1 /* $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $ */ 2 3 /*- 4 * SPDX-License-Identifier: BSD-4-Clause 5 * 6 * Copyright 2001 Wasabi Systems, Inc. 7 * All rights reserved. 8 * 9 * Written by Jason R. Thorpe for Wasabi Systems, Inc. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed for the NetBSD Project by 22 * Wasabi Systems, Inc. 23 * 4. The name of Wasabi Systems, Inc. may not be used to endorse 24 * or promote products derived from this software without specific prior 25 * written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC 31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 37 * POSSIBILITY OF SUCH DAMAGE. 38 */ 39 40 /* 41 * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net) 42 * All rights reserved. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 54 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 55 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 56 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, 57 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 58 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 59 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 61 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 62 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 63 * POSSIBILITY OF SUCH DAMAGE. 64 * 65 * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp 66 */ 67 68 /* 69 * Network interface bridge support. 70 * 71 * TODO: 72 * 73 * - Currently only supports Ethernet-like interfaces (Ethernet, 74 * 802.11, VLANs on Ethernet, etc.) Figure out a nice way 75 * to bridge other types of interfaces (maybe consider 76 * heterogeneous bridges). 77 */ 78 79 #include <sys/cdefs.h> 80 #include "opt_inet.h" 81 #include "opt_inet6.h" 82 83 #include <sys/param.h> 84 #include <sys/eventhandler.h> 85 #include <sys/mbuf.h> 86 #include <sys/malloc.h> 87 #include <sys/protosw.h> 88 #include <sys/systm.h> 89 #include <sys/jail.h> 90 #include <sys/time.h> 91 #include <sys/socket.h> /* for net/if.h */ 92 #include <sys/sockio.h> 93 #include <sys/ctype.h> /* string functions */ 94 #include <sys/kernel.h> 95 #include <sys/random.h> 96 #include <sys/syslog.h> 97 #include <sys/sysctl.h> 98 #include <vm/uma.h> 99 #include <sys/module.h> 100 #include <sys/priv.h> 101 #include <sys/proc.h> 102 #include <sys/lock.h> 103 #include <sys/mutex.h> 104 105 #include <net/bpf.h> 106 #include <net/if.h> 107 #include <net/if_clone.h> 108 #include <net/if_dl.h> 109 #include <net/if_types.h> 110 #include <net/if_var.h> 111 #include <net/if_private.h> 112 #include <net/pfil.h> 113 #include <net/vnet.h> 114 115 #include <netinet/in.h> 116 #include <netinet/in_systm.h> 117 #include <netinet/in_var.h> 118 #include <netinet/ip.h> 119 #include <netinet/ip_var.h> 120 #ifdef INET6 121 #include <netinet/ip6.h> 122 #include <netinet6/ip6_var.h> 123 #include <netinet6/in6_ifattach.h> 124 #endif 125 #if defined(INET) || defined(INET6) 126 #include <netinet/ip_carp.h> 127 #endif 128 #include <machine/in_cksum.h> 129 #include <netinet/if_ether.h> 130 #include <net/bridgestp.h> 131 #include <net/if_bridgevar.h> 132 #include <net/if_llc.h> 133 #include <net/if_vlan_var.h> 134 135 #include <net/route.h> 136 137 /* 138 * At various points in the code we need to know if we're hooked into the INET 139 * and/or INET6 pfil. Define some macros to do that based on which IP versions 140 * are enabled in the kernel. This avoids littering the rest of the code with 141 * #ifnet INET6 to avoid referencing V_inet6_pfil_head. 142 */ 143 #ifdef INET6 144 #define PFIL_HOOKED_IN_INET6 PFIL_HOOKED_IN(V_inet6_pfil_head) 145 #define PFIL_HOOKED_OUT_INET6 PFIL_HOOKED_OUT(V_inet6_pfil_head) 146 #else 147 #define PFIL_HOOKED_IN_INET6 false 148 #define PFIL_HOOKED_OUT_INET6 false 149 #endif 150 151 #ifdef INET 152 #define PFIL_HOOKED_IN_INET PFIL_HOOKED_IN(V_inet_pfil_head) 153 #define PFIL_HOOKED_OUT_INET PFIL_HOOKED_OUT(V_inet_pfil_head) 154 #else 155 #define PFIL_HOOKED_IN_INET false 156 #define PFIL_HOOKED_OUT_INET false 157 #endif 158 159 #define PFIL_HOOKED_IN_46 (PFIL_HOOKED_IN_INET6 || PFIL_HOOKED_IN_INET) 160 #define PFIL_HOOKED_OUT_46 (PFIL_HOOKED_OUT_INET6 || PFIL_HOOKED_OUT_INET) 161 162 /* 163 * Size of the route hash table. Must be a power of two. 164 */ 165 #ifndef BRIDGE_RTHASH_SIZE 166 #define BRIDGE_RTHASH_SIZE 1024 167 #endif 168 169 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1) 170 171 /* 172 * Default maximum number of addresses to cache. 173 */ 174 #ifndef BRIDGE_RTABLE_MAX 175 #define BRIDGE_RTABLE_MAX 2000 176 #endif 177 178 /* 179 * Timeout (in seconds) for entries learned dynamically. 180 */ 181 #ifndef BRIDGE_RTABLE_TIMEOUT 182 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */ 183 #endif 184 185 /* 186 * Number of seconds between walks of the route list. 187 */ 188 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD 189 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60) 190 #endif 191 192 /* 193 * List of capabilities to possibly mask on the member interface. 194 */ 195 #define BRIDGE_IFCAPS_MASK (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\ 196 IFCAP_TXCSUM_IPV6|IFCAP_MEXTPG) 197 198 /* 199 * List of capabilities to strip 200 */ 201 #define BRIDGE_IFCAPS_STRIP IFCAP_LRO 202 203 /* 204 * Bridge locking 205 * 206 * The bridge relies heavily on the epoch(9) system to protect its data 207 * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we 208 * must ensure there is only one writer at a time. 209 * 210 * That is: for read accesses we only need to be in NET_EPOCH, but for write 211 * accesses we must hold: 212 * 213 * - BRIDGE_RT_LOCK, for any change to bridge_rtnodes 214 * - BRIDGE_LOCK, for any other change 215 * 216 * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl() 217 * calls to bridge member interfaces and these ioctl()s can sleep. 218 * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes 219 * required while we're in NET_EPOCH and then we're not allowed to sleep. 220 */ 221 #define BRIDGE_LOCK_INIT(_sc) do { \ 222 sx_init(&(_sc)->sc_sx, "if_bridge"); \ 223 mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF); \ 224 } while (0) 225 #define BRIDGE_LOCK_DESTROY(_sc) do { \ 226 sx_destroy(&(_sc)->sc_sx); \ 227 mtx_destroy(&(_sc)->sc_rt_mtx); \ 228 } while (0) 229 #define BRIDGE_LOCK(_sc) sx_xlock(&(_sc)->sc_sx) 230 #define BRIDGE_UNLOCK(_sc) sx_xunlock(&(_sc)->sc_sx) 231 #define BRIDGE_LOCK_ASSERT(_sc) sx_assert(&(_sc)->sc_sx, SX_XLOCKED) 232 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc) \ 233 MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx)) 234 #define BRIDGE_UNLOCK_ASSERT(_sc) sx_assert(&(_sc)->sc_sx, SX_UNLOCKED) 235 #define BRIDGE_RT_LOCK(_sc) mtx_lock(&(_sc)->sc_rt_mtx) 236 #define BRIDGE_RT_UNLOCK(_sc) mtx_unlock(&(_sc)->sc_rt_mtx) 237 #define BRIDGE_RT_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED) 238 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \ 239 MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx)) 240 241 struct bridge_softc; 242 243 /* 244 * Bridge interface list entry. 245 */ 246 struct bridge_iflist { 247 CK_LIST_ENTRY(bridge_iflist) bif_next; 248 struct ifnet *bif_ifp; /* member if */ 249 struct bridge_softc *bif_sc; /* parent bridge */ 250 struct bstp_port bif_stp; /* STP state */ 251 uint32_t bif_flags; /* member if flags */ 252 int bif_savedcaps; /* saved capabilities */ 253 uint32_t bif_addrmax; /* max # of addresses */ 254 uint32_t bif_addrcnt; /* cur. # of addresses */ 255 uint32_t bif_addrexceeded;/* # of address violations */ 256 struct epoch_context bif_epoch_ctx; 257 ether_vlanid_t bif_untagged; /* untagged vlan id */ 258 ifbvlan_set_t bif_vlan_set; /* allowed tagged vlans */ 259 }; 260 261 /* 262 * Bridge route node. 263 */ 264 struct bridge_rtnode { 265 CK_LIST_ENTRY(bridge_rtnode) brt_hash; /* hash table linkage */ 266 CK_LIST_ENTRY(bridge_rtnode) brt_list; /* list linkage */ 267 struct bridge_iflist *brt_dst; /* destination if */ 268 unsigned long brt_expire; /* expiration time */ 269 uint8_t brt_flags; /* address flags */ 270 uint8_t brt_addr[ETHER_ADDR_LEN]; 271 ether_vlanid_t brt_vlan; /* vlan id */ 272 struct vnet *brt_vnet; 273 struct epoch_context brt_epoch_ctx; 274 }; 275 #define brt_ifp brt_dst->bif_ifp 276 277 /* 278 * Software state for each bridge. 279 */ 280 struct bridge_softc { 281 struct ifnet *sc_ifp; /* make this an interface */ 282 LIST_ENTRY(bridge_softc) sc_list; 283 struct sx sc_sx; 284 struct mtx sc_rt_mtx; 285 uint32_t sc_brtmax; /* max # of addresses */ 286 uint32_t sc_brtcnt; /* cur. # of addresses */ 287 uint32_t sc_brttimeout; /* rt timeout in seconds */ 288 struct callout sc_brcallout; /* bridge callout */ 289 CK_LIST_HEAD(, bridge_iflist) sc_iflist; /* member interface list */ 290 CK_LIST_HEAD(, bridge_rtnode) *sc_rthash; /* our forwarding table */ 291 CK_LIST_HEAD(, bridge_rtnode) sc_rtlist; /* list version of above */ 292 uint32_t sc_rthash_key; /* key for hash */ 293 CK_LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */ 294 struct bstp_state sc_stp; /* STP state */ 295 uint32_t sc_brtexceeded; /* # of cache drops */ 296 struct ifnet *sc_ifaddr; /* member mac copied from */ 297 struct ether_addr sc_defaddr; /* Default MAC address */ 298 if_input_fn_t sc_if_input; /* Saved copy of if_input */ 299 struct epoch_context sc_epoch_ctx; 300 }; 301 302 VNET_DEFINE_STATIC(struct sx, bridge_list_sx); 303 #define V_bridge_list_sx VNET(bridge_list_sx) 304 static eventhandler_tag bridge_detach_cookie; 305 306 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD; 307 308 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone); 309 #define V_bridge_rtnode_zone VNET(bridge_rtnode_zone) 310 311 static int bridge_clone_create(struct if_clone *, char *, size_t, 312 struct ifc_data *, struct ifnet **); 313 static int bridge_clone_destroy(struct if_clone *, struct ifnet *, uint32_t); 314 315 static int bridge_ioctl(struct ifnet *, u_long, caddr_t); 316 static void bridge_mutecaps(struct bridge_softc *); 317 static void bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *, 318 int); 319 static void bridge_ifdetach(void *arg __unused, struct ifnet *); 320 static void bridge_init(void *); 321 static void bridge_dummynet(struct mbuf *, struct ifnet *); 322 static bool bridge_same(const void *, const void *); 323 static void *bridge_get_softc(struct ifnet *); 324 static void bridge_stop(struct ifnet *, int); 325 static int bridge_transmit(struct ifnet *, struct mbuf *); 326 #ifdef ALTQ 327 static void bridge_altq_start(if_t); 328 static int bridge_altq_transmit(if_t, struct mbuf *); 329 #endif 330 static void bridge_qflush(struct ifnet *); 331 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *); 332 static void bridge_inject(struct ifnet *, struct mbuf *); 333 static int bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *, 334 struct rtentry *); 335 static int bridge_enqueue(struct bridge_softc *, struct ifnet *, 336 struct mbuf *, struct bridge_iflist *); 337 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int); 338 339 static void bridge_forward(struct bridge_softc *, struct bridge_iflist *, 340 struct mbuf *m); 341 static bool bridge_member_ifaddrs(void); 342 static void bridge_timer(void *); 343 344 static void bridge_broadcast(struct bridge_softc *, struct ifnet *, 345 struct mbuf *, int); 346 static void bridge_span(struct bridge_softc *, struct mbuf *); 347 348 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *, 349 ether_vlanid_t, struct bridge_iflist *, int, uint8_t); 350 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *, 351 ether_vlanid_t); 352 static void bridge_rttrim(struct bridge_softc *); 353 static void bridge_rtage(struct bridge_softc *); 354 static void bridge_rtflush(struct bridge_softc *, int); 355 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *, 356 ether_vlanid_t); 357 static bool bridge_vfilter_in(const struct bridge_iflist *, struct mbuf *); 358 static bool bridge_vfilter_out(const struct bridge_iflist *, 359 const struct mbuf *); 360 361 static void bridge_rtable_init(struct bridge_softc *); 362 static void bridge_rtable_fini(struct bridge_softc *); 363 364 static int bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *); 365 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *, 366 const uint8_t *, ether_vlanid_t); 367 static int bridge_rtnode_insert(struct bridge_softc *, 368 struct bridge_rtnode *); 369 static void bridge_rtnode_destroy(struct bridge_softc *, 370 struct bridge_rtnode *); 371 static void bridge_rtable_expire(struct ifnet *, int); 372 static void bridge_state_change(struct ifnet *, int); 373 374 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *, 375 const char *name); 376 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *, 377 struct ifnet *ifp); 378 static void bridge_delete_member(struct bridge_softc *, 379 struct bridge_iflist *, int); 380 static void bridge_delete_span(struct bridge_softc *, 381 struct bridge_iflist *); 382 383 static int bridge_ioctl_add(struct bridge_softc *, void *); 384 static int bridge_ioctl_del(struct bridge_softc *, void *); 385 static int bridge_ioctl_gifflags(struct bridge_softc *, void *); 386 static int bridge_ioctl_sifflags(struct bridge_softc *, void *); 387 static int bridge_ioctl_scache(struct bridge_softc *, void *); 388 static int bridge_ioctl_gcache(struct bridge_softc *, void *); 389 static int bridge_ioctl_gifs(struct bridge_softc *, void *); 390 static int bridge_ioctl_rts(struct bridge_softc *, void *); 391 static int bridge_ioctl_saddr(struct bridge_softc *, void *); 392 static int bridge_ioctl_sto(struct bridge_softc *, void *); 393 static int bridge_ioctl_gto(struct bridge_softc *, void *); 394 static int bridge_ioctl_daddr(struct bridge_softc *, void *); 395 static int bridge_ioctl_flush(struct bridge_softc *, void *); 396 static int bridge_ioctl_gpri(struct bridge_softc *, void *); 397 static int bridge_ioctl_spri(struct bridge_softc *, void *); 398 static int bridge_ioctl_ght(struct bridge_softc *, void *); 399 static int bridge_ioctl_sht(struct bridge_softc *, void *); 400 static int bridge_ioctl_gfd(struct bridge_softc *, void *); 401 static int bridge_ioctl_sfd(struct bridge_softc *, void *); 402 static int bridge_ioctl_gma(struct bridge_softc *, void *); 403 static int bridge_ioctl_sma(struct bridge_softc *, void *); 404 static int bridge_ioctl_sifprio(struct bridge_softc *, void *); 405 static int bridge_ioctl_sifcost(struct bridge_softc *, void *); 406 static int bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *); 407 static int bridge_ioctl_sifuntagged(struct bridge_softc *, void *); 408 static int bridge_ioctl_sifvlanset(struct bridge_softc *, void *); 409 static int bridge_ioctl_gifvlanset(struct bridge_softc *, void *); 410 static int bridge_ioctl_addspan(struct bridge_softc *, void *); 411 static int bridge_ioctl_delspan(struct bridge_softc *, void *); 412 static int bridge_ioctl_gbparam(struct bridge_softc *, void *); 413 static int bridge_ioctl_grte(struct bridge_softc *, void *); 414 static int bridge_ioctl_gifsstp(struct bridge_softc *, void *); 415 static int bridge_ioctl_sproto(struct bridge_softc *, void *); 416 static int bridge_ioctl_stxhc(struct bridge_softc *, void *); 417 static int bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *, 418 int); 419 #ifdef INET 420 static int bridge_ip_checkbasic(struct mbuf **mp); 421 static int bridge_fragment(struct ifnet *, struct mbuf **mp, 422 struct ether_header *, int, struct llc *); 423 #endif /* INET */ 424 #ifdef INET6 425 static int bridge_ip6_checkbasic(struct mbuf **mp); 426 #endif /* INET6 */ 427 static void bridge_linkstate(struct ifnet *ifp); 428 static void bridge_linkcheck(struct bridge_softc *sc); 429 430 /* 431 * Use the "null" value from IEEE 802.1Q-2014 Table 9-2 432 * to indicate untagged frames. 433 */ 434 #define VLANTAGOF(_m) \ 435 ((_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : DOT1Q_VID_NULL) 436 437 static struct bstp_cb_ops bridge_ops = { 438 .bcb_state = bridge_state_change, 439 .bcb_rtage = bridge_rtable_expire 440 }; 441 442 SYSCTL_DECL(_net_link); 443 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 444 "Bridge"); 445 446 /* only pass IP[46] packets when pfil is enabled */ 447 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1; 448 #define V_pfil_onlyip VNET(pfil_onlyip) 449 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, 450 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0, 451 "Only pass IP packets when pfil is enabled"); 452 453 /* run pfil hooks on the bridge interface */ 454 VNET_DEFINE_STATIC(int, pfil_bridge) = 0; 455 #define V_pfil_bridge VNET(pfil_bridge) 456 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, 457 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0, 458 "Packet filter on the bridge interface"); 459 460 /* layer2 filter with ipfw */ 461 VNET_DEFINE_STATIC(int, pfil_ipfw); 462 #define V_pfil_ipfw VNET(pfil_ipfw) 463 464 /* layer2 ARP filter with ipfw */ 465 VNET_DEFINE_STATIC(int, pfil_ipfw_arp); 466 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp) 467 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, 468 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0, 469 "Filter ARP packets through IPFW layer2"); 470 471 /* run pfil hooks on the member interface */ 472 VNET_DEFINE_STATIC(int, pfil_member) = 0; 473 #define V_pfil_member VNET(pfil_member) 474 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, 475 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0, 476 "Packet filter on the member interface"); 477 478 /* run pfil hooks on the physical interface for locally destined packets */ 479 VNET_DEFINE_STATIC(int, pfil_local_phys); 480 #define V_pfil_local_phys VNET(pfil_local_phys) 481 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, 482 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0, 483 "Packet filter on the physical interface for locally destined packets"); 484 485 /* log STP state changes */ 486 VNET_DEFINE_STATIC(int, log_stp); 487 #define V_log_stp VNET(log_stp) 488 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, 489 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0, 490 "Log STP state changes"); 491 492 /* share MAC with first bridge member */ 493 VNET_DEFINE_STATIC(int, bridge_inherit_mac); 494 #define V_bridge_inherit_mac VNET(bridge_inherit_mac) 495 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, 496 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0, 497 "Inherit MAC address from the first bridge member"); 498 499 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0; 500 #define V_allow_llz_overlap VNET(allow_llz_overlap) 501 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap, 502 CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0, 503 "Allow overlap of link-local scope " 504 "zones of a bridge interface and the member interfaces"); 505 506 /* log MAC address port flapping */ 507 VNET_DEFINE_STATIC(bool, log_mac_flap) = true; 508 #define V_log_mac_flap VNET(log_mac_flap) 509 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap, 510 CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(log_mac_flap), true, 511 "Log MAC address port flapping"); 512 513 /* allow IP addresses on bridge members */ 514 VNET_DEFINE_STATIC(bool, member_ifaddrs) = false; 515 #define V_member_ifaddrs VNET(member_ifaddrs) 516 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, member_ifaddrs, 517 CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(member_ifaddrs), false, 518 "Allow layer 3 addresses on bridge members"); 519 520 static bool 521 bridge_member_ifaddrs(void) 522 { 523 return (V_member_ifaddrs); 524 } 525 526 VNET_DEFINE_STATIC(int, log_interval) = 5; 527 VNET_DEFINE_STATIC(int, log_count) = 0; 528 VNET_DEFINE_STATIC(struct timeval, log_last) = { 0 }; 529 530 #define V_log_interval VNET(log_interval) 531 #define V_log_count VNET(log_count) 532 #define V_log_last VNET(log_last) 533 534 struct bridge_control { 535 int (*bc_func)(struct bridge_softc *, void *); 536 int bc_argsize; 537 int bc_flags; 538 }; 539 540 #define BC_F_COPYIN 0x01 /* copy arguments in */ 541 #define BC_F_COPYOUT 0x02 /* copy arguments out */ 542 #define BC_F_SUSER 0x04 /* do super-user check */ 543 544 static const struct bridge_control bridge_control_table[] = { 545 { bridge_ioctl_add, sizeof(struct ifbreq), 546 BC_F_COPYIN|BC_F_SUSER }, 547 { bridge_ioctl_del, sizeof(struct ifbreq), 548 BC_F_COPYIN|BC_F_SUSER }, 549 550 { bridge_ioctl_gifflags, sizeof(struct ifbreq), 551 BC_F_COPYIN|BC_F_COPYOUT }, 552 { bridge_ioctl_sifflags, sizeof(struct ifbreq), 553 BC_F_COPYIN|BC_F_SUSER }, 554 555 { bridge_ioctl_scache, sizeof(struct ifbrparam), 556 BC_F_COPYIN|BC_F_SUSER }, 557 { bridge_ioctl_gcache, sizeof(struct ifbrparam), 558 BC_F_COPYOUT }, 559 560 { bridge_ioctl_gifs, sizeof(struct ifbifconf), 561 BC_F_COPYIN|BC_F_COPYOUT }, 562 { bridge_ioctl_rts, sizeof(struct ifbaconf), 563 BC_F_COPYIN|BC_F_COPYOUT }, 564 565 { bridge_ioctl_saddr, sizeof(struct ifbareq), 566 BC_F_COPYIN|BC_F_SUSER }, 567 568 { bridge_ioctl_sto, sizeof(struct ifbrparam), 569 BC_F_COPYIN|BC_F_SUSER }, 570 { bridge_ioctl_gto, sizeof(struct ifbrparam), 571 BC_F_COPYOUT }, 572 573 { bridge_ioctl_daddr, sizeof(struct ifbareq), 574 BC_F_COPYIN|BC_F_SUSER }, 575 576 { bridge_ioctl_flush, sizeof(struct ifbreq), 577 BC_F_COPYIN|BC_F_SUSER }, 578 579 { bridge_ioctl_gpri, sizeof(struct ifbrparam), 580 BC_F_COPYOUT }, 581 { bridge_ioctl_spri, sizeof(struct ifbrparam), 582 BC_F_COPYIN|BC_F_SUSER }, 583 584 { bridge_ioctl_ght, sizeof(struct ifbrparam), 585 BC_F_COPYOUT }, 586 { bridge_ioctl_sht, sizeof(struct ifbrparam), 587 BC_F_COPYIN|BC_F_SUSER }, 588 589 { bridge_ioctl_gfd, sizeof(struct ifbrparam), 590 BC_F_COPYOUT }, 591 { bridge_ioctl_sfd, sizeof(struct ifbrparam), 592 BC_F_COPYIN|BC_F_SUSER }, 593 594 { bridge_ioctl_gma, sizeof(struct ifbrparam), 595 BC_F_COPYOUT }, 596 { bridge_ioctl_sma, sizeof(struct ifbrparam), 597 BC_F_COPYIN|BC_F_SUSER }, 598 599 { bridge_ioctl_sifprio, sizeof(struct ifbreq), 600 BC_F_COPYIN|BC_F_SUSER }, 601 602 { bridge_ioctl_sifcost, sizeof(struct ifbreq), 603 BC_F_COPYIN|BC_F_SUSER }, 604 605 { bridge_ioctl_addspan, sizeof(struct ifbreq), 606 BC_F_COPYIN|BC_F_SUSER }, 607 { bridge_ioctl_delspan, sizeof(struct ifbreq), 608 BC_F_COPYIN|BC_F_SUSER }, 609 610 { bridge_ioctl_gbparam, sizeof(struct ifbropreq), 611 BC_F_COPYOUT }, 612 613 { bridge_ioctl_grte, sizeof(struct ifbrparam), 614 BC_F_COPYOUT }, 615 616 { bridge_ioctl_gifsstp, sizeof(struct ifbpstpconf), 617 BC_F_COPYIN|BC_F_COPYOUT }, 618 619 { bridge_ioctl_sproto, sizeof(struct ifbrparam), 620 BC_F_COPYIN|BC_F_SUSER }, 621 622 { bridge_ioctl_stxhc, sizeof(struct ifbrparam), 623 BC_F_COPYIN|BC_F_SUSER }, 624 625 { bridge_ioctl_sifmaxaddr, sizeof(struct ifbreq), 626 BC_F_COPYIN|BC_F_SUSER }, 627 628 { bridge_ioctl_sifuntagged, sizeof(struct ifbreq), 629 BC_F_COPYIN|BC_F_SUSER }, 630 631 { bridge_ioctl_sifvlanset, sizeof(struct ifbif_vlan_req), 632 BC_F_COPYIN|BC_F_SUSER }, 633 634 { bridge_ioctl_gifvlanset, sizeof(struct ifbif_vlan_req), 635 BC_F_COPYIN|BC_F_COPYOUT }, 636 }; 637 static const int bridge_control_table_size = nitems(bridge_control_table); 638 639 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list) = 640 LIST_HEAD_INITIALIZER(); 641 #define V_bridge_list VNET(bridge_list) 642 #define BRIDGE_LIST_LOCK_INIT(x) sx_init(&V_bridge_list_sx, \ 643 "if_bridge list") 644 #define BRIDGE_LIST_LOCK_DESTROY(x) sx_destroy(&V_bridge_list_sx) 645 #define BRIDGE_LIST_LOCK(x) sx_xlock(&V_bridge_list_sx) 646 #define BRIDGE_LIST_UNLOCK(x) sx_xunlock(&V_bridge_list_sx) 647 648 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner); 649 #define V_bridge_cloner VNET(bridge_cloner) 650 651 static const char bridge_name[] = "bridge"; 652 653 static void 654 vnet_bridge_init(const void *unused __unused) 655 { 656 657 V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode", 658 sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL, 659 UMA_ALIGN_PTR, 0); 660 BRIDGE_LIST_LOCK_INIT(); 661 662 struct if_clone_addreq req = { 663 .create_f = bridge_clone_create, 664 .destroy_f = bridge_clone_destroy, 665 .flags = IFC_F_AUTOUNIT, 666 }; 667 V_bridge_cloner = ifc_attach_cloner(bridge_name, &req); 668 } 669 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY, 670 vnet_bridge_init, NULL); 671 672 static void 673 vnet_bridge_uninit(const void *unused __unused) 674 { 675 676 ifc_detach_cloner(V_bridge_cloner); 677 V_bridge_cloner = NULL; 678 BRIDGE_LIST_LOCK_DESTROY(); 679 680 /* Callbacks may use the UMA zone. */ 681 NET_EPOCH_DRAIN_CALLBACKS(); 682 683 uma_zdestroy(V_bridge_rtnode_zone); 684 } 685 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, 686 vnet_bridge_uninit, NULL); 687 688 static int 689 bridge_modevent(module_t mod, int type, void *data) 690 { 691 692 switch (type) { 693 case MOD_LOAD: 694 bridge_dn_p = bridge_dummynet; 695 bridge_same_p = bridge_same; 696 bridge_get_softc_p = bridge_get_softc; 697 bridge_member_ifaddrs_p = bridge_member_ifaddrs; 698 bridge_detach_cookie = EVENTHANDLER_REGISTER( 699 ifnet_departure_event, bridge_ifdetach, NULL, 700 EVENTHANDLER_PRI_ANY); 701 break; 702 case MOD_UNLOAD: 703 EVENTHANDLER_DEREGISTER(ifnet_departure_event, 704 bridge_detach_cookie); 705 bridge_dn_p = NULL; 706 bridge_same_p = NULL; 707 bridge_get_softc_p = NULL; 708 bridge_member_ifaddrs_p = NULL; 709 break; 710 default: 711 return (EOPNOTSUPP); 712 } 713 return (0); 714 } 715 716 static moduledata_t bridge_mod = { 717 "if_bridge", 718 bridge_modevent, 719 0 720 }; 721 722 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 723 MODULE_VERSION(if_bridge, 1); 724 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1); 725 726 /* 727 * handler for net.link.bridge.ipfw 728 */ 729 static int 730 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS) 731 { 732 int enable = V_pfil_ipfw; 733 int error; 734 735 error = sysctl_handle_int(oidp, &enable, 0, req); 736 enable &= 1; 737 738 if (enable != V_pfil_ipfw) { 739 V_pfil_ipfw = enable; 740 741 /* 742 * Disable pfil so that ipfw doesnt run twice, if the user 743 * really wants both then they can re-enable pfil_bridge and/or 744 * pfil_member. Also allow non-ip packets as ipfw can filter by 745 * layer2 type. 746 */ 747 if (V_pfil_ipfw) { 748 V_pfil_onlyip = 0; 749 V_pfil_bridge = 0; 750 V_pfil_member = 0; 751 } 752 } 753 754 return (error); 755 } 756 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, 757 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT, 758 &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I", 759 "Layer2 filter with IPFW"); 760 761 #ifdef VIMAGE 762 static void 763 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg) 764 { 765 struct bridge_softc *sc = ifp->if_softc; 766 struct bridge_iflist *bif; 767 768 BRIDGE_LOCK(sc); 769 770 while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL) 771 bridge_delete_member(sc, bif, 0); 772 773 while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) { 774 bridge_delete_span(sc, bif); 775 } 776 777 BRIDGE_UNLOCK(sc); 778 779 ether_reassign(ifp, newvnet, arg); 780 } 781 #endif 782 783 /* 784 * bridge_get_softc: 785 * 786 * Return the bridge softc for an ifnet. 787 */ 788 static void * 789 bridge_get_softc(struct ifnet *ifp) 790 { 791 struct bridge_iflist *bif; 792 793 NET_EPOCH_ASSERT(); 794 795 bif = ifp->if_bridge; 796 if (bif == NULL) 797 return (NULL); 798 return (bif->bif_sc); 799 } 800 801 /* 802 * bridge_same: 803 * 804 * Return true if two interfaces are in the same bridge. This is only used by 805 * bridgestp via bridge_same_p. 806 */ 807 static bool 808 bridge_same(const void *bifap, const void *bifbp) 809 { 810 const struct bridge_iflist *bifa = bifap, *bifb = bifbp; 811 812 NET_EPOCH_ASSERT(); 813 814 if (bifa == NULL || bifb == NULL) 815 return (false); 816 817 return (bifa->bif_sc == bifb->bif_sc); 818 } 819 820 /* 821 * bridge_clone_create: 822 * 823 * Create a new bridge instance. 824 */ 825 static int 826 bridge_clone_create(struct if_clone *ifc, char *name, size_t len, 827 struct ifc_data *ifd, struct ifnet **ifpp) 828 { 829 struct bridge_softc *sc; 830 struct ifnet *ifp; 831 832 sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO); 833 ifp = sc->sc_ifp = if_alloc(IFT_ETHER); 834 835 BRIDGE_LOCK_INIT(sc); 836 sc->sc_brtmax = BRIDGE_RTABLE_MAX; 837 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT; 838 839 /* Initialize our routing table. */ 840 bridge_rtable_init(sc); 841 842 callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0); 843 844 CK_LIST_INIT(&sc->sc_iflist); 845 CK_LIST_INIT(&sc->sc_spanlist); 846 847 ifp->if_softc = sc; 848 if_initname(ifp, bridge_name, ifd->unit); 849 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 850 ifp->if_capabilities = ifp->if_capenable = IFCAP_VLAN_HWTAGGING; 851 ifp->if_ioctl = bridge_ioctl; 852 #ifdef ALTQ 853 ifp->if_start = bridge_altq_start; 854 ifp->if_transmit = bridge_altq_transmit; 855 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 856 ifp->if_snd.ifq_drv_maxlen = 0; 857 IFQ_SET_READY(&ifp->if_snd); 858 #else 859 ifp->if_transmit = bridge_transmit; 860 #endif 861 ifp->if_qflush = bridge_qflush; 862 ifp->if_init = bridge_init; 863 ifp->if_type = IFT_BRIDGE; 864 865 ether_gen_addr(ifp, &sc->sc_defaddr); 866 867 bstp_attach(&sc->sc_stp, &bridge_ops); 868 ether_ifattach(ifp, sc->sc_defaddr.octet); 869 /* Now undo some of the damage... */ 870 ifp->if_baudrate = 0; 871 #ifdef VIMAGE 872 ifp->if_reassign = bridge_reassign; 873 #endif 874 sc->sc_if_input = ifp->if_input; /* ether_input */ 875 ifp->if_input = bridge_inject; 876 877 /* 878 * Allow BRIDGE_INPUT() to pass in packets originating from the bridge 879 * itself via bridge_inject(). This is required for netmap but 880 * otherwise has no effect. 881 */ 882 ifp->if_bridge_input = bridge_input; 883 884 BRIDGE_LIST_LOCK(); 885 LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list); 886 BRIDGE_LIST_UNLOCK(); 887 *ifpp = ifp; 888 889 return (0); 890 } 891 892 static void 893 bridge_clone_destroy_cb(struct epoch_context *ctx) 894 { 895 struct bridge_softc *sc; 896 897 sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx); 898 899 BRIDGE_LOCK_DESTROY(sc); 900 free(sc, M_DEVBUF); 901 } 902 903 /* 904 * bridge_clone_destroy: 905 * 906 * Destroy a bridge instance. 907 */ 908 static int 909 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags) 910 { 911 struct bridge_softc *sc = ifp->if_softc; 912 struct bridge_iflist *bif; 913 struct epoch_tracker et; 914 915 BRIDGE_LOCK(sc); 916 917 bridge_stop(ifp, 1); 918 ifp->if_flags &= ~IFF_UP; 919 920 while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL) 921 bridge_delete_member(sc, bif, 0); 922 923 while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) { 924 bridge_delete_span(sc, bif); 925 } 926 927 /* Tear down the routing table. */ 928 bridge_rtable_fini(sc); 929 930 BRIDGE_UNLOCK(sc); 931 932 NET_EPOCH_ENTER(et); 933 934 callout_drain(&sc->sc_brcallout); 935 936 BRIDGE_LIST_LOCK(); 937 LIST_REMOVE(sc, sc_list); 938 BRIDGE_LIST_UNLOCK(); 939 940 bstp_detach(&sc->sc_stp); 941 #ifdef ALTQ 942 IFQ_PURGE(&ifp->if_snd); 943 #endif 944 NET_EPOCH_EXIT(et); 945 946 ether_ifdetach(ifp); 947 if_free(ifp); 948 949 NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx); 950 951 return (0); 952 } 953 954 /* 955 * bridge_ioctl: 956 * 957 * Handle a control request from the operator. 958 */ 959 static int 960 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 961 { 962 struct bridge_softc *sc = ifp->if_softc; 963 struct ifreq *ifr = (struct ifreq *)data; 964 struct bridge_iflist *bif; 965 struct thread *td = curthread; 966 union { 967 struct ifbreq ifbreq; 968 struct ifbifconf ifbifconf; 969 struct ifbareq ifbareq; 970 struct ifbaconf ifbaconf; 971 struct ifbrparam ifbrparam; 972 struct ifbropreq ifbropreq; 973 struct ifbif_vlan_req ifvlanreq; 974 } args; 975 struct ifdrv *ifd = (struct ifdrv *) data; 976 const struct bridge_control *bc; 977 int error = 0, oldmtu; 978 979 BRIDGE_LOCK(sc); 980 981 switch (cmd) { 982 case SIOCADDMULTI: 983 case SIOCDELMULTI: 984 break; 985 986 case SIOCGDRVSPEC: 987 case SIOCSDRVSPEC: 988 if (ifd->ifd_cmd >= bridge_control_table_size) { 989 error = EINVAL; 990 break; 991 } 992 bc = &bridge_control_table[ifd->ifd_cmd]; 993 994 if (cmd == SIOCGDRVSPEC && 995 (bc->bc_flags & BC_F_COPYOUT) == 0) { 996 error = EINVAL; 997 break; 998 } 999 else if (cmd == SIOCSDRVSPEC && 1000 (bc->bc_flags & BC_F_COPYOUT) != 0) { 1001 error = EINVAL; 1002 break; 1003 } 1004 1005 if (bc->bc_flags & BC_F_SUSER) { 1006 error = priv_check(td, PRIV_NET_BRIDGE); 1007 if (error) 1008 break; 1009 } 1010 1011 if (ifd->ifd_len != bc->bc_argsize || 1012 ifd->ifd_len > sizeof(args)) { 1013 error = EINVAL; 1014 break; 1015 } 1016 1017 bzero(&args, sizeof(args)); 1018 if (bc->bc_flags & BC_F_COPYIN) { 1019 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); 1020 if (error) 1021 break; 1022 } 1023 1024 oldmtu = ifp->if_mtu; 1025 error = (*bc->bc_func)(sc, &args); 1026 if (error) 1027 break; 1028 1029 /* 1030 * Bridge MTU may change during addition of the first port. 1031 * If it did, do network layer specific procedure. 1032 */ 1033 if (ifp->if_mtu != oldmtu) 1034 if_notifymtu(ifp); 1035 1036 if (bc->bc_flags & BC_F_COPYOUT) 1037 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); 1038 1039 break; 1040 1041 case SIOCSIFFLAGS: 1042 if (!(ifp->if_flags & IFF_UP) && 1043 (ifp->if_drv_flags & IFF_DRV_RUNNING)) { 1044 /* 1045 * If interface is marked down and it is running, 1046 * then stop and disable it. 1047 */ 1048 bridge_stop(ifp, 1); 1049 } else if ((ifp->if_flags & IFF_UP) && 1050 !(ifp->if_drv_flags & IFF_DRV_RUNNING)) { 1051 /* 1052 * If interface is marked up and it is stopped, then 1053 * start it. 1054 */ 1055 BRIDGE_UNLOCK(sc); 1056 (*ifp->if_init)(sc); 1057 BRIDGE_LOCK(sc); 1058 } 1059 break; 1060 1061 case SIOCSIFMTU: 1062 oldmtu = sc->sc_ifp->if_mtu; 1063 1064 if (ifr->ifr_mtu < IF_MINMTU) { 1065 error = EINVAL; 1066 break; 1067 } 1068 if (CK_LIST_EMPTY(&sc->sc_iflist)) { 1069 sc->sc_ifp->if_mtu = ifr->ifr_mtu; 1070 break; 1071 } 1072 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 1073 error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp, 1074 SIOCSIFMTU, (caddr_t)ifr); 1075 if (error != 0) { 1076 log(LOG_NOTICE, "%s: invalid MTU: %u for" 1077 " member %s\n", sc->sc_ifp->if_xname, 1078 ifr->ifr_mtu, 1079 bif->bif_ifp->if_xname); 1080 error = EINVAL; 1081 break; 1082 } 1083 } 1084 if (error) { 1085 /* Restore the previous MTU on all member interfaces. */ 1086 ifr->ifr_mtu = oldmtu; 1087 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 1088 (*bif->bif_ifp->if_ioctl)(bif->bif_ifp, 1089 SIOCSIFMTU, (caddr_t)ifr); 1090 } 1091 } else { 1092 sc->sc_ifp->if_mtu = ifr->ifr_mtu; 1093 } 1094 break; 1095 default: 1096 /* 1097 * drop the lock as ether_ioctl() will call bridge_start() and 1098 * cause the lock to be recursed. 1099 */ 1100 BRIDGE_UNLOCK(sc); 1101 error = ether_ioctl(ifp, cmd, data); 1102 BRIDGE_LOCK(sc); 1103 break; 1104 } 1105 1106 BRIDGE_UNLOCK(sc); 1107 1108 return (error); 1109 } 1110 1111 /* 1112 * bridge_mutecaps: 1113 * 1114 * Clear or restore unwanted capabilities on the member interface 1115 */ 1116 static void 1117 bridge_mutecaps(struct bridge_softc *sc) 1118 { 1119 struct bridge_iflist *bif; 1120 int enabled, mask; 1121 1122 BRIDGE_LOCK_ASSERT(sc); 1123 1124 /* Initial bitmask of capabilities to test */ 1125 mask = BRIDGE_IFCAPS_MASK; 1126 1127 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 1128 /* Every member must support it or its disabled */ 1129 mask &= bif->bif_savedcaps; 1130 } 1131 1132 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 1133 enabled = bif->bif_ifp->if_capenable; 1134 enabled &= ~BRIDGE_IFCAPS_STRIP; 1135 /* strip off mask bits and enable them again if allowed */ 1136 enabled &= ~BRIDGE_IFCAPS_MASK; 1137 enabled |= mask; 1138 bridge_set_ifcap(sc, bif, enabled); 1139 } 1140 } 1141 1142 static void 1143 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set) 1144 { 1145 struct ifnet *ifp = bif->bif_ifp; 1146 struct ifreq ifr; 1147 int error, mask, stuck; 1148 1149 bzero(&ifr, sizeof(ifr)); 1150 ifr.ifr_reqcap = set; 1151 1152 if (ifp->if_capenable != set) { 1153 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr); 1154 if (error) 1155 if_printf(sc->sc_ifp, 1156 "error setting capabilities on %s: %d\n", 1157 ifp->if_xname, error); 1158 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP; 1159 stuck = ifp->if_capenable & mask & ~set; 1160 if (stuck != 0) 1161 if_printf(sc->sc_ifp, 1162 "can't disable some capabilities on %s: 0x%x\n", 1163 ifp->if_xname, stuck); 1164 } 1165 } 1166 1167 /* 1168 * bridge_lookup_member: 1169 * 1170 * Lookup a bridge member interface. 1171 */ 1172 static struct bridge_iflist * 1173 bridge_lookup_member(struct bridge_softc *sc, const char *name) 1174 { 1175 struct bridge_iflist *bif; 1176 struct ifnet *ifp; 1177 1178 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc); 1179 1180 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 1181 ifp = bif->bif_ifp; 1182 if (strcmp(ifp->if_xname, name) == 0) 1183 return (bif); 1184 } 1185 1186 return (NULL); 1187 } 1188 1189 /* 1190 * bridge_lookup_member_if: 1191 * 1192 * Lookup a bridge member interface by ifnet*. 1193 */ 1194 static struct bridge_iflist * 1195 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp) 1196 { 1197 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc); 1198 return (member_ifp->if_bridge); 1199 } 1200 1201 static void 1202 bridge_delete_member_cb(struct epoch_context *ctx) 1203 { 1204 struct bridge_iflist *bif; 1205 1206 bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx); 1207 1208 free(bif, M_DEVBUF); 1209 } 1210 1211 /* 1212 * bridge_delete_member: 1213 * 1214 * Delete the specified member interface. 1215 */ 1216 static void 1217 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif, 1218 int gone) 1219 { 1220 struct ifnet *ifs = bif->bif_ifp; 1221 struct ifnet *fif = NULL; 1222 struct bridge_iflist *bifl; 1223 1224 BRIDGE_LOCK_ASSERT(sc); 1225 1226 if (bif->bif_flags & IFBIF_STP) 1227 bstp_disable(&bif->bif_stp); 1228 1229 ifs->if_bridge = NULL; 1230 CK_LIST_REMOVE(bif, bif_next); 1231 1232 /* 1233 * If removing the interface that gave the bridge its mac address, set 1234 * the mac address of the bridge to the address of the next member, or 1235 * to its default address if no members are left. 1236 */ 1237 if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) { 1238 if (CK_LIST_EMPTY(&sc->sc_iflist)) { 1239 bcopy(&sc->sc_defaddr, 1240 IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN); 1241 sc->sc_ifaddr = NULL; 1242 } else { 1243 bifl = CK_LIST_FIRST(&sc->sc_iflist); 1244 fif = bifl->bif_ifp; 1245 bcopy(IF_LLADDR(fif), 1246 IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN); 1247 sc->sc_ifaddr = fif; 1248 } 1249 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp); 1250 } 1251 1252 bridge_linkcheck(sc); 1253 bridge_mutecaps(sc); /* recalcuate now this interface is removed */ 1254 BRIDGE_RT_LOCK(sc); 1255 bridge_rtdelete(sc, ifs, IFBF_FLUSHALL); 1256 BRIDGE_RT_UNLOCK(sc); 1257 KASSERT(bif->bif_addrcnt == 0, 1258 ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt)); 1259 1260 ifs->if_bridge_output = NULL; 1261 ifs->if_bridge_input = NULL; 1262 ifs->if_bridge_linkstate = NULL; 1263 if (!gone) { 1264 switch (ifs->if_type) { 1265 case IFT_ETHER: 1266 case IFT_L2VLAN: 1267 /* 1268 * Take the interface out of promiscuous mode, but only 1269 * if it was promiscuous in the first place. It might 1270 * not be if we're in the bridge_ioctl_add() error path. 1271 */ 1272 if (ifs->if_flags & IFF_PROMISC) 1273 (void) ifpromisc(ifs, 0); 1274 break; 1275 1276 case IFT_GIF: 1277 break; 1278 1279 default: 1280 #ifdef DIAGNOSTIC 1281 panic("bridge_delete_member: impossible"); 1282 #endif 1283 break; 1284 } 1285 /* reneable any interface capabilities */ 1286 bridge_set_ifcap(sc, bif, bif->bif_savedcaps); 1287 } 1288 bstp_destroy(&bif->bif_stp); /* prepare to free */ 1289 1290 NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx); 1291 } 1292 1293 /* 1294 * bridge_delete_span: 1295 * 1296 * Delete the specified span interface. 1297 */ 1298 static void 1299 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif) 1300 { 1301 BRIDGE_LOCK_ASSERT(sc); 1302 1303 KASSERT(bif->bif_ifp->if_bridge == NULL, 1304 ("%s: not a span interface", __func__)); 1305 1306 CK_LIST_REMOVE(bif, bif_next); 1307 1308 NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx); 1309 } 1310 1311 static int 1312 bridge_ioctl_add(struct bridge_softc *sc, void *arg) 1313 { 1314 struct ifbreq *req = arg; 1315 struct bridge_iflist *bif = NULL; 1316 struct ifnet *ifs; 1317 int error = 0; 1318 1319 ifs = ifunit(req->ifbr_ifsname); 1320 if (ifs == NULL) 1321 return (ENOENT); 1322 if (ifs->if_ioctl == NULL) /* must be supported */ 1323 return (EINVAL); 1324 1325 /* If it's in the span list, it can't be a member. */ 1326 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1327 if (ifs == bif->bif_ifp) 1328 return (EBUSY); 1329 1330 if (ifs->if_bridge) { 1331 struct bridge_iflist *sbif = ifs->if_bridge; 1332 if (sbif->bif_sc == sc) 1333 return (EEXIST); 1334 1335 return (EBUSY); 1336 } 1337 1338 switch (ifs->if_type) { 1339 case IFT_ETHER: 1340 case IFT_L2VLAN: 1341 case IFT_GIF: 1342 /* permitted interface types */ 1343 break; 1344 default: 1345 return (EINVAL); 1346 } 1347 1348 #ifdef INET6 1349 /* 1350 * Two valid inet6 addresses with link-local scope must not be 1351 * on the parent interface and the member interfaces at the 1352 * same time. This restriction is needed to prevent violation 1353 * of link-local scope zone. Attempts to add a member 1354 * interface which has inet6 addresses when the parent has 1355 * inet6 triggers removal of all inet6 addresses on the member 1356 * interface. 1357 */ 1358 1359 /* Check if the parent interface has a link-local scope addr. */ 1360 if (V_allow_llz_overlap == 0 && 1361 in6ifa_llaonifp(sc->sc_ifp) != NULL) { 1362 /* 1363 * If any, remove all inet6 addresses from the member 1364 * interfaces. 1365 */ 1366 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 1367 if (in6ifa_llaonifp(bif->bif_ifp)) { 1368 in6_ifdetach(bif->bif_ifp); 1369 if_printf(sc->sc_ifp, 1370 "IPv6 addresses on %s have been removed " 1371 "before adding it as a member to prevent " 1372 "IPv6 address scope violation.\n", 1373 bif->bif_ifp->if_xname); 1374 } 1375 } 1376 if (in6ifa_llaonifp(ifs)) { 1377 in6_ifdetach(ifs); 1378 if_printf(sc->sc_ifp, 1379 "IPv6 addresses on %s have been removed " 1380 "before adding it as a member to prevent " 1381 "IPv6 address scope violation.\n", 1382 ifs->if_xname); 1383 } 1384 } 1385 #endif 1386 1387 /* 1388 * If member_ifaddrs is disabled, do not allow an interface with 1389 * assigned IP addresses to be added to a bridge. 1390 */ 1391 if (!V_member_ifaddrs) { 1392 struct ifaddr *ifa; 1393 1394 CK_STAILQ_FOREACH(ifa, &ifs->if_addrhead, ifa_link) { 1395 #ifdef INET 1396 if (ifa->ifa_addr->sa_family == AF_INET) 1397 return (EINVAL); 1398 #endif 1399 #ifdef INET6 1400 if (ifa->ifa_addr->sa_family == AF_INET6) 1401 return (EINVAL); 1402 #endif 1403 } 1404 } 1405 1406 /* Allow the first Ethernet member to define the MTU */ 1407 if (CK_LIST_EMPTY(&sc->sc_iflist)) 1408 sc->sc_ifp->if_mtu = ifs->if_mtu; 1409 else if (sc->sc_ifp->if_mtu != ifs->if_mtu) { 1410 struct ifreq ifr; 1411 1412 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s", 1413 ifs->if_xname); 1414 ifr.ifr_mtu = sc->sc_ifp->if_mtu; 1415 1416 error = (*ifs->if_ioctl)(ifs, 1417 SIOCSIFMTU, (caddr_t)&ifr); 1418 if (error != 0) { 1419 log(LOG_NOTICE, "%s: invalid MTU: %u for" 1420 " new member %s\n", sc->sc_ifp->if_xname, 1421 ifr.ifr_mtu, 1422 ifs->if_xname); 1423 return (EINVAL); 1424 } 1425 } 1426 1427 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO); 1428 if (bif == NULL) 1429 return (ENOMEM); 1430 1431 bif->bif_sc = sc; 1432 bif->bif_ifp = ifs; 1433 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER; 1434 bif->bif_savedcaps = ifs->if_capenable; 1435 1436 /* 1437 * Assign the interface's MAC address to the bridge if it's the first 1438 * member and the MAC address of the bridge has not been changed from 1439 * the default randomly generated one. 1440 */ 1441 if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) && 1442 !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) { 1443 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN); 1444 sc->sc_ifaddr = ifs; 1445 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp); 1446 } 1447 1448 ifs->if_bridge = bif; 1449 ifs->if_bridge_output = bridge_output; 1450 ifs->if_bridge_input = bridge_input; 1451 ifs->if_bridge_linkstate = bridge_linkstate; 1452 bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp); 1453 /* 1454 * XXX: XLOCK HERE!?! 1455 * 1456 * NOTE: insert_***HEAD*** should be safe for the traversals. 1457 */ 1458 CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next); 1459 1460 /* Set interface capabilities to the intersection set of all members */ 1461 bridge_mutecaps(sc); 1462 bridge_linkcheck(sc); 1463 1464 /* Place the interface into promiscuous mode */ 1465 switch (ifs->if_type) { 1466 case IFT_ETHER: 1467 case IFT_L2VLAN: 1468 error = ifpromisc(ifs, 1); 1469 break; 1470 } 1471 1472 if (error) 1473 bridge_delete_member(sc, bif, 0); 1474 return (error); 1475 } 1476 1477 static int 1478 bridge_ioctl_del(struct bridge_softc *sc, void *arg) 1479 { 1480 struct ifbreq *req = arg; 1481 struct bridge_iflist *bif; 1482 1483 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1484 if (bif == NULL) 1485 return (ENOENT); 1486 1487 bridge_delete_member(sc, bif, 0); 1488 1489 return (0); 1490 } 1491 1492 static int 1493 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg) 1494 { 1495 struct ifbreq *req = arg; 1496 struct bridge_iflist *bif; 1497 struct bstp_port *bp; 1498 1499 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1500 if (bif == NULL) 1501 return (ENOENT); 1502 1503 bp = &bif->bif_stp; 1504 req->ifbr_ifsflags = bif->bif_flags; 1505 req->ifbr_state = bp->bp_state; 1506 req->ifbr_priority = bp->bp_priority; 1507 req->ifbr_path_cost = bp->bp_path_cost; 1508 req->ifbr_portno = bif->bif_ifp->if_index & 0xfff; 1509 req->ifbr_proto = bp->bp_protover; 1510 req->ifbr_role = bp->bp_role; 1511 req->ifbr_stpflags = bp->bp_flags; 1512 req->ifbr_addrcnt = bif->bif_addrcnt; 1513 req->ifbr_addrmax = bif->bif_addrmax; 1514 req->ifbr_addrexceeded = bif->bif_addrexceeded; 1515 req->ifbr_untagged = bif->bif_untagged; 1516 1517 /* Copy STP state options as flags */ 1518 if (bp->bp_operedge) 1519 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE; 1520 if (bp->bp_flags & BSTP_PORT_AUTOEDGE) 1521 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE; 1522 if (bp->bp_ptp_link) 1523 req->ifbr_ifsflags |= IFBIF_BSTP_PTP; 1524 if (bp->bp_flags & BSTP_PORT_AUTOPTP) 1525 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP; 1526 if (bp->bp_flags & BSTP_PORT_ADMEDGE) 1527 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE; 1528 if (bp->bp_flags & BSTP_PORT_ADMCOST) 1529 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST; 1530 return (0); 1531 } 1532 1533 static int 1534 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg) 1535 { 1536 struct epoch_tracker et; 1537 struct ifbreq *req = arg; 1538 struct bridge_iflist *bif; 1539 struct bstp_port *bp; 1540 int error; 1541 1542 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1543 if (bif == NULL) 1544 return (ENOENT); 1545 bp = &bif->bif_stp; 1546 1547 if (req->ifbr_ifsflags & IFBIF_SPAN) 1548 /* SPAN is readonly */ 1549 return (EINVAL); 1550 1551 NET_EPOCH_ENTER(et); 1552 1553 if (req->ifbr_ifsflags & IFBIF_STP) { 1554 if ((bif->bif_flags & IFBIF_STP) == 0) { 1555 error = bstp_enable(&bif->bif_stp); 1556 if (error) { 1557 NET_EPOCH_EXIT(et); 1558 return (error); 1559 } 1560 } 1561 } else { 1562 if ((bif->bif_flags & IFBIF_STP) != 0) 1563 bstp_disable(&bif->bif_stp); 1564 } 1565 1566 /* Pass on STP flags */ 1567 bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0); 1568 bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0); 1569 bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0); 1570 bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0); 1571 1572 /* Save the bits relating to the bridge */ 1573 bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK; 1574 1575 NET_EPOCH_EXIT(et); 1576 1577 return (0); 1578 } 1579 1580 static int 1581 bridge_ioctl_scache(struct bridge_softc *sc, void *arg) 1582 { 1583 struct ifbrparam *param = arg; 1584 1585 sc->sc_brtmax = param->ifbrp_csize; 1586 bridge_rttrim(sc); 1587 1588 return (0); 1589 } 1590 1591 static int 1592 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg) 1593 { 1594 struct ifbrparam *param = arg; 1595 1596 param->ifbrp_csize = sc->sc_brtmax; 1597 1598 return (0); 1599 } 1600 1601 static int 1602 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg) 1603 { 1604 struct ifbifconf *bifc = arg; 1605 struct bridge_iflist *bif; 1606 struct ifbreq breq; 1607 char *buf, *outbuf; 1608 int count, buflen, len, error = 0; 1609 1610 count = 0; 1611 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) 1612 count++; 1613 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1614 count++; 1615 1616 buflen = sizeof(breq) * count; 1617 if (bifc->ifbic_len == 0) { 1618 bifc->ifbic_len = buflen; 1619 return (0); 1620 } 1621 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO); 1622 if (outbuf == NULL) 1623 return (ENOMEM); 1624 1625 count = 0; 1626 buf = outbuf; 1627 len = min(bifc->ifbic_len, buflen); 1628 bzero(&breq, sizeof(breq)); 1629 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 1630 if (len < sizeof(breq)) 1631 break; 1632 1633 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname, 1634 sizeof(breq.ifbr_ifsname)); 1635 /* Fill in the ifbreq structure */ 1636 error = bridge_ioctl_gifflags(sc, &breq); 1637 if (error) 1638 break; 1639 memcpy(buf, &breq, sizeof(breq)); 1640 count++; 1641 buf += sizeof(breq); 1642 len -= sizeof(breq); 1643 } 1644 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) { 1645 if (len < sizeof(breq)) 1646 break; 1647 1648 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname, 1649 sizeof(breq.ifbr_ifsname)); 1650 breq.ifbr_ifsflags = bif->bif_flags; 1651 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff; 1652 memcpy(buf, &breq, sizeof(breq)); 1653 count++; 1654 buf += sizeof(breq); 1655 len -= sizeof(breq); 1656 } 1657 1658 bifc->ifbic_len = sizeof(breq) * count; 1659 error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len); 1660 free(outbuf, M_TEMP); 1661 return (error); 1662 } 1663 1664 static int 1665 bridge_ioctl_rts(struct bridge_softc *sc, void *arg) 1666 { 1667 struct ifbaconf *bac = arg; 1668 struct bridge_rtnode *brt; 1669 struct ifbareq bareq; 1670 char *buf, *outbuf; 1671 int count, buflen, len, error = 0; 1672 1673 if (bac->ifbac_len == 0) 1674 return (0); 1675 1676 count = 0; 1677 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) 1678 count++; 1679 buflen = sizeof(bareq) * count; 1680 1681 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO); 1682 if (outbuf == NULL) 1683 return (ENOMEM); 1684 1685 count = 0; 1686 buf = outbuf; 1687 len = min(bac->ifbac_len, buflen); 1688 bzero(&bareq, sizeof(bareq)); 1689 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) { 1690 if (len < sizeof(bareq)) 1691 goto out; 1692 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname, 1693 sizeof(bareq.ifba_ifsname)); 1694 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr)); 1695 bareq.ifba_vlan = brt->brt_vlan; 1696 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 1697 time_uptime < brt->brt_expire) 1698 bareq.ifba_expire = brt->brt_expire - time_uptime; 1699 else 1700 bareq.ifba_expire = 0; 1701 bareq.ifba_flags = brt->brt_flags; 1702 1703 memcpy(buf, &bareq, sizeof(bareq)); 1704 count++; 1705 buf += sizeof(bareq); 1706 len -= sizeof(bareq); 1707 } 1708 out: 1709 bac->ifbac_len = sizeof(bareq) * count; 1710 error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len); 1711 free(outbuf, M_TEMP); 1712 return (error); 1713 } 1714 1715 static int 1716 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg) 1717 { 1718 struct ifbareq *req = arg; 1719 struct bridge_iflist *bif; 1720 struct epoch_tracker et; 1721 int error; 1722 1723 NET_EPOCH_ENTER(et); 1724 bif = bridge_lookup_member(sc, req->ifba_ifsname); 1725 if (bif == NULL) { 1726 NET_EPOCH_EXIT(et); 1727 return (ENOENT); 1728 } 1729 1730 /* bridge_rtupdate() may acquire the lock. */ 1731 error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1, 1732 req->ifba_flags); 1733 NET_EPOCH_EXIT(et); 1734 1735 return (error); 1736 } 1737 1738 static int 1739 bridge_ioctl_sto(struct bridge_softc *sc, void *arg) 1740 { 1741 struct ifbrparam *param = arg; 1742 1743 sc->sc_brttimeout = param->ifbrp_ctime; 1744 return (0); 1745 } 1746 1747 static int 1748 bridge_ioctl_gto(struct bridge_softc *sc, void *arg) 1749 { 1750 struct ifbrparam *param = arg; 1751 1752 param->ifbrp_ctime = sc->sc_brttimeout; 1753 return (0); 1754 } 1755 1756 static int 1757 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg) 1758 { 1759 struct ifbareq *req = arg; 1760 int vlan = req->ifba_vlan; 1761 1762 /* Userspace uses '0' to mean 'any vlan' */ 1763 if (vlan == 0) 1764 vlan = DOT1Q_VID_RSVD_IMPL; 1765 1766 return (bridge_rtdaddr(sc, req->ifba_dst, vlan)); 1767 } 1768 1769 static int 1770 bridge_ioctl_flush(struct bridge_softc *sc, void *arg) 1771 { 1772 struct ifbreq *req = arg; 1773 1774 BRIDGE_RT_LOCK(sc); 1775 bridge_rtflush(sc, req->ifbr_ifsflags); 1776 BRIDGE_RT_UNLOCK(sc); 1777 1778 return (0); 1779 } 1780 1781 static int 1782 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg) 1783 { 1784 struct ifbrparam *param = arg; 1785 struct bstp_state *bs = &sc->sc_stp; 1786 1787 param->ifbrp_prio = bs->bs_bridge_priority; 1788 return (0); 1789 } 1790 1791 static int 1792 bridge_ioctl_spri(struct bridge_softc *sc, void *arg) 1793 { 1794 struct ifbrparam *param = arg; 1795 1796 return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio)); 1797 } 1798 1799 static int 1800 bridge_ioctl_ght(struct bridge_softc *sc, void *arg) 1801 { 1802 struct ifbrparam *param = arg; 1803 struct bstp_state *bs = &sc->sc_stp; 1804 1805 param->ifbrp_hellotime = bs->bs_bridge_htime >> 8; 1806 return (0); 1807 } 1808 1809 static int 1810 bridge_ioctl_sht(struct bridge_softc *sc, void *arg) 1811 { 1812 struct ifbrparam *param = arg; 1813 1814 return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime)); 1815 } 1816 1817 static int 1818 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg) 1819 { 1820 struct ifbrparam *param = arg; 1821 struct bstp_state *bs = &sc->sc_stp; 1822 1823 param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8; 1824 return (0); 1825 } 1826 1827 static int 1828 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg) 1829 { 1830 struct ifbrparam *param = arg; 1831 1832 return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay)); 1833 } 1834 1835 static int 1836 bridge_ioctl_gma(struct bridge_softc *sc, void *arg) 1837 { 1838 struct ifbrparam *param = arg; 1839 struct bstp_state *bs = &sc->sc_stp; 1840 1841 param->ifbrp_maxage = bs->bs_bridge_max_age >> 8; 1842 return (0); 1843 } 1844 1845 static int 1846 bridge_ioctl_sma(struct bridge_softc *sc, void *arg) 1847 { 1848 struct ifbrparam *param = arg; 1849 1850 return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage)); 1851 } 1852 1853 static int 1854 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg) 1855 { 1856 struct ifbreq *req = arg; 1857 struct bridge_iflist *bif; 1858 1859 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1860 if (bif == NULL) 1861 return (ENOENT); 1862 1863 return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority)); 1864 } 1865 1866 static int 1867 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg) 1868 { 1869 struct ifbreq *req = arg; 1870 struct bridge_iflist *bif; 1871 1872 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1873 if (bif == NULL) 1874 return (ENOENT); 1875 1876 return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost)); 1877 } 1878 1879 static int 1880 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg) 1881 { 1882 struct ifbreq *req = arg; 1883 struct bridge_iflist *bif; 1884 1885 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1886 if (bif == NULL) 1887 return (ENOENT); 1888 1889 bif->bif_addrmax = req->ifbr_addrmax; 1890 return (0); 1891 } 1892 1893 static int 1894 bridge_ioctl_sifuntagged(struct bridge_softc *sc, void *arg) 1895 { 1896 struct ifbreq *req = arg; 1897 struct bridge_iflist *bif; 1898 1899 bif = bridge_lookup_member(sc, req->ifbr_ifsname); 1900 if (bif == NULL) 1901 return (ENOENT); 1902 1903 if (req->ifbr_untagged > DOT1Q_VID_MAX) 1904 return (EINVAL); 1905 1906 if (req->ifbr_untagged != DOT1Q_VID_NULL) 1907 bif->bif_flags |= IFBIF_VLANFILTER; 1908 bif->bif_untagged = req->ifbr_untagged; 1909 return (0); 1910 } 1911 1912 static int 1913 bridge_ioctl_sifvlanset(struct bridge_softc *sc, void *arg) 1914 { 1915 struct ifbif_vlan_req *req = arg; 1916 struct bridge_iflist *bif; 1917 1918 bif = bridge_lookup_member(sc, req->bv_ifname); 1919 if (bif == NULL) 1920 return (ENOENT); 1921 1922 /* Reject invalid VIDs. */ 1923 if (BRVLAN_TEST(&req->bv_set, DOT1Q_VID_NULL) || 1924 BRVLAN_TEST(&req->bv_set, DOT1Q_VID_RSVD_IMPL)) 1925 return (EINVAL); 1926 1927 switch (req->bv_op) { 1928 /* Replace the existing vlan set with the new set */ 1929 case BRDG_VLAN_OP_SET: 1930 BIT_COPY(BRVLAN_SETSIZE, &req->bv_set, &bif->bif_vlan_set); 1931 break; 1932 1933 /* Modify the existing vlan set to add the given vlans */ 1934 case BRDG_VLAN_OP_ADD: 1935 BIT_OR(BRVLAN_SETSIZE, &bif->bif_vlan_set, &req->bv_set); 1936 break; 1937 1938 /* Modify the existing vlan set to remove the given vlans */ 1939 case BRDG_VLAN_OP_DEL: 1940 BIT_ANDNOT(BRVLAN_SETSIZE, &bif->bif_vlan_set, &req->bv_set); 1941 break; 1942 1943 /* Invalid or unknown operation */ 1944 default: 1945 return (EINVAL); 1946 } 1947 1948 /* 1949 * The only reason to modify the VLAN access list is to use VLAN 1950 * filtering on this interface, so enable it automatically. 1951 */ 1952 bif->bif_flags |= IFBIF_VLANFILTER; 1953 1954 return (0); 1955 } 1956 1957 static int 1958 bridge_ioctl_gifvlanset(struct bridge_softc *sc, void *arg) 1959 { 1960 struct ifbif_vlan_req *req = arg; 1961 struct bridge_iflist *bif; 1962 1963 bif = bridge_lookup_member(sc, req->bv_ifname); 1964 if (bif == NULL) 1965 return (ENOENT); 1966 1967 BIT_COPY(BRVLAN_SETSIZE, &bif->bif_vlan_set, &req->bv_set); 1968 return (0); 1969 } 1970 1971 static int 1972 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg) 1973 { 1974 struct ifbreq *req = arg; 1975 struct bridge_iflist *bif = NULL; 1976 struct ifnet *ifs; 1977 1978 ifs = ifunit(req->ifbr_ifsname); 1979 if (ifs == NULL) 1980 return (ENOENT); 1981 1982 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 1983 if (ifs == bif->bif_ifp) 1984 return (EBUSY); 1985 1986 if (ifs->if_bridge != NULL) 1987 return (EBUSY); 1988 1989 switch (ifs->if_type) { 1990 case IFT_ETHER: 1991 case IFT_GIF: 1992 case IFT_L2VLAN: 1993 break; 1994 default: 1995 return (EINVAL); 1996 } 1997 1998 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO); 1999 if (bif == NULL) 2000 return (ENOMEM); 2001 2002 bif->bif_ifp = ifs; 2003 bif->bif_flags = IFBIF_SPAN; 2004 2005 CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next); 2006 2007 return (0); 2008 } 2009 2010 static int 2011 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg) 2012 { 2013 struct ifbreq *req = arg; 2014 struct bridge_iflist *bif; 2015 struct ifnet *ifs; 2016 2017 ifs = ifunit(req->ifbr_ifsname); 2018 if (ifs == NULL) 2019 return (ENOENT); 2020 2021 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 2022 if (ifs == bif->bif_ifp) 2023 break; 2024 2025 if (bif == NULL) 2026 return (ENOENT); 2027 2028 bridge_delete_span(sc, bif); 2029 2030 return (0); 2031 } 2032 2033 static int 2034 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg) 2035 { 2036 struct ifbropreq *req = arg; 2037 struct bstp_state *bs = &sc->sc_stp; 2038 struct bstp_port *root_port; 2039 2040 req->ifbop_maxage = bs->bs_bridge_max_age >> 8; 2041 req->ifbop_hellotime = bs->bs_bridge_htime >> 8; 2042 req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8; 2043 2044 root_port = bs->bs_root_port; 2045 if (root_port == NULL) 2046 req->ifbop_root_port = 0; 2047 else 2048 req->ifbop_root_port = root_port->bp_ifp->if_index; 2049 2050 req->ifbop_holdcount = bs->bs_txholdcount; 2051 req->ifbop_priority = bs->bs_bridge_priority; 2052 req->ifbop_protocol = bs->bs_protover; 2053 req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost; 2054 req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id; 2055 req->ifbop_designated_root = bs->bs_root_pv.pv_root_id; 2056 req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id; 2057 req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec; 2058 req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec; 2059 2060 return (0); 2061 } 2062 2063 static int 2064 bridge_ioctl_grte(struct bridge_softc *sc, void *arg) 2065 { 2066 struct ifbrparam *param = arg; 2067 2068 param->ifbrp_cexceeded = sc->sc_brtexceeded; 2069 return (0); 2070 } 2071 2072 static int 2073 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg) 2074 { 2075 struct ifbpstpconf *bifstp = arg; 2076 struct bridge_iflist *bif; 2077 struct bstp_port *bp; 2078 struct ifbpstpreq bpreq; 2079 char *buf, *outbuf; 2080 int count, buflen, len, error = 0; 2081 2082 count = 0; 2083 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 2084 if ((bif->bif_flags & IFBIF_STP) != 0) 2085 count++; 2086 } 2087 2088 buflen = sizeof(bpreq) * count; 2089 if (bifstp->ifbpstp_len == 0) { 2090 bifstp->ifbpstp_len = buflen; 2091 return (0); 2092 } 2093 2094 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO); 2095 if (outbuf == NULL) 2096 return (ENOMEM); 2097 2098 count = 0; 2099 buf = outbuf; 2100 len = min(bifstp->ifbpstp_len, buflen); 2101 bzero(&bpreq, sizeof(bpreq)); 2102 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 2103 if (len < sizeof(bpreq)) 2104 break; 2105 2106 if ((bif->bif_flags & IFBIF_STP) == 0) 2107 continue; 2108 2109 bp = &bif->bif_stp; 2110 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff; 2111 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions; 2112 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost; 2113 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id; 2114 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id; 2115 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id; 2116 2117 memcpy(buf, &bpreq, sizeof(bpreq)); 2118 count++; 2119 buf += sizeof(bpreq); 2120 len -= sizeof(bpreq); 2121 } 2122 2123 bifstp->ifbpstp_len = sizeof(bpreq) * count; 2124 error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len); 2125 free(outbuf, M_TEMP); 2126 return (error); 2127 } 2128 2129 static int 2130 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg) 2131 { 2132 struct ifbrparam *param = arg; 2133 2134 return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto)); 2135 } 2136 2137 static int 2138 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg) 2139 { 2140 struct ifbrparam *param = arg; 2141 2142 return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc)); 2143 } 2144 2145 /* 2146 * bridge_ifdetach: 2147 * 2148 * Detach an interface from a bridge. Called when a member 2149 * interface is detaching. 2150 */ 2151 static void 2152 bridge_ifdetach(void *arg __unused, struct ifnet *ifp) 2153 { 2154 struct bridge_iflist *bif = ifp->if_bridge; 2155 struct bridge_softc *sc = NULL; 2156 2157 if (bif) 2158 sc = bif->bif_sc; 2159 2160 if (ifp->if_flags & IFF_RENAMING) 2161 return; 2162 if (V_bridge_cloner == NULL) { 2163 /* 2164 * This detach handler can be called after 2165 * vnet_bridge_uninit(). Just return in that case. 2166 */ 2167 return; 2168 } 2169 /* Check if the interface is a bridge member */ 2170 if (sc != NULL) { 2171 BRIDGE_LOCK(sc); 2172 bridge_delete_member(sc, bif, 1); 2173 BRIDGE_UNLOCK(sc); 2174 return; 2175 } 2176 2177 /* Check if the interface is a span port */ 2178 BRIDGE_LIST_LOCK(); 2179 LIST_FOREACH(sc, &V_bridge_list, sc_list) { 2180 BRIDGE_LOCK(sc); 2181 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) 2182 if (ifp == bif->bif_ifp) { 2183 bridge_delete_span(sc, bif); 2184 break; 2185 } 2186 2187 BRIDGE_UNLOCK(sc); 2188 } 2189 BRIDGE_LIST_UNLOCK(); 2190 } 2191 2192 /* 2193 * bridge_init: 2194 * 2195 * Initialize a bridge interface. 2196 */ 2197 static void 2198 bridge_init(void *xsc) 2199 { 2200 struct bridge_softc *sc = (struct bridge_softc *)xsc; 2201 struct ifnet *ifp = sc->sc_ifp; 2202 2203 if (ifp->if_drv_flags & IFF_DRV_RUNNING) 2204 return; 2205 2206 BRIDGE_LOCK(sc); 2207 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz, 2208 bridge_timer, sc); 2209 2210 ifp->if_drv_flags |= IFF_DRV_RUNNING; 2211 bstp_init(&sc->sc_stp); /* Initialize Spanning Tree */ 2212 2213 BRIDGE_UNLOCK(sc); 2214 } 2215 2216 /* 2217 * bridge_stop: 2218 * 2219 * Stop the bridge interface. 2220 */ 2221 static void 2222 bridge_stop(struct ifnet *ifp, int disable) 2223 { 2224 struct bridge_softc *sc = ifp->if_softc; 2225 2226 BRIDGE_LOCK_ASSERT(sc); 2227 2228 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 2229 return; 2230 2231 BRIDGE_RT_LOCK(sc); 2232 callout_stop(&sc->sc_brcallout); 2233 2234 bstp_stop(&sc->sc_stp); 2235 2236 bridge_rtflush(sc, IFBF_FLUSHDYN); 2237 BRIDGE_RT_UNLOCK(sc); 2238 2239 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 2240 } 2241 2242 /* 2243 * bridge_enqueue: 2244 * 2245 * Enqueue a packet on a bridge member interface. 2246 * 2247 */ 2248 static int 2249 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m, 2250 struct bridge_iflist *bif) 2251 { 2252 int len, err = 0; 2253 short mflags; 2254 struct mbuf *m0; 2255 2256 /* 2257 * Find the bridge member port this packet is being sent on, if the 2258 * caller didn't already provide it. 2259 */ 2260 if (bif == NULL) 2261 bif = bridge_lookup_member_if(sc, dst_ifp); 2262 if (bif == NULL) { 2263 /* Perhaps the interface was removed from the bridge */ 2264 m_freem(m); 2265 return (EINVAL); 2266 } 2267 2268 /* We may be sending a fragment so traverse the mbuf */ 2269 for (; m; m = m0) { 2270 m0 = m->m_nextpkt; 2271 m->m_nextpkt = NULL; 2272 len = m->m_pkthdr.len; 2273 mflags = m->m_flags; 2274 2275 /* 2276 * If VLAN filtering is enabled, and the native VLAN ID of the 2277 * outgoing interface matches the VLAN ID of the frame, remove 2278 * the VLAN header. 2279 */ 2280 if ((bif->bif_flags & IFBIF_VLANFILTER) && 2281 bif->bif_untagged != DOT1Q_VID_NULL && 2282 VLANTAGOF(m) == bif->bif_untagged) { 2283 m->m_flags &= ~M_VLANTAG; 2284 m->m_pkthdr.ether_vtag = 0; 2285 } 2286 2287 /* 2288 * If underlying interface can not do VLAN tag insertion itself 2289 * then attach a packet tag that holds it. 2290 */ 2291 if ((m->m_flags & M_VLANTAG) && 2292 (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) { 2293 m = ether_vlanencap(m, m->m_pkthdr.ether_vtag); 2294 if (m == NULL) { 2295 if_printf(dst_ifp, 2296 "unable to prepend VLAN header\n"); 2297 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1); 2298 continue; 2299 } 2300 m->m_flags &= ~M_VLANTAG; 2301 } 2302 2303 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */ 2304 if ((err = dst_ifp->if_transmit(dst_ifp, m))) { 2305 int n; 2306 2307 for (m = m0, n = 1; m != NULL; m = m0, n++) { 2308 m0 = m->m_nextpkt; 2309 m_freem(m); 2310 } 2311 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n); 2312 break; 2313 } 2314 2315 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1); 2316 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len); 2317 if (mflags & M_MCAST) 2318 if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1); 2319 } 2320 2321 return (err); 2322 } 2323 2324 /* 2325 * bridge_dummynet: 2326 * 2327 * Receive a queued packet from dummynet and pass it on to the output 2328 * interface. 2329 * 2330 * The mbuf has the Ethernet header already attached. 2331 */ 2332 static void 2333 bridge_dummynet(struct mbuf *m, struct ifnet *ifp) 2334 { 2335 struct bridge_iflist *bif = ifp->if_bridge; 2336 struct bridge_softc *sc = NULL; 2337 2338 if (bif) 2339 sc = bif->bif_sc; 2340 2341 /* 2342 * The packet didnt originate from a member interface. This should only 2343 * ever happen if a member interface is removed while packets are 2344 * queued for it. 2345 */ 2346 if (sc == NULL) { 2347 m_freem(m); 2348 return; 2349 } 2350 2351 if (PFIL_HOOKED_OUT_46) { 2352 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0) 2353 return; 2354 if (m == NULL) 2355 return; 2356 } 2357 2358 bridge_enqueue(sc, ifp, m, NULL); 2359 } 2360 2361 /* 2362 * bridge_output: 2363 * 2364 * Send output from a bridge member interface. This 2365 * performs the bridging function for locally originated 2366 * packets. 2367 * 2368 * The mbuf has the Ethernet header already attached. We must 2369 * enqueue or free the mbuf before returning. 2370 */ 2371 static int 2372 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa, 2373 struct rtentry *rt) 2374 { 2375 struct ether_header *eh; 2376 struct bridge_iflist *sbif; 2377 struct ifnet *bifp, *dst_if; 2378 struct bridge_softc *sc; 2379 ether_vlanid_t vlan; 2380 2381 NET_EPOCH_ASSERT(); 2382 2383 if (m->m_len < ETHER_HDR_LEN) { 2384 m = m_pullup(m, ETHER_HDR_LEN); 2385 if (m == NULL) 2386 return (0); 2387 } 2388 2389 sbif = ifp->if_bridge; 2390 sc = sbif->bif_sc; 2391 bifp = sc->sc_ifp; 2392 2393 eh = mtod(m, struct ether_header *); 2394 vlan = VLANTAGOF(m); 2395 2396 /* 2397 * If bridge is down, but the original output interface is up, 2398 * go ahead and send out that interface. Otherwise, the packet 2399 * is dropped below. 2400 */ 2401 if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 2402 dst_if = ifp; 2403 goto sendunicast; 2404 } 2405 2406 /* 2407 * If the packet is a multicast, or we don't know a better way to 2408 * get there, send to all interfaces. 2409 */ 2410 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 2411 dst_if = NULL; 2412 else 2413 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan); 2414 /* Tap any traffic not passing back out the originating interface */ 2415 if (dst_if != ifp) 2416 ETHER_BPF_MTAP(bifp, m); 2417 if (dst_if == NULL) { 2418 struct bridge_iflist *bif; 2419 struct mbuf *mc; 2420 int used = 0; 2421 2422 bridge_span(sc, m); 2423 2424 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 2425 dst_if = bif->bif_ifp; 2426 2427 if (dst_if->if_type == IFT_GIF) 2428 continue; 2429 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) 2430 continue; 2431 2432 /* 2433 * If this is not the original output interface, 2434 * and the interface is participating in spanning 2435 * tree, make sure the port is in a state that 2436 * allows forwarding. 2437 */ 2438 if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) && 2439 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) 2440 continue; 2441 2442 if (CK_LIST_NEXT(bif, bif_next) == NULL) { 2443 used = 1; 2444 mc = m; 2445 } else { 2446 mc = m_dup(m, M_NOWAIT); 2447 if (mc == NULL) { 2448 if_inc_counter(bifp, IFCOUNTER_OERRORS, 1); 2449 continue; 2450 } 2451 } 2452 2453 bridge_enqueue(sc, dst_if, mc, bif); 2454 } 2455 if (used == 0) 2456 m_freem(m); 2457 return (0); 2458 } 2459 2460 sendunicast: 2461 /* 2462 * XXX Spanning tree consideration here? 2463 */ 2464 2465 bridge_span(sc, m); 2466 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) { 2467 m_freem(m); 2468 return (0); 2469 } 2470 2471 bridge_enqueue(sc, dst_if, m, NULL); 2472 return (0); 2473 } 2474 2475 /* 2476 * bridge_transmit: 2477 * 2478 * Do output on a bridge. 2479 * 2480 */ 2481 static int 2482 bridge_transmit(struct ifnet *ifp, struct mbuf *m) 2483 { 2484 struct bridge_softc *sc; 2485 struct ether_header *eh; 2486 struct ifnet *dst_if; 2487 int error = 0; 2488 ether_vlanid_t vlan; 2489 2490 sc = ifp->if_softc; 2491 2492 ETHER_BPF_MTAP(ifp, m); 2493 2494 eh = mtod(m, struct ether_header *); 2495 vlan = VLANTAGOF(m); 2496 2497 if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) && 2498 (dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan)) != NULL) { 2499 error = bridge_enqueue(sc, dst_if, m, NULL); 2500 } else 2501 bridge_broadcast(sc, ifp, m, 0); 2502 2503 return (error); 2504 } 2505 2506 #ifdef ALTQ 2507 static void 2508 bridge_altq_start(if_t ifp) 2509 { 2510 struct ifaltq *ifq = &ifp->if_snd; 2511 struct mbuf *m; 2512 2513 IFQ_LOCK(ifq); 2514 IFQ_DEQUEUE_NOLOCK(ifq, m); 2515 while (m != NULL) { 2516 bridge_transmit(ifp, m); 2517 IFQ_DEQUEUE_NOLOCK(ifq, m); 2518 } 2519 IFQ_UNLOCK(ifq); 2520 } 2521 2522 static int 2523 bridge_altq_transmit(if_t ifp, struct mbuf *m) 2524 { 2525 int err; 2526 2527 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 2528 IFQ_ENQUEUE(&ifp->if_snd, m, err); 2529 if (err == 0) 2530 bridge_altq_start(ifp); 2531 } else 2532 err = bridge_transmit(ifp, m); 2533 2534 return (err); 2535 } 2536 #endif /* ALTQ */ 2537 2538 /* 2539 * The ifp->if_qflush entry point for if_bridge(4) is no-op. 2540 */ 2541 static void 2542 bridge_qflush(struct ifnet *ifp __unused) 2543 { 2544 } 2545 2546 /* 2547 * bridge_forward: 2548 * 2549 * The forwarding function of the bridge. 2550 * 2551 * NOTE: Releases the lock on return. 2552 */ 2553 static void 2554 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif, 2555 struct mbuf *m) 2556 { 2557 struct bridge_iflist *dbif; 2558 struct ifnet *src_if, *dst_if, *ifp; 2559 struct ether_header *eh; 2560 uint8_t *dst; 2561 int error; 2562 ether_vlanid_t vlan; 2563 2564 NET_EPOCH_ASSERT(); 2565 2566 src_if = m->m_pkthdr.rcvif; 2567 ifp = sc->sc_ifp; 2568 vlan = VLANTAGOF(m); 2569 2570 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); 2571 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 2572 2573 if ((sbif->bif_flags & IFBIF_STP) && 2574 sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) 2575 goto drop; 2576 2577 eh = mtod(m, struct ether_header *); 2578 dst = eh->ether_dhost; 2579 2580 /* If the interface is learning, record the address. */ 2581 if (sbif->bif_flags & IFBIF_LEARNING) { 2582 error = bridge_rtupdate(sc, eh->ether_shost, vlan, 2583 sbif, 0, IFBAF_DYNAMIC); 2584 /* 2585 * If the interface has addresses limits then deny any source 2586 * that is not in the cache. 2587 */ 2588 if (error && sbif->bif_addrmax) 2589 goto drop; 2590 } 2591 2592 if ((sbif->bif_flags & IFBIF_STP) != 0 && 2593 sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING) 2594 goto drop; 2595 2596 #ifdef DEV_NETMAP 2597 /* 2598 * Hand the packet to netmap only if it wasn't injected by netmap 2599 * itself. 2600 */ 2601 if ((m->m_flags & M_BRIDGE_INJECT) == 0 && 2602 (if_getcapenable(ifp) & IFCAP_NETMAP) != 0) { 2603 ifp->if_input(ifp, m); 2604 return; 2605 } 2606 m->m_flags &= ~M_BRIDGE_INJECT; 2607 #endif 2608 2609 /* 2610 * At this point, the port either doesn't participate 2611 * in spanning tree or it is in the forwarding state. 2612 */ 2613 2614 /* 2615 * If the packet is unicast, destined for someone on 2616 * "this" side of the bridge, drop it. 2617 */ 2618 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) { 2619 dst_if = bridge_rtlookup(sc, dst, vlan); 2620 if (src_if == dst_if) 2621 goto drop; 2622 } else { 2623 /* 2624 * Check if its a reserved multicast address, any address 2625 * listed in 802.1D section 7.12.6 may not be forwarded by the 2626 * bridge. 2627 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F 2628 */ 2629 if (dst[0] == 0x01 && dst[1] == 0x80 && 2630 dst[2] == 0xc2 && dst[3] == 0x00 && 2631 dst[4] == 0x00 && dst[5] <= 0x0f) 2632 goto drop; 2633 2634 /* ...forward it to all interfaces. */ 2635 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1); 2636 dst_if = NULL; 2637 } 2638 2639 /* 2640 * If we have a destination interface which is a member of our bridge, 2641 * OR this is a unicast packet, push it through the bpf(4) machinery. 2642 * For broadcast or multicast packets, don't bother because it will 2643 * be reinjected into ether_input. We do this before we pass the packets 2644 * through the pfil(9) framework, as it is possible that pfil(9) will 2645 * drop the packet, or possibly modify it, making it difficult to debug 2646 * firewall issues on the bridge. 2647 */ 2648 if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0) 2649 ETHER_BPF_MTAP(ifp, m); 2650 2651 /* run the packet filter */ 2652 if (PFIL_HOOKED_IN_46) { 2653 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0) 2654 return; 2655 if (m == NULL) 2656 return; 2657 } 2658 2659 if (dst_if == NULL) { 2660 bridge_broadcast(sc, src_if, m, 1); 2661 return; 2662 } 2663 2664 /* 2665 * At this point, we're dealing with a unicast frame 2666 * going to a different interface. 2667 */ 2668 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) 2669 goto drop; 2670 2671 dbif = bridge_lookup_member_if(sc, dst_if); 2672 if (dbif == NULL) 2673 /* Not a member of the bridge (anymore?) */ 2674 goto drop; 2675 2676 /* Private segments can not talk to each other */ 2677 if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE) 2678 goto drop; 2679 2680 /* Do VLAN filtering. */ 2681 if (!bridge_vfilter_out(dbif, m)) 2682 goto drop; 2683 2684 if ((dbif->bif_flags & IFBIF_STP) && 2685 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) 2686 goto drop; 2687 2688 if (PFIL_HOOKED_OUT_46) { 2689 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0) 2690 return; 2691 if (m == NULL) 2692 return; 2693 } 2694 2695 bridge_enqueue(sc, dst_if, m, dbif); 2696 return; 2697 2698 drop: 2699 m_freem(m); 2700 } 2701 2702 /* 2703 * bridge_input: 2704 * 2705 * Receive input from a member interface. Queue the packet for 2706 * bridging if it is not for us. 2707 */ 2708 static struct mbuf * 2709 bridge_input(struct ifnet *ifp, struct mbuf *m) 2710 { 2711 struct bridge_softc *sc = NULL; 2712 struct bridge_iflist *bif, *bif2; 2713 struct ifnet *bifp; 2714 struct ether_header *eh; 2715 struct mbuf *mc, *mc2; 2716 ether_vlanid_t vlan; 2717 int error; 2718 2719 NET_EPOCH_ASSERT(); 2720 2721 eh = mtod(m, struct ether_header *); 2722 vlan = VLANTAGOF(m); 2723 2724 bif = ifp->if_bridge; 2725 if (bif) 2726 sc = bif->bif_sc; 2727 2728 if (sc == NULL) { 2729 /* 2730 * This packet originated from the bridge itself, so it must 2731 * have been transmitted by netmap. Derive the "source" 2732 * interface from the source address and drop the packet if the 2733 * source address isn't known. 2734 */ 2735 KASSERT((m->m_flags & M_BRIDGE_INJECT) != 0, 2736 ("%s: ifnet %p missing a bridge softc", __func__, ifp)); 2737 sc = if_getsoftc(ifp); 2738 ifp = bridge_rtlookup(sc, eh->ether_shost, vlan); 2739 if (ifp == NULL) { 2740 if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1); 2741 m_freem(m); 2742 return (NULL); 2743 } 2744 m->m_pkthdr.rcvif = ifp; 2745 } 2746 bifp = sc->sc_ifp; 2747 if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) 2748 return (m); 2749 2750 /* 2751 * Implement support for bridge monitoring. If this flag has been 2752 * set on this interface, discard the packet once we push it through 2753 * the bpf(4) machinery, but before we do, increment the byte and 2754 * packet counters associated with this interface. 2755 */ 2756 if ((bifp->if_flags & IFF_MONITOR) != 0) { 2757 m->m_pkthdr.rcvif = bifp; 2758 ETHER_BPF_MTAP(bifp, m); 2759 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1); 2760 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 2761 m_freem(m); 2762 return (NULL); 2763 } 2764 2765 /* Do VLAN filtering. */ 2766 if (!bridge_vfilter_in(bif, m)) { 2767 if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1); 2768 m_freem(m); 2769 return (NULL); 2770 } 2771 /* bridge_vfilter_in() may add a tag */ 2772 vlan = VLANTAGOF(m); 2773 2774 bridge_span(sc, m); 2775 2776 if (m->m_flags & (M_BCAST|M_MCAST)) { 2777 /* Tap off 802.1D packets; they do not get forwarded. */ 2778 if (memcmp(eh->ether_dhost, bstp_etheraddr, 2779 ETHER_ADDR_LEN) == 0) { 2780 bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */ 2781 return (NULL); 2782 } 2783 2784 if ((bif->bif_flags & IFBIF_STP) && 2785 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) { 2786 return (m); 2787 } 2788 2789 /* 2790 * Make a deep copy of the packet and enqueue the copy 2791 * for bridge processing; return the original packet for 2792 * local processing. 2793 */ 2794 mc = m_dup(m, M_NOWAIT); 2795 if (mc == NULL) { 2796 return (m); 2797 } 2798 2799 /* Perform the bridge forwarding function with the copy. */ 2800 bridge_forward(sc, bif, mc); 2801 2802 #ifdef DEV_NETMAP 2803 /* 2804 * If netmap is enabled and has not already seen this packet, 2805 * then it will be consumed by bridge_forward(). 2806 */ 2807 if ((if_getcapenable(bifp) & IFCAP_NETMAP) != 0 && 2808 (m->m_flags & M_BRIDGE_INJECT) == 0) { 2809 m_freem(m); 2810 return (NULL); 2811 } 2812 #endif 2813 2814 /* 2815 * Reinject the mbuf as arriving on the bridge so we have a 2816 * chance at claiming multicast packets. We can not loop back 2817 * here from ether_input as a bridge is never a member of a 2818 * bridge. 2819 */ 2820 KASSERT(bifp->if_bridge == NULL, 2821 ("loop created in bridge_input")); 2822 mc2 = m_dup(m, M_NOWAIT); 2823 if (mc2 != NULL) { 2824 /* Keep the layer3 header aligned */ 2825 int i = min(mc2->m_pkthdr.len, max_protohdr); 2826 mc2 = m_copyup(mc2, i, ETHER_ALIGN); 2827 } 2828 if (mc2 != NULL) { 2829 mc2->m_pkthdr.rcvif = bifp; 2830 mc2->m_flags &= ~M_BRIDGE_INJECT; 2831 sc->sc_if_input(bifp, mc2); 2832 } 2833 2834 /* Return the original packet for local processing. */ 2835 return (m); 2836 } 2837 2838 if ((bif->bif_flags & IFBIF_STP) && 2839 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) { 2840 return (m); 2841 } 2842 2843 #if defined(INET) || defined(INET6) 2844 #define CARP_CHECK_WE_ARE_DST(iface) \ 2845 ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_dhost)) 2846 #define CARP_CHECK_WE_ARE_SRC(iface) \ 2847 ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_shost)) 2848 #else 2849 #define CARP_CHECK_WE_ARE_DST(iface) false 2850 #define CARP_CHECK_WE_ARE_SRC(iface) false 2851 #endif 2852 2853 #ifdef DEV_NETMAP 2854 #define GRAB_FOR_NETMAP(ifp, m) do { \ 2855 if ((if_getcapenable(ifp) & IFCAP_NETMAP) != 0 && \ 2856 ((m)->m_flags & M_BRIDGE_INJECT) == 0) { \ 2857 (ifp)->if_input(ifp, m); \ 2858 return (NULL); \ 2859 } \ 2860 } while (0) 2861 #else 2862 #define GRAB_FOR_NETMAP(ifp, m) 2863 #endif 2864 2865 #define GRAB_OUR_PACKETS(iface) \ 2866 if ((iface)->if_type == IFT_GIF) \ 2867 continue; \ 2868 /* It is destined for us. */ \ 2869 if (memcmp(IF_LLADDR(iface), eh->ether_dhost, ETHER_ADDR_LEN) == 0 || \ 2870 CARP_CHECK_WE_ARE_DST(iface)) { \ 2871 if (bif->bif_flags & IFBIF_LEARNING) { \ 2872 error = bridge_rtupdate(sc, eh->ether_shost, \ 2873 vlan, bif, 0, IFBAF_DYNAMIC); \ 2874 if (error && bif->bif_addrmax) { \ 2875 m_freem(m); \ 2876 return (NULL); \ 2877 } \ 2878 } \ 2879 m->m_pkthdr.rcvif = iface; \ 2880 if ((iface) == ifp) { \ 2881 /* Skip bridge processing... src == dest */ \ 2882 return (m); \ 2883 } \ 2884 /* It's passing over or to the bridge, locally. */ \ 2885 ETHER_BPF_MTAP(bifp, m); \ 2886 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1); \ 2887 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);\ 2888 /* Hand the packet over to netmap if necessary. */ \ 2889 GRAB_FOR_NETMAP(bifp, m); \ 2890 /* Filter on the physical interface. */ \ 2891 if (V_pfil_local_phys && PFIL_HOOKED_IN_46) { \ 2892 if (bridge_pfil(&m, NULL, ifp, \ 2893 PFIL_IN) != 0 || m == NULL) { \ 2894 return (NULL); \ 2895 } \ 2896 } \ 2897 if ((iface) != bifp) \ 2898 ETHER_BPF_MTAP(iface, m); \ 2899 /* Pass tagged packets to if_vlan, if it's loaded */ \ 2900 if (VLANTAGOF(m) != 0) { \ 2901 if (bifp->if_vlantrunk == NULL) { \ 2902 m_freem(m); \ 2903 return (NULL); \ 2904 } \ 2905 (*vlan_input_p)(bifp, m); \ 2906 return (NULL); \ 2907 } \ 2908 return (m); \ 2909 } \ 2910 \ 2911 /* We just received a packet that we sent out. */ \ 2912 if (memcmp(IF_LLADDR(iface), eh->ether_shost, ETHER_ADDR_LEN) == 0 || \ 2913 CARP_CHECK_WE_ARE_SRC(iface)) { \ 2914 m_freem(m); \ 2915 return (NULL); \ 2916 } 2917 2918 /* 2919 * Unicast. Make sure it's not for the bridge. 2920 */ 2921 do { GRAB_OUR_PACKETS(bifp) } while (0); 2922 2923 /* 2924 * Check the interface the packet arrived on. For tagged frames, 2925 * we need to do this even if member_ifaddrs is disabled because 2926 * vlan(4) might need to handle the traffic. 2927 */ 2928 if (V_member_ifaddrs || (vlan && ifp->if_vlantrunk)) 2929 do { GRAB_OUR_PACKETS(ifp) } while (0); 2930 2931 /* 2932 * We only need to check other members interface if member_ifaddrs 2933 * is enabled; otherwise we should have never traffic destined for 2934 * a member's lladdr. 2935 */ 2936 if (V_member_ifaddrs) { 2937 CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) { 2938 GRAB_OUR_PACKETS(bif2->bif_ifp) 2939 } 2940 } 2941 2942 #undef CARP_CHECK_WE_ARE_DST 2943 #undef CARP_CHECK_WE_ARE_SRC 2944 #undef GRAB_FOR_NETMAP 2945 #undef GRAB_OUR_PACKETS 2946 2947 /* Perform the bridge forwarding function. */ 2948 bridge_forward(sc, bif, m); 2949 2950 return (NULL); 2951 } 2952 2953 /* 2954 * Inject a packet back into the host ethernet stack. This will generally only 2955 * be used by netmap when an application writes to the host TX ring. The 2956 * M_BRIDGE_INJECT flag ensures that the packet is re-routed to the bridge 2957 * interface after ethernet processing. 2958 */ 2959 static void 2960 bridge_inject(struct ifnet *ifp, struct mbuf *m) 2961 { 2962 struct bridge_softc *sc; 2963 2964 if (ifp->if_type == IFT_L2VLAN) { 2965 /* 2966 * vlan(4) gives us the vlan ifnet, so we need to get the 2967 * bridge softc to get a pointer to ether_input to send the 2968 * packet to. 2969 */ 2970 struct ifnet *bifp = NULL; 2971 2972 if (vlan_trunkdev_p == NULL) { 2973 m_freem(m); 2974 return; 2975 } 2976 2977 bifp = vlan_trunkdev_p(ifp); 2978 if (bifp == NULL) { 2979 m_freem(m); 2980 return; 2981 } 2982 2983 sc = if_getsoftc(bifp); 2984 sc->sc_if_input(ifp, m); 2985 return; 2986 } 2987 2988 KASSERT((if_getcapenable(ifp) & IFCAP_NETMAP) != 0, 2989 ("%s: iface %s is not running in netmap mode", 2990 __func__, if_name(ifp))); 2991 KASSERT((m->m_flags & M_BRIDGE_INJECT) == 0, 2992 ("%s: mbuf %p has M_BRIDGE_INJECT set", __func__, m)); 2993 2994 m->m_flags |= M_BRIDGE_INJECT; 2995 sc = if_getsoftc(ifp); 2996 sc->sc_if_input(ifp, m); 2997 } 2998 2999 /* 3000 * bridge_broadcast: 3001 * 3002 * Send a frame to all interfaces that are members of 3003 * the bridge, except for the one on which the packet 3004 * arrived. 3005 * 3006 * NOTE: Releases the lock on return. 3007 */ 3008 static void 3009 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if, 3010 struct mbuf *m, int runfilt) 3011 { 3012 struct bridge_iflist *dbif, *sbif; 3013 struct mbuf *mc; 3014 struct ifnet *dst_if; 3015 int used = 0, i; 3016 3017 NET_EPOCH_ASSERT(); 3018 3019 sbif = bridge_lookup_member_if(sc, src_if); 3020 3021 /* Filter on the bridge interface before broadcasting */ 3022 if (runfilt && PFIL_HOOKED_OUT_46) { 3023 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0) 3024 return; 3025 if (m == NULL) 3026 return; 3027 } 3028 3029 CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) { 3030 dst_if = dbif->bif_ifp; 3031 if (dst_if == src_if) 3032 continue; 3033 3034 /* Private segments can not talk to each other */ 3035 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)) 3036 continue; 3037 3038 /* Do VLAN filtering. */ 3039 if (!bridge_vfilter_out(dbif, m)) 3040 continue; 3041 3042 if ((dbif->bif_flags & IFBIF_STP) && 3043 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) 3044 continue; 3045 3046 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 && 3047 (m->m_flags & (M_BCAST|M_MCAST)) == 0) 3048 continue; 3049 3050 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) 3051 continue; 3052 3053 if (CK_LIST_NEXT(dbif, bif_next) == NULL) { 3054 mc = m; 3055 used = 1; 3056 } else { 3057 mc = m_dup(m, M_NOWAIT); 3058 if (mc == NULL) { 3059 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1); 3060 continue; 3061 } 3062 } 3063 3064 /* 3065 * Filter on the output interface. Pass a NULL bridge interface 3066 * pointer so we do not redundantly filter on the bridge for 3067 * each interface we broadcast on. 3068 */ 3069 if (runfilt && PFIL_HOOKED_OUT_46) { 3070 if (used == 0) { 3071 /* Keep the layer3 header aligned */ 3072 i = min(mc->m_pkthdr.len, max_protohdr); 3073 mc = m_copyup(mc, i, ETHER_ALIGN); 3074 if (mc == NULL) { 3075 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1); 3076 continue; 3077 } 3078 } 3079 if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0) 3080 continue; 3081 if (mc == NULL) 3082 continue; 3083 } 3084 3085 bridge_enqueue(sc, dst_if, mc, dbif); 3086 } 3087 if (used == 0) 3088 m_freem(m); 3089 } 3090 3091 /* 3092 * bridge_span: 3093 * 3094 * Duplicate a packet out one or more interfaces that are in span mode, 3095 * the original mbuf is unmodified. 3096 */ 3097 static void 3098 bridge_span(struct bridge_softc *sc, struct mbuf *m) 3099 { 3100 struct bridge_iflist *bif; 3101 struct ifnet *dst_if; 3102 struct mbuf *mc; 3103 3104 NET_EPOCH_ASSERT(); 3105 3106 if (CK_LIST_EMPTY(&sc->sc_spanlist)) 3107 return; 3108 3109 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) { 3110 dst_if = bif->bif_ifp; 3111 3112 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) 3113 continue; 3114 3115 mc = m_dup(m, M_NOWAIT); 3116 if (mc == NULL) { 3117 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1); 3118 continue; 3119 } 3120 3121 bridge_enqueue(sc, dst_if, mc, bif); 3122 } 3123 } 3124 3125 /* 3126 * Incoming VLAN filtering. Given a frame and the member interface it was 3127 * received on, decide whether the port configuration allows it. 3128 */ 3129 static bool 3130 bridge_vfilter_in(const struct bridge_iflist *sbif, struct mbuf *m) 3131 { 3132 ether_vlanid_t vlan; 3133 3134 vlan = VLANTAGOF(m); 3135 /* Make sure the vlan id is reasonable. */ 3136 if (vlan > DOT1Q_VID_MAX) 3137 return (false); 3138 3139 /* If VLAN filtering isn't enabled, pass everything. */ 3140 if ((sbif->bif_flags & IFBIF_VLANFILTER) == 0) 3141 return (true); 3142 3143 if (vlan == DOT1Q_VID_NULL) { 3144 /* 3145 * The frame doesn't have a tag. If the interface does not 3146 * have an untagged vlan configured, drop the frame. 3147 */ 3148 if (sbif->bif_untagged == DOT1Q_VID_NULL) 3149 return (false); 3150 3151 /* 3152 * Otherwise, insert a new tag based on the interface's 3153 * untagged vlan id. 3154 */ 3155 m->m_pkthdr.ether_vtag = sbif->bif_untagged; 3156 m->m_flags |= M_VLANTAG; 3157 } else { 3158 /* 3159 * The frame has a tag, so check it matches the interface's 3160 * vlan access list. We explicitly do not accept tagged 3161 * frames for the untagged vlan id here (unless it's also 3162 * in the access list). 3163 */ 3164 if (!BRVLAN_TEST(&sbif->bif_vlan_set, vlan)) 3165 return (false); 3166 } 3167 3168 /* Accept the frame. */ 3169 return (true); 3170 } 3171 3172 /* 3173 * Outgoing VLAN filtering. Given a frame, its vlan, and the member interface 3174 * we intend to send it to, decide whether the port configuration allows it to 3175 * be sent. 3176 */ 3177 static bool 3178 bridge_vfilter_out(const struct bridge_iflist *dbif, const struct mbuf *m) 3179 { 3180 struct ether_header *eh; 3181 ether_vlanid_t vlan; 3182 3183 NET_EPOCH_ASSERT(); 3184 3185 /* If VLAN filtering isn't enabled, pass everything. */ 3186 if ((dbif->bif_flags & IFBIF_VLANFILTER) == 0) 3187 return (true); 3188 3189 vlan = VLANTAGOF(m); 3190 3191 /* 3192 * Always allow untagged 802.1D STP frames, even if they would 3193 * otherwise be dropped. This is required for STP to work on 3194 * a filtering bridge. 3195 * 3196 * Tagged STP (Cisco PVST+) is a non-standard extension, so 3197 * handle those frames via the normal filtering path. 3198 */ 3199 eh = mtod(m, struct ether_header *); 3200 if (vlan == DOT1Q_VID_NULL && 3201 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) 3202 return (true); 3203 3204 /* 3205 * If the frame wasn't assigned to a vlan at ingress, drop it. 3206 * We can't forward these frames to filtering ports because we 3207 * don't know what VLAN they're supposed to be in. 3208 */ 3209 if (vlan == DOT1Q_VID_NULL) 3210 return (false); 3211 3212 /* 3213 * If the frame's vlan matches the interfaces's untagged vlan, 3214 * allow it. 3215 */ 3216 if (vlan == dbif->bif_untagged) 3217 return (true); 3218 3219 /* 3220 * If the frame's vlan is on the interface's tagged access list, 3221 * allow it. 3222 */ 3223 if (BRVLAN_TEST(&dbif->bif_vlan_set, vlan)) 3224 return (true); 3225 3226 /* The frame was not permitted, so drop it. */ 3227 return (false); 3228 } 3229 3230 /* 3231 * bridge_rtupdate: 3232 * 3233 * Add a bridge routing entry. 3234 */ 3235 static int 3236 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, 3237 ether_vlanid_t vlan, struct bridge_iflist *bif, 3238 int setflags, uint8_t flags) 3239 { 3240 struct bridge_rtnode *brt; 3241 struct bridge_iflist *obif; 3242 int error; 3243 3244 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc); 3245 3246 /* Check the source address is valid and not multicast. */ 3247 if (ETHER_IS_MULTICAST(dst) || 3248 (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 && 3249 dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0) 3250 return (EINVAL); 3251 3252 /* 3253 * A route for this destination might already exist. If so, 3254 * update it, otherwise create a new one. 3255 */ 3256 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) { 3257 BRIDGE_RT_LOCK(sc); 3258 3259 /* Check again, now that we have the lock. There could have 3260 * been a race and we only want to insert this once. */ 3261 if (bridge_rtnode_lookup(sc, dst, vlan) != NULL) { 3262 BRIDGE_RT_UNLOCK(sc); 3263 return (0); 3264 } 3265 3266 if (sc->sc_brtcnt >= sc->sc_brtmax) { 3267 sc->sc_brtexceeded++; 3268 BRIDGE_RT_UNLOCK(sc); 3269 return (ENOSPC); 3270 } 3271 /* Check per interface address limits (if enabled) */ 3272 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) { 3273 bif->bif_addrexceeded++; 3274 BRIDGE_RT_UNLOCK(sc); 3275 return (ENOSPC); 3276 } 3277 3278 /* 3279 * Allocate a new bridge forwarding node, and 3280 * initialize the expiration time and Ethernet 3281 * address. 3282 */ 3283 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO); 3284 if (brt == NULL) { 3285 BRIDGE_RT_UNLOCK(sc); 3286 return (ENOMEM); 3287 } 3288 brt->brt_vnet = curvnet; 3289 3290 if (bif->bif_flags & IFBIF_STICKY) 3291 brt->brt_flags = IFBAF_STICKY; 3292 else 3293 brt->brt_flags = IFBAF_DYNAMIC; 3294 3295 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN); 3296 brt->brt_vlan = vlan; 3297 3298 brt->brt_dst = bif; 3299 if ((error = bridge_rtnode_insert(sc, brt)) != 0) { 3300 uma_zfree(V_bridge_rtnode_zone, brt); 3301 BRIDGE_RT_UNLOCK(sc); 3302 return (error); 3303 } 3304 bif->bif_addrcnt++; 3305 3306 BRIDGE_RT_UNLOCK(sc); 3307 } 3308 3309 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC && 3310 (obif = brt->brt_dst) != bif) { 3311 MPASS(obif != NULL); 3312 3313 BRIDGE_RT_LOCK(sc); 3314 brt->brt_dst->bif_addrcnt--; 3315 brt->brt_dst = bif; 3316 brt->brt_dst->bif_addrcnt++; 3317 BRIDGE_RT_UNLOCK(sc); 3318 3319 if (V_log_mac_flap && 3320 ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) { 3321 log(LOG_NOTICE, 3322 "%s: mac address %6D vlan %d moved from %s to %s\n", 3323 sc->sc_ifp->if_xname, 3324 &brt->brt_addr[0], ":", 3325 brt->brt_vlan, 3326 obif->bif_ifp->if_xname, 3327 bif->bif_ifp->if_xname); 3328 } 3329 } 3330 3331 if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 3332 brt->brt_expire = time_uptime + sc->sc_brttimeout; 3333 if (setflags) 3334 brt->brt_flags = flags; 3335 3336 return (0); 3337 } 3338 3339 /* 3340 * bridge_rtlookup: 3341 * 3342 * Lookup the destination interface for an address. 3343 */ 3344 static struct ifnet * 3345 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, 3346 ether_vlanid_t vlan) 3347 { 3348 struct bridge_rtnode *brt; 3349 3350 NET_EPOCH_ASSERT(); 3351 3352 if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL) 3353 return (NULL); 3354 3355 return (brt->brt_ifp); 3356 } 3357 3358 /* 3359 * bridge_rttrim: 3360 * 3361 * Trim the routine table so that we have a number 3362 * of routing entries less than or equal to the 3363 * maximum number. 3364 */ 3365 static void 3366 bridge_rttrim(struct bridge_softc *sc) 3367 { 3368 struct bridge_rtnode *brt, *nbrt; 3369 3370 NET_EPOCH_ASSERT(); 3371 BRIDGE_RT_LOCK_ASSERT(sc); 3372 3373 /* Make sure we actually need to do this. */ 3374 if (sc->sc_brtcnt <= sc->sc_brtmax) 3375 return; 3376 3377 /* Force an aging cycle; this might trim enough addresses. */ 3378 bridge_rtage(sc); 3379 if (sc->sc_brtcnt <= sc->sc_brtmax) 3380 return; 3381 3382 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) { 3383 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 3384 bridge_rtnode_destroy(sc, brt); 3385 if (sc->sc_brtcnt <= sc->sc_brtmax) 3386 return; 3387 } 3388 } 3389 } 3390 3391 /* 3392 * bridge_timer: 3393 * 3394 * Aging timer for the bridge. 3395 */ 3396 static void 3397 bridge_timer(void *arg) 3398 { 3399 struct bridge_softc *sc = arg; 3400 3401 BRIDGE_RT_LOCK_ASSERT(sc); 3402 3403 /* Destruction of rtnodes requires a proper vnet context */ 3404 CURVNET_SET(sc->sc_ifp->if_vnet); 3405 bridge_rtage(sc); 3406 3407 if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) 3408 callout_reset(&sc->sc_brcallout, 3409 bridge_rtable_prune_period * hz, bridge_timer, sc); 3410 CURVNET_RESTORE(); 3411 } 3412 3413 /* 3414 * bridge_rtage: 3415 * 3416 * Perform an aging cycle. 3417 */ 3418 static void 3419 bridge_rtage(struct bridge_softc *sc) 3420 { 3421 struct bridge_rtnode *brt, *nbrt; 3422 3423 BRIDGE_RT_LOCK_ASSERT(sc); 3424 3425 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) { 3426 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { 3427 if (time_uptime >= brt->brt_expire) 3428 bridge_rtnode_destroy(sc, brt); 3429 } 3430 } 3431 } 3432 3433 /* 3434 * bridge_rtflush: 3435 * 3436 * Remove all dynamic addresses from the bridge. 3437 */ 3438 static void 3439 bridge_rtflush(struct bridge_softc *sc, int full) 3440 { 3441 struct bridge_rtnode *brt, *nbrt; 3442 3443 BRIDGE_RT_LOCK_ASSERT(sc); 3444 3445 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) { 3446 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 3447 bridge_rtnode_destroy(sc, brt); 3448 } 3449 } 3450 3451 /* 3452 * bridge_rtdaddr: 3453 * 3454 * Remove an address from the table. 3455 */ 3456 static int 3457 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, 3458 ether_vlanid_t vlan) 3459 { 3460 struct bridge_rtnode *brt; 3461 int found = 0; 3462 3463 BRIDGE_RT_LOCK(sc); 3464 3465 /* 3466 * If vlan is DOT1Q_VID_RSVD_IMPL then we want to delete for all vlans 3467 * so the lookup may return more than one. 3468 */ 3469 while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) { 3470 bridge_rtnode_destroy(sc, brt); 3471 found = 1; 3472 } 3473 3474 BRIDGE_RT_UNLOCK(sc); 3475 3476 return (found ? 0 : ENOENT); 3477 } 3478 3479 /* 3480 * bridge_rtdelete: 3481 * 3482 * Delete routes to a speicifc member interface. 3483 */ 3484 static void 3485 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full) 3486 { 3487 struct bridge_rtnode *brt, *nbrt; 3488 3489 BRIDGE_RT_LOCK_ASSERT(sc); 3490 3491 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) { 3492 if (brt->brt_ifp == ifp && (full || 3493 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) 3494 bridge_rtnode_destroy(sc, brt); 3495 } 3496 } 3497 3498 /* 3499 * bridge_rtable_init: 3500 * 3501 * Initialize the route table for this bridge. 3502 */ 3503 static void 3504 bridge_rtable_init(struct bridge_softc *sc) 3505 { 3506 int i; 3507 3508 sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE, 3509 M_DEVBUF, M_WAITOK); 3510 3511 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++) 3512 CK_LIST_INIT(&sc->sc_rthash[i]); 3513 3514 sc->sc_rthash_key = arc4random(); 3515 CK_LIST_INIT(&sc->sc_rtlist); 3516 } 3517 3518 /* 3519 * bridge_rtable_fini: 3520 * 3521 * Deconstruct the route table for this bridge. 3522 */ 3523 static void 3524 bridge_rtable_fini(struct bridge_softc *sc) 3525 { 3526 3527 KASSERT(sc->sc_brtcnt == 0, 3528 ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt)); 3529 free(sc->sc_rthash, M_DEVBUF); 3530 } 3531 3532 /* 3533 * The following hash function is adapted from "Hash Functions" by Bob Jenkins 3534 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997). 3535 */ 3536 #define mix(a, b, c) \ 3537 do { \ 3538 a -= b; a -= c; a ^= (c >> 13); \ 3539 b -= c; b -= a; b ^= (a << 8); \ 3540 c -= a; c -= b; c ^= (b >> 13); \ 3541 a -= b; a -= c; a ^= (c >> 12); \ 3542 b -= c; b -= a; b ^= (a << 16); \ 3543 c -= a; c -= b; c ^= (b >> 5); \ 3544 a -= b; a -= c; a ^= (c >> 3); \ 3545 b -= c; b -= a; b ^= (a << 10); \ 3546 c -= a; c -= b; c ^= (b >> 15); \ 3547 } while (/*CONSTCOND*/0) 3548 3549 static __inline uint32_t 3550 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr) 3551 { 3552 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key; 3553 3554 b += addr[5] << 8; 3555 b += addr[4]; 3556 a += addr[3] << 24; 3557 a += addr[2] << 16; 3558 a += addr[1] << 8; 3559 a += addr[0]; 3560 3561 mix(a, b, c); 3562 3563 return (c & BRIDGE_RTHASH_MASK); 3564 } 3565 3566 #undef mix 3567 3568 static int 3569 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b) 3570 { 3571 int i, d; 3572 3573 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) { 3574 d = ((int)a[i]) - ((int)b[i]); 3575 } 3576 3577 return (d); 3578 } 3579 3580 /* 3581 * bridge_rtnode_lookup: 3582 * 3583 * Look up a bridge route node for the specified destination. Compare the 3584 * vlan id or if zero then just return the first match. 3585 */ 3586 static struct bridge_rtnode * 3587 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, 3588 ether_vlanid_t vlan) 3589 { 3590 struct bridge_rtnode *brt; 3591 uint32_t hash; 3592 int dir; 3593 3594 BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc); 3595 3596 hash = bridge_rthash(sc, addr); 3597 CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) { 3598 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr); 3599 if (dir == 0 && (brt->brt_vlan == vlan || vlan == DOT1Q_VID_RSVD_IMPL)) 3600 return (brt); 3601 if (dir > 0) 3602 return (NULL); 3603 } 3604 3605 return (NULL); 3606 } 3607 3608 /* 3609 * bridge_rtnode_insert: 3610 * 3611 * Insert the specified bridge node into the route table. We 3612 * assume the entry is not already in the table. 3613 */ 3614 static int 3615 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt) 3616 { 3617 struct bridge_rtnode *lbrt; 3618 uint32_t hash; 3619 int dir; 3620 3621 BRIDGE_RT_LOCK_ASSERT(sc); 3622 3623 hash = bridge_rthash(sc, brt->brt_addr); 3624 3625 lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]); 3626 if (lbrt == NULL) { 3627 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash); 3628 goto out; 3629 } 3630 3631 do { 3632 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr); 3633 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan) 3634 return (EEXIST); 3635 if (dir > 0) { 3636 CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash); 3637 goto out; 3638 } 3639 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) { 3640 CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash); 3641 goto out; 3642 } 3643 lbrt = CK_LIST_NEXT(lbrt, brt_hash); 3644 } while (lbrt != NULL); 3645 3646 #ifdef DIAGNOSTIC 3647 panic("bridge_rtnode_insert: impossible"); 3648 #endif 3649 3650 out: 3651 CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list); 3652 sc->sc_brtcnt++; 3653 3654 return (0); 3655 } 3656 3657 static void 3658 bridge_rtnode_destroy_cb(struct epoch_context *ctx) 3659 { 3660 struct bridge_rtnode *brt; 3661 3662 brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx); 3663 3664 CURVNET_SET(brt->brt_vnet); 3665 uma_zfree(V_bridge_rtnode_zone, brt); 3666 CURVNET_RESTORE(); 3667 } 3668 3669 /* 3670 * bridge_rtnode_destroy: 3671 * 3672 * Destroy a bridge rtnode. 3673 */ 3674 static void 3675 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt) 3676 { 3677 BRIDGE_RT_LOCK_ASSERT(sc); 3678 3679 CK_LIST_REMOVE(brt, brt_hash); 3680 3681 CK_LIST_REMOVE(brt, brt_list); 3682 sc->sc_brtcnt--; 3683 brt->brt_dst->bif_addrcnt--; 3684 3685 NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx); 3686 } 3687 3688 /* 3689 * bridge_rtable_expire: 3690 * 3691 * Set the expiry time for all routes on an interface. 3692 */ 3693 static void 3694 bridge_rtable_expire(struct ifnet *ifp, int age) 3695 { 3696 struct bridge_iflist *bif = NULL; 3697 struct bridge_softc *sc = NULL; 3698 struct bridge_rtnode *brt; 3699 3700 CURVNET_SET(ifp->if_vnet); 3701 3702 bif = ifp->if_bridge; 3703 if (bif) 3704 sc = bif->bif_sc; 3705 MPASS(sc != NULL); 3706 BRIDGE_RT_LOCK(sc); 3707 3708 /* 3709 * If the age is zero then flush, otherwise set all the expiry times to 3710 * age for the interface 3711 */ 3712 if (age == 0) 3713 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN); 3714 else { 3715 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) { 3716 /* Cap the expiry time to 'age' */ 3717 if (brt->brt_ifp == ifp && 3718 brt->brt_expire > time_uptime + age && 3719 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) 3720 brt->brt_expire = time_uptime + age; 3721 } 3722 } 3723 BRIDGE_RT_UNLOCK(sc); 3724 CURVNET_RESTORE(); 3725 } 3726 3727 /* 3728 * bridge_state_change: 3729 * 3730 * Callback from the bridgestp code when a port changes states. 3731 */ 3732 static void 3733 bridge_state_change(struct ifnet *ifp, int state) 3734 { 3735 struct bridge_iflist *bif = ifp->if_bridge; 3736 struct bridge_softc *sc = bif->bif_sc; 3737 static const char *stpstates[] = { 3738 "disabled", 3739 "listening", 3740 "learning", 3741 "forwarding", 3742 "blocking", 3743 "discarding" 3744 }; 3745 3746 CURVNET_SET(ifp->if_vnet); 3747 if (V_log_stp) 3748 log(LOG_NOTICE, "%s: state changed to %s on %s\n", 3749 sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname); 3750 CURVNET_RESTORE(); 3751 } 3752 3753 /* 3754 * Send bridge packets through pfil if they are one of the types pfil can deal 3755 * with, or if they are ARP or REVARP. (pfil will pass ARP and REVARP without 3756 * question.) If *bifp or *ifp are NULL then packet filtering is skipped for 3757 * that interface. 3758 */ 3759 static int 3760 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir) 3761 { 3762 int snap, error, i; 3763 struct ether_header *eh1, eh2; 3764 struct llc llc1; 3765 u_int16_t ether_type; 3766 pfil_return_t rv; 3767 #ifdef INET 3768 struct ip *ip = NULL; 3769 int hlen = 0; 3770 #endif 3771 3772 snap = 0; 3773 error = -1; /* Default error if not error == 0 */ 3774 3775 #if 0 3776 /* we may return with the IP fields swapped, ensure its not shared */ 3777 KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__)); 3778 #endif 3779 3780 if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0) 3781 return (0); /* filtering is disabled */ 3782 3783 i = min((*mp)->m_pkthdr.len, max_protohdr); 3784 if ((*mp)->m_len < i) { 3785 *mp = m_pullup(*mp, i); 3786 if (*mp == NULL) { 3787 printf("%s: m_pullup failed\n", __func__); 3788 return (-1); 3789 } 3790 } 3791 3792 eh1 = mtod(*mp, struct ether_header *); 3793 ether_type = ntohs(eh1->ether_type); 3794 3795 /* 3796 * Check for SNAP/LLC. 3797 */ 3798 if (ether_type < ETHERMTU) { 3799 struct llc *llc2 = (struct llc *)(eh1 + 1); 3800 3801 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 && 3802 llc2->llc_dsap == LLC_SNAP_LSAP && 3803 llc2->llc_ssap == LLC_SNAP_LSAP && 3804 llc2->llc_control == LLC_UI) { 3805 ether_type = htons(llc2->llc_un.type_snap.ether_type); 3806 snap = 1; 3807 } 3808 } 3809 3810 /* 3811 * If we're trying to filter bridge traffic, only look at traffic for 3812 * protocols available in the kernel (IPv4 and/or IPv6) to avoid 3813 * passing traffic for an unsupported protocol to the filter. This is 3814 * lame since if we really wanted, say, an AppleTalk filter, we are 3815 * hosed, but of course we don't have an AppleTalk filter to begin 3816 * with. (Note that since pfil doesn't understand ARP it will pass 3817 * *ALL* ARP traffic.) 3818 */ 3819 switch (ether_type) { 3820 #ifdef INET 3821 case ETHERTYPE_ARP: 3822 case ETHERTYPE_REVARP: 3823 if (V_pfil_ipfw_arp == 0) 3824 return (0); /* Automatically pass */ 3825 3826 /* FALLTHROUGH */ 3827 case ETHERTYPE_IP: 3828 #endif 3829 #ifdef INET6 3830 case ETHERTYPE_IPV6: 3831 #endif /* INET6 */ 3832 break; 3833 3834 default: 3835 /* 3836 * We get here if the packet isn't from a supported 3837 * protocol. Check to see if the user wants to pass 3838 * non-IP packets, these will not be checked by pfil(9) 3839 * and passed unconditionally so the default is to 3840 * drop. 3841 */ 3842 if (V_pfil_onlyip) 3843 goto bad; 3844 } 3845 3846 /* Run the packet through pfil before stripping link headers */ 3847 if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 && 3848 dir == PFIL_OUT && ifp != NULL) { 3849 switch (pfil_mbuf_out(V_link_pfil_head, mp, ifp, NULL)) { 3850 case PFIL_DROPPED: 3851 return (EACCES); 3852 case PFIL_CONSUMED: 3853 return (0); 3854 } 3855 } 3856 3857 /* Strip off the Ethernet header and keep a copy. */ 3858 m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2); 3859 m_adj(*mp, ETHER_HDR_LEN); 3860 3861 /* Strip off snap header, if present */ 3862 if (snap) { 3863 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1); 3864 m_adj(*mp, sizeof(struct llc)); 3865 } 3866 3867 /* 3868 * Check the IP header for alignment and errors 3869 */ 3870 if (dir == PFIL_IN) { 3871 switch (ether_type) { 3872 #ifdef INET 3873 case ETHERTYPE_IP: 3874 error = bridge_ip_checkbasic(mp); 3875 break; 3876 #endif 3877 #ifdef INET6 3878 case ETHERTYPE_IPV6: 3879 error = bridge_ip6_checkbasic(mp); 3880 break; 3881 #endif /* INET6 */ 3882 default: 3883 error = 0; 3884 } 3885 if (error) 3886 goto bad; 3887 } 3888 3889 error = 0; 3890 3891 /* 3892 * Run the packet through pfil 3893 */ 3894 rv = PFIL_PASS; 3895 switch (ether_type) { 3896 #ifdef INET 3897 case ETHERTYPE_IP: 3898 /* 3899 * Run pfil on the member interface and the bridge, both can 3900 * be skipped by clearing pfil_member or pfil_bridge. 3901 * 3902 * Keep the order: 3903 * in_if -> bridge_if -> out_if 3904 */ 3905 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv = 3906 pfil_mbuf_out(V_inet_pfil_head, mp, bifp, NULL)) != 3907 PFIL_PASS) 3908 break; 3909 3910 if (V_pfil_member && ifp != NULL) { 3911 rv = (dir == PFIL_OUT) ? 3912 pfil_mbuf_out(V_inet_pfil_head, mp, ifp, NULL) : 3913 pfil_mbuf_in(V_inet_pfil_head, mp, ifp, NULL); 3914 if (rv != PFIL_PASS) 3915 break; 3916 } 3917 3918 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv = 3919 pfil_mbuf_in(V_inet_pfil_head, mp, bifp, NULL)) != 3920 PFIL_PASS) 3921 break; 3922 3923 /* check if we need to fragment the packet */ 3924 /* bridge_fragment generates a mbuf chain of packets */ 3925 /* that already include eth headers */ 3926 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) { 3927 i = (*mp)->m_pkthdr.len; 3928 if (i > ifp->if_mtu) { 3929 error = bridge_fragment(ifp, mp, &eh2, snap, 3930 &llc1); 3931 return (error); 3932 } 3933 } 3934 3935 /* Recalculate the ip checksum. */ 3936 ip = mtod(*mp, struct ip *); 3937 hlen = ip->ip_hl << 2; 3938 if (hlen < sizeof(struct ip)) 3939 goto bad; 3940 if (hlen > (*mp)->m_len) { 3941 if ((*mp = m_pullup(*mp, hlen)) == NULL) 3942 goto bad; 3943 ip = mtod(*mp, struct ip *); 3944 if (ip == NULL) 3945 goto bad; 3946 } 3947 ip->ip_sum = 0; 3948 if (hlen == sizeof(struct ip)) 3949 ip->ip_sum = in_cksum_hdr(ip); 3950 else 3951 ip->ip_sum = in_cksum(*mp, hlen); 3952 3953 break; 3954 #endif /* INET */ 3955 #ifdef INET6 3956 case ETHERTYPE_IPV6: 3957 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv = 3958 pfil_mbuf_out(V_inet6_pfil_head, mp, bifp, NULL)) != 3959 PFIL_PASS) 3960 break; 3961 3962 if (V_pfil_member && ifp != NULL) { 3963 rv = (dir == PFIL_OUT) ? 3964 pfil_mbuf_out(V_inet6_pfil_head, mp, ifp, NULL) : 3965 pfil_mbuf_in(V_inet6_pfil_head, mp, ifp, NULL); 3966 if (rv != PFIL_PASS) 3967 break; 3968 } 3969 3970 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv = 3971 pfil_mbuf_in(V_inet6_pfil_head, mp, bifp, NULL)) != 3972 PFIL_PASS) 3973 break; 3974 break; 3975 #endif 3976 } 3977 3978 switch (rv) { 3979 case PFIL_CONSUMED: 3980 return (0); 3981 case PFIL_DROPPED: 3982 return (EACCES); 3983 default: 3984 break; 3985 } 3986 3987 error = -1; 3988 3989 /* 3990 * Finally, put everything back the way it was and return 3991 */ 3992 if (snap) { 3993 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT); 3994 if (*mp == NULL) 3995 return (error); 3996 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc)); 3997 } 3998 3999 M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT); 4000 if (*mp == NULL) 4001 return (error); 4002 bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN); 4003 4004 return (0); 4005 4006 bad: 4007 m_freem(*mp); 4008 *mp = NULL; 4009 return (error); 4010 } 4011 4012 #ifdef INET 4013 /* 4014 * Perform basic checks on header size since 4015 * pfil assumes ip_input has already processed 4016 * it for it. Cut-and-pasted from ip_input.c. 4017 * Given how simple the IPv6 version is, 4018 * does the IPv4 version really need to be 4019 * this complicated? 4020 * 4021 * XXX Should we update ipstat here, or not? 4022 * XXX Right now we update ipstat but not 4023 * XXX csum_counter. 4024 */ 4025 static int 4026 bridge_ip_checkbasic(struct mbuf **mp) 4027 { 4028 struct mbuf *m = *mp; 4029 struct ip *ip; 4030 int len, hlen; 4031 u_short sum; 4032 4033 if (*mp == NULL) 4034 return (-1); 4035 4036 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 4037 if ((m = m_copyup(m, sizeof(struct ip), 4038 (max_linkhdr + 3) & ~3)) == NULL) { 4039 /* XXXJRT new stat, please */ 4040 KMOD_IPSTAT_INC(ips_toosmall); 4041 goto bad; 4042 } 4043 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 4044 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 4045 KMOD_IPSTAT_INC(ips_toosmall); 4046 goto bad; 4047 } 4048 } 4049 ip = mtod(m, struct ip *); 4050 if (ip == NULL) goto bad; 4051 4052 if (ip->ip_v != IPVERSION) { 4053 KMOD_IPSTAT_INC(ips_badvers); 4054 goto bad; 4055 } 4056 hlen = ip->ip_hl << 2; 4057 if (hlen < sizeof(struct ip)) { /* minimum header length */ 4058 KMOD_IPSTAT_INC(ips_badhlen); 4059 goto bad; 4060 } 4061 if (hlen > m->m_len) { 4062 if ((m = m_pullup(m, hlen)) == NULL) { 4063 KMOD_IPSTAT_INC(ips_badhlen); 4064 goto bad; 4065 } 4066 ip = mtod(m, struct ip *); 4067 if (ip == NULL) goto bad; 4068 } 4069 4070 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 4071 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 4072 } else { 4073 if (hlen == sizeof(struct ip)) { 4074 sum = in_cksum_hdr(ip); 4075 } else { 4076 sum = in_cksum(m, hlen); 4077 } 4078 } 4079 if (sum) { 4080 KMOD_IPSTAT_INC(ips_badsum); 4081 goto bad; 4082 } 4083 4084 /* Retrieve the packet length. */ 4085 len = ntohs(ip->ip_len); 4086 4087 /* 4088 * Check for additional length bogosity 4089 */ 4090 if (len < hlen) { 4091 KMOD_IPSTAT_INC(ips_badlen); 4092 goto bad; 4093 } 4094 4095 /* 4096 * Check that the amount of data in the buffers 4097 * is as at least much as the IP header would have us expect. 4098 * Drop packet if shorter than we expect. 4099 */ 4100 if (m->m_pkthdr.len < len) { 4101 KMOD_IPSTAT_INC(ips_tooshort); 4102 goto bad; 4103 } 4104 4105 /* Checks out, proceed */ 4106 *mp = m; 4107 return (0); 4108 4109 bad: 4110 *mp = m; 4111 return (-1); 4112 } 4113 #endif /* INET */ 4114 4115 #ifdef INET6 4116 /* 4117 * Same as above, but for IPv6. 4118 * Cut-and-pasted from ip6_input.c. 4119 * XXX Should we update ip6stat, or not? 4120 */ 4121 static int 4122 bridge_ip6_checkbasic(struct mbuf **mp) 4123 { 4124 struct mbuf *m = *mp; 4125 struct ip6_hdr *ip6; 4126 4127 /* 4128 * If the IPv6 header is not aligned, slurp it up into a new 4129 * mbuf with space for link headers, in the event we forward 4130 * it. Otherwise, if it is aligned, make sure the entire base 4131 * IPv6 header is in the first mbuf of the chain. 4132 */ 4133 if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { 4134 struct ifnet *inifp = m->m_pkthdr.rcvif; 4135 if ((m = m_copyup(m, sizeof(struct ip6_hdr), 4136 (max_linkhdr + 3) & ~3)) == NULL) { 4137 /* XXXJRT new stat, please */ 4138 IP6STAT_INC(ip6s_toosmall); 4139 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 4140 goto bad; 4141 } 4142 } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) { 4143 struct ifnet *inifp = m->m_pkthdr.rcvif; 4144 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 4145 IP6STAT_INC(ip6s_toosmall); 4146 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 4147 goto bad; 4148 } 4149 } 4150 4151 ip6 = mtod(m, struct ip6_hdr *); 4152 4153 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 4154 IP6STAT_INC(ip6s_badvers); 4155 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 4156 goto bad; 4157 } 4158 4159 /* Checks out, proceed */ 4160 *mp = m; 4161 return (0); 4162 4163 bad: 4164 *mp = m; 4165 return (-1); 4166 } 4167 #endif /* INET6 */ 4168 4169 #ifdef INET 4170 /* 4171 * bridge_fragment: 4172 * 4173 * Fragment mbuf chain in multiple packets and prepend ethernet header. 4174 */ 4175 static int 4176 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh, 4177 int snap, struct llc *llc) 4178 { 4179 struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL; 4180 struct ip *ip; 4181 int error = -1; 4182 4183 if (m->m_len < sizeof(struct ip) && 4184 (m = m_pullup(m, sizeof(struct ip))) == NULL) 4185 goto dropit; 4186 ip = mtod(m, struct ip *); 4187 4188 m->m_pkthdr.csum_flags |= CSUM_IP; 4189 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist); 4190 if (error) 4191 goto dropit; 4192 4193 /* 4194 * Walk the chain and re-add the Ethernet header for 4195 * each mbuf packet. 4196 */ 4197 for (mcur = m; mcur; mcur = mcur->m_nextpkt) { 4198 nextpkt = mcur->m_nextpkt; 4199 mcur->m_nextpkt = NULL; 4200 if (snap) { 4201 M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT); 4202 if (mcur == NULL) { 4203 error = ENOBUFS; 4204 if (mprev != NULL) 4205 mprev->m_nextpkt = nextpkt; 4206 goto dropit; 4207 } 4208 bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc)); 4209 } 4210 4211 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT); 4212 if (mcur == NULL) { 4213 error = ENOBUFS; 4214 if (mprev != NULL) 4215 mprev->m_nextpkt = nextpkt; 4216 goto dropit; 4217 } 4218 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN); 4219 4220 /* 4221 * The previous two M_PREPEND could have inserted one or two 4222 * mbufs in front so we have to update the previous packet's 4223 * m_nextpkt. 4224 */ 4225 mcur->m_nextpkt = nextpkt; 4226 if (mprev != NULL) 4227 mprev->m_nextpkt = mcur; 4228 else { 4229 /* The first mbuf in the original chain needs to be 4230 * updated. */ 4231 *mp = mcur; 4232 } 4233 mprev = mcur; 4234 } 4235 4236 KMOD_IPSTAT_INC(ips_fragmented); 4237 return (error); 4238 4239 dropit: 4240 for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */ 4241 m = mcur->m_nextpkt; 4242 m_freem(mcur); 4243 } 4244 return (error); 4245 } 4246 #endif /* INET */ 4247 4248 static void 4249 bridge_linkstate(struct ifnet *ifp) 4250 { 4251 struct bridge_softc *sc = NULL; 4252 struct bridge_iflist *bif; 4253 struct epoch_tracker et; 4254 4255 NET_EPOCH_ENTER(et); 4256 4257 bif = ifp->if_bridge; 4258 if (bif) 4259 sc = bif->bif_sc; 4260 4261 if (sc != NULL) { 4262 bridge_linkcheck(sc); 4263 bstp_linkstate(&bif->bif_stp); 4264 } 4265 4266 NET_EPOCH_EXIT(et); 4267 } 4268 4269 static void 4270 bridge_linkcheck(struct bridge_softc *sc) 4271 { 4272 struct bridge_iflist *bif; 4273 int new_link, hasls; 4274 4275 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc); 4276 4277 new_link = LINK_STATE_DOWN; 4278 hasls = 0; 4279 /* Our link is considered up if at least one of our ports is active */ 4280 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { 4281 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE) 4282 hasls++; 4283 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) { 4284 new_link = LINK_STATE_UP; 4285 break; 4286 } 4287 } 4288 if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) { 4289 /* If no interfaces support link-state then we default to up */ 4290 new_link = LINK_STATE_UP; 4291 } 4292 if_link_state_change(sc->sc_ifp, new_link); 4293 } 4294