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         memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  45
  46         rcu_read_lock();
  47         nf_ops = rcu_dereference(nf_br_ops);
  48         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  49                 rcu_read_unlock();
  50                 return NETDEV_TX_OK;
  51         }
  52
  53         u64_stats_update_begin(&brstats->syncp);
  54         brstats->tx_packets++;
  55         brstats->tx_bytes += skb->len;
  56         u64_stats_update_end(&brstats->syncp);
  57
  58         br_switchdev_frame_unmark(skb);
  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         if (IS_ENABLED(CONFIG_INET) &&
  68             (eth_hdr(skb)->h_proto == htons(ETH_P_ARP) ||
  69              eth_hdr(skb)->h_proto == htons(ETH_P_RARP)) &&
  70             br->neigh_suppress_enabled) {
  71                 br_do_proxy_suppress_arp(skb, br, vid, NULL);
  72         } else if (IS_ENABLED(CONFIG_IPV6) &&
  73                    skb->protocol == htons(ETH_P_IPV6) &&
  74                    br->neigh_suppress_enabled &&
  75                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
  76                                  sizeof(struct nd_msg)) &&
  77                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
  78                         struct nd_msg *msg, _msg;
  79
  80                         msg = br_is_nd_neigh_msg(skb, &_msg);
  81                         if (msg)
  82                                 br_do_suppress_nd(skb, br, vid, NULL, msg);
  83         }
  84
  85         dest = eth_hdr(skb)->h_dest;
  86         if (is_broadcast_ether_addr(dest)) {
  87                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  88         } else if (is_multicast_ether_addr(dest)) {
  89                 if (unlikely(netpoll_tx_running(dev))) {
  90                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  91                         goto out;
  92                 }
  93                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  94                         kfree_skb(skb);
  95                         goto out;
  96                 }
  97
  98                 mdst = br_mdb_get(br, skb, vid);
  99                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
 100                     br_multicast_querier_exists(br, eth_hdr(skb)))
 101                         br_multicast_flood(mdst, skb, false, true);
 102                 else
 103                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
 104         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
 105                 br_forward(dst->dst, skb, false, true);
 106         } else {
 107                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
 108         }
 109 out:
 110         rcu_read_unlock();
 111         return NETDEV_TX_OK;
 112 }
 113
 114 static void br_set_lockdep_class(struct net_device *dev)
 115 {
 116         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
 117 }
 118
 119 static int br_dev_init(struct net_device *dev)
 120 {
 121         struct net_bridge *br = netdev_priv(dev);
 122         int err;
 123
 124         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 125         if (!br->stats)
 126                 return -ENOMEM;
 127
 128         err = br_fdb_hash_init(br);
 129         if (err) {
 130                 free_percpu(br->stats);
 131                 return err;
 132         }
 133
 134         err = br_vlan_init(br);
 135         if (err) {
 136                 free_percpu(br->stats);
 137                 br_fdb_hash_fini(br);
 138                 return err;
 139         }
 140
 141         err = br_multicast_init_stats(br);
 142         if (err) {
 143                 free_percpu(br->stats);
 144                 br_vlan_flush(br);
 145                 br_fdb_hash_fini(br);
 146         }
 147         br_set_lockdep_class(dev);
 148
 149         return err;
 150 }
 151
 152 static void br_dev_uninit(struct net_device *dev)
 153 {
 154         struct net_bridge *br = netdev_priv(dev);
 155
 156         br_multicast_dev_del(br);
 157         br_multicast_uninit_stats(br);
 158         br_vlan_flush(br);
 159         br_fdb_hash_fini(br);
 160         free_percpu(br->stats);
 161 }
 162
 163 static int br_dev_open(struct net_device *dev)
 164 {
 165         struct net_bridge *br = netdev_priv(dev);
 166
 167         netdev_update_features(dev);
 168         netif_start_queue(dev);
 169         br_stp_enable_bridge(br);
 170         br_multicast_open(br);
 171
 172         return 0;
 173 }
 174
 175 static void br_dev_set_multicast_list(struct net_device *dev)
 176 {
 177 }
 178
 179 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 180 {
 181         if (change & IFF_PROMISC)
 182                 br_manage_promisc(netdev_priv(dev));
 183 }
 184
 185 static int br_dev_stop(struct net_device *dev)
 186 {
 187         struct net_bridge *br = netdev_priv(dev);
 188
 189         br_stp_disable_bridge(br);
 190         br_multicast_stop(br);
 191
 192         netif_stop_queue(dev);
 193
 194         return 0;
 195 }
 196
 197 static void br_get_stats64(struct net_device *dev,
 198                            struct rtnl_link_stats64 *stats)
 199 {
 200         struct net_bridge *br = netdev_priv(dev);
 201         struct pcpu_sw_netstats tmp, sum = { 0 };
 202         unsigned int cpu;
 203
 204         for_each_possible_cpu(cpu) {
 205                 unsigned int start;
 206                 const struct pcpu_sw_netstats *bstats
 207                         = per_cpu_ptr(br->stats, cpu);
 208                 do {
 209                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 210                         memcpy(&tmp, bstats, sizeof(tmp));
 211                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 212                 sum.tx_bytes   += tmp.tx_bytes;
 213                 sum.tx_packets += tmp.tx_packets;
 214                 sum.rx_bytes   += tmp.rx_bytes;
 215                 sum.rx_packets += tmp.rx_packets;
 216         }
 217
 218         netdev_stats_to_stats64(stats, &dev->stats);
 219         stats->tx_bytes   = sum.tx_bytes;
 220         stats->tx_packets = sum.tx_packets;
 221         stats->rx_bytes   = sum.rx_bytes;
 222         stats->rx_packets = sum.rx_packets;
 223 }
 224
 225 static int br_change_mtu(struct net_device *dev, int new_mtu)
 226 {
 227         struct net_bridge *br = netdev_priv(dev);
 228
 229         dev->mtu = new_mtu;
 230
 231         /* this flag will be cleared if the MTU was automatically adjusted */
 232         br->mtu_set_by_user = true;
 233 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 234         /* remember the MTU in the rtable for PMTU */
 235         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 236 #endif
 237
 238         return 0;
 239 }
 240
 241 /* Allow setting mac address to any valid ethernet address. */
 242 static int br_set_mac_address(struct net_device *dev, void *p)
 243 {
 244         struct net_bridge *br = netdev_priv(dev);
 245         struct sockaddr *addr = p;
 246
 247         if (!is_valid_ether_addr(addr->sa_data))
 248                 return -EADDRNOTAVAIL;
 249
 250         /* dev_set_mac_addr() can be called by a master device on bridge's
 251          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 252          */
 253         if (dev->reg_state != NETREG_REGISTERED)
 254                 return -EBUSY;
 255
 256         spin_lock_bh(&br->lock);
 257         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 258                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 259                 br_stp_change_bridge_id(br, addr->sa_data);
 260         }
 261         spin_unlock_bh(&br->lock);
 262
 263         return 0;
 264 }
 265
 266 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 267 {
 268         strlcpy(info->driver, "bridge", sizeof(info->driver));
 269         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 270         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 271         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 272 }
 273
 274 static netdev_features_t br_fix_features(struct net_device *dev,
 275         netdev_features_t features)
 276 {
 277         struct net_bridge *br = netdev_priv(dev);
 278
 279         return br_features_recompute(br, features);
 280 }
 281
 282 #ifdef CONFIG_NET_POLL_CONTROLLER
 283 static void br_poll_controller(struct net_device *br_dev)
 284 {
 285 }
 286
 287 static void br_netpoll_cleanup(struct net_device *dev)
 288 {
 289         struct net_bridge *br = netdev_priv(dev);
 290         struct net_bridge_port *p;
 291
 292         list_for_each_entry(p, &br->port_list, list)
 293                 br_netpoll_disable(p);
 294 }
 295
 296 static int __br_netpoll_enable(struct net_bridge_port *p)
 297 {
 298         struct netpoll *np;
 299         int err;
 300
 301         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 302         if (!np)
 303                 return -ENOMEM;
 304
 305         err = __netpoll_setup(np, p->dev);
 306         if (err) {
 307                 kfree(np);
 308                 return err;
 309         }
 310
 311         p->np = np;
 312         return err;
 313 }
 314
 315 int br_netpoll_enable(struct net_bridge_port *p)
 316 {
 317         if (!p->br->dev->npinfo)
 318                 return 0;
 319
 320         return __br_netpoll_enable(p);
 321 }
 322
 323 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 324 {
 325         struct net_bridge *br = netdev_priv(dev);
 326         struct net_bridge_port *p;
 327         int err = 0;
 328
 329         list_for_each_entry(p, &br->port_list, list) {
 330                 if (!p->dev)
 331                         continue;
 332                 err = __br_netpoll_enable(p);
 333                 if (err)
 334                         goto fail;
 335         }
 336
 337 out:
 338         return err;
 339
 340 fail:
 341         br_netpoll_cleanup(dev);
 342         goto out;
 343 }
 344
 345 void br_netpoll_disable(struct net_bridge_port *p)
 346 {
 347         struct netpoll *np = p->np;
 348
 349         if (!np)
 350                 return;
 351
 352         p->np = NULL;
 353
 354         __netpoll_free_async(np);
 355 }
 356
 357 #endif
 358
 359 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 360                         struct netlink_ext_ack *extack)
 361
 362 {
 363         struct net_bridge *br = netdev_priv(dev);
 364
 365         return br_add_if(br, slave_dev, extack);
 366 }
 367
 368 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 369 {
 370         struct net_bridge *br = netdev_priv(dev);
 371
 372         return br_del_if(br, slave_dev);
 373 }
 374
 375 static const struct ethtool_ops br_ethtool_ops = {
 376         .get_drvinfo    = br_getinfo,
 377         .get_link       = ethtool_op_get_link,
 378 };
 379
 380 static const struct net_device_ops br_netdev_ops = {
 381         .ndo_open                = br_dev_open,
 382         .ndo_stop                = br_dev_stop,
 383         .ndo_init                = br_dev_init,
 384         .ndo_uninit              = br_dev_uninit,
 385         .ndo_start_xmit          = br_dev_xmit,
 386         .ndo_get_stats64         = br_get_stats64,
 387         .ndo_set_mac_address     = br_set_mac_address,
 388         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 389         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 390         .ndo_change_mtu          = br_change_mtu,
 391         .ndo_do_ioctl            = br_dev_ioctl,
 392 #ifdef CONFIG_NET_POLL_CONTROLLER
 393         .ndo_netpoll_setup       = br_netpoll_setup,
 394         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 395         .ndo_poll_controller     = br_poll_controller,
 396 #endif
 397         .ndo_add_slave           = br_add_slave,
 398         .ndo_del_slave           = br_del_slave,
 399         .ndo_fix_features        = br_fix_features,
 400         .ndo_fdb_add             = br_fdb_add,
 401         .ndo_fdb_del             = br_fdb_delete,
 402         .ndo_fdb_dump            = br_fdb_dump,
 403         .ndo_bridge_getlink      = br_getlink,
 404         .ndo_bridge_setlink      = br_setlink,
 405         .ndo_bridge_dellink      = br_dellink,
 406         .ndo_features_check      = passthru_features_check,
 407 };
 408
 409 static struct device_type br_type = {
 410         .name   = "bridge",
 411 };
 412
 413 void br_dev_setup(struct net_device *dev)
 414 {
 415         struct net_bridge *br = netdev_priv(dev);
 416
 417         eth_hw_addr_random(dev);
 418         ether_setup(dev);
 419
 420         dev->netdev_ops = &br_netdev_ops;
 421         dev->needs_free_netdev = true;
 422         dev->ethtool_ops = &br_ethtool_ops;
 423         SET_NETDEV_DEVTYPE(dev, &br_type);
 424         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 425
 426         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 427                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 428         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 429                            NETIF_F_HW_VLAN_STAG_TX;
 430         dev->vlan_features = COMMON_FEATURES;
 431
 432         br->dev = dev;
 433         spin_lock_init(&br->lock);
 434         INIT_LIST_HEAD(&br->port_list);
 435         INIT_HLIST_HEAD(&br->fdb_list);
 436         spin_lock_init(&br->hash_lock);
 437
 438         br->bridge_id.prio[0] = 0x80;
 439         br->bridge_id.prio[1] = 0x00;
 440
 441         ether_addr_copy(br->group_addr, eth_stp_addr);
 442
 443         br->stp_enabled = BR_NO_STP;
 444         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 445         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 446
 447         br->designated_root = br->bridge_id;
 448         br->bridge_max_age = br->max_age = 20 * HZ;
 449         br->bridge_hello_time = br->hello_time = 2 * HZ;
 450         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 451         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 452         dev->max_mtu = ETH_MAX_MTU;
 453
 454         br_netfilter_rtable_init(br);
 455         br_stp_timer_init(br);
 456         br_multicast_init(br);
 457         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 458 }