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 #ifdef CONFIG_NET_SWITCHDEV
  57         skb->offload_fwd_mark = 0;
  58 #endif
  59         BR_INPUT_SKB_CB(skb)->brdev = dev;
  60
  61         skb_reset_mac_header(skb);
  62         skb_pull(skb, ETH_HLEN);
  63
  64         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid))
  65                 goto out;
  66
  67         dest = eth_hdr(skb)->h_dest;
  68         if (is_broadcast_ether_addr(dest)) {
  69                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  70         } else if (is_multicast_ether_addr(dest)) {
  71                 if (unlikely(netpoll_tx_running(dev))) {
  72                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  73                         goto out;
  74                 }
  75                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  76                         kfree_skb(skb);
  77                         goto out;
  78                 }
  79
  80                 mdst = br_mdb_get(br, skb, vid);
  81                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  82                     br_multicast_querier_exists(br, eth_hdr(skb)))
  83                         br_multicast_flood(mdst, skb, false, true);
  84                 else
  85                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  86         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
  87                 br_forward(dst->dst, skb, false, true);
  88         } else {
  89                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
  90         }
  91 out:
  92         rcu_read_unlock();
  93         return NETDEV_TX_OK;
  94 }
  95
  96 static void br_set_lockdep_class(struct net_device *dev)
  97 {
  98         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
  99 }
 100
 101 static int br_dev_init(struct net_device *dev)
 102 {
 103         struct net_bridge *br = netdev_priv(dev);
 104         int err;
 105
 106         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 107         if (!br->stats)
 108                 return -ENOMEM;
 109
 110         err = br_vlan_init(br);
 111         if (err) {
 112                 free_percpu(br->stats);
 113                 return err;
 114         }
 115
 116         err = br_multicast_init_stats(br);
 117         if (err) {
 118                 free_percpu(br->stats);
 119                 br_vlan_flush(br);
 120         }
 121         br_set_lockdep_class(dev);
 122
 123         return err;
 124 }
 125
 126 static void br_dev_uninit(struct net_device *dev)
 127 {
 128         struct net_bridge *br = netdev_priv(dev);
 129
 130         br_multicast_dev_del(br);
 131         br_multicast_uninit_stats(br);
 132         br_vlan_flush(br);
 133         free_percpu(br->stats);
 134 }
 135
 136 static int br_dev_open(struct net_device *dev)
 137 {
 138         struct net_bridge *br = netdev_priv(dev);
 139
 140         netdev_update_features(dev);
 141         netif_start_queue(dev);
 142         br_stp_enable_bridge(br);
 143         br_multicast_open(br);
 144
 145         return 0;
 146 }
 147
 148 static void br_dev_set_multicast_list(struct net_device *dev)
 149 {
 150 }
 151
 152 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 153 {
 154         if (change & IFF_PROMISC)
 155                 br_manage_promisc(netdev_priv(dev));
 156 }
 157
 158 static int br_dev_stop(struct net_device *dev)
 159 {
 160         struct net_bridge *br = netdev_priv(dev);
 161
 162         br_stp_disable_bridge(br);
 163         br_multicast_stop(br);
 164
 165         netif_stop_queue(dev);
 166
 167         return 0;
 168 }
 169
 170 static void br_get_stats64(struct net_device *dev,
 171                            struct rtnl_link_stats64 *stats)
 172 {
 173         struct net_bridge *br = netdev_priv(dev);
 174         struct pcpu_sw_netstats tmp, sum = { 0 };
 175         unsigned int cpu;
 176
 177         for_each_possible_cpu(cpu) {
 178                 unsigned int start;
 179                 const struct pcpu_sw_netstats *bstats
 180                         = per_cpu_ptr(br->stats, cpu);
 181                 do {
 182                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 183                         memcpy(&tmp, bstats, sizeof(tmp));
 184                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 185                 sum.tx_bytes   += tmp.tx_bytes;
 186                 sum.tx_packets += tmp.tx_packets;
 187                 sum.rx_bytes   += tmp.rx_bytes;
 188                 sum.rx_packets += tmp.rx_packets;
 189         }
 190
 191         stats->tx_bytes   = sum.tx_bytes;
 192         stats->tx_packets = sum.tx_packets;
 193         stats->rx_bytes   = sum.rx_bytes;
 194         stats->rx_packets = sum.rx_packets;
 195 }
 196
 197 static int br_change_mtu(struct net_device *dev, int new_mtu)
 198 {
 199         struct net_bridge *br = netdev_priv(dev);
 200         if (new_mtu > br_min_mtu(br))
 201                 return -EINVAL;
 202
 203         dev->mtu = new_mtu;
 204
 205 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 206         /* remember the MTU in the rtable for PMTU */
 207         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 208 #endif
 209
 210         return 0;
 211 }
 212
 213 /* Allow setting mac address to any valid ethernet address. */
 214 static int br_set_mac_address(struct net_device *dev, void *p)
 215 {
 216         struct net_bridge *br = netdev_priv(dev);
 217         struct sockaddr *addr = p;
 218
 219         if (!is_valid_ether_addr(addr->sa_data))
 220                 return -EADDRNOTAVAIL;
 221
 222         spin_lock_bh(&br->lock);
 223         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 224                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 225                 br_stp_change_bridge_id(br, addr->sa_data);
 226         }
 227         spin_unlock_bh(&br->lock);
 228
 229         return 0;
 230 }
 231
 232 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 233 {
 234         strlcpy(info->driver, "bridge", sizeof(info->driver));
 235         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 236         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 237         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 238 }
 239
 240 static netdev_features_t br_fix_features(struct net_device *dev,
 241         netdev_features_t features)
 242 {
 243         struct net_bridge *br = netdev_priv(dev);
 244
 245         return br_features_recompute(br, features);
 246 }
 247
 248 #ifdef CONFIG_NET_POLL_CONTROLLER
 249 static void br_poll_controller(struct net_device *br_dev)
 250 {
 251 }
 252
 253 static void br_netpoll_cleanup(struct net_device *dev)
 254 {
 255         struct net_bridge *br = netdev_priv(dev);
 256         struct net_bridge_port *p;
 257
 258         list_for_each_entry(p, &br->port_list, list)
 259                 br_netpoll_disable(p);
 260 }
 261
 262 static int __br_netpoll_enable(struct net_bridge_port *p)
 263 {
 264         struct netpoll *np;
 265         int err;
 266
 267         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 268         if (!np)
 269                 return -ENOMEM;
 270
 271         err = __netpoll_setup(np, p->dev);
 272         if (err) {
 273                 kfree(np);
 274                 return err;
 275         }
 276
 277         p->np = np;
 278         return err;
 279 }
 280
 281 int br_netpoll_enable(struct net_bridge_port *p)
 282 {
 283         if (!p->br->dev->npinfo)
 284                 return 0;
 285
 286         return __br_netpoll_enable(p);
 287 }
 288
 289 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 290 {
 291         struct net_bridge *br = netdev_priv(dev);
 292         struct net_bridge_port *p;
 293         int err = 0;
 294
 295         list_for_each_entry(p, &br->port_list, list) {
 296                 if (!p->dev)
 297                         continue;
 298                 err = __br_netpoll_enable(p);
 299                 if (err)
 300                         goto fail;
 301         }
 302
 303 out:
 304         return err;
 305
 306 fail:
 307         br_netpoll_cleanup(dev);
 308         goto out;
 309 }
 310
 311 void br_netpoll_disable(struct net_bridge_port *p)
 312 {
 313         struct netpoll *np = p->np;
 314
 315         if (!np)
 316                 return;
 317
 318         p->np = NULL;
 319
 320         __netpoll_free_async(np);
 321 }
 322
 323 #endif
 324
 325 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev)
 326
 327 {
 328         struct net_bridge *br = netdev_priv(dev);
 329
 330         return br_add_if(br, slave_dev);
 331 }
 332
 333 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 334 {
 335         struct net_bridge *br = netdev_priv(dev);
 336
 337         return br_del_if(br, slave_dev);
 338 }
 339
 340 static const struct ethtool_ops br_ethtool_ops = {
 341         .get_drvinfo    = br_getinfo,
 342         .get_link       = ethtool_op_get_link,
 343 };
 344
 345 static const struct net_device_ops br_netdev_ops = {
 346         .ndo_open                = br_dev_open,
 347         .ndo_stop                = br_dev_stop,
 348         .ndo_init                = br_dev_init,
 349         .ndo_uninit              = br_dev_uninit,
 350         .ndo_start_xmit          = br_dev_xmit,
 351         .ndo_get_stats64         = br_get_stats64,
 352         .ndo_set_mac_address     = br_set_mac_address,
 353         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 354         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 355         .ndo_change_mtu          = br_change_mtu,
 356         .ndo_do_ioctl            = br_dev_ioctl,
 357 #ifdef CONFIG_NET_POLL_CONTROLLER
 358         .ndo_netpoll_setup       = br_netpoll_setup,
 359         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 360         .ndo_poll_controller     = br_poll_controller,
 361 #endif
 362         .ndo_add_slave           = br_add_slave,
 363         .ndo_del_slave           = br_del_slave,
 364         .ndo_fix_features        = br_fix_features,
 365         .ndo_fdb_add             = br_fdb_add,
 366         .ndo_fdb_del             = br_fdb_delete,
 367         .ndo_fdb_dump            = br_fdb_dump,
 368         .ndo_bridge_getlink      = br_getlink,
 369         .ndo_bridge_setlink      = br_setlink,
 370         .ndo_bridge_dellink      = br_dellink,
 371         .ndo_features_check      = passthru_features_check,
 372 };
 373
 374 static struct device_type br_type = {
 375         .name   = "bridge",
 376 };
 377
 378 void br_dev_setup(struct net_device *dev)
 379 {
 380         struct net_bridge *br = netdev_priv(dev);
 381
 382         eth_hw_addr_random(dev);
 383         ether_setup(dev);
 384
 385         dev->netdev_ops = &br_netdev_ops;
 386         dev->needs_free_netdev = true;
 387         dev->ethtool_ops = &br_ethtool_ops;
 388         SET_NETDEV_DEVTYPE(dev, &br_type);
 389         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 390
 391         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 392                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 393         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 394                            NETIF_F_HW_VLAN_STAG_TX;
 395         dev->vlan_features = COMMON_FEATURES;
 396
 397         br->dev = dev;
 398         spin_lock_init(&br->lock);
 399         INIT_LIST_HEAD(&br->port_list);
 400         spin_lock_init(&br->hash_lock);
 401
 402         br->bridge_id.prio[0] = 0x80;
 403         br->bridge_id.prio[1] = 0x00;
 404
 405         ether_addr_copy(br->group_addr, eth_reserved_addr_base);
 406
 407         br->stp_enabled = BR_NO_STP;
 408         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 409         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 410
 411         br->designated_root = br->bridge_id;
 412         br->bridge_max_age = br->max_age = 20 * HZ;
 413         br->bridge_hello_time = br->hello_time = 2 * HZ;
 414         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 415         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 416         dev->max_mtu = ETH_MAX_MTU;
 417
 418         br_netfilter_rtable_init(br);
 419         br_stp_timer_init(br);
 420         br_multicast_init(br);
 421         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 422 }