xref: /openbmc/linux/net/bridge/br_mdb.c (revision 39b6f3aa)
1 #include <linux/err.h>
2 #include <linux/igmp.h>
3 #include <linux/kernel.h>
4 #include <linux/netdevice.h>
5 #include <linux/rculist.h>
6 #include <linux/skbuff.h>
7 #include <linux/if_ether.h>
8 #include <net/ip.h>
9 #include <net/netlink.h>
10 #if IS_ENABLED(CONFIG_IPV6)
11 #include <net/ipv6.h>
12 #endif
13 
14 #include "br_private.h"
15 
16 static int br_rports_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
17 			       struct net_device *dev)
18 {
19 	struct net_bridge *br = netdev_priv(dev);
20 	struct net_bridge_port *p;
21 	struct nlattr *nest;
22 
23 	if (!br->multicast_router || hlist_empty(&br->router_list))
24 		return 0;
25 
26 	nest = nla_nest_start(skb, MDBA_ROUTER);
27 	if (nest == NULL)
28 		return -EMSGSIZE;
29 
30 	hlist_for_each_entry_rcu(p, &br->router_list, rlist) {
31 		if (p && nla_put_u32(skb, MDBA_ROUTER_PORT, p->dev->ifindex))
32 			goto fail;
33 	}
34 
35 	nla_nest_end(skb, nest);
36 	return 0;
37 fail:
38 	nla_nest_cancel(skb, nest);
39 	return -EMSGSIZE;
40 }
41 
42 static int br_mdb_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
43 			    struct net_device *dev)
44 {
45 	struct net_bridge *br = netdev_priv(dev);
46 	struct net_bridge_mdb_htable *mdb;
47 	struct nlattr *nest, *nest2;
48 	int i, err = 0;
49 	int idx = 0, s_idx = cb->args[1];
50 
51 	if (br->multicast_disabled)
52 		return 0;
53 
54 	mdb = rcu_dereference(br->mdb);
55 	if (!mdb)
56 		return 0;
57 
58 	nest = nla_nest_start(skb, MDBA_MDB);
59 	if (nest == NULL)
60 		return -EMSGSIZE;
61 
62 	for (i = 0; i < mdb->max; i++) {
63 		struct net_bridge_mdb_entry *mp;
64 		struct net_bridge_port_group *p, **pp;
65 		struct net_bridge_port *port;
66 
67 		hlist_for_each_entry_rcu(mp, &mdb->mhash[i], hlist[mdb->ver]) {
68 			if (idx < s_idx)
69 				goto skip;
70 
71 			nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
72 			if (nest2 == NULL) {
73 				err = -EMSGSIZE;
74 				goto out;
75 			}
76 
77 			for (pp = &mp->ports;
78 			     (p = rcu_dereference(*pp)) != NULL;
79 			      pp = &p->next) {
80 				port = p->port;
81 				if (port) {
82 					struct br_mdb_entry e;
83 					memset(&e, 0, sizeof(e));
84 					e.ifindex = port->dev->ifindex;
85 					e.state = p->state;
86 					if (p->addr.proto == htons(ETH_P_IP))
87 						e.addr.u.ip4 = p->addr.u.ip4;
88 #if IS_ENABLED(CONFIG_IPV6)
89 					if (p->addr.proto == htons(ETH_P_IPV6))
90 						e.addr.u.ip6 = p->addr.u.ip6;
91 #endif
92 					e.addr.proto = p->addr.proto;
93 					if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(e), &e)) {
94 						nla_nest_cancel(skb, nest2);
95 						err = -EMSGSIZE;
96 						goto out;
97 					}
98 				}
99 			}
100 			nla_nest_end(skb, nest2);
101 		skip:
102 			idx++;
103 		}
104 	}
105 
106 out:
107 	cb->args[1] = idx;
108 	nla_nest_end(skb, nest);
109 	return err;
110 }
111 
112 static int br_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
113 {
114 	struct net_device *dev;
115 	struct net *net = sock_net(skb->sk);
116 	struct nlmsghdr *nlh = NULL;
117 	int idx = 0, s_idx;
118 
119 	s_idx = cb->args[0];
120 
121 	rcu_read_lock();
122 
123 	/* In theory this could be wrapped to 0... */
124 	cb->seq = net->dev_base_seq + br_mdb_rehash_seq;
125 
126 	for_each_netdev_rcu(net, dev) {
127 		if (dev->priv_flags & IFF_EBRIDGE) {
128 			struct br_port_msg *bpm;
129 
130 			if (idx < s_idx)
131 				goto skip;
132 
133 			nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
134 					cb->nlh->nlmsg_seq, RTM_GETMDB,
135 					sizeof(*bpm), NLM_F_MULTI);
136 			if (nlh == NULL)
137 				break;
138 
139 			bpm = nlmsg_data(nlh);
140 			memset(bpm, 0, sizeof(*bpm));
141 			bpm->ifindex = dev->ifindex;
142 			if (br_mdb_fill_info(skb, cb, dev) < 0)
143 				goto out;
144 			if (br_rports_fill_info(skb, cb, dev) < 0)
145 				goto out;
146 
147 			cb->args[1] = 0;
148 			nlmsg_end(skb, nlh);
149 		skip:
150 			idx++;
151 		}
152 	}
153 
154 out:
155 	if (nlh)
156 		nlmsg_end(skb, nlh);
157 	rcu_read_unlock();
158 	cb->args[0] = idx;
159 	return skb->len;
160 }
161 
162 static int nlmsg_populate_mdb_fill(struct sk_buff *skb,
163 				   struct net_device *dev,
164 				   struct br_mdb_entry *entry, u32 pid,
165 				   u32 seq, int type, unsigned int flags)
166 {
167 	struct nlmsghdr *nlh;
168 	struct br_port_msg *bpm;
169 	struct nlattr *nest, *nest2;
170 
171 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*bpm), NLM_F_MULTI);
172 	if (!nlh)
173 		return -EMSGSIZE;
174 
175 	bpm = nlmsg_data(nlh);
176 	memset(bpm, 0, sizeof(*bpm));
177 	bpm->family  = AF_BRIDGE;
178 	bpm->ifindex = dev->ifindex;
179 	nest = nla_nest_start(skb, MDBA_MDB);
180 	if (nest == NULL)
181 		goto cancel;
182 	nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
183 	if (nest2 == NULL)
184 		goto end;
185 
186 	if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(*entry), entry))
187 		goto end;
188 
189 	nla_nest_end(skb, nest2);
190 	nla_nest_end(skb, nest);
191 	return nlmsg_end(skb, nlh);
192 
193 end:
194 	nla_nest_end(skb, nest);
195 cancel:
196 	nlmsg_cancel(skb, nlh);
197 	return -EMSGSIZE;
198 }
199 
200 static inline size_t rtnl_mdb_nlmsg_size(void)
201 {
202 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
203 		+ nla_total_size(sizeof(struct br_mdb_entry));
204 }
205 
206 static void __br_mdb_notify(struct net_device *dev, struct br_mdb_entry *entry,
207 			    int type)
208 {
209 	struct net *net = dev_net(dev);
210 	struct sk_buff *skb;
211 	int err = -ENOBUFS;
212 
213 	skb = nlmsg_new(rtnl_mdb_nlmsg_size(), GFP_ATOMIC);
214 	if (!skb)
215 		goto errout;
216 
217 	err = nlmsg_populate_mdb_fill(skb, dev, entry, 0, 0, type, NTF_SELF);
218 	if (err < 0) {
219 		kfree_skb(skb);
220 		goto errout;
221 	}
222 
223 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
224 	return;
225 errout:
226 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
227 }
228 
229 void br_mdb_notify(struct net_device *dev, struct net_bridge_port *port,
230 		   struct br_ip *group, int type)
231 {
232 	struct br_mdb_entry entry;
233 
234 	memset(&entry, 0, sizeof(entry));
235 	entry.ifindex = port->dev->ifindex;
236 	entry.addr.proto = group->proto;
237 	entry.addr.u.ip4 = group->u.ip4;
238 #if IS_ENABLED(CONFIG_IPV6)
239 	entry.addr.u.ip6 = group->u.ip6;
240 #endif
241 	__br_mdb_notify(dev, &entry, type);
242 }
243 
244 static bool is_valid_mdb_entry(struct br_mdb_entry *entry)
245 {
246 	if (entry->ifindex == 0)
247 		return false;
248 
249 	if (entry->addr.proto == htons(ETH_P_IP)) {
250 		if (!ipv4_is_multicast(entry->addr.u.ip4))
251 			return false;
252 		if (ipv4_is_local_multicast(entry->addr.u.ip4))
253 			return false;
254 #if IS_ENABLED(CONFIG_IPV6)
255 	} else if (entry->addr.proto == htons(ETH_P_IPV6)) {
256 		if (!ipv6_is_transient_multicast(&entry->addr.u.ip6))
257 			return false;
258 #endif
259 	} else
260 		return false;
261 	if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY)
262 		return false;
263 
264 	return true;
265 }
266 
267 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
268 			struct net_device **pdev, struct br_mdb_entry **pentry)
269 {
270 	struct net *net = sock_net(skb->sk);
271 	struct br_mdb_entry *entry;
272 	struct br_port_msg *bpm;
273 	struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
274 	struct net_device *dev;
275 	int err;
276 
277 	err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY, NULL);
278 	if (err < 0)
279 		return err;
280 
281 	bpm = nlmsg_data(nlh);
282 	if (bpm->ifindex == 0) {
283 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid ifindex\n");
284 		return -EINVAL;
285 	}
286 
287 	dev = __dev_get_by_index(net, bpm->ifindex);
288 	if (dev == NULL) {
289 		pr_info("PF_BRIDGE: br_mdb_parse() with unknown ifindex\n");
290 		return -ENODEV;
291 	}
292 
293 	if (!(dev->priv_flags & IFF_EBRIDGE)) {
294 		pr_info("PF_BRIDGE: br_mdb_parse() with non-bridge\n");
295 		return -EOPNOTSUPP;
296 	}
297 
298 	*pdev = dev;
299 
300 	if (!tb[MDBA_SET_ENTRY] ||
301 	    nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
302 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid attr\n");
303 		return -EINVAL;
304 	}
305 
306 	entry = nla_data(tb[MDBA_SET_ENTRY]);
307 	if (!is_valid_mdb_entry(entry)) {
308 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid entry\n");
309 		return -EINVAL;
310 	}
311 
312 	*pentry = entry;
313 	return 0;
314 }
315 
316 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
317 			    struct br_ip *group, unsigned char state)
318 {
319 	struct net_bridge_mdb_entry *mp;
320 	struct net_bridge_port_group *p;
321 	struct net_bridge_port_group __rcu **pp;
322 	struct net_bridge_mdb_htable *mdb;
323 	int err;
324 
325 	mdb = mlock_dereference(br->mdb, br);
326 	mp = br_mdb_ip_get(mdb, group);
327 	if (!mp) {
328 		mp = br_multicast_new_group(br, port, group);
329 		err = PTR_ERR(mp);
330 		if (IS_ERR(mp))
331 			return err;
332 	}
333 
334 	for (pp = &mp->ports;
335 	     (p = mlock_dereference(*pp, br)) != NULL;
336 	     pp = &p->next) {
337 		if (p->port == port)
338 			return -EEXIST;
339 		if ((unsigned long)p->port < (unsigned long)port)
340 			break;
341 	}
342 
343 	p = br_multicast_new_port_group(port, group, *pp, state);
344 	if (unlikely(!p))
345 		return -ENOMEM;
346 	rcu_assign_pointer(*pp, p);
347 
348 	br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
349 	return 0;
350 }
351 
352 static int __br_mdb_add(struct net *net, struct net_bridge *br,
353 			struct br_mdb_entry *entry)
354 {
355 	struct br_ip ip;
356 	struct net_device *dev;
357 	struct net_bridge_port *p;
358 	int ret;
359 
360 	if (!netif_running(br->dev) || br->multicast_disabled)
361 		return -EINVAL;
362 
363 	dev = __dev_get_by_index(net, entry->ifindex);
364 	if (!dev)
365 		return -ENODEV;
366 
367 	p = br_port_get_rtnl(dev);
368 	if (!p || p->br != br || p->state == BR_STATE_DISABLED)
369 		return -EINVAL;
370 
371 	ip.proto = entry->addr.proto;
372 	if (ip.proto == htons(ETH_P_IP))
373 		ip.u.ip4 = entry->addr.u.ip4;
374 #if IS_ENABLED(CONFIG_IPV6)
375 	else
376 		ip.u.ip6 = entry->addr.u.ip6;
377 #endif
378 
379 	spin_lock_bh(&br->multicast_lock);
380 	ret = br_mdb_add_group(br, p, &ip, entry->state);
381 	spin_unlock_bh(&br->multicast_lock);
382 	return ret;
383 }
384 
385 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
386 {
387 	struct net *net = sock_net(skb->sk);
388 	struct br_mdb_entry *entry;
389 	struct net_device *dev;
390 	struct net_bridge *br;
391 	int err;
392 
393 	err = br_mdb_parse(skb, nlh, &dev, &entry);
394 	if (err < 0)
395 		return err;
396 
397 	br = netdev_priv(dev);
398 
399 	err = __br_mdb_add(net, br, entry);
400 	if (!err)
401 		__br_mdb_notify(dev, entry, RTM_NEWMDB);
402 	return err;
403 }
404 
405 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry)
406 {
407 	struct net_bridge_mdb_htable *mdb;
408 	struct net_bridge_mdb_entry *mp;
409 	struct net_bridge_port_group *p;
410 	struct net_bridge_port_group __rcu **pp;
411 	struct br_ip ip;
412 	int err = -EINVAL;
413 
414 	if (!netif_running(br->dev) || br->multicast_disabled)
415 		return -EINVAL;
416 
417 	if (timer_pending(&br->multicast_querier_timer))
418 		return -EBUSY;
419 
420 	ip.proto = entry->addr.proto;
421 	if (ip.proto == htons(ETH_P_IP))
422 		ip.u.ip4 = entry->addr.u.ip4;
423 #if IS_ENABLED(CONFIG_IPV6)
424 	else
425 		ip.u.ip6 = entry->addr.u.ip6;
426 #endif
427 
428 	spin_lock_bh(&br->multicast_lock);
429 	mdb = mlock_dereference(br->mdb, br);
430 
431 	mp = br_mdb_ip_get(mdb, &ip);
432 	if (!mp)
433 		goto unlock;
434 
435 	for (pp = &mp->ports;
436 	     (p = mlock_dereference(*pp, br)) != NULL;
437 	     pp = &p->next) {
438 		if (!p->port || p->port->dev->ifindex != entry->ifindex)
439 			continue;
440 
441 		if (p->port->state == BR_STATE_DISABLED)
442 			goto unlock;
443 
444 		rcu_assign_pointer(*pp, p->next);
445 		hlist_del_init(&p->mglist);
446 		del_timer(&p->timer);
447 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
448 		err = 0;
449 
450 		if (!mp->ports && !mp->mglist &&
451 		    netif_running(br->dev))
452 			mod_timer(&mp->timer, jiffies);
453 		break;
454 	}
455 
456 unlock:
457 	spin_unlock_bh(&br->multicast_lock);
458 	return err;
459 }
460 
461 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
462 {
463 	struct net_device *dev;
464 	struct br_mdb_entry *entry;
465 	struct net_bridge *br;
466 	int err;
467 
468 	err = br_mdb_parse(skb, nlh, &dev, &entry);
469 	if (err < 0)
470 		return err;
471 
472 	br = netdev_priv(dev);
473 
474 	err = __br_mdb_del(br, entry);
475 	if (!err)
476 		__br_mdb_notify(dev, entry, RTM_DELMDB);
477 	return err;
478 }
479 
480 void br_mdb_init(void)
481 {
482 	rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, NULL);
483 	rtnl_register(PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, NULL);
484 	rtnl_register(PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, NULL);
485 }
486 
487 void br_mdb_uninit(void)
488 {
489 	rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
490 	rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
491 	rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
492 }
493