xref: /linux/net/bridge/br_input.c (revision e5c86679d5e864947a52fb31e45a425dea3e7fa9)
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 #include "br_private_tunnel.h"
25 
26 /* Hook for brouter */
27 br_should_route_hook_t __rcu *br_should_route_hook __read_mostly;
28 EXPORT_SYMBOL(br_should_route_hook);
29 
30 static int
31 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
32 {
33 	br_drop_fake_rtable(skb);
34 	return netif_receive_skb(skb);
35 }
36 
37 static int br_pass_frame_up(struct sk_buff *skb)
38 {
39 	struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
40 	struct net_bridge *br = netdev_priv(brdev);
41 	struct net_bridge_vlan_group *vg;
42 	struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
43 
44 	u64_stats_update_begin(&brstats->syncp);
45 	brstats->rx_packets++;
46 	brstats->rx_bytes += skb->len;
47 	u64_stats_update_end(&brstats->syncp);
48 
49 	vg = br_vlan_group_rcu(br);
50 	/* Bridge is just like any other port.  Make sure the
51 	 * packet is allowed except in promisc modue when someone
52 	 * may be running packet capture.
53 	 */
54 	if (!(brdev->flags & IFF_PROMISC) &&
55 	    !br_allowed_egress(vg, skb)) {
56 		kfree_skb(skb);
57 		return NET_RX_DROP;
58 	}
59 
60 	indev = skb->dev;
61 	skb->dev = brdev;
62 	skb = br_handle_vlan(br, NULL, vg, skb);
63 	if (!skb)
64 		return NET_RX_DROP;
65 	/* update the multicast stats if the packet is IGMP/MLD */
66 	br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
67 			   BR_MCAST_DIR_TX);
68 
69 	return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
70 		       dev_net(indev), NULL, skb, indev, NULL,
71 		       br_netif_receive_skb);
72 }
73 
74 static void br_do_proxy_arp(struct sk_buff *skb, struct net_bridge *br,
75 			    u16 vid, struct net_bridge_port *p)
76 {
77 	struct net_device *dev = br->dev;
78 	struct neighbour *n;
79 	struct arphdr *parp;
80 	u8 *arpptr, *sha;
81 	__be32 sip, tip;
82 
83 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = false;
84 
85 	if ((dev->flags & IFF_NOARP) ||
86 	    !pskb_may_pull(skb, arp_hdr_len(dev)))
87 		return;
88 
89 	parp = arp_hdr(skb);
90 
91 	if (parp->ar_pro != htons(ETH_P_IP) ||
92 	    parp->ar_op != htons(ARPOP_REQUEST) ||
93 	    parp->ar_hln != dev->addr_len ||
94 	    parp->ar_pln != 4)
95 		return;
96 
97 	arpptr = (u8 *)parp + sizeof(struct arphdr);
98 	sha = arpptr;
99 	arpptr += dev->addr_len;	/* sha */
100 	memcpy(&sip, arpptr, sizeof(sip));
101 	arpptr += sizeof(sip);
102 	arpptr += dev->addr_len;	/* tha */
103 	memcpy(&tip, arpptr, sizeof(tip));
104 
105 	if (ipv4_is_loopback(tip) ||
106 	    ipv4_is_multicast(tip))
107 		return;
108 
109 	n = neigh_lookup(&arp_tbl, &tip, dev);
110 	if (n) {
111 		struct net_bridge_fdb_entry *f;
112 
113 		if (!(n->nud_state & NUD_VALID)) {
114 			neigh_release(n);
115 			return;
116 		}
117 
118 		f = br_fdb_find_rcu(br, n->ha, vid);
119 		if (f && ((p->flags & BR_PROXYARP) ||
120 			  (f->dst && (f->dst->flags & BR_PROXYARP_WIFI)))) {
121 			arp_send(ARPOP_REPLY, ETH_P_ARP, sip, skb->dev, tip,
122 				 sha, n->ha, sha);
123 			BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
124 		}
125 
126 		neigh_release(n);
127 	}
128 }
129 
130 /* note: already called with rcu_read_lock */
131 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
132 {
133 	struct net_bridge_port *p = br_port_get_rcu(skb->dev);
134 	const unsigned char *dest = eth_hdr(skb)->h_dest;
135 	enum br_pkt_type pkt_type = BR_PKT_UNICAST;
136 	struct net_bridge_fdb_entry *dst = NULL;
137 	struct net_bridge_mdb_entry *mdst;
138 	bool local_rcv, mcast_hit = false;
139 	struct net_bridge *br;
140 	u16 vid = 0;
141 
142 	if (!p || p->state == BR_STATE_DISABLED)
143 		goto drop;
144 
145 	if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
146 		goto out;
147 
148 	nbp_switchdev_frame_mark(p, skb);
149 
150 	/* insert into forwarding database after filtering to avoid spoofing */
151 	br = p->br;
152 	if (p->flags & BR_LEARNING)
153 		br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
154 
155 	local_rcv = !!(br->dev->flags & IFF_PROMISC);
156 	if (is_multicast_ether_addr(dest)) {
157 		/* by definition the broadcast is also a multicast address */
158 		if (is_broadcast_ether_addr(dest)) {
159 			pkt_type = BR_PKT_BROADCAST;
160 			local_rcv = true;
161 		} else {
162 			pkt_type = BR_PKT_MULTICAST;
163 			if (br_multicast_rcv(br, p, skb, vid))
164 				goto drop;
165 		}
166 	}
167 
168 	if (p->state == BR_STATE_LEARNING)
169 		goto drop;
170 
171 	BR_INPUT_SKB_CB(skb)->brdev = br->dev;
172 
173 	if (IS_ENABLED(CONFIG_INET) && skb->protocol == htons(ETH_P_ARP))
174 		br_do_proxy_arp(skb, br, vid, p);
175 
176 	switch (pkt_type) {
177 	case BR_PKT_MULTICAST:
178 		mdst = br_mdb_get(br, skb, vid);
179 		if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
180 		    br_multicast_querier_exists(br, eth_hdr(skb))) {
181 			if ((mdst && mdst->mglist) ||
182 			    br_multicast_is_router(br)) {
183 				local_rcv = true;
184 				br->dev->stats.multicast++;
185 			}
186 			mcast_hit = true;
187 		} else {
188 			local_rcv = true;
189 			br->dev->stats.multicast++;
190 		}
191 		break;
192 	case BR_PKT_UNICAST:
193 		dst = br_fdb_find_rcu(br, dest, vid);
194 	default:
195 		break;
196 	}
197 
198 	if (dst) {
199 		unsigned long now = jiffies;
200 
201 		if (dst->is_local)
202 			return br_pass_frame_up(skb);
203 
204 		if (now != dst->used)
205 			dst->used = now;
206 		br_forward(dst->dst, skb, local_rcv, false);
207 	} else {
208 		if (!mcast_hit)
209 			br_flood(br, skb, pkt_type, local_rcv, false);
210 		else
211 			br_multicast_flood(mdst, skb, local_rcv, false);
212 	}
213 
214 	if (local_rcv)
215 		return br_pass_frame_up(skb);
216 
217 out:
218 	return 0;
219 drop:
220 	kfree_skb(skb);
221 	goto out;
222 }
223 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
224 
225 static void __br_handle_local_finish(struct sk_buff *skb)
226 {
227 	struct net_bridge_port *p = br_port_get_rcu(skb->dev);
228 	u16 vid = 0;
229 
230 	/* check if vlan is allowed, to avoid spoofing */
231 	if (p->flags & BR_LEARNING && br_should_learn(p, skb, &vid))
232 		br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
233 }
234 
235 /* note: already called with rcu_read_lock */
236 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
237 {
238 	struct net_bridge_port *p = br_port_get_rcu(skb->dev);
239 
240 	__br_handle_local_finish(skb);
241 
242 	BR_INPUT_SKB_CB(skb)->brdev = p->br->dev;
243 	br_pass_frame_up(skb);
244 	return 0;
245 }
246 
247 /*
248  * Return NULL if skb is handled
249  * note: already called with rcu_read_lock
250  */
251 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
252 {
253 	struct net_bridge_port *p;
254 	struct sk_buff *skb = *pskb;
255 	const unsigned char *dest = eth_hdr(skb)->h_dest;
256 	br_should_route_hook_t *rhook;
257 
258 	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
259 		return RX_HANDLER_PASS;
260 
261 	if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
262 		goto drop;
263 
264 	skb = skb_share_check(skb, GFP_ATOMIC);
265 	if (!skb)
266 		return RX_HANDLER_CONSUMED;
267 
268 	p = br_port_get_rcu(skb->dev);
269 	if (p->flags & BR_VLAN_TUNNEL) {
270 		if (br_handle_ingress_vlan_tunnel(skb, p,
271 						  nbp_vlan_group_rcu(p)))
272 			goto drop;
273 	}
274 
275 	if (unlikely(is_link_local_ether_addr(dest))) {
276 		u16 fwd_mask = p->br->group_fwd_mask_required;
277 
278 		/*
279 		 * See IEEE 802.1D Table 7-10 Reserved addresses
280 		 *
281 		 * Assignment		 		Value
282 		 * Bridge Group Address		01-80-C2-00-00-00
283 		 * (MAC Control) 802.3		01-80-C2-00-00-01
284 		 * (Link Aggregation) 802.3	01-80-C2-00-00-02
285 		 * 802.1X PAE address		01-80-C2-00-00-03
286 		 *
287 		 * 802.1AB LLDP 		01-80-C2-00-00-0E
288 		 *
289 		 * Others reserved for future standardization
290 		 */
291 		switch (dest[5]) {
292 		case 0x00:	/* Bridge Group Address */
293 			/* If STP is turned off,
294 			   then must forward to keep loop detection */
295 			if (p->br->stp_enabled == BR_NO_STP ||
296 			    fwd_mask & (1u << dest[5]))
297 				goto forward;
298 			*pskb = skb;
299 			__br_handle_local_finish(skb);
300 			return RX_HANDLER_PASS;
301 
302 		case 0x01:	/* IEEE MAC (Pause) */
303 			goto drop;
304 
305 		case 0x0E:	/* 802.1AB LLDP */
306 			fwd_mask |= p->br->group_fwd_mask;
307 			if (fwd_mask & (1u << dest[5]))
308 				goto forward;
309 			*pskb = skb;
310 			__br_handle_local_finish(skb);
311 			return RX_HANDLER_PASS;
312 
313 		default:
314 			/* Allow selective forwarding for most other protocols */
315 			fwd_mask |= p->br->group_fwd_mask;
316 			if (fwd_mask & (1u << dest[5]))
317 				goto forward;
318 		}
319 
320 		/* Deliver packet to local host only */
321 		NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN, dev_net(skb->dev),
322 			NULL, skb, skb->dev, NULL, br_handle_local_finish);
323 		return RX_HANDLER_CONSUMED;
324 	}
325 
326 forward:
327 	switch (p->state) {
328 	case BR_STATE_FORWARDING:
329 		rhook = rcu_dereference(br_should_route_hook);
330 		if (rhook) {
331 			if ((*rhook)(skb)) {
332 				*pskb = skb;
333 				return RX_HANDLER_PASS;
334 			}
335 			dest = eth_hdr(skb)->h_dest;
336 		}
337 		/* fall through */
338 	case BR_STATE_LEARNING:
339 		if (ether_addr_equal(p->br->dev->dev_addr, dest))
340 			skb->pkt_type = PACKET_HOST;
341 
342 		NF_HOOK(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
343 			dev_net(skb->dev), NULL, skb, skb->dev, NULL,
344 			br_handle_frame_finish);
345 		break;
346 	default:
347 drop:
348 		kfree_skb(skb);
349 	}
350 	return RX_HANDLER_CONSUMED;
351 }
352