xref: /linux/net/core/rtnetlink.c (revision d27656d02d85078c63f060fca9c5d99794791a75)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Routing netlink socket interface: protocol independent part.
8  *
9  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10  *
11  *	Fixes:
12  *	Vitaly E. Lavrov		RTA_OK arithmetic was wrong.
13  */
14 
15 #include <linux/bitops.h>
16 #include <linux/errno.h>
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/kernel.h>
21 #include <linux/timer.h>
22 #include <linux/string.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/fcntl.h>
26 #include <linux/mm.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/capability.h>
30 #include <linux/skbuff.h>
31 #include <linux/init.h>
32 #include <linux/security.h>
33 #include <linux/mutex.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_vlan.h>
37 #include <linux/pci.h>
38 #include <linux/etherdevice.h>
39 #include <linux/bpf.h>
40 
41 #include <linux/uaccess.h>
42 
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <net/ip.h>
46 #include <net/protocol.h>
47 #include <net/arp.h>
48 #include <net/route.h>
49 #include <net/udp.h>
50 #include <net/tcp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 
57 #define RTNL_MAX_TYPE		50
58 #define RTNL_SLAVE_MAX_TYPE	40
59 
60 struct rtnl_link {
61 	rtnl_doit_func		doit;
62 	rtnl_dumpit_func	dumpit;
63 	struct module		*owner;
64 	unsigned int		flags;
65 	struct rcu_head		rcu;
66 };
67 
68 static DEFINE_MUTEX(rtnl_mutex);
69 
70 void rtnl_lock(void)
71 {
72 	mutex_lock(&rtnl_mutex);
73 }
74 EXPORT_SYMBOL(rtnl_lock);
75 
76 int rtnl_lock_killable(void)
77 {
78 	return mutex_lock_killable(&rtnl_mutex);
79 }
80 EXPORT_SYMBOL(rtnl_lock_killable);
81 
82 static struct sk_buff *defer_kfree_skb_list;
83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
84 {
85 	if (head && tail) {
86 		tail->next = defer_kfree_skb_list;
87 		defer_kfree_skb_list = head;
88 	}
89 }
90 EXPORT_SYMBOL(rtnl_kfree_skbs);
91 
92 void __rtnl_unlock(void)
93 {
94 	struct sk_buff *head = defer_kfree_skb_list;
95 
96 	defer_kfree_skb_list = NULL;
97 
98 	mutex_unlock(&rtnl_mutex);
99 
100 	while (head) {
101 		struct sk_buff *next = head->next;
102 
103 		kfree_skb(head);
104 		cond_resched();
105 		head = next;
106 	}
107 }
108 
109 void rtnl_unlock(void)
110 {
111 	/* This fellow will unlock it for us. */
112 	netdev_run_todo();
113 }
114 EXPORT_SYMBOL(rtnl_unlock);
115 
116 int rtnl_trylock(void)
117 {
118 	return mutex_trylock(&rtnl_mutex);
119 }
120 EXPORT_SYMBOL(rtnl_trylock);
121 
122 int rtnl_is_locked(void)
123 {
124 	return mutex_is_locked(&rtnl_mutex);
125 }
126 EXPORT_SYMBOL(rtnl_is_locked);
127 
128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
129 {
130 	return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
131 }
132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
133 
134 #ifdef CONFIG_PROVE_LOCKING
135 bool lockdep_rtnl_is_held(void)
136 {
137 	return lockdep_is_held(&rtnl_mutex);
138 }
139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
141 
142 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
143 
144 static inline int rtm_msgindex(int msgtype)
145 {
146 	int msgindex = msgtype - RTM_BASE;
147 
148 	/*
149 	 * msgindex < 0 implies someone tried to register a netlink
150 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
151 	 * the message type has not been added to linux/rtnetlink.h
152 	 */
153 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
154 
155 	return msgindex;
156 }
157 
158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
159 {
160 	struct rtnl_link __rcu **tab;
161 
162 	if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
163 		protocol = PF_UNSPEC;
164 
165 	tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
166 	if (!tab)
167 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
168 
169 	return rcu_dereference_rtnl(tab[msgtype]);
170 }
171 
172 static int rtnl_register_internal(struct module *owner,
173 				  int protocol, int msgtype,
174 				  rtnl_doit_func doit, rtnl_dumpit_func dumpit,
175 				  unsigned int flags)
176 {
177 	struct rtnl_link *link, *old;
178 	struct rtnl_link __rcu **tab;
179 	int msgindex;
180 	int ret = -ENOBUFS;
181 
182 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
183 	msgindex = rtm_msgindex(msgtype);
184 
185 	rtnl_lock();
186 	tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
187 	if (tab == NULL) {
188 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
189 		if (!tab)
190 			goto unlock;
191 
192 		/* ensures we see the 0 stores */
193 		rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
194 	}
195 
196 	old = rtnl_dereference(tab[msgindex]);
197 	if (old) {
198 		link = kmemdup(old, sizeof(*old), GFP_KERNEL);
199 		if (!link)
200 			goto unlock;
201 	} else {
202 		link = kzalloc(sizeof(*link), GFP_KERNEL);
203 		if (!link)
204 			goto unlock;
205 	}
206 
207 	WARN_ON(link->owner && link->owner != owner);
208 	link->owner = owner;
209 
210 	WARN_ON(doit && link->doit && link->doit != doit);
211 	if (doit)
212 		link->doit = doit;
213 	WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
214 	if (dumpit)
215 		link->dumpit = dumpit;
216 
217 	link->flags |= flags;
218 
219 	/* publish protocol:msgtype */
220 	rcu_assign_pointer(tab[msgindex], link);
221 	ret = 0;
222 	if (old)
223 		kfree_rcu(old, rcu);
224 unlock:
225 	rtnl_unlock();
226 	return ret;
227 }
228 
229 /**
230  * rtnl_register_module - Register a rtnetlink message type
231  *
232  * @owner: module registering the hook (THIS_MODULE)
233  * @protocol: Protocol family or PF_UNSPEC
234  * @msgtype: rtnetlink message type
235  * @doit: Function pointer called for each request message
236  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
237  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
238  *
239  * Like rtnl_register, but for use by removable modules.
240  */
241 int rtnl_register_module(struct module *owner,
242 			 int protocol, int msgtype,
243 			 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
244 			 unsigned int flags)
245 {
246 	return rtnl_register_internal(owner, protocol, msgtype,
247 				      doit, dumpit, flags);
248 }
249 EXPORT_SYMBOL_GPL(rtnl_register_module);
250 
251 /**
252  * rtnl_register - Register a rtnetlink message type
253  * @protocol: Protocol family or PF_UNSPEC
254  * @msgtype: rtnetlink message type
255  * @doit: Function pointer called for each request message
256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
257  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
258  *
259  * Registers the specified function pointers (at least one of them has
260  * to be non-NULL) to be called whenever a request message for the
261  * specified protocol family and message type is received.
262  *
263  * The special protocol family PF_UNSPEC may be used to define fallback
264  * function pointers for the case when no entry for the specific protocol
265  * family exists.
266  */
267 void rtnl_register(int protocol, int msgtype,
268 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
269 		   unsigned int flags)
270 {
271 	int err;
272 
273 	err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
274 				     flags);
275 	if (err)
276 		pr_err("Unable to register rtnetlink message handler, "
277 		       "protocol = %d, message type = %d\n", protocol, msgtype);
278 }
279 
280 /**
281  * rtnl_unregister - Unregister a rtnetlink message type
282  * @protocol: Protocol family or PF_UNSPEC
283  * @msgtype: rtnetlink message type
284  *
285  * Returns 0 on success or a negative error code.
286  */
287 int rtnl_unregister(int protocol, int msgtype)
288 {
289 	struct rtnl_link __rcu **tab;
290 	struct rtnl_link *link;
291 	int msgindex;
292 
293 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
294 	msgindex = rtm_msgindex(msgtype);
295 
296 	rtnl_lock();
297 	tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
298 	if (!tab) {
299 		rtnl_unlock();
300 		return -ENOENT;
301 	}
302 
303 	link = rtnl_dereference(tab[msgindex]);
304 	RCU_INIT_POINTER(tab[msgindex], NULL);
305 	rtnl_unlock();
306 
307 	kfree_rcu(link, rcu);
308 
309 	return 0;
310 }
311 EXPORT_SYMBOL_GPL(rtnl_unregister);
312 
313 /**
314  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
315  * @protocol : Protocol family or PF_UNSPEC
316  *
317  * Identical to calling rtnl_unregster() for all registered message types
318  * of a certain protocol family.
319  */
320 void rtnl_unregister_all(int protocol)
321 {
322 	struct rtnl_link __rcu **tab;
323 	struct rtnl_link *link;
324 	int msgindex;
325 
326 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
327 
328 	rtnl_lock();
329 	tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
330 	if (!tab) {
331 		rtnl_unlock();
332 		return;
333 	}
334 	RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
335 	for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
336 		link = rtnl_dereference(tab[msgindex]);
337 		if (!link)
338 			continue;
339 
340 		RCU_INIT_POINTER(tab[msgindex], NULL);
341 		kfree_rcu(link, rcu);
342 	}
343 	rtnl_unlock();
344 
345 	synchronize_net();
346 
347 	kfree(tab);
348 }
349 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
350 
351 static LIST_HEAD(link_ops);
352 
353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
354 {
355 	const struct rtnl_link_ops *ops;
356 
357 	list_for_each_entry(ops, &link_ops, list) {
358 		if (!strcmp(ops->kind, kind))
359 			return ops;
360 	}
361 	return NULL;
362 }
363 
364 /**
365  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
366  * @ops: struct rtnl_link_ops * to register
367  *
368  * The caller must hold the rtnl_mutex. This function should be used
369  * by drivers that create devices during module initialization. It
370  * must be called before registering the devices.
371  *
372  * Returns 0 on success or a negative error code.
373  */
374 int __rtnl_link_register(struct rtnl_link_ops *ops)
375 {
376 	if (rtnl_link_ops_get(ops->kind))
377 		return -EEXIST;
378 
379 	/* The check for alloc/setup is here because if ops
380 	 * does not have that filled up, it is not possible
381 	 * to use the ops for creating device. So do not
382 	 * fill up dellink as well. That disables rtnl_dellink.
383 	 */
384 	if ((ops->alloc || ops->setup) && !ops->dellink)
385 		ops->dellink = unregister_netdevice_queue;
386 
387 	list_add_tail(&ops->list, &link_ops);
388 	return 0;
389 }
390 EXPORT_SYMBOL_GPL(__rtnl_link_register);
391 
392 /**
393  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
394  * @ops: struct rtnl_link_ops * to register
395  *
396  * Returns 0 on success or a negative error code.
397  */
398 int rtnl_link_register(struct rtnl_link_ops *ops)
399 {
400 	int err;
401 
402 	/* Sanity-check max sizes to avoid stack buffer overflow. */
403 	if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
404 		    ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
405 		return -EINVAL;
406 
407 	rtnl_lock();
408 	err = __rtnl_link_register(ops);
409 	rtnl_unlock();
410 	return err;
411 }
412 EXPORT_SYMBOL_GPL(rtnl_link_register);
413 
414 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
415 {
416 	struct net_device *dev;
417 	LIST_HEAD(list_kill);
418 
419 	for_each_netdev(net, dev) {
420 		if (dev->rtnl_link_ops == ops)
421 			ops->dellink(dev, &list_kill);
422 	}
423 	unregister_netdevice_many(&list_kill);
424 }
425 
426 /**
427  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
428  * @ops: struct rtnl_link_ops * to unregister
429  *
430  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
431  * integrity (hold pernet_ops_rwsem for writing to close the race
432  * with setup_net() and cleanup_net()).
433  */
434 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
435 {
436 	struct net *net;
437 
438 	for_each_net(net) {
439 		__rtnl_kill_links(net, ops);
440 	}
441 	list_del(&ops->list);
442 }
443 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
444 
445 /* Return with the rtnl_lock held when there are no network
446  * devices unregistering in any network namespace.
447  */
448 static void rtnl_lock_unregistering_all(void)
449 {
450 	struct net *net;
451 	bool unregistering;
452 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
453 
454 	add_wait_queue(&netdev_unregistering_wq, &wait);
455 	for (;;) {
456 		unregistering = false;
457 		rtnl_lock();
458 		/* We held write locked pernet_ops_rwsem, and parallel
459 		 * setup_net() and cleanup_net() are not possible.
460 		 */
461 		for_each_net(net) {
462 			if (atomic_read(&net->dev_unreg_count) > 0) {
463 				unregistering = true;
464 				break;
465 			}
466 		}
467 		if (!unregistering)
468 			break;
469 		__rtnl_unlock();
470 
471 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
472 	}
473 	remove_wait_queue(&netdev_unregistering_wq, &wait);
474 }
475 
476 /**
477  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
478  * @ops: struct rtnl_link_ops * to unregister
479  */
480 void rtnl_link_unregister(struct rtnl_link_ops *ops)
481 {
482 	/* Close the race with setup_net() and cleanup_net() */
483 	down_write(&pernet_ops_rwsem);
484 	rtnl_lock_unregistering_all();
485 	__rtnl_link_unregister(ops);
486 	rtnl_unlock();
487 	up_write(&pernet_ops_rwsem);
488 }
489 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
490 
491 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
492 {
493 	struct net_device *master_dev;
494 	const struct rtnl_link_ops *ops;
495 	size_t size = 0;
496 
497 	rcu_read_lock();
498 
499 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
500 	if (!master_dev)
501 		goto out;
502 
503 	ops = master_dev->rtnl_link_ops;
504 	if (!ops || !ops->get_slave_size)
505 		goto out;
506 	/* IFLA_INFO_SLAVE_DATA + nested data */
507 	size = nla_total_size(sizeof(struct nlattr)) +
508 	       ops->get_slave_size(master_dev, dev);
509 
510 out:
511 	rcu_read_unlock();
512 	return size;
513 }
514 
515 static size_t rtnl_link_get_size(const struct net_device *dev)
516 {
517 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
518 	size_t size;
519 
520 	if (!ops)
521 		return 0;
522 
523 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
524 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
525 
526 	if (ops->get_size)
527 		/* IFLA_INFO_DATA + nested data */
528 		size += nla_total_size(sizeof(struct nlattr)) +
529 			ops->get_size(dev);
530 
531 	if (ops->get_xstats_size)
532 		/* IFLA_INFO_XSTATS */
533 		size += nla_total_size(ops->get_xstats_size(dev));
534 
535 	size += rtnl_link_get_slave_info_data_size(dev);
536 
537 	return size;
538 }
539 
540 static LIST_HEAD(rtnl_af_ops);
541 
542 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
543 {
544 	const struct rtnl_af_ops *ops;
545 
546 	ASSERT_RTNL();
547 
548 	list_for_each_entry(ops, &rtnl_af_ops, list) {
549 		if (ops->family == family)
550 			return ops;
551 	}
552 
553 	return NULL;
554 }
555 
556 /**
557  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
558  * @ops: struct rtnl_af_ops * to register
559  *
560  * Returns 0 on success or a negative error code.
561  */
562 void rtnl_af_register(struct rtnl_af_ops *ops)
563 {
564 	rtnl_lock();
565 	list_add_tail_rcu(&ops->list, &rtnl_af_ops);
566 	rtnl_unlock();
567 }
568 EXPORT_SYMBOL_GPL(rtnl_af_register);
569 
570 /**
571  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
572  * @ops: struct rtnl_af_ops * to unregister
573  */
574 void rtnl_af_unregister(struct rtnl_af_ops *ops)
575 {
576 	rtnl_lock();
577 	list_del_rcu(&ops->list);
578 	rtnl_unlock();
579 
580 	synchronize_rcu();
581 }
582 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
583 
584 static size_t rtnl_link_get_af_size(const struct net_device *dev,
585 				    u32 ext_filter_mask)
586 {
587 	struct rtnl_af_ops *af_ops;
588 	size_t size;
589 
590 	/* IFLA_AF_SPEC */
591 	size = nla_total_size(sizeof(struct nlattr));
592 
593 	rcu_read_lock();
594 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
595 		if (af_ops->get_link_af_size) {
596 			/* AF_* + nested data */
597 			size += nla_total_size(sizeof(struct nlattr)) +
598 				af_ops->get_link_af_size(dev, ext_filter_mask);
599 		}
600 	}
601 	rcu_read_unlock();
602 
603 	return size;
604 }
605 
606 static bool rtnl_have_link_slave_info(const struct net_device *dev)
607 {
608 	struct net_device *master_dev;
609 	bool ret = false;
610 
611 	rcu_read_lock();
612 
613 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
614 	if (master_dev && master_dev->rtnl_link_ops)
615 		ret = true;
616 	rcu_read_unlock();
617 	return ret;
618 }
619 
620 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
621 				     const struct net_device *dev)
622 {
623 	struct net_device *master_dev;
624 	const struct rtnl_link_ops *ops;
625 	struct nlattr *slave_data;
626 	int err;
627 
628 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
629 	if (!master_dev)
630 		return 0;
631 	ops = master_dev->rtnl_link_ops;
632 	if (!ops)
633 		return 0;
634 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
635 		return -EMSGSIZE;
636 	if (ops->fill_slave_info) {
637 		slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
638 		if (!slave_data)
639 			return -EMSGSIZE;
640 		err = ops->fill_slave_info(skb, master_dev, dev);
641 		if (err < 0)
642 			goto err_cancel_slave_data;
643 		nla_nest_end(skb, slave_data);
644 	}
645 	return 0;
646 
647 err_cancel_slave_data:
648 	nla_nest_cancel(skb, slave_data);
649 	return err;
650 }
651 
652 static int rtnl_link_info_fill(struct sk_buff *skb,
653 			       const struct net_device *dev)
654 {
655 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
656 	struct nlattr *data;
657 	int err;
658 
659 	if (!ops)
660 		return 0;
661 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
662 		return -EMSGSIZE;
663 	if (ops->fill_xstats) {
664 		err = ops->fill_xstats(skb, dev);
665 		if (err < 0)
666 			return err;
667 	}
668 	if (ops->fill_info) {
669 		data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
670 		if (data == NULL)
671 			return -EMSGSIZE;
672 		err = ops->fill_info(skb, dev);
673 		if (err < 0)
674 			goto err_cancel_data;
675 		nla_nest_end(skb, data);
676 	}
677 	return 0;
678 
679 err_cancel_data:
680 	nla_nest_cancel(skb, data);
681 	return err;
682 }
683 
684 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
685 {
686 	struct nlattr *linkinfo;
687 	int err = -EMSGSIZE;
688 
689 	linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
690 	if (linkinfo == NULL)
691 		goto out;
692 
693 	err = rtnl_link_info_fill(skb, dev);
694 	if (err < 0)
695 		goto err_cancel_link;
696 
697 	err = rtnl_link_slave_info_fill(skb, dev);
698 	if (err < 0)
699 		goto err_cancel_link;
700 
701 	nla_nest_end(skb, linkinfo);
702 	return 0;
703 
704 err_cancel_link:
705 	nla_nest_cancel(skb, linkinfo);
706 out:
707 	return err;
708 }
709 
710 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
711 {
712 	struct sock *rtnl = net->rtnl;
713 
714 	return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
715 }
716 
717 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
718 {
719 	struct sock *rtnl = net->rtnl;
720 
721 	return nlmsg_unicast(rtnl, skb, pid);
722 }
723 EXPORT_SYMBOL(rtnl_unicast);
724 
725 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
726 		 struct nlmsghdr *nlh, gfp_t flags)
727 {
728 	struct sock *rtnl = net->rtnl;
729 
730 	nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
731 }
732 EXPORT_SYMBOL(rtnl_notify);
733 
734 void rtnl_set_sk_err(struct net *net, u32 group, int error)
735 {
736 	struct sock *rtnl = net->rtnl;
737 
738 	netlink_set_err(rtnl, 0, group, error);
739 }
740 EXPORT_SYMBOL(rtnl_set_sk_err);
741 
742 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
743 {
744 	struct nlattr *mx;
745 	int i, valid = 0;
746 
747 	/* nothing is dumped for dst_default_metrics, so just skip the loop */
748 	if (metrics == dst_default_metrics.metrics)
749 		return 0;
750 
751 	mx = nla_nest_start_noflag(skb, RTA_METRICS);
752 	if (mx == NULL)
753 		return -ENOBUFS;
754 
755 	for (i = 0; i < RTAX_MAX; i++) {
756 		if (metrics[i]) {
757 			if (i == RTAX_CC_ALGO - 1) {
758 				char tmp[TCP_CA_NAME_MAX], *name;
759 
760 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
761 				if (!name)
762 					continue;
763 				if (nla_put_string(skb, i + 1, name))
764 					goto nla_put_failure;
765 			} else if (i == RTAX_FEATURES - 1) {
766 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
767 
768 				if (!user_features)
769 					continue;
770 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
771 				if (nla_put_u32(skb, i + 1, user_features))
772 					goto nla_put_failure;
773 			} else {
774 				if (nla_put_u32(skb, i + 1, metrics[i]))
775 					goto nla_put_failure;
776 			}
777 			valid++;
778 		}
779 	}
780 
781 	if (!valid) {
782 		nla_nest_cancel(skb, mx);
783 		return 0;
784 	}
785 
786 	return nla_nest_end(skb, mx);
787 
788 nla_put_failure:
789 	nla_nest_cancel(skb, mx);
790 	return -EMSGSIZE;
791 }
792 EXPORT_SYMBOL(rtnetlink_put_metrics);
793 
794 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
795 		       long expires, u32 error)
796 {
797 	struct rta_cacheinfo ci = {
798 		.rta_error = error,
799 		.rta_id =  id,
800 	};
801 
802 	if (dst) {
803 		ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
804 		ci.rta_used = dst->__use;
805 		ci.rta_clntref = atomic_read(&dst->__refcnt);
806 	}
807 	if (expires) {
808 		unsigned long clock;
809 
810 		clock = jiffies_to_clock_t(abs(expires));
811 		clock = min_t(unsigned long, clock, INT_MAX);
812 		ci.rta_expires = (expires > 0) ? clock : -clock;
813 	}
814 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
815 }
816 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
817 
818 static void set_operstate(struct net_device *dev, unsigned char transition)
819 {
820 	unsigned char operstate = dev->operstate;
821 
822 	switch (transition) {
823 	case IF_OPER_UP:
824 		if ((operstate == IF_OPER_DORMANT ||
825 		     operstate == IF_OPER_TESTING ||
826 		     operstate == IF_OPER_UNKNOWN) &&
827 		    !netif_dormant(dev) && !netif_testing(dev))
828 			operstate = IF_OPER_UP;
829 		break;
830 
831 	case IF_OPER_TESTING:
832 		if (operstate == IF_OPER_UP ||
833 		    operstate == IF_OPER_UNKNOWN)
834 			operstate = IF_OPER_TESTING;
835 		break;
836 
837 	case IF_OPER_DORMANT:
838 		if (operstate == IF_OPER_UP ||
839 		    operstate == IF_OPER_UNKNOWN)
840 			operstate = IF_OPER_DORMANT;
841 		break;
842 	}
843 
844 	if (dev->operstate != operstate) {
845 		write_lock(&dev_base_lock);
846 		dev->operstate = operstate;
847 		write_unlock(&dev_base_lock);
848 		netdev_state_change(dev);
849 	}
850 }
851 
852 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
853 {
854 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
855 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
856 }
857 
858 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
859 					   const struct ifinfomsg *ifm)
860 {
861 	unsigned int flags = ifm->ifi_flags;
862 
863 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
864 	if (ifm->ifi_change)
865 		flags = (flags & ifm->ifi_change) |
866 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
867 
868 	return flags;
869 }
870 
871 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
872 				 const struct rtnl_link_stats64 *b)
873 {
874 	a->rx_packets = b->rx_packets;
875 	a->tx_packets = b->tx_packets;
876 	a->rx_bytes = b->rx_bytes;
877 	a->tx_bytes = b->tx_bytes;
878 	a->rx_errors = b->rx_errors;
879 	a->tx_errors = b->tx_errors;
880 	a->rx_dropped = b->rx_dropped;
881 	a->tx_dropped = b->tx_dropped;
882 
883 	a->multicast = b->multicast;
884 	a->collisions = b->collisions;
885 
886 	a->rx_length_errors = b->rx_length_errors;
887 	a->rx_over_errors = b->rx_over_errors;
888 	a->rx_crc_errors = b->rx_crc_errors;
889 	a->rx_frame_errors = b->rx_frame_errors;
890 	a->rx_fifo_errors = b->rx_fifo_errors;
891 	a->rx_missed_errors = b->rx_missed_errors;
892 
893 	a->tx_aborted_errors = b->tx_aborted_errors;
894 	a->tx_carrier_errors = b->tx_carrier_errors;
895 	a->tx_fifo_errors = b->tx_fifo_errors;
896 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
897 	a->tx_window_errors = b->tx_window_errors;
898 
899 	a->rx_compressed = b->rx_compressed;
900 	a->tx_compressed = b->tx_compressed;
901 
902 	a->rx_nohandler = b->rx_nohandler;
903 }
904 
905 /* All VF info */
906 static inline int rtnl_vfinfo_size(const struct net_device *dev,
907 				   u32 ext_filter_mask)
908 {
909 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
910 		int num_vfs = dev_num_vf(dev->dev.parent);
911 		size_t size = nla_total_size(0);
912 		size += num_vfs *
913 			(nla_total_size(0) +
914 			 nla_total_size(sizeof(struct ifla_vf_mac)) +
915 			 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
916 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
917 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
918 			 nla_total_size(MAX_VLAN_LIST_LEN *
919 					sizeof(struct ifla_vf_vlan_info)) +
920 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
921 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
922 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
923 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
924 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
925 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
926 			 /* IFLA_VF_STATS_RX_PACKETS */
927 			 nla_total_size_64bit(sizeof(__u64)) +
928 			 /* IFLA_VF_STATS_TX_PACKETS */
929 			 nla_total_size_64bit(sizeof(__u64)) +
930 			 /* IFLA_VF_STATS_RX_BYTES */
931 			 nla_total_size_64bit(sizeof(__u64)) +
932 			 /* IFLA_VF_STATS_TX_BYTES */
933 			 nla_total_size_64bit(sizeof(__u64)) +
934 			 /* IFLA_VF_STATS_BROADCAST */
935 			 nla_total_size_64bit(sizeof(__u64)) +
936 			 /* IFLA_VF_STATS_MULTICAST */
937 			 nla_total_size_64bit(sizeof(__u64)) +
938 			 /* IFLA_VF_STATS_RX_DROPPED */
939 			 nla_total_size_64bit(sizeof(__u64)) +
940 			 /* IFLA_VF_STATS_TX_DROPPED */
941 			 nla_total_size_64bit(sizeof(__u64)) +
942 			 nla_total_size(sizeof(struct ifla_vf_trust)));
943 		return size;
944 	} else
945 		return 0;
946 }
947 
948 static size_t rtnl_port_size(const struct net_device *dev,
949 			     u32 ext_filter_mask)
950 {
951 	size_t port_size = nla_total_size(4)		/* PORT_VF */
952 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
953 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
954 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
955 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
956 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
957 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
958 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
959 		+ port_size;
960 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
961 		+ port_size;
962 
963 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
964 	    !(ext_filter_mask & RTEXT_FILTER_VF))
965 		return 0;
966 	if (dev_num_vf(dev->dev.parent))
967 		return port_self_size + vf_ports_size +
968 			vf_port_size * dev_num_vf(dev->dev.parent);
969 	else
970 		return port_self_size;
971 }
972 
973 static size_t rtnl_xdp_size(void)
974 {
975 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
976 			  nla_total_size(1) +	/* XDP_ATTACHED */
977 			  nla_total_size(4) +	/* XDP_PROG_ID (or 1st mode) */
978 			  nla_total_size(4);	/* XDP_<mode>_PROG_ID */
979 
980 	return xdp_size;
981 }
982 
983 static size_t rtnl_prop_list_size(const struct net_device *dev)
984 {
985 	struct netdev_name_node *name_node;
986 	size_t size;
987 
988 	if (list_empty(&dev->name_node->list))
989 		return 0;
990 	size = nla_total_size(0);
991 	list_for_each_entry(name_node, &dev->name_node->list, list)
992 		size += nla_total_size(ALTIFNAMSIZ);
993 	return size;
994 }
995 
996 static size_t rtnl_proto_down_size(const struct net_device *dev)
997 {
998 	size_t size = nla_total_size(1);
999 
1000 	if (dev->proto_down_reason)
1001 		size += nla_total_size(0) + nla_total_size(4);
1002 
1003 	return size;
1004 }
1005 
1006 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1007 				     u32 ext_filter_mask)
1008 {
1009 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1010 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1011 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1012 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1013 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1014 	       + nla_total_size(sizeof(struct rtnl_link_stats))
1015 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1016 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1017 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1018 	       + nla_total_size(4) /* IFLA_TXQLEN */
1019 	       + nla_total_size(4) /* IFLA_WEIGHT */
1020 	       + nla_total_size(4) /* IFLA_MTU */
1021 	       + nla_total_size(4) /* IFLA_LINK */
1022 	       + nla_total_size(4) /* IFLA_MASTER */
1023 	       + nla_total_size(1) /* IFLA_CARRIER */
1024 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
1025 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1026 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1027 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1028 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1029 	       + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1030 	       + nla_total_size(1) /* IFLA_OPERSTATE */
1031 	       + nla_total_size(1) /* IFLA_LINKMODE */
1032 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1033 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
1034 	       + nla_total_size(4) /* IFLA_GROUP */
1035 	       + nla_total_size(ext_filter_mask
1036 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1037 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1038 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1039 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1040 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1041 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1042 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1043 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1044 	       + rtnl_xdp_size() /* IFLA_XDP */
1045 	       + nla_total_size(4)  /* IFLA_EVENT */
1046 	       + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1047 	       + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1048 	       + rtnl_proto_down_size(dev)  /* proto down */
1049 	       + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1050 	       + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1051 	       + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1052 	       + nla_total_size(4)  /* IFLA_MIN_MTU */
1053 	       + nla_total_size(4)  /* IFLA_MAX_MTU */
1054 	       + rtnl_prop_list_size(dev)
1055 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1056 	       + 0;
1057 }
1058 
1059 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1060 {
1061 	struct nlattr *vf_ports;
1062 	struct nlattr *vf_port;
1063 	int vf;
1064 	int err;
1065 
1066 	vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1067 	if (!vf_ports)
1068 		return -EMSGSIZE;
1069 
1070 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1071 		vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1072 		if (!vf_port)
1073 			goto nla_put_failure;
1074 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1075 			goto nla_put_failure;
1076 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1077 		if (err == -EMSGSIZE)
1078 			goto nla_put_failure;
1079 		if (err) {
1080 			nla_nest_cancel(skb, vf_port);
1081 			continue;
1082 		}
1083 		nla_nest_end(skb, vf_port);
1084 	}
1085 
1086 	nla_nest_end(skb, vf_ports);
1087 
1088 	return 0;
1089 
1090 nla_put_failure:
1091 	nla_nest_cancel(skb, vf_ports);
1092 	return -EMSGSIZE;
1093 }
1094 
1095 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1096 {
1097 	struct nlattr *port_self;
1098 	int err;
1099 
1100 	port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1101 	if (!port_self)
1102 		return -EMSGSIZE;
1103 
1104 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1105 	if (err) {
1106 		nla_nest_cancel(skb, port_self);
1107 		return (err == -EMSGSIZE) ? err : 0;
1108 	}
1109 
1110 	nla_nest_end(skb, port_self);
1111 
1112 	return 0;
1113 }
1114 
1115 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1116 			  u32 ext_filter_mask)
1117 {
1118 	int err;
1119 
1120 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1121 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1122 		return 0;
1123 
1124 	err = rtnl_port_self_fill(skb, dev);
1125 	if (err)
1126 		return err;
1127 
1128 	if (dev_num_vf(dev->dev.parent)) {
1129 		err = rtnl_vf_ports_fill(skb, dev);
1130 		if (err)
1131 			return err;
1132 	}
1133 
1134 	return 0;
1135 }
1136 
1137 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1138 {
1139 	int err;
1140 	struct netdev_phys_item_id ppid;
1141 
1142 	err = dev_get_phys_port_id(dev, &ppid);
1143 	if (err) {
1144 		if (err == -EOPNOTSUPP)
1145 			return 0;
1146 		return err;
1147 	}
1148 
1149 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1150 		return -EMSGSIZE;
1151 
1152 	return 0;
1153 }
1154 
1155 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1156 {
1157 	char name[IFNAMSIZ];
1158 	int err;
1159 
1160 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1161 	if (err) {
1162 		if (err == -EOPNOTSUPP)
1163 			return 0;
1164 		return err;
1165 	}
1166 
1167 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1168 		return -EMSGSIZE;
1169 
1170 	return 0;
1171 }
1172 
1173 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1174 {
1175 	struct netdev_phys_item_id ppid = { };
1176 	int err;
1177 
1178 	err = dev_get_port_parent_id(dev, &ppid, false);
1179 	if (err) {
1180 		if (err == -EOPNOTSUPP)
1181 			return 0;
1182 		return err;
1183 	}
1184 
1185 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1186 		return -EMSGSIZE;
1187 
1188 	return 0;
1189 }
1190 
1191 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1192 					      struct net_device *dev)
1193 {
1194 	struct rtnl_link_stats64 *sp;
1195 	struct nlattr *attr;
1196 
1197 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1198 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1199 	if (!attr)
1200 		return -EMSGSIZE;
1201 
1202 	sp = nla_data(attr);
1203 	dev_get_stats(dev, sp);
1204 
1205 	attr = nla_reserve(skb, IFLA_STATS,
1206 			   sizeof(struct rtnl_link_stats));
1207 	if (!attr)
1208 		return -EMSGSIZE;
1209 
1210 	copy_rtnl_link_stats(nla_data(attr), sp);
1211 
1212 	return 0;
1213 }
1214 
1215 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1216 					       struct net_device *dev,
1217 					       int vfs_num,
1218 					       struct nlattr *vfinfo)
1219 {
1220 	struct ifla_vf_rss_query_en vf_rss_query_en;
1221 	struct nlattr *vf, *vfstats, *vfvlanlist;
1222 	struct ifla_vf_link_state vf_linkstate;
1223 	struct ifla_vf_vlan_info vf_vlan_info;
1224 	struct ifla_vf_spoofchk vf_spoofchk;
1225 	struct ifla_vf_tx_rate vf_tx_rate;
1226 	struct ifla_vf_stats vf_stats;
1227 	struct ifla_vf_trust vf_trust;
1228 	struct ifla_vf_vlan vf_vlan;
1229 	struct ifla_vf_rate vf_rate;
1230 	struct ifla_vf_mac vf_mac;
1231 	struct ifla_vf_broadcast vf_broadcast;
1232 	struct ifla_vf_info ivi;
1233 	struct ifla_vf_guid node_guid;
1234 	struct ifla_vf_guid port_guid;
1235 
1236 	memset(&ivi, 0, sizeof(ivi));
1237 
1238 	/* Not all SR-IOV capable drivers support the
1239 	 * spoofcheck and "RSS query enable" query.  Preset to
1240 	 * -1 so the user space tool can detect that the driver
1241 	 * didn't report anything.
1242 	 */
1243 	ivi.spoofchk = -1;
1244 	ivi.rss_query_en = -1;
1245 	ivi.trusted = -1;
1246 	/* The default value for VF link state is "auto"
1247 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1248 	 */
1249 	ivi.linkstate = 0;
1250 	/* VLAN Protocol by default is 802.1Q */
1251 	ivi.vlan_proto = htons(ETH_P_8021Q);
1252 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1253 		return 0;
1254 
1255 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1256 	memset(&node_guid, 0, sizeof(node_guid));
1257 	memset(&port_guid, 0, sizeof(port_guid));
1258 
1259 	vf_mac.vf =
1260 		vf_vlan.vf =
1261 		vf_vlan_info.vf =
1262 		vf_rate.vf =
1263 		vf_tx_rate.vf =
1264 		vf_spoofchk.vf =
1265 		vf_linkstate.vf =
1266 		vf_rss_query_en.vf =
1267 		vf_trust.vf =
1268 		node_guid.vf =
1269 		port_guid.vf = ivi.vf;
1270 
1271 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1272 	memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1273 	vf_vlan.vlan = ivi.vlan;
1274 	vf_vlan.qos = ivi.qos;
1275 	vf_vlan_info.vlan = ivi.vlan;
1276 	vf_vlan_info.qos = ivi.qos;
1277 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1278 	vf_tx_rate.rate = ivi.max_tx_rate;
1279 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1280 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1281 	vf_spoofchk.setting = ivi.spoofchk;
1282 	vf_linkstate.link_state = ivi.linkstate;
1283 	vf_rss_query_en.setting = ivi.rss_query_en;
1284 	vf_trust.setting = ivi.trusted;
1285 	vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1286 	if (!vf)
1287 		goto nla_put_vfinfo_failure;
1288 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1289 	    nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1290 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1291 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1292 		    &vf_rate) ||
1293 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1294 		    &vf_tx_rate) ||
1295 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1296 		    &vf_spoofchk) ||
1297 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1298 		    &vf_linkstate) ||
1299 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1300 		    sizeof(vf_rss_query_en),
1301 		    &vf_rss_query_en) ||
1302 	    nla_put(skb, IFLA_VF_TRUST,
1303 		    sizeof(vf_trust), &vf_trust))
1304 		goto nla_put_vf_failure;
1305 
1306 	if (dev->netdev_ops->ndo_get_vf_guid &&
1307 	    !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1308 					      &port_guid)) {
1309 		if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1310 			    &node_guid) ||
1311 		    nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1312 			    &port_guid))
1313 			goto nla_put_vf_failure;
1314 	}
1315 	vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1316 	if (!vfvlanlist)
1317 		goto nla_put_vf_failure;
1318 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1319 		    &vf_vlan_info)) {
1320 		nla_nest_cancel(skb, vfvlanlist);
1321 		goto nla_put_vf_failure;
1322 	}
1323 	nla_nest_end(skb, vfvlanlist);
1324 	memset(&vf_stats, 0, sizeof(vf_stats));
1325 	if (dev->netdev_ops->ndo_get_vf_stats)
1326 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1327 						&vf_stats);
1328 	vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1329 	if (!vfstats)
1330 		goto nla_put_vf_failure;
1331 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1332 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1333 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1334 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1335 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1336 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1337 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1338 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1339 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1340 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1341 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1342 			      vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1343 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1344 			      vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1345 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1346 			      vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1347 		nla_nest_cancel(skb, vfstats);
1348 		goto nla_put_vf_failure;
1349 	}
1350 	nla_nest_end(skb, vfstats);
1351 	nla_nest_end(skb, vf);
1352 	return 0;
1353 
1354 nla_put_vf_failure:
1355 	nla_nest_cancel(skb, vf);
1356 nla_put_vfinfo_failure:
1357 	nla_nest_cancel(skb, vfinfo);
1358 	return -EMSGSIZE;
1359 }
1360 
1361 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1362 					   struct net_device *dev,
1363 					   u32 ext_filter_mask)
1364 {
1365 	struct nlattr *vfinfo;
1366 	int i, num_vfs;
1367 
1368 	if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1369 		return 0;
1370 
1371 	num_vfs = dev_num_vf(dev->dev.parent);
1372 	if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1373 		return -EMSGSIZE;
1374 
1375 	if (!dev->netdev_ops->ndo_get_vf_config)
1376 		return 0;
1377 
1378 	vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1379 	if (!vfinfo)
1380 		return -EMSGSIZE;
1381 
1382 	for (i = 0; i < num_vfs; i++) {
1383 		if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1384 			return -EMSGSIZE;
1385 	}
1386 
1387 	nla_nest_end(skb, vfinfo);
1388 	return 0;
1389 }
1390 
1391 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1392 {
1393 	struct rtnl_link_ifmap map;
1394 
1395 	memset(&map, 0, sizeof(map));
1396 	map.mem_start   = dev->mem_start;
1397 	map.mem_end     = dev->mem_end;
1398 	map.base_addr   = dev->base_addr;
1399 	map.irq         = dev->irq;
1400 	map.dma         = dev->dma;
1401 	map.port        = dev->if_port;
1402 
1403 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1404 		return -EMSGSIZE;
1405 
1406 	return 0;
1407 }
1408 
1409 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1410 {
1411 	const struct bpf_prog *generic_xdp_prog;
1412 
1413 	ASSERT_RTNL();
1414 
1415 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1416 	if (!generic_xdp_prog)
1417 		return 0;
1418 	return generic_xdp_prog->aux->id;
1419 }
1420 
1421 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1422 {
1423 	return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1424 }
1425 
1426 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1427 {
1428 	return dev_xdp_prog_id(dev, XDP_MODE_HW);
1429 }
1430 
1431 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1432 			       u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1433 			       u32 (*get_prog_id)(struct net_device *dev))
1434 {
1435 	u32 curr_id;
1436 	int err;
1437 
1438 	curr_id = get_prog_id(dev);
1439 	if (!curr_id)
1440 		return 0;
1441 
1442 	*prog_id = curr_id;
1443 	err = nla_put_u32(skb, attr, curr_id);
1444 	if (err)
1445 		return err;
1446 
1447 	if (*mode != XDP_ATTACHED_NONE)
1448 		*mode = XDP_ATTACHED_MULTI;
1449 	else
1450 		*mode = tgt_mode;
1451 
1452 	return 0;
1453 }
1454 
1455 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1456 {
1457 	struct nlattr *xdp;
1458 	u32 prog_id;
1459 	int err;
1460 	u8 mode;
1461 
1462 	xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1463 	if (!xdp)
1464 		return -EMSGSIZE;
1465 
1466 	prog_id = 0;
1467 	mode = XDP_ATTACHED_NONE;
1468 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1469 				  IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1470 	if (err)
1471 		goto err_cancel;
1472 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1473 				  IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1474 	if (err)
1475 		goto err_cancel;
1476 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1477 				  IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1478 	if (err)
1479 		goto err_cancel;
1480 
1481 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1482 	if (err)
1483 		goto err_cancel;
1484 
1485 	if (prog_id && mode != XDP_ATTACHED_MULTI) {
1486 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1487 		if (err)
1488 			goto err_cancel;
1489 	}
1490 
1491 	nla_nest_end(skb, xdp);
1492 	return 0;
1493 
1494 err_cancel:
1495 	nla_nest_cancel(skb, xdp);
1496 	return err;
1497 }
1498 
1499 static u32 rtnl_get_event(unsigned long event)
1500 {
1501 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1502 
1503 	switch (event) {
1504 	case NETDEV_REBOOT:
1505 		rtnl_event_type = IFLA_EVENT_REBOOT;
1506 		break;
1507 	case NETDEV_FEAT_CHANGE:
1508 		rtnl_event_type = IFLA_EVENT_FEATURES;
1509 		break;
1510 	case NETDEV_BONDING_FAILOVER:
1511 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1512 		break;
1513 	case NETDEV_NOTIFY_PEERS:
1514 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1515 		break;
1516 	case NETDEV_RESEND_IGMP:
1517 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1518 		break;
1519 	case NETDEV_CHANGEINFODATA:
1520 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1521 		break;
1522 	default:
1523 		break;
1524 	}
1525 
1526 	return rtnl_event_type;
1527 }
1528 
1529 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1530 {
1531 	const struct net_device *upper_dev;
1532 	int ret = 0;
1533 
1534 	rcu_read_lock();
1535 
1536 	upper_dev = netdev_master_upper_dev_get_rcu(dev);
1537 	if (upper_dev)
1538 		ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1539 
1540 	rcu_read_unlock();
1541 	return ret;
1542 }
1543 
1544 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1545 			  bool force)
1546 {
1547 	int ifindex = dev_get_iflink(dev);
1548 
1549 	if (force || dev->ifindex != ifindex)
1550 		return nla_put_u32(skb, IFLA_LINK, ifindex);
1551 
1552 	return 0;
1553 }
1554 
1555 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1556 					      struct net_device *dev)
1557 {
1558 	char buf[IFALIASZ];
1559 	int ret;
1560 
1561 	ret = dev_get_alias(dev, buf, sizeof(buf));
1562 	return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1563 }
1564 
1565 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1566 				  const struct net_device *dev,
1567 				  struct net *src_net, gfp_t gfp)
1568 {
1569 	bool put_iflink = false;
1570 
1571 	if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1572 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1573 
1574 		if (!net_eq(dev_net(dev), link_net)) {
1575 			int id = peernet2id_alloc(src_net, link_net, gfp);
1576 
1577 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1578 				return -EMSGSIZE;
1579 
1580 			put_iflink = true;
1581 		}
1582 	}
1583 
1584 	return nla_put_iflink(skb, dev, put_iflink);
1585 }
1586 
1587 static int rtnl_fill_link_af(struct sk_buff *skb,
1588 			     const struct net_device *dev,
1589 			     u32 ext_filter_mask)
1590 {
1591 	const struct rtnl_af_ops *af_ops;
1592 	struct nlattr *af_spec;
1593 
1594 	af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1595 	if (!af_spec)
1596 		return -EMSGSIZE;
1597 
1598 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1599 		struct nlattr *af;
1600 		int err;
1601 
1602 		if (!af_ops->fill_link_af)
1603 			continue;
1604 
1605 		af = nla_nest_start_noflag(skb, af_ops->family);
1606 		if (!af)
1607 			return -EMSGSIZE;
1608 
1609 		err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1610 		/*
1611 		 * Caller may return ENODATA to indicate that there
1612 		 * was no data to be dumped. This is not an error, it
1613 		 * means we should trim the attribute header and
1614 		 * continue.
1615 		 */
1616 		if (err == -ENODATA)
1617 			nla_nest_cancel(skb, af);
1618 		else if (err < 0)
1619 			return -EMSGSIZE;
1620 
1621 		nla_nest_end(skb, af);
1622 	}
1623 
1624 	nla_nest_end(skb, af_spec);
1625 	return 0;
1626 }
1627 
1628 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1629 				 const struct net_device *dev)
1630 {
1631 	struct netdev_name_node *name_node;
1632 	int count = 0;
1633 
1634 	list_for_each_entry(name_node, &dev->name_node->list, list) {
1635 		if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1636 			return -EMSGSIZE;
1637 		count++;
1638 	}
1639 	return count;
1640 }
1641 
1642 static int rtnl_fill_prop_list(struct sk_buff *skb,
1643 			       const struct net_device *dev)
1644 {
1645 	struct nlattr *prop_list;
1646 	int ret;
1647 
1648 	prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1649 	if (!prop_list)
1650 		return -EMSGSIZE;
1651 
1652 	ret = rtnl_fill_alt_ifnames(skb, dev);
1653 	if (ret <= 0)
1654 		goto nest_cancel;
1655 
1656 	nla_nest_end(skb, prop_list);
1657 	return 0;
1658 
1659 nest_cancel:
1660 	nla_nest_cancel(skb, prop_list);
1661 	return ret;
1662 }
1663 
1664 static int rtnl_fill_proto_down(struct sk_buff *skb,
1665 				const struct net_device *dev)
1666 {
1667 	struct nlattr *pr;
1668 	u32 preason;
1669 
1670 	if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1671 		goto nla_put_failure;
1672 
1673 	preason = dev->proto_down_reason;
1674 	if (!preason)
1675 		return 0;
1676 
1677 	pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1678 	if (!pr)
1679 		return -EMSGSIZE;
1680 
1681 	if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1682 		nla_nest_cancel(skb, pr);
1683 		goto nla_put_failure;
1684 	}
1685 
1686 	nla_nest_end(skb, pr);
1687 	return 0;
1688 
1689 nla_put_failure:
1690 	return -EMSGSIZE;
1691 }
1692 
1693 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1694 			    struct net_device *dev, struct net *src_net,
1695 			    int type, u32 pid, u32 seq, u32 change,
1696 			    unsigned int flags, u32 ext_filter_mask,
1697 			    u32 event, int *new_nsid, int new_ifindex,
1698 			    int tgt_netnsid, gfp_t gfp)
1699 {
1700 	struct ifinfomsg *ifm;
1701 	struct nlmsghdr *nlh;
1702 	struct Qdisc *qdisc;
1703 
1704 	ASSERT_RTNL();
1705 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1706 	if (nlh == NULL)
1707 		return -EMSGSIZE;
1708 
1709 	ifm = nlmsg_data(nlh);
1710 	ifm->ifi_family = AF_UNSPEC;
1711 	ifm->__ifi_pad = 0;
1712 	ifm->ifi_type = dev->type;
1713 	ifm->ifi_index = dev->ifindex;
1714 	ifm->ifi_flags = dev_get_flags(dev);
1715 	ifm->ifi_change = change;
1716 
1717 	if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1718 		goto nla_put_failure;
1719 
1720 	qdisc = rtnl_dereference(dev->qdisc);
1721 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1722 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1723 	    nla_put_u8(skb, IFLA_OPERSTATE,
1724 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1725 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1726 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1727 	    nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1728 	    nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1729 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1730 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1731 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1732 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1733 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1734 	    nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1735 #ifdef CONFIG_RPS
1736 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1737 #endif
1738 	    put_master_ifindex(skb, dev) ||
1739 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1740 	    (qdisc &&
1741 	     nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1742 	    nla_put_ifalias(skb, dev) ||
1743 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1744 			atomic_read(&dev->carrier_up_count) +
1745 			atomic_read(&dev->carrier_down_count)) ||
1746 	    nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1747 			atomic_read(&dev->carrier_up_count)) ||
1748 	    nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1749 			atomic_read(&dev->carrier_down_count)))
1750 		goto nla_put_failure;
1751 
1752 	if (rtnl_fill_proto_down(skb, dev))
1753 		goto nla_put_failure;
1754 
1755 	if (event != IFLA_EVENT_NONE) {
1756 		if (nla_put_u32(skb, IFLA_EVENT, event))
1757 			goto nla_put_failure;
1758 	}
1759 
1760 	if (rtnl_fill_link_ifmap(skb, dev))
1761 		goto nla_put_failure;
1762 
1763 	if (dev->addr_len) {
1764 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1765 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1766 			goto nla_put_failure;
1767 	}
1768 
1769 	if (rtnl_phys_port_id_fill(skb, dev))
1770 		goto nla_put_failure;
1771 
1772 	if (rtnl_phys_port_name_fill(skb, dev))
1773 		goto nla_put_failure;
1774 
1775 	if (rtnl_phys_switch_id_fill(skb, dev))
1776 		goto nla_put_failure;
1777 
1778 	if (rtnl_fill_stats(skb, dev))
1779 		goto nla_put_failure;
1780 
1781 	if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1782 		goto nla_put_failure;
1783 
1784 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1785 		goto nla_put_failure;
1786 
1787 	if (rtnl_xdp_fill(skb, dev))
1788 		goto nla_put_failure;
1789 
1790 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1791 		if (rtnl_link_fill(skb, dev) < 0)
1792 			goto nla_put_failure;
1793 	}
1794 
1795 	if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1796 		goto nla_put_failure;
1797 
1798 	if (new_nsid &&
1799 	    nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1800 		goto nla_put_failure;
1801 	if (new_ifindex &&
1802 	    nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1803 		goto nla_put_failure;
1804 
1805 	if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1806 	    nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1807 		goto nla_put_failure;
1808 
1809 	rcu_read_lock();
1810 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1811 		goto nla_put_failure_rcu;
1812 	rcu_read_unlock();
1813 
1814 	if (rtnl_fill_prop_list(skb, dev))
1815 		goto nla_put_failure;
1816 
1817 	if (dev->dev.parent &&
1818 	    nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1819 			   dev_name(dev->dev.parent)))
1820 		goto nla_put_failure;
1821 
1822 	if (dev->dev.parent && dev->dev.parent->bus &&
1823 	    nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1824 			   dev->dev.parent->bus->name))
1825 		goto nla_put_failure;
1826 
1827 	nlmsg_end(skb, nlh);
1828 	return 0;
1829 
1830 nla_put_failure_rcu:
1831 	rcu_read_unlock();
1832 nla_put_failure:
1833 	nlmsg_cancel(skb, nlh);
1834 	return -EMSGSIZE;
1835 }
1836 
1837 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1838 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1839 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1840 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1841 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1842 	[IFLA_MTU]		= { .type = NLA_U32 },
1843 	[IFLA_LINK]		= { .type = NLA_U32 },
1844 	[IFLA_MASTER]		= { .type = NLA_U32 },
1845 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1846 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1847 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1848 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1849 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1850 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1851 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1852 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1853 	/* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1854 	 * allow 0-length string (needed to remove an alias).
1855 	 */
1856 	[IFLA_IFALIAS]	        = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1857 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1858 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1859 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1860 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1861 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1862 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1863 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1864 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1865 	[IFLA_GSO_MAX_SEGS]	= { .type = NLA_U32 },
1866 	[IFLA_GSO_MAX_SIZE]	= { .type = NLA_U32 },
1867 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1868 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1869 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1870 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1871 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1872 	[IFLA_XDP]		= { .type = NLA_NESTED },
1873 	[IFLA_EVENT]		= { .type = NLA_U32 },
1874 	[IFLA_GROUP]		= { .type = NLA_U32 },
1875 	[IFLA_TARGET_NETNSID]	= { .type = NLA_S32 },
1876 	[IFLA_CARRIER_UP_COUNT]	= { .type = NLA_U32 },
1877 	[IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1878 	[IFLA_MIN_MTU]		= { .type = NLA_U32 },
1879 	[IFLA_MAX_MTU]		= { .type = NLA_U32 },
1880 	[IFLA_PROP_LIST]	= { .type = NLA_NESTED },
1881 	[IFLA_ALT_IFNAME]	= { .type = NLA_STRING,
1882 				    .len = ALTIFNAMSIZ - 1 },
1883 	[IFLA_PERM_ADDRESS]	= { .type = NLA_REJECT },
1884 	[IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1885 	[IFLA_NEW_IFINDEX]	= NLA_POLICY_MIN(NLA_S32, 1),
1886 	[IFLA_PARENT_DEV_NAME]	= { .type = NLA_NUL_STRING },
1887 	[IFLA_GRO_MAX_SIZE]	= { .type = NLA_U32 },
1888 };
1889 
1890 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1891 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1892 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1893 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1894 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1895 };
1896 
1897 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1898 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1899 	[IFLA_VF_BROADCAST]	= { .type = NLA_REJECT },
1900 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1901 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1902 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1903 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1904 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1905 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1906 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1907 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1908 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1909 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1910 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1911 };
1912 
1913 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1914 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1915 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1916 				    .len = PORT_PROFILE_MAX },
1917 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1918 				      .len = PORT_UUID_MAX },
1919 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1920 				    .len = PORT_UUID_MAX },
1921 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1922 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1923 
1924 	/* Unused, but we need to keep it here since user space could
1925 	 * fill it. It's also broken with regard to NLA_BINARY use in
1926 	 * combination with structs.
1927 	 */
1928 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1929 				    .len = sizeof(struct ifla_port_vsi) },
1930 };
1931 
1932 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1933 	[IFLA_XDP_UNSPEC]	= { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1934 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1935 	[IFLA_XDP_EXPECTED_FD]	= { .type = NLA_S32 },
1936 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1937 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1938 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1939 };
1940 
1941 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1942 {
1943 	const struct rtnl_link_ops *ops = NULL;
1944 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1945 
1946 	if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1947 		return NULL;
1948 
1949 	if (linfo[IFLA_INFO_KIND]) {
1950 		char kind[MODULE_NAME_LEN];
1951 
1952 		nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1953 		ops = rtnl_link_ops_get(kind);
1954 	}
1955 
1956 	return ops;
1957 }
1958 
1959 static bool link_master_filtered(struct net_device *dev, int master_idx)
1960 {
1961 	struct net_device *master;
1962 
1963 	if (!master_idx)
1964 		return false;
1965 
1966 	master = netdev_master_upper_dev_get(dev);
1967 
1968 	/* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
1969 	 * another invalid value for ifindex to denote "no master".
1970 	 */
1971 	if (master_idx == -1)
1972 		return !!master;
1973 
1974 	if (!master || master->ifindex != master_idx)
1975 		return true;
1976 
1977 	return false;
1978 }
1979 
1980 static bool link_kind_filtered(const struct net_device *dev,
1981 			       const struct rtnl_link_ops *kind_ops)
1982 {
1983 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1984 		return true;
1985 
1986 	return false;
1987 }
1988 
1989 static bool link_dump_filtered(struct net_device *dev,
1990 			       int master_idx,
1991 			       const struct rtnl_link_ops *kind_ops)
1992 {
1993 	if (link_master_filtered(dev, master_idx) ||
1994 	    link_kind_filtered(dev, kind_ops))
1995 		return true;
1996 
1997 	return false;
1998 }
1999 
2000 /**
2001  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2002  * @sk: netlink socket
2003  * @netnsid: network namespace identifier
2004  *
2005  * Returns the network namespace identified by netnsid on success or an error
2006  * pointer on failure.
2007  */
2008 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2009 {
2010 	struct net *net;
2011 
2012 	net = get_net_ns_by_id(sock_net(sk), netnsid);
2013 	if (!net)
2014 		return ERR_PTR(-EINVAL);
2015 
2016 	/* For now, the caller is required to have CAP_NET_ADMIN in
2017 	 * the user namespace owning the target net ns.
2018 	 */
2019 	if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2020 		put_net(net);
2021 		return ERR_PTR(-EACCES);
2022 	}
2023 	return net;
2024 }
2025 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2026 
2027 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2028 				      bool strict_check, struct nlattr **tb,
2029 				      struct netlink_ext_ack *extack)
2030 {
2031 	int hdrlen;
2032 
2033 	if (strict_check) {
2034 		struct ifinfomsg *ifm;
2035 
2036 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2037 			NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2038 			return -EINVAL;
2039 		}
2040 
2041 		ifm = nlmsg_data(nlh);
2042 		if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2043 		    ifm->ifi_change) {
2044 			NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2045 			return -EINVAL;
2046 		}
2047 		if (ifm->ifi_index) {
2048 			NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2049 			return -EINVAL;
2050 		}
2051 
2052 		return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2053 						     IFLA_MAX, ifla_policy,
2054 						     extack);
2055 	}
2056 
2057 	/* A hack to preserve kernel<->userspace interface.
2058 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2059 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2060 	 * what iproute2 < v3.9.0 used.
2061 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2062 	 * attribute, its netlink message is shorter than struct ifinfomsg.
2063 	 */
2064 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2065 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2066 
2067 	return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2068 				      extack);
2069 }
2070 
2071 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2072 {
2073 	struct netlink_ext_ack *extack = cb->extack;
2074 	const struct nlmsghdr *nlh = cb->nlh;
2075 	struct net *net = sock_net(skb->sk);
2076 	struct net *tgt_net = net;
2077 	int h, s_h;
2078 	int idx = 0, s_idx;
2079 	struct net_device *dev;
2080 	struct hlist_head *head;
2081 	struct nlattr *tb[IFLA_MAX+1];
2082 	u32 ext_filter_mask = 0;
2083 	const struct rtnl_link_ops *kind_ops = NULL;
2084 	unsigned int flags = NLM_F_MULTI;
2085 	int master_idx = 0;
2086 	int netnsid = -1;
2087 	int err, i;
2088 
2089 	s_h = cb->args[0];
2090 	s_idx = cb->args[1];
2091 
2092 	err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2093 	if (err < 0) {
2094 		if (cb->strict_check)
2095 			return err;
2096 
2097 		goto walk_entries;
2098 	}
2099 
2100 	for (i = 0; i <= IFLA_MAX; ++i) {
2101 		if (!tb[i])
2102 			continue;
2103 
2104 		/* new attributes should only be added with strict checking */
2105 		switch (i) {
2106 		case IFLA_TARGET_NETNSID:
2107 			netnsid = nla_get_s32(tb[i]);
2108 			tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2109 			if (IS_ERR(tgt_net)) {
2110 				NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2111 				return PTR_ERR(tgt_net);
2112 			}
2113 			break;
2114 		case IFLA_EXT_MASK:
2115 			ext_filter_mask = nla_get_u32(tb[i]);
2116 			break;
2117 		case IFLA_MASTER:
2118 			master_idx = nla_get_u32(tb[i]);
2119 			break;
2120 		case IFLA_LINKINFO:
2121 			kind_ops = linkinfo_to_kind_ops(tb[i]);
2122 			break;
2123 		default:
2124 			if (cb->strict_check) {
2125 				NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2126 				return -EINVAL;
2127 			}
2128 		}
2129 	}
2130 
2131 	if (master_idx || kind_ops)
2132 		flags |= NLM_F_DUMP_FILTERED;
2133 
2134 walk_entries:
2135 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2136 		idx = 0;
2137 		head = &tgt_net->dev_index_head[h];
2138 		hlist_for_each_entry(dev, head, index_hlist) {
2139 			if (link_dump_filtered(dev, master_idx, kind_ops))
2140 				goto cont;
2141 			if (idx < s_idx)
2142 				goto cont;
2143 			err = rtnl_fill_ifinfo(skb, dev, net,
2144 					       RTM_NEWLINK,
2145 					       NETLINK_CB(cb->skb).portid,
2146 					       nlh->nlmsg_seq, 0, flags,
2147 					       ext_filter_mask, 0, NULL, 0,
2148 					       netnsid, GFP_KERNEL);
2149 
2150 			if (err < 0) {
2151 				if (likely(skb->len))
2152 					goto out;
2153 
2154 				goto out_err;
2155 			}
2156 cont:
2157 			idx++;
2158 		}
2159 	}
2160 out:
2161 	err = skb->len;
2162 out_err:
2163 	cb->args[1] = idx;
2164 	cb->args[0] = h;
2165 	cb->seq = tgt_net->dev_base_seq;
2166 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2167 	if (netnsid >= 0)
2168 		put_net(tgt_net);
2169 
2170 	return err;
2171 }
2172 
2173 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2174 			struct netlink_ext_ack *exterr)
2175 {
2176 	return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2177 				    exterr);
2178 }
2179 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2180 
2181 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2182 {
2183 	struct net *net;
2184 	/* Examine the link attributes and figure out which
2185 	 * network namespace we are talking about.
2186 	 */
2187 	if (tb[IFLA_NET_NS_PID])
2188 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2189 	else if (tb[IFLA_NET_NS_FD])
2190 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2191 	else
2192 		net = get_net(src_net);
2193 	return net;
2194 }
2195 EXPORT_SYMBOL(rtnl_link_get_net);
2196 
2197 /* Figure out which network namespace we are talking about by
2198  * examining the link attributes in the following order:
2199  *
2200  * 1. IFLA_NET_NS_PID
2201  * 2. IFLA_NET_NS_FD
2202  * 3. IFLA_TARGET_NETNSID
2203  */
2204 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2205 					       struct nlattr *tb[])
2206 {
2207 	struct net *net;
2208 
2209 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2210 		return rtnl_link_get_net(src_net, tb);
2211 
2212 	if (!tb[IFLA_TARGET_NETNSID])
2213 		return get_net(src_net);
2214 
2215 	net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2216 	if (!net)
2217 		return ERR_PTR(-EINVAL);
2218 
2219 	return net;
2220 }
2221 
2222 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2223 					     struct net *src_net,
2224 					     struct nlattr *tb[], int cap)
2225 {
2226 	struct net *net;
2227 
2228 	net = rtnl_link_get_net_by_nlattr(src_net, tb);
2229 	if (IS_ERR(net))
2230 		return net;
2231 
2232 	if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2233 		put_net(net);
2234 		return ERR_PTR(-EPERM);
2235 	}
2236 
2237 	return net;
2238 }
2239 
2240 /* Verify that rtnetlink requests do not pass additional properties
2241  * potentially referring to different network namespaces.
2242  */
2243 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2244 				    struct netlink_ext_ack *extack,
2245 				    bool netns_id_only)
2246 {
2247 
2248 	if (netns_id_only) {
2249 		if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2250 			return 0;
2251 
2252 		NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2253 		return -EOPNOTSUPP;
2254 	}
2255 
2256 	if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2257 		goto invalid_attr;
2258 
2259 	if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2260 		goto invalid_attr;
2261 
2262 	if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2263 		goto invalid_attr;
2264 
2265 	return 0;
2266 
2267 invalid_attr:
2268 	NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2269 	return -EINVAL;
2270 }
2271 
2272 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2273 			    struct netlink_ext_ack *extack)
2274 {
2275 	if (dev) {
2276 		if (tb[IFLA_ADDRESS] &&
2277 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2278 			return -EINVAL;
2279 
2280 		if (tb[IFLA_BROADCAST] &&
2281 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2282 			return -EINVAL;
2283 	}
2284 
2285 	if (tb[IFLA_AF_SPEC]) {
2286 		struct nlattr *af;
2287 		int rem, err;
2288 
2289 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2290 			const struct rtnl_af_ops *af_ops;
2291 
2292 			af_ops = rtnl_af_lookup(nla_type(af));
2293 			if (!af_ops)
2294 				return -EAFNOSUPPORT;
2295 
2296 			if (!af_ops->set_link_af)
2297 				return -EOPNOTSUPP;
2298 
2299 			if (af_ops->validate_link_af) {
2300 				err = af_ops->validate_link_af(dev, af, extack);
2301 				if (err < 0)
2302 					return err;
2303 			}
2304 		}
2305 	}
2306 
2307 	if (tb[IFLA_GRO_MAX_SIZE]) {
2308 		u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2309 
2310 		if (gro_max_size > GRO_MAX_SIZE) {
2311 			NL_SET_ERR_MSG(extack, "too big gro_max_size");
2312 			return -EINVAL;
2313 		}
2314 	}
2315 	return 0;
2316 }
2317 
2318 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2319 				  int guid_type)
2320 {
2321 	const struct net_device_ops *ops = dev->netdev_ops;
2322 
2323 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2324 }
2325 
2326 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2327 {
2328 	if (dev->type != ARPHRD_INFINIBAND)
2329 		return -EOPNOTSUPP;
2330 
2331 	return handle_infiniband_guid(dev, ivt, guid_type);
2332 }
2333 
2334 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2335 {
2336 	const struct net_device_ops *ops = dev->netdev_ops;
2337 	int err = -EINVAL;
2338 
2339 	if (tb[IFLA_VF_MAC]) {
2340 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2341 
2342 		if (ivm->vf >= INT_MAX)
2343 			return -EINVAL;
2344 		err = -EOPNOTSUPP;
2345 		if (ops->ndo_set_vf_mac)
2346 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
2347 						  ivm->mac);
2348 		if (err < 0)
2349 			return err;
2350 	}
2351 
2352 	if (tb[IFLA_VF_VLAN]) {
2353 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2354 
2355 		if (ivv->vf >= INT_MAX)
2356 			return -EINVAL;
2357 		err = -EOPNOTSUPP;
2358 		if (ops->ndo_set_vf_vlan)
2359 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2360 						   ivv->qos,
2361 						   htons(ETH_P_8021Q));
2362 		if (err < 0)
2363 			return err;
2364 	}
2365 
2366 	if (tb[IFLA_VF_VLAN_LIST]) {
2367 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2368 		struct nlattr *attr;
2369 		int rem, len = 0;
2370 
2371 		err = -EOPNOTSUPP;
2372 		if (!ops->ndo_set_vf_vlan)
2373 			return err;
2374 
2375 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2376 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2377 			    nla_len(attr) < NLA_HDRLEN) {
2378 				return -EINVAL;
2379 			}
2380 			if (len >= MAX_VLAN_LIST_LEN)
2381 				return -EOPNOTSUPP;
2382 			ivvl[len] = nla_data(attr);
2383 
2384 			len++;
2385 		}
2386 		if (len == 0)
2387 			return -EINVAL;
2388 
2389 		if (ivvl[0]->vf >= INT_MAX)
2390 			return -EINVAL;
2391 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2392 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
2393 		if (err < 0)
2394 			return err;
2395 	}
2396 
2397 	if (tb[IFLA_VF_TX_RATE]) {
2398 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2399 		struct ifla_vf_info ivf;
2400 
2401 		if (ivt->vf >= INT_MAX)
2402 			return -EINVAL;
2403 		err = -EOPNOTSUPP;
2404 		if (ops->ndo_get_vf_config)
2405 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2406 		if (err < 0)
2407 			return err;
2408 
2409 		err = -EOPNOTSUPP;
2410 		if (ops->ndo_set_vf_rate)
2411 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2412 						   ivf.min_tx_rate,
2413 						   ivt->rate);
2414 		if (err < 0)
2415 			return err;
2416 	}
2417 
2418 	if (tb[IFLA_VF_RATE]) {
2419 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2420 
2421 		if (ivt->vf >= INT_MAX)
2422 			return -EINVAL;
2423 		err = -EOPNOTSUPP;
2424 		if (ops->ndo_set_vf_rate)
2425 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2426 						   ivt->min_tx_rate,
2427 						   ivt->max_tx_rate);
2428 		if (err < 0)
2429 			return err;
2430 	}
2431 
2432 	if (tb[IFLA_VF_SPOOFCHK]) {
2433 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2434 
2435 		if (ivs->vf >= INT_MAX)
2436 			return -EINVAL;
2437 		err = -EOPNOTSUPP;
2438 		if (ops->ndo_set_vf_spoofchk)
2439 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2440 						       ivs->setting);
2441 		if (err < 0)
2442 			return err;
2443 	}
2444 
2445 	if (tb[IFLA_VF_LINK_STATE]) {
2446 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2447 
2448 		if (ivl->vf >= INT_MAX)
2449 			return -EINVAL;
2450 		err = -EOPNOTSUPP;
2451 		if (ops->ndo_set_vf_link_state)
2452 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2453 							 ivl->link_state);
2454 		if (err < 0)
2455 			return err;
2456 	}
2457 
2458 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
2459 		struct ifla_vf_rss_query_en *ivrssq_en;
2460 
2461 		err = -EOPNOTSUPP;
2462 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2463 		if (ivrssq_en->vf >= INT_MAX)
2464 			return -EINVAL;
2465 		if (ops->ndo_set_vf_rss_query_en)
2466 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2467 							   ivrssq_en->setting);
2468 		if (err < 0)
2469 			return err;
2470 	}
2471 
2472 	if (tb[IFLA_VF_TRUST]) {
2473 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2474 
2475 		if (ivt->vf >= INT_MAX)
2476 			return -EINVAL;
2477 		err = -EOPNOTSUPP;
2478 		if (ops->ndo_set_vf_trust)
2479 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2480 		if (err < 0)
2481 			return err;
2482 	}
2483 
2484 	if (tb[IFLA_VF_IB_NODE_GUID]) {
2485 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2486 
2487 		if (ivt->vf >= INT_MAX)
2488 			return -EINVAL;
2489 		if (!ops->ndo_set_vf_guid)
2490 			return -EOPNOTSUPP;
2491 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2492 	}
2493 
2494 	if (tb[IFLA_VF_IB_PORT_GUID]) {
2495 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2496 
2497 		if (ivt->vf >= INT_MAX)
2498 			return -EINVAL;
2499 		if (!ops->ndo_set_vf_guid)
2500 			return -EOPNOTSUPP;
2501 
2502 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2503 	}
2504 
2505 	return err;
2506 }
2507 
2508 static int do_set_master(struct net_device *dev, int ifindex,
2509 			 struct netlink_ext_ack *extack)
2510 {
2511 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2512 	const struct net_device_ops *ops;
2513 	int err;
2514 
2515 	if (upper_dev) {
2516 		if (upper_dev->ifindex == ifindex)
2517 			return 0;
2518 		ops = upper_dev->netdev_ops;
2519 		if (ops->ndo_del_slave) {
2520 			err = ops->ndo_del_slave(upper_dev, dev);
2521 			if (err)
2522 				return err;
2523 		} else {
2524 			return -EOPNOTSUPP;
2525 		}
2526 	}
2527 
2528 	if (ifindex) {
2529 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2530 		if (!upper_dev)
2531 			return -EINVAL;
2532 		ops = upper_dev->netdev_ops;
2533 		if (ops->ndo_add_slave) {
2534 			err = ops->ndo_add_slave(upper_dev, dev, extack);
2535 			if (err)
2536 				return err;
2537 		} else {
2538 			return -EOPNOTSUPP;
2539 		}
2540 	}
2541 	return 0;
2542 }
2543 
2544 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2545 	[IFLA_PROTO_DOWN_REASON_MASK]	= { .type = NLA_U32 },
2546 	[IFLA_PROTO_DOWN_REASON_VALUE]	= { .type = NLA_U32 },
2547 };
2548 
2549 static int do_set_proto_down(struct net_device *dev,
2550 			     struct nlattr *nl_proto_down,
2551 			     struct nlattr *nl_proto_down_reason,
2552 			     struct netlink_ext_ack *extack)
2553 {
2554 	struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2555 	unsigned long mask = 0;
2556 	u32 value;
2557 	bool proto_down;
2558 	int err;
2559 
2560 	if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2561 		NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2562 		return -EOPNOTSUPP;
2563 	}
2564 
2565 	if (nl_proto_down_reason) {
2566 		err = nla_parse_nested_deprecated(pdreason,
2567 						  IFLA_PROTO_DOWN_REASON_MAX,
2568 						  nl_proto_down_reason,
2569 						  ifla_proto_down_reason_policy,
2570 						  NULL);
2571 		if (err < 0)
2572 			return err;
2573 
2574 		if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2575 			NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2576 			return -EINVAL;
2577 		}
2578 
2579 		value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2580 
2581 		if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2582 			mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2583 
2584 		dev_change_proto_down_reason(dev, mask, value);
2585 	}
2586 
2587 	if (nl_proto_down) {
2588 		proto_down = nla_get_u8(nl_proto_down);
2589 
2590 		/* Don't turn off protodown if there are active reasons */
2591 		if (!proto_down && dev->proto_down_reason) {
2592 			NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2593 			return -EBUSY;
2594 		}
2595 		err = dev_change_proto_down(dev,
2596 					    proto_down);
2597 		if (err)
2598 			return err;
2599 	}
2600 
2601 	return 0;
2602 }
2603 
2604 #define DO_SETLINK_MODIFIED	0x01
2605 /* notify flag means notify + modified. */
2606 #define DO_SETLINK_NOTIFY	0x03
2607 static int do_setlink(const struct sk_buff *skb,
2608 		      struct net_device *dev, struct ifinfomsg *ifm,
2609 		      struct netlink_ext_ack *extack,
2610 		      struct nlattr **tb, char *ifname, int status)
2611 {
2612 	const struct net_device_ops *ops = dev->netdev_ops;
2613 	int err;
2614 
2615 	err = validate_linkmsg(dev, tb, extack);
2616 	if (err < 0)
2617 		return err;
2618 
2619 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2620 		const char *pat = ifname && ifname[0] ? ifname : NULL;
2621 		struct net *net;
2622 		int new_ifindex;
2623 
2624 		net = rtnl_link_get_net_capable(skb, dev_net(dev),
2625 						tb, CAP_NET_ADMIN);
2626 		if (IS_ERR(net)) {
2627 			err = PTR_ERR(net);
2628 			goto errout;
2629 		}
2630 
2631 		if (tb[IFLA_NEW_IFINDEX])
2632 			new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2633 		else
2634 			new_ifindex = 0;
2635 
2636 		err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2637 		put_net(net);
2638 		if (err)
2639 			goto errout;
2640 		status |= DO_SETLINK_MODIFIED;
2641 	}
2642 
2643 	if (tb[IFLA_MAP]) {
2644 		struct rtnl_link_ifmap *u_map;
2645 		struct ifmap k_map;
2646 
2647 		if (!ops->ndo_set_config) {
2648 			err = -EOPNOTSUPP;
2649 			goto errout;
2650 		}
2651 
2652 		if (!netif_device_present(dev)) {
2653 			err = -ENODEV;
2654 			goto errout;
2655 		}
2656 
2657 		u_map = nla_data(tb[IFLA_MAP]);
2658 		k_map.mem_start = (unsigned long) u_map->mem_start;
2659 		k_map.mem_end = (unsigned long) u_map->mem_end;
2660 		k_map.base_addr = (unsigned short) u_map->base_addr;
2661 		k_map.irq = (unsigned char) u_map->irq;
2662 		k_map.dma = (unsigned char) u_map->dma;
2663 		k_map.port = (unsigned char) u_map->port;
2664 
2665 		err = ops->ndo_set_config(dev, &k_map);
2666 		if (err < 0)
2667 			goto errout;
2668 
2669 		status |= DO_SETLINK_NOTIFY;
2670 	}
2671 
2672 	if (tb[IFLA_ADDRESS]) {
2673 		struct sockaddr *sa;
2674 		int len;
2675 
2676 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2677 						  sizeof(*sa));
2678 		sa = kmalloc(len, GFP_KERNEL);
2679 		if (!sa) {
2680 			err = -ENOMEM;
2681 			goto errout;
2682 		}
2683 		sa->sa_family = dev->type;
2684 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2685 		       dev->addr_len);
2686 		err = dev_set_mac_address_user(dev, sa, extack);
2687 		kfree(sa);
2688 		if (err)
2689 			goto errout;
2690 		status |= DO_SETLINK_MODIFIED;
2691 	}
2692 
2693 	if (tb[IFLA_MTU]) {
2694 		err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2695 		if (err < 0)
2696 			goto errout;
2697 		status |= DO_SETLINK_MODIFIED;
2698 	}
2699 
2700 	if (tb[IFLA_GROUP]) {
2701 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2702 		status |= DO_SETLINK_NOTIFY;
2703 	}
2704 
2705 	/*
2706 	 * Interface selected by interface index but interface
2707 	 * name provided implies that a name change has been
2708 	 * requested.
2709 	 */
2710 	if (ifm->ifi_index > 0 && ifname[0]) {
2711 		err = dev_change_name(dev, ifname);
2712 		if (err < 0)
2713 			goto errout;
2714 		status |= DO_SETLINK_MODIFIED;
2715 	}
2716 
2717 	if (tb[IFLA_IFALIAS]) {
2718 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2719 				    nla_len(tb[IFLA_IFALIAS]));
2720 		if (err < 0)
2721 			goto errout;
2722 		status |= DO_SETLINK_NOTIFY;
2723 	}
2724 
2725 	if (tb[IFLA_BROADCAST]) {
2726 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2727 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2728 	}
2729 
2730 	if (ifm->ifi_flags || ifm->ifi_change) {
2731 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2732 				       extack);
2733 		if (err < 0)
2734 			goto errout;
2735 	}
2736 
2737 	if (tb[IFLA_MASTER]) {
2738 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2739 		if (err)
2740 			goto errout;
2741 		status |= DO_SETLINK_MODIFIED;
2742 	}
2743 
2744 	if (tb[IFLA_CARRIER]) {
2745 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2746 		if (err)
2747 			goto errout;
2748 		status |= DO_SETLINK_MODIFIED;
2749 	}
2750 
2751 	if (tb[IFLA_TXQLEN]) {
2752 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2753 
2754 		err = dev_change_tx_queue_len(dev, value);
2755 		if (err)
2756 			goto errout;
2757 		status |= DO_SETLINK_MODIFIED;
2758 	}
2759 
2760 	if (tb[IFLA_GSO_MAX_SIZE]) {
2761 		u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2762 
2763 		if (max_size > GSO_MAX_SIZE) {
2764 			err = -EINVAL;
2765 			goto errout;
2766 		}
2767 
2768 		if (dev->gso_max_size ^ max_size) {
2769 			netif_set_gso_max_size(dev, max_size);
2770 			status |= DO_SETLINK_MODIFIED;
2771 		}
2772 	}
2773 
2774 	if (tb[IFLA_GSO_MAX_SEGS]) {
2775 		u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2776 
2777 		if (max_segs > GSO_MAX_SEGS) {
2778 			err = -EINVAL;
2779 			goto errout;
2780 		}
2781 
2782 		if (dev->gso_max_segs ^ max_segs) {
2783 			netif_set_gso_max_segs(dev, max_segs);
2784 			status |= DO_SETLINK_MODIFIED;
2785 		}
2786 	}
2787 
2788 	if (tb[IFLA_GRO_MAX_SIZE]) {
2789 		u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2790 
2791 		if (dev->gro_max_size ^ gro_max_size) {
2792 			netif_set_gro_max_size(dev, gro_max_size);
2793 			status |= DO_SETLINK_MODIFIED;
2794 		}
2795 	}
2796 
2797 	if (tb[IFLA_OPERSTATE])
2798 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2799 
2800 	if (tb[IFLA_LINKMODE]) {
2801 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2802 
2803 		write_lock(&dev_base_lock);
2804 		if (dev->link_mode ^ value)
2805 			status |= DO_SETLINK_NOTIFY;
2806 		dev->link_mode = value;
2807 		write_unlock(&dev_base_lock);
2808 	}
2809 
2810 	if (tb[IFLA_VFINFO_LIST]) {
2811 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2812 		struct nlattr *attr;
2813 		int rem;
2814 
2815 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2816 			if (nla_type(attr) != IFLA_VF_INFO ||
2817 			    nla_len(attr) < NLA_HDRLEN) {
2818 				err = -EINVAL;
2819 				goto errout;
2820 			}
2821 			err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2822 							  attr,
2823 							  ifla_vf_policy,
2824 							  NULL);
2825 			if (err < 0)
2826 				goto errout;
2827 			err = do_setvfinfo(dev, vfinfo);
2828 			if (err < 0)
2829 				goto errout;
2830 			status |= DO_SETLINK_NOTIFY;
2831 		}
2832 	}
2833 	err = 0;
2834 
2835 	if (tb[IFLA_VF_PORTS]) {
2836 		struct nlattr *port[IFLA_PORT_MAX+1];
2837 		struct nlattr *attr;
2838 		int vf;
2839 		int rem;
2840 
2841 		err = -EOPNOTSUPP;
2842 		if (!ops->ndo_set_vf_port)
2843 			goto errout;
2844 
2845 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2846 			if (nla_type(attr) != IFLA_VF_PORT ||
2847 			    nla_len(attr) < NLA_HDRLEN) {
2848 				err = -EINVAL;
2849 				goto errout;
2850 			}
2851 			err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2852 							  attr,
2853 							  ifla_port_policy,
2854 							  NULL);
2855 			if (err < 0)
2856 				goto errout;
2857 			if (!port[IFLA_PORT_VF]) {
2858 				err = -EOPNOTSUPP;
2859 				goto errout;
2860 			}
2861 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2862 			err = ops->ndo_set_vf_port(dev, vf, port);
2863 			if (err < 0)
2864 				goto errout;
2865 			status |= DO_SETLINK_NOTIFY;
2866 		}
2867 	}
2868 	err = 0;
2869 
2870 	if (tb[IFLA_PORT_SELF]) {
2871 		struct nlattr *port[IFLA_PORT_MAX+1];
2872 
2873 		err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2874 						  tb[IFLA_PORT_SELF],
2875 						  ifla_port_policy, NULL);
2876 		if (err < 0)
2877 			goto errout;
2878 
2879 		err = -EOPNOTSUPP;
2880 		if (ops->ndo_set_vf_port)
2881 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2882 		if (err < 0)
2883 			goto errout;
2884 		status |= DO_SETLINK_NOTIFY;
2885 	}
2886 
2887 	if (tb[IFLA_AF_SPEC]) {
2888 		struct nlattr *af;
2889 		int rem;
2890 
2891 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2892 			const struct rtnl_af_ops *af_ops;
2893 
2894 			BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2895 
2896 			err = af_ops->set_link_af(dev, af, extack);
2897 			if (err < 0)
2898 				goto errout;
2899 
2900 			status |= DO_SETLINK_NOTIFY;
2901 		}
2902 	}
2903 	err = 0;
2904 
2905 	if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2906 		err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2907 					tb[IFLA_PROTO_DOWN_REASON], extack);
2908 		if (err)
2909 			goto errout;
2910 		status |= DO_SETLINK_NOTIFY;
2911 	}
2912 
2913 	if (tb[IFLA_XDP]) {
2914 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2915 		u32 xdp_flags = 0;
2916 
2917 		err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2918 						  tb[IFLA_XDP],
2919 						  ifla_xdp_policy, NULL);
2920 		if (err < 0)
2921 			goto errout;
2922 
2923 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2924 			err = -EINVAL;
2925 			goto errout;
2926 		}
2927 
2928 		if (xdp[IFLA_XDP_FLAGS]) {
2929 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2930 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2931 				err = -EINVAL;
2932 				goto errout;
2933 			}
2934 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2935 				err = -EINVAL;
2936 				goto errout;
2937 			}
2938 		}
2939 
2940 		if (xdp[IFLA_XDP_FD]) {
2941 			int expected_fd = -1;
2942 
2943 			if (xdp_flags & XDP_FLAGS_REPLACE) {
2944 				if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2945 					err = -EINVAL;
2946 					goto errout;
2947 				}
2948 				expected_fd =
2949 					nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2950 			}
2951 
2952 			err = dev_change_xdp_fd(dev, extack,
2953 						nla_get_s32(xdp[IFLA_XDP_FD]),
2954 						expected_fd,
2955 						xdp_flags);
2956 			if (err)
2957 				goto errout;
2958 			status |= DO_SETLINK_NOTIFY;
2959 		}
2960 	}
2961 
2962 errout:
2963 	if (status & DO_SETLINK_MODIFIED) {
2964 		if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2965 			netdev_state_change(dev);
2966 
2967 		if (err < 0)
2968 			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",
2969 					     dev->name);
2970 	}
2971 
2972 	return err;
2973 }
2974 
2975 static struct net_device *rtnl_dev_get(struct net *net,
2976 				       struct nlattr *ifname_attr,
2977 				       struct nlattr *altifname_attr,
2978 				       char *ifname)
2979 {
2980 	char buffer[ALTIFNAMSIZ];
2981 
2982 	if (!ifname) {
2983 		ifname = buffer;
2984 		if (ifname_attr)
2985 			nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2986 		else if (altifname_attr)
2987 			nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2988 		else
2989 			return NULL;
2990 	}
2991 
2992 	return __dev_get_by_name(net, ifname);
2993 }
2994 
2995 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2996 			struct netlink_ext_ack *extack)
2997 {
2998 	struct net *net = sock_net(skb->sk);
2999 	struct ifinfomsg *ifm;
3000 	struct net_device *dev;
3001 	int err;
3002 	struct nlattr *tb[IFLA_MAX+1];
3003 	char ifname[IFNAMSIZ];
3004 
3005 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3006 				     ifla_policy, extack);
3007 	if (err < 0)
3008 		goto errout;
3009 
3010 	err = rtnl_ensure_unique_netns(tb, extack, false);
3011 	if (err < 0)
3012 		goto errout;
3013 
3014 	if (tb[IFLA_IFNAME])
3015 		nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3016 	else
3017 		ifname[0] = '\0';
3018 
3019 	err = -EINVAL;
3020 	ifm = nlmsg_data(nlh);
3021 	if (ifm->ifi_index > 0)
3022 		dev = __dev_get_by_index(net, ifm->ifi_index);
3023 	else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3024 		dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3025 	else
3026 		goto errout;
3027 
3028 	if (dev == NULL) {
3029 		err = -ENODEV;
3030 		goto errout;
3031 	}
3032 
3033 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3034 errout:
3035 	return err;
3036 }
3037 
3038 static int rtnl_group_dellink(const struct net *net, int group)
3039 {
3040 	struct net_device *dev, *aux;
3041 	LIST_HEAD(list_kill);
3042 	bool found = false;
3043 
3044 	if (!group)
3045 		return -EPERM;
3046 
3047 	for_each_netdev(net, dev) {
3048 		if (dev->group == group) {
3049 			const struct rtnl_link_ops *ops;
3050 
3051 			found = true;
3052 			ops = dev->rtnl_link_ops;
3053 			if (!ops || !ops->dellink)
3054 				return -EOPNOTSUPP;
3055 		}
3056 	}
3057 
3058 	if (!found)
3059 		return -ENODEV;
3060 
3061 	for_each_netdev_safe(net, dev, aux) {
3062 		if (dev->group == group) {
3063 			const struct rtnl_link_ops *ops;
3064 
3065 			ops = dev->rtnl_link_ops;
3066 			ops->dellink(dev, &list_kill);
3067 		}
3068 	}
3069 	unregister_netdevice_many(&list_kill);
3070 
3071 	return 0;
3072 }
3073 
3074 int rtnl_delete_link(struct net_device *dev)
3075 {
3076 	const struct rtnl_link_ops *ops;
3077 	LIST_HEAD(list_kill);
3078 
3079 	ops = dev->rtnl_link_ops;
3080 	if (!ops || !ops->dellink)
3081 		return -EOPNOTSUPP;
3082 
3083 	ops->dellink(dev, &list_kill);
3084 	unregister_netdevice_many(&list_kill);
3085 
3086 	return 0;
3087 }
3088 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3089 
3090 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3091 			struct netlink_ext_ack *extack)
3092 {
3093 	struct net *net = sock_net(skb->sk);
3094 	struct net *tgt_net = net;
3095 	struct net_device *dev = NULL;
3096 	struct ifinfomsg *ifm;
3097 	struct nlattr *tb[IFLA_MAX+1];
3098 	int err;
3099 	int netnsid = -1;
3100 
3101 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3102 				     ifla_policy, extack);
3103 	if (err < 0)
3104 		return err;
3105 
3106 	err = rtnl_ensure_unique_netns(tb, extack, true);
3107 	if (err < 0)
3108 		return err;
3109 
3110 	if (tb[IFLA_TARGET_NETNSID]) {
3111 		netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3112 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3113 		if (IS_ERR(tgt_net))
3114 			return PTR_ERR(tgt_net);
3115 	}
3116 
3117 	err = -EINVAL;
3118 	ifm = nlmsg_data(nlh);
3119 	if (ifm->ifi_index > 0)
3120 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3121 	else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3122 		dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3123 				   tb[IFLA_ALT_IFNAME], NULL);
3124 	else if (tb[IFLA_GROUP])
3125 		err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3126 	else
3127 		goto out;
3128 
3129 	if (!dev) {
3130 		if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3131 			err = -ENODEV;
3132 
3133 		goto out;
3134 	}
3135 
3136 	err = rtnl_delete_link(dev);
3137 
3138 out:
3139 	if (netnsid >= 0)
3140 		put_net(tgt_net);
3141 
3142 	return err;
3143 }
3144 
3145 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3146 {
3147 	unsigned int old_flags;
3148 	int err;
3149 
3150 	old_flags = dev->flags;
3151 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3152 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3153 					 NULL);
3154 		if (err < 0)
3155 			return err;
3156 	}
3157 
3158 	if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3159 		__dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3160 	} else {
3161 		dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3162 		__dev_notify_flags(dev, old_flags, ~0U);
3163 	}
3164 	return 0;
3165 }
3166 EXPORT_SYMBOL(rtnl_configure_link);
3167 
3168 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3169 				    unsigned char name_assign_type,
3170 				    const struct rtnl_link_ops *ops,
3171 				    struct nlattr *tb[],
3172 				    struct netlink_ext_ack *extack)
3173 {
3174 	struct net_device *dev;
3175 	unsigned int num_tx_queues = 1;
3176 	unsigned int num_rx_queues = 1;
3177 
3178 	if (tb[IFLA_NUM_TX_QUEUES])
3179 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3180 	else if (ops->get_num_tx_queues)
3181 		num_tx_queues = ops->get_num_tx_queues();
3182 
3183 	if (tb[IFLA_NUM_RX_QUEUES])
3184 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3185 	else if (ops->get_num_rx_queues)
3186 		num_rx_queues = ops->get_num_rx_queues();
3187 
3188 	if (num_tx_queues < 1 || num_tx_queues > 4096) {
3189 		NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3190 		return ERR_PTR(-EINVAL);
3191 	}
3192 
3193 	if (num_rx_queues < 1 || num_rx_queues > 4096) {
3194 		NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3195 		return ERR_PTR(-EINVAL);
3196 	}
3197 
3198 	if (ops->alloc) {
3199 		dev = ops->alloc(tb, ifname, name_assign_type,
3200 				 num_tx_queues, num_rx_queues);
3201 		if (IS_ERR(dev))
3202 			return dev;
3203 	} else {
3204 		dev = alloc_netdev_mqs(ops->priv_size, ifname,
3205 				       name_assign_type, ops->setup,
3206 				       num_tx_queues, num_rx_queues);
3207 	}
3208 
3209 	if (!dev)
3210 		return ERR_PTR(-ENOMEM);
3211 
3212 	dev_net_set(dev, net);
3213 	dev->rtnl_link_ops = ops;
3214 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3215 
3216 	if (tb[IFLA_MTU]) {
3217 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3218 		int err;
3219 
3220 		err = dev_validate_mtu(dev, mtu, extack);
3221 		if (err) {
3222 			free_netdev(dev);
3223 			return ERR_PTR(err);
3224 		}
3225 		dev->mtu = mtu;
3226 	}
3227 	if (tb[IFLA_ADDRESS]) {
3228 		__dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3229 			       nla_len(tb[IFLA_ADDRESS]));
3230 		dev->addr_assign_type = NET_ADDR_SET;
3231 	}
3232 	if (tb[IFLA_BROADCAST])
3233 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3234 				nla_len(tb[IFLA_BROADCAST]));
3235 	if (tb[IFLA_TXQLEN])
3236 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3237 	if (tb[IFLA_OPERSTATE])
3238 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3239 	if (tb[IFLA_LINKMODE])
3240 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3241 	if (tb[IFLA_GROUP])
3242 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3243 	if (tb[IFLA_GSO_MAX_SIZE])
3244 		netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3245 	if (tb[IFLA_GSO_MAX_SEGS])
3246 		netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3247 	if (tb[IFLA_GRO_MAX_SIZE])
3248 		netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3249 
3250 	return dev;
3251 }
3252 EXPORT_SYMBOL(rtnl_create_link);
3253 
3254 static int rtnl_group_changelink(const struct sk_buff *skb,
3255 		struct net *net, int group,
3256 		struct ifinfomsg *ifm,
3257 		struct netlink_ext_ack *extack,
3258 		struct nlattr **tb)
3259 {
3260 	struct net_device *dev, *aux;
3261 	int err;
3262 
3263 	for_each_netdev_safe(net, dev, aux) {
3264 		if (dev->group == group) {
3265 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3266 			if (err < 0)
3267 				return err;
3268 		}
3269 	}
3270 
3271 	return 0;
3272 }
3273 
3274 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3275 			  struct nlattr **attr, struct netlink_ext_ack *extack)
3276 {
3277 	struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3278 	unsigned char name_assign_type = NET_NAME_USER;
3279 	struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3280 	const struct rtnl_link_ops *m_ops;
3281 	struct net_device *master_dev;
3282 	struct net *net = sock_net(skb->sk);
3283 	const struct rtnl_link_ops *ops;
3284 	struct nlattr *tb[IFLA_MAX + 1];
3285 	struct net *dest_net, *link_net;
3286 	struct nlattr **slave_data;
3287 	char kind[MODULE_NAME_LEN];
3288 	struct net_device *dev;
3289 	struct ifinfomsg *ifm;
3290 	char ifname[IFNAMSIZ];
3291 	struct nlattr **data;
3292 	int err;
3293 
3294 #ifdef CONFIG_MODULES
3295 replay:
3296 #endif
3297 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3298 				     ifla_policy, extack);
3299 	if (err < 0)
3300 		return err;
3301 
3302 	err = rtnl_ensure_unique_netns(tb, extack, false);
3303 	if (err < 0)
3304 		return err;
3305 
3306 	if (tb[IFLA_IFNAME])
3307 		nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3308 	else
3309 		ifname[0] = '\0';
3310 
3311 	ifm = nlmsg_data(nlh);
3312 	if (ifm->ifi_index > 0)
3313 		dev = __dev_get_by_index(net, ifm->ifi_index);
3314 	else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3315 		dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3316 	else
3317 		dev = NULL;
3318 
3319 	master_dev = NULL;
3320 	m_ops = NULL;
3321 	if (dev) {
3322 		master_dev = netdev_master_upper_dev_get(dev);
3323 		if (master_dev)
3324 			m_ops = master_dev->rtnl_link_ops;
3325 	}
3326 
3327 	err = validate_linkmsg(dev, tb, extack);
3328 	if (err < 0)
3329 		return err;
3330 
3331 	if (tb[IFLA_LINKINFO]) {
3332 		err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3333 						  tb[IFLA_LINKINFO],
3334 						  ifla_info_policy, NULL);
3335 		if (err < 0)
3336 			return err;
3337 	} else
3338 		memset(linkinfo, 0, sizeof(linkinfo));
3339 
3340 	if (linkinfo[IFLA_INFO_KIND]) {
3341 		nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3342 		ops = rtnl_link_ops_get(kind);
3343 	} else {
3344 		kind[0] = '\0';
3345 		ops = NULL;
3346 	}
3347 
3348 	data = NULL;
3349 	if (ops) {
3350 		if (ops->maxtype > RTNL_MAX_TYPE)
3351 			return -EINVAL;
3352 
3353 		if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3354 			err = nla_parse_nested_deprecated(attr, ops->maxtype,
3355 							  linkinfo[IFLA_INFO_DATA],
3356 							  ops->policy, extack);
3357 			if (err < 0)
3358 				return err;
3359 			data = attr;
3360 		}
3361 		if (ops->validate) {
3362 			err = ops->validate(tb, data, extack);
3363 			if (err < 0)
3364 				return err;
3365 		}
3366 	}
3367 
3368 	slave_data = NULL;
3369 	if (m_ops) {
3370 		if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3371 			return -EINVAL;
3372 
3373 		if (m_ops->slave_maxtype &&
3374 		    linkinfo[IFLA_INFO_SLAVE_DATA]) {
3375 			err = nla_parse_nested_deprecated(slave_attr,
3376 							  m_ops->slave_maxtype,
3377 							  linkinfo[IFLA_INFO_SLAVE_DATA],
3378 							  m_ops->slave_policy,
3379 							  extack);
3380 			if (err < 0)
3381 				return err;
3382 			slave_data = slave_attr;
3383 		}
3384 	}
3385 
3386 	if (dev) {
3387 		int status = 0;
3388 
3389 		if (nlh->nlmsg_flags & NLM_F_EXCL)
3390 			return -EEXIST;
3391 		if (nlh->nlmsg_flags & NLM_F_REPLACE)
3392 			return -EOPNOTSUPP;
3393 
3394 		if (linkinfo[IFLA_INFO_DATA]) {
3395 			if (!ops || ops != dev->rtnl_link_ops ||
3396 			    !ops->changelink)
3397 				return -EOPNOTSUPP;
3398 
3399 			err = ops->changelink(dev, tb, data, extack);
3400 			if (err < 0)
3401 				return err;
3402 			status |= DO_SETLINK_NOTIFY;
3403 		}
3404 
3405 		if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3406 			if (!m_ops || !m_ops->slave_changelink)
3407 				return -EOPNOTSUPP;
3408 
3409 			err = m_ops->slave_changelink(master_dev, dev, tb,
3410 						      slave_data, extack);
3411 			if (err < 0)
3412 				return err;
3413 			status |= DO_SETLINK_NOTIFY;
3414 		}
3415 
3416 		return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3417 	}
3418 
3419 	if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3420 		if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3421 			return rtnl_group_changelink(skb, net,
3422 						nla_get_u32(tb[IFLA_GROUP]),
3423 						ifm, extack, tb);
3424 		return -ENODEV;
3425 	}
3426 
3427 	if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3428 		return -EOPNOTSUPP;
3429 
3430 	if (!ops) {
3431 #ifdef CONFIG_MODULES
3432 		if (kind[0]) {
3433 			__rtnl_unlock();
3434 			request_module("rtnl-link-%s", kind);
3435 			rtnl_lock();
3436 			ops = rtnl_link_ops_get(kind);
3437 			if (ops)
3438 				goto replay;
3439 		}
3440 #endif
3441 		NL_SET_ERR_MSG(extack, "Unknown device type");
3442 		return -EOPNOTSUPP;
3443 	}
3444 
3445 	if (!ops->alloc && !ops->setup)
3446 		return -EOPNOTSUPP;
3447 
3448 	if (!ifname[0]) {
3449 		snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3450 		name_assign_type = NET_NAME_ENUM;
3451 	}
3452 
3453 	dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3454 	if (IS_ERR(dest_net))
3455 		return PTR_ERR(dest_net);
3456 
3457 	if (tb[IFLA_LINK_NETNSID]) {
3458 		int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3459 
3460 		link_net = get_net_ns_by_id(dest_net, id);
3461 		if (!link_net) {
3462 			NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3463 			err =  -EINVAL;
3464 			goto out;
3465 		}
3466 		err = -EPERM;
3467 		if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3468 			goto out;
3469 	} else {
3470 		link_net = NULL;
3471 	}
3472 
3473 	dev = rtnl_create_link(link_net ? : dest_net, ifname,
3474 			       name_assign_type, ops, tb, extack);
3475 	if (IS_ERR(dev)) {
3476 		err = PTR_ERR(dev);
3477 		goto out;
3478 	}
3479 
3480 	dev->ifindex = ifm->ifi_index;
3481 
3482 	if (ops->newlink)
3483 		err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3484 	else
3485 		err = register_netdevice(dev);
3486 	if (err < 0) {
3487 		free_netdev(dev);
3488 		goto out;
3489 	}
3490 
3491 	err = rtnl_configure_link(dev, ifm);
3492 	if (err < 0)
3493 		goto out_unregister;
3494 	if (link_net) {
3495 		err = dev_change_net_namespace(dev, dest_net, ifname);
3496 		if (err < 0)
3497 			goto out_unregister;
3498 	}
3499 	if (tb[IFLA_MASTER]) {
3500 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3501 		if (err)
3502 			goto out_unregister;
3503 	}
3504 out:
3505 	if (link_net)
3506 		put_net(link_net);
3507 	put_net(dest_net);
3508 	return err;
3509 out_unregister:
3510 	if (ops->newlink) {
3511 		LIST_HEAD(list_kill);
3512 
3513 		ops->dellink(dev, &list_kill);
3514 		unregister_netdevice_many(&list_kill);
3515 	} else {
3516 		unregister_netdevice(dev);
3517 	}
3518 	goto out;
3519 }
3520 
3521 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3522 			struct netlink_ext_ack *extack)
3523 {
3524 	struct nlattr **attr;
3525 	int ret;
3526 
3527 	attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3528 	if (!attr)
3529 		return -ENOMEM;
3530 
3531 	ret = __rtnl_newlink(skb, nlh, attr, extack);
3532 	kfree(attr);
3533 	return ret;
3534 }
3535 
3536 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3537 				  const struct nlmsghdr *nlh,
3538 				  struct nlattr **tb,
3539 				  struct netlink_ext_ack *extack)
3540 {
3541 	struct ifinfomsg *ifm;
3542 	int i, err;
3543 
3544 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3545 		NL_SET_ERR_MSG(extack, "Invalid header for get link");
3546 		return -EINVAL;
3547 	}
3548 
3549 	if (!netlink_strict_get_check(skb))
3550 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3551 					      ifla_policy, extack);
3552 
3553 	ifm = nlmsg_data(nlh);
3554 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3555 	    ifm->ifi_change) {
3556 		NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3557 		return -EINVAL;
3558 	}
3559 
3560 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3561 					    ifla_policy, extack);
3562 	if (err)
3563 		return err;
3564 
3565 	for (i = 0; i <= IFLA_MAX; i++) {
3566 		if (!tb[i])
3567 			continue;
3568 
3569 		switch (i) {
3570 		case IFLA_IFNAME:
3571 		case IFLA_ALT_IFNAME:
3572 		case IFLA_EXT_MASK:
3573 		case IFLA_TARGET_NETNSID:
3574 			break;
3575 		default:
3576 			NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3577 			return -EINVAL;
3578 		}
3579 	}
3580 
3581 	return 0;
3582 }
3583 
3584 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3585 			struct netlink_ext_ack *extack)
3586 {
3587 	struct net *net = sock_net(skb->sk);
3588 	struct net *tgt_net = net;
3589 	struct ifinfomsg *ifm;
3590 	struct nlattr *tb[IFLA_MAX+1];
3591 	struct net_device *dev = NULL;
3592 	struct sk_buff *nskb;
3593 	int netnsid = -1;
3594 	int err;
3595 	u32 ext_filter_mask = 0;
3596 
3597 	err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3598 	if (err < 0)
3599 		return err;
3600 
3601 	err = rtnl_ensure_unique_netns(tb, extack, true);
3602 	if (err < 0)
3603 		return err;
3604 
3605 	if (tb[IFLA_TARGET_NETNSID]) {
3606 		netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3607 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3608 		if (IS_ERR(tgt_net))
3609 			return PTR_ERR(tgt_net);
3610 	}
3611 
3612 	if (tb[IFLA_EXT_MASK])
3613 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3614 
3615 	err = -EINVAL;
3616 	ifm = nlmsg_data(nlh);
3617 	if (ifm->ifi_index > 0)
3618 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3619 	else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3620 		dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3621 				   tb[IFLA_ALT_IFNAME], NULL);
3622 	else
3623 		goto out;
3624 
3625 	err = -ENODEV;
3626 	if (dev == NULL)
3627 		goto out;
3628 
3629 	err = -ENOBUFS;
3630 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3631 	if (nskb == NULL)
3632 		goto out;
3633 
3634 	err = rtnl_fill_ifinfo(nskb, dev, net,
3635 			       RTM_NEWLINK, NETLINK_CB(skb).portid,
3636 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3637 			       0, NULL, 0, netnsid, GFP_KERNEL);
3638 	if (err < 0) {
3639 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
3640 		WARN_ON(err == -EMSGSIZE);
3641 		kfree_skb(nskb);
3642 	} else
3643 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3644 out:
3645 	if (netnsid >= 0)
3646 		put_net(tgt_net);
3647 
3648 	return err;
3649 }
3650 
3651 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3652 			   bool *changed, struct netlink_ext_ack *extack)
3653 {
3654 	char *alt_ifname;
3655 	int err;
3656 
3657 	err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3658 	if (err)
3659 		return err;
3660 
3661 	alt_ifname = nla_strdup(attr, GFP_KERNEL);
3662 	if (!alt_ifname)
3663 		return -ENOMEM;
3664 
3665 	if (cmd == RTM_NEWLINKPROP) {
3666 		err = netdev_name_node_alt_create(dev, alt_ifname);
3667 		if (!err)
3668 			alt_ifname = NULL;
3669 	} else if (cmd == RTM_DELLINKPROP) {
3670 		err = netdev_name_node_alt_destroy(dev, alt_ifname);
3671 	} else {
3672 		WARN_ON_ONCE(1);
3673 		err = -EINVAL;
3674 	}
3675 
3676 	kfree(alt_ifname);
3677 	if (!err)
3678 		*changed = true;
3679 	return err;
3680 }
3681 
3682 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3683 			 struct netlink_ext_ack *extack)
3684 {
3685 	struct net *net = sock_net(skb->sk);
3686 	struct nlattr *tb[IFLA_MAX + 1];
3687 	struct net_device *dev;
3688 	struct ifinfomsg *ifm;
3689 	bool changed = false;
3690 	struct nlattr *attr;
3691 	int err, rem;
3692 
3693 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3694 	if (err)
3695 		return err;
3696 
3697 	err = rtnl_ensure_unique_netns(tb, extack, true);
3698 	if (err)
3699 		return err;
3700 
3701 	ifm = nlmsg_data(nlh);
3702 	if (ifm->ifi_index > 0)
3703 		dev = __dev_get_by_index(net, ifm->ifi_index);
3704 	else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3705 		dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3706 				   tb[IFLA_ALT_IFNAME], NULL);
3707 	else
3708 		return -EINVAL;
3709 
3710 	if (!dev)
3711 		return -ENODEV;
3712 
3713 	if (!tb[IFLA_PROP_LIST])
3714 		return 0;
3715 
3716 	nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3717 		switch (nla_type(attr)) {
3718 		case IFLA_ALT_IFNAME:
3719 			err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3720 			if (err)
3721 				return err;
3722 			break;
3723 		}
3724 	}
3725 
3726 	if (changed)
3727 		netdev_state_change(dev);
3728 	return 0;
3729 }
3730 
3731 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3732 			    struct netlink_ext_ack *extack)
3733 {
3734 	return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3735 }
3736 
3737 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3738 			    struct netlink_ext_ack *extack)
3739 {
3740 	return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3741 }
3742 
3743 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3744 {
3745 	struct net *net = sock_net(skb->sk);
3746 	size_t min_ifinfo_dump_size = 0;
3747 	struct nlattr *tb[IFLA_MAX+1];
3748 	u32 ext_filter_mask = 0;
3749 	struct net_device *dev;
3750 	int hdrlen;
3751 
3752 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3753 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3754 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3755 
3756 	if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3757 		if (tb[IFLA_EXT_MASK])
3758 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3759 	}
3760 
3761 	if (!ext_filter_mask)
3762 		return NLMSG_GOODSIZE;
3763 	/*
3764 	 * traverse the list of net devices and compute the minimum
3765 	 * buffer size based upon the filter mask.
3766 	 */
3767 	rcu_read_lock();
3768 	for_each_netdev_rcu(net, dev) {
3769 		min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3770 					   if_nlmsg_size(dev, ext_filter_mask));
3771 	}
3772 	rcu_read_unlock();
3773 
3774 	return nlmsg_total_size(min_ifinfo_dump_size);
3775 }
3776 
3777 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3778 {
3779 	int idx;
3780 	int s_idx = cb->family;
3781 	int type = cb->nlh->nlmsg_type - RTM_BASE;
3782 	int ret = 0;
3783 
3784 	if (s_idx == 0)
3785 		s_idx = 1;
3786 
3787 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3788 		struct rtnl_link __rcu **tab;
3789 		struct rtnl_link *link;
3790 		rtnl_dumpit_func dumpit;
3791 
3792 		if (idx < s_idx || idx == PF_PACKET)
3793 			continue;
3794 
3795 		if (type < 0 || type >= RTM_NR_MSGTYPES)
3796 			continue;
3797 
3798 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3799 		if (!tab)
3800 			continue;
3801 
3802 		link = rcu_dereference_rtnl(tab[type]);
3803 		if (!link)
3804 			continue;
3805 
3806 		dumpit = link->dumpit;
3807 		if (!dumpit)
3808 			continue;
3809 
3810 		if (idx > s_idx) {
3811 			memset(&cb->args[0], 0, sizeof(cb->args));
3812 			cb->prev_seq = 0;
3813 			cb->seq = 0;
3814 		}
3815 		ret = dumpit(skb, cb);
3816 		if (ret)
3817 			break;
3818 	}
3819 	cb->family = idx;
3820 
3821 	return skb->len ? : ret;
3822 }
3823 
3824 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3825 				       unsigned int change,
3826 				       u32 event, gfp_t flags, int *new_nsid,
3827 				       int new_ifindex)
3828 {
3829 	struct net *net = dev_net(dev);
3830 	struct sk_buff *skb;
3831 	int err = -ENOBUFS;
3832 
3833 	skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
3834 	if (skb == NULL)
3835 		goto errout;
3836 
3837 	err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3838 			       type, 0, 0, change, 0, 0, event,
3839 			       new_nsid, new_ifindex, -1, flags);
3840 	if (err < 0) {
3841 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
3842 		WARN_ON(err == -EMSGSIZE);
3843 		kfree_skb(skb);
3844 		goto errout;
3845 	}
3846 	return skb;
3847 errout:
3848 	if (err < 0)
3849 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3850 	return NULL;
3851 }
3852 
3853 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3854 {
3855 	struct net *net = dev_net(dev);
3856 
3857 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3858 }
3859 
3860 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3861 			       unsigned int change, u32 event,
3862 			       gfp_t flags, int *new_nsid, int new_ifindex)
3863 {
3864 	struct sk_buff *skb;
3865 
3866 	if (dev->reg_state != NETREG_REGISTERED)
3867 		return;
3868 
3869 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3870 				     new_ifindex);
3871 	if (skb)
3872 		rtmsg_ifinfo_send(skb, dev, flags);
3873 }
3874 
3875 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3876 		  gfp_t flags)
3877 {
3878 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3879 			   NULL, 0);
3880 }
3881 
3882 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3883 			 gfp_t flags, int *new_nsid, int new_ifindex)
3884 {
3885 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3886 			   new_nsid, new_ifindex);
3887 }
3888 
3889 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3890 				   struct net_device *dev,
3891 				   u8 *addr, u16 vid, u32 pid, u32 seq,
3892 				   int type, unsigned int flags,
3893 				   int nlflags, u16 ndm_state)
3894 {
3895 	struct nlmsghdr *nlh;
3896 	struct ndmsg *ndm;
3897 
3898 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3899 	if (!nlh)
3900 		return -EMSGSIZE;
3901 
3902 	ndm = nlmsg_data(nlh);
3903 	ndm->ndm_family  = AF_BRIDGE;
3904 	ndm->ndm_pad1	 = 0;
3905 	ndm->ndm_pad2    = 0;
3906 	ndm->ndm_flags	 = flags;
3907 	ndm->ndm_type	 = 0;
3908 	ndm->ndm_ifindex = dev->ifindex;
3909 	ndm->ndm_state   = ndm_state;
3910 
3911 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3912 		goto nla_put_failure;
3913 	if (vid)
3914 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3915 			goto nla_put_failure;
3916 
3917 	nlmsg_end(skb, nlh);
3918 	return 0;
3919 
3920 nla_put_failure:
3921 	nlmsg_cancel(skb, nlh);
3922 	return -EMSGSIZE;
3923 }
3924 
3925 static inline size_t rtnl_fdb_nlmsg_size(void)
3926 {
3927 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3928 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
3929 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
3930 	       0;
3931 }
3932 
3933 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3934 			    u16 ndm_state)
3935 {
3936 	struct net *net = dev_net(dev);
3937 	struct sk_buff *skb;
3938 	int err = -ENOBUFS;
3939 
3940 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3941 	if (!skb)
3942 		goto errout;
3943 
3944 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3945 				      0, 0, type, NTF_SELF, 0, ndm_state);
3946 	if (err < 0) {
3947 		kfree_skb(skb);
3948 		goto errout;
3949 	}
3950 
3951 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3952 	return;
3953 errout:
3954 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3955 }
3956 
3957 /*
3958  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3959  */
3960 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3961 		     struct nlattr *tb[],
3962 		     struct net_device *dev,
3963 		     const unsigned char *addr, u16 vid,
3964 		     u16 flags)
3965 {
3966 	int err = -EINVAL;
3967 
3968 	/* If aging addresses are supported device will need to
3969 	 * implement its own handler for this.
3970 	 */
3971 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3972 		netdev_info(dev, "default FDB implementation only supports local addresses\n");
3973 		return err;
3974 	}
3975 
3976 	if (vid) {
3977 		netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
3978 		return err;
3979 	}
3980 
3981 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3982 		err = dev_uc_add_excl(dev, addr);
3983 	else if (is_multicast_ether_addr(addr))
3984 		err = dev_mc_add_excl(dev, addr);
3985 
3986 	/* Only return duplicate errors if NLM_F_EXCL is set */
3987 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3988 		err = 0;
3989 
3990 	return err;
3991 }
3992 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3993 
3994 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3995 			 struct netlink_ext_ack *extack)
3996 {
3997 	u16 vid = 0;
3998 
3999 	if (vlan_attr) {
4000 		if (nla_len(vlan_attr) != sizeof(u16)) {
4001 			NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4002 			return -EINVAL;
4003 		}
4004 
4005 		vid = nla_get_u16(vlan_attr);
4006 
4007 		if (!vid || vid >= VLAN_VID_MASK) {
4008 			NL_SET_ERR_MSG(extack, "invalid vlan id");
4009 			return -EINVAL;
4010 		}
4011 	}
4012 	*p_vid = vid;
4013 	return 0;
4014 }
4015 
4016 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4017 			struct netlink_ext_ack *extack)
4018 {
4019 	struct net *net = sock_net(skb->sk);
4020 	struct ndmsg *ndm;
4021 	struct nlattr *tb[NDA_MAX+1];
4022 	struct net_device *dev;
4023 	u8 *addr;
4024 	u16 vid;
4025 	int err;
4026 
4027 	err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4028 				     extack);
4029 	if (err < 0)
4030 		return err;
4031 
4032 	ndm = nlmsg_data(nlh);
4033 	if (ndm->ndm_ifindex == 0) {
4034 		NL_SET_ERR_MSG(extack, "invalid ifindex");
4035 		return -EINVAL;
4036 	}
4037 
4038 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4039 	if (dev == NULL) {
4040 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4041 		return -ENODEV;
4042 	}
4043 
4044 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4045 		NL_SET_ERR_MSG(extack, "invalid address");
4046 		return -EINVAL;
4047 	}
4048 
4049 	if (dev->type != ARPHRD_ETHER) {
4050 		NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4051 		return -EINVAL;
4052 	}
4053 
4054 	addr = nla_data(tb[NDA_LLADDR]);
4055 
4056 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4057 	if (err)
4058 		return err;
4059 
4060 	err = -EOPNOTSUPP;
4061 
4062 	/* Support fdb on master device the net/bridge default case */
4063 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4064 	    netif_is_bridge_port(dev)) {
4065 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4066 		const struct net_device_ops *ops = br_dev->netdev_ops;
4067 
4068 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4069 				       nlh->nlmsg_flags, extack);
4070 		if (err)
4071 			goto out;
4072 		else
4073 			ndm->ndm_flags &= ~NTF_MASTER;
4074 	}
4075 
4076 	/* Embedded bridge, macvlan, and any other device support */
4077 	if ((ndm->ndm_flags & NTF_SELF)) {
4078 		if (dev->netdev_ops->ndo_fdb_add)
4079 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4080 							   vid,
4081 							   nlh->nlmsg_flags,
4082 							   extack);
4083 		else
4084 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4085 					       nlh->nlmsg_flags);
4086 
4087 		if (!err) {
4088 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4089 					ndm->ndm_state);
4090 			ndm->ndm_flags &= ~NTF_SELF;
4091 		}
4092 	}
4093 out:
4094 	return err;
4095 }
4096 
4097 /*
4098  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4099  */
4100 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4101 		     struct nlattr *tb[],
4102 		     struct net_device *dev,
4103 		     const unsigned char *addr, u16 vid)
4104 {
4105 	int err = -EINVAL;
4106 
4107 	/* If aging addresses are supported device will need to
4108 	 * implement its own handler for this.
4109 	 */
4110 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
4111 		netdev_info(dev, "default FDB implementation only supports local addresses\n");
4112 		return err;
4113 	}
4114 
4115 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4116 		err = dev_uc_del(dev, addr);
4117 	else if (is_multicast_ether_addr(addr))
4118 		err = dev_mc_del(dev, addr);
4119 
4120 	return err;
4121 }
4122 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4123 
4124 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4125 			struct netlink_ext_ack *extack)
4126 {
4127 	struct net *net = sock_net(skb->sk);
4128 	struct ndmsg *ndm;
4129 	struct nlattr *tb[NDA_MAX+1];
4130 	struct net_device *dev;
4131 	__u8 *addr;
4132 	int err;
4133 	u16 vid;
4134 
4135 	if (!netlink_capable(skb, CAP_NET_ADMIN))
4136 		return -EPERM;
4137 
4138 	err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4139 				     extack);
4140 	if (err < 0)
4141 		return err;
4142 
4143 	ndm = nlmsg_data(nlh);
4144 	if (ndm->ndm_ifindex == 0) {
4145 		NL_SET_ERR_MSG(extack, "invalid ifindex");
4146 		return -EINVAL;
4147 	}
4148 
4149 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4150 	if (dev == NULL) {
4151 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4152 		return -ENODEV;
4153 	}
4154 
4155 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4156 		NL_SET_ERR_MSG(extack, "invalid address");
4157 		return -EINVAL;
4158 	}
4159 
4160 	if (dev->type != ARPHRD_ETHER) {
4161 		NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4162 		return -EINVAL;
4163 	}
4164 
4165 	addr = nla_data(tb[NDA_LLADDR]);
4166 
4167 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4168 	if (err)
4169 		return err;
4170 
4171 	err = -EOPNOTSUPP;
4172 
4173 	/* Support fdb on master device the net/bridge default case */
4174 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4175 	    netif_is_bridge_port(dev)) {
4176 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4177 		const struct net_device_ops *ops = br_dev->netdev_ops;
4178 
4179 		if (ops->ndo_fdb_del)
4180 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4181 
4182 		if (err)
4183 			goto out;
4184 		else
4185 			ndm->ndm_flags &= ~NTF_MASTER;
4186 	}
4187 
4188 	/* Embedded bridge, macvlan, and any other device support */
4189 	if (ndm->ndm_flags & NTF_SELF) {
4190 		if (dev->netdev_ops->ndo_fdb_del)
4191 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4192 							   vid);
4193 		else
4194 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4195 
4196 		if (!err) {
4197 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4198 					ndm->ndm_state);
4199 			ndm->ndm_flags &= ~NTF_SELF;
4200 		}
4201 	}
4202 out:
4203 	return err;
4204 }
4205 
4206 static int nlmsg_populate_fdb(struct sk_buff *skb,
4207 			      struct netlink_callback *cb,
4208 			      struct net_device *dev,
4209 			      int *idx,
4210 			      struct netdev_hw_addr_list *list)
4211 {
4212 	struct netdev_hw_addr *ha;
4213 	int err;
4214 	u32 portid, seq;
4215 
4216 	portid = NETLINK_CB(cb->skb).portid;
4217 	seq = cb->nlh->nlmsg_seq;
4218 
4219 	list_for_each_entry(ha, &list->list, list) {
4220 		if (*idx < cb->args[2])
4221 			goto skip;
4222 
4223 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4224 					      portid, seq,
4225 					      RTM_NEWNEIGH, NTF_SELF,
4226 					      NLM_F_MULTI, NUD_PERMANENT);
4227 		if (err < 0)
4228 			return err;
4229 skip:
4230 		*idx += 1;
4231 	}
4232 	return 0;
4233 }
4234 
4235 /**
4236  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4237  * @skb: socket buffer to store message in
4238  * @cb: netlink callback
4239  * @dev: netdevice
4240  * @filter_dev: ignored
4241  * @idx: the number of FDB table entries dumped is added to *@idx
4242  *
4243  * Default netdevice operation to dump the existing unicast address list.
4244  * Returns number of addresses from list put in skb.
4245  */
4246 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4247 		      struct netlink_callback *cb,
4248 		      struct net_device *dev,
4249 		      struct net_device *filter_dev,
4250 		      int *idx)
4251 {
4252 	int err;
4253 
4254 	if (dev->type != ARPHRD_ETHER)
4255 		return -EINVAL;
4256 
4257 	netif_addr_lock_bh(dev);
4258 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4259 	if (err)
4260 		goto out;
4261 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4262 out:
4263 	netif_addr_unlock_bh(dev);
4264 	return err;
4265 }
4266 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4267 
4268 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4269 				 int *br_idx, int *brport_idx,
4270 				 struct netlink_ext_ack *extack)
4271 {
4272 	struct nlattr *tb[NDA_MAX + 1];
4273 	struct ndmsg *ndm;
4274 	int err, i;
4275 
4276 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4277 		NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4278 		return -EINVAL;
4279 	}
4280 
4281 	ndm = nlmsg_data(nlh);
4282 	if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4283 	    ndm->ndm_flags || ndm->ndm_type) {
4284 		NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4285 		return -EINVAL;
4286 	}
4287 
4288 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4289 					    NDA_MAX, NULL, extack);
4290 	if (err < 0)
4291 		return err;
4292 
4293 	*brport_idx = ndm->ndm_ifindex;
4294 	for (i = 0; i <= NDA_MAX; ++i) {
4295 		if (!tb[i])
4296 			continue;
4297 
4298 		switch (i) {
4299 		case NDA_IFINDEX:
4300 			if (nla_len(tb[i]) != sizeof(u32)) {
4301 				NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4302 				return -EINVAL;
4303 			}
4304 			*brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4305 			break;
4306 		case NDA_MASTER:
4307 			if (nla_len(tb[i]) != sizeof(u32)) {
4308 				NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4309 				return -EINVAL;
4310 			}
4311 			*br_idx = nla_get_u32(tb[NDA_MASTER]);
4312 			break;
4313 		default:
4314 			NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4315 			return -EINVAL;
4316 		}
4317 	}
4318 
4319 	return 0;
4320 }
4321 
4322 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4323 				 int *br_idx, int *brport_idx,
4324 				 struct netlink_ext_ack *extack)
4325 {
4326 	struct nlattr *tb[IFLA_MAX+1];
4327 	int err;
4328 
4329 	/* A hack to preserve kernel<->userspace interface.
4330 	 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4331 	 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4332 	 * So, check for ndmsg with an optional u32 attribute (not used here).
4333 	 * Fortunately these sizes don't conflict with the size of ifinfomsg
4334 	 * with an optional attribute.
4335 	 */
4336 	if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4337 	    (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4338 	     nla_attr_size(sizeof(u32)))) {
4339 		struct ifinfomsg *ifm;
4340 
4341 		err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4342 					     tb, IFLA_MAX, ifla_policy,
4343 					     extack);
4344 		if (err < 0) {
4345 			return -EINVAL;
4346 		} else if (err == 0) {
4347 			if (tb[IFLA_MASTER])
4348 				*br_idx = nla_get_u32(tb[IFLA_MASTER]);
4349 		}
4350 
4351 		ifm = nlmsg_data(nlh);
4352 		*brport_idx = ifm->ifi_index;
4353 	}
4354 	return 0;
4355 }
4356 
4357 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4358 {
4359 	struct net_device *dev;
4360 	struct net_device *br_dev = NULL;
4361 	const struct net_device_ops *ops = NULL;
4362 	const struct net_device_ops *cops = NULL;
4363 	struct net *net = sock_net(skb->sk);
4364 	struct hlist_head *head;
4365 	int brport_idx = 0;
4366 	int br_idx = 0;
4367 	int h, s_h;
4368 	int idx = 0, s_idx;
4369 	int err = 0;
4370 	int fidx = 0;
4371 
4372 	if (cb->strict_check)
4373 		err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4374 					    cb->extack);
4375 	else
4376 		err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4377 					    cb->extack);
4378 	if (err < 0)
4379 		return err;
4380 
4381 	if (br_idx) {
4382 		br_dev = __dev_get_by_index(net, br_idx);
4383 		if (!br_dev)
4384 			return -ENODEV;
4385 
4386 		ops = br_dev->netdev_ops;
4387 	}
4388 
4389 	s_h = cb->args[0];
4390 	s_idx = cb->args[1];
4391 
4392 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4393 		idx = 0;
4394 		head = &net->dev_index_head[h];
4395 		hlist_for_each_entry(dev, head, index_hlist) {
4396 
4397 			if (brport_idx && (dev->ifindex != brport_idx))
4398 				continue;
4399 
4400 			if (!br_idx) { /* user did not specify a specific bridge */
4401 				if (netif_is_bridge_port(dev)) {
4402 					br_dev = netdev_master_upper_dev_get(dev);
4403 					cops = br_dev->netdev_ops;
4404 				}
4405 			} else {
4406 				if (dev != br_dev &&
4407 				    !netif_is_bridge_port(dev))
4408 					continue;
4409 
4410 				if (br_dev != netdev_master_upper_dev_get(dev) &&
4411 				    !netif_is_bridge_master(dev))
4412 					continue;
4413 				cops = ops;
4414 			}
4415 
4416 			if (idx < s_idx)
4417 				goto cont;
4418 
4419 			if (netif_is_bridge_port(dev)) {
4420 				if (cops && cops->ndo_fdb_dump) {
4421 					err = cops->ndo_fdb_dump(skb, cb,
4422 								br_dev, dev,
4423 								&fidx);
4424 					if (err == -EMSGSIZE)
4425 						goto out;
4426 				}
4427 			}
4428 
4429 			if (dev->netdev_ops->ndo_fdb_dump)
4430 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4431 								    dev, NULL,
4432 								    &fidx);
4433 			else
4434 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4435 							&fidx);
4436 			if (err == -EMSGSIZE)
4437 				goto out;
4438 
4439 			cops = NULL;
4440 
4441 			/* reset fdb offset to 0 for rest of the interfaces */
4442 			cb->args[2] = 0;
4443 			fidx = 0;
4444 cont:
4445 			idx++;
4446 		}
4447 	}
4448 
4449 out:
4450 	cb->args[0] = h;
4451 	cb->args[1] = idx;
4452 	cb->args[2] = fidx;
4453 
4454 	return skb->len;
4455 }
4456 
4457 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4458 				struct nlattr **tb, u8 *ndm_flags,
4459 				int *br_idx, int *brport_idx, u8 **addr,
4460 				u16 *vid, struct netlink_ext_ack *extack)
4461 {
4462 	struct ndmsg *ndm;
4463 	int err, i;
4464 
4465 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4466 		NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4467 		return -EINVAL;
4468 	}
4469 
4470 	ndm = nlmsg_data(nlh);
4471 	if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4472 	    ndm->ndm_type) {
4473 		NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4474 		return -EINVAL;
4475 	}
4476 
4477 	if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4478 		NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4479 		return -EINVAL;
4480 	}
4481 
4482 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4483 					    NDA_MAX, nda_policy, extack);
4484 	if (err < 0)
4485 		return err;
4486 
4487 	*ndm_flags = ndm->ndm_flags;
4488 	*brport_idx = ndm->ndm_ifindex;
4489 	for (i = 0; i <= NDA_MAX; ++i) {
4490 		if (!tb[i])
4491 			continue;
4492 
4493 		switch (i) {
4494 		case NDA_MASTER:
4495 			*br_idx = nla_get_u32(tb[i]);
4496 			break;
4497 		case NDA_LLADDR:
4498 			if (nla_len(tb[i]) != ETH_ALEN) {
4499 				NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4500 				return -EINVAL;
4501 			}
4502 			*addr = nla_data(tb[i]);
4503 			break;
4504 		case NDA_VLAN:
4505 			err = fdb_vid_parse(tb[i], vid, extack);
4506 			if (err)
4507 				return err;
4508 			break;
4509 		case NDA_VNI:
4510 			break;
4511 		default:
4512 			NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4513 			return -EINVAL;
4514 		}
4515 	}
4516 
4517 	return 0;
4518 }
4519 
4520 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4521 			struct netlink_ext_ack *extack)
4522 {
4523 	struct net_device *dev = NULL, *br_dev = NULL;
4524 	const struct net_device_ops *ops = NULL;
4525 	struct net *net = sock_net(in_skb->sk);
4526 	struct nlattr *tb[NDA_MAX + 1];
4527 	struct sk_buff *skb;
4528 	int brport_idx = 0;
4529 	u8 ndm_flags = 0;
4530 	int br_idx = 0;
4531 	u8 *addr = NULL;
4532 	u16 vid = 0;
4533 	int err;
4534 
4535 	err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4536 				   &brport_idx, &addr, &vid, extack);
4537 	if (err < 0)
4538 		return err;
4539 
4540 	if (!addr) {
4541 		NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4542 		return -EINVAL;
4543 	}
4544 
4545 	if (brport_idx) {
4546 		dev = __dev_get_by_index(net, brport_idx);
4547 		if (!dev) {
4548 			NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4549 			return -ENODEV;
4550 		}
4551 	}
4552 
4553 	if (br_idx) {
4554 		if (dev) {
4555 			NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4556 			return -EINVAL;
4557 		}
4558 
4559 		br_dev = __dev_get_by_index(net, br_idx);
4560 		if (!br_dev) {
4561 			NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4562 			return -EINVAL;
4563 		}
4564 		ops = br_dev->netdev_ops;
4565 	}
4566 
4567 	if (dev) {
4568 		if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4569 			if (!netif_is_bridge_port(dev)) {
4570 				NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4571 				return -EINVAL;
4572 			}
4573 			br_dev = netdev_master_upper_dev_get(dev);
4574 			if (!br_dev) {
4575 				NL_SET_ERR_MSG(extack, "Master of device not found");
4576 				return -EINVAL;
4577 			}
4578 			ops = br_dev->netdev_ops;
4579 		} else {
4580 			if (!(ndm_flags & NTF_SELF)) {
4581 				NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4582 				return -EINVAL;
4583 			}
4584 			ops = dev->netdev_ops;
4585 		}
4586 	}
4587 
4588 	if (!br_dev && !dev) {
4589 		NL_SET_ERR_MSG(extack, "No device specified");
4590 		return -ENODEV;
4591 	}
4592 
4593 	if (!ops || !ops->ndo_fdb_get) {
4594 		NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4595 		return -EOPNOTSUPP;
4596 	}
4597 
4598 	skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4599 	if (!skb)
4600 		return -ENOBUFS;
4601 
4602 	if (br_dev)
4603 		dev = br_dev;
4604 	err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4605 			       NETLINK_CB(in_skb).portid,
4606 			       nlh->nlmsg_seq, extack);
4607 	if (err)
4608 		goto out;
4609 
4610 	return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4611 out:
4612 	kfree_skb(skb);
4613 	return err;
4614 }
4615 
4616 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4617 			       unsigned int attrnum, unsigned int flag)
4618 {
4619 	if (mask & flag)
4620 		return nla_put_u8(skb, attrnum, !!(flags & flag));
4621 	return 0;
4622 }
4623 
4624 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4625 			    struct net_device *dev, u16 mode,
4626 			    u32 flags, u32 mask, int nlflags,
4627 			    u32 filter_mask,
4628 			    int (*vlan_fill)(struct sk_buff *skb,
4629 					     struct net_device *dev,
4630 					     u32 filter_mask))
4631 {
4632 	struct nlmsghdr *nlh;
4633 	struct ifinfomsg *ifm;
4634 	struct nlattr *br_afspec;
4635 	struct nlattr *protinfo;
4636 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4637 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4638 	int err = 0;
4639 
4640 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4641 	if (nlh == NULL)
4642 		return -EMSGSIZE;
4643 
4644 	ifm = nlmsg_data(nlh);
4645 	ifm->ifi_family = AF_BRIDGE;
4646 	ifm->__ifi_pad = 0;
4647 	ifm->ifi_type = dev->type;
4648 	ifm->ifi_index = dev->ifindex;
4649 	ifm->ifi_flags = dev_get_flags(dev);
4650 	ifm->ifi_change = 0;
4651 
4652 
4653 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4654 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4655 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4656 	    (br_dev &&
4657 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4658 	    (dev->addr_len &&
4659 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4660 	    (dev->ifindex != dev_get_iflink(dev) &&
4661 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4662 		goto nla_put_failure;
4663 
4664 	br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4665 	if (!br_afspec)
4666 		goto nla_put_failure;
4667 
4668 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4669 		nla_nest_cancel(skb, br_afspec);
4670 		goto nla_put_failure;
4671 	}
4672 
4673 	if (mode != BRIDGE_MODE_UNDEF) {
4674 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4675 			nla_nest_cancel(skb, br_afspec);
4676 			goto nla_put_failure;
4677 		}
4678 	}
4679 	if (vlan_fill) {
4680 		err = vlan_fill(skb, dev, filter_mask);
4681 		if (err) {
4682 			nla_nest_cancel(skb, br_afspec);
4683 			goto nla_put_failure;
4684 		}
4685 	}
4686 	nla_nest_end(skb, br_afspec);
4687 
4688 	protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4689 	if (!protinfo)
4690 		goto nla_put_failure;
4691 
4692 	if (brport_nla_put_flag(skb, flags, mask,
4693 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4694 	    brport_nla_put_flag(skb, flags, mask,
4695 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4696 	    brport_nla_put_flag(skb, flags, mask,
4697 				IFLA_BRPORT_FAST_LEAVE,
4698 				BR_MULTICAST_FAST_LEAVE) ||
4699 	    brport_nla_put_flag(skb, flags, mask,
4700 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4701 	    brport_nla_put_flag(skb, flags, mask,
4702 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4703 	    brport_nla_put_flag(skb, flags, mask,
4704 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4705 	    brport_nla_put_flag(skb, flags, mask,
4706 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4707 	    brport_nla_put_flag(skb, flags, mask,
4708 				IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4709 	    brport_nla_put_flag(skb, flags, mask,
4710 				IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4711 	    brport_nla_put_flag(skb, flags, mask,
4712 				IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4713 		nla_nest_cancel(skb, protinfo);
4714 		goto nla_put_failure;
4715 	}
4716 
4717 	nla_nest_end(skb, protinfo);
4718 
4719 	nlmsg_end(skb, nlh);
4720 	return 0;
4721 nla_put_failure:
4722 	nlmsg_cancel(skb, nlh);
4723 	return err ? err : -EMSGSIZE;
4724 }
4725 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4726 
4727 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4728 				    bool strict_check, u32 *filter_mask,
4729 				    struct netlink_ext_ack *extack)
4730 {
4731 	struct nlattr *tb[IFLA_MAX+1];
4732 	int err, i;
4733 
4734 	if (strict_check) {
4735 		struct ifinfomsg *ifm;
4736 
4737 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4738 			NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4739 			return -EINVAL;
4740 		}
4741 
4742 		ifm = nlmsg_data(nlh);
4743 		if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4744 		    ifm->ifi_change || ifm->ifi_index) {
4745 			NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4746 			return -EINVAL;
4747 		}
4748 
4749 		err = nlmsg_parse_deprecated_strict(nlh,
4750 						    sizeof(struct ifinfomsg),
4751 						    tb, IFLA_MAX, ifla_policy,
4752 						    extack);
4753 	} else {
4754 		err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4755 					     tb, IFLA_MAX, ifla_policy,
4756 					     extack);
4757 	}
4758 	if (err < 0)
4759 		return err;
4760 
4761 	/* new attributes should only be added with strict checking */
4762 	for (i = 0; i <= IFLA_MAX; ++i) {
4763 		if (!tb[i])
4764 			continue;
4765 
4766 		switch (i) {
4767 		case IFLA_EXT_MASK:
4768 			*filter_mask = nla_get_u32(tb[i]);
4769 			break;
4770 		default:
4771 			if (strict_check) {
4772 				NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4773 				return -EINVAL;
4774 			}
4775 		}
4776 	}
4777 
4778 	return 0;
4779 }
4780 
4781 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4782 {
4783 	const struct nlmsghdr *nlh = cb->nlh;
4784 	struct net *net = sock_net(skb->sk);
4785 	struct net_device *dev;
4786 	int idx = 0;
4787 	u32 portid = NETLINK_CB(cb->skb).portid;
4788 	u32 seq = nlh->nlmsg_seq;
4789 	u32 filter_mask = 0;
4790 	int err;
4791 
4792 	err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4793 				       cb->extack);
4794 	if (err < 0 && cb->strict_check)
4795 		return err;
4796 
4797 	rcu_read_lock();
4798 	for_each_netdev_rcu(net, dev) {
4799 		const struct net_device_ops *ops = dev->netdev_ops;
4800 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4801 
4802 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4803 			if (idx >= cb->args[0]) {
4804 				err = br_dev->netdev_ops->ndo_bridge_getlink(
4805 						skb, portid, seq, dev,
4806 						filter_mask, NLM_F_MULTI);
4807 				if (err < 0 && err != -EOPNOTSUPP) {
4808 					if (likely(skb->len))
4809 						break;
4810 
4811 					goto out_err;
4812 				}
4813 			}
4814 			idx++;
4815 		}
4816 
4817 		if (ops->ndo_bridge_getlink) {
4818 			if (idx >= cb->args[0]) {
4819 				err = ops->ndo_bridge_getlink(skb, portid,
4820 							      seq, dev,
4821 							      filter_mask,
4822 							      NLM_F_MULTI);
4823 				if (err < 0 && err != -EOPNOTSUPP) {
4824 					if (likely(skb->len))
4825 						break;
4826 
4827 					goto out_err;
4828 				}
4829 			}
4830 			idx++;
4831 		}
4832 	}
4833 	err = skb->len;
4834 out_err:
4835 	rcu_read_unlock();
4836 	cb->args[0] = idx;
4837 
4838 	return err;
4839 }
4840 
4841 static inline size_t bridge_nlmsg_size(void)
4842 {
4843 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4844 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
4845 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
4846 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
4847 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
4848 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
4849 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
4850 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
4851 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
4852 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
4853 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
4854 }
4855 
4856 static int rtnl_bridge_notify(struct net_device *dev)
4857 {
4858 	struct net *net = dev_net(dev);
4859 	struct sk_buff *skb;
4860 	int err = -EOPNOTSUPP;
4861 
4862 	if (!dev->netdev_ops->ndo_bridge_getlink)
4863 		return 0;
4864 
4865 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4866 	if (!skb) {
4867 		err = -ENOMEM;
4868 		goto errout;
4869 	}
4870 
4871 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4872 	if (err < 0)
4873 		goto errout;
4874 
4875 	/* Notification info is only filled for bridge ports, not the bridge
4876 	 * device itself. Therefore, a zero notification length is valid and
4877 	 * should not result in an error.
4878 	 */
4879 	if (!skb->len)
4880 		goto errout;
4881 
4882 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4883 	return 0;
4884 errout:
4885 	WARN_ON(err == -EMSGSIZE);
4886 	kfree_skb(skb);
4887 	if (err)
4888 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4889 	return err;
4890 }
4891 
4892 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4893 			       struct netlink_ext_ack *extack)
4894 {
4895 	struct net *net = sock_net(skb->sk);
4896 	struct ifinfomsg *ifm;
4897 	struct net_device *dev;
4898 	struct nlattr *br_spec, *attr = NULL;
4899 	int rem, err = -EOPNOTSUPP;
4900 	u16 flags = 0;
4901 	bool have_flags = false;
4902 
4903 	if (nlmsg_len(nlh) < sizeof(*ifm))
4904 		return -EINVAL;
4905 
4906 	ifm = nlmsg_data(nlh);
4907 	if (ifm->ifi_family != AF_BRIDGE)
4908 		return -EPFNOSUPPORT;
4909 
4910 	dev = __dev_get_by_index(net, ifm->ifi_index);
4911 	if (!dev) {
4912 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4913 		return -ENODEV;
4914 	}
4915 
4916 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4917 	if (br_spec) {
4918 		nla_for_each_nested(attr, br_spec, rem) {
4919 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4920 				if (nla_len(attr) < sizeof(flags))
4921 					return -EINVAL;
4922 
4923 				have_flags = true;
4924 				flags = nla_get_u16(attr);
4925 				break;
4926 			}
4927 		}
4928 	}
4929 
4930 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4931 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4932 
4933 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4934 			err = -EOPNOTSUPP;
4935 			goto out;
4936 		}
4937 
4938 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4939 							     extack);
4940 		if (err)
4941 			goto out;
4942 
4943 		flags &= ~BRIDGE_FLAGS_MASTER;
4944 	}
4945 
4946 	if ((flags & BRIDGE_FLAGS_SELF)) {
4947 		if (!dev->netdev_ops->ndo_bridge_setlink)
4948 			err = -EOPNOTSUPP;
4949 		else
4950 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4951 								  flags,
4952 								  extack);
4953 		if (!err) {
4954 			flags &= ~BRIDGE_FLAGS_SELF;
4955 
4956 			/* Generate event to notify upper layer of bridge
4957 			 * change
4958 			 */
4959 			err = rtnl_bridge_notify(dev);
4960 		}
4961 	}
4962 
4963 	if (have_flags)
4964 		memcpy(nla_data(attr), &flags, sizeof(flags));
4965 out:
4966 	return err;
4967 }
4968 
4969 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4970 			       struct netlink_ext_ack *extack)
4971 {
4972 	struct net *net = sock_net(skb->sk);
4973 	struct ifinfomsg *ifm;
4974 	struct net_device *dev;
4975 	struct nlattr *br_spec, *attr = NULL;
4976 	int rem, err = -EOPNOTSUPP;
4977 	u16 flags = 0;
4978 	bool have_flags = false;
4979 
4980 	if (nlmsg_len(nlh) < sizeof(*ifm))
4981 		return -EINVAL;
4982 
4983 	ifm = nlmsg_data(nlh);
4984 	if (ifm->ifi_family != AF_BRIDGE)
4985 		return -EPFNOSUPPORT;
4986 
4987 	dev = __dev_get_by_index(net, ifm->ifi_index);
4988 	if (!dev) {
4989 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4990 		return -ENODEV;
4991 	}
4992 
4993 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4994 	if (br_spec) {
4995 		nla_for_each_nested(attr, br_spec, rem) {
4996 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4997 				if (nla_len(attr) < sizeof(flags))
4998 					return -EINVAL;
4999 
5000 				have_flags = true;
5001 				flags = nla_get_u16(attr);
5002 				break;
5003 			}
5004 		}
5005 	}
5006 
5007 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5008 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5009 
5010 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5011 			err = -EOPNOTSUPP;
5012 			goto out;
5013 		}
5014 
5015 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5016 		if (err)
5017 			goto out;
5018 
5019 		flags &= ~BRIDGE_FLAGS_MASTER;
5020 	}
5021 
5022 	if ((flags & BRIDGE_FLAGS_SELF)) {
5023 		if (!dev->netdev_ops->ndo_bridge_dellink)
5024 			err = -EOPNOTSUPP;
5025 		else
5026 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5027 								  flags);
5028 
5029 		if (!err) {
5030 			flags &= ~BRIDGE_FLAGS_SELF;
5031 
5032 			/* Generate event to notify upper layer of bridge
5033 			 * change
5034 			 */
5035 			err = rtnl_bridge_notify(dev);
5036 		}
5037 	}
5038 
5039 	if (have_flags)
5040 		memcpy(nla_data(attr), &flags, sizeof(flags));
5041 out:
5042 	return err;
5043 }
5044 
5045 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5046 {
5047 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5048 	       (!idxattr || idxattr == attrid);
5049 }
5050 
5051 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5052 static int rtnl_get_offload_stats_attr_size(int attr_id)
5053 {
5054 	switch (attr_id) {
5055 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5056 		return sizeof(struct rtnl_link_stats64);
5057 	}
5058 
5059 	return 0;
5060 }
5061 
5062 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5063 				  int *prividx)
5064 {
5065 	struct nlattr *attr = NULL;
5066 	int attr_id, size;
5067 	void *attr_data;
5068 	int err;
5069 
5070 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5071 	      dev->netdev_ops->ndo_get_offload_stats))
5072 		return -ENODATA;
5073 
5074 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5075 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5076 		if (attr_id < *prividx)
5077 			continue;
5078 
5079 		size = rtnl_get_offload_stats_attr_size(attr_id);
5080 		if (!size)
5081 			continue;
5082 
5083 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5084 			continue;
5085 
5086 		attr = nla_reserve_64bit(skb, attr_id, size,
5087 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
5088 		if (!attr)
5089 			goto nla_put_failure;
5090 
5091 		attr_data = nla_data(attr);
5092 		memset(attr_data, 0, size);
5093 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5094 							     attr_data);
5095 		if (err)
5096 			goto get_offload_stats_failure;
5097 	}
5098 
5099 	if (!attr)
5100 		return -ENODATA;
5101 
5102 	*prividx = 0;
5103 	return 0;
5104 
5105 nla_put_failure:
5106 	err = -EMSGSIZE;
5107 get_offload_stats_failure:
5108 	*prividx = attr_id;
5109 	return err;
5110 }
5111 
5112 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5113 {
5114 	int nla_size = 0;
5115 	int attr_id;
5116 	int size;
5117 
5118 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5119 	      dev->netdev_ops->ndo_get_offload_stats))
5120 		return 0;
5121 
5122 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5123 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5124 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5125 			continue;
5126 		size = rtnl_get_offload_stats_attr_size(attr_id);
5127 		nla_size += nla_total_size_64bit(size);
5128 	}
5129 
5130 	if (nla_size != 0)
5131 		nla_size += nla_total_size(0);
5132 
5133 	return nla_size;
5134 }
5135 
5136 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5137 			       int type, u32 pid, u32 seq, u32 change,
5138 			       unsigned int flags, unsigned int filter_mask,
5139 			       int *idxattr, int *prividx)
5140 {
5141 	struct if_stats_msg *ifsm;
5142 	struct nlmsghdr *nlh;
5143 	struct nlattr *attr;
5144 	int s_prividx = *prividx;
5145 	int err;
5146 
5147 	ASSERT_RTNL();
5148 
5149 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5150 	if (!nlh)
5151 		return -EMSGSIZE;
5152 
5153 	ifsm = nlmsg_data(nlh);
5154 	ifsm->family = PF_UNSPEC;
5155 	ifsm->pad1 = 0;
5156 	ifsm->pad2 = 0;
5157 	ifsm->ifindex = dev->ifindex;
5158 	ifsm->filter_mask = filter_mask;
5159 
5160 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5161 		struct rtnl_link_stats64 *sp;
5162 
5163 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5164 					 sizeof(struct rtnl_link_stats64),
5165 					 IFLA_STATS_UNSPEC);
5166 		if (!attr)
5167 			goto nla_put_failure;
5168 
5169 		sp = nla_data(attr);
5170 		dev_get_stats(dev, sp);
5171 	}
5172 
5173 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5174 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5175 
5176 		if (ops && ops->fill_linkxstats) {
5177 			*idxattr = IFLA_STATS_LINK_XSTATS;
5178 			attr = nla_nest_start_noflag(skb,
5179 						     IFLA_STATS_LINK_XSTATS);
5180 			if (!attr)
5181 				goto nla_put_failure;
5182 
5183 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5184 			nla_nest_end(skb, attr);
5185 			if (err)
5186 				goto nla_put_failure;
5187 			*idxattr = 0;
5188 		}
5189 	}
5190 
5191 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5192 			     *idxattr)) {
5193 		const struct rtnl_link_ops *ops = NULL;
5194 		const struct net_device *master;
5195 
5196 		master = netdev_master_upper_dev_get(dev);
5197 		if (master)
5198 			ops = master->rtnl_link_ops;
5199 		if (ops && ops->fill_linkxstats) {
5200 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5201 			attr = nla_nest_start_noflag(skb,
5202 						     IFLA_STATS_LINK_XSTATS_SLAVE);
5203 			if (!attr)
5204 				goto nla_put_failure;
5205 
5206 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5207 			nla_nest_end(skb, attr);
5208 			if (err)
5209 				goto nla_put_failure;
5210 			*idxattr = 0;
5211 		}
5212 	}
5213 
5214 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5215 			     *idxattr)) {
5216 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5217 		attr = nla_nest_start_noflag(skb,
5218 					     IFLA_STATS_LINK_OFFLOAD_XSTATS);
5219 		if (!attr)
5220 			goto nla_put_failure;
5221 
5222 		err = rtnl_get_offload_stats(skb, dev, prividx);
5223 		if (err == -ENODATA)
5224 			nla_nest_cancel(skb, attr);
5225 		else
5226 			nla_nest_end(skb, attr);
5227 
5228 		if (err && err != -ENODATA)
5229 			goto nla_put_failure;
5230 		*idxattr = 0;
5231 	}
5232 
5233 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5234 		struct rtnl_af_ops *af_ops;
5235 
5236 		*idxattr = IFLA_STATS_AF_SPEC;
5237 		attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5238 		if (!attr)
5239 			goto nla_put_failure;
5240 
5241 		rcu_read_lock();
5242 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5243 			if (af_ops->fill_stats_af) {
5244 				struct nlattr *af;
5245 				int err;
5246 
5247 				af = nla_nest_start_noflag(skb,
5248 							   af_ops->family);
5249 				if (!af) {
5250 					rcu_read_unlock();
5251 					goto nla_put_failure;
5252 				}
5253 				err = af_ops->fill_stats_af(skb, dev);
5254 
5255 				if (err == -ENODATA) {
5256 					nla_nest_cancel(skb, af);
5257 				} else if (err < 0) {
5258 					rcu_read_unlock();
5259 					goto nla_put_failure;
5260 				}
5261 
5262 				nla_nest_end(skb, af);
5263 			}
5264 		}
5265 		rcu_read_unlock();
5266 
5267 		nla_nest_end(skb, attr);
5268 
5269 		*idxattr = 0;
5270 	}
5271 
5272 	nlmsg_end(skb, nlh);
5273 
5274 	return 0;
5275 
5276 nla_put_failure:
5277 	/* not a multi message or no progress mean a real error */
5278 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5279 		nlmsg_cancel(skb, nlh);
5280 	else
5281 		nlmsg_end(skb, nlh);
5282 
5283 	return -EMSGSIZE;
5284 }
5285 
5286 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5287 				  u32 filter_mask)
5288 {
5289 	size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5290 
5291 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5292 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5293 
5294 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5295 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5296 		int attr = IFLA_STATS_LINK_XSTATS;
5297 
5298 		if (ops && ops->get_linkxstats_size) {
5299 			size += nla_total_size(ops->get_linkxstats_size(dev,
5300 									attr));
5301 			/* for IFLA_STATS_LINK_XSTATS */
5302 			size += nla_total_size(0);
5303 		}
5304 	}
5305 
5306 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5307 		struct net_device *_dev = (struct net_device *)dev;
5308 		const struct rtnl_link_ops *ops = NULL;
5309 		const struct net_device *master;
5310 
5311 		/* netdev_master_upper_dev_get can't take const */
5312 		master = netdev_master_upper_dev_get(_dev);
5313 		if (master)
5314 			ops = master->rtnl_link_ops;
5315 		if (ops && ops->get_linkxstats_size) {
5316 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5317 
5318 			size += nla_total_size(ops->get_linkxstats_size(dev,
5319 									attr));
5320 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
5321 			size += nla_total_size(0);
5322 		}
5323 	}
5324 
5325 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5326 		size += rtnl_get_offload_stats_size(dev);
5327 
5328 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5329 		struct rtnl_af_ops *af_ops;
5330 
5331 		/* for IFLA_STATS_AF_SPEC */
5332 		size += nla_total_size(0);
5333 
5334 		rcu_read_lock();
5335 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5336 			if (af_ops->get_stats_af_size) {
5337 				size += nla_total_size(
5338 					af_ops->get_stats_af_size(dev));
5339 
5340 				/* for AF_* */
5341 				size += nla_total_size(0);
5342 			}
5343 		}
5344 		rcu_read_unlock();
5345 	}
5346 
5347 	return size;
5348 }
5349 
5350 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5351 				bool is_dump, struct netlink_ext_ack *extack)
5352 {
5353 	struct if_stats_msg *ifsm;
5354 
5355 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5356 		NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5357 		return -EINVAL;
5358 	}
5359 
5360 	if (!strict_check)
5361 		return 0;
5362 
5363 	ifsm = nlmsg_data(nlh);
5364 
5365 	/* only requests using strict checks can pass data to influence
5366 	 * the dump. The legacy exception is filter_mask.
5367 	 */
5368 	if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5369 		NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5370 		return -EINVAL;
5371 	}
5372 	if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5373 		NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5374 		return -EINVAL;
5375 	}
5376 	if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5377 		NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5378 		return -EINVAL;
5379 	}
5380 
5381 	return 0;
5382 }
5383 
5384 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5385 			  struct netlink_ext_ack *extack)
5386 {
5387 	struct net *net = sock_net(skb->sk);
5388 	struct net_device *dev = NULL;
5389 	int idxattr = 0, prividx = 0;
5390 	struct if_stats_msg *ifsm;
5391 	struct sk_buff *nskb;
5392 	u32 filter_mask;
5393 	int err;
5394 
5395 	err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5396 				   false, extack);
5397 	if (err)
5398 		return err;
5399 
5400 	ifsm = nlmsg_data(nlh);
5401 	if (ifsm->ifindex > 0)
5402 		dev = __dev_get_by_index(net, ifsm->ifindex);
5403 	else
5404 		return -EINVAL;
5405 
5406 	if (!dev)
5407 		return -ENODEV;
5408 
5409 	filter_mask = ifsm->filter_mask;
5410 	if (!filter_mask) {
5411 		NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5412 		return -EINVAL;
5413 	}
5414 
5415 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5416 	if (!nskb)
5417 		return -ENOBUFS;
5418 
5419 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5420 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5421 				  0, filter_mask, &idxattr, &prividx);
5422 	if (err < 0) {
5423 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5424 		WARN_ON(err == -EMSGSIZE);
5425 		kfree_skb(nskb);
5426 	} else {
5427 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5428 	}
5429 
5430 	return err;
5431 }
5432 
5433 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5434 {
5435 	struct netlink_ext_ack *extack = cb->extack;
5436 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
5437 	struct net *net = sock_net(skb->sk);
5438 	unsigned int flags = NLM_F_MULTI;
5439 	struct if_stats_msg *ifsm;
5440 	struct hlist_head *head;
5441 	struct net_device *dev;
5442 	u32 filter_mask = 0;
5443 	int idx = 0;
5444 
5445 	s_h = cb->args[0];
5446 	s_idx = cb->args[1];
5447 	s_idxattr = cb->args[2];
5448 	s_prividx = cb->args[3];
5449 
5450 	cb->seq = net->dev_base_seq;
5451 
5452 	err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5453 	if (err)
5454 		return err;
5455 
5456 	ifsm = nlmsg_data(cb->nlh);
5457 	filter_mask = ifsm->filter_mask;
5458 	if (!filter_mask) {
5459 		NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5460 		return -EINVAL;
5461 	}
5462 
5463 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5464 		idx = 0;
5465 		head = &net->dev_index_head[h];
5466 		hlist_for_each_entry(dev, head, index_hlist) {
5467 			if (idx < s_idx)
5468 				goto cont;
5469 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5470 						  NETLINK_CB(cb->skb).portid,
5471 						  cb->nlh->nlmsg_seq, 0,
5472 						  flags, filter_mask,
5473 						  &s_idxattr, &s_prividx);
5474 			/* If we ran out of room on the first message,
5475 			 * we're in trouble
5476 			 */
5477 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5478 
5479 			if (err < 0)
5480 				goto out;
5481 			s_prividx = 0;
5482 			s_idxattr = 0;
5483 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5484 cont:
5485 			idx++;
5486 		}
5487 	}
5488 out:
5489 	cb->args[3] = s_prividx;
5490 	cb->args[2] = s_idxattr;
5491 	cb->args[1] = idx;
5492 	cb->args[0] = h;
5493 
5494 	return skb->len;
5495 }
5496 
5497 /* Process one rtnetlink message. */
5498 
5499 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5500 			     struct netlink_ext_ack *extack)
5501 {
5502 	struct net *net = sock_net(skb->sk);
5503 	struct rtnl_link *link;
5504 	struct module *owner;
5505 	int err = -EOPNOTSUPP;
5506 	rtnl_doit_func doit;
5507 	unsigned int flags;
5508 	int kind;
5509 	int family;
5510 	int type;
5511 
5512 	type = nlh->nlmsg_type;
5513 	if (type > RTM_MAX)
5514 		return -EOPNOTSUPP;
5515 
5516 	type -= RTM_BASE;
5517 
5518 	/* All the messages must have at least 1 byte length */
5519 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5520 		return 0;
5521 
5522 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5523 	kind = type&3;
5524 
5525 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5526 		return -EPERM;
5527 
5528 	rcu_read_lock();
5529 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5530 		struct sock *rtnl;
5531 		rtnl_dumpit_func dumpit;
5532 		u32 min_dump_alloc = 0;
5533 
5534 		link = rtnl_get_link(family, type);
5535 		if (!link || !link->dumpit) {
5536 			family = PF_UNSPEC;
5537 			link = rtnl_get_link(family, type);
5538 			if (!link || !link->dumpit)
5539 				goto err_unlock;
5540 		}
5541 		owner = link->owner;
5542 		dumpit = link->dumpit;
5543 
5544 		if (type == RTM_GETLINK - RTM_BASE)
5545 			min_dump_alloc = rtnl_calcit(skb, nlh);
5546 
5547 		err = 0;
5548 		/* need to do this before rcu_read_unlock() */
5549 		if (!try_module_get(owner))
5550 			err = -EPROTONOSUPPORT;
5551 
5552 		rcu_read_unlock();
5553 
5554 		rtnl = net->rtnl;
5555 		if (err == 0) {
5556 			struct netlink_dump_control c = {
5557 				.dump		= dumpit,
5558 				.min_dump_alloc	= min_dump_alloc,
5559 				.module		= owner,
5560 			};
5561 			err = netlink_dump_start(rtnl, skb, nlh, &c);
5562 			/* netlink_dump_start() will keep a reference on
5563 			 * module if dump is still in progress.
5564 			 */
5565 			module_put(owner);
5566 		}
5567 		return err;
5568 	}
5569 
5570 	link = rtnl_get_link(family, type);
5571 	if (!link || !link->doit) {
5572 		family = PF_UNSPEC;
5573 		link = rtnl_get_link(PF_UNSPEC, type);
5574 		if (!link || !link->doit)
5575 			goto out_unlock;
5576 	}
5577 
5578 	owner = link->owner;
5579 	if (!try_module_get(owner)) {
5580 		err = -EPROTONOSUPPORT;
5581 		goto out_unlock;
5582 	}
5583 
5584 	flags = link->flags;
5585 	if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5586 		doit = link->doit;
5587 		rcu_read_unlock();
5588 		if (doit)
5589 			err = doit(skb, nlh, extack);
5590 		module_put(owner);
5591 		return err;
5592 	}
5593 	rcu_read_unlock();
5594 
5595 	rtnl_lock();
5596 	link = rtnl_get_link(family, type);
5597 	if (link && link->doit)
5598 		err = link->doit(skb, nlh, extack);
5599 	rtnl_unlock();
5600 
5601 	module_put(owner);
5602 
5603 	return err;
5604 
5605 out_unlock:
5606 	rcu_read_unlock();
5607 	return err;
5608 
5609 err_unlock:
5610 	rcu_read_unlock();
5611 	return -EOPNOTSUPP;
5612 }
5613 
5614 static void rtnetlink_rcv(struct sk_buff *skb)
5615 {
5616 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5617 }
5618 
5619 static int rtnetlink_bind(struct net *net, int group)
5620 {
5621 	switch (group) {
5622 	case RTNLGRP_IPV4_MROUTE_R:
5623 	case RTNLGRP_IPV6_MROUTE_R:
5624 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5625 			return -EPERM;
5626 		break;
5627 	}
5628 	return 0;
5629 }
5630 
5631 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5632 {
5633 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5634 
5635 	switch (event) {
5636 	case NETDEV_REBOOT:
5637 	case NETDEV_CHANGEMTU:
5638 	case NETDEV_CHANGEADDR:
5639 	case NETDEV_CHANGENAME:
5640 	case NETDEV_FEAT_CHANGE:
5641 	case NETDEV_BONDING_FAILOVER:
5642 	case NETDEV_POST_TYPE_CHANGE:
5643 	case NETDEV_NOTIFY_PEERS:
5644 	case NETDEV_CHANGEUPPER:
5645 	case NETDEV_RESEND_IGMP:
5646 	case NETDEV_CHANGEINFODATA:
5647 	case NETDEV_CHANGELOWERSTATE:
5648 	case NETDEV_CHANGE_TX_QUEUE_LEN:
5649 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5650 				   GFP_KERNEL, NULL, 0);
5651 		break;
5652 	default:
5653 		break;
5654 	}
5655 	return NOTIFY_DONE;
5656 }
5657 
5658 static struct notifier_block rtnetlink_dev_notifier = {
5659 	.notifier_call	= rtnetlink_event,
5660 };
5661 
5662 
5663 static int __net_init rtnetlink_net_init(struct net *net)
5664 {
5665 	struct sock *sk;
5666 	struct netlink_kernel_cfg cfg = {
5667 		.groups		= RTNLGRP_MAX,
5668 		.input		= rtnetlink_rcv,
5669 		.cb_mutex	= &rtnl_mutex,
5670 		.flags		= NL_CFG_F_NONROOT_RECV,
5671 		.bind		= rtnetlink_bind,
5672 	};
5673 
5674 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5675 	if (!sk)
5676 		return -ENOMEM;
5677 	net->rtnl = sk;
5678 	return 0;
5679 }
5680 
5681 static void __net_exit rtnetlink_net_exit(struct net *net)
5682 {
5683 	netlink_kernel_release(net->rtnl);
5684 	net->rtnl = NULL;
5685 }
5686 
5687 static struct pernet_operations rtnetlink_net_ops = {
5688 	.init = rtnetlink_net_init,
5689 	.exit = rtnetlink_net_exit,
5690 };
5691 
5692 void __init rtnetlink_init(void)
5693 {
5694 	if (register_pernet_subsys(&rtnetlink_net_ops))
5695 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
5696 
5697 	register_netdevice_notifier(&rtnetlink_dev_notifier);
5698 
5699 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5700 		      rtnl_dump_ifinfo, 0);
5701 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5702 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5703 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5704 
5705 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5706 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5707 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5708 
5709 	rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5710 	rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5711 
5712 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5713 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5714 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5715 
5716 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5717 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5718 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5719 
5720 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5721 		      0);
5722 }
5723