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