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