xref: /openbmc/linux/net/bridge/br_mdb.c (revision 034f90b3)
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 	nlmsg_end(skb, nlh);
194 	return 0;
195 
196 end:
197 	nla_nest_end(skb, nest);
198 cancel:
199 	nlmsg_cancel(skb, nlh);
200 	return -EMSGSIZE;
201 }
202 
203 static inline size_t rtnl_mdb_nlmsg_size(void)
204 {
205 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
206 		+ nla_total_size(sizeof(struct br_mdb_entry));
207 }
208 
209 static void __br_mdb_notify(struct net_device *dev, struct br_mdb_entry *entry,
210 			    int type)
211 {
212 	struct net *net = dev_net(dev);
213 	struct sk_buff *skb;
214 	int err = -ENOBUFS;
215 
216 	skb = nlmsg_new(rtnl_mdb_nlmsg_size(), GFP_ATOMIC);
217 	if (!skb)
218 		goto errout;
219 
220 	err = nlmsg_populate_mdb_fill(skb, dev, entry, 0, 0, type, NTF_SELF);
221 	if (err < 0) {
222 		kfree_skb(skb);
223 		goto errout;
224 	}
225 
226 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
227 	return;
228 errout:
229 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
230 }
231 
232 void br_mdb_notify(struct net_device *dev, struct net_bridge_port *port,
233 		   struct br_ip *group, int type)
234 {
235 	struct br_mdb_entry entry;
236 
237 	memset(&entry, 0, sizeof(entry));
238 	entry.ifindex = port->dev->ifindex;
239 	entry.addr.proto = group->proto;
240 	entry.addr.u.ip4 = group->u.ip4;
241 #if IS_ENABLED(CONFIG_IPV6)
242 	entry.addr.u.ip6 = group->u.ip6;
243 #endif
244 	__br_mdb_notify(dev, &entry, type);
245 }
246 
247 static bool is_valid_mdb_entry(struct br_mdb_entry *entry)
248 {
249 	if (entry->ifindex == 0)
250 		return false;
251 
252 	if (entry->addr.proto == htons(ETH_P_IP)) {
253 		if (!ipv4_is_multicast(entry->addr.u.ip4))
254 			return false;
255 		if (ipv4_is_local_multicast(entry->addr.u.ip4))
256 			return false;
257 #if IS_ENABLED(CONFIG_IPV6)
258 	} else if (entry->addr.proto == htons(ETH_P_IPV6)) {
259 		if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6))
260 			return false;
261 #endif
262 	} else
263 		return false;
264 	if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY)
265 		return false;
266 
267 	return true;
268 }
269 
270 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
271 			struct net_device **pdev, struct br_mdb_entry **pentry)
272 {
273 	struct net *net = sock_net(skb->sk);
274 	struct br_mdb_entry *entry;
275 	struct br_port_msg *bpm;
276 	struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
277 	struct net_device *dev;
278 	int err;
279 
280 	err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY_MAX, NULL);
281 	if (err < 0)
282 		return err;
283 
284 	bpm = nlmsg_data(nlh);
285 	if (bpm->ifindex == 0) {
286 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid ifindex\n");
287 		return -EINVAL;
288 	}
289 
290 	dev = __dev_get_by_index(net, bpm->ifindex);
291 	if (dev == NULL) {
292 		pr_info("PF_BRIDGE: br_mdb_parse() with unknown ifindex\n");
293 		return -ENODEV;
294 	}
295 
296 	if (!(dev->priv_flags & IFF_EBRIDGE)) {
297 		pr_info("PF_BRIDGE: br_mdb_parse() with non-bridge\n");
298 		return -EOPNOTSUPP;
299 	}
300 
301 	*pdev = dev;
302 
303 	if (!tb[MDBA_SET_ENTRY] ||
304 	    nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
305 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid attr\n");
306 		return -EINVAL;
307 	}
308 
309 	entry = nla_data(tb[MDBA_SET_ENTRY]);
310 	if (!is_valid_mdb_entry(entry)) {
311 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid entry\n");
312 		return -EINVAL;
313 	}
314 
315 	*pentry = entry;
316 	return 0;
317 }
318 
319 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
320 			    struct br_ip *group, unsigned char state)
321 {
322 	struct net_bridge_mdb_entry *mp;
323 	struct net_bridge_port_group *p;
324 	struct net_bridge_port_group __rcu **pp;
325 	struct net_bridge_mdb_htable *mdb;
326 	int err;
327 
328 	mdb = mlock_dereference(br->mdb, br);
329 	mp = br_mdb_ip_get(mdb, group);
330 	if (!mp) {
331 		mp = br_multicast_new_group(br, port, group);
332 		err = PTR_ERR(mp);
333 		if (IS_ERR(mp))
334 			return err;
335 	}
336 
337 	for (pp = &mp->ports;
338 	     (p = mlock_dereference(*pp, br)) != NULL;
339 	     pp = &p->next) {
340 		if (p->port == port)
341 			return -EEXIST;
342 		if ((unsigned long)p->port < (unsigned long)port)
343 			break;
344 	}
345 
346 	p = br_multicast_new_port_group(port, group, *pp, state);
347 	if (unlikely(!p))
348 		return -ENOMEM;
349 	rcu_assign_pointer(*pp, p);
350 
351 	br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
352 	return 0;
353 }
354 
355 static int __br_mdb_add(struct net *net, struct net_bridge *br,
356 			struct br_mdb_entry *entry)
357 {
358 	struct br_ip ip;
359 	struct net_device *dev;
360 	struct net_bridge_port *p;
361 	int ret;
362 
363 	if (!netif_running(br->dev) || br->multicast_disabled)
364 		return -EINVAL;
365 
366 	dev = __dev_get_by_index(net, entry->ifindex);
367 	if (!dev)
368 		return -ENODEV;
369 
370 	p = br_port_get_rtnl(dev);
371 	if (!p || p->br != br || p->state == BR_STATE_DISABLED)
372 		return -EINVAL;
373 
374 	ip.proto = entry->addr.proto;
375 	if (ip.proto == htons(ETH_P_IP))
376 		ip.u.ip4 = entry->addr.u.ip4;
377 #if IS_ENABLED(CONFIG_IPV6)
378 	else
379 		ip.u.ip6 = entry->addr.u.ip6;
380 #endif
381 
382 	spin_lock_bh(&br->multicast_lock);
383 	ret = br_mdb_add_group(br, p, &ip, entry->state);
384 	spin_unlock_bh(&br->multicast_lock);
385 	return ret;
386 }
387 
388 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
389 {
390 	struct net *net = sock_net(skb->sk);
391 	struct br_mdb_entry *entry;
392 	struct net_device *dev;
393 	struct net_bridge *br;
394 	int err;
395 
396 	err = br_mdb_parse(skb, nlh, &dev, &entry);
397 	if (err < 0)
398 		return err;
399 
400 	br = netdev_priv(dev);
401 
402 	err = __br_mdb_add(net, br, entry);
403 	if (!err)
404 		__br_mdb_notify(dev, entry, RTM_NEWMDB);
405 	return err;
406 }
407 
408 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry)
409 {
410 	struct net_bridge_mdb_htable *mdb;
411 	struct net_bridge_mdb_entry *mp;
412 	struct net_bridge_port_group *p;
413 	struct net_bridge_port_group __rcu **pp;
414 	struct br_ip ip;
415 	int err = -EINVAL;
416 
417 	if (!netif_running(br->dev) || br->multicast_disabled)
418 		return -EINVAL;
419 
420 	ip.proto = entry->addr.proto;
421 	if (ip.proto == htons(ETH_P_IP)) {
422 		if (timer_pending(&br->ip4_other_query.timer))
423 			return -EBUSY;
424 
425 		ip.u.ip4 = entry->addr.u.ip4;
426 #if IS_ENABLED(CONFIG_IPV6)
427 	} else {
428 		if (timer_pending(&br->ip6_other_query.timer))
429 			return -EBUSY;
430 
431 		ip.u.ip6 = entry->addr.u.ip6;
432 #endif
433 	}
434 
435 	spin_lock_bh(&br->multicast_lock);
436 	mdb = mlock_dereference(br->mdb, br);
437 
438 	mp = br_mdb_ip_get(mdb, &ip);
439 	if (!mp)
440 		goto unlock;
441 
442 	for (pp = &mp->ports;
443 	     (p = mlock_dereference(*pp, br)) != NULL;
444 	     pp = &p->next) {
445 		if (!p->port || p->port->dev->ifindex != entry->ifindex)
446 			continue;
447 
448 		if (p->port->state == BR_STATE_DISABLED)
449 			goto unlock;
450 
451 		rcu_assign_pointer(*pp, p->next);
452 		hlist_del_init(&p->mglist);
453 		del_timer(&p->timer);
454 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
455 		err = 0;
456 
457 		if (!mp->ports && !mp->mglist &&
458 		    netif_running(br->dev))
459 			mod_timer(&mp->timer, jiffies);
460 		break;
461 	}
462 
463 unlock:
464 	spin_unlock_bh(&br->multicast_lock);
465 	return err;
466 }
467 
468 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
469 {
470 	struct net_device *dev;
471 	struct br_mdb_entry *entry;
472 	struct net_bridge *br;
473 	int err;
474 
475 	err = br_mdb_parse(skb, nlh, &dev, &entry);
476 	if (err < 0)
477 		return err;
478 
479 	br = netdev_priv(dev);
480 
481 	err = __br_mdb_del(br, entry);
482 	if (!err)
483 		__br_mdb_notify(dev, entry, RTM_DELMDB);
484 	return err;
485 }
486 
487 void br_mdb_init(void)
488 {
489 	rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, NULL);
490 	rtnl_register(PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, NULL);
491 	rtnl_register(PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, NULL);
492 }
493 
494 void br_mdb_uninit(void)
495 {
496 	rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
497 	rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
498 	rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
499 }
500