xref: /openbmc/linux/net/bridge/br_device.c (revision d003d772)
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_opt_get(br, BROPT_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_opt_get(br, BROPT_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_fdb_hash_init(br);
129 	if (err) {
130 		free_percpu(br->stats);
131 		return err;
132 	}
133 
134 	err = br_mdb_hash_init(br);
135 	if (err) {
136 		free_percpu(br->stats);
137 		br_fdb_hash_fini(br);
138 		return err;
139 	}
140 
141 	err = br_vlan_init(br);
142 	if (err) {
143 		free_percpu(br->stats);
144 		br_mdb_hash_fini(br);
145 		br_fdb_hash_fini(br);
146 		return err;
147 	}
148 
149 	err = br_multicast_init_stats(br);
150 	if (err) {
151 		free_percpu(br->stats);
152 		br_vlan_flush(br);
153 		br_mdb_hash_fini(br);
154 		br_fdb_hash_fini(br);
155 	}
156 	br_set_lockdep_class(dev);
157 
158 	return err;
159 }
160 
161 static void br_dev_uninit(struct net_device *dev)
162 {
163 	struct net_bridge *br = netdev_priv(dev);
164 
165 	br_multicast_dev_del(br);
166 	br_multicast_uninit_stats(br);
167 	br_vlan_flush(br);
168 	br_mdb_hash_fini(br);
169 	br_fdb_hash_fini(br);
170 	free_percpu(br->stats);
171 }
172 
173 static int br_dev_open(struct net_device *dev)
174 {
175 	struct net_bridge *br = netdev_priv(dev);
176 
177 	netdev_update_features(dev);
178 	netif_start_queue(dev);
179 	br_stp_enable_bridge(br);
180 	br_multicast_open(br);
181 
182 	return 0;
183 }
184 
185 static void br_dev_set_multicast_list(struct net_device *dev)
186 {
187 }
188 
189 static void br_dev_change_rx_flags(struct net_device *dev, int change)
190 {
191 	if (change & IFF_PROMISC)
192 		br_manage_promisc(netdev_priv(dev));
193 }
194 
195 static int br_dev_stop(struct net_device *dev)
196 {
197 	struct net_bridge *br = netdev_priv(dev);
198 
199 	br_stp_disable_bridge(br);
200 	br_multicast_stop(br);
201 
202 	netif_stop_queue(dev);
203 
204 	return 0;
205 }
206 
207 static void br_get_stats64(struct net_device *dev,
208 			   struct rtnl_link_stats64 *stats)
209 {
210 	struct net_bridge *br = netdev_priv(dev);
211 	struct pcpu_sw_netstats tmp, sum = { 0 };
212 	unsigned int cpu;
213 
214 	for_each_possible_cpu(cpu) {
215 		unsigned int start;
216 		const struct pcpu_sw_netstats *bstats
217 			= per_cpu_ptr(br->stats, cpu);
218 		do {
219 			start = u64_stats_fetch_begin_irq(&bstats->syncp);
220 			memcpy(&tmp, bstats, sizeof(tmp));
221 		} while (u64_stats_fetch_retry_irq(&bstats->syncp, start));
222 		sum.tx_bytes   += tmp.tx_bytes;
223 		sum.tx_packets += tmp.tx_packets;
224 		sum.rx_bytes   += tmp.rx_bytes;
225 		sum.rx_packets += tmp.rx_packets;
226 	}
227 
228 	stats->tx_bytes   = sum.tx_bytes;
229 	stats->tx_packets = sum.tx_packets;
230 	stats->rx_bytes   = sum.rx_bytes;
231 	stats->rx_packets = sum.rx_packets;
232 }
233 
234 static int br_change_mtu(struct net_device *dev, int new_mtu)
235 {
236 	struct net_bridge *br = netdev_priv(dev);
237 
238 	dev->mtu = new_mtu;
239 
240 	/* this flag will be cleared if the MTU was automatically adjusted */
241 	br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
242 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
243 	/* remember the MTU in the rtable for PMTU */
244 	dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
245 #endif
246 
247 	return 0;
248 }
249 
250 /* Allow setting mac address to any valid ethernet address. */
251 static int br_set_mac_address(struct net_device *dev, void *p)
252 {
253 	struct net_bridge *br = netdev_priv(dev);
254 	struct sockaddr *addr = p;
255 
256 	if (!is_valid_ether_addr(addr->sa_data))
257 		return -EADDRNOTAVAIL;
258 
259 	spin_lock_bh(&br->lock);
260 	if (!ether_addr_equal(dev->dev_addr, addr->sa_data)) {
261 		/* Mac address will be changed in br_stp_change_bridge_id(). */
262 		br_stp_change_bridge_id(br, addr->sa_data);
263 	}
264 	spin_unlock_bh(&br->lock);
265 
266 	return 0;
267 }
268 
269 static void br_getinfo(struct net_device *dev, struct ethtool_drvinfo *info)
270 {
271 	strlcpy(info->driver, "bridge", sizeof(info->driver));
272 	strlcpy(info->version, BR_VERSION, sizeof(info->version));
273 	strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
274 	strlcpy(info->bus_info, "N/A", sizeof(info->bus_info));
275 }
276 
277 static netdev_features_t br_fix_features(struct net_device *dev,
278 	netdev_features_t features)
279 {
280 	struct net_bridge *br = netdev_priv(dev);
281 
282 	return br_features_recompute(br, features);
283 }
284 
285 #ifdef CONFIG_NET_POLL_CONTROLLER
286 static void br_poll_controller(struct net_device *br_dev)
287 {
288 }
289 
290 static void br_netpoll_cleanup(struct net_device *dev)
291 {
292 	struct net_bridge *br = netdev_priv(dev);
293 	struct net_bridge_port *p;
294 
295 	list_for_each_entry(p, &br->port_list, list)
296 		br_netpoll_disable(p);
297 }
298 
299 static int __br_netpoll_enable(struct net_bridge_port *p)
300 {
301 	struct netpoll *np;
302 	int err;
303 
304 	np = kzalloc(sizeof(*p->np), GFP_KERNEL);
305 	if (!np)
306 		return -ENOMEM;
307 
308 	err = __netpoll_setup(np, p->dev);
309 	if (err) {
310 		kfree(np);
311 		return err;
312 	}
313 
314 	p->np = np;
315 	return err;
316 }
317 
318 int br_netpoll_enable(struct net_bridge_port *p)
319 {
320 	if (!p->br->dev->npinfo)
321 		return 0;
322 
323 	return __br_netpoll_enable(p);
324 }
325 
326 static int br_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
327 {
328 	struct net_bridge *br = netdev_priv(dev);
329 	struct net_bridge_port *p;
330 	int err = 0;
331 
332 	list_for_each_entry(p, &br->port_list, list) {
333 		if (!p->dev)
334 			continue;
335 		err = __br_netpoll_enable(p);
336 		if (err)
337 			goto fail;
338 	}
339 
340 out:
341 	return err;
342 
343 fail:
344 	br_netpoll_cleanup(dev);
345 	goto out;
346 }
347 
348 void br_netpoll_disable(struct net_bridge_port *p)
349 {
350 	struct netpoll *np = p->np;
351 
352 	if (!np)
353 		return;
354 
355 	p->np = NULL;
356 
357 	__netpoll_free(np);
358 }
359 
360 #endif
361 
362 static int br_add_slave(struct net_device *dev, struct net_device *slave_dev,
363 			struct netlink_ext_ack *extack)
364 
365 {
366 	struct net_bridge *br = netdev_priv(dev);
367 
368 	return br_add_if(br, slave_dev, extack);
369 }
370 
371 static int br_del_slave(struct net_device *dev, struct net_device *slave_dev)
372 {
373 	struct net_bridge *br = netdev_priv(dev);
374 
375 	return br_del_if(br, slave_dev);
376 }
377 
378 static const struct ethtool_ops br_ethtool_ops = {
379 	.get_drvinfo    = br_getinfo,
380 	.get_link	= ethtool_op_get_link,
381 };
382 
383 static const struct net_device_ops br_netdev_ops = {
384 	.ndo_open		 = br_dev_open,
385 	.ndo_stop		 = br_dev_stop,
386 	.ndo_init		 = br_dev_init,
387 	.ndo_uninit		 = br_dev_uninit,
388 	.ndo_start_xmit		 = br_dev_xmit,
389 	.ndo_get_stats64	 = br_get_stats64,
390 	.ndo_set_mac_address	 = br_set_mac_address,
391 	.ndo_set_rx_mode	 = br_dev_set_multicast_list,
392 	.ndo_change_rx_flags	 = br_dev_change_rx_flags,
393 	.ndo_change_mtu		 = br_change_mtu,
394 	.ndo_do_ioctl		 = br_dev_ioctl,
395 #ifdef CONFIG_NET_POLL_CONTROLLER
396 	.ndo_netpoll_setup	 = br_netpoll_setup,
397 	.ndo_netpoll_cleanup	 = br_netpoll_cleanup,
398 	.ndo_poll_controller	 = br_poll_controller,
399 #endif
400 	.ndo_add_slave		 = br_add_slave,
401 	.ndo_del_slave		 = br_del_slave,
402 	.ndo_fix_features        = br_fix_features,
403 	.ndo_fdb_add		 = br_fdb_add,
404 	.ndo_fdb_del		 = br_fdb_delete,
405 	.ndo_fdb_dump		 = br_fdb_dump,
406 	.ndo_fdb_get		 = br_fdb_get,
407 	.ndo_bridge_getlink	 = br_getlink,
408 	.ndo_bridge_setlink	 = br_setlink,
409 	.ndo_bridge_dellink	 = br_dellink,
410 	.ndo_features_check	 = passthru_features_check,
411 };
412 
413 static struct device_type br_type = {
414 	.name	= "bridge",
415 };
416 
417 void br_dev_setup(struct net_device *dev)
418 {
419 	struct net_bridge *br = netdev_priv(dev);
420 
421 	eth_hw_addr_random(dev);
422 	ether_setup(dev);
423 
424 	dev->netdev_ops = &br_netdev_ops;
425 	dev->needs_free_netdev = true;
426 	dev->ethtool_ops = &br_ethtool_ops;
427 	SET_NETDEV_DEVTYPE(dev, &br_type);
428 	dev->priv_flags = IFF_EBRIDGE | IFF_NO_QUEUE;
429 
430 	dev->features = COMMON_FEATURES | NETIF_F_LLTX | NETIF_F_NETNS_LOCAL |
431 			NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
432 	dev->hw_features = COMMON_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
433 			   NETIF_F_HW_VLAN_STAG_TX;
434 	dev->vlan_features = COMMON_FEATURES;
435 
436 	br->dev = dev;
437 	spin_lock_init(&br->lock);
438 	INIT_LIST_HEAD(&br->port_list);
439 	INIT_HLIST_HEAD(&br->fdb_list);
440 	spin_lock_init(&br->hash_lock);
441 
442 	br->bridge_id.prio[0] = 0x80;
443 	br->bridge_id.prio[1] = 0x00;
444 
445 	ether_addr_copy(br->group_addr, eth_stp_addr);
446 
447 	br->stp_enabled = BR_NO_STP;
448 	br->group_fwd_mask = BR_GROUPFWD_DEFAULT;
449 	br->group_fwd_mask_required = BR_GROUPFWD_DEFAULT;
450 
451 	br->designated_root = br->bridge_id;
452 	br->bridge_max_age = br->max_age = 20 * HZ;
453 	br->bridge_hello_time = br->hello_time = 2 * HZ;
454 	br->bridge_forward_delay = br->forward_delay = 15 * HZ;
455 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
456 	dev->max_mtu = ETH_MAX_MTU;
457 
458 	br_netfilter_rtable_init(br);
459 	br_stp_timer_init(br);
460 	br_multicast_init(br);
461 	INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
462 }
463