1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Device handling code 4 * Linux ethernet bridge 5 * 6 * Authors: 7 * Lennert Buytenhek <buytenh@gnu.org> 8 */ 9 10 #include <linux/kernel.h> 11 #include <linux/netdevice.h> 12 #include <linux/netpoll.h> 13 #include <linux/etherdevice.h> 14 #include <linux/ethtool.h> 15 #include <linux/list.h> 16 #include <linux/netfilter_bridge.h> 17 18 #include <linux/uaccess.h> 19 #include "br_private.h" 20 21 #define COMMON_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA | \ 22 NETIF_F_GSO_MASK | NETIF_F_HW_CSUM) 23 24 const struct nf_br_ops __rcu *nf_br_ops __read_mostly; 25 EXPORT_SYMBOL_GPL(nf_br_ops); 26 27 /* net device transmit always called with BH disabled */ 28 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev) 29 { 30 enum skb_drop_reason reason = pskb_may_pull_reason(skb, ETH_HLEN); 31 struct net_bridge_mcast_port *pmctx_null = NULL; 32 struct net_bridge *br = netdev_priv(dev); 33 struct net_bridge_mcast *brmctx = &br->multicast_ctx; 34 struct net_bridge_fdb_entry *dst; 35 struct net_bridge_mdb_entry *mdst; 36 const struct nf_br_ops *nf_ops; 37 u8 state = BR_STATE_FORWARDING; 38 struct net_bridge_vlan *vlan; 39 const unsigned char *dest; 40 u16 vid = 0; 41 42 if (unlikely(reason != SKB_NOT_DROPPED_YET)) { 43 kfree_skb_reason(skb, reason); 44 return NETDEV_TX_OK; 45 } 46 47 memset(skb->cb, 0, sizeof(struct br_input_skb_cb)); 48 br_tc_skb_miss_set(skb, false); 49 50 rcu_read_lock(); 51 nf_ops = rcu_dereference(nf_br_ops); 52 if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) { 53 rcu_read_unlock(); 54 return NETDEV_TX_OK; 55 } 56 57 dev_sw_netstats_tx_add(dev, 1, skb->len); 58 59 br_switchdev_frame_unmark(skb); 60 BR_INPUT_SKB_CB(skb)->brdev = dev; 61 BR_INPUT_SKB_CB(skb)->frag_max_size = 0; 62 63 skb_reset_mac_header(skb); 64 skb_pull(skb, ETH_HLEN); 65 66 if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid, 67 &state, &vlan)) 68 goto out; 69 70 if (IS_ENABLED(CONFIG_INET) && 71 (eth_hdr(skb)->h_proto == htons(ETH_P_ARP) || 72 eth_hdr(skb)->h_proto == htons(ETH_P_RARP)) && 73 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) { 74 br_do_proxy_suppress_arp(skb, br, vid, NULL); 75 } else if (IS_ENABLED(CONFIG_IPV6) && 76 skb->protocol == htons(ETH_P_IPV6) && 77 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) && 78 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 79 sizeof(struct nd_msg)) && 80 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 81 struct nd_msg *msg, _msg; 82 83 msg = br_is_nd_neigh_msg(skb, &_msg); 84 if (msg) 85 br_do_suppress_nd(skb, br, vid, NULL, msg); 86 } 87 88 dest = eth_hdr(skb)->h_dest; 89 if (is_broadcast_ether_addr(dest)) { 90 br_flood(br, skb, BR_PKT_BROADCAST, false, true, vid); 91 } else if (is_multicast_ether_addr(dest)) { 92 if (unlikely(netpoll_tx_running(dev))) { 93 br_flood(br, skb, BR_PKT_MULTICAST, false, true, vid); 94 goto out; 95 } 96 if (br_multicast_rcv(&brmctx, &pmctx_null, vlan, skb, vid)) { 97 kfree_skb(skb); 98 goto out; 99 } 100 101 mdst = br_mdb_get(brmctx, skb, vid); 102 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) && 103 br_multicast_querier_exists(brmctx, eth_hdr(skb), mdst)) 104 br_multicast_flood(mdst, skb, brmctx, false, true); 105 else 106 br_flood(br, skb, BR_PKT_MULTICAST, false, true, vid); 107 } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) { 108 br_forward(dst->dst, skb, false, true); 109 } else { 110 br_flood(br, skb, BR_PKT_UNICAST, false, true, vid); 111 } 112 out: 113 rcu_read_unlock(); 114 return NETDEV_TX_OK; 115 } 116 117 static struct lock_class_key bridge_netdev_addr_lock_key; 118 119 static void br_set_lockdep_class(struct net_device *dev) 120 { 121 lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key); 122 } 123 124 static int br_dev_init(struct net_device *dev) 125 { 126 struct net_bridge *br = netdev_priv(dev); 127 int err; 128 129 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 130 if (!dev->tstats) 131 return -ENOMEM; 132 133 err = br_fdb_hash_init(br); 134 if (err) { 135 free_percpu(dev->tstats); 136 return err; 137 } 138 139 err = br_mdb_hash_init(br); 140 if (err) { 141 free_percpu(dev->tstats); 142 br_fdb_hash_fini(br); 143 return err; 144 } 145 146 err = br_vlan_init(br); 147 if (err) { 148 free_percpu(dev->tstats); 149 br_mdb_hash_fini(br); 150 br_fdb_hash_fini(br); 151 return err; 152 } 153 154 err = br_multicast_init_stats(br); 155 if (err) { 156 free_percpu(dev->tstats); 157 br_vlan_flush(br); 158 br_mdb_hash_fini(br); 159 br_fdb_hash_fini(br); 160 } 161 162 br_set_lockdep_class(dev); 163 return err; 164 } 165 166 static void br_dev_uninit(struct net_device *dev) 167 { 168 struct net_bridge *br = netdev_priv(dev); 169 170 br_multicast_dev_del(br); 171 br_multicast_uninit_stats(br); 172 br_vlan_flush(br); 173 br_mdb_hash_fini(br); 174 br_fdb_hash_fini(br); 175 free_percpu(dev->tstats); 176 } 177 178 static int br_dev_open(struct net_device *dev) 179 { 180 struct net_bridge *br = netdev_priv(dev); 181 182 netdev_update_features(dev); 183 netif_start_queue(dev); 184 br_stp_enable_bridge(br); 185 br_multicast_open(br); 186 187 if (br_opt_get(br, BROPT_MULTICAST_ENABLED)) 188 br_multicast_join_snoopers(br); 189 190 return 0; 191 } 192 193 static void br_dev_set_multicast_list(struct net_device *dev) 194 { 195 } 196 197 static void br_dev_change_rx_flags(struct net_device *dev, int change) 198 { 199 if (change & IFF_PROMISC) 200 br_manage_promisc(netdev_priv(dev)); 201 } 202 203 static int br_dev_stop(struct net_device *dev) 204 { 205 struct net_bridge *br = netdev_priv(dev); 206 207 br_stp_disable_bridge(br); 208 br_multicast_stop(br); 209 210 if (br_opt_get(br, BROPT_MULTICAST_ENABLED)) 211 br_multicast_leave_snoopers(br); 212 213 netif_stop_queue(dev); 214 215 return 0; 216 } 217 218 static int br_change_mtu(struct net_device *dev, int new_mtu) 219 { 220 struct net_bridge *br = netdev_priv(dev); 221 222 dev->mtu = new_mtu; 223 224 /* this flag will be cleared if the MTU was automatically adjusted */ 225 br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true); 226 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 227 /* remember the MTU in the rtable for PMTU */ 228 dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu); 229 #endif 230 231 return 0; 232 } 233 234 /* Allow setting mac address to any valid ethernet address. */ 235 static int br_set_mac_address(struct net_device *dev, void *p) 236 { 237 struct net_bridge *br = netdev_priv(dev); 238 struct sockaddr *addr = p; 239 240 if (!is_valid_ether_addr(addr->sa_data)) 241 return -EADDRNOTAVAIL; 242 243 /* dev_set_mac_addr() can be called by a master device on bridge's 244 * NETDEV_UNREGISTER, but since it's being destroyed do nothing 245 */ 246 if (dev->reg_state != NETREG_REGISTERED) 247 return -EBUSY; 248 249 spin_lock_bh(&br->lock); 250 if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) { 251 /* Mac address will be changed in br_stp_change_bridge_id(). */ 252 br_stp_change_bridge_id(br, addr->sa_data); 253 } 254 spin_unlock_bh(&br->lock); 255 256 return 0; 257 } 258 259 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info) 260 { 261 strscpy(info->driver, "bridge", sizeof(info->driver)); 262 strscpy(info->version, BR_VERSION, sizeof(info->version)); 263 strscpy(info->fw_version, "N/A", sizeof(info->fw_version)); 264 strscpy(info->bus_info, "N/A", sizeof(info->bus_info)); 265 } 266 267 static int br_get_link_ksettings(struct net_device *dev, 268 struct ethtool_link_ksettings *cmd) 269 { 270 struct net_bridge *br = netdev_priv(dev); 271 struct net_bridge_port *p; 272 273 cmd->base.duplex = DUPLEX_UNKNOWN; 274 cmd->base.port = PORT_OTHER; 275 cmd->base.speed = SPEED_UNKNOWN; 276 277 list_for_each_entry(p, &br->port_list, list) { 278 struct ethtool_link_ksettings ecmd; 279 struct net_device *pdev = p->dev; 280 281 if (!netif_running(pdev) || !netif_oper_up(pdev)) 282 continue; 283 284 if (__ethtool_get_link_ksettings(pdev, &ecmd)) 285 continue; 286 287 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN) 288 continue; 289 290 if (cmd->base.speed == (__u32)SPEED_UNKNOWN || 291 cmd->base.speed < ecmd.base.speed) 292 cmd->base.speed = ecmd.base.speed; 293 } 294 295 return 0; 296 } 297 298 static netdev_features_t br_fix_features(struct net_device *dev, 299 netdev_features_t features) 300 { 301 struct net_bridge *br = netdev_priv(dev); 302 303 return br_features_recompute(br, features); 304 } 305 306 #ifdef CONFIG_NET_POLL_CONTROLLER 307 static void br_poll_controller(struct net_device *br_dev) 308 { 309 } 310 311 static void br_netpoll_cleanup(struct net_device *dev) 312 { 313 struct net_bridge *br = netdev_priv(dev); 314 struct net_bridge_port *p; 315 316 list_for_each_entry(p, &br->port_list, list) 317 br_netpoll_disable(p); 318 } 319 320 static int __br_netpoll_enable(struct net_bridge_port *p) 321 { 322 struct netpoll *np; 323 int err; 324 325 np = kzalloc(sizeof(*p->np), GFP_KERNEL); 326 if (!np) 327 return -ENOMEM; 328 329 err = __netpoll_setup(np, p->dev); 330 if (err) { 331 kfree(np); 332 return err; 333 } 334 335 p->np = np; 336 return err; 337 } 338 339 int br_netpoll_enable(struct net_bridge_port *p) 340 { 341 if (!p->br->dev->npinfo) 342 return 0; 343 344 return __br_netpoll_enable(p); 345 } 346 347 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) 348 { 349 struct net_bridge *br = netdev_priv(dev); 350 struct net_bridge_port *p; 351 int err = 0; 352 353 list_for_each_entry(p, &br->port_list, list) { 354 if (!p->dev) 355 continue; 356 err = __br_netpoll_enable(p); 357 if (err) 358 goto fail; 359 } 360 361 out: 362 return err; 363 364 fail: 365 br_netpoll_cleanup(dev); 366 goto out; 367 } 368 369 void br_netpoll_disable(struct net_bridge_port *p) 370 { 371 struct netpoll *np = p->np; 372 373 if (!np) 374 return; 375 376 p->np = NULL; 377 378 __netpoll_free(np); 379 } 380 381 #endif 382 383 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev, 384 struct netlink_ext_ack *extack) 385 386 { 387 struct net_bridge *br = netdev_priv(dev); 388 389 return br_add_if(br, slave_dev, extack); 390 } 391 392 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev) 393 { 394 struct net_bridge *br = netdev_priv(dev); 395 396 return br_del_if(br, slave_dev); 397 } 398 399 static int br_fill_forward_path(struct net_device_path_ctx *ctx, 400 struct net_device_path *path) 401 { 402 struct net_bridge_fdb_entry *f; 403 struct net_bridge_port *dst; 404 struct net_bridge *br; 405 406 if (netif_is_bridge_port(ctx->dev)) 407 return -1; 408 409 br = netdev_priv(ctx->dev); 410 411 br_vlan_fill_forward_path_pvid(br, ctx, path); 412 413 f = br_fdb_find_rcu(br, ctx->daddr, path->bridge.vlan_id); 414 if (!f || !f->dst) 415 return -1; 416 417 dst = READ_ONCE(f->dst); 418 if (!dst) 419 return -1; 420 421 if (br_vlan_fill_forward_path_mode(br, dst, path)) 422 return -1; 423 424 path->type = DEV_PATH_BRIDGE; 425 path->dev = dst->br->dev; 426 ctx->dev = dst->dev; 427 428 switch (path->bridge.vlan_mode) { 429 case DEV_PATH_BR_VLAN_TAG: 430 if (ctx->num_vlans >= ARRAY_SIZE(ctx->vlan)) 431 return -ENOSPC; 432 ctx->vlan[ctx->num_vlans].id = path->bridge.vlan_id; 433 ctx->vlan[ctx->num_vlans].proto = path->bridge.vlan_proto; 434 ctx->num_vlans++; 435 break; 436 case DEV_PATH_BR_VLAN_UNTAG_HW: 437 case DEV_PATH_BR_VLAN_UNTAG: 438 ctx->num_vlans--; 439 break; 440 case DEV_PATH_BR_VLAN_KEEP: 441 break; 442 } 443 444 return 0; 445 } 446 447 static const struct ethtool_ops br_ethtool_ops = { 448 .get_drvinfo = br_getinfo, 449 .get_link = ethtool_op_get_link, 450 .get_link_ksettings = br_get_link_ksettings, 451 }; 452 453 static const struct net_device_ops br_netdev_ops = { 454 .ndo_open = br_dev_open, 455 .ndo_stop = br_dev_stop, 456 .ndo_init = br_dev_init, 457 .ndo_uninit = br_dev_uninit, 458 .ndo_start_xmit = br_dev_xmit, 459 .ndo_get_stats64 = dev_get_tstats64, 460 .ndo_set_mac_address = br_set_mac_address, 461 .ndo_set_rx_mode = br_dev_set_multicast_list, 462 .ndo_change_rx_flags = br_dev_change_rx_flags, 463 .ndo_change_mtu = br_change_mtu, 464 .ndo_siocdevprivate = br_dev_siocdevprivate, 465 #ifdef CONFIG_NET_POLL_CONTROLLER 466 .ndo_netpoll_setup = br_netpoll_setup, 467 .ndo_netpoll_cleanup = br_netpoll_cleanup, 468 .ndo_poll_controller = br_poll_controller, 469 #endif 470 .ndo_add_slave = br_add_slave, 471 .ndo_del_slave = br_del_slave, 472 .ndo_fix_features = br_fix_features, 473 .ndo_fdb_add = br_fdb_add, 474 .ndo_fdb_del = br_fdb_delete, 475 .ndo_fdb_del_bulk = br_fdb_delete_bulk, 476 .ndo_fdb_dump = br_fdb_dump, 477 .ndo_fdb_get = br_fdb_get, 478 .ndo_mdb_add = br_mdb_add, 479 .ndo_mdb_del = br_mdb_del, 480 .ndo_mdb_dump = br_mdb_dump, 481 .ndo_bridge_getlink = br_getlink, 482 .ndo_bridge_setlink = br_setlink, 483 .ndo_bridge_dellink = br_dellink, 484 .ndo_features_check = passthru_features_check, 485 .ndo_fill_forward_path = br_fill_forward_path, 486 }; 487 488 static struct device_type br_type = { 489 .name = "bridge", 490 }; 491 492 void br_dev_setup(struct net_device *dev) 493 { 494 struct net_bridge *br = netdev_priv(dev); 495 496 eth_hw_addr_random(dev); 497 ether_setup(dev); 498 499 dev->netdev_ops = &br_netdev_ops; 500 dev->needs_free_netdev = true; 501 dev->ethtool_ops = &br_ethtool_ops; 502 SET_NETDEV_DEVTYPE(dev, &br_type); 503 dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE; 504 505 dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL | 506 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; 507 dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | 508 NETIF_F_HW_VLAN_STAG_TX; 509 dev->vlan_features = COMMON_FEATURES; 510 511 br->dev = dev; 512 spin_lock_init(&br->lock); 513 INIT_LIST_HEAD(&br->port_list); 514 INIT_HLIST_HEAD(&br->fdb_list); 515 INIT_HLIST_HEAD(&br->frame_type_list); 516 #if IS_ENABLED(CONFIG_BRIDGE_MRP) 517 INIT_HLIST_HEAD(&br->mrp_list); 518 #endif 519 #if IS_ENABLED(CONFIG_BRIDGE_CFM) 520 INIT_HLIST_HEAD(&br->mep_list); 521 #endif 522 spin_lock_init(&br->hash_lock); 523 524 br->bridge_id.prio[0] = 0x80; 525 br->bridge_id.prio[1] = 0x00; 526 527 ether_addr_copy(br->group_addr, eth_stp_addr); 528 529 br->stp_enabled = BR_NO_STP; 530 br->group_fwd_mask = BR_GROUPFWD_DEFAULT; 531 br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT; 532 533 br->designated_root = br->bridge_id; 534 br->bridge_max_age = br->max_age = 20 * HZ; 535 br->bridge_hello_time = br->hello_time = 2 * HZ; 536 br->bridge_forward_delay = br->forward_delay = 15 * HZ; 537 br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME; 538 dev->max_mtu = ETH_MAX_MTU; 539 540 br_netfilter_rtable_init(br); 541 br_stp_timer_init(br); 542 br_multicast_init(br); 543 INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup); 544 } 545