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_find_rcu(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 void br_dev_uninit(struct net_device *dev)
 123 {
 124         struct net_bridge *br = netdev_priv(dev);
 125
 126         br_multicast_dev_del(br);
 127         br_multicast_uninit_stats(br);
 128         br_vlan_flush(br);
 129         free_percpu(br->stats);
 130 }
 131
 132 static int br_dev_open(struct net_device *dev)
 133 {
 134         struct net_bridge *br = netdev_priv(dev);
 135
 136         netdev_update_features(dev);
 137         netif_start_queue(dev);
 138         br_stp_enable_bridge(br);
 139         br_multicast_open(br);
 140
 141         return 0;
 142 }
 143
 144 static void br_dev_set_multicast_list(struct net_device *dev)
 145 {
 146 }
 147
 148 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 149 {
 150         if (change & IFF_PROMISC)
 151                 br_manage_promisc(netdev_priv(dev));
 152 }
 153
 154 static int br_dev_stop(struct net_device *dev)
 155 {
 156         struct net_bridge *br = netdev_priv(dev);
 157
 158         br_stp_disable_bridge(br);
 159         br_multicast_stop(br);
 160
 161         netif_stop_queue(dev);
 162
 163         return 0;
 164 }
 165
 166 static void br_get_stats64(struct net_device *dev,
 167                            struct rtnl_link_stats64 *stats)
 168 {
 169         struct net_bridge *br = netdev_priv(dev);
 170         struct pcpu_sw_netstats tmp, sum = { 0 };
 171         unsigned int cpu;
 172
 173         for_each_possible_cpu(cpu) {
 174                 unsigned int start;
 175                 const struct pcpu_sw_netstats *bstats
 176                         = per_cpu_ptr(br->stats, cpu);
 177                 do {
 178                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 179                         memcpy(&tmp, bstats, sizeof(tmp));
 180                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 181                 sum.tx_bytes   += tmp.tx_bytes;
 182                 sum.tx_packets += tmp.tx_packets;
 183                 sum.rx_bytes   += tmp.rx_bytes;
 184                 sum.rx_packets += tmp.rx_packets;
 185         }
 186
 187         stats->tx_bytes   = sum.tx_bytes;
 188         stats->tx_packets = sum.tx_packets;
 189         stats->rx_bytes   = sum.rx_bytes;
 190         stats->rx_packets = sum.rx_packets;
 191 }
 192
 193 static int br_change_mtu(struct net_device *dev, int new_mtu)
 194 {
 195         struct net_bridge *br = netdev_priv(dev);
 196         if (new_mtu > br_min_mtu(br))
 197                 return -EINVAL;
 198
 199         dev->mtu = new_mtu;
 200
 201 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 202         /* remember the MTU in the rtable for PMTU */
 203         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 204 #endif
 205
 206         return 0;
 207 }
 208
 209 /* Allow setting mac address to any valid ethernet address. */
 210 static int br_set_mac_address(struct net_device *dev, void *p)
 211 {
 212         struct net_bridge *br = netdev_priv(dev);
 213         struct sockaddr *addr = p;
 214
 215         if (!is_valid_ether_addr(addr->sa_data))
 216                 return -EADDRNOTAVAIL;
 217
 218         spin_lock_bh(&br->lock);
 219         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 220                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 221                 br_stp_change_bridge_id(br, addr->sa_data);
 222         }
 223         spin_unlock_bh(&br->lock);
 224
 225         return 0;
 226 }
 227
 228 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 229 {
 230         strlcpy(info->driver, "bridge", sizeof(info->driver));
 231         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 232         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 233         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 234 }
 235
 236 static netdev_features_t br_fix_features(struct net_device *dev,
 237         netdev_features_t features)
 238 {
 239         struct net_bridge *br = netdev_priv(dev);
 240
 241         return br_features_recompute(br, features);
 242 }
 243
 244 #ifdef CONFIG_NET_POLL_CONTROLLER
 245 static void br_poll_controller(struct net_device *br_dev)
 246 {
 247 }
 248
 249 static void br_netpoll_cleanup(struct net_device *dev)
 250 {
 251         struct net_bridge *br = netdev_priv(dev);
 252         struct net_bridge_port *p;
 253
 254         list_for_each_entry(p, &br->port_list, list)
 255                 br_netpoll_disable(p);
 256 }
 257
 258 static int __br_netpoll_enable(struct net_bridge_port *p)
 259 {
 260         struct netpoll *np;
 261         int err;
 262
 263         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 264         if (!np)
 265                 return -ENOMEM;
 266
 267         err = __netpoll_setup(np, p->dev);
 268         if (err) {
 269                 kfree(np);
 270                 return err;
 271         }
 272
 273         p->np = np;
 274         return err;
 275 }
 276
 277 int br_netpoll_enable(struct net_bridge_port *p)
 278 {
 279         if (!p->br->dev->npinfo)
 280                 return 0;
 281
 282         return __br_netpoll_enable(p);
 283 }
 284
 285 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 286 {
 287         struct net_bridge *br = netdev_priv(dev);
 288         struct net_bridge_port *p;
 289         int err = 0;
 290
 291         list_for_each_entry(p, &br->port_list, list) {
 292                 if (!p->dev)
 293                         continue;
 294                 err = __br_netpoll_enable(p);
 295                 if (err)
 296                         goto fail;
 297         }
 298
 299 out:
 300         return err;
 301
 302 fail:
 303         br_netpoll_cleanup(dev);
 304         goto out;
 305 }
 306
 307 void br_netpoll_disable(struct net_bridge_port *p)
 308 {
 309         struct netpoll *np = p->np;
 310
 311         if (!np)
 312                 return;
 313
 314         p->np = NULL;
 315
 316         __netpoll_free_async(np);
 317 }
 318
 319 #endif
 320
 321 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev)
 322
 323 {
 324         struct net_bridge *br = netdev_priv(dev);
 325
 326         return br_add_if(br, slave_dev);
 327 }
 328
 329 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 330 {
 331         struct net_bridge *br = netdev_priv(dev);
 332
 333         return br_del_if(br, slave_dev);
 334 }
 335
 336 static const struct ethtool_ops br_ethtool_ops = {
 337         .get_drvinfo    = br_getinfo,
 338         .get_link       = ethtool_op_get_link,
 339 };
 340
 341 static const struct net_device_ops br_netdev_ops = {
 342         .ndo_open                = br_dev_open,
 343         .ndo_stop                = br_dev_stop,
 344         .ndo_init                = br_dev_init,
 345         .ndo_uninit              = br_dev_uninit,
 346         .ndo_start_xmit          = br_dev_xmit,
 347         .ndo_get_stats64         = br_get_stats64,
 348         .ndo_set_mac_address     = br_set_mac_address,
 349         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 350         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 351         .ndo_change_mtu          = br_change_mtu,
 352         .ndo_do_ioctl            = br_dev_ioctl,
 353 #ifdef CONFIG_NET_POLL_CONTROLLER
 354         .ndo_netpoll_setup       = br_netpoll_setup,
 355         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 356         .ndo_poll_controller     = br_poll_controller,
 357 #endif
 358         .ndo_add_slave           = br_add_slave,
 359         .ndo_del_slave           = br_del_slave,
 360         .ndo_fix_features        = br_fix_features,
 361         .ndo_fdb_add             = br_fdb_add,
 362         .ndo_fdb_del             = br_fdb_delete,
 363         .ndo_fdb_dump            = br_fdb_dump,
 364         .ndo_bridge_getlink      = br_getlink,
 365         .ndo_bridge_setlink      = br_setlink,
 366         .ndo_bridge_dellink      = br_dellink,
 367         .ndo_features_check      = passthru_features_check,
 368 };
 369
 370 static struct device_type br_type = {
 371         .name   = "bridge",
 372 };
 373
 374 void br_dev_setup(struct net_device *dev)
 375 {
 376         struct net_bridge *br = netdev_priv(dev);
 377
 378         eth_hw_addr_random(dev);
 379         ether_setup(dev);
 380
 381         dev->netdev_ops = &br_netdev_ops;
 382         dev->destructor = free_netdev;
 383         dev->ethtool_ops = &br_ethtool_ops;
 384         SET_NETDEV_DEVTYPE(dev, &br_type);
 385         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 386
 387         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 388                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 389         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 390                            NETIF_F_HW_VLAN_STAG_TX;
 391         dev->vlan_features = COMMON_FEATURES;
 392
 393         br->dev = dev;
 394         spin_lock_init(&br->lock);
 395         INIT_LIST_HEAD(&br->port_list);
 396         spin_lock_init(&br->hash_lock);
 397
 398         br->bridge_id.prio[0] = 0x80;
 399         br->bridge_id.prio[1] = 0x00;
 400
 401         ether_addr_copy(br->group_addr, eth_reserved_addr_base);
 402
 403         br->stp_enabled = BR_NO_STP;
 404         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 405         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 406
 407         br->designated_root = br->bridge_id;
 408         br->bridge_max_age = br->max_age = 20 * HZ;
 409         br->bridge_hello_time = br->hello_time = 2 * HZ;
 410         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 411         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 412         dev->max_mtu = ETH_MAX_MTU;
 413
 414         br_netfilter_rtable_init(br);
 415         br_stp_timer_init(br);
 416         br_multicast_init(br);
 417         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 418 }