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         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         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         dest = eth_hdr(skb)->h_dest;
  65         if (is_broadcast_ether_addr(dest)) {
  66                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  67         } else if (is_multicast_ether_addr(dest)) {
  68                 if (unlikely(netpoll_tx_running(dev))) {
  69                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  70                         goto out;
  71                 }
  72                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  73                         kfree_skb(skb);
  74                         goto out;
  75                 }
  76
  77                 mdst = br_mdb_get(br, skb, vid);
  78                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  79                     br_multicast_querier_exists(br, eth_hdr(skb)))
  80                         br_multicast_flood(mdst, skb, false, true);
  81                 else
  82                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  83         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
  84                 br_forward(dst->dst, skb, false, true);
  85         } else {
  86                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
  87         }
  88 out:
  89         rcu_read_unlock();
  90         return NETDEV_TX_OK;
  91 }
  92
  93 static void br_set_lockdep_class(struct net_device *dev)
  94 {
  95         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
  96 }
  97
  98 static int br_dev_init(struct net_device *dev)
  99 {
 100         struct net_bridge *br = netdev_priv(dev);
 101         int err;
 102
 103         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 104         if (!br->stats)
 105                 return -ENOMEM;
 106
 107         err = br_vlan_init(br);
 108         if (err) {
 109                 free_percpu(br->stats);
 110                 return err;
 111         }
 112
 113         err = br_multicast_init_stats(br);
 114         if (err) {
 115                 free_percpu(br->stats);
 116                 br_vlan_flush(br);
 117         }
 118         br_set_lockdep_class(dev);
 119
 120         return err;
 121 }
 122
 123 static void br_dev_uninit(struct net_device *dev)
 124 {
 125         struct net_bridge *br = netdev_priv(dev);
 126
 127         br_multicast_dev_del(br);
 128         br_multicast_uninit_stats(br);
 129         br_vlan_flush(br);
 130         free_percpu(br->stats);
 131 }
 132
 133 static int br_dev_open(struct net_device *dev)
 134 {
 135         struct net_bridge *br = netdev_priv(dev);
 136
 137         netdev_update_features(dev);
 138         netif_start_queue(dev);
 139         br_stp_enable_bridge(br);
 140         br_multicast_open(br);
 141
 142         return 0;
 143 }
 144
 145 static void br_dev_set_multicast_list(struct net_device *dev)
 146 {
 147 }
 148
 149 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 150 {
 151         if (change & IFF_PROMISC)
 152                 br_manage_promisc(netdev_priv(dev));
 153 }
 154
 155 static int br_dev_stop(struct net_device *dev)
 156 {
 157         struct net_bridge *br = netdev_priv(dev);
 158
 159         br_stp_disable_bridge(br);
 160         br_multicast_stop(br);
 161
 162         netif_stop_queue(dev);
 163
 164         return 0;
 165 }
 166
 167 static void br_get_stats64(struct net_device *dev,
 168                            struct rtnl_link_stats64 *stats)
 169 {
 170         struct net_bridge *br = netdev_priv(dev);
 171         struct pcpu_sw_netstats tmp, sum = { 0 };
 172         unsigned int cpu;
 173
 174         for_each_possible_cpu(cpu) {
 175                 unsigned int start;
 176                 const struct pcpu_sw_netstats *bstats
 177                         = per_cpu_ptr(br->stats, cpu);
 178                 do {
 179                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 180                         memcpy(&tmp, bstats, sizeof(tmp));
 181                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 182                 sum.tx_bytes   += tmp.tx_bytes;
 183                 sum.tx_packets += tmp.tx_packets;
 184                 sum.rx_bytes   += tmp.rx_bytes;
 185                 sum.rx_packets += tmp.rx_packets;
 186         }
 187
 188         stats->tx_bytes   = sum.tx_bytes;
 189         stats->tx_packets = sum.tx_packets;
 190         stats->rx_bytes   = sum.rx_bytes;
 191         stats->rx_packets = sum.rx_packets;
 192 }
 193
 194 static int br_change_mtu(struct net_device *dev, int new_mtu)
 195 {
 196         struct net_bridge *br = netdev_priv(dev);
 197         if (new_mtu > br_min_mtu(br))
 198                 return -EINVAL;
 199
 200         dev->mtu = new_mtu;
 201
 202 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 203         /* remember the MTU in the rtable for PMTU */
 204         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 205 #endif
 206
 207         return 0;
 208 }
 209
 210 /* Allow setting mac address to any valid ethernet address. */
 211 static int br_set_mac_address(struct net_device *dev, void *p)
 212 {
 213         struct net_bridge *br = netdev_priv(dev);
 214         struct sockaddr *addr = p;
 215
 216         if (!is_valid_ether_addr(addr->sa_data))
 217                 return -EADDRNOTAVAIL;
 218
 219         spin_lock_bh(&br->lock);
 220         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 221                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 222                 br_stp_change_bridge_id(br, addr->sa_data);
 223         }
 224         spin_unlock_bh(&br->lock);
 225
 226         return 0;
 227 }
 228
 229 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 230 {
 231         strlcpy(info->driver, "bridge", sizeof(info->driver));
 232         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 233         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 234         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 235 }
 236
 237 static netdev_features_t br_fix_features(struct net_device *dev,
 238         netdev_features_t features)
 239 {
 240         struct net_bridge *br = netdev_priv(dev);
 241
 242         return br_features_recompute(br, features);
 243 }
 244
 245 #ifdef CONFIG_NET_POLL_CONTROLLER
 246 static void br_poll_controller(struct net_device *br_dev)
 247 {
 248 }
 249
 250 static void br_netpoll_cleanup(struct net_device *dev)
 251 {
 252         struct net_bridge *br = netdev_priv(dev);
 253         struct net_bridge_port *p;
 254
 255         list_for_each_entry(p, &br->port_list, list)
 256                 br_netpoll_disable(p);
 257 }
 258
 259 static int __br_netpoll_enable(struct net_bridge_port *p)
 260 {
 261         struct netpoll *np;
 262         int err;
 263
 264         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 265         if (!np)
 266                 return -ENOMEM;
 267
 268         err = __netpoll_setup(np, p->dev);
 269         if (err) {
 270                 kfree(np);
 271                 return err;
 272         }
 273
 274         p->np = np;
 275         return err;
 276 }
 277
 278 int br_netpoll_enable(struct net_bridge_port *p)
 279 {
 280         if (!p->br->dev->npinfo)
 281                 return 0;
 282
 283         return __br_netpoll_enable(p);
 284 }
 285
 286 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 287 {
 288         struct net_bridge *br = netdev_priv(dev);
 289         struct net_bridge_port *p;
 290         int err = 0;
 291
 292         list_for_each_entry(p, &br->port_list, list) {
 293                 if (!p->dev)
 294                         continue;
 295                 err = __br_netpoll_enable(p);
 296                 if (err)
 297                         goto fail;
 298         }
 299
 300 out:
 301         return err;
 302
 303 fail:
 304         br_netpoll_cleanup(dev);
 305         goto out;
 306 }
 307
 308 void br_netpoll_disable(struct net_bridge_port *p)
 309 {
 310         struct netpoll *np = p->np;
 311
 312         if (!np)
 313                 return;
 314
 315         p->np = NULL;
 316
 317         __netpoll_free_async(np);
 318 }
 319
 320 #endif
 321
 322 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev)
 323
 324 {
 325         struct net_bridge *br = netdev_priv(dev);
 326
 327         return br_add_if(br, slave_dev);
 328 }
 329
 330 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 331 {
 332         struct net_bridge *br = netdev_priv(dev);
 333
 334         return br_del_if(br, slave_dev);
 335 }
 336
 337 static const struct ethtool_ops br_ethtool_ops = {
 338         .get_drvinfo    = br_getinfo,
 339         .get_link       = ethtool_op_get_link,
 340 };
 341
 342 static const struct net_device_ops br_netdev_ops = {
 343         .ndo_open                = br_dev_open,
 344         .ndo_stop                = br_dev_stop,
 345         .ndo_init                = br_dev_init,
 346         .ndo_uninit              = br_dev_uninit,
 347         .ndo_start_xmit          = br_dev_xmit,
 348         .ndo_get_stats64         = br_get_stats64,
 349         .ndo_set_mac_address     = br_set_mac_address,
 350         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 351         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 352         .ndo_change_mtu          = br_change_mtu,
 353         .ndo_do_ioctl            = br_dev_ioctl,
 354 #ifdef CONFIG_NET_POLL_CONTROLLER
 355         .ndo_netpoll_setup       = br_netpoll_setup,
 356         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 357         .ndo_poll_controller     = br_poll_controller,
 358 #endif
 359         .ndo_add_slave           = br_add_slave,
 360         .ndo_del_slave           = br_del_slave,
 361         .ndo_fix_features        = br_fix_features,
 362         .ndo_fdb_add             = br_fdb_add,
 363         .ndo_fdb_del             = br_fdb_delete,
 364         .ndo_fdb_dump            = br_fdb_dump,
 365         .ndo_bridge_getlink      = br_getlink,
 366         .ndo_bridge_setlink      = br_setlink,
 367         .ndo_bridge_dellink      = br_dellink,
 368         .ndo_features_check      = passthru_features_check,
 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->needs_free_netdev = true;
 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         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 419 }