Loading...
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
25static int
26br_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
32static 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 */
70int 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
170out:
171 return 0;
172drop:
173 kfree_skb(skb);
174 goto out;
175}
176EXPORT_SYMBOL_GPL(br_handle_frame_finish);
177
178static 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 */
191static 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
199static 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 }
243frame_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 */
256rx_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
341forward:
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:
350drop:
351 kfree_skb(skb);
352 }
353 return RX_HANDLER_CONSUMED;
354}
1/*
2 * Handle incoming frames
3 * Linux ethernet bridge
4 *
5 * Authors:
6 * Lennert Buytenhek <buytenh@gnu.org>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#include <linux/slab.h>
15#include <linux/kernel.h>
16#include <linux/netdevice.h>
17#include <linux/etherdevice.h>
18#include <linux/netfilter_bridge.h>
19#include <linux/neighbour.h>
20#include <net/arp.h>
21#include <linux/export.h>
22#include <linux/rculist.h>
23#include "br_private.h"
24
25/* Hook for brouter */
26br_should_route_hook_t __rcu *br_should_route_hook __read_mostly;
27EXPORT_SYMBOL(br_should_route_hook);
28
29static int
30br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
31{
32 return netif_receive_skb(skb);
33}
34
35static int br_pass_frame_up(struct sk_buff *skb)
36{
37 struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
38 struct net_bridge *br = netdev_priv(brdev);
39 struct net_bridge_vlan_group *vg;
40 struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
41
42 u64_stats_update_begin(&brstats->syncp);
43 brstats->rx_packets++;
44 brstats->rx_bytes += skb->len;
45 u64_stats_update_end(&brstats->syncp);
46
47 vg = br_vlan_group_rcu(br);
48 /* Bridge is just like any other port. Make sure the
49 * packet is allowed except in promisc modue when someone
50 * may be running packet capture.
51 */
52 if (!(brdev->flags & IFF_PROMISC) &&
53 !br_allowed_egress(vg, skb)) {
54 kfree_skb(skb);
55 return NET_RX_DROP;
56 }
57
58 indev = skb->dev;
59 skb->dev = brdev;
60 skb = br_handle_vlan(br, vg, skb);
61 if (!skb)
62 return NET_RX_DROP;
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
69static void br_do_proxy_arp(struct sk_buff *skb, struct net_bridge *br,
70 u16 vid, struct net_bridge_port *p)
71{
72 struct net_device *dev = br->dev;
73 struct neighbour *n;
74 struct arphdr *parp;
75 u8 *arpptr, *sha;
76 __be32 sip, tip;
77
78 BR_INPUT_SKB_CB(skb)->proxyarp_replied = false;
79
80 if (dev->flags & IFF_NOARP)
81 return;
82
83 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
84 dev->stats.tx_dropped++;
85 return;
86 }
87 parp = arp_hdr(skb);
88
89 if (parp->ar_pro != htons(ETH_P_IP) ||
90 parp->ar_op != htons(ARPOP_REQUEST) ||
91 parp->ar_hln != dev->addr_len ||
92 parp->ar_pln != 4)
93 return;
94
95 arpptr = (u8 *)parp + sizeof(struct arphdr);
96 sha = arpptr;
97 arpptr += dev->addr_len; /* sha */
98 memcpy(&sip, arpptr, sizeof(sip));
99 arpptr += sizeof(sip);
100 arpptr += dev->addr_len; /* tha */
101 memcpy(&tip, arpptr, sizeof(tip));
102
103 if (ipv4_is_loopback(tip) ||
104 ipv4_is_multicast(tip))
105 return;
106
107 n = neigh_lookup(&arp_tbl, &tip, dev);
108 if (n) {
109 struct net_bridge_fdb_entry *f;
110
111 if (!(n->nud_state & NUD_VALID)) {
112 neigh_release(n);
113 return;
114 }
115
116 f = __br_fdb_get(br, n->ha, vid);
117 if (f && ((p->flags & BR_PROXYARP) ||
118 (f->dst && (f->dst->flags & BR_PROXYARP_WIFI)))) {
119 arp_send(ARPOP_REPLY, ETH_P_ARP, sip, skb->dev, tip,
120 sha, n->ha, sha);
121 BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
122 }
123
124 neigh_release(n);
125 }
126}
127
128/* note: already called with rcu_read_lock */
129int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
130{
131 const unsigned char *dest = eth_hdr(skb)->h_dest;
132 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
133 struct net_bridge *br;
134 struct net_bridge_fdb_entry *dst;
135 struct net_bridge_mdb_entry *mdst;
136 struct sk_buff *skb2;
137 bool unicast = true;
138 u16 vid = 0;
139
140 if (!p || p->state == BR_STATE_DISABLED)
141 goto drop;
142
143 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
144 goto out;
145
146 /* insert into forwarding database after filtering to avoid spoofing */
147 br = p->br;
148 if (p->flags & BR_LEARNING)
149 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
150
151 if (!is_broadcast_ether_addr(dest) && is_multicast_ether_addr(dest) &&
152 br_multicast_rcv(br, p, skb, vid))
153 goto drop;
154
155 if (p->state == BR_STATE_LEARNING)
156 goto drop;
157
158 BR_INPUT_SKB_CB(skb)->brdev = br->dev;
159
160 /* The packet skb2 goes to the local host (NULL to skip). */
161 skb2 = NULL;
162
163 if (br->dev->flags & IFF_PROMISC)
164 skb2 = skb;
165
166 dst = NULL;
167
168 if (IS_ENABLED(CONFIG_INET) && skb->protocol == htons(ETH_P_ARP))
169 br_do_proxy_arp(skb, br, vid, p);
170
171 if (is_broadcast_ether_addr(dest)) {
172 skb2 = skb;
173 unicast = false;
174 } else if (is_multicast_ether_addr(dest)) {
175 mdst = br_mdb_get(br, skb, vid);
176 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
177 br_multicast_querier_exists(br, eth_hdr(skb))) {
178 if ((mdst && mdst->mglist) ||
179 br_multicast_is_router(br))
180 skb2 = skb;
181 br_multicast_forward(mdst, skb, skb2);
182 skb = NULL;
183 if (!skb2)
184 goto out;
185 } else
186 skb2 = skb;
187
188 unicast = false;
189 br->dev->stats.multicast++;
190 } else if ((dst = __br_fdb_get(br, dest, vid)) &&
191 dst->is_local) {
192 skb2 = skb;
193 /* Do not forward the packet since it's local. */
194 skb = NULL;
195 }
196
197 if (skb) {
198 if (dst) {
199 dst->used = jiffies;
200 br_forward(dst->dst, skb, skb2);
201 } else
202 br_flood_forward(br, skb, skb2, unicast);
203 }
204
205 if (skb2)
206 return br_pass_frame_up(skb2);
207
208out:
209 return 0;
210drop:
211 kfree_skb(skb);
212 goto out;
213}
214EXPORT_SYMBOL_GPL(br_handle_frame_finish);
215
216/* note: already called with rcu_read_lock */
217static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
218{
219 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
220 u16 vid = 0;
221
222 /* check if vlan is allowed, to avoid spoofing */
223 if (p->flags & BR_LEARNING && br_should_learn(p, skb, &vid))
224 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
225
226 BR_INPUT_SKB_CB(skb)->brdev = p->br->dev;
227 br_pass_frame_up(skb);
228 return 0;
229}
230
231/*
232 * Return NULL if skb is handled
233 * note: already called with rcu_read_lock
234 */
235rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
236{
237 struct net_bridge_port *p;
238 struct sk_buff *skb = *pskb;
239 const unsigned char *dest = eth_hdr(skb)->h_dest;
240 br_should_route_hook_t *rhook;
241
242 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
243 return RX_HANDLER_PASS;
244
245 if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
246 goto drop;
247
248 skb = skb_share_check(skb, GFP_ATOMIC);
249 if (!skb)
250 return RX_HANDLER_CONSUMED;
251
252 p = br_port_get_rcu(skb->dev);
253
254 if (unlikely(is_link_local_ether_addr(dest))) {
255 u16 fwd_mask = p->br->group_fwd_mask_required;
256
257 /*
258 * See IEEE 802.1D Table 7-10 Reserved addresses
259 *
260 * Assignment Value
261 * Bridge Group Address 01-80-C2-00-00-00
262 * (MAC Control) 802.3 01-80-C2-00-00-01
263 * (Link Aggregation) 802.3 01-80-C2-00-00-02
264 * 802.1X PAE address 01-80-C2-00-00-03
265 *
266 * 802.1AB LLDP 01-80-C2-00-00-0E
267 *
268 * Others reserved for future standardization
269 */
270 switch (dest[5]) {
271 case 0x00: /* Bridge Group Address */
272 /* If STP is turned off,
273 then must forward to keep loop detection */
274 if (p->br->stp_enabled == BR_NO_STP ||
275 fwd_mask & (1u << dest[5]))
276 goto forward;
277 break;
278
279 case 0x01: /* IEEE MAC (Pause) */
280 goto drop;
281
282 default:
283 /* Allow selective forwarding for most other protocols */
284 fwd_mask |= p->br->group_fwd_mask;
285 if (fwd_mask & (1u << dest[5]))
286 goto forward;
287 }
288
289 /* Deliver packet to local host only */
290 NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN, dev_net(skb->dev),
291 NULL, skb, skb->dev, NULL, br_handle_local_finish);
292 return RX_HANDLER_CONSUMED;
293 }
294
295forward:
296 switch (p->state) {
297 case BR_STATE_FORWARDING:
298 rhook = rcu_dereference(br_should_route_hook);
299 if (rhook) {
300 if ((*rhook)(skb)) {
301 *pskb = skb;
302 return RX_HANDLER_PASS;
303 }
304 dest = eth_hdr(skb)->h_dest;
305 }
306 /* fall through */
307 case BR_STATE_LEARNING:
308 if (ether_addr_equal(p->br->dev->dev_addr, dest))
309 skb->pkt_type = PACKET_HOST;
310
311 NF_HOOK(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
312 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
313 br_handle_frame_finish);
314 break;
315 default:
316drop:
317 kfree_skb(skb);
318 }
319 return RX_HANDLER_CONSUMED;
320}