xref: /openbmc/linux/net/core/rtnetlink.c (revision 84d517f3)
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 
65 void rtnl_lock(void)
66 {
67 	mutex_lock(&rtnl_mutex);
68 }
69 EXPORT_SYMBOL(rtnl_lock);
70 
71 void __rtnl_unlock(void)
72 {
73 	mutex_unlock(&rtnl_mutex);
74 }
75 
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 
83 int rtnl_trylock(void)
84 {
85 	return mutex_trylock(&rtnl_mutex);
86 }
87 EXPORT_SYMBOL(rtnl_trylock);
88 
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
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 
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 
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 
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 
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  */
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  */
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  */
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  */
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 
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  */
297 int __rtnl_link_register(struct rtnl_link_ops *ops)
298 {
299 	if (rtnl_link_ops_get(ops->kind))
300 		return -EEXIST;
301 
302 	if (!ops->dellink)
303 		ops->dellink = unregister_netdevice_queue;
304 
305 	list_add_tail(&ops->list, &link_ops);
306 	return 0;
307 }
308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
309 
310 /**
311  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312  * @ops: struct rtnl_link_ops * to register
313  *
314  * Returns 0 on success or a negative error code.
315  */
316 int rtnl_link_register(struct rtnl_link_ops *ops)
317 {
318 	int err;
319 
320 	rtnl_lock();
321 	err = __rtnl_link_register(ops);
322 	rtnl_unlock();
323 	return err;
324 }
325 EXPORT_SYMBOL_GPL(rtnl_link_register);
326 
327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
328 {
329 	struct net_device *dev;
330 	LIST_HEAD(list_kill);
331 
332 	for_each_netdev(net, dev) {
333 		if (dev->rtnl_link_ops == ops)
334 			ops->dellink(dev, &list_kill);
335 	}
336 	unregister_netdevice_many(&list_kill);
337 }
338 
339 /**
340  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341  * @ops: struct rtnl_link_ops * to unregister
342  *
343  * The caller must hold the rtnl_mutex.
344  */
345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
346 {
347 	struct net *net;
348 
349 	for_each_net(net) {
350 		__rtnl_kill_links(net, ops);
351 	}
352 	list_del(&ops->list);
353 }
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
355 
356 /* Return with the rtnl_lock held when there are no network
357  * devices unregistering in any network namespace.
358  */
359 static void rtnl_lock_unregistering_all(void)
360 {
361 	struct net *net;
362 	bool unregistering;
363 	DEFINE_WAIT(wait);
364 
365 	for (;;) {
366 		prepare_to_wait(&netdev_unregistering_wq, &wait,
367 				TASK_UNINTERRUPTIBLE);
368 		unregistering = false;
369 		rtnl_lock();
370 		for_each_net(net) {
371 			if (net->dev_unreg_count > 0) {
372 				unregistering = true;
373 				break;
374 			}
375 		}
376 		if (!unregistering)
377 			break;
378 		__rtnl_unlock();
379 		schedule();
380 	}
381 	finish_wait(&netdev_unregistering_wq, &wait);
382 }
383 
384 /**
385  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
386  * @ops: struct rtnl_link_ops * to unregister
387  */
388 void rtnl_link_unregister(struct rtnl_link_ops *ops)
389 {
390 	/* Close the race with cleanup_net() */
391 	mutex_lock(&net_mutex);
392 	rtnl_lock_unregistering_all();
393 	__rtnl_link_unregister(ops);
394 	rtnl_unlock();
395 	mutex_unlock(&net_mutex);
396 }
397 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
398 
399 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
400 {
401 	struct net_device *master_dev;
402 	const struct rtnl_link_ops *ops;
403 
404 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
405 	if (!master_dev)
406 		return 0;
407 	ops = master_dev->rtnl_link_ops;
408 	if (!ops || !ops->get_slave_size)
409 		return 0;
410 	/* IFLA_INFO_SLAVE_DATA + nested data */
411 	return nla_total_size(sizeof(struct nlattr)) +
412 	       ops->get_slave_size(master_dev, dev);
413 }
414 
415 static size_t rtnl_link_get_size(const struct net_device *dev)
416 {
417 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
418 	size_t size;
419 
420 	if (!ops)
421 		return 0;
422 
423 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
424 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
425 
426 	if (ops->get_size)
427 		/* IFLA_INFO_DATA + nested data */
428 		size += nla_total_size(sizeof(struct nlattr)) +
429 			ops->get_size(dev);
430 
431 	if (ops->get_xstats_size)
432 		/* IFLA_INFO_XSTATS */
433 		size += nla_total_size(ops->get_xstats_size(dev));
434 
435 	size += rtnl_link_get_slave_info_data_size(dev);
436 
437 	return size;
438 }
439 
440 static LIST_HEAD(rtnl_af_ops);
441 
442 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
443 {
444 	const struct rtnl_af_ops *ops;
445 
446 	list_for_each_entry(ops, &rtnl_af_ops, list) {
447 		if (ops->family == family)
448 			return ops;
449 	}
450 
451 	return NULL;
452 }
453 
454 /**
455  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
456  * @ops: struct rtnl_af_ops * to register
457  *
458  * Returns 0 on success or a negative error code.
459  */
460 void rtnl_af_register(struct rtnl_af_ops *ops)
461 {
462 	rtnl_lock();
463 	list_add_tail(&ops->list, &rtnl_af_ops);
464 	rtnl_unlock();
465 }
466 EXPORT_SYMBOL_GPL(rtnl_af_register);
467 
468 /**
469  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
470  * @ops: struct rtnl_af_ops * to unregister
471  *
472  * The caller must hold the rtnl_mutex.
473  */
474 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
475 {
476 	list_del(&ops->list);
477 }
478 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
479 
480 /**
481  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
482  * @ops: struct rtnl_af_ops * to unregister
483  */
484 void rtnl_af_unregister(struct rtnl_af_ops *ops)
485 {
486 	rtnl_lock();
487 	__rtnl_af_unregister(ops);
488 	rtnl_unlock();
489 }
490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
491 
492 static size_t rtnl_link_get_af_size(const struct net_device *dev)
493 {
494 	struct rtnl_af_ops *af_ops;
495 	size_t size;
496 
497 	/* IFLA_AF_SPEC */
498 	size = nla_total_size(sizeof(struct nlattr));
499 
500 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
501 		if (af_ops->get_link_af_size) {
502 			/* AF_* + nested data */
503 			size += nla_total_size(sizeof(struct nlattr)) +
504 				af_ops->get_link_af_size(dev);
505 		}
506 	}
507 
508 	return size;
509 }
510 
511 static bool rtnl_have_link_slave_info(const struct net_device *dev)
512 {
513 	struct net_device *master_dev;
514 
515 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
516 	if (master_dev && master_dev->rtnl_link_ops)
517 		return true;
518 	return false;
519 }
520 
521 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
522 				     const struct net_device *dev)
523 {
524 	struct net_device *master_dev;
525 	const struct rtnl_link_ops *ops;
526 	struct nlattr *slave_data;
527 	int err;
528 
529 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
530 	if (!master_dev)
531 		return 0;
532 	ops = master_dev->rtnl_link_ops;
533 	if (!ops)
534 		return 0;
535 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
536 		return -EMSGSIZE;
537 	if (ops->fill_slave_info) {
538 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
539 		if (!slave_data)
540 			return -EMSGSIZE;
541 		err = ops->fill_slave_info(skb, master_dev, dev);
542 		if (err < 0)
543 			goto err_cancel_slave_data;
544 		nla_nest_end(skb, slave_data);
545 	}
546 	return 0;
547 
548 err_cancel_slave_data:
549 	nla_nest_cancel(skb, slave_data);
550 	return err;
551 }
552 
553 static int rtnl_link_info_fill(struct sk_buff *skb,
554 			       const struct net_device *dev)
555 {
556 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
557 	struct nlattr *data;
558 	int err;
559 
560 	if (!ops)
561 		return 0;
562 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
563 		return -EMSGSIZE;
564 	if (ops->fill_xstats) {
565 		err = ops->fill_xstats(skb, dev);
566 		if (err < 0)
567 			return err;
568 	}
569 	if (ops->fill_info) {
570 		data = nla_nest_start(skb, IFLA_INFO_DATA);
571 		if (data == NULL)
572 			return -EMSGSIZE;
573 		err = ops->fill_info(skb, dev);
574 		if (err < 0)
575 			goto err_cancel_data;
576 		nla_nest_end(skb, data);
577 	}
578 	return 0;
579 
580 err_cancel_data:
581 	nla_nest_cancel(skb, data);
582 	return err;
583 }
584 
585 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
586 {
587 	struct nlattr *linkinfo;
588 	int err = -EMSGSIZE;
589 
590 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
591 	if (linkinfo == NULL)
592 		goto out;
593 
594 	err = rtnl_link_info_fill(skb, dev);
595 	if (err < 0)
596 		goto err_cancel_link;
597 
598 	err = rtnl_link_slave_info_fill(skb, dev);
599 	if (err < 0)
600 		goto err_cancel_link;
601 
602 	nla_nest_end(skb, linkinfo);
603 	return 0;
604 
605 err_cancel_link:
606 	nla_nest_cancel(skb, linkinfo);
607 out:
608 	return err;
609 }
610 
611 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
612 {
613 	struct sock *rtnl = net->rtnl;
614 	int err = 0;
615 
616 	NETLINK_CB(skb).dst_group = group;
617 	if (echo)
618 		atomic_inc(&skb->users);
619 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
620 	if (echo)
621 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
622 	return err;
623 }
624 
625 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
626 {
627 	struct sock *rtnl = net->rtnl;
628 
629 	return nlmsg_unicast(rtnl, skb, pid);
630 }
631 EXPORT_SYMBOL(rtnl_unicast);
632 
633 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
634 		 struct nlmsghdr *nlh, gfp_t flags)
635 {
636 	struct sock *rtnl = net->rtnl;
637 	int report = 0;
638 
639 	if (nlh)
640 		report = nlmsg_report(nlh);
641 
642 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
643 }
644 EXPORT_SYMBOL(rtnl_notify);
645 
646 void rtnl_set_sk_err(struct net *net, u32 group, int error)
647 {
648 	struct sock *rtnl = net->rtnl;
649 
650 	netlink_set_err(rtnl, 0, group, error);
651 }
652 EXPORT_SYMBOL(rtnl_set_sk_err);
653 
654 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
655 {
656 	struct nlattr *mx;
657 	int i, valid = 0;
658 
659 	mx = nla_nest_start(skb, RTA_METRICS);
660 	if (mx == NULL)
661 		return -ENOBUFS;
662 
663 	for (i = 0; i < RTAX_MAX; i++) {
664 		if (metrics[i]) {
665 			valid++;
666 			if (nla_put_u32(skb, i+1, metrics[i]))
667 				goto nla_put_failure;
668 		}
669 	}
670 
671 	if (!valid) {
672 		nla_nest_cancel(skb, mx);
673 		return 0;
674 	}
675 
676 	return nla_nest_end(skb, mx);
677 
678 nla_put_failure:
679 	nla_nest_cancel(skb, mx);
680 	return -EMSGSIZE;
681 }
682 EXPORT_SYMBOL(rtnetlink_put_metrics);
683 
684 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
685 		       long expires, u32 error)
686 {
687 	struct rta_cacheinfo ci = {
688 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
689 		.rta_used = dst->__use,
690 		.rta_clntref = atomic_read(&(dst->__refcnt)),
691 		.rta_error = error,
692 		.rta_id =  id,
693 	};
694 
695 	if (expires) {
696 		unsigned long clock;
697 
698 		clock = jiffies_to_clock_t(abs(expires));
699 		clock = min_t(unsigned long, clock, INT_MAX);
700 		ci.rta_expires = (expires > 0) ? clock : -clock;
701 	}
702 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
703 }
704 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
705 
706 static void set_operstate(struct net_device *dev, unsigned char transition)
707 {
708 	unsigned char operstate = dev->operstate;
709 
710 	switch (transition) {
711 	case IF_OPER_UP:
712 		if ((operstate == IF_OPER_DORMANT ||
713 		     operstate == IF_OPER_UNKNOWN) &&
714 		    !netif_dormant(dev))
715 			operstate = IF_OPER_UP;
716 		break;
717 
718 	case IF_OPER_DORMANT:
719 		if (operstate == IF_OPER_UP ||
720 		    operstate == IF_OPER_UNKNOWN)
721 			operstate = IF_OPER_DORMANT;
722 		break;
723 	}
724 
725 	if (dev->operstate != operstate) {
726 		write_lock_bh(&dev_base_lock);
727 		dev->operstate = operstate;
728 		write_unlock_bh(&dev_base_lock);
729 		netdev_state_change(dev);
730 	}
731 }
732 
733 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
734 {
735 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
736 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
737 }
738 
739 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
740 					   const struct ifinfomsg *ifm)
741 {
742 	unsigned int flags = ifm->ifi_flags;
743 
744 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
745 	if (ifm->ifi_change)
746 		flags = (flags & ifm->ifi_change) |
747 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
748 
749 	return flags;
750 }
751 
752 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
753 				 const struct rtnl_link_stats64 *b)
754 {
755 	a->rx_packets = b->rx_packets;
756 	a->tx_packets = b->tx_packets;
757 	a->rx_bytes = b->rx_bytes;
758 	a->tx_bytes = b->tx_bytes;
759 	a->rx_errors = b->rx_errors;
760 	a->tx_errors = b->tx_errors;
761 	a->rx_dropped = b->rx_dropped;
762 	a->tx_dropped = b->tx_dropped;
763 
764 	a->multicast = b->multicast;
765 	a->collisions = b->collisions;
766 
767 	a->rx_length_errors = b->rx_length_errors;
768 	a->rx_over_errors = b->rx_over_errors;
769 	a->rx_crc_errors = b->rx_crc_errors;
770 	a->rx_frame_errors = b->rx_frame_errors;
771 	a->rx_fifo_errors = b->rx_fifo_errors;
772 	a->rx_missed_errors = b->rx_missed_errors;
773 
774 	a->tx_aborted_errors = b->tx_aborted_errors;
775 	a->tx_carrier_errors = b->tx_carrier_errors;
776 	a->tx_fifo_errors = b->tx_fifo_errors;
777 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
778 	a->tx_window_errors = b->tx_window_errors;
779 
780 	a->rx_compressed = b->rx_compressed;
781 	a->tx_compressed = b->tx_compressed;
782 }
783 
784 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
785 {
786 	memcpy(v, b, sizeof(*b));
787 }
788 
789 /* All VF info */
790 static inline int rtnl_vfinfo_size(const struct net_device *dev,
791 				   u32 ext_filter_mask)
792 {
793 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
794 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
795 		int num_vfs = dev_num_vf(dev->dev.parent);
796 		size_t size = nla_total_size(sizeof(struct nlattr));
797 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
798 		size += num_vfs *
799 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
800 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
801 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
802 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
803 		return size;
804 	} else
805 		return 0;
806 }
807 
808 static size_t rtnl_port_size(const struct net_device *dev,
809 			     u32 ext_filter_mask)
810 {
811 	size_t port_size = nla_total_size(4)		/* PORT_VF */
812 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
813 		+ nla_total_size(sizeof(struct ifla_port_vsi))
814 							/* PORT_VSI_TYPE */
815 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
816 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
817 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
818 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
819 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
820 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
821 		+ port_size;
822 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
823 		+ port_size;
824 
825 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
826 	    !(ext_filter_mask & RTEXT_FILTER_VF))
827 		return 0;
828 	if (dev_num_vf(dev->dev.parent))
829 		return port_self_size + vf_ports_size +
830 			vf_port_size * dev_num_vf(dev->dev.parent);
831 	else
832 		return port_self_size;
833 }
834 
835 static noinline size_t if_nlmsg_size(const struct net_device *dev,
836 				     u32 ext_filter_mask)
837 {
838 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
839 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
840 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
841 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
842 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
843 	       + nla_total_size(sizeof(struct rtnl_link_stats))
844 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
845 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
846 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
847 	       + nla_total_size(4) /* IFLA_TXQLEN */
848 	       + nla_total_size(4) /* IFLA_WEIGHT */
849 	       + nla_total_size(4) /* IFLA_MTU */
850 	       + nla_total_size(4) /* IFLA_LINK */
851 	       + nla_total_size(4) /* IFLA_MASTER */
852 	       + nla_total_size(1) /* IFLA_CARRIER */
853 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
854 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
855 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
856 	       + nla_total_size(1) /* IFLA_OPERSTATE */
857 	       + nla_total_size(1) /* IFLA_LINKMODE */
858 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
859 	       + nla_total_size(ext_filter_mask
860 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
861 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
862 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
863 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
864 	       + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
865 	       + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
866 }
867 
868 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
869 {
870 	struct nlattr *vf_ports;
871 	struct nlattr *vf_port;
872 	int vf;
873 	int err;
874 
875 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
876 	if (!vf_ports)
877 		return -EMSGSIZE;
878 
879 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
880 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
881 		if (!vf_port)
882 			goto nla_put_failure;
883 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
884 			goto nla_put_failure;
885 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
886 		if (err == -EMSGSIZE)
887 			goto nla_put_failure;
888 		if (err) {
889 			nla_nest_cancel(skb, vf_port);
890 			continue;
891 		}
892 		nla_nest_end(skb, vf_port);
893 	}
894 
895 	nla_nest_end(skb, vf_ports);
896 
897 	return 0;
898 
899 nla_put_failure:
900 	nla_nest_cancel(skb, vf_ports);
901 	return -EMSGSIZE;
902 }
903 
904 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
905 {
906 	struct nlattr *port_self;
907 	int err;
908 
909 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
910 	if (!port_self)
911 		return -EMSGSIZE;
912 
913 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
914 	if (err) {
915 		nla_nest_cancel(skb, port_self);
916 		return (err == -EMSGSIZE) ? err : 0;
917 	}
918 
919 	nla_nest_end(skb, port_self);
920 
921 	return 0;
922 }
923 
924 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
925 			  u32 ext_filter_mask)
926 {
927 	int err;
928 
929 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
930 	    !(ext_filter_mask & RTEXT_FILTER_VF))
931 		return 0;
932 
933 	err = rtnl_port_self_fill(skb, dev);
934 	if (err)
935 		return err;
936 
937 	if (dev_num_vf(dev->dev.parent)) {
938 		err = rtnl_vf_ports_fill(skb, dev);
939 		if (err)
940 			return err;
941 	}
942 
943 	return 0;
944 }
945 
946 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
947 {
948 	int err;
949 	struct netdev_phys_port_id ppid;
950 
951 	err = dev_get_phys_port_id(dev, &ppid);
952 	if (err) {
953 		if (err == -EOPNOTSUPP)
954 			return 0;
955 		return err;
956 	}
957 
958 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
959 		return -EMSGSIZE;
960 
961 	return 0;
962 }
963 
964 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
965 			    int type, u32 pid, u32 seq, u32 change,
966 			    unsigned int flags, u32 ext_filter_mask)
967 {
968 	struct ifinfomsg *ifm;
969 	struct nlmsghdr *nlh;
970 	struct rtnl_link_stats64 temp;
971 	const struct rtnl_link_stats64 *stats;
972 	struct nlattr *attr, *af_spec;
973 	struct rtnl_af_ops *af_ops;
974 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
975 
976 	ASSERT_RTNL();
977 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
978 	if (nlh == NULL)
979 		return -EMSGSIZE;
980 
981 	ifm = nlmsg_data(nlh);
982 	ifm->ifi_family = AF_UNSPEC;
983 	ifm->__ifi_pad = 0;
984 	ifm->ifi_type = dev->type;
985 	ifm->ifi_index = dev->ifindex;
986 	ifm->ifi_flags = dev_get_flags(dev);
987 	ifm->ifi_change = change;
988 
989 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
990 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
991 	    nla_put_u8(skb, IFLA_OPERSTATE,
992 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
993 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
994 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
995 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
996 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
997 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
998 #ifdef CONFIG_RPS
999 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1000 #endif
1001 	    (dev->ifindex != dev->iflink &&
1002 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1003 	    (upper_dev &&
1004 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1005 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1006 	    (dev->qdisc &&
1007 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1008 	    (dev->ifalias &&
1009 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1010 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1011 			atomic_read(&dev->carrier_changes)))
1012 		goto nla_put_failure;
1013 
1014 	if (1) {
1015 		struct rtnl_link_ifmap map = {
1016 			.mem_start   = dev->mem_start,
1017 			.mem_end     = dev->mem_end,
1018 			.base_addr   = dev->base_addr,
1019 			.irq         = dev->irq,
1020 			.dma         = dev->dma,
1021 			.port        = dev->if_port,
1022 		};
1023 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1024 			goto nla_put_failure;
1025 	}
1026 
1027 	if (dev->addr_len) {
1028 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1029 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1030 			goto nla_put_failure;
1031 	}
1032 
1033 	if (rtnl_phys_port_id_fill(skb, dev))
1034 		goto nla_put_failure;
1035 
1036 	attr = nla_reserve(skb, IFLA_STATS,
1037 			sizeof(struct rtnl_link_stats));
1038 	if (attr == NULL)
1039 		goto nla_put_failure;
1040 
1041 	stats = dev_get_stats(dev, &temp);
1042 	copy_rtnl_link_stats(nla_data(attr), stats);
1043 
1044 	attr = nla_reserve(skb, IFLA_STATS64,
1045 			sizeof(struct rtnl_link_stats64));
1046 	if (attr == NULL)
1047 		goto nla_put_failure;
1048 	copy_rtnl_link_stats64(nla_data(attr), stats);
1049 
1050 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1051 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1052 		goto nla_put_failure;
1053 
1054 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1055 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
1056 		int i;
1057 
1058 		struct nlattr *vfinfo, *vf;
1059 		int num_vfs = dev_num_vf(dev->dev.parent);
1060 
1061 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1062 		if (!vfinfo)
1063 			goto nla_put_failure;
1064 		for (i = 0; i < num_vfs; i++) {
1065 			struct ifla_vf_info ivi;
1066 			struct ifla_vf_mac vf_mac;
1067 			struct ifla_vf_vlan vf_vlan;
1068 			struct ifla_vf_tx_rate vf_tx_rate;
1069 			struct ifla_vf_spoofchk vf_spoofchk;
1070 			struct ifla_vf_link_state vf_linkstate;
1071 
1072 			/*
1073 			 * Not all SR-IOV capable drivers support the
1074 			 * spoofcheck query.  Preset to -1 so the user
1075 			 * space tool can detect that the driver didn't
1076 			 * report anything.
1077 			 */
1078 			ivi.spoofchk = -1;
1079 			memset(ivi.mac, 0, sizeof(ivi.mac));
1080 			/* The default value for VF link state is "auto"
1081 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
1082 			 */
1083 			ivi.linkstate = 0;
1084 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1085 				break;
1086 			vf_mac.vf =
1087 				vf_vlan.vf =
1088 				vf_tx_rate.vf =
1089 				vf_spoofchk.vf =
1090 				vf_linkstate.vf = ivi.vf;
1091 
1092 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1093 			vf_vlan.vlan = ivi.vlan;
1094 			vf_vlan.qos = ivi.qos;
1095 			vf_tx_rate.rate = ivi.tx_rate;
1096 			vf_spoofchk.setting = ivi.spoofchk;
1097 			vf_linkstate.link_state = ivi.linkstate;
1098 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1099 			if (!vf) {
1100 				nla_nest_cancel(skb, vfinfo);
1101 				goto nla_put_failure;
1102 			}
1103 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1104 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1105 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1106 				    &vf_tx_rate) ||
1107 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1108 				    &vf_spoofchk) ||
1109 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1110 				    &vf_linkstate))
1111 				goto nla_put_failure;
1112 			nla_nest_end(skb, vf);
1113 		}
1114 		nla_nest_end(skb, vfinfo);
1115 	}
1116 
1117 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1118 		goto nla_put_failure;
1119 
1120 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1121 		if (rtnl_link_fill(skb, dev) < 0)
1122 			goto nla_put_failure;
1123 	}
1124 
1125 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1126 		goto nla_put_failure;
1127 
1128 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1129 		if (af_ops->fill_link_af) {
1130 			struct nlattr *af;
1131 			int err;
1132 
1133 			if (!(af = nla_nest_start(skb, af_ops->family)))
1134 				goto nla_put_failure;
1135 
1136 			err = af_ops->fill_link_af(skb, dev);
1137 
1138 			/*
1139 			 * Caller may return ENODATA to indicate that there
1140 			 * was no data to be dumped. This is not an error, it
1141 			 * means we should trim the attribute header and
1142 			 * continue.
1143 			 */
1144 			if (err == -ENODATA)
1145 				nla_nest_cancel(skb, af);
1146 			else if (err < 0)
1147 				goto nla_put_failure;
1148 
1149 			nla_nest_end(skb, af);
1150 		}
1151 	}
1152 
1153 	nla_nest_end(skb, af_spec);
1154 
1155 	return nlmsg_end(skb, nlh);
1156 
1157 nla_put_failure:
1158 	nlmsg_cancel(skb, nlh);
1159 	return -EMSGSIZE;
1160 }
1161 
1162 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1163 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1164 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1165 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1166 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1167 	[IFLA_MTU]		= { .type = NLA_U32 },
1168 	[IFLA_LINK]		= { .type = NLA_U32 },
1169 	[IFLA_MASTER]		= { .type = NLA_U32 },
1170 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1171 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1172 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1173 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1174 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1175 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1176 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1177 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1178 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1179 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1180 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1181 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1182 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1183 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1184 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1185 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1186 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1187 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1188 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1189 };
1190 
1191 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1192 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1193 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1194 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1195 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1196 };
1197 
1198 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1199 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1200 };
1201 
1202 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1203 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1204 				    .len = sizeof(struct ifla_vf_mac) },
1205 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1206 				    .len = sizeof(struct ifla_vf_vlan) },
1207 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1208 				    .len = sizeof(struct ifla_vf_tx_rate) },
1209 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1210 				    .len = sizeof(struct ifla_vf_spoofchk) },
1211 };
1212 
1213 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1214 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1215 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1216 				    .len = PORT_PROFILE_MAX },
1217 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1218 				    .len = sizeof(struct ifla_port_vsi)},
1219 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1220 				      .len = PORT_UUID_MAX },
1221 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1222 				    .len = PORT_UUID_MAX },
1223 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1224 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1225 };
1226 
1227 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1228 {
1229 	struct net *net = sock_net(skb->sk);
1230 	int h, s_h;
1231 	int idx = 0, s_idx;
1232 	struct net_device *dev;
1233 	struct hlist_head *head;
1234 	struct nlattr *tb[IFLA_MAX+1];
1235 	u32 ext_filter_mask = 0;
1236 	int err;
1237 
1238 	s_h = cb->args[0];
1239 	s_idx = cb->args[1];
1240 
1241 	rcu_read_lock();
1242 	cb->seq = net->dev_base_seq;
1243 
1244 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1245 			ifla_policy) >= 0) {
1246 
1247 		if (tb[IFLA_EXT_MASK])
1248 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1249 	}
1250 
1251 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1252 		idx = 0;
1253 		head = &net->dev_index_head[h];
1254 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
1255 			if (idx < s_idx)
1256 				goto cont;
1257 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1258 					       NETLINK_CB(cb->skb).portid,
1259 					       cb->nlh->nlmsg_seq, 0,
1260 					       NLM_F_MULTI,
1261 					       ext_filter_mask);
1262 			/* If we ran out of room on the first message,
1263 			 * we're in trouble
1264 			 */
1265 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1266 
1267 			if (err <= 0)
1268 				goto out;
1269 
1270 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1271 cont:
1272 			idx++;
1273 		}
1274 	}
1275 out:
1276 	rcu_read_unlock();
1277 	cb->args[1] = idx;
1278 	cb->args[0] = h;
1279 
1280 	return skb->len;
1281 }
1282 
1283 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1284 {
1285 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1286 }
1287 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1288 
1289 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1290 {
1291 	struct net *net;
1292 	/* Examine the link attributes and figure out which
1293 	 * network namespace we are talking about.
1294 	 */
1295 	if (tb[IFLA_NET_NS_PID])
1296 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1297 	else if (tb[IFLA_NET_NS_FD])
1298 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1299 	else
1300 		net = get_net(src_net);
1301 	return net;
1302 }
1303 EXPORT_SYMBOL(rtnl_link_get_net);
1304 
1305 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1306 {
1307 	if (dev) {
1308 		if (tb[IFLA_ADDRESS] &&
1309 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1310 			return -EINVAL;
1311 
1312 		if (tb[IFLA_BROADCAST] &&
1313 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1314 			return -EINVAL;
1315 	}
1316 
1317 	if (tb[IFLA_AF_SPEC]) {
1318 		struct nlattr *af;
1319 		int rem, err;
1320 
1321 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1322 			const struct rtnl_af_ops *af_ops;
1323 
1324 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1325 				return -EAFNOSUPPORT;
1326 
1327 			if (!af_ops->set_link_af)
1328 				return -EOPNOTSUPP;
1329 
1330 			if (af_ops->validate_link_af) {
1331 				err = af_ops->validate_link_af(dev, af);
1332 				if (err < 0)
1333 					return err;
1334 			}
1335 		}
1336 	}
1337 
1338 	return 0;
1339 }
1340 
1341 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1342 {
1343 	int rem, err = -EINVAL;
1344 	struct nlattr *vf;
1345 	const struct net_device_ops *ops = dev->netdev_ops;
1346 
1347 	nla_for_each_nested(vf, attr, rem) {
1348 		switch (nla_type(vf)) {
1349 		case IFLA_VF_MAC: {
1350 			struct ifla_vf_mac *ivm;
1351 			ivm = nla_data(vf);
1352 			err = -EOPNOTSUPP;
1353 			if (ops->ndo_set_vf_mac)
1354 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1355 							  ivm->mac);
1356 			break;
1357 		}
1358 		case IFLA_VF_VLAN: {
1359 			struct ifla_vf_vlan *ivv;
1360 			ivv = nla_data(vf);
1361 			err = -EOPNOTSUPP;
1362 			if (ops->ndo_set_vf_vlan)
1363 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1364 							   ivv->vlan,
1365 							   ivv->qos);
1366 			break;
1367 		}
1368 		case IFLA_VF_TX_RATE: {
1369 			struct ifla_vf_tx_rate *ivt;
1370 			ivt = nla_data(vf);
1371 			err = -EOPNOTSUPP;
1372 			if (ops->ndo_set_vf_tx_rate)
1373 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1374 							      ivt->rate);
1375 			break;
1376 		}
1377 		case IFLA_VF_SPOOFCHK: {
1378 			struct ifla_vf_spoofchk *ivs;
1379 			ivs = nla_data(vf);
1380 			err = -EOPNOTSUPP;
1381 			if (ops->ndo_set_vf_spoofchk)
1382 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1383 							       ivs->setting);
1384 			break;
1385 		}
1386 		case IFLA_VF_LINK_STATE: {
1387 			struct ifla_vf_link_state *ivl;
1388 			ivl = nla_data(vf);
1389 			err = -EOPNOTSUPP;
1390 			if (ops->ndo_set_vf_link_state)
1391 				err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1392 								 ivl->link_state);
1393 			break;
1394 		}
1395 		default:
1396 			err = -EINVAL;
1397 			break;
1398 		}
1399 		if (err)
1400 			break;
1401 	}
1402 	return err;
1403 }
1404 
1405 static int do_set_master(struct net_device *dev, int ifindex)
1406 {
1407 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1408 	const struct net_device_ops *ops;
1409 	int err;
1410 
1411 	if (upper_dev) {
1412 		if (upper_dev->ifindex == ifindex)
1413 			return 0;
1414 		ops = upper_dev->netdev_ops;
1415 		if (ops->ndo_del_slave) {
1416 			err = ops->ndo_del_slave(upper_dev, dev);
1417 			if (err)
1418 				return err;
1419 		} else {
1420 			return -EOPNOTSUPP;
1421 		}
1422 	}
1423 
1424 	if (ifindex) {
1425 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1426 		if (!upper_dev)
1427 			return -EINVAL;
1428 		ops = upper_dev->netdev_ops;
1429 		if (ops->ndo_add_slave) {
1430 			err = ops->ndo_add_slave(upper_dev, dev);
1431 			if (err)
1432 				return err;
1433 		} else {
1434 			return -EOPNOTSUPP;
1435 		}
1436 	}
1437 	return 0;
1438 }
1439 
1440 static int do_setlink(const struct sk_buff *skb,
1441 		      struct net_device *dev, struct ifinfomsg *ifm,
1442 		      struct nlattr **tb, char *ifname, int modified)
1443 {
1444 	const struct net_device_ops *ops = dev->netdev_ops;
1445 	int err;
1446 
1447 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1448 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1449 		if (IS_ERR(net)) {
1450 			err = PTR_ERR(net);
1451 			goto errout;
1452 		}
1453 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1454 			err = -EPERM;
1455 			goto errout;
1456 		}
1457 		err = dev_change_net_namespace(dev, net, ifname);
1458 		put_net(net);
1459 		if (err)
1460 			goto errout;
1461 		modified = 1;
1462 	}
1463 
1464 	if (tb[IFLA_MAP]) {
1465 		struct rtnl_link_ifmap *u_map;
1466 		struct ifmap k_map;
1467 
1468 		if (!ops->ndo_set_config) {
1469 			err = -EOPNOTSUPP;
1470 			goto errout;
1471 		}
1472 
1473 		if (!netif_device_present(dev)) {
1474 			err = -ENODEV;
1475 			goto errout;
1476 		}
1477 
1478 		u_map = nla_data(tb[IFLA_MAP]);
1479 		k_map.mem_start = (unsigned long) u_map->mem_start;
1480 		k_map.mem_end = (unsigned long) u_map->mem_end;
1481 		k_map.base_addr = (unsigned short) u_map->base_addr;
1482 		k_map.irq = (unsigned char) u_map->irq;
1483 		k_map.dma = (unsigned char) u_map->dma;
1484 		k_map.port = (unsigned char) u_map->port;
1485 
1486 		err = ops->ndo_set_config(dev, &k_map);
1487 		if (err < 0)
1488 			goto errout;
1489 
1490 		modified = 1;
1491 	}
1492 
1493 	if (tb[IFLA_ADDRESS]) {
1494 		struct sockaddr *sa;
1495 		int len;
1496 
1497 		len = sizeof(sa_family_t) + dev->addr_len;
1498 		sa = kmalloc(len, GFP_KERNEL);
1499 		if (!sa) {
1500 			err = -ENOMEM;
1501 			goto errout;
1502 		}
1503 		sa->sa_family = dev->type;
1504 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1505 		       dev->addr_len);
1506 		err = dev_set_mac_address(dev, sa);
1507 		kfree(sa);
1508 		if (err)
1509 			goto errout;
1510 		modified = 1;
1511 	}
1512 
1513 	if (tb[IFLA_MTU]) {
1514 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1515 		if (err < 0)
1516 			goto errout;
1517 		modified = 1;
1518 	}
1519 
1520 	if (tb[IFLA_GROUP]) {
1521 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1522 		modified = 1;
1523 	}
1524 
1525 	/*
1526 	 * Interface selected by interface index but interface
1527 	 * name provided implies that a name change has been
1528 	 * requested.
1529 	 */
1530 	if (ifm->ifi_index > 0 && ifname[0]) {
1531 		err = dev_change_name(dev, ifname);
1532 		if (err < 0)
1533 			goto errout;
1534 		modified = 1;
1535 	}
1536 
1537 	if (tb[IFLA_IFALIAS]) {
1538 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1539 				    nla_len(tb[IFLA_IFALIAS]));
1540 		if (err < 0)
1541 			goto errout;
1542 		modified = 1;
1543 	}
1544 
1545 	if (tb[IFLA_BROADCAST]) {
1546 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1547 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1548 	}
1549 
1550 	if (ifm->ifi_flags || ifm->ifi_change) {
1551 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1552 		if (err < 0)
1553 			goto errout;
1554 	}
1555 
1556 	if (tb[IFLA_MASTER]) {
1557 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1558 		if (err)
1559 			goto errout;
1560 		modified = 1;
1561 	}
1562 
1563 	if (tb[IFLA_CARRIER]) {
1564 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1565 		if (err)
1566 			goto errout;
1567 		modified = 1;
1568 	}
1569 
1570 	if (tb[IFLA_TXQLEN])
1571 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1572 
1573 	if (tb[IFLA_OPERSTATE])
1574 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1575 
1576 	if (tb[IFLA_LINKMODE]) {
1577 		write_lock_bh(&dev_base_lock);
1578 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1579 		write_unlock_bh(&dev_base_lock);
1580 	}
1581 
1582 	if (tb[IFLA_VFINFO_LIST]) {
1583 		struct nlattr *attr;
1584 		int rem;
1585 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1586 			if (nla_type(attr) != IFLA_VF_INFO) {
1587 				err = -EINVAL;
1588 				goto errout;
1589 			}
1590 			err = do_setvfinfo(dev, attr);
1591 			if (err < 0)
1592 				goto errout;
1593 			modified = 1;
1594 		}
1595 	}
1596 	err = 0;
1597 
1598 	if (tb[IFLA_VF_PORTS]) {
1599 		struct nlattr *port[IFLA_PORT_MAX+1];
1600 		struct nlattr *attr;
1601 		int vf;
1602 		int rem;
1603 
1604 		err = -EOPNOTSUPP;
1605 		if (!ops->ndo_set_vf_port)
1606 			goto errout;
1607 
1608 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1609 			if (nla_type(attr) != IFLA_VF_PORT)
1610 				continue;
1611 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1612 				attr, ifla_port_policy);
1613 			if (err < 0)
1614 				goto errout;
1615 			if (!port[IFLA_PORT_VF]) {
1616 				err = -EOPNOTSUPP;
1617 				goto errout;
1618 			}
1619 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1620 			err = ops->ndo_set_vf_port(dev, vf, port);
1621 			if (err < 0)
1622 				goto errout;
1623 			modified = 1;
1624 		}
1625 	}
1626 	err = 0;
1627 
1628 	if (tb[IFLA_PORT_SELF]) {
1629 		struct nlattr *port[IFLA_PORT_MAX+1];
1630 
1631 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1632 			tb[IFLA_PORT_SELF], ifla_port_policy);
1633 		if (err < 0)
1634 			goto errout;
1635 
1636 		err = -EOPNOTSUPP;
1637 		if (ops->ndo_set_vf_port)
1638 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1639 		if (err < 0)
1640 			goto errout;
1641 		modified = 1;
1642 	}
1643 
1644 	if (tb[IFLA_AF_SPEC]) {
1645 		struct nlattr *af;
1646 		int rem;
1647 
1648 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1649 			const struct rtnl_af_ops *af_ops;
1650 
1651 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1652 				BUG();
1653 
1654 			err = af_ops->set_link_af(dev, af);
1655 			if (err < 0)
1656 				goto errout;
1657 
1658 			modified = 1;
1659 		}
1660 	}
1661 	err = 0;
1662 
1663 errout:
1664 	if (err < 0 && modified)
1665 		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",
1666 				     dev->name);
1667 
1668 	return err;
1669 }
1670 
1671 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1672 {
1673 	struct net *net = sock_net(skb->sk);
1674 	struct ifinfomsg *ifm;
1675 	struct net_device *dev;
1676 	int err;
1677 	struct nlattr *tb[IFLA_MAX+1];
1678 	char ifname[IFNAMSIZ];
1679 
1680 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1681 	if (err < 0)
1682 		goto errout;
1683 
1684 	if (tb[IFLA_IFNAME])
1685 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1686 	else
1687 		ifname[0] = '\0';
1688 
1689 	err = -EINVAL;
1690 	ifm = nlmsg_data(nlh);
1691 	if (ifm->ifi_index > 0)
1692 		dev = __dev_get_by_index(net, ifm->ifi_index);
1693 	else if (tb[IFLA_IFNAME])
1694 		dev = __dev_get_by_name(net, ifname);
1695 	else
1696 		goto errout;
1697 
1698 	if (dev == NULL) {
1699 		err = -ENODEV;
1700 		goto errout;
1701 	}
1702 
1703 	err = validate_linkmsg(dev, tb);
1704 	if (err < 0)
1705 		goto errout;
1706 
1707 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1708 errout:
1709 	return err;
1710 }
1711 
1712 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1713 {
1714 	struct net *net = sock_net(skb->sk);
1715 	const struct rtnl_link_ops *ops;
1716 	struct net_device *dev;
1717 	struct ifinfomsg *ifm;
1718 	char ifname[IFNAMSIZ];
1719 	struct nlattr *tb[IFLA_MAX+1];
1720 	int err;
1721 	LIST_HEAD(list_kill);
1722 
1723 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1724 	if (err < 0)
1725 		return err;
1726 
1727 	if (tb[IFLA_IFNAME])
1728 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1729 
1730 	ifm = nlmsg_data(nlh);
1731 	if (ifm->ifi_index > 0)
1732 		dev = __dev_get_by_index(net, ifm->ifi_index);
1733 	else if (tb[IFLA_IFNAME])
1734 		dev = __dev_get_by_name(net, ifname);
1735 	else
1736 		return -EINVAL;
1737 
1738 	if (!dev)
1739 		return -ENODEV;
1740 
1741 	ops = dev->rtnl_link_ops;
1742 	if (!ops)
1743 		return -EOPNOTSUPP;
1744 
1745 	ops->dellink(dev, &list_kill);
1746 	unregister_netdevice_many(&list_kill);
1747 	list_del(&list_kill);
1748 	return 0;
1749 }
1750 
1751 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1752 {
1753 	unsigned int old_flags;
1754 	int err;
1755 
1756 	old_flags = dev->flags;
1757 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1758 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1759 		if (err < 0)
1760 			return err;
1761 	}
1762 
1763 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1764 
1765 	__dev_notify_flags(dev, old_flags, ~0U);
1766 	return 0;
1767 }
1768 EXPORT_SYMBOL(rtnl_configure_link);
1769 
1770 struct net_device *rtnl_create_link(struct net *net,
1771 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1772 {
1773 	int err;
1774 	struct net_device *dev;
1775 	unsigned int num_tx_queues = 1;
1776 	unsigned int num_rx_queues = 1;
1777 
1778 	if (tb[IFLA_NUM_TX_QUEUES])
1779 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1780 	else if (ops->get_num_tx_queues)
1781 		num_tx_queues = ops->get_num_tx_queues();
1782 
1783 	if (tb[IFLA_NUM_RX_QUEUES])
1784 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1785 	else if (ops->get_num_rx_queues)
1786 		num_rx_queues = ops->get_num_rx_queues();
1787 
1788 	err = -ENOMEM;
1789 	dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1790 			       num_tx_queues, num_rx_queues);
1791 	if (!dev)
1792 		goto err;
1793 
1794 	dev_net_set(dev, net);
1795 	dev->rtnl_link_ops = ops;
1796 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1797 
1798 	if (tb[IFLA_MTU])
1799 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1800 	if (tb[IFLA_ADDRESS]) {
1801 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1802 				nla_len(tb[IFLA_ADDRESS]));
1803 		dev->addr_assign_type = NET_ADDR_SET;
1804 	}
1805 	if (tb[IFLA_BROADCAST])
1806 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1807 				nla_len(tb[IFLA_BROADCAST]));
1808 	if (tb[IFLA_TXQLEN])
1809 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1810 	if (tb[IFLA_OPERSTATE])
1811 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1812 	if (tb[IFLA_LINKMODE])
1813 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1814 	if (tb[IFLA_GROUP])
1815 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1816 
1817 	return dev;
1818 
1819 err:
1820 	return ERR_PTR(err);
1821 }
1822 EXPORT_SYMBOL(rtnl_create_link);
1823 
1824 static int rtnl_group_changelink(const struct sk_buff *skb,
1825 		struct net *net, int group,
1826 		struct ifinfomsg *ifm,
1827 		struct nlattr **tb)
1828 {
1829 	struct net_device *dev;
1830 	int err;
1831 
1832 	for_each_netdev(net, dev) {
1833 		if (dev->group == group) {
1834 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1835 			if (err < 0)
1836 				return err;
1837 		}
1838 	}
1839 
1840 	return 0;
1841 }
1842 
1843 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1844 {
1845 	struct net *net = sock_net(skb->sk);
1846 	const struct rtnl_link_ops *ops;
1847 	const struct rtnl_link_ops *m_ops = NULL;
1848 	struct net_device *dev;
1849 	struct net_device *master_dev = NULL;
1850 	struct ifinfomsg *ifm;
1851 	char kind[MODULE_NAME_LEN];
1852 	char ifname[IFNAMSIZ];
1853 	struct nlattr *tb[IFLA_MAX+1];
1854 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1855 	int err;
1856 
1857 #ifdef CONFIG_MODULES
1858 replay:
1859 #endif
1860 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1861 	if (err < 0)
1862 		return err;
1863 
1864 	if (tb[IFLA_IFNAME])
1865 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1866 	else
1867 		ifname[0] = '\0';
1868 
1869 	ifm = nlmsg_data(nlh);
1870 	if (ifm->ifi_index > 0)
1871 		dev = __dev_get_by_index(net, ifm->ifi_index);
1872 	else {
1873 		if (ifname[0])
1874 			dev = __dev_get_by_name(net, ifname);
1875 		else
1876 			dev = NULL;
1877 	}
1878 
1879 	if (dev) {
1880 		master_dev = netdev_master_upper_dev_get(dev);
1881 		if (master_dev)
1882 			m_ops = master_dev->rtnl_link_ops;
1883 	}
1884 
1885 	err = validate_linkmsg(dev, tb);
1886 	if (err < 0)
1887 		return err;
1888 
1889 	if (tb[IFLA_LINKINFO]) {
1890 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1891 				       tb[IFLA_LINKINFO], ifla_info_policy);
1892 		if (err < 0)
1893 			return err;
1894 	} else
1895 		memset(linkinfo, 0, sizeof(linkinfo));
1896 
1897 	if (linkinfo[IFLA_INFO_KIND]) {
1898 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1899 		ops = rtnl_link_ops_get(kind);
1900 	} else {
1901 		kind[0] = '\0';
1902 		ops = NULL;
1903 	}
1904 
1905 	if (1) {
1906 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1907 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1908 		struct nlattr **data = NULL;
1909 		struct nlattr **slave_data = NULL;
1910 		struct net *dest_net;
1911 
1912 		if (ops) {
1913 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1914 				err = nla_parse_nested(attr, ops->maxtype,
1915 						       linkinfo[IFLA_INFO_DATA],
1916 						       ops->policy);
1917 				if (err < 0)
1918 					return err;
1919 				data = attr;
1920 			}
1921 			if (ops->validate) {
1922 				err = ops->validate(tb, data);
1923 				if (err < 0)
1924 					return err;
1925 			}
1926 		}
1927 
1928 		if (m_ops) {
1929 			if (m_ops->slave_maxtype &&
1930 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
1931 				err = nla_parse_nested(slave_attr,
1932 						       m_ops->slave_maxtype,
1933 						       linkinfo[IFLA_INFO_SLAVE_DATA],
1934 						       m_ops->slave_policy);
1935 				if (err < 0)
1936 					return err;
1937 				slave_data = slave_attr;
1938 			}
1939 			if (m_ops->slave_validate) {
1940 				err = m_ops->slave_validate(tb, slave_data);
1941 				if (err < 0)
1942 					return err;
1943 			}
1944 		}
1945 
1946 		if (dev) {
1947 			int modified = 0;
1948 
1949 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1950 				return -EEXIST;
1951 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1952 				return -EOPNOTSUPP;
1953 
1954 			if (linkinfo[IFLA_INFO_DATA]) {
1955 				if (!ops || ops != dev->rtnl_link_ops ||
1956 				    !ops->changelink)
1957 					return -EOPNOTSUPP;
1958 
1959 				err = ops->changelink(dev, tb, data);
1960 				if (err < 0)
1961 					return err;
1962 				modified = 1;
1963 			}
1964 
1965 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1966 				if (!m_ops || !m_ops->slave_changelink)
1967 					return -EOPNOTSUPP;
1968 
1969 				err = m_ops->slave_changelink(master_dev, dev,
1970 							      tb, slave_data);
1971 				if (err < 0)
1972 					return err;
1973 				modified = 1;
1974 			}
1975 
1976 			return do_setlink(skb, dev, ifm, tb, ifname, modified);
1977 		}
1978 
1979 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1980 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1981 				return rtnl_group_changelink(skb, net,
1982 						nla_get_u32(tb[IFLA_GROUP]),
1983 						ifm, tb);
1984 			return -ENODEV;
1985 		}
1986 
1987 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1988 			return -EOPNOTSUPP;
1989 
1990 		if (!ops) {
1991 #ifdef CONFIG_MODULES
1992 			if (kind[0]) {
1993 				__rtnl_unlock();
1994 				request_module("rtnl-link-%s", kind);
1995 				rtnl_lock();
1996 				ops = rtnl_link_ops_get(kind);
1997 				if (ops)
1998 					goto replay;
1999 			}
2000 #endif
2001 			return -EOPNOTSUPP;
2002 		}
2003 
2004 		if (!ifname[0])
2005 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2006 
2007 		dest_net = rtnl_link_get_net(net, tb);
2008 		if (IS_ERR(dest_net))
2009 			return PTR_ERR(dest_net);
2010 
2011 		dev = rtnl_create_link(dest_net, ifname, ops, tb);
2012 		if (IS_ERR(dev)) {
2013 			err = PTR_ERR(dev);
2014 			goto out;
2015 		}
2016 
2017 		dev->ifindex = ifm->ifi_index;
2018 
2019 		if (ops->newlink) {
2020 			err = ops->newlink(net, dev, tb, data);
2021 			/* Drivers should call free_netdev() in ->destructor
2022 			 * and unregister it on failure so that device could be
2023 			 * finally freed in rtnl_unlock.
2024 			 */
2025 			if (err < 0)
2026 				goto out;
2027 		} else {
2028 			err = register_netdevice(dev);
2029 			if (err < 0) {
2030 				free_netdev(dev);
2031 				goto out;
2032 			}
2033 		}
2034 		err = rtnl_configure_link(dev, ifm);
2035 		if (err < 0)
2036 			unregister_netdevice(dev);
2037 out:
2038 		put_net(dest_net);
2039 		return err;
2040 	}
2041 }
2042 
2043 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2044 {
2045 	struct net *net = sock_net(skb->sk);
2046 	struct ifinfomsg *ifm;
2047 	char ifname[IFNAMSIZ];
2048 	struct nlattr *tb[IFLA_MAX+1];
2049 	struct net_device *dev = NULL;
2050 	struct sk_buff *nskb;
2051 	int err;
2052 	u32 ext_filter_mask = 0;
2053 
2054 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2055 	if (err < 0)
2056 		return err;
2057 
2058 	if (tb[IFLA_IFNAME])
2059 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2060 
2061 	if (tb[IFLA_EXT_MASK])
2062 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2063 
2064 	ifm = nlmsg_data(nlh);
2065 	if (ifm->ifi_index > 0)
2066 		dev = __dev_get_by_index(net, ifm->ifi_index);
2067 	else if (tb[IFLA_IFNAME])
2068 		dev = __dev_get_by_name(net, ifname);
2069 	else
2070 		return -EINVAL;
2071 
2072 	if (dev == NULL)
2073 		return -ENODEV;
2074 
2075 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2076 	if (nskb == NULL)
2077 		return -ENOBUFS;
2078 
2079 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2080 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2081 	if (err < 0) {
2082 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2083 		WARN_ON(err == -EMSGSIZE);
2084 		kfree_skb(nskb);
2085 	} else
2086 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2087 
2088 	return err;
2089 }
2090 
2091 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2092 {
2093 	struct net *net = sock_net(skb->sk);
2094 	struct net_device *dev;
2095 	struct nlattr *tb[IFLA_MAX+1];
2096 	u32 ext_filter_mask = 0;
2097 	u16 min_ifinfo_dump_size = 0;
2098 
2099 	if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2100 			ifla_policy) >= 0) {
2101 		if (tb[IFLA_EXT_MASK])
2102 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2103 	}
2104 
2105 	if (!ext_filter_mask)
2106 		return NLMSG_GOODSIZE;
2107 	/*
2108 	 * traverse the list of net devices and compute the minimum
2109 	 * buffer size based upon the filter mask.
2110 	 */
2111 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2112 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2113 					     if_nlmsg_size(dev,
2114 						           ext_filter_mask));
2115 	}
2116 
2117 	return min_ifinfo_dump_size;
2118 }
2119 
2120 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2121 {
2122 	int idx;
2123 	int s_idx = cb->family;
2124 
2125 	if (s_idx == 0)
2126 		s_idx = 1;
2127 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2128 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2129 		if (idx < s_idx || idx == PF_PACKET)
2130 			continue;
2131 		if (rtnl_msg_handlers[idx] == NULL ||
2132 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2133 			continue;
2134 		if (idx > s_idx) {
2135 			memset(&cb->args[0], 0, sizeof(cb->args));
2136 			cb->prev_seq = 0;
2137 			cb->seq = 0;
2138 		}
2139 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2140 			break;
2141 	}
2142 	cb->family = idx;
2143 
2144 	return skb->len;
2145 }
2146 
2147 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2148 		  gfp_t flags)
2149 {
2150 	struct net *net = dev_net(dev);
2151 	struct sk_buff *skb;
2152 	int err = -ENOBUFS;
2153 	size_t if_info_size;
2154 
2155 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2156 	if (skb == NULL)
2157 		goto errout;
2158 
2159 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2160 	if (err < 0) {
2161 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2162 		WARN_ON(err == -EMSGSIZE);
2163 		kfree_skb(skb);
2164 		goto errout;
2165 	}
2166 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2167 	return;
2168 errout:
2169 	if (err < 0)
2170 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2171 }
2172 EXPORT_SYMBOL(rtmsg_ifinfo);
2173 
2174 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2175 				   struct net_device *dev,
2176 				   u8 *addr, u32 pid, u32 seq,
2177 				   int type, unsigned int flags,
2178 				   int nlflags)
2179 {
2180 	struct nlmsghdr *nlh;
2181 	struct ndmsg *ndm;
2182 
2183 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2184 	if (!nlh)
2185 		return -EMSGSIZE;
2186 
2187 	ndm = nlmsg_data(nlh);
2188 	ndm->ndm_family  = AF_BRIDGE;
2189 	ndm->ndm_pad1	 = 0;
2190 	ndm->ndm_pad2    = 0;
2191 	ndm->ndm_flags	 = flags;
2192 	ndm->ndm_type	 = 0;
2193 	ndm->ndm_ifindex = dev->ifindex;
2194 	ndm->ndm_state   = NUD_PERMANENT;
2195 
2196 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2197 		goto nla_put_failure;
2198 
2199 	return nlmsg_end(skb, nlh);
2200 
2201 nla_put_failure:
2202 	nlmsg_cancel(skb, nlh);
2203 	return -EMSGSIZE;
2204 }
2205 
2206 static inline size_t rtnl_fdb_nlmsg_size(void)
2207 {
2208 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2209 }
2210 
2211 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2212 {
2213 	struct net *net = dev_net(dev);
2214 	struct sk_buff *skb;
2215 	int err = -ENOBUFS;
2216 
2217 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2218 	if (!skb)
2219 		goto errout;
2220 
2221 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2222 	if (err < 0) {
2223 		kfree_skb(skb);
2224 		goto errout;
2225 	}
2226 
2227 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2228 	return;
2229 errout:
2230 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2231 }
2232 
2233 /**
2234  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2235  */
2236 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2237 		     struct nlattr *tb[],
2238 		     struct net_device *dev,
2239 		     const unsigned char *addr,
2240 		     u16 flags)
2241 {
2242 	int err = -EINVAL;
2243 
2244 	/* If aging addresses are supported device will need to
2245 	 * implement its own handler for this.
2246 	 */
2247 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2248 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2249 		return err;
2250 	}
2251 
2252 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2253 		err = dev_uc_add_excl(dev, addr);
2254 	else if (is_multicast_ether_addr(addr))
2255 		err = dev_mc_add_excl(dev, addr);
2256 
2257 	/* Only return duplicate errors if NLM_F_EXCL is set */
2258 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2259 		err = 0;
2260 
2261 	return err;
2262 }
2263 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2264 
2265 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2266 {
2267 	struct net *net = sock_net(skb->sk);
2268 	struct ndmsg *ndm;
2269 	struct nlattr *tb[NDA_MAX+1];
2270 	struct net_device *dev;
2271 	u8 *addr;
2272 	int err;
2273 
2274 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2275 	if (err < 0)
2276 		return err;
2277 
2278 	ndm = nlmsg_data(nlh);
2279 	if (ndm->ndm_ifindex == 0) {
2280 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2281 		return -EINVAL;
2282 	}
2283 
2284 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2285 	if (dev == NULL) {
2286 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2287 		return -ENODEV;
2288 	}
2289 
2290 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2291 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2292 		return -EINVAL;
2293 	}
2294 
2295 	addr = nla_data(tb[NDA_LLADDR]);
2296 
2297 	err = -EOPNOTSUPP;
2298 
2299 	/* Support fdb on master device the net/bridge default case */
2300 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2301 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2302 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2303 		const struct net_device_ops *ops = br_dev->netdev_ops;
2304 
2305 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2306 		if (err)
2307 			goto out;
2308 		else
2309 			ndm->ndm_flags &= ~NTF_MASTER;
2310 	}
2311 
2312 	/* Embedded bridge, macvlan, and any other device support */
2313 	if ((ndm->ndm_flags & NTF_SELF)) {
2314 		if (dev->netdev_ops->ndo_fdb_add)
2315 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2316 							   nlh->nlmsg_flags);
2317 		else
2318 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2319 					       nlh->nlmsg_flags);
2320 
2321 		if (!err) {
2322 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2323 			ndm->ndm_flags &= ~NTF_SELF;
2324 		}
2325 	}
2326 out:
2327 	return err;
2328 }
2329 
2330 /**
2331  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2332  */
2333 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2334 		     struct nlattr *tb[],
2335 		     struct net_device *dev,
2336 		     const unsigned char *addr)
2337 {
2338 	int err = -EOPNOTSUPP;
2339 
2340 	/* If aging addresses are supported device will need to
2341 	 * implement its own handler for this.
2342 	 */
2343 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2344 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2345 		return -EINVAL;
2346 	}
2347 
2348 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2349 		err = dev_uc_del(dev, addr);
2350 	else if (is_multicast_ether_addr(addr))
2351 		err = dev_mc_del(dev, addr);
2352 	else
2353 		err = -EINVAL;
2354 
2355 	return err;
2356 }
2357 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2358 
2359 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2360 {
2361 	struct net *net = sock_net(skb->sk);
2362 	struct ndmsg *ndm;
2363 	struct nlattr *tb[NDA_MAX+1];
2364 	struct net_device *dev;
2365 	int err = -EINVAL;
2366 	__u8 *addr;
2367 
2368 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2369 		return -EPERM;
2370 
2371 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2372 	if (err < 0)
2373 		return err;
2374 
2375 	ndm = nlmsg_data(nlh);
2376 	if (ndm->ndm_ifindex == 0) {
2377 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2378 		return -EINVAL;
2379 	}
2380 
2381 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2382 	if (dev == NULL) {
2383 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2384 		return -ENODEV;
2385 	}
2386 
2387 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2388 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2389 		return -EINVAL;
2390 	}
2391 
2392 	addr = nla_data(tb[NDA_LLADDR]);
2393 
2394 	err = -EOPNOTSUPP;
2395 
2396 	/* Support fdb on master device the net/bridge default case */
2397 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2398 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2399 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2400 		const struct net_device_ops *ops = br_dev->netdev_ops;
2401 
2402 		if (ops->ndo_fdb_del)
2403 			err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2404 
2405 		if (err)
2406 			goto out;
2407 		else
2408 			ndm->ndm_flags &= ~NTF_MASTER;
2409 	}
2410 
2411 	/* Embedded bridge, macvlan, and any other device support */
2412 	if (ndm->ndm_flags & NTF_SELF) {
2413 		if (dev->netdev_ops->ndo_fdb_del)
2414 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2415 		else
2416 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2417 
2418 		if (!err) {
2419 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2420 			ndm->ndm_flags &= ~NTF_SELF;
2421 		}
2422 	}
2423 out:
2424 	return err;
2425 }
2426 
2427 static int nlmsg_populate_fdb(struct sk_buff *skb,
2428 			      struct netlink_callback *cb,
2429 			      struct net_device *dev,
2430 			      int *idx,
2431 			      struct netdev_hw_addr_list *list)
2432 {
2433 	struct netdev_hw_addr *ha;
2434 	int err;
2435 	u32 portid, seq;
2436 
2437 	portid = NETLINK_CB(cb->skb).portid;
2438 	seq = cb->nlh->nlmsg_seq;
2439 
2440 	list_for_each_entry(ha, &list->list, list) {
2441 		if (*idx < cb->args[0])
2442 			goto skip;
2443 
2444 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2445 					      portid, seq,
2446 					      RTM_NEWNEIGH, NTF_SELF,
2447 					      NLM_F_MULTI);
2448 		if (err < 0)
2449 			return err;
2450 skip:
2451 		*idx += 1;
2452 	}
2453 	return 0;
2454 }
2455 
2456 /**
2457  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2458  * @nlh: netlink message header
2459  * @dev: netdevice
2460  *
2461  * Default netdevice operation to dump the existing unicast address list.
2462  * Returns number of addresses from list put in skb.
2463  */
2464 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2465 		      struct netlink_callback *cb,
2466 		      struct net_device *dev,
2467 		      int idx)
2468 {
2469 	int err;
2470 
2471 	netif_addr_lock_bh(dev);
2472 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2473 	if (err)
2474 		goto out;
2475 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2476 out:
2477 	netif_addr_unlock_bh(dev);
2478 	return idx;
2479 }
2480 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2481 
2482 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2483 {
2484 	int idx = 0;
2485 	struct net *net = sock_net(skb->sk);
2486 	struct net_device *dev;
2487 
2488 	rcu_read_lock();
2489 	for_each_netdev_rcu(net, dev) {
2490 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2491 			struct net_device *br_dev;
2492 			const struct net_device_ops *ops;
2493 
2494 			br_dev = netdev_master_upper_dev_get(dev);
2495 			ops = br_dev->netdev_ops;
2496 			if (ops->ndo_fdb_dump)
2497 				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2498 		}
2499 
2500 		if (dev->netdev_ops->ndo_fdb_dump)
2501 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2502 		else
2503 			idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2504 	}
2505 	rcu_read_unlock();
2506 
2507 	cb->args[0] = idx;
2508 	return skb->len;
2509 }
2510 
2511 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2512 			    struct net_device *dev, u16 mode)
2513 {
2514 	struct nlmsghdr *nlh;
2515 	struct ifinfomsg *ifm;
2516 	struct nlattr *br_afspec;
2517 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2518 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2519 
2520 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2521 	if (nlh == NULL)
2522 		return -EMSGSIZE;
2523 
2524 	ifm = nlmsg_data(nlh);
2525 	ifm->ifi_family = AF_BRIDGE;
2526 	ifm->__ifi_pad = 0;
2527 	ifm->ifi_type = dev->type;
2528 	ifm->ifi_index = dev->ifindex;
2529 	ifm->ifi_flags = dev_get_flags(dev);
2530 	ifm->ifi_change = 0;
2531 
2532 
2533 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2534 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2535 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2536 	    (br_dev &&
2537 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2538 	    (dev->addr_len &&
2539 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2540 	    (dev->ifindex != dev->iflink &&
2541 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2542 		goto nla_put_failure;
2543 
2544 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2545 	if (!br_afspec)
2546 		goto nla_put_failure;
2547 
2548 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2549 	    nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2550 		nla_nest_cancel(skb, br_afspec);
2551 		goto nla_put_failure;
2552 	}
2553 	nla_nest_end(skb, br_afspec);
2554 
2555 	return nlmsg_end(skb, nlh);
2556 nla_put_failure:
2557 	nlmsg_cancel(skb, nlh);
2558 	return -EMSGSIZE;
2559 }
2560 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2561 
2562 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2563 {
2564 	struct net *net = sock_net(skb->sk);
2565 	struct net_device *dev;
2566 	int idx = 0;
2567 	u32 portid = NETLINK_CB(cb->skb).portid;
2568 	u32 seq = cb->nlh->nlmsg_seq;
2569 	struct nlattr *extfilt;
2570 	u32 filter_mask = 0;
2571 
2572 	extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2573 				  IFLA_EXT_MASK);
2574 	if (extfilt)
2575 		filter_mask = nla_get_u32(extfilt);
2576 
2577 	rcu_read_lock();
2578 	for_each_netdev_rcu(net, dev) {
2579 		const struct net_device_ops *ops = dev->netdev_ops;
2580 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2581 
2582 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2583 			if (idx >= cb->args[0] &&
2584 			    br_dev->netdev_ops->ndo_bridge_getlink(
2585 				    skb, portid, seq, dev, filter_mask) < 0)
2586 				break;
2587 			idx++;
2588 		}
2589 
2590 		if (ops->ndo_bridge_getlink) {
2591 			if (idx >= cb->args[0] &&
2592 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2593 						    filter_mask) < 0)
2594 				break;
2595 			idx++;
2596 		}
2597 	}
2598 	rcu_read_unlock();
2599 	cb->args[0] = idx;
2600 
2601 	return skb->len;
2602 }
2603 
2604 static inline size_t bridge_nlmsg_size(void)
2605 {
2606 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2607 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2608 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2609 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2610 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2611 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2612 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2613 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2614 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2615 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2616 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2617 }
2618 
2619 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2620 {
2621 	struct net *net = dev_net(dev);
2622 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2623 	struct sk_buff *skb;
2624 	int err = -EOPNOTSUPP;
2625 
2626 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2627 	if (!skb) {
2628 		err = -ENOMEM;
2629 		goto errout;
2630 	}
2631 
2632 	if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2633 	    br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2634 		err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2635 		if (err < 0)
2636 			goto errout;
2637 	}
2638 
2639 	if ((flags & BRIDGE_FLAGS_SELF) &&
2640 	    dev->netdev_ops->ndo_bridge_getlink) {
2641 		err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2642 		if (err < 0)
2643 			goto errout;
2644 	}
2645 
2646 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2647 	return 0;
2648 errout:
2649 	WARN_ON(err == -EMSGSIZE);
2650 	kfree_skb(skb);
2651 	rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2652 	return err;
2653 }
2654 
2655 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2656 {
2657 	struct net *net = sock_net(skb->sk);
2658 	struct ifinfomsg *ifm;
2659 	struct net_device *dev;
2660 	struct nlattr *br_spec, *attr = NULL;
2661 	int rem, err = -EOPNOTSUPP;
2662 	u16 oflags, flags = 0;
2663 	bool have_flags = false;
2664 
2665 	if (nlmsg_len(nlh) < sizeof(*ifm))
2666 		return -EINVAL;
2667 
2668 	ifm = nlmsg_data(nlh);
2669 	if (ifm->ifi_family != AF_BRIDGE)
2670 		return -EPFNOSUPPORT;
2671 
2672 	dev = __dev_get_by_index(net, ifm->ifi_index);
2673 	if (!dev) {
2674 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2675 		return -ENODEV;
2676 	}
2677 
2678 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2679 	if (br_spec) {
2680 		nla_for_each_nested(attr, br_spec, rem) {
2681 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2682 				have_flags = true;
2683 				flags = nla_get_u16(attr);
2684 				break;
2685 			}
2686 		}
2687 	}
2688 
2689 	oflags = flags;
2690 
2691 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2692 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2693 
2694 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2695 			err = -EOPNOTSUPP;
2696 			goto out;
2697 		}
2698 
2699 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2700 		if (err)
2701 			goto out;
2702 
2703 		flags &= ~BRIDGE_FLAGS_MASTER;
2704 	}
2705 
2706 	if ((flags & BRIDGE_FLAGS_SELF)) {
2707 		if (!dev->netdev_ops->ndo_bridge_setlink)
2708 			err = -EOPNOTSUPP;
2709 		else
2710 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2711 
2712 		if (!err)
2713 			flags &= ~BRIDGE_FLAGS_SELF;
2714 	}
2715 
2716 	if (have_flags)
2717 		memcpy(nla_data(attr), &flags, sizeof(flags));
2718 	/* Generate event to notify upper layer of bridge change */
2719 	if (!err)
2720 		err = rtnl_bridge_notify(dev, oflags);
2721 out:
2722 	return err;
2723 }
2724 
2725 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2726 {
2727 	struct net *net = sock_net(skb->sk);
2728 	struct ifinfomsg *ifm;
2729 	struct net_device *dev;
2730 	struct nlattr *br_spec, *attr = NULL;
2731 	int rem, err = -EOPNOTSUPP;
2732 	u16 oflags, flags = 0;
2733 	bool have_flags = false;
2734 
2735 	if (nlmsg_len(nlh) < sizeof(*ifm))
2736 		return -EINVAL;
2737 
2738 	ifm = nlmsg_data(nlh);
2739 	if (ifm->ifi_family != AF_BRIDGE)
2740 		return -EPFNOSUPPORT;
2741 
2742 	dev = __dev_get_by_index(net, ifm->ifi_index);
2743 	if (!dev) {
2744 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2745 		return -ENODEV;
2746 	}
2747 
2748 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2749 	if (br_spec) {
2750 		nla_for_each_nested(attr, br_spec, rem) {
2751 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2752 				have_flags = true;
2753 				flags = nla_get_u16(attr);
2754 				break;
2755 			}
2756 		}
2757 	}
2758 
2759 	oflags = flags;
2760 
2761 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2762 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2763 
2764 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2765 			err = -EOPNOTSUPP;
2766 			goto out;
2767 		}
2768 
2769 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2770 		if (err)
2771 			goto out;
2772 
2773 		flags &= ~BRIDGE_FLAGS_MASTER;
2774 	}
2775 
2776 	if ((flags & BRIDGE_FLAGS_SELF)) {
2777 		if (!dev->netdev_ops->ndo_bridge_dellink)
2778 			err = -EOPNOTSUPP;
2779 		else
2780 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2781 
2782 		if (!err)
2783 			flags &= ~BRIDGE_FLAGS_SELF;
2784 	}
2785 
2786 	if (have_flags)
2787 		memcpy(nla_data(attr), &flags, sizeof(flags));
2788 	/* Generate event to notify upper layer of bridge change */
2789 	if (!err)
2790 		err = rtnl_bridge_notify(dev, oflags);
2791 out:
2792 	return err;
2793 }
2794 
2795 /* Process one rtnetlink message. */
2796 
2797 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2798 {
2799 	struct net *net = sock_net(skb->sk);
2800 	rtnl_doit_func doit;
2801 	int sz_idx, kind;
2802 	int family;
2803 	int type;
2804 	int err;
2805 
2806 	type = nlh->nlmsg_type;
2807 	if (type > RTM_MAX)
2808 		return -EOPNOTSUPP;
2809 
2810 	type -= RTM_BASE;
2811 
2812 	/* All the messages must have at least 1 byte length */
2813 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2814 		return 0;
2815 
2816 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2817 	sz_idx = type>>2;
2818 	kind = type&3;
2819 
2820 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2821 		return -EPERM;
2822 
2823 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2824 		struct sock *rtnl;
2825 		rtnl_dumpit_func dumpit;
2826 		rtnl_calcit_func calcit;
2827 		u16 min_dump_alloc = 0;
2828 
2829 		dumpit = rtnl_get_dumpit(family, type);
2830 		if (dumpit == NULL)
2831 			return -EOPNOTSUPP;
2832 		calcit = rtnl_get_calcit(family, type);
2833 		if (calcit)
2834 			min_dump_alloc = calcit(skb, nlh);
2835 
2836 		__rtnl_unlock();
2837 		rtnl = net->rtnl;
2838 		{
2839 			struct netlink_dump_control c = {
2840 				.dump		= dumpit,
2841 				.min_dump_alloc	= min_dump_alloc,
2842 			};
2843 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2844 		}
2845 		rtnl_lock();
2846 		return err;
2847 	}
2848 
2849 	doit = rtnl_get_doit(family, type);
2850 	if (doit == NULL)
2851 		return -EOPNOTSUPP;
2852 
2853 	return doit(skb, nlh);
2854 }
2855 
2856 static void rtnetlink_rcv(struct sk_buff *skb)
2857 {
2858 	rtnl_lock();
2859 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2860 	rtnl_unlock();
2861 }
2862 
2863 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2864 {
2865 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2866 
2867 	switch (event) {
2868 	case NETDEV_UP:
2869 	case NETDEV_DOWN:
2870 	case NETDEV_PRE_UP:
2871 	case NETDEV_POST_INIT:
2872 	case NETDEV_REGISTER:
2873 	case NETDEV_CHANGE:
2874 	case NETDEV_PRE_TYPE_CHANGE:
2875 	case NETDEV_GOING_DOWN:
2876 	case NETDEV_UNREGISTER:
2877 	case NETDEV_UNREGISTER_FINAL:
2878 	case NETDEV_RELEASE:
2879 	case NETDEV_JOIN:
2880 		break;
2881 	default:
2882 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2883 		break;
2884 	}
2885 	return NOTIFY_DONE;
2886 }
2887 
2888 static struct notifier_block rtnetlink_dev_notifier = {
2889 	.notifier_call	= rtnetlink_event,
2890 };
2891 
2892 
2893 static int __net_init rtnetlink_net_init(struct net *net)
2894 {
2895 	struct sock *sk;
2896 	struct netlink_kernel_cfg cfg = {
2897 		.groups		= RTNLGRP_MAX,
2898 		.input		= rtnetlink_rcv,
2899 		.cb_mutex	= &rtnl_mutex,
2900 		.flags		= NL_CFG_F_NONROOT_RECV,
2901 	};
2902 
2903 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2904 	if (!sk)
2905 		return -ENOMEM;
2906 	net->rtnl = sk;
2907 	return 0;
2908 }
2909 
2910 static void __net_exit rtnetlink_net_exit(struct net *net)
2911 {
2912 	netlink_kernel_release(net->rtnl);
2913 	net->rtnl = NULL;
2914 }
2915 
2916 static struct pernet_operations rtnetlink_net_ops = {
2917 	.init = rtnetlink_net_init,
2918 	.exit = rtnetlink_net_exit,
2919 };
2920 
2921 void __init rtnetlink_init(void)
2922 {
2923 	if (register_pernet_subsys(&rtnetlink_net_ops))
2924 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2925 
2926 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2927 
2928 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2929 		      rtnl_dump_ifinfo, rtnl_calcit);
2930 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2931 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2932 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2933 
2934 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2935 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2936 
2937 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2938 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2939 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2940 
2941 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2942 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2943 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2944 }
2945 
2946