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                 return err;
 139         }
 140
 141         netdev_lockdep_set_classes(dev);
 142         return 0;
 143 }
 144
 145 static void br_dev_uninit(struct net_device *dev)
 146 {
 147         struct net_bridge *br = netdev_priv(dev);
 148
 149         br_multicast_dev_del(br);
 150         br_multicast_uninit_stats(br);
 151         br_vlan_flush(br);
 152         br_mdb_hash_fini(br);
 153         br_fdb_hash_fini(br);
 154 }
 155
 156 static int br_dev_open(struct net_device *dev)
 157 {
 158         struct net_bridge *br = netdev_priv(dev);
 159
 160         netdev_update_features(dev);
 161         netif_start_queue(dev);
 162         br_stp_enable_bridge(br);
 163         br_multicast_open(br);
 164
 165         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 166                 br_multicast_join_snoopers(br);
 167
 168         return 0;
 169 }
 170
 171 static void br_dev_set_multicast_list(struct net_device *dev)
 172 {
 173 }
 174
 175 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 176 {
 177         if (change & IFF_PROMISC)
 178                 br_manage_promisc(netdev_priv(dev));
 179 }
 180
 181 static int br_dev_stop(struct net_device *dev)
 182 {
 183         struct net_bridge *br = netdev_priv(dev);
 184
 185         br_stp_disable_bridge(br);
 186         br_multicast_stop(br);
 187
 188         if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
 189                 br_multicast_leave_snoopers(br);
 190
 191         netif_stop_queue(dev);
 192
 193         return 0;
 194 }
 195
 196 static int br_change_mtu(struct net_device *dev, int new_mtu)
 197 {
 198         struct net_bridge *br = netdev_priv(dev);
 199
 200         dev->mtu = new_mtu;
 201
 202         /* this flag will be cleared if the MTU was automatically adjusted */
 203         br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 204 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 205         /* remember the MTU in the rtable for PMTU */
 206         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 207 #endif
 208
 209         return 0;
 210 }
 211
 212 /* Allow setting mac address to any valid ethernet address. */
 213 static int br_set_mac_address(struct net_device *dev, void *p)
 214 {
 215         struct net_bridge *br = netdev_priv(dev);
 216         struct sockaddr *addr = p;
 217
 218         if (!is_valid_ether_addr(addr->sa_data))
 219                 return -EADDRNOTAVAIL;
 220
 221         /* dev_set_mac_addr() can be called by a master device on bridge's
 222          * NETDEV_UNREGISTER, but since it's being destroyed do nothing
 223          */
 224         if (dev->reg_state != NETREG_REGISTERED)
 225                 return -EBUSY;
 226
 227         spin_lock_bh(&br->lock);
 228         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 229                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 230                 br_stp_change_bridge_id(br, addr->sa_data);
 231         }
 232         spin_unlock_bh(&br->lock);
 233
 234         return 0;
 235 }
 236
 237 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 238 {
 239         strscpy(info->driver, "bridge", sizeof(info->driver));
 240         strscpy(info->version, BR_VERSION, sizeof(info->version));
 241         strscpy(info->fw_version, "N/A", sizeof(info->fw_version));
 242         strscpy(info->bus_info, "N/A", sizeof(info->bus_info));
 243 }
 244
 245 static int br_get_link_ksettings(struct net_device *dev,
 246                                  struct ethtool_link_ksettings *cmd)
 247 {
 248         struct net_bridge *br = netdev_priv(dev);
 249         struct net_bridge_port *p;
 250
 251         cmd->base.duplex = DUPLEX_UNKNOWN;
 252         cmd->base.port = PORT_OTHER;
 253         cmd->base.speed = SPEED_UNKNOWN;
 254
 255         list_for_each_entry(p, &br->port_list, list) {
 256                 struct ethtool_link_ksettings ecmd;
 257                 struct net_device *pdev = p->dev;
 258
 259                 if (!netif_running(pdev) || !netif_oper_up(pdev))
 260                         continue;
 261
 262                 if (__ethtool_get_link_ksettings(pdev, &ecmd))
 263                         continue;
 264
 265                 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN)
 266                         continue;
 267
 268                 if (cmd->base.speed == (__u32)SPEED_UNKNOWN ||
 269                     cmd->base.speed < ecmd.base.speed)
 270                         cmd->base.speed = ecmd.base.speed;
 271         }
 272
 273         return 0;
 274 }
 275
 276 static netdev_features_t br_fix_features(struct net_device *dev,
 277         netdev_features_t features)
 278 {
 279         struct net_bridge *br = netdev_priv(dev);
 280
 281         return br_features_recompute(br, features);
 282 }
 283
 284 #ifdef CONFIG_NET_POLL_CONTROLLER
 285 static void br_poll_controller(struct net_device *br_dev)
 286 {
 287 }
 288
 289 static void br_netpoll_cleanup(struct net_device *dev)
 290 {
 291         struct net_bridge *br = netdev_priv(dev);
 292         struct net_bridge_port *p;
 293
 294         list_for_each_entry(p, &br->port_list, list)
 295                 br_netpoll_disable(p);
 296 }
 297
 298 static int __br_netpoll_enable(struct net_bridge_port *p)
 299 {
 300         struct netpoll *np;
 301         int err;
 302
 303         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 304         if (!np)
 305                 return -ENOMEM;
 306
 307         err = __netpoll_setup(np, p->dev);
 308         if (err) {
 309                 kfree(np);
 310                 return err;
 311         }
 312
 313         p->np = np;
 314         return err;
 315 }
 316
 317 int br_netpoll_enable(struct net_bridge_port *p)
 318 {
 319         if (!p->br->dev->npinfo)
 320                 return 0;
 321
 322         return __br_netpoll_enable(p);
 323 }
 324
 325 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 326 {
 327         struct net_bridge *br = netdev_priv(dev);
 328         struct net_bridge_port *p;
 329         int err = 0;
 330
 331         list_for_each_entry(p, &br->port_list, list) {
 332                 if (!p->dev)
 333                         continue;
 334                 err = __br_netpoll_enable(p);
 335                 if (err)
 336                         goto fail;
 337         }
 338
 339 out:
 340         return err;
 341
 342 fail:
 343         br_netpoll_cleanup(dev);
 344         goto out;
 345 }
 346
 347 void br_netpoll_disable(struct net_bridge_port *p)
 348 {
 349         struct netpoll *np = p->np;
 350
 351         if (!np)
 352                 return;
 353
 354         p->np = NULL;
 355
 356         __netpoll_free(np);
 357 }
 358
 359 #endif
 360
 361 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 362                         struct netlink_ext_ack *extack)
 363
 364 {
 365         struct net_bridge *br = netdev_priv(dev);
 366
 367         return br_add_if(br, slave_dev, extack);
 368 }
 369
 370 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 371 {
 372         struct net_bridge *br = netdev_priv(dev);
 373
 374         return br_del_if(br, slave_dev);
 375 }
 376
 377 static int br_fill_forward_path(struct net_device_path_ctx *ctx,
 378                                 struct net_device_path *path)
 379 {
 380         struct net_bridge_fdb_entry *f;
 381         struct net_bridge_port *dst;
 382         struct net_bridge *br;
 383
 384         if (netif_is_bridge_port(ctx->dev))
 385                 return -1;
 386
 387         br = netdev_priv(ctx->dev);
 388
 389         br_vlan_fill_forward_path_pvid(br, ctx, path);
 390
 391         f = br_fdb_find_rcu(br, ctx->daddr, path->bridge.vlan_id);
 392         if (!f || !f->dst)
 393                 return -1;
 394
 395         dst = READ_ONCE(f->dst);
 396         if (!dst)
 397                 return -1;
 398
 399         if (br_vlan_fill_forward_path_mode(br, dst, path))
 400                 return -1;
 401
 402         path->type = DEV_PATH_BRIDGE;
 403         path->dev = dst->br->dev;
 404         ctx->dev = dst->dev;
 405
 406         switch (path->bridge.vlan_mode) {
 407         case DEV_PATH_BR_VLAN_TAG:
 408                 if (ctx->num_vlans >= ARRAY_SIZE(ctx->vlan))
 409                         return -ENOSPC;
 410                 ctx->vlan[ctx->num_vlans].id = path->bridge.vlan_id;
 411                 ctx->vlan[ctx->num_vlans].proto = path->bridge.vlan_proto;
 412                 ctx->num_vlans++;
 413                 break;
 414         case DEV_PATH_BR_VLAN_UNTAG_HW:
 415         case DEV_PATH_BR_VLAN_UNTAG:
 416                 ctx->num_vlans--;
 417                 break;
 418         case DEV_PATH_BR_VLAN_KEEP:
 419                 break;
 420         }
 421
 422         return 0;
 423 }
 424
 425 static const struct ethtool_ops br_ethtool_ops = {
 426         .get_drvinfo             = br_getinfo,
 427         .get_link                = ethtool_op_get_link,
 428         .get_link_ksettings      = br_get_link_ksettings,
 429 };
 430
 431 static const struct net_device_ops br_netdev_ops = {
 432         .ndo_open                = br_dev_open,
 433         .ndo_stop                = br_dev_stop,
 434         .ndo_init                = br_dev_init,
 435         .ndo_uninit              = br_dev_uninit,
 436         .ndo_start_xmit          = br_dev_xmit,
 437         .ndo_get_stats64         = dev_get_tstats64,
 438         .ndo_set_mac_address     = br_set_mac_address,
 439         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 440         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 441         .ndo_change_mtu          = br_change_mtu,
 442         .ndo_siocdevprivate      = br_dev_siocdevprivate,
 443 #ifdef CONFIG_NET_POLL_CONTROLLER
 444         .ndo_netpoll_setup       = br_netpoll_setup,
 445         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 446         .ndo_poll_controller     = br_poll_controller,
 447 #endif
 448         .ndo_add_slave           = br_add_slave,
 449         .ndo_del_slave           = br_del_slave,
 450         .ndo_fix_features        = br_fix_features,
 451         .ndo_fdb_add             = br_fdb_add,
 452         .ndo_fdb_del             = br_fdb_delete,
 453         .ndo_fdb_del_bulk        = br_fdb_delete_bulk,
 454         .ndo_fdb_dump            = br_fdb_dump,
 455         .ndo_fdb_get             = br_fdb_get,
 456         .ndo_mdb_add             = br_mdb_add,
 457         .ndo_mdb_del             = br_mdb_del,
 458         .ndo_mdb_del_bulk        = br_mdb_del_bulk,
 459         .ndo_mdb_dump            = br_mdb_dump,
 460         .ndo_mdb_get             = br_mdb_get,
 461         .ndo_bridge_getlink      = br_getlink,
 462         .ndo_bridge_setlink      = br_setlink,
 463         .ndo_bridge_dellink      = br_dellink,
 464         .ndo_features_check      = passthru_features_check,
 465         .ndo_fill_forward_path   = br_fill_forward_path,
 466 };
 467
 468 static const struct device_type br_type = {
 469         .name   = "bridge",
 470 };
 471
 472 void br_dev_setup(struct net_device *dev)
 473 {
 474         struct net_bridge *br = netdev_priv(dev);
 475
 476         eth_hw_addr_random(dev);
 477         ether_setup(dev);
 478
 479         dev->netdev_ops = &br_netdev_ops;
 480         dev->needs_free_netdev = true;
 481         dev->ethtool_ops = &br_ethtool_ops;
 482         SET_NETDEV_DEVTYPE(dev, &br_type);
 483         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 484
 485         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 486                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 487         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 488                            NETIF_F_HW_VLAN_STAG_TX;
 489         dev->vlan_features = COMMON_FEATURES;
 490         dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
 491
 492         br->dev = dev;
 493         spin_lock_init(&br->lock);
 494         INIT_LIST_HEAD(&br->port_list);
 495         INIT_HLIST_HEAD(&br->fdb_list);
 496         INIT_HLIST_HEAD(&br->frame_type_list);
 497 #if IS_ENABLED(CONFIG_BRIDGE_MRP)
 498         INIT_HLIST_HEAD(&br->mrp_list);
 499 #endif
 500 #if IS_ENABLED(CONFIG_BRIDGE_CFM)
 501         INIT_HLIST_HEAD(&br->mep_list);
 502 #endif
 503         spin_lock_init(&br->hash_lock);
 504
 505         br->bridge_id.prio[0] = 0x80;
 506         br->bridge_id.prio[1] = 0x00;
 507
 508         ether_addr_copy(br->group_addr, eth_stp_addr);
 509
 510         br->stp_enabled = BR_NO_STP;
 511         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 512         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 513
 514         br->designated_root = br->bridge_id;
 515         br->bridge_max_age = br->max_age = 20 * HZ;
 516         br->bridge_hello_time = br->hello_time = 2 * HZ;
 517         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 518         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 519         dev->max_mtu = ETH_MAX_MTU;
 520
 521         br_netfilter_rtable_init(br);
 522         br_stp_timer_init(br);
 523         br_multicast_init(br);
 524         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 525 }