1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1980, 1986, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)if.c 8.5 (Berkeley) 1/9/95 32 * $FreeBSD$ 33 */ 34 35 #include "opt_inet6.h" 36 #include "opt_inet.h" 37 38 #include <sys/param.h> 39 #include <sys/types.h> 40 #include <sys/conf.h> 41 #include <sys/malloc.h> 42 #include <sys/sbuf.h> 43 #include <sys/bus.h> 44 #include <sys/epoch.h> 45 #include <sys/mbuf.h> 46 #include <sys/systm.h> 47 #include <sys/priv.h> 48 #include <sys/proc.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/protosw.h> 52 #include <sys/kernel.h> 53 #include <sys/lock.h> 54 #include <sys/refcount.h> 55 #include <sys/module.h> 56 #include <sys/rwlock.h> 57 #include <sys/sockio.h> 58 #include <sys/syslog.h> 59 #include <sys/sysctl.h> 60 #include <sys/sysent.h> 61 #include <sys/taskqueue.h> 62 #include <sys/domain.h> 63 #include <sys/jail.h> 64 #include <sys/priv.h> 65 66 #include <machine/stdarg.h> 67 #include <vm/uma.h> 68 69 #include <net/bpf.h> 70 #include <net/ethernet.h> 71 #include <net/if.h> 72 #include <net/if_arp.h> 73 #include <net/if_clone.h> 74 #include <net/if_dl.h> 75 #include <net/if_types.h> 76 #include <net/if_var.h> 77 #include <net/if_media.h> 78 #include <net/if_vlan_var.h> 79 #include <net/radix.h> 80 #include <net/route.h> 81 #include <net/vnet.h> 82 83 #if defined(INET) || defined(INET6) 84 #include <net/ethernet.h> 85 #include <netinet/in.h> 86 #include <netinet/in_var.h> 87 #include <netinet/ip.h> 88 #include <netinet/ip_carp.h> 89 #ifdef INET 90 #include <netinet/if_ether.h> 91 #include <netinet/netdump/netdump.h> 92 #endif /* INET */ 93 #ifdef INET6 94 #include <netinet6/in6_var.h> 95 #include <netinet6/in6_ifattach.h> 96 #endif /* INET6 */ 97 #endif /* INET || INET6 */ 98 99 #include <security/mac/mac_framework.h> 100 101 /* 102 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 103 * and ifr_ifru when it is used in SIOCGIFCONF. 104 */ 105 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 106 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 107 108 __read_mostly epoch_t net_epoch_preempt; 109 __read_mostly epoch_t net_epoch; 110 #ifdef COMPAT_FREEBSD32 111 #include <sys/mount.h> 112 #include <compat/freebsd32/freebsd32.h> 113 114 struct ifreq_buffer32 { 115 uint32_t length; /* (size_t) */ 116 uint32_t buffer; /* (void *) */ 117 }; 118 119 /* 120 * Interface request structure used for socket 121 * ioctl's. All interface ioctl's must have parameter 122 * definitions which begin with ifr_name. The 123 * remainder may be interface specific. 124 */ 125 struct ifreq32 { 126 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 127 union { 128 struct sockaddr ifru_addr; 129 struct sockaddr ifru_dstaddr; 130 struct sockaddr ifru_broadaddr; 131 struct ifreq_buffer32 ifru_buffer; 132 short ifru_flags[2]; 133 short ifru_index; 134 int ifru_jid; 135 int ifru_metric; 136 int ifru_mtu; 137 int ifru_phys; 138 int ifru_media; 139 uint32_t ifru_data; 140 int ifru_cap[2]; 141 u_int ifru_fib; 142 u_char ifru_vlan_pcp; 143 } ifr_ifru; 144 }; 145 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 146 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 147 __offsetof(struct ifreq32, ifr_ifru)); 148 149 struct ifgroupreq32 { 150 char ifgr_name[IFNAMSIZ]; 151 u_int ifgr_len; 152 union { 153 char ifgru_group[IFNAMSIZ]; 154 uint32_t ifgru_groups; 155 } ifgr_ifgru; 156 }; 157 158 struct ifmediareq32 { 159 char ifm_name[IFNAMSIZ]; 160 int ifm_current; 161 int ifm_mask; 162 int ifm_status; 163 int ifm_active; 164 int ifm_count; 165 uint32_t ifm_ulist; /* (int *) */ 166 }; 167 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 168 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 169 170 #define _CASE_IOC_IFGROUPREQ_32(cmd) \ 171 case _IOC_NEWTYPE((cmd), struct ifgroupreq32): 172 #else /* !COMPAT_FREEBSD32 */ 173 #define _CASE_IOC_IFGROUPREQ_32(cmd) 174 #endif /* !COMPAT_FREEBSD32 */ 175 176 #define CASE_IOC_IFGROUPREQ(cmd) \ 177 _CASE_IOC_IFGROUPREQ_32(cmd) \ 178 case (cmd) 179 180 union ifreq_union { 181 struct ifreq ifr; 182 #ifdef COMPAT_FREEBSD32 183 struct ifreq32 ifr32; 184 #endif 185 }; 186 187 union ifgroupreq_union { 188 struct ifgroupreq ifgr; 189 #ifdef COMPAT_FREEBSD32 190 struct ifgroupreq32 ifgr32; 191 #endif 192 }; 193 194 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 195 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 196 197 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 198 &ifqmaxlen, 0, "max send queue size"); 199 200 /* Log link state change events */ 201 static int log_link_state_change = 1; 202 203 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 204 &log_link_state_change, 0, 205 "log interface link state change events"); 206 207 /* Log promiscuous mode change events */ 208 static int log_promisc_mode_change = 1; 209 210 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 211 &log_promisc_mode_change, 1, 212 "log promiscuous mode change events"); 213 214 /* Interface description */ 215 static unsigned int ifdescr_maxlen = 1024; 216 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 217 &ifdescr_maxlen, 0, 218 "administrative maximum length for interface description"); 219 220 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 221 222 /* global sx for non-critical path ifdescr */ 223 static struct sx ifdescr_sx; 224 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 225 226 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 227 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 228 /* These are external hooks for CARP. */ 229 void (*carp_linkstate_p)(struct ifnet *ifp); 230 void (*carp_demote_adj_p)(int, char *); 231 int (*carp_master_p)(struct ifaddr *); 232 #if defined(INET) || defined(INET6) 233 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 234 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 235 const struct sockaddr *sa); 236 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 237 int (*carp_attach_p)(struct ifaddr *, int); 238 void (*carp_detach_p)(struct ifaddr *, bool); 239 #endif 240 #ifdef INET 241 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 242 #endif 243 #ifdef INET6 244 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 245 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 246 const struct in6_addr *taddr); 247 #endif 248 249 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 250 251 /* 252 * XXX: Style; these should be sorted alphabetically, and unprototyped 253 * static functions should be prototyped. Currently they are sorted by 254 * declaration order. 255 */ 256 static void if_attachdomain(void *); 257 static void if_attachdomain1(struct ifnet *); 258 static int ifconf(u_long, caddr_t); 259 static void *if_grow(void); 260 static void if_input_default(struct ifnet *, struct mbuf *); 261 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 262 static void if_route(struct ifnet *, int flag, int fam); 263 static int if_setflag(struct ifnet *, int, int, int *, int); 264 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 265 static void if_unroute(struct ifnet *, int flag, int fam); 266 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 267 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 268 static void do_link_state_change(void *, int); 269 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 270 static int if_getgroupmembers(struct ifgroupreq *); 271 static void if_delgroups(struct ifnet *); 272 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 273 static int if_detach_internal(struct ifnet *, int, struct if_clone **); 274 #ifdef VIMAGE 275 static void if_vmove(struct ifnet *, struct vnet *); 276 #endif 277 278 #ifdef INET6 279 /* 280 * XXX: declare here to avoid to include many inet6 related files.. 281 * should be more generalized? 282 */ 283 extern void nd6_setmtu(struct ifnet *); 284 #endif 285 286 /* ipsec helper hooks */ 287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 289 290 VNET_DEFINE(int, if_index); 291 int ifqmaxlen = IFQ_MAXLEN; 292 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 293 VNET_DEFINE(struct ifgrouphead, ifg_head); 294 295 VNET_DEFINE_STATIC(int, if_indexlim) = 8; 296 297 /* Table of ifnet by index. */ 298 VNET_DEFINE(struct ifnet **, ifindex_table); 299 300 #define V_if_indexlim VNET(if_indexlim) 301 #define V_ifindex_table VNET(ifindex_table) 302 303 /* 304 * The global network interface list (V_ifnet) and related state (such as 305 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and 306 * an rwlock. Either may be acquired shared to stablize the list, but both 307 * must be acquired writable to modify the list. This model allows us to 308 * both stablize the interface list during interrupt thread processing, but 309 * also to stablize it over long-running ioctls, without introducing priority 310 * inversions and deadlocks. 311 */ 312 struct rwlock ifnet_rwlock; 313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE); 314 struct sx ifnet_sxlock; 315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 316 317 /* 318 * The allocation of network interfaces is a rather non-atomic affair; we 319 * need to select an index before we are ready to expose the interface for 320 * use, so will use this pointer value to indicate reservation. 321 */ 322 #define IFNET_HOLD (void *)(uintptr_t)(-1) 323 324 static if_com_alloc_t *if_com_alloc[256]; 325 static if_com_free_t *if_com_free[256]; 326 327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 330 331 struct ifnet * 332 ifnet_byindex_locked(u_short idx) 333 { 334 335 if (idx > V_if_index) 336 return (NULL); 337 if (V_ifindex_table[idx] == IFNET_HOLD) 338 return (NULL); 339 return (V_ifindex_table[idx]); 340 } 341 342 struct ifnet * 343 ifnet_byindex(u_short idx) 344 { 345 struct ifnet *ifp; 346 347 ifp = ifnet_byindex_locked(idx); 348 return (ifp); 349 } 350 351 struct ifnet * 352 ifnet_byindex_ref(u_short idx) 353 { 354 struct ifnet *ifp; 355 356 IFNET_RLOCK_NOSLEEP(); 357 ifp = ifnet_byindex_locked(idx); 358 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { 359 IFNET_RUNLOCK_NOSLEEP(); 360 return (NULL); 361 } 362 if_ref(ifp); 363 IFNET_RUNLOCK_NOSLEEP(); 364 return (ifp); 365 } 366 367 /* 368 * Allocate an ifindex array entry; return 0 on success or an error on 369 * failure. 370 */ 371 static u_short 372 ifindex_alloc(void **old) 373 { 374 u_short idx; 375 376 IFNET_WLOCK_ASSERT(); 377 /* 378 * Try to find an empty slot below V_if_index. If we fail, take the 379 * next slot. 380 */ 381 for (idx = 1; idx <= V_if_index; idx++) { 382 if (V_ifindex_table[idx] == NULL) 383 break; 384 } 385 386 /* Catch if_index overflow. */ 387 if (idx >= V_if_indexlim) { 388 *old = if_grow(); 389 return (USHRT_MAX); 390 } 391 if (idx > V_if_index) 392 V_if_index = idx; 393 return (idx); 394 } 395 396 static void 397 ifindex_free_locked(u_short idx) 398 { 399 400 IFNET_WLOCK_ASSERT(); 401 402 V_ifindex_table[idx] = NULL; 403 while (V_if_index > 0 && 404 V_ifindex_table[V_if_index] == NULL) 405 V_if_index--; 406 } 407 408 static void 409 ifindex_free(u_short idx) 410 { 411 412 IFNET_WLOCK(); 413 ifindex_free_locked(idx); 414 IFNET_WUNLOCK(); 415 } 416 417 static void 418 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 419 { 420 421 V_ifindex_table[idx] = ifp; 422 } 423 424 struct ifaddr * 425 ifaddr_byindex(u_short idx) 426 { 427 struct ifnet *ifp; 428 struct ifaddr *ifa = NULL; 429 430 IFNET_RLOCK_NOSLEEP(); 431 ifp = ifnet_byindex_locked(idx); 432 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 433 ifa_ref(ifa); 434 IFNET_RUNLOCK_NOSLEEP(); 435 return (ifa); 436 } 437 438 /* 439 * Network interface utility routines. 440 * 441 * Routines with ifa_ifwith* names take sockaddr *'s as 442 * parameters. 443 */ 444 445 static void 446 vnet_if_init(const void *unused __unused) 447 { 448 void *old; 449 450 CK_STAILQ_INIT(&V_ifnet); 451 CK_STAILQ_INIT(&V_ifg_head); 452 IFNET_WLOCK(); 453 old = if_grow(); /* create initial table */ 454 IFNET_WUNLOCK(); 455 epoch_wait_preempt(net_epoch_preempt); 456 free(old, M_IFNET); 457 vnet_if_clone_init(); 458 } 459 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 460 NULL); 461 462 #ifdef VIMAGE 463 static void 464 vnet_if_uninit(const void *unused __unused) 465 { 466 467 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 468 "not empty", __func__, __LINE__, &V_ifnet)); 469 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 470 "not empty", __func__, __LINE__, &V_ifg_head)); 471 472 free((caddr_t)V_ifindex_table, M_IFNET); 473 } 474 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 475 vnet_if_uninit, NULL); 476 477 static void 478 vnet_if_return(const void *unused __unused) 479 { 480 struct ifnet *ifp, *nifp; 481 482 /* Return all inherited interfaces to their parent vnets. */ 483 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 484 if (ifp->if_home_vnet != ifp->if_vnet) 485 if_vmove(ifp, ifp->if_home_vnet); 486 } 487 } 488 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 489 vnet_if_return, NULL); 490 #endif 491 492 493 static void * 494 if_grow(void) 495 { 496 int oldlim; 497 u_int n; 498 struct ifnet **e; 499 void *old; 500 501 old = NULL; 502 IFNET_WLOCK_ASSERT(); 503 oldlim = V_if_indexlim; 504 IFNET_WUNLOCK(); 505 n = (oldlim << 1) * sizeof(*e); 506 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 507 IFNET_WLOCK(); 508 if (V_if_indexlim != oldlim) { 509 free(e, M_IFNET); 510 return (NULL); 511 } 512 if (V_ifindex_table != NULL) { 513 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 514 old = V_ifindex_table; 515 } 516 V_if_indexlim <<= 1; 517 V_ifindex_table = e; 518 return (old); 519 } 520 521 /* 522 * Allocate a struct ifnet and an index for an interface. A layer 2 523 * common structure will also be allocated if an allocation routine is 524 * registered for the passed type. 525 */ 526 struct ifnet * 527 if_alloc(u_char type) 528 { 529 struct ifnet *ifp; 530 u_short idx; 531 void *old; 532 533 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); 534 restart: 535 IFNET_WLOCK(); 536 idx = ifindex_alloc(&old); 537 if (__predict_false(idx == USHRT_MAX)) { 538 IFNET_WUNLOCK(); 539 epoch_wait_preempt(net_epoch_preempt); 540 free(old, M_IFNET); 541 goto restart; 542 } 543 ifnet_setbyindex(idx, IFNET_HOLD); 544 IFNET_WUNLOCK(); 545 ifp->if_index = idx; 546 ifp->if_type = type; 547 ifp->if_alloctype = type; 548 #ifdef VIMAGE 549 ifp->if_vnet = curvnet; 550 #endif 551 if (if_com_alloc[type] != NULL) { 552 ifp->if_l2com = if_com_alloc[type](type, ifp); 553 if (ifp->if_l2com == NULL) { 554 free(ifp, M_IFNET); 555 ifindex_free(idx); 556 return (NULL); 557 } 558 } 559 560 IF_ADDR_LOCK_INIT(ifp); 561 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 562 ifp->if_afdata_initialized = 0; 563 IF_AFDATA_LOCK_INIT(ifp); 564 CK_STAILQ_INIT(&ifp->if_addrhead); 565 CK_STAILQ_INIT(&ifp->if_multiaddrs); 566 CK_STAILQ_INIT(&ifp->if_groups); 567 #ifdef MAC 568 mac_ifnet_init(ifp); 569 #endif 570 ifq_init(&ifp->if_snd, ifp); 571 572 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 573 for (int i = 0; i < IFCOUNTERS; i++) 574 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 575 ifp->if_get_counter = if_get_counter_default; 576 ifp->if_pcp = IFNET_PCP_NONE; 577 ifnet_setbyindex(ifp->if_index, ifp); 578 return (ifp); 579 } 580 581 /* 582 * Do the actual work of freeing a struct ifnet, and layer 2 common 583 * structure. This call is made when the last reference to an 584 * interface is released. 585 */ 586 static void 587 if_free_internal(struct ifnet *ifp) 588 { 589 590 KASSERT((ifp->if_flags & IFF_DYING), 591 ("if_free_internal: interface not dying")); 592 593 if (if_com_free[ifp->if_alloctype] != NULL) 594 if_com_free[ifp->if_alloctype](ifp->if_l2com, 595 ifp->if_alloctype); 596 597 #ifdef MAC 598 mac_ifnet_destroy(ifp); 599 #endif /* MAC */ 600 if (ifp->if_description != NULL) 601 free(ifp->if_description, M_IFDESCR); 602 IF_AFDATA_DESTROY(ifp); 603 IF_ADDR_LOCK_DESTROY(ifp); 604 ifq_delete(&ifp->if_snd); 605 606 for (int i = 0; i < IFCOUNTERS; i++) 607 counter_u64_free(ifp->if_counters[i]); 608 609 free(ifp, M_IFNET); 610 } 611 612 static void 613 if_destroy(epoch_context_t ctx) 614 { 615 struct ifnet *ifp; 616 617 ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 618 if_free_internal(ifp); 619 } 620 621 /* 622 * Deregister an interface and free the associated storage. 623 */ 624 void 625 if_free(struct ifnet *ifp) 626 { 627 628 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 629 630 CURVNET_SET_QUIET(ifp->if_vnet); 631 IFNET_WLOCK(); 632 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), 633 ("%s: freeing unallocated ifnet", ifp->if_xname)); 634 635 ifindex_free_locked(ifp->if_index); 636 IFNET_WUNLOCK(); 637 638 if (refcount_release(&ifp->if_refcount)) 639 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy); 640 CURVNET_RESTORE(); 641 } 642 643 /* 644 * Interfaces to keep an ifnet type-stable despite the possibility of the 645 * driver calling if_free(). If there are additional references, we defer 646 * freeing the underlying data structure. 647 */ 648 void 649 if_ref(struct ifnet *ifp) 650 { 651 652 /* We don't assert the ifnet list lock here, but arguably should. */ 653 refcount_acquire(&ifp->if_refcount); 654 } 655 656 void 657 if_rele(struct ifnet *ifp) 658 { 659 660 if (!refcount_release(&ifp->if_refcount)) 661 return; 662 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy); 663 } 664 665 void 666 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 667 { 668 669 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 670 671 if (ifq->ifq_maxlen == 0) 672 ifq->ifq_maxlen = ifqmaxlen; 673 674 ifq->altq_type = 0; 675 ifq->altq_disc = NULL; 676 ifq->altq_flags &= ALTQF_CANTCHANGE; 677 ifq->altq_tbr = NULL; 678 ifq->altq_ifp = ifp; 679 } 680 681 void 682 ifq_delete(struct ifaltq *ifq) 683 { 684 mtx_destroy(&ifq->ifq_mtx); 685 } 686 687 /* 688 * Perform generic interface initialization tasks and attach the interface 689 * to the list of "active" interfaces. If vmove flag is set on entry 690 * to if_attach_internal(), perform only a limited subset of initialization 691 * tasks, given that we are moving from one vnet to another an ifnet which 692 * has already been fully initialized. 693 * 694 * Note that if_detach_internal() removes group membership unconditionally 695 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 696 * Thus, when if_vmove() is applied to a cloned interface, group membership 697 * is lost while a cloned one always joins a group whose name is 698 * ifc->ifc_name. To recover this after if_detach_internal() and 699 * if_attach_internal(), the cloner should be specified to 700 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 701 * attempts to join a group whose name is ifc->ifc_name. 702 * 703 * XXX: 704 * - The decision to return void and thus require this function to 705 * succeed is questionable. 706 * - We should probably do more sanity checking. For instance we don't 707 * do anything to insure if_xname is unique or non-empty. 708 */ 709 void 710 if_attach(struct ifnet *ifp) 711 { 712 713 if_attach_internal(ifp, 0, NULL); 714 } 715 716 /* 717 * Compute the least common TSO limit. 718 */ 719 void 720 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 721 { 722 /* 723 * 1) If there is no limit currently, take the limit from 724 * the network adapter. 725 * 726 * 2) If the network adapter has a limit below the current 727 * limit, apply it. 728 */ 729 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 730 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 731 pmax->tsomaxbytes = ifp->if_hw_tsomax; 732 } 733 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 734 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 735 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 736 } 737 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 738 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 739 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 740 } 741 } 742 743 /* 744 * Update TSO limit of a network adapter. 745 * 746 * Returns zero if no change. Else non-zero. 747 */ 748 int 749 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 750 { 751 int retval = 0; 752 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 753 ifp->if_hw_tsomax = pmax->tsomaxbytes; 754 retval++; 755 } 756 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 757 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 758 retval++; 759 } 760 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 761 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 762 retval++; 763 } 764 return (retval); 765 } 766 767 static void 768 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 769 { 770 unsigned socksize, ifasize; 771 int namelen, masklen; 772 struct sockaddr_dl *sdl; 773 struct ifaddr *ifa; 774 775 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 776 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 777 ifp->if_xname); 778 779 #ifdef VIMAGE 780 ifp->if_vnet = curvnet; 781 if (ifp->if_home_vnet == NULL) 782 ifp->if_home_vnet = curvnet; 783 #endif 784 785 if_addgroup(ifp, IFG_ALL); 786 787 /* Restore group membership for cloned interfaces. */ 788 if (vmove && ifc != NULL) 789 if_clone_addgroup(ifp, ifc); 790 791 getmicrotime(&ifp->if_lastchange); 792 ifp->if_epoch = time_uptime; 793 794 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 795 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 796 ("transmit and qflush must both either be set or both be NULL")); 797 if (ifp->if_transmit == NULL) { 798 ifp->if_transmit = if_transmit; 799 ifp->if_qflush = if_qflush; 800 } 801 if (ifp->if_input == NULL) 802 ifp->if_input = if_input_default; 803 804 if (ifp->if_requestencap == NULL) 805 ifp->if_requestencap = if_requestencap_default; 806 807 if (!vmove) { 808 #ifdef MAC 809 mac_ifnet_create(ifp); 810 #endif 811 812 /* 813 * Create a Link Level name for this device. 814 */ 815 namelen = strlen(ifp->if_xname); 816 /* 817 * Always save enough space for any possiable name so we 818 * can do a rename in place later. 819 */ 820 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 821 socksize = masklen + ifp->if_addrlen; 822 if (socksize < sizeof(*sdl)) 823 socksize = sizeof(*sdl); 824 socksize = roundup2(socksize, sizeof(long)); 825 ifasize = sizeof(*ifa) + 2 * socksize; 826 ifa = ifa_alloc(ifasize, M_WAITOK); 827 sdl = (struct sockaddr_dl *)(ifa + 1); 828 sdl->sdl_len = socksize; 829 sdl->sdl_family = AF_LINK; 830 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 831 sdl->sdl_nlen = namelen; 832 sdl->sdl_index = ifp->if_index; 833 sdl->sdl_type = ifp->if_type; 834 ifp->if_addr = ifa; 835 ifa->ifa_ifp = ifp; 836 ifa->ifa_rtrequest = link_rtrequest; 837 ifa->ifa_addr = (struct sockaddr *)sdl; 838 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 839 ifa->ifa_netmask = (struct sockaddr *)sdl; 840 sdl->sdl_len = masklen; 841 while (namelen != 0) 842 sdl->sdl_data[--namelen] = 0xff; 843 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 844 /* Reliably crash if used uninitialized. */ 845 ifp->if_broadcastaddr = NULL; 846 847 if (ifp->if_type == IFT_ETHER) { 848 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 849 M_WAITOK | M_ZERO); 850 } 851 852 #if defined(INET) || defined(INET6) 853 /* Use defaults for TSO, if nothing is set */ 854 if (ifp->if_hw_tsomax == 0 && 855 ifp->if_hw_tsomaxsegcount == 0 && 856 ifp->if_hw_tsomaxsegsize == 0) { 857 /* 858 * The TSO defaults needs to be such that an 859 * NFS mbuf list of 35 mbufs totalling just 860 * below 64K works and that a chain of mbufs 861 * can be defragged into at most 32 segments: 862 */ 863 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 864 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 865 ifp->if_hw_tsomaxsegcount = 35; 866 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 867 868 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 869 if (ifp->if_capabilities & IFCAP_TSO) { 870 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 871 ifp->if_hw_tsomax, 872 ifp->if_hw_tsomaxsegcount, 873 ifp->if_hw_tsomaxsegsize); 874 } 875 } 876 #endif 877 } 878 #ifdef VIMAGE 879 else { 880 /* 881 * Update the interface index in the link layer address 882 * of the interface. 883 */ 884 for (ifa = ifp->if_addr; ifa != NULL; 885 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 886 if (ifa->ifa_addr->sa_family == AF_LINK) { 887 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 888 sdl->sdl_index = ifp->if_index; 889 } 890 } 891 } 892 #endif 893 894 IFNET_WLOCK(); 895 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 896 #ifdef VIMAGE 897 curvnet->vnet_ifcnt++; 898 #endif 899 IFNET_WUNLOCK(); 900 901 if (domain_init_status >= 2) 902 if_attachdomain1(ifp); 903 904 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 905 if (IS_DEFAULT_VNET(curvnet)) 906 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 907 908 /* Announce the interface. */ 909 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 910 } 911 912 static void 913 if_epochalloc(void *dummy __unused) 914 { 915 916 net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT); 917 net_epoch = epoch_alloc(0); 918 } 919 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY, 920 if_epochalloc, NULL); 921 922 static void 923 if_attachdomain(void *dummy) 924 { 925 struct ifnet *ifp; 926 927 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 928 if_attachdomain1(ifp); 929 } 930 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 931 if_attachdomain, NULL); 932 933 static void 934 if_attachdomain1(struct ifnet *ifp) 935 { 936 struct domain *dp; 937 938 /* 939 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 940 * cannot lock ifp->if_afdata initialization, entirely. 941 */ 942 IF_AFDATA_LOCK(ifp); 943 if (ifp->if_afdata_initialized >= domain_init_status) { 944 IF_AFDATA_UNLOCK(ifp); 945 log(LOG_WARNING, "%s called more than once on %s\n", 946 __func__, ifp->if_xname); 947 return; 948 } 949 ifp->if_afdata_initialized = domain_init_status; 950 IF_AFDATA_UNLOCK(ifp); 951 952 /* address family dependent data region */ 953 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 954 for (dp = domains; dp; dp = dp->dom_next) { 955 if (dp->dom_ifattach) 956 ifp->if_afdata[dp->dom_family] = 957 (*dp->dom_ifattach)(ifp); 958 } 959 } 960 961 /* 962 * Remove any unicast or broadcast network addresses from an interface. 963 */ 964 void 965 if_purgeaddrs(struct ifnet *ifp) 966 { 967 struct ifaddr *ifa; 968 969 while (1) { 970 NET_EPOCH_ENTER(); 971 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 972 if (ifa->ifa_addr->sa_family != AF_LINK) 973 break; 974 } 975 NET_EPOCH_EXIT(); 976 977 if (ifa == NULL) 978 break; 979 #ifdef INET 980 /* XXX: Ugly!! ad hoc just for INET */ 981 if (ifa->ifa_addr->sa_family == AF_INET) { 982 struct ifaliasreq ifr; 983 984 bzero(&ifr, sizeof(ifr)); 985 ifr.ifra_addr = *ifa->ifa_addr; 986 if (ifa->ifa_dstaddr) 987 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 988 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 989 NULL) == 0) 990 continue; 991 } 992 #endif /* INET */ 993 #ifdef INET6 994 if (ifa->ifa_addr->sa_family == AF_INET6) { 995 in6_purgeaddr(ifa); 996 /* ifp_addrhead is already updated */ 997 continue; 998 } 999 #endif /* INET6 */ 1000 IF_ADDR_WLOCK(ifp); 1001 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1002 IF_ADDR_WUNLOCK(ifp); 1003 ifa_free(ifa); 1004 } 1005 } 1006 1007 /* 1008 * Remove any multicast network addresses from an interface when an ifnet 1009 * is going away. 1010 */ 1011 static void 1012 if_purgemaddrs(struct ifnet *ifp) 1013 { 1014 struct ifmultiaddr *ifma; 1015 1016 IF_ADDR_WLOCK(ifp); 1017 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1018 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1019 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1020 if_delmulti_locked(ifp, ifma, 1); 1021 } 1022 IF_ADDR_WUNLOCK(ifp); 1023 } 1024 1025 /* 1026 * Detach an interface, removing it from the list of "active" interfaces. 1027 * If vmove flag is set on entry to if_detach_internal(), perform only a 1028 * limited subset of cleanup tasks, given that we are moving an ifnet from 1029 * one vnet to another, where it must be fully operational. 1030 * 1031 * XXXRW: There are some significant questions about event ordering, and 1032 * how to prevent things from starting to use the interface during detach. 1033 */ 1034 void 1035 if_detach(struct ifnet *ifp) 1036 { 1037 1038 CURVNET_SET_QUIET(ifp->if_vnet); 1039 if_detach_internal(ifp, 0, NULL); 1040 CURVNET_RESTORE(); 1041 } 1042 1043 /* 1044 * The vmove flag, if set, indicates that we are called from a callpath 1045 * that is moving an interface to a different vnet instance. 1046 * 1047 * The shutdown flag, if set, indicates that we are called in the 1048 * process of shutting down a vnet instance. Currently only the 1049 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1050 * on a vnet instance shutdown without this flag being set, e.g., when 1051 * the cloned interfaces are destoyed as first thing of teardown. 1052 */ 1053 static int 1054 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 1055 { 1056 struct ifaddr *ifa; 1057 int i; 1058 struct domain *dp; 1059 struct ifnet *iter; 1060 int found = 0; 1061 #ifdef VIMAGE 1062 int shutdown; 1063 1064 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1065 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1066 #endif 1067 IFNET_WLOCK(); 1068 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 1069 if (iter == ifp) { 1070 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 1071 if (!vmove) 1072 ifp->if_flags |= IFF_DYING; 1073 found = 1; 1074 break; 1075 } 1076 IFNET_WUNLOCK(); 1077 if (!found) { 1078 /* 1079 * While we would want to panic here, we cannot 1080 * guarantee that the interface is indeed still on 1081 * the list given we don't hold locks all the way. 1082 */ 1083 return (ENOENT); 1084 #if 0 1085 if (vmove) 1086 panic("%s: ifp=%p not on the ifnet tailq %p", 1087 __func__, ifp, &V_ifnet); 1088 else 1089 return; /* XXX this should panic as well? */ 1090 #endif 1091 } 1092 1093 /* 1094 * At this point we know the interface still was on the ifnet list 1095 * and we removed it so we are in a stable state. 1096 */ 1097 #ifdef VIMAGE 1098 curvnet->vnet_ifcnt--; 1099 #endif 1100 epoch_wait_preempt(net_epoch_preempt); 1101 /* 1102 * In any case (destroy or vmove) detach us from the groups 1103 * and remove/wait for pending events on the taskq. 1104 * XXX-BZ in theory an interface could still enqueue a taskq change? 1105 */ 1106 if_delgroups(ifp); 1107 1108 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1109 1110 /* 1111 * Check if this is a cloned interface or not. Must do even if 1112 * shutting down as a if_vmove_reclaim() would move the ifp and 1113 * the if_clone_addgroup() will have a corrupted string overwise 1114 * from a gibberish pointer. 1115 */ 1116 if (vmove && ifcp != NULL) 1117 *ifcp = if_clone_findifc(ifp); 1118 1119 if_down(ifp); 1120 1121 #ifdef VIMAGE 1122 /* 1123 * On VNET shutdown abort here as the stack teardown will do all 1124 * the work top-down for us. 1125 */ 1126 if (shutdown) { 1127 /* Give interface users the chance to clean up. */ 1128 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1129 1130 /* 1131 * In case of a vmove we are done here without error. 1132 * If we would signal an error it would lead to the same 1133 * abort as if we did not find the ifnet anymore. 1134 * if_detach() calls us in void context and does not care 1135 * about an early abort notification, so life is splendid :) 1136 */ 1137 goto finish_vnet_shutdown; 1138 } 1139 #endif 1140 1141 /* 1142 * At this point we are not tearing down a VNET and are either 1143 * going to destroy or vmove the interface and have to cleanup 1144 * accordingly. 1145 */ 1146 1147 /* 1148 * Remove routes and flush queues. 1149 */ 1150 #ifdef ALTQ 1151 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1152 altq_disable(&ifp->if_snd); 1153 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1154 altq_detach(&ifp->if_snd); 1155 #endif 1156 1157 if_purgeaddrs(ifp); 1158 1159 #ifdef INET 1160 in_ifdetach(ifp); 1161 #endif 1162 1163 #ifdef INET6 1164 /* 1165 * Remove all IPv6 kernel structs related to ifp. This should be done 1166 * before removing routing entries below, since IPv6 interface direct 1167 * routes are expected to be removed by the IPv6-specific kernel API. 1168 * Otherwise, the kernel will detect some inconsistency and bark it. 1169 */ 1170 in6_ifdetach(ifp); 1171 #endif 1172 if_purgemaddrs(ifp); 1173 1174 /* Announce that the interface is gone. */ 1175 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1176 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1177 if (IS_DEFAULT_VNET(curvnet)) 1178 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1179 1180 if (!vmove) { 1181 /* 1182 * Prevent further calls into the device driver via ifnet. 1183 */ 1184 if_dead(ifp); 1185 1186 /* 1187 * Remove link ifaddr pointer and maybe decrement if_index. 1188 * Clean up all addresses. 1189 */ 1190 free(ifp->if_hw_addr, M_IFADDR); 1191 ifp->if_hw_addr = NULL; 1192 ifp->if_addr = NULL; 1193 1194 /* We can now free link ifaddr. */ 1195 IF_ADDR_WLOCK(ifp); 1196 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1197 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1198 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1199 IF_ADDR_WUNLOCK(ifp); 1200 ifa_free(ifa); 1201 } else 1202 IF_ADDR_WUNLOCK(ifp); 1203 } 1204 1205 rt_flushifroutes(ifp); 1206 1207 #ifdef VIMAGE 1208 finish_vnet_shutdown: 1209 #endif 1210 /* 1211 * We cannot hold the lock over dom_ifdetach calls as they might 1212 * sleep, for example trying to drain a callout, thus open up the 1213 * theoretical race with re-attaching. 1214 */ 1215 IF_AFDATA_LOCK(ifp); 1216 i = ifp->if_afdata_initialized; 1217 ifp->if_afdata_initialized = 0; 1218 IF_AFDATA_UNLOCK(ifp); 1219 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1220 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1221 (*dp->dom_ifdetach)(ifp, 1222 ifp->if_afdata[dp->dom_family]); 1223 ifp->if_afdata[dp->dom_family] = NULL; 1224 } 1225 } 1226 1227 return (0); 1228 } 1229 1230 #ifdef VIMAGE 1231 /* 1232 * if_vmove() performs a limited version of if_detach() in current 1233 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1234 * An attempt is made to shrink if_index in current vnet, find an 1235 * unused if_index in target vnet and calls if_grow() if necessary, 1236 * and finally find an unused if_xname for the target vnet. 1237 */ 1238 static void 1239 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1240 { 1241 struct if_clone *ifc; 1242 u_int bif_dlt, bif_hdrlen; 1243 void *old; 1244 int rc; 1245 1246 /* 1247 * if_detach_internal() will call the eventhandler to notify 1248 * interface departure. That will detach if_bpf. We need to 1249 * safe the dlt and hdrlen so we can re-attach it later. 1250 */ 1251 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1252 1253 /* 1254 * Detach from current vnet, but preserve LLADDR info, do not 1255 * mark as dead etc. so that the ifnet can be reattached later. 1256 * If we cannot find it, we lost the race to someone else. 1257 */ 1258 rc = if_detach_internal(ifp, 1, &ifc); 1259 if (rc != 0) 1260 return; 1261 1262 /* 1263 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1264 * the if_index for that vnet if possible. 1265 * 1266 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1267 * or we'd lock on one vnet and unlock on another. 1268 */ 1269 IFNET_WLOCK(); 1270 ifindex_free_locked(ifp->if_index); 1271 IFNET_WUNLOCK(); 1272 1273 /* 1274 * Perform interface-specific reassignment tasks, if provided by 1275 * the driver. 1276 */ 1277 if (ifp->if_reassign != NULL) 1278 ifp->if_reassign(ifp, new_vnet, NULL); 1279 1280 /* 1281 * Switch to the context of the target vnet. 1282 */ 1283 CURVNET_SET_QUIET(new_vnet); 1284 restart: 1285 IFNET_WLOCK(); 1286 ifp->if_index = ifindex_alloc(&old); 1287 if (__predict_false(ifp->if_index == USHRT_MAX)) { 1288 IFNET_WUNLOCK(); 1289 epoch_wait_preempt(net_epoch_preempt); 1290 free(old, M_IFNET); 1291 goto restart; 1292 } 1293 ifnet_setbyindex(ifp->if_index, ifp); 1294 IFNET_WUNLOCK(); 1295 1296 if_attach_internal(ifp, 1, ifc); 1297 1298 if (ifp->if_bpf == NULL) 1299 bpfattach(ifp, bif_dlt, bif_hdrlen); 1300 1301 CURVNET_RESTORE(); 1302 } 1303 1304 /* 1305 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1306 */ 1307 static int 1308 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1309 { 1310 struct prison *pr; 1311 struct ifnet *difp; 1312 int shutdown; 1313 1314 /* Try to find the prison within our visibility. */ 1315 sx_slock(&allprison_lock); 1316 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1317 sx_sunlock(&allprison_lock); 1318 if (pr == NULL) 1319 return (ENXIO); 1320 prison_hold_locked(pr); 1321 mtx_unlock(&pr->pr_mtx); 1322 1323 /* Do not try to move the iface from and to the same prison. */ 1324 if (pr->pr_vnet == ifp->if_vnet) { 1325 prison_free(pr); 1326 return (EEXIST); 1327 } 1328 1329 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1330 /* XXX Lock interfaces to avoid races. */ 1331 CURVNET_SET_QUIET(pr->pr_vnet); 1332 difp = ifunit(ifname); 1333 if (difp != NULL) { 1334 CURVNET_RESTORE(); 1335 prison_free(pr); 1336 return (EEXIST); 1337 } 1338 1339 /* Make sure the VNET is stable. */ 1340 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1341 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1342 if (shutdown) { 1343 CURVNET_RESTORE(); 1344 prison_free(pr); 1345 return (EBUSY); 1346 } 1347 CURVNET_RESTORE(); 1348 1349 /* Move the interface into the child jail/vnet. */ 1350 if_vmove(ifp, pr->pr_vnet); 1351 1352 /* Report the new if_xname back to the userland. */ 1353 sprintf(ifname, "%s", ifp->if_xname); 1354 1355 prison_free(pr); 1356 return (0); 1357 } 1358 1359 static int 1360 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1361 { 1362 struct prison *pr; 1363 struct vnet *vnet_dst; 1364 struct ifnet *ifp; 1365 int shutdown; 1366 1367 /* Try to find the prison within our visibility. */ 1368 sx_slock(&allprison_lock); 1369 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1370 sx_sunlock(&allprison_lock); 1371 if (pr == NULL) 1372 return (ENXIO); 1373 prison_hold_locked(pr); 1374 mtx_unlock(&pr->pr_mtx); 1375 1376 /* Make sure the named iface exists in the source prison/vnet. */ 1377 CURVNET_SET(pr->pr_vnet); 1378 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1379 if (ifp == NULL) { 1380 CURVNET_RESTORE(); 1381 prison_free(pr); 1382 return (ENXIO); 1383 } 1384 1385 /* Do not try to move the iface from and to the same prison. */ 1386 vnet_dst = TD_TO_VNET(td); 1387 if (vnet_dst == ifp->if_vnet) { 1388 CURVNET_RESTORE(); 1389 prison_free(pr); 1390 return (EEXIST); 1391 } 1392 1393 /* Make sure the VNET is stable. */ 1394 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET && 1395 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 1396 if (shutdown) { 1397 CURVNET_RESTORE(); 1398 prison_free(pr); 1399 return (EBUSY); 1400 } 1401 1402 /* Get interface back from child jail/vnet. */ 1403 if_vmove(ifp, vnet_dst); 1404 CURVNET_RESTORE(); 1405 1406 /* Report the new if_xname back to the userland. */ 1407 sprintf(ifname, "%s", ifp->if_xname); 1408 1409 prison_free(pr); 1410 return (0); 1411 } 1412 #endif /* VIMAGE */ 1413 1414 /* 1415 * Add a group to an interface 1416 */ 1417 int 1418 if_addgroup(struct ifnet *ifp, const char *groupname) 1419 { 1420 struct ifg_list *ifgl; 1421 struct ifg_group *ifg = NULL; 1422 struct ifg_member *ifgm; 1423 int new = 0; 1424 1425 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1426 groupname[strlen(groupname) - 1] <= '9') 1427 return (EINVAL); 1428 1429 IFNET_WLOCK(); 1430 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1431 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1432 IFNET_WUNLOCK(); 1433 return (EEXIST); 1434 } 1435 1436 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, 1437 M_NOWAIT)) == NULL) { 1438 IFNET_WUNLOCK(); 1439 return (ENOMEM); 1440 } 1441 1442 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), 1443 M_TEMP, M_NOWAIT)) == NULL) { 1444 free(ifgl, M_TEMP); 1445 IFNET_WUNLOCK(); 1446 return (ENOMEM); 1447 } 1448 1449 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1450 if (!strcmp(ifg->ifg_group, groupname)) 1451 break; 1452 1453 if (ifg == NULL) { 1454 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), 1455 M_TEMP, M_NOWAIT)) == NULL) { 1456 free(ifgl, M_TEMP); 1457 free(ifgm, M_TEMP); 1458 IFNET_WUNLOCK(); 1459 return (ENOMEM); 1460 } 1461 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1462 ifg->ifg_refcnt = 0; 1463 CK_STAILQ_INIT(&ifg->ifg_members); 1464 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1465 new = 1; 1466 } 1467 1468 ifg->ifg_refcnt++; 1469 ifgl->ifgl_group = ifg; 1470 ifgm->ifgm_ifp = ifp; 1471 1472 IF_ADDR_WLOCK(ifp); 1473 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1474 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1475 IF_ADDR_WUNLOCK(ifp); 1476 1477 IFNET_WUNLOCK(); 1478 1479 if (new) 1480 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1481 EVENTHANDLER_INVOKE(group_change_event, groupname); 1482 1483 return (0); 1484 } 1485 1486 /* 1487 * Remove a group from an interface 1488 */ 1489 int 1490 if_delgroup(struct ifnet *ifp, const char *groupname) 1491 { 1492 struct ifg_list *ifgl; 1493 struct ifg_member *ifgm; 1494 int freeifgl; 1495 1496 IFNET_WLOCK(); 1497 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1498 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) 1499 break; 1500 if (ifgl == NULL) { 1501 IFNET_WUNLOCK(); 1502 return (ENOENT); 1503 } 1504 1505 freeifgl = 0; 1506 IF_ADDR_WLOCK(ifp); 1507 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1508 IF_ADDR_WUNLOCK(ifp); 1509 1510 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1511 if (ifgm->ifgm_ifp == ifp) 1512 break; 1513 1514 if (ifgm != NULL) 1515 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next); 1516 1517 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1518 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next); 1519 freeifgl = 1; 1520 } 1521 IFNET_WUNLOCK(); 1522 1523 epoch_wait_preempt(net_epoch_preempt); 1524 if (freeifgl) { 1525 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1526 free(ifgl->ifgl_group, M_TEMP); 1527 } 1528 free(ifgm, M_TEMP); 1529 free(ifgl, M_TEMP); 1530 1531 EVENTHANDLER_INVOKE(group_change_event, groupname); 1532 1533 return (0); 1534 } 1535 1536 /* 1537 * Remove an interface from all groups 1538 */ 1539 static void 1540 if_delgroups(struct ifnet *ifp) 1541 { 1542 struct ifg_list *ifgl; 1543 struct ifg_member *ifgm; 1544 char groupname[IFNAMSIZ]; 1545 int ifglfree; 1546 1547 IFNET_WLOCK(); 1548 while (!CK_STAILQ_EMPTY(&ifp->if_groups)) { 1549 ifgl = CK_STAILQ_FIRST(&ifp->if_groups); 1550 1551 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1552 1553 IF_ADDR_WLOCK(ifp); 1554 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1555 IF_ADDR_WUNLOCK(ifp); 1556 1557 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) 1558 if (ifgm->ifgm_ifp == ifp) 1559 break; 1560 1561 if (ifgm != NULL) 1562 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, 1563 ifgm_next); 1564 ifglfree = 0; 1565 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1566 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next); 1567 ifglfree = 1; 1568 } 1569 1570 IFNET_WUNLOCK(); 1571 epoch_wait_preempt(net_epoch_preempt); 1572 free(ifgm, M_TEMP); 1573 if (ifglfree) { 1574 EVENTHANDLER_INVOKE(group_detach_event, 1575 ifgl->ifgl_group); 1576 free(ifgl->ifgl_group, M_TEMP); 1577 } 1578 EVENTHANDLER_INVOKE(group_change_event, groupname); 1579 1580 IFNET_WLOCK(); 1581 } 1582 IFNET_WUNLOCK(); 1583 } 1584 1585 static char * 1586 ifgr_group_get(void *ifgrp) 1587 { 1588 union ifgroupreq_union *ifgrup; 1589 1590 ifgrup = ifgrp; 1591 #ifdef COMPAT_FREEBSD32 1592 if (SV_CURPROC_FLAG(SV_ILP32)) 1593 return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]); 1594 #endif 1595 return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]); 1596 } 1597 1598 static struct ifg_req * 1599 ifgr_groups_get(void *ifgrp) 1600 { 1601 union ifgroupreq_union *ifgrup; 1602 1603 ifgrup = ifgrp; 1604 #ifdef COMPAT_FREEBSD32 1605 if (SV_CURPROC_FLAG(SV_ILP32)) 1606 return ((struct ifg_req *)(uintptr_t) 1607 ifgrup->ifgr32.ifgr_ifgru.ifgru_groups); 1608 #endif 1609 return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups); 1610 } 1611 1612 /* 1613 * Stores all groups from an interface in memory pointed to by ifgr. 1614 */ 1615 static int 1616 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1617 { 1618 int len, error; 1619 struct ifg_list *ifgl; 1620 struct ifg_req ifgrq, *ifgp; 1621 1622 if (ifgr->ifgr_len == 0) { 1623 IF_ADDR_RLOCK(ifp); 1624 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1625 ifgr->ifgr_len += sizeof(struct ifg_req); 1626 IF_ADDR_RUNLOCK(ifp); 1627 return (0); 1628 } 1629 1630 len = ifgr->ifgr_len; 1631 ifgp = ifgr_groups_get(ifgr); 1632 /* XXX: wire */ 1633 IF_ADDR_RLOCK(ifp); 1634 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1635 if (len < sizeof(ifgrq)) { 1636 IF_ADDR_RUNLOCK(ifp); 1637 return (EINVAL); 1638 } 1639 bzero(&ifgrq, sizeof ifgrq); 1640 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1641 sizeof(ifgrq.ifgrq_group)); 1642 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1643 IF_ADDR_RUNLOCK(ifp); 1644 return (error); 1645 } 1646 len -= sizeof(ifgrq); 1647 ifgp++; 1648 } 1649 IF_ADDR_RUNLOCK(ifp); 1650 1651 return (0); 1652 } 1653 1654 /* 1655 * Stores all members of a group in memory pointed to by igfr 1656 */ 1657 static int 1658 if_getgroupmembers(struct ifgroupreq *ifgr) 1659 { 1660 struct ifg_group *ifg; 1661 struct ifg_member *ifgm; 1662 struct ifg_req ifgrq, *ifgp; 1663 int len, error; 1664 1665 IFNET_RLOCK(); 1666 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1667 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) 1668 break; 1669 if (ifg == NULL) { 1670 IFNET_RUNLOCK(); 1671 return (ENOENT); 1672 } 1673 1674 if (ifgr->ifgr_len == 0) { 1675 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1676 ifgr->ifgr_len += sizeof(ifgrq); 1677 IFNET_RUNLOCK(); 1678 return (0); 1679 } 1680 1681 len = ifgr->ifgr_len; 1682 ifgp = ifgr_groups_get(ifgr); 1683 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1684 if (len < sizeof(ifgrq)) { 1685 IFNET_RUNLOCK(); 1686 return (EINVAL); 1687 } 1688 bzero(&ifgrq, sizeof ifgrq); 1689 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1690 sizeof(ifgrq.ifgrq_member)); 1691 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1692 IFNET_RUNLOCK(); 1693 return (error); 1694 } 1695 len -= sizeof(ifgrq); 1696 ifgp++; 1697 } 1698 IFNET_RUNLOCK(); 1699 1700 return (0); 1701 } 1702 1703 /* 1704 * Return counter values from counter(9)s stored in ifnet. 1705 */ 1706 uint64_t 1707 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1708 { 1709 1710 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1711 1712 return (counter_u64_fetch(ifp->if_counters[cnt])); 1713 } 1714 1715 /* 1716 * Increase an ifnet counter. Usually used for counters shared 1717 * between the stack and a driver, but function supports them all. 1718 */ 1719 void 1720 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1721 { 1722 1723 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1724 1725 counter_u64_add(ifp->if_counters[cnt], inc); 1726 } 1727 1728 /* 1729 * Copy data from ifnet to userland API structure if_data. 1730 */ 1731 void 1732 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1733 { 1734 1735 ifd->ifi_type = ifp->if_type; 1736 ifd->ifi_physical = 0; 1737 ifd->ifi_addrlen = ifp->if_addrlen; 1738 ifd->ifi_hdrlen = ifp->if_hdrlen; 1739 ifd->ifi_link_state = ifp->if_link_state; 1740 ifd->ifi_vhid = 0; 1741 ifd->ifi_datalen = sizeof(struct if_data); 1742 ifd->ifi_mtu = ifp->if_mtu; 1743 ifd->ifi_metric = ifp->if_metric; 1744 ifd->ifi_baudrate = ifp->if_baudrate; 1745 ifd->ifi_hwassist = ifp->if_hwassist; 1746 ifd->ifi_epoch = ifp->if_epoch; 1747 ifd->ifi_lastchange = ifp->if_lastchange; 1748 1749 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1750 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1751 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1752 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1753 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1754 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1755 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1756 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1757 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1758 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1759 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1760 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1761 } 1762 1763 /* 1764 * Wrapper functions for struct ifnet address list locking macros. These are 1765 * used by kernel modules to avoid encoding programming interface or binary 1766 * interface assumptions that may be violated when kernel-internal locking 1767 * approaches change. 1768 */ 1769 void 1770 if_addr_rlock(struct ifnet *ifp) 1771 { 1772 1773 epoch_enter_preempt(net_epoch_preempt, curthread->td_et); 1774 } 1775 1776 void 1777 if_addr_runlock(struct ifnet *ifp) 1778 { 1779 1780 epoch_exit_preempt(net_epoch_preempt, curthread->td_et); 1781 } 1782 1783 void 1784 if_maddr_rlock(if_t ifp) 1785 { 1786 1787 epoch_enter_preempt(net_epoch_preempt, curthread->td_et); 1788 } 1789 1790 void 1791 if_maddr_runlock(if_t ifp) 1792 { 1793 1794 epoch_exit_preempt(net_epoch_preempt, curthread->td_et); 1795 } 1796 1797 /* 1798 * Initialization, destruction and refcounting functions for ifaddrs. 1799 */ 1800 struct ifaddr * 1801 ifa_alloc(size_t size, int flags) 1802 { 1803 struct ifaddr *ifa; 1804 1805 KASSERT(size >= sizeof(struct ifaddr), 1806 ("%s: invalid size %zu", __func__, size)); 1807 1808 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1809 if (ifa == NULL) 1810 return (NULL); 1811 1812 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1813 goto fail; 1814 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1815 goto fail; 1816 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1817 goto fail; 1818 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1819 goto fail; 1820 1821 refcount_init(&ifa->ifa_refcnt, 1); 1822 1823 return (ifa); 1824 1825 fail: 1826 /* free(NULL) is okay */ 1827 counter_u64_free(ifa->ifa_opackets); 1828 counter_u64_free(ifa->ifa_ipackets); 1829 counter_u64_free(ifa->ifa_obytes); 1830 counter_u64_free(ifa->ifa_ibytes); 1831 free(ifa, M_IFADDR); 1832 1833 return (NULL); 1834 } 1835 1836 void 1837 ifa_ref(struct ifaddr *ifa) 1838 { 1839 1840 refcount_acquire(&ifa->ifa_refcnt); 1841 } 1842 1843 static void 1844 ifa_destroy(epoch_context_t ctx) 1845 { 1846 struct ifaddr *ifa; 1847 1848 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1849 counter_u64_free(ifa->ifa_opackets); 1850 counter_u64_free(ifa->ifa_ipackets); 1851 counter_u64_free(ifa->ifa_obytes); 1852 counter_u64_free(ifa->ifa_ibytes); 1853 free(ifa, M_IFADDR); 1854 } 1855 1856 void 1857 ifa_free(struct ifaddr *ifa) 1858 { 1859 1860 if (refcount_release(&ifa->ifa_refcnt)) 1861 epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy); 1862 } 1863 1864 1865 static int 1866 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa, 1867 struct sockaddr *ia) 1868 { 1869 int error; 1870 struct rt_addrinfo info; 1871 struct sockaddr_dl null_sdl; 1872 struct ifnet *ifp; 1873 1874 ifp = ifa->ifa_ifp; 1875 1876 bzero(&info, sizeof(info)); 1877 if (cmd != RTM_DELETE) 1878 info.rti_ifp = V_loif; 1879 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED; 1880 info.rti_info[RTAX_DST] = ia; 1881 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; 1882 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type); 1883 1884 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib); 1885 1886 if (error != 0 && 1887 !(cmd == RTM_ADD && error == EEXIST) && 1888 !(cmd == RTM_DELETE && error == ENOENT)) 1889 if_printf(ifp, "%s failed: %d\n", otype, error); 1890 1891 return (error); 1892 } 1893 1894 int 1895 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1896 { 1897 1898 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia)); 1899 } 1900 1901 int 1902 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1903 { 1904 1905 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia)); 1906 } 1907 1908 int 1909 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) 1910 { 1911 1912 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia)); 1913 } 1914 1915 /* 1916 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1917 * structs used to represent other address families, it is necessary 1918 * to perform a different comparison. 1919 */ 1920 1921 #define sa_dl_equal(a1, a2) \ 1922 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1923 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1924 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1925 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1926 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1927 1928 /* 1929 * Locate an interface based on a complete address. 1930 */ 1931 /*ARGSUSED*/ 1932 struct ifaddr * 1933 ifa_ifwithaddr(const struct sockaddr *addr) 1934 { 1935 struct ifnet *ifp; 1936 struct ifaddr *ifa; 1937 1938 MPASS(in_epoch(net_epoch_preempt)); 1939 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1940 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1941 if (ifa->ifa_addr->sa_family != addr->sa_family) 1942 continue; 1943 if (sa_equal(addr, ifa->ifa_addr)) { 1944 goto done; 1945 } 1946 /* IP6 doesn't have broadcast */ 1947 if ((ifp->if_flags & IFF_BROADCAST) && 1948 ifa->ifa_broadaddr && 1949 ifa->ifa_broadaddr->sa_len != 0 && 1950 sa_equal(ifa->ifa_broadaddr, addr)) { 1951 goto done; 1952 } 1953 } 1954 } 1955 ifa = NULL; 1956 done: 1957 return (ifa); 1958 } 1959 1960 int 1961 ifa_ifwithaddr_check(const struct sockaddr *addr) 1962 { 1963 int rc; 1964 1965 NET_EPOCH_ENTER(); 1966 rc = (ifa_ifwithaddr(addr) != NULL); 1967 NET_EPOCH_EXIT(); 1968 return (rc); 1969 } 1970 1971 /* 1972 * Locate an interface based on the broadcast address. 1973 */ 1974 /* ARGSUSED */ 1975 struct ifaddr * 1976 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1977 { 1978 struct ifnet *ifp; 1979 struct ifaddr *ifa; 1980 1981 MPASS(in_epoch(net_epoch_preempt)); 1982 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1983 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1984 continue; 1985 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1986 if (ifa->ifa_addr->sa_family != addr->sa_family) 1987 continue; 1988 if ((ifp->if_flags & IFF_BROADCAST) && 1989 ifa->ifa_broadaddr && 1990 ifa->ifa_broadaddr->sa_len != 0 && 1991 sa_equal(ifa->ifa_broadaddr, addr)) { 1992 goto done; 1993 } 1994 } 1995 } 1996 ifa = NULL; 1997 done: 1998 return (ifa); 1999 } 2000 2001 /* 2002 * Locate the point to point interface with a given destination address. 2003 */ 2004 /*ARGSUSED*/ 2005 struct ifaddr * 2006 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 2007 { 2008 struct ifnet *ifp; 2009 struct ifaddr *ifa; 2010 2011 MPASS(in_epoch(net_epoch_preempt)); 2012 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2013 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 2014 continue; 2015 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2016 continue; 2017 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2018 if (ifa->ifa_addr->sa_family != addr->sa_family) 2019 continue; 2020 if (ifa->ifa_dstaddr != NULL && 2021 sa_equal(addr, ifa->ifa_dstaddr)) { 2022 goto done; 2023 } 2024 } 2025 } 2026 ifa = NULL; 2027 done: 2028 return (ifa); 2029 } 2030 2031 /* 2032 * Find an interface on a specific network. If many, choice 2033 * is most specific found. 2034 */ 2035 struct ifaddr * 2036 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 2037 { 2038 struct ifnet *ifp; 2039 struct ifaddr *ifa; 2040 struct ifaddr *ifa_maybe = NULL; 2041 u_int af = addr->sa_family; 2042 const char *addr_data = addr->sa_data, *cplim; 2043 2044 MPASS(in_epoch(net_epoch_preempt)); 2045 /* 2046 * AF_LINK addresses can be looked up directly by their index number, 2047 * so do that if we can. 2048 */ 2049 if (af == AF_LINK) { 2050 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 2051 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 2052 return (ifaddr_byindex(sdl->sdl_index)); 2053 } 2054 2055 /* 2056 * Scan though each interface, looking for ones that have addresses 2057 * in this address family and the requested fib. Maintain a reference 2058 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that 2059 * kept it stable when we move onto the next interface. 2060 */ 2061 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2062 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2063 continue; 2064 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2065 const char *cp, *cp2, *cp3; 2066 2067 if (ifa->ifa_addr->sa_family != af) 2068 next: continue; 2069 if (af == AF_INET && 2070 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 2071 /* 2072 * This is a bit broken as it doesn't 2073 * take into account that the remote end may 2074 * be a single node in the network we are 2075 * looking for. 2076 * The trouble is that we don't know the 2077 * netmask for the remote end. 2078 */ 2079 if (ifa->ifa_dstaddr != NULL && 2080 sa_equal(addr, ifa->ifa_dstaddr)) { 2081 goto done; 2082 } 2083 } else { 2084 /* 2085 * Scan all the bits in the ifa's address. 2086 * If a bit dissagrees with what we are 2087 * looking for, mask it with the netmask 2088 * to see if it really matters. 2089 * (A byte at a time) 2090 */ 2091 if (ifa->ifa_netmask == 0) 2092 continue; 2093 cp = addr_data; 2094 cp2 = ifa->ifa_addr->sa_data; 2095 cp3 = ifa->ifa_netmask->sa_data; 2096 cplim = ifa->ifa_netmask->sa_len 2097 + (char *)ifa->ifa_netmask; 2098 while (cp3 < cplim) 2099 if ((*cp++ ^ *cp2++) & *cp3++) 2100 goto next; /* next address! */ 2101 /* 2102 * If the netmask of what we just found 2103 * is more specific than what we had before 2104 * (if we had one), or if the virtual status 2105 * of new prefix is better than of the old one, 2106 * then remember the new one before continuing 2107 * to search for an even better one. 2108 */ 2109 if (ifa_maybe == NULL || 2110 ifa_preferred(ifa_maybe, ifa) || 2111 rn_refines((caddr_t)ifa->ifa_netmask, 2112 (caddr_t)ifa_maybe->ifa_netmask)) { 2113 ifa_maybe = ifa; 2114 } 2115 } 2116 } 2117 } 2118 ifa = ifa_maybe; 2119 ifa_maybe = NULL; 2120 done: 2121 return (ifa); 2122 } 2123 2124 /* 2125 * Find an interface address specific to an interface best matching 2126 * a given address. 2127 */ 2128 struct ifaddr * 2129 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2130 { 2131 struct ifaddr *ifa; 2132 const char *cp, *cp2, *cp3; 2133 char *cplim; 2134 struct ifaddr *ifa_maybe = NULL; 2135 u_int af = addr->sa_family; 2136 2137 if (af >= AF_MAX) 2138 return (NULL); 2139 2140 MPASS(in_epoch(net_epoch_preempt)); 2141 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2142 if (ifa->ifa_addr->sa_family != af) 2143 continue; 2144 if (ifa_maybe == NULL) 2145 ifa_maybe = ifa; 2146 if (ifa->ifa_netmask == 0) { 2147 if (sa_equal(addr, ifa->ifa_addr) || 2148 (ifa->ifa_dstaddr && 2149 sa_equal(addr, ifa->ifa_dstaddr))) 2150 goto done; 2151 continue; 2152 } 2153 if (ifp->if_flags & IFF_POINTOPOINT) { 2154 if (sa_equal(addr, ifa->ifa_dstaddr)) 2155 goto done; 2156 } else { 2157 cp = addr->sa_data; 2158 cp2 = ifa->ifa_addr->sa_data; 2159 cp3 = ifa->ifa_netmask->sa_data; 2160 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2161 for (; cp3 < cplim; cp3++) 2162 if ((*cp++ ^ *cp2++) & *cp3) 2163 break; 2164 if (cp3 == cplim) 2165 goto done; 2166 } 2167 } 2168 ifa = ifa_maybe; 2169 done: 2170 return (ifa); 2171 } 2172 2173 /* 2174 * See whether new ifa is better than current one: 2175 * 1) A non-virtual one is preferred over virtual. 2176 * 2) A virtual in master state preferred over any other state. 2177 * 2178 * Used in several address selecting functions. 2179 */ 2180 int 2181 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2182 { 2183 2184 return (cur->ifa_carp && (!next->ifa_carp || 2185 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2186 } 2187 2188 #include <net/if_llatbl.h> 2189 2190 /* 2191 * Default action when installing a route with a Link Level gateway. 2192 * Lookup an appropriate real ifa to point to. 2193 * This should be moved to /sys/net/link.c eventually. 2194 */ 2195 static void 2196 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 2197 { 2198 struct ifaddr *ifa, *oifa; 2199 struct sockaddr *dst; 2200 struct ifnet *ifp; 2201 2202 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) || 2203 ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL)) 2204 return; 2205 NET_EPOCH_ENTER(); 2206 ifa = ifaof_ifpforaddr(dst, ifp); 2207 if (ifa) { 2208 oifa = rt->rt_ifa; 2209 if (oifa != ifa) { 2210 ifa_free(oifa); 2211 ifa_ref(ifa); 2212 } 2213 rt->rt_ifa = ifa; 2214 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 2215 ifa->ifa_rtrequest(cmd, rt, info); 2216 } 2217 NET_EPOCH_EXIT(); 2218 } 2219 2220 struct sockaddr_dl * 2221 link_alloc_sdl(size_t size, int flags) 2222 { 2223 2224 return (malloc(size, M_TEMP, flags)); 2225 } 2226 2227 void 2228 link_free_sdl(struct sockaddr *sa) 2229 { 2230 free(sa, M_TEMP); 2231 } 2232 2233 /* 2234 * Fills in given sdl with interface basic info. 2235 * Returns pointer to filled sdl. 2236 */ 2237 struct sockaddr_dl * 2238 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2239 { 2240 struct sockaddr_dl *sdl; 2241 2242 sdl = (struct sockaddr_dl *)paddr; 2243 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2244 sdl->sdl_len = sizeof(struct sockaddr_dl); 2245 sdl->sdl_family = AF_LINK; 2246 sdl->sdl_index = ifp->if_index; 2247 sdl->sdl_type = iftype; 2248 2249 return (sdl); 2250 } 2251 2252 /* 2253 * Mark an interface down and notify protocols of 2254 * the transition. 2255 */ 2256 static void 2257 if_unroute(struct ifnet *ifp, int flag, int fam) 2258 { 2259 struct ifaddr *ifa; 2260 2261 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2262 2263 ifp->if_flags &= ~flag; 2264 getmicrotime(&ifp->if_lastchange); 2265 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2266 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2267 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2268 ifp->if_qflush(ifp); 2269 2270 if (ifp->if_carp) 2271 (*carp_linkstate_p)(ifp); 2272 rt_ifmsg(ifp); 2273 } 2274 2275 /* 2276 * Mark an interface up and notify protocols of 2277 * the transition. 2278 */ 2279 static void 2280 if_route(struct ifnet *ifp, int flag, int fam) 2281 { 2282 struct ifaddr *ifa; 2283 2284 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2285 2286 ifp->if_flags |= flag; 2287 getmicrotime(&ifp->if_lastchange); 2288 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2289 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2290 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2291 if (ifp->if_carp) 2292 (*carp_linkstate_p)(ifp); 2293 rt_ifmsg(ifp); 2294 #ifdef INET6 2295 in6_if_up(ifp); 2296 #endif 2297 } 2298 2299 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2300 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2301 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2302 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2303 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2304 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2305 int (*vlan_setcookie_p)(struct ifnet *, void *); 2306 void *(*vlan_cookie_p)(struct ifnet *); 2307 2308 /* 2309 * Handle a change in the interface link state. To avoid LORs 2310 * between driver lock and upper layer locks, as well as possible 2311 * recursions, we post event to taskqueue, and all job 2312 * is done in static do_link_state_change(). 2313 */ 2314 void 2315 if_link_state_change(struct ifnet *ifp, int link_state) 2316 { 2317 /* Return if state hasn't changed. */ 2318 if (ifp->if_link_state == link_state) 2319 return; 2320 2321 ifp->if_link_state = link_state; 2322 2323 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2324 } 2325 2326 static void 2327 do_link_state_change(void *arg, int pending) 2328 { 2329 struct ifnet *ifp = (struct ifnet *)arg; 2330 int link_state = ifp->if_link_state; 2331 CURVNET_SET(ifp->if_vnet); 2332 2333 /* Notify that the link state has changed. */ 2334 rt_ifmsg(ifp); 2335 if (ifp->if_vlantrunk != NULL) 2336 (*vlan_link_state_p)(ifp); 2337 2338 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2339 ifp->if_l2com != NULL) 2340 (*ng_ether_link_state_p)(ifp, link_state); 2341 if (ifp->if_carp) 2342 (*carp_linkstate_p)(ifp); 2343 if (ifp->if_bridge) 2344 ifp->if_bridge_linkstate(ifp); 2345 if (ifp->if_lagg) 2346 (*lagg_linkstate_p)(ifp, link_state); 2347 2348 if (IS_DEFAULT_VNET(curvnet)) 2349 devctl_notify("IFNET", ifp->if_xname, 2350 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2351 NULL); 2352 if (pending > 1) 2353 if_printf(ifp, "%d link states coalesced\n", pending); 2354 if (log_link_state_change) 2355 if_printf(ifp, "link state changed to %s\n", 2356 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2357 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2358 CURVNET_RESTORE(); 2359 } 2360 2361 /* 2362 * Mark an interface down and notify protocols of 2363 * the transition. 2364 */ 2365 void 2366 if_down(struct ifnet *ifp) 2367 { 2368 2369 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2370 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2371 } 2372 2373 /* 2374 * Mark an interface up and notify protocols of 2375 * the transition. 2376 */ 2377 void 2378 if_up(struct ifnet *ifp) 2379 { 2380 2381 if_route(ifp, IFF_UP, AF_UNSPEC); 2382 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2383 } 2384 2385 /* 2386 * Flush an interface queue. 2387 */ 2388 void 2389 if_qflush(struct ifnet *ifp) 2390 { 2391 struct mbuf *m, *n; 2392 struct ifaltq *ifq; 2393 2394 ifq = &ifp->if_snd; 2395 IFQ_LOCK(ifq); 2396 #ifdef ALTQ 2397 if (ALTQ_IS_ENABLED(ifq)) 2398 ALTQ_PURGE(ifq); 2399 #endif 2400 n = ifq->ifq_head; 2401 while ((m = n) != NULL) { 2402 n = m->m_nextpkt; 2403 m_freem(m); 2404 } 2405 ifq->ifq_head = 0; 2406 ifq->ifq_tail = 0; 2407 ifq->ifq_len = 0; 2408 IFQ_UNLOCK(ifq); 2409 } 2410 2411 /* 2412 * Map interface name to interface structure pointer, with or without 2413 * returning a reference. 2414 */ 2415 struct ifnet * 2416 ifunit_ref(const char *name) 2417 { 2418 struct ifnet *ifp; 2419 2420 IFNET_RLOCK_NOSLEEP(); 2421 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2422 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2423 !(ifp->if_flags & IFF_DYING)) 2424 break; 2425 } 2426 if (ifp != NULL) 2427 if_ref(ifp); 2428 IFNET_RUNLOCK_NOSLEEP(); 2429 return (ifp); 2430 } 2431 2432 struct ifnet * 2433 ifunit(const char *name) 2434 { 2435 struct ifnet *ifp; 2436 2437 IFNET_RLOCK_NOSLEEP(); 2438 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2439 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2440 break; 2441 } 2442 IFNET_RUNLOCK_NOSLEEP(); 2443 return (ifp); 2444 } 2445 2446 static void * 2447 ifr_buffer_get_buffer(void *data) 2448 { 2449 union ifreq_union *ifrup; 2450 2451 ifrup = data; 2452 #ifdef COMPAT_FREEBSD32 2453 if (SV_CURPROC_FLAG(SV_ILP32)) 2454 return ((void *)(uintptr_t) 2455 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2456 #endif 2457 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2458 } 2459 2460 static void 2461 ifr_buffer_set_buffer_null(void *data) 2462 { 2463 union ifreq_union *ifrup; 2464 2465 ifrup = data; 2466 #ifdef COMPAT_FREEBSD32 2467 if (SV_CURPROC_FLAG(SV_ILP32)) 2468 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2469 else 2470 #endif 2471 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2472 } 2473 2474 static size_t 2475 ifr_buffer_get_length(void *data) 2476 { 2477 union ifreq_union *ifrup; 2478 2479 ifrup = data; 2480 #ifdef COMPAT_FREEBSD32 2481 if (SV_CURPROC_FLAG(SV_ILP32)) 2482 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2483 #endif 2484 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2485 } 2486 2487 static void 2488 ifr_buffer_set_length(void *data, size_t len) 2489 { 2490 union ifreq_union *ifrup; 2491 2492 ifrup = data; 2493 #ifdef COMPAT_FREEBSD32 2494 if (SV_CURPROC_FLAG(SV_ILP32)) 2495 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2496 else 2497 #endif 2498 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2499 } 2500 2501 void * 2502 ifr_data_get_ptr(void *ifrp) 2503 { 2504 union ifreq_union *ifrup; 2505 2506 ifrup = ifrp; 2507 #ifdef COMPAT_FREEBSD32 2508 if (SV_CURPROC_FLAG(SV_ILP32)) 2509 return ((void *)(uintptr_t) 2510 ifrup->ifr32.ifr_ifru.ifru_data); 2511 #endif 2512 return (ifrup->ifr.ifr_ifru.ifru_data); 2513 } 2514 2515 /* 2516 * Hardware specific interface ioctls. 2517 */ 2518 int 2519 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2520 { 2521 struct ifreq *ifr; 2522 int error = 0, do_ifup = 0; 2523 int new_flags, temp_flags; 2524 size_t namelen, onamelen; 2525 size_t descrlen; 2526 char *descrbuf, *odescrbuf; 2527 char new_name[IFNAMSIZ]; 2528 struct ifaddr *ifa; 2529 struct sockaddr_dl *sdl; 2530 2531 ifr = (struct ifreq *)data; 2532 switch (cmd) { 2533 case SIOCGIFINDEX: 2534 ifr->ifr_index = ifp->if_index; 2535 break; 2536 2537 case SIOCGIFFLAGS: 2538 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2539 ifr->ifr_flags = temp_flags & 0xffff; 2540 ifr->ifr_flagshigh = temp_flags >> 16; 2541 break; 2542 2543 case SIOCGIFCAP: 2544 ifr->ifr_reqcap = ifp->if_capabilities; 2545 ifr->ifr_curcap = ifp->if_capenable; 2546 break; 2547 2548 #ifdef MAC 2549 case SIOCGIFMAC: 2550 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2551 break; 2552 #endif 2553 2554 case SIOCGIFMETRIC: 2555 ifr->ifr_metric = ifp->if_metric; 2556 break; 2557 2558 case SIOCGIFMTU: 2559 ifr->ifr_mtu = ifp->if_mtu; 2560 break; 2561 2562 case SIOCGIFPHYS: 2563 /* XXXGL: did this ever worked? */ 2564 ifr->ifr_phys = 0; 2565 break; 2566 2567 case SIOCGIFDESCR: 2568 error = 0; 2569 sx_slock(&ifdescr_sx); 2570 if (ifp->if_description == NULL) 2571 error = ENOMSG; 2572 else { 2573 /* space for terminating nul */ 2574 descrlen = strlen(ifp->if_description) + 1; 2575 if (ifr_buffer_get_length(ifr) < descrlen) 2576 ifr_buffer_set_buffer_null(ifr); 2577 else 2578 error = copyout(ifp->if_description, 2579 ifr_buffer_get_buffer(ifr), descrlen); 2580 ifr_buffer_set_length(ifr, descrlen); 2581 } 2582 sx_sunlock(&ifdescr_sx); 2583 break; 2584 2585 case SIOCSIFDESCR: 2586 error = priv_check(td, PRIV_NET_SETIFDESCR); 2587 if (error) 2588 return (error); 2589 2590 /* 2591 * Copy only (length-1) bytes to make sure that 2592 * if_description is always nul terminated. The 2593 * length parameter is supposed to count the 2594 * terminating nul in. 2595 */ 2596 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2597 return (ENAMETOOLONG); 2598 else if (ifr_buffer_get_length(ifr) == 0) 2599 descrbuf = NULL; 2600 else { 2601 descrbuf = malloc(ifr_buffer_get_length(ifr), 2602 M_IFDESCR, M_WAITOK | M_ZERO); 2603 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2604 ifr_buffer_get_length(ifr) - 1); 2605 if (error) { 2606 free(descrbuf, M_IFDESCR); 2607 break; 2608 } 2609 } 2610 2611 sx_xlock(&ifdescr_sx); 2612 odescrbuf = ifp->if_description; 2613 ifp->if_description = descrbuf; 2614 sx_xunlock(&ifdescr_sx); 2615 2616 getmicrotime(&ifp->if_lastchange); 2617 free(odescrbuf, M_IFDESCR); 2618 break; 2619 2620 case SIOCGIFFIB: 2621 ifr->ifr_fib = ifp->if_fib; 2622 break; 2623 2624 case SIOCSIFFIB: 2625 error = priv_check(td, PRIV_NET_SETIFFIB); 2626 if (error) 2627 return (error); 2628 if (ifr->ifr_fib >= rt_numfibs) 2629 return (EINVAL); 2630 2631 ifp->if_fib = ifr->ifr_fib; 2632 break; 2633 2634 case SIOCSIFFLAGS: 2635 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2636 if (error) 2637 return (error); 2638 /* 2639 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2640 * check, so we don't need special handling here yet. 2641 */ 2642 new_flags = (ifr->ifr_flags & 0xffff) | 2643 (ifr->ifr_flagshigh << 16); 2644 if (ifp->if_flags & IFF_UP && 2645 (new_flags & IFF_UP) == 0) { 2646 if_down(ifp); 2647 } else if (new_flags & IFF_UP && 2648 (ifp->if_flags & IFF_UP) == 0) { 2649 do_ifup = 1; 2650 } 2651 /* See if permanently promiscuous mode bit is about to flip */ 2652 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2653 if (new_flags & IFF_PPROMISC) 2654 ifp->if_flags |= IFF_PROMISC; 2655 else if (ifp->if_pcount == 0) 2656 ifp->if_flags &= ~IFF_PROMISC; 2657 if (log_promisc_mode_change) 2658 if_printf(ifp, "permanently promiscuous mode %s\n", 2659 ((new_flags & IFF_PPROMISC) ? 2660 "enabled" : "disabled")); 2661 } 2662 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2663 (new_flags &~ IFF_CANTCHANGE); 2664 if (ifp->if_ioctl) { 2665 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2666 } 2667 if (do_ifup) 2668 if_up(ifp); 2669 getmicrotime(&ifp->if_lastchange); 2670 break; 2671 2672 case SIOCSIFCAP: 2673 error = priv_check(td, PRIV_NET_SETIFCAP); 2674 if (error) 2675 return (error); 2676 if (ifp->if_ioctl == NULL) 2677 return (EOPNOTSUPP); 2678 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2679 return (EINVAL); 2680 error = (*ifp->if_ioctl)(ifp, cmd, data); 2681 if (error == 0) 2682 getmicrotime(&ifp->if_lastchange); 2683 break; 2684 2685 #ifdef MAC 2686 case SIOCSIFMAC: 2687 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2688 break; 2689 #endif 2690 2691 case SIOCSIFNAME: 2692 error = priv_check(td, PRIV_NET_SETIFNAME); 2693 if (error) 2694 return (error); 2695 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2696 NULL); 2697 if (error != 0) 2698 return (error); 2699 if (new_name[0] == '\0') 2700 return (EINVAL); 2701 if (new_name[IFNAMSIZ-1] != '\0') { 2702 new_name[IFNAMSIZ-1] = '\0'; 2703 if (strlen(new_name) == IFNAMSIZ-1) 2704 return (EINVAL); 2705 } 2706 if (ifunit(new_name) != NULL) 2707 return (EEXIST); 2708 2709 /* 2710 * XXX: Locking. Nothing else seems to lock if_flags, 2711 * and there are numerous other races with the 2712 * ifunit() checks not being atomic with namespace 2713 * changes (renames, vmoves, if_attach, etc). 2714 */ 2715 ifp->if_flags |= IFF_RENAMING; 2716 2717 /* Announce the departure of the interface. */ 2718 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2719 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2720 2721 if_printf(ifp, "changing name to '%s'\n", new_name); 2722 2723 IF_ADDR_WLOCK(ifp); 2724 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2725 ifa = ifp->if_addr; 2726 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2727 namelen = strlen(new_name); 2728 onamelen = sdl->sdl_nlen; 2729 /* 2730 * Move the address if needed. This is safe because we 2731 * allocate space for a name of length IFNAMSIZ when we 2732 * create this in if_attach(). 2733 */ 2734 if (namelen != onamelen) { 2735 bcopy(sdl->sdl_data + onamelen, 2736 sdl->sdl_data + namelen, sdl->sdl_alen); 2737 } 2738 bcopy(new_name, sdl->sdl_data, namelen); 2739 sdl->sdl_nlen = namelen; 2740 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2741 bzero(sdl->sdl_data, onamelen); 2742 while (namelen != 0) 2743 sdl->sdl_data[--namelen] = 0xff; 2744 IF_ADDR_WUNLOCK(ifp); 2745 2746 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2747 /* Announce the return of the interface. */ 2748 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2749 2750 ifp->if_flags &= ~IFF_RENAMING; 2751 break; 2752 2753 #ifdef VIMAGE 2754 case SIOCSIFVNET: 2755 error = priv_check(td, PRIV_NET_SETIFVNET); 2756 if (error) 2757 return (error); 2758 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2759 break; 2760 #endif 2761 2762 case SIOCSIFMETRIC: 2763 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2764 if (error) 2765 return (error); 2766 ifp->if_metric = ifr->ifr_metric; 2767 getmicrotime(&ifp->if_lastchange); 2768 break; 2769 2770 case SIOCSIFPHYS: 2771 error = priv_check(td, PRIV_NET_SETIFPHYS); 2772 if (error) 2773 return (error); 2774 if (ifp->if_ioctl == NULL) 2775 return (EOPNOTSUPP); 2776 error = (*ifp->if_ioctl)(ifp, cmd, data); 2777 if (error == 0) 2778 getmicrotime(&ifp->if_lastchange); 2779 break; 2780 2781 case SIOCSIFMTU: 2782 { 2783 u_long oldmtu = ifp->if_mtu; 2784 2785 error = priv_check(td, PRIV_NET_SETIFMTU); 2786 if (error) 2787 return (error); 2788 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2789 return (EINVAL); 2790 if (ifp->if_ioctl == NULL) 2791 return (EOPNOTSUPP); 2792 error = (*ifp->if_ioctl)(ifp, cmd, data); 2793 if (error == 0) { 2794 getmicrotime(&ifp->if_lastchange); 2795 rt_ifmsg(ifp); 2796 #ifdef INET 2797 NETDUMP_REINIT(ifp); 2798 #endif 2799 } 2800 /* 2801 * If the link MTU changed, do network layer specific procedure. 2802 */ 2803 if (ifp->if_mtu != oldmtu) { 2804 #ifdef INET6 2805 nd6_setmtu(ifp); 2806 #endif 2807 rt_updatemtu(ifp); 2808 } 2809 break; 2810 } 2811 2812 case SIOCADDMULTI: 2813 case SIOCDELMULTI: 2814 if (cmd == SIOCADDMULTI) 2815 error = priv_check(td, PRIV_NET_ADDMULTI); 2816 else 2817 error = priv_check(td, PRIV_NET_DELMULTI); 2818 if (error) 2819 return (error); 2820 2821 /* Don't allow group membership on non-multicast interfaces. */ 2822 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2823 return (EOPNOTSUPP); 2824 2825 /* Don't let users screw up protocols' entries. */ 2826 if (ifr->ifr_addr.sa_family != AF_LINK) 2827 return (EINVAL); 2828 2829 if (cmd == SIOCADDMULTI) { 2830 struct ifmultiaddr *ifma; 2831 2832 /* 2833 * Userland is only permitted to join groups once 2834 * via the if_addmulti() KPI, because it cannot hold 2835 * struct ifmultiaddr * between calls. It may also 2836 * lose a race while we check if the membership 2837 * already exists. 2838 */ 2839 IF_ADDR_RLOCK(ifp); 2840 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2841 IF_ADDR_RUNLOCK(ifp); 2842 if (ifma != NULL) 2843 error = EADDRINUSE; 2844 else 2845 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2846 } else { 2847 error = if_delmulti(ifp, &ifr->ifr_addr); 2848 } 2849 if (error == 0) 2850 getmicrotime(&ifp->if_lastchange); 2851 break; 2852 2853 case SIOCSIFPHYADDR: 2854 case SIOCDIFPHYADDR: 2855 #ifdef INET6 2856 case SIOCSIFPHYADDR_IN6: 2857 #endif 2858 case SIOCSIFMEDIA: 2859 case SIOCSIFGENERIC: 2860 error = priv_check(td, PRIV_NET_HWIOCTL); 2861 if (error) 2862 return (error); 2863 if (ifp->if_ioctl == NULL) 2864 return (EOPNOTSUPP); 2865 error = (*ifp->if_ioctl)(ifp, cmd, data); 2866 if (error == 0) 2867 getmicrotime(&ifp->if_lastchange); 2868 break; 2869 2870 case SIOCGIFSTATUS: 2871 case SIOCGIFPSRCADDR: 2872 case SIOCGIFPDSTADDR: 2873 case SIOCGIFMEDIA: 2874 case SIOCGIFXMEDIA: 2875 case SIOCGIFGENERIC: 2876 case SIOCGIFRSSKEY: 2877 case SIOCGIFRSSHASH: 2878 if (ifp->if_ioctl == NULL) 2879 return (EOPNOTSUPP); 2880 error = (*ifp->if_ioctl)(ifp, cmd, data); 2881 break; 2882 2883 case SIOCSIFLLADDR: 2884 error = priv_check(td, PRIV_NET_SETLLADDR); 2885 if (error) 2886 return (error); 2887 error = if_setlladdr(ifp, 2888 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2889 break; 2890 2891 case SIOCGHWADDR: 2892 error = if_gethwaddr(ifp, ifr); 2893 break; 2894 2895 CASE_IOC_IFGROUPREQ(SIOCAIFGROUP): 2896 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2897 if (error) 2898 return (error); 2899 if ((error = if_addgroup(ifp, 2900 ifgr_group_get((struct ifgroupreq *)data)))) 2901 return (error); 2902 break; 2903 2904 CASE_IOC_IFGROUPREQ(SIOCGIFGROUP): 2905 if ((error = if_getgroup((struct ifgroupreq *)data, ifp))) 2906 return (error); 2907 break; 2908 2909 CASE_IOC_IFGROUPREQ(SIOCDIFGROUP): 2910 error = priv_check(td, PRIV_NET_DELIFGROUP); 2911 if (error) 2912 return (error); 2913 if ((error = if_delgroup(ifp, 2914 ifgr_group_get((struct ifgroupreq *)data)))) 2915 return (error); 2916 break; 2917 2918 default: 2919 error = ENOIOCTL; 2920 break; 2921 } 2922 return (error); 2923 } 2924 2925 #ifdef COMPAT_FREEBSD32 2926 struct ifconf32 { 2927 int32_t ifc_len; 2928 union { 2929 uint32_t ifcu_buf; 2930 uint32_t ifcu_req; 2931 } ifc_ifcu; 2932 }; 2933 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2934 #endif 2935 2936 #ifdef COMPAT_FREEBSD32 2937 static void 2938 ifmr_init(struct ifmediareq *ifmr, caddr_t data) 2939 { 2940 struct ifmediareq32 *ifmr32; 2941 2942 ifmr32 = (struct ifmediareq32 *)data; 2943 memcpy(ifmr->ifm_name, ifmr32->ifm_name, 2944 sizeof(ifmr->ifm_name)); 2945 ifmr->ifm_current = ifmr32->ifm_current; 2946 ifmr->ifm_mask = ifmr32->ifm_mask; 2947 ifmr->ifm_status = ifmr32->ifm_status; 2948 ifmr->ifm_active = ifmr32->ifm_active; 2949 ifmr->ifm_count = ifmr32->ifm_count; 2950 ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist; 2951 } 2952 2953 static void 2954 ifmr_update(const struct ifmediareq *ifmr, caddr_t data) 2955 { 2956 struct ifmediareq32 *ifmr32; 2957 2958 ifmr32 = (struct ifmediareq32 *)data; 2959 ifmr32->ifm_current = ifmr->ifm_current; 2960 ifmr32->ifm_mask = ifmr->ifm_mask; 2961 ifmr32->ifm_status = ifmr->ifm_status; 2962 ifmr32->ifm_active = ifmr->ifm_active; 2963 ifmr32->ifm_count = ifmr->ifm_count; 2964 } 2965 #endif 2966 2967 /* 2968 * Interface ioctls. 2969 */ 2970 int 2971 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2972 { 2973 #ifdef COMPAT_FREEBSD32 2974 caddr_t saved_data = NULL; 2975 struct ifmediareq ifmr; 2976 struct ifmediareq *ifmrp; 2977 #endif 2978 struct ifnet *ifp; 2979 struct ifreq *ifr; 2980 int error; 2981 int oif_flags; 2982 #ifdef VIMAGE 2983 int shutdown; 2984 #endif 2985 2986 CURVNET_SET(so->so_vnet); 2987 #ifdef VIMAGE 2988 /* Make sure the VNET is stable. */ 2989 shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET && 2990 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0; 2991 if (shutdown) { 2992 CURVNET_RESTORE(); 2993 return (EBUSY); 2994 } 2995 #endif 2996 2997 2998 switch (cmd) { 2999 case SIOCGIFCONF: 3000 error = ifconf(cmd, data); 3001 CURVNET_RESTORE(); 3002 return (error); 3003 3004 #ifdef COMPAT_FREEBSD32 3005 case SIOCGIFCONF32: 3006 { 3007 struct ifconf32 *ifc32; 3008 struct ifconf ifc; 3009 3010 ifc32 = (struct ifconf32 *)data; 3011 ifc.ifc_len = ifc32->ifc_len; 3012 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 3013 3014 error = ifconf(SIOCGIFCONF, (void *)&ifc); 3015 CURVNET_RESTORE(); 3016 if (error == 0) 3017 ifc32->ifc_len = ifc.ifc_len; 3018 return (error); 3019 } 3020 #endif 3021 } 3022 3023 #ifdef COMPAT_FREEBSD32 3024 ifmrp = NULL; 3025 switch (cmd) { 3026 case SIOCGIFMEDIA32: 3027 case SIOCGIFXMEDIA32: 3028 ifmrp = &ifmr; 3029 ifmr_init(ifmrp, data); 3030 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 3031 saved_data = data; 3032 data = (caddr_t)ifmrp; 3033 } 3034 #endif 3035 3036 ifr = (struct ifreq *)data; 3037 switch (cmd) { 3038 #ifdef VIMAGE 3039 case SIOCSIFRVNET: 3040 error = priv_check(td, PRIV_NET_SETIFVNET); 3041 if (error == 0) 3042 error = if_vmove_reclaim(td, ifr->ifr_name, 3043 ifr->ifr_jid); 3044 goto out_noref; 3045 #endif 3046 case SIOCIFCREATE: 3047 case SIOCIFCREATE2: 3048 error = priv_check(td, PRIV_NET_IFCREATE); 3049 if (error == 0) 3050 error = if_clone_create(ifr->ifr_name, 3051 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3052 ifr_data_get_ptr(ifr) : NULL); 3053 goto out_noref; 3054 case SIOCIFDESTROY: 3055 error = priv_check(td, PRIV_NET_IFDESTROY); 3056 if (error == 0) 3057 error = if_clone_destroy(ifr->ifr_name); 3058 goto out_noref; 3059 3060 case SIOCIFGCLONERS: 3061 error = if_clone_list((struct if_clonereq *)data); 3062 goto out_noref; 3063 3064 CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB): 3065 error = if_getgroupmembers((struct ifgroupreq *)data); 3066 goto out_noref; 3067 3068 #if defined(INET) || defined(INET6) 3069 case SIOCSVH: 3070 case SIOCGVH: 3071 if (carp_ioctl_p == NULL) 3072 error = EPROTONOSUPPORT; 3073 else 3074 error = (*carp_ioctl_p)(ifr, cmd, td); 3075 goto out_noref; 3076 #endif 3077 } 3078 3079 ifp = ifunit_ref(ifr->ifr_name); 3080 if (ifp == NULL) { 3081 error = ENXIO; 3082 goto out_noref; 3083 } 3084 3085 error = ifhwioctl(cmd, ifp, data, td); 3086 if (error != ENOIOCTL) 3087 goto out_ref; 3088 3089 oif_flags = ifp->if_flags; 3090 if (so->so_proto == NULL) { 3091 error = EOPNOTSUPP; 3092 goto out_ref; 3093 } 3094 3095 /* 3096 * Pass the request on to the socket control method, and if the 3097 * latter returns EOPNOTSUPP, directly to the interface. 3098 * 3099 * Make an exception for the legacy SIOCSIF* requests. Drivers 3100 * trust SIOCSIFADDR et al to come from an already privileged 3101 * layer, and do not perform any credentials checks or input 3102 * validation. 3103 */ 3104 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 3105 ifp, td)); 3106 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3107 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3108 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3109 error = (*ifp->if_ioctl)(ifp, cmd, data); 3110 3111 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3112 #ifdef INET6 3113 if (ifp->if_flags & IFF_UP) 3114 in6_if_up(ifp); 3115 #endif 3116 } 3117 3118 out_ref: 3119 if_rele(ifp); 3120 out_noref: 3121 #ifdef COMPAT_FREEBSD32 3122 if (ifmrp != NULL) { 3123 KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA), 3124 ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx", 3125 cmd)); 3126 data = saved_data; 3127 ifmr_update(ifmrp, data); 3128 } 3129 #endif 3130 CURVNET_RESTORE(); 3131 return (error); 3132 } 3133 3134 /* 3135 * The code common to handling reference counted flags, 3136 * e.g., in ifpromisc() and if_allmulti(). 3137 * The "pflag" argument can specify a permanent mode flag to check, 3138 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3139 * 3140 * Only to be used on stack-owned flags, not driver-owned flags. 3141 */ 3142 static int 3143 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3144 { 3145 struct ifreq ifr; 3146 int error; 3147 int oldflags, oldcount; 3148 3149 /* Sanity checks to catch programming errors */ 3150 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3151 ("%s: setting driver-owned flag %d", __func__, flag)); 3152 3153 if (onswitch) 3154 KASSERT(*refcount >= 0, 3155 ("%s: increment negative refcount %d for flag %d", 3156 __func__, *refcount, flag)); 3157 else 3158 KASSERT(*refcount > 0, 3159 ("%s: decrement non-positive refcount %d for flag %d", 3160 __func__, *refcount, flag)); 3161 3162 /* In case this mode is permanent, just touch refcount */ 3163 if (ifp->if_flags & pflag) { 3164 *refcount += onswitch ? 1 : -1; 3165 return (0); 3166 } 3167 3168 /* Save ifnet parameters for if_ioctl() may fail */ 3169 oldcount = *refcount; 3170 oldflags = ifp->if_flags; 3171 3172 /* 3173 * See if we aren't the only and touching refcount is enough. 3174 * Actually toggle interface flag if we are the first or last. 3175 */ 3176 if (onswitch) { 3177 if ((*refcount)++) 3178 return (0); 3179 ifp->if_flags |= flag; 3180 } else { 3181 if (--(*refcount)) 3182 return (0); 3183 ifp->if_flags &= ~flag; 3184 } 3185 3186 /* Call down the driver since we've changed interface flags */ 3187 if (ifp->if_ioctl == NULL) { 3188 error = EOPNOTSUPP; 3189 goto recover; 3190 } 3191 ifr.ifr_flags = ifp->if_flags & 0xffff; 3192 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3193 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3194 if (error) 3195 goto recover; 3196 /* Notify userland that interface flags have changed */ 3197 rt_ifmsg(ifp); 3198 return (0); 3199 3200 recover: 3201 /* Recover after driver error */ 3202 *refcount = oldcount; 3203 ifp->if_flags = oldflags; 3204 return (error); 3205 } 3206 3207 /* 3208 * Set/clear promiscuous mode on interface ifp based on the truth value 3209 * of pswitch. The calls are reference counted so that only the first 3210 * "on" request actually has an effect, as does the final "off" request. 3211 * Results are undefined if the "off" and "on" requests are not matched. 3212 */ 3213 int 3214 ifpromisc(struct ifnet *ifp, int pswitch) 3215 { 3216 int error; 3217 int oldflags = ifp->if_flags; 3218 3219 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3220 &ifp->if_pcount, pswitch); 3221 /* If promiscuous mode status has changed, log a message */ 3222 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3223 log_promisc_mode_change) 3224 if_printf(ifp, "promiscuous mode %s\n", 3225 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3226 return (error); 3227 } 3228 3229 /* 3230 * Return interface configuration 3231 * of system. List may be used 3232 * in later ioctl's (above) to get 3233 * other information. 3234 */ 3235 /*ARGSUSED*/ 3236 static int 3237 ifconf(u_long cmd, caddr_t data) 3238 { 3239 struct ifconf *ifc = (struct ifconf *)data; 3240 struct ifnet *ifp; 3241 struct ifaddr *ifa; 3242 struct ifreq ifr; 3243 struct sbuf *sb; 3244 int error, full = 0, valid_len, max_len; 3245 3246 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ 3247 max_len = MAXPHYS - 1; 3248 3249 /* Prevent hostile input from being able to crash the system */ 3250 if (ifc->ifc_len <= 0) 3251 return (EINVAL); 3252 3253 again: 3254 if (ifc->ifc_len <= max_len) { 3255 max_len = ifc->ifc_len; 3256 full = 1; 3257 } 3258 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3259 max_len = 0; 3260 valid_len = 0; 3261 3262 IFNET_RLOCK(); 3263 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3264 int addrs; 3265 3266 /* 3267 * Zero the ifr to make sure we don't disclose the contents 3268 * of the stack. 3269 */ 3270 memset(&ifr, 0, sizeof(ifr)); 3271 3272 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3273 >= sizeof(ifr.ifr_name)) { 3274 sbuf_delete(sb); 3275 IFNET_RUNLOCK(); 3276 return (ENAMETOOLONG); 3277 } 3278 3279 addrs = 0; 3280 IF_ADDR_RLOCK(ifp); 3281 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3282 struct sockaddr *sa = ifa->ifa_addr; 3283 3284 if (prison_if(curthread->td_ucred, sa) != 0) 3285 continue; 3286 addrs++; 3287 if (sa->sa_len <= sizeof(*sa)) { 3288 if (sa->sa_len < sizeof(*sa)) { 3289 memset(&ifr.ifr_ifru.ifru_addr, 0, 3290 sizeof(ifr.ifr_ifru.ifru_addr)); 3291 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3292 sa->sa_len); 3293 } else 3294 ifr.ifr_ifru.ifru_addr = *sa; 3295 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3296 max_len += sizeof(ifr); 3297 } else { 3298 sbuf_bcat(sb, &ifr, 3299 offsetof(struct ifreq, ifr_addr)); 3300 max_len += offsetof(struct ifreq, ifr_addr); 3301 sbuf_bcat(sb, sa, sa->sa_len); 3302 max_len += sa->sa_len; 3303 } 3304 3305 if (sbuf_error(sb) == 0) 3306 valid_len = sbuf_len(sb); 3307 } 3308 IF_ADDR_RUNLOCK(ifp); 3309 if (addrs == 0) { 3310 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3311 max_len += sizeof(ifr); 3312 3313 if (sbuf_error(sb) == 0) 3314 valid_len = sbuf_len(sb); 3315 } 3316 } 3317 IFNET_RUNLOCK(); 3318 3319 /* 3320 * If we didn't allocate enough space (uncommon), try again. If 3321 * we have already allocated as much space as we are allowed, 3322 * return what we've got. 3323 */ 3324 if (valid_len != max_len && !full) { 3325 sbuf_delete(sb); 3326 goto again; 3327 } 3328 3329 ifc->ifc_len = valid_len; 3330 sbuf_finish(sb); 3331 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3332 sbuf_delete(sb); 3333 return (error); 3334 } 3335 3336 /* 3337 * Just like ifpromisc(), but for all-multicast-reception mode. 3338 */ 3339 int 3340 if_allmulti(struct ifnet *ifp, int onswitch) 3341 { 3342 3343 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3344 } 3345 3346 struct ifmultiaddr * 3347 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3348 { 3349 struct ifmultiaddr *ifma; 3350 3351 IF_ADDR_LOCK_ASSERT(ifp); 3352 3353 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3354 if (sa->sa_family == AF_LINK) { 3355 if (sa_dl_equal(ifma->ifma_addr, sa)) 3356 break; 3357 } else { 3358 if (sa_equal(ifma->ifma_addr, sa)) 3359 break; 3360 } 3361 } 3362 3363 return ifma; 3364 } 3365 3366 /* 3367 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3368 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3369 * the ifnet multicast address list here, so the caller must do that and 3370 * other setup work (such as notifying the device driver). The reference 3371 * count is initialized to 1. 3372 */ 3373 static struct ifmultiaddr * 3374 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3375 int mflags) 3376 { 3377 struct ifmultiaddr *ifma; 3378 struct sockaddr *dupsa; 3379 3380 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3381 M_ZERO); 3382 if (ifma == NULL) 3383 return (NULL); 3384 3385 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3386 if (dupsa == NULL) { 3387 free(ifma, M_IFMADDR); 3388 return (NULL); 3389 } 3390 bcopy(sa, dupsa, sa->sa_len); 3391 ifma->ifma_addr = dupsa; 3392 3393 ifma->ifma_ifp = ifp; 3394 ifma->ifma_refcount = 1; 3395 ifma->ifma_protospec = NULL; 3396 3397 if (llsa == NULL) { 3398 ifma->ifma_lladdr = NULL; 3399 return (ifma); 3400 } 3401 3402 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3403 if (dupsa == NULL) { 3404 free(ifma->ifma_addr, M_IFMADDR); 3405 free(ifma, M_IFMADDR); 3406 return (NULL); 3407 } 3408 bcopy(llsa, dupsa, llsa->sa_len); 3409 ifma->ifma_lladdr = dupsa; 3410 3411 return (ifma); 3412 } 3413 3414 /* 3415 * if_freemulti: free ifmultiaddr structure and possibly attached related 3416 * addresses. The caller is responsible for implementing reference 3417 * counting, notifying the driver, handling routing messages, and releasing 3418 * any dependent link layer state. 3419 */ 3420 #ifdef MCAST_VERBOSE 3421 extern void kdb_backtrace(void); 3422 #endif 3423 static void 3424 if_freemulti_internal(struct ifmultiaddr *ifma) 3425 { 3426 3427 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3428 ifma->ifma_refcount)); 3429 3430 if (ifma->ifma_lladdr != NULL) 3431 free(ifma->ifma_lladdr, M_IFMADDR); 3432 #ifdef MCAST_VERBOSE 3433 kdb_backtrace(); 3434 printf("%s freeing ifma: %p\n", __func__, ifma); 3435 #endif 3436 free(ifma->ifma_addr, M_IFMADDR); 3437 free(ifma, M_IFMADDR); 3438 } 3439 3440 static void 3441 if_destroymulti(epoch_context_t ctx) 3442 { 3443 struct ifmultiaddr *ifma; 3444 3445 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3446 if_freemulti_internal(ifma); 3447 } 3448 3449 void 3450 if_freemulti(struct ifmultiaddr *ifma) 3451 { 3452 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3453 ifma->ifma_refcount)); 3454 3455 epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti); 3456 } 3457 3458 3459 /* 3460 * Register an additional multicast address with a network interface. 3461 * 3462 * - If the address is already present, bump the reference count on the 3463 * address and return. 3464 * - If the address is not link-layer, look up a link layer address. 3465 * - Allocate address structures for one or both addresses, and attach to the 3466 * multicast address list on the interface. If automatically adding a link 3467 * layer address, the protocol address will own a reference to the link 3468 * layer address, to be freed when it is freed. 3469 * - Notify the network device driver of an addition to the multicast address 3470 * list. 3471 * 3472 * 'sa' points to caller-owned memory with the desired multicast address. 3473 * 3474 * 'retifma' will be used to return a pointer to the resulting multicast 3475 * address reference, if desired. 3476 */ 3477 int 3478 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3479 struct ifmultiaddr **retifma) 3480 { 3481 struct ifmultiaddr *ifma, *ll_ifma; 3482 struct sockaddr *llsa; 3483 struct sockaddr_dl sdl; 3484 int error; 3485 3486 #ifdef INET 3487 IN_MULTI_LIST_UNLOCK_ASSERT(); 3488 #endif 3489 #ifdef INET6 3490 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3491 #endif 3492 /* 3493 * If the address is already present, return a new reference to it; 3494 * otherwise, allocate storage and set up a new address. 3495 */ 3496 IF_ADDR_WLOCK(ifp); 3497 ifma = if_findmulti(ifp, sa); 3498 if (ifma != NULL) { 3499 ifma->ifma_refcount++; 3500 if (retifma != NULL) 3501 *retifma = ifma; 3502 IF_ADDR_WUNLOCK(ifp); 3503 return (0); 3504 } 3505 3506 /* 3507 * The address isn't already present; resolve the protocol address 3508 * into a link layer address, and then look that up, bump its 3509 * refcount or allocate an ifma for that also. 3510 * Most link layer resolving functions returns address data which 3511 * fits inside default sockaddr_dl structure. However callback 3512 * can allocate another sockaddr structure, in that case we need to 3513 * free it later. 3514 */ 3515 llsa = NULL; 3516 ll_ifma = NULL; 3517 if (ifp->if_resolvemulti != NULL) { 3518 /* Provide called function with buffer size information */ 3519 sdl.sdl_len = sizeof(sdl); 3520 llsa = (struct sockaddr *)&sdl; 3521 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3522 if (error) 3523 goto unlock_out; 3524 } 3525 3526 /* 3527 * Allocate the new address. Don't hook it up yet, as we may also 3528 * need to allocate a link layer multicast address. 3529 */ 3530 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3531 if (ifma == NULL) { 3532 error = ENOMEM; 3533 goto free_llsa_out; 3534 } 3535 3536 /* 3537 * If a link layer address is found, we'll need to see if it's 3538 * already present in the address list, or allocate is as well. 3539 * When this block finishes, the link layer address will be on the 3540 * list. 3541 */ 3542 if (llsa != NULL) { 3543 ll_ifma = if_findmulti(ifp, llsa); 3544 if (ll_ifma == NULL) { 3545 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3546 if (ll_ifma == NULL) { 3547 --ifma->ifma_refcount; 3548 if_freemulti(ifma); 3549 error = ENOMEM; 3550 goto free_llsa_out; 3551 } 3552 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3553 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3554 ifma_link); 3555 } else 3556 ll_ifma->ifma_refcount++; 3557 ifma->ifma_llifma = ll_ifma; 3558 } 3559 3560 /* 3561 * We now have a new multicast address, ifma, and possibly a new or 3562 * referenced link layer address. Add the primary address to the 3563 * ifnet address list. 3564 */ 3565 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3566 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3567 3568 if (retifma != NULL) 3569 *retifma = ifma; 3570 3571 /* 3572 * Must generate the message while holding the lock so that 'ifma' 3573 * pointer is still valid. 3574 */ 3575 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3576 IF_ADDR_WUNLOCK(ifp); 3577 3578 /* 3579 * We are certain we have added something, so call down to the 3580 * interface to let them know about it. 3581 */ 3582 if (ifp->if_ioctl != NULL) { 3583 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3584 } 3585 3586 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3587 link_free_sdl(llsa); 3588 3589 return (0); 3590 3591 free_llsa_out: 3592 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3593 link_free_sdl(llsa); 3594 3595 unlock_out: 3596 IF_ADDR_WUNLOCK(ifp); 3597 return (error); 3598 } 3599 3600 /* 3601 * Delete a multicast group membership by network-layer group address. 3602 * 3603 * Returns ENOENT if the entry could not be found. If ifp no longer 3604 * exists, results are undefined. This entry point should only be used 3605 * from subsystems which do appropriate locking to hold ifp for the 3606 * duration of the call. 3607 * Network-layer protocol domains must use if_delmulti_ifma(). 3608 */ 3609 int 3610 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3611 { 3612 struct ifmultiaddr *ifma; 3613 int lastref; 3614 #ifdef INVARIANTS 3615 struct ifnet *oifp; 3616 3617 IFNET_RLOCK_NOSLEEP(); 3618 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3619 if (ifp == oifp) 3620 break; 3621 if (ifp != oifp) 3622 ifp = NULL; 3623 IFNET_RUNLOCK_NOSLEEP(); 3624 3625 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); 3626 #endif 3627 if (ifp == NULL) 3628 return (ENOENT); 3629 3630 IF_ADDR_WLOCK(ifp); 3631 lastref = 0; 3632 ifma = if_findmulti(ifp, sa); 3633 if (ifma != NULL) 3634 lastref = if_delmulti_locked(ifp, ifma, 0); 3635 IF_ADDR_WUNLOCK(ifp); 3636 3637 if (ifma == NULL) 3638 return (ENOENT); 3639 3640 if (lastref && ifp->if_ioctl != NULL) { 3641 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3642 } 3643 3644 return (0); 3645 } 3646 3647 /* 3648 * Delete all multicast group membership for an interface. 3649 * Should be used to quickly flush all multicast filters. 3650 */ 3651 void 3652 if_delallmulti(struct ifnet *ifp) 3653 { 3654 struct ifmultiaddr *ifma; 3655 struct ifmultiaddr *next; 3656 3657 IF_ADDR_WLOCK(ifp); 3658 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3659 if_delmulti_locked(ifp, ifma, 0); 3660 IF_ADDR_WUNLOCK(ifp); 3661 } 3662 3663 void 3664 if_delmulti_ifma(struct ifmultiaddr *ifma) 3665 { 3666 if_delmulti_ifma_flags(ifma, 0); 3667 } 3668 3669 /* 3670 * Delete a multicast group membership by group membership pointer. 3671 * Network-layer protocol domains must use this routine. 3672 * 3673 * It is safe to call this routine if the ifp disappeared. 3674 */ 3675 void 3676 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3677 { 3678 struct ifnet *ifp; 3679 int lastref; 3680 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3681 #ifdef INET 3682 IN_MULTI_LIST_UNLOCK_ASSERT(); 3683 #endif 3684 ifp = ifma->ifma_ifp; 3685 #ifdef DIAGNOSTIC 3686 if (ifp == NULL) { 3687 printf("%s: ifma_ifp seems to be detached\n", __func__); 3688 } else { 3689 struct ifnet *oifp; 3690 3691 IFNET_RLOCK_NOSLEEP(); 3692 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3693 if (ifp == oifp) 3694 break; 3695 if (ifp != oifp) 3696 ifp = NULL; 3697 IFNET_RUNLOCK_NOSLEEP(); 3698 } 3699 #endif 3700 /* 3701 * If and only if the ifnet instance exists: Acquire the address lock. 3702 */ 3703 if (ifp != NULL) 3704 IF_ADDR_WLOCK(ifp); 3705 3706 lastref = if_delmulti_locked(ifp, ifma, flags); 3707 3708 if (ifp != NULL) { 3709 /* 3710 * If and only if the ifnet instance exists: 3711 * Release the address lock. 3712 * If the group was left: update the hardware hash filter. 3713 */ 3714 IF_ADDR_WUNLOCK(ifp); 3715 if (lastref && ifp->if_ioctl != NULL) { 3716 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3717 } 3718 } 3719 } 3720 3721 /* 3722 * Perform deletion of network-layer and/or link-layer multicast address. 3723 * 3724 * Return 0 if the reference count was decremented. 3725 * Return 1 if the final reference was released, indicating that the 3726 * hardware hash filter should be reprogrammed. 3727 */ 3728 static int 3729 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3730 { 3731 struct ifmultiaddr *ll_ifma; 3732 3733 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3734 KASSERT(ifma->ifma_ifp == ifp, 3735 ("%s: inconsistent ifp %p", __func__, ifp)); 3736 IF_ADDR_WLOCK_ASSERT(ifp); 3737 } 3738 3739 ifp = ifma->ifma_ifp; 3740 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3741 3742 /* 3743 * If the ifnet is detaching, null out references to ifnet, 3744 * so that upper protocol layers will notice, and not attempt 3745 * to obtain locks for an ifnet which no longer exists. The 3746 * routing socket announcement must happen before the ifnet 3747 * instance is detached from the system. 3748 */ 3749 if (detaching) { 3750 #ifdef DIAGNOSTIC 3751 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3752 #endif 3753 /* 3754 * ifp may already be nulled out if we are being reentered 3755 * to delete the ll_ifma. 3756 */ 3757 if (ifp != NULL) { 3758 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3759 ifma->ifma_ifp = NULL; 3760 } 3761 } 3762 3763 if (--ifma->ifma_refcount > 0) 3764 return 0; 3765 3766 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3767 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3768 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3769 } 3770 /* 3771 * If this ifma is a network-layer ifma, a link-layer ifma may 3772 * have been associated with it. Release it first if so. 3773 */ 3774 ll_ifma = ifma->ifma_llifma; 3775 if (ll_ifma != NULL) { 3776 KASSERT(ifma->ifma_lladdr != NULL, 3777 ("%s: llifma w/o lladdr", __func__)); 3778 if (detaching) 3779 ll_ifma->ifma_ifp = NULL; /* XXX */ 3780 if (--ll_ifma->ifma_refcount == 0) { 3781 if (ifp != NULL) { 3782 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3783 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3784 ifma_link); 3785 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3786 } 3787 } 3788 if_freemulti(ll_ifma); 3789 } 3790 } 3791 #ifdef INVARIANTS 3792 if (ifp) { 3793 struct ifmultiaddr *ifmatmp; 3794 3795 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3796 MPASS(ifma != ifmatmp); 3797 } 3798 #endif 3799 if_freemulti(ifma); 3800 /* 3801 * The last reference to this instance of struct ifmultiaddr 3802 * was released; the hardware should be notified of this change. 3803 */ 3804 return 1; 3805 } 3806 3807 /* 3808 * Set the link layer address on an interface. 3809 * 3810 * At this time we only support certain types of interfaces, 3811 * and we don't allow the length of the address to change. 3812 * 3813 * Set noinline to be dtrace-friendly 3814 */ 3815 __noinline int 3816 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3817 { 3818 struct sockaddr_dl *sdl; 3819 struct ifaddr *ifa; 3820 struct ifreq ifr; 3821 int rc; 3822 3823 rc = 0; 3824 NET_EPOCH_ENTER(); 3825 ifa = ifp->if_addr; 3826 if (ifa == NULL) { 3827 rc = EINVAL; 3828 goto out; 3829 } 3830 3831 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3832 if (sdl == NULL) { 3833 rc = EINVAL; 3834 goto out; 3835 } 3836 if (len != sdl->sdl_alen) { /* don't allow length to change */ 3837 rc = EINVAL; 3838 goto out; 3839 } 3840 switch (ifp->if_type) { 3841 case IFT_ETHER: 3842 case IFT_XETHER: 3843 case IFT_L2VLAN: 3844 case IFT_BRIDGE: 3845 case IFT_IEEE8023ADLAG: 3846 bcopy(lladdr, LLADDR(sdl), len); 3847 break; 3848 default: 3849 rc = ENODEV; 3850 goto out; 3851 } 3852 3853 /* 3854 * If the interface is already up, we need 3855 * to re-init it in order to reprogram its 3856 * address filter. 3857 */ 3858 NET_EPOCH_EXIT(); 3859 if ((ifp->if_flags & IFF_UP) != 0) { 3860 if (ifp->if_ioctl) { 3861 ifp->if_flags &= ~IFF_UP; 3862 ifr.ifr_flags = ifp->if_flags & 0xffff; 3863 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3864 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3865 ifp->if_flags |= IFF_UP; 3866 ifr.ifr_flags = ifp->if_flags & 0xffff; 3867 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3868 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3869 } 3870 } 3871 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3872 return (0); 3873 out: 3874 NET_EPOCH_EXIT(); 3875 return (rc); 3876 } 3877 3878 /* 3879 * Compat function for handling basic encapsulation requests. 3880 * Not converted stacks (FDDI, IB, ..) supports traditional 3881 * output model: ARP (and other similar L2 protocols) are handled 3882 * inside output routine, arpresolve/nd6_resolve() returns MAC 3883 * address instead of full prepend. 3884 * 3885 * This function creates calculated header==MAC for IPv4/IPv6 and 3886 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3887 * address families. 3888 */ 3889 static int 3890 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3891 { 3892 3893 if (req->rtype != IFENCAP_LL) 3894 return (EOPNOTSUPP); 3895 3896 if (req->bufsize < req->lladdr_len) 3897 return (ENOMEM); 3898 3899 switch (req->family) { 3900 case AF_INET: 3901 case AF_INET6: 3902 break; 3903 default: 3904 return (EAFNOSUPPORT); 3905 } 3906 3907 /* Copy lladdr to storage as is */ 3908 memmove(req->buf, req->lladdr, req->lladdr_len); 3909 req->bufsize = req->lladdr_len; 3910 req->lladdr_off = 0; 3911 3912 return (0); 3913 } 3914 3915 /* 3916 * Tunnel interfaces can nest, also they may cause infinite recursion 3917 * calls when misconfigured. We'll prevent this by detecting loops. 3918 * High nesting level may cause stack exhaustion. We'll prevent this 3919 * by introducing upper limit. 3920 * 3921 * Return 0, if tunnel nesting count is equal or less than limit. 3922 */ 3923 int 3924 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3925 int limit) 3926 { 3927 struct m_tag *mtag; 3928 int count; 3929 3930 count = 1; 3931 mtag = NULL; 3932 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3933 if (*(struct ifnet **)(mtag + 1) == ifp) { 3934 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3935 return (EIO); 3936 } 3937 count++; 3938 } 3939 if (count > limit) { 3940 log(LOG_NOTICE, 3941 "%s: if_output recursively called too many times(%d)\n", 3942 if_name(ifp), count); 3943 return (EIO); 3944 } 3945 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3946 if (mtag == NULL) 3947 return (ENOMEM); 3948 *(struct ifnet **)(mtag + 1) = ifp; 3949 m_tag_prepend(m, mtag); 3950 return (0); 3951 } 3952 3953 /* 3954 * Get the link layer address that was read from the hardware at attach. 3955 * 3956 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3957 * their component interfaces as IFT_IEEE8023ADLAG. 3958 */ 3959 int 3960 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3961 { 3962 3963 if (ifp->if_hw_addr == NULL) 3964 return (ENODEV); 3965 3966 switch (ifp->if_type) { 3967 case IFT_ETHER: 3968 case IFT_IEEE8023ADLAG: 3969 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3970 return (0); 3971 default: 3972 return (ENODEV); 3973 } 3974 } 3975 3976 /* 3977 * The name argument must be a pointer to storage which will last as 3978 * long as the interface does. For physical devices, the result of 3979 * device_get_name(dev) is a good choice and for pseudo-devices a 3980 * static string works well. 3981 */ 3982 void 3983 if_initname(struct ifnet *ifp, const char *name, int unit) 3984 { 3985 ifp->if_dname = name; 3986 ifp->if_dunit = unit; 3987 if (unit != IF_DUNIT_NONE) 3988 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3989 else 3990 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3991 } 3992 3993 int 3994 if_printf(struct ifnet *ifp, const char *fmt, ...) 3995 { 3996 char if_fmt[256]; 3997 va_list ap; 3998 3999 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 4000 va_start(ap, fmt); 4001 vlog(LOG_INFO, if_fmt, ap); 4002 va_end(ap); 4003 return (0); 4004 } 4005 4006 void 4007 if_start(struct ifnet *ifp) 4008 { 4009 4010 (*(ifp)->if_start)(ifp); 4011 } 4012 4013 /* 4014 * Backwards compatibility interface for drivers 4015 * that have not implemented it 4016 */ 4017 static int 4018 if_transmit(struct ifnet *ifp, struct mbuf *m) 4019 { 4020 int error; 4021 4022 IFQ_HANDOFF(ifp, m, error); 4023 return (error); 4024 } 4025 4026 static void 4027 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4028 { 4029 4030 m_freem(m); 4031 } 4032 4033 int 4034 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4035 { 4036 int active = 0; 4037 4038 IF_LOCK(ifq); 4039 if (_IF_QFULL(ifq)) { 4040 IF_UNLOCK(ifq); 4041 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4042 m_freem(m); 4043 return (0); 4044 } 4045 if (ifp != NULL) { 4046 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4047 if (m->m_flags & (M_BCAST|M_MCAST)) 4048 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4049 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4050 } 4051 _IF_ENQUEUE(ifq, m); 4052 IF_UNLOCK(ifq); 4053 if (ifp != NULL && !active) 4054 (*(ifp)->if_start)(ifp); 4055 return (1); 4056 } 4057 4058 void 4059 if_register_com_alloc(u_char type, 4060 if_com_alloc_t *a, if_com_free_t *f) 4061 { 4062 4063 KASSERT(if_com_alloc[type] == NULL, 4064 ("if_register_com_alloc: %d already registered", type)); 4065 KASSERT(if_com_free[type] == NULL, 4066 ("if_register_com_alloc: %d free already registered", type)); 4067 4068 if_com_alloc[type] = a; 4069 if_com_free[type] = f; 4070 } 4071 4072 void 4073 if_deregister_com_alloc(u_char type) 4074 { 4075 4076 KASSERT(if_com_alloc[type] != NULL, 4077 ("if_deregister_com_alloc: %d not registered", type)); 4078 KASSERT(if_com_free[type] != NULL, 4079 ("if_deregister_com_alloc: %d free not registered", type)); 4080 if_com_alloc[type] = NULL; 4081 if_com_free[type] = NULL; 4082 } 4083 4084 /* API for driver access to network stack owned ifnet.*/ 4085 uint64_t 4086 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4087 { 4088 uint64_t oldbrate; 4089 4090 oldbrate = ifp->if_baudrate; 4091 ifp->if_baudrate = baudrate; 4092 return (oldbrate); 4093 } 4094 4095 uint64_t 4096 if_getbaudrate(if_t ifp) 4097 { 4098 4099 return (((struct ifnet *)ifp)->if_baudrate); 4100 } 4101 4102 int 4103 if_setcapabilities(if_t ifp, int capabilities) 4104 { 4105 ((struct ifnet *)ifp)->if_capabilities = capabilities; 4106 return (0); 4107 } 4108 4109 int 4110 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4111 { 4112 ((struct ifnet *)ifp)->if_capabilities |= setbit; 4113 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 4114 4115 return (0); 4116 } 4117 4118 int 4119 if_getcapabilities(if_t ifp) 4120 { 4121 return ((struct ifnet *)ifp)->if_capabilities; 4122 } 4123 4124 int 4125 if_setcapenable(if_t ifp, int capabilities) 4126 { 4127 ((struct ifnet *)ifp)->if_capenable = capabilities; 4128 return (0); 4129 } 4130 4131 int 4132 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4133 { 4134 if(setcap) 4135 ((struct ifnet *)ifp)->if_capenable |= setcap; 4136 if(clearcap) 4137 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 4138 4139 return (0); 4140 } 4141 4142 const char * 4143 if_getdname(if_t ifp) 4144 { 4145 return ((struct ifnet *)ifp)->if_dname; 4146 } 4147 4148 int 4149 if_togglecapenable(if_t ifp, int togglecap) 4150 { 4151 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 4152 return (0); 4153 } 4154 4155 int 4156 if_getcapenable(if_t ifp) 4157 { 4158 return ((struct ifnet *)ifp)->if_capenable; 4159 } 4160 4161 /* 4162 * This is largely undesirable because it ties ifnet to a device, but does 4163 * provide flexiblity for an embedded product vendor. Should be used with 4164 * the understanding that it violates the interface boundaries, and should be 4165 * a last resort only. 4166 */ 4167 int 4168 if_setdev(if_t ifp, void *dev) 4169 { 4170 return (0); 4171 } 4172 4173 int 4174 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4175 { 4176 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 4177 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 4178 4179 return (0); 4180 } 4181 4182 int 4183 if_getdrvflags(if_t ifp) 4184 { 4185 return ((struct ifnet *)ifp)->if_drv_flags; 4186 } 4187 4188 int 4189 if_setdrvflags(if_t ifp, int flags) 4190 { 4191 ((struct ifnet *)ifp)->if_drv_flags = flags; 4192 return (0); 4193 } 4194 4195 4196 int 4197 if_setflags(if_t ifp, int flags) 4198 { 4199 ((struct ifnet *)ifp)->if_flags = flags; 4200 return (0); 4201 } 4202 4203 int 4204 if_setflagbits(if_t ifp, int set, int clear) 4205 { 4206 ((struct ifnet *)ifp)->if_flags |= set; 4207 ((struct ifnet *)ifp)->if_flags &= ~clear; 4208 4209 return (0); 4210 } 4211 4212 int 4213 if_getflags(if_t ifp) 4214 { 4215 return ((struct ifnet *)ifp)->if_flags; 4216 } 4217 4218 int 4219 if_clearhwassist(if_t ifp) 4220 { 4221 ((struct ifnet *)ifp)->if_hwassist = 0; 4222 return (0); 4223 } 4224 4225 int 4226 if_sethwassistbits(if_t ifp, int toset, int toclear) 4227 { 4228 ((struct ifnet *)ifp)->if_hwassist |= toset; 4229 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4230 4231 return (0); 4232 } 4233 4234 int 4235 if_sethwassist(if_t ifp, int hwassist_bit) 4236 { 4237 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4238 return (0); 4239 } 4240 4241 int 4242 if_gethwassist(if_t ifp) 4243 { 4244 return ((struct ifnet *)ifp)->if_hwassist; 4245 } 4246 4247 int 4248 if_setmtu(if_t ifp, int mtu) 4249 { 4250 ((struct ifnet *)ifp)->if_mtu = mtu; 4251 return (0); 4252 } 4253 4254 int 4255 if_getmtu(if_t ifp) 4256 { 4257 return ((struct ifnet *)ifp)->if_mtu; 4258 } 4259 4260 int 4261 if_getmtu_family(if_t ifp, int family) 4262 { 4263 struct domain *dp; 4264 4265 for (dp = domains; dp; dp = dp->dom_next) { 4266 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4267 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4268 } 4269 4270 return (((struct ifnet *)ifp)->if_mtu); 4271 } 4272 4273 int 4274 if_setsoftc(if_t ifp, void *softc) 4275 { 4276 ((struct ifnet *)ifp)->if_softc = softc; 4277 return (0); 4278 } 4279 4280 void * 4281 if_getsoftc(if_t ifp) 4282 { 4283 return ((struct ifnet *)ifp)->if_softc; 4284 } 4285 4286 void 4287 if_setrcvif(struct mbuf *m, if_t ifp) 4288 { 4289 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4290 } 4291 4292 void 4293 if_setvtag(struct mbuf *m, uint16_t tag) 4294 { 4295 m->m_pkthdr.ether_vtag = tag; 4296 } 4297 4298 uint16_t 4299 if_getvtag(struct mbuf *m) 4300 { 4301 4302 return (m->m_pkthdr.ether_vtag); 4303 } 4304 4305 int 4306 if_sendq_empty(if_t ifp) 4307 { 4308 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4309 } 4310 4311 struct ifaddr * 4312 if_getifaddr(if_t ifp) 4313 { 4314 return ((struct ifnet *)ifp)->if_addr; 4315 } 4316 4317 int 4318 if_getamcount(if_t ifp) 4319 { 4320 return ((struct ifnet *)ifp)->if_amcount; 4321 } 4322 4323 4324 int 4325 if_setsendqready(if_t ifp) 4326 { 4327 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4328 return (0); 4329 } 4330 4331 int 4332 if_setsendqlen(if_t ifp, int tx_desc_count) 4333 { 4334 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4335 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4336 4337 return (0); 4338 } 4339 4340 int 4341 if_vlantrunkinuse(if_t ifp) 4342 { 4343 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4344 } 4345 4346 int 4347 if_input(if_t ifp, struct mbuf* sendmp) 4348 { 4349 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4350 return (0); 4351 4352 } 4353 4354 /* XXX */ 4355 #ifndef ETH_ADDR_LEN 4356 #define ETH_ADDR_LEN 6 4357 #endif 4358 4359 int 4360 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max) 4361 { 4362 struct ifmultiaddr *ifma; 4363 uint8_t *lmta = (uint8_t *)mta; 4364 int mcnt = 0; 4365 4366 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4367 if (ifma->ifma_addr->sa_family != AF_LINK) 4368 continue; 4369 4370 if (mcnt == max) 4371 break; 4372 4373 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 4374 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 4375 mcnt++; 4376 } 4377 *cnt = mcnt; 4378 4379 return (0); 4380 } 4381 4382 int 4383 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max) 4384 { 4385 int error; 4386 4387 if_maddr_rlock(ifp); 4388 error = if_setupmultiaddr(ifp, mta, cnt, max); 4389 if_maddr_runlock(ifp); 4390 return (error); 4391 } 4392 4393 int 4394 if_multiaddr_count(if_t ifp, int max) 4395 { 4396 struct ifmultiaddr *ifma; 4397 int count; 4398 4399 count = 0; 4400 if_maddr_rlock(ifp); 4401 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { 4402 if (ifma->ifma_addr->sa_family != AF_LINK) 4403 continue; 4404 count++; 4405 if (count == max) 4406 break; 4407 } 4408 if_maddr_runlock(ifp); 4409 return (count); 4410 } 4411 4412 int 4413 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg) 4414 { 4415 struct ifmultiaddr *ifma; 4416 int cnt = 0; 4417 4418 if_maddr_rlock(ifp); 4419 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4420 cnt += filter(arg, ifma, cnt); 4421 if_maddr_runlock(ifp); 4422 return (cnt); 4423 } 4424 4425 struct mbuf * 4426 if_dequeue(if_t ifp) 4427 { 4428 struct mbuf *m; 4429 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4430 4431 return (m); 4432 } 4433 4434 int 4435 if_sendq_prepend(if_t ifp, struct mbuf *m) 4436 { 4437 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4438 return (0); 4439 } 4440 4441 int 4442 if_setifheaderlen(if_t ifp, int len) 4443 { 4444 ((struct ifnet *)ifp)->if_hdrlen = len; 4445 return (0); 4446 } 4447 4448 caddr_t 4449 if_getlladdr(if_t ifp) 4450 { 4451 return (IF_LLADDR((struct ifnet *)ifp)); 4452 } 4453 4454 void * 4455 if_gethandle(u_char type) 4456 { 4457 return (if_alloc(type)); 4458 } 4459 4460 void 4461 if_bpfmtap(if_t ifh, struct mbuf *m) 4462 { 4463 struct ifnet *ifp = (struct ifnet *)ifh; 4464 4465 BPF_MTAP(ifp, m); 4466 } 4467 4468 void 4469 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4470 { 4471 struct ifnet *ifp = (struct ifnet *)ifh; 4472 4473 ETHER_BPF_MTAP(ifp, m); 4474 } 4475 4476 void 4477 if_vlancap(if_t ifh) 4478 { 4479 struct ifnet *ifp = (struct ifnet *)ifh; 4480 VLAN_CAPABILITIES(ifp); 4481 } 4482 4483 int 4484 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4485 { 4486 4487 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax; 4488 return (0); 4489 } 4490 4491 int 4492 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4493 { 4494 4495 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4496 return (0); 4497 } 4498 4499 int 4500 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4501 { 4502 4503 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4504 return (0); 4505 } 4506 4507 u_int 4508 if_gethwtsomax(if_t ifp) 4509 { 4510 4511 return (((struct ifnet *)ifp)->if_hw_tsomax); 4512 } 4513 4514 u_int 4515 if_gethwtsomaxsegcount(if_t ifp) 4516 { 4517 4518 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount); 4519 } 4520 4521 u_int 4522 if_gethwtsomaxsegsize(if_t ifp) 4523 { 4524 4525 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize); 4526 } 4527 4528 void 4529 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4530 { 4531 ((struct ifnet *)ifp)->if_init = init_fn; 4532 } 4533 4534 void 4535 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4536 { 4537 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4538 } 4539 4540 void 4541 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4542 { 4543 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4544 } 4545 4546 void 4547 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4548 { 4549 ((struct ifnet *)ifp)->if_transmit = start_fn; 4550 } 4551 4552 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4553 { 4554 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4555 4556 } 4557 4558 void 4559 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4560 { 4561 4562 ifp->if_get_counter = fn; 4563 } 4564 4565 /* Revisit these - These are inline functions originally. */ 4566 int 4567 drbr_inuse_drv(if_t ifh, struct buf_ring *br) 4568 { 4569 return drbr_inuse(ifh, br); 4570 } 4571 4572 struct mbuf* 4573 drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 4574 { 4575 return drbr_dequeue(ifh, br); 4576 } 4577 4578 int 4579 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 4580 { 4581 return drbr_needs_enqueue(ifh, br); 4582 } 4583 4584 int 4585 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 4586 { 4587 return drbr_enqueue(ifh, br, m); 4588 4589 } 4590