xref: /openbmc/linux/net/bridge/br_device.c (revision 2eb3ed33e55d003d721d4d1a5e72fe323c12b4c0)
1 /*
2  *	Device handling code
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/kernel.h>
15 #include <linux/netdevice.h>
16 #include <linux/netpoll.h>
17 #include <linux/etherdevice.h>
18 #include <linux/ethtool.h>
19 #include <linux/list.h>
20 #include <linux/netfilter_bridge.h>
21 
22 #include <linux/uaccess.h>
23 #include "br_private.h"
24 
25 #define COMMON_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA | \
26 			 NETIF_F_GSO_MASK | NETIF_F_HW_CSUM)
27 
28 const struct nf_br_ops __rcu *nf_br_ops __read_mostly;
29 EXPORT_SYMBOL_GPL(nf_br_ops);
30 
31 static struct lock_class_key bridge_netdev_addr_lock_key;
32 
33 /* net device transmit always called with BH disabled */
34 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev)
35 {
36 	struct net_bridge *br = netdev_priv(dev);
37 	struct net_bridge_fdb_entry *dst;
38 	struct net_bridge_mdb_entry *mdst;
39 	struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
40 	const struct nf_br_ops *nf_ops;
41 	const unsigned char *dest;
42 	struct ethhdr *eth;
43 	u16 vid = 0;
44 
45 	rcu_read_lock();
46 	nf_ops = rcu_dereference(nf_br_ops);
47 	if (nf_ops && nf_ops->br_dev_xmit_hook(skb)) {
48 		rcu_read_unlock();
49 		return NETDEV_TX_OK;
50 	}
51 
52 	u64_stats_update_begin(&brstats->syncp);
53 	brstats->tx_packets++;
54 	brstats->tx_bytes += skb->len;
55 	u64_stats_update_end(&brstats->syncp);
56 
57 	br_switchdev_frame_unmark(skb);
58 	BR_INPUT_SKB_CB(skb)->brdev = dev;
59 
60 	skb_reset_mac_header(skb);
61 	eth = eth_hdr(skb);
62 	skb_pull(skb, ETH_HLEN);
63 
64 	if (!br_allowed_ingress(br, br_vlan_group_rcu(br), skb, &vid))
65 		goto out;
66 
67 	if (IS_ENABLED(CONFIG_INET) &&
68 	    (eth->h_proto == htons(ETH_P_ARP) ||
69 	     eth->h_proto == htons(ETH_P_RARP)) &&
70 	    br->neigh_suppress_enabled) {
71 		br_do_proxy_suppress_arp(skb, br, vid, NULL);
72 	} else if (IS_ENABLED(CONFIG_IPV6) &&
73 		   skb->protocol == htons(ETH_P_IPV6) &&
74 		   br->neigh_suppress_enabled &&
75 		   pskb_may_pull(skb, sizeof(struct ipv6hdr) +
76 				 sizeof(struct nd_msg)) &&
77 		   ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
78 			struct nd_msg *msg, _msg;
79 
80 			msg = br_is_nd_neigh_msg(skb, &_msg);
81 			if (msg)
82 				br_do_suppress_nd(skb, br, vid, NULL, msg);
83 	}
84 
85 	dest = eth_hdr(skb)->h_dest;
86 	if (is_broadcast_ether_addr(dest)) {
87 		br_flood(br, skb, BR_PKT_BROADCAST, false, true);
88 	} else if (is_multicast_ether_addr(dest)) {
89 		if (unlikely(netpoll_tx_running(dev))) {
90 			br_flood(br, skb, BR_PKT_MULTICAST, false, true);
91 			goto out;
92 		}
93 		if (br_multicast_rcv(br, NULL, skb, vid)) {
94 			kfree_skb(skb);
95 			goto out;
96 		}
97 
98 		mdst = br_mdb_get(br, skb, vid);
99 		if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
100 		    br_multicast_querier_exists(br, eth_hdr(skb)))
101 			br_multicast_flood(mdst, skb, false, true);
102 		else
103 			br_flood(br, skb, BR_PKT_MULTICAST, false, true);
104 	} else if ((dst = br_fdb_find_rcu(br, dest, vid)) != NULL) {
105 		br_forward(dst->dst, skb, false, true);
106 	} else {
107 		br_flood(br, skb, BR_PKT_UNICAST, false, true);
108 	}
109 out:
110 	rcu_read_unlock();
111 	return NETDEV_TX_OK;
112 }
113 
114 static void br_set_lockdep_class(struct net_device *dev)
115 {
116 	lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
117 }
118 
119 static int br_dev_init(struct net_device *dev)
120 {
121 	struct net_bridge *br = netdev_priv(dev);
122 	int err;
123 
124 	br->stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
125 	if (!br->stats)
126 		return -ENOMEM;
127 
128 	err = br_vlan_init(br);
129 	if (err) {
130 		free_percpu(br->stats);
131 		return err;
132 	}
133 
134 	err = br_multicast_init_stats(br);
135 	if (err) {
136 		free_percpu(br->stats);
137 		br_vlan_flush(br);
138 	}
139 	br_set_lockdep_class(dev);
140 
141 	return err;
142 }
143 
144 static void br_dev_uninit(struct net_device *dev)
145 {
146 	struct net_bridge *br = netdev_priv(dev);
147 
148 	br_multicast_dev_del(br);
149 	br_multicast_uninit_stats(br);
150 	br_vlan_flush(br);
151 	free_percpu(br->stats);
152 }
153 
154 static int br_dev_open(struct net_device *dev)
155 {
156 	struct net_bridge *br = netdev_priv(dev);
157 
158 	netdev_update_features(dev);
159 	netif_start_queue(dev);
160 	br_stp_enable_bridge(br);
161 	br_multicast_open(br);
162 
163 	return 0;
164 }
165 
166 static void br_dev_set_multicast_list(struct net_device *dev)
167 {
168 }
169 
170 static void br_dev_change_rx_flags(struct net_device *dev, int change)
171 {
172 	if (change & IFF_PROMISC)
173 		br_manage_promisc(netdev_priv(dev));
174 }
175 
176 static int br_dev_stop(struct net_device *dev)
177 {
178 	struct net_bridge *br = netdev_priv(dev);
179 
180 	br_stp_disable_bridge(br);
181 	br_multicast_stop(br);
182 
183 	netif_stop_queue(dev);
184 
185 	return 0;
186 }
187 
188 static void br_get_stats64(struct net_device *dev,
189 			   struct rtnl_link_stats64 *stats)
190 {
191 	struct net_bridge *br = netdev_priv(dev);
192 	struct pcpu_sw_netstats tmp, sum = { 0 };
193 	unsigned int cpu;
194 
195 	for_each_possible_cpu(cpu) {
196 		unsigned int start;
197 		const struct pcpu_sw_netstats *bstats
198 			= per_cpu_ptr(br->stats, cpu);
199 		do {
200 			start = u64_stats_fetch_begin_irq(&bstats->syncp);
201 			memcpy(&tmp, bstats, sizeof(tmp));
202 		} while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
203 		sum.tx_bytes   += tmp.tx_bytes;
204 		sum.tx_packets += tmp.tx_packets;
205 		sum.rx_bytes   += tmp.rx_bytes;
206 		sum.rx_packets += tmp.rx_packets;
207 	}
208 
209 	stats->tx_bytes   = sum.tx_bytes;
210 	stats->tx_packets = sum.tx_packets;
211 	stats->rx_bytes   = sum.rx_bytes;
212 	stats->rx_packets = sum.rx_packets;
213 }
214 
215 static int br_change_mtu(struct net_device *dev, int new_mtu)
216 {
217 	struct net_bridge *br = netdev_priv(dev);
218 	if (new_mtu > br_min_mtu(br))
219 		return -EINVAL;
220 
221 	dev->mtu = new_mtu;
222 
223 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
224 	/* remember the MTU in the rtable for PMTU */
225 	dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
226 #endif
227 
228 	return 0;
229 }
230 
231 /* Allow setting mac address to any valid ethernet address. */
232 static int br_set_mac_address(struct net_device *dev, void *p)
233 {
234 	struct net_bridge *br = netdev_priv(dev);
235 	struct sockaddr *addr = p;
236 
237 	if (!is_valid_ether_addr(addr->sa_data))
238 		return -EADDRNOTAVAIL;
239 
240 	spin_lock_bh(&br->lock);
241 	if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
242 		/* Mac address will be changed in br_stp_change_bridge_id(). */
243 		br_stp_change_bridge_id(br, addr->sa_data);
244 	}
245 	spin_unlock_bh(&br->lock);
246 
247 	return 0;
248 }
249 
250 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
251 {
252 	strlcpy(info->driver, "bridge", sizeof(info->driver));
253 	strlcpy(info->version, BR_VERSION, sizeof(info->version));
254 	strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
255 	strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
256 }
257 
258 static netdev_features_t br_fix_features(struct net_device *dev,
259 	netdev_features_t features)
260 {
261 	struct net_bridge *br = netdev_priv(dev);
262 
263 	return br_features_recompute(br, features);
264 }
265 
266 #ifdef CONFIG_NET_POLL_CONTROLLER
267 static void br_poll_controller(struct net_device *br_dev)
268 {
269 }
270 
271 static void br_netpoll_cleanup(struct net_device *dev)
272 {
273 	struct net_bridge *br = netdev_priv(dev);
274 	struct net_bridge_port *p;
275 
276 	list_for_each_entry(p, &br->port_list, list)
277 		br_netpoll_disable(p);
278 }
279 
280 static int __br_netpoll_enable(struct net_bridge_port *p)
281 {
282 	struct netpoll *np;
283 	int err;
284 
285 	np = kzalloc(sizeof(*p->np), GFP_KERNEL);
286 	if (!np)
287 		return -ENOMEM;
288 
289 	err = __netpoll_setup(np, p->dev);
290 	if (err) {
291 		kfree(np);
292 		return err;
293 	}
294 
295 	p->np = np;
296 	return err;
297 }
298 
299 int br_netpoll_enable(struct net_bridge_port *p)
300 {
301 	if (!p->br->dev->npinfo)
302 		return 0;
303 
304 	return __br_netpoll_enable(p);
305 }
306 
307 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
308 {
309 	struct net_bridge *br = netdev_priv(dev);
310 	struct net_bridge_port *p;
311 	int err = 0;
312 
313 	list_for_each_entry(p, &br->port_list, list) {
314 		if (!p->dev)
315 			continue;
316 		err = __br_netpoll_enable(p);
317 		if (err)
318 			goto fail;
319 	}
320 
321 out:
322 	return err;
323 
324 fail:
325 	br_netpoll_cleanup(dev);
326 	goto out;
327 }
328 
329 void br_netpoll_disable(struct net_bridge_port *p)
330 {
331 	struct netpoll *np = p->np;
332 
333 	if (!np)
334 		return;
335 
336 	p->np = NULL;
337 
338 	__netpoll_free_async(np);
339 }
340 
341 #endif
342 
343 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
344 			struct netlink_ext_ack *extack)
345 
346 {
347 	struct net_bridge *br = netdev_priv(dev);
348 
349 	return br_add_if(br, slave_dev, extack);
350 }
351 
352 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
353 {
354 	struct net_bridge *br = netdev_priv(dev);
355 
356 	return br_del_if(br, slave_dev);
357 }
358 
359 static const struct ethtool_ops br_ethtool_ops = {
360 	.get_drvinfo    = br_getinfo,
361 	.get_link	= ethtool_op_get_link,
362 };
363 
364 static const struct net_device_ops br_netdev_ops = {
365 	.ndo_open		 = br_dev_open,
366 	.ndo_stop		 = br_dev_stop,
367 	.ndo_init		 = br_dev_init,
368 	.ndo_uninit		 = br_dev_uninit,
369 	.ndo_start_xmit		 = br_dev_xmit,
370 	.ndo_get_stats64	 = br_get_stats64,
371 	.ndo_set_mac_address	 = br_set_mac_address,
372 	.ndo_set_rx_mode	 = br_dev_set_multicast_list,
373 	.ndo_change_rx_flags	 = br_dev_change_rx_flags,
374 	.ndo_change_mtu		 = br_change_mtu,
375 	.ndo_do_ioctl		 = br_dev_ioctl,
376 #ifdef CONFIG_NET_POLL_CONTROLLER
377 	.ndo_netpoll_setup	 = br_netpoll_setup,
378 	.ndo_netpoll_cleanup	 = br_netpoll_cleanup,
379 	.ndo_poll_controller	 = br_poll_controller,
380 #endif
381 	.ndo_add_slave		 = br_add_slave,
382 	.ndo_del_slave		 = br_del_slave,
383 	.ndo_fix_features        = br_fix_features,
384 	.ndo_fdb_add		 = br_fdb_add,
385 	.ndo_fdb_del		 = br_fdb_delete,
386 	.ndo_fdb_dump		 = br_fdb_dump,
387 	.ndo_bridge_getlink	 = br_getlink,
388 	.ndo_bridge_setlink	 = br_setlink,
389 	.ndo_bridge_dellink	 = br_dellink,
390 	.ndo_features_check	 = passthru_features_check,
391 };
392 
393 static struct device_type br_type = {
394 	.name	= "bridge",
395 };
396 
397 void br_dev_setup(struct net_device *dev)
398 {
399 	struct net_bridge *br = netdev_priv(dev);
400 
401 	eth_hw_addr_random(dev);
402 	ether_setup(dev);
403 
404 	dev->netdev_ops = &br_netdev_ops;
405 	dev->needs_free_netdev = true;
406 	dev->ethtool_ops = &br_ethtool_ops;
407 	SET_NETDEV_DEVTYPE(dev, &br_type);
408 	dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
409 
410 	dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
411 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
412 	dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
413 			   NETIF_F_HW_VLAN_STAG_TX;
414 	dev->vlan_features = COMMON_FEATURES;
415 
416 	br->dev = dev;
417 	spin_lock_init(&br->lock);
418 	INIT_LIST_HEAD(&br->port_list);
419 	spin_lock_init(&br->hash_lock);
420 
421 	br->bridge_id.prio[0] = 0x80;
422 	br->bridge_id.prio[1] = 0x00;
423 
424 	ether_addr_copy(br->group_addr, eth_stp_addr);
425 
426 	br->stp_enabled = BR_NO_STP;
427 	br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
428 	br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
429 
430 	br->designated_root = br->bridge_id;
431 	br->bridge_max_age = br->max_age = 20 * HZ;
432 	br->bridge_hello_time = br->hello_time = 2 * HZ;
433 	br->bridge_forward_delay = br->forward_delay = 15 * HZ;
434 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
435 	dev->max_mtu = ETH_MAX_MTU;
436 
437 	br_netfilter_rtable_init(br);
438 	br_stp_timer_init(br);
439 	br_multicast_init(br);
440 	INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
441 }
442