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_mcast_port *pmctx_null = NULL;
  31         struct net_bridge *br = netdev_priv(dev);
  32         struct net_bridge_mcast *brmctx = &br->multicast_ctx;
  33         struct net_bridge_fdb_entry *dst;
  34         struct net_bridge_mdb_entry *mdst;
  35         const struct nf_br_ops *nf_ops;
  36         u8 state = BR_STATE_FORWARDING;
  37         struct net_bridge_vlan *vlan;
  38         const unsigned char *dest;
  39         u16 vid = 0;
  40
  41         memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  42         br_tc_skb_miss_set(skb, false);
  43
  44         rcu_read_lock();
  45         nf_ops = rcu_dereference(nf_br_ops);
  46         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  47                 rcu_read_unlock();
  48                 return NETDEV_TX_OK;
  49         }
  50
  51         dev_sw_netstats_tx_add(dev, 1, skb->len);
  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,
  61                                 &state, &vlan))
  62                 goto out;
  63
  64         if (IS_ENABLED(CONFIG_INET) &&
  65             (eth_hdr(skb)->h_proto == htons(ETH_P_ARP) ||
  66              eth_hdr(skb)->h_proto == htons(ETH_P_RARP)) &&
  67             br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
  68                 br_do_proxy_suppress_arp(skb, br, vid, NULL);
  69         } else if (IS_ENABLED(CONFIG_IPV6) &&
  70                    skb->protocol == htons(ETH_P_IPV6) &&
  71                    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
  72                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
  73                                  sizeof(struct nd_msg)) &&
  74                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
  75                         struct nd_msg *msg, _msg;
  76
  77                         msg = br_is_nd_neigh_msg(skb, &_msg);
  78                         if (msg)
  79                                 br_do_suppress_nd(skb, br, vid, NULL, msg);
  80         }
  81
  82         dest = eth_hdr(skb)->h_dest;
  83         if (is_broadcast_ether_addr(dest)) {
  84                 br_flood(br, skb, BR_PKT_BROADCAST, false, true, vid);
  85         } else if (is_multicast_ether_addr(dest)) {
  86                 if (unlikely(netpoll_tx_running(dev))) {
  87                         br_flood(br, skb, BR_PKT_MULTICAST, false, true, vid);
  88                         goto out;
  89                 }
  90                 if (br_multicast_rcv(&brmctx, &pmctx_null, vlan, skb, vid)) {
  91                         kfree_skb(skb);
  92                         goto out;
  93                 }
  94
  95                 mdst = br_mdb_entry_skb_get(brmctx, skb, vid);
  96                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  97                     br_multicast_querier_exists(brmctx, eth_hdr(skb), mdst))
  98                         br_multicast_flood(mdst, skb, brmctx, false, true);
  99                 else
 100                         br_flood(br, skb, BR_PKT_MULTICAST, false, true, vid);
 101         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
 102                 br_forward(dst->dst, skb, false, true);
 103         } else {
 104                 br_flood(br, skb, BR_PKT_UNICAST, false, true, vid);
 105         }
 106 out:
 107         rcu_read_unlock();
 108         return NETDEV_TX_OK;
 109 }
 110
 111 static int br_dev_init(struct net_device *dev)
 112 {
 113         struct net_bridge *br = netdev_priv(dev);
 114         int err;
 115
 116         err = br_fdb_hash_init(br);
 117         if (err)
 118                 return err;
 119
 120         err = br_mdb_hash_init(br);
 121         if (err) {
 122                 br_fdb_hash_fini(br);
 123                 return err;
 124         }
 125
 126         err = br_vlan_init(br);
 127         if (err) {
 128                 br_mdb_hash_fini(br);
 129                 br_fdb_hash_fini(br);
 130                 return err;
 131         }
 132
 133         err = br_multicast_init_stats(br);
 134         if (err) {
 135                 br_vlan_flush(br);
 136                 br_mdb_hash_fini(br);
 137                 br_fdb_hash_fini(br);
 138         }
 139
 140         netdev_lockdep_set_classes(dev);
 141         return err;
 142 }
 143
 144 static void br_dev_uninit(struct net_device *dev)
 145 {
 146         struct net_bridge *br = netdev_priv(dev);
 147
 148         br_multicast_dev_del(br);
 149         br_multicast_uninit_stats(br);
 150         br_vlan_flush(br);
 151         br_mdb_hash_fini(br);
 152         br_fdb_hash_fini(br);
 153 }
 154
 155 static int br_dev_open(struct net_device *dev)
 156 {
 157         struct net_bridge *br = netdev_priv(dev);
 158
 159         netdev_update_features(dev);
 160         netif_start_queue(dev);
 161         br_stp_enable_bridge(br);
 162         br_multicast_open(br);
 163
 164         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 165                 br_multicast_join_snoopers(br);
 166
 167         return 0;
 168 }
 169
 170 static void br_dev_set_multicast_list(struct net_device *dev)
 171 {
 172 }
 173
 174 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 175 {
 176         if (change & IFF_PROMISC)
 177                 br_manage_promisc(netdev_priv(dev));
 178 }
 179
 180 static int br_dev_stop(struct net_device *dev)
 181 {
 182         struct net_bridge *br = netdev_priv(dev);
 183
 184         br_stp_disable_bridge(br);
 185         br_multicast_stop(br);
 186
 187         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 188                 br_multicast_leave_snoopers(br);
 189
 190         netif_stop_queue(dev);
 191
 192         return 0;
 193 }
 194
 195 static int br_change_mtu(struct net_device *dev, int new_mtu)
 196 {
 197         struct net_bridge *br = netdev_priv(dev);
 198
 199         dev->mtu = new_mtu;
 200
 201         /* this flag will be cleared if the MTU was automatically adjusted */
 202         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 203 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 204         /* remember the MTU in the rtable for PMTU */
 205         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 206 #endif
 207
 208         return 0;
 209 }
 210
 211 /* Allow setting mac address to any valid ethernet address. */
 212 static int br_set_mac_address(struct net_device *dev, void *p)
 213 {
 214         struct net_bridge *br = netdev_priv(dev);
 215         struct sockaddr *addr = p;
 216
 217         if (!is_valid_ether_addr(addr->sa_data))
 218                 return -EADDRNOTAVAIL;
 219
 220         /* dev_set_mac_addr() can be called by a master device on bridge's
 221          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 222          */
 223         if (dev->reg_state != NETREG_REGISTERED)
 224                 return -EBUSY;
 225
 226         spin_lock_bh(&br->lock);
 227         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 228                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 229                 br_stp_change_bridge_id(br, addr->sa_data);
 230         }
 231         spin_unlock_bh(&br->lock);
 232
 233         return 0;
 234 }
 235
 236 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 237 {
 238         strscpy(info->driver, "bridge", sizeof(info->driver));
 239         strscpy(info->version, BR_VERSION, sizeof(info->version));
 240         strscpy(info->fw_version, "N/A", sizeof(info->fw_version));
 241         strscpy(info->bus_info, "N/A", sizeof(info->bus_info));
 242 }
 243
 244 static int br_get_link_ksettings(struct net_device *dev,
 245                                  struct ethtool_link_ksettings *cmd)
 246 {
 247         struct net_bridge *br = netdev_priv(dev);
 248         struct net_bridge_port *p;
 249
 250         cmd->base.duplex = DUPLEX_UNKNOWN;
 251         cmd->base.port = PORT_OTHER;
 252         cmd->base.speed = SPEED_UNKNOWN;
 253
 254         list_for_each_entry(p, &br->port_list, list) {
 255                 struct ethtool_link_ksettings ecmd;
 256                 struct net_device *pdev = p->dev;
 257
 258                 if (!netif_running(pdev) || !netif_oper_up(pdev))
 259                         continue;
 260
 261                 if (__ethtool_get_link_ksettings(pdev, &ecmd))
 262                         continue;
 263
 264                 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN)
 265                         continue;
 266
 267                 if (cmd->base.speed == (__u32)SPEED_UNKNOWN ||
 268                     cmd->base.speed < ecmd.base.speed)
 269                         cmd->base.speed = ecmd.base.speed;
 270         }
 271
 272         return 0;
 273 }
 274
 275 static netdev_features_t br_fix_features(struct net_device *dev,
 276         netdev_features_t features)
 277 {
 278         struct net_bridge *br = netdev_priv(dev);
 279
 280         return br_features_recompute(br, features);
 281 }
 282
 283 #ifdef CONFIG_NET_POLL_CONTROLLER
 284 static void br_poll_controller(struct net_device *br_dev)
 285 {
 286 }
 287
 288 static void br_netpoll_cleanup(struct net_device *dev)
 289 {
 290         struct net_bridge *br = netdev_priv(dev);
 291         struct net_bridge_port *p;
 292
 293         list_for_each_entry(p, &br->port_list, list)
 294                 br_netpoll_disable(p);
 295 }
 296
 297 static int __br_netpoll_enable(struct net_bridge_port *p)
 298 {
 299         struct netpoll *np;
 300         int err;
 301
 302         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 303         if (!np)
 304                 return -ENOMEM;
 305
 306         err = __netpoll_setup(np, p->dev);
 307         if (err) {
 308                 kfree(np);
 309                 return err;
 310         }
 311
 312         p->np = np;
 313         return err;
 314 }
 315
 316 int br_netpoll_enable(struct net_bridge_port *p)
 317 {
 318         if (!p->br->dev->npinfo)
 319                 return 0;
 320
 321         return __br_netpoll_enable(p);
 322 }
 323
 324 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 325 {
 326         struct net_bridge *br = netdev_priv(dev);
 327         struct net_bridge_port *p;
 328         int err = 0;
 329
 330         list_for_each_entry(p, &br->port_list, list) {
 331                 if (!p->dev)
 332                         continue;
 333                 err = __br_netpoll_enable(p);
 334                 if (err)
 335                         goto fail;
 336         }
 337
 338 out:
 339         return err;
 340
 341 fail:
 342         br_netpoll_cleanup(dev);
 343         goto out;
 344 }
 345
 346 void br_netpoll_disable(struct net_bridge_port *p)
 347 {
 348         struct netpoll *np = p->np;
 349
 350         if (!np)
 351                 return;
 352
 353         p->np = NULL;
 354
 355         __netpoll_free(np);
 356 }
 357
 358 #endif
 359
 360 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 361                         struct netlink_ext_ack *extack)
 362
 363 {
 364         struct net_bridge *br = netdev_priv(dev);
 365
 366         return br_add_if(br, slave_dev, extack);
 367 }
 368
 369 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 370 {
 371         struct net_bridge *br = netdev_priv(dev);
 372
 373         return br_del_if(br, slave_dev);
 374 }
 375
 376 static int br_fill_forward_path(struct net_device_path_ctx *ctx,
 377                                 struct net_device_path *path)
 378 {
 379         struct net_bridge_fdb_entry *f;
 380         struct net_bridge_port *dst;
 381         struct net_bridge *br;
 382
 383         if (netif_is_bridge_port(ctx->dev))
 384                 return -1;
 385
 386         br = netdev_priv(ctx->dev);
 387
 388         br_vlan_fill_forward_path_pvid(br, ctx, path);
 389
 390         f = br_fdb_find_rcu(br, ctx->daddr, path->bridge.vlan_id);
 391         if (!f || !f->dst)
 392                 return -1;
 393
 394         dst = READ_ONCE(f->dst);
 395         if (!dst)
 396                 return -1;
 397
 398         if (br_vlan_fill_forward_path_mode(br, dst, path))
 399                 return -1;
 400
 401         path->type = DEV_PATH_BRIDGE;
 402         path->dev = dst->br->dev;
 403         ctx->dev = dst->dev;
 404
 405         switch (path->bridge.vlan_mode) {
 406         case DEV_PATH_BR_VLAN_TAG:
 407                 if (ctx->num_vlans >= ARRAY_SIZE(ctx->vlan))
 408                         return -ENOSPC;
 409                 ctx->vlan[ctx->num_vlans].id = path->bridge.vlan_id;
 410                 ctx->vlan[ctx->num_vlans].proto = path->bridge.vlan_proto;
 411                 ctx->num_vlans++;
 412                 break;
 413         case DEV_PATH_BR_VLAN_UNTAG_HW:
 414         case DEV_PATH_BR_VLAN_UNTAG:
 415                 ctx->num_vlans--;
 416                 break;
 417         case DEV_PATH_BR_VLAN_KEEP:
 418                 break;
 419         }
 420
 421         return 0;
 422 }
 423
 424 static const struct ethtool_ops br_ethtool_ops = {
 425         .get_drvinfo             = br_getinfo,
 426         .get_link                = ethtool_op_get_link,
 427         .get_link_ksettings      = br_get_link_ksettings,
 428 };
 429
 430 static const struct net_device_ops br_netdev_ops = {
 431         .ndo_open                = br_dev_open,
 432         .ndo_stop                = br_dev_stop,
 433         .ndo_init                = br_dev_init,
 434         .ndo_uninit              = br_dev_uninit,
 435         .ndo_start_xmit          = br_dev_xmit,
 436         .ndo_get_stats64         = dev_get_tstats64,
 437         .ndo_set_mac_address     = br_set_mac_address,
 438         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 439         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 440         .ndo_change_mtu          = br_change_mtu,
 441         .ndo_siocdevprivate      = br_dev_siocdevprivate,
 442 #ifdef CONFIG_NET_POLL_CONTROLLER
 443         .ndo_netpoll_setup       = br_netpoll_setup,
 444         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 445         .ndo_poll_controller     = br_poll_controller,
 446 #endif
 447         .ndo_add_slave           = br_add_slave,
 448         .ndo_del_slave           = br_del_slave,
 449         .ndo_fix_features        = br_fix_features,
 450         .ndo_fdb_add             = br_fdb_add,
 451         .ndo_fdb_del             = br_fdb_delete,
 452         .ndo_fdb_del_bulk        = br_fdb_delete_bulk,
 453         .ndo_fdb_dump            = br_fdb_dump,
 454         .ndo_fdb_get             = br_fdb_get,
 455         .ndo_mdb_add             = br_mdb_add,
 456         .ndo_mdb_del             = br_mdb_del,
 457         .ndo_mdb_del_bulk        = br_mdb_del_bulk,
 458         .ndo_mdb_dump            = br_mdb_dump,
 459         .ndo_mdb_get             = br_mdb_get,
 460         .ndo_bridge_getlink      = br_getlink,
 461         .ndo_bridge_setlink      = br_setlink,
 462         .ndo_bridge_dellink      = br_dellink,
 463         .ndo_features_check      = passthru_features_check,
 464         .ndo_fill_forward_path   = br_fill_forward_path,
 465 };
 466
 467 static const struct device_type br_type = {
 468         .name   = "bridge",
 469 };
 470
 471 void br_dev_setup(struct net_device *dev)
 472 {
 473         struct net_bridge *br = netdev_priv(dev);
 474
 475         eth_hw_addr_random(dev);
 476         ether_setup(dev);
 477
 478         dev->netdev_ops = &br_netdev_ops;
 479         dev->needs_free_netdev = true;
 480         dev->ethtool_ops = &br_ethtool_ops;
 481         SET_NETDEV_DEVTYPE(dev, &br_type);
 482         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 483
 484         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 485                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 486         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 487                            NETIF_F_HW_VLAN_STAG_TX;
 488         dev->vlan_features = COMMON_FEATURES;
 489         dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
 490
 491         br->dev = dev;
 492         spin_lock_init(&br->lock);
 493         INIT_LIST_HEAD(&br->port_list);
 494         INIT_HLIST_HEAD(&br->fdb_list);
 495         INIT_HLIST_HEAD(&br->frame_type_list);
 496 #if IS_ENABLED(CONFIG_BRIDGE_MRP)
 497         INIT_HLIST_HEAD(&br->mrp_list);
 498 #endif
 499 #if IS_ENABLED(CONFIG_BRIDGE_CFM)
 500         INIT_HLIST_HEAD(&br->mep_list);
 501 #endif
 502         spin_lock_init(&br->hash_lock);
 503
 504         br->bridge_id.prio[0] = 0x80;
 505         br->bridge_id.prio[1] = 0x00;
 506
 507         ether_addr_copy(br->group_addr, eth_stp_addr);
 508
 509         br->stp_enabled = BR_NO_STP;
 510         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 511         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 512
 513         br->designated_root = br->bridge_id;
 514         br->bridge_max_age = br->max_age = 20 * HZ;
 515         br->bridge_hello_time = br->hello_time = 2 * HZ;
 516         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 517         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 518         dev->max_mtu = ETH_MAX_MTU;
 519
 520         br_netfilter_rtable_init(br);
 521         br_stp_timer_init(br);
 522         br_multicast_init(br);
 523         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 524 }