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