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         enum skb_drop_reason reason = pskb_may_pull_reason(skb, ETH_HLEN);
  31         struct net_bridge_mcast_port *pmctx_null = NULL;
  32         struct net_bridge *br = netdev_priv(dev);
  33         struct net_bridge_mcast *brmctx = &br->multicast_ctx;
  34         struct net_bridge_fdb_entry *dst;
  35         struct net_bridge_mdb_entry *mdst;
  36         const struct nf_br_ops *nf_ops;
  37         u8 state = BR_STATE_FORWARDING;
  38         struct net_bridge_vlan *vlan;
  39         const unsigned char *dest;
  40         u16 vid = 0;
  41
  42         if (unlikely(reason != SKB_NOT_DROPPED_YET)) {
  43                 kfree_skb_reason(skb, reason);
  44                 return NETDEV_TX_OK;
  45         }
  46
  47         memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  48         br_tc_skb_miss_set(skb, false);
  49
  50         rcu_read_lock();
  51         nf_ops = rcu_dereference(nf_br_ops);
  52         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  53                 rcu_read_unlock();
  54                 return NETDEV_TX_OK;
  55         }
  56
  57         dev_sw_netstats_tx_add(dev, 1, skb->len);
  58
  59         br_switchdev_frame_unmark(skb);
  60         BR_INPUT_SKB_CB(skb)->brdev = dev;
  61         BR_INPUT_SKB_CB(skb)->frag_max_size = 0;
  62
  63         skb_reset_mac_header(skb);
  64         skb_pull(skb, ETH_HLEN);
  65
  66         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid,
  67                                 &state, &vlan))
  68                 goto out;
  69
  70         if (IS_ENABLED(CONFIG_INET) &&
  71             (eth_hdr(skb)->h_proto == htons(ETH_P_ARP) ||
  72              eth_hdr(skb)->h_proto == htons(ETH_P_RARP)) &&
  73             br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
  74                 br_do_proxy_suppress_arp(skb, br, vid, NULL);
  75         } else if (IS_ENABLED(CONFIG_IPV6) &&
  76                    skb->protocol == htons(ETH_P_IPV6) &&
  77                    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
  78                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
  79                                  sizeof(struct nd_msg)) &&
  80                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
  81                         struct nd_msg *msg, _msg;
  82
  83                         msg = br_is_nd_neigh_msg(skb, &_msg);
  84                         if (msg)
  85                                 br_do_suppress_nd(skb, br, vid, NULL, msg);
  86         }
  87
  88         dest = eth_hdr(skb)->h_dest;
  89         if (is_broadcast_ether_addr(dest)) {
  90                 br_flood(br, skb, BR_PKT_BROADCAST, false, true, vid);
  91         } else if (is_multicast_ether_addr(dest)) {
  92                 if (unlikely(netpoll_tx_running(dev))) {
  93                         br_flood(br, skb, BR_PKT_MULTICAST, false, true, vid);
  94                         goto out;
  95                 }
  96                 if (br_multicast_rcv(&brmctx, &pmctx_null, vlan, skb, vid)) {
  97                         kfree_skb(skb);
  98                         goto out;
  99                 }
 100
 101                 mdst = br_mdb_entry_skb_get(brmctx, skb, vid);
 102                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
 103                     br_multicast_querier_exists(brmctx, eth_hdr(skb), mdst))
 104                         br_multicast_flood(mdst, skb, brmctx, false, true);
 105                 else
 106                         br_flood(br, skb, BR_PKT_MULTICAST, false, true, vid);
 107         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
 108                 br_forward(dst->dst, skb, false, true);
 109         } else {
 110                 br_flood(br, skb, BR_PKT_UNICAST, false, true, vid);
 111         }
 112 out:
 113         rcu_read_unlock();
 114         return NETDEV_TX_OK;
 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         err = br_fdb_hash_init(br);
 123         if (err)
 124                 return err;
 125
 126         err = br_mdb_hash_init(br);
 127         if (err) {
 128                 br_fdb_hash_fini(br);
 129                 return err;
 130         }
 131
 132         err = br_vlan_init(br);
 133         if (err) {
 134                 br_mdb_hash_fini(br);
 135                 br_fdb_hash_fini(br);
 136                 return err;
 137         }
 138
 139         err = br_multicast_init_stats(br);
 140         if (err) {
 141                 br_vlan_flush(br);
 142                 br_mdb_hash_fini(br);
 143                 br_fdb_hash_fini(br);
 144                 return err;
 145         }
 146
 147         netdev_lockdep_set_classes(dev);
 148         return 0;
 149 }
 150
 151 static void br_dev_uninit(struct net_device *dev)
 152 {
 153         struct net_bridge *br = netdev_priv(dev);
 154
 155         br_multicast_dev_del(br);
 156         br_multicast_uninit_stats(br);
 157         br_vlan_flush(br);
 158         br_mdb_hash_fini(br);
 159         br_fdb_hash_fini(br);
 160 }
 161
 162 static int br_dev_open(struct net_device *dev)
 163 {
 164         struct net_bridge *br = netdev_priv(dev);
 165
 166         netdev_update_features(dev);
 167         netif_start_queue(dev);
 168         br_stp_enable_bridge(br);
 169         br_multicast_open(br);
 170
 171         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 172                 br_multicast_join_snoopers(br);
 173
 174         return 0;
 175 }
 176
 177 static void br_dev_set_multicast_list(struct net_device *dev)
 178 {
 179 }
 180
 181 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 182 {
 183         if (change & IFF_PROMISC)
 184                 br_manage_promisc(netdev_priv(dev));
 185 }
 186
 187 static int br_dev_stop(struct net_device *dev)
 188 {
 189         struct net_bridge *br = netdev_priv(dev);
 190
 191         br_stp_disable_bridge(br);
 192         br_multicast_stop(br);
 193
 194         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 195                 br_multicast_leave_snoopers(br);
 196
 197         netif_stop_queue(dev);
 198
 199         return 0;
 200 }
 201
 202 static int br_change_mtu(struct net_device *dev, int new_mtu)
 203 {
 204         struct net_bridge *br = netdev_priv(dev);
 205
 206         WRITE_ONCE(dev->mtu, new_mtu);
 207
 208         /* this flag will be cleared if the MTU was automatically adjusted */
 209         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 210 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 211         /* remember the MTU in the rtable for PMTU */
 212         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 213 #endif
 214
 215         return 0;
 216 }
 217
 218 /* Allow setting mac address to any valid ethernet address. */
 219 static int br_set_mac_address(struct net_device *dev, void *p)
 220 {
 221         struct net_bridge *br = netdev_priv(dev);
 222         struct sockaddr *addr = p;
 223
 224         if (!is_valid_ether_addr(addr->sa_data))
 225                 return -EADDRNOTAVAIL;
 226
 227         /* dev_set_mac_addr() can be called by a master device on bridge's
 228          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 229          */
 230         if (dev->reg_state != NETREG_REGISTERED)
 231                 return -EBUSY;
 232
 233         spin_lock_bh(&br->lock);
 234         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 235                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 236                 br_stp_change_bridge_id(br, addr->sa_data);
 237         }
 238         spin_unlock_bh(&br->lock);
 239
 240         return 0;
 241 }
 242
 243 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 244 {
 245         strscpy(info->driver, "bridge", sizeof(info->driver));
 246         strscpy(info->version, BR_VERSION, sizeof(info->version));
 247         strscpy(info->fw_version, "N/A", sizeof(info->fw_version));
 248         strscpy(info->bus_info, "N/A", sizeof(info->bus_info));
 249 }
 250
 251 static int br_get_link_ksettings(struct net_device *dev,
 252                                  struct ethtool_link_ksettings *cmd)
 253 {
 254         struct net_bridge *br = netdev_priv(dev);
 255         struct net_bridge_port *p;
 256
 257         cmd->base.duplex = DUPLEX_UNKNOWN;
 258         cmd->base.port = PORT_OTHER;
 259         cmd->base.speed = SPEED_UNKNOWN;
 260
 261         list_for_each_entry(p, &br->port_list, list) {
 262                 struct ethtool_link_ksettings ecmd;
 263                 struct net_device *pdev = p->dev;
 264
 265                 if (!netif_running(pdev) || !netif_oper_up(pdev))
 266                         continue;
 267
 268                 if (__ethtool_get_link_ksettings(pdev, &ecmd))
 269                         continue;
 270
 271                 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN)
 272                         continue;
 273
 274                 if (cmd->base.speed == (__u32)SPEED_UNKNOWN ||
 275                     cmd->base.speed < ecmd.base.speed)
 276                         cmd->base.speed = ecmd.base.speed;
 277         }
 278
 279         return 0;
 280 }
 281
 282 static netdev_features_t br_fix_features(struct net_device *dev,
 283         netdev_features_t features)
 284 {
 285         struct net_bridge *br = netdev_priv(dev);
 286
 287         return br_features_recompute(br, features);
 288 }
 289
 290 #ifdef CONFIG_NET_POLL_CONTROLLER
 291 static void br_poll_controller(struct net_device *br_dev)
 292 {
 293 }
 294
 295 static void br_netpoll_cleanup(struct net_device *dev)
 296 {
 297         struct net_bridge *br = netdev_priv(dev);
 298         struct net_bridge_port *p;
 299
 300         list_for_each_entry(p, &br->port_list, list)
 301                 br_netpoll_disable(p);
 302 }
 303
 304 static int __br_netpoll_enable(struct net_bridge_port *p)
 305 {
 306         struct netpoll *np;
 307         int err;
 308
 309         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 310         if (!np)
 311                 return -ENOMEM;
 312
 313         err = __netpoll_setup(np, p->dev);
 314         if (err) {
 315                 kfree(np);
 316                 return err;
 317         }
 318
 319         p->np = np;
 320         return err;
 321 }
 322
 323 int br_netpoll_enable(struct net_bridge_port *p)
 324 {
 325         if (!p->br->dev->npinfo)
 326                 return 0;
 327
 328         return __br_netpoll_enable(p);
 329 }
 330
 331 static int br_netpoll_setup(struct net_device *dev)
 332 {
 333         struct net_bridge *br = netdev_priv(dev);
 334         struct net_bridge_port *p;
 335         int err = 0;
 336
 337         list_for_each_entry(p, &br->port_list, list) {
 338                 if (!p->dev)
 339                         continue;
 340                 err = __br_netpoll_enable(p);
 341                 if (err)
 342                         goto fail;
 343         }
 344
 345 out:
 346         return err;
 347
 348 fail:
 349         br_netpoll_cleanup(dev);
 350         goto out;
 351 }
 352
 353 void br_netpoll_disable(struct net_bridge_port *p)
 354 {
 355         struct netpoll *np = p->np;
 356
 357         if (!np)
 358                 return;
 359
 360         p->np = NULL;
 361
 362         __netpoll_free(np);
 363 }
 364
 365 #endif
 366
 367 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 368                         struct netlink_ext_ack *extack)
 369
 370 {
 371         struct net_bridge *br = netdev_priv(dev);
 372
 373         return br_add_if(br, slave_dev, extack);
 374 }
 375
 376 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 377 {
 378         struct net_bridge *br = netdev_priv(dev);
 379
 380         return br_del_if(br, slave_dev);
 381 }
 382
 383 static int br_fill_forward_path(struct net_device_path_ctx *ctx,
 384                                 struct net_device_path *path)
 385 {
 386         struct net_bridge_fdb_entry *f;
 387         struct net_bridge_port *dst;
 388         struct net_bridge *br;
 389
 390         if (netif_is_bridge_port(ctx->dev))
 391                 return -1;
 392
 393         br = netdev_priv(ctx->dev);
 394
 395         br_vlan_fill_forward_path_pvid(br, ctx, path);
 396
 397         f = br_fdb_find_rcu(br, ctx->daddr, path->bridge.vlan_id);
 398         if (!f)
 399                 return -1;
 400
 401         dst = READ_ONCE(f->dst);
 402         if (!dst)
 403                 return -1;
 404
 405         if (br_vlan_fill_forward_path_mode(br, dst, path))
 406                 return -1;
 407
 408         path->type = DEV_PATH_BRIDGE;
 409         path->dev = dst->br->dev;
 410         ctx->dev = dst->dev;
 411
 412         switch (path->bridge.vlan_mode) {
 413         case DEV_PATH_BR_VLAN_TAG:
 414                 if (ctx->num_vlans >= ARRAY_SIZE(ctx->vlan))
 415                         return -ENOSPC;
 416                 ctx->vlan[ctx->num_vlans].id = path->bridge.vlan_id;
 417                 ctx->vlan[ctx->num_vlans].proto = path->bridge.vlan_proto;
 418                 ctx->num_vlans++;
 419                 break;
 420         case DEV_PATH_BR_VLAN_UNTAG_HW:
 421         case DEV_PATH_BR_VLAN_UNTAG:
 422                 ctx->num_vlans--;
 423                 break;
 424         case DEV_PATH_BR_VLAN_KEEP:
 425                 break;
 426         }
 427
 428         return 0;
 429 }
 430
 431 static const struct ethtool_ops br_ethtool_ops = {
 432         .get_drvinfo             = br_getinfo,
 433         .get_link                = ethtool_op_get_link,
 434         .get_link_ksettings      = br_get_link_ksettings,
 435 };
 436
 437 static const struct net_device_ops br_netdev_ops = {
 438         .ndo_open                = br_dev_open,
 439         .ndo_stop                = br_dev_stop,
 440         .ndo_init                = br_dev_init,
 441         .ndo_uninit              = br_dev_uninit,
 442         .ndo_start_xmit          = br_dev_xmit,
 443         .ndo_get_stats64         = dev_get_tstats64,
 444         .ndo_set_mac_address     = br_set_mac_address,
 445         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 446         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 447         .ndo_change_mtu          = br_change_mtu,
 448         .ndo_siocdevprivate      = br_dev_siocdevprivate,
 449 #ifdef CONFIG_NET_POLL_CONTROLLER
 450         .ndo_netpoll_setup       = br_netpoll_setup,
 451         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 452         .ndo_poll_controller     = br_poll_controller,
 453 #endif
 454         .ndo_add_slave           = br_add_slave,
 455         .ndo_del_slave           = br_del_slave,
 456         .ndo_fix_features        = br_fix_features,
 457         .ndo_fdb_add             = br_fdb_add,
 458         .ndo_fdb_del             = br_fdb_delete,
 459         .ndo_fdb_del_bulk        = br_fdb_delete_bulk,
 460         .ndo_fdb_dump            = br_fdb_dump,
 461         .ndo_fdb_get             = br_fdb_get,
 462         .ndo_mdb_add             = br_mdb_add,
 463         .ndo_mdb_del             = br_mdb_del,
 464         .ndo_mdb_del_bulk        = br_mdb_del_bulk,
 465         .ndo_mdb_dump            = br_mdb_dump,
 466         .ndo_mdb_get             = br_mdb_get,
 467         .ndo_bridge_getlink      = br_getlink,
 468         .ndo_bridge_setlink      = br_setlink,
 469         .ndo_bridge_dellink      = br_dellink,
 470         .ndo_features_check      = passthru_features_check,
 471         .ndo_fill_forward_path   = br_fill_forward_path,
 472 };
 473
 474 static const struct device_type br_type = {
 475         .name   = "bridge",
 476 };
 477
 478 void br_dev_setup(struct net_device *dev)
 479 {
 480         struct net_bridge *br = netdev_priv(dev);
 481
 482         eth_hw_addr_random(dev);
 483         ether_setup(dev);
 484
 485         dev->netdev_ops = &br_netdev_ops;
 486         dev->needs_free_netdev = true;
 487         dev->ethtool_ops = &br_ethtool_ops;
 488         SET_NETDEV_DEVTYPE(dev, &br_type);
 489         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 490         dev->lltx = true;
 491         dev->netns_local = true;
 492
 493         dev->features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 494                         NETIF_F_HW_VLAN_STAG_TX;
 495         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 496                            NETIF_F_HW_VLAN_STAG_TX;
 497         dev->vlan_features = COMMON_FEATURES;
 498         dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
 499
 500         br->dev = dev;
 501         spin_lock_init(&br->lock);
 502         INIT_LIST_HEAD(&br->port_list);
 503         INIT_HLIST_HEAD(&br->fdb_list);
 504         INIT_HLIST_HEAD(&br->frame_type_list);
 505 #if IS_ENABLED(CONFIG_BRIDGE_MRP)
 506         INIT_HLIST_HEAD(&br->mrp_list);
 507 #endif
 508 #if IS_ENABLED(CONFIG_BRIDGE_CFM)
 509         INIT_HLIST_HEAD(&br->mep_list);
 510 #endif
 511         spin_lock_init(&br->hash_lock);
 512
 513         br->bridge_id.prio[0] = 0x80;
 514         br->bridge_id.prio[1] = 0x00;
 515
 516         ether_addr_copy(br->group_addr, eth_stp_addr);
 517
 518         br->stp_enabled = BR_NO_STP;
 519         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 520         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 521
 522         br->designated_root = br->bridge_id;
 523         br->bridge_max_age = br->max_age = 20 * HZ;
 524         br->bridge_hello_time = br->hello_time = 2 * HZ;
 525         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 526         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 527         dev->max_mtu = ETH_MAX_MTU;
 528
 529         br_netfilter_rtable_init(br);
 530         br_stp_timer_init(br);
 531         br_multicast_init(br);
 532         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 533 }