xref: /linux/net/core/rtnetlink.c (revision 9cfc5c90ad38c8fc11bfd39de42a107da00871ba)
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/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
42 
43 #include <asm/uaccess.h>
44 
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
48 #include <net/ip.h>
49 #include <net/protocol.h>
50 #include <net/arp.h>
51 #include <net/route.h>
52 #include <net/udp.h>
53 #include <net/tcp.h>
54 #include <net/sock.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
59 
60 struct rtnl_link {
61 	rtnl_doit_func		doit;
62 	rtnl_dumpit_func	dumpit;
63 	rtnl_calcit_func 	calcit;
64 };
65 
66 static DEFINE_MUTEX(rtnl_mutex);
67 
68 void rtnl_lock(void)
69 {
70 	mutex_lock(&rtnl_mutex);
71 }
72 EXPORT_SYMBOL(rtnl_lock);
73 
74 void __rtnl_unlock(void)
75 {
76 	mutex_unlock(&rtnl_mutex);
77 }
78 
79 void rtnl_unlock(void)
80 {
81 	/* This fellow will unlock it for us. */
82 	netdev_run_todo();
83 }
84 EXPORT_SYMBOL(rtnl_unlock);
85 
86 int rtnl_trylock(void)
87 {
88 	return mutex_trylock(&rtnl_mutex);
89 }
90 EXPORT_SYMBOL(rtnl_trylock);
91 
92 int rtnl_is_locked(void)
93 {
94 	return mutex_is_locked(&rtnl_mutex);
95 }
96 EXPORT_SYMBOL(rtnl_is_locked);
97 
98 #ifdef CONFIG_PROVE_LOCKING
99 bool lockdep_rtnl_is_held(void)
100 {
101 	return lockdep_is_held(&rtnl_mutex);
102 }
103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
105 
106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
107 
108 static inline int rtm_msgindex(int msgtype)
109 {
110 	int msgindex = msgtype - RTM_BASE;
111 
112 	/*
113 	 * msgindex < 0 implies someone tried to register a netlink
114 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
115 	 * the message type has not been added to linux/rtnetlink.h
116 	 */
117 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
118 
119 	return msgindex;
120 }
121 
122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
123 {
124 	struct rtnl_link *tab;
125 
126 	if (protocol <= RTNL_FAMILY_MAX)
127 		tab = rtnl_msg_handlers[protocol];
128 	else
129 		tab = NULL;
130 
131 	if (tab == NULL || tab[msgindex].doit == NULL)
132 		tab = rtnl_msg_handlers[PF_UNSPEC];
133 
134 	return tab[msgindex].doit;
135 }
136 
137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
138 {
139 	struct rtnl_link *tab;
140 
141 	if (protocol <= RTNL_FAMILY_MAX)
142 		tab = rtnl_msg_handlers[protocol];
143 	else
144 		tab = NULL;
145 
146 	if (tab == NULL || tab[msgindex].dumpit == NULL)
147 		tab = rtnl_msg_handlers[PF_UNSPEC];
148 
149 	return tab[msgindex].dumpit;
150 }
151 
152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
153 {
154 	struct rtnl_link *tab;
155 
156 	if (protocol <= RTNL_FAMILY_MAX)
157 		tab = rtnl_msg_handlers[protocol];
158 	else
159 		tab = NULL;
160 
161 	if (tab == NULL || tab[msgindex].calcit == NULL)
162 		tab = rtnl_msg_handlers[PF_UNSPEC];
163 
164 	return tab[msgindex].calcit;
165 }
166 
167 /**
168  * __rtnl_register - Register a rtnetlink message type
169  * @protocol: Protocol family or PF_UNSPEC
170  * @msgtype: rtnetlink message type
171  * @doit: Function pointer called for each request message
172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
173  * @calcit: Function pointer to calc size of dump message
174  *
175  * Registers the specified function pointers (at least one of them has
176  * to be non-NULL) to be called whenever a request message for the
177  * specified protocol family and message type is received.
178  *
179  * The special protocol family PF_UNSPEC may be used to define fallback
180  * function pointers for the case when no entry for the specific protocol
181  * family exists.
182  *
183  * Returns 0 on success or a negative error code.
184  */
185 int __rtnl_register(int protocol, int msgtype,
186 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
187 		    rtnl_calcit_func calcit)
188 {
189 	struct rtnl_link *tab;
190 	int msgindex;
191 
192 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
193 	msgindex = rtm_msgindex(msgtype);
194 
195 	tab = rtnl_msg_handlers[protocol];
196 	if (tab == NULL) {
197 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
198 		if (tab == NULL)
199 			return -ENOBUFS;
200 
201 		rtnl_msg_handlers[protocol] = tab;
202 	}
203 
204 	if (doit)
205 		tab[msgindex].doit = doit;
206 
207 	if (dumpit)
208 		tab[msgindex].dumpit = dumpit;
209 
210 	if (calcit)
211 		tab[msgindex].calcit = calcit;
212 
213 	return 0;
214 }
215 EXPORT_SYMBOL_GPL(__rtnl_register);
216 
217 /**
218  * rtnl_register - Register a rtnetlink message type
219  *
220  * Identical to __rtnl_register() but panics on failure. This is useful
221  * as failure of this function is very unlikely, it can only happen due
222  * to lack of memory when allocating the chain to store all message
223  * handlers for a protocol. Meant for use in init functions where lack
224  * of memory implies no sense in continuing.
225  */
226 void rtnl_register(int protocol, int msgtype,
227 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
228 		   rtnl_calcit_func calcit)
229 {
230 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
231 		panic("Unable to register rtnetlink message handler, "
232 		      "protocol = %d, message type = %d\n",
233 		      protocol, msgtype);
234 }
235 EXPORT_SYMBOL_GPL(rtnl_register);
236 
237 /**
238  * rtnl_unregister - Unregister a rtnetlink message type
239  * @protocol: Protocol family or PF_UNSPEC
240  * @msgtype: rtnetlink message type
241  *
242  * Returns 0 on success or a negative error code.
243  */
244 int rtnl_unregister(int protocol, int msgtype)
245 {
246 	int msgindex;
247 
248 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
249 	msgindex = rtm_msgindex(msgtype);
250 
251 	if (rtnl_msg_handlers[protocol] == NULL)
252 		return -ENOENT;
253 
254 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
255 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
256 
257 	return 0;
258 }
259 EXPORT_SYMBOL_GPL(rtnl_unregister);
260 
261 /**
262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
263  * @protocol : Protocol family or PF_UNSPEC
264  *
265  * Identical to calling rtnl_unregster() for all registered message types
266  * of a certain protocol family.
267  */
268 void rtnl_unregister_all(int protocol)
269 {
270 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 
272 	kfree(rtnl_msg_handlers[protocol]);
273 	rtnl_msg_handlers[protocol] = NULL;
274 }
275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
276 
277 static LIST_HEAD(link_ops);
278 
279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
280 {
281 	const struct rtnl_link_ops *ops;
282 
283 	list_for_each_entry(ops, &link_ops, list) {
284 		if (!strcmp(ops->kind, kind))
285 			return ops;
286 	}
287 	return NULL;
288 }
289 
290 /**
291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
292  * @ops: struct rtnl_link_ops * to register
293  *
294  * The caller must hold the rtnl_mutex. This function should be used
295  * by drivers that create devices during module initialization. It
296  * must be called before registering the devices.
297  *
298  * Returns 0 on success or a negative error code.
299  */
300 int __rtnl_link_register(struct rtnl_link_ops *ops)
301 {
302 	if (rtnl_link_ops_get(ops->kind))
303 		return -EEXIST;
304 
305 	/* The check for setup is here because if ops
306 	 * does not have that filled up, it is not possible
307 	 * to use the ops for creating device. So do not
308 	 * fill up dellink as well. That disables rtnl_dellink.
309 	 */
310 	if (ops->setup && !ops->dellink)
311 		ops->dellink = unregister_netdevice_queue;
312 
313 	list_add_tail(&ops->list, &link_ops);
314 	return 0;
315 }
316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
317 
318 /**
319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
320  * @ops: struct rtnl_link_ops * to register
321  *
322  * Returns 0 on success or a negative error code.
323  */
324 int rtnl_link_register(struct rtnl_link_ops *ops)
325 {
326 	int err;
327 
328 	rtnl_lock();
329 	err = __rtnl_link_register(ops);
330 	rtnl_unlock();
331 	return err;
332 }
333 EXPORT_SYMBOL_GPL(rtnl_link_register);
334 
335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
336 {
337 	struct net_device *dev;
338 	LIST_HEAD(list_kill);
339 
340 	for_each_netdev(net, dev) {
341 		if (dev->rtnl_link_ops == ops)
342 			ops->dellink(dev, &list_kill);
343 	}
344 	unregister_netdevice_many(&list_kill);
345 }
346 
347 /**
348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
349  * @ops: struct rtnl_link_ops * to unregister
350  *
351  * The caller must hold the rtnl_mutex.
352  */
353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
354 {
355 	struct net *net;
356 
357 	for_each_net(net) {
358 		__rtnl_kill_links(net, ops);
359 	}
360 	list_del(&ops->list);
361 }
362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
363 
364 /* Return with the rtnl_lock held when there are no network
365  * devices unregistering in any network namespace.
366  */
367 static void rtnl_lock_unregistering_all(void)
368 {
369 	struct net *net;
370 	bool unregistering;
371 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
372 
373 	add_wait_queue(&netdev_unregistering_wq, &wait);
374 	for (;;) {
375 		unregistering = false;
376 		rtnl_lock();
377 		for_each_net(net) {
378 			if (net->dev_unreg_count > 0) {
379 				unregistering = true;
380 				break;
381 			}
382 		}
383 		if (!unregistering)
384 			break;
385 		__rtnl_unlock();
386 
387 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
388 	}
389 	remove_wait_queue(&netdev_unregistering_wq, &wait);
390 }
391 
392 /**
393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
394  * @ops: struct rtnl_link_ops * to unregister
395  */
396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
397 {
398 	/* Close the race with cleanup_net() */
399 	mutex_lock(&net_mutex);
400 	rtnl_lock_unregistering_all();
401 	__rtnl_link_unregister(ops);
402 	rtnl_unlock();
403 	mutex_unlock(&net_mutex);
404 }
405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
406 
407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
408 {
409 	struct net_device *master_dev;
410 	const struct rtnl_link_ops *ops;
411 
412 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
413 	if (!master_dev)
414 		return 0;
415 	ops = master_dev->rtnl_link_ops;
416 	if (!ops || !ops->get_slave_size)
417 		return 0;
418 	/* IFLA_INFO_SLAVE_DATA + nested data */
419 	return nla_total_size(sizeof(struct nlattr)) +
420 	       ops->get_slave_size(master_dev, dev);
421 }
422 
423 static size_t rtnl_link_get_size(const struct net_device *dev)
424 {
425 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
426 	size_t size;
427 
428 	if (!ops)
429 		return 0;
430 
431 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
432 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
433 
434 	if (ops->get_size)
435 		/* IFLA_INFO_DATA + nested data */
436 		size += nla_total_size(sizeof(struct nlattr)) +
437 			ops->get_size(dev);
438 
439 	if (ops->get_xstats_size)
440 		/* IFLA_INFO_XSTATS */
441 		size += nla_total_size(ops->get_xstats_size(dev));
442 
443 	size += rtnl_link_get_slave_info_data_size(dev);
444 
445 	return size;
446 }
447 
448 static LIST_HEAD(rtnl_af_ops);
449 
450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
451 {
452 	const struct rtnl_af_ops *ops;
453 
454 	list_for_each_entry(ops, &rtnl_af_ops, list) {
455 		if (ops->family == family)
456 			return ops;
457 	}
458 
459 	return NULL;
460 }
461 
462 /**
463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
464  * @ops: struct rtnl_af_ops * to register
465  *
466  * Returns 0 on success or a negative error code.
467  */
468 void rtnl_af_register(struct rtnl_af_ops *ops)
469 {
470 	rtnl_lock();
471 	list_add_tail(&ops->list, &rtnl_af_ops);
472 	rtnl_unlock();
473 }
474 EXPORT_SYMBOL_GPL(rtnl_af_register);
475 
476 /**
477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
478  * @ops: struct rtnl_af_ops * to unregister
479  *
480  * The caller must hold the rtnl_mutex.
481  */
482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
483 {
484 	list_del(&ops->list);
485 }
486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
487 
488 /**
489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
490  * @ops: struct rtnl_af_ops * to unregister
491  */
492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
493 {
494 	rtnl_lock();
495 	__rtnl_af_unregister(ops);
496 	rtnl_unlock();
497 }
498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
499 
500 static size_t rtnl_link_get_af_size(const struct net_device *dev,
501 				    u32 ext_filter_mask)
502 {
503 	struct rtnl_af_ops *af_ops;
504 	size_t size;
505 
506 	/* IFLA_AF_SPEC */
507 	size = nla_total_size(sizeof(struct nlattr));
508 
509 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
510 		if (af_ops->get_link_af_size) {
511 			/* AF_* + nested data */
512 			size += nla_total_size(sizeof(struct nlattr)) +
513 				af_ops->get_link_af_size(dev, ext_filter_mask);
514 		}
515 	}
516 
517 	return size;
518 }
519 
520 static bool rtnl_have_link_slave_info(const struct net_device *dev)
521 {
522 	struct net_device *master_dev;
523 
524 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
525 	if (master_dev && master_dev->rtnl_link_ops)
526 		return true;
527 	return false;
528 }
529 
530 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
531 				     const struct net_device *dev)
532 {
533 	struct net_device *master_dev;
534 	const struct rtnl_link_ops *ops;
535 	struct nlattr *slave_data;
536 	int err;
537 
538 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
539 	if (!master_dev)
540 		return 0;
541 	ops = master_dev->rtnl_link_ops;
542 	if (!ops)
543 		return 0;
544 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
545 		return -EMSGSIZE;
546 	if (ops->fill_slave_info) {
547 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
548 		if (!slave_data)
549 			return -EMSGSIZE;
550 		err = ops->fill_slave_info(skb, master_dev, dev);
551 		if (err < 0)
552 			goto err_cancel_slave_data;
553 		nla_nest_end(skb, slave_data);
554 	}
555 	return 0;
556 
557 err_cancel_slave_data:
558 	nla_nest_cancel(skb, slave_data);
559 	return err;
560 }
561 
562 static int rtnl_link_info_fill(struct sk_buff *skb,
563 			       const struct net_device *dev)
564 {
565 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
566 	struct nlattr *data;
567 	int err;
568 
569 	if (!ops)
570 		return 0;
571 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
572 		return -EMSGSIZE;
573 	if (ops->fill_xstats) {
574 		err = ops->fill_xstats(skb, dev);
575 		if (err < 0)
576 			return err;
577 	}
578 	if (ops->fill_info) {
579 		data = nla_nest_start(skb, IFLA_INFO_DATA);
580 		if (data == NULL)
581 			return -EMSGSIZE;
582 		err = ops->fill_info(skb, dev);
583 		if (err < 0)
584 			goto err_cancel_data;
585 		nla_nest_end(skb, data);
586 	}
587 	return 0;
588 
589 err_cancel_data:
590 	nla_nest_cancel(skb, data);
591 	return err;
592 }
593 
594 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
595 {
596 	struct nlattr *linkinfo;
597 	int err = -EMSGSIZE;
598 
599 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
600 	if (linkinfo == NULL)
601 		goto out;
602 
603 	err = rtnl_link_info_fill(skb, dev);
604 	if (err < 0)
605 		goto err_cancel_link;
606 
607 	err = rtnl_link_slave_info_fill(skb, dev);
608 	if (err < 0)
609 		goto err_cancel_link;
610 
611 	nla_nest_end(skb, linkinfo);
612 	return 0;
613 
614 err_cancel_link:
615 	nla_nest_cancel(skb, linkinfo);
616 out:
617 	return err;
618 }
619 
620 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
621 {
622 	struct sock *rtnl = net->rtnl;
623 	int err = 0;
624 
625 	NETLINK_CB(skb).dst_group = group;
626 	if (echo)
627 		atomic_inc(&skb->users);
628 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
629 	if (echo)
630 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
631 	return err;
632 }
633 
634 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
635 {
636 	struct sock *rtnl = net->rtnl;
637 
638 	return nlmsg_unicast(rtnl, skb, pid);
639 }
640 EXPORT_SYMBOL(rtnl_unicast);
641 
642 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
643 		 struct nlmsghdr *nlh, gfp_t flags)
644 {
645 	struct sock *rtnl = net->rtnl;
646 	int report = 0;
647 
648 	if (nlh)
649 		report = nlmsg_report(nlh);
650 
651 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
652 }
653 EXPORT_SYMBOL(rtnl_notify);
654 
655 void rtnl_set_sk_err(struct net *net, u32 group, int error)
656 {
657 	struct sock *rtnl = net->rtnl;
658 
659 	netlink_set_err(rtnl, 0, group, error);
660 }
661 EXPORT_SYMBOL(rtnl_set_sk_err);
662 
663 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
664 {
665 	struct nlattr *mx;
666 	int i, valid = 0;
667 
668 	mx = nla_nest_start(skb, RTA_METRICS);
669 	if (mx == NULL)
670 		return -ENOBUFS;
671 
672 	for (i = 0; i < RTAX_MAX; i++) {
673 		if (metrics[i]) {
674 			if (i == RTAX_CC_ALGO - 1) {
675 				char tmp[TCP_CA_NAME_MAX], *name;
676 
677 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
678 				if (!name)
679 					continue;
680 				if (nla_put_string(skb, i + 1, name))
681 					goto nla_put_failure;
682 			} else if (i == RTAX_FEATURES - 1) {
683 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
684 
685 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
686 				if (nla_put_u32(skb, i + 1, user_features))
687 					goto nla_put_failure;
688 			} else {
689 				if (nla_put_u32(skb, i + 1, metrics[i]))
690 					goto nla_put_failure;
691 			}
692 			valid++;
693 		}
694 	}
695 
696 	if (!valid) {
697 		nla_nest_cancel(skb, mx);
698 		return 0;
699 	}
700 
701 	return nla_nest_end(skb, mx);
702 
703 nla_put_failure:
704 	nla_nest_cancel(skb, mx);
705 	return -EMSGSIZE;
706 }
707 EXPORT_SYMBOL(rtnetlink_put_metrics);
708 
709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
710 		       long expires, u32 error)
711 {
712 	struct rta_cacheinfo ci = {
713 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
714 		.rta_used = dst->__use,
715 		.rta_clntref = atomic_read(&(dst->__refcnt)),
716 		.rta_error = error,
717 		.rta_id =  id,
718 	};
719 
720 	if (expires) {
721 		unsigned long clock;
722 
723 		clock = jiffies_to_clock_t(abs(expires));
724 		clock = min_t(unsigned long, clock, INT_MAX);
725 		ci.rta_expires = (expires > 0) ? clock : -clock;
726 	}
727 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
728 }
729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
730 
731 static void set_operstate(struct net_device *dev, unsigned char transition)
732 {
733 	unsigned char operstate = dev->operstate;
734 
735 	switch (transition) {
736 	case IF_OPER_UP:
737 		if ((operstate == IF_OPER_DORMANT ||
738 		     operstate == IF_OPER_UNKNOWN) &&
739 		    !netif_dormant(dev))
740 			operstate = IF_OPER_UP;
741 		break;
742 
743 	case IF_OPER_DORMANT:
744 		if (operstate == IF_OPER_UP ||
745 		    operstate == IF_OPER_UNKNOWN)
746 			operstate = IF_OPER_DORMANT;
747 		break;
748 	}
749 
750 	if (dev->operstate != operstate) {
751 		write_lock_bh(&dev_base_lock);
752 		dev->operstate = operstate;
753 		write_unlock_bh(&dev_base_lock);
754 		netdev_state_change(dev);
755 	}
756 }
757 
758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
759 {
760 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
761 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
762 }
763 
764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
765 					   const struct ifinfomsg *ifm)
766 {
767 	unsigned int flags = ifm->ifi_flags;
768 
769 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
770 	if (ifm->ifi_change)
771 		flags = (flags & ifm->ifi_change) |
772 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
773 
774 	return flags;
775 }
776 
777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
778 				 const struct rtnl_link_stats64 *b)
779 {
780 	a->rx_packets = b->rx_packets;
781 	a->tx_packets = b->tx_packets;
782 	a->rx_bytes = b->rx_bytes;
783 	a->tx_bytes = b->tx_bytes;
784 	a->rx_errors = b->rx_errors;
785 	a->tx_errors = b->tx_errors;
786 	a->rx_dropped = b->rx_dropped;
787 	a->tx_dropped = b->tx_dropped;
788 
789 	a->multicast = b->multicast;
790 	a->collisions = b->collisions;
791 
792 	a->rx_length_errors = b->rx_length_errors;
793 	a->rx_over_errors = b->rx_over_errors;
794 	a->rx_crc_errors = b->rx_crc_errors;
795 	a->rx_frame_errors = b->rx_frame_errors;
796 	a->rx_fifo_errors = b->rx_fifo_errors;
797 	a->rx_missed_errors = b->rx_missed_errors;
798 
799 	a->tx_aborted_errors = b->tx_aborted_errors;
800 	a->tx_carrier_errors = b->tx_carrier_errors;
801 	a->tx_fifo_errors = b->tx_fifo_errors;
802 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
803 	a->tx_window_errors = b->tx_window_errors;
804 
805 	a->rx_compressed = b->rx_compressed;
806 	a->tx_compressed = b->tx_compressed;
807 }
808 
809 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
810 {
811 	memcpy(v, b, sizeof(*b));
812 }
813 
814 /* All VF info */
815 static inline int rtnl_vfinfo_size(const struct net_device *dev,
816 				   u32 ext_filter_mask)
817 {
818 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
819 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
820 		int num_vfs = dev_num_vf(dev->dev.parent);
821 		size_t size = nla_total_size(sizeof(struct nlattr));
822 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
823 		size += num_vfs *
824 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
825 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
826 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
827 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
828 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
829 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
830 			 /* IFLA_VF_STATS_RX_PACKETS */
831 			 nla_total_size(sizeof(__u64)) +
832 			 /* IFLA_VF_STATS_TX_PACKETS */
833 			 nla_total_size(sizeof(__u64)) +
834 			 /* IFLA_VF_STATS_RX_BYTES */
835 			 nla_total_size(sizeof(__u64)) +
836 			 /* IFLA_VF_STATS_TX_BYTES */
837 			 nla_total_size(sizeof(__u64)) +
838 			 /* IFLA_VF_STATS_BROADCAST */
839 			 nla_total_size(sizeof(__u64)) +
840 			 /* IFLA_VF_STATS_MULTICAST */
841 			 nla_total_size(sizeof(__u64)) +
842 			 nla_total_size(sizeof(struct ifla_vf_trust)));
843 		return size;
844 	} else
845 		return 0;
846 }
847 
848 static size_t rtnl_port_size(const struct net_device *dev,
849 			     u32 ext_filter_mask)
850 {
851 	size_t port_size = nla_total_size(4)		/* PORT_VF */
852 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
853 		+ nla_total_size(sizeof(struct ifla_port_vsi))
854 							/* PORT_VSI_TYPE */
855 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
856 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
857 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
858 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
859 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
860 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
861 		+ port_size;
862 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
863 		+ port_size;
864 
865 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
866 	    !(ext_filter_mask & RTEXT_FILTER_VF))
867 		return 0;
868 	if (dev_num_vf(dev->dev.parent))
869 		return port_self_size + vf_ports_size +
870 			vf_port_size * dev_num_vf(dev->dev.parent);
871 	else
872 		return port_self_size;
873 }
874 
875 static noinline size_t if_nlmsg_size(const struct net_device *dev,
876 				     u32 ext_filter_mask)
877 {
878 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
879 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
880 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
881 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
882 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
883 	       + nla_total_size(sizeof(struct rtnl_link_stats))
884 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
885 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
886 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
887 	       + nla_total_size(4) /* IFLA_TXQLEN */
888 	       + nla_total_size(4) /* IFLA_WEIGHT */
889 	       + nla_total_size(4) /* IFLA_MTU */
890 	       + nla_total_size(4) /* IFLA_LINK */
891 	       + nla_total_size(4) /* IFLA_MASTER */
892 	       + nla_total_size(1) /* IFLA_CARRIER */
893 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
894 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
895 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
896 	       + nla_total_size(1) /* IFLA_OPERSTATE */
897 	       + nla_total_size(1) /* IFLA_LINKMODE */
898 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
899 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
900 	       + nla_total_size(ext_filter_mask
901 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
902 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
903 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
904 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
905 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
906 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
907 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
908 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
909 
910 }
911 
912 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
913 {
914 	struct nlattr *vf_ports;
915 	struct nlattr *vf_port;
916 	int vf;
917 	int err;
918 
919 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
920 	if (!vf_ports)
921 		return -EMSGSIZE;
922 
923 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
924 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
925 		if (!vf_port)
926 			goto nla_put_failure;
927 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
928 			goto nla_put_failure;
929 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
930 		if (err == -EMSGSIZE)
931 			goto nla_put_failure;
932 		if (err) {
933 			nla_nest_cancel(skb, vf_port);
934 			continue;
935 		}
936 		nla_nest_end(skb, vf_port);
937 	}
938 
939 	nla_nest_end(skb, vf_ports);
940 
941 	return 0;
942 
943 nla_put_failure:
944 	nla_nest_cancel(skb, vf_ports);
945 	return -EMSGSIZE;
946 }
947 
948 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
949 {
950 	struct nlattr *port_self;
951 	int err;
952 
953 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
954 	if (!port_self)
955 		return -EMSGSIZE;
956 
957 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
958 	if (err) {
959 		nla_nest_cancel(skb, port_self);
960 		return (err == -EMSGSIZE) ? err : 0;
961 	}
962 
963 	nla_nest_end(skb, port_self);
964 
965 	return 0;
966 }
967 
968 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
969 			  u32 ext_filter_mask)
970 {
971 	int err;
972 
973 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
974 	    !(ext_filter_mask & RTEXT_FILTER_VF))
975 		return 0;
976 
977 	err = rtnl_port_self_fill(skb, dev);
978 	if (err)
979 		return err;
980 
981 	if (dev_num_vf(dev->dev.parent)) {
982 		err = rtnl_vf_ports_fill(skb, dev);
983 		if (err)
984 			return err;
985 	}
986 
987 	return 0;
988 }
989 
990 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
991 {
992 	int err;
993 	struct netdev_phys_item_id ppid;
994 
995 	err = dev_get_phys_port_id(dev, &ppid);
996 	if (err) {
997 		if (err == -EOPNOTSUPP)
998 			return 0;
999 		return err;
1000 	}
1001 
1002 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1003 		return -EMSGSIZE;
1004 
1005 	return 0;
1006 }
1007 
1008 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1009 {
1010 	char name[IFNAMSIZ];
1011 	int err;
1012 
1013 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1014 	if (err) {
1015 		if (err == -EOPNOTSUPP)
1016 			return 0;
1017 		return err;
1018 	}
1019 
1020 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1021 		return -EMSGSIZE;
1022 
1023 	return 0;
1024 }
1025 
1026 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1027 {
1028 	int err;
1029 	struct switchdev_attr attr = {
1030 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1031 		.flags = SWITCHDEV_F_NO_RECURSE,
1032 	};
1033 
1034 	err = switchdev_port_attr_get(dev, &attr);
1035 	if (err) {
1036 		if (err == -EOPNOTSUPP)
1037 			return 0;
1038 		return err;
1039 	}
1040 
1041 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1042 		    attr.u.ppid.id))
1043 		return -EMSGSIZE;
1044 
1045 	return 0;
1046 }
1047 
1048 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1049 			    int type, u32 pid, u32 seq, u32 change,
1050 			    unsigned int flags, u32 ext_filter_mask)
1051 {
1052 	struct ifinfomsg *ifm;
1053 	struct nlmsghdr *nlh;
1054 	struct rtnl_link_stats64 temp;
1055 	const struct rtnl_link_stats64 *stats;
1056 	struct nlattr *attr, *af_spec;
1057 	struct rtnl_af_ops *af_ops;
1058 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1059 
1060 	ASSERT_RTNL();
1061 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1062 	if (nlh == NULL)
1063 		return -EMSGSIZE;
1064 
1065 	ifm = nlmsg_data(nlh);
1066 	ifm->ifi_family = AF_UNSPEC;
1067 	ifm->__ifi_pad = 0;
1068 	ifm->ifi_type = dev->type;
1069 	ifm->ifi_index = dev->ifindex;
1070 	ifm->ifi_flags = dev_get_flags(dev);
1071 	ifm->ifi_change = change;
1072 
1073 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1074 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1075 	    nla_put_u8(skb, IFLA_OPERSTATE,
1076 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1077 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1078 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1079 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1080 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1081 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1082 #ifdef CONFIG_RPS
1083 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1084 #endif
1085 	    (dev->ifindex != dev_get_iflink(dev) &&
1086 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1087 	    (upper_dev &&
1088 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1089 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1090 	    (dev->qdisc &&
1091 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1092 	    (dev->ifalias &&
1093 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1094 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1095 			atomic_read(&dev->carrier_changes)) ||
1096 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1097 		goto nla_put_failure;
1098 
1099 	if (1) {
1100 		struct rtnl_link_ifmap map = {
1101 			.mem_start   = dev->mem_start,
1102 			.mem_end     = dev->mem_end,
1103 			.base_addr   = dev->base_addr,
1104 			.irq         = dev->irq,
1105 			.dma         = dev->dma,
1106 			.port        = dev->if_port,
1107 		};
1108 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1109 			goto nla_put_failure;
1110 	}
1111 
1112 	if (dev->addr_len) {
1113 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1114 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1115 			goto nla_put_failure;
1116 	}
1117 
1118 	if (rtnl_phys_port_id_fill(skb, dev))
1119 		goto nla_put_failure;
1120 
1121 	if (rtnl_phys_port_name_fill(skb, dev))
1122 		goto nla_put_failure;
1123 
1124 	if (rtnl_phys_switch_id_fill(skb, dev))
1125 		goto nla_put_failure;
1126 
1127 	attr = nla_reserve(skb, IFLA_STATS,
1128 			sizeof(struct rtnl_link_stats));
1129 	if (attr == NULL)
1130 		goto nla_put_failure;
1131 
1132 	stats = dev_get_stats(dev, &temp);
1133 	copy_rtnl_link_stats(nla_data(attr), stats);
1134 
1135 	attr = nla_reserve(skb, IFLA_STATS64,
1136 			sizeof(struct rtnl_link_stats64));
1137 	if (attr == NULL)
1138 		goto nla_put_failure;
1139 	copy_rtnl_link_stats64(nla_data(attr), stats);
1140 
1141 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1142 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1143 		goto nla_put_failure;
1144 
1145 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1146 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
1147 		int i;
1148 
1149 		struct nlattr *vfinfo, *vf, *vfstats;
1150 		int num_vfs = dev_num_vf(dev->dev.parent);
1151 
1152 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1153 		if (!vfinfo)
1154 			goto nla_put_failure;
1155 		for (i = 0; i < num_vfs; i++) {
1156 			struct ifla_vf_info ivi;
1157 			struct ifla_vf_mac vf_mac;
1158 			struct ifla_vf_vlan vf_vlan;
1159 			struct ifla_vf_rate vf_rate;
1160 			struct ifla_vf_tx_rate vf_tx_rate;
1161 			struct ifla_vf_spoofchk vf_spoofchk;
1162 			struct ifla_vf_link_state vf_linkstate;
1163 			struct ifla_vf_rss_query_en vf_rss_query_en;
1164 			struct ifla_vf_stats vf_stats;
1165 			struct ifla_vf_trust vf_trust;
1166 
1167 			/*
1168 			 * Not all SR-IOV capable drivers support the
1169 			 * spoofcheck and "RSS query enable" query.  Preset to
1170 			 * -1 so the user space tool can detect that the driver
1171 			 * didn't report anything.
1172 			 */
1173 			ivi.spoofchk = -1;
1174 			ivi.rss_query_en = -1;
1175 			ivi.trusted = -1;
1176 			memset(ivi.mac, 0, sizeof(ivi.mac));
1177 			/* The default value for VF link state is "auto"
1178 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
1179 			 */
1180 			ivi.linkstate = 0;
1181 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1182 				break;
1183 			vf_mac.vf =
1184 				vf_vlan.vf =
1185 				vf_rate.vf =
1186 				vf_tx_rate.vf =
1187 				vf_spoofchk.vf =
1188 				vf_linkstate.vf =
1189 				vf_rss_query_en.vf =
1190 				vf_trust.vf = ivi.vf;
1191 
1192 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1193 			vf_vlan.vlan = ivi.vlan;
1194 			vf_vlan.qos = ivi.qos;
1195 			vf_tx_rate.rate = ivi.max_tx_rate;
1196 			vf_rate.min_tx_rate = ivi.min_tx_rate;
1197 			vf_rate.max_tx_rate = ivi.max_tx_rate;
1198 			vf_spoofchk.setting = ivi.spoofchk;
1199 			vf_linkstate.link_state = ivi.linkstate;
1200 			vf_rss_query_en.setting = ivi.rss_query_en;
1201 			vf_trust.setting = ivi.trusted;
1202 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1203 			if (!vf) {
1204 				nla_nest_cancel(skb, vfinfo);
1205 				goto nla_put_failure;
1206 			}
1207 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1208 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1209 			    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1210 				    &vf_rate) ||
1211 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1212 				    &vf_tx_rate) ||
1213 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1214 				    &vf_spoofchk) ||
1215 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1216 				    &vf_linkstate) ||
1217 			    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1218 				    sizeof(vf_rss_query_en),
1219 				    &vf_rss_query_en) ||
1220 			    nla_put(skb, IFLA_VF_TRUST,
1221 				    sizeof(vf_trust), &vf_trust))
1222 				goto nla_put_failure;
1223 			memset(&vf_stats, 0, sizeof(vf_stats));
1224 			if (dev->netdev_ops->ndo_get_vf_stats)
1225 				dev->netdev_ops->ndo_get_vf_stats(dev, i,
1226 								  &vf_stats);
1227 			vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1228 			if (!vfstats) {
1229 				nla_nest_cancel(skb, vf);
1230 				nla_nest_cancel(skb, vfinfo);
1231 				goto nla_put_failure;
1232 			}
1233 			if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1234 					vf_stats.rx_packets) ||
1235 			    nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1236 					vf_stats.tx_packets) ||
1237 			    nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1238 					vf_stats.rx_bytes) ||
1239 			    nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1240 					vf_stats.tx_bytes) ||
1241 			    nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1242 					vf_stats.broadcast) ||
1243 			    nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1244 					vf_stats.multicast))
1245 				goto nla_put_failure;
1246 			nla_nest_end(skb, vfstats);
1247 			nla_nest_end(skb, vf);
1248 		}
1249 		nla_nest_end(skb, vfinfo);
1250 	}
1251 
1252 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1253 		goto nla_put_failure;
1254 
1255 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1256 		if (rtnl_link_fill(skb, dev) < 0)
1257 			goto nla_put_failure;
1258 	}
1259 
1260 	if (dev->rtnl_link_ops &&
1261 	    dev->rtnl_link_ops->get_link_net) {
1262 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1263 
1264 		if (!net_eq(dev_net(dev), link_net)) {
1265 			int id = peernet2id_alloc(dev_net(dev), link_net);
1266 
1267 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1268 				goto nla_put_failure;
1269 		}
1270 	}
1271 
1272 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1273 		goto nla_put_failure;
1274 
1275 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1276 		if (af_ops->fill_link_af) {
1277 			struct nlattr *af;
1278 			int err;
1279 
1280 			if (!(af = nla_nest_start(skb, af_ops->family)))
1281 				goto nla_put_failure;
1282 
1283 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1284 
1285 			/*
1286 			 * Caller may return ENODATA to indicate that there
1287 			 * was no data to be dumped. This is not an error, it
1288 			 * means we should trim the attribute header and
1289 			 * continue.
1290 			 */
1291 			if (err == -ENODATA)
1292 				nla_nest_cancel(skb, af);
1293 			else if (err < 0)
1294 				goto nla_put_failure;
1295 
1296 			nla_nest_end(skb, af);
1297 		}
1298 	}
1299 
1300 	nla_nest_end(skb, af_spec);
1301 
1302 	nlmsg_end(skb, nlh);
1303 	return 0;
1304 
1305 nla_put_failure:
1306 	nlmsg_cancel(skb, nlh);
1307 	return -EMSGSIZE;
1308 }
1309 
1310 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1311 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1312 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1313 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1314 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1315 	[IFLA_MTU]		= { .type = NLA_U32 },
1316 	[IFLA_LINK]		= { .type = NLA_U32 },
1317 	[IFLA_MASTER]		= { .type = NLA_U32 },
1318 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1319 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1320 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1321 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1322 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1323 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1324 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1325 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1326 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1327 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1328 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1329 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1330 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1331 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1332 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1333 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1334 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1335 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1336 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1337 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1338 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1339 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1340 };
1341 
1342 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1343 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1344 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1345 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1346 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1347 };
1348 
1349 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1350 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1351 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1352 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1353 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1354 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1355 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1356 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1357 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1358 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1359 };
1360 
1361 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1362 	[IFLA_VF_STATS_RX_PACKETS]	= { .type = NLA_U64 },
1363 	[IFLA_VF_STATS_TX_PACKETS]	= { .type = NLA_U64 },
1364 	[IFLA_VF_STATS_RX_BYTES]	= { .type = NLA_U64 },
1365 	[IFLA_VF_STATS_TX_BYTES]	= { .type = NLA_U64 },
1366 	[IFLA_VF_STATS_BROADCAST]	= { .type = NLA_U64 },
1367 	[IFLA_VF_STATS_MULTICAST]	= { .type = NLA_U64 },
1368 };
1369 
1370 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1371 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1372 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1373 				    .len = PORT_PROFILE_MAX },
1374 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1375 				    .len = sizeof(struct ifla_port_vsi)},
1376 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1377 				      .len = PORT_UUID_MAX },
1378 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1379 				    .len = PORT_UUID_MAX },
1380 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1381 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1382 };
1383 
1384 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1385 {
1386 	struct net *net = sock_net(skb->sk);
1387 	int h, s_h;
1388 	int idx = 0, s_idx;
1389 	struct net_device *dev;
1390 	struct hlist_head *head;
1391 	struct nlattr *tb[IFLA_MAX+1];
1392 	u32 ext_filter_mask = 0;
1393 	int err;
1394 	int hdrlen;
1395 
1396 	s_h = cb->args[0];
1397 	s_idx = cb->args[1];
1398 
1399 	cb->seq = net->dev_base_seq;
1400 
1401 	/* A hack to preserve kernel<->userspace interface.
1402 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1403 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1404 	 * what iproute2 < v3.9.0 used.
1405 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1406 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1407 	 */
1408 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1409 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1410 
1411 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1412 
1413 		if (tb[IFLA_EXT_MASK])
1414 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1415 	}
1416 
1417 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1418 		idx = 0;
1419 		head = &net->dev_index_head[h];
1420 		hlist_for_each_entry(dev, head, index_hlist) {
1421 			if (idx < s_idx)
1422 				goto cont;
1423 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1424 					       NETLINK_CB(cb->skb).portid,
1425 					       cb->nlh->nlmsg_seq, 0,
1426 					       NLM_F_MULTI,
1427 					       ext_filter_mask);
1428 			/* If we ran out of room on the first message,
1429 			 * we're in trouble
1430 			 */
1431 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1432 
1433 			if (err < 0)
1434 				goto out;
1435 
1436 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1437 cont:
1438 			idx++;
1439 		}
1440 	}
1441 out:
1442 	cb->args[1] = idx;
1443 	cb->args[0] = h;
1444 
1445 	return skb->len;
1446 }
1447 
1448 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1449 {
1450 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1451 }
1452 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1453 
1454 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1455 {
1456 	struct net *net;
1457 	/* Examine the link attributes and figure out which
1458 	 * network namespace we are talking about.
1459 	 */
1460 	if (tb[IFLA_NET_NS_PID])
1461 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1462 	else if (tb[IFLA_NET_NS_FD])
1463 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1464 	else
1465 		net = get_net(src_net);
1466 	return net;
1467 }
1468 EXPORT_SYMBOL(rtnl_link_get_net);
1469 
1470 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1471 {
1472 	if (dev) {
1473 		if (tb[IFLA_ADDRESS] &&
1474 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1475 			return -EINVAL;
1476 
1477 		if (tb[IFLA_BROADCAST] &&
1478 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1479 			return -EINVAL;
1480 	}
1481 
1482 	if (tb[IFLA_AF_SPEC]) {
1483 		struct nlattr *af;
1484 		int rem, err;
1485 
1486 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1487 			const struct rtnl_af_ops *af_ops;
1488 
1489 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1490 				return -EAFNOSUPPORT;
1491 
1492 			if (!af_ops->set_link_af)
1493 				return -EOPNOTSUPP;
1494 
1495 			if (af_ops->validate_link_af) {
1496 				err = af_ops->validate_link_af(dev, af);
1497 				if (err < 0)
1498 					return err;
1499 			}
1500 		}
1501 	}
1502 
1503 	return 0;
1504 }
1505 
1506 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1507 {
1508 	const struct net_device_ops *ops = dev->netdev_ops;
1509 	int err = -EINVAL;
1510 
1511 	if (tb[IFLA_VF_MAC]) {
1512 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1513 
1514 		err = -EOPNOTSUPP;
1515 		if (ops->ndo_set_vf_mac)
1516 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1517 						  ivm->mac);
1518 		if (err < 0)
1519 			return err;
1520 	}
1521 
1522 	if (tb[IFLA_VF_VLAN]) {
1523 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1524 
1525 		err = -EOPNOTSUPP;
1526 		if (ops->ndo_set_vf_vlan)
1527 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1528 						   ivv->qos);
1529 		if (err < 0)
1530 			return err;
1531 	}
1532 
1533 	if (tb[IFLA_VF_TX_RATE]) {
1534 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1535 		struct ifla_vf_info ivf;
1536 
1537 		err = -EOPNOTSUPP;
1538 		if (ops->ndo_get_vf_config)
1539 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1540 		if (err < 0)
1541 			return err;
1542 
1543 		err = -EOPNOTSUPP;
1544 		if (ops->ndo_set_vf_rate)
1545 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1546 						   ivf.min_tx_rate,
1547 						   ivt->rate);
1548 		if (err < 0)
1549 			return err;
1550 	}
1551 
1552 	if (tb[IFLA_VF_RATE]) {
1553 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1554 
1555 		err = -EOPNOTSUPP;
1556 		if (ops->ndo_set_vf_rate)
1557 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1558 						   ivt->min_tx_rate,
1559 						   ivt->max_tx_rate);
1560 		if (err < 0)
1561 			return err;
1562 	}
1563 
1564 	if (tb[IFLA_VF_SPOOFCHK]) {
1565 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1566 
1567 		err = -EOPNOTSUPP;
1568 		if (ops->ndo_set_vf_spoofchk)
1569 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1570 						       ivs->setting);
1571 		if (err < 0)
1572 			return err;
1573 	}
1574 
1575 	if (tb[IFLA_VF_LINK_STATE]) {
1576 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1577 
1578 		err = -EOPNOTSUPP;
1579 		if (ops->ndo_set_vf_link_state)
1580 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1581 							 ivl->link_state);
1582 		if (err < 0)
1583 			return err;
1584 	}
1585 
1586 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1587 		struct ifla_vf_rss_query_en *ivrssq_en;
1588 
1589 		err = -EOPNOTSUPP;
1590 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1591 		if (ops->ndo_set_vf_rss_query_en)
1592 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1593 							   ivrssq_en->setting);
1594 		if (err < 0)
1595 			return err;
1596 	}
1597 
1598 	if (tb[IFLA_VF_TRUST]) {
1599 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1600 
1601 		err = -EOPNOTSUPP;
1602 		if (ops->ndo_set_vf_trust)
1603 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1604 		if (err < 0)
1605 			return err;
1606 	}
1607 
1608 	return err;
1609 }
1610 
1611 static int do_set_master(struct net_device *dev, int ifindex)
1612 {
1613 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1614 	const struct net_device_ops *ops;
1615 	int err;
1616 
1617 	if (upper_dev) {
1618 		if (upper_dev->ifindex == ifindex)
1619 			return 0;
1620 		ops = upper_dev->netdev_ops;
1621 		if (ops->ndo_del_slave) {
1622 			err = ops->ndo_del_slave(upper_dev, dev);
1623 			if (err)
1624 				return err;
1625 		} else {
1626 			return -EOPNOTSUPP;
1627 		}
1628 	}
1629 
1630 	if (ifindex) {
1631 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1632 		if (!upper_dev)
1633 			return -EINVAL;
1634 		ops = upper_dev->netdev_ops;
1635 		if (ops->ndo_add_slave) {
1636 			err = ops->ndo_add_slave(upper_dev, dev);
1637 			if (err)
1638 				return err;
1639 		} else {
1640 			return -EOPNOTSUPP;
1641 		}
1642 	}
1643 	return 0;
1644 }
1645 
1646 #define DO_SETLINK_MODIFIED	0x01
1647 /* notify flag means notify + modified. */
1648 #define DO_SETLINK_NOTIFY	0x03
1649 static int do_setlink(const struct sk_buff *skb,
1650 		      struct net_device *dev, struct ifinfomsg *ifm,
1651 		      struct nlattr **tb, char *ifname, int status)
1652 {
1653 	const struct net_device_ops *ops = dev->netdev_ops;
1654 	int err;
1655 
1656 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1657 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1658 		if (IS_ERR(net)) {
1659 			err = PTR_ERR(net);
1660 			goto errout;
1661 		}
1662 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1663 			put_net(net);
1664 			err = -EPERM;
1665 			goto errout;
1666 		}
1667 		err = dev_change_net_namespace(dev, net, ifname);
1668 		put_net(net);
1669 		if (err)
1670 			goto errout;
1671 		status |= DO_SETLINK_MODIFIED;
1672 	}
1673 
1674 	if (tb[IFLA_MAP]) {
1675 		struct rtnl_link_ifmap *u_map;
1676 		struct ifmap k_map;
1677 
1678 		if (!ops->ndo_set_config) {
1679 			err = -EOPNOTSUPP;
1680 			goto errout;
1681 		}
1682 
1683 		if (!netif_device_present(dev)) {
1684 			err = -ENODEV;
1685 			goto errout;
1686 		}
1687 
1688 		u_map = nla_data(tb[IFLA_MAP]);
1689 		k_map.mem_start = (unsigned long) u_map->mem_start;
1690 		k_map.mem_end = (unsigned long) u_map->mem_end;
1691 		k_map.base_addr = (unsigned short) u_map->base_addr;
1692 		k_map.irq = (unsigned char) u_map->irq;
1693 		k_map.dma = (unsigned char) u_map->dma;
1694 		k_map.port = (unsigned char) u_map->port;
1695 
1696 		err = ops->ndo_set_config(dev, &k_map);
1697 		if (err < 0)
1698 			goto errout;
1699 
1700 		status |= DO_SETLINK_NOTIFY;
1701 	}
1702 
1703 	if (tb[IFLA_ADDRESS]) {
1704 		struct sockaddr *sa;
1705 		int len;
1706 
1707 		len = sizeof(sa_family_t) + dev->addr_len;
1708 		sa = kmalloc(len, GFP_KERNEL);
1709 		if (!sa) {
1710 			err = -ENOMEM;
1711 			goto errout;
1712 		}
1713 		sa->sa_family = dev->type;
1714 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1715 		       dev->addr_len);
1716 		err = dev_set_mac_address(dev, sa);
1717 		kfree(sa);
1718 		if (err)
1719 			goto errout;
1720 		status |= DO_SETLINK_MODIFIED;
1721 	}
1722 
1723 	if (tb[IFLA_MTU]) {
1724 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1725 		if (err < 0)
1726 			goto errout;
1727 		status |= DO_SETLINK_MODIFIED;
1728 	}
1729 
1730 	if (tb[IFLA_GROUP]) {
1731 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1732 		status |= DO_SETLINK_NOTIFY;
1733 	}
1734 
1735 	/*
1736 	 * Interface selected by interface index but interface
1737 	 * name provided implies that a name change has been
1738 	 * requested.
1739 	 */
1740 	if (ifm->ifi_index > 0 && ifname[0]) {
1741 		err = dev_change_name(dev, ifname);
1742 		if (err < 0)
1743 			goto errout;
1744 		status |= DO_SETLINK_MODIFIED;
1745 	}
1746 
1747 	if (tb[IFLA_IFALIAS]) {
1748 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1749 				    nla_len(tb[IFLA_IFALIAS]));
1750 		if (err < 0)
1751 			goto errout;
1752 		status |= DO_SETLINK_NOTIFY;
1753 	}
1754 
1755 	if (tb[IFLA_BROADCAST]) {
1756 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1757 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1758 	}
1759 
1760 	if (ifm->ifi_flags || ifm->ifi_change) {
1761 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1762 		if (err < 0)
1763 			goto errout;
1764 	}
1765 
1766 	if (tb[IFLA_MASTER]) {
1767 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1768 		if (err)
1769 			goto errout;
1770 		status |= DO_SETLINK_MODIFIED;
1771 	}
1772 
1773 	if (tb[IFLA_CARRIER]) {
1774 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1775 		if (err)
1776 			goto errout;
1777 		status |= DO_SETLINK_MODIFIED;
1778 	}
1779 
1780 	if (tb[IFLA_TXQLEN]) {
1781 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1782 
1783 		if (dev->tx_queue_len ^ value)
1784 			status |= DO_SETLINK_NOTIFY;
1785 
1786 		dev->tx_queue_len = value;
1787 	}
1788 
1789 	if (tb[IFLA_OPERSTATE])
1790 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1791 
1792 	if (tb[IFLA_LINKMODE]) {
1793 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1794 
1795 		write_lock_bh(&dev_base_lock);
1796 		if (dev->link_mode ^ value)
1797 			status |= DO_SETLINK_NOTIFY;
1798 		dev->link_mode = value;
1799 		write_unlock_bh(&dev_base_lock);
1800 	}
1801 
1802 	if (tb[IFLA_VFINFO_LIST]) {
1803 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1804 		struct nlattr *attr;
1805 		int rem;
1806 
1807 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1808 			if (nla_type(attr) != IFLA_VF_INFO ||
1809 			    nla_len(attr) < NLA_HDRLEN) {
1810 				err = -EINVAL;
1811 				goto errout;
1812 			}
1813 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1814 					       ifla_vf_policy);
1815 			if (err < 0)
1816 				goto errout;
1817 			err = do_setvfinfo(dev, vfinfo);
1818 			if (err < 0)
1819 				goto errout;
1820 			status |= DO_SETLINK_NOTIFY;
1821 		}
1822 	}
1823 	err = 0;
1824 
1825 	if (tb[IFLA_VF_PORTS]) {
1826 		struct nlattr *port[IFLA_PORT_MAX+1];
1827 		struct nlattr *attr;
1828 		int vf;
1829 		int rem;
1830 
1831 		err = -EOPNOTSUPP;
1832 		if (!ops->ndo_set_vf_port)
1833 			goto errout;
1834 
1835 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1836 			if (nla_type(attr) != IFLA_VF_PORT ||
1837 			    nla_len(attr) < NLA_HDRLEN) {
1838 				err = -EINVAL;
1839 				goto errout;
1840 			}
1841 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1842 					       ifla_port_policy);
1843 			if (err < 0)
1844 				goto errout;
1845 			if (!port[IFLA_PORT_VF]) {
1846 				err = -EOPNOTSUPP;
1847 				goto errout;
1848 			}
1849 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1850 			err = ops->ndo_set_vf_port(dev, vf, port);
1851 			if (err < 0)
1852 				goto errout;
1853 			status |= DO_SETLINK_NOTIFY;
1854 		}
1855 	}
1856 	err = 0;
1857 
1858 	if (tb[IFLA_PORT_SELF]) {
1859 		struct nlattr *port[IFLA_PORT_MAX+1];
1860 
1861 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1862 			tb[IFLA_PORT_SELF], ifla_port_policy);
1863 		if (err < 0)
1864 			goto errout;
1865 
1866 		err = -EOPNOTSUPP;
1867 		if (ops->ndo_set_vf_port)
1868 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1869 		if (err < 0)
1870 			goto errout;
1871 		status |= DO_SETLINK_NOTIFY;
1872 	}
1873 
1874 	if (tb[IFLA_AF_SPEC]) {
1875 		struct nlattr *af;
1876 		int rem;
1877 
1878 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1879 			const struct rtnl_af_ops *af_ops;
1880 
1881 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1882 				BUG();
1883 
1884 			err = af_ops->set_link_af(dev, af);
1885 			if (err < 0)
1886 				goto errout;
1887 
1888 			status |= DO_SETLINK_NOTIFY;
1889 		}
1890 	}
1891 	err = 0;
1892 
1893 	if (tb[IFLA_PROTO_DOWN]) {
1894 		err = dev_change_proto_down(dev,
1895 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
1896 		if (err)
1897 			goto errout;
1898 		status |= DO_SETLINK_NOTIFY;
1899 	}
1900 
1901 errout:
1902 	if (status & DO_SETLINK_MODIFIED) {
1903 		if (status & DO_SETLINK_NOTIFY)
1904 			netdev_state_change(dev);
1905 
1906 		if (err < 0)
1907 			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",
1908 					     dev->name);
1909 	}
1910 
1911 	return err;
1912 }
1913 
1914 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1915 {
1916 	struct net *net = sock_net(skb->sk);
1917 	struct ifinfomsg *ifm;
1918 	struct net_device *dev;
1919 	int err;
1920 	struct nlattr *tb[IFLA_MAX+1];
1921 	char ifname[IFNAMSIZ];
1922 
1923 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1924 	if (err < 0)
1925 		goto errout;
1926 
1927 	if (tb[IFLA_IFNAME])
1928 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1929 	else
1930 		ifname[0] = '\0';
1931 
1932 	err = -EINVAL;
1933 	ifm = nlmsg_data(nlh);
1934 	if (ifm->ifi_index > 0)
1935 		dev = __dev_get_by_index(net, ifm->ifi_index);
1936 	else if (tb[IFLA_IFNAME])
1937 		dev = __dev_get_by_name(net, ifname);
1938 	else
1939 		goto errout;
1940 
1941 	if (dev == NULL) {
1942 		err = -ENODEV;
1943 		goto errout;
1944 	}
1945 
1946 	err = validate_linkmsg(dev, tb);
1947 	if (err < 0)
1948 		goto errout;
1949 
1950 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1951 errout:
1952 	return err;
1953 }
1954 
1955 static int rtnl_group_dellink(const struct net *net, int group)
1956 {
1957 	struct net_device *dev, *aux;
1958 	LIST_HEAD(list_kill);
1959 	bool found = false;
1960 
1961 	if (!group)
1962 		return -EPERM;
1963 
1964 	for_each_netdev(net, dev) {
1965 		if (dev->group == group) {
1966 			const struct rtnl_link_ops *ops;
1967 
1968 			found = true;
1969 			ops = dev->rtnl_link_ops;
1970 			if (!ops || !ops->dellink)
1971 				return -EOPNOTSUPP;
1972 		}
1973 	}
1974 
1975 	if (!found)
1976 		return -ENODEV;
1977 
1978 	for_each_netdev_safe(net, dev, aux) {
1979 		if (dev->group == group) {
1980 			const struct rtnl_link_ops *ops;
1981 
1982 			ops = dev->rtnl_link_ops;
1983 			ops->dellink(dev, &list_kill);
1984 		}
1985 	}
1986 	unregister_netdevice_many(&list_kill);
1987 
1988 	return 0;
1989 }
1990 
1991 int rtnl_delete_link(struct net_device *dev)
1992 {
1993 	const struct rtnl_link_ops *ops;
1994 	LIST_HEAD(list_kill);
1995 
1996 	ops = dev->rtnl_link_ops;
1997 	if (!ops || !ops->dellink)
1998 		return -EOPNOTSUPP;
1999 
2000 	ops->dellink(dev, &list_kill);
2001 	unregister_netdevice_many(&list_kill);
2002 
2003 	return 0;
2004 }
2005 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2006 
2007 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2008 {
2009 	struct net *net = sock_net(skb->sk);
2010 	struct net_device *dev;
2011 	struct ifinfomsg *ifm;
2012 	char ifname[IFNAMSIZ];
2013 	struct nlattr *tb[IFLA_MAX+1];
2014 	int err;
2015 
2016 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2017 	if (err < 0)
2018 		return err;
2019 
2020 	if (tb[IFLA_IFNAME])
2021 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2022 
2023 	ifm = nlmsg_data(nlh);
2024 	if (ifm->ifi_index > 0)
2025 		dev = __dev_get_by_index(net, ifm->ifi_index);
2026 	else if (tb[IFLA_IFNAME])
2027 		dev = __dev_get_by_name(net, ifname);
2028 	else if (tb[IFLA_GROUP])
2029 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2030 	else
2031 		return -EINVAL;
2032 
2033 	if (!dev)
2034 		return -ENODEV;
2035 
2036 	return rtnl_delete_link(dev);
2037 }
2038 
2039 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2040 {
2041 	unsigned int old_flags;
2042 	int err;
2043 
2044 	old_flags = dev->flags;
2045 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2046 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2047 		if (err < 0)
2048 			return err;
2049 	}
2050 
2051 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2052 
2053 	__dev_notify_flags(dev, old_flags, ~0U);
2054 	return 0;
2055 }
2056 EXPORT_SYMBOL(rtnl_configure_link);
2057 
2058 struct net_device *rtnl_create_link(struct net *net,
2059 	const char *ifname, unsigned char name_assign_type,
2060 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2061 {
2062 	int err;
2063 	struct net_device *dev;
2064 	unsigned int num_tx_queues = 1;
2065 	unsigned int num_rx_queues = 1;
2066 
2067 	if (tb[IFLA_NUM_TX_QUEUES])
2068 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2069 	else if (ops->get_num_tx_queues)
2070 		num_tx_queues = ops->get_num_tx_queues();
2071 
2072 	if (tb[IFLA_NUM_RX_QUEUES])
2073 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2074 	else if (ops->get_num_rx_queues)
2075 		num_rx_queues = ops->get_num_rx_queues();
2076 
2077 	err = -ENOMEM;
2078 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2079 			       ops->setup, num_tx_queues, num_rx_queues);
2080 	if (!dev)
2081 		goto err;
2082 
2083 	dev_net_set(dev, net);
2084 	dev->rtnl_link_ops = ops;
2085 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2086 
2087 	if (tb[IFLA_MTU])
2088 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2089 	if (tb[IFLA_ADDRESS]) {
2090 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2091 				nla_len(tb[IFLA_ADDRESS]));
2092 		dev->addr_assign_type = NET_ADDR_SET;
2093 	}
2094 	if (tb[IFLA_BROADCAST])
2095 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2096 				nla_len(tb[IFLA_BROADCAST]));
2097 	if (tb[IFLA_TXQLEN])
2098 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2099 	if (tb[IFLA_OPERSTATE])
2100 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2101 	if (tb[IFLA_LINKMODE])
2102 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2103 	if (tb[IFLA_GROUP])
2104 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2105 
2106 	return dev;
2107 
2108 err:
2109 	return ERR_PTR(err);
2110 }
2111 EXPORT_SYMBOL(rtnl_create_link);
2112 
2113 static int rtnl_group_changelink(const struct sk_buff *skb,
2114 		struct net *net, int group,
2115 		struct ifinfomsg *ifm,
2116 		struct nlattr **tb)
2117 {
2118 	struct net_device *dev, *aux;
2119 	int err;
2120 
2121 	for_each_netdev_safe(net, dev, aux) {
2122 		if (dev->group == group) {
2123 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2124 			if (err < 0)
2125 				return err;
2126 		}
2127 	}
2128 
2129 	return 0;
2130 }
2131 
2132 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2133 {
2134 	struct net *net = sock_net(skb->sk);
2135 	const struct rtnl_link_ops *ops;
2136 	const struct rtnl_link_ops *m_ops = NULL;
2137 	struct net_device *dev;
2138 	struct net_device *master_dev = NULL;
2139 	struct ifinfomsg *ifm;
2140 	char kind[MODULE_NAME_LEN];
2141 	char ifname[IFNAMSIZ];
2142 	struct nlattr *tb[IFLA_MAX+1];
2143 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2144 	unsigned char name_assign_type = NET_NAME_USER;
2145 	int err;
2146 
2147 #ifdef CONFIG_MODULES
2148 replay:
2149 #endif
2150 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2151 	if (err < 0)
2152 		return err;
2153 
2154 	if (tb[IFLA_IFNAME])
2155 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2156 	else
2157 		ifname[0] = '\0';
2158 
2159 	ifm = nlmsg_data(nlh);
2160 	if (ifm->ifi_index > 0)
2161 		dev = __dev_get_by_index(net, ifm->ifi_index);
2162 	else {
2163 		if (ifname[0])
2164 			dev = __dev_get_by_name(net, ifname);
2165 		else
2166 			dev = NULL;
2167 	}
2168 
2169 	if (dev) {
2170 		master_dev = netdev_master_upper_dev_get(dev);
2171 		if (master_dev)
2172 			m_ops = master_dev->rtnl_link_ops;
2173 	}
2174 
2175 	err = validate_linkmsg(dev, tb);
2176 	if (err < 0)
2177 		return err;
2178 
2179 	if (tb[IFLA_LINKINFO]) {
2180 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2181 				       tb[IFLA_LINKINFO], ifla_info_policy);
2182 		if (err < 0)
2183 			return err;
2184 	} else
2185 		memset(linkinfo, 0, sizeof(linkinfo));
2186 
2187 	if (linkinfo[IFLA_INFO_KIND]) {
2188 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2189 		ops = rtnl_link_ops_get(kind);
2190 	} else {
2191 		kind[0] = '\0';
2192 		ops = NULL;
2193 	}
2194 
2195 	if (1) {
2196 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2197 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2198 		struct nlattr **data = NULL;
2199 		struct nlattr **slave_data = NULL;
2200 		struct net *dest_net, *link_net = NULL;
2201 
2202 		if (ops) {
2203 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2204 				err = nla_parse_nested(attr, ops->maxtype,
2205 						       linkinfo[IFLA_INFO_DATA],
2206 						       ops->policy);
2207 				if (err < 0)
2208 					return err;
2209 				data = attr;
2210 			}
2211 			if (ops->validate) {
2212 				err = ops->validate(tb, data);
2213 				if (err < 0)
2214 					return err;
2215 			}
2216 		}
2217 
2218 		if (m_ops) {
2219 			if (m_ops->slave_maxtype &&
2220 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2221 				err = nla_parse_nested(slave_attr,
2222 						       m_ops->slave_maxtype,
2223 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2224 						       m_ops->slave_policy);
2225 				if (err < 0)
2226 					return err;
2227 				slave_data = slave_attr;
2228 			}
2229 			if (m_ops->slave_validate) {
2230 				err = m_ops->slave_validate(tb, slave_data);
2231 				if (err < 0)
2232 					return err;
2233 			}
2234 		}
2235 
2236 		if (dev) {
2237 			int status = 0;
2238 
2239 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2240 				return -EEXIST;
2241 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2242 				return -EOPNOTSUPP;
2243 
2244 			if (linkinfo[IFLA_INFO_DATA]) {
2245 				if (!ops || ops != dev->rtnl_link_ops ||
2246 				    !ops->changelink)
2247 					return -EOPNOTSUPP;
2248 
2249 				err = ops->changelink(dev, tb, data);
2250 				if (err < 0)
2251 					return err;
2252 				status |= DO_SETLINK_NOTIFY;
2253 			}
2254 
2255 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2256 				if (!m_ops || !m_ops->slave_changelink)
2257 					return -EOPNOTSUPP;
2258 
2259 				err = m_ops->slave_changelink(master_dev, dev,
2260 							      tb, slave_data);
2261 				if (err < 0)
2262 					return err;
2263 				status |= DO_SETLINK_NOTIFY;
2264 			}
2265 
2266 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2267 		}
2268 
2269 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2270 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2271 				return rtnl_group_changelink(skb, net,
2272 						nla_get_u32(tb[IFLA_GROUP]),
2273 						ifm, tb);
2274 			return -ENODEV;
2275 		}
2276 
2277 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2278 			return -EOPNOTSUPP;
2279 
2280 		if (!ops) {
2281 #ifdef CONFIG_MODULES
2282 			if (kind[0]) {
2283 				__rtnl_unlock();
2284 				request_module("rtnl-link-%s", kind);
2285 				rtnl_lock();
2286 				ops = rtnl_link_ops_get(kind);
2287 				if (ops)
2288 					goto replay;
2289 			}
2290 #endif
2291 			return -EOPNOTSUPP;
2292 		}
2293 
2294 		if (!ops->setup)
2295 			return -EOPNOTSUPP;
2296 
2297 		if (!ifname[0]) {
2298 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2299 			name_assign_type = NET_NAME_ENUM;
2300 		}
2301 
2302 		dest_net = rtnl_link_get_net(net, tb);
2303 		if (IS_ERR(dest_net))
2304 			return PTR_ERR(dest_net);
2305 
2306 		err = -EPERM;
2307 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2308 			goto out;
2309 
2310 		if (tb[IFLA_LINK_NETNSID]) {
2311 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2312 
2313 			link_net = get_net_ns_by_id(dest_net, id);
2314 			if (!link_net) {
2315 				err =  -EINVAL;
2316 				goto out;
2317 			}
2318 			err = -EPERM;
2319 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2320 				goto out;
2321 		}
2322 
2323 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2324 				       name_assign_type, ops, tb);
2325 		if (IS_ERR(dev)) {
2326 			err = PTR_ERR(dev);
2327 			goto out;
2328 		}
2329 
2330 		dev->ifindex = ifm->ifi_index;
2331 
2332 		if (ops->newlink) {
2333 			err = ops->newlink(link_net ? : net, dev, tb, data);
2334 			/* Drivers should call free_netdev() in ->destructor
2335 			 * and unregister it on failure after registration
2336 			 * so that device could be finally freed in rtnl_unlock.
2337 			 */
2338 			if (err < 0) {
2339 				/* If device is not registered at all, free it now */
2340 				if (dev->reg_state == NETREG_UNINITIALIZED)
2341 					free_netdev(dev);
2342 				goto out;
2343 			}
2344 		} else {
2345 			err = register_netdevice(dev);
2346 			if (err < 0) {
2347 				free_netdev(dev);
2348 				goto out;
2349 			}
2350 		}
2351 		err = rtnl_configure_link(dev, ifm);
2352 		if (err < 0)
2353 			goto out_unregister;
2354 		if (link_net) {
2355 			err = dev_change_net_namespace(dev, dest_net, ifname);
2356 			if (err < 0)
2357 				goto out_unregister;
2358 		}
2359 out:
2360 		if (link_net)
2361 			put_net(link_net);
2362 		put_net(dest_net);
2363 		return err;
2364 out_unregister:
2365 		if (ops->newlink) {
2366 			LIST_HEAD(list_kill);
2367 
2368 			ops->dellink(dev, &list_kill);
2369 			unregister_netdevice_many(&list_kill);
2370 		} else {
2371 			unregister_netdevice(dev);
2372 		}
2373 		goto out;
2374 	}
2375 }
2376 
2377 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2378 {
2379 	struct net *net = sock_net(skb->sk);
2380 	struct ifinfomsg *ifm;
2381 	char ifname[IFNAMSIZ];
2382 	struct nlattr *tb[IFLA_MAX+1];
2383 	struct net_device *dev = NULL;
2384 	struct sk_buff *nskb;
2385 	int err;
2386 	u32 ext_filter_mask = 0;
2387 
2388 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2389 	if (err < 0)
2390 		return err;
2391 
2392 	if (tb[IFLA_IFNAME])
2393 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2394 
2395 	if (tb[IFLA_EXT_MASK])
2396 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2397 
2398 	ifm = nlmsg_data(nlh);
2399 	if (ifm->ifi_index > 0)
2400 		dev = __dev_get_by_index(net, ifm->ifi_index);
2401 	else if (tb[IFLA_IFNAME])
2402 		dev = __dev_get_by_name(net, ifname);
2403 	else
2404 		return -EINVAL;
2405 
2406 	if (dev == NULL)
2407 		return -ENODEV;
2408 
2409 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2410 	if (nskb == NULL)
2411 		return -ENOBUFS;
2412 
2413 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2414 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2415 	if (err < 0) {
2416 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2417 		WARN_ON(err == -EMSGSIZE);
2418 		kfree_skb(nskb);
2419 	} else
2420 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2421 
2422 	return err;
2423 }
2424 
2425 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2426 {
2427 	struct net *net = sock_net(skb->sk);
2428 	struct net_device *dev;
2429 	struct nlattr *tb[IFLA_MAX+1];
2430 	u32 ext_filter_mask = 0;
2431 	u16 min_ifinfo_dump_size = 0;
2432 	int hdrlen;
2433 
2434 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2435 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2436 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2437 
2438 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2439 		if (tb[IFLA_EXT_MASK])
2440 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2441 	}
2442 
2443 	if (!ext_filter_mask)
2444 		return NLMSG_GOODSIZE;
2445 	/*
2446 	 * traverse the list of net devices and compute the minimum
2447 	 * buffer size based upon the filter mask.
2448 	 */
2449 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2450 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2451 					     if_nlmsg_size(dev,
2452 						           ext_filter_mask));
2453 	}
2454 
2455 	return min_ifinfo_dump_size;
2456 }
2457 
2458 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2459 {
2460 	int idx;
2461 	int s_idx = cb->family;
2462 
2463 	if (s_idx == 0)
2464 		s_idx = 1;
2465 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2466 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2467 		if (idx < s_idx || idx == PF_PACKET)
2468 			continue;
2469 		if (rtnl_msg_handlers[idx] == NULL ||
2470 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2471 			continue;
2472 		if (idx > s_idx) {
2473 			memset(&cb->args[0], 0, sizeof(cb->args));
2474 			cb->prev_seq = 0;
2475 			cb->seq = 0;
2476 		}
2477 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2478 			break;
2479 	}
2480 	cb->family = idx;
2481 
2482 	return skb->len;
2483 }
2484 
2485 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2486 				       unsigned int change, gfp_t flags)
2487 {
2488 	struct net *net = dev_net(dev);
2489 	struct sk_buff *skb;
2490 	int err = -ENOBUFS;
2491 	size_t if_info_size;
2492 
2493 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2494 	if (skb == NULL)
2495 		goto errout;
2496 
2497 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2498 	if (err < 0) {
2499 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2500 		WARN_ON(err == -EMSGSIZE);
2501 		kfree_skb(skb);
2502 		goto errout;
2503 	}
2504 	return skb;
2505 errout:
2506 	if (err < 0)
2507 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2508 	return NULL;
2509 }
2510 
2511 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2512 {
2513 	struct net *net = dev_net(dev);
2514 
2515 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2516 }
2517 
2518 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2519 		  gfp_t flags)
2520 {
2521 	struct sk_buff *skb;
2522 
2523 	if (dev->reg_state != NETREG_REGISTERED)
2524 		return;
2525 
2526 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2527 	if (skb)
2528 		rtmsg_ifinfo_send(skb, dev, flags);
2529 }
2530 EXPORT_SYMBOL(rtmsg_ifinfo);
2531 
2532 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2533 				   struct net_device *dev,
2534 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2535 				   int type, unsigned int flags,
2536 				   int nlflags)
2537 {
2538 	struct nlmsghdr *nlh;
2539 	struct ndmsg *ndm;
2540 
2541 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2542 	if (!nlh)
2543 		return -EMSGSIZE;
2544 
2545 	ndm = nlmsg_data(nlh);
2546 	ndm->ndm_family  = AF_BRIDGE;
2547 	ndm->ndm_pad1	 = 0;
2548 	ndm->ndm_pad2    = 0;
2549 	ndm->ndm_flags	 = flags;
2550 	ndm->ndm_type	 = 0;
2551 	ndm->ndm_ifindex = dev->ifindex;
2552 	ndm->ndm_state   = NUD_PERMANENT;
2553 
2554 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2555 		goto nla_put_failure;
2556 	if (vid)
2557 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2558 			goto nla_put_failure;
2559 
2560 	nlmsg_end(skb, nlh);
2561 	return 0;
2562 
2563 nla_put_failure:
2564 	nlmsg_cancel(skb, nlh);
2565 	return -EMSGSIZE;
2566 }
2567 
2568 static inline size_t rtnl_fdb_nlmsg_size(void)
2569 {
2570 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2571 }
2572 
2573 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2574 {
2575 	struct net *net = dev_net(dev);
2576 	struct sk_buff *skb;
2577 	int err = -ENOBUFS;
2578 
2579 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2580 	if (!skb)
2581 		goto errout;
2582 
2583 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2584 				      0, 0, type, NTF_SELF, 0);
2585 	if (err < 0) {
2586 		kfree_skb(skb);
2587 		goto errout;
2588 	}
2589 
2590 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2591 	return;
2592 errout:
2593 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2594 }
2595 
2596 /**
2597  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2598  */
2599 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2600 		     struct nlattr *tb[],
2601 		     struct net_device *dev,
2602 		     const unsigned char *addr, u16 vid,
2603 		     u16 flags)
2604 {
2605 	int err = -EINVAL;
2606 
2607 	/* If aging addresses are supported device will need to
2608 	 * implement its own handler for this.
2609 	 */
2610 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2611 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2612 		return err;
2613 	}
2614 
2615 	if (vid) {
2616 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2617 		return err;
2618 	}
2619 
2620 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2621 		err = dev_uc_add_excl(dev, addr);
2622 	else if (is_multicast_ether_addr(addr))
2623 		err = dev_mc_add_excl(dev, addr);
2624 
2625 	/* Only return duplicate errors if NLM_F_EXCL is set */
2626 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2627 		err = 0;
2628 
2629 	return err;
2630 }
2631 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2632 
2633 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2634 {
2635 	u16 vid = 0;
2636 
2637 	if (vlan_attr) {
2638 		if (nla_len(vlan_attr) != sizeof(u16)) {
2639 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2640 			return -EINVAL;
2641 		}
2642 
2643 		vid = nla_get_u16(vlan_attr);
2644 
2645 		if (!vid || vid >= VLAN_VID_MASK) {
2646 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2647 				vid);
2648 			return -EINVAL;
2649 		}
2650 	}
2651 	*p_vid = vid;
2652 	return 0;
2653 }
2654 
2655 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2656 {
2657 	struct net *net = sock_net(skb->sk);
2658 	struct ndmsg *ndm;
2659 	struct nlattr *tb[NDA_MAX+1];
2660 	struct net_device *dev;
2661 	u8 *addr;
2662 	u16 vid;
2663 	int err;
2664 
2665 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2666 	if (err < 0)
2667 		return err;
2668 
2669 	ndm = nlmsg_data(nlh);
2670 	if (ndm->ndm_ifindex == 0) {
2671 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2672 		return -EINVAL;
2673 	}
2674 
2675 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2676 	if (dev == NULL) {
2677 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2678 		return -ENODEV;
2679 	}
2680 
2681 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2682 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2683 		return -EINVAL;
2684 	}
2685 
2686 	addr = nla_data(tb[NDA_LLADDR]);
2687 
2688 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2689 	if (err)
2690 		return err;
2691 
2692 	err = -EOPNOTSUPP;
2693 
2694 	/* Support fdb on master device the net/bridge default case */
2695 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2696 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2697 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2698 		const struct net_device_ops *ops = br_dev->netdev_ops;
2699 
2700 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2701 				       nlh->nlmsg_flags);
2702 		if (err)
2703 			goto out;
2704 		else
2705 			ndm->ndm_flags &= ~NTF_MASTER;
2706 	}
2707 
2708 	/* Embedded bridge, macvlan, and any other device support */
2709 	if ((ndm->ndm_flags & NTF_SELF)) {
2710 		if (dev->netdev_ops->ndo_fdb_add)
2711 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2712 							   vid,
2713 							   nlh->nlmsg_flags);
2714 		else
2715 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2716 					       nlh->nlmsg_flags);
2717 
2718 		if (!err) {
2719 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2720 			ndm->ndm_flags &= ~NTF_SELF;
2721 		}
2722 	}
2723 out:
2724 	return err;
2725 }
2726 
2727 /**
2728  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2729  */
2730 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2731 		     struct nlattr *tb[],
2732 		     struct net_device *dev,
2733 		     const unsigned char *addr, u16 vid)
2734 {
2735 	int err = -EINVAL;
2736 
2737 	/* If aging addresses are supported device will need to
2738 	 * implement its own handler for this.
2739 	 */
2740 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2741 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2742 		return err;
2743 	}
2744 
2745 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2746 		err = dev_uc_del(dev, addr);
2747 	else if (is_multicast_ether_addr(addr))
2748 		err = dev_mc_del(dev, addr);
2749 
2750 	return err;
2751 }
2752 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2753 
2754 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2755 {
2756 	struct net *net = sock_net(skb->sk);
2757 	struct ndmsg *ndm;
2758 	struct nlattr *tb[NDA_MAX+1];
2759 	struct net_device *dev;
2760 	int err = -EINVAL;
2761 	__u8 *addr;
2762 	u16 vid;
2763 
2764 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2765 		return -EPERM;
2766 
2767 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2768 	if (err < 0)
2769 		return err;
2770 
2771 	ndm = nlmsg_data(nlh);
2772 	if (ndm->ndm_ifindex == 0) {
2773 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2774 		return -EINVAL;
2775 	}
2776 
2777 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2778 	if (dev == NULL) {
2779 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2780 		return -ENODEV;
2781 	}
2782 
2783 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2784 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2785 		return -EINVAL;
2786 	}
2787 
2788 	addr = nla_data(tb[NDA_LLADDR]);
2789 
2790 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2791 	if (err)
2792 		return err;
2793 
2794 	err = -EOPNOTSUPP;
2795 
2796 	/* Support fdb on master device the net/bridge default case */
2797 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2798 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2799 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2800 		const struct net_device_ops *ops = br_dev->netdev_ops;
2801 
2802 		if (ops->ndo_fdb_del)
2803 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2804 
2805 		if (err)
2806 			goto out;
2807 		else
2808 			ndm->ndm_flags &= ~NTF_MASTER;
2809 	}
2810 
2811 	/* Embedded bridge, macvlan, and any other device support */
2812 	if (ndm->ndm_flags & NTF_SELF) {
2813 		if (dev->netdev_ops->ndo_fdb_del)
2814 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2815 							   vid);
2816 		else
2817 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2818 
2819 		if (!err) {
2820 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2821 			ndm->ndm_flags &= ~NTF_SELF;
2822 		}
2823 	}
2824 out:
2825 	return err;
2826 }
2827 
2828 static int nlmsg_populate_fdb(struct sk_buff *skb,
2829 			      struct netlink_callback *cb,
2830 			      struct net_device *dev,
2831 			      int *idx,
2832 			      struct netdev_hw_addr_list *list)
2833 {
2834 	struct netdev_hw_addr *ha;
2835 	int err;
2836 	u32 portid, seq;
2837 
2838 	portid = NETLINK_CB(cb->skb).portid;
2839 	seq = cb->nlh->nlmsg_seq;
2840 
2841 	list_for_each_entry(ha, &list->list, list) {
2842 		if (*idx < cb->args[0])
2843 			goto skip;
2844 
2845 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2846 					      portid, seq,
2847 					      RTM_NEWNEIGH, NTF_SELF,
2848 					      NLM_F_MULTI);
2849 		if (err < 0)
2850 			return err;
2851 skip:
2852 		*idx += 1;
2853 	}
2854 	return 0;
2855 }
2856 
2857 /**
2858  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2859  * @nlh: netlink message header
2860  * @dev: netdevice
2861  *
2862  * Default netdevice operation to dump the existing unicast address list.
2863  * Returns number of addresses from list put in skb.
2864  */
2865 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2866 		      struct netlink_callback *cb,
2867 		      struct net_device *dev,
2868 		      struct net_device *filter_dev,
2869 		      int idx)
2870 {
2871 	int err;
2872 
2873 	netif_addr_lock_bh(dev);
2874 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2875 	if (err)
2876 		goto out;
2877 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2878 out:
2879 	netif_addr_unlock_bh(dev);
2880 	return idx;
2881 }
2882 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2883 
2884 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2885 {
2886 	struct net_device *dev;
2887 	struct nlattr *tb[IFLA_MAX+1];
2888 	struct net_device *br_dev = NULL;
2889 	const struct net_device_ops *ops = NULL;
2890 	const struct net_device_ops *cops = NULL;
2891 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2892 	struct net *net = sock_net(skb->sk);
2893 	int brport_idx = 0;
2894 	int br_idx = 0;
2895 	int idx = 0;
2896 
2897 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2898 			ifla_policy) == 0) {
2899 		if (tb[IFLA_MASTER])
2900 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
2901 	}
2902 
2903 	brport_idx = ifm->ifi_index;
2904 
2905 	if (br_idx) {
2906 		br_dev = __dev_get_by_index(net, br_idx);
2907 		if (!br_dev)
2908 			return -ENODEV;
2909 
2910 		ops = br_dev->netdev_ops;
2911 	}
2912 
2913 	for_each_netdev(net, dev) {
2914 		if (brport_idx && (dev->ifindex != brport_idx))
2915 			continue;
2916 
2917 		if (!br_idx) { /* user did not specify a specific bridge */
2918 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
2919 				br_dev = netdev_master_upper_dev_get(dev);
2920 				cops = br_dev->netdev_ops;
2921 			}
2922 
2923 		} else {
2924 			if (dev != br_dev &&
2925 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
2926 				continue;
2927 
2928 			if (br_dev != netdev_master_upper_dev_get(dev) &&
2929 			    !(dev->priv_flags & IFF_EBRIDGE))
2930 				continue;
2931 
2932 			cops = ops;
2933 		}
2934 
2935 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2936 			if (cops && cops->ndo_fdb_dump)
2937 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2938 							 idx);
2939 		}
2940 
2941 		if (dev->netdev_ops->ndo_fdb_dump)
2942 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2943 							    idx);
2944 		else
2945 			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2946 
2947 		cops = NULL;
2948 	}
2949 
2950 	cb->args[0] = idx;
2951 	return skb->len;
2952 }
2953 
2954 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2955 			       unsigned int attrnum, unsigned int flag)
2956 {
2957 	if (mask & flag)
2958 		return nla_put_u8(skb, attrnum, !!(flags & flag));
2959 	return 0;
2960 }
2961 
2962 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2963 			    struct net_device *dev, u16 mode,
2964 			    u32 flags, u32 mask, int nlflags,
2965 			    u32 filter_mask,
2966 			    int (*vlan_fill)(struct sk_buff *skb,
2967 					     struct net_device *dev,
2968 					     u32 filter_mask))
2969 {
2970 	struct nlmsghdr *nlh;
2971 	struct ifinfomsg *ifm;
2972 	struct nlattr *br_afspec;
2973 	struct nlattr *protinfo;
2974 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2975 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2976 	int err = 0;
2977 
2978 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2979 	if (nlh == NULL)
2980 		return -EMSGSIZE;
2981 
2982 	ifm = nlmsg_data(nlh);
2983 	ifm->ifi_family = AF_BRIDGE;
2984 	ifm->__ifi_pad = 0;
2985 	ifm->ifi_type = dev->type;
2986 	ifm->ifi_index = dev->ifindex;
2987 	ifm->ifi_flags = dev_get_flags(dev);
2988 	ifm->ifi_change = 0;
2989 
2990 
2991 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2992 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2993 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2994 	    (br_dev &&
2995 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2996 	    (dev->addr_len &&
2997 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2998 	    (dev->ifindex != dev_get_iflink(dev) &&
2999 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3000 		goto nla_put_failure;
3001 
3002 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3003 	if (!br_afspec)
3004 		goto nla_put_failure;
3005 
3006 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3007 		nla_nest_cancel(skb, br_afspec);
3008 		goto nla_put_failure;
3009 	}
3010 
3011 	if (mode != BRIDGE_MODE_UNDEF) {
3012 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3013 			nla_nest_cancel(skb, br_afspec);
3014 			goto nla_put_failure;
3015 		}
3016 	}
3017 	if (vlan_fill) {
3018 		err = vlan_fill(skb, dev, filter_mask);
3019 		if (err) {
3020 			nla_nest_cancel(skb, br_afspec);
3021 			goto nla_put_failure;
3022 		}
3023 	}
3024 	nla_nest_end(skb, br_afspec);
3025 
3026 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3027 	if (!protinfo)
3028 		goto nla_put_failure;
3029 
3030 	if (brport_nla_put_flag(skb, flags, mask,
3031 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3032 	    brport_nla_put_flag(skb, flags, mask,
3033 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3034 	    brport_nla_put_flag(skb, flags, mask,
3035 				IFLA_BRPORT_FAST_LEAVE,
3036 				BR_MULTICAST_FAST_LEAVE) ||
3037 	    brport_nla_put_flag(skb, flags, mask,
3038 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3039 	    brport_nla_put_flag(skb, flags, mask,
3040 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3041 	    brport_nla_put_flag(skb, flags, mask,
3042 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3043 	    brport_nla_put_flag(skb, flags, mask,
3044 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3045 	    brport_nla_put_flag(skb, flags, mask,
3046 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3047 		nla_nest_cancel(skb, protinfo);
3048 		goto nla_put_failure;
3049 	}
3050 
3051 	nla_nest_end(skb, protinfo);
3052 
3053 	nlmsg_end(skb, nlh);
3054 	return 0;
3055 nla_put_failure:
3056 	nlmsg_cancel(skb, nlh);
3057 	return err ? err : -EMSGSIZE;
3058 }
3059 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3060 
3061 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3062 {
3063 	struct net *net = sock_net(skb->sk);
3064 	struct net_device *dev;
3065 	int idx = 0;
3066 	u32 portid = NETLINK_CB(cb->skb).portid;
3067 	u32 seq = cb->nlh->nlmsg_seq;
3068 	u32 filter_mask = 0;
3069 	int err;
3070 
3071 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3072 		struct nlattr *extfilt;
3073 
3074 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3075 					  IFLA_EXT_MASK);
3076 		if (extfilt) {
3077 			if (nla_len(extfilt) < sizeof(filter_mask))
3078 				return -EINVAL;
3079 
3080 			filter_mask = nla_get_u32(extfilt);
3081 		}
3082 	}
3083 
3084 	rcu_read_lock();
3085 	for_each_netdev_rcu(net, dev) {
3086 		const struct net_device_ops *ops = dev->netdev_ops;
3087 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3088 
3089 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3090 			if (idx >= cb->args[0]) {
3091 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3092 						skb, portid, seq, dev,
3093 						filter_mask, NLM_F_MULTI);
3094 				if (err < 0 && err != -EOPNOTSUPP)
3095 					break;
3096 			}
3097 			idx++;
3098 		}
3099 
3100 		if (ops->ndo_bridge_getlink) {
3101 			if (idx >= cb->args[0]) {
3102 				err = ops->ndo_bridge_getlink(skb, portid,
3103 							      seq, dev,
3104 							      filter_mask,
3105 							      NLM_F_MULTI);
3106 				if (err < 0 && err != -EOPNOTSUPP)
3107 					break;
3108 			}
3109 			idx++;
3110 		}
3111 	}
3112 	rcu_read_unlock();
3113 	cb->args[0] = idx;
3114 
3115 	return skb->len;
3116 }
3117 
3118 static inline size_t bridge_nlmsg_size(void)
3119 {
3120 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3121 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3122 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3123 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3124 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3125 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3126 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3127 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3128 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3129 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3130 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3131 }
3132 
3133 static int rtnl_bridge_notify(struct net_device *dev)
3134 {
3135 	struct net *net = dev_net(dev);
3136 	struct sk_buff *skb;
3137 	int err = -EOPNOTSUPP;
3138 
3139 	if (!dev->netdev_ops->ndo_bridge_getlink)
3140 		return 0;
3141 
3142 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3143 	if (!skb) {
3144 		err = -ENOMEM;
3145 		goto errout;
3146 	}
3147 
3148 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3149 	if (err < 0)
3150 		goto errout;
3151 
3152 	if (!skb->len)
3153 		goto errout;
3154 
3155 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3156 	return 0;
3157 errout:
3158 	WARN_ON(err == -EMSGSIZE);
3159 	kfree_skb(skb);
3160 	if (err)
3161 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3162 	return err;
3163 }
3164 
3165 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3166 {
3167 	struct net *net = sock_net(skb->sk);
3168 	struct ifinfomsg *ifm;
3169 	struct net_device *dev;
3170 	struct nlattr *br_spec, *attr = NULL;
3171 	int rem, err = -EOPNOTSUPP;
3172 	u16 flags = 0;
3173 	bool have_flags = false;
3174 
3175 	if (nlmsg_len(nlh) < sizeof(*ifm))
3176 		return -EINVAL;
3177 
3178 	ifm = nlmsg_data(nlh);
3179 	if (ifm->ifi_family != AF_BRIDGE)
3180 		return -EPFNOSUPPORT;
3181 
3182 	dev = __dev_get_by_index(net, ifm->ifi_index);
3183 	if (!dev) {
3184 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3185 		return -ENODEV;
3186 	}
3187 
3188 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3189 	if (br_spec) {
3190 		nla_for_each_nested(attr, br_spec, rem) {
3191 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3192 				if (nla_len(attr) < sizeof(flags))
3193 					return -EINVAL;
3194 
3195 				have_flags = true;
3196 				flags = nla_get_u16(attr);
3197 				break;
3198 			}
3199 		}
3200 	}
3201 
3202 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3203 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3204 
3205 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3206 			err = -EOPNOTSUPP;
3207 			goto out;
3208 		}
3209 
3210 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3211 		if (err)
3212 			goto out;
3213 
3214 		flags &= ~BRIDGE_FLAGS_MASTER;
3215 	}
3216 
3217 	if ((flags & BRIDGE_FLAGS_SELF)) {
3218 		if (!dev->netdev_ops->ndo_bridge_setlink)
3219 			err = -EOPNOTSUPP;
3220 		else
3221 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3222 								  flags);
3223 		if (!err) {
3224 			flags &= ~BRIDGE_FLAGS_SELF;
3225 
3226 			/* Generate event to notify upper layer of bridge
3227 			 * change
3228 			 */
3229 			err = rtnl_bridge_notify(dev);
3230 		}
3231 	}
3232 
3233 	if (have_flags)
3234 		memcpy(nla_data(attr), &flags, sizeof(flags));
3235 out:
3236 	return err;
3237 }
3238 
3239 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3240 {
3241 	struct net *net = sock_net(skb->sk);
3242 	struct ifinfomsg *ifm;
3243 	struct net_device *dev;
3244 	struct nlattr *br_spec, *attr = NULL;
3245 	int rem, err = -EOPNOTSUPP;
3246 	u16 flags = 0;
3247 	bool have_flags = false;
3248 
3249 	if (nlmsg_len(nlh) < sizeof(*ifm))
3250 		return -EINVAL;
3251 
3252 	ifm = nlmsg_data(nlh);
3253 	if (ifm->ifi_family != AF_BRIDGE)
3254 		return -EPFNOSUPPORT;
3255 
3256 	dev = __dev_get_by_index(net, ifm->ifi_index);
3257 	if (!dev) {
3258 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3259 		return -ENODEV;
3260 	}
3261 
3262 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3263 	if (br_spec) {
3264 		nla_for_each_nested(attr, br_spec, rem) {
3265 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3266 				if (nla_len(attr) < sizeof(flags))
3267 					return -EINVAL;
3268 
3269 				have_flags = true;
3270 				flags = nla_get_u16(attr);
3271 				break;
3272 			}
3273 		}
3274 	}
3275 
3276 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3277 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3278 
3279 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3280 			err = -EOPNOTSUPP;
3281 			goto out;
3282 		}
3283 
3284 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3285 		if (err)
3286 			goto out;
3287 
3288 		flags &= ~BRIDGE_FLAGS_MASTER;
3289 	}
3290 
3291 	if ((flags & BRIDGE_FLAGS_SELF)) {
3292 		if (!dev->netdev_ops->ndo_bridge_dellink)
3293 			err = -EOPNOTSUPP;
3294 		else
3295 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3296 								  flags);
3297 
3298 		if (!err) {
3299 			flags &= ~BRIDGE_FLAGS_SELF;
3300 
3301 			/* Generate event to notify upper layer of bridge
3302 			 * change
3303 			 */
3304 			err = rtnl_bridge_notify(dev);
3305 		}
3306 	}
3307 
3308 	if (have_flags)
3309 		memcpy(nla_data(attr), &flags, sizeof(flags));
3310 out:
3311 	return err;
3312 }
3313 
3314 /* Process one rtnetlink message. */
3315 
3316 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3317 {
3318 	struct net *net = sock_net(skb->sk);
3319 	rtnl_doit_func doit;
3320 	int sz_idx, kind;
3321 	int family;
3322 	int type;
3323 	int err;
3324 
3325 	type = nlh->nlmsg_type;
3326 	if (type > RTM_MAX)
3327 		return -EOPNOTSUPP;
3328 
3329 	type -= RTM_BASE;
3330 
3331 	/* All the messages must have at least 1 byte length */
3332 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3333 		return 0;
3334 
3335 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3336 	sz_idx = type>>2;
3337 	kind = type&3;
3338 
3339 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3340 		return -EPERM;
3341 
3342 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3343 		struct sock *rtnl;
3344 		rtnl_dumpit_func dumpit;
3345 		rtnl_calcit_func calcit;
3346 		u16 min_dump_alloc = 0;
3347 
3348 		dumpit = rtnl_get_dumpit(family, type);
3349 		if (dumpit == NULL)
3350 			return -EOPNOTSUPP;
3351 		calcit = rtnl_get_calcit(family, type);
3352 		if (calcit)
3353 			min_dump_alloc = calcit(skb, nlh);
3354 
3355 		__rtnl_unlock();
3356 		rtnl = net->rtnl;
3357 		{
3358 			struct netlink_dump_control c = {
3359 				.dump		= dumpit,
3360 				.min_dump_alloc	= min_dump_alloc,
3361 			};
3362 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3363 		}
3364 		rtnl_lock();
3365 		return err;
3366 	}
3367 
3368 	doit = rtnl_get_doit(family, type);
3369 	if (doit == NULL)
3370 		return -EOPNOTSUPP;
3371 
3372 	return doit(skb, nlh);
3373 }
3374 
3375 static void rtnetlink_rcv(struct sk_buff *skb)
3376 {
3377 	rtnl_lock();
3378 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3379 	rtnl_unlock();
3380 }
3381 
3382 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3383 {
3384 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3385 
3386 	switch (event) {
3387 	case NETDEV_UP:
3388 	case NETDEV_DOWN:
3389 	case NETDEV_PRE_UP:
3390 	case NETDEV_POST_INIT:
3391 	case NETDEV_REGISTER:
3392 	case NETDEV_CHANGE:
3393 	case NETDEV_PRE_TYPE_CHANGE:
3394 	case NETDEV_GOING_DOWN:
3395 	case NETDEV_UNREGISTER:
3396 	case NETDEV_UNREGISTER_FINAL:
3397 	case NETDEV_RELEASE:
3398 	case NETDEV_JOIN:
3399 	case NETDEV_BONDING_INFO:
3400 		break;
3401 	default:
3402 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3403 		break;
3404 	}
3405 	return NOTIFY_DONE;
3406 }
3407 
3408 static struct notifier_block rtnetlink_dev_notifier = {
3409 	.notifier_call	= rtnetlink_event,
3410 };
3411 
3412 
3413 static int __net_init rtnetlink_net_init(struct net *net)
3414 {
3415 	struct sock *sk;
3416 	struct netlink_kernel_cfg cfg = {
3417 		.groups		= RTNLGRP_MAX,
3418 		.input		= rtnetlink_rcv,
3419 		.cb_mutex	= &rtnl_mutex,
3420 		.flags		= NL_CFG_F_NONROOT_RECV,
3421 	};
3422 
3423 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3424 	if (!sk)
3425 		return -ENOMEM;
3426 	net->rtnl = sk;
3427 	return 0;
3428 }
3429 
3430 static void __net_exit rtnetlink_net_exit(struct net *net)
3431 {
3432 	netlink_kernel_release(net->rtnl);
3433 	net->rtnl = NULL;
3434 }
3435 
3436 static struct pernet_operations rtnetlink_net_ops = {
3437 	.init = rtnetlink_net_init,
3438 	.exit = rtnetlink_net_exit,
3439 };
3440 
3441 void __init rtnetlink_init(void)
3442 {
3443 	if (register_pernet_subsys(&rtnetlink_net_ops))
3444 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3445 
3446 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3447 
3448 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3449 		      rtnl_dump_ifinfo, rtnl_calcit);
3450 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3451 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3452 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3453 
3454 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3455 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3456 
3457 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3458 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3459 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3460 
3461 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3462 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3463 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3464 }
3465