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