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 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 451 IFNET_WLOCK(); 452 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 453 if (iter == ifp) { 454 MPASS(refcount_load(&ifp->if_refcount) > 0); 455 MPASS(ifp->if_vnet == curvnet); 456 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 457 458 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 459 #ifdef VIMAGE 460 curvnet->vnet_ifcnt--; 461 #endif 462 if (!vmove) 463 ifp->if_flags |= IFF_DYING; 464 found = 1; 465 break; 466 } 467 IFNET_WUNLOCK(); 468 469 return (found); 470 } 471 472 #ifdef VIMAGE 473 static void 474 vnet_if_return(const void *unused __unused) 475 { 476 struct ifnet *ifp, *nifp; 477 struct ifnet **pending; 478 int found __diagused; 479 int i; 480 481 i = 0; 482 483 /* The lock has already been aquired in vnet_destroy() */ 484 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 485 /* 486 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 487 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 488 * if_detach_internal(), which waits for NET_EPOCH callbacks to 489 * complete. We can't do that from within NET_EPOCH. 490 * 491 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 492 * read/write lock. We cannot hold the lock as we call if_vmove() 493 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 494 * ctx lock. 495 */ 496 IFNET_WLOCK(); 497 498 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 499 M_IFNET, M_WAITOK | M_ZERO); 500 501 /* Return all inherited interfaces to their parent vnets. */ 502 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 503 if (ifp->if_home_vnet != ifp->if_vnet) { 504 found = if_unlink_ifnet(ifp, true); 505 MPASS(found); 506 507 pending[i++] = ifp; 508 } 509 } 510 IFNET_WUNLOCK(); 511 512 for (int j = 0; j < i; j++) { 513 if_vmove(pending[j], pending[j]->if_home_vnet); 514 } 515 516 free(pending, M_IFNET); 517 } 518 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 519 vnet_if_return, NULL); 520 #endif 521 522 /* 523 * Allocate a struct ifnet and an index for an interface. A layer 2 524 * common structure will also be allocated if an allocation routine is 525 * registered for the passed type. 526 */ 527 static struct ifnet * 528 if_alloc_domain(u_char type, int numa_domain) 529 { 530 struct ifnet *ifp; 531 u_short idx; 532 533 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 534 if (numa_domain == IF_NODOM) 535 ifp = malloc(sizeof(struct ifnet), M_IFNET, 536 M_WAITOK | M_ZERO); 537 else 538 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 539 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 540 ifp->if_type = type; 541 ifp->if_alloctype = type; 542 ifp->if_numa_domain = numa_domain; 543 #ifdef VIMAGE 544 ifp->if_vnet = curvnet; 545 #endif 546 if (if_com_alloc[type] != NULL) { 547 ifp->if_l2com = if_com_alloc[type](type, ifp); 548 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 549 type)); 550 } 551 552 IF_ADDR_LOCK_INIT(ifp); 553 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 554 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 555 CK_STAILQ_INIT(&ifp->if_addrhead); 556 CK_STAILQ_INIT(&ifp->if_multiaddrs); 557 CK_STAILQ_INIT(&ifp->if_groups); 558 #ifdef MAC 559 mac_ifnet_init(ifp); 560 #endif 561 ifq_init(&ifp->if_snd, ifp); 562 563 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 564 for (int i = 0; i < IFCOUNTERS; i++) 565 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 566 ifp->if_get_counter = if_get_counter_default; 567 ifp->if_pcp = IFNET_PCP_NONE; 568 569 /* Allocate an ifindex array entry. */ 570 IFNET_WLOCK(); 571 /* 572 * Try to find an empty slot below if_index. If we fail, take the 573 * next slot. 574 */ 575 for (idx = 1; idx <= if_index; idx++) { 576 if (ifindex_table[idx].ife_ifnet == NULL) 577 break; 578 } 579 580 /* Catch if_index overflow. */ 581 if (idx >= if_indexlim) { 582 struct ifindex_entry *new, *old; 583 int newlim; 584 585 newlim = if_indexlim * 2; 586 new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO); 587 memcpy(new, ifindex_table, if_indexlim * sizeof(*new)); 588 old = ifindex_table; 589 ck_pr_store_ptr(&ifindex_table, new); 590 if_indexlim = newlim; 591 NET_EPOCH_WAIT(); 592 free(old, M_IFNET); 593 } 594 if (idx > if_index) 595 if_index = idx; 596 597 ifp->if_index = idx; 598 ifp->if_idxgen = ifindex_table[idx].ife_gencnt; 599 ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp); 600 IFNET_WUNLOCK(); 601 602 return (ifp); 603 } 604 605 struct ifnet * 606 if_alloc_dev(u_char type, device_t dev) 607 { 608 int numa_domain; 609 610 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 611 return (if_alloc_domain(type, IF_NODOM)); 612 return (if_alloc_domain(type, numa_domain)); 613 } 614 615 struct ifnet * 616 if_alloc(u_char type) 617 { 618 619 return (if_alloc_domain(type, IF_NODOM)); 620 } 621 /* 622 * Do the actual work of freeing a struct ifnet, and layer 2 common 623 * structure. This call is made when the network epoch guarantees 624 * us that nobody holds a pointer to the interface. 625 */ 626 static void 627 if_free_deferred(epoch_context_t ctx) 628 { 629 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 630 631 KASSERT((ifp->if_flags & IFF_DYING), 632 ("%s: interface not dying", __func__)); 633 634 if (if_com_free[ifp->if_alloctype] != NULL) 635 if_com_free[ifp->if_alloctype](ifp->if_l2com, 636 ifp->if_alloctype); 637 638 #ifdef MAC 639 mac_ifnet_destroy(ifp); 640 #endif /* MAC */ 641 IF_ADDR_LOCK_DESTROY(ifp); 642 ifq_delete(&ifp->if_snd); 643 644 for (int i = 0; i < IFCOUNTERS; i++) 645 counter_u64_free(ifp->if_counters[i]); 646 647 if_freedescr(ifp->if_description); 648 free(ifp->if_hw_addr, M_IFADDR); 649 free(ifp, M_IFNET); 650 } 651 652 /* 653 * Deregister an interface and free the associated storage. 654 */ 655 void 656 if_free(struct ifnet *ifp) 657 { 658 659 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 660 661 /* 662 * XXXGL: An interface index is really an alias to ifp pointer. 663 * Why would we clear the alias now, and not in the deferred 664 * context? Indeed there is nothing wrong with some network 665 * thread obtaining ifp via ifnet_byindex() inside the network 666 * epoch and then dereferencing ifp while we perform if_free(), 667 * and after if_free() finished, too. 668 * 669 * This early index freeing was important back when ifindex was 670 * virtualized and interface would outlive the vnet. 671 */ 672 IFNET_WLOCK(); 673 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 674 ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL); 675 ifindex_table[ifp->if_index].ife_gencnt++; 676 while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL) 677 if_index--; 678 IFNET_WUNLOCK(); 679 680 if (refcount_release(&ifp->if_refcount)) 681 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 682 } 683 684 /* 685 * Interfaces to keep an ifnet type-stable despite the possibility of the 686 * driver calling if_free(). If there are additional references, we defer 687 * freeing the underlying data structure. 688 */ 689 void 690 if_ref(struct ifnet *ifp) 691 { 692 u_int old __diagused; 693 694 /* We don't assert the ifnet list lock here, but arguably should. */ 695 old = refcount_acquire(&ifp->if_refcount); 696 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 697 } 698 699 bool 700 if_try_ref(struct ifnet *ifp) 701 { 702 NET_EPOCH_ASSERT(); 703 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 704 } 705 706 void 707 if_rele(struct ifnet *ifp) 708 { 709 710 if (!refcount_release(&ifp->if_refcount)) 711 return; 712 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 713 } 714 715 void 716 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 717 { 718 719 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 720 721 if (ifq->ifq_maxlen == 0) 722 ifq->ifq_maxlen = ifqmaxlen; 723 724 ifq->altq_type = 0; 725 ifq->altq_disc = NULL; 726 ifq->altq_flags &= ALTQF_CANTCHANGE; 727 ifq->altq_tbr = NULL; 728 ifq->altq_ifp = ifp; 729 } 730 731 void 732 ifq_delete(struct ifaltq *ifq) 733 { 734 mtx_destroy(&ifq->ifq_mtx); 735 } 736 737 /* 738 * Perform generic interface initialization tasks and attach the interface 739 * to the list of "active" interfaces. If vmove flag is set on entry 740 * to if_attach_internal(), perform only a limited subset of initialization 741 * tasks, given that we are moving from one vnet to another an ifnet which 742 * has already been fully initialized. 743 * 744 * Note that if_detach_internal() removes group membership unconditionally 745 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 746 * Thus, when if_vmove() is applied to a cloned interface, group membership 747 * is lost while a cloned one always joins a group whose name is 748 * ifc->ifc_name. To recover this after if_detach_internal() and 749 * if_attach_internal(), the cloner should be specified to 750 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 751 * attempts to join a group whose name is ifc->ifc_name. 752 * 753 * XXX: 754 * - The decision to return void and thus require this function to 755 * succeed is questionable. 756 * - We should probably do more sanity checking. For instance we don't 757 * do anything to insure if_xname is unique or non-empty. 758 */ 759 void 760 if_attach(struct ifnet *ifp) 761 { 762 763 if_attach_internal(ifp, false); 764 } 765 766 /* 767 * Compute the least common TSO limit. 768 */ 769 void 770 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 771 { 772 /* 773 * 1) If there is no limit currently, take the limit from 774 * the network adapter. 775 * 776 * 2) If the network adapter has a limit below the current 777 * limit, apply it. 778 */ 779 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 780 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 781 pmax->tsomaxbytes = ifp->if_hw_tsomax; 782 } 783 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 784 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 785 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 786 } 787 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 788 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 789 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 790 } 791 } 792 793 /* 794 * Update TSO limit of a network adapter. 795 * 796 * Returns zero if no change. Else non-zero. 797 */ 798 int 799 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 800 { 801 int retval = 0; 802 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 803 ifp->if_hw_tsomax = pmax->tsomaxbytes; 804 retval++; 805 } 806 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 807 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 808 retval++; 809 } 810 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 811 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 812 retval++; 813 } 814 return (retval); 815 } 816 817 static void 818 if_attach_internal(struct ifnet *ifp, bool vmove) 819 { 820 unsigned socksize, ifasize; 821 int namelen, masklen; 822 struct sockaddr_dl *sdl; 823 struct ifaddr *ifa; 824 825 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 826 827 #ifdef VIMAGE 828 CURVNET_ASSERT_SET(); 829 ifp->if_vnet = curvnet; 830 if (ifp->if_home_vnet == NULL) 831 ifp->if_home_vnet = curvnet; 832 #endif 833 834 if_addgroup(ifp, IFG_ALL); 835 836 #ifdef VIMAGE 837 /* Restore group membership for cloned interface. */ 838 if (vmove) 839 if_clone_restoregroup(ifp); 840 #endif 841 842 getmicrotime(&ifp->if_lastchange); 843 ifp->if_epoch = time_uptime; 844 845 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 846 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 847 ("transmit and qflush must both either be set or both be NULL")); 848 if (ifp->if_transmit == NULL) { 849 ifp->if_transmit = if_transmit_default; 850 ifp->if_qflush = if_qflush; 851 } 852 if (ifp->if_input == NULL) 853 ifp->if_input = if_input_default; 854 855 if (ifp->if_requestencap == NULL) 856 ifp->if_requestencap = if_requestencap_default; 857 858 if (!vmove) { 859 #ifdef MAC 860 mac_ifnet_create(ifp); 861 #endif 862 863 /* 864 * Create a Link Level name for this device. 865 */ 866 namelen = strlen(ifp->if_xname); 867 /* 868 * Always save enough space for any possible name so we 869 * can do a rename in place later. 870 */ 871 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 872 socksize = masklen + ifp->if_addrlen; 873 if (socksize < sizeof(*sdl)) 874 socksize = sizeof(*sdl); 875 socksize = roundup2(socksize, sizeof(long)); 876 ifasize = sizeof(*ifa) + 2 * socksize; 877 ifa = ifa_alloc(ifasize, M_WAITOK); 878 sdl = (struct sockaddr_dl *)(ifa + 1); 879 sdl->sdl_len = socksize; 880 sdl->sdl_family = AF_LINK; 881 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 882 sdl->sdl_nlen = namelen; 883 sdl->sdl_index = ifp->if_index; 884 sdl->sdl_type = ifp->if_type; 885 ifp->if_addr = ifa; 886 ifa->ifa_ifp = ifp; 887 ifa->ifa_addr = (struct sockaddr *)sdl; 888 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 889 ifa->ifa_netmask = (struct sockaddr *)sdl; 890 sdl->sdl_len = masklen; 891 while (namelen != 0) 892 sdl->sdl_data[--namelen] = 0xff; 893 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 894 /* Reliably crash if used uninitialized. */ 895 ifp->if_broadcastaddr = NULL; 896 897 if (ifp->if_type == IFT_ETHER) { 898 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 899 M_WAITOK | M_ZERO); 900 } 901 902 #if defined(INET) || defined(INET6) 903 /* Use defaults for TSO, if nothing is set */ 904 if (ifp->if_hw_tsomax == 0 && 905 ifp->if_hw_tsomaxsegcount == 0 && 906 ifp->if_hw_tsomaxsegsize == 0) { 907 /* 908 * The TSO defaults needs to be such that an 909 * NFS mbuf list of 35 mbufs totalling just 910 * below 64K works and that a chain of mbufs 911 * can be defragged into at most 32 segments: 912 */ 913 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 914 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 915 ifp->if_hw_tsomaxsegcount = 35; 916 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 917 918 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 919 if (ifp->if_capabilities & IFCAP_TSO) { 920 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 921 ifp->if_hw_tsomax, 922 ifp->if_hw_tsomaxsegcount, 923 ifp->if_hw_tsomaxsegsize); 924 } 925 } 926 #endif 927 } 928 929 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 930 if_link_ifnet(ifp); 931 EVENTHANDLER_INVOKE(ifnet_attached_event, ifp); 932 if (IS_DEFAULT_VNET(curvnet)) 933 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 934 } 935 936 static void 937 if_epochalloc(void *dummy __unused) 938 { 939 940 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 941 } 942 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 943 944 /* 945 * Remove any unicast or broadcast network addresses from an interface. 946 */ 947 void 948 if_purgeaddrs(struct ifnet *ifp) 949 { 950 struct ifaddr *ifa; 951 952 #ifdef INET6 953 /* 954 * Need to leave multicast addresses of proxy NDP llentries 955 * before in6_purgeifaddr() because the llentries are keys 956 * for in6_multi objects of proxy NDP entries. 957 * in6_purgeifaddr()s clean up llentries including proxy NDPs 958 * then we would lose the keys if they are called earlier. 959 */ 960 in6_purge_proxy_ndp(ifp); 961 #endif 962 while (1) { 963 struct epoch_tracker et; 964 965 NET_EPOCH_ENTER(et); 966 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 967 if (ifa->ifa_addr->sa_family != AF_LINK) 968 break; 969 } 970 NET_EPOCH_EXIT(et); 971 972 if (ifa == NULL) 973 break; 974 #ifdef INET 975 /* XXX: Ugly!! ad hoc just for INET */ 976 if (ifa->ifa_addr->sa_family == AF_INET) { 977 struct ifreq ifr; 978 979 bzero(&ifr, sizeof(ifr)); 980 ifr.ifr_addr = *ifa->ifa_addr; 981 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 982 NULL) == 0) 983 continue; 984 } 985 #endif /* INET */ 986 #ifdef INET6 987 if (ifa->ifa_addr->sa_family == AF_INET6) { 988 in6_purgeifaddr((struct in6_ifaddr *)ifa); 989 /* ifp_addrhead is already updated */ 990 continue; 991 } 992 #endif /* INET6 */ 993 IF_ADDR_WLOCK(ifp); 994 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 995 IF_ADDR_WUNLOCK(ifp); 996 ifa_free(ifa); 997 } 998 } 999 1000 /* 1001 * Remove any multicast network addresses from an interface when an ifnet 1002 * is going away. 1003 */ 1004 static void 1005 if_purgemaddrs(struct ifnet *ifp) 1006 { 1007 struct ifmultiaddr *ifma; 1008 1009 IF_ADDR_WLOCK(ifp); 1010 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1011 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1012 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1013 if_delmulti_locked(ifp, ifma, 1); 1014 } 1015 IF_ADDR_WUNLOCK(ifp); 1016 } 1017 1018 /* 1019 * Detach an interface, removing it from the list of "active" interfaces. 1020 * If vmove flag is set on entry to if_detach_internal(), perform only a 1021 * limited subset of cleanup tasks, given that we are moving an ifnet from 1022 * one vnet to another, where it must be fully operational. 1023 * 1024 * XXXRW: There are some significant questions about event ordering, and 1025 * how to prevent things from starting to use the interface during detach. 1026 */ 1027 void 1028 if_detach(struct ifnet *ifp) 1029 { 1030 bool found; 1031 1032 /* 1033 * The driver private data holds a strong reference to the ifnet, and 1034 * it is actually the "owner", hence this routine shall never fail. 1035 * 1036 * Ideally we can loop retrying when we lose race with other threads 1037 * those run if_unlink_ifnet(). For simplicity, use ifnet_detach_sxlock 1038 * to serialize all the detach / vmove operations. 1039 */ 1040 sx_xlock(&ifnet_detach_sxlock); 1041 CURVNET_SET_QUIET(ifp->if_vnet); 1042 found = if_unlink_ifnet(ifp, false); 1043 if (! found) 1044 panic("%s: interface is not on the active list", 1045 ifp->if_xname); 1046 if_detach_internal(ifp, false); 1047 CURVNET_RESTORE(); 1048 sx_xunlock(&ifnet_detach_sxlock); 1049 } 1050 1051 /* 1052 * The vmove flag, if set, indicates that we are called from a callpath 1053 * that is moving an interface to a different vnet instance. 1054 * 1055 * The shutdown flag, if set, indicates that we are called in the 1056 * process of shutting down a vnet instance. Currently only the 1057 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1058 * on a vnet instance shutdown without this flag being set, e.g., when 1059 * the cloned interfaces are destoyed as first thing of teardown. 1060 */ 1061 static void 1062 if_detach_internal(struct ifnet *ifp, bool vmove) 1063 { 1064 struct ifaddr *ifa; 1065 #ifdef VIMAGE 1066 bool shutdown; 1067 1068 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1069 #endif 1070 1071 sx_assert(&ifnet_detach_sxlock, SX_XLOCKED); 1072 1073 /* 1074 * At this point we know the interface still was on the ifnet list 1075 * and we removed it so we are in a stable state. 1076 */ 1077 NET_EPOCH_WAIT(); 1078 1079 /* 1080 * Ensure all pending EPOCH(9) callbacks have been executed. This 1081 * fixes issues about late destruction of multicast options 1082 * which lead to leave group calls, which in turn access the 1083 * belonging ifnet structure: 1084 */ 1085 NET_EPOCH_DRAIN_CALLBACKS(); 1086 1087 /* 1088 * In any case (destroy or vmove) detach us from the groups 1089 * and remove/wait for pending events on the taskq. 1090 * XXX-BZ in theory an interface could still enqueue a taskq change? 1091 */ 1092 if_delgroups(ifp); 1093 1094 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1095 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1096 1097 if_down(ifp); 1098 1099 #ifdef VIMAGE 1100 /* 1101 * On VNET shutdown abort here as the stack teardown will do all 1102 * the work top-down for us. 1103 */ 1104 if (shutdown) { 1105 /* Give interface users the chance to clean up. */ 1106 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1107 1108 /* 1109 * In case of a vmove we are done here without error. 1110 * If we would signal an error it would lead to the same 1111 * abort as if we did not find the ifnet anymore. 1112 * if_detach() calls us in void context and does not care 1113 * about an early abort notification, so life is splendid :) 1114 */ 1115 return; 1116 } 1117 #endif 1118 1119 /* 1120 * At this point we are not tearing down a VNET and are either 1121 * going to destroy or vmove the interface and have to cleanup 1122 * accordingly. 1123 */ 1124 1125 /* 1126 * Remove routes and flush queues. 1127 */ 1128 #ifdef ALTQ 1129 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1130 altq_disable(&ifp->if_snd); 1131 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1132 altq_detach(&ifp->if_snd); 1133 #endif 1134 1135 rt_flushifroutes(ifp); 1136 1137 if_purgeaddrs(ifp); 1138 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1139 if_purgemaddrs(ifp); 1140 if (IS_DEFAULT_VNET(curvnet)) 1141 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1142 1143 if (!vmove) { 1144 /* 1145 * Prevent further calls into the device driver via ifnet. 1146 */ 1147 if_dead(ifp); 1148 1149 /* 1150 * Clean up all addresses. 1151 */ 1152 IF_ADDR_WLOCK(ifp); 1153 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1154 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1155 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1156 IF_ADDR_WUNLOCK(ifp); 1157 ifa_free(ifa); 1158 } else 1159 IF_ADDR_WUNLOCK(ifp); 1160 } 1161 } 1162 1163 #ifdef VIMAGE 1164 /* 1165 * if_vmove() performs a limited version of if_detach() in current 1166 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1167 */ 1168 static void 1169 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1170 { 1171 #ifdef DEV_BPF 1172 /* 1173 * Detach BPF file descriptors from its interface. 1174 */ 1175 bpf_ifdetach(ifp); 1176 #endif 1177 1178 /* 1179 * Detach from current vnet, but preserve LLADDR info, do not 1180 * mark as dead etc. so that the ifnet can be reattached later. 1181 */ 1182 if_detach_internal(ifp, true); 1183 1184 /* 1185 * Perform interface-specific reassignment tasks, if provided by 1186 * the driver. 1187 */ 1188 if (ifp->if_reassign != NULL) 1189 ifp->if_reassign(ifp, new_vnet, NULL); 1190 1191 /* 1192 * Switch to the context of the target vnet. 1193 */ 1194 CURVNET_SET_QUIET(new_vnet); 1195 if_attach_internal(ifp, true); 1196 bpf_vmove(ifp->if_bpf); 1197 CURVNET_RESTORE(); 1198 } 1199 1200 /* 1201 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1202 */ 1203 static int 1204 if_vmove_loan(struct thread *td, char *ifname, int jid) 1205 { 1206 struct prison *pr; 1207 struct ifnet *ifp, *difp; 1208 bool found; 1209 1210 MPASS(curthread == td); 1211 MPASS(curvnet == TD_TO_VNET(td)); 1212 1213 /* 1214 * We check the existence of the interface, and will later try to 1215 * unlink it from the "active" list, so it is sufficient to only 1216 * hold a weak reference to it. 1217 * Be aware that it is unsafe to access any member of it, until it 1218 * is proven to be safe to ( say it was on the "active" list ). 1219 */ 1220 ifp = ifunit(ifname); 1221 if (ifp == NULL) 1222 return (ENXIO); 1223 1224 /* Try to find the prison within our visibility. */ 1225 sx_slock(&allprison_lock); 1226 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1227 sx_sunlock(&allprison_lock); 1228 if (pr == NULL) 1229 return (ENXIO); 1230 /* Do not try to move the iface from and to the same vnet. */ 1231 if (pr->pr_vnet == TD_TO_VNET(td)) { 1232 mtx_unlock(&pr->pr_mtx); 1233 return (EEXIST); 1234 } 1235 prison_hold_locked(pr); 1236 mtx_unlock(&pr->pr_mtx); 1237 1238 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1239 /* XXX Lock interfaces to avoid races. */ 1240 CURVNET_SET_QUIET(pr->pr_vnet); 1241 difp = ifunit(ifname); 1242 CURVNET_RESTORE(); 1243 if (difp != NULL) { 1244 prison_free(pr); 1245 return (EEXIST); 1246 } 1247 sx_xlock(&ifnet_detach_sxlock); 1248 1249 found = if_unlink_ifnet(ifp, true); 1250 if (! found) { 1251 sx_xunlock(&ifnet_detach_sxlock); 1252 prison_free(pr); 1253 return (ENODEV); 1254 } 1255 1256 /* Move the interface into the child jail/vnet. */ 1257 if_vmove(ifp, pr->pr_vnet); 1258 1259 /* Report the new if_xname back to the userland. */ 1260 sprintf(ifname, "%s", ifp->if_xname); 1261 1262 sx_xunlock(&ifnet_detach_sxlock); 1263 1264 prison_free(pr); 1265 return (0); 1266 } 1267 1268 static int 1269 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1270 { 1271 struct prison *pr; 1272 struct vnet *vnet_dst; 1273 struct ifnet *ifp; 1274 int found; 1275 1276 /* Try to find the prison within our visibility. */ 1277 sx_slock(&allprison_lock); 1278 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1279 sx_sunlock(&allprison_lock); 1280 if (pr == NULL) 1281 return (ENXIO); 1282 prison_hold_locked(pr); 1283 mtx_unlock(&pr->pr_mtx); 1284 1285 /* Make sure the named iface exists in the source prison/vnet. */ 1286 CURVNET_SET(pr->pr_vnet); 1287 ifp = ifunit(ifname); 1288 if (ifp == NULL) { 1289 CURVNET_RESTORE(); 1290 prison_free(pr); 1291 return (ENXIO); 1292 } 1293 1294 /* Do not try to move the iface from and to the same vnet. */ 1295 vnet_dst = TD_TO_VNET(td); 1296 if (vnet_dst == pr->pr_vnet) { 1297 CURVNET_RESTORE(); 1298 prison_free(pr); 1299 return (EEXIST); 1300 } 1301 1302 /* Get interface back from child jail/vnet. */ 1303 sx_xlock(&ifnet_detach_sxlock); 1304 found = if_unlink_ifnet(ifp, true); 1305 if (! found) { 1306 sx_xunlock(&ifnet_detach_sxlock); 1307 CURVNET_RESTORE(); 1308 prison_free(pr); 1309 return (ENODEV); 1310 } 1311 if_vmove(ifp, vnet_dst); 1312 sx_xunlock(&ifnet_detach_sxlock); 1313 CURVNET_RESTORE(); 1314 1315 /* Report the new if_xname back to the userland. */ 1316 sprintf(ifname, "%s", ifp->if_xname); 1317 1318 prison_free(pr); 1319 return (0); 1320 } 1321 #endif /* VIMAGE */ 1322 1323 /* 1324 * Add a group to an interface 1325 */ 1326 int 1327 if_addgroup(struct ifnet *ifp, const char *groupname) 1328 { 1329 struct ifg_list *ifgl; 1330 struct ifg_group *ifg = NULL; 1331 struct ifg_member *ifgm; 1332 int new = 0; 1333 1334 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1335 groupname[strlen(groupname) - 1] <= '9') 1336 return (EINVAL); 1337 1338 IFNET_WLOCK(); 1339 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1340 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1341 IFNET_WUNLOCK(); 1342 return (EEXIST); 1343 } 1344 1345 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1346 IFNET_WUNLOCK(); 1347 return (ENOMEM); 1348 } 1349 1350 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1351 free(ifgl, M_TEMP); 1352 IFNET_WUNLOCK(); 1353 return (ENOMEM); 1354 } 1355 1356 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1357 if (!strcmp(ifg->ifg_group, groupname)) 1358 break; 1359 1360 if (ifg == NULL) { 1361 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1362 free(ifgl, M_TEMP); 1363 free(ifgm, M_TEMP); 1364 IFNET_WUNLOCK(); 1365 return (ENOMEM); 1366 } 1367 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1368 ifg->ifg_refcnt = 0; 1369 CK_STAILQ_INIT(&ifg->ifg_members); 1370 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1371 new = 1; 1372 } 1373 1374 ifg->ifg_refcnt++; 1375 ifgl->ifgl_group = ifg; 1376 ifgm->ifgm_ifp = ifp; 1377 1378 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1379 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1380 IFNET_WUNLOCK(); 1381 1382 if (new) 1383 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1384 EVENTHANDLER_INVOKE(group_change_event, groupname); 1385 1386 return (0); 1387 } 1388 1389 /* 1390 * Helper function to remove a group out of an interface. Expects the global 1391 * ifnet lock to be write-locked, and drops it before returning. 1392 */ 1393 static void 1394 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1395 const char *groupname) 1396 { 1397 struct ifg_member *ifgm; 1398 bool freeifgl; 1399 1400 IFNET_WLOCK_ASSERT(); 1401 1402 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1403 1404 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1405 if (ifgm->ifgm_ifp == ifp) { 1406 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1407 ifg_member, ifgm_next); 1408 break; 1409 } 1410 } 1411 1412 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1413 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1414 ifg_next); 1415 freeifgl = true; 1416 } else { 1417 freeifgl = false; 1418 } 1419 IFNET_WUNLOCK(); 1420 1421 NET_EPOCH_WAIT(); 1422 EVENTHANDLER_INVOKE(group_change_event, groupname); 1423 if (freeifgl) { 1424 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1425 free(ifgl->ifgl_group, M_TEMP); 1426 } 1427 free(ifgm, M_TEMP); 1428 free(ifgl, M_TEMP); 1429 } 1430 1431 /* 1432 * Remove a group from an interface 1433 */ 1434 int 1435 if_delgroup(struct ifnet *ifp, const char *groupname) 1436 { 1437 struct ifg_list *ifgl; 1438 1439 IFNET_WLOCK(); 1440 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1441 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1442 break; 1443 if (ifgl == NULL) { 1444 IFNET_WUNLOCK(); 1445 return (ENOENT); 1446 } 1447 1448 _if_delgroup_locked(ifp, ifgl, groupname); 1449 1450 return (0); 1451 } 1452 1453 /* 1454 * Remove an interface from all groups 1455 */ 1456 static void 1457 if_delgroups(struct ifnet *ifp) 1458 { 1459 struct ifg_list *ifgl; 1460 char groupname[IFNAMSIZ]; 1461 1462 IFNET_WLOCK(); 1463 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1464 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1465 _if_delgroup_locked(ifp, ifgl, groupname); 1466 IFNET_WLOCK(); 1467 } 1468 IFNET_WUNLOCK(); 1469 } 1470 1471 /* 1472 * Stores all groups from an interface in memory pointed to by ifgr. 1473 */ 1474 static int 1475 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1476 { 1477 struct ifg_list *ifgl; 1478 struct ifg_req ifgrq, *ifgp; 1479 int len, error; 1480 1481 IFNET_RLOCK(); 1482 if (ifgr->ifgr_len == 0) { 1483 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1484 ifgr->ifgr_len += sizeof(struct ifg_req); 1485 error = 0; 1486 } else { 1487 len = ifgr->ifgr_len; 1488 ifgp = ifgr->ifgr_groups; 1489 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1490 if (len < sizeof(ifgrq)) { 1491 error = EINVAL; 1492 break; 1493 } 1494 bzero(&ifgrq, sizeof ifgrq); 1495 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1496 sizeof(ifgrq.ifgrq_group)); 1497 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1498 break; 1499 len -= sizeof(ifgrq); 1500 ifgp++; 1501 } 1502 } 1503 IFNET_RUNLOCK(); 1504 1505 return (error); 1506 } 1507 1508 /* 1509 * Stores all members of a group in memory pointed to by igfr 1510 */ 1511 static int 1512 if_getgroupmembers(struct ifgroupreq *ifgr) 1513 { 1514 struct ifg_group *ifg; 1515 struct ifg_member *ifgm; 1516 struct ifg_req ifgrq, *ifgp; 1517 int len, error; 1518 1519 IFNET_RLOCK(); 1520 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1521 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1522 break; 1523 if (ifg == NULL) { 1524 IFNET_RUNLOCK(); 1525 return (ENOENT); 1526 } 1527 1528 if (ifgr->ifgr_len == 0) { 1529 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1530 ifgr->ifgr_len += sizeof(ifgrq); 1531 IFNET_RUNLOCK(); 1532 return (0); 1533 } 1534 1535 len = ifgr->ifgr_len; 1536 ifgp = ifgr->ifgr_groups; 1537 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1538 if (len < sizeof(ifgrq)) { 1539 IFNET_RUNLOCK(); 1540 return (EINVAL); 1541 } 1542 bzero(&ifgrq, sizeof ifgrq); 1543 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1544 sizeof(ifgrq.ifgrq_member)); 1545 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1546 IFNET_RUNLOCK(); 1547 return (error); 1548 } 1549 len -= sizeof(ifgrq); 1550 ifgp++; 1551 } 1552 IFNET_RUNLOCK(); 1553 1554 return (0); 1555 } 1556 1557 /* 1558 * Return counter values from counter(9)s stored in ifnet. 1559 */ 1560 uint64_t 1561 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1562 { 1563 1564 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1565 1566 return (counter_u64_fetch(ifp->if_counters[cnt])); 1567 } 1568 1569 /* 1570 * Increase an ifnet counter. Usually used for counters shared 1571 * between the stack and a driver, but function supports them all. 1572 */ 1573 void 1574 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1575 { 1576 1577 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1578 1579 counter_u64_add(ifp->if_counters[cnt], inc); 1580 } 1581 1582 /* 1583 * Copy data from ifnet to userland API structure if_data. 1584 */ 1585 void 1586 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1587 { 1588 1589 ifd->ifi_type = ifp->if_type; 1590 ifd->ifi_physical = 0; 1591 ifd->ifi_addrlen = ifp->if_addrlen; 1592 ifd->ifi_hdrlen = ifp->if_hdrlen; 1593 ifd->ifi_link_state = ifp->if_link_state; 1594 ifd->ifi_vhid = 0; 1595 ifd->ifi_datalen = sizeof(struct if_data); 1596 ifd->ifi_mtu = ifp->if_mtu; 1597 ifd->ifi_metric = ifp->if_metric; 1598 ifd->ifi_baudrate = ifp->if_baudrate; 1599 ifd->ifi_hwassist = ifp->if_hwassist; 1600 ifd->ifi_epoch = ifp->if_epoch; 1601 ifd->ifi_lastchange = ifp->if_lastchange; 1602 1603 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1604 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1605 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1606 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1607 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1608 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1609 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1610 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1611 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1612 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1613 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1614 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1615 } 1616 1617 /* 1618 * Initialization, destruction and refcounting functions for ifaddrs. 1619 */ 1620 struct ifaddr * 1621 ifa_alloc(size_t size, int flags) 1622 { 1623 struct ifaddr *ifa; 1624 1625 KASSERT(size >= sizeof(struct ifaddr), 1626 ("%s: invalid size %zu", __func__, size)); 1627 1628 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1629 if (ifa == NULL) 1630 return (NULL); 1631 1632 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1633 goto fail; 1634 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1635 goto fail; 1636 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1637 goto fail; 1638 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1639 goto fail; 1640 1641 refcount_init(&ifa->ifa_refcnt, 1); 1642 1643 return (ifa); 1644 1645 fail: 1646 /* free(NULL) is okay */ 1647 counter_u64_free(ifa->ifa_opackets); 1648 counter_u64_free(ifa->ifa_ipackets); 1649 counter_u64_free(ifa->ifa_obytes); 1650 counter_u64_free(ifa->ifa_ibytes); 1651 free(ifa, M_IFADDR); 1652 1653 return (NULL); 1654 } 1655 1656 void 1657 ifa_ref(struct ifaddr *ifa) 1658 { 1659 u_int old __diagused; 1660 1661 old = refcount_acquire(&ifa->ifa_refcnt); 1662 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1663 } 1664 1665 int 1666 ifa_try_ref(struct ifaddr *ifa) 1667 { 1668 1669 NET_EPOCH_ASSERT(); 1670 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1671 } 1672 1673 static void 1674 ifa_destroy(epoch_context_t ctx) 1675 { 1676 struct ifaddr *ifa; 1677 1678 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1679 counter_u64_free(ifa->ifa_opackets); 1680 counter_u64_free(ifa->ifa_ipackets); 1681 counter_u64_free(ifa->ifa_obytes); 1682 counter_u64_free(ifa->ifa_ibytes); 1683 free(ifa, M_IFADDR); 1684 } 1685 1686 void 1687 ifa_free(struct ifaddr *ifa) 1688 { 1689 1690 if (refcount_release(&ifa->ifa_refcnt)) 1691 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1692 } 1693 1694 /* 1695 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1696 * structs used to represent other address families, it is necessary 1697 * to perform a different comparison. 1698 */ 1699 static bool 1700 sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b) 1701 { 1702 const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a; 1703 const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b; 1704 1705 return (sdl1->sdl_len == sdl2->sdl_len && 1706 bcmp(sdl1->sdl_data + sdl1->sdl_nlen, 1707 sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0); 1708 } 1709 1710 /* 1711 * Locate an interface based on a complete address. 1712 */ 1713 /*ARGSUSED*/ 1714 struct ifaddr * 1715 ifa_ifwithaddr(const struct sockaddr *addr) 1716 { 1717 struct ifnet *ifp; 1718 struct ifaddr *ifa; 1719 1720 NET_EPOCH_ASSERT(); 1721 1722 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1723 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1724 if (ifa->ifa_addr->sa_family != addr->sa_family) 1725 continue; 1726 if (sa_equal(addr, ifa->ifa_addr)) { 1727 goto done; 1728 } 1729 /* IP6 doesn't have broadcast */ 1730 if ((ifp->if_flags & IFF_BROADCAST) && 1731 ifa->ifa_broadaddr && 1732 ifa->ifa_broadaddr->sa_len != 0 && 1733 sa_equal(ifa->ifa_broadaddr, addr)) { 1734 goto done; 1735 } 1736 } 1737 } 1738 ifa = NULL; 1739 done: 1740 return (ifa); 1741 } 1742 1743 int 1744 ifa_ifwithaddr_check(const struct sockaddr *addr) 1745 { 1746 struct epoch_tracker et; 1747 int rc; 1748 1749 NET_EPOCH_ENTER(et); 1750 rc = (ifa_ifwithaddr(addr) != NULL); 1751 NET_EPOCH_EXIT(et); 1752 return (rc); 1753 } 1754 1755 /* 1756 * Locate an interface based on the broadcast address. 1757 */ 1758 /* ARGSUSED */ 1759 struct ifaddr * 1760 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1761 { 1762 struct ifnet *ifp; 1763 struct ifaddr *ifa; 1764 1765 NET_EPOCH_ASSERT(); 1766 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1767 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1768 continue; 1769 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1770 if (ifa->ifa_addr->sa_family != addr->sa_family) 1771 continue; 1772 if ((ifp->if_flags & IFF_BROADCAST) && 1773 ifa->ifa_broadaddr && 1774 ifa->ifa_broadaddr->sa_len != 0 && 1775 sa_equal(ifa->ifa_broadaddr, addr)) { 1776 goto done; 1777 } 1778 } 1779 } 1780 ifa = NULL; 1781 done: 1782 return (ifa); 1783 } 1784 1785 /* 1786 * Locate the point to point interface with a given destination address. 1787 */ 1788 /*ARGSUSED*/ 1789 struct ifaddr * 1790 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1791 { 1792 struct ifnet *ifp; 1793 struct ifaddr *ifa; 1794 1795 NET_EPOCH_ASSERT(); 1796 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1797 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1798 continue; 1799 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1800 continue; 1801 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1802 if (ifa->ifa_addr->sa_family != addr->sa_family) 1803 continue; 1804 if (ifa->ifa_dstaddr != NULL && 1805 sa_equal(addr, ifa->ifa_dstaddr)) { 1806 goto done; 1807 } 1808 } 1809 } 1810 ifa = NULL; 1811 done: 1812 return (ifa); 1813 } 1814 1815 /* 1816 * Find an interface on a specific network. If many, choice 1817 * is most specific found. 1818 */ 1819 struct ifaddr * 1820 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 1821 { 1822 struct ifnet *ifp; 1823 struct ifaddr *ifa; 1824 struct ifaddr *ifa_maybe = NULL; 1825 u_int af = addr->sa_family; 1826 const char *addr_data = addr->sa_data, *cplim; 1827 1828 NET_EPOCH_ASSERT(); 1829 /* 1830 * AF_LINK addresses can be looked up directly by their index number, 1831 * so do that if we can. 1832 */ 1833 if (af == AF_LINK) { 1834 ifp = ifnet_byindex( 1835 ((const struct sockaddr_dl *)addr)->sdl_index); 1836 return (ifp ? ifp->if_addr : NULL); 1837 } 1838 1839 /* 1840 * Scan though each interface, looking for ones that have addresses 1841 * in this address family and the requested fib. 1842 */ 1843 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1844 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1845 continue; 1846 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1847 const char *cp, *cp2, *cp3; 1848 1849 if (ifa->ifa_addr->sa_family != af) 1850 next: continue; 1851 if (af == AF_INET && 1852 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 1853 /* 1854 * This is a bit broken as it doesn't 1855 * take into account that the remote end may 1856 * be a single node in the network we are 1857 * looking for. 1858 * The trouble is that we don't know the 1859 * netmask for the remote end. 1860 */ 1861 if (ifa->ifa_dstaddr != NULL && 1862 sa_equal(addr, ifa->ifa_dstaddr)) { 1863 goto done; 1864 } 1865 } else { 1866 /* 1867 * Scan all the bits in the ifa's address. 1868 * If a bit dissagrees with what we are 1869 * looking for, mask it with the netmask 1870 * to see if it really matters. 1871 * (A byte at a time) 1872 */ 1873 if (ifa->ifa_netmask == 0) 1874 continue; 1875 cp = addr_data; 1876 cp2 = ifa->ifa_addr->sa_data; 1877 cp3 = ifa->ifa_netmask->sa_data; 1878 cplim = ifa->ifa_netmask->sa_len 1879 + (char *)ifa->ifa_netmask; 1880 while (cp3 < cplim) 1881 if ((*cp++ ^ *cp2++) & *cp3++) 1882 goto next; /* next address! */ 1883 /* 1884 * If the netmask of what we just found 1885 * is more specific than what we had before 1886 * (if we had one), or if the virtual status 1887 * of new prefix is better than of the old one, 1888 * then remember the new one before continuing 1889 * to search for an even better one. 1890 */ 1891 if (ifa_maybe == NULL || 1892 ifa_preferred(ifa_maybe, ifa) || 1893 rn_refines((caddr_t)ifa->ifa_netmask, 1894 (caddr_t)ifa_maybe->ifa_netmask)) { 1895 ifa_maybe = ifa; 1896 } 1897 } 1898 } 1899 } 1900 ifa = ifa_maybe; 1901 ifa_maybe = NULL; 1902 done: 1903 return (ifa); 1904 } 1905 1906 /* 1907 * Find an interface address specific to an interface best matching 1908 * a given address. 1909 */ 1910 struct ifaddr * 1911 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 1912 { 1913 struct ifaddr *ifa; 1914 const char *cp, *cp2, *cp3; 1915 char *cplim; 1916 struct ifaddr *ifa_maybe = NULL; 1917 u_int af = addr->sa_family; 1918 1919 if (af >= AF_MAX) 1920 return (NULL); 1921 1922 NET_EPOCH_ASSERT(); 1923 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1924 if (ifa->ifa_addr->sa_family != af) 1925 continue; 1926 if (ifa_maybe == NULL) 1927 ifa_maybe = ifa; 1928 if (ifa->ifa_netmask == 0) { 1929 if (sa_equal(addr, ifa->ifa_addr) || 1930 (ifa->ifa_dstaddr && 1931 sa_equal(addr, ifa->ifa_dstaddr))) 1932 goto done; 1933 continue; 1934 } 1935 if (ifp->if_flags & IFF_POINTOPOINT) { 1936 if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr)) 1937 goto done; 1938 } else { 1939 cp = addr->sa_data; 1940 cp2 = ifa->ifa_addr->sa_data; 1941 cp3 = ifa->ifa_netmask->sa_data; 1942 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 1943 for (; cp3 < cplim; cp3++) 1944 if ((*cp++ ^ *cp2++) & *cp3) 1945 break; 1946 if (cp3 == cplim) 1947 goto done; 1948 } 1949 } 1950 ifa = ifa_maybe; 1951 done: 1952 return (ifa); 1953 } 1954 1955 /* 1956 * See whether new ifa is better than current one: 1957 * 1) A non-virtual one is preferred over virtual. 1958 * 2) A virtual in master state preferred over any other state. 1959 * 1960 * Used in several address selecting functions. 1961 */ 1962 int 1963 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 1964 { 1965 1966 return (cur->ifa_carp && (!next->ifa_carp || 1967 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 1968 } 1969 1970 struct sockaddr_dl * 1971 link_alloc_sdl(size_t size, int flags) 1972 { 1973 1974 return (malloc(size, M_TEMP, flags)); 1975 } 1976 1977 void 1978 link_free_sdl(struct sockaddr *sa) 1979 { 1980 free(sa, M_TEMP); 1981 } 1982 1983 /* 1984 * Fills in given sdl with interface basic info. 1985 * Returns pointer to filled sdl. 1986 */ 1987 struct sockaddr_dl * 1988 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 1989 { 1990 struct sockaddr_dl *sdl; 1991 1992 sdl = (struct sockaddr_dl *)paddr; 1993 memset(sdl, 0, sizeof(struct sockaddr_dl)); 1994 sdl->sdl_len = sizeof(struct sockaddr_dl); 1995 sdl->sdl_family = AF_LINK; 1996 sdl->sdl_index = ifp->if_index; 1997 sdl->sdl_type = iftype; 1998 1999 return (sdl); 2000 } 2001 2002 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2003 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2004 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2005 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2006 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2007 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2008 int (*vlan_setcookie_p)(struct ifnet *, void *); 2009 void *(*vlan_cookie_p)(struct ifnet *); 2010 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 2011 2012 /* 2013 * Handle a change in the interface link state. To avoid LORs 2014 * between driver lock and upper layer locks, as well as possible 2015 * recursions, we post event to taskqueue, and all job 2016 * is done in static do_link_state_change(). 2017 */ 2018 void 2019 if_link_state_change(struct ifnet *ifp, int link_state) 2020 { 2021 /* Return if state hasn't changed. */ 2022 if (ifp->if_link_state == link_state) 2023 return; 2024 2025 ifp->if_link_state = link_state; 2026 2027 /* XXXGL: reference ifp? */ 2028 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2029 } 2030 2031 static void 2032 do_link_state_change(void *arg, int pending) 2033 { 2034 struct ifnet *ifp; 2035 int link_state; 2036 2037 ifp = arg; 2038 link_state = ifp->if_link_state; 2039 2040 CURVNET_SET(ifp->if_vnet); 2041 rt_ifmsg(ifp, 0); 2042 if (ifp->if_vlantrunk != NULL) 2043 (*vlan_link_state_p)(ifp); 2044 if (ifp->if_carp) 2045 (*carp_linkstate_p)(ifp); 2046 if (ifp->if_bridge) 2047 ifp->if_bridge_linkstate(ifp); 2048 if (ifp->if_lagg) 2049 (*lagg_linkstate_p)(ifp, link_state); 2050 2051 if (IS_DEFAULT_VNET(curvnet)) 2052 devctl_notify("IFNET", ifp->if_xname, 2053 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2054 NULL); 2055 if (pending > 1) 2056 if_printf(ifp, "%d link states coalesced\n", pending); 2057 if (log_link_state_change) 2058 if_printf(ifp, "link state changed to %s\n", 2059 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2060 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2061 CURVNET_RESTORE(); 2062 } 2063 2064 /* 2065 * Mark an interface down and notify protocols of 2066 * the transition. 2067 */ 2068 void 2069 if_down(struct ifnet *ifp) 2070 { 2071 2072 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2073 2074 ifp->if_flags &= ~IFF_UP; 2075 getmicrotime(&ifp->if_lastchange); 2076 ifp->if_qflush(ifp); 2077 2078 if (ifp->if_carp) 2079 (*carp_linkstate_p)(ifp); 2080 rt_ifmsg(ifp, IFF_UP); 2081 } 2082 2083 /* 2084 * Mark an interface up and notify protocols of 2085 * the transition. 2086 */ 2087 void 2088 if_up(struct ifnet *ifp) 2089 { 2090 2091 ifp->if_flags |= IFF_UP; 2092 getmicrotime(&ifp->if_lastchange); 2093 if (ifp->if_carp) 2094 (*carp_linkstate_p)(ifp); 2095 rt_ifmsg(ifp, IFF_UP); 2096 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2097 } 2098 2099 /* 2100 * Flush an interface queue. 2101 */ 2102 void 2103 if_qflush(struct ifnet *ifp) 2104 { 2105 struct mbuf *m, *n; 2106 struct ifaltq *ifq; 2107 2108 ifq = &ifp->if_snd; 2109 IFQ_LOCK(ifq); 2110 #ifdef ALTQ 2111 if (ALTQ_IS_ENABLED(ifq)) 2112 ALTQ_PURGE(ifq); 2113 #endif 2114 n = ifq->ifq_head; 2115 while ((m = n) != NULL) { 2116 n = m->m_nextpkt; 2117 m_freem(m); 2118 } 2119 ifq->ifq_head = 0; 2120 ifq->ifq_tail = 0; 2121 ifq->ifq_len = 0; 2122 IFQ_UNLOCK(ifq); 2123 } 2124 2125 /* 2126 * Map interface name to interface structure pointer, with or without 2127 * returning a reference. 2128 */ 2129 struct ifnet * 2130 ifunit_ref(const char *name) 2131 { 2132 struct epoch_tracker et; 2133 struct ifnet *ifp; 2134 2135 NET_EPOCH_ENTER(et); 2136 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2137 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2138 !(ifp->if_flags & IFF_DYING)) { 2139 MPASS(ifp->if_vnet == curvnet); 2140 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2141 if_ref(ifp); 2142 break; 2143 } 2144 } 2145 NET_EPOCH_EXIT(et); 2146 return (ifp); 2147 } 2148 2149 struct ifnet * 2150 ifunit(const char *name) 2151 { 2152 struct epoch_tracker et; 2153 struct ifnet *ifp; 2154 2155 NET_EPOCH_ENTER(et); 2156 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2157 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) { 2158 MPASS(refcount_load(&ifp->if_refcount) > 0); 2159 MPASS(ifp->if_vnet == curvnet); 2160 MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp); 2161 break; 2162 } 2163 } 2164 NET_EPOCH_EXIT(et); 2165 return (ifp); 2166 } 2167 2168 void * 2169 ifr_buffer_get_buffer(void *data) 2170 { 2171 union ifreq_union *ifrup; 2172 2173 ifrup = data; 2174 #ifdef COMPAT_FREEBSD32 2175 if (SV_CURPROC_FLAG(SV_ILP32)) 2176 return ((void *)(uintptr_t) 2177 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2178 #endif 2179 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2180 } 2181 2182 static void 2183 ifr_buffer_set_buffer_null(void *data) 2184 { 2185 union ifreq_union *ifrup; 2186 2187 ifrup = data; 2188 #ifdef COMPAT_FREEBSD32 2189 if (SV_CURPROC_FLAG(SV_ILP32)) 2190 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2191 else 2192 #endif 2193 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2194 } 2195 2196 size_t 2197 ifr_buffer_get_length(void *data) 2198 { 2199 union ifreq_union *ifrup; 2200 2201 ifrup = data; 2202 #ifdef COMPAT_FREEBSD32 2203 if (SV_CURPROC_FLAG(SV_ILP32)) 2204 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2205 #endif 2206 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2207 } 2208 2209 static void 2210 ifr_buffer_set_length(void *data, size_t len) 2211 { 2212 union ifreq_union *ifrup; 2213 2214 ifrup = data; 2215 #ifdef COMPAT_FREEBSD32 2216 if (SV_CURPROC_FLAG(SV_ILP32)) 2217 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2218 else 2219 #endif 2220 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2221 } 2222 2223 void * 2224 ifr_data_get_ptr(void *ifrp) 2225 { 2226 union ifreq_union *ifrup; 2227 2228 ifrup = ifrp; 2229 #ifdef COMPAT_FREEBSD32 2230 if (SV_CURPROC_FLAG(SV_ILP32)) 2231 return ((void *)(uintptr_t) 2232 ifrup->ifr32.ifr_ifru.ifru_data); 2233 #endif 2234 return (ifrup->ifr.ifr_ifru.ifru_data); 2235 } 2236 2237 struct ifcap_nv_bit_name { 2238 uint64_t cap_bit; 2239 const char *cap_name; 2240 }; 2241 #define CAPNV(x) {.cap_bit = IFCAP_##x, \ 2242 .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) } 2243 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = { 2244 CAPNV(RXCSUM), 2245 CAPNV(TXCSUM), 2246 CAPNV(NETCONS), 2247 CAPNV(VLAN_MTU), 2248 CAPNV(VLAN_HWTAGGING), 2249 CAPNV(JUMBO_MTU), 2250 CAPNV(POLLING), 2251 CAPNV(VLAN_HWCSUM), 2252 CAPNV(TSO4), 2253 CAPNV(TSO6), 2254 CAPNV(LRO), 2255 CAPNV(WOL_UCAST), 2256 CAPNV(WOL_MCAST), 2257 CAPNV(WOL_MAGIC), 2258 CAPNV(TOE4), 2259 CAPNV(TOE6), 2260 CAPNV(VLAN_HWFILTER), 2261 CAPNV(VLAN_HWTSO), 2262 CAPNV(LINKSTATE), 2263 CAPNV(NETMAP), 2264 CAPNV(RXCSUM_IPV6), 2265 CAPNV(TXCSUM_IPV6), 2266 CAPNV(HWSTATS), 2267 CAPNV(TXRTLMT), 2268 CAPNV(HWRXTSTMP), 2269 CAPNV(MEXTPG), 2270 CAPNV(TXTLS4), 2271 CAPNV(TXTLS6), 2272 CAPNV(VXLAN_HWCSUM), 2273 CAPNV(VXLAN_HWTSO), 2274 CAPNV(TXTLS_RTLMT), 2275 {0, NULL} 2276 }; 2277 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \ 2278 .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) } 2279 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = { 2280 CAP2NV(RXTLS4), 2281 CAP2NV(RXTLS6), 2282 CAP2NV(IPSEC_OFFLOAD), 2283 CAP2NV(GENEVE_HWCSUM), 2284 CAP2NV(GENEVE_HWTSO), 2285 {0, NULL} 2286 }; 2287 #undef CAPNV 2288 #undef CAP2NV 2289 2290 int 2291 if_capnv_to_capint(const nvlist_t *nv, int *old_cap, 2292 const struct ifcap_nv_bit_name *nn, bool all) 2293 { 2294 int i, res; 2295 2296 res = 0; 2297 for (i = 0; nn[i].cap_name != NULL; i++) { 2298 if (nvlist_exists_bool(nv, nn[i].cap_name)) { 2299 if (all || nvlist_get_bool(nv, nn[i].cap_name)) 2300 res |= nn[i].cap_bit; 2301 } else { 2302 res |= *old_cap & nn[i].cap_bit; 2303 } 2304 } 2305 return (res); 2306 } 2307 2308 void 2309 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn, 2310 int ifr_cap, int ifr_req) 2311 { 2312 int i; 2313 2314 for (i = 0; nn[i].cap_name != NULL; i++) { 2315 if ((nn[i].cap_bit & ifr_cap) != 0) { 2316 nvlist_add_bool(nv, nn[i].cap_name, 2317 (nn[i].cap_bit & ifr_req) != 0); 2318 } 2319 } 2320 } 2321 2322 /* 2323 * Hardware specific interface ioctls. 2324 */ 2325 int 2326 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2327 { 2328 struct ifreq *ifr; 2329 int error = 0, do_ifup = 0; 2330 int new_flags, temp_flags; 2331 size_t descrlen, nvbuflen; 2332 char *descrbuf; 2333 char new_name[IFNAMSIZ]; 2334 void *buf; 2335 nvlist_t *nvcap; 2336 struct siocsifcapnv_driver_data drv_ioctl_data; 2337 2338 ifr = (struct ifreq *)data; 2339 switch (cmd) { 2340 case SIOCGIFINDEX: 2341 ifr->ifr_index = ifp->if_index; 2342 break; 2343 2344 case SIOCGIFFLAGS: 2345 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2346 ifr->ifr_flags = temp_flags & 0xffff; 2347 ifr->ifr_flagshigh = temp_flags >> 16; 2348 break; 2349 2350 case SIOCGIFCAP: 2351 ifr->ifr_reqcap = ifp->if_capabilities; 2352 ifr->ifr_curcap = ifp->if_capenable; 2353 break; 2354 2355 case SIOCGIFCAPNV: 2356 if ((ifp->if_capabilities & IFCAP_NV) == 0) { 2357 error = EINVAL; 2358 break; 2359 } 2360 buf = NULL; 2361 nvcap = nvlist_create(0); 2362 for (;;) { 2363 if_capint_to_capnv(nvcap, ifcap_nv_bit_names, 2364 ifp->if_capabilities, ifp->if_capenable); 2365 if_capint_to_capnv(nvcap, ifcap2_nv_bit_names, 2366 ifp->if_capabilities2, ifp->if_capenable2); 2367 error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV, 2368 __DECONST(caddr_t, nvcap)); 2369 if (error != 0) { 2370 if_printf(ifp, 2371 "SIOCGIFCAPNV driver mistake: nvlist error %d\n", 2372 error); 2373 break; 2374 } 2375 buf = nvlist_pack(nvcap, &nvbuflen); 2376 if (buf == NULL) { 2377 error = nvlist_error(nvcap); 2378 if (error == 0) 2379 error = EDOOFUS; 2380 break; 2381 } 2382 if (nvbuflen > ifr->ifr_cap_nv.buf_length) { 2383 ifr->ifr_cap_nv.length = nvbuflen; 2384 ifr->ifr_cap_nv.buffer = NULL; 2385 error = EFBIG; 2386 break; 2387 } 2388 ifr->ifr_cap_nv.length = nvbuflen; 2389 error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen); 2390 break; 2391 } 2392 free(buf, M_NVLIST); 2393 nvlist_destroy(nvcap); 2394 break; 2395 2396 case SIOCGIFDATA: 2397 { 2398 struct if_data ifd; 2399 2400 /* Ensure uninitialised padding is not leaked. */ 2401 memset(&ifd, 0, sizeof(ifd)); 2402 2403 if_data_copy(ifp, &ifd); 2404 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2405 break; 2406 } 2407 2408 #ifdef MAC 2409 case SIOCGIFMAC: 2410 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2411 break; 2412 #endif 2413 2414 case SIOCGIFMETRIC: 2415 ifr->ifr_metric = ifp->if_metric; 2416 break; 2417 2418 case SIOCGIFMTU: 2419 ifr->ifr_mtu = ifp->if_mtu; 2420 break; 2421 2422 case SIOCGIFPHYS: 2423 /* XXXGL: did this ever worked? */ 2424 ifr->ifr_phys = 0; 2425 break; 2426 2427 case SIOCGIFDESCR: 2428 error = 0; 2429 sx_slock(&ifdescr_sx); 2430 if (ifp->if_description == NULL) 2431 error = ENOMSG; 2432 else { 2433 /* space for terminating nul */ 2434 descrlen = strlen(ifp->if_description) + 1; 2435 if (ifr_buffer_get_length(ifr) < descrlen) 2436 ifr_buffer_set_buffer_null(ifr); 2437 else 2438 error = copyout(ifp->if_description, 2439 ifr_buffer_get_buffer(ifr), descrlen); 2440 ifr_buffer_set_length(ifr, descrlen); 2441 } 2442 sx_sunlock(&ifdescr_sx); 2443 break; 2444 2445 case SIOCSIFDESCR: 2446 error = priv_check(td, PRIV_NET_SETIFDESCR); 2447 if (error) 2448 return (error); 2449 2450 /* 2451 * Copy only (length-1) bytes to make sure that 2452 * if_description is always nul terminated. The 2453 * length parameter is supposed to count the 2454 * terminating nul in. 2455 */ 2456 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2457 return (ENAMETOOLONG); 2458 else if (ifr_buffer_get_length(ifr) == 0) 2459 descrbuf = NULL; 2460 else { 2461 descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK); 2462 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2463 ifr_buffer_get_length(ifr) - 1); 2464 if (error) { 2465 if_freedescr(descrbuf); 2466 break; 2467 } 2468 } 2469 2470 if_setdescr(ifp, descrbuf); 2471 getmicrotime(&ifp->if_lastchange); 2472 break; 2473 2474 case SIOCGIFFIB: 2475 ifr->ifr_fib = ifp->if_fib; 2476 break; 2477 2478 case SIOCSIFFIB: 2479 error = priv_check(td, PRIV_NET_SETIFFIB); 2480 if (error) 2481 return (error); 2482 if (ifr->ifr_fib >= rt_numfibs) 2483 return (EINVAL); 2484 2485 ifp->if_fib = ifr->ifr_fib; 2486 break; 2487 2488 case SIOCSIFFLAGS: 2489 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2490 if (error) 2491 return (error); 2492 /* 2493 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2494 * check, so we don't need special handling here yet. 2495 */ 2496 new_flags = (ifr->ifr_flags & 0xffff) | 2497 (ifr->ifr_flagshigh << 16); 2498 if (ifp->if_flags & IFF_UP && 2499 (new_flags & IFF_UP) == 0) { 2500 if_down(ifp); 2501 } else if (new_flags & IFF_UP && 2502 (ifp->if_flags & IFF_UP) == 0) { 2503 do_ifup = 1; 2504 } 2505 2506 /* 2507 * See if the promiscuous mode or allmulti bits are about to 2508 * flip. They require special handling because in-kernel 2509 * consumers may indepdently toggle them. 2510 */ 2511 if_setppromisc(ifp, new_flags & IFF_PPROMISC); 2512 if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) { 2513 if (new_flags & IFF_PALLMULTI) 2514 ifp->if_flags |= IFF_ALLMULTI; 2515 else if (ifp->if_amcount == 0) 2516 ifp->if_flags &= ~IFF_ALLMULTI; 2517 } 2518 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2519 (new_flags &~ IFF_CANTCHANGE); 2520 if (ifp->if_ioctl) { 2521 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2522 } 2523 if (do_ifup) 2524 if_up(ifp); 2525 getmicrotime(&ifp->if_lastchange); 2526 break; 2527 2528 case SIOCSIFCAP: 2529 error = priv_check(td, PRIV_NET_SETIFCAP); 2530 if (error != 0) 2531 return (error); 2532 if (ifp->if_ioctl == NULL) 2533 return (EOPNOTSUPP); 2534 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2535 return (EINVAL); 2536 error = (*ifp->if_ioctl)(ifp, cmd, data); 2537 if (error == 0) 2538 getmicrotime(&ifp->if_lastchange); 2539 break; 2540 2541 case SIOCSIFCAPNV: 2542 error = priv_check(td, PRIV_NET_SETIFCAP); 2543 if (error != 0) 2544 return (error); 2545 if (ifp->if_ioctl == NULL) 2546 return (EOPNOTSUPP); 2547 if ((ifp->if_capabilities & IFCAP_NV) == 0) 2548 return (EINVAL); 2549 if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE) 2550 return (EINVAL); 2551 nvcap = NULL; 2552 buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK); 2553 for (;;) { 2554 error = copyin(ifr->ifr_cap_nv.buffer, buf, 2555 ifr->ifr_cap_nv.length); 2556 if (error != 0) 2557 break; 2558 nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0); 2559 if (nvcap == NULL) { 2560 error = EINVAL; 2561 break; 2562 } 2563 drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap, 2564 &ifp->if_capenable, ifcap_nv_bit_names, false); 2565 if ((drv_ioctl_data.reqcap & 2566 ~ifp->if_capabilities) != 0) { 2567 error = EINVAL; 2568 break; 2569 } 2570 drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap, 2571 &ifp->if_capenable2, ifcap2_nv_bit_names, false); 2572 if ((drv_ioctl_data.reqcap2 & 2573 ~ifp->if_capabilities2) != 0) { 2574 error = EINVAL; 2575 break; 2576 } 2577 drv_ioctl_data.nvcap = nvcap; 2578 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV, 2579 (caddr_t)&drv_ioctl_data); 2580 break; 2581 } 2582 nvlist_destroy(nvcap); 2583 free(buf, M_TEMP); 2584 if (error == 0) 2585 getmicrotime(&ifp->if_lastchange); 2586 break; 2587 2588 #ifdef MAC 2589 case SIOCSIFMAC: 2590 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2591 break; 2592 #endif 2593 2594 case SIOCSIFNAME: 2595 error = priv_check(td, PRIV_NET_SETIFNAME); 2596 if (error) 2597 return (error); 2598 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2599 NULL); 2600 if (error != 0) 2601 return (error); 2602 error = if_rename(ifp, new_name); 2603 break; 2604 2605 case SIOCSIFMETRIC: 2606 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2607 if (error) 2608 return (error); 2609 ifp->if_metric = ifr->ifr_metric; 2610 getmicrotime(&ifp->if_lastchange); 2611 break; 2612 2613 case SIOCSIFPHYS: 2614 error = priv_check(td, PRIV_NET_SETIFPHYS); 2615 if (error) 2616 return (error); 2617 if (ifp->if_ioctl == NULL) 2618 return (EOPNOTSUPP); 2619 error = (*ifp->if_ioctl)(ifp, cmd, data); 2620 if (error == 0) 2621 getmicrotime(&ifp->if_lastchange); 2622 break; 2623 2624 case SIOCSIFMTU: 2625 { 2626 u_long oldmtu = ifp->if_mtu; 2627 2628 error = priv_check(td, PRIV_NET_SETIFMTU); 2629 if (error) 2630 return (error); 2631 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2632 return (EINVAL); 2633 if (ifp->if_ioctl == NULL) 2634 return (EOPNOTSUPP); 2635 /* Disallow MTU changes on bridge member interfaces. */ 2636 if (ifp->if_bridge) 2637 return (EOPNOTSUPP); 2638 error = (*ifp->if_ioctl)(ifp, cmd, data); 2639 if (error == 0) { 2640 getmicrotime(&ifp->if_lastchange); 2641 rt_ifmsg(ifp, 0); 2642 #ifdef INET 2643 DEBUGNET_NOTIFY_MTU(ifp); 2644 #endif 2645 } 2646 /* 2647 * If the link MTU changed, do network layer specific procedure. 2648 */ 2649 if (ifp->if_mtu != oldmtu) 2650 if_notifymtu(ifp); 2651 break; 2652 } 2653 2654 case SIOCADDMULTI: 2655 case SIOCDELMULTI: 2656 if (cmd == SIOCADDMULTI) 2657 error = priv_check(td, PRIV_NET_ADDMULTI); 2658 else 2659 error = priv_check(td, PRIV_NET_DELMULTI); 2660 if (error) 2661 return (error); 2662 2663 /* Don't allow group membership on non-multicast interfaces. */ 2664 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2665 return (EOPNOTSUPP); 2666 2667 /* Don't let users screw up protocols' entries. */ 2668 if (ifr->ifr_addr.sa_family != AF_LINK) 2669 return (EINVAL); 2670 2671 if (cmd == SIOCADDMULTI) { 2672 struct epoch_tracker et; 2673 struct ifmultiaddr *ifma; 2674 2675 /* 2676 * Userland is only permitted to join groups once 2677 * via the if_addmulti() KPI, because it cannot hold 2678 * struct ifmultiaddr * between calls. It may also 2679 * lose a race while we check if the membership 2680 * already exists. 2681 */ 2682 NET_EPOCH_ENTER(et); 2683 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2684 NET_EPOCH_EXIT(et); 2685 if (ifma != NULL) 2686 error = EADDRINUSE; 2687 else 2688 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2689 } else { 2690 error = if_delmulti(ifp, &ifr->ifr_addr); 2691 } 2692 if (error == 0) 2693 getmicrotime(&ifp->if_lastchange); 2694 break; 2695 2696 case SIOCSIFPHYADDR: 2697 case SIOCDIFPHYADDR: 2698 #ifdef INET6 2699 case SIOCSIFPHYADDR_IN6: 2700 #endif 2701 case SIOCSIFMEDIA: 2702 case SIOCSIFGENERIC: 2703 error = priv_check(td, PRIV_NET_HWIOCTL); 2704 if (error) 2705 return (error); 2706 if (ifp->if_ioctl == NULL) 2707 return (EOPNOTSUPP); 2708 error = (*ifp->if_ioctl)(ifp, cmd, data); 2709 if (error == 0) 2710 getmicrotime(&ifp->if_lastchange); 2711 break; 2712 2713 case SIOCGIFSTATUS: 2714 case SIOCGIFPSRCADDR: 2715 case SIOCGIFPDSTADDR: 2716 case SIOCGIFMEDIA: 2717 case SIOCGIFXMEDIA: 2718 case SIOCGIFGENERIC: 2719 case SIOCGIFRSSKEY: 2720 case SIOCGIFRSSHASH: 2721 case SIOCGIFDOWNREASON: 2722 if (ifp->if_ioctl == NULL) 2723 return (EOPNOTSUPP); 2724 error = (*ifp->if_ioctl)(ifp, cmd, data); 2725 break; 2726 2727 case SIOCSIFLLADDR: 2728 error = priv_check(td, PRIV_NET_SETLLADDR); 2729 if (error) 2730 return (error); 2731 error = if_setlladdr(ifp, 2732 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2733 break; 2734 2735 case SIOCGHWADDR: 2736 error = if_gethwaddr(ifp, ifr); 2737 break; 2738 2739 case SIOCAIFGROUP: 2740 { 2741 const char *groupname; 2742 2743 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2744 if (error) 2745 return (error); 2746 groupname = ((struct ifgroupreq *)data)->ifgr_group; 2747 if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ) 2748 return (EINVAL); 2749 error = if_addgroup(ifp, groupname); 2750 if (error != 0) 2751 return (error); 2752 break; 2753 } 2754 case SIOCGIFGROUP: 2755 error = if_getgroup((struct ifgroupreq *)data, ifp); 2756 break; 2757 2758 case SIOCDIFGROUP: 2759 { 2760 const char *groupname; 2761 2762 error = priv_check(td, PRIV_NET_DELIFGROUP); 2763 if (error) 2764 return (error); 2765 groupname = ((struct ifgroupreq *)data)->ifgr_group; 2766 if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ) 2767 return (EINVAL); 2768 error = if_delgroup(ifp, groupname); 2769 if (error != 0) 2770 return (error); 2771 break; 2772 } 2773 default: 2774 error = ENOIOCTL; 2775 break; 2776 } 2777 return (error); 2778 } 2779 2780 /* 2781 * Interface ioctls. 2782 */ 2783 int 2784 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2785 { 2786 #ifdef COMPAT_FREEBSD32 2787 union { 2788 struct ifconf ifc; 2789 struct ifdrv ifd; 2790 struct ifgroupreq ifgr; 2791 struct ifmediareq ifmr; 2792 } thunk; 2793 u_long saved_cmd; 2794 struct ifconf32 *ifc32; 2795 struct ifdrv32 *ifd32; 2796 struct ifgroupreq32 *ifgr32; 2797 struct ifmediareq32 *ifmr32; 2798 #endif 2799 struct ifnet *ifp; 2800 struct ifreq *ifr; 2801 int error; 2802 int oif_flags; 2803 #ifdef VIMAGE 2804 bool shutdown; 2805 #endif 2806 2807 CURVNET_SET(so->so_vnet); 2808 #ifdef VIMAGE 2809 /* Make sure the VNET is stable. */ 2810 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2811 if (shutdown) { 2812 CURVNET_RESTORE(); 2813 return (EBUSY); 2814 } 2815 #endif 2816 2817 #ifdef COMPAT_FREEBSD32 2818 saved_cmd = cmd; 2819 switch (cmd) { 2820 case SIOCGIFCONF32: 2821 ifc32 = (struct ifconf32 *)data; 2822 thunk.ifc.ifc_len = ifc32->ifc_len; 2823 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2824 data = (caddr_t)&thunk.ifc; 2825 cmd = SIOCGIFCONF; 2826 break; 2827 case SIOCGDRVSPEC32: 2828 case SIOCSDRVSPEC32: 2829 ifd32 = (struct ifdrv32 *)data; 2830 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2831 sizeof(thunk.ifd.ifd_name)); 2832 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2833 thunk.ifd.ifd_len = ifd32->ifd_len; 2834 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2835 data = (caddr_t)&thunk.ifd; 2836 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2837 break; 2838 case SIOCAIFGROUP32: 2839 case SIOCGIFGROUP32: 2840 case SIOCDIFGROUP32: 2841 case SIOCGIFGMEMB32: 2842 ifgr32 = (struct ifgroupreq32 *)data; 2843 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2844 sizeof(thunk.ifgr.ifgr_name)); 2845 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2846 switch (cmd) { 2847 case SIOCAIFGROUP32: 2848 case SIOCDIFGROUP32: 2849 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2850 sizeof(thunk.ifgr.ifgr_group)); 2851 break; 2852 case SIOCGIFGROUP32: 2853 case SIOCGIFGMEMB32: 2854 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2855 break; 2856 } 2857 data = (caddr_t)&thunk.ifgr; 2858 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2859 break; 2860 case SIOCGIFMEDIA32: 2861 case SIOCGIFXMEDIA32: 2862 ifmr32 = (struct ifmediareq32 *)data; 2863 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2864 sizeof(thunk.ifmr.ifm_name)); 2865 thunk.ifmr.ifm_current = ifmr32->ifm_current; 2866 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 2867 thunk.ifmr.ifm_status = ifmr32->ifm_status; 2868 thunk.ifmr.ifm_active = ifmr32->ifm_active; 2869 thunk.ifmr.ifm_count = ifmr32->ifm_count; 2870 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 2871 data = (caddr_t)&thunk.ifmr; 2872 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 2873 break; 2874 } 2875 #endif 2876 2877 switch (cmd) { 2878 case SIOCGIFCONF: 2879 error = ifconf(cmd, data); 2880 goto out_noref; 2881 } 2882 2883 ifr = (struct ifreq *)data; 2884 switch (cmd) { 2885 #ifdef VIMAGE 2886 case SIOCSIFVNET: 2887 error = priv_check(td, PRIV_NET_SETIFVNET); 2888 if (error == 0) 2889 error = if_vmove_loan(td, ifr->ifr_name, ifr->ifr_jid); 2890 goto out_noref; 2891 2892 case SIOCSIFRVNET: 2893 error = priv_check(td, PRIV_NET_SETIFVNET); 2894 if (error == 0) 2895 error = if_vmove_reclaim(td, ifr->ifr_name, 2896 ifr->ifr_jid); 2897 goto out_noref; 2898 #endif 2899 case SIOCIFCREATE: 2900 case SIOCIFCREATE2: 2901 error = priv_check(td, PRIV_NET_IFCREATE); 2902 if (error == 0) 2903 error = if_clone_create(ifr->ifr_name, 2904 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 2905 ifr_data_get_ptr(ifr) : NULL); 2906 goto out_noref; 2907 case SIOCIFDESTROY: 2908 error = priv_check(td, PRIV_NET_IFDESTROY); 2909 2910 if (error == 0) 2911 error = if_clone_destroy(ifr->ifr_name); 2912 goto out_noref; 2913 2914 case SIOCIFGCLONERS: 2915 error = if_clone_list((struct if_clonereq *)data); 2916 goto out_noref; 2917 2918 case SIOCGIFGMEMB: 2919 { 2920 struct ifgroupreq *req; 2921 2922 req = (struct ifgroupreq *)data; 2923 if (strnlen(req->ifgr_name, IFNAMSIZ) == IFNAMSIZ) { 2924 error = EINVAL; 2925 goto out_noref; 2926 } 2927 error = if_getgroupmembers(req); 2928 goto out_noref; 2929 } 2930 } 2931 2932 ifp = ifunit_ref(ifr->ifr_name); 2933 if (ifp == NULL) { 2934 error = ENXIO; 2935 goto out_noref; 2936 } 2937 2938 error = ifhwioctl(cmd, ifp, data, td); 2939 if (error != ENOIOCTL) 2940 goto out_ref; 2941 2942 oif_flags = ifp->if_flags; 2943 if (so->so_proto == NULL) { 2944 error = EOPNOTSUPP; 2945 goto out_ref; 2946 } 2947 2948 /* 2949 * Pass the request on to the socket control method, and if the 2950 * latter returns EOPNOTSUPP, directly to the interface. 2951 * 2952 * Make an exception for the legacy SIOCSIF* requests. Drivers 2953 * trust SIOCSIFADDR et al to come from an already privileged 2954 * layer, and do not perform any credentials checks or input 2955 * validation. 2956 */ 2957 error = so->so_proto->pr_control(so, cmd, data, ifp, td); 2958 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 2959 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 2960 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 2961 error = (*ifp->if_ioctl)(ifp, cmd, data); 2962 2963 if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP)) 2964 if_up(ifp); 2965 out_ref: 2966 if_rele(ifp); 2967 out_noref: 2968 CURVNET_RESTORE(); 2969 #ifdef COMPAT_FREEBSD32 2970 if (error != 0) 2971 return (error); 2972 switch (saved_cmd) { 2973 case SIOCGIFCONF32: 2974 ifc32->ifc_len = thunk.ifc.ifc_len; 2975 break; 2976 case SIOCGDRVSPEC32: 2977 /* 2978 * SIOCGDRVSPEC is IOWR, but nothing actually touches 2979 * the struct so just assert that ifd_len (the only 2980 * field it might make sense to update) hasn't 2981 * changed. 2982 */ 2983 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 2984 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 2985 thunk.ifd.ifd_len)); 2986 break; 2987 case SIOCGIFGROUP32: 2988 case SIOCGIFGMEMB32: 2989 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 2990 break; 2991 case SIOCGIFMEDIA32: 2992 case SIOCGIFXMEDIA32: 2993 ifmr32->ifm_current = thunk.ifmr.ifm_current; 2994 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 2995 ifmr32->ifm_status = thunk.ifmr.ifm_status; 2996 ifmr32->ifm_active = thunk.ifmr.ifm_active; 2997 ifmr32->ifm_count = thunk.ifmr.ifm_count; 2998 break; 2999 } 3000 #endif 3001 return (error); 3002 } 3003 3004 int 3005 if_rename(struct ifnet *ifp, char *new_name) 3006 { 3007 struct ifaddr *ifa; 3008 struct sockaddr_dl *sdl; 3009 size_t namelen, onamelen; 3010 char old_name[IFNAMSIZ]; 3011 char strbuf[IFNAMSIZ + 8]; 3012 3013 if (new_name[0] == '\0') 3014 return (EINVAL); 3015 if (strcmp(new_name, ifp->if_xname) == 0) 3016 return (0); 3017 if (ifunit(new_name) != NULL) 3018 return (EEXIST); 3019 3020 if_printf(ifp, "changing name to '%s'\n", new_name); 3021 3022 IF_ADDR_WLOCK(ifp); 3023 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 3024 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 3025 ifa = ifp->if_addr; 3026 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3027 namelen = strlen(new_name); 3028 onamelen = sdl->sdl_nlen; 3029 /* 3030 * Move the address if needed. This is safe because we 3031 * allocate space for a name of length IFNAMSIZ when we 3032 * create this in if_attach(). 3033 */ 3034 if (namelen != onamelen) { 3035 bcopy(sdl->sdl_data + onamelen, 3036 sdl->sdl_data + namelen, sdl->sdl_alen); 3037 } 3038 bcopy(new_name, sdl->sdl_data, namelen); 3039 sdl->sdl_nlen = namelen; 3040 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 3041 bzero(sdl->sdl_data, onamelen); 3042 while (namelen != 0) 3043 sdl->sdl_data[--namelen] = 0xff; 3044 IF_ADDR_WUNLOCK(ifp); 3045 3046 EVENTHANDLER_INVOKE(ifnet_rename_event, ifp, old_name); 3047 3048 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 3049 devctl_notify("IFNET", old_name, "RENAME", strbuf); 3050 3051 return (0); 3052 } 3053 3054 /* 3055 * The code common to handling reference counted flags, 3056 * e.g., in ifpromisc() and if_allmulti(). 3057 * The "pflag" argument can specify a permanent mode flag to check, 3058 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3059 * 3060 * Only to be used on stack-owned flags, not driver-owned flags. 3061 */ 3062 static int 3063 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3064 { 3065 struct ifreq ifr; 3066 int error; 3067 int oldflags, oldcount; 3068 3069 /* Sanity checks to catch programming errors */ 3070 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3071 ("%s: setting driver-owned flag %d", __func__, flag)); 3072 3073 if (onswitch) 3074 KASSERT(*refcount >= 0, 3075 ("%s: increment negative refcount %d for flag %d", 3076 __func__, *refcount, flag)); 3077 else 3078 KASSERT(*refcount > 0, 3079 ("%s: decrement non-positive refcount %d for flag %d", 3080 __func__, *refcount, flag)); 3081 3082 /* In case this mode is permanent, just touch refcount */ 3083 if (ifp->if_flags & pflag) { 3084 *refcount += onswitch ? 1 : -1; 3085 return (0); 3086 } 3087 3088 /* Save ifnet parameters for if_ioctl() may fail */ 3089 oldcount = *refcount; 3090 oldflags = ifp->if_flags; 3091 3092 /* 3093 * See if we aren't the only and touching refcount is enough. 3094 * Actually toggle interface flag if we are the first or last. 3095 */ 3096 if (onswitch) { 3097 if ((*refcount)++) 3098 return (0); 3099 ifp->if_flags |= flag; 3100 } else { 3101 if (--(*refcount)) 3102 return (0); 3103 ifp->if_flags &= ~flag; 3104 } 3105 3106 /* Call down the driver since we've changed interface flags */ 3107 if (ifp->if_ioctl == NULL) { 3108 error = EOPNOTSUPP; 3109 goto recover; 3110 } 3111 ifr.ifr_flags = ifp->if_flags & 0xffff; 3112 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3113 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3114 if (error) 3115 goto recover; 3116 /* Notify userland that interface flags have changed */ 3117 rt_ifmsg(ifp, flag); 3118 return (0); 3119 3120 recover: 3121 /* Recover after driver error */ 3122 *refcount = oldcount; 3123 ifp->if_flags = oldflags; 3124 return (error); 3125 } 3126 3127 /* 3128 * Set/clear promiscuous mode on interface ifp based on the truth value 3129 * of pswitch. The calls are reference counted so that only the first 3130 * "on" request actually has an effect, as does the final "off" request. 3131 * Results are undefined if the "off" and "on" requests are not matched. 3132 */ 3133 int 3134 ifpromisc(struct ifnet *ifp, int pswitch) 3135 { 3136 int error; 3137 int oldflags = ifp->if_flags; 3138 3139 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3140 &ifp->if_pcount, pswitch); 3141 /* If promiscuous mode status has changed, log a message */ 3142 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3143 log_promisc_mode_change) 3144 if_printf(ifp, "promiscuous mode %s\n", 3145 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3146 return (error); 3147 } 3148 3149 /* 3150 * Return interface configuration 3151 * of system. List may be used 3152 * in later ioctl's (above) to get 3153 * other information. 3154 */ 3155 /*ARGSUSED*/ 3156 static int 3157 ifconf(u_long cmd, caddr_t data) 3158 { 3159 struct ifconf *ifc = (struct ifconf *)data; 3160 struct ifnet *ifp; 3161 struct ifaddr *ifa; 3162 struct ifreq ifr; 3163 struct sbuf *sb; 3164 int error, full = 0, valid_len, max_len; 3165 3166 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3167 max_len = maxphys - 1; 3168 3169 /* Prevent hostile input from being able to crash the system */ 3170 if (ifc->ifc_len <= 0) 3171 return (EINVAL); 3172 3173 again: 3174 if (ifc->ifc_len <= max_len) { 3175 max_len = ifc->ifc_len; 3176 full = 1; 3177 } 3178 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3179 max_len = 0; 3180 valid_len = 0; 3181 3182 IFNET_RLOCK(); 3183 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3184 struct epoch_tracker et; 3185 int addrs; 3186 3187 /* 3188 * Zero the ifr to make sure we don't disclose the contents 3189 * of the stack. 3190 */ 3191 memset(&ifr, 0, sizeof(ifr)); 3192 3193 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3194 >= sizeof(ifr.ifr_name)) { 3195 sbuf_delete(sb); 3196 IFNET_RUNLOCK(); 3197 return (ENAMETOOLONG); 3198 } 3199 3200 addrs = 0; 3201 NET_EPOCH_ENTER(et); 3202 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3203 struct sockaddr *sa = ifa->ifa_addr; 3204 3205 if (prison_if(curthread->td_ucred, sa) != 0) 3206 continue; 3207 addrs++; 3208 if (sa->sa_len <= sizeof(*sa)) { 3209 if (sa->sa_len < sizeof(*sa)) { 3210 memset(&ifr.ifr_ifru.ifru_addr, 0, 3211 sizeof(ifr.ifr_ifru.ifru_addr)); 3212 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3213 sa->sa_len); 3214 } else 3215 ifr.ifr_ifru.ifru_addr = *sa; 3216 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3217 max_len += sizeof(ifr); 3218 } else { 3219 sbuf_bcat(sb, &ifr, 3220 offsetof(struct ifreq, ifr_addr)); 3221 max_len += offsetof(struct ifreq, ifr_addr); 3222 sbuf_bcat(sb, sa, sa->sa_len); 3223 max_len += sa->sa_len; 3224 } 3225 3226 if (sbuf_error(sb) == 0) 3227 valid_len = sbuf_len(sb); 3228 } 3229 NET_EPOCH_EXIT(et); 3230 if (addrs == 0) { 3231 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3232 max_len += sizeof(ifr); 3233 3234 if (sbuf_error(sb) == 0) 3235 valid_len = sbuf_len(sb); 3236 } 3237 } 3238 IFNET_RUNLOCK(); 3239 3240 /* 3241 * If we didn't allocate enough space (uncommon), try again. If 3242 * we have already allocated as much space as we are allowed, 3243 * return what we've got. 3244 */ 3245 if (valid_len != max_len && !full) { 3246 sbuf_delete(sb); 3247 goto again; 3248 } 3249 3250 ifc->ifc_len = valid_len; 3251 sbuf_finish(sb); 3252 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3253 sbuf_delete(sb); 3254 return (error); 3255 } 3256 3257 /* 3258 * Just like ifpromisc(), but for all-multicast-reception mode. 3259 */ 3260 int 3261 if_allmulti(struct ifnet *ifp, int onswitch) 3262 { 3263 3264 return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount, 3265 onswitch)); 3266 } 3267 3268 struct ifmultiaddr * 3269 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3270 { 3271 struct ifmultiaddr *ifma; 3272 3273 IF_ADDR_LOCK_ASSERT(ifp); 3274 3275 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3276 if (sa->sa_family == AF_LINK) { 3277 if (sa_dl_equal(ifma->ifma_addr, sa)) 3278 break; 3279 } else { 3280 if (sa_equal(ifma->ifma_addr, sa)) 3281 break; 3282 } 3283 } 3284 3285 return ifma; 3286 } 3287 3288 /* 3289 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3290 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3291 * the ifnet multicast address list here, so the caller must do that and 3292 * other setup work (such as notifying the device driver). The reference 3293 * count is initialized to 1. 3294 */ 3295 static struct ifmultiaddr * 3296 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3297 int mflags) 3298 { 3299 struct ifmultiaddr *ifma; 3300 struct sockaddr *dupsa; 3301 3302 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3303 M_ZERO); 3304 if (ifma == NULL) 3305 return (NULL); 3306 3307 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3308 if (dupsa == NULL) { 3309 free(ifma, M_IFMADDR); 3310 return (NULL); 3311 } 3312 bcopy(sa, dupsa, sa->sa_len); 3313 ifma->ifma_addr = dupsa; 3314 3315 ifma->ifma_ifp = ifp; 3316 ifma->ifma_refcount = 1; 3317 ifma->ifma_protospec = NULL; 3318 3319 if (llsa == NULL) { 3320 ifma->ifma_lladdr = NULL; 3321 return (ifma); 3322 } 3323 3324 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3325 if (dupsa == NULL) { 3326 free(ifma->ifma_addr, M_IFMADDR); 3327 free(ifma, M_IFMADDR); 3328 return (NULL); 3329 } 3330 bcopy(llsa, dupsa, llsa->sa_len); 3331 ifma->ifma_lladdr = dupsa; 3332 3333 return (ifma); 3334 } 3335 3336 /* 3337 * if_freemulti: free ifmultiaddr structure and possibly attached related 3338 * addresses. The caller is responsible for implementing reference 3339 * counting, notifying the driver, handling routing messages, and releasing 3340 * any dependent link layer state. 3341 */ 3342 #ifdef MCAST_VERBOSE 3343 extern void kdb_backtrace(void); 3344 #endif 3345 static void 3346 if_freemulti_internal(struct ifmultiaddr *ifma) 3347 { 3348 3349 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3350 ifma->ifma_refcount)); 3351 3352 if (ifma->ifma_lladdr != NULL) 3353 free(ifma->ifma_lladdr, M_IFMADDR); 3354 #ifdef MCAST_VERBOSE 3355 kdb_backtrace(); 3356 printf("%s freeing ifma: %p\n", __func__, ifma); 3357 #endif 3358 free(ifma->ifma_addr, M_IFMADDR); 3359 free(ifma, M_IFMADDR); 3360 } 3361 3362 static void 3363 if_destroymulti(epoch_context_t ctx) 3364 { 3365 struct ifmultiaddr *ifma; 3366 3367 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3368 if_freemulti_internal(ifma); 3369 } 3370 3371 void 3372 if_freemulti(struct ifmultiaddr *ifma) 3373 { 3374 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3375 ifma->ifma_refcount)); 3376 3377 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3378 } 3379 3380 /* 3381 * Register an additional multicast address with a network interface. 3382 * 3383 * - If the address is already present, bump the reference count on the 3384 * address and return. 3385 * - If the address is not link-layer, look up a link layer address. 3386 * - Allocate address structures for one or both addresses, and attach to the 3387 * multicast address list on the interface. If automatically adding a link 3388 * layer address, the protocol address will own a reference to the link 3389 * layer address, to be freed when it is freed. 3390 * - Notify the network device driver of an addition to the multicast address 3391 * list. 3392 * 3393 * 'sa' points to caller-owned memory with the desired multicast address. 3394 * 3395 * 'retifma' will be used to return a pointer to the resulting multicast 3396 * address reference, if desired. 3397 */ 3398 int 3399 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3400 struct ifmultiaddr **retifma) 3401 { 3402 struct ifmultiaddr *ifma, *ll_ifma; 3403 struct sockaddr *llsa; 3404 struct sockaddr_dl sdl; 3405 int error; 3406 3407 #ifdef INET 3408 IN_MULTI_LIST_UNLOCK_ASSERT(); 3409 #endif 3410 #ifdef INET6 3411 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3412 #endif 3413 /* 3414 * If the address is already present, return a new reference to it; 3415 * otherwise, allocate storage and set up a new address. 3416 */ 3417 IF_ADDR_WLOCK(ifp); 3418 ifma = if_findmulti(ifp, sa); 3419 if (ifma != NULL) { 3420 ifma->ifma_refcount++; 3421 if (retifma != NULL) 3422 *retifma = ifma; 3423 IF_ADDR_WUNLOCK(ifp); 3424 return (0); 3425 } 3426 3427 /* 3428 * The address isn't already present; resolve the protocol address 3429 * into a link layer address, and then look that up, bump its 3430 * refcount or allocate an ifma for that also. 3431 * Most link layer resolving functions returns address data which 3432 * fits inside default sockaddr_dl structure. However callback 3433 * can allocate another sockaddr structure, in that case we need to 3434 * free it later. 3435 */ 3436 llsa = NULL; 3437 ll_ifma = NULL; 3438 if (ifp->if_resolvemulti != NULL) { 3439 /* Provide called function with buffer size information */ 3440 sdl.sdl_len = sizeof(sdl); 3441 llsa = (struct sockaddr *)&sdl; 3442 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3443 if (error) 3444 goto unlock_out; 3445 } 3446 3447 /* 3448 * Allocate the new address. Don't hook it up yet, as we may also 3449 * need to allocate a link layer multicast address. 3450 */ 3451 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3452 if (ifma == NULL) { 3453 error = ENOMEM; 3454 goto free_llsa_out; 3455 } 3456 3457 /* 3458 * If a link layer address is found, we'll need to see if it's 3459 * already present in the address list, or allocate is as well. 3460 * When this block finishes, the link layer address will be on the 3461 * list. 3462 */ 3463 if (llsa != NULL) { 3464 ll_ifma = if_findmulti(ifp, llsa); 3465 if (ll_ifma == NULL) { 3466 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3467 if (ll_ifma == NULL) { 3468 --ifma->ifma_refcount; 3469 if_freemulti(ifma); 3470 error = ENOMEM; 3471 goto free_llsa_out; 3472 } 3473 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3474 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3475 ifma_link); 3476 } else 3477 ll_ifma->ifma_refcount++; 3478 ifma->ifma_llifma = ll_ifma; 3479 } 3480 3481 /* 3482 * We now have a new multicast address, ifma, and possibly a new or 3483 * referenced link layer address. Add the primary address to the 3484 * ifnet address list. 3485 */ 3486 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3487 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3488 3489 if (retifma != NULL) 3490 *retifma = ifma; 3491 3492 /* 3493 * Must generate the message while holding the lock so that 'ifma' 3494 * pointer is still valid. 3495 */ 3496 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3497 IF_ADDR_WUNLOCK(ifp); 3498 3499 /* 3500 * We are certain we have added something, so call down to the 3501 * interface to let them know about it. 3502 */ 3503 if (ifp->if_ioctl != NULL) { 3504 if (THREAD_CAN_SLEEP()) 3505 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3506 else 3507 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3508 } 3509 3510 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3511 link_free_sdl(llsa); 3512 3513 return (0); 3514 3515 free_llsa_out: 3516 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3517 link_free_sdl(llsa); 3518 3519 unlock_out: 3520 IF_ADDR_WUNLOCK(ifp); 3521 return (error); 3522 } 3523 3524 static void 3525 if_siocaddmulti(void *arg, int pending) 3526 { 3527 struct ifnet *ifp; 3528 3529 ifp = arg; 3530 #ifdef DIAGNOSTIC 3531 if (pending > 1) 3532 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3533 #endif 3534 CURVNET_SET(ifp->if_vnet); 3535 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3536 CURVNET_RESTORE(); 3537 } 3538 3539 /* 3540 * Delete a multicast group membership by network-layer group address. 3541 * 3542 * Returns ENOENT if the entry could not be found. If ifp no longer 3543 * exists, results are undefined. This entry point should only be used 3544 * from subsystems which do appropriate locking to hold ifp for the 3545 * duration of the call. 3546 * Network-layer protocol domains must use if_delmulti_ifma(). 3547 */ 3548 int 3549 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3550 { 3551 struct ifmultiaddr *ifma; 3552 int lastref; 3553 3554 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3555 3556 IF_ADDR_WLOCK(ifp); 3557 lastref = 0; 3558 ifma = if_findmulti(ifp, sa); 3559 if (ifma != NULL) 3560 lastref = if_delmulti_locked(ifp, ifma, 0); 3561 IF_ADDR_WUNLOCK(ifp); 3562 3563 if (ifma == NULL) 3564 return (ENOENT); 3565 3566 if (lastref && ifp->if_ioctl != NULL) { 3567 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3568 } 3569 3570 return (0); 3571 } 3572 3573 /* 3574 * Delete all multicast group membership for an interface. 3575 * Should be used to quickly flush all multicast filters. 3576 */ 3577 void 3578 if_delallmulti(struct ifnet *ifp) 3579 { 3580 struct ifmultiaddr *ifma; 3581 struct ifmultiaddr *next; 3582 3583 IF_ADDR_WLOCK(ifp); 3584 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3585 if_delmulti_locked(ifp, ifma, 0); 3586 IF_ADDR_WUNLOCK(ifp); 3587 } 3588 3589 void 3590 if_delmulti_ifma(struct ifmultiaddr *ifma) 3591 { 3592 if_delmulti_ifma_flags(ifma, 0); 3593 } 3594 3595 /* 3596 * Delete a multicast group membership by group membership pointer. 3597 * Network-layer protocol domains must use this routine. 3598 * 3599 * It is safe to call this routine if the ifp disappeared. 3600 */ 3601 void 3602 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3603 { 3604 struct ifnet *ifp; 3605 int lastref; 3606 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3607 #ifdef INET 3608 IN_MULTI_LIST_UNLOCK_ASSERT(); 3609 #endif 3610 ifp = ifma->ifma_ifp; 3611 #ifdef DIAGNOSTIC 3612 if (ifp == NULL) { 3613 printf("%s: ifma_ifp seems to be detached\n", __func__); 3614 } else { 3615 struct epoch_tracker et; 3616 struct ifnet *oifp; 3617 3618 NET_EPOCH_ENTER(et); 3619 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3620 if (ifp == oifp) 3621 break; 3622 NET_EPOCH_EXIT(et); 3623 if (ifp != oifp) 3624 ifp = NULL; 3625 } 3626 #endif 3627 /* 3628 * If and only if the ifnet instance exists: Acquire the address lock. 3629 */ 3630 if (ifp != NULL) 3631 IF_ADDR_WLOCK(ifp); 3632 3633 lastref = if_delmulti_locked(ifp, ifma, flags); 3634 3635 if (ifp != NULL) { 3636 /* 3637 * If and only if the ifnet instance exists: 3638 * Release the address lock. 3639 * If the group was left: update the hardware hash filter. 3640 */ 3641 IF_ADDR_WUNLOCK(ifp); 3642 if (lastref && ifp->if_ioctl != NULL) { 3643 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3644 } 3645 } 3646 } 3647 3648 /* 3649 * Perform deletion of network-layer and/or link-layer multicast address. 3650 * 3651 * Return 0 if the reference count was decremented. 3652 * Return 1 if the final reference was released, indicating that the 3653 * hardware hash filter should be reprogrammed. 3654 */ 3655 static int 3656 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3657 { 3658 struct ifmultiaddr *ll_ifma; 3659 3660 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3661 KASSERT(ifma->ifma_ifp == ifp, 3662 ("%s: inconsistent ifp %p", __func__, ifp)); 3663 IF_ADDR_WLOCK_ASSERT(ifp); 3664 } 3665 3666 ifp = ifma->ifma_ifp; 3667 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3668 3669 /* 3670 * If the ifnet is detaching, null out references to ifnet, 3671 * so that upper protocol layers will notice, and not attempt 3672 * to obtain locks for an ifnet which no longer exists. The 3673 * routing socket announcement must happen before the ifnet 3674 * instance is detached from the system. 3675 */ 3676 if (detaching) { 3677 #ifdef DIAGNOSTIC 3678 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3679 #endif 3680 /* 3681 * ifp may already be nulled out if we are being reentered 3682 * to delete the ll_ifma. 3683 */ 3684 if (ifp != NULL) { 3685 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3686 ifma->ifma_ifp = NULL; 3687 } 3688 } 3689 3690 if (--ifma->ifma_refcount > 0) 3691 return 0; 3692 3693 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3694 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3695 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3696 } 3697 /* 3698 * If this ifma is a network-layer ifma, a link-layer ifma may 3699 * have been associated with it. Release it first if so. 3700 */ 3701 ll_ifma = ifma->ifma_llifma; 3702 if (ll_ifma != NULL) { 3703 KASSERT(ifma->ifma_lladdr != NULL, 3704 ("%s: llifma w/o lladdr", __func__)); 3705 if (detaching) 3706 ll_ifma->ifma_ifp = NULL; /* XXX */ 3707 if (--ll_ifma->ifma_refcount == 0) { 3708 if (ifp != NULL) { 3709 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3710 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3711 ifma_link); 3712 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3713 } 3714 } 3715 if_freemulti(ll_ifma); 3716 } 3717 } 3718 #ifdef INVARIANTS 3719 if (ifp) { 3720 struct ifmultiaddr *ifmatmp; 3721 3722 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3723 MPASS(ifma != ifmatmp); 3724 } 3725 #endif 3726 if_freemulti(ifma); 3727 /* 3728 * The last reference to this instance of struct ifmultiaddr 3729 * was released; the hardware should be notified of this change. 3730 */ 3731 return 1; 3732 } 3733 3734 /* 3735 * Set the link layer address on an interface. 3736 * 3737 * At this time we only support certain types of interfaces, 3738 * and we don't allow the length of the address to change. 3739 * 3740 * Set noinline to be dtrace-friendly 3741 */ 3742 __noinline int 3743 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3744 { 3745 struct sockaddr_dl *sdl; 3746 struct ifaddr *ifa; 3747 struct ifreq ifr; 3748 3749 ifa = ifp->if_addr; 3750 if (ifa == NULL) 3751 return (EINVAL); 3752 3753 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3754 if (sdl == NULL) 3755 return (EINVAL); 3756 3757 if (len != sdl->sdl_alen) /* don't allow length to change */ 3758 return (EINVAL); 3759 3760 switch (ifp->if_type) { 3761 case IFT_ETHER: 3762 case IFT_XETHER: 3763 case IFT_L2VLAN: 3764 case IFT_BRIDGE: 3765 case IFT_IEEE8023ADLAG: 3766 bcopy(lladdr, LLADDR(sdl), len); 3767 break; 3768 default: 3769 return (ENODEV); 3770 } 3771 3772 /* 3773 * If the interface is already up, we need 3774 * to re-init it in order to reprogram its 3775 * address filter. 3776 */ 3777 if ((ifp->if_flags & IFF_UP) != 0) { 3778 if (ifp->if_ioctl) { 3779 ifp->if_flags &= ~IFF_UP; 3780 ifr.ifr_flags = ifp->if_flags & 0xffff; 3781 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3782 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3783 ifp->if_flags |= IFF_UP; 3784 ifr.ifr_flags = ifp->if_flags & 0xffff; 3785 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3786 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3787 } 3788 } 3789 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3790 3791 return (0); 3792 } 3793 3794 /* 3795 * Compat function for handling basic encapsulation requests. 3796 * Not converted stacks (FDDI, IB, ..) supports traditional 3797 * output model: ARP (and other similar L2 protocols) are handled 3798 * inside output routine, arpresolve/nd6_resolve() returns MAC 3799 * address instead of full prepend. 3800 * 3801 * This function creates calculated header==MAC for IPv4/IPv6 and 3802 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3803 * address families. 3804 */ 3805 static int 3806 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3807 { 3808 if (req->rtype != IFENCAP_LL) 3809 return (EOPNOTSUPP); 3810 3811 if (req->bufsize < req->lladdr_len) 3812 return (ENOMEM); 3813 3814 switch (req->family) { 3815 case AF_INET: 3816 case AF_INET6: 3817 break; 3818 default: 3819 return (EAFNOSUPPORT); 3820 } 3821 3822 /* Copy lladdr to storage as is */ 3823 memmove(req->buf, req->lladdr, req->lladdr_len); 3824 req->bufsize = req->lladdr_len; 3825 req->lladdr_off = 0; 3826 3827 return (0); 3828 } 3829 3830 /* 3831 * Tunnel interfaces can nest, also they may cause infinite recursion 3832 * calls when misconfigured. We'll prevent this by detecting loops. 3833 * High nesting level may cause stack exhaustion. We'll prevent this 3834 * by introducing upper limit. 3835 * 3836 * Return 0, if tunnel nesting count is equal or less than limit. 3837 */ 3838 int 3839 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3840 int limit) 3841 { 3842 struct m_tag *mtag; 3843 int count; 3844 3845 count = 1; 3846 mtag = NULL; 3847 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3848 if (*(struct ifnet **)(mtag + 1) == ifp) { 3849 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3850 return (EIO); 3851 } 3852 count++; 3853 } 3854 if (count > limit) { 3855 log(LOG_NOTICE, 3856 "%s: if_output recursively called too many times(%d)\n", 3857 if_name(ifp), count); 3858 return (EIO); 3859 } 3860 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3861 if (mtag == NULL) 3862 return (ENOMEM); 3863 *(struct ifnet **)(mtag + 1) = ifp; 3864 m_tag_prepend(m, mtag); 3865 return (0); 3866 } 3867 3868 /* 3869 * Get the link layer address that was read from the hardware at attach. 3870 * 3871 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3872 * their component interfaces as IFT_IEEE8023ADLAG. 3873 */ 3874 int 3875 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3876 { 3877 if (ifp->if_hw_addr == NULL) 3878 return (ENODEV); 3879 3880 switch (ifp->if_type) { 3881 case IFT_ETHER: 3882 case IFT_IEEE8023ADLAG: 3883 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3884 return (0); 3885 default: 3886 return (ENODEV); 3887 } 3888 } 3889 3890 /* 3891 * The name argument must be a pointer to storage which will last as 3892 * long as the interface does. For physical devices, the result of 3893 * device_get_name(dev) is a good choice and for pseudo-devices a 3894 * static string works well. 3895 */ 3896 void 3897 if_initname(struct ifnet *ifp, const char *name, int unit) 3898 { 3899 ifp->if_dname = name; 3900 ifp->if_dunit = unit; 3901 if (unit != IF_DUNIT_NONE) 3902 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3903 else 3904 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3905 } 3906 3907 static int 3908 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 3909 { 3910 char if_fmt[256]; 3911 3912 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 3913 vlog(pri, if_fmt, ap); 3914 return (0); 3915 } 3916 3917 3918 int 3919 if_printf(struct ifnet *ifp, const char *fmt, ...) 3920 { 3921 va_list ap; 3922 3923 va_start(ap, fmt); 3924 if_vlog(ifp, LOG_INFO, fmt, ap); 3925 va_end(ap); 3926 return (0); 3927 } 3928 3929 int 3930 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 3931 { 3932 va_list ap; 3933 3934 va_start(ap, fmt); 3935 if_vlog(ifp, pri, fmt, ap); 3936 va_end(ap); 3937 return (0); 3938 } 3939 3940 void 3941 if_start(struct ifnet *ifp) 3942 { 3943 3944 (*(ifp)->if_start)(ifp); 3945 } 3946 3947 /* 3948 * Backwards compatibility interface for drivers 3949 * that have not implemented it 3950 */ 3951 static int 3952 if_transmit_default(struct ifnet *ifp, struct mbuf *m) 3953 { 3954 int error; 3955 3956 IFQ_HANDOFF(ifp, m, error); 3957 return (error); 3958 } 3959 3960 static void 3961 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3962 { 3963 m_freem(m); 3964 } 3965 3966 int 3967 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 3968 { 3969 int active = 0; 3970 3971 IF_LOCK(ifq); 3972 if (_IF_QFULL(ifq)) { 3973 IF_UNLOCK(ifq); 3974 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 3975 m_freem(m); 3976 return (0); 3977 } 3978 if (ifp != NULL) { 3979 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 3980 if (m->m_flags & (M_BCAST|M_MCAST)) 3981 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 3982 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 3983 } 3984 _IF_ENQUEUE(ifq, m); 3985 IF_UNLOCK(ifq); 3986 if (ifp != NULL && !active) 3987 (*(ifp)->if_start)(ifp); 3988 return (1); 3989 } 3990 3991 void 3992 if_register_com_alloc(u_char type, 3993 if_com_alloc_t *a, if_com_free_t *f) 3994 { 3995 3996 KASSERT(if_com_alloc[type] == NULL, 3997 ("if_register_com_alloc: %d already registered", type)); 3998 KASSERT(if_com_free[type] == NULL, 3999 ("if_register_com_alloc: %d free already registered", type)); 4000 4001 if_com_alloc[type] = a; 4002 if_com_free[type] = f; 4003 } 4004 4005 void 4006 if_deregister_com_alloc(u_char type) 4007 { 4008 4009 KASSERT(if_com_alloc[type] != NULL, 4010 ("if_deregister_com_alloc: %d not registered", type)); 4011 KASSERT(if_com_free[type] != NULL, 4012 ("if_deregister_com_alloc: %d free not registered", type)); 4013 4014 /* 4015 * Ensure all pending EPOCH(9) callbacks have been executed. This 4016 * fixes issues about late invocation of if_destroy(), which leads 4017 * to memory leak from if_com_alloc[type] allocated if_l2com. 4018 */ 4019 NET_EPOCH_DRAIN_CALLBACKS(); 4020 4021 if_com_alloc[type] = NULL; 4022 if_com_free[type] = NULL; 4023 } 4024 4025 /* API for driver access to network stack owned ifnet.*/ 4026 uint64_t 4027 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4028 { 4029 uint64_t oldbrate; 4030 4031 oldbrate = ifp->if_baudrate; 4032 ifp->if_baudrate = baudrate; 4033 return (oldbrate); 4034 } 4035 4036 uint64_t 4037 if_getbaudrate(const if_t ifp) 4038 { 4039 return (ifp->if_baudrate); 4040 } 4041 4042 int 4043 if_setcapabilities(if_t ifp, int capabilities) 4044 { 4045 ifp->if_capabilities = capabilities; 4046 return (0); 4047 } 4048 4049 int 4050 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4051 { 4052 ifp->if_capabilities &= ~clearbit; 4053 ifp->if_capabilities |= setbit; 4054 return (0); 4055 } 4056 4057 int 4058 if_getcapabilities(const if_t ifp) 4059 { 4060 return (ifp->if_capabilities); 4061 } 4062 4063 int 4064 if_setcapenable(if_t ifp, int capabilities) 4065 { 4066 ifp->if_capenable = capabilities; 4067 return (0); 4068 } 4069 4070 int 4071 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4072 { 4073 ifp->if_capenable &= ~clearcap; 4074 ifp->if_capenable |= setcap; 4075 return (0); 4076 } 4077 4078 int 4079 if_setcapabilities2(if_t ifp, int capabilities) 4080 { 4081 ifp->if_capabilities2 = capabilities; 4082 return (0); 4083 } 4084 4085 int 4086 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit) 4087 { 4088 ifp->if_capabilities2 &= ~clearbit; 4089 ifp->if_capabilities2 |= setbit; 4090 return (0); 4091 } 4092 4093 int 4094 if_getcapabilities2(const if_t ifp) 4095 { 4096 return (ifp->if_capabilities2); 4097 } 4098 4099 int 4100 if_setcapenable2(if_t ifp, int capabilities2) 4101 { 4102 ifp->if_capenable2 = capabilities2; 4103 return (0); 4104 } 4105 4106 int 4107 if_setcapenable2bit(if_t ifp, int setcap, int clearcap) 4108 { 4109 ifp->if_capenable2 &= ~clearcap; 4110 ifp->if_capenable2 |= setcap; 4111 return (0); 4112 } 4113 4114 const char * 4115 if_getdname(const if_t ifp) 4116 { 4117 return (ifp->if_dname); 4118 } 4119 4120 void 4121 if_setdname(if_t ifp, const char *dname) 4122 { 4123 ifp->if_dname = dname; 4124 } 4125 4126 const char * 4127 if_name(if_t ifp) 4128 { 4129 return (ifp->if_xname); 4130 } 4131 4132 int 4133 if_setname(if_t ifp, const char *name) 4134 { 4135 if (strlen(name) > sizeof(ifp->if_xname) - 1) 4136 return (ENAMETOOLONG); 4137 strcpy(ifp->if_xname, name); 4138 4139 return (0); 4140 } 4141 4142 int 4143 if_togglecapenable(if_t ifp, int togglecap) 4144 { 4145 ifp->if_capenable ^= togglecap; 4146 return (0); 4147 } 4148 4149 int 4150 if_getcapenable(const if_t ifp) 4151 { 4152 return (ifp->if_capenable); 4153 } 4154 4155 int 4156 if_togglecapenable2(if_t ifp, int togglecap) 4157 { 4158 ifp->if_capenable2 ^= togglecap; 4159 return (0); 4160 } 4161 4162 int 4163 if_getcapenable2(const if_t ifp) 4164 { 4165 return (ifp->if_capenable2); 4166 } 4167 4168 int 4169 if_getdunit(const if_t ifp) 4170 { 4171 return (ifp->if_dunit); 4172 } 4173 4174 int 4175 if_getindex(const if_t ifp) 4176 { 4177 return (ifp->if_index); 4178 } 4179 4180 int 4181 if_getidxgen(const if_t ifp) 4182 { 4183 return (ifp->if_idxgen); 4184 } 4185 4186 const char * 4187 if_getdescr(if_t ifp) 4188 { 4189 return (ifp->if_description); 4190 } 4191 4192 void 4193 if_setdescr(if_t ifp, char *descrbuf) 4194 { 4195 sx_xlock(&ifdescr_sx); 4196 char *odescrbuf = ifp->if_description; 4197 ifp->if_description = descrbuf; 4198 sx_xunlock(&ifdescr_sx); 4199 4200 if_freedescr(odescrbuf); 4201 } 4202 4203 char * 4204 if_allocdescr(size_t sz, int malloc_flag) 4205 { 4206 malloc_flag &= (M_WAITOK | M_NOWAIT); 4207 return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag)); 4208 } 4209 4210 void 4211 if_freedescr(char *descrbuf) 4212 { 4213 free(descrbuf, M_IFDESCR); 4214 } 4215 4216 int 4217 if_getalloctype(const if_t ifp) 4218 { 4219 return (ifp->if_alloctype); 4220 } 4221 4222 void 4223 if_setlastchange(if_t ifp) 4224 { 4225 getmicrotime(&ifp->if_lastchange); 4226 } 4227 4228 /* 4229 * This is largely undesirable because it ties ifnet to a device, but does 4230 * provide flexiblity for an embedded product vendor. Should be used with 4231 * the understanding that it violates the interface boundaries, and should be 4232 * a last resort only. 4233 */ 4234 int 4235 if_setdev(if_t ifp, void *dev) 4236 { 4237 return (0); 4238 } 4239 4240 int 4241 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4242 { 4243 ifp->if_drv_flags &= ~clear_flags; 4244 ifp->if_drv_flags |= set_flags; 4245 4246 return (0); 4247 } 4248 4249 int 4250 if_getdrvflags(const if_t ifp) 4251 { 4252 return (ifp->if_drv_flags); 4253 } 4254 4255 int 4256 if_setdrvflags(if_t ifp, int flags) 4257 { 4258 ifp->if_drv_flags = flags; 4259 return (0); 4260 } 4261 4262 int 4263 if_setflags(if_t ifp, int flags) 4264 { 4265 ifp->if_flags = flags; 4266 return (0); 4267 } 4268 4269 int 4270 if_setflagbits(if_t ifp, int set, int clear) 4271 { 4272 ifp->if_flags &= ~clear; 4273 ifp->if_flags |= set; 4274 return (0); 4275 } 4276 4277 int 4278 if_getflags(const if_t ifp) 4279 { 4280 return (ifp->if_flags); 4281 } 4282 4283 int 4284 if_clearhwassist(if_t ifp) 4285 { 4286 ifp->if_hwassist = 0; 4287 return (0); 4288 } 4289 4290 int 4291 if_sethwassistbits(if_t ifp, int toset, int toclear) 4292 { 4293 ifp->if_hwassist &= ~toclear; 4294 ifp->if_hwassist |= toset; 4295 4296 return (0); 4297 } 4298 4299 int 4300 if_sethwassist(if_t ifp, int hwassist_bit) 4301 { 4302 ifp->if_hwassist = hwassist_bit; 4303 return (0); 4304 } 4305 4306 int 4307 if_gethwassist(const if_t ifp) 4308 { 4309 return (ifp->if_hwassist); 4310 } 4311 4312 int 4313 if_togglehwassist(if_t ifp, int toggle_bits) 4314 { 4315 ifp->if_hwassist ^= toggle_bits; 4316 return (0); 4317 } 4318 4319 int 4320 if_setmtu(if_t ifp, int mtu) 4321 { 4322 ifp->if_mtu = mtu; 4323 return (0); 4324 } 4325 4326 void 4327 if_notifymtu(if_t ifp) 4328 { 4329 #ifdef INET6 4330 nd6_setmtu(ifp); 4331 #endif 4332 rt_updatemtu(ifp); 4333 } 4334 4335 int 4336 if_getmtu(const if_t ifp) 4337 { 4338 return (ifp->if_mtu); 4339 } 4340 4341 void 4342 if_setppromisc(if_t ifp, bool ppromisc) 4343 { 4344 int new_flags; 4345 4346 if (ppromisc) 4347 new_flags = ifp->if_flags | IFF_PPROMISC; 4348 else 4349 new_flags = ifp->if_flags & ~IFF_PPROMISC; 4350 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 4351 if (new_flags & IFF_PPROMISC) 4352 new_flags |= IFF_PROMISC; 4353 /* 4354 * Only unset IFF_PROMISC if there are no more consumers of 4355 * promiscuity, i.e. the ifp->if_pcount refcount is 0. 4356 */ 4357 else if (ifp->if_pcount == 0) 4358 new_flags &= ~IFF_PROMISC; 4359 if (log_promisc_mode_change) 4360 if_printf(ifp, "permanently promiscuous mode %s\n", 4361 ((new_flags & IFF_PPROMISC) ? 4362 "enabled" : "disabled")); 4363 } 4364 ifp->if_flags = new_flags; 4365 } 4366 4367 /* 4368 * Methods for drivers to access interface unicast and multicast 4369 * link level addresses. Driver shall not know 'struct ifaddr' neither 4370 * 'struct ifmultiaddr'. 4371 */ 4372 u_int 4373 if_lladdr_count(if_t ifp) 4374 { 4375 struct epoch_tracker et; 4376 struct ifaddr *ifa; 4377 u_int count; 4378 4379 count = 0; 4380 NET_EPOCH_ENTER(et); 4381 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4382 if (ifa->ifa_addr->sa_family == AF_LINK) 4383 count++; 4384 NET_EPOCH_EXIT(et); 4385 4386 return (count); 4387 } 4388 4389 int 4390 if_foreach(if_foreach_cb_t cb, void *cb_arg) 4391 { 4392 if_t ifp; 4393 int error; 4394 4395 NET_EPOCH_ASSERT(); 4396 MPASS(cb); 4397 4398 error = 0; 4399 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4400 error = cb(ifp, cb_arg); 4401 if (error != 0) 4402 break; 4403 } 4404 4405 return (error); 4406 } 4407 4408 /* 4409 * Iterates over the list of interfaces, permitting callback function @cb to sleep. 4410 * Stops iteration if @cb returns non-zero error code. 4411 * Returns the last error code from @cb. 4412 * @match_cb: optional match callback limiting the iteration to only matched interfaces 4413 * @match_arg: argument to pass to @match_cb 4414 * @cb: iteration callback 4415 * @cb_arg: argument to pass to @cb 4416 */ 4417 int 4418 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb, 4419 void *cb_arg) 4420 { 4421 int match_count = 0, array_size = 16; /* 128 bytes for malloc */ 4422 struct ifnet **match_array = NULL; 4423 int error = 0; 4424 4425 MPASS(cb); 4426 4427 while (true) { 4428 struct ifnet **new_array; 4429 int new_size = array_size; 4430 struct epoch_tracker et; 4431 struct ifnet *ifp; 4432 4433 while (new_size < match_count) 4434 new_size *= 2; 4435 new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK); 4436 if (match_array != NULL) 4437 memcpy(new_array, match_array, array_size * sizeof(void *)); 4438 free(match_array, M_TEMP); 4439 match_array = new_array; 4440 array_size = new_size; 4441 4442 match_count = 0; 4443 NET_EPOCH_ENTER(et); 4444 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 4445 if (match_cb != NULL && !match_cb(ifp, match_arg)) 4446 continue; 4447 if (match_count < array_size) { 4448 if (if_try_ref(ifp)) 4449 match_array[match_count++] = ifp; 4450 } else 4451 match_count++; 4452 } 4453 NET_EPOCH_EXIT(et); 4454 4455 if (match_count > array_size) { 4456 for (int i = 0; i < array_size; i++) 4457 if_rele(match_array[i]); 4458 continue; 4459 } else { 4460 for (int i = 0; i < match_count; i++) { 4461 if (error == 0) 4462 error = cb(match_array[i], cb_arg); 4463 if_rele(match_array[i]); 4464 } 4465 free(match_array, M_TEMP); 4466 break; 4467 } 4468 } 4469 4470 return (error); 4471 } 4472 4473 4474 /* 4475 * Uses just 1 pointer of the 4 available in the public struct. 4476 */ 4477 if_t 4478 if_iter_start(struct if_iter *iter) 4479 { 4480 if_t ifp; 4481 4482 NET_EPOCH_ASSERT(); 4483 4484 bzero(iter, sizeof(*iter)); 4485 ifp = CK_STAILQ_FIRST(&V_ifnet); 4486 if (ifp != NULL) 4487 iter->context[0] = CK_STAILQ_NEXT(ifp, if_link); 4488 else 4489 iter->context[0] = NULL; 4490 return (ifp); 4491 } 4492 4493 if_t 4494 if_iter_next(struct if_iter *iter) 4495 { 4496 if_t cur_ifp = iter->context[0]; 4497 4498 if (cur_ifp != NULL) 4499 iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link); 4500 return (cur_ifp); 4501 } 4502 4503 void 4504 if_iter_finish(struct if_iter *iter) 4505 { 4506 /* Nothing to do here for now. */ 4507 } 4508 4509 u_int 4510 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4511 { 4512 struct epoch_tracker et; 4513 struct ifaddr *ifa; 4514 u_int count; 4515 4516 MPASS(cb); 4517 4518 count = 0; 4519 NET_EPOCH_ENTER(et); 4520 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4521 if (ifa->ifa_addr->sa_family != AF_LINK) 4522 continue; 4523 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4524 count); 4525 } 4526 NET_EPOCH_EXIT(et); 4527 4528 return (count); 4529 } 4530 4531 u_int 4532 if_llmaddr_count(if_t ifp) 4533 { 4534 struct epoch_tracker et; 4535 struct ifmultiaddr *ifma; 4536 int count; 4537 4538 count = 0; 4539 NET_EPOCH_ENTER(et); 4540 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4541 if (ifma->ifma_addr->sa_family == AF_LINK) 4542 count++; 4543 NET_EPOCH_EXIT(et); 4544 4545 return (count); 4546 } 4547 4548 bool 4549 if_maddr_empty(if_t ifp) 4550 { 4551 4552 return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs)); 4553 } 4554 4555 u_int 4556 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4557 { 4558 struct epoch_tracker et; 4559 struct ifmultiaddr *ifma; 4560 u_int count; 4561 4562 MPASS(cb); 4563 4564 count = 0; 4565 NET_EPOCH_ENTER(et); 4566 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4567 if (ifma->ifma_addr->sa_family != AF_LINK) 4568 continue; 4569 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4570 count); 4571 } 4572 NET_EPOCH_EXIT(et); 4573 4574 return (count); 4575 } 4576 4577 u_int 4578 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg) 4579 { 4580 struct epoch_tracker et; 4581 struct ifaddr *ifa; 4582 u_int count; 4583 4584 MPASS(cb); 4585 4586 count = 0; 4587 NET_EPOCH_ENTER(et); 4588 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4589 if (ifa->ifa_addr->sa_family != type) 4590 continue; 4591 count += (*cb)(cb_arg, ifa, count); 4592 } 4593 NET_EPOCH_EXIT(et); 4594 4595 return (count); 4596 } 4597 4598 struct ifaddr * 4599 ifa_iter_start(if_t ifp, struct ifa_iter *iter) 4600 { 4601 struct ifaddr *ifa; 4602 4603 NET_EPOCH_ASSERT(); 4604 4605 bzero(iter, sizeof(*iter)); 4606 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 4607 if (ifa != NULL) 4608 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4609 else 4610 iter->context[0] = NULL; 4611 return (ifa); 4612 } 4613 4614 struct ifaddr * 4615 ifa_iter_next(struct ifa_iter *iter) 4616 { 4617 struct ifaddr *ifa = iter->context[0]; 4618 4619 if (ifa != NULL) 4620 iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link); 4621 return (ifa); 4622 } 4623 4624 void 4625 ifa_iter_finish(struct ifa_iter *iter) 4626 { 4627 /* Nothing to do here for now. */ 4628 } 4629 4630 int 4631 if_setsoftc(if_t ifp, void *softc) 4632 { 4633 ifp->if_softc = softc; 4634 return (0); 4635 } 4636 4637 void * 4638 if_getsoftc(const if_t ifp) 4639 { 4640 return (ifp->if_softc); 4641 } 4642 4643 void 4644 if_setrcvif(struct mbuf *m, if_t ifp) 4645 { 4646 4647 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4648 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4649 } 4650 4651 void 4652 if_setvtag(struct mbuf *m, uint16_t tag) 4653 { 4654 m->m_pkthdr.ether_vtag = tag; 4655 } 4656 4657 uint16_t 4658 if_getvtag(struct mbuf *m) 4659 { 4660 return (m->m_pkthdr.ether_vtag); 4661 } 4662 4663 int 4664 if_sendq_empty(if_t ifp) 4665 { 4666 return (IFQ_DRV_IS_EMPTY(&ifp->if_snd)); 4667 } 4668 4669 struct ifaddr * 4670 if_getifaddr(const if_t ifp) 4671 { 4672 return (ifp->if_addr); 4673 } 4674 4675 int 4676 if_setsendqready(if_t ifp) 4677 { 4678 IFQ_SET_READY(&ifp->if_snd); 4679 return (0); 4680 } 4681 4682 int 4683 if_setsendqlen(if_t ifp, int tx_desc_count) 4684 { 4685 IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count); 4686 ifp->if_snd.ifq_drv_maxlen = tx_desc_count; 4687 return (0); 4688 } 4689 4690 void 4691 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na) 4692 { 4693 ifp->if_netmap = na; 4694 } 4695 4696 struct netmap_adapter * 4697 if_getnetmapadapter(if_t ifp) 4698 { 4699 return (ifp->if_netmap); 4700 } 4701 4702 int 4703 if_vlantrunkinuse(if_t ifp) 4704 { 4705 return (ifp->if_vlantrunk != NULL); 4706 } 4707 4708 void 4709 if_init(if_t ifp, void *ctx) 4710 { 4711 (*ifp->if_init)(ctx); 4712 } 4713 4714 void 4715 if_input(if_t ifp, struct mbuf* sendmp) 4716 { 4717 (*ifp->if_input)(ifp, sendmp); 4718 } 4719 4720 int 4721 if_transmit(if_t ifp, struct mbuf *m) 4722 { 4723 return ((*ifp->if_transmit)(ifp, m)); 4724 } 4725 4726 int 4727 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst) 4728 { 4729 if (ifp->if_resolvemulti == NULL) 4730 return (EOPNOTSUPP); 4731 4732 return (ifp->if_resolvemulti(ifp, srcs, dst)); 4733 } 4734 4735 int 4736 if_ioctl(if_t ifp, u_long cmd, void *data) 4737 { 4738 if (ifp->if_ioctl == NULL) 4739 return (EOPNOTSUPP); 4740 4741 return (ifp->if_ioctl(ifp, cmd, data)); 4742 } 4743 4744 struct mbuf * 4745 if_dequeue(if_t ifp) 4746 { 4747 struct mbuf *m; 4748 4749 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 4750 return (m); 4751 } 4752 4753 int 4754 if_sendq_prepend(if_t ifp, struct mbuf *m) 4755 { 4756 IFQ_DRV_PREPEND(&ifp->if_snd, m); 4757 return (0); 4758 } 4759 4760 int 4761 if_setifheaderlen(if_t ifp, int len) 4762 { 4763 ifp->if_hdrlen = len; 4764 return (0); 4765 } 4766 4767 char * 4768 if_getlladdr(const if_t ifp) 4769 { 4770 return (IF_LLADDR(ifp)); 4771 } 4772 4773 void * 4774 if_gethandle(u_char type) 4775 { 4776 return (if_alloc(type)); 4777 } 4778 4779 void 4780 if_vlancap(if_t ifp) 4781 { 4782 VLAN_CAPABILITIES(ifp); 4783 } 4784 4785 int 4786 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4787 { 4788 ifp->if_hw_tsomax = if_hw_tsomax; 4789 return (0); 4790 } 4791 4792 int 4793 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4794 { 4795 ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4796 return (0); 4797 } 4798 4799 int 4800 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4801 { 4802 ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4803 return (0); 4804 } 4805 4806 u_int 4807 if_gethwtsomax(const if_t ifp) 4808 { 4809 return (ifp->if_hw_tsomax); 4810 } 4811 4812 u_int 4813 if_gethwtsomaxsegcount(const if_t ifp) 4814 { 4815 return (ifp->if_hw_tsomaxsegcount); 4816 } 4817 4818 u_int 4819 if_gethwtsomaxsegsize(const if_t ifp) 4820 { 4821 return (ifp->if_hw_tsomaxsegsize); 4822 } 4823 4824 void 4825 if_setinitfn(if_t ifp, if_init_fn_t init_fn) 4826 { 4827 ifp->if_init = init_fn; 4828 } 4829 4830 void 4831 if_setinputfn(if_t ifp, if_input_fn_t input_fn) 4832 { 4833 ifp->if_input = input_fn; 4834 } 4835 4836 if_input_fn_t 4837 if_getinputfn(if_t ifp) 4838 { 4839 return (ifp->if_input); 4840 } 4841 4842 void 4843 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn) 4844 { 4845 ifp->if_ioctl = ioctl_fn; 4846 } 4847 4848 void 4849 if_setoutputfn(if_t ifp, if_output_fn_t output_fn) 4850 { 4851 ifp->if_output = output_fn; 4852 } 4853 4854 void 4855 if_setstartfn(if_t ifp, if_start_fn_t start_fn) 4856 { 4857 ifp->if_start = start_fn; 4858 } 4859 4860 if_start_fn_t 4861 if_getstartfn(if_t ifp) 4862 { 4863 return (ifp->if_start); 4864 } 4865 4866 void 4867 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4868 { 4869 ifp->if_transmit = start_fn; 4870 } 4871 4872 if_transmit_fn_t 4873 if_gettransmitfn(if_t ifp) 4874 { 4875 return (ifp->if_transmit); 4876 } 4877 4878 void 4879 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4880 { 4881 ifp->if_qflush = flush_fn; 4882 } 4883 4884 void 4885 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn) 4886 { 4887 ifp->if_snd_tag_alloc = alloc_fn; 4888 } 4889 4890 int 4891 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params, 4892 struct m_snd_tag **mstp) 4893 { 4894 if (ifp->if_snd_tag_alloc == NULL) 4895 return (EOPNOTSUPP); 4896 return (ifp->if_snd_tag_alloc(ifp, params, mstp)); 4897 } 4898 4899 void 4900 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4901 { 4902 ifp->if_get_counter = fn; 4903 } 4904 4905 void 4906 if_setreassignfn(if_t ifp, if_reassign_fn_t fn) 4907 { 4908 ifp->if_reassign = fn; 4909 } 4910 4911 void 4912 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn) 4913 { 4914 ifp->if_ratelimit_query = fn; 4915 } 4916 4917 void 4918 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m) 4919 { 4920 ifp->if_debugnet_methods = m; 4921 } 4922 4923 struct label * 4924 if_getmaclabel(if_t ifp) 4925 { 4926 return (ifp->if_label); 4927 } 4928 4929 void 4930 if_setmaclabel(if_t ifp, struct label *label) 4931 { 4932 ifp->if_label = label; 4933 } 4934 4935 int 4936 if_gettype(if_t ifp) 4937 { 4938 return (ifp->if_type); 4939 } 4940 4941 void * 4942 if_getllsoftc(if_t ifp) 4943 { 4944 return (ifp->if_llsoftc); 4945 } 4946 4947 void 4948 if_setllsoftc(if_t ifp, void *llsoftc) 4949 { 4950 ifp->if_llsoftc = llsoftc; 4951 }; 4952 4953 int 4954 if_getlinkstate(if_t ifp) 4955 { 4956 return (ifp->if_link_state); 4957 } 4958 4959 const uint8_t * 4960 if_getbroadcastaddr(if_t ifp) 4961 { 4962 return (ifp->if_broadcastaddr); 4963 } 4964 4965 void 4966 if_setbroadcastaddr(if_t ifp, const uint8_t *addr) 4967 { 4968 ifp->if_broadcastaddr = addr; 4969 } 4970 4971 int 4972 if_getnumadomain(if_t ifp) 4973 { 4974 return (ifp->if_numa_domain); 4975 } 4976 4977 uint64_t 4978 if_getcounter(if_t ifp, ift_counter counter) 4979 { 4980 return (ifp->if_get_counter(ifp, counter)); 4981 } 4982 4983 bool 4984 if_altq_is_enabled(if_t ifp) 4985 { 4986 return (ALTQ_IS_ENABLED(&ifp->if_snd)); 4987 } 4988 4989 struct vnet * 4990 if_getvnet(if_t ifp) 4991 { 4992 return (ifp->if_vnet); 4993 } 4994 4995 struct in_ifinfo * 4996 if_getinet(if_t ifp) 4997 { 4998 return (ifp->if_inet); 4999 } 5000 5001 struct in6_ifextra * 5002 if_getinet6(if_t ifp) 5003 { 5004 return (ifp->if_inet6); 5005 } 5006 5007 u_int 5008 if_getfib(if_t ifp) 5009 { 5010 return (ifp->if_fib); 5011 } 5012 5013 uint8_t 5014 if_getaddrlen(if_t ifp) 5015 { 5016 return (ifp->if_addrlen); 5017 } 5018 5019 struct bpf_if * 5020 if_getbpf(if_t ifp) 5021 { 5022 return (ifp->if_bpf); 5023 } 5024 5025 struct ifvlantrunk * 5026 if_getvlantrunk(if_t ifp) 5027 { 5028 return (ifp->if_vlantrunk); 5029 } 5030 5031 uint8_t 5032 if_getpcp(if_t ifp) 5033 { 5034 return (ifp->if_pcp); 5035 } 5036 5037 void * 5038 if_getl2com(if_t ifp) 5039 { 5040 return (ifp->if_l2com); 5041 } 5042 5043 void 5044 if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m) 5045 { 5046 ifp->if_ipsec_accel_m = m; 5047 } 5048 5049 #ifdef DDB 5050 static void 5051 if_show_ifnet(struct ifnet *ifp) 5052 { 5053 if (ifp == NULL) 5054 return; 5055 db_printf("%s:\n", ifp->if_xname); 5056 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 5057 IF_DB_PRINTF("%s", if_dname); 5058 IF_DB_PRINTF("%d", if_dunit); 5059 IF_DB_PRINTF("%s", if_description); 5060 IF_DB_PRINTF("%u", if_index); 5061 IF_DB_PRINTF("%d", if_idxgen); 5062 IF_DB_PRINTF("%u", if_refcount); 5063 IF_DB_PRINTF("%p", if_softc); 5064 IF_DB_PRINTF("%p", if_l2com); 5065 IF_DB_PRINTF("%p", if_llsoftc); 5066 IF_DB_PRINTF("%d", if_amcount); 5067 IF_DB_PRINTF("%p", if_addr); 5068 IF_DB_PRINTF("%p", if_broadcastaddr); 5069 IF_DB_PRINTF("%u", if_fib); 5070 IF_DB_PRINTF("%p", if_vnet); 5071 IF_DB_PRINTF("%p", if_home_vnet); 5072 IF_DB_PRINTF("%p", if_vlantrunk); 5073 IF_DB_PRINTF("%p", if_bpf); 5074 IF_DB_PRINTF("%u", if_pcount); 5075 IF_DB_PRINTF("%p", if_bridge); 5076 IF_DB_PRINTF("%p", if_lagg); 5077 IF_DB_PRINTF("%p", if_pf_kif); 5078 IF_DB_PRINTF("%p", if_carp); 5079 IF_DB_PRINTF("%p", if_label); 5080 IF_DB_PRINTF("%p", if_netmap); 5081 IF_DB_PRINTF("0x%08x", if_flags); 5082 IF_DB_PRINTF("0x%08x", if_drv_flags); 5083 IF_DB_PRINTF("0x%08x", if_capabilities); 5084 IF_DB_PRINTF("0x%08x", if_capenable); 5085 IF_DB_PRINTF("%p", if_snd.ifq_head); 5086 IF_DB_PRINTF("%p", if_snd.ifq_tail); 5087 IF_DB_PRINTF("%d", if_snd.ifq_len); 5088 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 5089 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 5090 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 5091 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 5092 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 5093 IF_DB_PRINTF("%d", if_snd.altq_type); 5094 IF_DB_PRINTF("%x", if_snd.altq_flags); 5095 #undef IF_DB_PRINTF 5096 } 5097 5098 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 5099 { 5100 if (!have_addr) { 5101 db_printf("usage: show ifnet <struct ifnet *>\n"); 5102 return; 5103 } 5104 5105 if_show_ifnet((struct ifnet *)addr); 5106 } 5107 5108 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 5109 { 5110 struct ifnet *ifp; 5111 u_short idx; 5112 5113 for (idx = 1; idx <= if_index; idx++) { 5114 ifp = ifindex_table[idx].ife_ifnet; 5115 if (ifp == NULL) 5116 continue; 5117 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 5118 if (db_pager_quit) 5119 break; 5120 } 5121 } 5122 #endif /* DDB */ 5123