1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Handle incoming frames
4 * Linux ethernet bridge
7 * Lennert Buytenhek <buytenh@gnu.org>
10 #include <linux/slab.h>
11 #include <linux/kernel.h>
12 #include <linux/netdevice.h>
13 #include <linux/etherdevice.h>
14 #include <linux/netfilter_bridge.h>
15 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
16 #include <net/netfilter/nf_queue.h>
18 #include <linux/neighbour.h>
20 #include <linux/export.h>
21 #include <linux/rculist.h>
22 #include "br_private.h"
23 #include "br_private_tunnel.h"
26 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
28 br_drop_fake_rtable(skb);
29 return netif_receive_skb(skb);
32 static int br_pass_frame_up(struct sk_buff *skb)
34 struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
35 struct net_bridge *br = netdev_priv(brdev);
36 struct net_bridge_vlan_group *vg;
37 struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
39 u64_stats_update_begin(&brstats->syncp);
40 brstats->rx_packets++;
41 brstats->rx_bytes += skb->len;
42 u64_stats_update_end(&brstats->syncp);
44 vg = br_vlan_group_rcu(br);
45 /* Bridge is just like any other port. Make sure the
46 * packet is allowed except in promisc modue when someone
47 * may be running packet capture.
49 if (!(brdev->flags & IFF_PROMISC) &&
50 !br_allowed_egress(vg, skb)) {
57 skb = br_handle_vlan(br, NULL, vg, skb);
60 /* update the multicast stats if the packet is IGMP/MLD */
61 br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
64 return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
65 dev_net(indev), NULL, skb, indev, NULL,
66 br_netif_receive_skb);
69 /* note: already called with rcu_read_lock */
70 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
72 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
73 enum br_pkt_type pkt_type = BR_PKT_UNICAST;
74 struct net_bridge_fdb_entry *dst = NULL;
75 struct net_bridge_mdb_entry *mdst;
76 bool local_rcv, mcast_hit = false;
77 struct net_bridge *br;
80 if (!p || p->state == BR_STATE_DISABLED)
83 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
86 nbp_switchdev_frame_mark(p, skb);
88 /* insert into forwarding database after filtering to avoid spoofing */
90 if (p->flags & BR_LEARNING)
91 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
93 local_rcv = !!(br->dev->flags & IFF_PROMISC);
94 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) {
95 /* by definition the broadcast is also a multicast address */
96 if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
97 pkt_type = BR_PKT_BROADCAST;
100 pkt_type = BR_PKT_MULTICAST;
101 if (br_multicast_rcv(br, p, skb, vid))
106 if (p->state == BR_STATE_LEARNING)
109 BR_INPUT_SKB_CB(skb)->brdev = br->dev;
110 BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
112 if (IS_ENABLED(CONFIG_INET) &&
113 (skb->protocol == htons(ETH_P_ARP) ||
114 skb->protocol == htons(ETH_P_RARP))) {
115 br_do_proxy_suppress_arp(skb, br, vid, p);
116 } else if (IS_ENABLED(CONFIG_IPV6) &&
117 skb->protocol == htons(ETH_P_IPV6) &&
118 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
119 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
120 sizeof(struct nd_msg)) &&
121 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
122 struct nd_msg *msg, _msg;
124 msg = br_is_nd_neigh_msg(skb, &_msg);
126 br_do_suppress_nd(skb, br, vid, p, msg);
130 case BR_PKT_MULTICAST:
131 mdst = br_mdb_get(br, skb, vid);
132 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
133 br_multicast_querier_exists(br, eth_hdr(skb))) {
134 if ((mdst && mdst->host_joined) ||
135 br_multicast_is_router(br)) {
137 br->dev->stats.multicast++;
142 br->dev->stats.multicast++;
146 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
152 unsigned long now = jiffies;
155 return br_pass_frame_up(skb);
157 if (now != dst->used)
159 br_forward(dst->dst, skb, local_rcv, false);
162 br_flood(br, skb, pkt_type, local_rcv, false);
164 br_multicast_flood(mdst, skb, local_rcv, false);
168 return br_pass_frame_up(skb);
176 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
178 static void __br_handle_local_finish(struct sk_buff *skb)
180 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
183 /* check if vlan is allowed, to avoid spoofing */
184 if ((p->flags & BR_LEARNING) &&
185 !br_opt_get(p->br, BROPT_NO_LL_LEARN) &&
186 br_should_learn(p, skb, &vid))
187 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
190 /* note: already called with rcu_read_lock */
191 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
193 __br_handle_local_finish(skb);
195 /* return 1 to signal the okfn() was called so it's ok to use the skb */
199 static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb)
201 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
202 struct nf_hook_entries *e = NULL;
203 struct nf_hook_state state;
204 unsigned int verdict, i;
208 net = dev_net(skb->dev);
209 #ifdef HAVE_JUMP_LABEL
210 if (!static_key_false(&nf_hooks_needed[NFPROTO_BRIDGE][NF_BR_PRE_ROUTING]))
214 e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]);
218 nf_hook_state_init(&state, NF_BR_PRE_ROUTING,
219 NFPROTO_BRIDGE, skb->dev, NULL, NULL,
220 net, br_handle_frame_finish);
222 for (i = 0; i < e->num_hook_entries; i++) {
223 verdict = nf_hook_entry_hookfn(&e->hooks[i], skb, &state);
224 switch (verdict & NF_VERDICT_MASK) {
226 if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) {
228 return RX_HANDLER_PASS;
233 return RX_HANDLER_CONSUMED;
235 ret = nf_queue(skb, &state, i, verdict);
238 return RX_HANDLER_CONSUMED;
239 default: /* STOLEN */
240 return RX_HANDLER_CONSUMED;
244 net = dev_net(skb->dev);
245 br_handle_frame_finish(net, NULL, skb);
247 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
249 return RX_HANDLER_CONSUMED;
253 * Return NULL if skb is handled
254 * note: already called with rcu_read_lock
256 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
258 struct net_bridge_port *p;
259 struct sk_buff *skb = *pskb;
260 const unsigned char *dest = eth_hdr(skb)->h_dest;
262 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
263 return RX_HANDLER_PASS;
265 if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
268 skb = skb_share_check(skb, GFP_ATOMIC);
270 return RX_HANDLER_CONSUMED;
272 memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
274 p = br_port_get_rcu(skb->dev);
275 if (p->flags & BR_VLAN_TUNNEL) {
276 if (br_handle_ingress_vlan_tunnel(skb, p,
277 nbp_vlan_group_rcu(p)))
281 if (unlikely(is_link_local_ether_addr(dest))) {
282 u16 fwd_mask = p->br->group_fwd_mask_required;
285 * See IEEE 802.1D Table 7-10 Reserved addresses
288 * Bridge Group Address 01-80-C2-00-00-00
289 * (MAC Control) 802.3 01-80-C2-00-00-01
290 * (Link Aggregation) 802.3 01-80-C2-00-00-02
291 * 802.1X PAE address 01-80-C2-00-00-03
293 * 802.1AB LLDP 01-80-C2-00-00-0E
295 * Others reserved for future standardization
297 fwd_mask |= p->group_fwd_mask;
299 case 0x00: /* Bridge Group Address */
300 /* If STP is turned off,
301 then must forward to keep loop detection */
302 if (p->br->stp_enabled == BR_NO_STP ||
303 fwd_mask & (1u << dest[5]))
306 __br_handle_local_finish(skb);
307 return RX_HANDLER_PASS;
309 case 0x01: /* IEEE MAC (Pause) */
312 case 0x0E: /* 802.1AB LLDP */
313 fwd_mask |= p->br->group_fwd_mask;
314 if (fwd_mask & (1u << dest[5]))
317 __br_handle_local_finish(skb);
318 return RX_HANDLER_PASS;
321 /* Allow selective forwarding for most other protocols */
322 fwd_mask |= p->br->group_fwd_mask;
323 if (fwd_mask & (1u << dest[5]))
327 /* The else clause should be hit when nf_hook():
328 * - returns < 0 (drop/error)
329 * - returns = 0 (stolen/nf_queue)
330 * Thus return 1 from the okfn() to signal the skb is ok to pass
332 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
333 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
334 br_handle_local_finish) == 1) {
335 return RX_HANDLER_PASS;
337 return RX_HANDLER_CONSUMED;
343 case BR_STATE_FORWARDING:
344 case BR_STATE_LEARNING:
345 if (ether_addr_equal(p->br->dev->dev_addr, dest))
346 skb->pkt_type = PACKET_HOST;
348 return nf_hook_bridge_pre(skb, pskb);
353 return RX_HANDLER_CONSUMED;