net/bridge/br_device.c

   1 /*
   2  *      Device handling code
   3  *      Linux ethernet bridge
   4  *
   5  *      Authors:
   6  *      Lennert Buytenhek               <buytenh@gnu.org>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 #include <linux/kernel.h>
  15 #include <linux/netdevice.h>
  16 #include <linux/netpoll.h>
  17 #include <linux/etherdevice.h>
  18 #include <linux/ethtool.h>
  19 #include <linux/list.h>
  20 #include <linux/netfilter_bridge.h>
  21
  22 #include <linux/uaccess.h>
  23 #include "br_private.h"
  24
  25 #define COMMON_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA | \
  26                          NETIF_F_GSO_MASK | NETIF_F_HW_CSUM)
  27
  28 const struct nf_br_ops __rcu *nf_br_ops __read_mostly;
  29 EXPORT_SYMBOL_GPL(nf_br_ops);
  30
  31 static struct lock_class_key bridge_netdev_addr_lock_key;
  32
  33 /* net device transmit always called with BH disabled */
  34 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev)
  35 {
  36         struct net_bridge *br = netdev_priv(dev);
  37         struct net_bridge_fdb_entry *dst;
  38         struct net_bridge_mdb_entry *mdst;
  39         struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
  40         const struct nf_br_ops *nf_ops;
  41         const unsigned char *dest;
  42         struct ethhdr *eth;
  43         u16 vid = 0;
  44
  45         rcu_read_lock();
  46         nf_ops = rcu_dereference(nf_br_ops);
  47         if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
  48                 rcu_read_unlock();
  49                 return NETDEV_TX_OK;
  50         }
  51
  52         u64_stats_update_begin(&brstats->syncp);
  53         brstats->tx_packets++;
  54         brstats->tx_bytes += skb->len;
  55         u64_stats_update_end(&brstats->syncp);
  56
  57         br_switchdev_frame_unmark(skb);
  58         BR_INPUT_SKB_CB(skb)->brdev = dev;
  59
  60         skb_reset_mac_header(skb);
  61         eth = eth_hdr(skb);
  62         skb_pull(skb, ETH_HLEN);
  63
  64         if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid))
  65                 goto out;
  66
  67         if (IS_ENABLED(CONFIG_INET) &&
  68             (eth->h_proto == htons(ETH_P_ARP) ||
  69              eth->h_proto == htons(ETH_P_RARP)) &&
  70             br->neigh_suppress_enabled) {
  71                 br_do_proxy_suppress_arp(skb, br, vid, NULL);
  72         }
  73
  74         dest = eth_hdr(skb)->h_dest;
  75         if (is_broadcast_ether_addr(dest)) {
  76                 br_flood(br, skb, BR_PKT_BROADCAST, false, true);
  77         } else if (is_multicast_ether_addr(dest)) {
  78                 if (unlikely(netpoll_tx_running(dev))) {
  79                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  80                         goto out;
  81                 }
  82                 if (br_multicast_rcv(br, NULL, skb, vid)) {
  83                         kfree_skb(skb);
  84                         goto out;
  85                 }
  86
  87                 mdst = br_mdb_get(br, skb, vid);
  88                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
  89                     br_multicast_querier_exists(br, eth_hdr(skb)))
  90                         br_multicast_flood(mdst, skb, false, true);
  91                 else
  92                         br_flood(br, skb, BR_PKT_MULTICAST, false, true);
  93         } else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
  94                 br_forward(dst->dst, skb, false, true);
  95         } else {
  96                 br_flood(br, skb, BR_PKT_UNICAST, false, true);
  97         }
  98 out:
  99         rcu_read_unlock();
 100         return NETDEV_TX_OK;
 101 }
 102
 103 static void br_set_lockdep_class(struct net_device *dev)
 104 {
 105         lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
 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_vlan_init(br);
 118         if (err) {
 119                 free_percpu(br->stats);
 120                 return err;
 121         }
 122
 123         err = br_multicast_init_stats(br);
 124         if (err) {
 125                 free_percpu(br->stats);
 126                 br_vlan_flush(br);
 127         }
 128         br_set_lockdep_class(dev);
 129
 130         return err;
 131 }
 132
 133 static void br_dev_uninit(struct net_device *dev)
 134 {
 135         struct net_bridge *br = netdev_priv(dev);
 136
 137         br_multicast_dev_del(br);
 138         br_multicast_uninit_stats(br);
 139         br_vlan_flush(br);
 140         free_percpu(br->stats);
 141 }
 142
 143 static int br_dev_open(struct net_device *dev)
 144 {
 145         struct net_bridge *br = netdev_priv(dev);
 146
 147         netdev_update_features(dev);
 148         netif_start_queue(dev);
 149         br_stp_enable_bridge(br);
 150         br_multicast_open(br);
 151
 152         return 0;
 153 }
 154
 155 static void br_dev_set_multicast_list(struct net_device *dev)
 156 {
 157 }
 158
 159 static void br_dev_change_rx_flags(struct net_device *dev, int change)
 160 {
 161         if (change & IFF_PROMISC)
 162                 br_manage_promisc(netdev_priv(dev));
 163 }
 164
 165 static int br_dev_stop(struct net_device *dev)
 166 {
 167         struct net_bridge *br = netdev_priv(dev);
 168
 169         br_stp_disable_bridge(br);
 170         br_multicast_stop(br);
 171
 172         netif_stop_queue(dev);
 173
 174         return 0;
 175 }
 176
 177 static void br_get_stats64(struct net_device *dev,
 178                            struct rtnl_link_stats64 *stats)
 179 {
 180         struct net_bridge *br = netdev_priv(dev);
 181         struct pcpu_sw_netstats tmp, sum = { 0 };
 182         unsigned int cpu;
 183
 184         for_each_possible_cpu(cpu) {
 185                 unsigned int start;
 186                 const struct pcpu_sw_netstats *bstats
 187                         = per_cpu_ptr(br->stats, cpu);
 188                 do {
 189                         start = u64_stats_fetch_begin_irq(&bstats->syncp);
 190                         memcpy(&tmp, bstats, sizeof(tmp));
 191                 } while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
 192                 sum.tx_bytes   += tmp.tx_bytes;
 193                 sum.tx_packets += tmp.tx_packets;
 194                 sum.rx_bytes   += tmp.rx_bytes;
 195                 sum.rx_packets += tmp.rx_packets;
 196         }
 197
 198         stats->tx_bytes   = sum.tx_bytes;
 199         stats->tx_packets = sum.tx_packets;
 200         stats->rx_bytes   = sum.rx_bytes;
 201         stats->rx_packets = sum.rx_packets;
 202 }
 203
 204 static int br_change_mtu(struct net_device *dev, int new_mtu)
 205 {
 206         struct net_bridge *br = netdev_priv(dev);
 207         if (new_mtu > br_min_mtu(br))
 208                 return -EINVAL;
 209
 210         dev->mtu = new_mtu;
 211
 212 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 213         /* remember the MTU in the rtable for PMTU */
 214         dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
 215 #endif
 216
 217         return 0;
 218 }
 219
 220 /* Allow setting mac address to any valid ethernet address. */
 221 static int br_set_mac_address(struct net_device *dev, void *p)
 222 {
 223         struct net_bridge *br = netdev_priv(dev);
 224         struct sockaddr *addr = p;
 225
 226         if (!is_valid_ether_addr(addr->sa_data))
 227                 return -EADDRNOTAVAIL;
 228
 229         spin_lock_bh(&br->lock);
 230         if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
 231                 /* Mac address will be changed in br_stp_change_bridge_id(). */
 232                 br_stp_change_bridge_id(br, addr->sa_data);
 233         }
 234         spin_unlock_bh(&br->lock);
 235
 236         return 0;
 237 }
 238
 239 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 240 {
 241         strlcpy(info->driver, "bridge", sizeof(info->driver));
 242         strlcpy(info->version, BR_VERSION, sizeof(info->version));
 243         strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 244         strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
 245 }
 246
 247 static netdev_features_t br_fix_features(struct net_device *dev,
 248         netdev_features_t features)
 249 {
 250         struct net_bridge *br = netdev_priv(dev);
 251
 252         return br_features_recompute(br, features);
 253 }
 254
 255 #ifdef CONFIG_NET_POLL_CONTROLLER
 256 static void br_poll_controller(struct net_device *br_dev)
 257 {
 258 }
 259
 260 static void br_netpoll_cleanup(struct net_device *dev)
 261 {
 262         struct net_bridge *br = netdev_priv(dev);
 263         struct net_bridge_port *p;
 264
 265         list_for_each_entry(p, &br->port_list, list)
 266                 br_netpoll_disable(p);
 267 }
 268
 269 static int __br_netpoll_enable(struct net_bridge_port *p)
 270 {
 271         struct netpoll *np;
 272         int err;
 273
 274         np = kzalloc(sizeof(*p->np), GFP_KERNEL);
 275         if (!np)
 276                 return -ENOMEM;
 277
 278         err = __netpoll_setup(np, p->dev);
 279         if (err) {
 280                 kfree(np);
 281                 return err;
 282         }
 283
 284         p->np = np;
 285         return err;
 286 }
 287
 288 int br_netpoll_enable(struct net_bridge_port *p)
 289 {
 290         if (!p->br->dev->npinfo)
 291                 return 0;
 292
 293         return __br_netpoll_enable(p);
 294 }
 295
 296 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
 297 {
 298         struct net_bridge *br = netdev_priv(dev);
 299         struct net_bridge_port *p;
 300         int err = 0;
 301
 302         list_for_each_entry(p, &br->port_list, list) {
 303                 if (!p->dev)
 304                         continue;
 305                 err = __br_netpoll_enable(p);
 306                 if (err)
 307                         goto fail;
 308         }
 309
 310 out:
 311         return err;
 312
 313 fail:
 314         br_netpoll_cleanup(dev);
 315         goto out;
 316 }
 317
 318 void br_netpoll_disable(struct net_bridge_port *p)
 319 {
 320         struct netpoll *np = p->np;
 321
 322         if (!np)
 323                 return;
 324
 325         p->np = NULL;
 326
 327         __netpoll_free_async(np);
 328 }
 329
 330 #endif
 331
 332 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
 333                         struct netlink_ext_ack *extack)
 334
 335 {
 336         struct net_bridge *br = netdev_priv(dev);
 337
 338         return br_add_if(br, slave_dev, extack);
 339 }
 340
 341 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
 342 {
 343         struct net_bridge *br = netdev_priv(dev);
 344
 345         return br_del_if(br, slave_dev);
 346 }
 347
 348 static const struct ethtool_ops br_ethtool_ops = {
 349         .get_drvinfo    = br_getinfo,
 350         .get_link       = ethtool_op_get_link,
 351 };
 352
 353 static const struct net_device_ops br_netdev_ops = {
 354         .ndo_open                = br_dev_open,
 355         .ndo_stop                = br_dev_stop,
 356         .ndo_init                = br_dev_init,
 357         .ndo_uninit              = br_dev_uninit,
 358         .ndo_start_xmit          = br_dev_xmit,
 359         .ndo_get_stats64         = br_get_stats64,
 360         .ndo_set_mac_address     = br_set_mac_address,
 361         .ndo_set_rx_mode         = br_dev_set_multicast_list,
 362         .ndo_change_rx_flags     = br_dev_change_rx_flags,
 363         .ndo_change_mtu          = br_change_mtu,
 364         .ndo_do_ioctl            = br_dev_ioctl,
 365 #ifdef CONFIG_NET_POLL_CONTROLLER
 366         .ndo_netpoll_setup       = br_netpoll_setup,
 367         .ndo_netpoll_cleanup     = br_netpoll_cleanup,
 368         .ndo_poll_controller     = br_poll_controller,
 369 #endif
 370         .ndo_add_slave           = br_add_slave,
 371         .ndo_del_slave           = br_del_slave,
 372         .ndo_fix_features        = br_fix_features,
 373         .ndo_fdb_add             = br_fdb_add,
 374         .ndo_fdb_del             = br_fdb_delete,
 375         .ndo_fdb_dump            = br_fdb_dump,
 376         .ndo_bridge_getlink      = br_getlink,
 377         .ndo_bridge_setlink      = br_setlink,
 378         .ndo_bridge_dellink      = br_dellink,
 379         .ndo_features_check      = passthru_features_check,
 380 };
 381
 382 static struct device_type br_type = {
 383         .name   = "bridge",
 384 };
 385
 386 void br_dev_setup(struct net_device *dev)
 387 {
 388         struct net_bridge *br = netdev_priv(dev);
 389
 390         eth_hw_addr_random(dev);
 391         ether_setup(dev);
 392
 393         dev->netdev_ops = &br_netdev_ops;
 394         dev->needs_free_netdev = true;
 395         dev->ethtool_ops = &br_ethtool_ops;
 396         SET_NETDEV_DEVTYPE(dev, &br_type);
 397         dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
 398
 399         dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
 400                         NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
 401         dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
 402                            NETIF_F_HW_VLAN_STAG_TX;
 403         dev->vlan_features = COMMON_FEATURES;
 404
 405         br->dev = dev;
 406         spin_lock_init(&br->lock);
 407         INIT_LIST_HEAD(&br->port_list);
 408         spin_lock_init(&br->hash_lock);
 409
 410         br->bridge_id.prio[0] = 0x80;
 411         br->bridge_id.prio[1] = 0x00;
 412
 413         ether_addr_copy(br->group_addr, eth_reserved_addr_base);
 414
 415         br->stp_enabled = BR_NO_STP;
 416         br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
 417         br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
 418
 419         br->designated_root = br->bridge_id;
 420         br->bridge_max_age = br->max_age = 20 * HZ;
 421         br->bridge_hello_time = br->hello_time = 2 * HZ;
 422         br->bridge_forward_delay = br->forward_delay = 15 * HZ;
 423         br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 424         dev->max_mtu = ETH_MAX_MTU;
 425
 426         br_netfilter_rtable_init(br);
 427         br_stp_timer_init(br);
 428         br_multicast_init(br);
 429         INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
 430 }