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 60 skb_reset_mac_header(skb); 61 eth = eth_hdr(skb); 62 skb_pull(skb, ETH_HLEN); 63 64 if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid)) 65 goto out; 66 67 if (IS_ENABLED(CONFIG_INET) && 68 (eth->h_proto == htons(ETH_P_ARP) || 69 eth->h_proto == htons(ETH_P_RARP)) && 70 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) { 71 br_do_proxy_suppress_arp(skb, br, vid, NULL); 72 } else if (IS_ENABLED(CONFIG_IPV6) && 73 skb->protocol == htons(ETH_P_IPV6) && 74 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) && 75 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 76 sizeof(struct nd_msg)) && 77 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 78 struct nd_msg *msg, _msg; 79 80 msg = br_is_nd_neigh_msg(skb, &_msg); 81 if (msg) 82 br_do_suppress_nd(skb, br, vid, NULL, msg); 83 } 84 85 dest = eth_hdr(skb)->h_dest; 86 if (is_broadcast_ether_addr(dest)) { 87 br_flood(br, skb, BR_PKT_BROADCAST, false, true); 88 } else if (is_multicast_ether_addr(dest)) { 89 if (unlikely(netpoll_tx_running(dev))) { 90 br_flood(br, skb, BR_PKT_MULTICAST, false, true); 91 goto out; 92 } 93 if (br_multicast_rcv(br, NULL, skb, vid)) { 94 kfree_skb(skb); 95 goto out; 96 } 97 98 mdst = br_mdb_get(br, skb, vid); 99 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) && 100 br_multicast_querier_exists(br, eth_hdr(skb))) 101 br_multicast_flood(mdst, skb, false, true); 102 else 103 br_flood(br, skb, BR_PKT_MULTICAST, false, true); 104 } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) { 105 br_forward(dst->dst, skb, false, true); 106 } else { 107 br_flood(br, skb, BR_PKT_UNICAST, false, true); 108 } 109 out: 110 rcu_read_unlock(); 111 return NETDEV_TX_OK; 112 } 113 114 static void br_set_lockdep_class(struct net_device *dev) 115 { 116 lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key); 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 br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 125 if (!br->stats) 126 return -ENOMEM; 127 128 err = br_fdb_hash_init(br); 129 if (err) { 130 free_percpu(br->stats); 131 return err; 132 } 133 134 err = br_vlan_init(br); 135 if (err) { 136 free_percpu(br->stats); 137 br_fdb_hash_fini(br); 138 return err; 139 } 140 141 err = br_multicast_init_stats(br); 142 if (err) { 143 free_percpu(br->stats); 144 br_vlan_flush(br); 145 br_fdb_hash_fini(br); 146 } 147 br_set_lockdep_class(dev); 148 149 return err; 150 } 151 152 static void br_dev_uninit(struct net_device *dev) 153 { 154 struct net_bridge *br = netdev_priv(dev); 155 156 br_multicast_dev_del(br); 157 br_multicast_uninit_stats(br); 158 br_vlan_flush(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 spin_lock_bh(&br->lock); 250 if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) { 251 /* Mac address will be changed in br_stp_change_bridge_id(). */ 252 br_stp_change_bridge_id(br, addr->sa_data); 253 } 254 spin_unlock_bh(&br->lock); 255 256 return 0; 257 } 258 259 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info) 260 { 261 strlcpy(info->driver, "bridge", sizeof(info->driver)); 262 strlcpy(info->version, BR_VERSION, sizeof(info->version)); 263 strlcpy(info->fw_version, "N/A", sizeof(info->fw_version)); 264 strlcpy(info->bus_info, "N/A", sizeof(info->bus_info)); 265 } 266 267 static netdev_features_t br_fix_features(struct net_device *dev, 268 netdev_features_t features) 269 { 270 struct net_bridge *br = netdev_priv(dev); 271 272 return br_features_recompute(br, features); 273 } 274 275 #ifdef CONFIG_NET_POLL_CONTROLLER 276 static void br_poll_controller(struct net_device *br_dev) 277 { 278 } 279 280 static void br_netpoll_cleanup(struct net_device *dev) 281 { 282 struct net_bridge *br = netdev_priv(dev); 283 struct net_bridge_port *p; 284 285 list_for_each_entry(p, &br->port_list, list) 286 br_netpoll_disable(p); 287 } 288 289 static int __br_netpoll_enable(struct net_bridge_port *p) 290 { 291 struct netpoll *np; 292 int err; 293 294 np = kzalloc(sizeof(*p->np), GFP_KERNEL); 295 if (!np) 296 return -ENOMEM; 297 298 err = __netpoll_setup(np, p->dev); 299 if (err) { 300 kfree(np); 301 return err; 302 } 303 304 p->np = np; 305 return err; 306 } 307 308 int br_netpoll_enable(struct net_bridge_port *p) 309 { 310 if (!p->br->dev->npinfo) 311 return 0; 312 313 return __br_netpoll_enable(p); 314 } 315 316 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) 317 { 318 struct net_bridge *br = netdev_priv(dev); 319 struct net_bridge_port *p; 320 int err = 0; 321 322 list_for_each_entry(p, &br->port_list, list) { 323 if (!p->dev) 324 continue; 325 err = __br_netpoll_enable(p); 326 if (err) 327 goto fail; 328 } 329 330 out: 331 return err; 332 333 fail: 334 br_netpoll_cleanup(dev); 335 goto out; 336 } 337 338 void br_netpoll_disable(struct net_bridge_port *p) 339 { 340 struct netpoll *np = p->np; 341 342 if (!np) 343 return; 344 345 p->np = NULL; 346 347 __netpoll_free(np); 348 } 349 350 #endif 351 352 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev, 353 struct netlink_ext_ack *extack) 354 355 { 356 struct net_bridge *br = netdev_priv(dev); 357 358 return br_add_if(br, slave_dev, extack); 359 } 360 361 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev) 362 { 363 struct net_bridge *br = netdev_priv(dev); 364 365 return br_del_if(br, slave_dev); 366 } 367 368 static const struct ethtool_ops br_ethtool_ops = { 369 .get_drvinfo = br_getinfo, 370 .get_link = ethtool_op_get_link, 371 }; 372 373 static const struct net_device_ops br_netdev_ops = { 374 .ndo_open = br_dev_open, 375 .ndo_stop = br_dev_stop, 376 .ndo_init = br_dev_init, 377 .ndo_uninit = br_dev_uninit, 378 .ndo_start_xmit = br_dev_xmit, 379 .ndo_get_stats64 = br_get_stats64, 380 .ndo_set_mac_address = br_set_mac_address, 381 .ndo_set_rx_mode = br_dev_set_multicast_list, 382 .ndo_change_rx_flags = br_dev_change_rx_flags, 383 .ndo_change_mtu = br_change_mtu, 384 .ndo_do_ioctl = br_dev_ioctl, 385 #ifdef CONFIG_NET_POLL_CONTROLLER 386 .ndo_netpoll_setup = br_netpoll_setup, 387 .ndo_netpoll_cleanup = br_netpoll_cleanup, 388 .ndo_poll_controller = br_poll_controller, 389 #endif 390 .ndo_add_slave = br_add_slave, 391 .ndo_del_slave = br_del_slave, 392 .ndo_fix_features = br_fix_features, 393 .ndo_fdb_add = br_fdb_add, 394 .ndo_fdb_del = br_fdb_delete, 395 .ndo_fdb_dump = br_fdb_dump, 396 .ndo_bridge_getlink = br_getlink, 397 .ndo_bridge_setlink = br_setlink, 398 .ndo_bridge_dellink = br_dellink, 399 .ndo_features_check = passthru_features_check, 400 }; 401 402 static struct device_type br_type = { 403 .name = "bridge", 404 }; 405 406 void br_dev_setup(struct net_device *dev) 407 { 408 struct net_bridge *br = netdev_priv(dev); 409 410 eth_hw_addr_random(dev); 411 ether_setup(dev); 412 413 dev->netdev_ops = &br_netdev_ops; 414 dev->needs_free_netdev = true; 415 dev->ethtool_ops = &br_ethtool_ops; 416 SET_NETDEV_DEVTYPE(dev, &br_type); 417 dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE; 418 419 dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL | 420 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; 421 dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | 422 NETIF_F_HW_VLAN_STAG_TX; 423 dev->vlan_features = COMMON_FEATURES; 424 425 br->dev = dev; 426 spin_lock_init(&br->lock); 427 INIT_LIST_HEAD(&br->port_list); 428 INIT_HLIST_HEAD(&br->fdb_list); 429 spin_lock_init(&br->hash_lock); 430 431 br->bridge_id.prio[0] = 0x80; 432 br->bridge_id.prio[1] = 0x00; 433 434 ether_addr_copy(br->group_addr, eth_stp_addr); 435 436 br->stp_enabled = BR_NO_STP; 437 br->group_fwd_mask = BR_GROUPFWD_DEFAULT; 438 br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT; 439 440 br->designated_root = br->bridge_id; 441 br->bridge_max_age = br->max_age = 20 * HZ; 442 br->bridge_hello_time = br->hello_time = 2 * HZ; 443 br->bridge_forward_delay = br->forward_delay = 15 * HZ; 444 br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME; 445 dev->max_mtu = ETH_MAX_MTU; 446 447 br_netfilter_rtable_init(br); 448 br_stp_timer_init(br); 449 br_multicast_init(br); 450 INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup); 451 } 452