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