xref: /openbmc/linux/net/bridge/br_mdb.c (revision 275876e2)
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 #include <net/addrconf.h>
13 #endif
14 
15 #include "br_private.h"
16 
17 static int br_rports_fill_info(struct sk_buff *skb, struct netlink_callback *cb,
18 			       struct net_device *dev)
19 {
20 	struct net_bridge *br = netdev_priv(dev);
21 	struct net_bridge_port *p;
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, &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 net_bridge_mdb_entry *mp;
65 		struct net_bridge_port_group *p;
66 		struct net_bridge_port_group __rcu **pp;
67 		struct net_bridge_port *port;
68 
69 		hlist_for_each_entry_rcu(mp, &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 					memset(&e, 0, sizeof(e));
86 					e.ifindex = port->dev->ifindex;
87 					e.state = p->state;
88 					if (p->addr.proto == htons(ETH_P_IP))
89 						e.addr.u.ip4 = p->addr.u.ip4;
90 #if IS_ENABLED(CONFIG_IPV6)
91 					if (p->addr.proto == htons(ETH_P_IPV6))
92 						e.addr.u.ip6 = p->addr.u.ip6;
93 #endif
94 					e.addr.proto = p->addr.proto;
95 					if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(e), &e)) {
96 						nla_nest_cancel(skb, nest2);
97 						err = -EMSGSIZE;
98 						goto out;
99 					}
100 				}
101 			}
102 			nla_nest_end(skb, nest2);
103 		skip:
104 			idx++;
105 		}
106 	}
107 
108 out:
109 	cb->args[1] = idx;
110 	nla_nest_end(skb, nest);
111 	return err;
112 }
113 
114 static int br_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
115 {
116 	struct net_device *dev;
117 	struct net *net = sock_net(skb->sk);
118 	struct nlmsghdr *nlh = NULL;
119 	int idx = 0, s_idx;
120 
121 	s_idx = cb->args[0];
122 
123 	rcu_read_lock();
124 
125 	/* In theory this could be wrapped to 0... */
126 	cb->seq = net->dev_base_seq + br_mdb_rehash_seq;
127 
128 	for_each_netdev_rcu(net, dev) {
129 		if (dev->priv_flags & IFF_EBRIDGE) {
130 			struct br_port_msg *bpm;
131 
132 			if (idx < s_idx)
133 				goto skip;
134 
135 			nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
136 					cb->nlh->nlmsg_seq, RTM_GETMDB,
137 					sizeof(*bpm), NLM_F_MULTI);
138 			if (nlh == NULL)
139 				break;
140 
141 			bpm = nlmsg_data(nlh);
142 			memset(bpm, 0, sizeof(*bpm));
143 			bpm->ifindex = dev->ifindex;
144 			if (br_mdb_fill_info(skb, cb, dev) < 0)
145 				goto out;
146 			if (br_rports_fill_info(skb, cb, dev) < 0)
147 				goto out;
148 
149 			cb->args[1] = 0;
150 			nlmsg_end(skb, nlh);
151 		skip:
152 			idx++;
153 		}
154 	}
155 
156 out:
157 	if (nlh)
158 		nlmsg_end(skb, nlh);
159 	rcu_read_unlock();
160 	cb->args[0] = idx;
161 	return skb->len;
162 }
163 
164 static int nlmsg_populate_mdb_fill(struct sk_buff *skb,
165 				   struct net_device *dev,
166 				   struct br_mdb_entry *entry, u32 pid,
167 				   u32 seq, int type, unsigned int flags)
168 {
169 	struct nlmsghdr *nlh;
170 	struct br_port_msg *bpm;
171 	struct nlattr *nest, *nest2;
172 
173 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*bpm), NLM_F_MULTI);
174 	if (!nlh)
175 		return -EMSGSIZE;
176 
177 	bpm = nlmsg_data(nlh);
178 	memset(bpm, 0, sizeof(*bpm));
179 	bpm->family  = AF_BRIDGE;
180 	bpm->ifindex = dev->ifindex;
181 	nest = nla_nest_start(skb, MDBA_MDB);
182 	if (nest == NULL)
183 		goto cancel;
184 	nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
185 	if (nest2 == NULL)
186 		goto end;
187 
188 	if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(*entry), entry))
189 		goto end;
190 
191 	nla_nest_end(skb, nest2);
192 	nla_nest_end(skb, nest);
193 	return nlmsg_end(skb, nlh);
194 
195 end:
196 	nla_nest_end(skb, nest);
197 cancel:
198 	nlmsg_cancel(skb, nlh);
199 	return -EMSGSIZE;
200 }
201 
202 static inline size_t rtnl_mdb_nlmsg_size(void)
203 {
204 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
205 		+ nla_total_size(sizeof(struct br_mdb_entry));
206 }
207 
208 static void __br_mdb_notify(struct net_device *dev, struct br_mdb_entry *entry,
209 			    int type)
210 {
211 	struct net *net = dev_net(dev);
212 	struct sk_buff *skb;
213 	int err = -ENOBUFS;
214 
215 	skb = nlmsg_new(rtnl_mdb_nlmsg_size(), GFP_ATOMIC);
216 	if (!skb)
217 		goto errout;
218 
219 	err = nlmsg_populate_mdb_fill(skb, dev, entry, 0, 0, type, NTF_SELF);
220 	if (err < 0) {
221 		kfree_skb(skb);
222 		goto errout;
223 	}
224 
225 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
226 	return;
227 errout:
228 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
229 }
230 
231 void br_mdb_notify(struct net_device *dev, struct net_bridge_port *port,
232 		   struct br_ip *group, int type)
233 {
234 	struct br_mdb_entry entry;
235 
236 	memset(&entry, 0, sizeof(entry));
237 	entry.ifindex = port->dev->ifindex;
238 	entry.addr.proto = group->proto;
239 	entry.addr.u.ip4 = group->u.ip4;
240 #if IS_ENABLED(CONFIG_IPV6)
241 	entry.addr.u.ip6 = group->u.ip6;
242 #endif
243 	__br_mdb_notify(dev, &entry, type);
244 }
245 
246 static bool is_valid_mdb_entry(struct br_mdb_entry *entry)
247 {
248 	if (entry->ifindex == 0)
249 		return false;
250 
251 	if (entry->addr.proto == htons(ETH_P_IP)) {
252 		if (!ipv4_is_multicast(entry->addr.u.ip4))
253 			return false;
254 		if (ipv4_is_local_multicast(entry->addr.u.ip4))
255 			return false;
256 #if IS_ENABLED(CONFIG_IPV6)
257 	} else if (entry->addr.proto == htons(ETH_P_IPV6)) {
258 		if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6))
259 			return false;
260 #endif
261 	} else
262 		return false;
263 	if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY)
264 		return false;
265 
266 	return true;
267 }
268 
269 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
270 			struct net_device **pdev, struct br_mdb_entry **pentry)
271 {
272 	struct net *net = sock_net(skb->sk);
273 	struct br_mdb_entry *entry;
274 	struct br_port_msg *bpm;
275 	struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
276 	struct net_device *dev;
277 	int err;
278 
279 	err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY, NULL);
280 	if (err < 0)
281 		return err;
282 
283 	bpm = nlmsg_data(nlh);
284 	if (bpm->ifindex == 0) {
285 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid ifindex\n");
286 		return -EINVAL;
287 	}
288 
289 	dev = __dev_get_by_index(net, bpm->ifindex);
290 	if (dev == NULL) {
291 		pr_info("PF_BRIDGE: br_mdb_parse() with unknown ifindex\n");
292 		return -ENODEV;
293 	}
294 
295 	if (!(dev->priv_flags & IFF_EBRIDGE)) {
296 		pr_info("PF_BRIDGE: br_mdb_parse() with non-bridge\n");
297 		return -EOPNOTSUPP;
298 	}
299 
300 	*pdev = dev;
301 
302 	if (!tb[MDBA_SET_ENTRY] ||
303 	    nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
304 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid attr\n");
305 		return -EINVAL;
306 	}
307 
308 	entry = nla_data(tb[MDBA_SET_ENTRY]);
309 	if (!is_valid_mdb_entry(entry)) {
310 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid entry\n");
311 		return -EINVAL;
312 	}
313 
314 	*pentry = entry;
315 	return 0;
316 }
317 
318 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
319 			    struct br_ip *group, unsigned char state)
320 {
321 	struct net_bridge_mdb_entry *mp;
322 	struct net_bridge_port_group *p;
323 	struct net_bridge_port_group __rcu **pp;
324 	struct net_bridge_mdb_htable *mdb;
325 	int err;
326 
327 	mdb = mlock_dereference(br->mdb, br);
328 	mp = br_mdb_ip_get(mdb, group);
329 	if (!mp) {
330 		mp = br_multicast_new_group(br, port, group);
331 		err = PTR_ERR(mp);
332 		if (IS_ERR(mp))
333 			return err;
334 	}
335 
336 	for (pp = &mp->ports;
337 	     (p = mlock_dereference(*pp, br)) != NULL;
338 	     pp = &p->next) {
339 		if (p->port == port)
340 			return -EEXIST;
341 		if ((unsigned long)p->port < (unsigned long)port)
342 			break;
343 	}
344 
345 	p = br_multicast_new_port_group(port, group, *pp, state);
346 	if (unlikely(!p))
347 		return -ENOMEM;
348 	rcu_assign_pointer(*pp, p);
349 
350 	br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
351 	return 0;
352 }
353 
354 static int __br_mdb_add(struct net *net, struct net_bridge *br,
355 			struct br_mdb_entry *entry)
356 {
357 	struct br_ip ip;
358 	struct net_device *dev;
359 	struct net_bridge_port *p;
360 	int ret;
361 
362 	if (!netif_running(br->dev) || br->multicast_disabled)
363 		return -EINVAL;
364 
365 	dev = __dev_get_by_index(net, entry->ifindex);
366 	if (!dev)
367 		return -ENODEV;
368 
369 	p = br_port_get_rtnl(dev);
370 	if (!p || p->br != br || p->state == BR_STATE_DISABLED)
371 		return -EINVAL;
372 
373 	ip.proto = entry->addr.proto;
374 	if (ip.proto == htons(ETH_P_IP))
375 		ip.u.ip4 = entry->addr.u.ip4;
376 #if IS_ENABLED(CONFIG_IPV6)
377 	else
378 		ip.u.ip6 = entry->addr.u.ip6;
379 #endif
380 
381 	spin_lock_bh(&br->multicast_lock);
382 	ret = br_mdb_add_group(br, p, &ip, entry->state);
383 	spin_unlock_bh(&br->multicast_lock);
384 	return ret;
385 }
386 
387 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
388 {
389 	struct net *net = sock_net(skb->sk);
390 	struct br_mdb_entry *entry;
391 	struct net_device *dev;
392 	struct net_bridge *br;
393 	int err;
394 
395 	err = br_mdb_parse(skb, nlh, &dev, &entry);
396 	if (err < 0)
397 		return err;
398 
399 	br = netdev_priv(dev);
400 
401 	err = __br_mdb_add(net, br, entry);
402 	if (!err)
403 		__br_mdb_notify(dev, entry, RTM_NEWMDB);
404 	return err;
405 }
406 
407 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry)
408 {
409 	struct net_bridge_mdb_htable *mdb;
410 	struct net_bridge_mdb_entry *mp;
411 	struct net_bridge_port_group *p;
412 	struct net_bridge_port_group __rcu **pp;
413 	struct br_ip ip;
414 	int err = -EINVAL;
415 
416 	if (!netif_running(br->dev) || br->multicast_disabled)
417 		return -EINVAL;
418 
419 	ip.proto = entry->addr.proto;
420 	if (ip.proto == htons(ETH_P_IP)) {
421 		if (timer_pending(&br->ip4_other_query.timer))
422 			return -EBUSY;
423 
424 		ip.u.ip4 = entry->addr.u.ip4;
425 #if IS_ENABLED(CONFIG_IPV6)
426 	} else {
427 		if (timer_pending(&br->ip6_other_query.timer))
428 			return -EBUSY;
429 
430 		ip.u.ip6 = entry->addr.u.ip6;
431 #endif
432 	}
433 
434 	spin_lock_bh(&br->multicast_lock);
435 	mdb = mlock_dereference(br->mdb, br);
436 
437 	mp = br_mdb_ip_get(mdb, &ip);
438 	if (!mp)
439 		goto unlock;
440 
441 	for (pp = &mp->ports;
442 	     (p = mlock_dereference(*pp, br)) != NULL;
443 	     pp = &p->next) {
444 		if (!p->port || p->port->dev->ifindex != entry->ifindex)
445 			continue;
446 
447 		if (p->port->state == BR_STATE_DISABLED)
448 			goto unlock;
449 
450 		rcu_assign_pointer(*pp, p->next);
451 		hlist_del_init(&p->mglist);
452 		del_timer(&p->timer);
453 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
454 		err = 0;
455 
456 		if (!mp->ports && !mp->mglist &&
457 		    netif_running(br->dev))
458 			mod_timer(&mp->timer, jiffies);
459 		break;
460 	}
461 
462 unlock:
463 	spin_unlock_bh(&br->multicast_lock);
464 	return err;
465 }
466 
467 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
468 {
469 	struct net_device *dev;
470 	struct br_mdb_entry *entry;
471 	struct net_bridge *br;
472 	int err;
473 
474 	err = br_mdb_parse(skb, nlh, &dev, &entry);
475 	if (err < 0)
476 		return err;
477 
478 	br = netdev_priv(dev);
479 
480 	err = __br_mdb_del(br, entry);
481 	if (!err)
482 		__br_mdb_notify(dev, entry, RTM_DELMDB);
483 	return err;
484 }
485 
486 void br_mdb_init(void)
487 {
488 	rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, NULL);
489 	rtnl_register(PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, NULL);
490 	rtnl_register(PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, NULL);
491 }
492 
493 void br_mdb_uninit(void)
494 {
495 	rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
496 	rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
497 	rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
498 }
499