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_inet6.h" 36 #include "opt_inet.h" 37 38 #include <sys/param.h> 39 #include <sys/types.h> 40 #include <sys/conf.h> 41 #include <sys/malloc.h> 42 #include <sys/sbuf.h> 43 #include <sys/bus.h> 44 #include <sys/epoch.h> 45 #include <sys/mbuf.h> 46 #include <sys/systm.h> 47 #include <sys/priv.h> 48 #include <sys/proc.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/protosw.h> 52 #include <sys/kernel.h> 53 #include <sys/lock.h> 54 #include <sys/refcount.h> 55 #include <sys/module.h> 56 #include <sys/rwlock.h> 57 #include <sys/sockio.h> 58 #include <sys/syslog.h> 59 #include <sys/sysctl.h> 60 #include <sys/sysent.h> 61 #include <sys/taskqueue.h> 62 #include <sys/domain.h> 63 #include <sys/jail.h> 64 #include <sys/priv.h> 65 66 #include <machine/stdarg.h> 67 #include <vm/uma.h> 68 69 #include <net/bpf.h> 70 #include <net/ethernet.h> 71 #include <net/if.h> 72 #include <net/if_arp.h> 73 #include <net/if_clone.h> 74 #include <net/if_dl.h> 75 #include <net/if_types.h> 76 #include <net/if_var.h> 77 #include <net/if_media.h> 78 #include <net/if_vlan_var.h> 79 #include <net/radix.h> 80 #include <net/route.h> 81 #include <net/vnet.h> 82 83 #if defined(INET) || defined(INET6) 84 #include <net/ethernet.h> 85 #include <netinet/in.h> 86 #include <netinet/in_var.h> 87 #include <netinet/ip.h> 88 #include <netinet/ip_carp.h> 89 #ifdef INET 90 #include <netinet/if_ether.h> 91 #include <netinet/netdump/netdump.h> 92 #endif /* INET */ 93 #ifdef INET6 94 #include <netinet6/in6_var.h> 95 #include <netinet6/in6_ifattach.h> 96 #endif /* INET6 */ 97 #endif /* INET || INET6 */ 98 99 #include <security/mac/mac_framework.h> 100 101 /* 102 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 103 * and ifr_ifru when it is used in SIOCGIFCONF. 104 */ 105 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 106 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 107 108 __read_mostly epoch_t net_epoch_preempt; 109 __read_mostly epoch_t net_epoch; 110 #ifdef COMPAT_FREEBSD32 111 #include <sys/mount.h> 112 #include <compat/freebsd32/freebsd32.h> 113 114 struct ifreq_buffer32 { 115 uint32_t length; /* (size_t) */ 116 uint32_t buffer; /* (void *) */ 117 }; 118 119 /* 120 * Interface request structure used for socket 121 * ioctl's. All interface ioctl's must have parameter 122 * definitions which begin with ifr_name. The 123 * remainder may be interface specific. 124 */ 125 struct ifreq32 { 126 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 127 union { 128 struct sockaddr ifru_addr; 129 struct sockaddr ifru_dstaddr; 130 struct sockaddr ifru_broadaddr; 131 struct ifreq_buffer32 ifru_buffer; 132 short ifru_flags[2]; 133 short ifru_index; 134 int ifru_jid; 135 int ifru_metric; 136 int ifru_mtu; 137 int ifru_phys; 138 int ifru_media; 139 uint32_t ifru_data; 140 int ifru_cap[2]; 141 u_int ifru_fib; 142 u_char ifru_vlan_pcp; 143 } ifr_ifru; 144 }; 145 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 146 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 147 __offsetof(struct ifreq32, ifr_ifru)); 148 149 struct ifgroupreq32 { 150 char ifgr_name[IFNAMSIZ]; 151 u_int ifgr_len; 152 union { 153 char ifgru_group[IFNAMSIZ]; 154 uint32_t ifgru_groups; 155 } ifgr_ifgru; 156 }; 157 158 struct ifmediareq32 { 159 char ifm_name[IFNAMSIZ]; 160 int ifm_current; 161 int ifm_mask; 162 int ifm_status; 163 int ifm_active; 164 int ifm_count; 165 uint32_t ifm_ulist; /* (int *) */ 166 }; 167 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 168 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 169 170 #define _CASE_IOC_IFGROUPREQ_32(cmd) \ 171 case _IOC_NEWTYPE((cmd), struct ifgroupreq32): 172 #else /* !COMPAT_FREEBSD32 */ 173 #define _CASE_IOC_IFGROUPREQ_32(cmd) 174 #endif /* !COMPAT_FREEBSD32 */ 175 176 #define CASE_IOC_IFGROUPREQ(cmd) \ 177 _CASE_IOC_IFGROUPREQ_32(cmd) \ 178 case (cmd) 179 180 union ifreq_union { 181 struct ifreq ifr; 182 #ifdef COMPAT_FREEBSD32 183 struct ifreq32 ifr32; 184 #endif 185 }; 186 187 union ifgroupreq_union { 188 struct ifgroupreq ifgr; 189 #ifdef COMPAT_FREEBSD32 190 struct ifgroupreq32 ifgr32; 191 #endif 192 }; 193 194 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 195 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 196 197 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 198 &ifqmaxlen, 0, "max send queue size"); 199 200 /* Log link state change events */ 201 static int log_link_state_change = 1; 202 203 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 204 &log_link_state_change, 0, 205 "log interface link state change events"); 206 207 /* Log promiscuous mode change events */ 208 static int log_promisc_mode_change = 1; 209 210 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 211 &log_promisc_mode_change, 1, 212 "log promiscuous mode change events"); 213 214 /* Interface description */ 215 static unsigned int ifdescr_maxlen = 1024; 216 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 217 &ifdescr_maxlen, 0, 218 "administrative maximum length for interface description"); 219 220 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 221 222 /* global sx for non-critical path ifdescr */ 223 static struct sx ifdescr_sx; 224 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 225 226 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 227 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 228 /* These are external hooks for CARP. */ 229 void (*carp_linkstate_p)(struct ifnet *ifp); 230 void (*carp_demote_adj_p)(int, char *); 231 int (*carp_master_p)(struct ifaddr *); 232 #if defined(INET) || defined(INET6) 233 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 234 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 235 const struct sockaddr *sa); 236 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 237 int (*carp_attach_p)(struct ifaddr *, int); 238 void (*carp_detach_p)(struct ifaddr *, bool); 239 #endif 240 #ifdef INET 241 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 242 #endif 243 #ifdef INET6 244 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 245 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 246 const struct in6_addr *taddr); 247 #endif 248 249 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 250 251 /* 252 * XXX: Style; these should be sorted alphabetically, and unprototyped 253 * static functions should be prototyped. Currently they are sorted by 254 * declaration order. 255 */ 256 static void if_attachdomain(void *); 257 static void if_attachdomain1(struct ifnet *); 258 static int ifconf(u_long, caddr_t); 259 static void *if_grow(void); 260 static void if_input_default(struct ifnet *, struct mbuf *); 261 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 262 static void if_route(struct ifnet *, int flag, int fam); 263 static int if_setflag(struct ifnet *, int, int, int *, int); 264 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 265 static void if_unroute(struct ifnet *, int flag, int fam); 266 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 267 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 268 static void do_link_state_change(void *, int); 269 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 270 static int if_getgroupmembers(struct ifgroupreq *); 271 static void if_delgroups(struct ifnet *); 272 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 273 static int if_detach_internal(struct ifnet *, int, struct if_clone **); 274 #ifdef VIMAGE 275 static void if_vmove(struct ifnet *, struct vnet *); 276 #endif 277 278 #ifdef INET6 279 /* 280 * XXX: declare here to avoid to include many inet6 related files.. 281 * should be more generalized? 282 */ 283 extern void nd6_setmtu(struct ifnet *); 284 #endif 285 286 /* ipsec helper hooks */ 287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 289 290 VNET_DEFINE(int, if_index); 291 int ifqmaxlen = IFQ_MAXLEN; 292 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 293 VNET_DEFINE(struct ifgrouphead, ifg_head); 294 295 VNET_DEFINE_STATIC(int, if_indexlim) = 8; 296 297 /* Table of ifnet by index. */ 298 VNET_DEFINE(struct ifnet **, ifindex_table); 299 300 #define V_if_indexlim VNET(if_indexlim) 301 #define V_ifindex_table VNET(ifindex_table) 302 303 /* 304 * The global network interface list (V_ifnet) and related state (such as 305 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 306 * an rwlock. Either may be acquired shared to stablize the list, but both 307 * must be acquired writable to modify the list. This model allows us to 308 * both stablize the interface list during interrupt thread processing, but 309 * also to stablize it over long-running ioctls, without introducing priority 310 * inversions and deadlocks. 311 */ 312 struct rwlock ifnet_rwlock; 313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE); 314 struct sx ifnet_sxlock; 315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 316 317 /* 318 * The allocation of network interfaces is a rather non-atomic affair; we 319 * need to select an index before we are ready to expose the interface for 320 * use, so will use this pointer value to indicate reservation. 321 */ 322 #define IFNET_HOLD (void *)(uintptr_t)(-1) 323 324 static if_com_alloc_t *if_com_alloc[256]; 325 static if_com_free_t *if_com_free[256]; 326 327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 330 331 struct ifnet * 332 ifnet_byindex_locked(u_short idx) 333 { 334 335 if (idx > V_if_index) 336 return (NULL); 337 if (V_ifindex_table[idx] == IFNET_HOLD) 338 return (NULL); 339 return (V_ifindex_table[idx]); 340 } 341 342 struct ifnet * 343 ifnet_byindex(u_short idx) 344 { 345 struct ifnet *ifp; 346 347 ifp = ifnet_byindex_locked(idx); 348 return (ifp); 349 } 350 351 struct ifnet * 352 ifnet_byindex_ref(u_short idx) 353 { 354 struct ifnet *ifp; 355 356 IFNET_RLOCK_NOSLEEP(); 357 ifp = ifnet_byindex_locked(idx); 358 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 359 IFNET_RUNLOCK_NOSLEEP(); 360 return (NULL); 361 } 362 if_ref(ifp); 363 IFNET_RUNLOCK_NOSLEEP(); 364 return (ifp); 365 } 366 367 /* 368 * Allocate an ifindex array entry; return 0 on success or an error on 369 * failure. 370 */ 371 static u_short 372 ifindex_alloc(void **old) 373 { 374 u_short idx; 375 376 IFNET_WLOCK_ASSERT(); 377 /* 378 * Try to find an empty slot below V_if_index. If we fail, take the 379 * next slot. 380 */ 381 for (idx = 1; idx <= V_if_index; idx++) { 382 if (V_ifindex_table[idx] == NULL) 383 break; 384 } 385 386 /* Catch if_index overflow. */ 387 if (idx >= V_if_indexlim) { 388 *old = if_grow(); 389 return (USHRT_MAX); 390 } 391 if (idx > V_if_index) 392 V_if_index = idx; 393 return (idx); 394 } 395 396 static void 397 ifindex_free_locked(u_short idx) 398 { 399 400 IFNET_WLOCK_ASSERT(); 401 402 V_ifindex_table[idx] = NULL; 403 while (V_if_index > 0 && 404 V_ifindex_table[V_if_index] == NULL) 405 V_if_index--; 406 } 407 408 static void 409 ifindex_free(u_short idx) 410 { 411 412 IFNET_WLOCK(); 413 ifindex_free_locked(idx); 414 IFNET_WUNLOCK(); 415 } 416 417 static void 418 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 419 { 420 421 V_ifindex_table[idx] = ifp; 422 } 423 424 struct ifaddr * 425 ifaddr_byindex(u_short idx) 426 { 427 struct ifnet *ifp; 428 struct ifaddr *ifa = NULL; 429 430 IFNET_RLOCK_NOSLEEP(); 431 ifp = ifnet_byindex_locked(idx); 432 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 433 ifa_ref(ifa); 434 IFNET_RUNLOCK_NOSLEEP(); 435 return (ifa); 436 } 437 438 /* 439 * Network interface utility routines. 440 * 441 * Routines with ifa_ifwith* names take sockaddr *'s as 442 * parameters. 443 */ 444 445 static void 446 vnet_if_init(const void *unused __unused) 447 { 448 void *old; 449 450 CK_STAILQ_INIT(&V_ifnet); 451 CK_STAILQ_INIT(&V_ifg_head); 452 IFNET_WLOCK(); 453 old = if_grow(); /* create initial table */ 454 IFNET_WUNLOCK(); 455 epoch_wait_preempt(net_epoch_preempt); 456 free(old, M_IFNET); 457 vnet_if_clone_init(); 458 } 459 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 460 NULL); 461 462 #ifdef VIMAGE 463 static void 464 vnet_if_uninit(const void *unused __unused) 465 { 466 467 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 468 "not empty", __func__, __LINE__, &V_ifnet)); 469 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 470 "not empty", __func__, __LINE__, &V_ifg_head)); 471 472 free((caddr_t)V_ifindex_table, M_IFNET); 473 } 474 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 475 vnet_if_uninit, NULL); 476 477 static void 478 vnet_if_return(const void *unused __unused) 479 { 480 struct ifnet *ifp, *nifp; 481 482 /* Return all inherited interfaces to their parent vnets. */ 483 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 484 if (ifp->if_home_vnet != ifp->if_vnet) 485 if_vmove(ifp, ifp->if_home_vnet); 486 } 487 } 488 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 489 vnet_if_return, NULL); 490 #endif 491 492 493 static void * 494 if_grow(void) 495 { 496 int oldlim; 497 u_int n; 498 struct ifnet **e; 499 void *old; 500 501 old = NULL; 502 IFNET_WLOCK_ASSERT(); 503 oldlim = V_if_indexlim; 504 IFNET_WUNLOCK(); 505 n = (oldlim << 1) * sizeof(*e); 506 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 507 IFNET_WLOCK(); 508 if (V_if_indexlim != oldlim) { 509 free(e, M_IFNET); 510 return (NULL); 511 } 512 if (V_ifindex_table != NULL) { 513 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 514 old = V_ifindex_table; 515 } 516 V_if_indexlim <<= 1; 517 V_ifindex_table = e; 518 return (old); 519 } 520 521 /* 522 * Allocate a struct ifnet and an index for an interface. A layer 2 523 * common structure will also be allocated if an allocation routine is 524 * registered for the passed type. 525 */ 526 struct ifnet * 527 if_alloc(u_char type) 528 { 529 struct ifnet *ifp; 530 u_short idx; 531 void *old; 532 533 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 534 restart: 535 IFNET_WLOCK(); 536 idx = ifindex_alloc(&old); 537 if (__predict_false(idx == USHRT_MAX)) { 538 IFNET_WUNLOCK(); 539 epoch_wait_preempt(net_epoch_preempt); 540 free(old, M_IFNET); 541 goto restart; 542 } 543 ifnet_setbyindex(idx, IFNET_HOLD); 544 IFNET_WUNLOCK(); 545 ifp->if_index = idx; 546 ifp->if_type = type; 547 ifp->if_alloctype = type; 548 #ifdef VIMAGE 549 ifp->if_vnet = curvnet; 550 #endif 551 if (if_com_alloc[type] != NULL) { 552 ifp->if_l2com = if_com_alloc[type](type, ifp); 553 if (ifp->if_l2com == NULL) { 554 free(ifp, M_IFNET); 555 ifindex_free(idx); 556 return (NULL); 557 } 558 } 559 560 IF_ADDR_LOCK_INIT(ifp); 561 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 562 ifp->if_afdata_initialized = 0; 563 IF_AFDATA_LOCK_INIT(ifp); 564 CK_STAILQ_INIT(&ifp->if_addrhead); 565 CK_STAILQ_INIT(&ifp->if_multiaddrs); 566 CK_STAILQ_INIT(&ifp->if_groups); 567 #ifdef MAC 568 mac_ifnet_init(ifp); 569 #endif 570 ifq_init(&ifp->if_snd, ifp); 571 572 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 573 for (int i = 0; i < IFCOUNTERS; i++) 574 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 575 ifp->if_get_counter = if_get_counter_default; 576 ifp->if_pcp = IFNET_PCP_NONE; 577 ifnet_setbyindex(ifp->if_index, ifp); 578 return (ifp); 579 } 580 581 /* 582 * Do the actual work of freeing a struct ifnet, and layer 2 common 583 * structure. This call is made when the last reference to an 584 * interface is released. 585 */ 586 static void 587 if_free_internal(struct ifnet *ifp) 588 { 589 590 KASSERT((ifp->if_flags & IFF_DYING), 591 ("if_free_internal: interface not dying")); 592 593 if (if_com_free[ifp->if_alloctype] != NULL) 594 if_com_free[ifp->if_alloctype](ifp->if_l2com, 595 ifp->if_alloctype); 596 597 #ifdef MAC 598 mac_ifnet_destroy(ifp); 599 #endif /* MAC */ 600 if (ifp->if_description != NULL) 601 free(ifp->if_description, M_IFDESCR); 602 IF_AFDATA_DESTROY(ifp); 603 IF_ADDR_LOCK_DESTROY(ifp); 604 ifq_delete(&ifp->if_snd); 605 606 for (int i = 0; i < IFCOUNTERS; i++) 607 counter_u64_free(ifp->if_counters[i]); 608 609 free(ifp, M_IFNET); 610 } 611 612 static void 613 if_destroy(epoch_context_t ctx) 614 { 615 struct ifnet *ifp; 616 617 ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 618 if_free_internal(ifp); 619 } 620 621 /* 622 * Deregister an interface and free the associated storage. 623 */ 624 void 625 if_free(struct ifnet *ifp) 626 { 627 628 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 629 630 CURVNET_SET_QUIET(ifp->if_vnet); 631 IFNET_WLOCK(); 632 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 633 ("%s: freeing unallocated ifnet", ifp->if_xname)); 634 635 ifindex_free_locked(ifp->if_index); 636 IFNET_WUNLOCK(); 637 638 if (refcount_release(&ifp->if_refcount)) 639 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy); 640 CURVNET_RESTORE(); 641 } 642 643 /* 644 * Interfaces to keep an ifnet type-stable despite the possibility of the 645 * driver calling if_free(). If there are additional references, we defer 646 * freeing the underlying data structure. 647 */ 648 void 649 if_ref(struct ifnet *ifp) 650 { 651 652 /* We don't assert the ifnet list lock here, but arguably should. */ 653 refcount_acquire(&ifp->if_refcount); 654 } 655 656 void 657 if_rele(struct ifnet *ifp) 658 { 659 660 if (!refcount_release(&ifp->if_refcount)) 661 return; 662 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy); 663 } 664 665 void 666 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 667 { 668 669 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 670 671 if (ifq->ifq_maxlen == 0) 672 ifq->ifq_maxlen = ifqmaxlen; 673 674 ifq->altq_type = 0; 675 ifq->altq_disc = NULL; 676 ifq->altq_flags &= ALTQF_CANTCHANGE; 677 ifq->altq_tbr = NULL; 678 ifq->altq_ifp = ifp; 679 } 680 681 void 682 ifq_delete(struct ifaltq *ifq) 683 { 684 mtx_destroy(&ifq->ifq_mtx); 685 } 686 687 /* 688 * Perform generic interface initialization tasks and attach the interface 689 * to the list of "active" interfaces. If vmove flag is set on entry 690 * to if_attach_internal(), perform only a limited subset of initialization 691 * tasks, given that we are moving from one vnet to another an ifnet which 692 * has already been fully initialized. 693 * 694 * Note that if_detach_internal() removes group membership unconditionally 695 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 696 * Thus, when if_vmove() is applied to a cloned interface, group membership 697 * is lost while a cloned one always joins a group whose name is 698 * ifc->ifc_name. To recover this after if_detach_internal() and 699 * if_attach_internal(), the cloner should be specified to 700 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 701 * attempts to join a group whose name is ifc->ifc_name. 702 * 703 * XXX: 704 * - The decision to return void and thus require this function to 705 * succeed is questionable. 706 * - We should probably do more sanity checking. For instance we don't 707 * do anything to insure if_xname is unique or non-empty. 708 */ 709 void 710 if_attach(struct ifnet *ifp) 711 { 712 713 if_attach_internal(ifp, 0, NULL); 714 } 715 716 /* 717 * Compute the least common TSO limit. 718 */ 719 void 720 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 721 { 722 /* 723 * 1) If there is no limit currently, take the limit from 724 * the network adapter. 725 * 726 * 2) If the network adapter has a limit below the current 727 * limit, apply it. 728 */ 729 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 730 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 731 pmax->tsomaxbytes = ifp->if_hw_tsomax; 732 } 733 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 734 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 735 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 736 } 737 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 738 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 739 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 740 } 741 } 742 743 /* 744 * Update TSO limit of a network adapter. 745 * 746 * Returns zero if no change. Else non-zero. 747 */ 748 int 749 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 750 { 751 int retval = 0; 752 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 753 ifp->if_hw_tsomax = pmax->tsomaxbytes; 754 retval++; 755 } 756 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 757 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 758 retval++; 759 } 760 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 761 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 762 retval++; 763 } 764 return (retval); 765 } 766 767 static void 768 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 769 { 770 unsigned socksize, ifasize; 771 int namelen, masklen; 772 struct sockaddr_dl *sdl; 773 struct ifaddr *ifa; 774 775 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 776 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 777 ifp->if_xname); 778 779 #ifdef VIMAGE 780 ifp->if_vnet = curvnet; 781 if (ifp->if_home_vnet == NULL) 782 ifp->if_home_vnet = curvnet; 783 #endif 784 785 if_addgroup(ifp, IFG_ALL); 786 787 /* Restore group membership for cloned interfaces. */ 788 if (vmove && ifc != NULL) 789 if_clone_addgroup(ifp, ifc); 790 791 getmicrotime(&ifp->if_lastchange); 792 ifp->if_epoch = time_uptime; 793 794 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 795 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 796 ("transmit and qflush must both either be set or both be NULL")); 797 if (ifp->if_transmit == NULL) { 798 ifp->if_transmit = if_transmit; 799 ifp->if_qflush = if_qflush; 800 } 801 if (ifp->if_input == NULL) 802 ifp->if_input = if_input_default; 803 804 if (ifp->if_requestencap == NULL) 805 ifp->if_requestencap = if_requestencap_default; 806 807 if (!vmove) { 808 #ifdef MAC 809 mac_ifnet_create(ifp); 810 #endif 811 812 /* 813 * Create a Link Level name for this device. 814 */ 815 namelen = strlen(ifp->if_xname); 816 /* 817 * Always save enough space for any possiable name so we 818 * can do a rename in place later. 819 */ 820 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 821 socksize = masklen + ifp->if_addrlen; 822 if (socksize < sizeof(*sdl)) 823 socksize = sizeof(*sdl); 824 socksize = roundup2(socksize, sizeof(long)); 825 ifasize = sizeof(*ifa) + 2 * socksize; 826 ifa = ifa_alloc(ifasize, M_WAITOK); 827 sdl = (struct sockaddr_dl *)(ifa + 1); 828 sdl->sdl_len = socksize; 829 sdl->sdl_family = AF_LINK; 830 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 831 sdl->sdl_nlen = namelen; 832 sdl->sdl_index = ifp->if_index; 833 sdl->sdl_type = ifp->if_type; 834 ifp->if_addr = ifa; 835 ifa->ifa_ifp = ifp; 836 ifa->ifa_rtrequest = link_rtrequest; 837 ifa->ifa_addr = (struct sockaddr *)sdl; 838 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 839 ifa->ifa_netmask = (struct sockaddr *)sdl; 840 sdl->sdl_len = masklen; 841 while (namelen != 0) 842 sdl->sdl_data[--namelen] = 0xff; 843 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 844 /* Reliably crash if used uninitialized. */ 845 ifp->if_broadcastaddr = NULL; 846 847 if (ifp->if_type == IFT_ETHER) { 848 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 849 M_WAITOK | M_ZERO); 850 } 851 852 #if defined(INET) || defined(INET6) 853 /* Use defaults for TSO, if nothing is set */ 854 if (ifp->if_hw_tsomax == 0 && 855 ifp->if_hw_tsomaxsegcount == 0 && 856 ifp->if_hw_tsomaxsegsize == 0) { 857 /* 858 * The TSO defaults needs to be such that an 859 * NFS mbuf list of 35 mbufs totalling just 860 * below 64K works and that a chain of mbufs 861 * can be defragged into at most 32 segments: 862 */ 863 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 864 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 865 ifp->if_hw_tsomaxsegcount = 35; 866 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 867 868 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 869 if (ifp->if_capabilities & IFCAP_TSO) { 870 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 871 ifp->if_hw_tsomax, 872 ifp->if_hw_tsomaxsegcount, 873 ifp->if_hw_tsomaxsegsize); 874 } 875 } 876 #endif 877 } 878 #ifdef VIMAGE 879 else { 880 /* 881 * Update the interface index in the link layer address 882 * of the interface. 883 */ 884 for (ifa = ifp->if_addr; ifa != NULL; 885 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 886 if (ifa->ifa_addr->sa_family == AF_LINK) { 887 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 888 sdl->sdl_index = ifp->if_index; 889 } 890 } 891 } 892 #endif 893 894 IFNET_WLOCK(); 895 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 896 #ifdef VIMAGE 897 curvnet->vnet_ifcnt++; 898 #endif 899 IFNET_WUNLOCK(); 900 901 if (domain_init_status >= 2) 902 if_attachdomain1(ifp); 903 904 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 905 if (IS_DEFAULT_VNET(curvnet)) 906 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 907 908 /* Announce the interface. */ 909 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 910 } 911 912 static void 913 if_epochalloc(void *dummy __unused) 914 { 915 916 net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT); 917 net_epoch = epoch_alloc(0); 918 } 919 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY, 920 if_epochalloc, NULL); 921 922 static void 923 if_attachdomain(void *dummy) 924 { 925 struct ifnet *ifp; 926 927 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 928 if_attachdomain1(ifp); 929 } 930 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 931 if_attachdomain, NULL); 932 933 static void 934 if_attachdomain1(struct ifnet *ifp) 935 { 936 struct domain *dp; 937 938 /* 939 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 940 * cannot lock ifp->if_afdata initialization, entirely. 941 */ 942 IF_AFDATA_LOCK(ifp); 943 if (ifp->if_afdata_initialized >= domain_init_status) { 944 IF_AFDATA_UNLOCK(ifp); 945 log(LOG_WARNING, "%s called more than once on %s\n", 946 __func__, ifp->if_xname); 947 return; 948 } 949 ifp->if_afdata_initialized = domain_init_status; 950 IF_AFDATA_UNLOCK(ifp); 951 952 /* address family dependent data region */ 953 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 954 for (dp = domains; dp; dp = dp->dom_next) { 955 if (dp->dom_ifattach) 956 ifp->if_afdata[dp->dom_family] = 957 (*dp->dom_ifattach)(ifp); 958 } 959 } 960 961 /* 962 * Remove any unicast or broadcast network addresses from an interface. 963 */ 964 void 965 if_purgeaddrs(struct ifnet *ifp) 966 { 967 struct ifaddr *ifa; 968 969 while (1) { 970 NET_EPOCH_ENTER(); 971 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 972 if (ifa->ifa_addr->sa_family != AF_LINK) 973 break; 974 } 975 NET_EPOCH_EXIT(); 976 977 if (ifa == NULL) 978 break; 979 #ifdef INET 980 /* XXX: Ugly!! ad hoc just for INET */ 981 if (ifa->ifa_addr->sa_family == AF_INET) { 982 struct ifaliasreq ifr; 983 984 bzero(&ifr, sizeof(ifr)); 985 ifr.ifra_addr = *ifa->ifa_addr; 986 if (ifa->ifa_dstaddr) 987 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 988 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 989 NULL) == 0) 990 continue; 991 } 992 #endif /* INET */ 993 #ifdef INET6 994 if (ifa->ifa_addr->sa_family == AF_INET6) { 995 in6_purgeaddr(ifa); 996 /* ifp_addrhead is already updated */ 997 continue; 998 } 999 #endif /* INET6 */ 1000 IF_ADDR_WLOCK(ifp); 1001 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1002 IF_ADDR_WUNLOCK(ifp); 1003 ifa_free(ifa); 1004 } 1005 } 1006 1007 /* 1008 * Remove any multicast network addresses from an interface when an ifnet 1009 * is going away. 1010 */ 1011 static void 1012 if_purgemaddrs(struct ifnet *ifp) 1013 { 1014 struct ifmultiaddr *ifma; 1015 1016 IF_ADDR_WLOCK(ifp); 1017 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1018 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1019 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1020 if_delmulti_locked(ifp, ifma, 1); 1021 } 1022 IF_ADDR_WUNLOCK(ifp); 1023 } 1024 1025 /* 1026 * Detach an interface, removing it from the list of "active" interfaces. 1027 * If vmove flag is set on entry to if_detach_internal(), perform only a 1028 * limited subset of cleanup tasks, given that we are moving an ifnet from 1029 * one vnet to another, where it must be fully operational. 1030 * 1031 * XXXRW: There are some significant questions about event ordering, and 1032 * how to prevent things from starting to use the interface during detach. 1033 */ 1034 void 1035 if_detach(struct ifnet *ifp) 1036 { 1037 1038 CURVNET_SET_QUIET(ifp->if_vnet); 1039 if_detach_internal(ifp, 0, NULL); 1040 CURVNET_RESTORE(); 1041 } 1042 1043 /* 1044 * The vmove flag, if set, indicates that we are called from a callpath 1045 * that is moving an interface to a different vnet instance. 1046 * 1047 * The shutdown flag, if set, indicates that we are called in the 1048 * process of shutting down a vnet instance. Currently only the 1049 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1050 * on a vnet instance shutdown without this flag being set, e.g., when 1051 * the cloned interfaces are destoyed as first thing of teardown. 1052 */ 1053 static int 1054 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 1055 { 1056 struct ifaddr *ifa; 1057 int i; 1058 struct domain *dp; 1059 struct ifnet *iter; 1060 int found = 0; 1061 #ifdef VIMAGE 1062 int shutdown; 1063 1064 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1065 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1066 #endif 1067 IFNET_WLOCK(); 1068 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 1069 if (iter == ifp) { 1070 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 1071 found = 1; 1072 break; 1073 } 1074 IFNET_WUNLOCK(); 1075 if (!found) { 1076 /* 1077 * While we would want to panic here, we cannot 1078 * guarantee that the interface is indeed still on 1079 * the list given we don't hold locks all the way. 1080 */ 1081 return (ENOENT); 1082 #if 0 1083 if (vmove) 1084 panic("%s: ifp=%p not on the ifnet tailq %p", 1085 __func__, ifp, &V_ifnet); 1086 else 1087 return; /* XXX this should panic as well? */ 1088 #endif 1089 } 1090 1091 /* 1092 * At this point we know the interface still was on the ifnet list 1093 * and we removed it so we are in a stable state. 1094 */ 1095 #ifdef VIMAGE 1096 curvnet->vnet_ifcnt--; 1097 #endif 1098 epoch_wait_preempt(net_epoch_preempt); 1099 /* 1100 * In any case (destroy or vmove) detach us from the groups 1101 * and remove/wait for pending events on the taskq. 1102 * XXX-BZ in theory an interface could still enqueue a taskq change? 1103 */ 1104 if_delgroups(ifp); 1105 1106 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1107 1108 /* 1109 * Check if this is a cloned interface or not. Must do even if 1110 * shutting down as a if_vmove_reclaim() would move the ifp and 1111 * the if_clone_addgroup() will have a corrupted string overwise 1112 * from a gibberish pointer. 1113 */ 1114 if (vmove && ifcp != NULL) 1115 *ifcp = if_clone_findifc(ifp); 1116 1117 if_down(ifp); 1118 1119 #ifdef VIMAGE 1120 /* 1121 * On VNET shutdown abort here as the stack teardown will do all 1122 * the work top-down for us. 1123 */ 1124 if (shutdown) { 1125 /* Give interface users the chance to clean up. */ 1126 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1127 1128 /* 1129 * In case of a vmove we are done here without error. 1130 * If we would signal an error it would lead to the same 1131 * abort as if we did not find the ifnet anymore. 1132 * if_detach() calls us in void context and does not care 1133 * about an early abort notification, so life is splendid :) 1134 */ 1135 goto finish_vnet_shutdown; 1136 } 1137 #endif 1138 1139 /* 1140 * At this point we are not tearing down a VNET and are either 1141 * going to destroy or vmove the interface and have to cleanup 1142 * accordingly. 1143 */ 1144 1145 /* 1146 * Remove routes and flush queues. 1147 */ 1148 #ifdef ALTQ 1149 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1150 altq_disable(&ifp->if_snd); 1151 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1152 altq_detach(&ifp->if_snd); 1153 #endif 1154 1155 if_purgeaddrs(ifp); 1156 1157 #ifdef INET 1158 in_ifdetach(ifp); 1159 #endif 1160 1161 #ifdef INET6 1162 /* 1163 * Remove all IPv6 kernel structs related to ifp. This should be done 1164 * before removing routing entries below, since IPv6 interface direct 1165 * routes are expected to be removed by the IPv6-specific kernel API. 1166 * Otherwise, the kernel will detect some inconsistency and bark it. 1167 */ 1168 in6_ifdetach(ifp); 1169 #endif 1170 if_purgemaddrs(ifp); 1171 1172 /* Announce that the interface is gone. */ 1173 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1174 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1175 if (IS_DEFAULT_VNET(curvnet)) 1176 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1177 1178 if (!vmove) { 1179 /* 1180 * Prevent further calls into the device driver via ifnet. 1181 */ 1182 if_dead(ifp); 1183 1184 /* 1185 * Remove link ifaddr pointer and maybe decrement if_index. 1186 * Clean up all addresses. 1187 */ 1188 free(ifp->if_hw_addr, M_IFADDR); 1189 ifp->if_hw_addr = NULL; 1190 ifp->if_addr = NULL; 1191 1192 /* We can now free link ifaddr. */ 1193 IF_ADDR_WLOCK(ifp); 1194 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1195 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1196 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1197 IF_ADDR_WUNLOCK(ifp); 1198 ifa_free(ifa); 1199 } else 1200 IF_ADDR_WUNLOCK(ifp); 1201 } 1202 1203 rt_flushifroutes(ifp); 1204 1205 #ifdef VIMAGE 1206 finish_vnet_shutdown: 1207 #endif 1208 /* 1209 * We cannot hold the lock over dom_ifdetach calls as they might 1210 * sleep, for example trying to drain a callout, thus open up the 1211 * theoretical race with re-attaching. 1212 */ 1213 IF_AFDATA_LOCK(ifp); 1214 i = ifp->if_afdata_initialized; 1215 ifp->if_afdata_initialized = 0; 1216 IF_AFDATA_UNLOCK(ifp); 1217 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1218 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1219 (*dp->dom_ifdetach)(ifp, 1220 ifp->if_afdata[dp->dom_family]); 1221 ifp->if_afdata[dp->dom_family] = NULL; 1222 } 1223 } 1224 1225 return (0); 1226 } 1227 1228 #ifdef VIMAGE 1229 /* 1230 * if_vmove() performs a limited version of if_detach() in current 1231 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1232 * An attempt is made to shrink if_index in current vnet, find an 1233 * unused if_index in target vnet and calls if_grow() if necessary, 1234 * and finally find an unused if_xname for the target vnet. 1235 */ 1236 static void 1237 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1238 { 1239 struct if_clone *ifc; 1240 u_int bif_dlt, bif_hdrlen; 1241 void *old; 1242 int rc; 1243 1244 /* 1245 * if_detach_internal() will call the eventhandler to notify 1246 * interface departure. That will detach if_bpf. We need to 1247 * safe the dlt and hdrlen so we can re-attach it later. 1248 */ 1249 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1250 1251 /* 1252 * Detach from current vnet, but preserve LLADDR info, do not 1253 * mark as dead etc. so that the ifnet can be reattached later. 1254 * If we cannot find it, we lost the race to someone else. 1255 */ 1256 rc = if_detach_internal(ifp, 1, &ifc); 1257 if (rc != 0) 1258 return; 1259 1260 /* 1261 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1262 * the if_index for that vnet if possible. 1263 * 1264 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1265 * or we'd lock on one vnet and unlock on another. 1266 */ 1267 IFNET_WLOCK(); 1268 ifindex_free_locked(ifp->if_index); 1269 IFNET_WUNLOCK(); 1270 1271 /* 1272 * Perform interface-specific reassignment tasks, if provided by 1273 * the driver. 1274 */ 1275 if (ifp->if_reassign != NULL) 1276 ifp->if_reassign(ifp, new_vnet, NULL); 1277 1278 /* 1279 * Switch to the context of the target vnet. 1280 */ 1281 CURVNET_SET_QUIET(new_vnet); 1282 restart: 1283 IFNET_WLOCK(); 1284 ifp->if_index = ifindex_alloc(&old); 1285 if (__predict_false(ifp->if_index == USHRT_MAX)) { 1286 IFNET_WUNLOCK(); 1287 epoch_wait_preempt(net_epoch_preempt); 1288 free(old, M_IFNET); 1289 goto restart; 1290 } 1291 ifnet_setbyindex(ifp->if_index, ifp); 1292 IFNET_WUNLOCK(); 1293 1294 if_attach_internal(ifp, 1, ifc); 1295 1296 if (ifp->if_bpf == NULL) 1297 bpfattach(ifp, bif_dlt, bif_hdrlen); 1298 1299 CURVNET_RESTORE(); 1300 } 1301 1302 /* 1303 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1304 */ 1305 static int 1306 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1307 { 1308 struct prison *pr; 1309 struct ifnet *difp; 1310 int shutdown; 1311 1312 /* Try to find the prison within our visibility. */ 1313 sx_slock(&allprison_lock); 1314 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1315 sx_sunlock(&allprison_lock); 1316 if (pr == NULL) 1317 return (ENXIO); 1318 prison_hold_locked(pr); 1319 mtx_unlock(&pr->pr_mtx); 1320 1321 /* Do not try to move the iface from and to the same prison. */ 1322 if (pr->pr_vnet == ifp->if_vnet) { 1323 prison_free(pr); 1324 return (EEXIST); 1325 } 1326 1327 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1328 /* XXX Lock interfaces to avoid races. */ 1329 CURVNET_SET_QUIET(pr->pr_vnet); 1330 difp = ifunit(ifname); 1331 if (difp != NULL) { 1332 CURVNET_RESTORE(); 1333 prison_free(pr); 1334 return (EEXIST); 1335 } 1336 1337 /* Make sure the VNET is stable. */ 1338 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1339 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1340 if (shutdown) { 1341 CURVNET_RESTORE(); 1342 prison_free(pr); 1343 return (EBUSY); 1344 } 1345 CURVNET_RESTORE(); 1346 1347 /* Move the interface into the child jail/vnet. */ 1348 if_vmove(ifp, pr->pr_vnet); 1349 1350 /* Report the new if_xname back to the userland. */ 1351 sprintf(ifname, "%s", ifp->if_xname); 1352 1353 prison_free(pr); 1354 return (0); 1355 } 1356 1357 static int 1358 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1359 { 1360 struct prison *pr; 1361 struct vnet *vnet_dst; 1362 struct ifnet *ifp; 1363 int shutdown; 1364 1365 /* Try to find the prison within our visibility. */ 1366 sx_slock(&allprison_lock); 1367 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1368 sx_sunlock(&allprison_lock); 1369 if (pr == NULL) 1370 return (ENXIO); 1371 prison_hold_locked(pr); 1372 mtx_unlock(&pr->pr_mtx); 1373 1374 /* Make sure the named iface exists in the source prison/vnet. */ 1375 CURVNET_SET(pr->pr_vnet); 1376 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1377 if (ifp == NULL) { 1378 CURVNET_RESTORE(); 1379 prison_free(pr); 1380 return (ENXIO); 1381 } 1382 1383 /* Do not try to move the iface from and to the same prison. */ 1384 vnet_dst = TD_TO_VNET(td); 1385 if (vnet_dst == ifp->if_vnet) { 1386 CURVNET_RESTORE(); 1387 prison_free(pr); 1388 return (EEXIST); 1389 } 1390 1391 /* Make sure the VNET is stable. */ 1392 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1393 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1394 if (shutdown) { 1395 CURVNET_RESTORE(); 1396 prison_free(pr); 1397 return (EBUSY); 1398 } 1399 1400 /* Get interface back from child jail/vnet. */ 1401 if_vmove(ifp, vnet_dst); 1402 CURVNET_RESTORE(); 1403 1404 /* Report the new if_xname back to the userland. */ 1405 sprintf(ifname, "%s", ifp->if_xname); 1406 1407 prison_free(pr); 1408 return (0); 1409 } 1410 #endif /* VIMAGE */ 1411 1412 /* 1413 * Add a group to an interface 1414 */ 1415 int 1416 if_addgroup(struct ifnet *ifp, const char *groupname) 1417 { 1418 struct ifg_list *ifgl; 1419 struct ifg_group *ifg = NULL; 1420 struct ifg_member *ifgm; 1421 int new = 0; 1422 1423 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1424 groupname[strlen(groupname) - 1] <= '9') 1425 return (EINVAL); 1426 1427 IFNET_WLOCK(); 1428 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1429 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1430 IFNET_WUNLOCK(); 1431 return (EEXIST); 1432 } 1433 1434 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1435 M_NOWAIT)) == NULL) { 1436 IFNET_WUNLOCK(); 1437 return (ENOMEM); 1438 } 1439 1440 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1441 M_TEMP, M_NOWAIT)) == NULL) { 1442 free(ifgl, M_TEMP); 1443 IFNET_WUNLOCK(); 1444 return (ENOMEM); 1445 } 1446 1447 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1448 if (!strcmp(ifg->ifg_group, groupname)) 1449 break; 1450 1451 if (ifg == NULL) { 1452 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1453 M_TEMP, M_NOWAIT)) == NULL) { 1454 free(ifgl, M_TEMP); 1455 free(ifgm, M_TEMP); 1456 IFNET_WUNLOCK(); 1457 return (ENOMEM); 1458 } 1459 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1460 ifg->ifg_refcnt = 0; 1461 CK_STAILQ_INIT(&ifg->ifg_members); 1462 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1463 new = 1; 1464 } 1465 1466 ifg->ifg_refcnt++; 1467 ifgl->ifgl_group = ifg; 1468 ifgm->ifgm_ifp = ifp; 1469 1470 IF_ADDR_WLOCK(ifp); 1471 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1472 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1473 IF_ADDR_WUNLOCK(ifp); 1474 1475 IFNET_WUNLOCK(); 1476 1477 if (new) 1478 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1479 EVENTHANDLER_INVOKE(group_change_event, groupname); 1480 1481 return (0); 1482 } 1483 1484 /* 1485 * Remove a group from an interface 1486 */ 1487 int 1488 if_delgroup(struct ifnet *ifp, const char *groupname) 1489 { 1490 struct ifg_list *ifgl; 1491 struct ifg_member *ifgm; 1492 int freeifgl; 1493 1494 IFNET_WLOCK(); 1495 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1496 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1497 break; 1498 if (ifgl == NULL) { 1499 IFNET_WUNLOCK(); 1500 return (ENOENT); 1501 } 1502 1503 freeifgl = 0; 1504 IF_ADDR_WLOCK(ifp); 1505 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1506 IF_ADDR_WUNLOCK(ifp); 1507 1508 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1509 if (ifgm->ifgm_ifp == ifp) 1510 break; 1511 1512 if (ifgm != NULL) 1513 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next); 1514 1515 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1516 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next); 1517 freeifgl = 1; 1518 } 1519 IFNET_WUNLOCK(); 1520 1521 epoch_wait_preempt(net_epoch_preempt); 1522 if (freeifgl) { 1523 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1524 free(ifgl->ifgl_group, M_TEMP); 1525 } 1526 free(ifgm, M_TEMP); 1527 free(ifgl, M_TEMP); 1528 1529 EVENTHANDLER_INVOKE(group_change_event, groupname); 1530 1531 return (0); 1532 } 1533 1534 /* 1535 * Remove an interface from all groups 1536 */ 1537 static void 1538 if_delgroups(struct ifnet *ifp) 1539 { 1540 struct ifg_list *ifgl; 1541 struct ifg_member *ifgm; 1542 char groupname[IFNAMSIZ]; 1543 int ifglfree; 1544 1545 IFNET_WLOCK(); 1546 while (!CK_STAILQ_EMPTY(&ifp->if_groups)) { 1547 ifgl = CK_STAILQ_FIRST(&ifp->if_groups); 1548 1549 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1550 1551 IF_ADDR_WLOCK(ifp); 1552 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1553 IF_ADDR_WUNLOCK(ifp); 1554 1555 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1556 if (ifgm->ifgm_ifp == ifp) 1557 break; 1558 1559 if (ifgm != NULL) 1560 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, 1561 ifgm_next); 1562 ifglfree = 0; 1563 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1564 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next); 1565 ifglfree = 1; 1566 } 1567 1568 IFNET_WUNLOCK(); 1569 epoch_wait_preempt(net_epoch_preempt); 1570 free(ifgm, M_TEMP); 1571 if (ifglfree) { 1572 EVENTHANDLER_INVOKE(group_detach_event, 1573 ifgl->ifgl_group); 1574 free(ifgl->ifgl_group, M_TEMP); 1575 } 1576 EVENTHANDLER_INVOKE(group_change_event, groupname); 1577 1578 IFNET_WLOCK(); 1579 } 1580 IFNET_WUNLOCK(); 1581 } 1582 1583 static char * 1584 ifgr_group_get(void *ifgrp) 1585 { 1586 union ifgroupreq_union *ifgrup; 1587 1588 ifgrup = ifgrp; 1589 #ifdef COMPAT_FREEBSD32 1590 if (SV_CURPROC_FLAG(SV_ILP32)) 1591 return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]); 1592 #endif 1593 return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]); 1594 } 1595 1596 static struct ifg_req * 1597 ifgr_groups_get(void *ifgrp) 1598 { 1599 union ifgroupreq_union *ifgrup; 1600 1601 ifgrup = ifgrp; 1602 #ifdef COMPAT_FREEBSD32 1603 if (SV_CURPROC_FLAG(SV_ILP32)) 1604 return ((struct ifg_req *)(uintptr_t) 1605 ifgrup->ifgr32.ifgr_ifgru.ifgru_groups); 1606 #endif 1607 return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups); 1608 } 1609 1610 /* 1611 * Stores all groups from an interface in memory pointed to by ifgr. 1612 */ 1613 static int 1614 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1615 { 1616 int len, error; 1617 struct ifg_list *ifgl; 1618 struct ifg_req ifgrq, *ifgp; 1619 1620 if (ifgr->ifgr_len == 0) { 1621 IF_ADDR_RLOCK(ifp); 1622 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1623 ifgr->ifgr_len += sizeof(struct ifg_req); 1624 IF_ADDR_RUNLOCK(ifp); 1625 return (0); 1626 } 1627 1628 len = ifgr->ifgr_len; 1629 ifgp = ifgr_groups_get(ifgr); 1630 /* XXX: wire */ 1631 IF_ADDR_RLOCK(ifp); 1632 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1633 if (len < sizeof(ifgrq)) { 1634 IF_ADDR_RUNLOCK(ifp); 1635 return (EINVAL); 1636 } 1637 bzero(&ifgrq, sizeof ifgrq); 1638 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1639 sizeof(ifgrq.ifgrq_group)); 1640 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1641 IF_ADDR_RUNLOCK(ifp); 1642 return (error); 1643 } 1644 len -= sizeof(ifgrq); 1645 ifgp++; 1646 } 1647 IF_ADDR_RUNLOCK(ifp); 1648 1649 return (0); 1650 } 1651 1652 /* 1653 * Stores all members of a group in memory pointed to by igfr 1654 */ 1655 static int 1656 if_getgroupmembers(struct ifgroupreq *ifgr) 1657 { 1658 struct ifg_group *ifg; 1659 struct ifg_member *ifgm; 1660 struct ifg_req ifgrq, *ifgp; 1661 int len, error; 1662 1663 IFNET_RLOCK(); 1664 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1665 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1666 break; 1667 if (ifg == NULL) { 1668 IFNET_RUNLOCK(); 1669 return (ENOENT); 1670 } 1671 1672 if (ifgr->ifgr_len == 0) { 1673 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1674 ifgr->ifgr_len += sizeof(ifgrq); 1675 IFNET_RUNLOCK(); 1676 return (0); 1677 } 1678 1679 len = ifgr->ifgr_len; 1680 ifgp = ifgr_groups_get(ifgr); 1681 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1682 if (len < sizeof(ifgrq)) { 1683 IFNET_RUNLOCK(); 1684 return (EINVAL); 1685 } 1686 bzero(&ifgrq, sizeof ifgrq); 1687 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1688 sizeof(ifgrq.ifgrq_member)); 1689 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1690 IFNET_RUNLOCK(); 1691 return (error); 1692 } 1693 len -= sizeof(ifgrq); 1694 ifgp++; 1695 } 1696 IFNET_RUNLOCK(); 1697 1698 return (0); 1699 } 1700 1701 /* 1702 * Return counter values from counter(9)s stored in ifnet. 1703 */ 1704 uint64_t 1705 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1706 { 1707 1708 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1709 1710 return (counter_u64_fetch(ifp->if_counters[cnt])); 1711 } 1712 1713 /* 1714 * Increase an ifnet counter. Usually used for counters shared 1715 * between the stack and a driver, but function supports them all. 1716 */ 1717 void 1718 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1719 { 1720 1721 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1722 1723 counter_u64_add(ifp->if_counters[cnt], inc); 1724 } 1725 1726 /* 1727 * Copy data from ifnet to userland API structure if_data. 1728 */ 1729 void 1730 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1731 { 1732 1733 ifd->ifi_type = ifp->if_type; 1734 ifd->ifi_physical = 0; 1735 ifd->ifi_addrlen = ifp->if_addrlen; 1736 ifd->ifi_hdrlen = ifp->if_hdrlen; 1737 ifd->ifi_link_state = ifp->if_link_state; 1738 ifd->ifi_vhid = 0; 1739 ifd->ifi_datalen = sizeof(struct if_data); 1740 ifd->ifi_mtu = ifp->if_mtu; 1741 ifd->ifi_metric = ifp->if_metric; 1742 ifd->ifi_baudrate = ifp->if_baudrate; 1743 ifd->ifi_hwassist = ifp->if_hwassist; 1744 ifd->ifi_epoch = ifp->if_epoch; 1745 ifd->ifi_lastchange = ifp->if_lastchange; 1746 1747 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1748 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1749 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1750 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1751 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1752 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1753 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1754 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1755 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1756 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1757 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1758 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1759 } 1760 1761 /* 1762 * Wrapper functions for struct ifnet address list locking macros. These are 1763 * used by kernel modules to avoid encoding programming interface or binary 1764 * interface assumptions that may be violated when kernel-internal locking 1765 * approaches change. 1766 */ 1767 void 1768 if_addr_rlock(struct ifnet *ifp) 1769 { 1770 MPASS(*(uint64_t *)&ifp->if_addr_et == 0); 1771 epoch_enter_preempt(net_epoch_preempt, &ifp->if_addr_et); 1772 } 1773 1774 void 1775 if_addr_runlock(struct ifnet *ifp) 1776 { 1777 epoch_exit_preempt(net_epoch_preempt, &ifp->if_addr_et); 1778 #ifdef INVARIANTS 1779 bzero(&ifp->if_addr_et, sizeof(struct epoch_tracker)); 1780 #endif 1781 } 1782 1783 void 1784 if_maddr_rlock(if_t ifp) 1785 { 1786 1787 MPASS(*(uint64_t *)&ifp->if_maddr_et == 0); 1788 epoch_enter_preempt(net_epoch_preempt, &ifp->if_maddr_et); 1789 } 1790 1791 void 1792 if_maddr_runlock(if_t ifp) 1793 { 1794 1795 epoch_exit_preempt(net_epoch_preempt, &ifp->if_maddr_et); 1796 #ifdef INVARIANTS 1797 bzero(&ifp->if_maddr_et, sizeof(struct epoch_tracker)); 1798 #endif 1799 } 1800 1801 /* 1802 * Initialization, destruction and refcounting functions for ifaddrs. 1803 */ 1804 struct ifaddr * 1805 ifa_alloc(size_t size, int flags) 1806 { 1807 struct ifaddr *ifa; 1808 1809 KASSERT(size >= sizeof(struct ifaddr), 1810 ("%s: invalid size %zu", __func__, size)); 1811 1812 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1813 if (ifa == NULL) 1814 return (NULL); 1815 1816 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1817 goto fail; 1818 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1819 goto fail; 1820 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1821 goto fail; 1822 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1823 goto fail; 1824 1825 refcount_init(&ifa->ifa_refcnt, 1); 1826 1827 return (ifa); 1828 1829 fail: 1830 /* free(NULL) is okay */ 1831 counter_u64_free(ifa->ifa_opackets); 1832 counter_u64_free(ifa->ifa_ipackets); 1833 counter_u64_free(ifa->ifa_obytes); 1834 counter_u64_free(ifa->ifa_ibytes); 1835 free(ifa, M_IFADDR); 1836 1837 return (NULL); 1838 } 1839 1840 void 1841 ifa_ref(struct ifaddr *ifa) 1842 { 1843 1844 refcount_acquire(&ifa->ifa_refcnt); 1845 } 1846 1847 static void 1848 ifa_destroy(epoch_context_t ctx) 1849 { 1850 struct ifaddr *ifa; 1851 1852 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1853 counter_u64_free(ifa->ifa_opackets); 1854 counter_u64_free(ifa->ifa_ipackets); 1855 counter_u64_free(ifa->ifa_obytes); 1856 counter_u64_free(ifa->ifa_ibytes); 1857 free(ifa, M_IFADDR); 1858 } 1859 1860 void 1861 ifa_free(struct ifaddr *ifa) 1862 { 1863 1864 if (refcount_release(&ifa->ifa_refcnt)) 1865 epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy); 1866 } 1867 1868 1869 static int 1870 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa, 1871 struct sockaddr *ia) 1872 { 1873 int error; 1874 struct rt_addrinfo info; 1875 struct sockaddr_dl null_sdl; 1876 struct ifnet *ifp; 1877 1878 ifp = ifa->ifa_ifp; 1879 1880 bzero(&info, sizeof(info)); 1881 if (cmd != RTM_DELETE) 1882 info.rti_ifp = V_loif; 1883 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED; 1884 info.rti_info[RTAX_DST] = ia; 1885 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1886 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type); 1887 1888 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib); 1889 1890 if (error != 0 && 1891 !(cmd == RTM_ADD && error == EEXIST) && 1892 !(cmd == RTM_DELETE && error == ENOENT)) 1893 if_printf(ifp, "%s failed: %d\n", otype, error); 1894 1895 return (error); 1896 } 1897 1898 int 1899 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1900 { 1901 1902 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia)); 1903 } 1904 1905 int 1906 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1907 { 1908 1909 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia)); 1910 } 1911 1912 int 1913 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1914 { 1915 1916 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia)); 1917 } 1918 1919 /* 1920 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1921 * structs used to represent other address families, it is necessary 1922 * to perform a different comparison. 1923 */ 1924 1925 #define sa_dl_equal(a1, a2) \ 1926 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1927 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1928 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1929 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1930 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1931 1932 /* 1933 * Locate an interface based on a complete address. 1934 */ 1935 /*ARGSUSED*/ 1936 struct ifaddr * 1937 ifa_ifwithaddr(const struct sockaddr *addr) 1938 { 1939 struct ifnet *ifp; 1940 struct ifaddr *ifa; 1941 1942 MPASS(in_epoch(net_epoch_preempt)); 1943 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1944 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1945 if (ifa->ifa_addr->sa_family != addr->sa_family) 1946 continue; 1947 if (sa_equal(addr, ifa->ifa_addr)) { 1948 goto done; 1949 } 1950 /* IP6 doesn't have broadcast */ 1951 if ((ifp->if_flags & IFF_BROADCAST) && 1952 ifa->ifa_broadaddr && 1953 ifa->ifa_broadaddr->sa_len != 0 && 1954 sa_equal(ifa->ifa_broadaddr, addr)) { 1955 goto done; 1956 } 1957 } 1958 } 1959 ifa = NULL; 1960 done: 1961 return (ifa); 1962 } 1963 1964 int 1965 ifa_ifwithaddr_check(const struct sockaddr *addr) 1966 { 1967 int rc; 1968 1969 NET_EPOCH_ENTER(); 1970 rc = (ifa_ifwithaddr(addr) != NULL); 1971 NET_EPOCH_EXIT(); 1972 return (rc); 1973 } 1974 1975 /* 1976 * Locate an interface based on the broadcast address. 1977 */ 1978 /* ARGSUSED */ 1979 struct ifaddr * 1980 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1981 { 1982 struct ifnet *ifp; 1983 struct ifaddr *ifa; 1984 1985 MPASS(in_epoch(net_epoch_preempt)); 1986 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 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 ((ifp->if_flags & IFF_BROADCAST) && 1993 ifa->ifa_broadaddr && 1994 ifa->ifa_broadaddr->sa_len != 0 && 1995 sa_equal(ifa->ifa_broadaddr, addr)) { 1996 goto done; 1997 } 1998 } 1999 } 2000 ifa = NULL; 2001 done: 2002 return (ifa); 2003 } 2004 2005 /* 2006 * Locate the point to point interface with a given destination address. 2007 */ 2008 /*ARGSUSED*/ 2009 struct ifaddr * 2010 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 2011 { 2012 struct ifnet *ifp; 2013 struct ifaddr *ifa; 2014 2015 MPASS(in_epoch(net_epoch_preempt)); 2016 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2017 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 2018 continue; 2019 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2020 continue; 2021 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2022 if (ifa->ifa_addr->sa_family != addr->sa_family) 2023 continue; 2024 if (ifa->ifa_dstaddr != NULL && 2025 sa_equal(addr, ifa->ifa_dstaddr)) { 2026 goto done; 2027 } 2028 } 2029 } 2030 ifa = NULL; 2031 done: 2032 return (ifa); 2033 } 2034 2035 /* 2036 * Find an interface on a specific network. If many, choice 2037 * is most specific found. 2038 */ 2039 struct ifaddr * 2040 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 2041 { 2042 struct ifnet *ifp; 2043 struct ifaddr *ifa; 2044 struct ifaddr *ifa_maybe = NULL; 2045 u_int af = addr->sa_family; 2046 const char *addr_data = addr->sa_data, *cplim; 2047 2048 MPASS(in_epoch(net_epoch_preempt)); 2049 /* 2050 * AF_LINK addresses can be looked up directly by their index number, 2051 * so do that if we can. 2052 */ 2053 if (af == AF_LINK) { 2054 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 2055 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 2056 return (ifaddr_byindex(sdl->sdl_index)); 2057 } 2058 2059 /* 2060 * Scan though each interface, looking for ones that have addresses 2061 * in this address family and the requested fib. Maintain a reference 2062 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 2063 * kept it stable when we move onto the next interface. 2064 */ 2065 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2066 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2067 continue; 2068 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2069 const char *cp, *cp2, *cp3; 2070 2071 if (ifa->ifa_addr->sa_family != af) 2072 next: continue; 2073 if (af == AF_INET && 2074 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 2075 /* 2076 * This is a bit broken as it doesn't 2077 * take into account that the remote end may 2078 * be a single node in the network we are 2079 * looking for. 2080 * The trouble is that we don't know the 2081 * netmask for the remote end. 2082 */ 2083 if (ifa->ifa_dstaddr != NULL && 2084 sa_equal(addr, ifa->ifa_dstaddr)) { 2085 goto done; 2086 } 2087 } else { 2088 /* 2089 * Scan all the bits in the ifa's address. 2090 * If a bit dissagrees with what we are 2091 * looking for, mask it with the netmask 2092 * to see if it really matters. 2093 * (A byte at a time) 2094 */ 2095 if (ifa->ifa_netmask == 0) 2096 continue; 2097 cp = addr_data; 2098 cp2 = ifa->ifa_addr->sa_data; 2099 cp3 = ifa->ifa_netmask->sa_data; 2100 cplim = ifa->ifa_netmask->sa_len 2101 + (char *)ifa->ifa_netmask; 2102 while (cp3 < cplim) 2103 if ((*cp++ ^ *cp2++) & *cp3++) 2104 goto next; /* next address! */ 2105 /* 2106 * If the netmask of what we just found 2107 * is more specific than what we had before 2108 * (if we had one), or if the virtual status 2109 * of new prefix is better than of the old one, 2110 * then remember the new one before continuing 2111 * to search for an even better one. 2112 */ 2113 if (ifa_maybe == NULL || 2114 ifa_preferred(ifa_maybe, ifa) || 2115 rn_refines((caddr_t)ifa->ifa_netmask, 2116 (caddr_t)ifa_maybe->ifa_netmask)) { 2117 ifa_maybe = ifa; 2118 } 2119 } 2120 } 2121 } 2122 ifa = ifa_maybe; 2123 ifa_maybe = NULL; 2124 done: 2125 return (ifa); 2126 } 2127 2128 /* 2129 * Find an interface address specific to an interface best matching 2130 * a given address. 2131 */ 2132 struct ifaddr * 2133 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2134 { 2135 struct ifaddr *ifa; 2136 const char *cp, *cp2, *cp3; 2137 char *cplim; 2138 struct ifaddr *ifa_maybe = NULL; 2139 u_int af = addr->sa_family; 2140 2141 if (af >= AF_MAX) 2142 return (NULL); 2143 2144 MPASS(in_epoch(net_epoch_preempt)); 2145 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2146 if (ifa->ifa_addr->sa_family != af) 2147 continue; 2148 if (ifa_maybe == NULL) 2149 ifa_maybe = ifa; 2150 if (ifa->ifa_netmask == 0) { 2151 if (sa_equal(addr, ifa->ifa_addr) || 2152 (ifa->ifa_dstaddr && 2153 sa_equal(addr, ifa->ifa_dstaddr))) 2154 goto done; 2155 continue; 2156 } 2157 if (ifp->if_flags & IFF_POINTOPOINT) { 2158 if (sa_equal(addr, ifa->ifa_dstaddr)) 2159 goto done; 2160 } else { 2161 cp = addr->sa_data; 2162 cp2 = ifa->ifa_addr->sa_data; 2163 cp3 = ifa->ifa_netmask->sa_data; 2164 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2165 for (; cp3 < cplim; cp3++) 2166 if ((*cp++ ^ *cp2++) & *cp3) 2167 break; 2168 if (cp3 == cplim) 2169 goto done; 2170 } 2171 } 2172 ifa = ifa_maybe; 2173 done: 2174 return (ifa); 2175 } 2176 2177 /* 2178 * See whether new ifa is better than current one: 2179 * 1) A non-virtual one is preferred over virtual. 2180 * 2) A virtual in master state preferred over any other state. 2181 * 2182 * Used in several address selecting functions. 2183 */ 2184 int 2185 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2186 { 2187 2188 return (cur->ifa_carp && (!next->ifa_carp || 2189 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2190 } 2191 2192 #include <net/if_llatbl.h> 2193 2194 /* 2195 * Default action when installing a route with a Link Level gateway. 2196 * Lookup an appropriate real ifa to point to. 2197 * This should be moved to /sys/net/link.c eventually. 2198 */ 2199 static void 2200 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 2201 { 2202 struct ifaddr *ifa, *oifa; 2203 struct sockaddr *dst; 2204 struct ifnet *ifp; 2205 2206 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) || 2207 ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL)) 2208 return; 2209 NET_EPOCH_ENTER(); 2210 ifa = ifaof_ifpforaddr(dst, ifp); 2211 if (ifa) { 2212 oifa = rt->rt_ifa; 2213 if (oifa != ifa) { 2214 ifa_free(oifa); 2215 ifa_ref(ifa); 2216 } 2217 rt->rt_ifa = ifa; 2218 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2219 ifa->ifa_rtrequest(cmd, rt, info); 2220 } 2221 NET_EPOCH_EXIT(); 2222 } 2223 2224 struct sockaddr_dl * 2225 link_alloc_sdl(size_t size, int flags) 2226 { 2227 2228 return (malloc(size, M_TEMP, flags)); 2229 } 2230 2231 void 2232 link_free_sdl(struct sockaddr *sa) 2233 { 2234 free(sa, M_TEMP); 2235 } 2236 2237 /* 2238 * Fills in given sdl with interface basic info. 2239 * Returns pointer to filled sdl. 2240 */ 2241 struct sockaddr_dl * 2242 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2243 { 2244 struct sockaddr_dl *sdl; 2245 2246 sdl = (struct sockaddr_dl *)paddr; 2247 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2248 sdl->sdl_len = sizeof(struct sockaddr_dl); 2249 sdl->sdl_family = AF_LINK; 2250 sdl->sdl_index = ifp->if_index; 2251 sdl->sdl_type = iftype; 2252 2253 return (sdl); 2254 } 2255 2256 /* 2257 * Mark an interface down and notify protocols of 2258 * the transition. 2259 */ 2260 static void 2261 if_unroute(struct ifnet *ifp, int flag, int fam) 2262 { 2263 struct ifaddr *ifa; 2264 2265 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2266 2267 ifp->if_flags &= ~flag; 2268 getmicrotime(&ifp->if_lastchange); 2269 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2270 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2271 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2272 ifp->if_qflush(ifp); 2273 2274 if (ifp->if_carp) 2275 (*carp_linkstate_p)(ifp); 2276 rt_ifmsg(ifp); 2277 } 2278 2279 /* 2280 * Mark an interface up and notify protocols of 2281 * the transition. 2282 */ 2283 static void 2284 if_route(struct ifnet *ifp, int flag, int fam) 2285 { 2286 struct ifaddr *ifa; 2287 2288 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2289 2290 ifp->if_flags |= flag; 2291 getmicrotime(&ifp->if_lastchange); 2292 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2293 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2294 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2295 if (ifp->if_carp) 2296 (*carp_linkstate_p)(ifp); 2297 rt_ifmsg(ifp); 2298 #ifdef INET6 2299 in6_if_up(ifp); 2300 #endif 2301 } 2302 2303 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2304 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2305 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2306 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2307 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2308 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2309 int (*vlan_setcookie_p)(struct ifnet *, void *); 2310 void *(*vlan_cookie_p)(struct ifnet *); 2311 2312 /* 2313 * Handle a change in the interface link state. To avoid LORs 2314 * between driver lock and upper layer locks, as well as possible 2315 * recursions, we post event to taskqueue, and all job 2316 * is done in static do_link_state_change(). 2317 */ 2318 void 2319 if_link_state_change(struct ifnet *ifp, int link_state) 2320 { 2321 /* Return if state hasn't changed. */ 2322 if (ifp->if_link_state == link_state) 2323 return; 2324 2325 ifp->if_link_state = link_state; 2326 2327 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2328 } 2329 2330 static void 2331 do_link_state_change(void *arg, int pending) 2332 { 2333 struct ifnet *ifp = (struct ifnet *)arg; 2334 int link_state = ifp->if_link_state; 2335 CURVNET_SET(ifp->if_vnet); 2336 2337 /* Notify that the link state has changed. */ 2338 rt_ifmsg(ifp); 2339 if (ifp->if_vlantrunk != NULL) 2340 (*vlan_link_state_p)(ifp); 2341 2342 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2343 ifp->if_l2com != NULL) 2344 (*ng_ether_link_state_p)(ifp, link_state); 2345 if (ifp->if_carp) 2346 (*carp_linkstate_p)(ifp); 2347 if (ifp->if_bridge) 2348 ifp->if_bridge_linkstate(ifp); 2349 if (ifp->if_lagg) 2350 (*lagg_linkstate_p)(ifp, link_state); 2351 2352 if (IS_DEFAULT_VNET(curvnet)) 2353 devctl_notify("IFNET", ifp->if_xname, 2354 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2355 NULL); 2356 if (pending > 1) 2357 if_printf(ifp, "%d link states coalesced\n", pending); 2358 if (log_link_state_change) 2359 if_printf(ifp, "link state changed to %s\n", 2360 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2361 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2362 CURVNET_RESTORE(); 2363 } 2364 2365 /* 2366 * Mark an interface down and notify protocols of 2367 * the transition. 2368 */ 2369 void 2370 if_down(struct ifnet *ifp) 2371 { 2372 2373 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2374 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2375 } 2376 2377 /* 2378 * Mark an interface up and notify protocols of 2379 * the transition. 2380 */ 2381 void 2382 if_up(struct ifnet *ifp) 2383 { 2384 2385 if_route(ifp, IFF_UP, AF_UNSPEC); 2386 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2387 } 2388 2389 /* 2390 * Flush an interface queue. 2391 */ 2392 void 2393 if_qflush(struct ifnet *ifp) 2394 { 2395 struct mbuf *m, *n; 2396 struct ifaltq *ifq; 2397 2398 ifq = &ifp->if_snd; 2399 IFQ_LOCK(ifq); 2400 #ifdef ALTQ 2401 if (ALTQ_IS_ENABLED(ifq)) 2402 ALTQ_PURGE(ifq); 2403 #endif 2404 n = ifq->ifq_head; 2405 while ((m = n) != NULL) { 2406 n = m->m_nextpkt; 2407 m_freem(m); 2408 } 2409 ifq->ifq_head = 0; 2410 ifq->ifq_tail = 0; 2411 ifq->ifq_len = 0; 2412 IFQ_UNLOCK(ifq); 2413 } 2414 2415 /* 2416 * Map interface name to interface structure pointer, with or without 2417 * returning a reference. 2418 */ 2419 struct ifnet * 2420 ifunit_ref(const char *name) 2421 { 2422 struct ifnet *ifp; 2423 2424 IFNET_RLOCK_NOSLEEP(); 2425 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2426 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2427 !(ifp->if_flags & IFF_DYING)) 2428 break; 2429 } 2430 if (ifp != NULL) 2431 if_ref(ifp); 2432 IFNET_RUNLOCK_NOSLEEP(); 2433 return (ifp); 2434 } 2435 2436 struct ifnet * 2437 ifunit(const char *name) 2438 { 2439 struct ifnet *ifp; 2440 2441 IFNET_RLOCK_NOSLEEP(); 2442 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2443 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2444 break; 2445 } 2446 IFNET_RUNLOCK_NOSLEEP(); 2447 return (ifp); 2448 } 2449 2450 static void * 2451 ifr_buffer_get_buffer(void *data) 2452 { 2453 union ifreq_union *ifrup; 2454 2455 ifrup = data; 2456 #ifdef COMPAT_FREEBSD32 2457 if (SV_CURPROC_FLAG(SV_ILP32)) 2458 return ((void *)(uintptr_t) 2459 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2460 #endif 2461 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2462 } 2463 2464 static void 2465 ifr_buffer_set_buffer_null(void *data) 2466 { 2467 union ifreq_union *ifrup; 2468 2469 ifrup = data; 2470 #ifdef COMPAT_FREEBSD32 2471 if (SV_CURPROC_FLAG(SV_ILP32)) 2472 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2473 else 2474 #endif 2475 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2476 } 2477 2478 static size_t 2479 ifr_buffer_get_length(void *data) 2480 { 2481 union ifreq_union *ifrup; 2482 2483 ifrup = data; 2484 #ifdef COMPAT_FREEBSD32 2485 if (SV_CURPROC_FLAG(SV_ILP32)) 2486 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2487 #endif 2488 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2489 } 2490 2491 static void 2492 ifr_buffer_set_length(void *data, size_t len) 2493 { 2494 union ifreq_union *ifrup; 2495 2496 ifrup = data; 2497 #ifdef COMPAT_FREEBSD32 2498 if (SV_CURPROC_FLAG(SV_ILP32)) 2499 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2500 else 2501 #endif 2502 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2503 } 2504 2505 void * 2506 ifr_data_get_ptr(void *ifrp) 2507 { 2508 union ifreq_union *ifrup; 2509 2510 ifrup = ifrp; 2511 #ifdef COMPAT_FREEBSD32 2512 if (SV_CURPROC_FLAG(SV_ILP32)) 2513 return ((void *)(uintptr_t) 2514 ifrup->ifr32.ifr_ifru.ifru_data); 2515 #endif 2516 return (ifrup->ifr.ifr_ifru.ifru_data); 2517 } 2518 2519 /* 2520 * Hardware specific interface ioctls. 2521 */ 2522 int 2523 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2524 { 2525 struct ifreq *ifr; 2526 int error = 0, do_ifup = 0; 2527 int new_flags, temp_flags; 2528 size_t namelen, onamelen; 2529 size_t descrlen; 2530 char *descrbuf, *odescrbuf; 2531 char new_name[IFNAMSIZ]; 2532 struct ifaddr *ifa; 2533 struct sockaddr_dl *sdl; 2534 2535 ifr = (struct ifreq *)data; 2536 switch (cmd) { 2537 case SIOCGIFINDEX: 2538 ifr->ifr_index = ifp->if_index; 2539 break; 2540 2541 case SIOCGIFFLAGS: 2542 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2543 ifr->ifr_flags = temp_flags & 0xffff; 2544 ifr->ifr_flagshigh = temp_flags >> 16; 2545 break; 2546 2547 case SIOCGIFCAP: 2548 ifr->ifr_reqcap = ifp->if_capabilities; 2549 ifr->ifr_curcap = ifp->if_capenable; 2550 break; 2551 2552 #ifdef MAC 2553 case SIOCGIFMAC: 2554 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2555 break; 2556 #endif 2557 2558 case SIOCGIFMETRIC: 2559 ifr->ifr_metric = ifp->if_metric; 2560 break; 2561 2562 case SIOCGIFMTU: 2563 ifr->ifr_mtu = ifp->if_mtu; 2564 break; 2565 2566 case SIOCGIFPHYS: 2567 /* XXXGL: did this ever worked? */ 2568 ifr->ifr_phys = 0; 2569 break; 2570 2571 case SIOCGIFDESCR: 2572 error = 0; 2573 sx_slock(&ifdescr_sx); 2574 if (ifp->if_description == NULL) 2575 error = ENOMSG; 2576 else { 2577 /* space for terminating nul */ 2578 descrlen = strlen(ifp->if_description) + 1; 2579 if (ifr_buffer_get_length(ifr) < descrlen) 2580 ifr_buffer_set_buffer_null(ifr); 2581 else 2582 error = copyout(ifp->if_description, 2583 ifr_buffer_get_buffer(ifr), descrlen); 2584 ifr_buffer_set_length(ifr, descrlen); 2585 } 2586 sx_sunlock(&ifdescr_sx); 2587 break; 2588 2589 case SIOCSIFDESCR: 2590 error = priv_check(td, PRIV_NET_SETIFDESCR); 2591 if (error) 2592 return (error); 2593 2594 /* 2595 * Copy only (length-1) bytes to make sure that 2596 * if_description is always nul terminated. The 2597 * length parameter is supposed to count the 2598 * terminating nul in. 2599 */ 2600 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2601 return (ENAMETOOLONG); 2602 else if (ifr_buffer_get_length(ifr) == 0) 2603 descrbuf = NULL; 2604 else { 2605 descrbuf = malloc(ifr_buffer_get_length(ifr), 2606 M_IFDESCR, M_WAITOK | M_ZERO); 2607 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2608 ifr_buffer_get_length(ifr) - 1); 2609 if (error) { 2610 free(descrbuf, M_IFDESCR); 2611 break; 2612 } 2613 } 2614 2615 sx_xlock(&ifdescr_sx); 2616 odescrbuf = ifp->if_description; 2617 ifp->if_description = descrbuf; 2618 sx_xunlock(&ifdescr_sx); 2619 2620 getmicrotime(&ifp->if_lastchange); 2621 free(odescrbuf, M_IFDESCR); 2622 break; 2623 2624 case SIOCGIFFIB: 2625 ifr->ifr_fib = ifp->if_fib; 2626 break; 2627 2628 case SIOCSIFFIB: 2629 error = priv_check(td, PRIV_NET_SETIFFIB); 2630 if (error) 2631 return (error); 2632 if (ifr->ifr_fib >= rt_numfibs) 2633 return (EINVAL); 2634 2635 ifp->if_fib = ifr->ifr_fib; 2636 break; 2637 2638 case SIOCSIFFLAGS: 2639 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2640 if (error) 2641 return (error); 2642 /* 2643 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2644 * check, so we don't need special handling here yet. 2645 */ 2646 new_flags = (ifr->ifr_flags & 0xffff) | 2647 (ifr->ifr_flagshigh << 16); 2648 if (ifp->if_flags & IFF_UP && 2649 (new_flags & IFF_UP) == 0) { 2650 if_down(ifp); 2651 } else if (new_flags & IFF_UP && 2652 (ifp->if_flags & IFF_UP) == 0) { 2653 do_ifup = 1; 2654 } 2655 /* See if permanently promiscuous mode bit is about to flip */ 2656 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2657 if (new_flags & IFF_PPROMISC) 2658 ifp->if_flags |= IFF_PROMISC; 2659 else if (ifp->if_pcount == 0) 2660 ifp->if_flags &= ~IFF_PROMISC; 2661 if (log_promisc_mode_change) 2662 if_printf(ifp, "permanently promiscuous mode %s\n", 2663 ((new_flags & IFF_PPROMISC) ? 2664 "enabled" : "disabled")); 2665 } 2666 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2667 (new_flags &~ IFF_CANTCHANGE); 2668 if (ifp->if_ioctl) { 2669 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2670 } 2671 if (do_ifup) 2672 if_up(ifp); 2673 getmicrotime(&ifp->if_lastchange); 2674 break; 2675 2676 case SIOCSIFCAP: 2677 error = priv_check(td, PRIV_NET_SETIFCAP); 2678 if (error) 2679 return (error); 2680 if (ifp->if_ioctl == NULL) 2681 return (EOPNOTSUPP); 2682 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2683 return (EINVAL); 2684 error = (*ifp->if_ioctl)(ifp, cmd, data); 2685 if (error == 0) 2686 getmicrotime(&ifp->if_lastchange); 2687 break; 2688 2689 #ifdef MAC 2690 case SIOCSIFMAC: 2691 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2692 break; 2693 #endif 2694 2695 case SIOCSIFNAME: 2696 error = priv_check(td, PRIV_NET_SETIFNAME); 2697 if (error) 2698 return (error); 2699 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2700 NULL); 2701 if (error != 0) 2702 return (error); 2703 if (new_name[0] == '\0') 2704 return (EINVAL); 2705 if (new_name[IFNAMSIZ-1] != '\0') { 2706 new_name[IFNAMSIZ-1] = '\0'; 2707 if (strlen(new_name) == IFNAMSIZ-1) 2708 return (EINVAL); 2709 } 2710 if (ifunit(new_name) != NULL) 2711 return (EEXIST); 2712 2713 /* 2714 * XXX: Locking. Nothing else seems to lock if_flags, 2715 * and there are numerous other races with the 2716 * ifunit() checks not being atomic with namespace 2717 * changes (renames, vmoves, if_attach, etc). 2718 */ 2719 ifp->if_flags |= IFF_RENAMING; 2720 2721 /* Announce the departure of the interface. */ 2722 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2723 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2724 2725 if_printf(ifp, "changing name to '%s'\n", new_name); 2726 2727 IF_ADDR_WLOCK(ifp); 2728 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2729 ifa = ifp->if_addr; 2730 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2731 namelen = strlen(new_name); 2732 onamelen = sdl->sdl_nlen; 2733 /* 2734 * Move the address if needed. This is safe because we 2735 * allocate space for a name of length IFNAMSIZ when we 2736 * create this in if_attach(). 2737 */ 2738 if (namelen != onamelen) { 2739 bcopy(sdl->sdl_data + onamelen, 2740 sdl->sdl_data + namelen, sdl->sdl_alen); 2741 } 2742 bcopy(new_name, sdl->sdl_data, namelen); 2743 sdl->sdl_nlen = namelen; 2744 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2745 bzero(sdl->sdl_data, onamelen); 2746 while (namelen != 0) 2747 sdl->sdl_data[--namelen] = 0xff; 2748 IF_ADDR_WUNLOCK(ifp); 2749 2750 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2751 /* Announce the return of the interface. */ 2752 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2753 2754 ifp->if_flags &= ~IFF_RENAMING; 2755 break; 2756 2757 #ifdef VIMAGE 2758 case SIOCSIFVNET: 2759 error = priv_check(td, PRIV_NET_SETIFVNET); 2760 if (error) 2761 return (error); 2762 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2763 break; 2764 #endif 2765 2766 case SIOCSIFMETRIC: 2767 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2768 if (error) 2769 return (error); 2770 ifp->if_metric = ifr->ifr_metric; 2771 getmicrotime(&ifp->if_lastchange); 2772 break; 2773 2774 case SIOCSIFPHYS: 2775 error = priv_check(td, PRIV_NET_SETIFPHYS); 2776 if (error) 2777 return (error); 2778 if (ifp->if_ioctl == NULL) 2779 return (EOPNOTSUPP); 2780 error = (*ifp->if_ioctl)(ifp, cmd, data); 2781 if (error == 0) 2782 getmicrotime(&ifp->if_lastchange); 2783 break; 2784 2785 case SIOCSIFMTU: 2786 { 2787 u_long oldmtu = ifp->if_mtu; 2788 2789 error = priv_check(td, PRIV_NET_SETIFMTU); 2790 if (error) 2791 return (error); 2792 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2793 return (EINVAL); 2794 if (ifp->if_ioctl == NULL) 2795 return (EOPNOTSUPP); 2796 error = (*ifp->if_ioctl)(ifp, cmd, data); 2797 if (error == 0) { 2798 getmicrotime(&ifp->if_lastchange); 2799 rt_ifmsg(ifp); 2800 #ifdef INET 2801 NETDUMP_REINIT(ifp); 2802 #endif 2803 } 2804 /* 2805 * If the link MTU changed, do network layer specific procedure. 2806 */ 2807 if (ifp->if_mtu != oldmtu) { 2808 #ifdef INET6 2809 nd6_setmtu(ifp); 2810 #endif 2811 rt_updatemtu(ifp); 2812 } 2813 break; 2814 } 2815 2816 case SIOCADDMULTI: 2817 case SIOCDELMULTI: 2818 if (cmd == SIOCADDMULTI) 2819 error = priv_check(td, PRIV_NET_ADDMULTI); 2820 else 2821 error = priv_check(td, PRIV_NET_DELMULTI); 2822 if (error) 2823 return (error); 2824 2825 /* Don't allow group membership on non-multicast interfaces. */ 2826 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2827 return (EOPNOTSUPP); 2828 2829 /* Don't let users screw up protocols' entries. */ 2830 if (ifr->ifr_addr.sa_family != AF_LINK) 2831 return (EINVAL); 2832 2833 if (cmd == SIOCADDMULTI) { 2834 struct ifmultiaddr *ifma; 2835 2836 /* 2837 * Userland is only permitted to join groups once 2838 * via the if_addmulti() KPI, because it cannot hold 2839 * struct ifmultiaddr * between calls. It may also 2840 * lose a race while we check if the membership 2841 * already exists. 2842 */ 2843 IF_ADDR_RLOCK(ifp); 2844 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2845 IF_ADDR_RUNLOCK(ifp); 2846 if (ifma != NULL) 2847 error = EADDRINUSE; 2848 else 2849 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2850 } else { 2851 error = if_delmulti(ifp, &ifr->ifr_addr); 2852 } 2853 if (error == 0) 2854 getmicrotime(&ifp->if_lastchange); 2855 break; 2856 2857 case SIOCSIFPHYADDR: 2858 case SIOCDIFPHYADDR: 2859 #ifdef INET6 2860 case SIOCSIFPHYADDR_IN6: 2861 #endif 2862 case SIOCSIFMEDIA: 2863 case SIOCSIFGENERIC: 2864 error = priv_check(td, PRIV_NET_HWIOCTL); 2865 if (error) 2866 return (error); 2867 if (ifp->if_ioctl == NULL) 2868 return (EOPNOTSUPP); 2869 error = (*ifp->if_ioctl)(ifp, cmd, data); 2870 if (error == 0) 2871 getmicrotime(&ifp->if_lastchange); 2872 break; 2873 2874 case SIOCGIFSTATUS: 2875 case SIOCGIFPSRCADDR: 2876 case SIOCGIFPDSTADDR: 2877 case SIOCGIFMEDIA: 2878 case SIOCGIFXMEDIA: 2879 case SIOCGIFGENERIC: 2880 case SIOCGIFRSSKEY: 2881 case SIOCGIFRSSHASH: 2882 if (ifp->if_ioctl == NULL) 2883 return (EOPNOTSUPP); 2884 error = (*ifp->if_ioctl)(ifp, cmd, data); 2885 break; 2886 2887 case SIOCSIFLLADDR: 2888 error = priv_check(td, PRIV_NET_SETLLADDR); 2889 if (error) 2890 return (error); 2891 error = if_setlladdr(ifp, 2892 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2893 break; 2894 2895 case SIOCGHWADDR: 2896 error = if_gethwaddr(ifp, ifr); 2897 break; 2898 2899 CASE_IOC_IFGROUPREQ(SIOCAIFGROUP): 2900 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2901 if (error) 2902 return (error); 2903 if ((error = if_addgroup(ifp, 2904 ifgr_group_get((struct ifgroupreq *)data)))) 2905 return (error); 2906 break; 2907 2908 CASE_IOC_IFGROUPREQ(SIOCGIFGROUP): 2909 if ((error = if_getgroup((struct ifgroupreq *)data, ifp))) 2910 return (error); 2911 break; 2912 2913 CASE_IOC_IFGROUPREQ(SIOCDIFGROUP): 2914 error = priv_check(td, PRIV_NET_DELIFGROUP); 2915 if (error) 2916 return (error); 2917 if ((error = if_delgroup(ifp, 2918 ifgr_group_get((struct ifgroupreq *)data)))) 2919 return (error); 2920 break; 2921 2922 default: 2923 error = ENOIOCTL; 2924 break; 2925 } 2926 return (error); 2927 } 2928 2929 #ifdef COMPAT_FREEBSD32 2930 struct ifconf32 { 2931 int32_t ifc_len; 2932 union { 2933 uint32_t ifcu_buf; 2934 uint32_t ifcu_req; 2935 } ifc_ifcu; 2936 }; 2937 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2938 #endif 2939 2940 #ifdef COMPAT_FREEBSD32 2941 static void 2942 ifmr_init(struct ifmediareq *ifmr, caddr_t data) 2943 { 2944 struct ifmediareq32 *ifmr32; 2945 2946 ifmr32 = (struct ifmediareq32 *)data; 2947 memcpy(ifmr->ifm_name, ifmr32->ifm_name, 2948 sizeof(ifmr->ifm_name)); 2949 ifmr->ifm_current = ifmr32->ifm_current; 2950 ifmr->ifm_mask = ifmr32->ifm_mask; 2951 ifmr->ifm_status = ifmr32->ifm_status; 2952 ifmr->ifm_active = ifmr32->ifm_active; 2953 ifmr->ifm_count = ifmr32->ifm_count; 2954 ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist; 2955 } 2956 2957 static void 2958 ifmr_update(const struct ifmediareq *ifmr, caddr_t data) 2959 { 2960 struct ifmediareq32 *ifmr32; 2961 2962 ifmr32 = (struct ifmediareq32 *)data; 2963 ifmr32->ifm_current = ifmr->ifm_current; 2964 ifmr32->ifm_mask = ifmr->ifm_mask; 2965 ifmr32->ifm_status = ifmr->ifm_status; 2966 ifmr32->ifm_active = ifmr->ifm_active; 2967 ifmr32->ifm_count = ifmr->ifm_count; 2968 } 2969 #endif 2970 2971 /* 2972 * Interface ioctls. 2973 */ 2974 int 2975 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2976 { 2977 #ifdef COMPAT_FREEBSD32 2978 caddr_t saved_data = NULL; 2979 struct ifmediareq ifmr; 2980 struct ifmediareq *ifmrp; 2981 #endif 2982 struct ifnet *ifp; 2983 struct ifreq *ifr; 2984 int error; 2985 int oif_flags; 2986 #ifdef VIMAGE 2987 int shutdown; 2988 #endif 2989 2990 CURVNET_SET(so->so_vnet); 2991 #ifdef VIMAGE 2992 /* Make sure the VNET is stable. */ 2993 shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET && 2994 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 2995 if (shutdown) { 2996 CURVNET_RESTORE(); 2997 return (EBUSY); 2998 } 2999 #endif 3000 3001 3002 switch (cmd) { 3003 case SIOCGIFCONF: 3004 error = ifconf(cmd, data); 3005 CURVNET_RESTORE(); 3006 return (error); 3007 3008 #ifdef COMPAT_FREEBSD32 3009 case SIOCGIFCONF32: 3010 { 3011 struct ifconf32 *ifc32; 3012 struct ifconf ifc; 3013 3014 ifc32 = (struct ifconf32 *)data; 3015 ifc.ifc_len = ifc32->ifc_len; 3016 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 3017 3018 error = ifconf(SIOCGIFCONF, (void *)&ifc); 3019 CURVNET_RESTORE(); 3020 if (error == 0) 3021 ifc32->ifc_len = ifc.ifc_len; 3022 return (error); 3023 } 3024 #endif 3025 } 3026 3027 #ifdef COMPAT_FREEBSD32 3028 ifmrp = NULL; 3029 switch (cmd) { 3030 case SIOCGIFMEDIA32: 3031 case SIOCGIFXMEDIA32: 3032 ifmrp = &ifmr; 3033 ifmr_init(ifmrp, data); 3034 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 3035 saved_data = data; 3036 data = (caddr_t)ifmrp; 3037 } 3038 #endif 3039 3040 ifr = (struct ifreq *)data; 3041 switch (cmd) { 3042 #ifdef VIMAGE 3043 case SIOCSIFRVNET: 3044 error = priv_check(td, PRIV_NET_SETIFVNET); 3045 if (error == 0) 3046 error = if_vmove_reclaim(td, ifr->ifr_name, 3047 ifr->ifr_jid); 3048 goto out_noref; 3049 #endif 3050 case SIOCIFCREATE: 3051 case SIOCIFCREATE2: 3052 error = priv_check(td, PRIV_NET_IFCREATE); 3053 if (error == 0) 3054 error = if_clone_create(ifr->ifr_name, 3055 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3056 ifr_data_get_ptr(ifr) : NULL); 3057 goto out_noref; 3058 case SIOCIFDESTROY: 3059 error = priv_check(td, PRIV_NET_IFDESTROY); 3060 if (error == 0) 3061 error = if_clone_destroy(ifr->ifr_name); 3062 goto out_noref; 3063 3064 case SIOCIFGCLONERS: 3065 error = if_clone_list((struct if_clonereq *)data); 3066 goto out_noref; 3067 3068 CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB): 3069 error = if_getgroupmembers((struct ifgroupreq *)data); 3070 goto out_noref; 3071 3072 #if defined(INET) || defined(INET6) 3073 case SIOCSVH: 3074 case SIOCGVH: 3075 if (carp_ioctl_p == NULL) 3076 error = EPROTONOSUPPORT; 3077 else 3078 error = (*carp_ioctl_p)(ifr, cmd, td); 3079 goto out_noref; 3080 #endif 3081 } 3082 3083 ifp = ifunit_ref(ifr->ifr_name); 3084 if (ifp == NULL) { 3085 error = ENXIO; 3086 goto out_noref; 3087 } 3088 3089 error = ifhwioctl(cmd, ifp, data, td); 3090 if (error != ENOIOCTL) 3091 goto out_ref; 3092 3093 oif_flags = ifp->if_flags; 3094 if (so->so_proto == NULL) { 3095 error = EOPNOTSUPP; 3096 goto out_ref; 3097 } 3098 3099 /* 3100 * Pass the request on to the socket control method, and if the 3101 * latter returns EOPNOTSUPP, directly to the interface. 3102 * 3103 * Make an exception for the legacy SIOCSIF* requests. Drivers 3104 * trust SIOCSIFADDR et al to come from an already privileged 3105 * layer, and do not perform any credentials checks or input 3106 * validation. 3107 */ 3108 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 3109 ifp, td)); 3110 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3111 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3112 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3113 error = (*ifp->if_ioctl)(ifp, cmd, data); 3114 3115 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3116 #ifdef INET6 3117 if (ifp->if_flags & IFF_UP) 3118 in6_if_up(ifp); 3119 #endif 3120 } 3121 3122 out_ref: 3123 if_rele(ifp); 3124 out_noref: 3125 #ifdef COMPAT_FREEBSD32 3126 if (ifmrp != NULL) { 3127 KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA), 3128 ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx", 3129 cmd)); 3130 data = saved_data; 3131 ifmr_update(ifmrp, data); 3132 } 3133 #endif 3134 CURVNET_RESTORE(); 3135 return (error); 3136 } 3137 3138 /* 3139 * The code common to handling reference counted flags, 3140 * e.g., in ifpromisc() and if_allmulti(). 3141 * The "pflag" argument can specify a permanent mode flag to check, 3142 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3143 * 3144 * Only to be used on stack-owned flags, not driver-owned flags. 3145 */ 3146 static int 3147 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3148 { 3149 struct ifreq ifr; 3150 int error; 3151 int oldflags, oldcount; 3152 3153 /* Sanity checks to catch programming errors */ 3154 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3155 ("%s: setting driver-owned flag %d", __func__, flag)); 3156 3157 if (onswitch) 3158 KASSERT(*refcount >= 0, 3159 ("%s: increment negative refcount %d for flag %d", 3160 __func__, *refcount, flag)); 3161 else 3162 KASSERT(*refcount > 0, 3163 ("%s: decrement non-positive refcount %d for flag %d", 3164 __func__, *refcount, flag)); 3165 3166 /* In case this mode is permanent, just touch refcount */ 3167 if (ifp->if_flags & pflag) { 3168 *refcount += onswitch ? 1 : -1; 3169 return (0); 3170 } 3171 3172 /* Save ifnet parameters for if_ioctl() may fail */ 3173 oldcount = *refcount; 3174 oldflags = ifp->if_flags; 3175 3176 /* 3177 * See if we aren't the only and touching refcount is enough. 3178 * Actually toggle interface flag if we are the first or last. 3179 */ 3180 if (onswitch) { 3181 if ((*refcount)++) 3182 return (0); 3183 ifp->if_flags |= flag; 3184 } else { 3185 if (--(*refcount)) 3186 return (0); 3187 ifp->if_flags &= ~flag; 3188 } 3189 3190 /* Call down the driver since we've changed interface flags */ 3191 if (ifp->if_ioctl == NULL) { 3192 error = EOPNOTSUPP; 3193 goto recover; 3194 } 3195 ifr.ifr_flags = ifp->if_flags & 0xffff; 3196 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3197 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3198 if (error) 3199 goto recover; 3200 /* Notify userland that interface flags have changed */ 3201 rt_ifmsg(ifp); 3202 return (0); 3203 3204 recover: 3205 /* Recover after driver error */ 3206 *refcount = oldcount; 3207 ifp->if_flags = oldflags; 3208 return (error); 3209 } 3210 3211 /* 3212 * Set/clear promiscuous mode on interface ifp based on the truth value 3213 * of pswitch. The calls are reference counted so that only the first 3214 * "on" request actually has an effect, as does the final "off" request. 3215 * Results are undefined if the "off" and "on" requests are not matched. 3216 */ 3217 int 3218 ifpromisc(struct ifnet *ifp, int pswitch) 3219 { 3220 int error; 3221 int oldflags = ifp->if_flags; 3222 3223 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3224 &ifp->if_pcount, pswitch); 3225 /* If promiscuous mode status has changed, log a message */ 3226 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3227 log_promisc_mode_change) 3228 if_printf(ifp, "promiscuous mode %s\n", 3229 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3230 return (error); 3231 } 3232 3233 /* 3234 * Return interface configuration 3235 * of system. List may be used 3236 * in later ioctl's (above) to get 3237 * other information. 3238 */ 3239 /*ARGSUSED*/ 3240 static int 3241 ifconf(u_long cmd, caddr_t data) 3242 { 3243 struct ifconf *ifc = (struct ifconf *)data; 3244 struct ifnet *ifp; 3245 struct ifaddr *ifa; 3246 struct ifreq ifr; 3247 struct sbuf *sb; 3248 int error, full = 0, valid_len, max_len; 3249 3250 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 3251 max_len = MAXPHYS - 1; 3252 3253 /* Prevent hostile input from being able to crash the system */ 3254 if (ifc->ifc_len <= 0) 3255 return (EINVAL); 3256 3257 again: 3258 if (ifc->ifc_len <= max_len) { 3259 max_len = ifc->ifc_len; 3260 full = 1; 3261 } 3262 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3263 max_len = 0; 3264 valid_len = 0; 3265 3266 IFNET_RLOCK(); 3267 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3268 int addrs; 3269 3270 /* 3271 * Zero the ifr to make sure we don't disclose the contents 3272 * of the stack. 3273 */ 3274 memset(&ifr, 0, sizeof(ifr)); 3275 3276 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3277 >= sizeof(ifr.ifr_name)) { 3278 sbuf_delete(sb); 3279 IFNET_RUNLOCK(); 3280 return (ENAMETOOLONG); 3281 } 3282 3283 addrs = 0; 3284 IF_ADDR_RLOCK(ifp); 3285 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3286 struct sockaddr *sa = ifa->ifa_addr; 3287 3288 if (prison_if(curthread->td_ucred, sa) != 0) 3289 continue; 3290 addrs++; 3291 if (sa->sa_len <= sizeof(*sa)) { 3292 if (sa->sa_len < sizeof(*sa)) { 3293 memset(&ifr.ifr_ifru.ifru_addr, 0, 3294 sizeof(ifr.ifr_ifru.ifru_addr)); 3295 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3296 sa->sa_len); 3297 } else 3298 ifr.ifr_ifru.ifru_addr = *sa; 3299 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3300 max_len += sizeof(ifr); 3301 } else { 3302 sbuf_bcat(sb, &ifr, 3303 offsetof(struct ifreq, ifr_addr)); 3304 max_len += offsetof(struct ifreq, ifr_addr); 3305 sbuf_bcat(sb, sa, sa->sa_len); 3306 max_len += sa->sa_len; 3307 } 3308 3309 if (sbuf_error(sb) == 0) 3310 valid_len = sbuf_len(sb); 3311 } 3312 IF_ADDR_RUNLOCK(ifp); 3313 if (addrs == 0) { 3314 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3315 max_len += sizeof(ifr); 3316 3317 if (sbuf_error(sb) == 0) 3318 valid_len = sbuf_len(sb); 3319 } 3320 } 3321 IFNET_RUNLOCK(); 3322 3323 /* 3324 * If we didn't allocate enough space (uncommon), try again. If 3325 * we have already allocated as much space as we are allowed, 3326 * return what we've got. 3327 */ 3328 if (valid_len != max_len && !full) { 3329 sbuf_delete(sb); 3330 goto again; 3331 } 3332 3333 ifc->ifc_len = valid_len; 3334 sbuf_finish(sb); 3335 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3336 sbuf_delete(sb); 3337 return (error); 3338 } 3339 3340 /* 3341 * Just like ifpromisc(), but for all-multicast-reception mode. 3342 */ 3343 int 3344 if_allmulti(struct ifnet *ifp, int onswitch) 3345 { 3346 3347 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3348 } 3349 3350 struct ifmultiaddr * 3351 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3352 { 3353 struct ifmultiaddr *ifma; 3354 3355 IF_ADDR_LOCK_ASSERT(ifp); 3356 3357 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3358 if (sa->sa_family == AF_LINK) { 3359 if (sa_dl_equal(ifma->ifma_addr, sa)) 3360 break; 3361 } else { 3362 if (sa_equal(ifma->ifma_addr, sa)) 3363 break; 3364 } 3365 } 3366 3367 return ifma; 3368 } 3369 3370 /* 3371 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3372 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3373 * the ifnet multicast address list here, so the caller must do that and 3374 * other setup work (such as notifying the device driver). The reference 3375 * count is initialized to 1. 3376 */ 3377 static struct ifmultiaddr * 3378 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3379 int mflags) 3380 { 3381 struct ifmultiaddr *ifma; 3382 struct sockaddr *dupsa; 3383 3384 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3385 M_ZERO); 3386 if (ifma == NULL) 3387 return (NULL); 3388 3389 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3390 if (dupsa == NULL) { 3391 free(ifma, M_IFMADDR); 3392 return (NULL); 3393 } 3394 bcopy(sa, dupsa, sa->sa_len); 3395 ifma->ifma_addr = dupsa; 3396 3397 ifma->ifma_ifp = ifp; 3398 ifma->ifma_refcount = 1; 3399 ifma->ifma_protospec = NULL; 3400 3401 if (llsa == NULL) { 3402 ifma->ifma_lladdr = NULL; 3403 return (ifma); 3404 } 3405 3406 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3407 if (dupsa == NULL) { 3408 free(ifma->ifma_addr, M_IFMADDR); 3409 free(ifma, M_IFMADDR); 3410 return (NULL); 3411 } 3412 bcopy(llsa, dupsa, llsa->sa_len); 3413 ifma->ifma_lladdr = dupsa; 3414 3415 return (ifma); 3416 } 3417 3418 /* 3419 * if_freemulti: free ifmultiaddr structure and possibly attached related 3420 * addresses. The caller is responsible for implementing reference 3421 * counting, notifying the driver, handling routing messages, and releasing 3422 * any dependent link layer state. 3423 */ 3424 #ifdef MCAST_VERBOSE 3425 extern void kdb_backtrace(void); 3426 #endif 3427 static void 3428 if_freemulti_internal(struct ifmultiaddr *ifma) 3429 { 3430 3431 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3432 ifma->ifma_refcount)); 3433 3434 if (ifma->ifma_lladdr != NULL) 3435 free(ifma->ifma_lladdr, M_IFMADDR); 3436 #ifdef MCAST_VERBOSE 3437 kdb_backtrace(); 3438 printf("%s freeing ifma: %p\n", __func__, ifma); 3439 #endif 3440 free(ifma->ifma_addr, M_IFMADDR); 3441 free(ifma, M_IFMADDR); 3442 } 3443 3444 static void 3445 if_destroymulti(epoch_context_t ctx) 3446 { 3447 struct ifmultiaddr *ifma; 3448 3449 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3450 if_freemulti_internal(ifma); 3451 } 3452 3453 void 3454 if_freemulti(struct ifmultiaddr *ifma) 3455 { 3456 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3457 ifma->ifma_refcount)); 3458 3459 epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti); 3460 } 3461 3462 3463 /* 3464 * Register an additional multicast address with a network interface. 3465 * 3466 * - If the address is already present, bump the reference count on the 3467 * address and return. 3468 * - If the address is not link-layer, look up a link layer address. 3469 * - Allocate address structures for one or both addresses, and attach to the 3470 * multicast address list on the interface. If automatically adding a link 3471 * layer address, the protocol address will own a reference to the link 3472 * layer address, to be freed when it is freed. 3473 * - Notify the network device driver of an addition to the multicast address 3474 * list. 3475 * 3476 * 'sa' points to caller-owned memory with the desired multicast address. 3477 * 3478 * 'retifma' will be used to return a pointer to the resulting multicast 3479 * address reference, if desired. 3480 */ 3481 int 3482 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3483 struct ifmultiaddr **retifma) 3484 { 3485 struct ifmultiaddr *ifma, *ll_ifma; 3486 struct sockaddr *llsa; 3487 struct sockaddr_dl sdl; 3488 int error; 3489 3490 #ifdef INET 3491 IN_MULTI_LIST_UNLOCK_ASSERT(); 3492 #endif 3493 #ifdef INET6 3494 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3495 #endif 3496 /* 3497 * If the address is already present, return a new reference to it; 3498 * otherwise, allocate storage and set up a new address. 3499 */ 3500 IF_ADDR_WLOCK(ifp); 3501 ifma = if_findmulti(ifp, sa); 3502 if (ifma != NULL) { 3503 ifma->ifma_refcount++; 3504 if (retifma != NULL) 3505 *retifma = ifma; 3506 IF_ADDR_WUNLOCK(ifp); 3507 return (0); 3508 } 3509 3510 /* 3511 * The address isn't already present; resolve the protocol address 3512 * into a link layer address, and then look that up, bump its 3513 * refcount or allocate an ifma for that also. 3514 * Most link layer resolving functions returns address data which 3515 * fits inside default sockaddr_dl structure. However callback 3516 * can allocate another sockaddr structure, in that case we need to 3517 * free it later. 3518 */ 3519 llsa = NULL; 3520 ll_ifma = NULL; 3521 if (ifp->if_resolvemulti != NULL) { 3522 /* Provide called function with buffer size information */ 3523 sdl.sdl_len = sizeof(sdl); 3524 llsa = (struct sockaddr *)&sdl; 3525 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3526 if (error) 3527 goto unlock_out; 3528 } 3529 3530 /* 3531 * Allocate the new address. Don't hook it up yet, as we may also 3532 * need to allocate a link layer multicast address. 3533 */ 3534 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3535 if (ifma == NULL) { 3536 error = ENOMEM; 3537 goto free_llsa_out; 3538 } 3539 3540 /* 3541 * If a link layer address is found, we'll need to see if it's 3542 * already present in the address list, or allocate is as well. 3543 * When this block finishes, the link layer address will be on the 3544 * list. 3545 */ 3546 if (llsa != NULL) { 3547 ll_ifma = if_findmulti(ifp, llsa); 3548 if (ll_ifma == NULL) { 3549 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3550 if (ll_ifma == NULL) { 3551 --ifma->ifma_refcount; 3552 if_freemulti(ifma); 3553 error = ENOMEM; 3554 goto free_llsa_out; 3555 } 3556 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3557 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3558 ifma_link); 3559 } else 3560 ll_ifma->ifma_refcount++; 3561 ifma->ifma_llifma = ll_ifma; 3562 } 3563 3564 /* 3565 * We now have a new multicast address, ifma, and possibly a new or 3566 * referenced link layer address. Add the primary address to the 3567 * ifnet address list. 3568 */ 3569 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3570 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3571 3572 if (retifma != NULL) 3573 *retifma = ifma; 3574 3575 /* 3576 * Must generate the message while holding the lock so that 'ifma' 3577 * pointer is still valid. 3578 */ 3579 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3580 IF_ADDR_WUNLOCK(ifp); 3581 3582 /* 3583 * We are certain we have added something, so call down to the 3584 * interface to let them know about it. 3585 */ 3586 if (ifp->if_ioctl != NULL) { 3587 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3588 } 3589 3590 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3591 link_free_sdl(llsa); 3592 3593 return (0); 3594 3595 free_llsa_out: 3596 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3597 link_free_sdl(llsa); 3598 3599 unlock_out: 3600 IF_ADDR_WUNLOCK(ifp); 3601 return (error); 3602 } 3603 3604 /* 3605 * Delete a multicast group membership by network-layer group address. 3606 * 3607 * Returns ENOENT if the entry could not be found. If ifp no longer 3608 * exists, results are undefined. This entry point should only be used 3609 * from subsystems which do appropriate locking to hold ifp for the 3610 * duration of the call. 3611 * Network-layer protocol domains must use if_delmulti_ifma(). 3612 */ 3613 int 3614 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3615 { 3616 struct ifmultiaddr *ifma; 3617 int lastref; 3618 #ifdef INVARIANTS 3619 struct ifnet *oifp; 3620 3621 IFNET_RLOCK_NOSLEEP(); 3622 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3623 if (ifp == oifp) 3624 break; 3625 if (ifp != oifp) 3626 ifp = NULL; 3627 IFNET_RUNLOCK_NOSLEEP(); 3628 3629 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3630 #endif 3631 if (ifp == NULL) 3632 return (ENOENT); 3633 3634 IF_ADDR_WLOCK(ifp); 3635 lastref = 0; 3636 ifma = if_findmulti(ifp, sa); 3637 if (ifma != NULL) 3638 lastref = if_delmulti_locked(ifp, ifma, 0); 3639 IF_ADDR_WUNLOCK(ifp); 3640 3641 if (ifma == NULL) 3642 return (ENOENT); 3643 3644 if (lastref && ifp->if_ioctl != NULL) { 3645 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3646 } 3647 3648 return (0); 3649 } 3650 3651 /* 3652 * Delete all multicast group membership for an interface. 3653 * Should be used to quickly flush all multicast filters. 3654 */ 3655 void 3656 if_delallmulti(struct ifnet *ifp) 3657 { 3658 struct ifmultiaddr *ifma; 3659 struct ifmultiaddr *next; 3660 3661 IF_ADDR_WLOCK(ifp); 3662 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3663 if_delmulti_locked(ifp, ifma, 0); 3664 IF_ADDR_WUNLOCK(ifp); 3665 } 3666 3667 void 3668 if_delmulti_ifma(struct ifmultiaddr *ifma) 3669 { 3670 if_delmulti_ifma_flags(ifma, 0); 3671 } 3672 3673 /* 3674 * Delete a multicast group membership by group membership pointer. 3675 * Network-layer protocol domains must use this routine. 3676 * 3677 * It is safe to call this routine if the ifp disappeared. 3678 */ 3679 void 3680 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3681 { 3682 struct ifnet *ifp; 3683 int lastref; 3684 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3685 #ifdef INET 3686 IN_MULTI_LIST_UNLOCK_ASSERT(); 3687 #endif 3688 ifp = ifma->ifma_ifp; 3689 #ifdef DIAGNOSTIC 3690 if (ifp == NULL) { 3691 printf("%s: ifma_ifp seems to be detached\n", __func__); 3692 } else { 3693 struct ifnet *oifp; 3694 3695 IFNET_RLOCK_NOSLEEP(); 3696 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3697 if (ifp == oifp) 3698 break; 3699 if (ifp != oifp) 3700 ifp = NULL; 3701 IFNET_RUNLOCK_NOSLEEP(); 3702 } 3703 #endif 3704 /* 3705 * If and only if the ifnet instance exists: Acquire the address lock. 3706 */ 3707 if (ifp != NULL) 3708 IF_ADDR_WLOCK(ifp); 3709 3710 lastref = if_delmulti_locked(ifp, ifma, flags); 3711 3712 if (ifp != NULL) { 3713 /* 3714 * If and only if the ifnet instance exists: 3715 * Release the address lock. 3716 * If the group was left: update the hardware hash filter. 3717 */ 3718 IF_ADDR_WUNLOCK(ifp); 3719 if (lastref && ifp->if_ioctl != NULL) { 3720 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3721 } 3722 } 3723 } 3724 3725 /* 3726 * Perform deletion of network-layer and/or link-layer multicast address. 3727 * 3728 * Return 0 if the reference count was decremented. 3729 * Return 1 if the final reference was released, indicating that the 3730 * hardware hash filter should be reprogrammed. 3731 */ 3732 static int 3733 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3734 { 3735 struct ifmultiaddr *ll_ifma; 3736 3737 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3738 KASSERT(ifma->ifma_ifp == ifp, 3739 ("%s: inconsistent ifp %p", __func__, ifp)); 3740 IF_ADDR_WLOCK_ASSERT(ifp); 3741 } 3742 3743 ifp = ifma->ifma_ifp; 3744 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3745 3746 /* 3747 * If the ifnet is detaching, null out references to ifnet, 3748 * so that upper protocol layers will notice, and not attempt 3749 * to obtain locks for an ifnet which no longer exists. The 3750 * routing socket announcement must happen before the ifnet 3751 * instance is detached from the system. 3752 */ 3753 if (detaching) { 3754 #ifdef DIAGNOSTIC 3755 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3756 #endif 3757 /* 3758 * ifp may already be nulled out if we are being reentered 3759 * to delete the ll_ifma. 3760 */ 3761 if (ifp != NULL) { 3762 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3763 ifma->ifma_ifp = NULL; 3764 } 3765 } 3766 3767 if (--ifma->ifma_refcount > 0) 3768 return 0; 3769 3770 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3771 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3772 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3773 } 3774 /* 3775 * If this ifma is a network-layer ifma, a link-layer ifma may 3776 * have been associated with it. Release it first if so. 3777 */ 3778 ll_ifma = ifma->ifma_llifma; 3779 if (ll_ifma != NULL) { 3780 KASSERT(ifma->ifma_lladdr != NULL, 3781 ("%s: llifma w/o lladdr", __func__)); 3782 if (detaching) 3783 ll_ifma->ifma_ifp = NULL; /* XXX */ 3784 if (--ll_ifma->ifma_refcount == 0) { 3785 if (ifp != NULL) { 3786 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3787 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3788 ifma_link); 3789 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3790 } 3791 } 3792 if_freemulti(ll_ifma); 3793 } 3794 } 3795 #ifdef INVARIANTS 3796 if (ifp) { 3797 struct ifmultiaddr *ifmatmp; 3798 3799 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3800 MPASS(ifma != ifmatmp); 3801 } 3802 #endif 3803 if_freemulti(ifma); 3804 /* 3805 * The last reference to this instance of struct ifmultiaddr 3806 * was released; the hardware should be notified of this change. 3807 */ 3808 return 1; 3809 } 3810 3811 /* 3812 * Set the link layer address on an interface. 3813 * 3814 * At this time we only support certain types of interfaces, 3815 * and we don't allow the length of the address to change. 3816 * 3817 * Set noinline to be dtrace-friendly 3818 */ 3819 __noinline int 3820 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3821 { 3822 struct sockaddr_dl *sdl; 3823 struct ifaddr *ifa; 3824 struct ifreq ifr; 3825 int rc; 3826 3827 rc = 0; 3828 NET_EPOCH_ENTER(); 3829 ifa = ifp->if_addr; 3830 if (ifa == NULL) { 3831 rc = EINVAL; 3832 goto out; 3833 } 3834 3835 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3836 if (sdl == NULL) { 3837 rc = EINVAL; 3838 goto out; 3839 } 3840 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3841 rc = EINVAL; 3842 goto out; 3843 } 3844 switch (ifp->if_type) { 3845 case IFT_ETHER: 3846 case IFT_XETHER: 3847 case IFT_L2VLAN: 3848 case IFT_BRIDGE: 3849 case IFT_IEEE8023ADLAG: 3850 bcopy(lladdr, LLADDR(sdl), len); 3851 break; 3852 default: 3853 rc = ENODEV; 3854 goto out; 3855 } 3856 3857 /* 3858 * If the interface is already up, we need 3859 * to re-init it in order to reprogram its 3860 * address filter. 3861 */ 3862 NET_EPOCH_EXIT(); 3863 if ((ifp->if_flags & IFF_UP) != 0) { 3864 if (ifp->if_ioctl) { 3865 ifp->if_flags &= ~IFF_UP; 3866 ifr.ifr_flags = ifp->if_flags & 0xffff; 3867 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3868 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3869 ifp->if_flags |= IFF_UP; 3870 ifr.ifr_flags = ifp->if_flags & 0xffff; 3871 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3872 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3873 } 3874 } 3875 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3876 return (0); 3877 out: 3878 NET_EPOCH_EXIT(); 3879 return (rc); 3880 } 3881 3882 /* 3883 * Compat function for handling basic encapsulation requests. 3884 * Not converted stacks (FDDI, IB, ..) supports traditional 3885 * output model: ARP (and other similar L2 protocols) are handled 3886 * inside output routine, arpresolve/nd6_resolve() returns MAC 3887 * address instead of full prepend. 3888 * 3889 * This function creates calculated header==MAC for IPv4/IPv6 and 3890 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3891 * address families. 3892 */ 3893 static int 3894 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3895 { 3896 3897 if (req->rtype != IFENCAP_LL) 3898 return (EOPNOTSUPP); 3899 3900 if (req->bufsize < req->lladdr_len) 3901 return (ENOMEM); 3902 3903 switch (req->family) { 3904 case AF_INET: 3905 case AF_INET6: 3906 break; 3907 default: 3908 return (EAFNOSUPPORT); 3909 } 3910 3911 /* Copy lladdr to storage as is */ 3912 memmove(req->buf, req->lladdr, req->lladdr_len); 3913 req->bufsize = req->lladdr_len; 3914 req->lladdr_off = 0; 3915 3916 return (0); 3917 } 3918 3919 /* 3920 * Tunnel interfaces can nest, also they may cause infinite recursion 3921 * calls when misconfigured. We'll prevent this by detecting loops. 3922 * High nesting level may cause stack exhaustion. We'll prevent this 3923 * by introducing upper limit. 3924 * 3925 * Return 0, if tunnel nesting count is equal or less than limit. 3926 */ 3927 int 3928 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3929 int limit) 3930 { 3931 struct m_tag *mtag; 3932 int count; 3933 3934 count = 1; 3935 mtag = NULL; 3936 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3937 if (*(struct ifnet **)(mtag + 1) == ifp) { 3938 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3939 return (EIO); 3940 } 3941 count++; 3942 } 3943 if (count > limit) { 3944 log(LOG_NOTICE, 3945 "%s: if_output recursively called too many times(%d)\n", 3946 if_name(ifp), count); 3947 return (EIO); 3948 } 3949 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3950 if (mtag == NULL) 3951 return (ENOMEM); 3952 *(struct ifnet **)(mtag + 1) = ifp; 3953 m_tag_prepend(m, mtag); 3954 return (0); 3955 } 3956 3957 /* 3958 * Get the link layer address that was read from the hardware at attach. 3959 * 3960 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3961 * their component interfaces as IFT_IEEE8023ADLAG. 3962 */ 3963 int 3964 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3965 { 3966 3967 if (ifp->if_hw_addr == NULL) 3968 return (ENODEV); 3969 3970 switch (ifp->if_type) { 3971 case IFT_ETHER: 3972 case IFT_IEEE8023ADLAG: 3973 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3974 return (0); 3975 default: 3976 return (ENODEV); 3977 } 3978 } 3979 3980 /* 3981 * The name argument must be a pointer to storage which will last as 3982 * long as the interface does. For physical devices, the result of 3983 * device_get_name(dev) is a good choice and for pseudo-devices a 3984 * static string works well. 3985 */ 3986 void 3987 if_initname(struct ifnet *ifp, const char *name, int unit) 3988 { 3989 ifp->if_dname = name; 3990 ifp->if_dunit = unit; 3991 if (unit != IF_DUNIT_NONE) 3992 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3993 else 3994 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3995 } 3996 3997 int 3998 if_printf(struct ifnet *ifp, const char *fmt, ...) 3999 { 4000 char if_fmt[256]; 4001 va_list ap; 4002 4003 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 4004 va_start(ap, fmt); 4005 vlog(LOG_INFO, if_fmt, ap); 4006 va_end(ap); 4007 return (0); 4008 } 4009 4010 void 4011 if_start(struct ifnet *ifp) 4012 { 4013 4014 (*(ifp)->if_start)(ifp); 4015 } 4016 4017 /* 4018 * Backwards compatibility interface for drivers 4019 * that have not implemented it 4020 */ 4021 static int 4022 if_transmit(struct ifnet *ifp, struct mbuf *m) 4023 { 4024 int error; 4025 4026 IFQ_HANDOFF(ifp, m, error); 4027 return (error); 4028 } 4029 4030 static void 4031 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4032 { 4033 4034 m_freem(m); 4035 } 4036 4037 int 4038 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4039 { 4040 int active = 0; 4041 4042 IF_LOCK(ifq); 4043 if (_IF_QFULL(ifq)) { 4044 IF_UNLOCK(ifq); 4045 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4046 m_freem(m); 4047 return (0); 4048 } 4049 if (ifp != NULL) { 4050 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4051 if (m->m_flags & (M_BCAST|M_MCAST)) 4052 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4053 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4054 } 4055 _IF_ENQUEUE(ifq, m); 4056 IF_UNLOCK(ifq); 4057 if (ifp != NULL && !active) 4058 (*(ifp)->if_start)(ifp); 4059 return (1); 4060 } 4061 4062 void 4063 if_register_com_alloc(u_char type, 4064 if_com_alloc_t *a, if_com_free_t *f) 4065 { 4066 4067 KASSERT(if_com_alloc[type] == NULL, 4068 ("if_register_com_alloc: %d already registered", type)); 4069 KASSERT(if_com_free[type] == NULL, 4070 ("if_register_com_alloc: %d free already registered", type)); 4071 4072 if_com_alloc[type] = a; 4073 if_com_free[type] = f; 4074 } 4075 4076 void 4077 if_deregister_com_alloc(u_char type) 4078 { 4079 4080 KASSERT(if_com_alloc[type] != NULL, 4081 ("if_deregister_com_alloc: %d not registered", type)); 4082 KASSERT(if_com_free[type] != NULL, 4083 ("if_deregister_com_alloc: %d free not registered", type)); 4084 if_com_alloc[type] = NULL; 4085 if_com_free[type] = NULL; 4086 } 4087 4088 /* API for driver access to network stack owned ifnet.*/ 4089 uint64_t 4090 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4091 { 4092 uint64_t oldbrate; 4093 4094 oldbrate = ifp->if_baudrate; 4095 ifp->if_baudrate = baudrate; 4096 return (oldbrate); 4097 } 4098 4099 uint64_t 4100 if_getbaudrate(if_t ifp) 4101 { 4102 4103 return (((struct ifnet *)ifp)->if_baudrate); 4104 } 4105 4106 int 4107 if_setcapabilities(if_t ifp, int capabilities) 4108 { 4109 ((struct ifnet *)ifp)->if_capabilities = capabilities; 4110 return (0); 4111 } 4112 4113 int 4114 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4115 { 4116 ((struct ifnet *)ifp)->if_capabilities |= setbit; 4117 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 4118 4119 return (0); 4120 } 4121 4122 int 4123 if_getcapabilities(if_t ifp) 4124 { 4125 return ((struct ifnet *)ifp)->if_capabilities; 4126 } 4127 4128 int 4129 if_setcapenable(if_t ifp, int capabilities) 4130 { 4131 ((struct ifnet *)ifp)->if_capenable = capabilities; 4132 return (0); 4133 } 4134 4135 int 4136 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4137 { 4138 if(setcap) 4139 ((struct ifnet *)ifp)->if_capenable |= setcap; 4140 if(clearcap) 4141 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 4142 4143 return (0); 4144 } 4145 4146 const char * 4147 if_getdname(if_t ifp) 4148 { 4149 return ((struct ifnet *)ifp)->if_dname; 4150 } 4151 4152 int 4153 if_togglecapenable(if_t ifp, int togglecap) 4154 { 4155 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 4156 return (0); 4157 } 4158 4159 int 4160 if_getcapenable(if_t ifp) 4161 { 4162 return ((struct ifnet *)ifp)->if_capenable; 4163 } 4164 4165 /* 4166 * This is largely undesirable because it ties ifnet to a device, but does 4167 * provide flexiblity for an embedded product vendor. Should be used with 4168 * the understanding that it violates the interface boundaries, and should be 4169 * a last resort only. 4170 */ 4171 int 4172 if_setdev(if_t ifp, void *dev) 4173 { 4174 return (0); 4175 } 4176 4177 int 4178 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4179 { 4180 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 4181 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 4182 4183 return (0); 4184 } 4185 4186 int 4187 if_getdrvflags(if_t ifp) 4188 { 4189 return ((struct ifnet *)ifp)->if_drv_flags; 4190 } 4191 4192 int 4193 if_setdrvflags(if_t ifp, int flags) 4194 { 4195 ((struct ifnet *)ifp)->if_drv_flags = flags; 4196 return (0); 4197 } 4198 4199 4200 int 4201 if_setflags(if_t ifp, int flags) 4202 { 4203 ((struct ifnet *)ifp)->if_flags = flags; 4204 return (0); 4205 } 4206 4207 int 4208 if_setflagbits(if_t ifp, int set, int clear) 4209 { 4210 ((struct ifnet *)ifp)->if_flags |= set; 4211 ((struct ifnet *)ifp)->if_flags &= ~clear; 4212 4213 return (0); 4214 } 4215 4216 int 4217 if_getflags(if_t ifp) 4218 { 4219 return ((struct ifnet *)ifp)->if_flags; 4220 } 4221 4222 int 4223 if_clearhwassist(if_t ifp) 4224 { 4225 ((struct ifnet *)ifp)->if_hwassist = 0; 4226 return (0); 4227 } 4228 4229 int 4230 if_sethwassistbits(if_t ifp, int toset, int toclear) 4231 { 4232 ((struct ifnet *)ifp)->if_hwassist |= toset; 4233 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4234 4235 return (0); 4236 } 4237 4238 int 4239 if_sethwassist(if_t ifp, int hwassist_bit) 4240 { 4241 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4242 return (0); 4243 } 4244 4245 int 4246 if_gethwassist(if_t ifp) 4247 { 4248 return ((struct ifnet *)ifp)->if_hwassist; 4249 } 4250 4251 int 4252 if_setmtu(if_t ifp, int mtu) 4253 { 4254 ((struct ifnet *)ifp)->if_mtu = mtu; 4255 return (0); 4256 } 4257 4258 int 4259 if_getmtu(if_t ifp) 4260 { 4261 return ((struct ifnet *)ifp)->if_mtu; 4262 } 4263 4264 int 4265 if_getmtu_family(if_t ifp, int family) 4266 { 4267 struct domain *dp; 4268 4269 for (dp = domains; dp; dp = dp->dom_next) { 4270 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4271 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4272 } 4273 4274 return (((struct ifnet *)ifp)->if_mtu); 4275 } 4276 4277 int 4278 if_setsoftc(if_t ifp, void *softc) 4279 { 4280 ((struct ifnet *)ifp)->if_softc = softc; 4281 return (0); 4282 } 4283 4284 void * 4285 if_getsoftc(if_t ifp) 4286 { 4287 return ((struct ifnet *)ifp)->if_softc; 4288 } 4289 4290 void 4291 if_setrcvif(struct mbuf *m, if_t ifp) 4292 { 4293 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4294 } 4295 4296 void 4297 if_setvtag(struct mbuf *m, uint16_t tag) 4298 { 4299 m->m_pkthdr.ether_vtag = tag; 4300 } 4301 4302 uint16_t 4303 if_getvtag(struct mbuf *m) 4304 { 4305 4306 return (m->m_pkthdr.ether_vtag); 4307 } 4308 4309 int 4310 if_sendq_empty(if_t ifp) 4311 { 4312 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4313 } 4314 4315 struct ifaddr * 4316 if_getifaddr(if_t ifp) 4317 { 4318 return ((struct ifnet *)ifp)->if_addr; 4319 } 4320 4321 int 4322 if_getamcount(if_t ifp) 4323 { 4324 return ((struct ifnet *)ifp)->if_amcount; 4325 } 4326 4327 4328 int 4329 if_setsendqready(if_t ifp) 4330 { 4331 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4332 return (0); 4333 } 4334 4335 int 4336 if_setsendqlen(if_t ifp, int tx_desc_count) 4337 { 4338 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4339 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4340 4341 return (0); 4342 } 4343 4344 int 4345 if_vlantrunkinuse(if_t ifp) 4346 { 4347 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4348 } 4349 4350 int 4351 if_input(if_t ifp, struct mbuf* sendmp) 4352 { 4353 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4354 return (0); 4355 4356 } 4357 4358 /* XXX */ 4359 #ifndef ETH_ADDR_LEN 4360 #define ETH_ADDR_LEN 6 4361 #endif 4362 4363 int 4364 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max) 4365 { 4366 struct ifmultiaddr *ifma; 4367 uint8_t *lmta = (uint8_t *)mta; 4368 int mcnt = 0; 4369 4370 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4371 if (ifma->ifma_addr->sa_family != AF_LINK) 4372 continue; 4373 4374 if (mcnt == max) 4375 break; 4376 4377 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 4378 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 4379 mcnt++; 4380 } 4381 *cnt = mcnt; 4382 4383 return (0); 4384 } 4385 4386 int 4387 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max) 4388 { 4389 int error; 4390 4391 if_maddr_rlock(ifp); 4392 error = if_setupmultiaddr(ifp, mta, cnt, max); 4393 if_maddr_runlock(ifp); 4394 return (error); 4395 } 4396 4397 int 4398 if_multiaddr_count(if_t ifp, int max) 4399 { 4400 struct ifmultiaddr *ifma; 4401 int count; 4402 4403 count = 0; 4404 if_maddr_rlock(ifp); 4405 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4406 if (ifma->ifma_addr->sa_family != AF_LINK) 4407 continue; 4408 count++; 4409 if (count == max) 4410 break; 4411 } 4412 if_maddr_runlock(ifp); 4413 return (count); 4414 } 4415 4416 int 4417 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg) 4418 { 4419 struct ifmultiaddr *ifma; 4420 int cnt = 0; 4421 4422 if_maddr_rlock(ifp); 4423 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4424 cnt += filter(arg, ifma, cnt); 4425 if_maddr_runlock(ifp); 4426 return (cnt); 4427 } 4428 4429 struct mbuf * 4430 if_dequeue(if_t ifp) 4431 { 4432 struct mbuf *m; 4433 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4434 4435 return (m); 4436 } 4437 4438 int 4439 if_sendq_prepend(if_t ifp, struct mbuf *m) 4440 { 4441 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4442 return (0); 4443 } 4444 4445 int 4446 if_setifheaderlen(if_t ifp, int len) 4447 { 4448 ((struct ifnet *)ifp)->if_hdrlen = len; 4449 return (0); 4450 } 4451 4452 caddr_t 4453 if_getlladdr(if_t ifp) 4454 { 4455 return (IF_LLADDR((struct ifnet *)ifp)); 4456 } 4457 4458 void * 4459 if_gethandle(u_char type) 4460 { 4461 return (if_alloc(type)); 4462 } 4463 4464 void 4465 if_bpfmtap(if_t ifh, struct mbuf *m) 4466 { 4467 struct ifnet *ifp = (struct ifnet *)ifh; 4468 4469 BPF_MTAP(ifp, m); 4470 } 4471 4472 void 4473 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4474 { 4475 struct ifnet *ifp = (struct ifnet *)ifh; 4476 4477 ETHER_BPF_MTAP(ifp, m); 4478 } 4479 4480 void 4481 if_vlancap(if_t ifh) 4482 { 4483 struct ifnet *ifp = (struct ifnet *)ifh; 4484 VLAN_CAPABILITIES(ifp); 4485 } 4486 4487 int 4488 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4489 { 4490 4491 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax; 4492 return (0); 4493 } 4494 4495 int 4496 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4497 { 4498 4499 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4500 return (0); 4501 } 4502 4503 int 4504 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4505 { 4506 4507 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4508 return (0); 4509 } 4510 4511 u_int 4512 if_gethwtsomax(if_t ifp) 4513 { 4514 4515 return (((struct ifnet *)ifp)->if_hw_tsomax); 4516 } 4517 4518 u_int 4519 if_gethwtsomaxsegcount(if_t ifp) 4520 { 4521 4522 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount); 4523 } 4524 4525 u_int 4526 if_gethwtsomaxsegsize(if_t ifp) 4527 { 4528 4529 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize); 4530 } 4531 4532 void 4533 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4534 { 4535 ((struct ifnet *)ifp)->if_init = init_fn; 4536 } 4537 4538 void 4539 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4540 { 4541 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4542 } 4543 4544 void 4545 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4546 { 4547 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4548 } 4549 4550 void 4551 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4552 { 4553 ((struct ifnet *)ifp)->if_transmit = start_fn; 4554 } 4555 4556 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4557 { 4558 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4559 4560 } 4561 4562 void 4563 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4564 { 4565 4566 ifp->if_get_counter = fn; 4567 } 4568 4569 /* Revisit these - These are inline functions originally. */ 4570 int 4571 drbr_inuse_drv(if_t ifh, struct buf_ring *br) 4572 { 4573 return drbr_inuse(ifh, br); 4574 } 4575 4576 struct mbuf* 4577 drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 4578 { 4579 return drbr_dequeue(ifh, br); 4580 } 4581 4582 int 4583 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 4584 { 4585 return drbr_needs_enqueue(ifh, br); 4586 } 4587 4588 int 4589 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 4590 { 4591 return drbr_enqueue(ifh, br, m); 4592 4593 } 4594