• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  *
15  *	Fixes:
16  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17  */
18 
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
29 #include <linux/mm.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/pci.h>
40 #include <linux/etherdevice.h>
41 
42 #include <asm/uaccess.h>
43 
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <net/arp.h>
49 #include <net/route.h>
50 #include <net/udp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 
57 struct rtnl_link {
58 	rtnl_doit_func		doit;
59 	rtnl_dumpit_func	dumpit;
60 	rtnl_calcit_func 	calcit;
61 };
62 
63 static DEFINE_MUTEX(rtnl_mutex);
64 
rtnl_lock(void)65 void rtnl_lock(void)
66 {
67 	mutex_lock(&rtnl_mutex);
68 }
69 EXPORT_SYMBOL(rtnl_lock);
70 
__rtnl_unlock(void)71 void __rtnl_unlock(void)
72 {
73 	mutex_unlock(&rtnl_mutex);
74 }
75 
rtnl_unlock(void)76 void rtnl_unlock(void)
77 {
78 	/* This fellow will unlock it for us. */
79 	netdev_run_todo();
80 }
81 EXPORT_SYMBOL(rtnl_unlock);
82 
rtnl_trylock(void)83 int rtnl_trylock(void)
84 {
85 	return mutex_trylock(&rtnl_mutex);
86 }
87 EXPORT_SYMBOL(rtnl_trylock);
88 
rtnl_is_locked(void)89 int rtnl_is_locked(void)
90 {
91 	return mutex_is_locked(&rtnl_mutex);
92 }
93 EXPORT_SYMBOL(rtnl_is_locked);
94 
95 #ifdef CONFIG_PROVE_LOCKING
lockdep_rtnl_is_held(void)96 int lockdep_rtnl_is_held(void)
97 {
98 	return lockdep_is_held(&rtnl_mutex);
99 }
100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
102 
103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
104 
rtm_msgindex(int msgtype)105 static inline int rtm_msgindex(int msgtype)
106 {
107 	int msgindex = msgtype - RTM_BASE;
108 
109 	/*
110 	 * msgindex < 0 implies someone tried to register a netlink
111 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 	 * the message type has not been added to linux/rtnetlink.h
113 	 */
114 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
115 
116 	return msgindex;
117 }
118 
rtnl_get_doit(int protocol,int msgindex)119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
120 {
121 	struct rtnl_link *tab;
122 
123 	if (protocol <= RTNL_FAMILY_MAX)
124 		tab = rtnl_msg_handlers[protocol];
125 	else
126 		tab = NULL;
127 
128 	if (tab == NULL || tab[msgindex].doit == NULL)
129 		tab = rtnl_msg_handlers[PF_UNSPEC];
130 
131 	return tab[msgindex].doit;
132 }
133 
rtnl_get_dumpit(int protocol,int msgindex)134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
135 {
136 	struct rtnl_link *tab;
137 
138 	if (protocol <= RTNL_FAMILY_MAX)
139 		tab = rtnl_msg_handlers[protocol];
140 	else
141 		tab = NULL;
142 
143 	if (tab == NULL || tab[msgindex].dumpit == NULL)
144 		tab = rtnl_msg_handlers[PF_UNSPEC];
145 
146 	return tab[msgindex].dumpit;
147 }
148 
rtnl_get_calcit(int protocol,int msgindex)149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
150 {
151 	struct rtnl_link *tab;
152 
153 	if (protocol <= RTNL_FAMILY_MAX)
154 		tab = rtnl_msg_handlers[protocol];
155 	else
156 		tab = NULL;
157 
158 	if (tab == NULL || tab[msgindex].calcit == NULL)
159 		tab = rtnl_msg_handlers[PF_UNSPEC];
160 
161 	return tab[msgindex].calcit;
162 }
163 
164 /**
165  * __rtnl_register - Register a rtnetlink message type
166  * @protocol: Protocol family or PF_UNSPEC
167  * @msgtype: rtnetlink message type
168  * @doit: Function pointer called for each request message
169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170  * @calcit: Function pointer to calc size of dump message
171  *
172  * Registers the specified function pointers (at least one of them has
173  * to be non-NULL) to be called whenever a request message for the
174  * specified protocol family and message type is received.
175  *
176  * The special protocol family PF_UNSPEC may be used to define fallback
177  * function pointers for the case when no entry for the specific protocol
178  * family exists.
179  *
180  * Returns 0 on success or a negative error code.
181  */
__rtnl_register(int protocol,int msgtype,rtnl_doit_func doit,rtnl_dumpit_func dumpit,rtnl_calcit_func calcit)182 int __rtnl_register(int protocol, int msgtype,
183 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
184 		    rtnl_calcit_func calcit)
185 {
186 	struct rtnl_link *tab;
187 	int msgindex;
188 
189 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
190 	msgindex = rtm_msgindex(msgtype);
191 
192 	tab = rtnl_msg_handlers[protocol];
193 	if (tab == NULL) {
194 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
195 		if (tab == NULL)
196 			return -ENOBUFS;
197 
198 		rtnl_msg_handlers[protocol] = tab;
199 	}
200 
201 	if (doit)
202 		tab[msgindex].doit = doit;
203 
204 	if (dumpit)
205 		tab[msgindex].dumpit = dumpit;
206 
207 	if (calcit)
208 		tab[msgindex].calcit = calcit;
209 
210 	return 0;
211 }
212 EXPORT_SYMBOL_GPL(__rtnl_register);
213 
214 /**
215  * rtnl_register - Register a rtnetlink message type
216  *
217  * Identical to __rtnl_register() but panics on failure. This is useful
218  * as failure of this function is very unlikely, it can only happen due
219  * to lack of memory when allocating the chain to store all message
220  * handlers for a protocol. Meant for use in init functions where lack
221  * of memory implies no sense in continuing.
222  */
rtnl_register(int protocol,int msgtype,rtnl_doit_func doit,rtnl_dumpit_func dumpit,rtnl_calcit_func calcit)223 void rtnl_register(int protocol, int msgtype,
224 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
225 		   rtnl_calcit_func calcit)
226 {
227 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
228 		panic("Unable to register rtnetlink message handler, "
229 		      "protocol = %d, message type = %d\n",
230 		      protocol, msgtype);
231 }
232 EXPORT_SYMBOL_GPL(rtnl_register);
233 
234 /**
235  * rtnl_unregister - Unregister a rtnetlink message type
236  * @protocol: Protocol family or PF_UNSPEC
237  * @msgtype: rtnetlink message type
238  *
239  * Returns 0 on success or a negative error code.
240  */
rtnl_unregister(int protocol,int msgtype)241 int rtnl_unregister(int protocol, int msgtype)
242 {
243 	int msgindex;
244 
245 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
246 	msgindex = rtm_msgindex(msgtype);
247 
248 	if (rtnl_msg_handlers[protocol] == NULL)
249 		return -ENOENT;
250 
251 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
252 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
253 
254 	return 0;
255 }
256 EXPORT_SYMBOL_GPL(rtnl_unregister);
257 
258 /**
259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260  * @protocol : Protocol family or PF_UNSPEC
261  *
262  * Identical to calling rtnl_unregster() for all registered message types
263  * of a certain protocol family.
264  */
rtnl_unregister_all(int protocol)265 void rtnl_unregister_all(int protocol)
266 {
267 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
268 
269 	kfree(rtnl_msg_handlers[protocol]);
270 	rtnl_msg_handlers[protocol] = NULL;
271 }
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
273 
274 static LIST_HEAD(link_ops);
275 
rtnl_link_ops_get(const char * kind)276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
277 {
278 	const struct rtnl_link_ops *ops;
279 
280 	list_for_each_entry(ops, &link_ops, list) {
281 		if (!strcmp(ops->kind, kind))
282 			return ops;
283 	}
284 	return NULL;
285 }
286 
287 /**
288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289  * @ops: struct rtnl_link_ops * to register
290  *
291  * The caller must hold the rtnl_mutex. This function should be used
292  * by drivers that create devices during module initialization. It
293  * must be called before registering the devices.
294  *
295  * Returns 0 on success or a negative error code.
296  */
__rtnl_link_register(struct rtnl_link_ops * ops)297 int __rtnl_link_register(struct rtnl_link_ops *ops)
298 {
299 	if (rtnl_link_ops_get(ops->kind))
300 		return -EEXIST;
301 
302 	/* The check for setup is here because if ops
303 	 * does not have that filled up, it is not possible
304 	 * to use the ops for creating device. So do not
305 	 * fill up dellink as well. That disables rtnl_dellink.
306 	 */
307 	if (ops->setup && !ops->dellink)
308 		ops->dellink = unregister_netdevice_queue;
309 
310 	list_add_tail(&ops->list, &link_ops);
311 	return 0;
312 }
313 EXPORT_SYMBOL_GPL(__rtnl_link_register);
314 
315 /**
316  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
317  * @ops: struct rtnl_link_ops * to register
318  *
319  * Returns 0 on success or a negative error code.
320  */
rtnl_link_register(struct rtnl_link_ops * ops)321 int rtnl_link_register(struct rtnl_link_ops *ops)
322 {
323 	int err;
324 
325 	rtnl_lock();
326 	err = __rtnl_link_register(ops);
327 	rtnl_unlock();
328 	return err;
329 }
330 EXPORT_SYMBOL_GPL(rtnl_link_register);
331 
__rtnl_kill_links(struct net * net,struct rtnl_link_ops * ops)332 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
333 {
334 	struct net_device *dev;
335 	LIST_HEAD(list_kill);
336 
337 	for_each_netdev(net, dev) {
338 		if (dev->rtnl_link_ops == ops)
339 			ops->dellink(dev, &list_kill);
340 	}
341 	unregister_netdevice_many(&list_kill);
342 }
343 
344 /**
345  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
346  * @ops: struct rtnl_link_ops * to unregister
347  *
348  * The caller must hold the rtnl_mutex.
349  */
__rtnl_link_unregister(struct rtnl_link_ops * ops)350 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
351 {
352 	struct net *net;
353 
354 	for_each_net(net) {
355 		__rtnl_kill_links(net, ops);
356 	}
357 	list_del(&ops->list);
358 }
359 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
360 
361 /* Return with the rtnl_lock held when there are no network
362  * devices unregistering in any network namespace.
363  */
rtnl_lock_unregistering_all(void)364 static void rtnl_lock_unregistering_all(void)
365 {
366 	struct net *net;
367 	bool unregistering;
368 	DEFINE_WAIT(wait);
369 
370 	for (;;) {
371 		prepare_to_wait(&netdev_unregistering_wq, &wait,
372 				TASK_UNINTERRUPTIBLE);
373 		unregistering = false;
374 		rtnl_lock();
375 		for_each_net(net) {
376 			if (net->dev_unreg_count > 0) {
377 				unregistering = true;
378 				break;
379 			}
380 		}
381 		if (!unregistering)
382 			break;
383 		__rtnl_unlock();
384 		schedule();
385 	}
386 	finish_wait(&netdev_unregistering_wq, &wait);
387 }
388 
389 /**
390  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
391  * @ops: struct rtnl_link_ops * to unregister
392  */
rtnl_link_unregister(struct rtnl_link_ops * ops)393 void rtnl_link_unregister(struct rtnl_link_ops *ops)
394 {
395 	/* Close the race with cleanup_net() */
396 	mutex_lock(&net_mutex);
397 	rtnl_lock_unregistering_all();
398 	__rtnl_link_unregister(ops);
399 	rtnl_unlock();
400 	mutex_unlock(&net_mutex);
401 }
402 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
403 
rtnl_link_get_slave_info_data_size(const struct net_device * dev)404 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
405 {
406 	struct net_device *master_dev;
407 	const struct rtnl_link_ops *ops;
408 
409 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
410 	if (!master_dev)
411 		return 0;
412 	ops = master_dev->rtnl_link_ops;
413 	if (!ops || !ops->get_slave_size)
414 		return 0;
415 	/* IFLA_INFO_SLAVE_DATA + nested data */
416 	return nla_total_size(sizeof(struct nlattr)) +
417 	       ops->get_slave_size(master_dev, dev);
418 }
419 
rtnl_link_get_size(const struct net_device * dev)420 static size_t rtnl_link_get_size(const struct net_device *dev)
421 {
422 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
423 	size_t size;
424 
425 	if (!ops)
426 		return 0;
427 
428 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
429 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
430 
431 	if (ops->get_size)
432 		/* IFLA_INFO_DATA + nested data */
433 		size += nla_total_size(sizeof(struct nlattr)) +
434 			ops->get_size(dev);
435 
436 	if (ops->get_xstats_size)
437 		/* IFLA_INFO_XSTATS */
438 		size += nla_total_size(ops->get_xstats_size(dev));
439 
440 	size += rtnl_link_get_slave_info_data_size(dev);
441 
442 	return size;
443 }
444 
445 static LIST_HEAD(rtnl_af_ops);
446 
rtnl_af_lookup(const int family)447 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
448 {
449 	const struct rtnl_af_ops *ops;
450 
451 	list_for_each_entry(ops, &rtnl_af_ops, list) {
452 		if (ops->family == family)
453 			return ops;
454 	}
455 
456 	return NULL;
457 }
458 
459 /**
460  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
461  * @ops: struct rtnl_af_ops * to register
462  *
463  * Returns 0 on success or a negative error code.
464  */
rtnl_af_register(struct rtnl_af_ops * ops)465 void rtnl_af_register(struct rtnl_af_ops *ops)
466 {
467 	rtnl_lock();
468 	list_add_tail(&ops->list, &rtnl_af_ops);
469 	rtnl_unlock();
470 }
471 EXPORT_SYMBOL_GPL(rtnl_af_register);
472 
473 /**
474  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
475  * @ops: struct rtnl_af_ops * to unregister
476  *
477  * The caller must hold the rtnl_mutex.
478  */
__rtnl_af_unregister(struct rtnl_af_ops * ops)479 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
480 {
481 	list_del(&ops->list);
482 }
483 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
484 
485 /**
486  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
487  * @ops: struct rtnl_af_ops * to unregister
488  */
rtnl_af_unregister(struct rtnl_af_ops * ops)489 void rtnl_af_unregister(struct rtnl_af_ops *ops)
490 {
491 	rtnl_lock();
492 	__rtnl_af_unregister(ops);
493 	rtnl_unlock();
494 }
495 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
496 
rtnl_link_get_af_size(const struct net_device * dev)497 static size_t rtnl_link_get_af_size(const struct net_device *dev)
498 {
499 	struct rtnl_af_ops *af_ops;
500 	size_t size;
501 
502 	/* IFLA_AF_SPEC */
503 	size = nla_total_size(sizeof(struct nlattr));
504 
505 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
506 		if (af_ops->get_link_af_size) {
507 			/* AF_* + nested data */
508 			size += nla_total_size(sizeof(struct nlattr)) +
509 				af_ops->get_link_af_size(dev);
510 		}
511 	}
512 
513 	return size;
514 }
515 
rtnl_have_link_slave_info(const struct net_device * dev)516 static bool rtnl_have_link_slave_info(const struct net_device *dev)
517 {
518 	struct net_device *master_dev;
519 
520 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
521 	if (master_dev && master_dev->rtnl_link_ops)
522 		return true;
523 	return false;
524 }
525 
rtnl_link_slave_info_fill(struct sk_buff * skb,const struct net_device * dev)526 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
527 				     const struct net_device *dev)
528 {
529 	struct net_device *master_dev;
530 	const struct rtnl_link_ops *ops;
531 	struct nlattr *slave_data;
532 	int err;
533 
534 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
535 	if (!master_dev)
536 		return 0;
537 	ops = master_dev->rtnl_link_ops;
538 	if (!ops)
539 		return 0;
540 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
541 		return -EMSGSIZE;
542 	if (ops->fill_slave_info) {
543 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
544 		if (!slave_data)
545 			return -EMSGSIZE;
546 		err = ops->fill_slave_info(skb, master_dev, dev);
547 		if (err < 0)
548 			goto err_cancel_slave_data;
549 		nla_nest_end(skb, slave_data);
550 	}
551 	return 0;
552 
553 err_cancel_slave_data:
554 	nla_nest_cancel(skb, slave_data);
555 	return err;
556 }
557 
rtnl_link_info_fill(struct sk_buff * skb,const struct net_device * dev)558 static int rtnl_link_info_fill(struct sk_buff *skb,
559 			       const struct net_device *dev)
560 {
561 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
562 	struct nlattr *data;
563 	int err;
564 
565 	if (!ops)
566 		return 0;
567 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
568 		return -EMSGSIZE;
569 	if (ops->fill_xstats) {
570 		err = ops->fill_xstats(skb, dev);
571 		if (err < 0)
572 			return err;
573 	}
574 	if (ops->fill_info) {
575 		data = nla_nest_start(skb, IFLA_INFO_DATA);
576 		if (data == NULL)
577 			return -EMSGSIZE;
578 		err = ops->fill_info(skb, dev);
579 		if (err < 0)
580 			goto err_cancel_data;
581 		nla_nest_end(skb, data);
582 	}
583 	return 0;
584 
585 err_cancel_data:
586 	nla_nest_cancel(skb, data);
587 	return err;
588 }
589 
rtnl_link_fill(struct sk_buff * skb,const struct net_device * dev)590 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
591 {
592 	struct nlattr *linkinfo;
593 	int err = -EMSGSIZE;
594 
595 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
596 	if (linkinfo == NULL)
597 		goto out;
598 
599 	err = rtnl_link_info_fill(skb, dev);
600 	if (err < 0)
601 		goto err_cancel_link;
602 
603 	err = rtnl_link_slave_info_fill(skb, dev);
604 	if (err < 0)
605 		goto err_cancel_link;
606 
607 	nla_nest_end(skb, linkinfo);
608 	return 0;
609 
610 err_cancel_link:
611 	nla_nest_cancel(skb, linkinfo);
612 out:
613 	return err;
614 }
615 
rtnetlink_send(struct sk_buff * skb,struct net * net,u32 pid,unsigned int group,int echo)616 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
617 {
618 	struct sock *rtnl = net->rtnl;
619 	int err = 0;
620 
621 	NETLINK_CB(skb).dst_group = group;
622 	if (echo)
623 		atomic_inc(&skb->users);
624 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
625 	if (echo)
626 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
627 	return err;
628 }
629 
rtnl_unicast(struct sk_buff * skb,struct net * net,u32 pid)630 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
631 {
632 	struct sock *rtnl = net->rtnl;
633 
634 	return nlmsg_unicast(rtnl, skb, pid);
635 }
636 EXPORT_SYMBOL(rtnl_unicast);
637 
rtnl_notify(struct sk_buff * skb,struct net * net,u32 pid,u32 group,struct nlmsghdr * nlh,gfp_t flags)638 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
639 		 struct nlmsghdr *nlh, gfp_t flags)
640 {
641 	struct sock *rtnl = net->rtnl;
642 	int report = 0;
643 
644 	if (nlh)
645 		report = nlmsg_report(nlh);
646 
647 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
648 }
649 EXPORT_SYMBOL(rtnl_notify);
650 
rtnl_set_sk_err(struct net * net,u32 group,int error)651 void rtnl_set_sk_err(struct net *net, u32 group, int error)
652 {
653 	struct sock *rtnl = net->rtnl;
654 
655 	netlink_set_err(rtnl, 0, group, error);
656 }
657 EXPORT_SYMBOL(rtnl_set_sk_err);
658 
rtnetlink_put_metrics(struct sk_buff * skb,u32 * metrics)659 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
660 {
661 	struct nlattr *mx;
662 	int i, valid = 0;
663 
664 	mx = nla_nest_start(skb, RTA_METRICS);
665 	if (mx == NULL)
666 		return -ENOBUFS;
667 
668 	for (i = 0; i < RTAX_MAX; i++) {
669 		if (metrics[i]) {
670 			valid++;
671 			if (nla_put_u32(skb, i+1, metrics[i]))
672 				goto nla_put_failure;
673 		}
674 	}
675 
676 	if (!valid) {
677 		nla_nest_cancel(skb, mx);
678 		return 0;
679 	}
680 
681 	return nla_nest_end(skb, mx);
682 
683 nla_put_failure:
684 	nla_nest_cancel(skb, mx);
685 	return -EMSGSIZE;
686 }
687 EXPORT_SYMBOL(rtnetlink_put_metrics);
688 
rtnl_put_cacheinfo(struct sk_buff * skb,struct dst_entry * dst,u32 id,long expires,u32 error)689 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
690 		       long expires, u32 error)
691 {
692 	struct rta_cacheinfo ci = {
693 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
694 		.rta_used = dst->__use,
695 		.rta_clntref = atomic_read(&(dst->__refcnt)),
696 		.rta_error = error,
697 		.rta_id =  id,
698 	};
699 
700 	if (expires) {
701 		unsigned long clock;
702 
703 		clock = jiffies_to_clock_t(abs(expires));
704 		clock = min_t(unsigned long, clock, INT_MAX);
705 		ci.rta_expires = (expires > 0) ? clock : -clock;
706 	}
707 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
708 }
709 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
710 
set_operstate(struct net_device * dev,unsigned char transition)711 static void set_operstate(struct net_device *dev, unsigned char transition)
712 {
713 	unsigned char operstate = dev->operstate;
714 
715 	switch (transition) {
716 	case IF_OPER_UP:
717 		if ((operstate == IF_OPER_DORMANT ||
718 		     operstate == IF_OPER_UNKNOWN) &&
719 		    !netif_dormant(dev))
720 			operstate = IF_OPER_UP;
721 		break;
722 
723 	case IF_OPER_DORMANT:
724 		if (operstate == IF_OPER_UP ||
725 		    operstate == IF_OPER_UNKNOWN)
726 			operstate = IF_OPER_DORMANT;
727 		break;
728 	}
729 
730 	if (dev->operstate != operstate) {
731 		write_lock_bh(&dev_base_lock);
732 		dev->operstate = operstate;
733 		write_unlock_bh(&dev_base_lock);
734 		netdev_state_change(dev);
735 	}
736 }
737 
rtnl_dev_get_flags(const struct net_device * dev)738 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
739 {
740 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
741 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
742 }
743 
rtnl_dev_combine_flags(const struct net_device * dev,const struct ifinfomsg * ifm)744 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
745 					   const struct ifinfomsg *ifm)
746 {
747 	unsigned int flags = ifm->ifi_flags;
748 
749 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
750 	if (ifm->ifi_change)
751 		flags = (flags & ifm->ifi_change) |
752 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
753 
754 	return flags;
755 }
756 
copy_rtnl_link_stats(struct rtnl_link_stats * a,const struct rtnl_link_stats64 * b)757 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
758 				 const struct rtnl_link_stats64 *b)
759 {
760 	a->rx_packets = b->rx_packets;
761 	a->tx_packets = b->tx_packets;
762 	a->rx_bytes = b->rx_bytes;
763 	a->tx_bytes = b->tx_bytes;
764 	a->rx_errors = b->rx_errors;
765 	a->tx_errors = b->tx_errors;
766 	a->rx_dropped = b->rx_dropped;
767 	a->tx_dropped = b->tx_dropped;
768 
769 	a->multicast = b->multicast;
770 	a->collisions = b->collisions;
771 
772 	a->rx_length_errors = b->rx_length_errors;
773 	a->rx_over_errors = b->rx_over_errors;
774 	a->rx_crc_errors = b->rx_crc_errors;
775 	a->rx_frame_errors = b->rx_frame_errors;
776 	a->rx_fifo_errors = b->rx_fifo_errors;
777 	a->rx_missed_errors = b->rx_missed_errors;
778 
779 	a->tx_aborted_errors = b->tx_aborted_errors;
780 	a->tx_carrier_errors = b->tx_carrier_errors;
781 	a->tx_fifo_errors = b->tx_fifo_errors;
782 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
783 	a->tx_window_errors = b->tx_window_errors;
784 
785 	a->rx_compressed = b->rx_compressed;
786 	a->tx_compressed = b->tx_compressed;
787 }
788 
copy_rtnl_link_stats64(void * v,const struct rtnl_link_stats64 * b)789 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
790 {
791 	memcpy(v, b, sizeof(*b));
792 }
793 
794 /* All VF info */
rtnl_vfinfo_size(const struct net_device * dev,u32 ext_filter_mask)795 static inline int rtnl_vfinfo_size(const struct net_device *dev,
796 				   u32 ext_filter_mask)
797 {
798 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
799 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
800 		int num_vfs = dev_num_vf(dev->dev.parent);
801 		size_t size = nla_total_size(sizeof(struct nlattr));
802 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
803 		size += num_vfs *
804 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
805 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
806 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
807 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
808 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
809 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)));
810 		return size;
811 	} else
812 		return 0;
813 }
814 
rtnl_port_size(const struct net_device * dev,u32 ext_filter_mask)815 static size_t rtnl_port_size(const struct net_device *dev,
816 			     u32 ext_filter_mask)
817 {
818 	size_t port_size = nla_total_size(4)		/* PORT_VF */
819 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
820 		+ nla_total_size(sizeof(struct ifla_port_vsi))
821 							/* PORT_VSI_TYPE */
822 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
823 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
824 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
825 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
826 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
827 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
828 		+ port_size;
829 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
830 		+ port_size;
831 
832 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
833 	    !(ext_filter_mask & RTEXT_FILTER_VF))
834 		return 0;
835 	if (dev_num_vf(dev->dev.parent))
836 		return port_self_size + vf_ports_size +
837 			vf_port_size * dev_num_vf(dev->dev.parent);
838 	else
839 		return port_self_size;
840 }
841 
if_nlmsg_size(const struct net_device * dev,u32 ext_filter_mask)842 static noinline size_t if_nlmsg_size(const struct net_device *dev,
843 				     u32 ext_filter_mask)
844 {
845 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
846 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
847 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
848 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
849 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
850 	       + nla_total_size(sizeof(struct rtnl_link_stats))
851 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
852 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
853 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
854 	       + nla_total_size(4) /* IFLA_TXQLEN */
855 	       + nla_total_size(4) /* IFLA_WEIGHT */
856 	       + nla_total_size(4) /* IFLA_MTU */
857 	       + nla_total_size(4) /* IFLA_LINK */
858 	       + nla_total_size(4) /* IFLA_MASTER */
859 	       + nla_total_size(1) /* IFLA_CARRIER */
860 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
861 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
862 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
863 	       + nla_total_size(1) /* IFLA_OPERSTATE */
864 	       + nla_total_size(1) /* IFLA_LINKMODE */
865 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
866 	       + nla_total_size(ext_filter_mask
867 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
868 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
869 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
870 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
871 	       + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
872 	       + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
873 }
874 
rtnl_vf_ports_fill(struct sk_buff * skb,struct net_device * dev)875 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
876 {
877 	struct nlattr *vf_ports;
878 	struct nlattr *vf_port;
879 	int vf;
880 	int err;
881 
882 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
883 	if (!vf_ports)
884 		return -EMSGSIZE;
885 
886 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
887 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
888 		if (!vf_port)
889 			goto nla_put_failure;
890 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
891 			goto nla_put_failure;
892 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
893 		if (err == -EMSGSIZE)
894 			goto nla_put_failure;
895 		if (err) {
896 			nla_nest_cancel(skb, vf_port);
897 			continue;
898 		}
899 		nla_nest_end(skb, vf_port);
900 	}
901 
902 	nla_nest_end(skb, vf_ports);
903 
904 	return 0;
905 
906 nla_put_failure:
907 	nla_nest_cancel(skb, vf_ports);
908 	return -EMSGSIZE;
909 }
910 
rtnl_port_self_fill(struct sk_buff * skb,struct net_device * dev)911 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
912 {
913 	struct nlattr *port_self;
914 	int err;
915 
916 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
917 	if (!port_self)
918 		return -EMSGSIZE;
919 
920 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
921 	if (err) {
922 		nla_nest_cancel(skb, port_self);
923 		return (err == -EMSGSIZE) ? err : 0;
924 	}
925 
926 	nla_nest_end(skb, port_self);
927 
928 	return 0;
929 }
930 
rtnl_port_fill(struct sk_buff * skb,struct net_device * dev,u32 ext_filter_mask)931 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
932 			  u32 ext_filter_mask)
933 {
934 	int err;
935 
936 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
937 	    !(ext_filter_mask & RTEXT_FILTER_VF))
938 		return 0;
939 
940 	err = rtnl_port_self_fill(skb, dev);
941 	if (err)
942 		return err;
943 
944 	if (dev_num_vf(dev->dev.parent)) {
945 		err = rtnl_vf_ports_fill(skb, dev);
946 		if (err)
947 			return err;
948 	}
949 
950 	return 0;
951 }
952 
rtnl_phys_port_id_fill(struct sk_buff * skb,struct net_device * dev)953 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
954 {
955 	int err;
956 	struct netdev_phys_port_id ppid;
957 
958 	err = dev_get_phys_port_id(dev, &ppid);
959 	if (err) {
960 		if (err == -EOPNOTSUPP)
961 			return 0;
962 		return err;
963 	}
964 
965 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
966 		return -EMSGSIZE;
967 
968 	return 0;
969 }
970 
rtnl_fill_ifinfo(struct sk_buff * skb,struct net_device * dev,int type,u32 pid,u32 seq,u32 change,unsigned int flags,u32 ext_filter_mask)971 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
972 			    int type, u32 pid, u32 seq, u32 change,
973 			    unsigned int flags, u32 ext_filter_mask)
974 {
975 	struct ifinfomsg *ifm;
976 	struct nlmsghdr *nlh;
977 	struct rtnl_link_stats64 temp;
978 	const struct rtnl_link_stats64 *stats;
979 	struct nlattr *attr, *af_spec;
980 	struct rtnl_af_ops *af_ops;
981 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
982 
983 	ASSERT_RTNL();
984 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
985 	if (nlh == NULL)
986 		return -EMSGSIZE;
987 
988 	ifm = nlmsg_data(nlh);
989 	ifm->ifi_family = AF_UNSPEC;
990 	ifm->__ifi_pad = 0;
991 	ifm->ifi_type = dev->type;
992 	ifm->ifi_index = dev->ifindex;
993 	ifm->ifi_flags = dev_get_flags(dev);
994 	ifm->ifi_change = change;
995 
996 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
997 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
998 	    nla_put_u8(skb, IFLA_OPERSTATE,
999 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1000 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1001 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1002 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1003 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1004 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1005 #ifdef CONFIG_RPS
1006 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1007 #endif
1008 	    (dev->ifindex != dev->iflink &&
1009 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1010 	    (upper_dev &&
1011 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1012 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1013 	    (dev->qdisc &&
1014 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1015 	    (dev->ifalias &&
1016 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1017 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1018 			atomic_read(&dev->carrier_changes)))
1019 		goto nla_put_failure;
1020 
1021 	if (1) {
1022 		struct rtnl_link_ifmap map;
1023 
1024 		memset(&map, 0, sizeof(map));
1025 		map.mem_start   = dev->mem_start;
1026 		map.mem_end     = dev->mem_end;
1027 		map.base_addr   = dev->base_addr;
1028 		map.irq         = dev->irq;
1029 		map.dma         = dev->dma;
1030 		map.port        = dev->if_port;
1031 
1032 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1033 			goto nla_put_failure;
1034 	}
1035 
1036 	if (dev->addr_len) {
1037 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1038 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1039 			goto nla_put_failure;
1040 	}
1041 
1042 	if (rtnl_phys_port_id_fill(skb, dev))
1043 		goto nla_put_failure;
1044 
1045 	attr = nla_reserve(skb, IFLA_STATS,
1046 			sizeof(struct rtnl_link_stats));
1047 	if (attr == NULL)
1048 		goto nla_put_failure;
1049 
1050 	stats = dev_get_stats(dev, &temp);
1051 	copy_rtnl_link_stats(nla_data(attr), stats);
1052 
1053 	attr = nla_reserve(skb, IFLA_STATS64,
1054 			sizeof(struct rtnl_link_stats64));
1055 	if (attr == NULL)
1056 		goto nla_put_failure;
1057 	copy_rtnl_link_stats64(nla_data(attr), stats);
1058 
1059 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1060 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1061 		goto nla_put_failure;
1062 
1063 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1064 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
1065 		int i;
1066 
1067 		struct nlattr *vfinfo, *vf;
1068 		int num_vfs = dev_num_vf(dev->dev.parent);
1069 
1070 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1071 		if (!vfinfo)
1072 			goto nla_put_failure;
1073 		for (i = 0; i < num_vfs; i++) {
1074 			struct ifla_vf_info ivi;
1075 			struct ifla_vf_mac vf_mac;
1076 			struct ifla_vf_vlan vf_vlan;
1077 			struct ifla_vf_rate vf_rate;
1078 			struct ifla_vf_tx_rate vf_tx_rate;
1079 			struct ifla_vf_spoofchk vf_spoofchk;
1080 			struct ifla_vf_link_state vf_linkstate;
1081 			struct ifla_vf_rss_query_en vf_rss_query_en;
1082 
1083 			/*
1084 			 * Not all SR-IOV capable drivers support the
1085 			 * spoofcheck and "RSS query enable" query.  Preset to
1086 			 * -1 so the user space tool can detect that the driver
1087 			 * didn't report anything.
1088 			 */
1089 			ivi.spoofchk = -1;
1090 			ivi.rss_query_en = -1;
1091 			memset(ivi.mac, 0, sizeof(ivi.mac));
1092 			/* The default value for VF link state is "auto"
1093 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
1094 			 */
1095 			ivi.linkstate = 0;
1096 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1097 				break;
1098 			vf_mac.vf =
1099 				vf_vlan.vf =
1100 				vf_rate.vf =
1101 				vf_tx_rate.vf =
1102 				vf_spoofchk.vf =
1103 				vf_linkstate.vf =
1104 				vf_rss_query_en.vf = ivi.vf;
1105 
1106 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1107 			vf_vlan.vlan = ivi.vlan;
1108 			vf_vlan.qos = ivi.qos;
1109 			vf_tx_rate.rate = ivi.max_tx_rate;
1110 			vf_rate.min_tx_rate = ivi.min_tx_rate;
1111 			vf_rate.max_tx_rate = ivi.max_tx_rate;
1112 			vf_spoofchk.setting = ivi.spoofchk;
1113 			vf_linkstate.link_state = ivi.linkstate;
1114 			vf_rss_query_en.setting = ivi.rss_query_en;
1115 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1116 			if (!vf) {
1117 				nla_nest_cancel(skb, vfinfo);
1118 				goto nla_put_failure;
1119 			}
1120 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1121 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1122 			    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1123 				    &vf_rate) ||
1124 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1125 				    &vf_tx_rate) ||
1126 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1127 				    &vf_spoofchk) ||
1128 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1129 				    &vf_linkstate) ||
1130 			    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1131 				    sizeof(vf_rss_query_en),
1132 				    &vf_rss_query_en))
1133 				goto nla_put_failure;
1134 			nla_nest_end(skb, vf);
1135 		}
1136 		nla_nest_end(skb, vfinfo);
1137 	}
1138 
1139 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1140 		goto nla_put_failure;
1141 
1142 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1143 		if (rtnl_link_fill(skb, dev) < 0)
1144 			goto nla_put_failure;
1145 	}
1146 
1147 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1148 		goto nla_put_failure;
1149 
1150 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1151 		if (af_ops->fill_link_af) {
1152 			struct nlattr *af;
1153 			int err;
1154 
1155 			if (!(af = nla_nest_start(skb, af_ops->family)))
1156 				goto nla_put_failure;
1157 
1158 			err = af_ops->fill_link_af(skb, dev);
1159 
1160 			/*
1161 			 * Caller may return ENODATA to indicate that there
1162 			 * was no data to be dumped. This is not an error, it
1163 			 * means we should trim the attribute header and
1164 			 * continue.
1165 			 */
1166 			if (err == -ENODATA)
1167 				nla_nest_cancel(skb, af);
1168 			else if (err < 0)
1169 				goto nla_put_failure;
1170 
1171 			nla_nest_end(skb, af);
1172 		}
1173 	}
1174 
1175 	nla_nest_end(skb, af_spec);
1176 
1177 	return nlmsg_end(skb, nlh);
1178 
1179 nla_put_failure:
1180 	nlmsg_cancel(skb, nlh);
1181 	return -EMSGSIZE;
1182 }
1183 
1184 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1185 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1186 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1187 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1188 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1189 	[IFLA_MTU]		= { .type = NLA_U32 },
1190 	[IFLA_LINK]		= { .type = NLA_U32 },
1191 	[IFLA_MASTER]		= { .type = NLA_U32 },
1192 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1193 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1194 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1195 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1196 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1197 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1198 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1199 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1200 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1201 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1202 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1203 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1204 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1205 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1206 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1207 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1208 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1209 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1210 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1211 };
1212 
1213 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1214 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1215 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1216 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1217 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1218 };
1219 
1220 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1221 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1222 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1223 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1224 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1225 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1226 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1227 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1228 };
1229 
1230 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1231 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1232 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1233 				    .len = PORT_PROFILE_MAX },
1234 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1235 				    .len = sizeof(struct ifla_port_vsi)},
1236 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1237 				      .len = PORT_UUID_MAX },
1238 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1239 				    .len = PORT_UUID_MAX },
1240 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1241 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1242 };
1243 
rtnl_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)1244 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1245 {
1246 	struct net *net = sock_net(skb->sk);
1247 	int h, s_h;
1248 	int idx = 0, s_idx;
1249 	struct net_device *dev;
1250 	struct hlist_head *head;
1251 	struct nlattr *tb[IFLA_MAX+1];
1252 	u32 ext_filter_mask = 0;
1253 	int err;
1254 	int hdrlen;
1255 
1256 	s_h = cb->args[0];
1257 	s_idx = cb->args[1];
1258 
1259 	cb->seq = net->dev_base_seq;
1260 
1261 	/* A hack to preserve kernel<->userspace interface.
1262 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1263 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1264 	 * what iproute2 < v3.9.0 used.
1265 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1266 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1267 	 */
1268 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1269 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1270 
1271 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1272 
1273 		if (tb[IFLA_EXT_MASK])
1274 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1275 	}
1276 
1277 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1278 		idx = 0;
1279 		head = &net->dev_index_head[h];
1280 		hlist_for_each_entry(dev, head, index_hlist) {
1281 			if (idx < s_idx)
1282 				goto cont;
1283 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1284 					       NETLINK_CB(cb->skb).portid,
1285 					       cb->nlh->nlmsg_seq, 0,
1286 					       NLM_F_MULTI,
1287 					       ext_filter_mask);
1288 			/* If we ran out of room on the first message,
1289 			 * we're in trouble
1290 			 */
1291 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1292 
1293 			if (err <= 0)
1294 				goto out;
1295 
1296 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1297 cont:
1298 			idx++;
1299 		}
1300 	}
1301 out:
1302 	cb->args[1] = idx;
1303 	cb->args[0] = h;
1304 
1305 	return skb->len;
1306 }
1307 
rtnl_nla_parse_ifla(struct nlattr ** tb,const struct nlattr * head,int len)1308 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1309 {
1310 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1311 }
1312 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1313 
rtnl_link_get_net(struct net * src_net,struct nlattr * tb[])1314 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1315 {
1316 	struct net *net;
1317 	/* Examine the link attributes and figure out which
1318 	 * network namespace we are talking about.
1319 	 */
1320 	if (tb[IFLA_NET_NS_PID])
1321 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1322 	else if (tb[IFLA_NET_NS_FD])
1323 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1324 	else
1325 		net = get_net(src_net);
1326 	return net;
1327 }
1328 EXPORT_SYMBOL(rtnl_link_get_net);
1329 
validate_linkmsg(struct net_device * dev,struct nlattr * tb[])1330 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1331 {
1332 	if (dev) {
1333 		if (tb[IFLA_ADDRESS] &&
1334 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1335 			return -EINVAL;
1336 
1337 		if (tb[IFLA_BROADCAST] &&
1338 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1339 			return -EINVAL;
1340 	}
1341 
1342 	if (tb[IFLA_AF_SPEC]) {
1343 		struct nlattr *af;
1344 		int rem, err;
1345 
1346 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1347 			const struct rtnl_af_ops *af_ops;
1348 
1349 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1350 				return -EAFNOSUPPORT;
1351 
1352 			if (!af_ops->set_link_af)
1353 				return -EOPNOTSUPP;
1354 
1355 			if (af_ops->validate_link_af) {
1356 				err = af_ops->validate_link_af(dev, af);
1357 				if (err < 0)
1358 					return err;
1359 			}
1360 		}
1361 	}
1362 
1363 	return 0;
1364 }
1365 
do_setvfinfo(struct net_device * dev,struct nlattr ** tb)1366 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1367 {
1368 	const struct net_device_ops *ops = dev->netdev_ops;
1369 	int err = -EINVAL;
1370 
1371 	if (tb[IFLA_VF_MAC]) {
1372 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1373 
1374 		err = -EOPNOTSUPP;
1375 		if (ops->ndo_set_vf_mac)
1376 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1377 						  ivm->mac);
1378 		if (err < 0)
1379 			return err;
1380 	}
1381 
1382 	if (tb[IFLA_VF_VLAN]) {
1383 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1384 
1385 		err = -EOPNOTSUPP;
1386 		if (ops->ndo_set_vf_vlan)
1387 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1388 						   ivv->qos);
1389 		if (err < 0)
1390 			return err;
1391 	}
1392 
1393 	if (tb[IFLA_VF_TX_RATE]) {
1394 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1395 		struct ifla_vf_info ivf;
1396 
1397 		err = -EOPNOTSUPP;
1398 		if (ops->ndo_get_vf_config)
1399 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1400 		if (err < 0)
1401 			return err;
1402 
1403 		err = -EOPNOTSUPP;
1404 		if (ops->ndo_set_vf_rate)
1405 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1406 						   ivf.min_tx_rate,
1407 						   ivt->rate);
1408 		if (err < 0)
1409 			return err;
1410 	}
1411 
1412 	if (tb[IFLA_VF_RATE]) {
1413 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1414 
1415 		err = -EOPNOTSUPP;
1416 		if (ops->ndo_set_vf_rate)
1417 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1418 						   ivt->min_tx_rate,
1419 						   ivt->max_tx_rate);
1420 		if (err < 0)
1421 			return err;
1422 	}
1423 
1424 	if (tb[IFLA_VF_SPOOFCHK]) {
1425 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1426 
1427 		err = -EOPNOTSUPP;
1428 		if (ops->ndo_set_vf_spoofchk)
1429 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1430 						       ivs->setting);
1431 		if (err < 0)
1432 			return err;
1433 	}
1434 
1435 	if (tb[IFLA_VF_LINK_STATE]) {
1436 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1437 
1438 		err = -EOPNOTSUPP;
1439 		if (ops->ndo_set_vf_link_state)
1440 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1441 							 ivl->link_state);
1442 		if (err < 0)
1443 			return err;
1444 	}
1445 
1446 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1447 		struct ifla_vf_rss_query_en *ivrssq_en;
1448 
1449 		err = -EOPNOTSUPP;
1450 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1451 		if (ops->ndo_set_vf_rss_query_en)
1452 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1453 							   ivrssq_en->setting);
1454 		if (err < 0)
1455 			return err;
1456 	}
1457 
1458 	return err;
1459 }
1460 
do_set_master(struct net_device * dev,int ifindex)1461 static int do_set_master(struct net_device *dev, int ifindex)
1462 {
1463 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1464 	const struct net_device_ops *ops;
1465 	int err;
1466 
1467 	if (upper_dev) {
1468 		if (upper_dev->ifindex == ifindex)
1469 			return 0;
1470 		ops = upper_dev->netdev_ops;
1471 		if (ops->ndo_del_slave) {
1472 			err = ops->ndo_del_slave(upper_dev, dev);
1473 			if (err)
1474 				return err;
1475 		} else {
1476 			return -EOPNOTSUPP;
1477 		}
1478 	}
1479 
1480 	if (ifindex) {
1481 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1482 		if (!upper_dev)
1483 			return -EINVAL;
1484 		ops = upper_dev->netdev_ops;
1485 		if (ops->ndo_add_slave) {
1486 			err = ops->ndo_add_slave(upper_dev, dev);
1487 			if (err)
1488 				return err;
1489 		} else {
1490 			return -EOPNOTSUPP;
1491 		}
1492 	}
1493 	return 0;
1494 }
1495 
1496 #define DO_SETLINK_MODIFIED	0x01
1497 /* notify flag means notify + modified. */
1498 #define DO_SETLINK_NOTIFY	0x03
do_setlink(const struct sk_buff * skb,struct net_device * dev,struct ifinfomsg * ifm,struct nlattr ** tb,char * ifname,int status)1499 static int do_setlink(const struct sk_buff *skb,
1500 		      struct net_device *dev, struct ifinfomsg *ifm,
1501 		      struct nlattr **tb, char *ifname, int status)
1502 {
1503 	const struct net_device_ops *ops = dev->netdev_ops;
1504 	int err;
1505 
1506 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1507 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1508 		if (IS_ERR(net)) {
1509 			err = PTR_ERR(net);
1510 			goto errout;
1511 		}
1512 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1513 			put_net(net);
1514 			err = -EPERM;
1515 			goto errout;
1516 		}
1517 		err = dev_change_net_namespace(dev, net, ifname);
1518 		put_net(net);
1519 		if (err)
1520 			goto errout;
1521 		status |= DO_SETLINK_MODIFIED;
1522 	}
1523 
1524 	if (tb[IFLA_MAP]) {
1525 		struct rtnl_link_ifmap *u_map;
1526 		struct ifmap k_map;
1527 
1528 		if (!ops->ndo_set_config) {
1529 			err = -EOPNOTSUPP;
1530 			goto errout;
1531 		}
1532 
1533 		if (!netif_device_present(dev)) {
1534 			err = -ENODEV;
1535 			goto errout;
1536 		}
1537 
1538 		u_map = nla_data(tb[IFLA_MAP]);
1539 		k_map.mem_start = (unsigned long) u_map->mem_start;
1540 		k_map.mem_end = (unsigned long) u_map->mem_end;
1541 		k_map.base_addr = (unsigned short) u_map->base_addr;
1542 		k_map.irq = (unsigned char) u_map->irq;
1543 		k_map.dma = (unsigned char) u_map->dma;
1544 		k_map.port = (unsigned char) u_map->port;
1545 
1546 		err = ops->ndo_set_config(dev, &k_map);
1547 		if (err < 0)
1548 			goto errout;
1549 
1550 		status |= DO_SETLINK_NOTIFY;
1551 	}
1552 
1553 	if (tb[IFLA_ADDRESS]) {
1554 		struct sockaddr *sa;
1555 		int len;
1556 
1557 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
1558 						  sizeof(*sa));
1559 		sa = kmalloc(len, GFP_KERNEL);
1560 		if (!sa) {
1561 			err = -ENOMEM;
1562 			goto errout;
1563 		}
1564 		sa->sa_family = dev->type;
1565 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1566 		       dev->addr_len);
1567 		err = dev_set_mac_address(dev, sa);
1568 		kfree(sa);
1569 		if (err)
1570 			goto errout;
1571 		status |= DO_SETLINK_MODIFIED;
1572 	}
1573 
1574 	if (tb[IFLA_MTU]) {
1575 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1576 		if (err < 0)
1577 			goto errout;
1578 		status |= DO_SETLINK_MODIFIED;
1579 	}
1580 
1581 	if (tb[IFLA_GROUP]) {
1582 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1583 		status |= DO_SETLINK_NOTIFY;
1584 	}
1585 
1586 	/*
1587 	 * Interface selected by interface index but interface
1588 	 * name provided implies that a name change has been
1589 	 * requested.
1590 	 */
1591 	if (ifm->ifi_index > 0 && ifname[0]) {
1592 		err = dev_change_name(dev, ifname);
1593 		if (err < 0)
1594 			goto errout;
1595 		status |= DO_SETLINK_MODIFIED;
1596 	}
1597 
1598 	if (tb[IFLA_IFALIAS]) {
1599 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1600 				    nla_len(tb[IFLA_IFALIAS]));
1601 		if (err < 0)
1602 			goto errout;
1603 		status |= DO_SETLINK_NOTIFY;
1604 	}
1605 
1606 	if (tb[IFLA_BROADCAST]) {
1607 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1608 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1609 	}
1610 
1611 	if (ifm->ifi_flags || ifm->ifi_change) {
1612 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1613 		if (err < 0)
1614 			goto errout;
1615 	}
1616 
1617 	if (tb[IFLA_MASTER]) {
1618 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1619 		if (err)
1620 			goto errout;
1621 		status |= DO_SETLINK_MODIFIED;
1622 	}
1623 
1624 	if (tb[IFLA_CARRIER]) {
1625 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1626 		if (err)
1627 			goto errout;
1628 		status |= DO_SETLINK_MODIFIED;
1629 	}
1630 
1631 	if (tb[IFLA_TXQLEN]) {
1632 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1633 
1634 		if (dev->tx_queue_len ^ value)
1635 			status |= DO_SETLINK_NOTIFY;
1636 
1637 		dev->tx_queue_len = value;
1638 	}
1639 
1640 	if (tb[IFLA_OPERSTATE])
1641 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1642 
1643 	if (tb[IFLA_LINKMODE]) {
1644 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1645 
1646 		write_lock_bh(&dev_base_lock);
1647 		if (dev->link_mode ^ value)
1648 			status |= DO_SETLINK_NOTIFY;
1649 		dev->link_mode = value;
1650 		write_unlock_bh(&dev_base_lock);
1651 	}
1652 
1653 	if (tb[IFLA_VFINFO_LIST]) {
1654 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1655 		struct nlattr *attr;
1656 		int rem;
1657 
1658 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1659 			if (nla_type(attr) != IFLA_VF_INFO ||
1660 			    nla_len(attr) < NLA_HDRLEN) {
1661 				err = -EINVAL;
1662 				goto errout;
1663 			}
1664 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1665 					       ifla_vf_policy);
1666 			if (err < 0)
1667 				goto errout;
1668 			err = do_setvfinfo(dev, vfinfo);
1669 			if (err < 0)
1670 				goto errout;
1671 			status |= DO_SETLINK_NOTIFY;
1672 		}
1673 	}
1674 	err = 0;
1675 
1676 	if (tb[IFLA_VF_PORTS]) {
1677 		struct nlattr *port[IFLA_PORT_MAX+1];
1678 		struct nlattr *attr;
1679 		int vf;
1680 		int rem;
1681 
1682 		err = -EOPNOTSUPP;
1683 		if (!ops->ndo_set_vf_port)
1684 			goto errout;
1685 
1686 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1687 			if (nla_type(attr) != IFLA_VF_PORT)
1688 				continue;
1689 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1690 				attr, ifla_port_policy);
1691 			if (err < 0)
1692 				goto errout;
1693 			if (!port[IFLA_PORT_VF]) {
1694 				err = -EOPNOTSUPP;
1695 				goto errout;
1696 			}
1697 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1698 			err = ops->ndo_set_vf_port(dev, vf, port);
1699 			if (err < 0)
1700 				goto errout;
1701 			status |= DO_SETLINK_NOTIFY;
1702 		}
1703 	}
1704 	err = 0;
1705 
1706 	if (tb[IFLA_PORT_SELF]) {
1707 		struct nlattr *port[IFLA_PORT_MAX+1];
1708 
1709 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1710 			tb[IFLA_PORT_SELF], ifla_port_policy);
1711 		if (err < 0)
1712 			goto errout;
1713 
1714 		err = -EOPNOTSUPP;
1715 		if (ops->ndo_set_vf_port)
1716 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1717 		if (err < 0)
1718 			goto errout;
1719 		status |= DO_SETLINK_NOTIFY;
1720 	}
1721 
1722 	if (tb[IFLA_AF_SPEC]) {
1723 		struct nlattr *af;
1724 		int rem;
1725 
1726 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1727 			const struct rtnl_af_ops *af_ops;
1728 
1729 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1730 				BUG();
1731 
1732 			err = af_ops->set_link_af(dev, af);
1733 			if (err < 0)
1734 				goto errout;
1735 
1736 			status |= DO_SETLINK_NOTIFY;
1737 		}
1738 	}
1739 	err = 0;
1740 
1741 errout:
1742 	if (status & DO_SETLINK_MODIFIED) {
1743 		if (status & DO_SETLINK_NOTIFY)
1744 			netdev_state_change(dev);
1745 
1746 		if (err < 0)
1747 			net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
1748 					     dev->name);
1749 	}
1750 
1751 	return err;
1752 }
1753 
rtnl_setlink(struct sk_buff * skb,struct nlmsghdr * nlh)1754 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1755 {
1756 	struct net *net = sock_net(skb->sk);
1757 	struct ifinfomsg *ifm;
1758 	struct net_device *dev;
1759 	int err;
1760 	struct nlattr *tb[IFLA_MAX+1];
1761 	char ifname[IFNAMSIZ];
1762 
1763 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1764 	if (err < 0)
1765 		goto errout;
1766 
1767 	if (tb[IFLA_IFNAME])
1768 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1769 	else
1770 		ifname[0] = '\0';
1771 
1772 	err = -EINVAL;
1773 	ifm = nlmsg_data(nlh);
1774 	if (ifm->ifi_index > 0)
1775 		dev = __dev_get_by_index(net, ifm->ifi_index);
1776 	else if (tb[IFLA_IFNAME])
1777 		dev = __dev_get_by_name(net, ifname);
1778 	else
1779 		goto errout;
1780 
1781 	if (dev == NULL) {
1782 		err = -ENODEV;
1783 		goto errout;
1784 	}
1785 
1786 	err = validate_linkmsg(dev, tb);
1787 	if (err < 0)
1788 		goto errout;
1789 
1790 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1791 errout:
1792 	return err;
1793 }
1794 
rtnl_dellink(struct sk_buff * skb,struct nlmsghdr * nlh)1795 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1796 {
1797 	struct net *net = sock_net(skb->sk);
1798 	const struct rtnl_link_ops *ops;
1799 	struct net_device *dev;
1800 	struct ifinfomsg *ifm;
1801 	char ifname[IFNAMSIZ];
1802 	struct nlattr *tb[IFLA_MAX+1];
1803 	int err;
1804 	LIST_HEAD(list_kill);
1805 
1806 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1807 	if (err < 0)
1808 		return err;
1809 
1810 	if (tb[IFLA_IFNAME])
1811 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1812 
1813 	ifm = nlmsg_data(nlh);
1814 	if (ifm->ifi_index > 0)
1815 		dev = __dev_get_by_index(net, ifm->ifi_index);
1816 	else if (tb[IFLA_IFNAME])
1817 		dev = __dev_get_by_name(net, ifname);
1818 	else
1819 		return -EINVAL;
1820 
1821 	if (!dev)
1822 		return -ENODEV;
1823 
1824 	ops = dev->rtnl_link_ops;
1825 	if (!ops || !ops->dellink)
1826 		return -EOPNOTSUPP;
1827 
1828 	ops->dellink(dev, &list_kill);
1829 	unregister_netdevice_many(&list_kill);
1830 	return 0;
1831 }
1832 
rtnl_configure_link(struct net_device * dev,const struct ifinfomsg * ifm)1833 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1834 {
1835 	unsigned int old_flags;
1836 	int err;
1837 
1838 	old_flags = dev->flags;
1839 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1840 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1841 		if (err < 0)
1842 			return err;
1843 	}
1844 
1845 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1846 
1847 	__dev_notify_flags(dev, old_flags, ~0U);
1848 	return 0;
1849 }
1850 EXPORT_SYMBOL(rtnl_configure_link);
1851 
rtnl_create_link(struct net * net,char * ifname,unsigned char name_assign_type,const struct rtnl_link_ops * ops,struct nlattr * tb[])1852 struct net_device *rtnl_create_link(struct net *net,
1853 	char *ifname, unsigned char name_assign_type,
1854 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
1855 {
1856 	int err;
1857 	struct net_device *dev;
1858 	unsigned int num_tx_queues = 1;
1859 	unsigned int num_rx_queues = 1;
1860 
1861 	if (tb[IFLA_NUM_TX_QUEUES])
1862 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1863 	else if (ops->get_num_tx_queues)
1864 		num_tx_queues = ops->get_num_tx_queues();
1865 
1866 	if (tb[IFLA_NUM_RX_QUEUES])
1867 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1868 	else if (ops->get_num_rx_queues)
1869 		num_rx_queues = ops->get_num_rx_queues();
1870 
1871 	err = -ENOMEM;
1872 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1873 			       ops->setup, num_tx_queues, num_rx_queues);
1874 	if (!dev)
1875 		goto err;
1876 
1877 	dev_net_set(dev, net);
1878 	dev->rtnl_link_ops = ops;
1879 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1880 
1881 	if (tb[IFLA_MTU])
1882 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1883 	if (tb[IFLA_ADDRESS]) {
1884 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1885 				nla_len(tb[IFLA_ADDRESS]));
1886 		dev->addr_assign_type = NET_ADDR_SET;
1887 	}
1888 	if (tb[IFLA_BROADCAST])
1889 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1890 				nla_len(tb[IFLA_BROADCAST]));
1891 	if (tb[IFLA_TXQLEN])
1892 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1893 	if (tb[IFLA_OPERSTATE])
1894 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1895 	if (tb[IFLA_LINKMODE])
1896 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1897 	if (tb[IFLA_GROUP])
1898 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1899 
1900 	return dev;
1901 
1902 err:
1903 	return ERR_PTR(err);
1904 }
1905 EXPORT_SYMBOL(rtnl_create_link);
1906 
rtnl_group_changelink(const struct sk_buff * skb,struct net * net,int group,struct ifinfomsg * ifm,struct nlattr ** tb)1907 static int rtnl_group_changelink(const struct sk_buff *skb,
1908 		struct net *net, int group,
1909 		struct ifinfomsg *ifm,
1910 		struct nlattr **tb)
1911 {
1912 	struct net_device *dev;
1913 	int err;
1914 
1915 	for_each_netdev(net, dev) {
1916 		if (dev->group == group) {
1917 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1918 			if (err < 0)
1919 				return err;
1920 		}
1921 	}
1922 
1923 	return 0;
1924 }
1925 
rtnl_newlink(struct sk_buff * skb,struct nlmsghdr * nlh)1926 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1927 {
1928 	struct net *net = sock_net(skb->sk);
1929 	const struct rtnl_link_ops *ops;
1930 	const struct rtnl_link_ops *m_ops = NULL;
1931 	struct net_device *dev;
1932 	struct net_device *master_dev = NULL;
1933 	struct ifinfomsg *ifm;
1934 	char kind[MODULE_NAME_LEN];
1935 	char ifname[IFNAMSIZ];
1936 	struct nlattr *tb[IFLA_MAX+1];
1937 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1938 	unsigned char name_assign_type = NET_NAME_USER;
1939 	int err;
1940 
1941 #ifdef CONFIG_MODULES
1942 replay:
1943 #endif
1944 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1945 	if (err < 0)
1946 		return err;
1947 
1948 	if (tb[IFLA_IFNAME])
1949 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1950 	else
1951 		ifname[0] = '\0';
1952 
1953 	ifm = nlmsg_data(nlh);
1954 	if (ifm->ifi_index > 0)
1955 		dev = __dev_get_by_index(net, ifm->ifi_index);
1956 	else {
1957 		if (ifname[0])
1958 			dev = __dev_get_by_name(net, ifname);
1959 		else
1960 			dev = NULL;
1961 	}
1962 
1963 	if (dev) {
1964 		master_dev = netdev_master_upper_dev_get(dev);
1965 		if (master_dev)
1966 			m_ops = master_dev->rtnl_link_ops;
1967 	}
1968 
1969 	err = validate_linkmsg(dev, tb);
1970 	if (err < 0)
1971 		return err;
1972 
1973 	if (tb[IFLA_LINKINFO]) {
1974 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1975 				       tb[IFLA_LINKINFO], ifla_info_policy);
1976 		if (err < 0)
1977 			return err;
1978 	} else
1979 		memset(linkinfo, 0, sizeof(linkinfo));
1980 
1981 	if (linkinfo[IFLA_INFO_KIND]) {
1982 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1983 		ops = rtnl_link_ops_get(kind);
1984 	} else {
1985 		kind[0] = '\0';
1986 		ops = NULL;
1987 	}
1988 
1989 	if (1) {
1990 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1991 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1992 		struct nlattr **data = NULL;
1993 		struct nlattr **slave_data = NULL;
1994 		struct net *dest_net;
1995 
1996 		if (ops) {
1997 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1998 				err = nla_parse_nested(attr, ops->maxtype,
1999 						       linkinfo[IFLA_INFO_DATA],
2000 						       ops->policy);
2001 				if (err < 0)
2002 					return err;
2003 				data = attr;
2004 			}
2005 			if (ops->validate) {
2006 				err = ops->validate(tb, data);
2007 				if (err < 0)
2008 					return err;
2009 			}
2010 		}
2011 
2012 		if (m_ops) {
2013 			if (m_ops->slave_maxtype &&
2014 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2015 				err = nla_parse_nested(slave_attr,
2016 						       m_ops->slave_maxtype,
2017 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2018 						       m_ops->slave_policy);
2019 				if (err < 0)
2020 					return err;
2021 				slave_data = slave_attr;
2022 			}
2023 			if (m_ops->slave_validate) {
2024 				err = m_ops->slave_validate(tb, slave_data);
2025 				if (err < 0)
2026 					return err;
2027 			}
2028 		}
2029 
2030 		if (dev) {
2031 			int status = 0;
2032 
2033 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2034 				return -EEXIST;
2035 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2036 				return -EOPNOTSUPP;
2037 
2038 			if (linkinfo[IFLA_INFO_DATA]) {
2039 				if (!ops || ops != dev->rtnl_link_ops ||
2040 				    !ops->changelink)
2041 					return -EOPNOTSUPP;
2042 
2043 				err = ops->changelink(dev, tb, data);
2044 				if (err < 0)
2045 					return err;
2046 				status |= DO_SETLINK_NOTIFY;
2047 			}
2048 
2049 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2050 				if (!m_ops || !m_ops->slave_changelink)
2051 					return -EOPNOTSUPP;
2052 
2053 				err = m_ops->slave_changelink(master_dev, dev,
2054 							      tb, slave_data);
2055 				if (err < 0)
2056 					return err;
2057 				status |= DO_SETLINK_NOTIFY;
2058 			}
2059 
2060 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2061 		}
2062 
2063 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2064 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2065 				return rtnl_group_changelink(skb, net,
2066 						nla_get_u32(tb[IFLA_GROUP]),
2067 						ifm, tb);
2068 			return -ENODEV;
2069 		}
2070 
2071 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2072 			return -EOPNOTSUPP;
2073 
2074 		if (!ops) {
2075 #ifdef CONFIG_MODULES
2076 			if (kind[0]) {
2077 				__rtnl_unlock();
2078 				request_module("rtnl-link-%s", kind);
2079 				rtnl_lock();
2080 				ops = rtnl_link_ops_get(kind);
2081 				if (ops)
2082 					goto replay;
2083 			}
2084 #endif
2085 			return -EOPNOTSUPP;
2086 		}
2087 
2088 		if (!ops->setup)
2089 			return -EOPNOTSUPP;
2090 
2091 		if (!ifname[0]) {
2092 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2093 			name_assign_type = NET_NAME_ENUM;
2094 		}
2095 
2096 		dest_net = rtnl_link_get_net(net, tb);
2097 		if (IS_ERR(dest_net))
2098 			return PTR_ERR(dest_net);
2099 
2100 		dev = rtnl_create_link(dest_net, ifname, name_assign_type, ops, tb);
2101 		if (IS_ERR(dev)) {
2102 			err = PTR_ERR(dev);
2103 			goto out;
2104 		}
2105 
2106 		dev->ifindex = ifm->ifi_index;
2107 
2108 		if (ops->newlink) {
2109 			err = ops->newlink(net, dev, tb, data);
2110 			/* Drivers should call free_netdev() in ->destructor
2111 			 * and unregister it on failure after registration
2112 			 * so that device could be finally freed in rtnl_unlock.
2113 			 */
2114 			if (err < 0) {
2115 				/* If device is not registered at all, free it now */
2116 				if (dev->reg_state == NETREG_UNINITIALIZED)
2117 					free_netdev(dev);
2118 				goto out;
2119 			}
2120 		} else {
2121 			err = register_netdevice(dev);
2122 			if (err < 0) {
2123 				free_netdev(dev);
2124 				goto out;
2125 			}
2126 		}
2127 		err = rtnl_configure_link(dev, ifm);
2128 		if (err < 0) {
2129 			if (ops->newlink) {
2130 				LIST_HEAD(list_kill);
2131 
2132 				ops->dellink(dev, &list_kill);
2133 				unregister_netdevice_many(&list_kill);
2134 			} else {
2135 				unregister_netdevice(dev);
2136 			}
2137 		}
2138 out:
2139 		put_net(dest_net);
2140 		return err;
2141 	}
2142 }
2143 
rtnl_getlink(struct sk_buff * skb,struct nlmsghdr * nlh)2144 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2145 {
2146 	struct net *net = sock_net(skb->sk);
2147 	struct ifinfomsg *ifm;
2148 	char ifname[IFNAMSIZ];
2149 	struct nlattr *tb[IFLA_MAX+1];
2150 	struct net_device *dev = NULL;
2151 	struct sk_buff *nskb;
2152 	int err;
2153 	u32 ext_filter_mask = 0;
2154 
2155 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2156 	if (err < 0)
2157 		return err;
2158 
2159 	if (tb[IFLA_IFNAME])
2160 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2161 
2162 	if (tb[IFLA_EXT_MASK])
2163 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2164 
2165 	ifm = nlmsg_data(nlh);
2166 	if (ifm->ifi_index > 0)
2167 		dev = __dev_get_by_index(net, ifm->ifi_index);
2168 	else if (tb[IFLA_IFNAME])
2169 		dev = __dev_get_by_name(net, ifname);
2170 	else
2171 		return -EINVAL;
2172 
2173 	if (dev == NULL)
2174 		return -ENODEV;
2175 
2176 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2177 	if (nskb == NULL)
2178 		return -ENOBUFS;
2179 
2180 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2181 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2182 	if (err < 0) {
2183 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2184 		WARN_ON(err == -EMSGSIZE);
2185 		kfree_skb(nskb);
2186 	} else
2187 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2188 
2189 	return err;
2190 }
2191 
rtnl_calcit(struct sk_buff * skb,struct nlmsghdr * nlh)2192 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2193 {
2194 	struct net *net = sock_net(skb->sk);
2195 	struct net_device *dev;
2196 	struct nlattr *tb[IFLA_MAX+1];
2197 	u32 ext_filter_mask = 0;
2198 	u16 min_ifinfo_dump_size = 0;
2199 	int hdrlen;
2200 
2201 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2202 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2203 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2204 
2205 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2206 		if (tb[IFLA_EXT_MASK])
2207 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2208 	}
2209 
2210 	if (!ext_filter_mask)
2211 		return NLMSG_GOODSIZE;
2212 	/*
2213 	 * traverse the list of net devices and compute the minimum
2214 	 * buffer size based upon the filter mask.
2215 	 */
2216 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2217 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2218 					     if_nlmsg_size(dev,
2219 						           ext_filter_mask));
2220 	}
2221 
2222 	return min_ifinfo_dump_size;
2223 }
2224 
rtnl_dump_all(struct sk_buff * skb,struct netlink_callback * cb)2225 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2226 {
2227 	int idx;
2228 	int s_idx = cb->family;
2229 
2230 	if (s_idx == 0)
2231 		s_idx = 1;
2232 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2233 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2234 		if (idx < s_idx || idx == PF_PACKET)
2235 			continue;
2236 		if (rtnl_msg_handlers[idx] == NULL ||
2237 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2238 			continue;
2239 		if (idx > s_idx) {
2240 			memset(&cb->args[0], 0, sizeof(cb->args));
2241 			cb->prev_seq = 0;
2242 			cb->seq = 0;
2243 		}
2244 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2245 			break;
2246 	}
2247 	cb->family = idx;
2248 
2249 	return skb->len;
2250 }
2251 
rtmsg_ifinfo(int type,struct net_device * dev,unsigned int change,gfp_t flags)2252 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2253 		  gfp_t flags)
2254 {
2255 	struct net *net = dev_net(dev);
2256 	struct sk_buff *skb;
2257 	int err = -ENOBUFS;
2258 	size_t if_info_size;
2259 
2260 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2261 	if (skb == NULL)
2262 		goto errout;
2263 
2264 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2265 	if (err < 0) {
2266 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2267 		WARN_ON(err == -EMSGSIZE);
2268 		kfree_skb(skb);
2269 		goto errout;
2270 	}
2271 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2272 	return;
2273 errout:
2274 	if (err < 0)
2275 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2276 }
2277 EXPORT_SYMBOL(rtmsg_ifinfo);
2278 
nlmsg_populate_fdb_fill(struct sk_buff * skb,struct net_device * dev,u8 * addr,u32 pid,u32 seq,int type,unsigned int flags,int nlflags)2279 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2280 				   struct net_device *dev,
2281 				   u8 *addr, u32 pid, u32 seq,
2282 				   int type, unsigned int flags,
2283 				   int nlflags)
2284 {
2285 	struct nlmsghdr *nlh;
2286 	struct ndmsg *ndm;
2287 
2288 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2289 	if (!nlh)
2290 		return -EMSGSIZE;
2291 
2292 	ndm = nlmsg_data(nlh);
2293 	ndm->ndm_family  = AF_BRIDGE;
2294 	ndm->ndm_pad1	 = 0;
2295 	ndm->ndm_pad2    = 0;
2296 	ndm->ndm_flags	 = flags;
2297 	ndm->ndm_type	 = 0;
2298 	ndm->ndm_ifindex = dev->ifindex;
2299 	ndm->ndm_state   = NUD_PERMANENT;
2300 
2301 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2302 		goto nla_put_failure;
2303 
2304 	return nlmsg_end(skb, nlh);
2305 
2306 nla_put_failure:
2307 	nlmsg_cancel(skb, nlh);
2308 	return -EMSGSIZE;
2309 }
2310 
rtnl_fdb_nlmsg_size(void)2311 static inline size_t rtnl_fdb_nlmsg_size(void)
2312 {
2313 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2314 }
2315 
rtnl_fdb_notify(struct net_device * dev,u8 * addr,int type)2316 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2317 {
2318 	struct net *net = dev_net(dev);
2319 	struct sk_buff *skb;
2320 	int err = -ENOBUFS;
2321 
2322 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2323 	if (!skb)
2324 		goto errout;
2325 
2326 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2327 	if (err < 0) {
2328 		kfree_skb(skb);
2329 		goto errout;
2330 	}
2331 
2332 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2333 	return;
2334 errout:
2335 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2336 }
2337 
2338 /**
2339  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2340  */
ndo_dflt_fdb_add(struct ndmsg * ndm,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr,u16 flags)2341 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2342 		     struct nlattr *tb[],
2343 		     struct net_device *dev,
2344 		     const unsigned char *addr,
2345 		     u16 flags)
2346 {
2347 	int err = -EINVAL;
2348 
2349 	/* If aging addresses are supported device will need to
2350 	 * implement its own handler for this.
2351 	 */
2352 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2353 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2354 		return err;
2355 	}
2356 
2357 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2358 		err = dev_uc_add_excl(dev, addr);
2359 	else if (is_multicast_ether_addr(addr))
2360 		err = dev_mc_add_excl(dev, addr);
2361 
2362 	/* Only return duplicate errors if NLM_F_EXCL is set */
2363 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2364 		err = 0;
2365 
2366 	return err;
2367 }
2368 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2369 
rtnl_fdb_add(struct sk_buff * skb,struct nlmsghdr * nlh)2370 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2371 {
2372 	struct net *net = sock_net(skb->sk);
2373 	struct ndmsg *ndm;
2374 	struct nlattr *tb[NDA_MAX+1];
2375 	struct net_device *dev;
2376 	u8 *addr;
2377 	int err;
2378 
2379 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2380 	if (err < 0)
2381 		return err;
2382 
2383 	ndm = nlmsg_data(nlh);
2384 	if (ndm->ndm_ifindex == 0) {
2385 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2386 		return -EINVAL;
2387 	}
2388 
2389 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2390 	if (dev == NULL) {
2391 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2392 		return -ENODEV;
2393 	}
2394 
2395 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2396 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2397 		return -EINVAL;
2398 	}
2399 
2400 	addr = nla_data(tb[NDA_LLADDR]);
2401 
2402 	err = -EOPNOTSUPP;
2403 
2404 	/* Support fdb on master device the net/bridge default case */
2405 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2406 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2407 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2408 		const struct net_device_ops *ops = br_dev->netdev_ops;
2409 
2410 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2411 		if (err)
2412 			goto out;
2413 		else
2414 			ndm->ndm_flags &= ~NTF_MASTER;
2415 	}
2416 
2417 	/* Embedded bridge, macvlan, and any other device support */
2418 	if ((ndm->ndm_flags & NTF_SELF)) {
2419 		if (dev->netdev_ops->ndo_fdb_add)
2420 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2421 							   nlh->nlmsg_flags);
2422 		else
2423 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2424 					       nlh->nlmsg_flags);
2425 
2426 		if (!err) {
2427 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2428 			ndm->ndm_flags &= ~NTF_SELF;
2429 		}
2430 	}
2431 out:
2432 	return err;
2433 }
2434 
2435 /**
2436  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2437  */
ndo_dflt_fdb_del(struct ndmsg * ndm,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr)2438 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2439 		     struct nlattr *tb[],
2440 		     struct net_device *dev,
2441 		     const unsigned char *addr)
2442 {
2443 	int err = -EINVAL;
2444 
2445 	/* If aging addresses are supported device will need to
2446 	 * implement its own handler for this.
2447 	 */
2448 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2449 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2450 		return err;
2451 	}
2452 
2453 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2454 		err = dev_uc_del(dev, addr);
2455 	else if (is_multicast_ether_addr(addr))
2456 		err = dev_mc_del(dev, addr);
2457 
2458 	return err;
2459 }
2460 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2461 
rtnl_fdb_del(struct sk_buff * skb,struct nlmsghdr * nlh)2462 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2463 {
2464 	struct net *net = sock_net(skb->sk);
2465 	struct ndmsg *ndm;
2466 	struct nlattr *tb[NDA_MAX+1];
2467 	struct net_device *dev;
2468 	int err = -EINVAL;
2469 	__u8 *addr;
2470 
2471 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2472 		return -EPERM;
2473 
2474 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2475 	if (err < 0)
2476 		return err;
2477 
2478 	ndm = nlmsg_data(nlh);
2479 	if (ndm->ndm_ifindex == 0) {
2480 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2481 		return -EINVAL;
2482 	}
2483 
2484 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2485 	if (dev == NULL) {
2486 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2487 		return -ENODEV;
2488 	}
2489 
2490 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2491 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2492 		return -EINVAL;
2493 	}
2494 
2495 	addr = nla_data(tb[NDA_LLADDR]);
2496 
2497 	err = -EOPNOTSUPP;
2498 
2499 	/* Support fdb on master device the net/bridge default case */
2500 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2501 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2502 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2503 		const struct net_device_ops *ops = br_dev->netdev_ops;
2504 
2505 		if (ops->ndo_fdb_del)
2506 			err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2507 
2508 		if (err)
2509 			goto out;
2510 		else
2511 			ndm->ndm_flags &= ~NTF_MASTER;
2512 	}
2513 
2514 	/* Embedded bridge, macvlan, and any other device support */
2515 	if (ndm->ndm_flags & NTF_SELF) {
2516 		if (dev->netdev_ops->ndo_fdb_del)
2517 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2518 		else
2519 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2520 
2521 		if (!err) {
2522 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2523 			ndm->ndm_flags &= ~NTF_SELF;
2524 		}
2525 	}
2526 out:
2527 	return err;
2528 }
2529 
nlmsg_populate_fdb(struct sk_buff * skb,struct netlink_callback * cb,struct net_device * dev,int * idx,struct netdev_hw_addr_list * list)2530 static int nlmsg_populate_fdb(struct sk_buff *skb,
2531 			      struct netlink_callback *cb,
2532 			      struct net_device *dev,
2533 			      int *idx,
2534 			      struct netdev_hw_addr_list *list)
2535 {
2536 	struct netdev_hw_addr *ha;
2537 	int err;
2538 	u32 portid, seq;
2539 
2540 	portid = NETLINK_CB(cb->skb).portid;
2541 	seq = cb->nlh->nlmsg_seq;
2542 
2543 	list_for_each_entry(ha, &list->list, list) {
2544 		if (*idx < cb->args[0])
2545 			goto skip;
2546 
2547 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2548 					      portid, seq,
2549 					      RTM_NEWNEIGH, NTF_SELF,
2550 					      NLM_F_MULTI);
2551 		if (err < 0)
2552 			return err;
2553 skip:
2554 		*idx += 1;
2555 	}
2556 	return 0;
2557 }
2558 
2559 /**
2560  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2561  * @nlh: netlink message header
2562  * @dev: netdevice
2563  *
2564  * Default netdevice operation to dump the existing unicast address list.
2565  * Returns number of addresses from list put in skb.
2566  */
ndo_dflt_fdb_dump(struct sk_buff * skb,struct netlink_callback * cb,struct net_device * dev,struct net_device * filter_dev,int idx)2567 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2568 		      struct netlink_callback *cb,
2569 		      struct net_device *dev,
2570 		      struct net_device *filter_dev,
2571 		      int idx)
2572 {
2573 	int err;
2574 
2575 	netif_addr_lock_bh(dev);
2576 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2577 	if (err)
2578 		goto out;
2579 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2580 out:
2581 	netif_addr_unlock_bh(dev);
2582 	return idx;
2583 }
2584 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2585 
rtnl_fdb_dump(struct sk_buff * skb,struct netlink_callback * cb)2586 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2587 {
2588 	struct net_device *dev;
2589 	struct nlattr *tb[IFLA_MAX+1];
2590 	struct net_device *bdev = NULL;
2591 	struct net_device *br_dev = NULL;
2592 	const struct net_device_ops *ops = NULL;
2593 	const struct net_device_ops *cops = NULL;
2594 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2595 	struct net *net = sock_net(skb->sk);
2596 	int brport_idx = 0;
2597 	int br_idx = 0;
2598 	int idx = 0;
2599 
2600 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2601 			ifla_policy) == 0) {
2602 		if (tb[IFLA_MASTER])
2603 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
2604 	}
2605 
2606 	brport_idx = ifm->ifi_index;
2607 
2608 	if (br_idx) {
2609 		br_dev = __dev_get_by_index(net, br_idx);
2610 		if (!br_dev)
2611 			return -ENODEV;
2612 
2613 		ops = br_dev->netdev_ops;
2614 		bdev = br_dev;
2615 	}
2616 
2617 	for_each_netdev(net, dev) {
2618 		if (brport_idx && (dev->ifindex != brport_idx))
2619 			continue;
2620 
2621 		if (!br_idx) { /* user did not specify a specific bridge */
2622 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
2623 				br_dev = netdev_master_upper_dev_get(dev);
2624 				cops = br_dev->netdev_ops;
2625 			}
2626 
2627 			bdev = dev;
2628 		} else {
2629 			if (dev != br_dev &&
2630 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
2631 				continue;
2632 
2633 			if (br_dev != netdev_master_upper_dev_get(dev) &&
2634 			    !(dev->priv_flags & IFF_EBRIDGE))
2635 				continue;
2636 
2637 			bdev = br_dev;
2638 			cops = ops;
2639 		}
2640 
2641 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2642 			if (cops && cops->ndo_fdb_dump)
2643 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2644 							 idx);
2645 		}
2646 
2647 		idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2648 		if (dev->netdev_ops->ndo_fdb_dump)
2649 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, bdev, dev,
2650 							    idx);
2651 
2652 		cops = NULL;
2653 	}
2654 
2655 	cb->args[0] = idx;
2656 	return skb->len;
2657 }
2658 
ndo_dflt_bridge_getlink(struct sk_buff * skb,u32 pid,u32 seq,struct net_device * dev,u16 mode)2659 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2660 			    struct net_device *dev, u16 mode)
2661 {
2662 	struct nlmsghdr *nlh;
2663 	struct ifinfomsg *ifm;
2664 	struct nlattr *br_afspec;
2665 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2666 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2667 
2668 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2669 	if (nlh == NULL)
2670 		return -EMSGSIZE;
2671 
2672 	ifm = nlmsg_data(nlh);
2673 	ifm->ifi_family = AF_BRIDGE;
2674 	ifm->__ifi_pad = 0;
2675 	ifm->ifi_type = dev->type;
2676 	ifm->ifi_index = dev->ifindex;
2677 	ifm->ifi_flags = dev_get_flags(dev);
2678 	ifm->ifi_change = 0;
2679 
2680 
2681 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2682 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2683 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2684 	    (br_dev &&
2685 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2686 	    (dev->addr_len &&
2687 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2688 	    (dev->ifindex != dev->iflink &&
2689 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2690 		goto nla_put_failure;
2691 
2692 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2693 	if (!br_afspec)
2694 		goto nla_put_failure;
2695 
2696 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2697 	    nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2698 		nla_nest_cancel(skb, br_afspec);
2699 		goto nla_put_failure;
2700 	}
2701 	nla_nest_end(skb, br_afspec);
2702 
2703 	return nlmsg_end(skb, nlh);
2704 nla_put_failure:
2705 	nlmsg_cancel(skb, nlh);
2706 	return -EMSGSIZE;
2707 }
2708 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2709 
rtnl_bridge_getlink(struct sk_buff * skb,struct netlink_callback * cb)2710 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2711 {
2712 	struct net *net = sock_net(skb->sk);
2713 	struct net_device *dev;
2714 	int idx = 0;
2715 	u32 portid = NETLINK_CB(cb->skb).portid;
2716 	u32 seq = cb->nlh->nlmsg_seq;
2717 	u32 filter_mask = 0;
2718 
2719 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2720 		struct nlattr *extfilt;
2721 
2722 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2723 					  IFLA_EXT_MASK);
2724 		if (extfilt) {
2725 			if (nla_len(extfilt) < sizeof(filter_mask))
2726 				return -EINVAL;
2727 
2728 			filter_mask = nla_get_u32(extfilt);
2729 		}
2730 	}
2731 
2732 	rcu_read_lock();
2733 	for_each_netdev_rcu(net, dev) {
2734 		const struct net_device_ops *ops = dev->netdev_ops;
2735 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2736 
2737 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2738 			if (idx >= cb->args[0] &&
2739 			    br_dev->netdev_ops->ndo_bridge_getlink(
2740 				    skb, portid, seq, dev, filter_mask) < 0)
2741 				break;
2742 			idx++;
2743 		}
2744 
2745 		if (ops->ndo_bridge_getlink) {
2746 			if (idx >= cb->args[0] &&
2747 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2748 						    filter_mask) < 0)
2749 				break;
2750 			idx++;
2751 		}
2752 	}
2753 	rcu_read_unlock();
2754 	cb->args[0] = idx;
2755 
2756 	return skb->len;
2757 }
2758 
bridge_nlmsg_size(void)2759 static inline size_t bridge_nlmsg_size(void)
2760 {
2761 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2762 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2763 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2764 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2765 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2766 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2767 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2768 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2769 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2770 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2771 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2772 }
2773 
rtnl_bridge_notify(struct net_device * dev,u16 flags)2774 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2775 {
2776 	struct net *net = dev_net(dev);
2777 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2778 	struct sk_buff *skb;
2779 	int err = -EOPNOTSUPP;
2780 
2781 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2782 	if (!skb) {
2783 		err = -ENOMEM;
2784 		goto errout;
2785 	}
2786 
2787 	if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2788 	    br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2789 		err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2790 		if (err < 0)
2791 			goto errout;
2792 	}
2793 
2794 	if ((flags & BRIDGE_FLAGS_SELF) &&
2795 	    dev->netdev_ops->ndo_bridge_getlink) {
2796 		err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2797 		if (err < 0)
2798 			goto errout;
2799 	}
2800 
2801 	if (!skb->len)
2802 		goto errout;
2803 
2804 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2805 	return 0;
2806 errout:
2807 	WARN_ON(err == -EMSGSIZE);
2808 	kfree_skb(skb);
2809 	if (err)
2810 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2811 	return err;
2812 }
2813 
rtnl_bridge_setlink(struct sk_buff * skb,struct nlmsghdr * nlh)2814 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2815 {
2816 	struct net *net = sock_net(skb->sk);
2817 	struct ifinfomsg *ifm;
2818 	struct net_device *dev;
2819 	struct nlattr *br_spec, *attr = NULL;
2820 	int rem, err = -EOPNOTSUPP;
2821 	u16 oflags, flags = 0;
2822 	bool have_flags = false;
2823 
2824 	if (nlmsg_len(nlh) < sizeof(*ifm))
2825 		return -EINVAL;
2826 
2827 	ifm = nlmsg_data(nlh);
2828 	if (ifm->ifi_family != AF_BRIDGE)
2829 		return -EPFNOSUPPORT;
2830 
2831 	dev = __dev_get_by_index(net, ifm->ifi_index);
2832 	if (!dev) {
2833 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2834 		return -ENODEV;
2835 	}
2836 
2837 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2838 	if (br_spec) {
2839 		nla_for_each_nested(attr, br_spec, rem) {
2840 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2841 				if (nla_len(attr) < sizeof(flags))
2842 					return -EINVAL;
2843 
2844 				have_flags = true;
2845 				flags = nla_get_u16(attr);
2846 				break;
2847 			}
2848 		}
2849 	}
2850 
2851 	oflags = flags;
2852 
2853 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2854 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2855 
2856 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2857 			err = -EOPNOTSUPP;
2858 			goto out;
2859 		}
2860 
2861 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2862 		if (err)
2863 			goto out;
2864 
2865 		flags &= ~BRIDGE_FLAGS_MASTER;
2866 	}
2867 
2868 	if ((flags & BRIDGE_FLAGS_SELF)) {
2869 		if (!dev->netdev_ops->ndo_bridge_setlink)
2870 			err = -EOPNOTSUPP;
2871 		else
2872 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2873 
2874 		if (!err)
2875 			flags &= ~BRIDGE_FLAGS_SELF;
2876 	}
2877 
2878 	if (have_flags)
2879 		memcpy(nla_data(attr), &flags, sizeof(flags));
2880 	/* Generate event to notify upper layer of bridge change */
2881 	if (!err)
2882 		err = rtnl_bridge_notify(dev, oflags);
2883 out:
2884 	return err;
2885 }
2886 
rtnl_bridge_dellink(struct sk_buff * skb,struct nlmsghdr * nlh)2887 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2888 {
2889 	struct net *net = sock_net(skb->sk);
2890 	struct ifinfomsg *ifm;
2891 	struct net_device *dev;
2892 	struct nlattr *br_spec, *attr = NULL;
2893 	int rem, err = -EOPNOTSUPP;
2894 	u16 oflags, flags = 0;
2895 	bool have_flags = false;
2896 
2897 	if (nlmsg_len(nlh) < sizeof(*ifm))
2898 		return -EINVAL;
2899 
2900 	ifm = nlmsg_data(nlh);
2901 	if (ifm->ifi_family != AF_BRIDGE)
2902 		return -EPFNOSUPPORT;
2903 
2904 	dev = __dev_get_by_index(net, ifm->ifi_index);
2905 	if (!dev) {
2906 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2907 		return -ENODEV;
2908 	}
2909 
2910 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2911 	if (br_spec) {
2912 		nla_for_each_nested(attr, br_spec, rem) {
2913 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2914 				if (nla_len(attr) < sizeof(flags))
2915 					return -EINVAL;
2916 
2917 				have_flags = true;
2918 				flags = nla_get_u16(attr);
2919 				break;
2920 			}
2921 		}
2922 	}
2923 
2924 	oflags = flags;
2925 
2926 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2927 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2928 
2929 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2930 			err = -EOPNOTSUPP;
2931 			goto out;
2932 		}
2933 
2934 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2935 		if (err)
2936 			goto out;
2937 
2938 		flags &= ~BRIDGE_FLAGS_MASTER;
2939 	}
2940 
2941 	if ((flags & BRIDGE_FLAGS_SELF)) {
2942 		if (!dev->netdev_ops->ndo_bridge_dellink)
2943 			err = -EOPNOTSUPP;
2944 		else
2945 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2946 
2947 		if (!err)
2948 			flags &= ~BRIDGE_FLAGS_SELF;
2949 	}
2950 
2951 	if (have_flags)
2952 		memcpy(nla_data(attr), &flags, sizeof(flags));
2953 	/* Generate event to notify upper layer of bridge change */
2954 	if (!err)
2955 		err = rtnl_bridge_notify(dev, oflags);
2956 out:
2957 	return err;
2958 }
2959 
2960 /* Process one rtnetlink message. */
2961 
rtnetlink_rcv_msg(struct sk_buff * skb,struct nlmsghdr * nlh)2962 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2963 {
2964 	struct net *net = sock_net(skb->sk);
2965 	rtnl_doit_func doit;
2966 	int sz_idx, kind;
2967 	int family;
2968 	int type;
2969 	int err;
2970 
2971 	type = nlh->nlmsg_type;
2972 	if (type > RTM_MAX)
2973 		return -EOPNOTSUPP;
2974 
2975 	type -= RTM_BASE;
2976 
2977 	/* All the messages must have at least 1 byte length */
2978 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2979 		return 0;
2980 
2981 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2982 	sz_idx = type>>2;
2983 	kind = type&3;
2984 
2985 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2986 		return -EPERM;
2987 
2988 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2989 		struct sock *rtnl;
2990 		rtnl_dumpit_func dumpit;
2991 		rtnl_calcit_func calcit;
2992 		u16 min_dump_alloc = 0;
2993 
2994 		dumpit = rtnl_get_dumpit(family, type);
2995 		if (dumpit == NULL)
2996 			return -EOPNOTSUPP;
2997 		calcit = rtnl_get_calcit(family, type);
2998 		if (calcit)
2999 			min_dump_alloc = calcit(skb, nlh);
3000 
3001 		__rtnl_unlock();
3002 		rtnl = net->rtnl;
3003 		{
3004 			struct netlink_dump_control c = {
3005 				.dump		= dumpit,
3006 				.min_dump_alloc	= min_dump_alloc,
3007 			};
3008 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3009 		}
3010 		rtnl_lock();
3011 		return err;
3012 	}
3013 
3014 	doit = rtnl_get_doit(family, type);
3015 	if (doit == NULL)
3016 		return -EOPNOTSUPP;
3017 
3018 	return doit(skb, nlh);
3019 }
3020 
rtnetlink_rcv(struct sk_buff * skb)3021 static void rtnetlink_rcv(struct sk_buff *skb)
3022 {
3023 	rtnl_lock();
3024 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3025 	rtnl_unlock();
3026 }
3027 
rtnetlink_event(struct notifier_block * this,unsigned long event,void * ptr)3028 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3029 {
3030 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3031 
3032 	switch (event) {
3033 	case NETDEV_UP:
3034 	case NETDEV_DOWN:
3035 	case NETDEV_PRE_UP:
3036 	case NETDEV_POST_INIT:
3037 	case NETDEV_REGISTER:
3038 	case NETDEV_CHANGE:
3039 	case NETDEV_PRE_TYPE_CHANGE:
3040 	case NETDEV_GOING_DOWN:
3041 	case NETDEV_UNREGISTER:
3042 	case NETDEV_UNREGISTER_FINAL:
3043 	case NETDEV_RELEASE:
3044 	case NETDEV_JOIN:
3045 		break;
3046 	default:
3047 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3048 		break;
3049 	}
3050 	return NOTIFY_DONE;
3051 }
3052 
3053 static struct notifier_block rtnetlink_dev_notifier = {
3054 	.notifier_call	= rtnetlink_event,
3055 };
3056 
3057 
rtnetlink_net_init(struct net * net)3058 static int __net_init rtnetlink_net_init(struct net *net)
3059 {
3060 	struct sock *sk;
3061 	struct netlink_kernel_cfg cfg = {
3062 		.groups		= RTNLGRP_MAX,
3063 		.input		= rtnetlink_rcv,
3064 		.cb_mutex	= &rtnl_mutex,
3065 		.flags		= NL_CFG_F_NONROOT_RECV,
3066 	};
3067 
3068 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3069 	if (!sk)
3070 		return -ENOMEM;
3071 	net->rtnl = sk;
3072 	return 0;
3073 }
3074 
rtnetlink_net_exit(struct net * net)3075 static void __net_exit rtnetlink_net_exit(struct net *net)
3076 {
3077 	netlink_kernel_release(net->rtnl);
3078 	net->rtnl = NULL;
3079 }
3080 
3081 static struct pernet_operations rtnetlink_net_ops = {
3082 	.init = rtnetlink_net_init,
3083 	.exit = rtnetlink_net_exit,
3084 };
3085 
rtnetlink_init(void)3086 void __init rtnetlink_init(void)
3087 {
3088 	if (register_pernet_subsys(&rtnetlink_net_ops))
3089 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3090 
3091 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3092 
3093 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3094 		      rtnl_dump_ifinfo, rtnl_calcit);
3095 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3096 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3097 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3098 
3099 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3100 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3101 
3102 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3103 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3104 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3105 
3106 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3107 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3108 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3109 }
3110 
3111