1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1980, 1986, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)if.c 8.5 (Berkeley) 1/9/95 32 * $FreeBSD$ 33 */ 34 35 #include "opt_bpf.h" 36 #include "opt_inet6.h" 37 #include "opt_inet.h" 38 39 #include <sys/param.h> 40 #include <sys/capsicum.h> 41 #include <sys/conf.h> 42 #include <sys/eventhandler.h> 43 #include <sys/malloc.h> 44 #include <sys/domainset.h> 45 #include <sys/sbuf.h> 46 #include <sys/bus.h> 47 #include <sys/epoch.h> 48 #include <sys/mbuf.h> 49 #include <sys/systm.h> 50 #include <sys/priv.h> 51 #include <sys/proc.h> 52 #include <sys/socket.h> 53 #include <sys/socketvar.h> 54 #include <sys/protosw.h> 55 #include <sys/kernel.h> 56 #include <sys/lock.h> 57 #include <sys/refcount.h> 58 #include <sys/module.h> 59 #include <sys/rwlock.h> 60 #include <sys/sockio.h> 61 #include <sys/syslog.h> 62 #include <sys/sysctl.h> 63 #include <sys/sysent.h> 64 #include <sys/taskqueue.h> 65 #include <sys/domain.h> 66 #include <sys/jail.h> 67 #include <sys/priv.h> 68 69 #include <machine/stdarg.h> 70 #include <vm/uma.h> 71 72 #include <net/bpf.h> 73 #include <net/ethernet.h> 74 #include <net/if.h> 75 #include <net/if_arp.h> 76 #include <net/if_clone.h> 77 #include <net/if_dl.h> 78 #include <net/if_types.h> 79 #include <net/if_var.h> 80 #include <net/if_media.h> 81 #include <net/if_vlan_var.h> 82 #include <net/radix.h> 83 #include <net/route.h> 84 #include <net/route/route_ctl.h> 85 #include <net/vnet.h> 86 87 #if defined(INET) || defined(INET6) 88 #include <net/ethernet.h> 89 #include <netinet/in.h> 90 #include <netinet/in_var.h> 91 #include <netinet/ip.h> 92 #include <netinet/ip_carp.h> 93 #ifdef INET 94 #include <net/debugnet.h> 95 #include <netinet/if_ether.h> 96 #endif /* INET */ 97 #ifdef INET6 98 #include <netinet6/in6_var.h> 99 #include <netinet6/in6_ifattach.h> 100 #endif /* INET6 */ 101 #endif /* INET || INET6 */ 102 103 #include <security/mac/mac_framework.h> 104 105 /* 106 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 107 * and ifr_ifru when it is used in SIOCGIFCONF. 108 */ 109 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 110 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 111 112 __read_mostly epoch_t net_epoch_preempt; 113 #ifdef COMPAT_FREEBSD32 114 #include <sys/mount.h> 115 #include <compat/freebsd32/freebsd32.h> 116 117 struct ifreq_buffer32 { 118 uint32_t length; /* (size_t) */ 119 uint32_t buffer; /* (void *) */ 120 }; 121 122 /* 123 * Interface request structure used for socket 124 * ioctl's. All interface ioctl's must have parameter 125 * definitions which begin with ifr_name. The 126 * remainder may be interface specific. 127 */ 128 struct ifreq32 { 129 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 130 union { 131 struct sockaddr ifru_addr; 132 struct sockaddr ifru_dstaddr; 133 struct sockaddr ifru_broadaddr; 134 struct ifreq_buffer32 ifru_buffer; 135 short ifru_flags[2]; 136 short ifru_index; 137 int ifru_jid; 138 int ifru_metric; 139 int ifru_mtu; 140 int ifru_phys; 141 int ifru_media; 142 uint32_t ifru_data; 143 int ifru_cap[2]; 144 u_int ifru_fib; 145 u_char ifru_vlan_pcp; 146 } ifr_ifru; 147 }; 148 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 149 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 150 __offsetof(struct ifreq32, ifr_ifru)); 151 152 struct ifconf32 { 153 int32_t ifc_len; 154 union { 155 uint32_t ifcu_buf; 156 uint32_t ifcu_req; 157 } ifc_ifcu; 158 }; 159 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 160 161 struct ifdrv32 { 162 char ifd_name[IFNAMSIZ]; 163 uint32_t ifd_cmd; 164 uint32_t ifd_len; 165 uint32_t ifd_data; 166 }; 167 #define SIOCSDRVSPEC32 _IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32) 168 #define SIOCGDRVSPEC32 _IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32) 169 170 struct ifgroupreq32 { 171 char ifgr_name[IFNAMSIZ]; 172 u_int ifgr_len; 173 union { 174 char ifgru_group[IFNAMSIZ]; 175 uint32_t ifgru_groups; 176 } ifgr_ifgru; 177 }; 178 #define SIOCAIFGROUP32 _IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32) 179 #define SIOCGIFGROUP32 _IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32) 180 #define SIOCDIFGROUP32 _IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32) 181 #define SIOCGIFGMEMB32 _IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32) 182 183 struct ifmediareq32 { 184 char ifm_name[IFNAMSIZ]; 185 int ifm_current; 186 int ifm_mask; 187 int ifm_status; 188 int ifm_active; 189 int ifm_count; 190 uint32_t ifm_ulist; /* (int *) */ 191 }; 192 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 193 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 194 #endif /* COMPAT_FREEBSD32 */ 195 196 union ifreq_union { 197 struct ifreq ifr; 198 #ifdef COMPAT_FREEBSD32 199 struct ifreq32 ifr32; 200 #endif 201 }; 202 203 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 204 "Link layers"); 205 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 206 "Generic link-management"); 207 208 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 209 &ifqmaxlen, 0, "max send queue size"); 210 211 /* Log link state change events */ 212 static int log_link_state_change = 1; 213 214 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 215 &log_link_state_change, 0, 216 "log interface link state change events"); 217 218 /* Log promiscuous mode change events */ 219 static int log_promisc_mode_change = 1; 220 221 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 222 &log_promisc_mode_change, 1, 223 "log promiscuous mode change events"); 224 225 /* Interface description */ 226 static unsigned int ifdescr_maxlen = 1024; 227 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 228 &ifdescr_maxlen, 0, 229 "administrative maximum length for interface description"); 230 231 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 232 233 /* global sx for non-critical path ifdescr */ 234 static struct sx ifdescr_sx; 235 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 236 237 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 238 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 239 /* These are external hooks for CARP. */ 240 void (*carp_linkstate_p)(struct ifnet *ifp); 241 void (*carp_demote_adj_p)(int, char *); 242 int (*carp_master_p)(struct ifaddr *); 243 #if defined(INET) || defined(INET6) 244 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 245 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 246 const struct sockaddr *sa); 247 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 248 int (*carp_attach_p)(struct ifaddr *, int); 249 void (*carp_detach_p)(struct ifaddr *, bool); 250 #endif 251 #ifdef INET 252 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 253 #endif 254 #ifdef INET6 255 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 256 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 257 const struct in6_addr *taddr); 258 #endif 259 260 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 261 262 /* 263 * XXX: Style; these should be sorted alphabetically, and unprototyped 264 * static functions should be prototyped. Currently they are sorted by 265 * declaration order. 266 */ 267 static void if_attachdomain(void *); 268 static void if_attachdomain1(struct ifnet *); 269 static int ifconf(u_long, caddr_t); 270 static void *if_grow(void); 271 static void if_input_default(struct ifnet *, struct mbuf *); 272 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 273 static void if_route(struct ifnet *, int flag, int fam); 274 static int if_setflag(struct ifnet *, int, int, int *, int); 275 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 276 static void if_unroute(struct ifnet *, int flag, int fam); 277 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 278 static void do_link_state_change(void *, int); 279 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 280 static int if_getgroupmembers(struct ifgroupreq *); 281 static void if_delgroups(struct ifnet *); 282 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 283 static int if_detach_internal(struct ifnet *, int, struct if_clone **); 284 static void if_siocaddmulti(void *, int); 285 static void if_link_ifnet(struct ifnet *); 286 static bool if_unlink_ifnet(struct ifnet *, bool); 287 #ifdef VIMAGE 288 static int if_vmove(struct ifnet *, struct vnet *); 289 #endif 290 291 #ifdef INET6 292 /* 293 * XXX: declare here to avoid to include many inet6 related files.. 294 * should be more generalized? 295 */ 296 extern void nd6_setmtu(struct ifnet *); 297 #endif 298 299 /* ipsec helper hooks */ 300 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 301 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 302 303 VNET_DEFINE(int, if_index); 304 int ifqmaxlen = IFQ_MAXLEN; 305 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 306 VNET_DEFINE(struct ifgrouphead, ifg_head); 307 308 VNET_DEFINE_STATIC(int, if_indexlim) = 8; 309 310 /* Table of ifnet by index. */ 311 VNET_DEFINE(struct ifnet **, ifindex_table); 312 313 #define V_if_indexlim VNET(if_indexlim) 314 #define V_ifindex_table VNET(ifindex_table) 315 316 /* 317 * The global network interface list (V_ifnet) and related state (such as 318 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 319 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 320 */ 321 struct sx ifnet_sxlock; 322 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 323 324 struct sx ifnet_detach_sxlock; 325 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 326 SX_RECURSE); 327 328 /* 329 * The allocation of network interfaces is a rather non-atomic affair; we 330 * need to select an index before we are ready to expose the interface for 331 * use, so will use this pointer value to indicate reservation. 332 */ 333 #define IFNET_HOLD (void *)(uintptr_t)(-1) 334 335 #ifdef VIMAGE 336 #define VNET_IS_SHUTTING_DOWN(_vnet) \ 337 ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE) 338 #endif 339 340 static if_com_alloc_t *if_com_alloc[256]; 341 static if_com_free_t *if_com_free[256]; 342 343 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 344 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 345 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 346 347 struct ifnet * 348 ifnet_byindex(u_short idx) 349 { 350 struct ifnet *ifp; 351 352 if (__predict_false(idx > V_if_index)) 353 return (NULL); 354 355 ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx); 356 return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp); 357 } 358 359 struct ifnet * 360 ifnet_byindex_ref(u_short idx) 361 { 362 struct ifnet *ifp; 363 364 NET_EPOCH_ASSERT(); 365 366 ifp = ifnet_byindex(idx); 367 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 368 return (NULL); 369 if (!if_try_ref(ifp)) 370 return (NULL); 371 return (ifp); 372 } 373 374 /* 375 * Allocate an ifindex array entry; return 0 on success or an error on 376 * failure. 377 */ 378 static u_short 379 ifindex_alloc(void **old) 380 { 381 u_short idx; 382 383 IFNET_WLOCK_ASSERT(); 384 /* 385 * Try to find an empty slot below V_if_index. If we fail, take the 386 * next slot. 387 */ 388 for (idx = 1; idx <= V_if_index; idx++) { 389 if (V_ifindex_table[idx] == NULL) 390 break; 391 } 392 393 /* Catch if_index overflow. */ 394 if (idx >= V_if_indexlim) { 395 *old = if_grow(); 396 return (USHRT_MAX); 397 } 398 if (idx > V_if_index) 399 V_if_index = idx; 400 return (idx); 401 } 402 403 static void 404 ifindex_free_locked(u_short idx) 405 { 406 407 IFNET_WLOCK_ASSERT(); 408 409 V_ifindex_table[idx] = NULL; 410 while (V_if_index > 0 && 411 V_ifindex_table[V_if_index] == NULL) 412 V_if_index--; 413 } 414 415 static void 416 ifindex_free(u_short idx) 417 { 418 419 IFNET_WLOCK(); 420 ifindex_free_locked(idx); 421 IFNET_WUNLOCK(); 422 } 423 424 static void 425 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 426 { 427 428 V_ifindex_table[idx] = ifp; 429 } 430 431 struct ifaddr * 432 ifaddr_byindex(u_short idx) 433 { 434 struct ifnet *ifp; 435 struct ifaddr *ifa = NULL; 436 437 NET_EPOCH_ASSERT(); 438 439 ifp = ifnet_byindex(idx); 440 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 441 ifa_ref(ifa); 442 return (ifa); 443 } 444 445 /* 446 * Network interface utility routines. 447 * 448 * Routines with ifa_ifwith* names take sockaddr *'s as 449 * parameters. 450 */ 451 452 static void 453 vnet_if_init(const void *unused __unused) 454 { 455 void *old; 456 457 CK_STAILQ_INIT(&V_ifnet); 458 CK_STAILQ_INIT(&V_ifg_head); 459 IFNET_WLOCK(); 460 old = if_grow(); /* create initial table */ 461 IFNET_WUNLOCK(); 462 epoch_wait_preempt(net_epoch_preempt); 463 free(old, M_IFNET); 464 vnet_if_clone_init(); 465 } 466 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 467 NULL); 468 469 #ifdef VIMAGE 470 static void 471 vnet_if_uninit(const void *unused __unused) 472 { 473 474 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 475 "not empty", __func__, __LINE__, &V_ifnet)); 476 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 477 "not empty", __func__, __LINE__, &V_ifg_head)); 478 479 free((caddr_t)V_ifindex_table, M_IFNET); 480 } 481 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 482 vnet_if_uninit, NULL); 483 #endif 484 485 static void 486 if_link_ifnet(struct ifnet *ifp) 487 { 488 489 IFNET_WLOCK(); 490 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 491 #ifdef VIMAGE 492 curvnet->vnet_ifcnt++; 493 #endif 494 IFNET_WUNLOCK(); 495 } 496 497 static bool 498 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 499 { 500 struct ifnet *iter; 501 int found = 0; 502 503 IFNET_WLOCK(); 504 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 505 if (iter == ifp) { 506 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 507 if (!vmove) 508 ifp->if_flags |= IFF_DYING; 509 found = 1; 510 break; 511 } 512 #ifdef VIMAGE 513 curvnet->vnet_ifcnt--; 514 #endif 515 IFNET_WUNLOCK(); 516 517 return (found); 518 } 519 520 #ifdef VIMAGE 521 static void 522 vnet_if_return(const void *unused __unused) 523 { 524 struct ifnet *ifp, *nifp; 525 struct ifnet **pending; 526 int found, i; 527 528 i = 0; 529 530 /* 531 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 532 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 533 * if_detach_internal(), which waits for NET_EPOCH callbacks to 534 * complete. We can't do that from within NET_EPOCH. 535 * 536 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 537 * read/write lock. We cannot hold the lock as we call if_vmove() 538 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 539 * ctx lock. 540 */ 541 IFNET_WLOCK(); 542 543 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 544 M_IFNET, M_WAITOK | M_ZERO); 545 546 /* Return all inherited interfaces to their parent vnets. */ 547 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 548 if (ifp->if_home_vnet != ifp->if_vnet) { 549 found = if_unlink_ifnet(ifp, true); 550 MPASS(found); 551 552 pending[i++] = ifp; 553 } 554 } 555 IFNET_WUNLOCK(); 556 557 for (int j = 0; j < i; j++) { 558 if_vmove(pending[j], pending[j]->if_home_vnet); 559 } 560 561 free(pending, M_IFNET); 562 } 563 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 564 vnet_if_return, NULL); 565 #endif 566 567 static void * 568 if_grow(void) 569 { 570 int oldlim; 571 u_int n; 572 struct ifnet **e; 573 void *old; 574 575 old = NULL; 576 IFNET_WLOCK_ASSERT(); 577 oldlim = V_if_indexlim; 578 IFNET_WUNLOCK(); 579 n = (oldlim << 1) * sizeof(*e); 580 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 581 IFNET_WLOCK(); 582 if (V_if_indexlim != oldlim) { 583 free(e, M_IFNET); 584 return (NULL); 585 } 586 if (V_ifindex_table != NULL) { 587 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 588 old = V_ifindex_table; 589 } 590 V_if_indexlim <<= 1; 591 V_ifindex_table = e; 592 return (old); 593 } 594 595 /* 596 * Allocate a struct ifnet and an index for an interface. A layer 2 597 * common structure will also be allocated if an allocation routine is 598 * registered for the passed type. 599 */ 600 struct ifnet * 601 if_alloc_domain(u_char type, int numa_domain) 602 { 603 struct ifnet *ifp; 604 u_short idx; 605 void *old; 606 607 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 608 if (numa_domain == IF_NODOM) 609 ifp = malloc(sizeof(struct ifnet), M_IFNET, 610 M_WAITOK | M_ZERO); 611 else 612 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 613 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 614 restart: 615 IFNET_WLOCK(); 616 idx = ifindex_alloc(&old); 617 if (__predict_false(idx == USHRT_MAX)) { 618 IFNET_WUNLOCK(); 619 epoch_wait_preempt(net_epoch_preempt); 620 free(old, M_IFNET); 621 goto restart; 622 } 623 ifnet_setbyindex(idx, IFNET_HOLD); 624 IFNET_WUNLOCK(); 625 ifp->if_index = idx; 626 ifp->if_type = type; 627 ifp->if_alloctype = type; 628 ifp->if_numa_domain = numa_domain; 629 #ifdef VIMAGE 630 ifp->if_vnet = curvnet; 631 #endif 632 if (if_com_alloc[type] != NULL) { 633 ifp->if_l2com = if_com_alloc[type](type, ifp); 634 if (ifp->if_l2com == NULL) { 635 free(ifp, M_IFNET); 636 ifindex_free(idx); 637 return (NULL); 638 } 639 } 640 641 IF_ADDR_LOCK_INIT(ifp); 642 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 643 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 644 ifp->if_afdata_initialized = 0; 645 IF_AFDATA_LOCK_INIT(ifp); 646 CK_STAILQ_INIT(&ifp->if_addrhead); 647 CK_STAILQ_INIT(&ifp->if_multiaddrs); 648 CK_STAILQ_INIT(&ifp->if_groups); 649 #ifdef MAC 650 mac_ifnet_init(ifp); 651 #endif 652 ifq_init(&ifp->if_snd, ifp); 653 654 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 655 for (int i = 0; i < IFCOUNTERS; i++) 656 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 657 ifp->if_get_counter = if_get_counter_default; 658 ifp->if_pcp = IFNET_PCP_NONE; 659 ifnet_setbyindex(ifp->if_index, ifp); 660 return (ifp); 661 } 662 663 struct ifnet * 664 if_alloc_dev(u_char type, device_t dev) 665 { 666 int numa_domain; 667 668 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 669 return (if_alloc_domain(type, IF_NODOM)); 670 return (if_alloc_domain(type, numa_domain)); 671 } 672 673 struct ifnet * 674 if_alloc(u_char type) 675 { 676 677 return (if_alloc_domain(type, IF_NODOM)); 678 } 679 /* 680 * Do the actual work of freeing a struct ifnet, and layer 2 common 681 * structure. This call is made when the last reference to an 682 * interface is released. 683 */ 684 static void 685 if_free_internal(struct ifnet *ifp) 686 { 687 688 KASSERT((ifp->if_flags & IFF_DYING), 689 ("if_free_internal: interface not dying")); 690 691 if (if_com_free[ifp->if_alloctype] != NULL) 692 if_com_free[ifp->if_alloctype](ifp->if_l2com, 693 ifp->if_alloctype); 694 695 #ifdef MAC 696 mac_ifnet_destroy(ifp); 697 #endif /* MAC */ 698 IF_AFDATA_DESTROY(ifp); 699 IF_ADDR_LOCK_DESTROY(ifp); 700 ifq_delete(&ifp->if_snd); 701 702 for (int i = 0; i < IFCOUNTERS; i++) 703 counter_u64_free(ifp->if_counters[i]); 704 705 free(ifp->if_description, M_IFDESCR); 706 free(ifp->if_hw_addr, M_IFADDR); 707 free(ifp, M_IFNET); 708 } 709 710 static void 711 if_destroy(epoch_context_t ctx) 712 { 713 struct ifnet *ifp; 714 715 ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 716 if_free_internal(ifp); 717 } 718 719 /* 720 * Deregister an interface and free the associated storage. 721 */ 722 void 723 if_free(struct ifnet *ifp) 724 { 725 726 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 727 728 CURVNET_SET_QUIET(ifp->if_vnet); 729 IFNET_WLOCK(); 730 KASSERT(ifp == ifnet_byindex(ifp->if_index), 731 ("%s: freeing unallocated ifnet", ifp->if_xname)); 732 733 ifindex_free_locked(ifp->if_index); 734 IFNET_WUNLOCK(); 735 736 if (refcount_release(&ifp->if_refcount)) 737 NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx); 738 CURVNET_RESTORE(); 739 } 740 741 /* 742 * Interfaces to keep an ifnet type-stable despite the possibility of the 743 * driver calling if_free(). If there are additional references, we defer 744 * freeing the underlying data structure. 745 */ 746 void 747 if_ref(struct ifnet *ifp) 748 { 749 u_int old; 750 751 /* We don't assert the ifnet list lock here, but arguably should. */ 752 old = refcount_acquire(&ifp->if_refcount); 753 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 754 } 755 756 bool 757 if_try_ref(struct ifnet *ifp) 758 { 759 NET_EPOCH_ASSERT(); 760 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 761 } 762 763 void 764 if_rele(struct ifnet *ifp) 765 { 766 767 if (!refcount_release(&ifp->if_refcount)) 768 return; 769 NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx); 770 } 771 772 void 773 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 774 { 775 776 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 777 778 if (ifq->ifq_maxlen == 0) 779 ifq->ifq_maxlen = ifqmaxlen; 780 781 ifq->altq_type = 0; 782 ifq->altq_disc = NULL; 783 ifq->altq_flags &= ALTQF_CANTCHANGE; 784 ifq->altq_tbr = NULL; 785 ifq->altq_ifp = ifp; 786 } 787 788 void 789 ifq_delete(struct ifaltq *ifq) 790 { 791 mtx_destroy(&ifq->ifq_mtx); 792 } 793 794 /* 795 * Perform generic interface initialization tasks and attach the interface 796 * to the list of "active" interfaces. If vmove flag is set on entry 797 * to if_attach_internal(), perform only a limited subset of initialization 798 * tasks, given that we are moving from one vnet to another an ifnet which 799 * has already been fully initialized. 800 * 801 * Note that if_detach_internal() removes group membership unconditionally 802 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 803 * Thus, when if_vmove() is applied to a cloned interface, group membership 804 * is lost while a cloned one always joins a group whose name is 805 * ifc->ifc_name. To recover this after if_detach_internal() and 806 * if_attach_internal(), the cloner should be specified to 807 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 808 * attempts to join a group whose name is ifc->ifc_name. 809 * 810 * XXX: 811 * - The decision to return void and thus require this function to 812 * succeed is questionable. 813 * - We should probably do more sanity checking. For instance we don't 814 * do anything to insure if_xname is unique or non-empty. 815 */ 816 void 817 if_attach(struct ifnet *ifp) 818 { 819 820 if_attach_internal(ifp, 0, NULL); 821 } 822 823 /* 824 * Compute the least common TSO limit. 825 */ 826 void 827 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 828 { 829 /* 830 * 1) If there is no limit currently, take the limit from 831 * the network adapter. 832 * 833 * 2) If the network adapter has a limit below the current 834 * limit, apply it. 835 */ 836 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 837 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 838 pmax->tsomaxbytes = ifp->if_hw_tsomax; 839 } 840 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 841 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 842 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 843 } 844 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 845 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 846 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 847 } 848 } 849 850 /* 851 * Update TSO limit of a network adapter. 852 * 853 * Returns zero if no change. Else non-zero. 854 */ 855 int 856 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 857 { 858 int retval = 0; 859 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 860 ifp->if_hw_tsomax = pmax->tsomaxbytes; 861 retval++; 862 } 863 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 864 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 865 retval++; 866 } 867 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 868 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 869 retval++; 870 } 871 return (retval); 872 } 873 874 static void 875 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 876 { 877 unsigned socksize, ifasize; 878 int namelen, masklen; 879 struct sockaddr_dl *sdl; 880 struct ifaddr *ifa; 881 882 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 883 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 884 ifp->if_xname); 885 886 #ifdef VIMAGE 887 ifp->if_vnet = curvnet; 888 if (ifp->if_home_vnet == NULL) 889 ifp->if_home_vnet = curvnet; 890 #endif 891 892 if_addgroup(ifp, IFG_ALL); 893 894 /* Restore group membership for cloned interfaces. */ 895 if (vmove && ifc != NULL) 896 if_clone_addgroup(ifp, ifc); 897 898 getmicrotime(&ifp->if_lastchange); 899 ifp->if_epoch = time_uptime; 900 901 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 902 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 903 ("transmit and qflush must both either be set or both be NULL")); 904 if (ifp->if_transmit == NULL) { 905 ifp->if_transmit = if_transmit; 906 ifp->if_qflush = if_qflush; 907 } 908 if (ifp->if_input == NULL) 909 ifp->if_input = if_input_default; 910 911 if (ifp->if_requestencap == NULL) 912 ifp->if_requestencap = if_requestencap_default; 913 914 if (!vmove) { 915 #ifdef MAC 916 mac_ifnet_create(ifp); 917 #endif 918 919 /* 920 * Create a Link Level name for this device. 921 */ 922 namelen = strlen(ifp->if_xname); 923 /* 924 * Always save enough space for any possiable name so we 925 * can do a rename in place later. 926 */ 927 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 928 socksize = masklen + ifp->if_addrlen; 929 if (socksize < sizeof(*sdl)) 930 socksize = sizeof(*sdl); 931 socksize = roundup2(socksize, sizeof(long)); 932 ifasize = sizeof(*ifa) + 2 * socksize; 933 ifa = ifa_alloc(ifasize, M_WAITOK); 934 sdl = (struct sockaddr_dl *)(ifa + 1); 935 sdl->sdl_len = socksize; 936 sdl->sdl_family = AF_LINK; 937 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 938 sdl->sdl_nlen = namelen; 939 sdl->sdl_index = ifp->if_index; 940 sdl->sdl_type = ifp->if_type; 941 ifp->if_addr = ifa; 942 ifa->ifa_ifp = ifp; 943 ifa->ifa_addr = (struct sockaddr *)sdl; 944 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 945 ifa->ifa_netmask = (struct sockaddr *)sdl; 946 sdl->sdl_len = masklen; 947 while (namelen != 0) 948 sdl->sdl_data[--namelen] = 0xff; 949 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 950 /* Reliably crash if used uninitialized. */ 951 ifp->if_broadcastaddr = NULL; 952 953 if (ifp->if_type == IFT_ETHER) { 954 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 955 M_WAITOK | M_ZERO); 956 } 957 958 #if defined(INET) || defined(INET6) 959 /* Use defaults for TSO, if nothing is set */ 960 if (ifp->if_hw_tsomax == 0 && 961 ifp->if_hw_tsomaxsegcount == 0 && 962 ifp->if_hw_tsomaxsegsize == 0) { 963 /* 964 * The TSO defaults needs to be such that an 965 * NFS mbuf list of 35 mbufs totalling just 966 * below 64K works and that a chain of mbufs 967 * can be defragged into at most 32 segments: 968 */ 969 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 970 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 971 ifp->if_hw_tsomaxsegcount = 35; 972 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 973 974 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 975 if (ifp->if_capabilities & IFCAP_TSO) { 976 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 977 ifp->if_hw_tsomax, 978 ifp->if_hw_tsomaxsegcount, 979 ifp->if_hw_tsomaxsegsize); 980 } 981 } 982 #endif 983 } 984 #ifdef VIMAGE 985 else { 986 /* 987 * Update the interface index in the link layer address 988 * of the interface. 989 */ 990 for (ifa = ifp->if_addr; ifa != NULL; 991 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 992 if (ifa->ifa_addr->sa_family == AF_LINK) { 993 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 994 sdl->sdl_index = ifp->if_index; 995 } 996 } 997 } 998 #endif 999 1000 if_link_ifnet(ifp); 1001 1002 if (domain_init_status >= 2) 1003 if_attachdomain1(ifp); 1004 1005 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 1006 if (IS_DEFAULT_VNET(curvnet)) 1007 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 1008 1009 /* Announce the interface. */ 1010 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 1011 } 1012 1013 static void 1014 if_epochalloc(void *dummy __unused) 1015 { 1016 1017 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 1018 } 1019 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 1020 1021 static void 1022 if_attachdomain(void *dummy) 1023 { 1024 struct ifnet *ifp; 1025 1026 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 1027 if_attachdomain1(ifp); 1028 } 1029 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 1030 if_attachdomain, NULL); 1031 1032 static void 1033 if_attachdomain1(struct ifnet *ifp) 1034 { 1035 struct domain *dp; 1036 1037 /* 1038 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 1039 * cannot lock ifp->if_afdata initialization, entirely. 1040 */ 1041 IF_AFDATA_LOCK(ifp); 1042 if (ifp->if_afdata_initialized >= domain_init_status) { 1043 IF_AFDATA_UNLOCK(ifp); 1044 log(LOG_WARNING, "%s called more than once on %s\n", 1045 __func__, ifp->if_xname); 1046 return; 1047 } 1048 ifp->if_afdata_initialized = domain_init_status; 1049 IF_AFDATA_UNLOCK(ifp); 1050 1051 /* address family dependent data region */ 1052 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 1053 for (dp = domains; dp; dp = dp->dom_next) { 1054 if (dp->dom_ifattach) 1055 ifp->if_afdata[dp->dom_family] = 1056 (*dp->dom_ifattach)(ifp); 1057 } 1058 } 1059 1060 /* 1061 * Remove any unicast or broadcast network addresses from an interface. 1062 */ 1063 void 1064 if_purgeaddrs(struct ifnet *ifp) 1065 { 1066 struct ifaddr *ifa; 1067 1068 while (1) { 1069 struct epoch_tracker et; 1070 1071 NET_EPOCH_ENTER(et); 1072 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1073 if (ifa->ifa_addr->sa_family != AF_LINK) 1074 break; 1075 } 1076 NET_EPOCH_EXIT(et); 1077 1078 if (ifa == NULL) 1079 break; 1080 #ifdef INET 1081 /* XXX: Ugly!! ad hoc just for INET */ 1082 if (ifa->ifa_addr->sa_family == AF_INET) { 1083 struct ifaliasreq ifr; 1084 1085 bzero(&ifr, sizeof(ifr)); 1086 ifr.ifra_addr = *ifa->ifa_addr; 1087 if (ifa->ifa_dstaddr) 1088 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1089 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1090 NULL) == 0) 1091 continue; 1092 } 1093 #endif /* INET */ 1094 #ifdef INET6 1095 if (ifa->ifa_addr->sa_family == AF_INET6) { 1096 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1097 /* ifp_addrhead is already updated */ 1098 continue; 1099 } 1100 #endif /* INET6 */ 1101 IF_ADDR_WLOCK(ifp); 1102 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1103 IF_ADDR_WUNLOCK(ifp); 1104 ifa_free(ifa); 1105 } 1106 } 1107 1108 /* 1109 * Remove any multicast network addresses from an interface when an ifnet 1110 * is going away. 1111 */ 1112 static void 1113 if_purgemaddrs(struct ifnet *ifp) 1114 { 1115 struct ifmultiaddr *ifma; 1116 1117 IF_ADDR_WLOCK(ifp); 1118 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1119 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1120 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1121 if_delmulti_locked(ifp, ifma, 1); 1122 } 1123 IF_ADDR_WUNLOCK(ifp); 1124 } 1125 1126 /* 1127 * Detach an interface, removing it from the list of "active" interfaces. 1128 * If vmove flag is set on entry to if_detach_internal(), perform only a 1129 * limited subset of cleanup tasks, given that we are moving an ifnet from 1130 * one vnet to another, where it must be fully operational. 1131 * 1132 * XXXRW: There are some significant questions about event ordering, and 1133 * how to prevent things from starting to use the interface during detach. 1134 */ 1135 void 1136 if_detach(struct ifnet *ifp) 1137 { 1138 bool found; 1139 1140 CURVNET_SET_QUIET(ifp->if_vnet); 1141 found = if_unlink_ifnet(ifp, false); 1142 if (found) { 1143 sx_xlock(&ifnet_detach_sxlock); 1144 if_detach_internal(ifp, 0, NULL); 1145 sx_xunlock(&ifnet_detach_sxlock); 1146 } 1147 CURVNET_RESTORE(); 1148 } 1149 1150 /* 1151 * The vmove flag, if set, indicates that we are called from a callpath 1152 * that is moving an interface to a different vnet instance. 1153 * 1154 * The shutdown flag, if set, indicates that we are called in the 1155 * process of shutting down a vnet instance. Currently only the 1156 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1157 * on a vnet instance shutdown without this flag being set, e.g., when 1158 * the cloned interfaces are destoyed as first thing of teardown. 1159 */ 1160 static int 1161 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 1162 { 1163 struct ifaddr *ifa; 1164 int i; 1165 struct domain *dp; 1166 #ifdef VIMAGE 1167 bool shutdown; 1168 1169 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1170 #endif 1171 1172 /* 1173 * At this point we know the interface still was on the ifnet list 1174 * and we removed it so we are in a stable state. 1175 */ 1176 epoch_wait_preempt(net_epoch_preempt); 1177 1178 /* 1179 * Ensure all pending EPOCH(9) callbacks have been executed. This 1180 * fixes issues about late destruction of multicast options 1181 * which lead to leave group calls, which in turn access the 1182 * belonging ifnet structure: 1183 */ 1184 epoch_drain_callbacks(net_epoch_preempt); 1185 1186 /* 1187 * In any case (destroy or vmove) detach us from the groups 1188 * and remove/wait for pending events on the taskq. 1189 * XXX-BZ in theory an interface could still enqueue a taskq change? 1190 */ 1191 if_delgroups(ifp); 1192 1193 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1194 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1195 1196 /* 1197 * Check if this is a cloned interface or not. Must do even if 1198 * shutting down as a if_vmove_reclaim() would move the ifp and 1199 * the if_clone_addgroup() will have a corrupted string overwise 1200 * from a gibberish pointer. 1201 */ 1202 if (vmove && ifcp != NULL) 1203 *ifcp = if_clone_findifc(ifp); 1204 1205 if_down(ifp); 1206 1207 #ifdef VIMAGE 1208 /* 1209 * On VNET shutdown abort here as the stack teardown will do all 1210 * the work top-down for us. 1211 */ 1212 if (shutdown) { 1213 /* Give interface users the chance to clean up. */ 1214 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1215 1216 /* 1217 * In case of a vmove we are done here without error. 1218 * If we would signal an error it would lead to the same 1219 * abort as if we did not find the ifnet anymore. 1220 * if_detach() calls us in void context and does not care 1221 * about an early abort notification, so life is splendid :) 1222 */ 1223 goto finish_vnet_shutdown; 1224 } 1225 #endif 1226 1227 /* 1228 * At this point we are not tearing down a VNET and are either 1229 * going to destroy or vmove the interface and have to cleanup 1230 * accordingly. 1231 */ 1232 1233 /* 1234 * Remove routes and flush queues. 1235 */ 1236 #ifdef ALTQ 1237 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1238 altq_disable(&ifp->if_snd); 1239 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1240 altq_detach(&ifp->if_snd); 1241 #endif 1242 1243 if_purgeaddrs(ifp); 1244 1245 #ifdef INET 1246 in_ifdetach(ifp); 1247 #endif 1248 1249 #ifdef INET6 1250 /* 1251 * Remove all IPv6 kernel structs related to ifp. This should be done 1252 * before removing routing entries below, since IPv6 interface direct 1253 * routes are expected to be removed by the IPv6-specific kernel API. 1254 * Otherwise, the kernel will detect some inconsistency and bark it. 1255 */ 1256 in6_ifdetach(ifp); 1257 #endif 1258 if_purgemaddrs(ifp); 1259 1260 /* Announce that the interface is gone. */ 1261 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1262 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1263 if (IS_DEFAULT_VNET(curvnet)) 1264 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1265 1266 if (!vmove) { 1267 /* 1268 * Prevent further calls into the device driver via ifnet. 1269 */ 1270 if_dead(ifp); 1271 1272 /* 1273 * Clean up all addresses. 1274 */ 1275 IF_ADDR_WLOCK(ifp); 1276 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1277 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1278 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1279 IF_ADDR_WUNLOCK(ifp); 1280 ifa_free(ifa); 1281 } else 1282 IF_ADDR_WUNLOCK(ifp); 1283 } 1284 1285 rt_flushifroutes(ifp); 1286 1287 #ifdef VIMAGE 1288 finish_vnet_shutdown: 1289 #endif 1290 /* 1291 * We cannot hold the lock over dom_ifdetach calls as they might 1292 * sleep, for example trying to drain a callout, thus open up the 1293 * theoretical race with re-attaching. 1294 */ 1295 IF_AFDATA_LOCK(ifp); 1296 i = ifp->if_afdata_initialized; 1297 ifp->if_afdata_initialized = 0; 1298 IF_AFDATA_UNLOCK(ifp); 1299 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1300 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1301 (*dp->dom_ifdetach)(ifp, 1302 ifp->if_afdata[dp->dom_family]); 1303 ifp->if_afdata[dp->dom_family] = NULL; 1304 } 1305 } 1306 1307 return (0); 1308 } 1309 1310 #ifdef VIMAGE 1311 /* 1312 * if_vmove() performs a limited version of if_detach() in current 1313 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1314 * An attempt is made to shrink if_index in current vnet, find an 1315 * unused if_index in target vnet and calls if_grow() if necessary, 1316 * and finally find an unused if_xname for the target vnet. 1317 */ 1318 static int 1319 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1320 { 1321 struct if_clone *ifc; 1322 #ifdef DEV_BPF 1323 u_int bif_dlt, bif_hdrlen; 1324 #endif 1325 void *old; 1326 int rc; 1327 1328 #ifdef DEV_BPF 1329 /* 1330 * if_detach_internal() will call the eventhandler to notify 1331 * interface departure. That will detach if_bpf. We need to 1332 * safe the dlt and hdrlen so we can re-attach it later. 1333 */ 1334 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1335 #endif 1336 1337 /* 1338 * Detach from current vnet, but preserve LLADDR info, do not 1339 * mark as dead etc. so that the ifnet can be reattached later. 1340 * If we cannot find it, we lost the race to someone else. 1341 */ 1342 rc = if_detach_internal(ifp, 1, &ifc); 1343 if (rc != 0) 1344 return (rc); 1345 1346 /* 1347 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1348 * the if_index for that vnet if possible. 1349 * 1350 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1351 * or we'd lock on one vnet and unlock on another. 1352 */ 1353 IFNET_WLOCK(); 1354 ifindex_free_locked(ifp->if_index); 1355 IFNET_WUNLOCK(); 1356 1357 /* 1358 * Perform interface-specific reassignment tasks, if provided by 1359 * the driver. 1360 */ 1361 if (ifp->if_reassign != NULL) 1362 ifp->if_reassign(ifp, new_vnet, NULL); 1363 1364 /* 1365 * Switch to the context of the target vnet. 1366 */ 1367 CURVNET_SET_QUIET(new_vnet); 1368 restart: 1369 IFNET_WLOCK(); 1370 ifp->if_index = ifindex_alloc(&old); 1371 if (__predict_false(ifp->if_index == USHRT_MAX)) { 1372 IFNET_WUNLOCK(); 1373 epoch_wait_preempt(net_epoch_preempt); 1374 free(old, M_IFNET); 1375 goto restart; 1376 } 1377 ifnet_setbyindex(ifp->if_index, ifp); 1378 IFNET_WUNLOCK(); 1379 1380 if_attach_internal(ifp, 1, ifc); 1381 1382 #ifdef DEV_BPF 1383 if (ifp->if_bpf == NULL) 1384 bpfattach(ifp, bif_dlt, bif_hdrlen); 1385 #endif 1386 1387 CURVNET_RESTORE(); 1388 return (0); 1389 } 1390 1391 /* 1392 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1393 */ 1394 static int 1395 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1396 { 1397 struct prison *pr; 1398 struct ifnet *difp; 1399 int error; 1400 bool found; 1401 bool shutdown; 1402 1403 /* Try to find the prison within our visibility. */ 1404 sx_slock(&allprison_lock); 1405 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1406 sx_sunlock(&allprison_lock); 1407 if (pr == NULL) 1408 return (ENXIO); 1409 prison_hold_locked(pr); 1410 mtx_unlock(&pr->pr_mtx); 1411 1412 /* Do not try to move the iface from and to the same prison. */ 1413 if (pr->pr_vnet == ifp->if_vnet) { 1414 prison_free(pr); 1415 return (EEXIST); 1416 } 1417 1418 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1419 /* XXX Lock interfaces to avoid races. */ 1420 CURVNET_SET_QUIET(pr->pr_vnet); 1421 difp = ifunit(ifname); 1422 if (difp != NULL) { 1423 CURVNET_RESTORE(); 1424 prison_free(pr); 1425 return (EEXIST); 1426 } 1427 1428 /* Make sure the VNET is stable. */ 1429 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1430 if (shutdown) { 1431 CURVNET_RESTORE(); 1432 prison_free(pr); 1433 return (EBUSY); 1434 } 1435 CURVNET_RESTORE(); 1436 1437 found = if_unlink_ifnet(ifp, true); 1438 MPASS(found); 1439 1440 /* Move the interface into the child jail/vnet. */ 1441 error = if_vmove(ifp, pr->pr_vnet); 1442 1443 /* Report the new if_xname back to the userland on success. */ 1444 if (error == 0) 1445 sprintf(ifname, "%s", ifp->if_xname); 1446 1447 prison_free(pr); 1448 return (error); 1449 } 1450 1451 static int 1452 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1453 { 1454 struct prison *pr; 1455 struct vnet *vnet_dst; 1456 struct ifnet *ifp; 1457 int error, found; 1458 bool shutdown; 1459 1460 /* Try to find the prison within our visibility. */ 1461 sx_slock(&allprison_lock); 1462 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1463 sx_sunlock(&allprison_lock); 1464 if (pr == NULL) 1465 return (ENXIO); 1466 prison_hold_locked(pr); 1467 mtx_unlock(&pr->pr_mtx); 1468 1469 /* Make sure the named iface exists in the source prison/vnet. */ 1470 CURVNET_SET(pr->pr_vnet); 1471 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1472 if (ifp == NULL) { 1473 CURVNET_RESTORE(); 1474 prison_free(pr); 1475 return (ENXIO); 1476 } 1477 1478 /* Do not try to move the iface from and to the same prison. */ 1479 vnet_dst = TD_TO_VNET(td); 1480 if (vnet_dst == ifp->if_vnet) { 1481 CURVNET_RESTORE(); 1482 prison_free(pr); 1483 return (EEXIST); 1484 } 1485 1486 /* Make sure the VNET is stable. */ 1487 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1488 if (shutdown) { 1489 CURVNET_RESTORE(); 1490 prison_free(pr); 1491 return (EBUSY); 1492 } 1493 1494 /* Get interface back from child jail/vnet. */ 1495 found = if_unlink_ifnet(ifp, true); 1496 MPASS(found); 1497 error = if_vmove(ifp, vnet_dst); 1498 CURVNET_RESTORE(); 1499 1500 /* Report the new if_xname back to the userland on success. */ 1501 if (error == 0) 1502 sprintf(ifname, "%s", ifp->if_xname); 1503 1504 prison_free(pr); 1505 return (error); 1506 } 1507 #endif /* VIMAGE */ 1508 1509 /* 1510 * Add a group to an interface 1511 */ 1512 int 1513 if_addgroup(struct ifnet *ifp, const char *groupname) 1514 { 1515 struct ifg_list *ifgl; 1516 struct ifg_group *ifg = NULL; 1517 struct ifg_member *ifgm; 1518 int new = 0; 1519 1520 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1521 groupname[strlen(groupname) - 1] <= '9') 1522 return (EINVAL); 1523 1524 IFNET_WLOCK(); 1525 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1526 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1527 IFNET_WUNLOCK(); 1528 return (EEXIST); 1529 } 1530 1531 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1532 IFNET_WUNLOCK(); 1533 return (ENOMEM); 1534 } 1535 1536 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1537 free(ifgl, M_TEMP); 1538 IFNET_WUNLOCK(); 1539 return (ENOMEM); 1540 } 1541 1542 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1543 if (!strcmp(ifg->ifg_group, groupname)) 1544 break; 1545 1546 if (ifg == NULL) { 1547 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1548 free(ifgl, M_TEMP); 1549 free(ifgm, M_TEMP); 1550 IFNET_WUNLOCK(); 1551 return (ENOMEM); 1552 } 1553 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1554 ifg->ifg_refcnt = 0; 1555 CK_STAILQ_INIT(&ifg->ifg_members); 1556 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1557 new = 1; 1558 } 1559 1560 ifg->ifg_refcnt++; 1561 ifgl->ifgl_group = ifg; 1562 ifgm->ifgm_ifp = ifp; 1563 1564 IF_ADDR_WLOCK(ifp); 1565 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1566 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1567 IF_ADDR_WUNLOCK(ifp); 1568 1569 IFNET_WUNLOCK(); 1570 1571 if (new) 1572 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1573 EVENTHANDLER_INVOKE(group_change_event, groupname); 1574 1575 return (0); 1576 } 1577 1578 /* 1579 * Helper function to remove a group out of an interface. Expects the global 1580 * ifnet lock to be write-locked, and drops it before returning. 1581 */ 1582 static void 1583 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1584 const char *groupname) 1585 { 1586 struct ifg_member *ifgm; 1587 bool freeifgl; 1588 1589 IFNET_WLOCK_ASSERT(); 1590 1591 IF_ADDR_WLOCK(ifp); 1592 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1593 IF_ADDR_WUNLOCK(ifp); 1594 1595 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1596 if (ifgm->ifgm_ifp == ifp) { 1597 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1598 ifg_member, ifgm_next); 1599 break; 1600 } 1601 } 1602 1603 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1604 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1605 ifg_next); 1606 freeifgl = true; 1607 } else { 1608 freeifgl = false; 1609 } 1610 IFNET_WUNLOCK(); 1611 1612 epoch_wait_preempt(net_epoch_preempt); 1613 if (freeifgl) { 1614 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1615 free(ifgl->ifgl_group, M_TEMP); 1616 } 1617 free(ifgm, M_TEMP); 1618 free(ifgl, M_TEMP); 1619 1620 EVENTHANDLER_INVOKE(group_change_event, groupname); 1621 } 1622 1623 /* 1624 * Remove a group from an interface 1625 */ 1626 int 1627 if_delgroup(struct ifnet *ifp, const char *groupname) 1628 { 1629 struct ifg_list *ifgl; 1630 1631 IFNET_WLOCK(); 1632 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1633 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1634 break; 1635 if (ifgl == NULL) { 1636 IFNET_WUNLOCK(); 1637 return (ENOENT); 1638 } 1639 1640 _if_delgroup_locked(ifp, ifgl, groupname); 1641 1642 return (0); 1643 } 1644 1645 /* 1646 * Remove an interface from all groups 1647 */ 1648 static void 1649 if_delgroups(struct ifnet *ifp) 1650 { 1651 struct ifg_list *ifgl; 1652 char groupname[IFNAMSIZ]; 1653 1654 IFNET_WLOCK(); 1655 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1656 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1657 _if_delgroup_locked(ifp, ifgl, groupname); 1658 IFNET_WLOCK(); 1659 } 1660 IFNET_WUNLOCK(); 1661 } 1662 1663 /* 1664 * Stores all groups from an interface in memory pointed to by ifgr. 1665 */ 1666 static int 1667 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1668 { 1669 int len, error; 1670 struct ifg_list *ifgl; 1671 struct ifg_req ifgrq, *ifgp; 1672 1673 NET_EPOCH_ASSERT(); 1674 1675 if (ifgr->ifgr_len == 0) { 1676 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1677 ifgr->ifgr_len += sizeof(struct ifg_req); 1678 return (0); 1679 } 1680 1681 len = ifgr->ifgr_len; 1682 ifgp = ifgr->ifgr_groups; 1683 /* XXX: wire */ 1684 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1685 if (len < sizeof(ifgrq)) 1686 return (EINVAL); 1687 bzero(&ifgrq, sizeof ifgrq); 1688 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1689 sizeof(ifgrq.ifgrq_group)); 1690 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1691 return (error); 1692 len -= sizeof(ifgrq); 1693 ifgp++; 1694 } 1695 1696 return (0); 1697 } 1698 1699 /* 1700 * Stores all members of a group in memory pointed to by igfr 1701 */ 1702 static int 1703 if_getgroupmembers(struct ifgroupreq *ifgr) 1704 { 1705 struct ifg_group *ifg; 1706 struct ifg_member *ifgm; 1707 struct ifg_req ifgrq, *ifgp; 1708 int len, error; 1709 1710 IFNET_RLOCK(); 1711 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1712 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1713 break; 1714 if (ifg == NULL) { 1715 IFNET_RUNLOCK(); 1716 return (ENOENT); 1717 } 1718 1719 if (ifgr->ifgr_len == 0) { 1720 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1721 ifgr->ifgr_len += sizeof(ifgrq); 1722 IFNET_RUNLOCK(); 1723 return (0); 1724 } 1725 1726 len = ifgr->ifgr_len; 1727 ifgp = ifgr->ifgr_groups; 1728 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1729 if (len < sizeof(ifgrq)) { 1730 IFNET_RUNLOCK(); 1731 return (EINVAL); 1732 } 1733 bzero(&ifgrq, sizeof ifgrq); 1734 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1735 sizeof(ifgrq.ifgrq_member)); 1736 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1737 IFNET_RUNLOCK(); 1738 return (error); 1739 } 1740 len -= sizeof(ifgrq); 1741 ifgp++; 1742 } 1743 IFNET_RUNLOCK(); 1744 1745 return (0); 1746 } 1747 1748 /* 1749 * Return counter values from counter(9)s stored in ifnet. 1750 */ 1751 uint64_t 1752 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1753 { 1754 1755 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1756 1757 return (counter_u64_fetch(ifp->if_counters[cnt])); 1758 } 1759 1760 /* 1761 * Increase an ifnet counter. Usually used for counters shared 1762 * between the stack and a driver, but function supports them all. 1763 */ 1764 void 1765 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1766 { 1767 1768 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1769 1770 counter_u64_add(ifp->if_counters[cnt], inc); 1771 } 1772 1773 /* 1774 * Copy data from ifnet to userland API structure if_data. 1775 */ 1776 void 1777 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1778 { 1779 1780 ifd->ifi_type = ifp->if_type; 1781 ifd->ifi_physical = 0; 1782 ifd->ifi_addrlen = ifp->if_addrlen; 1783 ifd->ifi_hdrlen = ifp->if_hdrlen; 1784 ifd->ifi_link_state = ifp->if_link_state; 1785 ifd->ifi_vhid = 0; 1786 ifd->ifi_datalen = sizeof(struct if_data); 1787 ifd->ifi_mtu = ifp->if_mtu; 1788 ifd->ifi_metric = ifp->if_metric; 1789 ifd->ifi_baudrate = ifp->if_baudrate; 1790 ifd->ifi_hwassist = ifp->if_hwassist; 1791 ifd->ifi_epoch = ifp->if_epoch; 1792 ifd->ifi_lastchange = ifp->if_lastchange; 1793 1794 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1795 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1796 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1797 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1798 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1799 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1800 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1801 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1802 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1803 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1804 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1805 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1806 } 1807 1808 /* 1809 * Initialization, destruction and refcounting functions for ifaddrs. 1810 */ 1811 struct ifaddr * 1812 ifa_alloc(size_t size, int flags) 1813 { 1814 struct ifaddr *ifa; 1815 1816 KASSERT(size >= sizeof(struct ifaddr), 1817 ("%s: invalid size %zu", __func__, size)); 1818 1819 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1820 if (ifa == NULL) 1821 return (NULL); 1822 1823 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1824 goto fail; 1825 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1826 goto fail; 1827 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1828 goto fail; 1829 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1830 goto fail; 1831 1832 refcount_init(&ifa->ifa_refcnt, 1); 1833 1834 return (ifa); 1835 1836 fail: 1837 /* free(NULL) is okay */ 1838 counter_u64_free(ifa->ifa_opackets); 1839 counter_u64_free(ifa->ifa_ipackets); 1840 counter_u64_free(ifa->ifa_obytes); 1841 counter_u64_free(ifa->ifa_ibytes); 1842 free(ifa, M_IFADDR); 1843 1844 return (NULL); 1845 } 1846 1847 void 1848 ifa_ref(struct ifaddr *ifa) 1849 { 1850 u_int old; 1851 1852 old = refcount_acquire(&ifa->ifa_refcnt); 1853 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1854 } 1855 1856 int 1857 ifa_try_ref(struct ifaddr *ifa) 1858 { 1859 1860 NET_EPOCH_ASSERT(); 1861 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1862 } 1863 1864 static void 1865 ifa_destroy(epoch_context_t ctx) 1866 { 1867 struct ifaddr *ifa; 1868 1869 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1870 counter_u64_free(ifa->ifa_opackets); 1871 counter_u64_free(ifa->ifa_ipackets); 1872 counter_u64_free(ifa->ifa_obytes); 1873 counter_u64_free(ifa->ifa_ibytes); 1874 free(ifa, M_IFADDR); 1875 } 1876 1877 void 1878 ifa_free(struct ifaddr *ifa) 1879 { 1880 1881 if (refcount_release(&ifa->ifa_refcnt)) 1882 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1883 } 1884 1885 /* 1886 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1887 * structs used to represent other address families, it is necessary 1888 * to perform a different comparison. 1889 */ 1890 1891 #define sa_dl_equal(a1, a2) \ 1892 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1893 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1894 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1895 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1896 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1897 1898 /* 1899 * Locate an interface based on a complete address. 1900 */ 1901 /*ARGSUSED*/ 1902 struct ifaddr * 1903 ifa_ifwithaddr(const struct sockaddr *addr) 1904 { 1905 struct ifnet *ifp; 1906 struct ifaddr *ifa; 1907 1908 NET_EPOCH_ASSERT(); 1909 1910 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1911 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1912 if (ifa->ifa_addr->sa_family != addr->sa_family) 1913 continue; 1914 if (sa_equal(addr, ifa->ifa_addr)) { 1915 goto done; 1916 } 1917 /* IP6 doesn't have broadcast */ 1918 if ((ifp->if_flags & IFF_BROADCAST) && 1919 ifa->ifa_broadaddr && 1920 ifa->ifa_broadaddr->sa_len != 0 && 1921 sa_equal(ifa->ifa_broadaddr, addr)) { 1922 goto done; 1923 } 1924 } 1925 } 1926 ifa = NULL; 1927 done: 1928 return (ifa); 1929 } 1930 1931 int 1932 ifa_ifwithaddr_check(const struct sockaddr *addr) 1933 { 1934 struct epoch_tracker et; 1935 int rc; 1936 1937 NET_EPOCH_ENTER(et); 1938 rc = (ifa_ifwithaddr(addr) != NULL); 1939 NET_EPOCH_EXIT(et); 1940 return (rc); 1941 } 1942 1943 /* 1944 * Locate an interface based on the broadcast address. 1945 */ 1946 /* ARGSUSED */ 1947 struct ifaddr * 1948 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1949 { 1950 struct ifnet *ifp; 1951 struct ifaddr *ifa; 1952 1953 NET_EPOCH_ASSERT(); 1954 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1955 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1956 continue; 1957 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1958 if (ifa->ifa_addr->sa_family != addr->sa_family) 1959 continue; 1960 if ((ifp->if_flags & IFF_BROADCAST) && 1961 ifa->ifa_broadaddr && 1962 ifa->ifa_broadaddr->sa_len != 0 && 1963 sa_equal(ifa->ifa_broadaddr, addr)) { 1964 goto done; 1965 } 1966 } 1967 } 1968 ifa = NULL; 1969 done: 1970 return (ifa); 1971 } 1972 1973 /* 1974 * Locate the point to point interface with a given destination address. 1975 */ 1976 /*ARGSUSED*/ 1977 struct ifaddr * 1978 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1979 { 1980 struct ifnet *ifp; 1981 struct ifaddr *ifa; 1982 1983 NET_EPOCH_ASSERT(); 1984 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1985 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1986 continue; 1987 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1988 continue; 1989 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1990 if (ifa->ifa_addr->sa_family != addr->sa_family) 1991 continue; 1992 if (ifa->ifa_dstaddr != NULL && 1993 sa_equal(addr, ifa->ifa_dstaddr)) { 1994 goto done; 1995 } 1996 } 1997 } 1998 ifa = NULL; 1999 done: 2000 return (ifa); 2001 } 2002 2003 /* 2004 * Find an interface on a specific network. If many, choice 2005 * is most specific found. 2006 */ 2007 struct ifaddr * 2008 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 2009 { 2010 struct ifnet *ifp; 2011 struct ifaddr *ifa; 2012 struct ifaddr *ifa_maybe = NULL; 2013 u_int af = addr->sa_family; 2014 const char *addr_data = addr->sa_data, *cplim; 2015 2016 NET_EPOCH_ASSERT(); 2017 /* 2018 * AF_LINK addresses can be looked up directly by their index number, 2019 * so do that if we can. 2020 */ 2021 if (af == AF_LINK) { 2022 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 2023 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 2024 return (ifaddr_byindex(sdl->sdl_index)); 2025 } 2026 2027 /* 2028 * Scan though each interface, looking for ones that have addresses 2029 * in this address family and the requested fib. 2030 */ 2031 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2032 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2033 continue; 2034 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2035 const char *cp, *cp2, *cp3; 2036 2037 if (ifa->ifa_addr->sa_family != af) 2038 next: continue; 2039 if (af == AF_INET && 2040 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 2041 /* 2042 * This is a bit broken as it doesn't 2043 * take into account that the remote end may 2044 * be a single node in the network we are 2045 * looking for. 2046 * The trouble is that we don't know the 2047 * netmask for the remote end. 2048 */ 2049 if (ifa->ifa_dstaddr != NULL && 2050 sa_equal(addr, ifa->ifa_dstaddr)) { 2051 goto done; 2052 } 2053 } else { 2054 /* 2055 * Scan all the bits in the ifa's address. 2056 * If a bit dissagrees with what we are 2057 * looking for, mask it with the netmask 2058 * to see if it really matters. 2059 * (A byte at a time) 2060 */ 2061 if (ifa->ifa_netmask == 0) 2062 continue; 2063 cp = addr_data; 2064 cp2 = ifa->ifa_addr->sa_data; 2065 cp3 = ifa->ifa_netmask->sa_data; 2066 cplim = ifa->ifa_netmask->sa_len 2067 + (char *)ifa->ifa_netmask; 2068 while (cp3 < cplim) 2069 if ((*cp++ ^ *cp2++) & *cp3++) 2070 goto next; /* next address! */ 2071 /* 2072 * If the netmask of what we just found 2073 * is more specific than what we had before 2074 * (if we had one), or if the virtual status 2075 * of new prefix is better than of the old one, 2076 * then remember the new one before continuing 2077 * to search for an even better one. 2078 */ 2079 if (ifa_maybe == NULL || 2080 ifa_preferred(ifa_maybe, ifa) || 2081 rn_refines((caddr_t)ifa->ifa_netmask, 2082 (caddr_t)ifa_maybe->ifa_netmask)) { 2083 ifa_maybe = ifa; 2084 } 2085 } 2086 } 2087 } 2088 ifa = ifa_maybe; 2089 ifa_maybe = NULL; 2090 done: 2091 return (ifa); 2092 } 2093 2094 /* 2095 * Find an interface address specific to an interface best matching 2096 * a given address. 2097 */ 2098 struct ifaddr * 2099 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2100 { 2101 struct ifaddr *ifa; 2102 const char *cp, *cp2, *cp3; 2103 char *cplim; 2104 struct ifaddr *ifa_maybe = NULL; 2105 u_int af = addr->sa_family; 2106 2107 if (af >= AF_MAX) 2108 return (NULL); 2109 2110 NET_EPOCH_ASSERT(); 2111 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2112 if (ifa->ifa_addr->sa_family != af) 2113 continue; 2114 if (ifa_maybe == NULL) 2115 ifa_maybe = ifa; 2116 if (ifa->ifa_netmask == 0) { 2117 if (sa_equal(addr, ifa->ifa_addr) || 2118 (ifa->ifa_dstaddr && 2119 sa_equal(addr, ifa->ifa_dstaddr))) 2120 goto done; 2121 continue; 2122 } 2123 if (ifp->if_flags & IFF_POINTOPOINT) { 2124 if (sa_equal(addr, ifa->ifa_dstaddr)) 2125 goto done; 2126 } else { 2127 cp = addr->sa_data; 2128 cp2 = ifa->ifa_addr->sa_data; 2129 cp3 = ifa->ifa_netmask->sa_data; 2130 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2131 for (; cp3 < cplim; cp3++) 2132 if ((*cp++ ^ *cp2++) & *cp3) 2133 break; 2134 if (cp3 == cplim) 2135 goto done; 2136 } 2137 } 2138 ifa = ifa_maybe; 2139 done: 2140 return (ifa); 2141 } 2142 2143 /* 2144 * See whether new ifa is better than current one: 2145 * 1) A non-virtual one is preferred over virtual. 2146 * 2) A virtual in master state preferred over any other state. 2147 * 2148 * Used in several address selecting functions. 2149 */ 2150 int 2151 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2152 { 2153 2154 return (cur->ifa_carp && (!next->ifa_carp || 2155 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2156 } 2157 2158 struct sockaddr_dl * 2159 link_alloc_sdl(size_t size, int flags) 2160 { 2161 2162 return (malloc(size, M_TEMP, flags)); 2163 } 2164 2165 void 2166 link_free_sdl(struct sockaddr *sa) 2167 { 2168 free(sa, M_TEMP); 2169 } 2170 2171 /* 2172 * Fills in given sdl with interface basic info. 2173 * Returns pointer to filled sdl. 2174 */ 2175 struct sockaddr_dl * 2176 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2177 { 2178 struct sockaddr_dl *sdl; 2179 2180 sdl = (struct sockaddr_dl *)paddr; 2181 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2182 sdl->sdl_len = sizeof(struct sockaddr_dl); 2183 sdl->sdl_family = AF_LINK; 2184 sdl->sdl_index = ifp->if_index; 2185 sdl->sdl_type = iftype; 2186 2187 return (sdl); 2188 } 2189 2190 /* 2191 * Mark an interface down and notify protocols of 2192 * the transition. 2193 */ 2194 static void 2195 if_unroute(struct ifnet *ifp, int flag, int fam) 2196 { 2197 struct ifaddr *ifa; 2198 struct epoch_tracker et; 2199 2200 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2201 2202 ifp->if_flags &= ~flag; 2203 getmicrotime(&ifp->if_lastchange); 2204 NET_EPOCH_ENTER(et); 2205 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2206 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2207 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2208 NET_EPOCH_EXIT(et); 2209 ifp->if_qflush(ifp); 2210 2211 if (ifp->if_carp) 2212 (*carp_linkstate_p)(ifp); 2213 rt_ifmsg(ifp); 2214 } 2215 2216 /* 2217 * Mark an interface up and notify protocols of 2218 * the transition. 2219 */ 2220 static void 2221 if_route(struct ifnet *ifp, int flag, int fam) 2222 { 2223 struct ifaddr *ifa; 2224 struct epoch_tracker et; 2225 2226 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2227 2228 ifp->if_flags |= flag; 2229 getmicrotime(&ifp->if_lastchange); 2230 NET_EPOCH_ENTER(et); 2231 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2232 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2233 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2234 NET_EPOCH_EXIT(et); 2235 if (ifp->if_carp) 2236 (*carp_linkstate_p)(ifp); 2237 rt_ifmsg(ifp); 2238 #ifdef INET6 2239 in6_if_up(ifp); 2240 #endif 2241 } 2242 2243 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2244 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2245 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2246 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2247 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2248 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2249 int (*vlan_setcookie_p)(struct ifnet *, void *); 2250 void *(*vlan_cookie_p)(struct ifnet *); 2251 2252 /* 2253 * Handle a change in the interface link state. To avoid LORs 2254 * between driver lock and upper layer locks, as well as possible 2255 * recursions, we post event to taskqueue, and all job 2256 * is done in static do_link_state_change(). 2257 */ 2258 void 2259 if_link_state_change(struct ifnet *ifp, int link_state) 2260 { 2261 /* Return if state hasn't changed. */ 2262 if (ifp->if_link_state == link_state) 2263 return; 2264 2265 ifp->if_link_state = link_state; 2266 2267 /* XXXGL: reference ifp? */ 2268 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2269 } 2270 2271 static void 2272 do_link_state_change(void *arg, int pending) 2273 { 2274 struct ifnet *ifp; 2275 int link_state; 2276 2277 ifp = arg; 2278 link_state = ifp->if_link_state; 2279 2280 CURVNET_SET(ifp->if_vnet); 2281 rt_ifmsg(ifp); 2282 if (ifp->if_vlantrunk != NULL) 2283 (*vlan_link_state_p)(ifp); 2284 2285 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2286 ifp->if_l2com != NULL) 2287 (*ng_ether_link_state_p)(ifp, link_state); 2288 if (ifp->if_carp) 2289 (*carp_linkstate_p)(ifp); 2290 if (ifp->if_bridge) 2291 ifp->if_bridge_linkstate(ifp); 2292 if (ifp->if_lagg) 2293 (*lagg_linkstate_p)(ifp, link_state); 2294 2295 if (IS_DEFAULT_VNET(curvnet)) 2296 devctl_notify("IFNET", ifp->if_xname, 2297 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2298 NULL); 2299 if (pending > 1) 2300 if_printf(ifp, "%d link states coalesced\n", pending); 2301 if (log_link_state_change) 2302 if_printf(ifp, "link state changed to %s\n", 2303 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2304 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2305 CURVNET_RESTORE(); 2306 } 2307 2308 /* 2309 * Mark an interface down and notify protocols of 2310 * the transition. 2311 */ 2312 void 2313 if_down(struct ifnet *ifp) 2314 { 2315 2316 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2317 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2318 } 2319 2320 /* 2321 * Mark an interface up and notify protocols of 2322 * the transition. 2323 */ 2324 void 2325 if_up(struct ifnet *ifp) 2326 { 2327 2328 if_route(ifp, IFF_UP, AF_UNSPEC); 2329 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2330 } 2331 2332 /* 2333 * Flush an interface queue. 2334 */ 2335 void 2336 if_qflush(struct ifnet *ifp) 2337 { 2338 struct mbuf *m, *n; 2339 struct ifaltq *ifq; 2340 2341 ifq = &ifp->if_snd; 2342 IFQ_LOCK(ifq); 2343 #ifdef ALTQ 2344 if (ALTQ_IS_ENABLED(ifq)) 2345 ALTQ_PURGE(ifq); 2346 #endif 2347 n = ifq->ifq_head; 2348 while ((m = n) != NULL) { 2349 n = m->m_nextpkt; 2350 m_freem(m); 2351 } 2352 ifq->ifq_head = 0; 2353 ifq->ifq_tail = 0; 2354 ifq->ifq_len = 0; 2355 IFQ_UNLOCK(ifq); 2356 } 2357 2358 /* 2359 * Map interface name to interface structure pointer, with or without 2360 * returning a reference. 2361 */ 2362 struct ifnet * 2363 ifunit_ref(const char *name) 2364 { 2365 struct epoch_tracker et; 2366 struct ifnet *ifp; 2367 2368 NET_EPOCH_ENTER(et); 2369 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2370 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2371 !(ifp->if_flags & IFF_DYING)) 2372 break; 2373 } 2374 if (ifp != NULL) 2375 if_ref(ifp); 2376 NET_EPOCH_EXIT(et); 2377 return (ifp); 2378 } 2379 2380 struct ifnet * 2381 ifunit(const char *name) 2382 { 2383 struct epoch_tracker et; 2384 struct ifnet *ifp; 2385 2386 NET_EPOCH_ENTER(et); 2387 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2388 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2389 break; 2390 } 2391 NET_EPOCH_EXIT(et); 2392 return (ifp); 2393 } 2394 2395 void * 2396 ifr_buffer_get_buffer(void *data) 2397 { 2398 union ifreq_union *ifrup; 2399 2400 ifrup = data; 2401 #ifdef COMPAT_FREEBSD32 2402 if (SV_CURPROC_FLAG(SV_ILP32)) 2403 return ((void *)(uintptr_t) 2404 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2405 #endif 2406 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2407 } 2408 2409 static void 2410 ifr_buffer_set_buffer_null(void *data) 2411 { 2412 union ifreq_union *ifrup; 2413 2414 ifrup = data; 2415 #ifdef COMPAT_FREEBSD32 2416 if (SV_CURPROC_FLAG(SV_ILP32)) 2417 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2418 else 2419 #endif 2420 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2421 } 2422 2423 size_t 2424 ifr_buffer_get_length(void *data) 2425 { 2426 union ifreq_union *ifrup; 2427 2428 ifrup = data; 2429 #ifdef COMPAT_FREEBSD32 2430 if (SV_CURPROC_FLAG(SV_ILP32)) 2431 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2432 #endif 2433 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2434 } 2435 2436 static void 2437 ifr_buffer_set_length(void *data, size_t len) 2438 { 2439 union ifreq_union *ifrup; 2440 2441 ifrup = data; 2442 #ifdef COMPAT_FREEBSD32 2443 if (SV_CURPROC_FLAG(SV_ILP32)) 2444 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2445 else 2446 #endif 2447 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2448 } 2449 2450 void * 2451 ifr_data_get_ptr(void *ifrp) 2452 { 2453 union ifreq_union *ifrup; 2454 2455 ifrup = ifrp; 2456 #ifdef COMPAT_FREEBSD32 2457 if (SV_CURPROC_FLAG(SV_ILP32)) 2458 return ((void *)(uintptr_t) 2459 ifrup->ifr32.ifr_ifru.ifru_data); 2460 #endif 2461 return (ifrup->ifr.ifr_ifru.ifru_data); 2462 } 2463 2464 /* 2465 * Hardware specific interface ioctls. 2466 */ 2467 int 2468 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2469 { 2470 struct ifreq *ifr; 2471 int error = 0, do_ifup = 0; 2472 int new_flags, temp_flags; 2473 size_t namelen, onamelen; 2474 size_t descrlen; 2475 char *descrbuf, *odescrbuf; 2476 char new_name[IFNAMSIZ]; 2477 char old_name[IFNAMSIZ], strbuf[IFNAMSIZ + 8]; 2478 struct ifaddr *ifa; 2479 struct sockaddr_dl *sdl; 2480 2481 ifr = (struct ifreq *)data; 2482 switch (cmd) { 2483 case SIOCGIFINDEX: 2484 ifr->ifr_index = ifp->if_index; 2485 break; 2486 2487 case SIOCGIFFLAGS: 2488 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2489 ifr->ifr_flags = temp_flags & 0xffff; 2490 ifr->ifr_flagshigh = temp_flags >> 16; 2491 break; 2492 2493 case SIOCGIFCAP: 2494 ifr->ifr_reqcap = ifp->if_capabilities; 2495 ifr->ifr_curcap = ifp->if_capenable; 2496 break; 2497 2498 case SIOCGIFDATA: 2499 { 2500 struct if_data ifd; 2501 2502 /* Ensure uninitialised padding is not leaked. */ 2503 memset(&ifd, 0, sizeof(ifd)); 2504 2505 if_data_copy(ifp, &ifd); 2506 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2507 break; 2508 } 2509 2510 #ifdef MAC 2511 case SIOCGIFMAC: 2512 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2513 break; 2514 #endif 2515 2516 case SIOCGIFMETRIC: 2517 ifr->ifr_metric = ifp->if_metric; 2518 break; 2519 2520 case SIOCGIFMTU: 2521 ifr->ifr_mtu = ifp->if_mtu; 2522 break; 2523 2524 case SIOCGIFPHYS: 2525 /* XXXGL: did this ever worked? */ 2526 ifr->ifr_phys = 0; 2527 break; 2528 2529 case SIOCGIFDESCR: 2530 error = 0; 2531 sx_slock(&ifdescr_sx); 2532 if (ifp->if_description == NULL) 2533 error = ENOMSG; 2534 else { 2535 /* space for terminating nul */ 2536 descrlen = strlen(ifp->if_description) + 1; 2537 if (ifr_buffer_get_length(ifr) < descrlen) 2538 ifr_buffer_set_buffer_null(ifr); 2539 else 2540 error = copyout(ifp->if_description, 2541 ifr_buffer_get_buffer(ifr), descrlen); 2542 ifr_buffer_set_length(ifr, descrlen); 2543 } 2544 sx_sunlock(&ifdescr_sx); 2545 break; 2546 2547 case SIOCSIFDESCR: 2548 error = priv_check(td, PRIV_NET_SETIFDESCR); 2549 if (error) 2550 return (error); 2551 2552 /* 2553 * Copy only (length-1) bytes to make sure that 2554 * if_description is always nul terminated. The 2555 * length parameter is supposed to count the 2556 * terminating nul in. 2557 */ 2558 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2559 return (ENAMETOOLONG); 2560 else if (ifr_buffer_get_length(ifr) == 0) 2561 descrbuf = NULL; 2562 else { 2563 descrbuf = malloc(ifr_buffer_get_length(ifr), 2564 M_IFDESCR, M_WAITOK | M_ZERO); 2565 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2566 ifr_buffer_get_length(ifr) - 1); 2567 if (error) { 2568 free(descrbuf, M_IFDESCR); 2569 break; 2570 } 2571 } 2572 2573 sx_xlock(&ifdescr_sx); 2574 odescrbuf = ifp->if_description; 2575 ifp->if_description = descrbuf; 2576 sx_xunlock(&ifdescr_sx); 2577 2578 getmicrotime(&ifp->if_lastchange); 2579 free(odescrbuf, M_IFDESCR); 2580 break; 2581 2582 case SIOCGIFFIB: 2583 ifr->ifr_fib = ifp->if_fib; 2584 break; 2585 2586 case SIOCSIFFIB: 2587 error = priv_check(td, PRIV_NET_SETIFFIB); 2588 if (error) 2589 return (error); 2590 if (ifr->ifr_fib >= rt_numfibs) 2591 return (EINVAL); 2592 2593 ifp->if_fib = ifr->ifr_fib; 2594 break; 2595 2596 case SIOCSIFFLAGS: 2597 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2598 if (error) 2599 return (error); 2600 /* 2601 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2602 * check, so we don't need special handling here yet. 2603 */ 2604 new_flags = (ifr->ifr_flags & 0xffff) | 2605 (ifr->ifr_flagshigh << 16); 2606 if (ifp->if_flags & IFF_UP && 2607 (new_flags & IFF_UP) == 0) { 2608 if_down(ifp); 2609 } else if (new_flags & IFF_UP && 2610 (ifp->if_flags & IFF_UP) == 0) { 2611 do_ifup = 1; 2612 } 2613 /* See if permanently promiscuous mode bit is about to flip */ 2614 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2615 if (new_flags & IFF_PPROMISC) 2616 ifp->if_flags |= IFF_PROMISC; 2617 else if (ifp->if_pcount == 0) 2618 ifp->if_flags &= ~IFF_PROMISC; 2619 if (log_promisc_mode_change) 2620 if_printf(ifp, "permanently promiscuous mode %s\n", 2621 ((new_flags & IFF_PPROMISC) ? 2622 "enabled" : "disabled")); 2623 } 2624 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2625 (new_flags &~ IFF_CANTCHANGE); 2626 if (ifp->if_ioctl) { 2627 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2628 } 2629 if (do_ifup) 2630 if_up(ifp); 2631 getmicrotime(&ifp->if_lastchange); 2632 break; 2633 2634 case SIOCSIFCAP: 2635 error = priv_check(td, PRIV_NET_SETIFCAP); 2636 if (error) 2637 return (error); 2638 if (ifp->if_ioctl == NULL) 2639 return (EOPNOTSUPP); 2640 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2641 return (EINVAL); 2642 error = (*ifp->if_ioctl)(ifp, cmd, data); 2643 if (error == 0) 2644 getmicrotime(&ifp->if_lastchange); 2645 break; 2646 2647 #ifdef MAC 2648 case SIOCSIFMAC: 2649 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2650 break; 2651 #endif 2652 2653 case SIOCSIFNAME: 2654 error = priv_check(td, PRIV_NET_SETIFNAME); 2655 if (error) 2656 return (error); 2657 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2658 NULL); 2659 if (error != 0) 2660 return (error); 2661 if (new_name[0] == '\0') 2662 return (EINVAL); 2663 if (strcmp(new_name, ifp->if_xname) == 0) 2664 break; 2665 if (ifunit(new_name) != NULL) 2666 return (EEXIST); 2667 2668 /* 2669 * XXX: Locking. Nothing else seems to lock if_flags, 2670 * and there are numerous other races with the 2671 * ifunit() checks not being atomic with namespace 2672 * changes (renames, vmoves, if_attach, etc). 2673 */ 2674 ifp->if_flags |= IFF_RENAMING; 2675 2676 /* Announce the departure of the interface. */ 2677 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2678 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2679 2680 if_printf(ifp, "changing name to '%s'\n", new_name); 2681 2682 IF_ADDR_WLOCK(ifp); 2683 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 2684 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2685 ifa = ifp->if_addr; 2686 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2687 namelen = strlen(new_name); 2688 onamelen = sdl->sdl_nlen; 2689 /* 2690 * Move the address if needed. This is safe because we 2691 * allocate space for a name of length IFNAMSIZ when we 2692 * create this in if_attach(). 2693 */ 2694 if (namelen != onamelen) { 2695 bcopy(sdl->sdl_data + onamelen, 2696 sdl->sdl_data + namelen, sdl->sdl_alen); 2697 } 2698 bcopy(new_name, sdl->sdl_data, namelen); 2699 sdl->sdl_nlen = namelen; 2700 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2701 bzero(sdl->sdl_data, onamelen); 2702 while (namelen != 0) 2703 sdl->sdl_data[--namelen] = 0xff; 2704 IF_ADDR_WUNLOCK(ifp); 2705 2706 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2707 /* Announce the return of the interface. */ 2708 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2709 2710 ifp->if_flags &= ~IFF_RENAMING; 2711 2712 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 2713 devctl_notify("IFNET", old_name, "RENAME", strbuf); 2714 break; 2715 2716 #ifdef VIMAGE 2717 case SIOCSIFVNET: 2718 error = priv_check(td, PRIV_NET_SETIFVNET); 2719 if (error) 2720 return (error); 2721 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2722 break; 2723 #endif 2724 2725 case SIOCSIFMETRIC: 2726 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2727 if (error) 2728 return (error); 2729 ifp->if_metric = ifr->ifr_metric; 2730 getmicrotime(&ifp->if_lastchange); 2731 break; 2732 2733 case SIOCSIFPHYS: 2734 error = priv_check(td, PRIV_NET_SETIFPHYS); 2735 if (error) 2736 return (error); 2737 if (ifp->if_ioctl == NULL) 2738 return (EOPNOTSUPP); 2739 error = (*ifp->if_ioctl)(ifp, cmd, data); 2740 if (error == 0) 2741 getmicrotime(&ifp->if_lastchange); 2742 break; 2743 2744 case SIOCSIFMTU: 2745 { 2746 u_long oldmtu = ifp->if_mtu; 2747 2748 error = priv_check(td, PRIV_NET_SETIFMTU); 2749 if (error) 2750 return (error); 2751 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2752 return (EINVAL); 2753 if (ifp->if_ioctl == NULL) 2754 return (EOPNOTSUPP); 2755 /* Disallow MTU changes on bridge member interfaces. */ 2756 if (ifp->if_bridge) 2757 return (EOPNOTSUPP); 2758 error = (*ifp->if_ioctl)(ifp, cmd, data); 2759 if (error == 0) { 2760 getmicrotime(&ifp->if_lastchange); 2761 rt_ifmsg(ifp); 2762 #ifdef INET 2763 DEBUGNET_NOTIFY_MTU(ifp); 2764 #endif 2765 } 2766 /* 2767 * If the link MTU changed, do network layer specific procedure. 2768 */ 2769 if (ifp->if_mtu != oldmtu) { 2770 #ifdef INET6 2771 nd6_setmtu(ifp); 2772 #endif 2773 rt_updatemtu(ifp); 2774 } 2775 break; 2776 } 2777 2778 case SIOCADDMULTI: 2779 case SIOCDELMULTI: 2780 if (cmd == SIOCADDMULTI) 2781 error = priv_check(td, PRIV_NET_ADDMULTI); 2782 else 2783 error = priv_check(td, PRIV_NET_DELMULTI); 2784 if (error) 2785 return (error); 2786 2787 /* Don't allow group membership on non-multicast interfaces. */ 2788 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2789 return (EOPNOTSUPP); 2790 2791 /* Don't let users screw up protocols' entries. */ 2792 if (ifr->ifr_addr.sa_family != AF_LINK) 2793 return (EINVAL); 2794 2795 if (cmd == SIOCADDMULTI) { 2796 struct epoch_tracker et; 2797 struct ifmultiaddr *ifma; 2798 2799 /* 2800 * Userland is only permitted to join groups once 2801 * via the if_addmulti() KPI, because it cannot hold 2802 * struct ifmultiaddr * between calls. It may also 2803 * lose a race while we check if the membership 2804 * already exists. 2805 */ 2806 NET_EPOCH_ENTER(et); 2807 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2808 NET_EPOCH_EXIT(et); 2809 if (ifma != NULL) 2810 error = EADDRINUSE; 2811 else 2812 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2813 } else { 2814 error = if_delmulti(ifp, &ifr->ifr_addr); 2815 } 2816 if (error == 0) 2817 getmicrotime(&ifp->if_lastchange); 2818 break; 2819 2820 case SIOCSIFPHYADDR: 2821 case SIOCDIFPHYADDR: 2822 #ifdef INET6 2823 case SIOCSIFPHYADDR_IN6: 2824 #endif 2825 case SIOCSIFMEDIA: 2826 case SIOCSIFGENERIC: 2827 error = priv_check(td, PRIV_NET_HWIOCTL); 2828 if (error) 2829 return (error); 2830 if (ifp->if_ioctl == NULL) 2831 return (EOPNOTSUPP); 2832 error = (*ifp->if_ioctl)(ifp, cmd, data); 2833 if (error == 0) 2834 getmicrotime(&ifp->if_lastchange); 2835 break; 2836 2837 case SIOCGIFSTATUS: 2838 case SIOCGIFPSRCADDR: 2839 case SIOCGIFPDSTADDR: 2840 case SIOCGIFMEDIA: 2841 case SIOCGIFXMEDIA: 2842 case SIOCGIFGENERIC: 2843 case SIOCGIFRSSKEY: 2844 case SIOCGIFRSSHASH: 2845 case SIOCGIFDOWNREASON: 2846 if (ifp->if_ioctl == NULL) 2847 return (EOPNOTSUPP); 2848 error = (*ifp->if_ioctl)(ifp, cmd, data); 2849 break; 2850 2851 case SIOCSIFLLADDR: 2852 error = priv_check(td, PRIV_NET_SETLLADDR); 2853 if (error) 2854 return (error); 2855 error = if_setlladdr(ifp, 2856 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2857 break; 2858 2859 case SIOCGHWADDR: 2860 error = if_gethwaddr(ifp, ifr); 2861 break; 2862 2863 case SIOCAIFGROUP: 2864 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2865 if (error) 2866 return (error); 2867 error = if_addgroup(ifp, 2868 ((struct ifgroupreq *)data)->ifgr_group); 2869 if (error != 0) 2870 return (error); 2871 break; 2872 2873 case SIOCGIFGROUP: 2874 { 2875 struct epoch_tracker et; 2876 2877 NET_EPOCH_ENTER(et); 2878 error = if_getgroup((struct ifgroupreq *)data, ifp); 2879 NET_EPOCH_EXIT(et); 2880 break; 2881 } 2882 2883 case SIOCDIFGROUP: 2884 error = priv_check(td, PRIV_NET_DELIFGROUP); 2885 if (error) 2886 return (error); 2887 error = if_delgroup(ifp, 2888 ((struct ifgroupreq *)data)->ifgr_group); 2889 if (error != 0) 2890 return (error); 2891 break; 2892 2893 default: 2894 error = ENOIOCTL; 2895 break; 2896 } 2897 return (error); 2898 } 2899 2900 /* 2901 * Interface ioctls. 2902 */ 2903 int 2904 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2905 { 2906 #ifdef COMPAT_FREEBSD32 2907 union { 2908 struct ifconf ifc; 2909 struct ifdrv ifd; 2910 struct ifgroupreq ifgr; 2911 struct ifmediareq ifmr; 2912 } thunk; 2913 u_long saved_cmd; 2914 struct ifconf32 *ifc32; 2915 struct ifdrv32 *ifd32; 2916 struct ifgroupreq32 *ifgr32; 2917 struct ifmediareq32 *ifmr32; 2918 #endif 2919 struct ifnet *ifp; 2920 struct ifreq *ifr; 2921 int error; 2922 int oif_flags; 2923 #ifdef VIMAGE 2924 bool shutdown; 2925 #endif 2926 2927 CURVNET_SET(so->so_vnet); 2928 #ifdef VIMAGE 2929 /* Make sure the VNET is stable. */ 2930 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2931 if (shutdown) { 2932 CURVNET_RESTORE(); 2933 return (EBUSY); 2934 } 2935 #endif 2936 2937 #ifdef COMPAT_FREEBSD32 2938 saved_cmd = cmd; 2939 switch (cmd) { 2940 case SIOCGIFCONF32: 2941 ifc32 = (struct ifconf32 *)data; 2942 thunk.ifc.ifc_len = ifc32->ifc_len; 2943 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2944 data = (caddr_t)&thunk.ifc; 2945 cmd = SIOCGIFCONF; 2946 break; 2947 case SIOCGDRVSPEC32: 2948 case SIOCSDRVSPEC32: 2949 ifd32 = (struct ifdrv32 *)data; 2950 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2951 sizeof(thunk.ifd.ifd_name)); 2952 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2953 thunk.ifd.ifd_len = ifd32->ifd_len; 2954 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2955 data = (caddr_t)&thunk.ifd; 2956 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2957 break; 2958 case SIOCAIFGROUP32: 2959 case SIOCGIFGROUP32: 2960 case SIOCDIFGROUP32: 2961 case SIOCGIFGMEMB32: 2962 ifgr32 = (struct ifgroupreq32 *)data; 2963 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2964 sizeof(thunk.ifgr.ifgr_name)); 2965 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2966 switch (cmd) { 2967 case SIOCAIFGROUP32: 2968 case SIOCDIFGROUP32: 2969 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2970 sizeof(thunk.ifgr.ifgr_group)); 2971 break; 2972 case SIOCGIFGROUP32: 2973 case SIOCGIFGMEMB32: 2974 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2975 break; 2976 } 2977 data = (caddr_t)&thunk.ifgr; 2978 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2979 break; 2980 case SIOCGIFMEDIA32: 2981 case SIOCGIFXMEDIA32: 2982 ifmr32 = (struct ifmediareq32 *)data; 2983 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2984 sizeof(thunk.ifmr.ifm_name)); 2985 thunk.ifmr.ifm_current = ifmr32->ifm_current; 2986 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 2987 thunk.ifmr.ifm_status = ifmr32->ifm_status; 2988 thunk.ifmr.ifm_active = ifmr32->ifm_active; 2989 thunk.ifmr.ifm_count = ifmr32->ifm_count; 2990 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 2991 data = (caddr_t)&thunk.ifmr; 2992 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 2993 break; 2994 } 2995 #endif 2996 2997 switch (cmd) { 2998 case SIOCGIFCONF: 2999 error = ifconf(cmd, data); 3000 goto out_noref; 3001 } 3002 3003 ifr = (struct ifreq *)data; 3004 switch (cmd) { 3005 #ifdef VIMAGE 3006 case SIOCSIFRVNET: 3007 error = priv_check(td, PRIV_NET_SETIFVNET); 3008 if (error == 0) 3009 error = if_vmove_reclaim(td, ifr->ifr_name, 3010 ifr->ifr_jid); 3011 goto out_noref; 3012 #endif 3013 case SIOCIFCREATE: 3014 case SIOCIFCREATE2: 3015 error = priv_check(td, PRIV_NET_IFCREATE); 3016 if (error == 0) 3017 error = if_clone_create(ifr->ifr_name, 3018 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3019 ifr_data_get_ptr(ifr) : NULL); 3020 goto out_noref; 3021 case SIOCIFDESTROY: 3022 error = priv_check(td, PRIV_NET_IFDESTROY); 3023 3024 if (error == 0) { 3025 sx_xlock(&ifnet_detach_sxlock); 3026 error = if_clone_destroy(ifr->ifr_name); 3027 sx_xunlock(&ifnet_detach_sxlock); 3028 } 3029 goto out_noref; 3030 3031 case SIOCIFGCLONERS: 3032 error = if_clone_list((struct if_clonereq *)data); 3033 goto out_noref; 3034 3035 case SIOCGIFGMEMB: 3036 error = if_getgroupmembers((struct ifgroupreq *)data); 3037 goto out_noref; 3038 3039 #if defined(INET) || defined(INET6) 3040 case SIOCSVH: 3041 case SIOCGVH: 3042 if (carp_ioctl_p == NULL) 3043 error = EPROTONOSUPPORT; 3044 else 3045 error = (*carp_ioctl_p)(ifr, cmd, td); 3046 goto out_noref; 3047 #endif 3048 } 3049 3050 ifp = ifunit_ref(ifr->ifr_name); 3051 if (ifp == NULL) { 3052 error = ENXIO; 3053 goto out_noref; 3054 } 3055 3056 error = ifhwioctl(cmd, ifp, data, td); 3057 if (error != ENOIOCTL) 3058 goto out_ref; 3059 3060 oif_flags = ifp->if_flags; 3061 if (so->so_proto == NULL) { 3062 error = EOPNOTSUPP; 3063 goto out_ref; 3064 } 3065 3066 /* 3067 * Pass the request on to the socket control method, and if the 3068 * latter returns EOPNOTSUPP, directly to the interface. 3069 * 3070 * Make an exception for the legacy SIOCSIF* requests. Drivers 3071 * trust SIOCSIFADDR et al to come from an already privileged 3072 * layer, and do not perform any credentials checks or input 3073 * validation. 3074 */ 3075 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 3076 ifp, td)); 3077 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3078 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3079 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3080 error = (*ifp->if_ioctl)(ifp, cmd, data); 3081 3082 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3083 #ifdef INET6 3084 if (ifp->if_flags & IFF_UP) 3085 in6_if_up(ifp); 3086 #endif 3087 } 3088 3089 out_ref: 3090 if_rele(ifp); 3091 out_noref: 3092 CURVNET_RESTORE(); 3093 #ifdef COMPAT_FREEBSD32 3094 if (error != 0) 3095 return (error); 3096 switch (saved_cmd) { 3097 case SIOCGIFCONF32: 3098 ifc32->ifc_len = thunk.ifc.ifc_len; 3099 break; 3100 case SIOCGDRVSPEC32: 3101 /* 3102 * SIOCGDRVSPEC is IOWR, but nothing actually touches 3103 * the struct so just assert that ifd_len (the only 3104 * field it might make sense to update) hasn't 3105 * changed. 3106 */ 3107 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 3108 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 3109 thunk.ifd.ifd_len)); 3110 break; 3111 case SIOCGIFGROUP32: 3112 case SIOCGIFGMEMB32: 3113 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 3114 break; 3115 case SIOCGIFMEDIA32: 3116 case SIOCGIFXMEDIA32: 3117 ifmr32->ifm_current = thunk.ifmr.ifm_current; 3118 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 3119 ifmr32->ifm_status = thunk.ifmr.ifm_status; 3120 ifmr32->ifm_active = thunk.ifmr.ifm_active; 3121 ifmr32->ifm_count = thunk.ifmr.ifm_count; 3122 break; 3123 } 3124 #endif 3125 return (error); 3126 } 3127 3128 /* 3129 * The code common to handling reference counted flags, 3130 * e.g., in ifpromisc() and if_allmulti(). 3131 * The "pflag" argument can specify a permanent mode flag to check, 3132 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3133 * 3134 * Only to be used on stack-owned flags, not driver-owned flags. 3135 */ 3136 static int 3137 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3138 { 3139 struct ifreq ifr; 3140 int error; 3141 int oldflags, oldcount; 3142 3143 /* Sanity checks to catch programming errors */ 3144 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3145 ("%s: setting driver-owned flag %d", __func__, flag)); 3146 3147 if (onswitch) 3148 KASSERT(*refcount >= 0, 3149 ("%s: increment negative refcount %d for flag %d", 3150 __func__, *refcount, flag)); 3151 else 3152 KASSERT(*refcount > 0, 3153 ("%s: decrement non-positive refcount %d for flag %d", 3154 __func__, *refcount, flag)); 3155 3156 /* In case this mode is permanent, just touch refcount */ 3157 if (ifp->if_flags & pflag) { 3158 *refcount += onswitch ? 1 : -1; 3159 return (0); 3160 } 3161 3162 /* Save ifnet parameters for if_ioctl() may fail */ 3163 oldcount = *refcount; 3164 oldflags = ifp->if_flags; 3165 3166 /* 3167 * See if we aren't the only and touching refcount is enough. 3168 * Actually toggle interface flag if we are the first or last. 3169 */ 3170 if (onswitch) { 3171 if ((*refcount)++) 3172 return (0); 3173 ifp->if_flags |= flag; 3174 } else { 3175 if (--(*refcount)) 3176 return (0); 3177 ifp->if_flags &= ~flag; 3178 } 3179 3180 /* Call down the driver since we've changed interface flags */ 3181 if (ifp->if_ioctl == NULL) { 3182 error = EOPNOTSUPP; 3183 goto recover; 3184 } 3185 ifr.ifr_flags = ifp->if_flags & 0xffff; 3186 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3187 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3188 if (error) 3189 goto recover; 3190 /* Notify userland that interface flags have changed */ 3191 rt_ifmsg(ifp); 3192 return (0); 3193 3194 recover: 3195 /* Recover after driver error */ 3196 *refcount = oldcount; 3197 ifp->if_flags = oldflags; 3198 return (error); 3199 } 3200 3201 /* 3202 * Set/clear promiscuous mode on interface ifp based on the truth value 3203 * of pswitch. The calls are reference counted so that only the first 3204 * "on" request actually has an effect, as does the final "off" request. 3205 * Results are undefined if the "off" and "on" requests are not matched. 3206 */ 3207 int 3208 ifpromisc(struct ifnet *ifp, int pswitch) 3209 { 3210 int error; 3211 int oldflags = ifp->if_flags; 3212 3213 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3214 &ifp->if_pcount, pswitch); 3215 /* If promiscuous mode status has changed, log a message */ 3216 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3217 log_promisc_mode_change) 3218 if_printf(ifp, "promiscuous mode %s\n", 3219 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3220 return (error); 3221 } 3222 3223 /* 3224 * Return interface configuration 3225 * of system. List may be used 3226 * in later ioctl's (above) to get 3227 * other information. 3228 */ 3229 /*ARGSUSED*/ 3230 static int 3231 ifconf(u_long cmd, caddr_t data) 3232 { 3233 struct ifconf *ifc = (struct ifconf *)data; 3234 struct ifnet *ifp; 3235 struct ifaddr *ifa; 3236 struct ifreq ifr; 3237 struct sbuf *sb; 3238 int error, full = 0, valid_len, max_len; 3239 3240 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3241 max_len = maxphys - 1; 3242 3243 /* Prevent hostile input from being able to crash the system */ 3244 if (ifc->ifc_len <= 0) 3245 return (EINVAL); 3246 3247 again: 3248 if (ifc->ifc_len <= max_len) { 3249 max_len = ifc->ifc_len; 3250 full = 1; 3251 } 3252 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3253 max_len = 0; 3254 valid_len = 0; 3255 3256 IFNET_RLOCK(); 3257 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3258 struct epoch_tracker et; 3259 int addrs; 3260 3261 /* 3262 * Zero the ifr to make sure we don't disclose the contents 3263 * of the stack. 3264 */ 3265 memset(&ifr, 0, sizeof(ifr)); 3266 3267 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3268 >= sizeof(ifr.ifr_name)) { 3269 sbuf_delete(sb); 3270 IFNET_RUNLOCK(); 3271 return (ENAMETOOLONG); 3272 } 3273 3274 addrs = 0; 3275 NET_EPOCH_ENTER(et); 3276 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3277 struct sockaddr *sa = ifa->ifa_addr; 3278 3279 if (prison_if(curthread->td_ucred, sa) != 0) 3280 continue; 3281 addrs++; 3282 if (sa->sa_len <= sizeof(*sa)) { 3283 if (sa->sa_len < sizeof(*sa)) { 3284 memset(&ifr.ifr_ifru.ifru_addr, 0, 3285 sizeof(ifr.ifr_ifru.ifru_addr)); 3286 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3287 sa->sa_len); 3288 } else 3289 ifr.ifr_ifru.ifru_addr = *sa; 3290 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3291 max_len += sizeof(ifr); 3292 } else { 3293 sbuf_bcat(sb, &ifr, 3294 offsetof(struct ifreq, ifr_addr)); 3295 max_len += offsetof(struct ifreq, ifr_addr); 3296 sbuf_bcat(sb, sa, sa->sa_len); 3297 max_len += sa->sa_len; 3298 } 3299 3300 if (sbuf_error(sb) == 0) 3301 valid_len = sbuf_len(sb); 3302 } 3303 NET_EPOCH_EXIT(et); 3304 if (addrs == 0) { 3305 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3306 max_len += sizeof(ifr); 3307 3308 if (sbuf_error(sb) == 0) 3309 valid_len = sbuf_len(sb); 3310 } 3311 } 3312 IFNET_RUNLOCK(); 3313 3314 /* 3315 * If we didn't allocate enough space (uncommon), try again. If 3316 * we have already allocated as much space as we are allowed, 3317 * return what we've got. 3318 */ 3319 if (valid_len != max_len && !full) { 3320 sbuf_delete(sb); 3321 goto again; 3322 } 3323 3324 ifc->ifc_len = valid_len; 3325 sbuf_finish(sb); 3326 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3327 sbuf_delete(sb); 3328 return (error); 3329 } 3330 3331 /* 3332 * Just like ifpromisc(), but for all-multicast-reception mode. 3333 */ 3334 int 3335 if_allmulti(struct ifnet *ifp, int onswitch) 3336 { 3337 3338 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3339 } 3340 3341 struct ifmultiaddr * 3342 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3343 { 3344 struct ifmultiaddr *ifma; 3345 3346 IF_ADDR_LOCK_ASSERT(ifp); 3347 3348 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3349 if (sa->sa_family == AF_LINK) { 3350 if (sa_dl_equal(ifma->ifma_addr, sa)) 3351 break; 3352 } else { 3353 if (sa_equal(ifma->ifma_addr, sa)) 3354 break; 3355 } 3356 } 3357 3358 return ifma; 3359 } 3360 3361 /* 3362 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3363 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3364 * the ifnet multicast address list here, so the caller must do that and 3365 * other setup work (such as notifying the device driver). The reference 3366 * count is initialized to 1. 3367 */ 3368 static struct ifmultiaddr * 3369 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3370 int mflags) 3371 { 3372 struct ifmultiaddr *ifma; 3373 struct sockaddr *dupsa; 3374 3375 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3376 M_ZERO); 3377 if (ifma == NULL) 3378 return (NULL); 3379 3380 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3381 if (dupsa == NULL) { 3382 free(ifma, M_IFMADDR); 3383 return (NULL); 3384 } 3385 bcopy(sa, dupsa, sa->sa_len); 3386 ifma->ifma_addr = dupsa; 3387 3388 ifma->ifma_ifp = ifp; 3389 ifma->ifma_refcount = 1; 3390 ifma->ifma_protospec = NULL; 3391 3392 if (llsa == NULL) { 3393 ifma->ifma_lladdr = NULL; 3394 return (ifma); 3395 } 3396 3397 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3398 if (dupsa == NULL) { 3399 free(ifma->ifma_addr, M_IFMADDR); 3400 free(ifma, M_IFMADDR); 3401 return (NULL); 3402 } 3403 bcopy(llsa, dupsa, llsa->sa_len); 3404 ifma->ifma_lladdr = dupsa; 3405 3406 return (ifma); 3407 } 3408 3409 /* 3410 * if_freemulti: free ifmultiaddr structure and possibly attached related 3411 * addresses. The caller is responsible for implementing reference 3412 * counting, notifying the driver, handling routing messages, and releasing 3413 * any dependent link layer state. 3414 */ 3415 #ifdef MCAST_VERBOSE 3416 extern void kdb_backtrace(void); 3417 #endif 3418 static void 3419 if_freemulti_internal(struct ifmultiaddr *ifma) 3420 { 3421 3422 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3423 ifma->ifma_refcount)); 3424 3425 if (ifma->ifma_lladdr != NULL) 3426 free(ifma->ifma_lladdr, M_IFMADDR); 3427 #ifdef MCAST_VERBOSE 3428 kdb_backtrace(); 3429 printf("%s freeing ifma: %p\n", __func__, ifma); 3430 #endif 3431 free(ifma->ifma_addr, M_IFMADDR); 3432 free(ifma, M_IFMADDR); 3433 } 3434 3435 static void 3436 if_destroymulti(epoch_context_t ctx) 3437 { 3438 struct ifmultiaddr *ifma; 3439 3440 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3441 if_freemulti_internal(ifma); 3442 } 3443 3444 void 3445 if_freemulti(struct ifmultiaddr *ifma) 3446 { 3447 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3448 ifma->ifma_refcount)); 3449 3450 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3451 } 3452 3453 /* 3454 * Register an additional multicast address with a network interface. 3455 * 3456 * - If the address is already present, bump the reference count on the 3457 * address and return. 3458 * - If the address is not link-layer, look up a link layer address. 3459 * - Allocate address structures for one or both addresses, and attach to the 3460 * multicast address list on the interface. If automatically adding a link 3461 * layer address, the protocol address will own a reference to the link 3462 * layer address, to be freed when it is freed. 3463 * - Notify the network device driver of an addition to the multicast address 3464 * list. 3465 * 3466 * 'sa' points to caller-owned memory with the desired multicast address. 3467 * 3468 * 'retifma' will be used to return a pointer to the resulting multicast 3469 * address reference, if desired. 3470 */ 3471 int 3472 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3473 struct ifmultiaddr **retifma) 3474 { 3475 struct ifmultiaddr *ifma, *ll_ifma; 3476 struct sockaddr *llsa; 3477 struct sockaddr_dl sdl; 3478 int error; 3479 3480 #ifdef INET 3481 IN_MULTI_LIST_UNLOCK_ASSERT(); 3482 #endif 3483 #ifdef INET6 3484 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3485 #endif 3486 /* 3487 * If the address is already present, return a new reference to it; 3488 * otherwise, allocate storage and set up a new address. 3489 */ 3490 IF_ADDR_WLOCK(ifp); 3491 ifma = if_findmulti(ifp, sa); 3492 if (ifma != NULL) { 3493 ifma->ifma_refcount++; 3494 if (retifma != NULL) 3495 *retifma = ifma; 3496 IF_ADDR_WUNLOCK(ifp); 3497 return (0); 3498 } 3499 3500 /* 3501 * The address isn't already present; resolve the protocol address 3502 * into a link layer address, and then look that up, bump its 3503 * refcount or allocate an ifma for that also. 3504 * Most link layer resolving functions returns address data which 3505 * fits inside default sockaddr_dl structure. However callback 3506 * can allocate another sockaddr structure, in that case we need to 3507 * free it later. 3508 */ 3509 llsa = NULL; 3510 ll_ifma = NULL; 3511 if (ifp->if_resolvemulti != NULL) { 3512 /* Provide called function with buffer size information */ 3513 sdl.sdl_len = sizeof(sdl); 3514 llsa = (struct sockaddr *)&sdl; 3515 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3516 if (error) 3517 goto unlock_out; 3518 } 3519 3520 /* 3521 * Allocate the new address. Don't hook it up yet, as we may also 3522 * need to allocate a link layer multicast address. 3523 */ 3524 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3525 if (ifma == NULL) { 3526 error = ENOMEM; 3527 goto free_llsa_out; 3528 } 3529 3530 /* 3531 * If a link layer address is found, we'll need to see if it's 3532 * already present in the address list, or allocate is as well. 3533 * When this block finishes, the link layer address will be on the 3534 * list. 3535 */ 3536 if (llsa != NULL) { 3537 ll_ifma = if_findmulti(ifp, llsa); 3538 if (ll_ifma == NULL) { 3539 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3540 if (ll_ifma == NULL) { 3541 --ifma->ifma_refcount; 3542 if_freemulti(ifma); 3543 error = ENOMEM; 3544 goto free_llsa_out; 3545 } 3546 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3547 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3548 ifma_link); 3549 } else 3550 ll_ifma->ifma_refcount++; 3551 ifma->ifma_llifma = ll_ifma; 3552 } 3553 3554 /* 3555 * We now have a new multicast address, ifma, and possibly a new or 3556 * referenced link layer address. Add the primary address to the 3557 * ifnet address list. 3558 */ 3559 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3560 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3561 3562 if (retifma != NULL) 3563 *retifma = ifma; 3564 3565 /* 3566 * Must generate the message while holding the lock so that 'ifma' 3567 * pointer is still valid. 3568 */ 3569 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3570 IF_ADDR_WUNLOCK(ifp); 3571 3572 /* 3573 * We are certain we have added something, so call down to the 3574 * interface to let them know about it. 3575 */ 3576 if (ifp->if_ioctl != NULL) { 3577 if (THREAD_CAN_SLEEP()) 3578 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3579 else 3580 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3581 } 3582 3583 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3584 link_free_sdl(llsa); 3585 3586 return (0); 3587 3588 free_llsa_out: 3589 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3590 link_free_sdl(llsa); 3591 3592 unlock_out: 3593 IF_ADDR_WUNLOCK(ifp); 3594 return (error); 3595 } 3596 3597 static void 3598 if_siocaddmulti(void *arg, int pending) 3599 { 3600 struct ifnet *ifp; 3601 3602 ifp = arg; 3603 #ifdef DIAGNOSTIC 3604 if (pending > 1) 3605 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3606 #endif 3607 CURVNET_SET(ifp->if_vnet); 3608 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3609 CURVNET_RESTORE(); 3610 } 3611 3612 /* 3613 * Delete a multicast group membership by network-layer group address. 3614 * 3615 * Returns ENOENT if the entry could not be found. If ifp no longer 3616 * exists, results are undefined. This entry point should only be used 3617 * from subsystems which do appropriate locking to hold ifp for the 3618 * duration of the call. 3619 * Network-layer protocol domains must use if_delmulti_ifma(). 3620 */ 3621 int 3622 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3623 { 3624 struct ifmultiaddr *ifma; 3625 int lastref; 3626 3627 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3628 3629 IF_ADDR_WLOCK(ifp); 3630 lastref = 0; 3631 ifma = if_findmulti(ifp, sa); 3632 if (ifma != NULL) 3633 lastref = if_delmulti_locked(ifp, ifma, 0); 3634 IF_ADDR_WUNLOCK(ifp); 3635 3636 if (ifma == NULL) 3637 return (ENOENT); 3638 3639 if (lastref && ifp->if_ioctl != NULL) { 3640 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3641 } 3642 3643 return (0); 3644 } 3645 3646 /* 3647 * Delete all multicast group membership for an interface. 3648 * Should be used to quickly flush all multicast filters. 3649 */ 3650 void 3651 if_delallmulti(struct ifnet *ifp) 3652 { 3653 struct ifmultiaddr *ifma; 3654 struct ifmultiaddr *next; 3655 3656 IF_ADDR_WLOCK(ifp); 3657 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3658 if_delmulti_locked(ifp, ifma, 0); 3659 IF_ADDR_WUNLOCK(ifp); 3660 } 3661 3662 void 3663 if_delmulti_ifma(struct ifmultiaddr *ifma) 3664 { 3665 if_delmulti_ifma_flags(ifma, 0); 3666 } 3667 3668 /* 3669 * Delete a multicast group membership by group membership pointer. 3670 * Network-layer protocol domains must use this routine. 3671 * 3672 * It is safe to call this routine if the ifp disappeared. 3673 */ 3674 void 3675 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3676 { 3677 struct ifnet *ifp; 3678 int lastref; 3679 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3680 #ifdef INET 3681 IN_MULTI_LIST_UNLOCK_ASSERT(); 3682 #endif 3683 ifp = ifma->ifma_ifp; 3684 #ifdef DIAGNOSTIC 3685 if (ifp == NULL) { 3686 printf("%s: ifma_ifp seems to be detached\n", __func__); 3687 } else { 3688 struct epoch_tracker et; 3689 struct ifnet *oifp; 3690 3691 NET_EPOCH_ENTER(et); 3692 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3693 if (ifp == oifp) 3694 break; 3695 NET_EPOCH_EXIT(et); 3696 if (ifp != oifp) 3697 ifp = NULL; 3698 } 3699 #endif 3700 /* 3701 * If and only if the ifnet instance exists: Acquire the address lock. 3702 */ 3703 if (ifp != NULL) 3704 IF_ADDR_WLOCK(ifp); 3705 3706 lastref = if_delmulti_locked(ifp, ifma, flags); 3707 3708 if (ifp != NULL) { 3709 /* 3710 * If and only if the ifnet instance exists: 3711 * Release the address lock. 3712 * If the group was left: update the hardware hash filter. 3713 */ 3714 IF_ADDR_WUNLOCK(ifp); 3715 if (lastref && ifp->if_ioctl != NULL) { 3716 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3717 } 3718 } 3719 } 3720 3721 /* 3722 * Perform deletion of network-layer and/or link-layer multicast address. 3723 * 3724 * Return 0 if the reference count was decremented. 3725 * Return 1 if the final reference was released, indicating that the 3726 * hardware hash filter should be reprogrammed. 3727 */ 3728 static int 3729 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3730 { 3731 struct ifmultiaddr *ll_ifma; 3732 3733 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3734 KASSERT(ifma->ifma_ifp == ifp, 3735 ("%s: inconsistent ifp %p", __func__, ifp)); 3736 IF_ADDR_WLOCK_ASSERT(ifp); 3737 } 3738 3739 ifp = ifma->ifma_ifp; 3740 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3741 3742 /* 3743 * If the ifnet is detaching, null out references to ifnet, 3744 * so that upper protocol layers will notice, and not attempt 3745 * to obtain locks for an ifnet which no longer exists. The 3746 * routing socket announcement must happen before the ifnet 3747 * instance is detached from the system. 3748 */ 3749 if (detaching) { 3750 #ifdef DIAGNOSTIC 3751 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3752 #endif 3753 /* 3754 * ifp may already be nulled out if we are being reentered 3755 * to delete the ll_ifma. 3756 */ 3757 if (ifp != NULL) { 3758 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3759 ifma->ifma_ifp = NULL; 3760 } 3761 } 3762 3763 if (--ifma->ifma_refcount > 0) 3764 return 0; 3765 3766 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3767 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3768 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3769 } 3770 /* 3771 * If this ifma is a network-layer ifma, a link-layer ifma may 3772 * have been associated with it. Release it first if so. 3773 */ 3774 ll_ifma = ifma->ifma_llifma; 3775 if (ll_ifma != NULL) { 3776 KASSERT(ifma->ifma_lladdr != NULL, 3777 ("%s: llifma w/o lladdr", __func__)); 3778 if (detaching) 3779 ll_ifma->ifma_ifp = NULL; /* XXX */ 3780 if (--ll_ifma->ifma_refcount == 0) { 3781 if (ifp != NULL) { 3782 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3783 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3784 ifma_link); 3785 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3786 } 3787 } 3788 if_freemulti(ll_ifma); 3789 } 3790 } 3791 #ifdef INVARIANTS 3792 if (ifp) { 3793 struct ifmultiaddr *ifmatmp; 3794 3795 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3796 MPASS(ifma != ifmatmp); 3797 } 3798 #endif 3799 if_freemulti(ifma); 3800 /* 3801 * The last reference to this instance of struct ifmultiaddr 3802 * was released; the hardware should be notified of this change. 3803 */ 3804 return 1; 3805 } 3806 3807 /* 3808 * Set the link layer address on an interface. 3809 * 3810 * At this time we only support certain types of interfaces, 3811 * and we don't allow the length of the address to change. 3812 * 3813 * Set noinline to be dtrace-friendly 3814 */ 3815 __noinline int 3816 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3817 { 3818 struct sockaddr_dl *sdl; 3819 struct ifaddr *ifa; 3820 struct ifreq ifr; 3821 3822 ifa = ifp->if_addr; 3823 if (ifa == NULL) 3824 return (EINVAL); 3825 3826 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3827 if (sdl == NULL) 3828 return (EINVAL); 3829 3830 if (len != sdl->sdl_alen) /* don't allow length to change */ 3831 return (EINVAL); 3832 3833 switch (ifp->if_type) { 3834 case IFT_ETHER: 3835 case IFT_XETHER: 3836 case IFT_L2VLAN: 3837 case IFT_BRIDGE: 3838 case IFT_IEEE8023ADLAG: 3839 bcopy(lladdr, LLADDR(sdl), len); 3840 break; 3841 default: 3842 return (ENODEV); 3843 } 3844 3845 /* 3846 * If the interface is already up, we need 3847 * to re-init it in order to reprogram its 3848 * address filter. 3849 */ 3850 if ((ifp->if_flags & IFF_UP) != 0) { 3851 if (ifp->if_ioctl) { 3852 ifp->if_flags &= ~IFF_UP; 3853 ifr.ifr_flags = ifp->if_flags & 0xffff; 3854 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3855 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3856 ifp->if_flags |= IFF_UP; 3857 ifr.ifr_flags = ifp->if_flags & 0xffff; 3858 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3859 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3860 } 3861 } 3862 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3863 3864 return (0); 3865 } 3866 3867 /* 3868 * Compat function for handling basic encapsulation requests. 3869 * Not converted stacks (FDDI, IB, ..) supports traditional 3870 * output model: ARP (and other similar L2 protocols) are handled 3871 * inside output routine, arpresolve/nd6_resolve() returns MAC 3872 * address instead of full prepend. 3873 * 3874 * This function creates calculated header==MAC for IPv4/IPv6 and 3875 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3876 * address families. 3877 */ 3878 static int 3879 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3880 { 3881 3882 if (req->rtype != IFENCAP_LL) 3883 return (EOPNOTSUPP); 3884 3885 if (req->bufsize < req->lladdr_len) 3886 return (ENOMEM); 3887 3888 switch (req->family) { 3889 case AF_INET: 3890 case AF_INET6: 3891 break; 3892 default: 3893 return (EAFNOSUPPORT); 3894 } 3895 3896 /* Copy lladdr to storage as is */ 3897 memmove(req->buf, req->lladdr, req->lladdr_len); 3898 req->bufsize = req->lladdr_len; 3899 req->lladdr_off = 0; 3900 3901 return (0); 3902 } 3903 3904 /* 3905 * Tunnel interfaces can nest, also they may cause infinite recursion 3906 * calls when misconfigured. We'll prevent this by detecting loops. 3907 * High nesting level may cause stack exhaustion. We'll prevent this 3908 * by introducing upper limit. 3909 * 3910 * Return 0, if tunnel nesting count is equal or less than limit. 3911 */ 3912 int 3913 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3914 int limit) 3915 { 3916 struct m_tag *mtag; 3917 int count; 3918 3919 count = 1; 3920 mtag = NULL; 3921 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3922 if (*(struct ifnet **)(mtag + 1) == ifp) { 3923 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3924 return (EIO); 3925 } 3926 count++; 3927 } 3928 if (count > limit) { 3929 log(LOG_NOTICE, 3930 "%s: if_output recursively called too many times(%d)\n", 3931 if_name(ifp), count); 3932 return (EIO); 3933 } 3934 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3935 if (mtag == NULL) 3936 return (ENOMEM); 3937 *(struct ifnet **)(mtag + 1) = ifp; 3938 m_tag_prepend(m, mtag); 3939 return (0); 3940 } 3941 3942 /* 3943 * Get the link layer address that was read from the hardware at attach. 3944 * 3945 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3946 * their component interfaces as IFT_IEEE8023ADLAG. 3947 */ 3948 int 3949 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3950 { 3951 3952 if (ifp->if_hw_addr == NULL) 3953 return (ENODEV); 3954 3955 switch (ifp->if_type) { 3956 case IFT_ETHER: 3957 case IFT_IEEE8023ADLAG: 3958 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3959 return (0); 3960 default: 3961 return (ENODEV); 3962 } 3963 } 3964 3965 /* 3966 * The name argument must be a pointer to storage which will last as 3967 * long as the interface does. For physical devices, the result of 3968 * device_get_name(dev) is a good choice and for pseudo-devices a 3969 * static string works well. 3970 */ 3971 void 3972 if_initname(struct ifnet *ifp, const char *name, int unit) 3973 { 3974 ifp->if_dname = name; 3975 ifp->if_dunit = unit; 3976 if (unit != IF_DUNIT_NONE) 3977 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3978 else 3979 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3980 } 3981 3982 static int 3983 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 3984 { 3985 char if_fmt[256]; 3986 3987 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 3988 vlog(pri, if_fmt, ap); 3989 return (0); 3990 } 3991 3992 3993 int 3994 if_printf(struct ifnet *ifp, const char *fmt, ...) 3995 { 3996 va_list ap; 3997 3998 va_start(ap, fmt); 3999 if_vlog(ifp, LOG_INFO, fmt, ap); 4000 va_end(ap); 4001 return (0); 4002 } 4003 4004 int 4005 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 4006 { 4007 va_list ap; 4008 4009 va_start(ap, fmt); 4010 if_vlog(ifp, pri, fmt, ap); 4011 va_end(ap); 4012 return (0); 4013 } 4014 4015 void 4016 if_start(struct ifnet *ifp) 4017 { 4018 4019 (*(ifp)->if_start)(ifp); 4020 } 4021 4022 /* 4023 * Backwards compatibility interface for drivers 4024 * that have not implemented it 4025 */ 4026 static int 4027 if_transmit(struct ifnet *ifp, struct mbuf *m) 4028 { 4029 int error; 4030 4031 IFQ_HANDOFF(ifp, m, error); 4032 return (error); 4033 } 4034 4035 static void 4036 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4037 { 4038 4039 m_freem(m); 4040 } 4041 4042 int 4043 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4044 { 4045 int active = 0; 4046 4047 IF_LOCK(ifq); 4048 if (_IF_QFULL(ifq)) { 4049 IF_UNLOCK(ifq); 4050 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4051 m_freem(m); 4052 return (0); 4053 } 4054 if (ifp != NULL) { 4055 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4056 if (m->m_flags & (M_BCAST|M_MCAST)) 4057 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4058 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4059 } 4060 _IF_ENQUEUE(ifq, m); 4061 IF_UNLOCK(ifq); 4062 if (ifp != NULL && !active) 4063 (*(ifp)->if_start)(ifp); 4064 return (1); 4065 } 4066 4067 void 4068 if_register_com_alloc(u_char type, 4069 if_com_alloc_t *a, if_com_free_t *f) 4070 { 4071 4072 KASSERT(if_com_alloc[type] == NULL, 4073 ("if_register_com_alloc: %d already registered", type)); 4074 KASSERT(if_com_free[type] == NULL, 4075 ("if_register_com_alloc: %d free already registered", type)); 4076 4077 if_com_alloc[type] = a; 4078 if_com_free[type] = f; 4079 } 4080 4081 void 4082 if_deregister_com_alloc(u_char type) 4083 { 4084 4085 KASSERT(if_com_alloc[type] != NULL, 4086 ("if_deregister_com_alloc: %d not registered", type)); 4087 KASSERT(if_com_free[type] != NULL, 4088 ("if_deregister_com_alloc: %d free not registered", type)); 4089 4090 /* 4091 * Ensure all pending EPOCH(9) callbacks have been executed. This 4092 * fixes issues about late invocation of if_destroy(), which leads 4093 * to memory leak from if_com_alloc[type] allocated if_l2com. 4094 */ 4095 epoch_drain_callbacks(net_epoch_preempt); 4096 4097 if_com_alloc[type] = NULL; 4098 if_com_free[type] = NULL; 4099 } 4100 4101 /* API for driver access to network stack owned ifnet.*/ 4102 uint64_t 4103 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4104 { 4105 uint64_t oldbrate; 4106 4107 oldbrate = ifp->if_baudrate; 4108 ifp->if_baudrate = baudrate; 4109 return (oldbrate); 4110 } 4111 4112 uint64_t 4113 if_getbaudrate(if_t ifp) 4114 { 4115 4116 return (((struct ifnet *)ifp)->if_baudrate); 4117 } 4118 4119 int 4120 if_setcapabilities(if_t ifp, int capabilities) 4121 { 4122 ((struct ifnet *)ifp)->if_capabilities = capabilities; 4123 return (0); 4124 } 4125 4126 int 4127 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4128 { 4129 ((struct ifnet *)ifp)->if_capabilities |= setbit; 4130 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 4131 4132 return (0); 4133 } 4134 4135 int 4136 if_getcapabilities(if_t ifp) 4137 { 4138 return ((struct ifnet *)ifp)->if_capabilities; 4139 } 4140 4141 int 4142 if_setcapenable(if_t ifp, int capabilities) 4143 { 4144 ((struct ifnet *)ifp)->if_capenable = capabilities; 4145 return (0); 4146 } 4147 4148 int 4149 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4150 { 4151 if(setcap) 4152 ((struct ifnet *)ifp)->if_capenable |= setcap; 4153 if(clearcap) 4154 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 4155 4156 return (0); 4157 } 4158 4159 const char * 4160 if_getdname(if_t ifp) 4161 { 4162 return ((struct ifnet *)ifp)->if_dname; 4163 } 4164 4165 int 4166 if_togglecapenable(if_t ifp, int togglecap) 4167 { 4168 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 4169 return (0); 4170 } 4171 4172 int 4173 if_getcapenable(if_t ifp) 4174 { 4175 return ((struct ifnet *)ifp)->if_capenable; 4176 } 4177 4178 /* 4179 * This is largely undesirable because it ties ifnet to a device, but does 4180 * provide flexiblity for an embedded product vendor. Should be used with 4181 * the understanding that it violates the interface boundaries, and should be 4182 * a last resort only. 4183 */ 4184 int 4185 if_setdev(if_t ifp, void *dev) 4186 { 4187 return (0); 4188 } 4189 4190 int 4191 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4192 { 4193 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 4194 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 4195 4196 return (0); 4197 } 4198 4199 int 4200 if_getdrvflags(if_t ifp) 4201 { 4202 return ((struct ifnet *)ifp)->if_drv_flags; 4203 } 4204 4205 int 4206 if_setdrvflags(if_t ifp, int flags) 4207 { 4208 ((struct ifnet *)ifp)->if_drv_flags = flags; 4209 return (0); 4210 } 4211 4212 int 4213 if_setflags(if_t ifp, int flags) 4214 { 4215 4216 ifp->if_flags = flags; 4217 return (0); 4218 } 4219 4220 int 4221 if_setflagbits(if_t ifp, int set, int clear) 4222 { 4223 ((struct ifnet *)ifp)->if_flags |= set; 4224 ((struct ifnet *)ifp)->if_flags &= ~clear; 4225 4226 return (0); 4227 } 4228 4229 int 4230 if_getflags(if_t ifp) 4231 { 4232 return ((struct ifnet *)ifp)->if_flags; 4233 } 4234 4235 int 4236 if_clearhwassist(if_t ifp) 4237 { 4238 ((struct ifnet *)ifp)->if_hwassist = 0; 4239 return (0); 4240 } 4241 4242 int 4243 if_sethwassistbits(if_t ifp, int toset, int toclear) 4244 { 4245 ((struct ifnet *)ifp)->if_hwassist |= toset; 4246 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4247 4248 return (0); 4249 } 4250 4251 int 4252 if_sethwassist(if_t ifp, int hwassist_bit) 4253 { 4254 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4255 return (0); 4256 } 4257 4258 int 4259 if_gethwassist(if_t ifp) 4260 { 4261 return ((struct ifnet *)ifp)->if_hwassist; 4262 } 4263 4264 int 4265 if_setmtu(if_t ifp, int mtu) 4266 { 4267 ((struct ifnet *)ifp)->if_mtu = mtu; 4268 return (0); 4269 } 4270 4271 int 4272 if_getmtu(if_t ifp) 4273 { 4274 return ((struct ifnet *)ifp)->if_mtu; 4275 } 4276 4277 int 4278 if_getmtu_family(if_t ifp, int family) 4279 { 4280 struct domain *dp; 4281 4282 for (dp = domains; dp; dp = dp->dom_next) { 4283 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4284 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4285 } 4286 4287 return (((struct ifnet *)ifp)->if_mtu); 4288 } 4289 4290 /* 4291 * Methods for drivers to access interface unicast and multicast 4292 * link level addresses. Driver shall not know 'struct ifaddr' neither 4293 * 'struct ifmultiaddr'. 4294 */ 4295 u_int 4296 if_lladdr_count(if_t ifp) 4297 { 4298 struct epoch_tracker et; 4299 struct ifaddr *ifa; 4300 u_int count; 4301 4302 count = 0; 4303 NET_EPOCH_ENTER(et); 4304 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4305 if (ifa->ifa_addr->sa_family == AF_LINK) 4306 count++; 4307 NET_EPOCH_EXIT(et); 4308 4309 return (count); 4310 } 4311 4312 u_int 4313 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4314 { 4315 struct epoch_tracker et; 4316 struct ifaddr *ifa; 4317 u_int count; 4318 4319 MPASS(cb); 4320 4321 count = 0; 4322 NET_EPOCH_ENTER(et); 4323 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4324 if (ifa->ifa_addr->sa_family != AF_LINK) 4325 continue; 4326 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4327 count); 4328 } 4329 NET_EPOCH_EXIT(et); 4330 4331 return (count); 4332 } 4333 4334 u_int 4335 if_llmaddr_count(if_t ifp) 4336 { 4337 struct epoch_tracker et; 4338 struct ifmultiaddr *ifma; 4339 int count; 4340 4341 count = 0; 4342 NET_EPOCH_ENTER(et); 4343 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4344 if (ifma->ifma_addr->sa_family == AF_LINK) 4345 count++; 4346 NET_EPOCH_EXIT(et); 4347 4348 return (count); 4349 } 4350 4351 u_int 4352 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4353 { 4354 struct epoch_tracker et; 4355 struct ifmultiaddr *ifma; 4356 u_int count; 4357 4358 MPASS(cb); 4359 4360 count = 0; 4361 NET_EPOCH_ENTER(et); 4362 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4363 if (ifma->ifma_addr->sa_family != AF_LINK) 4364 continue; 4365 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4366 count); 4367 } 4368 NET_EPOCH_EXIT(et); 4369 4370 return (count); 4371 } 4372 4373 int 4374 if_setsoftc(if_t ifp, void *softc) 4375 { 4376 ((struct ifnet *)ifp)->if_softc = softc; 4377 return (0); 4378 } 4379 4380 void * 4381 if_getsoftc(if_t ifp) 4382 { 4383 return ((struct ifnet *)ifp)->if_softc; 4384 } 4385 4386 void 4387 if_setrcvif(struct mbuf *m, if_t ifp) 4388 { 4389 4390 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4391 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4392 } 4393 4394 void 4395 if_setvtag(struct mbuf *m, uint16_t tag) 4396 { 4397 m->m_pkthdr.ether_vtag = tag; 4398 } 4399 4400 uint16_t 4401 if_getvtag(struct mbuf *m) 4402 { 4403 4404 return (m->m_pkthdr.ether_vtag); 4405 } 4406 4407 int 4408 if_sendq_empty(if_t ifp) 4409 { 4410 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4411 } 4412 4413 struct ifaddr * 4414 if_getifaddr(if_t ifp) 4415 { 4416 return ((struct ifnet *)ifp)->if_addr; 4417 } 4418 4419 int 4420 if_getamcount(if_t ifp) 4421 { 4422 return ((struct ifnet *)ifp)->if_amcount; 4423 } 4424 4425 int 4426 if_setsendqready(if_t ifp) 4427 { 4428 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4429 return (0); 4430 } 4431 4432 int 4433 if_setsendqlen(if_t ifp, int tx_desc_count) 4434 { 4435 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4436 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4437 4438 return (0); 4439 } 4440 4441 int 4442 if_vlantrunkinuse(if_t ifp) 4443 { 4444 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4445 } 4446 4447 int 4448 if_input(if_t ifp, struct mbuf* sendmp) 4449 { 4450 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4451 return (0); 4452 4453 } 4454 4455 struct mbuf * 4456 if_dequeue(if_t ifp) 4457 { 4458 struct mbuf *m; 4459 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4460 4461 return (m); 4462 } 4463 4464 int 4465 if_sendq_prepend(if_t ifp, struct mbuf *m) 4466 { 4467 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4468 return (0); 4469 } 4470 4471 int 4472 if_setifheaderlen(if_t ifp, int len) 4473 { 4474 ((struct ifnet *)ifp)->if_hdrlen = len; 4475 return (0); 4476 } 4477 4478 caddr_t 4479 if_getlladdr(if_t ifp) 4480 { 4481 return (IF_LLADDR((struct ifnet *)ifp)); 4482 } 4483 4484 void * 4485 if_gethandle(u_char type) 4486 { 4487 return (if_alloc(type)); 4488 } 4489 4490 void 4491 if_bpfmtap(if_t ifh, struct mbuf *m) 4492 { 4493 struct ifnet *ifp = (struct ifnet *)ifh; 4494 4495 BPF_MTAP(ifp, m); 4496 } 4497 4498 void 4499 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4500 { 4501 struct ifnet *ifp = (struct ifnet *)ifh; 4502 4503 ETHER_BPF_MTAP(ifp, m); 4504 } 4505 4506 void 4507 if_vlancap(if_t ifh) 4508 { 4509 struct ifnet *ifp = (struct ifnet *)ifh; 4510 VLAN_CAPABILITIES(ifp); 4511 } 4512 4513 int 4514 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4515 { 4516 4517 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax; 4518 return (0); 4519 } 4520 4521 int 4522 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4523 { 4524 4525 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4526 return (0); 4527 } 4528 4529 int 4530 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4531 { 4532 4533 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4534 return (0); 4535 } 4536 4537 u_int 4538 if_gethwtsomax(if_t ifp) 4539 { 4540 4541 return (((struct ifnet *)ifp)->if_hw_tsomax); 4542 } 4543 4544 u_int 4545 if_gethwtsomaxsegcount(if_t ifp) 4546 { 4547 4548 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount); 4549 } 4550 4551 u_int 4552 if_gethwtsomaxsegsize(if_t ifp) 4553 { 4554 4555 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize); 4556 } 4557 4558 void 4559 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4560 { 4561 ((struct ifnet *)ifp)->if_init = init_fn; 4562 } 4563 4564 void 4565 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4566 { 4567 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4568 } 4569 4570 void 4571 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4572 { 4573 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4574 } 4575 4576 void 4577 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4578 { 4579 ((struct ifnet *)ifp)->if_transmit = start_fn; 4580 } 4581 4582 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4583 { 4584 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4585 4586 } 4587 4588 void 4589 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4590 { 4591 4592 ifp->if_get_counter = fn; 4593 } 4594 4595 /* Revisit these - These are inline functions originally. */ 4596 int 4597 drbr_inuse_drv(if_t ifh, struct buf_ring *br) 4598 { 4599 return drbr_inuse(ifh, br); 4600 } 4601 4602 struct mbuf* 4603 drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 4604 { 4605 return drbr_dequeue(ifh, br); 4606 } 4607 4608 int 4609 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 4610 { 4611 return drbr_needs_enqueue(ifh, br); 4612 } 4613 4614 int 4615 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 4616 { 4617 return drbr_enqueue(ifh, br, m); 4618 4619 } 4620