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 static struct lock_class_key bridge_netdev_addr_lock_key;
  28
  29 /* net device transmit always called with BH disabled */
  30 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev)
  31 {
  32         struct net_bridge *br = netdev_priv(dev);
  33         struct net_bridge_fdb_entry *dst;
  34         struct net_bridge_mdb_entry *mdst;
  35         struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
  36         const struct nf_br_ops *nf_ops;
  37         const unsigned char *dest;
  38         struct ethhdr *eth;
  39         u16 vid = 0;
  40
  41         rcu_read_lock();
  42         nf_ops = rcu_dereference(nf_br_ops);
  43         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  44                 rcu_read_unlock();
  45                 return NETDEV_TX_OK;
  46         }
  47
  48         u64_stats_update_begin(&brstats->syncp);
  49         brstats->tx_packets++;
  50         brstats->tx_bytes += skb->len;
  51         u64_stats_update_end(&brstats->syncp);
  52
  53         br_switchdev_frame_unmark(skb);
  54         BR_INPUT_SKB_CB(skb)->brdev = dev;
  55
  56         skb_reset_mac_header(skb);
  57         eth = eth_hdr(skb);
  58         skb_pull(skb, ETH_HLEN);
  59
  60         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid))
  61                 goto out;
  62
  63         if (IS_ENABLED(CONFIG_INET) &&
  64             (eth->h_proto == htons(ETH_P_ARP) ||
  65              eth->h_proto == htons(ETH_P_RARP)) &&
  66             br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
  67                 br_do_proxy_suppress_arp(skb, br, vid, NULL);
  68         } else if (IS_ENABLED(CONFIG_IPV6) &&
  69                    skb->protocol == htons(ETH_P_IPV6) &&
  70                    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
  71                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
  72                                  sizeof(struct nd_msg)) &&
  73                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
  74                         struct nd_msg *msg, _msg;
  75
  76                         msg = br_is_nd_neigh_msg(skb, &_msg);
  77                         if (msg)
  78                                 br_do_suppress_nd(skb, br, vid, NULL, msg);
  79         }
  80
  81         dest = eth_hdr(skb)->h_dest;
  82         if (is_broadcast_ether_addr(dest)) {
  83                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  84         } else if (is_multicast_ether_addr(dest)) {
  85                 if (unlikely(netpoll_tx_running(dev))) {
  86                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  87                         goto out;
  88                 }
  89                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  90                         kfree_skb(skb);
  91                         goto out;
  92                 }
  93
  94                 mdst = br_mdb_get(br, skb, vid);
  95                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  96                     br_multicast_querier_exists(br, eth_hdr(skb)))
  97                         br_multicast_flood(mdst, skb, false, true);
  98                 else
  99                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
 100         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
 101                 br_forward(dst->dst, skb, false, true);
 102         } else {
 103                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
 104         }
 105 out:
 106         rcu_read_unlock();
 107         return NETDEV_TX_OK;
 108 }
 109
 110 static void br_set_lockdep_class(struct net_device *dev)
 111 {
 112         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
 113 }
 114
 115 static int br_dev_init(struct net_device *dev)
 116 {
 117         struct net_bridge *br = netdev_priv(dev);
 118         int err;
 119
 120         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 121         if (!br->stats)
 122                 return -ENOMEM;
 123
 124         err = br_fdb_hash_init(br);
 125         if (err) {
 126                 free_percpu(br->stats);
 127                 return err;
 128         }
 129
 130         err = br_mdb_hash_init(br);
 131         if (err) {
 132                 free_percpu(br->stats);
 133                 br_fdb_hash_fini(br);
 134                 return err;
 135         }
 136
 137         err = br_vlan_init(br);
 138         if (err) {
 139                 free_percpu(br->stats);
 140                 br_mdb_hash_fini(br);
 141                 br_fdb_hash_fini(br);
 142                 return err;
 143         }
 144
 145         err = br_multicast_init_stats(br);
 146         if (err) {
 147                 free_percpu(br->stats);
 148                 br_vlan_flush(br);
 149                 br_mdb_hash_fini(br);
 150                 br_fdb_hash_fini(br);
 151         }
 152         br_set_lockdep_class(dev);
 153
 154         return err;
 155 }
 156
 157 static void br_dev_uninit(struct net_device *dev)
 158 {
 159         struct net_bridge *br = netdev_priv(dev);
 160
 161         br_multicast_dev_del(br);
 162         br_multicast_uninit_stats(br);
 163         br_vlan_flush(br);
 164         br_mdb_hash_fini(br);
 165         br_fdb_hash_fini(br);
 166         free_percpu(br->stats);
 167 }
 168
 169 static int br_dev_open(struct net_device *dev)
 170 {
 171         struct net_bridge *br = netdev_priv(dev);
 172
 173         netdev_update_features(dev);
 174         netif_start_queue(dev);
 175         br_stp_enable_bridge(br);
 176         br_multicast_open(br);
 177
 178         return 0;
 179 }
 180
 181 static void br_dev_set_multicast_list(struct net_device *dev)
 182 {
 183 }
 184
 185 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 186 {
 187         if (change & IFF_PROMISC)
 188                 br_manage_promisc(netdev_priv(dev));
 189 }
 190
 191 static int br_dev_stop(struct net_device *dev)
 192 {
 193         struct net_bridge *br = netdev_priv(dev);
 194
 195         br_stp_disable_bridge(br);
 196         br_multicast_stop(br);
 197
 198         netif_stop_queue(dev);
 199
 200         return 0;
 201 }
 202
 203 static void br_get_stats64(struct net_device *dev,
 204                            struct rtnl_link_stats64 *stats)
 205 {
 206         struct net_bridge *br = netdev_priv(dev);
 207         struct pcpu_sw_netstats tmp, sum = { 0 };
 208         unsigned int cpu;
 209
 210         for_each_possible_cpu(cpu) {
 211                 unsigned int start;
 212                 const struct pcpu_sw_netstats *bstats
 213                         = per_cpu_ptr(br->stats, cpu);
 214                 do {
 215                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 216                         memcpy(&tmp, bstats, sizeof(tmp));
 217                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 218                 sum.tx_bytes   += tmp.tx_bytes;
 219                 sum.tx_packets += tmp.tx_packets;
 220                 sum.rx_bytes   += tmp.rx_bytes;
 221                 sum.rx_packets += tmp.rx_packets;
 222         }
 223
 224         stats->tx_bytes   = sum.tx_bytes;
 225         stats->tx_packets = sum.tx_packets;
 226         stats->rx_bytes   = sum.rx_bytes;
 227         stats->rx_packets = sum.rx_packets;
 228 }
 229
 230 static int br_change_mtu(struct net_device *dev, int new_mtu)
 231 {
 232         struct net_bridge *br = netdev_priv(dev);
 233
 234         dev->mtu = new_mtu;
 235
 236         /* this flag will be cleared if the MTU was automatically adjusted */
 237         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 238 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 239         /* remember the MTU in the rtable for PMTU */
 240         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 241 #endif
 242
 243         return 0;
 244 }
 245
 246 /* Allow setting mac address to any valid ethernet address. */
 247 static int br_set_mac_address(struct net_device *dev, void *p)
 248 {
 249         struct net_bridge *br = netdev_priv(dev);
 250         struct sockaddr *addr = p;
 251
 252         if (!is_valid_ether_addr(addr->sa_data))
 253                 return -EADDRNOTAVAIL;
 254
 255         spin_lock_bh(&br->lock);
 256         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 257                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 258                 br_stp_change_bridge_id(br, addr->sa_data);
 259         }
 260         spin_unlock_bh(&br->lock);
 261
 262         return 0;
 263 }
 264
 265 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 266 {
 267         strlcpy(info->driver, "bridge", sizeof(info->driver));
 268         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 269         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 270         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 271 }
 272
 273 static netdev_features_t br_fix_features(struct net_device *dev,
 274         netdev_features_t features)
 275 {
 276         struct net_bridge *br = netdev_priv(dev);
 277
 278         return br_features_recompute(br, features);
 279 }
 280
 281 #ifdef CONFIG_NET_POLL_CONTROLLER
 282 static void br_poll_controller(struct net_device *br_dev)
 283 {
 284 }
 285
 286 static void br_netpoll_cleanup(struct net_device *dev)
 287 {
 288         struct net_bridge *br = netdev_priv(dev);
 289         struct net_bridge_port *p;
 290
 291         list_for_each_entry(p, &br->port_list, list)
 292                 br_netpoll_disable(p);
 293 }
 294
 295 static int __br_netpoll_enable(struct net_bridge_port *p)
 296 {
 297         struct netpoll *np;
 298         int err;
 299
 300         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 301         if (!np)
 302                 return -ENOMEM;
 303
 304         err = __netpoll_setup(np, p->dev);
 305         if (err) {
 306                 kfree(np);
 307                 return err;
 308         }
 309
 310         p->np = np;
 311         return err;
 312 }
 313
 314 int br_netpoll_enable(struct net_bridge_port *p)
 315 {
 316         if (!p->br->dev->npinfo)
 317                 return 0;
 318
 319         return __br_netpoll_enable(p);
 320 }
 321
 322 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 323 {
 324         struct net_bridge *br = netdev_priv(dev);
 325         struct net_bridge_port *p;
 326         int err = 0;
 327
 328         list_for_each_entry(p, &br->port_list, list) {
 329                 if (!p->dev)
 330                         continue;
 331                 err = __br_netpoll_enable(p);
 332                 if (err)
 333                         goto fail;
 334         }
 335
 336 out:
 337         return err;
 338
 339 fail:
 340         br_netpoll_cleanup(dev);
 341         goto out;
 342 }
 343
 344 void br_netpoll_disable(struct net_bridge_port *p)
 345 {
 346         struct netpoll *np = p->np;
 347
 348         if (!np)
 349                 return;
 350
 351         p->np = NULL;
 352
 353         __netpoll_free(np);
 354 }
 355
 356 #endif
 357
 358 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 359                         struct netlink_ext_ack *extack)
 360
 361 {
 362         struct net_bridge *br = netdev_priv(dev);
 363
 364         return br_add_if(br, slave_dev, extack);
 365 }
 366
 367 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 368 {
 369         struct net_bridge *br = netdev_priv(dev);
 370
 371         return br_del_if(br, slave_dev);
 372 }
 373
 374 static const struct ethtool_ops br_ethtool_ops = {
 375         .get_drvinfo    = br_getinfo,
 376         .get_link       = ethtool_op_get_link,
 377 };
 378
 379 static const struct net_device_ops br_netdev_ops = {
 380         .ndo_open                = br_dev_open,
 381         .ndo_stop                = br_dev_stop,
 382         .ndo_init                = br_dev_init,
 383         .ndo_uninit              = br_dev_uninit,
 384         .ndo_start_xmit          = br_dev_xmit,
 385         .ndo_get_stats64         = br_get_stats64,
 386         .ndo_set_mac_address     = br_set_mac_address,
 387         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 388         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 389         .ndo_change_mtu          = br_change_mtu,
 390         .ndo_do_ioctl            = br_dev_ioctl,
 391 #ifdef CONFIG_NET_POLL_CONTROLLER
 392         .ndo_netpoll_setup       = br_netpoll_setup,
 393         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 394         .ndo_poll_controller     = br_poll_controller,
 395 #endif
 396         .ndo_add_slave           = br_add_slave,
 397         .ndo_del_slave           = br_del_slave,
 398         .ndo_fix_features        = br_fix_features,
 399         .ndo_fdb_add             = br_fdb_add,
 400         .ndo_fdb_del             = br_fdb_delete,
 401         .ndo_fdb_dump            = br_fdb_dump,
 402         .ndo_fdb_get             = br_fdb_get,
 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 }