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