Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
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#ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
20#include <net/netfilter/nf_queue.h>
21#endif
22#include <linux/neighbour.h>
23#include <net/arp.h>
24#include <linux/export.h>
25#include <linux/rculist.h>
26#include "br_private.h"
27#include "br_private_tunnel.h"
28
29static int
30br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
31{
32 br_drop_fake_rtable(skb);
33 return netif_receive_skb(skb);
34}
35
36static int br_pass_frame_up(struct sk_buff *skb)
37{
38 struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
39 struct net_bridge *br = netdev_priv(brdev);
40 struct net_bridge_vlan_group *vg;
41 struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
42
43 u64_stats_update_begin(&brstats->syncp);
44 brstats->rx_packets++;
45 brstats->rx_bytes += skb->len;
46 u64_stats_update_end(&brstats->syncp);
47
48 vg = br_vlan_group_rcu(br);
49 /* Bridge is just like any other port. Make sure the
50 * packet is allowed except in promisc modue when someone
51 * may be running packet capture.
52 */
53 if (!(brdev->flags & IFF_PROMISC) &&
54 !br_allowed_egress(vg, skb)) {
55 kfree_skb(skb);
56 return NET_RX_DROP;
57 }
58
59 indev = skb->dev;
60 skb->dev = brdev;
61 skb = br_handle_vlan(br, NULL, vg, skb);
62 if (!skb)
63 return NET_RX_DROP;
64 /* update the multicast stats if the packet is IGMP/MLD */
65 br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
66 BR_MCAST_DIR_TX);
67
68 return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
69 dev_net(indev), NULL, skb, indev, NULL,
70 br_netif_receive_skb);
71}
72
73/* note: already called with rcu_read_lock */
74int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
75{
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_mdb_entry *mdst;
80 bool local_rcv, mcast_hit = false;
81 const unsigned char *dest;
82 struct net_bridge *br;
83 u16 vid = 0;
84
85 if (!p || p->state == BR_STATE_DISABLED)
86 goto drop;
87
88 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
89 goto out;
90
91 nbp_switchdev_frame_mark(p, skb);
92
93 /* insert into forwarding database after filtering to avoid spoofing */
94 br = p->br;
95 if (p->flags & BR_LEARNING)
96 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
97
98 local_rcv = !!(br->dev->flags & IFF_PROMISC);
99 dest = eth_hdr(skb)->h_dest;
100 if (is_multicast_ether_addr(dest)) {
101 /* by definition the broadcast is also a multicast address */
102 if (is_broadcast_ether_addr(dest)) {
103 pkt_type = BR_PKT_BROADCAST;
104 local_rcv = true;
105 } else {
106 pkt_type = BR_PKT_MULTICAST;
107 if (br_multicast_rcv(br, p, skb, vid))
108 goto drop;
109 }
110 }
111
112 if (p->state == BR_STATE_LEARNING)
113 goto drop;
114
115 BR_INPUT_SKB_CB(skb)->brdev = br->dev;
116 BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
117
118 if (IS_ENABLED(CONFIG_INET) &&
119 (skb->protocol == htons(ETH_P_ARP) ||
120 skb->protocol == htons(ETH_P_RARP))) {
121 br_do_proxy_suppress_arp(skb, br, vid, p);
122 } else if (IS_ENABLED(CONFIG_IPV6) &&
123 skb->protocol == htons(ETH_P_IPV6) &&
124 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
125 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
126 sizeof(struct nd_msg)) &&
127 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
128 struct nd_msg *msg, _msg;
129
130 msg = br_is_nd_neigh_msg(skb, &_msg);
131 if (msg)
132 br_do_suppress_nd(skb, br, vid, p, msg);
133 }
134
135 switch (pkt_type) {
136 case BR_PKT_MULTICAST:
137 mdst = br_mdb_get(br, skb, vid);
138 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
139 br_multicast_querier_exists(br, eth_hdr(skb))) {
140 if ((mdst && mdst->host_joined) ||
141 br_multicast_is_router(br)) {
142 local_rcv = true;
143 br->dev->stats.multicast++;
144 }
145 mcast_hit = true;
146 } else {
147 local_rcv = true;
148 br->dev->stats.multicast++;
149 }
150 break;
151 case BR_PKT_UNICAST:
152 dst = br_fdb_find_rcu(br, dest, vid);
153 default:
154 break;
155 }
156
157 if (dst) {
158 unsigned long now = jiffies;
159
160 if (dst->is_local)
161 return br_pass_frame_up(skb);
162
163 if (now != dst->used)
164 dst->used = now;
165 br_forward(dst->dst, skb, local_rcv, false);
166 } else {
167 if (!mcast_hit)
168 br_flood(br, skb, pkt_type, local_rcv, false);
169 else
170 br_multicast_flood(mdst, skb, local_rcv, false);
171 }
172
173 if (local_rcv)
174 return br_pass_frame_up(skb);
175
176out:
177 return 0;
178drop:
179 kfree_skb(skb);
180 goto out;
181}
182EXPORT_SYMBOL_GPL(br_handle_frame_finish);
183
184static void __br_handle_local_finish(struct sk_buff *skb)
185{
186 struct net_bridge_port *p = br_port_get_rcu(skb->dev);
187 u16 vid = 0;
188
189 /* check if vlan is allowed, to avoid spoofing */
190 if ((p->flags & BR_LEARNING) &&
191 !br_opt_get(p->br, BROPT_NO_LL_LEARN) &&
192 br_should_learn(p, skb, &vid))
193 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
194}
195
196/* note: already called with rcu_read_lock */
197static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
198{
199 __br_handle_local_finish(skb);
200
201 /* return 1 to signal the okfn() was called so it's ok to use the skb */
202 return 1;
203}
204
205static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb)
206{
207#ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
208 struct nf_hook_entries *e = NULL;
209 struct nf_hook_state state;
210 unsigned int verdict, i;
211 struct net *net;
212 int ret;
213
214 net = dev_net(skb->dev);
215#ifdef HAVE_JUMP_LABEL
216 if (!static_key_false(&nf_hooks_needed[NFPROTO_BRIDGE][NF_BR_PRE_ROUTING]))
217 goto frame_finish;
218#endif
219
220 e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]);
221 if (!e)
222 goto frame_finish;
223
224 nf_hook_state_init(&state, NF_BR_PRE_ROUTING,
225 NFPROTO_BRIDGE, skb->dev, NULL, NULL,
226 net, br_handle_frame_finish);
227
228 for (i = 0; i < e->num_hook_entries; i++) {
229 verdict = nf_hook_entry_hookfn(&e->hooks[i], skb, &state);
230 switch (verdict & NF_VERDICT_MASK) {
231 case NF_ACCEPT:
232 if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) {
233 *pskb = skb;
234 return RX_HANDLER_PASS;
235 }
236 break;
237 case NF_DROP:
238 kfree_skb(skb);
239 return RX_HANDLER_CONSUMED;
240 case NF_QUEUE:
241 ret = nf_queue(skb, &state, e, i, verdict);
242 if (ret == 1)
243 continue;
244 return RX_HANDLER_CONSUMED;
245 default: /* STOLEN */
246 return RX_HANDLER_CONSUMED;
247 }
248 }
249frame_finish:
250 net = dev_net(skb->dev);
251 br_handle_frame_finish(net, NULL, skb);
252#else
253 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
254#endif
255 return RX_HANDLER_CONSUMED;
256}
257
258/*
259 * Return NULL if skb is handled
260 * note: already called with rcu_read_lock
261 */
262rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
263{
264 struct net_bridge_port *p;
265 struct sk_buff *skb = *pskb;
266 const unsigned char *dest = eth_hdr(skb)->h_dest;
267
268 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
269 return RX_HANDLER_PASS;
270
271 if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
272 goto drop;
273
274 skb = skb_share_check(skb, GFP_ATOMIC);
275 if (!skb)
276 return RX_HANDLER_CONSUMED;
277
278 memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
279
280 p = br_port_get_rcu(skb->dev);
281 if (p->flags & BR_VLAN_TUNNEL) {
282 if (br_handle_ingress_vlan_tunnel(skb, p,
283 nbp_vlan_group_rcu(p)))
284 goto drop;
285 }
286
287 if (unlikely(is_link_local_ether_addr(dest))) {
288 u16 fwd_mask = p->br->group_fwd_mask_required;
289
290 /*
291 * See IEEE 802.1D Table 7-10 Reserved addresses
292 *
293 * Assignment Value
294 * Bridge Group Address 01-80-C2-00-00-00
295 * (MAC Control) 802.3 01-80-C2-00-00-01
296 * (Link Aggregation) 802.3 01-80-C2-00-00-02
297 * 802.1X PAE address 01-80-C2-00-00-03
298 *
299 * 802.1AB LLDP 01-80-C2-00-00-0E
300 *
301 * Others reserved for future standardization
302 */
303 fwd_mask |= p->group_fwd_mask;
304 switch (dest[5]) {
305 case 0x00: /* Bridge Group Address */
306 /* If STP is turned off,
307 then must forward to keep loop detection */
308 if (p->br->stp_enabled == BR_NO_STP ||
309 fwd_mask & (1u << dest[5]))
310 goto forward;
311 *pskb = skb;
312 __br_handle_local_finish(skb);
313 return RX_HANDLER_PASS;
314
315 case 0x01: /* IEEE MAC (Pause) */
316 goto drop;
317
318 case 0x0E: /* 802.1AB LLDP */
319 fwd_mask |= p->br->group_fwd_mask;
320 if (fwd_mask & (1u << dest[5]))
321 goto forward;
322 *pskb = skb;
323 __br_handle_local_finish(skb);
324 return RX_HANDLER_PASS;
325
326 default:
327 /* Allow selective forwarding for most other protocols */
328 fwd_mask |= p->br->group_fwd_mask;
329 if (fwd_mask & (1u << dest[5]))
330 goto forward;
331 }
332
333 /* The else clause should be hit when nf_hook():
334 * - returns < 0 (drop/error)
335 * - returns = 0 (stolen/nf_queue)
336 * Thus return 1 from the okfn() to signal the skb is ok to pass
337 */
338 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
339 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
340 br_handle_local_finish) == 1) {
341 return RX_HANDLER_PASS;
342 } else {
343 return RX_HANDLER_CONSUMED;
344 }
345 }
346
347forward:
348 switch (p->state) {
349 case BR_STATE_FORWARDING:
350 case BR_STATE_LEARNING:
351 if (ether_addr_equal(p->br->dev->dev_addr, dest))
352 skb->pkt_type = PACKET_HOST;
353
354 return nf_hook_bridge_pre(skb, pskb);
355 default:
356drop:
357 kfree_skb(skb);
358 }
359 return RX_HANDLER_CONSUMED;
360}