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