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 IFNET_WLOCK(); 432 433 MPASS(refcount_load(&ifp->if_refcount) > 0); 434 MPASS(ifp->if_vnet == curvnet); 435 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 436 437 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 438 #ifdef VIMAGE 439 curvnet->vnet_ifcnt++; 440 #endif 441 IFNET_WUNLOCK(); 442 } 443 444 static bool 445 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 446 { 447 struct ifnet *iter; 448 int found = 0; 449 450 IFNET_WLOCK(); 451 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 452 if (iter == ifp) { 453 MPASS(refcount_load(&ifp->if_refcount) > 0); 454 MPASS(ifp->if_vnet == curvnet); 455 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 456 457 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 458 #ifdef VIMAGE 459 curvnet->vnet_ifcnt--; 460 #endif 461 if (!vmove) 462 ifp->if_flags |= IFF_DYING; 463 found = 1; 464 break; 465 } 466 IFNET_WUNLOCK(); 467 468 return (found); 469 } 470 471 #ifdef VIMAGE 472 static void 473 vnet_if_return(const void *unused __unused) 474 { 475 struct ifnet *ifp, *nifp; 476 struct ifnet **pending; 477 int found __diagused; 478 int i; 479 480 i = 0; 481 482 /* The lock has already been aquired in vnet_destroy() */ 483 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 484 /* 485 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 486 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 487 * if_detach_internal(), which waits for NET_EPOCH callbacks to 488 * complete. We can't do that from within NET_EPOCH. 489 * 490 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 491 * read/write lock. We cannot hold the lock as we call if_vmove() 492 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 493 * ctx lock. 494 */ 495 IFNET_WLOCK(); 496 497 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 498 M_IFNET, M_WAITOK | M_ZERO); 499 500 /* Return all inherited interfaces to their parent vnets. */ 501 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 502 if (ifp->if_home_vnet != ifp->if_vnet) { 503 found = if_unlink_ifnet(ifp, true); 504 MPASS(found); 505 506 pending[i++] = ifp; 507 } 508 } 509 IFNET_WUNLOCK(); 510 511 for (int j = 0; j < i; j++) { 512 if_vmove(pending[j], pending[j]->if_home_vnet); 513 } 514 515 free(pending, M_IFNET); 516 } 517 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 518 vnet_if_return, NULL); 519 #endif 520 521 /* 522 * Allocate a struct ifnet and an index for an interface. A layer 2 523 * common structure will also be allocated if an allocation routine is 524 * registered for the passed type. 525 */ 526 static struct ifnet * 527 if_alloc_domain(u_char type, int numa_domain) 528 { 529 struct ifnet *ifp; 530 u_short idx; 531 532 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 533 if (numa_domain == IF_NODOM) 534 ifp = malloc(sizeof(struct ifnet), M_IFNET, 535 M_WAITOK | M_ZERO); 536 else 537 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 538 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 539 ifp->if_type = type; 540 ifp->if_alloctype = type; 541 ifp->if_numa_domain = numa_domain; 542 #ifdef VIMAGE 543 ifp->if_vnet = curvnet; 544 #endif 545 if (if_com_alloc[type] != NULL) { 546 ifp->if_l2com = if_com_alloc[type](type, ifp); 547 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 548 type)); 549 } 550 551 IF_ADDR_LOCK_INIT(ifp); 552 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 553 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 554 CK_STAILQ_INIT(&ifp->if_addrhead); 555 CK_STAILQ_INIT(&ifp->if_multiaddrs); 556 CK_STAILQ_INIT(&ifp->if_groups); 557 #ifdef MAC 558 mac_ifnet_init(ifp); 559 #endif 560 ifq_init(&ifp->if_snd, ifp); 561 562 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 563 for (int i = 0; i < IFCOUNTERS; i++) 564 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 565 ifp->if_get_counter = if_get_counter_default; 566 ifp->if_pcp = IFNET_PCP_NONE; 567 568 /* Allocate an ifindex array entry. */ 569 IFNET_WLOCK(); 570 /* 571 * Try to find an empty slot below if_index. If we fail, take the 572 * next slot. 573 */ 574 for (idx = 1; idx <= if_index; idx++) { 575 if (ifindex_table[idx].ife_ifnet == NULL) 576 break; 577 } 578 579 /* Catch if_index overflow. */ 580 if (idx >= if_indexlim) { 581 struct ifindex_entry *new, *old; 582 int newlim; 583 584 newlim = if_indexlim * 2; 585 new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO); 586 memcpy(new, ifindex_table, if_indexlim * sizeof(*new)); 587 old = ifindex_table; 588 ck_pr_store_ptr(&ifindex_table, new); 589 if_indexlim = newlim; 590 NET_EPOCH_WAIT(); 591 free(old, M_IFNET); 592 } 593 if (idx > if_index) 594 if_index = idx; 595 596 ifp->if_index = idx; 597 ifp->if_idxgen = ifindex_table[idx].ife_gencnt; 598 ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp); 599 IFNET_WUNLOCK(); 600 601 return (ifp); 602 } 603 604 struct ifnet * 605 if_alloc_dev(u_char type, device_t dev) 606 { 607 int numa_domain; 608 609 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 610 return (if_alloc_domain(type, IF_NODOM)); 611 return (if_alloc_domain(type, numa_domain)); 612 } 613 614 struct ifnet * 615 if_alloc(u_char type) 616 { 617 618 return (if_alloc_domain(type, IF_NODOM)); 619 } 620 /* 621 * Do the actual work of freeing a struct ifnet, and layer 2 common 622 * structure. This call is made when the network epoch guarantees 623 * us that nobody holds a pointer to the interface. 624 */ 625 static void 626 if_free_deferred(epoch_context_t ctx) 627 { 628 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 629 630 KASSERT((ifp->if_flags & IFF_DYING), 631 ("%s: interface not dying", __func__)); 632 633 if (if_com_free[ifp->if_alloctype] != NULL) 634 if_com_free[ifp->if_alloctype](ifp->if_l2com, 635 ifp->if_alloctype); 636 637 #ifdef MAC 638 mac_ifnet_destroy(ifp); 639 #endif /* MAC */ 640 IF_ADDR_LOCK_DESTROY(ifp); 641 ifq_delete(&ifp->if_snd); 642 643 for (int i = 0; i < IFCOUNTERS; i++) 644 counter_u64_free(ifp->if_counters[i]); 645 646 if_freedescr(ifp->if_description); 647 free(ifp->if_hw_addr, M_IFADDR); 648 free(ifp, M_IFNET); 649 } 650 651 /* 652 * Deregister an interface and free the associated storage. 653 */ 654 void 655 if_free(struct ifnet *ifp) 656 { 657 658 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 659 660 /* 661 * XXXGL: An interface index is really an alias to ifp pointer. 662 * Why would we clear the alias now, and not in the deferred 663 * context? Indeed there is nothing wrong with some network 664 * thread obtaining ifp via ifnet_byindex() inside the network 665 * epoch and then dereferencing ifp while we perform if_free(), 666 * and after if_free() finished, too. 667 * 668 * This early index freeing was important back when ifindex was 669 * virtualized and interface would outlive the vnet. 670 */ 671 IFNET_WLOCK(); 672 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 673 ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL); 674 ifindex_table[ifp->if_index].ife_gencnt++; 675 while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL) 676 if_index--; 677 IFNET_WUNLOCK(); 678 679 if (refcount_release(&ifp->if_refcount)) 680 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 681 } 682 683 /* 684 * Interfaces to keep an ifnet type-stable despite the possibility of the 685 * driver calling if_free(). If there are additional references, we defer 686 * freeing the underlying data structure. 687 */ 688 void 689 if_ref(struct ifnet *ifp) 690 { 691 u_int old __diagused; 692 693 /* We don't assert the ifnet list lock here, but arguably should. */ 694 old = refcount_acquire(&ifp->if_refcount); 695 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 696 } 697 698 bool 699 if_try_ref(struct ifnet *ifp) 700 { 701 NET_EPOCH_ASSERT(); 702 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 703 } 704 705 void 706 if_rele(struct ifnet *ifp) 707 { 708 709 if (!refcount_release(&ifp->if_refcount)) 710 return; 711 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 712 } 713 714 void 715 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 716 { 717 718 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 719 720 if (ifq->ifq_maxlen == 0) 721 ifq->ifq_maxlen = ifqmaxlen; 722 723 ifq->altq_type = 0; 724 ifq->altq_disc = NULL; 725 ifq->altq_flags &= ALTQF_CANTCHANGE; 726 ifq->altq_tbr = NULL; 727 ifq->altq_ifp = ifp; 728 } 729 730 void 731 ifq_delete(struct ifaltq *ifq) 732 { 733 mtx_destroy(&ifq->ifq_mtx); 734 } 735 736 /* 737 * Perform generic interface initialization tasks and attach the interface 738 * to the list of "active" interfaces. If vmove flag is set on entry 739 * to if_attach_internal(), perform only a limited subset of initialization 740 * tasks, given that we are moving from one vnet to another an ifnet which 741 * has already been fully initialized. 742 * 743 * Note that if_detach_internal() removes group membership unconditionally 744 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 745 * Thus, when if_vmove() is applied to a cloned interface, group membership 746 * is lost while a cloned one always joins a group whose name is 747 * ifc->ifc_name. To recover this after if_detach_internal() and 748 * if_attach_internal(), the cloner should be specified to 749 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 750 * attempts to join a group whose name is ifc->ifc_name. 751 * 752 * XXX: 753 * - The decision to return void and thus require this function to 754 * succeed is questionable. 755 * - We should probably do more sanity checking. For instance we don't 756 * do anything to insure if_xname is unique or non-empty. 757 */ 758 void 759 if_attach(struct ifnet *ifp) 760 { 761 762 if_attach_internal(ifp, false); 763 } 764 765 /* 766 * Compute the least common TSO limit. 767 */ 768 void 769 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 770 { 771 /* 772 * 1) If there is no limit currently, take the limit from 773 * the network adapter. 774 * 775 * 2) If the network adapter has a limit below the current 776 * limit, apply it. 777 */ 778 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 779 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 780 pmax->tsomaxbytes = ifp->if_hw_tsomax; 781 } 782 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 783 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 784 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 785 } 786 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 787 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 788 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 789 } 790 } 791 792 /* 793 * Update TSO limit of a network adapter. 794 * 795 * Returns zero if no change. Else non-zero. 796 */ 797 int 798 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 799 { 800 int retval = 0; 801 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 802 ifp->if_hw_tsomax = pmax->tsomaxbytes; 803 retval++; 804 } 805 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 806 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 807 retval++; 808 } 809 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 810 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 811 retval++; 812 } 813 return (retval); 814 } 815 816 static void 817 if_attach_internal(struct ifnet *ifp, bool vmove) 818 { 819 unsigned socksize, ifasize; 820 int namelen, masklen; 821 struct sockaddr_dl *sdl; 822 struct ifaddr *ifa; 823 824 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 825 826 #ifdef VIMAGE 827 CURVNET_ASSERT_SET(); 828 ifp->if_vnet = curvnet; 829 if (ifp->if_home_vnet == NULL) 830 ifp->if_home_vnet = curvnet; 831 #endif 832 833 if_addgroup(ifp, IFG_ALL); 834 835 #ifdef VIMAGE 836 /* Restore group membership for cloned interface. */ 837 if (vmove) 838 if_clone_restoregroup(ifp); 839 #endif 840 841 getmicrotime(&ifp->if_lastchange); 842 ifp->if_epoch = time_uptime; 843 844 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 845 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 846 ("transmit and qflush must both either be set or both be NULL")); 847 if (ifp->if_transmit == NULL) { 848 ifp->if_transmit = if_transmit_default; 849 ifp->if_qflush = if_qflush; 850 } 851 if (ifp->if_input == NULL) 852 ifp->if_input = if_input_default; 853 854 if (ifp->if_requestencap == NULL) 855 ifp->if_requestencap = if_requestencap_default; 856 857 if (!vmove) { 858 #ifdef MAC 859 mac_ifnet_create(ifp); 860 #endif 861 862 /* 863 * Create a Link Level name for this device. 864 */ 865 namelen = strlen(ifp->if_xname); 866 /* 867 * Always save enough space for any possible name so we 868 * can do a rename in place later. 869 */ 870 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 871 socksize = masklen + ifp->if_addrlen; 872 if (socksize < sizeof(*sdl)) 873 socksize = sizeof(*sdl); 874 socksize = roundup2(socksize, sizeof(long)); 875 ifasize = sizeof(*ifa) + 2 * socksize; 876 ifa = ifa_alloc(ifasize, M_WAITOK); 877 sdl = (struct sockaddr_dl *)(ifa + 1); 878 sdl->sdl_len = socksize; 879 sdl->sdl_family = AF_LINK; 880 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 881 sdl->sdl_nlen = namelen; 882 sdl->sdl_index = ifp->if_index; 883 sdl->sdl_type = ifp->if_type; 884 ifp->if_addr = ifa; 885 ifa->ifa_ifp = ifp; 886 ifa->ifa_addr = (struct sockaddr *)sdl; 887 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 888 ifa->ifa_netmask = (struct sockaddr *)sdl; 889 sdl->sdl_len = masklen; 890 while (namelen != 0) 891 sdl->sdl_data[--namelen] = 0xff; 892 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 893 /* Reliably crash if used uninitialized. */ 894 ifp->if_broadcastaddr = NULL; 895 896 if (ifp->if_type == IFT_ETHER) { 897 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 898 M_WAITOK | M_ZERO); 899 } 900 901 #if defined(INET) || defined(INET6) 902 /* Use defaults for TSO, if nothing is set */ 903 if (ifp->if_hw_tsomax == 0 && 904 ifp->if_hw_tsomaxsegcount == 0 && 905 ifp->if_hw_tsomaxsegsize == 0) { 906 /* 907 * The TSO defaults needs to be such that an 908 * NFS mbuf list of 35 mbufs totalling just 909 * below 64K works and that a chain of mbufs 910 * can be defragged into at most 32 segments: 911 */ 912 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 913 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 914 ifp->if_hw_tsomaxsegcount = 35; 915 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 916 917 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 918 if (ifp->if_capabilities & IFCAP_TSO) { 919 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 920 ifp->if_hw_tsomax, 921 ifp->if_hw_tsomaxsegcount, 922 ifp->if_hw_tsomaxsegsize); 923 } 924 } 925 #endif 926 } 927 928 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 929 if_link_ifnet(ifp); 930 EVENTHANDLER_INVOKE(ifnet_attached_event, ifp); 931 if (IS_DEFAULT_VNET(curvnet)) 932 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 933 } 934 935 static void 936 if_epochalloc(void *dummy __unused) 937 { 938 939 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 940 } 941 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 942 943 /* 944 * Remove any unicast or broadcast network addresses from an interface. 945 */ 946 void 947 if_purgeaddrs(struct ifnet *ifp) 948 { 949 struct ifaddr *ifa; 950 951 #ifdef INET6 952 /* 953 * Need to leave multicast addresses of proxy NDP llentries 954 * before in6_purgeifaddr() because the llentries are keys 955 * for in6_multi objects of proxy NDP entries. 956 * in6_purgeifaddr()s clean up llentries including proxy NDPs 957 * then we would lose the keys if they are called earlier. 958 */ 959 in6_purge_proxy_ndp(ifp); 960 #endif 961 while (1) { 962 struct epoch_tracker et; 963 964 NET_EPOCH_ENTER(et); 965 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 966 if (ifa->ifa_addr->sa_family != AF_LINK) 967 break; 968 } 969 NET_EPOCH_EXIT(et); 970 971 if (ifa == NULL) 972 break; 973 #ifdef INET 974 /* XXX: Ugly!! ad hoc just for INET */ 975 if (ifa->ifa_addr->sa_family == AF_INET) { 976 struct ifreq ifr; 977 978 bzero(&ifr, sizeof(ifr)); 979 ifr.ifr_addr = *ifa->ifa_addr; 980 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 981 NULL) == 0) 982 continue; 983 } 984 #endif /* INET */ 985 #ifdef INET6 986 if (ifa->ifa_addr->sa_family == AF_INET6) { 987 in6_purgeifaddr((struct in6_ifaddr *)ifa); 988 /* ifp_addrhead is already updated */ 989 continue; 990 } 991 #endif /* INET6 */ 992 IF_ADDR_WLOCK(ifp); 993 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 994 IF_ADDR_WUNLOCK(ifp); 995 ifa_free(ifa); 996 } 997 } 998 999 /* 1000 * Remove any multicast network addresses from an interface when an ifnet 1001 * is going away. 1002 */ 1003 static void 1004 if_purgemaddrs(struct ifnet *ifp) 1005 { 1006 struct ifmultiaddr *ifma; 1007 1008 IF_ADDR_WLOCK(ifp); 1009 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1010 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1011 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1012 if_delmulti_locked(ifp, ifma, 1); 1013 } 1014 IF_ADDR_WUNLOCK(ifp); 1015 } 1016 1017 /* 1018 * Detach an interface, removing it from the list of "active" interfaces. 1019 * If vmove flag is set on entry to if_detach_internal(), perform only a 1020 * limited subset of cleanup tasks, given that we are moving an ifnet from 1021 * one vnet to another, where it must be fully operational. 1022 * 1023 * XXXRW: There are some significant questions about event ordering, and 1024 * how to prevent things from starting to use the interface during detach. 1025 */ 1026 void 1027 if_detach(struct ifnet *ifp) 1028 { 1029 bool found; 1030 1031 CURVNET_SET_QUIET(ifp->if_vnet); 1032 found = if_unlink_ifnet(ifp, false); 1033 if (found) { 1034 sx_xlock(&ifnet_detach_sxlock); 1035 if_detach_internal(ifp, false); 1036 sx_xunlock(&ifnet_detach_sxlock); 1037 } 1038 CURVNET_RESTORE(); 1039 } 1040 1041 /* 1042 * The vmove flag, if set, indicates that we are called from a callpath 1043 * that is moving an interface to a different vnet instance. 1044 * 1045 * The shutdown flag, if set, indicates that we are called in the 1046 * process of shutting down a vnet instance. Currently only the 1047 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1048 * on a vnet instance shutdown without this flag being set, e.g., when 1049 * the cloned interfaces are destoyed as first thing of teardown. 1050 */ 1051 static void 1052 if_detach_internal(struct ifnet *ifp, bool vmove) 1053 { 1054 struct ifaddr *ifa; 1055 #ifdef VIMAGE 1056 bool shutdown; 1057 1058 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1059 #endif 1060 1061 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 1062 1063 /* 1064 * At this point we know the interface still was on the ifnet list 1065 * and we removed it so we are in a stable state. 1066 */ 1067 NET_EPOCH_WAIT(); 1068 1069 /* 1070 * Ensure all pending EPOCH(9) callbacks have been executed. This 1071 * fixes issues about late destruction of multicast options 1072 * which lead to leave group calls, which in turn access the 1073 * belonging ifnet structure: 1074 */ 1075 NET_EPOCH_DRAIN_CALLBACKS(); 1076 1077 /* 1078 * In any case (destroy or vmove) detach us from the groups 1079 * and remove/wait for pending events on the taskq. 1080 * XXX-BZ in theory an interface could still enqueue a taskq change? 1081 */ 1082 if_delgroups(ifp); 1083 1084 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1085 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1086 1087 if_down(ifp); 1088 1089 #ifdef VIMAGE 1090 /* 1091 * On VNET shutdown abort here as the stack teardown will do all 1092 * the work top-down for us. 1093 */ 1094 if (shutdown) { 1095 /* Give interface users the chance to clean up. */ 1096 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1097 1098 /* 1099 * In case of a vmove we are done here without error. 1100 * If we would signal an error it would lead to the same 1101 * abort as if we did not find the ifnet anymore. 1102 * if_detach() calls us in void context and does not care 1103 * about an early abort notification, so life is splendid :) 1104 */ 1105 return; 1106 } 1107 #endif 1108 1109 /* 1110 * At this point we are not tearing down a VNET and are either 1111 * going to destroy or vmove the interface and have to cleanup 1112 * accordingly. 1113 */ 1114 1115 /* 1116 * Remove routes and flush queues. 1117 */ 1118 #ifdef ALTQ 1119 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1120 altq_disable(&ifp->if_snd); 1121 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1122 altq_detach(&ifp->if_snd); 1123 #endif 1124 1125 rt_flushifroutes(ifp); 1126 1127 if_purgeaddrs(ifp); 1128 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1129 if_purgemaddrs(ifp); 1130 if (IS_DEFAULT_VNET(curvnet)) 1131 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1132 1133 if (!vmove) { 1134 /* 1135 * Prevent further calls into the device driver via ifnet. 1136 */ 1137 if_dead(ifp); 1138 1139 /* 1140 * Clean up all addresses. 1141 */ 1142 IF_ADDR_WLOCK(ifp); 1143 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1144 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1145 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1146 IF_ADDR_WUNLOCK(ifp); 1147 ifa_free(ifa); 1148 } else 1149 IF_ADDR_WUNLOCK(ifp); 1150 } 1151 } 1152 1153 #ifdef VIMAGE 1154 /* 1155 * if_vmove() performs a limited version of if_detach() in current 1156 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1157 */ 1158 static void 1159 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1160 { 1161 #ifdef DEV_BPF 1162 /* 1163 * Detach BPF file descriptors from its interface. 1164 */ 1165 bpf_ifdetach(ifp); 1166 #endif 1167 1168 /* 1169 * Detach from current vnet, but preserve LLADDR info, do not 1170 * mark as dead etc. so that the ifnet can be reattached later. 1171 */ 1172 if_detach_internal(ifp, true); 1173 1174 /* 1175 * Perform interface-specific reassignment tasks, if provided by 1176 * the driver. 1177 */ 1178 if (ifp->if_reassign != NULL) 1179 ifp->if_reassign(ifp, new_vnet, NULL); 1180 1181 /* 1182 * Switch to the context of the target vnet. 1183 */ 1184 CURVNET_SET_QUIET(new_vnet); 1185 if_attach_internal(ifp, true); 1186 bpf_vmove(ifp->if_bpf); 1187 CURVNET_RESTORE(); 1188 } 1189 1190 /* 1191 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1192 */ 1193 static int 1194 if_vmove_loan(struct thread *td, char *ifname, int jid) 1195 { 1196 struct prison *pr; 1197 struct ifnet *ifp, *difp; 1198 bool found; 1199 1200 MPASS(curthread == td); 1201 MPASS(curvnet == TD_TO_VNET(td)); 1202 1203 /* 1204 * We check the existence of the interface, and will later try to 1205 * unlink it from the "active" list, so it is sufficient to only 1206 * hold a weak reference to it. 1207 * Be aware that it is unsafe to access any member of it, until it 1208 * is proven to be safe to ( say it was on the "active" list ). 1209 */ 1210 ifp = ifunit(ifname); 1211 if (ifp == NULL) 1212 return (ENXIO); 1213 1214 /* Try to find the prison within our visibility. */ 1215 sx_slock(&allprison_lock); 1216 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1217 sx_sunlock(&allprison_lock); 1218 if (pr == NULL) 1219 return (ENXIO); 1220 /* Do not try to move the iface from and to the same vnet. */ 1221 if (pr->pr_vnet == TD_TO_VNET(td)) { 1222 mtx_unlock(&pr->pr_mtx); 1223 return (EEXIST); 1224 } 1225 prison_hold_locked(pr); 1226 mtx_unlock(&pr->pr_mtx); 1227 1228 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1229 /* XXX Lock interfaces to avoid races. */ 1230 CURVNET_SET_QUIET(pr->pr_vnet); 1231 difp = ifunit(ifname); 1232 CURVNET_RESTORE(); 1233 if (difp != NULL) { 1234 prison_free(pr); 1235 return (EEXIST); 1236 } 1237 sx_xlock(&ifnet_detach_sxlock); 1238 1239 found = if_unlink_ifnet(ifp, true); 1240 if (! found) { 1241 sx_xunlock(&ifnet_detach_sxlock); 1242 prison_free(pr); 1243 return (ENODEV); 1244 } 1245 1246 /* Move the interface into the child jail/vnet. */ 1247 if_vmove(ifp, pr->pr_vnet); 1248 1249 /* Report the new if_xname back to the userland. */ 1250 sprintf(ifname, "%s", ifp->if_xname); 1251 1252 sx_xunlock(&ifnet_detach_sxlock); 1253 1254 prison_free(pr); 1255 return (0); 1256 } 1257 1258 static int 1259 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1260 { 1261 struct prison *pr; 1262 struct vnet *vnet_dst; 1263 struct ifnet *ifp; 1264 int found; 1265 1266 /* Try to find the prison within our visibility. */ 1267 sx_slock(&allprison_lock); 1268 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1269 sx_sunlock(&allprison_lock); 1270 if (pr == NULL) 1271 return (ENXIO); 1272 prison_hold_locked(pr); 1273 mtx_unlock(&pr->pr_mtx); 1274 1275 /* Make sure the named iface exists in the source prison/vnet. */ 1276 CURVNET_SET(pr->pr_vnet); 1277 ifp = ifunit(ifname); 1278 if (ifp == NULL) { 1279 CURVNET_RESTORE(); 1280 prison_free(pr); 1281 return (ENXIO); 1282 } 1283 1284 /* Do not try to move the iface from and to the same vnet. */ 1285 vnet_dst = TD_TO_VNET(td); 1286 if (vnet_dst == pr->pr_vnet) { 1287 CURVNET_RESTORE(); 1288 prison_free(pr); 1289 return (EEXIST); 1290 } 1291 1292 /* Get interface back from child jail/vnet. */ 1293 found = if_unlink_ifnet(ifp, true); 1294 if (! found) { 1295 CURVNET_RESTORE(); 1296 prison_free(pr); 1297 return (ENODEV); 1298 } 1299 sx_xlock(&ifnet_detach_sxlock); 1300 if_vmove(ifp, vnet_dst); 1301 sx_xunlock(&ifnet_detach_sxlock); 1302 CURVNET_RESTORE(); 1303 1304 /* Report the new if_xname back to the userland. */ 1305 sprintf(ifname, "%s", ifp->if_xname); 1306 1307 prison_free(pr); 1308 return (0); 1309 } 1310 #endif /* VIMAGE */ 1311 1312 /* 1313 * Add a group to an interface 1314 */ 1315 int 1316 if_addgroup(struct ifnet *ifp, const char *groupname) 1317 { 1318 struct ifg_list *ifgl; 1319 struct ifg_group *ifg = NULL; 1320 struct ifg_member *ifgm; 1321 int new = 0; 1322 1323 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1324 groupname[strlen(groupname) - 1] <= '9') 1325 return (EINVAL); 1326 1327 IFNET_WLOCK(); 1328 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1329 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1330 IFNET_WUNLOCK(); 1331 return (EEXIST); 1332 } 1333 1334 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1335 IFNET_WUNLOCK(); 1336 return (ENOMEM); 1337 } 1338 1339 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1340 free(ifgl, M_TEMP); 1341 IFNET_WUNLOCK(); 1342 return (ENOMEM); 1343 } 1344 1345 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1346 if (!strcmp(ifg->ifg_group, groupname)) 1347 break; 1348 1349 if (ifg == NULL) { 1350 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1351 free(ifgl, M_TEMP); 1352 free(ifgm, M_TEMP); 1353 IFNET_WUNLOCK(); 1354 return (ENOMEM); 1355 } 1356 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1357 ifg->ifg_refcnt = 0; 1358 CK_STAILQ_INIT(&ifg->ifg_members); 1359 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1360 new = 1; 1361 } 1362 1363 ifg->ifg_refcnt++; 1364 ifgl->ifgl_group = ifg; 1365 ifgm->ifgm_ifp = ifp; 1366 1367 IF_ADDR_WLOCK(ifp); 1368 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1369 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1370 IF_ADDR_WUNLOCK(ifp); 1371 1372 IFNET_WUNLOCK(); 1373 1374 if (new) 1375 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1376 EVENTHANDLER_INVOKE(group_change_event, groupname); 1377 1378 return (0); 1379 } 1380 1381 /* 1382 * Helper function to remove a group out of an interface. Expects the global 1383 * ifnet lock to be write-locked, and drops it before returning. 1384 */ 1385 static void 1386 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1387 const char *groupname) 1388 { 1389 struct ifg_member *ifgm; 1390 bool freeifgl; 1391 1392 IFNET_WLOCK_ASSERT(); 1393 1394 IF_ADDR_WLOCK(ifp); 1395 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1396 IF_ADDR_WUNLOCK(ifp); 1397 1398 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1399 if (ifgm->ifgm_ifp == ifp) { 1400 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1401 ifg_member, ifgm_next); 1402 break; 1403 } 1404 } 1405 1406 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1407 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1408 ifg_next); 1409 freeifgl = true; 1410 } else { 1411 freeifgl = false; 1412 } 1413 IFNET_WUNLOCK(); 1414 1415 NET_EPOCH_WAIT(); 1416 EVENTHANDLER_INVOKE(group_change_event, groupname); 1417 if (freeifgl) { 1418 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1419 free(ifgl->ifgl_group, M_TEMP); 1420 } 1421 free(ifgm, M_TEMP); 1422 free(ifgl, M_TEMP); 1423 } 1424 1425 /* 1426 * Remove a group from an interface 1427 */ 1428 int 1429 if_delgroup(struct ifnet *ifp, const char *groupname) 1430 { 1431 struct ifg_list *ifgl; 1432 1433 IFNET_WLOCK(); 1434 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1435 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1436 break; 1437 if (ifgl == NULL) { 1438 IFNET_WUNLOCK(); 1439 return (ENOENT); 1440 } 1441 1442 _if_delgroup_locked(ifp, ifgl, groupname); 1443 1444 return (0); 1445 } 1446 1447 /* 1448 * Remove an interface from all groups 1449 */ 1450 static void 1451 if_delgroups(struct ifnet *ifp) 1452 { 1453 struct ifg_list *ifgl; 1454 char groupname[IFNAMSIZ]; 1455 1456 IFNET_WLOCK(); 1457 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1458 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1459 _if_delgroup_locked(ifp, ifgl, groupname); 1460 IFNET_WLOCK(); 1461 } 1462 IFNET_WUNLOCK(); 1463 } 1464 1465 /* 1466 * Stores all groups from an interface in memory pointed to by ifgr. 1467 */ 1468 static int 1469 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1470 { 1471 int len, error; 1472 struct ifg_list *ifgl; 1473 struct ifg_req ifgrq, *ifgp; 1474 1475 NET_EPOCH_ASSERT(); 1476 1477 if (ifgr->ifgr_len == 0) { 1478 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1479 ifgr->ifgr_len += sizeof(struct ifg_req); 1480 return (0); 1481 } 1482 1483 len = ifgr->ifgr_len; 1484 ifgp = ifgr->ifgr_groups; 1485 /* XXX: wire */ 1486 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1487 if (len < sizeof(ifgrq)) 1488 return (EINVAL); 1489 bzero(&ifgrq, sizeof ifgrq); 1490 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1491 sizeof(ifgrq.ifgrq_group)); 1492 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1493 return (error); 1494 len -= sizeof(ifgrq); 1495 ifgp++; 1496 } 1497 1498 return (0); 1499 } 1500 1501 /* 1502 * Stores all members of a group in memory pointed to by igfr 1503 */ 1504 static int 1505 if_getgroupmembers(struct ifgroupreq *ifgr) 1506 { 1507 struct ifg_group *ifg; 1508 struct ifg_member *ifgm; 1509 struct ifg_req ifgrq, *ifgp; 1510 int len, error; 1511 1512 IFNET_RLOCK(); 1513 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1514 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1515 break; 1516 if (ifg == NULL) { 1517 IFNET_RUNLOCK(); 1518 return (ENOENT); 1519 } 1520 1521 if (ifgr->ifgr_len == 0) { 1522 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1523 ifgr->ifgr_len += sizeof(ifgrq); 1524 IFNET_RUNLOCK(); 1525 return (0); 1526 } 1527 1528 len = ifgr->ifgr_len; 1529 ifgp = ifgr->ifgr_groups; 1530 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1531 if (len < sizeof(ifgrq)) { 1532 IFNET_RUNLOCK(); 1533 return (EINVAL); 1534 } 1535 bzero(&ifgrq, sizeof ifgrq); 1536 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1537 sizeof(ifgrq.ifgrq_member)); 1538 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1539 IFNET_RUNLOCK(); 1540 return (error); 1541 } 1542 len -= sizeof(ifgrq); 1543 ifgp++; 1544 } 1545 IFNET_RUNLOCK(); 1546 1547 return (0); 1548 } 1549 1550 /* 1551 * Return counter values from counter(9)s stored in ifnet. 1552 */ 1553 uint64_t 1554 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1555 { 1556 1557 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1558 1559 return (counter_u64_fetch(ifp->if_counters[cnt])); 1560 } 1561 1562 /* 1563 * Increase an ifnet counter. Usually used for counters shared 1564 * between the stack and a driver, but function supports them all. 1565 */ 1566 void 1567 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1568 { 1569 1570 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1571 1572 counter_u64_add(ifp->if_counters[cnt], inc); 1573 } 1574 1575 /* 1576 * Copy data from ifnet to userland API structure if_data. 1577 */ 1578 void 1579 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1580 { 1581 1582 ifd->ifi_type = ifp->if_type; 1583 ifd->ifi_physical = 0; 1584 ifd->ifi_addrlen = ifp->if_addrlen; 1585 ifd->ifi_hdrlen = ifp->if_hdrlen; 1586 ifd->ifi_link_state = ifp->if_link_state; 1587 ifd->ifi_vhid = 0; 1588 ifd->ifi_datalen = sizeof(struct if_data); 1589 ifd->ifi_mtu = ifp->if_mtu; 1590 ifd->ifi_metric = ifp->if_metric; 1591 ifd->ifi_baudrate = ifp->if_baudrate; 1592 ifd->ifi_hwassist = ifp->if_hwassist; 1593 ifd->ifi_epoch = ifp->if_epoch; 1594 ifd->ifi_lastchange = ifp->if_lastchange; 1595 1596 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1597 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1598 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1599 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1600 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1601 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1602 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1603 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1604 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1605 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1606 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1607 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1608 } 1609 1610 /* 1611 * Initialization, destruction and refcounting functions for ifaddrs. 1612 */ 1613 struct ifaddr * 1614 ifa_alloc(size_t size, int flags) 1615 { 1616 struct ifaddr *ifa; 1617 1618 KASSERT(size >= sizeof(struct ifaddr), 1619 ("%s: invalid size %zu", __func__, size)); 1620 1621 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1622 if (ifa == NULL) 1623 return (NULL); 1624 1625 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1626 goto fail; 1627 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1628 goto fail; 1629 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1630 goto fail; 1631 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1632 goto fail; 1633 1634 refcount_init(&ifa->ifa_refcnt, 1); 1635 1636 return (ifa); 1637 1638 fail: 1639 /* free(NULL) is okay */ 1640 counter_u64_free(ifa->ifa_opackets); 1641 counter_u64_free(ifa->ifa_ipackets); 1642 counter_u64_free(ifa->ifa_obytes); 1643 counter_u64_free(ifa->ifa_ibytes); 1644 free(ifa, M_IFADDR); 1645 1646 return (NULL); 1647 } 1648 1649 void 1650 ifa_ref(struct ifaddr *ifa) 1651 { 1652 u_int old __diagused; 1653 1654 old = refcount_acquire(&ifa->ifa_refcnt); 1655 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1656 } 1657 1658 int 1659 ifa_try_ref(struct ifaddr *ifa) 1660 { 1661 1662 NET_EPOCH_ASSERT(); 1663 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1664 } 1665 1666 static void 1667 ifa_destroy(epoch_context_t ctx) 1668 { 1669 struct ifaddr *ifa; 1670 1671 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1672 counter_u64_free(ifa->ifa_opackets); 1673 counter_u64_free(ifa->ifa_ipackets); 1674 counter_u64_free(ifa->ifa_obytes); 1675 counter_u64_free(ifa->ifa_ibytes); 1676 free(ifa, M_IFADDR); 1677 } 1678 1679 void 1680 ifa_free(struct ifaddr *ifa) 1681 { 1682 1683 if (refcount_release(&ifa->ifa_refcnt)) 1684 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1685 } 1686 1687 /* 1688 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1689 * structs used to represent other address families, it is necessary 1690 * to perform a different comparison. 1691 */ 1692 static bool 1693 sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b) 1694 { 1695 const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a; 1696 const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b; 1697 1698 return (sdl1->sdl_len == sdl2->sdl_len && 1699 bcmp(sdl1->sdl_data + sdl1->sdl_nlen, 1700 sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0); 1701 } 1702 1703 /* 1704 * Locate an interface based on a complete address. 1705 */ 1706 /*ARGSUSED*/ 1707 struct ifaddr * 1708 ifa_ifwithaddr(const struct sockaddr *addr) 1709 { 1710 struct ifnet *ifp; 1711 struct ifaddr *ifa; 1712 1713 NET_EPOCH_ASSERT(); 1714 1715 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1716 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1717 if (ifa->ifa_addr->sa_family != addr->sa_family) 1718 continue; 1719 if (sa_equal(addr, ifa->ifa_addr)) { 1720 goto done; 1721 } 1722 /* IP6 doesn't have broadcast */ 1723 if ((ifp->if_flags & IFF_BROADCAST) && 1724 ifa->ifa_broadaddr && 1725 ifa->ifa_broadaddr->sa_len != 0 && 1726 sa_equal(ifa->ifa_broadaddr, addr)) { 1727 goto done; 1728 } 1729 } 1730 } 1731 ifa = NULL; 1732 done: 1733 return (ifa); 1734 } 1735 1736 int 1737 ifa_ifwithaddr_check(const struct sockaddr *addr) 1738 { 1739 struct epoch_tracker et; 1740 int rc; 1741 1742 NET_EPOCH_ENTER(et); 1743 rc = (ifa_ifwithaddr(addr) != NULL); 1744 NET_EPOCH_EXIT(et); 1745 return (rc); 1746 } 1747 1748 /* 1749 * Locate an interface based on the broadcast address. 1750 */ 1751 /* ARGSUSED */ 1752 struct ifaddr * 1753 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1754 { 1755 struct ifnet *ifp; 1756 struct ifaddr *ifa; 1757 1758 NET_EPOCH_ASSERT(); 1759 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1760 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1761 continue; 1762 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1763 if (ifa->ifa_addr->sa_family != addr->sa_family) 1764 continue; 1765 if ((ifp->if_flags & IFF_BROADCAST) && 1766 ifa->ifa_broadaddr && 1767 ifa->ifa_broadaddr->sa_len != 0 && 1768 sa_equal(ifa->ifa_broadaddr, addr)) { 1769 goto done; 1770 } 1771 } 1772 } 1773 ifa = NULL; 1774 done: 1775 return (ifa); 1776 } 1777 1778 /* 1779 * Locate the point to point interface with a given destination address. 1780 */ 1781 /*ARGSUSED*/ 1782 struct ifaddr * 1783 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1784 { 1785 struct ifnet *ifp; 1786 struct ifaddr *ifa; 1787 1788 NET_EPOCH_ASSERT(); 1789 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1790 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1791 continue; 1792 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1793 continue; 1794 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1795 if (ifa->ifa_addr->sa_family != addr->sa_family) 1796 continue; 1797 if (ifa->ifa_dstaddr != NULL && 1798 sa_equal(addr, ifa->ifa_dstaddr)) { 1799 goto done; 1800 } 1801 } 1802 } 1803 ifa = NULL; 1804 done: 1805 return (ifa); 1806 } 1807 1808 /* 1809 * Find an interface on a specific network. If many, choice 1810 * is most specific found. 1811 */ 1812 struct ifaddr * 1813 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1814 { 1815 struct ifnet *ifp; 1816 struct ifaddr *ifa; 1817 struct ifaddr *ifa_maybe = NULL; 1818 u_int af = addr->sa_family; 1819 const char *addr_data = addr->sa_data, *cplim; 1820 1821 NET_EPOCH_ASSERT(); 1822 /* 1823 * AF_LINK addresses can be looked up directly by their index number, 1824 * so do that if we can. 1825 */ 1826 if (af == AF_LINK) { 1827 ifp = ifnet_byindex( 1828 ((const struct sockaddr_dl *)addr)->sdl_index); 1829 return (ifp ? ifp->if_addr : NULL); 1830 } 1831 1832 /* 1833 * Scan though each interface, looking for ones that have addresses 1834 * in this address family and the requested fib. 1835 */ 1836 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1837 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1838 continue; 1839 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1840 const char *cp, *cp2, *cp3; 1841 1842 if (ifa->ifa_addr->sa_family != af) 1843 next: continue; 1844 if (af == AF_INET && 1845 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1846 /* 1847 * This is a bit broken as it doesn't 1848 * take into account that the remote end may 1849 * be a single node in the network we are 1850 * looking for. 1851 * The trouble is that we don't know the 1852 * netmask for the remote end. 1853 */ 1854 if (ifa->ifa_dstaddr != NULL && 1855 sa_equal(addr, ifa->ifa_dstaddr)) { 1856 goto done; 1857 } 1858 } else { 1859 /* 1860 * Scan all the bits in the ifa's address. 1861 * If a bit dissagrees with what we are 1862 * looking for, mask it with the netmask 1863 * to see if it really matters. 1864 * (A byte at a time) 1865 */ 1866 if (ifa->ifa_netmask == 0) 1867 continue; 1868 cp = addr_data; 1869 cp2 = ifa->ifa_addr->sa_data; 1870 cp3 = ifa->ifa_netmask->sa_data; 1871 cplim = ifa->ifa_netmask->sa_len 1872 + (char *)ifa->ifa_netmask; 1873 while (cp3 < cplim) 1874 if ((*cp++ ^ *cp2++) & *cp3++) 1875 goto next; /* next address! */ 1876 /* 1877 * If the netmask of what we just found 1878 * is more specific than what we had before 1879 * (if we had one), or if the virtual status 1880 * of new prefix is better than of the old one, 1881 * then remember the new one before continuing 1882 * to search for an even better one. 1883 */ 1884 if (ifa_maybe == NULL || 1885 ifa_preferred(ifa_maybe, ifa) || 1886 rn_refines((caddr_t)ifa->ifa_netmask, 1887 (caddr_t)ifa_maybe->ifa_netmask)) { 1888 ifa_maybe = ifa; 1889 } 1890 } 1891 } 1892 } 1893 ifa = ifa_maybe; 1894 ifa_maybe = NULL; 1895 done: 1896 return (ifa); 1897 } 1898 1899 /* 1900 * Find an interface address specific to an interface best matching 1901 * a given address. 1902 */ 1903 struct ifaddr * 1904 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1905 { 1906 struct ifaddr *ifa; 1907 const char *cp, *cp2, *cp3; 1908 char *cplim; 1909 struct ifaddr *ifa_maybe = NULL; 1910 u_int af = addr->sa_family; 1911 1912 if (af >= AF_MAX) 1913 return (NULL); 1914 1915 NET_EPOCH_ASSERT(); 1916 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1917 if (ifa->ifa_addr->sa_family != af) 1918 continue; 1919 if (ifa_maybe == NULL) 1920 ifa_maybe = ifa; 1921 if (ifa->ifa_netmask == 0) { 1922 if (sa_equal(addr, ifa->ifa_addr) || 1923 (ifa->ifa_dstaddr && 1924 sa_equal(addr, ifa->ifa_dstaddr))) 1925 goto done; 1926 continue; 1927 } 1928 if (ifp->if_flags & IFF_POINTOPOINT) { 1929 if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr)) 1930 goto done; 1931 } else { 1932 cp = addr->sa_data; 1933 cp2 = ifa->ifa_addr->sa_data; 1934 cp3 = ifa->ifa_netmask->sa_data; 1935 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1936 for (; cp3 < cplim; cp3++) 1937 if ((*cp++ ^ *cp2++) & *cp3) 1938 break; 1939 if (cp3 == cplim) 1940 goto done; 1941 } 1942 } 1943 ifa = ifa_maybe; 1944 done: 1945 return (ifa); 1946 } 1947 1948 /* 1949 * See whether new ifa is better than current one: 1950 * 1) A non-virtual one is preferred over virtual. 1951 * 2) A virtual in master state preferred over any other state. 1952 * 1953 * Used in several address selecting functions. 1954 */ 1955 int 1956 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 1957 { 1958 1959 return (cur->ifa_carp && (!next->ifa_carp || 1960 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 1961 } 1962 1963 struct sockaddr_dl * 1964 link_alloc_sdl(size_t size, int flags) 1965 { 1966 1967 return (malloc(size, M_TEMP, flags)); 1968 } 1969 1970 void 1971 link_free_sdl(struct sockaddr *sa) 1972 { 1973 free(sa, M_TEMP); 1974 } 1975 1976 /* 1977 * Fills in given sdl with interface basic info. 1978 * Returns pointer to filled sdl. 1979 */ 1980 struct sockaddr_dl * 1981 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 1982 { 1983 struct sockaddr_dl *sdl; 1984 1985 sdl = (struct sockaddr_dl *)paddr; 1986 memset(sdl, 0, sizeof(struct sockaddr_dl)); 1987 sdl->sdl_len = sizeof(struct sockaddr_dl); 1988 sdl->sdl_family = AF_LINK; 1989 sdl->sdl_index = ifp->if_index; 1990 sdl->sdl_type = iftype; 1991 1992 return (sdl); 1993 } 1994 1995 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 1996 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 1997 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 1998 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 1999 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2000 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2001 int (*vlan_setcookie_p)(struct ifnet *, void *); 2002 void *(*vlan_cookie_p)(struct ifnet *); 2003 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 2004 2005 /* 2006 * Handle a change in the interface link state. To avoid LORs 2007 * between driver lock and upper layer locks, as well as possible 2008 * recursions, we post event to taskqueue, and all job 2009 * is done in static do_link_state_change(). 2010 */ 2011 void 2012 if_link_state_change(struct ifnet *ifp, int link_state) 2013 { 2014 /* Return if state hasn't changed. */ 2015 if (ifp->if_link_state == link_state) 2016 return; 2017 2018 ifp->if_link_state = link_state; 2019 2020 /* XXXGL: reference ifp? */ 2021 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2022 } 2023 2024 static void 2025 do_link_state_change(void *arg, int pending) 2026 { 2027 struct ifnet *ifp; 2028 int link_state; 2029 2030 ifp = arg; 2031 link_state = ifp->if_link_state; 2032 2033 CURVNET_SET(ifp->if_vnet); 2034 rt_ifmsg(ifp, 0); 2035 if (ifp->if_vlantrunk != NULL) 2036 (*vlan_link_state_p)(ifp); 2037 if (ifp->if_carp) 2038 (*carp_linkstate_p)(ifp); 2039 if (ifp->if_bridge) 2040 ifp->if_bridge_linkstate(ifp); 2041 if (ifp->if_lagg) 2042 (*lagg_linkstate_p)(ifp, link_state); 2043 2044 if (IS_DEFAULT_VNET(curvnet)) 2045 devctl_notify("IFNET", ifp->if_xname, 2046 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2047 NULL); 2048 if (pending > 1) 2049 if_printf(ifp, "%d link states coalesced\n", pending); 2050 if (log_link_state_change) 2051 if_printf(ifp, "link state changed to %s\n", 2052 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2053 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2054 CURVNET_RESTORE(); 2055 } 2056 2057 /* 2058 * Mark an interface down and notify protocols of 2059 * the transition. 2060 */ 2061 void 2062 if_down(struct ifnet *ifp) 2063 { 2064 2065 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2066 2067 ifp->if_flags &= ~IFF_UP; 2068 getmicrotime(&ifp->if_lastchange); 2069 ifp->if_qflush(ifp); 2070 2071 if (ifp->if_carp) 2072 (*carp_linkstate_p)(ifp); 2073 rt_ifmsg(ifp, IFF_UP); 2074 } 2075 2076 /* 2077 * Mark an interface up and notify protocols of 2078 * the transition. 2079 */ 2080 void 2081 if_up(struct ifnet *ifp) 2082 { 2083 2084 ifp->if_flags |= IFF_UP; 2085 getmicrotime(&ifp->if_lastchange); 2086 if (ifp->if_carp) 2087 (*carp_linkstate_p)(ifp); 2088 rt_ifmsg(ifp, IFF_UP); 2089 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2090 } 2091 2092 /* 2093 * Flush an interface queue. 2094 */ 2095 void 2096 if_qflush(struct ifnet *ifp) 2097 { 2098 struct mbuf *m, *n; 2099 struct ifaltq *ifq; 2100 2101 ifq = &ifp->if_snd; 2102 IFQ_LOCK(ifq); 2103 #ifdef ALTQ 2104 if (ALTQ_IS_ENABLED(ifq)) 2105 ALTQ_PURGE(ifq); 2106 #endif 2107 n = ifq->ifq_head; 2108 while ((m = n) != NULL) { 2109 n = m->m_nextpkt; 2110 m_freem(m); 2111 } 2112 ifq->ifq_head = 0; 2113 ifq->ifq_tail = 0; 2114 ifq->ifq_len = 0; 2115 IFQ_UNLOCK(ifq); 2116 } 2117 2118 /* 2119 * Map interface name to interface structure pointer, with or without 2120 * returning a reference. 2121 */ 2122 struct ifnet * 2123 ifunit_ref(const char *name) 2124 { 2125 struct epoch_tracker et; 2126 struct ifnet *ifp; 2127 2128 NET_EPOCH_ENTER(et); 2129 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2130 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2131 !(ifp->if_flags & IFF_DYING)) { 2132 MPASS(ifp->if_vnet == curvnet); 2133 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2134 if_ref(ifp); 2135 break; 2136 } 2137 } 2138 NET_EPOCH_EXIT(et); 2139 return (ifp); 2140 } 2141 2142 struct ifnet * 2143 ifunit(const char *name) 2144 { 2145 struct epoch_tracker et; 2146 struct ifnet *ifp; 2147 2148 NET_EPOCH_ENTER(et); 2149 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2150 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) { 2151 MPASS(refcount_load(&ifp->if_refcount) > 0); 2152 MPASS(ifp->if_vnet == curvnet); 2153 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2154 break; 2155 } 2156 } 2157 NET_EPOCH_EXIT(et); 2158 return (ifp); 2159 } 2160 2161 void * 2162 ifr_buffer_get_buffer(void *data) 2163 { 2164 union ifreq_union *ifrup; 2165 2166 ifrup = data; 2167 #ifdef COMPAT_FREEBSD32 2168 if (SV_CURPROC_FLAG(SV_ILP32)) 2169 return ((void *)(uintptr_t) 2170 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2171 #endif 2172 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2173 } 2174 2175 static void 2176 ifr_buffer_set_buffer_null(void *data) 2177 { 2178 union ifreq_union *ifrup; 2179 2180 ifrup = data; 2181 #ifdef COMPAT_FREEBSD32 2182 if (SV_CURPROC_FLAG(SV_ILP32)) 2183 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2184 else 2185 #endif 2186 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2187 } 2188 2189 size_t 2190 ifr_buffer_get_length(void *data) 2191 { 2192 union ifreq_union *ifrup; 2193 2194 ifrup = data; 2195 #ifdef COMPAT_FREEBSD32 2196 if (SV_CURPROC_FLAG(SV_ILP32)) 2197 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2198 #endif 2199 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2200 } 2201 2202 static void 2203 ifr_buffer_set_length(void *data, size_t len) 2204 { 2205 union ifreq_union *ifrup; 2206 2207 ifrup = data; 2208 #ifdef COMPAT_FREEBSD32 2209 if (SV_CURPROC_FLAG(SV_ILP32)) 2210 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2211 else 2212 #endif 2213 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2214 } 2215 2216 void * 2217 ifr_data_get_ptr(void *ifrp) 2218 { 2219 union ifreq_union *ifrup; 2220 2221 ifrup = ifrp; 2222 #ifdef COMPAT_FREEBSD32 2223 if (SV_CURPROC_FLAG(SV_ILP32)) 2224 return ((void *)(uintptr_t) 2225 ifrup->ifr32.ifr_ifru.ifru_data); 2226 #endif 2227 return (ifrup->ifr.ifr_ifru.ifru_data); 2228 } 2229 2230 struct ifcap_nv_bit_name { 2231 uint64_t cap_bit; 2232 const char *cap_name; 2233 }; 2234 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2235 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2236 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2237 CAPNV(RXCSUM), 2238 CAPNV(TXCSUM), 2239 CAPNV(NETCONS), 2240 CAPNV(VLAN_MTU), 2241 CAPNV(VLAN_HWTAGGING), 2242 CAPNV(JUMBO_MTU), 2243 CAPNV(POLLING), 2244 CAPNV(VLAN_HWCSUM), 2245 CAPNV(TSO4), 2246 CAPNV(TSO6), 2247 CAPNV(LRO), 2248 CAPNV(WOL_UCAST), 2249 CAPNV(WOL_MCAST), 2250 CAPNV(WOL_MAGIC), 2251 CAPNV(TOE4), 2252 CAPNV(TOE6), 2253 CAPNV(VLAN_HWFILTER), 2254 CAPNV(VLAN_HWTSO), 2255 CAPNV(LINKSTATE), 2256 CAPNV(NETMAP), 2257 CAPNV(RXCSUM_IPV6), 2258 CAPNV(TXCSUM_IPV6), 2259 CAPNV(HWSTATS), 2260 CAPNV(TXRTLMT), 2261 CAPNV(HWRXTSTMP), 2262 CAPNV(MEXTPG), 2263 CAPNV(TXTLS4), 2264 CAPNV(TXTLS6), 2265 CAPNV(VXLAN_HWCSUM), 2266 CAPNV(VXLAN_HWTSO), 2267 CAPNV(TXTLS_RTLMT), 2268 {0, NULL} 2269 }; 2270 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \ 2271 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2272 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2273 CAP2NV(RXTLS4), 2274 CAP2NV(RXTLS6), 2275 CAP2NV(IPSEC_OFFLOAD), 2276 CAP2NV(GENEVE_HWCSUM), 2277 CAP2NV(GENEVE_HWTSO), 2278 {0, NULL} 2279 }; 2280 #undef CAPNV 2281 #undef CAP2NV 2282 2283 int 2284 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2285 const struct ifcap_nv_bit_name *nn, bool all) 2286 { 2287 int i, res; 2288 2289 res = 0; 2290 for (i = 0; nn[i].cap_name != NULL; i++) { 2291 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2292 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2293 res |= nn[i].cap_bit; 2294 } else { 2295 res |= *old_cap & nn[i].cap_bit; 2296 } 2297 } 2298 return (res); 2299 } 2300 2301 void 2302 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2303 int ifr_cap, int ifr_req) 2304 { 2305 int i; 2306 2307 for (i = 0; nn[i].cap_name != NULL; i++) { 2308 if ((nn[i].cap_bit & ifr_cap) != 0) { 2309 nvlist_add_bool(nv, nn[i].cap_name, 2310 (nn[i].cap_bit & ifr_req) != 0); 2311 } 2312 } 2313 } 2314 2315 /* 2316 * Hardware specific interface ioctls. 2317 */ 2318 int 2319 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2320 { 2321 struct ifreq *ifr; 2322 int error = 0, do_ifup = 0; 2323 int new_flags, temp_flags; 2324 size_t descrlen, nvbuflen; 2325 char *descrbuf; 2326 char new_name[IFNAMSIZ]; 2327 void *buf; 2328 nvlist_t *nvcap; 2329 struct siocsifcapnv_driver_data drv_ioctl_data; 2330 2331 ifr = (struct ifreq *)data; 2332 switch (cmd) { 2333 case SIOCGIFINDEX: 2334 ifr->ifr_index = ifp->if_index; 2335 break; 2336 2337 case SIOCGIFFLAGS: 2338 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2339 ifr->ifr_flags = temp_flags & 0xffff; 2340 ifr->ifr_flagshigh = temp_flags >> 16; 2341 break; 2342 2343 case SIOCGIFCAP: 2344 ifr->ifr_reqcap = ifp->if_capabilities; 2345 ifr->ifr_curcap = ifp->if_capenable; 2346 break; 2347 2348 case SIOCGIFCAPNV: 2349 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2350 error = EINVAL; 2351 break; 2352 } 2353 buf = NULL; 2354 nvcap = nvlist_create(0); 2355 for (;;) { 2356 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2357 ifp->if_capabilities, ifp->if_capenable); 2358 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2359 ifp->if_capabilities2, ifp->if_capenable2); 2360 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2361 __DECONST(caddr_t, nvcap)); 2362 if (error != 0) { 2363 if_printf(ifp, 2364 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2365 error); 2366 break; 2367 } 2368 buf = nvlist_pack(nvcap, &nvbuflen); 2369 if (buf == NULL) { 2370 error = nvlist_error(nvcap); 2371 if (error == 0) 2372 error = EDOOFUS; 2373 break; 2374 } 2375 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2376 ifr->ifr_cap_nv.length = nvbuflen; 2377 ifr->ifr_cap_nv.buffer = NULL; 2378 error = EFBIG; 2379 break; 2380 } 2381 ifr->ifr_cap_nv.length = nvbuflen; 2382 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2383 break; 2384 } 2385 free(buf, M_NVLIST); 2386 nvlist_destroy(nvcap); 2387 break; 2388 2389 case SIOCGIFDATA: 2390 { 2391 struct if_data ifd; 2392 2393 /* Ensure uninitialised padding is not leaked. */ 2394 memset(&ifd, 0, sizeof(ifd)); 2395 2396 if_data_copy(ifp, &ifd); 2397 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2398 break; 2399 } 2400 2401 #ifdef MAC 2402 case SIOCGIFMAC: 2403 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2404 break; 2405 #endif 2406 2407 case SIOCGIFMETRIC: 2408 ifr->ifr_metric = ifp->if_metric; 2409 break; 2410 2411 case SIOCGIFMTU: 2412 ifr->ifr_mtu = ifp->if_mtu; 2413 break; 2414 2415 case SIOCGIFPHYS: 2416 /* XXXGL: did this ever worked? */ 2417 ifr->ifr_phys = 0; 2418 break; 2419 2420 case SIOCGIFDESCR: 2421 error = 0; 2422 sx_slock(&ifdescr_sx); 2423 if (ifp->if_description == NULL) 2424 error = ENOMSG; 2425 else { 2426 /* space for terminating nul */ 2427 descrlen = strlen(ifp->if_description) + 1; 2428 if (ifr_buffer_get_length(ifr) < descrlen) 2429 ifr_buffer_set_buffer_null(ifr); 2430 else 2431 error = copyout(ifp->if_description, 2432 ifr_buffer_get_buffer(ifr), descrlen); 2433 ifr_buffer_set_length(ifr, descrlen); 2434 } 2435 sx_sunlock(&ifdescr_sx); 2436 break; 2437 2438 case SIOCSIFDESCR: 2439 error = priv_check(td, PRIV_NET_SETIFDESCR); 2440 if (error) 2441 return (error); 2442 2443 /* 2444 * Copy only (length-1) bytes to make sure that 2445 * if_description is always nul terminated. The 2446 * length parameter is supposed to count the 2447 * terminating nul in. 2448 */ 2449 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2450 return (ENAMETOOLONG); 2451 else if (ifr_buffer_get_length(ifr) == 0) 2452 descrbuf = NULL; 2453 else { 2454 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2455 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2456 ifr_buffer_get_length(ifr) - 1); 2457 if (error) { 2458 if_freedescr(descrbuf); 2459 break; 2460 } 2461 } 2462 2463 if_setdescr(ifp, descrbuf); 2464 getmicrotime(&ifp->if_lastchange); 2465 break; 2466 2467 case SIOCGIFFIB: 2468 ifr->ifr_fib = ifp->if_fib; 2469 break; 2470 2471 case SIOCSIFFIB: 2472 error = priv_check(td, PRIV_NET_SETIFFIB); 2473 if (error) 2474 return (error); 2475 if (ifr->ifr_fib >= rt_numfibs) 2476 return (EINVAL); 2477 2478 ifp->if_fib = ifr->ifr_fib; 2479 break; 2480 2481 case SIOCSIFFLAGS: 2482 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2483 if (error) 2484 return (error); 2485 /* 2486 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2487 * check, so we don't need special handling here yet. 2488 */ 2489 new_flags = (ifr->ifr_flags & 0xffff) | 2490 (ifr->ifr_flagshigh << 16); 2491 if (ifp->if_flags & IFF_UP && 2492 (new_flags & IFF_UP) == 0) { 2493 if_down(ifp); 2494 } else if (new_flags & IFF_UP && 2495 (ifp->if_flags & IFF_UP) == 0) { 2496 do_ifup = 1; 2497 } 2498 2499 /* 2500 * See if the promiscuous mode or allmulti bits are about to 2501 * flip. They require special handling because in-kernel 2502 * consumers may indepdently toggle them. 2503 */ 2504 if_setppromisc(ifp, new_flags & IFF_PPROMISC); 2505 if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) { 2506 if (new_flags & IFF_PALLMULTI) 2507 ifp->if_flags |= IFF_ALLMULTI; 2508 else if (ifp->if_amcount == 0) 2509 ifp->if_flags &= ~IFF_ALLMULTI; 2510 } 2511 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2512 (new_flags &~ IFF_CANTCHANGE); 2513 if (ifp->if_ioctl) { 2514 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2515 } 2516 if (do_ifup) 2517 if_up(ifp); 2518 getmicrotime(&ifp->if_lastchange); 2519 break; 2520 2521 case SIOCSIFCAP: 2522 error = priv_check(td, PRIV_NET_SETIFCAP); 2523 if (error != 0) 2524 return (error); 2525 if (ifp->if_ioctl == NULL) 2526 return (EOPNOTSUPP); 2527 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2528 return (EINVAL); 2529 error = (*ifp->if_ioctl)(ifp, cmd, data); 2530 if (error == 0) 2531 getmicrotime(&ifp->if_lastchange); 2532 break; 2533 2534 case SIOCSIFCAPNV: 2535 error = priv_check(td, PRIV_NET_SETIFCAP); 2536 if (error != 0) 2537 return (error); 2538 if (ifp->if_ioctl == NULL) 2539 return (EOPNOTSUPP); 2540 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2541 return (EINVAL); 2542 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2543 return (EINVAL); 2544 nvcap = NULL; 2545 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2546 for (;;) { 2547 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2548 ifr->ifr_cap_nv.length); 2549 if (error != 0) 2550 break; 2551 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2552 if (nvcap == NULL) { 2553 error = EINVAL; 2554 break; 2555 } 2556 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2557 &ifp->if_capenable, ifcap_nv_bit_names, false); 2558 if ((drv_ioctl_data.reqcap & 2559 ~ifp->if_capabilities) != 0) { 2560 error = EINVAL; 2561 break; 2562 } 2563 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2564 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2565 if ((drv_ioctl_data.reqcap2 & 2566 ~ifp->if_capabilities2) != 0) { 2567 error = EINVAL; 2568 break; 2569 } 2570 drv_ioctl_data.nvcap = nvcap; 2571 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2572 (caddr_t)&drv_ioctl_data); 2573 break; 2574 } 2575 nvlist_destroy(nvcap); 2576 free(buf, M_TEMP); 2577 if (error == 0) 2578 getmicrotime(&ifp->if_lastchange); 2579 break; 2580 2581 #ifdef MAC 2582 case SIOCSIFMAC: 2583 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2584 break; 2585 #endif 2586 2587 case SIOCSIFNAME: 2588 error = priv_check(td, PRIV_NET_SETIFNAME); 2589 if (error) 2590 return (error); 2591 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2592 NULL); 2593 if (error != 0) 2594 return (error); 2595 error = if_rename(ifp, new_name); 2596 break; 2597 2598 case SIOCSIFMETRIC: 2599 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2600 if (error) 2601 return (error); 2602 ifp->if_metric = ifr->ifr_metric; 2603 getmicrotime(&ifp->if_lastchange); 2604 break; 2605 2606 case SIOCSIFPHYS: 2607 error = priv_check(td, PRIV_NET_SETIFPHYS); 2608 if (error) 2609 return (error); 2610 if (ifp->if_ioctl == NULL) 2611 return (EOPNOTSUPP); 2612 error = (*ifp->if_ioctl)(ifp, cmd, data); 2613 if (error == 0) 2614 getmicrotime(&ifp->if_lastchange); 2615 break; 2616 2617 case SIOCSIFMTU: 2618 { 2619 u_long oldmtu = ifp->if_mtu; 2620 2621 error = priv_check(td, PRIV_NET_SETIFMTU); 2622 if (error) 2623 return (error); 2624 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2625 return (EINVAL); 2626 if (ifp->if_ioctl == NULL) 2627 return (EOPNOTSUPP); 2628 /* Disallow MTU changes on bridge member interfaces. */ 2629 if (ifp->if_bridge) 2630 return (EOPNOTSUPP); 2631 error = (*ifp->if_ioctl)(ifp, cmd, data); 2632 if (error == 0) { 2633 getmicrotime(&ifp->if_lastchange); 2634 rt_ifmsg(ifp, 0); 2635 #ifdef INET 2636 DEBUGNET_NOTIFY_MTU(ifp); 2637 #endif 2638 } 2639 /* 2640 * If the link MTU changed, do network layer specific procedure. 2641 */ 2642 if (ifp->if_mtu != oldmtu) 2643 if_notifymtu(ifp); 2644 break; 2645 } 2646 2647 case SIOCADDMULTI: 2648 case SIOCDELMULTI: 2649 if (cmd == SIOCADDMULTI) 2650 error = priv_check(td, PRIV_NET_ADDMULTI); 2651 else 2652 error = priv_check(td, PRIV_NET_DELMULTI); 2653 if (error) 2654 return (error); 2655 2656 /* Don't allow group membership on non-multicast interfaces. */ 2657 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2658 return (EOPNOTSUPP); 2659 2660 /* Don't let users screw up protocols' entries. */ 2661 if (ifr->ifr_addr.sa_family != AF_LINK) 2662 return (EINVAL); 2663 2664 if (cmd == SIOCADDMULTI) { 2665 struct epoch_tracker et; 2666 struct ifmultiaddr *ifma; 2667 2668 /* 2669 * Userland is only permitted to join groups once 2670 * via the if_addmulti() KPI, because it cannot hold 2671 * struct ifmultiaddr * between calls. It may also 2672 * lose a race while we check if the membership 2673 * already exists. 2674 */ 2675 NET_EPOCH_ENTER(et); 2676 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2677 NET_EPOCH_EXIT(et); 2678 if (ifma != NULL) 2679 error = EADDRINUSE; 2680 else 2681 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2682 } else { 2683 error = if_delmulti(ifp, &ifr->ifr_addr); 2684 } 2685 if (error == 0) 2686 getmicrotime(&ifp->if_lastchange); 2687 break; 2688 2689 case SIOCSIFPHYADDR: 2690 case SIOCDIFPHYADDR: 2691 #ifdef INET6 2692 case SIOCSIFPHYADDR_IN6: 2693 #endif 2694 case SIOCSIFMEDIA: 2695 case SIOCSIFGENERIC: 2696 error = priv_check(td, PRIV_NET_HWIOCTL); 2697 if (error) 2698 return (error); 2699 if (ifp->if_ioctl == NULL) 2700 return (EOPNOTSUPP); 2701 error = (*ifp->if_ioctl)(ifp, cmd, data); 2702 if (error == 0) 2703 getmicrotime(&ifp->if_lastchange); 2704 break; 2705 2706 case SIOCGIFSTATUS: 2707 case SIOCGIFPSRCADDR: 2708 case SIOCGIFPDSTADDR: 2709 case SIOCGIFMEDIA: 2710 case SIOCGIFXMEDIA: 2711 case SIOCGIFGENERIC: 2712 case SIOCGIFRSSKEY: 2713 case SIOCGIFRSSHASH: 2714 case SIOCGIFDOWNREASON: 2715 if (ifp->if_ioctl == NULL) 2716 return (EOPNOTSUPP); 2717 error = (*ifp->if_ioctl)(ifp, cmd, data); 2718 break; 2719 2720 case SIOCSIFLLADDR: 2721 error = priv_check(td, PRIV_NET_SETLLADDR); 2722 if (error) 2723 return (error); 2724 error = if_setlladdr(ifp, 2725 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2726 break; 2727 2728 case SIOCGHWADDR: 2729 error = if_gethwaddr(ifp, ifr); 2730 break; 2731 2732 case SIOCAIFGROUP: 2733 { 2734 const char *groupname; 2735 2736 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2737 if (error) 2738 return (error); 2739 groupname = ((struct ifgroupreq *)data)->ifgr_group; 2740 if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ) 2741 return (EINVAL); 2742 error = if_addgroup(ifp, groupname); 2743 if (error != 0) 2744 return (error); 2745 break; 2746 } 2747 case SIOCGIFGROUP: 2748 { 2749 struct epoch_tracker et; 2750 2751 NET_EPOCH_ENTER(et); 2752 error = if_getgroup((struct ifgroupreq *)data, ifp); 2753 NET_EPOCH_EXIT(et); 2754 break; 2755 } 2756 2757 case SIOCDIFGROUP: 2758 { 2759 const char *groupname; 2760 2761 error = priv_check(td, PRIV_NET_DELIFGROUP); 2762 if (error) 2763 return (error); 2764 groupname = ((struct ifgroupreq *)data)->ifgr_group; 2765 if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ) 2766 return (EINVAL); 2767 error = if_delgroup(ifp, groupname); 2768 if (error != 0) 2769 return (error); 2770 break; 2771 } 2772 default: 2773 error = ENOIOCTL; 2774 break; 2775 } 2776 return (error); 2777 } 2778 2779 /* 2780 * Interface ioctls. 2781 */ 2782 int 2783 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2784 { 2785 #ifdef COMPAT_FREEBSD32 2786 union { 2787 struct ifconf ifc; 2788 struct ifdrv ifd; 2789 struct ifgroupreq ifgr; 2790 struct ifmediareq ifmr; 2791 } thunk; 2792 u_long saved_cmd; 2793 struct ifconf32 *ifc32; 2794 struct ifdrv32 *ifd32; 2795 struct ifgroupreq32 *ifgr32; 2796 struct ifmediareq32 *ifmr32; 2797 #endif 2798 struct ifnet *ifp; 2799 struct ifreq *ifr; 2800 int error; 2801 int oif_flags; 2802 #ifdef VIMAGE 2803 bool shutdown; 2804 #endif 2805 2806 CURVNET_SET(so->so_vnet); 2807 #ifdef VIMAGE 2808 /* Make sure the VNET is stable. */ 2809 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2810 if (shutdown) { 2811 CURVNET_RESTORE(); 2812 return (EBUSY); 2813 } 2814 #endif 2815 2816 #ifdef COMPAT_FREEBSD32 2817 saved_cmd = cmd; 2818 switch (cmd) { 2819 case SIOCGIFCONF32: 2820 ifc32 = (struct ifconf32 *)data; 2821 thunk.ifc.ifc_len = ifc32->ifc_len; 2822 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2823 data = (caddr_t)&thunk.ifc; 2824 cmd = SIOCGIFCONF; 2825 break; 2826 case SIOCGDRVSPEC32: 2827 case SIOCSDRVSPEC32: 2828 ifd32 = (struct ifdrv32 *)data; 2829 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2830 sizeof(thunk.ifd.ifd_name)); 2831 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2832 thunk.ifd.ifd_len = ifd32->ifd_len; 2833 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2834 data = (caddr_t)&thunk.ifd; 2835 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2836 break; 2837 case SIOCAIFGROUP32: 2838 case SIOCGIFGROUP32: 2839 case SIOCDIFGROUP32: 2840 case SIOCGIFGMEMB32: 2841 ifgr32 = (struct ifgroupreq32 *)data; 2842 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2843 sizeof(thunk.ifgr.ifgr_name)); 2844 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2845 switch (cmd) { 2846 case SIOCAIFGROUP32: 2847 case SIOCDIFGROUP32: 2848 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2849 sizeof(thunk.ifgr.ifgr_group)); 2850 break; 2851 case SIOCGIFGROUP32: 2852 case SIOCGIFGMEMB32: 2853 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2854 break; 2855 } 2856 data = (caddr_t)&thunk.ifgr; 2857 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2858 break; 2859 case SIOCGIFMEDIA32: 2860 case SIOCGIFXMEDIA32: 2861 ifmr32 = (struct ifmediareq32 *)data; 2862 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2863 sizeof(thunk.ifmr.ifm_name)); 2864 thunk.ifmr.ifm_current = ifmr32->ifm_current; 2865 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 2866 thunk.ifmr.ifm_status = ifmr32->ifm_status; 2867 thunk.ifmr.ifm_active = ifmr32->ifm_active; 2868 thunk.ifmr.ifm_count = ifmr32->ifm_count; 2869 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 2870 data = (caddr_t)&thunk.ifmr; 2871 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 2872 break; 2873 } 2874 #endif 2875 2876 switch (cmd) { 2877 case SIOCGIFCONF: 2878 error = ifconf(cmd, data); 2879 goto out_noref; 2880 } 2881 2882 ifr = (struct ifreq *)data; 2883 switch (cmd) { 2884 #ifdef VIMAGE 2885 case SIOCSIFVNET: 2886 error = priv_check(td, PRIV_NET_SETIFVNET); 2887 if (error == 0) 2888 error = if_vmove_loan(td, ifr->ifr_name, ifr->ifr_jid); 2889 goto out_noref; 2890 2891 case SIOCSIFRVNET: 2892 error = priv_check(td, PRIV_NET_SETIFVNET); 2893 if (error == 0) 2894 error = if_vmove_reclaim(td, ifr->ifr_name, 2895 ifr->ifr_jid); 2896 goto out_noref; 2897 #endif 2898 case SIOCIFCREATE: 2899 case SIOCIFCREATE2: 2900 error = priv_check(td, PRIV_NET_IFCREATE); 2901 if (error == 0) 2902 error = if_clone_create(ifr->ifr_name, 2903 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 2904 ifr_data_get_ptr(ifr) : NULL); 2905 goto out_noref; 2906 case SIOCIFDESTROY: 2907 error = priv_check(td, PRIV_NET_IFDESTROY); 2908 2909 if (error == 0) 2910 error = if_clone_destroy(ifr->ifr_name); 2911 goto out_noref; 2912 2913 case SIOCIFGCLONERS: 2914 error = if_clone_list((struct if_clonereq *)data); 2915 goto out_noref; 2916 2917 case SIOCGIFGMEMB: 2918 { 2919 struct ifgroupreq *req; 2920 2921 req = (struct ifgroupreq *)data; 2922 if (strnlen(req->ifgr_name, IFNAMSIZ) == IFNAMSIZ) { 2923 error = EINVAL; 2924 goto out_noref; 2925 } 2926 error = if_getgroupmembers(req); 2927 goto out_noref; 2928 } 2929 } 2930 2931 ifp = ifunit_ref(ifr->ifr_name); 2932 if (ifp == NULL) { 2933 error = ENXIO; 2934 goto out_noref; 2935 } 2936 2937 error = ifhwioctl(cmd, ifp, data, td); 2938 if (error != ENOIOCTL) 2939 goto out_ref; 2940 2941 oif_flags = ifp->if_flags; 2942 if (so->so_proto == NULL) { 2943 error = EOPNOTSUPP; 2944 goto out_ref; 2945 } 2946 2947 /* 2948 * Pass the request on to the socket control method, and if the 2949 * latter returns EOPNOTSUPP, directly to the interface. 2950 * 2951 * Make an exception for the legacy SIOCSIF* requests. Drivers 2952 * trust SIOCSIFADDR et al to come from an already privileged 2953 * layer, and do not perform any credentials checks or input 2954 * validation. 2955 */ 2956 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 2957 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2958 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2959 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2960 error = (*ifp->if_ioctl)(ifp, cmd, data); 2961 2962 if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP)) 2963 if_up(ifp); 2964 out_ref: 2965 if_rele(ifp); 2966 out_noref: 2967 CURVNET_RESTORE(); 2968 #ifdef COMPAT_FREEBSD32 2969 if (error != 0) 2970 return (error); 2971 switch (saved_cmd) { 2972 case SIOCGIFCONF32: 2973 ifc32->ifc_len = thunk.ifc.ifc_len; 2974 break; 2975 case SIOCGDRVSPEC32: 2976 /* 2977 * SIOCGDRVSPEC is IOWR, but nothing actually touches 2978 * the struct so just assert that ifd_len (the only 2979 * field it might make sense to update) hasn't 2980 * changed. 2981 */ 2982 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 2983 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 2984 thunk.ifd.ifd_len)); 2985 break; 2986 case SIOCGIFGROUP32: 2987 case SIOCGIFGMEMB32: 2988 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 2989 break; 2990 case SIOCGIFMEDIA32: 2991 case SIOCGIFXMEDIA32: 2992 ifmr32->ifm_current = thunk.ifmr.ifm_current; 2993 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 2994 ifmr32->ifm_status = thunk.ifmr.ifm_status; 2995 ifmr32->ifm_active = thunk.ifmr.ifm_active; 2996 ifmr32->ifm_count = thunk.ifmr.ifm_count; 2997 break; 2998 } 2999 #endif 3000 return (error); 3001 } 3002 3003 int 3004 if_rename(struct ifnet *ifp, char *new_name) 3005 { 3006 struct ifaddr *ifa; 3007 struct sockaddr_dl *sdl; 3008 size_t namelen, onamelen; 3009 char old_name[IFNAMSIZ]; 3010 char strbuf[IFNAMSIZ + 8]; 3011 3012 if (new_name[0] == '\0') 3013 return (EINVAL); 3014 if (strcmp(new_name, ifp->if_xname) == 0) 3015 return (0); 3016 if (ifunit(new_name) != NULL) 3017 return (EEXIST); 3018 3019 if_printf(ifp, "changing name to '%s'\n", new_name); 3020 3021 IF_ADDR_WLOCK(ifp); 3022 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 3023 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 3024 ifa = ifp->if_addr; 3025 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3026 namelen = strlen(new_name); 3027 onamelen = sdl->sdl_nlen; 3028 /* 3029 * Move the address if needed. This is safe because we 3030 * allocate space for a name of length IFNAMSIZ when we 3031 * create this in if_attach(). 3032 */ 3033 if (namelen != onamelen) { 3034 bcopy(sdl->sdl_data + onamelen, 3035 sdl->sdl_data + namelen, sdl->sdl_alen); 3036 } 3037 bcopy(new_name, sdl->sdl_data, namelen); 3038 sdl->sdl_nlen = namelen; 3039 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 3040 bzero(sdl->sdl_data, onamelen); 3041 while (namelen != 0) 3042 sdl->sdl_data[--namelen] = 0xff; 3043 IF_ADDR_WUNLOCK(ifp); 3044 3045 EVENTHANDLER_INVOKE(ifnet_rename_event, ifp, old_name); 3046 3047 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 3048 devctl_notify("IFNET", old_name, "RENAME", strbuf); 3049 3050 return (0); 3051 } 3052 3053 /* 3054 * The code common to handling reference counted flags, 3055 * e.g., in ifpromisc() and if_allmulti(). 3056 * The "pflag" argument can specify a permanent mode flag to check, 3057 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3058 * 3059 * Only to be used on stack-owned flags, not driver-owned flags. 3060 */ 3061 static int 3062 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3063 { 3064 struct ifreq ifr; 3065 int error; 3066 int oldflags, oldcount; 3067 3068 /* Sanity checks to catch programming errors */ 3069 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3070 ("%s: setting driver-owned flag %d", __func__, flag)); 3071 3072 if (onswitch) 3073 KASSERT(*refcount >= 0, 3074 ("%s: increment negative refcount %d for flag %d", 3075 __func__, *refcount, flag)); 3076 else 3077 KASSERT(*refcount > 0, 3078 ("%s: decrement non-positive refcount %d for flag %d", 3079 __func__, *refcount, flag)); 3080 3081 /* In case this mode is permanent, just touch refcount */ 3082 if (ifp->if_flags & pflag) { 3083 *refcount += onswitch ? 1 : -1; 3084 return (0); 3085 } 3086 3087 /* Save ifnet parameters for if_ioctl() may fail */ 3088 oldcount = *refcount; 3089 oldflags = ifp->if_flags; 3090 3091 /* 3092 * See if we aren't the only and touching refcount is enough. 3093 * Actually toggle interface flag if we are the first or last. 3094 */ 3095 if (onswitch) { 3096 if ((*refcount)++) 3097 return (0); 3098 ifp->if_flags |= flag; 3099 } else { 3100 if (--(*refcount)) 3101 return (0); 3102 ifp->if_flags &= ~flag; 3103 } 3104 3105 /* Call down the driver since we've changed interface flags */ 3106 if (ifp->if_ioctl == NULL) { 3107 error = EOPNOTSUPP; 3108 goto recover; 3109 } 3110 ifr.ifr_flags = ifp->if_flags & 0xffff; 3111 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3112 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3113 if (error) 3114 goto recover; 3115 /* Notify userland that interface flags have changed */ 3116 rt_ifmsg(ifp, flag); 3117 return (0); 3118 3119 recover: 3120 /* Recover after driver error */ 3121 *refcount = oldcount; 3122 ifp->if_flags = oldflags; 3123 return (error); 3124 } 3125 3126 /* 3127 * Set/clear promiscuous mode on interface ifp based on the truth value 3128 * of pswitch. The calls are reference counted so that only the first 3129 * "on" request actually has an effect, as does the final "off" request. 3130 * Results are undefined if the "off" and "on" requests are not matched. 3131 */ 3132 int 3133 ifpromisc(struct ifnet *ifp, int pswitch) 3134 { 3135 int error; 3136 int oldflags = ifp->if_flags; 3137 3138 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3139 &ifp->if_pcount, pswitch); 3140 /* If promiscuous mode status has changed, log a message */ 3141 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3142 log_promisc_mode_change) 3143 if_printf(ifp, "promiscuous mode %s\n", 3144 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3145 return (error); 3146 } 3147 3148 /* 3149 * Return interface configuration 3150 * of system. List may be used 3151 * in later ioctl's (above) to get 3152 * other information. 3153 */ 3154 /*ARGSUSED*/ 3155 static int 3156 ifconf(u_long cmd, caddr_t data) 3157 { 3158 struct ifconf *ifc = (struct ifconf *)data; 3159 struct ifnet *ifp; 3160 struct ifaddr *ifa; 3161 struct ifreq ifr; 3162 struct sbuf *sb; 3163 int error, full = 0, valid_len, max_len; 3164 3165 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3166 max_len = maxphys - 1; 3167 3168 /* Prevent hostile input from being able to crash the system */ 3169 if (ifc->ifc_len <= 0) 3170 return (EINVAL); 3171 3172 again: 3173 if (ifc->ifc_len <= max_len) { 3174 max_len = ifc->ifc_len; 3175 full = 1; 3176 } 3177 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3178 max_len = 0; 3179 valid_len = 0; 3180 3181 IFNET_RLOCK(); 3182 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3183 struct epoch_tracker et; 3184 int addrs; 3185 3186 /* 3187 * Zero the ifr to make sure we don't disclose the contents 3188 * of the stack. 3189 */ 3190 memset(&ifr, 0, sizeof(ifr)); 3191 3192 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3193 >= sizeof(ifr.ifr_name)) { 3194 sbuf_delete(sb); 3195 IFNET_RUNLOCK(); 3196 return (ENAMETOOLONG); 3197 } 3198 3199 addrs = 0; 3200 NET_EPOCH_ENTER(et); 3201 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3202 struct sockaddr *sa = ifa->ifa_addr; 3203 3204 if (prison_if(curthread->td_ucred, sa) != 0) 3205 continue; 3206 addrs++; 3207 if (sa->sa_len <= sizeof(*sa)) { 3208 if (sa->sa_len < sizeof(*sa)) { 3209 memset(&ifr.ifr_ifru.ifru_addr, 0, 3210 sizeof(ifr.ifr_ifru.ifru_addr)); 3211 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3212 sa->sa_len); 3213 } else 3214 ifr.ifr_ifru.ifru_addr = *sa; 3215 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3216 max_len += sizeof(ifr); 3217 } else { 3218 sbuf_bcat(sb, &ifr, 3219 offsetof(struct ifreq, ifr_addr)); 3220 max_len += offsetof(struct ifreq, ifr_addr); 3221 sbuf_bcat(sb, sa, sa->sa_len); 3222 max_len += sa->sa_len; 3223 } 3224 3225 if (sbuf_error(sb) == 0) 3226 valid_len = sbuf_len(sb); 3227 } 3228 NET_EPOCH_EXIT(et); 3229 if (addrs == 0) { 3230 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3231 max_len += sizeof(ifr); 3232 3233 if (sbuf_error(sb) == 0) 3234 valid_len = sbuf_len(sb); 3235 } 3236 } 3237 IFNET_RUNLOCK(); 3238 3239 /* 3240 * If we didn't allocate enough space (uncommon), try again. If 3241 * we have already allocated as much space as we are allowed, 3242 * return what we've got. 3243 */ 3244 if (valid_len != max_len && !full) { 3245 sbuf_delete(sb); 3246 goto again; 3247 } 3248 3249 ifc->ifc_len = valid_len; 3250 sbuf_finish(sb); 3251 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3252 sbuf_delete(sb); 3253 return (error); 3254 } 3255 3256 /* 3257 * Just like ifpromisc(), but for all-multicast-reception mode. 3258 */ 3259 int 3260 if_allmulti(struct ifnet *ifp, int onswitch) 3261 { 3262 3263 return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount, 3264 onswitch)); 3265 } 3266 3267 struct ifmultiaddr * 3268 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3269 { 3270 struct ifmultiaddr *ifma; 3271 3272 IF_ADDR_LOCK_ASSERT(ifp); 3273 3274 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3275 if (sa->sa_family == AF_LINK) { 3276 if (sa_dl_equal(ifma->ifma_addr, sa)) 3277 break; 3278 } else { 3279 if (sa_equal(ifma->ifma_addr, sa)) 3280 break; 3281 } 3282 } 3283 3284 return ifma; 3285 } 3286 3287 /* 3288 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3289 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3290 * the ifnet multicast address list here, so the caller must do that and 3291 * other setup work (such as notifying the device driver). The reference 3292 * count is initialized to 1. 3293 */ 3294 static struct ifmultiaddr * 3295 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3296 int mflags) 3297 { 3298 struct ifmultiaddr *ifma; 3299 struct sockaddr *dupsa; 3300 3301 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3302 M_ZERO); 3303 if (ifma == NULL) 3304 return (NULL); 3305 3306 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3307 if (dupsa == NULL) { 3308 free(ifma, M_IFMADDR); 3309 return (NULL); 3310 } 3311 bcopy(sa, dupsa, sa->sa_len); 3312 ifma->ifma_addr = dupsa; 3313 3314 ifma->ifma_ifp = ifp; 3315 ifma->ifma_refcount = 1; 3316 ifma->ifma_protospec = NULL; 3317 3318 if (llsa == NULL) { 3319 ifma->ifma_lladdr = NULL; 3320 return (ifma); 3321 } 3322 3323 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3324 if (dupsa == NULL) { 3325 free(ifma->ifma_addr, M_IFMADDR); 3326 free(ifma, M_IFMADDR); 3327 return (NULL); 3328 } 3329 bcopy(llsa, dupsa, llsa->sa_len); 3330 ifma->ifma_lladdr = dupsa; 3331 3332 return (ifma); 3333 } 3334 3335 /* 3336 * if_freemulti: free ifmultiaddr structure and possibly attached related 3337 * addresses. The caller is responsible for implementing reference 3338 * counting, notifying the driver, handling routing messages, and releasing 3339 * any dependent link layer state. 3340 */ 3341 #ifdef MCAST_VERBOSE 3342 extern void kdb_backtrace(void); 3343 #endif 3344 static void 3345 if_freemulti_internal(struct ifmultiaddr *ifma) 3346 { 3347 3348 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3349 ifma->ifma_refcount)); 3350 3351 if (ifma->ifma_lladdr != NULL) 3352 free(ifma->ifma_lladdr, M_IFMADDR); 3353 #ifdef MCAST_VERBOSE 3354 kdb_backtrace(); 3355 printf("%s freeing ifma: %p\n", __func__, ifma); 3356 #endif 3357 free(ifma->ifma_addr, M_IFMADDR); 3358 free(ifma, M_IFMADDR); 3359 } 3360 3361 static void 3362 if_destroymulti(epoch_context_t ctx) 3363 { 3364 struct ifmultiaddr *ifma; 3365 3366 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3367 if_freemulti_internal(ifma); 3368 } 3369 3370 void 3371 if_freemulti(struct ifmultiaddr *ifma) 3372 { 3373 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3374 ifma->ifma_refcount)); 3375 3376 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3377 } 3378 3379 /* 3380 * Register an additional multicast address with a network interface. 3381 * 3382 * - If the address is already present, bump the reference count on the 3383 * address and return. 3384 * - If the address is not link-layer, look up a link layer address. 3385 * - Allocate address structures for one or both addresses, and attach to the 3386 * multicast address list on the interface. If automatically adding a link 3387 * layer address, the protocol address will own a reference to the link 3388 * layer address, to be freed when it is freed. 3389 * - Notify the network device driver of an addition to the multicast address 3390 * list. 3391 * 3392 * 'sa' points to caller-owned memory with the desired multicast address. 3393 * 3394 * 'retifma' will be used to return a pointer to the resulting multicast 3395 * address reference, if desired. 3396 */ 3397 int 3398 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3399 struct ifmultiaddr **retifma) 3400 { 3401 struct ifmultiaddr *ifma, *ll_ifma; 3402 struct sockaddr *llsa; 3403 struct sockaddr_dl sdl; 3404 int error; 3405 3406 #ifdef INET 3407 IN_MULTI_LIST_UNLOCK_ASSERT(); 3408 #endif 3409 #ifdef INET6 3410 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3411 #endif 3412 /* 3413 * If the address is already present, return a new reference to it; 3414 * otherwise, allocate storage and set up a new address. 3415 */ 3416 IF_ADDR_WLOCK(ifp); 3417 ifma = if_findmulti(ifp, sa); 3418 if (ifma != NULL) { 3419 ifma->ifma_refcount++; 3420 if (retifma != NULL) 3421 *retifma = ifma; 3422 IF_ADDR_WUNLOCK(ifp); 3423 return (0); 3424 } 3425 3426 /* 3427 * The address isn't already present; resolve the protocol address 3428 * into a link layer address, and then look that up, bump its 3429 * refcount or allocate an ifma for that also. 3430 * Most link layer resolving functions returns address data which 3431 * fits inside default sockaddr_dl structure. However callback 3432 * can allocate another sockaddr structure, in that case we need to 3433 * free it later. 3434 */ 3435 llsa = NULL; 3436 ll_ifma = NULL; 3437 if (ifp->if_resolvemulti != NULL) { 3438 /* Provide called function with buffer size information */ 3439 sdl.sdl_len = sizeof(sdl); 3440 llsa = (struct sockaddr *)&sdl; 3441 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3442 if (error) 3443 goto unlock_out; 3444 } 3445 3446 /* 3447 * Allocate the new address. Don't hook it up yet, as we may also 3448 * need to allocate a link layer multicast address. 3449 */ 3450 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3451 if (ifma == NULL) { 3452 error = ENOMEM; 3453 goto free_llsa_out; 3454 } 3455 3456 /* 3457 * If a link layer address is found, we'll need to see if it's 3458 * already present in the address list, or allocate is as well. 3459 * When this block finishes, the link layer address will be on the 3460 * list. 3461 */ 3462 if (llsa != NULL) { 3463 ll_ifma = if_findmulti(ifp, llsa); 3464 if (ll_ifma == NULL) { 3465 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3466 if (ll_ifma == NULL) { 3467 --ifma->ifma_refcount; 3468 if_freemulti(ifma); 3469 error = ENOMEM; 3470 goto free_llsa_out; 3471 } 3472 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3473 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3474 ifma_link); 3475 } else 3476 ll_ifma->ifma_refcount++; 3477 ifma->ifma_llifma = ll_ifma; 3478 } 3479 3480 /* 3481 * We now have a new multicast address, ifma, and possibly a new or 3482 * referenced link layer address. Add the primary address to the 3483 * ifnet address list. 3484 */ 3485 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3486 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3487 3488 if (retifma != NULL) 3489 *retifma = ifma; 3490 3491 /* 3492 * Must generate the message while holding the lock so that 'ifma' 3493 * pointer is still valid. 3494 */ 3495 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3496 IF_ADDR_WUNLOCK(ifp); 3497 3498 /* 3499 * We are certain we have added something, so call down to the 3500 * interface to let them know about it. 3501 */ 3502 if (ifp->if_ioctl != NULL) { 3503 if (THREAD_CAN_SLEEP()) 3504 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3505 else 3506 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3507 } 3508 3509 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3510 link_free_sdl(llsa); 3511 3512 return (0); 3513 3514 free_llsa_out: 3515 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3516 link_free_sdl(llsa); 3517 3518 unlock_out: 3519 IF_ADDR_WUNLOCK(ifp); 3520 return (error); 3521 } 3522 3523 static void 3524 if_siocaddmulti(void *arg, int pending) 3525 { 3526 struct ifnet *ifp; 3527 3528 ifp = arg; 3529 #ifdef DIAGNOSTIC 3530 if (pending > 1) 3531 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3532 #endif 3533 CURVNET_SET(ifp->if_vnet); 3534 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3535 CURVNET_RESTORE(); 3536 } 3537 3538 /* 3539 * Delete a multicast group membership by network-layer group address. 3540 * 3541 * Returns ENOENT if the entry could not be found. If ifp no longer 3542 * exists, results are undefined. This entry point should only be used 3543 * from subsystems which do appropriate locking to hold ifp for the 3544 * duration of the call. 3545 * Network-layer protocol domains must use if_delmulti_ifma(). 3546 */ 3547 int 3548 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3549 { 3550 struct ifmultiaddr *ifma; 3551 int lastref; 3552 3553 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3554 3555 IF_ADDR_WLOCK(ifp); 3556 lastref = 0; 3557 ifma = if_findmulti(ifp, sa); 3558 if (ifma != NULL) 3559 lastref = if_delmulti_locked(ifp, ifma, 0); 3560 IF_ADDR_WUNLOCK(ifp); 3561 3562 if (ifma == NULL) 3563 return (ENOENT); 3564 3565 if (lastref && ifp->if_ioctl != NULL) { 3566 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3567 } 3568 3569 return (0); 3570 } 3571 3572 /* 3573 * Delete all multicast group membership for an interface. 3574 * Should be used to quickly flush all multicast filters. 3575 */ 3576 void 3577 if_delallmulti(struct ifnet *ifp) 3578 { 3579 struct ifmultiaddr *ifma; 3580 struct ifmultiaddr *next; 3581 3582 IF_ADDR_WLOCK(ifp); 3583 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3584 if_delmulti_locked(ifp, ifma, 0); 3585 IF_ADDR_WUNLOCK(ifp); 3586 } 3587 3588 void 3589 if_delmulti_ifma(struct ifmultiaddr *ifma) 3590 { 3591 if_delmulti_ifma_flags(ifma, 0); 3592 } 3593 3594 /* 3595 * Delete a multicast group membership by group membership pointer. 3596 * Network-layer protocol domains must use this routine. 3597 * 3598 * It is safe to call this routine if the ifp disappeared. 3599 */ 3600 void 3601 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3602 { 3603 struct ifnet *ifp; 3604 int lastref; 3605 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3606 #ifdef INET 3607 IN_MULTI_LIST_UNLOCK_ASSERT(); 3608 #endif 3609 ifp = ifma->ifma_ifp; 3610 #ifdef DIAGNOSTIC 3611 if (ifp == NULL) { 3612 printf("%s: ifma_ifp seems to be detached\n", __func__); 3613 } else { 3614 struct epoch_tracker et; 3615 struct ifnet *oifp; 3616 3617 NET_EPOCH_ENTER(et); 3618 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3619 if (ifp == oifp) 3620 break; 3621 NET_EPOCH_EXIT(et); 3622 if (ifp != oifp) 3623 ifp = NULL; 3624 } 3625 #endif 3626 /* 3627 * If and only if the ifnet instance exists: Acquire the address lock. 3628 */ 3629 if (ifp != NULL) 3630 IF_ADDR_WLOCK(ifp); 3631 3632 lastref = if_delmulti_locked(ifp, ifma, flags); 3633 3634 if (ifp != NULL) { 3635 /* 3636 * If and only if the ifnet instance exists: 3637 * Release the address lock. 3638 * If the group was left: update the hardware hash filter. 3639 */ 3640 IF_ADDR_WUNLOCK(ifp); 3641 if (lastref && ifp->if_ioctl != NULL) { 3642 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3643 } 3644 } 3645 } 3646 3647 /* 3648 * Perform deletion of network-layer and/or link-layer multicast address. 3649 * 3650 * Return 0 if the reference count was decremented. 3651 * Return 1 if the final reference was released, indicating that the 3652 * hardware hash filter should be reprogrammed. 3653 */ 3654 static int 3655 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3656 { 3657 struct ifmultiaddr *ll_ifma; 3658 3659 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3660 KASSERT(ifma->ifma_ifp == ifp, 3661 ("%s: inconsistent ifp %p", __func__, ifp)); 3662 IF_ADDR_WLOCK_ASSERT(ifp); 3663 } 3664 3665 ifp = ifma->ifma_ifp; 3666 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3667 3668 /* 3669 * If the ifnet is detaching, null out references to ifnet, 3670 * so that upper protocol layers will notice, and not attempt 3671 * to obtain locks for an ifnet which no longer exists. The 3672 * routing socket announcement must happen before the ifnet 3673 * instance is detached from the system. 3674 */ 3675 if (detaching) { 3676 #ifdef DIAGNOSTIC 3677 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3678 #endif 3679 /* 3680 * ifp may already be nulled out if we are being reentered 3681 * to delete the ll_ifma. 3682 */ 3683 if (ifp != NULL) { 3684 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3685 ifma->ifma_ifp = NULL; 3686 } 3687 } 3688 3689 if (--ifma->ifma_refcount > 0) 3690 return 0; 3691 3692 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3693 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3694 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3695 } 3696 /* 3697 * If this ifma is a network-layer ifma, a link-layer ifma may 3698 * have been associated with it. Release it first if so. 3699 */ 3700 ll_ifma = ifma->ifma_llifma; 3701 if (ll_ifma != NULL) { 3702 KASSERT(ifma->ifma_lladdr != NULL, 3703 ("%s: llifma w/o lladdr", __func__)); 3704 if (detaching) 3705 ll_ifma->ifma_ifp = NULL; /* XXX */ 3706 if (--ll_ifma->ifma_refcount == 0) { 3707 if (ifp != NULL) { 3708 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3709 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3710 ifma_link); 3711 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3712 } 3713 } 3714 if_freemulti(ll_ifma); 3715 } 3716 } 3717 #ifdef INVARIANTS 3718 if (ifp) { 3719 struct ifmultiaddr *ifmatmp; 3720 3721 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3722 MPASS(ifma != ifmatmp); 3723 } 3724 #endif 3725 if_freemulti(ifma); 3726 /* 3727 * The last reference to this instance of struct ifmultiaddr 3728 * was released; the hardware should be notified of this change. 3729 */ 3730 return 1; 3731 } 3732 3733 /* 3734 * Set the link layer address on an interface. 3735 * 3736 * At this time we only support certain types of interfaces, 3737 * and we don't allow the length of the address to change. 3738 * 3739 * Set noinline to be dtrace-friendly 3740 */ 3741 __noinline int 3742 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3743 { 3744 struct sockaddr_dl *sdl; 3745 struct ifaddr *ifa; 3746 struct ifreq ifr; 3747 3748 ifa = ifp->if_addr; 3749 if (ifa == NULL) 3750 return (EINVAL); 3751 3752 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3753 if (sdl == NULL) 3754 return (EINVAL); 3755 3756 if (len != sdl->sdl_alen) /* don't allow length to change */ 3757 return (EINVAL); 3758 3759 switch (ifp->if_type) { 3760 case IFT_ETHER: 3761 case IFT_XETHER: 3762 case IFT_L2VLAN: 3763 case IFT_BRIDGE: 3764 case IFT_IEEE8023ADLAG: 3765 bcopy(lladdr, LLADDR(sdl), len); 3766 break; 3767 default: 3768 return (ENODEV); 3769 } 3770 3771 /* 3772 * If the interface is already up, we need 3773 * to re-init it in order to reprogram its 3774 * address filter. 3775 */ 3776 if ((ifp->if_flags & IFF_UP) != 0) { 3777 if (ifp->if_ioctl) { 3778 ifp->if_flags &= ~IFF_UP; 3779 ifr.ifr_flags = ifp->if_flags & 0xffff; 3780 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3781 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3782 ifp->if_flags |= IFF_UP; 3783 ifr.ifr_flags = ifp->if_flags & 0xffff; 3784 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3785 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3786 } 3787 } 3788 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3789 3790 return (0); 3791 } 3792 3793 /* 3794 * Compat function for handling basic encapsulation requests. 3795 * Not converted stacks (FDDI, IB, ..) supports traditional 3796 * output model: ARP (and other similar L2 protocols) are handled 3797 * inside output routine, arpresolve/nd6_resolve() returns MAC 3798 * address instead of full prepend. 3799 * 3800 * This function creates calculated header==MAC for IPv4/IPv6 and 3801 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3802 * address families. 3803 */ 3804 static int 3805 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3806 { 3807 if (req->rtype != IFENCAP_LL) 3808 return (EOPNOTSUPP); 3809 3810 if (req->bufsize < req->lladdr_len) 3811 return (ENOMEM); 3812 3813 switch (req->family) { 3814 case AF_INET: 3815 case AF_INET6: 3816 break; 3817 default: 3818 return (EAFNOSUPPORT); 3819 } 3820 3821 /* Copy lladdr to storage as is */ 3822 memmove(req->buf, req->lladdr, req->lladdr_len); 3823 req->bufsize = req->lladdr_len; 3824 req->lladdr_off = 0; 3825 3826 return (0); 3827 } 3828 3829 /* 3830 * Tunnel interfaces can nest, also they may cause infinite recursion 3831 * calls when misconfigured. We'll prevent this by detecting loops. 3832 * High nesting level may cause stack exhaustion. We'll prevent this 3833 * by introducing upper limit. 3834 * 3835 * Return 0, if tunnel nesting count is equal or less than limit. 3836 */ 3837 int 3838 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3839 int limit) 3840 { 3841 struct m_tag *mtag; 3842 int count; 3843 3844 count = 1; 3845 mtag = NULL; 3846 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3847 if (*(struct ifnet **)(mtag + 1) == ifp) { 3848 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3849 return (EIO); 3850 } 3851 count++; 3852 } 3853 if (count > limit) { 3854 log(LOG_NOTICE, 3855 "%s: if_output recursively called too many times(%d)\n", 3856 if_name(ifp), count); 3857 return (EIO); 3858 } 3859 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3860 if (mtag == NULL) 3861 return (ENOMEM); 3862 *(struct ifnet **)(mtag + 1) = ifp; 3863 m_tag_prepend(m, mtag); 3864 return (0); 3865 } 3866 3867 /* 3868 * Get the link layer address that was read from the hardware at attach. 3869 * 3870 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3871 * their component interfaces as IFT_IEEE8023ADLAG. 3872 */ 3873 int 3874 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3875 { 3876 if (ifp->if_hw_addr == NULL) 3877 return (ENODEV); 3878 3879 switch (ifp->if_type) { 3880 case IFT_ETHER: 3881 case IFT_IEEE8023ADLAG: 3882 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3883 return (0); 3884 default: 3885 return (ENODEV); 3886 } 3887 } 3888 3889 /* 3890 * The name argument must be a pointer to storage which will last as 3891 * long as the interface does. For physical devices, the result of 3892 * device_get_name(dev) is a good choice and for pseudo-devices a 3893 * static string works well. 3894 */ 3895 void 3896 if_initname(struct ifnet *ifp, const char *name, int unit) 3897 { 3898 ifp->if_dname = name; 3899 ifp->if_dunit = unit; 3900 if (unit != IF_DUNIT_NONE) 3901 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3902 else 3903 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3904 } 3905 3906 static int 3907 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 3908 { 3909 char if_fmt[256]; 3910 3911 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 3912 vlog(pri, if_fmt, ap); 3913 return (0); 3914 } 3915 3916 3917 int 3918 if_printf(struct ifnet *ifp, const char *fmt, ...) 3919 { 3920 va_list ap; 3921 3922 va_start(ap, fmt); 3923 if_vlog(ifp, LOG_INFO, fmt, ap); 3924 va_end(ap); 3925 return (0); 3926 } 3927 3928 int 3929 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 3930 { 3931 va_list ap; 3932 3933 va_start(ap, fmt); 3934 if_vlog(ifp, pri, fmt, ap); 3935 va_end(ap); 3936 return (0); 3937 } 3938 3939 void 3940 if_start(struct ifnet *ifp) 3941 { 3942 3943 (*(ifp)->if_start)(ifp); 3944 } 3945 3946 /* 3947 * Backwards compatibility interface for drivers 3948 * that have not implemented it 3949 */ 3950 static int 3951 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 3952 { 3953 int error; 3954 3955 IFQ_HANDOFF(ifp, m, error); 3956 return (error); 3957 } 3958 3959 static void 3960 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3961 { 3962 m_freem(m); 3963 } 3964 3965 int 3966 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3967 { 3968 int active = 0; 3969 3970 IF_LOCK(ifq); 3971 if (_IF_QFULL(ifq)) { 3972 IF_UNLOCK(ifq); 3973 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 3974 m_freem(m); 3975 return (0); 3976 } 3977 if (ifp != NULL) { 3978 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 3979 if (m->m_flags & (M_BCAST|M_MCAST)) 3980 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 3981 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3982 } 3983 _IF_ENQUEUE(ifq, m); 3984 IF_UNLOCK(ifq); 3985 if (ifp != NULL && !active) 3986 (*(ifp)->if_start)(ifp); 3987 return (1); 3988 } 3989 3990 void 3991 if_register_com_alloc(u_char type, 3992 if_com_alloc_t *a, if_com_free_t *f) 3993 { 3994 3995 KASSERT(if_com_alloc[type] == NULL, 3996 ("if_register_com_alloc: %d already registered", type)); 3997 KASSERT(if_com_free[type] == NULL, 3998 ("if_register_com_alloc: %d free already registered", type)); 3999 4000 if_com_alloc[type] = a; 4001 if_com_free[type] = f; 4002 } 4003 4004 void 4005 if_deregister_com_alloc(u_char type) 4006 { 4007 4008 KASSERT(if_com_alloc[type] != NULL, 4009 ("if_deregister_com_alloc: %d not registered", type)); 4010 KASSERT(if_com_free[type] != NULL, 4011 ("if_deregister_com_alloc: %d free not registered", type)); 4012 4013 /* 4014 * Ensure all pending EPOCH(9) callbacks have been executed. This 4015 * fixes issues about late invocation of if_destroy(), which leads 4016 * to memory leak from if_com_alloc[type] allocated if_l2com. 4017 */ 4018 NET_EPOCH_DRAIN_CALLBACKS(); 4019 4020 if_com_alloc[type] = NULL; 4021 if_com_free[type] = NULL; 4022 } 4023 4024 /* API for driver access to network stack owned ifnet.*/ 4025 uint64_t 4026 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4027 { 4028 uint64_t oldbrate; 4029 4030 oldbrate = ifp->if_baudrate; 4031 ifp->if_baudrate = baudrate; 4032 return (oldbrate); 4033 } 4034 4035 uint64_t 4036 if_getbaudrate(const if_t ifp) 4037 { 4038 return (ifp->if_baudrate); 4039 } 4040 4041 int 4042 if_setcapabilities(if_t ifp, int capabilities) 4043 { 4044 ifp->if_capabilities = capabilities; 4045 return (0); 4046 } 4047 4048 int 4049 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4050 { 4051 ifp->if_capabilities &= ~clearbit; 4052 ifp->if_capabilities |= setbit; 4053 return (0); 4054 } 4055 4056 int 4057 if_getcapabilities(const if_t ifp) 4058 { 4059 return (ifp->if_capabilities); 4060 } 4061 4062 int 4063 if_setcapenable(if_t ifp, int capabilities) 4064 { 4065 ifp->if_capenable = capabilities; 4066 return (0); 4067 } 4068 4069 int 4070 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4071 { 4072 ifp->if_capenable &= ~clearcap; 4073 ifp->if_capenable |= setcap; 4074 return (0); 4075 } 4076 4077 int 4078 if_setcapabilities2(if_t ifp, int capabilities) 4079 { 4080 ifp->if_capabilities2 = capabilities; 4081 return (0); 4082 } 4083 4084 int 4085 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit) 4086 { 4087 ifp->if_capabilities2 &= ~clearbit; 4088 ifp->if_capabilities2 |= setbit; 4089 return (0); 4090 } 4091 4092 int 4093 if_getcapabilities2(const if_t ifp) 4094 { 4095 return (ifp->if_capabilities2); 4096 } 4097 4098 int 4099 if_setcapenable2(if_t ifp, int capabilities2) 4100 { 4101 ifp->if_capenable2 = capabilities2; 4102 return (0); 4103 } 4104 4105 int 4106 if_setcapenable2bit(if_t ifp, int setcap, int clearcap) 4107 { 4108 ifp->if_capenable2 &= ~clearcap; 4109 ifp->if_capenable2 |= setcap; 4110 return (0); 4111 } 4112 4113 const char * 4114 if_getdname(const if_t ifp) 4115 { 4116 return (ifp->if_dname); 4117 } 4118 4119 void 4120 if_setdname(if_t ifp, const char *dname) 4121 { 4122 ifp->if_dname = dname; 4123 } 4124 4125 const char * 4126 if_name(if_t ifp) 4127 { 4128 return (ifp->if_xname); 4129 } 4130 4131 int 4132 if_setname(if_t ifp, const char *name) 4133 { 4134 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4135 return (ENAMETOOLONG); 4136 strcpy(ifp->if_xname, name); 4137 4138 return (0); 4139 } 4140 4141 int 4142 if_togglecapenable(if_t ifp, int togglecap) 4143 { 4144 ifp->if_capenable ^= togglecap; 4145 return (0); 4146 } 4147 4148 int 4149 if_getcapenable(const if_t ifp) 4150 { 4151 return (ifp->if_capenable); 4152 } 4153 4154 int 4155 if_togglecapenable2(if_t ifp, int togglecap) 4156 { 4157 ifp->if_capenable2 ^= togglecap; 4158 return (0); 4159 } 4160 4161 int 4162 if_getcapenable2(const if_t ifp) 4163 { 4164 return (ifp->if_capenable2); 4165 } 4166 4167 int 4168 if_getdunit(const if_t ifp) 4169 { 4170 return (ifp->if_dunit); 4171 } 4172 4173 int 4174 if_getindex(const if_t ifp) 4175 { 4176 return (ifp->if_index); 4177 } 4178 4179 int 4180 if_getidxgen(const if_t ifp) 4181 { 4182 return (ifp->if_idxgen); 4183 } 4184 4185 const char * 4186 if_getdescr(if_t ifp) 4187 { 4188 return (ifp->if_description); 4189 } 4190 4191 void 4192 if_setdescr(if_t ifp, char *descrbuf) 4193 { 4194 sx_xlock(&ifdescr_sx); 4195 char *odescrbuf = ifp->if_description; 4196 ifp->if_description = descrbuf; 4197 sx_xunlock(&ifdescr_sx); 4198 4199 if_freedescr(odescrbuf); 4200 } 4201 4202 char * 4203 if_allocdescr(size_t sz, int malloc_flag) 4204 { 4205 malloc_flag &= (M_WAITOK | M_NOWAIT); 4206 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4207 } 4208 4209 void 4210 if_freedescr(char *descrbuf) 4211 { 4212 free(descrbuf, M_IFDESCR); 4213 } 4214 4215 int 4216 if_getalloctype(const if_t ifp) 4217 { 4218 return (ifp->if_alloctype); 4219 } 4220 4221 void 4222 if_setlastchange(if_t ifp) 4223 { 4224 getmicrotime(&ifp->if_lastchange); 4225 } 4226 4227 /* 4228 * This is largely undesirable because it ties ifnet to a device, but does 4229 * provide flexiblity for an embedded product vendor. Should be used with 4230 * the understanding that it violates the interface boundaries, and should be 4231 * a last resort only. 4232 */ 4233 int 4234 if_setdev(if_t ifp, void *dev) 4235 { 4236 return (0); 4237 } 4238 4239 int 4240 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4241 { 4242 ifp->if_drv_flags &= ~clear_flags; 4243 ifp->if_drv_flags |= set_flags; 4244 4245 return (0); 4246 } 4247 4248 int 4249 if_getdrvflags(const if_t ifp) 4250 { 4251 return (ifp->if_drv_flags); 4252 } 4253 4254 int 4255 if_setdrvflags(if_t ifp, int flags) 4256 { 4257 ifp->if_drv_flags = flags; 4258 return (0); 4259 } 4260 4261 int 4262 if_setflags(if_t ifp, int flags) 4263 { 4264 ifp->if_flags = flags; 4265 return (0); 4266 } 4267 4268 int 4269 if_setflagbits(if_t ifp, int set, int clear) 4270 { 4271 ifp->if_flags &= ~clear; 4272 ifp->if_flags |= set; 4273 return (0); 4274 } 4275 4276 int 4277 if_getflags(const if_t ifp) 4278 { 4279 return (ifp->if_flags); 4280 } 4281 4282 int 4283 if_clearhwassist(if_t ifp) 4284 { 4285 ifp->if_hwassist = 0; 4286 return (0); 4287 } 4288 4289 int 4290 if_sethwassistbits(if_t ifp, int toset, int toclear) 4291 { 4292 ifp->if_hwassist &= ~toclear; 4293 ifp->if_hwassist |= toset; 4294 4295 return (0); 4296 } 4297 4298 int 4299 if_sethwassist(if_t ifp, int hwassist_bit) 4300 { 4301 ifp->if_hwassist = hwassist_bit; 4302 return (0); 4303 } 4304 4305 int 4306 if_gethwassist(const if_t ifp) 4307 { 4308 return (ifp->if_hwassist); 4309 } 4310 4311 int 4312 if_togglehwassist(if_t ifp, int toggle_bits) 4313 { 4314 ifp->if_hwassist ^= toggle_bits; 4315 return (0); 4316 } 4317 4318 int 4319 if_setmtu(if_t ifp, int mtu) 4320 { 4321 ifp->if_mtu = mtu; 4322 return (0); 4323 } 4324 4325 void 4326 if_notifymtu(if_t ifp) 4327 { 4328 #ifdef INET6 4329 nd6_setmtu(ifp); 4330 #endif 4331 rt_updatemtu(ifp); 4332 } 4333 4334 int 4335 if_getmtu(const if_t ifp) 4336 { 4337 return (ifp->if_mtu); 4338 } 4339 4340 void 4341 if_setppromisc(if_t ifp, bool ppromisc) 4342 { 4343 int new_flags; 4344 4345 if (ppromisc) 4346 new_flags = ifp->if_flags | IFF_PPROMISC; 4347 else 4348 new_flags = ifp->if_flags & ~IFF_PPROMISC; 4349 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 4350 if (new_flags & IFF_PPROMISC) 4351 new_flags |= IFF_PROMISC; 4352 /* 4353 * Only unset IFF_PROMISC if there are no more consumers of 4354 * promiscuity, i.e. the ifp->if_pcount refcount is 0. 4355 */ 4356 else if (ifp->if_pcount == 0) 4357 new_flags &= ~IFF_PROMISC; 4358 if (log_promisc_mode_change) 4359 if_printf(ifp, "permanently promiscuous mode %s\n", 4360 ((new_flags & IFF_PPROMISC) ? 4361 "enabled" : "disabled")); 4362 } 4363 ifp->if_flags = new_flags; 4364 } 4365 4366 /* 4367 * Methods for drivers to access interface unicast and multicast 4368 * link level addresses. Driver shall not know 'struct ifaddr' neither 4369 * 'struct ifmultiaddr'. 4370 */ 4371 u_int 4372 if_lladdr_count(if_t ifp) 4373 { 4374 struct epoch_tracker et; 4375 struct ifaddr *ifa; 4376 u_int count; 4377 4378 count = 0; 4379 NET_EPOCH_ENTER(et); 4380 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4381 if (ifa->ifa_addr->sa_family == AF_LINK) 4382 count++; 4383 NET_EPOCH_EXIT(et); 4384 4385 return (count); 4386 } 4387 4388 int 4389 if_foreach(if_foreach_cb_t cb, void *cb_arg) 4390 { 4391 if_t ifp; 4392 int error; 4393 4394 NET_EPOCH_ASSERT(); 4395 MPASS(cb); 4396 4397 error = 0; 4398 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4399 error = cb(ifp, cb_arg); 4400 if (error != 0) 4401 break; 4402 } 4403 4404 return (error); 4405 } 4406 4407 /* 4408 * Iterates over the list of interfaces, permitting callback function @cb to sleep. 4409 * Stops iteration if @cb returns non-zero error code. 4410 * Returns the last error code from @cb. 4411 * @match_cb: optional match callback limiting the iteration to only matched interfaces 4412 * @match_arg: argument to pass to @match_cb 4413 * @cb: iteration callback 4414 * @cb_arg: argument to pass to @cb 4415 */ 4416 int 4417 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb, 4418 void *cb_arg) 4419 { 4420 int match_count = 0, array_size = 16; /* 128 bytes for malloc */ 4421 struct ifnet **match_array = NULL; 4422 int error = 0; 4423 4424 MPASS(cb); 4425 4426 while (true) { 4427 struct ifnet **new_array; 4428 int new_size = array_size; 4429 struct epoch_tracker et; 4430 struct ifnet *ifp; 4431 4432 while (new_size < match_count) 4433 new_size *= 2; 4434 new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK); 4435 if (match_array != NULL) 4436 memcpy(new_array, match_array, array_size * sizeof(void *)); 4437 free(match_array, M_TEMP); 4438 match_array = new_array; 4439 array_size = new_size; 4440 4441 match_count = 0; 4442 NET_EPOCH_ENTER(et); 4443 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4444 if (match_cb != NULL && !match_cb(ifp, match_arg)) 4445 continue; 4446 if (match_count < array_size) { 4447 if (if_try_ref(ifp)) 4448 match_array[match_count++] = ifp; 4449 } else 4450 match_count++; 4451 } 4452 NET_EPOCH_EXIT(et); 4453 4454 if (match_count > array_size) { 4455 for (int i = 0; i < array_size; i++) 4456 if_rele(match_array[i]); 4457 continue; 4458 } else { 4459 for (int i = 0; i < match_count; i++) { 4460 if (error == 0) 4461 error = cb(match_array[i], cb_arg); 4462 if_rele(match_array[i]); 4463 } 4464 free(match_array, M_TEMP); 4465 break; 4466 } 4467 } 4468 4469 return (error); 4470 } 4471 4472 4473 /* 4474 * Uses just 1 pointer of the 4 available in the public struct. 4475 */ 4476 if_t 4477 if_iter_start(struct if_iter *iter) 4478 { 4479 if_t ifp; 4480 4481 NET_EPOCH_ASSERT(); 4482 4483 bzero(iter, sizeof(*iter)); 4484 ifp = CK_STAILQ_FIRST(&V_ifnet); 4485 if (ifp != NULL) 4486 iter->context[0] = CK_STAILQ_NEXT(ifp, if_link); 4487 else 4488 iter->context[0] = NULL; 4489 return (ifp); 4490 } 4491 4492 if_t 4493 if_iter_next(struct if_iter *iter) 4494 { 4495 if_t cur_ifp = iter->context[0]; 4496 4497 if (cur_ifp != NULL) 4498 iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link); 4499 return (cur_ifp); 4500 } 4501 4502 void 4503 if_iter_finish(struct if_iter *iter) 4504 { 4505 /* Nothing to do here for now. */ 4506 } 4507 4508 u_int 4509 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4510 { 4511 struct epoch_tracker et; 4512 struct ifaddr *ifa; 4513 u_int count; 4514 4515 MPASS(cb); 4516 4517 count = 0; 4518 NET_EPOCH_ENTER(et); 4519 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4520 if (ifa->ifa_addr->sa_family != AF_LINK) 4521 continue; 4522 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4523 count); 4524 } 4525 NET_EPOCH_EXIT(et); 4526 4527 return (count); 4528 } 4529 4530 u_int 4531 if_llmaddr_count(if_t ifp) 4532 { 4533 struct epoch_tracker et; 4534 struct ifmultiaddr *ifma; 4535 int count; 4536 4537 count = 0; 4538 NET_EPOCH_ENTER(et); 4539 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4540 if (ifma->ifma_addr->sa_family == AF_LINK) 4541 count++; 4542 NET_EPOCH_EXIT(et); 4543 4544 return (count); 4545 } 4546 4547 bool 4548 if_maddr_empty(if_t ifp) 4549 { 4550 4551 return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs)); 4552 } 4553 4554 u_int 4555 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4556 { 4557 struct epoch_tracker et; 4558 struct ifmultiaddr *ifma; 4559 u_int count; 4560 4561 MPASS(cb); 4562 4563 count = 0; 4564 NET_EPOCH_ENTER(et); 4565 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4566 if (ifma->ifma_addr->sa_family != AF_LINK) 4567 continue; 4568 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4569 count); 4570 } 4571 NET_EPOCH_EXIT(et); 4572 4573 return (count); 4574 } 4575 4576 u_int 4577 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4578 { 4579 struct epoch_tracker et; 4580 struct ifaddr *ifa; 4581 u_int count; 4582 4583 MPASS(cb); 4584 4585 count = 0; 4586 NET_EPOCH_ENTER(et); 4587 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4588 if (ifa->ifa_addr->sa_family != type) 4589 continue; 4590 count += (*cb)(cb_arg, ifa, count); 4591 } 4592 NET_EPOCH_EXIT(et); 4593 4594 return (count); 4595 } 4596 4597 struct ifaddr * 4598 ifa_iter_start(if_t ifp, struct ifa_iter *iter) 4599 { 4600 struct ifaddr *ifa; 4601 4602 NET_EPOCH_ASSERT(); 4603 4604 bzero(iter, sizeof(*iter)); 4605 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 4606 if (ifa != NULL) 4607 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4608 else 4609 iter->context[0] = NULL; 4610 return (ifa); 4611 } 4612 4613 struct ifaddr * 4614 ifa_iter_next(struct ifa_iter *iter) 4615 { 4616 struct ifaddr *ifa = iter->context[0]; 4617 4618 if (ifa != NULL) 4619 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4620 return (ifa); 4621 } 4622 4623 void 4624 ifa_iter_finish(struct ifa_iter *iter) 4625 { 4626 /* Nothing to do here for now. */ 4627 } 4628 4629 int 4630 if_setsoftc(if_t ifp, void *softc) 4631 { 4632 ifp->if_softc = softc; 4633 return (0); 4634 } 4635 4636 void * 4637 if_getsoftc(const if_t ifp) 4638 { 4639 return (ifp->if_softc); 4640 } 4641 4642 void 4643 if_setrcvif(struct mbuf *m, if_t ifp) 4644 { 4645 4646 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4647 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4648 } 4649 4650 void 4651 if_setvtag(struct mbuf *m, uint16_t tag) 4652 { 4653 m->m_pkthdr.ether_vtag = tag; 4654 } 4655 4656 uint16_t 4657 if_getvtag(struct mbuf *m) 4658 { 4659 return (m->m_pkthdr.ether_vtag); 4660 } 4661 4662 int 4663 if_sendq_empty(if_t ifp) 4664 { 4665 return (IFQ_DRV_IS_EMPTY(&ifp->if_snd)); 4666 } 4667 4668 struct ifaddr * 4669 if_getifaddr(const if_t ifp) 4670 { 4671 return (ifp->if_addr); 4672 } 4673 4674 int 4675 if_setsendqready(if_t ifp) 4676 { 4677 IFQ_SET_READY(&ifp->if_snd); 4678 return (0); 4679 } 4680 4681 int 4682 if_setsendqlen(if_t ifp, int tx_desc_count) 4683 { 4684 IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count); 4685 ifp->if_snd.ifq_drv_maxlen = tx_desc_count; 4686 return (0); 4687 } 4688 4689 void 4690 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na) 4691 { 4692 ifp->if_netmap = na; 4693 } 4694 4695 struct netmap_adapter * 4696 if_getnetmapadapter(if_t ifp) 4697 { 4698 return (ifp->if_netmap); 4699 } 4700 4701 int 4702 if_vlantrunkinuse(if_t ifp) 4703 { 4704 return (ifp->if_vlantrunk != NULL); 4705 } 4706 4707 void 4708 if_init(if_t ifp, void *ctx) 4709 { 4710 (*ifp->if_init)(ctx); 4711 } 4712 4713 void 4714 if_input(if_t ifp, struct mbuf* sendmp) 4715 { 4716 (*ifp->if_input)(ifp, sendmp); 4717 } 4718 4719 int 4720 if_transmit(if_t ifp, struct mbuf *m) 4721 { 4722 return ((*ifp->if_transmit)(ifp, m)); 4723 } 4724 4725 int 4726 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst) 4727 { 4728 if (ifp->if_resolvemulti == NULL) 4729 return (EOPNOTSUPP); 4730 4731 return (ifp->if_resolvemulti(ifp, srcs, dst)); 4732 } 4733 4734 int 4735 if_ioctl(if_t ifp, u_long cmd, void *data) 4736 { 4737 if (ifp->if_ioctl == NULL) 4738 return (EOPNOTSUPP); 4739 4740 return (ifp->if_ioctl(ifp, cmd, data)); 4741 } 4742 4743 struct mbuf * 4744 if_dequeue(if_t ifp) 4745 { 4746 struct mbuf *m; 4747 4748 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 4749 return (m); 4750 } 4751 4752 int 4753 if_sendq_prepend(if_t ifp, struct mbuf *m) 4754 { 4755 IFQ_DRV_PREPEND(&ifp->if_snd, m); 4756 return (0); 4757 } 4758 4759 int 4760 if_setifheaderlen(if_t ifp, int len) 4761 { 4762 ifp->if_hdrlen = len; 4763 return (0); 4764 } 4765 4766 char * 4767 if_getlladdr(const if_t ifp) 4768 { 4769 return (IF_LLADDR(ifp)); 4770 } 4771 4772 void * 4773 if_gethandle(u_char type) 4774 { 4775 return (if_alloc(type)); 4776 } 4777 4778 void 4779 if_vlancap(if_t ifp) 4780 { 4781 VLAN_CAPABILITIES(ifp); 4782 } 4783 4784 int 4785 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4786 { 4787 ifp->if_hw_tsomax = if_hw_tsomax; 4788 return (0); 4789 } 4790 4791 int 4792 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4793 { 4794 ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4795 return (0); 4796 } 4797 4798 int 4799 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4800 { 4801 ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4802 return (0); 4803 } 4804 4805 u_int 4806 if_gethwtsomax(const if_t ifp) 4807 { 4808 return (ifp->if_hw_tsomax); 4809 } 4810 4811 u_int 4812 if_gethwtsomaxsegcount(const if_t ifp) 4813 { 4814 return (ifp->if_hw_tsomaxsegcount); 4815 } 4816 4817 u_int 4818 if_gethwtsomaxsegsize(const if_t ifp) 4819 { 4820 return (ifp->if_hw_tsomaxsegsize); 4821 } 4822 4823 void 4824 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4825 { 4826 ifp->if_init = init_fn; 4827 } 4828 4829 void 4830 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4831 { 4832 ifp->if_input = input_fn; 4833 } 4834 4835 if_input_fn_t 4836 if_getinputfn(if_t ifp) 4837 { 4838 return (ifp->if_input); 4839 } 4840 4841 void 4842 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4843 { 4844 ifp->if_ioctl = ioctl_fn; 4845 } 4846 4847 void 4848 if_setoutputfn(if_t ifp, if_output_fn_t output_fn) 4849 { 4850 ifp->if_output = output_fn; 4851 } 4852 4853 void 4854 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4855 { 4856 ifp->if_start = start_fn; 4857 } 4858 4859 if_start_fn_t 4860 if_getstartfn(if_t ifp) 4861 { 4862 return (ifp->if_start); 4863 } 4864 4865 void 4866 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4867 { 4868 ifp->if_transmit = start_fn; 4869 } 4870 4871 if_transmit_fn_t 4872 if_gettransmitfn(if_t ifp) 4873 { 4874 return (ifp->if_transmit); 4875 } 4876 4877 void 4878 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4879 { 4880 ifp->if_qflush = flush_fn; 4881 } 4882 4883 void 4884 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 4885 { 4886 ifp->if_snd_tag_alloc = alloc_fn; 4887 } 4888 4889 int 4890 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params, 4891 struct m_snd_tag **mstp) 4892 { 4893 if (ifp->if_snd_tag_alloc == NULL) 4894 return (EOPNOTSUPP); 4895 return (ifp->if_snd_tag_alloc(ifp, params, mstp)); 4896 } 4897 4898 void 4899 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4900 { 4901 ifp->if_get_counter = fn; 4902 } 4903 4904 void 4905 if_setreassignfn(if_t ifp, if_reassign_fn_t fn) 4906 { 4907 ifp->if_reassign = fn; 4908 } 4909 4910 void 4911 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn) 4912 { 4913 ifp->if_ratelimit_query = fn; 4914 } 4915 4916 void 4917 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m) 4918 { 4919 ifp->if_debugnet_methods = m; 4920 } 4921 4922 struct label * 4923 if_getmaclabel(if_t ifp) 4924 { 4925 return (ifp->if_label); 4926 } 4927 4928 void 4929 if_setmaclabel(if_t ifp, struct label *label) 4930 { 4931 ifp->if_label = label; 4932 } 4933 4934 int 4935 if_gettype(if_t ifp) 4936 { 4937 return (ifp->if_type); 4938 } 4939 4940 void * 4941 if_getllsoftc(if_t ifp) 4942 { 4943 return (ifp->if_llsoftc); 4944 } 4945 4946 void 4947 if_setllsoftc(if_t ifp, void *llsoftc) 4948 { 4949 ifp->if_llsoftc = llsoftc; 4950 }; 4951 4952 int 4953 if_getlinkstate(if_t ifp) 4954 { 4955 return (ifp->if_link_state); 4956 } 4957 4958 const uint8_t * 4959 if_getbroadcastaddr(if_t ifp) 4960 { 4961 return (ifp->if_broadcastaddr); 4962 } 4963 4964 void 4965 if_setbroadcastaddr(if_t ifp, const uint8_t *addr) 4966 { 4967 ifp->if_broadcastaddr = addr; 4968 } 4969 4970 int 4971 if_getnumadomain(if_t ifp) 4972 { 4973 return (ifp->if_numa_domain); 4974 } 4975 4976 uint64_t 4977 if_getcounter(if_t ifp, ift_counter counter) 4978 { 4979 return (ifp->if_get_counter(ifp, counter)); 4980 } 4981 4982 bool 4983 if_altq_is_enabled(if_t ifp) 4984 { 4985 return (ALTQ_IS_ENABLED(&ifp->if_snd)); 4986 } 4987 4988 struct vnet * 4989 if_getvnet(if_t ifp) 4990 { 4991 return (ifp->if_vnet); 4992 } 4993 4994 struct in_ifinfo * 4995 if_getinet(if_t ifp) 4996 { 4997 return (ifp->if_inet); 4998 } 4999 5000 struct in6_ifextra * 5001 if_getinet6(if_t ifp) 5002 { 5003 return (ifp->if_inet6); 5004 } 5005 5006 u_int 5007 if_getfib(if_t ifp) 5008 { 5009 return (ifp->if_fib); 5010 } 5011 5012 uint8_t 5013 if_getaddrlen(if_t ifp) 5014 { 5015 return (ifp->if_addrlen); 5016 } 5017 5018 struct bpf_if * 5019 if_getbpf(if_t ifp) 5020 { 5021 return (ifp->if_bpf); 5022 } 5023 5024 struct ifvlantrunk * 5025 if_getvlantrunk(if_t ifp) 5026 { 5027 return (ifp->if_vlantrunk); 5028 } 5029 5030 uint8_t 5031 if_getpcp(if_t ifp) 5032 { 5033 return (ifp->if_pcp); 5034 } 5035 5036 void * 5037 if_getl2com(if_t ifp) 5038 { 5039 return (ifp->if_l2com); 5040 } 5041 5042 void 5043 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m) 5044 { 5045 ifp->if_ipsec_accel_m = m; 5046 } 5047 5048 #ifdef DDB 5049 static void 5050 if_show_ifnet(struct ifnet *ifp) 5051 { 5052 if (ifp == NULL) 5053 return; 5054 db_printf("%s:\n", ifp->if_xname); 5055 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 5056 IF_DB_PRINTF("%s", if_dname); 5057 IF_DB_PRINTF("%d", if_dunit); 5058 IF_DB_PRINTF("%s", if_description); 5059 IF_DB_PRINTF("%u", if_index); 5060 IF_DB_PRINTF("%d", if_idxgen); 5061 IF_DB_PRINTF("%u", if_refcount); 5062 IF_DB_PRINTF("%p", if_softc); 5063 IF_DB_PRINTF("%p", if_l2com); 5064 IF_DB_PRINTF("%p", if_llsoftc); 5065 IF_DB_PRINTF("%d", if_amcount); 5066 IF_DB_PRINTF("%p", if_addr); 5067 IF_DB_PRINTF("%p", if_broadcastaddr); 5068 IF_DB_PRINTF("%u", if_fib); 5069 IF_DB_PRINTF("%p", if_vnet); 5070 IF_DB_PRINTF("%p", if_home_vnet); 5071 IF_DB_PRINTF("%p", if_vlantrunk); 5072 IF_DB_PRINTF("%p", if_bpf); 5073 IF_DB_PRINTF("%u", if_pcount); 5074 IF_DB_PRINTF("%p", if_bridge); 5075 IF_DB_PRINTF("%p", if_lagg); 5076 IF_DB_PRINTF("%p", if_pf_kif); 5077 IF_DB_PRINTF("%p", if_carp); 5078 IF_DB_PRINTF("%p", if_label); 5079 IF_DB_PRINTF("%p", if_netmap); 5080 IF_DB_PRINTF("0x%08x", if_flags); 5081 IF_DB_PRINTF("0x%08x", if_drv_flags); 5082 IF_DB_PRINTF("0x%08x", if_capabilities); 5083 IF_DB_PRINTF("0x%08x", if_capenable); 5084 IF_DB_PRINTF("%p", if_snd.ifq_head); 5085 IF_DB_PRINTF("%p", if_snd.ifq_tail); 5086 IF_DB_PRINTF("%d", if_snd.ifq_len); 5087 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 5088 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 5089 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 5090 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 5091 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 5092 IF_DB_PRINTF("%d", if_snd.altq_type); 5093 IF_DB_PRINTF("%x", if_snd.altq_flags); 5094 #undef IF_DB_PRINTF 5095 } 5096 5097 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 5098 { 5099 if (!have_addr) { 5100 db_printf("usage: show ifnet <struct ifnet *>\n"); 5101 return; 5102 } 5103 5104 if_show_ifnet((struct ifnet *)addr); 5105 } 5106 5107 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 5108 { 5109 struct ifnet *ifp; 5110 u_short idx; 5111 5112 for (idx = 1; idx <= if_index; idx++) { 5113 ifp = ifindex_table[idx].ife_ifnet; 5114 if (ifp == NULL) 5115 continue; 5116 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 5117 if (db_pager_quit) 5118 break; 5119 } 5120 } 5121 #endif /* DDB */ 5122