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 ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 268 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 269 static void do_link_state_change(void *, int); 270 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 271 static int if_getgroupmembers(struct ifgroupreq *); 272 static void if_delgroups(struct ifnet *); 273 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 274 static int if_detach_internal(struct ifnet *, int, struct if_clone **); 275 #ifdef VIMAGE 276 static void if_vmove(struct ifnet *, struct vnet *); 277 #endif 278 279 #ifdef INET6 280 /* 281 * XXX: declare here to avoid to include many inet6 related files.. 282 * should be more generalized? 283 */ 284 extern void nd6_setmtu(struct ifnet *); 285 #endif 286 287 /* ipsec helper hooks */ 288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 289 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 290 291 VNET_DEFINE(int, if_index); 292 int ifqmaxlen = IFQ_MAXLEN; 293 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 294 VNET_DEFINE(struct ifgrouphead, ifg_head); 295 296 VNET_DEFINE_STATIC(int, if_indexlim) = 8; 297 298 /* Table of ifnet by index. */ 299 VNET_DEFINE(struct ifnet **, ifindex_table); 300 301 #define V_if_indexlim VNET(if_indexlim) 302 #define V_ifindex_table VNET(ifindex_table) 303 304 /* 305 * The global network interface list (V_ifnet) and related state (such as 306 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 307 * an rwlock. Either may be acquired shared to stablize the list, but both 308 * must be acquired writable to modify the list. This model allows us to 309 * both stablize the interface list during interrupt thread processing, but 310 * also to stablize it over long-running ioctls, without introducing priority 311 * inversions and deadlocks. 312 */ 313 struct rwlock ifnet_rwlock; 314 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE); 315 struct sx ifnet_sxlock; 316 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 317 318 /* 319 * The allocation of network interfaces is a rather non-atomic affair; we 320 * need to select an index before we are ready to expose the interface for 321 * use, so will use this pointer value to indicate reservation. 322 */ 323 #define IFNET_HOLD (void *)(uintptr_t)(-1) 324 325 static if_com_alloc_t *if_com_alloc[256]; 326 static if_com_free_t *if_com_free[256]; 327 328 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 329 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 330 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 331 332 struct ifnet * 333 ifnet_byindex_locked(u_short idx) 334 { 335 336 if (idx > V_if_index) 337 return (NULL); 338 if (V_ifindex_table[idx] == IFNET_HOLD) 339 return (NULL); 340 return (V_ifindex_table[idx]); 341 } 342 343 struct ifnet * 344 ifnet_byindex(u_short idx) 345 { 346 struct ifnet *ifp; 347 348 ifp = ifnet_byindex_locked(idx); 349 return (ifp); 350 } 351 352 struct ifnet * 353 ifnet_byindex_ref(u_short idx) 354 { 355 struct ifnet *ifp; 356 357 IFNET_RLOCK_NOSLEEP(); 358 ifp = ifnet_byindex_locked(idx); 359 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 360 IFNET_RUNLOCK_NOSLEEP(); 361 return (NULL); 362 } 363 if_ref(ifp); 364 IFNET_RUNLOCK_NOSLEEP(); 365 return (ifp); 366 } 367 368 /* 369 * Allocate an ifindex array entry; return 0 on success or an error on 370 * failure. 371 */ 372 static u_short 373 ifindex_alloc(void **old) 374 { 375 u_short idx; 376 377 IFNET_WLOCK_ASSERT(); 378 /* 379 * Try to find an empty slot below V_if_index. If we fail, take the 380 * next slot. 381 */ 382 for (idx = 1; idx <= V_if_index; idx++) { 383 if (V_ifindex_table[idx] == NULL) 384 break; 385 } 386 387 /* Catch if_index overflow. */ 388 if (idx >= V_if_indexlim) { 389 *old = if_grow(); 390 return (USHRT_MAX); 391 } 392 if (idx > V_if_index) 393 V_if_index = idx; 394 return (idx); 395 } 396 397 static void 398 ifindex_free_locked(u_short idx) 399 { 400 401 IFNET_WLOCK_ASSERT(); 402 403 V_ifindex_table[idx] = NULL; 404 while (V_if_index > 0 && 405 V_ifindex_table[V_if_index] == NULL) 406 V_if_index--; 407 } 408 409 static void 410 ifindex_free(u_short idx) 411 { 412 413 IFNET_WLOCK(); 414 ifindex_free_locked(idx); 415 IFNET_WUNLOCK(); 416 } 417 418 static void 419 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 420 { 421 422 V_ifindex_table[idx] = ifp; 423 } 424 425 struct ifaddr * 426 ifaddr_byindex(u_short idx) 427 { 428 struct ifnet *ifp; 429 struct ifaddr *ifa = NULL; 430 431 IFNET_RLOCK_NOSLEEP(); 432 ifp = ifnet_byindex_locked(idx); 433 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 434 ifa_ref(ifa); 435 IFNET_RUNLOCK_NOSLEEP(); 436 return (ifa); 437 } 438 439 /* 440 * Network interface utility routines. 441 * 442 * Routines with ifa_ifwith* names take sockaddr *'s as 443 * parameters. 444 */ 445 446 static void 447 vnet_if_init(const void *unused __unused) 448 { 449 void *old; 450 451 CK_STAILQ_INIT(&V_ifnet); 452 CK_STAILQ_INIT(&V_ifg_head); 453 IFNET_WLOCK(); 454 old = if_grow(); /* create initial table */ 455 IFNET_WUNLOCK(); 456 epoch_wait_preempt(net_epoch_preempt); 457 free(old, M_IFNET); 458 vnet_if_clone_init(); 459 } 460 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 461 NULL); 462 463 #ifdef VIMAGE 464 static void 465 vnet_if_uninit(const void *unused __unused) 466 { 467 468 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 469 "not empty", __func__, __LINE__, &V_ifnet)); 470 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 471 "not empty", __func__, __LINE__, &V_ifg_head)); 472 473 free((caddr_t)V_ifindex_table, M_IFNET); 474 } 475 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 476 vnet_if_uninit, NULL); 477 478 static void 479 vnet_if_return(const void *unused __unused) 480 { 481 struct ifnet *ifp, *nifp; 482 483 /* Return all inherited interfaces to their parent vnets. */ 484 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 485 if (ifp->if_home_vnet != ifp->if_vnet) 486 if_vmove(ifp, ifp->if_home_vnet); 487 } 488 } 489 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 490 vnet_if_return, NULL); 491 #endif 492 493 494 static void * 495 if_grow(void) 496 { 497 int oldlim; 498 u_int n; 499 struct ifnet **e; 500 void *old; 501 502 old = NULL; 503 IFNET_WLOCK_ASSERT(); 504 oldlim = V_if_indexlim; 505 IFNET_WUNLOCK(); 506 n = (oldlim << 1) * sizeof(*e); 507 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 508 IFNET_WLOCK(); 509 if (V_if_indexlim != oldlim) { 510 free(e, M_IFNET); 511 return (NULL); 512 } 513 if (V_ifindex_table != NULL) { 514 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 515 old = V_ifindex_table; 516 } 517 V_if_indexlim <<= 1; 518 V_ifindex_table = e; 519 return (old); 520 } 521 522 /* 523 * Allocate a struct ifnet and an index for an interface. A layer 2 524 * common structure will also be allocated if an allocation routine is 525 * registered for the passed type. 526 */ 527 struct ifnet * 528 if_alloc(u_char type) 529 { 530 struct ifnet *ifp; 531 u_short idx; 532 void *old; 533 534 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 535 restart: 536 IFNET_WLOCK(); 537 idx = ifindex_alloc(&old); 538 if (__predict_false(idx == USHRT_MAX)) { 539 IFNET_WUNLOCK(); 540 epoch_wait_preempt(net_epoch_preempt); 541 free(old, M_IFNET); 542 goto restart; 543 } 544 ifnet_setbyindex(idx, IFNET_HOLD); 545 IFNET_WUNLOCK(); 546 ifp->if_index = idx; 547 ifp->if_type = type; 548 ifp->if_alloctype = type; 549 #ifdef VIMAGE 550 ifp->if_vnet = curvnet; 551 #endif 552 if (if_com_alloc[type] != NULL) { 553 ifp->if_l2com = if_com_alloc[type](type, ifp); 554 if (ifp->if_l2com == NULL) { 555 free(ifp, M_IFNET); 556 ifindex_free(idx); 557 return (NULL); 558 } 559 } 560 561 IF_ADDR_LOCK_INIT(ifp); 562 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 563 ifp->if_afdata_initialized = 0; 564 IF_AFDATA_LOCK_INIT(ifp); 565 CK_STAILQ_INIT(&ifp->if_addrhead); 566 CK_STAILQ_INIT(&ifp->if_multiaddrs); 567 CK_STAILQ_INIT(&ifp->if_groups); 568 #ifdef MAC 569 mac_ifnet_init(ifp); 570 #endif 571 ifq_init(&ifp->if_snd, ifp); 572 573 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 574 for (int i = 0; i < IFCOUNTERS; i++) 575 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 576 ifp->if_get_counter = if_get_counter_default; 577 ifp->if_pcp = IFNET_PCP_NONE; 578 ifnet_setbyindex(ifp->if_index, ifp); 579 return (ifp); 580 } 581 582 /* 583 * Do the actual work of freeing a struct ifnet, and layer 2 common 584 * structure. This call is made when the last reference to an 585 * interface is released. 586 */ 587 static void 588 if_free_internal(struct ifnet *ifp) 589 { 590 591 KASSERT((ifp->if_flags & IFF_DYING), 592 ("if_free_internal: interface not dying")); 593 594 if (if_com_free[ifp->if_alloctype] != NULL) 595 if_com_free[ifp->if_alloctype](ifp->if_l2com, 596 ifp->if_alloctype); 597 598 #ifdef MAC 599 mac_ifnet_destroy(ifp); 600 #endif /* MAC */ 601 if (ifp->if_description != NULL) 602 free(ifp->if_description, M_IFDESCR); 603 IF_AFDATA_DESTROY(ifp); 604 IF_ADDR_LOCK_DESTROY(ifp); 605 ifq_delete(&ifp->if_snd); 606 607 for (int i = 0; i < IFCOUNTERS; i++) 608 counter_u64_free(ifp->if_counters[i]); 609 610 free(ifp, M_IFNET); 611 } 612 613 static void 614 if_destroy(epoch_context_t ctx) 615 { 616 struct ifnet *ifp; 617 618 ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 619 if_free_internal(ifp); 620 } 621 622 /* 623 * Deregister an interface and free the associated storage. 624 */ 625 void 626 if_free(struct ifnet *ifp) 627 { 628 629 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 630 631 CURVNET_SET_QUIET(ifp->if_vnet); 632 IFNET_WLOCK(); 633 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 634 ("%s: freeing unallocated ifnet", ifp->if_xname)); 635 636 ifindex_free_locked(ifp->if_index); 637 IFNET_WUNLOCK(); 638 639 if (refcount_release(&ifp->if_refcount)) 640 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy); 641 CURVNET_RESTORE(); 642 } 643 644 /* 645 * Interfaces to keep an ifnet type-stable despite the possibility of the 646 * driver calling if_free(). If there are additional references, we defer 647 * freeing the underlying data structure. 648 */ 649 void 650 if_ref(struct ifnet *ifp) 651 { 652 653 /* We don't assert the ifnet list lock here, but arguably should. */ 654 refcount_acquire(&ifp->if_refcount); 655 } 656 657 void 658 if_rele(struct ifnet *ifp) 659 { 660 661 if (!refcount_release(&ifp->if_refcount)) 662 return; 663 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy); 664 } 665 666 void 667 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 668 { 669 670 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 671 672 if (ifq->ifq_maxlen == 0) 673 ifq->ifq_maxlen = ifqmaxlen; 674 675 ifq->altq_type = 0; 676 ifq->altq_disc = NULL; 677 ifq->altq_flags &= ALTQF_CANTCHANGE; 678 ifq->altq_tbr = NULL; 679 ifq->altq_ifp = ifp; 680 } 681 682 void 683 ifq_delete(struct ifaltq *ifq) 684 { 685 mtx_destroy(&ifq->ifq_mtx); 686 } 687 688 /* 689 * Perform generic interface initialization tasks and attach the interface 690 * to the list of "active" interfaces. If vmove flag is set on entry 691 * to if_attach_internal(), perform only a limited subset of initialization 692 * tasks, given that we are moving from one vnet to another an ifnet which 693 * has already been fully initialized. 694 * 695 * Note that if_detach_internal() removes group membership unconditionally 696 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 697 * Thus, when if_vmove() is applied to a cloned interface, group membership 698 * is lost while a cloned one always joins a group whose name is 699 * ifc->ifc_name. To recover this after if_detach_internal() and 700 * if_attach_internal(), the cloner should be specified to 701 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 702 * attempts to join a group whose name is ifc->ifc_name. 703 * 704 * XXX: 705 * - The decision to return void and thus require this function to 706 * succeed is questionable. 707 * - We should probably do more sanity checking. For instance we don't 708 * do anything to insure if_xname is unique or non-empty. 709 */ 710 void 711 if_attach(struct ifnet *ifp) 712 { 713 714 if_attach_internal(ifp, 0, NULL); 715 } 716 717 /* 718 * Compute the least common TSO limit. 719 */ 720 void 721 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 722 { 723 /* 724 * 1) If there is no limit currently, take the limit from 725 * the network adapter. 726 * 727 * 2) If the network adapter has a limit below the current 728 * limit, apply it. 729 */ 730 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 731 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 732 pmax->tsomaxbytes = ifp->if_hw_tsomax; 733 } 734 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 735 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 736 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 737 } 738 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 739 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 740 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 741 } 742 } 743 744 /* 745 * Update TSO limit of a network adapter. 746 * 747 * Returns zero if no change. Else non-zero. 748 */ 749 int 750 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 751 { 752 int retval = 0; 753 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 754 ifp->if_hw_tsomax = pmax->tsomaxbytes; 755 retval++; 756 } 757 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 758 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 759 retval++; 760 } 761 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 762 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 763 retval++; 764 } 765 return (retval); 766 } 767 768 static void 769 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 770 { 771 unsigned socksize, ifasize; 772 int namelen, masklen; 773 struct sockaddr_dl *sdl; 774 struct ifaddr *ifa; 775 776 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 777 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 778 ifp->if_xname); 779 780 #ifdef VIMAGE 781 ifp->if_vnet = curvnet; 782 if (ifp->if_home_vnet == NULL) 783 ifp->if_home_vnet = curvnet; 784 #endif 785 786 if_addgroup(ifp, IFG_ALL); 787 788 /* Restore group membership for cloned interfaces. */ 789 if (vmove && ifc != NULL) 790 if_clone_addgroup(ifp, ifc); 791 792 getmicrotime(&ifp->if_lastchange); 793 ifp->if_epoch = time_uptime; 794 795 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 796 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 797 ("transmit and qflush must both either be set or both be NULL")); 798 if (ifp->if_transmit == NULL) { 799 ifp->if_transmit = if_transmit; 800 ifp->if_qflush = if_qflush; 801 } 802 if (ifp->if_input == NULL) 803 ifp->if_input = if_input_default; 804 805 if (ifp->if_requestencap == NULL) 806 ifp->if_requestencap = if_requestencap_default; 807 808 if (!vmove) { 809 #ifdef MAC 810 mac_ifnet_create(ifp); 811 #endif 812 813 /* 814 * Create a Link Level name for this device. 815 */ 816 namelen = strlen(ifp->if_xname); 817 /* 818 * Always save enough space for any possiable name so we 819 * can do a rename in place later. 820 */ 821 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 822 socksize = masklen + ifp->if_addrlen; 823 if (socksize < sizeof(*sdl)) 824 socksize = sizeof(*sdl); 825 socksize = roundup2(socksize, sizeof(long)); 826 ifasize = sizeof(*ifa) + 2 * socksize; 827 ifa = ifa_alloc(ifasize, M_WAITOK); 828 sdl = (struct sockaddr_dl *)(ifa + 1); 829 sdl->sdl_len = socksize; 830 sdl->sdl_family = AF_LINK; 831 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 832 sdl->sdl_nlen = namelen; 833 sdl->sdl_index = ifp->if_index; 834 sdl->sdl_type = ifp->if_type; 835 ifp->if_addr = ifa; 836 ifa->ifa_ifp = ifp; 837 ifa->ifa_rtrequest = link_rtrequest; 838 ifa->ifa_addr = (struct sockaddr *)sdl; 839 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 840 ifa->ifa_netmask = (struct sockaddr *)sdl; 841 sdl->sdl_len = masklen; 842 while (namelen != 0) 843 sdl->sdl_data[--namelen] = 0xff; 844 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 845 /* Reliably crash if used uninitialized. */ 846 ifp->if_broadcastaddr = NULL; 847 848 if (ifp->if_type == IFT_ETHER) { 849 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 850 M_WAITOK | M_ZERO); 851 } 852 853 #if defined(INET) || defined(INET6) 854 /* Use defaults for TSO, if nothing is set */ 855 if (ifp->if_hw_tsomax == 0 && 856 ifp->if_hw_tsomaxsegcount == 0 && 857 ifp->if_hw_tsomaxsegsize == 0) { 858 /* 859 * The TSO defaults needs to be such that an 860 * NFS mbuf list of 35 mbufs totalling just 861 * below 64K works and that a chain of mbufs 862 * can be defragged into at most 32 segments: 863 */ 864 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 865 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 866 ifp->if_hw_tsomaxsegcount = 35; 867 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 868 869 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 870 if (ifp->if_capabilities & IFCAP_TSO) { 871 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 872 ifp->if_hw_tsomax, 873 ifp->if_hw_tsomaxsegcount, 874 ifp->if_hw_tsomaxsegsize); 875 } 876 } 877 #endif 878 } 879 #ifdef VIMAGE 880 else { 881 /* 882 * Update the interface index in the link layer address 883 * of the interface. 884 */ 885 for (ifa = ifp->if_addr; ifa != NULL; 886 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 887 if (ifa->ifa_addr->sa_family == AF_LINK) { 888 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 889 sdl->sdl_index = ifp->if_index; 890 } 891 } 892 } 893 #endif 894 895 IFNET_WLOCK(); 896 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 897 #ifdef VIMAGE 898 curvnet->vnet_ifcnt++; 899 #endif 900 IFNET_WUNLOCK(); 901 902 if (domain_init_status >= 2) 903 if_attachdomain1(ifp); 904 905 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 906 if (IS_DEFAULT_VNET(curvnet)) 907 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 908 909 /* Announce the interface. */ 910 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 911 } 912 913 static void 914 if_epochalloc(void *dummy __unused) 915 { 916 917 net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT); 918 net_epoch = epoch_alloc(0); 919 } 920 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY, 921 if_epochalloc, NULL); 922 923 static void 924 if_attachdomain(void *dummy) 925 { 926 struct ifnet *ifp; 927 928 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 929 if_attachdomain1(ifp); 930 } 931 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 932 if_attachdomain, NULL); 933 934 static void 935 if_attachdomain1(struct ifnet *ifp) 936 { 937 struct domain *dp; 938 939 /* 940 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 941 * cannot lock ifp->if_afdata initialization, entirely. 942 */ 943 IF_AFDATA_LOCK(ifp); 944 if (ifp->if_afdata_initialized >= domain_init_status) { 945 IF_AFDATA_UNLOCK(ifp); 946 log(LOG_WARNING, "%s called more than once on %s\n", 947 __func__, ifp->if_xname); 948 return; 949 } 950 ifp->if_afdata_initialized = domain_init_status; 951 IF_AFDATA_UNLOCK(ifp); 952 953 /* address family dependent data region */ 954 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 955 for (dp = domains; dp; dp = dp->dom_next) { 956 if (dp->dom_ifattach) 957 ifp->if_afdata[dp->dom_family] = 958 (*dp->dom_ifattach)(ifp); 959 } 960 } 961 962 /* 963 * Remove any unicast or broadcast network addresses from an interface. 964 */ 965 void 966 if_purgeaddrs(struct ifnet *ifp) 967 { 968 struct ifaddr *ifa, *next; 969 970 NET_EPOCH_ENTER(); 971 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 972 if (ifa->ifa_addr->sa_family == AF_LINK) 973 continue; 974 #ifdef INET 975 /* XXX: Ugly!! ad hoc just for INET */ 976 if (ifa->ifa_addr->sa_family == AF_INET) { 977 struct ifaliasreq ifr; 978 979 bzero(&ifr, sizeof(ifr)); 980 ifr.ifra_addr = *ifa->ifa_addr; 981 if (ifa->ifa_dstaddr) 982 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 983 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 984 NULL) == 0) 985 continue; 986 } 987 #endif /* INET */ 988 #ifdef INET6 989 if (ifa->ifa_addr->sa_family == AF_INET6) { 990 in6_purgeaddr(ifa); 991 /* ifp_addrhead is already updated */ 992 continue; 993 } 994 #endif /* INET6 */ 995 IF_ADDR_WLOCK(ifp); 996 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 997 IF_ADDR_WUNLOCK(ifp); 998 ifa_free(ifa); 999 } 1000 NET_EPOCH_EXIT(); 1001 } 1002 1003 /* 1004 * Remove any multicast network addresses from an interface when an ifnet 1005 * is going away. 1006 */ 1007 static void 1008 if_purgemaddrs(struct ifnet *ifp) 1009 { 1010 struct ifmultiaddr *ifma; 1011 1012 IF_ADDR_WLOCK(ifp); 1013 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1014 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1015 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1016 if_delmulti_locked(ifp, ifma, 1); 1017 } 1018 IF_ADDR_WUNLOCK(ifp); 1019 } 1020 1021 /* 1022 * Detach an interface, removing it from the list of "active" interfaces. 1023 * If vmove flag is set on entry to if_detach_internal(), perform only a 1024 * limited subset of cleanup tasks, given that we are moving an ifnet from 1025 * one vnet to another, where it must be fully operational. 1026 * 1027 * XXXRW: There are some significant questions about event ordering, and 1028 * how to prevent things from starting to use the interface during detach. 1029 */ 1030 void 1031 if_detach(struct ifnet *ifp) 1032 { 1033 1034 CURVNET_SET_QUIET(ifp->if_vnet); 1035 if_detach_internal(ifp, 0, NULL); 1036 CURVNET_RESTORE(); 1037 } 1038 1039 /* 1040 * The vmove flag, if set, indicates that we are called from a callpath 1041 * that is moving an interface to a different vnet instance. 1042 * 1043 * The shutdown flag, if set, indicates that we are called in the 1044 * process of shutting down a vnet instance. Currently only the 1045 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1046 * on a vnet instance shutdown without this flag being set, e.g., when 1047 * the cloned interfaces are destoyed as first thing of teardown. 1048 */ 1049 static int 1050 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 1051 { 1052 struct ifaddr *ifa; 1053 int i; 1054 struct domain *dp; 1055 struct ifnet *iter; 1056 int found = 0; 1057 #ifdef VIMAGE 1058 int shutdown; 1059 1060 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1061 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1062 #endif 1063 IFNET_WLOCK(); 1064 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 1065 if (iter == ifp) { 1066 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 1067 found = 1; 1068 break; 1069 } 1070 IFNET_WUNLOCK(); 1071 if (!found) { 1072 /* 1073 * While we would want to panic here, we cannot 1074 * guarantee that the interface is indeed still on 1075 * the list given we don't hold locks all the way. 1076 */ 1077 return (ENOENT); 1078 #if 0 1079 if (vmove) 1080 panic("%s: ifp=%p not on the ifnet tailq %p", 1081 __func__, ifp, &V_ifnet); 1082 else 1083 return; /* XXX this should panic as well? */ 1084 #endif 1085 } 1086 1087 /* 1088 * At this point we know the interface still was on the ifnet list 1089 * and we removed it so we are in a stable state. 1090 */ 1091 #ifdef VIMAGE 1092 curvnet->vnet_ifcnt--; 1093 #endif 1094 epoch_wait_preempt(net_epoch_preempt); 1095 /* 1096 * In any case (destroy or vmove) detach us from the groups 1097 * and remove/wait for pending events on the taskq. 1098 * XXX-BZ in theory an interface could still enqueue a taskq change? 1099 */ 1100 if_delgroups(ifp); 1101 1102 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1103 1104 /* 1105 * Check if this is a cloned interface or not. Must do even if 1106 * shutting down as a if_vmove_reclaim() would move the ifp and 1107 * the if_clone_addgroup() will have a corrupted string overwise 1108 * from a gibberish pointer. 1109 */ 1110 if (vmove && ifcp != NULL) 1111 *ifcp = if_clone_findifc(ifp); 1112 1113 if_down(ifp); 1114 1115 #ifdef VIMAGE 1116 /* 1117 * On VNET shutdown abort here as the stack teardown will do all 1118 * the work top-down for us. 1119 */ 1120 if (shutdown) { 1121 /* 1122 * In case of a vmove we are done here without error. 1123 * If we would signal an error it would lead to the same 1124 * abort as if we did not find the ifnet anymore. 1125 * if_detach() calls us in void context and does not care 1126 * about an early abort notification, so life is splendid :) 1127 */ 1128 goto finish_vnet_shutdown; 1129 } 1130 #endif 1131 1132 /* 1133 * At this point we are not tearing down a VNET and are either 1134 * going to destroy or vmove the interface and have to cleanup 1135 * accordingly. 1136 */ 1137 1138 /* 1139 * Remove routes and flush queues. 1140 */ 1141 #ifdef ALTQ 1142 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1143 altq_disable(&ifp->if_snd); 1144 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1145 altq_detach(&ifp->if_snd); 1146 #endif 1147 1148 if_purgeaddrs(ifp); 1149 1150 #ifdef INET 1151 in_ifdetach(ifp); 1152 #endif 1153 1154 #ifdef INET6 1155 /* 1156 * Remove all IPv6 kernel structs related to ifp. This should be done 1157 * before removing routing entries below, since IPv6 interface direct 1158 * routes are expected to be removed by the IPv6-specific kernel API. 1159 * Otherwise, the kernel will detect some inconsistency and bark it. 1160 */ 1161 in6_ifdetach(ifp); 1162 #endif 1163 if_purgemaddrs(ifp); 1164 1165 /* Announce that the interface is gone. */ 1166 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1167 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1168 if (IS_DEFAULT_VNET(curvnet)) 1169 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1170 1171 if (!vmove) { 1172 /* 1173 * Prevent further calls into the device driver via ifnet. 1174 */ 1175 if_dead(ifp); 1176 1177 /* 1178 * Remove link ifaddr pointer and maybe decrement if_index. 1179 * Clean up all addresses. 1180 */ 1181 free(ifp->if_hw_addr, M_IFADDR); 1182 ifp->if_hw_addr = NULL; 1183 ifp->if_addr = NULL; 1184 1185 /* We can now free link ifaddr. */ 1186 IF_ADDR_WLOCK(ifp); 1187 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1188 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1189 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1190 IF_ADDR_WUNLOCK(ifp); 1191 ifa_free(ifa); 1192 } else 1193 IF_ADDR_WUNLOCK(ifp); 1194 } 1195 1196 rt_flushifroutes(ifp); 1197 1198 #ifdef VIMAGE 1199 finish_vnet_shutdown: 1200 #endif 1201 /* 1202 * We cannot hold the lock over dom_ifdetach calls as they might 1203 * sleep, for example trying to drain a callout, thus open up the 1204 * theoretical race with re-attaching. 1205 */ 1206 IF_AFDATA_LOCK(ifp); 1207 i = ifp->if_afdata_initialized; 1208 ifp->if_afdata_initialized = 0; 1209 IF_AFDATA_UNLOCK(ifp); 1210 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1211 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1212 (*dp->dom_ifdetach)(ifp, 1213 ifp->if_afdata[dp->dom_family]); 1214 ifp->if_afdata[dp->dom_family] = NULL; 1215 } 1216 } 1217 1218 return (0); 1219 } 1220 1221 #ifdef VIMAGE 1222 /* 1223 * if_vmove() performs a limited version of if_detach() in current 1224 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1225 * An attempt is made to shrink if_index in current vnet, find an 1226 * unused if_index in target vnet and calls if_grow() if necessary, 1227 * and finally find an unused if_xname for the target vnet. 1228 */ 1229 static void 1230 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1231 { 1232 struct if_clone *ifc; 1233 u_int bif_dlt, bif_hdrlen; 1234 void *old; 1235 int rc; 1236 1237 /* 1238 * if_detach_internal() will call the eventhandler to notify 1239 * interface departure. That will detach if_bpf. We need to 1240 * safe the dlt and hdrlen so we can re-attach it later. 1241 */ 1242 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1243 1244 /* 1245 * Detach from current vnet, but preserve LLADDR info, do not 1246 * mark as dead etc. so that the ifnet can be reattached later. 1247 * If we cannot find it, we lost the race to someone else. 1248 */ 1249 rc = if_detach_internal(ifp, 1, &ifc); 1250 if (rc != 0) 1251 return; 1252 1253 /* 1254 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1255 * the if_index for that vnet if possible. 1256 * 1257 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1258 * or we'd lock on one vnet and unlock on another. 1259 */ 1260 IFNET_WLOCK(); 1261 ifindex_free_locked(ifp->if_index); 1262 IFNET_WUNLOCK(); 1263 1264 /* 1265 * Perform interface-specific reassignment tasks, if provided by 1266 * the driver. 1267 */ 1268 if (ifp->if_reassign != NULL) 1269 ifp->if_reassign(ifp, new_vnet, NULL); 1270 1271 /* 1272 * Switch to the context of the target vnet. 1273 */ 1274 CURVNET_SET_QUIET(new_vnet); 1275 restart: 1276 IFNET_WLOCK(); 1277 ifp->if_index = ifindex_alloc(&old); 1278 if (__predict_false(ifp->if_index == USHRT_MAX)) { 1279 IFNET_WUNLOCK(); 1280 epoch_wait_preempt(net_epoch_preempt); 1281 free(old, M_IFNET); 1282 goto restart; 1283 } 1284 ifnet_setbyindex(ifp->if_index, ifp); 1285 IFNET_WUNLOCK(); 1286 1287 if_attach_internal(ifp, 1, ifc); 1288 1289 if (ifp->if_bpf == NULL) 1290 bpfattach(ifp, bif_dlt, bif_hdrlen); 1291 1292 CURVNET_RESTORE(); 1293 } 1294 1295 /* 1296 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1297 */ 1298 static int 1299 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1300 { 1301 struct prison *pr; 1302 struct ifnet *difp; 1303 int shutdown; 1304 1305 /* Try to find the prison within our visibility. */ 1306 sx_slock(&allprison_lock); 1307 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1308 sx_sunlock(&allprison_lock); 1309 if (pr == NULL) 1310 return (ENXIO); 1311 prison_hold_locked(pr); 1312 mtx_unlock(&pr->pr_mtx); 1313 1314 /* Do not try to move the iface from and to the same prison. */ 1315 if (pr->pr_vnet == ifp->if_vnet) { 1316 prison_free(pr); 1317 return (EEXIST); 1318 } 1319 1320 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1321 /* XXX Lock interfaces to avoid races. */ 1322 CURVNET_SET_QUIET(pr->pr_vnet); 1323 difp = ifunit(ifname); 1324 if (difp != NULL) { 1325 CURVNET_RESTORE(); 1326 prison_free(pr); 1327 return (EEXIST); 1328 } 1329 1330 /* Make sure the VNET is stable. */ 1331 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1332 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1333 if (shutdown) { 1334 CURVNET_RESTORE(); 1335 prison_free(pr); 1336 return (EBUSY); 1337 } 1338 CURVNET_RESTORE(); 1339 1340 /* Move the interface into the child jail/vnet. */ 1341 if_vmove(ifp, pr->pr_vnet); 1342 1343 /* Report the new if_xname back to the userland. */ 1344 sprintf(ifname, "%s", ifp->if_xname); 1345 1346 prison_free(pr); 1347 return (0); 1348 } 1349 1350 static int 1351 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1352 { 1353 struct prison *pr; 1354 struct vnet *vnet_dst; 1355 struct ifnet *ifp; 1356 int shutdown; 1357 1358 /* Try to find the prison within our visibility. */ 1359 sx_slock(&allprison_lock); 1360 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1361 sx_sunlock(&allprison_lock); 1362 if (pr == NULL) 1363 return (ENXIO); 1364 prison_hold_locked(pr); 1365 mtx_unlock(&pr->pr_mtx); 1366 1367 /* Make sure the named iface exists in the source prison/vnet. */ 1368 CURVNET_SET(pr->pr_vnet); 1369 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1370 if (ifp == NULL) { 1371 CURVNET_RESTORE(); 1372 prison_free(pr); 1373 return (ENXIO); 1374 } 1375 1376 /* Do not try to move the iface from and to the same prison. */ 1377 vnet_dst = TD_TO_VNET(td); 1378 if (vnet_dst == ifp->if_vnet) { 1379 CURVNET_RESTORE(); 1380 prison_free(pr); 1381 return (EEXIST); 1382 } 1383 1384 /* Make sure the VNET is stable. */ 1385 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1386 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1387 if (shutdown) { 1388 CURVNET_RESTORE(); 1389 prison_free(pr); 1390 return (EBUSY); 1391 } 1392 1393 /* Get interface back from child jail/vnet. */ 1394 if_vmove(ifp, vnet_dst); 1395 CURVNET_RESTORE(); 1396 1397 /* Report the new if_xname back to the userland. */ 1398 sprintf(ifname, "%s", ifp->if_xname); 1399 1400 prison_free(pr); 1401 return (0); 1402 } 1403 #endif /* VIMAGE */ 1404 1405 /* 1406 * Add a group to an interface 1407 */ 1408 int 1409 if_addgroup(struct ifnet *ifp, const char *groupname) 1410 { 1411 struct ifg_list *ifgl; 1412 struct ifg_group *ifg = NULL; 1413 struct ifg_member *ifgm; 1414 int new = 0; 1415 1416 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1417 groupname[strlen(groupname) - 1] <= '9') 1418 return (EINVAL); 1419 1420 IFNET_WLOCK(); 1421 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1422 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1423 IFNET_WUNLOCK(); 1424 return (EEXIST); 1425 } 1426 1427 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1428 M_NOWAIT)) == NULL) { 1429 IFNET_WUNLOCK(); 1430 return (ENOMEM); 1431 } 1432 1433 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1434 M_TEMP, M_NOWAIT)) == NULL) { 1435 free(ifgl, M_TEMP); 1436 IFNET_WUNLOCK(); 1437 return (ENOMEM); 1438 } 1439 1440 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1441 if (!strcmp(ifg->ifg_group, groupname)) 1442 break; 1443 1444 if (ifg == NULL) { 1445 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1446 M_TEMP, M_NOWAIT)) == NULL) { 1447 free(ifgl, M_TEMP); 1448 free(ifgm, M_TEMP); 1449 IFNET_WUNLOCK(); 1450 return (ENOMEM); 1451 } 1452 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1453 ifg->ifg_refcnt = 0; 1454 CK_STAILQ_INIT(&ifg->ifg_members); 1455 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1456 new = 1; 1457 } 1458 1459 ifg->ifg_refcnt++; 1460 ifgl->ifgl_group = ifg; 1461 ifgm->ifgm_ifp = ifp; 1462 1463 IF_ADDR_WLOCK(ifp); 1464 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1465 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1466 IF_ADDR_WUNLOCK(ifp); 1467 1468 IFNET_WUNLOCK(); 1469 1470 if (new) 1471 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1472 EVENTHANDLER_INVOKE(group_change_event, groupname); 1473 1474 return (0); 1475 } 1476 1477 /* 1478 * Remove a group from an interface 1479 */ 1480 int 1481 if_delgroup(struct ifnet *ifp, const char *groupname) 1482 { 1483 struct ifg_list *ifgl; 1484 struct ifg_member *ifgm; 1485 int freeifgl; 1486 1487 IFNET_WLOCK(); 1488 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1489 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1490 break; 1491 if (ifgl == NULL) { 1492 IFNET_WUNLOCK(); 1493 return (ENOENT); 1494 } 1495 1496 freeifgl = 0; 1497 IF_ADDR_WLOCK(ifp); 1498 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1499 IF_ADDR_WUNLOCK(ifp); 1500 1501 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1502 if (ifgm->ifgm_ifp == ifp) 1503 break; 1504 1505 if (ifgm != NULL) 1506 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next); 1507 1508 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1509 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next); 1510 freeifgl = 1; 1511 } 1512 IFNET_WUNLOCK(); 1513 1514 epoch_wait_preempt(net_epoch_preempt); 1515 if (freeifgl) { 1516 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1517 free(ifgl->ifgl_group, M_TEMP); 1518 } 1519 free(ifgm, M_TEMP); 1520 free(ifgl, M_TEMP); 1521 1522 EVENTHANDLER_INVOKE(group_change_event, groupname); 1523 1524 return (0); 1525 } 1526 1527 /* 1528 * Remove an interface from all groups 1529 */ 1530 static void 1531 if_delgroups(struct ifnet *ifp) 1532 { 1533 struct ifg_list *ifgl; 1534 struct ifg_member *ifgm; 1535 char groupname[IFNAMSIZ]; 1536 int ifglfree; 1537 1538 IFNET_WLOCK(); 1539 while (!CK_STAILQ_EMPTY(&ifp->if_groups)) { 1540 ifgl = CK_STAILQ_FIRST(&ifp->if_groups); 1541 1542 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1543 1544 IF_ADDR_WLOCK(ifp); 1545 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1546 IF_ADDR_WUNLOCK(ifp); 1547 1548 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1549 if (ifgm->ifgm_ifp == ifp) 1550 break; 1551 1552 if (ifgm != NULL) 1553 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, 1554 ifgm_next); 1555 ifglfree = 0; 1556 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1557 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next); 1558 ifglfree = 1; 1559 } 1560 1561 IFNET_WUNLOCK(); 1562 epoch_wait_preempt(net_epoch_preempt); 1563 free(ifgm, M_TEMP); 1564 if (ifglfree) { 1565 EVENTHANDLER_INVOKE(group_detach_event, 1566 ifgl->ifgl_group); 1567 free(ifgl->ifgl_group, M_TEMP); 1568 } 1569 EVENTHANDLER_INVOKE(group_change_event, groupname); 1570 1571 IFNET_WLOCK(); 1572 } 1573 IFNET_WUNLOCK(); 1574 } 1575 1576 static char * 1577 ifgr_group_get(void *ifgrp) 1578 { 1579 union ifgroupreq_union *ifgrup; 1580 1581 ifgrup = ifgrp; 1582 #ifdef COMPAT_FREEBSD32 1583 if (SV_CURPROC_FLAG(SV_ILP32)) 1584 return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]); 1585 #endif 1586 return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]); 1587 } 1588 1589 static struct ifg_req * 1590 ifgr_groups_get(void *ifgrp) 1591 { 1592 union ifgroupreq_union *ifgrup; 1593 1594 ifgrup = ifgrp; 1595 #ifdef COMPAT_FREEBSD32 1596 if (SV_CURPROC_FLAG(SV_ILP32)) 1597 return ((struct ifg_req *)(uintptr_t) 1598 ifgrup->ifgr32.ifgr_ifgru.ifgru_groups); 1599 #endif 1600 return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups); 1601 } 1602 1603 /* 1604 * Stores all groups from an interface in memory pointed to by ifgr. 1605 */ 1606 static int 1607 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1608 { 1609 int len, error; 1610 struct ifg_list *ifgl; 1611 struct ifg_req ifgrq, *ifgp; 1612 1613 if (ifgr->ifgr_len == 0) { 1614 IF_ADDR_RLOCK(ifp); 1615 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1616 ifgr->ifgr_len += sizeof(struct ifg_req); 1617 IF_ADDR_RUNLOCK(ifp); 1618 return (0); 1619 } 1620 1621 len = ifgr->ifgr_len; 1622 ifgp = ifgr_groups_get(ifgr); 1623 /* XXX: wire */ 1624 IF_ADDR_RLOCK(ifp); 1625 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1626 if (len < sizeof(ifgrq)) { 1627 IF_ADDR_RUNLOCK(ifp); 1628 return (EINVAL); 1629 } 1630 bzero(&ifgrq, sizeof ifgrq); 1631 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1632 sizeof(ifgrq.ifgrq_group)); 1633 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1634 IF_ADDR_RUNLOCK(ifp); 1635 return (error); 1636 } 1637 len -= sizeof(ifgrq); 1638 ifgp++; 1639 } 1640 IF_ADDR_RUNLOCK(ifp); 1641 1642 return (0); 1643 } 1644 1645 /* 1646 * Stores all members of a group in memory pointed to by igfr 1647 */ 1648 static int 1649 if_getgroupmembers(struct ifgroupreq *ifgr) 1650 { 1651 struct ifg_group *ifg; 1652 struct ifg_member *ifgm; 1653 struct ifg_req ifgrq, *ifgp; 1654 int len, error; 1655 1656 IFNET_RLOCK(); 1657 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1658 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1659 break; 1660 if (ifg == NULL) { 1661 IFNET_RUNLOCK(); 1662 return (ENOENT); 1663 } 1664 1665 if (ifgr->ifgr_len == 0) { 1666 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1667 ifgr->ifgr_len += sizeof(ifgrq); 1668 IFNET_RUNLOCK(); 1669 return (0); 1670 } 1671 1672 len = ifgr->ifgr_len; 1673 ifgp = ifgr_groups_get(ifgr); 1674 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1675 if (len < sizeof(ifgrq)) { 1676 IFNET_RUNLOCK(); 1677 return (EINVAL); 1678 } 1679 bzero(&ifgrq, sizeof ifgrq); 1680 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1681 sizeof(ifgrq.ifgrq_member)); 1682 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1683 IFNET_RUNLOCK(); 1684 return (error); 1685 } 1686 len -= sizeof(ifgrq); 1687 ifgp++; 1688 } 1689 IFNET_RUNLOCK(); 1690 1691 return (0); 1692 } 1693 1694 /* 1695 * Return counter values from counter(9)s stored in ifnet. 1696 */ 1697 uint64_t 1698 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1699 { 1700 1701 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1702 1703 return (counter_u64_fetch(ifp->if_counters[cnt])); 1704 } 1705 1706 /* 1707 * Increase an ifnet counter. Usually used for counters shared 1708 * between the stack and a driver, but function supports them all. 1709 */ 1710 void 1711 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1712 { 1713 1714 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1715 1716 counter_u64_add(ifp->if_counters[cnt], inc); 1717 } 1718 1719 /* 1720 * Copy data from ifnet to userland API structure if_data. 1721 */ 1722 void 1723 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1724 { 1725 1726 ifd->ifi_type = ifp->if_type; 1727 ifd->ifi_physical = 0; 1728 ifd->ifi_addrlen = ifp->if_addrlen; 1729 ifd->ifi_hdrlen = ifp->if_hdrlen; 1730 ifd->ifi_link_state = ifp->if_link_state; 1731 ifd->ifi_vhid = 0; 1732 ifd->ifi_datalen = sizeof(struct if_data); 1733 ifd->ifi_mtu = ifp->if_mtu; 1734 ifd->ifi_metric = ifp->if_metric; 1735 ifd->ifi_baudrate = ifp->if_baudrate; 1736 ifd->ifi_hwassist = ifp->if_hwassist; 1737 ifd->ifi_epoch = ifp->if_epoch; 1738 ifd->ifi_lastchange = ifp->if_lastchange; 1739 1740 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1741 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1742 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1743 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1744 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1745 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1746 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1747 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1748 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1749 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1750 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1751 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1752 } 1753 1754 /* 1755 * Wrapper functions for struct ifnet address list locking macros. These are 1756 * used by kernel modules to avoid encoding programming interface or binary 1757 * interface assumptions that may be violated when kernel-internal locking 1758 * approaches change. 1759 */ 1760 void 1761 if_addr_rlock(struct ifnet *ifp) 1762 { 1763 MPASS(*(uint64_t *)&ifp->if_addr_et == 0); 1764 epoch_enter_preempt(net_epoch_preempt, &ifp->if_addr_et); 1765 } 1766 1767 void 1768 if_addr_runlock(struct ifnet *ifp) 1769 { 1770 epoch_exit_preempt(net_epoch_preempt, &ifp->if_addr_et); 1771 #ifdef INVARIANTS 1772 bzero(&ifp->if_addr_et, sizeof(struct epoch_tracker)); 1773 #endif 1774 } 1775 1776 void 1777 if_maddr_rlock(if_t ifp) 1778 { 1779 1780 MPASS(*(uint64_t *)&ifp->if_maddr_et == 0); 1781 epoch_enter_preempt(net_epoch_preempt, &ifp->if_maddr_et); 1782 } 1783 1784 void 1785 if_maddr_runlock(if_t ifp) 1786 { 1787 1788 epoch_exit_preempt(net_epoch_preempt, &ifp->if_maddr_et); 1789 #ifdef INVARIANTS 1790 bzero(&ifp->if_maddr_et, sizeof(struct epoch_tracker)); 1791 #endif 1792 } 1793 1794 /* 1795 * Initialization, destruction and refcounting functions for ifaddrs. 1796 */ 1797 struct ifaddr * 1798 ifa_alloc(size_t size, int flags) 1799 { 1800 struct ifaddr *ifa; 1801 1802 KASSERT(size >= sizeof(struct ifaddr), 1803 ("%s: invalid size %zu", __func__, size)); 1804 1805 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1806 if (ifa == NULL) 1807 return (NULL); 1808 1809 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1810 goto fail; 1811 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1812 goto fail; 1813 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1814 goto fail; 1815 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1816 goto fail; 1817 1818 refcount_init(&ifa->ifa_refcnt, 1); 1819 1820 return (ifa); 1821 1822 fail: 1823 /* free(NULL) is okay */ 1824 counter_u64_free(ifa->ifa_opackets); 1825 counter_u64_free(ifa->ifa_ipackets); 1826 counter_u64_free(ifa->ifa_obytes); 1827 counter_u64_free(ifa->ifa_ibytes); 1828 free(ifa, M_IFADDR); 1829 1830 return (NULL); 1831 } 1832 1833 void 1834 ifa_ref(struct ifaddr *ifa) 1835 { 1836 1837 refcount_acquire(&ifa->ifa_refcnt); 1838 } 1839 1840 static void 1841 ifa_destroy(epoch_context_t ctx) 1842 { 1843 struct ifaddr *ifa; 1844 1845 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1846 counter_u64_free(ifa->ifa_opackets); 1847 counter_u64_free(ifa->ifa_ipackets); 1848 counter_u64_free(ifa->ifa_obytes); 1849 counter_u64_free(ifa->ifa_ibytes); 1850 free(ifa, M_IFADDR); 1851 } 1852 1853 void 1854 ifa_free(struct ifaddr *ifa) 1855 { 1856 1857 if (refcount_release(&ifa->ifa_refcnt)) 1858 epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy); 1859 } 1860 1861 1862 static int 1863 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa, 1864 struct sockaddr *ia) 1865 { 1866 int error; 1867 struct rt_addrinfo info; 1868 struct sockaddr_dl null_sdl; 1869 struct ifnet *ifp; 1870 1871 ifp = ifa->ifa_ifp; 1872 1873 bzero(&info, sizeof(info)); 1874 if (cmd != RTM_DELETE) 1875 info.rti_ifp = V_loif; 1876 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED; 1877 info.rti_info[RTAX_DST] = ia; 1878 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1879 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type); 1880 1881 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib); 1882 1883 if (error != 0 && 1884 !(cmd == RTM_ADD && error == EEXIST) && 1885 !(cmd == RTM_DELETE && error == ENOENT)) 1886 if_printf(ifp, "%s failed: %d\n", otype, error); 1887 1888 return (error); 1889 } 1890 1891 int 1892 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1893 { 1894 1895 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia)); 1896 } 1897 1898 int 1899 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1900 { 1901 1902 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia)); 1903 } 1904 1905 int 1906 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1907 { 1908 1909 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia)); 1910 } 1911 1912 /* 1913 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1914 * structs used to represent other address families, it is necessary 1915 * to perform a different comparison. 1916 */ 1917 1918 #define sa_dl_equal(a1, a2) \ 1919 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1920 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1921 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1922 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1923 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1924 1925 /* 1926 * Locate an interface based on a complete address. 1927 */ 1928 /*ARGSUSED*/ 1929 struct ifaddr * 1930 ifa_ifwithaddr(const struct sockaddr *addr) 1931 { 1932 struct ifnet *ifp; 1933 struct ifaddr *ifa; 1934 1935 MPASS(in_epoch(net_epoch_preempt)); 1936 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1937 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1938 if (ifa->ifa_addr->sa_family != addr->sa_family) 1939 continue; 1940 if (sa_equal(addr, ifa->ifa_addr)) { 1941 goto done; 1942 } 1943 /* IP6 doesn't have broadcast */ 1944 if ((ifp->if_flags & IFF_BROADCAST) && 1945 ifa->ifa_broadaddr && 1946 ifa->ifa_broadaddr->sa_len != 0 && 1947 sa_equal(ifa->ifa_broadaddr, addr)) { 1948 goto done; 1949 } 1950 } 1951 } 1952 ifa = NULL; 1953 done: 1954 return (ifa); 1955 } 1956 1957 int 1958 ifa_ifwithaddr_check(const struct sockaddr *addr) 1959 { 1960 int rc; 1961 1962 NET_EPOCH_ENTER(); 1963 rc = (ifa_ifwithaddr(addr) != NULL); 1964 NET_EPOCH_EXIT(); 1965 return (rc); 1966 } 1967 1968 /* 1969 * Locate an interface based on the broadcast address. 1970 */ 1971 /* ARGSUSED */ 1972 struct ifaddr * 1973 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1974 { 1975 struct ifnet *ifp; 1976 struct ifaddr *ifa; 1977 1978 MPASS(in_epoch(net_epoch_preempt)); 1979 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1980 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1981 continue; 1982 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1983 if (ifa->ifa_addr->sa_family != addr->sa_family) 1984 continue; 1985 if ((ifp->if_flags & IFF_BROADCAST) && 1986 ifa->ifa_broadaddr && 1987 ifa->ifa_broadaddr->sa_len != 0 && 1988 sa_equal(ifa->ifa_broadaddr, addr)) { 1989 goto done; 1990 } 1991 } 1992 } 1993 ifa = NULL; 1994 done: 1995 return (ifa); 1996 } 1997 1998 /* 1999 * Locate the point to point interface with a given destination address. 2000 */ 2001 /*ARGSUSED*/ 2002 struct ifaddr * 2003 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 2004 { 2005 struct ifnet *ifp; 2006 struct ifaddr *ifa; 2007 2008 MPASS(in_epoch(net_epoch_preempt)); 2009 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2010 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 2011 continue; 2012 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2013 continue; 2014 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2015 if (ifa->ifa_addr->sa_family != addr->sa_family) 2016 continue; 2017 if (ifa->ifa_dstaddr != NULL && 2018 sa_equal(addr, ifa->ifa_dstaddr)) { 2019 goto done; 2020 } 2021 } 2022 } 2023 ifa = NULL; 2024 done: 2025 return (ifa); 2026 } 2027 2028 /* 2029 * Find an interface on a specific network. If many, choice 2030 * is most specific found. 2031 */ 2032 struct ifaddr * 2033 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 2034 { 2035 struct ifnet *ifp; 2036 struct ifaddr *ifa; 2037 struct ifaddr *ifa_maybe = NULL; 2038 u_int af = addr->sa_family; 2039 const char *addr_data = addr->sa_data, *cplim; 2040 2041 MPASS(in_epoch(net_epoch_preempt)); 2042 /* 2043 * AF_LINK addresses can be looked up directly by their index number, 2044 * so do that if we can. 2045 */ 2046 if (af == AF_LINK) { 2047 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 2048 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 2049 return (ifaddr_byindex(sdl->sdl_index)); 2050 } 2051 2052 /* 2053 * Scan though each interface, looking for ones that have addresses 2054 * in this address family and the requested fib. Maintain a reference 2055 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 2056 * kept it stable when we move onto the next interface. 2057 */ 2058 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2059 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2060 continue; 2061 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2062 const char *cp, *cp2, *cp3; 2063 2064 if (ifa->ifa_addr->sa_family != af) 2065 next: continue; 2066 if (af == AF_INET && 2067 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 2068 /* 2069 * This is a bit broken as it doesn't 2070 * take into account that the remote end may 2071 * be a single node in the network we are 2072 * looking for. 2073 * The trouble is that we don't know the 2074 * netmask for the remote end. 2075 */ 2076 if (ifa->ifa_dstaddr != NULL && 2077 sa_equal(addr, ifa->ifa_dstaddr)) { 2078 goto done; 2079 } 2080 } else { 2081 /* 2082 * Scan all the bits in the ifa's address. 2083 * If a bit dissagrees with what we are 2084 * looking for, mask it with the netmask 2085 * to see if it really matters. 2086 * (A byte at a time) 2087 */ 2088 if (ifa->ifa_netmask == 0) 2089 continue; 2090 cp = addr_data; 2091 cp2 = ifa->ifa_addr->sa_data; 2092 cp3 = ifa->ifa_netmask->sa_data; 2093 cplim = ifa->ifa_netmask->sa_len 2094 + (char *)ifa->ifa_netmask; 2095 while (cp3 < cplim) 2096 if ((*cp++ ^ *cp2++) & *cp3++) 2097 goto next; /* next address! */ 2098 /* 2099 * If the netmask of what we just found 2100 * is more specific than what we had before 2101 * (if we had one), or if the virtual status 2102 * of new prefix is better than of the old one, 2103 * then remember the new one before continuing 2104 * to search for an even better one. 2105 */ 2106 if (ifa_maybe == NULL || 2107 ifa_preferred(ifa_maybe, ifa) || 2108 rn_refines((caddr_t)ifa->ifa_netmask, 2109 (caddr_t)ifa_maybe->ifa_netmask)) { 2110 ifa_maybe = ifa; 2111 } 2112 } 2113 } 2114 } 2115 ifa = ifa_maybe; 2116 ifa_maybe = NULL; 2117 done: 2118 return (ifa); 2119 } 2120 2121 /* 2122 * Find an interface address specific to an interface best matching 2123 * a given address. 2124 */ 2125 struct ifaddr * 2126 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2127 { 2128 struct ifaddr *ifa; 2129 const char *cp, *cp2, *cp3; 2130 char *cplim; 2131 struct ifaddr *ifa_maybe = NULL; 2132 u_int af = addr->sa_family; 2133 2134 if (af >= AF_MAX) 2135 return (NULL); 2136 2137 MPASS(in_epoch(net_epoch_preempt)); 2138 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2139 if (ifa->ifa_addr->sa_family != af) 2140 continue; 2141 if (ifa_maybe == NULL) 2142 ifa_maybe = ifa; 2143 if (ifa->ifa_netmask == 0) { 2144 if (sa_equal(addr, ifa->ifa_addr) || 2145 (ifa->ifa_dstaddr && 2146 sa_equal(addr, ifa->ifa_dstaddr))) 2147 goto done; 2148 continue; 2149 } 2150 if (ifp->if_flags & IFF_POINTOPOINT) { 2151 if (sa_equal(addr, ifa->ifa_dstaddr)) 2152 goto done; 2153 } else { 2154 cp = addr->sa_data; 2155 cp2 = ifa->ifa_addr->sa_data; 2156 cp3 = ifa->ifa_netmask->sa_data; 2157 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2158 for (; cp3 < cplim; cp3++) 2159 if ((*cp++ ^ *cp2++) & *cp3) 2160 break; 2161 if (cp3 == cplim) 2162 goto done; 2163 } 2164 } 2165 ifa = ifa_maybe; 2166 done: 2167 return (ifa); 2168 } 2169 2170 /* 2171 * See whether new ifa is better than current one: 2172 * 1) A non-virtual one is preferred over virtual. 2173 * 2) A virtual in master state preferred over any other state. 2174 * 2175 * Used in several address selecting functions. 2176 */ 2177 int 2178 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2179 { 2180 2181 return (cur->ifa_carp && (!next->ifa_carp || 2182 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2183 } 2184 2185 #include <net/if_llatbl.h> 2186 2187 /* 2188 * Default action when installing a route with a Link Level gateway. 2189 * Lookup an appropriate real ifa to point to. 2190 * This should be moved to /sys/net/link.c eventually. 2191 */ 2192 static void 2193 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 2194 { 2195 struct ifaddr *ifa, *oifa; 2196 struct sockaddr *dst; 2197 struct ifnet *ifp; 2198 2199 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) || 2200 ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL)) 2201 return; 2202 NET_EPOCH_ENTER(); 2203 ifa = ifaof_ifpforaddr(dst, ifp); 2204 if (ifa) { 2205 oifa = rt->rt_ifa; 2206 if (oifa != ifa) { 2207 ifa_free(oifa); 2208 ifa_ref(ifa); 2209 } 2210 rt->rt_ifa = ifa; 2211 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2212 ifa->ifa_rtrequest(cmd, rt, info); 2213 } 2214 NET_EPOCH_EXIT(); 2215 } 2216 2217 struct sockaddr_dl * 2218 link_alloc_sdl(size_t size, int flags) 2219 { 2220 2221 return (malloc(size, M_TEMP, flags)); 2222 } 2223 2224 void 2225 link_free_sdl(struct sockaddr *sa) 2226 { 2227 free(sa, M_TEMP); 2228 } 2229 2230 /* 2231 * Fills in given sdl with interface basic info. 2232 * Returns pointer to filled sdl. 2233 */ 2234 struct sockaddr_dl * 2235 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2236 { 2237 struct sockaddr_dl *sdl; 2238 2239 sdl = (struct sockaddr_dl *)paddr; 2240 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2241 sdl->sdl_len = sizeof(struct sockaddr_dl); 2242 sdl->sdl_family = AF_LINK; 2243 sdl->sdl_index = ifp->if_index; 2244 sdl->sdl_type = iftype; 2245 2246 return (sdl); 2247 } 2248 2249 /* 2250 * Mark an interface down and notify protocols of 2251 * the transition. 2252 */ 2253 static void 2254 if_unroute(struct ifnet *ifp, int flag, int fam) 2255 { 2256 struct ifaddr *ifa; 2257 2258 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2259 2260 ifp->if_flags &= ~flag; 2261 getmicrotime(&ifp->if_lastchange); 2262 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2263 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2264 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2265 ifp->if_qflush(ifp); 2266 2267 if (ifp->if_carp) 2268 (*carp_linkstate_p)(ifp); 2269 rt_ifmsg(ifp); 2270 } 2271 2272 /* 2273 * Mark an interface up and notify protocols of 2274 * the transition. 2275 */ 2276 static void 2277 if_route(struct ifnet *ifp, int flag, int fam) 2278 { 2279 struct ifaddr *ifa; 2280 2281 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2282 2283 ifp->if_flags |= flag; 2284 getmicrotime(&ifp->if_lastchange); 2285 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2286 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2287 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2288 if (ifp->if_carp) 2289 (*carp_linkstate_p)(ifp); 2290 rt_ifmsg(ifp); 2291 #ifdef INET6 2292 in6_if_up(ifp); 2293 #endif 2294 } 2295 2296 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2297 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2298 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2299 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2300 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2301 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2302 int (*vlan_setcookie_p)(struct ifnet *, void *); 2303 void *(*vlan_cookie_p)(struct ifnet *); 2304 2305 /* 2306 * Handle a change in the interface link state. To avoid LORs 2307 * between driver lock and upper layer locks, as well as possible 2308 * recursions, we post event to taskqueue, and all job 2309 * is done in static do_link_state_change(). 2310 */ 2311 void 2312 if_link_state_change(struct ifnet *ifp, int link_state) 2313 { 2314 /* Return if state hasn't changed. */ 2315 if (ifp->if_link_state == link_state) 2316 return; 2317 2318 ifp->if_link_state = link_state; 2319 2320 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2321 } 2322 2323 static void 2324 do_link_state_change(void *arg, int pending) 2325 { 2326 struct ifnet *ifp = (struct ifnet *)arg; 2327 int link_state = ifp->if_link_state; 2328 CURVNET_SET(ifp->if_vnet); 2329 2330 /* Notify that the link state has changed. */ 2331 rt_ifmsg(ifp); 2332 if (ifp->if_vlantrunk != NULL) 2333 (*vlan_link_state_p)(ifp); 2334 2335 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2336 ifp->if_l2com != NULL) 2337 (*ng_ether_link_state_p)(ifp, link_state); 2338 if (ifp->if_carp) 2339 (*carp_linkstate_p)(ifp); 2340 if (ifp->if_bridge) 2341 ifp->if_bridge_linkstate(ifp); 2342 if (ifp->if_lagg) 2343 (*lagg_linkstate_p)(ifp, link_state); 2344 2345 if (IS_DEFAULT_VNET(curvnet)) 2346 devctl_notify("IFNET", ifp->if_xname, 2347 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2348 NULL); 2349 if (pending > 1) 2350 if_printf(ifp, "%d link states coalesced\n", pending); 2351 if (log_link_state_change) 2352 if_printf(ifp, "link state changed to %s\n", 2353 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2354 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2355 CURVNET_RESTORE(); 2356 } 2357 2358 /* 2359 * Mark an interface down and notify protocols of 2360 * the transition. 2361 */ 2362 void 2363 if_down(struct ifnet *ifp) 2364 { 2365 2366 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2367 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2368 } 2369 2370 /* 2371 * Mark an interface up and notify protocols of 2372 * the transition. 2373 */ 2374 void 2375 if_up(struct ifnet *ifp) 2376 { 2377 2378 if_route(ifp, IFF_UP, AF_UNSPEC); 2379 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2380 } 2381 2382 /* 2383 * Flush an interface queue. 2384 */ 2385 void 2386 if_qflush(struct ifnet *ifp) 2387 { 2388 struct mbuf *m, *n; 2389 struct ifaltq *ifq; 2390 2391 ifq = &ifp->if_snd; 2392 IFQ_LOCK(ifq); 2393 #ifdef ALTQ 2394 if (ALTQ_IS_ENABLED(ifq)) 2395 ALTQ_PURGE(ifq); 2396 #endif 2397 n = ifq->ifq_head; 2398 while ((m = n) != NULL) { 2399 n = m->m_nextpkt; 2400 m_freem(m); 2401 } 2402 ifq->ifq_head = 0; 2403 ifq->ifq_tail = 0; 2404 ifq->ifq_len = 0; 2405 IFQ_UNLOCK(ifq); 2406 } 2407 2408 /* 2409 * Map interface name to interface structure pointer, with or without 2410 * returning a reference. 2411 */ 2412 struct ifnet * 2413 ifunit_ref(const char *name) 2414 { 2415 struct ifnet *ifp; 2416 2417 IFNET_RLOCK_NOSLEEP(); 2418 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2419 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2420 !(ifp->if_flags & IFF_DYING)) 2421 break; 2422 } 2423 if (ifp != NULL) 2424 if_ref(ifp); 2425 IFNET_RUNLOCK_NOSLEEP(); 2426 return (ifp); 2427 } 2428 2429 struct ifnet * 2430 ifunit(const char *name) 2431 { 2432 struct ifnet *ifp; 2433 2434 IFNET_RLOCK_NOSLEEP(); 2435 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2436 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2437 break; 2438 } 2439 IFNET_RUNLOCK_NOSLEEP(); 2440 return (ifp); 2441 } 2442 2443 static void * 2444 ifr_buffer_get_buffer(void *data) 2445 { 2446 union ifreq_union *ifrup; 2447 2448 ifrup = data; 2449 #ifdef COMPAT_FREEBSD32 2450 if (SV_CURPROC_FLAG(SV_ILP32)) 2451 return ((void *)(uintptr_t) 2452 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2453 #endif 2454 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2455 } 2456 2457 static void 2458 ifr_buffer_set_buffer_null(void *data) 2459 { 2460 union ifreq_union *ifrup; 2461 2462 ifrup = data; 2463 #ifdef COMPAT_FREEBSD32 2464 if (SV_CURPROC_FLAG(SV_ILP32)) 2465 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2466 else 2467 #endif 2468 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2469 } 2470 2471 static size_t 2472 ifr_buffer_get_length(void *data) 2473 { 2474 union ifreq_union *ifrup; 2475 2476 ifrup = data; 2477 #ifdef COMPAT_FREEBSD32 2478 if (SV_CURPROC_FLAG(SV_ILP32)) 2479 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2480 #endif 2481 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2482 } 2483 2484 static void 2485 ifr_buffer_set_length(void *data, size_t len) 2486 { 2487 union ifreq_union *ifrup; 2488 2489 ifrup = data; 2490 #ifdef COMPAT_FREEBSD32 2491 if (SV_CURPROC_FLAG(SV_ILP32)) 2492 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2493 else 2494 #endif 2495 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2496 } 2497 2498 void * 2499 ifr_data_get_ptr(void *ifrp) 2500 { 2501 union ifreq_union *ifrup; 2502 2503 ifrup = ifrp; 2504 #ifdef COMPAT_FREEBSD32 2505 if (SV_CURPROC_FLAG(SV_ILP32)) 2506 return ((void *)(uintptr_t) 2507 ifrup->ifr32.ifr_ifru.ifru_data); 2508 #endif 2509 return (ifrup->ifr.ifr_ifru.ifru_data); 2510 } 2511 2512 /* 2513 * Hardware specific interface ioctls. 2514 */ 2515 static int 2516 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2517 { 2518 struct ifreq *ifr; 2519 int error = 0, do_ifup = 0; 2520 int new_flags, temp_flags; 2521 size_t namelen, onamelen; 2522 size_t descrlen; 2523 char *descrbuf, *odescrbuf; 2524 char new_name[IFNAMSIZ]; 2525 struct ifaddr *ifa; 2526 struct sockaddr_dl *sdl; 2527 2528 ifr = (struct ifreq *)data; 2529 switch (cmd) { 2530 case SIOCGIFINDEX: 2531 ifr->ifr_index = ifp->if_index; 2532 break; 2533 2534 case SIOCGIFFLAGS: 2535 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2536 ifr->ifr_flags = temp_flags & 0xffff; 2537 ifr->ifr_flagshigh = temp_flags >> 16; 2538 break; 2539 2540 case SIOCGIFCAP: 2541 ifr->ifr_reqcap = ifp->if_capabilities; 2542 ifr->ifr_curcap = ifp->if_capenable; 2543 break; 2544 2545 #ifdef MAC 2546 case SIOCGIFMAC: 2547 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2548 break; 2549 #endif 2550 2551 case SIOCGIFMETRIC: 2552 ifr->ifr_metric = ifp->if_metric; 2553 break; 2554 2555 case SIOCGIFMTU: 2556 ifr->ifr_mtu = ifp->if_mtu; 2557 break; 2558 2559 case SIOCGIFPHYS: 2560 /* XXXGL: did this ever worked? */ 2561 ifr->ifr_phys = 0; 2562 break; 2563 2564 case SIOCGIFDESCR: 2565 error = 0; 2566 sx_slock(&ifdescr_sx); 2567 if (ifp->if_description == NULL) 2568 error = ENOMSG; 2569 else { 2570 /* space for terminating nul */ 2571 descrlen = strlen(ifp->if_description) + 1; 2572 if (ifr_buffer_get_length(ifr) < descrlen) 2573 ifr_buffer_set_buffer_null(ifr); 2574 else 2575 error = copyout(ifp->if_description, 2576 ifr_buffer_get_buffer(ifr), descrlen); 2577 ifr_buffer_set_length(ifr, descrlen); 2578 } 2579 sx_sunlock(&ifdescr_sx); 2580 break; 2581 2582 case SIOCSIFDESCR: 2583 error = priv_check(td, PRIV_NET_SETIFDESCR); 2584 if (error) 2585 return (error); 2586 2587 /* 2588 * Copy only (length-1) bytes to make sure that 2589 * if_description is always nul terminated. The 2590 * length parameter is supposed to count the 2591 * terminating nul in. 2592 */ 2593 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2594 return (ENAMETOOLONG); 2595 else if (ifr_buffer_get_length(ifr) == 0) 2596 descrbuf = NULL; 2597 else { 2598 descrbuf = malloc(ifr_buffer_get_length(ifr), 2599 M_IFDESCR, M_WAITOK | M_ZERO); 2600 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2601 ifr_buffer_get_length(ifr) - 1); 2602 if (error) { 2603 free(descrbuf, M_IFDESCR); 2604 break; 2605 } 2606 } 2607 2608 sx_xlock(&ifdescr_sx); 2609 odescrbuf = ifp->if_description; 2610 ifp->if_description = descrbuf; 2611 sx_xunlock(&ifdescr_sx); 2612 2613 getmicrotime(&ifp->if_lastchange); 2614 free(odescrbuf, M_IFDESCR); 2615 break; 2616 2617 case SIOCGIFFIB: 2618 ifr->ifr_fib = ifp->if_fib; 2619 break; 2620 2621 case SIOCSIFFIB: 2622 error = priv_check(td, PRIV_NET_SETIFFIB); 2623 if (error) 2624 return (error); 2625 if (ifr->ifr_fib >= rt_numfibs) 2626 return (EINVAL); 2627 2628 ifp->if_fib = ifr->ifr_fib; 2629 break; 2630 2631 case SIOCSIFFLAGS: 2632 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2633 if (error) 2634 return (error); 2635 /* 2636 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2637 * check, so we don't need special handling here yet. 2638 */ 2639 new_flags = (ifr->ifr_flags & 0xffff) | 2640 (ifr->ifr_flagshigh << 16); 2641 if (ifp->if_flags & IFF_UP && 2642 (new_flags & IFF_UP) == 0) { 2643 if_down(ifp); 2644 } else if (new_flags & IFF_UP && 2645 (ifp->if_flags & IFF_UP) == 0) { 2646 do_ifup = 1; 2647 } 2648 /* See if permanently promiscuous mode bit is about to flip */ 2649 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2650 if (new_flags & IFF_PPROMISC) 2651 ifp->if_flags |= IFF_PROMISC; 2652 else if (ifp->if_pcount == 0) 2653 ifp->if_flags &= ~IFF_PROMISC; 2654 if (log_promisc_mode_change) 2655 if_printf(ifp, "permanently promiscuous mode %s\n", 2656 ((new_flags & IFF_PPROMISC) ? 2657 "enabled" : "disabled")); 2658 } 2659 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2660 (new_flags &~ IFF_CANTCHANGE); 2661 if (ifp->if_ioctl) { 2662 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2663 } 2664 if (do_ifup) 2665 if_up(ifp); 2666 getmicrotime(&ifp->if_lastchange); 2667 break; 2668 2669 case SIOCSIFCAP: 2670 error = priv_check(td, PRIV_NET_SETIFCAP); 2671 if (error) 2672 return (error); 2673 if (ifp->if_ioctl == NULL) 2674 return (EOPNOTSUPP); 2675 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2676 return (EINVAL); 2677 error = (*ifp->if_ioctl)(ifp, cmd, data); 2678 if (error == 0) 2679 getmicrotime(&ifp->if_lastchange); 2680 break; 2681 2682 #ifdef MAC 2683 case SIOCSIFMAC: 2684 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2685 break; 2686 #endif 2687 2688 case SIOCSIFNAME: 2689 error = priv_check(td, PRIV_NET_SETIFNAME); 2690 if (error) 2691 return (error); 2692 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2693 NULL); 2694 if (error != 0) 2695 return (error); 2696 if (new_name[0] == '\0') 2697 return (EINVAL); 2698 if (new_name[IFNAMSIZ-1] != '\0') { 2699 new_name[IFNAMSIZ-1] = '\0'; 2700 if (strlen(new_name) == IFNAMSIZ-1) 2701 return (EINVAL); 2702 } 2703 if (ifunit(new_name) != NULL) 2704 return (EEXIST); 2705 2706 /* 2707 * XXX: Locking. Nothing else seems to lock if_flags, 2708 * and there are numerous other races with the 2709 * ifunit() checks not being atomic with namespace 2710 * changes (renames, vmoves, if_attach, etc). 2711 */ 2712 ifp->if_flags |= IFF_RENAMING; 2713 2714 /* Announce the departure of the interface. */ 2715 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2716 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2717 2718 if_printf(ifp, "changing name to '%s'\n", new_name); 2719 2720 IF_ADDR_WLOCK(ifp); 2721 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2722 ifa = ifp->if_addr; 2723 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2724 namelen = strlen(new_name); 2725 onamelen = sdl->sdl_nlen; 2726 /* 2727 * Move the address if needed. This is safe because we 2728 * allocate space for a name of length IFNAMSIZ when we 2729 * create this in if_attach(). 2730 */ 2731 if (namelen != onamelen) { 2732 bcopy(sdl->sdl_data + onamelen, 2733 sdl->sdl_data + namelen, sdl->sdl_alen); 2734 } 2735 bcopy(new_name, sdl->sdl_data, namelen); 2736 sdl->sdl_nlen = namelen; 2737 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2738 bzero(sdl->sdl_data, onamelen); 2739 while (namelen != 0) 2740 sdl->sdl_data[--namelen] = 0xff; 2741 IF_ADDR_WUNLOCK(ifp); 2742 2743 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2744 /* Announce the return of the interface. */ 2745 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2746 2747 ifp->if_flags &= ~IFF_RENAMING; 2748 break; 2749 2750 #ifdef VIMAGE 2751 case SIOCSIFVNET: 2752 error = priv_check(td, PRIV_NET_SETIFVNET); 2753 if (error) 2754 return (error); 2755 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2756 break; 2757 #endif 2758 2759 case SIOCSIFMETRIC: 2760 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2761 if (error) 2762 return (error); 2763 ifp->if_metric = ifr->ifr_metric; 2764 getmicrotime(&ifp->if_lastchange); 2765 break; 2766 2767 case SIOCSIFPHYS: 2768 error = priv_check(td, PRIV_NET_SETIFPHYS); 2769 if (error) 2770 return (error); 2771 if (ifp->if_ioctl == NULL) 2772 return (EOPNOTSUPP); 2773 error = (*ifp->if_ioctl)(ifp, cmd, data); 2774 if (error == 0) 2775 getmicrotime(&ifp->if_lastchange); 2776 break; 2777 2778 case SIOCSIFMTU: 2779 { 2780 u_long oldmtu = ifp->if_mtu; 2781 2782 error = priv_check(td, PRIV_NET_SETIFMTU); 2783 if (error) 2784 return (error); 2785 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2786 return (EINVAL); 2787 if (ifp->if_ioctl == NULL) 2788 return (EOPNOTSUPP); 2789 error = (*ifp->if_ioctl)(ifp, cmd, data); 2790 if (error == 0) { 2791 getmicrotime(&ifp->if_lastchange); 2792 rt_ifmsg(ifp); 2793 #ifdef INET 2794 NETDUMP_REINIT(ifp); 2795 #endif 2796 } 2797 /* 2798 * If the link MTU changed, do network layer specific procedure. 2799 */ 2800 if (ifp->if_mtu != oldmtu) { 2801 #ifdef INET6 2802 nd6_setmtu(ifp); 2803 #endif 2804 rt_updatemtu(ifp); 2805 } 2806 break; 2807 } 2808 2809 case SIOCADDMULTI: 2810 case SIOCDELMULTI: 2811 if (cmd == SIOCADDMULTI) 2812 error = priv_check(td, PRIV_NET_ADDMULTI); 2813 else 2814 error = priv_check(td, PRIV_NET_DELMULTI); 2815 if (error) 2816 return (error); 2817 2818 /* Don't allow group membership on non-multicast interfaces. */ 2819 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2820 return (EOPNOTSUPP); 2821 2822 /* Don't let users screw up protocols' entries. */ 2823 if (ifr->ifr_addr.sa_family != AF_LINK) 2824 return (EINVAL); 2825 2826 if (cmd == SIOCADDMULTI) { 2827 struct ifmultiaddr *ifma; 2828 2829 /* 2830 * Userland is only permitted to join groups once 2831 * via the if_addmulti() KPI, because it cannot hold 2832 * struct ifmultiaddr * between calls. It may also 2833 * lose a race while we check if the membership 2834 * already exists. 2835 */ 2836 IF_ADDR_RLOCK(ifp); 2837 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2838 IF_ADDR_RUNLOCK(ifp); 2839 if (ifma != NULL) 2840 error = EADDRINUSE; 2841 else 2842 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2843 } else { 2844 error = if_delmulti(ifp, &ifr->ifr_addr); 2845 } 2846 if (error == 0) 2847 getmicrotime(&ifp->if_lastchange); 2848 break; 2849 2850 case SIOCSIFPHYADDR: 2851 case SIOCDIFPHYADDR: 2852 #ifdef INET6 2853 case SIOCSIFPHYADDR_IN6: 2854 #endif 2855 case SIOCSIFMEDIA: 2856 case SIOCSIFGENERIC: 2857 error = priv_check(td, PRIV_NET_HWIOCTL); 2858 if (error) 2859 return (error); 2860 if (ifp->if_ioctl == NULL) 2861 return (EOPNOTSUPP); 2862 error = (*ifp->if_ioctl)(ifp, cmd, data); 2863 if (error == 0) 2864 getmicrotime(&ifp->if_lastchange); 2865 break; 2866 2867 case SIOCGIFSTATUS: 2868 case SIOCGIFPSRCADDR: 2869 case SIOCGIFPDSTADDR: 2870 case SIOCGIFMEDIA: 2871 case SIOCGIFXMEDIA: 2872 case SIOCGIFGENERIC: 2873 case SIOCGIFRSSKEY: 2874 case SIOCGIFRSSHASH: 2875 if (ifp->if_ioctl == NULL) 2876 return (EOPNOTSUPP); 2877 error = (*ifp->if_ioctl)(ifp, cmd, data); 2878 break; 2879 2880 case SIOCSIFLLADDR: 2881 error = priv_check(td, PRIV_NET_SETLLADDR); 2882 if (error) 2883 return (error); 2884 error = if_setlladdr(ifp, 2885 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2886 break; 2887 2888 case SIOCGHWADDR: 2889 error = if_gethwaddr(ifp, ifr); 2890 break; 2891 2892 CASE_IOC_IFGROUPREQ(SIOCAIFGROUP): 2893 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2894 if (error) 2895 return (error); 2896 if ((error = if_addgroup(ifp, 2897 ifgr_group_get((struct ifgroupreq *)data)))) 2898 return (error); 2899 break; 2900 2901 CASE_IOC_IFGROUPREQ(SIOCGIFGROUP): 2902 if ((error = if_getgroup((struct ifgroupreq *)data, ifp))) 2903 return (error); 2904 break; 2905 2906 CASE_IOC_IFGROUPREQ(SIOCDIFGROUP): 2907 error = priv_check(td, PRIV_NET_DELIFGROUP); 2908 if (error) 2909 return (error); 2910 if ((error = if_delgroup(ifp, 2911 ifgr_group_get((struct ifgroupreq *)data)))) 2912 return (error); 2913 break; 2914 2915 default: 2916 error = ENOIOCTL; 2917 break; 2918 } 2919 return (error); 2920 } 2921 2922 #ifdef COMPAT_FREEBSD32 2923 struct ifconf32 { 2924 int32_t ifc_len; 2925 union { 2926 uint32_t ifcu_buf; 2927 uint32_t ifcu_req; 2928 } ifc_ifcu; 2929 }; 2930 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2931 #endif 2932 2933 #ifdef COMPAT_FREEBSD32 2934 static void 2935 ifmr_init(struct ifmediareq *ifmr, caddr_t data) 2936 { 2937 struct ifmediareq32 *ifmr32; 2938 2939 ifmr32 = (struct ifmediareq32 *)data; 2940 memcpy(ifmr->ifm_name, ifmr32->ifm_name, 2941 sizeof(ifmr->ifm_name)); 2942 ifmr->ifm_current = ifmr32->ifm_current; 2943 ifmr->ifm_mask = ifmr32->ifm_mask; 2944 ifmr->ifm_status = ifmr32->ifm_status; 2945 ifmr->ifm_active = ifmr32->ifm_active; 2946 ifmr->ifm_count = ifmr32->ifm_count; 2947 ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist; 2948 } 2949 2950 static void 2951 ifmr_update(const struct ifmediareq *ifmr, caddr_t data) 2952 { 2953 struct ifmediareq32 *ifmr32; 2954 2955 ifmr32 = (struct ifmediareq32 *)data; 2956 ifmr32->ifm_current = ifmr->ifm_current; 2957 ifmr32->ifm_mask = ifmr->ifm_mask; 2958 ifmr32->ifm_status = ifmr->ifm_status; 2959 ifmr32->ifm_active = ifmr->ifm_active; 2960 ifmr32->ifm_count = ifmr->ifm_count; 2961 } 2962 #endif 2963 2964 /* 2965 * Interface ioctls. 2966 */ 2967 int 2968 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2969 { 2970 #ifdef COMPAT_FREEBSD32 2971 caddr_t saved_data = NULL; 2972 struct ifmediareq ifmr; 2973 struct ifmediareq *ifmrp; 2974 #endif 2975 struct ifnet *ifp; 2976 struct ifreq *ifr; 2977 int error; 2978 int oif_flags; 2979 #ifdef VIMAGE 2980 int shutdown; 2981 #endif 2982 2983 CURVNET_SET(so->so_vnet); 2984 #ifdef VIMAGE 2985 /* Make sure the VNET is stable. */ 2986 shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET && 2987 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 2988 if (shutdown) { 2989 CURVNET_RESTORE(); 2990 return (EBUSY); 2991 } 2992 #endif 2993 2994 2995 switch (cmd) { 2996 case SIOCGIFCONF: 2997 error = ifconf(cmd, data); 2998 CURVNET_RESTORE(); 2999 return (error); 3000 3001 #ifdef COMPAT_FREEBSD32 3002 case SIOCGIFCONF32: 3003 { 3004 struct ifconf32 *ifc32; 3005 struct ifconf ifc; 3006 3007 ifc32 = (struct ifconf32 *)data; 3008 ifc.ifc_len = ifc32->ifc_len; 3009 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 3010 3011 error = ifconf(SIOCGIFCONF, (void *)&ifc); 3012 CURVNET_RESTORE(); 3013 if (error == 0) 3014 ifc32->ifc_len = ifc.ifc_len; 3015 return (error); 3016 } 3017 #endif 3018 } 3019 3020 #ifdef COMPAT_FREEBSD32 3021 ifmrp = NULL; 3022 switch (cmd) { 3023 case SIOCGIFMEDIA32: 3024 case SIOCGIFXMEDIA32: 3025 ifmrp = &ifmr; 3026 ifmr_init(ifmrp, data); 3027 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 3028 saved_data = data; 3029 data = (caddr_t)ifmrp; 3030 } 3031 #endif 3032 3033 ifr = (struct ifreq *)data; 3034 switch (cmd) { 3035 #ifdef VIMAGE 3036 case SIOCSIFRVNET: 3037 error = priv_check(td, PRIV_NET_SETIFVNET); 3038 if (error == 0) 3039 error = if_vmove_reclaim(td, ifr->ifr_name, 3040 ifr->ifr_jid); 3041 goto out_noref; 3042 #endif 3043 case SIOCIFCREATE: 3044 case SIOCIFCREATE2: 3045 error = priv_check(td, PRIV_NET_IFCREATE); 3046 if (error == 0) 3047 error = if_clone_create(ifr->ifr_name, 3048 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3049 ifr_data_get_ptr(ifr) : NULL); 3050 goto out_noref; 3051 case SIOCIFDESTROY: 3052 error = priv_check(td, PRIV_NET_IFDESTROY); 3053 if (error == 0) 3054 error = if_clone_destroy(ifr->ifr_name); 3055 goto out_noref; 3056 3057 case SIOCIFGCLONERS: 3058 error = if_clone_list((struct if_clonereq *)data); 3059 goto out_noref; 3060 3061 CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB): 3062 error = if_getgroupmembers((struct ifgroupreq *)data); 3063 goto out_noref; 3064 3065 #if defined(INET) || defined(INET6) 3066 case SIOCSVH: 3067 case SIOCGVH: 3068 if (carp_ioctl_p == NULL) 3069 error = EPROTONOSUPPORT; 3070 else 3071 error = (*carp_ioctl_p)(ifr, cmd, td); 3072 goto out_noref; 3073 #endif 3074 } 3075 3076 ifp = ifunit_ref(ifr->ifr_name); 3077 if (ifp == NULL) { 3078 error = ENXIO; 3079 goto out_noref; 3080 } 3081 3082 error = ifhwioctl(cmd, ifp, data, td); 3083 if (error != ENOIOCTL) 3084 goto out_ref; 3085 3086 oif_flags = ifp->if_flags; 3087 if (so->so_proto == NULL) { 3088 error = EOPNOTSUPP; 3089 goto out_ref; 3090 } 3091 3092 /* 3093 * Pass the request on to the socket control method, and if the 3094 * latter returns EOPNOTSUPP, directly to the interface. 3095 * 3096 * Make an exception for the legacy SIOCSIF* requests. Drivers 3097 * trust SIOCSIFADDR et al to come from an already privileged 3098 * layer, and do not perform any credentials checks or input 3099 * validation. 3100 */ 3101 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 3102 ifp, td)); 3103 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3104 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3105 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3106 error = (*ifp->if_ioctl)(ifp, cmd, data); 3107 3108 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3109 #ifdef INET6 3110 if (ifp->if_flags & IFF_UP) 3111 in6_if_up(ifp); 3112 #endif 3113 } 3114 3115 out_ref: 3116 if_rele(ifp); 3117 out_noref: 3118 #ifdef COMPAT_FREEBSD32 3119 if (ifmrp != NULL) { 3120 KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA), 3121 ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx", 3122 cmd)); 3123 data = saved_data; 3124 ifmr_update(ifmrp, data); 3125 } 3126 #endif 3127 CURVNET_RESTORE(); 3128 return (error); 3129 } 3130 3131 /* 3132 * The code common to handling reference counted flags, 3133 * e.g., in ifpromisc() and if_allmulti(). 3134 * The "pflag" argument can specify a permanent mode flag to check, 3135 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3136 * 3137 * Only to be used on stack-owned flags, not driver-owned flags. 3138 */ 3139 static int 3140 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3141 { 3142 struct ifreq ifr; 3143 int error; 3144 int oldflags, oldcount; 3145 3146 /* Sanity checks to catch programming errors */ 3147 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3148 ("%s: setting driver-owned flag %d", __func__, flag)); 3149 3150 if (onswitch) 3151 KASSERT(*refcount >= 0, 3152 ("%s: increment negative refcount %d for flag %d", 3153 __func__, *refcount, flag)); 3154 else 3155 KASSERT(*refcount > 0, 3156 ("%s: decrement non-positive refcount %d for flag %d", 3157 __func__, *refcount, flag)); 3158 3159 /* In case this mode is permanent, just touch refcount */ 3160 if (ifp->if_flags & pflag) { 3161 *refcount += onswitch ? 1 : -1; 3162 return (0); 3163 } 3164 3165 /* Save ifnet parameters for if_ioctl() may fail */ 3166 oldcount = *refcount; 3167 oldflags = ifp->if_flags; 3168 3169 /* 3170 * See if we aren't the only and touching refcount is enough. 3171 * Actually toggle interface flag if we are the first or last. 3172 */ 3173 if (onswitch) { 3174 if ((*refcount)++) 3175 return (0); 3176 ifp->if_flags |= flag; 3177 } else { 3178 if (--(*refcount)) 3179 return (0); 3180 ifp->if_flags &= ~flag; 3181 } 3182 3183 /* Call down the driver since we've changed interface flags */ 3184 if (ifp->if_ioctl == NULL) { 3185 error = EOPNOTSUPP; 3186 goto recover; 3187 } 3188 ifr.ifr_flags = ifp->if_flags & 0xffff; 3189 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3190 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3191 if (error) 3192 goto recover; 3193 /* Notify userland that interface flags have changed */ 3194 rt_ifmsg(ifp); 3195 return (0); 3196 3197 recover: 3198 /* Recover after driver error */ 3199 *refcount = oldcount; 3200 ifp->if_flags = oldflags; 3201 return (error); 3202 } 3203 3204 /* 3205 * Set/clear promiscuous mode on interface ifp based on the truth value 3206 * of pswitch. The calls are reference counted so that only the first 3207 * "on" request actually has an effect, as does the final "off" request. 3208 * Results are undefined if the "off" and "on" requests are not matched. 3209 */ 3210 int 3211 ifpromisc(struct ifnet *ifp, int pswitch) 3212 { 3213 int error; 3214 int oldflags = ifp->if_flags; 3215 3216 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3217 &ifp->if_pcount, pswitch); 3218 /* If promiscuous mode status has changed, log a message */ 3219 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3220 log_promisc_mode_change) 3221 if_printf(ifp, "promiscuous mode %s\n", 3222 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3223 return (error); 3224 } 3225 3226 /* 3227 * Return interface configuration 3228 * of system. List may be used 3229 * in later ioctl's (above) to get 3230 * other information. 3231 */ 3232 /*ARGSUSED*/ 3233 static int 3234 ifconf(u_long cmd, caddr_t data) 3235 { 3236 struct ifconf *ifc = (struct ifconf *)data; 3237 struct ifnet *ifp; 3238 struct ifaddr *ifa; 3239 struct ifreq ifr; 3240 struct sbuf *sb; 3241 int error, full = 0, valid_len, max_len; 3242 3243 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 3244 max_len = MAXPHYS - 1; 3245 3246 /* Prevent hostile input from being able to crash the system */ 3247 if (ifc->ifc_len <= 0) 3248 return (EINVAL); 3249 3250 again: 3251 if (ifc->ifc_len <= max_len) { 3252 max_len = ifc->ifc_len; 3253 full = 1; 3254 } 3255 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3256 max_len = 0; 3257 valid_len = 0; 3258 3259 IFNET_RLOCK(); 3260 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3261 int addrs; 3262 3263 /* 3264 * Zero the ifr to make sure we don't disclose the contents 3265 * of the stack. 3266 */ 3267 memset(&ifr, 0, sizeof(ifr)); 3268 3269 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3270 >= sizeof(ifr.ifr_name)) { 3271 sbuf_delete(sb); 3272 IFNET_RUNLOCK(); 3273 return (ENAMETOOLONG); 3274 } 3275 3276 addrs = 0; 3277 IF_ADDR_RLOCK(ifp); 3278 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3279 struct sockaddr *sa = ifa->ifa_addr; 3280 3281 if (prison_if(curthread->td_ucred, sa) != 0) 3282 continue; 3283 addrs++; 3284 if (sa->sa_len <= sizeof(*sa)) { 3285 if (sa->sa_len < sizeof(*sa)) { 3286 memset(&ifr.ifr_ifru.ifru_addr, 0, 3287 sizeof(ifr.ifr_ifru.ifru_addr)); 3288 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3289 sa->sa_len); 3290 } else 3291 ifr.ifr_ifru.ifru_addr = *sa; 3292 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3293 max_len += sizeof(ifr); 3294 } else { 3295 sbuf_bcat(sb, &ifr, 3296 offsetof(struct ifreq, ifr_addr)); 3297 max_len += offsetof(struct ifreq, ifr_addr); 3298 sbuf_bcat(sb, sa, sa->sa_len); 3299 max_len += sa->sa_len; 3300 } 3301 3302 if (sbuf_error(sb) == 0) 3303 valid_len = sbuf_len(sb); 3304 } 3305 IF_ADDR_RUNLOCK(ifp); 3306 if (addrs == 0) { 3307 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3308 max_len += sizeof(ifr); 3309 3310 if (sbuf_error(sb) == 0) 3311 valid_len = sbuf_len(sb); 3312 } 3313 } 3314 IFNET_RUNLOCK(); 3315 3316 /* 3317 * If we didn't allocate enough space (uncommon), try again. If 3318 * we have already allocated as much space as we are allowed, 3319 * return what we've got. 3320 */ 3321 if (valid_len != max_len && !full) { 3322 sbuf_delete(sb); 3323 goto again; 3324 } 3325 3326 ifc->ifc_len = valid_len; 3327 sbuf_finish(sb); 3328 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3329 sbuf_delete(sb); 3330 return (error); 3331 } 3332 3333 /* 3334 * Just like ifpromisc(), but for all-multicast-reception mode. 3335 */ 3336 int 3337 if_allmulti(struct ifnet *ifp, int onswitch) 3338 { 3339 3340 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3341 } 3342 3343 struct ifmultiaddr * 3344 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3345 { 3346 struct ifmultiaddr *ifma; 3347 3348 IF_ADDR_LOCK_ASSERT(ifp); 3349 3350 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3351 if (sa->sa_family == AF_LINK) { 3352 if (sa_dl_equal(ifma->ifma_addr, sa)) 3353 break; 3354 } else { 3355 if (sa_equal(ifma->ifma_addr, sa)) 3356 break; 3357 } 3358 } 3359 3360 return ifma; 3361 } 3362 3363 /* 3364 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3365 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3366 * the ifnet multicast address list here, so the caller must do that and 3367 * other setup work (such as notifying the device driver). The reference 3368 * count is initialized to 1. 3369 */ 3370 static struct ifmultiaddr * 3371 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3372 int mflags) 3373 { 3374 struct ifmultiaddr *ifma; 3375 struct sockaddr *dupsa; 3376 3377 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3378 M_ZERO); 3379 if (ifma == NULL) 3380 return (NULL); 3381 3382 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3383 if (dupsa == NULL) { 3384 free(ifma, M_IFMADDR); 3385 return (NULL); 3386 } 3387 bcopy(sa, dupsa, sa->sa_len); 3388 ifma->ifma_addr = dupsa; 3389 3390 ifma->ifma_ifp = ifp; 3391 ifma->ifma_refcount = 1; 3392 ifma->ifma_protospec = NULL; 3393 3394 if (llsa == NULL) { 3395 ifma->ifma_lladdr = NULL; 3396 return (ifma); 3397 } 3398 3399 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3400 if (dupsa == NULL) { 3401 free(ifma->ifma_addr, M_IFMADDR); 3402 free(ifma, M_IFMADDR); 3403 return (NULL); 3404 } 3405 bcopy(llsa, dupsa, llsa->sa_len); 3406 ifma->ifma_lladdr = dupsa; 3407 3408 return (ifma); 3409 } 3410 3411 /* 3412 * if_freemulti: free ifmultiaddr structure and possibly attached related 3413 * addresses. The caller is responsible for implementing reference 3414 * counting, notifying the driver, handling routing messages, and releasing 3415 * any dependent link layer state. 3416 */ 3417 #ifdef MCAST_VERBOSE 3418 extern void kdb_backtrace(void); 3419 #endif 3420 static void 3421 if_freemulti_internal(struct ifmultiaddr *ifma) 3422 { 3423 3424 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3425 ifma->ifma_refcount)); 3426 3427 if (ifma->ifma_lladdr != NULL) 3428 free(ifma->ifma_lladdr, M_IFMADDR); 3429 #ifdef MCAST_VERBOSE 3430 kdb_backtrace(); 3431 printf("%s freeing ifma: %p\n", __func__, ifma); 3432 #endif 3433 free(ifma->ifma_addr, M_IFMADDR); 3434 free(ifma, M_IFMADDR); 3435 } 3436 3437 static void 3438 if_destroymulti(epoch_context_t ctx) 3439 { 3440 struct ifmultiaddr *ifma; 3441 3442 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3443 if_freemulti_internal(ifma); 3444 } 3445 3446 void 3447 if_freemulti(struct ifmultiaddr *ifma) 3448 { 3449 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3450 ifma->ifma_refcount)); 3451 3452 epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti); 3453 } 3454 3455 3456 /* 3457 * Register an additional multicast address with a network interface. 3458 * 3459 * - If the address is already present, bump the reference count on the 3460 * address and return. 3461 * - If the address is not link-layer, look up a link layer address. 3462 * - Allocate address structures for one or both addresses, and attach to the 3463 * multicast address list on the interface. If automatically adding a link 3464 * layer address, the protocol address will own a reference to the link 3465 * layer address, to be freed when it is freed. 3466 * - Notify the network device driver of an addition to the multicast address 3467 * list. 3468 * 3469 * 'sa' points to caller-owned memory with the desired multicast address. 3470 * 3471 * 'retifma' will be used to return a pointer to the resulting multicast 3472 * address reference, if desired. 3473 */ 3474 int 3475 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3476 struct ifmultiaddr **retifma) 3477 { 3478 struct ifmultiaddr *ifma, *ll_ifma; 3479 struct sockaddr *llsa; 3480 struct sockaddr_dl sdl; 3481 int error; 3482 3483 #ifdef INET 3484 IN_MULTI_LIST_UNLOCK_ASSERT(); 3485 #endif 3486 #ifdef INET6 3487 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3488 #endif 3489 /* 3490 * If the address is already present, return a new reference to it; 3491 * otherwise, allocate storage and set up a new address. 3492 */ 3493 IF_ADDR_WLOCK(ifp); 3494 ifma = if_findmulti(ifp, sa); 3495 if (ifma != NULL) { 3496 ifma->ifma_refcount++; 3497 if (retifma != NULL) 3498 *retifma = ifma; 3499 IF_ADDR_WUNLOCK(ifp); 3500 return (0); 3501 } 3502 3503 /* 3504 * The address isn't already present; resolve the protocol address 3505 * into a link layer address, and then look that up, bump its 3506 * refcount or allocate an ifma for that also. 3507 * Most link layer resolving functions returns address data which 3508 * fits inside default sockaddr_dl structure. However callback 3509 * can allocate another sockaddr structure, in that case we need to 3510 * free it later. 3511 */ 3512 llsa = NULL; 3513 ll_ifma = NULL; 3514 if (ifp->if_resolvemulti != NULL) { 3515 /* Provide called function with buffer size information */ 3516 sdl.sdl_len = sizeof(sdl); 3517 llsa = (struct sockaddr *)&sdl; 3518 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3519 if (error) 3520 goto unlock_out; 3521 } 3522 3523 /* 3524 * Allocate the new address. Don't hook it up yet, as we may also 3525 * need to allocate a link layer multicast address. 3526 */ 3527 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3528 if (ifma == NULL) { 3529 error = ENOMEM; 3530 goto free_llsa_out; 3531 } 3532 3533 /* 3534 * If a link layer address is found, we'll need to see if it's 3535 * already present in the address list, or allocate is as well. 3536 * When this block finishes, the link layer address will be on the 3537 * list. 3538 */ 3539 if (llsa != NULL) { 3540 ll_ifma = if_findmulti(ifp, llsa); 3541 if (ll_ifma == NULL) { 3542 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3543 if (ll_ifma == NULL) { 3544 --ifma->ifma_refcount; 3545 if_freemulti(ifma); 3546 error = ENOMEM; 3547 goto free_llsa_out; 3548 } 3549 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3550 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3551 ifma_link); 3552 } else 3553 ll_ifma->ifma_refcount++; 3554 ifma->ifma_llifma = ll_ifma; 3555 } 3556 3557 /* 3558 * We now have a new multicast address, ifma, and possibly a new or 3559 * referenced link layer address. Add the primary address to the 3560 * ifnet address list. 3561 */ 3562 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3563 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3564 3565 if (retifma != NULL) 3566 *retifma = ifma; 3567 3568 /* 3569 * Must generate the message while holding the lock so that 'ifma' 3570 * pointer is still valid. 3571 */ 3572 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3573 IF_ADDR_WUNLOCK(ifp); 3574 3575 /* 3576 * We are certain we have added something, so call down to the 3577 * interface to let them know about it. 3578 */ 3579 if (ifp->if_ioctl != NULL) { 3580 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3581 } 3582 3583 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3584 link_free_sdl(llsa); 3585 3586 return (0); 3587 3588 free_llsa_out: 3589 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3590 link_free_sdl(llsa); 3591 3592 unlock_out: 3593 IF_ADDR_WUNLOCK(ifp); 3594 return (error); 3595 } 3596 3597 /* 3598 * Delete a multicast group membership by network-layer group address. 3599 * 3600 * Returns ENOENT if the entry could not be found. If ifp no longer 3601 * exists, results are undefined. This entry point should only be used 3602 * from subsystems which do appropriate locking to hold ifp for the 3603 * duration of the call. 3604 * Network-layer protocol domains must use if_delmulti_ifma(). 3605 */ 3606 int 3607 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3608 { 3609 struct ifmultiaddr *ifma; 3610 int lastref; 3611 #ifdef INVARIANTS 3612 struct ifnet *oifp; 3613 3614 IFNET_RLOCK_NOSLEEP(); 3615 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3616 if (ifp == oifp) 3617 break; 3618 if (ifp != oifp) 3619 ifp = NULL; 3620 IFNET_RUNLOCK_NOSLEEP(); 3621 3622 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3623 #endif 3624 if (ifp == NULL) 3625 return (ENOENT); 3626 3627 IF_ADDR_WLOCK(ifp); 3628 lastref = 0; 3629 ifma = if_findmulti(ifp, sa); 3630 if (ifma != NULL) 3631 lastref = if_delmulti_locked(ifp, ifma, 0); 3632 IF_ADDR_WUNLOCK(ifp); 3633 3634 if (ifma == NULL) 3635 return (ENOENT); 3636 3637 if (lastref && ifp->if_ioctl != NULL) { 3638 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3639 } 3640 3641 return (0); 3642 } 3643 3644 /* 3645 * Delete all multicast group membership for an interface. 3646 * Should be used to quickly flush all multicast filters. 3647 */ 3648 void 3649 if_delallmulti(struct ifnet *ifp) 3650 { 3651 struct ifmultiaddr *ifma; 3652 struct ifmultiaddr *next; 3653 3654 IF_ADDR_WLOCK(ifp); 3655 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3656 if_delmulti_locked(ifp, ifma, 0); 3657 IF_ADDR_WUNLOCK(ifp); 3658 } 3659 3660 void 3661 if_delmulti_ifma(struct ifmultiaddr *ifma) 3662 { 3663 if_delmulti_ifma_flags(ifma, 0); 3664 } 3665 3666 /* 3667 * Delete a multicast group membership by group membership pointer. 3668 * Network-layer protocol domains must use this routine. 3669 * 3670 * It is safe to call this routine if the ifp disappeared. 3671 */ 3672 void 3673 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3674 { 3675 struct ifnet *ifp; 3676 int lastref; 3677 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3678 #ifdef INET 3679 IN_MULTI_LIST_UNLOCK_ASSERT(); 3680 #endif 3681 ifp = ifma->ifma_ifp; 3682 #ifdef DIAGNOSTIC 3683 if (ifp == NULL) { 3684 printf("%s: ifma_ifp seems to be detached\n", __func__); 3685 } else { 3686 struct ifnet *oifp; 3687 3688 IFNET_RLOCK_NOSLEEP(); 3689 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3690 if (ifp == oifp) 3691 break; 3692 if (ifp != oifp) 3693 ifp = NULL; 3694 IFNET_RUNLOCK_NOSLEEP(); 3695 } 3696 #endif 3697 /* 3698 * If and only if the ifnet instance exists: Acquire the address lock. 3699 */ 3700 if (ifp != NULL) 3701 IF_ADDR_WLOCK(ifp); 3702 3703 lastref = if_delmulti_locked(ifp, ifma, flags); 3704 3705 if (ifp != NULL) { 3706 /* 3707 * If and only if the ifnet instance exists: 3708 * Release the address lock. 3709 * If the group was left: update the hardware hash filter. 3710 */ 3711 IF_ADDR_WUNLOCK(ifp); 3712 if (lastref && ifp->if_ioctl != NULL) { 3713 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3714 } 3715 } 3716 } 3717 3718 /* 3719 * Perform deletion of network-layer and/or link-layer multicast address. 3720 * 3721 * Return 0 if the reference count was decremented. 3722 * Return 1 if the final reference was released, indicating that the 3723 * hardware hash filter should be reprogrammed. 3724 */ 3725 static int 3726 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3727 { 3728 struct ifmultiaddr *ll_ifma; 3729 3730 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3731 KASSERT(ifma->ifma_ifp == ifp, 3732 ("%s: inconsistent ifp %p", __func__, ifp)); 3733 IF_ADDR_WLOCK_ASSERT(ifp); 3734 } 3735 3736 ifp = ifma->ifma_ifp; 3737 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3738 3739 /* 3740 * If the ifnet is detaching, null out references to ifnet, 3741 * so that upper protocol layers will notice, and not attempt 3742 * to obtain locks for an ifnet which no longer exists. The 3743 * routing socket announcement must happen before the ifnet 3744 * instance is detached from the system. 3745 */ 3746 if (detaching) { 3747 #ifdef DIAGNOSTIC 3748 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3749 #endif 3750 /* 3751 * ifp may already be nulled out if we are being reentered 3752 * to delete the ll_ifma. 3753 */ 3754 if (ifp != NULL) { 3755 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3756 ifma->ifma_ifp = NULL; 3757 } 3758 } 3759 3760 if (--ifma->ifma_refcount > 0) 3761 return 0; 3762 3763 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3764 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3765 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3766 } 3767 /* 3768 * If this ifma is a network-layer ifma, a link-layer ifma may 3769 * have been associated with it. Release it first if so. 3770 */ 3771 ll_ifma = ifma->ifma_llifma; 3772 if (ll_ifma != NULL) { 3773 KASSERT(ifma->ifma_lladdr != NULL, 3774 ("%s: llifma w/o lladdr", __func__)); 3775 if (detaching) 3776 ll_ifma->ifma_ifp = NULL; /* XXX */ 3777 if (--ll_ifma->ifma_refcount == 0) { 3778 if (ifp != NULL) { 3779 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3780 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3781 ifma_link); 3782 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3783 } 3784 } 3785 if_freemulti(ll_ifma); 3786 } 3787 } 3788 #ifdef INVARIANTS 3789 if (ifp) { 3790 struct ifmultiaddr *ifmatmp; 3791 3792 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3793 MPASS(ifma != ifmatmp); 3794 } 3795 #endif 3796 if_freemulti(ifma); 3797 /* 3798 * The last reference to this instance of struct ifmultiaddr 3799 * was released; the hardware should be notified of this change. 3800 */ 3801 return 1; 3802 } 3803 3804 /* 3805 * Set the link layer address on an interface. 3806 * 3807 * At this time we only support certain types of interfaces, 3808 * and we don't allow the length of the address to change. 3809 * 3810 * Set noinline to be dtrace-friendly 3811 */ 3812 __noinline int 3813 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3814 { 3815 struct sockaddr_dl *sdl; 3816 struct ifaddr *ifa; 3817 struct ifreq ifr; 3818 int rc; 3819 3820 rc = 0; 3821 NET_EPOCH_ENTER(); 3822 ifa = ifp->if_addr; 3823 if (ifa == NULL) { 3824 rc = EINVAL; 3825 goto out; 3826 } 3827 3828 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3829 if (sdl == NULL) { 3830 rc = EINVAL; 3831 goto out; 3832 } 3833 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3834 rc = EINVAL; 3835 goto out; 3836 } 3837 switch (ifp->if_type) { 3838 case IFT_ETHER: 3839 case IFT_XETHER: 3840 case IFT_L2VLAN: 3841 case IFT_BRIDGE: 3842 case IFT_IEEE8023ADLAG: 3843 bcopy(lladdr, LLADDR(sdl), len); 3844 break; 3845 default: 3846 rc = ENODEV; 3847 goto out; 3848 } 3849 3850 /* 3851 * If the interface is already up, we need 3852 * to re-init it in order to reprogram its 3853 * address filter. 3854 */ 3855 NET_EPOCH_EXIT(); 3856 if ((ifp->if_flags & IFF_UP) != 0) { 3857 if (ifp->if_ioctl) { 3858 ifp->if_flags &= ~IFF_UP; 3859 ifr.ifr_flags = ifp->if_flags & 0xffff; 3860 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3861 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3862 ifp->if_flags |= IFF_UP; 3863 ifr.ifr_flags = ifp->if_flags & 0xffff; 3864 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3865 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3866 } 3867 } 3868 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3869 return (0); 3870 out: 3871 NET_EPOCH_EXIT(); 3872 return (rc); 3873 } 3874 3875 /* 3876 * Compat function for handling basic encapsulation requests. 3877 * Not converted stacks (FDDI, IB, ..) supports traditional 3878 * output model: ARP (and other similar L2 protocols) are handled 3879 * inside output routine, arpresolve/nd6_resolve() returns MAC 3880 * address instead of full prepend. 3881 * 3882 * This function creates calculated header==MAC for IPv4/IPv6 and 3883 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3884 * address families. 3885 */ 3886 static int 3887 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3888 { 3889 3890 if (req->rtype != IFENCAP_LL) 3891 return (EOPNOTSUPP); 3892 3893 if (req->bufsize < req->lladdr_len) 3894 return (ENOMEM); 3895 3896 switch (req->family) { 3897 case AF_INET: 3898 case AF_INET6: 3899 break; 3900 default: 3901 return (EAFNOSUPPORT); 3902 } 3903 3904 /* Copy lladdr to storage as is */ 3905 memmove(req->buf, req->lladdr, req->lladdr_len); 3906 req->bufsize = req->lladdr_len; 3907 req->lladdr_off = 0; 3908 3909 return (0); 3910 } 3911 3912 /* 3913 * Tunnel interfaces can nest, also they may cause infinite recursion 3914 * calls when misconfigured. We'll prevent this by detecting loops. 3915 * High nesting level may cause stack exhaustion. We'll prevent this 3916 * by introducing upper limit. 3917 * 3918 * Return 0, if tunnel nesting count is equal or less than limit. 3919 */ 3920 int 3921 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3922 int limit) 3923 { 3924 struct m_tag *mtag; 3925 int count; 3926 3927 count = 1; 3928 mtag = NULL; 3929 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3930 if (*(struct ifnet **)(mtag + 1) == ifp) { 3931 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3932 return (EIO); 3933 } 3934 count++; 3935 } 3936 if (count > limit) { 3937 log(LOG_NOTICE, 3938 "%s: if_output recursively called too many times(%d)\n", 3939 if_name(ifp), count); 3940 return (EIO); 3941 } 3942 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3943 if (mtag == NULL) 3944 return (ENOMEM); 3945 *(struct ifnet **)(mtag + 1) = ifp; 3946 m_tag_prepend(m, mtag); 3947 return (0); 3948 } 3949 3950 /* 3951 * Get the link layer address that was read from the hardware at attach. 3952 * 3953 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3954 * their component interfaces as IFT_IEEE8023ADLAG. 3955 */ 3956 int 3957 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3958 { 3959 3960 if (ifp->if_hw_addr == NULL) 3961 return (ENODEV); 3962 3963 switch (ifp->if_type) { 3964 case IFT_ETHER: 3965 case IFT_IEEE8023ADLAG: 3966 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3967 return (0); 3968 default: 3969 return (ENODEV); 3970 } 3971 } 3972 3973 /* 3974 * The name argument must be a pointer to storage which will last as 3975 * long as the interface does. For physical devices, the result of 3976 * device_get_name(dev) is a good choice and for pseudo-devices a 3977 * static string works well. 3978 */ 3979 void 3980 if_initname(struct ifnet *ifp, const char *name, int unit) 3981 { 3982 ifp->if_dname = name; 3983 ifp->if_dunit = unit; 3984 if (unit != IF_DUNIT_NONE) 3985 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3986 else 3987 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3988 } 3989 3990 int 3991 if_printf(struct ifnet *ifp, const char *fmt, ...) 3992 { 3993 char if_fmt[256]; 3994 va_list ap; 3995 3996 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 3997 va_start(ap, fmt); 3998 vlog(LOG_INFO, if_fmt, ap); 3999 va_end(ap); 4000 return (0); 4001 } 4002 4003 void 4004 if_start(struct ifnet *ifp) 4005 { 4006 4007 (*(ifp)->if_start)(ifp); 4008 } 4009 4010 /* 4011 * Backwards compatibility interface for drivers 4012 * that have not implemented it 4013 */ 4014 static int 4015 if_transmit(struct ifnet *ifp, struct mbuf *m) 4016 { 4017 int error; 4018 4019 IFQ_HANDOFF(ifp, m, error); 4020 return (error); 4021 } 4022 4023 static void 4024 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4025 { 4026 4027 m_freem(m); 4028 } 4029 4030 int 4031 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4032 { 4033 int active = 0; 4034 4035 IF_LOCK(ifq); 4036 if (_IF_QFULL(ifq)) { 4037 IF_UNLOCK(ifq); 4038 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4039 m_freem(m); 4040 return (0); 4041 } 4042 if (ifp != NULL) { 4043 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4044 if (m->m_flags & (M_BCAST|M_MCAST)) 4045 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4046 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4047 } 4048 _IF_ENQUEUE(ifq, m); 4049 IF_UNLOCK(ifq); 4050 if (ifp != NULL && !active) 4051 (*(ifp)->if_start)(ifp); 4052 return (1); 4053 } 4054 4055 void 4056 if_register_com_alloc(u_char type, 4057 if_com_alloc_t *a, if_com_free_t *f) 4058 { 4059 4060 KASSERT(if_com_alloc[type] == NULL, 4061 ("if_register_com_alloc: %d already registered", type)); 4062 KASSERT(if_com_free[type] == NULL, 4063 ("if_register_com_alloc: %d free already registered", type)); 4064 4065 if_com_alloc[type] = a; 4066 if_com_free[type] = f; 4067 } 4068 4069 void 4070 if_deregister_com_alloc(u_char type) 4071 { 4072 4073 KASSERT(if_com_alloc[type] != NULL, 4074 ("if_deregister_com_alloc: %d not registered", type)); 4075 KASSERT(if_com_free[type] != NULL, 4076 ("if_deregister_com_alloc: %d free not registered", type)); 4077 if_com_alloc[type] = NULL; 4078 if_com_free[type] = NULL; 4079 } 4080 4081 /* API for driver access to network stack owned ifnet.*/ 4082 uint64_t 4083 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4084 { 4085 uint64_t oldbrate; 4086 4087 oldbrate = ifp->if_baudrate; 4088 ifp->if_baudrate = baudrate; 4089 return (oldbrate); 4090 } 4091 4092 uint64_t 4093 if_getbaudrate(if_t ifp) 4094 { 4095 4096 return (((struct ifnet *)ifp)->if_baudrate); 4097 } 4098 4099 int 4100 if_setcapabilities(if_t ifp, int capabilities) 4101 { 4102 ((struct ifnet *)ifp)->if_capabilities = capabilities; 4103 return (0); 4104 } 4105 4106 int 4107 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4108 { 4109 ((struct ifnet *)ifp)->if_capabilities |= setbit; 4110 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 4111 4112 return (0); 4113 } 4114 4115 int 4116 if_getcapabilities(if_t ifp) 4117 { 4118 return ((struct ifnet *)ifp)->if_capabilities; 4119 } 4120 4121 int 4122 if_setcapenable(if_t ifp, int capabilities) 4123 { 4124 ((struct ifnet *)ifp)->if_capenable = capabilities; 4125 return (0); 4126 } 4127 4128 int 4129 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4130 { 4131 if(setcap) 4132 ((struct ifnet *)ifp)->if_capenable |= setcap; 4133 if(clearcap) 4134 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 4135 4136 return (0); 4137 } 4138 4139 const char * 4140 if_getdname(if_t ifp) 4141 { 4142 return ((struct ifnet *)ifp)->if_dname; 4143 } 4144 4145 int 4146 if_togglecapenable(if_t ifp, int togglecap) 4147 { 4148 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 4149 return (0); 4150 } 4151 4152 int 4153 if_getcapenable(if_t ifp) 4154 { 4155 return ((struct ifnet *)ifp)->if_capenable; 4156 } 4157 4158 /* 4159 * This is largely undesirable because it ties ifnet to a device, but does 4160 * provide flexiblity for an embedded product vendor. Should be used with 4161 * the understanding that it violates the interface boundaries, and should be 4162 * a last resort only. 4163 */ 4164 int 4165 if_setdev(if_t ifp, void *dev) 4166 { 4167 return (0); 4168 } 4169 4170 int 4171 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4172 { 4173 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 4174 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 4175 4176 return (0); 4177 } 4178 4179 int 4180 if_getdrvflags(if_t ifp) 4181 { 4182 return ((struct ifnet *)ifp)->if_drv_flags; 4183 } 4184 4185 int 4186 if_setdrvflags(if_t ifp, int flags) 4187 { 4188 ((struct ifnet *)ifp)->if_drv_flags = flags; 4189 return (0); 4190 } 4191 4192 4193 int 4194 if_setflags(if_t ifp, int flags) 4195 { 4196 ((struct ifnet *)ifp)->if_flags = flags; 4197 return (0); 4198 } 4199 4200 int 4201 if_setflagbits(if_t ifp, int set, int clear) 4202 { 4203 ((struct ifnet *)ifp)->if_flags |= set; 4204 ((struct ifnet *)ifp)->if_flags &= ~clear; 4205 4206 return (0); 4207 } 4208 4209 int 4210 if_getflags(if_t ifp) 4211 { 4212 return ((struct ifnet *)ifp)->if_flags; 4213 } 4214 4215 int 4216 if_clearhwassist(if_t ifp) 4217 { 4218 ((struct ifnet *)ifp)->if_hwassist = 0; 4219 return (0); 4220 } 4221 4222 int 4223 if_sethwassistbits(if_t ifp, int toset, int toclear) 4224 { 4225 ((struct ifnet *)ifp)->if_hwassist |= toset; 4226 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4227 4228 return (0); 4229 } 4230 4231 int 4232 if_sethwassist(if_t ifp, int hwassist_bit) 4233 { 4234 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4235 return (0); 4236 } 4237 4238 int 4239 if_gethwassist(if_t ifp) 4240 { 4241 return ((struct ifnet *)ifp)->if_hwassist; 4242 } 4243 4244 int 4245 if_setmtu(if_t ifp, int mtu) 4246 { 4247 ((struct ifnet *)ifp)->if_mtu = mtu; 4248 return (0); 4249 } 4250 4251 int 4252 if_getmtu(if_t ifp) 4253 { 4254 return ((struct ifnet *)ifp)->if_mtu; 4255 } 4256 4257 int 4258 if_getmtu_family(if_t ifp, int family) 4259 { 4260 struct domain *dp; 4261 4262 for (dp = domains; dp; dp = dp->dom_next) { 4263 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4264 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4265 } 4266 4267 return (((struct ifnet *)ifp)->if_mtu); 4268 } 4269 4270 int 4271 if_setsoftc(if_t ifp, void *softc) 4272 { 4273 ((struct ifnet *)ifp)->if_softc = softc; 4274 return (0); 4275 } 4276 4277 void * 4278 if_getsoftc(if_t ifp) 4279 { 4280 return ((struct ifnet *)ifp)->if_softc; 4281 } 4282 4283 void 4284 if_setrcvif(struct mbuf *m, if_t ifp) 4285 { 4286 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4287 } 4288 4289 void 4290 if_setvtag(struct mbuf *m, uint16_t tag) 4291 { 4292 m->m_pkthdr.ether_vtag = tag; 4293 } 4294 4295 uint16_t 4296 if_getvtag(struct mbuf *m) 4297 { 4298 4299 return (m->m_pkthdr.ether_vtag); 4300 } 4301 4302 int 4303 if_sendq_empty(if_t ifp) 4304 { 4305 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4306 } 4307 4308 struct ifaddr * 4309 if_getifaddr(if_t ifp) 4310 { 4311 return ((struct ifnet *)ifp)->if_addr; 4312 } 4313 4314 int 4315 if_getamcount(if_t ifp) 4316 { 4317 return ((struct ifnet *)ifp)->if_amcount; 4318 } 4319 4320 4321 int 4322 if_setsendqready(if_t ifp) 4323 { 4324 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4325 return (0); 4326 } 4327 4328 int 4329 if_setsendqlen(if_t ifp, int tx_desc_count) 4330 { 4331 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4332 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4333 4334 return (0); 4335 } 4336 4337 int 4338 if_vlantrunkinuse(if_t ifp) 4339 { 4340 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4341 } 4342 4343 int 4344 if_input(if_t ifp, struct mbuf* sendmp) 4345 { 4346 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4347 return (0); 4348 4349 } 4350 4351 /* XXX */ 4352 #ifndef ETH_ADDR_LEN 4353 #define ETH_ADDR_LEN 6 4354 #endif 4355 4356 int 4357 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max) 4358 { 4359 struct ifmultiaddr *ifma; 4360 uint8_t *lmta = (uint8_t *)mta; 4361 int mcnt = 0; 4362 4363 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4364 if (ifma->ifma_addr->sa_family != AF_LINK) 4365 continue; 4366 4367 if (mcnt == max) 4368 break; 4369 4370 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 4371 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 4372 mcnt++; 4373 } 4374 *cnt = mcnt; 4375 4376 return (0); 4377 } 4378 4379 int 4380 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max) 4381 { 4382 int error; 4383 4384 if_maddr_rlock(ifp); 4385 error = if_setupmultiaddr(ifp, mta, cnt, max); 4386 if_maddr_runlock(ifp); 4387 return (error); 4388 } 4389 4390 int 4391 if_multiaddr_count(if_t ifp, int max) 4392 { 4393 struct ifmultiaddr *ifma; 4394 int count; 4395 4396 count = 0; 4397 if_maddr_rlock(ifp); 4398 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4399 if (ifma->ifma_addr->sa_family != AF_LINK) 4400 continue; 4401 count++; 4402 if (count == max) 4403 break; 4404 } 4405 if_maddr_runlock(ifp); 4406 return (count); 4407 } 4408 4409 int 4410 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg) 4411 { 4412 struct ifmultiaddr *ifma; 4413 int cnt = 0; 4414 4415 if_maddr_rlock(ifp); 4416 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4417 cnt += filter(arg, ifma, cnt); 4418 if_maddr_runlock(ifp); 4419 return (cnt); 4420 } 4421 4422 struct mbuf * 4423 if_dequeue(if_t ifp) 4424 { 4425 struct mbuf *m; 4426 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4427 4428 return (m); 4429 } 4430 4431 int 4432 if_sendq_prepend(if_t ifp, struct mbuf *m) 4433 { 4434 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4435 return (0); 4436 } 4437 4438 int 4439 if_setifheaderlen(if_t ifp, int len) 4440 { 4441 ((struct ifnet *)ifp)->if_hdrlen = len; 4442 return (0); 4443 } 4444 4445 caddr_t 4446 if_getlladdr(if_t ifp) 4447 { 4448 return (IF_LLADDR((struct ifnet *)ifp)); 4449 } 4450 4451 void * 4452 if_gethandle(u_char type) 4453 { 4454 return (if_alloc(type)); 4455 } 4456 4457 void 4458 if_bpfmtap(if_t ifh, struct mbuf *m) 4459 { 4460 struct ifnet *ifp = (struct ifnet *)ifh; 4461 4462 BPF_MTAP(ifp, m); 4463 } 4464 4465 void 4466 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4467 { 4468 struct ifnet *ifp = (struct ifnet *)ifh; 4469 4470 ETHER_BPF_MTAP(ifp, m); 4471 } 4472 4473 void 4474 if_vlancap(if_t ifh) 4475 { 4476 struct ifnet *ifp = (struct ifnet *)ifh; 4477 VLAN_CAPABILITIES(ifp); 4478 } 4479 4480 int 4481 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4482 { 4483 4484 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax; 4485 return (0); 4486 } 4487 4488 int 4489 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4490 { 4491 4492 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4493 return (0); 4494 } 4495 4496 int 4497 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4498 { 4499 4500 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4501 return (0); 4502 } 4503 4504 u_int 4505 if_gethwtsomax(if_t ifp) 4506 { 4507 4508 return (((struct ifnet *)ifp)->if_hw_tsomax); 4509 } 4510 4511 u_int 4512 if_gethwtsomaxsegcount(if_t ifp) 4513 { 4514 4515 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount); 4516 } 4517 4518 u_int 4519 if_gethwtsomaxsegsize(if_t ifp) 4520 { 4521 4522 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize); 4523 } 4524 4525 void 4526 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4527 { 4528 ((struct ifnet *)ifp)->if_init = init_fn; 4529 } 4530 4531 void 4532 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4533 { 4534 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4535 } 4536 4537 void 4538 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4539 { 4540 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4541 } 4542 4543 void 4544 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4545 { 4546 ((struct ifnet *)ifp)->if_transmit = start_fn; 4547 } 4548 4549 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4550 { 4551 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4552 4553 } 4554 4555 void 4556 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4557 { 4558 4559 ifp->if_get_counter = fn; 4560 } 4561 4562 /* Revisit these - These are inline functions originally. */ 4563 int 4564 drbr_inuse_drv(if_t ifh, struct buf_ring *br) 4565 { 4566 return drbr_inuse(ifh, br); 4567 } 4568 4569 struct mbuf* 4570 drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 4571 { 4572 return drbr_dequeue(ifh, br); 4573 } 4574 4575 int 4576 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 4577 { 4578 return drbr_needs_enqueue(ifh, br); 4579 } 4580 4581 int 4582 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 4583 { 4584 return drbr_enqueue(ifh, br, m); 4585 4586 } 4587