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 #endif /* INET6 */ 106 #endif /* INET || INET6 */ 107 108 #include <security/mac/mac_framework.h> 109 110 /* 111 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 112 * and ifr_ifru when it is used in SIOCGIFCONF. 113 */ 114 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 115 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 116 117 __read_mostly epoch_t net_epoch_preempt; 118 #ifdef COMPAT_FREEBSD32 119 #include <sys/mount.h> 120 #include <compat/freebsd32/freebsd32.h> 121 122 struct ifreq_buffer32 { 123 uint32_t length; /* (size_t) */ 124 uint32_t buffer; /* (void *) */ 125 }; 126 127 /* 128 * Interface request structure used for socket 129 * ioctl's. All interface ioctl's must have parameter 130 * definitions which begin with ifr_name. The 131 * remainder may be interface specific. 132 */ 133 struct ifreq32 { 134 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 135 union { 136 struct sockaddr ifru_addr; 137 struct sockaddr ifru_dstaddr; 138 struct sockaddr ifru_broadaddr; 139 struct ifreq_buffer32 ifru_buffer; 140 short ifru_flags[2]; 141 short ifru_index; 142 int ifru_jid; 143 int ifru_metric; 144 int ifru_mtu; 145 int ifru_phys; 146 int ifru_media; 147 uint32_t ifru_data; 148 int ifru_cap[2]; 149 u_int ifru_fib; 150 u_char ifru_vlan_pcp; 151 } ifr_ifru; 152 }; 153 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 154 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 155 __offsetof(struct ifreq32, ifr_ifru)); 156 157 struct ifconf32 { 158 int32_t ifc_len; 159 union { 160 uint32_t ifcu_buf; 161 uint32_t ifcu_req; 162 } ifc_ifcu; 163 }; 164 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 165 166 struct ifdrv32 { 167 char ifd_name[IFNAMSIZ]; 168 uint32_t ifd_cmd; 169 uint32_t ifd_len; 170 uint32_t ifd_data; 171 }; 172 #define SIOCSDRVSPEC32 _IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32) 173 #define SIOCGDRVSPEC32 _IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32) 174 175 struct ifgroupreq32 { 176 char ifgr_name[IFNAMSIZ]; 177 u_int ifgr_len; 178 union { 179 char ifgru_group[IFNAMSIZ]; 180 uint32_t ifgru_groups; 181 } ifgr_ifgru; 182 }; 183 #define SIOCAIFGROUP32 _IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32) 184 #define SIOCGIFGROUP32 _IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32) 185 #define SIOCDIFGROUP32 _IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32) 186 #define SIOCGIFGMEMB32 _IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32) 187 188 struct ifmediareq32 { 189 char ifm_name[IFNAMSIZ]; 190 int ifm_current; 191 int ifm_mask; 192 int ifm_status; 193 int ifm_active; 194 int ifm_count; 195 uint32_t ifm_ulist; /* (int *) */ 196 }; 197 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 198 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 199 #endif /* COMPAT_FREEBSD32 */ 200 201 union ifreq_union { 202 struct ifreq ifr; 203 #ifdef COMPAT_FREEBSD32 204 struct ifreq32 ifr32; 205 #endif 206 }; 207 208 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 209 "Link layers"); 210 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 211 "Generic link-management"); 212 213 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 214 &ifqmaxlen, 0, "max send queue size"); 215 216 /* Log link state change events */ 217 static int log_link_state_change = 1; 218 219 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 220 &log_link_state_change, 0, 221 "log interface link state change events"); 222 223 /* Log promiscuous mode change events */ 224 static int log_promisc_mode_change = 1; 225 226 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 227 &log_promisc_mode_change, 1, 228 "log promiscuous mode change events"); 229 230 /* Interface description */ 231 static unsigned int ifdescr_maxlen = 1024; 232 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 233 &ifdescr_maxlen, 0, 234 "administrative maximum length for interface description"); 235 236 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 237 238 /* global sx for non-critical path ifdescr */ 239 static struct sx ifdescr_sx; 240 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 241 242 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 243 /* These are external hooks for CARP. */ 244 void (*carp_linkstate_p)(struct ifnet *ifp); 245 void (*carp_demote_adj_p)(int, char *); 246 int (*carp_master_p)(struct ifaddr *); 247 #if defined(INET) || defined(INET6) 248 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 249 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 250 const struct sockaddr *sa); 251 int (*carp_attach_p)(struct ifaddr *, int); 252 void (*carp_detach_p)(struct ifaddr *, bool); 253 #endif 254 #ifdef INET 255 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 256 #endif 257 #ifdef INET6 258 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 259 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 260 const struct in6_addr *taddr); 261 #endif 262 263 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 264 265 /* 266 * XXX: Style; these should be sorted alphabetically, and unprototyped 267 * static functions should be prototyped. Currently they are sorted by 268 * declaration order. 269 */ 270 static int ifconf(u_long, caddr_t); 271 static void if_input_default(struct ifnet *, struct mbuf *); 272 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 273 static int if_setflag(struct ifnet *, int, int, int *, int); 274 static int if_transmit_default(struct ifnet *ifp, struct mbuf *m); 275 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 276 static void do_link_state_change(void *, int); 277 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 278 static int if_getgroupmembers(struct ifgroupreq *); 279 static void if_delgroups(struct ifnet *); 280 static void if_attach_internal(struct ifnet *, bool); 281 static void if_detach_internal(struct ifnet *, bool); 282 static void if_siocaddmulti(void *, int); 283 static void if_link_ifnet(struct ifnet *); 284 static bool if_unlink_ifnet(struct ifnet *, bool); 285 #ifdef VIMAGE 286 static void if_vmove(struct ifnet *, struct vnet *); 287 #endif 288 289 #ifdef INET6 290 /* 291 * XXX: declare here to avoid to include many inet6 related files.. 292 * should be more generalized? 293 */ 294 extern void nd6_setmtu(struct ifnet *); 295 #endif 296 297 /* ipsec helper hooks */ 298 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 299 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 300 301 int ifqmaxlen = IFQ_MAXLEN; 302 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 303 VNET_DEFINE(struct ifgrouphead, ifg_head); 304 305 /* Table of ifnet by index. */ 306 static int if_index; 307 static int if_indexlim = 8; 308 static struct ifindex_entry { 309 struct ifnet *ife_ifnet; 310 uint16_t ife_gencnt; 311 } *ifindex_table; 312 313 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system, 314 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 315 "Variables global to all interfaces"); 316 static int 317 sysctl_ifcount(SYSCTL_HANDLER_ARGS) 318 { 319 int rv = 0; 320 321 IFNET_RLOCK(); 322 for (int i = 1; i <= if_index; i++) 323 if (ifindex_table[i].ife_ifnet != NULL && 324 ifindex_table[i].ife_ifnet->if_vnet == curvnet) 325 rv = i; 326 IFNET_RUNLOCK(); 327 328 return (sysctl_handle_int(oidp, &rv, 0, req)); 329 } 330 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount, 331 CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I", 332 "Maximum known interface index"); 333 334 /* 335 * The global network interface list (V_ifnet) and related state (such as 336 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 337 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 338 */ 339 struct sx ifnet_sxlock; 340 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 341 342 struct sx ifnet_detach_sxlock; 343 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 344 SX_RECURSE); 345 346 static if_com_alloc_t *if_com_alloc[256]; 347 static if_com_free_t *if_com_free[256]; 348 349 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 350 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 351 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 352 353 struct ifnet * 354 ifnet_byindex(u_int idx) 355 { 356 struct ifnet *ifp; 357 358 NET_EPOCH_ASSERT(); 359 360 if (__predict_false(idx > if_index)) 361 return (NULL); 362 363 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 364 365 if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet) 366 ifp = NULL; 367 368 return (ifp); 369 } 370 371 struct ifnet * 372 ifnet_byindex_ref(u_int idx) 373 { 374 struct ifnet *ifp; 375 376 ifp = ifnet_byindex(idx); 377 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 378 return (NULL); 379 if (!if_try_ref(ifp)) 380 return (NULL); 381 return (ifp); 382 } 383 384 struct ifnet * 385 ifnet_byindexgen(uint16_t idx, uint16_t gen) 386 { 387 struct ifnet *ifp; 388 389 NET_EPOCH_ASSERT(); 390 391 if (__predict_false(idx > if_index)) 392 return (NULL); 393 394 ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet); 395 396 if (ifindex_table[idx].ife_gencnt == gen) 397 return (ifp); 398 else 399 return (NULL); 400 } 401 402 /* 403 * Network interface utility routines. 404 * 405 * Routines with ifa_ifwith* names take sockaddr *'s as 406 * parameters. 407 */ 408 409 static void 410 if_init_idxtable(void *arg __unused) 411 { 412 413 ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table), 414 M_IFNET, M_WAITOK | M_ZERO); 415 } 416 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL); 417 418 static void 419 vnet_if_init(const void *unused __unused) 420 { 421 422 CK_STAILQ_INIT(&V_ifnet); 423 CK_STAILQ_INIT(&V_ifg_head); 424 } 425 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 426 NULL); 427 428 static void 429 if_link_ifnet(struct ifnet *ifp) 430 { 431 432 IFNET_WLOCK(); 433 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 434 #ifdef VIMAGE 435 curvnet->vnet_ifcnt++; 436 #endif 437 IFNET_WUNLOCK(); 438 } 439 440 static bool 441 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 442 { 443 struct ifnet *iter; 444 int found = 0; 445 446 IFNET_WLOCK(); 447 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 448 if (iter == ifp) { 449 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 450 #ifdef VIMAGE 451 curvnet->vnet_ifcnt--; 452 #endif 453 if (!vmove) 454 ifp->if_flags |= IFF_DYING; 455 found = 1; 456 break; 457 } 458 IFNET_WUNLOCK(); 459 460 return (found); 461 } 462 463 #ifdef VIMAGE 464 static void 465 vnet_if_return(const void *unused __unused) 466 { 467 struct ifnet *ifp, *nifp; 468 struct ifnet **pending; 469 int found __diagused; 470 int i; 471 472 i = 0; 473 474 /* 475 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 476 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 477 * if_detach_internal(), which waits for NET_EPOCH callbacks to 478 * complete. We can't do that from within NET_EPOCH. 479 * 480 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 481 * read/write lock. We cannot hold the lock as we call if_vmove() 482 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 483 * ctx lock. 484 */ 485 IFNET_WLOCK(); 486 487 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 488 M_IFNET, M_WAITOK | M_ZERO); 489 490 /* Return all inherited interfaces to their parent vnets. */ 491 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 492 if (ifp->if_home_vnet != ifp->if_vnet) { 493 found = if_unlink_ifnet(ifp, true); 494 MPASS(found); 495 496 pending[i++] = ifp; 497 } 498 } 499 IFNET_WUNLOCK(); 500 501 for (int j = 0; j < i; j++) { 502 sx_xlock(&ifnet_detach_sxlock); 503 if_vmove(pending[j], pending[j]->if_home_vnet); 504 sx_xunlock(&ifnet_detach_sxlock); 505 } 506 507 free(pending, M_IFNET); 508 } 509 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 510 vnet_if_return, NULL); 511 #endif 512 513 /* 514 * Allocate a struct ifnet and an index for an interface. A layer 2 515 * common structure will also be allocated if an allocation routine is 516 * registered for the passed type. 517 */ 518 static struct ifnet * 519 if_alloc_domain(u_char type, int numa_domain) 520 { 521 struct ifnet *ifp; 522 u_short idx; 523 524 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 525 if (numa_domain == IF_NODOM) 526 ifp = malloc(sizeof(struct ifnet), M_IFNET, 527 M_WAITOK | M_ZERO); 528 else 529 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 530 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 531 ifp->if_type = type; 532 ifp->if_alloctype = type; 533 ifp->if_numa_domain = numa_domain; 534 #ifdef VIMAGE 535 ifp->if_vnet = curvnet; 536 #endif 537 if (if_com_alloc[type] != NULL) { 538 ifp->if_l2com = if_com_alloc[type](type, ifp); 539 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 540 type)); 541 } 542 543 IF_ADDR_LOCK_INIT(ifp); 544 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 545 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 546 CK_STAILQ_INIT(&ifp->if_addrhead); 547 CK_STAILQ_INIT(&ifp->if_multiaddrs); 548 CK_STAILQ_INIT(&ifp->if_groups); 549 #ifdef MAC 550 mac_ifnet_init(ifp); 551 #endif 552 ifq_init(&ifp->if_snd, ifp); 553 554 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 555 for (int i = 0; i < IFCOUNTERS; i++) 556 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 557 ifp->if_get_counter = if_get_counter_default; 558 ifp->if_pcp = IFNET_PCP_NONE; 559 560 /* Allocate an ifindex array entry. */ 561 IFNET_WLOCK(); 562 /* 563 * Try to find an empty slot below if_index. If we fail, take the 564 * next slot. 565 */ 566 for (idx = 1; idx <= if_index; idx++) { 567 if (ifindex_table[idx].ife_ifnet == NULL) 568 break; 569 } 570 571 /* Catch if_index overflow. */ 572 if (idx >= if_indexlim) { 573 struct ifindex_entry *new, *old; 574 int newlim; 575 576 newlim = if_indexlim * 2; 577 new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO); 578 memcpy(new, ifindex_table, if_indexlim * sizeof(*new)); 579 old = ifindex_table; 580 ck_pr_store_ptr(&ifindex_table, new); 581 if_indexlim = newlim; 582 NET_EPOCH_WAIT(); 583 free(old, M_IFNET); 584 } 585 if (idx > if_index) 586 if_index = idx; 587 588 ifp->if_index = idx; 589 ifp->if_idxgen = ifindex_table[idx].ife_gencnt; 590 ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp); 591 IFNET_WUNLOCK(); 592 593 return (ifp); 594 } 595 596 struct ifnet * 597 if_alloc_dev(u_char type, device_t dev) 598 { 599 int numa_domain; 600 601 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 602 return (if_alloc_domain(type, IF_NODOM)); 603 return (if_alloc_domain(type, numa_domain)); 604 } 605 606 struct ifnet * 607 if_alloc(u_char type) 608 { 609 610 return (if_alloc_domain(type, IF_NODOM)); 611 } 612 /* 613 * Do the actual work of freeing a struct ifnet, and layer 2 common 614 * structure. This call is made when the network epoch guarantees 615 * us that nobody holds a pointer to the interface. 616 */ 617 static void 618 if_free_deferred(epoch_context_t ctx) 619 { 620 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 621 622 KASSERT((ifp->if_flags & IFF_DYING), 623 ("%s: interface not dying", __func__)); 624 625 if (if_com_free[ifp->if_alloctype] != NULL) 626 if_com_free[ifp->if_alloctype](ifp->if_l2com, 627 ifp->if_alloctype); 628 629 #ifdef MAC 630 mac_ifnet_destroy(ifp); 631 #endif /* MAC */ 632 IF_ADDR_LOCK_DESTROY(ifp); 633 ifq_delete(&ifp->if_snd); 634 635 for (int i = 0; i < IFCOUNTERS; i++) 636 counter_u64_free(ifp->if_counters[i]); 637 638 if_freedescr(ifp->if_description); 639 free(ifp->if_hw_addr, M_IFADDR); 640 free(ifp, M_IFNET); 641 } 642 643 /* 644 * Deregister an interface and free the associated storage. 645 */ 646 void 647 if_free(struct ifnet *ifp) 648 { 649 650 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 651 652 /* 653 * XXXGL: An interface index is really an alias to ifp pointer. 654 * Why would we clear the alias now, and not in the deferred 655 * context? Indeed there is nothing wrong with some network 656 * thread obtaining ifp via ifnet_byindex() inside the network 657 * epoch and then dereferencing ifp while we perform if_free(), 658 * and after if_free() finished, too. 659 * 660 * This early index freeing was important back when ifindex was 661 * virtualized and interface would outlive the vnet. 662 */ 663 IFNET_WLOCK(); 664 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 665 ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL); 666 ifindex_table[ifp->if_index].ife_gencnt++; 667 while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL) 668 if_index--; 669 IFNET_WUNLOCK(); 670 671 if (refcount_release(&ifp->if_refcount)) 672 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 673 } 674 675 /* 676 * Interfaces to keep an ifnet type-stable despite the possibility of the 677 * driver calling if_free(). If there are additional references, we defer 678 * freeing the underlying data structure. 679 */ 680 void 681 if_ref(struct ifnet *ifp) 682 { 683 u_int old __diagused; 684 685 /* We don't assert the ifnet list lock here, but arguably should. */ 686 old = refcount_acquire(&ifp->if_refcount); 687 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 688 } 689 690 bool 691 if_try_ref(struct ifnet *ifp) 692 { 693 NET_EPOCH_ASSERT(); 694 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 695 } 696 697 void 698 if_rele(struct ifnet *ifp) 699 { 700 701 if (!refcount_release(&ifp->if_refcount)) 702 return; 703 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 704 } 705 706 void 707 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 708 { 709 710 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 711 712 if (ifq->ifq_maxlen == 0) 713 ifq->ifq_maxlen = ifqmaxlen; 714 715 ifq->altq_type = 0; 716 ifq->altq_disc = NULL; 717 ifq->altq_flags &= ALTQF_CANTCHANGE; 718 ifq->altq_tbr = NULL; 719 ifq->altq_ifp = ifp; 720 } 721 722 void 723 ifq_delete(struct ifaltq *ifq) 724 { 725 mtx_destroy(&ifq->ifq_mtx); 726 } 727 728 /* 729 * Perform generic interface initialization tasks and attach the interface 730 * to the list of "active" interfaces. If vmove flag is set on entry 731 * to if_attach_internal(), perform only a limited subset of initialization 732 * tasks, given that we are moving from one vnet to another an ifnet which 733 * has already been fully initialized. 734 * 735 * Note that if_detach_internal() removes group membership unconditionally 736 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 737 * Thus, when if_vmove() is applied to a cloned interface, group membership 738 * is lost while a cloned one always joins a group whose name is 739 * ifc->ifc_name. To recover this after if_detach_internal() and 740 * if_attach_internal(), the cloner should be specified to 741 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 742 * attempts to join a group whose name is ifc->ifc_name. 743 * 744 * XXX: 745 * - The decision to return void and thus require this function to 746 * succeed is questionable. 747 * - We should probably do more sanity checking. For instance we don't 748 * do anything to insure if_xname is unique or non-empty. 749 */ 750 void 751 if_attach(struct ifnet *ifp) 752 { 753 754 if_attach_internal(ifp, false); 755 } 756 757 /* 758 * Compute the least common TSO limit. 759 */ 760 void 761 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 762 { 763 /* 764 * 1) If there is no limit currently, take the limit from 765 * the network adapter. 766 * 767 * 2) If the network adapter has a limit below the current 768 * limit, apply it. 769 */ 770 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 771 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 772 pmax->tsomaxbytes = ifp->if_hw_tsomax; 773 } 774 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 775 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 776 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 777 } 778 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 779 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 780 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 781 } 782 } 783 784 /* 785 * Update TSO limit of a network adapter. 786 * 787 * Returns zero if no change. Else non-zero. 788 */ 789 int 790 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 791 { 792 int retval = 0; 793 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 794 ifp->if_hw_tsomax = pmax->tsomaxbytes; 795 retval++; 796 } 797 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 798 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 799 retval++; 800 } 801 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 802 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 803 retval++; 804 } 805 return (retval); 806 } 807 808 static void 809 if_attach_internal(struct ifnet *ifp, bool vmove) 810 { 811 unsigned socksize, ifasize; 812 int namelen, masklen; 813 struct sockaddr_dl *sdl; 814 struct ifaddr *ifa; 815 816 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 817 818 #ifdef VIMAGE 819 CURVNET_ASSERT_SET(); 820 ifp->if_vnet = curvnet; 821 if (ifp->if_home_vnet == NULL) 822 ifp->if_home_vnet = curvnet; 823 #endif 824 825 if_addgroup(ifp, IFG_ALL); 826 827 #ifdef VIMAGE 828 /* Restore group membership for cloned interface. */ 829 if (vmove) 830 if_clone_restoregroup(ifp); 831 #endif 832 833 getmicrotime(&ifp->if_lastchange); 834 ifp->if_epoch = time_uptime; 835 836 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 837 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 838 ("transmit and qflush must both either be set or both be NULL")); 839 if (ifp->if_transmit == NULL) { 840 ifp->if_transmit = if_transmit_default; 841 ifp->if_qflush = if_qflush; 842 } 843 if (ifp->if_input == NULL) 844 ifp->if_input = if_input_default; 845 846 if (ifp->if_requestencap == NULL) 847 ifp->if_requestencap = if_requestencap_default; 848 849 if (!vmove) { 850 #ifdef MAC 851 mac_ifnet_create(ifp); 852 #endif 853 854 /* 855 * Create a Link Level name for this device. 856 */ 857 namelen = strlen(ifp->if_xname); 858 /* 859 * Always save enough space for any possible name so we 860 * can do a rename in place later. 861 */ 862 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 863 socksize = masklen + ifp->if_addrlen; 864 if (socksize < sizeof(*sdl)) 865 socksize = sizeof(*sdl); 866 socksize = roundup2(socksize, sizeof(long)); 867 ifasize = sizeof(*ifa) + 2 * socksize; 868 ifa = ifa_alloc(ifasize, M_WAITOK); 869 sdl = (struct sockaddr_dl *)(ifa + 1); 870 sdl->sdl_len = socksize; 871 sdl->sdl_family = AF_LINK; 872 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 873 sdl->sdl_nlen = namelen; 874 sdl->sdl_index = ifp->if_index; 875 sdl->sdl_type = ifp->if_type; 876 ifp->if_addr = ifa; 877 ifa->ifa_ifp = ifp; 878 ifa->ifa_addr = (struct sockaddr *)sdl; 879 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 880 ifa->ifa_netmask = (struct sockaddr *)sdl; 881 sdl->sdl_len = masklen; 882 while (namelen != 0) 883 sdl->sdl_data[--namelen] = 0xff; 884 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 885 /* Reliably crash if used uninitialized. */ 886 ifp->if_broadcastaddr = NULL; 887 888 if (ifp->if_type == IFT_ETHER) { 889 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 890 M_WAITOK | M_ZERO); 891 } 892 893 #if defined(INET) || defined(INET6) 894 /* Use defaults for TSO, if nothing is set */ 895 if (ifp->if_hw_tsomax == 0 && 896 ifp->if_hw_tsomaxsegcount == 0 && 897 ifp->if_hw_tsomaxsegsize == 0) { 898 /* 899 * The TSO defaults needs to be such that an 900 * NFS mbuf list of 35 mbufs totalling just 901 * below 64K works and that a chain of mbufs 902 * can be defragged into at most 32 segments: 903 */ 904 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 905 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 906 ifp->if_hw_tsomaxsegcount = 35; 907 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 908 909 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 910 if (ifp->if_capabilities & IFCAP_TSO) { 911 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 912 ifp->if_hw_tsomax, 913 ifp->if_hw_tsomaxsegcount, 914 ifp->if_hw_tsomaxsegsize); 915 } 916 } 917 #endif 918 } 919 920 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 921 if_link_ifnet(ifp); 922 EVENTHANDLER_INVOKE(ifnet_attached_event, ifp); 923 if (IS_DEFAULT_VNET(curvnet)) 924 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 925 } 926 927 static void 928 if_epochalloc(void *dummy __unused) 929 { 930 931 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 932 } 933 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 934 935 /* 936 * Remove any unicast or broadcast network addresses from an interface. 937 */ 938 void 939 if_purgeaddrs(struct ifnet *ifp) 940 { 941 struct ifaddr *ifa; 942 943 #ifdef INET6 944 /* 945 * Need to leave multicast addresses of proxy NDP llentries 946 * before in6_purgeifaddr() because the llentries are keys 947 * for in6_multi objects of proxy NDP entries. 948 * in6_purgeifaddr()s clean up llentries including proxy NDPs 949 * then we would lose the keys if they are called earlier. 950 */ 951 in6_purge_proxy_ndp(ifp); 952 #endif 953 while (1) { 954 struct epoch_tracker et; 955 956 NET_EPOCH_ENTER(et); 957 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 958 if (ifa->ifa_addr->sa_family != AF_LINK) 959 break; 960 } 961 NET_EPOCH_EXIT(et); 962 963 if (ifa == NULL) 964 break; 965 #ifdef INET 966 /* XXX: Ugly!! ad hoc just for INET */ 967 if (ifa->ifa_addr->sa_family == AF_INET) { 968 struct ifreq ifr; 969 970 bzero(&ifr, sizeof(ifr)); 971 ifr.ifr_addr = *ifa->ifa_addr; 972 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 973 NULL) == 0) 974 continue; 975 } 976 #endif /* INET */ 977 #ifdef INET6 978 if (ifa->ifa_addr->sa_family == AF_INET6) { 979 in6_purgeifaddr((struct in6_ifaddr *)ifa); 980 /* ifp_addrhead is already updated */ 981 continue; 982 } 983 #endif /* INET6 */ 984 IF_ADDR_WLOCK(ifp); 985 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 986 IF_ADDR_WUNLOCK(ifp); 987 ifa_free(ifa); 988 } 989 } 990 991 /* 992 * Remove any multicast network addresses from an interface when an ifnet 993 * is going away. 994 */ 995 static void 996 if_purgemaddrs(struct ifnet *ifp) 997 { 998 struct ifmultiaddr *ifma; 999 1000 IF_ADDR_WLOCK(ifp); 1001 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1002 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1003 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1004 if_delmulti_locked(ifp, ifma, 1); 1005 } 1006 IF_ADDR_WUNLOCK(ifp); 1007 } 1008 1009 /* 1010 * Detach an interface, removing it from the list of "active" interfaces. 1011 * If vmove flag is set on entry to if_detach_internal(), perform only a 1012 * limited subset of cleanup tasks, given that we are moving an ifnet from 1013 * one vnet to another, where it must be fully operational. 1014 * 1015 * XXXRW: There are some significant questions about event ordering, and 1016 * how to prevent things from starting to use the interface during detach. 1017 */ 1018 void 1019 if_detach(struct ifnet *ifp) 1020 { 1021 bool found; 1022 1023 CURVNET_SET_QUIET(ifp->if_vnet); 1024 found = if_unlink_ifnet(ifp, false); 1025 if (found) { 1026 sx_xlock(&ifnet_detach_sxlock); 1027 if_detach_internal(ifp, false); 1028 sx_xunlock(&ifnet_detach_sxlock); 1029 } 1030 CURVNET_RESTORE(); 1031 } 1032 1033 /* 1034 * The vmove flag, if set, indicates that we are called from a callpath 1035 * that is moving an interface to a different vnet instance. 1036 * 1037 * The shutdown flag, if set, indicates that we are called in the 1038 * process of shutting down a vnet instance. Currently only the 1039 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1040 * on a vnet instance shutdown without this flag being set, e.g., when 1041 * the cloned interfaces are destoyed as first thing of teardown. 1042 */ 1043 static void 1044 if_detach_internal(struct ifnet *ifp, bool vmove) 1045 { 1046 struct ifaddr *ifa; 1047 #ifdef VIMAGE 1048 bool shutdown; 1049 1050 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1051 #endif 1052 1053 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 1054 1055 /* 1056 * At this point we know the interface still was on the ifnet list 1057 * and we removed it so we are in a stable state. 1058 */ 1059 NET_EPOCH_WAIT(); 1060 1061 /* 1062 * Ensure all pending EPOCH(9) callbacks have been executed. This 1063 * fixes issues about late destruction of multicast options 1064 * which lead to leave group calls, which in turn access the 1065 * belonging ifnet structure: 1066 */ 1067 NET_EPOCH_DRAIN_CALLBACKS(); 1068 1069 /* 1070 * In any case (destroy or vmove) detach us from the groups 1071 * and remove/wait for pending events on the taskq. 1072 * XXX-BZ in theory an interface could still enqueue a taskq change? 1073 */ 1074 if_delgroups(ifp); 1075 1076 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1077 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1078 1079 if_down(ifp); 1080 1081 #ifdef VIMAGE 1082 /* 1083 * On VNET shutdown abort here as the stack teardown will do all 1084 * the work top-down for us. 1085 */ 1086 if (shutdown) { 1087 /* Give interface users the chance to clean up. */ 1088 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1089 1090 /* 1091 * In case of a vmove we are done here without error. 1092 * If we would signal an error it would lead to the same 1093 * abort as if we did not find the ifnet anymore. 1094 * if_detach() calls us in void context and does not care 1095 * about an early abort notification, so life is splendid :) 1096 */ 1097 return; 1098 } 1099 #endif 1100 1101 /* 1102 * At this point we are not tearing down a VNET and are either 1103 * going to destroy or vmove the interface and have to cleanup 1104 * accordingly. 1105 */ 1106 1107 /* 1108 * Remove routes and flush queues. 1109 */ 1110 #ifdef ALTQ 1111 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1112 altq_disable(&ifp->if_snd); 1113 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1114 altq_detach(&ifp->if_snd); 1115 #endif 1116 1117 rt_flushifroutes(ifp); 1118 1119 if_purgeaddrs(ifp); 1120 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1121 if_purgemaddrs(ifp); 1122 if (IS_DEFAULT_VNET(curvnet)) 1123 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1124 1125 if (!vmove) { 1126 /* 1127 * Prevent further calls into the device driver via ifnet. 1128 */ 1129 if_dead(ifp); 1130 1131 /* 1132 * Clean up all addresses. 1133 */ 1134 IF_ADDR_WLOCK(ifp); 1135 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1136 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1137 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1138 IF_ADDR_WUNLOCK(ifp); 1139 ifa_free(ifa); 1140 } else 1141 IF_ADDR_WUNLOCK(ifp); 1142 } 1143 } 1144 1145 #ifdef VIMAGE 1146 /* 1147 * if_vmove() performs a limited version of if_detach() in current 1148 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1149 */ 1150 static void 1151 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1152 { 1153 #ifdef DEV_BPF 1154 /* 1155 * Detach BPF file descriptors from its interface. 1156 */ 1157 bpf_ifdetach(ifp); 1158 #endif 1159 1160 /* 1161 * Detach from current vnet, but preserve LLADDR info, do not 1162 * mark as dead etc. so that the ifnet can be reattached later. 1163 */ 1164 if_detach_internal(ifp, true); 1165 1166 /* 1167 * Perform interface-specific reassignment tasks, if provided by 1168 * the driver. 1169 */ 1170 if (ifp->if_reassign != NULL) 1171 ifp->if_reassign(ifp, new_vnet, NULL); 1172 1173 /* 1174 * Switch to the context of the target vnet. 1175 */ 1176 CURVNET_SET_QUIET(new_vnet); 1177 if_attach_internal(ifp, true); 1178 bpf_vmove(ifp->if_bpf); 1179 CURVNET_RESTORE(); 1180 } 1181 1182 /* 1183 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1184 */ 1185 static int 1186 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1187 { 1188 struct prison *pr; 1189 struct ifnet *difp; 1190 bool found; 1191 1192 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 1193 1194 /* Try to find the prison within our visibility. */ 1195 sx_slock(&allprison_lock); 1196 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1197 sx_sunlock(&allprison_lock); 1198 if (pr == NULL) 1199 return (ENXIO); 1200 prison_hold_locked(pr); 1201 mtx_unlock(&pr->pr_mtx); 1202 1203 /* Do not try to move the iface from and to the same prison. */ 1204 if (pr->pr_vnet == ifp->if_vnet) { 1205 prison_free(pr); 1206 return (EEXIST); 1207 } 1208 1209 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1210 /* XXX Lock interfaces to avoid races. */ 1211 CURVNET_SET_QUIET(pr->pr_vnet); 1212 difp = ifunit(ifname); 1213 CURVNET_RESTORE(); 1214 if (difp != NULL) { 1215 prison_free(pr); 1216 return (EEXIST); 1217 } 1218 sx_xlock(&ifnet_detach_sxlock); 1219 1220 found = if_unlink_ifnet(ifp, true); 1221 if (! found) { 1222 sx_xunlock(&ifnet_detach_sxlock); 1223 prison_free(pr); 1224 return (ENODEV); 1225 } 1226 1227 /* Move the interface into the child jail/vnet. */ 1228 if_vmove(ifp, pr->pr_vnet); 1229 1230 /* Report the new if_xname back to the userland. */ 1231 sprintf(ifname, "%s", ifp->if_xname); 1232 1233 sx_xunlock(&ifnet_detach_sxlock); 1234 1235 prison_free(pr); 1236 return (0); 1237 } 1238 1239 static int 1240 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1241 { 1242 struct prison *pr; 1243 struct vnet *vnet_dst; 1244 struct ifnet *ifp; 1245 int found __diagused; 1246 1247 /* Try to find the prison within our visibility. */ 1248 sx_slock(&allprison_lock); 1249 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1250 sx_sunlock(&allprison_lock); 1251 if (pr == NULL) 1252 return (ENXIO); 1253 prison_hold_locked(pr); 1254 mtx_unlock(&pr->pr_mtx); 1255 1256 /* Make sure the named iface exists in the source prison/vnet. */ 1257 CURVNET_SET(pr->pr_vnet); 1258 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1259 if (ifp == NULL) { 1260 CURVNET_RESTORE(); 1261 prison_free(pr); 1262 return (ENXIO); 1263 } 1264 1265 /* Do not try to move the iface from and to the same prison. */ 1266 vnet_dst = TD_TO_VNET(td); 1267 if (vnet_dst == ifp->if_vnet) { 1268 CURVNET_RESTORE(); 1269 prison_free(pr); 1270 return (EEXIST); 1271 } 1272 1273 /* Get interface back from child jail/vnet. */ 1274 found = if_unlink_ifnet(ifp, true); 1275 MPASS(found); 1276 sx_xlock(&ifnet_detach_sxlock); 1277 if_vmove(ifp, vnet_dst); 1278 sx_xunlock(&ifnet_detach_sxlock); 1279 CURVNET_RESTORE(); 1280 1281 /* Report the new if_xname back to the userland. */ 1282 sprintf(ifname, "%s", ifp->if_xname); 1283 1284 prison_free(pr); 1285 return (0); 1286 } 1287 #endif /* VIMAGE */ 1288 1289 /* 1290 * Add a group to an interface 1291 */ 1292 int 1293 if_addgroup(struct ifnet *ifp, const char *groupname) 1294 { 1295 struct ifg_list *ifgl; 1296 struct ifg_group *ifg = NULL; 1297 struct ifg_member *ifgm; 1298 int new = 0; 1299 1300 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1301 groupname[strlen(groupname) - 1] <= '9') 1302 return (EINVAL); 1303 1304 IFNET_WLOCK(); 1305 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1306 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1307 IFNET_WUNLOCK(); 1308 return (EEXIST); 1309 } 1310 1311 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1312 IFNET_WUNLOCK(); 1313 return (ENOMEM); 1314 } 1315 1316 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1317 free(ifgl, M_TEMP); 1318 IFNET_WUNLOCK(); 1319 return (ENOMEM); 1320 } 1321 1322 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1323 if (!strcmp(ifg->ifg_group, groupname)) 1324 break; 1325 1326 if (ifg == NULL) { 1327 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1328 free(ifgl, M_TEMP); 1329 free(ifgm, M_TEMP); 1330 IFNET_WUNLOCK(); 1331 return (ENOMEM); 1332 } 1333 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1334 ifg->ifg_refcnt = 0; 1335 CK_STAILQ_INIT(&ifg->ifg_members); 1336 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1337 new = 1; 1338 } 1339 1340 ifg->ifg_refcnt++; 1341 ifgl->ifgl_group = ifg; 1342 ifgm->ifgm_ifp = ifp; 1343 1344 IF_ADDR_WLOCK(ifp); 1345 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1346 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1347 IF_ADDR_WUNLOCK(ifp); 1348 1349 IFNET_WUNLOCK(); 1350 1351 if (new) 1352 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1353 EVENTHANDLER_INVOKE(group_change_event, groupname); 1354 1355 return (0); 1356 } 1357 1358 /* 1359 * Helper function to remove a group out of an interface. Expects the global 1360 * ifnet lock to be write-locked, and drops it before returning. 1361 */ 1362 static void 1363 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1364 const char *groupname) 1365 { 1366 struct ifg_member *ifgm; 1367 bool freeifgl; 1368 1369 IFNET_WLOCK_ASSERT(); 1370 1371 IF_ADDR_WLOCK(ifp); 1372 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1373 IF_ADDR_WUNLOCK(ifp); 1374 1375 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1376 if (ifgm->ifgm_ifp == ifp) { 1377 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1378 ifg_member, ifgm_next); 1379 break; 1380 } 1381 } 1382 1383 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1384 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1385 ifg_next); 1386 freeifgl = true; 1387 } else { 1388 freeifgl = false; 1389 } 1390 IFNET_WUNLOCK(); 1391 1392 NET_EPOCH_WAIT(); 1393 EVENTHANDLER_INVOKE(group_change_event, groupname); 1394 if (freeifgl) { 1395 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1396 free(ifgl->ifgl_group, M_TEMP); 1397 } 1398 free(ifgm, M_TEMP); 1399 free(ifgl, M_TEMP); 1400 } 1401 1402 /* 1403 * Remove a group from an interface 1404 */ 1405 int 1406 if_delgroup(struct ifnet *ifp, const char *groupname) 1407 { 1408 struct ifg_list *ifgl; 1409 1410 IFNET_WLOCK(); 1411 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1412 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1413 break; 1414 if (ifgl == NULL) { 1415 IFNET_WUNLOCK(); 1416 return (ENOENT); 1417 } 1418 1419 _if_delgroup_locked(ifp, ifgl, groupname); 1420 1421 return (0); 1422 } 1423 1424 /* 1425 * Remove an interface from all groups 1426 */ 1427 static void 1428 if_delgroups(struct ifnet *ifp) 1429 { 1430 struct ifg_list *ifgl; 1431 char groupname[IFNAMSIZ]; 1432 1433 IFNET_WLOCK(); 1434 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1435 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1436 _if_delgroup_locked(ifp, ifgl, groupname); 1437 IFNET_WLOCK(); 1438 } 1439 IFNET_WUNLOCK(); 1440 } 1441 1442 /* 1443 * Stores all groups from an interface in memory pointed to by ifgr. 1444 */ 1445 static int 1446 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1447 { 1448 int len, error; 1449 struct ifg_list *ifgl; 1450 struct ifg_req ifgrq, *ifgp; 1451 1452 NET_EPOCH_ASSERT(); 1453 1454 if (ifgr->ifgr_len == 0) { 1455 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1456 ifgr->ifgr_len += sizeof(struct ifg_req); 1457 return (0); 1458 } 1459 1460 len = ifgr->ifgr_len; 1461 ifgp = ifgr->ifgr_groups; 1462 /* XXX: wire */ 1463 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1464 if (len < sizeof(ifgrq)) 1465 return (EINVAL); 1466 bzero(&ifgrq, sizeof ifgrq); 1467 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1468 sizeof(ifgrq.ifgrq_group)); 1469 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1470 return (error); 1471 len -= sizeof(ifgrq); 1472 ifgp++; 1473 } 1474 1475 return (0); 1476 } 1477 1478 /* 1479 * Stores all members of a group in memory pointed to by igfr 1480 */ 1481 static int 1482 if_getgroupmembers(struct ifgroupreq *ifgr) 1483 { 1484 struct ifg_group *ifg; 1485 struct ifg_member *ifgm; 1486 struct ifg_req ifgrq, *ifgp; 1487 int len, error; 1488 1489 IFNET_RLOCK(); 1490 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1491 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1492 break; 1493 if (ifg == NULL) { 1494 IFNET_RUNLOCK(); 1495 return (ENOENT); 1496 } 1497 1498 if (ifgr->ifgr_len == 0) { 1499 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1500 ifgr->ifgr_len += sizeof(ifgrq); 1501 IFNET_RUNLOCK(); 1502 return (0); 1503 } 1504 1505 len = ifgr->ifgr_len; 1506 ifgp = ifgr->ifgr_groups; 1507 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1508 if (len < sizeof(ifgrq)) { 1509 IFNET_RUNLOCK(); 1510 return (EINVAL); 1511 } 1512 bzero(&ifgrq, sizeof ifgrq); 1513 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1514 sizeof(ifgrq.ifgrq_member)); 1515 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1516 IFNET_RUNLOCK(); 1517 return (error); 1518 } 1519 len -= sizeof(ifgrq); 1520 ifgp++; 1521 } 1522 IFNET_RUNLOCK(); 1523 1524 return (0); 1525 } 1526 1527 /* 1528 * Return counter values from counter(9)s stored in ifnet. 1529 */ 1530 uint64_t 1531 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1532 { 1533 1534 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1535 1536 return (counter_u64_fetch(ifp->if_counters[cnt])); 1537 } 1538 1539 /* 1540 * Increase an ifnet counter. Usually used for counters shared 1541 * between the stack and a driver, but function supports them all. 1542 */ 1543 void 1544 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1545 { 1546 1547 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1548 1549 counter_u64_add(ifp->if_counters[cnt], inc); 1550 } 1551 1552 /* 1553 * Copy data from ifnet to userland API structure if_data. 1554 */ 1555 void 1556 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1557 { 1558 1559 ifd->ifi_type = ifp->if_type; 1560 ifd->ifi_physical = 0; 1561 ifd->ifi_addrlen = ifp->if_addrlen; 1562 ifd->ifi_hdrlen = ifp->if_hdrlen; 1563 ifd->ifi_link_state = ifp->if_link_state; 1564 ifd->ifi_vhid = 0; 1565 ifd->ifi_datalen = sizeof(struct if_data); 1566 ifd->ifi_mtu = ifp->if_mtu; 1567 ifd->ifi_metric = ifp->if_metric; 1568 ifd->ifi_baudrate = ifp->if_baudrate; 1569 ifd->ifi_hwassist = ifp->if_hwassist; 1570 ifd->ifi_epoch = ifp->if_epoch; 1571 ifd->ifi_lastchange = ifp->if_lastchange; 1572 1573 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1574 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1575 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1576 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1577 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1578 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1579 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1580 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1581 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1582 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1583 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1584 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1585 } 1586 1587 /* 1588 * Initialization, destruction and refcounting functions for ifaddrs. 1589 */ 1590 struct ifaddr * 1591 ifa_alloc(size_t size, int flags) 1592 { 1593 struct ifaddr *ifa; 1594 1595 KASSERT(size >= sizeof(struct ifaddr), 1596 ("%s: invalid size %zu", __func__, size)); 1597 1598 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1599 if (ifa == NULL) 1600 return (NULL); 1601 1602 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1603 goto fail; 1604 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1605 goto fail; 1606 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1607 goto fail; 1608 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1609 goto fail; 1610 1611 refcount_init(&ifa->ifa_refcnt, 1); 1612 1613 return (ifa); 1614 1615 fail: 1616 /* free(NULL) is okay */ 1617 counter_u64_free(ifa->ifa_opackets); 1618 counter_u64_free(ifa->ifa_ipackets); 1619 counter_u64_free(ifa->ifa_obytes); 1620 counter_u64_free(ifa->ifa_ibytes); 1621 free(ifa, M_IFADDR); 1622 1623 return (NULL); 1624 } 1625 1626 void 1627 ifa_ref(struct ifaddr *ifa) 1628 { 1629 u_int old __diagused; 1630 1631 old = refcount_acquire(&ifa->ifa_refcnt); 1632 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1633 } 1634 1635 int 1636 ifa_try_ref(struct ifaddr *ifa) 1637 { 1638 1639 NET_EPOCH_ASSERT(); 1640 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1641 } 1642 1643 static void 1644 ifa_destroy(epoch_context_t ctx) 1645 { 1646 struct ifaddr *ifa; 1647 1648 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1649 counter_u64_free(ifa->ifa_opackets); 1650 counter_u64_free(ifa->ifa_ipackets); 1651 counter_u64_free(ifa->ifa_obytes); 1652 counter_u64_free(ifa->ifa_ibytes); 1653 free(ifa, M_IFADDR); 1654 } 1655 1656 void 1657 ifa_free(struct ifaddr *ifa) 1658 { 1659 1660 if (refcount_release(&ifa->ifa_refcnt)) 1661 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1662 } 1663 1664 /* 1665 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1666 * structs used to represent other address families, it is necessary 1667 * to perform a different comparison. 1668 */ 1669 static bool 1670 sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b) 1671 { 1672 const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a; 1673 const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b; 1674 1675 return (sdl1->sdl_len == sdl2->sdl_len && 1676 bcmp(sdl1->sdl_data + sdl1->sdl_nlen, 1677 sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0); 1678 } 1679 1680 /* 1681 * Locate an interface based on a complete address. 1682 */ 1683 /*ARGSUSED*/ 1684 struct ifaddr * 1685 ifa_ifwithaddr(const struct sockaddr *addr) 1686 { 1687 struct ifnet *ifp; 1688 struct ifaddr *ifa; 1689 1690 NET_EPOCH_ASSERT(); 1691 1692 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1693 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1694 if (ifa->ifa_addr->sa_family != addr->sa_family) 1695 continue; 1696 if (sa_equal(addr, ifa->ifa_addr)) { 1697 goto done; 1698 } 1699 /* IP6 doesn't have broadcast */ 1700 if ((ifp->if_flags & IFF_BROADCAST) && 1701 ifa->ifa_broadaddr && 1702 ifa->ifa_broadaddr->sa_len != 0 && 1703 sa_equal(ifa->ifa_broadaddr, addr)) { 1704 goto done; 1705 } 1706 } 1707 } 1708 ifa = NULL; 1709 done: 1710 return (ifa); 1711 } 1712 1713 int 1714 ifa_ifwithaddr_check(const struct sockaddr *addr) 1715 { 1716 struct epoch_tracker et; 1717 int rc; 1718 1719 NET_EPOCH_ENTER(et); 1720 rc = (ifa_ifwithaddr(addr) != NULL); 1721 NET_EPOCH_EXIT(et); 1722 return (rc); 1723 } 1724 1725 /* 1726 * Locate an interface based on the broadcast address. 1727 */ 1728 /* ARGSUSED */ 1729 struct ifaddr * 1730 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1731 { 1732 struct ifnet *ifp; 1733 struct ifaddr *ifa; 1734 1735 NET_EPOCH_ASSERT(); 1736 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1737 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1738 continue; 1739 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1740 if (ifa->ifa_addr->sa_family != addr->sa_family) 1741 continue; 1742 if ((ifp->if_flags & IFF_BROADCAST) && 1743 ifa->ifa_broadaddr && 1744 ifa->ifa_broadaddr->sa_len != 0 && 1745 sa_equal(ifa->ifa_broadaddr, addr)) { 1746 goto done; 1747 } 1748 } 1749 } 1750 ifa = NULL; 1751 done: 1752 return (ifa); 1753 } 1754 1755 /* 1756 * Locate the point to point interface with a given destination address. 1757 */ 1758 /*ARGSUSED*/ 1759 struct ifaddr * 1760 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1761 { 1762 struct ifnet *ifp; 1763 struct ifaddr *ifa; 1764 1765 NET_EPOCH_ASSERT(); 1766 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1767 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1768 continue; 1769 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1770 continue; 1771 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1772 if (ifa->ifa_addr->sa_family != addr->sa_family) 1773 continue; 1774 if (ifa->ifa_dstaddr != NULL && 1775 sa_equal(addr, ifa->ifa_dstaddr)) { 1776 goto done; 1777 } 1778 } 1779 } 1780 ifa = NULL; 1781 done: 1782 return (ifa); 1783 } 1784 1785 /* 1786 * Find an interface on a specific network. If many, choice 1787 * is most specific found. 1788 */ 1789 struct ifaddr * 1790 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1791 { 1792 struct ifnet *ifp; 1793 struct ifaddr *ifa; 1794 struct ifaddr *ifa_maybe = NULL; 1795 u_int af = addr->sa_family; 1796 const char *addr_data = addr->sa_data, *cplim; 1797 1798 NET_EPOCH_ASSERT(); 1799 /* 1800 * AF_LINK addresses can be looked up directly by their index number, 1801 * so do that if we can. 1802 */ 1803 if (af == AF_LINK) { 1804 ifp = ifnet_byindex( 1805 ((const struct sockaddr_dl *)addr)->sdl_index); 1806 return (ifp ? ifp->if_addr : NULL); 1807 } 1808 1809 /* 1810 * Scan though each interface, looking for ones that have addresses 1811 * in this address family and the requested fib. 1812 */ 1813 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1814 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1815 continue; 1816 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1817 const char *cp, *cp2, *cp3; 1818 1819 if (ifa->ifa_addr->sa_family != af) 1820 next: continue; 1821 if (af == AF_INET && 1822 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1823 /* 1824 * This is a bit broken as it doesn't 1825 * take into account that the remote end may 1826 * be a single node in the network we are 1827 * looking for. 1828 * The trouble is that we don't know the 1829 * netmask for the remote end. 1830 */ 1831 if (ifa->ifa_dstaddr != NULL && 1832 sa_equal(addr, ifa->ifa_dstaddr)) { 1833 goto done; 1834 } 1835 } else { 1836 /* 1837 * Scan all the bits in the ifa's address. 1838 * If a bit dissagrees with what we are 1839 * looking for, mask it with the netmask 1840 * to see if it really matters. 1841 * (A byte at a time) 1842 */ 1843 if (ifa->ifa_netmask == 0) 1844 continue; 1845 cp = addr_data; 1846 cp2 = ifa->ifa_addr->sa_data; 1847 cp3 = ifa->ifa_netmask->sa_data; 1848 cplim = ifa->ifa_netmask->sa_len 1849 + (char *)ifa->ifa_netmask; 1850 while (cp3 < cplim) 1851 if ((*cp++ ^ *cp2++) & *cp3++) 1852 goto next; /* next address! */ 1853 /* 1854 * If the netmask of what we just found 1855 * is more specific than what we had before 1856 * (if we had one), or if the virtual status 1857 * of new prefix is better than of the old one, 1858 * then remember the new one before continuing 1859 * to search for an even better one. 1860 */ 1861 if (ifa_maybe == NULL || 1862 ifa_preferred(ifa_maybe, ifa) || 1863 rn_refines((caddr_t)ifa->ifa_netmask, 1864 (caddr_t)ifa_maybe->ifa_netmask)) { 1865 ifa_maybe = ifa; 1866 } 1867 } 1868 } 1869 } 1870 ifa = ifa_maybe; 1871 ifa_maybe = NULL; 1872 done: 1873 return (ifa); 1874 } 1875 1876 /* 1877 * Find an interface address specific to an interface best matching 1878 * a given address. 1879 */ 1880 struct ifaddr * 1881 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1882 { 1883 struct ifaddr *ifa; 1884 const char *cp, *cp2, *cp3; 1885 char *cplim; 1886 struct ifaddr *ifa_maybe = NULL; 1887 u_int af = addr->sa_family; 1888 1889 if (af >= AF_MAX) 1890 return (NULL); 1891 1892 NET_EPOCH_ASSERT(); 1893 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1894 if (ifa->ifa_addr->sa_family != af) 1895 continue; 1896 if (ifa_maybe == NULL) 1897 ifa_maybe = ifa; 1898 if (ifa->ifa_netmask == 0) { 1899 if (sa_equal(addr, ifa->ifa_addr) || 1900 (ifa->ifa_dstaddr && 1901 sa_equal(addr, ifa->ifa_dstaddr))) 1902 goto done; 1903 continue; 1904 } 1905 if (ifp->if_flags & IFF_POINTOPOINT) { 1906 if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr)) 1907 goto done; 1908 } else { 1909 cp = addr->sa_data; 1910 cp2 = ifa->ifa_addr->sa_data; 1911 cp3 = ifa->ifa_netmask->sa_data; 1912 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1913 for (; cp3 < cplim; cp3++) 1914 if ((*cp++ ^ *cp2++) & *cp3) 1915 break; 1916 if (cp3 == cplim) 1917 goto done; 1918 } 1919 } 1920 ifa = ifa_maybe; 1921 done: 1922 return (ifa); 1923 } 1924 1925 /* 1926 * See whether new ifa is better than current one: 1927 * 1) A non-virtual one is preferred over virtual. 1928 * 2) A virtual in master state preferred over any other state. 1929 * 1930 * Used in several address selecting functions. 1931 */ 1932 int 1933 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 1934 { 1935 1936 return (cur->ifa_carp && (!next->ifa_carp || 1937 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 1938 } 1939 1940 struct sockaddr_dl * 1941 link_alloc_sdl(size_t size, int flags) 1942 { 1943 1944 return (malloc(size, M_TEMP, flags)); 1945 } 1946 1947 void 1948 link_free_sdl(struct sockaddr *sa) 1949 { 1950 free(sa, M_TEMP); 1951 } 1952 1953 /* 1954 * Fills in given sdl with interface basic info. 1955 * Returns pointer to filled sdl. 1956 */ 1957 struct sockaddr_dl * 1958 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 1959 { 1960 struct sockaddr_dl *sdl; 1961 1962 sdl = (struct sockaddr_dl *)paddr; 1963 memset(sdl, 0, sizeof(struct sockaddr_dl)); 1964 sdl->sdl_len = sizeof(struct sockaddr_dl); 1965 sdl->sdl_family = AF_LINK; 1966 sdl->sdl_index = ifp->if_index; 1967 sdl->sdl_type = iftype; 1968 1969 return (sdl); 1970 } 1971 1972 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 1973 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1974 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 1975 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 1976 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 1977 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 1978 int (*vlan_setcookie_p)(struct ifnet *, void *); 1979 void *(*vlan_cookie_p)(struct ifnet *); 1980 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 1981 1982 /* 1983 * Handle a change in the interface link state. To avoid LORs 1984 * between driver lock and upper layer locks, as well as possible 1985 * recursions, we post event to taskqueue, and all job 1986 * is done in static do_link_state_change(). 1987 */ 1988 void 1989 if_link_state_change(struct ifnet *ifp, int link_state) 1990 { 1991 /* Return if state hasn't changed. */ 1992 if (ifp->if_link_state == link_state) 1993 return; 1994 1995 ifp->if_link_state = link_state; 1996 1997 /* XXXGL: reference ifp? */ 1998 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 1999 } 2000 2001 static void 2002 do_link_state_change(void *arg, int pending) 2003 { 2004 struct ifnet *ifp; 2005 int link_state; 2006 2007 ifp = arg; 2008 link_state = ifp->if_link_state; 2009 2010 CURVNET_SET(ifp->if_vnet); 2011 rt_ifmsg(ifp, 0); 2012 if (ifp->if_vlantrunk != NULL) 2013 (*vlan_link_state_p)(ifp); 2014 if (ifp->if_carp) 2015 (*carp_linkstate_p)(ifp); 2016 if (ifp->if_bridge) 2017 ifp->if_bridge_linkstate(ifp); 2018 if (ifp->if_lagg) 2019 (*lagg_linkstate_p)(ifp, link_state); 2020 2021 if (IS_DEFAULT_VNET(curvnet)) 2022 devctl_notify("IFNET", ifp->if_xname, 2023 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2024 NULL); 2025 if (pending > 1) 2026 if_printf(ifp, "%d link states coalesced\n", pending); 2027 if (log_link_state_change) 2028 if_printf(ifp, "link state changed to %s\n", 2029 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2030 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2031 CURVNET_RESTORE(); 2032 } 2033 2034 /* 2035 * Mark an interface down and notify protocols of 2036 * the transition. 2037 */ 2038 void 2039 if_down(struct ifnet *ifp) 2040 { 2041 2042 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2043 2044 ifp->if_flags &= ~IFF_UP; 2045 getmicrotime(&ifp->if_lastchange); 2046 ifp->if_qflush(ifp); 2047 2048 if (ifp->if_carp) 2049 (*carp_linkstate_p)(ifp); 2050 rt_ifmsg(ifp, IFF_UP); 2051 } 2052 2053 /* 2054 * Mark an interface up and notify protocols of 2055 * the transition. 2056 */ 2057 void 2058 if_up(struct ifnet *ifp) 2059 { 2060 2061 ifp->if_flags |= IFF_UP; 2062 getmicrotime(&ifp->if_lastchange); 2063 if (ifp->if_carp) 2064 (*carp_linkstate_p)(ifp); 2065 rt_ifmsg(ifp, IFF_UP); 2066 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2067 } 2068 2069 /* 2070 * Flush an interface queue. 2071 */ 2072 void 2073 if_qflush(struct ifnet *ifp) 2074 { 2075 struct mbuf *m, *n; 2076 struct ifaltq *ifq; 2077 2078 ifq = &ifp->if_snd; 2079 IFQ_LOCK(ifq); 2080 #ifdef ALTQ 2081 if (ALTQ_IS_ENABLED(ifq)) 2082 ALTQ_PURGE(ifq); 2083 #endif 2084 n = ifq->ifq_head; 2085 while ((m = n) != NULL) { 2086 n = m->m_nextpkt; 2087 m_freem(m); 2088 } 2089 ifq->ifq_head = 0; 2090 ifq->ifq_tail = 0; 2091 ifq->ifq_len = 0; 2092 IFQ_UNLOCK(ifq); 2093 } 2094 2095 /* 2096 * Map interface name to interface structure pointer, with or without 2097 * returning a reference. 2098 */ 2099 struct ifnet * 2100 ifunit_ref(const char *name) 2101 { 2102 struct epoch_tracker et; 2103 struct ifnet *ifp; 2104 2105 NET_EPOCH_ENTER(et); 2106 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2107 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2108 !(ifp->if_flags & IFF_DYING)) 2109 break; 2110 } 2111 if (ifp != NULL) { 2112 if_ref(ifp); 2113 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2114 } 2115 2116 NET_EPOCH_EXIT(et); 2117 return (ifp); 2118 } 2119 2120 struct ifnet * 2121 ifunit(const char *name) 2122 { 2123 struct epoch_tracker et; 2124 struct ifnet *ifp; 2125 2126 NET_EPOCH_ENTER(et); 2127 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2128 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2129 break; 2130 } 2131 NET_EPOCH_EXIT(et); 2132 return (ifp); 2133 } 2134 2135 void * 2136 ifr_buffer_get_buffer(void *data) 2137 { 2138 union ifreq_union *ifrup; 2139 2140 ifrup = data; 2141 #ifdef COMPAT_FREEBSD32 2142 if (SV_CURPROC_FLAG(SV_ILP32)) 2143 return ((void *)(uintptr_t) 2144 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2145 #endif 2146 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2147 } 2148 2149 static void 2150 ifr_buffer_set_buffer_null(void *data) 2151 { 2152 union ifreq_union *ifrup; 2153 2154 ifrup = data; 2155 #ifdef COMPAT_FREEBSD32 2156 if (SV_CURPROC_FLAG(SV_ILP32)) 2157 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2158 else 2159 #endif 2160 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2161 } 2162 2163 size_t 2164 ifr_buffer_get_length(void *data) 2165 { 2166 union ifreq_union *ifrup; 2167 2168 ifrup = data; 2169 #ifdef COMPAT_FREEBSD32 2170 if (SV_CURPROC_FLAG(SV_ILP32)) 2171 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2172 #endif 2173 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2174 } 2175 2176 static void 2177 ifr_buffer_set_length(void *data, size_t len) 2178 { 2179 union ifreq_union *ifrup; 2180 2181 ifrup = data; 2182 #ifdef COMPAT_FREEBSD32 2183 if (SV_CURPROC_FLAG(SV_ILP32)) 2184 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2185 else 2186 #endif 2187 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2188 } 2189 2190 void * 2191 ifr_data_get_ptr(void *ifrp) 2192 { 2193 union ifreq_union *ifrup; 2194 2195 ifrup = ifrp; 2196 #ifdef COMPAT_FREEBSD32 2197 if (SV_CURPROC_FLAG(SV_ILP32)) 2198 return ((void *)(uintptr_t) 2199 ifrup->ifr32.ifr_ifru.ifru_data); 2200 #endif 2201 return (ifrup->ifr.ifr_ifru.ifru_data); 2202 } 2203 2204 struct ifcap_nv_bit_name { 2205 uint64_t cap_bit; 2206 const char *cap_name; 2207 }; 2208 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2209 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2210 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2211 CAPNV(RXCSUM), 2212 CAPNV(TXCSUM), 2213 CAPNV(NETCONS), 2214 CAPNV(VLAN_MTU), 2215 CAPNV(VLAN_HWTAGGING), 2216 CAPNV(JUMBO_MTU), 2217 CAPNV(POLLING), 2218 CAPNV(VLAN_HWCSUM), 2219 CAPNV(TSO4), 2220 CAPNV(TSO6), 2221 CAPNV(LRO), 2222 CAPNV(WOL_UCAST), 2223 CAPNV(WOL_MCAST), 2224 CAPNV(WOL_MAGIC), 2225 CAPNV(TOE4), 2226 CAPNV(TOE6), 2227 CAPNV(VLAN_HWFILTER), 2228 CAPNV(VLAN_HWTSO), 2229 CAPNV(LINKSTATE), 2230 CAPNV(NETMAP), 2231 CAPNV(RXCSUM_IPV6), 2232 CAPNV(TXCSUM_IPV6), 2233 CAPNV(HWSTATS), 2234 CAPNV(TXRTLMT), 2235 CAPNV(HWRXTSTMP), 2236 CAPNV(MEXTPG), 2237 CAPNV(TXTLS4), 2238 CAPNV(TXTLS6), 2239 CAPNV(VXLAN_HWCSUM), 2240 CAPNV(VXLAN_HWTSO), 2241 CAPNV(TXTLS_RTLMT), 2242 {0, NULL} 2243 }; 2244 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \ 2245 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2246 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2247 CAP2NV(RXTLS4), 2248 CAP2NV(RXTLS6), 2249 CAP2NV(IPSEC_OFFLOAD), 2250 {0, NULL} 2251 }; 2252 #undef CAPNV 2253 #undef CAP2NV 2254 2255 int 2256 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2257 const struct ifcap_nv_bit_name *nn, bool all) 2258 { 2259 int i, res; 2260 2261 res = 0; 2262 for (i = 0; nn[i].cap_name != NULL; i++) { 2263 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2264 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2265 res |= nn[i].cap_bit; 2266 } else { 2267 res |= *old_cap & nn[i].cap_bit; 2268 } 2269 } 2270 return (res); 2271 } 2272 2273 void 2274 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2275 int ifr_cap, int ifr_req) 2276 { 2277 int i; 2278 2279 for (i = 0; nn[i].cap_name != NULL; i++) { 2280 if ((nn[i].cap_bit & ifr_cap) != 0) { 2281 nvlist_add_bool(nv, nn[i].cap_name, 2282 (nn[i].cap_bit & ifr_req) != 0); 2283 } 2284 } 2285 } 2286 2287 /* 2288 * Hardware specific interface ioctls. 2289 */ 2290 int 2291 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2292 { 2293 struct ifreq *ifr; 2294 int error = 0, do_ifup = 0; 2295 int new_flags, temp_flags; 2296 size_t descrlen, nvbuflen; 2297 char *descrbuf; 2298 char new_name[IFNAMSIZ]; 2299 void *buf; 2300 nvlist_t *nvcap; 2301 struct siocsifcapnv_driver_data drv_ioctl_data; 2302 2303 ifr = (struct ifreq *)data; 2304 switch (cmd) { 2305 case SIOCGIFINDEX: 2306 ifr->ifr_index = ifp->if_index; 2307 break; 2308 2309 case SIOCGIFFLAGS: 2310 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2311 ifr->ifr_flags = temp_flags & 0xffff; 2312 ifr->ifr_flagshigh = temp_flags >> 16; 2313 break; 2314 2315 case SIOCGIFCAP: 2316 ifr->ifr_reqcap = ifp->if_capabilities; 2317 ifr->ifr_curcap = ifp->if_capenable; 2318 break; 2319 2320 case SIOCGIFCAPNV: 2321 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2322 error = EINVAL; 2323 break; 2324 } 2325 buf = NULL; 2326 nvcap = nvlist_create(0); 2327 for (;;) { 2328 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2329 ifp->if_capabilities, ifp->if_capenable); 2330 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2331 ifp->if_capabilities2, ifp->if_capenable2); 2332 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2333 __DECONST(caddr_t, nvcap)); 2334 if (error != 0) { 2335 if_printf(ifp, 2336 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2337 error); 2338 break; 2339 } 2340 buf = nvlist_pack(nvcap, &nvbuflen); 2341 if (buf == NULL) { 2342 error = nvlist_error(nvcap); 2343 if (error == 0) 2344 error = EDOOFUS; 2345 break; 2346 } 2347 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2348 ifr->ifr_cap_nv.length = nvbuflen; 2349 ifr->ifr_cap_nv.buffer = NULL; 2350 error = EFBIG; 2351 break; 2352 } 2353 ifr->ifr_cap_nv.length = nvbuflen; 2354 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2355 break; 2356 } 2357 free(buf, M_NVLIST); 2358 nvlist_destroy(nvcap); 2359 break; 2360 2361 case SIOCGIFDATA: 2362 { 2363 struct if_data ifd; 2364 2365 /* Ensure uninitialised padding is not leaked. */ 2366 memset(&ifd, 0, sizeof(ifd)); 2367 2368 if_data_copy(ifp, &ifd); 2369 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2370 break; 2371 } 2372 2373 #ifdef MAC 2374 case SIOCGIFMAC: 2375 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2376 break; 2377 #endif 2378 2379 case SIOCGIFMETRIC: 2380 ifr->ifr_metric = ifp->if_metric; 2381 break; 2382 2383 case SIOCGIFMTU: 2384 ifr->ifr_mtu = ifp->if_mtu; 2385 break; 2386 2387 case SIOCGIFPHYS: 2388 /* XXXGL: did this ever worked? */ 2389 ifr->ifr_phys = 0; 2390 break; 2391 2392 case SIOCGIFDESCR: 2393 error = 0; 2394 sx_slock(&ifdescr_sx); 2395 if (ifp->if_description == NULL) 2396 error = ENOMSG; 2397 else { 2398 /* space for terminating nul */ 2399 descrlen = strlen(ifp->if_description) + 1; 2400 if (ifr_buffer_get_length(ifr) < descrlen) 2401 ifr_buffer_set_buffer_null(ifr); 2402 else 2403 error = copyout(ifp->if_description, 2404 ifr_buffer_get_buffer(ifr), descrlen); 2405 ifr_buffer_set_length(ifr, descrlen); 2406 } 2407 sx_sunlock(&ifdescr_sx); 2408 break; 2409 2410 case SIOCSIFDESCR: 2411 error = priv_check(td, PRIV_NET_SETIFDESCR); 2412 if (error) 2413 return (error); 2414 2415 /* 2416 * Copy only (length-1) bytes to make sure that 2417 * if_description is always nul terminated. The 2418 * length parameter is supposed to count the 2419 * terminating nul in. 2420 */ 2421 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2422 return (ENAMETOOLONG); 2423 else if (ifr_buffer_get_length(ifr) == 0) 2424 descrbuf = NULL; 2425 else { 2426 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2427 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2428 ifr_buffer_get_length(ifr) - 1); 2429 if (error) { 2430 if_freedescr(descrbuf); 2431 break; 2432 } 2433 } 2434 2435 if_setdescr(ifp, descrbuf); 2436 getmicrotime(&ifp->if_lastchange); 2437 break; 2438 2439 case SIOCGIFFIB: 2440 ifr->ifr_fib = ifp->if_fib; 2441 break; 2442 2443 case SIOCSIFFIB: 2444 error = priv_check(td, PRIV_NET_SETIFFIB); 2445 if (error) 2446 return (error); 2447 if (ifr->ifr_fib >= rt_numfibs) 2448 return (EINVAL); 2449 2450 ifp->if_fib = ifr->ifr_fib; 2451 break; 2452 2453 case SIOCSIFFLAGS: 2454 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2455 if (error) 2456 return (error); 2457 /* 2458 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2459 * check, so we don't need special handling here yet. 2460 */ 2461 new_flags = (ifr->ifr_flags & 0xffff) | 2462 (ifr->ifr_flagshigh << 16); 2463 if (ifp->if_flags & IFF_UP && 2464 (new_flags & IFF_UP) == 0) { 2465 if_down(ifp); 2466 } else if (new_flags & IFF_UP && 2467 (ifp->if_flags & IFF_UP) == 0) { 2468 do_ifup = 1; 2469 } 2470 2471 /* 2472 * See if the promiscuous mode or allmulti bits are about to 2473 * flip. They require special handling because in-kernel 2474 * consumers may indepdently toggle them. 2475 */ 2476 if_setppromisc(ifp, new_flags & IFF_PPROMISC); 2477 if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) { 2478 if (new_flags & IFF_PALLMULTI) 2479 ifp->if_flags |= IFF_ALLMULTI; 2480 else if (ifp->if_amcount == 0) 2481 ifp->if_flags &= ~IFF_ALLMULTI; 2482 } 2483 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2484 (new_flags &~ IFF_CANTCHANGE); 2485 if (ifp->if_ioctl) { 2486 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2487 } 2488 if (do_ifup) 2489 if_up(ifp); 2490 getmicrotime(&ifp->if_lastchange); 2491 break; 2492 2493 case SIOCSIFCAP: 2494 error = priv_check(td, PRIV_NET_SETIFCAP); 2495 if (error != 0) 2496 return (error); 2497 if (ifp->if_ioctl == NULL) 2498 return (EOPNOTSUPP); 2499 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2500 return (EINVAL); 2501 error = (*ifp->if_ioctl)(ifp, cmd, data); 2502 if (error == 0) 2503 getmicrotime(&ifp->if_lastchange); 2504 break; 2505 2506 case SIOCSIFCAPNV: 2507 error = priv_check(td, PRIV_NET_SETIFCAP); 2508 if (error != 0) 2509 return (error); 2510 if (ifp->if_ioctl == NULL) 2511 return (EOPNOTSUPP); 2512 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2513 return (EINVAL); 2514 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2515 return (EINVAL); 2516 nvcap = NULL; 2517 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2518 for (;;) { 2519 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2520 ifr->ifr_cap_nv.length); 2521 if (error != 0) 2522 break; 2523 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2524 if (nvcap == NULL) { 2525 error = EINVAL; 2526 break; 2527 } 2528 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2529 &ifp->if_capenable, ifcap_nv_bit_names, false); 2530 if ((drv_ioctl_data.reqcap & 2531 ~ifp->if_capabilities) != 0) { 2532 error = EINVAL; 2533 break; 2534 } 2535 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2536 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2537 if ((drv_ioctl_data.reqcap2 & 2538 ~ifp->if_capabilities2) != 0) { 2539 error = EINVAL; 2540 break; 2541 } 2542 drv_ioctl_data.nvcap = nvcap; 2543 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2544 (caddr_t)&drv_ioctl_data); 2545 break; 2546 } 2547 nvlist_destroy(nvcap); 2548 free(buf, M_TEMP); 2549 if (error == 0) 2550 getmicrotime(&ifp->if_lastchange); 2551 break; 2552 2553 #ifdef MAC 2554 case SIOCSIFMAC: 2555 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2556 break; 2557 #endif 2558 2559 case SIOCSIFNAME: 2560 error = priv_check(td, PRIV_NET_SETIFNAME); 2561 if (error) 2562 return (error); 2563 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2564 NULL); 2565 if (error != 0) 2566 return (error); 2567 error = if_rename(ifp, new_name); 2568 break; 2569 2570 #ifdef VIMAGE 2571 case SIOCSIFVNET: 2572 error = priv_check(td, PRIV_NET_SETIFVNET); 2573 if (error) 2574 return (error); 2575 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2576 break; 2577 #endif 2578 2579 case SIOCSIFMETRIC: 2580 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2581 if (error) 2582 return (error); 2583 ifp->if_metric = ifr->ifr_metric; 2584 getmicrotime(&ifp->if_lastchange); 2585 break; 2586 2587 case SIOCSIFPHYS: 2588 error = priv_check(td, PRIV_NET_SETIFPHYS); 2589 if (error) 2590 return (error); 2591 if (ifp->if_ioctl == NULL) 2592 return (EOPNOTSUPP); 2593 error = (*ifp->if_ioctl)(ifp, cmd, data); 2594 if (error == 0) 2595 getmicrotime(&ifp->if_lastchange); 2596 break; 2597 2598 case SIOCSIFMTU: 2599 { 2600 u_long oldmtu = ifp->if_mtu; 2601 2602 error = priv_check(td, PRIV_NET_SETIFMTU); 2603 if (error) 2604 return (error); 2605 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2606 return (EINVAL); 2607 if (ifp->if_ioctl == NULL) 2608 return (EOPNOTSUPP); 2609 /* Disallow MTU changes on bridge member interfaces. */ 2610 if (ifp->if_bridge) 2611 return (EOPNOTSUPP); 2612 error = (*ifp->if_ioctl)(ifp, cmd, data); 2613 if (error == 0) { 2614 getmicrotime(&ifp->if_lastchange); 2615 rt_ifmsg(ifp, 0); 2616 #ifdef INET 2617 DEBUGNET_NOTIFY_MTU(ifp); 2618 #endif 2619 } 2620 /* 2621 * If the link MTU changed, do network layer specific procedure. 2622 */ 2623 if (ifp->if_mtu != oldmtu) 2624 if_notifymtu(ifp); 2625 break; 2626 } 2627 2628 case SIOCADDMULTI: 2629 case SIOCDELMULTI: 2630 if (cmd == SIOCADDMULTI) 2631 error = priv_check(td, PRIV_NET_ADDMULTI); 2632 else 2633 error = priv_check(td, PRIV_NET_DELMULTI); 2634 if (error) 2635 return (error); 2636 2637 /* Don't allow group membership on non-multicast interfaces. */ 2638 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2639 return (EOPNOTSUPP); 2640 2641 /* Don't let users screw up protocols' entries. */ 2642 if (ifr->ifr_addr.sa_family != AF_LINK) 2643 return (EINVAL); 2644 2645 if (cmd == SIOCADDMULTI) { 2646 struct epoch_tracker et; 2647 struct ifmultiaddr *ifma; 2648 2649 /* 2650 * Userland is only permitted to join groups once 2651 * via the if_addmulti() KPI, because it cannot hold 2652 * struct ifmultiaddr * between calls. It may also 2653 * lose a race while we check if the membership 2654 * already exists. 2655 */ 2656 NET_EPOCH_ENTER(et); 2657 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2658 NET_EPOCH_EXIT(et); 2659 if (ifma != NULL) 2660 error = EADDRINUSE; 2661 else 2662 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2663 } else { 2664 error = if_delmulti(ifp, &ifr->ifr_addr); 2665 } 2666 if (error == 0) 2667 getmicrotime(&ifp->if_lastchange); 2668 break; 2669 2670 case SIOCSIFPHYADDR: 2671 case SIOCDIFPHYADDR: 2672 #ifdef INET6 2673 case SIOCSIFPHYADDR_IN6: 2674 #endif 2675 case SIOCSIFMEDIA: 2676 case SIOCSIFGENERIC: 2677 error = priv_check(td, PRIV_NET_HWIOCTL); 2678 if (error) 2679 return (error); 2680 if (ifp->if_ioctl == NULL) 2681 return (EOPNOTSUPP); 2682 error = (*ifp->if_ioctl)(ifp, cmd, data); 2683 if (error == 0) 2684 getmicrotime(&ifp->if_lastchange); 2685 break; 2686 2687 case SIOCGIFSTATUS: 2688 case SIOCGIFPSRCADDR: 2689 case SIOCGIFPDSTADDR: 2690 case SIOCGIFMEDIA: 2691 case SIOCGIFXMEDIA: 2692 case SIOCGIFGENERIC: 2693 case SIOCGIFRSSKEY: 2694 case SIOCGIFRSSHASH: 2695 case SIOCGIFDOWNREASON: 2696 if (ifp->if_ioctl == NULL) 2697 return (EOPNOTSUPP); 2698 error = (*ifp->if_ioctl)(ifp, cmd, data); 2699 break; 2700 2701 case SIOCSIFLLADDR: 2702 error = priv_check(td, PRIV_NET_SETLLADDR); 2703 if (error) 2704 return (error); 2705 error = if_setlladdr(ifp, 2706 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2707 break; 2708 2709 case SIOCGHWADDR: 2710 error = if_gethwaddr(ifp, ifr); 2711 break; 2712 2713 case SIOCAIFGROUP: 2714 { 2715 const char *groupname; 2716 2717 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2718 if (error) 2719 return (error); 2720 groupname = ((struct ifgroupreq *)data)->ifgr_group; 2721 if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ) 2722 return (EINVAL); 2723 error = if_addgroup(ifp, groupname); 2724 if (error != 0) 2725 return (error); 2726 break; 2727 } 2728 case SIOCGIFGROUP: 2729 { 2730 struct epoch_tracker et; 2731 2732 NET_EPOCH_ENTER(et); 2733 error = if_getgroup((struct ifgroupreq *)data, ifp); 2734 NET_EPOCH_EXIT(et); 2735 break; 2736 } 2737 2738 case SIOCDIFGROUP: 2739 { 2740 const char *groupname; 2741 2742 error = priv_check(td, PRIV_NET_DELIFGROUP); 2743 if (error) 2744 return (error); 2745 groupname = ((struct ifgroupreq *)data)->ifgr_group; 2746 if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ) 2747 return (EINVAL); 2748 error = if_delgroup(ifp, groupname); 2749 if (error != 0) 2750 return (error); 2751 break; 2752 } 2753 default: 2754 error = ENOIOCTL; 2755 break; 2756 } 2757 return (error); 2758 } 2759 2760 /* 2761 * Interface ioctls. 2762 */ 2763 int 2764 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2765 { 2766 #ifdef COMPAT_FREEBSD32 2767 union { 2768 struct ifconf ifc; 2769 struct ifdrv ifd; 2770 struct ifgroupreq ifgr; 2771 struct ifmediareq ifmr; 2772 } thunk; 2773 u_long saved_cmd; 2774 struct ifconf32 *ifc32; 2775 struct ifdrv32 *ifd32; 2776 struct ifgroupreq32 *ifgr32; 2777 struct ifmediareq32 *ifmr32; 2778 #endif 2779 struct ifnet *ifp; 2780 struct ifreq *ifr; 2781 int error; 2782 int oif_flags; 2783 #ifdef VIMAGE 2784 bool shutdown; 2785 #endif 2786 2787 CURVNET_SET(so->so_vnet); 2788 #ifdef VIMAGE 2789 /* Make sure the VNET is stable. */ 2790 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2791 if (shutdown) { 2792 CURVNET_RESTORE(); 2793 return (EBUSY); 2794 } 2795 #endif 2796 2797 #ifdef COMPAT_FREEBSD32 2798 saved_cmd = cmd; 2799 switch (cmd) { 2800 case SIOCGIFCONF32: 2801 ifc32 = (struct ifconf32 *)data; 2802 thunk.ifc.ifc_len = ifc32->ifc_len; 2803 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2804 data = (caddr_t)&thunk.ifc; 2805 cmd = SIOCGIFCONF; 2806 break; 2807 case SIOCGDRVSPEC32: 2808 case SIOCSDRVSPEC32: 2809 ifd32 = (struct ifdrv32 *)data; 2810 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2811 sizeof(thunk.ifd.ifd_name)); 2812 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2813 thunk.ifd.ifd_len = ifd32->ifd_len; 2814 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2815 data = (caddr_t)&thunk.ifd; 2816 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2817 break; 2818 case SIOCAIFGROUP32: 2819 case SIOCGIFGROUP32: 2820 case SIOCDIFGROUP32: 2821 case SIOCGIFGMEMB32: 2822 ifgr32 = (struct ifgroupreq32 *)data; 2823 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2824 sizeof(thunk.ifgr.ifgr_name)); 2825 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2826 switch (cmd) { 2827 case SIOCAIFGROUP32: 2828 case SIOCDIFGROUP32: 2829 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2830 sizeof(thunk.ifgr.ifgr_group)); 2831 break; 2832 case SIOCGIFGROUP32: 2833 case SIOCGIFGMEMB32: 2834 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2835 break; 2836 } 2837 data = (caddr_t)&thunk.ifgr; 2838 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2839 break; 2840 case SIOCGIFMEDIA32: 2841 case SIOCGIFXMEDIA32: 2842 ifmr32 = (struct ifmediareq32 *)data; 2843 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2844 sizeof(thunk.ifmr.ifm_name)); 2845 thunk.ifmr.ifm_current = ifmr32->ifm_current; 2846 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 2847 thunk.ifmr.ifm_status = ifmr32->ifm_status; 2848 thunk.ifmr.ifm_active = ifmr32->ifm_active; 2849 thunk.ifmr.ifm_count = ifmr32->ifm_count; 2850 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 2851 data = (caddr_t)&thunk.ifmr; 2852 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 2853 break; 2854 } 2855 #endif 2856 2857 switch (cmd) { 2858 case SIOCGIFCONF: 2859 error = ifconf(cmd, data); 2860 goto out_noref; 2861 } 2862 2863 ifr = (struct ifreq *)data; 2864 switch (cmd) { 2865 #ifdef VIMAGE 2866 case SIOCSIFRVNET: 2867 error = priv_check(td, PRIV_NET_SETIFVNET); 2868 if (error == 0) 2869 error = if_vmove_reclaim(td, ifr->ifr_name, 2870 ifr->ifr_jid); 2871 goto out_noref; 2872 #endif 2873 case SIOCIFCREATE: 2874 case SIOCIFCREATE2: 2875 error = priv_check(td, PRIV_NET_IFCREATE); 2876 if (error == 0) 2877 error = if_clone_create(ifr->ifr_name, 2878 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 2879 ifr_data_get_ptr(ifr) : NULL); 2880 goto out_noref; 2881 case SIOCIFDESTROY: 2882 error = priv_check(td, PRIV_NET_IFDESTROY); 2883 2884 if (error == 0) { 2885 sx_xlock(&ifnet_detach_sxlock); 2886 error = if_clone_destroy(ifr->ifr_name); 2887 sx_xunlock(&ifnet_detach_sxlock); 2888 } 2889 goto out_noref; 2890 2891 case SIOCIFGCLONERS: 2892 error = if_clone_list((struct if_clonereq *)data); 2893 goto out_noref; 2894 2895 case SIOCGIFGMEMB: 2896 { 2897 struct ifgroupreq *req; 2898 2899 req = (struct ifgroupreq *)data; 2900 if (strnlen(req->ifgr_name, IFNAMSIZ) == IFNAMSIZ) { 2901 error = EINVAL; 2902 goto out_noref; 2903 } 2904 error = if_getgroupmembers(req); 2905 goto out_noref; 2906 } 2907 } 2908 2909 ifp = ifunit_ref(ifr->ifr_name); 2910 if (ifp == NULL) { 2911 error = ENXIO; 2912 goto out_noref; 2913 } 2914 2915 error = ifhwioctl(cmd, ifp, data, td); 2916 if (error != ENOIOCTL) 2917 goto out_ref; 2918 2919 oif_flags = ifp->if_flags; 2920 if (so->so_proto == NULL) { 2921 error = EOPNOTSUPP; 2922 goto out_ref; 2923 } 2924 2925 /* 2926 * Pass the request on to the socket control method, and if the 2927 * latter returns EOPNOTSUPP, directly to the interface. 2928 * 2929 * Make an exception for the legacy SIOCSIF* requests. Drivers 2930 * trust SIOCSIFADDR et al to come from an already privileged 2931 * layer, and do not perform any credentials checks or input 2932 * validation. 2933 */ 2934 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 2935 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2936 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2937 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2938 error = (*ifp->if_ioctl)(ifp, cmd, data); 2939 2940 if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP)) 2941 if_up(ifp); 2942 out_ref: 2943 if_rele(ifp); 2944 out_noref: 2945 CURVNET_RESTORE(); 2946 #ifdef COMPAT_FREEBSD32 2947 if (error != 0) 2948 return (error); 2949 switch (saved_cmd) { 2950 case SIOCGIFCONF32: 2951 ifc32->ifc_len = thunk.ifc.ifc_len; 2952 break; 2953 case SIOCGDRVSPEC32: 2954 /* 2955 * SIOCGDRVSPEC is IOWR, but nothing actually touches 2956 * the struct so just assert that ifd_len (the only 2957 * field it might make sense to update) hasn't 2958 * changed. 2959 */ 2960 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 2961 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 2962 thunk.ifd.ifd_len)); 2963 break; 2964 case SIOCGIFGROUP32: 2965 case SIOCGIFGMEMB32: 2966 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 2967 break; 2968 case SIOCGIFMEDIA32: 2969 case SIOCGIFXMEDIA32: 2970 ifmr32->ifm_current = thunk.ifmr.ifm_current; 2971 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 2972 ifmr32->ifm_status = thunk.ifmr.ifm_status; 2973 ifmr32->ifm_active = thunk.ifmr.ifm_active; 2974 ifmr32->ifm_count = thunk.ifmr.ifm_count; 2975 break; 2976 } 2977 #endif 2978 return (error); 2979 } 2980 2981 int 2982 if_rename(struct ifnet *ifp, char *new_name) 2983 { 2984 struct ifaddr *ifa; 2985 struct sockaddr_dl *sdl; 2986 size_t namelen, onamelen; 2987 char old_name[IFNAMSIZ]; 2988 char strbuf[IFNAMSIZ + 8]; 2989 2990 if (new_name[0] == '\0') 2991 return (EINVAL); 2992 if (strcmp(new_name, ifp->if_xname) == 0) 2993 return (0); 2994 if (ifunit(new_name) != NULL) 2995 return (EEXIST); 2996 2997 if_printf(ifp, "changing name to '%s'\n", new_name); 2998 2999 IF_ADDR_WLOCK(ifp); 3000 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 3001 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 3002 ifa = ifp->if_addr; 3003 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3004 namelen = strlen(new_name); 3005 onamelen = sdl->sdl_nlen; 3006 /* 3007 * Move the address if needed. This is safe because we 3008 * allocate space for a name of length IFNAMSIZ when we 3009 * create this in if_attach(). 3010 */ 3011 if (namelen != onamelen) { 3012 bcopy(sdl->sdl_data + onamelen, 3013 sdl->sdl_data + namelen, sdl->sdl_alen); 3014 } 3015 bcopy(new_name, sdl->sdl_data, namelen); 3016 sdl->sdl_nlen = namelen; 3017 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 3018 bzero(sdl->sdl_data, onamelen); 3019 while (namelen != 0) 3020 sdl->sdl_data[--namelen] = 0xff; 3021 IF_ADDR_WUNLOCK(ifp); 3022 3023 EVENTHANDLER_INVOKE(ifnet_rename_event, ifp, old_name); 3024 3025 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 3026 devctl_notify("IFNET", old_name, "RENAME", strbuf); 3027 3028 return (0); 3029 } 3030 3031 /* 3032 * The code common to handling reference counted flags, 3033 * e.g., in ifpromisc() and if_allmulti(). 3034 * The "pflag" argument can specify a permanent mode flag to check, 3035 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3036 * 3037 * Only to be used on stack-owned flags, not driver-owned flags. 3038 */ 3039 static int 3040 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3041 { 3042 struct ifreq ifr; 3043 int error; 3044 int oldflags, oldcount; 3045 3046 /* Sanity checks to catch programming errors */ 3047 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3048 ("%s: setting driver-owned flag %d", __func__, flag)); 3049 3050 if (onswitch) 3051 KASSERT(*refcount >= 0, 3052 ("%s: increment negative refcount %d for flag %d", 3053 __func__, *refcount, flag)); 3054 else 3055 KASSERT(*refcount > 0, 3056 ("%s: decrement non-positive refcount %d for flag %d", 3057 __func__, *refcount, flag)); 3058 3059 /* In case this mode is permanent, just touch refcount */ 3060 if (ifp->if_flags & pflag) { 3061 *refcount += onswitch ? 1 : -1; 3062 return (0); 3063 } 3064 3065 /* Save ifnet parameters for if_ioctl() may fail */ 3066 oldcount = *refcount; 3067 oldflags = ifp->if_flags; 3068 3069 /* 3070 * See if we aren't the only and touching refcount is enough. 3071 * Actually toggle interface flag if we are the first or last. 3072 */ 3073 if (onswitch) { 3074 if ((*refcount)++) 3075 return (0); 3076 ifp->if_flags |= flag; 3077 } else { 3078 if (--(*refcount)) 3079 return (0); 3080 ifp->if_flags &= ~flag; 3081 } 3082 3083 /* Call down the driver since we've changed interface flags */ 3084 if (ifp->if_ioctl == NULL) { 3085 error = EOPNOTSUPP; 3086 goto recover; 3087 } 3088 ifr.ifr_flags = ifp->if_flags & 0xffff; 3089 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3090 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3091 if (error) 3092 goto recover; 3093 /* Notify userland that interface flags have changed */ 3094 rt_ifmsg(ifp, flag); 3095 return (0); 3096 3097 recover: 3098 /* Recover after driver error */ 3099 *refcount = oldcount; 3100 ifp->if_flags = oldflags; 3101 return (error); 3102 } 3103 3104 /* 3105 * Set/clear promiscuous mode on interface ifp based on the truth value 3106 * of pswitch. The calls are reference counted so that only the first 3107 * "on" request actually has an effect, as does the final "off" request. 3108 * Results are undefined if the "off" and "on" requests are not matched. 3109 */ 3110 int 3111 ifpromisc(struct ifnet *ifp, int pswitch) 3112 { 3113 int error; 3114 int oldflags = ifp->if_flags; 3115 3116 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3117 &ifp->if_pcount, pswitch); 3118 /* If promiscuous mode status has changed, log a message */ 3119 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3120 log_promisc_mode_change) 3121 if_printf(ifp, "promiscuous mode %s\n", 3122 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3123 return (error); 3124 } 3125 3126 /* 3127 * Return interface configuration 3128 * of system. List may be used 3129 * in later ioctl's (above) to get 3130 * other information. 3131 */ 3132 /*ARGSUSED*/ 3133 static int 3134 ifconf(u_long cmd, caddr_t data) 3135 { 3136 struct ifconf *ifc = (struct ifconf *)data; 3137 struct ifnet *ifp; 3138 struct ifaddr *ifa; 3139 struct ifreq ifr; 3140 struct sbuf *sb; 3141 int error, full = 0, valid_len, max_len; 3142 3143 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3144 max_len = maxphys - 1; 3145 3146 /* Prevent hostile input from being able to crash the system */ 3147 if (ifc->ifc_len <= 0) 3148 return (EINVAL); 3149 3150 again: 3151 if (ifc->ifc_len <= max_len) { 3152 max_len = ifc->ifc_len; 3153 full = 1; 3154 } 3155 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3156 max_len = 0; 3157 valid_len = 0; 3158 3159 IFNET_RLOCK(); 3160 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3161 struct epoch_tracker et; 3162 int addrs; 3163 3164 /* 3165 * Zero the ifr to make sure we don't disclose the contents 3166 * of the stack. 3167 */ 3168 memset(&ifr, 0, sizeof(ifr)); 3169 3170 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3171 >= sizeof(ifr.ifr_name)) { 3172 sbuf_delete(sb); 3173 IFNET_RUNLOCK(); 3174 return (ENAMETOOLONG); 3175 } 3176 3177 addrs = 0; 3178 NET_EPOCH_ENTER(et); 3179 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3180 struct sockaddr *sa = ifa->ifa_addr; 3181 3182 if (prison_if(curthread->td_ucred, sa) != 0) 3183 continue; 3184 addrs++; 3185 if (sa->sa_len <= sizeof(*sa)) { 3186 if (sa->sa_len < sizeof(*sa)) { 3187 memset(&ifr.ifr_ifru.ifru_addr, 0, 3188 sizeof(ifr.ifr_ifru.ifru_addr)); 3189 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3190 sa->sa_len); 3191 } else 3192 ifr.ifr_ifru.ifru_addr = *sa; 3193 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3194 max_len += sizeof(ifr); 3195 } else { 3196 sbuf_bcat(sb, &ifr, 3197 offsetof(struct ifreq, ifr_addr)); 3198 max_len += offsetof(struct ifreq, ifr_addr); 3199 sbuf_bcat(sb, sa, sa->sa_len); 3200 max_len += sa->sa_len; 3201 } 3202 3203 if (sbuf_error(sb) == 0) 3204 valid_len = sbuf_len(sb); 3205 } 3206 NET_EPOCH_EXIT(et); 3207 if (addrs == 0) { 3208 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3209 max_len += sizeof(ifr); 3210 3211 if (sbuf_error(sb) == 0) 3212 valid_len = sbuf_len(sb); 3213 } 3214 } 3215 IFNET_RUNLOCK(); 3216 3217 /* 3218 * If we didn't allocate enough space (uncommon), try again. If 3219 * we have already allocated as much space as we are allowed, 3220 * return what we've got. 3221 */ 3222 if (valid_len != max_len && !full) { 3223 sbuf_delete(sb); 3224 goto again; 3225 } 3226 3227 ifc->ifc_len = valid_len; 3228 sbuf_finish(sb); 3229 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3230 sbuf_delete(sb); 3231 return (error); 3232 } 3233 3234 /* 3235 * Just like ifpromisc(), but for all-multicast-reception mode. 3236 */ 3237 int 3238 if_allmulti(struct ifnet *ifp, int onswitch) 3239 { 3240 3241 return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount, 3242 onswitch)); 3243 } 3244 3245 struct ifmultiaddr * 3246 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3247 { 3248 struct ifmultiaddr *ifma; 3249 3250 IF_ADDR_LOCK_ASSERT(ifp); 3251 3252 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3253 if (sa->sa_family == AF_LINK) { 3254 if (sa_dl_equal(ifma->ifma_addr, sa)) 3255 break; 3256 } else { 3257 if (sa_equal(ifma->ifma_addr, sa)) 3258 break; 3259 } 3260 } 3261 3262 return ifma; 3263 } 3264 3265 /* 3266 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3267 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3268 * the ifnet multicast address list here, so the caller must do that and 3269 * other setup work (such as notifying the device driver). The reference 3270 * count is initialized to 1. 3271 */ 3272 static struct ifmultiaddr * 3273 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3274 int mflags) 3275 { 3276 struct ifmultiaddr *ifma; 3277 struct sockaddr *dupsa; 3278 3279 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3280 M_ZERO); 3281 if (ifma == NULL) 3282 return (NULL); 3283 3284 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3285 if (dupsa == NULL) { 3286 free(ifma, M_IFMADDR); 3287 return (NULL); 3288 } 3289 bcopy(sa, dupsa, sa->sa_len); 3290 ifma->ifma_addr = dupsa; 3291 3292 ifma->ifma_ifp = ifp; 3293 ifma->ifma_refcount = 1; 3294 ifma->ifma_protospec = NULL; 3295 3296 if (llsa == NULL) { 3297 ifma->ifma_lladdr = NULL; 3298 return (ifma); 3299 } 3300 3301 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3302 if (dupsa == NULL) { 3303 free(ifma->ifma_addr, M_IFMADDR); 3304 free(ifma, M_IFMADDR); 3305 return (NULL); 3306 } 3307 bcopy(llsa, dupsa, llsa->sa_len); 3308 ifma->ifma_lladdr = dupsa; 3309 3310 return (ifma); 3311 } 3312 3313 /* 3314 * if_freemulti: free ifmultiaddr structure and possibly attached related 3315 * addresses. The caller is responsible for implementing reference 3316 * counting, notifying the driver, handling routing messages, and releasing 3317 * any dependent link layer state. 3318 */ 3319 #ifdef MCAST_VERBOSE 3320 extern void kdb_backtrace(void); 3321 #endif 3322 static void 3323 if_freemulti_internal(struct ifmultiaddr *ifma) 3324 { 3325 3326 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3327 ifma->ifma_refcount)); 3328 3329 if (ifma->ifma_lladdr != NULL) 3330 free(ifma->ifma_lladdr, M_IFMADDR); 3331 #ifdef MCAST_VERBOSE 3332 kdb_backtrace(); 3333 printf("%s freeing ifma: %p\n", __func__, ifma); 3334 #endif 3335 free(ifma->ifma_addr, M_IFMADDR); 3336 free(ifma, M_IFMADDR); 3337 } 3338 3339 static void 3340 if_destroymulti(epoch_context_t ctx) 3341 { 3342 struct ifmultiaddr *ifma; 3343 3344 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3345 if_freemulti_internal(ifma); 3346 } 3347 3348 void 3349 if_freemulti(struct ifmultiaddr *ifma) 3350 { 3351 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3352 ifma->ifma_refcount)); 3353 3354 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3355 } 3356 3357 /* 3358 * Register an additional multicast address with a network interface. 3359 * 3360 * - If the address is already present, bump the reference count on the 3361 * address and return. 3362 * - If the address is not link-layer, look up a link layer address. 3363 * - Allocate address structures for one or both addresses, and attach to the 3364 * multicast address list on the interface. If automatically adding a link 3365 * layer address, the protocol address will own a reference to the link 3366 * layer address, to be freed when it is freed. 3367 * - Notify the network device driver of an addition to the multicast address 3368 * list. 3369 * 3370 * 'sa' points to caller-owned memory with the desired multicast address. 3371 * 3372 * 'retifma' will be used to return a pointer to the resulting multicast 3373 * address reference, if desired. 3374 */ 3375 int 3376 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3377 struct ifmultiaddr **retifma) 3378 { 3379 struct ifmultiaddr *ifma, *ll_ifma; 3380 struct sockaddr *llsa; 3381 struct sockaddr_dl sdl; 3382 int error; 3383 3384 #ifdef INET 3385 IN_MULTI_LIST_UNLOCK_ASSERT(); 3386 #endif 3387 #ifdef INET6 3388 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3389 #endif 3390 /* 3391 * If the address is already present, return a new reference to it; 3392 * otherwise, allocate storage and set up a new address. 3393 */ 3394 IF_ADDR_WLOCK(ifp); 3395 ifma = if_findmulti(ifp, sa); 3396 if (ifma != NULL) { 3397 ifma->ifma_refcount++; 3398 if (retifma != NULL) 3399 *retifma = ifma; 3400 IF_ADDR_WUNLOCK(ifp); 3401 return (0); 3402 } 3403 3404 /* 3405 * The address isn't already present; resolve the protocol address 3406 * into a link layer address, and then look that up, bump its 3407 * refcount or allocate an ifma for that also. 3408 * Most link layer resolving functions returns address data which 3409 * fits inside default sockaddr_dl structure. However callback 3410 * can allocate another sockaddr structure, in that case we need to 3411 * free it later. 3412 */ 3413 llsa = NULL; 3414 ll_ifma = NULL; 3415 if (ifp->if_resolvemulti != NULL) { 3416 /* Provide called function with buffer size information */ 3417 sdl.sdl_len = sizeof(sdl); 3418 llsa = (struct sockaddr *)&sdl; 3419 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3420 if (error) 3421 goto unlock_out; 3422 } 3423 3424 /* 3425 * Allocate the new address. Don't hook it up yet, as we may also 3426 * need to allocate a link layer multicast address. 3427 */ 3428 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3429 if (ifma == NULL) { 3430 error = ENOMEM; 3431 goto free_llsa_out; 3432 } 3433 3434 /* 3435 * If a link layer address is found, we'll need to see if it's 3436 * already present in the address list, or allocate is as well. 3437 * When this block finishes, the link layer address will be on the 3438 * list. 3439 */ 3440 if (llsa != NULL) { 3441 ll_ifma = if_findmulti(ifp, llsa); 3442 if (ll_ifma == NULL) { 3443 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3444 if (ll_ifma == NULL) { 3445 --ifma->ifma_refcount; 3446 if_freemulti(ifma); 3447 error = ENOMEM; 3448 goto free_llsa_out; 3449 } 3450 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3451 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3452 ifma_link); 3453 } else 3454 ll_ifma->ifma_refcount++; 3455 ifma->ifma_llifma = ll_ifma; 3456 } 3457 3458 /* 3459 * We now have a new multicast address, ifma, and possibly a new or 3460 * referenced link layer address. Add the primary address to the 3461 * ifnet address list. 3462 */ 3463 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3464 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3465 3466 if (retifma != NULL) 3467 *retifma = ifma; 3468 3469 /* 3470 * Must generate the message while holding the lock so that 'ifma' 3471 * pointer is still valid. 3472 */ 3473 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3474 IF_ADDR_WUNLOCK(ifp); 3475 3476 /* 3477 * We are certain we have added something, so call down to the 3478 * interface to let them know about it. 3479 */ 3480 if (ifp->if_ioctl != NULL) { 3481 if (THREAD_CAN_SLEEP()) 3482 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3483 else 3484 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3485 } 3486 3487 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3488 link_free_sdl(llsa); 3489 3490 return (0); 3491 3492 free_llsa_out: 3493 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3494 link_free_sdl(llsa); 3495 3496 unlock_out: 3497 IF_ADDR_WUNLOCK(ifp); 3498 return (error); 3499 } 3500 3501 static void 3502 if_siocaddmulti(void *arg, int pending) 3503 { 3504 struct ifnet *ifp; 3505 3506 ifp = arg; 3507 #ifdef DIAGNOSTIC 3508 if (pending > 1) 3509 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3510 #endif 3511 CURVNET_SET(ifp->if_vnet); 3512 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3513 CURVNET_RESTORE(); 3514 } 3515 3516 /* 3517 * Delete a multicast group membership by network-layer group address. 3518 * 3519 * Returns ENOENT if the entry could not be found. If ifp no longer 3520 * exists, results are undefined. This entry point should only be used 3521 * from subsystems which do appropriate locking to hold ifp for the 3522 * duration of the call. 3523 * Network-layer protocol domains must use if_delmulti_ifma(). 3524 */ 3525 int 3526 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3527 { 3528 struct ifmultiaddr *ifma; 3529 int lastref; 3530 3531 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3532 3533 IF_ADDR_WLOCK(ifp); 3534 lastref = 0; 3535 ifma = if_findmulti(ifp, sa); 3536 if (ifma != NULL) 3537 lastref = if_delmulti_locked(ifp, ifma, 0); 3538 IF_ADDR_WUNLOCK(ifp); 3539 3540 if (ifma == NULL) 3541 return (ENOENT); 3542 3543 if (lastref && ifp->if_ioctl != NULL) { 3544 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3545 } 3546 3547 return (0); 3548 } 3549 3550 /* 3551 * Delete all multicast group membership for an interface. 3552 * Should be used to quickly flush all multicast filters. 3553 */ 3554 void 3555 if_delallmulti(struct ifnet *ifp) 3556 { 3557 struct ifmultiaddr *ifma; 3558 struct ifmultiaddr *next; 3559 3560 IF_ADDR_WLOCK(ifp); 3561 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3562 if_delmulti_locked(ifp, ifma, 0); 3563 IF_ADDR_WUNLOCK(ifp); 3564 } 3565 3566 void 3567 if_delmulti_ifma(struct ifmultiaddr *ifma) 3568 { 3569 if_delmulti_ifma_flags(ifma, 0); 3570 } 3571 3572 /* 3573 * Delete a multicast group membership by group membership pointer. 3574 * Network-layer protocol domains must use this routine. 3575 * 3576 * It is safe to call this routine if the ifp disappeared. 3577 */ 3578 void 3579 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3580 { 3581 struct ifnet *ifp; 3582 int lastref; 3583 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3584 #ifdef INET 3585 IN_MULTI_LIST_UNLOCK_ASSERT(); 3586 #endif 3587 ifp = ifma->ifma_ifp; 3588 #ifdef DIAGNOSTIC 3589 if (ifp == NULL) { 3590 printf("%s: ifma_ifp seems to be detached\n", __func__); 3591 } else { 3592 struct epoch_tracker et; 3593 struct ifnet *oifp; 3594 3595 NET_EPOCH_ENTER(et); 3596 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3597 if (ifp == oifp) 3598 break; 3599 NET_EPOCH_EXIT(et); 3600 if (ifp != oifp) 3601 ifp = NULL; 3602 } 3603 #endif 3604 /* 3605 * If and only if the ifnet instance exists: Acquire the address lock. 3606 */ 3607 if (ifp != NULL) 3608 IF_ADDR_WLOCK(ifp); 3609 3610 lastref = if_delmulti_locked(ifp, ifma, flags); 3611 3612 if (ifp != NULL) { 3613 /* 3614 * If and only if the ifnet instance exists: 3615 * Release the address lock. 3616 * If the group was left: update the hardware hash filter. 3617 */ 3618 IF_ADDR_WUNLOCK(ifp); 3619 if (lastref && ifp->if_ioctl != NULL) { 3620 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3621 } 3622 } 3623 } 3624 3625 /* 3626 * Perform deletion of network-layer and/or link-layer multicast address. 3627 * 3628 * Return 0 if the reference count was decremented. 3629 * Return 1 if the final reference was released, indicating that the 3630 * hardware hash filter should be reprogrammed. 3631 */ 3632 static int 3633 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3634 { 3635 struct ifmultiaddr *ll_ifma; 3636 3637 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3638 KASSERT(ifma->ifma_ifp == ifp, 3639 ("%s: inconsistent ifp %p", __func__, ifp)); 3640 IF_ADDR_WLOCK_ASSERT(ifp); 3641 } 3642 3643 ifp = ifma->ifma_ifp; 3644 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3645 3646 /* 3647 * If the ifnet is detaching, null out references to ifnet, 3648 * so that upper protocol layers will notice, and not attempt 3649 * to obtain locks for an ifnet which no longer exists. The 3650 * routing socket announcement must happen before the ifnet 3651 * instance is detached from the system. 3652 */ 3653 if (detaching) { 3654 #ifdef DIAGNOSTIC 3655 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3656 #endif 3657 /* 3658 * ifp may already be nulled out if we are being reentered 3659 * to delete the ll_ifma. 3660 */ 3661 if (ifp != NULL) { 3662 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3663 ifma->ifma_ifp = NULL; 3664 } 3665 } 3666 3667 if (--ifma->ifma_refcount > 0) 3668 return 0; 3669 3670 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3671 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3672 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3673 } 3674 /* 3675 * If this ifma is a network-layer ifma, a link-layer ifma may 3676 * have been associated with it. Release it first if so. 3677 */ 3678 ll_ifma = ifma->ifma_llifma; 3679 if (ll_ifma != NULL) { 3680 KASSERT(ifma->ifma_lladdr != NULL, 3681 ("%s: llifma w/o lladdr", __func__)); 3682 if (detaching) 3683 ll_ifma->ifma_ifp = NULL; /* XXX */ 3684 if (--ll_ifma->ifma_refcount == 0) { 3685 if (ifp != NULL) { 3686 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3687 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3688 ifma_link); 3689 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3690 } 3691 } 3692 if_freemulti(ll_ifma); 3693 } 3694 } 3695 #ifdef INVARIANTS 3696 if (ifp) { 3697 struct ifmultiaddr *ifmatmp; 3698 3699 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3700 MPASS(ifma != ifmatmp); 3701 } 3702 #endif 3703 if_freemulti(ifma); 3704 /* 3705 * The last reference to this instance of struct ifmultiaddr 3706 * was released; the hardware should be notified of this change. 3707 */ 3708 return 1; 3709 } 3710 3711 /* 3712 * Set the link layer address on an interface. 3713 * 3714 * At this time we only support certain types of interfaces, 3715 * and we don't allow the length of the address to change. 3716 * 3717 * Set noinline to be dtrace-friendly 3718 */ 3719 __noinline int 3720 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3721 { 3722 struct sockaddr_dl *sdl; 3723 struct ifaddr *ifa; 3724 struct ifreq ifr; 3725 3726 ifa = ifp->if_addr; 3727 if (ifa == NULL) 3728 return (EINVAL); 3729 3730 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3731 if (sdl == NULL) 3732 return (EINVAL); 3733 3734 if (len != sdl->sdl_alen) /* don't allow length to change */ 3735 return (EINVAL); 3736 3737 switch (ifp->if_type) { 3738 case IFT_ETHER: 3739 case IFT_XETHER: 3740 case IFT_L2VLAN: 3741 case IFT_BRIDGE: 3742 case IFT_IEEE8023ADLAG: 3743 bcopy(lladdr, LLADDR(sdl), len); 3744 break; 3745 default: 3746 return (ENODEV); 3747 } 3748 3749 /* 3750 * If the interface is already up, we need 3751 * to re-init it in order to reprogram its 3752 * address filter. 3753 */ 3754 if ((ifp->if_flags & IFF_UP) != 0) { 3755 if (ifp->if_ioctl) { 3756 ifp->if_flags &= ~IFF_UP; 3757 ifr.ifr_flags = ifp->if_flags & 0xffff; 3758 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3759 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3760 ifp->if_flags |= IFF_UP; 3761 ifr.ifr_flags = ifp->if_flags & 0xffff; 3762 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3763 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3764 } 3765 } 3766 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3767 3768 return (0); 3769 } 3770 3771 /* 3772 * Compat function for handling basic encapsulation requests. 3773 * Not converted stacks (FDDI, IB, ..) supports traditional 3774 * output model: ARP (and other similar L2 protocols) are handled 3775 * inside output routine, arpresolve/nd6_resolve() returns MAC 3776 * address instead of full prepend. 3777 * 3778 * This function creates calculated header==MAC for IPv4/IPv6 and 3779 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3780 * address families. 3781 */ 3782 static int 3783 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3784 { 3785 if (req->rtype != IFENCAP_LL) 3786 return (EOPNOTSUPP); 3787 3788 if (req->bufsize < req->lladdr_len) 3789 return (ENOMEM); 3790 3791 switch (req->family) { 3792 case AF_INET: 3793 case AF_INET6: 3794 break; 3795 default: 3796 return (EAFNOSUPPORT); 3797 } 3798 3799 /* Copy lladdr to storage as is */ 3800 memmove(req->buf, req->lladdr, req->lladdr_len); 3801 req->bufsize = req->lladdr_len; 3802 req->lladdr_off = 0; 3803 3804 return (0); 3805 } 3806 3807 /* 3808 * Tunnel interfaces can nest, also they may cause infinite recursion 3809 * calls when misconfigured. We'll prevent this by detecting loops. 3810 * High nesting level may cause stack exhaustion. We'll prevent this 3811 * by introducing upper limit. 3812 * 3813 * Return 0, if tunnel nesting count is equal or less than limit. 3814 */ 3815 int 3816 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3817 int limit) 3818 { 3819 struct m_tag *mtag; 3820 int count; 3821 3822 count = 1; 3823 mtag = NULL; 3824 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3825 if (*(struct ifnet **)(mtag + 1) == ifp) { 3826 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3827 return (EIO); 3828 } 3829 count++; 3830 } 3831 if (count > limit) { 3832 log(LOG_NOTICE, 3833 "%s: if_output recursively called too many times(%d)\n", 3834 if_name(ifp), count); 3835 return (EIO); 3836 } 3837 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3838 if (mtag == NULL) 3839 return (ENOMEM); 3840 *(struct ifnet **)(mtag + 1) = ifp; 3841 m_tag_prepend(m, mtag); 3842 return (0); 3843 } 3844 3845 /* 3846 * Get the link layer address that was read from the hardware at attach. 3847 * 3848 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3849 * their component interfaces as IFT_IEEE8023ADLAG. 3850 */ 3851 int 3852 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3853 { 3854 if (ifp->if_hw_addr == NULL) 3855 return (ENODEV); 3856 3857 switch (ifp->if_type) { 3858 case IFT_ETHER: 3859 case IFT_IEEE8023ADLAG: 3860 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3861 return (0); 3862 default: 3863 return (ENODEV); 3864 } 3865 } 3866 3867 /* 3868 * The name argument must be a pointer to storage which will last as 3869 * long as the interface does. For physical devices, the result of 3870 * device_get_name(dev) is a good choice and for pseudo-devices a 3871 * static string works well. 3872 */ 3873 void 3874 if_initname(struct ifnet *ifp, const char *name, int unit) 3875 { 3876 ifp->if_dname = name; 3877 ifp->if_dunit = unit; 3878 if (unit != IF_DUNIT_NONE) 3879 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3880 else 3881 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3882 } 3883 3884 static int 3885 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 3886 { 3887 char if_fmt[256]; 3888 3889 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 3890 vlog(pri, if_fmt, ap); 3891 return (0); 3892 } 3893 3894 3895 int 3896 if_printf(struct ifnet *ifp, const char *fmt, ...) 3897 { 3898 va_list ap; 3899 3900 va_start(ap, fmt); 3901 if_vlog(ifp, LOG_INFO, fmt, ap); 3902 va_end(ap); 3903 return (0); 3904 } 3905 3906 int 3907 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 3908 { 3909 va_list ap; 3910 3911 va_start(ap, fmt); 3912 if_vlog(ifp, pri, fmt, ap); 3913 va_end(ap); 3914 return (0); 3915 } 3916 3917 void 3918 if_start(struct ifnet *ifp) 3919 { 3920 3921 (*(ifp)->if_start)(ifp); 3922 } 3923 3924 /* 3925 * Backwards compatibility interface for drivers 3926 * that have not implemented it 3927 */ 3928 static int 3929 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 3930 { 3931 int error; 3932 3933 IFQ_HANDOFF(ifp, m, error); 3934 return (error); 3935 } 3936 3937 static void 3938 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3939 { 3940 m_freem(m); 3941 } 3942 3943 int 3944 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3945 { 3946 int active = 0; 3947 3948 IF_LOCK(ifq); 3949 if (_IF_QFULL(ifq)) { 3950 IF_UNLOCK(ifq); 3951 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 3952 m_freem(m); 3953 return (0); 3954 } 3955 if (ifp != NULL) { 3956 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 3957 if (m->m_flags & (M_BCAST|M_MCAST)) 3958 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 3959 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3960 } 3961 _IF_ENQUEUE(ifq, m); 3962 IF_UNLOCK(ifq); 3963 if (ifp != NULL && !active) 3964 (*(ifp)->if_start)(ifp); 3965 return (1); 3966 } 3967 3968 void 3969 if_register_com_alloc(u_char type, 3970 if_com_alloc_t *a, if_com_free_t *f) 3971 { 3972 3973 KASSERT(if_com_alloc[type] == NULL, 3974 ("if_register_com_alloc: %d already registered", type)); 3975 KASSERT(if_com_free[type] == NULL, 3976 ("if_register_com_alloc: %d free already registered", type)); 3977 3978 if_com_alloc[type] = a; 3979 if_com_free[type] = f; 3980 } 3981 3982 void 3983 if_deregister_com_alloc(u_char type) 3984 { 3985 3986 KASSERT(if_com_alloc[type] != NULL, 3987 ("if_deregister_com_alloc: %d not registered", type)); 3988 KASSERT(if_com_free[type] != NULL, 3989 ("if_deregister_com_alloc: %d free not registered", type)); 3990 3991 /* 3992 * Ensure all pending EPOCH(9) callbacks have been executed. This 3993 * fixes issues about late invocation of if_destroy(), which leads 3994 * to memory leak from if_com_alloc[type] allocated if_l2com. 3995 */ 3996 NET_EPOCH_DRAIN_CALLBACKS(); 3997 3998 if_com_alloc[type] = NULL; 3999 if_com_free[type] = NULL; 4000 } 4001 4002 /* API for driver access to network stack owned ifnet.*/ 4003 uint64_t 4004 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4005 { 4006 uint64_t oldbrate; 4007 4008 oldbrate = ifp->if_baudrate; 4009 ifp->if_baudrate = baudrate; 4010 return (oldbrate); 4011 } 4012 4013 uint64_t 4014 if_getbaudrate(const if_t ifp) 4015 { 4016 return (ifp->if_baudrate); 4017 } 4018 4019 int 4020 if_setcapabilities(if_t ifp, int capabilities) 4021 { 4022 ifp->if_capabilities = capabilities; 4023 return (0); 4024 } 4025 4026 int 4027 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4028 { 4029 ifp->if_capabilities &= ~clearbit; 4030 ifp->if_capabilities |= setbit; 4031 return (0); 4032 } 4033 4034 int 4035 if_getcapabilities(const if_t ifp) 4036 { 4037 return (ifp->if_capabilities); 4038 } 4039 4040 int 4041 if_setcapenable(if_t ifp, int capabilities) 4042 { 4043 ifp->if_capenable = capabilities; 4044 return (0); 4045 } 4046 4047 int 4048 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4049 { 4050 ifp->if_capenable &= ~clearcap; 4051 ifp->if_capenable |= setcap; 4052 return (0); 4053 } 4054 4055 int 4056 if_setcapabilities2(if_t ifp, int capabilities) 4057 { 4058 ifp->if_capabilities2 = capabilities; 4059 return (0); 4060 } 4061 4062 int 4063 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit) 4064 { 4065 ifp->if_capabilities2 &= ~clearbit; 4066 ifp->if_capabilities2 |= setbit; 4067 return (0); 4068 } 4069 4070 int 4071 if_getcapabilities2(const if_t ifp) 4072 { 4073 return (ifp->if_capabilities2); 4074 } 4075 4076 int 4077 if_setcapenable2(if_t ifp, int capabilities2) 4078 { 4079 ifp->if_capenable2 = capabilities2; 4080 return (0); 4081 } 4082 4083 int 4084 if_setcapenable2bit(if_t ifp, int setcap, int clearcap) 4085 { 4086 ifp->if_capenable2 &= ~clearcap; 4087 ifp->if_capenable2 |= setcap; 4088 return (0); 4089 } 4090 4091 const char * 4092 if_getdname(const if_t ifp) 4093 { 4094 return (ifp->if_dname); 4095 } 4096 4097 void 4098 if_setdname(if_t ifp, const char *dname) 4099 { 4100 ifp->if_dname = dname; 4101 } 4102 4103 const char * 4104 if_name(if_t ifp) 4105 { 4106 return (ifp->if_xname); 4107 } 4108 4109 int 4110 if_setname(if_t ifp, const char *name) 4111 { 4112 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4113 return (ENAMETOOLONG); 4114 strcpy(ifp->if_xname, name); 4115 4116 return (0); 4117 } 4118 4119 int 4120 if_togglecapenable(if_t ifp, int togglecap) 4121 { 4122 ifp->if_capenable ^= togglecap; 4123 return (0); 4124 } 4125 4126 int 4127 if_getcapenable(const if_t ifp) 4128 { 4129 return (ifp->if_capenable); 4130 } 4131 4132 int 4133 if_togglecapenable2(if_t ifp, int togglecap) 4134 { 4135 ifp->if_capenable2 ^= togglecap; 4136 return (0); 4137 } 4138 4139 int 4140 if_getcapenable2(const if_t ifp) 4141 { 4142 return (ifp->if_capenable2); 4143 } 4144 4145 int 4146 if_getdunit(const if_t ifp) 4147 { 4148 return (ifp->if_dunit); 4149 } 4150 4151 int 4152 if_getindex(const if_t ifp) 4153 { 4154 return (ifp->if_index); 4155 } 4156 4157 int 4158 if_getidxgen(const if_t ifp) 4159 { 4160 return (ifp->if_idxgen); 4161 } 4162 4163 const char * 4164 if_getdescr(if_t ifp) 4165 { 4166 return (ifp->if_description); 4167 } 4168 4169 void 4170 if_setdescr(if_t ifp, char *descrbuf) 4171 { 4172 sx_xlock(&ifdescr_sx); 4173 char *odescrbuf = ifp->if_description; 4174 ifp->if_description = descrbuf; 4175 sx_xunlock(&ifdescr_sx); 4176 4177 if_freedescr(odescrbuf); 4178 } 4179 4180 char * 4181 if_allocdescr(size_t sz, int malloc_flag) 4182 { 4183 malloc_flag &= (M_WAITOK | M_NOWAIT); 4184 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4185 } 4186 4187 void 4188 if_freedescr(char *descrbuf) 4189 { 4190 free(descrbuf, M_IFDESCR); 4191 } 4192 4193 int 4194 if_getalloctype(const if_t ifp) 4195 { 4196 return (ifp->if_alloctype); 4197 } 4198 4199 void 4200 if_setlastchange(if_t ifp) 4201 { 4202 getmicrotime(&ifp->if_lastchange); 4203 } 4204 4205 /* 4206 * This is largely undesirable because it ties ifnet to a device, but does 4207 * provide flexiblity for an embedded product vendor. Should be used with 4208 * the understanding that it violates the interface boundaries, and should be 4209 * a last resort only. 4210 */ 4211 int 4212 if_setdev(if_t ifp, void *dev) 4213 { 4214 return (0); 4215 } 4216 4217 int 4218 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4219 { 4220 ifp->if_drv_flags &= ~clear_flags; 4221 ifp->if_drv_flags |= set_flags; 4222 4223 return (0); 4224 } 4225 4226 int 4227 if_getdrvflags(const if_t ifp) 4228 { 4229 return (ifp->if_drv_flags); 4230 } 4231 4232 int 4233 if_setdrvflags(if_t ifp, int flags) 4234 { 4235 ifp->if_drv_flags = flags; 4236 return (0); 4237 } 4238 4239 int 4240 if_setflags(if_t ifp, int flags) 4241 { 4242 ifp->if_flags = flags; 4243 return (0); 4244 } 4245 4246 int 4247 if_setflagbits(if_t ifp, int set, int clear) 4248 { 4249 ifp->if_flags &= ~clear; 4250 ifp->if_flags |= set; 4251 return (0); 4252 } 4253 4254 int 4255 if_getflags(const if_t ifp) 4256 { 4257 return (ifp->if_flags); 4258 } 4259 4260 int 4261 if_clearhwassist(if_t ifp) 4262 { 4263 ifp->if_hwassist = 0; 4264 return (0); 4265 } 4266 4267 int 4268 if_sethwassistbits(if_t ifp, int toset, int toclear) 4269 { 4270 ifp->if_hwassist &= ~toclear; 4271 ifp->if_hwassist |= toset; 4272 4273 return (0); 4274 } 4275 4276 int 4277 if_sethwassist(if_t ifp, int hwassist_bit) 4278 { 4279 ifp->if_hwassist = hwassist_bit; 4280 return (0); 4281 } 4282 4283 int 4284 if_gethwassist(const if_t ifp) 4285 { 4286 return (ifp->if_hwassist); 4287 } 4288 4289 int 4290 if_togglehwassist(if_t ifp, int toggle_bits) 4291 { 4292 ifp->if_hwassist ^= toggle_bits; 4293 return (0); 4294 } 4295 4296 int 4297 if_setmtu(if_t ifp, int mtu) 4298 { 4299 ifp->if_mtu = mtu; 4300 return (0); 4301 } 4302 4303 void 4304 if_notifymtu(if_t ifp) 4305 { 4306 #ifdef INET6 4307 nd6_setmtu(ifp); 4308 #endif 4309 rt_updatemtu(ifp); 4310 } 4311 4312 int 4313 if_getmtu(const if_t ifp) 4314 { 4315 return (ifp->if_mtu); 4316 } 4317 4318 void 4319 if_setppromisc(if_t ifp, bool ppromisc) 4320 { 4321 int new_flags; 4322 4323 if (ppromisc) 4324 new_flags = ifp->if_flags | IFF_PPROMISC; 4325 else 4326 new_flags = ifp->if_flags & ~IFF_PPROMISC; 4327 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 4328 if (new_flags & IFF_PPROMISC) 4329 new_flags |= IFF_PROMISC; 4330 /* 4331 * Only unset IFF_PROMISC if there are no more consumers of 4332 * promiscuity, i.e. the ifp->if_pcount refcount is 0. 4333 */ 4334 else if (ifp->if_pcount == 0) 4335 new_flags &= ~IFF_PROMISC; 4336 if (log_promisc_mode_change) 4337 if_printf(ifp, "permanently promiscuous mode %s\n", 4338 ((new_flags & IFF_PPROMISC) ? 4339 "enabled" : "disabled")); 4340 } 4341 ifp->if_flags = new_flags; 4342 } 4343 4344 /* 4345 * Methods for drivers to access interface unicast and multicast 4346 * link level addresses. Driver shall not know 'struct ifaddr' neither 4347 * 'struct ifmultiaddr'. 4348 */ 4349 u_int 4350 if_lladdr_count(if_t ifp) 4351 { 4352 struct epoch_tracker et; 4353 struct ifaddr *ifa; 4354 u_int count; 4355 4356 count = 0; 4357 NET_EPOCH_ENTER(et); 4358 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4359 if (ifa->ifa_addr->sa_family == AF_LINK) 4360 count++; 4361 NET_EPOCH_EXIT(et); 4362 4363 return (count); 4364 } 4365 4366 int 4367 if_foreach(if_foreach_cb_t cb, void *cb_arg) 4368 { 4369 if_t ifp; 4370 int error; 4371 4372 NET_EPOCH_ASSERT(); 4373 MPASS(cb); 4374 4375 error = 0; 4376 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4377 error = cb(ifp, cb_arg); 4378 if (error != 0) 4379 break; 4380 } 4381 4382 return (error); 4383 } 4384 4385 /* 4386 * Iterates over the list of interfaces, permitting callback function @cb to sleep. 4387 * Stops iteration if @cb returns non-zero error code. 4388 * Returns the last error code from @cb. 4389 * @match_cb: optional match callback limiting the iteration to only matched interfaces 4390 * @match_arg: argument to pass to @match_cb 4391 * @cb: iteration callback 4392 * @cb_arg: argument to pass to @cb 4393 */ 4394 int 4395 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb, 4396 void *cb_arg) 4397 { 4398 int match_count = 0, array_size = 16; /* 128 bytes for malloc */ 4399 struct ifnet **match_array = NULL; 4400 int error = 0; 4401 4402 MPASS(cb); 4403 4404 while (true) { 4405 struct ifnet **new_array; 4406 int new_size = array_size; 4407 struct epoch_tracker et; 4408 struct ifnet *ifp; 4409 4410 while (new_size < match_count) 4411 new_size *= 2; 4412 new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK); 4413 if (match_array != NULL) 4414 memcpy(new_array, match_array, array_size * sizeof(void *)); 4415 free(match_array, M_TEMP); 4416 match_array = new_array; 4417 array_size = new_size; 4418 4419 match_count = 0; 4420 NET_EPOCH_ENTER(et); 4421 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4422 if (match_cb != NULL && !match_cb(ifp, match_arg)) 4423 continue; 4424 if (match_count < array_size) { 4425 if (if_try_ref(ifp)) 4426 match_array[match_count++] = ifp; 4427 } else 4428 match_count++; 4429 } 4430 NET_EPOCH_EXIT(et); 4431 4432 if (match_count > array_size) { 4433 for (int i = 0; i < array_size; i++) 4434 if_rele(match_array[i]); 4435 continue; 4436 } else { 4437 for (int i = 0; i < match_count; i++) { 4438 if (error == 0) 4439 error = cb(match_array[i], cb_arg); 4440 if_rele(match_array[i]); 4441 } 4442 free(match_array, M_TEMP); 4443 break; 4444 } 4445 } 4446 4447 return (error); 4448 } 4449 4450 4451 /* 4452 * Uses just 1 pointer of the 4 available in the public struct. 4453 */ 4454 if_t 4455 if_iter_start(struct if_iter *iter) 4456 { 4457 if_t ifp; 4458 4459 NET_EPOCH_ASSERT(); 4460 4461 bzero(iter, sizeof(*iter)); 4462 ifp = CK_STAILQ_FIRST(&V_ifnet); 4463 if (ifp != NULL) 4464 iter->context[0] = CK_STAILQ_NEXT(ifp, if_link); 4465 else 4466 iter->context[0] = NULL; 4467 return (ifp); 4468 } 4469 4470 if_t 4471 if_iter_next(struct if_iter *iter) 4472 { 4473 if_t cur_ifp = iter->context[0]; 4474 4475 if (cur_ifp != NULL) 4476 iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link); 4477 return (cur_ifp); 4478 } 4479 4480 void 4481 if_iter_finish(struct if_iter *iter) 4482 { 4483 /* Nothing to do here for now. */ 4484 } 4485 4486 u_int 4487 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4488 { 4489 struct epoch_tracker et; 4490 struct ifaddr *ifa; 4491 u_int count; 4492 4493 MPASS(cb); 4494 4495 count = 0; 4496 NET_EPOCH_ENTER(et); 4497 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4498 if (ifa->ifa_addr->sa_family != AF_LINK) 4499 continue; 4500 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4501 count); 4502 } 4503 NET_EPOCH_EXIT(et); 4504 4505 return (count); 4506 } 4507 4508 u_int 4509 if_llmaddr_count(if_t ifp) 4510 { 4511 struct epoch_tracker et; 4512 struct ifmultiaddr *ifma; 4513 int count; 4514 4515 count = 0; 4516 NET_EPOCH_ENTER(et); 4517 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4518 if (ifma->ifma_addr->sa_family == AF_LINK) 4519 count++; 4520 NET_EPOCH_EXIT(et); 4521 4522 return (count); 4523 } 4524 4525 bool 4526 if_maddr_empty(if_t ifp) 4527 { 4528 4529 return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs)); 4530 } 4531 4532 u_int 4533 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4534 { 4535 struct epoch_tracker et; 4536 struct ifmultiaddr *ifma; 4537 u_int count; 4538 4539 MPASS(cb); 4540 4541 count = 0; 4542 NET_EPOCH_ENTER(et); 4543 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4544 if (ifma->ifma_addr->sa_family != AF_LINK) 4545 continue; 4546 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4547 count); 4548 } 4549 NET_EPOCH_EXIT(et); 4550 4551 return (count); 4552 } 4553 4554 u_int 4555 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4556 { 4557 struct epoch_tracker et; 4558 struct ifaddr *ifa; 4559 u_int count; 4560 4561 MPASS(cb); 4562 4563 count = 0; 4564 NET_EPOCH_ENTER(et); 4565 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4566 if (ifa->ifa_addr->sa_family != type) 4567 continue; 4568 count += (*cb)(cb_arg, ifa, count); 4569 } 4570 NET_EPOCH_EXIT(et); 4571 4572 return (count); 4573 } 4574 4575 struct ifaddr * 4576 ifa_iter_start(if_t ifp, struct ifa_iter *iter) 4577 { 4578 struct ifaddr *ifa; 4579 4580 NET_EPOCH_ASSERT(); 4581 4582 bzero(iter, sizeof(*iter)); 4583 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 4584 if (ifa != NULL) 4585 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4586 else 4587 iter->context[0] = NULL; 4588 return (ifa); 4589 } 4590 4591 struct ifaddr * 4592 ifa_iter_next(struct ifa_iter *iter) 4593 { 4594 struct ifaddr *ifa = iter->context[0]; 4595 4596 if (ifa != NULL) 4597 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4598 return (ifa); 4599 } 4600 4601 void 4602 ifa_iter_finish(struct ifa_iter *iter) 4603 { 4604 /* Nothing to do here for now. */ 4605 } 4606 4607 int 4608 if_setsoftc(if_t ifp, void *softc) 4609 { 4610 ifp->if_softc = softc; 4611 return (0); 4612 } 4613 4614 void * 4615 if_getsoftc(const if_t ifp) 4616 { 4617 return (ifp->if_softc); 4618 } 4619 4620 void 4621 if_setrcvif(struct mbuf *m, if_t ifp) 4622 { 4623 4624 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4625 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4626 } 4627 4628 void 4629 if_setvtag(struct mbuf *m, uint16_t tag) 4630 { 4631 m->m_pkthdr.ether_vtag = tag; 4632 } 4633 4634 uint16_t 4635 if_getvtag(struct mbuf *m) 4636 { 4637 return (m->m_pkthdr.ether_vtag); 4638 } 4639 4640 int 4641 if_sendq_empty(if_t ifp) 4642 { 4643 return (IFQ_DRV_IS_EMPTY(&ifp->if_snd)); 4644 } 4645 4646 struct ifaddr * 4647 if_getifaddr(const if_t ifp) 4648 { 4649 return (ifp->if_addr); 4650 } 4651 4652 int 4653 if_setsendqready(if_t ifp) 4654 { 4655 IFQ_SET_READY(&ifp->if_snd); 4656 return (0); 4657 } 4658 4659 int 4660 if_setsendqlen(if_t ifp, int tx_desc_count) 4661 { 4662 IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count); 4663 ifp->if_snd.ifq_drv_maxlen = tx_desc_count; 4664 return (0); 4665 } 4666 4667 void 4668 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na) 4669 { 4670 ifp->if_netmap = na; 4671 } 4672 4673 struct netmap_adapter * 4674 if_getnetmapadapter(if_t ifp) 4675 { 4676 return (ifp->if_netmap); 4677 } 4678 4679 int 4680 if_vlantrunkinuse(if_t ifp) 4681 { 4682 return (ifp->if_vlantrunk != NULL); 4683 } 4684 4685 void 4686 if_init(if_t ifp, void *ctx) 4687 { 4688 (*ifp->if_init)(ctx); 4689 } 4690 4691 void 4692 if_input(if_t ifp, struct mbuf* sendmp) 4693 { 4694 (*ifp->if_input)(ifp, sendmp); 4695 } 4696 4697 int 4698 if_transmit(if_t ifp, struct mbuf *m) 4699 { 4700 return ((*ifp->if_transmit)(ifp, m)); 4701 } 4702 4703 int 4704 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst) 4705 { 4706 if (ifp->if_resolvemulti == NULL) 4707 return (EOPNOTSUPP); 4708 4709 return (ifp->if_resolvemulti(ifp, srcs, dst)); 4710 } 4711 4712 int 4713 if_ioctl(if_t ifp, u_long cmd, void *data) 4714 { 4715 if (ifp->if_ioctl == NULL) 4716 return (EOPNOTSUPP); 4717 4718 return (ifp->if_ioctl(ifp, cmd, data)); 4719 } 4720 4721 struct mbuf * 4722 if_dequeue(if_t ifp) 4723 { 4724 struct mbuf *m; 4725 4726 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 4727 return (m); 4728 } 4729 4730 int 4731 if_sendq_prepend(if_t ifp, struct mbuf *m) 4732 { 4733 IFQ_DRV_PREPEND(&ifp->if_snd, m); 4734 return (0); 4735 } 4736 4737 int 4738 if_setifheaderlen(if_t ifp, int len) 4739 { 4740 ifp->if_hdrlen = len; 4741 return (0); 4742 } 4743 4744 char * 4745 if_getlladdr(const if_t ifp) 4746 { 4747 return (IF_LLADDR(ifp)); 4748 } 4749 4750 void * 4751 if_gethandle(u_char type) 4752 { 4753 return (if_alloc(type)); 4754 } 4755 4756 void 4757 if_vlancap(if_t ifp) 4758 { 4759 VLAN_CAPABILITIES(ifp); 4760 } 4761 4762 int 4763 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4764 { 4765 ifp->if_hw_tsomax = if_hw_tsomax; 4766 return (0); 4767 } 4768 4769 int 4770 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4771 { 4772 ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4773 return (0); 4774 } 4775 4776 int 4777 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4778 { 4779 ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4780 return (0); 4781 } 4782 4783 u_int 4784 if_gethwtsomax(const if_t ifp) 4785 { 4786 return (ifp->if_hw_tsomax); 4787 } 4788 4789 u_int 4790 if_gethwtsomaxsegcount(const if_t ifp) 4791 { 4792 return (ifp->if_hw_tsomaxsegcount); 4793 } 4794 4795 u_int 4796 if_gethwtsomaxsegsize(const if_t ifp) 4797 { 4798 return (ifp->if_hw_tsomaxsegsize); 4799 } 4800 4801 void 4802 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4803 { 4804 ifp->if_init = init_fn; 4805 } 4806 4807 void 4808 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4809 { 4810 ifp->if_input = input_fn; 4811 } 4812 4813 if_input_fn_t 4814 if_getinputfn(if_t ifp) 4815 { 4816 return (ifp->if_input); 4817 } 4818 4819 void 4820 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4821 { 4822 ifp->if_ioctl = ioctl_fn; 4823 } 4824 4825 void 4826 if_setoutputfn(if_t ifp, if_output_fn_t output_fn) 4827 { 4828 ifp->if_output = output_fn; 4829 } 4830 4831 void 4832 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4833 { 4834 ifp->if_start = start_fn; 4835 } 4836 4837 if_start_fn_t 4838 if_getstartfn(if_t ifp) 4839 { 4840 return (ifp->if_start); 4841 } 4842 4843 void 4844 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4845 { 4846 ifp->if_transmit = start_fn; 4847 } 4848 4849 if_transmit_fn_t 4850 if_gettransmitfn(if_t ifp) 4851 { 4852 return (ifp->if_transmit); 4853 } 4854 4855 void 4856 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4857 { 4858 ifp->if_qflush = flush_fn; 4859 } 4860 4861 void 4862 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 4863 { 4864 ifp->if_snd_tag_alloc = alloc_fn; 4865 } 4866 4867 int 4868 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params, 4869 struct m_snd_tag **mstp) 4870 { 4871 if (ifp->if_snd_tag_alloc == NULL) 4872 return (EOPNOTSUPP); 4873 return (ifp->if_snd_tag_alloc(ifp, params, mstp)); 4874 } 4875 4876 void 4877 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4878 { 4879 ifp->if_get_counter = fn; 4880 } 4881 4882 void 4883 if_setreassignfn(if_t ifp, if_reassign_fn_t fn) 4884 { 4885 ifp->if_reassign = fn; 4886 } 4887 4888 void 4889 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn) 4890 { 4891 ifp->if_ratelimit_query = fn; 4892 } 4893 4894 void 4895 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m) 4896 { 4897 ifp->if_debugnet_methods = m; 4898 } 4899 4900 struct label * 4901 if_getmaclabel(if_t ifp) 4902 { 4903 return (ifp->if_label); 4904 } 4905 4906 void 4907 if_setmaclabel(if_t ifp, struct label *label) 4908 { 4909 ifp->if_label = label; 4910 } 4911 4912 int 4913 if_gettype(if_t ifp) 4914 { 4915 return (ifp->if_type); 4916 } 4917 4918 void * 4919 if_getllsoftc(if_t ifp) 4920 { 4921 return (ifp->if_llsoftc); 4922 } 4923 4924 void 4925 if_setllsoftc(if_t ifp, void *llsoftc) 4926 { 4927 ifp->if_llsoftc = llsoftc; 4928 }; 4929 4930 int 4931 if_getlinkstate(if_t ifp) 4932 { 4933 return (ifp->if_link_state); 4934 } 4935 4936 const uint8_t * 4937 if_getbroadcastaddr(if_t ifp) 4938 { 4939 return (ifp->if_broadcastaddr); 4940 } 4941 4942 void 4943 if_setbroadcastaddr(if_t ifp, const uint8_t *addr) 4944 { 4945 ifp->if_broadcastaddr = addr; 4946 } 4947 4948 int 4949 if_getnumadomain(if_t ifp) 4950 { 4951 return (ifp->if_numa_domain); 4952 } 4953 4954 uint64_t 4955 if_getcounter(if_t ifp, ift_counter counter) 4956 { 4957 return (ifp->if_get_counter(ifp, counter)); 4958 } 4959 4960 bool 4961 if_altq_is_enabled(if_t ifp) 4962 { 4963 return (ALTQ_IS_ENABLED(&ifp->if_snd)); 4964 } 4965 4966 struct vnet * 4967 if_getvnet(if_t ifp) 4968 { 4969 return (ifp->if_vnet); 4970 } 4971 4972 struct in_ifinfo * 4973 if_getinet(if_t ifp) 4974 { 4975 return (ifp->if_inet); 4976 } 4977 4978 struct in6_ifextra * 4979 if_getinet6(if_t ifp) 4980 { 4981 return (ifp->if_inet6); 4982 } 4983 4984 u_int 4985 if_getfib(if_t ifp) 4986 { 4987 return (ifp->if_fib); 4988 } 4989 4990 uint8_t 4991 if_getaddrlen(if_t ifp) 4992 { 4993 return (ifp->if_addrlen); 4994 } 4995 4996 struct bpf_if * 4997 if_getbpf(if_t ifp) 4998 { 4999 return (ifp->if_bpf); 5000 } 5001 5002 struct ifvlantrunk * 5003 if_getvlantrunk(if_t ifp) 5004 { 5005 return (ifp->if_vlantrunk); 5006 } 5007 5008 uint8_t 5009 if_getpcp(if_t ifp) 5010 { 5011 return (ifp->if_pcp); 5012 } 5013 5014 void * 5015 if_getl2com(if_t ifp) 5016 { 5017 return (ifp->if_l2com); 5018 } 5019 5020 void 5021 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m) 5022 { 5023 ifp->if_ipsec_accel_m = m; 5024 } 5025 5026 #ifdef DDB 5027 static void 5028 if_show_ifnet(struct ifnet *ifp) 5029 { 5030 if (ifp == NULL) 5031 return; 5032 db_printf("%s:\n", ifp->if_xname); 5033 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 5034 IF_DB_PRINTF("%s", if_dname); 5035 IF_DB_PRINTF("%d", if_dunit); 5036 IF_DB_PRINTF("%s", if_description); 5037 IF_DB_PRINTF("%u", if_index); 5038 IF_DB_PRINTF("%d", if_idxgen); 5039 IF_DB_PRINTF("%u", if_refcount); 5040 IF_DB_PRINTF("%p", if_softc); 5041 IF_DB_PRINTF("%p", if_l2com); 5042 IF_DB_PRINTF("%p", if_llsoftc); 5043 IF_DB_PRINTF("%d", if_amcount); 5044 IF_DB_PRINTF("%p", if_addr); 5045 IF_DB_PRINTF("%p", if_broadcastaddr); 5046 IF_DB_PRINTF("%u", if_fib); 5047 IF_DB_PRINTF("%p", if_vnet); 5048 IF_DB_PRINTF("%p", if_home_vnet); 5049 IF_DB_PRINTF("%p", if_vlantrunk); 5050 IF_DB_PRINTF("%p", if_bpf); 5051 IF_DB_PRINTF("%u", if_pcount); 5052 IF_DB_PRINTF("%p", if_bridge); 5053 IF_DB_PRINTF("%p", if_lagg); 5054 IF_DB_PRINTF("%p", if_pf_kif); 5055 IF_DB_PRINTF("%p", if_carp); 5056 IF_DB_PRINTF("%p", if_label); 5057 IF_DB_PRINTF("%p", if_netmap); 5058 IF_DB_PRINTF("0x%08x", if_flags); 5059 IF_DB_PRINTF("0x%08x", if_drv_flags); 5060 IF_DB_PRINTF("0x%08x", if_capabilities); 5061 IF_DB_PRINTF("0x%08x", if_capenable); 5062 IF_DB_PRINTF("%p", if_snd.ifq_head); 5063 IF_DB_PRINTF("%p", if_snd.ifq_tail); 5064 IF_DB_PRINTF("%d", if_snd.ifq_len); 5065 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 5066 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 5067 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 5068 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 5069 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 5070 IF_DB_PRINTF("%d", if_snd.altq_type); 5071 IF_DB_PRINTF("%x", if_snd.altq_flags); 5072 #undef IF_DB_PRINTF 5073 } 5074 5075 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 5076 { 5077 if (!have_addr) { 5078 db_printf("usage: show ifnet <struct ifnet *>\n"); 5079 return; 5080 } 5081 5082 if_show_ifnet((struct ifnet *)addr); 5083 } 5084 5085 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 5086 { 5087 struct ifnet *ifp; 5088 u_short idx; 5089 5090 for (idx = 1; idx <= if_index; idx++) { 5091 ifp = ifindex_table[idx].ife_ifnet; 5092 if (ifp == NULL) 5093 continue; 5094 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 5095 if (db_pager_quit) 5096 break; 5097 } 5098 } 5099 #endif /* DDB */ 5100