net/bridge/br_device.c

   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         const unsigned char *dest = skb->data;
  38         struct net_bridge_fdb_entry *dst;
  39         struct net_bridge_mdb_entry *mdst;
  40         struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
  41         const struct nf_br_ops *nf_ops;
  42         u16 vid = 0;
  43
  44         rcu_read_lock();
  45         nf_ops = rcu_dereference(nf_br_ops);
  46         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  47                 rcu_read_unlock();
  48                 return NETDEV_TX_OK;
  49         }
  50
  51         u64_stats_update_begin(&brstats->syncp);
  52         brstats->tx_packets++;
  53         brstats->tx_bytes += skb->len;
  54         u64_stats_update_end(&brstats->syncp);
  55
  56         BR_INPUT_SKB_CB(skb)->brdev = dev;
  57
  58         skb_reset_mac_header(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_broadcast_ether_addr(dest)) {
  65                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  66         } else if (is_multicast_ether_addr(dest)) {
  67                 if (unlikely(netpoll_tx_running(dev))) {
  68                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  69                         goto out;
  70                 }
  71                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  72                         kfree_skb(skb);
  73                         goto out;
  74                 }
  75
  76                 mdst = br_mdb_get(br, skb, vid);
  77                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  78                     br_multicast_querier_exists(br, eth_hdr(skb)))
  79                         br_multicast_flood(mdst, skb, false, true);
  80                 else
  81                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  82         } else if ((dst = __br_fdb_get(br, dest, vid)) != NULL) {
  83                 br_forward(dst->dst, skb, false, true);
  84         } else {
  85                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
  86         }
  87 out:
  88         rcu_read_unlock();
  89         return NETDEV_TX_OK;
  90 }
  91
  92 static void br_set_lockdep_class(struct net_device *dev)
  93 {
  94         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
  95 }
  96
  97 static int br_dev_init(struct net_device *dev)
  98 {
  99         struct net_bridge *br = netdev_priv(dev);
 100         int err;
 101
 102         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 103         if (!br->stats)
 104                 return -ENOMEM;
 105
 106         err = br_vlan_init(br);
 107         if (err) {
 108                 free_percpu(br->stats);
 109                 return err;
 110         }
 111
 112         err = br_multicast_init_stats(br);
 113         if (err) {
 114                 free_percpu(br->stats);
 115                 br_vlan_flush(br);
 116         }
 117         br_set_lockdep_class(dev);
 118
 119         return err;
 120 }
 121
 122 static int br_dev_open(struct net_device *dev)
 123 {
 124         struct net_bridge *br = netdev_priv(dev);
 125
 126         netdev_update_features(dev);
 127         netif_start_queue(dev);
 128         br_stp_enable_bridge(br);
 129         br_multicast_open(br);
 130
 131         return 0;
 132 }
 133
 134 static void br_dev_set_multicast_list(struct net_device *dev)
 135 {
 136 }
 137
 138 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 139 {
 140         if (change & IFF_PROMISC)
 141                 br_manage_promisc(netdev_priv(dev));
 142 }
 143
 144 static int br_dev_stop(struct net_device *dev)
 145 {
 146         struct net_bridge *br = netdev_priv(dev);
 147
 148         br_stp_disable_bridge(br);
 149         br_multicast_stop(br);
 150
 151         netif_stop_queue(dev);
 152
 153         return 0;
 154 }
 155
 156 static struct rtnl_link_stats64 *br_get_stats64(struct net_device *dev,
 157                                                 struct rtnl_link_stats64 *stats)
 158 {
 159         struct net_bridge *br = netdev_priv(dev);
 160         struct pcpu_sw_netstats tmp, sum = { 0 };
 161         unsigned int cpu;
 162
 163         for_each_possible_cpu(cpu) {
 164                 unsigned int start;
 165                 const struct pcpu_sw_netstats *bstats
 166                         = per_cpu_ptr(br->stats, cpu);
 167                 do {
 168                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 169                         memcpy(&tmp, bstats, sizeof(tmp));
 170                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 171                 sum.tx_bytes   += tmp.tx_bytes;
 172                 sum.tx_packets += tmp.tx_packets;
 173                 sum.rx_bytes   += tmp.rx_bytes;
 174                 sum.rx_packets += tmp.rx_packets;
 175         }
 176
 177         stats->tx_bytes   = sum.tx_bytes;
 178         stats->tx_packets = sum.tx_packets;
 179         stats->rx_bytes   = sum.rx_bytes;
 180         stats->rx_packets = sum.rx_packets;
 181
 182         return stats;
 183 }
 184
 185 static int br_change_mtu(struct net_device *dev, int new_mtu)
 186 {
 187         struct net_bridge *br = netdev_priv(dev);
 188         if (new_mtu > br_min_mtu(br))
 189                 return -EINVAL;
 190
 191         dev->mtu = new_mtu;
 192
 193 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 194         /* remember the MTU in the rtable for PMTU */
 195         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 196 #endif
 197
 198         return 0;
 199 }
 200
 201 /* Allow setting mac address to any valid ethernet address. */
 202 static int br_set_mac_address(struct net_device *dev, void *p)
 203 {
 204         struct net_bridge *br = netdev_priv(dev);
 205         struct sockaddr *addr = p;
 206
 207         if (!is_valid_ether_addr(addr->sa_data))
 208                 return -EADDRNOTAVAIL;
 209
 210         spin_lock_bh(&br->lock);
 211         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 212                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 213                 br_stp_change_bridge_id(br, addr->sa_data);
 214         }
 215         spin_unlock_bh(&br->lock);
 216
 217         return 0;
 218 }
 219
 220 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 221 {
 222         strlcpy(info->driver, "bridge", sizeof(info->driver));
 223         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 224         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 225         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 226 }
 227
 228 static netdev_features_t br_fix_features(struct net_device *dev,
 229         netdev_features_t features)
 230 {
 231         struct net_bridge *br = netdev_priv(dev);
 232
 233         return br_features_recompute(br, features);
 234 }
 235
 236 #ifdef CONFIG_NET_POLL_CONTROLLER
 237 static void br_poll_controller(struct net_device *br_dev)
 238 {
 239 }
 240
 241 static void br_netpoll_cleanup(struct net_device *dev)
 242 {
 243         struct net_bridge *br = netdev_priv(dev);
 244         struct net_bridge_port *p;
 245
 246         list_for_each_entry(p, &br->port_list, list)
 247                 br_netpoll_disable(p);
 248 }
 249
 250 static int __br_netpoll_enable(struct net_bridge_port *p)
 251 {
 252         struct netpoll *np;
 253         int err;
 254
 255         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 256         if (!np)
 257                 return -ENOMEM;
 258
 259         err = __netpoll_setup(np, p->dev);
 260         if (err) {
 261                 kfree(np);
 262                 return err;
 263         }
 264
 265         p->np = np;
 266         return err;
 267 }
 268
 269 int br_netpoll_enable(struct net_bridge_port *p)
 270 {
 271         if (!p->br->dev->npinfo)
 272                 return 0;
 273
 274         return __br_netpoll_enable(p);
 275 }
 276
 277 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 278 {
 279         struct net_bridge *br = netdev_priv(dev);
 280         struct net_bridge_port *p;
 281         int err = 0;
 282
 283         list_for_each_entry(p, &br->port_list, list) {
 284                 if (!p->dev)
 285                         continue;
 286                 err = __br_netpoll_enable(p);
 287                 if (err)
 288                         goto fail;
 289         }
 290
 291 out:
 292         return err;
 293
 294 fail:
 295         br_netpoll_cleanup(dev);
 296         goto out;
 297 }
 298
 299 void br_netpoll_disable(struct net_bridge_port *p)
 300 {
 301         struct netpoll *np = p->np;
 302
 303         if (!np)
 304                 return;
 305
 306         p->np = NULL;
 307
 308         __netpoll_free_async(np);
 309 }
 310
 311 #endif
 312
 313 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev)
 314
 315 {
 316         struct net_bridge *br = netdev_priv(dev);
 317
 318         return br_add_if(br, slave_dev);
 319 }
 320
 321 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 322 {
 323         struct net_bridge *br = netdev_priv(dev);
 324
 325         return br_del_if(br, slave_dev);
 326 }
 327
 328 static const struct ethtool_ops br_ethtool_ops = {
 329         .get_drvinfo    = br_getinfo,
 330         .get_link       = ethtool_op_get_link,
 331 };
 332
 333 static const struct net_device_ops br_netdev_ops = {
 334         .ndo_open                = br_dev_open,
 335         .ndo_stop                = br_dev_stop,
 336         .ndo_init                = br_dev_init,
 337         .ndo_start_xmit          = br_dev_xmit,
 338         .ndo_get_stats64         = br_get_stats64,
 339         .ndo_set_mac_address     = br_set_mac_address,
 340         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 341         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 342         .ndo_change_mtu          = br_change_mtu,
 343         .ndo_do_ioctl            = br_dev_ioctl,
 344 #ifdef CONFIG_NET_POLL_CONTROLLER
 345         .ndo_netpoll_setup       = br_netpoll_setup,
 346         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 347         .ndo_poll_controller     = br_poll_controller,
 348 #endif
 349         .ndo_add_slave           = br_add_slave,
 350         .ndo_del_slave           = br_del_slave,
 351         .ndo_fix_features        = br_fix_features,
 352         .ndo_neigh_construct     = netdev_default_l2upper_neigh_construct,
 353         .ndo_neigh_destroy       = netdev_default_l2upper_neigh_destroy,
 354         .ndo_fdb_add             = br_fdb_add,
 355         .ndo_fdb_del             = br_fdb_delete,
 356         .ndo_fdb_dump            = br_fdb_dump,
 357         .ndo_bridge_getlink      = br_getlink,
 358         .ndo_bridge_setlink      = br_setlink,
 359         .ndo_bridge_dellink      = br_dellink,
 360         .ndo_features_check      = passthru_features_check,
 361 };
 362
 363 static void br_dev_free(struct net_device *dev)
 364 {
 365         struct net_bridge *br = netdev_priv(dev);
 366
 367         free_percpu(br->stats);
 368         free_netdev(dev);
 369 }
 370
 371 static struct device_type br_type = {
 372         .name   = "bridge",
 373 };
 374
 375 void br_dev_setup(struct net_device *dev)
 376 {
 377         struct net_bridge *br = netdev_priv(dev);
 378
 379         eth_hw_addr_random(dev);
 380         ether_setup(dev);
 381
 382         dev->netdev_ops = &br_netdev_ops;
 383         dev->destructor = br_dev_free;
 384         dev->ethtool_ops = &br_ethtool_ops;
 385         SET_NETDEV_DEVTYPE(dev, &br_type);
 386         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 387
 388         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 389                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 390         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 391                            NETIF_F_HW_VLAN_STAG_TX;
 392         dev->vlan_features = COMMON_FEATURES;
 393
 394         br->dev = dev;
 395         spin_lock_init(&br->lock);
 396         INIT_LIST_HEAD(&br->port_list);
 397         spin_lock_init(&br->hash_lock);
 398
 399         br->bridge_id.prio[0] = 0x80;
 400         br->bridge_id.prio[1] = 0x00;
 401
 402         ether_addr_copy(br->group_addr, eth_reserved_addr_base);
 403
 404         br->stp_enabled = BR_NO_STP;
 405         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 406         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 407
 408         br->designated_root = br->bridge_id;
 409         br->bridge_max_age = br->max_age = 20 * HZ;
 410         br->bridge_hello_time = br->hello_time = 2 * HZ;
 411         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 412         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 413         dev->max_mtu = ETH_MAX_MTU;
 414
 415         br_netfilter_rtable_init(br);
 416         br_stp_timer_init(br);
 417         br_multicast_init(br);
 418 }