1 /* 2 * Copyright 2001 Wasabi Systems, Inc. 3 * All rights reserved. 4 * 5 * Written by Jason R. Thorpe for Wasabi Systems, Inc. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed for the NetBSD Project by 18 * Wasabi Systems, Inc. 19 * 4. The name of Wasabi Systems, Inc. may not be used to endorse 20 * or promote products derived from this software without specific prior 21 * written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 25 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC 27 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 33 * POSSIBILITY OF SUCH DAMAGE. 34 */ 35 36 /* 37 * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net) 38 * All rights reserved. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by Jason L. Wright 51 * 4. The name of the author may not be used to endorse or promote products 52 * derived from this software without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 55 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 56 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 57 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, 58 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 59 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 60 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 62 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 63 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 64 * POSSIBILITY OF SUCH DAMAGE. 65 * 66 * $OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp $ 67 * $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $ 68 * $FreeBSD: src/sys/net/if_bridge.c,v 1.26 2005/10/13 23:05:55 thompsa Exp $ 69 * $DragonFly: src/sys/net/bridge/if_bridge.c,v 1.60 2008/11/26 12:49:43 sephe Exp $ 70 */ 71 72 /* 73 * Network interface bridge support. 74 * 75 * TODO: 76 * 77 * - Currently only supports Ethernet-like interfaces (Ethernet, 78 * 802.11, VLANs on Ethernet, etc.) Figure out a nice way 79 * to bridge other types of interfaces (FDDI-FDDI, and maybe 80 * consider heterogenous bridges). 81 * 82 * 83 * Bridge's route information is duplicated to each CPUs: 84 * 85 * CPU0 CPU1 CPU2 CPU3 86 * +-----------+ +-----------+ +-----------+ +-----------+ 87 * | rtnode | | rtnode | | rtnode | | rtnode | 88 * | | | | | | | | 89 * | dst eaddr | | dst eaddr | | dst eaddr | | dst eaddr | 90 * +-----------+ +-----------+ +-----------+ +-----------+ 91 * | | | | 92 * | | | | 93 * | | +----------+ | | 94 * | | | rtinfo | | | 95 * | +---->| |<---+ | 96 * | | flags | | 97 * +-------------->| timeout |<-------------+ 98 * | dst_ifp | 99 * +----------+ 100 * 101 * We choose to put timeout and dst_ifp into shared part, so updating 102 * them will be cheaper than using message forwarding. Also there is 103 * not need to use spinlock to protect the updating: timeout and dst_ifp 104 * is not related and specific field's updating order has no importance. 105 * The cache pollution by the share part should not be heavy: in a stable 106 * setup, dst_ifp probably will be not changed in rtnode's life time, 107 * while timeout is refreshed once per second; most of the time, timeout 108 * and dst_ifp are read-only accessed. 109 * 110 * 111 * Bridge route information installation on bridge_input path: 112 * 113 * CPU0 CPU1 CPU2 CPU3 114 * 115 * tcp_thread2 116 * | 117 * alloc nmsg 118 * snd nmsg | 119 * w/o rtinfo | 120 * ifnet0<-----------------------+ 121 * | : 122 * lookup dst : 123 * rtnode exists?(Y)free nmsg : 124 * |(N) : 125 * | 126 * alloc rtinfo 127 * alloc rtnode 128 * install rtnode 129 * | 130 * +---------->ifnet1 131 * : fwd nmsg | 132 * : w/ rtinfo | 133 * : | 134 * : | 135 * alloc rtnode 136 * (w/ nmsg's rtinfo) 137 * install rtnode 138 * | 139 * +---------->ifnet2 140 * : fwd nmsg | 141 * : w/ rtinfo | 142 * : | 143 * : same as ifnet1 144 * | 145 * +---------->ifnet3 146 * : fwd nmsg | 147 * : w/ rtinfo | 148 * : | 149 * : same as ifnet1 150 * free nmsg 151 * : 152 * : 153 * 154 * The netmsgs forwarded between protocol threads and ifnet threads are 155 * allocated with (M_WAITOK|M_NULLOK), so it will not fail under most 156 * cases (route information is too precious to be not installed :). 157 * Since multiple threads may try to install route information for the 158 * same dst eaddr, we look up route information in ifnet0. However, this 159 * looking up only need to be performed on ifnet0, which is the start 160 * point of the route information installation process. 161 * 162 * 163 * Bridge route information deleting/flushing: 164 * 165 * CPU0 CPU1 CPU2 CPU3 166 * 167 * netisr0 168 * | 169 * find suitable rtnodes, 170 * mark their rtinfo dead 171 * | 172 * | domsg <------------------------------------------+ 173 * | | replymsg 174 * | | 175 * V fwdmsg fwdmsg fwdmsg | 176 * ifnet0 --------> ifnet1 --------> ifnet2 --------> ifnet3 177 * delete rtnodes delete rtnodes delete rtnodes delete rtnodes 178 * w/ dead rtinfo w/ dead rtinfo w/ dead rtinfo w/ dead rtinfo 179 * free dead rtinfos 180 * 181 * All deleting/flushing operations are serialized by netisr0, so each 182 * operation only reaps the route information marked dead by itself. 183 * 184 * 185 * Bridge route information adding/deleting/flushing: 186 * Since all operation is serialized by the fixed message flow between 187 * ifnet threads, it is not possible to create corrupted per-cpu route 188 * information. 189 * 190 * 191 * 192 * Percpu member interface list iteration with blocking operation: 193 * Since one bridge could only delete one member interface at a time and 194 * the deleted member interface is not freed after netmsg_service_sync(), 195 * following way is used to make sure that even if the certain member 196 * interface is ripped from the percpu list during the blocking operation, 197 * the iteration still could keep going: 198 * 199 * LIST_FOREACH_MUTABLE(bif, sc->sc_iflists[mycpuid], bif_next, nbif) { 200 * blocking operation; 201 * blocking operation; 202 * ... 203 * ... 204 * if (nbif != NULL && !nbif->bif_onlist) { 205 * KKASSERT(bif->bif_onlist); 206 * nbif = LIST_NEXT(bif, bif_next); 207 * } 208 * } 209 * 210 * As mentioned above only one member interface could be unlinked from the 211 * percpu member interface list, so either bif or nbif may be not on the list, 212 * but _not_ both. To keep the list iteration, we don't care about bif, but 213 * only nbif. Since removed member interface will only be freed after we 214 * finish our work, it is safe to access any field in an unlinked bif (here 215 * bif_onlist). If nbif is no longer on the list, then bif must be on the 216 * list, so we change nbif to the next element of bif and keep going. 217 */ 218 219 #include "opt_inet.h" 220 #include "opt_inet6.h" 221 222 #include <sys/param.h> 223 #include <sys/mbuf.h> 224 #include <sys/malloc.h> 225 #include <sys/protosw.h> 226 #include <sys/systm.h> 227 #include <sys/time.h> 228 #include <sys/socket.h> /* for net/if.h */ 229 #include <sys/sockio.h> 230 #include <sys/ctype.h> /* string functions */ 231 #include <sys/kernel.h> 232 #include <sys/random.h> 233 #include <sys/sysctl.h> 234 #include <sys/module.h> 235 #include <sys/proc.h> 236 #include <sys/priv.h> 237 #include <sys/lock.h> 238 #include <sys/thread.h> 239 #include <sys/thread2.h> 240 #include <sys/mpipe.h> 241 242 #include <net/bpf.h> 243 #include <net/if.h> 244 #include <net/if_dl.h> 245 #include <net/if_types.h> 246 #include <net/if_var.h> 247 #include <net/pfil.h> 248 #include <net/ifq_var.h> 249 #include <net/if_clone.h> 250 251 #include <netinet/in.h> /* for struct arpcom */ 252 #include <netinet/in_systm.h> 253 #include <netinet/in_var.h> 254 #include <netinet/ip.h> 255 #include <netinet/ip_var.h> 256 #ifdef INET6 257 #include <netinet/ip6.h> 258 #include <netinet6/ip6_var.h> 259 #endif 260 #include <netinet/if_ether.h> /* for struct arpcom */ 261 #include <net/bridge/if_bridgevar.h> 262 #include <net/if_llc.h> 263 #include <net/netmsg2.h> 264 265 #include <net/route.h> 266 #include <sys/in_cksum.h> 267 268 /* 269 * Size of the route hash table. Must be a power of two. 270 */ 271 #ifndef BRIDGE_RTHASH_SIZE 272 #define BRIDGE_RTHASH_SIZE 1024 273 #endif 274 275 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1) 276 277 /* 278 * Maximum number of addresses to cache. 279 */ 280 #ifndef BRIDGE_RTABLE_MAX 281 #define BRIDGE_RTABLE_MAX 100 282 #endif 283 284 /* 285 * Spanning tree defaults. 286 */ 287 #define BSTP_DEFAULT_MAX_AGE (20 * 256) 288 #define BSTP_DEFAULT_HELLO_TIME (2 * 256) 289 #define BSTP_DEFAULT_FORWARD_DELAY (15 * 256) 290 #define BSTP_DEFAULT_HOLD_TIME (1 * 256) 291 #define BSTP_DEFAULT_BRIDGE_PRIORITY 0x8000 292 #define BSTP_DEFAULT_PORT_PRIORITY 0x80 293 #define BSTP_DEFAULT_PATH_COST 55 294 295 /* 296 * Timeout (in seconds) for entries learned dynamically. 297 */ 298 #ifndef BRIDGE_RTABLE_TIMEOUT 299 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */ 300 #endif 301 302 /* 303 * Number of seconds between walks of the route list. 304 */ 305 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD 306 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60) 307 #endif 308 309 /* 310 * List of capabilities to mask on the member interface. 311 */ 312 #define BRIDGE_IFCAPS_MASK IFCAP_TXCSUM 313 314 typedef int (*bridge_ctl_t)(struct bridge_softc *, void *); 315 316 struct netmsg_brctl { 317 struct netmsg_base base; 318 bridge_ctl_t bc_func; 319 struct bridge_softc *bc_sc; 320 void *bc_arg; 321 }; 322 323 struct netmsg_brsaddr { 324 struct netmsg_base base; 325 struct bridge_softc *br_softc; 326 struct ifnet *br_dst_if; 327 struct bridge_rtinfo *br_rtinfo; 328 int br_setflags; 329 uint8_t br_dst[ETHER_ADDR_LEN]; 330 uint8_t br_flags; 331 }; 332 333 struct netmsg_braddbif { 334 struct netmsg_base base; 335 struct bridge_softc *br_softc; 336 struct bridge_ifinfo *br_bif_info; 337 struct ifnet *br_bif_ifp; 338 }; 339 340 struct netmsg_brdelbif { 341 struct netmsg_base base; 342 struct bridge_softc *br_softc; 343 struct bridge_ifinfo *br_bif_info; 344 struct bridge_iflist_head *br_bif_list; 345 }; 346 347 struct netmsg_brsflags { 348 struct netmsg_base base; 349 struct bridge_softc *br_softc; 350 struct bridge_ifinfo *br_bif_info; 351 uint32_t br_bif_flags; 352 }; 353 354 eventhandler_tag bridge_detach_cookie = NULL; 355 356 extern struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 357 extern int (*bridge_output_p)(struct ifnet *, struct mbuf *); 358 extern void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 359 360 static int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD; 361 362 static int bridge_clone_create(struct if_clone *, int, caddr_t); 363 static int bridge_clone_destroy(struct ifnet *); 364 365 static int bridge_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *); 366 static void bridge_mutecaps(struct bridge_ifinfo *, struct ifnet *, int); 367 static void bridge_ifdetach(void *, struct ifnet *); 368 static void bridge_init(void *); 369 static void bridge_stop(struct ifnet *); 370 static void bridge_start(struct ifnet *); 371 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *); 372 static int bridge_output(struct ifnet *, struct mbuf *); 373 374 static void bridge_forward(struct bridge_softc *, struct mbuf *m); 375 376 static void bridge_timer_handler(netmsg_t); 377 static void bridge_timer(void *); 378 379 static void bridge_start_bcast(struct bridge_softc *, struct mbuf *); 380 static void bridge_broadcast(struct bridge_softc *, struct ifnet *, 381 struct mbuf *); 382 static void bridge_span(struct bridge_softc *, struct mbuf *); 383 384 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *, 385 struct ifnet *, uint8_t); 386 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *); 387 static void bridge_rtreap(struct bridge_softc *); 388 static void bridge_rtreap_async(struct bridge_softc *); 389 static void bridge_rttrim(struct bridge_softc *); 390 static int bridge_rtage_finddead(struct bridge_softc *); 391 static void bridge_rtage(struct bridge_softc *); 392 static void bridge_rtflush(struct bridge_softc *, int); 393 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *); 394 static int bridge_rtsaddr(struct bridge_softc *, const uint8_t *, 395 struct ifnet *, uint8_t); 396 static void bridge_rtmsg_sync(struct bridge_softc *sc); 397 static void bridge_rtreap_handler(netmsg_t); 398 static void bridge_rtinstall_handler(netmsg_t); 399 static int bridge_rtinstall_oncpu(struct bridge_softc *, const uint8_t *, 400 struct ifnet *, int, uint8_t, struct bridge_rtinfo **); 401 402 static void bridge_rtable_init(struct bridge_softc *); 403 static void bridge_rtable_fini(struct bridge_softc *); 404 405 static int bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *); 406 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *, 407 const uint8_t *); 408 static void bridge_rtnode_insert(struct bridge_softc *, 409 struct bridge_rtnode *); 410 static void bridge_rtnode_destroy(struct bridge_softc *, 411 struct bridge_rtnode *); 412 413 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *, 414 const char *name); 415 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *, 416 struct ifnet *ifp); 417 static struct bridge_iflist *bridge_lookup_member_ifinfo(struct bridge_softc *, 418 struct bridge_ifinfo *); 419 static void bridge_delete_member(struct bridge_softc *, 420 struct bridge_iflist *, int); 421 static void bridge_delete_span(struct bridge_softc *, 422 struct bridge_iflist *); 423 424 static int bridge_control(struct bridge_softc *, u_long, 425 bridge_ctl_t, void *); 426 static int bridge_ioctl_init(struct bridge_softc *, void *); 427 static int bridge_ioctl_stop(struct bridge_softc *, void *); 428 static int bridge_ioctl_add(struct bridge_softc *, void *); 429 static int bridge_ioctl_del(struct bridge_softc *, void *); 430 static int bridge_ioctl_gifflags(struct bridge_softc *, void *); 431 static int bridge_ioctl_sifflags(struct bridge_softc *, void *); 432 static int bridge_ioctl_scache(struct bridge_softc *, void *); 433 static int bridge_ioctl_gcache(struct bridge_softc *, void *); 434 static int bridge_ioctl_gifs(struct bridge_softc *, void *); 435 static int bridge_ioctl_rts(struct bridge_softc *, void *); 436 static int bridge_ioctl_saddr(struct bridge_softc *, void *); 437 static int bridge_ioctl_sto(struct bridge_softc *, void *); 438 static int bridge_ioctl_gto(struct bridge_softc *, void *); 439 static int bridge_ioctl_daddr(struct bridge_softc *, void *); 440 static int bridge_ioctl_flush(struct bridge_softc *, void *); 441 static int bridge_ioctl_gpri(struct bridge_softc *, void *); 442 static int bridge_ioctl_spri(struct bridge_softc *, void *); 443 static int bridge_ioctl_ght(struct bridge_softc *, void *); 444 static int bridge_ioctl_sht(struct bridge_softc *, void *); 445 static int bridge_ioctl_gfd(struct bridge_softc *, void *); 446 static int bridge_ioctl_sfd(struct bridge_softc *, void *); 447 static int bridge_ioctl_gma(struct bridge_softc *, void *); 448 static int bridge_ioctl_sma(struct bridge_softc *, void *); 449 static int bridge_ioctl_sifprio(struct bridge_softc *, void *); 450 static int bridge_ioctl_sifcost(struct bridge_softc *, void *); 451 static int bridge_ioctl_addspan(struct bridge_softc *, void *); 452 static int bridge_ioctl_delspan(struct bridge_softc *, void *); 453 static int bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *, 454 int); 455 static int bridge_ip_checkbasic(struct mbuf **mp); 456 #ifdef INET6 457 static int bridge_ip6_checkbasic(struct mbuf **mp); 458 #endif /* INET6 */ 459 static int bridge_fragment(struct ifnet *, struct mbuf *, 460 struct ether_header *, int, struct llc *); 461 static void bridge_enqueue_handler(netmsg_t); 462 static void bridge_handoff(struct ifnet *, struct mbuf *); 463 464 static void bridge_del_bif_handler(netmsg_t); 465 static void bridge_add_bif_handler(netmsg_t); 466 static void bridge_set_bifflags_handler(netmsg_t); 467 static void bridge_del_bif(struct bridge_softc *, struct bridge_ifinfo *, 468 struct bridge_iflist_head *); 469 static void bridge_add_bif(struct bridge_softc *, struct bridge_ifinfo *, 470 struct ifnet *); 471 static void bridge_set_bifflags(struct bridge_softc *, 472 struct bridge_ifinfo *, uint32_t); 473 474 SYSCTL_DECL(_net_link); 475 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge"); 476 477 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */ 478 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */ 479 static int pfil_member = 1; /* run pfil hooks on the member interface */ 480 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW, 481 &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled"); 482 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW, 483 &pfil_bridge, 0, "Packet filter on the bridge interface"); 484 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW, 485 &pfil_member, 0, "Packet filter on the member interface"); 486 487 struct bridge_control_arg { 488 union { 489 struct ifbreq ifbreq; 490 struct ifbifconf ifbifconf; 491 struct ifbareq ifbareq; 492 struct ifbaconf ifbaconf; 493 struct ifbrparam ifbrparam; 494 } bca_u; 495 int bca_len; 496 void *bca_uptr; 497 void *bca_kptr; 498 }; 499 500 struct bridge_control { 501 bridge_ctl_t bc_func; 502 int bc_argsize; 503 int bc_flags; 504 }; 505 506 #define BC_F_COPYIN 0x01 /* copy arguments in */ 507 #define BC_F_COPYOUT 0x02 /* copy arguments out */ 508 #define BC_F_SUSER 0x04 /* do super-user check */ 509 510 const struct bridge_control bridge_control_table[] = { 511 { bridge_ioctl_add, sizeof(struct ifbreq), 512 BC_F_COPYIN|BC_F_SUSER }, 513 { bridge_ioctl_del, sizeof(struct ifbreq), 514 BC_F_COPYIN|BC_F_SUSER }, 515 516 { bridge_ioctl_gifflags, sizeof(struct ifbreq), 517 BC_F_COPYIN|BC_F_COPYOUT }, 518 { bridge_ioctl_sifflags, sizeof(struct ifbreq), 519 BC_F_COPYIN|BC_F_SUSER }, 520 521 { bridge_ioctl_scache, sizeof(struct ifbrparam), 522 BC_F_COPYIN|BC_F_SUSER }, 523 { bridge_ioctl_gcache, sizeof(struct ifbrparam), 524 BC_F_COPYOUT }, 525 526 { bridge_ioctl_gifs, sizeof(struct ifbifconf), 527 BC_F_COPYIN|BC_F_COPYOUT }, 528 { bridge_ioctl_rts, sizeof(struct ifbaconf), 529 BC_F_COPYIN|BC_F_COPYOUT }, 530 531 { bridge_ioctl_saddr, sizeof(struct ifbareq), 532 BC_F_COPYIN|BC_F_SUSER }, 533 534 { bridge_ioctl_sto, sizeof(struct ifbrparam), 535 BC_F_COPYIN|BC_F_SUSER }, 536 { bridge_ioctl_gto, sizeof(struct ifbrparam), 537 BC_F_COPYOUT }, 538 539 { bridge_ioctl_daddr, sizeof(struct ifbareq), 540 BC_F_COPYIN|BC_F_SUSER }, 541 542 { bridge_ioctl_flush, sizeof(struct ifbreq), 543 BC_F_COPYIN|BC_F_SUSER }, 544 545 { bridge_ioctl_gpri, sizeof(struct ifbrparam), 546 BC_F_COPYOUT }, 547 { bridge_ioctl_spri, sizeof(struct ifbrparam), 548 BC_F_COPYIN|BC_F_SUSER }, 549 550 { bridge_ioctl_ght, sizeof(struct ifbrparam), 551 BC_F_COPYOUT }, 552 { bridge_ioctl_sht, sizeof(struct ifbrparam), 553 BC_F_COPYIN|BC_F_SUSER }, 554 555 { bridge_ioctl_gfd, sizeof(struct ifbrparam), 556 BC_F_COPYOUT }, 557 { bridge_ioctl_sfd, sizeof(struct ifbrparam), 558 BC_F_COPYIN|BC_F_SUSER }, 559 560 { bridge_ioctl_gma, sizeof(struct ifbrparam), 561 BC_F_COPYOUT }, 562 { bridge_ioctl_sma, sizeof(struct ifbrparam), 563 BC_F_COPYIN|BC_F_SUSER }, 564 565 { bridge_ioctl_sifprio, sizeof(struct ifbreq), 566 BC_F_COPYIN|BC_F_SUSER }, 567 568 { bridge_ioctl_sifcost, sizeof(struct ifbreq), 569 BC_F_COPYIN|BC_F_SUSER }, 570 571 { bridge_ioctl_addspan, sizeof(struct ifbreq), 572 BC_F_COPYIN|BC_F_SUSER }, 573 { bridge_ioctl_delspan, sizeof(struct ifbreq), 574 BC_F_COPYIN|BC_F_SUSER }, 575 }; 576 static const int bridge_control_table_size = 577 sizeof(bridge_control_table) / sizeof(bridge_control_table[0]); 578 579 LIST_HEAD(, bridge_softc) bridge_list; 580 581 struct if_clone bridge_cloner = IF_CLONE_INITIALIZER("bridge", 582 bridge_clone_create, 583 bridge_clone_destroy, 0, IF_MAXUNIT); 584 585 static int 586 bridge_modevent(module_t mod, int type, void *data) 587 { 588 switch (type) { 589 case MOD_LOAD: 590 LIST_INIT(&bridge_list); 591 if_clone_attach(&bridge_cloner); 592 bridge_input_p = bridge_input; 593 bridge_output_p = bridge_output; 594 bridge_detach_cookie = EVENTHANDLER_REGISTER( 595 ifnet_detach_event, bridge_ifdetach, NULL, 596 EVENTHANDLER_PRI_ANY); 597 #if notyet 598 bstp_linkstate_p = bstp_linkstate; 599 #endif 600 break; 601 case MOD_UNLOAD: 602 if (!LIST_EMPTY(&bridge_list)) 603 return (EBUSY); 604 EVENTHANDLER_DEREGISTER(ifnet_detach_event, 605 bridge_detach_cookie); 606 if_clone_detach(&bridge_cloner); 607 bridge_input_p = NULL; 608 bridge_output_p = NULL; 609 #if notyet 610 bstp_linkstate_p = NULL; 611 #endif 612 break; 613 default: 614 return (EOPNOTSUPP); 615 } 616 return (0); 617 } 618 619 static moduledata_t bridge_mod = { 620 "if_bridge", 621 bridge_modevent, 622 0 623 }; 624 625 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 626 627 628 /* 629 * bridge_clone_create: 630 * 631 * Create a new bridge instance. 632 */ 633 static int 634 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t param __unused) 635 { 636 struct bridge_softc *sc; 637 struct ifnet *ifp; 638 u_char eaddr[6]; 639 int cpu, rnd; 640 641 sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO); 642 ifp = sc->sc_ifp = &sc->sc_if; 643 644 sc->sc_brtmax = BRIDGE_RTABLE_MAX; 645 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT; 646 sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE; 647 sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME; 648 sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY; 649 sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY; 650 sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME; 651 652 /* Initialize our routing table. */ 653 bridge_rtable_init(sc); 654 655 callout_init(&sc->sc_brcallout); 656 netmsg_init(&sc->sc_brtimemsg, NULL, &netisr_adone_rport, 657 MSGF_DROPABLE, bridge_timer_handler); 658 sc->sc_brtimemsg.lmsg.u.ms_resultp = sc; 659 660 callout_init(&sc->sc_bstpcallout); 661 netmsg_init(&sc->sc_bstptimemsg, NULL, &netisr_adone_rport, 662 MSGF_DROPABLE, bstp_tick_handler); 663 sc->sc_bstptimemsg.lmsg.u.ms_resultp = sc; 664 665 /* Initialize per-cpu member iface lists */ 666 sc->sc_iflists = kmalloc(sizeof(*sc->sc_iflists) * ncpus, 667 M_DEVBUF, M_WAITOK); 668 for (cpu = 0; cpu < ncpus; ++cpu) 669 LIST_INIT(&sc->sc_iflists[cpu]); 670 671 LIST_INIT(&sc->sc_spanlist); 672 673 ifp->if_softc = sc; 674 if_initname(ifp, ifc->ifc_name, unit); 675 ifp->if_mtu = ETHERMTU; 676 ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST; 677 ifp->if_ioctl = bridge_ioctl; 678 ifp->if_start = bridge_start; 679 ifp->if_init = bridge_init; 680 ifp->if_type = IFT_BRIDGE; 681 ifq_set_maxlen(&ifp->if_snd, ifqmaxlen); 682 ifq_set_ready(&ifp->if_snd); 683 ifp->if_hdrlen = ETHER_HDR_LEN; 684 685 /* 686 * Generate a random ethernet address and use the private AC:DE:48 687 * OUI code. 688 */ 689 rnd = karc4random(); 690 bcopy(&rnd, &eaddr[0], 4); /* ETHER_ADDR_LEN == 6 */ 691 rnd = karc4random(); 692 bcopy(&rnd, &eaddr[2], 4); /* ETHER_ADDR_LEN == 6 */ 693 694 eaddr[0] &= ~1; /* clear multicast bit */ 695 eaddr[0] |= 2; /* set the LAA bit */ 696 697 ether_ifattach(ifp, eaddr, NULL); 698 /* Now undo some of the damage... */ 699 ifp->if_baudrate = 0; 700 ifp->if_type = IFT_BRIDGE; 701 702 crit_enter(); /* XXX MP */ 703 LIST_INSERT_HEAD(&bridge_list, sc, sc_list); 704 crit_exit(); 705 706 return (0); 707 } 708 709 static void 710 bridge_delete_dispatch(netmsg_t msg) 711 { 712 struct bridge_softc *sc = msg->lmsg.u.ms_resultp; 713 struct ifnet *bifp = sc->sc_ifp; 714 struct bridge_iflist *bif; 715 716 ifnet_serialize_all(bifp); 717 718 while ((bif = LIST_FIRST(&sc->sc_iflists[mycpuid])) != NULL) 719 bridge_delete_member(sc, bif, 0); 720 721 while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) 722 bridge_delete_span(sc, bif); 723 724 ifnet_deserialize_all(bifp); 725 726 lwkt_replymsg(&msg->lmsg, 0); 727 } 728 729 /* 730 * bridge_clone_destroy: 731 * 732 * Destroy a bridge instance. 733 */ 734 static int 735 bridge_clone_destroy(struct ifnet *ifp) 736 { 737 struct bridge_softc *sc = ifp->if_softc; 738 struct netmsg_base msg; 739 740 ifnet_serialize_all(ifp); 741 742 bridge_stop(ifp); 743 ifp->if_flags &= ~IFF_UP; 744 745 ifnet_deserialize_all(ifp); 746 747 netmsg_init(&msg, NULL, &curthread->td_msgport, 748 0, bridge_delete_dispatch); 749 msg.lmsg.u.ms_resultp = sc; 750 lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0); 751 752 crit_enter(); /* XXX MP */ 753 LIST_REMOVE(sc, sc_list); 754 crit_exit(); 755 756 ether_ifdetach(ifp); 757 758 /* Tear down the routing table. */ 759 bridge_rtable_fini(sc); 760 761 /* Free per-cpu member iface lists */ 762 kfree(sc->sc_iflists, M_DEVBUF); 763 764 kfree(sc, M_DEVBUF); 765 766 return 0; 767 } 768 769 /* 770 * bridge_ioctl: 771 * 772 * Handle a control request from the operator. 773 */ 774 static int 775 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr) 776 { 777 struct bridge_softc *sc = ifp->if_softc; 778 struct bridge_control_arg args; 779 struct ifdrv *ifd = (struct ifdrv *) data; 780 const struct bridge_control *bc; 781 int error = 0; 782 783 ASSERT_IFNET_SERIALIZED_ALL(ifp); 784 785 switch (cmd) { 786 case SIOCADDMULTI: 787 case SIOCDELMULTI: 788 break; 789 790 case SIOCGDRVSPEC: 791 case SIOCSDRVSPEC: 792 if (ifd->ifd_cmd >= bridge_control_table_size) { 793 error = EINVAL; 794 break; 795 } 796 bc = &bridge_control_table[ifd->ifd_cmd]; 797 798 if (cmd == SIOCGDRVSPEC && 799 (bc->bc_flags & BC_F_COPYOUT) == 0) { 800 error = EINVAL; 801 break; 802 } else if (cmd == SIOCSDRVSPEC && 803 (bc->bc_flags & BC_F_COPYOUT)) { 804 error = EINVAL; 805 break; 806 } 807 808 if (bc->bc_flags & BC_F_SUSER) { 809 error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY); 810 if (error) 811 break; 812 } 813 814 if (ifd->ifd_len != bc->bc_argsize || 815 ifd->ifd_len > sizeof(args.bca_u)) { 816 error = EINVAL; 817 break; 818 } 819 820 memset(&args, 0, sizeof(args)); 821 if (bc->bc_flags & BC_F_COPYIN) { 822 error = copyin(ifd->ifd_data, &args.bca_u, 823 ifd->ifd_len); 824 if (error) 825 break; 826 } 827 828 error = bridge_control(sc, cmd, bc->bc_func, &args); 829 if (error) { 830 KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL); 831 break; 832 } 833 834 if (bc->bc_flags & BC_F_COPYOUT) { 835 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); 836 if (args.bca_len != 0) { 837 KKASSERT(args.bca_kptr != NULL); 838 if (!error) { 839 error = copyout(args.bca_kptr, 840 args.bca_uptr, args.bca_len); 841 } 842 kfree(args.bca_kptr, M_TEMP); 843 } else { 844 KKASSERT(args.bca_kptr == NULL); 845 } 846 } else { 847 KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL); 848 } 849 break; 850 851 case SIOCSIFFLAGS: 852 if (!(ifp->if_flags & IFF_UP) && 853 (ifp->if_flags & IFF_RUNNING)) { 854 /* 855 * If interface is marked down and it is running, 856 * then stop it. 857 */ 858 bridge_stop(ifp); 859 } else if ((ifp->if_flags & IFF_UP) && 860 !(ifp->if_flags & IFF_RUNNING)) { 861 /* 862 * If interface is marked up and it is stopped, then 863 * start it. 864 */ 865 ifp->if_init(sc); 866 } 867 break; 868 869 case SIOCSIFMTU: 870 /* Do not allow the MTU to be changed on the bridge */ 871 error = EINVAL; 872 break; 873 874 default: 875 error = ether_ioctl(ifp, cmd, data); 876 break; 877 } 878 return (error); 879 } 880 881 /* 882 * bridge_mutecaps: 883 * 884 * Clear or restore unwanted capabilities on the member interface 885 */ 886 static void 887 bridge_mutecaps(struct bridge_ifinfo *bif_info, struct ifnet *ifp, int mute) 888 { 889 struct ifreq ifr; 890 int error; 891 892 if (ifp->if_ioctl == NULL) 893 return; 894 895 bzero(&ifr, sizeof(ifr)); 896 ifr.ifr_reqcap = ifp->if_capenable; 897 898 if (mute) { 899 /* mask off and save capabilities */ 900 bif_info->bifi_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK; 901 if (bif_info->bifi_mutecap != 0) 902 ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK; 903 } else { 904 /* restore muted capabilities */ 905 ifr.ifr_reqcap |= bif_info->bifi_mutecap; 906 } 907 908 if (bif_info->bifi_mutecap != 0) { 909 ifnet_serialize_all(ifp); 910 error = ifp->if_ioctl(ifp, SIOCSIFCAP, (caddr_t)&ifr, NULL); 911 ifnet_deserialize_all(ifp); 912 } 913 } 914 915 /* 916 * bridge_lookup_member: 917 * 918 * Lookup a bridge member interface. 919 */ 920 static struct bridge_iflist * 921 bridge_lookup_member(struct bridge_softc *sc, const char *name) 922 { 923 struct bridge_iflist *bif; 924 925 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) { 926 if (strcmp(bif->bif_ifp->if_xname, name) == 0) 927 return (bif); 928 } 929 return (NULL); 930 } 931 932 /* 933 * bridge_lookup_member_if: 934 * 935 * Lookup a bridge member interface by ifnet*. 936 */ 937 static struct bridge_iflist * 938 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp) 939 { 940 struct bridge_iflist *bif; 941 942 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) { 943 if (bif->bif_ifp == member_ifp) 944 return (bif); 945 } 946 return (NULL); 947 } 948 949 /* 950 * bridge_lookup_member_ifinfo: 951 * 952 * Lookup a bridge member interface by bridge_ifinfo. 953 */ 954 static struct bridge_iflist * 955 bridge_lookup_member_ifinfo(struct bridge_softc *sc, 956 struct bridge_ifinfo *bif_info) 957 { 958 struct bridge_iflist *bif; 959 960 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) { 961 if (bif->bif_info == bif_info) 962 return (bif); 963 } 964 return (NULL); 965 } 966 967 /* 968 * bridge_delete_member: 969 * 970 * Delete the specified member interface. 971 */ 972 static void 973 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif, 974 int gone) 975 { 976 struct ifnet *ifs = bif->bif_ifp; 977 struct ifnet *bifp = sc->sc_ifp; 978 struct bridge_ifinfo *bif_info = bif->bif_info; 979 struct bridge_iflist_head saved_bifs; 980 981 ASSERT_IFNET_SERIALIZED_ALL(bifp); 982 KKASSERT(bif_info != NULL); 983 984 ifs->if_bridge = NULL; 985 986 /* 987 * Release bridge interface's serializer: 988 * - To avoid possible dead lock. 989 * - Various sync operation will block the current thread. 990 */ 991 ifnet_deserialize_all(bifp); 992 993 if (!gone) { 994 switch (ifs->if_type) { 995 case IFT_ETHER: 996 case IFT_L2VLAN: 997 /* 998 * Take the interface out of promiscuous mode. 999 */ 1000 ifpromisc(ifs, 0); 1001 bridge_mutecaps(bif_info, ifs, 0); 1002 break; 1003 1004 case IFT_GIF: 1005 break; 1006 1007 default: 1008 panic("bridge_delete_member: impossible"); 1009 break; 1010 } 1011 } 1012 1013 /* 1014 * Remove bifs from percpu linked list. 1015 * 1016 * Removed bifs are not freed immediately, instead, 1017 * they are saved in saved_bifs. They will be freed 1018 * after we make sure that no one is accessing them, 1019 * i.e. after following netmsg_service_sync() 1020 */ 1021 LIST_INIT(&saved_bifs); 1022 bridge_del_bif(sc, bif_info, &saved_bifs); 1023 1024 /* 1025 * Make sure that all protocol threads: 1026 * o see 'ifs' if_bridge is changed 1027 * o know that bif is removed from the percpu linked list 1028 */ 1029 netmsg_service_sync(); 1030 1031 /* 1032 * Free the removed bifs 1033 */ 1034 KKASSERT(!LIST_EMPTY(&saved_bifs)); 1035 while ((bif = LIST_FIRST(&saved_bifs)) != NULL) { 1036 LIST_REMOVE(bif, bif_next); 1037 kfree(bif, M_DEVBUF); 1038 } 1039 1040 /* See the comment in bridge_ioctl_stop() */ 1041 bridge_rtmsg_sync(sc); 1042 bridge_rtdelete(sc, ifs, IFBF_FLUSHALL | IFBF_FLUSHSYNC); 1043 1044 ifnet_serialize_all(bifp); 1045 1046 if (bifp->if_flags & IFF_RUNNING) 1047 bstp_initialization(sc); 1048 1049 /* 1050 * Free the bif_info after bstp_initialization(), so that 1051 * bridge_softc.sc_root_port will not reference a dangling 1052 * pointer. 1053 */ 1054 kfree(bif_info, M_DEVBUF); 1055 } 1056 1057 /* 1058 * bridge_delete_span: 1059 * 1060 * Delete the specified span interface. 1061 */ 1062 static void 1063 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif) 1064 { 1065 KASSERT(bif->bif_ifp->if_bridge == NULL, 1066 ("%s: not a span interface", __func__)); 1067 1068 LIST_REMOVE(bif, bif_next); 1069 kfree(bif, M_DEVBUF); 1070 } 1071 1072 static int 1073 bridge_ioctl_init(struct bridge_softc *sc, void *arg __unused) 1074 { 1075 struct ifnet *ifp = sc->sc_ifp; 1076 1077 if (ifp->if_flags & IFF_RUNNING) 1078 return 0; 1079 1080 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz, 1081 bridge_timer, sc); 1082 1083 ifp->if_flags |= IFF_RUNNING; 1084 bstp_initialization(sc); 1085 return 0; 1086 } 1087 1088 static int 1089 bridge_ioctl_stop(struct bridge_softc *sc, void *arg __unused) 1090 { 1091 struct ifnet *ifp = sc->sc_ifp; 1092 struct lwkt_msg *lmsg; 1093 1094 if ((ifp->if_flags & IFF_RUNNING) == 0) 1095 return 0; 1096 1097 callout_stop(&sc->sc_brcallout); 1098 1099 crit_enter(); 1100 lmsg = &sc->sc_brtimemsg.lmsg; 1101 if ((lmsg->ms_flags & MSGF_DONE) == 0) { 1102 /* Pending to be processed; drop it */ 1103 lwkt_dropmsg(lmsg); 1104 } 1105 crit_exit(); 1106 1107 bstp_stop(sc); 1108 1109 ifp->if_flags &= ~IFF_RUNNING; 1110 1111 ifnet_deserialize_all(ifp); 1112 1113 /* Let everyone know that we are stopped */ 1114 netmsg_service_sync(); 1115 1116 /* 1117 * Sync ifnetX msgports in the order we forward rtnode 1118 * installation message. This is used to make sure that 1119 * all rtnode installation messages sent by bridge_rtupdate() 1120 * during above netmsg_service_sync() are flushed. 1121 */ 1122 bridge_rtmsg_sync(sc); 1123 bridge_rtflush(sc, IFBF_FLUSHDYN | IFBF_FLUSHSYNC); 1124 1125 ifnet_serialize_all(ifp); 1126 return 0; 1127 } 1128 1129 static int 1130 bridge_ioctl_add(struct bridge_softc *sc, void *arg) 1131 { 1132 struct ifbreq *req = arg; 1133 struct bridge_iflist *bif; 1134 struct bridge_ifinfo *bif_info; 1135 struct ifnet *ifs, *bifp; 1136 int error = 0; 1137 1138 bifp = sc->sc_ifp; 1139 ASSERT_IFNET_SERIALIZED_ALL(bifp); 1140 1141 ifs = ifunit(req->ifbr_ifsname); 1142 if (ifs == NULL) 1143 return (ENOENT); 1144 1145 /* If it's in the span list, it can't be a member. */ 1146 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1147 if (ifs == bif->bif_ifp) 1148 return (EBUSY); 1149 1150 /* Allow the first Ethernet member to define the MTU */ 1151 if (ifs->if_type != IFT_GIF) { 1152 if (LIST_EMPTY(&sc->sc_iflists[mycpuid])) { 1153 bifp->if_mtu = ifs->if_mtu; 1154 } else if (bifp->if_mtu != ifs->if_mtu) { 1155 if_printf(bifp, "invalid MTU for %s\n", ifs->if_xname); 1156 return (EINVAL); 1157 } 1158 } 1159 1160 if (ifs->if_bridge == sc) 1161 return (EEXIST); 1162 1163 if (ifs->if_bridge != NULL) 1164 return (EBUSY); 1165 1166 bif_info = kmalloc(sizeof(*bif_info), M_DEVBUF, M_WAITOK | M_ZERO); 1167 bif_info->bifi_priority = BSTP_DEFAULT_PORT_PRIORITY; 1168 bif_info->bifi_path_cost = BSTP_DEFAULT_PATH_COST; 1169 bif_info->bifi_ifp = ifs; 1170 1171 /* 1172 * Release bridge interface's serializer: 1173 * - To avoid possible dead lock. 1174 * - Various sync operation will block the current thread. 1175 */ 1176 ifnet_deserialize_all(bifp); 1177 1178 switch (ifs->if_type) { 1179 case IFT_ETHER: 1180 case IFT_L2VLAN: 1181 /* 1182 * Place the interface into promiscuous mode. 1183 */ 1184 error = ifpromisc(ifs, 1); 1185 if (error) { 1186 ifnet_serialize_all(bifp); 1187 goto out; 1188 } 1189 bridge_mutecaps(bif_info, ifs, 1); 1190 break; 1191 1192 case IFT_GIF: /* :^) */ 1193 break; 1194 1195 default: 1196 error = EINVAL; 1197 ifnet_serialize_all(bifp); 1198 goto out; 1199 } 1200 1201 /* 1202 * Add bifs to percpu linked lists 1203 */ 1204 bridge_add_bif(sc, bif_info, ifs); 1205 1206 ifnet_serialize_all(bifp); 1207 1208 if (bifp->if_flags & IFF_RUNNING) 1209 bstp_initialization(sc); 1210 else 1211 bstp_stop(sc); 1212 1213 /* 1214 * Everything has been setup, so let the member interface 1215 * deliver packets to this bridge on its input/output path. 1216 */ 1217 ifs->if_bridge = sc; 1218 out: 1219 if (error) { 1220 if (bif_info != NULL) 1221 kfree(bif_info, M_DEVBUF); 1222 } 1223 return (error); 1224 } 1225 1226 static int 1227 bridge_ioctl_del(struct bridge_softc *sc, void *arg) 1228 { 1229 struct ifbreq *req = arg; 1230 struct bridge_iflist *bif; 1231 1232 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1233 if (bif == NULL) 1234 return (ENOENT); 1235 1236 bridge_delete_member(sc, bif, 0); 1237 1238 return (0); 1239 } 1240 1241 static int 1242 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg) 1243 { 1244 struct ifbreq *req = arg; 1245 struct bridge_iflist *bif; 1246 1247 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1248 if (bif == NULL) 1249 return (ENOENT); 1250 1251 req->ifbr_ifsflags = bif->bif_flags; 1252 req->ifbr_state = bif->bif_state; 1253 req->ifbr_priority = bif->bif_priority; 1254 req->ifbr_path_cost = bif->bif_path_cost; 1255 req->ifbr_portno = bif->bif_ifp->if_index & 0xff; 1256 1257 return (0); 1258 } 1259 1260 static int 1261 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg) 1262 { 1263 struct ifbreq *req = arg; 1264 struct bridge_iflist *bif; 1265 struct ifnet *bifp = sc->sc_ifp; 1266 1267 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1268 if (bif == NULL) 1269 return (ENOENT); 1270 1271 if (req->ifbr_ifsflags & IFBIF_SPAN) { 1272 /* SPAN is readonly */ 1273 return (EINVAL); 1274 } 1275 1276 if (req->ifbr_ifsflags & IFBIF_STP) { 1277 switch (bif->bif_ifp->if_type) { 1278 case IFT_ETHER: 1279 /* These can do spanning tree. */ 1280 break; 1281 1282 default: 1283 /* Nothing else can. */ 1284 return (EINVAL); 1285 } 1286 } 1287 1288 ifnet_deserialize_all(bifp); 1289 bridge_set_bifflags(sc, bif->bif_info, req->ifbr_ifsflags); 1290 ifnet_serialize_all(bifp); 1291 1292 if (bifp->if_flags & IFF_RUNNING) 1293 bstp_initialization(sc); 1294 1295 return (0); 1296 } 1297 1298 static int 1299 bridge_ioctl_scache(struct bridge_softc *sc, void *arg) 1300 { 1301 struct ifbrparam *param = arg; 1302 struct ifnet *ifp = sc->sc_ifp; 1303 1304 sc->sc_brtmax = param->ifbrp_csize; 1305 1306 ifnet_deserialize_all(ifp); 1307 bridge_rttrim(sc); 1308 ifnet_serialize_all(ifp); 1309 1310 return (0); 1311 } 1312 1313 static int 1314 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg) 1315 { 1316 struct ifbrparam *param = arg; 1317 1318 param->ifbrp_csize = sc->sc_brtmax; 1319 1320 return (0); 1321 } 1322 1323 static int 1324 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg) 1325 { 1326 struct bridge_control_arg *bc_arg = arg; 1327 struct ifbifconf *bifc = arg; 1328 struct bridge_iflist *bif; 1329 struct ifbreq *breq; 1330 int count, len; 1331 1332 count = 0; 1333 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) 1334 count++; 1335 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1336 count++; 1337 1338 if (bifc->ifbic_len == 0) { 1339 bifc->ifbic_len = sizeof(*breq) * count; 1340 return 0; 1341 } else if (count == 0 || bifc->ifbic_len < sizeof(*breq)) { 1342 bifc->ifbic_len = 0; 1343 return 0; 1344 } 1345 1346 len = min(bifc->ifbic_len, sizeof(*breq) * count); 1347 KKASSERT(len >= sizeof(*breq)); 1348 1349 breq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO); 1350 if (breq == NULL) { 1351 bifc->ifbic_len = 0; 1352 return ENOMEM; 1353 } 1354 bc_arg->bca_kptr = breq; 1355 1356 count = 0; 1357 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) { 1358 if (len < sizeof(*breq)) 1359 break; 1360 1361 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname, 1362 sizeof(breq->ifbr_ifsname)); 1363 breq->ifbr_ifsflags = bif->bif_flags; 1364 breq->ifbr_state = bif->bif_state; 1365 breq->ifbr_priority = bif->bif_priority; 1366 breq->ifbr_path_cost = bif->bif_path_cost; 1367 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff; 1368 breq++; 1369 count++; 1370 len -= sizeof(*breq); 1371 } 1372 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) { 1373 if (len < sizeof(*breq)) 1374 break; 1375 1376 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname, 1377 sizeof(breq->ifbr_ifsname)); 1378 breq->ifbr_ifsflags = bif->bif_flags; 1379 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff; 1380 breq++; 1381 count++; 1382 len -= sizeof(*breq); 1383 } 1384 1385 bifc->ifbic_len = sizeof(*breq) * count; 1386 KKASSERT(bifc->ifbic_len > 0); 1387 1388 bc_arg->bca_len = bifc->ifbic_len; 1389 bc_arg->bca_uptr = bifc->ifbic_req; 1390 return 0; 1391 } 1392 1393 static int 1394 bridge_ioctl_rts(struct bridge_softc *sc, void *arg) 1395 { 1396 struct bridge_control_arg *bc_arg = arg; 1397 struct ifbaconf *bac = arg; 1398 struct bridge_rtnode *brt; 1399 struct ifbareq *bareq; 1400 int count, len; 1401 1402 count = 0; 1403 LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) 1404 count++; 1405 1406 if (bac->ifbac_len == 0) { 1407 bac->ifbac_len = sizeof(*bareq) * count; 1408 return 0; 1409 } else if (count == 0 || bac->ifbac_len < sizeof(*bareq)) { 1410 bac->ifbac_len = 0; 1411 return 0; 1412 } 1413 1414 len = min(bac->ifbac_len, sizeof(*bareq) * count); 1415 KKASSERT(len >= sizeof(*bareq)); 1416 1417 bareq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO); 1418 if (bareq == NULL) { 1419 bac->ifbac_len = 0; 1420 return ENOMEM; 1421 } 1422 bc_arg->bca_kptr = bareq; 1423 1424 count = 0; 1425 LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) { 1426 struct bridge_rtinfo *bri = brt->brt_info; 1427 unsigned long expire; 1428 1429 if (len < sizeof(*bareq)) 1430 break; 1431 1432 strlcpy(bareq->ifba_ifsname, bri->bri_ifp->if_xname, 1433 sizeof(bareq->ifba_ifsname)); 1434 memcpy(bareq->ifba_dst, brt->brt_addr, sizeof(brt->brt_addr)); 1435 expire = bri->bri_expire; 1436 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 1437 time_second < expire) 1438 bareq->ifba_expire = expire - time_second; 1439 else 1440 bareq->ifba_expire = 0; 1441 bareq->ifba_flags = bri->bri_flags; 1442 bareq++; 1443 count++; 1444 len -= sizeof(*bareq); 1445 } 1446 1447 bac->ifbac_len = sizeof(*bareq) * count; 1448 KKASSERT(bac->ifbac_len > 0); 1449 1450 bc_arg->bca_len = bac->ifbac_len; 1451 bc_arg->bca_uptr = bac->ifbac_req; 1452 return 0; 1453 } 1454 1455 static int 1456 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg) 1457 { 1458 struct ifbareq *req = arg; 1459 struct bridge_iflist *bif; 1460 struct ifnet *ifp = sc->sc_ifp; 1461 int error; 1462 1463 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1464 1465 bif = bridge_lookup_member(sc, req->ifba_ifsname); 1466 if (bif == NULL) 1467 return (ENOENT); 1468 1469 ifnet_deserialize_all(ifp); 1470 error = bridge_rtsaddr(sc, req->ifba_dst, bif->bif_ifp, 1471 req->ifba_flags); 1472 ifnet_serialize_all(ifp); 1473 return (error); 1474 } 1475 1476 static int 1477 bridge_ioctl_sto(struct bridge_softc *sc, void *arg) 1478 { 1479 struct ifbrparam *param = arg; 1480 1481 sc->sc_brttimeout = param->ifbrp_ctime; 1482 1483 return (0); 1484 } 1485 1486 static int 1487 bridge_ioctl_gto(struct bridge_softc *sc, void *arg) 1488 { 1489 struct ifbrparam *param = arg; 1490 1491 param->ifbrp_ctime = sc->sc_brttimeout; 1492 1493 return (0); 1494 } 1495 1496 static int 1497 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg) 1498 { 1499 struct ifbareq *req = arg; 1500 struct ifnet *ifp = sc->sc_ifp; 1501 int error; 1502 1503 ifnet_deserialize_all(ifp); 1504 error = bridge_rtdaddr(sc, req->ifba_dst); 1505 ifnet_serialize_all(ifp); 1506 return error; 1507 } 1508 1509 static int 1510 bridge_ioctl_flush(struct bridge_softc *sc, void *arg) 1511 { 1512 struct ifbreq *req = arg; 1513 struct ifnet *ifp = sc->sc_ifp; 1514 1515 ifnet_deserialize_all(ifp); 1516 bridge_rtflush(sc, req->ifbr_ifsflags | IFBF_FLUSHSYNC); 1517 ifnet_serialize_all(ifp); 1518 1519 return (0); 1520 } 1521 1522 static int 1523 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg) 1524 { 1525 struct ifbrparam *param = arg; 1526 1527 param->ifbrp_prio = sc->sc_bridge_priority; 1528 1529 return (0); 1530 } 1531 1532 static int 1533 bridge_ioctl_spri(struct bridge_softc *sc, void *arg) 1534 { 1535 struct ifbrparam *param = arg; 1536 1537 sc->sc_bridge_priority = param->ifbrp_prio; 1538 1539 if (sc->sc_ifp->if_flags & IFF_RUNNING) 1540 bstp_initialization(sc); 1541 1542 return (0); 1543 } 1544 1545 static int 1546 bridge_ioctl_ght(struct bridge_softc *sc, void *arg) 1547 { 1548 struct ifbrparam *param = arg; 1549 1550 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8; 1551 1552 return (0); 1553 } 1554 1555 static int 1556 bridge_ioctl_sht(struct bridge_softc *sc, void *arg) 1557 { 1558 struct ifbrparam *param = arg; 1559 1560 if (param->ifbrp_hellotime == 0) 1561 return (EINVAL); 1562 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8; 1563 1564 if (sc->sc_ifp->if_flags & IFF_RUNNING) 1565 bstp_initialization(sc); 1566 1567 return (0); 1568 } 1569 1570 static int 1571 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg) 1572 { 1573 struct ifbrparam *param = arg; 1574 1575 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8; 1576 1577 return (0); 1578 } 1579 1580 static int 1581 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg) 1582 { 1583 struct ifbrparam *param = arg; 1584 1585 if (param->ifbrp_fwddelay == 0) 1586 return (EINVAL); 1587 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8; 1588 1589 if (sc->sc_ifp->if_flags & IFF_RUNNING) 1590 bstp_initialization(sc); 1591 1592 return (0); 1593 } 1594 1595 static int 1596 bridge_ioctl_gma(struct bridge_softc *sc, void *arg) 1597 { 1598 struct ifbrparam *param = arg; 1599 1600 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8; 1601 1602 return (0); 1603 } 1604 1605 static int 1606 bridge_ioctl_sma(struct bridge_softc *sc, void *arg) 1607 { 1608 struct ifbrparam *param = arg; 1609 1610 if (param->ifbrp_maxage == 0) 1611 return (EINVAL); 1612 sc->sc_bridge_max_age = param->ifbrp_maxage << 8; 1613 1614 if (sc->sc_ifp->if_flags & IFF_RUNNING) 1615 bstp_initialization(sc); 1616 1617 return (0); 1618 } 1619 1620 static int 1621 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg) 1622 { 1623 struct ifbreq *req = arg; 1624 struct bridge_iflist *bif; 1625 1626 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1627 if (bif == NULL) 1628 return (ENOENT); 1629 1630 bif->bif_priority = req->ifbr_priority; 1631 1632 if (sc->sc_ifp->if_flags & IFF_RUNNING) 1633 bstp_initialization(sc); 1634 1635 return (0); 1636 } 1637 1638 static int 1639 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg) 1640 { 1641 struct ifbreq *req = arg; 1642 struct bridge_iflist *bif; 1643 1644 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1645 if (bif == NULL) 1646 return (ENOENT); 1647 1648 bif->bif_path_cost = req->ifbr_path_cost; 1649 1650 if (sc->sc_ifp->if_flags & IFF_RUNNING) 1651 bstp_initialization(sc); 1652 1653 return (0); 1654 } 1655 1656 static int 1657 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg) 1658 { 1659 struct ifbreq *req = arg; 1660 struct bridge_iflist *bif; 1661 struct ifnet *ifs; 1662 1663 ifs = ifunit(req->ifbr_ifsname); 1664 if (ifs == NULL) 1665 return (ENOENT); 1666 1667 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1668 if (ifs == bif->bif_ifp) 1669 return (EBUSY); 1670 1671 if (ifs->if_bridge != NULL) 1672 return (EBUSY); 1673 1674 switch (ifs->if_type) { 1675 case IFT_ETHER: 1676 case IFT_GIF: 1677 case IFT_L2VLAN: 1678 break; 1679 1680 default: 1681 return (EINVAL); 1682 } 1683 1684 bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO); 1685 bif->bif_ifp = ifs; 1686 bif->bif_flags = IFBIF_SPAN; 1687 /* NOTE: span bif does not need bridge_ifinfo */ 1688 1689 LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next); 1690 1691 sc->sc_span = 1; 1692 1693 return (0); 1694 } 1695 1696 static int 1697 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg) 1698 { 1699 struct ifbreq *req = arg; 1700 struct bridge_iflist *bif; 1701 struct ifnet *ifs; 1702 1703 ifs = ifunit(req->ifbr_ifsname); 1704 if (ifs == NULL) 1705 return (ENOENT); 1706 1707 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1708 if (ifs == bif->bif_ifp) 1709 break; 1710 1711 if (bif == NULL) 1712 return (ENOENT); 1713 1714 bridge_delete_span(sc, bif); 1715 1716 if (LIST_EMPTY(&sc->sc_spanlist)) 1717 sc->sc_span = 0; 1718 1719 return (0); 1720 } 1721 1722 static void 1723 bridge_ifdetach_dispatch(netmsg_t msg) 1724 { 1725 struct ifnet *ifp, *bifp; 1726 struct bridge_softc *sc; 1727 struct bridge_iflist *bif; 1728 1729 ifp = msg->lmsg.u.ms_resultp; 1730 sc = ifp->if_bridge; 1731 1732 /* Check if the interface is a bridge member */ 1733 if (sc != NULL) { 1734 bifp = sc->sc_ifp; 1735 1736 ifnet_serialize_all(bifp); 1737 1738 bif = bridge_lookup_member_if(sc, ifp); 1739 if (bif != NULL) { 1740 bridge_delete_member(sc, bif, 1); 1741 } else { 1742 /* XXX Why bif will be NULL? */ 1743 } 1744 1745 ifnet_deserialize_all(bifp); 1746 goto reply; 1747 } 1748 1749 crit_enter(); /* XXX MP */ 1750 1751 /* Check if the interface is a span port */ 1752 LIST_FOREACH(sc, &bridge_list, sc_list) { 1753 bifp = sc->sc_ifp; 1754 1755 ifnet_serialize_all(bifp); 1756 1757 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1758 if (ifp == bif->bif_ifp) { 1759 bridge_delete_span(sc, bif); 1760 break; 1761 } 1762 1763 ifnet_deserialize_all(bifp); 1764 } 1765 1766 crit_exit(); 1767 1768 reply: 1769 lwkt_replymsg(&msg->lmsg, 0); 1770 } 1771 1772 /* 1773 * bridge_ifdetach: 1774 * 1775 * Detach an interface from a bridge. Called when a member 1776 * interface is detaching. 1777 */ 1778 static void 1779 bridge_ifdetach(void *arg __unused, struct ifnet *ifp) 1780 { 1781 struct netmsg_base msg; 1782 1783 netmsg_init(&msg, NULL, &curthread->td_msgport, 1784 0, bridge_ifdetach_dispatch); 1785 msg.lmsg.u.ms_resultp = ifp; 1786 1787 lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0); 1788 } 1789 1790 /* 1791 * bridge_init: 1792 * 1793 * Initialize a bridge interface. 1794 */ 1795 static void 1796 bridge_init(void *xsc) 1797 { 1798 bridge_control(xsc, SIOCSIFFLAGS, bridge_ioctl_init, NULL); 1799 } 1800 1801 /* 1802 * bridge_stop: 1803 * 1804 * Stop the bridge interface. 1805 */ 1806 static void 1807 bridge_stop(struct ifnet *ifp) 1808 { 1809 bridge_control(ifp->if_softc, SIOCSIFFLAGS, bridge_ioctl_stop, NULL); 1810 } 1811 1812 /* 1813 * bridge_enqueue: 1814 * 1815 * Enqueue a packet on a bridge member interface. 1816 * 1817 */ 1818 void 1819 bridge_enqueue(struct ifnet *dst_ifp, struct mbuf *m) 1820 { 1821 struct netmsg_packet *nmp; 1822 1823 nmp = &m->m_hdr.mh_netmsg; 1824 netmsg_init(&nmp->base, NULL, &netisr_apanic_rport, 1825 0, bridge_enqueue_handler); 1826 nmp->nm_packet = m; 1827 nmp->base.lmsg.u.ms_resultp = dst_ifp; 1828 1829 lwkt_sendmsg(ifnet_portfn(mycpu->gd_cpuid), &nmp->base.lmsg); 1830 } 1831 1832 /* 1833 * bridge_output: 1834 * 1835 * Send output from a bridge member interface. This 1836 * performs the bridging function for locally originated 1837 * packets. 1838 * 1839 * The mbuf has the Ethernet header already attached. We must 1840 * enqueue or free the mbuf before returning. 1841 */ 1842 static int 1843 bridge_output(struct ifnet *ifp, struct mbuf *m) 1844 { 1845 struct bridge_softc *sc = ifp->if_bridge; 1846 struct ether_header *eh; 1847 struct ifnet *dst_if, *bifp; 1848 1849 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 1850 1851 /* 1852 * Make sure that we are still a member of a bridge interface. 1853 */ 1854 if (sc == NULL) { 1855 m_freem(m); 1856 return (0); 1857 } 1858 bifp = sc->sc_ifp; 1859 1860 if (m->m_len < ETHER_HDR_LEN) { 1861 m = m_pullup(m, ETHER_HDR_LEN); 1862 if (m == NULL) 1863 return (0); 1864 } 1865 eh = mtod(m, struct ether_header *); 1866 1867 /* 1868 * If bridge is down, but the original output interface is up, 1869 * go ahead and send out that interface. Otherwise, the packet 1870 * is dropped below. 1871 */ 1872 if ((bifp->if_flags & IFF_RUNNING) == 0) { 1873 dst_if = ifp; 1874 goto sendunicast; 1875 } 1876 1877 /* 1878 * If the packet is a multicast, or we don't know a better way to 1879 * get there, send to all interfaces. 1880 */ 1881 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1882 dst_if = NULL; 1883 else 1884 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1885 if (dst_if == NULL) { 1886 struct bridge_iflist *bif, *nbif; 1887 struct mbuf *mc; 1888 int used = 0; 1889 1890 if (sc->sc_span) 1891 bridge_span(sc, m); 1892 1893 LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], 1894 bif_next, nbif) { 1895 dst_if = bif->bif_ifp; 1896 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1897 continue; 1898 1899 /* 1900 * If this is not the original output interface, 1901 * and the interface is participating in spanning 1902 * tree, make sure the port is in a state that 1903 * allows forwarding. 1904 */ 1905 if (dst_if != ifp && 1906 (bif->bif_flags & IFBIF_STP) != 0) { 1907 switch (bif->bif_state) { 1908 case BSTP_IFSTATE_BLOCKING: 1909 case BSTP_IFSTATE_LISTENING: 1910 case BSTP_IFSTATE_DISABLED: 1911 continue; 1912 } 1913 } 1914 1915 if (LIST_NEXT(bif, bif_next) == NULL) { 1916 used = 1; 1917 mc = m; 1918 } else { 1919 mc = m_copypacket(m, MB_DONTWAIT); 1920 if (mc == NULL) { 1921 bifp->if_oerrors++; 1922 continue; 1923 } 1924 } 1925 bridge_handoff(dst_if, mc); 1926 1927 if (nbif != NULL && !nbif->bif_onlist) { 1928 KKASSERT(bif->bif_onlist); 1929 nbif = LIST_NEXT(bif, bif_next); 1930 } 1931 } 1932 if (used == 0) 1933 m_freem(m); 1934 return (0); 1935 } 1936 1937 sendunicast: 1938 /* 1939 * XXX Spanning tree consideration here? 1940 */ 1941 if (sc->sc_span) 1942 bridge_span(sc, m); 1943 if ((dst_if->if_flags & IFF_RUNNING) == 0) 1944 m_freem(m); 1945 else 1946 bridge_handoff(dst_if, m); 1947 return (0); 1948 } 1949 1950 /* 1951 * bridge_start: 1952 * 1953 * Start output on a bridge. 1954 * 1955 */ 1956 static void 1957 bridge_start(struct ifnet *ifp) 1958 { 1959 struct bridge_softc *sc = ifp->if_softc; 1960 1961 ASSERT_IFNET_SERIALIZED_TX(ifp); 1962 1963 ifp->if_flags |= IFF_OACTIVE; 1964 for (;;) { 1965 struct ifnet *dst_if = NULL; 1966 struct ether_header *eh; 1967 struct mbuf *m; 1968 1969 m = ifq_dequeue(&ifp->if_snd, NULL); 1970 if (m == NULL) 1971 break; 1972 1973 if (m->m_len < sizeof(*eh)) { 1974 m = m_pullup(m, sizeof(*eh)); 1975 if (m == NULL) { 1976 ifp->if_oerrors++; 1977 continue; 1978 } 1979 } 1980 eh = mtod(m, struct ether_header *); 1981 1982 BPF_MTAP(ifp, m); 1983 ifp->if_opackets++; 1984 1985 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) 1986 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 1987 1988 if (dst_if == NULL) 1989 bridge_start_bcast(sc, m); 1990 else 1991 bridge_enqueue(dst_if, m); 1992 } 1993 ifp->if_flags &= ~IFF_OACTIVE; 1994 } 1995 1996 /* 1997 * bridge_forward: 1998 * 1999 * The forwarding function of the bridge. 2000 */ 2001 static void 2002 bridge_forward(struct bridge_softc *sc, struct mbuf *m) 2003 { 2004 struct bridge_iflist *bif; 2005 struct ifnet *src_if, *dst_if, *ifp; 2006 struct ether_header *eh; 2007 2008 src_if = m->m_pkthdr.rcvif; 2009 ifp = sc->sc_ifp; 2010 2011 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 2012 2013 ifp->if_ipackets++; 2014 ifp->if_ibytes += m->m_pkthdr.len; 2015 2016 /* 2017 * Look up the bridge_iflist. 2018 */ 2019 bif = bridge_lookup_member_if(sc, src_if); 2020 if (bif == NULL) { 2021 /* Interface is not a bridge member (anymore?) */ 2022 m_freem(m); 2023 return; 2024 } 2025 2026 if (bif->bif_flags & IFBIF_STP) { 2027 switch (bif->bif_state) { 2028 case BSTP_IFSTATE_BLOCKING: 2029 case BSTP_IFSTATE_LISTENING: 2030 case BSTP_IFSTATE_DISABLED: 2031 m_freem(m); 2032 return; 2033 } 2034 } 2035 2036 eh = mtod(m, struct ether_header *); 2037 2038 /* 2039 * If the interface is learning, and the source 2040 * address is valid and not multicast, record 2041 * the address. 2042 */ 2043 if ((bif->bif_flags & IFBIF_LEARNING) != 0 && 2044 ETHER_IS_MULTICAST(eh->ether_shost) == 0 && 2045 (eh->ether_shost[0] == 0 && 2046 eh->ether_shost[1] == 0 && 2047 eh->ether_shost[2] == 0 && 2048 eh->ether_shost[3] == 0 && 2049 eh->ether_shost[4] == 0 && 2050 eh->ether_shost[5] == 0) == 0) 2051 bridge_rtupdate(sc, eh->ether_shost, src_if, IFBAF_DYNAMIC); 2052 2053 if ((bif->bif_flags & IFBIF_STP) != 0 && 2054 bif->bif_state == BSTP_IFSTATE_LEARNING) { 2055 m_freem(m); 2056 return; 2057 } 2058 2059 /* 2060 * At this point, the port either doesn't participate 2061 * in spanning tree or it is in the forwarding state. 2062 */ 2063 2064 /* 2065 * If the packet is unicast, destined for someone on 2066 * "this" side of the bridge, drop it. 2067 */ 2068 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) { 2069 dst_if = bridge_rtlookup(sc, eh->ether_dhost); 2070 if (src_if == dst_if) { 2071 m_freem(m); 2072 return; 2073 } 2074 } else { 2075 /* ...forward it to all interfaces. */ 2076 ifp->if_imcasts++; 2077 dst_if = NULL; 2078 } 2079 2080 if (dst_if == NULL) { 2081 bridge_broadcast(sc, src_if, m); 2082 return; 2083 } 2084 2085 /* 2086 * At this point, we're dealing with a unicast frame 2087 * going to a different interface. 2088 */ 2089 if ((dst_if->if_flags & IFF_RUNNING) == 0) { 2090 m_freem(m); 2091 return; 2092 } 2093 bif = bridge_lookup_member_if(sc, dst_if); 2094 if (bif == NULL) { 2095 /* Not a member of the bridge (anymore?) */ 2096 m_freem(m); 2097 return; 2098 } 2099 2100 if (bif->bif_flags & IFBIF_STP) { 2101 switch (bif->bif_state) { 2102 case BSTP_IFSTATE_DISABLED: 2103 case BSTP_IFSTATE_BLOCKING: 2104 m_freem(m); 2105 return; 2106 } 2107 } 2108 2109 if (inet_pfil_hook.ph_hashooks > 0 2110 #ifdef INET6 2111 || inet6_pfil_hook.ph_hashooks > 0 2112 #endif 2113 ) { 2114 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0) 2115 return; 2116 if (m == NULL) 2117 return; 2118 2119 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0) 2120 return; 2121 if (m == NULL) 2122 return; 2123 } 2124 bridge_handoff(dst_if, m); 2125 } 2126 2127 /* 2128 * bridge_input: 2129 * 2130 * Receive input from a member interface. Queue the packet for 2131 * bridging if it is not for us. 2132 */ 2133 static struct mbuf * 2134 bridge_input(struct ifnet *ifp, struct mbuf *m) 2135 { 2136 struct bridge_softc *sc = ifp->if_bridge; 2137 struct bridge_iflist *bif; 2138 struct ifnet *bifp, *new_ifp; 2139 struct ether_header *eh; 2140 struct mbuf *mc, *mc2; 2141 2142 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 2143 2144 /* 2145 * Make sure that we are still a member of a bridge interface. 2146 */ 2147 if (sc == NULL) 2148 return m; 2149 2150 new_ifp = NULL; 2151 bifp = sc->sc_ifp; 2152 2153 if ((bifp->if_flags & IFF_RUNNING) == 0) 2154 goto out; 2155 2156 /* 2157 * Implement support for bridge monitoring. If this flag has been 2158 * set on this interface, discard the packet once we push it through 2159 * the bpf(4) machinery, but before we do, increment various counters 2160 * associated with this bridge. 2161 */ 2162 if (bifp->if_flags & IFF_MONITOR) { 2163 /* Change input interface to this bridge */ 2164 m->m_pkthdr.rcvif = bifp; 2165 2166 BPF_MTAP(bifp, m); 2167 2168 /* Update bridge's ifnet statistics */ 2169 bifp->if_ipackets++; 2170 bifp->if_ibytes += m->m_pkthdr.len; 2171 if (m->m_flags & (M_MCAST | M_BCAST)) 2172 bifp->if_imcasts++; 2173 2174 m_freem(m); 2175 m = NULL; 2176 goto out; 2177 } 2178 2179 eh = mtod(m, struct ether_header *); 2180 2181 if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) { 2182 /* 2183 * If the packet is for us, set the packets source as the 2184 * bridge, and return the packet back to ifnet.if_input for 2185 * local processing. 2186 */ 2187 KASSERT(bifp->if_bridge == NULL, 2188 ("loop created in bridge_input")); 2189 if (pfil_member != 0) { 2190 if (inet_pfil_hook.ph_hashooks > 0 2191 #ifdef INET6 2192 || inet6_pfil_hook.ph_hashooks > 0 2193 #endif 2194 ) { 2195 if (bridge_pfil(&m, NULL, ifp, PFIL_IN) != 0) 2196 goto out; 2197 if (m == NULL) 2198 goto out; 2199 } 2200 } 2201 new_ifp = bifp; 2202 goto out; 2203 } 2204 2205 /* 2206 * Tap all packets arriving on the bridge, no matter if 2207 * they are local destinations or not. In is in. 2208 */ 2209 BPF_MTAP(bifp, m); 2210 2211 bif = bridge_lookup_member_if(sc, ifp); 2212 if (bif == NULL) 2213 goto out; 2214 2215 if (sc->sc_span) 2216 bridge_span(sc, m); 2217 2218 if (m->m_flags & (M_BCAST | M_MCAST)) { 2219 /* Tap off 802.1D packets; they do not get forwarded. */ 2220 if (memcmp(eh->ether_dhost, bstp_etheraddr, 2221 ETHER_ADDR_LEN) == 0) { 2222 ifnet_serialize_all(bifp); 2223 bstp_input(sc, bif, m); 2224 ifnet_deserialize_all(bifp); 2225 2226 /* m is freed by bstp_input */ 2227 m = NULL; 2228 goto out; 2229 } 2230 2231 if (bif->bif_flags & IFBIF_STP) { 2232 switch (bif->bif_state) { 2233 case BSTP_IFSTATE_BLOCKING: 2234 case BSTP_IFSTATE_LISTENING: 2235 case BSTP_IFSTATE_DISABLED: 2236 goto out; 2237 } 2238 } 2239 2240 /* 2241 * Make a deep copy of the packet and enqueue the copy 2242 * for bridge processing; return the original packet for 2243 * local processing. 2244 */ 2245 mc = m_dup(m, MB_DONTWAIT); 2246 if (mc == NULL) 2247 goto out; 2248 2249 bridge_forward(sc, mc); 2250 2251 /* 2252 * Reinject the mbuf as arriving on the bridge so we have a 2253 * chance at claiming multicast packets. We can not loop back 2254 * here from ether_input as a bridge is never a member of a 2255 * bridge. 2256 */ 2257 KASSERT(bifp->if_bridge == NULL, 2258 ("loop created in bridge_input")); 2259 mc2 = m_dup(m, MB_DONTWAIT); 2260 #ifdef notyet 2261 if (mc2 != NULL) { 2262 /* Keep the layer3 header aligned */ 2263 int i = min(mc2->m_pkthdr.len, max_protohdr); 2264 mc2 = m_copyup(mc2, i, ETHER_ALIGN); 2265 } 2266 #endif 2267 if (mc2 != NULL) { 2268 /* 2269 * Don't tap to bpf(4) again; we have 2270 * already done the tapping. 2271 */ 2272 ether_reinput_oncpu(bifp, mc2, 0); 2273 } 2274 2275 /* Return the original packet for local processing. */ 2276 goto out; 2277 } 2278 2279 if (bif->bif_flags & IFBIF_STP) { 2280 switch (bif->bif_state) { 2281 case BSTP_IFSTATE_BLOCKING: 2282 case BSTP_IFSTATE_LISTENING: 2283 case BSTP_IFSTATE_DISABLED: 2284 goto out; 2285 } 2286 } 2287 2288 /* 2289 * Unicast. Make sure it's not for us. 2290 * 2291 * This loop is MPSAFE; the only blocking operation (bridge_rtupdate) 2292 * is followed by breaking out of the loop. 2293 */ 2294 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) { 2295 if (bif->bif_ifp->if_type != IFT_ETHER) 2296 continue; 2297 2298 /* It is destined for us. */ 2299 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost, 2300 ETHER_ADDR_LEN) == 0) { 2301 if (bif->bif_ifp != ifp) { 2302 /* XXX loop prevention */ 2303 m->m_flags |= M_ETHER_BRIDGED; 2304 new_ifp = bif->bif_ifp; 2305 } 2306 if (bif->bif_flags & IFBIF_LEARNING) { 2307 bridge_rtupdate(sc, eh->ether_shost, 2308 ifp, IFBAF_DYNAMIC); 2309 } 2310 goto out; 2311 } 2312 2313 /* We just received a packet that we sent out. */ 2314 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost, 2315 ETHER_ADDR_LEN) == 0) { 2316 m_freem(m); 2317 m = NULL; 2318 goto out; 2319 } 2320 } 2321 2322 /* Perform the bridge forwarding function. */ 2323 bridge_forward(sc, m); 2324 m = NULL; 2325 out: 2326 if (new_ifp != NULL) { 2327 ether_reinput_oncpu(new_ifp, m, 1); 2328 m = NULL; 2329 } 2330 return (m); 2331 } 2332 2333 /* 2334 * bridge_start_bcast: 2335 * 2336 * Broadcast the packet sent from bridge to all member 2337 * interfaces. 2338 * This is a simplified version of bridge_broadcast(), however, 2339 * this function expects caller to hold bridge's serializer. 2340 */ 2341 static void 2342 bridge_start_bcast(struct bridge_softc *sc, struct mbuf *m) 2343 { 2344 struct bridge_iflist *bif; 2345 struct mbuf *mc; 2346 struct ifnet *dst_if, *bifp; 2347 int used = 0; 2348 2349 bifp = sc->sc_ifp; 2350 ASSERT_IFNET_SERIALIZED_ALL(bifp); 2351 2352 /* 2353 * Following loop is MPSAFE; nothing is blocking 2354 * in the loop body. 2355 */ 2356 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) { 2357 dst_if = bif->bif_ifp; 2358 2359 if (bif->bif_flags & IFBIF_STP) { 2360 switch (bif->bif_state) { 2361 case BSTP_IFSTATE_BLOCKING: 2362 case BSTP_IFSTATE_DISABLED: 2363 continue; 2364 } 2365 } 2366 2367 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && 2368 (m->m_flags & (M_BCAST|M_MCAST)) == 0) 2369 continue; 2370 2371 if ((dst_if->if_flags & IFF_RUNNING) == 0) 2372 continue; 2373 2374 if (LIST_NEXT(bif, bif_next) == NULL) { 2375 mc = m; 2376 used = 1; 2377 } else { 2378 mc = m_copypacket(m, MB_DONTWAIT); 2379 if (mc == NULL) { 2380 bifp->if_oerrors++; 2381 continue; 2382 } 2383 } 2384 bridge_enqueue(dst_if, mc); 2385 } 2386 if (used == 0) 2387 m_freem(m); 2388 } 2389 2390 /* 2391 * bridge_broadcast: 2392 * 2393 * Send a frame to all interfaces that are members of 2394 * the bridge, except for the one on which the packet 2395 * arrived. 2396 */ 2397 static void 2398 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if, 2399 struct mbuf *m) 2400 { 2401 struct bridge_iflist *bif, *nbif; 2402 struct mbuf *mc; 2403 struct ifnet *dst_if, *bifp; 2404 int used = 0; 2405 2406 bifp = sc->sc_ifp; 2407 ASSERT_IFNET_NOT_SERIALIZED_ALL(bifp); 2408 2409 if (inet_pfil_hook.ph_hashooks > 0 2410 #ifdef INET6 2411 || inet6_pfil_hook.ph_hashooks > 0 2412 #endif 2413 ) { 2414 if (bridge_pfil(&m, bifp, src_if, PFIL_IN) != 0) 2415 return; 2416 if (m == NULL) 2417 return; 2418 2419 /* Filter on the bridge interface before broadcasting */ 2420 if (bridge_pfil(&m, bifp, NULL, PFIL_OUT) != 0) 2421 return; 2422 if (m == NULL) 2423 return; 2424 } 2425 2426 LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) { 2427 dst_if = bif->bif_ifp; 2428 if (dst_if == src_if) 2429 continue; 2430 2431 if (bif->bif_flags & IFBIF_STP) { 2432 switch (bif->bif_state) { 2433 case BSTP_IFSTATE_BLOCKING: 2434 case BSTP_IFSTATE_DISABLED: 2435 continue; 2436 } 2437 } 2438 2439 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && 2440 (m->m_flags & (M_BCAST|M_MCAST)) == 0) 2441 continue; 2442 2443 if ((dst_if->if_flags & IFF_RUNNING) == 0) 2444 continue; 2445 2446 if (LIST_NEXT(bif, bif_next) == NULL) { 2447 mc = m; 2448 used = 1; 2449 } else { 2450 mc = m_copypacket(m, MB_DONTWAIT); 2451 if (mc == NULL) { 2452 sc->sc_ifp->if_oerrors++; 2453 continue; 2454 } 2455 } 2456 2457 /* 2458 * Filter on the output interface. Pass a NULL bridge 2459 * interface pointer so we do not redundantly filter on 2460 * the bridge for each interface we broadcast on. 2461 */ 2462 if (inet_pfil_hook.ph_hashooks > 0 2463 #ifdef INET6 2464 || inet6_pfil_hook.ph_hashooks > 0 2465 #endif 2466 ) { 2467 if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0) 2468 continue; 2469 if (mc == NULL) 2470 continue; 2471 } 2472 bridge_handoff(dst_if, mc); 2473 2474 if (nbif != NULL && !nbif->bif_onlist) { 2475 KKASSERT(bif->bif_onlist); 2476 nbif = LIST_NEXT(bif, bif_next); 2477 } 2478 } 2479 if (used == 0) 2480 m_freem(m); 2481 } 2482 2483 /* 2484 * bridge_span: 2485 * 2486 * Duplicate a packet out one or more interfaces that are in span mode, 2487 * the original mbuf is unmodified. 2488 */ 2489 static void 2490 bridge_span(struct bridge_softc *sc, struct mbuf *m) 2491 { 2492 struct bridge_iflist *bif; 2493 struct ifnet *dst_if, *bifp; 2494 struct mbuf *mc; 2495 2496 bifp = sc->sc_ifp; 2497 ifnet_serialize_all(bifp); 2498 2499 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) { 2500 dst_if = bif->bif_ifp; 2501 2502 if ((dst_if->if_flags & IFF_RUNNING) == 0) 2503 continue; 2504 2505 mc = m_copypacket(m, MB_DONTWAIT); 2506 if (mc == NULL) { 2507 sc->sc_ifp->if_oerrors++; 2508 continue; 2509 } 2510 bridge_enqueue(dst_if, mc); 2511 } 2512 2513 ifnet_deserialize_all(bifp); 2514 } 2515 2516 static void 2517 bridge_rtmsg_sync_handler(netmsg_t msg) 2518 { 2519 ifnet_forwardmsg(&msg->lmsg, mycpuid + 1); 2520 } 2521 2522 static void 2523 bridge_rtmsg_sync(struct bridge_softc *sc) 2524 { 2525 struct netmsg_base msg; 2526 2527 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 2528 2529 netmsg_init(&msg, NULL, &curthread->td_msgport, 2530 0, bridge_rtmsg_sync_handler); 2531 ifnet_domsg(&msg.lmsg, 0); 2532 } 2533 2534 static __inline void 2535 bridge_rtinfo_update(struct bridge_rtinfo *bri, struct ifnet *dst_if, 2536 int setflags, uint8_t flags, uint32_t timeo) 2537 { 2538 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 2539 bri->bri_ifp != dst_if) 2540 bri->bri_ifp = dst_if; 2541 if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 2542 bri->bri_expire != time_second + timeo) 2543 bri->bri_expire = time_second + timeo; 2544 if (setflags) 2545 bri->bri_flags = flags; 2546 } 2547 2548 static int 2549 bridge_rtinstall_oncpu(struct bridge_softc *sc, const uint8_t *dst, 2550 struct ifnet *dst_if, int setflags, uint8_t flags, 2551 struct bridge_rtinfo **bri0) 2552 { 2553 struct bridge_rtnode *brt; 2554 struct bridge_rtinfo *bri; 2555 2556 if (mycpuid == 0) { 2557 brt = bridge_rtnode_lookup(sc, dst); 2558 if (brt != NULL) { 2559 /* 2560 * rtnode for 'dst' already exists. We inform the 2561 * caller about this by leaving bri0 as NULL. The 2562 * caller will terminate the intallation upon getting 2563 * NULL bri0. However, we still need to update the 2564 * rtinfo. 2565 */ 2566 KKASSERT(*bri0 == NULL); 2567 2568 /* Update rtinfo */ 2569 bridge_rtinfo_update(brt->brt_info, dst_if, setflags, 2570 flags, sc->sc_brttimeout); 2571 return 0; 2572 } 2573 2574 /* 2575 * We only need to check brtcnt on CPU0, since if limit 2576 * is to be exceeded, ENOSPC is returned. Caller knows 2577 * this and will terminate the installation. 2578 */ 2579 if (sc->sc_brtcnt >= sc->sc_brtmax) 2580 return ENOSPC; 2581 2582 KKASSERT(*bri0 == NULL); 2583 bri = kmalloc(sizeof(struct bridge_rtinfo), M_DEVBUF, 2584 M_WAITOK | M_ZERO); 2585 *bri0 = bri; 2586 2587 /* Setup rtinfo */ 2588 bri->bri_flags = IFBAF_DYNAMIC; 2589 bridge_rtinfo_update(bri, dst_if, setflags, flags, 2590 sc->sc_brttimeout); 2591 } else { 2592 bri = *bri0; 2593 KKASSERT(bri != NULL); 2594 } 2595 2596 brt = kmalloc(sizeof(struct bridge_rtnode), M_DEVBUF, 2597 M_WAITOK | M_ZERO); 2598 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN); 2599 brt->brt_info = bri; 2600 2601 bridge_rtnode_insert(sc, brt); 2602 return 0; 2603 } 2604 2605 static void 2606 bridge_rtinstall_handler(netmsg_t msg) 2607 { 2608 struct netmsg_brsaddr *brmsg = (struct netmsg_brsaddr *)msg; 2609 int error; 2610 2611 error = bridge_rtinstall_oncpu(brmsg->br_softc, 2612 brmsg->br_dst, brmsg->br_dst_if, 2613 brmsg->br_setflags, brmsg->br_flags, 2614 &brmsg->br_rtinfo); 2615 if (error) { 2616 KKASSERT(mycpuid == 0 && brmsg->br_rtinfo == NULL); 2617 lwkt_replymsg(&brmsg->base.lmsg, error); 2618 return; 2619 } else if (brmsg->br_rtinfo == NULL) { 2620 /* rtnode already exists for 'dst' */ 2621 KKASSERT(mycpuid == 0); 2622 lwkt_replymsg(&brmsg->base.lmsg, 0); 2623 return; 2624 } 2625 ifnet_forwardmsg(&brmsg->base.lmsg, mycpuid + 1); 2626 } 2627 2628 /* 2629 * bridge_rtupdate: 2630 * 2631 * Add/Update a bridge routing entry. 2632 */ 2633 static int 2634 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, 2635 struct ifnet *dst_if, uint8_t flags) 2636 { 2637 struct bridge_rtnode *brt; 2638 2639 /* 2640 * A route for this destination might already exist. If so, 2641 * update it, otherwise create a new one. 2642 */ 2643 if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) { 2644 struct netmsg_brsaddr *brmsg; 2645 2646 if (sc->sc_brtcnt >= sc->sc_brtmax) 2647 return ENOSPC; 2648 2649 brmsg = kmalloc(sizeof(*brmsg), M_LWKTMSG, M_WAITOK | M_NULLOK); 2650 if (brmsg == NULL) 2651 return ENOMEM; 2652 2653 netmsg_init(&brmsg->base, NULL, &netisr_afree_rport, 2654 0, bridge_rtinstall_handler); 2655 memcpy(brmsg->br_dst, dst, ETHER_ADDR_LEN); 2656 brmsg->br_dst_if = dst_if; 2657 brmsg->br_flags = flags; 2658 brmsg->br_setflags = 0; 2659 brmsg->br_softc = sc; 2660 brmsg->br_rtinfo = NULL; 2661 2662 ifnet_sendmsg(&brmsg->base.lmsg, 0); 2663 return 0; 2664 } 2665 bridge_rtinfo_update(brt->brt_info, dst_if, 0, flags, 2666 sc->sc_brttimeout); 2667 return 0; 2668 } 2669 2670 static int 2671 bridge_rtsaddr(struct bridge_softc *sc, const uint8_t *dst, 2672 struct ifnet *dst_if, uint8_t flags) 2673 { 2674 struct netmsg_brsaddr brmsg; 2675 2676 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 2677 2678 netmsg_init(&brmsg.base, NULL, &curthread->td_msgport, 2679 0, bridge_rtinstall_handler); 2680 memcpy(brmsg.br_dst, dst, ETHER_ADDR_LEN); 2681 brmsg.br_dst_if = dst_if; 2682 brmsg.br_flags = flags; 2683 brmsg.br_setflags = 1; 2684 brmsg.br_softc = sc; 2685 brmsg.br_rtinfo = NULL; 2686 2687 return ifnet_domsg(&brmsg.base.lmsg, 0); 2688 } 2689 2690 /* 2691 * bridge_rtlookup: 2692 * 2693 * Lookup the destination interface for an address. 2694 */ 2695 static struct ifnet * 2696 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr) 2697 { 2698 struct bridge_rtnode *brt; 2699 2700 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) 2701 return NULL; 2702 return brt->brt_info->bri_ifp; 2703 } 2704 2705 static void 2706 bridge_rtreap_handler(netmsg_t msg) 2707 { 2708 struct bridge_softc *sc = msg->lmsg.u.ms_resultp; 2709 struct bridge_rtnode *brt, *nbrt; 2710 2711 LIST_FOREACH_MUTABLE(brt, &sc->sc_rtlists[mycpuid], brt_list, nbrt) { 2712 if (brt->brt_info->bri_dead) 2713 bridge_rtnode_destroy(sc, brt); 2714 } 2715 ifnet_forwardmsg(&msg->lmsg, mycpuid + 1); 2716 } 2717 2718 static void 2719 bridge_rtreap(struct bridge_softc *sc) 2720 { 2721 struct netmsg_base msg; 2722 2723 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 2724 2725 netmsg_init(&msg, NULL, &curthread->td_msgport, 2726 0, bridge_rtreap_handler); 2727 msg.lmsg.u.ms_resultp = sc; 2728 2729 ifnet_domsg(&msg.lmsg, 0); 2730 } 2731 2732 static void 2733 bridge_rtreap_async(struct bridge_softc *sc) 2734 { 2735 struct netmsg_base *msg; 2736 2737 msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_WAITOK); 2738 2739 netmsg_init(msg, NULL, &netisr_afree_rport, 2740 0, bridge_rtreap_handler); 2741 msg->lmsg.u.ms_resultp = sc; 2742 2743 ifnet_sendmsg(&msg->lmsg, 0); 2744 } 2745 2746 /* 2747 * bridge_rttrim: 2748 * 2749 * Trim the routine table so that we have a number 2750 * of routing entries less than or equal to the 2751 * maximum number. 2752 */ 2753 static void 2754 bridge_rttrim(struct bridge_softc *sc) 2755 { 2756 struct bridge_rtnode *brt; 2757 int dead; 2758 2759 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 2760 2761 /* Make sure we actually need to do this. */ 2762 if (sc->sc_brtcnt <= sc->sc_brtmax) 2763 return; 2764 2765 /* 2766 * Find out how many rtnodes are dead 2767 */ 2768 dead = bridge_rtage_finddead(sc); 2769 KKASSERT(dead <= sc->sc_brtcnt); 2770 2771 if (sc->sc_brtcnt - dead <= sc->sc_brtmax) { 2772 /* Enough dead rtnodes are found */ 2773 bridge_rtreap(sc); 2774 return; 2775 } 2776 2777 /* 2778 * Kill some dynamic rtnodes to meet the brtmax 2779 */ 2780 LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) { 2781 struct bridge_rtinfo *bri = brt->brt_info; 2782 2783 if (bri->bri_dead) { 2784 /* 2785 * We have counted this rtnode in 2786 * bridge_rtage_finddead() 2787 */ 2788 continue; 2789 } 2790 2791 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 2792 bri->bri_dead = 1; 2793 ++dead; 2794 KKASSERT(dead <= sc->sc_brtcnt); 2795 2796 if (sc->sc_brtcnt - dead <= sc->sc_brtmax) { 2797 /* Enough rtnodes are collected */ 2798 break; 2799 } 2800 } 2801 } 2802 if (dead) 2803 bridge_rtreap(sc); 2804 } 2805 2806 /* 2807 * bridge_timer: 2808 * 2809 * Aging timer for the bridge. 2810 */ 2811 static void 2812 bridge_timer(void *arg) 2813 { 2814 struct bridge_softc *sc = arg; 2815 struct netmsg_base *msg; 2816 2817 KKASSERT(mycpuid == BRIDGE_CFGCPU); 2818 2819 crit_enter(); 2820 2821 if (callout_pending(&sc->sc_brcallout) || 2822 !callout_active(&sc->sc_brcallout)) { 2823 crit_exit(); 2824 return; 2825 } 2826 callout_deactivate(&sc->sc_brcallout); 2827 2828 msg = &sc->sc_brtimemsg; 2829 KKASSERT(msg->lmsg.ms_flags & MSGF_DONE); 2830 lwkt_sendmsg(BRIDGE_CFGPORT, &msg->lmsg); 2831 2832 crit_exit(); 2833 } 2834 2835 static void 2836 bridge_timer_handler(netmsg_t msg) 2837 { 2838 struct bridge_softc *sc = msg->lmsg.u.ms_resultp; 2839 2840 KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT); 2841 2842 crit_enter(); 2843 /* Reply ASAP */ 2844 lwkt_replymsg(&msg->lmsg, 0); 2845 crit_exit(); 2846 2847 bridge_rtage(sc); 2848 if (sc->sc_ifp->if_flags & IFF_RUNNING) { 2849 callout_reset(&sc->sc_brcallout, 2850 bridge_rtable_prune_period * hz, bridge_timer, sc); 2851 } 2852 } 2853 2854 static int 2855 bridge_rtage_finddead(struct bridge_softc *sc) 2856 { 2857 struct bridge_rtnode *brt; 2858 int dead = 0; 2859 2860 LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) { 2861 struct bridge_rtinfo *bri = brt->brt_info; 2862 2863 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 2864 time_second >= bri->bri_expire) { 2865 bri->bri_dead = 1; 2866 ++dead; 2867 KKASSERT(dead <= sc->sc_brtcnt); 2868 } 2869 } 2870 return dead; 2871 } 2872 2873 /* 2874 * bridge_rtage: 2875 * 2876 * Perform an aging cycle. 2877 */ 2878 static void 2879 bridge_rtage(struct bridge_softc *sc) 2880 { 2881 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 2882 2883 if (bridge_rtage_finddead(sc)) 2884 bridge_rtreap(sc); 2885 } 2886 2887 /* 2888 * bridge_rtflush: 2889 * 2890 * Remove all dynamic addresses from the bridge. 2891 */ 2892 static void 2893 bridge_rtflush(struct bridge_softc *sc, int bf) 2894 { 2895 struct bridge_rtnode *brt; 2896 int reap; 2897 2898 reap = 0; 2899 LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) { 2900 struct bridge_rtinfo *bri = brt->brt_info; 2901 2902 if ((bf & IFBF_FLUSHALL) || 2903 (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 2904 bri->bri_dead = 1; 2905 reap = 1; 2906 } 2907 } 2908 if (reap) { 2909 if (bf & IFBF_FLUSHSYNC) 2910 bridge_rtreap(sc); 2911 else 2912 bridge_rtreap_async(sc); 2913 } 2914 } 2915 2916 /* 2917 * bridge_rtdaddr: 2918 * 2919 * Remove an address from the table. 2920 */ 2921 static int 2922 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr) 2923 { 2924 struct bridge_rtnode *brt; 2925 2926 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 2927 2928 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) 2929 return (ENOENT); 2930 2931 /* TODO: add a cheaper delete operation */ 2932 brt->brt_info->bri_dead = 1; 2933 bridge_rtreap(sc); 2934 return (0); 2935 } 2936 2937 /* 2938 * bridge_rtdelete: 2939 * 2940 * Delete routes to a speicifc member interface. 2941 */ 2942 void 2943 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int bf) 2944 { 2945 struct bridge_rtnode *brt; 2946 int reap; 2947 2948 reap = 0; 2949 LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) { 2950 struct bridge_rtinfo *bri = brt->brt_info; 2951 2952 if (bri->bri_ifp == ifp && 2953 ((bf & IFBF_FLUSHALL) || 2954 (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) { 2955 bri->bri_dead = 1; 2956 reap = 1; 2957 } 2958 } 2959 if (reap) { 2960 if (bf & IFBF_FLUSHSYNC) 2961 bridge_rtreap(sc); 2962 else 2963 bridge_rtreap_async(sc); 2964 } 2965 } 2966 2967 /* 2968 * bridge_rtable_init: 2969 * 2970 * Initialize the route table for this bridge. 2971 */ 2972 static void 2973 bridge_rtable_init(struct bridge_softc *sc) 2974 { 2975 int cpu; 2976 2977 /* 2978 * Initialize per-cpu hash tables 2979 */ 2980 sc->sc_rthashs = kmalloc(sizeof(*sc->sc_rthashs) * ncpus, 2981 M_DEVBUF, M_WAITOK); 2982 for (cpu = 0; cpu < ncpus; ++cpu) { 2983 int i; 2984 2985 sc->sc_rthashs[cpu] = 2986 kmalloc(sizeof(struct bridge_rtnode_head) * BRIDGE_RTHASH_SIZE, 2987 M_DEVBUF, M_WAITOK); 2988 2989 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++) 2990 LIST_INIT(&sc->sc_rthashs[cpu][i]); 2991 } 2992 sc->sc_rthash_key = karc4random(); 2993 2994 /* 2995 * Initialize per-cpu lists 2996 */ 2997 sc->sc_rtlists = kmalloc(sizeof(struct bridge_rtnode_head) * ncpus, 2998 M_DEVBUF, M_WAITOK); 2999 for (cpu = 0; cpu < ncpus; ++cpu) 3000 LIST_INIT(&sc->sc_rtlists[cpu]); 3001 } 3002 3003 /* 3004 * bridge_rtable_fini: 3005 * 3006 * Deconstruct the route table for this bridge. 3007 */ 3008 static void 3009 bridge_rtable_fini(struct bridge_softc *sc) 3010 { 3011 int cpu; 3012 3013 /* 3014 * Free per-cpu hash tables 3015 */ 3016 for (cpu = 0; cpu < ncpus; ++cpu) 3017 kfree(sc->sc_rthashs[cpu], M_DEVBUF); 3018 kfree(sc->sc_rthashs, M_DEVBUF); 3019 3020 /* 3021 * Free per-cpu lists 3022 */ 3023 kfree(sc->sc_rtlists, M_DEVBUF); 3024 } 3025 3026 /* 3027 * The following hash function is adapted from "Hash Functions" by Bob Jenkins 3028 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997). 3029 */ 3030 #define mix(a, b, c) \ 3031 do { \ 3032 a -= b; a -= c; a ^= (c >> 13); \ 3033 b -= c; b -= a; b ^= (a << 8); \ 3034 c -= a; c -= b; c ^= (b >> 13); \ 3035 a -= b; a -= c; a ^= (c >> 12); \ 3036 b -= c; b -= a; b ^= (a << 16); \ 3037 c -= a; c -= b; c ^= (b >> 5); \ 3038 a -= b; a -= c; a ^= (c >> 3); \ 3039 b -= c; b -= a; b ^= (a << 10); \ 3040 c -= a; c -= b; c ^= (b >> 15); \ 3041 } while (/*CONSTCOND*/0) 3042 3043 static __inline uint32_t 3044 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr) 3045 { 3046 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key; 3047 3048 b += addr[5] << 8; 3049 b += addr[4]; 3050 a += addr[3] << 24; 3051 a += addr[2] << 16; 3052 a += addr[1] << 8; 3053 a += addr[0]; 3054 3055 mix(a, b, c); 3056 3057 return (c & BRIDGE_RTHASH_MASK); 3058 } 3059 3060 #undef mix 3061 3062 static int 3063 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b) 3064 { 3065 int i, d; 3066 3067 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) { 3068 d = ((int)a[i]) - ((int)b[i]); 3069 } 3070 3071 return (d); 3072 } 3073 3074 /* 3075 * bridge_rtnode_lookup: 3076 * 3077 * Look up a bridge route node for the specified destination. 3078 */ 3079 static struct bridge_rtnode * 3080 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr) 3081 { 3082 struct bridge_rtnode *brt; 3083 uint32_t hash; 3084 int dir; 3085 3086 hash = bridge_rthash(sc, addr); 3087 LIST_FOREACH(brt, &sc->sc_rthashs[mycpuid][hash], brt_hash) { 3088 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr); 3089 if (dir == 0) 3090 return (brt); 3091 if (dir > 0) 3092 return (NULL); 3093 } 3094 3095 return (NULL); 3096 } 3097 3098 /* 3099 * bridge_rtnode_insert: 3100 * 3101 * Insert the specified bridge node into the route table. 3102 * Caller has to make sure that rtnode does not exist. 3103 */ 3104 static void 3105 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt) 3106 { 3107 struct bridge_rtnode *lbrt; 3108 uint32_t hash; 3109 int dir; 3110 3111 hash = bridge_rthash(sc, brt->brt_addr); 3112 3113 lbrt = LIST_FIRST(&sc->sc_rthashs[mycpuid][hash]); 3114 if (lbrt == NULL) { 3115 LIST_INSERT_HEAD(&sc->sc_rthashs[mycpuid][hash], brt, brt_hash); 3116 goto out; 3117 } 3118 3119 do { 3120 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr); 3121 KASSERT(dir != 0, ("rtnode already exist\n")); 3122 3123 if (dir > 0) { 3124 LIST_INSERT_BEFORE(lbrt, brt, brt_hash); 3125 goto out; 3126 } 3127 if (LIST_NEXT(lbrt, brt_hash) == NULL) { 3128 LIST_INSERT_AFTER(lbrt, brt, brt_hash); 3129 goto out; 3130 } 3131 lbrt = LIST_NEXT(lbrt, brt_hash); 3132 } while (lbrt != NULL); 3133 3134 panic("no suitable position found for rtnode\n"); 3135 out: 3136 LIST_INSERT_HEAD(&sc->sc_rtlists[mycpuid], brt, brt_list); 3137 if (mycpuid == 0) { 3138 /* 3139 * Update the brtcnt. 3140 * We only need to do it once and we do it on CPU0. 3141 */ 3142 sc->sc_brtcnt++; 3143 } 3144 } 3145 3146 /* 3147 * bridge_rtnode_destroy: 3148 * 3149 * Destroy a bridge rtnode. 3150 */ 3151 static void 3152 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt) 3153 { 3154 LIST_REMOVE(brt, brt_hash); 3155 LIST_REMOVE(brt, brt_list); 3156 3157 if (mycpuid + 1 == ncpus) { 3158 /* Free rtinfo associated with rtnode on the last cpu */ 3159 kfree(brt->brt_info, M_DEVBUF); 3160 } 3161 kfree(brt, M_DEVBUF); 3162 3163 if (mycpuid == 0) { 3164 /* Update brtcnt only on CPU0 */ 3165 sc->sc_brtcnt--; 3166 } 3167 } 3168 3169 static __inline int 3170 bridge_post_pfil(struct mbuf *m) 3171 { 3172 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) 3173 return EOPNOTSUPP; 3174 3175 /* Not yet */ 3176 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) 3177 return EOPNOTSUPP; 3178 3179 return 0; 3180 } 3181 3182 /* 3183 * Send bridge packets through pfil if they are one of the types pfil can deal 3184 * with, or if they are ARP or REVARP. (pfil will pass ARP and REVARP without 3185 * question.) If *bifp or *ifp are NULL then packet filtering is skipped for 3186 * that interface. 3187 */ 3188 static int 3189 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir) 3190 { 3191 int snap, error, i, hlen; 3192 struct ether_header *eh1, eh2; 3193 struct ip *ip; 3194 struct llc llc1; 3195 u_int16_t ether_type; 3196 3197 snap = 0; 3198 error = -1; /* Default error if not error == 0 */ 3199 3200 if (pfil_bridge == 0 && pfil_member == 0) 3201 return (0); /* filtering is disabled */ 3202 3203 i = min((*mp)->m_pkthdr.len, max_protohdr); 3204 if ((*mp)->m_len < i) { 3205 *mp = m_pullup(*mp, i); 3206 if (*mp == NULL) { 3207 kprintf("%s: m_pullup failed\n", __func__); 3208 return (-1); 3209 } 3210 } 3211 3212 eh1 = mtod(*mp, struct ether_header *); 3213 ether_type = ntohs(eh1->ether_type); 3214 3215 /* 3216 * Check for SNAP/LLC. 3217 */ 3218 if (ether_type < ETHERMTU) { 3219 struct llc *llc2 = (struct llc *)(eh1 + 1); 3220 3221 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 && 3222 llc2->llc_dsap == LLC_SNAP_LSAP && 3223 llc2->llc_ssap == LLC_SNAP_LSAP && 3224 llc2->llc_control == LLC_UI) { 3225 ether_type = htons(llc2->llc_un.type_snap.ether_type); 3226 snap = 1; 3227 } 3228 } 3229 3230 /* 3231 * If we're trying to filter bridge traffic, don't look at anything 3232 * other than IP and ARP traffic. If the filter doesn't understand 3233 * IPv6, don't allow IPv6 through the bridge either. This is lame 3234 * since if we really wanted, say, an AppleTalk filter, we are hosed, 3235 * but of course we don't have an AppleTalk filter to begin with. 3236 * (Note that since pfil doesn't understand ARP it will pass *ALL* 3237 * ARP traffic.) 3238 */ 3239 switch (ether_type) { 3240 case ETHERTYPE_ARP: 3241 case ETHERTYPE_REVARP: 3242 return (0); /* Automatically pass */ 3243 3244 case ETHERTYPE_IP: 3245 #ifdef INET6 3246 case ETHERTYPE_IPV6: 3247 #endif /* INET6 */ 3248 break; 3249 3250 default: 3251 /* 3252 * Check to see if the user wants to pass non-ip 3253 * packets, these will not be checked by pfil(9) 3254 * and passed unconditionally so the default is to drop. 3255 */ 3256 if (pfil_onlyip) 3257 goto bad; 3258 } 3259 3260 /* Strip off the Ethernet header and keep a copy. */ 3261 m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2); 3262 m_adj(*mp, ETHER_HDR_LEN); 3263 3264 /* Strip off snap header, if present */ 3265 if (snap) { 3266 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1); 3267 m_adj(*mp, sizeof(struct llc)); 3268 } 3269 3270 /* 3271 * Check the IP header for alignment and errors 3272 */ 3273 if (dir == PFIL_IN) { 3274 switch (ether_type) { 3275 case ETHERTYPE_IP: 3276 error = bridge_ip_checkbasic(mp); 3277 break; 3278 #ifdef INET6 3279 case ETHERTYPE_IPV6: 3280 error = bridge_ip6_checkbasic(mp); 3281 break; 3282 #endif /* INET6 */ 3283 default: 3284 error = 0; 3285 } 3286 if (error) 3287 goto bad; 3288 } 3289 3290 error = 0; 3291 3292 /* 3293 * Run the packet through pfil 3294 */ 3295 switch (ether_type) { 3296 case ETHERTYPE_IP: 3297 /* 3298 * before calling the firewall, swap fields the same as 3299 * IP does. here we assume the header is contiguous 3300 */ 3301 ip = mtod(*mp, struct ip *); 3302 3303 ip->ip_len = ntohs(ip->ip_len); 3304 ip->ip_off = ntohs(ip->ip_off); 3305 3306 /* 3307 * Run pfil on the member interface and the bridge, both can 3308 * be skipped by clearing pfil_member or pfil_bridge. 3309 * 3310 * Keep the order: 3311 * in_if -> bridge_if -> out_if 3312 */ 3313 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL) { 3314 error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir); 3315 if (*mp == NULL || error != 0) /* filter may consume */ 3316 break; 3317 error = bridge_post_pfil(*mp); 3318 if (error) 3319 break; 3320 } 3321 3322 if (pfil_member && ifp != NULL) { 3323 error = pfil_run_hooks(&inet_pfil_hook, mp, ifp, dir); 3324 if (*mp == NULL || error != 0) /* filter may consume */ 3325 break; 3326 error = bridge_post_pfil(*mp); 3327 if (error) 3328 break; 3329 } 3330 3331 if (pfil_bridge && dir == PFIL_IN && bifp != NULL) { 3332 error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir); 3333 if (*mp == NULL || error != 0) /* filter may consume */ 3334 break; 3335 error = bridge_post_pfil(*mp); 3336 if (error) 3337 break; 3338 } 3339 3340 /* check if we need to fragment the packet */ 3341 if (pfil_member && ifp != NULL && dir == PFIL_OUT) { 3342 i = (*mp)->m_pkthdr.len; 3343 if (i > ifp->if_mtu) { 3344 error = bridge_fragment(ifp, *mp, &eh2, snap, 3345 &llc1); 3346 return (error); 3347 } 3348 } 3349 3350 /* Recalculate the ip checksum and restore byte ordering */ 3351 ip = mtod(*mp, struct ip *); 3352 hlen = ip->ip_hl << 2; 3353 if (hlen < sizeof(struct ip)) 3354 goto bad; 3355 if (hlen > (*mp)->m_len) { 3356 if ((*mp = m_pullup(*mp, hlen)) == 0) 3357 goto bad; 3358 ip = mtod(*mp, struct ip *); 3359 if (ip == NULL) 3360 goto bad; 3361 } 3362 ip->ip_len = htons(ip->ip_len); 3363 ip->ip_off = htons(ip->ip_off); 3364 ip->ip_sum = 0; 3365 if (hlen == sizeof(struct ip)) 3366 ip->ip_sum = in_cksum_hdr(ip); 3367 else 3368 ip->ip_sum = in_cksum(*mp, hlen); 3369 3370 break; 3371 #ifdef INET6 3372 case ETHERTYPE_IPV6: 3373 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL) 3374 error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp, 3375 dir); 3376 3377 if (*mp == NULL || error != 0) /* filter may consume */ 3378 break; 3379 3380 if (pfil_member && ifp != NULL) 3381 error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp, 3382 dir); 3383 3384 if (*mp == NULL || error != 0) /* filter may consume */ 3385 break; 3386 3387 if (pfil_bridge && dir == PFIL_IN && bifp != NULL) 3388 error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp, 3389 dir); 3390 break; 3391 #endif 3392 default: 3393 error = 0; 3394 break; 3395 } 3396 3397 if (*mp == NULL) 3398 return (error); 3399 if (error != 0) 3400 goto bad; 3401 3402 error = -1; 3403 3404 /* 3405 * Finally, put everything back the way it was and return 3406 */ 3407 if (snap) { 3408 M_PREPEND(*mp, sizeof(struct llc), MB_DONTWAIT); 3409 if (*mp == NULL) 3410 return (error); 3411 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc)); 3412 } 3413 3414 M_PREPEND(*mp, ETHER_HDR_LEN, MB_DONTWAIT); 3415 if (*mp == NULL) 3416 return (error); 3417 bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN); 3418 3419 return (0); 3420 3421 bad: 3422 m_freem(*mp); 3423 *mp = NULL; 3424 return (error); 3425 } 3426 3427 /* 3428 * Perform basic checks on header size since 3429 * pfil assumes ip_input has already processed 3430 * it for it. Cut-and-pasted from ip_input.c. 3431 * Given how simple the IPv6 version is, 3432 * does the IPv4 version really need to be 3433 * this complicated? 3434 * 3435 * XXX Should we update ipstat here, or not? 3436 * XXX Right now we update ipstat but not 3437 * XXX csum_counter. 3438 */ 3439 static int 3440 bridge_ip_checkbasic(struct mbuf **mp) 3441 { 3442 struct mbuf *m = *mp; 3443 struct ip *ip; 3444 int len, hlen; 3445 u_short sum; 3446 3447 if (*mp == NULL) 3448 return (-1); 3449 #if notyet 3450 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 3451 if ((m = m_copyup(m, sizeof(struct ip), 3452 (max_linkhdr + 3) & ~3)) == NULL) { 3453 /* XXXJRT new stat, please */ 3454 ipstat.ips_toosmall++; 3455 goto bad; 3456 } 3457 } else 3458 #endif 3459 #ifndef __predict_false 3460 #define __predict_false(x) x 3461 #endif 3462 if (__predict_false(m->m_len < sizeof (struct ip))) { 3463 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 3464 ipstat.ips_toosmall++; 3465 goto bad; 3466 } 3467 } 3468 ip = mtod(m, struct ip *); 3469 if (ip == NULL) goto bad; 3470 3471 if (ip->ip_v != IPVERSION) { 3472 ipstat.ips_badvers++; 3473 goto bad; 3474 } 3475 hlen = ip->ip_hl << 2; 3476 if (hlen < sizeof(struct ip)) { /* minimum header length */ 3477 ipstat.ips_badhlen++; 3478 goto bad; 3479 } 3480 if (hlen > m->m_len) { 3481 if ((m = m_pullup(m, hlen)) == 0) { 3482 ipstat.ips_badhlen++; 3483 goto bad; 3484 } 3485 ip = mtod(m, struct ip *); 3486 if (ip == NULL) goto bad; 3487 } 3488 3489 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 3490 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 3491 } else { 3492 if (hlen == sizeof(struct ip)) { 3493 sum = in_cksum_hdr(ip); 3494 } else { 3495 sum = in_cksum(m, hlen); 3496 } 3497 } 3498 if (sum) { 3499 ipstat.ips_badsum++; 3500 goto bad; 3501 } 3502 3503 /* Retrieve the packet length. */ 3504 len = ntohs(ip->ip_len); 3505 3506 /* 3507 * Check for additional length bogosity 3508 */ 3509 if (len < hlen) { 3510 ipstat.ips_badlen++; 3511 goto bad; 3512 } 3513 3514 /* 3515 * Check that the amount of data in the buffers 3516 * is as at least much as the IP header would have us expect. 3517 * Drop packet if shorter than we expect. 3518 */ 3519 if (m->m_pkthdr.len < len) { 3520 ipstat.ips_tooshort++; 3521 goto bad; 3522 } 3523 3524 /* Checks out, proceed */ 3525 *mp = m; 3526 return (0); 3527 3528 bad: 3529 *mp = m; 3530 return (-1); 3531 } 3532 3533 #ifdef INET6 3534 /* 3535 * Same as above, but for IPv6. 3536 * Cut-and-pasted from ip6_input.c. 3537 * XXX Should we update ip6stat, or not? 3538 */ 3539 static int 3540 bridge_ip6_checkbasic(struct mbuf **mp) 3541 { 3542 struct mbuf *m = *mp; 3543 struct ip6_hdr *ip6; 3544 3545 /* 3546 * If the IPv6 header is not aligned, slurp it up into a new 3547 * mbuf with space for link headers, in the event we forward 3548 * it. Otherwise, if it is aligned, make sure the entire base 3549 * IPv6 header is in the first mbuf of the chain. 3550 */ 3551 #if notyet 3552 if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 3553 struct ifnet *inifp = m->m_pkthdr.rcvif; 3554 if ((m = m_copyup(m, sizeof(struct ip6_hdr), 3555 (max_linkhdr + 3) & ~3)) == NULL) { 3556 /* XXXJRT new stat, please */ 3557 ip6stat.ip6s_toosmall++; 3558 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 3559 goto bad; 3560 } 3561 } else 3562 #endif 3563 if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) { 3564 struct ifnet *inifp = m->m_pkthdr.rcvif; 3565 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 3566 ip6stat.ip6s_toosmall++; 3567 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 3568 goto bad; 3569 } 3570 } 3571 3572 ip6 = mtod(m, struct ip6_hdr *); 3573 3574 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 3575 ip6stat.ip6s_badvers++; 3576 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 3577 goto bad; 3578 } 3579 3580 /* Checks out, proceed */ 3581 *mp = m; 3582 return (0); 3583 3584 bad: 3585 *mp = m; 3586 return (-1); 3587 } 3588 #endif /* INET6 */ 3589 3590 /* 3591 * bridge_fragment: 3592 * 3593 * Return a fragmented mbuf chain. 3594 */ 3595 static int 3596 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh, 3597 int snap, struct llc *llc) 3598 { 3599 struct mbuf *m0; 3600 struct ip *ip; 3601 int error = -1; 3602 3603 if (m->m_len < sizeof(struct ip) && 3604 (m = m_pullup(m, sizeof(struct ip))) == NULL) 3605 goto out; 3606 ip = mtod(m, struct ip *); 3607 3608 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, 3609 CSUM_DELAY_IP); 3610 if (error) 3611 goto out; 3612 3613 /* walk the chain and re-add the Ethernet header */ 3614 for (m0 = m; m0; m0 = m0->m_nextpkt) { 3615 if (error == 0) { 3616 if (snap) { 3617 M_PREPEND(m0, sizeof(struct llc), MB_DONTWAIT); 3618 if (m0 == NULL) { 3619 error = ENOBUFS; 3620 continue; 3621 } 3622 bcopy(llc, mtod(m0, caddr_t), 3623 sizeof(struct llc)); 3624 } 3625 M_PREPEND(m0, ETHER_HDR_LEN, MB_DONTWAIT); 3626 if (m0 == NULL) { 3627 error = ENOBUFS; 3628 continue; 3629 } 3630 bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN); 3631 } else 3632 m_freem(m); 3633 } 3634 3635 if (error == 0) 3636 ipstat.ips_fragmented++; 3637 3638 return (error); 3639 3640 out: 3641 if (m != NULL) 3642 m_freem(m); 3643 return (error); 3644 } 3645 3646 static void 3647 bridge_enqueue_handler(netmsg_t msg) 3648 { 3649 struct netmsg_packet *nmp; 3650 struct ifnet *dst_ifp; 3651 struct mbuf *m; 3652 3653 nmp = &msg->packet; 3654 m = nmp->nm_packet; 3655 dst_ifp = nmp->base.lmsg.u.ms_resultp; 3656 3657 bridge_handoff(dst_ifp, m); 3658 } 3659 3660 static void 3661 bridge_handoff(struct ifnet *dst_ifp, struct mbuf *m) 3662 { 3663 struct mbuf *m0; 3664 3665 /* We may be sending a fragment so traverse the mbuf */ 3666 for (; m; m = m0) { 3667 struct altq_pktattr pktattr; 3668 3669 m0 = m->m_nextpkt; 3670 m->m_nextpkt = NULL; 3671 3672 if (ifq_is_enabled(&dst_ifp->if_snd)) 3673 altq_etherclassify(&dst_ifp->if_snd, m, &pktattr); 3674 3675 ifq_dispatch(dst_ifp, m, &pktattr); 3676 } 3677 } 3678 3679 static void 3680 bridge_control_dispatch(netmsg_t msg) 3681 { 3682 struct netmsg_brctl *bc_msg = (struct netmsg_brctl *)msg; 3683 struct ifnet *bifp = bc_msg->bc_sc->sc_ifp; 3684 int error; 3685 3686 ifnet_serialize_all(bifp); 3687 error = bc_msg->bc_func(bc_msg->bc_sc, bc_msg->bc_arg); 3688 ifnet_deserialize_all(bifp); 3689 3690 lwkt_replymsg(&bc_msg->base.lmsg, error); 3691 } 3692 3693 static int 3694 bridge_control(struct bridge_softc *sc, u_long cmd, 3695 bridge_ctl_t bc_func, void *bc_arg) 3696 { 3697 struct ifnet *bifp = sc->sc_ifp; 3698 struct netmsg_brctl bc_msg; 3699 int error; 3700 3701 ASSERT_IFNET_SERIALIZED_ALL(bifp); 3702 3703 bzero(&bc_msg, sizeof(bc_msg)); 3704 3705 netmsg_init(&bc_msg.base, NULL, &curthread->td_msgport, 3706 0, bridge_control_dispatch); 3707 bc_msg.bc_func = bc_func; 3708 bc_msg.bc_sc = sc; 3709 bc_msg.bc_arg = bc_arg; 3710 3711 ifnet_deserialize_all(bifp); 3712 error = lwkt_domsg(BRIDGE_CFGPORT, &bc_msg.base.lmsg, 0); 3713 ifnet_serialize_all(bifp); 3714 return error; 3715 } 3716 3717 static void 3718 bridge_add_bif_handler(netmsg_t msg) 3719 { 3720 struct netmsg_braddbif *amsg = (struct netmsg_braddbif *)msg; 3721 struct bridge_softc *sc; 3722 struct bridge_iflist *bif; 3723 3724 sc = amsg->br_softc; 3725 3726 bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO); 3727 bif->bif_ifp = amsg->br_bif_ifp; 3728 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER; 3729 bif->bif_onlist = 1; 3730 bif->bif_info = amsg->br_bif_info; 3731 3732 LIST_INSERT_HEAD(&sc->sc_iflists[mycpuid], bif, bif_next); 3733 3734 ifnet_forwardmsg(&amsg->base.lmsg, mycpuid + 1); 3735 } 3736 3737 static void 3738 bridge_add_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info, 3739 struct ifnet *ifp) 3740 { 3741 struct netmsg_braddbif amsg; 3742 3743 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 3744 3745 netmsg_init(&amsg.base, NULL, &curthread->td_msgport, 3746 0, bridge_add_bif_handler); 3747 amsg.br_softc = sc; 3748 amsg.br_bif_info = bif_info; 3749 amsg.br_bif_ifp = ifp; 3750 3751 ifnet_domsg(&amsg.base.lmsg, 0); 3752 } 3753 3754 static void 3755 bridge_del_bif_handler(netmsg_t msg) 3756 { 3757 struct netmsg_brdelbif *dmsg = (struct netmsg_brdelbif *)msg; 3758 struct bridge_softc *sc; 3759 struct bridge_iflist *bif; 3760 3761 sc = dmsg->br_softc; 3762 3763 /* 3764 * Locate the bif associated with the br_bif_info 3765 * on the current CPU 3766 */ 3767 bif = bridge_lookup_member_ifinfo(sc, dmsg->br_bif_info); 3768 KKASSERT(bif != NULL && bif->bif_onlist); 3769 3770 /* Remove the bif from the current CPU's iflist */ 3771 bif->bif_onlist = 0; 3772 LIST_REMOVE(bif, bif_next); 3773 3774 /* Save the removed bif for later freeing */ 3775 LIST_INSERT_HEAD(dmsg->br_bif_list, bif, bif_next); 3776 3777 ifnet_forwardmsg(&dmsg->base.lmsg, mycpuid + 1); 3778 } 3779 3780 static void 3781 bridge_del_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info, 3782 struct bridge_iflist_head *saved_bifs) 3783 { 3784 struct netmsg_brdelbif dmsg; 3785 3786 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 3787 3788 netmsg_init(&dmsg.base, NULL, &curthread->td_msgport, 3789 0, bridge_del_bif_handler); 3790 dmsg.br_softc = sc; 3791 dmsg.br_bif_info = bif_info; 3792 dmsg.br_bif_list = saved_bifs; 3793 3794 ifnet_domsg(&dmsg.base.lmsg, 0); 3795 } 3796 3797 static void 3798 bridge_set_bifflags_handler(netmsg_t msg) 3799 { 3800 struct netmsg_brsflags *smsg = (struct netmsg_brsflags *)msg; 3801 struct bridge_softc *sc; 3802 struct bridge_iflist *bif; 3803 3804 sc = smsg->br_softc; 3805 3806 /* 3807 * Locate the bif associated with the br_bif_info 3808 * on the current CPU 3809 */ 3810 bif = bridge_lookup_member_ifinfo(sc, smsg->br_bif_info); 3811 KKASSERT(bif != NULL && bif->bif_onlist); 3812 3813 bif->bif_flags = smsg->br_bif_flags; 3814 3815 ifnet_forwardmsg(&smsg->base.lmsg, mycpuid + 1); 3816 } 3817 3818 static void 3819 bridge_set_bifflags(struct bridge_softc *sc, struct bridge_ifinfo *bif_info, 3820 uint32_t bif_flags) 3821 { 3822 struct netmsg_brsflags smsg; 3823 3824 ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp); 3825 3826 netmsg_init(&smsg.base, NULL, &curthread->td_msgport, 3827 0, bridge_set_bifflags_handler); 3828 smsg.br_softc = sc; 3829 smsg.br_bif_info = bif_info; 3830 smsg.br_bif_flags = bif_flags; 3831 3832 ifnet_domsg(&smsg.base.lmsg, 0); 3833 } 3834