1 /* $NetBSD: if_bridge.c,v 1.174 2020/08/01 06:50:43 maxv 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@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 1.174 2020/08/01 06:50:43 maxv 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 *); 246 247 static void bridge_input(struct ifnet *, struct mbuf *); 248 static void bridge_forward(struct bridge_softc *, struct mbuf *); 249 250 static void bridge_timer(void *); 251 252 static void bridge_broadcast(struct bridge_softc *, struct ifnet *, 253 struct mbuf *); 254 255 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *, 256 struct ifnet *, int, uint8_t); 257 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *); 258 static void bridge_rttrim(struct bridge_softc *); 259 static void bridge_rtage(struct bridge_softc *); 260 static void bridge_rtage_work(struct work *, void *); 261 static void bridge_rtflush(struct bridge_softc *, int); 262 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *); 263 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp); 264 265 static void bridge_rtable_init(struct bridge_softc *); 266 static void bridge_rtable_fini(struct bridge_softc *); 267 268 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *, 269 const uint8_t *); 270 static int bridge_rtnode_insert(struct bridge_softc *, 271 struct bridge_rtnode *); 272 static void bridge_rtnode_remove(struct bridge_softc *, 273 struct bridge_rtnode *); 274 static void bridge_rtnode_destroy(struct bridge_rtnode *); 275 276 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *, 277 const char *name, 278 struct psref *); 279 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *, 280 struct ifnet *ifp, 281 struct psref *); 282 static void bridge_release_member(struct bridge_softc *, struct bridge_iflist *, 283 struct psref *); 284 static void bridge_delete_member(struct bridge_softc *, 285 struct bridge_iflist *); 286 static void bridge_acquire_member(struct bridge_softc *sc, 287 struct bridge_iflist *, 288 struct psref *); 289 290 static int bridge_ioctl_add(struct bridge_softc *, void *); 291 static int bridge_ioctl_del(struct bridge_softc *, void *); 292 static int bridge_ioctl_gifflags(struct bridge_softc *, void *); 293 static int bridge_ioctl_sifflags(struct bridge_softc *, void *); 294 static int bridge_ioctl_scache(struct bridge_softc *, void *); 295 static int bridge_ioctl_gcache(struct bridge_softc *, void *); 296 static int bridge_ioctl_gifs(struct bridge_softc *, void *); 297 static int bridge_ioctl_rts(struct bridge_softc *, void *); 298 static int bridge_ioctl_saddr(struct bridge_softc *, void *); 299 static int bridge_ioctl_sto(struct bridge_softc *, void *); 300 static int bridge_ioctl_gto(struct bridge_softc *, void *); 301 static int bridge_ioctl_daddr(struct bridge_softc *, void *); 302 static int bridge_ioctl_flush(struct bridge_softc *, void *); 303 static int bridge_ioctl_gpri(struct bridge_softc *, void *); 304 static int bridge_ioctl_spri(struct bridge_softc *, void *); 305 static int bridge_ioctl_ght(struct bridge_softc *, void *); 306 static int bridge_ioctl_sht(struct bridge_softc *, void *); 307 static int bridge_ioctl_gfd(struct bridge_softc *, void *); 308 static int bridge_ioctl_sfd(struct bridge_softc *, void *); 309 static int bridge_ioctl_gma(struct bridge_softc *, void *); 310 static int bridge_ioctl_sma(struct bridge_softc *, void *); 311 static int bridge_ioctl_sifprio(struct bridge_softc *, void *); 312 static int bridge_ioctl_sifcost(struct bridge_softc *, void *); 313 static int bridge_ioctl_gfilt(struct bridge_softc *, void *); 314 static int bridge_ioctl_sfilt(struct bridge_softc *, void *); 315 static int bridge_ipf(void *, struct mbuf **, struct ifnet *, int); 316 static int bridge_ip_checkbasic(struct mbuf **mp); 317 # ifdef INET6 318 static int bridge_ip6_checkbasic(struct mbuf **mp); 319 # endif /* INET6 */ 320 321 struct bridge_control { 322 int (*bc_func)(struct bridge_softc *, void *); 323 int bc_argsize; 324 int bc_flags; 325 }; 326 327 #define BC_F_COPYIN 0x01 /* copy arguments in */ 328 #define BC_F_COPYOUT 0x02 /* copy arguments out */ 329 #define BC_F_SUSER 0x04 /* do super-user check */ 330 #define BC_F_XLATEIN 0x08 /* xlate arguments in */ 331 #define BC_F_XLATEOUT 0x10 /* xlate arguments out */ 332 333 static const struct bridge_control bridge_control_table[] = { 334 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 335 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 336 337 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT}, 338 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 339 340 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 341 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT}, 342 343 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT}, 344 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT}, 345 346 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER}, 347 348 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 349 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT}, 350 351 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER}, 352 353 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 354 355 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT}, 356 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 357 358 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT}, 359 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 360 361 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT}, 362 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 363 364 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT}, 365 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 366 367 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 368 369 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER}, 370 371 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT}, 372 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER}, 373 374 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT}, 375 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT}, 376 }; 377 378 static const int bridge_control_table_size = __arraycount(bridge_control_table); 379 380 static struct if_clone bridge_cloner = 381 IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy); 382 383 /* 384 * bridgeattach: 385 * 386 * Pseudo-device attach routine. 387 */ 388 void 389 bridgeattach(int n) 390 { 391 392 pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode), 393 0, 0, 0, "brtpl", NULL, IPL_NET); 394 395 bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET); 396 397 if_clone_attach(&bridge_cloner); 398 } 399 400 /* 401 * bridge_clone_create: 402 * 403 * Create a new bridge instance. 404 */ 405 static int 406 bridge_clone_create(struct if_clone *ifc, int unit) 407 { 408 struct bridge_softc *sc; 409 struct ifnet *ifp; 410 int error; 411 412 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP); 413 ifp = &sc->sc_if; 414 415 sc->sc_brtmax = BRIDGE_RTABLE_MAX; 416 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT; 417 sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE; 418 sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME; 419 sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY; 420 sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY; 421 sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME; 422 sc->sc_filter_flags = 0; 423 424 /* Initialize our routing table. */ 425 bridge_rtable_init(sc); 426 427 error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage", 428 bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE); 429 if (error) 430 panic("%s: workqueue_create %d\n", __func__, error); 431 432 callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE); 433 callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE); 434 435 mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE); 436 PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist); 437 sc->sc_iflist_psref.bip_psz = pserialize_create(); 438 439 if_initname(ifp, ifc->ifc_name, unit); 440 ifp->if_softc = sc; 441 ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE; 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; 455 456 error = if_initialize(ifp); 457 if (error != 0) { 458 pserialize_destroy(sc->sc_iflist_psref.bip_psz); 459 mutex_destroy(&sc->sc_iflist_psref.bip_lock); 460 callout_destroy(&sc->sc_brcallout); 461 callout_destroy(&sc->sc_bstpcallout); 462 workqueue_destroy(sc->sc_rtage_wq); 463 bridge_rtable_fini(sc); 464 kmem_free(sc, sizeof(*sc)); 465 466 return error; 467 } 468 if_alloc_sadl(ifp); 469 if_register(ifp); 470 471 return 0; 472 } 473 474 /* 475 * bridge_clone_destroy: 476 * 477 * Destroy a bridge instance. 478 */ 479 static int 480 bridge_clone_destroy(struct ifnet *ifp) 481 { 482 struct bridge_softc *sc = ifp->if_softc; 483 struct bridge_iflist *bif; 484 485 if ((ifp->if_flags & IFF_RUNNING) != 0) 486 bridge_stop(ifp, 1); 487 488 BRIDGE_LOCK(sc); 489 for (;;) { 490 bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist, 491 bif_next); 492 if (bif == NULL) 493 break; 494 bridge_delete_member(sc, bif); 495 } 496 PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist); 497 BRIDGE_UNLOCK(sc); 498 499 if_detach(ifp); 500 501 /* Tear down the routing table. */ 502 bridge_rtable_fini(sc); 503 504 pserialize_destroy(sc->sc_iflist_psref.bip_psz); 505 mutex_destroy(&sc->sc_iflist_psref.bip_lock); 506 callout_destroy(&sc->sc_brcallout); 507 callout_destroy(&sc->sc_bstpcallout); 508 workqueue_destroy(sc->sc_rtage_wq); 509 kmem_free(sc, sizeof(*sc)); 510 511 return 0; 512 } 513 514 /* 515 * bridge_ioctl: 516 * 517 * Handle a control request from the operator. 518 */ 519 static int 520 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data) 521 { 522 struct bridge_softc *sc = ifp->if_softc; 523 struct lwp *l = curlwp; /* XXX */ 524 union { 525 struct ifbreq ifbreq; 526 struct ifbifconf ifbifconf; 527 struct ifbareq ifbareq; 528 struct ifbaconf ifbaconf; 529 struct ifbrparam ifbrparam; 530 } args; 531 struct ifdrv *ifd = (struct ifdrv *) data; 532 const struct bridge_control *bc = NULL; /* XXXGCC */ 533 int s, error = 0; 534 535 /* Authorize command before calling splsoftnet(). */ 536 switch (cmd) { 537 case SIOCGDRVSPEC: 538 case SIOCSDRVSPEC: 539 if (ifd->ifd_cmd >= bridge_control_table_size 540 || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) { 541 error = EINVAL; 542 return error; 543 } 544 545 /* We only care about BC_F_SUSER at this point. */ 546 if ((bc->bc_flags & BC_F_SUSER) == 0) 547 break; 548 549 error = kauth_authorize_network(l->l_cred, 550 KAUTH_NETWORK_INTERFACE_BRIDGE, 551 cmd == SIOCGDRVSPEC ? 552 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV : 553 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV, 554 ifd, NULL, NULL); 555 if (error) 556 return error; 557 558 break; 559 } 560 561 s = splsoftnet(); 562 563 switch (cmd) { 564 case SIOCGDRVSPEC: 565 case SIOCSDRVSPEC: 566 KASSERT(bc != NULL); 567 if (cmd == SIOCGDRVSPEC && 568 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) { 569 error = EINVAL; 570 break; 571 } 572 else if (cmd == SIOCSDRVSPEC && 573 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) { 574 error = EINVAL; 575 break; 576 } 577 578 /* BC_F_SUSER is checked above, before splsoftnet(). */ 579 580 if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0 581 && (ifd->ifd_len != bc->bc_argsize 582 || ifd->ifd_len > sizeof(args))) { 583 error = EINVAL; 584 break; 585 } 586 587 memset(&args, 0, sizeof(args)); 588 if (bc->bc_flags & BC_F_COPYIN) { 589 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); 590 if (error) 591 break; 592 } else if (bc->bc_flags & BC_F_XLATEIN) { 593 args.ifbifconf.ifbic_len = ifd->ifd_len; 594 args.ifbifconf.ifbic_buf = ifd->ifd_data; 595 } 596 597 error = (*bc->bc_func)(sc, &args); 598 if (error) 599 break; 600 601 if (bc->bc_flags & BC_F_COPYOUT) { 602 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); 603 } else if (bc->bc_flags & BC_F_XLATEOUT) { 604 ifd->ifd_len = args.ifbifconf.ifbic_len; 605 ifd->ifd_data = args.ifbifconf.ifbic_buf; 606 } 607 break; 608 609 case SIOCSIFFLAGS: 610 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 611 break; 612 switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) { 613 case IFF_RUNNING: 614 /* 615 * If interface is marked down and it is running, 616 * then stop and disable it. 617 */ 618 (*ifp->if_stop)(ifp, 1); 619 break; 620 case IFF_UP: 621 /* 622 * If interface is marked up and it is stopped, then 623 * start it. 624 */ 625 error = (*ifp->if_init)(ifp); 626 break; 627 default: 628 break; 629 } 630 break; 631 632 case SIOCSIFMTU: 633 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET) 634 error = 0; 635 break; 636 637 case SIOCGIFCAP: 638 { 639 struct ifcapreq *ifcr = (struct ifcapreq *)data; 640 ifcr->ifcr_capabilities = sc->sc_capenable; 641 ifcr->ifcr_capenable = sc->sc_capenable; 642 break; 643 } 644 645 default: 646 error = ifioctl_common(ifp, cmd, data); 647 break; 648 } 649 650 splx(s); 651 652 return error; 653 } 654 655 /* 656 * bridge_lookup_member: 657 * 658 * Lookup a bridge member interface. 659 */ 660 static struct bridge_iflist * 661 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref) 662 { 663 struct bridge_iflist *bif; 664 struct ifnet *ifp; 665 int s; 666 667 BRIDGE_PSZ_RENTER(s); 668 669 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 670 ifp = bif->bif_ifp; 671 if (strcmp(ifp->if_xname, name) == 0) 672 break; 673 } 674 if (bif != NULL) 675 bridge_acquire_member(sc, bif, psref); 676 677 BRIDGE_PSZ_REXIT(s); 678 679 return bif; 680 } 681 682 /* 683 * bridge_lookup_member_if: 684 * 685 * Lookup a bridge member interface by ifnet*. 686 */ 687 static struct bridge_iflist * 688 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp, 689 struct psref *psref) 690 { 691 struct bridge_iflist *bif; 692 int s; 693 694 BRIDGE_PSZ_RENTER(s); 695 696 bif = member_ifp->if_bridgeif; 697 if (bif != NULL) { 698 psref_acquire(psref, &bif->bif_psref, 699 bridge_psref_class); 700 } 701 702 BRIDGE_PSZ_REXIT(s); 703 704 return bif; 705 } 706 707 static void 708 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif, 709 struct psref *psref) 710 { 711 712 psref_acquire(psref, &bif->bif_psref, bridge_psref_class); 713 } 714 715 /* 716 * bridge_release_member: 717 * 718 * Release the specified member interface. 719 */ 720 static void 721 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif, 722 struct psref *psref) 723 { 724 725 psref_release(psref, &bif->bif_psref, bridge_psref_class); 726 } 727 728 /* 729 * bridge_delete_member: 730 * 731 * Delete the specified member interface. 732 */ 733 static void 734 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif) 735 { 736 struct ifnet *ifs = bif->bif_ifp; 737 738 KASSERT(BRIDGE_LOCKED(sc)); 739 740 ifs->_if_input = ether_input; 741 ifs->if_bridge = NULL; 742 ifs->if_bridgeif = NULL; 743 744 PSLIST_WRITER_REMOVE(bif, bif_next); 745 BRIDGE_PSZ_PERFORM(sc); 746 BRIDGE_UNLOCK(sc); 747 748 switch (ifs->if_type) { 749 case IFT_ETHER: 750 case IFT_L2TP: 751 /* 752 * Take the interface out of promiscuous mode. 753 * Don't call it with holding a spin lock. 754 */ 755 (void) ifpromisc(ifs, 0); 756 IFNET_LOCK(ifs); 757 (void) ether_disable_vlan_mtu(ifs); 758 IFNET_UNLOCK(ifs); 759 break; 760 default: 761 #ifdef DIAGNOSTIC 762 panic("%s: impossible", __func__); 763 #endif 764 break; 765 } 766 767 psref_target_destroy(&bif->bif_psref, bridge_psref_class); 768 769 PSLIST_ENTRY_DESTROY(bif, bif_next); 770 kmem_free(bif, sizeof(*bif)); 771 772 BRIDGE_LOCK(sc); 773 } 774 775 /* 776 * bridge_calc_csum_flags: 777 * 778 * Calculate logical and b/w csum flags each member interface supports. 779 */ 780 void 781 bridge_calc_csum_flags(struct bridge_softc *sc) 782 { 783 struct bridge_iflist *bif; 784 struct ifnet *ifs = NULL; 785 int flags = ~0; 786 int capenable = ~0; 787 788 BRIDGE_LOCK(sc); 789 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 790 ifs = bif->bif_ifp; 791 flags &= ifs->if_csum_flags_tx; 792 capenable &= ifs->if_capenable; 793 } 794 sc->sc_csum_flags_tx = flags; 795 sc->sc_capenable = (ifs != NULL) ? capenable : 0; 796 BRIDGE_UNLOCK(sc); 797 } 798 799 static int 800 bridge_ioctl_add(struct bridge_softc *sc, void *arg) 801 { 802 struct ifbreq *req = arg; 803 struct bridge_iflist *bif = NULL; 804 struct ifnet *ifs; 805 int error = 0; 806 struct psref psref; 807 808 ifs = if_get(req->ifbr_ifsname, &psref); 809 if (ifs == NULL) 810 return ENOENT; 811 812 if (ifs->if_bridge == sc) { 813 error = EEXIST; 814 goto out; 815 } 816 817 if (ifs->if_bridge != NULL) { 818 error = EBUSY; 819 goto out; 820 } 821 822 if (ifs->_if_input != ether_input) { 823 error = EINVAL; 824 goto out; 825 } 826 827 /* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */ 828 if ((ifs->if_flags & IFF_SIMPLEX) == 0) { 829 error = EINVAL; 830 goto out; 831 } 832 833 bif = kmem_alloc(sizeof(*bif), KM_SLEEP); 834 835 switch (ifs->if_type) { 836 case IFT_ETHER: 837 if (sc->sc_if.if_mtu != ifs->if_mtu) { 838 /* Change MTU of added interface to bridge MTU */ 839 struct ifreq ifr; 840 memset(&ifr, 0, sizeof(ifr)); 841 ifr.ifr_mtu = sc->sc_if.if_mtu; 842 IFNET_LOCK(ifs); 843 error = ether_ioctl(ifs, SIOCSIFMTU, &ifr); 844 IFNET_UNLOCK(ifs); 845 if (error != 0) 846 goto out; 847 } 848 /* FALLTHROUGH */ 849 case IFT_L2TP: 850 IFNET_LOCK(ifs); 851 error = ether_enable_vlan_mtu(ifs); 852 IFNET_UNLOCK(ifs); 853 if (error > 0) 854 goto out; 855 /* 856 * Place the interface into promiscuous mode. 857 */ 858 error = ifpromisc(ifs, 1); 859 if (error) 860 goto out; 861 break; 862 default: 863 error = EINVAL; 864 goto out; 865 } 866 867 bif->bif_ifp = ifs; 868 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER; 869 bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY; 870 bif->bif_path_cost = BSTP_DEFAULT_PATH_COST; 871 PSLIST_ENTRY_INIT(bif, bif_next); 872 psref_target_init(&bif->bif_psref, bridge_psref_class); 873 874 BRIDGE_LOCK(sc); 875 876 ifs->if_bridge = sc; 877 ifs->if_bridgeif = bif; 878 PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next); 879 ifs->_if_input = bridge_input; 880 881 BRIDGE_UNLOCK(sc); 882 883 bridge_calc_csum_flags(sc); 884 885 if (sc->sc_if.if_flags & IFF_RUNNING) 886 bstp_initialization(sc); 887 else 888 bstp_stop(sc); 889 890 out: 891 if_put(ifs, &psref); 892 if (error) { 893 if (bif != NULL) 894 kmem_free(bif, sizeof(*bif)); 895 } 896 return error; 897 } 898 899 static int 900 bridge_ioctl_del(struct bridge_softc *sc, void *arg) 901 { 902 struct ifbreq *req = arg; 903 const char *name = req->ifbr_ifsname; 904 struct bridge_iflist *bif; 905 struct ifnet *ifs; 906 907 BRIDGE_LOCK(sc); 908 909 /* 910 * Don't use bridge_lookup_member. We want to get a member 911 * with bif_refs == 0. 912 */ 913 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) { 914 ifs = bif->bif_ifp; 915 if (strcmp(ifs->if_xname, name) == 0) 916 break; 917 } 918 919 if (bif == NULL) { 920 BRIDGE_UNLOCK(sc); 921 return ENOENT; 922 } 923 924 bridge_delete_member(sc, bif); 925 926 BRIDGE_UNLOCK(sc); 927 928 bridge_rtdelete(sc, ifs); 929 bridge_calc_csum_flags(sc); 930 931 if (sc->sc_if.if_flags & IFF_RUNNING) 932 bstp_initialization(sc); 933 934 return 0; 935 } 936 937 static int 938 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg) 939 { 940 struct ifbreq *req = arg; 941 struct bridge_iflist *bif; 942 struct psref psref; 943 944 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 945 if (bif == NULL) 946 return ENOENT; 947 948 req->ifbr_ifsflags = bif->bif_flags; 949 req->ifbr_state = bif->bif_state; 950 req->ifbr_priority = bif->bif_priority; 951 req->ifbr_path_cost = bif->bif_path_cost; 952 req->ifbr_portno = bif->bif_ifp->if_index & 0xff; 953 954 bridge_release_member(sc, bif, &psref); 955 956 return 0; 957 } 958 959 static int 960 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg) 961 { 962 struct ifbreq *req = arg; 963 struct bridge_iflist *bif; 964 struct psref psref; 965 966 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 967 if (bif == NULL) 968 return ENOENT; 969 970 if (req->ifbr_ifsflags & IFBIF_STP) { 971 switch (bif->bif_ifp->if_type) { 972 case IFT_ETHER: 973 case IFT_L2TP: 974 /* These can do spanning tree. */ 975 break; 976 977 default: 978 /* Nothing else can. */ 979 bridge_release_member(sc, bif, &psref); 980 return EINVAL; 981 } 982 } 983 984 bif->bif_flags = req->ifbr_ifsflags; 985 986 bridge_release_member(sc, bif, &psref); 987 988 if (sc->sc_if.if_flags & IFF_RUNNING) 989 bstp_initialization(sc); 990 991 return 0; 992 } 993 994 static int 995 bridge_ioctl_scache(struct bridge_softc *sc, void *arg) 996 { 997 struct ifbrparam *param = arg; 998 999 sc->sc_brtmax = param->ifbrp_csize; 1000 bridge_rttrim(sc); 1001 1002 return 0; 1003 } 1004 1005 static int 1006 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg) 1007 { 1008 struct ifbrparam *param = arg; 1009 1010 param->ifbrp_csize = sc->sc_brtmax; 1011 1012 return 0; 1013 } 1014 1015 static int 1016 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg) 1017 { 1018 struct ifbifconf *bifc = arg; 1019 struct bridge_iflist *bif; 1020 struct ifbreq *breqs; 1021 int i, count, error = 0; 1022 1023 retry: 1024 BRIDGE_LOCK(sc); 1025 count = 0; 1026 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) 1027 count++; 1028 BRIDGE_UNLOCK(sc); 1029 1030 if (count == 0) { 1031 bifc->ifbic_len = 0; 1032 return 0; 1033 } 1034 1035 if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) { 1036 /* Tell that a larger buffer is needed */ 1037 bifc->ifbic_len = sizeof(*breqs) * count; 1038 return 0; 1039 } 1040 1041 breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP); 1042 1043 BRIDGE_LOCK(sc); 1044 1045 i = 0; 1046 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) 1047 i++; 1048 if (i > count) { 1049 /* 1050 * The number of members has been increased. 1051 * We need more memory! 1052 */ 1053 BRIDGE_UNLOCK(sc); 1054 kmem_free(breqs, sizeof(*breqs) * count); 1055 goto retry; 1056 } 1057 1058 i = 0; 1059 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) { 1060 struct ifbreq *breq = &breqs[i++]; 1061 memset(breq, 0, sizeof(*breq)); 1062 1063 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname, 1064 sizeof(breq->ifbr_ifsname)); 1065 breq->ifbr_ifsflags = bif->bif_flags; 1066 breq->ifbr_state = bif->bif_state; 1067 breq->ifbr_priority = bif->bif_priority; 1068 breq->ifbr_path_cost = bif->bif_path_cost; 1069 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff; 1070 } 1071 1072 /* Don't call copyout with holding the mutex */ 1073 BRIDGE_UNLOCK(sc); 1074 1075 for (i = 0; i < count; i++) { 1076 error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs)); 1077 if (error) 1078 break; 1079 } 1080 bifc->ifbic_len = sizeof(*breqs) * i; 1081 1082 kmem_free(breqs, sizeof(*breqs) * count); 1083 1084 return error; 1085 } 1086 1087 static int 1088 bridge_ioctl_rts(struct bridge_softc *sc, void *arg) 1089 { 1090 struct ifbaconf *bac = arg; 1091 struct bridge_rtnode *brt; 1092 struct ifbareq bareq; 1093 int count = 0, error = 0, len; 1094 1095 if (bac->ifbac_len == 0) 1096 return 0; 1097 1098 BRIDGE_RT_LOCK(sc); 1099 1100 /* The passed buffer is not enough, tell a required size. */ 1101 if (bac->ifbac_len < (sizeof(bareq) * sc->sc_brtcnt)) { 1102 count = sc->sc_brtcnt; 1103 goto out; 1104 } 1105 1106 len = bac->ifbac_len; 1107 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 1108 if (len < sizeof(bareq)) 1109 goto out; 1110 memset(&bareq, 0, sizeof(bareq)); 1111 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname, 1112 sizeof(bareq.ifba_ifsname)); 1113 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr)); 1114 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 1115 bareq.ifba_expire = brt->brt_expire - time_uptime; 1116 } else 1117 bareq.ifba_expire = 0; 1118 bareq.ifba_flags = brt->brt_flags; 1119 1120 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq)); 1121 if (error) 1122 goto out; 1123 count++; 1124 len -= sizeof(bareq); 1125 } 1126 out: 1127 BRIDGE_RT_UNLOCK(sc); 1128 1129 bac->ifbac_len = sizeof(bareq) * count; 1130 return error; 1131 } 1132 1133 static int 1134 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg) 1135 { 1136 struct ifbareq *req = arg; 1137 struct bridge_iflist *bif; 1138 int error; 1139 struct psref psref; 1140 1141 bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref); 1142 if (bif == NULL) 1143 return ENOENT; 1144 1145 error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1, 1146 req->ifba_flags); 1147 1148 bridge_release_member(sc, bif, &psref); 1149 1150 return error; 1151 } 1152 1153 static int 1154 bridge_ioctl_sto(struct bridge_softc *sc, void *arg) 1155 { 1156 struct ifbrparam *param = arg; 1157 1158 sc->sc_brttimeout = param->ifbrp_ctime; 1159 1160 return 0; 1161 } 1162 1163 static int 1164 bridge_ioctl_gto(struct bridge_softc *sc, void *arg) 1165 { 1166 struct ifbrparam *param = arg; 1167 1168 param->ifbrp_ctime = sc->sc_brttimeout; 1169 1170 return 0; 1171 } 1172 1173 static int 1174 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg) 1175 { 1176 struct ifbareq *req = arg; 1177 1178 return (bridge_rtdaddr(sc, req->ifba_dst)); 1179 } 1180 1181 static int 1182 bridge_ioctl_flush(struct bridge_softc *sc, void *arg) 1183 { 1184 struct ifbreq *req = arg; 1185 1186 bridge_rtflush(sc, req->ifbr_ifsflags); 1187 1188 return 0; 1189 } 1190 1191 static int 1192 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg) 1193 { 1194 struct ifbrparam *param = arg; 1195 1196 param->ifbrp_prio = sc->sc_bridge_priority; 1197 1198 return 0; 1199 } 1200 1201 static int 1202 bridge_ioctl_spri(struct bridge_softc *sc, void *arg) 1203 { 1204 struct ifbrparam *param = arg; 1205 1206 sc->sc_bridge_priority = param->ifbrp_prio; 1207 1208 if (sc->sc_if.if_flags & IFF_RUNNING) 1209 bstp_initialization(sc); 1210 1211 return 0; 1212 } 1213 1214 static int 1215 bridge_ioctl_ght(struct bridge_softc *sc, void *arg) 1216 { 1217 struct ifbrparam *param = arg; 1218 1219 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8; 1220 1221 return 0; 1222 } 1223 1224 static int 1225 bridge_ioctl_sht(struct bridge_softc *sc, void *arg) 1226 { 1227 struct ifbrparam *param = arg; 1228 1229 if (param->ifbrp_hellotime == 0) 1230 return EINVAL; 1231 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8; 1232 1233 if (sc->sc_if.if_flags & IFF_RUNNING) 1234 bstp_initialization(sc); 1235 1236 return 0; 1237 } 1238 1239 static int 1240 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg) 1241 { 1242 struct ifbrparam *param = arg; 1243 1244 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8; 1245 1246 return 0; 1247 } 1248 1249 static int 1250 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg) 1251 { 1252 struct ifbrparam *param = arg; 1253 1254 if (param->ifbrp_fwddelay == 0) 1255 return EINVAL; 1256 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8; 1257 1258 if (sc->sc_if.if_flags & IFF_RUNNING) 1259 bstp_initialization(sc); 1260 1261 return 0; 1262 } 1263 1264 static int 1265 bridge_ioctl_gma(struct bridge_softc *sc, void *arg) 1266 { 1267 struct ifbrparam *param = arg; 1268 1269 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8; 1270 1271 return 0; 1272 } 1273 1274 static int 1275 bridge_ioctl_sma(struct bridge_softc *sc, void *arg) 1276 { 1277 struct ifbrparam *param = arg; 1278 1279 if (param->ifbrp_maxage == 0) 1280 return EINVAL; 1281 sc->sc_bridge_max_age = param->ifbrp_maxage << 8; 1282 1283 if (sc->sc_if.if_flags & IFF_RUNNING) 1284 bstp_initialization(sc); 1285 1286 return 0; 1287 } 1288 1289 static int 1290 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg) 1291 { 1292 struct ifbreq *req = arg; 1293 struct bridge_iflist *bif; 1294 struct psref psref; 1295 1296 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1297 if (bif == NULL) 1298 return ENOENT; 1299 1300 bif->bif_priority = req->ifbr_priority; 1301 1302 if (sc->sc_if.if_flags & IFF_RUNNING) 1303 bstp_initialization(sc); 1304 1305 bridge_release_member(sc, bif, &psref); 1306 1307 return 0; 1308 } 1309 1310 static int 1311 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg) 1312 { 1313 struct ifbrparam *param = arg; 1314 1315 param->ifbrp_filter = sc->sc_filter_flags; 1316 1317 return 0; 1318 } 1319 1320 static int 1321 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg) 1322 { 1323 struct ifbrparam *param = arg; 1324 uint32_t nflags, oflags; 1325 1326 if (param->ifbrp_filter & ~IFBF_FILT_MASK) 1327 return EINVAL; 1328 1329 nflags = param->ifbrp_filter; 1330 oflags = sc->sc_filter_flags; 1331 1332 if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) { 1333 pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT, 1334 sc->sc_if.if_pfil); 1335 } 1336 if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) { 1337 pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT, 1338 sc->sc_if.if_pfil); 1339 } 1340 1341 sc->sc_filter_flags = nflags; 1342 1343 return 0; 1344 } 1345 1346 static int 1347 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg) 1348 { 1349 struct ifbreq *req = arg; 1350 struct bridge_iflist *bif; 1351 struct psref psref; 1352 1353 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref); 1354 if (bif == NULL) 1355 return ENOENT; 1356 1357 bif->bif_path_cost = req->ifbr_path_cost; 1358 1359 if (sc->sc_if.if_flags & IFF_RUNNING) 1360 bstp_initialization(sc); 1361 1362 bridge_release_member(sc, bif, &psref); 1363 1364 return 0; 1365 } 1366 1367 /* 1368 * bridge_ifdetach: 1369 * 1370 * Detach an interface from a bridge. Called when a member 1371 * interface is detaching. 1372 */ 1373 void 1374 bridge_ifdetach(struct ifnet *ifp) 1375 { 1376 struct bridge_softc *sc = ifp->if_bridge; 1377 struct ifbreq breq; 1378 1379 /* ioctl_lock should prevent this from happening */ 1380 KASSERT(sc != NULL); 1381 1382 memset(&breq, 0, sizeof(breq)); 1383 strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname)); 1384 1385 (void) bridge_ioctl_del(sc, &breq); 1386 } 1387 1388 /* 1389 * bridge_init: 1390 * 1391 * Initialize a bridge interface. 1392 */ 1393 static int 1394 bridge_init(struct ifnet *ifp) 1395 { 1396 struct bridge_softc *sc = ifp->if_softc; 1397 1398 KASSERT((ifp->if_flags & IFF_RUNNING) == 0); 1399 1400 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz, 1401 bridge_timer, sc); 1402 bstp_initialization(sc); 1403 1404 ifp->if_flags |= IFF_RUNNING; 1405 return 0; 1406 } 1407 1408 /* 1409 * bridge_stop: 1410 * 1411 * Stop the bridge interface. 1412 */ 1413 static void 1414 bridge_stop(struct ifnet *ifp, int disable) 1415 { 1416 struct bridge_softc *sc = ifp->if_softc; 1417 1418 KASSERT((ifp->if_flags & IFF_RUNNING) != 0); 1419 ifp->if_flags &= ~IFF_RUNNING; 1420 1421 callout_halt(&sc->sc_brcallout, NULL); 1422 workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk); 1423 bstp_stop(sc); 1424 bridge_rtflush(sc, IFBF_FLUSHDYN); 1425 } 1426 1427 /* 1428 * bridge_enqueue: 1429 * 1430 * Enqueue a packet on a bridge member interface. 1431 */ 1432 void 1433 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m, 1434 int runfilt) 1435 { 1436 int len, error; 1437 short mflags; 1438 1439 if (runfilt) { 1440 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, 1441 dst_ifp, PFIL_OUT) != 0) { 1442 if (m != NULL) 1443 m_freem(m); 1444 return; 1445 } 1446 if (m == NULL) 1447 return; 1448 } 1449 1450 #ifdef ALTQ 1451 KERNEL_LOCK(1, NULL); 1452 /* 1453 * If ALTQ is enabled on the member interface, do 1454 * classification; the queueing discipline might 1455 * not require classification, but might require 1456 * the address family/header pointer in the pktattr. 1457 */ 1458 if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) { 1459 /* XXX IFT_ETHER */ 1460 altq_etherclassify(&dst_ifp->if_snd, m); 1461 } 1462 KERNEL_UNLOCK_ONE(NULL); 1463 #endif /* ALTQ */ 1464 1465 len = m->m_pkthdr.len; 1466 mflags = m->m_flags; 1467 1468 error = if_transmit_lock(dst_ifp, m); 1469 if (error) { 1470 /* mbuf is already freed */ 1471 if_statinc(&sc->sc_if, if_oerrors); 1472 return; 1473 } 1474 1475 net_stat_ref_t nsr = IF_STAT_GETREF(&sc->sc_if); 1476 if_statinc_ref(nsr, if_opackets); 1477 if_statadd_ref(nsr, if_obytes, len); 1478 if (mflags & M_MCAST) 1479 if_statinc_ref(nsr, if_omcasts); 1480 IF_STAT_PUTREF(&sc->sc_if); 1481 } 1482 1483 /* 1484 * bridge_output: 1485 * 1486 * Send output from a bridge member interface. This 1487 * performs the bridging function for locally originated 1488 * packets. 1489 * 1490 * The mbuf has the Ethernet header already attached. We must 1491 * enqueue or free the mbuf before returning. 1492 */ 1493 int 1494 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, 1495 const struct rtentry *rt) 1496 { 1497 struct ether_header *eh; 1498 struct ifnet *dst_if; 1499 struct bridge_softc *sc; 1500 struct mbuf *n; 1501 int s; 1502 1503 /* 1504 * bridge_output() is called from ether_output(), furthermore 1505 * ifp argument doesn't point to bridge(4). So, don't assert 1506 * IFEF_MPSAFE here. 1507 */ 1508 1509 KASSERT(m->m_len >= ETHER_HDR_LEN); 1510 1511 eh = mtod(m, struct ether_header *); 1512 sc = ifp->if_bridge; 1513 1514 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1515 if (memcmp(etherbroadcastaddr, 1516 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 1517 m->m_flags |= M_BCAST; 1518 else 1519 m->m_flags |= M_MCAST; 1520 } 1521 1522 /* 1523 * If bridge is down, but the original output interface is up, 1524 * go ahead and send out that interface. Otherwise, the packet 1525 * is dropped below. 1526 */ 1527 if (__predict_false(sc == NULL) || 1528 (sc->sc_if.if_flags & IFF_RUNNING) == 0) { 1529 dst_if = ifp; 1530 goto unicast_asis; 1531 } 1532 1533 /* 1534 * If the packet is a multicast, or we don't know a better way to 1535 * get there, send to all interfaces. 1536 */ 1537 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0) 1538 dst_if = NULL; 1539 else 1540 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1541 1542 /* 1543 * In general, we need to handle TX offload in software before 1544 * enqueueing a packet. However, we can send it as is in the 1545 * cases of unicast via (1) the source interface, or (2) an 1546 * interface which supports the specified offload options. 1547 * For multicast or broadcast, send it as is only if (3) all 1548 * the member interfaces support the specified options. 1549 */ 1550 1551 /* 1552 * Unicast via the source interface. 1553 */ 1554 if (dst_if == ifp) 1555 goto unicast_asis; 1556 1557 /* 1558 * Unicast via other interface. 1559 */ 1560 if (dst_if != NULL) { 1561 KASSERT(m->m_flags & M_PKTHDR); 1562 if (TX_OFFLOAD_SUPPORTED(dst_if->if_csum_flags_tx, 1563 m->m_pkthdr.csum_flags)) { 1564 /* 1565 * Unicast via an interface which supports the 1566 * specified offload options. 1567 */ 1568 goto unicast_asis; 1569 } 1570 1571 /* 1572 * Handle TX offload in software. For TSO, a packet is 1573 * split into multiple chunks. Thus, the return value of 1574 * ether_sw_offload_tx() is mbuf queue consists of them. 1575 */ 1576 m = ether_sw_offload_tx(ifp, m); 1577 if (m == NULL) 1578 return 0; 1579 1580 do { 1581 n = m->m_nextpkt; 1582 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1583 m_freem(m); 1584 else 1585 bridge_enqueue(sc, dst_if, m, 0); 1586 m = n; 1587 } while (m != NULL); 1588 1589 return 0; 1590 } 1591 1592 /* 1593 * Multicast or broadcast. 1594 */ 1595 if (TX_OFFLOAD_SUPPORTED(sc->sc_csum_flags_tx, 1596 m->m_pkthdr.csum_flags)) { 1597 /* 1598 * Specified TX offload options are supported by all 1599 * the member interfaces of this bridge. 1600 */ 1601 m->m_nextpkt = NULL; /* XXX */ 1602 } else { 1603 /* 1604 * Otherwise, handle TX offload in software. 1605 */ 1606 m = ether_sw_offload_tx(ifp, m); 1607 if (m == NULL) 1608 return 0; 1609 } 1610 1611 do { 1612 /* XXX Should call bridge_broadcast, but there are locking 1613 * issues which need resolving first. */ 1614 struct bridge_iflist *bif; 1615 struct mbuf *mc; 1616 bool used = false; 1617 1618 n = m->m_nextpkt; 1619 1620 BRIDGE_PSZ_RENTER(s); 1621 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 1622 struct psref psref; 1623 1624 bridge_acquire_member(sc, bif, &psref); 1625 BRIDGE_PSZ_REXIT(s); 1626 1627 dst_if = bif->bif_ifp; 1628 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1629 goto next; 1630 1631 /* 1632 * If this is not the original output interface, 1633 * and the interface is participating in spanning 1634 * tree, make sure the port is in a state that 1635 * allows forwarding. 1636 */ 1637 if (dst_if != ifp && 1638 (bif->bif_flags & IFBIF_STP) != 0) { 1639 switch (bif->bif_state) { 1640 case BSTP_IFSTATE_BLOCKING: 1641 case BSTP_IFSTATE_LISTENING: 1642 case BSTP_IFSTATE_DISABLED: 1643 goto next; 1644 } 1645 } 1646 1647 if (PSLIST_READER_NEXT(bif, struct bridge_iflist, 1648 bif_next) == NULL && 1649 ((m->m_flags & (M_MCAST | M_BCAST)) == 0 || 1650 dst_if == ifp)) 1651 { 1652 used = true; 1653 mc = m; 1654 } else { 1655 mc = m_copypacket(m, M_DONTWAIT); 1656 if (mc == NULL) { 1657 if_statinc(&sc->sc_if, if_oerrors); 1658 goto next; 1659 } 1660 } 1661 1662 bridge_enqueue(sc, dst_if, mc, 0); 1663 1664 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 && 1665 dst_if != ifp) 1666 { 1667 if (PSLIST_READER_NEXT(bif, 1668 struct bridge_iflist, bif_next) == NULL) 1669 { 1670 used = true; 1671 mc = m; 1672 } else { 1673 mc = m_copypacket(m, M_DONTWAIT); 1674 if (mc == NULL) { 1675 if_statinc(&sc->sc_if, 1676 if_oerrors); 1677 goto next; 1678 } 1679 } 1680 1681 m_set_rcvif(mc, dst_if); 1682 mc->m_flags &= ~M_PROMISC; 1683 1684 s = splsoftnet(); 1685 KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if); 1686 ether_input(dst_if, mc); 1687 KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if); 1688 splx(s); 1689 } 1690 1691 next: 1692 BRIDGE_PSZ_RENTER(s); 1693 bridge_release_member(sc, bif, &psref); 1694 1695 /* Guarantee we don't re-enter the loop as we already 1696 * decided we're at the end. */ 1697 if (used) 1698 break; 1699 } 1700 BRIDGE_PSZ_REXIT(s); 1701 1702 if (!used) 1703 m_freem(m); 1704 1705 m = n; 1706 } while (m != NULL); 1707 return 0; 1708 1709 unicast_asis: 1710 /* 1711 * XXX Spanning tree consideration here? 1712 */ 1713 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1714 m_freem(m); 1715 else 1716 bridge_enqueue(sc, dst_if, m, 0); 1717 return 0; 1718 } 1719 1720 /* 1721 * bridge_start: 1722 * 1723 * Start output on a bridge. 1724 * 1725 * NOTE: This routine should never be called in this implementation. 1726 */ 1727 static void 1728 bridge_start(struct ifnet *ifp) 1729 { 1730 1731 printf("%s: bridge_start() called\n", ifp->if_xname); 1732 } 1733 1734 /* 1735 * bridge_forward: 1736 * 1737 * The forwarding function of the bridge. 1738 */ 1739 static void 1740 bridge_forward(struct bridge_softc *sc, struct mbuf *m) 1741 { 1742 struct bridge_iflist *bif; 1743 struct ifnet *src_if, *dst_if; 1744 struct ether_header *eh; 1745 struct psref psref; 1746 struct psref psref_src; 1747 DECLARE_LOCK_VARIABLE; 1748 1749 if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) 1750 return; 1751 1752 src_if = m_get_rcvif_psref(m, &psref_src); 1753 if (src_if == NULL) { 1754 /* Interface is being destroyed? */ 1755 m_freem(m); 1756 goto out; 1757 } 1758 1759 if_statadd2(&sc->sc_if, if_ipackets, 1, if_ibytes, m->m_pkthdr.len); 1760 1761 /* 1762 * Look up the bridge_iflist. 1763 */ 1764 bif = bridge_lookup_member_if(sc, src_if, &psref); 1765 if (bif == NULL) { 1766 /* Interface is not a bridge member (anymore?) */ 1767 m_freem(m); 1768 goto out; 1769 } 1770 1771 if (bif->bif_flags & IFBIF_STP) { 1772 switch (bif->bif_state) { 1773 case BSTP_IFSTATE_BLOCKING: 1774 case BSTP_IFSTATE_LISTENING: 1775 case BSTP_IFSTATE_DISABLED: 1776 m_freem(m); 1777 bridge_release_member(sc, bif, &psref); 1778 goto out; 1779 } 1780 } 1781 1782 eh = mtod(m, struct ether_header *); 1783 1784 /* 1785 * If the interface is learning, and the source 1786 * address is valid and not multicast, record 1787 * the address. 1788 */ 1789 if ((bif->bif_flags & IFBIF_LEARNING) != 0 && 1790 ETHER_IS_MULTICAST(eh->ether_shost) == 0 && 1791 (eh->ether_shost[0] == 0 && 1792 eh->ether_shost[1] == 0 && 1793 eh->ether_shost[2] == 0 && 1794 eh->ether_shost[3] == 0 && 1795 eh->ether_shost[4] == 0 && 1796 eh->ether_shost[5] == 0) == 0) { 1797 (void) bridge_rtupdate(sc, eh->ether_shost, 1798 src_if, 0, IFBAF_DYNAMIC); 1799 } 1800 1801 if ((bif->bif_flags & IFBIF_STP) != 0 && 1802 bif->bif_state == BSTP_IFSTATE_LEARNING) { 1803 m_freem(m); 1804 bridge_release_member(sc, bif, &psref); 1805 goto out; 1806 } 1807 1808 bridge_release_member(sc, bif, &psref); 1809 1810 /* 1811 * At this point, the port either doesn't participate 1812 * in spanning tree or it is in the forwarding state. 1813 */ 1814 1815 /* 1816 * If the packet is unicast, destined for someone on 1817 * "this" side of the bridge, drop it. 1818 */ 1819 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) { 1820 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1821 if (src_if == dst_if) { 1822 m_freem(m); 1823 goto out; 1824 } 1825 } else { 1826 /* ...forward it to all interfaces. */ 1827 if_statinc(&sc->sc_if, if_imcasts); 1828 dst_if = NULL; 1829 } 1830 1831 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) { 1832 if (m != NULL) 1833 m_freem(m); 1834 goto out; 1835 } 1836 if (m == NULL) 1837 goto out; 1838 1839 if (dst_if == NULL) { 1840 bridge_broadcast(sc, src_if, m); 1841 goto out; 1842 } 1843 1844 m_put_rcvif_psref(src_if, &psref_src); 1845 src_if = NULL; 1846 1847 /* 1848 * At this point, we're dealing with a unicast frame 1849 * going to a different interface. 1850 */ 1851 if ((dst_if->if_flags & IFF_RUNNING) == 0) { 1852 m_freem(m); 1853 goto out; 1854 } 1855 1856 bif = bridge_lookup_member_if(sc, dst_if, &psref); 1857 if (bif == NULL) { 1858 /* Not a member of the bridge (anymore?) */ 1859 m_freem(m); 1860 goto out; 1861 } 1862 1863 if (bif->bif_flags & IFBIF_STP) { 1864 switch (bif->bif_state) { 1865 case BSTP_IFSTATE_DISABLED: 1866 case BSTP_IFSTATE_BLOCKING: 1867 m_freem(m); 1868 bridge_release_member(sc, bif, &psref); 1869 goto out; 1870 } 1871 } 1872 1873 bridge_release_member(sc, bif, &psref); 1874 1875 /* 1876 * Before enqueueing this packet to the destination interface, 1877 * clear any in-bound checksum flags to prevent them from being 1878 * misused as out-bound flags. 1879 */ 1880 m->m_pkthdr.csum_flags = 0; 1881 1882 ACQUIRE_GLOBAL_LOCKS(); 1883 bridge_enqueue(sc, dst_if, m, 1); 1884 RELEASE_GLOBAL_LOCKS(); 1885 out: 1886 if (src_if != NULL) 1887 m_put_rcvif_psref(src_if, &psref_src); 1888 return; 1889 } 1890 1891 static bool 1892 bstp_state_before_learning(struct bridge_iflist *bif) 1893 { 1894 if (bif->bif_flags & IFBIF_STP) { 1895 switch (bif->bif_state) { 1896 case BSTP_IFSTATE_BLOCKING: 1897 case BSTP_IFSTATE_LISTENING: 1898 case BSTP_IFSTATE_DISABLED: 1899 return true; 1900 } 1901 } 1902 return false; 1903 } 1904 1905 static bool 1906 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src) 1907 { 1908 uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost; 1909 1910 if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0 1911 #if NCARP > 0 1912 || (bif->bif_ifp->if_carp && 1913 carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL) 1914 #endif /* NCARP > 0 */ 1915 ) 1916 return true; 1917 1918 return false; 1919 } 1920 1921 /* 1922 * bridge_input: 1923 * 1924 * Receive input from a member interface. Queue the packet for 1925 * bridging if it is not for us. 1926 */ 1927 static void 1928 bridge_input(struct ifnet *ifp, struct mbuf *m) 1929 { 1930 struct bridge_softc *sc = ifp->if_bridge; 1931 struct bridge_iflist *bif; 1932 struct ether_header *eh; 1933 struct psref psref; 1934 int bound; 1935 DECLARE_LOCK_VARIABLE; 1936 1937 KASSERT(!cpu_intr_p()); 1938 1939 if (__predict_false(sc == NULL) || 1940 (sc->sc_if.if_flags & IFF_RUNNING) == 0) { 1941 ACQUIRE_GLOBAL_LOCKS(); 1942 ether_input(ifp, m); 1943 RELEASE_GLOBAL_LOCKS(); 1944 return; 1945 } 1946 1947 bound = curlwp_bind(); 1948 bif = bridge_lookup_member_if(sc, ifp, &psref); 1949 if (bif == NULL) { 1950 curlwp_bindx(bound); 1951 ACQUIRE_GLOBAL_LOCKS(); 1952 ether_input(ifp, m); 1953 RELEASE_GLOBAL_LOCKS(); 1954 return; 1955 } 1956 1957 eh = mtod(m, struct ether_header *); 1958 1959 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1960 if (memcmp(etherbroadcastaddr, 1961 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 1962 m->m_flags |= M_BCAST; 1963 else 1964 m->m_flags |= M_MCAST; 1965 } 1966 1967 /* 1968 * A 'fast' path for packets addressed to interfaces that are 1969 * part of this bridge. 1970 */ 1971 if (!(m->m_flags & (M_BCAST|M_MCAST)) && 1972 !bstp_state_before_learning(bif)) { 1973 struct bridge_iflist *_bif; 1974 struct ifnet *_ifp = NULL; 1975 int s; 1976 struct psref _psref; 1977 1978 BRIDGE_PSZ_RENTER(s); 1979 BRIDGE_IFLIST_READER_FOREACH(_bif, sc) { 1980 /* It is destined for us. */ 1981 if (bridge_ourether(_bif, eh, 0)) { 1982 bridge_acquire_member(sc, _bif, &_psref); 1983 BRIDGE_PSZ_REXIT(s); 1984 if (_bif->bif_flags & IFBIF_LEARNING) 1985 (void) bridge_rtupdate(sc, 1986 eh->ether_shost, ifp, 0, IFBAF_DYNAMIC); 1987 m_set_rcvif(m, _bif->bif_ifp); 1988 _ifp = _bif->bif_ifp; 1989 bridge_release_member(sc, _bif, &_psref); 1990 goto out; 1991 } 1992 1993 /* We just received a packet that we sent out. */ 1994 if (bridge_ourether(_bif, eh, 1)) 1995 break; 1996 } 1997 BRIDGE_PSZ_REXIT(s); 1998 out: 1999 2000 if (_bif != NULL) { 2001 bridge_release_member(sc, bif, &psref); 2002 curlwp_bindx(bound); 2003 if (_ifp != NULL) { 2004 m->m_flags &= ~M_PROMISC; 2005 ACQUIRE_GLOBAL_LOCKS(); 2006 ether_input(_ifp, m); 2007 RELEASE_GLOBAL_LOCKS(); 2008 } else 2009 m_freem(m); 2010 return; 2011 } 2012 } 2013 2014 /* Tap off 802.1D packets; they do not get forwarded. */ 2015 if (bif->bif_flags & IFBIF_STP && 2016 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) { 2017 bstp_input(sc, bif, m); 2018 bridge_release_member(sc, bif, &psref); 2019 curlwp_bindx(bound); 2020 return; 2021 } 2022 2023 /* 2024 * A normal switch would discard the packet here, but that's not what 2025 * we've done historically. This also prevents some obnoxious behaviour. 2026 */ 2027 if (bstp_state_before_learning(bif)) { 2028 bridge_release_member(sc, bif, &psref); 2029 curlwp_bindx(bound); 2030 ACQUIRE_GLOBAL_LOCKS(); 2031 ether_input(ifp, m); 2032 RELEASE_GLOBAL_LOCKS(); 2033 return; 2034 } 2035 2036 bridge_release_member(sc, bif, &psref); 2037 2038 bridge_forward(sc, m); 2039 2040 curlwp_bindx(bound); 2041 } 2042 2043 /* 2044 * bridge_broadcast: 2045 * 2046 * Send a frame to all interfaces that are members of 2047 * the bridge, except for the one on which the packet 2048 * arrived. 2049 */ 2050 static void 2051 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if, 2052 struct mbuf *m) 2053 { 2054 struct bridge_iflist *bif; 2055 struct mbuf *mc; 2056 struct ifnet *dst_if; 2057 bool bmcast; 2058 int s; 2059 DECLARE_LOCK_VARIABLE; 2060 2061 bmcast = m->m_flags & (M_BCAST|M_MCAST); 2062 2063 BRIDGE_PSZ_RENTER(s); 2064 BRIDGE_IFLIST_READER_FOREACH(bif, sc) { 2065 struct psref psref; 2066 2067 bridge_acquire_member(sc, bif, &psref); 2068 BRIDGE_PSZ_REXIT(s); 2069 2070 dst_if = bif->bif_ifp; 2071 2072 if (bif->bif_flags & IFBIF_STP) { 2073 switch (bif->bif_state) { 2074 case BSTP_IFSTATE_BLOCKING: 2075 case BSTP_IFSTATE_DISABLED: 2076 goto next; 2077 } 2078 } 2079 2080 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast) 2081 goto next; 2082 2083 if ((dst_if->if_flags & IFF_RUNNING) == 0) 2084 goto next; 2085 2086 if (dst_if != src_if) { 2087 mc = m_copypacket(m, M_DONTWAIT); 2088 if (mc == NULL) { 2089 if_statinc(&sc->sc_if, if_oerrors); 2090 goto next; 2091 } 2092 /* 2093 * Before enqueueing this packet to the destination 2094 * interface, clear any in-bound checksum flags to 2095 * prevent them from being misused as out-bound flags. 2096 */ 2097 mc->m_pkthdr.csum_flags = 0; 2098 2099 ACQUIRE_GLOBAL_LOCKS(); 2100 bridge_enqueue(sc, dst_if, mc, 1); 2101 RELEASE_GLOBAL_LOCKS(); 2102 } 2103 2104 if (bmcast) { 2105 mc = m_copypacket(m, M_DONTWAIT); 2106 if (mc == NULL) { 2107 if_statinc(&sc->sc_if, if_oerrors); 2108 goto next; 2109 } 2110 /* 2111 * Before enqueueing this packet to the destination 2112 * interface, clear any in-bound checksum flags to 2113 * prevent them from being misused as out-bound flags. 2114 */ 2115 mc->m_pkthdr.csum_flags = 0; 2116 2117 m_set_rcvif(mc, dst_if); 2118 mc->m_flags &= ~M_PROMISC; 2119 2120 ACQUIRE_GLOBAL_LOCKS(); 2121 ether_input(dst_if, mc); 2122 RELEASE_GLOBAL_LOCKS(); 2123 } 2124 next: 2125 BRIDGE_PSZ_RENTER(s); 2126 bridge_release_member(sc, bif, &psref); 2127 } 2128 BRIDGE_PSZ_REXIT(s); 2129 2130 m_freem(m); 2131 } 2132 2133 static int 2134 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst, 2135 struct bridge_rtnode **brtp) 2136 { 2137 struct bridge_rtnode *brt; 2138 int error; 2139 2140 if (sc->sc_brtcnt >= sc->sc_brtmax) 2141 return ENOSPC; 2142 2143 /* 2144 * Allocate a new bridge forwarding node, and 2145 * initialize the expiration time and Ethernet 2146 * address. 2147 */ 2148 brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT); 2149 if (brt == NULL) 2150 return ENOMEM; 2151 2152 memset(brt, 0, sizeof(*brt)); 2153 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2154 brt->brt_flags = IFBAF_DYNAMIC; 2155 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN); 2156 PSLIST_ENTRY_INIT(brt, brt_list); 2157 PSLIST_ENTRY_INIT(brt, brt_hash); 2158 2159 BRIDGE_RT_LOCK(sc); 2160 error = bridge_rtnode_insert(sc, brt); 2161 BRIDGE_RT_UNLOCK(sc); 2162 2163 if (error != 0) { 2164 pool_put(&bridge_rtnode_pool, brt); 2165 return error; 2166 } 2167 2168 *brtp = brt; 2169 return 0; 2170 } 2171 2172 /* 2173 * bridge_rtupdate: 2174 * 2175 * Add a bridge routing entry. 2176 */ 2177 static int 2178 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, 2179 struct ifnet *dst_if, int setflags, uint8_t flags) 2180 { 2181 struct bridge_rtnode *brt; 2182 int s; 2183 2184 again: 2185 /* 2186 * A route for this destination might already exist. If so, 2187 * update it, otherwise create a new one. 2188 */ 2189 BRIDGE_RT_RENTER(s); 2190 brt = bridge_rtnode_lookup(sc, dst); 2191 2192 if (brt != NULL) { 2193 brt->brt_ifp = dst_if; 2194 if (setflags) { 2195 brt->brt_flags = flags; 2196 if (flags & IFBAF_STATIC) 2197 brt->brt_expire = 0; 2198 else 2199 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2200 } else { 2201 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 2202 brt->brt_expire = time_uptime + sc->sc_brttimeout; 2203 } 2204 } 2205 BRIDGE_RT_REXIT(s); 2206 2207 if (brt == NULL) { 2208 int r; 2209 2210 r = bridge_rtalloc(sc, dst, &brt); 2211 if (r != 0) 2212 return r; 2213 goto again; 2214 } 2215 2216 return 0; 2217 } 2218 2219 /* 2220 * bridge_rtlookup: 2221 * 2222 * Lookup the destination interface for an address. 2223 */ 2224 static struct ifnet * 2225 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr) 2226 { 2227 struct bridge_rtnode *brt; 2228 struct ifnet *ifs = NULL; 2229 int s; 2230 2231 BRIDGE_RT_RENTER(s); 2232 brt = bridge_rtnode_lookup(sc, addr); 2233 if (brt != NULL) 2234 ifs = brt->brt_ifp; 2235 BRIDGE_RT_REXIT(s); 2236 2237 return ifs; 2238 } 2239 2240 typedef bool (*bridge_iterate_cb_t) 2241 (struct bridge_softc *, struct bridge_rtnode *, bool *, void *); 2242 2243 /* 2244 * bridge_rtlist_iterate_remove: 2245 * 2246 * It iterates on sc->sc_rtlist and removes rtnodes of it which func 2247 * callback judges to remove. Removals of rtnodes are done in a manner 2248 * of pserialize. To this end, all kmem_* operations are placed out of 2249 * mutexes. 2250 */ 2251 static void 2252 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg) 2253 { 2254 struct bridge_rtnode *brt; 2255 struct bridge_rtnode **brt_list; 2256 int i, count; 2257 2258 retry: 2259 count = sc->sc_brtcnt; 2260 if (count == 0) 2261 return; 2262 brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP); 2263 2264 BRIDGE_RT_LOCK(sc); 2265 if (__predict_false(sc->sc_brtcnt > count)) { 2266 /* The rtnodes increased, we need more memory */ 2267 BRIDGE_RT_UNLOCK(sc); 2268 kmem_free(brt_list, sizeof(*brt_list) * count); 2269 goto retry; 2270 } 2271 2272 i = 0; 2273 /* 2274 * We don't need to use a _SAFE variant here because we know 2275 * that a removed item keeps its next pointer as-is thanks to 2276 * pslist(9) and isn't freed in the loop. 2277 */ 2278 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 2279 bool need_break = false; 2280 if (func(sc, brt, &need_break, arg)) { 2281 bridge_rtnode_remove(sc, brt); 2282 brt_list[i++] = brt; 2283 } 2284 if (need_break) 2285 break; 2286 } 2287 2288 if (i > 0) 2289 BRIDGE_RT_PSZ_PERFORM(sc); 2290 BRIDGE_RT_UNLOCK(sc); 2291 2292 while (--i >= 0) 2293 bridge_rtnode_destroy(brt_list[i]); 2294 2295 kmem_free(brt_list, sizeof(*brt_list) * count); 2296 } 2297 2298 static bool 2299 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2300 bool *need_break, void *arg) 2301 { 2302 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 2303 /* Take into account of the subsequent removal */ 2304 if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax) 2305 *need_break = true; 2306 return true; 2307 } else 2308 return false; 2309 } 2310 2311 static void 2312 bridge_rttrim0(struct bridge_softc *sc) 2313 { 2314 bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL); 2315 } 2316 2317 /* 2318 * bridge_rttrim: 2319 * 2320 * Trim the routine table so that we have a number 2321 * of routing entries less than or equal to the 2322 * maximum number. 2323 */ 2324 static void 2325 bridge_rttrim(struct bridge_softc *sc) 2326 { 2327 2328 /* Make sure we actually need to do this. */ 2329 if (sc->sc_brtcnt <= sc->sc_brtmax) 2330 return; 2331 2332 /* Force an aging cycle; this might trim enough addresses. */ 2333 bridge_rtage(sc); 2334 if (sc->sc_brtcnt <= sc->sc_brtmax) 2335 return; 2336 2337 bridge_rttrim0(sc); 2338 2339 return; 2340 } 2341 2342 /* 2343 * bridge_timer: 2344 * 2345 * Aging timer for the bridge. 2346 */ 2347 static void 2348 bridge_timer(void *arg) 2349 { 2350 struct bridge_softc *sc = arg; 2351 2352 workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL); 2353 } 2354 2355 static void 2356 bridge_rtage_work(struct work *wk, void *arg) 2357 { 2358 struct bridge_softc *sc = arg; 2359 2360 KASSERT(wk == &sc->sc_rtage_wk); 2361 2362 bridge_rtage(sc); 2363 2364 if (sc->sc_if.if_flags & IFF_RUNNING) 2365 callout_reset(&sc->sc_brcallout, 2366 bridge_rtable_prune_period * hz, bridge_timer, sc); 2367 } 2368 2369 static bool 2370 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2371 bool *need_break, void *arg) 2372 { 2373 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 2374 time_uptime >= brt->brt_expire) 2375 return true; 2376 else 2377 return false; 2378 } 2379 2380 /* 2381 * bridge_rtage: 2382 * 2383 * Perform an aging cycle. 2384 */ 2385 static void 2386 bridge_rtage(struct bridge_softc *sc) 2387 { 2388 bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL); 2389 } 2390 2391 2392 static bool 2393 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt, 2394 bool *need_break, void *arg) 2395 { 2396 int full = *(int*)arg; 2397 2398 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 2399 return true; 2400 else 2401 return false; 2402 } 2403 2404 /* 2405 * bridge_rtflush: 2406 * 2407 * Remove all dynamic addresses from the bridge. 2408 */ 2409 static void 2410 bridge_rtflush(struct bridge_softc *sc, int full) 2411 { 2412 bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full); 2413 } 2414 2415 /* 2416 * bridge_rtdaddr: 2417 * 2418 * Remove an address from the table. 2419 */ 2420 static int 2421 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr) 2422 { 2423 struct bridge_rtnode *brt; 2424 2425 BRIDGE_RT_LOCK(sc); 2426 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) { 2427 BRIDGE_RT_UNLOCK(sc); 2428 return ENOENT; 2429 } 2430 bridge_rtnode_remove(sc, brt); 2431 BRIDGE_RT_PSZ_PERFORM(sc); 2432 BRIDGE_RT_UNLOCK(sc); 2433 2434 bridge_rtnode_destroy(brt); 2435 2436 return 0; 2437 } 2438 2439 /* 2440 * bridge_rtdelete: 2441 * 2442 * Delete routes to a speicifc member interface. 2443 */ 2444 static void 2445 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp) 2446 { 2447 struct bridge_rtnode *brt; 2448 2449 /* XXX pserialize_perform for each entry is slow */ 2450 again: 2451 BRIDGE_RT_LOCK(sc); 2452 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) { 2453 if (brt->brt_ifp == ifp) 2454 break; 2455 } 2456 if (brt == NULL) { 2457 BRIDGE_RT_UNLOCK(sc); 2458 return; 2459 } 2460 bridge_rtnode_remove(sc, brt); 2461 BRIDGE_RT_PSZ_PERFORM(sc); 2462 BRIDGE_RT_UNLOCK(sc); 2463 2464 bridge_rtnode_destroy(brt); 2465 2466 goto again; 2467 } 2468 2469 /* 2470 * bridge_rtable_init: 2471 * 2472 * Initialize the route table for this bridge. 2473 */ 2474 static void 2475 bridge_rtable_init(struct bridge_softc *sc) 2476 { 2477 int i; 2478 2479 sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE, 2480 KM_SLEEP); 2481 2482 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++) 2483 PSLIST_INIT(&sc->sc_rthash[i]); 2484 2485 sc->sc_rthash_key = cprng_fast32(); 2486 2487 PSLIST_INIT(&sc->sc_rtlist); 2488 2489 sc->sc_rtlist_psz = pserialize_create(); 2490 sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET); 2491 } 2492 2493 /* 2494 * bridge_rtable_fini: 2495 * 2496 * Deconstruct the route table for this bridge. 2497 */ 2498 static void 2499 bridge_rtable_fini(struct bridge_softc *sc) 2500 { 2501 2502 kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE); 2503 mutex_obj_free(sc->sc_rtlist_lock); 2504 pserialize_destroy(sc->sc_rtlist_psz); 2505 } 2506 2507 /* 2508 * The following hash function is adapted from "Hash Functions" by Bob Jenkins 2509 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997). 2510 */ 2511 #define mix(a, b, c) \ 2512 do { \ 2513 a -= b; a -= c; a ^= (c >> 13); \ 2514 b -= c; b -= a; b ^= (a << 8); \ 2515 c -= a; c -= b; c ^= (b >> 13); \ 2516 a -= b; a -= c; a ^= (c >> 12); \ 2517 b -= c; b -= a; b ^= (a << 16); \ 2518 c -= a; c -= b; c ^= (b >> 5); \ 2519 a -= b; a -= c; a ^= (c >> 3); \ 2520 b -= c; b -= a; b ^= (a << 10); \ 2521 c -= a; c -= b; c ^= (b >> 15); \ 2522 } while (/*CONSTCOND*/0) 2523 2524 static inline uint32_t 2525 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr) 2526 { 2527 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key; 2528 2529 b += addr[5] << 8; 2530 b += addr[4]; 2531 a += (uint32_t)addr[3] << 24; 2532 a += addr[2] << 16; 2533 a += addr[1] << 8; 2534 a += addr[0]; 2535 2536 mix(a, b, c); 2537 2538 return (c & BRIDGE_RTHASH_MASK); 2539 } 2540 2541 #undef mix 2542 2543 /* 2544 * bridge_rtnode_lookup: 2545 * 2546 * Look up a bridge route node for the specified destination. 2547 */ 2548 static struct bridge_rtnode * 2549 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr) 2550 { 2551 struct bridge_rtnode *brt; 2552 uint32_t hash; 2553 int dir; 2554 2555 hash = bridge_rthash(sc, addr); 2556 BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) { 2557 dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN); 2558 if (dir == 0) 2559 return brt; 2560 if (dir > 0) 2561 return NULL; 2562 } 2563 2564 return NULL; 2565 } 2566 2567 /* 2568 * bridge_rtnode_insert: 2569 * 2570 * Insert the specified bridge node into the route table. We 2571 * assume the entry is not already in the table. 2572 */ 2573 static int 2574 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt) 2575 { 2576 struct bridge_rtnode *lbrt, *prev = NULL; 2577 uint32_t hash; 2578 2579 KASSERT(BRIDGE_RT_LOCKED(sc)); 2580 2581 hash = bridge_rthash(sc, brt->brt_addr); 2582 BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) { 2583 int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN); 2584 if (dir == 0) 2585 return EEXIST; 2586 if (dir > 0) 2587 break; 2588 prev = lbrt; 2589 } 2590 if (prev == NULL) 2591 BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt); 2592 else 2593 BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt); 2594 2595 BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt); 2596 sc->sc_brtcnt++; 2597 2598 return 0; 2599 } 2600 2601 /* 2602 * bridge_rtnode_remove: 2603 * 2604 * Remove a bridge rtnode from the rthash and the rtlist of a bridge. 2605 */ 2606 static void 2607 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt) 2608 { 2609 2610 KASSERT(BRIDGE_RT_LOCKED(sc)); 2611 2612 BRIDGE_RTHASH_WRITER_REMOVE(brt); 2613 BRIDGE_RTLIST_WRITER_REMOVE(brt); 2614 sc->sc_brtcnt--; 2615 } 2616 2617 /* 2618 * bridge_rtnode_destroy: 2619 * 2620 * Destroy a bridge rtnode. 2621 */ 2622 static void 2623 bridge_rtnode_destroy(struct bridge_rtnode *brt) 2624 { 2625 2626 PSLIST_ENTRY_DESTROY(brt, brt_list); 2627 PSLIST_ENTRY_DESTROY(brt, brt_hash); 2628 pool_put(&bridge_rtnode_pool, brt); 2629 } 2630 2631 extern pfil_head_t *inet_pfil_hook; /* XXX */ 2632 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 2633 2634 /* 2635 * Send bridge packets through IPF if they are one of the types IPF can deal 2636 * with, or if they are ARP or REVARP. (IPF will pass ARP and REVARP without 2637 * question.) 2638 */ 2639 static int 2640 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir) 2641 { 2642 int snap, error; 2643 struct ether_header *eh1, eh2; 2644 struct llc llc1; 2645 uint16_t ether_type; 2646 2647 snap = 0; 2648 error = -1; /* Default error if not error == 0 */ 2649 eh1 = mtod(*mp, struct ether_header *); 2650 ether_type = ntohs(eh1->ether_type); 2651 2652 /* 2653 * Check for SNAP/LLC. 2654 */ 2655 if (ether_type < ETHERMTU) { 2656 struct llc *llc2 = (struct llc *)(eh1 + 1); 2657 2658 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 && 2659 llc2->llc_dsap == LLC_SNAP_LSAP && 2660 llc2->llc_ssap == LLC_SNAP_LSAP && 2661 llc2->llc_control == LLC_UI) { 2662 ether_type = htons(llc2->llc_un.type_snap.ether_type); 2663 snap = 1; 2664 } 2665 } 2666 2667 /* 2668 * If we're trying to filter bridge traffic, don't look at anything 2669 * other than IP and ARP traffic. If the filter doesn't understand 2670 * IPv6, don't allow IPv6 through the bridge either. This is lame 2671 * since if we really wanted, say, an AppleTalk filter, we are hosed, 2672 * but of course we don't have an AppleTalk filter to begin with. 2673 * (Note that since IPF doesn't understand ARP it will pass *ALL* 2674 * ARP traffic.) 2675 */ 2676 switch (ether_type) { 2677 case ETHERTYPE_ARP: 2678 case ETHERTYPE_REVARP: 2679 return 0; /* Automatically pass */ 2680 case ETHERTYPE_IP: 2681 # ifdef INET6 2682 case ETHERTYPE_IPV6: 2683 # endif /* INET6 */ 2684 break; 2685 default: 2686 goto bad; 2687 } 2688 2689 /* Strip off the Ethernet header and keep a copy. */ 2690 m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2); 2691 m_adj(*mp, ETHER_HDR_LEN); 2692 2693 /* Strip off snap header, if present */ 2694 if (snap) { 2695 m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1); 2696 m_adj(*mp, sizeof(struct llc)); 2697 } 2698 2699 /* 2700 * Check basic packet sanity and run IPF through pfil. 2701 */ 2702 KASSERT(!cpu_intr_p()); 2703 switch (ether_type) 2704 { 2705 case ETHERTYPE_IP : 2706 error = bridge_ip_checkbasic(mp); 2707 if (error == 0) 2708 error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir); 2709 break; 2710 # ifdef INET6 2711 case ETHERTYPE_IPV6 : 2712 error = bridge_ip6_checkbasic(mp); 2713 if (error == 0) 2714 error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir); 2715 break; 2716 # endif 2717 default : 2718 error = 0; 2719 break; 2720 } 2721 2722 if (*mp == NULL) 2723 return error; 2724 if (error != 0) 2725 goto bad; 2726 2727 error = -1; 2728 2729 /* 2730 * Finally, put everything back the way it was and return 2731 */ 2732 if (snap) { 2733 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT); 2734 if (*mp == NULL) 2735 return error; 2736 bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc)); 2737 } 2738 2739 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT); 2740 if (*mp == NULL) 2741 return error; 2742 bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN); 2743 2744 return 0; 2745 2746 bad: 2747 m_freem(*mp); 2748 *mp = NULL; 2749 return error; 2750 } 2751 2752 /* 2753 * Perform basic checks on header size since 2754 * IPF assumes ip_input has already processed 2755 * it for it. Cut-and-pasted from ip_input.c. 2756 * Given how simple the IPv6 version is, 2757 * does the IPv4 version really need to be 2758 * this complicated? 2759 * 2760 * XXX Should we update ipstat here, or not? 2761 * XXX Right now we update ipstat but not 2762 * XXX csum_counter. 2763 */ 2764 static int 2765 bridge_ip_checkbasic(struct mbuf **mp) 2766 { 2767 struct mbuf *m = *mp; 2768 struct ip *ip; 2769 int len, hlen; 2770 2771 if (*mp == NULL) 2772 return -1; 2773 2774 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 2775 if ((m = m_copyup(m, sizeof(struct ip), 2776 (max_linkhdr + 3) & ~3)) == NULL) { 2777 /* XXXJRT new stat, please */ 2778 ip_statinc(IP_STAT_TOOSMALL); 2779 goto bad; 2780 } 2781 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 2782 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 2783 ip_statinc(IP_STAT_TOOSMALL); 2784 goto bad; 2785 } 2786 } 2787 ip = mtod(m, struct ip *); 2788 if (ip == NULL) goto bad; 2789 2790 if (ip->ip_v != IPVERSION) { 2791 ip_statinc(IP_STAT_BADVERS); 2792 goto bad; 2793 } 2794 hlen = ip->ip_hl << 2; 2795 if (hlen < sizeof(struct ip)) { /* minimum header length */ 2796 ip_statinc(IP_STAT_BADHLEN); 2797 goto bad; 2798 } 2799 if (hlen > m->m_len) { 2800 if ((m = m_pullup(m, hlen)) == 0) { 2801 ip_statinc(IP_STAT_BADHLEN); 2802 goto bad; 2803 } 2804 ip = mtod(m, struct ip *); 2805 if (ip == NULL) goto bad; 2806 } 2807 2808 switch (m->m_pkthdr.csum_flags & 2809 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) | 2810 M_CSUM_IPv4_BAD)) { 2811 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 2812 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */ 2813 goto bad; 2814 2815 case M_CSUM_IPv4: 2816 /* Checksum was okay. */ 2817 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */ 2818 break; 2819 2820 default: 2821 /* Must compute it ourselves. */ 2822 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */ 2823 if (in_cksum(m, hlen) != 0) 2824 goto bad; 2825 break; 2826 } 2827 2828 /* Retrieve the packet length. */ 2829 len = ntohs(ip->ip_len); 2830 2831 /* 2832 * Check for additional length bogosity 2833 */ 2834 if (len < hlen) { 2835 ip_statinc(IP_STAT_BADLEN); 2836 goto bad; 2837 } 2838 2839 /* 2840 * Check that the amount of data in the buffers 2841 * is as at least much as the IP header would have us expect. 2842 * Drop packet if shorter than we expect. 2843 */ 2844 if (m->m_pkthdr.len < len) { 2845 ip_statinc(IP_STAT_TOOSHORT); 2846 goto bad; 2847 } 2848 2849 /* Checks out, proceed */ 2850 *mp = m; 2851 return 0; 2852 2853 bad: 2854 *mp = m; 2855 return -1; 2856 } 2857 2858 # ifdef INET6 2859 /* 2860 * Same as above, but for IPv6. 2861 * Cut-and-pasted from ip6_input.c. 2862 * XXX Should we update ip6stat, or not? 2863 */ 2864 static int 2865 bridge_ip6_checkbasic(struct mbuf **mp) 2866 { 2867 struct mbuf *m = *mp; 2868 struct ip6_hdr *ip6; 2869 2870 /* 2871 * If the IPv6 header is not aligned, slurp it up into a new 2872 * mbuf with space for link headers, in the event we forward 2873 * it. Otherwise, if it is aligned, make sure the entire base 2874 * IPv6 header is in the first mbuf of the chain. 2875 */ 2876 if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 2877 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m); 2878 if ((m = m_copyup(m, sizeof(struct ip6_hdr), 2879 (max_linkhdr + 3) & ~3)) == NULL) { 2880 /* XXXJRT new stat, please */ 2881 ip6_statinc(IP6_STAT_TOOSMALL); 2882 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 2883 goto bad; 2884 } 2885 } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) { 2886 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m); 2887 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 2888 ip6_statinc(IP6_STAT_TOOSMALL); 2889 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 2890 goto bad; 2891 } 2892 } 2893 2894 ip6 = mtod(m, struct ip6_hdr *); 2895 2896 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 2897 ip6_statinc(IP6_STAT_BADVERS); 2898 in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr); 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 # endif /* INET6 */ 2911