1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org> 5 * Copyright (c) 1980, 1986, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 #include "opt_bpf.h" 34 #include "opt_inet6.h" 35 #include "opt_inet.h" 36 #include "opt_ddb.h" 37 38 #include <sys/param.h> 39 #include <sys/capsicum.h> 40 #include <sys/conf.h> 41 #include <sys/eventhandler.h> 42 #include <sys/malloc.h> 43 #include <sys/domainset.h> 44 #include <sys/sbuf.h> 45 #include <sys/bus.h> 46 #include <sys/epoch.h> 47 #include <sys/mbuf.h> 48 #include <sys/systm.h> 49 #include <sys/priv.h> 50 #include <sys/proc.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/protosw.h> 54 #include <sys/kernel.h> 55 #include <sys/lock.h> 56 #include <sys/refcount.h> 57 #include <sys/module.h> 58 #include <sys/nv.h> 59 #include <sys/rwlock.h> 60 #include <sys/sockio.h> 61 #include <sys/stdarg.h> 62 #include <sys/syslog.h> 63 #include <sys/sysctl.h> 64 #include <sys/sysent.h> 65 #include <sys/taskqueue.h> 66 #include <sys/domain.h> 67 #include <sys/jail.h> 68 #include <sys/priv.h> 69 70 #ifdef DDB 71 #include <ddb/ddb.h> 72 #endif 73 74 #include <vm/uma.h> 75 76 #include <net/bpf.h> 77 #include <net/if.h> 78 #include <net/if_arp.h> 79 #include <net/if_clone.h> 80 #include <net/if_dl.h> 81 #include <net/if_strings.h> 82 #include <net/if_types.h> 83 #include <net/if_var.h> 84 #include <net/if_media.h> 85 #include <net/if_mib.h> 86 #include <net/if_private.h> 87 #include <net/if_vlan_var.h> 88 #include <net/radix.h> 89 #include <net/route.h> 90 #include <net/route/route_ctl.h> 91 #include <net/vnet.h> 92 93 #if defined(INET) || defined(INET6) 94 #include <net/ethernet.h> 95 #include <netinet/in.h> 96 #include <netinet/in_var.h> 97 #include <netinet/ip.h> 98 #include <netinet/ip_carp.h> 99 #ifdef INET 100 #include <net/debugnet.h> 101 #include <netinet/if_ether.h> 102 #endif /* INET */ 103 #ifdef INET6 104 #include <netinet6/in6_var.h> 105 #include <netinet6/in6_ifattach.h> 106 #endif /* INET6 */ 107 #endif /* INET || INET6 */ 108 109 #include <security/mac/mac_framework.h> 110 111 /* 112 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 113 * and ifr_ifru when it is used in SIOCGIFCONF. 114 */ 115 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 116 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 117 118 __read_mostly epoch_t net_epoch_preempt; 119 #ifdef COMPAT_FREEBSD32 120 #include <sys/mount.h> 121 #include <compat/freebsd32/freebsd32.h> 122 123 struct ifreq_buffer32 { 124 uint32_t length; /* (size_t) */ 125 uint32_t buffer; /* (void *) */ 126 }; 127 128 /* 129 * Interface request structure used for socket 130 * ioctl's. All interface ioctl's must have parameter 131 * definitions which begin with ifr_name. The 132 * remainder may be interface specific. 133 */ 134 struct ifreq32 { 135 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 136 union { 137 struct sockaddr ifru_addr; 138 struct sockaddr ifru_dstaddr; 139 struct sockaddr ifru_broadaddr; 140 struct ifreq_buffer32 ifru_buffer; 141 short ifru_flags[2]; 142 short ifru_index; 143 int ifru_jid; 144 int ifru_metric; 145 int ifru_mtu; 146 int ifru_phys; 147 int ifru_media; 148 uint32_t ifru_data; 149 int ifru_cap[2]; 150 u_int ifru_fib; 151 u_char ifru_vlan_pcp; 152 } ifr_ifru; 153 }; 154 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 155 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 156 __offsetof(struct ifreq32, ifr_ifru)); 157 158 struct ifconf32 { 159 int32_t ifc_len; 160 union { 161 uint32_t ifcu_buf; 162 uint32_t ifcu_req; 163 } ifc_ifcu; 164 }; 165 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 166 167 struct ifdrv32 { 168 char ifd_name[IFNAMSIZ]; 169 uint32_t ifd_cmd; 170 uint32_t ifd_len; 171 uint32_t ifd_data; 172 }; 173 #define SIOCSDRVSPEC32 _IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32) 174 #define SIOCGDRVSPEC32 _IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32) 175 176 struct ifgroupreq32 { 177 char ifgr_name[IFNAMSIZ]; 178 u_int ifgr_len; 179 union { 180 char ifgru_group[IFNAMSIZ]; 181 uint32_t ifgru_groups; 182 } ifgr_ifgru; 183 }; 184 #define SIOCAIFGROUP32 _IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32) 185 #define SIOCGIFGROUP32 _IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32) 186 #define SIOCDIFGROUP32 _IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32) 187 #define SIOCGIFGMEMB32 _IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32) 188 189 struct ifmediareq32 { 190 char ifm_name[IFNAMSIZ]; 191 int ifm_current; 192 int ifm_mask; 193 int ifm_status; 194 int ifm_active; 195 int ifm_count; 196 uint32_t ifm_ulist; /* (int *) */ 197 }; 198 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 199 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 200 #endif /* COMPAT_FREEBSD32 */ 201 202 union ifreq_union { 203 struct ifreq ifr; 204 #ifdef COMPAT_FREEBSD32 205 struct ifreq32 ifr32; 206 #endif 207 }; 208 209 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 210 "Link layers"); 211 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 212 "Generic link-management"); 213 214 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 215 &ifqmaxlen, 0, "max send queue size"); 216 217 /* Log link state change events */ 218 static int log_link_state_change = 1; 219 220 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 221 &log_link_state_change, 0, 222 "log interface link state change events"); 223 224 /* Log promiscuous mode change events */ 225 static int log_promisc_mode_change = 1; 226 227 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 228 &log_promisc_mode_change, 1, 229 "log promiscuous mode change events"); 230 231 /* Interface description */ 232 static unsigned int ifdescr_maxlen = 1024; 233 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 234 &ifdescr_maxlen, 0, 235 "administrative maximum length for interface description"); 236 237 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 238 239 /* global sx for non-critical path ifdescr */ 240 static struct sx ifdescr_sx; 241 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 242 243 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 244 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 245 /* These are external hooks for CARP. */ 246 void (*carp_linkstate_p)(struct ifnet *ifp); 247 void (*carp_demote_adj_p)(int, char *); 248 int (*carp_master_p)(struct ifaddr *); 249 #if defined(INET) || defined(INET6) 250 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 251 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 252 const struct sockaddr *sa); 253 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 254 int (*carp_attach_p)(struct ifaddr *, int); 255 void (*carp_detach_p)(struct ifaddr *, bool); 256 #endif 257 #ifdef INET 258 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 259 #endif 260 #ifdef INET6 261 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 262 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 263 const struct in6_addr *taddr); 264 #endif 265 266 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 267 268 /* 269 * XXX: Style; these should be sorted alphabetically, and unprototyped 270 * static functions should be prototyped. Currently they are sorted by 271 * declaration order. 272 */ 273 static void if_attachdomain(void *); 274 static void if_attachdomain1(struct ifnet *); 275 static int ifconf(u_long, caddr_t); 276 static void if_input_default(struct ifnet *, struct mbuf *); 277 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 278 static int if_setflag(struct ifnet *, int, int, int *, int); 279 static int if_transmit_default(struct ifnet *ifp, struct mbuf *m); 280 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 281 static void do_link_state_change(void *, int); 282 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 283 static int if_getgroupmembers(struct ifgroupreq *); 284 static void if_delgroups(struct ifnet *); 285 static void if_attach_internal(struct ifnet *, bool); 286 static void if_detach_internal(struct ifnet *, bool); 287 static void if_siocaddmulti(void *, int); 288 static void if_link_ifnet(struct ifnet *); 289 static bool if_unlink_ifnet(struct ifnet *, bool); 290 #ifdef VIMAGE 291 static void if_vmove(struct ifnet *, struct vnet *); 292 #endif 293 294 #ifdef INET6 295 /* 296 * XXX: declare here to avoid to include many inet6 related files.. 297 * should be more generalized? 298 */ 299 extern void nd6_setmtu(struct ifnet *); 300 #endif 301 302 /* ipsec helper hooks */ 303 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 304 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 305 306 int ifqmaxlen = IFQ_MAXLEN; 307 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 308 VNET_DEFINE(struct ifgrouphead, ifg_head); 309 310 /* Table of ifnet by index. */ 311 static int if_index; 312 static int if_indexlim = 8; 313 static struct ifindex_entry { 314 struct ifnet *ife_ifnet; 315 uint16_t ife_gencnt; 316 } *ifindex_table; 317 318 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system, 319 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 320 "Variables global to all interfaces"); 321 static int 322 sysctl_ifcount(SYSCTL_HANDLER_ARGS) 323 { 324 int rv = 0; 325 326 IFNET_RLOCK(); 327 for (int i = 1; i <= if_index; i++) 328 if (ifindex_table[i].ife_ifnet != NULL && 329 ifindex_table[i].ife_ifnet->if_vnet == curvnet) 330 rv = i; 331 IFNET_RUNLOCK(); 332 333 return (sysctl_handle_int(oidp, &rv, 0, req)); 334 } 335 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount, 336 CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I", 337 "Maximum known interface index"); 338 339 /* 340 * The global network interface list (V_ifnet) and related state (such as 341 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 342 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 343 */ 344 struct sx ifnet_sxlock; 345 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 346 347 struct sx ifnet_detach_sxlock; 348 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 349 SX_RECURSE); 350 351 #ifdef VIMAGE 352 #define VNET_IS_SHUTTING_DOWN(_vnet) \ 353 ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE) 354 #endif 355 356 static if_com_alloc_t *if_com_alloc[256]; 357 static if_com_free_t *if_com_free[256]; 358 359 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 360 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 361 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 362 363 struct ifnet * 364 ifnet_byindex(u_int idx) 365 { 366 struct ifnet *ifp; 367 368 NET_EPOCH_ASSERT(); 369 370 if (__predict_false(idx > if_index)) 371 return (NULL); 372 373 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 374 375 if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet) 376 ifp = NULL; 377 378 return (ifp); 379 } 380 381 struct ifnet * 382 ifnet_byindex_ref(u_int idx) 383 { 384 struct ifnet *ifp; 385 386 ifp = ifnet_byindex(idx); 387 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 388 return (NULL); 389 if (!if_try_ref(ifp)) 390 return (NULL); 391 return (ifp); 392 } 393 394 struct ifnet * 395 ifnet_byindexgen(uint16_t idx, uint16_t gen) 396 { 397 struct ifnet *ifp; 398 399 NET_EPOCH_ASSERT(); 400 401 if (__predict_false(idx > if_index)) 402 return (NULL); 403 404 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 405 406 if (ifindex_table[idx].ife_gencnt == gen) 407 return (ifp); 408 else 409 return (NULL); 410 } 411 412 /* 413 * Network interface utility routines. 414 * 415 * Routines with ifa_ifwith* names take sockaddr *'s as 416 * parameters. 417 */ 418 419 static void 420 if_init_idxtable(void *arg __unused) 421 { 422 423 ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table), 424 M_IFNET, M_WAITOK | M_ZERO); 425 } 426 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL); 427 428 static void 429 vnet_if_init(const void *unused __unused) 430 { 431 432 CK_STAILQ_INIT(&V_ifnet); 433 CK_STAILQ_INIT(&V_ifg_head); 434 } 435 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 436 NULL); 437 438 static void 439 if_link_ifnet(struct ifnet *ifp) 440 { 441 442 IFNET_WLOCK(); 443 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 444 #ifdef VIMAGE 445 curvnet->vnet_ifcnt++; 446 #endif 447 IFNET_WUNLOCK(); 448 } 449 450 static bool 451 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 452 { 453 struct ifnet *iter; 454 int found = 0; 455 456 IFNET_WLOCK(); 457 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 458 if (iter == ifp) { 459 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 460 if (!vmove) 461 ifp->if_flags |= IFF_DYING; 462 found = 1; 463 break; 464 } 465 #ifdef VIMAGE 466 curvnet->vnet_ifcnt--; 467 #endif 468 IFNET_WUNLOCK(); 469 470 return (found); 471 } 472 473 #ifdef VIMAGE 474 static void 475 vnet_if_return(const void *unused __unused) 476 { 477 struct ifnet *ifp, *nifp; 478 struct ifnet **pending; 479 int found __diagused; 480 int i; 481 482 i = 0; 483 484 /* 485 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 486 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 487 * if_detach_internal(), which waits for NET_EPOCH callbacks to 488 * complete. We can't do that from within NET_EPOCH. 489 * 490 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 491 * read/write lock. We cannot hold the lock as we call if_vmove() 492 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 493 * ctx lock. 494 */ 495 IFNET_WLOCK(); 496 497 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 498 M_IFNET, M_WAITOK | M_ZERO); 499 500 /* Return all inherited interfaces to their parent vnets. */ 501 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 502 if (ifp->if_home_vnet != ifp->if_vnet) { 503 found = if_unlink_ifnet(ifp, true); 504 MPASS(found); 505 506 pending[i++] = ifp; 507 } 508 } 509 IFNET_WUNLOCK(); 510 511 for (int j = 0; j < i; j++) { 512 sx_xlock(&ifnet_detach_sxlock); 513 if_vmove(pending[j], pending[j]->if_home_vnet); 514 sx_xunlock(&ifnet_detach_sxlock); 515 } 516 517 free(pending, M_IFNET); 518 } 519 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 520 vnet_if_return, NULL); 521 #endif 522 523 /* 524 * Allocate a struct ifnet and an index for an interface. A layer 2 525 * common structure will also be allocated if an allocation routine is 526 * registered for the passed type. 527 */ 528 static struct ifnet * 529 if_alloc_domain(u_char type, int numa_domain) 530 { 531 struct ifnet *ifp; 532 u_short idx; 533 534 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 535 if (numa_domain == IF_NODOM) 536 ifp = malloc(sizeof(struct ifnet), M_IFNET, 537 M_WAITOK | M_ZERO); 538 else 539 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 540 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 541 ifp->if_type = type; 542 ifp->if_alloctype = type; 543 ifp->if_numa_domain = numa_domain; 544 #ifdef VIMAGE 545 ifp->if_vnet = curvnet; 546 #endif 547 if (if_com_alloc[type] != NULL) { 548 ifp->if_l2com = if_com_alloc[type](type, ifp); 549 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 550 type)); 551 } 552 553 IF_ADDR_LOCK_INIT(ifp); 554 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 555 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 556 ifp->if_afdata_initialized = 0; 557 IF_AFDATA_LOCK_INIT(ifp); 558 CK_STAILQ_INIT(&ifp->if_addrhead); 559 CK_STAILQ_INIT(&ifp->if_multiaddrs); 560 CK_STAILQ_INIT(&ifp->if_groups); 561 #ifdef MAC 562 mac_ifnet_init(ifp); 563 #endif 564 ifq_init(&ifp->if_snd, ifp); 565 566 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 567 for (int i = 0; i < IFCOUNTERS; i++) 568 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 569 ifp->if_get_counter = if_get_counter_default; 570 ifp->if_pcp = IFNET_PCP_NONE; 571 572 /* Allocate an ifindex array entry. */ 573 IFNET_WLOCK(); 574 /* 575 * Try to find an empty slot below if_index. If we fail, take the 576 * next slot. 577 */ 578 for (idx = 1; idx <= if_index; idx++) { 579 if (ifindex_table[idx].ife_ifnet == NULL) 580 break; 581 } 582 583 /* Catch if_index overflow. */ 584 if (idx >= if_indexlim) { 585 struct ifindex_entry *new, *old; 586 int newlim; 587 588 newlim = if_indexlim * 2; 589 new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO); 590 memcpy(new, ifindex_table, if_indexlim * sizeof(*new)); 591 old = ifindex_table; 592 ck_pr_store_ptr(&ifindex_table, new); 593 if_indexlim = newlim; 594 NET_EPOCH_WAIT(); 595 free(old, M_IFNET); 596 } 597 if (idx > if_index) 598 if_index = idx; 599 600 ifp->if_index = idx; 601 ifp->if_idxgen = ifindex_table[idx].ife_gencnt; 602 ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp); 603 IFNET_WUNLOCK(); 604 605 return (ifp); 606 } 607 608 struct ifnet * 609 if_alloc_dev(u_char type, device_t dev) 610 { 611 int numa_domain; 612 613 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 614 return (if_alloc_domain(type, IF_NODOM)); 615 return (if_alloc_domain(type, numa_domain)); 616 } 617 618 struct ifnet * 619 if_alloc(u_char type) 620 { 621 622 return (if_alloc_domain(type, IF_NODOM)); 623 } 624 /* 625 * Do the actual work of freeing a struct ifnet, and layer 2 common 626 * structure. This call is made when the network epoch guarantees 627 * us that nobody holds a pointer to the interface. 628 */ 629 static void 630 if_free_deferred(epoch_context_t ctx) 631 { 632 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 633 634 KASSERT((ifp->if_flags & IFF_DYING), 635 ("%s: interface not dying", __func__)); 636 637 if (if_com_free[ifp->if_alloctype] != NULL) 638 if_com_free[ifp->if_alloctype](ifp->if_l2com, 639 ifp->if_alloctype); 640 641 #ifdef MAC 642 mac_ifnet_destroy(ifp); 643 #endif /* MAC */ 644 IF_AFDATA_DESTROY(ifp); 645 IF_ADDR_LOCK_DESTROY(ifp); 646 ifq_delete(&ifp->if_snd); 647 648 for (int i = 0; i < IFCOUNTERS; i++) 649 counter_u64_free(ifp->if_counters[i]); 650 651 if_freedescr(ifp->if_description); 652 free(ifp->if_hw_addr, M_IFADDR); 653 free(ifp, M_IFNET); 654 } 655 656 /* 657 * Deregister an interface and free the associated storage. 658 */ 659 void 660 if_free(struct ifnet *ifp) 661 { 662 663 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 664 665 /* 666 * XXXGL: An interface index is really an alias to ifp pointer. 667 * Why would we clear the alias now, and not in the deferred 668 * context? Indeed there is nothing wrong with some network 669 * thread obtaining ifp via ifnet_byindex() inside the network 670 * epoch and then dereferencing ifp while we perform if_free(), 671 * and after if_free() finished, too. 672 * 673 * This early index freeing was important back when ifindex was 674 * virtualized and interface would outlive the vnet. 675 */ 676 IFNET_WLOCK(); 677 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 678 ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL); 679 ifindex_table[ifp->if_index].ife_gencnt++; 680 while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL) 681 if_index--; 682 IFNET_WUNLOCK(); 683 684 if (refcount_release(&ifp->if_refcount)) 685 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 686 } 687 688 /* 689 * Interfaces to keep an ifnet type-stable despite the possibility of the 690 * driver calling if_free(). If there are additional references, we defer 691 * freeing the underlying data structure. 692 */ 693 void 694 if_ref(struct ifnet *ifp) 695 { 696 u_int old __diagused; 697 698 /* We don't assert the ifnet list lock here, but arguably should. */ 699 old = refcount_acquire(&ifp->if_refcount); 700 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 701 } 702 703 bool 704 if_try_ref(struct ifnet *ifp) 705 { 706 NET_EPOCH_ASSERT(); 707 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 708 } 709 710 void 711 if_rele(struct ifnet *ifp) 712 { 713 714 if (!refcount_release(&ifp->if_refcount)) 715 return; 716 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 717 } 718 719 void 720 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 721 { 722 723 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 724 725 if (ifq->ifq_maxlen == 0) 726 ifq->ifq_maxlen = ifqmaxlen; 727 728 ifq->altq_type = 0; 729 ifq->altq_disc = NULL; 730 ifq->altq_flags &= ALTQF_CANTCHANGE; 731 ifq->altq_tbr = NULL; 732 ifq->altq_ifp = ifp; 733 } 734 735 void 736 ifq_delete(struct ifaltq *ifq) 737 { 738 mtx_destroy(&ifq->ifq_mtx); 739 } 740 741 /* 742 * Perform generic interface initialization tasks and attach the interface 743 * to the list of "active" interfaces. If vmove flag is set on entry 744 * to if_attach_internal(), perform only a limited subset of initialization 745 * tasks, given that we are moving from one vnet to another an ifnet which 746 * has already been fully initialized. 747 * 748 * Note that if_detach_internal() removes group membership unconditionally 749 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 750 * Thus, when if_vmove() is applied to a cloned interface, group membership 751 * is lost while a cloned one always joins a group whose name is 752 * ifc->ifc_name. To recover this after if_detach_internal() and 753 * if_attach_internal(), the cloner should be specified to 754 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 755 * attempts to join a group whose name is ifc->ifc_name. 756 * 757 * XXX: 758 * - The decision to return void and thus require this function to 759 * succeed is questionable. 760 * - We should probably do more sanity checking. For instance we don't 761 * do anything to insure if_xname is unique or non-empty. 762 */ 763 void 764 if_attach(struct ifnet *ifp) 765 { 766 767 if_attach_internal(ifp, false); 768 } 769 770 /* 771 * Compute the least common TSO limit. 772 */ 773 void 774 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 775 { 776 /* 777 * 1) If there is no limit currently, take the limit from 778 * the network adapter. 779 * 780 * 2) If the network adapter has a limit below the current 781 * limit, apply it. 782 */ 783 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 784 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 785 pmax->tsomaxbytes = ifp->if_hw_tsomax; 786 } 787 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 788 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 789 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 790 } 791 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 792 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 793 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 794 } 795 } 796 797 /* 798 * Update TSO limit of a network adapter. 799 * 800 * Returns zero if no change. Else non-zero. 801 */ 802 int 803 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 804 { 805 int retval = 0; 806 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 807 ifp->if_hw_tsomax = pmax->tsomaxbytes; 808 retval++; 809 } 810 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 811 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 812 retval++; 813 } 814 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 815 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 816 retval++; 817 } 818 return (retval); 819 } 820 821 static void 822 if_attach_internal(struct ifnet *ifp, bool vmove) 823 { 824 unsigned socksize, ifasize; 825 int namelen, masklen; 826 struct sockaddr_dl *sdl; 827 struct ifaddr *ifa; 828 829 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 830 831 #ifdef VIMAGE 832 CURVNET_ASSERT_SET(); 833 ifp->if_vnet = curvnet; 834 if (ifp->if_home_vnet == NULL) 835 ifp->if_home_vnet = curvnet; 836 #endif 837 838 if_addgroup(ifp, IFG_ALL); 839 840 #ifdef VIMAGE 841 /* Restore group membership for cloned interface. */ 842 if (vmove) 843 if_clone_restoregroup(ifp); 844 #endif 845 846 getmicrotime(&ifp->if_lastchange); 847 ifp->if_epoch = time_uptime; 848 849 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 850 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 851 ("transmit and qflush must both either be set or both be NULL")); 852 if (ifp->if_transmit == NULL) { 853 ifp->if_transmit = if_transmit_default; 854 ifp->if_qflush = if_qflush; 855 } 856 if (ifp->if_input == NULL) 857 ifp->if_input = if_input_default; 858 859 if (ifp->if_requestencap == NULL) 860 ifp->if_requestencap = if_requestencap_default; 861 862 if (!vmove) { 863 #ifdef MAC 864 mac_ifnet_create(ifp); 865 #endif 866 867 /* 868 * Create a Link Level name for this device. 869 */ 870 namelen = strlen(ifp->if_xname); 871 /* 872 * Always save enough space for any possible name so we 873 * can do a rename in place later. 874 */ 875 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 876 socksize = masklen + ifp->if_addrlen; 877 if (socksize < sizeof(*sdl)) 878 socksize = sizeof(*sdl); 879 socksize = roundup2(socksize, sizeof(long)); 880 ifasize = sizeof(*ifa) + 2 * socksize; 881 ifa = ifa_alloc(ifasize, M_WAITOK); 882 sdl = (struct sockaddr_dl *)(ifa + 1); 883 sdl->sdl_len = socksize; 884 sdl->sdl_family = AF_LINK; 885 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 886 sdl->sdl_nlen = namelen; 887 sdl->sdl_index = ifp->if_index; 888 sdl->sdl_type = ifp->if_type; 889 ifp->if_addr = ifa; 890 ifa->ifa_ifp = ifp; 891 ifa->ifa_addr = (struct sockaddr *)sdl; 892 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 893 ifa->ifa_netmask = (struct sockaddr *)sdl; 894 sdl->sdl_len = masklen; 895 while (namelen != 0) 896 sdl->sdl_data[--namelen] = 0xff; 897 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 898 /* Reliably crash if used uninitialized. */ 899 ifp->if_broadcastaddr = NULL; 900 901 if (ifp->if_type == IFT_ETHER) { 902 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 903 M_WAITOK | M_ZERO); 904 } 905 906 #if defined(INET) || defined(INET6) 907 /* Use defaults for TSO, if nothing is set */ 908 if (ifp->if_hw_tsomax == 0 && 909 ifp->if_hw_tsomaxsegcount == 0 && 910 ifp->if_hw_tsomaxsegsize == 0) { 911 /* 912 * The TSO defaults needs to be such that an 913 * NFS mbuf list of 35 mbufs totalling just 914 * below 64K works and that a chain of mbufs 915 * can be defragged into at most 32 segments: 916 */ 917 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 918 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 919 ifp->if_hw_tsomaxsegcount = 35; 920 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 921 922 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 923 if (ifp->if_capabilities & IFCAP_TSO) { 924 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 925 ifp->if_hw_tsomax, 926 ifp->if_hw_tsomaxsegcount, 927 ifp->if_hw_tsomaxsegsize); 928 } 929 } 930 #endif 931 } 932 933 if (domain_init_status >= 2) 934 if_attachdomain1(ifp); 935 936 if_link_ifnet(ifp); 937 938 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 939 if (IS_DEFAULT_VNET(curvnet)) 940 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 941 } 942 943 static void 944 if_epochalloc(void *dummy __unused) 945 { 946 947 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 948 } 949 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 950 951 static void 952 if_attachdomain(void *dummy) 953 { 954 struct ifnet *ifp; 955 956 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 957 if_attachdomain1(ifp); 958 } 959 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 960 if_attachdomain, NULL); 961 962 static void 963 if_attachdomain1(struct ifnet *ifp) 964 { 965 struct domain *dp; 966 967 /* 968 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 969 * cannot lock ifp->if_afdata initialization, entirely. 970 */ 971 IF_AFDATA_LOCK(ifp); 972 if (ifp->if_afdata_initialized >= domain_init_status) { 973 IF_AFDATA_UNLOCK(ifp); 974 log(LOG_WARNING, "%s called more than once on %s\n", 975 __func__, ifp->if_xname); 976 return; 977 } 978 ifp->if_afdata_initialized = domain_init_status; 979 IF_AFDATA_UNLOCK(ifp); 980 981 /* address family dependent data region */ 982 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 983 SLIST_FOREACH(dp, &domains, dom_next) { 984 if (dp->dom_ifattach) 985 ifp->if_afdata[dp->dom_family] = 986 (*dp->dom_ifattach)(ifp); 987 } 988 } 989 990 /* 991 * Remove any unicast or broadcast network addresses from an interface. 992 */ 993 void 994 if_purgeaddrs(struct ifnet *ifp) 995 { 996 struct ifaddr *ifa; 997 998 #ifdef INET6 999 /* 1000 * Need to leave multicast addresses of proxy NDP llentries 1001 * before in6_purgeifaddr() because the llentries are keys 1002 * for in6_multi objects of proxy NDP entries. 1003 * in6_purgeifaddr()s clean up llentries including proxy NDPs 1004 * then we would lose the keys if they are called earlier. 1005 */ 1006 in6_purge_proxy_ndp(ifp); 1007 #endif 1008 while (1) { 1009 struct epoch_tracker et; 1010 1011 NET_EPOCH_ENTER(et); 1012 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1013 if (ifa->ifa_addr->sa_family != AF_LINK) 1014 break; 1015 } 1016 NET_EPOCH_EXIT(et); 1017 1018 if (ifa == NULL) 1019 break; 1020 #ifdef INET 1021 /* XXX: Ugly!! ad hoc just for INET */ 1022 if (ifa->ifa_addr->sa_family == AF_INET) { 1023 struct ifreq ifr; 1024 1025 bzero(&ifr, sizeof(ifr)); 1026 ifr.ifr_addr = *ifa->ifa_addr; 1027 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1028 NULL) == 0) 1029 continue; 1030 } 1031 #endif /* INET */ 1032 #ifdef INET6 1033 if (ifa->ifa_addr->sa_family == AF_INET6) { 1034 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1035 /* ifp_addrhead is already updated */ 1036 continue; 1037 } 1038 #endif /* INET6 */ 1039 IF_ADDR_WLOCK(ifp); 1040 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1041 IF_ADDR_WUNLOCK(ifp); 1042 ifa_free(ifa); 1043 } 1044 } 1045 1046 /* 1047 * Remove any multicast network addresses from an interface when an ifnet 1048 * is going away. 1049 */ 1050 static void 1051 if_purgemaddrs(struct ifnet *ifp) 1052 { 1053 struct ifmultiaddr *ifma; 1054 1055 IF_ADDR_WLOCK(ifp); 1056 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1057 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1058 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1059 if_delmulti_locked(ifp, ifma, 1); 1060 } 1061 IF_ADDR_WUNLOCK(ifp); 1062 } 1063 1064 /* 1065 * Detach an interface, removing it from the list of "active" interfaces. 1066 * If vmove flag is set on entry to if_detach_internal(), perform only a 1067 * limited subset of cleanup tasks, given that we are moving an ifnet from 1068 * one vnet to another, where it must be fully operational. 1069 * 1070 * XXXRW: There are some significant questions about event ordering, and 1071 * how to prevent things from starting to use the interface during detach. 1072 */ 1073 void 1074 if_detach(struct ifnet *ifp) 1075 { 1076 bool found; 1077 1078 CURVNET_SET_QUIET(ifp->if_vnet); 1079 found = if_unlink_ifnet(ifp, false); 1080 if (found) { 1081 sx_xlock(&ifnet_detach_sxlock); 1082 if_detach_internal(ifp, false); 1083 sx_xunlock(&ifnet_detach_sxlock); 1084 } 1085 CURVNET_RESTORE(); 1086 } 1087 1088 /* 1089 * The vmove flag, if set, indicates that we are called from a callpath 1090 * that is moving an interface to a different vnet instance. 1091 * 1092 * The shutdown flag, if set, indicates that we are called in the 1093 * process of shutting down a vnet instance. Currently only the 1094 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1095 * on a vnet instance shutdown without this flag being set, e.g., when 1096 * the cloned interfaces are destoyed as first thing of teardown. 1097 */ 1098 static void 1099 if_detach_internal(struct ifnet *ifp, bool vmove) 1100 { 1101 struct ifaddr *ifa; 1102 int i; 1103 struct domain *dp; 1104 #ifdef VIMAGE 1105 bool shutdown; 1106 1107 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1108 #endif 1109 1110 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 1111 1112 /* 1113 * At this point we know the interface still was on the ifnet list 1114 * and we removed it so we are in a stable state. 1115 */ 1116 NET_EPOCH_WAIT(); 1117 1118 /* 1119 * Ensure all pending EPOCH(9) callbacks have been executed. This 1120 * fixes issues about late destruction of multicast options 1121 * which lead to leave group calls, which in turn access the 1122 * belonging ifnet structure: 1123 */ 1124 NET_EPOCH_DRAIN_CALLBACKS(); 1125 1126 /* 1127 * In any case (destroy or vmove) detach us from the groups 1128 * and remove/wait for pending events on the taskq. 1129 * XXX-BZ in theory an interface could still enqueue a taskq change? 1130 */ 1131 if_delgroups(ifp); 1132 1133 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1134 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1135 1136 if_down(ifp); 1137 1138 #ifdef VIMAGE 1139 /* 1140 * On VNET shutdown abort here as the stack teardown will do all 1141 * the work top-down for us. 1142 */ 1143 if (shutdown) { 1144 /* Give interface users the chance to clean up. */ 1145 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1146 1147 /* 1148 * In case of a vmove we are done here without error. 1149 * If we would signal an error it would lead to the same 1150 * abort as if we did not find the ifnet anymore. 1151 * if_detach() calls us in void context and does not care 1152 * about an early abort notification, so life is splendid :) 1153 */ 1154 goto finish_vnet_shutdown; 1155 } 1156 #endif 1157 1158 /* 1159 * At this point we are not tearing down a VNET and are either 1160 * going to destroy or vmove the interface and have to cleanup 1161 * accordingly. 1162 */ 1163 1164 /* 1165 * Remove routes and flush queues. 1166 */ 1167 #ifdef ALTQ 1168 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1169 altq_disable(&ifp->if_snd); 1170 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1171 altq_detach(&ifp->if_snd); 1172 #endif 1173 1174 if_purgeaddrs(ifp); 1175 1176 #ifdef INET 1177 in_ifdetach(ifp); 1178 #endif 1179 1180 #ifdef INET6 1181 /* 1182 * Remove all IPv6 kernel structs related to ifp. This should be done 1183 * before removing routing entries below, since IPv6 interface direct 1184 * routes are expected to be removed by the IPv6-specific kernel API. 1185 * Otherwise, the kernel will detect some inconsistency and bark it. 1186 */ 1187 in6_ifdetach(ifp); 1188 #endif 1189 if_purgemaddrs(ifp); 1190 1191 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1192 if (IS_DEFAULT_VNET(curvnet)) 1193 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1194 1195 if (!vmove) { 1196 /* 1197 * Prevent further calls into the device driver via ifnet. 1198 */ 1199 if_dead(ifp); 1200 1201 /* 1202 * Clean up all addresses. 1203 */ 1204 IF_ADDR_WLOCK(ifp); 1205 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1206 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1207 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1208 IF_ADDR_WUNLOCK(ifp); 1209 ifa_free(ifa); 1210 } else 1211 IF_ADDR_WUNLOCK(ifp); 1212 } 1213 1214 rt_flushifroutes(ifp); 1215 1216 #ifdef VIMAGE 1217 finish_vnet_shutdown: 1218 #endif 1219 /* 1220 * We cannot hold the lock over dom_ifdetach calls as they might 1221 * sleep, for example trying to drain a callout, thus open up the 1222 * theoretical race with re-attaching. 1223 */ 1224 IF_AFDATA_LOCK(ifp); 1225 i = ifp->if_afdata_initialized; 1226 ifp->if_afdata_initialized = 0; 1227 IF_AFDATA_UNLOCK(ifp); 1228 if (i == 0) 1229 return; 1230 SLIST_FOREACH(dp, &domains, dom_next) { 1231 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1232 (*dp->dom_ifdetach)(ifp, 1233 ifp->if_afdata[dp->dom_family]); 1234 ifp->if_afdata[dp->dom_family] = NULL; 1235 } 1236 } 1237 } 1238 1239 #ifdef VIMAGE 1240 /* 1241 * if_vmove() performs a limited version of if_detach() in current 1242 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1243 */ 1244 static void 1245 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1246 { 1247 #ifdef DEV_BPF 1248 /* 1249 * Detach BPF file descriptors from its interface. 1250 */ 1251 bpf_ifdetach(ifp); 1252 #endif 1253 1254 /* 1255 * Detach from current vnet, but preserve LLADDR info, do not 1256 * mark as dead etc. so that the ifnet can be reattached later. 1257 */ 1258 if_detach_internal(ifp, true); 1259 1260 /* 1261 * Perform interface-specific reassignment tasks, if provided by 1262 * the driver. 1263 */ 1264 if (ifp->if_reassign != NULL) 1265 ifp->if_reassign(ifp, new_vnet, NULL); 1266 1267 /* 1268 * Switch to the context of the target vnet. 1269 */ 1270 CURVNET_SET_QUIET(new_vnet); 1271 if_attach_internal(ifp, true); 1272 CURVNET_RESTORE(); 1273 } 1274 1275 /* 1276 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1277 */ 1278 static int 1279 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1280 { 1281 struct prison *pr; 1282 struct ifnet *difp; 1283 bool found; 1284 bool shutdown; 1285 1286 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 1287 1288 /* Try to find the prison within our visibility. */ 1289 sx_slock(&allprison_lock); 1290 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1291 sx_sunlock(&allprison_lock); 1292 if (pr == NULL) 1293 return (ENXIO); 1294 prison_hold_locked(pr); 1295 mtx_unlock(&pr->pr_mtx); 1296 1297 /* Do not try to move the iface from and to the same prison. */ 1298 if (pr->pr_vnet == ifp->if_vnet) { 1299 prison_free(pr); 1300 return (EEXIST); 1301 } 1302 1303 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1304 /* XXX Lock interfaces to avoid races. */ 1305 CURVNET_SET_QUIET(pr->pr_vnet); 1306 difp = ifunit(ifname); 1307 CURVNET_RESTORE(); 1308 if (difp != NULL) { 1309 prison_free(pr); 1310 return (EEXIST); 1311 } 1312 sx_xlock(&ifnet_detach_sxlock); 1313 1314 /* Make sure the VNET is stable. */ 1315 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1316 if (shutdown) { 1317 sx_xunlock(&ifnet_detach_sxlock); 1318 prison_free(pr); 1319 return (EBUSY); 1320 } 1321 1322 found = if_unlink_ifnet(ifp, true); 1323 if (! found) { 1324 sx_xunlock(&ifnet_detach_sxlock); 1325 prison_free(pr); 1326 return (ENODEV); 1327 } 1328 1329 /* Move the interface into the child jail/vnet. */ 1330 if_vmove(ifp, pr->pr_vnet); 1331 1332 /* Report the new if_xname back to the userland. */ 1333 sprintf(ifname, "%s", ifp->if_xname); 1334 1335 sx_xunlock(&ifnet_detach_sxlock); 1336 1337 prison_free(pr); 1338 return (0); 1339 } 1340 1341 static int 1342 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1343 { 1344 struct prison *pr; 1345 struct vnet *vnet_dst; 1346 struct ifnet *ifp; 1347 int found __diagused; 1348 bool shutdown; 1349 1350 /* Try to find the prison within our visibility. */ 1351 sx_slock(&allprison_lock); 1352 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1353 sx_sunlock(&allprison_lock); 1354 if (pr == NULL) 1355 return (ENXIO); 1356 prison_hold_locked(pr); 1357 mtx_unlock(&pr->pr_mtx); 1358 1359 /* Make sure the named iface exists in the source prison/vnet. */ 1360 CURVNET_SET(pr->pr_vnet); 1361 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1362 if (ifp == NULL) { 1363 CURVNET_RESTORE(); 1364 prison_free(pr); 1365 return (ENXIO); 1366 } 1367 1368 /* Do not try to move the iface from and to the same prison. */ 1369 vnet_dst = TD_TO_VNET(td); 1370 if (vnet_dst == ifp->if_vnet) { 1371 CURVNET_RESTORE(); 1372 prison_free(pr); 1373 return (EEXIST); 1374 } 1375 1376 /* Make sure the VNET is stable. */ 1377 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1378 if (shutdown) { 1379 CURVNET_RESTORE(); 1380 prison_free(pr); 1381 return (EBUSY); 1382 } 1383 1384 /* Get interface back from child jail/vnet. */ 1385 found = if_unlink_ifnet(ifp, true); 1386 MPASS(found); 1387 sx_xlock(&ifnet_detach_sxlock); 1388 if_vmove(ifp, vnet_dst); 1389 sx_xunlock(&ifnet_detach_sxlock); 1390 CURVNET_RESTORE(); 1391 1392 /* Report the new if_xname back to the userland. */ 1393 sprintf(ifname, "%s", ifp->if_xname); 1394 1395 prison_free(pr); 1396 return (0); 1397 } 1398 #endif /* VIMAGE */ 1399 1400 /* 1401 * Add a group to an interface 1402 */ 1403 int 1404 if_addgroup(struct ifnet *ifp, const char *groupname) 1405 { 1406 struct ifg_list *ifgl; 1407 struct ifg_group *ifg = NULL; 1408 struct ifg_member *ifgm; 1409 int new = 0; 1410 1411 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1412 groupname[strlen(groupname) - 1] <= '9') 1413 return (EINVAL); 1414 1415 IFNET_WLOCK(); 1416 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1417 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1418 IFNET_WUNLOCK(); 1419 return (EEXIST); 1420 } 1421 1422 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1423 IFNET_WUNLOCK(); 1424 return (ENOMEM); 1425 } 1426 1427 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1428 free(ifgl, M_TEMP); 1429 IFNET_WUNLOCK(); 1430 return (ENOMEM); 1431 } 1432 1433 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1434 if (!strcmp(ifg->ifg_group, groupname)) 1435 break; 1436 1437 if (ifg == NULL) { 1438 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1439 free(ifgl, M_TEMP); 1440 free(ifgm, M_TEMP); 1441 IFNET_WUNLOCK(); 1442 return (ENOMEM); 1443 } 1444 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1445 ifg->ifg_refcnt = 0; 1446 CK_STAILQ_INIT(&ifg->ifg_members); 1447 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1448 new = 1; 1449 } 1450 1451 ifg->ifg_refcnt++; 1452 ifgl->ifgl_group = ifg; 1453 ifgm->ifgm_ifp = ifp; 1454 1455 IF_ADDR_WLOCK(ifp); 1456 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1457 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1458 IF_ADDR_WUNLOCK(ifp); 1459 1460 IFNET_WUNLOCK(); 1461 1462 if (new) 1463 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1464 EVENTHANDLER_INVOKE(group_change_event, groupname); 1465 1466 return (0); 1467 } 1468 1469 /* 1470 * Helper function to remove a group out of an interface. Expects the global 1471 * ifnet lock to be write-locked, and drops it before returning. 1472 */ 1473 static void 1474 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1475 const char *groupname) 1476 { 1477 struct ifg_member *ifgm; 1478 bool freeifgl; 1479 1480 IFNET_WLOCK_ASSERT(); 1481 1482 IF_ADDR_WLOCK(ifp); 1483 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1484 IF_ADDR_WUNLOCK(ifp); 1485 1486 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1487 if (ifgm->ifgm_ifp == ifp) { 1488 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1489 ifg_member, ifgm_next); 1490 break; 1491 } 1492 } 1493 1494 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1495 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1496 ifg_next); 1497 freeifgl = true; 1498 } else { 1499 freeifgl = false; 1500 } 1501 IFNET_WUNLOCK(); 1502 1503 NET_EPOCH_WAIT(); 1504 EVENTHANDLER_INVOKE(group_change_event, groupname); 1505 if (freeifgl) { 1506 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1507 free(ifgl->ifgl_group, M_TEMP); 1508 } 1509 free(ifgm, M_TEMP); 1510 free(ifgl, M_TEMP); 1511 } 1512 1513 /* 1514 * Remove a group from an interface 1515 */ 1516 int 1517 if_delgroup(struct ifnet *ifp, const char *groupname) 1518 { 1519 struct ifg_list *ifgl; 1520 1521 IFNET_WLOCK(); 1522 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1523 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1524 break; 1525 if (ifgl == NULL) { 1526 IFNET_WUNLOCK(); 1527 return (ENOENT); 1528 } 1529 1530 _if_delgroup_locked(ifp, ifgl, groupname); 1531 1532 return (0); 1533 } 1534 1535 /* 1536 * Remove an interface from all groups 1537 */ 1538 static void 1539 if_delgroups(struct ifnet *ifp) 1540 { 1541 struct ifg_list *ifgl; 1542 char groupname[IFNAMSIZ]; 1543 1544 IFNET_WLOCK(); 1545 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1546 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1547 _if_delgroup_locked(ifp, ifgl, groupname); 1548 IFNET_WLOCK(); 1549 } 1550 IFNET_WUNLOCK(); 1551 } 1552 1553 /* 1554 * Stores all groups from an interface in memory pointed to by ifgr. 1555 */ 1556 static int 1557 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1558 { 1559 int len, error; 1560 struct ifg_list *ifgl; 1561 struct ifg_req ifgrq, *ifgp; 1562 1563 NET_EPOCH_ASSERT(); 1564 1565 if (ifgr->ifgr_len == 0) { 1566 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1567 ifgr->ifgr_len += sizeof(struct ifg_req); 1568 return (0); 1569 } 1570 1571 len = ifgr->ifgr_len; 1572 ifgp = ifgr->ifgr_groups; 1573 /* XXX: wire */ 1574 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1575 if (len < sizeof(ifgrq)) 1576 return (EINVAL); 1577 bzero(&ifgrq, sizeof ifgrq); 1578 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1579 sizeof(ifgrq.ifgrq_group)); 1580 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1581 return (error); 1582 len -= sizeof(ifgrq); 1583 ifgp++; 1584 } 1585 1586 return (0); 1587 } 1588 1589 /* 1590 * Stores all members of a group in memory pointed to by igfr 1591 */ 1592 static int 1593 if_getgroupmembers(struct ifgroupreq *ifgr) 1594 { 1595 struct ifg_group *ifg; 1596 struct ifg_member *ifgm; 1597 struct ifg_req ifgrq, *ifgp; 1598 int len, error; 1599 1600 IFNET_RLOCK(); 1601 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1602 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1603 break; 1604 if (ifg == NULL) { 1605 IFNET_RUNLOCK(); 1606 return (ENOENT); 1607 } 1608 1609 if (ifgr->ifgr_len == 0) { 1610 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1611 ifgr->ifgr_len += sizeof(ifgrq); 1612 IFNET_RUNLOCK(); 1613 return (0); 1614 } 1615 1616 len = ifgr->ifgr_len; 1617 ifgp = ifgr->ifgr_groups; 1618 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1619 if (len < sizeof(ifgrq)) { 1620 IFNET_RUNLOCK(); 1621 return (EINVAL); 1622 } 1623 bzero(&ifgrq, sizeof ifgrq); 1624 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1625 sizeof(ifgrq.ifgrq_member)); 1626 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1627 IFNET_RUNLOCK(); 1628 return (error); 1629 } 1630 len -= sizeof(ifgrq); 1631 ifgp++; 1632 } 1633 IFNET_RUNLOCK(); 1634 1635 return (0); 1636 } 1637 1638 /* 1639 * Return counter values from counter(9)s stored in ifnet. 1640 */ 1641 uint64_t 1642 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1643 { 1644 1645 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1646 1647 return (counter_u64_fetch(ifp->if_counters[cnt])); 1648 } 1649 1650 /* 1651 * Increase an ifnet counter. Usually used for counters shared 1652 * between the stack and a driver, but function supports them all. 1653 */ 1654 void 1655 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1656 { 1657 1658 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1659 1660 counter_u64_add(ifp->if_counters[cnt], inc); 1661 } 1662 1663 /* 1664 * Copy data from ifnet to userland API structure if_data. 1665 */ 1666 void 1667 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1668 { 1669 1670 ifd->ifi_type = ifp->if_type; 1671 ifd->ifi_physical = 0; 1672 ifd->ifi_addrlen = ifp->if_addrlen; 1673 ifd->ifi_hdrlen = ifp->if_hdrlen; 1674 ifd->ifi_link_state = ifp->if_link_state; 1675 ifd->ifi_vhid = 0; 1676 ifd->ifi_datalen = sizeof(struct if_data); 1677 ifd->ifi_mtu = ifp->if_mtu; 1678 ifd->ifi_metric = ifp->if_metric; 1679 ifd->ifi_baudrate = ifp->if_baudrate; 1680 ifd->ifi_hwassist = ifp->if_hwassist; 1681 ifd->ifi_epoch = ifp->if_epoch; 1682 ifd->ifi_lastchange = ifp->if_lastchange; 1683 1684 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1685 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1686 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1687 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1688 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1689 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1690 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1691 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1692 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1693 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1694 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1695 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1696 } 1697 1698 /* 1699 * Initialization, destruction and refcounting functions for ifaddrs. 1700 */ 1701 struct ifaddr * 1702 ifa_alloc(size_t size, int flags) 1703 { 1704 struct ifaddr *ifa; 1705 1706 KASSERT(size >= sizeof(struct ifaddr), 1707 ("%s: invalid size %zu", __func__, size)); 1708 1709 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1710 if (ifa == NULL) 1711 return (NULL); 1712 1713 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1714 goto fail; 1715 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1716 goto fail; 1717 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1718 goto fail; 1719 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1720 goto fail; 1721 1722 refcount_init(&ifa->ifa_refcnt, 1); 1723 1724 return (ifa); 1725 1726 fail: 1727 /* free(NULL) is okay */ 1728 counter_u64_free(ifa->ifa_opackets); 1729 counter_u64_free(ifa->ifa_ipackets); 1730 counter_u64_free(ifa->ifa_obytes); 1731 counter_u64_free(ifa->ifa_ibytes); 1732 free(ifa, M_IFADDR); 1733 1734 return (NULL); 1735 } 1736 1737 void 1738 ifa_ref(struct ifaddr *ifa) 1739 { 1740 u_int old __diagused; 1741 1742 old = refcount_acquire(&ifa->ifa_refcnt); 1743 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1744 } 1745 1746 int 1747 ifa_try_ref(struct ifaddr *ifa) 1748 { 1749 1750 NET_EPOCH_ASSERT(); 1751 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1752 } 1753 1754 static void 1755 ifa_destroy(epoch_context_t ctx) 1756 { 1757 struct ifaddr *ifa; 1758 1759 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1760 counter_u64_free(ifa->ifa_opackets); 1761 counter_u64_free(ifa->ifa_ipackets); 1762 counter_u64_free(ifa->ifa_obytes); 1763 counter_u64_free(ifa->ifa_ibytes); 1764 free(ifa, M_IFADDR); 1765 } 1766 1767 void 1768 ifa_free(struct ifaddr *ifa) 1769 { 1770 1771 if (refcount_release(&ifa->ifa_refcnt)) 1772 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1773 } 1774 1775 /* 1776 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1777 * structs used to represent other address families, it is necessary 1778 * to perform a different comparison. 1779 */ 1780 static bool 1781 sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b) 1782 { 1783 const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a; 1784 const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b; 1785 1786 return (sdl1->sdl_len == sdl2->sdl_len && 1787 bcmp(sdl1->sdl_data + sdl1->sdl_nlen, 1788 sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0); 1789 } 1790 1791 /* 1792 * Locate an interface based on a complete address. 1793 */ 1794 /*ARGSUSED*/ 1795 struct ifaddr * 1796 ifa_ifwithaddr(const struct sockaddr *addr) 1797 { 1798 struct ifnet *ifp; 1799 struct ifaddr *ifa; 1800 1801 NET_EPOCH_ASSERT(); 1802 1803 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1804 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1805 if (ifa->ifa_addr->sa_family != addr->sa_family) 1806 continue; 1807 if (sa_equal(addr, ifa->ifa_addr)) { 1808 goto done; 1809 } 1810 /* IP6 doesn't have broadcast */ 1811 if ((ifp->if_flags & IFF_BROADCAST) && 1812 ifa->ifa_broadaddr && 1813 ifa->ifa_broadaddr->sa_len != 0 && 1814 sa_equal(ifa->ifa_broadaddr, addr)) { 1815 goto done; 1816 } 1817 } 1818 } 1819 ifa = NULL; 1820 done: 1821 return (ifa); 1822 } 1823 1824 int 1825 ifa_ifwithaddr_check(const struct sockaddr *addr) 1826 { 1827 struct epoch_tracker et; 1828 int rc; 1829 1830 NET_EPOCH_ENTER(et); 1831 rc = (ifa_ifwithaddr(addr) != NULL); 1832 NET_EPOCH_EXIT(et); 1833 return (rc); 1834 } 1835 1836 /* 1837 * Locate an interface based on the broadcast address. 1838 */ 1839 /* ARGSUSED */ 1840 struct ifaddr * 1841 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1842 { 1843 struct ifnet *ifp; 1844 struct ifaddr *ifa; 1845 1846 NET_EPOCH_ASSERT(); 1847 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1848 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1849 continue; 1850 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1851 if (ifa->ifa_addr->sa_family != addr->sa_family) 1852 continue; 1853 if ((ifp->if_flags & IFF_BROADCAST) && 1854 ifa->ifa_broadaddr && 1855 ifa->ifa_broadaddr->sa_len != 0 && 1856 sa_equal(ifa->ifa_broadaddr, addr)) { 1857 goto done; 1858 } 1859 } 1860 } 1861 ifa = NULL; 1862 done: 1863 return (ifa); 1864 } 1865 1866 /* 1867 * Locate the point to point interface with a given destination address. 1868 */ 1869 /*ARGSUSED*/ 1870 struct ifaddr * 1871 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1872 { 1873 struct ifnet *ifp; 1874 struct ifaddr *ifa; 1875 1876 NET_EPOCH_ASSERT(); 1877 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1878 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1879 continue; 1880 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1881 continue; 1882 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1883 if (ifa->ifa_addr->sa_family != addr->sa_family) 1884 continue; 1885 if (ifa->ifa_dstaddr != NULL && 1886 sa_equal(addr, ifa->ifa_dstaddr)) { 1887 goto done; 1888 } 1889 } 1890 } 1891 ifa = NULL; 1892 done: 1893 return (ifa); 1894 } 1895 1896 /* 1897 * Find an interface on a specific network. If many, choice 1898 * is most specific found. 1899 */ 1900 struct ifaddr * 1901 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1902 { 1903 struct ifnet *ifp; 1904 struct ifaddr *ifa; 1905 struct ifaddr *ifa_maybe = NULL; 1906 u_int af = addr->sa_family; 1907 const char *addr_data = addr->sa_data, *cplim; 1908 1909 NET_EPOCH_ASSERT(); 1910 /* 1911 * AF_LINK addresses can be looked up directly by their index number, 1912 * so do that if we can. 1913 */ 1914 if (af == AF_LINK) { 1915 ifp = ifnet_byindex( 1916 ((const struct sockaddr_dl *)addr)->sdl_index); 1917 return (ifp ? ifp->if_addr : NULL); 1918 } 1919 1920 /* 1921 * Scan though each interface, looking for ones that have addresses 1922 * in this address family and the requested fib. 1923 */ 1924 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1925 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1926 continue; 1927 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1928 const char *cp, *cp2, *cp3; 1929 1930 if (ifa->ifa_addr->sa_family != af) 1931 next: continue; 1932 if (af == AF_INET && 1933 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1934 /* 1935 * This is a bit broken as it doesn't 1936 * take into account that the remote end may 1937 * be a single node in the network we are 1938 * looking for. 1939 * The trouble is that we don't know the 1940 * netmask for the remote end. 1941 */ 1942 if (ifa->ifa_dstaddr != NULL && 1943 sa_equal(addr, ifa->ifa_dstaddr)) { 1944 goto done; 1945 } 1946 } else { 1947 /* 1948 * Scan all the bits in the ifa's address. 1949 * If a bit dissagrees with what we are 1950 * looking for, mask it with the netmask 1951 * to see if it really matters. 1952 * (A byte at a time) 1953 */ 1954 if (ifa->ifa_netmask == 0) 1955 continue; 1956 cp = addr_data; 1957 cp2 = ifa->ifa_addr->sa_data; 1958 cp3 = ifa->ifa_netmask->sa_data; 1959 cplim = ifa->ifa_netmask->sa_len 1960 + (char *)ifa->ifa_netmask; 1961 while (cp3 < cplim) 1962 if ((*cp++ ^ *cp2++) & *cp3++) 1963 goto next; /* next address! */ 1964 /* 1965 * If the netmask of what we just found 1966 * is more specific than what we had before 1967 * (if we had one), or if the virtual status 1968 * of new prefix is better than of the old one, 1969 * then remember the new one before continuing 1970 * to search for an even better one. 1971 */ 1972 if (ifa_maybe == NULL || 1973 ifa_preferred(ifa_maybe, ifa) || 1974 rn_refines((caddr_t)ifa->ifa_netmask, 1975 (caddr_t)ifa_maybe->ifa_netmask)) { 1976 ifa_maybe = ifa; 1977 } 1978 } 1979 } 1980 } 1981 ifa = ifa_maybe; 1982 ifa_maybe = NULL; 1983 done: 1984 return (ifa); 1985 } 1986 1987 /* 1988 * Find an interface address specific to an interface best matching 1989 * a given address. 1990 */ 1991 struct ifaddr * 1992 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1993 { 1994 struct ifaddr *ifa; 1995 const char *cp, *cp2, *cp3; 1996 char *cplim; 1997 struct ifaddr *ifa_maybe = NULL; 1998 u_int af = addr->sa_family; 1999 2000 if (af >= AF_MAX) 2001 return (NULL); 2002 2003 NET_EPOCH_ASSERT(); 2004 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2005 if (ifa->ifa_addr->sa_family != af) 2006 continue; 2007 if (ifa_maybe == NULL) 2008 ifa_maybe = ifa; 2009 if (ifa->ifa_netmask == 0) { 2010 if (sa_equal(addr, ifa->ifa_addr) || 2011 (ifa->ifa_dstaddr && 2012 sa_equal(addr, ifa->ifa_dstaddr))) 2013 goto done; 2014 continue; 2015 } 2016 if (ifp->if_flags & IFF_POINTOPOINT) { 2017 if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr)) 2018 goto done; 2019 } else { 2020 cp = addr->sa_data; 2021 cp2 = ifa->ifa_addr->sa_data; 2022 cp3 = ifa->ifa_netmask->sa_data; 2023 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2024 for (; cp3 < cplim; cp3++) 2025 if ((*cp++ ^ *cp2++) & *cp3) 2026 break; 2027 if (cp3 == cplim) 2028 goto done; 2029 } 2030 } 2031 ifa = ifa_maybe; 2032 done: 2033 return (ifa); 2034 } 2035 2036 /* 2037 * See whether new ifa is better than current one: 2038 * 1) A non-virtual one is preferred over virtual. 2039 * 2) A virtual in master state preferred over any other state. 2040 * 2041 * Used in several address selecting functions. 2042 */ 2043 int 2044 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2045 { 2046 2047 return (cur->ifa_carp && (!next->ifa_carp || 2048 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2049 } 2050 2051 struct sockaddr_dl * 2052 link_alloc_sdl(size_t size, int flags) 2053 { 2054 2055 return (malloc(size, M_TEMP, flags)); 2056 } 2057 2058 void 2059 link_free_sdl(struct sockaddr *sa) 2060 { 2061 free(sa, M_TEMP); 2062 } 2063 2064 /* 2065 * Fills in given sdl with interface basic info. 2066 * Returns pointer to filled sdl. 2067 */ 2068 struct sockaddr_dl * 2069 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2070 { 2071 struct sockaddr_dl *sdl; 2072 2073 sdl = (struct sockaddr_dl *)paddr; 2074 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2075 sdl->sdl_len = sizeof(struct sockaddr_dl); 2076 sdl->sdl_family = AF_LINK; 2077 sdl->sdl_index = ifp->if_index; 2078 sdl->sdl_type = iftype; 2079 2080 return (sdl); 2081 } 2082 2083 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2084 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2085 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2086 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2087 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2088 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2089 int (*vlan_setcookie_p)(struct ifnet *, void *); 2090 void *(*vlan_cookie_p)(struct ifnet *); 2091 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 2092 2093 /* 2094 * Handle a change in the interface link state. To avoid LORs 2095 * between driver lock and upper layer locks, as well as possible 2096 * recursions, we post event to taskqueue, and all job 2097 * is done in static do_link_state_change(). 2098 */ 2099 void 2100 if_link_state_change(struct ifnet *ifp, int link_state) 2101 { 2102 /* Return if state hasn't changed. */ 2103 if (ifp->if_link_state == link_state) 2104 return; 2105 2106 ifp->if_link_state = link_state; 2107 2108 /* XXXGL: reference ifp? */ 2109 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2110 } 2111 2112 static void 2113 do_link_state_change(void *arg, int pending) 2114 { 2115 struct ifnet *ifp; 2116 int link_state; 2117 2118 ifp = arg; 2119 link_state = ifp->if_link_state; 2120 2121 CURVNET_SET(ifp->if_vnet); 2122 rt_ifmsg(ifp, 0); 2123 if (ifp->if_vlantrunk != NULL) 2124 (*vlan_link_state_p)(ifp); 2125 2126 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2127 ifp->if_l2com != NULL) 2128 (*ng_ether_link_state_p)(ifp, link_state); 2129 if (ifp->if_carp) 2130 (*carp_linkstate_p)(ifp); 2131 if (ifp->if_bridge) 2132 ifp->if_bridge_linkstate(ifp); 2133 if (ifp->if_lagg) 2134 (*lagg_linkstate_p)(ifp, link_state); 2135 2136 if (IS_DEFAULT_VNET(curvnet)) 2137 devctl_notify("IFNET", ifp->if_xname, 2138 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2139 NULL); 2140 if (pending > 1) 2141 if_printf(ifp, "%d link states coalesced\n", pending); 2142 if (log_link_state_change) 2143 if_printf(ifp, "link state changed to %s\n", 2144 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2145 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2146 CURVNET_RESTORE(); 2147 } 2148 2149 /* 2150 * Mark an interface down and notify protocols of 2151 * the transition. 2152 */ 2153 void 2154 if_down(struct ifnet *ifp) 2155 { 2156 2157 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2158 2159 ifp->if_flags &= ~IFF_UP; 2160 getmicrotime(&ifp->if_lastchange); 2161 ifp->if_qflush(ifp); 2162 2163 if (ifp->if_carp) 2164 (*carp_linkstate_p)(ifp); 2165 rt_ifmsg(ifp, IFF_UP); 2166 } 2167 2168 /* 2169 * Mark an interface up and notify protocols of 2170 * the transition. 2171 */ 2172 void 2173 if_up(struct ifnet *ifp) 2174 { 2175 2176 ifp->if_flags |= IFF_UP; 2177 getmicrotime(&ifp->if_lastchange); 2178 if (ifp->if_carp) 2179 (*carp_linkstate_p)(ifp); 2180 rt_ifmsg(ifp, IFF_UP); 2181 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2182 } 2183 2184 /* 2185 * Flush an interface queue. 2186 */ 2187 void 2188 if_qflush(struct ifnet *ifp) 2189 { 2190 struct mbuf *m, *n; 2191 struct ifaltq *ifq; 2192 2193 ifq = &ifp->if_snd; 2194 IFQ_LOCK(ifq); 2195 #ifdef ALTQ 2196 if (ALTQ_IS_ENABLED(ifq)) 2197 ALTQ_PURGE(ifq); 2198 #endif 2199 n = ifq->ifq_head; 2200 while ((m = n) != NULL) { 2201 n = m->m_nextpkt; 2202 m_freem(m); 2203 } 2204 ifq->ifq_head = 0; 2205 ifq->ifq_tail = 0; 2206 ifq->ifq_len = 0; 2207 IFQ_UNLOCK(ifq); 2208 } 2209 2210 /* 2211 * Map interface name to interface structure pointer, with or without 2212 * returning a reference. 2213 */ 2214 struct ifnet * 2215 ifunit_ref(const char *name) 2216 { 2217 struct epoch_tracker et; 2218 struct ifnet *ifp; 2219 2220 NET_EPOCH_ENTER(et); 2221 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2222 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2223 !(ifp->if_flags & IFF_DYING)) 2224 break; 2225 } 2226 if (ifp != NULL) { 2227 if_ref(ifp); 2228 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2229 } 2230 2231 NET_EPOCH_EXIT(et); 2232 return (ifp); 2233 } 2234 2235 struct ifnet * 2236 ifunit(const char *name) 2237 { 2238 struct epoch_tracker et; 2239 struct ifnet *ifp; 2240 2241 NET_EPOCH_ENTER(et); 2242 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2243 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2244 break; 2245 } 2246 NET_EPOCH_EXIT(et); 2247 return (ifp); 2248 } 2249 2250 void * 2251 ifr_buffer_get_buffer(void *data) 2252 { 2253 union ifreq_union *ifrup; 2254 2255 ifrup = data; 2256 #ifdef COMPAT_FREEBSD32 2257 if (SV_CURPROC_FLAG(SV_ILP32)) 2258 return ((void *)(uintptr_t) 2259 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2260 #endif 2261 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2262 } 2263 2264 static void 2265 ifr_buffer_set_buffer_null(void *data) 2266 { 2267 union ifreq_union *ifrup; 2268 2269 ifrup = data; 2270 #ifdef COMPAT_FREEBSD32 2271 if (SV_CURPROC_FLAG(SV_ILP32)) 2272 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2273 else 2274 #endif 2275 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2276 } 2277 2278 size_t 2279 ifr_buffer_get_length(void *data) 2280 { 2281 union ifreq_union *ifrup; 2282 2283 ifrup = data; 2284 #ifdef COMPAT_FREEBSD32 2285 if (SV_CURPROC_FLAG(SV_ILP32)) 2286 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2287 #endif 2288 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2289 } 2290 2291 static void 2292 ifr_buffer_set_length(void *data, size_t len) 2293 { 2294 union ifreq_union *ifrup; 2295 2296 ifrup = data; 2297 #ifdef COMPAT_FREEBSD32 2298 if (SV_CURPROC_FLAG(SV_ILP32)) 2299 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2300 else 2301 #endif 2302 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2303 } 2304 2305 void * 2306 ifr_data_get_ptr(void *ifrp) 2307 { 2308 union ifreq_union *ifrup; 2309 2310 ifrup = ifrp; 2311 #ifdef COMPAT_FREEBSD32 2312 if (SV_CURPROC_FLAG(SV_ILP32)) 2313 return ((void *)(uintptr_t) 2314 ifrup->ifr32.ifr_ifru.ifru_data); 2315 #endif 2316 return (ifrup->ifr.ifr_ifru.ifru_data); 2317 } 2318 2319 struct ifcap_nv_bit_name { 2320 uint64_t cap_bit; 2321 const char *cap_name; 2322 }; 2323 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2324 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2325 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2326 CAPNV(RXCSUM), 2327 CAPNV(TXCSUM), 2328 CAPNV(NETCONS), 2329 CAPNV(VLAN_MTU), 2330 CAPNV(VLAN_HWTAGGING), 2331 CAPNV(JUMBO_MTU), 2332 CAPNV(POLLING), 2333 CAPNV(VLAN_HWCSUM), 2334 CAPNV(TSO4), 2335 CAPNV(TSO6), 2336 CAPNV(LRO), 2337 CAPNV(WOL_UCAST), 2338 CAPNV(WOL_MCAST), 2339 CAPNV(WOL_MAGIC), 2340 CAPNV(TOE4), 2341 CAPNV(TOE6), 2342 CAPNV(VLAN_HWFILTER), 2343 CAPNV(VLAN_HWTSO), 2344 CAPNV(LINKSTATE), 2345 CAPNV(NETMAP), 2346 CAPNV(RXCSUM_IPV6), 2347 CAPNV(TXCSUM_IPV6), 2348 CAPNV(HWSTATS), 2349 CAPNV(TXRTLMT), 2350 CAPNV(HWRXTSTMP), 2351 CAPNV(MEXTPG), 2352 CAPNV(TXTLS4), 2353 CAPNV(TXTLS6), 2354 CAPNV(VXLAN_HWCSUM), 2355 CAPNV(VXLAN_HWTSO), 2356 CAPNV(TXTLS_RTLMT), 2357 {0, NULL} 2358 }; 2359 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \ 2360 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2361 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2362 CAP2NV(RXTLS4), 2363 CAP2NV(RXTLS6), 2364 CAP2NV(IPSEC_OFFLOAD), 2365 {0, NULL} 2366 }; 2367 #undef CAPNV 2368 #undef CAP2NV 2369 2370 int 2371 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2372 const struct ifcap_nv_bit_name *nn, bool all) 2373 { 2374 int i, res; 2375 2376 res = 0; 2377 for (i = 0; nn[i].cap_name != NULL; i++) { 2378 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2379 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2380 res |= nn[i].cap_bit; 2381 } else { 2382 res |= *old_cap & nn[i].cap_bit; 2383 } 2384 } 2385 return (res); 2386 } 2387 2388 void 2389 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2390 int ifr_cap, int ifr_req) 2391 { 2392 int i; 2393 2394 for (i = 0; nn[i].cap_name != NULL; i++) { 2395 if ((nn[i].cap_bit & ifr_cap) != 0) { 2396 nvlist_add_bool(nv, nn[i].cap_name, 2397 (nn[i].cap_bit & ifr_req) != 0); 2398 } 2399 } 2400 } 2401 2402 /* 2403 * Hardware specific interface ioctls. 2404 */ 2405 int 2406 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2407 { 2408 struct ifreq *ifr; 2409 int error = 0, do_ifup = 0; 2410 int new_flags, temp_flags; 2411 size_t descrlen, nvbuflen; 2412 char *descrbuf; 2413 char new_name[IFNAMSIZ]; 2414 void *buf; 2415 nvlist_t *nvcap; 2416 struct siocsifcapnv_driver_data drv_ioctl_data; 2417 2418 ifr = (struct ifreq *)data; 2419 switch (cmd) { 2420 case SIOCGIFINDEX: 2421 ifr->ifr_index = ifp->if_index; 2422 break; 2423 2424 case SIOCGIFFLAGS: 2425 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2426 ifr->ifr_flags = temp_flags & 0xffff; 2427 ifr->ifr_flagshigh = temp_flags >> 16; 2428 break; 2429 2430 case SIOCGIFCAP: 2431 ifr->ifr_reqcap = ifp->if_capabilities; 2432 ifr->ifr_curcap = ifp->if_capenable; 2433 break; 2434 2435 case SIOCGIFCAPNV: 2436 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2437 error = EINVAL; 2438 break; 2439 } 2440 buf = NULL; 2441 nvcap = nvlist_create(0); 2442 for (;;) { 2443 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2444 ifp->if_capabilities, ifp->if_capenable); 2445 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2446 ifp->if_capabilities2, ifp->if_capenable2); 2447 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2448 __DECONST(caddr_t, nvcap)); 2449 if (error != 0) { 2450 if_printf(ifp, 2451 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2452 error); 2453 break; 2454 } 2455 buf = nvlist_pack(nvcap, &nvbuflen); 2456 if (buf == NULL) { 2457 error = nvlist_error(nvcap); 2458 if (error == 0) 2459 error = EDOOFUS; 2460 break; 2461 } 2462 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2463 ifr->ifr_cap_nv.length = nvbuflen; 2464 ifr->ifr_cap_nv.buffer = NULL; 2465 error = EFBIG; 2466 break; 2467 } 2468 ifr->ifr_cap_nv.length = nvbuflen; 2469 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2470 break; 2471 } 2472 free(buf, M_NVLIST); 2473 nvlist_destroy(nvcap); 2474 break; 2475 2476 case SIOCGIFDATA: 2477 { 2478 struct if_data ifd; 2479 2480 /* Ensure uninitialised padding is not leaked. */ 2481 memset(&ifd, 0, sizeof(ifd)); 2482 2483 if_data_copy(ifp, &ifd); 2484 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2485 break; 2486 } 2487 2488 #ifdef MAC 2489 case SIOCGIFMAC: 2490 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2491 break; 2492 #endif 2493 2494 case SIOCGIFMETRIC: 2495 ifr->ifr_metric = ifp->if_metric; 2496 break; 2497 2498 case SIOCGIFMTU: 2499 ifr->ifr_mtu = ifp->if_mtu; 2500 break; 2501 2502 case SIOCGIFPHYS: 2503 /* XXXGL: did this ever worked? */ 2504 ifr->ifr_phys = 0; 2505 break; 2506 2507 case SIOCGIFDESCR: 2508 error = 0; 2509 sx_slock(&ifdescr_sx); 2510 if (ifp->if_description == NULL) 2511 error = ENOMSG; 2512 else { 2513 /* space for terminating nul */ 2514 descrlen = strlen(ifp->if_description) + 1; 2515 if (ifr_buffer_get_length(ifr) < descrlen) 2516 ifr_buffer_set_buffer_null(ifr); 2517 else 2518 error = copyout(ifp->if_description, 2519 ifr_buffer_get_buffer(ifr), descrlen); 2520 ifr_buffer_set_length(ifr, descrlen); 2521 } 2522 sx_sunlock(&ifdescr_sx); 2523 break; 2524 2525 case SIOCSIFDESCR: 2526 error = priv_check(td, PRIV_NET_SETIFDESCR); 2527 if (error) 2528 return (error); 2529 2530 /* 2531 * Copy only (length-1) bytes to make sure that 2532 * if_description is always nul terminated. The 2533 * length parameter is supposed to count the 2534 * terminating nul in. 2535 */ 2536 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2537 return (ENAMETOOLONG); 2538 else if (ifr_buffer_get_length(ifr) == 0) 2539 descrbuf = NULL; 2540 else { 2541 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2542 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2543 ifr_buffer_get_length(ifr) - 1); 2544 if (error) { 2545 if_freedescr(descrbuf); 2546 break; 2547 } 2548 } 2549 2550 if_setdescr(ifp, descrbuf); 2551 getmicrotime(&ifp->if_lastchange); 2552 break; 2553 2554 case SIOCGIFFIB: 2555 ifr->ifr_fib = ifp->if_fib; 2556 break; 2557 2558 case SIOCSIFFIB: 2559 error = priv_check(td, PRIV_NET_SETIFFIB); 2560 if (error) 2561 return (error); 2562 if (ifr->ifr_fib >= rt_numfibs) 2563 return (EINVAL); 2564 2565 ifp->if_fib = ifr->ifr_fib; 2566 break; 2567 2568 case SIOCSIFFLAGS: 2569 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2570 if (error) 2571 return (error); 2572 /* 2573 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2574 * check, so we don't need special handling here yet. 2575 */ 2576 new_flags = (ifr->ifr_flags & 0xffff) | 2577 (ifr->ifr_flagshigh << 16); 2578 if (ifp->if_flags & IFF_UP && 2579 (new_flags & IFF_UP) == 0) { 2580 if_down(ifp); 2581 } else if (new_flags & IFF_UP && 2582 (ifp->if_flags & IFF_UP) == 0) { 2583 do_ifup = 1; 2584 } 2585 2586 /* 2587 * See if the promiscuous mode or allmulti bits are about to 2588 * flip. They require special handling because in-kernel 2589 * consumers may indepdently toggle them. 2590 */ 2591 if_setppromisc(ifp, new_flags & IFF_PPROMISC); 2592 if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) { 2593 if (new_flags & IFF_PALLMULTI) 2594 ifp->if_flags |= IFF_ALLMULTI; 2595 else if (ifp->if_amcount == 0) 2596 ifp->if_flags &= ~IFF_ALLMULTI; 2597 } 2598 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2599 (new_flags &~ IFF_CANTCHANGE); 2600 if (ifp->if_ioctl) { 2601 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2602 } 2603 if (do_ifup) 2604 if_up(ifp); 2605 getmicrotime(&ifp->if_lastchange); 2606 break; 2607 2608 case SIOCSIFCAP: 2609 error = priv_check(td, PRIV_NET_SETIFCAP); 2610 if (error != 0) 2611 return (error); 2612 if (ifp->if_ioctl == NULL) 2613 return (EOPNOTSUPP); 2614 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2615 return (EINVAL); 2616 error = (*ifp->if_ioctl)(ifp, cmd, data); 2617 if (error == 0) 2618 getmicrotime(&ifp->if_lastchange); 2619 break; 2620 2621 case SIOCSIFCAPNV: 2622 error = priv_check(td, PRIV_NET_SETIFCAP); 2623 if (error != 0) 2624 return (error); 2625 if (ifp->if_ioctl == NULL) 2626 return (EOPNOTSUPP); 2627 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2628 return (EINVAL); 2629 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2630 return (EINVAL); 2631 nvcap = NULL; 2632 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2633 for (;;) { 2634 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2635 ifr->ifr_cap_nv.length); 2636 if (error != 0) 2637 break; 2638 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2639 if (nvcap == NULL) { 2640 error = EINVAL; 2641 break; 2642 } 2643 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2644 &ifp->if_capenable, ifcap_nv_bit_names, false); 2645 if ((drv_ioctl_data.reqcap & 2646 ~ifp->if_capabilities) != 0) { 2647 error = EINVAL; 2648 break; 2649 } 2650 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2651 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2652 if ((drv_ioctl_data.reqcap2 & 2653 ~ifp->if_capabilities2) != 0) { 2654 error = EINVAL; 2655 break; 2656 } 2657 drv_ioctl_data.nvcap = nvcap; 2658 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2659 (caddr_t)&drv_ioctl_data); 2660 break; 2661 } 2662 nvlist_destroy(nvcap); 2663 free(buf, M_TEMP); 2664 if (error == 0) 2665 getmicrotime(&ifp->if_lastchange); 2666 break; 2667 2668 #ifdef MAC 2669 case SIOCSIFMAC: 2670 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2671 break; 2672 #endif 2673 2674 case SIOCSIFNAME: 2675 error = priv_check(td, PRIV_NET_SETIFNAME); 2676 if (error) 2677 return (error); 2678 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2679 NULL); 2680 if (error != 0) 2681 return (error); 2682 error = if_rename(ifp, new_name); 2683 break; 2684 2685 #ifdef VIMAGE 2686 case SIOCSIFVNET: 2687 error = priv_check(td, PRIV_NET_SETIFVNET); 2688 if (error) 2689 return (error); 2690 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2691 break; 2692 #endif 2693 2694 case SIOCSIFMETRIC: 2695 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2696 if (error) 2697 return (error); 2698 ifp->if_metric = ifr->ifr_metric; 2699 getmicrotime(&ifp->if_lastchange); 2700 break; 2701 2702 case SIOCSIFPHYS: 2703 error = priv_check(td, PRIV_NET_SETIFPHYS); 2704 if (error) 2705 return (error); 2706 if (ifp->if_ioctl == NULL) 2707 return (EOPNOTSUPP); 2708 error = (*ifp->if_ioctl)(ifp, cmd, data); 2709 if (error == 0) 2710 getmicrotime(&ifp->if_lastchange); 2711 break; 2712 2713 case SIOCSIFMTU: 2714 { 2715 u_long oldmtu = ifp->if_mtu; 2716 2717 error = priv_check(td, PRIV_NET_SETIFMTU); 2718 if (error) 2719 return (error); 2720 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2721 return (EINVAL); 2722 if (ifp->if_ioctl == NULL) 2723 return (EOPNOTSUPP); 2724 /* Disallow MTU changes on bridge member interfaces. */ 2725 if (ifp->if_bridge) 2726 return (EOPNOTSUPP); 2727 error = (*ifp->if_ioctl)(ifp, cmd, data); 2728 if (error == 0) { 2729 getmicrotime(&ifp->if_lastchange); 2730 rt_ifmsg(ifp, 0); 2731 #ifdef INET 2732 DEBUGNET_NOTIFY_MTU(ifp); 2733 #endif 2734 } 2735 /* 2736 * If the link MTU changed, do network layer specific procedure. 2737 */ 2738 if (ifp->if_mtu != oldmtu) 2739 if_notifymtu(ifp); 2740 break; 2741 } 2742 2743 case SIOCADDMULTI: 2744 case SIOCDELMULTI: 2745 if (cmd == SIOCADDMULTI) 2746 error = priv_check(td, PRIV_NET_ADDMULTI); 2747 else 2748 error = priv_check(td, PRIV_NET_DELMULTI); 2749 if (error) 2750 return (error); 2751 2752 /* Don't allow group membership on non-multicast interfaces. */ 2753 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2754 return (EOPNOTSUPP); 2755 2756 /* Don't let users screw up protocols' entries. */ 2757 if (ifr->ifr_addr.sa_family != AF_LINK) 2758 return (EINVAL); 2759 2760 if (cmd == SIOCADDMULTI) { 2761 struct epoch_tracker et; 2762 struct ifmultiaddr *ifma; 2763 2764 /* 2765 * Userland is only permitted to join groups once 2766 * via the if_addmulti() KPI, because it cannot hold 2767 * struct ifmultiaddr * between calls. It may also 2768 * lose a race while we check if the membership 2769 * already exists. 2770 */ 2771 NET_EPOCH_ENTER(et); 2772 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2773 NET_EPOCH_EXIT(et); 2774 if (ifma != NULL) 2775 error = EADDRINUSE; 2776 else 2777 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2778 } else { 2779 error = if_delmulti(ifp, &ifr->ifr_addr); 2780 } 2781 if (error == 0) 2782 getmicrotime(&ifp->if_lastchange); 2783 break; 2784 2785 case SIOCSIFPHYADDR: 2786 case SIOCDIFPHYADDR: 2787 #ifdef INET6 2788 case SIOCSIFPHYADDR_IN6: 2789 #endif 2790 case SIOCSIFMEDIA: 2791 case SIOCSIFGENERIC: 2792 error = priv_check(td, PRIV_NET_HWIOCTL); 2793 if (error) 2794 return (error); 2795 if (ifp->if_ioctl == NULL) 2796 return (EOPNOTSUPP); 2797 error = (*ifp->if_ioctl)(ifp, cmd, data); 2798 if (error == 0) 2799 getmicrotime(&ifp->if_lastchange); 2800 break; 2801 2802 case SIOCGIFSTATUS: 2803 case SIOCGIFPSRCADDR: 2804 case SIOCGIFPDSTADDR: 2805 case SIOCGIFMEDIA: 2806 case SIOCGIFXMEDIA: 2807 case SIOCGIFGENERIC: 2808 case SIOCGIFRSSKEY: 2809 case SIOCGIFRSSHASH: 2810 case SIOCGIFDOWNREASON: 2811 if (ifp->if_ioctl == NULL) 2812 return (EOPNOTSUPP); 2813 error = (*ifp->if_ioctl)(ifp, cmd, data); 2814 break; 2815 2816 case SIOCSIFLLADDR: 2817 error = priv_check(td, PRIV_NET_SETLLADDR); 2818 if (error) 2819 return (error); 2820 error = if_setlladdr(ifp, 2821 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2822 break; 2823 2824 case SIOCGHWADDR: 2825 error = if_gethwaddr(ifp, ifr); 2826 break; 2827 2828 case SIOCAIFGROUP: 2829 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2830 if (error) 2831 return (error); 2832 error = if_addgroup(ifp, 2833 ((struct ifgroupreq *)data)->ifgr_group); 2834 if (error != 0) 2835 return (error); 2836 break; 2837 2838 case SIOCGIFGROUP: 2839 { 2840 struct epoch_tracker et; 2841 2842 NET_EPOCH_ENTER(et); 2843 error = if_getgroup((struct ifgroupreq *)data, ifp); 2844 NET_EPOCH_EXIT(et); 2845 break; 2846 } 2847 2848 case SIOCDIFGROUP: 2849 error = priv_check(td, PRIV_NET_DELIFGROUP); 2850 if (error) 2851 return (error); 2852 error = if_delgroup(ifp, 2853 ((struct ifgroupreq *)data)->ifgr_group); 2854 if (error != 0) 2855 return (error); 2856 break; 2857 2858 default: 2859 error = ENOIOCTL; 2860 break; 2861 } 2862 return (error); 2863 } 2864 2865 /* 2866 * Interface ioctls. 2867 */ 2868 int 2869 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2870 { 2871 #ifdef COMPAT_FREEBSD32 2872 union { 2873 struct ifconf ifc; 2874 struct ifdrv ifd; 2875 struct ifgroupreq ifgr; 2876 struct ifmediareq ifmr; 2877 } thunk; 2878 u_long saved_cmd; 2879 struct ifconf32 *ifc32; 2880 struct ifdrv32 *ifd32; 2881 struct ifgroupreq32 *ifgr32; 2882 struct ifmediareq32 *ifmr32; 2883 #endif 2884 struct ifnet *ifp; 2885 struct ifreq *ifr; 2886 int error; 2887 int oif_flags; 2888 #ifdef VIMAGE 2889 bool shutdown; 2890 #endif 2891 2892 CURVNET_SET(so->so_vnet); 2893 #ifdef VIMAGE 2894 /* Make sure the VNET is stable. */ 2895 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2896 if (shutdown) { 2897 CURVNET_RESTORE(); 2898 return (EBUSY); 2899 } 2900 #endif 2901 2902 #ifdef COMPAT_FREEBSD32 2903 saved_cmd = cmd; 2904 switch (cmd) { 2905 case SIOCGIFCONF32: 2906 ifc32 = (struct ifconf32 *)data; 2907 thunk.ifc.ifc_len = ifc32->ifc_len; 2908 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2909 data = (caddr_t)&thunk.ifc; 2910 cmd = SIOCGIFCONF; 2911 break; 2912 case SIOCGDRVSPEC32: 2913 case SIOCSDRVSPEC32: 2914 ifd32 = (struct ifdrv32 *)data; 2915 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2916 sizeof(thunk.ifd.ifd_name)); 2917 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2918 thunk.ifd.ifd_len = ifd32->ifd_len; 2919 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2920 data = (caddr_t)&thunk.ifd; 2921 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2922 break; 2923 case SIOCAIFGROUP32: 2924 case SIOCGIFGROUP32: 2925 case SIOCDIFGROUP32: 2926 case SIOCGIFGMEMB32: 2927 ifgr32 = (struct ifgroupreq32 *)data; 2928 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2929 sizeof(thunk.ifgr.ifgr_name)); 2930 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2931 switch (cmd) { 2932 case SIOCAIFGROUP32: 2933 case SIOCDIFGROUP32: 2934 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2935 sizeof(thunk.ifgr.ifgr_group)); 2936 break; 2937 case SIOCGIFGROUP32: 2938 case SIOCGIFGMEMB32: 2939 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2940 break; 2941 } 2942 data = (caddr_t)&thunk.ifgr; 2943 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2944 break; 2945 case SIOCGIFMEDIA32: 2946 case SIOCGIFXMEDIA32: 2947 ifmr32 = (struct ifmediareq32 *)data; 2948 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2949 sizeof(thunk.ifmr.ifm_name)); 2950 thunk.ifmr.ifm_current = ifmr32->ifm_current; 2951 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 2952 thunk.ifmr.ifm_status = ifmr32->ifm_status; 2953 thunk.ifmr.ifm_active = ifmr32->ifm_active; 2954 thunk.ifmr.ifm_count = ifmr32->ifm_count; 2955 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 2956 data = (caddr_t)&thunk.ifmr; 2957 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 2958 break; 2959 } 2960 #endif 2961 2962 switch (cmd) { 2963 case SIOCGIFCONF: 2964 error = ifconf(cmd, data); 2965 goto out_noref; 2966 } 2967 2968 ifr = (struct ifreq *)data; 2969 switch (cmd) { 2970 #ifdef VIMAGE 2971 case SIOCSIFRVNET: 2972 error = priv_check(td, PRIV_NET_SETIFVNET); 2973 if (error == 0) 2974 error = if_vmove_reclaim(td, ifr->ifr_name, 2975 ifr->ifr_jid); 2976 goto out_noref; 2977 #endif 2978 case SIOCIFCREATE: 2979 case SIOCIFCREATE2: 2980 error = priv_check(td, PRIV_NET_IFCREATE); 2981 if (error == 0) 2982 error = if_clone_create(ifr->ifr_name, 2983 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 2984 ifr_data_get_ptr(ifr) : NULL); 2985 goto out_noref; 2986 case SIOCIFDESTROY: 2987 error = priv_check(td, PRIV_NET_IFDESTROY); 2988 2989 if (error == 0) { 2990 sx_xlock(&ifnet_detach_sxlock); 2991 error = if_clone_destroy(ifr->ifr_name); 2992 sx_xunlock(&ifnet_detach_sxlock); 2993 } 2994 goto out_noref; 2995 2996 case SIOCIFGCLONERS: 2997 error = if_clone_list((struct if_clonereq *)data); 2998 goto out_noref; 2999 3000 case SIOCGIFGMEMB: 3001 error = if_getgroupmembers((struct ifgroupreq *)data); 3002 goto out_noref; 3003 3004 #if defined(INET) || defined(INET6) 3005 case SIOCSVH: 3006 case SIOCGVH: 3007 if (carp_ioctl_p == NULL) 3008 error = EPROTONOSUPPORT; 3009 else 3010 error = (*carp_ioctl_p)(ifr, cmd, td); 3011 goto out_noref; 3012 #endif 3013 } 3014 3015 ifp = ifunit_ref(ifr->ifr_name); 3016 if (ifp == NULL) { 3017 error = ENXIO; 3018 goto out_noref; 3019 } 3020 3021 error = ifhwioctl(cmd, ifp, data, td); 3022 if (error != ENOIOCTL) 3023 goto out_ref; 3024 3025 oif_flags = ifp->if_flags; 3026 if (so->so_proto == NULL) { 3027 error = EOPNOTSUPP; 3028 goto out_ref; 3029 } 3030 3031 /* 3032 * Pass the request on to the socket control method, and if the 3033 * latter returns EOPNOTSUPP, directly to the interface. 3034 * 3035 * Make an exception for the legacy SIOCSIF* requests. Drivers 3036 * trust SIOCSIFADDR et al to come from an already privileged 3037 * layer, and do not perform any credentials checks or input 3038 * validation. 3039 */ 3040 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 3041 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3042 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3043 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3044 error = (*ifp->if_ioctl)(ifp, cmd, data); 3045 3046 if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP)) 3047 if_up(ifp); 3048 out_ref: 3049 if_rele(ifp); 3050 out_noref: 3051 CURVNET_RESTORE(); 3052 #ifdef COMPAT_FREEBSD32 3053 if (error != 0) 3054 return (error); 3055 switch (saved_cmd) { 3056 case SIOCGIFCONF32: 3057 ifc32->ifc_len = thunk.ifc.ifc_len; 3058 break; 3059 case SIOCGDRVSPEC32: 3060 /* 3061 * SIOCGDRVSPEC is IOWR, but nothing actually touches 3062 * the struct so just assert that ifd_len (the only 3063 * field it might make sense to update) hasn't 3064 * changed. 3065 */ 3066 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 3067 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 3068 thunk.ifd.ifd_len)); 3069 break; 3070 case SIOCGIFGROUP32: 3071 case SIOCGIFGMEMB32: 3072 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 3073 break; 3074 case SIOCGIFMEDIA32: 3075 case SIOCGIFXMEDIA32: 3076 ifmr32->ifm_current = thunk.ifmr.ifm_current; 3077 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 3078 ifmr32->ifm_status = thunk.ifmr.ifm_status; 3079 ifmr32->ifm_active = thunk.ifmr.ifm_active; 3080 ifmr32->ifm_count = thunk.ifmr.ifm_count; 3081 break; 3082 } 3083 #endif 3084 return (error); 3085 } 3086 3087 int 3088 if_rename(struct ifnet *ifp, char *new_name) 3089 { 3090 struct ifaddr *ifa; 3091 struct sockaddr_dl *sdl; 3092 size_t namelen, onamelen; 3093 char old_name[IFNAMSIZ]; 3094 char strbuf[IFNAMSIZ + 8]; 3095 3096 if (new_name[0] == '\0') 3097 return (EINVAL); 3098 if (strcmp(new_name, ifp->if_xname) == 0) 3099 return (0); 3100 if (ifunit(new_name) != NULL) 3101 return (EEXIST); 3102 3103 /* 3104 * XXX: Locking. Nothing else seems to lock if_flags, 3105 * and there are numerous other races with the 3106 * ifunit() checks not being atomic with namespace 3107 * changes (renames, vmoves, if_attach, etc). 3108 */ 3109 ifp->if_flags |= IFF_RENAMING; 3110 3111 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 3112 3113 if_printf(ifp, "changing name to '%s'\n", new_name); 3114 3115 IF_ADDR_WLOCK(ifp); 3116 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 3117 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 3118 ifa = ifp->if_addr; 3119 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3120 namelen = strlen(new_name); 3121 onamelen = sdl->sdl_nlen; 3122 /* 3123 * Move the address if needed. This is safe because we 3124 * allocate space for a name of length IFNAMSIZ when we 3125 * create this in if_attach(). 3126 */ 3127 if (namelen != onamelen) { 3128 bcopy(sdl->sdl_data + onamelen, 3129 sdl->sdl_data + namelen, sdl->sdl_alen); 3130 } 3131 bcopy(new_name, sdl->sdl_data, namelen); 3132 sdl->sdl_nlen = namelen; 3133 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 3134 bzero(sdl->sdl_data, onamelen); 3135 while (namelen != 0) 3136 sdl->sdl_data[--namelen] = 0xff; 3137 IF_ADDR_WUNLOCK(ifp); 3138 3139 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 3140 3141 ifp->if_flags &= ~IFF_RENAMING; 3142 3143 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 3144 devctl_notify("IFNET", old_name, "RENAME", strbuf); 3145 3146 return (0); 3147 } 3148 3149 /* 3150 * The code common to handling reference counted flags, 3151 * e.g., in ifpromisc() and if_allmulti(). 3152 * The "pflag" argument can specify a permanent mode flag to check, 3153 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3154 * 3155 * Only to be used on stack-owned flags, not driver-owned flags. 3156 */ 3157 static int 3158 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3159 { 3160 struct ifreq ifr; 3161 int error; 3162 int oldflags, oldcount; 3163 3164 /* Sanity checks to catch programming errors */ 3165 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3166 ("%s: setting driver-owned flag %d", __func__, flag)); 3167 3168 if (onswitch) 3169 KASSERT(*refcount >= 0, 3170 ("%s: increment negative refcount %d for flag %d", 3171 __func__, *refcount, flag)); 3172 else 3173 KASSERT(*refcount > 0, 3174 ("%s: decrement non-positive refcount %d for flag %d", 3175 __func__, *refcount, flag)); 3176 3177 /* In case this mode is permanent, just touch refcount */ 3178 if (ifp->if_flags & pflag) { 3179 *refcount += onswitch ? 1 : -1; 3180 return (0); 3181 } 3182 3183 /* Save ifnet parameters for if_ioctl() may fail */ 3184 oldcount = *refcount; 3185 oldflags = ifp->if_flags; 3186 3187 /* 3188 * See if we aren't the only and touching refcount is enough. 3189 * Actually toggle interface flag if we are the first or last. 3190 */ 3191 if (onswitch) { 3192 if ((*refcount)++) 3193 return (0); 3194 ifp->if_flags |= flag; 3195 } else { 3196 if (--(*refcount)) 3197 return (0); 3198 ifp->if_flags &= ~flag; 3199 } 3200 3201 /* Call down the driver since we've changed interface flags */ 3202 if (ifp->if_ioctl == NULL) { 3203 error = EOPNOTSUPP; 3204 goto recover; 3205 } 3206 ifr.ifr_flags = ifp->if_flags & 0xffff; 3207 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3208 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3209 if (error) 3210 goto recover; 3211 /* Notify userland that interface flags have changed */ 3212 rt_ifmsg(ifp, flag); 3213 return (0); 3214 3215 recover: 3216 /* Recover after driver error */ 3217 *refcount = oldcount; 3218 ifp->if_flags = oldflags; 3219 return (error); 3220 } 3221 3222 /* 3223 * Set/clear promiscuous mode on interface ifp based on the truth value 3224 * of pswitch. The calls are reference counted so that only the first 3225 * "on" request actually has an effect, as does the final "off" request. 3226 * Results are undefined if the "off" and "on" requests are not matched. 3227 */ 3228 int 3229 ifpromisc(struct ifnet *ifp, int pswitch) 3230 { 3231 int error; 3232 int oldflags = ifp->if_flags; 3233 3234 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3235 &ifp->if_pcount, pswitch); 3236 /* If promiscuous mode status has changed, log a message */ 3237 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3238 log_promisc_mode_change) 3239 if_printf(ifp, "promiscuous mode %s\n", 3240 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3241 return (error); 3242 } 3243 3244 /* 3245 * Return interface configuration 3246 * of system. List may be used 3247 * in later ioctl's (above) to get 3248 * other information. 3249 */ 3250 /*ARGSUSED*/ 3251 static int 3252 ifconf(u_long cmd, caddr_t data) 3253 { 3254 struct ifconf *ifc = (struct ifconf *)data; 3255 struct ifnet *ifp; 3256 struct ifaddr *ifa; 3257 struct ifreq ifr; 3258 struct sbuf *sb; 3259 int error, full = 0, valid_len, max_len; 3260 3261 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3262 max_len = maxphys - 1; 3263 3264 /* Prevent hostile input from being able to crash the system */ 3265 if (ifc->ifc_len <= 0) 3266 return (EINVAL); 3267 3268 again: 3269 if (ifc->ifc_len <= max_len) { 3270 max_len = ifc->ifc_len; 3271 full = 1; 3272 } 3273 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3274 max_len = 0; 3275 valid_len = 0; 3276 3277 IFNET_RLOCK(); 3278 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3279 struct epoch_tracker et; 3280 int addrs; 3281 3282 /* 3283 * Zero the ifr to make sure we don't disclose the contents 3284 * of the stack. 3285 */ 3286 memset(&ifr, 0, sizeof(ifr)); 3287 3288 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3289 >= sizeof(ifr.ifr_name)) { 3290 sbuf_delete(sb); 3291 IFNET_RUNLOCK(); 3292 return (ENAMETOOLONG); 3293 } 3294 3295 addrs = 0; 3296 NET_EPOCH_ENTER(et); 3297 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3298 struct sockaddr *sa = ifa->ifa_addr; 3299 3300 if (prison_if(curthread->td_ucred, sa) != 0) 3301 continue; 3302 addrs++; 3303 if (sa->sa_len <= sizeof(*sa)) { 3304 if (sa->sa_len < sizeof(*sa)) { 3305 memset(&ifr.ifr_ifru.ifru_addr, 0, 3306 sizeof(ifr.ifr_ifru.ifru_addr)); 3307 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3308 sa->sa_len); 3309 } else 3310 ifr.ifr_ifru.ifru_addr = *sa; 3311 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3312 max_len += sizeof(ifr); 3313 } else { 3314 sbuf_bcat(sb, &ifr, 3315 offsetof(struct ifreq, ifr_addr)); 3316 max_len += offsetof(struct ifreq, ifr_addr); 3317 sbuf_bcat(sb, sa, sa->sa_len); 3318 max_len += sa->sa_len; 3319 } 3320 3321 if (sbuf_error(sb) == 0) 3322 valid_len = sbuf_len(sb); 3323 } 3324 NET_EPOCH_EXIT(et); 3325 if (addrs == 0) { 3326 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3327 max_len += sizeof(ifr); 3328 3329 if (sbuf_error(sb) == 0) 3330 valid_len = sbuf_len(sb); 3331 } 3332 } 3333 IFNET_RUNLOCK(); 3334 3335 /* 3336 * If we didn't allocate enough space (uncommon), try again. If 3337 * we have already allocated as much space as we are allowed, 3338 * return what we've got. 3339 */ 3340 if (valid_len != max_len && !full) { 3341 sbuf_delete(sb); 3342 goto again; 3343 } 3344 3345 ifc->ifc_len = valid_len; 3346 sbuf_finish(sb); 3347 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3348 sbuf_delete(sb); 3349 return (error); 3350 } 3351 3352 /* 3353 * Just like ifpromisc(), but for all-multicast-reception mode. 3354 */ 3355 int 3356 if_allmulti(struct ifnet *ifp, int onswitch) 3357 { 3358 3359 return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount, 3360 onswitch)); 3361 } 3362 3363 struct ifmultiaddr * 3364 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3365 { 3366 struct ifmultiaddr *ifma; 3367 3368 IF_ADDR_LOCK_ASSERT(ifp); 3369 3370 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3371 if (sa->sa_family == AF_LINK) { 3372 if (sa_dl_equal(ifma->ifma_addr, sa)) 3373 break; 3374 } else { 3375 if (sa_equal(ifma->ifma_addr, sa)) 3376 break; 3377 } 3378 } 3379 3380 return ifma; 3381 } 3382 3383 /* 3384 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3385 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3386 * the ifnet multicast address list here, so the caller must do that and 3387 * other setup work (such as notifying the device driver). The reference 3388 * count is initialized to 1. 3389 */ 3390 static struct ifmultiaddr * 3391 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3392 int mflags) 3393 { 3394 struct ifmultiaddr *ifma; 3395 struct sockaddr *dupsa; 3396 3397 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3398 M_ZERO); 3399 if (ifma == NULL) 3400 return (NULL); 3401 3402 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3403 if (dupsa == NULL) { 3404 free(ifma, M_IFMADDR); 3405 return (NULL); 3406 } 3407 bcopy(sa, dupsa, sa->sa_len); 3408 ifma->ifma_addr = dupsa; 3409 3410 ifma->ifma_ifp = ifp; 3411 ifma->ifma_refcount = 1; 3412 ifma->ifma_protospec = NULL; 3413 3414 if (llsa == NULL) { 3415 ifma->ifma_lladdr = NULL; 3416 return (ifma); 3417 } 3418 3419 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3420 if (dupsa == NULL) { 3421 free(ifma->ifma_addr, M_IFMADDR); 3422 free(ifma, M_IFMADDR); 3423 return (NULL); 3424 } 3425 bcopy(llsa, dupsa, llsa->sa_len); 3426 ifma->ifma_lladdr = dupsa; 3427 3428 return (ifma); 3429 } 3430 3431 /* 3432 * if_freemulti: free ifmultiaddr structure and possibly attached related 3433 * addresses. The caller is responsible for implementing reference 3434 * counting, notifying the driver, handling routing messages, and releasing 3435 * any dependent link layer state. 3436 */ 3437 #ifdef MCAST_VERBOSE 3438 extern void kdb_backtrace(void); 3439 #endif 3440 static void 3441 if_freemulti_internal(struct ifmultiaddr *ifma) 3442 { 3443 3444 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3445 ifma->ifma_refcount)); 3446 3447 if (ifma->ifma_lladdr != NULL) 3448 free(ifma->ifma_lladdr, M_IFMADDR); 3449 #ifdef MCAST_VERBOSE 3450 kdb_backtrace(); 3451 printf("%s freeing ifma: %p\n", __func__, ifma); 3452 #endif 3453 free(ifma->ifma_addr, M_IFMADDR); 3454 free(ifma, M_IFMADDR); 3455 } 3456 3457 static void 3458 if_destroymulti(epoch_context_t ctx) 3459 { 3460 struct ifmultiaddr *ifma; 3461 3462 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3463 if_freemulti_internal(ifma); 3464 } 3465 3466 void 3467 if_freemulti(struct ifmultiaddr *ifma) 3468 { 3469 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3470 ifma->ifma_refcount)); 3471 3472 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3473 } 3474 3475 /* 3476 * Register an additional multicast address with a network interface. 3477 * 3478 * - If the address is already present, bump the reference count on the 3479 * address and return. 3480 * - If the address is not link-layer, look up a link layer address. 3481 * - Allocate address structures for one or both addresses, and attach to the 3482 * multicast address list on the interface. If automatically adding a link 3483 * layer address, the protocol address will own a reference to the link 3484 * layer address, to be freed when it is freed. 3485 * - Notify the network device driver of an addition to the multicast address 3486 * list. 3487 * 3488 * 'sa' points to caller-owned memory with the desired multicast address. 3489 * 3490 * 'retifma' will be used to return a pointer to the resulting multicast 3491 * address reference, if desired. 3492 */ 3493 int 3494 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3495 struct ifmultiaddr **retifma) 3496 { 3497 struct ifmultiaddr *ifma, *ll_ifma; 3498 struct sockaddr *llsa; 3499 struct sockaddr_dl sdl; 3500 int error; 3501 3502 #ifdef INET 3503 IN_MULTI_LIST_UNLOCK_ASSERT(); 3504 #endif 3505 #ifdef INET6 3506 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3507 #endif 3508 /* 3509 * If the address is already present, return a new reference to it; 3510 * otherwise, allocate storage and set up a new address. 3511 */ 3512 IF_ADDR_WLOCK(ifp); 3513 ifma = if_findmulti(ifp, sa); 3514 if (ifma != NULL) { 3515 ifma->ifma_refcount++; 3516 if (retifma != NULL) 3517 *retifma = ifma; 3518 IF_ADDR_WUNLOCK(ifp); 3519 return (0); 3520 } 3521 3522 /* 3523 * The address isn't already present; resolve the protocol address 3524 * into a link layer address, and then look that up, bump its 3525 * refcount or allocate an ifma for that also. 3526 * Most link layer resolving functions returns address data which 3527 * fits inside default sockaddr_dl structure. However callback 3528 * can allocate another sockaddr structure, in that case we need to 3529 * free it later. 3530 */ 3531 llsa = NULL; 3532 ll_ifma = NULL; 3533 if (ifp->if_resolvemulti != NULL) { 3534 /* Provide called function with buffer size information */ 3535 sdl.sdl_len = sizeof(sdl); 3536 llsa = (struct sockaddr *)&sdl; 3537 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3538 if (error) 3539 goto unlock_out; 3540 } 3541 3542 /* 3543 * Allocate the new address. Don't hook it up yet, as we may also 3544 * need to allocate a link layer multicast address. 3545 */ 3546 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3547 if (ifma == NULL) { 3548 error = ENOMEM; 3549 goto free_llsa_out; 3550 } 3551 3552 /* 3553 * If a link layer address is found, we'll need to see if it's 3554 * already present in the address list, or allocate is as well. 3555 * When this block finishes, the link layer address will be on the 3556 * list. 3557 */ 3558 if (llsa != NULL) { 3559 ll_ifma = if_findmulti(ifp, llsa); 3560 if (ll_ifma == NULL) { 3561 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3562 if (ll_ifma == NULL) { 3563 --ifma->ifma_refcount; 3564 if_freemulti(ifma); 3565 error = ENOMEM; 3566 goto free_llsa_out; 3567 } 3568 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3569 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3570 ifma_link); 3571 } else 3572 ll_ifma->ifma_refcount++; 3573 ifma->ifma_llifma = ll_ifma; 3574 } 3575 3576 /* 3577 * We now have a new multicast address, ifma, and possibly a new or 3578 * referenced link layer address. Add the primary address to the 3579 * ifnet address list. 3580 */ 3581 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3582 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3583 3584 if (retifma != NULL) 3585 *retifma = ifma; 3586 3587 /* 3588 * Must generate the message while holding the lock so that 'ifma' 3589 * pointer is still valid. 3590 */ 3591 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3592 IF_ADDR_WUNLOCK(ifp); 3593 3594 /* 3595 * We are certain we have added something, so call down to the 3596 * interface to let them know about it. 3597 */ 3598 if (ifp->if_ioctl != NULL) { 3599 if (THREAD_CAN_SLEEP()) 3600 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3601 else 3602 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3603 } 3604 3605 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3606 link_free_sdl(llsa); 3607 3608 return (0); 3609 3610 free_llsa_out: 3611 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3612 link_free_sdl(llsa); 3613 3614 unlock_out: 3615 IF_ADDR_WUNLOCK(ifp); 3616 return (error); 3617 } 3618 3619 static void 3620 if_siocaddmulti(void *arg, int pending) 3621 { 3622 struct ifnet *ifp; 3623 3624 ifp = arg; 3625 #ifdef DIAGNOSTIC 3626 if (pending > 1) 3627 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3628 #endif 3629 CURVNET_SET(ifp->if_vnet); 3630 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3631 CURVNET_RESTORE(); 3632 } 3633 3634 /* 3635 * Delete a multicast group membership by network-layer group address. 3636 * 3637 * Returns ENOENT if the entry could not be found. If ifp no longer 3638 * exists, results are undefined. This entry point should only be used 3639 * from subsystems which do appropriate locking to hold ifp for the 3640 * duration of the call. 3641 * Network-layer protocol domains must use if_delmulti_ifma(). 3642 */ 3643 int 3644 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3645 { 3646 struct ifmultiaddr *ifma; 3647 int lastref; 3648 3649 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3650 3651 IF_ADDR_WLOCK(ifp); 3652 lastref = 0; 3653 ifma = if_findmulti(ifp, sa); 3654 if (ifma != NULL) 3655 lastref = if_delmulti_locked(ifp, ifma, 0); 3656 IF_ADDR_WUNLOCK(ifp); 3657 3658 if (ifma == NULL) 3659 return (ENOENT); 3660 3661 if (lastref && ifp->if_ioctl != NULL) { 3662 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3663 } 3664 3665 return (0); 3666 } 3667 3668 /* 3669 * Delete all multicast group membership for an interface. 3670 * Should be used to quickly flush all multicast filters. 3671 */ 3672 void 3673 if_delallmulti(struct ifnet *ifp) 3674 { 3675 struct ifmultiaddr *ifma; 3676 struct ifmultiaddr *next; 3677 3678 IF_ADDR_WLOCK(ifp); 3679 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3680 if_delmulti_locked(ifp, ifma, 0); 3681 IF_ADDR_WUNLOCK(ifp); 3682 } 3683 3684 void 3685 if_delmulti_ifma(struct ifmultiaddr *ifma) 3686 { 3687 if_delmulti_ifma_flags(ifma, 0); 3688 } 3689 3690 /* 3691 * Delete a multicast group membership by group membership pointer. 3692 * Network-layer protocol domains must use this routine. 3693 * 3694 * It is safe to call this routine if the ifp disappeared. 3695 */ 3696 void 3697 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3698 { 3699 struct ifnet *ifp; 3700 int lastref; 3701 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3702 #ifdef INET 3703 IN_MULTI_LIST_UNLOCK_ASSERT(); 3704 #endif 3705 ifp = ifma->ifma_ifp; 3706 #ifdef DIAGNOSTIC 3707 if (ifp == NULL) { 3708 printf("%s: ifma_ifp seems to be detached\n", __func__); 3709 } else { 3710 struct epoch_tracker et; 3711 struct ifnet *oifp; 3712 3713 NET_EPOCH_ENTER(et); 3714 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3715 if (ifp == oifp) 3716 break; 3717 NET_EPOCH_EXIT(et); 3718 if (ifp != oifp) 3719 ifp = NULL; 3720 } 3721 #endif 3722 /* 3723 * If and only if the ifnet instance exists: Acquire the address lock. 3724 */ 3725 if (ifp != NULL) 3726 IF_ADDR_WLOCK(ifp); 3727 3728 lastref = if_delmulti_locked(ifp, ifma, flags); 3729 3730 if (ifp != NULL) { 3731 /* 3732 * If and only if the ifnet instance exists: 3733 * Release the address lock. 3734 * If the group was left: update the hardware hash filter. 3735 */ 3736 IF_ADDR_WUNLOCK(ifp); 3737 if (lastref && ifp->if_ioctl != NULL) { 3738 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3739 } 3740 } 3741 } 3742 3743 /* 3744 * Perform deletion of network-layer and/or link-layer multicast address. 3745 * 3746 * Return 0 if the reference count was decremented. 3747 * Return 1 if the final reference was released, indicating that the 3748 * hardware hash filter should be reprogrammed. 3749 */ 3750 static int 3751 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3752 { 3753 struct ifmultiaddr *ll_ifma; 3754 3755 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3756 KASSERT(ifma->ifma_ifp == ifp, 3757 ("%s: inconsistent ifp %p", __func__, ifp)); 3758 IF_ADDR_WLOCK_ASSERT(ifp); 3759 } 3760 3761 ifp = ifma->ifma_ifp; 3762 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3763 3764 /* 3765 * If the ifnet is detaching, null out references to ifnet, 3766 * so that upper protocol layers will notice, and not attempt 3767 * to obtain locks for an ifnet which no longer exists. The 3768 * routing socket announcement must happen before the ifnet 3769 * instance is detached from the system. 3770 */ 3771 if (detaching) { 3772 #ifdef DIAGNOSTIC 3773 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3774 #endif 3775 /* 3776 * ifp may already be nulled out if we are being reentered 3777 * to delete the ll_ifma. 3778 */ 3779 if (ifp != NULL) { 3780 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3781 ifma->ifma_ifp = NULL; 3782 } 3783 } 3784 3785 if (--ifma->ifma_refcount > 0) 3786 return 0; 3787 3788 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3789 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3790 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3791 } 3792 /* 3793 * If this ifma is a network-layer ifma, a link-layer ifma may 3794 * have been associated with it. Release it first if so. 3795 */ 3796 ll_ifma = ifma->ifma_llifma; 3797 if (ll_ifma != NULL) { 3798 KASSERT(ifma->ifma_lladdr != NULL, 3799 ("%s: llifma w/o lladdr", __func__)); 3800 if (detaching) 3801 ll_ifma->ifma_ifp = NULL; /* XXX */ 3802 if (--ll_ifma->ifma_refcount == 0) { 3803 if (ifp != NULL) { 3804 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3805 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3806 ifma_link); 3807 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3808 } 3809 } 3810 if_freemulti(ll_ifma); 3811 } 3812 } 3813 #ifdef INVARIANTS 3814 if (ifp) { 3815 struct ifmultiaddr *ifmatmp; 3816 3817 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3818 MPASS(ifma != ifmatmp); 3819 } 3820 #endif 3821 if_freemulti(ifma); 3822 /* 3823 * The last reference to this instance of struct ifmultiaddr 3824 * was released; the hardware should be notified of this change. 3825 */ 3826 return 1; 3827 } 3828 3829 /* 3830 * Set the link layer address on an interface. 3831 * 3832 * At this time we only support certain types of interfaces, 3833 * and we don't allow the length of the address to change. 3834 * 3835 * Set noinline to be dtrace-friendly 3836 */ 3837 __noinline int 3838 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3839 { 3840 struct sockaddr_dl *sdl; 3841 struct ifaddr *ifa; 3842 struct ifreq ifr; 3843 3844 ifa = ifp->if_addr; 3845 if (ifa == NULL) 3846 return (EINVAL); 3847 3848 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3849 if (sdl == NULL) 3850 return (EINVAL); 3851 3852 if (len != sdl->sdl_alen) /* don't allow length to change */ 3853 return (EINVAL); 3854 3855 switch (ifp->if_type) { 3856 case IFT_ETHER: 3857 case IFT_XETHER: 3858 case IFT_L2VLAN: 3859 case IFT_BRIDGE: 3860 case IFT_IEEE8023ADLAG: 3861 bcopy(lladdr, LLADDR(sdl), len); 3862 break; 3863 default: 3864 return (ENODEV); 3865 } 3866 3867 /* 3868 * If the interface is already up, we need 3869 * to re-init it in order to reprogram its 3870 * address filter. 3871 */ 3872 if ((ifp->if_flags & IFF_UP) != 0) { 3873 if (ifp->if_ioctl) { 3874 ifp->if_flags &= ~IFF_UP; 3875 ifr.ifr_flags = ifp->if_flags & 0xffff; 3876 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3877 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3878 ifp->if_flags |= IFF_UP; 3879 ifr.ifr_flags = ifp->if_flags & 0xffff; 3880 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3881 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3882 } 3883 } 3884 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3885 3886 return (0); 3887 } 3888 3889 /* 3890 * Compat function for handling basic encapsulation requests. 3891 * Not converted stacks (FDDI, IB, ..) supports traditional 3892 * output model: ARP (and other similar L2 protocols) are handled 3893 * inside output routine, arpresolve/nd6_resolve() returns MAC 3894 * address instead of full prepend. 3895 * 3896 * This function creates calculated header==MAC for IPv4/IPv6 and 3897 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3898 * address families. 3899 */ 3900 static int 3901 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3902 { 3903 if (req->rtype != IFENCAP_LL) 3904 return (EOPNOTSUPP); 3905 3906 if (req->bufsize < req->lladdr_len) 3907 return (ENOMEM); 3908 3909 switch (req->family) { 3910 case AF_INET: 3911 case AF_INET6: 3912 break; 3913 default: 3914 return (EAFNOSUPPORT); 3915 } 3916 3917 /* Copy lladdr to storage as is */ 3918 memmove(req->buf, req->lladdr, req->lladdr_len); 3919 req->bufsize = req->lladdr_len; 3920 req->lladdr_off = 0; 3921 3922 return (0); 3923 } 3924 3925 /* 3926 * Tunnel interfaces can nest, also they may cause infinite recursion 3927 * calls when misconfigured. We'll prevent this by detecting loops. 3928 * High nesting level may cause stack exhaustion. We'll prevent this 3929 * by introducing upper limit. 3930 * 3931 * Return 0, if tunnel nesting count is equal or less than limit. 3932 */ 3933 int 3934 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3935 int limit) 3936 { 3937 struct m_tag *mtag; 3938 int count; 3939 3940 count = 1; 3941 mtag = NULL; 3942 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3943 if (*(struct ifnet **)(mtag + 1) == ifp) { 3944 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3945 return (EIO); 3946 } 3947 count++; 3948 } 3949 if (count > limit) { 3950 log(LOG_NOTICE, 3951 "%s: if_output recursively called too many times(%d)\n", 3952 if_name(ifp), count); 3953 return (EIO); 3954 } 3955 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3956 if (mtag == NULL) 3957 return (ENOMEM); 3958 *(struct ifnet **)(mtag + 1) = ifp; 3959 m_tag_prepend(m, mtag); 3960 return (0); 3961 } 3962 3963 /* 3964 * Get the link layer address that was read from the hardware at attach. 3965 * 3966 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3967 * their component interfaces as IFT_IEEE8023ADLAG. 3968 */ 3969 int 3970 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3971 { 3972 if (ifp->if_hw_addr == NULL) 3973 return (ENODEV); 3974 3975 switch (ifp->if_type) { 3976 case IFT_ETHER: 3977 case IFT_IEEE8023ADLAG: 3978 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3979 return (0); 3980 default: 3981 return (ENODEV); 3982 } 3983 } 3984 3985 /* 3986 * The name argument must be a pointer to storage which will last as 3987 * long as the interface does. For physical devices, the result of 3988 * device_get_name(dev) is a good choice and for pseudo-devices a 3989 * static string works well. 3990 */ 3991 void 3992 if_initname(struct ifnet *ifp, const char *name, int unit) 3993 { 3994 ifp->if_dname = name; 3995 ifp->if_dunit = unit; 3996 if (unit != IF_DUNIT_NONE) 3997 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3998 else 3999 strlcpy(ifp->if_xname, name, IFNAMSIZ); 4000 } 4001 4002 static int 4003 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 4004 { 4005 char if_fmt[256]; 4006 4007 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 4008 vlog(pri, if_fmt, ap); 4009 return (0); 4010 } 4011 4012 4013 int 4014 if_printf(struct ifnet *ifp, const char *fmt, ...) 4015 { 4016 va_list ap; 4017 4018 va_start(ap, fmt); 4019 if_vlog(ifp, LOG_INFO, fmt, ap); 4020 va_end(ap); 4021 return (0); 4022 } 4023 4024 int 4025 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 4026 { 4027 va_list ap; 4028 4029 va_start(ap, fmt); 4030 if_vlog(ifp, pri, fmt, ap); 4031 va_end(ap); 4032 return (0); 4033 } 4034 4035 void 4036 if_start(struct ifnet *ifp) 4037 { 4038 4039 (*(ifp)->if_start)(ifp); 4040 } 4041 4042 /* 4043 * Backwards compatibility interface for drivers 4044 * that have not implemented it 4045 */ 4046 static int 4047 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 4048 { 4049 int error; 4050 4051 IFQ_HANDOFF(ifp, m, error); 4052 return (error); 4053 } 4054 4055 static void 4056 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4057 { 4058 m_freem(m); 4059 } 4060 4061 int 4062 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4063 { 4064 int active = 0; 4065 4066 IF_LOCK(ifq); 4067 if (_IF_QFULL(ifq)) { 4068 IF_UNLOCK(ifq); 4069 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4070 m_freem(m); 4071 return (0); 4072 } 4073 if (ifp != NULL) { 4074 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4075 if (m->m_flags & (M_BCAST|M_MCAST)) 4076 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4077 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4078 } 4079 _IF_ENQUEUE(ifq, m); 4080 IF_UNLOCK(ifq); 4081 if (ifp != NULL && !active) 4082 (*(ifp)->if_start)(ifp); 4083 return (1); 4084 } 4085 4086 void 4087 if_register_com_alloc(u_char type, 4088 if_com_alloc_t *a, if_com_free_t *f) 4089 { 4090 4091 KASSERT(if_com_alloc[type] == NULL, 4092 ("if_register_com_alloc: %d already registered", type)); 4093 KASSERT(if_com_free[type] == NULL, 4094 ("if_register_com_alloc: %d free already registered", type)); 4095 4096 if_com_alloc[type] = a; 4097 if_com_free[type] = f; 4098 } 4099 4100 void 4101 if_deregister_com_alloc(u_char type) 4102 { 4103 4104 KASSERT(if_com_alloc[type] != NULL, 4105 ("if_deregister_com_alloc: %d not registered", type)); 4106 KASSERT(if_com_free[type] != NULL, 4107 ("if_deregister_com_alloc: %d free not registered", type)); 4108 4109 /* 4110 * Ensure all pending EPOCH(9) callbacks have been executed. This 4111 * fixes issues about late invocation of if_destroy(), which leads 4112 * to memory leak from if_com_alloc[type] allocated if_l2com. 4113 */ 4114 NET_EPOCH_DRAIN_CALLBACKS(); 4115 4116 if_com_alloc[type] = NULL; 4117 if_com_free[type] = NULL; 4118 } 4119 4120 /* API for driver access to network stack owned ifnet.*/ 4121 uint64_t 4122 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4123 { 4124 uint64_t oldbrate; 4125 4126 oldbrate = ifp->if_baudrate; 4127 ifp->if_baudrate = baudrate; 4128 return (oldbrate); 4129 } 4130 4131 uint64_t 4132 if_getbaudrate(const if_t ifp) 4133 { 4134 return (ifp->if_baudrate); 4135 } 4136 4137 int 4138 if_setcapabilities(if_t ifp, int capabilities) 4139 { 4140 ifp->if_capabilities = capabilities; 4141 return (0); 4142 } 4143 4144 int 4145 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4146 { 4147 ifp->if_capabilities &= ~clearbit; 4148 ifp->if_capabilities |= setbit; 4149 return (0); 4150 } 4151 4152 int 4153 if_getcapabilities(const if_t ifp) 4154 { 4155 return (ifp->if_capabilities); 4156 } 4157 4158 int 4159 if_setcapenable(if_t ifp, int capabilities) 4160 { 4161 ifp->if_capenable = capabilities; 4162 return (0); 4163 } 4164 4165 int 4166 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4167 { 4168 ifp->if_capenable &= ~clearcap; 4169 ifp->if_capenable |= setcap; 4170 return (0); 4171 } 4172 4173 int 4174 if_setcapabilities2(if_t ifp, int capabilities) 4175 { 4176 ifp->if_capabilities2 = capabilities; 4177 return (0); 4178 } 4179 4180 int 4181 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit) 4182 { 4183 ifp->if_capabilities2 &= ~clearbit; 4184 ifp->if_capabilities2 |= setbit; 4185 return (0); 4186 } 4187 4188 int 4189 if_getcapabilities2(const if_t ifp) 4190 { 4191 return (ifp->if_capabilities2); 4192 } 4193 4194 int 4195 if_setcapenable2(if_t ifp, int capabilities2) 4196 { 4197 ifp->if_capenable2 = capabilities2; 4198 return (0); 4199 } 4200 4201 int 4202 if_setcapenable2bit(if_t ifp, int setcap, int clearcap) 4203 { 4204 ifp->if_capenable2 &= ~clearcap; 4205 ifp->if_capenable2 |= setcap; 4206 return (0); 4207 } 4208 4209 const char * 4210 if_getdname(const if_t ifp) 4211 { 4212 return (ifp->if_dname); 4213 } 4214 4215 void 4216 if_setdname(if_t ifp, const char *dname) 4217 { 4218 ifp->if_dname = dname; 4219 } 4220 4221 const char * 4222 if_name(if_t ifp) 4223 { 4224 return (ifp->if_xname); 4225 } 4226 4227 int 4228 if_setname(if_t ifp, const char *name) 4229 { 4230 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4231 return (ENAMETOOLONG); 4232 strcpy(ifp->if_xname, name); 4233 4234 return (0); 4235 } 4236 4237 int 4238 if_togglecapenable(if_t ifp, int togglecap) 4239 { 4240 ifp->if_capenable ^= togglecap; 4241 return (0); 4242 } 4243 4244 int 4245 if_getcapenable(const if_t ifp) 4246 { 4247 return (ifp->if_capenable); 4248 } 4249 4250 int 4251 if_togglecapenable2(if_t ifp, int togglecap) 4252 { 4253 ifp->if_capenable2 ^= togglecap; 4254 return (0); 4255 } 4256 4257 int 4258 if_getcapenable2(const if_t ifp) 4259 { 4260 return (ifp->if_capenable2); 4261 } 4262 4263 int 4264 if_getdunit(const if_t ifp) 4265 { 4266 return (ifp->if_dunit); 4267 } 4268 4269 int 4270 if_getindex(const if_t ifp) 4271 { 4272 return (ifp->if_index); 4273 } 4274 4275 int 4276 if_getidxgen(const if_t ifp) 4277 { 4278 return (ifp->if_idxgen); 4279 } 4280 4281 const char * 4282 if_getdescr(if_t ifp) 4283 { 4284 return (ifp->if_description); 4285 } 4286 4287 void 4288 if_setdescr(if_t ifp, char *descrbuf) 4289 { 4290 sx_xlock(&ifdescr_sx); 4291 char *odescrbuf = ifp->if_description; 4292 ifp->if_description = descrbuf; 4293 sx_xunlock(&ifdescr_sx); 4294 4295 if_freedescr(odescrbuf); 4296 } 4297 4298 char * 4299 if_allocdescr(size_t sz, int malloc_flag) 4300 { 4301 malloc_flag &= (M_WAITOK | M_NOWAIT); 4302 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4303 } 4304 4305 void 4306 if_freedescr(char *descrbuf) 4307 { 4308 free(descrbuf, M_IFDESCR); 4309 } 4310 4311 int 4312 if_getalloctype(const if_t ifp) 4313 { 4314 return (ifp->if_alloctype); 4315 } 4316 4317 void 4318 if_setlastchange(if_t ifp) 4319 { 4320 getmicrotime(&ifp->if_lastchange); 4321 } 4322 4323 /* 4324 * This is largely undesirable because it ties ifnet to a device, but does 4325 * provide flexiblity for an embedded product vendor. Should be used with 4326 * the understanding that it violates the interface boundaries, and should be 4327 * a last resort only. 4328 */ 4329 int 4330 if_setdev(if_t ifp, void *dev) 4331 { 4332 return (0); 4333 } 4334 4335 int 4336 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4337 { 4338 ifp->if_drv_flags &= ~clear_flags; 4339 ifp->if_drv_flags |= set_flags; 4340 4341 return (0); 4342 } 4343 4344 int 4345 if_getdrvflags(const if_t ifp) 4346 { 4347 return (ifp->if_drv_flags); 4348 } 4349 4350 int 4351 if_setdrvflags(if_t ifp, int flags) 4352 { 4353 ifp->if_drv_flags = flags; 4354 return (0); 4355 } 4356 4357 int 4358 if_setflags(if_t ifp, int flags) 4359 { 4360 ifp->if_flags = flags; 4361 return (0); 4362 } 4363 4364 int 4365 if_setflagbits(if_t ifp, int set, int clear) 4366 { 4367 ifp->if_flags &= ~clear; 4368 ifp->if_flags |= set; 4369 return (0); 4370 } 4371 4372 int 4373 if_getflags(const if_t ifp) 4374 { 4375 return (ifp->if_flags); 4376 } 4377 4378 int 4379 if_clearhwassist(if_t ifp) 4380 { 4381 ifp->if_hwassist = 0; 4382 return (0); 4383 } 4384 4385 int 4386 if_sethwassistbits(if_t ifp, int toset, int toclear) 4387 { 4388 ifp->if_hwassist &= ~toclear; 4389 ifp->if_hwassist |= toset; 4390 4391 return (0); 4392 } 4393 4394 int 4395 if_sethwassist(if_t ifp, int hwassist_bit) 4396 { 4397 ifp->if_hwassist = hwassist_bit; 4398 return (0); 4399 } 4400 4401 int 4402 if_gethwassist(const if_t ifp) 4403 { 4404 return (ifp->if_hwassist); 4405 } 4406 4407 int 4408 if_togglehwassist(if_t ifp, int toggle_bits) 4409 { 4410 ifp->if_hwassist ^= toggle_bits; 4411 return (0); 4412 } 4413 4414 int 4415 if_setmtu(if_t ifp, int mtu) 4416 { 4417 ifp->if_mtu = mtu; 4418 return (0); 4419 } 4420 4421 void 4422 if_notifymtu(if_t ifp) 4423 { 4424 #ifdef INET6 4425 nd6_setmtu(ifp); 4426 #endif 4427 rt_updatemtu(ifp); 4428 } 4429 4430 int 4431 if_getmtu(const if_t ifp) 4432 { 4433 return (ifp->if_mtu); 4434 } 4435 4436 int 4437 if_getmtu_family(const if_t ifp, int family) 4438 { 4439 struct domain *dp; 4440 4441 SLIST_FOREACH(dp, &domains, dom_next) { 4442 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4443 return (dp->dom_ifmtu(ifp)); 4444 } 4445 4446 return (ifp->if_mtu); 4447 } 4448 4449 void 4450 if_setppromisc(if_t ifp, bool ppromisc) 4451 { 4452 int new_flags; 4453 4454 if (ppromisc) 4455 new_flags = ifp->if_flags | IFF_PPROMISC; 4456 else 4457 new_flags = ifp->if_flags & ~IFF_PPROMISC; 4458 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 4459 if (new_flags & IFF_PPROMISC) 4460 new_flags |= IFF_PROMISC; 4461 /* 4462 * Only unset IFF_PROMISC if there are no more consumers of 4463 * promiscuity, i.e. the ifp->if_pcount refcount is 0. 4464 */ 4465 else if (ifp->if_pcount == 0) 4466 new_flags &= ~IFF_PROMISC; 4467 if (log_promisc_mode_change) 4468 if_printf(ifp, "permanently promiscuous mode %s\n", 4469 ((new_flags & IFF_PPROMISC) ? 4470 "enabled" : "disabled")); 4471 } 4472 ifp->if_flags = new_flags; 4473 } 4474 4475 /* 4476 * Methods for drivers to access interface unicast and multicast 4477 * link level addresses. Driver shall not know 'struct ifaddr' neither 4478 * 'struct ifmultiaddr'. 4479 */ 4480 u_int 4481 if_lladdr_count(if_t ifp) 4482 { 4483 struct epoch_tracker et; 4484 struct ifaddr *ifa; 4485 u_int count; 4486 4487 count = 0; 4488 NET_EPOCH_ENTER(et); 4489 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4490 if (ifa->ifa_addr->sa_family == AF_LINK) 4491 count++; 4492 NET_EPOCH_EXIT(et); 4493 4494 return (count); 4495 } 4496 4497 int 4498 if_foreach(if_foreach_cb_t cb, void *cb_arg) 4499 { 4500 if_t ifp; 4501 int error; 4502 4503 NET_EPOCH_ASSERT(); 4504 MPASS(cb); 4505 4506 error = 0; 4507 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4508 error = cb(ifp, cb_arg); 4509 if (error != 0) 4510 break; 4511 } 4512 4513 return (error); 4514 } 4515 4516 /* 4517 * Iterates over the list of interfaces, permitting callback function @cb to sleep. 4518 * Stops iteration if @cb returns non-zero error code. 4519 * Returns the last error code from @cb. 4520 * @match_cb: optional match callback limiting the iteration to only matched interfaces 4521 * @match_arg: argument to pass to @match_cb 4522 * @cb: iteration callback 4523 * @cb_arg: argument to pass to @cb 4524 */ 4525 int 4526 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb, 4527 void *cb_arg) 4528 { 4529 int match_count = 0, array_size = 16; /* 128 bytes for malloc */ 4530 struct ifnet **match_array = NULL; 4531 int error = 0; 4532 4533 MPASS(cb); 4534 4535 while (true) { 4536 struct ifnet **new_array; 4537 int new_size = array_size; 4538 struct epoch_tracker et; 4539 struct ifnet *ifp; 4540 4541 while (new_size < match_count) 4542 new_size *= 2; 4543 new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK); 4544 if (match_array != NULL) 4545 memcpy(new_array, match_array, array_size * sizeof(void *)); 4546 free(match_array, M_TEMP); 4547 match_array = new_array; 4548 array_size = new_size; 4549 4550 match_count = 0; 4551 NET_EPOCH_ENTER(et); 4552 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4553 if (match_cb != NULL && !match_cb(ifp, match_arg)) 4554 continue; 4555 if (match_count < array_size) { 4556 if (if_try_ref(ifp)) 4557 match_array[match_count++] = ifp; 4558 } else 4559 match_count++; 4560 } 4561 NET_EPOCH_EXIT(et); 4562 4563 if (match_count > array_size) { 4564 for (int i = 0; i < array_size; i++) 4565 if_rele(match_array[i]); 4566 continue; 4567 } else { 4568 for (int i = 0; i < match_count; i++) { 4569 if (error == 0) 4570 error = cb(match_array[i], cb_arg); 4571 if_rele(match_array[i]); 4572 } 4573 free(match_array, M_TEMP); 4574 break; 4575 } 4576 } 4577 4578 return (error); 4579 } 4580 4581 4582 /* 4583 * Uses just 1 pointer of the 4 available in the public struct. 4584 */ 4585 if_t 4586 if_iter_start(struct if_iter *iter) 4587 { 4588 if_t ifp; 4589 4590 NET_EPOCH_ASSERT(); 4591 4592 bzero(iter, sizeof(*iter)); 4593 ifp = CK_STAILQ_FIRST(&V_ifnet); 4594 if (ifp != NULL) 4595 iter->context[0] = CK_STAILQ_NEXT(ifp, if_link); 4596 else 4597 iter->context[0] = NULL; 4598 return (ifp); 4599 } 4600 4601 if_t 4602 if_iter_next(struct if_iter *iter) 4603 { 4604 if_t cur_ifp = iter->context[0]; 4605 4606 if (cur_ifp != NULL) 4607 iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link); 4608 return (cur_ifp); 4609 } 4610 4611 void 4612 if_iter_finish(struct if_iter *iter) 4613 { 4614 /* Nothing to do here for now. */ 4615 } 4616 4617 u_int 4618 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4619 { 4620 struct epoch_tracker et; 4621 struct ifaddr *ifa; 4622 u_int count; 4623 4624 MPASS(cb); 4625 4626 count = 0; 4627 NET_EPOCH_ENTER(et); 4628 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4629 if (ifa->ifa_addr->sa_family != AF_LINK) 4630 continue; 4631 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4632 count); 4633 } 4634 NET_EPOCH_EXIT(et); 4635 4636 return (count); 4637 } 4638 4639 u_int 4640 if_llmaddr_count(if_t ifp) 4641 { 4642 struct epoch_tracker et; 4643 struct ifmultiaddr *ifma; 4644 int count; 4645 4646 count = 0; 4647 NET_EPOCH_ENTER(et); 4648 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4649 if (ifma->ifma_addr->sa_family == AF_LINK) 4650 count++; 4651 NET_EPOCH_EXIT(et); 4652 4653 return (count); 4654 } 4655 4656 bool 4657 if_maddr_empty(if_t ifp) 4658 { 4659 4660 return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs)); 4661 } 4662 4663 u_int 4664 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4665 { 4666 struct epoch_tracker et; 4667 struct ifmultiaddr *ifma; 4668 u_int count; 4669 4670 MPASS(cb); 4671 4672 count = 0; 4673 NET_EPOCH_ENTER(et); 4674 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4675 if (ifma->ifma_addr->sa_family != AF_LINK) 4676 continue; 4677 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4678 count); 4679 } 4680 NET_EPOCH_EXIT(et); 4681 4682 return (count); 4683 } 4684 4685 u_int 4686 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4687 { 4688 struct epoch_tracker et; 4689 struct ifaddr *ifa; 4690 u_int count; 4691 4692 MPASS(cb); 4693 4694 count = 0; 4695 NET_EPOCH_ENTER(et); 4696 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4697 if (ifa->ifa_addr->sa_family != type) 4698 continue; 4699 count += (*cb)(cb_arg, ifa, count); 4700 } 4701 NET_EPOCH_EXIT(et); 4702 4703 return (count); 4704 } 4705 4706 struct ifaddr * 4707 ifa_iter_start(if_t ifp, struct ifa_iter *iter) 4708 { 4709 struct ifaddr *ifa; 4710 4711 NET_EPOCH_ASSERT(); 4712 4713 bzero(iter, sizeof(*iter)); 4714 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 4715 if (ifa != NULL) 4716 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4717 else 4718 iter->context[0] = NULL; 4719 return (ifa); 4720 } 4721 4722 struct ifaddr * 4723 ifa_iter_next(struct ifa_iter *iter) 4724 { 4725 struct ifaddr *ifa = iter->context[0]; 4726 4727 if (ifa != NULL) 4728 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4729 return (ifa); 4730 } 4731 4732 void 4733 ifa_iter_finish(struct ifa_iter *iter) 4734 { 4735 /* Nothing to do here for now. */ 4736 } 4737 4738 int 4739 if_setsoftc(if_t ifp, void *softc) 4740 { 4741 ifp->if_softc = softc; 4742 return (0); 4743 } 4744 4745 void * 4746 if_getsoftc(const if_t ifp) 4747 { 4748 return (ifp->if_softc); 4749 } 4750 4751 void 4752 if_setrcvif(struct mbuf *m, if_t ifp) 4753 { 4754 4755 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4756 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4757 } 4758 4759 void 4760 if_setvtag(struct mbuf *m, uint16_t tag) 4761 { 4762 m->m_pkthdr.ether_vtag = tag; 4763 } 4764 4765 uint16_t 4766 if_getvtag(struct mbuf *m) 4767 { 4768 return (m->m_pkthdr.ether_vtag); 4769 } 4770 4771 int 4772 if_sendq_empty(if_t ifp) 4773 { 4774 return (IFQ_DRV_IS_EMPTY(&ifp->if_snd)); 4775 } 4776 4777 struct ifaddr * 4778 if_getifaddr(const if_t ifp) 4779 { 4780 return (ifp->if_addr); 4781 } 4782 4783 int 4784 if_setsendqready(if_t ifp) 4785 { 4786 IFQ_SET_READY(&ifp->if_snd); 4787 return (0); 4788 } 4789 4790 int 4791 if_setsendqlen(if_t ifp, int tx_desc_count) 4792 { 4793 IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count); 4794 ifp->if_snd.ifq_drv_maxlen = tx_desc_count; 4795 return (0); 4796 } 4797 4798 void 4799 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na) 4800 { 4801 ifp->if_netmap = na; 4802 } 4803 4804 struct netmap_adapter * 4805 if_getnetmapadapter(if_t ifp) 4806 { 4807 return (ifp->if_netmap); 4808 } 4809 4810 int 4811 if_vlantrunkinuse(if_t ifp) 4812 { 4813 return (ifp->if_vlantrunk != NULL); 4814 } 4815 4816 void 4817 if_init(if_t ifp, void *ctx) 4818 { 4819 (*ifp->if_init)(ctx); 4820 } 4821 4822 void 4823 if_input(if_t ifp, struct mbuf* sendmp) 4824 { 4825 (*ifp->if_input)(ifp, sendmp); 4826 } 4827 4828 int 4829 if_transmit(if_t ifp, struct mbuf *m) 4830 { 4831 return ((*ifp->if_transmit)(ifp, m)); 4832 } 4833 4834 int 4835 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst) 4836 { 4837 if (ifp->if_resolvemulti == NULL) 4838 return (EOPNOTSUPP); 4839 4840 return (ifp->if_resolvemulti(ifp, srcs, dst)); 4841 } 4842 4843 int 4844 if_ioctl(if_t ifp, u_long cmd, void *data) 4845 { 4846 if (ifp->if_ioctl == NULL) 4847 return (EOPNOTSUPP); 4848 4849 return (ifp->if_ioctl(ifp, cmd, data)); 4850 } 4851 4852 struct mbuf * 4853 if_dequeue(if_t ifp) 4854 { 4855 struct mbuf *m; 4856 4857 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 4858 return (m); 4859 } 4860 4861 int 4862 if_sendq_prepend(if_t ifp, struct mbuf *m) 4863 { 4864 IFQ_DRV_PREPEND(&ifp->if_snd, m); 4865 return (0); 4866 } 4867 4868 int 4869 if_setifheaderlen(if_t ifp, int len) 4870 { 4871 ifp->if_hdrlen = len; 4872 return (0); 4873 } 4874 4875 char * 4876 if_getlladdr(const if_t ifp) 4877 { 4878 return (IF_LLADDR(ifp)); 4879 } 4880 4881 void * 4882 if_gethandle(u_char type) 4883 { 4884 return (if_alloc(type)); 4885 } 4886 4887 void 4888 if_vlancap(if_t ifp) 4889 { 4890 VLAN_CAPABILITIES(ifp); 4891 } 4892 4893 int 4894 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4895 { 4896 ifp->if_hw_tsomax = if_hw_tsomax; 4897 return (0); 4898 } 4899 4900 int 4901 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4902 { 4903 ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4904 return (0); 4905 } 4906 4907 int 4908 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4909 { 4910 ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4911 return (0); 4912 } 4913 4914 u_int 4915 if_gethwtsomax(const if_t ifp) 4916 { 4917 return (ifp->if_hw_tsomax); 4918 } 4919 4920 u_int 4921 if_gethwtsomaxsegcount(const if_t ifp) 4922 { 4923 return (ifp->if_hw_tsomaxsegcount); 4924 } 4925 4926 u_int 4927 if_gethwtsomaxsegsize(const if_t ifp) 4928 { 4929 return (ifp->if_hw_tsomaxsegsize); 4930 } 4931 4932 void 4933 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4934 { 4935 ifp->if_init = init_fn; 4936 } 4937 4938 void 4939 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4940 { 4941 ifp->if_input = input_fn; 4942 } 4943 4944 if_input_fn_t 4945 if_getinputfn(if_t ifp) 4946 { 4947 return (ifp->if_input); 4948 } 4949 4950 void 4951 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4952 { 4953 ifp->if_ioctl = ioctl_fn; 4954 } 4955 4956 void 4957 if_setoutputfn(if_t ifp, if_output_fn_t output_fn) 4958 { 4959 ifp->if_output = output_fn; 4960 } 4961 4962 void 4963 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4964 { 4965 ifp->if_start = start_fn; 4966 } 4967 4968 if_start_fn_t 4969 if_getstartfn(if_t ifp) 4970 { 4971 return (ifp->if_start); 4972 } 4973 4974 void 4975 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4976 { 4977 ifp->if_transmit = start_fn; 4978 } 4979 4980 if_transmit_fn_t 4981 if_gettransmitfn(if_t ifp) 4982 { 4983 return (ifp->if_transmit); 4984 } 4985 4986 void 4987 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4988 { 4989 ifp->if_qflush = flush_fn; 4990 } 4991 4992 void 4993 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 4994 { 4995 ifp->if_snd_tag_alloc = alloc_fn; 4996 } 4997 4998 int 4999 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params, 5000 struct m_snd_tag **mstp) 5001 { 5002 if (ifp->if_snd_tag_alloc == NULL) 5003 return (EOPNOTSUPP); 5004 return (ifp->if_snd_tag_alloc(ifp, params, mstp)); 5005 } 5006 5007 void 5008 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 5009 { 5010 ifp->if_get_counter = fn; 5011 } 5012 5013 void 5014 if_setreassignfn(if_t ifp, if_reassign_fn_t fn) 5015 { 5016 ifp->if_reassign = fn; 5017 } 5018 5019 void 5020 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn) 5021 { 5022 ifp->if_ratelimit_query = fn; 5023 } 5024 5025 void 5026 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m) 5027 { 5028 ifp->if_debugnet_methods = m; 5029 } 5030 5031 struct label * 5032 if_getmaclabel(if_t ifp) 5033 { 5034 return (ifp->if_label); 5035 } 5036 5037 void 5038 if_setmaclabel(if_t ifp, struct label *label) 5039 { 5040 ifp->if_label = label; 5041 } 5042 5043 int 5044 if_gettype(if_t ifp) 5045 { 5046 return (ifp->if_type); 5047 } 5048 5049 void * 5050 if_getllsoftc(if_t ifp) 5051 { 5052 return (ifp->if_llsoftc); 5053 } 5054 5055 void 5056 if_setllsoftc(if_t ifp, void *llsoftc) 5057 { 5058 ifp->if_llsoftc = llsoftc; 5059 }; 5060 5061 int 5062 if_getlinkstate(if_t ifp) 5063 { 5064 return (ifp->if_link_state); 5065 } 5066 5067 const uint8_t * 5068 if_getbroadcastaddr(if_t ifp) 5069 { 5070 return (ifp->if_broadcastaddr); 5071 } 5072 5073 void 5074 if_setbroadcastaddr(if_t ifp, const uint8_t *addr) 5075 { 5076 ifp->if_broadcastaddr = addr; 5077 } 5078 5079 int 5080 if_getnumadomain(if_t ifp) 5081 { 5082 return (ifp->if_numa_domain); 5083 } 5084 5085 uint64_t 5086 if_getcounter(if_t ifp, ift_counter counter) 5087 { 5088 return (ifp->if_get_counter(ifp, counter)); 5089 } 5090 5091 bool 5092 if_altq_is_enabled(if_t ifp) 5093 { 5094 return (ALTQ_IS_ENABLED(&ifp->if_snd)); 5095 } 5096 5097 struct vnet * 5098 if_getvnet(if_t ifp) 5099 { 5100 return (ifp->if_vnet); 5101 } 5102 5103 void * 5104 if_getafdata(if_t ifp, int af) 5105 { 5106 return (ifp->if_afdata[af]); 5107 } 5108 5109 u_int 5110 if_getfib(if_t ifp) 5111 { 5112 return (ifp->if_fib); 5113 } 5114 5115 uint8_t 5116 if_getaddrlen(if_t ifp) 5117 { 5118 return (ifp->if_addrlen); 5119 } 5120 5121 struct bpf_if * 5122 if_getbpf(if_t ifp) 5123 { 5124 return (ifp->if_bpf); 5125 } 5126 5127 struct ifvlantrunk * 5128 if_getvlantrunk(if_t ifp) 5129 { 5130 return (ifp->if_vlantrunk); 5131 } 5132 5133 uint8_t 5134 if_getpcp(if_t ifp) 5135 { 5136 return (ifp->if_pcp); 5137 } 5138 5139 void * 5140 if_getl2com(if_t ifp) 5141 { 5142 return (ifp->if_l2com); 5143 } 5144 5145 void 5146 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m) 5147 { 5148 ifp->if_ipsec_accel_m = m; 5149 } 5150 5151 #ifdef DDB 5152 static void 5153 if_show_ifnet(struct ifnet *ifp) 5154 { 5155 if (ifp == NULL) 5156 return; 5157 db_printf("%s:\n", ifp->if_xname); 5158 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 5159 IF_DB_PRINTF("%s", if_dname); 5160 IF_DB_PRINTF("%d", if_dunit); 5161 IF_DB_PRINTF("%s", if_description); 5162 IF_DB_PRINTF("%u", if_index); 5163 IF_DB_PRINTF("%d", if_idxgen); 5164 IF_DB_PRINTF("%u", if_refcount); 5165 IF_DB_PRINTF("%p", if_softc); 5166 IF_DB_PRINTF("%p", if_l2com); 5167 IF_DB_PRINTF("%p", if_llsoftc); 5168 IF_DB_PRINTF("%d", if_amcount); 5169 IF_DB_PRINTF("%p", if_addr); 5170 IF_DB_PRINTF("%p", if_broadcastaddr); 5171 IF_DB_PRINTF("%p", if_afdata); 5172 IF_DB_PRINTF("%d", if_afdata_initialized); 5173 IF_DB_PRINTF("%u", if_fib); 5174 IF_DB_PRINTF("%p", if_vnet); 5175 IF_DB_PRINTF("%p", if_home_vnet); 5176 IF_DB_PRINTF("%p", if_vlantrunk); 5177 IF_DB_PRINTF("%p", if_bpf); 5178 IF_DB_PRINTF("%u", if_pcount); 5179 IF_DB_PRINTF("%p", if_bridge); 5180 IF_DB_PRINTF("%p", if_lagg); 5181 IF_DB_PRINTF("%p", if_pf_kif); 5182 IF_DB_PRINTF("%p", if_carp); 5183 IF_DB_PRINTF("%p", if_label); 5184 IF_DB_PRINTF("%p", if_netmap); 5185 IF_DB_PRINTF("0x%08x", if_flags); 5186 IF_DB_PRINTF("0x%08x", if_drv_flags); 5187 IF_DB_PRINTF("0x%08x", if_capabilities); 5188 IF_DB_PRINTF("0x%08x", if_capenable); 5189 IF_DB_PRINTF("%p", if_snd.ifq_head); 5190 IF_DB_PRINTF("%p", if_snd.ifq_tail); 5191 IF_DB_PRINTF("%d", if_snd.ifq_len); 5192 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 5193 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 5194 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 5195 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 5196 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 5197 IF_DB_PRINTF("%d", if_snd.altq_type); 5198 IF_DB_PRINTF("%x", if_snd.altq_flags); 5199 #undef IF_DB_PRINTF 5200 } 5201 5202 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 5203 { 5204 if (!have_addr) { 5205 db_printf("usage: show ifnet <struct ifnet *>\n"); 5206 return; 5207 } 5208 5209 if_show_ifnet((struct ifnet *)addr); 5210 } 5211 5212 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 5213 { 5214 struct ifnet *ifp; 5215 u_short idx; 5216 5217 for (idx = 1; idx <= if_index; idx++) { 5218 ifp = ifindex_table[idx].ife_ifnet; 5219 if (ifp == NULL) 5220 continue; 5221 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 5222 if (db_pager_quit) 5223 break; 5224 } 5225 } 5226 #endif /* DDB */ 5227