1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Handle incoming frames
4 * Linux ethernet bridge
5 *
6 * Authors:
7 * Lennert Buytenhek <buytenh@gnu.org>
8 */
9
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>
17 #endif
18 #include <linux/neighbour.h>
19 #include <net/arp.h>
20 #include <linux/export.h>
21 #include <linux/rculist.h>
22 #include "br_private.h"
23 #include "br_private_tunnel.h"
24
25 static int
br_netif_receive_skb(struct net * net,struct sock * sk,struct sk_buff * skb)26 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
27 {
28 br_drop_fake_rtable(skb);
29 return netif_receive_skb(skb);
30 }
31
br_pass_frame_up(struct sk_buff * skb)32 static int br_pass_frame_up(struct sk_buff *skb)
33 {
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);
38
39 u64_stats_update_begin(&brstats->syncp);
40 brstats->rx_packets++;
41 brstats->rx_bytes += skb->len;
42 u64_stats_update_end(&brstats->syncp);
43
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.
48 */
49 if (!(brdev->flags & IFF_PROMISC) &&
50 !br_allowed_egress(vg, skb)) {
51 kfree_skb(skb);
52 return NET_RX_DROP;
53 }
54
55 indev = skb->dev;
56 skb->dev = brdev;
57 skb = br_handle_vlan(br, NULL, vg, skb);
58 if (!skb)
59 return NET_RX_DROP;
60 /* update the multicast stats if the packet is IGMP/MLD */
61 br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
62 BR_MCAST_DIR_TX);
63
64 return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
65 dev_net(indev), NULL, skb, indev, NULL,
66 br_netif_receive_skb);
67 }
68
69 /* note: already called with rcu_read_lock */
br_handle_frame_finish(struct net * net,struct sock * sk,struct sk_buff * skb)70 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
71 {
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;
78 u16 vid = 0;
79
80 if (!p || p->state == BR_STATE_DISABLED)
81 goto drop;
82
83 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
84 goto out;
85
86 nbp_switchdev_frame_mark(p, skb);
87
88 /* insert into forwarding database after filtering to avoid spoofing */
89 br = p->br;
90 if (p->flags & BR_LEARNING)
91 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
92
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;
98 local_rcv = true;
99 } else {
100 pkt_type = BR_PKT_MULTICAST;
101 if (br_multicast_rcv(br, p, skb, vid))
102 goto drop;
103 }
104 }
105
106 if (p->state == BR_STATE_LEARNING)
107 goto drop;
108
109 BR_INPUT_SKB_CB(skb)->brdev = br->dev;
110 BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
111
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;
123
124 msg = br_is_nd_neigh_msg(skb, &_msg);
125 if (msg)
126 br_do_suppress_nd(skb, br, vid, p, msg);
127 }
128
129 switch (pkt_type) {
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)) {
136 local_rcv = true;
137 br->dev->stats.multicast++;
138 }
139 mcast_hit = true;
140 } else {
141 local_rcv = true;
142 br->dev->stats.multicast++;
143 }
144 break;
145 case BR_PKT_UNICAST:
146 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
147 default:
148 break;
149 }
150
151 if (dst) {
152 unsigned long now = jiffies;
153
154 if (dst->is_local)
155 return br_pass_frame_up(skb);
156
157 if (now != dst->used)
158 dst->used = now;
159 br_forward(dst->dst, skb, local_rcv, false);
160 } else {
161 if (!mcast_hit)
162 br_flood(br, skb, pkt_type, local_rcv, false);
163 else
164 br_multicast_flood(mdst, skb, local_rcv, false);
165 }
166
167 if (local_rcv)
168 return br_pass_frame_up(skb);
169
170 out:
171 return 0;
172 drop:
173 kfree_skb(skb);
174 goto out;
175 }
176 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
177
__br_handle_local_finish(struct sk_buff * skb)178 static void __br_handle_local_finish(struct sk_buff *skb)
179 {
180 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
181 u16 vid = 0;
182
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);
188 }
189
190 /* note: already called with rcu_read_lock */
br_handle_local_finish(struct net * net,struct sock * sk,struct sk_buff * skb)191 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
192 {
193 __br_handle_local_finish(skb);
194
195 /* return 1 to signal the okfn() was called so it's ok to use the skb */
196 return 1;
197 }
198
nf_hook_bridge_pre(struct sk_buff * skb,struct sk_buff ** pskb)199 static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb)
200 {
201 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
202 struct nf_hook_entries *e = NULL;
203 struct nf_hook_state state;
204 unsigned int verdict, i;
205 struct net *net;
206 int ret;
207
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]))
211 goto frame_finish;
212 #endif
213
214 e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]);
215 if (!e)
216 goto frame_finish;
217
218 nf_hook_state_init(&state, NF_BR_PRE_ROUTING,
219 NFPROTO_BRIDGE, skb->dev, NULL, NULL,
220 net, br_handle_frame_finish);
221
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) {
225 case NF_ACCEPT:
226 if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) {
227 *pskb = skb;
228 return RX_HANDLER_PASS;
229 }
230 break;
231 case NF_DROP:
232 kfree_skb(skb);
233 return RX_HANDLER_CONSUMED;
234 case NF_QUEUE:
235 ret = nf_queue(skb, &state, i, verdict);
236 if (ret == 1)
237 continue;
238 return RX_HANDLER_CONSUMED;
239 default: /* STOLEN */
240 return RX_HANDLER_CONSUMED;
241 }
242 }
243 frame_finish:
244 net = dev_net(skb->dev);
245 br_handle_frame_finish(net, NULL, skb);
246 #else
247 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
248 #endif
249 return RX_HANDLER_CONSUMED;
250 }
251
252 /*
253 * Return NULL if skb is handled
254 * note: already called with rcu_read_lock
255 */
br_handle_frame(struct sk_buff ** pskb)256 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
257 {
258 struct net_bridge_port *p;
259 struct sk_buff *skb = *pskb;
260 const unsigned char *dest = eth_hdr(skb)->h_dest;
261
262 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
263 return RX_HANDLER_PASS;
264
265 if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
266 goto drop;
267
268 skb = skb_share_check(skb, GFP_ATOMIC);
269 if (!skb)
270 return RX_HANDLER_CONSUMED;
271
272 memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
273
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)))
278 goto drop;
279 }
280
281 if (unlikely(is_link_local_ether_addr(dest))) {
282 u16 fwd_mask = p->br->group_fwd_mask_required;
283
284 /*
285 * See IEEE 802.1D Table 7-10 Reserved addresses
286 *
287 * Assignment Value
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
292 *
293 * 802.1AB LLDP 01-80-C2-00-00-0E
294 *
295 * Others reserved for future standardization
296 */
297 fwd_mask |= p->group_fwd_mask;
298 switch (dest[5]) {
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]))
304 goto forward;
305 *pskb = skb;
306 __br_handle_local_finish(skb);
307 return RX_HANDLER_PASS;
308
309 case 0x01: /* IEEE MAC (Pause) */
310 goto drop;
311
312 case 0x0E: /* 802.1AB LLDP */
313 fwd_mask |= p->br->group_fwd_mask;
314 if (fwd_mask & (1u << dest[5]))
315 goto forward;
316 *pskb = skb;
317 __br_handle_local_finish(skb);
318 return RX_HANDLER_PASS;
319
320 default:
321 /* Allow selective forwarding for most other protocols */
322 fwd_mask |= p->br->group_fwd_mask;
323 if (fwd_mask & (1u << dest[5]))
324 goto forward;
325 }
326
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
331 */
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;
336 } else {
337 return RX_HANDLER_CONSUMED;
338 }
339 }
340
341 forward:
342 switch (p->state) {
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;
347
348 return nf_hook_bridge_pre(skb, pskb);
349 default:
350 drop:
351 kfree_skb(skb);
352 }
353 return RX_HANDLER_CONSUMED;
354 }
355