xref: /openbmc/linux/net/bridge/br_mdb.c (revision 9d749629)
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 hlist_node *n;
22 	struct nlattr *nest;
23 
24 	if (!br->multicast_router || hlist_empty(&br->router_list))
25 		return 0;
26 
27 	nest = nla_nest_start(skb, MDBA_ROUTER);
28 	if (nest == NULL)
29 		return -EMSGSIZE;
30 
31 	hlist_for_each_entry_rcu(p, n, &br->router_list, rlist) {
32 		if (p && nla_put_u32(skb, MDBA_ROUTER_PORT, p->dev->ifindex))
33 			goto fail;
34 	}
35 
36 	nla_nest_end(skb, nest);
37 	return 0;
38 fail:
39 	nla_nest_cancel(skb, nest);
40 	return -EMSGSIZE;
41 }
42 
43 static int br_mdb_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
44 			    struct net_device *dev)
45 {
46 	struct net_bridge *br = netdev_priv(dev);
47 	struct net_bridge_mdb_htable *mdb;
48 	struct nlattr *nest, *nest2;
49 	int i, err = 0;
50 	int idx = 0, s_idx = cb->args[1];
51 
52 	if (br->multicast_disabled)
53 		return 0;
54 
55 	mdb = rcu_dereference(br->mdb);
56 	if (!mdb)
57 		return 0;
58 
59 	nest = nla_nest_start(skb, MDBA_MDB);
60 	if (nest == NULL)
61 		return -EMSGSIZE;
62 
63 	for (i = 0; i < mdb->max; i++) {
64 		struct hlist_node *h;
65 		struct net_bridge_mdb_entry *mp;
66 		struct net_bridge_port_group *p, **pp;
67 		struct net_bridge_port *port;
68 
69 		hlist_for_each_entry_rcu(mp, h, &mdb->mhash[i], hlist[mdb->ver]) {
70 			if (idx < s_idx)
71 				goto skip;
72 
73 			nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
74 			if (nest2 == NULL) {
75 				err = -EMSGSIZE;
76 				goto out;
77 			}
78 
79 			for (pp = &mp->ports;
80 			     (p = rcu_dereference(*pp)) != NULL;
81 			      pp = &p->next) {
82 				port = p->port;
83 				if (port) {
84 					struct br_mdb_entry e;
85 					e.ifindex = port->dev->ifindex;
86 					e.state = p->state;
87 					if (p->addr.proto == htons(ETH_P_IP))
88 						e.addr.u.ip4 = p->addr.u.ip4;
89 #if IS_ENABLED(CONFIG_IPV6)
90 					if (p->addr.proto == htons(ETH_P_IPV6))
91 						e.addr.u.ip6 = p->addr.u.ip6;
92 #endif
93 					e.addr.proto = p->addr.proto;
94 					if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(e), &e)) {
95 						nla_nest_cancel(skb, nest2);
96 						err = -EMSGSIZE;
97 						goto out;
98 					}
99 				}
100 			}
101 			nla_nest_end(skb, nest2);
102 		skip:
103 			idx++;
104 		}
105 	}
106 
107 out:
108 	cb->args[1] = idx;
109 	nla_nest_end(skb, nest);
110 	return err;
111 }
112 
113 static int br_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
114 {
115 	struct net_device *dev;
116 	struct net *net = sock_net(skb->sk);
117 	struct nlmsghdr *nlh = NULL;
118 	int idx = 0, s_idx;
119 
120 	s_idx = cb->args[0];
121 
122 	rcu_read_lock();
123 
124 	/* In theory this could be wrapped to 0... */
125 	cb->seq = net->dev_base_seq + br_mdb_rehash_seq;
126 
127 	for_each_netdev_rcu(net, dev) {
128 		if (dev->priv_flags & IFF_EBRIDGE) {
129 			struct br_port_msg *bpm;
130 
131 			if (idx < s_idx)
132 				goto skip;
133 
134 			nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
135 					cb->nlh->nlmsg_seq, RTM_GETMDB,
136 					sizeof(*bpm), NLM_F_MULTI);
137 			if (nlh == NULL)
138 				break;
139 
140 			bpm = nlmsg_data(nlh);
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 	bpm->family  = AF_BRIDGE;
177 	bpm->ifindex = dev->ifindex;
178 	nest = nla_nest_start(skb, MDBA_MDB);
179 	if (nest == NULL)
180 		goto cancel;
181 	nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
182 	if (nest2 == NULL)
183 		goto end;
184 
185 	if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(*entry), entry))
186 		goto end;
187 
188 	nla_nest_end(skb, nest2);
189 	nla_nest_end(skb, nest);
190 	return nlmsg_end(skb, nlh);
191 
192 end:
193 	nla_nest_end(skb, nest);
194 cancel:
195 	nlmsg_cancel(skb, nlh);
196 	return -EMSGSIZE;
197 }
198 
199 static inline size_t rtnl_mdb_nlmsg_size(void)
200 {
201 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
202 		+ nla_total_size(sizeof(struct br_mdb_entry));
203 }
204 
205 static void __br_mdb_notify(struct net_device *dev, struct br_mdb_entry *entry,
206 			    int type)
207 {
208 	struct net *net = dev_net(dev);
209 	struct sk_buff *skb;
210 	int err = -ENOBUFS;
211 
212 	skb = nlmsg_new(rtnl_mdb_nlmsg_size(), GFP_ATOMIC);
213 	if (!skb)
214 		goto errout;
215 
216 	err = nlmsg_populate_mdb_fill(skb, dev, entry, 0, 0, type, NTF_SELF);
217 	if (err < 0) {
218 		kfree_skb(skb);
219 		goto errout;
220 	}
221 
222 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
223 	return;
224 errout:
225 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
226 }
227 
228 void br_mdb_notify(struct net_device *dev, struct net_bridge_port *port,
229 		   struct br_ip *group, int type)
230 {
231 	struct br_mdb_entry entry;
232 
233 	entry.ifindex = port->dev->ifindex;
234 	entry.addr.proto = group->proto;
235 	entry.addr.u.ip4 = group->u.ip4;
236 #if IS_ENABLED(CONFIG_IPV6)
237 	entry.addr.u.ip6 = group->u.ip6;
238 #endif
239 	__br_mdb_notify(dev, &entry, type);
240 }
241 
242 static bool is_valid_mdb_entry(struct br_mdb_entry *entry)
243 {
244 	if (entry->ifindex == 0)
245 		return false;
246 
247 	if (entry->addr.proto == htons(ETH_P_IP)) {
248 		if (!ipv4_is_multicast(entry->addr.u.ip4))
249 			return false;
250 		if (ipv4_is_local_multicast(entry->addr.u.ip4))
251 			return false;
252 #if IS_ENABLED(CONFIG_IPV6)
253 	} else if (entry->addr.proto == htons(ETH_P_IPV6)) {
254 		if (!ipv6_is_transient_multicast(&entry->addr.u.ip6))
255 			return false;
256 #endif
257 	} else
258 		return false;
259 	if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY)
260 		return false;
261 
262 	return true;
263 }
264 
265 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
266 			struct net_device **pdev, struct br_mdb_entry **pentry)
267 {
268 	struct net *net = sock_net(skb->sk);
269 	struct br_mdb_entry *entry;
270 	struct br_port_msg *bpm;
271 	struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
272 	struct net_device *dev;
273 	int err;
274 
275 	err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY, NULL);
276 	if (err < 0)
277 		return err;
278 
279 	bpm = nlmsg_data(nlh);
280 	if (bpm->ifindex == 0) {
281 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid ifindex\n");
282 		return -EINVAL;
283 	}
284 
285 	dev = __dev_get_by_index(net, bpm->ifindex);
286 	if (dev == NULL) {
287 		pr_info("PF_BRIDGE: br_mdb_parse() with unknown ifindex\n");
288 		return -ENODEV;
289 	}
290 
291 	if (!(dev->priv_flags & IFF_EBRIDGE)) {
292 		pr_info("PF_BRIDGE: br_mdb_parse() with non-bridge\n");
293 		return -EOPNOTSUPP;
294 	}
295 
296 	*pdev = dev;
297 
298 	if (!tb[MDBA_SET_ENTRY] ||
299 	    nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
300 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid attr\n");
301 		return -EINVAL;
302 	}
303 
304 	entry = nla_data(tb[MDBA_SET_ENTRY]);
305 	if (!is_valid_mdb_entry(entry)) {
306 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid entry\n");
307 		return -EINVAL;
308 	}
309 
310 	*pentry = entry;
311 	return 0;
312 }
313 
314 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
315 			    struct br_ip *group, unsigned char state)
316 {
317 	struct net_bridge_mdb_entry *mp;
318 	struct net_bridge_port_group *p;
319 	struct net_bridge_port_group __rcu **pp;
320 	struct net_bridge_mdb_htable *mdb;
321 	int err;
322 
323 	mdb = mlock_dereference(br->mdb, br);
324 	mp = br_mdb_ip_get(mdb, group);
325 	if (!mp) {
326 		mp = br_multicast_new_group(br, port, group);
327 		err = PTR_ERR(mp);
328 		if (IS_ERR(mp))
329 			return err;
330 	}
331 
332 	for (pp = &mp->ports;
333 	     (p = mlock_dereference(*pp, br)) != NULL;
334 	     pp = &p->next) {
335 		if (p->port == port)
336 			return -EEXIST;
337 		if ((unsigned long)p->port < (unsigned long)port)
338 			break;
339 	}
340 
341 	p = br_multicast_new_port_group(port, group, *pp, state);
342 	if (unlikely(!p))
343 		return -ENOMEM;
344 	rcu_assign_pointer(*pp, p);
345 
346 	br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
347 	return 0;
348 }
349 
350 static int __br_mdb_add(struct net *net, struct net_bridge *br,
351 			struct br_mdb_entry *entry)
352 {
353 	struct br_ip ip;
354 	struct net_device *dev;
355 	struct net_bridge_port *p;
356 	int ret;
357 
358 	if (!netif_running(br->dev) || br->multicast_disabled)
359 		return -EINVAL;
360 
361 	dev = __dev_get_by_index(net, entry->ifindex);
362 	if (!dev)
363 		return -ENODEV;
364 
365 	p = br_port_get_rtnl(dev);
366 	if (!p || p->br != br || p->state == BR_STATE_DISABLED)
367 		return -EINVAL;
368 
369 	ip.proto = entry->addr.proto;
370 	if (ip.proto == htons(ETH_P_IP))
371 		ip.u.ip4 = entry->addr.u.ip4;
372 #if IS_ENABLED(CONFIG_IPV6)
373 	else
374 		ip.u.ip6 = entry->addr.u.ip6;
375 #endif
376 
377 	spin_lock_bh(&br->multicast_lock);
378 	ret = br_mdb_add_group(br, p, &ip, entry->state);
379 	spin_unlock_bh(&br->multicast_lock);
380 	return ret;
381 }
382 
383 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
384 {
385 	struct net *net = sock_net(skb->sk);
386 	struct br_mdb_entry *entry;
387 	struct net_device *dev;
388 	struct net_bridge *br;
389 	int err;
390 
391 	err = br_mdb_parse(skb, nlh, &dev, &entry);
392 	if (err < 0)
393 		return err;
394 
395 	br = netdev_priv(dev);
396 
397 	err = __br_mdb_add(net, br, entry);
398 	if (!err)
399 		__br_mdb_notify(dev, entry, RTM_NEWMDB);
400 	return err;
401 }
402 
403 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry)
404 {
405 	struct net_bridge_mdb_htable *mdb;
406 	struct net_bridge_mdb_entry *mp;
407 	struct net_bridge_port_group *p;
408 	struct net_bridge_port_group __rcu **pp;
409 	struct br_ip ip;
410 	int err = -EINVAL;
411 
412 	if (!netif_running(br->dev) || br->multicast_disabled)
413 		return -EINVAL;
414 
415 	if (timer_pending(&br->multicast_querier_timer))
416 		return -EBUSY;
417 
418 	ip.proto = entry->addr.proto;
419 	if (ip.proto == htons(ETH_P_IP))
420 		ip.u.ip4 = entry->addr.u.ip4;
421 #if IS_ENABLED(CONFIG_IPV6)
422 	else
423 		ip.u.ip6 = entry->addr.u.ip6;
424 #endif
425 
426 	spin_lock_bh(&br->multicast_lock);
427 	mdb = mlock_dereference(br->mdb, br);
428 
429 	mp = br_mdb_ip_get(mdb, &ip);
430 	if (!mp)
431 		goto unlock;
432 
433 	for (pp = &mp->ports;
434 	     (p = mlock_dereference(*pp, br)) != NULL;
435 	     pp = &p->next) {
436 		if (!p->port || p->port->dev->ifindex != entry->ifindex)
437 			continue;
438 
439 		if (p->port->state == BR_STATE_DISABLED)
440 			goto unlock;
441 
442 		rcu_assign_pointer(*pp, p->next);
443 		hlist_del_init(&p->mglist);
444 		del_timer(&p->timer);
445 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
446 		err = 0;
447 
448 		if (!mp->ports && !mp->mglist &&
449 		    netif_running(br->dev))
450 			mod_timer(&mp->timer, jiffies);
451 		break;
452 	}
453 
454 unlock:
455 	spin_unlock_bh(&br->multicast_lock);
456 	return err;
457 }
458 
459 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
460 {
461 	struct net_device *dev;
462 	struct br_mdb_entry *entry;
463 	struct net_bridge *br;
464 	int err;
465 
466 	err = br_mdb_parse(skb, nlh, &dev, &entry);
467 	if (err < 0)
468 		return err;
469 
470 	br = netdev_priv(dev);
471 
472 	err = __br_mdb_del(br, entry);
473 	if (!err)
474 		__br_mdb_notify(dev, entry, RTM_DELMDB);
475 	return err;
476 }
477 
478 void br_mdb_init(void)
479 {
480 	rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, NULL);
481 	rtnl_register(PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, NULL);
482 	rtnl_register(PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, NULL);
483 }
484 
485 void br_mdb_uninit(void)
486 {
487 	rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
488 	rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
489 	rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
490 }
491