1 // SPDX-License-Identifier: GPL-2.0-or-later
4 * Linux ethernet bridge
7 * Lennert Buytenhek <buytenh@gnu.org>
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>
18 #include <linux/uaccess.h>
19 #include "br_private.h"
21 #define COMMON_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA | \
22 NETIF_F_GSO_MASK | NETIF_F_HW_CSUM)
24 const struct nf_br_ops __rcu *nf_br_ops __read_mostly;
25 EXPORT_SYMBOL_GPL(nf_br_ops);
27 /* net device transmit always called with BH disabled */
28 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev)
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;
41 memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
42 br_tc_skb_miss_set(skb, false);
45 nf_ops = rcu_dereference(nf_br_ops);
46 if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
51 dev_sw_netstats_tx_add(dev, 1, skb->len);
53 br_switchdev_frame_unmark(skb);
54 BR_INPUT_SKB_CB(skb)->brdev = dev;
55 BR_INPUT_SKB_CB(skb)->frag_max_size = 0;
57 skb_reset_mac_header(skb);
58 skb_pull(skb, ETH_HLEN);
60 if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid,
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;
77 msg = br_is_nd_neigh_msg(skb, &_msg);
79 br_do_suppress_nd(skb, br, vid, NULL, msg);
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);
90 if (br_multicast_rcv(&brmctx, &pmctx_null, vlan, skb, vid)) {
95 mdst = br_mdb_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);
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);
104 br_flood(br, skb, BR_PKT_UNICAST, false, true, vid);
111 static struct lock_class_key bridge_netdev_addr_lock_key;
113 static void br_set_lockdep_class(struct net_device *dev)
115 lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
118 static int br_dev_init(struct net_device *dev)
120 struct net_bridge *br = netdev_priv(dev);
123 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
127 err = br_fdb_hash_init(br);
129 free_percpu(dev->tstats);
133 err = br_mdb_hash_init(br);
135 free_percpu(dev->tstats);
136 br_fdb_hash_fini(br);
140 err = br_vlan_init(br);
142 free_percpu(dev->tstats);
143 br_mdb_hash_fini(br);
144 br_fdb_hash_fini(br);
148 err = br_multicast_init_stats(br);
150 free_percpu(dev->tstats);
152 br_mdb_hash_fini(br);
153 br_fdb_hash_fini(br);
156 br_set_lockdep_class(dev);
160 static void br_dev_uninit(struct net_device *dev)
162 struct net_bridge *br = netdev_priv(dev);
164 br_multicast_dev_del(br);
165 br_multicast_uninit_stats(br);
167 br_mdb_hash_fini(br);
168 br_fdb_hash_fini(br);
169 free_percpu(dev->tstats);
172 static int br_dev_open(struct net_device *dev)
174 struct net_bridge *br = netdev_priv(dev);
176 netdev_update_features(dev);
177 netif_start_queue(dev);
178 br_stp_enable_bridge(br);
179 br_multicast_open(br);
181 if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
182 br_multicast_join_snoopers(br);
187 static void br_dev_set_multicast_list(struct net_device *dev)
191 static void br_dev_change_rx_flags(struct net_device *dev, int change)
193 if (change & IFF_PROMISC)
194 br_manage_promisc(netdev_priv(dev));
197 static int br_dev_stop(struct net_device *dev)
199 struct net_bridge *br = netdev_priv(dev);
201 br_stp_disable_bridge(br);
202 br_multicast_stop(br);
204 if (br_opt_get(br, BROPT_MULTICAST_ENABLED))
205 br_multicast_leave_snoopers(br);
207 netif_stop_queue(dev);
212 static int br_change_mtu(struct net_device *dev, int new_mtu)
214 struct net_bridge *br = netdev_priv(dev);
218 /* this flag will be cleared if the MTU was automatically adjusted */
219 br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
220 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
221 /* remember the MTU in the rtable for PMTU */
222 dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
228 /* Allow setting mac address to any valid ethernet address. */
229 static int br_set_mac_address(struct net_device *dev, void *p)
231 struct net_bridge *br = netdev_priv(dev);
232 struct sockaddr *addr = p;
234 if (!is_valid_ether_addr(addr->sa_data))
235 return -EADDRNOTAVAIL;
237 /* dev_set_mac_addr() can be called by a master device on bridge's
238 * NETDEV_UNREGISTER, but since it's being destroyed do nothing
240 if (dev->reg_state != NETREG_REGISTERED)
243 spin_lock_bh(&br->lock);
244 if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
245 /* Mac address will be changed in br_stp_change_bridge_id(). */
246 br_stp_change_bridge_id(br, addr->sa_data);
248 spin_unlock_bh(&br->lock);
253 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
255 strscpy(info->driver, "bridge", sizeof(info->driver));
256 strscpy(info->version, BR_VERSION, sizeof(info->version));
257 strscpy(info->fw_version, "N/A", sizeof(info->fw_version));
258 strscpy(info->bus_info, "N/A", sizeof(info->bus_info));
261 static int br_get_link_ksettings(struct net_device *dev,
262 struct ethtool_link_ksettings *cmd)
264 struct net_bridge *br = netdev_priv(dev);
265 struct net_bridge_port *p;
267 cmd->base.duplex = DUPLEX_UNKNOWN;
268 cmd->base.port = PORT_OTHER;
269 cmd->base.speed = SPEED_UNKNOWN;
271 list_for_each_entry(p, &br->port_list, list) {
272 struct ethtool_link_ksettings ecmd;
273 struct net_device *pdev = p->dev;
275 if (!netif_running(pdev) || !netif_oper_up(pdev))
278 if (__ethtool_get_link_ksettings(pdev, &ecmd))
281 if (ecmd.base.speed == (__u32)SPEED_UNKNOWN)
284 if (cmd->base.speed == (__u32)SPEED_UNKNOWN ||
285 cmd->base.speed < ecmd.base.speed)
286 cmd->base.speed = ecmd.base.speed;
292 static netdev_features_t br_fix_features(struct net_device *dev,
293 netdev_features_t features)
295 struct net_bridge *br = netdev_priv(dev);
297 return br_features_recompute(br, features);
300 #ifdef CONFIG_NET_POLL_CONTROLLER
301 static void br_poll_controller(struct net_device *br_dev)
305 static void br_netpoll_cleanup(struct net_device *dev)
307 struct net_bridge *br = netdev_priv(dev);
308 struct net_bridge_port *p;
310 list_for_each_entry(p, &br->port_list, list)
311 br_netpoll_disable(p);
314 static int __br_netpoll_enable(struct net_bridge_port *p)
319 np = kzalloc(sizeof(*p->np), GFP_KERNEL);
323 err = __netpoll_setup(np, p->dev);
333 int br_netpoll_enable(struct net_bridge_port *p)
335 if (!p->br->dev->npinfo)
338 return __br_netpoll_enable(p);
341 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
343 struct net_bridge *br = netdev_priv(dev);
344 struct net_bridge_port *p;
347 list_for_each_entry(p, &br->port_list, list) {
350 err = __br_netpoll_enable(p);
359 br_netpoll_cleanup(dev);
363 void br_netpoll_disable(struct net_bridge_port *p)
365 struct netpoll *np = p->np;
377 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
378 struct netlink_ext_ack *extack)
381 struct net_bridge *br = netdev_priv(dev);
383 return br_add_if(br, slave_dev, extack);
386 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
388 struct net_bridge *br = netdev_priv(dev);
390 return br_del_if(br, slave_dev);
393 static int br_fill_forward_path(struct net_device_path_ctx *ctx,
394 struct net_device_path *path)
396 struct net_bridge_fdb_entry *f;
397 struct net_bridge_port *dst;
398 struct net_bridge *br;
400 if (netif_is_bridge_port(ctx->dev))
403 br = netdev_priv(ctx->dev);
405 br_vlan_fill_forward_path_pvid(br, ctx, path);
407 f = br_fdb_find_rcu(br, ctx->daddr, path->bridge.vlan_id);
411 dst = READ_ONCE(f->dst);
415 if (br_vlan_fill_forward_path_mode(br, dst, path))
418 path->type = DEV_PATH_BRIDGE;
419 path->dev = dst->br->dev;
422 switch (path->bridge.vlan_mode) {
423 case DEV_PATH_BR_VLAN_TAG:
424 if (ctx->num_vlans >= ARRAY_SIZE(ctx->vlan))
426 ctx->vlan[ctx->num_vlans].id = path->bridge.vlan_id;
427 ctx->vlan[ctx->num_vlans].proto = path->bridge.vlan_proto;
430 case DEV_PATH_BR_VLAN_UNTAG_HW:
431 case DEV_PATH_BR_VLAN_UNTAG:
434 case DEV_PATH_BR_VLAN_KEEP:
441 static const struct ethtool_ops br_ethtool_ops = {
442 .get_drvinfo = br_getinfo,
443 .get_link = ethtool_op_get_link,
444 .get_link_ksettings = br_get_link_ksettings,
447 static const struct net_device_ops br_netdev_ops = {
448 .ndo_open = br_dev_open,
449 .ndo_stop = br_dev_stop,
450 .ndo_init = br_dev_init,
451 .ndo_uninit = br_dev_uninit,
452 .ndo_start_xmit = br_dev_xmit,
453 .ndo_get_stats64 = dev_get_tstats64,
454 .ndo_set_mac_address = br_set_mac_address,
455 .ndo_set_rx_mode = br_dev_set_multicast_list,
456 .ndo_change_rx_flags = br_dev_change_rx_flags,
457 .ndo_change_mtu = br_change_mtu,
458 .ndo_siocdevprivate = br_dev_siocdevprivate,
459 #ifdef CONFIG_NET_POLL_CONTROLLER
460 .ndo_netpoll_setup = br_netpoll_setup,
461 .ndo_netpoll_cleanup = br_netpoll_cleanup,
462 .ndo_poll_controller = br_poll_controller,
464 .ndo_add_slave = br_add_slave,
465 .ndo_del_slave = br_del_slave,
466 .ndo_fix_features = br_fix_features,
467 .ndo_fdb_add = br_fdb_add,
468 .ndo_fdb_del = br_fdb_delete,
469 .ndo_fdb_del_bulk = br_fdb_delete_bulk,
470 .ndo_fdb_dump = br_fdb_dump,
471 .ndo_fdb_get = br_fdb_get,
472 .ndo_mdb_add = br_mdb_add,
473 .ndo_mdb_del = br_mdb_del,
474 .ndo_mdb_dump = br_mdb_dump,
475 .ndo_bridge_getlink = br_getlink,
476 .ndo_bridge_setlink = br_setlink,
477 .ndo_bridge_dellink = br_dellink,
478 .ndo_features_check = passthru_features_check,
479 .ndo_fill_forward_path = br_fill_forward_path,
482 static struct device_type br_type = {
486 void br_dev_setup(struct net_device *dev)
488 struct net_bridge *br = netdev_priv(dev);
490 eth_hw_addr_random(dev);
493 dev->netdev_ops = &br_netdev_ops;
494 dev->needs_free_netdev = true;
495 dev->ethtool_ops = &br_ethtool_ops;
496 SET_NETDEV_DEVTYPE(dev, &br_type);
497 dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
499 dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
500 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
501 dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
502 NETIF_F_HW_VLAN_STAG_TX;
503 dev->vlan_features = COMMON_FEATURES;
506 spin_lock_init(&br->lock);
507 INIT_LIST_HEAD(&br->port_list);
508 INIT_HLIST_HEAD(&br->fdb_list);
509 INIT_HLIST_HEAD(&br->frame_type_list);
510 #if IS_ENABLED(CONFIG_BRIDGE_MRP)
511 INIT_HLIST_HEAD(&br->mrp_list);
513 #if IS_ENABLED(CONFIG_BRIDGE_CFM)
514 INIT_HLIST_HEAD(&br->mep_list);
516 spin_lock_init(&br->hash_lock);
518 br->bridge_id.prio[0] = 0x80;
519 br->bridge_id.prio[1] = 0x00;
521 ether_addr_copy(br->group_addr, eth_stp_addr);
523 br->stp_enabled = BR_NO_STP;
524 br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
525 br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
527 br->designated_root = br->bridge_id;
528 br->bridge_max_age = br->max_age = 20 * HZ;
529 br->bridge_hello_time = br->hello_time = 2 * HZ;
530 br->bridge_forward_delay = br->forward_delay = 15 * HZ;
531 br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
532 dev->max_mtu = ETH_MAX_MTU;
534 br_netfilter_rtable_init(br);
535 br_stp_timer_init(br);
536 br_multicast_init(br);
537 INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);