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>
21 #include <linux/export.h>
22 #include <linux/rculist.h>
23 #include "br_private.h"
24 #include "br_private_tunnel.h"
27 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
29 br_drop_fake_rtable(skb);
30 return netif_receive_skb(skb);
33 static int br_pass_frame_up(struct sk_buff *skb)
35 struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
36 struct net_bridge *br = netdev_priv(brdev);
37 struct net_bridge_vlan_group *vg;
39 dev_sw_netstats_rx_add(brdev, skb->len);
41 vg = br_vlan_group_rcu(br);
43 /* Reset the offload_fwd_mark because there could be a stacked
44 * bridge above, and it should not think this bridge it doing
45 * that bridge's work forwarding out its ports.
47 br_switchdev_frame_unmark(skb);
49 /* Bridge is just like any other port. Make sure the
50 * packet is allowed except in promisc mode when someone
51 * may be running packet capture.
53 if (!(brdev->flags & IFF_PROMISC) &&
54 !br_allowed_egress(vg, skb)) {
61 skb = br_handle_vlan(br, NULL, vg, skb);
64 /* update the multicast stats if the packet is IGMP/MLD */
65 br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
68 return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
69 dev_net(indev), NULL, skb, indev, NULL,
70 br_netif_receive_skb);
73 /* note: already called with rcu_read_lock */
74 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
76 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
77 enum br_pkt_type pkt_type = BR_PKT_UNICAST;
78 struct net_bridge_fdb_entry *dst = NULL;
79 struct net_bridge_mcast_port *pmctx;
80 struct net_bridge_mdb_entry *mdst;
81 bool local_rcv, mcast_hit = false;
82 struct net_bridge_mcast *brmctx;
83 struct net_bridge_vlan *vlan;
84 struct net_bridge *br;
88 if (!p || p->state == BR_STATE_DISABLED)
91 brmctx = &p->br->multicast_ctx;
92 pmctx = &p->multicast_ctx;
94 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid,
98 nbp_switchdev_frame_mark(p, skb);
100 /* insert into forwarding database after filtering to avoid spoofing */
102 if (p->flags & BR_LEARNING)
103 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, 0);
105 local_rcv = !!(br->dev->flags & IFF_PROMISC);
106 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) {
107 /* by definition the broadcast is also a multicast address */
108 if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
109 pkt_type = BR_PKT_BROADCAST;
112 pkt_type = BR_PKT_MULTICAST;
113 if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid))
118 if (state == BR_STATE_LEARNING)
121 BR_INPUT_SKB_CB(skb)->brdev = br->dev;
122 BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
124 if (IS_ENABLED(CONFIG_INET) &&
125 (skb->protocol == htons(ETH_P_ARP) ||
126 skb->protocol == htons(ETH_P_RARP))) {
127 br_do_proxy_suppress_arp(skb, br, vid, p);
128 } else if (IS_ENABLED(CONFIG_IPV6) &&
129 skb->protocol == htons(ETH_P_IPV6) &&
130 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
131 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
132 sizeof(struct nd_msg)) &&
133 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
134 struct nd_msg *msg, _msg;
136 msg = br_is_nd_neigh_msg(skb, &_msg);
138 br_do_suppress_nd(skb, br, vid, p, msg);
142 case BR_PKT_MULTICAST:
143 mdst = br_mdb_get(brmctx, skb, vid);
144 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
145 br_multicast_querier_exists(brmctx, eth_hdr(skb), mdst)) {
146 if ((mdst && mdst->host_joined) ||
147 br_multicast_is_router(brmctx, skb)) {
149 br->dev->stats.multicast++;
154 br->dev->stats.multicast++;
158 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
165 unsigned long now = jiffies;
167 if (test_bit(BR_FDB_LOCAL, &dst->flags))
168 return br_pass_frame_up(skb);
170 if (now != dst->used)
172 br_forward(dst->dst, skb, local_rcv, false);
175 br_flood(br, skb, pkt_type, local_rcv, false);
177 br_multicast_flood(mdst, skb, brmctx, local_rcv, false);
181 return br_pass_frame_up(skb);
189 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
191 static void __br_handle_local_finish(struct sk_buff *skb)
193 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
196 /* check if vlan is allowed, to avoid spoofing */
197 if ((p->flags & BR_LEARNING) &&
198 nbp_state_should_learn(p) &&
199 !br_opt_get(p->br, BROPT_NO_LL_LEARN) &&
200 br_should_learn(p, skb, &vid))
201 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, 0);
204 /* note: already called with rcu_read_lock */
205 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
207 __br_handle_local_finish(skb);
209 /* return 1 to signal the okfn() was called so it's ok to use the skb */
213 static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb)
215 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
216 struct nf_hook_entries *e = NULL;
217 struct nf_hook_state state;
218 unsigned int verdict, i;
222 net = dev_net(skb->dev);
223 #ifdef HAVE_JUMP_LABEL
224 if (!static_key_false(&nf_hooks_needed[NFPROTO_BRIDGE][NF_BR_PRE_ROUTING]))
228 e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]);
232 nf_hook_state_init(&state, NF_BR_PRE_ROUTING,
233 NFPROTO_BRIDGE, skb->dev, NULL, NULL,
234 net, br_handle_frame_finish);
236 for (i = 0; i < e->num_hook_entries; i++) {
237 verdict = nf_hook_entry_hookfn(&e->hooks[i], skb, &state);
238 switch (verdict & NF_VERDICT_MASK) {
240 if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) {
242 return RX_HANDLER_PASS;
247 return RX_HANDLER_CONSUMED;
249 ret = nf_queue(skb, &state, i, verdict);
252 return RX_HANDLER_CONSUMED;
253 default: /* STOLEN */
254 return RX_HANDLER_CONSUMED;
258 net = dev_net(skb->dev);
259 br_handle_frame_finish(net, NULL, skb);
261 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
263 return RX_HANDLER_CONSUMED;
266 /* Return 0 if the frame was not processed otherwise 1
267 * note: already called with rcu_read_lock
269 static int br_process_frame_type(struct net_bridge_port *p,
272 struct br_frame_type *tmp;
274 hlist_for_each_entry_rcu(tmp, &p->br->frame_type_list, list)
275 if (unlikely(tmp->type == skb->protocol))
276 return tmp->frame_handler(p, skb);
282 * Return NULL if skb is handled
283 * note: already called with rcu_read_lock
285 static rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
287 struct net_bridge_port *p;
288 struct sk_buff *skb = *pskb;
289 const unsigned char *dest = eth_hdr(skb)->h_dest;
291 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
292 return RX_HANDLER_PASS;
294 if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
297 skb = skb_share_check(skb, GFP_ATOMIC);
299 return RX_HANDLER_CONSUMED;
301 memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
303 p = br_port_get_rcu(skb->dev);
304 if (p->flags & BR_VLAN_TUNNEL)
305 br_handle_ingress_vlan_tunnel(skb, p, nbp_vlan_group_rcu(p));
307 if (unlikely(is_link_local_ether_addr(dest))) {
308 u16 fwd_mask = p->br->group_fwd_mask_required;
311 * See IEEE 802.1D Table 7-10 Reserved addresses
314 * Bridge Group Address 01-80-C2-00-00-00
315 * (MAC Control) 802.3 01-80-C2-00-00-01
316 * (Link Aggregation) 802.3 01-80-C2-00-00-02
317 * 802.1X PAE address 01-80-C2-00-00-03
319 * 802.1AB LLDP 01-80-C2-00-00-0E
321 * Others reserved for future standardization
323 fwd_mask |= p->group_fwd_mask;
325 case 0x00: /* Bridge Group Address */
326 /* If STP is turned off,
327 then must forward to keep loop detection */
328 if (p->br->stp_enabled == BR_NO_STP ||
329 fwd_mask & (1u << dest[5]))
332 __br_handle_local_finish(skb);
333 return RX_HANDLER_PASS;
335 case 0x01: /* IEEE MAC (Pause) */
338 case 0x0E: /* 802.1AB LLDP */
339 fwd_mask |= p->br->group_fwd_mask;
340 if (fwd_mask & (1u << dest[5]))
343 __br_handle_local_finish(skb);
344 return RX_HANDLER_PASS;
347 /* Allow selective forwarding for most other protocols */
348 fwd_mask |= p->br->group_fwd_mask;
349 if (fwd_mask & (1u << dest[5]))
353 /* The else clause should be hit when nf_hook():
354 * - returns < 0 (drop/error)
355 * - returns = 0 (stolen/nf_queue)
356 * Thus return 1 from the okfn() to signal the skb is ok to pass
358 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
359 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
360 br_handle_local_finish) == 1) {
361 return RX_HANDLER_PASS;
363 return RX_HANDLER_CONSUMED;
367 if (unlikely(br_process_frame_type(p, skb)))
368 return RX_HANDLER_PASS;
372 case BR_STATE_FORWARDING:
373 case BR_STATE_LEARNING:
374 if (ether_addr_equal(p->br->dev->dev_addr, dest))
375 skb->pkt_type = PACKET_HOST;
377 return nf_hook_bridge_pre(skb, pskb);
382 return RX_HANDLER_CONSUMED;
385 /* This function has no purpose other than to appease the br_port_get_rcu/rtnl
386 * helpers which identify bridged ports according to the rx_handler installed
387 * on them (so there _needs_ to be a bridge rx_handler even if we don't need it
388 * to do anything useful). This bridge won't support traffic to/from the stack,
389 * but only hardware bridging. So return RX_HANDLER_PASS so we don't steal
390 * frames from the ETH_P_XDSA packet_type handler.
392 static rx_handler_result_t br_handle_frame_dummy(struct sk_buff **pskb)
394 return RX_HANDLER_PASS;
397 rx_handler_func_t *br_get_rx_handler(const struct net_device *dev)
399 if (netdev_uses_dsa(dev))
400 return br_handle_frame_dummy;
402 return br_handle_frame;
405 void br_add_frame(struct net_bridge *br, struct br_frame_type *ft)
407 hlist_add_head_rcu(&ft->list, &br->frame_type_list);
410 void br_del_frame(struct net_bridge *br, struct br_frame_type *ft)
412 struct br_frame_type *tmp;
414 hlist_for_each_entry(tmp, &br->frame_type_list, list)
416 hlist_del_rcu(&ft->list);