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 struct net_bridge *br = netdev_priv(dev); 31 struct net_bridge_fdb_entry *dst; 32 struct net_bridge_mdb_entry *mdst; 33 struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats); 34 const struct nf_br_ops *nf_ops; 35 u8 state = BR_STATE_FORWARDING; 36 const unsigned char *dest; 37 struct ethhdr *eth; 38 u16 vid = 0; 39 40 rcu_read_lock(); 41 nf_ops = rcu_dereference(nf_br_ops); 42 if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) { 43 rcu_read_unlock(); 44 return NETDEV_TX_OK; 45 } 46 47 u64_stats_update_begin(&brstats->syncp); 48 brstats->tx_packets++; 49 brstats->tx_bytes += skb->len; 50 u64_stats_update_end(&brstats->syncp); 51 52 br_switchdev_frame_unmark(skb); 53 BR_INPUT_SKB_CB(skb)->brdev = dev; 54 BR_INPUT_SKB_CB(skb)->frag_max_size = 0; 55 56 skb_reset_mac_header(skb); 57 eth = eth_hdr(skb); 58 skb_pull(skb, ETH_HLEN); 59 60 if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid, &state)) 61 goto out; 62 63 if (IS_ENABLED(CONFIG_INET) && 64 (eth->h_proto == htons(ETH_P_ARP) || 65 eth->h_proto == htons(ETH_P_RARP)) && 66 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) { 67 br_do_proxy_suppress_arp(skb, br, vid, NULL); 68 } else if (IS_ENABLED(CONFIG_IPV6) && 69 skb->protocol == htons(ETH_P_IPV6) && 70 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) && 71 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 72 sizeof(struct nd_msg)) && 73 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 74 struct nd_msg *msg, _msg; 75 76 msg = br_is_nd_neigh_msg(skb, &_msg); 77 if (msg) 78 br_do_suppress_nd(skb, br, vid, NULL, msg); 79 } 80 81 dest = eth_hdr(skb)->h_dest; 82 if (is_broadcast_ether_addr(dest)) { 83 br_flood(br, skb, BR_PKT_BROADCAST, false, true); 84 } else if (is_multicast_ether_addr(dest)) { 85 if (unlikely(netpoll_tx_running(dev))) { 86 br_flood(br, skb, BR_PKT_MULTICAST, false, true); 87 goto out; 88 } 89 if (br_multicast_rcv(br, NULL, skb, vid)) { 90 kfree_skb(skb); 91 goto out; 92 } 93 94 mdst = br_mdb_get(br, skb, vid); 95 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) && 96 br_multicast_querier_exists(br, eth_hdr(skb))) 97 br_multicast_flood(mdst, skb, false, true); 98 else 99 br_flood(br, skb, BR_PKT_MULTICAST, false, true); 100 } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) { 101 br_forward(dst->dst, skb, false, true); 102 } else { 103 br_flood(br, skb, BR_PKT_UNICAST, false, true); 104 } 105 out: 106 rcu_read_unlock(); 107 return NETDEV_TX_OK; 108 } 109 110 static int br_dev_init(struct net_device *dev) 111 { 112 struct net_bridge *br = netdev_priv(dev); 113 int err; 114 115 br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 116 if (!br->stats) 117 return -ENOMEM; 118 119 err = br_fdb_hash_init(br); 120 if (err) { 121 free_percpu(br->stats); 122 return err; 123 } 124 125 err = br_mdb_hash_init(br); 126 if (err) { 127 free_percpu(br->stats); 128 br_fdb_hash_fini(br); 129 return err; 130 } 131 132 err = br_vlan_init(br); 133 if (err) { 134 free_percpu(br->stats); 135 br_mdb_hash_fini(br); 136 br_fdb_hash_fini(br); 137 return err; 138 } 139 140 err = br_multicast_init_stats(br); 141 if (err) { 142 free_percpu(br->stats); 143 br_vlan_flush(br); 144 br_mdb_hash_fini(br); 145 br_fdb_hash_fini(br); 146 } 147 148 return err; 149 } 150 151 static void br_dev_uninit(struct net_device *dev) 152 { 153 struct net_bridge *br = netdev_priv(dev); 154 155 br_multicast_dev_del(br); 156 br_multicast_uninit_stats(br); 157 br_vlan_flush(br); 158 br_mdb_hash_fini(br); 159 br_fdb_hash_fini(br); 160 free_percpu(br->stats); 161 } 162 163 static int br_dev_open(struct net_device *dev) 164 { 165 struct net_bridge *br = netdev_priv(dev); 166 167 netdev_update_features(dev); 168 netif_start_queue(dev); 169 br_stp_enable_bridge(br); 170 br_multicast_open(br); 171 172 return 0; 173 } 174 175 static void br_dev_set_multicast_list(struct net_device *dev) 176 { 177 } 178 179 static void br_dev_change_rx_flags(struct net_device *dev, int change) 180 { 181 if (change & IFF_PROMISC) 182 br_manage_promisc(netdev_priv(dev)); 183 } 184 185 static int br_dev_stop(struct net_device *dev) 186 { 187 struct net_bridge *br = netdev_priv(dev); 188 189 br_stp_disable_bridge(br); 190 br_multicast_stop(br); 191 192 netif_stop_queue(dev); 193 194 return 0; 195 } 196 197 static void br_get_stats64(struct net_device *dev, 198 struct rtnl_link_stats64 *stats) 199 { 200 struct net_bridge *br = netdev_priv(dev); 201 struct pcpu_sw_netstats tmp, sum = { 0 }; 202 unsigned int cpu; 203 204 for_each_possible_cpu(cpu) { 205 unsigned int start; 206 const struct pcpu_sw_netstats *bstats 207 = per_cpu_ptr(br->stats, cpu); 208 do { 209 start = u64_stats_fetch_begin_irq(&bstats->syncp); 210 memcpy(&tmp, bstats, sizeof(tmp)); 211 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start)); 212 sum.tx_bytes += tmp.tx_bytes; 213 sum.tx_packets += tmp.tx_packets; 214 sum.rx_bytes += tmp.rx_bytes; 215 sum.rx_packets += tmp.rx_packets; 216 } 217 218 stats->tx_bytes = sum.tx_bytes; 219 stats->tx_packets = sum.tx_packets; 220 stats->rx_bytes = sum.rx_bytes; 221 stats->rx_packets = sum.rx_packets; 222 } 223 224 static int br_change_mtu(struct net_device *dev, int new_mtu) 225 { 226 struct net_bridge *br = netdev_priv(dev); 227 228 dev->mtu = new_mtu; 229 230 /* this flag will be cleared if the MTU was automatically adjusted */ 231 br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true); 232 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 233 /* remember the MTU in the rtable for PMTU */ 234 dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu); 235 #endif 236 237 return 0; 238 } 239 240 /* Allow setting mac address to any valid ethernet address. */ 241 static int br_set_mac_address(struct net_device *dev, void *p) 242 { 243 struct net_bridge *br = netdev_priv(dev); 244 struct sockaddr *addr = p; 245 246 if (!is_valid_ether_addr(addr->sa_data)) 247 return -EADDRNOTAVAIL; 248 249 /* dev_set_mac_addr() can be called by a master device on bridge's 250 * NETDEV_UNREGISTER, but since it's being destroyed do nothing 251 */ 252 if (dev->reg_state != NETREG_REGISTERED) 253 return -EBUSY; 254 255 spin_lock_bh(&br->lock); 256 if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) { 257 /* Mac address will be changed in br_stp_change_bridge_id(). */ 258 br_stp_change_bridge_id(br, addr->sa_data); 259 } 260 spin_unlock_bh(&br->lock); 261 262 return 0; 263 } 264 265 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info) 266 { 267 strlcpy(info->driver, "bridge", sizeof(info->driver)); 268 strlcpy(info->version, BR_VERSION, sizeof(info->version)); 269 strlcpy(info->fw_version, "N/A", sizeof(info->fw_version)); 270 strlcpy(info->bus_info, "N/A", sizeof(info->bus_info)); 271 } 272 273 static int br_get_link_ksettings(struct net_device *dev, 274 struct ethtool_link_ksettings *cmd) 275 { 276 struct net_bridge *br = netdev_priv(dev); 277 struct net_bridge_port *p; 278 279 cmd->base.duplex = DUPLEX_UNKNOWN; 280 cmd->base.port = PORT_OTHER; 281 cmd->base.speed = SPEED_UNKNOWN; 282 283 list_for_each_entry(p, &br->port_list, list) { 284 struct ethtool_link_ksettings ecmd; 285 struct net_device *pdev = p->dev; 286 287 if (!netif_running(pdev) || !netif_oper_up(pdev)) 288 continue; 289 290 if (__ethtool_get_link_ksettings(pdev, &ecmd)) 291 continue; 292 293 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN) 294 continue; 295 296 if (cmd->base.speed == (__u32)SPEED_UNKNOWN || 297 cmd->base.speed < ecmd.base.speed) 298 cmd->base.speed = ecmd.base.speed; 299 } 300 301 return 0; 302 } 303 304 static netdev_features_t br_fix_features(struct net_device *dev, 305 netdev_features_t features) 306 { 307 struct net_bridge *br = netdev_priv(dev); 308 309 return br_features_recompute(br, features); 310 } 311 312 #ifdef CONFIG_NET_POLL_CONTROLLER 313 static void br_poll_controller(struct net_device *br_dev) 314 { 315 } 316 317 static void br_netpoll_cleanup(struct net_device *dev) 318 { 319 struct net_bridge *br = netdev_priv(dev); 320 struct net_bridge_port *p; 321 322 list_for_each_entry(p, &br->port_list, list) 323 br_netpoll_disable(p); 324 } 325 326 static int __br_netpoll_enable(struct net_bridge_port *p) 327 { 328 struct netpoll *np; 329 int err; 330 331 np = kzalloc(sizeof(*p->np), GFP_KERNEL); 332 if (!np) 333 return -ENOMEM; 334 335 err = __netpoll_setup(np, p->dev); 336 if (err) { 337 kfree(np); 338 return err; 339 } 340 341 p->np = np; 342 return err; 343 } 344 345 int br_netpoll_enable(struct net_bridge_port *p) 346 { 347 if (!p->br->dev->npinfo) 348 return 0; 349 350 return __br_netpoll_enable(p); 351 } 352 353 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) 354 { 355 struct net_bridge *br = netdev_priv(dev); 356 struct net_bridge_port *p; 357 int err = 0; 358 359 list_for_each_entry(p, &br->port_list, list) { 360 if (!p->dev) 361 continue; 362 err = __br_netpoll_enable(p); 363 if (err) 364 goto fail; 365 } 366 367 out: 368 return err; 369 370 fail: 371 br_netpoll_cleanup(dev); 372 goto out; 373 } 374 375 void br_netpoll_disable(struct net_bridge_port *p) 376 { 377 struct netpoll *np = p->np; 378 379 if (!np) 380 return; 381 382 p->np = NULL; 383 384 __netpoll_free(np); 385 } 386 387 #endif 388 389 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev, 390 struct netlink_ext_ack *extack) 391 392 { 393 struct net_bridge *br = netdev_priv(dev); 394 395 return br_add_if(br, slave_dev, extack); 396 } 397 398 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev) 399 { 400 struct net_bridge *br = netdev_priv(dev); 401 402 return br_del_if(br, slave_dev); 403 } 404 405 static const struct ethtool_ops br_ethtool_ops = { 406 .get_drvinfo = br_getinfo, 407 .get_link = ethtool_op_get_link, 408 .get_link_ksettings = br_get_link_ksettings, 409 }; 410 411 static const struct net_device_ops br_netdev_ops = { 412 .ndo_open = br_dev_open, 413 .ndo_stop = br_dev_stop, 414 .ndo_init = br_dev_init, 415 .ndo_uninit = br_dev_uninit, 416 .ndo_start_xmit = br_dev_xmit, 417 .ndo_get_stats64 = br_get_stats64, 418 .ndo_set_mac_address = br_set_mac_address, 419 .ndo_set_rx_mode = br_dev_set_multicast_list, 420 .ndo_change_rx_flags = br_dev_change_rx_flags, 421 .ndo_change_mtu = br_change_mtu, 422 .ndo_do_ioctl = br_dev_ioctl, 423 #ifdef CONFIG_NET_POLL_CONTROLLER 424 .ndo_netpoll_setup = br_netpoll_setup, 425 .ndo_netpoll_cleanup = br_netpoll_cleanup, 426 .ndo_poll_controller = br_poll_controller, 427 #endif 428 .ndo_add_slave = br_add_slave, 429 .ndo_del_slave = br_del_slave, 430 .ndo_fix_features = br_fix_features, 431 .ndo_fdb_add = br_fdb_add, 432 .ndo_fdb_del = br_fdb_delete, 433 .ndo_fdb_dump = br_fdb_dump, 434 .ndo_fdb_get = br_fdb_get, 435 .ndo_bridge_getlink = br_getlink, 436 .ndo_bridge_setlink = br_setlink, 437 .ndo_bridge_dellink = br_dellink, 438 .ndo_features_check = passthru_features_check, 439 }; 440 441 static struct device_type br_type = { 442 .name = "bridge", 443 }; 444 445 void br_dev_setup(struct net_device *dev) 446 { 447 struct net_bridge *br = netdev_priv(dev); 448 449 eth_hw_addr_random(dev); 450 ether_setup(dev); 451 452 dev->netdev_ops = &br_netdev_ops; 453 dev->needs_free_netdev = true; 454 dev->ethtool_ops = &br_ethtool_ops; 455 SET_NETDEV_DEVTYPE(dev, &br_type); 456 dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE; 457 458 dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL | 459 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; 460 dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | 461 NETIF_F_HW_VLAN_STAG_TX; 462 dev->vlan_features = COMMON_FEATURES; 463 464 br->dev = dev; 465 spin_lock_init(&br->lock); 466 INIT_LIST_HEAD(&br->port_list); 467 INIT_HLIST_HEAD(&br->fdb_list); 468 spin_lock_init(&br->hash_lock); 469 470 br->bridge_id.prio[0] = 0x80; 471 br->bridge_id.prio[1] = 0x00; 472 473 ether_addr_copy(br->group_addr, eth_stp_addr); 474 475 br->stp_enabled = BR_NO_STP; 476 br->group_fwd_mask = BR_GROUPFWD_DEFAULT; 477 br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT; 478 479 br->designated_root = br->bridge_id; 480 br->bridge_max_age = br->max_age = 20 * HZ; 481 br->bridge_hello_time = br->hello_time = 2 * HZ; 482 br->bridge_forward_delay = br->forward_delay = 15 * HZ; 483 br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME; 484 dev->max_mtu = ETH_MAX_MTU; 485 486 br_netfilter_rtable_init(br); 487 br_stp_timer_init(br); 488 br_multicast_init(br); 489 INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup); 490 } 491