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         u8 state = BR_STATE_FORWARDING;
  36         const unsigned char *dest;
  37         u16 vid = 0;
  38
  39         rcu_read_lock();
  40         nf_ops = rcu_dereference(nf_br_ops);
  41         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  42                 rcu_read_unlock();
  43                 return NETDEV_TX_OK;
  44         }
  45
  46         u64_stats_update_begin(&brstats->syncp);
  47         brstats->tx_packets++;
  48         brstats->tx_bytes += skb->len;
  49         u64_stats_update_end(&brstats->syncp);
  50
  51         br_switchdev_frame_unmark(skb);
  52         BR_INPUT_SKB_CB(skb)->brdev = dev;
  53         BR_INPUT_SKB_CB(skb)->frag_max_size = 0;
  54
  55         skb_reset_mac_header(skb);
  56         skb_pull(skb, ETH_HLEN);
  57
  58         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid, &state))
  59                 goto out;
  60
  61         if (IS_ENABLED(CONFIG_INET) &&
  62             (eth_hdr(skb)->h_proto == htons(ETH_P_ARP) ||
  63              eth_hdr(skb)->h_proto == htons(ETH_P_RARP)) &&
  64             br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
  65                 br_do_proxy_suppress_arp(skb, br, vid, NULL);
  66         } else if (IS_ENABLED(CONFIG_IPV6) &&
  67                    skb->protocol == htons(ETH_P_IPV6) &&
  68                    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
  69                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
  70                                  sizeof(struct nd_msg)) &&
  71                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
  72                         struct nd_msg *msg, _msg;
  73
  74                         msg = br_is_nd_neigh_msg(skb, &_msg);
  75                         if (msg)
  76                                 br_do_suppress_nd(skb, br, vid, NULL, msg);
  77         }
  78
  79         dest = eth_hdr(skb)->h_dest;
  80         if (is_broadcast_ether_addr(dest)) {
  81                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  82         } else if (is_multicast_ether_addr(dest)) {
  83                 if (unlikely(netpoll_tx_running(dev))) {
  84                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  85                         goto out;
  86                 }
  87                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  88                         kfree_skb(skb);
  89                         goto out;
  90                 }
  91
  92                 mdst = br_mdb_get(br, skb, vid);
  93                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  94                     br_multicast_querier_exists(br, eth_hdr(skb)))
  95                         br_multicast_flood(mdst, skb, false, true);
  96                 else
  97                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  98         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
  99                 br_forward(dst->dst, skb, false, true);
 100         } else {
 101                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
 102         }
 103 out:
 104         rcu_read_unlock();
 105         return NETDEV_TX_OK;
 106 }
 107
 108 static int br_dev_init(struct net_device *dev)
 109 {
 110         struct net_bridge *br = netdev_priv(dev);
 111         int err;
 112
 113         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 114         if (!br->stats)
 115                 return -ENOMEM;
 116
 117         err = br_fdb_hash_init(br);
 118         if (err) {
 119                 free_percpu(br->stats);
 120                 return err;
 121         }
 122
 123         err = br_mdb_hash_init(br);
 124         if (err) {
 125                 free_percpu(br->stats);
 126                 br_fdb_hash_fini(br);
 127                 return err;
 128         }
 129
 130         err = br_vlan_init(br);
 131         if (err) {
 132                 free_percpu(br->stats);
 133                 br_mdb_hash_fini(br);
 134                 br_fdb_hash_fini(br);
 135                 return err;
 136         }
 137
 138         err = br_multicast_init_stats(br);
 139         if (err) {
 140                 free_percpu(br->stats);
 141                 br_vlan_flush(br);
 142                 br_mdb_hash_fini(br);
 143                 br_fdb_hash_fini(br);
 144         }
 145
 146         return err;
 147 }
 148
 149 static void br_dev_uninit(struct net_device *dev)
 150 {
 151         struct net_bridge *br = netdev_priv(dev);
 152
 153         br_multicast_dev_del(br);
 154         br_multicast_uninit_stats(br);
 155         br_vlan_flush(br);
 156         br_mdb_hash_fini(br);
 157         br_fdb_hash_fini(br);
 158         free_percpu(br->stats);
 159 }
 160
 161 static int br_dev_open(struct net_device *dev)
 162 {
 163         struct net_bridge *br = netdev_priv(dev);
 164
 165         netdev_update_features(dev);
 166         netif_start_queue(dev);
 167         br_stp_enable_bridge(br);
 168         br_multicast_open(br);
 169
 170         return 0;
 171 }
 172
 173 static void br_dev_set_multicast_list(struct net_device *dev)
 174 {
 175 }
 176
 177 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 178 {
 179         if (change & IFF_PROMISC)
 180                 br_manage_promisc(netdev_priv(dev));
 181 }
 182
 183 static int br_dev_stop(struct net_device *dev)
 184 {
 185         struct net_bridge *br = netdev_priv(dev);
 186
 187         br_stp_disable_bridge(br);
 188         br_multicast_stop(br);
 189
 190         netif_stop_queue(dev);
 191
 192         return 0;
 193 }
 194
 195 static void br_get_stats64(struct net_device *dev,
 196                            struct rtnl_link_stats64 *stats)
 197 {
 198         struct net_bridge *br = netdev_priv(dev);
 199         struct pcpu_sw_netstats tmp, sum = { 0 };
 200         unsigned int cpu;
 201
 202         for_each_possible_cpu(cpu) {
 203                 unsigned int start;
 204                 const struct pcpu_sw_netstats *bstats
 205                         = per_cpu_ptr(br->stats, cpu);
 206                 do {
 207                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 208                         memcpy(&tmp, bstats, sizeof(tmp));
 209                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 210                 sum.tx_bytes   += tmp.tx_bytes;
 211                 sum.tx_packets += tmp.tx_packets;
 212                 sum.rx_bytes   += tmp.rx_bytes;
 213                 sum.rx_packets += tmp.rx_packets;
 214         }
 215
 216         stats->tx_bytes   = sum.tx_bytes;
 217         stats->tx_packets = sum.tx_packets;
 218         stats->rx_bytes   = sum.rx_bytes;
 219         stats->rx_packets = sum.rx_packets;
 220 }
 221
 222 static int br_change_mtu(struct net_device *dev, int new_mtu)
 223 {
 224         struct net_bridge *br = netdev_priv(dev);
 225
 226         dev->mtu = new_mtu;
 227
 228         /* this flag will be cleared if the MTU was automatically adjusted */
 229         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 230 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 231         /* remember the MTU in the rtable for PMTU */
 232         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 233 #endif
 234
 235         return 0;
 236 }
 237
 238 /* Allow setting mac address to any valid ethernet address. */
 239 static int br_set_mac_address(struct net_device *dev, void *p)
 240 {
 241         struct net_bridge *br = netdev_priv(dev);
 242         struct sockaddr *addr = p;
 243
 244         if (!is_valid_ether_addr(addr->sa_data))
 245                 return -EADDRNOTAVAIL;
 246
 247         /* dev_set_mac_addr() can be called by a master device on bridge's
 248          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 249          */
 250         if (dev->reg_state != NETREG_REGISTERED)
 251                 return -EBUSY;
 252
 253         spin_lock_bh(&br->lock);
 254         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 255                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 256                 br_stp_change_bridge_id(br, addr->sa_data);
 257         }
 258         spin_unlock_bh(&br->lock);
 259
 260         return 0;
 261 }
 262
 263 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 264 {
 265         strlcpy(info->driver, "bridge", sizeof(info->driver));
 266         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 267         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 268         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 269 }
 270
 271 static int br_get_link_ksettings(struct net_device *dev,
 272                                  struct ethtool_link_ksettings *cmd)
 273 {
 274         struct net_bridge *br = netdev_priv(dev);
 275         struct net_bridge_port *p;
 276
 277         cmd->base.duplex = DUPLEX_UNKNOWN;
 278         cmd->base.port = PORT_OTHER;
 279         cmd->base.speed = SPEED_UNKNOWN;
 280
 281         list_for_each_entry(p, &br->port_list, list) {
 282                 struct ethtool_link_ksettings ecmd;
 283                 struct net_device *pdev = p->dev;
 284
 285                 if (!netif_running(pdev) || !netif_oper_up(pdev))
 286                         continue;
 287
 288                 if (__ethtool_get_link_ksettings(pdev, &ecmd))
 289                         continue;
 290
 291                 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN)
 292                         continue;
 293
 294                 if (cmd->base.speed == (__u32)SPEED_UNKNOWN ||
 295                     cmd->base.speed < ecmd.base.speed)
 296                         cmd->base.speed = ecmd.base.speed;
 297         }
 298
 299         return 0;
 300 }
 301
 302 static netdev_features_t br_fix_features(struct net_device *dev,
 303         netdev_features_t features)
 304 {
 305         struct net_bridge *br = netdev_priv(dev);
 306
 307         return br_features_recompute(br, features);
 308 }
 309
 310 #ifdef CONFIG_NET_POLL_CONTROLLER
 311 static void br_poll_controller(struct net_device *br_dev)
 312 {
 313 }
 314
 315 static void br_netpoll_cleanup(struct net_device *dev)
 316 {
 317         struct net_bridge *br = netdev_priv(dev);
 318         struct net_bridge_port *p;
 319
 320         list_for_each_entry(p, &br->port_list, list)
 321                 br_netpoll_disable(p);
 322 }
 323
 324 static int __br_netpoll_enable(struct net_bridge_port *p)
 325 {
 326         struct netpoll *np;
 327         int err;
 328
 329         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 330         if (!np)
 331                 return -ENOMEM;
 332
 333         err = __netpoll_setup(np, p->dev);
 334         if (err) {
 335                 kfree(np);
 336                 return err;
 337         }
 338
 339         p->np = np;
 340         return err;
 341 }
 342
 343 int br_netpoll_enable(struct net_bridge_port *p)
 344 {
 345         if (!p->br->dev->npinfo)
 346                 return 0;
 347
 348         return __br_netpoll_enable(p);
 349 }
 350
 351 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 352 {
 353         struct net_bridge *br = netdev_priv(dev);
 354         struct net_bridge_port *p;
 355         int err = 0;
 356
 357         list_for_each_entry(p, &br->port_list, list) {
 358                 if (!p->dev)
 359                         continue;
 360                 err = __br_netpoll_enable(p);
 361                 if (err)
 362                         goto fail;
 363         }
 364
 365 out:
 366         return err;
 367
 368 fail:
 369         br_netpoll_cleanup(dev);
 370         goto out;
 371 }
 372
 373 void br_netpoll_disable(struct net_bridge_port *p)
 374 {
 375         struct netpoll *np = p->np;
 376
 377         if (!np)
 378                 return;
 379
 380         p->np = NULL;
 381
 382         __netpoll_free(np);
 383 }
 384
 385 #endif
 386
 387 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 388                         struct netlink_ext_ack *extack)
 389
 390 {
 391         struct net_bridge *br = netdev_priv(dev);
 392
 393         return br_add_if(br, slave_dev, extack);
 394 }
 395
 396 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 397 {
 398         struct net_bridge *br = netdev_priv(dev);
 399
 400         return br_del_if(br, slave_dev);
 401 }
 402
 403 static const struct ethtool_ops br_ethtool_ops = {
 404         .get_drvinfo             = br_getinfo,
 405         .get_link                = ethtool_op_get_link,
 406         .get_link_ksettings      = br_get_link_ksettings,
 407 };
 408
 409 static const struct net_device_ops br_netdev_ops = {
 410         .ndo_open                = br_dev_open,
 411         .ndo_stop                = br_dev_stop,
 412         .ndo_init                = br_dev_init,
 413         .ndo_uninit              = br_dev_uninit,
 414         .ndo_start_xmit          = br_dev_xmit,
 415         .ndo_get_stats64         = br_get_stats64,
 416         .ndo_set_mac_address     = br_set_mac_address,
 417         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 418         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 419         .ndo_change_mtu          = br_change_mtu,
 420         .ndo_do_ioctl            = br_dev_ioctl,
 421 #ifdef CONFIG_NET_POLL_CONTROLLER
 422         .ndo_netpoll_setup       = br_netpoll_setup,
 423         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 424         .ndo_poll_controller     = br_poll_controller,
 425 #endif
 426         .ndo_add_slave           = br_add_slave,
 427         .ndo_del_slave           = br_del_slave,
 428         .ndo_fix_features        = br_fix_features,
 429         .ndo_fdb_add             = br_fdb_add,
 430         .ndo_fdb_del             = br_fdb_delete,
 431         .ndo_fdb_dump            = br_fdb_dump,
 432         .ndo_fdb_get             = br_fdb_get,
 433         .ndo_bridge_getlink      = br_getlink,
 434         .ndo_bridge_setlink      = br_setlink,
 435         .ndo_bridge_dellink      = br_dellink,
 436         .ndo_features_check      = passthru_features_check,
 437 };
 438
 439 static struct device_type br_type = {
 440         .name   = "bridge",
 441 };
 442
 443 void br_dev_setup(struct net_device *dev)
 444 {
 445         struct net_bridge *br = netdev_priv(dev);
 446
 447         eth_hw_addr_random(dev);
 448         ether_setup(dev);
 449
 450         dev->netdev_ops = &br_netdev_ops;
 451         dev->needs_free_netdev = true;
 452         dev->ethtool_ops = &br_ethtool_ops;
 453         SET_NETDEV_DEVTYPE(dev, &br_type);
 454         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 455
 456         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 457                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 458         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 459                            NETIF_F_HW_VLAN_STAG_TX;
 460         dev->vlan_features = COMMON_FEATURES;
 461
 462         br->dev = dev;
 463         spin_lock_init(&br->lock);
 464         INIT_LIST_HEAD(&br->port_list);
 465         INIT_HLIST_HEAD(&br->fdb_list);
 466         spin_lock_init(&br->hash_lock);
 467
 468         br->bridge_id.prio[0] = 0x80;
 469         br->bridge_id.prio[1] = 0x00;
 470
 471         ether_addr_copy(br->group_addr, eth_stp_addr);
 472
 473         br->stp_enabled = BR_NO_STP;
 474         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 475         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 476
 477         br->designated_root = br->bridge_id;
 478         br->bridge_max_age = br->max_age = 20 * HZ;
 479         br->bridge_hello_time = br->hello_time = 2 * HZ;
 480         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 481         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 482         dev->max_mtu = ETH_MAX_MTU;
 483
 484         br_netfilter_rtable_init(br);
 485         br_stp_timer_init(br);
 486         br_multicast_init(br);
 487         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 488 }