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 epoch_wait_preempt(net_epoch_preempt); 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 ifaliasreq ifr; 1041 1042 bzero(&ifr, sizeof(ifr)); 1043 ifr.ifra_addr = *ifa->ifa_addr; 1044 if (ifa->ifa_dstaddr) 1045 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1046 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1047 NULL) == 0) 1048 continue; 1049 } 1050 #endif /* INET */ 1051 #ifdef INET6 1052 if (ifa->ifa_addr->sa_family == AF_INET6) { 1053 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1054 /* ifp_addrhead is already updated */ 1055 continue; 1056 } 1057 #endif /* INET6 */ 1058 IF_ADDR_WLOCK(ifp); 1059 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1060 IF_ADDR_WUNLOCK(ifp); 1061 ifa_free(ifa); 1062 } 1063 } 1064 1065 /* 1066 * Remove any multicast network addresses from an interface when an ifnet 1067 * is going away. 1068 */ 1069 static void 1070 if_purgemaddrs(struct ifnet *ifp) 1071 { 1072 struct ifmultiaddr *ifma; 1073 1074 IF_ADDR_WLOCK(ifp); 1075 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1076 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1077 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1078 if_delmulti_locked(ifp, ifma, 1); 1079 } 1080 IF_ADDR_WUNLOCK(ifp); 1081 } 1082 1083 /* 1084 * Detach an interface, removing it from the list of "active" interfaces. 1085 * If vmove flag is set on entry to if_detach_internal(), perform only a 1086 * limited subset of cleanup tasks, given that we are moving an ifnet from 1087 * one vnet to another, where it must be fully operational. 1088 * 1089 * XXXRW: There are some significant questions about event ordering, and 1090 * how to prevent things from starting to use the interface during detach. 1091 */ 1092 void 1093 if_detach(struct ifnet *ifp) 1094 { 1095 bool found; 1096 1097 CURVNET_SET_QUIET(ifp->if_vnet); 1098 found = if_unlink_ifnet(ifp, false); 1099 if (found) { 1100 sx_xlock(&ifnet_detach_sxlock); 1101 if_detach_internal(ifp, false); 1102 sx_xunlock(&ifnet_detach_sxlock); 1103 } 1104 CURVNET_RESTORE(); 1105 } 1106 1107 /* 1108 * The vmove flag, if set, indicates that we are called from a callpath 1109 * that is moving an interface to a different vnet instance. 1110 * 1111 * The shutdown flag, if set, indicates that we are called in the 1112 * process of shutting down a vnet instance. Currently only the 1113 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1114 * on a vnet instance shutdown without this flag being set, e.g., when 1115 * the cloned interfaces are destoyed as first thing of teardown. 1116 */ 1117 static int 1118 if_detach_internal(struct ifnet *ifp, bool vmove) 1119 { 1120 struct ifaddr *ifa; 1121 int i; 1122 struct domain *dp; 1123 #ifdef VIMAGE 1124 bool shutdown; 1125 1126 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1127 #endif 1128 1129 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 1130 1131 /* 1132 * At this point we know the interface still was on the ifnet list 1133 * and we removed it so we are in a stable state. 1134 */ 1135 epoch_wait_preempt(net_epoch_preempt); 1136 1137 /* 1138 * Ensure all pending EPOCH(9) callbacks have been executed. This 1139 * fixes issues about late destruction of multicast options 1140 * which lead to leave group calls, which in turn access the 1141 * belonging ifnet structure: 1142 */ 1143 NET_EPOCH_DRAIN_CALLBACKS(); 1144 1145 /* 1146 * In any case (destroy or vmove) detach us from the groups 1147 * and remove/wait for pending events on the taskq. 1148 * XXX-BZ in theory an interface could still enqueue a taskq change? 1149 */ 1150 if_delgroups(ifp); 1151 1152 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1153 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1154 1155 if_down(ifp); 1156 1157 #ifdef VIMAGE 1158 /* 1159 * On VNET shutdown abort here as the stack teardown will do all 1160 * the work top-down for us. 1161 */ 1162 if (shutdown) { 1163 /* Give interface users the chance to clean up. */ 1164 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1165 1166 /* 1167 * In case of a vmove we are done here without error. 1168 * If we would signal an error it would lead to the same 1169 * abort as if we did not find the ifnet anymore. 1170 * if_detach() calls us in void context and does not care 1171 * about an early abort notification, so life is splendid :) 1172 */ 1173 goto finish_vnet_shutdown; 1174 } 1175 #endif 1176 1177 /* 1178 * At this point we are not tearing down a VNET and are either 1179 * going to destroy or vmove the interface and have to cleanup 1180 * accordingly. 1181 */ 1182 1183 /* 1184 * Remove routes and flush queues. 1185 */ 1186 #ifdef ALTQ 1187 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1188 altq_disable(&ifp->if_snd); 1189 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1190 altq_detach(&ifp->if_snd); 1191 #endif 1192 1193 if_purgeaddrs(ifp); 1194 1195 #ifdef INET 1196 in_ifdetach(ifp); 1197 #endif 1198 1199 #ifdef INET6 1200 /* 1201 * Remove all IPv6 kernel structs related to ifp. This should be done 1202 * before removing routing entries below, since IPv6 interface direct 1203 * routes are expected to be removed by the IPv6-specific kernel API. 1204 * Otherwise, the kernel will detect some inconsistency and bark it. 1205 */ 1206 in6_ifdetach(ifp); 1207 #endif 1208 if_purgemaddrs(ifp); 1209 1210 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1211 if (IS_DEFAULT_VNET(curvnet)) 1212 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1213 1214 if (!vmove) { 1215 /* 1216 * Prevent further calls into the device driver via ifnet. 1217 */ 1218 if_dead(ifp); 1219 1220 /* 1221 * Clean up all addresses. 1222 */ 1223 IF_ADDR_WLOCK(ifp); 1224 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1225 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1226 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1227 IF_ADDR_WUNLOCK(ifp); 1228 ifa_free(ifa); 1229 } else 1230 IF_ADDR_WUNLOCK(ifp); 1231 } 1232 1233 rt_flushifroutes(ifp); 1234 1235 #ifdef VIMAGE 1236 finish_vnet_shutdown: 1237 #endif 1238 /* 1239 * We cannot hold the lock over dom_ifdetach calls as they might 1240 * sleep, for example trying to drain a callout, thus open up the 1241 * theoretical race with re-attaching. 1242 */ 1243 IF_AFDATA_LOCK(ifp); 1244 i = ifp->if_afdata_initialized; 1245 ifp->if_afdata_initialized = 0; 1246 IF_AFDATA_UNLOCK(ifp); 1247 if (i == 0) 1248 return (0); 1249 SLIST_FOREACH(dp, &domains, dom_next) { 1250 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1251 (*dp->dom_ifdetach)(ifp, 1252 ifp->if_afdata[dp->dom_family]); 1253 ifp->if_afdata[dp->dom_family] = NULL; 1254 } 1255 } 1256 1257 return (0); 1258 } 1259 1260 #ifdef VIMAGE 1261 /* 1262 * if_vmove() performs a limited version of if_detach() in current 1263 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1264 */ 1265 static int 1266 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1267 { 1268 #ifdef DEV_BPF 1269 u_int bif_dlt, bif_hdrlen; 1270 #endif 1271 int rc; 1272 1273 #ifdef DEV_BPF 1274 /* 1275 * if_detach_internal() will call the eventhandler to notify 1276 * interface departure. That will detach if_bpf. We need to 1277 * safe the dlt and hdrlen so we can re-attach it later. 1278 */ 1279 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1280 #endif 1281 1282 /* 1283 * Detach from current vnet, but preserve LLADDR info, do not 1284 * mark as dead etc. so that the ifnet can be reattached later. 1285 * If we cannot find it, we lost the race to someone else. 1286 */ 1287 rc = if_detach_internal(ifp, true); 1288 if (rc != 0) 1289 return (rc); 1290 1291 /* 1292 * Perform interface-specific reassignment tasks, if provided by 1293 * the driver. 1294 */ 1295 if (ifp->if_reassign != NULL) 1296 ifp->if_reassign(ifp, new_vnet, NULL); 1297 1298 /* 1299 * Switch to the context of the target vnet. 1300 */ 1301 CURVNET_SET_QUIET(new_vnet); 1302 if_attach_internal(ifp, true); 1303 1304 #ifdef DEV_BPF 1305 if (ifp->if_bpf == NULL) 1306 bpfattach(ifp, bif_dlt, bif_hdrlen); 1307 #endif 1308 1309 CURVNET_RESTORE(); 1310 return (0); 1311 } 1312 1313 /* 1314 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1315 */ 1316 static int 1317 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1318 { 1319 struct prison *pr; 1320 struct ifnet *difp; 1321 int error; 1322 bool found __diagused; 1323 bool shutdown; 1324 1325 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 1326 1327 /* Try to find the prison within our visibility. */ 1328 sx_slock(&allprison_lock); 1329 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1330 sx_sunlock(&allprison_lock); 1331 if (pr == NULL) 1332 return (ENXIO); 1333 prison_hold_locked(pr); 1334 mtx_unlock(&pr->pr_mtx); 1335 1336 /* Do not try to move the iface from and to the same prison. */ 1337 if (pr->pr_vnet == ifp->if_vnet) { 1338 prison_free(pr); 1339 return (EEXIST); 1340 } 1341 1342 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1343 /* XXX Lock interfaces to avoid races. */ 1344 CURVNET_SET_QUIET(pr->pr_vnet); 1345 difp = ifunit(ifname); 1346 if (difp != NULL) { 1347 CURVNET_RESTORE(); 1348 prison_free(pr); 1349 return (EEXIST); 1350 } 1351 sx_xlock(&ifnet_detach_sxlock); 1352 1353 /* Make sure the VNET is stable. */ 1354 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1355 if (shutdown) { 1356 sx_xunlock(&ifnet_detach_sxlock); 1357 CURVNET_RESTORE(); 1358 prison_free(pr); 1359 return (EBUSY); 1360 } 1361 CURVNET_RESTORE(); 1362 1363 found = if_unlink_ifnet(ifp, true); 1364 if (! found) { 1365 sx_xunlock(&ifnet_detach_sxlock); 1366 CURVNET_RESTORE(); 1367 prison_free(pr); 1368 return (ENODEV); 1369 } 1370 1371 /* Move the interface into the child jail/vnet. */ 1372 error = if_vmove(ifp, pr->pr_vnet); 1373 1374 /* Report the new if_xname back to the userland on success. */ 1375 if (error == 0) 1376 sprintf(ifname, "%s", ifp->if_xname); 1377 1378 sx_xunlock(&ifnet_detach_sxlock); 1379 1380 prison_free(pr); 1381 return (error); 1382 } 1383 1384 static int 1385 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1386 { 1387 struct prison *pr; 1388 struct vnet *vnet_dst; 1389 struct ifnet *ifp; 1390 int error, found __diagused; 1391 bool shutdown; 1392 1393 /* Try to find the prison within our visibility. */ 1394 sx_slock(&allprison_lock); 1395 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1396 sx_sunlock(&allprison_lock); 1397 if (pr == NULL) 1398 return (ENXIO); 1399 prison_hold_locked(pr); 1400 mtx_unlock(&pr->pr_mtx); 1401 1402 /* Make sure the named iface exists in the source prison/vnet. */ 1403 CURVNET_SET(pr->pr_vnet); 1404 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1405 if (ifp == NULL) { 1406 CURVNET_RESTORE(); 1407 prison_free(pr); 1408 return (ENXIO); 1409 } 1410 1411 /* Do not try to move the iface from and to the same prison. */ 1412 vnet_dst = TD_TO_VNET(td); 1413 if (vnet_dst == ifp->if_vnet) { 1414 CURVNET_RESTORE(); 1415 prison_free(pr); 1416 return (EEXIST); 1417 } 1418 1419 /* Make sure the VNET is stable. */ 1420 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1421 if (shutdown) { 1422 CURVNET_RESTORE(); 1423 prison_free(pr); 1424 return (EBUSY); 1425 } 1426 1427 /* Get interface back from child jail/vnet. */ 1428 found = if_unlink_ifnet(ifp, true); 1429 MPASS(found); 1430 sx_xlock(&ifnet_detach_sxlock); 1431 error = if_vmove(ifp, vnet_dst); 1432 sx_xunlock(&ifnet_detach_sxlock); 1433 CURVNET_RESTORE(); 1434 1435 /* Report the new if_xname back to the userland on success. */ 1436 if (error == 0) 1437 sprintf(ifname, "%s", ifp->if_xname); 1438 1439 prison_free(pr); 1440 return (error); 1441 } 1442 #endif /* VIMAGE */ 1443 1444 /* 1445 * Add a group to an interface 1446 */ 1447 int 1448 if_addgroup(struct ifnet *ifp, const char *groupname) 1449 { 1450 struct ifg_list *ifgl; 1451 struct ifg_group *ifg = NULL; 1452 struct ifg_member *ifgm; 1453 int new = 0; 1454 1455 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1456 groupname[strlen(groupname) - 1] <= '9') 1457 return (EINVAL); 1458 1459 IFNET_WLOCK(); 1460 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1461 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1462 IFNET_WUNLOCK(); 1463 return (EEXIST); 1464 } 1465 1466 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1467 IFNET_WUNLOCK(); 1468 return (ENOMEM); 1469 } 1470 1471 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1472 free(ifgl, M_TEMP); 1473 IFNET_WUNLOCK(); 1474 return (ENOMEM); 1475 } 1476 1477 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1478 if (!strcmp(ifg->ifg_group, groupname)) 1479 break; 1480 1481 if (ifg == NULL) { 1482 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1483 free(ifgl, M_TEMP); 1484 free(ifgm, M_TEMP); 1485 IFNET_WUNLOCK(); 1486 return (ENOMEM); 1487 } 1488 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1489 ifg->ifg_refcnt = 0; 1490 CK_STAILQ_INIT(&ifg->ifg_members); 1491 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1492 new = 1; 1493 } 1494 1495 ifg->ifg_refcnt++; 1496 ifgl->ifgl_group = ifg; 1497 ifgm->ifgm_ifp = ifp; 1498 1499 IF_ADDR_WLOCK(ifp); 1500 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1501 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1502 IF_ADDR_WUNLOCK(ifp); 1503 1504 IFNET_WUNLOCK(); 1505 1506 if (new) 1507 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1508 EVENTHANDLER_INVOKE(group_change_event, groupname); 1509 1510 return (0); 1511 } 1512 1513 /* 1514 * Helper function to remove a group out of an interface. Expects the global 1515 * ifnet lock to be write-locked, and drops it before returning. 1516 */ 1517 static void 1518 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1519 const char *groupname) 1520 { 1521 struct ifg_member *ifgm; 1522 bool freeifgl; 1523 1524 IFNET_WLOCK_ASSERT(); 1525 1526 IF_ADDR_WLOCK(ifp); 1527 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1528 IF_ADDR_WUNLOCK(ifp); 1529 1530 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1531 if (ifgm->ifgm_ifp == ifp) { 1532 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1533 ifg_member, ifgm_next); 1534 break; 1535 } 1536 } 1537 1538 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1539 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1540 ifg_next); 1541 freeifgl = true; 1542 } else { 1543 freeifgl = false; 1544 } 1545 IFNET_WUNLOCK(); 1546 1547 epoch_wait_preempt(net_epoch_preempt); 1548 EVENTHANDLER_INVOKE(group_change_event, groupname); 1549 if (freeifgl) { 1550 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1551 free(ifgl->ifgl_group, M_TEMP); 1552 } 1553 free(ifgm, M_TEMP); 1554 free(ifgl, M_TEMP); 1555 } 1556 1557 /* 1558 * Remove a group from an interface 1559 */ 1560 int 1561 if_delgroup(struct ifnet *ifp, const char *groupname) 1562 { 1563 struct ifg_list *ifgl; 1564 1565 IFNET_WLOCK(); 1566 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1567 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1568 break; 1569 if (ifgl == NULL) { 1570 IFNET_WUNLOCK(); 1571 return (ENOENT); 1572 } 1573 1574 _if_delgroup_locked(ifp, ifgl, groupname); 1575 1576 return (0); 1577 } 1578 1579 /* 1580 * Remove an interface from all groups 1581 */ 1582 static void 1583 if_delgroups(struct ifnet *ifp) 1584 { 1585 struct ifg_list *ifgl; 1586 char groupname[IFNAMSIZ]; 1587 1588 IFNET_WLOCK(); 1589 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1590 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1591 _if_delgroup_locked(ifp, ifgl, groupname); 1592 IFNET_WLOCK(); 1593 } 1594 IFNET_WUNLOCK(); 1595 } 1596 1597 /* 1598 * Stores all groups from an interface in memory pointed to by ifgr. 1599 */ 1600 static int 1601 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1602 { 1603 int len, error; 1604 struct ifg_list *ifgl; 1605 struct ifg_req ifgrq, *ifgp; 1606 1607 NET_EPOCH_ASSERT(); 1608 1609 if (ifgr->ifgr_len == 0) { 1610 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1611 ifgr->ifgr_len += sizeof(struct ifg_req); 1612 return (0); 1613 } 1614 1615 len = ifgr->ifgr_len; 1616 ifgp = ifgr->ifgr_groups; 1617 /* XXX: wire */ 1618 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1619 if (len < sizeof(ifgrq)) 1620 return (EINVAL); 1621 bzero(&ifgrq, sizeof ifgrq); 1622 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1623 sizeof(ifgrq.ifgrq_group)); 1624 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1625 return (error); 1626 len -= sizeof(ifgrq); 1627 ifgp++; 1628 } 1629 1630 return (0); 1631 } 1632 1633 /* 1634 * Stores all members of a group in memory pointed to by igfr 1635 */ 1636 static int 1637 if_getgroupmembers(struct ifgroupreq *ifgr) 1638 { 1639 struct ifg_group *ifg; 1640 struct ifg_member *ifgm; 1641 struct ifg_req ifgrq, *ifgp; 1642 int len, error; 1643 1644 IFNET_RLOCK(); 1645 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1646 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1647 break; 1648 if (ifg == NULL) { 1649 IFNET_RUNLOCK(); 1650 return (ENOENT); 1651 } 1652 1653 if (ifgr->ifgr_len == 0) { 1654 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1655 ifgr->ifgr_len += sizeof(ifgrq); 1656 IFNET_RUNLOCK(); 1657 return (0); 1658 } 1659 1660 len = ifgr->ifgr_len; 1661 ifgp = ifgr->ifgr_groups; 1662 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1663 if (len < sizeof(ifgrq)) { 1664 IFNET_RUNLOCK(); 1665 return (EINVAL); 1666 } 1667 bzero(&ifgrq, sizeof ifgrq); 1668 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1669 sizeof(ifgrq.ifgrq_member)); 1670 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1671 IFNET_RUNLOCK(); 1672 return (error); 1673 } 1674 len -= sizeof(ifgrq); 1675 ifgp++; 1676 } 1677 IFNET_RUNLOCK(); 1678 1679 return (0); 1680 } 1681 1682 /* 1683 * Return counter values from counter(9)s stored in ifnet. 1684 */ 1685 uint64_t 1686 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1687 { 1688 1689 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1690 1691 return (counter_u64_fetch(ifp->if_counters[cnt])); 1692 } 1693 1694 /* 1695 * Increase an ifnet counter. Usually used for counters shared 1696 * between the stack and a driver, but function supports them all. 1697 */ 1698 void 1699 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1700 { 1701 1702 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1703 1704 counter_u64_add(ifp->if_counters[cnt], inc); 1705 } 1706 1707 /* 1708 * Copy data from ifnet to userland API structure if_data. 1709 */ 1710 void 1711 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1712 { 1713 1714 ifd->ifi_type = ifp->if_type; 1715 ifd->ifi_physical = 0; 1716 ifd->ifi_addrlen = ifp->if_addrlen; 1717 ifd->ifi_hdrlen = ifp->if_hdrlen; 1718 ifd->ifi_link_state = ifp->if_link_state; 1719 ifd->ifi_vhid = 0; 1720 ifd->ifi_datalen = sizeof(struct if_data); 1721 ifd->ifi_mtu = ifp->if_mtu; 1722 ifd->ifi_metric = ifp->if_metric; 1723 ifd->ifi_baudrate = ifp->if_baudrate; 1724 ifd->ifi_hwassist = ifp->if_hwassist; 1725 ifd->ifi_epoch = ifp->if_epoch; 1726 ifd->ifi_lastchange = ifp->if_lastchange; 1727 1728 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1729 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1730 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1731 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1732 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1733 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1734 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1735 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1736 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1737 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1738 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1739 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1740 } 1741 1742 /* 1743 * Initialization, destruction and refcounting functions for ifaddrs. 1744 */ 1745 struct ifaddr * 1746 ifa_alloc(size_t size, int flags) 1747 { 1748 struct ifaddr *ifa; 1749 1750 KASSERT(size >= sizeof(struct ifaddr), 1751 ("%s: invalid size %zu", __func__, size)); 1752 1753 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1754 if (ifa == NULL) 1755 return (NULL); 1756 1757 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1758 goto fail; 1759 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1760 goto fail; 1761 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1762 goto fail; 1763 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1764 goto fail; 1765 1766 refcount_init(&ifa->ifa_refcnt, 1); 1767 1768 return (ifa); 1769 1770 fail: 1771 /* free(NULL) is okay */ 1772 counter_u64_free(ifa->ifa_opackets); 1773 counter_u64_free(ifa->ifa_ipackets); 1774 counter_u64_free(ifa->ifa_obytes); 1775 counter_u64_free(ifa->ifa_ibytes); 1776 free(ifa, M_IFADDR); 1777 1778 return (NULL); 1779 } 1780 1781 void 1782 ifa_ref(struct ifaddr *ifa) 1783 { 1784 u_int old __diagused; 1785 1786 old = refcount_acquire(&ifa->ifa_refcnt); 1787 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1788 } 1789 1790 int 1791 ifa_try_ref(struct ifaddr *ifa) 1792 { 1793 1794 NET_EPOCH_ASSERT(); 1795 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1796 } 1797 1798 static void 1799 ifa_destroy(epoch_context_t ctx) 1800 { 1801 struct ifaddr *ifa; 1802 1803 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1804 counter_u64_free(ifa->ifa_opackets); 1805 counter_u64_free(ifa->ifa_ipackets); 1806 counter_u64_free(ifa->ifa_obytes); 1807 counter_u64_free(ifa->ifa_ibytes); 1808 free(ifa, M_IFADDR); 1809 } 1810 1811 void 1812 ifa_free(struct ifaddr *ifa) 1813 { 1814 1815 if (refcount_release(&ifa->ifa_refcnt)) 1816 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1817 } 1818 1819 /* 1820 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1821 * structs used to represent other address families, it is necessary 1822 * to perform a different comparison. 1823 */ 1824 1825 #define sa_dl_equal(a1, a2) \ 1826 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1827 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1828 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1829 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1830 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1831 1832 /* 1833 * Locate an interface based on a complete address. 1834 */ 1835 /*ARGSUSED*/ 1836 struct ifaddr * 1837 ifa_ifwithaddr(const struct sockaddr *addr) 1838 { 1839 struct ifnet *ifp; 1840 struct ifaddr *ifa; 1841 1842 NET_EPOCH_ASSERT(); 1843 1844 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1845 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1846 if (ifa->ifa_addr->sa_family != addr->sa_family) 1847 continue; 1848 if (sa_equal(addr, ifa->ifa_addr)) { 1849 goto done; 1850 } 1851 /* IP6 doesn't have broadcast */ 1852 if ((ifp->if_flags & IFF_BROADCAST) && 1853 ifa->ifa_broadaddr && 1854 ifa->ifa_broadaddr->sa_len != 0 && 1855 sa_equal(ifa->ifa_broadaddr, addr)) { 1856 goto done; 1857 } 1858 } 1859 } 1860 ifa = NULL; 1861 done: 1862 return (ifa); 1863 } 1864 1865 int 1866 ifa_ifwithaddr_check(const struct sockaddr *addr) 1867 { 1868 struct epoch_tracker et; 1869 int rc; 1870 1871 NET_EPOCH_ENTER(et); 1872 rc = (ifa_ifwithaddr(addr) != NULL); 1873 NET_EPOCH_EXIT(et); 1874 return (rc); 1875 } 1876 1877 /* 1878 * Locate an interface based on the broadcast address. 1879 */ 1880 /* ARGSUSED */ 1881 struct ifaddr * 1882 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1883 { 1884 struct ifnet *ifp; 1885 struct ifaddr *ifa; 1886 1887 NET_EPOCH_ASSERT(); 1888 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1889 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1890 continue; 1891 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1892 if (ifa->ifa_addr->sa_family != addr->sa_family) 1893 continue; 1894 if ((ifp->if_flags & IFF_BROADCAST) && 1895 ifa->ifa_broadaddr && 1896 ifa->ifa_broadaddr->sa_len != 0 && 1897 sa_equal(ifa->ifa_broadaddr, addr)) { 1898 goto done; 1899 } 1900 } 1901 } 1902 ifa = NULL; 1903 done: 1904 return (ifa); 1905 } 1906 1907 /* 1908 * Locate the point to point interface with a given destination address. 1909 */ 1910 /*ARGSUSED*/ 1911 struct ifaddr * 1912 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1913 { 1914 struct ifnet *ifp; 1915 struct ifaddr *ifa; 1916 1917 NET_EPOCH_ASSERT(); 1918 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1919 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1920 continue; 1921 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1922 continue; 1923 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1924 if (ifa->ifa_addr->sa_family != addr->sa_family) 1925 continue; 1926 if (ifa->ifa_dstaddr != NULL && 1927 sa_equal(addr, ifa->ifa_dstaddr)) { 1928 goto done; 1929 } 1930 } 1931 } 1932 ifa = NULL; 1933 done: 1934 return (ifa); 1935 } 1936 1937 /* 1938 * Find an interface on a specific network. If many, choice 1939 * is most specific found. 1940 */ 1941 struct ifaddr * 1942 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1943 { 1944 struct ifnet *ifp; 1945 struct ifaddr *ifa; 1946 struct ifaddr *ifa_maybe = NULL; 1947 u_int af = addr->sa_family; 1948 const char *addr_data = addr->sa_data, *cplim; 1949 1950 NET_EPOCH_ASSERT(); 1951 /* 1952 * AF_LINK addresses can be looked up directly by their index number, 1953 * so do that if we can. 1954 */ 1955 if (af == AF_LINK) { 1956 ifp = ifnet_byindex( 1957 ((const struct sockaddr_dl *)addr)->sdl_index); 1958 return (ifp ? ifp->if_addr : NULL); 1959 } 1960 1961 /* 1962 * Scan though each interface, looking for ones that have addresses 1963 * in this address family and the requested fib. 1964 */ 1965 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1966 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1967 continue; 1968 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1969 const char *cp, *cp2, *cp3; 1970 1971 if (ifa->ifa_addr->sa_family != af) 1972 next: continue; 1973 if (af == AF_INET && 1974 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1975 /* 1976 * This is a bit broken as it doesn't 1977 * take into account that the remote end may 1978 * be a single node in the network we are 1979 * looking for. 1980 * The trouble is that we don't know the 1981 * netmask for the remote end. 1982 */ 1983 if (ifa->ifa_dstaddr != NULL && 1984 sa_equal(addr, ifa->ifa_dstaddr)) { 1985 goto done; 1986 } 1987 } else { 1988 /* 1989 * Scan all the bits in the ifa's address. 1990 * If a bit dissagrees with what we are 1991 * looking for, mask it with the netmask 1992 * to see if it really matters. 1993 * (A byte at a time) 1994 */ 1995 if (ifa->ifa_netmask == 0) 1996 continue; 1997 cp = addr_data; 1998 cp2 = ifa->ifa_addr->sa_data; 1999 cp3 = ifa->ifa_netmask->sa_data; 2000 cplim = ifa->ifa_netmask->sa_len 2001 + (char *)ifa->ifa_netmask; 2002 while (cp3 < cplim) 2003 if ((*cp++ ^ *cp2++) & *cp3++) 2004 goto next; /* next address! */ 2005 /* 2006 * If the netmask of what we just found 2007 * is more specific than what we had before 2008 * (if we had one), or if the virtual status 2009 * of new prefix is better than of the old one, 2010 * then remember the new one before continuing 2011 * to search for an even better one. 2012 */ 2013 if (ifa_maybe == NULL || 2014 ifa_preferred(ifa_maybe, ifa) || 2015 rn_refines((caddr_t)ifa->ifa_netmask, 2016 (caddr_t)ifa_maybe->ifa_netmask)) { 2017 ifa_maybe = ifa; 2018 } 2019 } 2020 } 2021 } 2022 ifa = ifa_maybe; 2023 ifa_maybe = NULL; 2024 done: 2025 return (ifa); 2026 } 2027 2028 /* 2029 * Find an interface address specific to an interface best matching 2030 * a given address. 2031 */ 2032 struct ifaddr * 2033 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2034 { 2035 struct ifaddr *ifa; 2036 const char *cp, *cp2, *cp3; 2037 char *cplim; 2038 struct ifaddr *ifa_maybe = NULL; 2039 u_int af = addr->sa_family; 2040 2041 if (af >= AF_MAX) 2042 return (NULL); 2043 2044 NET_EPOCH_ASSERT(); 2045 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2046 if (ifa->ifa_addr->sa_family != af) 2047 continue; 2048 if (ifa_maybe == NULL) 2049 ifa_maybe = ifa; 2050 if (ifa->ifa_netmask == 0) { 2051 if (sa_equal(addr, ifa->ifa_addr) || 2052 (ifa->ifa_dstaddr && 2053 sa_equal(addr, ifa->ifa_dstaddr))) 2054 goto done; 2055 continue; 2056 } 2057 if (ifp->if_flags & IFF_POINTOPOINT) { 2058 if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr)) 2059 goto done; 2060 } else { 2061 cp = addr->sa_data; 2062 cp2 = ifa->ifa_addr->sa_data; 2063 cp3 = ifa->ifa_netmask->sa_data; 2064 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2065 for (; cp3 < cplim; cp3++) 2066 if ((*cp++ ^ *cp2++) & *cp3) 2067 break; 2068 if (cp3 == cplim) 2069 goto done; 2070 } 2071 } 2072 ifa = ifa_maybe; 2073 done: 2074 return (ifa); 2075 } 2076 2077 /* 2078 * See whether new ifa is better than current one: 2079 * 1) A non-virtual one is preferred over virtual. 2080 * 2) A virtual in master state preferred over any other state. 2081 * 2082 * Used in several address selecting functions. 2083 */ 2084 int 2085 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2086 { 2087 2088 return (cur->ifa_carp && (!next->ifa_carp || 2089 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2090 } 2091 2092 struct sockaddr_dl * 2093 link_alloc_sdl(size_t size, int flags) 2094 { 2095 2096 return (malloc(size, M_TEMP, flags)); 2097 } 2098 2099 void 2100 link_free_sdl(struct sockaddr *sa) 2101 { 2102 free(sa, M_TEMP); 2103 } 2104 2105 /* 2106 * Fills in given sdl with interface basic info. 2107 * Returns pointer to filled sdl. 2108 */ 2109 struct sockaddr_dl * 2110 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2111 { 2112 struct sockaddr_dl *sdl; 2113 2114 sdl = (struct sockaddr_dl *)paddr; 2115 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2116 sdl->sdl_len = sizeof(struct sockaddr_dl); 2117 sdl->sdl_family = AF_LINK; 2118 sdl->sdl_index = ifp->if_index; 2119 sdl->sdl_type = iftype; 2120 2121 return (sdl); 2122 } 2123 2124 /* 2125 * Mark an interface down and notify protocols of 2126 * the transition. 2127 */ 2128 static void 2129 if_unroute(struct ifnet *ifp, int flag, int fam) 2130 { 2131 2132 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2133 2134 ifp->if_flags &= ~flag; 2135 getmicrotime(&ifp->if_lastchange); 2136 ifp->if_qflush(ifp); 2137 2138 if (ifp->if_carp) 2139 (*carp_linkstate_p)(ifp); 2140 rt_ifmsg(ifp, IFF_UP); 2141 } 2142 2143 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2144 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2145 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2146 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2147 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2148 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2149 int (*vlan_setcookie_p)(struct ifnet *, void *); 2150 void *(*vlan_cookie_p)(struct ifnet *); 2151 2152 /* 2153 * Handle a change in the interface link state. To avoid LORs 2154 * between driver lock and upper layer locks, as well as possible 2155 * recursions, we post event to taskqueue, and all job 2156 * is done in static do_link_state_change(). 2157 */ 2158 void 2159 if_link_state_change(struct ifnet *ifp, int link_state) 2160 { 2161 /* Return if state hasn't changed. */ 2162 if (ifp->if_link_state == link_state) 2163 return; 2164 2165 ifp->if_link_state = link_state; 2166 2167 /* XXXGL: reference ifp? */ 2168 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2169 } 2170 2171 static void 2172 do_link_state_change(void *arg, int pending) 2173 { 2174 struct ifnet *ifp; 2175 int link_state; 2176 2177 ifp = arg; 2178 link_state = ifp->if_link_state; 2179 2180 CURVNET_SET(ifp->if_vnet); 2181 rt_ifmsg(ifp, 0); 2182 if (ifp->if_vlantrunk != NULL) 2183 (*vlan_link_state_p)(ifp); 2184 2185 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2186 ifp->if_l2com != NULL) 2187 (*ng_ether_link_state_p)(ifp, link_state); 2188 if (ifp->if_carp) 2189 (*carp_linkstate_p)(ifp); 2190 if (ifp->if_bridge) 2191 ifp->if_bridge_linkstate(ifp); 2192 if (ifp->if_lagg) 2193 (*lagg_linkstate_p)(ifp, link_state); 2194 2195 if (IS_DEFAULT_VNET(curvnet)) 2196 devctl_notify("IFNET", ifp->if_xname, 2197 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2198 NULL); 2199 if (pending > 1) 2200 if_printf(ifp, "%d link states coalesced\n", pending); 2201 if (log_link_state_change) 2202 if_printf(ifp, "link state changed to %s\n", 2203 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2204 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2205 CURVNET_RESTORE(); 2206 } 2207 2208 /* 2209 * Mark an interface down and notify protocols of 2210 * the transition. 2211 */ 2212 void 2213 if_down(struct ifnet *ifp) 2214 { 2215 2216 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2217 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2218 } 2219 2220 /* 2221 * Mark an interface up and notify protocols of 2222 * the transition. 2223 */ 2224 void 2225 if_up(struct ifnet *ifp) 2226 { 2227 2228 ifp->if_flags |= IFF_UP; 2229 getmicrotime(&ifp->if_lastchange); 2230 if (ifp->if_carp) 2231 (*carp_linkstate_p)(ifp); 2232 rt_ifmsg(ifp, IFF_UP); 2233 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2234 } 2235 2236 /* 2237 * Flush an interface queue. 2238 */ 2239 void 2240 if_qflush(struct ifnet *ifp) 2241 { 2242 struct mbuf *m, *n; 2243 struct ifaltq *ifq; 2244 2245 ifq = &ifp->if_snd; 2246 IFQ_LOCK(ifq); 2247 #ifdef ALTQ 2248 if (ALTQ_IS_ENABLED(ifq)) 2249 ALTQ_PURGE(ifq); 2250 #endif 2251 n = ifq->ifq_head; 2252 while ((m = n) != NULL) { 2253 n = m->m_nextpkt; 2254 m_freem(m); 2255 } 2256 ifq->ifq_head = 0; 2257 ifq->ifq_tail = 0; 2258 ifq->ifq_len = 0; 2259 IFQ_UNLOCK(ifq); 2260 } 2261 2262 /* 2263 * Map interface name to interface structure pointer, with or without 2264 * returning a reference. 2265 */ 2266 struct ifnet * 2267 ifunit_ref(const char *name) 2268 { 2269 struct epoch_tracker et; 2270 struct ifnet *ifp; 2271 2272 NET_EPOCH_ENTER(et); 2273 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2274 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2275 !(ifp->if_flags & IFF_DYING)) 2276 break; 2277 } 2278 if (ifp != NULL) { 2279 if_ref(ifp); 2280 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2281 } 2282 2283 NET_EPOCH_EXIT(et); 2284 return (ifp); 2285 } 2286 2287 struct ifnet * 2288 ifunit(const char *name) 2289 { 2290 struct epoch_tracker et; 2291 struct ifnet *ifp; 2292 2293 NET_EPOCH_ENTER(et); 2294 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2295 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2296 break; 2297 } 2298 NET_EPOCH_EXIT(et); 2299 return (ifp); 2300 } 2301 2302 void * 2303 ifr_buffer_get_buffer(void *data) 2304 { 2305 union ifreq_union *ifrup; 2306 2307 ifrup = data; 2308 #ifdef COMPAT_FREEBSD32 2309 if (SV_CURPROC_FLAG(SV_ILP32)) 2310 return ((void *)(uintptr_t) 2311 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2312 #endif 2313 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2314 } 2315 2316 static void 2317 ifr_buffer_set_buffer_null(void *data) 2318 { 2319 union ifreq_union *ifrup; 2320 2321 ifrup = data; 2322 #ifdef COMPAT_FREEBSD32 2323 if (SV_CURPROC_FLAG(SV_ILP32)) 2324 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2325 else 2326 #endif 2327 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2328 } 2329 2330 size_t 2331 ifr_buffer_get_length(void *data) 2332 { 2333 union ifreq_union *ifrup; 2334 2335 ifrup = data; 2336 #ifdef COMPAT_FREEBSD32 2337 if (SV_CURPROC_FLAG(SV_ILP32)) 2338 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2339 #endif 2340 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2341 } 2342 2343 static void 2344 ifr_buffer_set_length(void *data, size_t len) 2345 { 2346 union ifreq_union *ifrup; 2347 2348 ifrup = data; 2349 #ifdef COMPAT_FREEBSD32 2350 if (SV_CURPROC_FLAG(SV_ILP32)) 2351 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2352 else 2353 #endif 2354 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2355 } 2356 2357 void * 2358 ifr_data_get_ptr(void *ifrp) 2359 { 2360 union ifreq_union *ifrup; 2361 2362 ifrup = ifrp; 2363 #ifdef COMPAT_FREEBSD32 2364 if (SV_CURPROC_FLAG(SV_ILP32)) 2365 return ((void *)(uintptr_t) 2366 ifrup->ifr32.ifr_ifru.ifru_data); 2367 #endif 2368 return (ifrup->ifr.ifr_ifru.ifru_data); 2369 } 2370 2371 struct ifcap_nv_bit_name { 2372 uint64_t cap_bit; 2373 const char *cap_name; 2374 }; 2375 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2376 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2377 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2378 CAPNV(RXCSUM), 2379 CAPNV(TXCSUM), 2380 CAPNV(NETCONS), 2381 CAPNV(VLAN_MTU), 2382 CAPNV(VLAN_HWTAGGING), 2383 CAPNV(JUMBO_MTU), 2384 CAPNV(POLLING), 2385 CAPNV(VLAN_HWCSUM), 2386 CAPNV(TSO4), 2387 CAPNV(TSO6), 2388 CAPNV(LRO), 2389 CAPNV(WOL_UCAST), 2390 CAPNV(WOL_MCAST), 2391 CAPNV(WOL_MAGIC), 2392 CAPNV(TOE4), 2393 CAPNV(TOE6), 2394 CAPNV(VLAN_HWFILTER), 2395 CAPNV(VLAN_HWTSO), 2396 CAPNV(LINKSTATE), 2397 CAPNV(NETMAP), 2398 CAPNV(RXCSUM_IPV6), 2399 CAPNV(TXCSUM_IPV6), 2400 CAPNV(HWSTATS), 2401 CAPNV(TXRTLMT), 2402 CAPNV(HWRXTSTMP), 2403 CAPNV(MEXTPG), 2404 CAPNV(TXTLS4), 2405 CAPNV(TXTLS6), 2406 CAPNV(VXLAN_HWCSUM), 2407 CAPNV(VXLAN_HWTSO), 2408 CAPNV(TXTLS_RTLMT), 2409 {0, NULL} 2410 }; 2411 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \ 2412 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2413 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2414 CAP2NV(RXTLS4), 2415 CAP2NV(RXTLS6), 2416 {0, NULL} 2417 }; 2418 #undef CAPNV 2419 #undef CAP2NV 2420 2421 int 2422 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2423 const struct ifcap_nv_bit_name *nn, bool all) 2424 { 2425 int i, res; 2426 2427 res = 0; 2428 for (i = 0; nn[i].cap_name != NULL; i++) { 2429 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2430 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2431 res |= nn[i].cap_bit; 2432 } else { 2433 res |= *old_cap & nn[i].cap_bit; 2434 } 2435 } 2436 return (res); 2437 } 2438 2439 void 2440 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2441 int ifr_cap, int ifr_req) 2442 { 2443 int i; 2444 2445 for (i = 0; nn[i].cap_name != NULL; i++) { 2446 if ((nn[i].cap_bit & ifr_cap) != 0) { 2447 nvlist_add_bool(nv, nn[i].cap_name, 2448 (nn[i].cap_bit & ifr_req) != 0); 2449 } 2450 } 2451 } 2452 2453 /* 2454 * Hardware specific interface ioctls. 2455 */ 2456 int 2457 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2458 { 2459 struct ifreq *ifr; 2460 int error = 0, do_ifup = 0; 2461 int new_flags, temp_flags; 2462 size_t descrlen, nvbuflen; 2463 char *descrbuf; 2464 char new_name[IFNAMSIZ]; 2465 void *buf; 2466 nvlist_t *nvcap; 2467 struct siocsifcapnv_driver_data drv_ioctl_data; 2468 2469 ifr = (struct ifreq *)data; 2470 switch (cmd) { 2471 case SIOCGIFINDEX: 2472 ifr->ifr_index = ifp->if_index; 2473 break; 2474 2475 case SIOCGIFFLAGS: 2476 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2477 ifr->ifr_flags = temp_flags & 0xffff; 2478 ifr->ifr_flagshigh = temp_flags >> 16; 2479 break; 2480 2481 case SIOCGIFCAP: 2482 ifr->ifr_reqcap = ifp->if_capabilities; 2483 ifr->ifr_curcap = ifp->if_capenable; 2484 break; 2485 2486 case SIOCGIFCAPNV: 2487 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2488 error = EINVAL; 2489 break; 2490 } 2491 buf = NULL; 2492 nvcap = nvlist_create(0); 2493 for (;;) { 2494 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2495 ifp->if_capabilities, ifp->if_capenable); 2496 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2497 ifp->if_capabilities2, ifp->if_capenable2); 2498 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2499 __DECONST(caddr_t, nvcap)); 2500 if (error != 0) { 2501 if_printf(ifp, 2502 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2503 error); 2504 break; 2505 } 2506 buf = nvlist_pack(nvcap, &nvbuflen); 2507 if (buf == NULL) { 2508 error = nvlist_error(nvcap); 2509 if (error == 0) 2510 error = EDOOFUS; 2511 break; 2512 } 2513 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2514 ifr->ifr_cap_nv.length = nvbuflen; 2515 ifr->ifr_cap_nv.buffer = NULL; 2516 error = EFBIG; 2517 break; 2518 } 2519 ifr->ifr_cap_nv.length = nvbuflen; 2520 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2521 break; 2522 } 2523 free(buf, M_NVLIST); 2524 nvlist_destroy(nvcap); 2525 break; 2526 2527 case SIOCGIFDATA: 2528 { 2529 struct if_data ifd; 2530 2531 /* Ensure uninitialised padding is not leaked. */ 2532 memset(&ifd, 0, sizeof(ifd)); 2533 2534 if_data_copy(ifp, &ifd); 2535 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2536 break; 2537 } 2538 2539 #ifdef MAC 2540 case SIOCGIFMAC: 2541 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2542 break; 2543 #endif 2544 2545 case SIOCGIFMETRIC: 2546 ifr->ifr_metric = ifp->if_metric; 2547 break; 2548 2549 case SIOCGIFMTU: 2550 ifr->ifr_mtu = ifp->if_mtu; 2551 break; 2552 2553 case SIOCGIFPHYS: 2554 /* XXXGL: did this ever worked? */ 2555 ifr->ifr_phys = 0; 2556 break; 2557 2558 case SIOCGIFDESCR: 2559 error = 0; 2560 sx_slock(&ifdescr_sx); 2561 if (ifp->if_description == NULL) 2562 error = ENOMSG; 2563 else { 2564 /* space for terminating nul */ 2565 descrlen = strlen(ifp->if_description) + 1; 2566 if (ifr_buffer_get_length(ifr) < descrlen) 2567 ifr_buffer_set_buffer_null(ifr); 2568 else 2569 error = copyout(ifp->if_description, 2570 ifr_buffer_get_buffer(ifr), descrlen); 2571 ifr_buffer_set_length(ifr, descrlen); 2572 } 2573 sx_sunlock(&ifdescr_sx); 2574 break; 2575 2576 case SIOCSIFDESCR: 2577 error = priv_check(td, PRIV_NET_SETIFDESCR); 2578 if (error) 2579 return (error); 2580 2581 /* 2582 * Copy only (length-1) bytes to make sure that 2583 * if_description is always nul terminated. The 2584 * length parameter is supposed to count the 2585 * terminating nul in. 2586 */ 2587 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2588 return (ENAMETOOLONG); 2589 else if (ifr_buffer_get_length(ifr) == 0) 2590 descrbuf = NULL; 2591 else { 2592 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2593 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2594 ifr_buffer_get_length(ifr) - 1); 2595 if (error) { 2596 if_freedescr(descrbuf); 2597 break; 2598 } 2599 } 2600 2601 if_setdescr(ifp, descrbuf); 2602 getmicrotime(&ifp->if_lastchange); 2603 break; 2604 2605 case SIOCGIFFIB: 2606 ifr->ifr_fib = ifp->if_fib; 2607 break; 2608 2609 case SIOCSIFFIB: 2610 error = priv_check(td, PRIV_NET_SETIFFIB); 2611 if (error) 2612 return (error); 2613 if (ifr->ifr_fib >= rt_numfibs) 2614 return (EINVAL); 2615 2616 ifp->if_fib = ifr->ifr_fib; 2617 break; 2618 2619 case SIOCSIFFLAGS: 2620 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2621 if (error) 2622 return (error); 2623 /* 2624 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2625 * check, so we don't need special handling here yet. 2626 */ 2627 new_flags = (ifr->ifr_flags & 0xffff) | 2628 (ifr->ifr_flagshigh << 16); 2629 if (ifp->if_flags & IFF_UP && 2630 (new_flags & IFF_UP) == 0) { 2631 if_down(ifp); 2632 } else if (new_flags & IFF_UP && 2633 (ifp->if_flags & IFF_UP) == 0) { 2634 do_ifup = 1; 2635 } 2636 /* See if permanently promiscuous mode bit is about to flip */ 2637 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2638 if (new_flags & IFF_PPROMISC) 2639 ifp->if_flags |= IFF_PROMISC; 2640 else if (ifp->if_pcount == 0) 2641 ifp->if_flags &= ~IFF_PROMISC; 2642 if (log_promisc_mode_change) 2643 if_printf(ifp, "permanently promiscuous mode %s\n", 2644 ((new_flags & IFF_PPROMISC) ? 2645 "enabled" : "disabled")); 2646 } 2647 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2648 (new_flags &~ IFF_CANTCHANGE); 2649 if (ifp->if_ioctl) { 2650 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2651 } 2652 if (do_ifup) 2653 if_up(ifp); 2654 getmicrotime(&ifp->if_lastchange); 2655 break; 2656 2657 case SIOCSIFCAP: 2658 error = priv_check(td, PRIV_NET_SETIFCAP); 2659 if (error != 0) 2660 return (error); 2661 if (ifp->if_ioctl == NULL) 2662 return (EOPNOTSUPP); 2663 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2664 return (EINVAL); 2665 error = (*ifp->if_ioctl)(ifp, cmd, data); 2666 if (error == 0) 2667 getmicrotime(&ifp->if_lastchange); 2668 break; 2669 2670 case SIOCSIFCAPNV: 2671 error = priv_check(td, PRIV_NET_SETIFCAP); 2672 if (error != 0) 2673 return (error); 2674 if (ifp->if_ioctl == NULL) 2675 return (EOPNOTSUPP); 2676 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2677 return (EINVAL); 2678 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2679 return (EINVAL); 2680 nvcap = NULL; 2681 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2682 for (;;) { 2683 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2684 ifr->ifr_cap_nv.length); 2685 if (error != 0) 2686 break; 2687 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2688 if (nvcap == NULL) { 2689 error = EINVAL; 2690 break; 2691 } 2692 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2693 &ifp->if_capenable, ifcap_nv_bit_names, false); 2694 if ((drv_ioctl_data.reqcap & 2695 ~ifp->if_capabilities) != 0) { 2696 error = EINVAL; 2697 break; 2698 } 2699 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2700 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2701 if ((drv_ioctl_data.reqcap2 & 2702 ~ifp->if_capabilities2) != 0) { 2703 error = EINVAL; 2704 break; 2705 } 2706 drv_ioctl_data.nvcap = nvcap; 2707 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2708 (caddr_t)&drv_ioctl_data); 2709 break; 2710 } 2711 nvlist_destroy(nvcap); 2712 free(buf, M_TEMP); 2713 if (error == 0) 2714 getmicrotime(&ifp->if_lastchange); 2715 break; 2716 2717 #ifdef MAC 2718 case SIOCSIFMAC: 2719 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2720 break; 2721 #endif 2722 2723 case SIOCSIFNAME: 2724 error = priv_check(td, PRIV_NET_SETIFNAME); 2725 if (error) 2726 return (error); 2727 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2728 NULL); 2729 if (error != 0) 2730 return (error); 2731 error = if_rename(ifp, new_name); 2732 break; 2733 2734 #ifdef VIMAGE 2735 case SIOCSIFVNET: 2736 error = priv_check(td, PRIV_NET_SETIFVNET); 2737 if (error) 2738 return (error); 2739 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2740 break; 2741 #endif 2742 2743 case SIOCSIFMETRIC: 2744 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2745 if (error) 2746 return (error); 2747 ifp->if_metric = ifr->ifr_metric; 2748 getmicrotime(&ifp->if_lastchange); 2749 break; 2750 2751 case SIOCSIFPHYS: 2752 error = priv_check(td, PRIV_NET_SETIFPHYS); 2753 if (error) 2754 return (error); 2755 if (ifp->if_ioctl == NULL) 2756 return (EOPNOTSUPP); 2757 error = (*ifp->if_ioctl)(ifp, cmd, data); 2758 if (error == 0) 2759 getmicrotime(&ifp->if_lastchange); 2760 break; 2761 2762 case SIOCSIFMTU: 2763 { 2764 u_long oldmtu = ifp->if_mtu; 2765 2766 error = priv_check(td, PRIV_NET_SETIFMTU); 2767 if (error) 2768 return (error); 2769 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2770 return (EINVAL); 2771 if (ifp->if_ioctl == NULL) 2772 return (EOPNOTSUPP); 2773 /* Disallow MTU changes on bridge member interfaces. */ 2774 if (ifp->if_bridge) 2775 return (EOPNOTSUPP); 2776 error = (*ifp->if_ioctl)(ifp, cmd, data); 2777 if (error == 0) { 2778 getmicrotime(&ifp->if_lastchange); 2779 rt_ifmsg(ifp, 0); 2780 #ifdef INET 2781 DEBUGNET_NOTIFY_MTU(ifp); 2782 #endif 2783 } 2784 /* 2785 * If the link MTU changed, do network layer specific procedure. 2786 */ 2787 if (ifp->if_mtu != oldmtu) 2788 if_notifymtu(ifp); 2789 break; 2790 } 2791 2792 case SIOCADDMULTI: 2793 case SIOCDELMULTI: 2794 if (cmd == SIOCADDMULTI) 2795 error = priv_check(td, PRIV_NET_ADDMULTI); 2796 else 2797 error = priv_check(td, PRIV_NET_DELMULTI); 2798 if (error) 2799 return (error); 2800 2801 /* Don't allow group membership on non-multicast interfaces. */ 2802 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2803 return (EOPNOTSUPP); 2804 2805 /* Don't let users screw up protocols' entries. */ 2806 if (ifr->ifr_addr.sa_family != AF_LINK) 2807 return (EINVAL); 2808 2809 if (cmd == SIOCADDMULTI) { 2810 struct epoch_tracker et; 2811 struct ifmultiaddr *ifma; 2812 2813 /* 2814 * Userland is only permitted to join groups once 2815 * via the if_addmulti() KPI, because it cannot hold 2816 * struct ifmultiaddr * between calls. It may also 2817 * lose a race while we check if the membership 2818 * already exists. 2819 */ 2820 NET_EPOCH_ENTER(et); 2821 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2822 NET_EPOCH_EXIT(et); 2823 if (ifma != NULL) 2824 error = EADDRINUSE; 2825 else 2826 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2827 } else { 2828 error = if_delmulti(ifp, &ifr->ifr_addr); 2829 } 2830 if (error == 0) 2831 getmicrotime(&ifp->if_lastchange); 2832 break; 2833 2834 case SIOCSIFPHYADDR: 2835 case SIOCDIFPHYADDR: 2836 #ifdef INET6 2837 case SIOCSIFPHYADDR_IN6: 2838 #endif 2839 case SIOCSIFMEDIA: 2840 case SIOCSIFGENERIC: 2841 error = priv_check(td, PRIV_NET_HWIOCTL); 2842 if (error) 2843 return (error); 2844 if (ifp->if_ioctl == NULL) 2845 return (EOPNOTSUPP); 2846 error = (*ifp->if_ioctl)(ifp, cmd, data); 2847 if (error == 0) 2848 getmicrotime(&ifp->if_lastchange); 2849 break; 2850 2851 case SIOCGIFSTATUS: 2852 case SIOCGIFPSRCADDR: 2853 case SIOCGIFPDSTADDR: 2854 case SIOCGIFMEDIA: 2855 case SIOCGIFXMEDIA: 2856 case SIOCGIFGENERIC: 2857 case SIOCGIFRSSKEY: 2858 case SIOCGIFRSSHASH: 2859 case SIOCGIFDOWNREASON: 2860 if (ifp->if_ioctl == NULL) 2861 return (EOPNOTSUPP); 2862 error = (*ifp->if_ioctl)(ifp, cmd, data); 2863 break; 2864 2865 case SIOCSIFLLADDR: 2866 error = priv_check(td, PRIV_NET_SETLLADDR); 2867 if (error) 2868 return (error); 2869 error = if_setlladdr(ifp, 2870 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2871 break; 2872 2873 case SIOCGHWADDR: 2874 error = if_gethwaddr(ifp, ifr); 2875 break; 2876 2877 case SIOCAIFGROUP: 2878 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2879 if (error) 2880 return (error); 2881 error = if_addgroup(ifp, 2882 ((struct ifgroupreq *)data)->ifgr_group); 2883 if (error != 0) 2884 return (error); 2885 break; 2886 2887 case SIOCGIFGROUP: 2888 { 2889 struct epoch_tracker et; 2890 2891 NET_EPOCH_ENTER(et); 2892 error = if_getgroup((struct ifgroupreq *)data, ifp); 2893 NET_EPOCH_EXIT(et); 2894 break; 2895 } 2896 2897 case SIOCDIFGROUP: 2898 error = priv_check(td, PRIV_NET_DELIFGROUP); 2899 if (error) 2900 return (error); 2901 error = if_delgroup(ifp, 2902 ((struct ifgroupreq *)data)->ifgr_group); 2903 if (error != 0) 2904 return (error); 2905 break; 2906 2907 default: 2908 error = ENOIOCTL; 2909 break; 2910 } 2911 return (error); 2912 } 2913 2914 /* 2915 * Interface ioctls. 2916 */ 2917 int 2918 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2919 { 2920 #ifdef COMPAT_FREEBSD32 2921 union { 2922 struct ifconf ifc; 2923 struct ifdrv ifd; 2924 struct ifgroupreq ifgr; 2925 struct ifmediareq ifmr; 2926 } thunk; 2927 u_long saved_cmd; 2928 struct ifconf32 *ifc32; 2929 struct ifdrv32 *ifd32; 2930 struct ifgroupreq32 *ifgr32; 2931 struct ifmediareq32 *ifmr32; 2932 #endif 2933 struct ifnet *ifp; 2934 struct ifreq *ifr; 2935 int error; 2936 int oif_flags; 2937 #ifdef VIMAGE 2938 bool shutdown; 2939 #endif 2940 2941 CURVNET_SET(so->so_vnet); 2942 #ifdef VIMAGE 2943 /* Make sure the VNET is stable. */ 2944 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2945 if (shutdown) { 2946 CURVNET_RESTORE(); 2947 return (EBUSY); 2948 } 2949 #endif 2950 2951 #ifdef COMPAT_FREEBSD32 2952 saved_cmd = cmd; 2953 switch (cmd) { 2954 case SIOCGIFCONF32: 2955 ifc32 = (struct ifconf32 *)data; 2956 thunk.ifc.ifc_len = ifc32->ifc_len; 2957 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2958 data = (caddr_t)&thunk.ifc; 2959 cmd = SIOCGIFCONF; 2960 break; 2961 case SIOCGDRVSPEC32: 2962 case SIOCSDRVSPEC32: 2963 ifd32 = (struct ifdrv32 *)data; 2964 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2965 sizeof(thunk.ifd.ifd_name)); 2966 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2967 thunk.ifd.ifd_len = ifd32->ifd_len; 2968 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2969 data = (caddr_t)&thunk.ifd; 2970 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2971 break; 2972 case SIOCAIFGROUP32: 2973 case SIOCGIFGROUP32: 2974 case SIOCDIFGROUP32: 2975 case SIOCGIFGMEMB32: 2976 ifgr32 = (struct ifgroupreq32 *)data; 2977 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2978 sizeof(thunk.ifgr.ifgr_name)); 2979 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2980 switch (cmd) { 2981 case SIOCAIFGROUP32: 2982 case SIOCDIFGROUP32: 2983 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2984 sizeof(thunk.ifgr.ifgr_group)); 2985 break; 2986 case SIOCGIFGROUP32: 2987 case SIOCGIFGMEMB32: 2988 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2989 break; 2990 } 2991 data = (caddr_t)&thunk.ifgr; 2992 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2993 break; 2994 case SIOCGIFMEDIA32: 2995 case SIOCGIFXMEDIA32: 2996 ifmr32 = (struct ifmediareq32 *)data; 2997 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2998 sizeof(thunk.ifmr.ifm_name)); 2999 thunk.ifmr.ifm_current = ifmr32->ifm_current; 3000 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 3001 thunk.ifmr.ifm_status = ifmr32->ifm_status; 3002 thunk.ifmr.ifm_active = ifmr32->ifm_active; 3003 thunk.ifmr.ifm_count = ifmr32->ifm_count; 3004 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 3005 data = (caddr_t)&thunk.ifmr; 3006 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 3007 break; 3008 } 3009 #endif 3010 3011 switch (cmd) { 3012 case SIOCGIFCONF: 3013 error = ifconf(cmd, data); 3014 goto out_noref; 3015 } 3016 3017 ifr = (struct ifreq *)data; 3018 switch (cmd) { 3019 #ifdef VIMAGE 3020 case SIOCSIFRVNET: 3021 error = priv_check(td, PRIV_NET_SETIFVNET); 3022 if (error == 0) 3023 error = if_vmove_reclaim(td, ifr->ifr_name, 3024 ifr->ifr_jid); 3025 goto out_noref; 3026 #endif 3027 case SIOCIFCREATE: 3028 case SIOCIFCREATE2: 3029 error = priv_check(td, PRIV_NET_IFCREATE); 3030 if (error == 0) 3031 error = if_clone_create(ifr->ifr_name, 3032 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3033 ifr_data_get_ptr(ifr) : NULL); 3034 goto out_noref; 3035 case SIOCIFDESTROY: 3036 error = priv_check(td, PRIV_NET_IFDESTROY); 3037 3038 if (error == 0) { 3039 sx_xlock(&ifnet_detach_sxlock); 3040 error = if_clone_destroy(ifr->ifr_name); 3041 sx_xunlock(&ifnet_detach_sxlock); 3042 } 3043 goto out_noref; 3044 3045 case SIOCIFGCLONERS: 3046 error = if_clone_list((struct if_clonereq *)data); 3047 goto out_noref; 3048 3049 case SIOCGIFGMEMB: 3050 error = if_getgroupmembers((struct ifgroupreq *)data); 3051 goto out_noref; 3052 3053 #if defined(INET) || defined(INET6) 3054 case SIOCSVH: 3055 case SIOCGVH: 3056 if (carp_ioctl_p == NULL) 3057 error = EPROTONOSUPPORT; 3058 else 3059 error = (*carp_ioctl_p)(ifr, cmd, td); 3060 goto out_noref; 3061 #endif 3062 } 3063 3064 ifp = ifunit_ref(ifr->ifr_name); 3065 if (ifp == NULL) { 3066 error = ENXIO; 3067 goto out_noref; 3068 } 3069 3070 error = ifhwioctl(cmd, ifp, data, td); 3071 if (error != ENOIOCTL) 3072 goto out_ref; 3073 3074 oif_flags = ifp->if_flags; 3075 if (so->so_proto == NULL) { 3076 error = EOPNOTSUPP; 3077 goto out_ref; 3078 } 3079 3080 /* 3081 * Pass the request on to the socket control method, and if the 3082 * latter returns EOPNOTSUPP, directly to the interface. 3083 * 3084 * Make an exception for the legacy SIOCSIF* requests. Drivers 3085 * trust SIOCSIFADDR et al to come from an already privileged 3086 * layer, and do not perform any credentials checks or input 3087 * validation. 3088 */ 3089 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 3090 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3091 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3092 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3093 error = (*ifp->if_ioctl)(ifp, cmd, data); 3094 3095 if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP)) 3096 if_up(ifp); 3097 out_ref: 3098 if_rele(ifp); 3099 out_noref: 3100 CURVNET_RESTORE(); 3101 #ifdef COMPAT_FREEBSD32 3102 if (error != 0) 3103 return (error); 3104 switch (saved_cmd) { 3105 case SIOCGIFCONF32: 3106 ifc32->ifc_len = thunk.ifc.ifc_len; 3107 break; 3108 case SIOCGDRVSPEC32: 3109 /* 3110 * SIOCGDRVSPEC is IOWR, but nothing actually touches 3111 * the struct so just assert that ifd_len (the only 3112 * field it might make sense to update) hasn't 3113 * changed. 3114 */ 3115 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 3116 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 3117 thunk.ifd.ifd_len)); 3118 break; 3119 case SIOCGIFGROUP32: 3120 case SIOCGIFGMEMB32: 3121 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 3122 break; 3123 case SIOCGIFMEDIA32: 3124 case SIOCGIFXMEDIA32: 3125 ifmr32->ifm_current = thunk.ifmr.ifm_current; 3126 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 3127 ifmr32->ifm_status = thunk.ifmr.ifm_status; 3128 ifmr32->ifm_active = thunk.ifmr.ifm_active; 3129 ifmr32->ifm_count = thunk.ifmr.ifm_count; 3130 break; 3131 } 3132 #endif 3133 return (error); 3134 } 3135 3136 int 3137 if_rename(struct ifnet *ifp, char *new_name) 3138 { 3139 struct ifaddr *ifa; 3140 struct sockaddr_dl *sdl; 3141 size_t namelen, onamelen; 3142 char old_name[IFNAMSIZ]; 3143 char strbuf[IFNAMSIZ + 8]; 3144 3145 if (new_name[0] == '\0') 3146 return (EINVAL); 3147 if (strcmp(new_name, ifp->if_xname) == 0) 3148 return (0); 3149 if (ifunit(new_name) != NULL) 3150 return (EEXIST); 3151 3152 /* 3153 * XXX: Locking. Nothing else seems to lock if_flags, 3154 * and there are numerous other races with the 3155 * ifunit() checks not being atomic with namespace 3156 * changes (renames, vmoves, if_attach, etc). 3157 */ 3158 ifp->if_flags |= IFF_RENAMING; 3159 3160 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 3161 3162 if_printf(ifp, "changing name to '%s'\n", new_name); 3163 3164 IF_ADDR_WLOCK(ifp); 3165 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 3166 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 3167 ifa = ifp->if_addr; 3168 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3169 namelen = strlen(new_name); 3170 onamelen = sdl->sdl_nlen; 3171 /* 3172 * Move the address if needed. This is safe because we 3173 * allocate space for a name of length IFNAMSIZ when we 3174 * create this in if_attach(). 3175 */ 3176 if (namelen != onamelen) { 3177 bcopy(sdl->sdl_data + onamelen, 3178 sdl->sdl_data + namelen, sdl->sdl_alen); 3179 } 3180 bcopy(new_name, sdl->sdl_data, namelen); 3181 sdl->sdl_nlen = namelen; 3182 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 3183 bzero(sdl->sdl_data, onamelen); 3184 while (namelen != 0) 3185 sdl->sdl_data[--namelen] = 0xff; 3186 IF_ADDR_WUNLOCK(ifp); 3187 3188 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 3189 3190 ifp->if_flags &= ~IFF_RENAMING; 3191 3192 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 3193 devctl_notify("IFNET", old_name, "RENAME", strbuf); 3194 3195 return (0); 3196 } 3197 3198 /* 3199 * The code common to handling reference counted flags, 3200 * e.g., in ifpromisc() and if_allmulti(). 3201 * The "pflag" argument can specify a permanent mode flag to check, 3202 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3203 * 3204 * Only to be used on stack-owned flags, not driver-owned flags. 3205 */ 3206 static int 3207 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3208 { 3209 struct ifreq ifr; 3210 int error; 3211 int oldflags, oldcount; 3212 3213 /* Sanity checks to catch programming errors */ 3214 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3215 ("%s: setting driver-owned flag %d", __func__, flag)); 3216 3217 if (onswitch) 3218 KASSERT(*refcount >= 0, 3219 ("%s: increment negative refcount %d for flag %d", 3220 __func__, *refcount, flag)); 3221 else 3222 KASSERT(*refcount > 0, 3223 ("%s: decrement non-positive refcount %d for flag %d", 3224 __func__, *refcount, flag)); 3225 3226 /* In case this mode is permanent, just touch refcount */ 3227 if (ifp->if_flags & pflag) { 3228 *refcount += onswitch ? 1 : -1; 3229 return (0); 3230 } 3231 3232 /* Save ifnet parameters for if_ioctl() may fail */ 3233 oldcount = *refcount; 3234 oldflags = ifp->if_flags; 3235 3236 /* 3237 * See if we aren't the only and touching refcount is enough. 3238 * Actually toggle interface flag if we are the first or last. 3239 */ 3240 if (onswitch) { 3241 if ((*refcount)++) 3242 return (0); 3243 ifp->if_flags |= flag; 3244 } else { 3245 if (--(*refcount)) 3246 return (0); 3247 ifp->if_flags &= ~flag; 3248 } 3249 3250 /* Call down the driver since we've changed interface flags */ 3251 if (ifp->if_ioctl == NULL) { 3252 error = EOPNOTSUPP; 3253 goto recover; 3254 } 3255 ifr.ifr_flags = ifp->if_flags & 0xffff; 3256 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3257 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3258 if (error) 3259 goto recover; 3260 /* Notify userland that interface flags have changed */ 3261 rt_ifmsg(ifp, flag); 3262 return (0); 3263 3264 recover: 3265 /* Recover after driver error */ 3266 *refcount = oldcount; 3267 ifp->if_flags = oldflags; 3268 return (error); 3269 } 3270 3271 /* 3272 * Set/clear promiscuous mode on interface ifp based on the truth value 3273 * of pswitch. The calls are reference counted so that only the first 3274 * "on" request actually has an effect, as does the final "off" request. 3275 * Results are undefined if the "off" and "on" requests are not matched. 3276 */ 3277 int 3278 ifpromisc(struct ifnet *ifp, int pswitch) 3279 { 3280 int error; 3281 int oldflags = ifp->if_flags; 3282 3283 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3284 &ifp->if_pcount, pswitch); 3285 /* If promiscuous mode status has changed, log a message */ 3286 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3287 log_promisc_mode_change) 3288 if_printf(ifp, "promiscuous mode %s\n", 3289 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3290 return (error); 3291 } 3292 3293 /* 3294 * Return interface configuration 3295 * of system. List may be used 3296 * in later ioctl's (above) to get 3297 * other information. 3298 */ 3299 /*ARGSUSED*/ 3300 static int 3301 ifconf(u_long cmd, caddr_t data) 3302 { 3303 struct ifconf *ifc = (struct ifconf *)data; 3304 struct ifnet *ifp; 3305 struct ifaddr *ifa; 3306 struct ifreq ifr; 3307 struct sbuf *sb; 3308 int error, full = 0, valid_len, max_len; 3309 3310 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3311 max_len = maxphys - 1; 3312 3313 /* Prevent hostile input from being able to crash the system */ 3314 if (ifc->ifc_len <= 0) 3315 return (EINVAL); 3316 3317 again: 3318 if (ifc->ifc_len <= max_len) { 3319 max_len = ifc->ifc_len; 3320 full = 1; 3321 } 3322 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3323 max_len = 0; 3324 valid_len = 0; 3325 3326 IFNET_RLOCK(); 3327 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3328 struct epoch_tracker et; 3329 int addrs; 3330 3331 /* 3332 * Zero the ifr to make sure we don't disclose the contents 3333 * of the stack. 3334 */ 3335 memset(&ifr, 0, sizeof(ifr)); 3336 3337 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3338 >= sizeof(ifr.ifr_name)) { 3339 sbuf_delete(sb); 3340 IFNET_RUNLOCK(); 3341 return (ENAMETOOLONG); 3342 } 3343 3344 addrs = 0; 3345 NET_EPOCH_ENTER(et); 3346 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3347 struct sockaddr *sa = ifa->ifa_addr; 3348 3349 if (prison_if(curthread->td_ucred, sa) != 0) 3350 continue; 3351 addrs++; 3352 if (sa->sa_len <= sizeof(*sa)) { 3353 if (sa->sa_len < sizeof(*sa)) { 3354 memset(&ifr.ifr_ifru.ifru_addr, 0, 3355 sizeof(ifr.ifr_ifru.ifru_addr)); 3356 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3357 sa->sa_len); 3358 } else 3359 ifr.ifr_ifru.ifru_addr = *sa; 3360 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3361 max_len += sizeof(ifr); 3362 } else { 3363 sbuf_bcat(sb, &ifr, 3364 offsetof(struct ifreq, ifr_addr)); 3365 max_len += offsetof(struct ifreq, ifr_addr); 3366 sbuf_bcat(sb, sa, sa->sa_len); 3367 max_len += sa->sa_len; 3368 } 3369 3370 if (sbuf_error(sb) == 0) 3371 valid_len = sbuf_len(sb); 3372 } 3373 NET_EPOCH_EXIT(et); 3374 if (addrs == 0) { 3375 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3376 max_len += sizeof(ifr); 3377 3378 if (sbuf_error(sb) == 0) 3379 valid_len = sbuf_len(sb); 3380 } 3381 } 3382 IFNET_RUNLOCK(); 3383 3384 /* 3385 * If we didn't allocate enough space (uncommon), try again. If 3386 * we have already allocated as much space as we are allowed, 3387 * return what we've got. 3388 */ 3389 if (valid_len != max_len && !full) { 3390 sbuf_delete(sb); 3391 goto again; 3392 } 3393 3394 ifc->ifc_len = valid_len; 3395 sbuf_finish(sb); 3396 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3397 sbuf_delete(sb); 3398 return (error); 3399 } 3400 3401 /* 3402 * Just like ifpromisc(), but for all-multicast-reception mode. 3403 */ 3404 int 3405 if_allmulti(struct ifnet *ifp, int onswitch) 3406 { 3407 3408 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3409 } 3410 3411 struct ifmultiaddr * 3412 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3413 { 3414 struct ifmultiaddr *ifma; 3415 3416 IF_ADDR_LOCK_ASSERT(ifp); 3417 3418 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3419 if (sa->sa_family == AF_LINK) { 3420 if (sa_dl_equal(ifma->ifma_addr, sa)) 3421 break; 3422 } else { 3423 if (sa_equal(ifma->ifma_addr, sa)) 3424 break; 3425 } 3426 } 3427 3428 return ifma; 3429 } 3430 3431 /* 3432 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3433 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3434 * the ifnet multicast address list here, so the caller must do that and 3435 * other setup work (such as notifying the device driver). The reference 3436 * count is initialized to 1. 3437 */ 3438 static struct ifmultiaddr * 3439 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3440 int mflags) 3441 { 3442 struct ifmultiaddr *ifma; 3443 struct sockaddr *dupsa; 3444 3445 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3446 M_ZERO); 3447 if (ifma == NULL) 3448 return (NULL); 3449 3450 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3451 if (dupsa == NULL) { 3452 free(ifma, M_IFMADDR); 3453 return (NULL); 3454 } 3455 bcopy(sa, dupsa, sa->sa_len); 3456 ifma->ifma_addr = dupsa; 3457 3458 ifma->ifma_ifp = ifp; 3459 ifma->ifma_refcount = 1; 3460 ifma->ifma_protospec = NULL; 3461 3462 if (llsa == NULL) { 3463 ifma->ifma_lladdr = NULL; 3464 return (ifma); 3465 } 3466 3467 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3468 if (dupsa == NULL) { 3469 free(ifma->ifma_addr, M_IFMADDR); 3470 free(ifma, M_IFMADDR); 3471 return (NULL); 3472 } 3473 bcopy(llsa, dupsa, llsa->sa_len); 3474 ifma->ifma_lladdr = dupsa; 3475 3476 return (ifma); 3477 } 3478 3479 /* 3480 * if_freemulti: free ifmultiaddr structure and possibly attached related 3481 * addresses. The caller is responsible for implementing reference 3482 * counting, notifying the driver, handling routing messages, and releasing 3483 * any dependent link layer state. 3484 */ 3485 #ifdef MCAST_VERBOSE 3486 extern void kdb_backtrace(void); 3487 #endif 3488 static void 3489 if_freemulti_internal(struct ifmultiaddr *ifma) 3490 { 3491 3492 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3493 ifma->ifma_refcount)); 3494 3495 if (ifma->ifma_lladdr != NULL) 3496 free(ifma->ifma_lladdr, M_IFMADDR); 3497 #ifdef MCAST_VERBOSE 3498 kdb_backtrace(); 3499 printf("%s freeing ifma: %p\n", __func__, ifma); 3500 #endif 3501 free(ifma->ifma_addr, M_IFMADDR); 3502 free(ifma, M_IFMADDR); 3503 } 3504 3505 static void 3506 if_destroymulti(epoch_context_t ctx) 3507 { 3508 struct ifmultiaddr *ifma; 3509 3510 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3511 if_freemulti_internal(ifma); 3512 } 3513 3514 void 3515 if_freemulti(struct ifmultiaddr *ifma) 3516 { 3517 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3518 ifma->ifma_refcount)); 3519 3520 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3521 } 3522 3523 /* 3524 * Register an additional multicast address with a network interface. 3525 * 3526 * - If the address is already present, bump the reference count on the 3527 * address and return. 3528 * - If the address is not link-layer, look up a link layer address. 3529 * - Allocate address structures for one or both addresses, and attach to the 3530 * multicast address list on the interface. If automatically adding a link 3531 * layer address, the protocol address will own a reference to the link 3532 * layer address, to be freed when it is freed. 3533 * - Notify the network device driver of an addition to the multicast address 3534 * list. 3535 * 3536 * 'sa' points to caller-owned memory with the desired multicast address. 3537 * 3538 * 'retifma' will be used to return a pointer to the resulting multicast 3539 * address reference, if desired. 3540 */ 3541 int 3542 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3543 struct ifmultiaddr **retifma) 3544 { 3545 struct ifmultiaddr *ifma, *ll_ifma; 3546 struct sockaddr *llsa; 3547 struct sockaddr_dl sdl; 3548 int error; 3549 3550 #ifdef INET 3551 IN_MULTI_LIST_UNLOCK_ASSERT(); 3552 #endif 3553 #ifdef INET6 3554 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3555 #endif 3556 /* 3557 * If the address is already present, return a new reference to it; 3558 * otherwise, allocate storage and set up a new address. 3559 */ 3560 IF_ADDR_WLOCK(ifp); 3561 ifma = if_findmulti(ifp, sa); 3562 if (ifma != NULL) { 3563 ifma->ifma_refcount++; 3564 if (retifma != NULL) 3565 *retifma = ifma; 3566 IF_ADDR_WUNLOCK(ifp); 3567 return (0); 3568 } 3569 3570 /* 3571 * The address isn't already present; resolve the protocol address 3572 * into a link layer address, and then look that up, bump its 3573 * refcount or allocate an ifma for that also. 3574 * Most link layer resolving functions returns address data which 3575 * fits inside default sockaddr_dl structure. However callback 3576 * can allocate another sockaddr structure, in that case we need to 3577 * free it later. 3578 */ 3579 llsa = NULL; 3580 ll_ifma = NULL; 3581 if (ifp->if_resolvemulti != NULL) { 3582 /* Provide called function with buffer size information */ 3583 sdl.sdl_len = sizeof(sdl); 3584 llsa = (struct sockaddr *)&sdl; 3585 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3586 if (error) 3587 goto unlock_out; 3588 } 3589 3590 /* 3591 * Allocate the new address. Don't hook it up yet, as we may also 3592 * need to allocate a link layer multicast address. 3593 */ 3594 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3595 if (ifma == NULL) { 3596 error = ENOMEM; 3597 goto free_llsa_out; 3598 } 3599 3600 /* 3601 * If a link layer address is found, we'll need to see if it's 3602 * already present in the address list, or allocate is as well. 3603 * When this block finishes, the link layer address will be on the 3604 * list. 3605 */ 3606 if (llsa != NULL) { 3607 ll_ifma = if_findmulti(ifp, llsa); 3608 if (ll_ifma == NULL) { 3609 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3610 if (ll_ifma == NULL) { 3611 --ifma->ifma_refcount; 3612 if_freemulti(ifma); 3613 error = ENOMEM; 3614 goto free_llsa_out; 3615 } 3616 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3617 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3618 ifma_link); 3619 } else 3620 ll_ifma->ifma_refcount++; 3621 ifma->ifma_llifma = ll_ifma; 3622 } 3623 3624 /* 3625 * We now have a new multicast address, ifma, and possibly a new or 3626 * referenced link layer address. Add the primary address to the 3627 * ifnet address list. 3628 */ 3629 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3630 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3631 3632 if (retifma != NULL) 3633 *retifma = ifma; 3634 3635 /* 3636 * Must generate the message while holding the lock so that 'ifma' 3637 * pointer is still valid. 3638 */ 3639 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3640 IF_ADDR_WUNLOCK(ifp); 3641 3642 /* 3643 * We are certain we have added something, so call down to the 3644 * interface to let them know about it. 3645 */ 3646 if (ifp->if_ioctl != NULL) { 3647 if (THREAD_CAN_SLEEP()) 3648 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3649 else 3650 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3651 } 3652 3653 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3654 link_free_sdl(llsa); 3655 3656 return (0); 3657 3658 free_llsa_out: 3659 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3660 link_free_sdl(llsa); 3661 3662 unlock_out: 3663 IF_ADDR_WUNLOCK(ifp); 3664 return (error); 3665 } 3666 3667 static void 3668 if_siocaddmulti(void *arg, int pending) 3669 { 3670 struct ifnet *ifp; 3671 3672 ifp = arg; 3673 #ifdef DIAGNOSTIC 3674 if (pending > 1) 3675 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3676 #endif 3677 CURVNET_SET(ifp->if_vnet); 3678 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3679 CURVNET_RESTORE(); 3680 } 3681 3682 /* 3683 * Delete a multicast group membership by network-layer group address. 3684 * 3685 * Returns ENOENT if the entry could not be found. If ifp no longer 3686 * exists, results are undefined. This entry point should only be used 3687 * from subsystems which do appropriate locking to hold ifp for the 3688 * duration of the call. 3689 * Network-layer protocol domains must use if_delmulti_ifma(). 3690 */ 3691 int 3692 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3693 { 3694 struct ifmultiaddr *ifma; 3695 int lastref; 3696 3697 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3698 3699 IF_ADDR_WLOCK(ifp); 3700 lastref = 0; 3701 ifma = if_findmulti(ifp, sa); 3702 if (ifma != NULL) 3703 lastref = if_delmulti_locked(ifp, ifma, 0); 3704 IF_ADDR_WUNLOCK(ifp); 3705 3706 if (ifma == NULL) 3707 return (ENOENT); 3708 3709 if (lastref && ifp->if_ioctl != NULL) { 3710 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3711 } 3712 3713 return (0); 3714 } 3715 3716 /* 3717 * Delete all multicast group membership for an interface. 3718 * Should be used to quickly flush all multicast filters. 3719 */ 3720 void 3721 if_delallmulti(struct ifnet *ifp) 3722 { 3723 struct ifmultiaddr *ifma; 3724 struct ifmultiaddr *next; 3725 3726 IF_ADDR_WLOCK(ifp); 3727 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3728 if_delmulti_locked(ifp, ifma, 0); 3729 IF_ADDR_WUNLOCK(ifp); 3730 } 3731 3732 void 3733 if_delmulti_ifma(struct ifmultiaddr *ifma) 3734 { 3735 if_delmulti_ifma_flags(ifma, 0); 3736 } 3737 3738 /* 3739 * Delete a multicast group membership by group membership pointer. 3740 * Network-layer protocol domains must use this routine. 3741 * 3742 * It is safe to call this routine if the ifp disappeared. 3743 */ 3744 void 3745 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3746 { 3747 struct ifnet *ifp; 3748 int lastref; 3749 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3750 #ifdef INET 3751 IN_MULTI_LIST_UNLOCK_ASSERT(); 3752 #endif 3753 ifp = ifma->ifma_ifp; 3754 #ifdef DIAGNOSTIC 3755 if (ifp == NULL) { 3756 printf("%s: ifma_ifp seems to be detached\n", __func__); 3757 } else { 3758 struct epoch_tracker et; 3759 struct ifnet *oifp; 3760 3761 NET_EPOCH_ENTER(et); 3762 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3763 if (ifp == oifp) 3764 break; 3765 NET_EPOCH_EXIT(et); 3766 if (ifp != oifp) 3767 ifp = NULL; 3768 } 3769 #endif 3770 /* 3771 * If and only if the ifnet instance exists: Acquire the address lock. 3772 */ 3773 if (ifp != NULL) 3774 IF_ADDR_WLOCK(ifp); 3775 3776 lastref = if_delmulti_locked(ifp, ifma, flags); 3777 3778 if (ifp != NULL) { 3779 /* 3780 * If and only if the ifnet instance exists: 3781 * Release the address lock. 3782 * If the group was left: update the hardware hash filter. 3783 */ 3784 IF_ADDR_WUNLOCK(ifp); 3785 if (lastref && ifp->if_ioctl != NULL) { 3786 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3787 } 3788 } 3789 } 3790 3791 /* 3792 * Perform deletion of network-layer and/or link-layer multicast address. 3793 * 3794 * Return 0 if the reference count was decremented. 3795 * Return 1 if the final reference was released, indicating that the 3796 * hardware hash filter should be reprogrammed. 3797 */ 3798 static int 3799 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3800 { 3801 struct ifmultiaddr *ll_ifma; 3802 3803 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3804 KASSERT(ifma->ifma_ifp == ifp, 3805 ("%s: inconsistent ifp %p", __func__, ifp)); 3806 IF_ADDR_WLOCK_ASSERT(ifp); 3807 } 3808 3809 ifp = ifma->ifma_ifp; 3810 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3811 3812 /* 3813 * If the ifnet is detaching, null out references to ifnet, 3814 * so that upper protocol layers will notice, and not attempt 3815 * to obtain locks for an ifnet which no longer exists. The 3816 * routing socket announcement must happen before the ifnet 3817 * instance is detached from the system. 3818 */ 3819 if (detaching) { 3820 #ifdef DIAGNOSTIC 3821 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3822 #endif 3823 /* 3824 * ifp may already be nulled out if we are being reentered 3825 * to delete the ll_ifma. 3826 */ 3827 if (ifp != NULL) { 3828 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3829 ifma->ifma_ifp = NULL; 3830 } 3831 } 3832 3833 if (--ifma->ifma_refcount > 0) 3834 return 0; 3835 3836 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3837 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3838 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3839 } 3840 /* 3841 * If this ifma is a network-layer ifma, a link-layer ifma may 3842 * have been associated with it. Release it first if so. 3843 */ 3844 ll_ifma = ifma->ifma_llifma; 3845 if (ll_ifma != NULL) { 3846 KASSERT(ifma->ifma_lladdr != NULL, 3847 ("%s: llifma w/o lladdr", __func__)); 3848 if (detaching) 3849 ll_ifma->ifma_ifp = NULL; /* XXX */ 3850 if (--ll_ifma->ifma_refcount == 0) { 3851 if (ifp != NULL) { 3852 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3853 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3854 ifma_link); 3855 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3856 } 3857 } 3858 if_freemulti(ll_ifma); 3859 } 3860 } 3861 #ifdef INVARIANTS 3862 if (ifp) { 3863 struct ifmultiaddr *ifmatmp; 3864 3865 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3866 MPASS(ifma != ifmatmp); 3867 } 3868 #endif 3869 if_freemulti(ifma); 3870 /* 3871 * The last reference to this instance of struct ifmultiaddr 3872 * was released; the hardware should be notified of this change. 3873 */ 3874 return 1; 3875 } 3876 3877 /* 3878 * Set the link layer address on an interface. 3879 * 3880 * At this time we only support certain types of interfaces, 3881 * and we don't allow the length of the address to change. 3882 * 3883 * Set noinline to be dtrace-friendly 3884 */ 3885 __noinline int 3886 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3887 { 3888 struct sockaddr_dl *sdl; 3889 struct ifaddr *ifa; 3890 struct ifreq ifr; 3891 3892 ifa = ifp->if_addr; 3893 if (ifa == NULL) 3894 return (EINVAL); 3895 3896 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3897 if (sdl == NULL) 3898 return (EINVAL); 3899 3900 if (len != sdl->sdl_alen) /* don't allow length to change */ 3901 return (EINVAL); 3902 3903 switch (ifp->if_type) { 3904 case IFT_ETHER: 3905 case IFT_XETHER: 3906 case IFT_L2VLAN: 3907 case IFT_BRIDGE: 3908 case IFT_IEEE8023ADLAG: 3909 bcopy(lladdr, LLADDR(sdl), len); 3910 break; 3911 default: 3912 return (ENODEV); 3913 } 3914 3915 /* 3916 * If the interface is already up, we need 3917 * to re-init it in order to reprogram its 3918 * address filter. 3919 */ 3920 if ((ifp->if_flags & IFF_UP) != 0) { 3921 if (ifp->if_ioctl) { 3922 ifp->if_flags &= ~IFF_UP; 3923 ifr.ifr_flags = ifp->if_flags & 0xffff; 3924 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3925 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3926 ifp->if_flags |= IFF_UP; 3927 ifr.ifr_flags = ifp->if_flags & 0xffff; 3928 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3929 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3930 } 3931 } 3932 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3933 3934 return (0); 3935 } 3936 3937 /* 3938 * Compat function for handling basic encapsulation requests. 3939 * Not converted stacks (FDDI, IB, ..) supports traditional 3940 * output model: ARP (and other similar L2 protocols) are handled 3941 * inside output routine, arpresolve/nd6_resolve() returns MAC 3942 * address instead of full prepend. 3943 * 3944 * This function creates calculated header==MAC for IPv4/IPv6 and 3945 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3946 * address families. 3947 */ 3948 static int 3949 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3950 { 3951 if (req->rtype != IFENCAP_LL) 3952 return (EOPNOTSUPP); 3953 3954 if (req->bufsize < req->lladdr_len) 3955 return (ENOMEM); 3956 3957 switch (req->family) { 3958 case AF_INET: 3959 case AF_INET6: 3960 break; 3961 default: 3962 return (EAFNOSUPPORT); 3963 } 3964 3965 /* Copy lladdr to storage as is */ 3966 memmove(req->buf, req->lladdr, req->lladdr_len); 3967 req->bufsize = req->lladdr_len; 3968 req->lladdr_off = 0; 3969 3970 return (0); 3971 } 3972 3973 /* 3974 * Tunnel interfaces can nest, also they may cause infinite recursion 3975 * calls when misconfigured. We'll prevent this by detecting loops. 3976 * High nesting level may cause stack exhaustion. We'll prevent this 3977 * by introducing upper limit. 3978 * 3979 * Return 0, if tunnel nesting count is equal or less than limit. 3980 */ 3981 int 3982 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3983 int limit) 3984 { 3985 struct m_tag *mtag; 3986 int count; 3987 3988 count = 1; 3989 mtag = NULL; 3990 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3991 if (*(struct ifnet **)(mtag + 1) == ifp) { 3992 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3993 return (EIO); 3994 } 3995 count++; 3996 } 3997 if (count > limit) { 3998 log(LOG_NOTICE, 3999 "%s: if_output recursively called too many times(%d)\n", 4000 if_name(ifp), count); 4001 return (EIO); 4002 } 4003 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 4004 if (mtag == NULL) 4005 return (ENOMEM); 4006 *(struct ifnet **)(mtag + 1) = ifp; 4007 m_tag_prepend(m, mtag); 4008 return (0); 4009 } 4010 4011 /* 4012 * Get the link layer address that was read from the hardware at attach. 4013 * 4014 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 4015 * their component interfaces as IFT_IEEE8023ADLAG. 4016 */ 4017 int 4018 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 4019 { 4020 if (ifp->if_hw_addr == NULL) 4021 return (ENODEV); 4022 4023 switch (ifp->if_type) { 4024 case IFT_ETHER: 4025 case IFT_IEEE8023ADLAG: 4026 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 4027 return (0); 4028 default: 4029 return (ENODEV); 4030 } 4031 } 4032 4033 /* 4034 * The name argument must be a pointer to storage which will last as 4035 * long as the interface does. For physical devices, the result of 4036 * device_get_name(dev) is a good choice and for pseudo-devices a 4037 * static string works well. 4038 */ 4039 void 4040 if_initname(struct ifnet *ifp, const char *name, int unit) 4041 { 4042 ifp->if_dname = name; 4043 ifp->if_dunit = unit; 4044 if (unit != IF_DUNIT_NONE) 4045 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 4046 else 4047 strlcpy(ifp->if_xname, name, IFNAMSIZ); 4048 } 4049 4050 static int 4051 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 4052 { 4053 char if_fmt[256]; 4054 4055 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 4056 vlog(pri, if_fmt, ap); 4057 return (0); 4058 } 4059 4060 4061 int 4062 if_printf(struct ifnet *ifp, const char *fmt, ...) 4063 { 4064 va_list ap; 4065 4066 va_start(ap, fmt); 4067 if_vlog(ifp, LOG_INFO, fmt, ap); 4068 va_end(ap); 4069 return (0); 4070 } 4071 4072 int 4073 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 4074 { 4075 va_list ap; 4076 4077 va_start(ap, fmt); 4078 if_vlog(ifp, pri, fmt, ap); 4079 va_end(ap); 4080 return (0); 4081 } 4082 4083 void 4084 if_start(struct ifnet *ifp) 4085 { 4086 4087 (*(ifp)->if_start)(ifp); 4088 } 4089 4090 /* 4091 * Backwards compatibility interface for drivers 4092 * that have not implemented it 4093 */ 4094 static int 4095 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 4096 { 4097 int error; 4098 4099 IFQ_HANDOFF(ifp, m, error); 4100 return (error); 4101 } 4102 4103 static void 4104 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4105 { 4106 m_freem(m); 4107 } 4108 4109 int 4110 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4111 { 4112 int active = 0; 4113 4114 IF_LOCK(ifq); 4115 if (_IF_QFULL(ifq)) { 4116 IF_UNLOCK(ifq); 4117 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4118 m_freem(m); 4119 return (0); 4120 } 4121 if (ifp != NULL) { 4122 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4123 if (m->m_flags & (M_BCAST|M_MCAST)) 4124 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4125 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4126 } 4127 _IF_ENQUEUE(ifq, m); 4128 IF_UNLOCK(ifq); 4129 if (ifp != NULL && !active) 4130 (*(ifp)->if_start)(ifp); 4131 return (1); 4132 } 4133 4134 void 4135 if_register_com_alloc(u_char type, 4136 if_com_alloc_t *a, if_com_free_t *f) 4137 { 4138 4139 KASSERT(if_com_alloc[type] == NULL, 4140 ("if_register_com_alloc: %d already registered", type)); 4141 KASSERT(if_com_free[type] == NULL, 4142 ("if_register_com_alloc: %d free already registered", type)); 4143 4144 if_com_alloc[type] = a; 4145 if_com_free[type] = f; 4146 } 4147 4148 void 4149 if_deregister_com_alloc(u_char type) 4150 { 4151 4152 KASSERT(if_com_alloc[type] != NULL, 4153 ("if_deregister_com_alloc: %d not registered", type)); 4154 KASSERT(if_com_free[type] != NULL, 4155 ("if_deregister_com_alloc: %d free not registered", type)); 4156 4157 /* 4158 * Ensure all pending EPOCH(9) callbacks have been executed. This 4159 * fixes issues about late invocation of if_destroy(), which leads 4160 * to memory leak from if_com_alloc[type] allocated if_l2com. 4161 */ 4162 NET_EPOCH_DRAIN_CALLBACKS(); 4163 4164 if_com_alloc[type] = NULL; 4165 if_com_free[type] = NULL; 4166 } 4167 4168 /* API for driver access to network stack owned ifnet.*/ 4169 uint64_t 4170 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4171 { 4172 uint64_t oldbrate; 4173 4174 oldbrate = ifp->if_baudrate; 4175 ifp->if_baudrate = baudrate; 4176 return (oldbrate); 4177 } 4178 4179 uint64_t 4180 if_getbaudrate(const if_t ifp) 4181 { 4182 return (ifp->if_baudrate); 4183 } 4184 4185 int 4186 if_setcapabilities(if_t ifp, int capabilities) 4187 { 4188 ifp->if_capabilities = capabilities; 4189 return (0); 4190 } 4191 4192 int 4193 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4194 { 4195 ifp->if_capabilities &= ~clearbit; 4196 ifp->if_capabilities |= setbit; 4197 return (0); 4198 } 4199 4200 int 4201 if_getcapabilities(const if_t ifp) 4202 { 4203 return (ifp->if_capabilities); 4204 } 4205 4206 int 4207 if_setcapenable(if_t ifp, int capabilities) 4208 { 4209 ifp->if_capenable = capabilities; 4210 return (0); 4211 } 4212 4213 int 4214 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4215 { 4216 ifp->if_capenable &= ~clearcap; 4217 ifp->if_capenable |= setcap; 4218 return (0); 4219 } 4220 4221 int 4222 if_setcapabilities2(if_t ifp, int capabilities) 4223 { 4224 ifp->if_capabilities2 = capabilities; 4225 return (0); 4226 } 4227 4228 int 4229 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit) 4230 { 4231 ifp->if_capabilities2 &= ~clearbit; 4232 ifp->if_capabilities2 |= setbit; 4233 return (0); 4234 } 4235 4236 int 4237 if_getcapabilities2(const if_t ifp) 4238 { 4239 return (ifp->if_capabilities2); 4240 } 4241 4242 int 4243 if_setcapenable2(if_t ifp, int capabilities2) 4244 { 4245 ifp->if_capenable2 = capabilities2; 4246 return (0); 4247 } 4248 4249 int 4250 if_setcapenable2bit(if_t ifp, int setcap, int clearcap) 4251 { 4252 ifp->if_capenable2 &= ~clearcap; 4253 ifp->if_capenable2 |= setcap; 4254 return (0); 4255 } 4256 4257 const char * 4258 if_getdname(const if_t ifp) 4259 { 4260 return (ifp->if_dname); 4261 } 4262 4263 void 4264 if_setdname(if_t ifp, const char *dname) 4265 { 4266 ifp->if_dname = dname; 4267 } 4268 4269 const char * 4270 if_name(if_t ifp) 4271 { 4272 return (ifp->if_xname); 4273 } 4274 4275 int 4276 if_setname(if_t ifp, const char *name) 4277 { 4278 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4279 return (ENAMETOOLONG); 4280 strcpy(ifp->if_xname, name); 4281 4282 return (0); 4283 } 4284 4285 int 4286 if_togglecapenable(if_t ifp, int togglecap) 4287 { 4288 ifp->if_capenable ^= togglecap; 4289 return (0); 4290 } 4291 4292 int 4293 if_getcapenable(const if_t ifp) 4294 { 4295 return (ifp->if_capenable); 4296 } 4297 4298 int 4299 if_togglecapenable2(if_t ifp, int togglecap) 4300 { 4301 ifp->if_capenable2 ^= togglecap; 4302 return (0); 4303 } 4304 4305 int 4306 if_getcapenable2(const if_t ifp) 4307 { 4308 return (ifp->if_capenable2); 4309 } 4310 4311 int 4312 if_getdunit(const if_t ifp) 4313 { 4314 return (ifp->if_dunit); 4315 } 4316 4317 int 4318 if_getindex(const if_t ifp) 4319 { 4320 return (ifp->if_index); 4321 } 4322 4323 int 4324 if_getidxgen(const if_t ifp) 4325 { 4326 return (ifp->if_idxgen); 4327 } 4328 4329 const char * 4330 if_getdescr(if_t ifp) 4331 { 4332 return (ifp->if_description); 4333 } 4334 4335 void 4336 if_setdescr(if_t ifp, char *descrbuf) 4337 { 4338 sx_xlock(&ifdescr_sx); 4339 char *odescrbuf = ifp->if_description; 4340 ifp->if_description = descrbuf; 4341 sx_xunlock(&ifdescr_sx); 4342 4343 if_freedescr(odescrbuf); 4344 } 4345 4346 char * 4347 if_allocdescr(size_t sz, int malloc_flag) 4348 { 4349 malloc_flag &= (M_WAITOK | M_NOWAIT); 4350 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4351 } 4352 4353 void 4354 if_freedescr(char *descrbuf) 4355 { 4356 free(descrbuf, M_IFDESCR); 4357 } 4358 4359 int 4360 if_getalloctype(const if_t ifp) 4361 { 4362 return (ifp->if_alloctype); 4363 } 4364 4365 void 4366 if_setlastchange(if_t ifp) 4367 { 4368 getmicrotime(&ifp->if_lastchange); 4369 } 4370 4371 /* 4372 * This is largely undesirable because it ties ifnet to a device, but does 4373 * provide flexiblity for an embedded product vendor. Should be used with 4374 * the understanding that it violates the interface boundaries, and should be 4375 * a last resort only. 4376 */ 4377 int 4378 if_setdev(if_t ifp, void *dev) 4379 { 4380 return (0); 4381 } 4382 4383 int 4384 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4385 { 4386 ifp->if_drv_flags &= ~clear_flags; 4387 ifp->if_drv_flags |= set_flags; 4388 4389 return (0); 4390 } 4391 4392 int 4393 if_getdrvflags(const if_t ifp) 4394 { 4395 return (ifp->if_drv_flags); 4396 } 4397 4398 int 4399 if_setdrvflags(if_t ifp, int flags) 4400 { 4401 ifp->if_drv_flags = flags; 4402 return (0); 4403 } 4404 4405 int 4406 if_setflags(if_t ifp, int flags) 4407 { 4408 ifp->if_flags = flags; 4409 return (0); 4410 } 4411 4412 int 4413 if_setflagbits(if_t ifp, int set, int clear) 4414 { 4415 ifp->if_flags &= ~clear; 4416 ifp->if_flags |= set; 4417 return (0); 4418 } 4419 4420 int 4421 if_getflags(const if_t ifp) 4422 { 4423 return (ifp->if_flags); 4424 } 4425 4426 int 4427 if_clearhwassist(if_t ifp) 4428 { 4429 ifp->if_hwassist = 0; 4430 return (0); 4431 } 4432 4433 int 4434 if_sethwassistbits(if_t ifp, int toset, int toclear) 4435 { 4436 ifp->if_hwassist &= ~toclear; 4437 ifp->if_hwassist |= toset; 4438 4439 return (0); 4440 } 4441 4442 int 4443 if_sethwassist(if_t ifp, int hwassist_bit) 4444 { 4445 ifp->if_hwassist = hwassist_bit; 4446 return (0); 4447 } 4448 4449 int 4450 if_gethwassist(const if_t ifp) 4451 { 4452 return (ifp->if_hwassist); 4453 } 4454 4455 int 4456 if_togglehwassist(if_t ifp, int toggle_bits) 4457 { 4458 ifp->if_hwassist ^= toggle_bits; 4459 return (0); 4460 } 4461 4462 int 4463 if_setmtu(if_t ifp, int mtu) 4464 { 4465 ifp->if_mtu = mtu; 4466 return (0); 4467 } 4468 4469 void 4470 if_notifymtu(if_t ifp) 4471 { 4472 #ifdef INET6 4473 nd6_setmtu(ifp); 4474 #endif 4475 rt_updatemtu(ifp); 4476 } 4477 4478 int 4479 if_getmtu(const if_t ifp) 4480 { 4481 return (ifp->if_mtu); 4482 } 4483 4484 int 4485 if_getmtu_family(const if_t ifp, int family) 4486 { 4487 struct domain *dp; 4488 4489 SLIST_FOREACH(dp, &domains, dom_next) { 4490 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4491 return (dp->dom_ifmtu(ifp)); 4492 } 4493 4494 return (ifp->if_mtu); 4495 } 4496 4497 /* 4498 * Methods for drivers to access interface unicast and multicast 4499 * link level addresses. Driver shall not know 'struct ifaddr' neither 4500 * 'struct ifmultiaddr'. 4501 */ 4502 u_int 4503 if_lladdr_count(if_t ifp) 4504 { 4505 struct epoch_tracker et; 4506 struct ifaddr *ifa; 4507 u_int count; 4508 4509 count = 0; 4510 NET_EPOCH_ENTER(et); 4511 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4512 if (ifa->ifa_addr->sa_family == AF_LINK) 4513 count++; 4514 NET_EPOCH_EXIT(et); 4515 4516 return (count); 4517 } 4518 4519 int 4520 if_foreach(if_foreach_cb_t cb, void *cb_arg) 4521 { 4522 if_t ifp; 4523 int error; 4524 4525 NET_EPOCH_ASSERT(); 4526 MPASS(cb); 4527 4528 error = 0; 4529 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4530 error = cb(ifp, cb_arg); 4531 if (error != 0) 4532 break; 4533 } 4534 4535 return (error); 4536 } 4537 4538 /* 4539 * Iterates over the list of interfaces, permitting callback function @cb to sleep. 4540 * Stops iteration if @cb returns non-zero error code. 4541 * Returns the last error code from @cb. 4542 * @match_cb: optional match callback limiting the iteration to only matched interfaces 4543 * @match_arg: argument to pass to @match_cb 4544 * @cb: iteration callback 4545 * @cb_arg: argument to pass to @cb 4546 */ 4547 int 4548 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb, 4549 void *cb_arg) 4550 { 4551 int match_count = 0, array_size = 16; /* 128 bytes for malloc */ 4552 struct ifnet **match_array = NULL; 4553 int error = 0; 4554 4555 MPASS(cb); 4556 4557 while (true) { 4558 struct ifnet **new_array; 4559 int new_size = array_size; 4560 struct epoch_tracker et; 4561 struct ifnet *ifp; 4562 4563 while (new_size < match_count) 4564 new_size *= 2; 4565 new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK); 4566 if (match_array != NULL) 4567 memcpy(new_array, match_array, array_size * sizeof(void *)); 4568 free(match_array, M_TEMP); 4569 match_array = new_array; 4570 array_size = new_size; 4571 4572 match_count = 0; 4573 NET_EPOCH_ENTER(et); 4574 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4575 if (match_cb != NULL && !match_cb(ifp, match_arg)) 4576 continue; 4577 if (match_count < array_size) { 4578 if (if_try_ref(ifp)) 4579 match_array[match_count++] = ifp; 4580 } else 4581 match_count++; 4582 } 4583 NET_EPOCH_EXIT(et); 4584 4585 if (match_count > array_size) { 4586 for (int i = 0; i < array_size; i++) 4587 if_rele(match_array[i]); 4588 continue; 4589 } else { 4590 for (int i = 0; i < match_count; i++) { 4591 if (error == 0) 4592 error = cb(match_array[i], cb_arg); 4593 if_rele(match_array[i]); 4594 } 4595 free(match_array, M_TEMP); 4596 break; 4597 } 4598 } 4599 4600 return (error); 4601 } 4602 4603 4604 /* 4605 * Uses just 1 pointer of the 4 available in the public struct. 4606 */ 4607 if_t 4608 if_iter_start(struct if_iter *iter) 4609 { 4610 if_t ifp; 4611 4612 NET_EPOCH_ASSERT(); 4613 4614 bzero(iter, sizeof(*iter)); 4615 ifp = CK_STAILQ_FIRST(&V_ifnet); 4616 if (ifp != NULL) 4617 iter->context[0] = CK_STAILQ_NEXT(ifp, if_link); 4618 else 4619 iter->context[0] = NULL; 4620 return (ifp); 4621 } 4622 4623 if_t 4624 if_iter_next(struct if_iter *iter) 4625 { 4626 if_t cur_ifp = iter->context[0]; 4627 4628 if (cur_ifp != NULL) 4629 iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link); 4630 return (cur_ifp); 4631 } 4632 4633 void 4634 if_iter_finish(struct if_iter *iter) 4635 { 4636 /* Nothing to do here for now. */ 4637 } 4638 4639 u_int 4640 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4641 { 4642 struct epoch_tracker et; 4643 struct ifaddr *ifa; 4644 u_int count; 4645 4646 MPASS(cb); 4647 4648 count = 0; 4649 NET_EPOCH_ENTER(et); 4650 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4651 if (ifa->ifa_addr->sa_family != AF_LINK) 4652 continue; 4653 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4654 count); 4655 } 4656 NET_EPOCH_EXIT(et); 4657 4658 return (count); 4659 } 4660 4661 u_int 4662 if_llmaddr_count(if_t ifp) 4663 { 4664 struct epoch_tracker et; 4665 struct ifmultiaddr *ifma; 4666 int count; 4667 4668 count = 0; 4669 NET_EPOCH_ENTER(et); 4670 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4671 if (ifma->ifma_addr->sa_family == AF_LINK) 4672 count++; 4673 NET_EPOCH_EXIT(et); 4674 4675 return (count); 4676 } 4677 4678 bool 4679 if_maddr_empty(if_t ifp) 4680 { 4681 4682 return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs)); 4683 } 4684 4685 u_int 4686 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4687 { 4688 struct epoch_tracker et; 4689 struct ifmultiaddr *ifma; 4690 u_int count; 4691 4692 MPASS(cb); 4693 4694 count = 0; 4695 NET_EPOCH_ENTER(et); 4696 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4697 if (ifma->ifma_addr->sa_family != AF_LINK) 4698 continue; 4699 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4700 count); 4701 } 4702 NET_EPOCH_EXIT(et); 4703 4704 return (count); 4705 } 4706 4707 u_int 4708 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4709 { 4710 struct epoch_tracker et; 4711 struct ifaddr *ifa; 4712 u_int count; 4713 4714 MPASS(cb); 4715 4716 count = 0; 4717 NET_EPOCH_ENTER(et); 4718 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4719 if (ifa->ifa_addr->sa_family != type) 4720 continue; 4721 count += (*cb)(cb_arg, ifa, count); 4722 } 4723 NET_EPOCH_EXIT(et); 4724 4725 return (count); 4726 } 4727 4728 struct ifaddr * 4729 ifa_iter_start(if_t ifp, struct ifa_iter *iter) 4730 { 4731 struct ifaddr *ifa; 4732 4733 NET_EPOCH_ASSERT(); 4734 4735 bzero(iter, sizeof(*iter)); 4736 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 4737 if (ifa != NULL) 4738 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4739 else 4740 iter->context[0] = NULL; 4741 return (ifa); 4742 } 4743 4744 struct ifaddr * 4745 ifa_iter_next(struct ifa_iter *iter) 4746 { 4747 struct ifaddr *ifa = iter->context[0]; 4748 4749 if (ifa != NULL) 4750 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4751 return (ifa); 4752 } 4753 4754 void 4755 ifa_iter_finish(struct ifa_iter *iter) 4756 { 4757 /* Nothing to do here for now. */ 4758 } 4759 4760 int 4761 if_setsoftc(if_t ifp, void *softc) 4762 { 4763 ifp->if_softc = softc; 4764 return (0); 4765 } 4766 4767 void * 4768 if_getsoftc(const if_t ifp) 4769 { 4770 return (ifp->if_softc); 4771 } 4772 4773 void 4774 if_setrcvif(struct mbuf *m, if_t ifp) 4775 { 4776 4777 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4778 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4779 } 4780 4781 void 4782 if_setvtag(struct mbuf *m, uint16_t tag) 4783 { 4784 m->m_pkthdr.ether_vtag = tag; 4785 } 4786 4787 uint16_t 4788 if_getvtag(struct mbuf *m) 4789 { 4790 return (m->m_pkthdr.ether_vtag); 4791 } 4792 4793 int 4794 if_sendq_empty(if_t ifp) 4795 { 4796 return (IFQ_DRV_IS_EMPTY(&ifp->if_snd)); 4797 } 4798 4799 struct ifaddr * 4800 if_getifaddr(const if_t ifp) 4801 { 4802 return (ifp->if_addr); 4803 } 4804 4805 int 4806 if_getamcount(const if_t ifp) 4807 { 4808 return (ifp->if_amcount); 4809 } 4810 4811 int 4812 if_setsendqready(if_t ifp) 4813 { 4814 IFQ_SET_READY(&ifp->if_snd); 4815 return (0); 4816 } 4817 4818 int 4819 if_setsendqlen(if_t ifp, int tx_desc_count) 4820 { 4821 IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count); 4822 ifp->if_snd.ifq_drv_maxlen = tx_desc_count; 4823 return (0); 4824 } 4825 4826 void 4827 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na) 4828 { 4829 ifp->if_netmap = na; 4830 } 4831 4832 struct netmap_adapter * 4833 if_getnetmapadapter(if_t ifp) 4834 { 4835 return (ifp->if_netmap); 4836 } 4837 4838 int 4839 if_vlantrunkinuse(if_t ifp) 4840 { 4841 return (ifp->if_vlantrunk != NULL); 4842 } 4843 4844 void 4845 if_init(if_t ifp, void *ctx) 4846 { 4847 (*ifp->if_init)(ctx); 4848 } 4849 4850 void 4851 if_input(if_t ifp, struct mbuf* sendmp) 4852 { 4853 (*ifp->if_input)(ifp, sendmp); 4854 } 4855 4856 int 4857 if_transmit(if_t ifp, struct mbuf *m) 4858 { 4859 return ((*ifp->if_transmit)(ifp, m)); 4860 } 4861 4862 int 4863 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst) 4864 { 4865 if (ifp->if_resolvemulti == NULL) 4866 return (EOPNOTSUPP); 4867 4868 return (ifp->if_resolvemulti(ifp, srcs, dst)); 4869 } 4870 4871 int 4872 if_ioctl(if_t ifp, u_long cmd, void *data) 4873 { 4874 if (ifp->if_ioctl == NULL) 4875 return (EOPNOTSUPP); 4876 4877 return (ifp->if_ioctl(ifp, cmd, data)); 4878 } 4879 4880 struct mbuf * 4881 if_dequeue(if_t ifp) 4882 { 4883 struct mbuf *m; 4884 4885 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 4886 return (m); 4887 } 4888 4889 int 4890 if_sendq_prepend(if_t ifp, struct mbuf *m) 4891 { 4892 IFQ_DRV_PREPEND(&ifp->if_snd, m); 4893 return (0); 4894 } 4895 4896 int 4897 if_setifheaderlen(if_t ifp, int len) 4898 { 4899 ifp->if_hdrlen = len; 4900 return (0); 4901 } 4902 4903 caddr_t 4904 if_getlladdr(const if_t ifp) 4905 { 4906 return (IF_LLADDR(ifp)); 4907 } 4908 4909 void * 4910 if_gethandle(u_char type) 4911 { 4912 return (if_alloc(type)); 4913 } 4914 4915 void 4916 if_vlancap(if_t ifp) 4917 { 4918 VLAN_CAPABILITIES(ifp); 4919 } 4920 4921 int 4922 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4923 { 4924 ifp->if_hw_tsomax = if_hw_tsomax; 4925 return (0); 4926 } 4927 4928 int 4929 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4930 { 4931 ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4932 return (0); 4933 } 4934 4935 int 4936 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4937 { 4938 ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4939 return (0); 4940 } 4941 4942 u_int 4943 if_gethwtsomax(const if_t ifp) 4944 { 4945 return (ifp->if_hw_tsomax); 4946 } 4947 4948 u_int 4949 if_gethwtsomaxsegcount(const if_t ifp) 4950 { 4951 return (ifp->if_hw_tsomaxsegcount); 4952 } 4953 4954 u_int 4955 if_gethwtsomaxsegsize(const if_t ifp) 4956 { 4957 return (ifp->if_hw_tsomaxsegsize); 4958 } 4959 4960 void 4961 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4962 { 4963 ifp->if_init = init_fn; 4964 } 4965 4966 void 4967 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4968 { 4969 ifp->if_input = input_fn; 4970 } 4971 4972 if_input_fn_t 4973 if_getinputfn(if_t ifp) 4974 { 4975 return (ifp->if_input); 4976 } 4977 4978 void 4979 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4980 { 4981 ifp->if_ioctl = ioctl_fn; 4982 } 4983 4984 void 4985 if_setoutputfn(if_t ifp, if_output_fn_t output_fn) 4986 { 4987 ifp->if_output = output_fn; 4988 } 4989 4990 void 4991 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4992 { 4993 ifp->if_start = start_fn; 4994 } 4995 4996 if_start_fn_t 4997 if_getstartfn(if_t ifp) 4998 { 4999 return (ifp->if_start); 5000 } 5001 5002 void 5003 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 5004 { 5005 ifp->if_transmit = start_fn; 5006 } 5007 5008 if_transmit_fn_t 5009 if_gettransmitfn(if_t ifp) 5010 { 5011 return (ifp->if_transmit); 5012 } 5013 5014 void 5015 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 5016 { 5017 ifp->if_qflush = flush_fn; 5018 } 5019 5020 void 5021 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 5022 { 5023 ifp->if_snd_tag_alloc = alloc_fn; 5024 } 5025 5026 int 5027 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params, 5028 struct m_snd_tag **mstp) 5029 { 5030 if (ifp->if_snd_tag_alloc == NULL) 5031 return (EOPNOTSUPP); 5032 return (ifp->if_snd_tag_alloc(ifp, params, mstp)); 5033 } 5034 5035 void 5036 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 5037 { 5038 ifp->if_get_counter = fn; 5039 } 5040 5041 void 5042 if_setreassignfn(if_t ifp, if_reassign_fn_t fn) 5043 { 5044 ifp->if_reassign = fn; 5045 } 5046 5047 void 5048 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn) 5049 { 5050 ifp->if_ratelimit_query = fn; 5051 } 5052 5053 void 5054 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m) 5055 { 5056 ifp->if_debugnet_methods = m; 5057 } 5058 5059 struct label * 5060 if_getmaclabel(if_t ifp) 5061 { 5062 return (ifp->if_label); 5063 } 5064 5065 void 5066 if_setmaclabel(if_t ifp, struct label *label) 5067 { 5068 ifp->if_label = label; 5069 } 5070 5071 int 5072 if_gettype(if_t ifp) 5073 { 5074 return (ifp->if_type); 5075 } 5076 5077 void * 5078 if_getllsoftc(if_t ifp) 5079 { 5080 return (ifp->if_llsoftc); 5081 } 5082 5083 void 5084 if_setllsoftc(if_t ifp, void *llsoftc) 5085 { 5086 ifp->if_llsoftc = llsoftc; 5087 }; 5088 5089 int 5090 if_getlinkstate(if_t ifp) 5091 { 5092 return (ifp->if_link_state); 5093 } 5094 5095 const uint8_t * 5096 if_getbroadcastaddr(if_t ifp) 5097 { 5098 return (ifp->if_broadcastaddr); 5099 } 5100 5101 void 5102 if_setbroadcastaddr(if_t ifp, const uint8_t *addr) 5103 { 5104 ifp->if_broadcastaddr = addr; 5105 } 5106 5107 int 5108 if_getnumadomain(if_t ifp) 5109 { 5110 return (ifp->if_numa_domain); 5111 } 5112 5113 uint64_t 5114 if_getcounter(if_t ifp, ift_counter counter) 5115 { 5116 return (ifp->if_get_counter(ifp, counter)); 5117 } 5118 5119 bool 5120 if_altq_is_enabled(if_t ifp) 5121 { 5122 return (ALTQ_IS_ENABLED(&ifp->if_snd)); 5123 } 5124 5125 struct vnet * 5126 if_getvnet(if_t ifp) 5127 { 5128 return (ifp->if_vnet); 5129 } 5130 5131 void * 5132 if_getafdata(if_t ifp, int af) 5133 { 5134 return (ifp->if_afdata[af]); 5135 } 5136 5137 u_int 5138 if_getfib(if_t ifp) 5139 { 5140 return (ifp->if_fib); 5141 } 5142 5143 uint8_t 5144 if_getaddrlen(if_t ifp) 5145 { 5146 return (ifp->if_addrlen); 5147 } 5148 5149 struct bpf_if * 5150 if_getbpf(if_t ifp) 5151 { 5152 return (ifp->if_bpf); 5153 } 5154 5155 struct ifvlantrunk * 5156 if_getvlantrunk(if_t ifp) 5157 { 5158 return (ifp->if_vlantrunk); 5159 } 5160 5161 uint8_t 5162 if_getpcp(if_t ifp) 5163 { 5164 return (ifp->if_pcp); 5165 } 5166 5167 void * 5168 if_getl2com(if_t ifp) 5169 { 5170 return (ifp->if_l2com); 5171 } 5172 5173 #ifdef DDB 5174 static void 5175 if_show_ifnet(struct ifnet *ifp) 5176 { 5177 if (ifp == NULL) 5178 return; 5179 db_printf("%s:\n", ifp->if_xname); 5180 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 5181 IF_DB_PRINTF("%s", if_dname); 5182 IF_DB_PRINTF("%d", if_dunit); 5183 IF_DB_PRINTF("%s", if_description); 5184 IF_DB_PRINTF("%u", if_index); 5185 IF_DB_PRINTF("%d", if_idxgen); 5186 IF_DB_PRINTF("%u", if_refcount); 5187 IF_DB_PRINTF("%p", if_softc); 5188 IF_DB_PRINTF("%p", if_l2com); 5189 IF_DB_PRINTF("%p", if_llsoftc); 5190 IF_DB_PRINTF("%d", if_amcount); 5191 IF_DB_PRINTF("%p", if_addr); 5192 IF_DB_PRINTF("%p", if_broadcastaddr); 5193 IF_DB_PRINTF("%p", if_afdata); 5194 IF_DB_PRINTF("%d", if_afdata_initialized); 5195 IF_DB_PRINTF("%u", if_fib); 5196 IF_DB_PRINTF("%p", if_vnet); 5197 IF_DB_PRINTF("%p", if_home_vnet); 5198 IF_DB_PRINTF("%p", if_vlantrunk); 5199 IF_DB_PRINTF("%p", if_bpf); 5200 IF_DB_PRINTF("%u", if_pcount); 5201 IF_DB_PRINTF("%p", if_bridge); 5202 IF_DB_PRINTF("%p", if_lagg); 5203 IF_DB_PRINTF("%p", if_pf_kif); 5204 IF_DB_PRINTF("%p", if_carp); 5205 IF_DB_PRINTF("%p", if_label); 5206 IF_DB_PRINTF("%p", if_netmap); 5207 IF_DB_PRINTF("0x%08x", if_flags); 5208 IF_DB_PRINTF("0x%08x", if_drv_flags); 5209 IF_DB_PRINTF("0x%08x", if_capabilities); 5210 IF_DB_PRINTF("0x%08x", if_capenable); 5211 IF_DB_PRINTF("%p", if_snd.ifq_head); 5212 IF_DB_PRINTF("%p", if_snd.ifq_tail); 5213 IF_DB_PRINTF("%d", if_snd.ifq_len); 5214 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 5215 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 5216 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 5217 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 5218 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 5219 IF_DB_PRINTF("%d", if_snd.altq_type); 5220 IF_DB_PRINTF("%x", if_snd.altq_flags); 5221 #undef IF_DB_PRINTF 5222 } 5223 5224 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 5225 { 5226 if (!have_addr) { 5227 db_printf("usage: show ifnet <struct ifnet *>\n"); 5228 return; 5229 } 5230 5231 if_show_ifnet((struct ifnet *)addr); 5232 } 5233 5234 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 5235 { 5236 struct ifnet *ifp; 5237 u_short idx; 5238 5239 for (idx = 1; idx <= if_index; idx++) { 5240 ifp = ifindex_table[idx].ife_ifnet; 5241 if (ifp == NULL) 5242 continue; 5243 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 5244 if (db_pager_quit) 5245 break; 5246 } 5247 } 5248 #endif /* DDB */ 5249