if (ifma == NULL) 3439 return (NULL); 3440 3441 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3442 if (dupsa == NULL) { 3443 free(ifma, M_IFMADDR); 3444 return (NULL); 3445 } 3446 bcopy(sa, dupsa, sa->sa_len); 3447 ifma->ifma_addr = dupsa; 3448 3449 ifma->ifma_ifp = ifp; 3450 ifma->ifma_refcount = 1; 3451 ifma->ifma_protospec = NULL; 3452 3453 if (llsa == NULL) { 3454 ifma->ifma_lladdr = NULL; 3455 return (ifma); 3456 } 3457 3458 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3459 if (dupsa == NULL) { 3460 free(ifma->ifma_addr, M_IFMADDR); 3461 free(ifma, M_IFMADDR); 3462 return (NULL); 3463 } 3464 bcopy(llsa, dupsa, llsa->sa_len); 3465 ifma->ifma_lladdr = dupsa; 3466 3467 return (ifma); 3468 } 3469 3470 /* 3471 * if_freemulti: free ifmultiaddr structure and possibly attached related 3472 * addresses. The caller is responsible for implementing reference 3473 * counting, notifying the driver, handling routing messages, and releasing 3474 * any dependent link layer state. 3475 */ 3476 #ifdef MCAST_VERBOSE 3477 extern void kdb_backtrace(void); 3478 #endif 3479 static void 3480 if_freemulti_internal(struct ifmultiaddr *ifma) 3481 { 3482 3483 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3484 ifma->ifma_refcount)); 3485 3486 if (ifma->ifma_lladdr != NULL) 3487 free(ifma->ifma_lladdr, M_IFMADDR); 3488 #ifdef MCAST_VERBOSE 3489 kdb_backtrace(); 3490 printf("%s freeing ifma: %p\n", __func__, ifma); 3491 #endif 3492 free(ifma->ifma_addr, M_IFMADDR); 3493 free(ifma, M_IFMADDR); 3494 } 3495 3496 static void 3497 if_destroymulti(epoch_context_t ctx) 3498 { 3499 struct ifmultiaddr *ifma; 3500 3501 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3502 if_freemulti_internal(ifma); 3503 } 3504 3505 void 3506 if_freemulti(struct ifmultiaddr *ifma) 3507 { 3508 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3509 ifma->ifma_refcount)); 3510 3511 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3512 } 3513 3514 /* 3515 * Register an additional multicast address with a network interface. 3516 * 3517 * - If the address is already present, bump the reference count on the 3518 * address and return. 3519 * - If the address is not link-layer, look up a link layer address. 3520 * - Allocate address structures for one or both addresses, and attach to the 3521 * multicast address list on the interface. If automatically adding a link 3522 * layer address, the protocol address will own a reference to the link 3523 * layer address, to be freed when it is freed. 3524 * - Notify the network device driver of an addition to the multicast address 3525 * list. 3526 * 3527 * 'sa' points to caller-owned memory with the desired multicast address. 3528 * 3529 * 'retifma' will be used to return a pointer to the resulting multicast 3530 * address reference, if desired. 3531 */ 3532 int 3533 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3534 struct ifmultiaddr **retifma) 3535 { 3536 struct ifmultiaddr *ifma, *ll_ifma; 3537 struct sockaddr *llsa; 3538 struct sockaddr_dl sdl; 3539 int error; 3540 3541 #ifdef INET 3542 IN_MULTI_LIST_UNLOCK_ASSERT(); 3543 #endif 3544 #ifdef INET6 3545 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3546 #endif 3547 /* 3548 * If the address is already present, return a new reference to it; 3549 * otherwise, allocate storage and set up a new address. 3550 */ 3551 IF_ADDR_WLOCK(ifp); 3552 ifma = if_findmulti(ifp, sa); 3553 if (ifma != NULL) { 3554 ifma->ifma_refcount++; 3555 if (retifma != NULL) 3556 *retifma = ifma; 3557 IF_ADDR_WUNLOCK(ifp); 3558 return (0); 3559 } 3560 3561 /* 3562 * The address isn't already present; resolve the protocol address 3563 * into a link layer address, and then look that up, bump its 3564 * refcount or allocate an ifma for that also. 3565 * Most link layer resolving functions returns address data which 3566 * fits inside default sockaddr_dl structure. However callback 3567 * can allocate another sockaddr structure, in that case we need to 3568 * free it later. 3569 */ 3570 llsa = NULL; 3571 ll_ifma = NULL; 3572 if (ifp->if_resolvemulti != NULL) { 3573 /* Provide called function with buffer size information */ 3574 sdl.sdl_len = sizeof(sdl); 3575 llsa = (struct sockaddr *)&sdl; 3576 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3577 if (error) 3578 goto unlock_out; 3579 } 3580 3581 /* 3582 * Allocate the new address. Don't hook it up yet, as we may also 3583 * need to allocate a link layer multicast address. 3584 */ 3585 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3586 if (ifma == NULL) { 3587 error = ENOMEM; 3588 goto free_llsa_out; 3589 } 3590 3591 /* 3592 * If a link layer address is found, we'll need to see if it's 3593 * already present in the address list, or allocate is as well. 3594 * When this block finishes, the link layer address will be on the 3595 * list. 3596 */ 3597 if (llsa != NULL) { 3598 ll_ifma = if_findmulti(ifp, llsa); 3599 if (ll_ifma == NULL) { 3600 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3601 if (ll_ifma == NULL) { 3602 --ifma->ifma_refcount; 3603 if_freemulti(ifma); 3604 error = ENOMEM; 3605 goto free_llsa_out; 3606 } 3607 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3608 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3609 ifma_link); 3610 } else 3611 ll_ifma->ifma_refcount++; 3612 ifma->ifma_llifma = ll_ifma; 3613 } 3614 3615 /* 3616 * We now have a new multicast address, ifma, and possibly a new or 3617 * referenced link layer address. Add the primary address to the 3618 * ifnet address list. 3619 */ 3620 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3621 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3622 3623 if (retifma != NULL) 3624 *retifma = ifma; 3625 3626 /* 3627 * Must generate the message while holding the lock so that 'ifma' 3628 * pointer is still valid. 3629 */ 3630 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3631 IF_ADDR_WUNLOCK(ifp); 3632 3633 /* 3634 * We are certain we have added something, so call down to the 3635 * interface to let them know about it. 3636 */ 3637 if (ifp->if_ioctl != NULL) { 3638 if (THREAD_CAN_SLEEP()) 3639 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3640 else 3641 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3642 } 3643 3644 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3645 link_free_sdl(llsa); 3646 3647 return (0); 3648 3649 free_llsa_out: 3650 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3651 link_free_sdl(llsa); 3652 3653 unlock_out: 3654 IF_ADDR_WUNLOCK(ifp); 3655 return (error); 3656 } 3657 3658 static void 3659 if_siocaddmulti(void *arg, int pending) 3660 { 3661 struct ifnet *ifp; 3662 3663 ifp = arg; 3664 #ifdef DIAGNOSTIC 3665 if (pending > 1) 3666 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3667 #endif 3668 CURVNET_SET(ifp->if_vnet); 3669 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3670 CURVNET_RESTORE(); 3671 } 3672 3673 /* 3674 * Delete a multicast group membership by network-layer group address. 3675 * 3676 * Returns ENOENT if the entry could not be found. If ifp no longer 3677 * exists, results are undefined. This entry point should only be used 3678 * from subsystems which do appropriate locking to hold ifp for the 3679 * duration of the call. 3680 * Network-layer protocol domains must use if_delmulti_ifma(). 3681 */ 3682 int 3683 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3684 { 3685 struct ifmultiaddr *ifma; 3686 int lastref; 3687 3688 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3689 3690 IF_ADDR_WLOCK(ifp); 3691 lastref = 0; 3692 ifma = if_findmulti(ifp, sa); 3693 if (ifma != NULL) 3694 lastref = if_delmulti_locked(ifp, ifma, 0); 3695 IF_ADDR_WUNLOCK(ifp); 3696 3697 if (ifma == NULL) 3698 return (ENOENT); 3699 3700 if (lastref && ifp->if_ioctl != NULL) { 3701 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3702 } 3703 3704 return (0); 3705 } 3706 3707 /* 3708 * Delete all multicast group membership for an interface. 3709 * Should be used to quickly flush all multicast filters. 3710 */ 3711 void 3712 if_delallmulti(struct ifnet *ifp) 3713 { 3714 struct ifmultiaddr *ifma; 3715 struct ifmultiaddr *next; 3716 3717 IF_ADDR_WLOCK(ifp); 3718 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3719 if_delmulti_locked(ifp, ifma, 0); 3720 IF_ADDR_WUNLOCK(ifp); 3721 } 3722 3723 void 3724 if_delmulti_ifma(struct ifmultiaddr *ifma) 3725 { 3726 if_delmulti_ifma_flags(ifma, 0); 3727 } 3728 3729 /* 3730 * Delete a multicast group membership by group membership pointer. 3731 * Network-layer protocol domains must use this routine. 3732 * 3733 * It is safe to call this routine if the ifp disappeared. 3734 */ 3735 void 3736 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3737 { 3738 struct ifnet *ifp; 3739 int lastref; 3740 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3741 #ifdef INET 3742 IN_MULTI_LIST_UNLOCK_ASSERT(); 3743 #endif 3744 ifp = ifma->ifma_ifp; 3745 #ifdef DIAGNOSTIC 3746 if (ifp == NULL) { 3747 printf("%s: ifma_ifp seems to be detached\n", __func__); 3748 } else { 3749 struct epoch_tracker et; 3750 struct ifnet *oifp; 3751 3752 NET_EPOCH_ENTER(et); 3753 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3754 if (ifp == oifp) 3755 break; 3756 NET_EPOCH_EXIT(et); 3757 if (ifp != oifp) 3758 ifp = NULL; 3759 } 3760 #endif 3761 /* 3762 * If and only if the ifnet instance exists: Acquire the address lock. 3763 */ 3764 if (ifp != NULL) 3765 IF_ADDR_WLOCK(ifp); 3766 3767 lastref = if_delmulti_locked(ifp, ifma, flags); 3768 3769 if (ifp != NULL) { 3770 /* 3771 * If and only if the ifnet instance exists: 3772 * Release the address lock. 3773 * If the group was left: update the hardware hash filter. 3774 */ 3775 IF_ADDR_WUNLOCK(ifp); 3776 if (lastref && ifp->if_ioctl != NULL) { 3777 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3778 } 3779 } 3780 } 3781 3782 /* 3783 * Perform deletion of network-layer and/or link-layer multicast address. 3784 * 3785 * Return 0 if the reference count was decremented. 3786 * Return 1 if the final reference was released, indicating that the 3787 * hardware hash filter should be reprogrammed. 3788 */ 3789 static int 3790 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3791 { 3792 struct ifmultiaddr *ll_ifma; 3793 3794 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3795 KASSERT(ifma->ifma_ifp == ifp, 3796 ("%s: inconsistent ifp %p", __func__, ifp)); 3797 IF_ADDR_WLOCK_ASSERT(ifp); 3798 } 3799 3800 ifp = ifma->ifma_ifp; 3801 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3802 3803 /* 3804 * If the ifnet is detaching, null out references to ifnet, 3805 * so that upper protocol layers will notice, and not attempt 3806 * to obtain locks for an ifnet which no longer exists. The 3807 * routing socket announcement must happen before the ifnet 3808 * instance is detached from the system. 3809 */ 3810 if (detaching) { 3811 #ifdef DIAGNOSTIC 3812 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3813 #endif 3814 /* 3815 * ifp may already be nulled out if we are being reentered 3816 * to delete the ll_ifma. 3817 */ 3818 if (ifp != NULL) { 3819 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3820 ifma->ifma_ifp = NULL; 3821 } 3822 } 3823 3824 if (--ifma->ifma_refcount > 0) 3825 return 0; 3826 3827 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3828 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3829 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3830 } 3831 /* 3832 * If this ifma is a network-layer ifma, a link-layer ifma may 3833 * have been associated with it. Release it first if so. 3834 */ 3835 ll_ifma = ifma->ifma_llifma; 3836 if (ll_ifma != NULL) { 3837 KASSERT(ifma->ifma_lladdr != NULL, 3838 ("%s: llifma w/o lladdr", __func__)); 3839 if (detaching) 3840 ll_ifma->ifma_ifp = NULL; /* XXX */ 3841 if (--ll_ifma->ifma_refcount == 0) { 3842 if (ifp != NULL) { 3843 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3844 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3845 ifma_link); 3846 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3847 } 3848 } 3849 if_freemulti(ll_ifma); 3850 } 3851 } 3852 #ifdef INVARIANTS 3853 if (ifp) { 3854 struct ifmultiaddr *ifmatmp; 3855 3856 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3857 MPASS(ifma != ifmatmp); 3858 } 3859 #endif 3860 if_freemulti(ifma); 3861 /* 3862 * The last reference to this instance of struct ifmultiaddr 3863 * was released; the hardware should be notified of this change. 3864 */ 3865 return 1; 3866 } 3867 3868 /* 3869 * Set the link layer address on an interface. 3870 * 3871 * At this time we only support certain types of interfaces, 3872 * and we don't allow the length of the address to change. 3873 * 3874 * Set noinline to be dtrace-friendly 3875 */ 3876 __noinline int 3877 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3878 { 3879 struct sockaddr_dl *sdl; 3880 struct ifaddr *ifa; 3881 struct ifreq ifr; 3882 3883 ifa = ifp->if_addr; 3884 if (ifa == NULL) 3885 return (EINVAL); 3886 3887 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3888 if (sdl == NULL) 3889 return (EINVAL); 3890 3891 if (len != sdl->sdl_alen) /* don't allow length to change */ 3892 return (EINVAL); 3893 3894 switch (ifp->if_type) { 3895 case IFT_ETHER: 3896 case IFT_XETHER: 3897 case IFT_L2VLAN: 3898 case IFT_BRIDGE: 3899 case IFT_IEEE8023ADLAG: 3900 bcopy(lladdr, LLADDR(sdl), len); 3901 break; 3902 default: 3903 return (ENODEV); 3904 } 3905 3906 /* 3907 * If the interface is already up, we need 3908 * to re-init it in order to reprogram its 3909 * address filter. 3910 */ 3911 if ((ifp->if_flags & IFF_UP) != 0) { 3912 if (ifp->if_ioctl) { 3913 ifp->if_flags &= ~IFF_UP; 3914 ifr.ifr_flags = ifp->if_flags & 0xffff; 3915 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3916 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3917 ifp->if_flags |= IFF_UP; 3918 ifr.ifr_flags = ifp->if_flags & 0xffff; 3919 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3920 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3921 } 3922 } 3923 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3924 3925 return (0); 3926 } 3927 3928 /* 3929 * Compat function for handling basic encapsulation requests. 3930 * Not converted stacks (FDDI, IB, ..) supports traditional 3931 * output model: ARP (and other similar L2 protocols) are handled 3932 * inside output routine, arpresolve/nd6_resolve() returns MAC 3933 * address instead of full prepend. 3934 * 3935 * This function creates calculated header==MAC for IPv4/IPv6 and 3936 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3937 * address families. 3938 */ 3939 static int 3940 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3941 { 3942 if (req->rtype != IFENCAP_LL) 3943 return (EOPNOTSUPP); 3944 3945 if (req->bufsize < req->lladdr_len) 3946 return (ENOMEM); 3947 3948 switch (req->family) { 3949 case AF_INET: 3950 case AF_INET6: 3951 break; 3952 default: 3953 return (EAFNOSUPPORT); 3954 } 3955 3956 /* Copy lladdr to storage as is */ 3957 memmove(req->buf, req->lladdr, req->lladdr_len); 3958 req->bufsize = req->lladdr_len; 3959 req->lladdr_off = 0; 3960 3961 return (0); 3962 } 3963 3964 /* 3965 * Tunnel interfaces can nest, also they may cause infinite recursion 3966 * calls when misconfigured. We'll prevent this by detecting loops. 3967 * High nesting level may cause stack exhaustion. We'll prevent this 3968 * by introducing upper limit. 3969 * 3970 * Return 0, if tunnel nesting count is equal or less than limit. 3971 */ 3972 int 3973 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3974 int limit) 3975 { 3976 struct m_tag *mtag; 3977 int count; 3978 3979 count = 1; 3980 mtag = NULL; 3981 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3982 if (*(struct ifnet **)(mtag + 1) == ifp) { 3983 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3984 return (EIO); 3985 } 3986 count++; 3987 } 3988 if (count > limit) { 3989 log(LOG_NOTICE, 3990 "%s: if_output recursively called too many times(%d)\n", 3991 if_name(ifp), count); 3992 return (EIO); 3993 } 3994 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3995 if (mtag == NULL) 3996 return (ENOMEM); 3997 *(struct ifnet **)(mtag + 1) = ifp; 3998 m_tag_prepend(m, mtag); 3999 return (0); 4000 } 4001 4002 /* 4003 * Get the link layer address that was read from the hardware at attach. 4004 * 4005 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 4006 * their component interfaces as IFT_IEEE8023ADLAG. 4007 */ 4008 int 4009 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 4010 { 4011 if (ifp->if_hw_addr == NULL) 4012 return (ENODEV); 4013 4014 switch (ifp->if_type) { 4015 case IFT_ETHER: 4016 case IFT_IEEE8023ADLAG: 4017 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 4018 return (0); 4019 default: 4020 return (ENODEV); 4021 } 4022 } 4023 4024 /* 4025 * The name argument must be a pointer to storage which will last as 4026 * long as the interface does. For physical devices, the result of 4027 * device_get_name(dev) is a good choice and for pseudo-devices a 4028 * static string works well. 4029 */ 4030 void 4031 if_initname(struct ifnet *ifp, const char *name, int unit) 4032 { 4033 ifp->if_dname = name; 4034 ifp->if_dunit = unit; 4035 if (unit != IF_DUNIT_NONE) 4036 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 4037 else 4038 strlcpy(ifp->if_xname, name, IFNAMSIZ); 4039 } 4040 4041 static int 4042 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 4043 { 4044 char if_fmt[256]; 4045 4046 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 4047 vlog(pri, if_fmt, ap); 4048 return (0); 4049 } 4050 4051 4052 int 4053 if_printf(struct ifnet *ifp, const char *fmt, ...) 4054 { 4055 va_list ap; 4056 4057 va_start(ap, fmt); 4058 if_vlog(ifp, LOG_INFO, fmt, ap); 4059 va_end(ap); 4060 return (0); 4061 } 4062 4063 int 4064 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 4065 { 4066 va_list ap; 4067 4068 va_start(ap, fmt); 4069 if_vlog(ifp, pri, fmt, ap); 4070 va_end(ap); 4071 return (0); 4072 } 4073 4074 void 4075 if_start(struct ifnet *ifp) 4076 { 4077 4078 (*(ifp)->if_start)(ifp); 4079 } 4080 4081 /* 4082 * Backwards compatibility interface for drivers 4083 * that have not implemented it 4084 */ 4085 static int 4086 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 4087 { 4088 int error; 4089 4090 IFQ_HANDOFF(ifp, m, error); 4091 return (error); 4092 } 4093 4094 static void 4095 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4096 { 4097 m_freem(m); 4098 } 4099 4100 int 4101 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4102 { 4103 int active = 0; 4104 4105 IF_LOCK(ifq); 4106 if (_IF_QFULL(ifq)) { 4107 IF_UNLOCK(ifq); 4108 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4109 m_freem(m); 4110 return (0); 4111 } 4112 if (ifp != NULL) { 4113 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4114 if (m->m_flags & (M_BCAST|M_MCAST)) 4115 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4116 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4117 } 4118 _IF_ENQUEUE(ifq, m); 4119 IF_UNLOCK(ifq); 4120 if (ifp != NULL && !active) 4121 (*(ifp)->if_start)(ifp); 4122 return (1); 4123 } 4124 4125 void 4126 if_register_com_alloc(u_char type, 4127 if_com_alloc_t *a, if_com_free_t *f) 4128 { 4129 4130 KASSERT(if_com_alloc[type] == NULL, 4131 ("if_register_com_alloc: %d already registered", type)); 4132 KASSERT(if_com_free[type] == NULL, 4133 ("if_register_com_alloc: %d free already registered", type)); 4134 4135 if_com_alloc[type] = a; 4136 if_com_free[type] = f; 4137 } 4138 4139 void 4140 if_deregister_com_alloc(u_char type) 4141 { 4142 4143 KASSERT(if_com_alloc[type] != NULL, 4144 ("if_deregister_com_alloc: %d not registered", type)); 4145 KASSERT(if_com_free[type] != NULL, 4146 ("if_deregister_com_alloc: %d free not registered", type)); 4147 4148 /* 4149 * Ensure all pending EPOCH(9) callbacks have been executed. This 4150 * fixes issues about late invocation of if_destroy(), which leads 4151 * to memory leak from if_com_alloc[type] allocated if_l2com. 4152 */ 4153 NET_EPOCH_DRAIN_CALLBACKS(); 4154 4155 if_com_alloc[type] = NULL; 4156 if_com_free[type] = NULL; 4157 } 4158 4159 /* API for driver access to network stack owned ifnet.*/ 4160 uint64_t 4161 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4162 { 4163 uint64_t oldbrate; 4164 4165 oldbrate = ifp->if_baudrate; 4166 ifp->if_baudrate = baudrate; 4167 return (oldbrate); 4168 } 4169 4170 uint64_t 4171 if_getbaudrate(const if_t ifp) 4172 { 4173 return (ifp->if_baudrate); 4174 } 4175 4176 int 4177 if_setcapabilities(if_t ifp, int capabilities) 4178 { 4179 ifp->if_capabilities = capabilities; 4180 return (0); 4181 } 4182 4183 int 4184 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4185 { 4186 ifp->if_capabilities &= ~clearbit; 4187 ifp->if_capabilities |= setbit; 4188 return (0); 4189 } 4190 4191 int 4192 if_getcapabilities(const if_t ifp) 4193 { 4194 return (ifp->if_capabilities); 4195 } 4196 4197 int 4198 if_setcapenable(if_t ifp, int capabilities) 4199 { 4200 ifp->if_capenable = capabilities; 4201 return (0); 4202 } 4203 4204 int 4205 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4206 { 4207 ifp->if_capenable &= ~clearcap; 4208 ifp->if_capenable |= setcap; 4209 return (0); 4210 } 4211 4212 int 4213 if_setcapabilities2(if_t ifp, int capabilities) 4214 { 4215 ifp->if_capabilities2 = capabilities; 4216 return (0); 4217 } 4218 4219 int 4220 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit) 4221 { 4222 ifp->if_capabilities2 &= ~clearbit; 4223 ifp->if_capabilities2 |= setbit; 4224 return (0); 4225 } 4226 4227 int 4228 if_getcapabilities2(const if_t ifp) 4229 { 4230 return (ifp->if_capabilities2); 4231 } 4232 4233 int 4234 if_setcapenable2(if_t ifp, int capabilities2) 4235 { 4236 ifp->if_capenable2 = capabilities2; 4237 return (0); 4238 } 4239 4240 int 4241 if_setcapenable2bit(if_t ifp, int setcap, int clearcap) 4242 { 4243 ifp->if_capenable2 &= ~clearcap; 4244 ifp->if_capenable2 |= setcap; 4245 return (0); 4246 } 4247 4248 const char * 4249 if_getdname(const if_t ifp) 4250 { 4251 return (ifp->if_dname); 4252 } 4253 4254 void 4255 if_setdname(if_t ifp, const char *dname) 4256 { 4257 ifp->if_dname = dname; 4258 } 4259 4260 const char * 4261 if_name(if_t ifp) 4262 { 4263 return (ifp->if_xname); 4264 } 4265 4266 int 4267 if_setname(if_t ifp, const char *name) 4268 { 4269 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4270 return (ENAMETOOLONG); 4271 strcpy(ifp->if_xname, name); 4272 4273 return (0); 4274 } 4275 4276 int 4277 if_togglecapenable(if_t ifp, int togglecap) 4278 { 4279 ifp->if_capenable ^= togglecap; 4280 return (0); 4281 } 4282 4283 int 4284 if_getcapenable(const if_t ifp) 4285 { 4286 return (ifp->if_capenable); 4287 } 4288 4289 int 4290 if_togglecapenable2(if_t ifp, int togglecap) 4291 { 4292 ifp->if_capenable2 ^= togglecap; 4293 return (0); 4294 } 4295 4296 int 4297 if_getcapenable2(const if_t ifp) 4298 { 4299 return (ifp->if_capenable2); 4300 } 4301 4302 int 4303 if_getdunit(const if_t ifp) 4304 { 4305 return (ifp->if_dunit); 4306 } 4307 4308 int 4309 if_getindex(const if_t ifp) 4310 { 4311 return (ifp->if_index); 4312 } 4313 4314 int 4315 if_getidxgen(const if_t ifp) 4316 { 4317 return (ifp->if_idxgen); 4318 } 4319 4320 const char * 4321 if_getdescr(if_t ifp) 4322 { 4323 return (ifp->if_description); 4324 } 4325 4326 void 4327 if_setdescr(if_t ifp, char *descrbuf) 4328 { 4329 sx_xlock(&ifdescr_sx); 4330 char *odescrbuf = ifp->if_description; 4331 ifp->if_description = descrbuf; 4332 sx_xunlock(&ifdescr_sx); 4333 4334 if_freedescr(odescrbuf); 4335 } 4336 4337 char * 4338 if_allocdescr(size_t sz, int malloc_flag) 4339 { 4340 malloc_flag &= (M_WAITOK | M_NOWAIT); 4341 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4342 } 4343 4344 void 4345 if_freedescr(char *descrbuf) 4346 { 4347 free(descrbuf, M_IFDESCR); 4348 } 4349 4350 int 4351 if_getalloctype(const if_t ifp) 4352 { 4353 return (ifp->if_alloctype); 4354 } 4355 4356 void 4357 if_setlastchange(if_t ifp) 4358 { 4359 getmicrotime(&ifp->if_lastchange); 4360 } 4361 4362 /* 4363 * This is largely undesirable because it ties ifnet to a device, but does 4364 * provide flexiblity for an embedded product vendor. Should be used with 4365 * the understanding that it violates the interface boundaries, and should be 4366 * a last resort only. 4367 */ 4368 int 4369 if_setdev(if_t ifp, void *dev) 4370 { 4371 return (0); 4372 } 4373 4374 int 4375 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4376 { 4377 ifp->if_drv_flags &= ~clear_flags; 4378 ifp->if_drv_flags |= set_flags; 4379 4380 return (0); 4381 } 4382 4383 int 4384 if_getdrvflags(const if_t ifp) 4385 { 4386 return (ifp->if_drv_flags); 4387 } 4388 4389 int 4390 if_setdrvflags(if_t ifp, int flags) 4391 { 4392 ifp->if_drv_flags = flags; 4393 return (0); 4394 } 4395 4396 int 4397 if_setflags(if_t ifp, int flags) 4398 { 4399 ifp->if_flags = flags; 4400 return (0); 4401 } 4402 4403 int 4404 if_setflagbits(if_t ifp, int set, int clear) 4405 { 4406 ifp->if_flags &= ~clear; 4407 ifp->if_flags |= set; 4408 return (0); 4409 } 4410 4411 int 4412 if_getflags(const if_t ifp) 4413 { 4414 return (ifp->if_flags); 4415 } 4416 4417 int 4418 if_clearhwassist(if_t ifp) 4419 { 4420 ifp->if_hwassist = 0; 4421 return (0); 4422 } 4423 4424 int 4425 if_sethwassistbits(if_t ifp, int toset, int toclear) 4426 { 4427 ifp->if_hwassist &= ~toclear; 4428 ifp->if_hwassist |= toset; 4429 4430 return (0); 4431 } 4432 4433 int 4434 if_sethwassist(if_t ifp, int hwassist_bit) 4435 { 4436 ifp->if_hwassist = hwassist_bit; 4437 return (0); 4438 } 4439 4440 int 4441 if_gethwassist(const if_t ifp) 4442 { 4443 return (ifp->if_hwassist); 4444 } 4445 4446 int 4447 if_togglehwassist(if_t ifp, int toggle_bits) 4448 { 4449 ifp->if_hwassist ^= toggle_bits; 4450 return (0); 4451 } 4452 4453 int 4454 if_setmtu(if_t ifp, int mtu) 4455 { 4456 ifp->if_mtu = mtu; 4457 return (0); 4458 } 4459 4460 void 4461 if_notifymtu(if_t ifp) 4462 { 4463 #ifdef INET6 4464 nd6_setmtu(ifp); 4465 #endif 4466 rt_updatemtu(ifp); 4467 } 4468 4469 int 4470 if_getmtu(const if_t ifp) 4471 { 4472 return (ifp->if_mtu); 4473 } 4474 4475 int 4476 if_getmtu_family(const if_t ifp, int family) 4477 { 4478 struct domain *dp; 4479 4480 SLIST_FOREACH(dp, &domains, dom_next) { 4481 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4482 return (dp->dom_ifmtu(ifp)); 4483 } 4484 4485 return (ifp->if_mtu); 4486 } 4487 4488 /* 4489 * Methods for drivers to access interface unicast and multicast 4490 * link level addresses. Driver shall not know 'struct ifaddr' neither 4491 * 'struct ifmultiaddr'. 4492 */ 4493 u_int 4494 if_lladdr_count(if_t ifp) 4495 { 4496 struct epoch_tracker et; 4497 struct ifaddr *ifa; 4498 u_int count; 4499 4500 count = 0; 4501 NET_EPOCH_ENTER(et); 4502 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4503 if (ifa->ifa_addr->sa_family == AF_LINK) 4504 count++; 4505 NET_EPOCH_EXIT(et); 4506 4507 return (count); 4508 } 4509 4510 int 4511 if_foreach(if_foreach_cb_t cb, void *cb_arg) 4512 { 4513 if_t ifp; 4514 int error; 4515 4516 NET_EPOCH_ASSERT(); 4517 MPASS(cb); 4518 4519 error = 0; 4520 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4521 error = cb(ifp, cb_arg); 4522 if (error != 0) 4523 break; 4524 } 4525 4526 return (error); 4527 } 4528 4529 /* 4530 * Iterates over the list of interfaces, permitting callback function @cb to sleep. 4531 * Stops iteration if @cb returns non-zero error code. 4532 * Returns the last error code from @cb. 4533 * @match_cb: optional match callback limiting the iteration to only matched interfaces 4534 * @match_arg: argument to pass to @match_cb 4535 * @cb: iteration callback 4536 * @cb_arg: argument to pass to @cb 4537 */ 4538 int 4539 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb, 4540 void *cb_arg) 4541 { 4542 int match_count = 0, array_size = 16; /* 128 bytes for malloc */ 4543 struct ifnet **match_array = NULL; 4544 int error = 0; 4545 4546 MPASS(cb); 4547 4548 while (true) { 4549 struct ifnet **new_array; 4550 int new_size = array_size; 4551 struct epoch_tracker et; 4552 struct ifnet *ifp; 4553 4554 while (new_size < match_count) 4555 new_size *= 2; 4556 new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK); 4557 if (match_array != NULL) 4558 memcpy(new_array, match_array, array_size * sizeof(void *)); 4559 free(match_array, M_TEMP); 4560 match_array = new_array; 4561 array_size = new_size; 4562 4563 match_count = 0; 4564 NET_EPOCH_ENTER(et); 4565 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4566 if (match_cb != NULL && !match_cb(ifp, match_arg)) 4567 continue; 4568 if (match_count < array_size) { 4569 if (if_try_ref(ifp)) 4570 match_array[match_count++] = ifp; 4571 } else 4572 match_count++; 4573 } 4574 NET_EPOCH_EXIT(et); 4575 4576 if (match_count > array_size) { 4577 for (int i = 0; i < array_size; i++) 4578 if_rele(match_array[i]); 4579 continue; 4580 } else { 4581 for (int i = 0; i < match_count; i++) { 4582 if (error == 0) 4583 error = cb(match_array[i], cb_arg); 4584 if_rele(match_array[i]); 4585 } 4586 free(match_array, M_TEMP); 4587 break; 4588 } 4589 } 4590 4591 return (error); 4592 } 4593 4594 4595 /* 4596 * Uses just 1 pointer of the 4 available in the public struct. 4597 */ 4598 if_t 4599 if_iter_start(struct if_iter *iter) 4600 { 4601 if_t ifp; 4602 4603 NET_EPOCH_ASSERT(); 4604 4605 bzero(iter, sizeof(*iter)); 4606 ifp = CK_STAILQ_FIRST(&V_ifnet); 4607 if (ifp != NULL) 4608 iter->context[0] = CK_STAILQ_NEXT(ifp, if_link); 4609 else 4610 iter->context[0] = NULL; 4611 return (ifp); 4612 } 4613 4614 if_t 4615 if_iter_next(struct if_iter *iter) 4616 { 4617 if_t cur_ifp = iter->context[0]; 4618 4619 if (cur_ifp != NULL) 4620 iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link); 4621 return (cur_ifp); 4622 } 4623 4624 void 4625 if_iter_finish(struct if_iter *iter) 4626 { 4627 /* Nothing to do here for now. */ 4628 } 4629 4630 u_int 4631 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4632 { 4633 struct epoch_tracker et; 4634 struct ifaddr *ifa; 4635 u_int count; 4636 4637 MPASS(cb); 4638 4639 count = 0; 4640 NET_EPOCH_ENTER(et); 4641 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4642 if (ifa->ifa_addr->sa_family != AF_LINK) 4643 continue; 4644 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4645 count); 4646 } 4647 NET_EPOCH_EXIT(et); 4648 4649 return (count); 4650 } 4651 4652 u_int 4653 if_llmaddr_count(if_t ifp) 4654 { 4655 struct epoch_tracker et; 4656 struct ifmultiaddr *ifma; 4657 int count; 4658 4659 count = 0; 4660 NET_EPOCH_ENTER(et); 4661 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4662 if (ifma->ifma_addr->sa_family == AF_LINK) 4663 count++; 4664 NET_EPOCH_EXIT(et); 4665 4666 return (count); 4667 } 4668 4669 bool 4670 if_maddr_empty(if_t ifp) 4671 { 4672 4673 return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs)); 4674 } 4675 4676 u_int 4677 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4678 { 4679 struct epoch_tracker et; 4680 struct ifmultiaddr *ifma; 4681 u_int count; 4682 4683 MPASS(cb); 4684 4685 count = 0; 4686 NET_EPOCH_ENTER(et); 4687 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4688 if (ifma->ifma_addr->sa_family != AF_LINK) 4689 continue; 4690 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4691 count); 4692 } 4693 NET_EPOCH_EXIT(et); 4694 4695 return (count); 4696 } 4697 4698 u_int 4699 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4700 { 4701 struct epoch_tracker et; 4702 struct ifaddr *ifa; 4703 u_int count; 4704 4705 MPASS(cb); 4706 4707 count = 0; 4708 NET_EPOCH_ENTER(et); 4709 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4710 if (ifa->ifa_addr->sa_family != type) 4711 continue; 4712 count += (*cb)(cb_arg, ifa, count); 4713 } 4714 NET_EPOCH_EXIT(et); 4715 4716 return (count); 4717 } 4718 4719 struct ifaddr * 4720 ifa_iter_start(if_t ifp, struct ifa_iter *iter) 4721 { 4722 struct ifaddr *ifa; 4723 4724 NET_EPOCH_ASSERT(); 4725 4726 bzero(iter, sizeof(*iter)); 4727 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 4728 if (ifa != NULL) 4729 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4730 else 4731 iter->context[0] = NULL; 4732 return (ifa); 4733 } 4734 4735 struct ifaddr * 4736 ifa_iter_next(struct ifa_iter *iter) 4737 { 4738 struct ifaddr *ifa = iter->context[0]; 4739 4740 if (ifa != NULL) 4741 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4742 return (ifa); 4743 } 4744 4745 void 4746 ifa_iter_finish(struct ifa_iter *iter) 4747 { 4748 /* Nothing to do here for now. */ 4749 } 4750 4751 int 4752 if_setsoftc(if_t ifp, void *softc) 4753 { 4754 ifp->if_softc = softc; 4755 return (0); 4756 } 4757 4758 void * 4759 if_getsoftc(const if_t ifp) 4760 { 4761 return (ifp->if_softc); 4762 } 4763 4764 void 4765 if_setrcvif(struct mbuf *m, if_t ifp) 4766 { 4767 4768 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4769 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4770 } 4771 4772 void 4773 if_setvtag(struct mbuf *m, uint16_t tag) 4774 { 4775 m->m_pkthdr.ether_vtag = tag; 4776 } 4777 4778 uint16_t 4779 if_getvtag(struct mbuf *m) 4780 { 4781 return (m->m_pkthdr.ether_vtag); 4782 } 4783 4784 int 4785 if_sendq_empty(if_t ifp) 4786 { 4787 return (IFQ_DRV_IS_EMPTY(&ifp->if_snd)); 4788 } 4789 4790 struct ifaddr * 4791 if_getifaddr(const if_t ifp) 4792 { 4793 return (ifp->if_addr); 4794 } 4795 4796 int 4797 if_setsendqready(if_t ifp) 4798 { 4799 IFQ_SET_READY(&ifp->if_snd); 4800 return (0); 4801 } 4802 4803 int 4804 if_setsendqlen(if_t ifp, int tx_desc_count) 4805 { 4806 IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count); 4807 ifp->if_snd.ifq_drv_maxlen = tx_desc_count; 4808 return (0); 4809 } 4810 4811 void 4812 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na) 4813 { 4814 ifp->if_netmap = na; 4815 } 4816 4817 struct netmap_adapter * 4818 if_getnetmapadapter(if_t ifp) 4819 { 4820 return (ifp->if_netmap); 4821 } 4822 4823 int 4824 if_vlantrunkinuse(if_t ifp) 4825 { 4826 return (ifp->if_vlantrunk != NULL); 4827 } 4828 4829 void 4830 if_init(if_t ifp, void *ctx) 4831 { 4832 (*ifp->if_init)(ctx); 4833 } 4834 4835 void 4836 if_input(if_t ifp, struct mbuf* sendmp) 4837 { 4838 (*ifp->if_input)(ifp, sendmp); 4839 } 4840 4841 int 4842 if_transmit(if_t ifp, struct mbuf *m) 4843 { 4844 return ((*ifp->if_transmit)(ifp, m)); 4845 } 4846 4847 int 4848 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst) 4849 { 4850 if (ifp->if_resolvemulti == NULL) 4851 return (EOPNOTSUPP); 4852 4853 return (ifp->if_resolvemulti(ifp, srcs, dst)); 4854 } 4855 4856 int 4857 if_ioctl(if_t ifp, u_long cmd, void *data) 4858 { 4859 if (ifp->if_ioctl == NULL) 4860 return (EOPNOTSUPP); 4861 4862 return (ifp->if_ioctl(ifp, cmd, data)); 4863 } 4864 4865 struct mbuf * 4866 if_dequeue(if_t ifp) 4867 { 4868 struct mbuf *m; 4869 4870 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 4871 return (m); 4872 } 4873 4874 int 4875 if_sendq_prepend(if_t ifp, struct mbuf *m) 4876 { 4877 IFQ_DRV_PREPEND(&ifp->if_snd, m); 4878 return (0); 4879 } 4880 4881 int 4882 if_setifheaderlen(if_t ifp, int len) 4883 { 4884 ifp->if_hdrlen = len; 4885 return (0); 4886 } 4887 4888 caddr_t 4889 if_getlladdr(const if_t ifp) 4890 { 4891 return (IF_LLADDR(ifp)); 4892 } 4893 4894 void * 4895 if_gethandle(u_char type) 4896 { 4897 return (if_alloc(type)); 4898 } 4899 4900 void 4901 if_vlancap(if_t ifp) 4902 { 4903 VLAN_CAPABILITIES(ifp); 4904 } 4905 4906 int 4907 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4908 { 4909 ifp->if_hw_tsomax = if_hw_tsomax; 4910 return (0); 4911 } 4912 4913 int 4914 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4915 { 4916 ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4917 return (0); 4918 } 4919 4920 int 4921 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4922 { 4923 ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4924 return (0); 4925 } 4926 4927 u_int 4928 if_gethwtsomax(const if_t ifp) 4929 { 4930 return (ifp->if_hw_tsomax); 4931 } 4932 4933 u_int 4934 if_gethwtsomaxsegcount(const if_t ifp) 4935 { 4936 return (ifp->if_hw_tsomaxsegcount); 4937 } 4938 4939 u_int 4940 if_gethwtsomaxsegsize(const if_t ifp) 4941 { 4942 return (ifp->if_hw_tsomaxsegsize); 4943 } 4944 4945 void 4946 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4947 { 4948 ifp->if_init = init_fn; 4949 } 4950 4951 void 4952 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4953 { 4954 ifp->if_input = input_fn; 4955 } 4956 4957 if_input_fn_t 4958 if_getinputfn(if_t ifp) 4959 { 4960 return (ifp->if_input); 4961 } 4962 4963 void 4964 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4965 { 4966 ifp->if_ioctl = ioctl_fn; 4967 } 4968 4969 void 4970 if_setoutputfn(if_t ifp, if_output_fn_t output_fn) 4971 { 4972 ifp->if_output = output_fn; 4973 } 4974 4975 void 4976 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4977 { 4978 ifp->if_start = start_fn; 4979 } 4980 4981 if_start_fn_t 4982 if_getstartfn(if_t ifp) 4983 { 4984 return (ifp->if_start); 4985 } 4986 4987 void 4988 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4989 { 4990 ifp->if_transmit = start_fn; 4991 } 4992 4993 if_transmit_fn_t 4994 if_gettransmitfn(if_t ifp) 4995 { 4996 return (ifp->if_transmit); 4997 } 4998 4999 void 5000 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 5001 { 5002 ifp->if_qflush = flush_fn; 5003 } 5004 5005 void 5006 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 5007 { 5008 ifp->if_snd_tag_alloc = alloc_fn; 5009 } 5010 5011 int 5012 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params, 5013 struct m_snd_tag **mstp) 5014 { 5015 if (ifp->if_snd_tag_alloc == NULL) 5016 return (EOPNOTSUPP); 5017 return (ifp->if_snd_tag_alloc(ifp, params, mstp)); 5018 } 5019 5020 void 5021 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 5022 { 5023 ifp->if_get_counter = fn; 5024 } 5025 5026 void 5027 if_setreassignfn(if_t ifp, if_reassign_fn_t fn) 5028 { 5029 ifp->if_reassign = fn; 5030 } 5031 5032 void 5033 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn) 5034 { 5035 ifp->if_ratelimit_query = fn; 5036 } 5037 5038 void 5039 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m) 5040 { 5041 ifp->if_debugnet_methods = m; 5042 } 5043 5044 struct label * 5045 if_getmaclabel(if_t ifp) 5046 { 5047 return (ifp->if_label); 5048 } 5049 5050 void 5051 if_setmaclabel(if_t ifp, struct label *label) 5052 { 5053 ifp->if_label = label; 5054 } 5055 5056 int 5057 if_gettype(if_t ifp) 5058 { 5059 return (ifp->if_type); 5060 } 5061 5062 void * 5063 if_getllsoftc(if_t ifp) 5064 { 5065 return (ifp->if_llsoftc); 5066 } 5067 5068 void 5069 if_setllsoftc(if_t ifp, void *llsoftc) 5070 { 5071 ifp->if_llsoftc = llsoftc; 5072 }; 5073 5074 int 5075 if_getlinkstate(if_t ifp) 5076 { 5077 return (ifp->if_link_state); 5078 } 5079 5080 const uint8_t * 5081 if_getbroadcastaddr(if_t ifp) 5082 { 5083 return (ifp->if_broadcastaddr); 5084 } 5085 5086 void 5087 if_setbroadcastaddr(if_t ifp, const uint8_t *addr) 5088 { 5089 ifp->if_broadcastaddr = addr; 5090 } 5091 5092 int 5093 if_getnumadomain(if_t ifp) 5094 { 5095 return (ifp->if_numa_domain); 5096 } 5097 5098 uint64_t 5099 if_getcounter(if_t ifp, ift_counter counter) 5100 { 5101 return (ifp->if_get_counter(ifp, counter)); 5102 } 5103 5104 bool 5105 if_altq_is_enabled(if_t ifp) 5106 { 5107 return (ALTQ_IS_ENABLED(&ifp->if_snd)); 5108 } 5109 5110 struct vnet * 5111 if_getvnet(if_t ifp) 5112 { 5113 return (ifp->if_vnet); 5114 } 5115 5116 void * 5117 if_getafdata(if_t ifp, int af) 5118 { 5119 return (ifp->if_afdata[af]); 5120 } 5121 5122 u_int 5123 if_getfib(if_t ifp) 5124 { 5125 return (ifp->if_fib); 5126 } 5127 5128 uint8_t 5129 if_getaddrlen(if_t ifp) 5130 { 5131 return (ifp->if_addrlen); 5132 } 5133 5134 struct bpf_if * 5135 if_getbpf(if_t ifp) 5136 { 5137 return (ifp->if_bpf); 5138 } 5139 5140 struct ifvlantrunk * 5141 if_getvlantrunk(if_t ifp) 5142 { 5143 return (ifp->if_vlantrunk); 5144 } 5145 5146 uint8_t 5147 if_getpcp(if_t ifp) 5148 { 5149 return (ifp->if_pcp); 5150 } 5151 5152 void * 5153 if_getl2com(if_t ifp) 5154 { 5155 return (ifp->if_l2com); 5156 } 5157 5158 void 5159 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m) 5160 { 5161 ifp->if_ipsec_accel_m = m; 5162 } 5163 5164 #ifdef DDB 5165 static void 5166 if_show_ifnet(struct ifnet *ifp) 5167 { 5168 if (ifp == NULL) 5169 return; 5170 db_printf("%s:\n", ifp->if_xname); 5171 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 5172 IF_DB_PRINTF("%s", if_dname); 5173 IF_DB_PRINTF("%d", if_dunit); 5174 IF_DB_PRINTF("%s", if_description); 5175 IF_DB_PRINTF("%u", if_index); 5176 IF_DB_PRINTF("%d", if_idxgen); 5177 IF_DB_PRINTF("%u", if_refcount); 5178 IF_DB_PRINTF("%p", if_softc); 5179 IF_DB_PRINTF("%p", if_l2com); 5180 IF_DB_PRINTF("%p", if_llsoftc); 5181 IF_DB_PRINTF("%d", if_amcount); 5182 IF_DB_PRINTF("%p", if_addr); 5183 IF_DB_PRINTF("%p", if_broadcastaddr); 5184 IF_DB_PRINTF("%p", if_afdata); 5185 IF_DB_PRINTF("%d", if_afdata_initialized); 5186 IF_DB_PRINTF("%u", if_fib); 5187 IF_DB_PRINTF("%p", if_vnet); 5188 IF_DB_PRINTF("%p", if_home_vnet); 5189 IF_DB_PRINTF("%p", if_vlantrunk); 5190 IF_DB_PRINTF("%p", if_bpf); 5191 IF_DB_PRINTF("%u", if_pcount); 5192 IF_DB_PRINTF("%p", if_bridge); 5193 IF_DB_PRINTF("%p", if_lagg); 5194 IF_DB_PRINTF("%p", if_pf_kif); 5195 IF_DB_PRINTF("%p", if_carp); 5196 IF_DB_PRINTF("%p", if_label); 5197 IF_DB_PRINTF("%p", if_netmap); 5198 IF_DB_PRINTF("0x%08x", if_flags); 5199 IF_DB_PRINTF("0x%08x", if_drv_flags); 5200 IF_DB_PRINTF("0x%08x", if_capabilities); 5201 IF_DB_PRINTF("0x%08x", if_capenable); 5202 IF_DB_PRINTF("%p", if_snd.ifq_head); 5203 IF_DB_PRINTF("%p", if_snd.ifq_tail); 5204 IF_DB_PRINTF("%d", if_snd.ifq_len); 5205 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 5206 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 5207 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 5208 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 5209 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 5210 IF_DB_PRINTF("%d", if_snd.altq_type); 5211 IF_DB_PRINTF("%x", if_snd.altq_flags); 5212 #undef IF_DB_PRINTF 5213 } 5214 5215 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 5216 { 5217 if (!have_addr) { 5218 db_printf("usage: show ifnet <struct ifnet *>\n"); 5219 return; 5220 } 5221 5222 if_show_ifnet((struct ifnet *)addr); 5223 } 5224 5225 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 5226 { 5227 struct ifnet *ifp; 5228 u_short idx; 5229 5230 for (idx = 1; idx <= if_index; idx++) { 5231 ifp = ifindex_table[idx].ife_ifnet; 5232 if (ifp == NULL) 5233 continue; 5234 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 5235 if (db_pager_quit) 5236 break; 5237 } 5238 } 5239 #endif /* DDB */ 5240