xref: /openbmc/linux/net/bridge/br_mdb.c (revision 3821a065)
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 					e.vid = p->addr.vid;
89 					if (p->addr.proto == htons(ETH_P_IP))
90 						e.addr.u.ip4 = p->addr.u.ip4;
91 #if IS_ENABLED(CONFIG_IPV6)
92 					if (p->addr.proto == htons(ETH_P_IPV6))
93 						e.addr.u.ip6 = p->addr.u.ip6;
94 #endif
95 					e.addr.proto = p->addr.proto;
96 					if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(e), &e)) {
97 						nla_nest_cancel(skb, nest2);
98 						err = -EMSGSIZE;
99 						goto out;
100 					}
101 				}
102 			}
103 			nla_nest_end(skb, nest2);
104 		skip:
105 			idx++;
106 		}
107 	}
108 
109 out:
110 	cb->args[1] = idx;
111 	nla_nest_end(skb, nest);
112 	return err;
113 }
114 
115 static int br_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
116 {
117 	struct net_device *dev;
118 	struct net *net = sock_net(skb->sk);
119 	struct nlmsghdr *nlh = NULL;
120 	int idx = 0, s_idx;
121 
122 	s_idx = cb->args[0];
123 
124 	rcu_read_lock();
125 
126 	/* In theory this could be wrapped to 0... */
127 	cb->seq = net->dev_base_seq + br_mdb_rehash_seq;
128 
129 	for_each_netdev_rcu(net, dev) {
130 		if (dev->priv_flags & IFF_EBRIDGE) {
131 			struct br_port_msg *bpm;
132 
133 			if (idx < s_idx)
134 				goto skip;
135 
136 			nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
137 					cb->nlh->nlmsg_seq, RTM_GETMDB,
138 					sizeof(*bpm), NLM_F_MULTI);
139 			if (nlh == NULL)
140 				break;
141 
142 			bpm = nlmsg_data(nlh);
143 			memset(bpm, 0, sizeof(*bpm));
144 			bpm->ifindex = dev->ifindex;
145 			if (br_mdb_fill_info(skb, cb, dev) < 0)
146 				goto out;
147 			if (br_rports_fill_info(skb, cb, dev) < 0)
148 				goto out;
149 
150 			cb->args[1] = 0;
151 			nlmsg_end(skb, nlh);
152 		skip:
153 			idx++;
154 		}
155 	}
156 
157 out:
158 	if (nlh)
159 		nlmsg_end(skb, nlh);
160 	rcu_read_unlock();
161 	cb->args[0] = idx;
162 	return skb->len;
163 }
164 
165 static int nlmsg_populate_mdb_fill(struct sk_buff *skb,
166 				   struct net_device *dev,
167 				   struct br_mdb_entry *entry, u32 pid,
168 				   u32 seq, int type, unsigned int flags)
169 {
170 	struct nlmsghdr *nlh;
171 	struct br_port_msg *bpm;
172 	struct nlattr *nest, *nest2;
173 
174 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*bpm), 0);
175 	if (!nlh)
176 		return -EMSGSIZE;
177 
178 	bpm = nlmsg_data(nlh);
179 	memset(bpm, 0, sizeof(*bpm));
180 	bpm->family  = AF_BRIDGE;
181 	bpm->ifindex = dev->ifindex;
182 	nest = nla_nest_start(skb, MDBA_MDB);
183 	if (nest == NULL)
184 		goto cancel;
185 	nest2 = nla_nest_start(skb, MDBA_MDB_ENTRY);
186 	if (nest2 == NULL)
187 		goto end;
188 
189 	if (nla_put(skb, MDBA_MDB_ENTRY_INFO, sizeof(*entry), entry))
190 		goto end;
191 
192 	nla_nest_end(skb, nest2);
193 	nla_nest_end(skb, nest);
194 	nlmsg_end(skb, nlh);
195 	return 0;
196 
197 end:
198 	nla_nest_end(skb, nest);
199 cancel:
200 	nlmsg_cancel(skb, nlh);
201 	return -EMSGSIZE;
202 }
203 
204 static inline size_t rtnl_mdb_nlmsg_size(void)
205 {
206 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
207 		+ nla_total_size(sizeof(struct br_mdb_entry));
208 }
209 
210 static void __br_mdb_notify(struct net_device *dev, struct br_mdb_entry *entry,
211 			    int type)
212 {
213 	struct net *net = dev_net(dev);
214 	struct sk_buff *skb;
215 	int err = -ENOBUFS;
216 
217 	skb = nlmsg_new(rtnl_mdb_nlmsg_size(), GFP_ATOMIC);
218 	if (!skb)
219 		goto errout;
220 
221 	err = nlmsg_populate_mdb_fill(skb, dev, entry, 0, 0, type, NTF_SELF);
222 	if (err < 0) {
223 		kfree_skb(skb);
224 		goto errout;
225 	}
226 
227 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
228 	return;
229 errout:
230 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
231 }
232 
233 void br_mdb_notify(struct net_device *dev, struct net_bridge_port *port,
234 		   struct br_ip *group, int type, u8 state)
235 {
236 	struct br_mdb_entry entry;
237 
238 	memset(&entry, 0, sizeof(entry));
239 	entry.ifindex = port->dev->ifindex;
240 	entry.addr.proto = group->proto;
241 	entry.addr.u.ip4 = group->u.ip4;
242 #if IS_ENABLED(CONFIG_IPV6)
243 	entry.addr.u.ip6 = group->u.ip6;
244 #endif
245 	entry.state = state;
246 	entry.vid = group->vid;
247 	__br_mdb_notify(dev, &entry, type);
248 }
249 
250 static int nlmsg_populate_rtr_fill(struct sk_buff *skb,
251 				   struct net_device *dev,
252 				   int ifindex, u32 pid,
253 				   u32 seq, int type, unsigned int flags)
254 {
255 	struct br_port_msg *bpm;
256 	struct nlmsghdr *nlh;
257 	struct nlattr *nest;
258 
259 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*bpm), NLM_F_MULTI);
260 	if (!nlh)
261 		return -EMSGSIZE;
262 
263 	bpm = nlmsg_data(nlh);
264 	memset(bpm, 0, sizeof(*bpm));
265 	bpm->family = AF_BRIDGE;
266 	bpm->ifindex = dev->ifindex;
267 	nest = nla_nest_start(skb, MDBA_ROUTER);
268 	if (!nest)
269 		goto cancel;
270 
271 	if (nla_put_u32(skb, MDBA_ROUTER_PORT, ifindex))
272 		goto end;
273 
274 	nla_nest_end(skb, nest);
275 	nlmsg_end(skb, nlh);
276 	return 0;
277 
278 end:
279 	nla_nest_end(skb, nest);
280 cancel:
281 	nlmsg_cancel(skb, nlh);
282 	return -EMSGSIZE;
283 }
284 
285 static inline size_t rtnl_rtr_nlmsg_size(void)
286 {
287 	return NLMSG_ALIGN(sizeof(struct br_port_msg))
288 		+ nla_total_size(sizeof(__u32));
289 }
290 
291 void br_rtr_notify(struct net_device *dev, struct net_bridge_port *port,
292 		   int type)
293 {
294 	struct net *net = dev_net(dev);
295 	struct sk_buff *skb;
296 	int err = -ENOBUFS;
297 	int ifindex;
298 
299 	ifindex = port ? port->dev->ifindex : 0;
300 	skb = nlmsg_new(rtnl_rtr_nlmsg_size(), GFP_ATOMIC);
301 	if (!skb)
302 		goto errout;
303 
304 	err = nlmsg_populate_rtr_fill(skb, dev, ifindex, 0, 0, type, NTF_SELF);
305 	if (err < 0) {
306 		kfree_skb(skb);
307 		goto errout;
308 	}
309 
310 	rtnl_notify(skb, net, 0, RTNLGRP_MDB, NULL, GFP_ATOMIC);
311 	return;
312 
313 errout:
314 	rtnl_set_sk_err(net, RTNLGRP_MDB, err);
315 }
316 
317 static bool is_valid_mdb_entry(struct br_mdb_entry *entry)
318 {
319 	if (entry->ifindex == 0)
320 		return false;
321 
322 	if (entry->addr.proto == htons(ETH_P_IP)) {
323 		if (!ipv4_is_multicast(entry->addr.u.ip4))
324 			return false;
325 		if (ipv4_is_local_multicast(entry->addr.u.ip4))
326 			return false;
327 #if IS_ENABLED(CONFIG_IPV6)
328 	} else if (entry->addr.proto == htons(ETH_P_IPV6)) {
329 		if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6))
330 			return false;
331 #endif
332 	} else
333 		return false;
334 	if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY)
335 		return false;
336 	if (entry->vid >= VLAN_VID_MASK)
337 		return false;
338 
339 	return true;
340 }
341 
342 static int br_mdb_parse(struct sk_buff *skb, struct nlmsghdr *nlh,
343 			struct net_device **pdev, struct br_mdb_entry **pentry)
344 {
345 	struct net *net = sock_net(skb->sk);
346 	struct br_mdb_entry *entry;
347 	struct br_port_msg *bpm;
348 	struct nlattr *tb[MDBA_SET_ENTRY_MAX+1];
349 	struct net_device *dev;
350 	int err;
351 
352 	err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY_MAX, NULL);
353 	if (err < 0)
354 		return err;
355 
356 	bpm = nlmsg_data(nlh);
357 	if (bpm->ifindex == 0) {
358 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid ifindex\n");
359 		return -EINVAL;
360 	}
361 
362 	dev = __dev_get_by_index(net, bpm->ifindex);
363 	if (dev == NULL) {
364 		pr_info("PF_BRIDGE: br_mdb_parse() with unknown ifindex\n");
365 		return -ENODEV;
366 	}
367 
368 	if (!(dev->priv_flags & IFF_EBRIDGE)) {
369 		pr_info("PF_BRIDGE: br_mdb_parse() with non-bridge\n");
370 		return -EOPNOTSUPP;
371 	}
372 
373 	*pdev = dev;
374 
375 	if (!tb[MDBA_SET_ENTRY] ||
376 	    nla_len(tb[MDBA_SET_ENTRY]) != sizeof(struct br_mdb_entry)) {
377 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid attr\n");
378 		return -EINVAL;
379 	}
380 
381 	entry = nla_data(tb[MDBA_SET_ENTRY]);
382 	if (!is_valid_mdb_entry(entry)) {
383 		pr_info("PF_BRIDGE: br_mdb_parse() with invalid entry\n");
384 		return -EINVAL;
385 	}
386 
387 	*pentry = entry;
388 	return 0;
389 }
390 
391 static int br_mdb_add_group(struct net_bridge *br, struct net_bridge_port *port,
392 			    struct br_ip *group, unsigned char state)
393 {
394 	struct net_bridge_mdb_entry *mp;
395 	struct net_bridge_port_group *p;
396 	struct net_bridge_port_group __rcu **pp;
397 	struct net_bridge_mdb_htable *mdb;
398 	unsigned long now = jiffies;
399 	int err;
400 
401 	mdb = mlock_dereference(br->mdb, br);
402 	mp = br_mdb_ip_get(mdb, group);
403 	if (!mp) {
404 		mp = br_multicast_new_group(br, port, group);
405 		err = PTR_ERR(mp);
406 		if (IS_ERR(mp))
407 			return err;
408 	}
409 
410 	for (pp = &mp->ports;
411 	     (p = mlock_dereference(*pp, br)) != NULL;
412 	     pp = &p->next) {
413 		if (p->port == port)
414 			return -EEXIST;
415 		if ((unsigned long)p->port < (unsigned long)port)
416 			break;
417 	}
418 
419 	p = br_multicast_new_port_group(port, group, *pp, state);
420 	if (unlikely(!p))
421 		return -ENOMEM;
422 	rcu_assign_pointer(*pp, p);
423 	if (state == MDB_TEMPORARY)
424 		mod_timer(&p->timer, now + br->multicast_membership_interval);
425 
426 	return 0;
427 }
428 
429 static int __br_mdb_add(struct net *net, struct net_bridge *br,
430 			struct br_mdb_entry *entry)
431 {
432 	struct br_ip ip;
433 	struct net_device *dev;
434 	struct net_bridge_port *p;
435 	int ret;
436 
437 	if (!netif_running(br->dev) || br->multicast_disabled)
438 		return -EINVAL;
439 
440 	dev = __dev_get_by_index(net, entry->ifindex);
441 	if (!dev)
442 		return -ENODEV;
443 
444 	p = br_port_get_rtnl(dev);
445 	if (!p || p->br != br || p->state == BR_STATE_DISABLED)
446 		return -EINVAL;
447 
448 	memset(&ip, 0, sizeof(ip));
449 	ip.vid = entry->vid;
450 	ip.proto = entry->addr.proto;
451 	if (ip.proto == htons(ETH_P_IP))
452 		ip.u.ip4 = entry->addr.u.ip4;
453 #if IS_ENABLED(CONFIG_IPV6)
454 	else
455 		ip.u.ip6 = entry->addr.u.ip6;
456 #endif
457 
458 	spin_lock_bh(&br->multicast_lock);
459 	ret = br_mdb_add_group(br, p, &ip, entry->state);
460 	spin_unlock_bh(&br->multicast_lock);
461 	return ret;
462 }
463 
464 static int br_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
465 {
466 	struct net *net = sock_net(skb->sk);
467 	unsigned short vid = VLAN_N_VID;
468 	struct net_device *dev, *pdev;
469 	struct br_mdb_entry *entry;
470 	struct net_bridge_port *p;
471 	struct net_port_vlans *pv;
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 	/* If vlan filtering is enabled and VLAN is not specified
482 	 * install mdb entry on all vlans configured on the port.
483 	 */
484 	pdev = __dev_get_by_index(net, entry->ifindex);
485 	if (!pdev)
486 		return -ENODEV;
487 
488 	p = br_port_get_rtnl(pdev);
489 	if (!p || p->br != br || p->state == BR_STATE_DISABLED)
490 		return -EINVAL;
491 
492 	pv = nbp_get_vlan_info(p);
493 	if (br_vlan_enabled(br) && pv && entry->vid == 0) {
494 		for_each_set_bit(vid, pv->vlan_bitmap, VLAN_N_VID) {
495 			entry->vid = vid;
496 			err = __br_mdb_add(net, br, entry);
497 			if (err)
498 				break;
499 			__br_mdb_notify(dev, entry, RTM_NEWMDB);
500 		}
501 	} else {
502 		err = __br_mdb_add(net, br, entry);
503 		if (!err)
504 			__br_mdb_notify(dev, entry, RTM_NEWMDB);
505 	}
506 
507 	return err;
508 }
509 
510 static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry)
511 {
512 	struct net_bridge_mdb_htable *mdb;
513 	struct net_bridge_mdb_entry *mp;
514 	struct net_bridge_port_group *p;
515 	struct net_bridge_port_group __rcu **pp;
516 	struct br_ip ip;
517 	int err = -EINVAL;
518 
519 	if (!netif_running(br->dev) || br->multicast_disabled)
520 		return -EINVAL;
521 
522 	memset(&ip, 0, sizeof(ip));
523 	ip.vid = entry->vid;
524 	ip.proto = entry->addr.proto;
525 	if (ip.proto == htons(ETH_P_IP))
526 		ip.u.ip4 = entry->addr.u.ip4;
527 #if IS_ENABLED(CONFIG_IPV6)
528 	else
529 		ip.u.ip6 = entry->addr.u.ip6;
530 #endif
531 
532 	spin_lock_bh(&br->multicast_lock);
533 	mdb = mlock_dereference(br->mdb, br);
534 
535 	mp = br_mdb_ip_get(mdb, &ip);
536 	if (!mp)
537 		goto unlock;
538 
539 	for (pp = &mp->ports;
540 	     (p = mlock_dereference(*pp, br)) != NULL;
541 	     pp = &p->next) {
542 		if (!p->port || p->port->dev->ifindex != entry->ifindex)
543 			continue;
544 
545 		if (p->port->state == BR_STATE_DISABLED)
546 			goto unlock;
547 
548 		entry->state = p->state;
549 		rcu_assign_pointer(*pp, p->next);
550 		hlist_del_init(&p->mglist);
551 		del_timer(&p->timer);
552 		call_rcu_bh(&p->rcu, br_multicast_free_pg);
553 		err = 0;
554 
555 		if (!mp->ports && !mp->mglist &&
556 		    netif_running(br->dev))
557 			mod_timer(&mp->timer, jiffies);
558 		break;
559 	}
560 
561 unlock:
562 	spin_unlock_bh(&br->multicast_lock);
563 	return err;
564 }
565 
566 static int br_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
567 {
568 	struct net *net = sock_net(skb->sk);
569 	unsigned short vid = VLAN_N_VID;
570 	struct net_device *dev, *pdev;
571 	struct br_mdb_entry *entry;
572 	struct net_bridge_port *p;
573 	struct net_port_vlans *pv;
574 	struct net_bridge *br;
575 	int err;
576 
577 	err = br_mdb_parse(skb, nlh, &dev, &entry);
578 	if (err < 0)
579 		return err;
580 
581 	br = netdev_priv(dev);
582 
583 	/* If vlan filtering is enabled and VLAN is not specified
584 	 * delete mdb entry on all vlans configured on the port.
585 	 */
586 	pdev = __dev_get_by_index(net, entry->ifindex);
587 	if (!pdev)
588 		return -ENODEV;
589 
590 	p = br_port_get_rtnl(pdev);
591 	if (!p || p->br != br || p->state == BR_STATE_DISABLED)
592 		return -EINVAL;
593 
594 	pv = nbp_get_vlan_info(p);
595 	if (br_vlan_enabled(br) && pv && entry->vid == 0) {
596 		for_each_set_bit(vid, pv->vlan_bitmap, VLAN_N_VID) {
597 			entry->vid = vid;
598 			err = __br_mdb_del(br, entry);
599 			if (!err)
600 				__br_mdb_notify(dev, entry, RTM_DELMDB);
601 		}
602 	} else {
603 		err = __br_mdb_del(br, entry);
604 		if (!err)
605 			__br_mdb_notify(dev, entry, RTM_DELMDB);
606 	}
607 
608 	return err;
609 }
610 
611 void br_mdb_init(void)
612 {
613 	rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, br_mdb_dump, NULL);
614 	rtnl_register(PF_BRIDGE, RTM_NEWMDB, br_mdb_add, NULL, NULL);
615 	rtnl_register(PF_BRIDGE, RTM_DELMDB, br_mdb_del, NULL, NULL);
616 }
617 
618 void br_mdb_uninit(void)
619 {
620 	rtnl_unregister(PF_BRIDGE, RTM_GETMDB);
621 	rtnl_unregister(PF_BRIDGE, RTM_NEWMDB);
622 	rtnl_unregister(PF_BRIDGE, RTM_DELMDB);
623 }
624