1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1980, 1986, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)if.c 8.5 (Berkeley) 1/9/95 32 * $FreeBSD$ 33 */ 34 35 #include "opt_bpf.h" 36 #include "opt_inet6.h" 37 #include "opt_inet.h" 38 39 #include <sys/param.h> 40 #include <sys/conf.h> 41 #include <sys/eventhandler.h> 42 #include <sys/malloc.h> 43 #include <sys/domainset.h> 44 #include <sys/sbuf.h> 45 #include <sys/bus.h> 46 #include <sys/epoch.h> 47 #include <sys/mbuf.h> 48 #include <sys/systm.h> 49 #include <sys/priv.h> 50 #include <sys/proc.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/protosw.h> 54 #include <sys/kernel.h> 55 #include <sys/lock.h> 56 #include <sys/refcount.h> 57 #include <sys/module.h> 58 #include <sys/rwlock.h> 59 #include <sys/sockio.h> 60 #include <sys/syslog.h> 61 #include <sys/sysctl.h> 62 #include <sys/sysent.h> 63 #include <sys/taskqueue.h> 64 #include <sys/domain.h> 65 #include <sys/jail.h> 66 #include <sys/priv.h> 67 68 #include <machine/stdarg.h> 69 #include <vm/uma.h> 70 71 #include <net/bpf.h> 72 #include <net/ethernet.h> 73 #include <net/if.h> 74 #include <net/if_arp.h> 75 #include <net/if_clone.h> 76 #include <net/if_dl.h> 77 #include <net/if_types.h> 78 #include <net/if_var.h> 79 #include <net/if_media.h> 80 #include <net/if_vlan_var.h> 81 #include <net/radix.h> 82 #include <net/route.h> 83 #include <net/route/route_ctl.h> 84 #include <net/vnet.h> 85 86 #if defined(INET) || defined(INET6) 87 #include <net/ethernet.h> 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip.h> 91 #include <netinet/ip_carp.h> 92 #ifdef INET 93 #include <net/debugnet.h> 94 #include <netinet/if_ether.h> 95 #endif /* INET */ 96 #ifdef INET6 97 #include <netinet6/in6_var.h> 98 #include <netinet6/in6_ifattach.h> 99 #endif /* INET6 */ 100 #endif /* INET || INET6 */ 101 102 #include <security/mac/mac_framework.h> 103 104 /* 105 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name 106 * and ifr_ifru when it is used in SIOCGIFCONF. 107 */ 108 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) == 109 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru"); 110 111 __read_mostly epoch_t net_epoch_preempt; 112 #ifdef COMPAT_FREEBSD32 113 #include <sys/mount.h> 114 #include <compat/freebsd32/freebsd32.h> 115 116 struct ifreq_buffer32 { 117 uint32_t length; /* (size_t) */ 118 uint32_t buffer; /* (void *) */ 119 }; 120 121 /* 122 * Interface request structure used for socket 123 * ioctl's. All interface ioctl's must have parameter 124 * definitions which begin with ifr_name. The 125 * remainder may be interface specific. 126 */ 127 struct ifreq32 { 128 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */ 129 union { 130 struct sockaddr ifru_addr; 131 struct sockaddr ifru_dstaddr; 132 struct sockaddr ifru_broadaddr; 133 struct ifreq_buffer32 ifru_buffer; 134 short ifru_flags[2]; 135 short ifru_index; 136 int ifru_jid; 137 int ifru_metric; 138 int ifru_mtu; 139 int ifru_phys; 140 int ifru_media; 141 uint32_t ifru_data; 142 int ifru_cap[2]; 143 u_int ifru_fib; 144 u_char ifru_vlan_pcp; 145 } ifr_ifru; 146 }; 147 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32)); 148 CTASSERT(__offsetof(struct ifreq, ifr_ifru) == 149 __offsetof(struct ifreq32, ifr_ifru)); 150 151 struct ifgroupreq32 { 152 char ifgr_name[IFNAMSIZ]; 153 u_int ifgr_len; 154 union { 155 char ifgru_group[IFNAMSIZ]; 156 uint32_t ifgru_groups; 157 } ifgr_ifgru; 158 }; 159 160 struct ifmediareq32 { 161 char ifm_name[IFNAMSIZ]; 162 int ifm_current; 163 int ifm_mask; 164 int ifm_status; 165 int ifm_active; 166 int ifm_count; 167 uint32_t ifm_ulist; /* (int *) */ 168 }; 169 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32) 170 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32) 171 172 #define _CASE_IOC_IFGROUPREQ_32(cmd) \ 173 _IOC_NEWTYPE((cmd), struct ifgroupreq32): case 174 #else /* !COMPAT_FREEBSD32 */ 175 #define _CASE_IOC_IFGROUPREQ_32(cmd) 176 #endif /* !COMPAT_FREEBSD32 */ 177 178 #define CASE_IOC_IFGROUPREQ(cmd) \ 179 _CASE_IOC_IFGROUPREQ_32(cmd) \ 180 (cmd) 181 182 union ifreq_union { 183 struct ifreq ifr; 184 #ifdef COMPAT_FREEBSD32 185 struct ifreq32 ifr32; 186 #endif 187 }; 188 189 union ifgroupreq_union { 190 struct ifgroupreq ifgr; 191 #ifdef COMPAT_FREEBSD32 192 struct ifgroupreq32 ifgr32; 193 #endif 194 }; 195 196 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 197 "Link layers"); 198 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 199 "Generic link-management"); 200 201 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, 202 &ifqmaxlen, 0, "max send queue size"); 203 204 /* Log link state change events */ 205 static int log_link_state_change = 1; 206 207 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, 208 &log_link_state_change, 0, 209 "log interface link state change events"); 210 211 /* Log promiscuous mode change events */ 212 static int log_promisc_mode_change = 1; 213 214 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN, 215 &log_promisc_mode_change, 1, 216 "log promiscuous mode change events"); 217 218 /* Interface description */ 219 static unsigned int ifdescr_maxlen = 1024; 220 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, 221 &ifdescr_maxlen, 0, 222 "administrative maximum length for interface description"); 223 224 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); 225 226 /* global sx for non-critical path ifdescr */ 227 static struct sx ifdescr_sx; 228 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); 229 230 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); 231 void (*lagg_linkstate_p)(struct ifnet *ifp, int state); 232 /* These are external hooks for CARP. */ 233 void (*carp_linkstate_p)(struct ifnet *ifp); 234 void (*carp_demote_adj_p)(int, char *); 235 int (*carp_master_p)(struct ifaddr *); 236 #if defined(INET) || defined(INET6) 237 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); 238 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, 239 const struct sockaddr *sa); 240 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); 241 int (*carp_attach_p)(struct ifaddr *, int); 242 void (*carp_detach_p)(struct ifaddr *, bool); 243 #endif 244 #ifdef INET 245 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); 246 #endif 247 #ifdef INET6 248 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); 249 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, 250 const struct in6_addr *taddr); 251 #endif 252 253 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 254 255 /* 256 * XXX: Style; these should be sorted alphabetically, and unprototyped 257 * static functions should be prototyped. Currently they are sorted by 258 * declaration order. 259 */ 260 static void if_attachdomain(void *); 261 static void if_attachdomain1(struct ifnet *); 262 static int ifconf(u_long, caddr_t); 263 static void *if_grow(void); 264 static void if_input_default(struct ifnet *, struct mbuf *); 265 static int if_requestencap_default(struct ifnet *, struct if_encap_req *); 266 static void if_route(struct ifnet *, int flag, int fam); 267 static int if_setflag(struct ifnet *, int, int, int *, int); 268 static int if_transmit(struct ifnet *ifp, struct mbuf *m); 269 static void if_unroute(struct ifnet *, int flag, int fam); 270 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); 271 static void do_link_state_change(void *, int); 272 static int if_getgroup(struct ifgroupreq *, struct ifnet *); 273 static int if_getgroupmembers(struct ifgroupreq *); 274 static void if_delgroups(struct ifnet *); 275 static void if_attach_internal(struct ifnet *, int, struct if_clone *); 276 static int if_detach_internal(struct ifnet *, int, struct if_clone **); 277 static void if_siocaddmulti(void *, int); 278 static void if_link_ifnet(struct ifnet *); 279 static bool if_unlink_ifnet(struct ifnet *, bool); 280 #ifdef VIMAGE 281 static int if_vmove(struct ifnet *, struct vnet *); 282 #endif 283 284 #ifdef INET6 285 /* 286 * XXX: declare here to avoid to include many inet6 related files.. 287 * should be more generalized? 288 */ 289 extern void nd6_setmtu(struct ifnet *); 290 #endif 291 292 /* ipsec helper hooks */ 293 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); 294 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); 295 296 VNET_DEFINE(int, if_index); 297 int ifqmaxlen = IFQ_MAXLEN; 298 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ 299 VNET_DEFINE(struct ifgrouphead, ifg_head); 300 301 VNET_DEFINE_STATIC(int, if_indexlim) = 8; 302 303 /* Table of ifnet by index. */ 304 VNET_DEFINE(struct ifnet **, ifindex_table); 305 306 #define V_if_indexlim VNET(if_indexlim) 307 #define V_ifindex_table VNET(ifindex_table) 308 309 /* 310 * The global network interface list (V_ifnet) and related state (such as 311 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock. 312 * This may be acquired to stabilise the list, or we may rely on NET_EPOCH. 313 */ 314 struct sx ifnet_sxlock; 315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); 316 317 struct sx ifnet_detach_sxlock; 318 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx", 319 SX_RECURSE); 320 321 /* 322 * The allocation of network interfaces is a rather non-atomic affair; we 323 * need to select an index before we are ready to expose the interface for 324 * use, so will use this pointer value to indicate reservation. 325 */ 326 #define IFNET_HOLD (void *)(uintptr_t)(-1) 327 328 #ifdef VIMAGE 329 #define VNET_IS_SHUTTING_DOWN(_vnet) \ 330 ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE) 331 #endif 332 333 static if_com_alloc_t *if_com_alloc[256]; 334 static if_com_free_t *if_com_free[256]; 335 336 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); 337 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 338 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 339 340 struct ifnet * 341 ifnet_byindex(u_short idx) 342 { 343 struct ifnet *ifp; 344 345 if (__predict_false(idx > V_if_index)) 346 return (NULL); 347 348 ifp = *(struct ifnet * const volatile *)(V_ifindex_table + idx); 349 return (__predict_false(ifp == IFNET_HOLD) ? NULL : ifp); 350 } 351 352 struct ifnet * 353 ifnet_byindex_ref(u_short idx) 354 { 355 struct ifnet *ifp; 356 357 NET_EPOCH_ASSERT(); 358 359 ifp = ifnet_byindex(idx); 360 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) 361 return (NULL); 362 if_ref(ifp); 363 return (ifp); 364 } 365 366 /* 367 * Allocate an ifindex array entry; return 0 on success or an error on 368 * failure. 369 */ 370 static u_short 371 ifindex_alloc(void **old) 372 { 373 u_short idx; 374 375 IFNET_WLOCK_ASSERT(); 376 /* 377 * Try to find an empty slot below V_if_index. If we fail, take the 378 * next slot. 379 */ 380 for (idx = 1; idx <= V_if_index; idx++) { 381 if (V_ifindex_table[idx] == NULL) 382 break; 383 } 384 385 /* Catch if_index overflow. */ 386 if (idx >= V_if_indexlim) { 387 *old = if_grow(); 388 return (USHRT_MAX); 389 } 390 if (idx > V_if_index) 391 V_if_index = idx; 392 return (idx); 393 } 394 395 static void 396 ifindex_free_locked(u_short idx) 397 { 398 399 IFNET_WLOCK_ASSERT(); 400 401 V_ifindex_table[idx] = NULL; 402 while (V_if_index > 0 && 403 V_ifindex_table[V_if_index] == NULL) 404 V_if_index--; 405 } 406 407 static void 408 ifindex_free(u_short idx) 409 { 410 411 IFNET_WLOCK(); 412 ifindex_free_locked(idx); 413 IFNET_WUNLOCK(); 414 } 415 416 static void 417 ifnet_setbyindex(u_short idx, struct ifnet *ifp) 418 { 419 420 V_ifindex_table[idx] = ifp; 421 } 422 423 struct ifaddr * 424 ifaddr_byindex(u_short idx) 425 { 426 struct ifnet *ifp; 427 struct ifaddr *ifa = NULL; 428 429 NET_EPOCH_ASSERT(); 430 431 ifp = ifnet_byindex(idx); 432 if (ifp != NULL && (ifa = ifp->if_addr) != NULL) 433 ifa_ref(ifa); 434 return (ifa); 435 } 436 437 /* 438 * Network interface utility routines. 439 * 440 * Routines with ifa_ifwith* names take sockaddr *'s as 441 * parameters. 442 */ 443 444 static void 445 vnet_if_init(const void *unused __unused) 446 { 447 void *old; 448 449 CK_STAILQ_INIT(&V_ifnet); 450 CK_STAILQ_INIT(&V_ifg_head); 451 IFNET_WLOCK(); 452 old = if_grow(); /* create initial table */ 453 IFNET_WUNLOCK(); 454 epoch_wait_preempt(net_epoch_preempt); 455 free(old, M_IFNET); 456 vnet_if_clone_init(); 457 } 458 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, 459 NULL); 460 461 #ifdef VIMAGE 462 static void 463 vnet_if_uninit(const void *unused __unused) 464 { 465 466 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " 467 "not empty", __func__, __LINE__, &V_ifnet)); 468 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " 469 "not empty", __func__, __LINE__, &V_ifg_head)); 470 471 free((caddr_t)V_ifindex_table, M_IFNET); 472 } 473 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, 474 vnet_if_uninit, NULL); 475 #endif 476 477 static void 478 if_link_ifnet(struct ifnet *ifp) 479 { 480 481 IFNET_WLOCK(); 482 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); 483 #ifdef VIMAGE 484 curvnet->vnet_ifcnt++; 485 #endif 486 IFNET_WUNLOCK(); 487 } 488 489 static bool 490 if_unlink_ifnet(struct ifnet *ifp, bool vmove) 491 { 492 struct ifnet *iter; 493 int found = 0; 494 495 IFNET_WLOCK(); 496 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link) 497 if (iter == ifp) { 498 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link); 499 if (!vmove) 500 ifp->if_flags |= IFF_DYING; 501 found = 1; 502 break; 503 } 504 #ifdef VIMAGE 505 curvnet->vnet_ifcnt--; 506 #endif 507 IFNET_WUNLOCK(); 508 509 return (found); 510 } 511 512 #ifdef VIMAGE 513 static void 514 vnet_if_return(const void *unused __unused) 515 { 516 struct ifnet *ifp, *nifp; 517 struct ifnet **pending; 518 int found, i; 519 520 i = 0; 521 522 /* 523 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd 524 * enter NET_EPOCH, but that's not possible, because if_vmove() calls 525 * if_detach_internal(), which waits for NET_EPOCH callbacks to 526 * complete. We can't do that from within NET_EPOCH. 527 * 528 * However, we can also use the IFNET_xLOCK, which is the V_ifnet 529 * read/write lock. We cannot hold the lock as we call if_vmove() 530 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib 531 * ctx lock. 532 */ 533 IFNET_WLOCK(); 534 535 pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt, 536 M_IFNET, M_WAITOK | M_ZERO); 537 538 /* Return all inherited interfaces to their parent vnets. */ 539 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 540 if (ifp->if_home_vnet != ifp->if_vnet) { 541 found = if_unlink_ifnet(ifp, true); 542 MPASS(found); 543 544 pending[i++] = ifp; 545 } 546 } 547 IFNET_WUNLOCK(); 548 549 for (int j = 0; j < i; j++) { 550 if_vmove(pending[j], pending[j]->if_home_vnet); 551 } 552 553 free(pending, M_IFNET); 554 } 555 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, 556 vnet_if_return, NULL); 557 #endif 558 559 static void * 560 if_grow(void) 561 { 562 int oldlim; 563 u_int n; 564 struct ifnet **e; 565 void *old; 566 567 old = NULL; 568 IFNET_WLOCK_ASSERT(); 569 oldlim = V_if_indexlim; 570 IFNET_WUNLOCK(); 571 n = (oldlim << 1) * sizeof(*e); 572 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); 573 IFNET_WLOCK(); 574 if (V_if_indexlim != oldlim) { 575 free(e, M_IFNET); 576 return (NULL); 577 } 578 if (V_ifindex_table != NULL) { 579 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); 580 old = V_ifindex_table; 581 } 582 V_if_indexlim <<= 1; 583 V_ifindex_table = e; 584 return (old); 585 } 586 587 /* 588 * Allocate a struct ifnet and an index for an interface. A layer 2 589 * common structure will also be allocated if an allocation routine is 590 * registered for the passed type. 591 */ 592 struct ifnet * 593 if_alloc_domain(u_char type, int numa_domain) 594 { 595 struct ifnet *ifp; 596 u_short idx; 597 void *old; 598 599 KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large")); 600 if (numa_domain == IF_NODOM) 601 ifp = malloc(sizeof(struct ifnet), M_IFNET, 602 M_WAITOK | M_ZERO); 603 else 604 ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET, 605 DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO); 606 restart: 607 IFNET_WLOCK(); 608 idx = ifindex_alloc(&old); 609 if (__predict_false(idx == USHRT_MAX)) { 610 IFNET_WUNLOCK(); 611 epoch_wait_preempt(net_epoch_preempt); 612 free(old, M_IFNET); 613 goto restart; 614 } 615 ifnet_setbyindex(idx, IFNET_HOLD); 616 IFNET_WUNLOCK(); 617 ifp->if_index = idx; 618 ifp->if_type = type; 619 ifp->if_alloctype = type; 620 ifp->if_numa_domain = numa_domain; 621 #ifdef VIMAGE 622 ifp->if_vnet = curvnet; 623 #endif 624 if (if_com_alloc[type] != NULL) { 625 ifp->if_l2com = if_com_alloc[type](type, ifp); 626 if (ifp->if_l2com == NULL) { 627 free(ifp, M_IFNET); 628 ifindex_free(idx); 629 return (NULL); 630 } 631 } 632 633 IF_ADDR_LOCK_INIT(ifp); 634 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); 635 TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp); 636 ifp->if_afdata_initialized = 0; 637 IF_AFDATA_LOCK_INIT(ifp); 638 CK_STAILQ_INIT(&ifp->if_addrhead); 639 CK_STAILQ_INIT(&ifp->if_multiaddrs); 640 CK_STAILQ_INIT(&ifp->if_groups); 641 #ifdef MAC 642 mac_ifnet_init(ifp); 643 #endif 644 ifq_init(&ifp->if_snd, ifp); 645 646 refcount_init(&ifp->if_refcount, 1); /* Index reference. */ 647 for (int i = 0; i < IFCOUNTERS; i++) 648 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); 649 ifp->if_get_counter = if_get_counter_default; 650 ifp->if_pcp = IFNET_PCP_NONE; 651 ifnet_setbyindex(ifp->if_index, ifp); 652 return (ifp); 653 } 654 655 struct ifnet * 656 if_alloc_dev(u_char type, device_t dev) 657 { 658 int numa_domain; 659 660 if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0) 661 return (if_alloc_domain(type, IF_NODOM)); 662 return (if_alloc_domain(type, numa_domain)); 663 } 664 665 struct ifnet * 666 if_alloc(u_char type) 667 { 668 669 return (if_alloc_domain(type, IF_NODOM)); 670 } 671 /* 672 * Do the actual work of freeing a struct ifnet, and layer 2 common 673 * structure. This call is made when the last reference to an 674 * interface is released. 675 */ 676 static void 677 if_free_internal(struct ifnet *ifp) 678 { 679 680 KASSERT((ifp->if_flags & IFF_DYING), 681 ("if_free_internal: interface not dying")); 682 683 if (if_com_free[ifp->if_alloctype] != NULL) 684 if_com_free[ifp->if_alloctype](ifp->if_l2com, 685 ifp->if_alloctype); 686 687 #ifdef MAC 688 mac_ifnet_destroy(ifp); 689 #endif /* MAC */ 690 IF_AFDATA_DESTROY(ifp); 691 IF_ADDR_LOCK_DESTROY(ifp); 692 ifq_delete(&ifp->if_snd); 693 694 for (int i = 0; i < IFCOUNTERS; i++) 695 counter_u64_free(ifp->if_counters[i]); 696 697 free(ifp->if_description, M_IFDESCR); 698 free(ifp->if_hw_addr, M_IFADDR); 699 free(ifp, M_IFNET); 700 } 701 702 static void 703 if_destroy(epoch_context_t ctx) 704 { 705 struct ifnet *ifp; 706 707 ifp = __containerof(ctx, struct ifnet, if_epoch_ctx); 708 if_free_internal(ifp); 709 } 710 711 /* 712 * Deregister an interface and free the associated storage. 713 */ 714 void 715 if_free(struct ifnet *ifp) 716 { 717 718 ifp->if_flags |= IFF_DYING; /* XXX: Locking */ 719 720 CURVNET_SET_QUIET(ifp->if_vnet); 721 IFNET_WLOCK(); 722 KASSERT(ifp == ifnet_byindex(ifp->if_index), 723 ("%s: freeing unallocated ifnet", ifp->if_xname)); 724 725 ifindex_free_locked(ifp->if_index); 726 IFNET_WUNLOCK(); 727 728 if (refcount_release(&ifp->if_refcount)) 729 NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx); 730 CURVNET_RESTORE(); 731 } 732 733 /* 734 * Interfaces to keep an ifnet type-stable despite the possibility of the 735 * driver calling if_free(). If there are additional references, we defer 736 * freeing the underlying data structure. 737 */ 738 void 739 if_ref(struct ifnet *ifp) 740 { 741 742 /* We don't assert the ifnet list lock here, but arguably should. */ 743 refcount_acquire(&ifp->if_refcount); 744 } 745 746 void 747 if_rele(struct ifnet *ifp) 748 { 749 750 if (!refcount_release(&ifp->if_refcount)) 751 return; 752 NET_EPOCH_CALL(if_destroy, &ifp->if_epoch_ctx); 753 } 754 755 void 756 ifq_init(struct ifaltq *ifq, struct ifnet *ifp) 757 { 758 759 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 760 761 if (ifq->ifq_maxlen == 0) 762 ifq->ifq_maxlen = ifqmaxlen; 763 764 ifq->altq_type = 0; 765 ifq->altq_disc = NULL; 766 ifq->altq_flags &= ALTQF_CANTCHANGE; 767 ifq->altq_tbr = NULL; 768 ifq->altq_ifp = ifp; 769 } 770 771 void 772 ifq_delete(struct ifaltq *ifq) 773 { 774 mtx_destroy(&ifq->ifq_mtx); 775 } 776 777 /* 778 * Perform generic interface initialization tasks and attach the interface 779 * to the list of "active" interfaces. If vmove flag is set on entry 780 * to if_attach_internal(), perform only a limited subset of initialization 781 * tasks, given that we are moving from one vnet to another an ifnet which 782 * has already been fully initialized. 783 * 784 * Note that if_detach_internal() removes group membership unconditionally 785 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. 786 * Thus, when if_vmove() is applied to a cloned interface, group membership 787 * is lost while a cloned one always joins a group whose name is 788 * ifc->ifc_name. To recover this after if_detach_internal() and 789 * if_attach_internal(), the cloner should be specified to 790 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() 791 * attempts to join a group whose name is ifc->ifc_name. 792 * 793 * XXX: 794 * - The decision to return void and thus require this function to 795 * succeed is questionable. 796 * - We should probably do more sanity checking. For instance we don't 797 * do anything to insure if_xname is unique or non-empty. 798 */ 799 void 800 if_attach(struct ifnet *ifp) 801 { 802 803 if_attach_internal(ifp, 0, NULL); 804 } 805 806 /* 807 * Compute the least common TSO limit. 808 */ 809 void 810 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) 811 { 812 /* 813 * 1) If there is no limit currently, take the limit from 814 * the network adapter. 815 * 816 * 2) If the network adapter has a limit below the current 817 * limit, apply it. 818 */ 819 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && 820 ifp->if_hw_tsomax < pmax->tsomaxbytes)) { 821 pmax->tsomaxbytes = ifp->if_hw_tsomax; 822 } 823 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && 824 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { 825 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; 826 } 827 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && 828 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { 829 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; 830 } 831 } 832 833 /* 834 * Update TSO limit of a network adapter. 835 * 836 * Returns zero if no change. Else non-zero. 837 */ 838 int 839 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) 840 { 841 int retval = 0; 842 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { 843 ifp->if_hw_tsomax = pmax->tsomaxbytes; 844 retval++; 845 } 846 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { 847 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; 848 retval++; 849 } 850 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { 851 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; 852 retval++; 853 } 854 return (retval); 855 } 856 857 static void 858 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) 859 { 860 unsigned socksize, ifasize; 861 int namelen, masklen; 862 struct sockaddr_dl *sdl; 863 struct ifaddr *ifa; 864 865 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) 866 panic ("%s: BUG: if_attach called without if_alloc'd input()\n", 867 ifp->if_xname); 868 869 #ifdef VIMAGE 870 ifp->if_vnet = curvnet; 871 if (ifp->if_home_vnet == NULL) 872 ifp->if_home_vnet = curvnet; 873 #endif 874 875 if_addgroup(ifp, IFG_ALL); 876 877 /* Restore group membership for cloned interfaces. */ 878 if (vmove && ifc != NULL) 879 if_clone_addgroup(ifp, ifc); 880 881 getmicrotime(&ifp->if_lastchange); 882 ifp->if_epoch = time_uptime; 883 884 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || 885 (ifp->if_transmit != NULL && ifp->if_qflush != NULL), 886 ("transmit and qflush must both either be set or both be NULL")); 887 if (ifp->if_transmit == NULL) { 888 ifp->if_transmit = if_transmit; 889 ifp->if_qflush = if_qflush; 890 } 891 if (ifp->if_input == NULL) 892 ifp->if_input = if_input_default; 893 894 if (ifp->if_requestencap == NULL) 895 ifp->if_requestencap = if_requestencap_default; 896 897 if (!vmove) { 898 #ifdef MAC 899 mac_ifnet_create(ifp); 900 #endif 901 902 /* 903 * Create a Link Level name for this device. 904 */ 905 namelen = strlen(ifp->if_xname); 906 /* 907 * Always save enough space for any possiable name so we 908 * can do a rename in place later. 909 */ 910 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 911 socksize = masklen + ifp->if_addrlen; 912 if (socksize < sizeof(*sdl)) 913 socksize = sizeof(*sdl); 914 socksize = roundup2(socksize, sizeof(long)); 915 ifasize = sizeof(*ifa) + 2 * socksize; 916 ifa = ifa_alloc(ifasize, M_WAITOK); 917 sdl = (struct sockaddr_dl *)(ifa + 1); 918 sdl->sdl_len = socksize; 919 sdl->sdl_family = AF_LINK; 920 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 921 sdl->sdl_nlen = namelen; 922 sdl->sdl_index = ifp->if_index; 923 sdl->sdl_type = ifp->if_type; 924 ifp->if_addr = ifa; 925 ifa->ifa_ifp = ifp; 926 ifa->ifa_addr = (struct sockaddr *)sdl; 927 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 928 ifa->ifa_netmask = (struct sockaddr *)sdl; 929 sdl->sdl_len = masklen; 930 while (namelen != 0) 931 sdl->sdl_data[--namelen] = 0xff; 932 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 933 /* Reliably crash if used uninitialized. */ 934 ifp->if_broadcastaddr = NULL; 935 936 if (ifp->if_type == IFT_ETHER) { 937 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR, 938 M_WAITOK | M_ZERO); 939 } 940 941 #if defined(INET) || defined(INET6) 942 /* Use defaults for TSO, if nothing is set */ 943 if (ifp->if_hw_tsomax == 0 && 944 ifp->if_hw_tsomaxsegcount == 0 && 945 ifp->if_hw_tsomaxsegsize == 0) { 946 /* 947 * The TSO defaults needs to be such that an 948 * NFS mbuf list of 35 mbufs totalling just 949 * below 64K works and that a chain of mbufs 950 * can be defragged into at most 32 segments: 951 */ 952 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - 953 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); 954 ifp->if_hw_tsomaxsegcount = 35; 955 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ 956 957 /* XXX some drivers set IFCAP_TSO after ethernet attach */ 958 if (ifp->if_capabilities & IFCAP_TSO) { 959 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", 960 ifp->if_hw_tsomax, 961 ifp->if_hw_tsomaxsegcount, 962 ifp->if_hw_tsomaxsegsize); 963 } 964 } 965 #endif 966 } 967 #ifdef VIMAGE 968 else { 969 /* 970 * Update the interface index in the link layer address 971 * of the interface. 972 */ 973 for (ifa = ifp->if_addr; ifa != NULL; 974 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) { 975 if (ifa->ifa_addr->sa_family == AF_LINK) { 976 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 977 sdl->sdl_index = ifp->if_index; 978 } 979 } 980 } 981 #endif 982 983 if_link_ifnet(ifp); 984 985 if (domain_init_status >= 2) 986 if_attachdomain1(ifp); 987 988 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 989 if (IS_DEFAULT_VNET(curvnet)) 990 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); 991 992 /* Announce the interface. */ 993 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 994 } 995 996 static void 997 if_epochalloc(void *dummy __unused) 998 { 999 1000 net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT); 1001 } 1002 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL); 1003 1004 static void 1005 if_attachdomain(void *dummy) 1006 { 1007 struct ifnet *ifp; 1008 1009 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 1010 if_attachdomain1(ifp); 1011 } 1012 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, 1013 if_attachdomain, NULL); 1014 1015 static void 1016 if_attachdomain1(struct ifnet *ifp) 1017 { 1018 struct domain *dp; 1019 1020 /* 1021 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 1022 * cannot lock ifp->if_afdata initialization, entirely. 1023 */ 1024 IF_AFDATA_LOCK(ifp); 1025 if (ifp->if_afdata_initialized >= domain_init_status) { 1026 IF_AFDATA_UNLOCK(ifp); 1027 log(LOG_WARNING, "%s called more than once on %s\n", 1028 __func__, ifp->if_xname); 1029 return; 1030 } 1031 ifp->if_afdata_initialized = domain_init_status; 1032 IF_AFDATA_UNLOCK(ifp); 1033 1034 /* address family dependent data region */ 1035 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 1036 for (dp = domains; dp; dp = dp->dom_next) { 1037 if (dp->dom_ifattach) 1038 ifp->if_afdata[dp->dom_family] = 1039 (*dp->dom_ifattach)(ifp); 1040 } 1041 } 1042 1043 /* 1044 * Remove any unicast or broadcast network addresses from an interface. 1045 */ 1046 void 1047 if_purgeaddrs(struct ifnet *ifp) 1048 { 1049 struct ifaddr *ifa; 1050 1051 while (1) { 1052 struct epoch_tracker et; 1053 1054 NET_EPOCH_ENTER(et); 1055 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1056 if (ifa->ifa_addr->sa_family != AF_LINK) 1057 break; 1058 } 1059 NET_EPOCH_EXIT(et); 1060 1061 if (ifa == NULL) 1062 break; 1063 #ifdef INET 1064 /* XXX: Ugly!! ad hoc just for INET */ 1065 if (ifa->ifa_addr->sa_family == AF_INET) { 1066 struct ifaliasreq ifr; 1067 1068 bzero(&ifr, sizeof(ifr)); 1069 ifr.ifra_addr = *ifa->ifa_addr; 1070 if (ifa->ifa_dstaddr) 1071 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1072 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 1073 NULL) == 0) 1074 continue; 1075 } 1076 #endif /* INET */ 1077 #ifdef INET6 1078 if (ifa->ifa_addr->sa_family == AF_INET6) { 1079 in6_purgeifaddr((struct in6_ifaddr *)ifa); 1080 /* ifp_addrhead is already updated */ 1081 continue; 1082 } 1083 #endif /* INET6 */ 1084 IF_ADDR_WLOCK(ifp); 1085 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1086 IF_ADDR_WUNLOCK(ifp); 1087 ifa_free(ifa); 1088 } 1089 } 1090 1091 /* 1092 * Remove any multicast network addresses from an interface when an ifnet 1093 * is going away. 1094 */ 1095 static void 1096 if_purgemaddrs(struct ifnet *ifp) 1097 { 1098 struct ifmultiaddr *ifma; 1099 1100 IF_ADDR_WLOCK(ifp); 1101 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) { 1102 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs); 1103 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 1104 if_delmulti_locked(ifp, ifma, 1); 1105 } 1106 IF_ADDR_WUNLOCK(ifp); 1107 } 1108 1109 /* 1110 * Detach an interface, removing it from the list of "active" interfaces. 1111 * If vmove flag is set on entry to if_detach_internal(), perform only a 1112 * limited subset of cleanup tasks, given that we are moving an ifnet from 1113 * one vnet to another, where it must be fully operational. 1114 * 1115 * XXXRW: There are some significant questions about event ordering, and 1116 * how to prevent things from starting to use the interface during detach. 1117 */ 1118 void 1119 if_detach(struct ifnet *ifp) 1120 { 1121 bool found; 1122 1123 CURVNET_SET_QUIET(ifp->if_vnet); 1124 found = if_unlink_ifnet(ifp, false); 1125 if (found) { 1126 sx_xlock(&ifnet_detach_sxlock); 1127 if_detach_internal(ifp, 0, NULL); 1128 sx_xunlock(&ifnet_detach_sxlock); 1129 } 1130 CURVNET_RESTORE(); 1131 } 1132 1133 /* 1134 * The vmove flag, if set, indicates that we are called from a callpath 1135 * that is moving an interface to a different vnet instance. 1136 * 1137 * The shutdown flag, if set, indicates that we are called in the 1138 * process of shutting down a vnet instance. Currently only the 1139 * vnet_if_return SYSUNINIT function sets it. Note: we can be called 1140 * on a vnet instance shutdown without this flag being set, e.g., when 1141 * the cloned interfaces are destoyed as first thing of teardown. 1142 */ 1143 static int 1144 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) 1145 { 1146 struct ifaddr *ifa; 1147 int i; 1148 struct domain *dp; 1149 #ifdef VIMAGE 1150 bool shutdown; 1151 1152 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1153 #endif 1154 1155 /* 1156 * At this point we know the interface still was on the ifnet list 1157 * and we removed it so we are in a stable state. 1158 */ 1159 epoch_wait_preempt(net_epoch_preempt); 1160 1161 /* 1162 * Ensure all pending EPOCH(9) callbacks have been executed. This 1163 * fixes issues about late destruction of multicast options 1164 * which lead to leave group calls, which in turn access the 1165 * belonging ifnet structure: 1166 */ 1167 epoch_drain_callbacks(net_epoch_preempt); 1168 1169 /* 1170 * In any case (destroy or vmove) detach us from the groups 1171 * and remove/wait for pending events on the taskq. 1172 * XXX-BZ in theory an interface could still enqueue a taskq change? 1173 */ 1174 if_delgroups(ifp); 1175 1176 taskqueue_drain(taskqueue_swi, &ifp->if_linktask); 1177 taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask); 1178 1179 /* 1180 * Check if this is a cloned interface or not. Must do even if 1181 * shutting down as a if_vmove_reclaim() would move the ifp and 1182 * the if_clone_addgroup() will have a corrupted string overwise 1183 * from a gibberish pointer. 1184 */ 1185 if (vmove && ifcp != NULL) 1186 *ifcp = if_clone_findifc(ifp); 1187 1188 if_down(ifp); 1189 1190 #ifdef VIMAGE 1191 /* 1192 * On VNET shutdown abort here as the stack teardown will do all 1193 * the work top-down for us. 1194 */ 1195 if (shutdown) { 1196 /* Give interface users the chance to clean up. */ 1197 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1198 1199 /* 1200 * In case of a vmove we are done here without error. 1201 * If we would signal an error it would lead to the same 1202 * abort as if we did not find the ifnet anymore. 1203 * if_detach() calls us in void context and does not care 1204 * about an early abort notification, so life is splendid :) 1205 */ 1206 goto finish_vnet_shutdown; 1207 } 1208 #endif 1209 1210 /* 1211 * At this point we are not tearing down a VNET and are either 1212 * going to destroy or vmove the interface and have to cleanup 1213 * accordingly. 1214 */ 1215 1216 /* 1217 * Remove routes and flush queues. 1218 */ 1219 #ifdef ALTQ 1220 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 1221 altq_disable(&ifp->if_snd); 1222 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 1223 altq_detach(&ifp->if_snd); 1224 #endif 1225 1226 if_purgeaddrs(ifp); 1227 1228 #ifdef INET 1229 in_ifdetach(ifp); 1230 #endif 1231 1232 #ifdef INET6 1233 /* 1234 * Remove all IPv6 kernel structs related to ifp. This should be done 1235 * before removing routing entries below, since IPv6 interface direct 1236 * routes are expected to be removed by the IPv6-specific kernel API. 1237 * Otherwise, the kernel will detect some inconsistency and bark it. 1238 */ 1239 in6_ifdetach(ifp); 1240 #endif 1241 if_purgemaddrs(ifp); 1242 1243 /* Announce that the interface is gone. */ 1244 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1245 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1246 if (IS_DEFAULT_VNET(curvnet)) 1247 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); 1248 1249 if (!vmove) { 1250 /* 1251 * Prevent further calls into the device driver via ifnet. 1252 */ 1253 if_dead(ifp); 1254 1255 /* 1256 * Clean up all addresses. 1257 */ 1258 IF_ADDR_WLOCK(ifp); 1259 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) { 1260 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead); 1261 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link); 1262 IF_ADDR_WUNLOCK(ifp); 1263 ifa_free(ifa); 1264 } else 1265 IF_ADDR_WUNLOCK(ifp); 1266 } 1267 1268 rt_flushifroutes(ifp); 1269 1270 #ifdef VIMAGE 1271 finish_vnet_shutdown: 1272 #endif 1273 /* 1274 * We cannot hold the lock over dom_ifdetach calls as they might 1275 * sleep, for example trying to drain a callout, thus open up the 1276 * theoretical race with re-attaching. 1277 */ 1278 IF_AFDATA_LOCK(ifp); 1279 i = ifp->if_afdata_initialized; 1280 ifp->if_afdata_initialized = 0; 1281 IF_AFDATA_UNLOCK(ifp); 1282 for (dp = domains; i > 0 && dp; dp = dp->dom_next) { 1283 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) { 1284 (*dp->dom_ifdetach)(ifp, 1285 ifp->if_afdata[dp->dom_family]); 1286 ifp->if_afdata[dp->dom_family] = NULL; 1287 } 1288 } 1289 1290 return (0); 1291 } 1292 1293 #ifdef VIMAGE 1294 /* 1295 * if_vmove() performs a limited version of if_detach() in current 1296 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. 1297 * An attempt is made to shrink if_index in current vnet, find an 1298 * unused if_index in target vnet and calls if_grow() if necessary, 1299 * and finally find an unused if_xname for the target vnet. 1300 */ 1301 static int 1302 if_vmove(struct ifnet *ifp, struct vnet *new_vnet) 1303 { 1304 struct if_clone *ifc; 1305 #ifdef DEV_BPF 1306 u_int bif_dlt, bif_hdrlen; 1307 #endif 1308 void *old; 1309 int rc; 1310 1311 #ifdef DEV_BPF 1312 /* 1313 * if_detach_internal() will call the eventhandler to notify 1314 * interface departure. That will detach if_bpf. We need to 1315 * safe the dlt and hdrlen so we can re-attach it later. 1316 */ 1317 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen); 1318 #endif 1319 1320 /* 1321 * Detach from current vnet, but preserve LLADDR info, do not 1322 * mark as dead etc. so that the ifnet can be reattached later. 1323 * If we cannot find it, we lost the race to someone else. 1324 */ 1325 rc = if_detach_internal(ifp, 1, &ifc); 1326 if (rc != 0) 1327 return (rc); 1328 1329 /* 1330 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink 1331 * the if_index for that vnet if possible. 1332 * 1333 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, 1334 * or we'd lock on one vnet and unlock on another. 1335 */ 1336 IFNET_WLOCK(); 1337 ifindex_free_locked(ifp->if_index); 1338 IFNET_WUNLOCK(); 1339 1340 /* 1341 * Perform interface-specific reassignment tasks, if provided by 1342 * the driver. 1343 */ 1344 if (ifp->if_reassign != NULL) 1345 ifp->if_reassign(ifp, new_vnet, NULL); 1346 1347 /* 1348 * Switch to the context of the target vnet. 1349 */ 1350 CURVNET_SET_QUIET(new_vnet); 1351 restart: 1352 IFNET_WLOCK(); 1353 ifp->if_index = ifindex_alloc(&old); 1354 if (__predict_false(ifp->if_index == USHRT_MAX)) { 1355 IFNET_WUNLOCK(); 1356 epoch_wait_preempt(net_epoch_preempt); 1357 free(old, M_IFNET); 1358 goto restart; 1359 } 1360 ifnet_setbyindex(ifp->if_index, ifp); 1361 IFNET_WUNLOCK(); 1362 1363 if_attach_internal(ifp, 1, ifc); 1364 1365 #ifdef DEV_BPF 1366 if (ifp->if_bpf == NULL) 1367 bpfattach(ifp, bif_dlt, bif_hdrlen); 1368 #endif 1369 1370 CURVNET_RESTORE(); 1371 return (0); 1372 } 1373 1374 /* 1375 * Move an ifnet to or from another child prison/vnet, specified by the jail id. 1376 */ 1377 static int 1378 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) 1379 { 1380 struct prison *pr; 1381 struct ifnet *difp; 1382 int error; 1383 bool found; 1384 bool shutdown; 1385 1386 /* Try to find the prison within our visibility. */ 1387 sx_slock(&allprison_lock); 1388 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1389 sx_sunlock(&allprison_lock); 1390 if (pr == NULL) 1391 return (ENXIO); 1392 prison_hold_locked(pr); 1393 mtx_unlock(&pr->pr_mtx); 1394 1395 /* Do not try to move the iface from and to the same prison. */ 1396 if (pr->pr_vnet == ifp->if_vnet) { 1397 prison_free(pr); 1398 return (EEXIST); 1399 } 1400 1401 /* Make sure the named iface does not exists in the dst. prison/vnet. */ 1402 /* XXX Lock interfaces to avoid races. */ 1403 CURVNET_SET_QUIET(pr->pr_vnet); 1404 difp = ifunit(ifname); 1405 if (difp != NULL) { 1406 CURVNET_RESTORE(); 1407 prison_free(pr); 1408 return (EEXIST); 1409 } 1410 1411 /* Make sure the VNET is stable. */ 1412 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1413 if (shutdown) { 1414 CURVNET_RESTORE(); 1415 prison_free(pr); 1416 return (EBUSY); 1417 } 1418 CURVNET_RESTORE(); 1419 1420 found = if_unlink_ifnet(ifp, true); 1421 MPASS(found); 1422 1423 /* Move the interface into the child jail/vnet. */ 1424 error = if_vmove(ifp, pr->pr_vnet); 1425 1426 /* Report the new if_xname back to the userland on success. */ 1427 if (error == 0) 1428 sprintf(ifname, "%s", ifp->if_xname); 1429 1430 prison_free(pr); 1431 return (error); 1432 } 1433 1434 static int 1435 if_vmove_reclaim(struct thread *td, char *ifname, int jid) 1436 { 1437 struct prison *pr; 1438 struct vnet *vnet_dst; 1439 struct ifnet *ifp; 1440 int error, found; 1441 bool shutdown; 1442 1443 /* Try to find the prison within our visibility. */ 1444 sx_slock(&allprison_lock); 1445 pr = prison_find_child(td->td_ucred->cr_prison, jid); 1446 sx_sunlock(&allprison_lock); 1447 if (pr == NULL) 1448 return (ENXIO); 1449 prison_hold_locked(pr); 1450 mtx_unlock(&pr->pr_mtx); 1451 1452 /* Make sure the named iface exists in the source prison/vnet. */ 1453 CURVNET_SET(pr->pr_vnet); 1454 ifp = ifunit(ifname); /* XXX Lock to avoid races. */ 1455 if (ifp == NULL) { 1456 CURVNET_RESTORE(); 1457 prison_free(pr); 1458 return (ENXIO); 1459 } 1460 1461 /* Do not try to move the iface from and to the same prison. */ 1462 vnet_dst = TD_TO_VNET(td); 1463 if (vnet_dst == ifp->if_vnet) { 1464 CURVNET_RESTORE(); 1465 prison_free(pr); 1466 return (EEXIST); 1467 } 1468 1469 /* Make sure the VNET is stable. */ 1470 shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet); 1471 if (shutdown) { 1472 CURVNET_RESTORE(); 1473 prison_free(pr); 1474 return (EBUSY); 1475 } 1476 1477 /* Get interface back from child jail/vnet. */ 1478 found = if_unlink_ifnet(ifp, true); 1479 MPASS(found); 1480 error = if_vmove(ifp, vnet_dst); 1481 CURVNET_RESTORE(); 1482 1483 /* Report the new if_xname back to the userland on success. */ 1484 if (error == 0) 1485 sprintf(ifname, "%s", ifp->if_xname); 1486 1487 prison_free(pr); 1488 return (error); 1489 } 1490 #endif /* VIMAGE */ 1491 1492 /* 1493 * Add a group to an interface 1494 */ 1495 int 1496 if_addgroup(struct ifnet *ifp, const char *groupname) 1497 { 1498 struct ifg_list *ifgl; 1499 struct ifg_group *ifg = NULL; 1500 struct ifg_member *ifgm; 1501 int new = 0; 1502 1503 if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && 1504 groupname[strlen(groupname) - 1] <= '9') 1505 return (EINVAL); 1506 1507 IFNET_WLOCK(); 1508 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1509 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { 1510 IFNET_WUNLOCK(); 1511 return (EEXIST); 1512 } 1513 1514 if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) { 1515 IFNET_WUNLOCK(); 1516 return (ENOMEM); 1517 } 1518 1519 if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) { 1520 free(ifgl, M_TEMP); 1521 IFNET_WUNLOCK(); 1522 return (ENOMEM); 1523 } 1524 1525 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1526 if (!strcmp(ifg->ifg_group, groupname)) 1527 break; 1528 1529 if (ifg == NULL) { 1530 if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) { 1531 free(ifgl, M_TEMP); 1532 free(ifgm, M_TEMP); 1533 IFNET_WUNLOCK(); 1534 return (ENOMEM); 1535 } 1536 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); 1537 ifg->ifg_refcnt = 0; 1538 CK_STAILQ_INIT(&ifg->ifg_members); 1539 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); 1540 new = 1; 1541 } 1542 1543 ifg->ifg_refcnt++; 1544 ifgl->ifgl_group = ifg; 1545 ifgm->ifgm_ifp = ifp; 1546 1547 IF_ADDR_WLOCK(ifp); 1548 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); 1549 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); 1550 IF_ADDR_WUNLOCK(ifp); 1551 1552 IFNET_WUNLOCK(); 1553 1554 if (new) 1555 EVENTHANDLER_INVOKE(group_attach_event, ifg); 1556 EVENTHANDLER_INVOKE(group_change_event, groupname); 1557 1558 return (0); 1559 } 1560 1561 /* 1562 * Helper function to remove a group out of an interface. Expects the global 1563 * ifnet lock to be write-locked, and drops it before returning. 1564 */ 1565 static void 1566 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl, 1567 const char *groupname) 1568 { 1569 struct ifg_member *ifgm; 1570 bool freeifgl; 1571 1572 IFNET_WLOCK_ASSERT(); 1573 1574 IF_ADDR_WLOCK(ifp); 1575 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next); 1576 IF_ADDR_WUNLOCK(ifp); 1577 1578 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) { 1579 if (ifgm->ifgm_ifp == ifp) { 1580 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, 1581 ifg_member, ifgm_next); 1582 break; 1583 } 1584 } 1585 1586 if (--ifgl->ifgl_group->ifg_refcnt == 0) { 1587 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, 1588 ifg_next); 1589 freeifgl = true; 1590 } else { 1591 freeifgl = false; 1592 } 1593 IFNET_WUNLOCK(); 1594 1595 epoch_wait_preempt(net_epoch_preempt); 1596 if (freeifgl) { 1597 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); 1598 free(ifgl->ifgl_group, M_TEMP); 1599 } 1600 free(ifgm, M_TEMP); 1601 free(ifgl, M_TEMP); 1602 1603 EVENTHANDLER_INVOKE(group_change_event, groupname); 1604 } 1605 1606 /* 1607 * Remove a group from an interface 1608 */ 1609 int 1610 if_delgroup(struct ifnet *ifp, const char *groupname) 1611 { 1612 struct ifg_list *ifgl; 1613 1614 IFNET_WLOCK(); 1615 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1616 if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0) 1617 break; 1618 if (ifgl == NULL) { 1619 IFNET_WUNLOCK(); 1620 return (ENOENT); 1621 } 1622 1623 _if_delgroup_locked(ifp, ifgl, groupname); 1624 1625 return (0); 1626 } 1627 1628 /* 1629 * Remove an interface from all groups 1630 */ 1631 static void 1632 if_delgroups(struct ifnet *ifp) 1633 { 1634 struct ifg_list *ifgl; 1635 char groupname[IFNAMSIZ]; 1636 1637 IFNET_WLOCK(); 1638 while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) { 1639 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); 1640 _if_delgroup_locked(ifp, ifgl, groupname); 1641 IFNET_WLOCK(); 1642 } 1643 IFNET_WUNLOCK(); 1644 } 1645 1646 static char * 1647 ifgr_group_get(void *ifgrp) 1648 { 1649 union ifgroupreq_union *ifgrup; 1650 1651 ifgrup = ifgrp; 1652 #ifdef COMPAT_FREEBSD32 1653 if (SV_CURPROC_FLAG(SV_ILP32)) 1654 return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]); 1655 #endif 1656 return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]); 1657 } 1658 1659 static struct ifg_req * 1660 ifgr_groups_get(void *ifgrp) 1661 { 1662 union ifgroupreq_union *ifgrup; 1663 1664 ifgrup = ifgrp; 1665 #ifdef COMPAT_FREEBSD32 1666 if (SV_CURPROC_FLAG(SV_ILP32)) 1667 return ((struct ifg_req *)(uintptr_t) 1668 ifgrup->ifgr32.ifgr_ifgru.ifgru_groups); 1669 #endif 1670 return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups); 1671 } 1672 1673 /* 1674 * Stores all groups from an interface in memory pointed to by ifgr. 1675 */ 1676 static int 1677 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp) 1678 { 1679 int len, error; 1680 struct ifg_list *ifgl; 1681 struct ifg_req ifgrq, *ifgp; 1682 1683 NET_EPOCH_ASSERT(); 1684 1685 if (ifgr->ifgr_len == 0) { 1686 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) 1687 ifgr->ifgr_len += sizeof(struct ifg_req); 1688 return (0); 1689 } 1690 1691 len = ifgr->ifgr_len; 1692 ifgp = ifgr_groups_get(ifgr); 1693 /* XXX: wire */ 1694 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { 1695 if (len < sizeof(ifgrq)) 1696 return (EINVAL); 1697 bzero(&ifgrq, sizeof ifgrq); 1698 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, 1699 sizeof(ifgrq.ifgrq_group)); 1700 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) 1701 return (error); 1702 len -= sizeof(ifgrq); 1703 ifgp++; 1704 } 1705 1706 return (0); 1707 } 1708 1709 /* 1710 * Stores all members of a group in memory pointed to by igfr 1711 */ 1712 static int 1713 if_getgroupmembers(struct ifgroupreq *ifgr) 1714 { 1715 struct ifg_group *ifg; 1716 struct ifg_member *ifgm; 1717 struct ifg_req ifgrq, *ifgp; 1718 int len, error; 1719 1720 IFNET_RLOCK(); 1721 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) 1722 if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0) 1723 break; 1724 if (ifg == NULL) { 1725 IFNET_RUNLOCK(); 1726 return (ENOENT); 1727 } 1728 1729 if (ifgr->ifgr_len == 0) { 1730 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) 1731 ifgr->ifgr_len += sizeof(ifgrq); 1732 IFNET_RUNLOCK(); 1733 return (0); 1734 } 1735 1736 len = ifgr->ifgr_len; 1737 ifgp = ifgr_groups_get(ifgr); 1738 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { 1739 if (len < sizeof(ifgrq)) { 1740 IFNET_RUNLOCK(); 1741 return (EINVAL); 1742 } 1743 bzero(&ifgrq, sizeof ifgrq); 1744 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, 1745 sizeof(ifgrq.ifgrq_member)); 1746 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { 1747 IFNET_RUNLOCK(); 1748 return (error); 1749 } 1750 len -= sizeof(ifgrq); 1751 ifgp++; 1752 } 1753 IFNET_RUNLOCK(); 1754 1755 return (0); 1756 } 1757 1758 /* 1759 * Return counter values from counter(9)s stored in ifnet. 1760 */ 1761 uint64_t 1762 if_get_counter_default(struct ifnet *ifp, ift_counter cnt) 1763 { 1764 1765 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1766 1767 return (counter_u64_fetch(ifp->if_counters[cnt])); 1768 } 1769 1770 /* 1771 * Increase an ifnet counter. Usually used for counters shared 1772 * between the stack and a driver, but function supports them all. 1773 */ 1774 void 1775 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) 1776 { 1777 1778 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); 1779 1780 counter_u64_add(ifp->if_counters[cnt], inc); 1781 } 1782 1783 /* 1784 * Copy data from ifnet to userland API structure if_data. 1785 */ 1786 void 1787 if_data_copy(struct ifnet *ifp, struct if_data *ifd) 1788 { 1789 1790 ifd->ifi_type = ifp->if_type; 1791 ifd->ifi_physical = 0; 1792 ifd->ifi_addrlen = ifp->if_addrlen; 1793 ifd->ifi_hdrlen = ifp->if_hdrlen; 1794 ifd->ifi_link_state = ifp->if_link_state; 1795 ifd->ifi_vhid = 0; 1796 ifd->ifi_datalen = sizeof(struct if_data); 1797 ifd->ifi_mtu = ifp->if_mtu; 1798 ifd->ifi_metric = ifp->if_metric; 1799 ifd->ifi_baudrate = ifp->if_baudrate; 1800 ifd->ifi_hwassist = ifp->if_hwassist; 1801 ifd->ifi_epoch = ifp->if_epoch; 1802 ifd->ifi_lastchange = ifp->if_lastchange; 1803 1804 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); 1805 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); 1806 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); 1807 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); 1808 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); 1809 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); 1810 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); 1811 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); 1812 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); 1813 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); 1814 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); 1815 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); 1816 } 1817 1818 /* 1819 * Initialization, destruction and refcounting functions for ifaddrs. 1820 */ 1821 struct ifaddr * 1822 ifa_alloc(size_t size, int flags) 1823 { 1824 struct ifaddr *ifa; 1825 1826 KASSERT(size >= sizeof(struct ifaddr), 1827 ("%s: invalid size %zu", __func__, size)); 1828 1829 ifa = malloc(size, M_IFADDR, M_ZERO | flags); 1830 if (ifa == NULL) 1831 return (NULL); 1832 1833 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) 1834 goto fail; 1835 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) 1836 goto fail; 1837 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) 1838 goto fail; 1839 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) 1840 goto fail; 1841 1842 refcount_init(&ifa->ifa_refcnt, 1); 1843 1844 return (ifa); 1845 1846 fail: 1847 /* free(NULL) is okay */ 1848 counter_u64_free(ifa->ifa_opackets); 1849 counter_u64_free(ifa->ifa_ipackets); 1850 counter_u64_free(ifa->ifa_obytes); 1851 counter_u64_free(ifa->ifa_ibytes); 1852 free(ifa, M_IFADDR); 1853 1854 return (NULL); 1855 } 1856 1857 void 1858 ifa_ref(struct ifaddr *ifa) 1859 { 1860 u_int old; 1861 1862 old = refcount_acquire(&ifa->ifa_refcnt); 1863 KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa)); 1864 } 1865 1866 int 1867 ifa_try_ref(struct ifaddr *ifa) 1868 { 1869 1870 NET_EPOCH_ASSERT(); 1871 return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt)); 1872 } 1873 1874 static void 1875 ifa_destroy(epoch_context_t ctx) 1876 { 1877 struct ifaddr *ifa; 1878 1879 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx); 1880 counter_u64_free(ifa->ifa_opackets); 1881 counter_u64_free(ifa->ifa_ipackets); 1882 counter_u64_free(ifa->ifa_obytes); 1883 counter_u64_free(ifa->ifa_ibytes); 1884 free(ifa, M_IFADDR); 1885 } 1886 1887 void 1888 ifa_free(struct ifaddr *ifa) 1889 { 1890 1891 if (refcount_release(&ifa->ifa_refcnt)) 1892 NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx); 1893 } 1894 1895 /* 1896 * XXX: Because sockaddr_dl has deeper structure than the sockaddr 1897 * structs used to represent other address families, it is necessary 1898 * to perform a different comparison. 1899 */ 1900 1901 #define sa_dl_equal(a1, a2) \ 1902 ((((const struct sockaddr_dl *)(a1))->sdl_len == \ 1903 ((const struct sockaddr_dl *)(a2))->sdl_len) && \ 1904 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ 1905 CLLADDR((const struct sockaddr_dl *)(a2)), \ 1906 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) 1907 1908 /* 1909 * Locate an interface based on a complete address. 1910 */ 1911 /*ARGSUSED*/ 1912 struct ifaddr * 1913 ifa_ifwithaddr(const struct sockaddr *addr) 1914 { 1915 struct ifnet *ifp; 1916 struct ifaddr *ifa; 1917 1918 NET_EPOCH_ASSERT(); 1919 1920 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1921 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1922 if (ifa->ifa_addr->sa_family != addr->sa_family) 1923 continue; 1924 if (sa_equal(addr, ifa->ifa_addr)) { 1925 goto done; 1926 } 1927 /* IP6 doesn't have broadcast */ 1928 if ((ifp->if_flags & IFF_BROADCAST) && 1929 ifa->ifa_broadaddr && 1930 ifa->ifa_broadaddr->sa_len != 0 && 1931 sa_equal(ifa->ifa_broadaddr, addr)) { 1932 goto done; 1933 } 1934 } 1935 } 1936 ifa = NULL; 1937 done: 1938 return (ifa); 1939 } 1940 1941 int 1942 ifa_ifwithaddr_check(const struct sockaddr *addr) 1943 { 1944 struct epoch_tracker et; 1945 int rc; 1946 1947 NET_EPOCH_ENTER(et); 1948 rc = (ifa_ifwithaddr(addr) != NULL); 1949 NET_EPOCH_EXIT(et); 1950 return (rc); 1951 } 1952 1953 /* 1954 * Locate an interface based on the broadcast address. 1955 */ 1956 /* ARGSUSED */ 1957 struct ifaddr * 1958 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) 1959 { 1960 struct ifnet *ifp; 1961 struct ifaddr *ifa; 1962 1963 NET_EPOCH_ASSERT(); 1964 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1965 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1966 continue; 1967 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1968 if (ifa->ifa_addr->sa_family != addr->sa_family) 1969 continue; 1970 if ((ifp->if_flags & IFF_BROADCAST) && 1971 ifa->ifa_broadaddr && 1972 ifa->ifa_broadaddr->sa_len != 0 && 1973 sa_equal(ifa->ifa_broadaddr, addr)) { 1974 goto done; 1975 } 1976 } 1977 } 1978 ifa = NULL; 1979 done: 1980 return (ifa); 1981 } 1982 1983 /* 1984 * Locate the point to point interface with a given destination address. 1985 */ 1986 /*ARGSUSED*/ 1987 struct ifaddr * 1988 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) 1989 { 1990 struct ifnet *ifp; 1991 struct ifaddr *ifa; 1992 1993 NET_EPOCH_ASSERT(); 1994 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1995 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 1996 continue; 1997 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 1998 continue; 1999 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2000 if (ifa->ifa_addr->sa_family != addr->sa_family) 2001 continue; 2002 if (ifa->ifa_dstaddr != NULL && 2003 sa_equal(addr, ifa->ifa_dstaddr)) { 2004 goto done; 2005 } 2006 } 2007 } 2008 ifa = NULL; 2009 done: 2010 return (ifa); 2011 } 2012 2013 /* 2014 * Find an interface on a specific network. If many, choice 2015 * is most specific found. 2016 */ 2017 struct ifaddr * 2018 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) 2019 { 2020 struct ifnet *ifp; 2021 struct ifaddr *ifa; 2022 struct ifaddr *ifa_maybe = NULL; 2023 u_int af = addr->sa_family; 2024 const char *addr_data = addr->sa_data, *cplim; 2025 2026 NET_EPOCH_ASSERT(); 2027 /* 2028 * AF_LINK addresses can be looked up directly by their index number, 2029 * so do that if we can. 2030 */ 2031 if (af == AF_LINK) { 2032 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; 2033 if (sdl->sdl_index && sdl->sdl_index <= V_if_index) 2034 return (ifaddr_byindex(sdl->sdl_index)); 2035 } 2036 2037 /* 2038 * Scan though each interface, looking for ones that have addresses 2039 * in this address family and the requested fib. 2040 */ 2041 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2042 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) 2043 continue; 2044 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2045 const char *cp, *cp2, *cp3; 2046 2047 if (ifa->ifa_addr->sa_family != af) 2048 next: continue; 2049 if (af == AF_INET && 2050 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { 2051 /* 2052 * This is a bit broken as it doesn't 2053 * take into account that the remote end may 2054 * be a single node in the network we are 2055 * looking for. 2056 * The trouble is that we don't know the 2057 * netmask for the remote end. 2058 */ 2059 if (ifa->ifa_dstaddr != NULL && 2060 sa_equal(addr, ifa->ifa_dstaddr)) { 2061 goto done; 2062 } 2063 } else { 2064 /* 2065 * Scan all the bits in the ifa's address. 2066 * If a bit dissagrees with what we are 2067 * looking for, mask it with the netmask 2068 * to see if it really matters. 2069 * (A byte at a time) 2070 */ 2071 if (ifa->ifa_netmask == 0) 2072 continue; 2073 cp = addr_data; 2074 cp2 = ifa->ifa_addr->sa_data; 2075 cp3 = ifa->ifa_netmask->sa_data; 2076 cplim = ifa->ifa_netmask->sa_len 2077 + (char *)ifa->ifa_netmask; 2078 while (cp3 < cplim) 2079 if ((*cp++ ^ *cp2++) & *cp3++) 2080 goto next; /* next address! */ 2081 /* 2082 * If the netmask of what we just found 2083 * is more specific than what we had before 2084 * (if we had one), or if the virtual status 2085 * of new prefix is better than of the old one, 2086 * then remember the new one before continuing 2087 * to search for an even better one. 2088 */ 2089 if (ifa_maybe == NULL || 2090 ifa_preferred(ifa_maybe, ifa) || 2091 rn_refines((caddr_t)ifa->ifa_netmask, 2092 (caddr_t)ifa_maybe->ifa_netmask)) { 2093 ifa_maybe = ifa; 2094 } 2095 } 2096 } 2097 } 2098 ifa = ifa_maybe; 2099 ifa_maybe = NULL; 2100 done: 2101 return (ifa); 2102 } 2103 2104 /* 2105 * Find an interface address specific to an interface best matching 2106 * a given address. 2107 */ 2108 struct ifaddr * 2109 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) 2110 { 2111 struct ifaddr *ifa; 2112 const char *cp, *cp2, *cp3; 2113 char *cplim; 2114 struct ifaddr *ifa_maybe = NULL; 2115 u_int af = addr->sa_family; 2116 2117 if (af >= AF_MAX) 2118 return (NULL); 2119 2120 NET_EPOCH_ASSERT(); 2121 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 2122 if (ifa->ifa_addr->sa_family != af) 2123 continue; 2124 if (ifa_maybe == NULL) 2125 ifa_maybe = ifa; 2126 if (ifa->ifa_netmask == 0) { 2127 if (sa_equal(addr, ifa->ifa_addr) || 2128 (ifa->ifa_dstaddr && 2129 sa_equal(addr, ifa->ifa_dstaddr))) 2130 goto done; 2131 continue; 2132 } 2133 if (ifp->if_flags & IFF_POINTOPOINT) { 2134 if (sa_equal(addr, ifa->ifa_dstaddr)) 2135 goto done; 2136 } else { 2137 cp = addr->sa_data; 2138 cp2 = ifa->ifa_addr->sa_data; 2139 cp3 = ifa->ifa_netmask->sa_data; 2140 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 2141 for (; cp3 < cplim; cp3++) 2142 if ((*cp++ ^ *cp2++) & *cp3) 2143 break; 2144 if (cp3 == cplim) 2145 goto done; 2146 } 2147 } 2148 ifa = ifa_maybe; 2149 done: 2150 return (ifa); 2151 } 2152 2153 /* 2154 * See whether new ifa is better than current one: 2155 * 1) A non-virtual one is preferred over virtual. 2156 * 2) A virtual in master state preferred over any other state. 2157 * 2158 * Used in several address selecting functions. 2159 */ 2160 int 2161 ifa_preferred(struct ifaddr *cur, struct ifaddr *next) 2162 { 2163 2164 return (cur->ifa_carp && (!next->ifa_carp || 2165 ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); 2166 } 2167 2168 struct sockaddr_dl * 2169 link_alloc_sdl(size_t size, int flags) 2170 { 2171 2172 return (malloc(size, M_TEMP, flags)); 2173 } 2174 2175 void 2176 link_free_sdl(struct sockaddr *sa) 2177 { 2178 free(sa, M_TEMP); 2179 } 2180 2181 /* 2182 * Fills in given sdl with interface basic info. 2183 * Returns pointer to filled sdl. 2184 */ 2185 struct sockaddr_dl * 2186 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) 2187 { 2188 struct sockaddr_dl *sdl; 2189 2190 sdl = (struct sockaddr_dl *)paddr; 2191 memset(sdl, 0, sizeof(struct sockaddr_dl)); 2192 sdl->sdl_len = sizeof(struct sockaddr_dl); 2193 sdl->sdl_family = AF_LINK; 2194 sdl->sdl_index = ifp->if_index; 2195 sdl->sdl_type = iftype; 2196 2197 return (sdl); 2198 } 2199 2200 /* 2201 * Mark an interface down and notify protocols of 2202 * the transition. 2203 */ 2204 static void 2205 if_unroute(struct ifnet *ifp, int flag, int fam) 2206 { 2207 struct ifaddr *ifa; 2208 2209 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); 2210 2211 ifp->if_flags &= ~flag; 2212 getmicrotime(&ifp->if_lastchange); 2213 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2214 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2215 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 2216 ifp->if_qflush(ifp); 2217 2218 if (ifp->if_carp) 2219 (*carp_linkstate_p)(ifp); 2220 rt_ifmsg(ifp); 2221 } 2222 2223 /* 2224 * Mark an interface up and notify protocols of 2225 * the transition. 2226 */ 2227 static void 2228 if_route(struct ifnet *ifp, int flag, int fam) 2229 { 2230 struct ifaddr *ifa; 2231 2232 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); 2233 2234 ifp->if_flags |= flag; 2235 getmicrotime(&ifp->if_lastchange); 2236 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 2237 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 2238 pfctlinput(PRC_IFUP, ifa->ifa_addr); 2239 if (ifp->if_carp) 2240 (*carp_linkstate_p)(ifp); 2241 rt_ifmsg(ifp); 2242 #ifdef INET6 2243 in6_if_up(ifp); 2244 #endif 2245 } 2246 2247 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ 2248 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ 2249 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); 2250 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); 2251 int (*vlan_tag_p)(struct ifnet *, uint16_t *); 2252 int (*vlan_pcp_p)(struct ifnet *, uint16_t *); 2253 int (*vlan_setcookie_p)(struct ifnet *, void *); 2254 void *(*vlan_cookie_p)(struct ifnet *); 2255 2256 /* 2257 * Handle a change in the interface link state. To avoid LORs 2258 * between driver lock and upper layer locks, as well as possible 2259 * recursions, we post event to taskqueue, and all job 2260 * is done in static do_link_state_change(). 2261 */ 2262 void 2263 if_link_state_change(struct ifnet *ifp, int link_state) 2264 { 2265 /* Return if state hasn't changed. */ 2266 if (ifp->if_link_state == link_state) 2267 return; 2268 2269 ifp->if_link_state = link_state; 2270 2271 /* XXXGL: reference ifp? */ 2272 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); 2273 } 2274 2275 static void 2276 do_link_state_change(void *arg, int pending) 2277 { 2278 struct ifnet *ifp; 2279 int link_state; 2280 2281 ifp = arg; 2282 link_state = ifp->if_link_state; 2283 2284 CURVNET_SET(ifp->if_vnet); 2285 rt_ifmsg(ifp); 2286 if (ifp->if_vlantrunk != NULL) 2287 (*vlan_link_state_p)(ifp); 2288 2289 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && 2290 ifp->if_l2com != NULL) 2291 (*ng_ether_link_state_p)(ifp, link_state); 2292 if (ifp->if_carp) 2293 (*carp_linkstate_p)(ifp); 2294 if (ifp->if_bridge) 2295 ifp->if_bridge_linkstate(ifp); 2296 if (ifp->if_lagg) 2297 (*lagg_linkstate_p)(ifp, link_state); 2298 2299 if (IS_DEFAULT_VNET(curvnet)) 2300 devctl_notify("IFNET", ifp->if_xname, 2301 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", 2302 NULL); 2303 if (pending > 1) 2304 if_printf(ifp, "%d link states coalesced\n", pending); 2305 if (log_link_state_change) 2306 if_printf(ifp, "link state changed to %s\n", 2307 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); 2308 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state); 2309 CURVNET_RESTORE(); 2310 } 2311 2312 /* 2313 * Mark an interface down and notify protocols of 2314 * the transition. 2315 */ 2316 void 2317 if_down(struct ifnet *ifp) 2318 { 2319 2320 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN); 2321 if_unroute(ifp, IFF_UP, AF_UNSPEC); 2322 } 2323 2324 /* 2325 * Mark an interface up and notify protocols of 2326 * the transition. 2327 */ 2328 void 2329 if_up(struct ifnet *ifp) 2330 { 2331 2332 if_route(ifp, IFF_UP, AF_UNSPEC); 2333 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP); 2334 } 2335 2336 /* 2337 * Flush an interface queue. 2338 */ 2339 void 2340 if_qflush(struct ifnet *ifp) 2341 { 2342 struct mbuf *m, *n; 2343 struct ifaltq *ifq; 2344 2345 ifq = &ifp->if_snd; 2346 IFQ_LOCK(ifq); 2347 #ifdef ALTQ 2348 if (ALTQ_IS_ENABLED(ifq)) 2349 ALTQ_PURGE(ifq); 2350 #endif 2351 n = ifq->ifq_head; 2352 while ((m = n) != NULL) { 2353 n = m->m_nextpkt; 2354 m_freem(m); 2355 } 2356 ifq->ifq_head = 0; 2357 ifq->ifq_tail = 0; 2358 ifq->ifq_len = 0; 2359 IFQ_UNLOCK(ifq); 2360 } 2361 2362 /* 2363 * Map interface name to interface structure pointer, with or without 2364 * returning a reference. 2365 */ 2366 struct ifnet * 2367 ifunit_ref(const char *name) 2368 { 2369 struct epoch_tracker et; 2370 struct ifnet *ifp; 2371 2372 NET_EPOCH_ENTER(et); 2373 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2374 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && 2375 !(ifp->if_flags & IFF_DYING)) 2376 break; 2377 } 2378 if (ifp != NULL) 2379 if_ref(ifp); 2380 NET_EPOCH_EXIT(et); 2381 return (ifp); 2382 } 2383 2384 struct ifnet * 2385 ifunit(const char *name) 2386 { 2387 struct epoch_tracker et; 2388 struct ifnet *ifp; 2389 2390 NET_EPOCH_ENTER(et); 2391 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2392 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 2393 break; 2394 } 2395 NET_EPOCH_EXIT(et); 2396 return (ifp); 2397 } 2398 2399 void * 2400 ifr_buffer_get_buffer(void *data) 2401 { 2402 union ifreq_union *ifrup; 2403 2404 ifrup = data; 2405 #ifdef COMPAT_FREEBSD32 2406 if (SV_CURPROC_FLAG(SV_ILP32)) 2407 return ((void *)(uintptr_t) 2408 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer); 2409 #endif 2410 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer); 2411 } 2412 2413 static void 2414 ifr_buffer_set_buffer_null(void *data) 2415 { 2416 union ifreq_union *ifrup; 2417 2418 ifrup = data; 2419 #ifdef COMPAT_FREEBSD32 2420 if (SV_CURPROC_FLAG(SV_ILP32)) 2421 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0; 2422 else 2423 #endif 2424 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL; 2425 } 2426 2427 size_t 2428 ifr_buffer_get_length(void *data) 2429 { 2430 union ifreq_union *ifrup; 2431 2432 ifrup = data; 2433 #ifdef COMPAT_FREEBSD32 2434 if (SV_CURPROC_FLAG(SV_ILP32)) 2435 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length); 2436 #endif 2437 return (ifrup->ifr.ifr_ifru.ifru_buffer.length); 2438 } 2439 2440 static void 2441 ifr_buffer_set_length(void *data, size_t len) 2442 { 2443 union ifreq_union *ifrup; 2444 2445 ifrup = data; 2446 #ifdef COMPAT_FREEBSD32 2447 if (SV_CURPROC_FLAG(SV_ILP32)) 2448 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len; 2449 else 2450 #endif 2451 ifrup->ifr.ifr_ifru.ifru_buffer.length = len; 2452 } 2453 2454 void * 2455 ifr_data_get_ptr(void *ifrp) 2456 { 2457 union ifreq_union *ifrup; 2458 2459 ifrup = ifrp; 2460 #ifdef COMPAT_FREEBSD32 2461 if (SV_CURPROC_FLAG(SV_ILP32)) 2462 return ((void *)(uintptr_t) 2463 ifrup->ifr32.ifr_ifru.ifru_data); 2464 #endif 2465 return (ifrup->ifr.ifr_ifru.ifru_data); 2466 } 2467 2468 /* 2469 * Hardware specific interface ioctls. 2470 */ 2471 int 2472 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 2473 { 2474 struct ifreq *ifr; 2475 int error = 0, do_ifup = 0; 2476 int new_flags, temp_flags; 2477 size_t namelen, onamelen; 2478 size_t descrlen; 2479 char *descrbuf, *odescrbuf; 2480 char new_name[IFNAMSIZ]; 2481 struct ifaddr *ifa; 2482 struct sockaddr_dl *sdl; 2483 2484 ifr = (struct ifreq *)data; 2485 switch (cmd) { 2486 case SIOCGIFINDEX: 2487 ifr->ifr_index = ifp->if_index; 2488 break; 2489 2490 case SIOCGIFFLAGS: 2491 temp_flags = ifp->if_flags | ifp->if_drv_flags; 2492 ifr->ifr_flags = temp_flags & 0xffff; 2493 ifr->ifr_flagshigh = temp_flags >> 16; 2494 break; 2495 2496 case SIOCGIFCAP: 2497 ifr->ifr_reqcap = ifp->if_capabilities; 2498 ifr->ifr_curcap = ifp->if_capenable; 2499 break; 2500 2501 case SIOCGIFDATA: 2502 { 2503 struct if_data ifd; 2504 2505 /* Ensure uninitialised padding is not leaked. */ 2506 memset(&ifd, 0, sizeof(ifd)); 2507 2508 if_data_copy(ifp, &ifd); 2509 error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd)); 2510 break; 2511 } 2512 2513 #ifdef MAC 2514 case SIOCGIFMAC: 2515 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); 2516 break; 2517 #endif 2518 2519 case SIOCGIFMETRIC: 2520 ifr->ifr_metric = ifp->if_metric; 2521 break; 2522 2523 case SIOCGIFMTU: 2524 ifr->ifr_mtu = ifp->if_mtu; 2525 break; 2526 2527 case SIOCGIFPHYS: 2528 /* XXXGL: did this ever worked? */ 2529 ifr->ifr_phys = 0; 2530 break; 2531 2532 case SIOCGIFDESCR: 2533 error = 0; 2534 sx_slock(&ifdescr_sx); 2535 if (ifp->if_description == NULL) 2536 error = ENOMSG; 2537 else { 2538 /* space for terminating nul */ 2539 descrlen = strlen(ifp->if_description) + 1; 2540 if (ifr_buffer_get_length(ifr) < descrlen) 2541 ifr_buffer_set_buffer_null(ifr); 2542 else 2543 error = copyout(ifp->if_description, 2544 ifr_buffer_get_buffer(ifr), descrlen); 2545 ifr_buffer_set_length(ifr, descrlen); 2546 } 2547 sx_sunlock(&ifdescr_sx); 2548 break; 2549 2550 case SIOCSIFDESCR: 2551 error = priv_check(td, PRIV_NET_SETIFDESCR); 2552 if (error) 2553 return (error); 2554 2555 /* 2556 * Copy only (length-1) bytes to make sure that 2557 * if_description is always nul terminated. The 2558 * length parameter is supposed to count the 2559 * terminating nul in. 2560 */ 2561 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen) 2562 return (ENAMETOOLONG); 2563 else if (ifr_buffer_get_length(ifr) == 0) 2564 descrbuf = NULL; 2565 else { 2566 descrbuf = malloc(ifr_buffer_get_length(ifr), 2567 M_IFDESCR, M_WAITOK | M_ZERO); 2568 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf, 2569 ifr_buffer_get_length(ifr) - 1); 2570 if (error) { 2571 free(descrbuf, M_IFDESCR); 2572 break; 2573 } 2574 } 2575 2576 sx_xlock(&ifdescr_sx); 2577 odescrbuf = ifp->if_description; 2578 ifp->if_description = descrbuf; 2579 sx_xunlock(&ifdescr_sx); 2580 2581 getmicrotime(&ifp->if_lastchange); 2582 free(odescrbuf, M_IFDESCR); 2583 break; 2584 2585 case SIOCGIFFIB: 2586 ifr->ifr_fib = ifp->if_fib; 2587 break; 2588 2589 case SIOCSIFFIB: 2590 error = priv_check(td, PRIV_NET_SETIFFIB); 2591 if (error) 2592 return (error); 2593 if (ifr->ifr_fib >= rt_numfibs) 2594 return (EINVAL); 2595 2596 ifp->if_fib = ifr->ifr_fib; 2597 break; 2598 2599 case SIOCSIFFLAGS: 2600 error = priv_check(td, PRIV_NET_SETIFFLAGS); 2601 if (error) 2602 return (error); 2603 /* 2604 * Currently, no driver owned flags pass the IFF_CANTCHANGE 2605 * check, so we don't need special handling here yet. 2606 */ 2607 new_flags = (ifr->ifr_flags & 0xffff) | 2608 (ifr->ifr_flagshigh << 16); 2609 if (ifp->if_flags & IFF_UP && 2610 (new_flags & IFF_UP) == 0) { 2611 if_down(ifp); 2612 } else if (new_flags & IFF_UP && 2613 (ifp->if_flags & IFF_UP) == 0) { 2614 do_ifup = 1; 2615 } 2616 /* See if permanently promiscuous mode bit is about to flip */ 2617 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { 2618 if (new_flags & IFF_PPROMISC) 2619 ifp->if_flags |= IFF_PROMISC; 2620 else if (ifp->if_pcount == 0) 2621 ifp->if_flags &= ~IFF_PROMISC; 2622 if (log_promisc_mode_change) 2623 if_printf(ifp, "permanently promiscuous mode %s\n", 2624 ((new_flags & IFF_PPROMISC) ? 2625 "enabled" : "disabled")); 2626 } 2627 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 2628 (new_flags &~ IFF_CANTCHANGE); 2629 if (ifp->if_ioctl) { 2630 (void) (*ifp->if_ioctl)(ifp, cmd, data); 2631 } 2632 if (do_ifup) 2633 if_up(ifp); 2634 getmicrotime(&ifp->if_lastchange); 2635 break; 2636 2637 case SIOCSIFCAP: 2638 error = priv_check(td, PRIV_NET_SETIFCAP); 2639 if (error) 2640 return (error); 2641 if (ifp->if_ioctl == NULL) 2642 return (EOPNOTSUPP); 2643 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 2644 return (EINVAL); 2645 error = (*ifp->if_ioctl)(ifp, cmd, data); 2646 if (error == 0) 2647 getmicrotime(&ifp->if_lastchange); 2648 break; 2649 2650 #ifdef MAC 2651 case SIOCSIFMAC: 2652 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); 2653 break; 2654 #endif 2655 2656 case SIOCSIFNAME: 2657 error = priv_check(td, PRIV_NET_SETIFNAME); 2658 if (error) 2659 return (error); 2660 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ, 2661 NULL); 2662 if (error != 0) 2663 return (error); 2664 if (new_name[0] == '\0') 2665 return (EINVAL); 2666 if (new_name[IFNAMSIZ-1] != '\0') { 2667 new_name[IFNAMSIZ-1] = '\0'; 2668 if (strlen(new_name) == IFNAMSIZ-1) 2669 return (EINVAL); 2670 } 2671 if (strcmp(new_name, ifp->if_xname) == 0) 2672 break; 2673 if (ifunit(new_name) != NULL) 2674 return (EEXIST); 2675 2676 /* 2677 * XXX: Locking. Nothing else seems to lock if_flags, 2678 * and there are numerous other races with the 2679 * ifunit() checks not being atomic with namespace 2680 * changes (renames, vmoves, if_attach, etc). 2681 */ 2682 ifp->if_flags |= IFF_RENAMING; 2683 2684 /* Announce the departure of the interface. */ 2685 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 2686 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 2687 2688 if_printf(ifp, "changing name to '%s'\n", new_name); 2689 2690 IF_ADDR_WLOCK(ifp); 2691 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 2692 ifa = ifp->if_addr; 2693 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 2694 namelen = strlen(new_name); 2695 onamelen = sdl->sdl_nlen; 2696 /* 2697 * Move the address if needed. This is safe because we 2698 * allocate space for a name of length IFNAMSIZ when we 2699 * create this in if_attach(). 2700 */ 2701 if (namelen != onamelen) { 2702 bcopy(sdl->sdl_data + onamelen, 2703 sdl->sdl_data + namelen, sdl->sdl_alen); 2704 } 2705 bcopy(new_name, sdl->sdl_data, namelen); 2706 sdl->sdl_nlen = namelen; 2707 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 2708 bzero(sdl->sdl_data, onamelen); 2709 while (namelen != 0) 2710 sdl->sdl_data[--namelen] = 0xff; 2711 IF_ADDR_WUNLOCK(ifp); 2712 2713 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 2714 /* Announce the return of the interface. */ 2715 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 2716 2717 ifp->if_flags &= ~IFF_RENAMING; 2718 break; 2719 2720 #ifdef VIMAGE 2721 case SIOCSIFVNET: 2722 error = priv_check(td, PRIV_NET_SETIFVNET); 2723 if (error) 2724 return (error); 2725 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); 2726 break; 2727 #endif 2728 2729 case SIOCSIFMETRIC: 2730 error = priv_check(td, PRIV_NET_SETIFMETRIC); 2731 if (error) 2732 return (error); 2733 ifp->if_metric = ifr->ifr_metric; 2734 getmicrotime(&ifp->if_lastchange); 2735 break; 2736 2737 case SIOCSIFPHYS: 2738 error = priv_check(td, PRIV_NET_SETIFPHYS); 2739 if (error) 2740 return (error); 2741 if (ifp->if_ioctl == NULL) 2742 return (EOPNOTSUPP); 2743 error = (*ifp->if_ioctl)(ifp, cmd, data); 2744 if (error == 0) 2745 getmicrotime(&ifp->if_lastchange); 2746 break; 2747 2748 case SIOCSIFMTU: 2749 { 2750 u_long oldmtu = ifp->if_mtu; 2751 2752 error = priv_check(td, PRIV_NET_SETIFMTU); 2753 if (error) 2754 return (error); 2755 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 2756 return (EINVAL); 2757 if (ifp->if_ioctl == NULL) 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 CASE_IOC_IFGROUPREQ(SIOCAIFGROUP): 2865 error = priv_check(td, PRIV_NET_ADDIFGROUP); 2866 if (error) 2867 return (error); 2868 if ((error = if_addgroup(ifp, 2869 ifgr_group_get((struct ifgroupreq *)data)))) 2870 return (error); 2871 break; 2872 2873 case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP): 2874 { 2875 struct epoch_tracker et; 2876 2877 NET_EPOCH_ENTER(et); 2878 error = if_getgroup((struct ifgroupreq *)data, ifp); 2879 NET_EPOCH_EXIT(et); 2880 break; 2881 } 2882 2883 case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP): 2884 error = priv_check(td, PRIV_NET_DELIFGROUP); 2885 if (error) 2886 return (error); 2887 if ((error = if_delgroup(ifp, 2888 ifgr_group_get((struct ifgroupreq *)data)))) 2889 return (error); 2890 break; 2891 2892 default: 2893 error = ENOIOCTL; 2894 break; 2895 } 2896 return (error); 2897 } 2898 2899 #ifdef COMPAT_FREEBSD32 2900 struct ifconf32 { 2901 int32_t ifc_len; 2902 union { 2903 uint32_t ifcu_buf; 2904 uint32_t ifcu_req; 2905 } ifc_ifcu; 2906 }; 2907 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) 2908 #endif 2909 2910 #ifdef COMPAT_FREEBSD32 2911 static void 2912 ifmr_init(struct ifmediareq *ifmr, caddr_t data) 2913 { 2914 struct ifmediareq32 *ifmr32; 2915 2916 ifmr32 = (struct ifmediareq32 *)data; 2917 memcpy(ifmr->ifm_name, ifmr32->ifm_name, 2918 sizeof(ifmr->ifm_name)); 2919 ifmr->ifm_current = ifmr32->ifm_current; 2920 ifmr->ifm_mask = ifmr32->ifm_mask; 2921 ifmr->ifm_status = ifmr32->ifm_status; 2922 ifmr->ifm_active = ifmr32->ifm_active; 2923 ifmr->ifm_count = ifmr32->ifm_count; 2924 ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist; 2925 } 2926 2927 static void 2928 ifmr_update(const struct ifmediareq *ifmr, caddr_t data) 2929 { 2930 struct ifmediareq32 *ifmr32; 2931 2932 ifmr32 = (struct ifmediareq32 *)data; 2933 ifmr32->ifm_current = ifmr->ifm_current; 2934 ifmr32->ifm_mask = ifmr->ifm_mask; 2935 ifmr32->ifm_status = ifmr->ifm_status; 2936 ifmr32->ifm_active = ifmr->ifm_active; 2937 ifmr32->ifm_count = ifmr->ifm_count; 2938 } 2939 #endif 2940 2941 /* 2942 * Interface ioctls. 2943 */ 2944 int 2945 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 2946 { 2947 #ifdef COMPAT_FREEBSD32 2948 caddr_t saved_data = NULL; 2949 struct ifmediareq ifmr; 2950 struct ifmediareq *ifmrp = NULL; 2951 #endif 2952 struct ifnet *ifp; 2953 struct ifreq *ifr; 2954 int error; 2955 int oif_flags; 2956 #ifdef VIMAGE 2957 bool shutdown; 2958 #endif 2959 2960 CURVNET_SET(so->so_vnet); 2961 #ifdef VIMAGE 2962 /* Make sure the VNET is stable. */ 2963 shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet); 2964 if (shutdown) { 2965 CURVNET_RESTORE(); 2966 return (EBUSY); 2967 } 2968 #endif 2969 2970 switch (cmd) { 2971 case SIOCGIFCONF: 2972 error = ifconf(cmd, data); 2973 goto out_noref; 2974 2975 #ifdef COMPAT_FREEBSD32 2976 case SIOCGIFCONF32: 2977 { 2978 struct ifconf32 *ifc32; 2979 struct ifconf ifc; 2980 2981 ifc32 = (struct ifconf32 *)data; 2982 ifc.ifc_len = ifc32->ifc_len; 2983 ifc.ifc_buf = PTRIN(ifc32->ifc_buf); 2984 2985 error = ifconf(SIOCGIFCONF, (void *)&ifc); 2986 if (error == 0) 2987 ifc32->ifc_len = ifc.ifc_len; 2988 goto out_noref; 2989 } 2990 #endif 2991 } 2992 2993 #ifdef COMPAT_FREEBSD32 2994 switch (cmd) { 2995 case SIOCGIFMEDIA32: 2996 case SIOCGIFXMEDIA32: 2997 ifmrp = &ifmr; 2998 ifmr_init(ifmrp, data); 2999 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq); 3000 saved_data = data; 3001 data = (caddr_t)ifmrp; 3002 } 3003 #endif 3004 3005 ifr = (struct ifreq *)data; 3006 switch (cmd) { 3007 #ifdef VIMAGE 3008 case SIOCSIFRVNET: 3009 error = priv_check(td, PRIV_NET_SETIFVNET); 3010 if (error == 0) 3011 error = if_vmove_reclaim(td, ifr->ifr_name, 3012 ifr->ifr_jid); 3013 goto out_noref; 3014 #endif 3015 case SIOCIFCREATE: 3016 case SIOCIFCREATE2: 3017 error = priv_check(td, PRIV_NET_IFCREATE); 3018 if (error == 0) 3019 error = if_clone_create(ifr->ifr_name, 3020 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? 3021 ifr_data_get_ptr(ifr) : NULL); 3022 goto out_noref; 3023 case SIOCIFDESTROY: 3024 error = priv_check(td, PRIV_NET_IFDESTROY); 3025 3026 if (error == 0) { 3027 sx_xlock(&ifnet_detach_sxlock); 3028 error = if_clone_destroy(ifr->ifr_name); 3029 sx_xunlock(&ifnet_detach_sxlock); 3030 } 3031 goto out_noref; 3032 3033 case SIOCIFGCLONERS: 3034 error = if_clone_list((struct if_clonereq *)data); 3035 goto out_noref; 3036 3037 case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB): 3038 error = if_getgroupmembers((struct ifgroupreq *)data); 3039 goto out_noref; 3040 3041 #if defined(INET) || defined(INET6) 3042 case SIOCSVH: 3043 case SIOCGVH: 3044 if (carp_ioctl_p == NULL) 3045 error = EPROTONOSUPPORT; 3046 else 3047 error = (*carp_ioctl_p)(ifr, cmd, td); 3048 goto out_noref; 3049 #endif 3050 } 3051 3052 ifp = ifunit_ref(ifr->ifr_name); 3053 if (ifp == NULL) { 3054 error = ENXIO; 3055 goto out_noref; 3056 } 3057 3058 error = ifhwioctl(cmd, ifp, data, td); 3059 if (error != ENOIOCTL) 3060 goto out_ref; 3061 3062 oif_flags = ifp->if_flags; 3063 if (so->so_proto == NULL) { 3064 error = EOPNOTSUPP; 3065 goto out_ref; 3066 } 3067 3068 /* 3069 * Pass the request on to the socket control method, and if the 3070 * latter returns EOPNOTSUPP, directly to the interface. 3071 * 3072 * Make an exception for the legacy SIOCSIF* requests. Drivers 3073 * trust SIOCSIFADDR et al to come from an already privileged 3074 * layer, and do not perform any credentials checks or input 3075 * validation. 3076 */ 3077 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, 3078 ifp, td)); 3079 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && 3080 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && 3081 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) 3082 error = (*ifp->if_ioctl)(ifp, cmd, data); 3083 3084 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 3085 #ifdef INET6 3086 if (ifp->if_flags & IFF_UP) 3087 in6_if_up(ifp); 3088 #endif 3089 } 3090 3091 out_ref: 3092 if_rele(ifp); 3093 out_noref: 3094 #ifdef COMPAT_FREEBSD32 3095 if (ifmrp != NULL) { 3096 KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA), 3097 ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx", 3098 cmd)); 3099 data = saved_data; 3100 ifmr_update(ifmrp, data); 3101 } 3102 #endif 3103 CURVNET_RESTORE(); 3104 return (error); 3105 } 3106 3107 /* 3108 * The code common to handling reference counted flags, 3109 * e.g., in ifpromisc() and if_allmulti(). 3110 * The "pflag" argument can specify a permanent mode flag to check, 3111 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. 3112 * 3113 * Only to be used on stack-owned flags, not driver-owned flags. 3114 */ 3115 static int 3116 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) 3117 { 3118 struct ifreq ifr; 3119 int error; 3120 int oldflags, oldcount; 3121 3122 /* Sanity checks to catch programming errors */ 3123 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, 3124 ("%s: setting driver-owned flag %d", __func__, flag)); 3125 3126 if (onswitch) 3127 KASSERT(*refcount >= 0, 3128 ("%s: increment negative refcount %d for flag %d", 3129 __func__, *refcount, flag)); 3130 else 3131 KASSERT(*refcount > 0, 3132 ("%s: decrement non-positive refcount %d for flag %d", 3133 __func__, *refcount, flag)); 3134 3135 /* In case this mode is permanent, just touch refcount */ 3136 if (ifp->if_flags & pflag) { 3137 *refcount += onswitch ? 1 : -1; 3138 return (0); 3139 } 3140 3141 /* Save ifnet parameters for if_ioctl() may fail */ 3142 oldcount = *refcount; 3143 oldflags = ifp->if_flags; 3144 3145 /* 3146 * See if we aren't the only and touching refcount is enough. 3147 * Actually toggle interface flag if we are the first or last. 3148 */ 3149 if (onswitch) { 3150 if ((*refcount)++) 3151 return (0); 3152 ifp->if_flags |= flag; 3153 } else { 3154 if (--(*refcount)) 3155 return (0); 3156 ifp->if_flags &= ~flag; 3157 } 3158 3159 /* Call down the driver since we've changed interface flags */ 3160 if (ifp->if_ioctl == NULL) { 3161 error = EOPNOTSUPP; 3162 goto recover; 3163 } 3164 ifr.ifr_flags = ifp->if_flags & 0xffff; 3165 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3166 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3167 if (error) 3168 goto recover; 3169 /* Notify userland that interface flags have changed */ 3170 rt_ifmsg(ifp); 3171 return (0); 3172 3173 recover: 3174 /* Recover after driver error */ 3175 *refcount = oldcount; 3176 ifp->if_flags = oldflags; 3177 return (error); 3178 } 3179 3180 /* 3181 * Set/clear promiscuous mode on interface ifp based on the truth value 3182 * of pswitch. The calls are reference counted so that only the first 3183 * "on" request actually has an effect, as does the final "off" request. 3184 * Results are undefined if the "off" and "on" requests are not matched. 3185 */ 3186 int 3187 ifpromisc(struct ifnet *ifp, int pswitch) 3188 { 3189 int error; 3190 int oldflags = ifp->if_flags; 3191 3192 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, 3193 &ifp->if_pcount, pswitch); 3194 /* If promiscuous mode status has changed, log a message */ 3195 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) && 3196 log_promisc_mode_change) 3197 if_printf(ifp, "promiscuous mode %s\n", 3198 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 3199 return (error); 3200 } 3201 3202 /* 3203 * Return interface configuration 3204 * of system. List may be used 3205 * in later ioctl's (above) to get 3206 * other information. 3207 */ 3208 /*ARGSUSED*/ 3209 static int 3210 ifconf(u_long cmd, caddr_t data) 3211 { 3212 struct ifconf *ifc = (struct ifconf *)data; 3213 struct ifnet *ifp; 3214 struct ifaddr *ifa; 3215 struct ifreq ifr; 3216 struct sbuf *sb; 3217 int error, full = 0, valid_len, max_len; 3218 3219 /* Limit initial buffer size to maxphys to avoid DoS from userspace. */ 3220 max_len = maxphys - 1; 3221 3222 /* Prevent hostile input from being able to crash the system */ 3223 if (ifc->ifc_len <= 0) 3224 return (EINVAL); 3225 3226 again: 3227 if (ifc->ifc_len <= max_len) { 3228 max_len = ifc->ifc_len; 3229 full = 1; 3230 } 3231 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); 3232 max_len = 0; 3233 valid_len = 0; 3234 3235 IFNET_RLOCK(); 3236 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 3237 struct epoch_tracker et; 3238 int addrs; 3239 3240 /* 3241 * Zero the ifr to make sure we don't disclose the contents 3242 * of the stack. 3243 */ 3244 memset(&ifr, 0, sizeof(ifr)); 3245 3246 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 3247 >= sizeof(ifr.ifr_name)) { 3248 sbuf_delete(sb); 3249 IFNET_RUNLOCK(); 3250 return (ENAMETOOLONG); 3251 } 3252 3253 addrs = 0; 3254 NET_EPOCH_ENTER(et); 3255 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 3256 struct sockaddr *sa = ifa->ifa_addr; 3257 3258 if (prison_if(curthread->td_ucred, sa) != 0) 3259 continue; 3260 addrs++; 3261 if (sa->sa_len <= sizeof(*sa)) { 3262 if (sa->sa_len < sizeof(*sa)) { 3263 memset(&ifr.ifr_ifru.ifru_addr, 0, 3264 sizeof(ifr.ifr_ifru.ifru_addr)); 3265 memcpy(&ifr.ifr_ifru.ifru_addr, sa, 3266 sa->sa_len); 3267 } else 3268 ifr.ifr_ifru.ifru_addr = *sa; 3269 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3270 max_len += sizeof(ifr); 3271 } else { 3272 sbuf_bcat(sb, &ifr, 3273 offsetof(struct ifreq, ifr_addr)); 3274 max_len += offsetof(struct ifreq, ifr_addr); 3275 sbuf_bcat(sb, sa, sa->sa_len); 3276 max_len += sa->sa_len; 3277 } 3278 3279 if (sbuf_error(sb) == 0) 3280 valid_len = sbuf_len(sb); 3281 } 3282 NET_EPOCH_EXIT(et); 3283 if (addrs == 0) { 3284 sbuf_bcat(sb, &ifr, sizeof(ifr)); 3285 max_len += sizeof(ifr); 3286 3287 if (sbuf_error(sb) == 0) 3288 valid_len = sbuf_len(sb); 3289 } 3290 } 3291 IFNET_RUNLOCK(); 3292 3293 /* 3294 * If we didn't allocate enough space (uncommon), try again. If 3295 * we have already allocated as much space as we are allowed, 3296 * return what we've got. 3297 */ 3298 if (valid_len != max_len && !full) { 3299 sbuf_delete(sb); 3300 goto again; 3301 } 3302 3303 ifc->ifc_len = valid_len; 3304 sbuf_finish(sb); 3305 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); 3306 sbuf_delete(sb); 3307 return (error); 3308 } 3309 3310 /* 3311 * Just like ifpromisc(), but for all-multicast-reception mode. 3312 */ 3313 int 3314 if_allmulti(struct ifnet *ifp, int onswitch) 3315 { 3316 3317 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); 3318 } 3319 3320 struct ifmultiaddr * 3321 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) 3322 { 3323 struct ifmultiaddr *ifma; 3324 3325 IF_ADDR_LOCK_ASSERT(ifp); 3326 3327 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 3328 if (sa->sa_family == AF_LINK) { 3329 if (sa_dl_equal(ifma->ifma_addr, sa)) 3330 break; 3331 } else { 3332 if (sa_equal(ifma->ifma_addr, sa)) 3333 break; 3334 } 3335 } 3336 3337 return ifma; 3338 } 3339 3340 /* 3341 * Allocate a new ifmultiaddr and initialize based on passed arguments. We 3342 * make copies of passed sockaddrs. The ifmultiaddr will not be added to 3343 * the ifnet multicast address list here, so the caller must do that and 3344 * other setup work (such as notifying the device driver). The reference 3345 * count is initialized to 1. 3346 */ 3347 static struct ifmultiaddr * 3348 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, 3349 int mflags) 3350 { 3351 struct ifmultiaddr *ifma; 3352 struct sockaddr *dupsa; 3353 3354 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | 3355 M_ZERO); 3356 if (ifma == NULL) 3357 return (NULL); 3358 3359 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); 3360 if (dupsa == NULL) { 3361 free(ifma, M_IFMADDR); 3362 return (NULL); 3363 } 3364 bcopy(sa, dupsa, sa->sa_len); 3365 ifma->ifma_addr = dupsa; 3366 3367 ifma->ifma_ifp = ifp; 3368 ifma->ifma_refcount = 1; 3369 ifma->ifma_protospec = NULL; 3370 3371 if (llsa == NULL) { 3372 ifma->ifma_lladdr = NULL; 3373 return (ifma); 3374 } 3375 3376 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); 3377 if (dupsa == NULL) { 3378 free(ifma->ifma_addr, M_IFMADDR); 3379 free(ifma, M_IFMADDR); 3380 return (NULL); 3381 } 3382 bcopy(llsa, dupsa, llsa->sa_len); 3383 ifma->ifma_lladdr = dupsa; 3384 3385 return (ifma); 3386 } 3387 3388 /* 3389 * if_freemulti: free ifmultiaddr structure and possibly attached related 3390 * addresses. The caller is responsible for implementing reference 3391 * counting, notifying the driver, handling routing messages, and releasing 3392 * any dependent link layer state. 3393 */ 3394 #ifdef MCAST_VERBOSE 3395 extern void kdb_backtrace(void); 3396 #endif 3397 static void 3398 if_freemulti_internal(struct ifmultiaddr *ifma) 3399 { 3400 3401 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", 3402 ifma->ifma_refcount)); 3403 3404 if (ifma->ifma_lladdr != NULL) 3405 free(ifma->ifma_lladdr, M_IFMADDR); 3406 #ifdef MCAST_VERBOSE 3407 kdb_backtrace(); 3408 printf("%s freeing ifma: %p\n", __func__, ifma); 3409 #endif 3410 free(ifma->ifma_addr, M_IFMADDR); 3411 free(ifma, M_IFMADDR); 3412 } 3413 3414 static void 3415 if_destroymulti(epoch_context_t ctx) 3416 { 3417 struct ifmultiaddr *ifma; 3418 3419 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx); 3420 if_freemulti_internal(ifma); 3421 } 3422 3423 void 3424 if_freemulti(struct ifmultiaddr *ifma) 3425 { 3426 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d", 3427 ifma->ifma_refcount)); 3428 3429 NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx); 3430 } 3431 3432 /* 3433 * Register an additional multicast address with a network interface. 3434 * 3435 * - If the address is already present, bump the reference count on the 3436 * address and return. 3437 * - If the address is not link-layer, look up a link layer address. 3438 * - Allocate address structures for one or both addresses, and attach to the 3439 * multicast address list on the interface. If automatically adding a link 3440 * layer address, the protocol address will own a reference to the link 3441 * layer address, to be freed when it is freed. 3442 * - Notify the network device driver of an addition to the multicast address 3443 * list. 3444 * 3445 * 'sa' points to caller-owned memory with the desired multicast address. 3446 * 3447 * 'retifma' will be used to return a pointer to the resulting multicast 3448 * address reference, if desired. 3449 */ 3450 int 3451 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, 3452 struct ifmultiaddr **retifma) 3453 { 3454 struct ifmultiaddr *ifma, *ll_ifma; 3455 struct sockaddr *llsa; 3456 struct sockaddr_dl sdl; 3457 int error; 3458 3459 #ifdef INET 3460 IN_MULTI_LIST_UNLOCK_ASSERT(); 3461 #endif 3462 #ifdef INET6 3463 IN6_MULTI_LIST_UNLOCK_ASSERT(); 3464 #endif 3465 /* 3466 * If the address is already present, return a new reference to it; 3467 * otherwise, allocate storage and set up a new address. 3468 */ 3469 IF_ADDR_WLOCK(ifp); 3470 ifma = if_findmulti(ifp, sa); 3471 if (ifma != NULL) { 3472 ifma->ifma_refcount++; 3473 if (retifma != NULL) 3474 *retifma = ifma; 3475 IF_ADDR_WUNLOCK(ifp); 3476 return (0); 3477 } 3478 3479 /* 3480 * The address isn't already present; resolve the protocol address 3481 * into a link layer address, and then look that up, bump its 3482 * refcount or allocate an ifma for that also. 3483 * Most link layer resolving functions returns address data which 3484 * fits inside default sockaddr_dl structure. However callback 3485 * can allocate another sockaddr structure, in that case we need to 3486 * free it later. 3487 */ 3488 llsa = NULL; 3489 ll_ifma = NULL; 3490 if (ifp->if_resolvemulti != NULL) { 3491 /* Provide called function with buffer size information */ 3492 sdl.sdl_len = sizeof(sdl); 3493 llsa = (struct sockaddr *)&sdl; 3494 error = ifp->if_resolvemulti(ifp, &llsa, sa); 3495 if (error) 3496 goto unlock_out; 3497 } 3498 3499 /* 3500 * Allocate the new address. Don't hook it up yet, as we may also 3501 * need to allocate a link layer multicast address. 3502 */ 3503 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); 3504 if (ifma == NULL) { 3505 error = ENOMEM; 3506 goto free_llsa_out; 3507 } 3508 3509 /* 3510 * If a link layer address is found, we'll need to see if it's 3511 * already present in the address list, or allocate is as well. 3512 * When this block finishes, the link layer address will be on the 3513 * list. 3514 */ 3515 if (llsa != NULL) { 3516 ll_ifma = if_findmulti(ifp, llsa); 3517 if (ll_ifma == NULL) { 3518 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); 3519 if (ll_ifma == NULL) { 3520 --ifma->ifma_refcount; 3521 if_freemulti(ifma); 3522 error = ENOMEM; 3523 goto free_llsa_out; 3524 } 3525 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED; 3526 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, 3527 ifma_link); 3528 } else 3529 ll_ifma->ifma_refcount++; 3530 ifma->ifma_llifma = ll_ifma; 3531 } 3532 3533 /* 3534 * We now have a new multicast address, ifma, and possibly a new or 3535 * referenced link layer address. Add the primary address to the 3536 * ifnet address list. 3537 */ 3538 ifma->ifma_flags |= IFMA_F_ENQUEUED; 3539 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 3540 3541 if (retifma != NULL) 3542 *retifma = ifma; 3543 3544 /* 3545 * Must generate the message while holding the lock so that 'ifma' 3546 * pointer is still valid. 3547 */ 3548 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 3549 IF_ADDR_WUNLOCK(ifp); 3550 3551 /* 3552 * We are certain we have added something, so call down to the 3553 * interface to let them know about it. 3554 */ 3555 if (ifp->if_ioctl != NULL) { 3556 if (THREAD_CAN_SLEEP()) 3557 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3558 else 3559 taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask); 3560 } 3561 3562 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3563 link_free_sdl(llsa); 3564 3565 return (0); 3566 3567 free_llsa_out: 3568 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) 3569 link_free_sdl(llsa); 3570 3571 unlock_out: 3572 IF_ADDR_WUNLOCK(ifp); 3573 return (error); 3574 } 3575 3576 static void 3577 if_siocaddmulti(void *arg, int pending) 3578 { 3579 struct ifnet *ifp; 3580 3581 ifp = arg; 3582 #ifdef DIAGNOSTIC 3583 if (pending > 1) 3584 if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending); 3585 #endif 3586 CURVNET_SET(ifp->if_vnet); 3587 (void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); 3588 CURVNET_RESTORE(); 3589 } 3590 3591 /* 3592 * Delete a multicast group membership by network-layer group address. 3593 * 3594 * Returns ENOENT if the entry could not be found. If ifp no longer 3595 * exists, results are undefined. This entry point should only be used 3596 * from subsystems which do appropriate locking to hold ifp for the 3597 * duration of the call. 3598 * Network-layer protocol domains must use if_delmulti_ifma(). 3599 */ 3600 int 3601 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 3602 { 3603 struct ifmultiaddr *ifma; 3604 int lastref; 3605 3606 KASSERT(ifp, ("%s: NULL ifp", __func__)); 3607 3608 IF_ADDR_WLOCK(ifp); 3609 lastref = 0; 3610 ifma = if_findmulti(ifp, sa); 3611 if (ifma != NULL) 3612 lastref = if_delmulti_locked(ifp, ifma, 0); 3613 IF_ADDR_WUNLOCK(ifp); 3614 3615 if (ifma == NULL) 3616 return (ENOENT); 3617 3618 if (lastref && ifp->if_ioctl != NULL) { 3619 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3620 } 3621 3622 return (0); 3623 } 3624 3625 /* 3626 * Delete all multicast group membership for an interface. 3627 * Should be used to quickly flush all multicast filters. 3628 */ 3629 void 3630 if_delallmulti(struct ifnet *ifp) 3631 { 3632 struct ifmultiaddr *ifma; 3633 struct ifmultiaddr *next; 3634 3635 IF_ADDR_WLOCK(ifp); 3636 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) 3637 if_delmulti_locked(ifp, ifma, 0); 3638 IF_ADDR_WUNLOCK(ifp); 3639 } 3640 3641 void 3642 if_delmulti_ifma(struct ifmultiaddr *ifma) 3643 { 3644 if_delmulti_ifma_flags(ifma, 0); 3645 } 3646 3647 /* 3648 * Delete a multicast group membership by group membership pointer. 3649 * Network-layer protocol domains must use this routine. 3650 * 3651 * It is safe to call this routine if the ifp disappeared. 3652 */ 3653 void 3654 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags) 3655 { 3656 struct ifnet *ifp; 3657 int lastref; 3658 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma); 3659 #ifdef INET 3660 IN_MULTI_LIST_UNLOCK_ASSERT(); 3661 #endif 3662 ifp = ifma->ifma_ifp; 3663 #ifdef DIAGNOSTIC 3664 if (ifp == NULL) { 3665 printf("%s: ifma_ifp seems to be detached\n", __func__); 3666 } else { 3667 struct epoch_tracker et; 3668 struct ifnet *oifp; 3669 3670 NET_EPOCH_ENTER(et); 3671 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link) 3672 if (ifp == oifp) 3673 break; 3674 NET_EPOCH_EXIT(et); 3675 if (ifp != oifp) 3676 ifp = NULL; 3677 } 3678 #endif 3679 /* 3680 * If and only if the ifnet instance exists: Acquire the address lock. 3681 */ 3682 if (ifp != NULL) 3683 IF_ADDR_WLOCK(ifp); 3684 3685 lastref = if_delmulti_locked(ifp, ifma, flags); 3686 3687 if (ifp != NULL) { 3688 /* 3689 * If and only if the ifnet instance exists: 3690 * Release the address lock. 3691 * If the group was left: update the hardware hash filter. 3692 */ 3693 IF_ADDR_WUNLOCK(ifp); 3694 if (lastref && ifp->if_ioctl != NULL) { 3695 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); 3696 } 3697 } 3698 } 3699 3700 /* 3701 * Perform deletion of network-layer and/or link-layer multicast address. 3702 * 3703 * Return 0 if the reference count was decremented. 3704 * Return 1 if the final reference was released, indicating that the 3705 * hardware hash filter should be reprogrammed. 3706 */ 3707 static int 3708 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) 3709 { 3710 struct ifmultiaddr *ll_ifma; 3711 3712 if (ifp != NULL && ifma->ifma_ifp != NULL) { 3713 KASSERT(ifma->ifma_ifp == ifp, 3714 ("%s: inconsistent ifp %p", __func__, ifp)); 3715 IF_ADDR_WLOCK_ASSERT(ifp); 3716 } 3717 3718 ifp = ifma->ifma_ifp; 3719 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : ""); 3720 3721 /* 3722 * If the ifnet is detaching, null out references to ifnet, 3723 * so that upper protocol layers will notice, and not attempt 3724 * to obtain locks for an ifnet which no longer exists. The 3725 * routing socket announcement must happen before the ifnet 3726 * instance is detached from the system. 3727 */ 3728 if (detaching) { 3729 #ifdef DIAGNOSTIC 3730 printf("%s: detaching ifnet instance %p\n", __func__, ifp); 3731 #endif 3732 /* 3733 * ifp may already be nulled out if we are being reentered 3734 * to delete the ll_ifma. 3735 */ 3736 if (ifp != NULL) { 3737 rt_newmaddrmsg(RTM_DELMADDR, ifma); 3738 ifma->ifma_ifp = NULL; 3739 } 3740 } 3741 3742 if (--ifma->ifma_refcount > 0) 3743 return 0; 3744 3745 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) { 3746 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link); 3747 ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3748 } 3749 /* 3750 * If this ifma is a network-layer ifma, a link-layer ifma may 3751 * have been associated with it. Release it first if so. 3752 */ 3753 ll_ifma = ifma->ifma_llifma; 3754 if (ll_ifma != NULL) { 3755 KASSERT(ifma->ifma_lladdr != NULL, 3756 ("%s: llifma w/o lladdr", __func__)); 3757 if (detaching) 3758 ll_ifma->ifma_ifp = NULL; /* XXX */ 3759 if (--ll_ifma->ifma_refcount == 0) { 3760 if (ifp != NULL) { 3761 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) { 3762 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr, 3763 ifma_link); 3764 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED; 3765 } 3766 } 3767 if_freemulti(ll_ifma); 3768 } 3769 } 3770 #ifdef INVARIANTS 3771 if (ifp) { 3772 struct ifmultiaddr *ifmatmp; 3773 3774 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link) 3775 MPASS(ifma != ifmatmp); 3776 } 3777 #endif 3778 if_freemulti(ifma); 3779 /* 3780 * The last reference to this instance of struct ifmultiaddr 3781 * was released; the hardware should be notified of this change. 3782 */ 3783 return 1; 3784 } 3785 3786 /* 3787 * Set the link layer address on an interface. 3788 * 3789 * At this time we only support certain types of interfaces, 3790 * and we don't allow the length of the address to change. 3791 * 3792 * Set noinline to be dtrace-friendly 3793 */ 3794 __noinline int 3795 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 3796 { 3797 struct sockaddr_dl *sdl; 3798 struct ifaddr *ifa; 3799 struct ifreq ifr; 3800 3801 ifa = ifp->if_addr; 3802 if (ifa == NULL) 3803 return (EINVAL); 3804 3805 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 3806 if (sdl == NULL) 3807 return (EINVAL); 3808 3809 if (len != sdl->sdl_alen) /* don't allow length to change */ 3810 return (EINVAL); 3811 3812 switch (ifp->if_type) { 3813 case IFT_ETHER: 3814 case IFT_XETHER: 3815 case IFT_L2VLAN: 3816 case IFT_BRIDGE: 3817 case IFT_IEEE8023ADLAG: 3818 bcopy(lladdr, LLADDR(sdl), len); 3819 break; 3820 default: 3821 return (ENODEV); 3822 } 3823 3824 /* 3825 * If the interface is already up, we need 3826 * to re-init it in order to reprogram its 3827 * address filter. 3828 */ 3829 if ((ifp->if_flags & IFF_UP) != 0) { 3830 if (ifp->if_ioctl) { 3831 ifp->if_flags &= ~IFF_UP; 3832 ifr.ifr_flags = ifp->if_flags & 0xffff; 3833 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3834 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3835 ifp->if_flags |= IFF_UP; 3836 ifr.ifr_flags = ifp->if_flags & 0xffff; 3837 ifr.ifr_flagshigh = ifp->if_flags >> 16; 3838 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 3839 } 3840 } 3841 EVENTHANDLER_INVOKE(iflladdr_event, ifp); 3842 3843 return (0); 3844 } 3845 3846 /* 3847 * Compat function for handling basic encapsulation requests. 3848 * Not converted stacks (FDDI, IB, ..) supports traditional 3849 * output model: ARP (and other similar L2 protocols) are handled 3850 * inside output routine, arpresolve/nd6_resolve() returns MAC 3851 * address instead of full prepend. 3852 * 3853 * This function creates calculated header==MAC for IPv4/IPv6 and 3854 * returns EAFNOSUPPORT (which is then handled in ARP code) for other 3855 * address families. 3856 */ 3857 static int 3858 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req) 3859 { 3860 3861 if (req->rtype != IFENCAP_LL) 3862 return (EOPNOTSUPP); 3863 3864 if (req->bufsize < req->lladdr_len) 3865 return (ENOMEM); 3866 3867 switch (req->family) { 3868 case AF_INET: 3869 case AF_INET6: 3870 break; 3871 default: 3872 return (EAFNOSUPPORT); 3873 } 3874 3875 /* Copy lladdr to storage as is */ 3876 memmove(req->buf, req->lladdr, req->lladdr_len); 3877 req->bufsize = req->lladdr_len; 3878 req->lladdr_off = 0; 3879 3880 return (0); 3881 } 3882 3883 /* 3884 * Tunnel interfaces can nest, also they may cause infinite recursion 3885 * calls when misconfigured. We'll prevent this by detecting loops. 3886 * High nesting level may cause stack exhaustion. We'll prevent this 3887 * by introducing upper limit. 3888 * 3889 * Return 0, if tunnel nesting count is equal or less than limit. 3890 */ 3891 int 3892 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie, 3893 int limit) 3894 { 3895 struct m_tag *mtag; 3896 int count; 3897 3898 count = 1; 3899 mtag = NULL; 3900 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) { 3901 if (*(struct ifnet **)(mtag + 1) == ifp) { 3902 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp)); 3903 return (EIO); 3904 } 3905 count++; 3906 } 3907 if (count > limit) { 3908 log(LOG_NOTICE, 3909 "%s: if_output recursively called too many times(%d)\n", 3910 if_name(ifp), count); 3911 return (EIO); 3912 } 3913 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT); 3914 if (mtag == NULL) 3915 return (ENOMEM); 3916 *(struct ifnet **)(mtag + 1) = ifp; 3917 m_tag_prepend(m, mtag); 3918 return (0); 3919 } 3920 3921 /* 3922 * Get the link layer address that was read from the hardware at attach. 3923 * 3924 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type 3925 * their component interfaces as IFT_IEEE8023ADLAG. 3926 */ 3927 int 3928 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr) 3929 { 3930 3931 if (ifp->if_hw_addr == NULL) 3932 return (ENODEV); 3933 3934 switch (ifp->if_type) { 3935 case IFT_ETHER: 3936 case IFT_IEEE8023ADLAG: 3937 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen); 3938 return (0); 3939 default: 3940 return (ENODEV); 3941 } 3942 } 3943 3944 /* 3945 * The name argument must be a pointer to storage which will last as 3946 * long as the interface does. For physical devices, the result of 3947 * device_get_name(dev) is a good choice and for pseudo-devices a 3948 * static string works well. 3949 */ 3950 void 3951 if_initname(struct ifnet *ifp, const char *name, int unit) 3952 { 3953 ifp->if_dname = name; 3954 ifp->if_dunit = unit; 3955 if (unit != IF_DUNIT_NONE) 3956 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 3957 else 3958 strlcpy(ifp->if_xname, name, IFNAMSIZ); 3959 } 3960 3961 int 3962 if_printf(struct ifnet *ifp, const char *fmt, ...) 3963 { 3964 char if_fmt[256]; 3965 va_list ap; 3966 3967 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt); 3968 va_start(ap, fmt); 3969 vlog(LOG_INFO, if_fmt, ap); 3970 va_end(ap); 3971 return (0); 3972 } 3973 3974 void 3975 if_start(struct ifnet *ifp) 3976 { 3977 3978 (*(ifp)->if_start)(ifp); 3979 } 3980 3981 /* 3982 * Backwards compatibility interface for drivers 3983 * that have not implemented it 3984 */ 3985 static int 3986 if_transmit(struct ifnet *ifp, struct mbuf *m) 3987 { 3988 int error; 3989 3990 IFQ_HANDOFF(ifp, m, error); 3991 return (error); 3992 } 3993 3994 static void 3995 if_input_default(struct ifnet *ifp __unused, struct mbuf *m) 3996 { 3997 3998 m_freem(m); 3999 } 4000 4001 int 4002 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) 4003 { 4004 int active = 0; 4005 4006 IF_LOCK(ifq); 4007 if (_IF_QFULL(ifq)) { 4008 IF_UNLOCK(ifq); 4009 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); 4010 m_freem(m); 4011 return (0); 4012 } 4013 if (ifp != NULL) { 4014 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); 4015 if (m->m_flags & (M_BCAST|M_MCAST)) 4016 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); 4017 active = ifp->if_drv_flags & IFF_DRV_OACTIVE; 4018 } 4019 _IF_ENQUEUE(ifq, m); 4020 IF_UNLOCK(ifq); 4021 if (ifp != NULL && !active) 4022 (*(ifp)->if_start)(ifp); 4023 return (1); 4024 } 4025 4026 void 4027 if_register_com_alloc(u_char type, 4028 if_com_alloc_t *a, if_com_free_t *f) 4029 { 4030 4031 KASSERT(if_com_alloc[type] == NULL, 4032 ("if_register_com_alloc: %d already registered", type)); 4033 KASSERT(if_com_free[type] == NULL, 4034 ("if_register_com_alloc: %d free already registered", type)); 4035 4036 if_com_alloc[type] = a; 4037 if_com_free[type] = f; 4038 } 4039 4040 void 4041 if_deregister_com_alloc(u_char type) 4042 { 4043 4044 KASSERT(if_com_alloc[type] != NULL, 4045 ("if_deregister_com_alloc: %d not registered", type)); 4046 KASSERT(if_com_free[type] != NULL, 4047 ("if_deregister_com_alloc: %d free not registered", type)); 4048 if_com_alloc[type] = NULL; 4049 if_com_free[type] = NULL; 4050 } 4051 4052 /* API for driver access to network stack owned ifnet.*/ 4053 uint64_t 4054 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) 4055 { 4056 uint64_t oldbrate; 4057 4058 oldbrate = ifp->if_baudrate; 4059 ifp->if_baudrate = baudrate; 4060 return (oldbrate); 4061 } 4062 4063 uint64_t 4064 if_getbaudrate(if_t ifp) 4065 { 4066 4067 return (((struct ifnet *)ifp)->if_baudrate); 4068 } 4069 4070 int 4071 if_setcapabilities(if_t ifp, int capabilities) 4072 { 4073 ((struct ifnet *)ifp)->if_capabilities = capabilities; 4074 return (0); 4075 } 4076 4077 int 4078 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) 4079 { 4080 ((struct ifnet *)ifp)->if_capabilities |= setbit; 4081 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; 4082 4083 return (0); 4084 } 4085 4086 int 4087 if_getcapabilities(if_t ifp) 4088 { 4089 return ((struct ifnet *)ifp)->if_capabilities; 4090 } 4091 4092 int 4093 if_setcapenable(if_t ifp, int capabilities) 4094 { 4095 ((struct ifnet *)ifp)->if_capenable = capabilities; 4096 return (0); 4097 } 4098 4099 int 4100 if_setcapenablebit(if_t ifp, int setcap, int clearcap) 4101 { 4102 if(setcap) 4103 ((struct ifnet *)ifp)->if_capenable |= setcap; 4104 if(clearcap) 4105 ((struct ifnet *)ifp)->if_capenable &= ~clearcap; 4106 4107 return (0); 4108 } 4109 4110 const char * 4111 if_getdname(if_t ifp) 4112 { 4113 return ((struct ifnet *)ifp)->if_dname; 4114 } 4115 4116 int 4117 if_togglecapenable(if_t ifp, int togglecap) 4118 { 4119 ((struct ifnet *)ifp)->if_capenable ^= togglecap; 4120 return (0); 4121 } 4122 4123 int 4124 if_getcapenable(if_t ifp) 4125 { 4126 return ((struct ifnet *)ifp)->if_capenable; 4127 } 4128 4129 /* 4130 * This is largely undesirable because it ties ifnet to a device, but does 4131 * provide flexiblity for an embedded product vendor. Should be used with 4132 * the understanding that it violates the interface boundaries, and should be 4133 * a last resort only. 4134 */ 4135 int 4136 if_setdev(if_t ifp, void *dev) 4137 { 4138 return (0); 4139 } 4140 4141 int 4142 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) 4143 { 4144 ((struct ifnet *)ifp)->if_drv_flags |= set_flags; 4145 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; 4146 4147 return (0); 4148 } 4149 4150 int 4151 if_getdrvflags(if_t ifp) 4152 { 4153 return ((struct ifnet *)ifp)->if_drv_flags; 4154 } 4155 4156 int 4157 if_setdrvflags(if_t ifp, int flags) 4158 { 4159 ((struct ifnet *)ifp)->if_drv_flags = flags; 4160 return (0); 4161 } 4162 4163 int 4164 if_setflags(if_t ifp, int flags) 4165 { 4166 4167 ifp->if_flags = flags; 4168 return (0); 4169 } 4170 4171 int 4172 if_setflagbits(if_t ifp, int set, int clear) 4173 { 4174 ((struct ifnet *)ifp)->if_flags |= set; 4175 ((struct ifnet *)ifp)->if_flags &= ~clear; 4176 4177 return (0); 4178 } 4179 4180 int 4181 if_getflags(if_t ifp) 4182 { 4183 return ((struct ifnet *)ifp)->if_flags; 4184 } 4185 4186 int 4187 if_clearhwassist(if_t ifp) 4188 { 4189 ((struct ifnet *)ifp)->if_hwassist = 0; 4190 return (0); 4191 } 4192 4193 int 4194 if_sethwassistbits(if_t ifp, int toset, int toclear) 4195 { 4196 ((struct ifnet *)ifp)->if_hwassist |= toset; 4197 ((struct ifnet *)ifp)->if_hwassist &= ~toclear; 4198 4199 return (0); 4200 } 4201 4202 int 4203 if_sethwassist(if_t ifp, int hwassist_bit) 4204 { 4205 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; 4206 return (0); 4207 } 4208 4209 int 4210 if_gethwassist(if_t ifp) 4211 { 4212 return ((struct ifnet *)ifp)->if_hwassist; 4213 } 4214 4215 int 4216 if_setmtu(if_t ifp, int mtu) 4217 { 4218 ((struct ifnet *)ifp)->if_mtu = mtu; 4219 return (0); 4220 } 4221 4222 int 4223 if_getmtu(if_t ifp) 4224 { 4225 return ((struct ifnet *)ifp)->if_mtu; 4226 } 4227 4228 int 4229 if_getmtu_family(if_t ifp, int family) 4230 { 4231 struct domain *dp; 4232 4233 for (dp = domains; dp; dp = dp->dom_next) { 4234 if (dp->dom_family == family && dp->dom_ifmtu != NULL) 4235 return (dp->dom_ifmtu((struct ifnet *)ifp)); 4236 } 4237 4238 return (((struct ifnet *)ifp)->if_mtu); 4239 } 4240 4241 /* 4242 * Methods for drivers to access interface unicast and multicast 4243 * link level addresses. Driver shall not know 'struct ifaddr' neither 4244 * 'struct ifmultiaddr'. 4245 */ 4246 u_int 4247 if_lladdr_count(if_t ifp) 4248 { 4249 struct epoch_tracker et; 4250 struct ifaddr *ifa; 4251 u_int count; 4252 4253 count = 0; 4254 NET_EPOCH_ENTER(et); 4255 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 4256 if (ifa->ifa_addr->sa_family == AF_LINK) 4257 count++; 4258 NET_EPOCH_EXIT(et); 4259 4260 return (count); 4261 } 4262 4263 u_int 4264 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4265 { 4266 struct epoch_tracker et; 4267 struct ifaddr *ifa; 4268 u_int count; 4269 4270 MPASS(cb); 4271 4272 count = 0; 4273 NET_EPOCH_ENTER(et); 4274 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 4275 if (ifa->ifa_addr->sa_family != AF_LINK) 4276 continue; 4277 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr, 4278 count); 4279 } 4280 NET_EPOCH_EXIT(et); 4281 4282 return (count); 4283 } 4284 4285 u_int 4286 if_llmaddr_count(if_t ifp) 4287 { 4288 struct epoch_tracker et; 4289 struct ifmultiaddr *ifma; 4290 int count; 4291 4292 count = 0; 4293 NET_EPOCH_ENTER(et); 4294 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 4295 if (ifma->ifma_addr->sa_family == AF_LINK) 4296 count++; 4297 NET_EPOCH_EXIT(et); 4298 4299 return (count); 4300 } 4301 4302 u_int 4303 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg) 4304 { 4305 struct epoch_tracker et; 4306 struct ifmultiaddr *ifma; 4307 u_int count; 4308 4309 MPASS(cb); 4310 4311 count = 0; 4312 NET_EPOCH_ENTER(et); 4313 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 4314 if (ifma->ifma_addr->sa_family != AF_LINK) 4315 continue; 4316 count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr, 4317 count); 4318 } 4319 NET_EPOCH_EXIT(et); 4320 4321 return (count); 4322 } 4323 4324 int 4325 if_setsoftc(if_t ifp, void *softc) 4326 { 4327 ((struct ifnet *)ifp)->if_softc = softc; 4328 return (0); 4329 } 4330 4331 void * 4332 if_getsoftc(if_t ifp) 4333 { 4334 return ((struct ifnet *)ifp)->if_softc; 4335 } 4336 4337 void 4338 if_setrcvif(struct mbuf *m, if_t ifp) 4339 { 4340 4341 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 4342 m->m_pkthdr.rcvif = (struct ifnet *)ifp; 4343 } 4344 4345 void 4346 if_setvtag(struct mbuf *m, uint16_t tag) 4347 { 4348 m->m_pkthdr.ether_vtag = tag; 4349 } 4350 4351 uint16_t 4352 if_getvtag(struct mbuf *m) 4353 { 4354 4355 return (m->m_pkthdr.ether_vtag); 4356 } 4357 4358 int 4359 if_sendq_empty(if_t ifp) 4360 { 4361 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); 4362 } 4363 4364 struct ifaddr * 4365 if_getifaddr(if_t ifp) 4366 { 4367 return ((struct ifnet *)ifp)->if_addr; 4368 } 4369 4370 int 4371 if_getamcount(if_t ifp) 4372 { 4373 return ((struct ifnet *)ifp)->if_amcount; 4374 } 4375 4376 int 4377 if_setsendqready(if_t ifp) 4378 { 4379 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); 4380 return (0); 4381 } 4382 4383 int 4384 if_setsendqlen(if_t ifp, int tx_desc_count) 4385 { 4386 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); 4387 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; 4388 4389 return (0); 4390 } 4391 4392 int 4393 if_vlantrunkinuse(if_t ifp) 4394 { 4395 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; 4396 } 4397 4398 int 4399 if_input(if_t ifp, struct mbuf* sendmp) 4400 { 4401 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); 4402 return (0); 4403 4404 } 4405 4406 struct mbuf * 4407 if_dequeue(if_t ifp) 4408 { 4409 struct mbuf *m; 4410 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); 4411 4412 return (m); 4413 } 4414 4415 int 4416 if_sendq_prepend(if_t ifp, struct mbuf *m) 4417 { 4418 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); 4419 return (0); 4420 } 4421 4422 int 4423 if_setifheaderlen(if_t ifp, int len) 4424 { 4425 ((struct ifnet *)ifp)->if_hdrlen = len; 4426 return (0); 4427 } 4428 4429 caddr_t 4430 if_getlladdr(if_t ifp) 4431 { 4432 return (IF_LLADDR((struct ifnet *)ifp)); 4433 } 4434 4435 void * 4436 if_gethandle(u_char type) 4437 { 4438 return (if_alloc(type)); 4439 } 4440 4441 void 4442 if_bpfmtap(if_t ifh, struct mbuf *m) 4443 { 4444 struct ifnet *ifp = (struct ifnet *)ifh; 4445 4446 BPF_MTAP(ifp, m); 4447 } 4448 4449 void 4450 if_etherbpfmtap(if_t ifh, struct mbuf *m) 4451 { 4452 struct ifnet *ifp = (struct ifnet *)ifh; 4453 4454 ETHER_BPF_MTAP(ifp, m); 4455 } 4456 4457 void 4458 if_vlancap(if_t ifh) 4459 { 4460 struct ifnet *ifp = (struct ifnet *)ifh; 4461 VLAN_CAPABILITIES(ifp); 4462 } 4463 4464 int 4465 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax) 4466 { 4467 4468 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax; 4469 return (0); 4470 } 4471 4472 int 4473 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount) 4474 { 4475 4476 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount; 4477 return (0); 4478 } 4479 4480 int 4481 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize) 4482 { 4483 4484 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize; 4485 return (0); 4486 } 4487 4488 u_int 4489 if_gethwtsomax(if_t ifp) 4490 { 4491 4492 return (((struct ifnet *)ifp)->if_hw_tsomax); 4493 } 4494 4495 u_int 4496 if_gethwtsomaxsegcount(if_t ifp) 4497 { 4498 4499 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount); 4500 } 4501 4502 u_int 4503 if_gethwtsomaxsegsize(if_t ifp) 4504 { 4505 4506 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize); 4507 } 4508 4509 void 4510 if_setinitfn(if_t ifp, void (*init_fn)(void *)) 4511 { 4512 ((struct ifnet *)ifp)->if_init = init_fn; 4513 } 4514 4515 void 4516 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) 4517 { 4518 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; 4519 } 4520 4521 void 4522 if_setstartfn(if_t ifp, void (*start_fn)(if_t)) 4523 { 4524 ((struct ifnet *)ifp)->if_start = (void *)start_fn; 4525 } 4526 4527 void 4528 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) 4529 { 4530 ((struct ifnet *)ifp)->if_transmit = start_fn; 4531 } 4532 4533 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) 4534 { 4535 ((struct ifnet *)ifp)->if_qflush = flush_fn; 4536 4537 } 4538 4539 void 4540 if_setgetcounterfn(if_t ifp, if_get_counter_t fn) 4541 { 4542 4543 ifp->if_get_counter = fn; 4544 } 4545 4546 /* Revisit these - These are inline functions originally. */ 4547 int 4548 drbr_inuse_drv(if_t ifh, struct buf_ring *br) 4549 { 4550 return drbr_inuse(ifh, br); 4551 } 4552 4553 struct mbuf* 4554 drbr_dequeue_drv(if_t ifh, struct buf_ring *br) 4555 { 4556 return drbr_dequeue(ifh, br); 4557 } 4558 4559 int 4560 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) 4561 { 4562 return drbr_needs_enqueue(ifh, br); 4563 } 4564 4565 int 4566 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) 4567 { 4568 return drbr_enqueue(ifh, br, m); 4569 4570 } 4571