1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org> 5 * Copyright (c) 1980, 1986, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)if.c 8.5 (Berkeley) 1/9/95 33 * $FreeBSD$ 34 */ 35 36 #include "opt_bpf.h" 37 #include "opt_inet6.h" 38 #include "opt_inet.h" 39 40 #include <sys/param.h> 41 #include <sys/capsicum.h> 42 #include <sys/conf.h> 43 #include <sys/eventhandler.h> 44 #include <sys/malloc.h> 45 #include <sys/domainset.h> 46 #include <sys/sbuf.h> 47 #include <sys/bus.h> 48 #include <sys/epoch.h> 49 #include <sys/mbuf.h> 50 #include <sys/systm.h> 51 #include <sys/priv.h> 52 #include <sys/proc.h> 53 #include <sys/socket.h> 54 #include <sys/socketvar.h> 55 #include <sys/protosw.h> 56 #include <sys/kernel.h> 57 #include <sys/lock.h> 58 #include <sys/refcount.h> 59 #include <sys/module.h> 60 #include <sys/rwlock.h> 61 #include <sys/sockio.h> 62 #include <sys/syslog.h> 63 #include <sys/sysctl.h> 64 #include <sys/sysent.h> 65 #include <sys/taskqueue.h> 66 #include <sys/domain.h> 67 #include <sys/jail.h> 68 #include <sys/priv.h> 69 70 #ifdef DDB 71 #include <ddb/ddb.h> 72 #endif 73 74 #include <machine/stdarg.h> 75 #include <vm/uma.h> 76 77 #include <net/bpf.h> 78 #include <net/ethernet.h> 79 #include <net/if.h> 80 #include <net/if_arp.h> 81 #include <net/if_clone.h> 82 #include <net/if_dl.h> 83 #include <net/if_types.h> 84 #include <net/if_var.h> 85 #include <net/if_media.h> 86 #include <net/if_vlan_var.h> 87 #include <net/radix.h> 88 #include <net/route.h> 89 #include <net/route/route_ctl.h> 90 #include <net/vnet.h> 91 92 #if defined(INET) || defined(INET6) 93 #include <net/ethernet.h> 94 #include <netinet/in.h> 95 #include <netinet/in_var.h> 96 #include <netinet/ip.h> 97 #include <netinet/ip_carp.h> 98 #ifdef INET 99 #include <net/debugnet.h> 100 #include <netinet/if_ether.h> 101 #endif /* INET */ 102 #ifdef INET6 103 #include <netinet6/in6_var.h> 104 #include <netinet6/in6_ifattach.h> 105 #endif /* INET6 */ 106 #endif /* INET || INET6 */ 107 108 #include <security/mac/mac_framework.h> 109 110 /* 111 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 112 * and ifr_ifru when it is used in SIOCGIFCONF. 113 */ 114 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 115 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 116 117 __read_mostly epoch_t net_epoch_preempt; 118 #ifdef COMPAT_FREEBSD32 119 #include <sys/mount.h> 120 #include <compat/freebsd32/freebsd32.h> 121 122 struct ifreq_buffer32 { 123 uint32_t length; /* (size_t) */ 124 uint32_t buffer; /* (void *) */ 125 }; 126 127 /* 128 * Interface request structure used for socket 129 * ioctl's. All interface ioctl's must have parameter 130 * definitions which begin with ifr_name. The 131 * remainder may be interface specific. 132 */ 133 struct ifreq32 { 134 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 135 union { 136 struct sockaddr ifru_addr; 137 struct sockaddr ifru_dstaddr; 138 struct sockaddr ifru_broadaddr; 139 struct ifreq_buffer32 ifru_buffer; 140 short ifru_flags[2]; 141 short ifru_index; 142 int ifru_jid; 143 int ifru_metric; 144 int ifru_mtu; 145 int ifru_phys; 146 int ifru_media; 147 uint32_t ifru_data; 148 int ifru_cap[2]; 149 u_int ifru_fib; 150 u_char ifru_vlan_pcp; 151 } ifr_ifru; 152 }; 153 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 154 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 155 __offsetof(struct ifreq32, ifr_ifru)); 156 157 struct ifconf32 { 158 int32_t ifc_len; 159 union { 160 uint32_t ifcu_buf; 161 uint32_t ifcu_req; 162 } ifc_ifcu; 163 }; 164 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 165 166 struct ifdrv32 { 167 char ifd_name[IFNAMSIZ]; 168 uint32_t ifd_cmd; 169 uint32_t ifd_len; 170 uint32_t ifd_data; 171 }; 172 #define SIOCSDRVSPEC32 _IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32) 173 #define SIOCGDRVSPEC32 _IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32) 174 175 struct ifgroupreq32 { 176 char ifgr_name[IFNAMSIZ]; 177 u_int ifgr_len; 178 union { 179 char ifgru_group[IFNAMSIZ]; 180 uint32_t ifgru_groups; 181 } ifgr_ifgru; 182 }; 183 #define SIOCAIFGROUP32 _IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32) 184 #define SIOCGIFGROUP32 _IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32) 185 #define SIOCDIFGROUP32 _IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32) 186 #define SIOCGIFGMEMB32 _IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32) 187 188 struct ifmediareq32 { 189 char ifm_name[IFNAMSIZ]; 190 int ifm_current; 191 int ifm_mask; 192 int ifm_status; 193 int ifm_active; 194 int ifm_count; 195 uint32_t ifm_ulist; /* (int *) */ 196 }; 197 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 198 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 199 #endif /* COMPAT_FREEBSD32 */ 200 201 union ifreq_union { 202 struct ifreq ifr; 203 #ifdef COMPAT_FREEBSD32 204 struct ifreq32 ifr32; 205 #endif 206 }; 207 208 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 209 "Link layers"); 210 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 211 "Generic link-management"); 212 213 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 214 &ifqmaxlen, 0, "max send queue size"); 215 216 /* Log link state change events */ 217 static int log_link_state_change = 1; 218 219 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 220 &log_link_state_change, 0, 221 "log interface link state change events"); 222 223 /* Log promiscuous mode change events */ 224 static int log_promisc_mode_change = 1; 225 226 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 227 &log_promisc_mode_change, 1, 228 "log promiscuous mode change events"); 229 230 /* Interface description */ 231 static unsigned int ifdescr_maxlen = 1024; 232 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 233 &ifdescr_maxlen, 0, 234 "administrative maximum length for interface description"); 235 236 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 237 238 /* global sx for non-critical path ifdescr */ 239 static struct sx ifdescr_sx; 240 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 241 242 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 243 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 244 /* These are external hooks for CARP. */ 245 void (*carp_linkstate_p)(struct ifnet *ifp); 246 void (*carp_demote_adj_p)(int, char *); 247 int (*carp_master_p)(struct ifaddr *); 248 #if defined(INET) || defined(INET6) 249 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 250 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 251 const struct sockaddr *sa); 252 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 253 int (*carp_attach_p)(struct ifaddr *, int); 254 void (*carp_detach_p)(struct ifaddr *, bool); 255 #endif 256 #ifdef INET 257 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 258 #endif 259 #ifdef INET6 260 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 261 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 262 const struct in6_addr *taddr); 263 #endif 264 265 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 266 267 /* 268 * XXX: Style; these should be sorted alphabetically, and unprototyped 269 * static functions should be prototyped. Currently they are sorted by 270 * declaration order. 271 */ 272 static void if_attachdomain(void *); 273 static void if_attachdomain1(struct ifnet *); 274 static int ifconf(u_long, caddr_t); 275 static void *if_grow(void); 276 static void if_input_default(struct ifnet *, struct mbuf *); 277 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 278 static void if_route(struct ifnet *, int flag, int fam); 279 static int if_setflag(struct ifnet *, int, int, int *, int); 280 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 281 static void if_unroute(struct ifnet *, int flag, int fam); 282 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 283 static void do_link_state_change(void *, int); 284 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 285 static int if_getgroupmembers(struct ifgroupreq *); 286 static void if_delgroups(struct ifnet *); 287 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 288 static int if_detach_internal(struct ifnet *, int, struct if_clone **); 289 static void if_siocaddmulti(void *, int); 290 static void if_link_ifnet(struct ifnet *); 291 static bool if_unlink_ifnet(struct ifnet *, bool); 292 #ifdef VIMAGE 293 static int if_vmove(struct ifnet *, struct vnet *); 294 #endif 295 296 #ifdef INET6 297 /* 298 * XXX: declare here to avoid to include many inet6 related files.. 299 * should be more generalized? 300 */ 301 extern void nd6_setmtu(struct ifnet *); 302 #endif 303 304 /* ipsec helper hooks */ 305 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 306 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 307 308 VNET_DEFINE(int, if_index); 309 int ifqmaxlen = IFQ_MAXLEN; 310 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 311 VNET_DEFINE(struct ifgrouphead, ifg_head); 312 313 VNET_DEFINE_STATIC(int, if_indexlim) = 8; 314 315 /* Table of ifnet by index. */ 316 VNET_DEFINE_STATIC(struct ifnet **, ifindex_table); 317 318 #define V_if_indexlim VNET(if_indexlim) 319 #define V_ifindex_table VNET(ifindex_table) 320 321 /* 322 * The global network interface list (V_ifnet) and related state (such as 323 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 324 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 325 */ 326 struct sx ifnet_sxlock; 327 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 328 329 struct sx ifnet_detach_sxlock; 330 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 331 SX_RECURSE); 332 333 /* 334 * The allocation of network interfaces is a rather non-atomic affair; we 335 * need to select an index before we are ready to expose the interface for 336 * use, so will use this pointer value to indicate reservation. 337 */ 338 #define IFNET_HOLD (void *)(uintptr_t)(-1) 339 340 #ifdef VIMAGE 341 #define VNET_IS_SHUTTING_DOWN(_vnet) \ 342 ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE) 343 #endif 344 345 static if_com_alloc_t *if_com_alloc[256]; 346 static if_com_free_t *if_com_free[256]; 347 348 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 349 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 350 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 351 352 struct ifnet * 353 ifnet_byindex(u_short idx) 354 { 355 struct ifnet *ifp; 356 357 if (__predict_false(idx > V_if_index)) 358 return (NULL); 359 360 ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx); 361 return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp); 362 } 363 364 struct ifnet * 365 ifnet_byindex_ref(u_short idx) 366 { 367 struct ifnet *ifp; 368 369 NET_EPOCH_ASSERT(); 370 371 ifp = ifnet_byindex(idx); 372 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 373 return (NULL); 374 if (!if_try_ref(ifp)) 375 return (NULL); 376 return (ifp); 377 } 378 379 /* 380 * Allocate an ifindex array entry; return 0 on success or an error on 381 * failure. 382 */ 383 static u_short 384 ifindex_alloc(void **old) 385 { 386 u_short idx; 387 388 IFNET_WLOCK_ASSERT(); 389 /* 390 * Try to find an empty slot below V_if_index. If we fail, take the 391 * next slot. 392 */ 393 for (idx = 1; idx <= V_if_index; idx++) { 394 if (V_ifindex_table[idx] == NULL) 395 break; 396 } 397 398 /* Catch if_index overflow. */ 399 if (idx >= V_if_indexlim) { 400 *old = if_grow(); 401 return (USHRT_MAX); 402 } 403 if (idx > V_if_index) 404 V_if_index = idx; 405 return (idx); 406 } 407 408 static void 409 ifindex_free(u_short idx) 410 { 411 412 IFNET_WLOCK_ASSERT(); 413 414 V_ifindex_table[idx] = NULL; 415 while (V_if_index > 0 && 416 V_ifindex_table[V_if_index] == NULL) 417 V_if_index--; 418 } 419 420 static void 421 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 422 { 423 424 V_ifindex_table[idx] = ifp; 425 } 426 427 struct ifaddr * 428 ifaddr_byindex(u_short idx) 429 { 430 struct ifnet *ifp; 431 struct ifaddr *ifa = NULL; 432 433 NET_EPOCH_ASSERT(); 434 435 ifp = ifnet_byindex(idx); 436 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 437 ifa_ref(ifa); 438 return (ifa); 439 } 440 441 /* 442 * Network interface utility routines. 443 * 444 * Routines with ifa_ifwith* names take sockaddr *'s as 445 * parameters. 446 */ 447 448 static void 449 vnet_if_init(const void *unused __unused) 450 { 451 void *old; 452 453 CK_STAILQ_INIT(&V_ifnet); 454 CK_STAILQ_INIT(&V_ifg_head); 455 IFNET_WLOCK(); 456 old = if_grow(); /* create initial table */ 457 IFNET_WUNLOCK(); 458 epoch_wait_preempt(net_epoch_preempt); 459 free(old, M_IFNET); 460 vnet_if_clone_init(); 461 } 462 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 463 NULL); 464 465 #ifdef VIMAGE 466 static void 467 vnet_if_uninit(const void *unused __unused) 468 { 469 470 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 471 "not empty", __func__, __LINE__, &V_ifnet)); 472 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 473 "not empty", __func__, __LINE__, &V_ifg_head)); 474 475 free((caddr_t)V_ifindex_table, M_IFNET); 476 } 477 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 478 vnet_if_uninit, NULL); 479 #endif 480 481 static void 482 if_link_ifnet(struct ifnet *ifp) 483 { 484 485 IFNET_WLOCK(); 486 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 487 #ifdef VIMAGE 488 curvnet->vnet_ifcnt++; 489 #endif 490 IFNET_WUNLOCK(); 491 } 492 493 static bool 494 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 495 { 496 struct ifnet *iter; 497 int found = 0; 498 499 IFNET_WLOCK(); 500 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 501 if (iter == ifp) { 502 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 503 if (!vmove) 504 ifp->if_flags |= IFF_DYING; 505 found = 1; 506 break; 507 } 508 #ifdef VIMAGE 509 curvnet->vnet_ifcnt--; 510 #endif 511 IFNET_WUNLOCK(); 512 513 return (found); 514 } 515 516 #ifdef VIMAGE 517 static void 518 vnet_if_return(const void *unused __unused) 519 { 520 struct ifnet *ifp, *nifp; 521 struct ifnet **pending; 522 int found, i; 523 524 i = 0; 525 526 /* 527 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 528 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 529 * if_detach_internal(), which waits for NET_EPOCH callbacks to 530 * complete. We can't do that from within NET_EPOCH. 531 * 532 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 533 * read/write lock. We cannot hold the lock as we call if_vmove() 534 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 535 * ctx lock. 536 */ 537 IFNET_WLOCK(); 538 539 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 540 M_IFNET, M_WAITOK | M_ZERO); 541 542 /* Return all inherited interfaces to their parent vnets. */ 543 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 544 if (ifp->if_home_vnet != ifp->if_vnet) { 545 found = if_unlink_ifnet(ifp, true); 546 MPASS(found); 547 548 pending[i++] = ifp; 549 } 550 } 551 IFNET_WUNLOCK(); 552 553 for (int j = 0; j < i; j++) { 554 if_vmove(pending[j], pending[j]->if_home_vnet); 555 } 556 557 free(pending, M_IFNET); 558 } 559 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 560 vnet_if_return, NULL); 561 #endif 562 563 static void * 564 if_grow(void) 565 { 566 int oldlim; 567 u_int n; 568 struct ifnet **e; 569 void *old; 570 571 old = NULL; 572 IFNET_WLOCK_ASSERT(); 573 oldlim = V_if_indexlim; 574 IFNET_WUNLOCK(); 575 n = (oldlim << 1) * sizeof(*e); 576 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 577 IFNET_WLOCK(); 578 if (V_if_indexlim != oldlim) { 579 free(e, M_IFNET); 580 return (NULL); 581 } 582 if (V_ifindex_table != NULL) { 583 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 584 old = V_ifindex_table; 585 } 586 V_if_indexlim <<= 1; 587 V_ifindex_table = e; 588 return (old); 589 } 590 591 /* 592 * Allocate a struct ifnet and an index for an interface. A layer 2 593 * common structure will also be allocated if an allocation routine is 594 * registered for the passed type. 595 */ 596 static struct ifnet * 597 if_alloc_domain(u_char type, int numa_domain) 598 { 599 struct ifnet *ifp; 600 u_short idx; 601 void *old; 602 603 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 604 if (numa_domain == IF_NODOM) 605 ifp = malloc(sizeof(struct ifnet), M_IFNET, 606 M_WAITOK | M_ZERO); 607 else 608 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 609 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 610 restart: 611 IFNET_WLOCK(); 612 idx = ifindex_alloc(&old); 613 if (__predict_false(idx == USHRT_MAX)) { 614 IFNET_WUNLOCK(); 615 epoch_wait_preempt(net_epoch_preempt); 616 free(old, M_IFNET); 617 goto restart; 618 } 619 ifnet_setbyindex(idx, IFNET_HOLD); 620 IFNET_WUNLOCK(); 621 ifp->if_index = idx; 622 ifp->if_type = type; 623 ifp->if_alloctype = type; 624 ifp->if_numa_domain = numa_domain; 625 #ifdef VIMAGE 626 ifp->if_vnet = curvnet; 627 #endif 628 if (if_com_alloc[type] != NULL) { 629 ifp->if_l2com = if_com_alloc[type](type, ifp); 630 KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__, 631 type)); 632 } 633 634 IF_ADDR_LOCK_INIT(ifp); 635 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 636 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 637 ifp->if_afdata_initialized = 0; 638 IF_AFDATA_LOCK_INIT(ifp); 639 CK_STAILQ_INIT(&ifp->if_addrhead); 640 CK_STAILQ_INIT(&ifp->if_multiaddrs); 641 CK_STAILQ_INIT(&ifp->if_groups); 642 #ifdef MAC 643 mac_ifnet_init(ifp); 644 #endif 645 ifq_init(&ifp->if_snd, ifp); 646 647 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 648 for (int i = 0; i < IFCOUNTERS; i++) 649 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 650 ifp->if_get_counter = if_get_counter_default; 651 ifp->if_pcp = IFNET_PCP_NONE; 652 ifnet_setbyindex(ifp->if_index, ifp); 653 return (ifp); 654 } 655 656 struct ifnet * 657 if_alloc_dev(u_char type, device_t dev) 658 { 659 int numa_domain; 660 661 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 662 return (if_alloc_domain(type, IF_NODOM)); 663 return (if_alloc_domain(type, numa_domain)); 664 } 665 666 struct ifnet * 667 if_alloc(u_char type) 668 { 669 670 return (if_alloc_domain(type, IF_NODOM)); 671 } 672 /* 673 * Do the actual work of freeing a struct ifnet, and layer 2 common 674 * structure. This call is made when the network epoch guarantees 675 * us that nobody holds a pointer to the interface. 676 */ 677 static void 678 if_free_deferred(epoch_context_t ctx) 679 { 680 struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 681 682 KASSERT((ifp->if_flags & IFF_DYING), 683 ("%s: interface not dying", __func__)); 684 685 if (if_com_free[ifp->if_alloctype] != NULL) 686 if_com_free[ifp->if_alloctype](ifp->if_l2com, 687 ifp->if_alloctype); 688 689 #ifdef MAC 690 mac_ifnet_destroy(ifp); 691 #endif /* MAC */ 692 IF_AFDATA_DESTROY(ifp); 693 IF_ADDR_LOCK_DESTROY(ifp); 694 ifq_delete(&ifp->if_snd); 695 696 for (int i = 0; i < IFCOUNTERS; i++) 697 counter_u64_free(ifp->if_counters[i]); 698 699 free(ifp->if_description, M_IFDESCR); 700 free(ifp->if_hw_addr, M_IFADDR); 701 free(ifp, M_IFNET); 702 } 703 704 /* 705 * Deregister an interface and free the associated storage. 706 */ 707 void 708 if_free(struct ifnet *ifp) 709 { 710 711 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 712 713 /* 714 * XXXGL: An interface index is really an alias to ifp pointer. 715 * Why would we clear the alias now, and not in the deferred 716 * context? Indeed there is nothing wrong with some network 717 * thread obtaining ifp via ifnet_byindex() inside the network 718 * epoch and then dereferencing ifp while we peform if_free(), 719 * and after if_free() finished, too. 720 * 721 * The reason is the VIMAGE. For some reason it was designed 722 * to require all sockets drained before destroying, but not all 723 * ifnets. A vnet destruction calls if_vmove() on ifnet, which 724 * causes ID change. But ID change and a possible misidentification 725 * of an ifnet later is a lesser problem, as it doesn't crash kernel. 726 * A worse problem is that removed interface may outlive the vnet it 727 * belongs too! The if_free_deferred() would see ifp->if_vnet freed. 728 */ 729 CURVNET_SET_QUIET(ifp->if_vnet); 730 IFNET_WLOCK(); 731 KASSERT(ifp == ifnet_byindex(ifp->if_index), 732 ("%s: freeing unallocated ifnet", ifp->if_xname)); 733 734 ifindex_free(ifp->if_index); 735 IFNET_WUNLOCK(); 736 737 if (refcount_release(&ifp->if_refcount)) 738 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 739 CURVNET_RESTORE(); 740 } 741 742 /* 743 * Interfaces to keep an ifnet type-stable despite the possibility of the 744 * driver calling if_free(). If there are additional references, we defer 745 * freeing the underlying data structure. 746 */ 747 void 748 if_ref(struct ifnet *ifp) 749 { 750 u_int old __diagused; 751 752 /* We don't assert the ifnet list lock here, but arguably should. */ 753 old = refcount_acquire(&ifp->if_refcount); 754 KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp)); 755 } 756 757 bool 758 if_try_ref(struct ifnet *ifp) 759 { 760 NET_EPOCH_ASSERT(); 761 return (refcount_acquire_if_not_zero(&ifp->if_refcount)); 762 } 763 764 void 765 if_rele(struct ifnet *ifp) 766 { 767 768 if (!refcount_release(&ifp->if_refcount)) 769 return; 770 NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx); 771 } 772 773 void 774 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 775 { 776 777 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 778 779 if (ifq->ifq_maxlen == 0) 780 ifq->ifq_maxlen = ifqmaxlen; 781 782 ifq->altq_type = 0; 783 ifq->altq_disc = NULL; 784 ifq->altq_flags &= ALTQF_CANTCHANGE; 785 ifq->altq_tbr = NULL; 786 ifq->altq_ifp = ifp; 787 } 788 789 void 790 ifq_delete(struct ifaltq *ifq) 791 { 792 mtx_destroy(&ifq->ifq_mtx); 793 } 794 795 /* 796 * Perform generic interface initialization tasks and attach the interface 797 * to the list of "active" interfaces. If vmove flag is set on entry 798 * to if_attach_internal(), perform only a limited subset of initialization 799 * tasks, given that we are moving from one vnet to another an ifnet which 800 * has already been fully initialized. 801 * 802 * Note that if_detach_internal() removes group membership unconditionally 803 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 804 * Thus, when if_vmove() is applied to a cloned interface, group membership 805 * is lost while a cloned one always joins a group whose name is 806 * ifc->ifc_name. To recover this after if_detach_internal() and 807 * if_attach_internal(), the cloner should be specified to 808 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 809 * attempts to join a group whose name is ifc->ifc_name. 810 * 811 * XXX: 812 * - The decision to return void and thus require this function to 813 * succeed is questionable. 814 * - We should probably do more sanity checking. For instance we don't 815 * do anything to insure if_xname is unique or non-empty. 816 */ 817 void 818 if_attach(struct ifnet *ifp) 819 { 820 821 if_attach_internal(ifp, 0, NULL); 822 } 823 824 /* 825 * Compute the least common TSO limit. 826 */ 827 void 828 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 829 { 830 /* 831 * 1) If there is no limit currently, take the limit from 832 * the network adapter. 833 * 834 * 2) If the network adapter has a limit below the current 835 * limit, apply it. 836 */ 837 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 838 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 839 pmax->tsomaxbytes = ifp->if_hw_tsomax; 840 } 841 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 842 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 843 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 844 } 845 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 846 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 847 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 848 } 849 } 850 851 /* 852 * Update TSO limit of a network adapter. 853 * 854 * Returns zero if no change. Else non-zero. 855 */ 856 int 857 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 858 { 859 int retval = 0; 860 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 861 ifp->if_hw_tsomax = pmax->tsomaxbytes; 862 retval++; 863 } 864 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 865 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 866 retval++; 867 } 868 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 869 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 870 retval++; 871 } 872 return (retval); 873 } 874 875 static void 876 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 877 { 878 unsigned socksize, ifasize; 879 int namelen, masklen; 880 struct sockaddr_dl *sdl; 881 struct ifaddr *ifa; 882 883 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 884 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 885 ifp->if_xname); 886 887 #ifdef VIMAGE 888 ifp->if_vnet = curvnet; 889 if (ifp->if_home_vnet == NULL) 890 ifp->if_home_vnet = curvnet; 891 #endif 892 893 if_addgroup(ifp, IFG_ALL); 894 895 /* Restore group membership for cloned interfaces. */ 896 if (vmove && ifc != NULL) 897 if_clone_addgroup(ifp, ifc); 898 899 getmicrotime(&ifp->if_lastchange); 900 ifp->if_epoch = time_uptime; 901 902 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 903 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 904 ("transmit and qflush must both either be set or both be NULL")); 905 if (ifp->if_transmit == NULL) { 906 ifp->if_transmit = if_transmit; 907 ifp->if_qflush = if_qflush; 908 } 909 if (ifp->if_input == NULL) 910 ifp->if_input = if_input_default; 911 912 if (ifp->if_requestencap == NULL) 913 ifp->if_requestencap = if_requestencap_default; 914 915 if (!vmove) { 916 #ifdef MAC 917 mac_ifnet_create(ifp); 918 #endif 919 920 /* 921 * Create a Link Level name for this device. 922 */ 923 namelen = strlen(ifp->if_xname); 924 /* 925 * Always save enough space for any possiable name so we 926 * can do a rename in place later. 927 */ 928 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 929 socksize = masklen + ifp->if_addrlen; 930 if (socksize < sizeof(*sdl)) 931 socksize = sizeof(*sdl); 932 socksize = roundup2(socksize, sizeof(long)); 933 ifasize = sizeof(*ifa) + 2 * socksize; 934 ifa = ifa_alloc(ifasize, M_WAITOK); 935 sdl = (struct sockaddr_dl *)(ifa + 1); 936 sdl->sdl_len = socksize; 937 sdl->sdl_family = AF_LINK; 938 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 939 sdl->sdl_nlen = namelen; 940 sdl->sdl_index = ifp->if_index; 941 sdl->sdl_type = ifp->if_type; 942 ifp->if_addr = ifa; 943 ifa->ifa_ifp = ifp; 944 ifa->ifa_addr = (struct sockaddr *)sdl; 945 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 946 ifa->ifa_netmask = (struct sockaddr *)sdl; 947 sdl->sdl_len = masklen; 948 while (namelen != 0) 949 sdl->sdl_data[--namelen] = 0xff; 950 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 951 /* Reliably crash if used uninitialized. */ 952 ifp->if_broadcastaddr = NULL; 953 954 if (ifp->if_type == IFT_ETHER) { 955 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 956 M_WAITOK | M_ZERO); 957 } 958 959 #if defined(INET) || defined(INET6) 960 /* Use defaults for TSO, if nothing is set */ 961 if (ifp->if_hw_tsomax == 0 && 962 ifp->if_hw_tsomaxsegcount == 0 && 963 ifp->if_hw_tsomaxsegsize == 0) { 964 /* 965 * The TSO defaults needs to be such that an 966 * NFS mbuf list of 35 mbufs totalling just 967 * below 64K works and that a chain of mbufs 968 * can be defragged into at most 32 segments: 969 */ 970 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 971 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 972 ifp->if_hw_tsomaxsegcount = 35; 973 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 974 975 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 976 if (ifp->if_capabilities & IFCAP_TSO) { 977 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 978 ifp->if_hw_tsomax, 979 ifp->if_hw_tsomaxsegcount, 980 ifp->if_hw_tsomaxsegsize); 981 } 982 } 983 #endif 984 } 985 #ifdef VIMAGE 986 else { 987 /* 988 * Update the interface index in the link layer address 989 * of the interface. 990 */ 991 for (ifa = ifp->if_addr; ifa != NULL; 992 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 993 if (ifa->ifa_addr->sa_family == AF_LINK) { 994 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 995 sdl->sdl_index = ifp->if_index; 996 } 997 } 998 } 999 #endif 1000 1001 if_link_ifnet(ifp); 1002 1003 if (domain_init_status >= 2) 1004 if_attachdomain1(ifp); 1005 1006 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 1007 if (IS_DEFAULT_VNET(curvnet)) 1008 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 1009 1010 /* Announce the interface. */ 1011 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 1012 } 1013 1014 static void 1015 if_epochalloc(void *dummy __unused) 1016 { 1017 1018 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 1019 } 1020 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 1021 1022 static void 1023 if_attachdomain(void *dummy) 1024 { 1025 struct ifnet *ifp; 1026 1027 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 1028 if_attachdomain1(ifp); 1029 } 1030 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 1031 if_attachdomain, NULL); 1032 1033 static void 1034 if_attachdomain1(struct ifnet *ifp) 1035 { 1036 struct domain *dp; 1037 1038 /* 1039 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 1040 * cannot lock ifp->if_afdata initialization, entirely. 1041 */ 1042 IF_AFDATA_LOCK(ifp); 1043 if (ifp->if_afdata_initialized >= domain_init_status) { 1044 IF_AFDATA_UNLOCK(ifp); 1045 log(LOG_WARNING, "%s called more than once on %s\n", 1046 __func__, ifp->if_xname); 1047 return; 1048 } 1049 ifp->if_afdata_initialized = domain_init_status; 1050 IF_AFDATA_UNLOCK(ifp); 1051 1052 /* address family dependent data region */ 1053 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 1054 for (dp = domains; dp; dp = dp->dom_next) { 1055 if (dp->dom_ifattach) 1056 ifp->if_afdata[dp->dom_family] = 1057 (*dp->dom_ifattach)(ifp); 1058 } 1059 } 1060 1061 /* 1062 * Remove any unicast or broadcast network addresses from an interface. 1063 */ 1064 void 1065 if_purgeaddrs(struct ifnet *ifp) 1066 { 1067 struct ifaddr *ifa; 1068 1069 while (1) { 1070 struct epoch_tracker et; 1071 1072 NET_EPOCH_ENTER(et); 1073 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1074 if (ifa->ifa_addr->sa_family != AF_LINK) 1075 break; 1076 } 1077 NET_EPOCH_EXIT(et); 1078 1079 if (ifa == NULL) 1080 break; 1081 #ifdef INET 1082 /* XXX: Ugly!! ad hoc just for INET */ 1083 if (ifa->ifa_addr->sa_family == AF_INET) { 1084 struct ifaliasreq ifr; 1085 1086 bzero(&ifr, sizeof(ifr)); 1087 ifr.ifra_addr = *ifa->ifa_addr; 1088 if (ifa->ifa_dstaddr) 1089 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1090 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1091 NULL) == 0) 1092 continue; 1093 } 1094 #endif /* INET */ 1095 #ifdef INET6 1096 if (ifa->ifa_addr->sa_family == AF_INET6) { 1097 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1098 /* ifp_addrhead is already updated */ 1099 continue; 1100 } 1101 #endif /* INET6 */ 1102 IF_ADDR_WLOCK(ifp); 1103 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1104 IF_ADDR_WUNLOCK(ifp); 1105 ifa_free(ifa); 1106 } 1107 } 1108 1109 /* 1110 * Remove any multicast network addresses from an interface when an ifnet 1111 * is going away. 1112 */ 1113 static void 1114 if_purgemaddrs(struct ifnet *ifp) 1115 { 1116 struct ifmultiaddr *ifma; 1117 1118 IF_ADDR_WLOCK(ifp); 1119 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1120 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1121 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1122 if_delmulti_locked(ifp, ifma, 1); 1123 } 1124 IF_ADDR_WUNLOCK(ifp); 1125 } 1126 1127 /* 1128 * Detach an interface, removing it from the list of "active" interfaces. 1129 * If vmove flag is set on entry to if_detach_internal(), perform only a 1130 * limited subset of cleanup tasks, given that we are moving an ifnet from 1131 * one vnet to another, where it must be fully operational. 1132 * 1133 * XXXRW: There are some significant questions about event ordering, and 1134 * how to prevent things from starting to use the interface during detach. 1135 */ 1136 void 1137 if_detach(struct ifnet *ifp) 1138 { 1139 bool found; 1140 1141 CURVNET_SET_QUIET(ifp->if_vnet); 1142 found = if_unlink_ifnet(ifp, false); 1143 if (found) { 1144 sx_xlock(&ifnet_detach_sxlock); 1145 if_detach_internal(ifp, 0, NULL); 1146 sx_xunlock(&ifnet_detach_sxlock); 1147 } 1148 CURVNET_RESTORE(); 1149 } 1150 1151 /* 1152 * The vmove flag, if set, indicates that we are called from a callpath 1153 * that is moving an interface to a different vnet instance. 1154 * 1155 * The shutdown flag, if set, indicates that we are called in the 1156 * process of shutting down a vnet instance. Currently only the 1157 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1158 * on a vnet instance shutdown without this flag being set, e.g., when 1159 * the cloned interfaces are destoyed as first thing of teardown. 1160 */ 1161 static int 1162 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 1163 { 1164 struct ifaddr *ifa; 1165 int i; 1166 struct domain *dp; 1167 #ifdef VIMAGE 1168 bool shutdown; 1169 1170 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1171 #endif 1172 1173 /* 1174 * At this point we know the interface still was on the ifnet list 1175 * and we removed it so we are in a stable state. 1176 */ 1177 epoch_wait_preempt(net_epoch_preempt); 1178 1179 /* 1180 * Ensure all pending EPOCH(9) callbacks have been executed. This 1181 * fixes issues about late destruction of multicast options 1182 * which lead to leave group calls, which in turn access the 1183 * belonging ifnet structure: 1184 */ 1185 epoch_drain_callbacks(net_epoch_preempt); 1186 1187 /* 1188 * In any case (destroy or vmove) detach us from the groups 1189 * and remove/wait for pending events on the taskq. 1190 * XXX-BZ in theory an interface could still enqueue a taskq change? 1191 */ 1192 if_delgroups(ifp); 1193 1194 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1195 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1196 1197 /* 1198 * Check if this is a cloned interface or not. Must do even if 1199 * shutting down as a if_vmove_reclaim() would move the ifp and 1200 * the if_clone_addgroup() will have a corrupted string overwise 1201 * from a gibberish pointer. 1202 */ 1203 if (vmove && ifcp != NULL) 1204 *ifcp = if_clone_findifc(ifp); 1205 1206 if_down(ifp); 1207 1208 #ifdef VIMAGE 1209 /* 1210 * On VNET shutdown abort here as the stack teardown will do all 1211 * the work top-down for us. 1212 */ 1213 if (shutdown) { 1214 /* Give interface users the chance to clean up. */ 1215 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1216 1217 /* 1218 * In case of a vmove we are done here without error. 1219 * If we would signal an error it would lead to the same 1220 * abort as if we did not find the ifnet anymore. 1221 * if_detach() calls us in void context and does not care 1222 * about an early abort notification, so life is splendid :) 1223 */ 1224 goto finish_vnet_shutdown; 1225 } 1226 #endif 1227 1228 /* 1229 * At this point we are not tearing down a VNET and are either 1230 * going to destroy or vmove the interface and have to cleanup 1231 * accordingly. 1232 */ 1233 1234 /* 1235 * Remove routes and flush queues. 1236 */ 1237 #ifdef ALTQ 1238 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1239 altq_disable(&ifp->if_snd); 1240 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1241 altq_detach(&ifp->if_snd); 1242 #endif 1243 1244 if_purgeaddrs(ifp); 1245 1246 #ifdef INET 1247 in_ifdetach(ifp); 1248 #endif 1249 1250 #ifdef INET6 1251 /* 1252 * Remove all IPv6 kernel structs related to ifp. This should be done 1253 * before removing routing entries below, since IPv6 interface direct 1254 * routes are expected to be removed by the IPv6-specific kernel API. 1255 * Otherwise, the kernel will detect some inconsistency and bark it. 1256 */ 1257 in6_ifdetach(ifp); 1258 #endif 1259 if_purgemaddrs(ifp); 1260 1261 /* Announce that the interface is gone. */ 1262 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1263 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1264 if (IS_DEFAULT_VNET(curvnet)) 1265 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1266 1267 if (!vmove) { 1268 /* 1269 * Prevent further calls into the device driver via ifnet. 1270 */ 1271 if_dead(ifp); 1272 1273 /* 1274 * Clean up all addresses. 1275 */ 1276 IF_ADDR_WLOCK(ifp); 1277 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1278 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1279 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1280 IF_ADDR_WUNLOCK(ifp); 1281 ifa_free(ifa); 1282 } else 1283 IF_ADDR_WUNLOCK(ifp); 1284 } 1285 1286 rt_flushifroutes(ifp); 1287 1288 #ifdef VIMAGE 1289 finish_vnet_shutdown: 1290 #endif 1291 /* 1292 * We cannot hold the lock over dom_ifdetach calls as they might 1293 * sleep, for example trying to drain a callout, thus open up the 1294 * theoretical race with re-attaching. 1295 */ 1296 IF_AFDATA_LOCK(ifp); 1297 i = ifp->if_afdata_initialized; 1298 ifp->if_afdata_initialized = 0; 1299 IF_AFDATA_UNLOCK(ifp); 1300 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1301 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1302 (*dp->dom_ifdetach)(ifp, 1303 ifp->if_afdata[dp->dom_family]); 1304 ifp->if_afdata[dp->dom_family] = NULL; 1305 } 1306 } 1307 1308 return (0); 1309 } 1310 1311 #ifdef VIMAGE 1312 /* 1313 * if_vmove() performs a limited version of if_detach() in current 1314 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1315 * An attempt is made to shrink if_index in current vnet, find an 1316 * unused if_index in target vnet and calls if_grow() if necessary, 1317 * and finally find an unused if_xname for the target vnet. 1318 */ 1319 static int 1320 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1321 { 1322 struct if_clone *ifc; 1323 #ifdef DEV_BPF 1324 u_int bif_dlt, bif_hdrlen; 1325 #endif 1326 void *old; 1327 int rc; 1328 1329 #ifdef DEV_BPF 1330 /* 1331 * if_detach_internal() will call the eventhandler to notify 1332 * interface departure. That will detach if_bpf. We need to 1333 * safe the dlt and hdrlen so we can re-attach it later. 1334 */ 1335 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1336 #endif 1337 1338 /* 1339 * Detach from current vnet, but preserve LLADDR info, do not 1340 * mark as dead etc. so that the ifnet can be reattached later. 1341 * If we cannot find it, we lost the race to someone else. 1342 */ 1343 rc = if_detach_internal(ifp, 1, &ifc); 1344 if (rc != 0) 1345 return (rc); 1346 1347 /* 1348 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1349 * the if_index for that vnet if possible. 1350 * 1351 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1352 * or we'd lock on one vnet and unlock on another. 1353 */ 1354 IFNET_WLOCK(); 1355 ifindex_free(ifp->if_index); 1356 IFNET_WUNLOCK(); 1357 1358 /* 1359 * Perform interface-specific reassignment tasks, if provided by 1360 * the driver. 1361 */ 1362 if (ifp->if_reassign != NULL) 1363 ifp->if_reassign(ifp, new_vnet, NULL); 1364 1365 /* 1366 * Switch to the context of the target vnet. 1367 */ 1368 CURVNET_SET_QUIET(new_vnet); 1369 restart: 1370 IFNET_WLOCK(); 1371 ifp->if_index = ifindex_alloc(&old); 1372 if (__predict_false(ifp->if_index == USHRT_MAX)) { 1373 IFNET_WUNLOCK(); 1374 epoch_wait_preempt(net_epoch_preempt); 1375 free(old, M_IFNET); 1376 goto restart; 1377 } 1378 ifnet_setbyindex(ifp->if_index, ifp); 1379 IFNET_WUNLOCK(); 1380 1381 if_attach_internal(ifp, 1, ifc); 1382 1383 #ifdef DEV_BPF 1384 if (ifp->if_bpf == NULL) 1385 bpfattach(ifp, bif_dlt, bif_hdrlen); 1386 #endif 1387 1388 CURVNET_RESTORE(); 1389 return (0); 1390 } 1391 1392 /* 1393 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1394 */ 1395 static int 1396 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1397 { 1398 struct prison *pr; 1399 struct ifnet *difp; 1400 int error; 1401 bool found; 1402 bool shutdown; 1403 1404 /* Try to find the prison within our visibility. */ 1405 sx_slock(&allprison_lock); 1406 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1407 sx_sunlock(&allprison_lock); 1408 if (pr == NULL) 1409 return (ENXIO); 1410 prison_hold_locked(pr); 1411 mtx_unlock(&pr->pr_mtx); 1412 1413 /* Do not try to move the iface from and to the same prison. */ 1414 if (pr->pr_vnet == ifp->if_vnet) { 1415 prison_free(pr); 1416 return (EEXIST); 1417 } 1418 1419 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1420 /* XXX Lock interfaces to avoid races. */ 1421 CURVNET_SET_QUIET(pr->pr_vnet); 1422 difp = ifunit(ifname); 1423 if (difp != NULL) { 1424 CURVNET_RESTORE(); 1425 prison_free(pr); 1426 return (EEXIST); 1427 } 1428 1429 /* Make sure the VNET is stable. */ 1430 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1431 if (shutdown) { 1432 CURVNET_RESTORE(); 1433 prison_free(pr); 1434 return (EBUSY); 1435 } 1436 CURVNET_RESTORE(); 1437 1438 found = if_unlink_ifnet(ifp, true); 1439 MPASS(found); 1440 1441 /* Move the interface into the child jail/vnet. */ 1442 error = if_vmove(ifp, pr->pr_vnet); 1443 1444 /* Report the new if_xname back to the userland on success. */ 1445 if (error == 0) 1446 sprintf(ifname, "%s", ifp->if_xname); 1447 1448 prison_free(pr); 1449 return (error); 1450 } 1451 1452 static int 1453 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1454 { 1455 struct prison *pr; 1456 struct vnet *vnet_dst; 1457 struct ifnet *ifp; 1458 int error, found; 1459 bool shutdown; 1460 1461 /* Try to find the prison within our visibility. */ 1462 sx_slock(&allprison_lock); 1463 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1464 sx_sunlock(&allprison_lock); 1465 if (pr == NULL) 1466 return (ENXIO); 1467 prison_hold_locked(pr); 1468 mtx_unlock(&pr->pr_mtx); 1469 1470 /* Make sure the named iface exists in the source prison/vnet. */ 1471 CURVNET_SET(pr->pr_vnet); 1472 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1473 if (ifp == NULL) { 1474 CURVNET_RESTORE(); 1475 prison_free(pr); 1476 return (ENXIO); 1477 } 1478 1479 /* Do not try to move the iface from and to the same prison. */ 1480 vnet_dst = TD_TO_VNET(td); 1481 if (vnet_dst == ifp->if_vnet) { 1482 CURVNET_RESTORE(); 1483 prison_free(pr); 1484 return (EEXIST); 1485 } 1486 1487 /* Make sure the VNET is stable. */ 1488 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1489 if (shutdown) { 1490 CURVNET_RESTORE(); 1491 prison_free(pr); 1492 return (EBUSY); 1493 } 1494 1495 /* Get interface back from child jail/vnet. */ 1496 found = if_unlink_ifnet(ifp, true); 1497 MPASS(found); 1498 error = if_vmove(ifp, vnet_dst); 1499 CURVNET_RESTORE(); 1500 1501 /* Report the new if_xname back to the userland on success. */ 1502 if (error == 0) 1503 sprintf(ifname, "%s", ifp->if_xname); 1504 1505 prison_free(pr); 1506 return (error); 1507 } 1508 #endif /* VIMAGE */ 1509 1510 /* 1511 * Add a group to an interface 1512 */ 1513 int 1514 if_addgroup(struct ifnet *ifp, const char *groupname) 1515 { 1516 struct ifg_list *ifgl; 1517 struct ifg_group *ifg = NULL; 1518 struct ifg_member *ifgm; 1519 int new = 0; 1520 1521 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1522 groupname[strlen(groupname) - 1] <= '9') 1523 return (EINVAL); 1524 1525 IFNET_WLOCK(); 1526 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1527 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1528 IFNET_WUNLOCK(); 1529 return (EEXIST); 1530 } 1531 1532 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1533 IFNET_WUNLOCK(); 1534 return (ENOMEM); 1535 } 1536 1537 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1538 free(ifgl, M_TEMP); 1539 IFNET_WUNLOCK(); 1540 return (ENOMEM); 1541 } 1542 1543 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1544 if (!strcmp(ifg->ifg_group, groupname)) 1545 break; 1546 1547 if (ifg == NULL) { 1548 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1549 free(ifgl, M_TEMP); 1550 free(ifgm, M_TEMP); 1551 IFNET_WUNLOCK(); 1552 return (ENOMEM); 1553 } 1554 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1555 ifg->ifg_refcnt = 0; 1556 CK_STAILQ_INIT(&ifg->ifg_members); 1557 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1558 new = 1; 1559 } 1560 1561 ifg->ifg_refcnt++; 1562 ifgl->ifgl_group = ifg; 1563 ifgm->ifgm_ifp = ifp; 1564 1565 IF_ADDR_WLOCK(ifp); 1566 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1567 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1568 IF_ADDR_WUNLOCK(ifp); 1569 1570 IFNET_WUNLOCK(); 1571 1572 if (new) 1573 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1574 EVENTHANDLER_INVOKE(group_change_event, groupname); 1575 1576 return (0); 1577 } 1578 1579 /* 1580 * Helper function to remove a group out of an interface. Expects the global 1581 * ifnet lock to be write-locked, and drops it before returning. 1582 */ 1583 static void 1584 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1585 const char *groupname) 1586 { 1587 struct ifg_member *ifgm; 1588 bool freeifgl; 1589 1590 IFNET_WLOCK_ASSERT(); 1591 1592 IF_ADDR_WLOCK(ifp); 1593 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1594 IF_ADDR_WUNLOCK(ifp); 1595 1596 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1597 if (ifgm->ifgm_ifp == ifp) { 1598 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1599 ifg_member, ifgm_next); 1600 break; 1601 } 1602 } 1603 1604 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1605 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1606 ifg_next); 1607 freeifgl = true; 1608 } else { 1609 freeifgl = false; 1610 } 1611 IFNET_WUNLOCK(); 1612 1613 epoch_wait_preempt(net_epoch_preempt); 1614 if (freeifgl) { 1615 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1616 free(ifgl->ifgl_group, M_TEMP); 1617 } 1618 free(ifgm, M_TEMP); 1619 free(ifgl, M_TEMP); 1620 1621 EVENTHANDLER_INVOKE(group_change_event, groupname); 1622 } 1623 1624 /* 1625 * Remove a group from an interface 1626 */ 1627 int 1628 if_delgroup(struct ifnet *ifp, const char *groupname) 1629 { 1630 struct ifg_list *ifgl; 1631 1632 IFNET_WLOCK(); 1633 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1634 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1635 break; 1636 if (ifgl == NULL) { 1637 IFNET_WUNLOCK(); 1638 return (ENOENT); 1639 } 1640 1641 _if_delgroup_locked(ifp, ifgl, groupname); 1642 1643 return (0); 1644 } 1645 1646 /* 1647 * Remove an interface from all groups 1648 */ 1649 static void 1650 if_delgroups(struct ifnet *ifp) 1651 { 1652 struct ifg_list *ifgl; 1653 char groupname[IFNAMSIZ]; 1654 1655 IFNET_WLOCK(); 1656 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1657 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1658 _if_delgroup_locked(ifp, ifgl, groupname); 1659 IFNET_WLOCK(); 1660 } 1661 IFNET_WUNLOCK(); 1662 } 1663 1664 /* 1665 * Stores all groups from an interface in memory pointed to by ifgr. 1666 */ 1667 static int 1668 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1669 { 1670 int len, error; 1671 struct ifg_list *ifgl; 1672 struct ifg_req ifgrq, *ifgp; 1673 1674 NET_EPOCH_ASSERT(); 1675 1676 if (ifgr->ifgr_len == 0) { 1677 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1678 ifgr->ifgr_len += sizeof(struct ifg_req); 1679 return (0); 1680 } 1681 1682 len = ifgr->ifgr_len; 1683 ifgp = ifgr->ifgr_groups; 1684 /* XXX: wire */ 1685 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1686 if (len < sizeof(ifgrq)) 1687 return (EINVAL); 1688 bzero(&ifgrq, sizeof ifgrq); 1689 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1690 sizeof(ifgrq.ifgrq_group)); 1691 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1692 return (error); 1693 len -= sizeof(ifgrq); 1694 ifgp++; 1695 } 1696 1697 return (0); 1698 } 1699 1700 /* 1701 * Stores all members of a group in memory pointed to by igfr 1702 */ 1703 static int 1704 if_getgroupmembers(struct ifgroupreq *ifgr) 1705 { 1706 struct ifg_group *ifg; 1707 struct ifg_member *ifgm; 1708 struct ifg_req ifgrq, *ifgp; 1709 int len, error; 1710 1711 IFNET_RLOCK(); 1712 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1713 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1714 break; 1715 if (ifg == NULL) { 1716 IFNET_RUNLOCK(); 1717 return (ENOENT); 1718 } 1719 1720 if (ifgr->ifgr_len == 0) { 1721 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1722 ifgr->ifgr_len += sizeof(ifgrq); 1723 IFNET_RUNLOCK(); 1724 return (0); 1725 } 1726 1727 len = ifgr->ifgr_len; 1728 ifgp = ifgr->ifgr_groups; 1729 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1730 if (len < sizeof(ifgrq)) { 1731 IFNET_RUNLOCK(); 1732 return (EINVAL); 1733 } 1734 bzero(&ifgrq, sizeof ifgrq); 1735 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1736 sizeof(ifgrq.ifgrq_member)); 1737 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1738 IFNET_RUNLOCK(); 1739 return (error); 1740 } 1741 len -= sizeof(ifgrq); 1742 ifgp++; 1743 } 1744 IFNET_RUNLOCK(); 1745 1746 return (0); 1747 } 1748 1749 /* 1750 * Return counter values from counter(9)s stored in ifnet. 1751 */ 1752 uint64_t 1753 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1754 { 1755 1756 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1757 1758 return (counter_u64_fetch(ifp->if_counters[cnt])); 1759 } 1760 1761 /* 1762 * Increase an ifnet counter. Usually used for counters shared 1763 * between the stack and a driver, but function supports them all. 1764 */ 1765 void 1766 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1767 { 1768 1769 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1770 1771 counter_u64_add(ifp->if_counters[cnt], inc); 1772 } 1773 1774 /* 1775 * Copy data from ifnet to userland API structure if_data. 1776 */ 1777 void 1778 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1779 { 1780 1781 ifd->ifi_type = ifp->if_type; 1782 ifd->ifi_physical = 0; 1783 ifd->ifi_addrlen = ifp->if_addrlen; 1784 ifd->ifi_hdrlen = ifp->if_hdrlen; 1785 ifd->ifi_link_state = ifp->if_link_state; 1786 ifd->ifi_vhid = 0; 1787 ifd->ifi_datalen = sizeof(struct if_data); 1788 ifd->ifi_mtu = ifp->if_mtu; 1789 ifd->ifi_metric = ifp->if_metric; 1790 ifd->ifi_baudrate = ifp->if_baudrate; 1791 ifd->ifi_hwassist = ifp->if_hwassist; 1792 ifd->ifi_epoch = ifp->if_epoch; 1793 ifd->ifi_lastchange = ifp->if_lastchange; 1794 1795 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1796 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1797 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1798 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1799 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1800 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1801 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1802 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1803 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1804 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1805 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1806 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1807 } 1808 1809 /* 1810 * Initialization, destruction and refcounting functions for ifaddrs. 1811 */ 1812 struct ifaddr * 1813 ifa_alloc(size_t size, int flags) 1814 { 1815 struct ifaddr *ifa; 1816 1817 KASSERT(size >= sizeof(struct ifaddr), 1818 ("%s: invalid size %zu", __func__, size)); 1819 1820 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1821 if (ifa == NULL) 1822 return (NULL); 1823 1824 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1825 goto fail; 1826 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1827 goto fail; 1828 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1829 goto fail; 1830 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1831 goto fail; 1832 1833 refcount_init(&ifa->ifa_refcnt, 1); 1834 1835 return (ifa); 1836 1837 fail: 1838 /* free(NULL) is okay */ 1839 counter_u64_free(ifa->ifa_opackets); 1840 counter_u64_free(ifa->ifa_ipackets); 1841 counter_u64_free(ifa->ifa_obytes); 1842 counter_u64_free(ifa->ifa_ibytes); 1843 free(ifa, M_IFADDR); 1844 1845 return (NULL); 1846 } 1847 1848 void 1849 ifa_ref(struct ifaddr *ifa) 1850 { 1851 u_int old __diagused; 1852 1853 old = refcount_acquire(&ifa->ifa_refcnt); 1854 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1855 } 1856 1857 int 1858 ifa_try_ref(struct ifaddr *ifa) 1859 { 1860 1861 NET_EPOCH_ASSERT(); 1862 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1863 } 1864 1865 static void 1866 ifa_destroy(epoch_context_t ctx) 1867 { 1868 struct ifaddr *ifa; 1869 1870 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1871 counter_u64_free(ifa->ifa_opackets); 1872 counter_u64_free(ifa->ifa_ipackets); 1873 counter_u64_free(ifa->ifa_obytes); 1874 counter_u64_free(ifa->ifa_ibytes); 1875 free(ifa, M_IFADDR); 1876 } 1877 1878 void 1879 ifa_free(struct ifaddr *ifa) 1880 { 1881 1882 if (refcount_release(&ifa->ifa_refcnt)) 1883 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1884 } 1885 1886 /* 1887 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1888 * structs used to represent other address families, it is necessary 1889 * to perform a different comparison. 1890 */ 1891 1892 #define sa_dl_equal(a1, a2) \ 1893 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1894 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1895 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1896 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1897 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1898 1899 /* 1900 * Locate an interface based on a complete address. 1901 */ 1902 /*ARGSUSED*/ 1903 struct ifaddr * 1904 ifa_ifwithaddr(const struct sockaddr *addr) 1905 { 1906 struct ifnet *ifp; 1907 struct ifaddr *ifa; 1908 1909 NET_EPOCH_ASSERT(); 1910 1911 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1912 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1913 if (ifa->ifa_addr->sa_family != addr->sa_family) 1914 continue; 1915 if (sa_equal(addr, ifa->ifa_addr)) { 1916 goto done; 1917 } 1918 /* IP6 doesn't have broadcast */ 1919 if ((ifp->if_flags & IFF_BROADCAST) && 1920 ifa->ifa_broadaddr && 1921 ifa->ifa_broadaddr->sa_len != 0 && 1922 sa_equal(ifa->ifa_broadaddr, addr)) { 1923 goto done; 1924 } 1925 } 1926 } 1927 ifa = NULL; 1928 done: 1929 return (ifa); 1930 } 1931 1932 int 1933 ifa_ifwithaddr_check(const struct sockaddr *addr) 1934 { 1935 struct epoch_tracker et; 1936 int rc; 1937 1938 NET_EPOCH_ENTER(et); 1939 rc = (ifa_ifwithaddr(addr) != NULL); 1940 NET_EPOCH_EXIT(et); 1941 return (rc); 1942 } 1943 1944 /* 1945 * Locate an interface based on the broadcast address. 1946 */ 1947 /* ARGSUSED */ 1948 struct ifaddr * 1949 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1950 { 1951 struct ifnet *ifp; 1952 struct ifaddr *ifa; 1953 1954 NET_EPOCH_ASSERT(); 1955 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1956 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1957 continue; 1958 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1959 if (ifa->ifa_addr->sa_family != addr->sa_family) 1960 continue; 1961 if ((ifp->if_flags & IFF_BROADCAST) && 1962 ifa->ifa_broadaddr && 1963 ifa->ifa_broadaddr->sa_len != 0 && 1964 sa_equal(ifa->ifa_broadaddr, addr)) { 1965 goto done; 1966 } 1967 } 1968 } 1969 ifa = NULL; 1970 done: 1971 return (ifa); 1972 } 1973 1974 /* 1975 * Locate the point to point interface with a given destination address. 1976 */ 1977 /*ARGSUSED*/ 1978 struct ifaddr * 1979 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1980 { 1981 struct ifnet *ifp; 1982 struct ifaddr *ifa; 1983 1984 NET_EPOCH_ASSERT(); 1985 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1986 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1987 continue; 1988 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1989 continue; 1990 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1991 if (ifa->ifa_addr->sa_family != addr->sa_family) 1992 continue; 1993 if (ifa->ifa_dstaddr != NULL && 1994 sa_equal(addr, ifa->ifa_dstaddr)) { 1995 goto done; 1996 } 1997 } 1998 } 1999 ifa = NULL; 2000 done: 2001 return (ifa); 2002 } 2003 2004 /* 2005 * Find an interface on a specific network. If many, choice 2006 * is most specific found. 2007 */ 2008 struct ifaddr * 2009 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 2010 { 2011 struct ifnet *ifp; 2012 struct ifaddr *ifa; 2013 struct ifaddr *ifa_maybe = NULL; 2014 u_int af = addr->sa_family; 2015 const char *addr_data = addr->sa_data, *cplim; 2016 2017 NET_EPOCH_ASSERT(); 2018 /* 2019 * AF_LINK addresses can be looked up directly by their index number, 2020 * so do that if we can. 2021 */ 2022 if (af == AF_LINK) { 2023 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 2024 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 2025 return (ifaddr_byindex(sdl->sdl_index)); 2026 } 2027 2028 /* 2029 * Scan though each interface, looking for ones that have addresses 2030 * in this address family and the requested fib. 2031 */ 2032 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2033 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2034 continue; 2035 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2036 const char *cp, *cp2, *cp3; 2037 2038 if (ifa->ifa_addr->sa_family != af) 2039 next: continue; 2040 if (af == AF_INET && 2041 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 2042 /* 2043 * This is a bit broken as it doesn't 2044 * take into account that the remote end may 2045 * be a single node in the network we are 2046 * looking for. 2047 * The trouble is that we don't know the 2048 * netmask for the remote end. 2049 */ 2050 if (ifa->ifa_dstaddr != NULL && 2051 sa_equal(addr, ifa->ifa_dstaddr)) { 2052 goto done; 2053 } 2054 } else { 2055 /* 2056 * Scan all the bits in the ifa's address. 2057 * If a bit dissagrees with what we are 2058 * looking for, mask it with the netmask 2059 * to see if it really matters. 2060 * (A byte at a time) 2061 */ 2062 if (ifa->ifa_netmask == 0) 2063 continue; 2064 cp = addr_data; 2065 cp2 = ifa->ifa_addr->sa_data; 2066 cp3 = ifa->ifa_netmask->sa_data; 2067 cplim = ifa->ifa_netmask->sa_len 2068 + (char *)ifa->ifa_netmask; 2069 while (cp3 < cplim) 2070 if ((*cp++ ^ *cp2++) & *cp3++) 2071 goto next; /* next address! */ 2072 /* 2073 * If the netmask of what we just found 2074 * is more specific than what we had before 2075 * (if we had one), or if the virtual status 2076 * of new prefix is better than of the old one, 2077 * then remember the new one before continuing 2078 * to search for an even better one. 2079 */ 2080 if (ifa_maybe == NULL || 2081 ifa_preferred(ifa_maybe, ifa) || 2082 rn_refines((caddr_t)ifa->ifa_netmask, 2083 (caddr_t)ifa_maybe->ifa_netmask)) { 2084 ifa_maybe = ifa; 2085 } 2086 } 2087 } 2088 } 2089 ifa = ifa_maybe; 2090 ifa_maybe = NULL; 2091 done: 2092 return (ifa); 2093 } 2094 2095 /* 2096 * Find an interface address specific to an interface best matching 2097 * a given address. 2098 */ 2099 struct ifaddr * 2100 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2101 { 2102 struct ifaddr *ifa; 2103 const char *cp, *cp2, *cp3; 2104 char *cplim; 2105 struct ifaddr *ifa_maybe = NULL; 2106 u_int af = addr->sa_family; 2107 2108 if (af >= AF_MAX) 2109 return (NULL); 2110 2111 NET_EPOCH_ASSERT(); 2112 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2113 if (ifa->ifa_addr->sa_family != af) 2114 continue; 2115 if (ifa_maybe == NULL) 2116 ifa_maybe = ifa; 2117 if (ifa->ifa_netmask == 0) { 2118 if (sa_equal(addr, ifa->ifa_addr) || 2119 (ifa->ifa_dstaddr && 2120 sa_equal(addr, ifa->ifa_dstaddr))) 2121 goto done; 2122 continue; 2123 } 2124 if (ifp->if_flags & IFF_POINTOPOINT) { 2125 if (sa_equal(addr, ifa->ifa_dstaddr)) 2126 goto done; 2127 } else { 2128 cp = addr->sa_data; 2129 cp2 = ifa->ifa_addr->sa_data; 2130 cp3 = ifa->ifa_netmask->sa_data; 2131 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2132 for (; cp3 < cplim; cp3++) 2133 if ((*cp++ ^ *cp2++) & *cp3) 2134 break; 2135 if (cp3 == cplim) 2136 goto done; 2137 } 2138 } 2139 ifa = ifa_maybe; 2140 done: 2141 return (ifa); 2142 } 2143 2144 /* 2145 * See whether new ifa is better than current one: 2146 * 1) A non-virtual one is preferred over virtual. 2147 * 2) A virtual in master state preferred over any other state. 2148 * 2149 * Used in several address selecting functions. 2150 */ 2151 int 2152 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2153 { 2154 2155 return (cur->ifa_carp && (!next->ifa_carp || 2156 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2157 } 2158 2159 struct sockaddr_dl * 2160 link_alloc_sdl(size_t size, int flags) 2161 { 2162 2163 return (malloc(size, M_TEMP, flags)); 2164 } 2165 2166 void 2167 link_free_sdl(struct sockaddr *sa) 2168 { 2169 free(sa, M_TEMP); 2170 } 2171 2172 /* 2173 * Fills in given sdl with interface basic info. 2174 * Returns pointer to filled sdl. 2175 */ 2176 struct sockaddr_dl * 2177 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2178 { 2179 struct sockaddr_dl *sdl; 2180 2181 sdl = (struct sockaddr_dl *)paddr; 2182 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2183 sdl->sdl_len = sizeof(struct sockaddr_dl); 2184 sdl->sdl_family = AF_LINK; 2185 sdl->sdl_index = ifp->if_index; 2186 sdl->sdl_type = iftype; 2187 2188 return (sdl); 2189 } 2190 2191 /* 2192 * Mark an interface down and notify protocols of 2193 * the transition. 2194 */ 2195 static void 2196 if_unroute(struct ifnet *ifp, int flag, int fam) 2197 { 2198 struct ifaddr *ifa; 2199 struct epoch_tracker et; 2200 2201 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2202 2203 ifp->if_flags &= ~flag; 2204 getmicrotime(&ifp->if_lastchange); 2205 NET_EPOCH_ENTER(et); 2206 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2207 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2208 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2209 NET_EPOCH_EXIT(et); 2210 ifp->if_qflush(ifp); 2211 2212 if (ifp->if_carp) 2213 (*carp_linkstate_p)(ifp); 2214 rt_ifmsg(ifp); 2215 } 2216 2217 /* 2218 * Mark an interface up and notify protocols of 2219 * the transition. 2220 */ 2221 static void 2222 if_route(struct ifnet *ifp, int flag, int fam) 2223 { 2224 struct ifaddr *ifa; 2225 struct epoch_tracker et; 2226 2227 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2228 2229 ifp->if_flags |= flag; 2230 getmicrotime(&ifp->if_lastchange); 2231 NET_EPOCH_ENTER(et); 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 NET_EPOCH_EXIT(et); 2236 if (ifp->if_carp) 2237 (*carp_linkstate_p)(ifp); 2238 rt_ifmsg(ifp); 2239 #ifdef INET6 2240 in6_if_up(ifp); 2241 #endif 2242 } 2243 2244 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2245 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2246 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2247 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2248 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2249 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2250 int (*vlan_setcookie_p)(struct ifnet *, void *); 2251 void *(*vlan_cookie_p)(struct ifnet *); 2252 2253 /* 2254 * Handle a change in the interface link state. To avoid LORs 2255 * between driver lock and upper layer locks, as well as possible 2256 * recursions, we post event to taskqueue, and all job 2257 * is done in static do_link_state_change(). 2258 */ 2259 void 2260 if_link_state_change(struct ifnet *ifp, int link_state) 2261 { 2262 /* Return if state hasn't changed. */ 2263 if (ifp->if_link_state == link_state) 2264 return; 2265 2266 ifp->if_link_state = link_state; 2267 2268 /* XXXGL: reference ifp? */ 2269 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2270 } 2271 2272 static void 2273 do_link_state_change(void *arg, int pending) 2274 { 2275 struct ifnet *ifp; 2276 int link_state; 2277 2278 ifp = arg; 2279 link_state = ifp->if_link_state; 2280 2281 CURVNET_SET(ifp->if_vnet); 2282 rt_ifmsg(ifp); 2283 if (ifp->if_vlantrunk != NULL) 2284 (*vlan_link_state_p)(ifp); 2285 2286 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2287 ifp->if_l2com != NULL) 2288 (*ng_ether_link_state_p)(ifp, link_state); 2289 if (ifp->if_carp) 2290 (*carp_linkstate_p)(ifp); 2291 if (ifp->if_bridge) 2292 ifp->if_bridge_linkstate(ifp); 2293 if (ifp->if_lagg) 2294 (*lagg_linkstate_p)(ifp, link_state); 2295 2296 if (IS_DEFAULT_VNET(curvnet)) 2297 devctl_notify("IFNET", ifp->if_xname, 2298 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2299 NULL); 2300 if (pending > 1) 2301 if_printf(ifp, "%d link states coalesced\n", pending); 2302 if (log_link_state_change) 2303 if_printf(ifp, "link state changed to %s\n", 2304 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2305 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2306 CURVNET_RESTORE(); 2307 } 2308 2309 /* 2310 * Mark an interface down and notify protocols of 2311 * the transition. 2312 */ 2313 void 2314 if_down(struct ifnet *ifp) 2315 { 2316 2317 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2318 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2319 } 2320 2321 /* 2322 * Mark an interface up and notify protocols of 2323 * the transition. 2324 */ 2325 void 2326 if_up(struct ifnet *ifp) 2327 { 2328 2329 if_route(ifp, IFF_UP, AF_UNSPEC); 2330 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2331 } 2332 2333 /* 2334 * Flush an interface queue. 2335 */ 2336 void 2337 if_qflush(struct ifnet *ifp) 2338 { 2339 struct mbuf *m, *n; 2340 struct ifaltq *ifq; 2341 2342 ifq = &ifp->if_snd; 2343 IFQ_LOCK(ifq); 2344 #ifdef ALTQ 2345 if (ALTQ_IS_ENABLED(ifq)) 2346 ALTQ_PURGE(ifq); 2347 #endif 2348 n = ifq->ifq_head; 2349 while ((m = n) != NULL) { 2350 n = m->m_nextpkt; 2351 m_freem(m); 2352 } 2353 ifq->ifq_head = 0; 2354 ifq->ifq_tail = 0; 2355 ifq->ifq_len = 0; 2356 IFQ_UNLOCK(ifq); 2357 } 2358 2359 /* 2360 * Map interface name to interface structure pointer, with or without 2361 * returning a reference. 2362 */ 2363 struct ifnet * 2364 ifunit_ref(const char *name) 2365 { 2366 struct epoch_tracker et; 2367 struct ifnet *ifp; 2368 2369 NET_EPOCH_ENTER(et); 2370 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2371 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2372 !(ifp->if_flags & IFF_DYING)) 2373 break; 2374 } 2375 if (ifp != NULL) 2376 if_ref(ifp); 2377 NET_EPOCH_EXIT(et); 2378 return (ifp); 2379 } 2380 2381 struct ifnet * 2382 ifunit(const char *name) 2383 { 2384 struct epoch_tracker et; 2385 struct ifnet *ifp; 2386 2387 NET_EPOCH_ENTER(et); 2388 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2389 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2390 break; 2391 } 2392 NET_EPOCH_EXIT(et); 2393 return (ifp); 2394 } 2395 2396 void * 2397 ifr_buffer_get_buffer(void *data) 2398 { 2399 union ifreq_union *ifrup; 2400 2401 ifrup = data; 2402 #ifdef COMPAT_FREEBSD32 2403 if (SV_CURPROC_FLAG(SV_ILP32)) 2404 return ((void *)(uintptr_t) 2405 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2406 #endif 2407 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2408 } 2409 2410 static void 2411 ifr_buffer_set_buffer_null(void *data) 2412 { 2413 union ifreq_union *ifrup; 2414 2415 ifrup = data; 2416 #ifdef COMPAT_FREEBSD32 2417 if (SV_CURPROC_FLAG(SV_ILP32)) 2418 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2419 else 2420 #endif 2421 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2422 } 2423 2424 size_t 2425 ifr_buffer_get_length(void *data) 2426 { 2427 union ifreq_union *ifrup; 2428 2429 ifrup = data; 2430 #ifdef COMPAT_FREEBSD32 2431 if (SV_CURPROC_FLAG(SV_ILP32)) 2432 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2433 #endif 2434 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2435 } 2436 2437 static void 2438 ifr_buffer_set_length(void *data, size_t len) 2439 { 2440 union ifreq_union *ifrup; 2441 2442 ifrup = data; 2443 #ifdef COMPAT_FREEBSD32 2444 if (SV_CURPROC_FLAG(SV_ILP32)) 2445 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2446 else 2447 #endif 2448 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2449 } 2450 2451 void * 2452 ifr_data_get_ptr(void *ifrp) 2453 { 2454 union ifreq_union *ifrup; 2455 2456 ifrup = ifrp; 2457 #ifdef COMPAT_FREEBSD32 2458 if (SV_CURPROC_FLAG(SV_ILP32)) 2459 return ((void *)(uintptr_t) 2460 ifrup->ifr32.ifr_ifru.ifru_data); 2461 #endif 2462 return (ifrup->ifr.ifr_ifru.ifru_data); 2463 } 2464 2465 /* 2466 * Hardware specific interface ioctls. 2467 */ 2468 int 2469 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2470 { 2471 struct ifreq *ifr; 2472 int error = 0, do_ifup = 0; 2473 int new_flags, temp_flags; 2474 size_t namelen, onamelen; 2475 size_t descrlen; 2476 char *descrbuf, *odescrbuf; 2477 char new_name[IFNAMSIZ]; 2478 char old_name[IFNAMSIZ], strbuf[IFNAMSIZ + 8]; 2479 struct ifaddr *ifa; 2480 struct sockaddr_dl *sdl; 2481 2482 ifr = (struct ifreq *)data; 2483 switch (cmd) { 2484 case SIOCGIFINDEX: 2485 ifr->ifr_index = ifp->if_index; 2486 break; 2487 2488 case SIOCGIFFLAGS: 2489 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2490 ifr->ifr_flags = temp_flags & 0xffff; 2491 ifr->ifr_flagshigh = temp_flags >> 16; 2492 break; 2493 2494 case SIOCGIFCAP: 2495 ifr->ifr_reqcap = ifp->if_capabilities; 2496 ifr->ifr_curcap = ifp->if_capenable; 2497 break; 2498 2499 case SIOCGIFDATA: 2500 { 2501 struct if_data ifd; 2502 2503 /* Ensure uninitialised padding is not leaked. */ 2504 memset(&ifd, 0, sizeof(ifd)); 2505 2506 if_data_copy(ifp, &ifd); 2507 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2508 break; 2509 } 2510 2511 #ifdef MAC 2512 case SIOCGIFMAC: 2513 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2514 break; 2515 #endif 2516 2517 case SIOCGIFMETRIC: 2518 ifr->ifr_metric = ifp->if_metric; 2519 break; 2520 2521 case SIOCGIFMTU: 2522 ifr->ifr_mtu = ifp->if_mtu; 2523 break; 2524 2525 case SIOCGIFPHYS: 2526 /* XXXGL: did this ever worked? */ 2527 ifr->ifr_phys = 0; 2528 break; 2529 2530 case SIOCGIFDESCR: 2531 error = 0; 2532 sx_slock(&ifdescr_sx); 2533 if (ifp->if_description == NULL) 2534 error = ENOMSG; 2535 else { 2536 /* space for terminating nul */ 2537 descrlen = strlen(ifp->if_description) + 1; 2538 if (ifr_buffer_get_length(ifr) < descrlen) 2539 ifr_buffer_set_buffer_null(ifr); 2540 else 2541 error = copyout(ifp->if_description, 2542 ifr_buffer_get_buffer(ifr), descrlen); 2543 ifr_buffer_set_length(ifr, descrlen); 2544 } 2545 sx_sunlock(&ifdescr_sx); 2546 break; 2547 2548 case SIOCSIFDESCR: 2549 error = priv_check(td, PRIV_NET_SETIFDESCR); 2550 if (error) 2551 return (error); 2552 2553 /* 2554 * Copy only (length-1) bytes to make sure that 2555 * if_description is always nul terminated. The 2556 * length parameter is supposed to count the 2557 * terminating nul in. 2558 */ 2559 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2560 return (ENAMETOOLONG); 2561 else if (ifr_buffer_get_length(ifr) == 0) 2562 descrbuf = NULL; 2563 else { 2564 descrbuf = malloc(ifr_buffer_get_length(ifr), 2565 M_IFDESCR, M_WAITOK | M_ZERO); 2566 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2567 ifr_buffer_get_length(ifr) - 1); 2568 if (error) { 2569 free(descrbuf, M_IFDESCR); 2570 break; 2571 } 2572 } 2573 2574 sx_xlock(&ifdescr_sx); 2575 odescrbuf = ifp->if_description; 2576 ifp->if_description = descrbuf; 2577 sx_xunlock(&ifdescr_sx); 2578 2579 getmicrotime(&ifp->if_lastchange); 2580 free(odescrbuf, M_IFDESCR); 2581 break; 2582 2583 case SIOCGIFFIB: 2584 ifr->ifr_fib = ifp->if_fib; 2585 break; 2586 2587 case SIOCSIFFIB: 2588 error = priv_check(td, PRIV_NET_SETIFFIB); 2589 if (error) 2590 return (error); 2591 if (ifr->ifr_fib >= rt_numfibs) 2592 return (EINVAL); 2593 2594 ifp->if_fib = ifr->ifr_fib; 2595 break; 2596 2597 case SIOCSIFFLAGS: 2598 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2599 if (error) 2600 return (error); 2601 /* 2602 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2603 * check, so we don't need special handling here yet. 2604 */ 2605 new_flags = (ifr->ifr_flags & 0xffff) | 2606 (ifr->ifr_flagshigh << 16); 2607 if (ifp->if_flags & IFF_UP && 2608 (new_flags & IFF_UP) == 0) { 2609 if_down(ifp); 2610 } else if (new_flags & IFF_UP && 2611 (ifp->if_flags & IFF_UP) == 0) { 2612 do_ifup = 1; 2613 } 2614 /* See if permanently promiscuous mode bit is about to flip */ 2615 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2616 if (new_flags & IFF_PPROMISC) 2617 ifp->if_flags |= IFF_PROMISC; 2618 else if (ifp->if_pcount == 0) 2619 ifp->if_flags &= ~IFF_PROMISC; 2620 if (log_promisc_mode_change) 2621 if_printf(ifp, "permanently promiscuous mode %s\n", 2622 ((new_flags & IFF_PPROMISC) ? 2623 "enabled" : "disabled")); 2624 } 2625 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2626 (new_flags &~ IFF_CANTCHANGE); 2627 if (ifp->if_ioctl) { 2628 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2629 } 2630 if (do_ifup) 2631 if_up(ifp); 2632 getmicrotime(&ifp->if_lastchange); 2633 break; 2634 2635 case SIOCSIFCAP: 2636 error = priv_check(td, PRIV_NET_SETIFCAP); 2637 if (error) 2638 return (error); 2639 if (ifp->if_ioctl == NULL) 2640 return (EOPNOTSUPP); 2641 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2642 return (EINVAL); 2643 error = (*ifp->if_ioctl)(ifp, cmd, data); 2644 if (error == 0) 2645 getmicrotime(&ifp->if_lastchange); 2646 break; 2647 2648 #ifdef MAC 2649 case SIOCSIFMAC: 2650 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2651 break; 2652 #endif 2653 2654 case SIOCSIFNAME: 2655 error = priv_check(td, PRIV_NET_SETIFNAME); 2656 if (error) 2657 return (error); 2658 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2659 NULL); 2660 if (error != 0) 2661 return (error); 2662 if (new_name[0] == '\0') 2663 return (EINVAL); 2664 if (strcmp(new_name, ifp->if_xname) == 0) 2665 break; 2666 if (ifunit(new_name) != NULL) 2667 return (EEXIST); 2668 2669 /* 2670 * XXX: Locking. Nothing else seems to lock if_flags, 2671 * and there are numerous other races with the 2672 * ifunit() checks not being atomic with namespace 2673 * changes (renames, vmoves, if_attach, etc). 2674 */ 2675 ifp->if_flags |= IFF_RENAMING; 2676 2677 /* Announce the departure of the interface. */ 2678 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2679 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2680 2681 if_printf(ifp, "changing name to '%s'\n", new_name); 2682 2683 IF_ADDR_WLOCK(ifp); 2684 strlcpy(old_name, ifp->if_xname, sizeof(old_name)); 2685 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2686 ifa = ifp->if_addr; 2687 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2688 namelen = strlen(new_name); 2689 onamelen = sdl->sdl_nlen; 2690 /* 2691 * Move the address if needed. This is safe because we 2692 * allocate space for a name of length IFNAMSIZ when we 2693 * create this in if_attach(). 2694 */ 2695 if (namelen != onamelen) { 2696 bcopy(sdl->sdl_data + onamelen, 2697 sdl->sdl_data + namelen, sdl->sdl_alen); 2698 } 2699 bcopy(new_name, sdl->sdl_data, namelen); 2700 sdl->sdl_nlen = namelen; 2701 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2702 bzero(sdl->sdl_data, onamelen); 2703 while (namelen != 0) 2704 sdl->sdl_data[--namelen] = 0xff; 2705 IF_ADDR_WUNLOCK(ifp); 2706 2707 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2708 /* Announce the return of the interface. */ 2709 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2710 2711 ifp->if_flags &= ~IFF_RENAMING; 2712 2713 snprintf(strbuf, sizeof(strbuf), "name=%s", new_name); 2714 devctl_notify("IFNET", old_name, "RENAME", strbuf); 2715 break; 2716 2717 #ifdef VIMAGE 2718 case SIOCSIFVNET: 2719 error = priv_check(td, PRIV_NET_SETIFVNET); 2720 if (error) 2721 return (error); 2722 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2723 break; 2724 #endif 2725 2726 case SIOCSIFMETRIC: 2727 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2728 if (error) 2729 return (error); 2730 ifp->if_metric = ifr->ifr_metric; 2731 getmicrotime(&ifp->if_lastchange); 2732 break; 2733 2734 case SIOCSIFPHYS: 2735 error = priv_check(td, PRIV_NET_SETIFPHYS); 2736 if (error) 2737 return (error); 2738 if (ifp->if_ioctl == NULL) 2739 return (EOPNOTSUPP); 2740 error = (*ifp->if_ioctl)(ifp, cmd, data); 2741 if (error == 0) 2742 getmicrotime(&ifp->if_lastchange); 2743 break; 2744 2745 case SIOCSIFMTU: 2746 { 2747 u_long oldmtu = ifp->if_mtu; 2748 2749 error = priv_check(td, PRIV_NET_SETIFMTU); 2750 if (error) 2751 return (error); 2752 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2753 return (EINVAL); 2754 if (ifp->if_ioctl == NULL) 2755 return (EOPNOTSUPP); 2756 /* Disallow MTU changes on bridge member interfaces. */ 2757 if (ifp->if_bridge) 2758 return (EOPNOTSUPP); 2759 error = (*ifp->if_ioctl)(ifp, cmd, data); 2760 if (error == 0) { 2761 getmicrotime(&ifp->if_lastchange); 2762 rt_ifmsg(ifp); 2763 #ifdef INET 2764 DEBUGNET_NOTIFY_MTU(ifp); 2765 #endif 2766 } 2767 /* 2768 * If the link MTU changed, do network layer specific procedure. 2769 */ 2770 if (ifp->if_mtu != oldmtu) { 2771 #ifdef INET6 2772 nd6_setmtu(ifp); 2773 #endif 2774 rt_updatemtu(ifp); 2775 } 2776 break; 2777 } 2778 2779 case SIOCADDMULTI: 2780 case SIOCDELMULTI: 2781 if (cmd == SIOCADDMULTI) 2782 error = priv_check(td, PRIV_NET_ADDMULTI); 2783 else 2784 error = priv_check(td, PRIV_NET_DELMULTI); 2785 if (error) 2786 return (error); 2787 2788 /* Don't allow group membership on non-multicast interfaces. */ 2789 if ((ifp->if_flags & IFF_MULTICAST) == 0) 2790 return (EOPNOTSUPP); 2791 2792 /* Don't let users screw up protocols' entries. */ 2793 if (ifr->ifr_addr.sa_family != AF_LINK) 2794 return (EINVAL); 2795 2796 if (cmd == SIOCADDMULTI) { 2797 struct epoch_tracker et; 2798 struct ifmultiaddr *ifma; 2799 2800 /* 2801 * Userland is only permitted to join groups once 2802 * via the if_addmulti() KPI, because it cannot hold 2803 * struct ifmultiaddr * between calls. It may also 2804 * lose a race while we check if the membership 2805 * already exists. 2806 */ 2807 NET_EPOCH_ENTER(et); 2808 ifma = if_findmulti(ifp, &ifr->ifr_addr); 2809 NET_EPOCH_EXIT(et); 2810 if (ifma != NULL) 2811 error = EADDRINUSE; 2812 else 2813 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 2814 } else { 2815 error = if_delmulti(ifp, &ifr->ifr_addr); 2816 } 2817 if (error == 0) 2818 getmicrotime(&ifp->if_lastchange); 2819 break; 2820 2821 case SIOCSIFPHYADDR: 2822 case SIOCDIFPHYADDR: 2823 #ifdef INET6 2824 case SIOCSIFPHYADDR_IN6: 2825 #endif 2826 case SIOCSIFMEDIA: 2827 case SIOCSIFGENERIC: 2828 error = priv_check(td, PRIV_NET_HWIOCTL); 2829 if (error) 2830 return (error); 2831 if (ifp->if_ioctl == NULL) 2832 return (EOPNOTSUPP); 2833 error = (*ifp->if_ioctl)(ifp, cmd, data); 2834 if (error == 0) 2835 getmicrotime(&ifp->if_lastchange); 2836 break; 2837 2838 case SIOCGIFSTATUS: 2839 case SIOCGIFPSRCADDR: 2840 case SIOCGIFPDSTADDR: 2841 case SIOCGIFMEDIA: 2842 case SIOCGIFXMEDIA: 2843 case SIOCGIFGENERIC: 2844 case SIOCGIFRSSKEY: 2845 case SIOCGIFRSSHASH: 2846 case SIOCGIFDOWNREASON: 2847 if (ifp->if_ioctl == NULL) 2848 return (EOPNOTSUPP); 2849 error = (*ifp->if_ioctl)(ifp, cmd, data); 2850 break; 2851 2852 case SIOCSIFLLADDR: 2853 error = priv_check(td, PRIV_NET_SETLLADDR); 2854 if (error) 2855 return (error); 2856 error = if_setlladdr(ifp, 2857 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 2858 break; 2859 2860 case SIOCGHWADDR: 2861 error = if_gethwaddr(ifp, ifr); 2862 break; 2863 2864 case SIOCAIFGROUP: 2865 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2866 if (error) 2867 return (error); 2868 error = if_addgroup(ifp, 2869 ((struct ifgroupreq *)data)->ifgr_group); 2870 if (error != 0) 2871 return (error); 2872 break; 2873 2874 case SIOCGIFGROUP: 2875 { 2876 struct epoch_tracker et; 2877 2878 NET_EPOCH_ENTER(et); 2879 error = if_getgroup((struct ifgroupreq *)data, ifp); 2880 NET_EPOCH_EXIT(et); 2881 break; 2882 } 2883 2884 case SIOCDIFGROUP: 2885 error = priv_check(td, PRIV_NET_DELIFGROUP); 2886 if (error) 2887 return (error); 2888 error = if_delgroup(ifp, 2889 ((struct ifgroupreq *)data)->ifgr_group); 2890 if (error != 0) 2891 return (error); 2892 break; 2893 2894 default: 2895 error = ENOIOCTL; 2896 break; 2897 } 2898 return (error); 2899 } 2900 2901 /* 2902 * Interface ioctls. 2903 */ 2904 int 2905 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2906 { 2907 #ifdef COMPAT_FREEBSD32 2908 union { 2909 struct ifconf ifc; 2910 struct ifdrv ifd; 2911 struct ifgroupreq ifgr; 2912 struct ifmediareq ifmr; 2913 } thunk; 2914 u_long saved_cmd; 2915 struct ifconf32 *ifc32; 2916 struct ifdrv32 *ifd32; 2917 struct ifgroupreq32 *ifgr32; 2918 struct ifmediareq32 *ifmr32; 2919 #endif 2920 struct ifnet *ifp; 2921 struct ifreq *ifr; 2922 int error; 2923 int oif_flags; 2924 #ifdef VIMAGE 2925 bool shutdown; 2926 #endif 2927 2928 CURVNET_SET(so->so_vnet); 2929 #ifdef VIMAGE 2930 /* Make sure the VNET is stable. */ 2931 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2932 if (shutdown) { 2933 CURVNET_RESTORE(); 2934 return (EBUSY); 2935 } 2936 #endif 2937 2938 #ifdef COMPAT_FREEBSD32 2939 saved_cmd = cmd; 2940 switch (cmd) { 2941 case SIOCGIFCONF32: 2942 ifc32 = (struct ifconf32 *)data; 2943 thunk.ifc.ifc_len = ifc32->ifc_len; 2944 thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2945 data = (caddr_t)&thunk.ifc; 2946 cmd = SIOCGIFCONF; 2947 break; 2948 case SIOCGDRVSPEC32: 2949 case SIOCSDRVSPEC32: 2950 ifd32 = (struct ifdrv32 *)data; 2951 memcpy(thunk.ifd.ifd_name, ifd32->ifd_name, 2952 sizeof(thunk.ifd.ifd_name)); 2953 thunk.ifd.ifd_cmd = ifd32->ifd_cmd; 2954 thunk.ifd.ifd_len = ifd32->ifd_len; 2955 thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data); 2956 data = (caddr_t)&thunk.ifd; 2957 cmd = _IOC_NEWTYPE(cmd, struct ifdrv); 2958 break; 2959 case SIOCAIFGROUP32: 2960 case SIOCGIFGROUP32: 2961 case SIOCDIFGROUP32: 2962 case SIOCGIFGMEMB32: 2963 ifgr32 = (struct ifgroupreq32 *)data; 2964 memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name, 2965 sizeof(thunk.ifgr.ifgr_name)); 2966 thunk.ifgr.ifgr_len = ifgr32->ifgr_len; 2967 switch (cmd) { 2968 case SIOCAIFGROUP32: 2969 case SIOCDIFGROUP32: 2970 memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group, 2971 sizeof(thunk.ifgr.ifgr_group)); 2972 break; 2973 case SIOCGIFGROUP32: 2974 case SIOCGIFGMEMB32: 2975 thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups); 2976 break; 2977 } 2978 data = (caddr_t)&thunk.ifgr; 2979 cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq); 2980 break; 2981 case SIOCGIFMEDIA32: 2982 case SIOCGIFXMEDIA32: 2983 ifmr32 = (struct ifmediareq32 *)data; 2984 memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name, 2985 sizeof(thunk.ifmr.ifm_name)); 2986 thunk.ifmr.ifm_current = ifmr32->ifm_current; 2987 thunk.ifmr.ifm_mask = ifmr32->ifm_mask; 2988 thunk.ifmr.ifm_status = ifmr32->ifm_status; 2989 thunk.ifmr.ifm_active = ifmr32->ifm_active; 2990 thunk.ifmr.ifm_count = ifmr32->ifm_count; 2991 thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist); 2992 data = (caddr_t)&thunk.ifmr; 2993 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 2994 break; 2995 } 2996 #endif 2997 2998 switch (cmd) { 2999 case SIOCGIFCONF: 3000 error = ifconf(cmd, data); 3001 goto out_noref; 3002 } 3003 3004 ifr = (struct ifreq *)data; 3005 switch (cmd) { 3006 #ifdef VIMAGE 3007 case SIOCSIFRVNET: 3008 error = priv_check(td, PRIV_NET_SETIFVNET); 3009 if (error == 0) 3010 error = if_vmove_reclaim(td, ifr->ifr_name, 3011 ifr->ifr_jid); 3012 goto out_noref; 3013 #endif 3014 case SIOCIFCREATE: 3015 case SIOCIFCREATE2: 3016 error = priv_check(td, PRIV_NET_IFCREATE); 3017 if (error == 0) 3018 error = if_clone_create(ifr->ifr_name, 3019 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3020 ifr_data_get_ptr(ifr) : NULL); 3021 goto out_noref; 3022 case SIOCIFDESTROY: 3023 error = priv_check(td, PRIV_NET_IFDESTROY); 3024 3025 if (error == 0) { 3026 sx_xlock(&ifnet_detach_sxlock); 3027 error = if_clone_destroy(ifr->ifr_name); 3028 sx_xunlock(&ifnet_detach_sxlock); 3029 } 3030 goto out_noref; 3031 3032 case SIOCIFGCLONERS: 3033 error = if_clone_list((struct if_clonereq *)data); 3034 goto out_noref; 3035 3036 case SIOCGIFGMEMB: 3037 error = if_getgroupmembers((struct ifgroupreq *)data); 3038 goto out_noref; 3039 3040 #if defined(INET) || defined(INET6) 3041 case SIOCSVH: 3042 case SIOCGVH: 3043 if (carp_ioctl_p == NULL) 3044 error = EPROTONOSUPPORT; 3045 else 3046 error = (*carp_ioctl_p)(ifr, cmd, td); 3047 goto out_noref; 3048 #endif 3049 } 3050 3051 ifp = ifunit_ref(ifr->ifr_name); 3052 if (ifp == NULL) { 3053 error = ENXIO; 3054 goto out_noref; 3055 } 3056 3057 error = ifhwioctl(cmd, ifp, data, td); 3058 if (error != ENOIOCTL) 3059 goto out_ref; 3060 3061 oif_flags = ifp->if_flags; 3062 if (so->so_proto == NULL) { 3063 error = EOPNOTSUPP; 3064 goto out_ref; 3065 } 3066 3067 /* 3068 * Pass the request on to the socket control method, and if the 3069 * latter returns EOPNOTSUPP, directly to the interface. 3070 * 3071 * Make an exception for the legacy SIOCSIF* requests. Drivers 3072 * trust SIOCSIFADDR et al to come from an already privileged 3073 * layer, and do not perform any credentials checks or input 3074 * validation. 3075 */ 3076 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 3077 ifp, td)); 3078 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3079 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3080 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3081 error = (*ifp->if_ioctl)(ifp, cmd, data); 3082 3083 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3084 #ifdef INET6 3085 if (ifp->if_flags & IFF_UP) 3086 in6_if_up(ifp); 3087 #endif 3088 } 3089 3090 out_ref: 3091 if_rele(ifp); 3092 out_noref: 3093 CURVNET_RESTORE(); 3094 #ifdef COMPAT_FREEBSD32 3095 if (error != 0) 3096 return (error); 3097 switch (saved_cmd) { 3098 case SIOCGIFCONF32: 3099 ifc32->ifc_len = thunk.ifc.ifc_len; 3100 break; 3101 case SIOCGDRVSPEC32: 3102 /* 3103 * SIOCGDRVSPEC is IOWR, but nothing actually touches 3104 * the struct so just assert that ifd_len (the only 3105 * field it might make sense to update) hasn't 3106 * changed. 3107 */ 3108 KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len, 3109 ("ifd_len was updated %u -> %zu", ifd32->ifd_len, 3110 thunk.ifd.ifd_len)); 3111 break; 3112 case SIOCGIFGROUP32: 3113 case SIOCGIFGMEMB32: 3114 ifgr32->ifgr_len = thunk.ifgr.ifgr_len; 3115 break; 3116 case SIOCGIFMEDIA32: 3117 case SIOCGIFXMEDIA32: 3118 ifmr32->ifm_current = thunk.ifmr.ifm_current; 3119 ifmr32->ifm_mask = thunk.ifmr.ifm_mask; 3120 ifmr32->ifm_status = thunk.ifmr.ifm_status; 3121 ifmr32->ifm_active = thunk.ifmr.ifm_active; 3122 ifmr32->ifm_count = thunk.ifmr.ifm_count; 3123 break; 3124 } 3125 #endif 3126 return (error); 3127 } 3128 3129 /* 3130 * The code common to handling reference counted flags, 3131 * e.g., in ifpromisc() and if_allmulti(). 3132 * The "pflag" argument can specify a permanent mode flag to check, 3133 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3134 * 3135 * Only to be used on stack-owned flags, not driver-owned flags. 3136 */ 3137 static int 3138 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3139 { 3140 struct ifreq ifr; 3141 int error; 3142 int oldflags, oldcount; 3143 3144 /* Sanity checks to catch programming errors */ 3145 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3146 ("%s: setting driver-owned flag %d", __func__, flag)); 3147 3148 if (onswitch) 3149 KASSERT(*refcount >= 0, 3150 ("%s: increment negative refcount %d for flag %d", 3151 __func__, *refcount, flag)); 3152 else 3153 KASSERT(*refcount > 0, 3154 ("%s: decrement non-positive refcount %d for flag %d", 3155 __func__, *refcount, flag)); 3156 3157 /* In case this mode is permanent, just touch refcount */ 3158 if (ifp->if_flags & pflag) { 3159 *refcount += onswitch ? 1 : -1; 3160 return (0); 3161 } 3162 3163 /* Save ifnet parameters for if_ioctl() may fail */ 3164 oldcount = *refcount; 3165 oldflags = ifp->if_flags; 3166 3167 /* 3168 * See if we aren't the only and touching refcount is enough. 3169 * Actually toggle interface flag if we are the first or last. 3170 */ 3171 if (onswitch) { 3172 if ((*refcount)++) 3173 return (0); 3174 ifp->if_flags |= flag; 3175 } else { 3176 if (--(*refcount)) 3177 return (0); 3178 ifp->if_flags &= ~flag; 3179 } 3180 3181 /* Call down the driver since we've changed interface flags */ 3182 if (ifp->if_ioctl == NULL) { 3183 error = EOPNOTSUPP; 3184 goto recover; 3185 } 3186 ifr.ifr_flags = ifp->if_flags & 0xffff; 3187 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3188 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3189 if (error) 3190 goto recover; 3191 /* Notify userland that interface flags have changed */ 3192 rt_ifmsg(ifp); 3193 return (0); 3194 3195 recover: 3196 /* Recover after driver error */ 3197 *refcount = oldcount; 3198 ifp->if_flags = oldflags; 3199 return (error); 3200 } 3201 3202 /* 3203 * Set/clear promiscuous mode on interface ifp based on the truth value 3204 * of pswitch. The calls are reference counted so that only the first 3205 * "on" request actually has an effect, as does the final "off" request. 3206 * Results are undefined if the "off" and "on" requests are not matched. 3207 */ 3208 int 3209 ifpromisc(struct ifnet *ifp, int pswitch) 3210 { 3211 int error; 3212 int oldflags = ifp->if_flags; 3213 3214 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3215 &ifp->if_pcount, pswitch); 3216 /* If promiscuous mode status has changed, log a message */ 3217 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3218 log_promisc_mode_change) 3219 if_printf(ifp, "promiscuous mode %s\n", 3220 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3221 return (error); 3222 } 3223 3224 /* 3225 * Return interface configuration 3226 * of system. List may be used 3227 * in later ioctl's (above) to get 3228 * other information. 3229 */ 3230 /*ARGSUSED*/ 3231 static int 3232 ifconf(u_long cmd, caddr_t data) 3233 { 3234 struct ifconf *ifc = (struct ifconf *)data; 3235 struct ifnet *ifp; 3236 struct ifaddr *ifa; 3237 struct ifreq ifr; 3238 struct sbuf *sb; 3239 int error, full = 0, valid_len, max_len; 3240 3241 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3242 max_len = maxphys - 1; 3243 3244 /* Prevent hostile input from being able to crash the system */ 3245 if (ifc->ifc_len <= 0) 3246 return (EINVAL); 3247 3248 again: 3249 if (ifc->ifc_len <= max_len) { 3250 max_len = ifc->ifc_len; 3251 full = 1; 3252 } 3253 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3254 max_len = 0; 3255 valid_len = 0; 3256 3257 IFNET_RLOCK(); 3258 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3259 struct epoch_tracker et; 3260 int addrs; 3261 3262 /* 3263 * Zero the ifr to make sure we don't disclose the contents 3264 * of the stack. 3265 */ 3266 memset(&ifr, 0, sizeof(ifr)); 3267 3268 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3269 >= sizeof(ifr.ifr_name)) { 3270 sbuf_delete(sb); 3271 IFNET_RUNLOCK(); 3272 return (ENAMETOOLONG); 3273 } 3274 3275 addrs = 0; 3276 NET_EPOCH_ENTER(et); 3277 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3278 struct sockaddr *sa = ifa->ifa_addr; 3279 3280 if (prison_if(curthread->td_ucred, sa) != 0) 3281 continue; 3282 addrs++; 3283 if (sa->sa_len <= sizeof(*sa)) { 3284 if (sa->sa_len < sizeof(*sa)) { 3285 memset(&ifr.ifr_ifru.ifru_addr, 0, 3286 sizeof(ifr.ifr_ifru.ifru_addr)); 3287 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3288 sa->sa_len); 3289 } else 3290 ifr.ifr_ifru.ifru_addr = *sa; 3291 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3292 max_len += sizeof(ifr); 3293 } else { 3294 sbuf_bcat(sb, &ifr, 3295 offsetof(struct ifreq, ifr_addr)); 3296 max_len += offsetof(struct ifreq, ifr_addr); 3297 sbuf_bcat(sb, sa, sa->sa_len); 3298 max_len += sa->sa_len; 3299 } 3300 3301 if (sbuf_error(sb) == 0) 3302 valid_len = sbuf_len(sb); 3303 } 3304 NET_EPOCH_EXIT(et); 3305 if (addrs == 0) { 3306 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3307 max_len += sizeof(ifr); 3308 3309 if (sbuf_error(sb) == 0) 3310 valid_len = sbuf_len(sb); 3311 } 3312 } 3313 IFNET_RUNLOCK(); 3314 3315 /* 3316 * If we didn't allocate enough space (uncommon), try again. If 3317 * we have already allocated as much space as we are allowed, 3318 * return what we've got. 3319 */ 3320 if (valid_len != max_len && !full) { 3321 sbuf_delete(sb); 3322 goto again; 3323 } 3324 3325 ifc->ifc_len = valid_len; 3326 sbuf_finish(sb); 3327 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3328 sbuf_delete(sb); 3329 return (error); 3330 } 3331 3332 /* 3333 * Just like ifpromisc(), but for all-multicast-reception mode. 3334 */ 3335 int 3336 if_allmulti(struct ifnet *ifp, int onswitch) 3337 { 3338 3339 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3340 } 3341 3342 struct ifmultiaddr * 3343 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3344 { 3345 struct ifmultiaddr *ifma; 3346 3347 IF_ADDR_LOCK_ASSERT(ifp); 3348 3349 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3350 if (sa->sa_family == AF_LINK) { 3351 if (sa_dl_equal(ifma->ifma_addr, sa)) 3352 break; 3353 } else { 3354 if (sa_equal(ifma->ifma_addr, sa)) 3355 break; 3356 } 3357 } 3358 3359 return ifma; 3360 } 3361 3362 /* 3363 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3364 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3365 * the ifnet multicast address list here, so the caller must do that and 3366 * other setup work (such as notifying the device driver). The reference 3367 * count is initialized to 1. 3368 */ 3369 static struct ifmultiaddr * 3370 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3371 int mflags) 3372 { 3373 struct ifmultiaddr *ifma; 3374 struct sockaddr *dupsa; 3375 3376 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3377 M_ZERO); 3378 if (ifma == NULL) 3379 return (NULL); 3380 3381 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3382 if (dupsa == NULL) { 3383 free(ifma, M_IFMADDR); 3384 return (NULL); 3385 } 3386 bcopy(sa, dupsa, sa->sa_len); 3387 ifma->ifma_addr = dupsa; 3388 3389 ifma->ifma_ifp = ifp; 3390 ifma->ifma_refcount = 1; 3391 ifma->ifma_protospec = NULL; 3392 3393 if (llsa == NULL) { 3394 ifma->ifma_lladdr = NULL; 3395 return (ifma); 3396 } 3397 3398 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3399 if (dupsa == NULL) { 3400 free(ifma->ifma_addr, M_IFMADDR); 3401 free(ifma, M_IFMADDR); 3402 return (NULL); 3403 } 3404 bcopy(llsa, dupsa, llsa->sa_len); 3405 ifma->ifma_lladdr = dupsa; 3406 3407 return (ifma); 3408 } 3409 3410 /* 3411 * if_freemulti: free ifmultiaddr structure and possibly attached related 3412 * addresses. The caller is responsible for implementing reference 3413 * counting, notifying the driver, handling routing messages, and releasing 3414 * any dependent link layer state. 3415 */ 3416 #ifdef MCAST_VERBOSE 3417 extern void kdb_backtrace(void); 3418 #endif 3419 static void 3420 if_freemulti_internal(struct ifmultiaddr *ifma) 3421 { 3422 3423 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3424 ifma->ifma_refcount)); 3425 3426 if (ifma->ifma_lladdr != NULL) 3427 free(ifma->ifma_lladdr, M_IFMADDR); 3428 #ifdef MCAST_VERBOSE 3429 kdb_backtrace(); 3430 printf("%s freeing ifma: %p\n", __func__, ifma); 3431 #endif 3432 free(ifma->ifma_addr, M_IFMADDR); 3433 free(ifma, M_IFMADDR); 3434 } 3435 3436 static void 3437 if_destroymulti(epoch_context_t ctx) 3438 { 3439 struct ifmultiaddr *ifma; 3440 3441 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3442 if_freemulti_internal(ifma); 3443 } 3444 3445 void 3446 if_freemulti(struct ifmultiaddr *ifma) 3447 { 3448 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3449 ifma->ifma_refcount)); 3450 3451 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3452 } 3453 3454 /* 3455 * Register an additional multicast address with a network interface. 3456 * 3457 * - If the address is already present, bump the reference count on the 3458 * address and return. 3459 * - If the address is not link-layer, look up a link layer address. 3460 * - Allocate address structures for one or both addresses, and attach to the 3461 * multicast address list on the interface. If automatically adding a link 3462 * layer address, the protocol address will own a reference to the link 3463 * layer address, to be freed when it is freed. 3464 * - Notify the network device driver of an addition to the multicast address 3465 * list. 3466 * 3467 * 'sa' points to caller-owned memory with the desired multicast address. 3468 * 3469 * 'retifma' will be used to return a pointer to the resulting multicast 3470 * address reference, if desired. 3471 */ 3472 int 3473 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3474 struct ifmultiaddr **retifma) 3475 { 3476 struct ifmultiaddr *ifma, *ll_ifma; 3477 struct sockaddr *llsa; 3478 struct sockaddr_dl sdl; 3479 int error; 3480 3481 #ifdef INET 3482 IN_MULTI_LIST_UNLOCK_ASSERT(); 3483 #endif 3484 #ifdef INET6 3485 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3486 #endif 3487 /* 3488 * If the address is already present, return a new reference to it; 3489 * otherwise, allocate storage and set up a new address. 3490 */ 3491 IF_ADDR_WLOCK(ifp); 3492 ifma = if_findmulti(ifp, sa); 3493 if (ifma != NULL) { 3494 ifma->ifma_refcount++; 3495 if (retifma != NULL) 3496 *retifma = ifma; 3497 IF_ADDR_WUNLOCK(ifp); 3498 return (0); 3499 } 3500 3501 /* 3502 * The address isn't already present; resolve the protocol address 3503 * into a link layer address, and then look that up, bump its 3504 * refcount or allocate an ifma for that also. 3505 * Most link layer resolving functions returns address data which 3506 * fits inside default sockaddr_dl structure. However callback 3507 * can allocate another sockaddr structure, in that case we need to 3508 * free it later. 3509 */ 3510 llsa = NULL; 3511 ll_ifma = NULL; 3512 if (ifp->if_resolvemulti != NULL) { 3513 /* Provide called function with buffer size information */ 3514 sdl.sdl_len = sizeof(sdl); 3515 llsa = (struct sockaddr *)&sdl; 3516 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3517 if (error) 3518 goto unlock_out; 3519 } 3520 3521 /* 3522 * Allocate the new address. Don't hook it up yet, as we may also 3523 * need to allocate a link layer multicast address. 3524 */ 3525 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3526 if (ifma == NULL) { 3527 error = ENOMEM; 3528 goto free_llsa_out; 3529 } 3530 3531 /* 3532 * If a link layer address is found, we'll need to see if it's 3533 * already present in the address list, or allocate is as well. 3534 * When this block finishes, the link layer address will be on the 3535 * list. 3536 */ 3537 if (llsa != NULL) { 3538 ll_ifma = if_findmulti(ifp, llsa); 3539 if (ll_ifma == NULL) { 3540 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3541 if (ll_ifma == NULL) { 3542 --ifma->ifma_refcount; 3543 if_freemulti(ifma); 3544 error = ENOMEM; 3545 goto free_llsa_out; 3546 } 3547 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3548 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3549 ifma_link); 3550 } else 3551 ll_ifma->ifma_refcount++; 3552 ifma->ifma_llifma = ll_ifma; 3553 } 3554 3555 /* 3556 * We now have a new multicast address, ifma, and possibly a new or 3557 * referenced link layer address. Add the primary address to the 3558 * ifnet address list. 3559 */ 3560 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3561 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3562 3563 if (retifma != NULL) 3564 *retifma = ifma; 3565 3566 /* 3567 * Must generate the message while holding the lock so that 'ifma' 3568 * pointer is still valid. 3569 */ 3570 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3571 IF_ADDR_WUNLOCK(ifp); 3572 3573 /* 3574 * We are certain we have added something, so call down to the 3575 * interface to let them know about it. 3576 */ 3577 if (ifp->if_ioctl != NULL) { 3578 if (THREAD_CAN_SLEEP()) 3579 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3580 else 3581 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3582 } 3583 3584 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3585 link_free_sdl(llsa); 3586 3587 return (0); 3588 3589 free_llsa_out: 3590 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3591 link_free_sdl(llsa); 3592 3593 unlock_out: 3594 IF_ADDR_WUNLOCK(ifp); 3595 return (error); 3596 } 3597 3598 static void 3599 if_siocaddmulti(void *arg, int pending) 3600 { 3601 struct ifnet *ifp; 3602 3603 ifp = arg; 3604 #ifdef DIAGNOSTIC 3605 if (pending > 1) 3606 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3607 #endif 3608 CURVNET_SET(ifp->if_vnet); 3609 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3610 CURVNET_RESTORE(); 3611 } 3612 3613 /* 3614 * Delete a multicast group membership by network-layer group address. 3615 * 3616 * Returns ENOENT if the entry could not be found. If ifp no longer 3617 * exists, results are undefined. This entry point should only be used 3618 * from subsystems which do appropriate locking to hold ifp for the 3619 * duration of the call. 3620 * Network-layer protocol domains must use if_delmulti_ifma(). 3621 */ 3622 int 3623 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3624 { 3625 struct ifmultiaddr *ifma; 3626 int lastref; 3627 3628 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3629 3630 IF_ADDR_WLOCK(ifp); 3631 lastref = 0; 3632 ifma = if_findmulti(ifp, sa); 3633 if (ifma != NULL) 3634 lastref = if_delmulti_locked(ifp, ifma, 0); 3635 IF_ADDR_WUNLOCK(ifp); 3636 3637 if (ifma == NULL) 3638 return (ENOENT); 3639 3640 if (lastref && ifp->if_ioctl != NULL) { 3641 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3642 } 3643 3644 return (0); 3645 } 3646 3647 /* 3648 * Delete all multicast group membership for an interface. 3649 * Should be used to quickly flush all multicast filters. 3650 */ 3651 void 3652 if_delallmulti(struct ifnet *ifp) 3653 { 3654 struct ifmultiaddr *ifma; 3655 struct ifmultiaddr *next; 3656 3657 IF_ADDR_WLOCK(ifp); 3658 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3659 if_delmulti_locked(ifp, ifma, 0); 3660 IF_ADDR_WUNLOCK(ifp); 3661 } 3662 3663 void 3664 if_delmulti_ifma(struct ifmultiaddr *ifma) 3665 { 3666 if_delmulti_ifma_flags(ifma, 0); 3667 } 3668 3669 /* 3670 * Delete a multicast group membership by group membership pointer. 3671 * Network-layer protocol domains must use this routine. 3672 * 3673 * It is safe to call this routine if the ifp disappeared. 3674 */ 3675 void 3676 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3677 { 3678 struct ifnet *ifp; 3679 int lastref; 3680 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3681 #ifdef INET 3682 IN_MULTI_LIST_UNLOCK_ASSERT(); 3683 #endif 3684 ifp = ifma->ifma_ifp; 3685 #ifdef DIAGNOSTIC 3686 if (ifp == NULL) { 3687 printf("%s: ifma_ifp seems to be detached\n", __func__); 3688 } else { 3689 struct epoch_tracker et; 3690 struct ifnet *oifp; 3691 3692 NET_EPOCH_ENTER(et); 3693 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3694 if (ifp == oifp) 3695 break; 3696 NET_EPOCH_EXIT(et); 3697 if (ifp != oifp) 3698 ifp = NULL; 3699 } 3700 #endif 3701 /* 3702 * If and only if the ifnet instance exists: Acquire the address lock. 3703 */ 3704 if (ifp != NULL) 3705 IF_ADDR_WLOCK(ifp); 3706 3707 lastref = if_delmulti_locked(ifp, ifma, flags); 3708 3709 if (ifp != NULL) { 3710 /* 3711 * If and only if the ifnet instance exists: 3712 * Release the address lock. 3713 * If the group was left: update the hardware hash filter. 3714 */ 3715 IF_ADDR_WUNLOCK(ifp); 3716 if (lastref && ifp->if_ioctl != NULL) { 3717 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3718 } 3719 } 3720 } 3721 3722 /* 3723 * Perform deletion of network-layer and/or link-layer multicast address. 3724 * 3725 * Return 0 if the reference count was decremented. 3726 * Return 1 if the final reference was released, indicating that the 3727 * hardware hash filter should be reprogrammed. 3728 */ 3729 static int 3730 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3731 { 3732 struct ifmultiaddr *ll_ifma; 3733 3734 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3735 KASSERT(ifma->ifma_ifp == ifp, 3736 ("%s: inconsistent ifp %p", __func__, ifp)); 3737 IF_ADDR_WLOCK_ASSERT(ifp); 3738 } 3739 3740 ifp = ifma->ifma_ifp; 3741 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3742 3743 /* 3744 * If the ifnet is detaching, null out references to ifnet, 3745 * so that upper protocol layers will notice, and not attempt 3746 * to obtain locks for an ifnet which no longer exists. The 3747 * routing socket announcement must happen before the ifnet 3748 * instance is detached from the system. 3749 */ 3750 if (detaching) { 3751 #ifdef DIAGNOSTIC 3752 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3753 #endif 3754 /* 3755 * ifp may already be nulled out if we are being reentered 3756 * to delete the ll_ifma. 3757 */ 3758 if (ifp != NULL) { 3759 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3760 ifma->ifma_ifp = NULL; 3761 } 3762 } 3763 3764 if (--ifma->ifma_refcount > 0) 3765 return 0; 3766 3767 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3768 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3769 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3770 } 3771 /* 3772 * If this ifma is a network-layer ifma, a link-layer ifma may 3773 * have been associated with it. Release it first if so. 3774 */ 3775 ll_ifma = ifma->ifma_llifma; 3776 if (ll_ifma != NULL) { 3777 KASSERT(ifma->ifma_lladdr != NULL, 3778 ("%s: llifma w/o lladdr", __func__)); 3779 if (detaching) 3780 ll_ifma->ifma_ifp = NULL; /* XXX */ 3781 if (--ll_ifma->ifma_refcount == 0) { 3782 if (ifp != NULL) { 3783 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3784 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3785 ifma_link); 3786 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3787 } 3788 } 3789 if_freemulti(ll_ifma); 3790 } 3791 } 3792 #ifdef INVARIANTS 3793 if (ifp) { 3794 struct ifmultiaddr *ifmatmp; 3795 3796 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3797 MPASS(ifma != ifmatmp); 3798 } 3799 #endif 3800 if_freemulti(ifma); 3801 /* 3802 * The last reference to this instance of struct ifmultiaddr 3803 * was released; the hardware should be notified of this change. 3804 */ 3805 return 1; 3806 } 3807 3808 /* 3809 * Set the link layer address on an interface. 3810 * 3811 * At this time we only support certain types of interfaces, 3812 * and we don't allow the length of the address to change. 3813 * 3814 * Set noinline to be dtrace-friendly 3815 */ 3816 __noinline int 3817 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3818 { 3819 struct sockaddr_dl *sdl; 3820 struct ifaddr *ifa; 3821 struct ifreq ifr; 3822 3823 ifa = ifp->if_addr; 3824 if (ifa == NULL) 3825 return (EINVAL); 3826 3827 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3828 if (sdl == NULL) 3829 return (EINVAL); 3830 3831 if (len != sdl->sdl_alen) /* don't allow length to change */ 3832 return (EINVAL); 3833 3834 switch (ifp->if_type) { 3835 case IFT_ETHER: 3836 case IFT_XETHER: 3837 case IFT_L2VLAN: 3838 case IFT_BRIDGE: 3839 case IFT_IEEE8023ADLAG: 3840 bcopy(lladdr, LLADDR(sdl), len); 3841 break; 3842 default: 3843 return (ENODEV); 3844 } 3845 3846 /* 3847 * If the interface is already up, we need 3848 * to re-init it in order to reprogram its 3849 * address filter. 3850 */ 3851 if ((ifp->if_flags & IFF_UP) != 0) { 3852 if (ifp->if_ioctl) { 3853 ifp->if_flags &= ~IFF_UP; 3854 ifr.ifr_flags = ifp->if_flags & 0xffff; 3855 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3856 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3857 ifp->if_flags |= IFF_UP; 3858 ifr.ifr_flags = ifp->if_flags & 0xffff; 3859 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3860 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3861 } 3862 } 3863 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3864 3865 return (0); 3866 } 3867 3868 /* 3869 * Compat function for handling basic encapsulation requests. 3870 * Not converted stacks (FDDI, IB, ..) supports traditional 3871 * output model: ARP (and other similar L2 protocols) are handled 3872 * inside output routine, arpresolve/nd6_resolve() returns MAC 3873 * address instead of full prepend. 3874 * 3875 * This function creates calculated header==MAC for IPv4/IPv6 and 3876 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3877 * address families. 3878 */ 3879 static int 3880 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3881 { 3882 3883 if (req->rtype != IFENCAP_LL) 3884 return (EOPNOTSUPP); 3885 3886 if (req->bufsize < req->lladdr_len) 3887 return (ENOMEM); 3888 3889 switch (req->family) { 3890 case AF_INET: 3891 case AF_INET6: 3892 break; 3893 default: 3894 return (EAFNOSUPPORT); 3895 } 3896 3897 /* Copy lladdr to storage as is */ 3898 memmove(req->buf, req->lladdr, req->lladdr_len); 3899 req->bufsize = req->lladdr_len; 3900 req->lladdr_off = 0; 3901 3902 return (0); 3903 } 3904 3905 /* 3906 * Tunnel interfaces can nest, also they may cause infinite recursion 3907 * calls when misconfigured. We'll prevent this by detecting loops. 3908 * High nesting level may cause stack exhaustion. We'll prevent this 3909 * by introducing upper limit. 3910 * 3911 * Return 0, if tunnel nesting count is equal or less than limit. 3912 */ 3913 int 3914 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3915 int limit) 3916 { 3917 struct m_tag *mtag; 3918 int count; 3919 3920 count = 1; 3921 mtag = NULL; 3922 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3923 if (*(struct ifnet **)(mtag + 1) == ifp) { 3924 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3925 return (EIO); 3926 } 3927 count++; 3928 } 3929 if (count > limit) { 3930 log(LOG_NOTICE, 3931 "%s: if_output recursively called too many times(%d)\n", 3932 if_name(ifp), count); 3933 return (EIO); 3934 } 3935 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3936 if (mtag == NULL) 3937 return (ENOMEM); 3938 *(struct ifnet **)(mtag + 1) = ifp; 3939 m_tag_prepend(m, mtag); 3940 return (0); 3941 } 3942 3943 /* 3944 * Get the link layer address that was read from the hardware at attach. 3945 * 3946 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3947 * their component interfaces as IFT_IEEE8023ADLAG. 3948 */ 3949 int 3950 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3951 { 3952 3953 if (ifp->if_hw_addr == NULL) 3954 return (ENODEV); 3955 3956 switch (ifp->if_type) { 3957 case IFT_ETHER: 3958 case IFT_IEEE8023ADLAG: 3959 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3960 return (0); 3961 default: 3962 return (ENODEV); 3963 } 3964 } 3965 3966 /* 3967 * The name argument must be a pointer to storage which will last as 3968 * long as the interface does. For physical devices, the result of 3969 * device_get_name(dev) is a good choice and for pseudo-devices a 3970 * static string works well. 3971 */ 3972 void 3973 if_initname(struct ifnet *ifp, const char *name, int unit) 3974 { 3975 ifp->if_dname = name; 3976 ifp->if_dunit = unit; 3977 if (unit != IF_DUNIT_NONE) 3978 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3979 else 3980 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3981 } 3982 3983 static int 3984 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap) 3985 { 3986 char if_fmt[256]; 3987 3988 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 3989 vlog(pri, if_fmt, ap); 3990 return (0); 3991 } 3992 3993 3994 int 3995 if_printf(struct ifnet *ifp, const char *fmt, ...) 3996 { 3997 va_list ap; 3998 3999 va_start(ap, fmt); 4000 if_vlog(ifp, LOG_INFO, fmt, ap); 4001 va_end(ap); 4002 return (0); 4003 } 4004 4005 int 4006 if_log(struct ifnet *ifp, int pri, const char *fmt, ...) 4007 { 4008 va_list ap; 4009 4010 va_start(ap, fmt); 4011 if_vlog(ifp, pri, fmt, ap); 4012 va_end(ap); 4013 return (0); 4014 } 4015 4016 void 4017 if_start(struct ifnet *ifp) 4018 { 4019 4020 (*(ifp)->if_start)(ifp); 4021 } 4022 4023 /* 4024 * Backwards compatibility interface for drivers 4025 * that have not implemented it 4026 */ 4027 static int 4028 if_transmit(struct ifnet *ifp, struct mbuf *m) 4029 { 4030 int error; 4031 4032 IFQ_HANDOFF(ifp, m, error); 4033 return (error); 4034 } 4035 4036 static void 4037 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 4038 { 4039 4040 m_freem(m); 4041 } 4042 4043 int 4044 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4045 { 4046 int active = 0; 4047 4048 IF_LOCK(ifq); 4049 if (_IF_QFULL(ifq)) { 4050 IF_UNLOCK(ifq); 4051 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4052 m_freem(m); 4053 return (0); 4054 } 4055 if (ifp != NULL) { 4056 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4057 if (m->m_flags & (M_BCAST|M_MCAST)) 4058 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4059 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4060 } 4061 _IF_ENQUEUE(ifq, m); 4062 IF_UNLOCK(ifq); 4063 if (ifp != NULL && !active) 4064 (*(ifp)->if_start)(ifp); 4065 return (1); 4066 } 4067 4068 void 4069 if_register_com_alloc(u_char type, 4070 if_com_alloc_t *a, if_com_free_t *f) 4071 { 4072 4073 KASSERT(if_com_alloc[type] == NULL, 4074 ("if_register_com_alloc: %d already registered", type)); 4075 KASSERT(if_com_free[type] == NULL, 4076 ("if_register_com_alloc: %d free already registered", type)); 4077 4078 if_com_alloc[type] = a; 4079 if_com_free[type] = f; 4080 } 4081 4082 void 4083 if_deregister_com_alloc(u_char type) 4084 { 4085 4086 KASSERT(if_com_alloc[type] != NULL, 4087 ("if_deregister_com_alloc: %d not registered", type)); 4088 KASSERT(if_com_free[type] != NULL, 4089 ("if_deregister_com_alloc: %d free not registered", type)); 4090 4091 /* 4092 * Ensure all pending EPOCH(9) callbacks have been executed. This 4093 * fixes issues about late invocation of if_destroy(), which leads 4094 * to memory leak from if_com_alloc[type] allocated if_l2com. 4095 */ 4096 epoch_drain_callbacks(net_epoch_preempt); 4097 4098 if_com_alloc[type] = NULL; 4099 if_com_free[type] = NULL; 4100 } 4101 4102 /* API for driver access to network stack owned ifnet.*/ 4103 uint64_t 4104 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4105 { 4106 uint64_t oldbrate; 4107 4108 oldbrate = ifp->if_baudrate; 4109 ifp->if_baudrate = baudrate; 4110 return (oldbrate); 4111 } 4112 4113 uint64_t 4114 if_getbaudrate(if_t ifp) 4115 { 4116 4117 return (((struct ifnet *)ifp)->if_baudrate); 4118 } 4119 4120 int 4121 if_setcapabilities(if_t ifp, int capabilities) 4122 { 4123 ((struct ifnet *)ifp)->if_capabilities = capabilities; 4124 return (0); 4125 } 4126 4127 int 4128 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4129 { 4130 ((struct ifnet *)ifp)->if_capabilities |= setbit; 4131 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 4132 4133 return (0); 4134 } 4135 4136 int 4137 if_getcapabilities(if_t ifp) 4138 { 4139 return ((struct ifnet *)ifp)->if_capabilities; 4140 } 4141 4142 int 4143 if_setcapenable(if_t ifp, int capabilities) 4144 { 4145 ((struct ifnet *)ifp)->if_capenable = capabilities; 4146 return (0); 4147 } 4148 4149 int 4150 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4151 { 4152 if(setcap) 4153 ((struct ifnet *)ifp)->if_capenable |= setcap; 4154 if(clearcap) 4155 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 4156 4157 return (0); 4158 } 4159 4160 const char * 4161 if_getdname(if_t ifp) 4162 { 4163 return ((struct ifnet *)ifp)->if_dname; 4164 } 4165 4166 int 4167 if_togglecapenable(if_t ifp, int togglecap) 4168 { 4169 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 4170 return (0); 4171 } 4172 4173 int 4174 if_getcapenable(if_t ifp) 4175 { 4176 return ((struct ifnet *)ifp)->if_capenable; 4177 } 4178 4179 /* 4180 * This is largely undesirable because it ties ifnet to a device, but does 4181 * provide flexiblity for an embedded product vendor. Should be used with 4182 * the understanding that it violates the interface boundaries, and should be 4183 * a last resort only. 4184 */ 4185 int 4186 if_setdev(if_t ifp, void *dev) 4187 { 4188 return (0); 4189 } 4190 4191 int 4192 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4193 { 4194 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 4195 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 4196 4197 return (0); 4198 } 4199 4200 int 4201 if_getdrvflags(if_t ifp) 4202 { 4203 return ((struct ifnet *)ifp)->if_drv_flags; 4204 } 4205 4206 int 4207 if_setdrvflags(if_t ifp, int flags) 4208 { 4209 ((struct ifnet *)ifp)->if_drv_flags = flags; 4210 return (0); 4211 } 4212 4213 int 4214 if_setflags(if_t ifp, int flags) 4215 { 4216 4217 ifp->if_flags = flags; 4218 return (0); 4219 } 4220 4221 int 4222 if_setflagbits(if_t ifp, int set, int clear) 4223 { 4224 ((struct ifnet *)ifp)->if_flags |= set; 4225 ((struct ifnet *)ifp)->if_flags &= ~clear; 4226 4227 return (0); 4228 } 4229 4230 int 4231 if_getflags(if_t ifp) 4232 { 4233 return ((struct ifnet *)ifp)->if_flags; 4234 } 4235 4236 int 4237 if_clearhwassist(if_t ifp) 4238 { 4239 ((struct ifnet *)ifp)->if_hwassist = 0; 4240 return (0); 4241 } 4242 4243 int 4244 if_sethwassistbits(if_t ifp, int toset, int toclear) 4245 { 4246 ((struct ifnet *)ifp)->if_hwassist |= toset; 4247 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4248 4249 return (0); 4250 } 4251 4252 int 4253 if_sethwassist(if_t ifp, int hwassist_bit) 4254 { 4255 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4256 return (0); 4257 } 4258 4259 int 4260 if_gethwassist(if_t ifp) 4261 { 4262 return ((struct ifnet *)ifp)->if_hwassist; 4263 } 4264 4265 int 4266 if_setmtu(if_t ifp, int mtu) 4267 { 4268 ((struct ifnet *)ifp)->if_mtu = mtu; 4269 return (0); 4270 } 4271 4272 int 4273 if_getmtu(if_t ifp) 4274 { 4275 return ((struct ifnet *)ifp)->if_mtu; 4276 } 4277 4278 int 4279 if_getmtu_family(if_t ifp, int family) 4280 { 4281 struct domain *dp; 4282 4283 for (dp = domains; dp; dp = dp->dom_next) { 4284 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4285 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4286 } 4287 4288 return (((struct ifnet *)ifp)->if_mtu); 4289 } 4290 4291 /* 4292 * Methods for drivers to access interface unicast and multicast 4293 * link level addresses. Driver shall not know 'struct ifaddr' neither 4294 * 'struct ifmultiaddr'. 4295 */ 4296 u_int 4297 if_lladdr_count(if_t ifp) 4298 { 4299 struct epoch_tracker et; 4300 struct ifaddr *ifa; 4301 u_int count; 4302 4303 count = 0; 4304 NET_EPOCH_ENTER(et); 4305 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4306 if (ifa->ifa_addr->sa_family == AF_LINK) 4307 count++; 4308 NET_EPOCH_EXIT(et); 4309 4310 return (count); 4311 } 4312 4313 u_int 4314 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4315 { 4316 struct epoch_tracker et; 4317 struct ifaddr *ifa; 4318 u_int count; 4319 4320 MPASS(cb); 4321 4322 count = 0; 4323 NET_EPOCH_ENTER(et); 4324 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4325 if (ifa->ifa_addr->sa_family != AF_LINK) 4326 continue; 4327 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4328 count); 4329 } 4330 NET_EPOCH_EXIT(et); 4331 4332 return (count); 4333 } 4334 4335 u_int 4336 if_llmaddr_count(if_t ifp) 4337 { 4338 struct epoch_tracker et; 4339 struct ifmultiaddr *ifma; 4340 int count; 4341 4342 count = 0; 4343 NET_EPOCH_ENTER(et); 4344 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4345 if (ifma->ifma_addr->sa_family == AF_LINK) 4346 count++; 4347 NET_EPOCH_EXIT(et); 4348 4349 return (count); 4350 } 4351 4352 u_int 4353 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4354 { 4355 struct epoch_tracker et; 4356 struct ifmultiaddr *ifma; 4357 u_int count; 4358 4359 MPASS(cb); 4360 4361 count = 0; 4362 NET_EPOCH_ENTER(et); 4363 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4364 if (ifma->ifma_addr->sa_family != AF_LINK) 4365 continue; 4366 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4367 count); 4368 } 4369 NET_EPOCH_EXIT(et); 4370 4371 return (count); 4372 } 4373 4374 int 4375 if_setsoftc(if_t ifp, void *softc) 4376 { 4377 ((struct ifnet *)ifp)->if_softc = softc; 4378 return (0); 4379 } 4380 4381 void * 4382 if_getsoftc(if_t ifp) 4383 { 4384 return ((struct ifnet *)ifp)->if_softc; 4385 } 4386 4387 void 4388 if_setrcvif(struct mbuf *m, if_t ifp) 4389 { 4390 4391 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4392 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4393 } 4394 4395 void 4396 if_setvtag(struct mbuf *m, uint16_t tag) 4397 { 4398 m->m_pkthdr.ether_vtag = tag; 4399 } 4400 4401 uint16_t 4402 if_getvtag(struct mbuf *m) 4403 { 4404 4405 return (m->m_pkthdr.ether_vtag); 4406 } 4407 4408 int 4409 if_sendq_empty(if_t ifp) 4410 { 4411 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4412 } 4413 4414 struct ifaddr * 4415 if_getifaddr(if_t ifp) 4416 { 4417 return ((struct ifnet *)ifp)->if_addr; 4418 } 4419 4420 int 4421 if_getamcount(if_t ifp) 4422 { 4423 return ((struct ifnet *)ifp)->if_amcount; 4424 } 4425 4426 int 4427 if_setsendqready(if_t ifp) 4428 { 4429 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4430 return (0); 4431 } 4432 4433 int 4434 if_setsendqlen(if_t ifp, int tx_desc_count) 4435 { 4436 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4437 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4438 4439 return (0); 4440 } 4441 4442 int 4443 if_vlantrunkinuse(if_t ifp) 4444 { 4445 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4446 } 4447 4448 int 4449 if_input(if_t ifp, struct mbuf* sendmp) 4450 { 4451 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4452 return (0); 4453 4454 } 4455 4456 struct mbuf * 4457 if_dequeue(if_t ifp) 4458 { 4459 struct mbuf *m; 4460 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4461 4462 return (m); 4463 } 4464 4465 int 4466 if_sendq_prepend(if_t ifp, struct mbuf *m) 4467 { 4468 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4469 return (0); 4470 } 4471 4472 int 4473 if_setifheaderlen(if_t ifp, int len) 4474 { 4475 ((struct ifnet *)ifp)->if_hdrlen = len; 4476 return (0); 4477 } 4478 4479 caddr_t 4480 if_getlladdr(if_t ifp) 4481 { 4482 return (IF_LLADDR((struct ifnet *)ifp)); 4483 } 4484 4485 void * 4486 if_gethandle(u_char type) 4487 { 4488 return (if_alloc(type)); 4489 } 4490 4491 void 4492 if_bpfmtap(if_t ifh, struct mbuf *m) 4493 { 4494 struct ifnet *ifp = (struct ifnet *)ifh; 4495 4496 BPF_MTAP(ifp, m); 4497 } 4498 4499 void 4500 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4501 { 4502 struct ifnet *ifp = (struct ifnet *)ifh; 4503 4504 ETHER_BPF_MTAP(ifp, m); 4505 } 4506 4507 void 4508 if_vlancap(if_t ifh) 4509 { 4510 struct ifnet *ifp = (struct ifnet *)ifh; 4511 VLAN_CAPABILITIES(ifp); 4512 } 4513 4514 int 4515 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4516 { 4517 4518 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax; 4519 return (0); 4520 } 4521 4522 int 4523 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4524 { 4525 4526 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4527 return (0); 4528 } 4529 4530 int 4531 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4532 { 4533 4534 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4535 return (0); 4536 } 4537 4538 u_int 4539 if_gethwtsomax(if_t ifp) 4540 { 4541 4542 return (((struct ifnet *)ifp)->if_hw_tsomax); 4543 } 4544 4545 u_int 4546 if_gethwtsomaxsegcount(if_t ifp) 4547 { 4548 4549 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount); 4550 } 4551 4552 u_int 4553 if_gethwtsomaxsegsize(if_t ifp) 4554 { 4555 4556 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize); 4557 } 4558 4559 void 4560 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4561 { 4562 ((struct ifnet *)ifp)->if_init = init_fn; 4563 } 4564 4565 void 4566 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4567 { 4568 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4569 } 4570 4571 void 4572 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4573 { 4574 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4575 } 4576 4577 void 4578 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4579 { 4580 ((struct ifnet *)ifp)->if_transmit = start_fn; 4581 } 4582 4583 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4584 { 4585 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4586 4587 } 4588 4589 void 4590 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4591 { 4592 4593 ifp->if_get_counter = fn; 4594 } 4595 4596 #ifdef DDB 4597 static void 4598 if_show_ifnet(struct ifnet *ifp) 4599 { 4600 4601 if (ifp == NULL) 4602 return; 4603 db_printf("%s:\n", ifp->if_xname); 4604 #define IF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, ifp->e); 4605 IF_DB_PRINTF("%s", if_dname); 4606 IF_DB_PRINTF("%d", if_dunit); 4607 IF_DB_PRINTF("%s", if_description); 4608 IF_DB_PRINTF("%u", if_index); 4609 IF_DB_PRINTF("%u", if_refcount); 4610 IF_DB_PRINTF("%d", if_index_reserved); 4611 IF_DB_PRINTF("%p", if_softc); 4612 IF_DB_PRINTF("%p", if_l2com); 4613 IF_DB_PRINTF("%p", if_llsoftc); 4614 IF_DB_PRINTF("%d", if_amcount); 4615 IF_DB_PRINTF("%p", if_addr); 4616 IF_DB_PRINTF("%p", if_broadcastaddr); 4617 IF_DB_PRINTF("%p", if_afdata); 4618 IF_DB_PRINTF("%d", if_afdata_initialized); 4619 IF_DB_PRINTF("%u", if_fib); 4620 IF_DB_PRINTF("%p", if_vnet); 4621 IF_DB_PRINTF("%p", if_home_vnet); 4622 IF_DB_PRINTF("%p", if_vlantrunk); 4623 IF_DB_PRINTF("%p", if_bpf); 4624 IF_DB_PRINTF("%u", if_pcount); 4625 IF_DB_PRINTF("%p", if_bridge); 4626 IF_DB_PRINTF("%p", if_lagg); 4627 IF_DB_PRINTF("%p", if_pf_kif); 4628 IF_DB_PRINTF("%p", if_carp); 4629 IF_DB_PRINTF("%p", if_label); 4630 IF_DB_PRINTF("%p", if_netmap); 4631 IF_DB_PRINTF("0x%08x", if_flags); 4632 IF_DB_PRINTF("0x%08x", if_drv_flags); 4633 IF_DB_PRINTF("0x%08x", if_capabilities); 4634 IF_DB_PRINTF("0x%08x", if_capenable); 4635 IF_DB_PRINTF("%p", if_snd.ifq_head); 4636 IF_DB_PRINTF("%p", if_snd.ifq_tail); 4637 IF_DB_PRINTF("%d", if_snd.ifq_len); 4638 IF_DB_PRINTF("%d", if_snd.ifq_maxlen); 4639 IF_DB_PRINTF("%p", if_snd.ifq_drv_head); 4640 IF_DB_PRINTF("%p", if_snd.ifq_drv_tail); 4641 IF_DB_PRINTF("%d", if_snd.ifq_drv_len); 4642 IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen); 4643 IF_DB_PRINTF("%d", if_snd.altq_type); 4644 IF_DB_PRINTF("%x", if_snd.altq_flags); 4645 #undef IF_DB_PRINTF 4646 } 4647 4648 DB_SHOW_COMMAND(ifnet, db_show_ifnet) 4649 { 4650 4651 if (!have_addr) { 4652 db_printf("usage: show ifnet <struct ifnet *>\n"); 4653 return; 4654 } 4655 4656 if_show_ifnet((struct ifnet *)addr); 4657 } 4658 4659 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets) 4660 { 4661 VNET_ITERATOR_DECL(vnet_iter); 4662 struct ifnet *ifp; 4663 u_short idx; 4664 4665 VNET_FOREACH(vnet_iter) { 4666 CURVNET_SET_QUIET(vnet_iter); 4667 #ifdef VIMAGE 4668 db_printf("vnet=%p\n", curvnet); 4669 #endif 4670 for (idx = 1; idx <= V_if_index; idx++) { 4671 ifp = V_ifindex_table[idx].ife_ifnet; 4672 if (ifp == NULL) 4673 continue; 4674 db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp); 4675 if (db_pager_quit) 4676 break; 4677 } 4678 CURVNET_RESTORE(); 4679 } 4680 } 4681 #endif /* DDB */ 4682