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