if_bridge.c revision 1.189.4.1
1 /* $NetBSD: if_bridge.c,v 1.189.4.1 2024/07/20 15:54:10 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 1.189.4.1 2024/07/20 15:54:10 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> 105 106 #include <net/bpf.h> 107 #include <net/if.h> 108 #include <net/if_dl.h> 109 #include <net/if_types.h> 110 #include <net/if_llc.h> 111 112 #include <net/if_ether.h> 113 #include <net/if_bridgevar.h> 114 #include <net/ether_sw_offload.h> 115 116 /* Used for bridge_ip[6]_checkbasic */ 117 #include <netinet/in.h> 118 #include <netinet/in_systm.h> 119 #include <netinet/ip.h> 120 #include <netinet/ip_var.h> 121 #include <netinet/ip_private.h> /* XXX */ 122 #include <netinet/ip6.h> 123 #include <netinet6/in6_var.h> 124 #include <netinet6/ip6_var.h> 125 #include <netinet6/ip6_private.h> /* XXX */ 126 127 /* 128 * Size of the route hash table. Must be a power of two. 129 */ 130 #ifndef BRIDGE_RTHASH_SIZE 131 #define BRIDGE_RTHASH_SIZE 1024 132 #endif 133 134 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1) 135 136 #include "carp.h" 137 #if NCARP > 0 138 #include <netinet/in.h> 139 #include <netinet/in_var.h> 140 #include <netinet/ip_carp.h> 141 #endif 142 143 #include "ioconf.h" 144 145 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf)); 146 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len)); 147 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf)); 148 149 /* 150 * Maximum number of addresses to cache. 151 */ 152 #ifndef BRIDGE_RTABLE_MAX 153 #define BRIDGE_RTABLE_MAX 100 154 #endif 155 156 /* 157 * Spanning tree defaults. 158 */ 159 #define BSTP_DEFAULT_MAX_AGE (20 * 256) 160 #define BSTP_DEFAULT_HELLO_TIME (2 * 256) 161 #define BSTP_DEFAULT_FORWARD_DELAY (15 * 256) 162 #define BSTP_DEFAULT_HOLD_TIME (1 * 256) 163 #define BSTP_DEFAULT_BRIDGE_PRIORITY 0x8000 164 #define BSTP_DEFAULT_PORT_PRIORITY 0x80 165 #define BSTP_DEFAULT_PATH_COST 55 166 167 /* 168 * Timeout (in seconds) for entries learned dynamically. 169 */ 170 #ifndef BRIDGE_RTABLE_TIMEOUT 171 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */ 172 #endif 173 174 /* 175 * Number of seconds between walks of the route list. 176 */ 177 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD 178 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60) 179 #endif 180 181 #define BRIDGE_RT_LOCK(_sc) mutex_enter((_sc)->sc_rtlist_lock) 182 #define BRIDGE_RT_UNLOCK(_sc) mutex_exit((_sc)->sc_rtlist_lock) 183 #define BRIDGE_RT_LOCKED(_sc) mutex_owned((_sc)->sc_rtlist_lock) 184 185 #define BRIDGE_RT_PSZ_PERFORM(_sc) \ 186 pserialize_perform((_sc)->sc_rtlist_psz) 187 188 #define BRIDGE_RT_RENTER(__s) do { __s = pserialize_read_enter(); } while (0) 189 #define BRIDGE_RT_REXIT(__s) do { pserialize_read_exit(__s); } while (0) 190 191 #define BRIDGE_RTLIST_READER_FOREACH(_brt, _sc) \ 192 PSLIST_READER_FOREACH((_brt), &((_sc)->sc_rtlist), \ 193 struct bridge_rtnode, brt_list) 194 #define BRIDGE_RTLIST_WRITER_FOREACH(_brt, _sc) \ 195 PSLIST_WRITER_FOREACH((_brt), &((_sc)->sc_rtlist), \ 196 struct bridge_rtnode, brt_list) 197 #define BRIDGE_RTLIST_WRITER_INSERT_HEAD(_sc, _brt) \ 198 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rtlist, brt, brt_list) 199 #define BRIDGE_RTLIST_WRITER_REMOVE(_brt) \ 200 PSLIST_WRITER_REMOVE((_brt), brt_list) 201 202 #define BRIDGE_RTHASH_READER_FOREACH(_brt, _sc, _hash) \ 203 PSLIST_READER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \ 204 struct bridge_rtnode, brt_hash) 205 #define BRIDGE_RTHASH_WRITER_FOREACH(_brt, _sc, _hash) \ 206 PSLIST_WRITER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \ 207 struct bridge_rtnode, brt_hash) 208 #define BRIDGE_RTHASH_WRITER_INSERT_HEAD(_sc, _hash, _brt) \ 209 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rthash[(_hash)], brt, brt_hash) 210 #define BRIDGE_RTHASH_WRITER_INSERT_AFTER(_brt, _new) \ 211 PSLIST_WRITER_INSERT_AFTER((_brt), (_new), brt_hash) 212 #define BRIDGE_RTHASH_WRITER_REMOVE(_brt) \ 213 PSLIST_WRITER_REMOVE((_brt), brt_hash) 214 215 #ifdef NET_MPSAFE 216 #define DECLARE_LOCK_VARIABLE 217 #define ACQUIRE_GLOBAL_LOCKS() do { } while (0) 218 #define RELEASE_GLOBAL_LOCKS() do { } while (0) 219 #else 220 #define DECLARE_LOCK_VARIABLE int __s 221 #define ACQUIRE_GLOBAL_LOCKS() do { \ 222 KERNEL_LOCK(1, NULL); \ 223 mutex_enter(softnet_lock); \ 224 __s = splsoftnet(); \ 225 } while (0) 226 #define RELEASE_GLOBAL_LOCKS() do { \ 227 splx(__s); \ 228 mutex_exit(softnet_lock); \ 229 KERNEL_UNLOCK_ONE(NULL); \ 230 } while (0) 231 #endif 232 233 struct psref_class *bridge_psref_class __read_mostly; 234 235 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD; 236 237 static struct pool bridge_rtnode_pool; 238 239 static int bridge_clone_create(struct if_clone *, int); 240 static int bridge_clone_destroy(struct ifnet *); 241 242 static int bridge_ioctl(struct ifnet *, u_long, void *); 243 static int bridge_init(struct ifnet *); 244 static void bridge_stop(struct ifnet *, int); 245 static void bridge_start(struct ifnet *); 246 static void bridge_ifdetach(void *); 247 248 static void bridge_input(struct ifnet *, struct mbuf *); 249 static void bridge_forward(struct bridge_softc *, struct mbuf *); 250 251 static void bridge_timer(void *); 252 253 static void bridge_broadcast(struct bridge_softc *, struct ifnet *, 254 struct mbuf *); 255 256 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *, 257 struct ifnet *, int, uint8_t); 258 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *); 259 static void bridge_rttrim(struct bridge_softc *); 260 static void bridge_rtage(struct bridge_softc *); 261 static void bridge_rtage_work(struct work *, void *); 262 static void bridge_rtflush(struct bridge_softc *, int); 263 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *); 264 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp); 265 266 static void bridge_rtable_init(struct bridge_softc *); 267 static void bridge_rtable_fini(struct bridge_softc *); 268 269 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *, 270 const uint8_t *); 271 static int bridge_rtnode_insert(struct bridge_softc *, 272 struct bridge_rtnode *); 273 static void bridge_rtnode_remove(struct bridge_softc *, 274 struct bridge_rtnode *); 275 static void bridge_rtnode_destroy(struct bridge_rtnode *); 276 277 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *, 278 const char *name, 279 struct psref *); 280 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *, 281 struct ifnet *ifp, 282 struct psref *); 283 static void bridge_release_member(struct bridge_softc *, struct bridge_iflist *, 284 struct psref *); 285 static void bridge_delete_member(struct bridge_softc *, 286 struct bridge_iflist *); 287 static void bridge_acquire_member(struct bridge_softc *sc, 288 struct bridge_iflist *, 289 struct psref *); 290 291 static int bridge_ioctl_add(struct bridge_softc *, void *); 292 static int bridge_ioctl_del(struct bridge_softc *, void *); 293 static int bridge_ioctl_gifflags(struct bridge_softc *, void *); 294 static int bridge_ioctl_sifflags(struct bridge_softc *, void *); 295 static int bridge_ioctl_scache(struct bridge_softc *, void *); 296 static int bridge_ioctl_gcache(struct bridge_softc *, void *); 297 static int bridge_ioctl_gifs(struct bridge_softc *, void *); 298 static int bridge_ioctl_rts(struct bridge_softc *, void *); 299 static int bridge_ioctl_saddr(struct bridge_softc *, void *); 300 static int bridge_ioctl_sto(struct bridge_softc *, void *); 301 static int bridge_ioctl_gto(struct bridge_softc *, void *); 302 static int bridge_ioctl_daddr(struct bridge_softc *, void *); 303 static int bridge_ioctl_flush(struct bridge_softc *, void *); 304 static int bridge_ioctl_gpri(struct bridge_softc *, void *); 305 static int bridge_ioctl_spri(struct bridge_softc *, void *); 306 static int bridge_ioctl_ght(struct bridge_softc *, void *); 307 static int bridge_ioctl_sht(struct bridge_softc *, void *); 308 static int bridge_ioctl_gfd(struct bridge_softc *, void *); 309 static int bridge_ioctl_sfd(struct bridge_softc *, void *); 310 static int bridge_ioctl_gma(struct bridge_softc *, void *); 311 static int bridge_ioctl_sma(struct bridge_softc *, void *); 312 static int bridge_ioctl_sifprio(struct bridge_softc *, void *); 313 static int bridge_ioctl_sifcost(struct bridge_softc *, void *); 314 static int bridge_ioctl_gfilt(struct bridge_softc *, void *); 315 static int bridge_ioctl_sfilt(struct bridge_softc *, void *); 316 static int bridge_ipf(void *, struct mbuf **, struct ifnet *, int); 317 static int bridge_ip_checkbasic(struct mbuf **mp); 318 # ifdef INET6 319 static int bridge_ip6_checkbasic(struct mbuf **mp); 320 # endif /* INET6 */ 321 322 struct bridge_control { 323 int (*bc_func)(struct bridge_softc *, void *); 324 int bc_argsize; 325 int bc_flags; 326 }; 327 328 #define BC_F_COPYIN 0x01 /* copy arguments in */ 329 #define BC_F_COPYOUT 0x02 /* copy arguments out */ 330 #define BC_F_SUSER 0x04 /* do super-user check */ 331 #define BC_F_XLATEIN 0x08 /* xlate arguments in */ 332 #define BC_F_XLATEOUT 0x10 /* xlate arguments out */ 333 334 static const struct bridge_control bridge_control_table[] = { 335 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 336 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 337 338 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT}, 339 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 340 341 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 342 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT}, 343 344 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT}, 345 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT}, 346 347 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER}, 348 349 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 350 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT}, 351 352 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER}, 353 354 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 355 356 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT}, 357 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 358 359 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT}, 360 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 361 362 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT}, 363 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 364 365 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT}, 366 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 367 368 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 369 370 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 371 372 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT}, 373 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 374 375 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT}, 376 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT}, 377 }; 378 379 static const int bridge_control_table_size = __arraycount(bridge_control_table); 380 381 static struct if_clone bridge_cloner = 382 IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy); 383 384 /* 385 * bridgeattach: 386 * 387 * Pseudo-device attach routine. 388 */ 389 void 390 bridgeattach(int n) 391 { 392 393 pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode), 394 0, 0, 0, "brtpl", NULL, IPL_NET); 395 396 bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET); 397 398 if_clone_attach(&bridge_cloner); 399 } 400 401 /* 402 * bridge_clone_create: 403 * 404 * Create a new bridge instance. 405 */ 406 static int 407 bridge_clone_create(struct if_clone *ifc, int unit) 408 { 409 struct bridge_softc *sc; 410 struct ifnet *ifp; 411 int error; 412 413 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP); 414 ifp = &sc->sc_if; 415 416 sc->sc_brtmax = BRIDGE_RTABLE_MAX; 417 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT; 418 sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE; 419 sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME; 420 sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY; 421 sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY; 422 sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME; 423 sc->sc_filter_flags = 0; 424 425 /* Initialize our routing table. */ 426 bridge_rtable_init(sc); 427 428 error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage", 429 bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE); 430 if (error) 431 panic("%s: workqueue_create %d\n", __func__, error); 432 433 callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE); 434 callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE); 435 436 mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE); 437 PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist); 438 sc->sc_iflist_psref.bip_psz = pserialize_create(); 439 440 if_initname(ifp, ifc->ifc_name, unit); 441 ifp->if_softc = sc; 442 #ifdef NET_MPSAFE 443 ifp->if_extflags = IFEF_MPSAFE; 444 #endif 445 ifp->if_mtu = ETHERMTU; 446 ifp->if_ioctl = bridge_ioctl; 447 ifp->if_output = bridge_output; 448 ifp->if_start = bridge_start; 449 ifp->if_stop = bridge_stop; 450 ifp->if_init = bridge_init; 451 ifp->if_type = IFT_BRIDGE; 452 ifp->if_addrlen = 0; 453 ifp->if_dlt = DLT_EN10MB; 454 ifp->if_hdrlen = ETHER_HDR_LEN; 455 if_initialize(ifp); 456 457 /* 458 * Set the link state to down. 459 * When interfaces are added the link state will reflect 460 * the best link state of the combined interfaces. 461 */ 462 ifp->if_link_state = LINK_STATE_DOWN; 463 464 if_alloc_sadl(ifp); 465 if_register(ifp); 466 467 return 0; 468 } 469 470 /* 471 * bridge_clone_destroy: 472 * 473 * Destroy a bridge instance. 474 */ 475 static int 476 bridge_clone_destroy(struct ifnet *ifp) 477 { 478 struct bridge_softc *sc = ifp->if_softc; 479 struct bridge_iflist *bif; 480 481 if ((ifp->if_flags & IFF_RUNNING) != 0) 482 bridge_stop(ifp, 1); 483 484 BRIDGE_LOCK(sc); 485 for (;;) { 486 bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist, 487 bif_next); 488 if (bif == NULL) 489 break; 490 bridge_delete_member(sc, bif); 491 } 492 PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist); 493 BRIDGE_UNLOCK(sc); 494 495 if_detach(ifp); 496 497 /* Tear down the routing table. */ 498 bridge_rtable_fini(sc); 499 500 pserialize_destroy(sc->sc_iflist_psref.bip_psz); 501 mutex_destroy(&sc->sc_iflist_psref.bip_lock); 502 callout_destroy(&sc->sc_brcallout); 503 callout_destroy(&sc->sc_bstpcallout); 504 workqueue_destroy(sc->sc_rtage_wq); 505 kmem_free(sc, sizeof(*sc)); 506 507 return 0; 508 } 509 510 /* 511 * bridge_ioctl: 512 * 513 * Handle a control request from the operator. 514 */ 515 static int 516 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data) 517 { 518 struct bridge_softc *sc = ifp->if_softc; 519 struct lwp *l = curlwp; /* XXX */ 520 union { 521 struct ifbreq ifbreq; 522 struct ifbifconf ifbifconf; 523 struct ifbareq ifbareq; 524 struct ifbaconf ifbaconf; 525 struct ifbrparam ifbrparam; 526 } args; 527 struct ifdrv *ifd = (struct ifdrv *) data; 528 const struct bridge_control *bc = NULL; /* XXXGCC */ 529 int error = 0; 530 531 /* Authorize command before calling splsoftnet(). */ 532 switch (cmd) { 533 case SIOCGDRVSPEC: 534 case SIOCSDRVSPEC: 535 if (ifd->ifd_cmd >= bridge_control_table_size 536 || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) { 537 error = EINVAL; 538 return error; 539 } 540 541 /* We only care about BC_F_SUSER at this point. */ 542 if ((bc->bc_flags & BC_F_SUSER) == 0) 543 break; 544 545 error = kauth_authorize_network(l->l_cred, 546 KAUTH_NETWORK_INTERFACE_BRIDGE, 547 cmd == SIOCGDRVSPEC ? 548 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV : 549 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV, 550 ifd, NULL, NULL); 551 if (error) 552 return error; 553 554 break; 555 } 556 557 const int s = splsoftnet(); 558 559 switch (cmd) { 560 case SIOCGDRVSPEC: 561 case SIOCSDRVSPEC: 562 KASSERT(bc != NULL); 563 if (cmd == SIOCGDRVSPEC && 564 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) { 565 error = EINVAL; 566 break; 567 } 568 else if (cmd == SIOCSDRVSPEC && 569 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) { 570 error = EINVAL; 571 break; 572 } 573 574 /* BC_F_SUSER is checked above, before splsoftnet(). */ 575 576 if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0 577 && (ifd->ifd_len != bc->bc_argsize 578 || ifd->ifd_len > sizeof(args))) { 579 error = EINVAL; 580 break; 581 } 582 583 memset(&args, 0, sizeof(args)); 584 if (bc->bc_flags & BC_F_COPYIN) { 585 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); 586 if (error) 587 break; 588 } else if (bc->bc_flags & BC_F_XLATEIN) { 589 args.ifbifconf.ifbic_len = ifd->ifd_len; 590 args.ifbifconf.ifbic_buf = ifd->ifd_data; 591 } 592 593 error = (*bc->bc_func)(sc, &args); 594 if (error) 595 break; 596 597 if (bc->bc_flags & BC_F_COPYOUT) { 598 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); 599 } else if (bc->bc_flags & BC_F_XLATEOUT) { 600 ifd->ifd_len = args.ifbifconf.ifbic_len; 601 ifd->ifd_data = args.ifbifconf.ifbic_buf; 602 } 603 break; 604 605 case SIOCSIFFLAGS: 606 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 607 break; 608 switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) { 609 case IFF_RUNNING: 610 /* 611 * If interface is marked down and it is running, 612 * then stop and disable it. 613 */ 614 if_stop(ifp, 1); 615 break; 616 case IFF_UP: 617 /* 618 * If interface is marked up and it is stopped, then 619 * start it. 620 */ 621 error = if_init(ifp); 622 break; 623 default: 624 break; 625 } 626 break; 627 628 case SIOCSIFMTU: 629 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET) 630 error = 0; 631 break; 632 633 case SIOCGIFCAP: 634 { 635 struct ifcapreq *ifcr = (struct ifcapreq *)data; 636 ifcr->ifcr_capabilities = sc->sc_capenable; 637 ifcr->ifcr_capenable = sc->sc_capenable; 638 break; 639 } 640 641 default: 642 error = ifioctl_common(ifp, cmd, data); 643 break; 644 } 645 646 splx(s); 647 648 return error; 649 } 650 651 /* 652 * bridge_lookup_member: 653 * 654 * Lookup a bridge member interface. 655 */ 656 static struct bridge_iflist * 657 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref) 658 { 659 struct bridge_iflist *bif; 660 struct ifnet *ifp; 661 int s; 662 663 BRIDGE_PSZ_RENTER(s); 664 665 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 666 ifp = bif->bif_ifp; 667 if (strcmp(ifp->if_xname, name) == 0) 668 break; 669 } 670 if (bif != NULL) 671 bridge_acquire_member(sc, bif, psref); 672 673 BRIDGE_PSZ_REXIT(s); 674 675 return bif; 676 } 677 678 /* 679 * bridge_lookup_member_if: 680 * 681 * Lookup a bridge member interface by ifnet*. 682 */ 683 static struct bridge_iflist * 684 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp, 685 struct psref *psref) 686 { 687 struct bridge_iflist *bif; 688 int s; 689 690 BRIDGE_PSZ_RENTER(s); 691 692 bif = member_ifp->if_bridgeif; 693 if (bif != NULL) { 694 psref_acquire(psref, &bif->bif_psref, 695 bridge_psref_class); 696 } 697 698 BRIDGE_PSZ_REXIT(s); 699 700 return bif; 701 } 702 703 static void 704 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif, 705 struct psref *psref) 706 { 707 708 psref_acquire(psref, &bif->bif_psref, bridge_psref_class); 709 } 710 711 /* 712 * bridge_release_member: 713 * 714 * Release the specified member interface. 715 */ 716 static void 717 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif, 718 struct psref *psref) 719 { 720 721 psref_release(psref, &bif->bif_psref, bridge_psref_class); 722 } 723 724 /* 725 * bridge_delete_member: 726 * 727 * Delete the specified member interface. 728 */ 729 static void 730 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif) 731 { 732 struct ifnet *ifs = bif->bif_ifp; 733 734 KASSERT(BRIDGE_LOCKED(sc)); 735 736 ifs->_if_input = ether_input; 737 ifs->if_bridge = NULL; 738 ifs->if_bridgeif = NULL; 739 740 PSLIST_WRITER_REMOVE(bif, bif_next); 741 BRIDGE_PSZ_PERFORM(sc); 742 743 if_linkstate_change_disestablish(ifs, 744 bif->bif_linkstate_hook, BRIDGE_LOCK_OBJ(sc)); 745 ether_ifdetachhook_disestablish(ifs, 746 bif->bif_ifdetach_hook, BRIDGE_LOCK_OBJ(sc)); 747 748 BRIDGE_UNLOCK(sc); 749 750 switch (ifs->if_type) { 751 case IFT_ETHER: 752 case IFT_L2TP: 753 /* 754 * Take the interface out of promiscuous mode. 755 * Don't call it with holding a spin lock. 756 */ 757 (void) ifpromisc(ifs, 0); 758 IFNET_LOCK(ifs); 759 (void) ether_disable_vlan_mtu(ifs); 760 IFNET_UNLOCK(ifs); 761 break; 762 default: 763 #ifdef DIAGNOSTIC 764 panic("%s: impossible", __func__); 765 #endif 766 break; 767 } 768 769 psref_target_destroy(&bif->bif_psref, bridge_psref_class); 770 771 PSLIST_ENTRY_DESTROY(bif, bif_next); 772 kmem_free(bif, sizeof(*bif)); 773 774 BRIDGE_LOCK(sc); 775 } 776 777 /* 778 * bridge_calc_csum_flags: 779 * 780 * Calculate logical and b/w csum flags each member interface supports. 781 */ 782 void 783 bridge_calc_csum_flags(struct bridge_softc *sc) 784 { 785 struct bridge_iflist *bif; 786 struct ifnet *ifs = NULL; 787 int flags = ~0; 788 int capenable = ~0; 789 790 BRIDGE_LOCK(sc); 791 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 792 ifs = bif->bif_ifp; 793 flags &= ifs->if_csum_flags_tx; 794 capenable &= ifs->if_capenable; 795 } 796 sc->sc_csum_flags_tx = flags; 797 sc->sc_capenable = (ifs != NULL) ? capenable : 0; 798 BRIDGE_UNLOCK(sc); 799 } 800 801 /* 802 * bridge_calc_link_state: 803 * 804 * Calculate the link state based on each member interface. 805 */ 806 static void 807 bridge_calc_link_state(void *xsc) 808 { 809 struct bridge_softc *sc = xsc; 810 struct bridge_iflist *bif; 811 struct ifnet *ifs; 812 int link_state = LINK_STATE_DOWN; 813 814 BRIDGE_LOCK(sc); 815 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 816 ifs = bif->bif_ifp; 817 if (ifs->if_link_state == LINK_STATE_UP) { 818 link_state = LINK_STATE_UP; 819 break; 820 } 821 if (ifs->if_link_state == LINK_STATE_UNKNOWN) 822 link_state = LINK_STATE_UNKNOWN; 823 } 824 if_link_state_change(&sc->sc_if, link_state); 825 BRIDGE_UNLOCK(sc); 826 } 827 828 static int 829 bridge_ioctl_add(struct bridge_softc *sc, void *arg) 830 { 831 struct ifbreq *req = arg; 832 struct bridge_iflist *bif = NULL; 833 struct ifnet *ifs; 834 int error = 0; 835 struct psref psref; 836 837 ifs = if_get(req->ifbr_ifsname, &psref); 838 if (ifs == NULL) 839 return ENOENT; 840 841 if (ifs->if_bridge == sc) { 842 error = EEXIST; 843 goto out; 844 } 845 846 if (ifs->if_bridge != NULL) { 847 error = EBUSY; 848 goto out; 849 } 850 851 if (ifs->_if_input != ether_input) { 852 error = EINVAL; 853 goto out; 854 } 855 856 /* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */ 857 if ((ifs->if_flags & IFF_SIMPLEX) == 0) { 858 error = EINVAL; 859 goto out; 860 } 861 862 bif = kmem_alloc(sizeof(*bif), KM_SLEEP); 863 864 switch (ifs->if_type) { 865 case IFT_ETHER: 866 if (sc->sc_if.if_mtu != ifs->if_mtu) { 867 /* Change MTU of added interface to bridge MTU */ 868 struct ifreq ifr; 869 memset(&ifr, 0, sizeof(ifr)); 870 ifr.ifr_mtu = sc->sc_if.if_mtu; 871 IFNET_LOCK(ifs); 872 error = if_ioctl(ifs, SIOCSIFMTU, &ifr); 873 IFNET_UNLOCK(ifs); 874 if (error != 0) 875 goto out; 876 } 877 /* FALLTHROUGH */ 878 case IFT_L2TP: 879 IFNET_LOCK(ifs); 880 error = ether_enable_vlan_mtu(ifs); 881 IFNET_UNLOCK(ifs); 882 if (error > 0) 883 goto out; 884 /* 885 * Place the interface into promiscuous mode. 886 */ 887 error = ifpromisc(ifs, 1); 888 if (error) 889 goto out; 890 break; 891 default: 892 error = EINVAL; 893 goto out; 894 } 895 896 bif->bif_ifp = ifs; 897 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER; 898 bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY; 899 bif->bif_path_cost = BSTP_DEFAULT_PATH_COST; 900 bif->bif_linkstate_hook = if_linkstate_change_establish(ifs, 901 bridge_calc_link_state, sc); 902 PSLIST_ENTRY_INIT(bif, bif_next); 903 psref_target_init(&bif->bif_psref, bridge_psref_class); 904 905 BRIDGE_LOCK(sc); 906 907 ifs->if_bridge = sc; 908 ifs->if_bridgeif = bif; 909 PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next); 910 ifs->_if_input = bridge_input; 911 912 BRIDGE_UNLOCK(sc); 913 914 bif->bif_ifdetach_hook = ether_ifdetachhook_establish(ifs, 915 bridge_ifdetach, (void *)ifs); 916 917 bridge_calc_csum_flags(sc); 918 bridge_calc_link_state(sc); 919 920 if (sc->sc_if.if_flags & IFF_RUNNING) 921 bstp_initialization(sc); 922 else 923 bstp_stop(sc); 924 925 out: 926 if_put(ifs, &psref); 927 if (error) { 928 if (bif != NULL) 929 kmem_free(bif, sizeof(*bif)); 930 } 931 return error; 932 } 933 934 static int 935 bridge_ioctl_del(struct bridge_softc *sc, void *arg) 936 { 937 struct ifbreq *req = arg; 938 const char *name = req->ifbr_ifsname; 939 struct bridge_iflist *bif; 940 struct ifnet *ifs; 941 942 BRIDGE_LOCK(sc); 943 944 /* 945 * Don't use bridge_lookup_member. We want to get a member 946 * with bif_refs == 0. 947 */ 948 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) { 949 ifs = bif->bif_ifp; 950 if (strcmp(ifs->if_xname, name) == 0) 951 break; 952 } 953 954 if (bif == NULL) { 955 BRIDGE_UNLOCK(sc); 956 return ENOENT; 957 } 958 959 bridge_delete_member(sc, bif); 960 961 BRIDGE_UNLOCK(sc); 962 963 bridge_rtdelete(sc, ifs); 964 bridge_calc_csum_flags(sc); 965 bridge_calc_link_state(sc); 966 967 if (sc->sc_if.if_flags & IFF_RUNNING) 968 bstp_initialization(sc); 969 970 return 0; 971 } 972 973 static int 974 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg) 975 { 976 struct ifbreq *req = arg; 977 struct bridge_iflist *bif; 978 struct psref psref; 979 980 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 981 if (bif == NULL) 982 return ENOENT; 983 984 req->ifbr_ifsflags = bif->bif_flags; 985 req->ifbr_state = bif->bif_state; 986 req->ifbr_priority = bif->bif_priority; 987 req->ifbr_path_cost = bif->bif_path_cost; 988 req->ifbr_portno = bif->bif_ifp->if_index & 0xff; 989 990 bridge_release_member(sc, bif, &psref); 991 992 return 0; 993 } 994 995 static int 996 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg) 997 { 998 struct ifbreq *req = arg; 999 struct bridge_iflist *bif; 1000 struct psref psref; 1001 1002 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1003 if (bif == NULL) 1004 return ENOENT; 1005 1006 if (req->ifbr_ifsflags & IFBIF_STP) { 1007 switch (bif->bif_ifp->if_type) { 1008 case IFT_ETHER: 1009 case IFT_L2TP: 1010 /* These can do spanning tree. */ 1011 break; 1012 1013 default: 1014 /* Nothing else can. */ 1015 bridge_release_member(sc, bif, &psref); 1016 return EINVAL; 1017 } 1018 } 1019 1020 bif->bif_flags = req->ifbr_ifsflags; 1021 1022 bridge_release_member(sc, bif, &psref); 1023 1024 if (sc->sc_if.if_flags & IFF_RUNNING) 1025 bstp_initialization(sc); 1026 1027 return 0; 1028 } 1029 1030 static int 1031 bridge_ioctl_scache(struct bridge_softc *sc, void *arg) 1032 { 1033 struct ifbrparam *param = arg; 1034 1035 sc->sc_brtmax = param->ifbrp_csize; 1036 bridge_rttrim(sc); 1037 1038 return 0; 1039 } 1040 1041 static int 1042 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg) 1043 { 1044 struct ifbrparam *param = arg; 1045 1046 param->ifbrp_csize = sc->sc_brtmax; 1047 1048 return 0; 1049 } 1050 1051 static int 1052 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg) 1053 { 1054 struct ifbifconf *bifc = arg; 1055 struct bridge_iflist *bif; 1056 struct ifbreq *breqs; 1057 int i, count, error = 0; 1058 1059 retry: 1060 BRIDGE_LOCK(sc); 1061 count = 0; 1062 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) 1063 count++; 1064 BRIDGE_UNLOCK(sc); 1065 1066 if (count == 0) { 1067 bifc->ifbic_len = 0; 1068 return 0; 1069 } 1070 1071 if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) { 1072 /* Tell that a larger buffer is needed */ 1073 bifc->ifbic_len = sizeof(*breqs) * count; 1074 return 0; 1075 } 1076 1077 breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP); 1078 1079 BRIDGE_LOCK(sc); 1080 1081 i = 0; 1082 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) 1083 i++; 1084 if (i > count) { 1085 /* 1086 * The number of members has been increased. 1087 * We need more memory! 1088 */ 1089 BRIDGE_UNLOCK(sc); 1090 kmem_free(breqs, sizeof(*breqs) * count); 1091 goto retry; 1092 } 1093 1094 i = 0; 1095 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) { 1096 struct ifbreq *breq = &breqs[i++]; 1097 memset(breq, 0, sizeof(*breq)); 1098 1099 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname, 1100 sizeof(breq->ifbr_ifsname)); 1101 breq->ifbr_ifsflags = bif->bif_flags; 1102 breq->ifbr_state = bif->bif_state; 1103 breq->ifbr_priority = bif->bif_priority; 1104 breq->ifbr_path_cost = bif->bif_path_cost; 1105 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff; 1106 } 1107 1108 /* Don't call copyout with holding the mutex */ 1109 BRIDGE_UNLOCK(sc); 1110 1111 for (i = 0; i < count; i++) { 1112 error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs)); 1113 if (error) 1114 break; 1115 } 1116 bifc->ifbic_len = sizeof(*breqs) * i; 1117 1118 kmem_free(breqs, sizeof(*breqs) * count); 1119 1120 return error; 1121 } 1122 1123 static int 1124 bridge_ioctl_rts(struct bridge_softc *sc, void *arg) 1125 { 1126 struct ifbaconf *bac = arg; 1127 struct bridge_rtnode *brt; 1128 struct ifbareq bareq; 1129 int count = 0, error = 0, len; 1130 1131 if (bac->ifbac_len == 0) 1132 return 0; 1133 1134 BRIDGE_RT_LOCK(sc); 1135 1136 /* The passed buffer is not enough, tell a required size. */ 1137 if (bac->ifbac_len < (sizeof(bareq) * sc->sc_brtcnt)) { 1138 count = sc->sc_brtcnt; 1139 goto out; 1140 } 1141 1142 len = bac->ifbac_len; 1143 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 1144 if (len < sizeof(bareq)) 1145 goto out; 1146 memset(&bareq, 0, sizeof(bareq)); 1147 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname, 1148 sizeof(bareq.ifba_ifsname)); 1149 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr)); 1150 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 1151 bareq.ifba_expire = brt->brt_expire - time_uptime; 1152 } else 1153 bareq.ifba_expire = 0; 1154 bareq.ifba_flags = brt->brt_flags; 1155 1156 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq)); 1157 if (error) 1158 goto out; 1159 count++; 1160 len -= sizeof(bareq); 1161 } 1162 out: 1163 BRIDGE_RT_UNLOCK(sc); 1164 1165 bac->ifbac_len = sizeof(bareq) * count; 1166 return error; 1167 } 1168 1169 static int 1170 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg) 1171 { 1172 struct ifbareq *req = arg; 1173 struct bridge_iflist *bif; 1174 int error; 1175 struct psref psref; 1176 1177 bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref); 1178 if (bif == NULL) 1179 return ENOENT; 1180 1181 error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1, 1182 req->ifba_flags); 1183 1184 bridge_release_member(sc, bif, &psref); 1185 1186 return error; 1187 } 1188 1189 static int 1190 bridge_ioctl_sto(struct bridge_softc *sc, void *arg) 1191 { 1192 struct ifbrparam *param = arg; 1193 1194 sc->sc_brttimeout = param->ifbrp_ctime; 1195 1196 return 0; 1197 } 1198 1199 static int 1200 bridge_ioctl_gto(struct bridge_softc *sc, void *arg) 1201 { 1202 struct ifbrparam *param = arg; 1203 1204 param->ifbrp_ctime = sc->sc_brttimeout; 1205 1206 return 0; 1207 } 1208 1209 static int 1210 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg) 1211 { 1212 struct ifbareq *req = arg; 1213 1214 return (bridge_rtdaddr(sc, req->ifba_dst)); 1215 } 1216 1217 static int 1218 bridge_ioctl_flush(struct bridge_softc *sc, void *arg) 1219 { 1220 struct ifbreq *req = arg; 1221 1222 bridge_rtflush(sc, req->ifbr_ifsflags); 1223 1224 return 0; 1225 } 1226 1227 static int 1228 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg) 1229 { 1230 struct ifbrparam *param = arg; 1231 1232 param->ifbrp_prio = sc->sc_bridge_priority; 1233 1234 return 0; 1235 } 1236 1237 static int 1238 bridge_ioctl_spri(struct bridge_softc *sc, void *arg) 1239 { 1240 struct ifbrparam *param = arg; 1241 1242 sc->sc_bridge_priority = param->ifbrp_prio; 1243 1244 if (sc->sc_if.if_flags & IFF_RUNNING) 1245 bstp_initialization(sc); 1246 1247 return 0; 1248 } 1249 1250 static int 1251 bridge_ioctl_ght(struct bridge_softc *sc, void *arg) 1252 { 1253 struct ifbrparam *param = arg; 1254 1255 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8; 1256 1257 return 0; 1258 } 1259 1260 static int 1261 bridge_ioctl_sht(struct bridge_softc *sc, void *arg) 1262 { 1263 struct ifbrparam *param = arg; 1264 1265 if (param->ifbrp_hellotime == 0) 1266 return EINVAL; 1267 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8; 1268 1269 if (sc->sc_if.if_flags & IFF_RUNNING) 1270 bstp_initialization(sc); 1271 1272 return 0; 1273 } 1274 1275 static int 1276 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg) 1277 { 1278 struct ifbrparam *param = arg; 1279 1280 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8; 1281 1282 return 0; 1283 } 1284 1285 static int 1286 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg) 1287 { 1288 struct ifbrparam *param = arg; 1289 1290 if (param->ifbrp_fwddelay == 0) 1291 return EINVAL; 1292 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8; 1293 1294 if (sc->sc_if.if_flags & IFF_RUNNING) 1295 bstp_initialization(sc); 1296 1297 return 0; 1298 } 1299 1300 static int 1301 bridge_ioctl_gma(struct bridge_softc *sc, void *arg) 1302 { 1303 struct ifbrparam *param = arg; 1304 1305 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8; 1306 1307 return 0; 1308 } 1309 1310 static int 1311 bridge_ioctl_sma(struct bridge_softc *sc, void *arg) 1312 { 1313 struct ifbrparam *param = arg; 1314 1315 if (param->ifbrp_maxage == 0) 1316 return EINVAL; 1317 sc->sc_bridge_max_age = param->ifbrp_maxage << 8; 1318 1319 if (sc->sc_if.if_flags & IFF_RUNNING) 1320 bstp_initialization(sc); 1321 1322 return 0; 1323 } 1324 1325 static int 1326 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg) 1327 { 1328 struct ifbreq *req = arg; 1329 struct bridge_iflist *bif; 1330 struct psref psref; 1331 1332 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1333 if (bif == NULL) 1334 return ENOENT; 1335 1336 bif->bif_priority = req->ifbr_priority; 1337 1338 if (sc->sc_if.if_flags & IFF_RUNNING) 1339 bstp_initialization(sc); 1340 1341 bridge_release_member(sc, bif, &psref); 1342 1343 return 0; 1344 } 1345 1346 static int 1347 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg) 1348 { 1349 struct ifbrparam *param = arg; 1350 1351 param->ifbrp_filter = sc->sc_filter_flags; 1352 1353 return 0; 1354 } 1355 1356 static int 1357 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg) 1358 { 1359 struct ifbrparam *param = arg; 1360 uint32_t nflags, oflags; 1361 1362 if (param->ifbrp_filter & ~IFBF_FILT_MASK) 1363 return EINVAL; 1364 1365 nflags = param->ifbrp_filter; 1366 oflags = sc->sc_filter_flags; 1367 1368 if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) { 1369 pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT, 1370 sc->sc_if.if_pfil); 1371 } 1372 if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) { 1373 pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT, 1374 sc->sc_if.if_pfil); 1375 } 1376 1377 sc->sc_filter_flags = nflags; 1378 1379 return 0; 1380 } 1381 1382 static int 1383 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg) 1384 { 1385 struct ifbreq *req = arg; 1386 struct bridge_iflist *bif; 1387 struct psref psref; 1388 1389 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1390 if (bif == NULL) 1391 return ENOENT; 1392 1393 bif->bif_path_cost = req->ifbr_path_cost; 1394 1395 if (sc->sc_if.if_flags & IFF_RUNNING) 1396 bstp_initialization(sc); 1397 1398 bridge_release_member(sc, bif, &psref); 1399 1400 return 0; 1401 } 1402 1403 /* 1404 * bridge_ifdetach: 1405 * 1406 * Detach an interface from a bridge. Called when a member 1407 * interface is detaching. 1408 */ 1409 static void 1410 bridge_ifdetach(void *xifs) 1411 { 1412 struct ifnet *ifs; 1413 struct bridge_softc *sc; 1414 struct ifbreq breq; 1415 1416 ifs = (struct ifnet *)xifs; 1417 sc = ifs->if_bridge; 1418 1419 /* ioctl_lock should prevent this from happening */ 1420 KASSERT(sc != NULL); 1421 1422 memset(&breq, 0, sizeof(breq)); 1423 strlcpy(breq.ifbr_ifsname, ifs->if_xname, sizeof(breq.ifbr_ifsname)); 1424 1425 (void) bridge_ioctl_del(sc, &breq); 1426 } 1427 1428 /* 1429 * bridge_init: 1430 * 1431 * Initialize a bridge interface. 1432 */ 1433 static int 1434 bridge_init(struct ifnet *ifp) 1435 { 1436 struct bridge_softc *sc = ifp->if_softc; 1437 1438 KASSERT((ifp->if_flags & IFF_RUNNING) == 0); 1439 1440 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz, 1441 bridge_timer, sc); 1442 bstp_initialization(sc); 1443 1444 ifp->if_flags |= IFF_RUNNING; 1445 return 0; 1446 } 1447 1448 /* 1449 * bridge_stop: 1450 * 1451 * Stop the bridge interface. 1452 */ 1453 static void 1454 bridge_stop(struct ifnet *ifp, int disable) 1455 { 1456 struct bridge_softc *sc = ifp->if_softc; 1457 1458 KASSERT((ifp->if_flags & IFF_RUNNING) != 0); 1459 ifp->if_flags &= ~IFF_RUNNING; 1460 1461 callout_halt(&sc->sc_brcallout, NULL); 1462 workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk); 1463 bstp_stop(sc); 1464 bridge_rtflush(sc, IFBF_FLUSHDYN); 1465 } 1466 1467 /* 1468 * bridge_enqueue: 1469 * 1470 * Enqueue a packet on a bridge member interface. 1471 */ 1472 void 1473 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m, 1474 int runfilt) 1475 { 1476 int len, error; 1477 short mflags; 1478 1479 if (runfilt) { 1480 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, 1481 dst_ifp, PFIL_OUT) != 0) { 1482 if (m != NULL) 1483 m_freem(m); 1484 return; 1485 } 1486 if (m == NULL) 1487 return; 1488 } 1489 1490 #ifdef ALTQ 1491 KERNEL_LOCK(1, NULL); 1492 /* 1493 * If ALTQ is enabled on the member interface, do 1494 * classification; the queueing discipline might 1495 * not require classification, but might require 1496 * the address family/header pointer in the pktattr. 1497 */ 1498 if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) { 1499 /* XXX IFT_ETHER */ 1500 altq_etherclassify(&dst_ifp->if_snd, m); 1501 } 1502 KERNEL_UNLOCK_ONE(NULL); 1503 #endif /* ALTQ */ 1504 1505 if (vlan_has_tag(m) && 1506 !vlan_is_hwtag_enabled(dst_ifp)) { 1507 (void)ether_inject_vlantag(&m, ETHERTYPE_VLAN, 1508 vlan_get_tag(m)); 1509 if (m == NULL) { 1510 if_statinc(&sc->sc_if, if_oerrors); 1511 return; 1512 } 1513 } 1514 1515 len = m->m_pkthdr.len; 1516 mflags = m->m_flags; 1517 1518 error = if_transmit_lock(dst_ifp, m); 1519 if (error) { 1520 /* mbuf is already freed */ 1521 if_statinc(&sc->sc_if, if_oerrors); 1522 return; 1523 } 1524 1525 net_stat_ref_t nsr = IF_STAT_GETREF(&sc->sc_if); 1526 if_statinc_ref(nsr, if_opackets); 1527 if_statadd_ref(nsr, if_obytes, len); 1528 if (mflags & M_MCAST) 1529 if_statinc_ref(nsr, if_omcasts); 1530 IF_STAT_PUTREF(&sc->sc_if); 1531 } 1532 1533 /* 1534 * bridge_output: 1535 * 1536 * Send output from a bridge member interface. This 1537 * performs the bridging function for locally originated 1538 * packets. 1539 * 1540 * The mbuf has the Ethernet header already attached. We must 1541 * enqueue or free the mbuf before returning. 1542 */ 1543 int 1544 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, 1545 const struct rtentry *rt) 1546 { 1547 struct ether_header *eh; 1548 struct ifnet *dst_if; 1549 struct bridge_softc *sc; 1550 struct mbuf *n; 1551 int s, bound; 1552 1553 /* 1554 * bridge_output() is called from ether_output(), furthermore 1555 * ifp argument doesn't point to bridge(4). So, don't assert 1556 * IFEF_MPSAFE here. 1557 */ 1558 1559 KASSERT(m->m_len >= ETHER_HDR_LEN); 1560 1561 eh = mtod(m, struct ether_header *); 1562 sc = ifp->if_bridge; 1563 1564 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1565 if (memcmp(etherbroadcastaddr, 1566 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 1567 m->m_flags |= M_BCAST; 1568 else 1569 m->m_flags |= M_MCAST; 1570 } 1571 1572 /* 1573 * If bridge is down, but the original output interface is up, 1574 * go ahead and send out that interface. Otherwise, the packet 1575 * is dropped below. 1576 */ 1577 if (__predict_false(sc == NULL) || 1578 (sc->sc_if.if_flags & IFF_RUNNING) == 0) { 1579 dst_if = ifp; 1580 goto unicast_asis; 1581 } 1582 1583 /* 1584 * If the packet is a multicast, or we don't know a better way to 1585 * get there, send to all interfaces. 1586 */ 1587 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0) 1588 dst_if = NULL; 1589 else 1590 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1591 1592 /* 1593 * In general, we need to handle TX offload in software before 1594 * enqueueing a packet. However, we can send it as is in the 1595 * cases of unicast via (1) the source interface, or (2) an 1596 * interface which supports the specified offload options. 1597 * For multicast or broadcast, send it as is only if (3) all 1598 * the member interfaces support the specified options. 1599 */ 1600 1601 /* 1602 * Unicast via the source interface. 1603 */ 1604 if (dst_if == ifp) 1605 goto unicast_asis; 1606 1607 /* 1608 * Unicast via other interface. 1609 */ 1610 if (dst_if != NULL) { 1611 KASSERT(m->m_flags & M_PKTHDR); 1612 if (TX_OFFLOAD_SUPPORTED(dst_if->if_csum_flags_tx, 1613 m->m_pkthdr.csum_flags)) { 1614 /* 1615 * Unicast via an interface which supports the 1616 * specified offload options. 1617 */ 1618 goto unicast_asis; 1619 } 1620 1621 /* 1622 * Handle TX offload in software. For TSO, a packet is 1623 * split into multiple chunks. Thus, the return value of 1624 * ether_sw_offload_tx() is mbuf queue consists of them. 1625 */ 1626 m = ether_sw_offload_tx(ifp, m); 1627 if (m == NULL) 1628 return 0; 1629 1630 do { 1631 n = m->m_nextpkt; 1632 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1633 m_freem(m); 1634 else 1635 bridge_enqueue(sc, dst_if, m, 0); 1636 m = n; 1637 } while (m != NULL); 1638 1639 return 0; 1640 } 1641 1642 /* 1643 * Multicast or broadcast. 1644 */ 1645 if (TX_OFFLOAD_SUPPORTED(sc->sc_csum_flags_tx, 1646 m->m_pkthdr.csum_flags)) { 1647 /* 1648 * Specified TX offload options are supported by all 1649 * the member interfaces of this bridge. 1650 */ 1651 m->m_nextpkt = NULL; /* XXX */ 1652 } else { 1653 /* 1654 * Otherwise, handle TX offload in software. 1655 */ 1656 m = ether_sw_offload_tx(ifp, m); 1657 if (m == NULL) 1658 return 0; 1659 } 1660 1661 /* 1662 * When we use pppoe over bridge, bridge_output() can be called 1663 * in a lwp context by pppoe_timeout_wk(). 1664 */ 1665 bound = curlwp_bind(); 1666 do { 1667 /* XXX Should call bridge_broadcast, but there are locking 1668 * issues which need resolving first. */ 1669 struct bridge_iflist *bif; 1670 struct mbuf *mc; 1671 bool used = false; 1672 1673 n = m->m_nextpkt; 1674 1675 BRIDGE_PSZ_RENTER(s); 1676 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 1677 struct psref psref; 1678 1679 bridge_acquire_member(sc, bif, &psref); 1680 BRIDGE_PSZ_REXIT(s); 1681 1682 dst_if = bif->bif_ifp; 1683 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1684 goto next; 1685 1686 /* 1687 * If this is not the original output interface, 1688 * and the interface is participating in spanning 1689 * tree, make sure the port is in a state that 1690 * allows forwarding. 1691 */ 1692 if (dst_if != ifp && 1693 (bif->bif_flags & IFBIF_STP) != 0) { 1694 switch (bif->bif_state) { 1695 case BSTP_IFSTATE_BLOCKING: 1696 case BSTP_IFSTATE_LISTENING: 1697 case BSTP_IFSTATE_DISABLED: 1698 goto next; 1699 } 1700 } 1701 1702 if (PSLIST_READER_NEXT(bif, struct bridge_iflist, 1703 bif_next) == NULL && 1704 ((m->m_flags & (M_MCAST | M_BCAST)) == 0 || 1705 dst_if == ifp)) 1706 { 1707 used = true; 1708 mc = m; 1709 } else { 1710 mc = m_copypacket(m, M_DONTWAIT); 1711 if (mc == NULL) { 1712 if_statinc(&sc->sc_if, if_oerrors); 1713 goto next; 1714 } 1715 } 1716 1717 bridge_enqueue(sc, dst_if, mc, 0); 1718 1719 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 && 1720 dst_if != ifp) 1721 { 1722 if (PSLIST_READER_NEXT(bif, 1723 struct bridge_iflist, bif_next) == NULL) 1724 { 1725 used = true; 1726 mc = m; 1727 } else { 1728 mc = m_copypacket(m, M_DONTWAIT); 1729 if (mc == NULL) { 1730 if_statinc(&sc->sc_if, 1731 if_oerrors); 1732 goto next; 1733 } 1734 } 1735 1736 m_set_rcvif(mc, dst_if); 1737 mc->m_flags &= ~M_PROMISC; 1738 1739 const int _s = splsoftnet(); 1740 KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if); 1741 ether_input(dst_if, mc); 1742 KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if); 1743 splx(_s); 1744 } 1745 1746 next: 1747 BRIDGE_PSZ_RENTER(s); 1748 bridge_release_member(sc, bif, &psref); 1749 1750 /* Guarantee we don't re-enter the loop as we already 1751 * decided we're at the end. */ 1752 if (used) 1753 break; 1754 } 1755 BRIDGE_PSZ_REXIT(s); 1756 1757 if (!used) 1758 m_freem(m); 1759 1760 m = n; 1761 } while (m != NULL); 1762 curlwp_bindx(bound); 1763 1764 return 0; 1765 1766 unicast_asis: 1767 /* 1768 * XXX Spanning tree consideration here? 1769 */ 1770 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1771 m_freem(m); 1772 else 1773 bridge_enqueue(sc, dst_if, m, 0); 1774 return 0; 1775 } 1776 1777 /* 1778 * bridge_start: 1779 * 1780 * Start output on a bridge. 1781 * 1782 * NOTE: This routine should never be called in this implementation. 1783 */ 1784 static void 1785 bridge_start(struct ifnet *ifp) 1786 { 1787 1788 printf("%s: bridge_start() called\n", ifp->if_xname); 1789 } 1790 1791 /* 1792 * bridge_forward: 1793 * 1794 * The forwarding function of the bridge. 1795 */ 1796 static void 1797 bridge_forward(struct bridge_softc *sc, struct mbuf *m) 1798 { 1799 struct bridge_iflist *bif; 1800 struct ifnet *src_if, *dst_if; 1801 struct ether_header *eh; 1802 struct psref psref; 1803 struct psref psref_src; 1804 DECLARE_LOCK_VARIABLE; 1805 1806 if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) 1807 return; 1808 1809 src_if = m_get_rcvif_psref(m, &psref_src); 1810 if (src_if == NULL) { 1811 /* Interface is being destroyed? */ 1812 m_freem(m); 1813 goto out; 1814 } 1815 1816 if_statadd2(&sc->sc_if, if_ipackets, 1, if_ibytes, m->m_pkthdr.len); 1817 1818 /* 1819 * Look up the bridge_iflist. 1820 */ 1821 bif = bridge_lookup_member_if(sc, src_if, &psref); 1822 if (bif == NULL) { 1823 /* Interface is not a bridge member (anymore?) */ 1824 m_freem(m); 1825 goto out; 1826 } 1827 1828 if (bif->bif_flags & IFBIF_STP) { 1829 switch (bif->bif_state) { 1830 case BSTP_IFSTATE_BLOCKING: 1831 case BSTP_IFSTATE_LISTENING: 1832 case BSTP_IFSTATE_DISABLED: 1833 m_freem(m); 1834 bridge_release_member(sc, bif, &psref); 1835 goto out; 1836 } 1837 } 1838 1839 eh = mtod(m, struct ether_header *); 1840 1841 /* 1842 * If the interface is learning, and the source 1843 * address is valid and not multicast, record 1844 * the address. 1845 */ 1846 if ((bif->bif_flags & IFBIF_LEARNING) != 0 && 1847 ETHER_IS_MULTICAST(eh->ether_shost) == 0 && 1848 (eh->ether_shost[0] == 0 && 1849 eh->ether_shost[1] == 0 && 1850 eh->ether_shost[2] == 0 && 1851 eh->ether_shost[3] == 0 && 1852 eh->ether_shost[4] == 0 && 1853 eh->ether_shost[5] == 0) == 0) { 1854 (void) bridge_rtupdate(sc, eh->ether_shost, 1855 src_if, 0, IFBAF_DYNAMIC); 1856 } 1857 1858 if ((bif->bif_flags & IFBIF_STP) != 0 && 1859 bif->bif_state == BSTP_IFSTATE_LEARNING) { 1860 m_freem(m); 1861 bridge_release_member(sc, bif, &psref); 1862 goto out; 1863 } 1864 1865 bridge_release_member(sc, bif, &psref); 1866 1867 /* 1868 * At this point, the port either doesn't participate 1869 * in spanning tree or it is in the forwarding state. 1870 */ 1871 1872 /* 1873 * If the packet is unicast, destined for someone on 1874 * "this" side of the bridge, drop it. 1875 */ 1876 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) { 1877 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1878 if (src_if == dst_if) { 1879 m_freem(m); 1880 goto out; 1881 } 1882 } else { 1883 /* ...forward it to all interfaces. */ 1884 if_statinc(&sc->sc_if, if_imcasts); 1885 dst_if = NULL; 1886 } 1887 1888 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) { 1889 if (m != NULL) 1890 m_freem(m); 1891 goto out; 1892 } 1893 if (m == NULL) 1894 goto out; 1895 1896 if (dst_if == NULL) { 1897 bridge_broadcast(sc, src_if, m); 1898 goto out; 1899 } 1900 1901 m_put_rcvif_psref(src_if, &psref_src); 1902 src_if = NULL; 1903 1904 /* 1905 * At this point, we're dealing with a unicast frame 1906 * going to a different interface. 1907 */ 1908 if ((dst_if->if_flags & IFF_RUNNING) == 0) { 1909 m_freem(m); 1910 goto out; 1911 } 1912 1913 bif = bridge_lookup_member_if(sc, dst_if, &psref); 1914 if (bif == NULL) { 1915 /* Not a member of the bridge (anymore?) */ 1916 m_freem(m); 1917 goto out; 1918 } 1919 1920 if (bif->bif_flags & IFBIF_STP) { 1921 switch (bif->bif_state) { 1922 case BSTP_IFSTATE_DISABLED: 1923 case BSTP_IFSTATE_BLOCKING: 1924 m_freem(m); 1925 bridge_release_member(sc, bif, &psref); 1926 goto out; 1927 } 1928 } 1929 1930 bridge_release_member(sc, bif, &psref); 1931 1932 /* 1933 * Before enqueueing this packet to the destination interface, 1934 * clear any in-bound checksum flags to prevent them from being 1935 * misused as out-bound flags. 1936 */ 1937 m->m_pkthdr.csum_flags = 0; 1938 1939 ACQUIRE_GLOBAL_LOCKS(); 1940 bridge_enqueue(sc, dst_if, m, 1); 1941 RELEASE_GLOBAL_LOCKS(); 1942 out: 1943 if (src_if != NULL) 1944 m_put_rcvif_psref(src_if, &psref_src); 1945 return; 1946 } 1947 1948 static bool 1949 bstp_state_before_learning(struct bridge_iflist *bif) 1950 { 1951 if (bif->bif_flags & IFBIF_STP) { 1952 switch (bif->bif_state) { 1953 case BSTP_IFSTATE_BLOCKING: 1954 case BSTP_IFSTATE_LISTENING: 1955 case BSTP_IFSTATE_DISABLED: 1956 return true; 1957 } 1958 } 1959 return false; 1960 } 1961 1962 static bool 1963 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src) 1964 { 1965 uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost; 1966 1967 if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0 1968 #if NCARP > 0 1969 || (bif->bif_ifp->if_carp && 1970 carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL) 1971 #endif /* NCARP > 0 */ 1972 ) 1973 return true; 1974 1975 return false; 1976 } 1977 1978 /* 1979 * bridge_input: 1980 * 1981 * Receive input from a member interface. Queue the packet for 1982 * bridging if it is not for us. 1983 */ 1984 static void 1985 bridge_input(struct ifnet *ifp, struct mbuf *m) 1986 { 1987 struct bridge_softc *sc = ifp->if_bridge; 1988 struct bridge_iflist *bif; 1989 struct ether_header *eh; 1990 struct psref psref; 1991 int bound; 1992 DECLARE_LOCK_VARIABLE; 1993 1994 KASSERT(!cpu_intr_p()); 1995 1996 if (__predict_false(sc == NULL) || 1997 (sc->sc_if.if_flags & IFF_RUNNING) == 0) { 1998 ACQUIRE_GLOBAL_LOCKS(); 1999 ether_input(ifp, m); 2000 RELEASE_GLOBAL_LOCKS(); 2001 return; 2002 } 2003 2004 bound = curlwp_bind(); 2005 bif = bridge_lookup_member_if(sc, ifp, &psref); 2006 if (bif == NULL) { 2007 curlwp_bindx(bound); 2008 ACQUIRE_GLOBAL_LOCKS(); 2009 ether_input(ifp, m); 2010 RELEASE_GLOBAL_LOCKS(); 2011 return; 2012 } 2013 2014 eh = mtod(m, struct ether_header *); 2015 2016 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 2017 if (memcmp(etherbroadcastaddr, 2018 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 2019 m->m_flags |= M_BCAST; 2020 else 2021 m->m_flags |= M_MCAST; 2022 } 2023 2024 /* 2025 * A 'fast' path for packets addressed to interfaces that are 2026 * part of this bridge. 2027 */ 2028 if (!(m->m_flags & (M_BCAST|M_MCAST)) && 2029 !bstp_state_before_learning(bif)) { 2030 struct bridge_iflist *_bif; 2031 struct ifnet *_ifp = NULL; 2032 int s; 2033 struct psref _psref; 2034 2035 BRIDGE_PSZ_RENTER(s); 2036 BRIDGE_IFLIST_READER_FOREACH(_bif, sc) { 2037 /* It is destined for us. */ 2038 if (bridge_ourether(_bif, eh, 0)) { 2039 bridge_acquire_member(sc, _bif, &_psref); 2040 BRIDGE_PSZ_REXIT(s); 2041 if (_bif->bif_flags & IFBIF_LEARNING) 2042 (void) bridge_rtupdate(sc, 2043 eh->ether_shost, ifp, 0, IFBAF_DYNAMIC); 2044 m_set_rcvif(m, _bif->bif_ifp); 2045 _ifp = _bif->bif_ifp; 2046 bridge_release_member(sc, _bif, &_psref); 2047 goto out; 2048 } 2049 2050 /* We just received a packet that we sent out. */ 2051 if (bridge_ourether(_bif, eh, 1)) 2052 break; 2053 } 2054 BRIDGE_PSZ_REXIT(s); 2055 out: 2056 2057 if (_bif != NULL) { 2058 bridge_release_member(sc, bif, &psref); 2059 curlwp_bindx(bound); 2060 if (_ifp != NULL) { 2061 m->m_flags &= ~M_PROMISC; 2062 ACQUIRE_GLOBAL_LOCKS(); 2063 ether_input(_ifp, m); 2064 RELEASE_GLOBAL_LOCKS(); 2065 } else 2066 m_freem(m); 2067 return; 2068 } 2069 } 2070 2071 /* Tap off 802.1D packets; they do not get forwarded. */ 2072 if (bif->bif_flags & IFBIF_STP && 2073 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) { 2074 bstp_input(sc, bif, m); 2075 bridge_release_member(sc, bif, &psref); 2076 curlwp_bindx(bound); 2077 return; 2078 } 2079 2080 /* 2081 * A normal switch would discard the packet here, but that's not what 2082 * we've done historically. This also prevents some obnoxious behaviour. 2083 */ 2084 if (bstp_state_before_learning(bif)) { 2085 bridge_release_member(sc, bif, &psref); 2086 curlwp_bindx(bound); 2087 ACQUIRE_GLOBAL_LOCKS(); 2088 ether_input(ifp, m); 2089 RELEASE_GLOBAL_LOCKS(); 2090 return; 2091 } 2092 2093 bridge_release_member(sc, bif, &psref); 2094 2095 bridge_forward(sc, m); 2096 2097 curlwp_bindx(bound); 2098 } 2099 2100 /* 2101 * bridge_broadcast: 2102 * 2103 * Send a frame to all interfaces that are members of 2104 * the bridge, except for the one on which the packet 2105 * arrived. 2106 */ 2107 static void 2108 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if, 2109 struct mbuf *m) 2110 { 2111 struct bridge_iflist *bif; 2112 struct mbuf *mc; 2113 struct ifnet *dst_if; 2114 bool bmcast; 2115 int s; 2116 DECLARE_LOCK_VARIABLE; 2117 2118 bmcast = m->m_flags & (M_BCAST|M_MCAST); 2119 2120 BRIDGE_PSZ_RENTER(s); 2121 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 2122 struct psref psref; 2123 2124 bridge_acquire_member(sc, bif, &psref); 2125 BRIDGE_PSZ_REXIT(s); 2126 2127 dst_if = bif->bif_ifp; 2128 2129 if (bif->bif_flags & IFBIF_STP) { 2130 switch (bif->bif_state) { 2131 case BSTP_IFSTATE_BLOCKING: 2132 case BSTP_IFSTATE_DISABLED: 2133 goto next; 2134 } 2135 } 2136 2137 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast) 2138 goto next; 2139 2140 if ((dst_if->if_flags & IFF_RUNNING) == 0) 2141 goto next; 2142 2143 if (dst_if != src_if) { 2144 mc = m_copypacket(m, M_DONTWAIT); 2145 if (mc == NULL) { 2146 if_statinc(&sc->sc_if, if_oerrors); 2147 goto next; 2148 } 2149 /* 2150 * Before enqueueing this packet to the destination 2151 * interface, clear any in-bound checksum flags to 2152 * prevent them from being misused as out-bound flags. 2153 */ 2154 mc->m_pkthdr.csum_flags = 0; 2155 2156 ACQUIRE_GLOBAL_LOCKS(); 2157 bridge_enqueue(sc, dst_if, mc, 1); 2158 RELEASE_GLOBAL_LOCKS(); 2159 } 2160 2161 if (bmcast) { 2162 mc = m_copypacket(m, M_DONTWAIT); 2163 if (mc == NULL) { 2164 if_statinc(&sc->sc_if, if_oerrors); 2165 goto next; 2166 } 2167 /* 2168 * Before enqueueing this packet to the destination 2169 * interface, clear any in-bound checksum flags to 2170 * prevent them from being misused as out-bound flags. 2171 */ 2172 mc->m_pkthdr.csum_flags = 0; 2173 2174 m_set_rcvif(mc, dst_if); 2175 mc->m_flags &= ~M_PROMISC; 2176 2177 ACQUIRE_GLOBAL_LOCKS(); 2178 ether_input(dst_if, mc); 2179 RELEASE_GLOBAL_LOCKS(); 2180 } 2181 next: 2182 BRIDGE_PSZ_RENTER(s); 2183 bridge_release_member(sc, bif, &psref); 2184 } 2185 BRIDGE_PSZ_REXIT(s); 2186 2187 m_freem(m); 2188 } 2189 2190 static int 2191 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst, 2192 struct bridge_rtnode **brtp) 2193 { 2194 struct bridge_rtnode *brt; 2195 int error; 2196 2197 if (sc->sc_brtcnt >= sc->sc_brtmax) 2198 return ENOSPC; 2199 2200 /* 2201 * Allocate a new bridge forwarding node, and 2202 * initialize the expiration time and Ethernet 2203 * address. 2204 */ 2205 brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT); 2206 if (brt == NULL) 2207 return ENOMEM; 2208 2209 memset(brt, 0, sizeof(*brt)); 2210 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2211 brt->brt_flags = IFBAF_DYNAMIC; 2212 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN); 2213 PSLIST_ENTRY_INIT(brt, brt_list); 2214 PSLIST_ENTRY_INIT(brt, brt_hash); 2215 2216 BRIDGE_RT_LOCK(sc); 2217 error = bridge_rtnode_insert(sc, brt); 2218 BRIDGE_RT_UNLOCK(sc); 2219 2220 if (error != 0) { 2221 pool_put(&bridge_rtnode_pool, brt); 2222 return error; 2223 } 2224 2225 *brtp = brt; 2226 return 0; 2227 } 2228 2229 /* 2230 * bridge_rtupdate: 2231 * 2232 * Add a bridge routing entry. 2233 */ 2234 static int 2235 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, 2236 struct ifnet *dst_if, int setflags, uint8_t flags) 2237 { 2238 struct bridge_rtnode *brt; 2239 int s; 2240 2241 again: 2242 /* 2243 * A route for this destination might already exist. If so, 2244 * update it, otherwise create a new one. 2245 */ 2246 BRIDGE_RT_RENTER(s); 2247 brt = bridge_rtnode_lookup(sc, dst); 2248 2249 if (brt != NULL) { 2250 brt->brt_ifp = dst_if; 2251 if (setflags) { 2252 brt->brt_flags = flags; 2253 if (flags & IFBAF_STATIC) 2254 brt->brt_expire = 0; 2255 else 2256 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2257 } else { 2258 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 2259 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2260 } 2261 } 2262 BRIDGE_RT_REXIT(s); 2263 2264 if (brt == NULL) { 2265 int r; 2266 2267 r = bridge_rtalloc(sc, dst, &brt); 2268 if (r != 0) 2269 return r; 2270 goto again; 2271 } 2272 2273 return 0; 2274 } 2275 2276 /* 2277 * bridge_rtlookup: 2278 * 2279 * Lookup the destination interface for an address. 2280 */ 2281 static struct ifnet * 2282 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr) 2283 { 2284 struct bridge_rtnode *brt; 2285 struct ifnet *ifs = NULL; 2286 int s; 2287 2288 BRIDGE_RT_RENTER(s); 2289 brt = bridge_rtnode_lookup(sc, addr); 2290 if (brt != NULL) 2291 ifs = brt->brt_ifp; 2292 BRIDGE_RT_REXIT(s); 2293 2294 return ifs; 2295 } 2296 2297 typedef bool (*bridge_iterate_cb_t) 2298 (struct bridge_softc *, struct bridge_rtnode *, bool *, void *); 2299 2300 /* 2301 * bridge_rtlist_iterate_remove: 2302 * 2303 * It iterates on sc->sc_rtlist and removes rtnodes of it which func 2304 * callback judges to remove. Removals of rtnodes are done in a manner 2305 * of pserialize. To this end, all kmem_* operations are placed out of 2306 * mutexes. 2307 */ 2308 static void 2309 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg) 2310 { 2311 struct bridge_rtnode *brt; 2312 struct bridge_rtnode **brt_list; 2313 int i, count; 2314 2315 retry: 2316 count = sc->sc_brtcnt; 2317 if (count == 0) 2318 return; 2319 brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP); 2320 2321 BRIDGE_RT_LOCK(sc); 2322 if (__predict_false(sc->sc_brtcnt > count)) { 2323 /* The rtnodes increased, we need more memory */ 2324 BRIDGE_RT_UNLOCK(sc); 2325 kmem_free(brt_list, sizeof(*brt_list) * count); 2326 goto retry; 2327 } 2328 2329 i = 0; 2330 /* 2331 * We don't need to use a _SAFE variant here because we know 2332 * that a removed item keeps its next pointer as-is thanks to 2333 * pslist(9) and isn't freed in the loop. 2334 */ 2335 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 2336 bool need_break = false; 2337 if (func(sc, brt, &need_break, arg)) { 2338 bridge_rtnode_remove(sc, brt); 2339 brt_list[i++] = brt; 2340 } 2341 if (need_break) 2342 break; 2343 } 2344 2345 if (i > 0) 2346 BRIDGE_RT_PSZ_PERFORM(sc); 2347 BRIDGE_RT_UNLOCK(sc); 2348 2349 while (--i >= 0) 2350 bridge_rtnode_destroy(brt_list[i]); 2351 2352 kmem_free(brt_list, sizeof(*brt_list) * count); 2353 } 2354 2355 static bool 2356 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2357 bool *need_break, void *arg) 2358 { 2359 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 2360 /* Take into account of the subsequent removal */ 2361 if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax) 2362 *need_break = true; 2363 return true; 2364 } else 2365 return false; 2366 } 2367 2368 static void 2369 bridge_rttrim0(struct bridge_softc *sc) 2370 { 2371 bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL); 2372 } 2373 2374 /* 2375 * bridge_rttrim: 2376 * 2377 * Trim the routine table so that we have a number 2378 * of routing entries less than or equal to the 2379 * maximum number. 2380 */ 2381 static void 2382 bridge_rttrim(struct bridge_softc *sc) 2383 { 2384 2385 /* Make sure we actually need to do this. */ 2386 if (sc->sc_brtcnt <= sc->sc_brtmax) 2387 return; 2388 2389 /* Force an aging cycle; this might trim enough addresses. */ 2390 bridge_rtage(sc); 2391 if (sc->sc_brtcnt <= sc->sc_brtmax) 2392 return; 2393 2394 bridge_rttrim0(sc); 2395 2396 return; 2397 } 2398 2399 /* 2400 * bridge_timer: 2401 * 2402 * Aging timer for the bridge. 2403 */ 2404 static void 2405 bridge_timer(void *arg) 2406 { 2407 struct bridge_softc *sc = arg; 2408 2409 workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL); 2410 } 2411 2412 static void 2413 bridge_rtage_work(struct work *wk, void *arg) 2414 { 2415 struct bridge_softc *sc = arg; 2416 2417 KASSERT(wk == &sc->sc_rtage_wk); 2418 2419 bridge_rtage(sc); 2420 2421 if (sc->sc_if.if_flags & IFF_RUNNING) 2422 callout_reset(&sc->sc_brcallout, 2423 bridge_rtable_prune_period * hz, bridge_timer, sc); 2424 } 2425 2426 static bool 2427 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2428 bool *need_break, void *arg) 2429 { 2430 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 2431 time_uptime >= brt->brt_expire) 2432 return true; 2433 else 2434 return false; 2435 } 2436 2437 /* 2438 * bridge_rtage: 2439 * 2440 * Perform an aging cycle. 2441 */ 2442 static void 2443 bridge_rtage(struct bridge_softc *sc) 2444 { 2445 bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL); 2446 } 2447 2448 2449 static bool 2450 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2451 bool *need_break, void *arg) 2452 { 2453 int full = *(int*)arg; 2454 2455 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 2456 return true; 2457 else 2458 return false; 2459 } 2460 2461 /* 2462 * bridge_rtflush: 2463 * 2464 * Remove all dynamic addresses from the bridge. 2465 */ 2466 static void 2467 bridge_rtflush(struct bridge_softc *sc, int full) 2468 { 2469 bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full); 2470 } 2471 2472 /* 2473 * bridge_rtdaddr: 2474 * 2475 * Remove an address from the table. 2476 */ 2477 static int 2478 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr) 2479 { 2480 struct bridge_rtnode *brt; 2481 2482 BRIDGE_RT_LOCK(sc); 2483 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) { 2484 BRIDGE_RT_UNLOCK(sc); 2485 return ENOENT; 2486 } 2487 bridge_rtnode_remove(sc, brt); 2488 BRIDGE_RT_PSZ_PERFORM(sc); 2489 BRIDGE_RT_UNLOCK(sc); 2490 2491 bridge_rtnode_destroy(brt); 2492 2493 return 0; 2494 } 2495 2496 /* 2497 * bridge_rtdelete: 2498 * 2499 * Delete routes to a speicifc member interface. 2500 */ 2501 static void 2502 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp) 2503 { 2504 struct bridge_rtnode *brt; 2505 2506 /* XXX pserialize_perform for each entry is slow */ 2507 again: 2508 BRIDGE_RT_LOCK(sc); 2509 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 2510 if (brt->brt_ifp == ifp) 2511 break; 2512 } 2513 if (brt == NULL) { 2514 BRIDGE_RT_UNLOCK(sc); 2515 return; 2516 } 2517 bridge_rtnode_remove(sc, brt); 2518 BRIDGE_RT_PSZ_PERFORM(sc); 2519 BRIDGE_RT_UNLOCK(sc); 2520 2521 bridge_rtnode_destroy(brt); 2522 2523 goto again; 2524 } 2525 2526 /* 2527 * bridge_rtable_init: 2528 * 2529 * Initialize the route table for this bridge. 2530 */ 2531 static void 2532 bridge_rtable_init(struct bridge_softc *sc) 2533 { 2534 int i; 2535 2536 sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE, 2537 KM_SLEEP); 2538 2539 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++) 2540 PSLIST_INIT(&sc->sc_rthash[i]); 2541 2542 sc->sc_rthash_key = cprng_fast32(); 2543 2544 PSLIST_INIT(&sc->sc_rtlist); 2545 2546 sc->sc_rtlist_psz = pserialize_create(); 2547 sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET); 2548 } 2549 2550 /* 2551 * bridge_rtable_fini: 2552 * 2553 * Deconstruct the route table for this bridge. 2554 */ 2555 static void 2556 bridge_rtable_fini(struct bridge_softc *sc) 2557 { 2558 2559 kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE); 2560 mutex_obj_free(sc->sc_rtlist_lock); 2561 pserialize_destroy(sc->sc_rtlist_psz); 2562 } 2563 2564 /* 2565 * The following hash function is adapted from "Hash Functions" by Bob Jenkins 2566 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997). 2567 */ 2568 #define mix(a, b, c) \ 2569 do { \ 2570 a -= b; a -= c; a ^= (c >> 13); \ 2571 b -= c; b -= a; b ^= (a << 8); \ 2572 c -= a; c -= b; c ^= (b >> 13); \ 2573 a -= b; a -= c; a ^= (c >> 12); \ 2574 b -= c; b -= a; b ^= (a << 16); \ 2575 c -= a; c -= b; c ^= (b >> 5); \ 2576 a -= b; a -= c; a ^= (c >> 3); \ 2577 b -= c; b -= a; b ^= (a << 10); \ 2578 c -= a; c -= b; c ^= (b >> 15); \ 2579 } while (/*CONSTCOND*/0) 2580 2581 static inline uint32_t 2582 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr) 2583 { 2584 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key; 2585 2586 b += addr[5] << 8; 2587 b += addr[4]; 2588 a += (uint32_t)addr[3] << 24; 2589 a += addr[2] << 16; 2590 a += addr[1] << 8; 2591 a += addr[0]; 2592 2593 mix(a, b, c); 2594 2595 return (c & BRIDGE_RTHASH_MASK); 2596 } 2597 2598 #undef mix 2599 2600 /* 2601 * bridge_rtnode_lookup: 2602 * 2603 * Look up a bridge route node for the specified destination. 2604 */ 2605 static struct bridge_rtnode * 2606 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr) 2607 { 2608 struct bridge_rtnode *brt; 2609 uint32_t hash; 2610 int dir; 2611 2612 hash = bridge_rthash(sc, addr); 2613 BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) { 2614 dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN); 2615 if (dir == 0) 2616 return brt; 2617 if (dir > 0) 2618 return NULL; 2619 } 2620 2621 return NULL; 2622 } 2623 2624 /* 2625 * bridge_rtnode_insert: 2626 * 2627 * Insert the specified bridge node into the route table. We 2628 * assume the entry is not already in the table. 2629 */ 2630 static int 2631 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt) 2632 { 2633 struct bridge_rtnode *lbrt, *prev = NULL; 2634 uint32_t hash; 2635 2636 KASSERT(BRIDGE_RT_LOCKED(sc)); 2637 2638 hash = bridge_rthash(sc, brt->brt_addr); 2639 BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) { 2640 int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN); 2641 if (dir == 0) 2642 return EEXIST; 2643 if (dir > 0) 2644 break; 2645 prev = lbrt; 2646 } 2647 if (prev == NULL) 2648 BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt); 2649 else 2650 BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt); 2651 2652 BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt); 2653 sc->sc_brtcnt++; 2654 2655 return 0; 2656 } 2657 2658 /* 2659 * bridge_rtnode_remove: 2660 * 2661 * Remove a bridge rtnode from the rthash and the rtlist of a bridge. 2662 */ 2663 static void 2664 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt) 2665 { 2666 2667 KASSERT(BRIDGE_RT_LOCKED(sc)); 2668 2669 BRIDGE_RTHASH_WRITER_REMOVE(brt); 2670 BRIDGE_RTLIST_WRITER_REMOVE(brt); 2671 sc->sc_brtcnt--; 2672 } 2673 2674 /* 2675 * bridge_rtnode_destroy: 2676 * 2677 * Destroy a bridge rtnode. 2678 */ 2679 static void 2680 bridge_rtnode_destroy(struct bridge_rtnode *brt) 2681 { 2682 2683 PSLIST_ENTRY_DESTROY(brt, brt_list); 2684 PSLIST_ENTRY_DESTROY(brt, brt_hash); 2685 pool_put(&bridge_rtnode_pool, brt); 2686 } 2687 2688 extern pfil_head_t *inet_pfil_hook; /* XXX */ 2689 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 2690 2691 /* 2692 * Send bridge packets through IPF if they are one of the types IPF can deal 2693 * with, or if they are ARP or REVARP. (IPF will pass ARP and REVARP without 2694 * question.) 2695 */ 2696 static int 2697 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir) 2698 { 2699 int snap, error; 2700 struct ether_header *eh1, eh2; 2701 struct llc llc1; 2702 uint16_t ether_type; 2703 2704 snap = 0; 2705 error = -1; /* Default error if not error == 0 */ 2706 eh1 = mtod(*mp, struct ether_header *); 2707 ether_type = ntohs(eh1->ether_type); 2708 2709 /* 2710 * Check for SNAP/LLC. 2711 */ 2712 if (ether_type < ETHERMTU) { 2713 struct llc *llc2 = (struct llc *)(eh1 + 1); 2714 2715 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 && 2716 llc2->llc_dsap == LLC_SNAP_LSAP && 2717 llc2->llc_ssap == LLC_SNAP_LSAP && 2718 llc2->llc_control == LLC_UI) { 2719 ether_type = htons(llc2->llc_un.type_snap.ether_type); 2720 snap = 1; 2721 } 2722 } 2723 2724 /* drop VLAN traffic untagged by hardware offloading */ 2725 if (vlan_has_tag(*mp)) 2726 goto bad; 2727 2728 /* 2729 * If we're trying to filter bridge traffic, don't look at anything 2730 * other than IP and ARP traffic. If the filter doesn't understand 2731 * IPv6, don't allow IPv6 through the bridge either. This is lame 2732 * since if we really wanted, say, an AppleTalk filter, we are hosed, 2733 * but of course we don't have an AppleTalk filter to begin with. 2734 * (Note that since IPF doesn't understand ARP it will pass *ALL* 2735 * ARP traffic.) 2736 */ 2737 switch (ether_type) { 2738 case ETHERTYPE_ARP: 2739 case ETHERTYPE_REVARP: 2740 return 0; /* Automatically pass */ 2741 case ETHERTYPE_IP: 2742 # ifdef INET6 2743 case ETHERTYPE_IPV6: 2744 # endif /* INET6 */ 2745 break; 2746 default: 2747 goto bad; 2748 } 2749 2750 /* Strip off the Ethernet header and keep a copy. */ 2751 m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2); 2752 m_adj(*mp, ETHER_HDR_LEN); 2753 2754 /* Strip off snap header, if present */ 2755 if (snap) { 2756 m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1); 2757 m_adj(*mp, sizeof(struct llc)); 2758 } 2759 2760 /* 2761 * Check basic packet sanity and run IPF through pfil. 2762 */ 2763 KASSERT(!cpu_intr_p()); 2764 switch (ether_type) 2765 { 2766 case ETHERTYPE_IP : 2767 error = bridge_ip_checkbasic(mp); 2768 if (error == 0) 2769 error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir); 2770 break; 2771 # ifdef INET6 2772 case ETHERTYPE_IPV6 : 2773 error = bridge_ip6_checkbasic(mp); 2774 if (error == 0) 2775 error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir); 2776 break; 2777 # endif 2778 default : 2779 error = 0; 2780 break; 2781 } 2782 2783 if (*mp == NULL) 2784 return error; 2785 if (error != 0) 2786 goto bad; 2787 2788 error = -1; 2789 2790 /* 2791 * Finally, put everything back the way it was and return 2792 */ 2793 if (snap) { 2794 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT); 2795 if (*mp == NULL) 2796 return error; 2797 bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc)); 2798 } 2799 2800 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT); 2801 if (*mp == NULL) 2802 return error; 2803 bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN); 2804 2805 return 0; 2806 2807 bad: 2808 m_freem(*mp); 2809 *mp = NULL; 2810 return error; 2811 } 2812 2813 /* 2814 * Perform basic checks on header size since 2815 * IPF assumes ip_input has already processed 2816 * it for it. Cut-and-pasted from ip_input.c. 2817 * Given how simple the IPv6 version is, 2818 * does the IPv4 version really need to be 2819 * this complicated? 2820 * 2821 * XXX Should we update ipstat here, or not? 2822 * XXX Right now we update ipstat but not 2823 * XXX csum_counter. 2824 */ 2825 static int 2826 bridge_ip_checkbasic(struct mbuf **mp) 2827 { 2828 struct mbuf *m = *mp; 2829 struct ip *ip; 2830 int len, hlen; 2831 2832 if (*mp == NULL) 2833 return -1; 2834 2835 if (M_GET_ALIGNED_HDR(&m, struct ip, true) != 0) { 2836 /* XXXJRT new stat, please */ 2837 ip_statinc(IP_STAT_TOOSMALL); 2838 goto bad; 2839 } 2840 ip = mtod(m, struct ip *); 2841 if (ip == NULL) goto bad; 2842 2843 if (ip->ip_v != IPVERSION) { 2844 ip_statinc(IP_STAT_BADVERS); 2845 goto bad; 2846 } 2847 hlen = ip->ip_hl << 2; 2848 if (hlen < sizeof(struct ip)) { /* minimum header length */ 2849 ip_statinc(IP_STAT_BADHLEN); 2850 goto bad; 2851 } 2852 if (hlen > m->m_len) { 2853 if ((m = m_pullup(m, hlen)) == 0) { 2854 ip_statinc(IP_STAT_BADHLEN); 2855 goto bad; 2856 } 2857 ip = mtod(m, struct ip *); 2858 if (ip == NULL) goto bad; 2859 } 2860 2861 switch (m->m_pkthdr.csum_flags & 2862 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) | 2863 M_CSUM_IPv4_BAD)) { 2864 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 2865 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */ 2866 goto bad; 2867 2868 case M_CSUM_IPv4: 2869 /* Checksum was okay. */ 2870 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */ 2871 break; 2872 2873 default: 2874 /* Must compute it ourselves. */ 2875 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */ 2876 if (in_cksum(m, hlen) != 0) 2877 goto bad; 2878 break; 2879 } 2880 2881 /* Retrieve the packet length. */ 2882 len = ntohs(ip->ip_len); 2883 2884 /* 2885 * Check for additional length bogosity 2886 */ 2887 if (len < hlen) { 2888 ip_statinc(IP_STAT_BADLEN); 2889 goto bad; 2890 } 2891 2892 /* 2893 * Check that the amount of data in the buffers 2894 * is as at least much as the IP header would have us expect. 2895 * Drop packet if shorter than we expect. 2896 */ 2897 if (m->m_pkthdr.len < len) { 2898 ip_statinc(IP_STAT_TOOSHORT); 2899 goto bad; 2900 } 2901 2902 /* Checks out, proceed */ 2903 *mp = m; 2904 return 0; 2905 2906 bad: 2907 *mp = m; 2908 return -1; 2909 } 2910 2911 # ifdef INET6 2912 /* 2913 * Same as above, but for IPv6. 2914 * Cut-and-pasted from ip6_input.c. 2915 * XXX Should we update ip6stat, or not? 2916 */ 2917 static int 2918 bridge_ip6_checkbasic(struct mbuf **mp) 2919 { 2920 struct mbuf *m = *mp; 2921 struct ip6_hdr *ip6; 2922 2923 /* 2924 * If the IPv6 header is not aligned, slurp it up into a new 2925 * mbuf with space for link headers, in the event we forward 2926 * it. Otherwise, if it is aligned, make sure the entire base 2927 * IPv6 header is in the first mbuf of the chain. 2928 */ 2929 if (M_GET_ALIGNED_HDR(&m, struct ip6_hdr, true) != 0) { 2930 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m); 2931 /* XXXJRT new stat, please */ 2932 ip6_statinc(IP6_STAT_TOOSMALL); 2933 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 2934 goto bad; 2935 } 2936 2937 ip6 = mtod(m, struct ip6_hdr *); 2938 2939 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 2940 ip6_statinc(IP6_STAT_BADVERS); 2941 in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr); 2942 goto bad; 2943 } 2944 2945 /* Checks out, proceed */ 2946 *mp = m; 2947 return 0; 2948 2949 bad: 2950 *mp = m; 2951 return -1; 2952 } 2953 # endif /* INET6 */ 2954
Indexes created Thu Oct 02 14:10:14 GMT 2025