1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org> 5 * Copyright (c) 1980, 1986, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)if.c 8.5 (Berkeley) 1/9/95 33 * $FreeBSD$ 34 */ 35 36 #include "opt_bpf.h" 37 #include "opt_inet6.h" 38 #include "opt_inet.h" 39 #include "opt_ddb.h" 40 41 #include <sys/param.h> 42 #include <sys/capsicum.h> 43 #include <sys/conf.h> 44 #include <sys/eventhandler.h> 45 #include <sys/malloc.h> 46 #include <sys/domainset.h> 47 #include <sys/sbuf.h> 48 #include <sys/bus.h> 49 #include <sys/epoch.h> 50 #include <sys/mbuf.h> 51 #include <sys/systm.h> 52 #include <sys/priv.h> 53 #include <sys/proc.h> 54 #include <sys/socket.h> 55 #include <sys/socketvar.h> 56 #include <sys/protosw.h> 57 #include <sys/kernel.h> 58 #include <sys/lock.h> 59 #include <sys/refcount.h> 60 #include <sys/module.h> 61 #include <sys/nv.h> 62 #include <sys/rwlock.h> 63 #include <sys/sockio.h> 64 #include <sys/syslog.h> 65 #include <sys/sysctl.h> 66 #include <sys/sysent.h> 67 #include <sys/taskqueue.h> 68 #include <sys/domain.h> 69 #include <sys/jail.h> 70 #include <sys/priv.h> 71 72 #ifdef DDB 73 #include <ddb/ddb.h> 74 #endif 75 76 #include <machine/stdarg.h> 77 #include <vm/uma.h> 78 79 #include <net/bpf.h> 80 #include <net/ethernet.h> 81 #include <net/if.h> 82 #include <net/if_arp.h> 83 #include <net/if_clone.h> 84 #include <net/if_dl.h> 85 #include <net/if_types.h> 86 #include <net/if_var.h> 87 #include <net/if_media.h> 88 #include <net/if_mib.h> 89 #include <net/if_vlan_var.h> 90 #include <net/radix.h> 91 #include <net/route.h> 92 #include <net/route/route_ctl.h> 93 #include <net/vnet.h> 94 95 #if defined(INET) || defined(INET6) 96 #include <net/ethernet.h> 97 #include <netinet/in.h> 98 #include <netinet/in_var.h> 99 #include <netinet/ip.h> 100 #include <netinet/ip_carp.h> 101 #ifdef INET 102 #include <net/debugnet.h> 103 #include <netinet/if_ether.h> 104 #endif /* INET */ 105 #ifdef INET6 106 #include <netinet6/in6_var.h> 107 #include <netinet6/in6_ifattach.h> 108 #endif /* INET6 */ 109 #endif /* INET || INET6 */ 110 111 #include <security/mac/mac_framework.h> 112 113 /* 114 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 115 * and ifr_ifru when it is used in SIOCGIFCONF. 116 */ 117 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 118 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 119 120 __read_mostly epoch_t net_epoch_preempt; 121 #ifdef COMPAT_FREEBSD32 122 #include <sys/mount.h> 123 #include <compat/freebsd32/freebsd32.h> 124 125 struct ifreq_buffer32 { 126 uint32_t length; /* (size_t) */ 127 uint32_t buffer; /* (void *) */ 128 }; 129 130 /* 131 * Interface request structure used for socket 132 * ioctl's. All interface ioctl's must have parameter 133 * definitions which begin with ifr_name. The 134 * remainder may be interface specific. 135 */ 136 struct ifreq32 { 137 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 138 union { 139 struct sockaddr ifru_addr; 140 struct sockaddr ifru_dstaddr; 141 struct sockaddr ifru_broadaddr; 142 struct ifreq_buffer32 ifru_buffer; 143 short ifru_flags[2]; 144 short ifru_index; 145 int ifru_jid; 146 int ifru_metric; 147 int ifru_mtu; 148 int ifru_phys; 149 int ifru_media; 150 uint32_t ifru_data; 151 int ifru_cap[2]; 152 u_int ifru_fib; 153 u_char ifru_vlan_pcp; 154 } ifr_ifru; 155 }; 156 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 157 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 158 __offsetof(struct ifreq32, ifr_ifru)); 159 160 struct ifconf32 { 161 int32_t ifc_len; 162 union { 163 uint32_t ifcu_buf; 164 uint32_t ifcu_req; 165 } ifc_ifcu; 166 }; 167 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 168 169 struct ifdrv32 { 170 char ifd_name[IFNAMSIZ]; 171 uint32_t ifd_cmd; 172 uint32_t ifd_len; 173 uint32_t ifd_data; 174 }; 175 #define SIOCSDRVSPEC32 _IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32) 176 #define SIOCGDRVSPEC32 _IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32) 177 178 struct ifgroupreq32 { 179 char ifgr_name[IFNAMSIZ]; 180 u_int ifgr_len; 181 union { 182 char ifgru_group[IFNAMSIZ]; 183 uint32_t ifgru_groups; 184 } ifgr_ifgru; 185 }; 186 #define SIOCAIFGROUP32 _IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32) 187 #define SIOCGIFGROUP32 _IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32) 188 #define SIOCDIFGROUP32 _IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32) 189 #define SIOCGIFGMEMB32 _IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32) 190 191 struct ifmediareq32 { 192 char ifm_name[IFNAMSIZ]; 193 int ifm_current; 194 int ifm_mask; 195 int ifm_status; 196 int ifm_active; 197 int ifm_count; 198 uint32_t ifm_ulist; /* (int *) */ 199 }; 200 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 201 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 202 #endif /* COMPAT_FREEBSD32 */ 203 204 union ifreq_union { 205 struct ifreq ifr; 206 #ifdef COMPAT_FREEBSD32 207 struct ifreq32 ifr32; 208 #endif 209 }; 210 211 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 212 "Link layers"); 213 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 214 "Generic link-management"); 215 216 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 217 &ifqmaxlen, 0, "max send queue size"); 218 219 /* Log link state change events */ 220 static int log_link_state_change = 1; 221 222 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 223 &log_link_state_change, 0, 224 "log interface link state change events"); 225 226 /* Log promiscuous mode change events */ 227 static int log_promisc_mode_change = 1; 228 229 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 230 &log_promisc_mode_change, 1, 231 "log promiscuous mode change events"); 232 233 /* Interface description */ 234 static unsigned int ifdescr_maxlen = 1024; 235 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 236 &ifdescr_maxlen, 0, 237 "administrative maximum length for interface description"); 238 239 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 240 241 /* global sx for non-critical path ifdescr */ 242 static struct sx ifdescr_sx; 243 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 244 245 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 246 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 247 /* These are external hooks for CARP. */ 248 void (*carp_linkstate_p)(struct ifnet *ifp); 249 void (*carp_demote_adj_p)(int, char *); 250 int (*carp_master_p)(struct ifaddr *); 251 #if defined(INET) || defined(INET6) 252 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 253 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 254 const struct sockaddr *sa); 255 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 256 int (*carp_attach_p)(struct ifaddr *, int); 257 void (*carp_detach_p)(struct ifaddr *, bool); 258 #endif 259 #ifdef INET 260 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 261 #endif 262 #ifdef INET6 263 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 264 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 265 const struct in6_addr *taddr); 266 #endif 267 268 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 269 270 /* 271 * XXX: Style; these should be sorted alphabetically, and unprototyped 272 * static functions should be prototyped. Currently they are sorted by 273 * declaration order. 274 */ 275 static void if_attachdomain(void *); 276 static void if_attachdomain1(struct ifnet *); 277 static int ifconf(u_long, caddr_t); 278 static void if_input_default(struct ifnet *, struct mbuf *); 279 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 280 static void if_route(struct ifnet *, int flag, int fam); 281 static int if_setflag(struct ifnet *, int, int, int *, int); 282 static int if_transmit_default(struct ifnet *ifp, struct mbuf *m); 283 static void if_unroute(struct ifnet *, int flag, int fam); 284 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 285 static void do_link_state_change(void *, int); 286 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 287 static int if_getgroupmembers(struct ifgroupreq *); 288 static void if_delgroups(struct ifnet *); 289 static void if_attach_internal(struct ifnet *, bool); 290 static int if_detach_internal(struct ifnet *, bool); 291 static void if_siocaddmulti(void *, int); 292 static void if_link_ifnet(struct ifnet *); 293 static bool if_unlink_ifnet(struct ifnet *, bool); 294 #ifdef VIMAGE 295 static int if_vmove(struct ifnet *, struct vnet *); 296 #endif 297 298 #ifdef INET6 299 /* 300 * XXX: declare here to avoid to include many inet6 related files.. 301 * should be more generalized? 302 */ 303 extern void nd6_setmtu(struct ifnet *); 304 #endif 305 306 /* ipsec helper hooks */ 307 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 308 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 309 310 int ifqmaxlen = IFQ_MAXLEN; 311 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 312 VNET_DEFINE(struct ifgrouphead, ifg_head); 313 314 /* Table of ifnet by index. */ 315 static int if_index; 316 static int if_indexlim = 8; 317 static struct ifindex_entry { 318 struct ifnet *ife_ifnet; 319 uint16_t ife_gencnt; 320 } *ifindex_table; 321 322 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system, 323 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 324 "Variables global to all interfaces"); 325 static int 326 sysctl_ifcount(SYSCTL_HANDLER_ARGS) 327 { 328 int rv = 0; 329 330 IFNET_RLOCK(); 331 for (int i = 1; i <= if_index; i++) 332 if (ifindex_table[i].ife_ifnet != NULL && 333 ifindex_table[i].ife_ifnet->if_vnet == curvnet) 334 rv = i; 335 IFNET_RUNLOCK(); 336 337 return (sysctl_handle_int(oidp, &rv, 0, req)); 338 } 339 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount, 340 CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I", 341 "Maximum known interface index"); 342 343 /* 344 * The global network interface list (V_ifnet) and related state (such as 345 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 346 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 347 */ 348 struct sx ifnet_sxlock; 349 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 350 351 struct sx ifnet_detach_sxlock; 352 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 353 SX_RECURSE); 354 355 #ifdef VIMAGE 356 #define VNET_IS_SHUTTING_DOWN(_vnet) \ 357 ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE) 358 #endif 359 360 static if_com_alloc_t *if_com_alloc[256]; 361 static if_com_free_t *if_com_free[256]; 362 363 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 364 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 365 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 366 367 struct ifnet * 368 ifnet_byindex(u_int idx) 369 { 370 struct ifnet *ifp; 371 372 NET_EPOCH_ASSERT(); 373 374 if (__predict_false(idx > if_index)) 375 return (NULL); 376 377 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 378 379 if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet) 380 ifp = NULL; 381 382 return (ifp); 383 } 384 385 struct ifnet * 386 ifnet_byindex_ref(u_int idx) 387 { 388 struct ifnet *ifp; 389 390 ifp = ifnet_byindex(idx); 391 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 392 return (NULL); 393 if (!if_try_ref(ifp)) 394 return (NULL); 395 return (ifp); 396 } 397 398 struct ifnet * 399 ifnet_byindexgen(uint16_t idx, uint16_t gen) 400 { 401 struct ifnet *ifp; 402 403 NET_EPOCH_ASSERT(); 404 405 if (__predict_false(idx > if_index)) 406 return (NULL); 407 408 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 409 410 if (ifindex_table[idx].ife_gencnt == gen) 411 return (ifp); 412 else 413 return (NULL); 414 } 415 416 /* 417 * Network interface utility routines. 418 * 419 * Routines with ifa_ifwith* names take sockaddr *'s as 420 * parameters. 421 */ 422 423 static void 424 if_init_idxtable(void *arg __unused) 425 { 426 427 ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table), 428 M_IFNET, M_WAITOK | M_ZERO); 429 } 430 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL); 431 432 static void 433 vnet_if_init(const void *unused __unused) 434 { 435 436 CK_STAILQ_INIT(&V_ifnet); 437 CK_STAILQ_INIT(&V_ifg_head); 438 vnet_if_clone_init(); 439 } 440 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 441 NULL); 442 443 static void 444 if_link_ifnet(struct ifnet *ifp) 445 { 446 447 IFNET_WLOCK(); 448 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 449 #ifdef VIMAGE 450 curvnet->vnet_ifcnt++; 451 #endif 452 IFNET_WUNLOCK(); 453 } 454 455 static bool 456 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 457 { 458 struct ifnet *iter; 459 int found = 0; 460 461 IFNET_WLOCK(); 462 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 463 if (iter == ifp) { 464 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 465 if (!vmove) 466 ifp->if_flags |= IFF_DYING; 467 found = 1; 468 break; 469 } 470 #ifdef VIMAGE 471 curvnet->vnet_ifcnt--; 472 #endif 473 IFNET_WUNLOCK(); 474 475 return (found); 476 } 477 478 #ifdef VIMAGE 479 static void 480 vnet_if_return(const void *unused __unused) 481 { 482 struct ifnet *ifp, *nifp; 483 struct ifnet **pending; 484 int found __diagused; 485 int i; 486 487 i = 0; 488 489 /* 490 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 491 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 492 * if_detach_internal(), which waits for NET_EPOCH callbacks to 493 * complete. We can't do that from within NET_EPOCH. 494 * 495 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 496 * read/write lock. We cannot hold the lock as we call if_vmove() 497 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 498 * ctx lock. 499 */ 500 IFNET_WLOCK(); 501 502 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 503 M_IFNET, M_WAITOK | M_ZERO); 504 505 /* Return all inherited interfaces to their parent vnets. */ 506 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 507 if (ifp->if_home_vnet != ifp->if_vnet) { 508 found = if_unlink_ifnet(ifp, true); 509 MPASS(found); 510 511 pending[i++] = ifp; 512 } 513 } 514 IFNET_WUNLOCK(); 515 516 for (int j = 0; j < i; j++) { 517 sx_xlock(&ifnet_detach_sxlock); 518 if_vmove(pending[j], pending[j]->if_home_vnet); 519 sx_xunlock(&ifnet_detach_sxlock); 520 } 521 522 free(pending, M_IFNET); 523 } 524 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 525 vnet_if_return, NULL); 526 #endif 527 528 /* 529 * Allocate a struct ifnet and an index for an interface. A layer 2 530 * common structure will also be allocated if an allocation routine is 531 * registered for the passed type. 532 */ 533 static struct ifnet * 534 if_alloc_domain(u_char type, int numa_domain) 535 { 536 struct ifnet *ifp; 537 u_short idx; 538 539 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 540 if (numa_domain == IF_NODOM) 541 ifp = malloc(sizeof(struct ifnet), M_IFNET, 542 M_WAITOK | M_ZERO); 543 else 544 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 545 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 546 ifp->if_type = type; 547 ifp->if_alloctype = type; 548 ifp->if_numa_domain = numa_domain; 549 #ifdef VIMAGE 550 ifp->if_vnet = curvnet; 551 #endif 552 if (if_com_alloc[type] != NULL) { 553 ifp->if_l2com = if_com_alloc[type](type, ifp); 554 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 555 type)); 556 } 557 558 IF_ADDR_LOCK_INIT(ifp); 559 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 560 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 561 ifp->if_afdata_initialized = 0; 562 IF_AFDATA_LOCK_INIT(ifp); 563 CK_STAILQ_INIT(&ifp->if_addrhead); 564 CK_STAILQ_INIT(&ifp->if_multiaddrs); 565 CK_STAILQ_INIT(&ifp->if_groups); 566 #ifdef MAC 567 mac_ifnet_init(ifp); 568 #endif 569 ifq_init(&ifp->if_snd, ifp); 570 571 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 572 for (int i = 0; i < IFCOUNTERS; i++) 573 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 574 ifp->if_get_counter = if_get_counter_default; 575 ifp->if_pcp = IFNET_PCP_NONE; 576 577 /* Allocate an ifindex array entry. */ 578 IFNET_WLOCK(); 579 /* 580 * Try to find an empty slot below if_index. If we fail, take the 581 * next slot. 582 */ 583 for (idx = 1; idx <= if_index; idx++) { 584 if (ifindex_table[idx].ife_ifnet == NULL) 585 break; 586 } 587 588 /* Catch if_index overflow. */ 589 if (idx >= if_indexlim) { 590 struct ifindex_entry *new, *old; 591 int newlim; 592 593 newlim = if_indexlim * 2; 594 new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO); 595 memcpy(new, ifindex_table, if_indexlim * sizeof(*new)); 596 old = ifindex_table; 597 ck_pr_store_ptr(&ifindex_table, new); 598 if_indexlim = newlim; 599 epoch_wait_preempt(net_epoch_preempt); 600 free(old, M_IFNET); 601 } 602 if (idx > if_index) 603 if_index = idx; 604 605 ifp->if_index = idx; 606 ifp->if_idxgen = ifindex_table[idx].ife_gencnt; 607 ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp); 608 IFNET_WUNLOCK(); 609 610 return (ifp); 611 } 612 613 struct ifnet * 614 if_alloc_dev(u_char type, device_t dev) 615 { 616 int numa_domain; 617 618 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 619 return (if_alloc_domain(type, IF_NODOM)); 620 return (if_alloc_domain(type, numa_domain)); 621 } 622 623 struct ifnet * 624 if_alloc(u_char type) 625 { 626 627 return (if_alloc_domain(type, IF_NODOM)); 628 } 629 /* 630 * Do the actual work of freeing a struct ifnet, and layer 2 common 631 * structure. This call is made when the network epoch guarantees 632 * us that nobody holds a pointer to the interface. 633 */ 634 static void 635 if_free_deferred(epoch_context_t ctx) 636 { 637 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 638 639 KASSERT((ifp->if_flags & IFF_DYING), 640 ("%s: interface not dying", __func__)); 641 642 if (if_com_free[ifp->if_alloctype] != NULL) 643 if_com_free[ifp->if_alloctype](ifp->if_l2com, 644 ifp->if_alloctype); 645 646 #ifdef MAC 647 mac_ifnet_destroy(ifp); 648 #endif /* MAC */ 649 IF_AFDATA_DESTROY(ifp); 650 IF_ADDR_LOCK_DESTROY(ifp); 651 ifq_delete(&ifp->if_snd); 652 653 for (int i = 0; i < IFCOUNTERS; i++) 654 counter_u64_free(ifp->if_counters[i]); 655 656 if_freedescr(ifp->if_description); 657 free(ifp->if_hw_addr, M_IFADDR); 658 free(ifp, M_IFNET); 659 } 660 661 /* 662 * Deregister an interface and free the associated storage. 663 */ 664 void 665 if_free(struct ifnet *ifp) 666 { 667 668 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 669 670 /* 671 * XXXGL: An interface index is really an alias to ifp pointer. 672 * Why would we clear the alias now, and not in the deferred 673 * context? Indeed there is nothing wrong with some network 674 * thread obtaining ifp via ifnet_byindex() inside the network 675 * epoch and then dereferencing ifp while we perform if_free(), 676 * and after if_free() finished, too. 677 * 678 * This early index freeing was important back when ifindex was 679 * virtualized and interface would outlive the vnet. 680 */ 681 IFNET_WLOCK(); 682 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 683 ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL); 684 ifindex_table[ifp->if_index].ife_gencnt++; 685 while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL) 686 if_index--; 687 IFNET_WUNLOCK(); 688 689 if (refcount_release(&ifp->if_refcount)) 690 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 691 } 692 693 /* 694 * Interfaces to keep an ifnet type-stable despite the possibility of the 695 * driver calling if_free(). If there are additional references, we defer 696 * freeing the underlying data structure. 697 */ 698 void 699 if_ref(struct ifnet *ifp) 700 { 701 u_int old __diagused; 702 703 /* We don't assert the ifnet list lock here, but arguably should. */ 704 old = refcount_acquire(&ifp->if_refcount); 705 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 706 } 707 708 bool 709 if_try_ref(struct ifnet *ifp) 710 { 711 NET_EPOCH_ASSERT(); 712 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 713 } 714 715 void 716 if_rele(struct ifnet *ifp) 717 { 718 719 if (!refcount_release(&ifp->if_refcount)) 720 return; 721 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 722 } 723 724 void 725 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 726 { 727 728 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 729 730 if (ifq->ifq_maxlen == 0) 731 ifq->ifq_maxlen = ifqmaxlen; 732 733 ifq->altq_type = 0; 734 ifq->altq_disc = NULL; 735 ifq->altq_flags &= ALTQF_CANTCHANGE; 736 ifq->altq_tbr = NULL; 737 ifq->altq_ifp = ifp; 738 } 739 740 void 741 ifq_delete(struct ifaltq *ifq) 742 { 743 mtx_destroy(&ifq->ifq_mtx); 744 } 745 746 /* 747 * Perform generic interface initialization tasks and attach the interface 748 * to the list of "active" interfaces. If vmove flag is set on entry 749 * to if_attach_internal(), perform only a limited subset of initialization 750 * tasks, given that we are moving from one vnet to another an ifnet which 751 * has already been fully initialized. 752 * 753 * Note that if_detach_internal() removes group membership unconditionally 754 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 755 * Thus, when if_vmove() is applied to a cloned interface, group membership 756 * is lost while a cloned one always joins a group whose name is 757 * ifc->ifc_name. To recover this after if_detach_internal() and 758 * if_attach_internal(), the cloner should be specified to 759 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 760 * attempts to join a group whose name is ifc->ifc_name. 761 * 762 * XXX: 763 * - The decision to return void and thus require this function to 764 * succeed is questionable. 765 * - We should probably do more sanity checking. For instance we don't 766 * do anything to insure if_xname is unique or non-empty. 767 */ 768 void 769 if_attach(struct ifnet *ifp) 770 { 771 772 if_attach_internal(ifp, false); 773 } 774 775 /* 776 * Compute the least common TSO limit. 777 */ 778 void 779 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 780 { 781 /* 782 * 1) If there is no limit currently, take the limit from 783 * the network adapter. 784 * 785 * 2) If the network adapter has a limit below the current 786 * limit, apply it. 787 */ 788 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 789 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 790 pmax->tsomaxbytes = ifp->if_hw_tsomax; 791 } 792 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 793 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 794 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 795 } 796 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 797 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 798 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 799 } 800 } 801 802 /* 803 * Update TSO limit of a network adapter. 804 * 805 * Returns zero if no change. Else non-zero. 806 */ 807 int 808 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 809 { 810 int retval = 0; 811 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 812 ifp->if_hw_tsomax = pmax->tsomaxbytes; 813 retval++; 814 } 815 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 816 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 817 retval++; 818 } 819 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 820 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 821 retval++; 822 } 823 return (retval); 824 } 825 826 static void 827 if_attach_internal(struct ifnet *ifp, bool vmove) 828 { 829 unsigned socksize, ifasize; 830 int namelen, masklen; 831 struct sockaddr_dl *sdl; 832 struct ifaddr *ifa; 833 834 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 835 836 #ifdef VIMAGE 837 ifp->if_vnet = curvnet; 838 if (ifp->if_home_vnet == NULL) 839 ifp->if_home_vnet = curvnet; 840 #endif 841 842 if_addgroup(ifp, IFG_ALL); 843 844 #ifdef VIMAGE 845 /* Restore group membership for cloned interface. */ 846 if (vmove) 847 if_clone_restoregroup(ifp); 848 #endif 849 850 getmicrotime(&ifp->if_lastchange); 851 ifp->if_epoch = time_uptime; 852 853 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 854 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 855 ("transmit and qflush must both either be set or both be NULL")); 856 if (ifp->if_transmit == NULL) { 857 ifp->if_transmit = if_transmit_default; 858 ifp->if_qflush = if_qflush; 859 } 860 if (ifp->if_input == NULL) 861 ifp->if_input = if_input_default; 862 863 if (ifp->if_requestencap == NULL) 864 ifp->if_requestencap = if_requestencap_default; 865 866 if (!vmove) { 867 #ifdef MAC 868 mac_ifnet_create(ifp); 869 #endif 870 871 /* 872 * Create a Link Level name for this device. 873 */ 874 namelen = strlen(ifp->if_xname); 875 /* 876 * Always save enough space for any possiable name so we 877 * can do a rename in place later. 878 */ 879 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 880 socksize = masklen + ifp->if_addrlen; 881 if (socksize < sizeof(*sdl)) 882 socksize = sizeof(*sdl); 883 socksize = roundup2(socksize, sizeof(long)); 884 ifasize = sizeof(*ifa) + 2 * socksize; 885 ifa = ifa_alloc(ifasize, M_WAITOK); 886 sdl = (struct sockaddr_dl *)(ifa + 1); 887 sdl->sdl_len = socksize; 888 sdl->sdl_family = AF_LINK; 889 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 890 sdl->sdl_nlen = namelen; 891 sdl->sdl_index = ifp->if_index; 892 sdl->sdl_type = ifp->if_type; 893 ifp->if_addr = ifa; 894 ifa->ifa_ifp = ifp; 895 ifa->ifa_addr = (struct sockaddr *)sdl; 896 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 897 ifa->ifa_netmask = (struct sockaddr *)sdl; 898 sdl->sdl_len = masklen; 899 while (namelen != 0) 900 sdl->sdl_data[--namelen] = 0xff; 901 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 902 /* Reliably crash if used uninitialized. */ 903 ifp->if_broadcastaddr = NULL; 904 905 if (ifp->if_type == IFT_ETHER) { 906 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 907 M_WAITOK | M_ZERO); 908 } 909 910 #if defined(INET) || defined(INET6) 911 /* Use defaults for TSO, if nothing is set */ 912 if (ifp->if_hw_tsomax == 0 && 913 ifp->if_hw_tsomaxsegcount == 0 && 914 ifp->if_hw_tsomaxsegsize == 0) { 915 /* 916 * The TSO defaults needs to be such that an 917 * NFS mbuf list of 35 mbufs totalling just 918 * below 64K works and that a chain of mbufs 919 * can be defragged into at most 32 segments: 920 */ 921 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 922 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 923 ifp->if_hw_tsomaxsegcount = 35; 924 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 925 926 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 927 if (ifp->if_capabilities & IFCAP_TSO) { 928 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 929 ifp->if_hw_tsomax, 930 ifp->if_hw_tsomaxsegcount, 931 ifp->if_hw_tsomaxsegsize); 932 } 933 } 934 #endif 935 } 936 #ifdef VIMAGE 937 else { 938 /* 939 * Update the interface index in the link layer address 940 * of the interface. 941 */ 942 for (ifa = ifp->if_addr; ifa != NULL; 943 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 944 if (ifa->ifa_addr->sa_family == AF_LINK) { 945 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 946 sdl->sdl_index = ifp->if_index; 947 } 948 } 949 } 950 #endif 951 952 if_link_ifnet(ifp); 953 954 if (domain_init_status >= 2) 955 if_attachdomain1(ifp); 956 957 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 958 if (IS_DEFAULT_VNET(curvnet)) 959 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 960 } 961 962 static void 963 if_epochalloc(void *dummy __unused) 964 { 965 966 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 967 } 968 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 969 970 static void 971 if_attachdomain(void *dummy) 972 { 973 struct ifnet *ifp; 974 975 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 976 if_attachdomain1(ifp); 977 } 978 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 979 if_attachdomain, NULL); 980 981 static void 982 if_attachdomain1(struct ifnet *ifp) 983 { 984 struct domain *dp; 985 986 /* 987 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 988 * cannot lock ifp->if_afdata initialization, entirely. 989 */ 990 IF_AFDATA_LOCK(ifp); 991 if (ifp->if_afdata_initialized >= domain_init_status) { 992 IF_AFDATA_UNLOCK(ifp); 993 log(LOG_WARNING, "%s called more than once on %s\n", 994 __func__, ifp->if_xname); 995 return; 996 } 997 ifp->if_afdata_initialized = domain_init_status; 998 IF_AFDATA_UNLOCK(ifp); 999 1000 /* address family dependent data region */ 1001 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 1002 SLIST_FOREACH(dp, &domains, dom_next) { 1003 if (dp->dom_ifattach) 1004 ifp->if_afdata[dp->dom_family] = 1005 (*dp->dom_ifattach)(ifp); 1006 } 1007 } 1008 1009 /* 1010 * Remove any unicast or broadcast network addresses from an interface. 1011 */ 1012 void 1013 if_purgeaddrs(struct ifnet *ifp) 1014 { 1015 struct ifaddr *ifa; 1016 1017 #ifdef INET6 1018 /* 1019 * Need to leave multicast addresses of proxy NDP llentries 1020 * before in6_purgeifaddr() because the llentries are keys 1021 * for in6_multi objects of proxy NDP entries. 1022 * in6_purgeifaddr()s clean up llentries including proxy NDPs 1023 * then we would lose the keys if they are called earlier. 1024 */ 1025 in6_purge_proxy_ndp(ifp); 1026 #endif 1027 while (1) { 1028 struct epoch_tracker et; 1029 1030 NET_EPOCH_ENTER(et); 1031 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1032 if (ifa->ifa_addr->sa_family != AF_LINK) 1033 break; 1034 } 1035 NET_EPOCH_EXIT(et); 1036 1037 if (ifa == NULL) 1038 break; 1039 #ifdef INET 1040 /* XXX: Ugly!! ad hoc just for INET */ 1041 if (ifa->ifa_addr->sa_family == AF_INET) { 1042 struct ifaliasreq ifr; 1043 1044 bzero(&ifr, sizeof(ifr)); 1045 ifr.ifra_addr = *ifa->ifa_addr; 1046 if (ifa->ifa_dstaddr) 1047 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1048 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1049 NULL) == 0) 1050 continue; 1051 } 1052 #endif /* INET */ 1053 #ifdef INET6 1054 if (ifa->ifa_addr->sa_family == AF_INET6) { 1055 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1056 /* ifp_addrhead is already updated */ 1057 continue; 1058 } 1059 #endif /* INET6 */ 1060 IF_ADDR_WLOCK(ifp); 1061 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1062 IF_ADDR_WUNLOCK(ifp); 1063 ifa_free(ifa); 1064 } 1065 } 1066 1067 /* 1068 * Remove any multicast network addresses from an interface when an ifnet 1069 * is going away. 1070 */ 1071 static void 1072 if_purgemaddrs(struct ifnet *ifp) 1073 { 1074 struct ifmultiaddr *ifma; 1075 1076 IF_ADDR_WLOCK(ifp); 1077 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1078 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1079 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1080 if_delmulti_locked(ifp, ifma, 1); 1081 } 1082 IF_ADDR_WUNLOCK(ifp); 1083 } 1084 1085 /* 1086 * Detach an interface, removing it from the list of "active" interfaces. 1087 * If vmove flag is set on entry to if_detach_internal(), perform only a 1088 * limited subset of cleanup tasks, given that we are moving an ifnet from 1089 * one vnet to another, where it must be fully operational. 1090 * 1091 * XXXRW: There are some significant questions about event ordering, and 1092 * how to prevent things from starting to use the interface during detach. 1093 */ 1094 void 1095 if_detach(struct ifnet *ifp) 1096 { 1097 bool found; 1098 1099 CURVNET_SET_QUIET(ifp->if_vnet); 1100 found = if_unlink_ifnet(ifp, false); 1101 if (found) { 1102 sx_xlock(&ifnet_detach_sxlock); 1103 if_detach_internal(ifp, false); 1104 sx_xunlock(&ifnet_detach_sxlock); 1105 } 1106 CURVNET_RESTORE(); 1107 } 1108 1109 /* 1110 * The vmove flag, if set, indicates that we are called from a callpath 1111 * that is moving an interface to a different vnet instance. 1112 * 1113 * The shutdown flag, if set, indicates that we are called in the 1114 * process of shutting down a vnet instance. Currently only the 1115 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1116 * on a vnet instance shutdown without this flag being set, e.g., when 1117 * the cloned interfaces are destoyed as first thing of teardown. 1118 */ 1119 static int 1120 if_detach_internal(struct ifnet *ifp, bool vmove) 1121 { 1122 struct ifaddr *ifa; 1123 int i; 1124 struct domain *dp; 1125 #ifdef VIMAGE 1126 bool shutdown; 1127 1128 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1129 #endif 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 (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 /* 2144 * Mark an interface up and notify protocols of 2145 * the transition. 2146 */ 2147 static void 2148 if_route(struct ifnet *ifp, int flag, int fam) 2149 { 2150 2151 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2152 2153 ifp->if_flags |= flag; 2154 getmicrotime(&ifp->if_lastchange); 2155 if (ifp->if_carp) 2156 (*carp_linkstate_p)(ifp); 2157 rt_ifmsg(ifp, IFF_UP); 2158 #ifdef INET6 2159 in6_if_up(ifp); 2160 #endif 2161 } 2162 2163 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2164 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2165 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2166 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2167 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2168 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2169 int (*vlan_setcookie_p)(struct ifnet *, void *); 2170 void *(*vlan_cookie_p)(struct ifnet *); 2171 2172 /* 2173 * Handle a change in the interface link state. To avoid LORs 2174 * between driver lock and upper layer locks, as well as possible 2175 * recursions, we post event to taskqueue, and all job 2176 * is done in static do_link_state_change(). 2177 */ 2178 void 2179 if_link_state_change(struct ifnet *ifp, int link_state) 2180 { 2181 /* Return if state hasn't changed. */ 2182 if (ifp->if_link_state == link_state) 2183 return; 2184 2185 ifp->if_link_state = link_state; 2186 2187 /* XXXGL: reference ifp? */ 2188 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2189 } 2190 2191 static void 2192 do_link_state_change(void *arg, int pending) 2193 { 2194 struct ifnet *ifp; 2195 int link_state; 2196 2197 ifp = arg; 2198 link_state = ifp->if_link_state; 2199 2200 CURVNET_SET(ifp->if_vnet); 2201 rt_ifmsg(ifp, 0); 2202 if (ifp->if_vlantrunk != NULL) 2203 (*vlan_link_state_p)(ifp); 2204 2205 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2206 ifp->if_l2com != NULL) 2207 (*ng_ether_link_state_p)(ifp, link_state); 2208 if (ifp->if_carp) 2209 (*carp_linkstate_p)(ifp); 2210 if (ifp->if_bridge) 2211 ifp->if_bridge_linkstate(ifp); 2212 if (ifp->if_lagg) 2213 (*lagg_linkstate_p)(ifp, link_state); 2214 2215 if (IS_DEFAULT_VNET(curvnet)) 2216 devctl_notify("IFNET", ifp->if_xname, 2217 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2218 NULL); 2219 if (pending > 1) 2220 if_printf(ifp, "%d link states coalesced\n", pending); 2221 if (log_link_state_change) 2222 if_printf(ifp, "link state changed to %s\n", 2223 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2224 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2225 CURVNET_RESTORE(); 2226 } 2227 2228 /* 2229 * Mark an interface down and notify protocols of 2230 * the transition. 2231 */ 2232 void 2233 if_down(struct ifnet *ifp) 2234 { 2235 2236 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2237 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2238 } 2239 2240 /* 2241 * Mark an interface up and notify protocols of 2242 * the transition. 2243 */ 2244 void 2245 if_up(struct ifnet *ifp) 2246 { 2247 2248 if_route(ifp, IFF_UP, AF_UNSPEC); 2249 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2250 } 2251 2252 /* 2253 * Flush an interface queue. 2254 */ 2255 void 2256 if_qflush(struct ifnet *ifp) 2257 { 2258 struct mbuf *m, *n; 2259 struct ifaltq *ifq; 2260 2261 ifq = &ifp->if_snd; 2262 IFQ_LOCK(ifq); 2263 #ifdef ALTQ 2264 if (ALTQ_IS_ENABLED(ifq)) 2265 ALTQ_PURGE(ifq); 2266 #endif 2267 n = ifq->ifq_head; 2268 while ((m = n) != NULL) { 2269 n = m->m_nextpkt; 2270 m_freem(m); 2271 } 2272 ifq->ifq_head = 0; 2273 ifq->ifq_tail = 0; 2274 ifq->ifq_len = 0; 2275 IFQ_UNLOCK(ifq); 2276 } 2277 2278 /* 2279 * Map interface name to interface structure pointer, with or without 2280 * returning a reference. 2281 */ 2282 struct ifnet * 2283 ifunit_ref(const char *name) 2284 { 2285 struct epoch_tracker et; 2286 struct ifnet *ifp; 2287 2288 NET_EPOCH_ENTER(et); 2289 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2290 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2291 !(ifp->if_flags & IFF_DYING)) 2292 break; 2293 } 2294 if (ifp != NULL) { 2295 if_ref(ifp); 2296 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2297 } 2298 2299 NET_EPOCH_EXIT(et); 2300 return (ifp); 2301 } 2302 2303 struct ifnet * 2304 ifunit(const char *name) 2305 { 2306 struct epoch_tracker et; 2307 struct ifnet *ifp; 2308 2309 NET_EPOCH_ENTER(et); 2310 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2311 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2312 break; 2313 } 2314 NET_EPOCH_EXIT(et); 2315 return (ifp); 2316 } 2317 2318 void * 2319 ifr_buffer_get_buffer(void *data) 2320 { 2321 union ifreq_union *ifrup; 2322 2323 ifrup = data; 2324 #ifdef COMPAT_FREEBSD32 2325 if (SV_CURPROC_FLAG(SV_ILP32)) 2326 return ((void *)(uintptr_t) 2327 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2328 #endif 2329 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2330 } 2331 2332 static void 2333 ifr_buffer_set_buffer_null(void *data) 2334 { 2335 union ifreq_union *ifrup; 2336 2337 ifrup = data; 2338 #ifdef COMPAT_FREEBSD32 2339 if (SV_CURPROC_FLAG(SV_ILP32)) 2340 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2341 else 2342 #endif 2343 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2344 } 2345 2346 size_t 2347 ifr_buffer_get_length(void *data) 2348 { 2349 union ifreq_union *ifrup; 2350 2351 ifrup = data; 2352 #ifdef COMPAT_FREEBSD32 2353 if (SV_CURPROC_FLAG(SV_ILP32)) 2354 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2355 #endif 2356 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2357 } 2358 2359 static void 2360 ifr_buffer_set_length(void *data, size_t len) 2361 { 2362 union ifreq_union *ifrup; 2363 2364 ifrup = data; 2365 #ifdef COMPAT_FREEBSD32 2366 if (SV_CURPROC_FLAG(SV_ILP32)) 2367 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2368 else 2369 #endif 2370 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2371 } 2372 2373 void * 2374 ifr_data_get_ptr(void *ifrp) 2375 { 2376 union ifreq_union *ifrup; 2377 2378 ifrup = ifrp; 2379 #ifdef COMPAT_FREEBSD32 2380 if (SV_CURPROC_FLAG(SV_ILP32)) 2381 return ((void *)(uintptr_t) 2382 ifrup->ifr32.ifr_ifru.ifru_data); 2383 #endif 2384 return (ifrup->ifr.ifr_ifru.ifru_data); 2385 } 2386 2387 struct ifcap_nv_bit_name { 2388 uint64_t cap_bit; 2389 const char *cap_name; 2390 }; 2391 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2392 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2393 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2394 CAPNV(RXCSUM), 2395 CAPNV(TXCSUM), 2396 CAPNV(NETCONS), 2397 CAPNV(VLAN_MTU), 2398 CAPNV(VLAN_HWTAGGING), 2399 CAPNV(JUMBO_MTU), 2400 CAPNV(POLLING), 2401 CAPNV(VLAN_HWCSUM), 2402 CAPNV(TSO4), 2403 CAPNV(TSO6), 2404 CAPNV(LRO), 2405 CAPNV(WOL_UCAST), 2406 CAPNV(WOL_MCAST), 2407 CAPNV(WOL_MAGIC), 2408 CAPNV(TOE4), 2409 CAPNV(TOE6), 2410 CAPNV(VLAN_HWFILTER), 2411 CAPNV(VLAN_HWTSO), 2412 CAPNV(LINKSTATE), 2413 CAPNV(NETMAP), 2414 CAPNV(RXCSUM_IPV6), 2415 CAPNV(TXCSUM_IPV6), 2416 CAPNV(HWSTATS), 2417 CAPNV(TXRTLMT), 2418 CAPNV(HWRXTSTMP), 2419 CAPNV(MEXTPG), 2420 CAPNV(TXTLS4), 2421 CAPNV(TXTLS6), 2422 CAPNV(VXLAN_HWCSUM), 2423 CAPNV(VXLAN_HWTSO), 2424 CAPNV(TXTLS_RTLMT), 2425 {0, NULL} 2426 }; 2427 #define CAP2NV(x) {.cap_bit = IFCAP2_##x, \ 2428 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2429 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2430 CAP2NV(RXTLS4), 2431 CAP2NV(RXTLS6), 2432 {0, NULL} 2433 }; 2434 #undef CAPNV 2435 #undef CAP2NV 2436 2437 int 2438 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2439 const struct ifcap_nv_bit_name *nn, bool all) 2440 { 2441 int i, res; 2442 2443 res = 0; 2444 for (i = 0; nn[i].cap_name != NULL; i++) { 2445 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2446 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2447 res |= nn[i].cap_bit; 2448 } else { 2449 res |= *old_cap & nn[i].cap_bit; 2450 } 2451 } 2452 return (res); 2453 } 2454 2455 void 2456 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2457 int ifr_cap, int ifr_req) 2458 { 2459 int i; 2460 2461 for (i = 0; nn[i].cap_name != NULL; i++) { 2462 if ((nn[i].cap_bit & ifr_cap) != 0) { 2463 nvlist_add_bool(nv, nn[i].cap_name, 2464 (nn[i].cap_bit & ifr_req) != 0); 2465 } 2466 } 2467 } 2468 2469 /* 2470 * Hardware specific interface ioctls. 2471 */ 2472 int 2473 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2474 { 2475 struct ifreq *ifr; 2476 int error = 0, do_ifup = 0; 2477 int new_flags, temp_flags; 2478 size_t namelen, onamelen; 2479 size_t descrlen, nvbuflen; 2480 char *descrbuf; 2481 char new_name[IFNAMSIZ]; 2482 char old_name[IFNAMSIZ], strbuf[IFNAMSIZ + 8]; 2483 struct ifaddr *ifa; 2484 struct sockaddr_dl *sdl; 2485 void *buf; 2486 nvlist_t *nvcap; 2487 struct siocsifcapnv_driver_data drv_ioctl_data; 2488 2489 ifr = (struct ifreq *)data; 2490 switch (cmd) { 2491 case SIOCGIFINDEX: 2492 ifr->ifr_index = ifp->if_index; 2493 break; 2494 2495 case SIOCGIFFLAGS: 2496 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2497 ifr->ifr_flags = temp_flags & 0xffff; 2498 ifr->ifr_flagshigh = temp_flags >> 16; 2499 break; 2500 2501 case SIOCGIFCAP: 2502 ifr->ifr_reqcap = ifp->if_capabilities; 2503 ifr->ifr_curcap = ifp->if_capenable; 2504 break; 2505 2506 case SIOCGIFCAPNV: 2507 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2508 error = EINVAL; 2509 break; 2510 } 2511 buf = NULL; 2512 nvcap = nvlist_create(0); 2513 for (;;) { 2514 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2515 ifp->if_capabilities, ifp->if_capenable); 2516 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2517 ifp->if_capabilities2, ifp->if_capenable2); 2518 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2519 __DECONST(caddr_t, nvcap)); 2520 if (error != 0) { 2521 if_printf(ifp, 2522 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2523 error); 2524 break; 2525 } 2526 buf = nvlist_pack(nvcap, &nvbuflen); 2527 if (buf == NULL) { 2528 error = nvlist_error(nvcap); 2529 if (error == 0) 2530 error = EDOOFUS; 2531 break; 2532 } 2533 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2534 ifr->ifr_cap_nv.length = nvbuflen; 2535 ifr->ifr_cap_nv.buffer = NULL; 2536 error = EFBIG; 2537 break; 2538 } 2539 ifr->ifr_cap_nv.length = nvbuflen; 2540 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2541 break; 2542 } 2543 free(buf, M_NVLIST); 2544 nvlist_destroy(nvcap); 2545 break; 2546 2547 case SIOCGIFDATA: 2548 { 2549 struct if_data ifd; 2550 2551 /* Ensure uninitialised padding is not leaked. */ 2552 memset(&ifd, 0, sizeof(ifd)); 2553 2554 if_data_copy(ifp, &ifd); 2555 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2556 break; 2557 } 2558 2559 #ifdef MAC 2560 case SIOCGIFMAC: 2561 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2562 break; 2563 #endif 2564 2565 case SIOCGIFMETRIC: 2566 ifr->ifr_metric = ifp->if_metric; 2567 break; 2568 2569 case SIOCGIFMTU: 2570 ifr->ifr_mtu = ifp->if_mtu; 2571 break; 2572 2573 case SIOCGIFPHYS: 2574 /* XXXGL: did this ever worked? */ 2575 ifr->ifr_phys = 0; 2576 break; 2577 2578 case SIOCGIFDESCR: 2579 error = 0; 2580 sx_slock(&ifdescr_sx); 2581 if (ifp->if_description == NULL) 2582 error = ENOMSG; 2583 else { 2584 /* space for terminating nul */ 2585 descrlen = strlen(ifp->if_description) + 1; 2586 if (ifr_buffer_get_length(ifr) < descrlen) 2587 ifr_buffer_set_buffer_null(ifr); 2588 else 2589 error = copyout(ifp->if_description, 2590 ifr_buffer_get_buffer(ifr), descrlen); 2591 ifr_buffer_set_length(ifr, descrlen); 2592 } 2593 sx_sunlock(&ifdescr_sx); 2594 break; 2595 2596 case SIOCSIFDESCR: 2597 error = priv_check(td, PRIV_NET_SETIFDESCR); 2598 if (error) 2599 return (error); 2600 2601 /* 2602 * Copy only (length-1) bytes to make sure that 2603 * if_description is always nul terminated. The 2604 * length parameter is supposed to count the 2605 * terminating nul in. 2606 */ 2607 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2608 return (ENAMETOOLONG); 2609 else if (ifr_buffer_get_length(ifr) == 0) 2610 descrbuf = NULL; 2611 else { 2612 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2613 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2614 ifr_buffer_get_length(ifr) - 1); 2615 if (error) { 2616 if_freedescr(descrbuf); 2617 break; 2618 } 2619 } 2620 2621 if_setdescr(ifp, descrbuf); 2622 getmicrotime(&ifp->if_lastchange); 2623 break; 2624 2625 case SIOCGIFFIB: 2626 ifr->ifr_fib = ifp->if_fib; 2627 break; 2628 2629 case SIOCSIFFIB: 2630 error = priv_check(td, PRIV_NET_SETIFFIB); 2631 if (error) 2632 return (error); 2633 if (ifr->ifr_fib >= rt_numfibs) 2634 return (EINVAL); 2635 2636 ifp->if_fib = ifr->ifr_fib; 2637 break; 2638 2639 case SIOCSIFFLAGS: 2640 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2641 if (error) 2642 return (error); 2643 /* 2644 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2645 * check, so we don't need special handling here yet. 2646 */ 2647 new_flags = (ifr->ifr_flags & 0xffff) | 2648 (ifr->ifr_flagshigh << 16); 2649 if (ifp->if_flags & IFF_UP && 2650 (new_flags & IFF_UP) == 0) { 2651 if_down(ifp); 2652 } else if (new_flags & IFF_UP && 2653 (ifp->if_flags & IFF_UP) == 0) { 2654 do_ifup = 1; 2655 } 2656 /* See if permanently promiscuous mode bit is about to flip */ 2657 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2658 if (new_flags & IFF_PPROMISC) 2659 ifp->if_flags |= IFF_PROMISC; 2660 else if (ifp->if_pcount == 0) 2661 ifp->if_flags &= ~IFF_PROMISC; 2662 if (log_promisc_mode_change) 2663 if_printf(ifp, "permanently promiscuous mode %s\n", 2664 ((new_flags & IFF_PPROMISC) ? 2665 "enabled" : "disabled")); 2666 } 2667 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2668 (new_flags &~ IFF_CANTCHANGE); 2669 if (ifp->if_ioctl) { 2670 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2671 } 2672 if (do_ifup) 2673 if_up(ifp); 2674 getmicrotime(&ifp->if_lastchange); 2675 break; 2676 2677 case SIOCSIFCAP: 2678 error = priv_check(td, PRIV_NET_SETIFCAP); 2679 if (error != 0) 2680 return (error); 2681 if (ifp->if_ioctl == NULL) 2682 return (EOPNOTSUPP); 2683 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2684 return (EINVAL); 2685 error = (*ifp->if_ioctl)(ifp, cmd, data); 2686 if (error == 0) 2687 getmicrotime(&ifp->if_lastchange); 2688 break; 2689 2690 case SIOCSIFCAPNV: 2691 error = priv_check(td, PRIV_NET_SETIFCAP); 2692 if (error != 0) 2693 return (error); 2694 if (ifp->if_ioctl == NULL) 2695 return (EOPNOTSUPP); 2696 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2697 return (EINVAL); 2698 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2699 return (EINVAL); 2700 nvcap = NULL; 2701 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2702 for (;;) { 2703 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2704 ifr->ifr_cap_nv.length); 2705 if (error != 0) 2706 break; 2707 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2708 if (nvcap == NULL) { 2709 error = EINVAL; 2710 break; 2711 } 2712 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2713 &ifp->if_capenable, ifcap_nv_bit_names, false); 2714 if ((drv_ioctl_data.reqcap & 2715 ~ifp->if_capabilities) != 0) { 2716 error = EINVAL; 2717 break; 2718 } 2719 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2720 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2721 if ((drv_ioctl_data.reqcap2 & 2722 ~ifp->if_capabilities2) != 0) { 2723 error = EINVAL; 2724 break; 2725 } 2726 drv_ioctl_data.nvcap = nvcap; 2727 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2728 (caddr_t)&drv_ioctl_data); 2729 break; 2730 } 2731 nvlist_destroy(nvcap); 2732 free(buf, M_TEMP); 2733 if (error == 0) 2734 getmicrotime(&ifp->if_lastchange); 2735 break; 2736 2737 #ifdef MAC 2738 case SIOCSIFMAC: 2739 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2740 break; 2741 #endif 2742 2743 case SIOCSIFNAME: 2744 error = priv_check(td, PRIV_NET_SETIFNAME); 2745 if (error) 2746 return (error); 2747 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2748 NULL); 2749 if (error != 0) 2750 return (error); 2751 if (new_name[0] == '\0') 2752 return (EINVAL); 2753 if (strcmp(new_name, ifp->if_xname) == 0) 2754 break; 2755 if (ifunit(new_name) != NULL) 2756 return (EEXIST); 2757 2758 /* 2759 * XXX: Locking. Nothing else seems to lock if_flags, 2760 * and there are numerous other races with the 2761 * ifunit() checks not being atomic with namespace 2762 * changes (renames, vmoves, if_attach, etc). 2763 */ 2764 ifp->if_flags |= IFF_RENAMING; 2765 2766 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2767 2768 if_printf(ifp, "changing name to '%s'\n", new_name); 2769 2770 IF_ADDR_WLOCK(ifp); 2771 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 2772 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2773 ifa = ifp->if_addr; 2774 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2775 namelen = strlen(new_name); 2776 onamelen = sdl->sdl_nlen; 2777 /* 2778 * Move the address if needed. This is safe because we 2779 * allocate space for a name of length IFNAMSIZ when we 2780 * create this in if_attach(). 2781 */ 2782 if (namelen != onamelen) { 2783 bcopy(sdl->sdl_data + onamelen, 2784 sdl->sdl_data + namelen, sdl->sdl_alen); 2785 } 2786 bcopy(new_name, sdl->sdl_data, namelen); 2787 sdl->sdl_nlen = namelen; 2788 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2789 bzero(sdl->sdl_data, onamelen); 2790 while (namelen != 0) 2791 sdl->sdl_data[--namelen] = 0xff; 2792 IF_ADDR_WUNLOCK(ifp); 2793 2794 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2795 2796 ifp->if_flags &= ~IFF_RENAMING; 2797 2798 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 2799 devctl_notify("IFNET", old_name, "RENAME", strbuf); 2800 break; 2801 2802 #ifdef VIMAGE 2803 case SIOCSIFVNET: 2804 error = priv_check(td, PRIV_NET_SETIFVNET); 2805 if (error) 2806 return (error); 2807 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2808 break; 2809 #endif 2810 2811 case SIOCSIFMETRIC: 2812 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2813 if (error) 2814 return (error); 2815 ifp->if_metric = ifr->ifr_metric; 2816 getmicrotime(&ifp->if_lastchange); 2817 break; 2818 2819 case SIOCSIFPHYS: 2820 error = priv_check(td, PRIV_NET_SETIFPHYS); 2821 if (error) 2822 return (error); 2823 if (ifp->if_ioctl == NULL) 2824 return (EOPNOTSUPP); 2825 error = (*ifp->if_ioctl)(ifp, cmd, data); 2826 if (error == 0) 2827 getmicrotime(&ifp->if_lastchange); 2828 break; 2829 2830 case SIOCSIFMTU: 2831 { 2832 u_long oldmtu = ifp->if_mtu; 2833 2834 error = priv_check(td, PRIV_NET_SETIFMTU); 2835 if (error) 2836 return (error); 2837 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2838 return (EINVAL); 2839 if (ifp->if_ioctl == NULL) 2840 return (EOPNOTSUPP); 2841 /* Disallow MTU changes on bridge member interfaces. */ 2842 if (ifp->if_bridge) 2843 return (EOPNOTSUPP); 2844 error = (*ifp->if_ioctl)(ifp, cmd, data); 2845 if (error == 0) { 2846 getmicrotime(&ifp->if_lastchange); 2847 rt_ifmsg(ifp, 0); 2848 #ifdef INET 2849 DEBUGNET_NOTIFY_MTU(ifp); 2850 #endif 2851 } 2852 /* 2853 * If the link MTU changed, do network layer specific procedure. 2854 */ 2855 if (ifp->if_mtu != oldmtu) { 2856 #ifdef INET6 2857 nd6_setmtu(ifp); 2858 #endif 2859 rt_updatemtu(ifp); 2860 } 2861 break; 2862 } 2863 2864 case SIOCADDMULTI: 2865 case SIOCDELMULTI: 2866 if (cmd == SIOCADDMULTI) 2867 error = priv_check(td, PRIV_NET_ADDMULTI); 2868 else 2869 error = priv_check(td, PRIV_NET_DELMULTI); 2870 if (error) 2871 return (error); 2872 2873 /* Don't allow group membership on non-multicast interfaces. */ 2874 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2875 return (EOPNOTSUPP); 2876 2877 /* Don't let users screw up protocols' entries. */ 2878 if (ifr->ifr_addr.sa_family != AF_LINK) 2879 return (EINVAL); 2880 2881 if (cmd == SIOCADDMULTI) { 2882 struct epoch_tracker et; 2883 struct ifmultiaddr *ifma; 2884 2885 /* 2886 * Userland is only permitted to join groups once 2887 * via the if_addmulti() KPI, because it cannot hold 2888 * struct ifmultiaddr * between calls. It may also 2889 * lose a race while we check if the membership 2890 * already exists. 2891 */ 2892 NET_EPOCH_ENTER(et); 2893 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2894 NET_EPOCH_EXIT(et); 2895 if (ifma != NULL) 2896 error = EADDRINUSE; 2897 else 2898 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2899 } else { 2900 error = if_delmulti(ifp, &ifr->ifr_addr); 2901 } 2902 if (error == 0) 2903 getmicrotime(&ifp->if_lastchange); 2904 break; 2905 2906 case SIOCSIFPHYADDR: 2907 case SIOCDIFPHYADDR: 2908 #ifdef INET6 2909 case SIOCSIFPHYADDR_IN6: 2910 #endif 2911 case SIOCSIFMEDIA: 2912 case SIOCSIFGENERIC: 2913 error = priv_check(td, PRIV_NET_HWIOCTL); 2914 if (error) 2915 return (error); 2916 if (ifp->if_ioctl == NULL) 2917 return (EOPNOTSUPP); 2918 error = (*ifp->if_ioctl)(ifp, cmd, data); 2919 if (error == 0) 2920 getmicrotime(&ifp->if_lastchange); 2921 break; 2922 2923 case SIOCGIFSTATUS: 2924 case SIOCGIFPSRCADDR: 2925 case SIOCGIFPDSTADDR: 2926 case SIOCGIFMEDIA: 2927 case SIOCGIFXMEDIA: 2928 case SIOCGIFGENERIC: 2929 case SIOCGIFRSSKEY: 2930 case SIOCGIFRSSHASH: 2931 case SIOCGIFDOWNREASON: 2932 if (ifp->if_ioctl == NULL) 2933 return (EOPNOTSUPP); 2934 error = (*ifp->if_ioctl)(ifp, cmd, data); 2935 break; 2936 2937 case SIOCSIFLLADDR: 2938 error = priv_check(td, PRIV_NET_SETLLADDR); 2939 if (error) 2940 return (error); 2941 error = if_setlladdr(ifp, 2942 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2943 break; 2944 2945 case SIOCGHWADDR: 2946 error = if_gethwaddr(ifp, ifr); 2947 break; 2948 2949 case SIOCAIFGROUP: 2950 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2951 if (error) 2952 return (error); 2953 error = if_addgroup(ifp, 2954 ((struct ifgroupreq *)data)->ifgr_group); 2955 if (error != 0) 2956 return (error); 2957 break; 2958 2959 case SIOCGIFGROUP: 2960 { 2961 struct epoch_tracker et; 2962 2963 NET_EPOCH_ENTER(et); 2964 error = if_getgroup((struct ifgroupreq *)data, ifp); 2965 NET_EPOCH_EXIT(et); 2966 break; 2967 } 2968 2969 case SIOCDIFGROUP: 2970 error = priv_check(td, PRIV_NET_DELIFGROUP); 2971 if (error) 2972 return (error); 2973 error = if_delgroup(ifp, 2974 ((struct ifgroupreq *)data)->ifgr_group); 2975 if (error != 0) 2976 return (error); 2977 break; 2978 2979 default: 2980 error = ENOIOCTL; 2981 break; 2982 } 2983 return (error); 2984 } 2985 2986 /* 2987 * Interface ioctls. 2988 */ 2989 int 2990 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2991 { 2992 #ifdef COMPAT_FREEBSD32 2993 union { 2994 struct ifconf ifc; 2995 struct ifdrv ifd; 2996 struct ifgroupreq ifgr; 2997 struct ifmediareq ifmr; 2998 } thunk; 2999 u_long saved_cmd; 3000 struct ifconf32 *ifc32; 3001 struct ifdrv32 *ifd32; 3002 struct ifgroupreq32 *ifgr32; 3003 struct ifmediareq32 *ifmr32; 3004 #endif 3005 struct ifnet *ifp; 3006 struct ifreq *ifr; 3007 int error; 3008 int oif_flags; 3009 #ifdef VIMAGE 3010 bool shutdown; 3011 #endif 3012 3013 CURVNET_SET(so->so_vnet); 3014 #ifdef VIMAGE 3015 /* Make sure the VNET is stable. */ 3016 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 3017 if (shutdown) { 3018 CURVNET_RESTORE(); 3019 return (EBUSY); 3020 } 3021 #endif 3022 3023 #ifdef COMPAT_FREEBSD32 3024 saved_cmd = cmd; 3025 switch (cmd) { 3026 case SIOCGIFCONF32: 3027 ifc32 = (struct ifconf32 *)data; 3028 thunk.ifc.ifc_len = ifc32->ifc_len; 3029 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 3030 data = (caddr_t)&thunk.ifc; 3031 cmd = SIOCGIFCONF; 3032 break; 3033 case SIOCGDRVSPEC32: 3034 case SIOCSDRVSPEC32: 3035 ifd32 = (struct ifdrv32 *)data; 3036 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 3037 sizeof(thunk.ifd.ifd_name)); 3038 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 3039 thunk.ifd.ifd_len = ifd32->ifd_len; 3040 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 3041 data = (caddr_t)&thunk.ifd; 3042 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 3043 break; 3044 case SIOCAIFGROUP32: 3045 case SIOCGIFGROUP32: 3046 case SIOCDIFGROUP32: 3047 case SIOCGIFGMEMB32: 3048 ifgr32 = (struct ifgroupreq32 *)data; 3049 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 3050 sizeof(thunk.ifgr.ifgr_name)); 3051 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 3052 switch (cmd) { 3053 case SIOCAIFGROUP32: 3054 case SIOCDIFGROUP32: 3055 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 3056 sizeof(thunk.ifgr.ifgr_group)); 3057 break; 3058 case SIOCGIFGROUP32: 3059 case SIOCGIFGMEMB32: 3060 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 3061 break; 3062 } 3063 data = (caddr_t)&thunk.ifgr; 3064 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 3065 break; 3066 case SIOCGIFMEDIA32: 3067 case SIOCGIFXMEDIA32: 3068 ifmr32 = (struct ifmediareq32 *)data; 3069 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 3070 sizeof(thunk.ifmr.ifm_name)); 3071 thunk.ifmr.ifm_current = ifmr32->ifm_current; 3072 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 3073 thunk.ifmr.ifm_status = ifmr32->ifm_status; 3074 thunk.ifmr.ifm_active = ifmr32->ifm_active; 3075 thunk.ifmr.ifm_count = ifmr32->ifm_count; 3076 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 3077 data = (caddr_t)&thunk.ifmr; 3078 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 3079 break; 3080 } 3081 #endif 3082 3083 switch (cmd) { 3084 case SIOCGIFCONF: 3085 error = ifconf(cmd, data); 3086 goto out_noref; 3087 } 3088 3089 ifr = (struct ifreq *)data; 3090 switch (cmd) { 3091 #ifdef VIMAGE 3092 case SIOCSIFRVNET: 3093 error = priv_check(td, PRIV_NET_SETIFVNET); 3094 if (error == 0) 3095 error = if_vmove_reclaim(td, ifr->ifr_name, 3096 ifr->ifr_jid); 3097 goto out_noref; 3098 #endif 3099 case SIOCIFCREATE: 3100 case SIOCIFCREATE2: 3101 error = priv_check(td, PRIV_NET_IFCREATE); 3102 if (error == 0) 3103 error = if_clone_create(ifr->ifr_name, 3104 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3105 ifr_data_get_ptr(ifr) : NULL); 3106 goto out_noref; 3107 case SIOCIFDESTROY: 3108 error = priv_check(td, PRIV_NET_IFDESTROY); 3109 3110 if (error == 0) { 3111 sx_xlock(&ifnet_detach_sxlock); 3112 error = if_clone_destroy(ifr->ifr_name); 3113 sx_xunlock(&ifnet_detach_sxlock); 3114 } 3115 goto out_noref; 3116 3117 case SIOCIFGCLONERS: 3118 error = if_clone_list((struct if_clonereq *)data); 3119 goto out_noref; 3120 3121 case SIOCGIFGMEMB: 3122 error = if_getgroupmembers((struct ifgroupreq *)data); 3123 goto out_noref; 3124 3125 #if defined(INET) || defined(INET6) 3126 case SIOCSVH: 3127 case SIOCGVH: 3128 if (carp_ioctl_p == NULL) 3129 error = EPROTONOSUPPORT; 3130 else 3131 error = (*carp_ioctl_p)(ifr, cmd, td); 3132 goto out_noref; 3133 #endif 3134 } 3135 3136 ifp = ifunit_ref(ifr->ifr_name); 3137 if (ifp == NULL) { 3138 error = ENXIO; 3139 goto out_noref; 3140 } 3141 3142 error = ifhwioctl(cmd, ifp, data, td); 3143 if (error != ENOIOCTL) 3144 goto out_ref; 3145 3146 oif_flags = ifp->if_flags; 3147 if (so->so_proto == NULL) { 3148 error = EOPNOTSUPP; 3149 goto out_ref; 3150 } 3151 3152 /* 3153 * Pass the request on to the socket control method, and if the 3154 * latter returns EOPNOTSUPP, directly to the interface. 3155 * 3156 * Make an exception for the legacy SIOCSIF* requests. Drivers 3157 * trust SIOCSIFADDR et al to come from an already privileged 3158 * layer, and do not perform any credentials checks or input 3159 * validation. 3160 */ 3161 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 3162 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3163 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3164 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3165 error = (*ifp->if_ioctl)(ifp, cmd, data); 3166 3167 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3168 #ifdef INET6 3169 if (ifp->if_flags & IFF_UP) 3170 in6_if_up(ifp); 3171 #endif 3172 } 3173 3174 out_ref: 3175 if_rele(ifp); 3176 out_noref: 3177 CURVNET_RESTORE(); 3178 #ifdef COMPAT_FREEBSD32 3179 if (error != 0) 3180 return (error); 3181 switch (saved_cmd) { 3182 case SIOCGIFCONF32: 3183 ifc32->ifc_len = thunk.ifc.ifc_len; 3184 break; 3185 case SIOCGDRVSPEC32: 3186 /* 3187 * SIOCGDRVSPEC is IOWR, but nothing actually touches 3188 * the struct so just assert that ifd_len (the only 3189 * field it might make sense to update) hasn't 3190 * changed. 3191 */ 3192 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 3193 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 3194 thunk.ifd.ifd_len)); 3195 break; 3196 case SIOCGIFGROUP32: 3197 case SIOCGIFGMEMB32: 3198 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 3199 break; 3200 case SIOCGIFMEDIA32: 3201 case SIOCGIFXMEDIA32: 3202 ifmr32->ifm_current = thunk.ifmr.ifm_current; 3203 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 3204 ifmr32->ifm_status = thunk.ifmr.ifm_status; 3205 ifmr32->ifm_active = thunk.ifmr.ifm_active; 3206 ifmr32->ifm_count = thunk.ifmr.ifm_count; 3207 break; 3208 } 3209 #endif 3210 return (error); 3211 } 3212 3213 /* 3214 * The code common to handling reference counted flags, 3215 * e.g., in ifpromisc() and if_allmulti(). 3216 * The "pflag" argument can specify a permanent mode flag to check, 3217 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3218 * 3219 * Only to be used on stack-owned flags, not driver-owned flags. 3220 */ 3221 static int 3222 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3223 { 3224 struct ifreq ifr; 3225 int error; 3226 int oldflags, oldcount; 3227 3228 /* Sanity checks to catch programming errors */ 3229 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3230 ("%s: setting driver-owned flag %d", __func__, flag)); 3231 3232 if (onswitch) 3233 KASSERT(*refcount >= 0, 3234 ("%s: increment negative refcount %d for flag %d", 3235 __func__, *refcount, flag)); 3236 else 3237 KASSERT(*refcount > 0, 3238 ("%s: decrement non-positive refcount %d for flag %d", 3239 __func__, *refcount, flag)); 3240 3241 /* In case this mode is permanent, just touch refcount */ 3242 if (ifp->if_flags & pflag) { 3243 *refcount += onswitch ? 1 : -1; 3244 return (0); 3245 } 3246 3247 /* Save ifnet parameters for if_ioctl() may fail */ 3248 oldcount = *refcount; 3249 oldflags = ifp->if_flags; 3250 3251 /* 3252 * See if we aren't the only and touching refcount is enough. 3253 * Actually toggle interface flag if we are the first or last. 3254 */ 3255 if (onswitch) { 3256 if ((*refcount)++) 3257 return (0); 3258 ifp->if_flags |= flag; 3259 } else { 3260 if (--(*refcount)) 3261 return (0); 3262 ifp->if_flags &= ~flag; 3263 } 3264 3265 /* Call down the driver since we've changed interface flags */ 3266 if (ifp->if_ioctl == NULL) { 3267 error = EOPNOTSUPP; 3268 goto recover; 3269 } 3270 ifr.ifr_flags = ifp->if_flags & 0xffff; 3271 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3272 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3273 if (error) 3274 goto recover; 3275 /* Notify userland that interface flags have changed */ 3276 rt_ifmsg(ifp, flag); 3277 return (0); 3278 3279 recover: 3280 /* Recover after driver error */ 3281 *refcount = oldcount; 3282 ifp->if_flags = oldflags; 3283 return (error); 3284 } 3285 3286 /* 3287 * Set/clear promiscuous mode on interface ifp based on the truth value 3288 * of pswitch. The calls are reference counted so that only the first 3289 * "on" request actually has an effect, as does the final "off" request. 3290 * Results are undefined if the "off" and "on" requests are not matched. 3291 */ 3292 int 3293 ifpromisc(struct ifnet *ifp, int pswitch) 3294 { 3295 int error; 3296 int oldflags = ifp->if_flags; 3297 3298 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3299 &ifp->if_pcount, pswitch); 3300 /* If promiscuous mode status has changed, log a message */ 3301 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3302 log_promisc_mode_change) 3303 if_printf(ifp, "promiscuous mode %s\n", 3304 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3305 return (error); 3306 } 3307 3308 /* 3309 * Return interface configuration 3310 * of system. List may be used 3311 * in later ioctl's (above) to get 3312 * other information. 3313 */ 3314 /*ARGSUSED*/ 3315 static int 3316 ifconf(u_long cmd, caddr_t data) 3317 { 3318 struct ifconf *ifc = (struct ifconf *)data; 3319 struct ifnet *ifp; 3320 struct ifaddr *ifa; 3321 struct ifreq ifr; 3322 struct sbuf *sb; 3323 int error, full = 0, valid_len, max_len; 3324 3325 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3326 max_len = maxphys - 1; 3327 3328 /* Prevent hostile input from being able to crash the system */ 3329 if (ifc->ifc_len <= 0) 3330 return (EINVAL); 3331 3332 again: 3333 if (ifc->ifc_len <= max_len) { 3334 max_len = ifc->ifc_len; 3335 full = 1; 3336 } 3337 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3338 max_len = 0; 3339 valid_len = 0; 3340 3341 IFNET_RLOCK(); 3342 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3343 struct epoch_tracker et; 3344 int addrs; 3345 3346 /* 3347 * Zero the ifr to make sure we don't disclose the contents 3348 * of the stack. 3349 */ 3350 memset(&ifr, 0, sizeof(ifr)); 3351 3352 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3353 >= sizeof(ifr.ifr_name)) { 3354 sbuf_delete(sb); 3355 IFNET_RUNLOCK(); 3356 return (ENAMETOOLONG); 3357 } 3358 3359 addrs = 0; 3360 NET_EPOCH_ENTER(et); 3361 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3362 struct sockaddr *sa = ifa->ifa_addr; 3363 3364 if (prison_if(curthread->td_ucred, sa) != 0) 3365 continue; 3366 addrs++; 3367 if (sa->sa_len <= sizeof(*sa)) { 3368 if (sa->sa_len < sizeof(*sa)) { 3369 memset(&ifr.ifr_ifru.ifru_addr, 0, 3370 sizeof(ifr.ifr_ifru.ifru_addr)); 3371 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3372 sa->sa_len); 3373 } else 3374 ifr.ifr_ifru.ifru_addr = *sa; 3375 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3376 max_len += sizeof(ifr); 3377 } else { 3378 sbuf_bcat(sb, &ifr, 3379 offsetof(struct ifreq, ifr_addr)); 3380 max_len += offsetof(struct ifreq, ifr_addr); 3381 sbuf_bcat(sb, sa, sa->sa_len); 3382 max_len += sa->sa_len; 3383 } 3384 3385 if (sbuf_error(sb) == 0) 3386 valid_len = sbuf_len(sb); 3387 } 3388 NET_EPOCH_EXIT(et); 3389 if (addrs == 0) { 3390 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3391 max_len += sizeof(ifr); 3392 3393 if (sbuf_error(sb) == 0) 3394 valid_len = sbuf_len(sb); 3395 } 3396 } 3397 IFNET_RUNLOCK(); 3398 3399 /* 3400 * If we didn't allocate enough space (uncommon), try again. If 3401 * we have already allocated as much space as we are allowed, 3402 * return what we've got. 3403 */ 3404 if (valid_len != max_len && !full) { 3405 sbuf_delete(sb); 3406 goto again; 3407 } 3408 3409 ifc->ifc_len = valid_len; 3410 sbuf_finish(sb); 3411 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3412 sbuf_delete(sb); 3413 return (error); 3414 } 3415 3416 /* 3417 * Just like ifpromisc(), but for all-multicast-reception mode. 3418 */ 3419 int 3420 if_allmulti(struct ifnet *ifp, int onswitch) 3421 { 3422 3423 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3424 } 3425 3426 struct ifmultiaddr * 3427 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3428 { 3429 struct ifmultiaddr *ifma; 3430 3431 IF_ADDR_LOCK_ASSERT(ifp); 3432 3433 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3434 if (sa->sa_family == AF_LINK) { 3435 if (sa_dl_equal(ifma->ifma_addr, sa)) 3436 break; 3437 } else { 3438 if (sa_equal(ifma->ifma_addr, sa)) 3439 break; 3440 } 3441 } 3442 3443 return ifma; 3444 } 3445 3446 /* 3447 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3448 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3449 * the ifnet multicast address list here, so the caller must do that and 3450 * other setup work (such as notifying the device driver). The reference 3451 * count is initialized to 1. 3452 */ 3453 static struct ifmultiaddr * 3454 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3455 int mflags) 3456 { 3457 struct ifmultiaddr *ifma; 3458 struct sockaddr *dupsa; 3459 3460 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3461 M_ZERO); 3462 if (ifma == NULL) 3463 return (NULL); 3464 3465 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3466 if (dupsa == NULL) { 3467 free(ifma, M_IFMADDR); 3468 return (NULL); 3469 } 3470 bcopy(sa, dupsa, sa->sa_len); 3471 ifma->ifma_addr = dupsa; 3472 3473 ifma->ifma_ifp = ifp; 3474 ifma->ifma_refcount = 1; 3475 ifma->ifma_protospec = NULL; 3476 3477 if (llsa == NULL) { 3478 ifma->ifma_lladdr = NULL; 3479 return (ifma); 3480 } 3481 3482 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3483 if (dupsa == NULL) { 3484 free(ifma->ifma_addr, M_IFMADDR); 3485 free(ifma, M_IFMADDR); 3486 return (NULL); 3487 } 3488 bcopy(llsa, dupsa, llsa->sa_len); 3489 ifma->ifma_lladdr = dupsa; 3490 3491 return (ifma); 3492 } 3493 3494 /* 3495 * if_freemulti: free ifmultiaddr structure and possibly attached related 3496 * addresses. The caller is responsible for implementing reference 3497 * counting, notifying the driver, handling routing messages, and releasing 3498 * any dependent link layer state. 3499 */ 3500 #ifdef MCAST_VERBOSE 3501 extern void kdb_backtrace(void); 3502 #endif 3503 static void 3504 if_freemulti_internal(struct ifmultiaddr *ifma) 3505 { 3506 3507 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3508 ifma->ifma_refcount)); 3509 3510 if (ifma->ifma_lladdr != NULL) 3511 free(ifma->ifma_lladdr, M_IFMADDR); 3512 #ifdef MCAST_VERBOSE 3513 kdb_backtrace(); 3514 printf("%s freeing ifma: %p\n", __func__, ifma); 3515 #endif 3516 free(ifma->ifma_addr, M_IFMADDR); 3517 free(ifma, M_IFMADDR); 3518 } 3519 3520 static void 3521 if_destroymulti(epoch_context_t ctx) 3522 { 3523 struct ifmultiaddr *ifma; 3524 3525 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3526 if_freemulti_internal(ifma); 3527 } 3528 3529 void 3530 if_freemulti(struct ifmultiaddr *ifma) 3531 { 3532 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3533 ifma->ifma_refcount)); 3534 3535 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3536 } 3537 3538 /* 3539 * Register an additional multicast address with a network interface. 3540 * 3541 * - If the address is already present, bump the reference count on the 3542 * address and return. 3543 * - If the address is not link-layer, look up a link layer address. 3544 * - Allocate address structures for one or both addresses, and attach to the 3545 * multicast address list on the interface. If automatically adding a link 3546 * layer address, the protocol address will own a reference to the link 3547 * layer address, to be freed when it is freed. 3548 * - Notify the network device driver of an addition to the multicast address 3549 * list. 3550 * 3551 * 'sa' points to caller-owned memory with the desired multicast address. 3552 * 3553 * 'retifma' will be used to return a pointer to the resulting multicast 3554 * address reference, if desired. 3555 */ 3556 int 3557 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3558 struct ifmultiaddr **retifma) 3559 { 3560 struct ifmultiaddr *ifma, *ll_ifma; 3561 struct sockaddr *llsa; 3562 struct sockaddr_dl sdl; 3563 int error; 3564 3565 #ifdef INET 3566 IN_MULTI_LIST_UNLOCK_ASSERT(); 3567 #endif 3568 #ifdef INET6 3569 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3570 #endif 3571 /* 3572 * If the address is already present, return a new reference to it; 3573 * otherwise, allocate storage and set up a new address. 3574 */ 3575 IF_ADDR_WLOCK(ifp); 3576 ifma = if_findmulti(ifp, sa); 3577 if (ifma != NULL) { 3578 ifma->ifma_refcount++; 3579 if (retifma != NULL) 3580 *retifma = ifma; 3581 IF_ADDR_WUNLOCK(ifp); 3582 return (0); 3583 } 3584 3585 /* 3586 * The address isn't already present; resolve the protocol address 3587 * into a link layer address, and then look that up, bump its 3588 * refcount or allocate an ifma for that also. 3589 * Most link layer resolving functions returns address data which 3590 * fits inside default sockaddr_dl structure. However callback 3591 * can allocate another sockaddr structure, in that case we need to 3592 * free it later. 3593 */ 3594 llsa = NULL; 3595 ll_ifma = NULL; 3596 if (ifp->if_resolvemulti != NULL) { 3597 /* Provide called function with buffer size information */ 3598 sdl.sdl_len = sizeof(sdl); 3599 llsa = (struct sockaddr *)&sdl; 3600 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3601 if (error) 3602 goto unlock_out; 3603 } 3604 3605 /* 3606 * Allocate the new address. Don't hook it up yet, as we may also 3607 * need to allocate a link layer multicast address. 3608 */ 3609 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3610 if (ifma == NULL) { 3611 error = ENOMEM; 3612 goto free_llsa_out; 3613 } 3614 3615 /* 3616 * If a link layer address is found, we'll need to see if it's 3617 * already present in the address list, or allocate is as well. 3618 * When this block finishes, the link layer address will be on the 3619 * list. 3620 */ 3621 if (llsa != NULL) { 3622 ll_ifma = if_findmulti(ifp, llsa); 3623 if (ll_ifma == NULL) { 3624 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3625 if (ll_ifma == NULL) { 3626 --ifma->ifma_refcount; 3627 if_freemulti(ifma); 3628 error = ENOMEM; 3629 goto free_llsa_out; 3630 } 3631 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3632 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3633 ifma_link); 3634 } else 3635 ll_ifma->ifma_refcount++; 3636 ifma->ifma_llifma = ll_ifma; 3637 } 3638 3639 /* 3640 * We now have a new multicast address, ifma, and possibly a new or 3641 * referenced link layer address. Add the primary address to the 3642 * ifnet address list. 3643 */ 3644 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3645 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3646 3647 if (retifma != NULL) 3648 *retifma = ifma; 3649 3650 /* 3651 * Must generate the message while holding the lock so that 'ifma' 3652 * pointer is still valid. 3653 */ 3654 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3655 IF_ADDR_WUNLOCK(ifp); 3656 3657 /* 3658 * We are certain we have added something, so call down to the 3659 * interface to let them know about it. 3660 */ 3661 if (ifp->if_ioctl != NULL) { 3662 if (THREAD_CAN_SLEEP()) 3663 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3664 else 3665 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3666 } 3667 3668 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3669 link_free_sdl(llsa); 3670 3671 return (0); 3672 3673 free_llsa_out: 3674 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3675 link_free_sdl(llsa); 3676 3677 unlock_out: 3678 IF_ADDR_WUNLOCK(ifp); 3679 return (error); 3680 } 3681 3682 static void 3683 if_siocaddmulti(void *arg, int pending) 3684 { 3685 struct ifnet *ifp; 3686 3687 ifp = arg; 3688 #ifdef DIAGNOSTIC 3689 if (pending > 1) 3690 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3691 #endif 3692 CURVNET_SET(ifp->if_vnet); 3693 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3694 CURVNET_RESTORE(); 3695 } 3696 3697 /* 3698 * Delete a multicast group membership by network-layer group address. 3699 * 3700 * Returns ENOENT if the entry could not be found. If ifp no longer 3701 * exists, results are undefined. This entry point should only be used 3702 * from subsystems which do appropriate locking to hold ifp for the 3703 * duration of the call. 3704 * Network-layer protocol domains must use if_delmulti_ifma(). 3705 */ 3706 int 3707 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3708 { 3709 struct ifmultiaddr *ifma; 3710 int lastref; 3711 3712 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3713 3714 IF_ADDR_WLOCK(ifp); 3715 lastref = 0; 3716 ifma = if_findmulti(ifp, sa); 3717 if (ifma != NULL) 3718 lastref = if_delmulti_locked(ifp, ifma, 0); 3719 IF_ADDR_WUNLOCK(ifp); 3720 3721 if (ifma == NULL) 3722 return (ENOENT); 3723 3724 if (lastref && ifp->if_ioctl != NULL) { 3725 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3726 } 3727 3728 return (0); 3729 } 3730 3731 /* 3732 * Delete all multicast group membership for an interface. 3733 * Should be used to quickly flush all multicast filters. 3734 */ 3735 void 3736 if_delallmulti(struct ifnet *ifp) 3737 { 3738 struct ifmultiaddr *ifma; 3739 struct ifmultiaddr *next; 3740 3741 IF_ADDR_WLOCK(ifp); 3742 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3743 if_delmulti_locked(ifp, ifma, 0); 3744 IF_ADDR_WUNLOCK(ifp); 3745 } 3746 3747 void 3748 if_delmulti_ifma(struct ifmultiaddr *ifma) 3749 { 3750 if_delmulti_ifma_flags(ifma, 0); 3751 } 3752 3753 /* 3754 * Delete a multicast group membership by group membership pointer. 3755 * Network-layer protocol domains must use this routine. 3756 * 3757 * It is safe to call this routine if the ifp disappeared. 3758 */ 3759 void 3760 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3761 { 3762 struct ifnet *ifp; 3763 int lastref; 3764 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3765 #ifdef INET 3766 IN_MULTI_LIST_UNLOCK_ASSERT(); 3767 #endif 3768 ifp = ifma->ifma_ifp; 3769 #ifdef DIAGNOSTIC 3770 if (ifp == NULL) { 3771 printf("%s: ifma_ifp seems to be detached\n", __func__); 3772 } else { 3773 struct epoch_tracker et; 3774 struct ifnet *oifp; 3775 3776 NET_EPOCH_ENTER(et); 3777 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3778 if (ifp == oifp) 3779 break; 3780 NET_EPOCH_EXIT(et); 3781 if (ifp != oifp) 3782 ifp = NULL; 3783 } 3784 #endif 3785 /* 3786 * If and only if the ifnet instance exists: Acquire the address lock. 3787 */ 3788 if (ifp != NULL) 3789 IF_ADDR_WLOCK(ifp); 3790 3791 lastref = if_delmulti_locked(ifp, ifma, flags); 3792 3793 if (ifp != NULL) { 3794 /* 3795 * If and only if the ifnet instance exists: 3796 * Release the address lock. 3797 * If the group was left: update the hardware hash filter. 3798 */ 3799 IF_ADDR_WUNLOCK(ifp); 3800 if (lastref && ifp->if_ioctl != NULL) { 3801 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3802 } 3803 } 3804 } 3805 3806 /* 3807 * Perform deletion of network-layer and/or link-layer multicast address. 3808 * 3809 * Return 0 if the reference count was decremented. 3810 * Return 1 if the final reference was released, indicating that the 3811 * hardware hash filter should be reprogrammed. 3812 */ 3813 static int 3814 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3815 { 3816 struct ifmultiaddr *ll_ifma; 3817 3818 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3819 KASSERT(ifma->ifma_ifp == ifp, 3820 ("%s: inconsistent ifp %p", __func__, ifp)); 3821 IF_ADDR_WLOCK_ASSERT(ifp); 3822 } 3823 3824 ifp = ifma->ifma_ifp; 3825 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3826 3827 /* 3828 * If the ifnet is detaching, null out references to ifnet, 3829 * so that upper protocol layers will notice, and not attempt 3830 * to obtain locks for an ifnet which no longer exists. The 3831 * routing socket announcement must happen before the ifnet 3832 * instance is detached from the system. 3833 */ 3834 if (detaching) { 3835 #ifdef DIAGNOSTIC 3836 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3837 #endif 3838 /* 3839 * ifp may already be nulled out if we are being reentered 3840 * to delete the ll_ifma. 3841 */ 3842 if (ifp != NULL) { 3843 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3844 ifma->ifma_ifp = NULL; 3845 } 3846 } 3847 3848 if (--ifma->ifma_refcount > 0) 3849 return 0; 3850 3851 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3852 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3853 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3854 } 3855 /* 3856 * If this ifma is a network-layer ifma, a link-layer ifma may 3857 * have been associated with it. Release it first if so. 3858 */ 3859 ll_ifma = ifma->ifma_llifma; 3860 if (ll_ifma != NULL) { 3861 KASSERT(ifma->ifma_lladdr != NULL, 3862 ("%s: llifma w/o lladdr", __func__)); 3863 if (detaching) 3864 ll_ifma->ifma_ifp = NULL; /* XXX */ 3865 if (--ll_ifma->ifma_refcount == 0) { 3866 if (ifp != NULL) { 3867 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3868 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3869 ifma_link); 3870 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3871 } 3872 } 3873 if_freemulti(ll_ifma); 3874 } 3875 } 3876 #ifdef INVARIANTS 3877 if (ifp) { 3878 struct ifmultiaddr *ifmatmp; 3879 3880 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3881 MPASS(ifma != ifmatmp); 3882 } 3883 #endif 3884 if_freemulti(ifma); 3885 /* 3886 * The last reference to this instance of struct ifmultiaddr 3887 * was released; the hardware should be notified of this change. 3888 */ 3889 return 1; 3890 } 3891 3892 /* 3893 * Set the link layer address on an interface. 3894 * 3895 * At this time we only support certain types of interfaces, 3896 * and we don't allow the length of the address to change. 3897 * 3898 * Set noinline to be dtrace-friendly 3899 */ 3900 __noinline int 3901 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3902 { 3903 struct sockaddr_dl *sdl; 3904 struct ifaddr *ifa; 3905 struct ifreq ifr; 3906 3907 ifa = ifp->if_addr; 3908 if (ifa == NULL) 3909 return (EINVAL); 3910 3911 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3912 if (sdl == NULL) 3913 return (EINVAL); 3914 3915 if (len != sdl->sdl_alen) /* don't allow length to change */ 3916 return (EINVAL); 3917 3918 switch (ifp->if_type) { 3919 case IFT_ETHER: 3920 case IFT_XETHER: 3921 case IFT_L2VLAN: 3922 case IFT_BRIDGE: 3923 case IFT_IEEE8023ADLAG: 3924 bcopy(lladdr, LLADDR(sdl), len); 3925 break; 3926 default: 3927 return (ENODEV); 3928 } 3929 3930 /* 3931 * If the interface is already up, we need 3932 * to re-init it in order to reprogram its 3933 * address filter. 3934 */ 3935 if ((ifp->if_flags & IFF_UP) != 0) { 3936 if (ifp->if_ioctl) { 3937 ifp->if_flags &= ~IFF_UP; 3938 ifr.ifr_flags = ifp->if_flags & 0xffff; 3939 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3940 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3941 ifp->if_flags |= IFF_UP; 3942 ifr.ifr_flags = ifp->if_flags & 0xffff; 3943 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3944 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3945 } 3946 } 3947 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3948 3949 return (0); 3950 } 3951 3952 /* 3953 * Compat function for handling basic encapsulation requests. 3954 * Not converted stacks (FDDI, IB, ..) supports traditional 3955 * output model: ARP (and other similar L2 protocols) are handled 3956 * inside output routine, arpresolve/nd6_resolve() returns MAC 3957 * address instead of full prepend. 3958 * 3959 * This function creates calculated header==MAC for IPv4/IPv6 and 3960 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3961 * address families. 3962 */ 3963 static int 3964 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3965 { 3966 3967 if (req->rtype != IFENCAP_LL) 3968 return (EOPNOTSUPP); 3969 3970 if (req->bufsize < req->lladdr_len) 3971 return (ENOMEM); 3972 3973 switch (req->family) { 3974 case AF_INET: 3975 case AF_INET6: 3976 break; 3977 default: 3978 return (EAFNOSUPPORT); 3979 } 3980 3981 /* Copy lladdr to storage as is */ 3982 memmove(req->buf, req->lladdr, req->lladdr_len); 3983 req->bufsize = req->lladdr_len; 3984 req->lladdr_off = 0; 3985 3986 return (0); 3987 } 3988 3989 /* 3990 * Tunnel interfaces can nest, also they may cause infinite recursion 3991 * calls when misconfigured. We'll prevent this by detecting loops. 3992 * High nesting level may cause stack exhaustion. We'll prevent this 3993 * by introducing upper limit. 3994 * 3995 * Return 0, if tunnel nesting count is equal or less than limit. 3996 */ 3997 int 3998 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3999 int limit) 4000 { 4001 struct m_tag *mtag; 4002 int count; 4003 4004 count = 1; 4005 mtag = NULL; 4006 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 4007 if (*(struct ifnet **)(mtag + 1) == ifp) { 4008 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 4009 return (EIO); 4010 } 4011 count++; 4012 } 4013 if (count > limit) { 4014 log(LOG_NOTICE, 4015 "%s: if_output recursively called too many times(%d)\n", 4016 if_name(ifp), count); 4017 return (EIO); 4018 } 4019 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 4020 if (mtag == NULL) 4021 return (ENOMEM); 4022 *(struct ifnet **)(mtag + 1) = ifp; 4023 m_tag_prepend(m, mtag); 4024 return (0); 4025 } 4026 4027 /* 4028 * Get the link layer address that was read from the hardware at attach. 4029 * 4030 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 4031 * their component interfaces as IFT_IEEE8023ADLAG. 4032 */ 4033 int 4034 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 4035 { 4036 4037 if (ifp->if_hw_addr == NULL) 4038 return (ENODEV); 4039 4040 switch (ifp->if_type) { 4041 case IFT_ETHER: 4042 case IFT_IEEE8023ADLAG: 4043 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 4044 return (0); 4045 default: 4046 return (ENODEV); 4047 } 4048 } 4049 4050 /* 4051 * The name argument must be a pointer to storage which will last as 4052 * long as the interface does. For physical devices, the result of 4053 * device_get_name(dev) is a good choice and for pseudo-devices a 4054 * static string works well. 4055 */ 4056 void 4057 if_initname(struct ifnet *ifp, const char *name, int unit) 4058 { 4059 ifp->if_dname = name; 4060 ifp->if_dunit = unit; 4061 if (unit != IF_DUNIT_NONE) 4062 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 4063 else 4064 strlcpy(ifp->if_xname, name, IFNAMSIZ); 4065 } 4066 4067 static int 4068 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 4069 { 4070 char if_fmt[256]; 4071 4072 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 4073 vlog(pri, if_fmt, ap); 4074 return (0); 4075 } 4076 4077 4078 int 4079 if_printf(struct ifnet *ifp, const char *fmt, ...) 4080 { 4081 va_list ap; 4082 4083 va_start(ap, fmt); 4084 if_vlog(ifp, LOG_INFO, fmt, ap); 4085 va_end(ap); 4086 return (0); 4087 } 4088 4089 int 4090 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 4091 { 4092 va_list ap; 4093 4094 va_start(ap, fmt); 4095 if_vlog(ifp, pri, fmt, ap); 4096 va_end(ap); 4097 return (0); 4098 } 4099 4100 void 4101 if_start(struct ifnet *ifp) 4102 { 4103 4104 (*(ifp)->if_start)(ifp); 4105 } 4106 4107 /* 4108 * Backwards compatibility interface for drivers 4109 * that have not implemented it 4110 */ 4111 static int 4112 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 4113 { 4114 int error; 4115 4116 IFQ_HANDOFF(ifp, m, error); 4117 return (error); 4118 } 4119 4120 static void 4121 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4122 { 4123 4124 m_freem(m); 4125 } 4126 4127 int 4128 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4129 { 4130 int active = 0; 4131 4132 IF_LOCK(ifq); 4133 if (_IF_QFULL(ifq)) { 4134 IF_UNLOCK(ifq); 4135 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4136 m_freem(m); 4137 return (0); 4138 } 4139 if (ifp != NULL) { 4140 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4141 if (m->m_flags & (M_BCAST|M_MCAST)) 4142 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4143 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4144 } 4145 _IF_ENQUEUE(ifq, m); 4146 IF_UNLOCK(ifq); 4147 if (ifp != NULL && !active) 4148 (*(ifp)->if_start)(ifp); 4149 return (1); 4150 } 4151 4152 void 4153 if_register_com_alloc(u_char type, 4154 if_com_alloc_t *a, if_com_free_t *f) 4155 { 4156 4157 KASSERT(if_com_alloc[type] == NULL, 4158 ("if_register_com_alloc: %d already registered", type)); 4159 KASSERT(if_com_free[type] == NULL, 4160 ("if_register_com_alloc: %d free already registered", type)); 4161 4162 if_com_alloc[type] = a; 4163 if_com_free[type] = f; 4164 } 4165 4166 void 4167 if_deregister_com_alloc(u_char type) 4168 { 4169 4170 KASSERT(if_com_alloc[type] != NULL, 4171 ("if_deregister_com_alloc: %d not registered", type)); 4172 KASSERT(if_com_free[type] != NULL, 4173 ("if_deregister_com_alloc: %d free not registered", type)); 4174 4175 /* 4176 * Ensure all pending EPOCH(9) callbacks have been executed. This 4177 * fixes issues about late invocation of if_destroy(), which leads 4178 * to memory leak from if_com_alloc[type] allocated if_l2com. 4179 */ 4180 NET_EPOCH_DRAIN_CALLBACKS(); 4181 4182 if_com_alloc[type] = NULL; 4183 if_com_free[type] = NULL; 4184 } 4185 4186 /* API for driver access to network stack owned ifnet.*/ 4187 uint64_t 4188 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4189 { 4190 uint64_t oldbrate; 4191 4192 oldbrate = ifp->if_baudrate; 4193 ifp->if_baudrate = baudrate; 4194 return (oldbrate); 4195 } 4196 4197 uint64_t 4198 if_getbaudrate(const if_t ifp) 4199 { 4200 4201 return (((struct ifnet *)ifp)->if_baudrate); 4202 } 4203 4204 int 4205 if_setcapabilities(if_t ifp, int capabilities) 4206 { 4207 ((struct ifnet *)ifp)->if_capabilities = capabilities; 4208 return (0); 4209 } 4210 4211 int 4212 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4213 { 4214 ((struct ifnet *)ifp)->if_capabilities |= setbit; 4215 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 4216 4217 return (0); 4218 } 4219 4220 int 4221 if_getcapabilities(const if_t ifp) 4222 { 4223 return ((struct ifnet *)ifp)->if_capabilities; 4224 } 4225 4226 int 4227 if_setcapenable(if_t ifp, int capabilities) 4228 { 4229 ((struct ifnet *)ifp)->if_capenable = capabilities; 4230 return (0); 4231 } 4232 4233 int 4234 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4235 { 4236 if(setcap) 4237 ((struct ifnet *)ifp)->if_capenable |= setcap; 4238 if(clearcap) 4239 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 4240 4241 return (0); 4242 } 4243 4244 const char * 4245 if_getdname(const if_t ifp) 4246 { 4247 return ((struct ifnet *)ifp)->if_dname; 4248 } 4249 4250 void 4251 if_setdname(if_t ifp, const char *dname) 4252 { 4253 ((struct ifnet *)ifp)->if_dname = dname; 4254 } 4255 4256 const char * 4257 if_name(if_t ifp) 4258 { 4259 return ((struct ifnet *)ifp)->if_xname; 4260 } 4261 4262 int 4263 if_setname(if_t ifp, const char *name) 4264 { 4265 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4266 return (ENAMETOOLONG); 4267 strlcpy(ifp->if_xname, name, sizeof(ifp->if_xname)); 4268 4269 return (0); 4270 } 4271 4272 int 4273 if_togglecapenable(if_t ifp, int togglecap) 4274 { 4275 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 4276 return (0); 4277 } 4278 4279 int 4280 if_getcapenable(const if_t ifp) 4281 { 4282 return ((struct ifnet *)ifp)->if_capenable; 4283 } 4284 4285 int 4286 if_getdunit(const if_t ifp) 4287 { 4288 return ((struct ifnet *)ifp)->if_dunit; 4289 } 4290 4291 int 4292 if_getindex(const if_t ifp) 4293 { 4294 return ((struct ifnet *)ifp)->if_index; 4295 } 4296 4297 void 4298 if_setdescr(if_t ifp, char *descrbuf) 4299 { 4300 sx_xlock(&ifdescr_sx); 4301 char *odescrbuf = ifp->if_description; 4302 ifp->if_description = descrbuf; 4303 sx_xunlock(&ifdescr_sx); 4304 4305 if_freedescr(odescrbuf); 4306 } 4307 4308 char * 4309 if_allocdescr(size_t sz, int malloc_flag) 4310 { 4311 malloc_flag &= (M_WAITOK | M_NOWAIT); 4312 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4313 } 4314 4315 void 4316 if_freedescr(char *descrbuf) 4317 { 4318 free(descrbuf, M_IFDESCR); 4319 } 4320 4321 int 4322 if_getalloctype(const if_t ifp) 4323 { 4324 return ((struct ifnet *)ifp)->if_alloctype; 4325 } 4326 4327 /* 4328 * This is largely undesirable because it ties ifnet to a device, but does 4329 * provide flexiblity for an embedded product vendor. Should be used with 4330 * the understanding that it violates the interface boundaries, and should be 4331 * a last resort only. 4332 */ 4333 int 4334 if_setdev(if_t ifp, void *dev) 4335 { 4336 return (0); 4337 } 4338 4339 int 4340 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4341 { 4342 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 4343 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 4344 4345 return (0); 4346 } 4347 4348 int 4349 if_getdrvflags(const if_t ifp) 4350 { 4351 return ((struct ifnet *)ifp)->if_drv_flags; 4352 } 4353 4354 int 4355 if_setdrvflags(if_t ifp, int flags) 4356 { 4357 ((struct ifnet *)ifp)->if_drv_flags = flags; 4358 return (0); 4359 } 4360 4361 int 4362 if_setflags(if_t ifp, int flags) 4363 { 4364 4365 ifp->if_flags = flags; 4366 return (0); 4367 } 4368 4369 int 4370 if_setflagbits(if_t ifp, int set, int clear) 4371 { 4372 ((struct ifnet *)ifp)->if_flags |= set; 4373 ((struct ifnet *)ifp)->if_flags &= ~clear; 4374 4375 return (0); 4376 } 4377 4378 int 4379 if_getflags(const if_t ifp) 4380 { 4381 return ((struct ifnet *)ifp)->if_flags; 4382 } 4383 4384 int 4385 if_clearhwassist(if_t ifp) 4386 { 4387 ((struct ifnet *)ifp)->if_hwassist = 0; 4388 return (0); 4389 } 4390 4391 int 4392 if_sethwassistbits(if_t ifp, int toset, int toclear) 4393 { 4394 ((struct ifnet *)ifp)->if_hwassist |= toset; 4395 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4396 4397 return (0); 4398 } 4399 4400 int 4401 if_sethwassist(if_t ifp, int hwassist_bit) 4402 { 4403 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4404 return (0); 4405 } 4406 4407 int 4408 if_gethwassist(const if_t ifp) 4409 { 4410 return ((struct ifnet *)ifp)->if_hwassist; 4411 } 4412 4413 int 4414 if_togglehwassist(if_t ifp, int toggle_bits) 4415 { 4416 ((struct ifnet *)ifp)->if_hwassist ^= toggle_bits; 4417 return (0); 4418 } 4419 4420 int 4421 if_setmtu(if_t ifp, int mtu) 4422 { 4423 ((struct ifnet *)ifp)->if_mtu = mtu; 4424 return (0); 4425 } 4426 4427 int 4428 if_getmtu(const if_t ifp) 4429 { 4430 return ((struct ifnet *)ifp)->if_mtu; 4431 } 4432 4433 int 4434 if_getmtu_family(const if_t ifp, int family) 4435 { 4436 struct domain *dp; 4437 4438 SLIST_FOREACH(dp, &domains, dom_next) { 4439 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4440 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4441 } 4442 4443 return (((struct ifnet *)ifp)->if_mtu); 4444 } 4445 4446 /* 4447 * Methods for drivers to access interface unicast and multicast 4448 * link level addresses. Driver shall not know 'struct ifaddr' neither 4449 * 'struct ifmultiaddr'. 4450 */ 4451 u_int 4452 if_lladdr_count(if_t ifp) 4453 { 4454 struct epoch_tracker et; 4455 struct ifaddr *ifa; 4456 u_int count; 4457 4458 count = 0; 4459 NET_EPOCH_ENTER(et); 4460 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4461 if (ifa->ifa_addr->sa_family == AF_LINK) 4462 count++; 4463 NET_EPOCH_EXIT(et); 4464 4465 return (count); 4466 } 4467 4468 u_int 4469 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4470 { 4471 struct epoch_tracker et; 4472 struct ifaddr *ifa; 4473 u_int count; 4474 4475 MPASS(cb); 4476 4477 count = 0; 4478 NET_EPOCH_ENTER(et); 4479 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4480 if (ifa->ifa_addr->sa_family != AF_LINK) 4481 continue; 4482 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4483 count); 4484 } 4485 NET_EPOCH_EXIT(et); 4486 4487 return (count); 4488 } 4489 4490 u_int 4491 if_llmaddr_count(if_t ifp) 4492 { 4493 struct epoch_tracker et; 4494 struct ifmultiaddr *ifma; 4495 int count; 4496 4497 count = 0; 4498 NET_EPOCH_ENTER(et); 4499 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4500 if (ifma->ifma_addr->sa_family == AF_LINK) 4501 count++; 4502 NET_EPOCH_EXIT(et); 4503 4504 return (count); 4505 } 4506 4507 u_int 4508 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4509 { 4510 struct epoch_tracker et; 4511 struct ifmultiaddr *ifma; 4512 u_int count; 4513 4514 MPASS(cb); 4515 4516 count = 0; 4517 NET_EPOCH_ENTER(et); 4518 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4519 if (ifma->ifma_addr->sa_family != AF_LINK) 4520 continue; 4521 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4522 count); 4523 } 4524 NET_EPOCH_EXIT(et); 4525 4526 return (count); 4527 } 4528 4529 u_int 4530 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4531 { 4532 struct epoch_tracker et; 4533 struct ifaddr *ifa; 4534 u_int count; 4535 4536 MPASS(cb); 4537 4538 count = 0; 4539 NET_EPOCH_ENTER(et); 4540 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4541 if (ifa->ifa_addr->sa_family != type) 4542 continue; 4543 count += (*cb)(cb_arg, ifa, count); 4544 } 4545 NET_EPOCH_EXIT(et); 4546 4547 return (count); 4548 } 4549 4550 int 4551 if_setsoftc(if_t ifp, void *softc) 4552 { 4553 ((struct ifnet *)ifp)->if_softc = softc; 4554 return (0); 4555 } 4556 4557 void * 4558 if_getsoftc(const if_t ifp) 4559 { 4560 return ((struct ifnet *)ifp)->if_softc; 4561 } 4562 4563 void 4564 if_setrcvif(struct mbuf *m, if_t ifp) 4565 { 4566 4567 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4568 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4569 } 4570 4571 void 4572 if_setvtag(struct mbuf *m, uint16_t tag) 4573 { 4574 m->m_pkthdr.ether_vtag = tag; 4575 } 4576 4577 uint16_t 4578 if_getvtag(struct mbuf *m) 4579 { 4580 4581 return (m->m_pkthdr.ether_vtag); 4582 } 4583 4584 int 4585 if_sendq_empty(if_t ifp) 4586 { 4587 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4588 } 4589 4590 struct ifaddr * 4591 if_getifaddr(const if_t ifp) 4592 { 4593 return ((struct ifnet *)ifp)->if_addr; 4594 } 4595 4596 int 4597 if_getamcount(const if_t ifp) 4598 { 4599 return ((struct ifnet *)ifp)->if_amcount; 4600 } 4601 4602 int 4603 if_setsendqready(if_t ifp) 4604 { 4605 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4606 return (0); 4607 } 4608 4609 int 4610 if_setsendqlen(if_t ifp, int tx_desc_count) 4611 { 4612 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4613 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4614 4615 return (0); 4616 } 4617 4618 int 4619 if_vlantrunkinuse(if_t ifp) 4620 { 4621 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4622 } 4623 4624 int 4625 if_init(if_t ifp) 4626 { 4627 (*((struct ifnet *)ifp)->if_init)((struct ifnet *)ifp); 4628 return (0); 4629 } 4630 4631 int 4632 if_input(if_t ifp, struct mbuf* sendmp) 4633 { 4634 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4635 return (0); 4636 4637 } 4638 4639 int 4640 if_transmit(if_t ifp, struct mbuf *m) 4641 { 4642 (*((struct ifnet *)ifp)->if_transmit)((struct ifnet *)ifp, m); 4643 return (0); 4644 } 4645 4646 struct mbuf * 4647 if_dequeue(if_t ifp) 4648 { 4649 struct mbuf *m; 4650 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4651 4652 return (m); 4653 } 4654 4655 int 4656 if_sendq_prepend(if_t ifp, struct mbuf *m) 4657 { 4658 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4659 return (0); 4660 } 4661 4662 int 4663 if_setifheaderlen(if_t ifp, int len) 4664 { 4665 ((struct ifnet *)ifp)->if_hdrlen = len; 4666 return (0); 4667 } 4668 4669 caddr_t 4670 if_getlladdr(const if_t ifp) 4671 { 4672 return (IF_LLADDR((struct ifnet *)ifp)); 4673 } 4674 4675 void * 4676 if_gethandle(u_char type) 4677 { 4678 return (if_alloc(type)); 4679 } 4680 4681 void 4682 if_bpfmtap(if_t ifh, struct mbuf *m) 4683 { 4684 struct ifnet *ifp = (struct ifnet *)ifh; 4685 4686 BPF_MTAP(ifp, m); 4687 } 4688 4689 void 4690 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4691 { 4692 struct ifnet *ifp = (struct ifnet *)ifh; 4693 4694 ETHER_BPF_MTAP(ifp, m); 4695 } 4696 4697 void 4698 if_vlancap(if_t ifh) 4699 { 4700 struct ifnet *ifp = (struct ifnet *)ifh; 4701 VLAN_CAPABILITIES(ifp); 4702 } 4703 4704 int 4705 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4706 { 4707 4708 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax; 4709 return (0); 4710 } 4711 4712 int 4713 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4714 { 4715 4716 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4717 return (0); 4718 } 4719 4720 int 4721 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4722 { 4723 4724 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4725 return (0); 4726 } 4727 4728 u_int 4729 if_gethwtsomax(const if_t ifp) 4730 { 4731 4732 return (((struct ifnet *)ifp)->if_hw_tsomax); 4733 } 4734 4735 u_int 4736 if_gethwtsomaxsegcount(const if_t ifp) 4737 { 4738 4739 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount); 4740 } 4741 4742 u_int 4743 if_gethwtsomaxsegsize(const if_t ifp) 4744 { 4745 4746 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize); 4747 } 4748 4749 void 4750 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4751 { 4752 ((struct ifnet *)ifp)->if_init = init_fn; 4753 } 4754 4755 void 4756 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4757 { 4758 ((struct ifnet *)ifp)->if_input = input_fn; 4759 } 4760 4761 void 4762 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4763 { 4764 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4765 } 4766 4767 void 4768 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4769 { 4770 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4771 } 4772 4773 void 4774 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4775 { 4776 ((struct ifnet *)ifp)->if_transmit = start_fn; 4777 } 4778 4779 void 4780 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4781 { 4782 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4783 4784 } 4785 4786 void 4787 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 4788 { 4789 ((struct ifnet *)ifp)->if_snd_tag_alloc = alloc_fn; 4790 } 4791 4792 void 4793 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4794 { 4795 4796 ifp->if_get_counter = fn; 4797 } 4798 4799 #ifdef DDB 4800 static void 4801 if_show_ifnet(struct ifnet *ifp) 4802 { 4803 4804 if (ifp == NULL) 4805 return; 4806 db_printf("%s:\n", ifp->if_xname); 4807 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 4808 IF_DB_PRINTF("%s", if_dname); 4809 IF_DB_PRINTF("%d", if_dunit); 4810 IF_DB_PRINTF("%s", if_description); 4811 IF_DB_PRINTF("%u", if_index); 4812 IF_DB_PRINTF("%d", if_idxgen); 4813 IF_DB_PRINTF("%u", if_refcount); 4814 IF_DB_PRINTF("%p", if_softc); 4815 IF_DB_PRINTF("%p", if_l2com); 4816 IF_DB_PRINTF("%p", if_llsoftc); 4817 IF_DB_PRINTF("%d", if_amcount); 4818 IF_DB_PRINTF("%p", if_addr); 4819 IF_DB_PRINTF("%p", if_broadcastaddr); 4820 IF_DB_PRINTF("%p", if_afdata); 4821 IF_DB_PRINTF("%d", if_afdata_initialized); 4822 IF_DB_PRINTF("%u", if_fib); 4823 IF_DB_PRINTF("%p", if_vnet); 4824 IF_DB_PRINTF("%p", if_home_vnet); 4825 IF_DB_PRINTF("%p", if_vlantrunk); 4826 IF_DB_PRINTF("%p", if_bpf); 4827 IF_DB_PRINTF("%u", if_pcount); 4828 IF_DB_PRINTF("%p", if_bridge); 4829 IF_DB_PRINTF("%p", if_lagg); 4830 IF_DB_PRINTF("%p", if_pf_kif); 4831 IF_DB_PRINTF("%p", if_carp); 4832 IF_DB_PRINTF("%p", if_label); 4833 IF_DB_PRINTF("%p", if_netmap); 4834 IF_DB_PRINTF("0x%08x", if_flags); 4835 IF_DB_PRINTF("0x%08x", if_drv_flags); 4836 IF_DB_PRINTF("0x%08x", if_capabilities); 4837 IF_DB_PRINTF("0x%08x", if_capenable); 4838 IF_DB_PRINTF("%p", if_snd.ifq_head); 4839 IF_DB_PRINTF("%p", if_snd.ifq_tail); 4840 IF_DB_PRINTF("%d", if_snd.ifq_len); 4841 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 4842 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 4843 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 4844 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 4845 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 4846 IF_DB_PRINTF("%d", if_snd.altq_type); 4847 IF_DB_PRINTF("%x", if_snd.altq_flags); 4848 #undef IF_DB_PRINTF 4849 } 4850 4851 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 4852 { 4853 4854 if (!have_addr) { 4855 db_printf("usage: show ifnet <struct ifnet *>\n"); 4856 return; 4857 } 4858 4859 if_show_ifnet((struct ifnet *)addr); 4860 } 4861 4862 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 4863 { 4864 struct ifnet *ifp; 4865 u_short idx; 4866 4867 for (idx = 1; idx <= if_index; idx++) { 4868 ifp = ifindex_table[idx].ife_ifnet; 4869 if (ifp == NULL) 4870 continue; 4871 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 4872 if (db_pager_quit) 4873 break; 4874 } 4875 } 4876 #endif /* DDB */ 4877