net/bridge/br_device.c

   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         u16 vid = 0;
  38
  39         memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  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         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_hdr(skb)->h_proto == htons(ETH_P_ARP) ||
  65              eth_hdr(skb)->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 struct lock_class_key bridge_netdev_addr_lock_key;
 111
 112 static void br_set_lockdep_class(struct net_device *dev)
 113 {
 114         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
 115 }
 116
 117 static int br_dev_init(struct net_device *dev)
 118 {
 119         struct net_bridge *br = netdev_priv(dev);
 120         int err;
 121
 122         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 123         if (!br->stats)
 124                 return -ENOMEM;
 125
 126         err = br_fdb_hash_init(br);
 127         if (err) {
 128                 free_percpu(br->stats);
 129                 return err;
 130         }
 131
 132         err = br_mdb_hash_init(br);
 133         if (err) {
 134                 free_percpu(br->stats);
 135                 br_fdb_hash_fini(br);
 136                 return err;
 137         }
 138
 139         err = br_vlan_init(br);
 140         if (err) {
 141                 free_percpu(br->stats);
 142                 br_mdb_hash_fini(br);
 143                 br_fdb_hash_fini(br);
 144                 return err;
 145         }
 146
 147         err = br_multicast_init_stats(br);
 148         if (err) {
 149                 free_percpu(br->stats);
 150                 br_vlan_flush(br);
 151                 br_mdb_hash_fini(br);
 152                 br_fdb_hash_fini(br);
 153         }
 154
 155         br_set_lockdep_class(dev);
 156         return err;
 157 }
 158
 159 static void br_dev_uninit(struct net_device *dev)
 160 {
 161         struct net_bridge *br = netdev_priv(dev);
 162
 163         br_multicast_dev_del(br);
 164         br_multicast_uninit_stats(br);
 165         br_vlan_flush(br);
 166         br_mdb_hash_fini(br);
 167         br_fdb_hash_fini(br);
 168         free_percpu(br->stats);
 169 }
 170
 171 static int br_dev_open(struct net_device *dev)
 172 {
 173         struct net_bridge *br = netdev_priv(dev);
 174
 175         netdev_update_features(dev);
 176         netif_start_queue(dev);
 177         br_stp_enable_bridge(br);
 178         br_multicast_open(br);
 179
 180         return 0;
 181 }
 182
 183 static void br_dev_set_multicast_list(struct net_device *dev)
 184 {
 185 }
 186
 187 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 188 {
 189         if (change & IFF_PROMISC)
 190                 br_manage_promisc(netdev_priv(dev));
 191 }
 192
 193 static int br_dev_stop(struct net_device *dev)
 194 {
 195         struct net_bridge *br = netdev_priv(dev);
 196
 197         br_stp_disable_bridge(br);
 198         br_multicast_stop(br);
 199
 200         netif_stop_queue(dev);
 201
 202         return 0;
 203 }
 204
 205 static void br_get_stats64(struct net_device *dev,
 206                            struct rtnl_link_stats64 *stats)
 207 {
 208         struct net_bridge *br = netdev_priv(dev);
 209         struct pcpu_sw_netstats tmp, sum = { 0 };
 210         unsigned int cpu;
 211
 212         for_each_possible_cpu(cpu) {
 213                 unsigned int start;
 214                 const struct pcpu_sw_netstats *bstats
 215                         = per_cpu_ptr(br->stats, cpu);
 216                 do {
 217                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 218                         memcpy(&tmp, bstats, sizeof(tmp));
 219                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 220                 sum.tx_bytes   += tmp.tx_bytes;
 221                 sum.tx_packets += tmp.tx_packets;
 222                 sum.rx_bytes   += tmp.rx_bytes;
 223                 sum.rx_packets += tmp.rx_packets;
 224         }
 225
 226         stats->tx_bytes   = sum.tx_bytes;
 227         stats->tx_packets = sum.tx_packets;
 228         stats->rx_bytes   = sum.rx_bytes;
 229         stats->rx_packets = sum.rx_packets;
 230 }
 231
 232 static int br_change_mtu(struct net_device *dev, int new_mtu)
 233 {
 234         struct net_bridge *br = netdev_priv(dev);
 235
 236         dev->mtu = new_mtu;
 237
 238         /* this flag will be cleared if the MTU was automatically adjusted */
 239         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 240 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 241         /* remember the MTU in the rtable for PMTU */
 242         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 243 #endif
 244
 245         return 0;
 246 }
 247
 248 /* Allow setting mac address to any valid ethernet address. */
 249 static int br_set_mac_address(struct net_device *dev, void *p)
 250 {
 251         struct net_bridge *br = netdev_priv(dev);
 252         struct sockaddr *addr = p;
 253
 254         if (!is_valid_ether_addr(addr->sa_data))
 255                 return -EADDRNOTAVAIL;
 256
 257         /* dev_set_mac_addr() can be called by a master device on bridge's
 258          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 259          */
 260         if (dev->reg_state != NETREG_REGISTERED)
 261                 return -EBUSY;
 262
 263         spin_lock_bh(&br->lock);
 264         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 265                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 266                 br_stp_change_bridge_id(br, addr->sa_data);
 267         }
 268         spin_unlock_bh(&br->lock);
 269
 270         return 0;
 271 }
 272
 273 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 274 {
 275         strlcpy(info->driver, "bridge", sizeof(info->driver));
 276         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 277         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 278         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 279 }
 280
 281 static int br_get_link_ksettings(struct net_device *dev,
 282                                  struct ethtool_link_ksettings *cmd)
 283 {
 284         struct net_bridge *br = netdev_priv(dev);
 285         struct net_bridge_port *p;
 286
 287         cmd->base.duplex = DUPLEX_UNKNOWN;
 288         cmd->base.port = PORT_OTHER;
 289         cmd->base.speed = SPEED_UNKNOWN;
 290
 291         list_for_each_entry(p, &br->port_list, list) {
 292                 struct ethtool_link_ksettings ecmd;
 293                 struct net_device *pdev = p->dev;
 294
 295                 if (!netif_running(pdev) || !netif_oper_up(pdev))
 296                         continue;
 297
 298                 if (__ethtool_get_link_ksettings(pdev, &ecmd))
 299                         continue;
 300
 301                 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN)
 302                         continue;
 303
 304                 if (cmd->base.speed == (__u32)SPEED_UNKNOWN ||
 305                     cmd->base.speed < ecmd.base.speed)
 306                         cmd->base.speed = ecmd.base.speed;
 307         }
 308
 309         return 0;
 310 }
 311
 312 static netdev_features_t br_fix_features(struct net_device *dev,
 313         netdev_features_t features)
 314 {
 315         struct net_bridge *br = netdev_priv(dev);
 316
 317         return br_features_recompute(br, features);
 318 }
 319
 320 #ifdef CONFIG_NET_POLL_CONTROLLER
 321 static void br_poll_controller(struct net_device *br_dev)
 322 {
 323 }
 324
 325 static void br_netpoll_cleanup(struct net_device *dev)
 326 {
 327         struct net_bridge *br = netdev_priv(dev);
 328         struct net_bridge_port *p;
 329
 330         list_for_each_entry(p, &br->port_list, list)
 331                 br_netpoll_disable(p);
 332 }
 333
 334 static int __br_netpoll_enable(struct net_bridge_port *p)
 335 {
 336         struct netpoll *np;
 337         int err;
 338
 339         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 340         if (!np)
 341                 return -ENOMEM;
 342
 343         err = __netpoll_setup(np, p->dev);
 344         if (err) {
 345                 kfree(np);
 346                 return err;
 347         }
 348
 349         p->np = np;
 350         return err;
 351 }
 352
 353 int br_netpoll_enable(struct net_bridge_port *p)
 354 {
 355         if (!p->br->dev->npinfo)
 356                 return 0;
 357
 358         return __br_netpoll_enable(p);
 359 }
 360
 361 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 362 {
 363         struct net_bridge *br = netdev_priv(dev);
 364         struct net_bridge_port *p;
 365         int err = 0;
 366
 367         list_for_each_entry(p, &br->port_list, list) {
 368                 if (!p->dev)
 369                         continue;
 370                 err = __br_netpoll_enable(p);
 371                 if (err)
 372                         goto fail;
 373         }
 374
 375 out:
 376         return err;
 377
 378 fail:
 379         br_netpoll_cleanup(dev);
 380         goto out;
 381 }
 382
 383 void br_netpoll_disable(struct net_bridge_port *p)
 384 {
 385         struct netpoll *np = p->np;
 386
 387         if (!np)
 388                 return;
 389
 390         p->np = NULL;
 391
 392         __netpoll_free(np);
 393 }
 394
 395 #endif
 396
 397 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 398                         struct netlink_ext_ack *extack)
 399
 400 {
 401         struct net_bridge *br = netdev_priv(dev);
 402
 403         return br_add_if(br, slave_dev, extack);
 404 }
 405
 406 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 407 {
 408         struct net_bridge *br = netdev_priv(dev);
 409
 410         return br_del_if(br, slave_dev);
 411 }
 412
 413 static const struct ethtool_ops br_ethtool_ops = {
 414         .get_drvinfo             = br_getinfo,
 415         .get_link                = ethtool_op_get_link,
 416         .get_link_ksettings      = br_get_link_ksettings,
 417 };
 418
 419 static const struct net_device_ops br_netdev_ops = {
 420         .ndo_open                = br_dev_open,
 421         .ndo_stop                = br_dev_stop,
 422         .ndo_init                = br_dev_init,
 423         .ndo_uninit              = br_dev_uninit,
 424         .ndo_start_xmit          = br_dev_xmit,
 425         .ndo_get_stats64         = br_get_stats64,
 426         .ndo_set_mac_address     = br_set_mac_address,
 427         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 428         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 429         .ndo_change_mtu          = br_change_mtu,
 430         .ndo_do_ioctl            = br_dev_ioctl,
 431 #ifdef CONFIG_NET_POLL_CONTROLLER
 432         .ndo_netpoll_setup       = br_netpoll_setup,
 433         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 434         .ndo_poll_controller     = br_poll_controller,
 435 #endif
 436         .ndo_add_slave           = br_add_slave,
 437         .ndo_del_slave           = br_del_slave,
 438         .ndo_fix_features        = br_fix_features,
 439         .ndo_fdb_add             = br_fdb_add,
 440         .ndo_fdb_del             = br_fdb_delete,
 441         .ndo_fdb_dump            = br_fdb_dump,
 442         .ndo_fdb_get             = br_fdb_get,
 443         .ndo_bridge_getlink      = br_getlink,
 444         .ndo_bridge_setlink      = br_setlink,
 445         .ndo_bridge_dellink      = br_dellink,
 446         .ndo_features_check      = passthru_features_check,
 447 };
 448
 449 static struct device_type br_type = {
 450         .name   = "bridge",
 451 };
 452
 453 void br_dev_setup(struct net_device *dev)
 454 {
 455         struct net_bridge *br = netdev_priv(dev);
 456
 457         eth_hw_addr_random(dev);
 458         ether_setup(dev);
 459
 460         dev->netdev_ops = &br_netdev_ops;
 461         dev->needs_free_netdev = true;
 462         dev->ethtool_ops = &br_ethtool_ops;
 463         SET_NETDEV_DEVTYPE(dev, &br_type);
 464         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 465
 466         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 467                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 468         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 469                            NETIF_F_HW_VLAN_STAG_TX;
 470         dev->vlan_features = COMMON_FEATURES;
 471
 472         br->dev = dev;
 473         spin_lock_init(&br->lock);
 474         INIT_LIST_HEAD(&br->port_list);
 475         INIT_HLIST_HEAD(&br->fdb_list);
 476 #if IS_ENABLED(CONFIG_BRIDGE_MRP)
 477         INIT_LIST_HEAD(&br->mrp_list);
 478 #endif
 479         spin_lock_init(&br->hash_lock);
 480
 481         br->bridge_id.prio[0] = 0x80;
 482         br->bridge_id.prio[1] = 0x00;
 483
 484         ether_addr_copy(br->group_addr, eth_stp_addr);
 485
 486         br->stp_enabled = BR_NO_STP;
 487         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 488         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 489
 490         br->designated_root = br->bridge_id;
 491         br->bridge_max_age = br->max_age = 20 * HZ;
 492         br->bridge_hello_time = br->hello_time = 2 * HZ;
 493         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 494         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 495         dev->max_mtu = ETH_MAX_MTU;
 496
 497         br_netfilter_rtable_init(br);
 498         br_stp_timer_init(br);
 499         br_multicast_init(br);
 500         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 501 }