xref: /linux/net/core/rtnetlink.c (revision 791d3ef2e11100449837dc0b6fe884e60ca3a484)
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/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44 
45 #include <linux/uaccess.h>
46 
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
50 #include <net/ip.h>
51 #include <net/protocol.h>
52 #include <net/arp.h>
53 #include <net/route.h>
54 #include <net/udp.h>
55 #include <net/tcp.h>
56 #include <net/sock.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
61 
62 #define RTNL_MAX_TYPE		48
63 #define RTNL_SLAVE_MAX_TYPE	36
64 
65 struct rtnl_link {
66 	rtnl_doit_func		doit;
67 	rtnl_dumpit_func	dumpit;
68 	struct module		*owner;
69 	unsigned int		flags;
70 	struct rcu_head		rcu;
71 };
72 
73 static DEFINE_MUTEX(rtnl_mutex);
74 
75 void rtnl_lock(void)
76 {
77 	mutex_lock(&rtnl_mutex);
78 }
79 EXPORT_SYMBOL(rtnl_lock);
80 
81 int rtnl_lock_killable(void)
82 {
83 	return mutex_lock_killable(&rtnl_mutex);
84 }
85 EXPORT_SYMBOL(rtnl_lock_killable);
86 
87 static struct sk_buff *defer_kfree_skb_list;
88 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
89 {
90 	if (head && tail) {
91 		tail->next = defer_kfree_skb_list;
92 		defer_kfree_skb_list = head;
93 	}
94 }
95 EXPORT_SYMBOL(rtnl_kfree_skbs);
96 
97 void __rtnl_unlock(void)
98 {
99 	struct sk_buff *head = defer_kfree_skb_list;
100 
101 	defer_kfree_skb_list = NULL;
102 
103 	mutex_unlock(&rtnl_mutex);
104 
105 	while (head) {
106 		struct sk_buff *next = head->next;
107 
108 		kfree_skb(head);
109 		cond_resched();
110 		head = next;
111 	}
112 }
113 
114 void rtnl_unlock(void)
115 {
116 	/* This fellow will unlock it for us. */
117 	netdev_run_todo();
118 }
119 EXPORT_SYMBOL(rtnl_unlock);
120 
121 int rtnl_trylock(void)
122 {
123 	return mutex_trylock(&rtnl_mutex);
124 }
125 EXPORT_SYMBOL(rtnl_trylock);
126 
127 int rtnl_is_locked(void)
128 {
129 	return mutex_is_locked(&rtnl_mutex);
130 }
131 EXPORT_SYMBOL(rtnl_is_locked);
132 
133 #ifdef CONFIG_PROVE_LOCKING
134 bool lockdep_rtnl_is_held(void)
135 {
136 	return lockdep_is_held(&rtnl_mutex);
137 }
138 EXPORT_SYMBOL(lockdep_rtnl_is_held);
139 #endif /* #ifdef CONFIG_PROVE_LOCKING */
140 
141 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
142 
143 static inline int rtm_msgindex(int msgtype)
144 {
145 	int msgindex = msgtype - RTM_BASE;
146 
147 	/*
148 	 * msgindex < 0 implies someone tried to register a netlink
149 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
150 	 * the message type has not been added to linux/rtnetlink.h
151 	 */
152 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
153 
154 	return msgindex;
155 }
156 
157 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
158 {
159 	struct rtnl_link **tab;
160 
161 	if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
162 		protocol = PF_UNSPEC;
163 
164 	tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
165 	if (!tab)
166 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
167 
168 	return tab[msgtype];
169 }
170 
171 static int rtnl_register_internal(struct module *owner,
172 				  int protocol, int msgtype,
173 				  rtnl_doit_func doit, rtnl_dumpit_func dumpit,
174 				  unsigned int flags)
175 {
176 	struct rtnl_link *link, *old;
177 	struct rtnl_link __rcu **tab;
178 	int msgindex;
179 	int ret = -ENOBUFS;
180 
181 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
182 	msgindex = rtm_msgindex(msgtype);
183 
184 	rtnl_lock();
185 	tab = rtnl_msg_handlers[protocol];
186 	if (tab == NULL) {
187 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
188 		if (!tab)
189 			goto unlock;
190 
191 		/* ensures we see the 0 stores */
192 		rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
193 	}
194 
195 	old = rtnl_dereference(tab[msgindex]);
196 	if (old) {
197 		link = kmemdup(old, sizeof(*old), GFP_KERNEL);
198 		if (!link)
199 			goto unlock;
200 	} else {
201 		link = kzalloc(sizeof(*link), GFP_KERNEL);
202 		if (!link)
203 			goto unlock;
204 	}
205 
206 	WARN_ON(link->owner && link->owner != owner);
207 	link->owner = owner;
208 
209 	WARN_ON(doit && link->doit && link->doit != doit);
210 	if (doit)
211 		link->doit = doit;
212 	WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
213 	if (dumpit)
214 		link->dumpit = dumpit;
215 
216 	link->flags |= flags;
217 
218 	/* publish protocol:msgtype */
219 	rcu_assign_pointer(tab[msgindex], link);
220 	ret = 0;
221 	if (old)
222 		kfree_rcu(old, rcu);
223 unlock:
224 	rtnl_unlock();
225 	return ret;
226 }
227 
228 /**
229  * rtnl_register_module - Register a rtnetlink message type
230  *
231  * @owner: module registering the hook (THIS_MODULE)
232  * @protocol: Protocol family or PF_UNSPEC
233  * @msgtype: rtnetlink message type
234  * @doit: Function pointer called for each request message
235  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
236  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
237  *
238  * Like rtnl_register, but for use by removable modules.
239  */
240 int rtnl_register_module(struct module *owner,
241 			 int protocol, int msgtype,
242 			 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
243 			 unsigned int flags)
244 {
245 	return rtnl_register_internal(owner, protocol, msgtype,
246 				      doit, dumpit, flags);
247 }
248 EXPORT_SYMBOL_GPL(rtnl_register_module);
249 
250 /**
251  * rtnl_register - Register a rtnetlink message type
252  * @protocol: Protocol family or PF_UNSPEC
253  * @msgtype: rtnetlink message type
254  * @doit: Function pointer called for each request message
255  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
256  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
257  *
258  * Registers the specified function pointers (at least one of them has
259  * to be non-NULL) to be called whenever a request message for the
260  * specified protocol family and message type is received.
261  *
262  * The special protocol family PF_UNSPEC may be used to define fallback
263  * function pointers for the case when no entry for the specific protocol
264  * family exists.
265  */
266 void rtnl_register(int protocol, int msgtype,
267 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
268 		   unsigned int flags)
269 {
270 	int err;
271 
272 	err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
273 				     flags);
274 	if (err)
275 		pr_err("Unable to register rtnetlink message handler, "
276 		       "protocol = %d, message type = %d\n", protocol, msgtype);
277 }
278 
279 /**
280  * rtnl_unregister - Unregister a rtnetlink message type
281  * @protocol: Protocol family or PF_UNSPEC
282  * @msgtype: rtnetlink message type
283  *
284  * Returns 0 on success or a negative error code.
285  */
286 int rtnl_unregister(int protocol, int msgtype)
287 {
288 	struct rtnl_link **tab, *link;
289 	int msgindex;
290 
291 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
292 	msgindex = rtm_msgindex(msgtype);
293 
294 	rtnl_lock();
295 	tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
296 	if (!tab) {
297 		rtnl_unlock();
298 		return -ENOENT;
299 	}
300 
301 	link = tab[msgindex];
302 	rcu_assign_pointer(tab[msgindex], NULL);
303 	rtnl_unlock();
304 
305 	kfree_rcu(link, rcu);
306 
307 	return 0;
308 }
309 EXPORT_SYMBOL_GPL(rtnl_unregister);
310 
311 /**
312  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
313  * @protocol : Protocol family or PF_UNSPEC
314  *
315  * Identical to calling rtnl_unregster() for all registered message types
316  * of a certain protocol family.
317  */
318 void rtnl_unregister_all(int protocol)
319 {
320 	struct rtnl_link **tab, *link;
321 	int msgindex;
322 
323 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
324 
325 	rtnl_lock();
326 	tab = rtnl_msg_handlers[protocol];
327 	RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
328 	for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
329 		link = tab[msgindex];
330 		if (!link)
331 			continue;
332 
333 		rcu_assign_pointer(tab[msgindex], NULL);
334 		kfree_rcu(link, rcu);
335 	}
336 	rtnl_unlock();
337 
338 	synchronize_net();
339 
340 	kfree(tab);
341 }
342 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
343 
344 static LIST_HEAD(link_ops);
345 
346 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
347 {
348 	const struct rtnl_link_ops *ops;
349 
350 	list_for_each_entry(ops, &link_ops, list) {
351 		if (!strcmp(ops->kind, kind))
352 			return ops;
353 	}
354 	return NULL;
355 }
356 
357 /**
358  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
359  * @ops: struct rtnl_link_ops * to register
360  *
361  * The caller must hold the rtnl_mutex. This function should be used
362  * by drivers that create devices during module initialization. It
363  * must be called before registering the devices.
364  *
365  * Returns 0 on success or a negative error code.
366  */
367 int __rtnl_link_register(struct rtnl_link_ops *ops)
368 {
369 	if (rtnl_link_ops_get(ops->kind))
370 		return -EEXIST;
371 
372 	/* The check for setup is here because if ops
373 	 * does not have that filled up, it is not possible
374 	 * to use the ops for creating device. So do not
375 	 * fill up dellink as well. That disables rtnl_dellink.
376 	 */
377 	if (ops->setup && !ops->dellink)
378 		ops->dellink = unregister_netdevice_queue;
379 
380 	list_add_tail(&ops->list, &link_ops);
381 	return 0;
382 }
383 EXPORT_SYMBOL_GPL(__rtnl_link_register);
384 
385 /**
386  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
387  * @ops: struct rtnl_link_ops * to register
388  *
389  * Returns 0 on success or a negative error code.
390  */
391 int rtnl_link_register(struct rtnl_link_ops *ops)
392 {
393 	int err;
394 
395 	/* Sanity-check max sizes to avoid stack buffer overflow. */
396 	if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
397 		    ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
398 		return -EINVAL;
399 
400 	rtnl_lock();
401 	err = __rtnl_link_register(ops);
402 	rtnl_unlock();
403 	return err;
404 }
405 EXPORT_SYMBOL_GPL(rtnl_link_register);
406 
407 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
408 {
409 	struct net_device *dev;
410 	LIST_HEAD(list_kill);
411 
412 	for_each_netdev(net, dev) {
413 		if (dev->rtnl_link_ops == ops)
414 			ops->dellink(dev, &list_kill);
415 	}
416 	unregister_netdevice_many(&list_kill);
417 }
418 
419 /**
420  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
421  * @ops: struct rtnl_link_ops * to unregister
422  *
423  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
424  * integrity (hold pernet_ops_rwsem for writing to close the race
425  * with setup_net() and cleanup_net()).
426  */
427 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
428 {
429 	struct net *net;
430 
431 	for_each_net(net) {
432 		__rtnl_kill_links(net, ops);
433 	}
434 	list_del(&ops->list);
435 }
436 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
437 
438 /* Return with the rtnl_lock held when there are no network
439  * devices unregistering in any network namespace.
440  */
441 static void rtnl_lock_unregistering_all(void)
442 {
443 	struct net *net;
444 	bool unregistering;
445 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
446 
447 	add_wait_queue(&netdev_unregistering_wq, &wait);
448 	for (;;) {
449 		unregistering = false;
450 		rtnl_lock();
451 		/* We held write locked pernet_ops_rwsem, and parallel
452 		 * setup_net() and cleanup_net() are not possible.
453 		 */
454 		for_each_net(net) {
455 			if (net->dev_unreg_count > 0) {
456 				unregistering = true;
457 				break;
458 			}
459 		}
460 		if (!unregistering)
461 			break;
462 		__rtnl_unlock();
463 
464 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
465 	}
466 	remove_wait_queue(&netdev_unregistering_wq, &wait);
467 }
468 
469 /**
470  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
471  * @ops: struct rtnl_link_ops * to unregister
472  */
473 void rtnl_link_unregister(struct rtnl_link_ops *ops)
474 {
475 	/* Close the race with setup_net() and cleanup_net() */
476 	down_write(&pernet_ops_rwsem);
477 	rtnl_lock_unregistering_all();
478 	__rtnl_link_unregister(ops);
479 	rtnl_unlock();
480 	up_write(&pernet_ops_rwsem);
481 }
482 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
483 
484 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
485 {
486 	struct net_device *master_dev;
487 	const struct rtnl_link_ops *ops;
488 	size_t size = 0;
489 
490 	rcu_read_lock();
491 
492 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
493 	if (!master_dev)
494 		goto out;
495 
496 	ops = master_dev->rtnl_link_ops;
497 	if (!ops || !ops->get_slave_size)
498 		goto out;
499 	/* IFLA_INFO_SLAVE_DATA + nested data */
500 	size = nla_total_size(sizeof(struct nlattr)) +
501 	       ops->get_slave_size(master_dev, dev);
502 
503 out:
504 	rcu_read_unlock();
505 	return size;
506 }
507 
508 static size_t rtnl_link_get_size(const struct net_device *dev)
509 {
510 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
511 	size_t size;
512 
513 	if (!ops)
514 		return 0;
515 
516 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
517 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
518 
519 	if (ops->get_size)
520 		/* IFLA_INFO_DATA + nested data */
521 		size += nla_total_size(sizeof(struct nlattr)) +
522 			ops->get_size(dev);
523 
524 	if (ops->get_xstats_size)
525 		/* IFLA_INFO_XSTATS */
526 		size += nla_total_size(ops->get_xstats_size(dev));
527 
528 	size += rtnl_link_get_slave_info_data_size(dev);
529 
530 	return size;
531 }
532 
533 static LIST_HEAD(rtnl_af_ops);
534 
535 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
536 {
537 	const struct rtnl_af_ops *ops;
538 
539 	list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
540 		if (ops->family == family)
541 			return ops;
542 	}
543 
544 	return NULL;
545 }
546 
547 /**
548  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
549  * @ops: struct rtnl_af_ops * to register
550  *
551  * Returns 0 on success or a negative error code.
552  */
553 void rtnl_af_register(struct rtnl_af_ops *ops)
554 {
555 	rtnl_lock();
556 	list_add_tail_rcu(&ops->list, &rtnl_af_ops);
557 	rtnl_unlock();
558 }
559 EXPORT_SYMBOL_GPL(rtnl_af_register);
560 
561 /**
562  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
563  * @ops: struct rtnl_af_ops * to unregister
564  */
565 void rtnl_af_unregister(struct rtnl_af_ops *ops)
566 {
567 	rtnl_lock();
568 	list_del_rcu(&ops->list);
569 	rtnl_unlock();
570 
571 	synchronize_rcu();
572 }
573 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
574 
575 static size_t rtnl_link_get_af_size(const struct net_device *dev,
576 				    u32 ext_filter_mask)
577 {
578 	struct rtnl_af_ops *af_ops;
579 	size_t size;
580 
581 	/* IFLA_AF_SPEC */
582 	size = nla_total_size(sizeof(struct nlattr));
583 
584 	rcu_read_lock();
585 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
586 		if (af_ops->get_link_af_size) {
587 			/* AF_* + nested data */
588 			size += nla_total_size(sizeof(struct nlattr)) +
589 				af_ops->get_link_af_size(dev, ext_filter_mask);
590 		}
591 	}
592 	rcu_read_unlock();
593 
594 	return size;
595 }
596 
597 static bool rtnl_have_link_slave_info(const struct net_device *dev)
598 {
599 	struct net_device *master_dev;
600 	bool ret = false;
601 
602 	rcu_read_lock();
603 
604 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
605 	if (master_dev && master_dev->rtnl_link_ops)
606 		ret = true;
607 	rcu_read_unlock();
608 	return ret;
609 }
610 
611 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
612 				     const struct net_device *dev)
613 {
614 	struct net_device *master_dev;
615 	const struct rtnl_link_ops *ops;
616 	struct nlattr *slave_data;
617 	int err;
618 
619 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
620 	if (!master_dev)
621 		return 0;
622 	ops = master_dev->rtnl_link_ops;
623 	if (!ops)
624 		return 0;
625 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
626 		return -EMSGSIZE;
627 	if (ops->fill_slave_info) {
628 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
629 		if (!slave_data)
630 			return -EMSGSIZE;
631 		err = ops->fill_slave_info(skb, master_dev, dev);
632 		if (err < 0)
633 			goto err_cancel_slave_data;
634 		nla_nest_end(skb, slave_data);
635 	}
636 	return 0;
637 
638 err_cancel_slave_data:
639 	nla_nest_cancel(skb, slave_data);
640 	return err;
641 }
642 
643 static int rtnl_link_info_fill(struct sk_buff *skb,
644 			       const struct net_device *dev)
645 {
646 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
647 	struct nlattr *data;
648 	int err;
649 
650 	if (!ops)
651 		return 0;
652 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
653 		return -EMSGSIZE;
654 	if (ops->fill_xstats) {
655 		err = ops->fill_xstats(skb, dev);
656 		if (err < 0)
657 			return err;
658 	}
659 	if (ops->fill_info) {
660 		data = nla_nest_start(skb, IFLA_INFO_DATA);
661 		if (data == NULL)
662 			return -EMSGSIZE;
663 		err = ops->fill_info(skb, dev);
664 		if (err < 0)
665 			goto err_cancel_data;
666 		nla_nest_end(skb, data);
667 	}
668 	return 0;
669 
670 err_cancel_data:
671 	nla_nest_cancel(skb, data);
672 	return err;
673 }
674 
675 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
676 {
677 	struct nlattr *linkinfo;
678 	int err = -EMSGSIZE;
679 
680 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
681 	if (linkinfo == NULL)
682 		goto out;
683 
684 	err = rtnl_link_info_fill(skb, dev);
685 	if (err < 0)
686 		goto err_cancel_link;
687 
688 	err = rtnl_link_slave_info_fill(skb, dev);
689 	if (err < 0)
690 		goto err_cancel_link;
691 
692 	nla_nest_end(skb, linkinfo);
693 	return 0;
694 
695 err_cancel_link:
696 	nla_nest_cancel(skb, linkinfo);
697 out:
698 	return err;
699 }
700 
701 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
702 {
703 	struct sock *rtnl = net->rtnl;
704 	int err = 0;
705 
706 	NETLINK_CB(skb).dst_group = group;
707 	if (echo)
708 		refcount_inc(&skb->users);
709 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
710 	if (echo)
711 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
712 	return err;
713 }
714 
715 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
716 {
717 	struct sock *rtnl = net->rtnl;
718 
719 	return nlmsg_unicast(rtnl, skb, pid);
720 }
721 EXPORT_SYMBOL(rtnl_unicast);
722 
723 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
724 		 struct nlmsghdr *nlh, gfp_t flags)
725 {
726 	struct sock *rtnl = net->rtnl;
727 	int report = 0;
728 
729 	if (nlh)
730 		report = nlmsg_report(nlh);
731 
732 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
733 }
734 EXPORT_SYMBOL(rtnl_notify);
735 
736 void rtnl_set_sk_err(struct net *net, u32 group, int error)
737 {
738 	struct sock *rtnl = net->rtnl;
739 
740 	netlink_set_err(rtnl, 0, group, error);
741 }
742 EXPORT_SYMBOL(rtnl_set_sk_err);
743 
744 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
745 {
746 	struct nlattr *mx;
747 	int i, valid = 0;
748 
749 	mx = nla_nest_start(skb, RTA_METRICS);
750 	if (mx == NULL)
751 		return -ENOBUFS;
752 
753 	for (i = 0; i < RTAX_MAX; i++) {
754 		if (metrics[i]) {
755 			if (i == RTAX_CC_ALGO - 1) {
756 				char tmp[TCP_CA_NAME_MAX], *name;
757 
758 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
759 				if (!name)
760 					continue;
761 				if (nla_put_string(skb, i + 1, name))
762 					goto nla_put_failure;
763 			} else if (i == RTAX_FEATURES - 1) {
764 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
765 
766 				if (!user_features)
767 					continue;
768 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
769 				if (nla_put_u32(skb, i + 1, user_features))
770 					goto nla_put_failure;
771 			} else {
772 				if (nla_put_u32(skb, i + 1, metrics[i]))
773 					goto nla_put_failure;
774 			}
775 			valid++;
776 		}
777 	}
778 
779 	if (!valid) {
780 		nla_nest_cancel(skb, mx);
781 		return 0;
782 	}
783 
784 	return nla_nest_end(skb, mx);
785 
786 nla_put_failure:
787 	nla_nest_cancel(skb, mx);
788 	return -EMSGSIZE;
789 }
790 EXPORT_SYMBOL(rtnetlink_put_metrics);
791 
792 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
793 		       long expires, u32 error)
794 {
795 	struct rta_cacheinfo ci = {
796 		.rta_error = error,
797 		.rta_id =  id,
798 	};
799 
800 	if (dst) {
801 		ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
802 		ci.rta_used = dst->__use;
803 		ci.rta_clntref = atomic_read(&dst->__refcnt);
804 	}
805 	if (expires) {
806 		unsigned long clock;
807 
808 		clock = jiffies_to_clock_t(abs(expires));
809 		clock = min_t(unsigned long, clock, INT_MAX);
810 		ci.rta_expires = (expires > 0) ? clock : -clock;
811 	}
812 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
813 }
814 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
815 
816 static void set_operstate(struct net_device *dev, unsigned char transition)
817 {
818 	unsigned char operstate = dev->operstate;
819 
820 	switch (transition) {
821 	case IF_OPER_UP:
822 		if ((operstate == IF_OPER_DORMANT ||
823 		     operstate == IF_OPER_UNKNOWN) &&
824 		    !netif_dormant(dev))
825 			operstate = IF_OPER_UP;
826 		break;
827 
828 	case IF_OPER_DORMANT:
829 		if (operstate == IF_OPER_UP ||
830 		    operstate == IF_OPER_UNKNOWN)
831 			operstate = IF_OPER_DORMANT;
832 		break;
833 	}
834 
835 	if (dev->operstate != operstate) {
836 		write_lock_bh(&dev_base_lock);
837 		dev->operstate = operstate;
838 		write_unlock_bh(&dev_base_lock);
839 		netdev_state_change(dev);
840 	}
841 }
842 
843 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
844 {
845 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
846 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
847 }
848 
849 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
850 					   const struct ifinfomsg *ifm)
851 {
852 	unsigned int flags = ifm->ifi_flags;
853 
854 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
855 	if (ifm->ifi_change)
856 		flags = (flags & ifm->ifi_change) |
857 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
858 
859 	return flags;
860 }
861 
862 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
863 				 const struct rtnl_link_stats64 *b)
864 {
865 	a->rx_packets = b->rx_packets;
866 	a->tx_packets = b->tx_packets;
867 	a->rx_bytes = b->rx_bytes;
868 	a->tx_bytes = b->tx_bytes;
869 	a->rx_errors = b->rx_errors;
870 	a->tx_errors = b->tx_errors;
871 	a->rx_dropped = b->rx_dropped;
872 	a->tx_dropped = b->tx_dropped;
873 
874 	a->multicast = b->multicast;
875 	a->collisions = b->collisions;
876 
877 	a->rx_length_errors = b->rx_length_errors;
878 	a->rx_over_errors = b->rx_over_errors;
879 	a->rx_crc_errors = b->rx_crc_errors;
880 	a->rx_frame_errors = b->rx_frame_errors;
881 	a->rx_fifo_errors = b->rx_fifo_errors;
882 	a->rx_missed_errors = b->rx_missed_errors;
883 
884 	a->tx_aborted_errors = b->tx_aborted_errors;
885 	a->tx_carrier_errors = b->tx_carrier_errors;
886 	a->tx_fifo_errors = b->tx_fifo_errors;
887 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
888 	a->tx_window_errors = b->tx_window_errors;
889 
890 	a->rx_compressed = b->rx_compressed;
891 	a->tx_compressed = b->tx_compressed;
892 
893 	a->rx_nohandler = b->rx_nohandler;
894 }
895 
896 /* All VF info */
897 static inline int rtnl_vfinfo_size(const struct net_device *dev,
898 				   u32 ext_filter_mask)
899 {
900 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
901 		int num_vfs = dev_num_vf(dev->dev.parent);
902 		size_t size = nla_total_size(0);
903 		size += num_vfs *
904 			(nla_total_size(0) +
905 			 nla_total_size(sizeof(struct ifla_vf_mac)) +
906 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
907 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
908 			 nla_total_size(MAX_VLAN_LIST_LEN *
909 					sizeof(struct ifla_vf_vlan_info)) +
910 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
911 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
912 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
913 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
914 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
915 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
916 			 /* IFLA_VF_STATS_RX_PACKETS */
917 			 nla_total_size_64bit(sizeof(__u64)) +
918 			 /* IFLA_VF_STATS_TX_PACKETS */
919 			 nla_total_size_64bit(sizeof(__u64)) +
920 			 /* IFLA_VF_STATS_RX_BYTES */
921 			 nla_total_size_64bit(sizeof(__u64)) +
922 			 /* IFLA_VF_STATS_TX_BYTES */
923 			 nla_total_size_64bit(sizeof(__u64)) +
924 			 /* IFLA_VF_STATS_BROADCAST */
925 			 nla_total_size_64bit(sizeof(__u64)) +
926 			 /* IFLA_VF_STATS_MULTICAST */
927 			 nla_total_size_64bit(sizeof(__u64)) +
928 			 /* IFLA_VF_STATS_RX_DROPPED */
929 			 nla_total_size_64bit(sizeof(__u64)) +
930 			 /* IFLA_VF_STATS_TX_DROPPED */
931 			 nla_total_size_64bit(sizeof(__u64)) +
932 			 nla_total_size(sizeof(struct ifla_vf_trust)));
933 		return size;
934 	} else
935 		return 0;
936 }
937 
938 static size_t rtnl_port_size(const struct net_device *dev,
939 			     u32 ext_filter_mask)
940 {
941 	size_t port_size = nla_total_size(4)		/* PORT_VF */
942 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
943 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
944 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
945 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
946 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
947 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
948 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
949 		+ port_size;
950 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
951 		+ port_size;
952 
953 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
954 	    !(ext_filter_mask & RTEXT_FILTER_VF))
955 		return 0;
956 	if (dev_num_vf(dev->dev.parent))
957 		return port_self_size + vf_ports_size +
958 			vf_port_size * dev_num_vf(dev->dev.parent);
959 	else
960 		return port_self_size;
961 }
962 
963 static size_t rtnl_xdp_size(void)
964 {
965 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
966 			  nla_total_size(1) +	/* XDP_ATTACHED */
967 			  nla_total_size(4);	/* XDP_PROG_ID */
968 
969 	return xdp_size;
970 }
971 
972 static noinline size_t if_nlmsg_size(const struct net_device *dev,
973 				     u32 ext_filter_mask)
974 {
975 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
976 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
977 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
978 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
979 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
980 	       + nla_total_size(sizeof(struct rtnl_link_stats))
981 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
982 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
983 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
984 	       + nla_total_size(4) /* IFLA_TXQLEN */
985 	       + nla_total_size(4) /* IFLA_WEIGHT */
986 	       + nla_total_size(4) /* IFLA_MTU */
987 	       + nla_total_size(4) /* IFLA_LINK */
988 	       + nla_total_size(4) /* IFLA_MASTER */
989 	       + nla_total_size(1) /* IFLA_CARRIER */
990 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
991 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
992 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
993 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
994 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
995 	       + nla_total_size(1) /* IFLA_OPERSTATE */
996 	       + nla_total_size(1) /* IFLA_LINKMODE */
997 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
998 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
999 	       + nla_total_size(4) /* IFLA_GROUP */
1000 	       + nla_total_size(ext_filter_mask
1001 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1002 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1003 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1004 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1005 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1006 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1007 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1008 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1009 	       + rtnl_xdp_size() /* IFLA_XDP */
1010 	       + nla_total_size(4)  /* IFLA_EVENT */
1011 	       + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1012 	       + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1013 	       + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1014 	       + nla_total_size(4)  /* IFLA_IF_NETNSID */
1015 	       + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1016 	       + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1017 	       + 0;
1018 }
1019 
1020 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1021 {
1022 	struct nlattr *vf_ports;
1023 	struct nlattr *vf_port;
1024 	int vf;
1025 	int err;
1026 
1027 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
1028 	if (!vf_ports)
1029 		return -EMSGSIZE;
1030 
1031 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1032 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
1033 		if (!vf_port)
1034 			goto nla_put_failure;
1035 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1036 			goto nla_put_failure;
1037 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1038 		if (err == -EMSGSIZE)
1039 			goto nla_put_failure;
1040 		if (err) {
1041 			nla_nest_cancel(skb, vf_port);
1042 			continue;
1043 		}
1044 		nla_nest_end(skb, vf_port);
1045 	}
1046 
1047 	nla_nest_end(skb, vf_ports);
1048 
1049 	return 0;
1050 
1051 nla_put_failure:
1052 	nla_nest_cancel(skb, vf_ports);
1053 	return -EMSGSIZE;
1054 }
1055 
1056 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1057 {
1058 	struct nlattr *port_self;
1059 	int err;
1060 
1061 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
1062 	if (!port_self)
1063 		return -EMSGSIZE;
1064 
1065 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1066 	if (err) {
1067 		nla_nest_cancel(skb, port_self);
1068 		return (err == -EMSGSIZE) ? err : 0;
1069 	}
1070 
1071 	nla_nest_end(skb, port_self);
1072 
1073 	return 0;
1074 }
1075 
1076 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1077 			  u32 ext_filter_mask)
1078 {
1079 	int err;
1080 
1081 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1082 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1083 		return 0;
1084 
1085 	err = rtnl_port_self_fill(skb, dev);
1086 	if (err)
1087 		return err;
1088 
1089 	if (dev_num_vf(dev->dev.parent)) {
1090 		err = rtnl_vf_ports_fill(skb, dev);
1091 		if (err)
1092 			return err;
1093 	}
1094 
1095 	return 0;
1096 }
1097 
1098 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1099 {
1100 	int err;
1101 	struct netdev_phys_item_id ppid;
1102 
1103 	err = dev_get_phys_port_id(dev, &ppid);
1104 	if (err) {
1105 		if (err == -EOPNOTSUPP)
1106 			return 0;
1107 		return err;
1108 	}
1109 
1110 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1111 		return -EMSGSIZE;
1112 
1113 	return 0;
1114 }
1115 
1116 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1117 {
1118 	char name[IFNAMSIZ];
1119 	int err;
1120 
1121 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1122 	if (err) {
1123 		if (err == -EOPNOTSUPP)
1124 			return 0;
1125 		return err;
1126 	}
1127 
1128 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1129 		return -EMSGSIZE;
1130 
1131 	return 0;
1132 }
1133 
1134 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1135 {
1136 	int err;
1137 	struct switchdev_attr attr = {
1138 		.orig_dev = dev,
1139 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1140 		.flags = SWITCHDEV_F_NO_RECURSE,
1141 	};
1142 
1143 	err = switchdev_port_attr_get(dev, &attr);
1144 	if (err) {
1145 		if (err == -EOPNOTSUPP)
1146 			return 0;
1147 		return err;
1148 	}
1149 
1150 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1151 		    attr.u.ppid.id))
1152 		return -EMSGSIZE;
1153 
1154 	return 0;
1155 }
1156 
1157 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1158 					      struct net_device *dev)
1159 {
1160 	struct rtnl_link_stats64 *sp;
1161 	struct nlattr *attr;
1162 
1163 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1164 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1165 	if (!attr)
1166 		return -EMSGSIZE;
1167 
1168 	sp = nla_data(attr);
1169 	dev_get_stats(dev, sp);
1170 
1171 	attr = nla_reserve(skb, IFLA_STATS,
1172 			   sizeof(struct rtnl_link_stats));
1173 	if (!attr)
1174 		return -EMSGSIZE;
1175 
1176 	copy_rtnl_link_stats(nla_data(attr), sp);
1177 
1178 	return 0;
1179 }
1180 
1181 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1182 					       struct net_device *dev,
1183 					       int vfs_num,
1184 					       struct nlattr *vfinfo)
1185 {
1186 	struct ifla_vf_rss_query_en vf_rss_query_en;
1187 	struct nlattr *vf, *vfstats, *vfvlanlist;
1188 	struct ifla_vf_link_state vf_linkstate;
1189 	struct ifla_vf_vlan_info vf_vlan_info;
1190 	struct ifla_vf_spoofchk vf_spoofchk;
1191 	struct ifla_vf_tx_rate vf_tx_rate;
1192 	struct ifla_vf_stats vf_stats;
1193 	struct ifla_vf_trust vf_trust;
1194 	struct ifla_vf_vlan vf_vlan;
1195 	struct ifla_vf_rate vf_rate;
1196 	struct ifla_vf_mac vf_mac;
1197 	struct ifla_vf_info ivi;
1198 
1199 	memset(&ivi, 0, sizeof(ivi));
1200 
1201 	/* Not all SR-IOV capable drivers support the
1202 	 * spoofcheck and "RSS query enable" query.  Preset to
1203 	 * -1 so the user space tool can detect that the driver
1204 	 * didn't report anything.
1205 	 */
1206 	ivi.spoofchk = -1;
1207 	ivi.rss_query_en = -1;
1208 	ivi.trusted = -1;
1209 	/* The default value for VF link state is "auto"
1210 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1211 	 */
1212 	ivi.linkstate = 0;
1213 	/* VLAN Protocol by default is 802.1Q */
1214 	ivi.vlan_proto = htons(ETH_P_8021Q);
1215 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1216 		return 0;
1217 
1218 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1219 
1220 	vf_mac.vf =
1221 		vf_vlan.vf =
1222 		vf_vlan_info.vf =
1223 		vf_rate.vf =
1224 		vf_tx_rate.vf =
1225 		vf_spoofchk.vf =
1226 		vf_linkstate.vf =
1227 		vf_rss_query_en.vf =
1228 		vf_trust.vf = ivi.vf;
1229 
1230 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1231 	vf_vlan.vlan = ivi.vlan;
1232 	vf_vlan.qos = ivi.qos;
1233 	vf_vlan_info.vlan = ivi.vlan;
1234 	vf_vlan_info.qos = ivi.qos;
1235 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1236 	vf_tx_rate.rate = ivi.max_tx_rate;
1237 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1238 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1239 	vf_spoofchk.setting = ivi.spoofchk;
1240 	vf_linkstate.link_state = ivi.linkstate;
1241 	vf_rss_query_en.setting = ivi.rss_query_en;
1242 	vf_trust.setting = ivi.trusted;
1243 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1244 	if (!vf)
1245 		goto nla_put_vfinfo_failure;
1246 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1247 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1248 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1249 		    &vf_rate) ||
1250 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1251 		    &vf_tx_rate) ||
1252 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1253 		    &vf_spoofchk) ||
1254 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1255 		    &vf_linkstate) ||
1256 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1257 		    sizeof(vf_rss_query_en),
1258 		    &vf_rss_query_en) ||
1259 	    nla_put(skb, IFLA_VF_TRUST,
1260 		    sizeof(vf_trust), &vf_trust))
1261 		goto nla_put_vf_failure;
1262 	vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1263 	if (!vfvlanlist)
1264 		goto nla_put_vf_failure;
1265 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1266 		    &vf_vlan_info)) {
1267 		nla_nest_cancel(skb, vfvlanlist);
1268 		goto nla_put_vf_failure;
1269 	}
1270 	nla_nest_end(skb, vfvlanlist);
1271 	memset(&vf_stats, 0, sizeof(vf_stats));
1272 	if (dev->netdev_ops->ndo_get_vf_stats)
1273 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1274 						&vf_stats);
1275 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1276 	if (!vfstats)
1277 		goto nla_put_vf_failure;
1278 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1279 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1280 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1281 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1282 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1283 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1284 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1285 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1286 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1287 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1288 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1289 			      vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1290 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1291 			      vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1292 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1293 			      vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1294 		nla_nest_cancel(skb, vfstats);
1295 		goto nla_put_vf_failure;
1296 	}
1297 	nla_nest_end(skb, vfstats);
1298 	nla_nest_end(skb, vf);
1299 	return 0;
1300 
1301 nla_put_vf_failure:
1302 	nla_nest_cancel(skb, vf);
1303 nla_put_vfinfo_failure:
1304 	nla_nest_cancel(skb, vfinfo);
1305 	return -EMSGSIZE;
1306 }
1307 
1308 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1309 					   struct net_device *dev,
1310 					   u32 ext_filter_mask)
1311 {
1312 	struct nlattr *vfinfo;
1313 	int i, num_vfs;
1314 
1315 	if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1316 		return 0;
1317 
1318 	num_vfs = dev_num_vf(dev->dev.parent);
1319 	if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1320 		return -EMSGSIZE;
1321 
1322 	if (!dev->netdev_ops->ndo_get_vf_config)
1323 		return 0;
1324 
1325 	vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1326 	if (!vfinfo)
1327 		return -EMSGSIZE;
1328 
1329 	for (i = 0; i < num_vfs; i++) {
1330 		if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1331 			return -EMSGSIZE;
1332 	}
1333 
1334 	nla_nest_end(skb, vfinfo);
1335 	return 0;
1336 }
1337 
1338 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1339 {
1340 	struct rtnl_link_ifmap map;
1341 
1342 	memset(&map, 0, sizeof(map));
1343 	map.mem_start   = dev->mem_start;
1344 	map.mem_end     = dev->mem_end;
1345 	map.base_addr   = dev->base_addr;
1346 	map.irq         = dev->irq;
1347 	map.dma         = dev->dma;
1348 	map.port        = dev->if_port;
1349 
1350 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1351 		return -EMSGSIZE;
1352 
1353 	return 0;
1354 }
1355 
1356 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1357 {
1358 	const struct net_device_ops *ops = dev->netdev_ops;
1359 	const struct bpf_prog *generic_xdp_prog;
1360 	struct netdev_bpf xdp;
1361 
1362 	ASSERT_RTNL();
1363 
1364 	*prog_id = 0;
1365 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1366 	if (generic_xdp_prog) {
1367 		*prog_id = generic_xdp_prog->aux->id;
1368 		return XDP_ATTACHED_SKB;
1369 	}
1370 	if (!ops->ndo_bpf)
1371 		return XDP_ATTACHED_NONE;
1372 
1373 	__dev_xdp_query(dev, ops->ndo_bpf, &xdp);
1374 	*prog_id = xdp.prog_id;
1375 
1376 	return xdp.prog_attached;
1377 }
1378 
1379 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1380 {
1381 	struct nlattr *xdp;
1382 	u32 prog_id;
1383 	int err;
1384 
1385 	xdp = nla_nest_start(skb, IFLA_XDP);
1386 	if (!xdp)
1387 		return -EMSGSIZE;
1388 
1389 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1390 			 rtnl_xdp_attached_mode(dev, &prog_id));
1391 	if (err)
1392 		goto err_cancel;
1393 
1394 	if (prog_id) {
1395 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1396 		if (err)
1397 			goto err_cancel;
1398 	}
1399 
1400 	nla_nest_end(skb, xdp);
1401 	return 0;
1402 
1403 err_cancel:
1404 	nla_nest_cancel(skb, xdp);
1405 	return err;
1406 }
1407 
1408 static u32 rtnl_get_event(unsigned long event)
1409 {
1410 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1411 
1412 	switch (event) {
1413 	case NETDEV_REBOOT:
1414 		rtnl_event_type = IFLA_EVENT_REBOOT;
1415 		break;
1416 	case NETDEV_FEAT_CHANGE:
1417 		rtnl_event_type = IFLA_EVENT_FEATURES;
1418 		break;
1419 	case NETDEV_BONDING_FAILOVER:
1420 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1421 		break;
1422 	case NETDEV_NOTIFY_PEERS:
1423 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1424 		break;
1425 	case NETDEV_RESEND_IGMP:
1426 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1427 		break;
1428 	case NETDEV_CHANGEINFODATA:
1429 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1430 		break;
1431 	default:
1432 		break;
1433 	}
1434 
1435 	return rtnl_event_type;
1436 }
1437 
1438 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1439 {
1440 	const struct net_device *upper_dev;
1441 	int ret = 0;
1442 
1443 	rcu_read_lock();
1444 
1445 	upper_dev = netdev_master_upper_dev_get_rcu(dev);
1446 	if (upper_dev)
1447 		ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1448 
1449 	rcu_read_unlock();
1450 	return ret;
1451 }
1452 
1453 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1454 {
1455 	int ifindex = dev_get_iflink(dev);
1456 
1457 	if (dev->ifindex == ifindex)
1458 		return 0;
1459 
1460 	return nla_put_u32(skb, IFLA_LINK, ifindex);
1461 }
1462 
1463 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1464 					      struct net_device *dev)
1465 {
1466 	char buf[IFALIASZ];
1467 	int ret;
1468 
1469 	ret = dev_get_alias(dev, buf, sizeof(buf));
1470 	return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1471 }
1472 
1473 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1474 				  const struct net_device *dev,
1475 				  struct net *src_net)
1476 {
1477 	if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1478 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1479 
1480 		if (!net_eq(dev_net(dev), link_net)) {
1481 			int id = peernet2id_alloc(src_net, link_net);
1482 
1483 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1484 				return -EMSGSIZE;
1485 		}
1486 	}
1487 
1488 	return 0;
1489 }
1490 
1491 static int rtnl_fill_link_af(struct sk_buff *skb,
1492 			     const struct net_device *dev,
1493 			     u32 ext_filter_mask)
1494 {
1495 	const struct rtnl_af_ops *af_ops;
1496 	struct nlattr *af_spec;
1497 
1498 	af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1499 	if (!af_spec)
1500 		return -EMSGSIZE;
1501 
1502 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1503 		struct nlattr *af;
1504 		int err;
1505 
1506 		if (!af_ops->fill_link_af)
1507 			continue;
1508 
1509 		af = nla_nest_start(skb, af_ops->family);
1510 		if (!af)
1511 			return -EMSGSIZE;
1512 
1513 		err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1514 		/*
1515 		 * Caller may return ENODATA to indicate that there
1516 		 * was no data to be dumped. This is not an error, it
1517 		 * means we should trim the attribute header and
1518 		 * continue.
1519 		 */
1520 		if (err == -ENODATA)
1521 			nla_nest_cancel(skb, af);
1522 		else if (err < 0)
1523 			return -EMSGSIZE;
1524 
1525 		nla_nest_end(skb, af);
1526 	}
1527 
1528 	nla_nest_end(skb, af_spec);
1529 	return 0;
1530 }
1531 
1532 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1533 			    struct net_device *dev, struct net *src_net,
1534 			    int type, u32 pid, u32 seq, u32 change,
1535 			    unsigned int flags, u32 ext_filter_mask,
1536 			    u32 event, int *new_nsid, int new_ifindex,
1537 			    int tgt_netnsid)
1538 {
1539 	struct ifinfomsg *ifm;
1540 	struct nlmsghdr *nlh;
1541 
1542 	ASSERT_RTNL();
1543 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1544 	if (nlh == NULL)
1545 		return -EMSGSIZE;
1546 
1547 	ifm = nlmsg_data(nlh);
1548 	ifm->ifi_family = AF_UNSPEC;
1549 	ifm->__ifi_pad = 0;
1550 	ifm->ifi_type = dev->type;
1551 	ifm->ifi_index = dev->ifindex;
1552 	ifm->ifi_flags = dev_get_flags(dev);
1553 	ifm->ifi_change = change;
1554 
1555 	if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid))
1556 		goto nla_put_failure;
1557 
1558 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1559 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1560 	    nla_put_u8(skb, IFLA_OPERSTATE,
1561 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1562 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1563 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1564 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1565 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1566 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1567 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1568 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1569 #ifdef CONFIG_RPS
1570 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1571 #endif
1572 	    nla_put_iflink(skb, dev) ||
1573 	    put_master_ifindex(skb, dev) ||
1574 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1575 	    (dev->qdisc &&
1576 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1577 	    nla_put_ifalias(skb, dev) ||
1578 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1579 			atomic_read(&dev->carrier_up_count) +
1580 			atomic_read(&dev->carrier_down_count)) ||
1581 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1582 	    nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1583 			atomic_read(&dev->carrier_up_count)) ||
1584 	    nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1585 			atomic_read(&dev->carrier_down_count)))
1586 		goto nla_put_failure;
1587 
1588 	if (event != IFLA_EVENT_NONE) {
1589 		if (nla_put_u32(skb, IFLA_EVENT, event))
1590 			goto nla_put_failure;
1591 	}
1592 
1593 	if (rtnl_fill_link_ifmap(skb, dev))
1594 		goto nla_put_failure;
1595 
1596 	if (dev->addr_len) {
1597 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1598 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1599 			goto nla_put_failure;
1600 	}
1601 
1602 	if (rtnl_phys_port_id_fill(skb, dev))
1603 		goto nla_put_failure;
1604 
1605 	if (rtnl_phys_port_name_fill(skb, dev))
1606 		goto nla_put_failure;
1607 
1608 	if (rtnl_phys_switch_id_fill(skb, dev))
1609 		goto nla_put_failure;
1610 
1611 	if (rtnl_fill_stats(skb, dev))
1612 		goto nla_put_failure;
1613 
1614 	if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1615 		goto nla_put_failure;
1616 
1617 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1618 		goto nla_put_failure;
1619 
1620 	if (rtnl_xdp_fill(skb, dev))
1621 		goto nla_put_failure;
1622 
1623 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1624 		if (rtnl_link_fill(skb, dev) < 0)
1625 			goto nla_put_failure;
1626 	}
1627 
1628 	if (rtnl_fill_link_netnsid(skb, dev, src_net))
1629 		goto nla_put_failure;
1630 
1631 	if (new_nsid &&
1632 	    nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1633 		goto nla_put_failure;
1634 	if (new_ifindex &&
1635 	    nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1636 		goto nla_put_failure;
1637 
1638 
1639 	rcu_read_lock();
1640 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1641 		goto nla_put_failure_rcu;
1642 	rcu_read_unlock();
1643 
1644 	nlmsg_end(skb, nlh);
1645 	return 0;
1646 
1647 nla_put_failure_rcu:
1648 	rcu_read_unlock();
1649 nla_put_failure:
1650 	nlmsg_cancel(skb, nlh);
1651 	return -EMSGSIZE;
1652 }
1653 
1654 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1655 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1656 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1657 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1658 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1659 	[IFLA_MTU]		= { .type = NLA_U32 },
1660 	[IFLA_LINK]		= { .type = NLA_U32 },
1661 	[IFLA_MASTER]		= { .type = NLA_U32 },
1662 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1663 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1664 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1665 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1666 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1667 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1668 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1669 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1670 	/* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1671 	 * allow 0-length string (needed to remove an alias).
1672 	 */
1673 	[IFLA_IFALIAS]	        = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1674 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1675 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1676 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1677 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1678 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1679 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1680 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1681 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1682 	[IFLA_GSO_MAX_SEGS]	= { .type = NLA_U32 },
1683 	[IFLA_GSO_MAX_SIZE]	= { .type = NLA_U32 },
1684 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1685 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1686 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1687 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1688 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1689 	[IFLA_XDP]		= { .type = NLA_NESTED },
1690 	[IFLA_EVENT]		= { .type = NLA_U32 },
1691 	[IFLA_GROUP]		= { .type = NLA_U32 },
1692 	[IFLA_IF_NETNSID]	= { .type = NLA_S32 },
1693 	[IFLA_CARRIER_UP_COUNT]	= { .type = NLA_U32 },
1694 	[IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1695 };
1696 
1697 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1698 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1699 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1700 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1701 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1702 };
1703 
1704 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1705 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1706 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1707 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1708 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1709 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1710 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1711 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1712 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1713 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1714 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1715 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1716 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1717 };
1718 
1719 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1720 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1721 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1722 				    .len = PORT_PROFILE_MAX },
1723 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1724 				      .len = PORT_UUID_MAX },
1725 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1726 				    .len = PORT_UUID_MAX },
1727 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1728 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1729 
1730 	/* Unused, but we need to keep it here since user space could
1731 	 * fill it. It's also broken with regard to NLA_BINARY use in
1732 	 * combination with structs.
1733 	 */
1734 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1735 				    .len = sizeof(struct ifla_port_vsi) },
1736 };
1737 
1738 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1739 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1740 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1741 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1742 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1743 };
1744 
1745 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1746 {
1747 	const struct rtnl_link_ops *ops = NULL;
1748 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1749 
1750 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1751 			     ifla_info_policy, NULL) < 0)
1752 		return NULL;
1753 
1754 	if (linfo[IFLA_INFO_KIND]) {
1755 		char kind[MODULE_NAME_LEN];
1756 
1757 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1758 		ops = rtnl_link_ops_get(kind);
1759 	}
1760 
1761 	return ops;
1762 }
1763 
1764 static bool link_master_filtered(struct net_device *dev, int master_idx)
1765 {
1766 	struct net_device *master;
1767 
1768 	if (!master_idx)
1769 		return false;
1770 
1771 	master = netdev_master_upper_dev_get(dev);
1772 	if (!master || master->ifindex != master_idx)
1773 		return true;
1774 
1775 	return false;
1776 }
1777 
1778 static bool link_kind_filtered(const struct net_device *dev,
1779 			       const struct rtnl_link_ops *kind_ops)
1780 {
1781 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1782 		return true;
1783 
1784 	return false;
1785 }
1786 
1787 static bool link_dump_filtered(struct net_device *dev,
1788 			       int master_idx,
1789 			       const struct rtnl_link_ops *kind_ops)
1790 {
1791 	if (link_master_filtered(dev, master_idx) ||
1792 	    link_kind_filtered(dev, kind_ops))
1793 		return true;
1794 
1795 	return false;
1796 }
1797 
1798 static struct net *get_target_net(struct sock *sk, int netnsid)
1799 {
1800 	struct net *net;
1801 
1802 	net = get_net_ns_by_id(sock_net(sk), netnsid);
1803 	if (!net)
1804 		return ERR_PTR(-EINVAL);
1805 
1806 	/* For now, the caller is required to have CAP_NET_ADMIN in
1807 	 * the user namespace owning the target net ns.
1808 	 */
1809 	if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1810 		put_net(net);
1811 		return ERR_PTR(-EACCES);
1812 	}
1813 	return net;
1814 }
1815 
1816 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1817 {
1818 	struct net *net = sock_net(skb->sk);
1819 	struct net *tgt_net = net;
1820 	int h, s_h;
1821 	int idx = 0, s_idx;
1822 	struct net_device *dev;
1823 	struct hlist_head *head;
1824 	struct nlattr *tb[IFLA_MAX+1];
1825 	u32 ext_filter_mask = 0;
1826 	const struct rtnl_link_ops *kind_ops = NULL;
1827 	unsigned int flags = NLM_F_MULTI;
1828 	int master_idx = 0;
1829 	int netnsid = -1;
1830 	int err;
1831 	int hdrlen;
1832 
1833 	s_h = cb->args[0];
1834 	s_idx = cb->args[1];
1835 
1836 	/* A hack to preserve kernel<->userspace interface.
1837 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1838 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1839 	 * what iproute2 < v3.9.0 used.
1840 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1841 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1842 	 */
1843 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1844 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1845 
1846 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1847 			ifla_policy, NULL) >= 0) {
1848 		if (tb[IFLA_IF_NETNSID]) {
1849 			netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
1850 			tgt_net = get_target_net(skb->sk, netnsid);
1851 			if (IS_ERR(tgt_net)) {
1852 				tgt_net = net;
1853 				netnsid = -1;
1854 			}
1855 		}
1856 
1857 		if (tb[IFLA_EXT_MASK])
1858 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1859 
1860 		if (tb[IFLA_MASTER])
1861 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1862 
1863 		if (tb[IFLA_LINKINFO])
1864 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1865 
1866 		if (master_idx || kind_ops)
1867 			flags |= NLM_F_DUMP_FILTERED;
1868 	}
1869 
1870 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1871 		idx = 0;
1872 		head = &tgt_net->dev_index_head[h];
1873 		hlist_for_each_entry(dev, head, index_hlist) {
1874 			if (link_dump_filtered(dev, master_idx, kind_ops))
1875 				goto cont;
1876 			if (idx < s_idx)
1877 				goto cont;
1878 			err = rtnl_fill_ifinfo(skb, dev, net,
1879 					       RTM_NEWLINK,
1880 					       NETLINK_CB(cb->skb).portid,
1881 					       cb->nlh->nlmsg_seq, 0,
1882 					       flags,
1883 					       ext_filter_mask, 0, NULL, 0,
1884 					       netnsid);
1885 
1886 			if (err < 0) {
1887 				if (likely(skb->len))
1888 					goto out;
1889 
1890 				goto out_err;
1891 			}
1892 cont:
1893 			idx++;
1894 		}
1895 	}
1896 out:
1897 	err = skb->len;
1898 out_err:
1899 	cb->args[1] = idx;
1900 	cb->args[0] = h;
1901 	cb->seq = net->dev_base_seq;
1902 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1903 	if (netnsid >= 0)
1904 		put_net(tgt_net);
1905 
1906 	return err;
1907 }
1908 
1909 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1910 			struct netlink_ext_ack *exterr)
1911 {
1912 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1913 }
1914 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1915 
1916 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1917 {
1918 	struct net *net;
1919 	/* Examine the link attributes and figure out which
1920 	 * network namespace we are talking about.
1921 	 */
1922 	if (tb[IFLA_NET_NS_PID])
1923 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1924 	else if (tb[IFLA_NET_NS_FD])
1925 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1926 	else
1927 		net = get_net(src_net);
1928 	return net;
1929 }
1930 EXPORT_SYMBOL(rtnl_link_get_net);
1931 
1932 /* Figure out which network namespace we are talking about by
1933  * examining the link attributes in the following order:
1934  *
1935  * 1. IFLA_NET_NS_PID
1936  * 2. IFLA_NET_NS_FD
1937  * 3. IFLA_IF_NETNSID
1938  */
1939 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
1940 					       struct nlattr *tb[])
1941 {
1942 	struct net *net;
1943 
1944 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
1945 		return rtnl_link_get_net(src_net, tb);
1946 
1947 	if (!tb[IFLA_IF_NETNSID])
1948 		return get_net(src_net);
1949 
1950 	net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID]));
1951 	if (!net)
1952 		return ERR_PTR(-EINVAL);
1953 
1954 	return net;
1955 }
1956 
1957 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
1958 					     struct net *src_net,
1959 					     struct nlattr *tb[], int cap)
1960 {
1961 	struct net *net;
1962 
1963 	net = rtnl_link_get_net_by_nlattr(src_net, tb);
1964 	if (IS_ERR(net))
1965 		return net;
1966 
1967 	if (!netlink_ns_capable(skb, net->user_ns, cap)) {
1968 		put_net(net);
1969 		return ERR_PTR(-EPERM);
1970 	}
1971 
1972 	return net;
1973 }
1974 
1975 /* Verify that rtnetlink requests do not pass additional properties
1976  * potentially referring to different network namespaces.
1977  */
1978 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
1979 				    struct netlink_ext_ack *extack,
1980 				    bool netns_id_only)
1981 {
1982 
1983 	if (netns_id_only) {
1984 		if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
1985 			return 0;
1986 
1987 		NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
1988 		return -EOPNOTSUPP;
1989 	}
1990 
1991 	if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
1992 		goto invalid_attr;
1993 
1994 	if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD]))
1995 		goto invalid_attr;
1996 
1997 	if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID]))
1998 		goto invalid_attr;
1999 
2000 	return 0;
2001 
2002 invalid_attr:
2003 	NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2004 	return -EINVAL;
2005 }
2006 
2007 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2008 {
2009 	if (dev) {
2010 		if (tb[IFLA_ADDRESS] &&
2011 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2012 			return -EINVAL;
2013 
2014 		if (tb[IFLA_BROADCAST] &&
2015 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2016 			return -EINVAL;
2017 	}
2018 
2019 	if (tb[IFLA_AF_SPEC]) {
2020 		struct nlattr *af;
2021 		int rem, err;
2022 
2023 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2024 			const struct rtnl_af_ops *af_ops;
2025 
2026 			rcu_read_lock();
2027 			af_ops = rtnl_af_lookup(nla_type(af));
2028 			if (!af_ops) {
2029 				rcu_read_unlock();
2030 				return -EAFNOSUPPORT;
2031 			}
2032 
2033 			if (!af_ops->set_link_af) {
2034 				rcu_read_unlock();
2035 				return -EOPNOTSUPP;
2036 			}
2037 
2038 			if (af_ops->validate_link_af) {
2039 				err = af_ops->validate_link_af(dev, af);
2040 				if (err < 0) {
2041 					rcu_read_unlock();
2042 					return err;
2043 				}
2044 			}
2045 
2046 			rcu_read_unlock();
2047 		}
2048 	}
2049 
2050 	return 0;
2051 }
2052 
2053 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2054 				  int guid_type)
2055 {
2056 	const struct net_device_ops *ops = dev->netdev_ops;
2057 
2058 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2059 }
2060 
2061 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2062 {
2063 	if (dev->type != ARPHRD_INFINIBAND)
2064 		return -EOPNOTSUPP;
2065 
2066 	return handle_infiniband_guid(dev, ivt, guid_type);
2067 }
2068 
2069 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2070 {
2071 	const struct net_device_ops *ops = dev->netdev_ops;
2072 	int err = -EINVAL;
2073 
2074 	if (tb[IFLA_VF_MAC]) {
2075 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2076 
2077 		err = -EOPNOTSUPP;
2078 		if (ops->ndo_set_vf_mac)
2079 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
2080 						  ivm->mac);
2081 		if (err < 0)
2082 			return err;
2083 	}
2084 
2085 	if (tb[IFLA_VF_VLAN]) {
2086 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2087 
2088 		err = -EOPNOTSUPP;
2089 		if (ops->ndo_set_vf_vlan)
2090 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2091 						   ivv->qos,
2092 						   htons(ETH_P_8021Q));
2093 		if (err < 0)
2094 			return err;
2095 	}
2096 
2097 	if (tb[IFLA_VF_VLAN_LIST]) {
2098 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2099 		struct nlattr *attr;
2100 		int rem, len = 0;
2101 
2102 		err = -EOPNOTSUPP;
2103 		if (!ops->ndo_set_vf_vlan)
2104 			return err;
2105 
2106 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2107 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2108 			    nla_len(attr) < NLA_HDRLEN) {
2109 				return -EINVAL;
2110 			}
2111 			if (len >= MAX_VLAN_LIST_LEN)
2112 				return -EOPNOTSUPP;
2113 			ivvl[len] = nla_data(attr);
2114 
2115 			len++;
2116 		}
2117 		if (len == 0)
2118 			return -EINVAL;
2119 
2120 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2121 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
2122 		if (err < 0)
2123 			return err;
2124 	}
2125 
2126 	if (tb[IFLA_VF_TX_RATE]) {
2127 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2128 		struct ifla_vf_info ivf;
2129 
2130 		err = -EOPNOTSUPP;
2131 		if (ops->ndo_get_vf_config)
2132 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2133 		if (err < 0)
2134 			return err;
2135 
2136 		err = -EOPNOTSUPP;
2137 		if (ops->ndo_set_vf_rate)
2138 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2139 						   ivf.min_tx_rate,
2140 						   ivt->rate);
2141 		if (err < 0)
2142 			return err;
2143 	}
2144 
2145 	if (tb[IFLA_VF_RATE]) {
2146 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2147 
2148 		err = -EOPNOTSUPP;
2149 		if (ops->ndo_set_vf_rate)
2150 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2151 						   ivt->min_tx_rate,
2152 						   ivt->max_tx_rate);
2153 		if (err < 0)
2154 			return err;
2155 	}
2156 
2157 	if (tb[IFLA_VF_SPOOFCHK]) {
2158 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2159 
2160 		err = -EOPNOTSUPP;
2161 		if (ops->ndo_set_vf_spoofchk)
2162 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2163 						       ivs->setting);
2164 		if (err < 0)
2165 			return err;
2166 	}
2167 
2168 	if (tb[IFLA_VF_LINK_STATE]) {
2169 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2170 
2171 		err = -EOPNOTSUPP;
2172 		if (ops->ndo_set_vf_link_state)
2173 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2174 							 ivl->link_state);
2175 		if (err < 0)
2176 			return err;
2177 	}
2178 
2179 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
2180 		struct ifla_vf_rss_query_en *ivrssq_en;
2181 
2182 		err = -EOPNOTSUPP;
2183 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2184 		if (ops->ndo_set_vf_rss_query_en)
2185 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2186 							   ivrssq_en->setting);
2187 		if (err < 0)
2188 			return err;
2189 	}
2190 
2191 	if (tb[IFLA_VF_TRUST]) {
2192 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2193 
2194 		err = -EOPNOTSUPP;
2195 		if (ops->ndo_set_vf_trust)
2196 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2197 		if (err < 0)
2198 			return err;
2199 	}
2200 
2201 	if (tb[IFLA_VF_IB_NODE_GUID]) {
2202 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2203 
2204 		if (!ops->ndo_set_vf_guid)
2205 			return -EOPNOTSUPP;
2206 
2207 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2208 	}
2209 
2210 	if (tb[IFLA_VF_IB_PORT_GUID]) {
2211 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2212 
2213 		if (!ops->ndo_set_vf_guid)
2214 			return -EOPNOTSUPP;
2215 
2216 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2217 	}
2218 
2219 	return err;
2220 }
2221 
2222 static int do_set_master(struct net_device *dev, int ifindex,
2223 			 struct netlink_ext_ack *extack)
2224 {
2225 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2226 	const struct net_device_ops *ops;
2227 	int err;
2228 
2229 	if (upper_dev) {
2230 		if (upper_dev->ifindex == ifindex)
2231 			return 0;
2232 		ops = upper_dev->netdev_ops;
2233 		if (ops->ndo_del_slave) {
2234 			err = ops->ndo_del_slave(upper_dev, dev);
2235 			if (err)
2236 				return err;
2237 		} else {
2238 			return -EOPNOTSUPP;
2239 		}
2240 	}
2241 
2242 	if (ifindex) {
2243 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2244 		if (!upper_dev)
2245 			return -EINVAL;
2246 		ops = upper_dev->netdev_ops;
2247 		if (ops->ndo_add_slave) {
2248 			err = ops->ndo_add_slave(upper_dev, dev, extack);
2249 			if (err)
2250 				return err;
2251 		} else {
2252 			return -EOPNOTSUPP;
2253 		}
2254 	}
2255 	return 0;
2256 }
2257 
2258 #define DO_SETLINK_MODIFIED	0x01
2259 /* notify flag means notify + modified. */
2260 #define DO_SETLINK_NOTIFY	0x03
2261 static int do_setlink(const struct sk_buff *skb,
2262 		      struct net_device *dev, struct ifinfomsg *ifm,
2263 		      struct netlink_ext_ack *extack,
2264 		      struct nlattr **tb, char *ifname, int status)
2265 {
2266 	const struct net_device_ops *ops = dev->netdev_ops;
2267 	int err;
2268 
2269 	err = validate_linkmsg(dev, tb);
2270 	if (err < 0)
2271 		return err;
2272 
2273 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) {
2274 		struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2275 							    tb, CAP_NET_ADMIN);
2276 		if (IS_ERR(net)) {
2277 			err = PTR_ERR(net);
2278 			goto errout;
2279 		}
2280 
2281 		err = dev_change_net_namespace(dev, net, ifname);
2282 		put_net(net);
2283 		if (err)
2284 			goto errout;
2285 		status |= DO_SETLINK_MODIFIED;
2286 	}
2287 
2288 	if (tb[IFLA_MAP]) {
2289 		struct rtnl_link_ifmap *u_map;
2290 		struct ifmap k_map;
2291 
2292 		if (!ops->ndo_set_config) {
2293 			err = -EOPNOTSUPP;
2294 			goto errout;
2295 		}
2296 
2297 		if (!netif_device_present(dev)) {
2298 			err = -ENODEV;
2299 			goto errout;
2300 		}
2301 
2302 		u_map = nla_data(tb[IFLA_MAP]);
2303 		k_map.mem_start = (unsigned long) u_map->mem_start;
2304 		k_map.mem_end = (unsigned long) u_map->mem_end;
2305 		k_map.base_addr = (unsigned short) u_map->base_addr;
2306 		k_map.irq = (unsigned char) u_map->irq;
2307 		k_map.dma = (unsigned char) u_map->dma;
2308 		k_map.port = (unsigned char) u_map->port;
2309 
2310 		err = ops->ndo_set_config(dev, &k_map);
2311 		if (err < 0)
2312 			goto errout;
2313 
2314 		status |= DO_SETLINK_NOTIFY;
2315 	}
2316 
2317 	if (tb[IFLA_ADDRESS]) {
2318 		struct sockaddr *sa;
2319 		int len;
2320 
2321 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2322 						  sizeof(*sa));
2323 		sa = kmalloc(len, GFP_KERNEL);
2324 		if (!sa) {
2325 			err = -ENOMEM;
2326 			goto errout;
2327 		}
2328 		sa->sa_family = dev->type;
2329 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2330 		       dev->addr_len);
2331 		err = dev_set_mac_address(dev, sa);
2332 		kfree(sa);
2333 		if (err)
2334 			goto errout;
2335 		status |= DO_SETLINK_MODIFIED;
2336 	}
2337 
2338 	if (tb[IFLA_MTU]) {
2339 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2340 		if (err < 0)
2341 			goto errout;
2342 		status |= DO_SETLINK_MODIFIED;
2343 	}
2344 
2345 	if (tb[IFLA_GROUP]) {
2346 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2347 		status |= DO_SETLINK_NOTIFY;
2348 	}
2349 
2350 	/*
2351 	 * Interface selected by interface index but interface
2352 	 * name provided implies that a name change has been
2353 	 * requested.
2354 	 */
2355 	if (ifm->ifi_index > 0 && ifname[0]) {
2356 		err = dev_change_name(dev, ifname);
2357 		if (err < 0)
2358 			goto errout;
2359 		status |= DO_SETLINK_MODIFIED;
2360 	}
2361 
2362 	if (tb[IFLA_IFALIAS]) {
2363 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2364 				    nla_len(tb[IFLA_IFALIAS]));
2365 		if (err < 0)
2366 			goto errout;
2367 		status |= DO_SETLINK_NOTIFY;
2368 	}
2369 
2370 	if (tb[IFLA_BROADCAST]) {
2371 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2372 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2373 	}
2374 
2375 	if (ifm->ifi_flags || ifm->ifi_change) {
2376 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2377 		if (err < 0)
2378 			goto errout;
2379 	}
2380 
2381 	if (tb[IFLA_MASTER]) {
2382 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2383 		if (err)
2384 			goto errout;
2385 		status |= DO_SETLINK_MODIFIED;
2386 	}
2387 
2388 	if (tb[IFLA_CARRIER]) {
2389 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2390 		if (err)
2391 			goto errout;
2392 		status |= DO_SETLINK_MODIFIED;
2393 	}
2394 
2395 	if (tb[IFLA_TXQLEN]) {
2396 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2397 
2398 		err = dev_change_tx_queue_len(dev, value);
2399 		if (err)
2400 			goto errout;
2401 		status |= DO_SETLINK_MODIFIED;
2402 	}
2403 
2404 	if (tb[IFLA_GSO_MAX_SIZE]) {
2405 		u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2406 
2407 		if (max_size > GSO_MAX_SIZE) {
2408 			err = -EINVAL;
2409 			goto errout;
2410 		}
2411 
2412 		if (dev->gso_max_size ^ max_size) {
2413 			netif_set_gso_max_size(dev, max_size);
2414 			status |= DO_SETLINK_MODIFIED;
2415 		}
2416 	}
2417 
2418 	if (tb[IFLA_GSO_MAX_SEGS]) {
2419 		u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2420 
2421 		if (max_segs > GSO_MAX_SEGS) {
2422 			err = -EINVAL;
2423 			goto errout;
2424 		}
2425 
2426 		if (dev->gso_max_segs ^ max_segs) {
2427 			dev->gso_max_segs = max_segs;
2428 			status |= DO_SETLINK_MODIFIED;
2429 		}
2430 	}
2431 
2432 	if (tb[IFLA_OPERSTATE])
2433 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2434 
2435 	if (tb[IFLA_LINKMODE]) {
2436 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2437 
2438 		write_lock_bh(&dev_base_lock);
2439 		if (dev->link_mode ^ value)
2440 			status |= DO_SETLINK_NOTIFY;
2441 		dev->link_mode = value;
2442 		write_unlock_bh(&dev_base_lock);
2443 	}
2444 
2445 	if (tb[IFLA_VFINFO_LIST]) {
2446 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2447 		struct nlattr *attr;
2448 		int rem;
2449 
2450 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2451 			if (nla_type(attr) != IFLA_VF_INFO ||
2452 			    nla_len(attr) < NLA_HDRLEN) {
2453 				err = -EINVAL;
2454 				goto errout;
2455 			}
2456 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2457 					       ifla_vf_policy, NULL);
2458 			if (err < 0)
2459 				goto errout;
2460 			err = do_setvfinfo(dev, vfinfo);
2461 			if (err < 0)
2462 				goto errout;
2463 			status |= DO_SETLINK_NOTIFY;
2464 		}
2465 	}
2466 	err = 0;
2467 
2468 	if (tb[IFLA_VF_PORTS]) {
2469 		struct nlattr *port[IFLA_PORT_MAX+1];
2470 		struct nlattr *attr;
2471 		int vf;
2472 		int rem;
2473 
2474 		err = -EOPNOTSUPP;
2475 		if (!ops->ndo_set_vf_port)
2476 			goto errout;
2477 
2478 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2479 			if (nla_type(attr) != IFLA_VF_PORT ||
2480 			    nla_len(attr) < NLA_HDRLEN) {
2481 				err = -EINVAL;
2482 				goto errout;
2483 			}
2484 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2485 					       ifla_port_policy, NULL);
2486 			if (err < 0)
2487 				goto errout;
2488 			if (!port[IFLA_PORT_VF]) {
2489 				err = -EOPNOTSUPP;
2490 				goto errout;
2491 			}
2492 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2493 			err = ops->ndo_set_vf_port(dev, vf, port);
2494 			if (err < 0)
2495 				goto errout;
2496 			status |= DO_SETLINK_NOTIFY;
2497 		}
2498 	}
2499 	err = 0;
2500 
2501 	if (tb[IFLA_PORT_SELF]) {
2502 		struct nlattr *port[IFLA_PORT_MAX+1];
2503 
2504 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2505 				       tb[IFLA_PORT_SELF], ifla_port_policy,
2506 				       NULL);
2507 		if (err < 0)
2508 			goto errout;
2509 
2510 		err = -EOPNOTSUPP;
2511 		if (ops->ndo_set_vf_port)
2512 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2513 		if (err < 0)
2514 			goto errout;
2515 		status |= DO_SETLINK_NOTIFY;
2516 	}
2517 
2518 	if (tb[IFLA_AF_SPEC]) {
2519 		struct nlattr *af;
2520 		int rem;
2521 
2522 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2523 			const struct rtnl_af_ops *af_ops;
2524 
2525 			rcu_read_lock();
2526 
2527 			BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2528 
2529 			err = af_ops->set_link_af(dev, af);
2530 			if (err < 0) {
2531 				rcu_read_unlock();
2532 				goto errout;
2533 			}
2534 
2535 			rcu_read_unlock();
2536 			status |= DO_SETLINK_NOTIFY;
2537 		}
2538 	}
2539 	err = 0;
2540 
2541 	if (tb[IFLA_PROTO_DOWN]) {
2542 		err = dev_change_proto_down(dev,
2543 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2544 		if (err)
2545 			goto errout;
2546 		status |= DO_SETLINK_NOTIFY;
2547 	}
2548 
2549 	if (tb[IFLA_XDP]) {
2550 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2551 		u32 xdp_flags = 0;
2552 
2553 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2554 				       ifla_xdp_policy, NULL);
2555 		if (err < 0)
2556 			goto errout;
2557 
2558 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2559 			err = -EINVAL;
2560 			goto errout;
2561 		}
2562 
2563 		if (xdp[IFLA_XDP_FLAGS]) {
2564 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2565 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2566 				err = -EINVAL;
2567 				goto errout;
2568 			}
2569 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2570 				err = -EINVAL;
2571 				goto errout;
2572 			}
2573 		}
2574 
2575 		if (xdp[IFLA_XDP_FD]) {
2576 			err = dev_change_xdp_fd(dev, extack,
2577 						nla_get_s32(xdp[IFLA_XDP_FD]),
2578 						xdp_flags);
2579 			if (err)
2580 				goto errout;
2581 			status |= DO_SETLINK_NOTIFY;
2582 		}
2583 	}
2584 
2585 errout:
2586 	if (status & DO_SETLINK_MODIFIED) {
2587 		if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2588 			netdev_state_change(dev);
2589 
2590 		if (err < 0)
2591 			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",
2592 					     dev->name);
2593 	}
2594 
2595 	return err;
2596 }
2597 
2598 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2599 			struct netlink_ext_ack *extack)
2600 {
2601 	struct net *net = sock_net(skb->sk);
2602 	struct ifinfomsg *ifm;
2603 	struct net_device *dev;
2604 	int err;
2605 	struct nlattr *tb[IFLA_MAX+1];
2606 	char ifname[IFNAMSIZ];
2607 
2608 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2609 			  extack);
2610 	if (err < 0)
2611 		goto errout;
2612 
2613 	err = rtnl_ensure_unique_netns(tb, extack, false);
2614 	if (err < 0)
2615 		goto errout;
2616 
2617 	if (tb[IFLA_IFNAME])
2618 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2619 	else
2620 		ifname[0] = '\0';
2621 
2622 	err = -EINVAL;
2623 	ifm = nlmsg_data(nlh);
2624 	if (ifm->ifi_index > 0)
2625 		dev = __dev_get_by_index(net, ifm->ifi_index);
2626 	else if (tb[IFLA_IFNAME])
2627 		dev = __dev_get_by_name(net, ifname);
2628 	else
2629 		goto errout;
2630 
2631 	if (dev == NULL) {
2632 		err = -ENODEV;
2633 		goto errout;
2634 	}
2635 
2636 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2637 errout:
2638 	return err;
2639 }
2640 
2641 static int rtnl_group_dellink(const struct net *net, int group)
2642 {
2643 	struct net_device *dev, *aux;
2644 	LIST_HEAD(list_kill);
2645 	bool found = false;
2646 
2647 	if (!group)
2648 		return -EPERM;
2649 
2650 	for_each_netdev(net, dev) {
2651 		if (dev->group == group) {
2652 			const struct rtnl_link_ops *ops;
2653 
2654 			found = true;
2655 			ops = dev->rtnl_link_ops;
2656 			if (!ops || !ops->dellink)
2657 				return -EOPNOTSUPP;
2658 		}
2659 	}
2660 
2661 	if (!found)
2662 		return -ENODEV;
2663 
2664 	for_each_netdev_safe(net, dev, aux) {
2665 		if (dev->group == group) {
2666 			const struct rtnl_link_ops *ops;
2667 
2668 			ops = dev->rtnl_link_ops;
2669 			ops->dellink(dev, &list_kill);
2670 		}
2671 	}
2672 	unregister_netdevice_many(&list_kill);
2673 
2674 	return 0;
2675 }
2676 
2677 int rtnl_delete_link(struct net_device *dev)
2678 {
2679 	const struct rtnl_link_ops *ops;
2680 	LIST_HEAD(list_kill);
2681 
2682 	ops = dev->rtnl_link_ops;
2683 	if (!ops || !ops->dellink)
2684 		return -EOPNOTSUPP;
2685 
2686 	ops->dellink(dev, &list_kill);
2687 	unregister_netdevice_many(&list_kill);
2688 
2689 	return 0;
2690 }
2691 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2692 
2693 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2694 			struct netlink_ext_ack *extack)
2695 {
2696 	struct net *net = sock_net(skb->sk);
2697 	struct net *tgt_net = net;
2698 	struct net_device *dev = NULL;
2699 	struct ifinfomsg *ifm;
2700 	char ifname[IFNAMSIZ];
2701 	struct nlattr *tb[IFLA_MAX+1];
2702 	int err;
2703 	int netnsid = -1;
2704 
2705 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2706 	if (err < 0)
2707 		return err;
2708 
2709 	err = rtnl_ensure_unique_netns(tb, extack, true);
2710 	if (err < 0)
2711 		return err;
2712 
2713 	if (tb[IFLA_IFNAME])
2714 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2715 
2716 	if (tb[IFLA_IF_NETNSID]) {
2717 		netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2718 		tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2719 		if (IS_ERR(tgt_net))
2720 			return PTR_ERR(tgt_net);
2721 	}
2722 
2723 	err = -EINVAL;
2724 	ifm = nlmsg_data(nlh);
2725 	if (ifm->ifi_index > 0)
2726 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2727 	else if (tb[IFLA_IFNAME])
2728 		dev = __dev_get_by_name(tgt_net, ifname);
2729 	else if (tb[IFLA_GROUP])
2730 		err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2731 	else
2732 		goto out;
2733 
2734 	if (!dev) {
2735 		if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2736 			err = -ENODEV;
2737 
2738 		goto out;
2739 	}
2740 
2741 	err = rtnl_delete_link(dev);
2742 
2743 out:
2744 	if (netnsid >= 0)
2745 		put_net(tgt_net);
2746 
2747 	return err;
2748 }
2749 
2750 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2751 {
2752 	unsigned int old_flags;
2753 	int err;
2754 
2755 	old_flags = dev->flags;
2756 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2757 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2758 		if (err < 0)
2759 			return err;
2760 	}
2761 
2762 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2763 
2764 	__dev_notify_flags(dev, old_flags, ~0U);
2765 	return 0;
2766 }
2767 EXPORT_SYMBOL(rtnl_configure_link);
2768 
2769 struct net_device *rtnl_create_link(struct net *net,
2770 	const char *ifname, unsigned char name_assign_type,
2771 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2772 {
2773 	struct net_device *dev;
2774 	unsigned int num_tx_queues = 1;
2775 	unsigned int num_rx_queues = 1;
2776 
2777 	if (tb[IFLA_NUM_TX_QUEUES])
2778 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2779 	else if (ops->get_num_tx_queues)
2780 		num_tx_queues = ops->get_num_tx_queues();
2781 
2782 	if (tb[IFLA_NUM_RX_QUEUES])
2783 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2784 	else if (ops->get_num_rx_queues)
2785 		num_rx_queues = ops->get_num_rx_queues();
2786 
2787 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2788 			       ops->setup, num_tx_queues, num_rx_queues);
2789 	if (!dev)
2790 		return ERR_PTR(-ENOMEM);
2791 
2792 	dev_net_set(dev, net);
2793 	dev->rtnl_link_ops = ops;
2794 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2795 
2796 	if (tb[IFLA_MTU])
2797 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2798 	if (tb[IFLA_ADDRESS]) {
2799 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2800 				nla_len(tb[IFLA_ADDRESS]));
2801 		dev->addr_assign_type = NET_ADDR_SET;
2802 	}
2803 	if (tb[IFLA_BROADCAST])
2804 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2805 				nla_len(tb[IFLA_BROADCAST]));
2806 	if (tb[IFLA_TXQLEN])
2807 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2808 	if (tb[IFLA_OPERSTATE])
2809 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2810 	if (tb[IFLA_LINKMODE])
2811 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2812 	if (tb[IFLA_GROUP])
2813 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2814 	if (tb[IFLA_GSO_MAX_SIZE])
2815 		netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2816 	if (tb[IFLA_GSO_MAX_SEGS])
2817 		dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2818 
2819 	return dev;
2820 }
2821 EXPORT_SYMBOL(rtnl_create_link);
2822 
2823 static int rtnl_group_changelink(const struct sk_buff *skb,
2824 		struct net *net, int group,
2825 		struct ifinfomsg *ifm,
2826 		struct netlink_ext_ack *extack,
2827 		struct nlattr **tb)
2828 {
2829 	struct net_device *dev, *aux;
2830 	int err;
2831 
2832 	for_each_netdev_safe(net, dev, aux) {
2833 		if (dev->group == group) {
2834 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2835 			if (err < 0)
2836 				return err;
2837 		}
2838 	}
2839 
2840 	return 0;
2841 }
2842 
2843 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2844 			struct netlink_ext_ack *extack)
2845 {
2846 	struct net *net = sock_net(skb->sk);
2847 	const struct rtnl_link_ops *ops;
2848 	const struct rtnl_link_ops *m_ops = NULL;
2849 	struct net_device *dev;
2850 	struct net_device *master_dev = NULL;
2851 	struct ifinfomsg *ifm;
2852 	char kind[MODULE_NAME_LEN];
2853 	char ifname[IFNAMSIZ];
2854 	struct nlattr *tb[IFLA_MAX+1];
2855 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2856 	unsigned char name_assign_type = NET_NAME_USER;
2857 	int err;
2858 
2859 #ifdef CONFIG_MODULES
2860 replay:
2861 #endif
2862 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2863 	if (err < 0)
2864 		return err;
2865 
2866 	err = rtnl_ensure_unique_netns(tb, extack, false);
2867 	if (err < 0)
2868 		return err;
2869 
2870 	if (tb[IFLA_IFNAME])
2871 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2872 	else
2873 		ifname[0] = '\0';
2874 
2875 	ifm = nlmsg_data(nlh);
2876 	if (ifm->ifi_index > 0)
2877 		dev = __dev_get_by_index(net, ifm->ifi_index);
2878 	else {
2879 		if (ifname[0])
2880 			dev = __dev_get_by_name(net, ifname);
2881 		else
2882 			dev = NULL;
2883 	}
2884 
2885 	if (dev) {
2886 		master_dev = netdev_master_upper_dev_get(dev);
2887 		if (master_dev)
2888 			m_ops = master_dev->rtnl_link_ops;
2889 	}
2890 
2891 	err = validate_linkmsg(dev, tb);
2892 	if (err < 0)
2893 		return err;
2894 
2895 	if (tb[IFLA_LINKINFO]) {
2896 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2897 				       tb[IFLA_LINKINFO], ifla_info_policy,
2898 				       NULL);
2899 		if (err < 0)
2900 			return err;
2901 	} else
2902 		memset(linkinfo, 0, sizeof(linkinfo));
2903 
2904 	if (linkinfo[IFLA_INFO_KIND]) {
2905 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2906 		ops = rtnl_link_ops_get(kind);
2907 	} else {
2908 		kind[0] = '\0';
2909 		ops = NULL;
2910 	}
2911 
2912 	if (1) {
2913 		struct nlattr *attr[RTNL_MAX_TYPE + 1];
2914 		struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
2915 		struct nlattr **data = NULL;
2916 		struct nlattr **slave_data = NULL;
2917 		struct net *dest_net, *link_net = NULL;
2918 
2919 		if (ops) {
2920 			if (ops->maxtype > RTNL_MAX_TYPE)
2921 				return -EINVAL;
2922 
2923 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2924 				err = nla_parse_nested(attr, ops->maxtype,
2925 						       linkinfo[IFLA_INFO_DATA],
2926 						       ops->policy, NULL);
2927 				if (err < 0)
2928 					return err;
2929 				data = attr;
2930 			}
2931 			if (ops->validate) {
2932 				err = ops->validate(tb, data, extack);
2933 				if (err < 0)
2934 					return err;
2935 			}
2936 		}
2937 
2938 		if (m_ops) {
2939 			if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
2940 				return -EINVAL;
2941 
2942 			if (m_ops->slave_maxtype &&
2943 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2944 				err = nla_parse_nested(slave_attr,
2945 						       m_ops->slave_maxtype,
2946 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2947 						       m_ops->slave_policy,
2948 						       NULL);
2949 				if (err < 0)
2950 					return err;
2951 				slave_data = slave_attr;
2952 			}
2953 		}
2954 
2955 		if (dev) {
2956 			int status = 0;
2957 
2958 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2959 				return -EEXIST;
2960 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2961 				return -EOPNOTSUPP;
2962 
2963 			if (linkinfo[IFLA_INFO_DATA]) {
2964 				if (!ops || ops != dev->rtnl_link_ops ||
2965 				    !ops->changelink)
2966 					return -EOPNOTSUPP;
2967 
2968 				err = ops->changelink(dev, tb, data, extack);
2969 				if (err < 0)
2970 					return err;
2971 				status |= DO_SETLINK_NOTIFY;
2972 			}
2973 
2974 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2975 				if (!m_ops || !m_ops->slave_changelink)
2976 					return -EOPNOTSUPP;
2977 
2978 				err = m_ops->slave_changelink(master_dev, dev,
2979 							      tb, slave_data,
2980 							      extack);
2981 				if (err < 0)
2982 					return err;
2983 				status |= DO_SETLINK_NOTIFY;
2984 			}
2985 
2986 			return do_setlink(skb, dev, ifm, extack, tb, ifname,
2987 					  status);
2988 		}
2989 
2990 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2991 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2992 				return rtnl_group_changelink(skb, net,
2993 						nla_get_u32(tb[IFLA_GROUP]),
2994 						ifm, extack, tb);
2995 			return -ENODEV;
2996 		}
2997 
2998 		if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2999 			return -EOPNOTSUPP;
3000 
3001 		if (!ops) {
3002 #ifdef CONFIG_MODULES
3003 			if (kind[0]) {
3004 				__rtnl_unlock();
3005 				request_module("rtnl-link-%s", kind);
3006 				rtnl_lock();
3007 				ops = rtnl_link_ops_get(kind);
3008 				if (ops)
3009 					goto replay;
3010 			}
3011 #endif
3012 			return -EOPNOTSUPP;
3013 		}
3014 
3015 		if (!ops->setup)
3016 			return -EOPNOTSUPP;
3017 
3018 		if (!ifname[0]) {
3019 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3020 			name_assign_type = NET_NAME_ENUM;
3021 		}
3022 
3023 		dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3024 		if (IS_ERR(dest_net))
3025 			return PTR_ERR(dest_net);
3026 
3027 		if (tb[IFLA_LINK_NETNSID]) {
3028 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3029 
3030 			link_net = get_net_ns_by_id(dest_net, id);
3031 			if (!link_net) {
3032 				err =  -EINVAL;
3033 				goto out;
3034 			}
3035 			err = -EPERM;
3036 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3037 				goto out;
3038 		}
3039 
3040 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
3041 				       name_assign_type, ops, tb);
3042 		if (IS_ERR(dev)) {
3043 			err = PTR_ERR(dev);
3044 			goto out;
3045 		}
3046 
3047 		dev->ifindex = ifm->ifi_index;
3048 
3049 		if (ops->newlink) {
3050 			err = ops->newlink(link_net ? : net, dev, tb, data,
3051 					   extack);
3052 			/* Drivers should call free_netdev() in ->destructor
3053 			 * and unregister it on failure after registration
3054 			 * so that device could be finally freed in rtnl_unlock.
3055 			 */
3056 			if (err < 0) {
3057 				/* If device is not registered at all, free it now */
3058 				if (dev->reg_state == NETREG_UNINITIALIZED)
3059 					free_netdev(dev);
3060 				goto out;
3061 			}
3062 		} else {
3063 			err = register_netdevice(dev);
3064 			if (err < 0) {
3065 				free_netdev(dev);
3066 				goto out;
3067 			}
3068 		}
3069 		err = rtnl_configure_link(dev, ifm);
3070 		if (err < 0)
3071 			goto out_unregister;
3072 		if (link_net) {
3073 			err = dev_change_net_namespace(dev, dest_net, ifname);
3074 			if (err < 0)
3075 				goto out_unregister;
3076 		}
3077 		if (tb[IFLA_MASTER]) {
3078 			err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
3079 					    extack);
3080 			if (err)
3081 				goto out_unregister;
3082 		}
3083 out:
3084 		if (link_net)
3085 			put_net(link_net);
3086 		put_net(dest_net);
3087 		return err;
3088 out_unregister:
3089 		if (ops->newlink) {
3090 			LIST_HEAD(list_kill);
3091 
3092 			ops->dellink(dev, &list_kill);
3093 			unregister_netdevice_many(&list_kill);
3094 		} else {
3095 			unregister_netdevice(dev);
3096 		}
3097 		goto out;
3098 	}
3099 }
3100 
3101 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3102 			struct netlink_ext_ack *extack)
3103 {
3104 	struct net *net = sock_net(skb->sk);
3105 	struct net *tgt_net = net;
3106 	struct ifinfomsg *ifm;
3107 	char ifname[IFNAMSIZ];
3108 	struct nlattr *tb[IFLA_MAX+1];
3109 	struct net_device *dev = NULL;
3110 	struct sk_buff *nskb;
3111 	int netnsid = -1;
3112 	int err;
3113 	u32 ext_filter_mask = 0;
3114 
3115 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3116 	if (err < 0)
3117 		return err;
3118 
3119 	err = rtnl_ensure_unique_netns(tb, extack, true);
3120 	if (err < 0)
3121 		return err;
3122 
3123 	if (tb[IFLA_IF_NETNSID]) {
3124 		netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
3125 		tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
3126 		if (IS_ERR(tgt_net))
3127 			return PTR_ERR(tgt_net);
3128 	}
3129 
3130 	if (tb[IFLA_IFNAME])
3131 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3132 
3133 	if (tb[IFLA_EXT_MASK])
3134 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3135 
3136 	err = -EINVAL;
3137 	ifm = nlmsg_data(nlh);
3138 	if (ifm->ifi_index > 0)
3139 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3140 	else if (tb[IFLA_IFNAME])
3141 		dev = __dev_get_by_name(tgt_net, ifname);
3142 	else
3143 		goto out;
3144 
3145 	err = -ENODEV;
3146 	if (dev == NULL)
3147 		goto out;
3148 
3149 	err = -ENOBUFS;
3150 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3151 	if (nskb == NULL)
3152 		goto out;
3153 
3154 	err = rtnl_fill_ifinfo(nskb, dev, net,
3155 			       RTM_NEWLINK, NETLINK_CB(skb).portid,
3156 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3157 			       0, NULL, 0, netnsid);
3158 	if (err < 0) {
3159 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
3160 		WARN_ON(err == -EMSGSIZE);
3161 		kfree_skb(nskb);
3162 	} else
3163 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3164 out:
3165 	if (netnsid >= 0)
3166 		put_net(tgt_net);
3167 
3168 	return err;
3169 }
3170 
3171 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3172 {
3173 	struct net *net = sock_net(skb->sk);
3174 	struct net_device *dev;
3175 	struct nlattr *tb[IFLA_MAX+1];
3176 	u32 ext_filter_mask = 0;
3177 	u16 min_ifinfo_dump_size = 0;
3178 	int hdrlen;
3179 
3180 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3181 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3182 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3183 
3184 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3185 		if (tb[IFLA_EXT_MASK])
3186 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3187 	}
3188 
3189 	if (!ext_filter_mask)
3190 		return NLMSG_GOODSIZE;
3191 	/*
3192 	 * traverse the list of net devices and compute the minimum
3193 	 * buffer size based upon the filter mask.
3194 	 */
3195 	rcu_read_lock();
3196 	for_each_netdev_rcu(net, dev) {
3197 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3198 					     if_nlmsg_size(dev,
3199 						           ext_filter_mask));
3200 	}
3201 	rcu_read_unlock();
3202 
3203 	return nlmsg_total_size(min_ifinfo_dump_size);
3204 }
3205 
3206 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3207 {
3208 	int idx;
3209 	int s_idx = cb->family;
3210 
3211 	if (s_idx == 0)
3212 		s_idx = 1;
3213 
3214 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3215 		struct rtnl_link **tab;
3216 		int type = cb->nlh->nlmsg_type-RTM_BASE;
3217 		struct rtnl_link *link;
3218 		rtnl_dumpit_func dumpit;
3219 
3220 		if (idx < s_idx || idx == PF_PACKET)
3221 			continue;
3222 
3223 		if (type < 0 || type >= RTM_NR_MSGTYPES)
3224 			continue;
3225 
3226 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3227 		if (!tab)
3228 			continue;
3229 
3230 		link = tab[type];
3231 		if (!link)
3232 			continue;
3233 
3234 		dumpit = link->dumpit;
3235 		if (!dumpit)
3236 			continue;
3237 
3238 		if (idx > s_idx) {
3239 			memset(&cb->args[0], 0, sizeof(cb->args));
3240 			cb->prev_seq = 0;
3241 			cb->seq = 0;
3242 		}
3243 		if (dumpit(skb, cb))
3244 			break;
3245 	}
3246 	cb->family = idx;
3247 
3248 	return skb->len;
3249 }
3250 
3251 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3252 				       unsigned int change,
3253 				       u32 event, gfp_t flags, int *new_nsid,
3254 				       int new_ifindex)
3255 {
3256 	struct net *net = dev_net(dev);
3257 	struct sk_buff *skb;
3258 	int err = -ENOBUFS;
3259 	size_t if_info_size;
3260 
3261 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3262 	if (skb == NULL)
3263 		goto errout;
3264 
3265 	err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3266 			       type, 0, 0, change, 0, 0, event,
3267 			       new_nsid, new_ifindex, -1);
3268 	if (err < 0) {
3269 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
3270 		WARN_ON(err == -EMSGSIZE);
3271 		kfree_skb(skb);
3272 		goto errout;
3273 	}
3274 	return skb;
3275 errout:
3276 	if (err < 0)
3277 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3278 	return NULL;
3279 }
3280 
3281 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3282 {
3283 	struct net *net = dev_net(dev);
3284 
3285 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3286 }
3287 
3288 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3289 			       unsigned int change, u32 event,
3290 			       gfp_t flags, int *new_nsid, int new_ifindex)
3291 {
3292 	struct sk_buff *skb;
3293 
3294 	if (dev->reg_state != NETREG_REGISTERED)
3295 		return;
3296 
3297 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3298 				     new_ifindex);
3299 	if (skb)
3300 		rtmsg_ifinfo_send(skb, dev, flags);
3301 }
3302 
3303 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3304 		  gfp_t flags)
3305 {
3306 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3307 			   NULL, 0);
3308 }
3309 
3310 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3311 			 gfp_t flags, int *new_nsid, int new_ifindex)
3312 {
3313 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3314 			   new_nsid, new_ifindex);
3315 }
3316 
3317 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3318 				   struct net_device *dev,
3319 				   u8 *addr, u16 vid, u32 pid, u32 seq,
3320 				   int type, unsigned int flags,
3321 				   int nlflags, u16 ndm_state)
3322 {
3323 	struct nlmsghdr *nlh;
3324 	struct ndmsg *ndm;
3325 
3326 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3327 	if (!nlh)
3328 		return -EMSGSIZE;
3329 
3330 	ndm = nlmsg_data(nlh);
3331 	ndm->ndm_family  = AF_BRIDGE;
3332 	ndm->ndm_pad1	 = 0;
3333 	ndm->ndm_pad2    = 0;
3334 	ndm->ndm_flags	 = flags;
3335 	ndm->ndm_type	 = 0;
3336 	ndm->ndm_ifindex = dev->ifindex;
3337 	ndm->ndm_state   = ndm_state;
3338 
3339 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3340 		goto nla_put_failure;
3341 	if (vid)
3342 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3343 			goto nla_put_failure;
3344 
3345 	nlmsg_end(skb, nlh);
3346 	return 0;
3347 
3348 nla_put_failure:
3349 	nlmsg_cancel(skb, nlh);
3350 	return -EMSGSIZE;
3351 }
3352 
3353 static inline size_t rtnl_fdb_nlmsg_size(void)
3354 {
3355 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3356 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
3357 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
3358 	       0;
3359 }
3360 
3361 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3362 			    u16 ndm_state)
3363 {
3364 	struct net *net = dev_net(dev);
3365 	struct sk_buff *skb;
3366 	int err = -ENOBUFS;
3367 
3368 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3369 	if (!skb)
3370 		goto errout;
3371 
3372 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3373 				      0, 0, type, NTF_SELF, 0, ndm_state);
3374 	if (err < 0) {
3375 		kfree_skb(skb);
3376 		goto errout;
3377 	}
3378 
3379 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3380 	return;
3381 errout:
3382 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3383 }
3384 
3385 /**
3386  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3387  */
3388 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3389 		     struct nlattr *tb[],
3390 		     struct net_device *dev,
3391 		     const unsigned char *addr, u16 vid,
3392 		     u16 flags)
3393 {
3394 	int err = -EINVAL;
3395 
3396 	/* If aging addresses are supported device will need to
3397 	 * implement its own handler for this.
3398 	 */
3399 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3400 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3401 		return err;
3402 	}
3403 
3404 	if (vid) {
3405 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3406 		return err;
3407 	}
3408 
3409 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3410 		err = dev_uc_add_excl(dev, addr);
3411 	else if (is_multicast_ether_addr(addr))
3412 		err = dev_mc_add_excl(dev, addr);
3413 
3414 	/* Only return duplicate errors if NLM_F_EXCL is set */
3415 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3416 		err = 0;
3417 
3418 	return err;
3419 }
3420 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3421 
3422 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3423 			 struct netlink_ext_ack *extack)
3424 {
3425 	u16 vid = 0;
3426 
3427 	if (vlan_attr) {
3428 		if (nla_len(vlan_attr) != sizeof(u16)) {
3429 			NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3430 			return -EINVAL;
3431 		}
3432 
3433 		vid = nla_get_u16(vlan_attr);
3434 
3435 		if (!vid || vid >= VLAN_VID_MASK) {
3436 			NL_SET_ERR_MSG(extack, "invalid vlan id");
3437 			return -EINVAL;
3438 		}
3439 	}
3440 	*p_vid = vid;
3441 	return 0;
3442 }
3443 
3444 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3445 			struct netlink_ext_ack *extack)
3446 {
3447 	struct net *net = sock_net(skb->sk);
3448 	struct ndmsg *ndm;
3449 	struct nlattr *tb[NDA_MAX+1];
3450 	struct net_device *dev;
3451 	u8 *addr;
3452 	u16 vid;
3453 	int err;
3454 
3455 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3456 	if (err < 0)
3457 		return err;
3458 
3459 	ndm = nlmsg_data(nlh);
3460 	if (ndm->ndm_ifindex == 0) {
3461 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3462 		return -EINVAL;
3463 	}
3464 
3465 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3466 	if (dev == NULL) {
3467 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3468 		return -ENODEV;
3469 	}
3470 
3471 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3472 		NL_SET_ERR_MSG(extack, "invalid address");
3473 		return -EINVAL;
3474 	}
3475 
3476 	addr = nla_data(tb[NDA_LLADDR]);
3477 
3478 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3479 	if (err)
3480 		return err;
3481 
3482 	err = -EOPNOTSUPP;
3483 
3484 	/* Support fdb on master device the net/bridge default case */
3485 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3486 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3487 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3488 		const struct net_device_ops *ops = br_dev->netdev_ops;
3489 
3490 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3491 				       nlh->nlmsg_flags);
3492 		if (err)
3493 			goto out;
3494 		else
3495 			ndm->ndm_flags &= ~NTF_MASTER;
3496 	}
3497 
3498 	/* Embedded bridge, macvlan, and any other device support */
3499 	if ((ndm->ndm_flags & NTF_SELF)) {
3500 		if (dev->netdev_ops->ndo_fdb_add)
3501 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3502 							   vid,
3503 							   nlh->nlmsg_flags);
3504 		else
3505 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3506 					       nlh->nlmsg_flags);
3507 
3508 		if (!err) {
3509 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3510 					ndm->ndm_state);
3511 			ndm->ndm_flags &= ~NTF_SELF;
3512 		}
3513 	}
3514 out:
3515 	return err;
3516 }
3517 
3518 /**
3519  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3520  */
3521 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3522 		     struct nlattr *tb[],
3523 		     struct net_device *dev,
3524 		     const unsigned char *addr, u16 vid)
3525 {
3526 	int err = -EINVAL;
3527 
3528 	/* If aging addresses are supported device will need to
3529 	 * implement its own handler for this.
3530 	 */
3531 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3532 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3533 		return err;
3534 	}
3535 
3536 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3537 		err = dev_uc_del(dev, addr);
3538 	else if (is_multicast_ether_addr(addr))
3539 		err = dev_mc_del(dev, addr);
3540 
3541 	return err;
3542 }
3543 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3544 
3545 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3546 			struct netlink_ext_ack *extack)
3547 {
3548 	struct net *net = sock_net(skb->sk);
3549 	struct ndmsg *ndm;
3550 	struct nlattr *tb[NDA_MAX+1];
3551 	struct net_device *dev;
3552 	int err = -EINVAL;
3553 	__u8 *addr;
3554 	u16 vid;
3555 
3556 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3557 		return -EPERM;
3558 
3559 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3560 	if (err < 0)
3561 		return err;
3562 
3563 	ndm = nlmsg_data(nlh);
3564 	if (ndm->ndm_ifindex == 0) {
3565 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3566 		return -EINVAL;
3567 	}
3568 
3569 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3570 	if (dev == NULL) {
3571 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3572 		return -ENODEV;
3573 	}
3574 
3575 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3576 		NL_SET_ERR_MSG(extack, "invalid address");
3577 		return -EINVAL;
3578 	}
3579 
3580 	addr = nla_data(tb[NDA_LLADDR]);
3581 
3582 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3583 	if (err)
3584 		return err;
3585 
3586 	err = -EOPNOTSUPP;
3587 
3588 	/* Support fdb on master device the net/bridge default case */
3589 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3590 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3591 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3592 		const struct net_device_ops *ops = br_dev->netdev_ops;
3593 
3594 		if (ops->ndo_fdb_del)
3595 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3596 
3597 		if (err)
3598 			goto out;
3599 		else
3600 			ndm->ndm_flags &= ~NTF_MASTER;
3601 	}
3602 
3603 	/* Embedded bridge, macvlan, and any other device support */
3604 	if (ndm->ndm_flags & NTF_SELF) {
3605 		if (dev->netdev_ops->ndo_fdb_del)
3606 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3607 							   vid);
3608 		else
3609 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3610 
3611 		if (!err) {
3612 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3613 					ndm->ndm_state);
3614 			ndm->ndm_flags &= ~NTF_SELF;
3615 		}
3616 	}
3617 out:
3618 	return err;
3619 }
3620 
3621 static int nlmsg_populate_fdb(struct sk_buff *skb,
3622 			      struct netlink_callback *cb,
3623 			      struct net_device *dev,
3624 			      int *idx,
3625 			      struct netdev_hw_addr_list *list)
3626 {
3627 	struct netdev_hw_addr *ha;
3628 	int err;
3629 	u32 portid, seq;
3630 
3631 	portid = NETLINK_CB(cb->skb).portid;
3632 	seq = cb->nlh->nlmsg_seq;
3633 
3634 	list_for_each_entry(ha, &list->list, list) {
3635 		if (*idx < cb->args[2])
3636 			goto skip;
3637 
3638 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3639 					      portid, seq,
3640 					      RTM_NEWNEIGH, NTF_SELF,
3641 					      NLM_F_MULTI, NUD_PERMANENT);
3642 		if (err < 0)
3643 			return err;
3644 skip:
3645 		*idx += 1;
3646 	}
3647 	return 0;
3648 }
3649 
3650 /**
3651  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3652  * @nlh: netlink message header
3653  * @dev: netdevice
3654  *
3655  * Default netdevice operation to dump the existing unicast address list.
3656  * Returns number of addresses from list put in skb.
3657  */
3658 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3659 		      struct netlink_callback *cb,
3660 		      struct net_device *dev,
3661 		      struct net_device *filter_dev,
3662 		      int *idx)
3663 {
3664 	int err;
3665 
3666 	netif_addr_lock_bh(dev);
3667 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3668 	if (err)
3669 		goto out;
3670 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3671 out:
3672 	netif_addr_unlock_bh(dev);
3673 	return err;
3674 }
3675 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3676 
3677 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3678 {
3679 	struct net_device *dev;
3680 	struct nlattr *tb[IFLA_MAX+1];
3681 	struct net_device *br_dev = NULL;
3682 	const struct net_device_ops *ops = NULL;
3683 	const struct net_device_ops *cops = NULL;
3684 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3685 	struct net *net = sock_net(skb->sk);
3686 	struct hlist_head *head;
3687 	int brport_idx = 0;
3688 	int br_idx = 0;
3689 	int h, s_h;
3690 	int idx = 0, s_idx;
3691 	int err = 0;
3692 	int fidx = 0;
3693 
3694 	err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3695 			  IFLA_MAX, ifla_policy, NULL);
3696 	if (err < 0) {
3697 		return -EINVAL;
3698 	} else if (err == 0) {
3699 		if (tb[IFLA_MASTER])
3700 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3701 	}
3702 
3703 	brport_idx = ifm->ifi_index;
3704 
3705 	if (br_idx) {
3706 		br_dev = __dev_get_by_index(net, br_idx);
3707 		if (!br_dev)
3708 			return -ENODEV;
3709 
3710 		ops = br_dev->netdev_ops;
3711 	}
3712 
3713 	s_h = cb->args[0];
3714 	s_idx = cb->args[1];
3715 
3716 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3717 		idx = 0;
3718 		head = &net->dev_index_head[h];
3719 		hlist_for_each_entry(dev, head, index_hlist) {
3720 
3721 			if (brport_idx && (dev->ifindex != brport_idx))
3722 				continue;
3723 
3724 			if (!br_idx) { /* user did not specify a specific bridge */
3725 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3726 					br_dev = netdev_master_upper_dev_get(dev);
3727 					cops = br_dev->netdev_ops;
3728 				}
3729 			} else {
3730 				if (dev != br_dev &&
3731 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3732 					continue;
3733 
3734 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3735 				    !(dev->priv_flags & IFF_EBRIDGE))
3736 					continue;
3737 				cops = ops;
3738 			}
3739 
3740 			if (idx < s_idx)
3741 				goto cont;
3742 
3743 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3744 				if (cops && cops->ndo_fdb_dump) {
3745 					err = cops->ndo_fdb_dump(skb, cb,
3746 								br_dev, dev,
3747 								&fidx);
3748 					if (err == -EMSGSIZE)
3749 						goto out;
3750 				}
3751 			}
3752 
3753 			if (dev->netdev_ops->ndo_fdb_dump)
3754 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3755 								    dev, NULL,
3756 								    &fidx);
3757 			else
3758 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3759 							&fidx);
3760 			if (err == -EMSGSIZE)
3761 				goto out;
3762 
3763 			cops = NULL;
3764 
3765 			/* reset fdb offset to 0 for rest of the interfaces */
3766 			cb->args[2] = 0;
3767 			fidx = 0;
3768 cont:
3769 			idx++;
3770 		}
3771 	}
3772 
3773 out:
3774 	cb->args[0] = h;
3775 	cb->args[1] = idx;
3776 	cb->args[2] = fidx;
3777 
3778 	return skb->len;
3779 }
3780 
3781 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3782 			       unsigned int attrnum, unsigned int flag)
3783 {
3784 	if (mask & flag)
3785 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3786 	return 0;
3787 }
3788 
3789 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3790 			    struct net_device *dev, u16 mode,
3791 			    u32 flags, u32 mask, int nlflags,
3792 			    u32 filter_mask,
3793 			    int (*vlan_fill)(struct sk_buff *skb,
3794 					     struct net_device *dev,
3795 					     u32 filter_mask))
3796 {
3797 	struct nlmsghdr *nlh;
3798 	struct ifinfomsg *ifm;
3799 	struct nlattr *br_afspec;
3800 	struct nlattr *protinfo;
3801 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3802 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3803 	int err = 0;
3804 
3805 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3806 	if (nlh == NULL)
3807 		return -EMSGSIZE;
3808 
3809 	ifm = nlmsg_data(nlh);
3810 	ifm->ifi_family = AF_BRIDGE;
3811 	ifm->__ifi_pad = 0;
3812 	ifm->ifi_type = dev->type;
3813 	ifm->ifi_index = dev->ifindex;
3814 	ifm->ifi_flags = dev_get_flags(dev);
3815 	ifm->ifi_change = 0;
3816 
3817 
3818 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3819 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3820 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3821 	    (br_dev &&
3822 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3823 	    (dev->addr_len &&
3824 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3825 	    (dev->ifindex != dev_get_iflink(dev) &&
3826 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3827 		goto nla_put_failure;
3828 
3829 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3830 	if (!br_afspec)
3831 		goto nla_put_failure;
3832 
3833 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3834 		nla_nest_cancel(skb, br_afspec);
3835 		goto nla_put_failure;
3836 	}
3837 
3838 	if (mode != BRIDGE_MODE_UNDEF) {
3839 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3840 			nla_nest_cancel(skb, br_afspec);
3841 			goto nla_put_failure;
3842 		}
3843 	}
3844 	if (vlan_fill) {
3845 		err = vlan_fill(skb, dev, filter_mask);
3846 		if (err) {
3847 			nla_nest_cancel(skb, br_afspec);
3848 			goto nla_put_failure;
3849 		}
3850 	}
3851 	nla_nest_end(skb, br_afspec);
3852 
3853 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3854 	if (!protinfo)
3855 		goto nla_put_failure;
3856 
3857 	if (brport_nla_put_flag(skb, flags, mask,
3858 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3859 	    brport_nla_put_flag(skb, flags, mask,
3860 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3861 	    brport_nla_put_flag(skb, flags, mask,
3862 				IFLA_BRPORT_FAST_LEAVE,
3863 				BR_MULTICAST_FAST_LEAVE) ||
3864 	    brport_nla_put_flag(skb, flags, mask,
3865 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3866 	    brport_nla_put_flag(skb, flags, mask,
3867 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3868 	    brport_nla_put_flag(skb, flags, mask,
3869 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3870 	    brport_nla_put_flag(skb, flags, mask,
3871 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3872 	    brport_nla_put_flag(skb, flags, mask,
3873 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3874 		nla_nest_cancel(skb, protinfo);
3875 		goto nla_put_failure;
3876 	}
3877 
3878 	nla_nest_end(skb, protinfo);
3879 
3880 	nlmsg_end(skb, nlh);
3881 	return 0;
3882 nla_put_failure:
3883 	nlmsg_cancel(skb, nlh);
3884 	return err ? err : -EMSGSIZE;
3885 }
3886 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3887 
3888 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3889 {
3890 	struct net *net = sock_net(skb->sk);
3891 	struct net_device *dev;
3892 	int idx = 0;
3893 	u32 portid = NETLINK_CB(cb->skb).portid;
3894 	u32 seq = cb->nlh->nlmsg_seq;
3895 	u32 filter_mask = 0;
3896 	int err;
3897 
3898 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3899 		struct nlattr *extfilt;
3900 
3901 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3902 					  IFLA_EXT_MASK);
3903 		if (extfilt) {
3904 			if (nla_len(extfilt) < sizeof(filter_mask))
3905 				return -EINVAL;
3906 
3907 			filter_mask = nla_get_u32(extfilt);
3908 		}
3909 	}
3910 
3911 	rcu_read_lock();
3912 	for_each_netdev_rcu(net, dev) {
3913 		const struct net_device_ops *ops = dev->netdev_ops;
3914 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3915 
3916 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3917 			if (idx >= cb->args[0]) {
3918 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3919 						skb, portid, seq, dev,
3920 						filter_mask, NLM_F_MULTI);
3921 				if (err < 0 && err != -EOPNOTSUPP) {
3922 					if (likely(skb->len))
3923 						break;
3924 
3925 					goto out_err;
3926 				}
3927 			}
3928 			idx++;
3929 		}
3930 
3931 		if (ops->ndo_bridge_getlink) {
3932 			if (idx >= cb->args[0]) {
3933 				err = ops->ndo_bridge_getlink(skb, portid,
3934 							      seq, dev,
3935 							      filter_mask,
3936 							      NLM_F_MULTI);
3937 				if (err < 0 && err != -EOPNOTSUPP) {
3938 					if (likely(skb->len))
3939 						break;
3940 
3941 					goto out_err;
3942 				}
3943 			}
3944 			idx++;
3945 		}
3946 	}
3947 	err = skb->len;
3948 out_err:
3949 	rcu_read_unlock();
3950 	cb->args[0] = idx;
3951 
3952 	return err;
3953 }
3954 
3955 static inline size_t bridge_nlmsg_size(void)
3956 {
3957 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3958 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3959 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3960 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3961 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3962 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3963 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3964 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3965 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3966 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3967 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3968 }
3969 
3970 static int rtnl_bridge_notify(struct net_device *dev)
3971 {
3972 	struct net *net = dev_net(dev);
3973 	struct sk_buff *skb;
3974 	int err = -EOPNOTSUPP;
3975 
3976 	if (!dev->netdev_ops->ndo_bridge_getlink)
3977 		return 0;
3978 
3979 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3980 	if (!skb) {
3981 		err = -ENOMEM;
3982 		goto errout;
3983 	}
3984 
3985 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3986 	if (err < 0)
3987 		goto errout;
3988 
3989 	if (!skb->len)
3990 		goto errout;
3991 
3992 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3993 	return 0;
3994 errout:
3995 	WARN_ON(err == -EMSGSIZE);
3996 	kfree_skb(skb);
3997 	if (err)
3998 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3999 	return err;
4000 }
4001 
4002 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4003 			       struct netlink_ext_ack *extack)
4004 {
4005 	struct net *net = sock_net(skb->sk);
4006 	struct ifinfomsg *ifm;
4007 	struct net_device *dev;
4008 	struct nlattr *br_spec, *attr = NULL;
4009 	int rem, err = -EOPNOTSUPP;
4010 	u16 flags = 0;
4011 	bool have_flags = false;
4012 
4013 	if (nlmsg_len(nlh) < sizeof(*ifm))
4014 		return -EINVAL;
4015 
4016 	ifm = nlmsg_data(nlh);
4017 	if (ifm->ifi_family != AF_BRIDGE)
4018 		return -EPFNOSUPPORT;
4019 
4020 	dev = __dev_get_by_index(net, ifm->ifi_index);
4021 	if (!dev) {
4022 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4023 		return -ENODEV;
4024 	}
4025 
4026 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4027 	if (br_spec) {
4028 		nla_for_each_nested(attr, br_spec, rem) {
4029 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4030 				if (nla_len(attr) < sizeof(flags))
4031 					return -EINVAL;
4032 
4033 				have_flags = true;
4034 				flags = nla_get_u16(attr);
4035 				break;
4036 			}
4037 		}
4038 	}
4039 
4040 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4041 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4042 
4043 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4044 			err = -EOPNOTSUPP;
4045 			goto out;
4046 		}
4047 
4048 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
4049 		if (err)
4050 			goto out;
4051 
4052 		flags &= ~BRIDGE_FLAGS_MASTER;
4053 	}
4054 
4055 	if ((flags & BRIDGE_FLAGS_SELF)) {
4056 		if (!dev->netdev_ops->ndo_bridge_setlink)
4057 			err = -EOPNOTSUPP;
4058 		else
4059 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4060 								  flags);
4061 		if (!err) {
4062 			flags &= ~BRIDGE_FLAGS_SELF;
4063 
4064 			/* Generate event to notify upper layer of bridge
4065 			 * change
4066 			 */
4067 			err = rtnl_bridge_notify(dev);
4068 		}
4069 	}
4070 
4071 	if (have_flags)
4072 		memcpy(nla_data(attr), &flags, sizeof(flags));
4073 out:
4074 	return err;
4075 }
4076 
4077 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4078 			       struct netlink_ext_ack *extack)
4079 {
4080 	struct net *net = sock_net(skb->sk);
4081 	struct ifinfomsg *ifm;
4082 	struct net_device *dev;
4083 	struct nlattr *br_spec, *attr = NULL;
4084 	int rem, err = -EOPNOTSUPP;
4085 	u16 flags = 0;
4086 	bool have_flags = false;
4087 
4088 	if (nlmsg_len(nlh) < sizeof(*ifm))
4089 		return -EINVAL;
4090 
4091 	ifm = nlmsg_data(nlh);
4092 	if (ifm->ifi_family != AF_BRIDGE)
4093 		return -EPFNOSUPPORT;
4094 
4095 	dev = __dev_get_by_index(net, ifm->ifi_index);
4096 	if (!dev) {
4097 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4098 		return -ENODEV;
4099 	}
4100 
4101 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4102 	if (br_spec) {
4103 		nla_for_each_nested(attr, br_spec, rem) {
4104 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4105 				if (nla_len(attr) < sizeof(flags))
4106 					return -EINVAL;
4107 
4108 				have_flags = true;
4109 				flags = nla_get_u16(attr);
4110 				break;
4111 			}
4112 		}
4113 	}
4114 
4115 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4116 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4117 
4118 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4119 			err = -EOPNOTSUPP;
4120 			goto out;
4121 		}
4122 
4123 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4124 		if (err)
4125 			goto out;
4126 
4127 		flags &= ~BRIDGE_FLAGS_MASTER;
4128 	}
4129 
4130 	if ((flags & BRIDGE_FLAGS_SELF)) {
4131 		if (!dev->netdev_ops->ndo_bridge_dellink)
4132 			err = -EOPNOTSUPP;
4133 		else
4134 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4135 								  flags);
4136 
4137 		if (!err) {
4138 			flags &= ~BRIDGE_FLAGS_SELF;
4139 
4140 			/* Generate event to notify upper layer of bridge
4141 			 * change
4142 			 */
4143 			err = rtnl_bridge_notify(dev);
4144 		}
4145 	}
4146 
4147 	if (have_flags)
4148 		memcpy(nla_data(attr), &flags, sizeof(flags));
4149 out:
4150 	return err;
4151 }
4152 
4153 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4154 {
4155 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4156 	       (!idxattr || idxattr == attrid);
4157 }
4158 
4159 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4160 static int rtnl_get_offload_stats_attr_size(int attr_id)
4161 {
4162 	switch (attr_id) {
4163 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4164 		return sizeof(struct rtnl_link_stats64);
4165 	}
4166 
4167 	return 0;
4168 }
4169 
4170 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4171 				  int *prividx)
4172 {
4173 	struct nlattr *attr = NULL;
4174 	int attr_id, size;
4175 	void *attr_data;
4176 	int err;
4177 
4178 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4179 	      dev->netdev_ops->ndo_get_offload_stats))
4180 		return -ENODATA;
4181 
4182 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4183 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4184 		if (attr_id < *prividx)
4185 			continue;
4186 
4187 		size = rtnl_get_offload_stats_attr_size(attr_id);
4188 		if (!size)
4189 			continue;
4190 
4191 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4192 			continue;
4193 
4194 		attr = nla_reserve_64bit(skb, attr_id, size,
4195 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
4196 		if (!attr)
4197 			goto nla_put_failure;
4198 
4199 		attr_data = nla_data(attr);
4200 		memset(attr_data, 0, size);
4201 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4202 							     attr_data);
4203 		if (err)
4204 			goto get_offload_stats_failure;
4205 	}
4206 
4207 	if (!attr)
4208 		return -ENODATA;
4209 
4210 	*prividx = 0;
4211 	return 0;
4212 
4213 nla_put_failure:
4214 	err = -EMSGSIZE;
4215 get_offload_stats_failure:
4216 	*prividx = attr_id;
4217 	return err;
4218 }
4219 
4220 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4221 {
4222 	int nla_size = 0;
4223 	int attr_id;
4224 	int size;
4225 
4226 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4227 	      dev->netdev_ops->ndo_get_offload_stats))
4228 		return 0;
4229 
4230 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4231 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4232 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4233 			continue;
4234 		size = rtnl_get_offload_stats_attr_size(attr_id);
4235 		nla_size += nla_total_size_64bit(size);
4236 	}
4237 
4238 	if (nla_size != 0)
4239 		nla_size += nla_total_size(0);
4240 
4241 	return nla_size;
4242 }
4243 
4244 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4245 			       int type, u32 pid, u32 seq, u32 change,
4246 			       unsigned int flags, unsigned int filter_mask,
4247 			       int *idxattr, int *prividx)
4248 {
4249 	struct if_stats_msg *ifsm;
4250 	struct nlmsghdr *nlh;
4251 	struct nlattr *attr;
4252 	int s_prividx = *prividx;
4253 	int err;
4254 
4255 	ASSERT_RTNL();
4256 
4257 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4258 	if (!nlh)
4259 		return -EMSGSIZE;
4260 
4261 	ifsm = nlmsg_data(nlh);
4262 	ifsm->family = PF_UNSPEC;
4263 	ifsm->pad1 = 0;
4264 	ifsm->pad2 = 0;
4265 	ifsm->ifindex = dev->ifindex;
4266 	ifsm->filter_mask = filter_mask;
4267 
4268 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4269 		struct rtnl_link_stats64 *sp;
4270 
4271 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4272 					 sizeof(struct rtnl_link_stats64),
4273 					 IFLA_STATS_UNSPEC);
4274 		if (!attr)
4275 			goto nla_put_failure;
4276 
4277 		sp = nla_data(attr);
4278 		dev_get_stats(dev, sp);
4279 	}
4280 
4281 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4282 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4283 
4284 		if (ops && ops->fill_linkxstats) {
4285 			*idxattr = IFLA_STATS_LINK_XSTATS;
4286 			attr = nla_nest_start(skb,
4287 					      IFLA_STATS_LINK_XSTATS);
4288 			if (!attr)
4289 				goto nla_put_failure;
4290 
4291 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4292 			nla_nest_end(skb, attr);
4293 			if (err)
4294 				goto nla_put_failure;
4295 			*idxattr = 0;
4296 		}
4297 	}
4298 
4299 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4300 			     *idxattr)) {
4301 		const struct rtnl_link_ops *ops = NULL;
4302 		const struct net_device *master;
4303 
4304 		master = netdev_master_upper_dev_get(dev);
4305 		if (master)
4306 			ops = master->rtnl_link_ops;
4307 		if (ops && ops->fill_linkxstats) {
4308 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4309 			attr = nla_nest_start(skb,
4310 					      IFLA_STATS_LINK_XSTATS_SLAVE);
4311 			if (!attr)
4312 				goto nla_put_failure;
4313 
4314 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4315 			nla_nest_end(skb, attr);
4316 			if (err)
4317 				goto nla_put_failure;
4318 			*idxattr = 0;
4319 		}
4320 	}
4321 
4322 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4323 			     *idxattr)) {
4324 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4325 		attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4326 		if (!attr)
4327 			goto nla_put_failure;
4328 
4329 		err = rtnl_get_offload_stats(skb, dev, prividx);
4330 		if (err == -ENODATA)
4331 			nla_nest_cancel(skb, attr);
4332 		else
4333 			nla_nest_end(skb, attr);
4334 
4335 		if (err && err != -ENODATA)
4336 			goto nla_put_failure;
4337 		*idxattr = 0;
4338 	}
4339 
4340 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4341 		struct rtnl_af_ops *af_ops;
4342 
4343 		*idxattr = IFLA_STATS_AF_SPEC;
4344 		attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4345 		if (!attr)
4346 			goto nla_put_failure;
4347 
4348 		rcu_read_lock();
4349 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4350 			if (af_ops->fill_stats_af) {
4351 				struct nlattr *af;
4352 				int err;
4353 
4354 				af = nla_nest_start(skb, af_ops->family);
4355 				if (!af) {
4356 					rcu_read_unlock();
4357 					goto nla_put_failure;
4358 				}
4359 				err = af_ops->fill_stats_af(skb, dev);
4360 
4361 				if (err == -ENODATA) {
4362 					nla_nest_cancel(skb, af);
4363 				} else if (err < 0) {
4364 					rcu_read_unlock();
4365 					goto nla_put_failure;
4366 				}
4367 
4368 				nla_nest_end(skb, af);
4369 			}
4370 		}
4371 		rcu_read_unlock();
4372 
4373 		nla_nest_end(skb, attr);
4374 
4375 		*idxattr = 0;
4376 	}
4377 
4378 	nlmsg_end(skb, nlh);
4379 
4380 	return 0;
4381 
4382 nla_put_failure:
4383 	/* not a multi message or no progress mean a real error */
4384 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4385 		nlmsg_cancel(skb, nlh);
4386 	else
4387 		nlmsg_end(skb, nlh);
4388 
4389 	return -EMSGSIZE;
4390 }
4391 
4392 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4393 				  u32 filter_mask)
4394 {
4395 	size_t size = 0;
4396 
4397 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4398 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4399 
4400 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4401 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4402 		int attr = IFLA_STATS_LINK_XSTATS;
4403 
4404 		if (ops && ops->get_linkxstats_size) {
4405 			size += nla_total_size(ops->get_linkxstats_size(dev,
4406 									attr));
4407 			/* for IFLA_STATS_LINK_XSTATS */
4408 			size += nla_total_size(0);
4409 		}
4410 	}
4411 
4412 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4413 		struct net_device *_dev = (struct net_device *)dev;
4414 		const struct rtnl_link_ops *ops = NULL;
4415 		const struct net_device *master;
4416 
4417 		/* netdev_master_upper_dev_get can't take const */
4418 		master = netdev_master_upper_dev_get(_dev);
4419 		if (master)
4420 			ops = master->rtnl_link_ops;
4421 		if (ops && ops->get_linkxstats_size) {
4422 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4423 
4424 			size += nla_total_size(ops->get_linkxstats_size(dev,
4425 									attr));
4426 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
4427 			size += nla_total_size(0);
4428 		}
4429 	}
4430 
4431 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4432 		size += rtnl_get_offload_stats_size(dev);
4433 
4434 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4435 		struct rtnl_af_ops *af_ops;
4436 
4437 		/* for IFLA_STATS_AF_SPEC */
4438 		size += nla_total_size(0);
4439 
4440 		rcu_read_lock();
4441 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4442 			if (af_ops->get_stats_af_size) {
4443 				size += nla_total_size(
4444 					af_ops->get_stats_af_size(dev));
4445 
4446 				/* for AF_* */
4447 				size += nla_total_size(0);
4448 			}
4449 		}
4450 		rcu_read_unlock();
4451 	}
4452 
4453 	return size;
4454 }
4455 
4456 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4457 			  struct netlink_ext_ack *extack)
4458 {
4459 	struct net *net = sock_net(skb->sk);
4460 	struct net_device *dev = NULL;
4461 	int idxattr = 0, prividx = 0;
4462 	struct if_stats_msg *ifsm;
4463 	struct sk_buff *nskb;
4464 	u32 filter_mask;
4465 	int err;
4466 
4467 	if (nlmsg_len(nlh) < sizeof(*ifsm))
4468 		return -EINVAL;
4469 
4470 	ifsm = nlmsg_data(nlh);
4471 	if (ifsm->ifindex > 0)
4472 		dev = __dev_get_by_index(net, ifsm->ifindex);
4473 	else
4474 		return -EINVAL;
4475 
4476 	if (!dev)
4477 		return -ENODEV;
4478 
4479 	filter_mask = ifsm->filter_mask;
4480 	if (!filter_mask)
4481 		return -EINVAL;
4482 
4483 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4484 	if (!nskb)
4485 		return -ENOBUFS;
4486 
4487 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4488 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4489 				  0, filter_mask, &idxattr, &prividx);
4490 	if (err < 0) {
4491 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4492 		WARN_ON(err == -EMSGSIZE);
4493 		kfree_skb(nskb);
4494 	} else {
4495 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4496 	}
4497 
4498 	return err;
4499 }
4500 
4501 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4502 {
4503 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
4504 	struct net *net = sock_net(skb->sk);
4505 	unsigned int flags = NLM_F_MULTI;
4506 	struct if_stats_msg *ifsm;
4507 	struct hlist_head *head;
4508 	struct net_device *dev;
4509 	u32 filter_mask = 0;
4510 	int idx = 0;
4511 
4512 	s_h = cb->args[0];
4513 	s_idx = cb->args[1];
4514 	s_idxattr = cb->args[2];
4515 	s_prividx = cb->args[3];
4516 
4517 	cb->seq = net->dev_base_seq;
4518 
4519 	if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4520 		return -EINVAL;
4521 
4522 	ifsm = nlmsg_data(cb->nlh);
4523 	filter_mask = ifsm->filter_mask;
4524 	if (!filter_mask)
4525 		return -EINVAL;
4526 
4527 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4528 		idx = 0;
4529 		head = &net->dev_index_head[h];
4530 		hlist_for_each_entry(dev, head, index_hlist) {
4531 			if (idx < s_idx)
4532 				goto cont;
4533 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4534 						  NETLINK_CB(cb->skb).portid,
4535 						  cb->nlh->nlmsg_seq, 0,
4536 						  flags, filter_mask,
4537 						  &s_idxattr, &s_prividx);
4538 			/* If we ran out of room on the first message,
4539 			 * we're in trouble
4540 			 */
4541 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4542 
4543 			if (err < 0)
4544 				goto out;
4545 			s_prividx = 0;
4546 			s_idxattr = 0;
4547 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4548 cont:
4549 			idx++;
4550 		}
4551 	}
4552 out:
4553 	cb->args[3] = s_prividx;
4554 	cb->args[2] = s_idxattr;
4555 	cb->args[1] = idx;
4556 	cb->args[0] = h;
4557 
4558 	return skb->len;
4559 }
4560 
4561 /* Process one rtnetlink message. */
4562 
4563 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4564 			     struct netlink_ext_ack *extack)
4565 {
4566 	struct net *net = sock_net(skb->sk);
4567 	struct rtnl_link *link;
4568 	struct module *owner;
4569 	int err = -EOPNOTSUPP;
4570 	rtnl_doit_func doit;
4571 	unsigned int flags;
4572 	int kind;
4573 	int family;
4574 	int type;
4575 
4576 	type = nlh->nlmsg_type;
4577 	if (type > RTM_MAX)
4578 		return -EOPNOTSUPP;
4579 
4580 	type -= RTM_BASE;
4581 
4582 	/* All the messages must have at least 1 byte length */
4583 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4584 		return 0;
4585 
4586 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4587 	kind = type&3;
4588 
4589 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4590 		return -EPERM;
4591 
4592 	rcu_read_lock();
4593 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4594 		struct sock *rtnl;
4595 		rtnl_dumpit_func dumpit;
4596 		u16 min_dump_alloc = 0;
4597 
4598 		link = rtnl_get_link(family, type);
4599 		if (!link || !link->dumpit) {
4600 			family = PF_UNSPEC;
4601 			link = rtnl_get_link(family, type);
4602 			if (!link || !link->dumpit)
4603 				goto err_unlock;
4604 		}
4605 		owner = link->owner;
4606 		dumpit = link->dumpit;
4607 
4608 		if (type == RTM_GETLINK - RTM_BASE)
4609 			min_dump_alloc = rtnl_calcit(skb, nlh);
4610 
4611 		err = 0;
4612 		/* need to do this before rcu_read_unlock() */
4613 		if (!try_module_get(owner))
4614 			err = -EPROTONOSUPPORT;
4615 
4616 		rcu_read_unlock();
4617 
4618 		rtnl = net->rtnl;
4619 		if (err == 0) {
4620 			struct netlink_dump_control c = {
4621 				.dump		= dumpit,
4622 				.min_dump_alloc	= min_dump_alloc,
4623 				.module		= owner,
4624 			};
4625 			err = netlink_dump_start(rtnl, skb, nlh, &c);
4626 			/* netlink_dump_start() will keep a reference on
4627 			 * module if dump is still in progress.
4628 			 */
4629 			module_put(owner);
4630 		}
4631 		return err;
4632 	}
4633 
4634 	link = rtnl_get_link(family, type);
4635 	if (!link || !link->doit) {
4636 		family = PF_UNSPEC;
4637 		link = rtnl_get_link(PF_UNSPEC, type);
4638 		if (!link || !link->doit)
4639 			goto out_unlock;
4640 	}
4641 
4642 	owner = link->owner;
4643 	if (!try_module_get(owner)) {
4644 		err = -EPROTONOSUPPORT;
4645 		goto out_unlock;
4646 	}
4647 
4648 	flags = link->flags;
4649 	if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4650 		doit = link->doit;
4651 		rcu_read_unlock();
4652 		if (doit)
4653 			err = doit(skb, nlh, extack);
4654 		module_put(owner);
4655 		return err;
4656 	}
4657 	rcu_read_unlock();
4658 
4659 	rtnl_lock();
4660 	link = rtnl_get_link(family, type);
4661 	if (link && link->doit)
4662 		err = link->doit(skb, nlh, extack);
4663 	rtnl_unlock();
4664 
4665 	module_put(owner);
4666 
4667 	return err;
4668 
4669 out_unlock:
4670 	rcu_read_unlock();
4671 	return err;
4672 
4673 err_unlock:
4674 	rcu_read_unlock();
4675 	return -EOPNOTSUPP;
4676 }
4677 
4678 static void rtnetlink_rcv(struct sk_buff *skb)
4679 {
4680 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4681 }
4682 
4683 static int rtnetlink_bind(struct net *net, int group)
4684 {
4685 	switch (group) {
4686 	case RTNLGRP_IPV4_MROUTE_R:
4687 	case RTNLGRP_IPV6_MROUTE_R:
4688 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4689 			return -EPERM;
4690 		break;
4691 	}
4692 	return 0;
4693 }
4694 
4695 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4696 {
4697 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4698 
4699 	switch (event) {
4700 	case NETDEV_REBOOT:
4701 	case NETDEV_CHANGEMTU:
4702 	case NETDEV_CHANGEADDR:
4703 	case NETDEV_CHANGENAME:
4704 	case NETDEV_FEAT_CHANGE:
4705 	case NETDEV_BONDING_FAILOVER:
4706 	case NETDEV_POST_TYPE_CHANGE:
4707 	case NETDEV_NOTIFY_PEERS:
4708 	case NETDEV_CHANGEUPPER:
4709 	case NETDEV_RESEND_IGMP:
4710 	case NETDEV_CHANGEINFODATA:
4711 	case NETDEV_CHANGELOWERSTATE:
4712 	case NETDEV_CHANGE_TX_QUEUE_LEN:
4713 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4714 				   GFP_KERNEL, NULL, 0);
4715 		break;
4716 	default:
4717 		break;
4718 	}
4719 	return NOTIFY_DONE;
4720 }
4721 
4722 static struct notifier_block rtnetlink_dev_notifier = {
4723 	.notifier_call	= rtnetlink_event,
4724 };
4725 
4726 
4727 static int __net_init rtnetlink_net_init(struct net *net)
4728 {
4729 	struct sock *sk;
4730 	struct netlink_kernel_cfg cfg = {
4731 		.groups		= RTNLGRP_MAX,
4732 		.input		= rtnetlink_rcv,
4733 		.cb_mutex	= &rtnl_mutex,
4734 		.flags		= NL_CFG_F_NONROOT_RECV,
4735 		.bind		= rtnetlink_bind,
4736 	};
4737 
4738 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4739 	if (!sk)
4740 		return -ENOMEM;
4741 	net->rtnl = sk;
4742 	return 0;
4743 }
4744 
4745 static void __net_exit rtnetlink_net_exit(struct net *net)
4746 {
4747 	netlink_kernel_release(net->rtnl);
4748 	net->rtnl = NULL;
4749 }
4750 
4751 static struct pernet_operations rtnetlink_net_ops = {
4752 	.init = rtnetlink_net_init,
4753 	.exit = rtnetlink_net_exit,
4754 };
4755 
4756 void __init rtnetlink_init(void)
4757 {
4758 	if (register_pernet_subsys(&rtnetlink_net_ops))
4759 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
4760 
4761 	register_netdevice_notifier(&rtnetlink_dev_notifier);
4762 
4763 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4764 		      rtnl_dump_ifinfo, 0);
4765 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4766 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4767 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4768 
4769 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4770 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4771 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4772 
4773 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4774 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4775 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4776 
4777 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4778 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4779 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4780 
4781 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4782 		      0);
4783 }
4784