net/bridge/br_device.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      Device handling code
   4  *      Linux ethernet bridge
   5  *
   6  *      Authors:
   7  *      Lennert Buytenhek               <buytenh@gnu.org>
   8  */
   9
  10 #include <linux/kernel.h>
  11 #include <linux/netdevice.h>
  12 #include <linux/netpoll.h>
  13 #include <linux/etherdevice.h>
  14 #include <linux/ethtool.h>
  15 #include <linux/list.h>
  16 #include <linux/netfilter_bridge.h>
  17
  18 #include <linux/uaccess.h>
  19 #include "br_private.h"
  20
  21 #define COMMON_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA | \
  22                          NETIF_F_GSO_MASK | NETIF_F_HW_CSUM)
  23
  24 const struct nf_br_ops __rcu *nf_br_ops __read_mostly;
  25 EXPORT_SYMBOL_GPL(nf_br_ops);
  26
  27 /* net device transmit always called with BH disabled */
  28 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev)
  29 {
  30         struct net_bridge *br = netdev_priv(dev);
  31         struct net_bridge_fdb_entry *dst;
  32         struct net_bridge_mdb_entry *mdst;
  33         struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
  34         const struct nf_br_ops *nf_ops;
  35         const unsigned char *dest;
  36         u16 vid = 0;
  37
  38         rcu_read_lock();
  39         nf_ops = rcu_dereference(nf_br_ops);
  40         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  41                 rcu_read_unlock();
  42                 return NETDEV_TX_OK;
  43         }
  44
  45         u64_stats_update_begin(&brstats->syncp);
  46         brstats->tx_packets++;
  47         brstats->tx_bytes += skb->len;
  48         u64_stats_update_end(&brstats->syncp);
  49
  50         br_switchdev_frame_unmark(skb);
  51         BR_INPUT_SKB_CB(skb)->brdev = dev;
  52         BR_INPUT_SKB_CB(skb)->frag_max_size = 0;
  53
  54         skb_reset_mac_header(skb);
  55         skb_pull(skb, ETH_HLEN);
  56
  57         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid))
  58                 goto out;
  59
  60         if (IS_ENABLED(CONFIG_INET) &&
  61             (eth_hdr(skb)->h_proto == htons(ETH_P_ARP) ||
  62              eth_hdr(skb)->h_proto == htons(ETH_P_RARP)) &&
  63             br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
  64                 br_do_proxy_suppress_arp(skb, br, vid, NULL);
  65         } else if (IS_ENABLED(CONFIG_IPV6) &&
  66                    skb->protocol == htons(ETH_P_IPV6) &&
  67                    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
  68                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
  69                                  sizeof(struct nd_msg)) &&
  70                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
  71                         struct nd_msg *msg, _msg;
  72
  73                         msg = br_is_nd_neigh_msg(skb, &_msg);
  74                         if (msg)
  75                                 br_do_suppress_nd(skb, br, vid, NULL, msg);
  76         }
  77
  78         dest = eth_hdr(skb)->h_dest;
  79         if (is_broadcast_ether_addr(dest)) {
  80                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  81         } else if (is_multicast_ether_addr(dest)) {
  82                 if (unlikely(netpoll_tx_running(dev))) {
  83                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  84                         goto out;
  85                 }
  86                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  87                         kfree_skb(skb);
  88                         goto out;
  89                 }
  90
  91                 mdst = br_mdb_get(br, skb, vid);
  92                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  93                     br_multicast_querier_exists(br, eth_hdr(skb)))
  94                         br_multicast_flood(mdst, skb, false, true);
  95                 else
  96                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  97         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
  98                 br_forward(dst->dst, skb, false, true);
  99         } else {
 100                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
 101         }
 102 out:
 103         rcu_read_unlock();
 104         return NETDEV_TX_OK;
 105 }
 106
 107 static int br_dev_init(struct net_device *dev)
 108 {
 109         struct net_bridge *br = netdev_priv(dev);
 110         int err;
 111
 112         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 113         if (!br->stats)
 114                 return -ENOMEM;
 115
 116         err = br_fdb_hash_init(br);
 117         if (err) {
 118                 free_percpu(br->stats);
 119                 return err;
 120         }
 121
 122         err = br_mdb_hash_init(br);
 123         if (err) {
 124                 free_percpu(br->stats);
 125                 br_fdb_hash_fini(br);
 126                 return err;
 127         }
 128
 129         err = br_vlan_init(br);
 130         if (err) {
 131                 free_percpu(br->stats);
 132                 br_mdb_hash_fini(br);
 133                 br_fdb_hash_fini(br);
 134                 return err;
 135         }
 136
 137         err = br_multicast_init_stats(br);
 138         if (err) {
 139                 free_percpu(br->stats);
 140                 br_vlan_flush(br);
 141                 br_mdb_hash_fini(br);
 142                 br_fdb_hash_fini(br);
 143         }
 144
 145         return err;
 146 }
 147
 148 static void br_dev_uninit(struct net_device *dev)
 149 {
 150         struct net_bridge *br = netdev_priv(dev);
 151
 152         br_multicast_dev_del(br);
 153         br_multicast_uninit_stats(br);
 154         br_vlan_flush(br);
 155         br_mdb_hash_fini(br);
 156         br_fdb_hash_fini(br);
 157         free_percpu(br->stats);
 158 }
 159
 160 static int br_dev_open(struct net_device *dev)
 161 {
 162         struct net_bridge *br = netdev_priv(dev);
 163
 164         netdev_update_features(dev);
 165         netif_start_queue(dev);
 166         br_stp_enable_bridge(br);
 167         br_multicast_open(br);
 168
 169         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 170                 br_multicast_join_snoopers(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         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 193                 br_multicast_leave_snoopers(br);
 194
 195         netif_stop_queue(dev);
 196
 197         return 0;
 198 }
 199
 200 static void br_get_stats64(struct net_device *dev,
 201                            struct rtnl_link_stats64 *stats)
 202 {
 203         struct net_bridge *br = netdev_priv(dev);
 204         struct pcpu_sw_netstats tmp, sum = { 0 };
 205         unsigned int cpu;
 206
 207         for_each_possible_cpu(cpu) {
 208                 unsigned int start;
 209                 const struct pcpu_sw_netstats *bstats
 210                         = per_cpu_ptr(br->stats, cpu);
 211                 do {
 212                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 213                         memcpy(&tmp, bstats, sizeof(tmp));
 214                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 215                 sum.tx_bytes   += tmp.tx_bytes;
 216                 sum.tx_packets += tmp.tx_packets;
 217                 sum.rx_bytes   += tmp.rx_bytes;
 218                 sum.rx_packets += tmp.rx_packets;
 219         }
 220
 221         netdev_stats_to_stats64(stats, &dev->stats);
 222         stats->tx_bytes   = sum.tx_bytes;
 223         stats->tx_packets = sum.tx_packets;
 224         stats->rx_bytes   = sum.rx_bytes;
 225         stats->rx_packets = sum.rx_packets;
 226 }
 227
 228 static int br_change_mtu(struct net_device *dev, int new_mtu)
 229 {
 230         struct net_bridge *br = netdev_priv(dev);
 231
 232         dev->mtu = new_mtu;
 233
 234         /* this flag will be cleared if the MTU was automatically adjusted */
 235         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 236 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 237         /* remember the MTU in the rtable for PMTU */
 238         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 239 #endif
 240
 241         return 0;
 242 }
 243
 244 /* Allow setting mac address to any valid ethernet address. */
 245 static int br_set_mac_address(struct net_device *dev, void *p)
 246 {
 247         struct net_bridge *br = netdev_priv(dev);
 248         struct sockaddr *addr = p;
 249
 250         if (!is_valid_ether_addr(addr->sa_data))
 251                 return -EADDRNOTAVAIL;
 252
 253         /* dev_set_mac_addr() can be called by a master device on bridge's
 254          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 255          */
 256         if (dev->reg_state != NETREG_REGISTERED)
 257                 return -EBUSY;
 258
 259         spin_lock_bh(&br->lock);
 260         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 261                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 262                 br_stp_change_bridge_id(br, addr->sa_data);
 263         }
 264         spin_unlock_bh(&br->lock);
 265
 266         return 0;
 267 }
 268
 269 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 270 {
 271         strlcpy(info->driver, "bridge", sizeof(info->driver));
 272         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 273         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 274         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 275 }
 276
 277 static netdev_features_t br_fix_features(struct net_device *dev,
 278         netdev_features_t features)
 279 {
 280         struct net_bridge *br = netdev_priv(dev);
 281
 282         return br_features_recompute(br, features);
 283 }
 284
 285 #ifdef CONFIG_NET_POLL_CONTROLLER
 286 static void br_poll_controller(struct net_device *br_dev)
 287 {
 288 }
 289
 290 static void br_netpoll_cleanup(struct net_device *dev)
 291 {
 292         struct net_bridge *br = netdev_priv(dev);
 293         struct net_bridge_port *p;
 294
 295         list_for_each_entry(p, &br->port_list, list)
 296                 br_netpoll_disable(p);
 297 }
 298
 299 static int __br_netpoll_enable(struct net_bridge_port *p)
 300 {
 301         struct netpoll *np;
 302         int err;
 303
 304         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 305         if (!np)
 306                 return -ENOMEM;
 307
 308         err = __netpoll_setup(np, p->dev);
 309         if (err) {
 310                 kfree(np);
 311                 return err;
 312         }
 313
 314         p->np = np;
 315         return err;
 316 }
 317
 318 int br_netpoll_enable(struct net_bridge_port *p)
 319 {
 320         if (!p->br->dev->npinfo)
 321                 return 0;
 322
 323         return __br_netpoll_enable(p);
 324 }
 325
 326 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 327 {
 328         struct net_bridge *br = netdev_priv(dev);
 329         struct net_bridge_port *p;
 330         int err = 0;
 331
 332         list_for_each_entry(p, &br->port_list, list) {
 333                 if (!p->dev)
 334                         continue;
 335                 err = __br_netpoll_enable(p);
 336                 if (err)
 337                         goto fail;
 338         }
 339
 340 out:
 341         return err;
 342
 343 fail:
 344         br_netpoll_cleanup(dev);
 345         goto out;
 346 }
 347
 348 void br_netpoll_disable(struct net_bridge_port *p)
 349 {
 350         struct netpoll *np = p->np;
 351
 352         if (!np)
 353                 return;
 354
 355         p->np = NULL;
 356
 357         __netpoll_free(np);
 358 }
 359
 360 #endif
 361
 362 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 363                         struct netlink_ext_ack *extack)
 364
 365 {
 366         struct net_bridge *br = netdev_priv(dev);
 367
 368         return br_add_if(br, slave_dev, extack);
 369 }
 370
 371 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 372 {
 373         struct net_bridge *br = netdev_priv(dev);
 374
 375         return br_del_if(br, slave_dev);
 376 }
 377
 378 static const struct ethtool_ops br_ethtool_ops = {
 379         .get_drvinfo    = br_getinfo,
 380         .get_link       = ethtool_op_get_link,
 381 };
 382
 383 static const struct net_device_ops br_netdev_ops = {
 384         .ndo_open                = br_dev_open,
 385         .ndo_stop                = br_dev_stop,
 386         .ndo_init                = br_dev_init,
 387         .ndo_uninit              = br_dev_uninit,
 388         .ndo_start_xmit          = br_dev_xmit,
 389         .ndo_get_stats64         = br_get_stats64,
 390         .ndo_set_mac_address     = br_set_mac_address,
 391         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 392         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 393         .ndo_change_mtu          = br_change_mtu,
 394         .ndo_do_ioctl            = br_dev_ioctl,
 395 #ifdef CONFIG_NET_POLL_CONTROLLER
 396         .ndo_netpoll_setup       = br_netpoll_setup,
 397         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 398         .ndo_poll_controller     = br_poll_controller,
 399 #endif
 400         .ndo_add_slave           = br_add_slave,
 401         .ndo_del_slave           = br_del_slave,
 402         .ndo_fix_features        = br_fix_features,
 403         .ndo_fdb_add             = br_fdb_add,
 404         .ndo_fdb_del             = br_fdb_delete,
 405         .ndo_fdb_dump            = br_fdb_dump,
 406         .ndo_fdb_get             = br_fdb_get,
 407         .ndo_bridge_getlink      = br_getlink,
 408         .ndo_bridge_setlink      = br_setlink,
 409         .ndo_bridge_dellink      = br_dellink,
 410         .ndo_features_check      = passthru_features_check,
 411 };
 412
 413 static struct device_type br_type = {
 414         .name   = "bridge",
 415 };
 416
 417 void br_dev_setup(struct net_device *dev)
 418 {
 419         struct net_bridge *br = netdev_priv(dev);
 420
 421         eth_hw_addr_random(dev);
 422         ether_setup(dev);
 423
 424         dev->netdev_ops = &br_netdev_ops;
 425         dev->needs_free_netdev = true;
 426         dev->ethtool_ops = &br_ethtool_ops;
 427         SET_NETDEV_DEVTYPE(dev, &br_type);
 428         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 429
 430         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 431                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 432         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 433                            NETIF_F_HW_VLAN_STAG_TX;
 434         dev->vlan_features = COMMON_FEATURES;
 435
 436         br->dev = dev;
 437         spin_lock_init(&br->lock);
 438         INIT_LIST_HEAD(&br->port_list);
 439         INIT_HLIST_HEAD(&br->fdb_list);
 440         spin_lock_init(&br->hash_lock);
 441
 442         br->bridge_id.prio[0] = 0x80;
 443         br->bridge_id.prio[1] = 0x00;
 444
 445         ether_addr_copy(br->group_addr, eth_stp_addr);
 446
 447         br->stp_enabled = BR_NO_STP;
 448         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 449         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 450
 451         br->designated_root = br->bridge_id;
 452         br->bridge_max_age = br->max_age = 20 * HZ;
 453         br->bridge_hello_time = br->hello_time = 2 * HZ;
 454         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 455         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 456         dev->max_mtu = ETH_MAX_MTU;
 457
 458         br_netfilter_rtable_init(br);
 459         br_stp_timer_init(br);
 460         br_multicast_init(br);
 461         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 462 }