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         BR_INPUT_SKB_CB(skb)->brdev = dev;
  52
  53         skb_reset_mac_header(skb);
  54         skb_pull(skb, ETH_HLEN);
  55
  56         u64_stats_update_begin(&brstats->syncp);
  57         brstats->tx_packets++;
  58         /* Exclude ETH_HLEN from byte stats for consistency with Rx chain */
  59         brstats->tx_bytes += skb->len;
  60         u64_stats_update_end(&brstats->syncp);
  61
  62         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid))
  63                 goto out;
  64
  65         if (is_broadcast_ether_addr(dest))
  66                 br_flood_deliver(br, skb, false);
  67         else if (is_multicast_ether_addr(dest)) {
  68                 if (unlikely(netpoll_tx_running(dev))) {
  69                         br_flood_deliver(br, skb, false);
  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_deliver(mdst, skb);
  81                 else
  82                         br_flood_deliver(br, skb, false);
  83         } else if ((dst = __br_fdb_get(br, dest, vid)) != NULL)
  84                 br_deliver(dst->dst, skb);
  85         else
  86                 br_flood_deliver(br, skb, 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         br_set_lockdep_class(dev);
 111
 112         return err;
 113 }
 114
 115 static int br_dev_open(struct net_device *dev)
 116 {
 117         struct net_bridge *br = netdev_priv(dev);
 118
 119         netdev_update_features(dev);
 120         netif_start_queue(dev);
 121         br_stp_enable_bridge(br);
 122         br_multicast_open(br);
 123
 124         return 0;
 125 }
 126
 127 static void br_dev_set_multicast_list(struct net_device *dev)
 128 {
 129 }
 130
 131 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 132 {
 133         if (change & IFF_PROMISC)
 134                 br_manage_promisc(netdev_priv(dev));
 135 }
 136
 137 static int br_dev_stop(struct net_device *dev)
 138 {
 139         struct net_bridge *br = netdev_priv(dev);
 140
 141         br_stp_disable_bridge(br);
 142         br_multicast_stop(br);
 143
 144         netif_stop_queue(dev);
 145
 146         return 0;
 147 }
 148
 149 static struct rtnl_link_stats64 *br_get_stats64(struct net_device *dev,
 150                                                 struct rtnl_link_stats64 *stats)
 151 {
 152         struct net_bridge *br = netdev_priv(dev);
 153         struct pcpu_sw_netstats tmp, sum = { 0 };
 154         unsigned int cpu;
 155
 156         for_each_possible_cpu(cpu) {
 157                 unsigned int start;
 158                 const struct pcpu_sw_netstats *bstats
 159                         = per_cpu_ptr(br->stats, cpu);
 160                 do {
 161                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 162                         memcpy(&tmp, bstats, sizeof(tmp));
 163                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 164                 sum.tx_bytes   += tmp.tx_bytes;
 165                 sum.tx_packets += tmp.tx_packets;
 166                 sum.rx_bytes   += tmp.rx_bytes;
 167                 sum.rx_packets += tmp.rx_packets;
 168         }
 169
 170         stats->tx_bytes   = sum.tx_bytes;
 171         stats->tx_packets = sum.tx_packets;
 172         stats->rx_bytes   = sum.rx_bytes;
 173         stats->rx_packets = sum.rx_packets;
 174
 175         return stats;
 176 }
 177
 178 static int br_change_mtu(struct net_device *dev, int new_mtu)
 179 {
 180         struct net_bridge *br = netdev_priv(dev);
 181         if (new_mtu < 68 || new_mtu > br_min_mtu(br))
 182                 return -EINVAL;
 183
 184         dev->mtu = new_mtu;
 185
 186 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 187         /* remember the MTU in the rtable for PMTU */
 188         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 189 #endif
 190
 191         return 0;
 192 }
 193
 194 /* Allow setting mac address to any valid ethernet address. */
 195 static int br_set_mac_address(struct net_device *dev, void *p)
 196 {
 197         struct net_bridge *br = netdev_priv(dev);
 198         struct sockaddr *addr = p;
 199
 200         if (!is_valid_ether_addr(addr->sa_data))
 201                 return -EADDRNOTAVAIL;
 202
 203         spin_lock_bh(&br->lock);
 204         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 205                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 206                 br_stp_change_bridge_id(br, addr->sa_data);
 207         }
 208         spin_unlock_bh(&br->lock);
 209
 210         return 0;
 211 }
 212
 213 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 214 {
 215         strlcpy(info->driver, "bridge", sizeof(info->driver));
 216         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 217         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 218         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 219 }
 220
 221 static netdev_features_t br_fix_features(struct net_device *dev,
 222         netdev_features_t features)
 223 {
 224         struct net_bridge *br = netdev_priv(dev);
 225
 226         return br_features_recompute(br, features);
 227 }
 228
 229 #ifdef CONFIG_NET_POLL_CONTROLLER
 230 static void br_poll_controller(struct net_device *br_dev)
 231 {
 232 }
 233
 234 static void br_netpoll_cleanup(struct net_device *dev)
 235 {
 236         struct net_bridge *br = netdev_priv(dev);
 237         struct net_bridge_port *p;
 238
 239         list_for_each_entry(p, &br->port_list, list)
 240                 br_netpoll_disable(p);
 241 }
 242
 243 static int __br_netpoll_enable(struct net_bridge_port *p)
 244 {
 245         struct netpoll *np;
 246         int err;
 247
 248         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 249         if (!np)
 250                 return -ENOMEM;
 251
 252         err = __netpoll_setup(np, p->dev);
 253         if (err) {
 254                 kfree(np);
 255                 return err;
 256         }
 257
 258         p->np = np;
 259         return err;
 260 }
 261
 262 int br_netpoll_enable(struct net_bridge_port *p)
 263 {
 264         if (!p->br->dev->npinfo)
 265                 return 0;
 266
 267         return __br_netpoll_enable(p);
 268 }
 269
 270 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 271 {
 272         struct net_bridge *br = netdev_priv(dev);
 273         struct net_bridge_port *p;
 274         int err = 0;
 275
 276         list_for_each_entry(p, &br->port_list, list) {
 277                 if (!p->dev)
 278                         continue;
 279                 err = __br_netpoll_enable(p);
 280                 if (err)
 281                         goto fail;
 282         }
 283
 284 out:
 285         return err;
 286
 287 fail:
 288         br_netpoll_cleanup(dev);
 289         goto out;
 290 }
 291
 292 void br_netpoll_disable(struct net_bridge_port *p)
 293 {
 294         struct netpoll *np = p->np;
 295
 296         if (!np)
 297                 return;
 298
 299         p->np = NULL;
 300
 301         __netpoll_free_async(np);
 302 }
 303
 304 #endif
 305
 306 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev)
 307
 308 {
 309         struct net_bridge *br = netdev_priv(dev);
 310
 311         return br_add_if(br, slave_dev);
 312 }
 313
 314 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 315 {
 316         struct net_bridge *br = netdev_priv(dev);
 317
 318         return br_del_if(br, slave_dev);
 319 }
 320
 321 static const struct ethtool_ops br_ethtool_ops = {
 322         .get_drvinfo    = br_getinfo,
 323         .get_link       = ethtool_op_get_link,
 324 };
 325
 326 static const struct net_device_ops br_netdev_ops = {
 327         .ndo_open                = br_dev_open,
 328         .ndo_stop                = br_dev_stop,
 329         .ndo_init                = br_dev_init,
 330         .ndo_start_xmit          = br_dev_xmit,
 331         .ndo_get_stats64         = br_get_stats64,
 332         .ndo_set_mac_address     = br_set_mac_address,
 333         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 334         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 335         .ndo_change_mtu          = br_change_mtu,
 336         .ndo_do_ioctl            = br_dev_ioctl,
 337 #ifdef CONFIG_NET_POLL_CONTROLLER
 338         .ndo_netpoll_setup       = br_netpoll_setup,
 339         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 340         .ndo_poll_controller     = br_poll_controller,
 341 #endif
 342         .ndo_add_slave           = br_add_slave,
 343         .ndo_del_slave           = br_del_slave,
 344         .ndo_fix_features        = br_fix_features,
 345         .ndo_fdb_add             = br_fdb_add,
 346         .ndo_fdb_del             = br_fdb_delete,
 347         .ndo_fdb_dump            = br_fdb_dump,
 348         .ndo_bridge_getlink      = br_getlink,
 349         .ndo_bridge_setlink      = br_setlink,
 350         .ndo_bridge_dellink      = br_dellink,
 351         .ndo_features_check      = passthru_features_check,
 352 };
 353
 354 static void br_dev_free(struct net_device *dev)
 355 {
 356         struct net_bridge *br = netdev_priv(dev);
 357
 358         free_percpu(br->stats);
 359         free_netdev(dev);
 360 }
 361
 362 static struct device_type br_type = {
 363         .name   = "bridge",
 364 };
 365
 366 void br_dev_setup(struct net_device *dev)
 367 {
 368         struct net_bridge *br = netdev_priv(dev);
 369
 370         eth_hw_addr_random(dev);
 371         ether_setup(dev);
 372
 373         dev->netdev_ops = &br_netdev_ops;
 374         dev->destructor = br_dev_free;
 375         dev->ethtool_ops = &br_ethtool_ops;
 376         SET_NETDEV_DEVTYPE(dev, &br_type);
 377         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 378
 379         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 380                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 381         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 382                            NETIF_F_HW_VLAN_STAG_TX;
 383         dev->vlan_features = COMMON_FEATURES;
 384
 385         br->dev = dev;
 386         spin_lock_init(&br->lock);
 387         INIT_LIST_HEAD(&br->port_list);
 388         spin_lock_init(&br->hash_lock);
 389
 390         br->bridge_id.prio[0] = 0x80;
 391         br->bridge_id.prio[1] = 0x00;
 392
 393         ether_addr_copy(br->group_addr, eth_reserved_addr_base);
 394
 395         br->stp_enabled = BR_NO_STP;
 396         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 397         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 398
 399         br->designated_root = br->bridge_id;
 400         br->bridge_max_age = br->max_age = 20 * HZ;
 401         br->bridge_hello_time = br->hello_time = 2 * HZ;
 402         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 403         br->ageing_time = BR_DEFAULT_AGEING_TIME;
 404
 405         br_netfilter_rtable_init(br);
 406         br_stp_timer_init(br);
 407         br_multicast_init(br);
 408 }