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