1 /* 2 * Device handling code 3 * Linux ethernet bridge 4 * 5 * Authors: 6 * Lennert Buytenhek <buytenh@gnu.org> 7 * 8 * This program is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU General Public License 10 * as published by the Free Software Foundation; either version 11 * 2 of the License, or (at your option) any later version. 12 */ 13 14 #include <linux/kernel.h> 15 #include <linux/netdevice.h> 16 #include <linux/netpoll.h> 17 #include <linux/etherdevice.h> 18 #include <linux/ethtool.h> 19 #include <linux/list.h> 20 #include <linux/netfilter_bridge.h> 21 22 #include <linux/uaccess.h> 23 #include "br_private.h" 24 25 #define COMMON_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA | \ 26 NETIF_F_GSO_MASK | NETIF_F_HW_CSUM) 27 28 const struct nf_br_ops __rcu *nf_br_ops __read_mostly; 29 EXPORT_SYMBOL_GPL(nf_br_ops); 30 31 static struct lock_class_key bridge_netdev_addr_lock_key; 32 33 /* net device transmit always called with BH disabled */ 34 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev) 35 { 36 struct net_bridge *br = netdev_priv(dev); 37 struct net_bridge_fdb_entry *dst; 38 struct net_bridge_mdb_entry *mdst; 39 struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats); 40 const struct nf_br_ops *nf_ops; 41 const unsigned char *dest; 42 struct ethhdr *eth; 43 u16 vid = 0; 44 45 rcu_read_lock(); 46 nf_ops = rcu_dereference(nf_br_ops); 47 if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) { 48 rcu_read_unlock(); 49 return NETDEV_TX_OK; 50 } 51 52 u64_stats_update_begin(&brstats->syncp); 53 brstats->tx_packets++; 54 brstats->tx_bytes += skb->len; 55 u64_stats_update_end(&brstats->syncp); 56 57 br_switchdev_frame_unmark(skb); 58 BR_INPUT_SKB_CB(skb)->brdev = dev; 59 BR_INPUT_SKB_CB(skb)->frag_max_size = 0; 60 61 skb_reset_mac_header(skb); 62 eth = eth_hdr(skb); 63 skb_pull(skb, ETH_HLEN); 64 65 if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid)) 66 goto out; 67 68 if (IS_ENABLED(CONFIG_INET) && 69 (eth->h_proto == htons(ETH_P_ARP) || 70 eth->h_proto == htons(ETH_P_RARP)) && 71 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) { 72 br_do_proxy_suppress_arp(skb, br, vid, NULL); 73 } else if (IS_ENABLED(CONFIG_IPV6) && 74 skb->protocol == htons(ETH_P_IPV6) && 75 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) && 76 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 77 sizeof(struct nd_msg)) && 78 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 79 struct nd_msg *msg, _msg; 80 81 msg = br_is_nd_neigh_msg(skb, &_msg); 82 if (msg) 83 br_do_suppress_nd(skb, br, vid, NULL, msg); 84 } 85 86 dest = eth_hdr(skb)->h_dest; 87 if (is_broadcast_ether_addr(dest)) { 88 br_flood(br, skb, BR_PKT_BROADCAST, false, true); 89 } else if (is_multicast_ether_addr(dest)) { 90 if (unlikely(netpoll_tx_running(dev))) { 91 br_flood(br, skb, BR_PKT_MULTICAST, false, true); 92 goto out; 93 } 94 if (br_multicast_rcv(br, NULL, skb, vid)) { 95 kfree_skb(skb); 96 goto out; 97 } 98 99 mdst = br_mdb_get(br, skb, vid); 100 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) && 101 br_multicast_querier_exists(br, eth_hdr(skb))) 102 br_multicast_flood(mdst, skb, false, true); 103 else 104 br_flood(br, skb, BR_PKT_MULTICAST, false, true); 105 } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) { 106 br_forward(dst->dst, skb, false, true); 107 } else { 108 br_flood(br, skb, BR_PKT_UNICAST, false, true); 109 } 110 out: 111 rcu_read_unlock(); 112 return NETDEV_TX_OK; 113 } 114 115 static void br_set_lockdep_class(struct net_device *dev) 116 { 117 lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key); 118 } 119 120 static int br_dev_init(struct net_device *dev) 121 { 122 struct net_bridge *br = netdev_priv(dev); 123 int err; 124 125 br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 126 if (!br->stats) 127 return -ENOMEM; 128 129 err = br_fdb_hash_init(br); 130 if (err) { 131 free_percpu(br->stats); 132 return err; 133 } 134 135 err = br_mdb_hash_init(br); 136 if (err) { 137 free_percpu(br->stats); 138 br_fdb_hash_fini(br); 139 return err; 140 } 141 142 err = br_vlan_init(br); 143 if (err) { 144 free_percpu(br->stats); 145 br_mdb_hash_fini(br); 146 br_fdb_hash_fini(br); 147 return err; 148 } 149 150 err = br_multicast_init_stats(br); 151 if (err) { 152 free_percpu(br->stats); 153 br_vlan_flush(br); 154 br_mdb_hash_fini(br); 155 br_fdb_hash_fini(br); 156 } 157 br_set_lockdep_class(dev); 158 159 return err; 160 } 161 162 static void br_dev_uninit(struct net_device *dev) 163 { 164 struct net_bridge *br = netdev_priv(dev); 165 166 br_multicast_dev_del(br); 167 br_multicast_uninit_stats(br); 168 br_vlan_flush(br); 169 br_mdb_hash_fini(br); 170 br_fdb_hash_fini(br); 171 free_percpu(br->stats); 172 } 173 174 static int br_dev_open(struct net_device *dev) 175 { 176 struct net_bridge *br = netdev_priv(dev); 177 178 netdev_update_features(dev); 179 netif_start_queue(dev); 180 br_stp_enable_bridge(br); 181 br_multicast_open(br); 182 183 return 0; 184 } 185 186 static void br_dev_set_multicast_list(struct net_device *dev) 187 { 188 } 189 190 static void br_dev_change_rx_flags(struct net_device *dev, int change) 191 { 192 if (change & IFF_PROMISC) 193 br_manage_promisc(netdev_priv(dev)); 194 } 195 196 static int br_dev_stop(struct net_device *dev) 197 { 198 struct net_bridge *br = netdev_priv(dev); 199 200 br_stp_disable_bridge(br); 201 br_multicast_stop(br); 202 203 netif_stop_queue(dev); 204 205 return 0; 206 } 207 208 static void br_get_stats64(struct net_device *dev, 209 struct rtnl_link_stats64 *stats) 210 { 211 struct net_bridge *br = netdev_priv(dev); 212 struct pcpu_sw_netstats tmp, sum = { 0 }; 213 unsigned int cpu; 214 215 for_each_possible_cpu(cpu) { 216 unsigned int start; 217 const struct pcpu_sw_netstats *bstats 218 = per_cpu_ptr(br->stats, cpu); 219 do { 220 start = u64_stats_fetch_begin_irq(&bstats->syncp); 221 memcpy(&tmp, bstats, sizeof(tmp)); 222 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start)); 223 sum.tx_bytes += tmp.tx_bytes; 224 sum.tx_packets += tmp.tx_packets; 225 sum.rx_bytes += tmp.rx_bytes; 226 sum.rx_packets += tmp.rx_packets; 227 } 228 229 stats->tx_bytes = sum.tx_bytes; 230 stats->tx_packets = sum.tx_packets; 231 stats->rx_bytes = sum.rx_bytes; 232 stats->rx_packets = sum.rx_packets; 233 } 234 235 static int br_change_mtu(struct net_device *dev, int new_mtu) 236 { 237 struct net_bridge *br = netdev_priv(dev); 238 239 dev->mtu = new_mtu; 240 241 /* this flag will be cleared if the MTU was automatically adjusted */ 242 br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true); 243 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 244 /* remember the MTU in the rtable for PMTU */ 245 dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu); 246 #endif 247 248 return 0; 249 } 250 251 /* Allow setting mac address to any valid ethernet address. */ 252 static int br_set_mac_address(struct net_device *dev, void *p) 253 { 254 struct net_bridge *br = netdev_priv(dev); 255 struct sockaddr *addr = p; 256 257 if (!is_valid_ether_addr(addr->sa_data)) 258 return -EADDRNOTAVAIL; 259 260 spin_lock_bh(&br->lock); 261 if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) { 262 /* Mac address will be changed in br_stp_change_bridge_id(). */ 263 br_stp_change_bridge_id(br, addr->sa_data); 264 } 265 spin_unlock_bh(&br->lock); 266 267 return 0; 268 } 269 270 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info) 271 { 272 strlcpy(info->driver, "bridge", sizeof(info->driver)); 273 strlcpy(info->version, BR_VERSION, sizeof(info->version)); 274 strlcpy(info->fw_version, "N/A", sizeof(info->fw_version)); 275 strlcpy(info->bus_info, "N/A", sizeof(info->bus_info)); 276 } 277 278 static netdev_features_t br_fix_features(struct net_device *dev, 279 netdev_features_t features) 280 { 281 struct net_bridge *br = netdev_priv(dev); 282 283 return br_features_recompute(br, features); 284 } 285 286 #ifdef CONFIG_NET_POLL_CONTROLLER 287 static void br_poll_controller(struct net_device *br_dev) 288 { 289 } 290 291 static void br_netpoll_cleanup(struct net_device *dev) 292 { 293 struct net_bridge *br = netdev_priv(dev); 294 struct net_bridge_port *p; 295 296 list_for_each_entry(p, &br->port_list, list) 297 br_netpoll_disable(p); 298 } 299 300 static int __br_netpoll_enable(struct net_bridge_port *p) 301 { 302 struct netpoll *np; 303 int err; 304 305 np = kzalloc(sizeof(*p->np), GFP_KERNEL); 306 if (!np) 307 return -ENOMEM; 308 309 err = __netpoll_setup(np, p->dev); 310 if (err) { 311 kfree(np); 312 return err; 313 } 314 315 p->np = np; 316 return err; 317 } 318 319 int br_netpoll_enable(struct net_bridge_port *p) 320 { 321 if (!p->br->dev->npinfo) 322 return 0; 323 324 return __br_netpoll_enable(p); 325 } 326 327 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) 328 { 329 struct net_bridge *br = netdev_priv(dev); 330 struct net_bridge_port *p; 331 int err = 0; 332 333 list_for_each_entry(p, &br->port_list, list) { 334 if (!p->dev) 335 continue; 336 err = __br_netpoll_enable(p); 337 if (err) 338 goto fail; 339 } 340 341 out: 342 return err; 343 344 fail: 345 br_netpoll_cleanup(dev); 346 goto out; 347 } 348 349 void br_netpoll_disable(struct net_bridge_port *p) 350 { 351 struct netpoll *np = p->np; 352 353 if (!np) 354 return; 355 356 p->np = NULL; 357 358 __netpoll_free(np); 359 } 360 361 #endif 362 363 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev, 364 struct netlink_ext_ack *extack) 365 366 { 367 struct net_bridge *br = netdev_priv(dev); 368 369 return br_add_if(br, slave_dev, extack); 370 } 371 372 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev) 373 { 374 struct net_bridge *br = netdev_priv(dev); 375 376 return br_del_if(br, slave_dev); 377 } 378 379 static const struct ethtool_ops br_ethtool_ops = { 380 .get_drvinfo = br_getinfo, 381 .get_link = ethtool_op_get_link, 382 }; 383 384 static const struct net_device_ops br_netdev_ops = { 385 .ndo_open = br_dev_open, 386 .ndo_stop = br_dev_stop, 387 .ndo_init = br_dev_init, 388 .ndo_uninit = br_dev_uninit, 389 .ndo_start_xmit = br_dev_xmit, 390 .ndo_get_stats64 = br_get_stats64, 391 .ndo_set_mac_address = br_set_mac_address, 392 .ndo_set_rx_mode = br_dev_set_multicast_list, 393 .ndo_change_rx_flags = br_dev_change_rx_flags, 394 .ndo_change_mtu = br_change_mtu, 395 .ndo_do_ioctl = br_dev_ioctl, 396 #ifdef CONFIG_NET_POLL_CONTROLLER 397 .ndo_netpoll_setup = br_netpoll_setup, 398 .ndo_netpoll_cleanup = br_netpoll_cleanup, 399 .ndo_poll_controller = br_poll_controller, 400 #endif 401 .ndo_add_slave = br_add_slave, 402 .ndo_del_slave = br_del_slave, 403 .ndo_fix_features = br_fix_features, 404 .ndo_fdb_add = br_fdb_add, 405 .ndo_fdb_del = br_fdb_delete, 406 .ndo_fdb_dump = br_fdb_dump, 407 .ndo_fdb_get = br_fdb_get, 408 .ndo_bridge_getlink = br_getlink, 409 .ndo_bridge_setlink = br_setlink, 410 .ndo_bridge_dellink = br_dellink, 411 .ndo_features_check = passthru_features_check, 412 }; 413 414 static struct device_type br_type = { 415 .name = "bridge", 416 }; 417 418 void br_dev_setup(struct net_device *dev) 419 { 420 struct net_bridge *br = netdev_priv(dev); 421 422 eth_hw_addr_random(dev); 423 ether_setup(dev); 424 425 dev->netdev_ops = &br_netdev_ops; 426 dev->needs_free_netdev = true; 427 dev->ethtool_ops = &br_ethtool_ops; 428 SET_NETDEV_DEVTYPE(dev, &br_type); 429 dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE; 430 431 dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL | 432 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; 433 dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | 434 NETIF_F_HW_VLAN_STAG_TX; 435 dev->vlan_features = COMMON_FEATURES; 436 437 br->dev = dev; 438 spin_lock_init(&br->lock); 439 INIT_LIST_HEAD(&br->port_list); 440 INIT_HLIST_HEAD(&br->fdb_list); 441 spin_lock_init(&br->hash_lock); 442 443 br->bridge_id.prio[0] = 0x80; 444 br->bridge_id.prio[1] = 0x00; 445 446 ether_addr_copy(br->group_addr, eth_stp_addr); 447 448 br->stp_enabled = BR_NO_STP; 449 br->group_fwd_mask = BR_GROUPFWD_DEFAULT; 450 br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT; 451 452 br->designated_root = br->bridge_id; 453 br->bridge_max_age = br->max_age = 20 * HZ; 454 br->bridge_hello_time = br->hello_time = 2 * HZ; 455 br->bridge_forward_delay = br->forward_delay = 15 * HZ; 456 br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME; 457 dev->max_mtu = ETH_MAX_MTU; 458 459 br_netfilter_rtable_init(br); 460 br_stp_timer_init(br); 461 br_multicast_init(br); 462 INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup); 463 } 464