if_bridge.c revision 1.189.4.3
1 /* $NetBSD: if_bridge.c,v 1.189.4.3 2025/05/15 17:58:18 martin Exp $ */ 2 3 /* 4 * Copyright 2001 Wasabi Systems, Inc. 5 * All rights reserved. 6 * 7 * Written by Jason R. Thorpe for Wasabi Systems, Inc. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. All advertising materials mentioning features or use of this software 18 * must display the following acknowledgement: 19 * This product includes software developed for the NetBSD Project by 20 * Wasabi Systems, Inc. 21 * 4. The name of Wasabi Systems, Inc. may not be used to endorse 22 * or promote products derived from this software without specific prior 23 * written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 27 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 28 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC 29 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 35 * POSSIBILITY OF SUCH DAMAGE. 36 */ 37 38 /* 39 * Copyright (c) 1999, 2000 Jason L. Wright (jason (at) thought.net) 40 * All rights reserved. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. All advertising materials mentioning features or use of this software 51 * must display the following acknowledgement: 52 * This product includes software developed by Jason L. Wright 53 * 4. The name of the author may not be used to endorse or promote products 54 * derived from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 57 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 58 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 59 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, 60 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 61 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 62 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 64 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 65 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 * 68 * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp 69 */ 70 71 /* 72 * Network interface bridge support. 73 * 74 * TODO: 75 * 76 * - Currently only supports Ethernet-like interfaces (Ethernet, 77 * 802.11, VLANs on Ethernet, etc.) Figure out a nice way 78 * to bridge other types of interfaces (FDDI-FDDI, and maybe 79 * consider heterogenous bridges). 80 */ 81 82 #include <sys/cdefs.h> 83 __KERNEL_RCSID(0, "$NetBSD: if_bridge.c,v 1.189.4.3 2025/05/15 17:58:18 martin Exp $"); 84 85 #ifdef _KERNEL_OPT 86 #include "opt_inet.h" 87 #include "opt_net_mpsafe.h" 88 #endif /* _KERNEL_OPT */ 89 90 #include <sys/param.h> 91 #include <sys/kernel.h> 92 #include <sys/mbuf.h> 93 #include <sys/queue.h> 94 #include <sys/socket.h> 95 #include <sys/socketvar.h> /* for softnet_lock */ 96 #include <sys/sockio.h> 97 #include <sys/systm.h> 98 #include <sys/proc.h> 99 #include <sys/pool.h> 100 #include <sys/kauth.h> 101 #include <sys/cpu.h> 102 #include <sys/cprng.h> 103 #include <sys/mutex.h> 104 #include <sys/kmem.h> 105 #include <sys/syslog.h> 106 107 #include <net/bpf.h> 108 #include <net/if.h> 109 #include <net/if_dl.h> 110 #include <net/if_types.h> 111 #include <net/if_llc.h> 112 113 #include <net/if_ether.h> 114 #include <net/if_bridgevar.h> 115 #include <net/ether_sw_offload.h> 116 117 /* Used for bridge_ip[6]_checkbasic */ 118 #include <netinet/in.h> 119 #include <netinet/in_systm.h> 120 #include <netinet/ip.h> 121 #include <netinet/ip_var.h> 122 #include <netinet/ip_private.h> /* XXX */ 123 #include <netinet/ip6.h> 124 #include <netinet6/in6_var.h> 125 #include <netinet6/ip6_var.h> 126 #include <netinet6/ip6_private.h> /* XXX */ 127 128 /* 129 * Size of the route hash table. Must be a power of two. 130 */ 131 #ifndef BRIDGE_RTHASH_SIZE 132 #define BRIDGE_RTHASH_SIZE 1024 133 #endif 134 135 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1) 136 137 #include "carp.h" 138 #if NCARP > 0 139 #include <netinet/in.h> 140 #include <netinet/in_var.h> 141 #include <netinet/ip_carp.h> 142 #endif 143 144 #include "ioconf.h" 145 146 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf)); 147 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len)); 148 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf)); 149 150 /* 151 * Maximum number of addresses to cache. 152 */ 153 #ifndef BRIDGE_RTABLE_MAX 154 #define BRIDGE_RTABLE_MAX 100 155 #endif 156 157 /* 158 * Spanning tree defaults. 159 */ 160 #define BSTP_DEFAULT_MAX_AGE (20 * 256) 161 #define BSTP_DEFAULT_HELLO_TIME (2 * 256) 162 #define BSTP_DEFAULT_FORWARD_DELAY (15 * 256) 163 #define BSTP_DEFAULT_HOLD_TIME (1 * 256) 164 #define BSTP_DEFAULT_BRIDGE_PRIORITY 0x8000 165 #define BSTP_DEFAULT_PORT_PRIORITY 0x80 166 #define BSTP_DEFAULT_PATH_COST 55 167 168 /* 169 * Timeout (in seconds) for entries learned dynamically. 170 */ 171 #ifndef BRIDGE_RTABLE_TIMEOUT 172 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */ 173 #endif 174 175 /* 176 * Number of seconds between walks of the route list. 177 */ 178 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD 179 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60) 180 #endif 181 182 #define BRIDGE_RT_LOCK(_sc) mutex_enter((_sc)->sc_rtlist_lock) 183 #define BRIDGE_RT_UNLOCK(_sc) mutex_exit((_sc)->sc_rtlist_lock) 184 #define BRIDGE_RT_LOCKED(_sc) mutex_owned((_sc)->sc_rtlist_lock) 185 186 #define BRIDGE_RT_PSZ_PERFORM(_sc) \ 187 pserialize_perform((_sc)->sc_rtlist_psz) 188 189 #define BRIDGE_RT_RENTER(__s) do { __s = pserialize_read_enter(); } while (0) 190 #define BRIDGE_RT_REXIT(__s) do { pserialize_read_exit(__s); } while (0) 191 192 #define BRIDGE_RTLIST_READER_FOREACH(_brt, _sc) \ 193 PSLIST_READER_FOREACH((_brt), &((_sc)->sc_rtlist), \ 194 struct bridge_rtnode, brt_list) 195 #define BRIDGE_RTLIST_WRITER_FOREACH(_brt, _sc) \ 196 PSLIST_WRITER_FOREACH((_brt), &((_sc)->sc_rtlist), \ 197 struct bridge_rtnode, brt_list) 198 #define BRIDGE_RTLIST_WRITER_INSERT_HEAD(_sc, _brt) \ 199 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rtlist, brt, brt_list) 200 #define BRIDGE_RTLIST_WRITER_REMOVE(_brt) \ 201 PSLIST_WRITER_REMOVE((_brt), brt_list) 202 203 #define BRIDGE_RTHASH_READER_FOREACH(_brt, _sc, _hash) \ 204 PSLIST_READER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \ 205 struct bridge_rtnode, brt_hash) 206 #define BRIDGE_RTHASH_WRITER_FOREACH(_brt, _sc, _hash) \ 207 PSLIST_WRITER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \ 208 struct bridge_rtnode, brt_hash) 209 #define BRIDGE_RTHASH_WRITER_INSERT_HEAD(_sc, _hash, _brt) \ 210 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rthash[(_hash)], brt, brt_hash) 211 #define BRIDGE_RTHASH_WRITER_INSERT_AFTER(_brt, _new) \ 212 PSLIST_WRITER_INSERT_AFTER((_brt), (_new), brt_hash) 213 #define BRIDGE_RTHASH_WRITER_REMOVE(_brt) \ 214 PSLIST_WRITER_REMOVE((_brt), brt_hash) 215 216 #ifdef NET_MPSAFE 217 #define DECLARE_LOCK_VARIABLE 218 #define ACQUIRE_GLOBAL_LOCKS() do { } while (0) 219 #define RELEASE_GLOBAL_LOCKS() do { } while (0) 220 #else 221 #define DECLARE_LOCK_VARIABLE int __s 222 #define ACQUIRE_GLOBAL_LOCKS() do { \ 223 KERNEL_LOCK(1, NULL); \ 224 mutex_enter(softnet_lock); \ 225 __s = splsoftnet(); \ 226 } while (0) 227 #define RELEASE_GLOBAL_LOCKS() do { \ 228 splx(__s); \ 229 mutex_exit(softnet_lock); \ 230 KERNEL_UNLOCK_ONE(NULL); \ 231 } while (0) 232 #endif 233 234 struct psref_class *bridge_psref_class __read_mostly; 235 236 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD; 237 238 static struct pool bridge_rtnode_pool; 239 240 static int bridge_clone_create(struct if_clone *, int); 241 static int bridge_clone_destroy(struct ifnet *); 242 243 static int bridge_ioctl(struct ifnet *, u_long, void *); 244 static int bridge_init(struct ifnet *); 245 static void bridge_stop(struct ifnet *, int); 246 static void bridge_start(struct ifnet *); 247 static void bridge_ifdetach(void *); 248 249 static void bridge_input(struct ifnet *, struct mbuf *); 250 static void bridge_forward(struct bridge_softc *, struct mbuf *); 251 252 static void bridge_timer(void *); 253 254 static void bridge_broadcast(struct bridge_softc *, struct ifnet *, bool, 255 struct mbuf *); 256 257 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *, 258 struct ifnet *, int, uint8_t); 259 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *); 260 static void bridge_rttrim(struct bridge_softc *); 261 static void bridge_rtage(struct bridge_softc *); 262 static void bridge_rtage_work(struct work *, void *); 263 static void bridge_rtflush(struct bridge_softc *, int); 264 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *); 265 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp); 266 267 static void bridge_rtable_init(struct bridge_softc *); 268 static void bridge_rtable_fini(struct bridge_softc *); 269 270 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *, 271 const uint8_t *); 272 static int bridge_rtnode_insert(struct bridge_softc *, 273 struct bridge_rtnode *); 274 static void bridge_rtnode_remove(struct bridge_softc *, 275 struct bridge_rtnode *); 276 static void bridge_rtnode_destroy(struct bridge_rtnode *); 277 278 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *, 279 const char *name, 280 struct psref *); 281 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *, 282 struct ifnet *ifp, 283 struct psref *); 284 static void bridge_release_member(struct bridge_softc *, struct bridge_iflist *, 285 struct psref *); 286 static void bridge_delete_member(struct bridge_softc *, 287 struct bridge_iflist *); 288 static void bridge_acquire_member(struct bridge_softc *sc, 289 struct bridge_iflist *, 290 struct psref *); 291 292 static int bridge_ioctl_add(struct bridge_softc *, void *); 293 static int bridge_ioctl_del(struct bridge_softc *, void *); 294 static int bridge_ioctl_gifflags(struct bridge_softc *, void *); 295 static int bridge_ioctl_sifflags(struct bridge_softc *, void *); 296 static int bridge_ioctl_scache(struct bridge_softc *, void *); 297 static int bridge_ioctl_gcache(struct bridge_softc *, void *); 298 static int bridge_ioctl_gifs(struct bridge_softc *, void *); 299 static int bridge_ioctl_rts(struct bridge_softc *, void *); 300 static int bridge_ioctl_saddr(struct bridge_softc *, void *); 301 static int bridge_ioctl_sto(struct bridge_softc *, void *); 302 static int bridge_ioctl_gto(struct bridge_softc *, void *); 303 static int bridge_ioctl_daddr(struct bridge_softc *, void *); 304 static int bridge_ioctl_flush(struct bridge_softc *, void *); 305 static int bridge_ioctl_gpri(struct bridge_softc *, void *); 306 static int bridge_ioctl_spri(struct bridge_softc *, void *); 307 static int bridge_ioctl_ght(struct bridge_softc *, void *); 308 static int bridge_ioctl_sht(struct bridge_softc *, void *); 309 static int bridge_ioctl_gfd(struct bridge_softc *, void *); 310 static int bridge_ioctl_sfd(struct bridge_softc *, void *); 311 static int bridge_ioctl_gma(struct bridge_softc *, void *); 312 static int bridge_ioctl_sma(struct bridge_softc *, void *); 313 static int bridge_ioctl_sifprio(struct bridge_softc *, void *); 314 static int bridge_ioctl_sifcost(struct bridge_softc *, void *); 315 static int bridge_ioctl_gfilt(struct bridge_softc *, void *); 316 static int bridge_ioctl_sfilt(struct bridge_softc *, void *); 317 static int bridge_ipf(void *, struct mbuf **, struct ifnet *, int); 318 static int bridge_ip_checkbasic(struct mbuf **mp); 319 # ifdef INET6 320 static int bridge_ip6_checkbasic(struct mbuf **mp); 321 # endif /* INET6 */ 322 323 struct bridge_control { 324 int (*bc_func)(struct bridge_softc *, void *); 325 int bc_argsize; 326 int bc_flags; 327 }; 328 329 #define BC_F_COPYIN 0x01 /* copy arguments in */ 330 #define BC_F_COPYOUT 0x02 /* copy arguments out */ 331 #define BC_F_SUSER 0x04 /* do super-user check */ 332 #define BC_F_XLATEIN 0x08 /* xlate arguments in */ 333 #define BC_F_XLATEOUT 0x10 /* xlate arguments out */ 334 335 static const struct bridge_control bridge_control_table[] = { 336 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 337 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 338 339 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT}, 340 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 341 342 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 343 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT}, 344 345 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT}, 346 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT}, 347 348 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER}, 349 350 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 351 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT}, 352 353 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER}, 354 355 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 356 357 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT}, 358 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 359 360 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT}, 361 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 362 363 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT}, 364 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 365 366 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT}, 367 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 368 369 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 370 371 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 372 373 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT}, 374 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 375 376 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT}, 377 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT}, 378 }; 379 380 static const int bridge_control_table_size = __arraycount(bridge_control_table); 381 382 static struct if_clone bridge_cloner = 383 IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy); 384 385 /* 386 * bridgeattach: 387 * 388 * Pseudo-device attach routine. 389 */ 390 void 391 bridgeattach(int n) 392 { 393 394 pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode), 395 0, 0, 0, "brtpl", NULL, IPL_NET); 396 397 bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET); 398 399 if_clone_attach(&bridge_cloner); 400 } 401 402 /* 403 * bridge_clone_create: 404 * 405 * Create a new bridge instance. 406 */ 407 static int 408 bridge_clone_create(struct if_clone *ifc, int unit) 409 { 410 struct bridge_softc *sc; 411 struct ifnet *ifp; 412 int error; 413 414 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP); 415 ifp = &sc->sc_if; 416 417 sc->sc_brtmax = BRIDGE_RTABLE_MAX; 418 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT; 419 sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE; 420 sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME; 421 sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY; 422 sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY; 423 sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME; 424 sc->sc_filter_flags = 0; 425 426 /* Initialize our routing table. */ 427 bridge_rtable_init(sc); 428 429 error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage", 430 bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE); 431 if (error) 432 panic("%s: workqueue_create %d\n", __func__, error); 433 434 callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE); 435 callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE); 436 437 mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE); 438 PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist); 439 sc->sc_iflist_psref.bip_psz = pserialize_create(); 440 441 if_initname(ifp, ifc->ifc_name, unit); 442 ifp->if_softc = sc; 443 #ifdef NET_MPSAFE 444 ifp->if_extflags = IFEF_MPSAFE; 445 #endif 446 ifp->if_mtu = ETHERMTU; 447 ifp->if_ioctl = bridge_ioctl; 448 ifp->if_output = bridge_output; 449 ifp->if_start = bridge_start; 450 ifp->if_stop = bridge_stop; 451 ifp->if_init = bridge_init; 452 ifp->if_type = IFT_BRIDGE; 453 ifp->if_addrlen = 0; 454 ifp->if_dlt = DLT_EN10MB; 455 ifp->if_hdrlen = ETHER_HDR_LEN; 456 if_initialize(ifp); 457 458 /* 459 * Set the link state to down. 460 * When interfaces are added the link state will reflect 461 * the best link state of the combined interfaces. 462 */ 463 ifp->if_link_state = LINK_STATE_DOWN; 464 465 if_alloc_sadl(ifp); 466 if_register(ifp); 467 468 return 0; 469 } 470 471 /* 472 * bridge_clone_destroy: 473 * 474 * Destroy a bridge instance. 475 */ 476 static int 477 bridge_clone_destroy(struct ifnet *ifp) 478 { 479 struct bridge_softc *sc = ifp->if_softc; 480 struct bridge_iflist *bif; 481 482 if ((ifp->if_flags & IFF_RUNNING) != 0) 483 bridge_stop(ifp, 1); 484 485 BRIDGE_LOCK(sc); 486 for (;;) { 487 bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist, 488 bif_next); 489 if (bif == NULL) 490 break; 491 bridge_delete_member(sc, bif); 492 } 493 PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist); 494 BRIDGE_UNLOCK(sc); 495 496 if_detach(ifp); 497 498 /* Tear down the routing table. */ 499 bridge_rtable_fini(sc); 500 501 pserialize_destroy(sc->sc_iflist_psref.bip_psz); 502 mutex_destroy(&sc->sc_iflist_psref.bip_lock); 503 callout_destroy(&sc->sc_brcallout); 504 callout_destroy(&sc->sc_bstpcallout); 505 workqueue_destroy(sc->sc_rtage_wq); 506 kmem_free(sc, sizeof(*sc)); 507 508 return 0; 509 } 510 511 /* 512 * bridge_ioctl: 513 * 514 * Handle a control request from the operator. 515 */ 516 static int 517 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data) 518 { 519 struct bridge_softc *sc = ifp->if_softc; 520 struct lwp *l = curlwp; /* XXX */ 521 union { 522 struct ifbreq ifbreq; 523 struct ifbifconf ifbifconf; 524 struct ifbareq ifbareq; 525 struct ifbaconf ifbaconf; 526 struct ifbrparam ifbrparam; 527 } args; 528 struct ifdrv *ifd = (struct ifdrv *) data; 529 const struct bridge_control *bc = NULL; /* XXXGCC */ 530 int error = 0; 531 532 /* Authorize command before calling splsoftnet(). */ 533 switch (cmd) { 534 case SIOCGDRVSPEC: 535 case SIOCSDRVSPEC: 536 if (ifd->ifd_cmd >= bridge_control_table_size 537 || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) { 538 error = EINVAL; 539 return error; 540 } 541 542 /* We only care about BC_F_SUSER at this point. */ 543 if ((bc->bc_flags & BC_F_SUSER) == 0) 544 break; 545 546 error = kauth_authorize_network(l->l_cred, 547 KAUTH_NETWORK_INTERFACE_BRIDGE, 548 cmd == SIOCGDRVSPEC ? 549 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV : 550 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV, 551 ifd, NULL, NULL); 552 if (error) 553 return error; 554 555 break; 556 } 557 558 const int s = splsoftnet(); 559 560 switch (cmd) { 561 case SIOCGDRVSPEC: 562 case SIOCSDRVSPEC: 563 KASSERT(bc != NULL); 564 if (cmd == SIOCGDRVSPEC && 565 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) { 566 error = EINVAL; 567 break; 568 } 569 else if (cmd == SIOCSDRVSPEC && 570 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) { 571 error = EINVAL; 572 break; 573 } 574 575 /* BC_F_SUSER is checked above, before splsoftnet(). */ 576 577 if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0 578 && (ifd->ifd_len != bc->bc_argsize 579 || ifd->ifd_len > sizeof(args))) { 580 error = EINVAL; 581 break; 582 } 583 584 memset(&args, 0, sizeof(args)); 585 if (bc->bc_flags & BC_F_COPYIN) { 586 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); 587 if (error) 588 break; 589 } else if (bc->bc_flags & BC_F_XLATEIN) { 590 args.ifbifconf.ifbic_len = ifd->ifd_len; 591 args.ifbifconf.ifbic_buf = ifd->ifd_data; 592 } 593 594 error = (*bc->bc_func)(sc, &args); 595 if (error) 596 break; 597 598 if (bc->bc_flags & BC_F_COPYOUT) { 599 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); 600 } else if (bc->bc_flags & BC_F_XLATEOUT) { 601 ifd->ifd_len = args.ifbifconf.ifbic_len; 602 ifd->ifd_data = args.ifbifconf.ifbic_buf; 603 } 604 break; 605 606 case SIOCSIFFLAGS: 607 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 608 break; 609 switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) { 610 case IFF_RUNNING: 611 /* 612 * If interface is marked down and it is running, 613 * then stop and disable it. 614 */ 615 if_stop(ifp, 1); 616 break; 617 case IFF_UP: 618 /* 619 * If interface is marked up and it is stopped, then 620 * start it. 621 */ 622 error = if_init(ifp); 623 break; 624 default: 625 break; 626 } 627 break; 628 629 case SIOCSIFMTU: 630 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET) 631 error = 0; 632 break; 633 634 case SIOCGIFCAP: 635 { 636 struct ifcapreq *ifcr = (struct ifcapreq *)data; 637 ifcr->ifcr_capabilities = sc->sc_capenable; 638 ifcr->ifcr_capenable = sc->sc_capenable; 639 break; 640 } 641 642 default: 643 error = ifioctl_common(ifp, cmd, data); 644 break; 645 } 646 647 splx(s); 648 649 return error; 650 } 651 652 /* 653 * bridge_lookup_member: 654 * 655 * Lookup a bridge member interface. 656 */ 657 static struct bridge_iflist * 658 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref) 659 { 660 struct bridge_iflist *bif; 661 struct ifnet *ifp; 662 int s; 663 664 BRIDGE_PSZ_RENTER(s); 665 666 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 667 ifp = bif->bif_ifp; 668 if (strcmp(ifp->if_xname, name) == 0) 669 break; 670 } 671 if (bif != NULL) 672 bridge_acquire_member(sc, bif, psref); 673 674 BRIDGE_PSZ_REXIT(s); 675 676 return bif; 677 } 678 679 /* 680 * bridge_lookup_member_if: 681 * 682 * Lookup a bridge member interface by ifnet*. 683 */ 684 static struct bridge_iflist * 685 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp, 686 struct psref *psref) 687 { 688 struct bridge_iflist *bif; 689 int s; 690 691 BRIDGE_PSZ_RENTER(s); 692 693 bif = member_ifp->if_bridgeif; 694 if (bif != NULL) { 695 psref_acquire(psref, &bif->bif_psref, 696 bridge_psref_class); 697 } 698 699 BRIDGE_PSZ_REXIT(s); 700 701 return bif; 702 } 703 704 static void 705 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif, 706 struct psref *psref) 707 { 708 709 psref_acquire(psref, &bif->bif_psref, bridge_psref_class); 710 } 711 712 /* 713 * bridge_release_member: 714 * 715 * Release the specified member interface. 716 */ 717 static void 718 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif, 719 struct psref *psref) 720 { 721 722 psref_release(psref, &bif->bif_psref, bridge_psref_class); 723 } 724 725 /* 726 * bridge_delete_member: 727 * 728 * Delete the specified member interface. 729 */ 730 static void 731 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif) 732 { 733 struct ifnet *ifs = bif->bif_ifp; 734 735 KASSERT(BRIDGE_LOCKED(sc)); 736 737 ifs->_if_input = ether_input; 738 ifs->if_bridge = NULL; 739 ifs->if_bridgeif = NULL; 740 741 PSLIST_WRITER_REMOVE(bif, bif_next); 742 BRIDGE_PSZ_PERFORM(sc); 743 744 if_linkstate_change_disestablish(ifs, 745 bif->bif_linkstate_hook, BRIDGE_LOCK_OBJ(sc)); 746 ether_ifdetachhook_disestablish(ifs, 747 bif->bif_ifdetach_hook, BRIDGE_LOCK_OBJ(sc)); 748 749 BRIDGE_UNLOCK(sc); 750 751 switch (ifs->if_type) { 752 case IFT_ETHER: 753 case IFT_L2TP: 754 /* 755 * Take the interface out of promiscuous mode. 756 * Don't call it with holding a spin lock. 757 */ 758 (void) ifpromisc(ifs, 0); 759 IFNET_LOCK(ifs); 760 (void) ether_disable_vlan_mtu(ifs); 761 IFNET_UNLOCK(ifs); 762 break; 763 default: 764 #ifdef DIAGNOSTIC 765 panic("%s: impossible", __func__); 766 #endif 767 break; 768 } 769 770 psref_target_destroy(&bif->bif_psref, bridge_psref_class); 771 772 PSLIST_ENTRY_DESTROY(bif, bif_next); 773 kmem_free(bif, sizeof(*bif)); 774 775 BRIDGE_LOCK(sc); 776 } 777 778 /* 779 * bridge_calc_csum_flags: 780 * 781 * Calculate logical and b/w csum flags each member interface supports. 782 */ 783 void 784 bridge_calc_csum_flags(struct bridge_softc *sc) 785 { 786 struct bridge_iflist *bif; 787 struct ifnet *ifs = NULL; 788 int flags = ~0; 789 int capenable = ~0; 790 791 BRIDGE_LOCK(sc); 792 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 793 ifs = bif->bif_ifp; 794 flags &= ifs->if_csum_flags_tx; 795 capenable &= ifs->if_capenable; 796 } 797 sc->sc_csum_flags_tx = flags; 798 sc->sc_capenable = (ifs != NULL) ? capenable : 0; 799 BRIDGE_UNLOCK(sc); 800 } 801 802 /* 803 * bridge_calc_link_state: 804 * 805 * Calculate the link state based on each member interface. 806 */ 807 static void 808 bridge_calc_link_state(void *xsc) 809 { 810 struct bridge_softc *sc = xsc; 811 struct bridge_iflist *bif; 812 struct ifnet *ifs; 813 int link_state = LINK_STATE_DOWN; 814 815 BRIDGE_LOCK(sc); 816 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 817 ifs = bif->bif_ifp; 818 if (ifs->if_link_state == LINK_STATE_UP) { 819 link_state = LINK_STATE_UP; 820 break; 821 } 822 if (ifs->if_link_state == LINK_STATE_UNKNOWN) 823 link_state = LINK_STATE_UNKNOWN; 824 } 825 if_link_state_change(&sc->sc_if, link_state); 826 BRIDGE_UNLOCK(sc); 827 } 828 829 static int 830 bridge_ioctl_add(struct bridge_softc *sc, void *arg) 831 { 832 struct ifbreq *req = arg; 833 struct bridge_iflist *bif = NULL; 834 struct ifnet *ifs; 835 int error = 0; 836 struct psref psref; 837 838 ifs = if_get(req->ifbr_ifsname, &psref); 839 if (ifs == NULL) 840 return ENOENT; 841 842 if (ifs->if_bridge == sc) { 843 error = EEXIST; 844 goto out; 845 } 846 847 if (ifs->if_bridge != NULL) { 848 error = EBUSY; 849 goto out; 850 } 851 852 if (ifs->_if_input != ether_input) { 853 error = EINVAL; 854 goto out; 855 } 856 857 /* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */ 858 if ((ifs->if_flags & IFF_SIMPLEX) == 0) { 859 error = EINVAL; 860 goto out; 861 } 862 863 bif = kmem_alloc(sizeof(*bif), KM_SLEEP); 864 865 switch (ifs->if_type) { 866 case IFT_ETHER: 867 if (sc->sc_if.if_mtu != ifs->if_mtu) { 868 /* Change MTU of added interface to bridge MTU */ 869 struct ifreq ifr; 870 memset(&ifr, 0, sizeof(ifr)); 871 ifr.ifr_mtu = sc->sc_if.if_mtu; 872 IFNET_LOCK(ifs); 873 error = if_ioctl(ifs, SIOCSIFMTU, &ifr); 874 IFNET_UNLOCK(ifs); 875 if (error != 0) 876 goto out; 877 } 878 /* FALLTHROUGH */ 879 case IFT_L2TP: 880 IFNET_LOCK(ifs); 881 error = ether_enable_vlan_mtu(ifs); 882 IFNET_UNLOCK(ifs); 883 if (error > 0) 884 goto out; 885 /* 886 * Place the interface into promiscuous mode. 887 */ 888 error = ifpromisc(ifs, 1); 889 if (error) 890 goto out; 891 break; 892 default: 893 error = EINVAL; 894 goto out; 895 } 896 897 bif->bif_ifp = ifs; 898 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER; 899 bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY; 900 bif->bif_path_cost = BSTP_DEFAULT_PATH_COST; 901 bif->bif_linkstate_hook = if_linkstate_change_establish(ifs, 902 bridge_calc_link_state, sc); 903 PSLIST_ENTRY_INIT(bif, bif_next); 904 psref_target_init(&bif->bif_psref, bridge_psref_class); 905 906 BRIDGE_LOCK(sc); 907 908 ifs->if_bridge = sc; 909 ifs->if_bridgeif = bif; 910 PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next); 911 ifs->_if_input = bridge_input; 912 913 BRIDGE_UNLOCK(sc); 914 915 bif->bif_ifdetach_hook = ether_ifdetachhook_establish(ifs, 916 bridge_ifdetach, (void *)ifs); 917 918 bridge_calc_csum_flags(sc); 919 bridge_calc_link_state(sc); 920 921 if (sc->sc_if.if_flags & IFF_RUNNING) 922 bstp_initialization(sc); 923 else 924 bstp_stop(sc); 925 926 out: 927 if_put(ifs, &psref); 928 if (error) { 929 if (bif != NULL) 930 kmem_free(bif, sizeof(*bif)); 931 } 932 return error; 933 } 934 935 static int 936 bridge_ioctl_del(struct bridge_softc *sc, void *arg) 937 { 938 struct ifbreq *req = arg; 939 const char *name = req->ifbr_ifsname; 940 struct bridge_iflist *bif; 941 struct ifnet *ifs; 942 943 BRIDGE_LOCK(sc); 944 945 /* 946 * Don't use bridge_lookup_member. We want to get a member 947 * with bif_refs == 0. 948 */ 949 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) { 950 ifs = bif->bif_ifp; 951 if (strcmp(ifs->if_xname, name) == 0) 952 break; 953 } 954 955 if (bif == NULL) { 956 BRIDGE_UNLOCK(sc); 957 return ENOENT; 958 } 959 960 bridge_delete_member(sc, bif); 961 962 BRIDGE_UNLOCK(sc); 963 964 bridge_rtdelete(sc, ifs); 965 bridge_calc_csum_flags(sc); 966 bridge_calc_link_state(sc); 967 968 if (sc->sc_if.if_flags & IFF_RUNNING) 969 bstp_initialization(sc); 970 971 return 0; 972 } 973 974 static int 975 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg) 976 { 977 struct ifbreq *req = arg; 978 struct bridge_iflist *bif; 979 struct psref psref; 980 981 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 982 if (bif == NULL) 983 return ENOENT; 984 985 req->ifbr_ifsflags = bif->bif_flags; 986 req->ifbr_state = bif->bif_state; 987 req->ifbr_priority = bif->bif_priority; 988 req->ifbr_path_cost = bif->bif_path_cost; 989 req->ifbr_portno = bif->bif_ifp->if_index & 0xff; 990 991 bridge_release_member(sc, bif, &psref); 992 993 return 0; 994 } 995 996 static int 997 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg) 998 { 999 struct ifbreq *req = arg; 1000 struct bridge_iflist *bif; 1001 struct psref psref; 1002 1003 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1004 if (bif == NULL) 1005 return ENOENT; 1006 1007 if (req->ifbr_ifsflags & IFBIF_STP) { 1008 switch (bif->bif_ifp->if_type) { 1009 case IFT_ETHER: 1010 case IFT_L2TP: 1011 /* These can do spanning tree. */ 1012 break; 1013 1014 default: 1015 /* Nothing else can. */ 1016 bridge_release_member(sc, bif, &psref); 1017 return EINVAL; 1018 } 1019 } 1020 1021 if (bif->bif_flags & IFBIF_PROTECTED) { 1022 if ((req->ifbr_ifsflags & IFBIF_PROTECTED) == 0) { 1023 log(LOG_INFO, "%s: disabling protection on %s\n", 1024 sc->sc_if.if_xname, bif->bif_ifp->if_xname); 1025 } 1026 } else { 1027 if (req->ifbr_ifsflags & IFBIF_PROTECTED) { 1028 log(LOG_INFO, "%s: enabling protection on %s\n", 1029 sc->sc_if.if_xname, bif->bif_ifp->if_xname); 1030 } 1031 } 1032 1033 bif->bif_flags = req->ifbr_ifsflags; 1034 1035 bridge_release_member(sc, bif, &psref); 1036 1037 if (sc->sc_if.if_flags & IFF_RUNNING) 1038 bstp_initialization(sc); 1039 1040 return 0; 1041 } 1042 1043 static int 1044 bridge_ioctl_scache(struct bridge_softc *sc, void *arg) 1045 { 1046 struct ifbrparam *param = arg; 1047 1048 sc->sc_brtmax = param->ifbrp_csize; 1049 bridge_rttrim(sc); 1050 1051 return 0; 1052 } 1053 1054 static int 1055 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg) 1056 { 1057 struct ifbrparam *param = arg; 1058 1059 param->ifbrp_csize = sc->sc_brtmax; 1060 1061 return 0; 1062 } 1063 1064 static int 1065 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg) 1066 { 1067 struct ifbifconf *bifc = arg; 1068 struct bridge_iflist *bif; 1069 struct ifbreq *breqs; 1070 int i, count, error = 0; 1071 1072 retry: 1073 BRIDGE_LOCK(sc); 1074 count = 0; 1075 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) 1076 count++; 1077 BRIDGE_UNLOCK(sc); 1078 1079 if (count == 0) { 1080 bifc->ifbic_len = 0; 1081 return 0; 1082 } 1083 1084 if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) { 1085 /* Tell that a larger buffer is needed */ 1086 bifc->ifbic_len = sizeof(*breqs) * count; 1087 return 0; 1088 } 1089 1090 breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP); 1091 1092 BRIDGE_LOCK(sc); 1093 1094 i = 0; 1095 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) 1096 i++; 1097 if (i > count) { 1098 /* 1099 * The number of members has been increased. 1100 * We need more memory! 1101 */ 1102 BRIDGE_UNLOCK(sc); 1103 kmem_free(breqs, sizeof(*breqs) * count); 1104 goto retry; 1105 } 1106 1107 i = 0; 1108 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) { 1109 struct ifbreq *breq = &breqs[i++]; 1110 memset(breq, 0, sizeof(*breq)); 1111 1112 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname, 1113 sizeof(breq->ifbr_ifsname)); 1114 breq->ifbr_ifsflags = bif->bif_flags; 1115 breq->ifbr_state = bif->bif_state; 1116 breq->ifbr_priority = bif->bif_priority; 1117 breq->ifbr_path_cost = bif->bif_path_cost; 1118 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff; 1119 } 1120 1121 /* Don't call copyout with holding the mutex */ 1122 BRIDGE_UNLOCK(sc); 1123 1124 for (i = 0; i < count; i++) { 1125 error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs)); 1126 if (error) 1127 break; 1128 } 1129 bifc->ifbic_len = sizeof(*breqs) * i; 1130 1131 kmem_free(breqs, sizeof(*breqs) * count); 1132 1133 return error; 1134 } 1135 1136 static int 1137 bridge_ioctl_rts(struct bridge_softc *sc, void *arg) 1138 { 1139 struct ifbaconf *bac = arg; 1140 struct bridge_rtnode *brt; 1141 struct ifbareq bareq; 1142 int count = 0, error = 0, len; 1143 1144 if (bac->ifbac_len == 0) 1145 return 0; 1146 1147 BRIDGE_RT_LOCK(sc); 1148 1149 /* The passed buffer is not enough, tell a required size. */ 1150 if (bac->ifbac_len < (sizeof(bareq) * sc->sc_brtcnt)) { 1151 count = sc->sc_brtcnt; 1152 goto out; 1153 } 1154 1155 len = bac->ifbac_len; 1156 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 1157 if (len < sizeof(bareq)) 1158 goto out; 1159 memset(&bareq, 0, sizeof(bareq)); 1160 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname, 1161 sizeof(bareq.ifba_ifsname)); 1162 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr)); 1163 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 1164 bareq.ifba_expire = brt->brt_expire - time_uptime; 1165 } else 1166 bareq.ifba_expire = 0; 1167 bareq.ifba_flags = brt->brt_flags; 1168 1169 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq)); 1170 if (error) 1171 goto out; 1172 count++; 1173 len -= sizeof(bareq); 1174 } 1175 out: 1176 BRIDGE_RT_UNLOCK(sc); 1177 1178 bac->ifbac_len = sizeof(bareq) * count; 1179 return error; 1180 } 1181 1182 static int 1183 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg) 1184 { 1185 struct ifbareq *req = arg; 1186 struct bridge_iflist *bif; 1187 int error; 1188 struct psref psref; 1189 1190 bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref); 1191 if (bif == NULL) 1192 return ENOENT; 1193 1194 error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1, 1195 req->ifba_flags); 1196 1197 bridge_release_member(sc, bif, &psref); 1198 1199 return error; 1200 } 1201 1202 static int 1203 bridge_ioctl_sto(struct bridge_softc *sc, void *arg) 1204 { 1205 struct ifbrparam *param = arg; 1206 1207 sc->sc_brttimeout = param->ifbrp_ctime; 1208 1209 return 0; 1210 } 1211 1212 static int 1213 bridge_ioctl_gto(struct bridge_softc *sc, void *arg) 1214 { 1215 struct ifbrparam *param = arg; 1216 1217 param->ifbrp_ctime = sc->sc_brttimeout; 1218 1219 return 0; 1220 } 1221 1222 static int 1223 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg) 1224 { 1225 struct ifbareq *req = arg; 1226 1227 return (bridge_rtdaddr(sc, req->ifba_dst)); 1228 } 1229 1230 static int 1231 bridge_ioctl_flush(struct bridge_softc *sc, void *arg) 1232 { 1233 struct ifbreq *req = arg; 1234 1235 bridge_rtflush(sc, req->ifbr_ifsflags); 1236 1237 return 0; 1238 } 1239 1240 static int 1241 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg) 1242 { 1243 struct ifbrparam *param = arg; 1244 1245 param->ifbrp_prio = sc->sc_bridge_priority; 1246 1247 return 0; 1248 } 1249 1250 static int 1251 bridge_ioctl_spri(struct bridge_softc *sc, void *arg) 1252 { 1253 struct ifbrparam *param = arg; 1254 1255 sc->sc_bridge_priority = param->ifbrp_prio; 1256 1257 if (sc->sc_if.if_flags & IFF_RUNNING) 1258 bstp_initialization(sc); 1259 1260 return 0; 1261 } 1262 1263 static int 1264 bridge_ioctl_ght(struct bridge_softc *sc, void *arg) 1265 { 1266 struct ifbrparam *param = arg; 1267 1268 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8; 1269 1270 return 0; 1271 } 1272 1273 static int 1274 bridge_ioctl_sht(struct bridge_softc *sc, void *arg) 1275 { 1276 struct ifbrparam *param = arg; 1277 1278 if (param->ifbrp_hellotime == 0) 1279 return EINVAL; 1280 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8; 1281 1282 if (sc->sc_if.if_flags & IFF_RUNNING) 1283 bstp_initialization(sc); 1284 1285 return 0; 1286 } 1287 1288 static int 1289 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg) 1290 { 1291 struct ifbrparam *param = arg; 1292 1293 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8; 1294 1295 return 0; 1296 } 1297 1298 static int 1299 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg) 1300 { 1301 struct ifbrparam *param = arg; 1302 1303 if (param->ifbrp_fwddelay == 0) 1304 return EINVAL; 1305 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8; 1306 1307 if (sc->sc_if.if_flags & IFF_RUNNING) 1308 bstp_initialization(sc); 1309 1310 return 0; 1311 } 1312 1313 static int 1314 bridge_ioctl_gma(struct bridge_softc *sc, void *arg) 1315 { 1316 struct ifbrparam *param = arg; 1317 1318 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8; 1319 1320 return 0; 1321 } 1322 1323 static int 1324 bridge_ioctl_sma(struct bridge_softc *sc, void *arg) 1325 { 1326 struct ifbrparam *param = arg; 1327 1328 if (param->ifbrp_maxage == 0) 1329 return EINVAL; 1330 sc->sc_bridge_max_age = param->ifbrp_maxage << 8; 1331 1332 if (sc->sc_if.if_flags & IFF_RUNNING) 1333 bstp_initialization(sc); 1334 1335 return 0; 1336 } 1337 1338 static int 1339 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg) 1340 { 1341 struct ifbreq *req = arg; 1342 struct bridge_iflist *bif; 1343 struct psref psref; 1344 1345 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1346 if (bif == NULL) 1347 return ENOENT; 1348 1349 bif->bif_priority = req->ifbr_priority; 1350 1351 if (sc->sc_if.if_flags & IFF_RUNNING) 1352 bstp_initialization(sc); 1353 1354 bridge_release_member(sc, bif, &psref); 1355 1356 return 0; 1357 } 1358 1359 static int 1360 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg) 1361 { 1362 struct ifbrparam *param = arg; 1363 1364 param->ifbrp_filter = sc->sc_filter_flags; 1365 1366 return 0; 1367 } 1368 1369 static int 1370 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg) 1371 { 1372 struct ifbrparam *param = arg; 1373 uint32_t nflags, oflags; 1374 1375 if (param->ifbrp_filter & ~IFBF_FILT_MASK) 1376 return EINVAL; 1377 1378 nflags = param->ifbrp_filter; 1379 oflags = sc->sc_filter_flags; 1380 1381 if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) { 1382 pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT, 1383 sc->sc_if.if_pfil); 1384 } 1385 if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) { 1386 pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT, 1387 sc->sc_if.if_pfil); 1388 } 1389 1390 sc->sc_filter_flags = nflags; 1391 1392 return 0; 1393 } 1394 1395 static int 1396 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg) 1397 { 1398 struct ifbreq *req = arg; 1399 struct bridge_iflist *bif; 1400 struct psref psref; 1401 1402 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1403 if (bif == NULL) 1404 return ENOENT; 1405 1406 bif->bif_path_cost = req->ifbr_path_cost; 1407 1408 if (sc->sc_if.if_flags & IFF_RUNNING) 1409 bstp_initialization(sc); 1410 1411 bridge_release_member(sc, bif, &psref); 1412 1413 return 0; 1414 } 1415 1416 /* 1417 * bridge_ifdetach: 1418 * 1419 * Detach an interface from a bridge. Called when a member 1420 * interface is detaching. 1421 */ 1422 static void 1423 bridge_ifdetach(void *xifs) 1424 { 1425 struct ifnet *ifs; 1426 struct bridge_softc *sc; 1427 struct ifbreq breq; 1428 1429 ifs = (struct ifnet *)xifs; 1430 sc = ifs->if_bridge; 1431 1432 /* ioctl_lock should prevent this from happening */ 1433 KASSERT(sc != NULL); 1434 1435 memset(&breq, 0, sizeof(breq)); 1436 strlcpy(breq.ifbr_ifsname, ifs->if_xname, sizeof(breq.ifbr_ifsname)); 1437 1438 (void) bridge_ioctl_del(sc, &breq); 1439 } 1440 1441 /* 1442 * bridge_init: 1443 * 1444 * Initialize a bridge interface. 1445 */ 1446 static int 1447 bridge_init(struct ifnet *ifp) 1448 { 1449 struct bridge_softc *sc = ifp->if_softc; 1450 1451 KASSERT((ifp->if_flags & IFF_RUNNING) == 0); 1452 1453 BRIDGE_LOCK(sc); 1454 sc->sc_stopping = false; 1455 BRIDGE_UNLOCK(sc); 1456 1457 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz, 1458 bridge_timer, sc); 1459 bstp_initialization(sc); 1460 1461 ifp->if_flags |= IFF_RUNNING; 1462 return 0; 1463 } 1464 1465 /* 1466 * bridge_stop: 1467 * 1468 * Stop the bridge interface. 1469 */ 1470 static void 1471 bridge_stop(struct ifnet *ifp, int disable) 1472 { 1473 struct bridge_softc *sc = ifp->if_softc; 1474 1475 KASSERT((ifp->if_flags & IFF_RUNNING) != 0); 1476 ifp->if_flags &= ~IFF_RUNNING; 1477 1478 BRIDGE_LOCK(sc); 1479 sc->sc_stopping = true; 1480 BRIDGE_UNLOCK(sc); 1481 1482 callout_halt(&sc->sc_brcallout, NULL); 1483 workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk); 1484 bstp_stop(sc); 1485 bridge_rtflush(sc, IFBF_FLUSHDYN); 1486 } 1487 1488 /* 1489 * bridge_enqueue: 1490 * 1491 * Enqueue a packet on a bridge member interface. 1492 */ 1493 void 1494 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m, 1495 int runfilt) 1496 { 1497 int len, error; 1498 short mflags; 1499 1500 if (runfilt) { 1501 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, 1502 dst_ifp, PFIL_OUT) != 0) { 1503 if (m != NULL) 1504 m_freem(m); 1505 return; 1506 } 1507 if (m == NULL) 1508 return; 1509 } 1510 1511 #ifdef ALTQ 1512 KERNEL_LOCK(1, NULL); 1513 /* 1514 * If ALTQ is enabled on the member interface, do 1515 * classification; the queueing discipline might 1516 * not require classification, but might require 1517 * the address family/header pointer in the pktattr. 1518 */ 1519 if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) { 1520 /* XXX IFT_ETHER */ 1521 altq_etherclassify(&dst_ifp->if_snd, m); 1522 } 1523 KERNEL_UNLOCK_ONE(NULL); 1524 #endif /* ALTQ */ 1525 1526 if (vlan_has_tag(m) && 1527 !vlan_is_hwtag_enabled(dst_ifp)) { 1528 (void)ether_inject_vlantag(&m, ETHERTYPE_VLAN, 1529 vlan_get_tag(m)); 1530 if (m == NULL) { 1531 if_statinc(&sc->sc_if, if_oerrors); 1532 return; 1533 } 1534 } 1535 1536 len = m->m_pkthdr.len; 1537 mflags = m->m_flags; 1538 1539 error = if_transmit_lock(dst_ifp, m); 1540 if (error) { 1541 /* mbuf is already freed */ 1542 if_statinc(&sc->sc_if, if_oerrors); 1543 return; 1544 } 1545 1546 net_stat_ref_t nsr = IF_STAT_GETREF(&sc->sc_if); 1547 if_statinc_ref(nsr, if_opackets); 1548 if_statadd_ref(nsr, if_obytes, len); 1549 if (mflags & M_MCAST) 1550 if_statinc_ref(nsr, if_omcasts); 1551 IF_STAT_PUTREF(&sc->sc_if); 1552 } 1553 1554 /* 1555 * bridge_output: 1556 * 1557 * Send output from a bridge member interface. This 1558 * performs the bridging function for locally originated 1559 * packets. 1560 * 1561 * The mbuf has the Ethernet header already attached. We must 1562 * enqueue or free the mbuf before returning. 1563 */ 1564 int 1565 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, 1566 const struct rtentry *rt) 1567 { 1568 struct ether_header *eh; 1569 struct ifnet *dst_if; 1570 struct bridge_softc *sc; 1571 struct mbuf *n; 1572 int s, bound; 1573 1574 /* 1575 * bridge_output() is called from ether_output(), furthermore 1576 * ifp argument doesn't point to bridge(4). So, don't assert 1577 * IFEF_MPSAFE here. 1578 */ 1579 1580 KASSERT(m->m_len >= ETHER_HDR_LEN); 1581 1582 eh = mtod(m, struct ether_header *); 1583 sc = ifp->if_bridge; 1584 1585 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1586 if (memcmp(etherbroadcastaddr, 1587 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 1588 m->m_flags |= M_BCAST; 1589 else 1590 m->m_flags |= M_MCAST; 1591 } 1592 1593 /* 1594 * If bridge is down, but the original output interface is up, 1595 * go ahead and send out that interface. Otherwise, the packet 1596 * is dropped below. 1597 */ 1598 if (__predict_false(sc == NULL) || 1599 (sc->sc_if.if_flags & IFF_RUNNING) == 0) { 1600 dst_if = ifp; 1601 goto unicast_asis; 1602 } 1603 1604 /* 1605 * If the packet is a multicast, or we don't know a better way to 1606 * get there, send to all interfaces. 1607 */ 1608 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0) 1609 dst_if = NULL; 1610 else 1611 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1612 1613 /* 1614 * In general, we need to handle TX offload in software before 1615 * enqueueing a packet. However, we can send it as is in the 1616 * cases of unicast via (1) the source interface, or (2) an 1617 * interface which supports the specified offload options. 1618 * For multicast or broadcast, send it as is only if (3) all 1619 * the member interfaces support the specified options. 1620 */ 1621 1622 /* 1623 * Unicast via the source interface. 1624 */ 1625 if (dst_if == ifp) 1626 goto unicast_asis; 1627 1628 /* 1629 * Unicast via other interface. 1630 */ 1631 if (dst_if != NULL) { 1632 KASSERT(m->m_flags & M_PKTHDR); 1633 if (TX_OFFLOAD_SUPPORTED(dst_if->if_csum_flags_tx, 1634 m->m_pkthdr.csum_flags)) { 1635 /* 1636 * Unicast via an interface which supports the 1637 * specified offload options. 1638 */ 1639 goto unicast_asis; 1640 } 1641 1642 /* 1643 * Handle TX offload in software. For TSO, a packet is 1644 * split into multiple chunks. Thus, the return value of 1645 * ether_sw_offload_tx() is mbuf queue consists of them. 1646 */ 1647 m = ether_sw_offload_tx(ifp, m); 1648 if (m == NULL) 1649 return 0; 1650 1651 do { 1652 n = m->m_nextpkt; 1653 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1654 m_freem(m); 1655 else 1656 bridge_enqueue(sc, dst_if, m, 0); 1657 m = n; 1658 } while (m != NULL); 1659 1660 return 0; 1661 } 1662 1663 /* 1664 * Multicast or broadcast. 1665 */ 1666 if (TX_OFFLOAD_SUPPORTED(sc->sc_csum_flags_tx, 1667 m->m_pkthdr.csum_flags)) { 1668 /* 1669 * Specified TX offload options are supported by all 1670 * the member interfaces of this bridge. 1671 */ 1672 m->m_nextpkt = NULL; /* XXX */ 1673 } else { 1674 /* 1675 * Otherwise, handle TX offload in software. 1676 */ 1677 m = ether_sw_offload_tx(ifp, m); 1678 if (m == NULL) 1679 return 0; 1680 } 1681 1682 /* 1683 * When we use pppoe over bridge, bridge_output() can be called 1684 * in a lwp context by pppoe_timeout_wk(). 1685 */ 1686 bound = curlwp_bind(); 1687 do { 1688 /* XXX Should call bridge_broadcast, but there are locking 1689 * issues which need resolving first. */ 1690 struct bridge_iflist *bif; 1691 struct mbuf *mc; 1692 bool used = false; 1693 1694 n = m->m_nextpkt; 1695 1696 BRIDGE_PSZ_RENTER(s); 1697 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 1698 struct psref psref; 1699 1700 bridge_acquire_member(sc, bif, &psref); 1701 BRIDGE_PSZ_REXIT(s); 1702 1703 dst_if = bif->bif_ifp; 1704 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1705 goto next; 1706 1707 /* 1708 * If this is not the original output interface, 1709 * and the interface is participating in spanning 1710 * tree, make sure the port is in a state that 1711 * allows forwarding. 1712 */ 1713 if (dst_if != ifp && 1714 (bif->bif_flags & IFBIF_STP) != 0) { 1715 switch (bif->bif_state) { 1716 case BSTP_IFSTATE_BLOCKING: 1717 case BSTP_IFSTATE_LISTENING: 1718 case BSTP_IFSTATE_DISABLED: 1719 goto next; 1720 } 1721 } 1722 1723 if (PSLIST_READER_NEXT(bif, struct bridge_iflist, 1724 bif_next) == NULL && 1725 ((m->m_flags & (M_MCAST | M_BCAST)) == 0 || 1726 dst_if == ifp)) 1727 { 1728 used = true; 1729 mc = m; 1730 } else { 1731 mc = m_copypacket(m, M_DONTWAIT); 1732 if (mc == NULL) { 1733 if_statinc(&sc->sc_if, if_oerrors); 1734 goto next; 1735 } 1736 } 1737 1738 bridge_enqueue(sc, dst_if, mc, 0); 1739 1740 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 && 1741 dst_if != ifp) 1742 { 1743 if (PSLIST_READER_NEXT(bif, 1744 struct bridge_iflist, bif_next) == NULL) 1745 { 1746 used = true; 1747 mc = m; 1748 } else { 1749 mc = m_copypacket(m, M_DONTWAIT); 1750 if (mc == NULL) { 1751 if_statinc(&sc->sc_if, 1752 if_oerrors); 1753 goto next; 1754 } 1755 } 1756 1757 m_set_rcvif(mc, dst_if); 1758 mc->m_flags &= ~M_PROMISC; 1759 1760 const int _s = splsoftnet(); 1761 KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if); 1762 ether_input(dst_if, mc); 1763 KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if); 1764 splx(_s); 1765 } 1766 1767 next: 1768 BRIDGE_PSZ_RENTER(s); 1769 bridge_release_member(sc, bif, &psref); 1770 1771 /* Guarantee we don't re-enter the loop as we already 1772 * decided we're at the end. */ 1773 if (used) 1774 break; 1775 } 1776 BRIDGE_PSZ_REXIT(s); 1777 1778 if (!used) 1779 m_freem(m); 1780 1781 m = n; 1782 } while (m != NULL); 1783 curlwp_bindx(bound); 1784 1785 return 0; 1786 1787 unicast_asis: 1788 /* 1789 * XXX Spanning tree consideration here? 1790 */ 1791 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1792 m_freem(m); 1793 else 1794 bridge_enqueue(sc, dst_if, m, 0); 1795 return 0; 1796 } 1797 1798 /* 1799 * bridge_start: 1800 * 1801 * Start output on a bridge. 1802 * 1803 * NOTE: This routine should never be called in this implementation. 1804 */ 1805 static void 1806 bridge_start(struct ifnet *ifp) 1807 { 1808 1809 printf("%s: bridge_start() called\n", ifp->if_xname); 1810 } 1811 1812 /* 1813 * bridge_forward: 1814 * 1815 * The forwarding function of the bridge. 1816 */ 1817 static void 1818 bridge_forward(struct bridge_softc *sc, struct mbuf *m) 1819 { 1820 struct bridge_iflist *bif; 1821 struct ifnet *src_if, *dst_if; 1822 struct ether_header *eh; 1823 struct psref psref; 1824 struct psref psref_src; 1825 DECLARE_LOCK_VARIABLE; 1826 bool src_if_protected; 1827 1828 if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) 1829 return; 1830 1831 src_if = m_get_rcvif_psref(m, &psref_src); 1832 if (src_if == NULL) { 1833 /* Interface is being destroyed? */ 1834 m_freem(m); 1835 goto out; 1836 } 1837 1838 if_statadd2(&sc->sc_if, if_ipackets, 1, if_ibytes, m->m_pkthdr.len); 1839 1840 /* 1841 * Look up the bridge_iflist. 1842 */ 1843 bif = bridge_lookup_member_if(sc, src_if, &psref); 1844 if (bif == NULL) { 1845 /* Interface is not a bridge member (anymore?) */ 1846 m_freem(m); 1847 goto out; 1848 } 1849 1850 if (bif->bif_flags & IFBIF_STP) { 1851 switch (bif->bif_state) { 1852 case BSTP_IFSTATE_BLOCKING: 1853 case BSTP_IFSTATE_LISTENING: 1854 case BSTP_IFSTATE_DISABLED: 1855 m_freem(m); 1856 bridge_release_member(sc, bif, &psref); 1857 goto out; 1858 } 1859 } 1860 1861 eh = mtod(m, struct ether_header *); 1862 1863 /* 1864 * If the interface is learning, and the source 1865 * address is valid and not multicast, record 1866 * the address. 1867 */ 1868 if ((bif->bif_flags & IFBIF_LEARNING) != 0 && 1869 ETHER_IS_MULTICAST(eh->ether_shost) == 0 && 1870 (eh->ether_shost[0] == 0 && 1871 eh->ether_shost[1] == 0 && 1872 eh->ether_shost[2] == 0 && 1873 eh->ether_shost[3] == 0 && 1874 eh->ether_shost[4] == 0 && 1875 eh->ether_shost[5] == 0) == 0) { 1876 (void) bridge_rtupdate(sc, eh->ether_shost, 1877 src_if, 0, IFBAF_DYNAMIC); 1878 } 1879 1880 if ((bif->bif_flags & IFBIF_STP) != 0 && 1881 bif->bif_state == BSTP_IFSTATE_LEARNING) { 1882 m_freem(m); 1883 bridge_release_member(sc, bif, &psref); 1884 goto out; 1885 } 1886 1887 src_if_protected = ((bif->bif_flags & IFBIF_PROTECTED) != 0); 1888 1889 bridge_release_member(sc, bif, &psref); 1890 1891 /* 1892 * At this point, the port either doesn't participate 1893 * in spanning tree or it is in the forwarding state. 1894 */ 1895 1896 /* 1897 * If the packet is unicast, destined for someone on 1898 * "this" side of the bridge, drop it. 1899 */ 1900 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) { 1901 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1902 if (src_if == dst_if) { 1903 m_freem(m); 1904 goto out; 1905 } 1906 } else { 1907 /* ...forward it to all interfaces. */ 1908 if_statinc(&sc->sc_if, if_imcasts); 1909 dst_if = NULL; 1910 } 1911 1912 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) { 1913 if (m != NULL) 1914 m_freem(m); 1915 goto out; 1916 } 1917 if (m == NULL) 1918 goto out; 1919 1920 if (dst_if == NULL) { 1921 bridge_broadcast(sc, src_if, src_if_protected, m); 1922 goto out; 1923 } 1924 1925 m_put_rcvif_psref(src_if, &psref_src); 1926 src_if = NULL; 1927 1928 /* 1929 * At this point, we're dealing with a unicast frame 1930 * going to a different interface. 1931 */ 1932 if ((dst_if->if_flags & IFF_RUNNING) == 0) { 1933 m_freem(m); 1934 goto out; 1935 } 1936 1937 bif = bridge_lookup_member_if(sc, dst_if, &psref); 1938 if (bif == NULL) { 1939 /* Not a member of the bridge (anymore?) */ 1940 m_freem(m); 1941 goto out; 1942 } 1943 1944 if (bif->bif_flags & IFBIF_STP) { 1945 switch (bif->bif_state) { 1946 case BSTP_IFSTATE_DISABLED: 1947 case BSTP_IFSTATE_BLOCKING: 1948 m_freem(m); 1949 bridge_release_member(sc, bif, &psref); 1950 goto out; 1951 } 1952 } 1953 1954 if ((bif->bif_flags & IFBIF_PROTECTED) && src_if_protected) { 1955 m_freem(m); 1956 bridge_release_member(sc, bif, &psref); 1957 goto out; 1958 } 1959 1960 bridge_release_member(sc, bif, &psref); 1961 1962 /* 1963 * Before enqueueing this packet to the destination interface, 1964 * clear any in-bound checksum flags to prevent them from being 1965 * misused as out-bound flags. 1966 */ 1967 m->m_pkthdr.csum_flags = 0; 1968 1969 ACQUIRE_GLOBAL_LOCKS(); 1970 bridge_enqueue(sc, dst_if, m, 1); 1971 RELEASE_GLOBAL_LOCKS(); 1972 out: 1973 if (src_if != NULL) 1974 m_put_rcvif_psref(src_if, &psref_src); 1975 return; 1976 } 1977 1978 static bool 1979 bstp_state_before_learning(struct bridge_iflist *bif) 1980 { 1981 if (bif->bif_flags & IFBIF_STP) { 1982 switch (bif->bif_state) { 1983 case BSTP_IFSTATE_BLOCKING: 1984 case BSTP_IFSTATE_LISTENING: 1985 case BSTP_IFSTATE_DISABLED: 1986 return true; 1987 } 1988 } 1989 return false; 1990 } 1991 1992 static bool 1993 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src) 1994 { 1995 uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost; 1996 1997 if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0 1998 #if NCARP > 0 1999 || (bif->bif_ifp->if_carp && 2000 carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL) 2001 #endif /* NCARP > 0 */ 2002 ) 2003 return true; 2004 2005 return false; 2006 } 2007 2008 /* 2009 * bridge_input: 2010 * 2011 * Receive input from a member interface. Queue the packet for 2012 * bridging if it is not for us. 2013 */ 2014 static void 2015 bridge_input(struct ifnet *ifp, struct mbuf *m) 2016 { 2017 struct bridge_softc *sc = ifp->if_bridge; 2018 struct bridge_iflist *bif; 2019 struct ether_header *eh; 2020 struct psref psref; 2021 int bound; 2022 DECLARE_LOCK_VARIABLE; 2023 2024 KASSERT(!cpu_intr_p()); 2025 2026 if (__predict_false(sc == NULL) || 2027 (sc->sc_if.if_flags & IFF_RUNNING) == 0) { 2028 ACQUIRE_GLOBAL_LOCKS(); 2029 ether_input(ifp, m); 2030 RELEASE_GLOBAL_LOCKS(); 2031 return; 2032 } 2033 2034 bound = curlwp_bind(); 2035 bif = bridge_lookup_member_if(sc, ifp, &psref); 2036 if (bif == NULL) { 2037 curlwp_bindx(bound); 2038 ACQUIRE_GLOBAL_LOCKS(); 2039 ether_input(ifp, m); 2040 RELEASE_GLOBAL_LOCKS(); 2041 return; 2042 } 2043 2044 eh = mtod(m, struct ether_header *); 2045 2046 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 2047 if (memcmp(etherbroadcastaddr, 2048 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 2049 m->m_flags |= M_BCAST; 2050 else 2051 m->m_flags |= M_MCAST; 2052 } 2053 2054 /* 2055 * A 'fast' path for packets addressed to interfaces that are 2056 * part of this bridge. 2057 */ 2058 if (!(m->m_flags & (M_BCAST|M_MCAST)) && 2059 !bstp_state_before_learning(bif)) { 2060 struct bridge_iflist *_bif; 2061 struct ifnet *_ifp = NULL; 2062 int s; 2063 struct psref _psref; 2064 2065 BRIDGE_PSZ_RENTER(s); 2066 BRIDGE_IFLIST_READER_FOREACH(_bif, sc) { 2067 /* It is destined for us. */ 2068 if (bridge_ourether(_bif, eh, 0)) { 2069 bridge_acquire_member(sc, _bif, &_psref); 2070 BRIDGE_PSZ_REXIT(s); 2071 if (_bif->bif_flags & IFBIF_LEARNING) 2072 (void) bridge_rtupdate(sc, 2073 eh->ether_shost, ifp, 0, IFBAF_DYNAMIC); 2074 m_set_rcvif(m, _bif->bif_ifp); 2075 _ifp = _bif->bif_ifp; 2076 bridge_release_member(sc, _bif, &_psref); 2077 goto out; 2078 } 2079 2080 /* We just received a packet that we sent out. */ 2081 if (bridge_ourether(_bif, eh, 1)) 2082 break; 2083 } 2084 BRIDGE_PSZ_REXIT(s); 2085 out: 2086 2087 if (_bif != NULL) { 2088 bridge_release_member(sc, bif, &psref); 2089 curlwp_bindx(bound); 2090 if (_ifp != NULL) { 2091 m->m_flags &= ~M_PROMISC; 2092 ACQUIRE_GLOBAL_LOCKS(); 2093 ether_input(_ifp, m); 2094 RELEASE_GLOBAL_LOCKS(); 2095 } else 2096 m_freem(m); 2097 return; 2098 } 2099 } 2100 2101 /* Tap off 802.1D packets; they do not get forwarded. */ 2102 if (bif->bif_flags & IFBIF_STP && 2103 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) { 2104 bstp_input(sc, bif, m); 2105 bridge_release_member(sc, bif, &psref); 2106 curlwp_bindx(bound); 2107 return; 2108 } 2109 2110 /* 2111 * A normal switch would discard the packet here, but that's not what 2112 * we've done historically. This also prevents some obnoxious behaviour. 2113 */ 2114 if (bstp_state_before_learning(bif)) { 2115 bridge_release_member(sc, bif, &psref); 2116 curlwp_bindx(bound); 2117 ACQUIRE_GLOBAL_LOCKS(); 2118 ether_input(ifp, m); 2119 RELEASE_GLOBAL_LOCKS(); 2120 return; 2121 } 2122 2123 bridge_release_member(sc, bif, &psref); 2124 2125 bridge_forward(sc, m); 2126 2127 curlwp_bindx(bound); 2128 } 2129 2130 /* 2131 * bridge_broadcast: 2132 * 2133 * Send a frame to all interfaces that are members of 2134 * the bridge, except for the one on which the packet 2135 * arrived. 2136 */ 2137 static void 2138 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if, 2139 bool src_if_protected, struct mbuf *m) 2140 { 2141 struct bridge_iflist *bif; 2142 struct mbuf *mc; 2143 struct ifnet *dst_if; 2144 bool bmcast; 2145 int s; 2146 DECLARE_LOCK_VARIABLE; 2147 2148 bmcast = m->m_flags & (M_BCAST|M_MCAST); 2149 2150 BRIDGE_PSZ_RENTER(s); 2151 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 2152 struct psref psref; 2153 2154 bridge_acquire_member(sc, bif, &psref); 2155 BRIDGE_PSZ_REXIT(s); 2156 2157 dst_if = bif->bif_ifp; 2158 2159 if (bif->bif_flags & IFBIF_STP) { 2160 switch (bif->bif_state) { 2161 case BSTP_IFSTATE_BLOCKING: 2162 case BSTP_IFSTATE_DISABLED: 2163 goto next; 2164 } 2165 } 2166 2167 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast) 2168 goto next; 2169 2170 if ((dst_if->if_flags & IFF_RUNNING) == 0) 2171 goto next; 2172 2173 if (dst_if != src_if) { 2174 if ((bif->bif_flags & IFBIF_PROTECTED) && 2175 src_if_protected) { 2176 goto next; 2177 } 2178 2179 mc = m_copypacket(m, M_DONTWAIT); 2180 if (mc == NULL) { 2181 if_statinc(&sc->sc_if, if_oerrors); 2182 goto next; 2183 } 2184 /* 2185 * Before enqueueing this packet to the destination 2186 * interface, clear any in-bound checksum flags to 2187 * prevent them from being misused as out-bound flags. 2188 */ 2189 mc->m_pkthdr.csum_flags = 0; 2190 2191 ACQUIRE_GLOBAL_LOCKS(); 2192 bridge_enqueue(sc, dst_if, mc, 1); 2193 RELEASE_GLOBAL_LOCKS(); 2194 } 2195 2196 if (bmcast) { 2197 mc = m_copypacket(m, M_DONTWAIT); 2198 if (mc == NULL) { 2199 if_statinc(&sc->sc_if, if_oerrors); 2200 goto next; 2201 } 2202 /* 2203 * Before enqueueing this packet to the destination 2204 * interface, clear any in-bound checksum flags to 2205 * prevent them from being misused as out-bound flags. 2206 */ 2207 mc->m_pkthdr.csum_flags = 0; 2208 2209 m_set_rcvif(mc, dst_if); 2210 mc->m_flags &= ~M_PROMISC; 2211 2212 ACQUIRE_GLOBAL_LOCKS(); 2213 ether_input(dst_if, mc); 2214 RELEASE_GLOBAL_LOCKS(); 2215 } 2216 next: 2217 BRIDGE_PSZ_RENTER(s); 2218 bridge_release_member(sc, bif, &psref); 2219 } 2220 BRIDGE_PSZ_REXIT(s); 2221 2222 m_freem(m); 2223 } 2224 2225 static int 2226 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst, 2227 struct bridge_rtnode **brtp) 2228 { 2229 struct bridge_rtnode *brt; 2230 int error; 2231 2232 if (sc->sc_brtcnt >= sc->sc_brtmax) 2233 return ENOSPC; 2234 2235 /* 2236 * Allocate a new bridge forwarding node, and 2237 * initialize the expiration time and Ethernet 2238 * address. 2239 */ 2240 brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT); 2241 if (brt == NULL) 2242 return ENOMEM; 2243 2244 memset(brt, 0, sizeof(*brt)); 2245 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2246 brt->brt_flags = IFBAF_DYNAMIC; 2247 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN); 2248 PSLIST_ENTRY_INIT(brt, brt_list); 2249 PSLIST_ENTRY_INIT(brt, brt_hash); 2250 2251 BRIDGE_RT_LOCK(sc); 2252 error = bridge_rtnode_insert(sc, brt); 2253 BRIDGE_RT_UNLOCK(sc); 2254 2255 if (error != 0) { 2256 pool_put(&bridge_rtnode_pool, brt); 2257 return error; 2258 } 2259 2260 *brtp = brt; 2261 return 0; 2262 } 2263 2264 /* 2265 * bridge_rtupdate: 2266 * 2267 * Add a bridge routing entry. 2268 */ 2269 static int 2270 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, 2271 struct ifnet *dst_if, int setflags, uint8_t flags) 2272 { 2273 struct bridge_rtnode *brt; 2274 int s; 2275 2276 again: 2277 /* 2278 * A route for this destination might already exist. If so, 2279 * update it, otherwise create a new one. 2280 */ 2281 BRIDGE_RT_RENTER(s); 2282 brt = bridge_rtnode_lookup(sc, dst); 2283 2284 if (brt != NULL) { 2285 brt->brt_ifp = dst_if; 2286 if (setflags) { 2287 brt->brt_flags = flags; 2288 if (flags & IFBAF_STATIC) 2289 brt->brt_expire = 0; 2290 else 2291 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2292 } else { 2293 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 2294 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2295 } 2296 } 2297 BRIDGE_RT_REXIT(s); 2298 2299 if (brt == NULL) { 2300 int r; 2301 2302 r = bridge_rtalloc(sc, dst, &brt); 2303 if (r != 0) 2304 return r; 2305 goto again; 2306 } 2307 2308 return 0; 2309 } 2310 2311 /* 2312 * bridge_rtlookup: 2313 * 2314 * Lookup the destination interface for an address. 2315 */ 2316 static struct ifnet * 2317 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr) 2318 { 2319 struct bridge_rtnode *brt; 2320 struct ifnet *ifs = NULL; 2321 int s; 2322 2323 BRIDGE_RT_RENTER(s); 2324 brt = bridge_rtnode_lookup(sc, addr); 2325 if (brt != NULL) 2326 ifs = brt->brt_ifp; 2327 BRIDGE_RT_REXIT(s); 2328 2329 return ifs; 2330 } 2331 2332 typedef bool (*bridge_iterate_cb_t) 2333 (struct bridge_softc *, struct bridge_rtnode *, bool *, void *); 2334 2335 /* 2336 * bridge_rtlist_iterate_remove: 2337 * 2338 * It iterates on sc->sc_rtlist and removes rtnodes of it which func 2339 * callback judges to remove. Removals of rtnodes are done in a manner 2340 * of pserialize. To this end, all kmem_* operations are placed out of 2341 * mutexes. 2342 */ 2343 static void 2344 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg) 2345 { 2346 struct bridge_rtnode *brt; 2347 struct bridge_rtnode **brt_list; 2348 int i, count; 2349 2350 retry: 2351 count = sc->sc_brtcnt; 2352 if (count == 0) 2353 return; 2354 brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP); 2355 2356 BRIDGE_RT_LOCK(sc); 2357 if (__predict_false(sc->sc_brtcnt > count)) { 2358 /* The rtnodes increased, we need more memory */ 2359 BRIDGE_RT_UNLOCK(sc); 2360 kmem_free(brt_list, sizeof(*brt_list) * count); 2361 goto retry; 2362 } 2363 2364 i = 0; 2365 /* 2366 * We don't need to use a _SAFE variant here because we know 2367 * that a removed item keeps its next pointer as-is thanks to 2368 * pslist(9) and isn't freed in the loop. 2369 */ 2370 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 2371 bool need_break = false; 2372 if (func(sc, brt, &need_break, arg)) { 2373 bridge_rtnode_remove(sc, brt); 2374 brt_list[i++] = brt; 2375 } 2376 if (need_break) 2377 break; 2378 } 2379 2380 if (i > 0) 2381 BRIDGE_RT_PSZ_PERFORM(sc); 2382 BRIDGE_RT_UNLOCK(sc); 2383 2384 while (--i >= 0) 2385 bridge_rtnode_destroy(brt_list[i]); 2386 2387 kmem_free(brt_list, sizeof(*brt_list) * count); 2388 } 2389 2390 static bool 2391 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2392 bool *need_break, void *arg) 2393 { 2394 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 2395 /* Take into account of the subsequent removal */ 2396 if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax) 2397 *need_break = true; 2398 return true; 2399 } else 2400 return false; 2401 } 2402 2403 static void 2404 bridge_rttrim0(struct bridge_softc *sc) 2405 { 2406 bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL); 2407 } 2408 2409 /* 2410 * bridge_rttrim: 2411 * 2412 * Trim the routine table so that we have a number 2413 * of routing entries less than or equal to the 2414 * maximum number. 2415 */ 2416 static void 2417 bridge_rttrim(struct bridge_softc *sc) 2418 { 2419 2420 /* Make sure we actually need to do this. */ 2421 if (sc->sc_brtcnt <= sc->sc_brtmax) 2422 return; 2423 2424 /* Force an aging cycle; this might trim enough addresses. */ 2425 bridge_rtage(sc); 2426 if (sc->sc_brtcnt <= sc->sc_brtmax) 2427 return; 2428 2429 bridge_rttrim0(sc); 2430 2431 return; 2432 } 2433 2434 /* 2435 * bridge_timer: 2436 * 2437 * Aging timer for the bridge. 2438 */ 2439 static void 2440 bridge_timer(void *arg) 2441 { 2442 struct bridge_softc *sc = arg; 2443 2444 workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL); 2445 } 2446 2447 static void 2448 bridge_rtage_work(struct work *wk, void *arg) 2449 { 2450 struct bridge_softc *sc = arg; 2451 2452 KASSERT(wk == &sc->sc_rtage_wk); 2453 2454 bridge_rtage(sc); 2455 2456 BRIDGE_LOCK(sc); 2457 if (!sc->sc_stopping) { 2458 callout_reset(&sc->sc_brcallout, 2459 bridge_rtable_prune_period * hz, bridge_timer, sc); 2460 } 2461 BRIDGE_UNLOCK(sc); 2462 } 2463 2464 static bool 2465 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2466 bool *need_break, void *arg) 2467 { 2468 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 2469 time_uptime >= brt->brt_expire) 2470 return true; 2471 else 2472 return false; 2473 } 2474 2475 /* 2476 * bridge_rtage: 2477 * 2478 * Perform an aging cycle. 2479 */ 2480 static void 2481 bridge_rtage(struct bridge_softc *sc) 2482 { 2483 bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL); 2484 } 2485 2486 2487 static bool 2488 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2489 bool *need_break, void *arg) 2490 { 2491 int full = *(int*)arg; 2492 2493 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 2494 return true; 2495 else 2496 return false; 2497 } 2498 2499 /* 2500 * bridge_rtflush: 2501 * 2502 * Remove all dynamic addresses from the bridge. 2503 */ 2504 static void 2505 bridge_rtflush(struct bridge_softc *sc, int full) 2506 { 2507 bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full); 2508 } 2509 2510 /* 2511 * bridge_rtdaddr: 2512 * 2513 * Remove an address from the table. 2514 */ 2515 static int 2516 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr) 2517 { 2518 struct bridge_rtnode *brt; 2519 2520 BRIDGE_RT_LOCK(sc); 2521 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) { 2522 BRIDGE_RT_UNLOCK(sc); 2523 return ENOENT; 2524 } 2525 bridge_rtnode_remove(sc, brt); 2526 BRIDGE_RT_PSZ_PERFORM(sc); 2527 BRIDGE_RT_UNLOCK(sc); 2528 2529 bridge_rtnode_destroy(brt); 2530 2531 return 0; 2532 } 2533 2534 /* 2535 * bridge_rtdelete: 2536 * 2537 * Delete routes to a speicifc member interface. 2538 */ 2539 static void 2540 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp) 2541 { 2542 struct bridge_rtnode *brt; 2543 2544 /* XXX pserialize_perform for each entry is slow */ 2545 again: 2546 BRIDGE_RT_LOCK(sc); 2547 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 2548 if (brt->brt_ifp == ifp) 2549 break; 2550 } 2551 if (brt == NULL) { 2552 BRIDGE_RT_UNLOCK(sc); 2553 return; 2554 } 2555 bridge_rtnode_remove(sc, brt); 2556 BRIDGE_RT_PSZ_PERFORM(sc); 2557 BRIDGE_RT_UNLOCK(sc); 2558 2559 bridge_rtnode_destroy(brt); 2560 2561 goto again; 2562 } 2563 2564 /* 2565 * bridge_rtable_init: 2566 * 2567 * Initialize the route table for this bridge. 2568 */ 2569 static void 2570 bridge_rtable_init(struct bridge_softc *sc) 2571 { 2572 int i; 2573 2574 sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE, 2575 KM_SLEEP); 2576 2577 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++) 2578 PSLIST_INIT(&sc->sc_rthash[i]); 2579 2580 sc->sc_rthash_key = cprng_fast32(); 2581 2582 PSLIST_INIT(&sc->sc_rtlist); 2583 2584 sc->sc_rtlist_psz = pserialize_create(); 2585 sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET); 2586 } 2587 2588 /* 2589 * bridge_rtable_fini: 2590 * 2591 * Deconstruct the route table for this bridge. 2592 */ 2593 static void 2594 bridge_rtable_fini(struct bridge_softc *sc) 2595 { 2596 2597 kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE); 2598 mutex_obj_free(sc->sc_rtlist_lock); 2599 pserialize_destroy(sc->sc_rtlist_psz); 2600 } 2601 2602 /* 2603 * The following hash function is adapted from "Hash Functions" by Bob Jenkins 2604 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997). 2605 */ 2606 #define mix(a, b, c) \ 2607 do { \ 2608 a -= b; a -= c; a ^= (c >> 13); \ 2609 b -= c; b -= a; b ^= (a << 8); \ 2610 c -= a; c -= b; c ^= (b >> 13); \ 2611 a -= b; a -= c; a ^= (c >> 12); \ 2612 b -= c; b -= a; b ^= (a << 16); \ 2613 c -= a; c -= b; c ^= (b >> 5); \ 2614 a -= b; a -= c; a ^= (c >> 3); \ 2615 b -= c; b -= a; b ^= (a << 10); \ 2616 c -= a; c -= b; c ^= (b >> 15); \ 2617 } while (/*CONSTCOND*/0) 2618 2619 static inline uint32_t 2620 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr) 2621 { 2622 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key; 2623 2624 b += addr[5] << 8; 2625 b += addr[4]; 2626 a += (uint32_t)addr[3] << 24; 2627 a += addr[2] << 16; 2628 a += addr[1] << 8; 2629 a += addr[0]; 2630 2631 mix(a, b, c); 2632 2633 return (c & BRIDGE_RTHASH_MASK); 2634 } 2635 2636 #undef mix 2637 2638 /* 2639 * bridge_rtnode_lookup: 2640 * 2641 * Look up a bridge route node for the specified destination. 2642 */ 2643 static struct bridge_rtnode * 2644 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr) 2645 { 2646 struct bridge_rtnode *brt; 2647 uint32_t hash; 2648 int dir; 2649 2650 hash = bridge_rthash(sc, addr); 2651 BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) { 2652 dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN); 2653 if (dir == 0) 2654 return brt; 2655 if (dir > 0) 2656 return NULL; 2657 } 2658 2659 return NULL; 2660 } 2661 2662 /* 2663 * bridge_rtnode_insert: 2664 * 2665 * Insert the specified bridge node into the route table. We 2666 * assume the entry is not already in the table. 2667 */ 2668 static int 2669 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt) 2670 { 2671 struct bridge_rtnode *lbrt, *prev = NULL; 2672 uint32_t hash; 2673 2674 KASSERT(BRIDGE_RT_LOCKED(sc)); 2675 2676 hash = bridge_rthash(sc, brt->brt_addr); 2677 BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) { 2678 int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN); 2679 if (dir == 0) 2680 return EEXIST; 2681 if (dir > 0) 2682 break; 2683 prev = lbrt; 2684 } 2685 if (prev == NULL) 2686 BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt); 2687 else 2688 BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt); 2689 2690 BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt); 2691 sc->sc_brtcnt++; 2692 2693 return 0; 2694 } 2695 2696 /* 2697 * bridge_rtnode_remove: 2698 * 2699 * Remove a bridge rtnode from the rthash and the rtlist of a bridge. 2700 */ 2701 static void 2702 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt) 2703 { 2704 2705 KASSERT(BRIDGE_RT_LOCKED(sc)); 2706 2707 BRIDGE_RTHASH_WRITER_REMOVE(brt); 2708 BRIDGE_RTLIST_WRITER_REMOVE(brt); 2709 sc->sc_brtcnt--; 2710 } 2711 2712 /* 2713 * bridge_rtnode_destroy: 2714 * 2715 * Destroy a bridge rtnode. 2716 */ 2717 static void 2718 bridge_rtnode_destroy(struct bridge_rtnode *brt) 2719 { 2720 2721 PSLIST_ENTRY_DESTROY(brt, brt_list); 2722 PSLIST_ENTRY_DESTROY(brt, brt_hash); 2723 pool_put(&bridge_rtnode_pool, brt); 2724 } 2725 2726 extern pfil_head_t *inet_pfil_hook; /* XXX */ 2727 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 2728 2729 /* 2730 * Send bridge packets through IPF if they are one of the types IPF can deal 2731 * with, or if they are ARP or REVARP. (IPF will pass ARP and REVARP without 2732 * question.) 2733 */ 2734 static int 2735 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir) 2736 { 2737 int snap, error; 2738 struct ether_header *eh1, eh2; 2739 struct llc llc1; 2740 uint16_t ether_type; 2741 2742 snap = 0; 2743 error = -1; /* Default error if not error == 0 */ 2744 eh1 = mtod(*mp, struct ether_header *); 2745 ether_type = ntohs(eh1->ether_type); 2746 2747 /* 2748 * Check for SNAP/LLC. 2749 */ 2750 if (ether_type < ETHERMTU) { 2751 struct llc *llc2 = (struct llc *)(eh1 + 1); 2752 2753 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 && 2754 llc2->llc_dsap == LLC_SNAP_LSAP && 2755 llc2->llc_ssap == LLC_SNAP_LSAP && 2756 llc2->llc_control == LLC_UI) { 2757 ether_type = htons(llc2->llc_un.type_snap.ether_type); 2758 snap = 1; 2759 } 2760 } 2761 2762 /* drop VLAN traffic untagged by hardware offloading */ 2763 if (vlan_has_tag(*mp)) 2764 goto bad; 2765 2766 /* 2767 * If we're trying to filter bridge traffic, don't look at anything 2768 * other than IP and ARP traffic. If the filter doesn't understand 2769 * IPv6, don't allow IPv6 through the bridge either. This is lame 2770 * since if we really wanted, say, an AppleTalk filter, we are hosed, 2771 * but of course we don't have an AppleTalk filter to begin with. 2772 * (Note that since IPF doesn't understand ARP it will pass *ALL* 2773 * ARP traffic.) 2774 */ 2775 switch (ether_type) { 2776 case ETHERTYPE_ARP: 2777 case ETHERTYPE_REVARP: 2778 return 0; /* Automatically pass */ 2779 case ETHERTYPE_IP: 2780 # ifdef INET6 2781 case ETHERTYPE_IPV6: 2782 # endif /* INET6 */ 2783 break; 2784 default: 2785 goto bad; 2786 } 2787 2788 /* Strip off the Ethernet header and keep a copy. */ 2789 m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2); 2790 m_adj(*mp, ETHER_HDR_LEN); 2791 2792 /* Strip off snap header, if present */ 2793 if (snap) { 2794 m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1); 2795 m_adj(*mp, sizeof(struct llc)); 2796 } 2797 2798 /* 2799 * Check basic packet sanity and run IPF through pfil. 2800 */ 2801 KASSERT(!cpu_intr_p()); 2802 switch (ether_type) 2803 { 2804 case ETHERTYPE_IP : 2805 error = bridge_ip_checkbasic(mp); 2806 if (error == 0) 2807 error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir); 2808 break; 2809 # ifdef INET6 2810 case ETHERTYPE_IPV6 : 2811 error = bridge_ip6_checkbasic(mp); 2812 if (error == 0) 2813 error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir); 2814 break; 2815 # endif 2816 default : 2817 error = 0; 2818 break; 2819 } 2820 2821 if (*mp == NULL) 2822 return error; 2823 if (error != 0) 2824 goto bad; 2825 2826 error = -1; 2827 2828 /* 2829 * Finally, put everything back the way it was and return 2830 */ 2831 if (snap) { 2832 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT); 2833 if (*mp == NULL) 2834 return error; 2835 bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc)); 2836 } 2837 2838 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT); 2839 if (*mp == NULL) 2840 return error; 2841 bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN); 2842 2843 return 0; 2844 2845 bad: 2846 m_freem(*mp); 2847 *mp = NULL; 2848 return error; 2849 } 2850 2851 /* 2852 * Perform basic checks on header size since 2853 * IPF assumes ip_input has already processed 2854 * it for it. Cut-and-pasted from ip_input.c. 2855 * Given how simple the IPv6 version is, 2856 * does the IPv4 version really need to be 2857 * this complicated? 2858 * 2859 * XXX Should we update ipstat here, or not? 2860 * XXX Right now we update ipstat but not 2861 * XXX csum_counter. 2862 */ 2863 static int 2864 bridge_ip_checkbasic(struct mbuf **mp) 2865 { 2866 struct mbuf *m = *mp; 2867 struct ip *ip; 2868 int len, hlen; 2869 2870 if (*mp == NULL) 2871 return -1; 2872 2873 if (M_GET_ALIGNED_HDR(&m, struct ip, true) != 0) { 2874 /* XXXJRT new stat, please */ 2875 ip_statinc(IP_STAT_TOOSMALL); 2876 goto bad; 2877 } 2878 ip = mtod(m, struct ip *); 2879 if (ip == NULL) goto bad; 2880 2881 if (ip->ip_v != IPVERSION) { 2882 ip_statinc(IP_STAT_BADVERS); 2883 goto bad; 2884 } 2885 hlen = ip->ip_hl << 2; 2886 if (hlen < sizeof(struct ip)) { /* minimum header length */ 2887 ip_statinc(IP_STAT_BADHLEN); 2888 goto bad; 2889 } 2890 if (hlen > m->m_len) { 2891 if ((m = m_pullup(m, hlen)) == 0) { 2892 ip_statinc(IP_STAT_BADHLEN); 2893 goto bad; 2894 } 2895 ip = mtod(m, struct ip *); 2896 if (ip == NULL) goto bad; 2897 } 2898 2899 switch (m->m_pkthdr.csum_flags & 2900 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) | 2901 M_CSUM_IPv4_BAD)) { 2902 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 2903 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */ 2904 goto bad; 2905 2906 case M_CSUM_IPv4: 2907 /* Checksum was okay. */ 2908 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */ 2909 break; 2910 2911 default: 2912 /* Must compute it ourselves. */ 2913 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */ 2914 if (in_cksum(m, hlen) != 0) 2915 goto bad; 2916 break; 2917 } 2918 2919 /* Retrieve the packet length. */ 2920 len = ntohs(ip->ip_len); 2921 2922 /* 2923 * Check for additional length bogosity 2924 */ 2925 if (len < hlen) { 2926 ip_statinc(IP_STAT_BADLEN); 2927 goto bad; 2928 } 2929 2930 /* 2931 * Check that the amount of data in the buffers 2932 * is as at least much as the IP header would have us expect. 2933 * Drop packet if shorter than we expect. 2934 */ 2935 if (m->m_pkthdr.len < len) { 2936 ip_statinc(IP_STAT_TOOSHORT); 2937 goto bad; 2938 } 2939 2940 /* Checks out, proceed */ 2941 *mp = m; 2942 return 0; 2943 2944 bad: 2945 *mp = m; 2946 return -1; 2947 } 2948 2949 # ifdef INET6 2950 /* 2951 * Same as above, but for IPv6. 2952 * Cut-and-pasted from ip6_input.c. 2953 * XXX Should we update ip6stat, or not? 2954 */ 2955 static int 2956 bridge_ip6_checkbasic(struct mbuf **mp) 2957 { 2958 struct mbuf *m = *mp; 2959 struct ip6_hdr *ip6; 2960 2961 /* 2962 * If the IPv6 header is not aligned, slurp it up into a new 2963 * mbuf with space for link headers, in the event we forward 2964 * it. Otherwise, if it is aligned, make sure the entire base 2965 * IPv6 header is in the first mbuf of the chain. 2966 */ 2967 if (M_GET_ALIGNED_HDR(&m, struct ip6_hdr, true) != 0) { 2968 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m); 2969 /* XXXJRT new stat, please */ 2970 ip6_statinc(IP6_STAT_TOOSMALL); 2971 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 2972 goto bad; 2973 } 2974 2975 ip6 = mtod(m, struct ip6_hdr *); 2976 2977 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 2978 ip6_statinc(IP6_STAT_BADVERS); 2979 in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr); 2980 goto bad; 2981 } 2982 2983 /* Checks out, proceed */ 2984 *mp = m; 2985 return 0; 2986 2987 bad: 2988 *mp = m; 2989 return -1; 2990 } 2991 # endif /* INET6 */ 2992
Indexes created Thu Oct 02 14:10:14 GMT 2025