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         memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  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         BR_INPUT_SKB_CB(skb)->frag_max_size = 0;
  56
  57         skb_reset_mac_header(skb);
  58         skb_pull(skb, ETH_HLEN);
  59
  60         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid, &state))
  61                 goto out;
  62
  63         if (IS_ENABLED(CONFIG_INET) &&
  64             (eth_hdr(skb)->h_proto == htons(ETH_P_ARP) ||
  65              eth_hdr(skb)->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 struct lock_class_key bridge_netdev_addr_lock_key;
 111
 112 static void br_set_lockdep_class(struct net_device *dev)
 113 {
 114         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
 115 }
 116
 117 static int br_dev_init(struct net_device *dev)
 118 {
 119         struct net_bridge *br = netdev_priv(dev);
 120         int err;
 121
 122         br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 123         if (!br->stats)
 124                 return -ENOMEM;
 125
 126         err = br_fdb_hash_init(br);
 127         if (err) {
 128                 free_percpu(br->stats);
 129                 return err;
 130         }
 131
 132         err = br_mdb_hash_init(br);
 133         if (err) {
 134                 free_percpu(br->stats);
 135                 br_fdb_hash_fini(br);
 136                 return err;
 137         }
 138
 139         err = br_vlan_init(br);
 140         if (err) {
 141                 free_percpu(br->stats);
 142                 br_mdb_hash_fini(br);
 143                 br_fdb_hash_fini(br);
 144                 return err;
 145         }
 146
 147         err = br_multicast_init_stats(br);
 148         if (err) {
 149                 free_percpu(br->stats);
 150                 br_vlan_flush(br);
 151                 br_mdb_hash_fini(br);
 152                 br_fdb_hash_fini(br);
 153         }
 154
 155         br_set_lockdep_class(dev);
 156         return err;
 157 }
 158
 159 static void br_dev_uninit(struct net_device *dev)
 160 {
 161         struct net_bridge *br = netdev_priv(dev);
 162
 163         br_multicast_dev_del(br);
 164         br_multicast_uninit_stats(br);
 165         br_vlan_flush(br);
 166         br_mdb_hash_fini(br);
 167         br_fdb_hash_fini(br);
 168         free_percpu(br->stats);
 169 }
 170
 171 static int br_dev_open(struct net_device *dev)
 172 {
 173         struct net_bridge *br = netdev_priv(dev);
 174
 175         netdev_update_features(dev);
 176         netif_start_queue(dev);
 177         br_stp_enable_bridge(br);
 178         br_multicast_open(br);
 179
 180         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 181                 br_multicast_join_snoopers(br);
 182
 183         return 0;
 184 }
 185
 186 static void br_dev_set_multicast_list(struct net_device *dev)
 187 {
 188 }
 189
 190 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 191 {
 192         if (change & IFF_PROMISC)
 193                 br_manage_promisc(netdev_priv(dev));
 194 }
 195
 196 static int br_dev_stop(struct net_device *dev)
 197 {
 198         struct net_bridge *br = netdev_priv(dev);
 199
 200         br_stp_disable_bridge(br);
 201         br_multicast_stop(br);
 202
 203         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 204                 br_multicast_leave_snoopers(br);
 205
 206         netif_stop_queue(dev);
 207
 208         return 0;
 209 }
 210
 211 static void br_get_stats64(struct net_device *dev,
 212                            struct rtnl_link_stats64 *stats)
 213 {
 214         struct net_bridge *br = netdev_priv(dev);
 215
 216         netdev_stats_to_stats64(stats, &dev->stats);
 217         dev_fetch_sw_netstats(stats, br->stats);
 218 }
 219
 220 static int br_change_mtu(struct net_device *dev, int new_mtu)
 221 {
 222         struct net_bridge *br = netdev_priv(dev);
 223
 224         dev->mtu = new_mtu;
 225
 226         /* this flag will be cleared if the MTU was automatically adjusted */
 227         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 228 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 229         /* remember the MTU in the rtable for PMTU */
 230         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 231 #endif
 232
 233         return 0;
 234 }
 235
 236 /* Allow setting mac address to any valid ethernet address. */
 237 static int br_set_mac_address(struct net_device *dev, void *p)
 238 {
 239         struct net_bridge *br = netdev_priv(dev);
 240         struct sockaddr *addr = p;
 241
 242         if (!is_valid_ether_addr(addr->sa_data))
 243                 return -EADDRNOTAVAIL;
 244
 245         /* dev_set_mac_addr() can be called by a master device on bridge's
 246          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 247          */
 248         if (dev->reg_state != NETREG_REGISTERED)
 249                 return -EBUSY;
 250
 251         spin_lock_bh(&br->lock);
 252         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 253                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 254                 br_stp_change_bridge_id(br, addr->sa_data);
 255         }
 256         spin_unlock_bh(&br->lock);
 257
 258         return 0;
 259 }
 260
 261 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 262 {
 263         strlcpy(info->driver, "bridge", sizeof(info->driver));
 264         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 265         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 266         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 267 }
 268
 269 static int br_get_link_ksettings(struct net_device *dev,
 270                                  struct ethtool_link_ksettings *cmd)
 271 {
 272         struct net_bridge *br = netdev_priv(dev);
 273         struct net_bridge_port *p;
 274
 275         cmd->base.duplex = DUPLEX_UNKNOWN;
 276         cmd->base.port = PORT_OTHER;
 277         cmd->base.speed = SPEED_UNKNOWN;
 278
 279         list_for_each_entry(p, &br->port_list, list) {
 280                 struct ethtool_link_ksettings ecmd;
 281                 struct net_device *pdev = p->dev;
 282
 283                 if (!netif_running(pdev) || !netif_oper_up(pdev))
 284                         continue;
 285
 286                 if (__ethtool_get_link_ksettings(pdev, &ecmd))
 287                         continue;
 288
 289                 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN)
 290                         continue;
 291
 292                 if (cmd->base.speed == (__u32)SPEED_UNKNOWN ||
 293                     cmd->base.speed < ecmd.base.speed)
 294                         cmd->base.speed = ecmd.base.speed;
 295         }
 296
 297         return 0;
 298 }
 299
 300 static netdev_features_t br_fix_features(struct net_device *dev,
 301         netdev_features_t features)
 302 {
 303         struct net_bridge *br = netdev_priv(dev);
 304
 305         return br_features_recompute(br, features);
 306 }
 307
 308 #ifdef CONFIG_NET_POLL_CONTROLLER
 309 static void br_poll_controller(struct net_device *br_dev)
 310 {
 311 }
 312
 313 static void br_netpoll_cleanup(struct net_device *dev)
 314 {
 315         struct net_bridge *br = netdev_priv(dev);
 316         struct net_bridge_port *p;
 317
 318         list_for_each_entry(p, &br->port_list, list)
 319                 br_netpoll_disable(p);
 320 }
 321
 322 static int __br_netpoll_enable(struct net_bridge_port *p)
 323 {
 324         struct netpoll *np;
 325         int err;
 326
 327         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 328         if (!np)
 329                 return -ENOMEM;
 330
 331         err = __netpoll_setup(np, p->dev);
 332         if (err) {
 333                 kfree(np);
 334                 return err;
 335         }
 336
 337         p->np = np;
 338         return err;
 339 }
 340
 341 int br_netpoll_enable(struct net_bridge_port *p)
 342 {
 343         if (!p->br->dev->npinfo)
 344                 return 0;
 345
 346         return __br_netpoll_enable(p);
 347 }
 348
 349 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 350 {
 351         struct net_bridge *br = netdev_priv(dev);
 352         struct net_bridge_port *p;
 353         int err = 0;
 354
 355         list_for_each_entry(p, &br->port_list, list) {
 356                 if (!p->dev)
 357                         continue;
 358                 err = __br_netpoll_enable(p);
 359                 if (err)
 360                         goto fail;
 361         }
 362
 363 out:
 364         return err;
 365
 366 fail:
 367         br_netpoll_cleanup(dev);
 368         goto out;
 369 }
 370
 371 void br_netpoll_disable(struct net_bridge_port *p)
 372 {
 373         struct netpoll *np = p->np;
 374
 375         if (!np)
 376                 return;
 377
 378         p->np = NULL;
 379
 380         __netpoll_free(np);
 381 }
 382
 383 #endif
 384
 385 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 386                         struct netlink_ext_ack *extack)
 387
 388 {
 389         struct net_bridge *br = netdev_priv(dev);
 390
 391         return br_add_if(br, slave_dev, extack);
 392 }
 393
 394 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 395 {
 396         struct net_bridge *br = netdev_priv(dev);
 397
 398         return br_del_if(br, slave_dev);
 399 }
 400
 401 static const struct ethtool_ops br_ethtool_ops = {
 402         .get_drvinfo             = br_getinfo,
 403         .get_link                = ethtool_op_get_link,
 404         .get_link_ksettings      = br_get_link_ksettings,
 405 };
 406
 407 static const struct net_device_ops br_netdev_ops = {
 408         .ndo_open                = br_dev_open,
 409         .ndo_stop                = br_dev_stop,
 410         .ndo_init                = br_dev_init,
 411         .ndo_uninit              = br_dev_uninit,
 412         .ndo_start_xmit          = br_dev_xmit,
 413         .ndo_get_stats64         = br_get_stats64,
 414         .ndo_set_mac_address     = br_set_mac_address,
 415         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 416         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 417         .ndo_change_mtu          = br_change_mtu,
 418         .ndo_do_ioctl            = br_dev_ioctl,
 419 #ifdef CONFIG_NET_POLL_CONTROLLER
 420         .ndo_netpoll_setup       = br_netpoll_setup,
 421         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 422         .ndo_poll_controller     = br_poll_controller,
 423 #endif
 424         .ndo_add_slave           = br_add_slave,
 425         .ndo_del_slave           = br_del_slave,
 426         .ndo_fix_features        = br_fix_features,
 427         .ndo_fdb_add             = br_fdb_add,
 428         .ndo_fdb_del             = br_fdb_delete,
 429         .ndo_fdb_dump            = br_fdb_dump,
 430         .ndo_fdb_get             = br_fdb_get,
 431         .ndo_bridge_getlink      = br_getlink,
 432         .ndo_bridge_setlink      = br_setlink,
 433         .ndo_bridge_dellink      = br_dellink,
 434         .ndo_features_check      = passthru_features_check,
 435 };
 436
 437 static struct device_type br_type = {
 438         .name   = "bridge",
 439 };
 440
 441 void br_dev_setup(struct net_device *dev)
 442 {
 443         struct net_bridge *br = netdev_priv(dev);
 444
 445         eth_hw_addr_random(dev);
 446         ether_setup(dev);
 447
 448         dev->netdev_ops = &br_netdev_ops;
 449         dev->needs_free_netdev = true;
 450         dev->ethtool_ops = &br_ethtool_ops;
 451         SET_NETDEV_DEVTYPE(dev, &br_type);
 452         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 453
 454         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 455                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 456         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 457                            NETIF_F_HW_VLAN_STAG_TX;
 458         dev->vlan_features = COMMON_FEATURES;
 459
 460         br->dev = dev;
 461         spin_lock_init(&br->lock);
 462         INIT_LIST_HEAD(&br->port_list);
 463         INIT_HLIST_HEAD(&br->fdb_list);
 464 #if IS_ENABLED(CONFIG_BRIDGE_MRP)
 465         INIT_LIST_HEAD(&br->mrp_list);
 466 #endif
 467         spin_lock_init(&br->hash_lock);
 468
 469         br->bridge_id.prio[0] = 0x80;
 470         br->bridge_id.prio[1] = 0x00;
 471
 472         ether_addr_copy(br->group_addr, eth_stp_addr);
 473
 474         br->stp_enabled = BR_NO_STP;
 475         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 476         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 477
 478         br->designated_root = br->bridge_id;
 479         br->bridge_max_age = br->max_age = 20 * HZ;
 480         br->bridge_hello_time = br->hello_time = 2 * HZ;
 481         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 482         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 483         dev->max_mtu = ETH_MAX_MTU;
 484
 485         br_netfilter_rtable_init(br);
 486         br_stp_timer_init(br);
 487         br_multicast_init(br);
 488         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 489 }