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;
93 if (br_mst_is_enabled(br)) {
94 state = BR_STATE_FORWARDING;
96 if (p->state == BR_STATE_DISABLED)
102 brmctx = &p->br->multicast_ctx;
103 pmctx = &p->multicast_ctx;
104 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid,
108 if (p->flags & BR_PORT_LOCKED) {
109 struct net_bridge_fdb_entry *fdb_src =
110 br_fdb_find_rcu(br, eth_hdr(skb)->h_source, vid);
112 if (!fdb_src || READ_ONCE(fdb_src->dst) != p ||
113 test_bit(BR_FDB_LOCAL, &fdb_src->flags))
117 nbp_switchdev_frame_mark(p, skb);
119 /* insert into forwarding database after filtering to avoid spoofing */
120 if (p->flags & BR_LEARNING)
121 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, 0);
123 local_rcv = !!(br->dev->flags & IFF_PROMISC);
124 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) {
125 /* by definition the broadcast is also a multicast address */
126 if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
127 pkt_type = BR_PKT_BROADCAST;
130 pkt_type = BR_PKT_MULTICAST;
131 if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid))
136 if (state == BR_STATE_LEARNING)
139 BR_INPUT_SKB_CB(skb)->brdev = br->dev;
140 BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
142 if (IS_ENABLED(CONFIG_INET) &&
143 (skb->protocol == htons(ETH_P_ARP) ||
144 skb->protocol == htons(ETH_P_RARP))) {
145 br_do_proxy_suppress_arp(skb, br, vid, p);
146 } else if (IS_ENABLED(CONFIG_IPV6) &&
147 skb->protocol == htons(ETH_P_IPV6) &&
148 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
149 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
150 sizeof(struct nd_msg)) &&
151 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
152 struct nd_msg *msg, _msg;
154 msg = br_is_nd_neigh_msg(skb, &_msg);
156 br_do_suppress_nd(skb, br, vid, p, msg);
160 case BR_PKT_MULTICAST:
161 mdst = br_mdb_get(brmctx, skb, vid);
162 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
163 br_multicast_querier_exists(brmctx, eth_hdr(skb), mdst)) {
164 if ((mdst && mdst->host_joined) ||
165 br_multicast_is_router(brmctx, skb)) {
167 br->dev->stats.multicast++;
172 br->dev->stats.multicast++;
176 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
183 unsigned long now = jiffies;
185 if (test_bit(BR_FDB_LOCAL, &dst->flags))
186 return br_pass_frame_up(skb);
188 if (now != dst->used)
190 br_forward(dst->dst, skb, local_rcv, false);
193 br_flood(br, skb, pkt_type, local_rcv, false);
195 br_multicast_flood(mdst, skb, brmctx, local_rcv, false);
199 return br_pass_frame_up(skb);
207 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
209 static void __br_handle_local_finish(struct sk_buff *skb)
211 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
214 /* check if vlan is allowed, to avoid spoofing */
215 if ((p->flags & BR_LEARNING) &&
216 nbp_state_should_learn(p) &&
217 !br_opt_get(p->br, BROPT_NO_LL_LEARN) &&
218 br_should_learn(p, skb, &vid))
219 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, 0);
222 /* note: already called with rcu_read_lock */
223 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
225 __br_handle_local_finish(skb);
227 /* return 1 to signal the okfn() was called so it's ok to use the skb */
231 static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb)
233 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
234 struct nf_hook_entries *e = NULL;
235 struct nf_hook_state state;
236 unsigned int verdict, i;
240 net = dev_net(skb->dev);
241 #ifdef HAVE_JUMP_LABEL
242 if (!static_key_false(&nf_hooks_needed[NFPROTO_BRIDGE][NF_BR_PRE_ROUTING]))
246 e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]);
250 nf_hook_state_init(&state, NF_BR_PRE_ROUTING,
251 NFPROTO_BRIDGE, skb->dev, NULL, NULL,
252 net, br_handle_frame_finish);
254 for (i = 0; i < e->num_hook_entries; i++) {
255 verdict = nf_hook_entry_hookfn(&e->hooks[i], skb, &state);
256 switch (verdict & NF_VERDICT_MASK) {
258 if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) {
260 return RX_HANDLER_PASS;
265 return RX_HANDLER_CONSUMED;
267 ret = nf_queue(skb, &state, i, verdict);
270 return RX_HANDLER_CONSUMED;
271 default: /* STOLEN */
272 return RX_HANDLER_CONSUMED;
276 net = dev_net(skb->dev);
277 br_handle_frame_finish(net, NULL, skb);
279 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
281 return RX_HANDLER_CONSUMED;
284 /* Return 0 if the frame was not processed otherwise 1
285 * note: already called with rcu_read_lock
287 static int br_process_frame_type(struct net_bridge_port *p,
290 struct br_frame_type *tmp;
292 hlist_for_each_entry_rcu(tmp, &p->br->frame_type_list, list)
293 if (unlikely(tmp->type == skb->protocol))
294 return tmp->frame_handler(p, skb);
300 * Return NULL if skb is handled
301 * note: already called with rcu_read_lock
303 static rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
305 struct net_bridge_port *p;
306 struct sk_buff *skb = *pskb;
307 const unsigned char *dest = eth_hdr(skb)->h_dest;
309 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
310 return RX_HANDLER_PASS;
312 if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
315 skb = skb_share_check(skb, GFP_ATOMIC);
317 return RX_HANDLER_CONSUMED;
319 memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
321 p = br_port_get_rcu(skb->dev);
322 if (p->flags & BR_VLAN_TUNNEL)
323 br_handle_ingress_vlan_tunnel(skb, p, nbp_vlan_group_rcu(p));
325 if (unlikely(is_link_local_ether_addr(dest))) {
326 u16 fwd_mask = p->br->group_fwd_mask_required;
329 * See IEEE 802.1D Table 7-10 Reserved addresses
332 * Bridge Group Address 01-80-C2-00-00-00
333 * (MAC Control) 802.3 01-80-C2-00-00-01
334 * (Link Aggregation) 802.3 01-80-C2-00-00-02
335 * 802.1X PAE address 01-80-C2-00-00-03
337 * 802.1AB LLDP 01-80-C2-00-00-0E
339 * Others reserved for future standardization
341 fwd_mask |= p->group_fwd_mask;
343 case 0x00: /* Bridge Group Address */
344 /* If STP is turned off,
345 then must forward to keep loop detection */
346 if (p->br->stp_enabled == BR_NO_STP ||
347 fwd_mask & (1u << dest[5]))
350 __br_handle_local_finish(skb);
351 return RX_HANDLER_PASS;
353 case 0x01: /* IEEE MAC (Pause) */
356 case 0x0E: /* 802.1AB LLDP */
357 fwd_mask |= p->br->group_fwd_mask;
358 if (fwd_mask & (1u << dest[5]))
361 __br_handle_local_finish(skb);
362 return RX_HANDLER_PASS;
365 /* Allow selective forwarding for most other protocols */
366 fwd_mask |= p->br->group_fwd_mask;
367 if (fwd_mask & (1u << dest[5]))
371 /* The else clause should be hit when nf_hook():
372 * - returns < 0 (drop/error)
373 * - returns = 0 (stolen/nf_queue)
374 * Thus return 1 from the okfn() to signal the skb is ok to pass
376 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
377 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
378 br_handle_local_finish) == 1) {
379 return RX_HANDLER_PASS;
381 return RX_HANDLER_CONSUMED;
385 if (unlikely(br_process_frame_type(p, skb)))
386 return RX_HANDLER_PASS;
389 if (br_mst_is_enabled(p->br))
390 goto defer_stp_filtering;
393 case BR_STATE_FORWARDING:
394 case BR_STATE_LEARNING:
396 if (ether_addr_equal(p->br->dev->dev_addr, dest))
397 skb->pkt_type = PACKET_HOST;
399 return nf_hook_bridge_pre(skb, pskb);
404 return RX_HANDLER_CONSUMED;
407 /* This function has no purpose other than to appease the br_port_get_rcu/rtnl
408 * helpers which identify bridged ports according to the rx_handler installed
409 * on them (so there _needs_ to be a bridge rx_handler even if we don't need it
410 * to do anything useful). This bridge won't support traffic to/from the stack,
411 * but only hardware bridging. So return RX_HANDLER_PASS so we don't steal
412 * frames from the ETH_P_XDSA packet_type handler.
414 static rx_handler_result_t br_handle_frame_dummy(struct sk_buff **pskb)
416 return RX_HANDLER_PASS;
419 rx_handler_func_t *br_get_rx_handler(const struct net_device *dev)
421 if (netdev_uses_dsa(dev))
422 return br_handle_frame_dummy;
424 return br_handle_frame;
427 void br_add_frame(struct net_bridge *br, struct br_frame_type *ft)
429 hlist_add_head_rcu(&ft->list, &br->frame_type_list);
432 void br_del_frame(struct net_bridge *br, struct br_frame_type *ft)
434 struct br_frame_type *tmp;
436 hlist_for_each_entry(tmp, &br->frame_type_list, list)
438 hlist_del_rcu(&ft->list);