xref: /openbmc/linux/net/core/rtnetlink.c (revision ae213c44)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  *
15  *	Fixes:
16  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17  */
18 
19 #include <linux/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44 
45 #include <linux/uaccess.h>
46 
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/ip.h>
50 #include <net/protocol.h>
51 #include <net/arp.h>
52 #include <net/route.h>
53 #include <net/udp.h>
54 #include <net/tcp.h>
55 #include <net/sock.h>
56 #include <net/pkt_sched.h>
57 #include <net/fib_rules.h>
58 #include <net/rtnetlink.h>
59 #include <net/net_namespace.h>
60 
61 #define RTNL_MAX_TYPE		50
62 #define RTNL_SLAVE_MAX_TYPE	36
63 
64 struct rtnl_link {
65 	rtnl_doit_func		doit;
66 	rtnl_dumpit_func	dumpit;
67 	struct module		*owner;
68 	unsigned int		flags;
69 	struct rcu_head		rcu;
70 };
71 
72 static DEFINE_MUTEX(rtnl_mutex);
73 
74 void rtnl_lock(void)
75 {
76 	mutex_lock(&rtnl_mutex);
77 }
78 EXPORT_SYMBOL(rtnl_lock);
79 
80 int rtnl_lock_killable(void)
81 {
82 	return mutex_lock_killable(&rtnl_mutex);
83 }
84 EXPORT_SYMBOL(rtnl_lock_killable);
85 
86 static struct sk_buff *defer_kfree_skb_list;
87 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
88 {
89 	if (head && tail) {
90 		tail->next = defer_kfree_skb_list;
91 		defer_kfree_skb_list = head;
92 	}
93 }
94 EXPORT_SYMBOL(rtnl_kfree_skbs);
95 
96 void __rtnl_unlock(void)
97 {
98 	struct sk_buff *head = defer_kfree_skb_list;
99 
100 	defer_kfree_skb_list = NULL;
101 
102 	mutex_unlock(&rtnl_mutex);
103 
104 	while (head) {
105 		struct sk_buff *next = head->next;
106 
107 		kfree_skb(head);
108 		cond_resched();
109 		head = next;
110 	}
111 }
112 
113 void rtnl_unlock(void)
114 {
115 	/* This fellow will unlock it for us. */
116 	netdev_run_todo();
117 }
118 EXPORT_SYMBOL(rtnl_unlock);
119 
120 int rtnl_trylock(void)
121 {
122 	return mutex_trylock(&rtnl_mutex);
123 }
124 EXPORT_SYMBOL(rtnl_trylock);
125 
126 int rtnl_is_locked(void)
127 {
128 	return mutex_is_locked(&rtnl_mutex);
129 }
130 EXPORT_SYMBOL(rtnl_is_locked);
131 
132 bool refcount_dec_and_rtnl_lock(refcount_t *r)
133 {
134 	return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
135 }
136 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
137 
138 #ifdef CONFIG_PROVE_LOCKING
139 bool lockdep_rtnl_is_held(void)
140 {
141 	return lockdep_is_held(&rtnl_mutex);
142 }
143 EXPORT_SYMBOL(lockdep_rtnl_is_held);
144 #endif /* #ifdef CONFIG_PROVE_LOCKING */
145 
146 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
147 
148 static inline int rtm_msgindex(int msgtype)
149 {
150 	int msgindex = msgtype - RTM_BASE;
151 
152 	/*
153 	 * msgindex < 0 implies someone tried to register a netlink
154 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
155 	 * the message type has not been added to linux/rtnetlink.h
156 	 */
157 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
158 
159 	return msgindex;
160 }
161 
162 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
163 {
164 	struct rtnl_link **tab;
165 
166 	if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
167 		protocol = PF_UNSPEC;
168 
169 	tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
170 	if (!tab)
171 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
172 
173 	return tab[msgtype];
174 }
175 
176 static int rtnl_register_internal(struct module *owner,
177 				  int protocol, int msgtype,
178 				  rtnl_doit_func doit, rtnl_dumpit_func dumpit,
179 				  unsigned int flags)
180 {
181 	struct rtnl_link *link, *old;
182 	struct rtnl_link __rcu **tab;
183 	int msgindex;
184 	int ret = -ENOBUFS;
185 
186 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
187 	msgindex = rtm_msgindex(msgtype);
188 
189 	rtnl_lock();
190 	tab = rtnl_msg_handlers[protocol];
191 	if (tab == NULL) {
192 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
193 		if (!tab)
194 			goto unlock;
195 
196 		/* ensures we see the 0 stores */
197 		rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
198 	}
199 
200 	old = rtnl_dereference(tab[msgindex]);
201 	if (old) {
202 		link = kmemdup(old, sizeof(*old), GFP_KERNEL);
203 		if (!link)
204 			goto unlock;
205 	} else {
206 		link = kzalloc(sizeof(*link), GFP_KERNEL);
207 		if (!link)
208 			goto unlock;
209 	}
210 
211 	WARN_ON(link->owner && link->owner != owner);
212 	link->owner = owner;
213 
214 	WARN_ON(doit && link->doit && link->doit != doit);
215 	if (doit)
216 		link->doit = doit;
217 	WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
218 	if (dumpit)
219 		link->dumpit = dumpit;
220 
221 	link->flags |= flags;
222 
223 	/* publish protocol:msgtype */
224 	rcu_assign_pointer(tab[msgindex], link);
225 	ret = 0;
226 	if (old)
227 		kfree_rcu(old, rcu);
228 unlock:
229 	rtnl_unlock();
230 	return ret;
231 }
232 
233 /**
234  * rtnl_register_module - Register a rtnetlink message type
235  *
236  * @owner: module registering the hook (THIS_MODULE)
237  * @protocol: Protocol family or PF_UNSPEC
238  * @msgtype: rtnetlink message type
239  * @doit: Function pointer called for each request message
240  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
241  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
242  *
243  * Like rtnl_register, but for use by removable modules.
244  */
245 int rtnl_register_module(struct module *owner,
246 			 int protocol, int msgtype,
247 			 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
248 			 unsigned int flags)
249 {
250 	return rtnl_register_internal(owner, protocol, msgtype,
251 				      doit, dumpit, flags);
252 }
253 EXPORT_SYMBOL_GPL(rtnl_register_module);
254 
255 /**
256  * rtnl_register - Register a rtnetlink message type
257  * @protocol: Protocol family or PF_UNSPEC
258  * @msgtype: rtnetlink message type
259  * @doit: Function pointer called for each request message
260  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
261  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
262  *
263  * Registers the specified function pointers (at least one of them has
264  * to be non-NULL) to be called whenever a request message for the
265  * specified protocol family and message type is received.
266  *
267  * The special protocol family PF_UNSPEC may be used to define fallback
268  * function pointers for the case when no entry for the specific protocol
269  * family exists.
270  */
271 void rtnl_register(int protocol, int msgtype,
272 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
273 		   unsigned int flags)
274 {
275 	int err;
276 
277 	err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
278 				     flags);
279 	if (err)
280 		pr_err("Unable to register rtnetlink message handler, "
281 		       "protocol = %d, message type = %d\n", protocol, msgtype);
282 }
283 
284 /**
285  * rtnl_unregister - Unregister a rtnetlink message type
286  * @protocol: Protocol family or PF_UNSPEC
287  * @msgtype: rtnetlink message type
288  *
289  * Returns 0 on success or a negative error code.
290  */
291 int rtnl_unregister(int protocol, int msgtype)
292 {
293 	struct rtnl_link **tab, *link;
294 	int msgindex;
295 
296 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
297 	msgindex = rtm_msgindex(msgtype);
298 
299 	rtnl_lock();
300 	tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
301 	if (!tab) {
302 		rtnl_unlock();
303 		return -ENOENT;
304 	}
305 
306 	link = tab[msgindex];
307 	rcu_assign_pointer(tab[msgindex], NULL);
308 	rtnl_unlock();
309 
310 	kfree_rcu(link, rcu);
311 
312 	return 0;
313 }
314 EXPORT_SYMBOL_GPL(rtnl_unregister);
315 
316 /**
317  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
318  * @protocol : Protocol family or PF_UNSPEC
319  *
320  * Identical to calling rtnl_unregster() for all registered message types
321  * of a certain protocol family.
322  */
323 void rtnl_unregister_all(int protocol)
324 {
325 	struct rtnl_link **tab, *link;
326 	int msgindex;
327 
328 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
329 
330 	rtnl_lock();
331 	tab = rtnl_msg_handlers[protocol];
332 	if (!tab) {
333 		rtnl_unlock();
334 		return;
335 	}
336 	RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
337 	for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
338 		link = tab[msgindex];
339 		if (!link)
340 			continue;
341 
342 		rcu_assign_pointer(tab[msgindex], NULL);
343 		kfree_rcu(link, rcu);
344 	}
345 	rtnl_unlock();
346 
347 	synchronize_net();
348 
349 	kfree(tab);
350 }
351 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
352 
353 static LIST_HEAD(link_ops);
354 
355 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
356 {
357 	const struct rtnl_link_ops *ops;
358 
359 	list_for_each_entry(ops, &link_ops, list) {
360 		if (!strcmp(ops->kind, kind))
361 			return ops;
362 	}
363 	return NULL;
364 }
365 
366 /**
367  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
368  * @ops: struct rtnl_link_ops * to register
369  *
370  * The caller must hold the rtnl_mutex. This function should be used
371  * by drivers that create devices during module initialization. It
372  * must be called before registering the devices.
373  *
374  * Returns 0 on success or a negative error code.
375  */
376 int __rtnl_link_register(struct rtnl_link_ops *ops)
377 {
378 	if (rtnl_link_ops_get(ops->kind))
379 		return -EEXIST;
380 
381 	/* The check for setup is here because if ops
382 	 * does not have that filled up, it is not possible
383 	 * to use the ops for creating device. So do not
384 	 * fill up dellink as well. That disables rtnl_dellink.
385 	 */
386 	if (ops->setup && !ops->dellink)
387 		ops->dellink = unregister_netdevice_queue;
388 
389 	list_add_tail(&ops->list, &link_ops);
390 	return 0;
391 }
392 EXPORT_SYMBOL_GPL(__rtnl_link_register);
393 
394 /**
395  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
396  * @ops: struct rtnl_link_ops * to register
397  *
398  * Returns 0 on success or a negative error code.
399  */
400 int rtnl_link_register(struct rtnl_link_ops *ops)
401 {
402 	int err;
403 
404 	/* Sanity-check max sizes to avoid stack buffer overflow. */
405 	if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
406 		    ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
407 		return -EINVAL;
408 
409 	rtnl_lock();
410 	err = __rtnl_link_register(ops);
411 	rtnl_unlock();
412 	return err;
413 }
414 EXPORT_SYMBOL_GPL(rtnl_link_register);
415 
416 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
417 {
418 	struct net_device *dev;
419 	LIST_HEAD(list_kill);
420 
421 	for_each_netdev(net, dev) {
422 		if (dev->rtnl_link_ops == ops)
423 			ops->dellink(dev, &list_kill);
424 	}
425 	unregister_netdevice_many(&list_kill);
426 }
427 
428 /**
429  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
430  * @ops: struct rtnl_link_ops * to unregister
431  *
432  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
433  * integrity (hold pernet_ops_rwsem for writing to close the race
434  * with setup_net() and cleanup_net()).
435  */
436 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
437 {
438 	struct net *net;
439 
440 	for_each_net(net) {
441 		__rtnl_kill_links(net, ops);
442 	}
443 	list_del(&ops->list);
444 }
445 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
446 
447 /* Return with the rtnl_lock held when there are no network
448  * devices unregistering in any network namespace.
449  */
450 static void rtnl_lock_unregistering_all(void)
451 {
452 	struct net *net;
453 	bool unregistering;
454 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
455 
456 	add_wait_queue(&netdev_unregistering_wq, &wait);
457 	for (;;) {
458 		unregistering = false;
459 		rtnl_lock();
460 		/* We held write locked pernet_ops_rwsem, and parallel
461 		 * setup_net() and cleanup_net() are not possible.
462 		 */
463 		for_each_net(net) {
464 			if (net->dev_unreg_count > 0) {
465 				unregistering = true;
466 				break;
467 			}
468 		}
469 		if (!unregistering)
470 			break;
471 		__rtnl_unlock();
472 
473 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
474 	}
475 	remove_wait_queue(&netdev_unregistering_wq, &wait);
476 }
477 
478 /**
479  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
480  * @ops: struct rtnl_link_ops * to unregister
481  */
482 void rtnl_link_unregister(struct rtnl_link_ops *ops)
483 {
484 	/* Close the race with setup_net() and cleanup_net() */
485 	down_write(&pernet_ops_rwsem);
486 	rtnl_lock_unregistering_all();
487 	__rtnl_link_unregister(ops);
488 	rtnl_unlock();
489 	up_write(&pernet_ops_rwsem);
490 }
491 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
492 
493 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
494 {
495 	struct net_device *master_dev;
496 	const struct rtnl_link_ops *ops;
497 	size_t size = 0;
498 
499 	rcu_read_lock();
500 
501 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
502 	if (!master_dev)
503 		goto out;
504 
505 	ops = master_dev->rtnl_link_ops;
506 	if (!ops || !ops->get_slave_size)
507 		goto out;
508 	/* IFLA_INFO_SLAVE_DATA + nested data */
509 	size = nla_total_size(sizeof(struct nlattr)) +
510 	       ops->get_slave_size(master_dev, dev);
511 
512 out:
513 	rcu_read_unlock();
514 	return size;
515 }
516 
517 static size_t rtnl_link_get_size(const struct net_device *dev)
518 {
519 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
520 	size_t size;
521 
522 	if (!ops)
523 		return 0;
524 
525 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
526 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
527 
528 	if (ops->get_size)
529 		/* IFLA_INFO_DATA + nested data */
530 		size += nla_total_size(sizeof(struct nlattr)) +
531 			ops->get_size(dev);
532 
533 	if (ops->get_xstats_size)
534 		/* IFLA_INFO_XSTATS */
535 		size += nla_total_size(ops->get_xstats_size(dev));
536 
537 	size += rtnl_link_get_slave_info_data_size(dev);
538 
539 	return size;
540 }
541 
542 static LIST_HEAD(rtnl_af_ops);
543 
544 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
545 {
546 	const struct rtnl_af_ops *ops;
547 
548 	list_for_each_entry_rcu(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 	int err = 0;
714 
715 	NETLINK_CB(skb).dst_group = group;
716 	if (echo)
717 		refcount_inc(&skb->users);
718 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
719 	if (echo)
720 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
721 	return err;
722 }
723 
724 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
725 {
726 	struct sock *rtnl = net->rtnl;
727 
728 	return nlmsg_unicast(rtnl, skb, pid);
729 }
730 EXPORT_SYMBOL(rtnl_unicast);
731 
732 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
733 		 struct nlmsghdr *nlh, gfp_t flags)
734 {
735 	struct sock *rtnl = net->rtnl;
736 	int report = 0;
737 
738 	if (nlh)
739 		report = nlmsg_report(nlh);
740 
741 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
742 }
743 EXPORT_SYMBOL(rtnl_notify);
744 
745 void rtnl_set_sk_err(struct net *net, u32 group, int error)
746 {
747 	struct sock *rtnl = net->rtnl;
748 
749 	netlink_set_err(rtnl, 0, group, error);
750 }
751 EXPORT_SYMBOL(rtnl_set_sk_err);
752 
753 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
754 {
755 	struct nlattr *mx;
756 	int i, valid = 0;
757 
758 	mx = nla_nest_start_noflag(skb, RTA_METRICS);
759 	if (mx == NULL)
760 		return -ENOBUFS;
761 
762 	for (i = 0; i < RTAX_MAX; i++) {
763 		if (metrics[i]) {
764 			if (i == RTAX_CC_ALGO - 1) {
765 				char tmp[TCP_CA_NAME_MAX], *name;
766 
767 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
768 				if (!name)
769 					continue;
770 				if (nla_put_string(skb, i + 1, name))
771 					goto nla_put_failure;
772 			} else if (i == RTAX_FEATURES - 1) {
773 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
774 
775 				if (!user_features)
776 					continue;
777 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
778 				if (nla_put_u32(skb, i + 1, user_features))
779 					goto nla_put_failure;
780 			} else {
781 				if (nla_put_u32(skb, i + 1, metrics[i]))
782 					goto nla_put_failure;
783 			}
784 			valid++;
785 		}
786 	}
787 
788 	if (!valid) {
789 		nla_nest_cancel(skb, mx);
790 		return 0;
791 	}
792 
793 	return nla_nest_end(skb, mx);
794 
795 nla_put_failure:
796 	nla_nest_cancel(skb, mx);
797 	return -EMSGSIZE;
798 }
799 EXPORT_SYMBOL(rtnetlink_put_metrics);
800 
801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
802 		       long expires, u32 error)
803 {
804 	struct rta_cacheinfo ci = {
805 		.rta_error = error,
806 		.rta_id =  id,
807 	};
808 
809 	if (dst) {
810 		ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
811 		ci.rta_used = dst->__use;
812 		ci.rta_clntref = atomic_read(&dst->__refcnt);
813 	}
814 	if (expires) {
815 		unsigned long clock;
816 
817 		clock = jiffies_to_clock_t(abs(expires));
818 		clock = min_t(unsigned long, clock, INT_MAX);
819 		ci.rta_expires = (expires > 0) ? clock : -clock;
820 	}
821 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
822 }
823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
824 
825 static void set_operstate(struct net_device *dev, unsigned char transition)
826 {
827 	unsigned char operstate = dev->operstate;
828 
829 	switch (transition) {
830 	case IF_OPER_UP:
831 		if ((operstate == IF_OPER_DORMANT ||
832 		     operstate == IF_OPER_UNKNOWN) &&
833 		    !netif_dormant(dev))
834 			operstate = IF_OPER_UP;
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_bh(&dev_base_lock);
846 		dev->operstate = operstate;
847 		write_unlock_bh(&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_vlan)) +
916 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
917 			 nla_total_size(MAX_VLAN_LIST_LEN *
918 					sizeof(struct ifla_vf_vlan_info)) +
919 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
920 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
921 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
922 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
923 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
924 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
925 			 /* IFLA_VF_STATS_RX_PACKETS */
926 			 nla_total_size_64bit(sizeof(__u64)) +
927 			 /* IFLA_VF_STATS_TX_PACKETS */
928 			 nla_total_size_64bit(sizeof(__u64)) +
929 			 /* IFLA_VF_STATS_RX_BYTES */
930 			 nla_total_size_64bit(sizeof(__u64)) +
931 			 /* IFLA_VF_STATS_TX_BYTES */
932 			 nla_total_size_64bit(sizeof(__u64)) +
933 			 /* IFLA_VF_STATS_BROADCAST */
934 			 nla_total_size_64bit(sizeof(__u64)) +
935 			 /* IFLA_VF_STATS_MULTICAST */
936 			 nla_total_size_64bit(sizeof(__u64)) +
937 			 /* IFLA_VF_STATS_RX_DROPPED */
938 			 nla_total_size_64bit(sizeof(__u64)) +
939 			 /* IFLA_VF_STATS_TX_DROPPED */
940 			 nla_total_size_64bit(sizeof(__u64)) +
941 			 nla_total_size(sizeof(struct ifla_vf_trust)));
942 		return size;
943 	} else
944 		return 0;
945 }
946 
947 static size_t rtnl_port_size(const struct net_device *dev,
948 			     u32 ext_filter_mask)
949 {
950 	size_t port_size = nla_total_size(4)		/* PORT_VF */
951 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
952 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
953 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
954 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
955 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
956 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
957 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
958 		+ port_size;
959 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
960 		+ port_size;
961 
962 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
963 	    !(ext_filter_mask & RTEXT_FILTER_VF))
964 		return 0;
965 	if (dev_num_vf(dev->dev.parent))
966 		return port_self_size + vf_ports_size +
967 			vf_port_size * dev_num_vf(dev->dev.parent);
968 	else
969 		return port_self_size;
970 }
971 
972 static size_t rtnl_xdp_size(void)
973 {
974 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
975 			  nla_total_size(1) +	/* XDP_ATTACHED */
976 			  nla_total_size(4) +	/* XDP_PROG_ID (or 1st mode) */
977 			  nla_total_size(4);	/* XDP_<mode>_PROG_ID */
978 
979 	return xdp_size;
980 }
981 
982 static noinline size_t if_nlmsg_size(const struct net_device *dev,
983 				     u32 ext_filter_mask)
984 {
985 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
986 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
987 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
988 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
989 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
990 	       + nla_total_size(sizeof(struct rtnl_link_stats))
991 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
992 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
993 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
994 	       + nla_total_size(4) /* IFLA_TXQLEN */
995 	       + nla_total_size(4) /* IFLA_WEIGHT */
996 	       + nla_total_size(4) /* IFLA_MTU */
997 	       + nla_total_size(4) /* IFLA_LINK */
998 	       + nla_total_size(4) /* IFLA_MASTER */
999 	       + nla_total_size(1) /* IFLA_CARRIER */
1000 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
1001 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1002 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1003 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1004 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1005 	       + nla_total_size(1) /* IFLA_OPERSTATE */
1006 	       + nla_total_size(1) /* IFLA_LINKMODE */
1007 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1008 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
1009 	       + nla_total_size(4) /* IFLA_GROUP */
1010 	       + nla_total_size(ext_filter_mask
1011 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1012 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1013 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1014 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1015 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1016 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1017 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1018 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1019 	       + rtnl_xdp_size() /* IFLA_XDP */
1020 	       + nla_total_size(4)  /* IFLA_EVENT */
1021 	       + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1022 	       + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1023 	       + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1024 	       + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1025 	       + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1026 	       + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1027 	       + nla_total_size(4)  /* IFLA_MIN_MTU */
1028 	       + nla_total_size(4)  /* IFLA_MAX_MTU */
1029 	       + 0;
1030 }
1031 
1032 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1033 {
1034 	struct nlattr *vf_ports;
1035 	struct nlattr *vf_port;
1036 	int vf;
1037 	int err;
1038 
1039 	vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1040 	if (!vf_ports)
1041 		return -EMSGSIZE;
1042 
1043 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1044 		vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1045 		if (!vf_port)
1046 			goto nla_put_failure;
1047 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1048 			goto nla_put_failure;
1049 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1050 		if (err == -EMSGSIZE)
1051 			goto nla_put_failure;
1052 		if (err) {
1053 			nla_nest_cancel(skb, vf_port);
1054 			continue;
1055 		}
1056 		nla_nest_end(skb, vf_port);
1057 	}
1058 
1059 	nla_nest_end(skb, vf_ports);
1060 
1061 	return 0;
1062 
1063 nla_put_failure:
1064 	nla_nest_cancel(skb, vf_ports);
1065 	return -EMSGSIZE;
1066 }
1067 
1068 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1069 {
1070 	struct nlattr *port_self;
1071 	int err;
1072 
1073 	port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1074 	if (!port_self)
1075 		return -EMSGSIZE;
1076 
1077 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1078 	if (err) {
1079 		nla_nest_cancel(skb, port_self);
1080 		return (err == -EMSGSIZE) ? err : 0;
1081 	}
1082 
1083 	nla_nest_end(skb, port_self);
1084 
1085 	return 0;
1086 }
1087 
1088 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1089 			  u32 ext_filter_mask)
1090 {
1091 	int err;
1092 
1093 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1094 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1095 		return 0;
1096 
1097 	err = rtnl_port_self_fill(skb, dev);
1098 	if (err)
1099 		return err;
1100 
1101 	if (dev_num_vf(dev->dev.parent)) {
1102 		err = rtnl_vf_ports_fill(skb, dev);
1103 		if (err)
1104 			return err;
1105 	}
1106 
1107 	return 0;
1108 }
1109 
1110 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1111 {
1112 	int err;
1113 	struct netdev_phys_item_id ppid;
1114 
1115 	err = dev_get_phys_port_id(dev, &ppid);
1116 	if (err) {
1117 		if (err == -EOPNOTSUPP)
1118 			return 0;
1119 		return err;
1120 	}
1121 
1122 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1123 		return -EMSGSIZE;
1124 
1125 	return 0;
1126 }
1127 
1128 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1129 {
1130 	char name[IFNAMSIZ];
1131 	int err;
1132 
1133 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1134 	if (err) {
1135 		if (err == -EOPNOTSUPP)
1136 			return 0;
1137 		return err;
1138 	}
1139 
1140 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1141 		return -EMSGSIZE;
1142 
1143 	return 0;
1144 }
1145 
1146 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1147 {
1148 	struct netdev_phys_item_id ppid = { };
1149 	int err;
1150 
1151 	err = dev_get_port_parent_id(dev, &ppid, false);
1152 	if (err) {
1153 		if (err == -EOPNOTSUPP)
1154 			return 0;
1155 		return err;
1156 	}
1157 
1158 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1159 		return -EMSGSIZE;
1160 
1161 	return 0;
1162 }
1163 
1164 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1165 					      struct net_device *dev)
1166 {
1167 	struct rtnl_link_stats64 *sp;
1168 	struct nlattr *attr;
1169 
1170 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1171 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1172 	if (!attr)
1173 		return -EMSGSIZE;
1174 
1175 	sp = nla_data(attr);
1176 	dev_get_stats(dev, sp);
1177 
1178 	attr = nla_reserve(skb, IFLA_STATS,
1179 			   sizeof(struct rtnl_link_stats));
1180 	if (!attr)
1181 		return -EMSGSIZE;
1182 
1183 	copy_rtnl_link_stats(nla_data(attr), sp);
1184 
1185 	return 0;
1186 }
1187 
1188 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1189 					       struct net_device *dev,
1190 					       int vfs_num,
1191 					       struct nlattr *vfinfo)
1192 {
1193 	struct ifla_vf_rss_query_en vf_rss_query_en;
1194 	struct nlattr *vf, *vfstats, *vfvlanlist;
1195 	struct ifla_vf_link_state vf_linkstate;
1196 	struct ifla_vf_vlan_info vf_vlan_info;
1197 	struct ifla_vf_spoofchk vf_spoofchk;
1198 	struct ifla_vf_tx_rate vf_tx_rate;
1199 	struct ifla_vf_stats vf_stats;
1200 	struct ifla_vf_trust vf_trust;
1201 	struct ifla_vf_vlan vf_vlan;
1202 	struct ifla_vf_rate vf_rate;
1203 	struct ifla_vf_mac vf_mac;
1204 	struct ifla_vf_info ivi;
1205 
1206 	memset(&ivi, 0, sizeof(ivi));
1207 
1208 	/* Not all SR-IOV capable drivers support the
1209 	 * spoofcheck and "RSS query enable" query.  Preset to
1210 	 * -1 so the user space tool can detect that the driver
1211 	 * didn't report anything.
1212 	 */
1213 	ivi.spoofchk = -1;
1214 	ivi.rss_query_en = -1;
1215 	ivi.trusted = -1;
1216 	/* The default value for VF link state is "auto"
1217 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1218 	 */
1219 	ivi.linkstate = 0;
1220 	/* VLAN Protocol by default is 802.1Q */
1221 	ivi.vlan_proto = htons(ETH_P_8021Q);
1222 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1223 		return 0;
1224 
1225 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1226 
1227 	vf_mac.vf =
1228 		vf_vlan.vf =
1229 		vf_vlan_info.vf =
1230 		vf_rate.vf =
1231 		vf_tx_rate.vf =
1232 		vf_spoofchk.vf =
1233 		vf_linkstate.vf =
1234 		vf_rss_query_en.vf =
1235 		vf_trust.vf = ivi.vf;
1236 
1237 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1238 	vf_vlan.vlan = ivi.vlan;
1239 	vf_vlan.qos = ivi.qos;
1240 	vf_vlan_info.vlan = ivi.vlan;
1241 	vf_vlan_info.qos = ivi.qos;
1242 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1243 	vf_tx_rate.rate = ivi.max_tx_rate;
1244 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1245 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1246 	vf_spoofchk.setting = ivi.spoofchk;
1247 	vf_linkstate.link_state = ivi.linkstate;
1248 	vf_rss_query_en.setting = ivi.rss_query_en;
1249 	vf_trust.setting = ivi.trusted;
1250 	vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1251 	if (!vf)
1252 		goto nla_put_vfinfo_failure;
1253 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1254 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1255 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1256 		    &vf_rate) ||
1257 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1258 		    &vf_tx_rate) ||
1259 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1260 		    &vf_spoofchk) ||
1261 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1262 		    &vf_linkstate) ||
1263 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1264 		    sizeof(vf_rss_query_en),
1265 		    &vf_rss_query_en) ||
1266 	    nla_put(skb, IFLA_VF_TRUST,
1267 		    sizeof(vf_trust), &vf_trust))
1268 		goto nla_put_vf_failure;
1269 	vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1270 	if (!vfvlanlist)
1271 		goto nla_put_vf_failure;
1272 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1273 		    &vf_vlan_info)) {
1274 		nla_nest_cancel(skb, vfvlanlist);
1275 		goto nla_put_vf_failure;
1276 	}
1277 	nla_nest_end(skb, vfvlanlist);
1278 	memset(&vf_stats, 0, sizeof(vf_stats));
1279 	if (dev->netdev_ops->ndo_get_vf_stats)
1280 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1281 						&vf_stats);
1282 	vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1283 	if (!vfstats)
1284 		goto nla_put_vf_failure;
1285 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1286 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1287 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1288 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1289 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1290 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1291 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1292 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1293 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1294 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1295 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1296 			      vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1297 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1298 			      vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1299 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1300 			      vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1301 		nla_nest_cancel(skb, vfstats);
1302 		goto nla_put_vf_failure;
1303 	}
1304 	nla_nest_end(skb, vfstats);
1305 	nla_nest_end(skb, vf);
1306 	return 0;
1307 
1308 nla_put_vf_failure:
1309 	nla_nest_cancel(skb, vf);
1310 nla_put_vfinfo_failure:
1311 	nla_nest_cancel(skb, vfinfo);
1312 	return -EMSGSIZE;
1313 }
1314 
1315 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1316 					   struct net_device *dev,
1317 					   u32 ext_filter_mask)
1318 {
1319 	struct nlattr *vfinfo;
1320 	int i, num_vfs;
1321 
1322 	if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1323 		return 0;
1324 
1325 	num_vfs = dev_num_vf(dev->dev.parent);
1326 	if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1327 		return -EMSGSIZE;
1328 
1329 	if (!dev->netdev_ops->ndo_get_vf_config)
1330 		return 0;
1331 
1332 	vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1333 	if (!vfinfo)
1334 		return -EMSGSIZE;
1335 
1336 	for (i = 0; i < num_vfs; i++) {
1337 		if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1338 			return -EMSGSIZE;
1339 	}
1340 
1341 	nla_nest_end(skb, vfinfo);
1342 	return 0;
1343 }
1344 
1345 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1346 {
1347 	struct rtnl_link_ifmap map;
1348 
1349 	memset(&map, 0, sizeof(map));
1350 	map.mem_start   = dev->mem_start;
1351 	map.mem_end     = dev->mem_end;
1352 	map.base_addr   = dev->base_addr;
1353 	map.irq         = dev->irq;
1354 	map.dma         = dev->dma;
1355 	map.port        = dev->if_port;
1356 
1357 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1358 		return -EMSGSIZE;
1359 
1360 	return 0;
1361 }
1362 
1363 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1364 {
1365 	const struct bpf_prog *generic_xdp_prog;
1366 
1367 	ASSERT_RTNL();
1368 
1369 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1370 	if (!generic_xdp_prog)
1371 		return 0;
1372 	return generic_xdp_prog->aux->id;
1373 }
1374 
1375 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1376 {
1377 	return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG);
1378 }
1379 
1380 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1381 {
1382 	return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf,
1383 			       XDP_QUERY_PROG_HW);
1384 }
1385 
1386 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1387 			       u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1388 			       u32 (*get_prog_id)(struct net_device *dev))
1389 {
1390 	u32 curr_id;
1391 	int err;
1392 
1393 	curr_id = get_prog_id(dev);
1394 	if (!curr_id)
1395 		return 0;
1396 
1397 	*prog_id = curr_id;
1398 	err = nla_put_u32(skb, attr, curr_id);
1399 	if (err)
1400 		return err;
1401 
1402 	if (*mode != XDP_ATTACHED_NONE)
1403 		*mode = XDP_ATTACHED_MULTI;
1404 	else
1405 		*mode = tgt_mode;
1406 
1407 	return 0;
1408 }
1409 
1410 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1411 {
1412 	struct nlattr *xdp;
1413 	u32 prog_id;
1414 	int err;
1415 	u8 mode;
1416 
1417 	xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1418 	if (!xdp)
1419 		return -EMSGSIZE;
1420 
1421 	prog_id = 0;
1422 	mode = XDP_ATTACHED_NONE;
1423 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1424 				  IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1425 	if (err)
1426 		goto err_cancel;
1427 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1428 				  IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1429 	if (err)
1430 		goto err_cancel;
1431 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1432 				  IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1433 	if (err)
1434 		goto err_cancel;
1435 
1436 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1437 	if (err)
1438 		goto err_cancel;
1439 
1440 	if (prog_id && mode != XDP_ATTACHED_MULTI) {
1441 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1442 		if (err)
1443 			goto err_cancel;
1444 	}
1445 
1446 	nla_nest_end(skb, xdp);
1447 	return 0;
1448 
1449 err_cancel:
1450 	nla_nest_cancel(skb, xdp);
1451 	return err;
1452 }
1453 
1454 static u32 rtnl_get_event(unsigned long event)
1455 {
1456 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1457 
1458 	switch (event) {
1459 	case NETDEV_REBOOT:
1460 		rtnl_event_type = IFLA_EVENT_REBOOT;
1461 		break;
1462 	case NETDEV_FEAT_CHANGE:
1463 		rtnl_event_type = IFLA_EVENT_FEATURES;
1464 		break;
1465 	case NETDEV_BONDING_FAILOVER:
1466 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1467 		break;
1468 	case NETDEV_NOTIFY_PEERS:
1469 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1470 		break;
1471 	case NETDEV_RESEND_IGMP:
1472 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1473 		break;
1474 	case NETDEV_CHANGEINFODATA:
1475 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1476 		break;
1477 	default:
1478 		break;
1479 	}
1480 
1481 	return rtnl_event_type;
1482 }
1483 
1484 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1485 {
1486 	const struct net_device *upper_dev;
1487 	int ret = 0;
1488 
1489 	rcu_read_lock();
1490 
1491 	upper_dev = netdev_master_upper_dev_get_rcu(dev);
1492 	if (upper_dev)
1493 		ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1494 
1495 	rcu_read_unlock();
1496 	return ret;
1497 }
1498 
1499 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1500 			  bool force)
1501 {
1502 	int ifindex = dev_get_iflink(dev);
1503 
1504 	if (force || dev->ifindex != ifindex)
1505 		return nla_put_u32(skb, IFLA_LINK, ifindex);
1506 
1507 	return 0;
1508 }
1509 
1510 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1511 					      struct net_device *dev)
1512 {
1513 	char buf[IFALIASZ];
1514 	int ret;
1515 
1516 	ret = dev_get_alias(dev, buf, sizeof(buf));
1517 	return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1518 }
1519 
1520 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1521 				  const struct net_device *dev,
1522 				  struct net *src_net)
1523 {
1524 	bool put_iflink = false;
1525 
1526 	if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1527 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1528 
1529 		if (!net_eq(dev_net(dev), link_net)) {
1530 			int id = peernet2id_alloc(src_net, link_net);
1531 
1532 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1533 				return -EMSGSIZE;
1534 
1535 			put_iflink = true;
1536 		}
1537 	}
1538 
1539 	return nla_put_iflink(skb, dev, put_iflink);
1540 }
1541 
1542 static int rtnl_fill_link_af(struct sk_buff *skb,
1543 			     const struct net_device *dev,
1544 			     u32 ext_filter_mask)
1545 {
1546 	const struct rtnl_af_ops *af_ops;
1547 	struct nlattr *af_spec;
1548 
1549 	af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1550 	if (!af_spec)
1551 		return -EMSGSIZE;
1552 
1553 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1554 		struct nlattr *af;
1555 		int err;
1556 
1557 		if (!af_ops->fill_link_af)
1558 			continue;
1559 
1560 		af = nla_nest_start_noflag(skb, af_ops->family);
1561 		if (!af)
1562 			return -EMSGSIZE;
1563 
1564 		err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1565 		/*
1566 		 * Caller may return ENODATA to indicate that there
1567 		 * was no data to be dumped. This is not an error, it
1568 		 * means we should trim the attribute header and
1569 		 * continue.
1570 		 */
1571 		if (err == -ENODATA)
1572 			nla_nest_cancel(skb, af);
1573 		else if (err < 0)
1574 			return -EMSGSIZE;
1575 
1576 		nla_nest_end(skb, af);
1577 	}
1578 
1579 	nla_nest_end(skb, af_spec);
1580 	return 0;
1581 }
1582 
1583 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1584 			    struct net_device *dev, struct net *src_net,
1585 			    int type, u32 pid, u32 seq, u32 change,
1586 			    unsigned int flags, u32 ext_filter_mask,
1587 			    u32 event, int *new_nsid, int new_ifindex,
1588 			    int tgt_netnsid)
1589 {
1590 	struct ifinfomsg *ifm;
1591 	struct nlmsghdr *nlh;
1592 
1593 	ASSERT_RTNL();
1594 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1595 	if (nlh == NULL)
1596 		return -EMSGSIZE;
1597 
1598 	ifm = nlmsg_data(nlh);
1599 	ifm->ifi_family = AF_UNSPEC;
1600 	ifm->__ifi_pad = 0;
1601 	ifm->ifi_type = dev->type;
1602 	ifm->ifi_index = dev->ifindex;
1603 	ifm->ifi_flags = dev_get_flags(dev);
1604 	ifm->ifi_change = change;
1605 
1606 	if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1607 		goto nla_put_failure;
1608 
1609 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1610 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1611 	    nla_put_u8(skb, IFLA_OPERSTATE,
1612 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1613 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1614 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1615 	    nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1616 	    nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1617 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1618 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1619 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1620 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1621 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1622 #ifdef CONFIG_RPS
1623 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1624 #endif
1625 	    put_master_ifindex(skb, dev) ||
1626 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1627 	    (dev->qdisc &&
1628 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1629 	    nla_put_ifalias(skb, dev) ||
1630 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1631 			atomic_read(&dev->carrier_up_count) +
1632 			atomic_read(&dev->carrier_down_count)) ||
1633 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1634 	    nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1635 			atomic_read(&dev->carrier_up_count)) ||
1636 	    nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1637 			atomic_read(&dev->carrier_down_count)))
1638 		goto nla_put_failure;
1639 
1640 	if (event != IFLA_EVENT_NONE) {
1641 		if (nla_put_u32(skb, IFLA_EVENT, event))
1642 			goto nla_put_failure;
1643 	}
1644 
1645 	if (rtnl_fill_link_ifmap(skb, dev))
1646 		goto nla_put_failure;
1647 
1648 	if (dev->addr_len) {
1649 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1650 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1651 			goto nla_put_failure;
1652 	}
1653 
1654 	if (rtnl_phys_port_id_fill(skb, dev))
1655 		goto nla_put_failure;
1656 
1657 	if (rtnl_phys_port_name_fill(skb, dev))
1658 		goto nla_put_failure;
1659 
1660 	if (rtnl_phys_switch_id_fill(skb, dev))
1661 		goto nla_put_failure;
1662 
1663 	if (rtnl_fill_stats(skb, dev))
1664 		goto nla_put_failure;
1665 
1666 	if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1667 		goto nla_put_failure;
1668 
1669 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1670 		goto nla_put_failure;
1671 
1672 	if (rtnl_xdp_fill(skb, dev))
1673 		goto nla_put_failure;
1674 
1675 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1676 		if (rtnl_link_fill(skb, dev) < 0)
1677 			goto nla_put_failure;
1678 	}
1679 
1680 	if (rtnl_fill_link_netnsid(skb, dev, src_net))
1681 		goto nla_put_failure;
1682 
1683 	if (new_nsid &&
1684 	    nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1685 		goto nla_put_failure;
1686 	if (new_ifindex &&
1687 	    nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1688 		goto nla_put_failure;
1689 
1690 
1691 	rcu_read_lock();
1692 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1693 		goto nla_put_failure_rcu;
1694 	rcu_read_unlock();
1695 
1696 	nlmsg_end(skb, nlh);
1697 	return 0;
1698 
1699 nla_put_failure_rcu:
1700 	rcu_read_unlock();
1701 nla_put_failure:
1702 	nlmsg_cancel(skb, nlh);
1703 	return -EMSGSIZE;
1704 }
1705 
1706 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1707 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1708 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1709 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1710 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1711 	[IFLA_MTU]		= { .type = NLA_U32 },
1712 	[IFLA_LINK]		= { .type = NLA_U32 },
1713 	[IFLA_MASTER]		= { .type = NLA_U32 },
1714 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1715 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1716 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1717 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1718 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1719 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1720 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1721 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1722 	/* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1723 	 * allow 0-length string (needed to remove an alias).
1724 	 */
1725 	[IFLA_IFALIAS]	        = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1726 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1727 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1728 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1729 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1730 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1731 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1732 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1733 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1734 	[IFLA_GSO_MAX_SEGS]	= { .type = NLA_U32 },
1735 	[IFLA_GSO_MAX_SIZE]	= { .type = NLA_U32 },
1736 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1737 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1738 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1739 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1740 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1741 	[IFLA_XDP]		= { .type = NLA_NESTED },
1742 	[IFLA_EVENT]		= { .type = NLA_U32 },
1743 	[IFLA_GROUP]		= { .type = NLA_U32 },
1744 	[IFLA_TARGET_NETNSID]	= { .type = NLA_S32 },
1745 	[IFLA_CARRIER_UP_COUNT]	= { .type = NLA_U32 },
1746 	[IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1747 	[IFLA_MIN_MTU]		= { .type = NLA_U32 },
1748 	[IFLA_MAX_MTU]		= { .type = NLA_U32 },
1749 };
1750 
1751 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1752 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1753 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1754 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1755 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1756 };
1757 
1758 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1759 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1760 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1761 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1762 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1763 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1764 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1765 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1766 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1767 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1768 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1769 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1770 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1771 };
1772 
1773 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1774 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1775 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1776 				    .len = PORT_PROFILE_MAX },
1777 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1778 				      .len = PORT_UUID_MAX },
1779 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1780 				    .len = PORT_UUID_MAX },
1781 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1782 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1783 
1784 	/* Unused, but we need to keep it here since user space could
1785 	 * fill it. It's also broken with regard to NLA_BINARY use in
1786 	 * combination with structs.
1787 	 */
1788 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1789 				    .len = sizeof(struct ifla_port_vsi) },
1790 };
1791 
1792 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1793 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1794 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1795 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1796 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1797 };
1798 
1799 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1800 {
1801 	const struct rtnl_link_ops *ops = NULL;
1802 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1803 
1804 	if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1805 		return NULL;
1806 
1807 	if (linfo[IFLA_INFO_KIND]) {
1808 		char kind[MODULE_NAME_LEN];
1809 
1810 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1811 		ops = rtnl_link_ops_get(kind);
1812 	}
1813 
1814 	return ops;
1815 }
1816 
1817 static bool link_master_filtered(struct net_device *dev, int master_idx)
1818 {
1819 	struct net_device *master;
1820 
1821 	if (!master_idx)
1822 		return false;
1823 
1824 	master = netdev_master_upper_dev_get(dev);
1825 	if (!master || master->ifindex != master_idx)
1826 		return true;
1827 
1828 	return false;
1829 }
1830 
1831 static bool link_kind_filtered(const struct net_device *dev,
1832 			       const struct rtnl_link_ops *kind_ops)
1833 {
1834 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1835 		return true;
1836 
1837 	return false;
1838 }
1839 
1840 static bool link_dump_filtered(struct net_device *dev,
1841 			       int master_idx,
1842 			       const struct rtnl_link_ops *kind_ops)
1843 {
1844 	if (link_master_filtered(dev, master_idx) ||
1845 	    link_kind_filtered(dev, kind_ops))
1846 		return true;
1847 
1848 	return false;
1849 }
1850 
1851 /**
1852  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1853  * @sk: netlink socket
1854  * @netnsid: network namespace identifier
1855  *
1856  * Returns the network namespace identified by netnsid on success or an error
1857  * pointer on failure.
1858  */
1859 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1860 {
1861 	struct net *net;
1862 
1863 	net = get_net_ns_by_id(sock_net(sk), netnsid);
1864 	if (!net)
1865 		return ERR_PTR(-EINVAL);
1866 
1867 	/* For now, the caller is required to have CAP_NET_ADMIN in
1868 	 * the user namespace owning the target net ns.
1869 	 */
1870 	if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1871 		put_net(net);
1872 		return ERR_PTR(-EACCES);
1873 	}
1874 	return net;
1875 }
1876 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
1877 
1878 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
1879 				      bool strict_check, struct nlattr **tb,
1880 				      struct netlink_ext_ack *extack)
1881 {
1882 	int hdrlen;
1883 
1884 	if (strict_check) {
1885 		struct ifinfomsg *ifm;
1886 
1887 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1888 			NL_SET_ERR_MSG(extack, "Invalid header for link dump");
1889 			return -EINVAL;
1890 		}
1891 
1892 		ifm = nlmsg_data(nlh);
1893 		if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
1894 		    ifm->ifi_change) {
1895 			NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
1896 			return -EINVAL;
1897 		}
1898 		if (ifm->ifi_index) {
1899 			NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
1900 			return -EINVAL;
1901 		}
1902 
1903 		return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
1904 						     IFLA_MAX, ifla_policy,
1905 						     extack);
1906 	}
1907 
1908 	/* A hack to preserve kernel<->userspace interface.
1909 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1910 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1911 	 * what iproute2 < v3.9.0 used.
1912 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1913 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1914 	 */
1915 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1916 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1917 
1918 	return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
1919 				      extack);
1920 }
1921 
1922 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1923 {
1924 	struct netlink_ext_ack *extack = cb->extack;
1925 	const struct nlmsghdr *nlh = cb->nlh;
1926 	struct net *net = sock_net(skb->sk);
1927 	struct net *tgt_net = net;
1928 	int h, s_h;
1929 	int idx = 0, s_idx;
1930 	struct net_device *dev;
1931 	struct hlist_head *head;
1932 	struct nlattr *tb[IFLA_MAX+1];
1933 	u32 ext_filter_mask = 0;
1934 	const struct rtnl_link_ops *kind_ops = NULL;
1935 	unsigned int flags = NLM_F_MULTI;
1936 	int master_idx = 0;
1937 	int netnsid = -1;
1938 	int err, i;
1939 
1940 	s_h = cb->args[0];
1941 	s_idx = cb->args[1];
1942 
1943 	err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
1944 	if (err < 0) {
1945 		if (cb->strict_check)
1946 			return err;
1947 
1948 		goto walk_entries;
1949 	}
1950 
1951 	for (i = 0; i <= IFLA_MAX; ++i) {
1952 		if (!tb[i])
1953 			continue;
1954 
1955 		/* new attributes should only be added with strict checking */
1956 		switch (i) {
1957 		case IFLA_TARGET_NETNSID:
1958 			netnsid = nla_get_s32(tb[i]);
1959 			tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
1960 			if (IS_ERR(tgt_net)) {
1961 				NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
1962 				return PTR_ERR(tgt_net);
1963 			}
1964 			break;
1965 		case IFLA_EXT_MASK:
1966 			ext_filter_mask = nla_get_u32(tb[i]);
1967 			break;
1968 		case IFLA_MASTER:
1969 			master_idx = nla_get_u32(tb[i]);
1970 			break;
1971 		case IFLA_LINKINFO:
1972 			kind_ops = linkinfo_to_kind_ops(tb[i]);
1973 			break;
1974 		default:
1975 			if (cb->strict_check) {
1976 				NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
1977 				return -EINVAL;
1978 			}
1979 		}
1980 	}
1981 
1982 	if (master_idx || kind_ops)
1983 		flags |= NLM_F_DUMP_FILTERED;
1984 
1985 walk_entries:
1986 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1987 		idx = 0;
1988 		head = &tgt_net->dev_index_head[h];
1989 		hlist_for_each_entry(dev, head, index_hlist) {
1990 			if (link_dump_filtered(dev, master_idx, kind_ops))
1991 				goto cont;
1992 			if (idx < s_idx)
1993 				goto cont;
1994 			err = rtnl_fill_ifinfo(skb, dev, net,
1995 					       RTM_NEWLINK,
1996 					       NETLINK_CB(cb->skb).portid,
1997 					       nlh->nlmsg_seq, 0, flags,
1998 					       ext_filter_mask, 0, NULL, 0,
1999 					       netnsid);
2000 
2001 			if (err < 0) {
2002 				if (likely(skb->len))
2003 					goto out;
2004 
2005 				goto out_err;
2006 			}
2007 cont:
2008 			idx++;
2009 		}
2010 	}
2011 out:
2012 	err = skb->len;
2013 out_err:
2014 	cb->args[1] = idx;
2015 	cb->args[0] = h;
2016 	cb->seq = net->dev_base_seq;
2017 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2018 	if (netnsid >= 0)
2019 		put_net(tgt_net);
2020 
2021 	return err;
2022 }
2023 
2024 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2025 			struct netlink_ext_ack *exterr)
2026 {
2027 	return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2028 				    exterr);
2029 }
2030 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2031 
2032 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2033 {
2034 	struct net *net;
2035 	/* Examine the link attributes and figure out which
2036 	 * network namespace we are talking about.
2037 	 */
2038 	if (tb[IFLA_NET_NS_PID])
2039 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2040 	else if (tb[IFLA_NET_NS_FD])
2041 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2042 	else
2043 		net = get_net(src_net);
2044 	return net;
2045 }
2046 EXPORT_SYMBOL(rtnl_link_get_net);
2047 
2048 /* Figure out which network namespace we are talking about by
2049  * examining the link attributes in the following order:
2050  *
2051  * 1. IFLA_NET_NS_PID
2052  * 2. IFLA_NET_NS_FD
2053  * 3. IFLA_TARGET_NETNSID
2054  */
2055 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2056 					       struct nlattr *tb[])
2057 {
2058 	struct net *net;
2059 
2060 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2061 		return rtnl_link_get_net(src_net, tb);
2062 
2063 	if (!tb[IFLA_TARGET_NETNSID])
2064 		return get_net(src_net);
2065 
2066 	net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2067 	if (!net)
2068 		return ERR_PTR(-EINVAL);
2069 
2070 	return net;
2071 }
2072 
2073 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2074 					     struct net *src_net,
2075 					     struct nlattr *tb[], int cap)
2076 {
2077 	struct net *net;
2078 
2079 	net = rtnl_link_get_net_by_nlattr(src_net, tb);
2080 	if (IS_ERR(net))
2081 		return net;
2082 
2083 	if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2084 		put_net(net);
2085 		return ERR_PTR(-EPERM);
2086 	}
2087 
2088 	return net;
2089 }
2090 
2091 /* Verify that rtnetlink requests do not pass additional properties
2092  * potentially referring to different network namespaces.
2093  */
2094 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2095 				    struct netlink_ext_ack *extack,
2096 				    bool netns_id_only)
2097 {
2098 
2099 	if (netns_id_only) {
2100 		if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2101 			return 0;
2102 
2103 		NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2104 		return -EOPNOTSUPP;
2105 	}
2106 
2107 	if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2108 		goto invalid_attr;
2109 
2110 	if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2111 		goto invalid_attr;
2112 
2113 	if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2114 		goto invalid_attr;
2115 
2116 	return 0;
2117 
2118 invalid_attr:
2119 	NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2120 	return -EINVAL;
2121 }
2122 
2123 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2124 {
2125 	if (dev) {
2126 		if (tb[IFLA_ADDRESS] &&
2127 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2128 			return -EINVAL;
2129 
2130 		if (tb[IFLA_BROADCAST] &&
2131 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2132 			return -EINVAL;
2133 	}
2134 
2135 	if (tb[IFLA_AF_SPEC]) {
2136 		struct nlattr *af;
2137 		int rem, err;
2138 
2139 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2140 			const struct rtnl_af_ops *af_ops;
2141 
2142 			rcu_read_lock();
2143 			af_ops = rtnl_af_lookup(nla_type(af));
2144 			if (!af_ops) {
2145 				rcu_read_unlock();
2146 				return -EAFNOSUPPORT;
2147 			}
2148 
2149 			if (!af_ops->set_link_af) {
2150 				rcu_read_unlock();
2151 				return -EOPNOTSUPP;
2152 			}
2153 
2154 			if (af_ops->validate_link_af) {
2155 				err = af_ops->validate_link_af(dev, af);
2156 				if (err < 0) {
2157 					rcu_read_unlock();
2158 					return err;
2159 				}
2160 			}
2161 
2162 			rcu_read_unlock();
2163 		}
2164 	}
2165 
2166 	return 0;
2167 }
2168 
2169 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2170 				  int guid_type)
2171 {
2172 	const struct net_device_ops *ops = dev->netdev_ops;
2173 
2174 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2175 }
2176 
2177 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2178 {
2179 	if (dev->type != ARPHRD_INFINIBAND)
2180 		return -EOPNOTSUPP;
2181 
2182 	return handle_infiniband_guid(dev, ivt, guid_type);
2183 }
2184 
2185 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2186 {
2187 	const struct net_device_ops *ops = dev->netdev_ops;
2188 	int err = -EINVAL;
2189 
2190 	if (tb[IFLA_VF_MAC]) {
2191 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2192 
2193 		err = -EOPNOTSUPP;
2194 		if (ops->ndo_set_vf_mac)
2195 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
2196 						  ivm->mac);
2197 		if (err < 0)
2198 			return err;
2199 	}
2200 
2201 	if (tb[IFLA_VF_VLAN]) {
2202 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2203 
2204 		err = -EOPNOTSUPP;
2205 		if (ops->ndo_set_vf_vlan)
2206 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2207 						   ivv->qos,
2208 						   htons(ETH_P_8021Q));
2209 		if (err < 0)
2210 			return err;
2211 	}
2212 
2213 	if (tb[IFLA_VF_VLAN_LIST]) {
2214 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2215 		struct nlattr *attr;
2216 		int rem, len = 0;
2217 
2218 		err = -EOPNOTSUPP;
2219 		if (!ops->ndo_set_vf_vlan)
2220 			return err;
2221 
2222 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2223 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2224 			    nla_len(attr) < NLA_HDRLEN) {
2225 				return -EINVAL;
2226 			}
2227 			if (len >= MAX_VLAN_LIST_LEN)
2228 				return -EOPNOTSUPP;
2229 			ivvl[len] = nla_data(attr);
2230 
2231 			len++;
2232 		}
2233 		if (len == 0)
2234 			return -EINVAL;
2235 
2236 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2237 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
2238 		if (err < 0)
2239 			return err;
2240 	}
2241 
2242 	if (tb[IFLA_VF_TX_RATE]) {
2243 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2244 		struct ifla_vf_info ivf;
2245 
2246 		err = -EOPNOTSUPP;
2247 		if (ops->ndo_get_vf_config)
2248 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2249 		if (err < 0)
2250 			return err;
2251 
2252 		err = -EOPNOTSUPP;
2253 		if (ops->ndo_set_vf_rate)
2254 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2255 						   ivf.min_tx_rate,
2256 						   ivt->rate);
2257 		if (err < 0)
2258 			return err;
2259 	}
2260 
2261 	if (tb[IFLA_VF_RATE]) {
2262 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2263 
2264 		err = -EOPNOTSUPP;
2265 		if (ops->ndo_set_vf_rate)
2266 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2267 						   ivt->min_tx_rate,
2268 						   ivt->max_tx_rate);
2269 		if (err < 0)
2270 			return err;
2271 	}
2272 
2273 	if (tb[IFLA_VF_SPOOFCHK]) {
2274 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2275 
2276 		err = -EOPNOTSUPP;
2277 		if (ops->ndo_set_vf_spoofchk)
2278 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2279 						       ivs->setting);
2280 		if (err < 0)
2281 			return err;
2282 	}
2283 
2284 	if (tb[IFLA_VF_LINK_STATE]) {
2285 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2286 
2287 		err = -EOPNOTSUPP;
2288 		if (ops->ndo_set_vf_link_state)
2289 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2290 							 ivl->link_state);
2291 		if (err < 0)
2292 			return err;
2293 	}
2294 
2295 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
2296 		struct ifla_vf_rss_query_en *ivrssq_en;
2297 
2298 		err = -EOPNOTSUPP;
2299 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2300 		if (ops->ndo_set_vf_rss_query_en)
2301 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2302 							   ivrssq_en->setting);
2303 		if (err < 0)
2304 			return err;
2305 	}
2306 
2307 	if (tb[IFLA_VF_TRUST]) {
2308 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2309 
2310 		err = -EOPNOTSUPP;
2311 		if (ops->ndo_set_vf_trust)
2312 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2313 		if (err < 0)
2314 			return err;
2315 	}
2316 
2317 	if (tb[IFLA_VF_IB_NODE_GUID]) {
2318 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2319 
2320 		if (!ops->ndo_set_vf_guid)
2321 			return -EOPNOTSUPP;
2322 
2323 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2324 	}
2325 
2326 	if (tb[IFLA_VF_IB_PORT_GUID]) {
2327 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2328 
2329 		if (!ops->ndo_set_vf_guid)
2330 			return -EOPNOTSUPP;
2331 
2332 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2333 	}
2334 
2335 	return err;
2336 }
2337 
2338 static int do_set_master(struct net_device *dev, int ifindex,
2339 			 struct netlink_ext_ack *extack)
2340 {
2341 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2342 	const struct net_device_ops *ops;
2343 	int err;
2344 
2345 	if (upper_dev) {
2346 		if (upper_dev->ifindex == ifindex)
2347 			return 0;
2348 		ops = upper_dev->netdev_ops;
2349 		if (ops->ndo_del_slave) {
2350 			err = ops->ndo_del_slave(upper_dev, dev);
2351 			if (err)
2352 				return err;
2353 		} else {
2354 			return -EOPNOTSUPP;
2355 		}
2356 	}
2357 
2358 	if (ifindex) {
2359 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2360 		if (!upper_dev)
2361 			return -EINVAL;
2362 		ops = upper_dev->netdev_ops;
2363 		if (ops->ndo_add_slave) {
2364 			err = ops->ndo_add_slave(upper_dev, dev, extack);
2365 			if (err)
2366 				return err;
2367 		} else {
2368 			return -EOPNOTSUPP;
2369 		}
2370 	}
2371 	return 0;
2372 }
2373 
2374 #define DO_SETLINK_MODIFIED	0x01
2375 /* notify flag means notify + modified. */
2376 #define DO_SETLINK_NOTIFY	0x03
2377 static int do_setlink(const struct sk_buff *skb,
2378 		      struct net_device *dev, struct ifinfomsg *ifm,
2379 		      struct netlink_ext_ack *extack,
2380 		      struct nlattr **tb, char *ifname, int status)
2381 {
2382 	const struct net_device_ops *ops = dev->netdev_ops;
2383 	int err;
2384 
2385 	err = validate_linkmsg(dev, tb);
2386 	if (err < 0)
2387 		return err;
2388 
2389 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2390 		struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2391 							    tb, CAP_NET_ADMIN);
2392 		if (IS_ERR(net)) {
2393 			err = PTR_ERR(net);
2394 			goto errout;
2395 		}
2396 
2397 		err = dev_change_net_namespace(dev, net, ifname);
2398 		put_net(net);
2399 		if (err)
2400 			goto errout;
2401 		status |= DO_SETLINK_MODIFIED;
2402 	}
2403 
2404 	if (tb[IFLA_MAP]) {
2405 		struct rtnl_link_ifmap *u_map;
2406 		struct ifmap k_map;
2407 
2408 		if (!ops->ndo_set_config) {
2409 			err = -EOPNOTSUPP;
2410 			goto errout;
2411 		}
2412 
2413 		if (!netif_device_present(dev)) {
2414 			err = -ENODEV;
2415 			goto errout;
2416 		}
2417 
2418 		u_map = nla_data(tb[IFLA_MAP]);
2419 		k_map.mem_start = (unsigned long) u_map->mem_start;
2420 		k_map.mem_end = (unsigned long) u_map->mem_end;
2421 		k_map.base_addr = (unsigned short) u_map->base_addr;
2422 		k_map.irq = (unsigned char) u_map->irq;
2423 		k_map.dma = (unsigned char) u_map->dma;
2424 		k_map.port = (unsigned char) u_map->port;
2425 
2426 		err = ops->ndo_set_config(dev, &k_map);
2427 		if (err < 0)
2428 			goto errout;
2429 
2430 		status |= DO_SETLINK_NOTIFY;
2431 	}
2432 
2433 	if (tb[IFLA_ADDRESS]) {
2434 		struct sockaddr *sa;
2435 		int len;
2436 
2437 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2438 						  sizeof(*sa));
2439 		sa = kmalloc(len, GFP_KERNEL);
2440 		if (!sa) {
2441 			err = -ENOMEM;
2442 			goto errout;
2443 		}
2444 		sa->sa_family = dev->type;
2445 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2446 		       dev->addr_len);
2447 		err = dev_set_mac_address(dev, sa, extack);
2448 		kfree(sa);
2449 		if (err)
2450 			goto errout;
2451 		status |= DO_SETLINK_MODIFIED;
2452 	}
2453 
2454 	if (tb[IFLA_MTU]) {
2455 		err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2456 		if (err < 0)
2457 			goto errout;
2458 		status |= DO_SETLINK_MODIFIED;
2459 	}
2460 
2461 	if (tb[IFLA_GROUP]) {
2462 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2463 		status |= DO_SETLINK_NOTIFY;
2464 	}
2465 
2466 	/*
2467 	 * Interface selected by interface index but interface
2468 	 * name provided implies that a name change has been
2469 	 * requested.
2470 	 */
2471 	if (ifm->ifi_index > 0 && ifname[0]) {
2472 		err = dev_change_name(dev, ifname);
2473 		if (err < 0)
2474 			goto errout;
2475 		status |= DO_SETLINK_MODIFIED;
2476 	}
2477 
2478 	if (tb[IFLA_IFALIAS]) {
2479 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2480 				    nla_len(tb[IFLA_IFALIAS]));
2481 		if (err < 0)
2482 			goto errout;
2483 		status |= DO_SETLINK_NOTIFY;
2484 	}
2485 
2486 	if (tb[IFLA_BROADCAST]) {
2487 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2488 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2489 	}
2490 
2491 	if (ifm->ifi_flags || ifm->ifi_change) {
2492 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2493 				       extack);
2494 		if (err < 0)
2495 			goto errout;
2496 	}
2497 
2498 	if (tb[IFLA_MASTER]) {
2499 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2500 		if (err)
2501 			goto errout;
2502 		status |= DO_SETLINK_MODIFIED;
2503 	}
2504 
2505 	if (tb[IFLA_CARRIER]) {
2506 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2507 		if (err)
2508 			goto errout;
2509 		status |= DO_SETLINK_MODIFIED;
2510 	}
2511 
2512 	if (tb[IFLA_TXQLEN]) {
2513 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2514 
2515 		err = dev_change_tx_queue_len(dev, value);
2516 		if (err)
2517 			goto errout;
2518 		status |= DO_SETLINK_MODIFIED;
2519 	}
2520 
2521 	if (tb[IFLA_GSO_MAX_SIZE]) {
2522 		u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2523 
2524 		if (max_size > GSO_MAX_SIZE) {
2525 			err = -EINVAL;
2526 			goto errout;
2527 		}
2528 
2529 		if (dev->gso_max_size ^ max_size) {
2530 			netif_set_gso_max_size(dev, max_size);
2531 			status |= DO_SETLINK_MODIFIED;
2532 		}
2533 	}
2534 
2535 	if (tb[IFLA_GSO_MAX_SEGS]) {
2536 		u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2537 
2538 		if (max_segs > GSO_MAX_SEGS) {
2539 			err = -EINVAL;
2540 			goto errout;
2541 		}
2542 
2543 		if (dev->gso_max_segs ^ max_segs) {
2544 			dev->gso_max_segs = max_segs;
2545 			status |= DO_SETLINK_MODIFIED;
2546 		}
2547 	}
2548 
2549 	if (tb[IFLA_OPERSTATE])
2550 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2551 
2552 	if (tb[IFLA_LINKMODE]) {
2553 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2554 
2555 		write_lock_bh(&dev_base_lock);
2556 		if (dev->link_mode ^ value)
2557 			status |= DO_SETLINK_NOTIFY;
2558 		dev->link_mode = value;
2559 		write_unlock_bh(&dev_base_lock);
2560 	}
2561 
2562 	if (tb[IFLA_VFINFO_LIST]) {
2563 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2564 		struct nlattr *attr;
2565 		int rem;
2566 
2567 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2568 			if (nla_type(attr) != IFLA_VF_INFO ||
2569 			    nla_len(attr) < NLA_HDRLEN) {
2570 				err = -EINVAL;
2571 				goto errout;
2572 			}
2573 			err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2574 							  attr,
2575 							  ifla_vf_policy,
2576 							  NULL);
2577 			if (err < 0)
2578 				goto errout;
2579 			err = do_setvfinfo(dev, vfinfo);
2580 			if (err < 0)
2581 				goto errout;
2582 			status |= DO_SETLINK_NOTIFY;
2583 		}
2584 	}
2585 	err = 0;
2586 
2587 	if (tb[IFLA_VF_PORTS]) {
2588 		struct nlattr *port[IFLA_PORT_MAX+1];
2589 		struct nlattr *attr;
2590 		int vf;
2591 		int rem;
2592 
2593 		err = -EOPNOTSUPP;
2594 		if (!ops->ndo_set_vf_port)
2595 			goto errout;
2596 
2597 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2598 			if (nla_type(attr) != IFLA_VF_PORT ||
2599 			    nla_len(attr) < NLA_HDRLEN) {
2600 				err = -EINVAL;
2601 				goto errout;
2602 			}
2603 			err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2604 							  attr,
2605 							  ifla_port_policy,
2606 							  NULL);
2607 			if (err < 0)
2608 				goto errout;
2609 			if (!port[IFLA_PORT_VF]) {
2610 				err = -EOPNOTSUPP;
2611 				goto errout;
2612 			}
2613 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2614 			err = ops->ndo_set_vf_port(dev, vf, port);
2615 			if (err < 0)
2616 				goto errout;
2617 			status |= DO_SETLINK_NOTIFY;
2618 		}
2619 	}
2620 	err = 0;
2621 
2622 	if (tb[IFLA_PORT_SELF]) {
2623 		struct nlattr *port[IFLA_PORT_MAX+1];
2624 
2625 		err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2626 						  tb[IFLA_PORT_SELF],
2627 						  ifla_port_policy, NULL);
2628 		if (err < 0)
2629 			goto errout;
2630 
2631 		err = -EOPNOTSUPP;
2632 		if (ops->ndo_set_vf_port)
2633 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2634 		if (err < 0)
2635 			goto errout;
2636 		status |= DO_SETLINK_NOTIFY;
2637 	}
2638 
2639 	if (tb[IFLA_AF_SPEC]) {
2640 		struct nlattr *af;
2641 		int rem;
2642 
2643 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2644 			const struct rtnl_af_ops *af_ops;
2645 
2646 			rcu_read_lock();
2647 
2648 			BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2649 
2650 			err = af_ops->set_link_af(dev, af);
2651 			if (err < 0) {
2652 				rcu_read_unlock();
2653 				goto errout;
2654 			}
2655 
2656 			rcu_read_unlock();
2657 			status |= DO_SETLINK_NOTIFY;
2658 		}
2659 	}
2660 	err = 0;
2661 
2662 	if (tb[IFLA_PROTO_DOWN]) {
2663 		err = dev_change_proto_down(dev,
2664 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2665 		if (err)
2666 			goto errout;
2667 		status |= DO_SETLINK_NOTIFY;
2668 	}
2669 
2670 	if (tb[IFLA_XDP]) {
2671 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2672 		u32 xdp_flags = 0;
2673 
2674 		err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2675 						  tb[IFLA_XDP],
2676 						  ifla_xdp_policy, NULL);
2677 		if (err < 0)
2678 			goto errout;
2679 
2680 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2681 			err = -EINVAL;
2682 			goto errout;
2683 		}
2684 
2685 		if (xdp[IFLA_XDP_FLAGS]) {
2686 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2687 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2688 				err = -EINVAL;
2689 				goto errout;
2690 			}
2691 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2692 				err = -EINVAL;
2693 				goto errout;
2694 			}
2695 		}
2696 
2697 		if (xdp[IFLA_XDP_FD]) {
2698 			err = dev_change_xdp_fd(dev, extack,
2699 						nla_get_s32(xdp[IFLA_XDP_FD]),
2700 						xdp_flags);
2701 			if (err)
2702 				goto errout;
2703 			status |= DO_SETLINK_NOTIFY;
2704 		}
2705 	}
2706 
2707 errout:
2708 	if (status & DO_SETLINK_MODIFIED) {
2709 		if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2710 			netdev_state_change(dev);
2711 
2712 		if (err < 0)
2713 			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",
2714 					     dev->name);
2715 	}
2716 
2717 	return err;
2718 }
2719 
2720 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2721 			struct netlink_ext_ack *extack)
2722 {
2723 	struct net *net = sock_net(skb->sk);
2724 	struct ifinfomsg *ifm;
2725 	struct net_device *dev;
2726 	int err;
2727 	struct nlattr *tb[IFLA_MAX+1];
2728 	char ifname[IFNAMSIZ];
2729 
2730 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2731 				     ifla_policy, extack);
2732 	if (err < 0)
2733 		goto errout;
2734 
2735 	err = rtnl_ensure_unique_netns(tb, extack, false);
2736 	if (err < 0)
2737 		goto errout;
2738 
2739 	if (tb[IFLA_IFNAME])
2740 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2741 	else
2742 		ifname[0] = '\0';
2743 
2744 	err = -EINVAL;
2745 	ifm = nlmsg_data(nlh);
2746 	if (ifm->ifi_index > 0)
2747 		dev = __dev_get_by_index(net, ifm->ifi_index);
2748 	else if (tb[IFLA_IFNAME])
2749 		dev = __dev_get_by_name(net, ifname);
2750 	else
2751 		goto errout;
2752 
2753 	if (dev == NULL) {
2754 		err = -ENODEV;
2755 		goto errout;
2756 	}
2757 
2758 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2759 errout:
2760 	return err;
2761 }
2762 
2763 static int rtnl_group_dellink(const struct net *net, int group)
2764 {
2765 	struct net_device *dev, *aux;
2766 	LIST_HEAD(list_kill);
2767 	bool found = false;
2768 
2769 	if (!group)
2770 		return -EPERM;
2771 
2772 	for_each_netdev(net, dev) {
2773 		if (dev->group == group) {
2774 			const struct rtnl_link_ops *ops;
2775 
2776 			found = true;
2777 			ops = dev->rtnl_link_ops;
2778 			if (!ops || !ops->dellink)
2779 				return -EOPNOTSUPP;
2780 		}
2781 	}
2782 
2783 	if (!found)
2784 		return -ENODEV;
2785 
2786 	for_each_netdev_safe(net, dev, aux) {
2787 		if (dev->group == group) {
2788 			const struct rtnl_link_ops *ops;
2789 
2790 			ops = dev->rtnl_link_ops;
2791 			ops->dellink(dev, &list_kill);
2792 		}
2793 	}
2794 	unregister_netdevice_many(&list_kill);
2795 
2796 	return 0;
2797 }
2798 
2799 int rtnl_delete_link(struct net_device *dev)
2800 {
2801 	const struct rtnl_link_ops *ops;
2802 	LIST_HEAD(list_kill);
2803 
2804 	ops = dev->rtnl_link_ops;
2805 	if (!ops || !ops->dellink)
2806 		return -EOPNOTSUPP;
2807 
2808 	ops->dellink(dev, &list_kill);
2809 	unregister_netdevice_many(&list_kill);
2810 
2811 	return 0;
2812 }
2813 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2814 
2815 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2816 			struct netlink_ext_ack *extack)
2817 {
2818 	struct net *net = sock_net(skb->sk);
2819 	struct net *tgt_net = net;
2820 	struct net_device *dev = NULL;
2821 	struct ifinfomsg *ifm;
2822 	char ifname[IFNAMSIZ];
2823 	struct nlattr *tb[IFLA_MAX+1];
2824 	int err;
2825 	int netnsid = -1;
2826 
2827 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2828 				     ifla_policy, extack);
2829 	if (err < 0)
2830 		return err;
2831 
2832 	err = rtnl_ensure_unique_netns(tb, extack, true);
2833 	if (err < 0)
2834 		return err;
2835 
2836 	if (tb[IFLA_IFNAME])
2837 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2838 
2839 	if (tb[IFLA_TARGET_NETNSID]) {
2840 		netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
2841 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
2842 		if (IS_ERR(tgt_net))
2843 			return PTR_ERR(tgt_net);
2844 	}
2845 
2846 	err = -EINVAL;
2847 	ifm = nlmsg_data(nlh);
2848 	if (ifm->ifi_index > 0)
2849 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2850 	else if (tb[IFLA_IFNAME])
2851 		dev = __dev_get_by_name(tgt_net, ifname);
2852 	else if (tb[IFLA_GROUP])
2853 		err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2854 	else
2855 		goto out;
2856 
2857 	if (!dev) {
2858 		if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2859 			err = -ENODEV;
2860 
2861 		goto out;
2862 	}
2863 
2864 	err = rtnl_delete_link(dev);
2865 
2866 out:
2867 	if (netnsid >= 0)
2868 		put_net(tgt_net);
2869 
2870 	return err;
2871 }
2872 
2873 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2874 {
2875 	unsigned int old_flags;
2876 	int err;
2877 
2878 	old_flags = dev->flags;
2879 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2880 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2881 					 NULL);
2882 		if (err < 0)
2883 			return err;
2884 	}
2885 
2886 	if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2887 		__dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2888 	} else {
2889 		dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2890 		__dev_notify_flags(dev, old_flags, ~0U);
2891 	}
2892 	return 0;
2893 }
2894 EXPORT_SYMBOL(rtnl_configure_link);
2895 
2896 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
2897 				    unsigned char name_assign_type,
2898 				    const struct rtnl_link_ops *ops,
2899 				    struct nlattr *tb[],
2900 				    struct netlink_ext_ack *extack)
2901 {
2902 	struct net_device *dev;
2903 	unsigned int num_tx_queues = 1;
2904 	unsigned int num_rx_queues = 1;
2905 
2906 	if (tb[IFLA_NUM_TX_QUEUES])
2907 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2908 	else if (ops->get_num_tx_queues)
2909 		num_tx_queues = ops->get_num_tx_queues();
2910 
2911 	if (tb[IFLA_NUM_RX_QUEUES])
2912 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2913 	else if (ops->get_num_rx_queues)
2914 		num_rx_queues = ops->get_num_rx_queues();
2915 
2916 	if (num_tx_queues < 1 || num_tx_queues > 4096) {
2917 		NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
2918 		return ERR_PTR(-EINVAL);
2919 	}
2920 
2921 	if (num_rx_queues < 1 || num_rx_queues > 4096) {
2922 		NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
2923 		return ERR_PTR(-EINVAL);
2924 	}
2925 
2926 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2927 			       ops->setup, num_tx_queues, num_rx_queues);
2928 	if (!dev)
2929 		return ERR_PTR(-ENOMEM);
2930 
2931 	dev_net_set(dev, net);
2932 	dev->rtnl_link_ops = ops;
2933 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2934 
2935 	if (tb[IFLA_MTU])
2936 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2937 	if (tb[IFLA_ADDRESS]) {
2938 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2939 				nla_len(tb[IFLA_ADDRESS]));
2940 		dev->addr_assign_type = NET_ADDR_SET;
2941 	}
2942 	if (tb[IFLA_BROADCAST])
2943 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2944 				nla_len(tb[IFLA_BROADCAST]));
2945 	if (tb[IFLA_TXQLEN])
2946 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2947 	if (tb[IFLA_OPERSTATE])
2948 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2949 	if (tb[IFLA_LINKMODE])
2950 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2951 	if (tb[IFLA_GROUP])
2952 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2953 	if (tb[IFLA_GSO_MAX_SIZE])
2954 		netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2955 	if (tb[IFLA_GSO_MAX_SEGS])
2956 		dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2957 
2958 	return dev;
2959 }
2960 EXPORT_SYMBOL(rtnl_create_link);
2961 
2962 static int rtnl_group_changelink(const struct sk_buff *skb,
2963 		struct net *net, int group,
2964 		struct ifinfomsg *ifm,
2965 		struct netlink_ext_ack *extack,
2966 		struct nlattr **tb)
2967 {
2968 	struct net_device *dev, *aux;
2969 	int err;
2970 
2971 	for_each_netdev_safe(net, dev, aux) {
2972 		if (dev->group == group) {
2973 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2974 			if (err < 0)
2975 				return err;
2976 		}
2977 	}
2978 
2979 	return 0;
2980 }
2981 
2982 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2983 			  struct nlattr **attr, struct netlink_ext_ack *extack)
2984 {
2985 	struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
2986 	unsigned char name_assign_type = NET_NAME_USER;
2987 	struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
2988 	const struct rtnl_link_ops *m_ops = NULL;
2989 	struct net_device *master_dev = NULL;
2990 	struct net *net = sock_net(skb->sk);
2991 	const struct rtnl_link_ops *ops;
2992 	struct nlattr *tb[IFLA_MAX + 1];
2993 	struct net *dest_net, *link_net;
2994 	struct nlattr **slave_data;
2995 	char kind[MODULE_NAME_LEN];
2996 	struct net_device *dev;
2997 	struct ifinfomsg *ifm;
2998 	char ifname[IFNAMSIZ];
2999 	struct nlattr **data;
3000 	int err;
3001 
3002 #ifdef CONFIG_MODULES
3003 replay:
3004 #endif
3005 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3006 				     ifla_policy, extack);
3007 	if (err < 0)
3008 		return err;
3009 
3010 	err = rtnl_ensure_unique_netns(tb, extack, false);
3011 	if (err < 0)
3012 		return err;
3013 
3014 	if (tb[IFLA_IFNAME])
3015 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3016 	else
3017 		ifname[0] = '\0';
3018 
3019 	ifm = nlmsg_data(nlh);
3020 	if (ifm->ifi_index > 0)
3021 		dev = __dev_get_by_index(net, ifm->ifi_index);
3022 	else {
3023 		if (ifname[0])
3024 			dev = __dev_get_by_name(net, ifname);
3025 		else
3026 			dev = NULL;
3027 	}
3028 
3029 	if (dev) {
3030 		master_dev = netdev_master_upper_dev_get(dev);
3031 		if (master_dev)
3032 			m_ops = master_dev->rtnl_link_ops;
3033 	}
3034 
3035 	err = validate_linkmsg(dev, tb);
3036 	if (err < 0)
3037 		return err;
3038 
3039 	if (tb[IFLA_LINKINFO]) {
3040 		err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3041 						  tb[IFLA_LINKINFO],
3042 						  ifla_info_policy, NULL);
3043 		if (err < 0)
3044 			return err;
3045 	} else
3046 		memset(linkinfo, 0, sizeof(linkinfo));
3047 
3048 	if (linkinfo[IFLA_INFO_KIND]) {
3049 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3050 		ops = rtnl_link_ops_get(kind);
3051 	} else {
3052 		kind[0] = '\0';
3053 		ops = NULL;
3054 	}
3055 
3056 	data = NULL;
3057 	if (ops) {
3058 		if (ops->maxtype > RTNL_MAX_TYPE)
3059 			return -EINVAL;
3060 
3061 		if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3062 			err = nla_parse_nested_deprecated(attr, ops->maxtype,
3063 							  linkinfo[IFLA_INFO_DATA],
3064 							  ops->policy, extack);
3065 			if (err < 0)
3066 				return err;
3067 			data = attr;
3068 		}
3069 		if (ops->validate) {
3070 			err = ops->validate(tb, data, extack);
3071 			if (err < 0)
3072 				return err;
3073 		}
3074 	}
3075 
3076 	slave_data = NULL;
3077 	if (m_ops) {
3078 		if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3079 			return -EINVAL;
3080 
3081 		if (m_ops->slave_maxtype &&
3082 		    linkinfo[IFLA_INFO_SLAVE_DATA]) {
3083 			err = nla_parse_nested_deprecated(slave_attr,
3084 							  m_ops->slave_maxtype,
3085 							  linkinfo[IFLA_INFO_SLAVE_DATA],
3086 							  m_ops->slave_policy,
3087 							  extack);
3088 			if (err < 0)
3089 				return err;
3090 			slave_data = slave_attr;
3091 		}
3092 	}
3093 
3094 	if (dev) {
3095 		int status = 0;
3096 
3097 		if (nlh->nlmsg_flags & NLM_F_EXCL)
3098 			return -EEXIST;
3099 		if (nlh->nlmsg_flags & NLM_F_REPLACE)
3100 			return -EOPNOTSUPP;
3101 
3102 		if (linkinfo[IFLA_INFO_DATA]) {
3103 			if (!ops || ops != dev->rtnl_link_ops ||
3104 			    !ops->changelink)
3105 				return -EOPNOTSUPP;
3106 
3107 			err = ops->changelink(dev, tb, data, extack);
3108 			if (err < 0)
3109 				return err;
3110 			status |= DO_SETLINK_NOTIFY;
3111 		}
3112 
3113 		if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3114 			if (!m_ops || !m_ops->slave_changelink)
3115 				return -EOPNOTSUPP;
3116 
3117 			err = m_ops->slave_changelink(master_dev, dev, tb,
3118 						      slave_data, extack);
3119 			if (err < 0)
3120 				return err;
3121 			status |= DO_SETLINK_NOTIFY;
3122 		}
3123 
3124 		return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3125 	}
3126 
3127 	if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3128 		if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3129 			return rtnl_group_changelink(skb, net,
3130 						nla_get_u32(tb[IFLA_GROUP]),
3131 						ifm, extack, tb);
3132 		return -ENODEV;
3133 	}
3134 
3135 	if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3136 		return -EOPNOTSUPP;
3137 
3138 	if (!ops) {
3139 #ifdef CONFIG_MODULES
3140 		if (kind[0]) {
3141 			__rtnl_unlock();
3142 			request_module("rtnl-link-%s", kind);
3143 			rtnl_lock();
3144 			ops = rtnl_link_ops_get(kind);
3145 			if (ops)
3146 				goto replay;
3147 		}
3148 #endif
3149 		NL_SET_ERR_MSG(extack, "Unknown device type");
3150 		return -EOPNOTSUPP;
3151 	}
3152 
3153 	if (!ops->setup)
3154 		return -EOPNOTSUPP;
3155 
3156 	if (!ifname[0]) {
3157 		snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3158 		name_assign_type = NET_NAME_ENUM;
3159 	}
3160 
3161 	dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3162 	if (IS_ERR(dest_net))
3163 		return PTR_ERR(dest_net);
3164 
3165 	if (tb[IFLA_LINK_NETNSID]) {
3166 		int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3167 
3168 		link_net = get_net_ns_by_id(dest_net, id);
3169 		if (!link_net) {
3170 			NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3171 			err =  -EINVAL;
3172 			goto out;
3173 		}
3174 		err = -EPERM;
3175 		if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3176 			goto out;
3177 	} else {
3178 		link_net = NULL;
3179 	}
3180 
3181 	dev = rtnl_create_link(link_net ? : dest_net, ifname,
3182 			       name_assign_type, ops, tb, extack);
3183 	if (IS_ERR(dev)) {
3184 		err = PTR_ERR(dev);
3185 		goto out;
3186 	}
3187 
3188 	dev->ifindex = ifm->ifi_index;
3189 
3190 	if (ops->newlink) {
3191 		err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3192 		/* Drivers should call free_netdev() in ->destructor
3193 		 * and unregister it on failure after registration
3194 		 * so that device could be finally freed in rtnl_unlock.
3195 		 */
3196 		if (err < 0) {
3197 			/* If device is not registered at all, free it now */
3198 			if (dev->reg_state == NETREG_UNINITIALIZED)
3199 				free_netdev(dev);
3200 			goto out;
3201 		}
3202 	} else {
3203 		err = register_netdevice(dev);
3204 		if (err < 0) {
3205 			free_netdev(dev);
3206 			goto out;
3207 		}
3208 	}
3209 	err = rtnl_configure_link(dev, ifm);
3210 	if (err < 0)
3211 		goto out_unregister;
3212 	if (link_net) {
3213 		err = dev_change_net_namespace(dev, dest_net, ifname);
3214 		if (err < 0)
3215 			goto out_unregister;
3216 	}
3217 	if (tb[IFLA_MASTER]) {
3218 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3219 		if (err)
3220 			goto out_unregister;
3221 	}
3222 out:
3223 	if (link_net)
3224 		put_net(link_net);
3225 	put_net(dest_net);
3226 	return err;
3227 out_unregister:
3228 	if (ops->newlink) {
3229 		LIST_HEAD(list_kill);
3230 
3231 		ops->dellink(dev, &list_kill);
3232 		unregister_netdevice_many(&list_kill);
3233 	} else {
3234 		unregister_netdevice(dev);
3235 	}
3236 	goto out;
3237 }
3238 
3239 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3240 			struct netlink_ext_ack *extack)
3241 {
3242 	struct nlattr **attr;
3243 	int ret;
3244 
3245 	attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3246 	if (!attr)
3247 		return -ENOMEM;
3248 
3249 	ret = __rtnl_newlink(skb, nlh, attr, extack);
3250 	kfree(attr);
3251 	return ret;
3252 }
3253 
3254 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3255 				  const struct nlmsghdr *nlh,
3256 				  struct nlattr **tb,
3257 				  struct netlink_ext_ack *extack)
3258 {
3259 	struct ifinfomsg *ifm;
3260 	int i, err;
3261 
3262 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3263 		NL_SET_ERR_MSG(extack, "Invalid header for get link");
3264 		return -EINVAL;
3265 	}
3266 
3267 	if (!netlink_strict_get_check(skb))
3268 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3269 					      ifla_policy, extack);
3270 
3271 	ifm = nlmsg_data(nlh);
3272 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3273 	    ifm->ifi_change) {
3274 		NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3275 		return -EINVAL;
3276 	}
3277 
3278 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3279 					    ifla_policy, extack);
3280 	if (err)
3281 		return err;
3282 
3283 	for (i = 0; i <= IFLA_MAX; i++) {
3284 		if (!tb[i])
3285 			continue;
3286 
3287 		switch (i) {
3288 		case IFLA_IFNAME:
3289 		case IFLA_EXT_MASK:
3290 		case IFLA_TARGET_NETNSID:
3291 			break;
3292 		default:
3293 			NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3294 			return -EINVAL;
3295 		}
3296 	}
3297 
3298 	return 0;
3299 }
3300 
3301 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3302 			struct netlink_ext_ack *extack)
3303 {
3304 	struct net *net = sock_net(skb->sk);
3305 	struct net *tgt_net = net;
3306 	struct ifinfomsg *ifm;
3307 	char ifname[IFNAMSIZ];
3308 	struct nlattr *tb[IFLA_MAX+1];
3309 	struct net_device *dev = NULL;
3310 	struct sk_buff *nskb;
3311 	int netnsid = -1;
3312 	int err;
3313 	u32 ext_filter_mask = 0;
3314 
3315 	err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3316 	if (err < 0)
3317 		return err;
3318 
3319 	err = rtnl_ensure_unique_netns(tb, extack, true);
3320 	if (err < 0)
3321 		return err;
3322 
3323 	if (tb[IFLA_TARGET_NETNSID]) {
3324 		netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3325 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3326 		if (IS_ERR(tgt_net))
3327 			return PTR_ERR(tgt_net);
3328 	}
3329 
3330 	if (tb[IFLA_IFNAME])
3331 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3332 
3333 	if (tb[IFLA_EXT_MASK])
3334 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3335 
3336 	err = -EINVAL;
3337 	ifm = nlmsg_data(nlh);
3338 	if (ifm->ifi_index > 0)
3339 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3340 	else if (tb[IFLA_IFNAME])
3341 		dev = __dev_get_by_name(tgt_net, ifname);
3342 	else
3343 		goto out;
3344 
3345 	err = -ENODEV;
3346 	if (dev == NULL)
3347 		goto out;
3348 
3349 	err = -ENOBUFS;
3350 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3351 	if (nskb == NULL)
3352 		goto out;
3353 
3354 	err = rtnl_fill_ifinfo(nskb, dev, net,
3355 			       RTM_NEWLINK, NETLINK_CB(skb).portid,
3356 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3357 			       0, NULL, 0, netnsid);
3358 	if (err < 0) {
3359 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
3360 		WARN_ON(err == -EMSGSIZE);
3361 		kfree_skb(nskb);
3362 	} else
3363 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3364 out:
3365 	if (netnsid >= 0)
3366 		put_net(tgt_net);
3367 
3368 	return err;
3369 }
3370 
3371 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3372 {
3373 	struct net *net = sock_net(skb->sk);
3374 	struct net_device *dev;
3375 	struct nlattr *tb[IFLA_MAX+1];
3376 	u32 ext_filter_mask = 0;
3377 	u16 min_ifinfo_dump_size = 0;
3378 	int hdrlen;
3379 
3380 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3381 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3382 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3383 
3384 	if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3385 		if (tb[IFLA_EXT_MASK])
3386 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3387 	}
3388 
3389 	if (!ext_filter_mask)
3390 		return NLMSG_GOODSIZE;
3391 	/*
3392 	 * traverse the list of net devices and compute the minimum
3393 	 * buffer size based upon the filter mask.
3394 	 */
3395 	rcu_read_lock();
3396 	for_each_netdev_rcu(net, dev) {
3397 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3398 					     if_nlmsg_size(dev,
3399 						           ext_filter_mask));
3400 	}
3401 	rcu_read_unlock();
3402 
3403 	return nlmsg_total_size(min_ifinfo_dump_size);
3404 }
3405 
3406 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3407 {
3408 	int idx;
3409 	int s_idx = cb->family;
3410 	int type = cb->nlh->nlmsg_type - RTM_BASE;
3411 	int ret = 0;
3412 
3413 	if (s_idx == 0)
3414 		s_idx = 1;
3415 
3416 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3417 		struct rtnl_link **tab;
3418 		struct rtnl_link *link;
3419 		rtnl_dumpit_func dumpit;
3420 
3421 		if (idx < s_idx || idx == PF_PACKET)
3422 			continue;
3423 
3424 		if (type < 0 || type >= RTM_NR_MSGTYPES)
3425 			continue;
3426 
3427 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3428 		if (!tab)
3429 			continue;
3430 
3431 		link = tab[type];
3432 		if (!link)
3433 			continue;
3434 
3435 		dumpit = link->dumpit;
3436 		if (!dumpit)
3437 			continue;
3438 
3439 		if (idx > s_idx) {
3440 			memset(&cb->args[0], 0, sizeof(cb->args));
3441 			cb->prev_seq = 0;
3442 			cb->seq = 0;
3443 		}
3444 		ret = dumpit(skb, cb);
3445 		if (ret)
3446 			break;
3447 	}
3448 	cb->family = idx;
3449 
3450 	return skb->len ? : ret;
3451 }
3452 
3453 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3454 				       unsigned int change,
3455 				       u32 event, gfp_t flags, int *new_nsid,
3456 				       int new_ifindex)
3457 {
3458 	struct net *net = dev_net(dev);
3459 	struct sk_buff *skb;
3460 	int err = -ENOBUFS;
3461 	size_t if_info_size;
3462 
3463 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3464 	if (skb == NULL)
3465 		goto errout;
3466 
3467 	err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3468 			       type, 0, 0, change, 0, 0, event,
3469 			       new_nsid, new_ifindex, -1);
3470 	if (err < 0) {
3471 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
3472 		WARN_ON(err == -EMSGSIZE);
3473 		kfree_skb(skb);
3474 		goto errout;
3475 	}
3476 	return skb;
3477 errout:
3478 	if (err < 0)
3479 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3480 	return NULL;
3481 }
3482 
3483 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3484 {
3485 	struct net *net = dev_net(dev);
3486 
3487 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3488 }
3489 
3490 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3491 			       unsigned int change, u32 event,
3492 			       gfp_t flags, int *new_nsid, int new_ifindex)
3493 {
3494 	struct sk_buff *skb;
3495 
3496 	if (dev->reg_state != NETREG_REGISTERED)
3497 		return;
3498 
3499 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3500 				     new_ifindex);
3501 	if (skb)
3502 		rtmsg_ifinfo_send(skb, dev, flags);
3503 }
3504 
3505 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3506 		  gfp_t flags)
3507 {
3508 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3509 			   NULL, 0);
3510 }
3511 
3512 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3513 			 gfp_t flags, int *new_nsid, int new_ifindex)
3514 {
3515 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3516 			   new_nsid, new_ifindex);
3517 }
3518 
3519 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3520 				   struct net_device *dev,
3521 				   u8 *addr, u16 vid, u32 pid, u32 seq,
3522 				   int type, unsigned int flags,
3523 				   int nlflags, u16 ndm_state)
3524 {
3525 	struct nlmsghdr *nlh;
3526 	struct ndmsg *ndm;
3527 
3528 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3529 	if (!nlh)
3530 		return -EMSGSIZE;
3531 
3532 	ndm = nlmsg_data(nlh);
3533 	ndm->ndm_family  = AF_BRIDGE;
3534 	ndm->ndm_pad1	 = 0;
3535 	ndm->ndm_pad2    = 0;
3536 	ndm->ndm_flags	 = flags;
3537 	ndm->ndm_type	 = 0;
3538 	ndm->ndm_ifindex = dev->ifindex;
3539 	ndm->ndm_state   = ndm_state;
3540 
3541 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3542 		goto nla_put_failure;
3543 	if (vid)
3544 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3545 			goto nla_put_failure;
3546 
3547 	nlmsg_end(skb, nlh);
3548 	return 0;
3549 
3550 nla_put_failure:
3551 	nlmsg_cancel(skb, nlh);
3552 	return -EMSGSIZE;
3553 }
3554 
3555 static inline size_t rtnl_fdb_nlmsg_size(void)
3556 {
3557 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3558 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
3559 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
3560 	       0;
3561 }
3562 
3563 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3564 			    u16 ndm_state)
3565 {
3566 	struct net *net = dev_net(dev);
3567 	struct sk_buff *skb;
3568 	int err = -ENOBUFS;
3569 
3570 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3571 	if (!skb)
3572 		goto errout;
3573 
3574 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3575 				      0, 0, type, NTF_SELF, 0, ndm_state);
3576 	if (err < 0) {
3577 		kfree_skb(skb);
3578 		goto errout;
3579 	}
3580 
3581 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3582 	return;
3583 errout:
3584 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3585 }
3586 
3587 /*
3588  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3589  */
3590 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3591 		     struct nlattr *tb[],
3592 		     struct net_device *dev,
3593 		     const unsigned char *addr, u16 vid,
3594 		     u16 flags)
3595 {
3596 	int err = -EINVAL;
3597 
3598 	/* If aging addresses are supported device will need to
3599 	 * implement its own handler for this.
3600 	 */
3601 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3602 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3603 		return err;
3604 	}
3605 
3606 	if (vid) {
3607 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3608 		return err;
3609 	}
3610 
3611 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3612 		err = dev_uc_add_excl(dev, addr);
3613 	else if (is_multicast_ether_addr(addr))
3614 		err = dev_mc_add_excl(dev, addr);
3615 
3616 	/* Only return duplicate errors if NLM_F_EXCL is set */
3617 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3618 		err = 0;
3619 
3620 	return err;
3621 }
3622 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3623 
3624 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3625 			 struct netlink_ext_ack *extack)
3626 {
3627 	u16 vid = 0;
3628 
3629 	if (vlan_attr) {
3630 		if (nla_len(vlan_attr) != sizeof(u16)) {
3631 			NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3632 			return -EINVAL;
3633 		}
3634 
3635 		vid = nla_get_u16(vlan_attr);
3636 
3637 		if (!vid || vid >= VLAN_VID_MASK) {
3638 			NL_SET_ERR_MSG(extack, "invalid vlan id");
3639 			return -EINVAL;
3640 		}
3641 	}
3642 	*p_vid = vid;
3643 	return 0;
3644 }
3645 
3646 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3647 			struct netlink_ext_ack *extack)
3648 {
3649 	struct net *net = sock_net(skb->sk);
3650 	struct ndmsg *ndm;
3651 	struct nlattr *tb[NDA_MAX+1];
3652 	struct net_device *dev;
3653 	u8 *addr;
3654 	u16 vid;
3655 	int err;
3656 
3657 	err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3658 				     extack);
3659 	if (err < 0)
3660 		return err;
3661 
3662 	ndm = nlmsg_data(nlh);
3663 	if (ndm->ndm_ifindex == 0) {
3664 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3665 		return -EINVAL;
3666 	}
3667 
3668 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3669 	if (dev == NULL) {
3670 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3671 		return -ENODEV;
3672 	}
3673 
3674 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3675 		NL_SET_ERR_MSG(extack, "invalid address");
3676 		return -EINVAL;
3677 	}
3678 
3679 	if (dev->type != ARPHRD_ETHER) {
3680 		NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3681 		return -EINVAL;
3682 	}
3683 
3684 	addr = nla_data(tb[NDA_LLADDR]);
3685 
3686 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3687 	if (err)
3688 		return err;
3689 
3690 	err = -EOPNOTSUPP;
3691 
3692 	/* Support fdb on master device the net/bridge default case */
3693 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3694 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3695 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3696 		const struct net_device_ops *ops = br_dev->netdev_ops;
3697 
3698 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3699 				       nlh->nlmsg_flags, extack);
3700 		if (err)
3701 			goto out;
3702 		else
3703 			ndm->ndm_flags &= ~NTF_MASTER;
3704 	}
3705 
3706 	/* Embedded bridge, macvlan, and any other device support */
3707 	if ((ndm->ndm_flags & NTF_SELF)) {
3708 		if (dev->netdev_ops->ndo_fdb_add)
3709 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3710 							   vid,
3711 							   nlh->nlmsg_flags,
3712 							   extack);
3713 		else
3714 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3715 					       nlh->nlmsg_flags);
3716 
3717 		if (!err) {
3718 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3719 					ndm->ndm_state);
3720 			ndm->ndm_flags &= ~NTF_SELF;
3721 		}
3722 	}
3723 out:
3724 	return err;
3725 }
3726 
3727 /*
3728  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3729  */
3730 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3731 		     struct nlattr *tb[],
3732 		     struct net_device *dev,
3733 		     const unsigned char *addr, u16 vid)
3734 {
3735 	int err = -EINVAL;
3736 
3737 	/* If aging addresses are supported device will need to
3738 	 * implement its own handler for this.
3739 	 */
3740 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3741 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3742 		return err;
3743 	}
3744 
3745 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3746 		err = dev_uc_del(dev, addr);
3747 	else if (is_multicast_ether_addr(addr))
3748 		err = dev_mc_del(dev, addr);
3749 
3750 	return err;
3751 }
3752 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3753 
3754 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3755 			struct netlink_ext_ack *extack)
3756 {
3757 	struct net *net = sock_net(skb->sk);
3758 	struct ndmsg *ndm;
3759 	struct nlattr *tb[NDA_MAX+1];
3760 	struct net_device *dev;
3761 	int err = -EINVAL;
3762 	__u8 *addr;
3763 	u16 vid;
3764 
3765 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3766 		return -EPERM;
3767 
3768 	err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3769 				     extack);
3770 	if (err < 0)
3771 		return err;
3772 
3773 	ndm = nlmsg_data(nlh);
3774 	if (ndm->ndm_ifindex == 0) {
3775 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3776 		return -EINVAL;
3777 	}
3778 
3779 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3780 	if (dev == NULL) {
3781 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3782 		return -ENODEV;
3783 	}
3784 
3785 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3786 		NL_SET_ERR_MSG(extack, "invalid address");
3787 		return -EINVAL;
3788 	}
3789 
3790 	if (dev->type != ARPHRD_ETHER) {
3791 		NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
3792 		return -EINVAL;
3793 	}
3794 
3795 	addr = nla_data(tb[NDA_LLADDR]);
3796 
3797 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3798 	if (err)
3799 		return err;
3800 
3801 	err = -EOPNOTSUPP;
3802 
3803 	/* Support fdb on master device the net/bridge default case */
3804 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3805 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3806 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3807 		const struct net_device_ops *ops = br_dev->netdev_ops;
3808 
3809 		if (ops->ndo_fdb_del)
3810 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3811 
3812 		if (err)
3813 			goto out;
3814 		else
3815 			ndm->ndm_flags &= ~NTF_MASTER;
3816 	}
3817 
3818 	/* Embedded bridge, macvlan, and any other device support */
3819 	if (ndm->ndm_flags & NTF_SELF) {
3820 		if (dev->netdev_ops->ndo_fdb_del)
3821 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3822 							   vid);
3823 		else
3824 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3825 
3826 		if (!err) {
3827 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3828 					ndm->ndm_state);
3829 			ndm->ndm_flags &= ~NTF_SELF;
3830 		}
3831 	}
3832 out:
3833 	return err;
3834 }
3835 
3836 static int nlmsg_populate_fdb(struct sk_buff *skb,
3837 			      struct netlink_callback *cb,
3838 			      struct net_device *dev,
3839 			      int *idx,
3840 			      struct netdev_hw_addr_list *list)
3841 {
3842 	struct netdev_hw_addr *ha;
3843 	int err;
3844 	u32 portid, seq;
3845 
3846 	portid = NETLINK_CB(cb->skb).portid;
3847 	seq = cb->nlh->nlmsg_seq;
3848 
3849 	list_for_each_entry(ha, &list->list, list) {
3850 		if (*idx < cb->args[2])
3851 			goto skip;
3852 
3853 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3854 					      portid, seq,
3855 					      RTM_NEWNEIGH, NTF_SELF,
3856 					      NLM_F_MULTI, NUD_PERMANENT);
3857 		if (err < 0)
3858 			return err;
3859 skip:
3860 		*idx += 1;
3861 	}
3862 	return 0;
3863 }
3864 
3865 /**
3866  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3867  * @skb: socket buffer to store message in
3868  * @cb: netlink callback
3869  * @dev: netdevice
3870  * @filter_dev: ignored
3871  * @idx: the number of FDB table entries dumped is added to *@idx
3872  *
3873  * Default netdevice operation to dump the existing unicast address list.
3874  * Returns number of addresses from list put in skb.
3875  */
3876 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3877 		      struct netlink_callback *cb,
3878 		      struct net_device *dev,
3879 		      struct net_device *filter_dev,
3880 		      int *idx)
3881 {
3882 	int err;
3883 
3884 	if (dev->type != ARPHRD_ETHER)
3885 		return -EINVAL;
3886 
3887 	netif_addr_lock_bh(dev);
3888 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3889 	if (err)
3890 		goto out;
3891 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3892 out:
3893 	netif_addr_unlock_bh(dev);
3894 	return err;
3895 }
3896 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3897 
3898 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
3899 				 int *br_idx, int *brport_idx,
3900 				 struct netlink_ext_ack *extack)
3901 {
3902 	struct nlattr *tb[NDA_MAX + 1];
3903 	struct ndmsg *ndm;
3904 	int err, i;
3905 
3906 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
3907 		NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
3908 		return -EINVAL;
3909 	}
3910 
3911 	ndm = nlmsg_data(nlh);
3912 	if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
3913 	    ndm->ndm_flags || ndm->ndm_type) {
3914 		NL_SET_ERR_MSG(extack, "Invalid values in header for fbd dump request");
3915 		return -EINVAL;
3916 	}
3917 
3918 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
3919 					    NDA_MAX, NULL, extack);
3920 	if (err < 0)
3921 		return err;
3922 
3923 	*brport_idx = ndm->ndm_ifindex;
3924 	for (i = 0; i <= NDA_MAX; ++i) {
3925 		if (!tb[i])
3926 			continue;
3927 
3928 		switch (i) {
3929 		case NDA_IFINDEX:
3930 			if (nla_len(tb[i]) != sizeof(u32)) {
3931 				NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
3932 				return -EINVAL;
3933 			}
3934 			*brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
3935 			break;
3936 		case NDA_MASTER:
3937 			if (nla_len(tb[i]) != sizeof(u32)) {
3938 				NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
3939 				return -EINVAL;
3940 			}
3941 			*br_idx = nla_get_u32(tb[NDA_MASTER]);
3942 			break;
3943 		default:
3944 			NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
3945 			return -EINVAL;
3946 		}
3947 	}
3948 
3949 	return 0;
3950 }
3951 
3952 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
3953 				 int *br_idx, int *brport_idx,
3954 				 struct netlink_ext_ack *extack)
3955 {
3956 	struct nlattr *tb[IFLA_MAX+1];
3957 	int err;
3958 
3959 	/* A hack to preserve kernel<->userspace interface.
3960 	 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
3961 	 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
3962 	 * So, check for ndmsg with an optional u32 attribute (not used here).
3963 	 * Fortunately these sizes don't conflict with the size of ifinfomsg
3964 	 * with an optional attribute.
3965 	 */
3966 	if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
3967 	    (nlmsg_len(nlh) != sizeof(struct ndmsg) +
3968 	     nla_attr_size(sizeof(u32)))) {
3969 		struct ifinfomsg *ifm;
3970 
3971 		err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
3972 					     tb, IFLA_MAX, ifla_policy,
3973 					     extack);
3974 		if (err < 0) {
3975 			return -EINVAL;
3976 		} else if (err == 0) {
3977 			if (tb[IFLA_MASTER])
3978 				*br_idx = nla_get_u32(tb[IFLA_MASTER]);
3979 		}
3980 
3981 		ifm = nlmsg_data(nlh);
3982 		*brport_idx = ifm->ifi_index;
3983 	}
3984 	return 0;
3985 }
3986 
3987 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3988 {
3989 	struct net_device *dev;
3990 	struct net_device *br_dev = NULL;
3991 	const struct net_device_ops *ops = NULL;
3992 	const struct net_device_ops *cops = NULL;
3993 	struct net *net = sock_net(skb->sk);
3994 	struct hlist_head *head;
3995 	int brport_idx = 0;
3996 	int br_idx = 0;
3997 	int h, s_h;
3998 	int idx = 0, s_idx;
3999 	int err = 0;
4000 	int fidx = 0;
4001 
4002 	if (cb->strict_check)
4003 		err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4004 					    cb->extack);
4005 	else
4006 		err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4007 					    cb->extack);
4008 	if (err < 0)
4009 		return err;
4010 
4011 	if (br_idx) {
4012 		br_dev = __dev_get_by_index(net, br_idx);
4013 		if (!br_dev)
4014 			return -ENODEV;
4015 
4016 		ops = br_dev->netdev_ops;
4017 	}
4018 
4019 	s_h = cb->args[0];
4020 	s_idx = cb->args[1];
4021 
4022 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4023 		idx = 0;
4024 		head = &net->dev_index_head[h];
4025 		hlist_for_each_entry(dev, head, index_hlist) {
4026 
4027 			if (brport_idx && (dev->ifindex != brport_idx))
4028 				continue;
4029 
4030 			if (!br_idx) { /* user did not specify a specific bridge */
4031 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
4032 					br_dev = netdev_master_upper_dev_get(dev);
4033 					cops = br_dev->netdev_ops;
4034 				}
4035 			} else {
4036 				if (dev != br_dev &&
4037 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
4038 					continue;
4039 
4040 				if (br_dev != netdev_master_upper_dev_get(dev) &&
4041 				    !(dev->priv_flags & IFF_EBRIDGE))
4042 					continue;
4043 				cops = ops;
4044 			}
4045 
4046 			if (idx < s_idx)
4047 				goto cont;
4048 
4049 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
4050 				if (cops && cops->ndo_fdb_dump) {
4051 					err = cops->ndo_fdb_dump(skb, cb,
4052 								br_dev, dev,
4053 								&fidx);
4054 					if (err == -EMSGSIZE)
4055 						goto out;
4056 				}
4057 			}
4058 
4059 			if (dev->netdev_ops->ndo_fdb_dump)
4060 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4061 								    dev, NULL,
4062 								    &fidx);
4063 			else
4064 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4065 							&fidx);
4066 			if (err == -EMSGSIZE)
4067 				goto out;
4068 
4069 			cops = NULL;
4070 
4071 			/* reset fdb offset to 0 for rest of the interfaces */
4072 			cb->args[2] = 0;
4073 			fidx = 0;
4074 cont:
4075 			idx++;
4076 		}
4077 	}
4078 
4079 out:
4080 	cb->args[0] = h;
4081 	cb->args[1] = idx;
4082 	cb->args[2] = fidx;
4083 
4084 	return skb->len;
4085 }
4086 
4087 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4088 				struct nlattr **tb, u8 *ndm_flags,
4089 				int *br_idx, int *brport_idx, u8 **addr,
4090 				u16 *vid, struct netlink_ext_ack *extack)
4091 {
4092 	struct ndmsg *ndm;
4093 	int err, i;
4094 
4095 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4096 		NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4097 		return -EINVAL;
4098 	}
4099 
4100 	ndm = nlmsg_data(nlh);
4101 	if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4102 	    ndm->ndm_type) {
4103 		NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4104 		return -EINVAL;
4105 	}
4106 
4107 	if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4108 		NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4109 		return -EINVAL;
4110 	}
4111 
4112 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4113 					    NDA_MAX, nda_policy, extack);
4114 	if (err < 0)
4115 		return err;
4116 
4117 	*ndm_flags = ndm->ndm_flags;
4118 	*brport_idx = ndm->ndm_ifindex;
4119 	for (i = 0; i <= NDA_MAX; ++i) {
4120 		if (!tb[i])
4121 			continue;
4122 
4123 		switch (i) {
4124 		case NDA_MASTER:
4125 			*br_idx = nla_get_u32(tb[i]);
4126 			break;
4127 		case NDA_LLADDR:
4128 			if (nla_len(tb[i]) != ETH_ALEN) {
4129 				NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4130 				return -EINVAL;
4131 			}
4132 			*addr = nla_data(tb[i]);
4133 			break;
4134 		case NDA_VLAN:
4135 			err = fdb_vid_parse(tb[i], vid, extack);
4136 			if (err)
4137 				return err;
4138 			break;
4139 		case NDA_VNI:
4140 			break;
4141 		default:
4142 			NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4143 			return -EINVAL;
4144 		}
4145 	}
4146 
4147 	return 0;
4148 }
4149 
4150 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4151 			struct netlink_ext_ack *extack)
4152 {
4153 	struct net_device *dev = NULL, *br_dev = NULL;
4154 	const struct net_device_ops *ops = NULL;
4155 	struct net *net = sock_net(in_skb->sk);
4156 	struct nlattr *tb[NDA_MAX + 1];
4157 	struct sk_buff *skb;
4158 	int brport_idx = 0;
4159 	u8 ndm_flags = 0;
4160 	int br_idx = 0;
4161 	u8 *addr = NULL;
4162 	u16 vid = 0;
4163 	int err;
4164 
4165 	err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4166 				   &brport_idx, &addr, &vid, extack);
4167 	if (err < 0)
4168 		return err;
4169 
4170 	if (!addr) {
4171 		NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4172 		return -EINVAL;
4173 	}
4174 
4175 	if (brport_idx) {
4176 		dev = __dev_get_by_index(net, brport_idx);
4177 		if (!dev) {
4178 			NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4179 			return -ENODEV;
4180 		}
4181 	}
4182 
4183 	if (br_idx) {
4184 		if (dev) {
4185 			NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4186 			return -EINVAL;
4187 		}
4188 
4189 		br_dev = __dev_get_by_index(net, br_idx);
4190 		if (!br_dev) {
4191 			NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4192 			return -EINVAL;
4193 		}
4194 		ops = br_dev->netdev_ops;
4195 	}
4196 
4197 	if (dev) {
4198 		if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4199 			if (!(dev->priv_flags & IFF_BRIDGE_PORT)) {
4200 				NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4201 				return -EINVAL;
4202 			}
4203 			br_dev = netdev_master_upper_dev_get(dev);
4204 			if (!br_dev) {
4205 				NL_SET_ERR_MSG(extack, "Master of device not found");
4206 				return -EINVAL;
4207 			}
4208 			ops = br_dev->netdev_ops;
4209 		} else {
4210 			if (!(ndm_flags & NTF_SELF)) {
4211 				NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4212 				return -EINVAL;
4213 			}
4214 			ops = dev->netdev_ops;
4215 		}
4216 	}
4217 
4218 	if (!br_dev && !dev) {
4219 		NL_SET_ERR_MSG(extack, "No device specified");
4220 		return -ENODEV;
4221 	}
4222 
4223 	if (!ops || !ops->ndo_fdb_get) {
4224 		NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4225 		return -EOPNOTSUPP;
4226 	}
4227 
4228 	skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4229 	if (!skb)
4230 		return -ENOBUFS;
4231 
4232 	if (br_dev)
4233 		dev = br_dev;
4234 	err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4235 			       NETLINK_CB(in_skb).portid,
4236 			       nlh->nlmsg_seq, extack);
4237 	if (err)
4238 		goto out;
4239 
4240 	return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4241 out:
4242 	kfree_skb(skb);
4243 	return err;
4244 }
4245 
4246 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4247 			       unsigned int attrnum, unsigned int flag)
4248 {
4249 	if (mask & flag)
4250 		return nla_put_u8(skb, attrnum, !!(flags & flag));
4251 	return 0;
4252 }
4253 
4254 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4255 			    struct net_device *dev, u16 mode,
4256 			    u32 flags, u32 mask, int nlflags,
4257 			    u32 filter_mask,
4258 			    int (*vlan_fill)(struct sk_buff *skb,
4259 					     struct net_device *dev,
4260 					     u32 filter_mask))
4261 {
4262 	struct nlmsghdr *nlh;
4263 	struct ifinfomsg *ifm;
4264 	struct nlattr *br_afspec;
4265 	struct nlattr *protinfo;
4266 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4267 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4268 	int err = 0;
4269 
4270 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4271 	if (nlh == NULL)
4272 		return -EMSGSIZE;
4273 
4274 	ifm = nlmsg_data(nlh);
4275 	ifm->ifi_family = AF_BRIDGE;
4276 	ifm->__ifi_pad = 0;
4277 	ifm->ifi_type = dev->type;
4278 	ifm->ifi_index = dev->ifindex;
4279 	ifm->ifi_flags = dev_get_flags(dev);
4280 	ifm->ifi_change = 0;
4281 
4282 
4283 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4284 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4285 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4286 	    (br_dev &&
4287 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4288 	    (dev->addr_len &&
4289 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4290 	    (dev->ifindex != dev_get_iflink(dev) &&
4291 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4292 		goto nla_put_failure;
4293 
4294 	br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4295 	if (!br_afspec)
4296 		goto nla_put_failure;
4297 
4298 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4299 		nla_nest_cancel(skb, br_afspec);
4300 		goto nla_put_failure;
4301 	}
4302 
4303 	if (mode != BRIDGE_MODE_UNDEF) {
4304 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4305 			nla_nest_cancel(skb, br_afspec);
4306 			goto nla_put_failure;
4307 		}
4308 	}
4309 	if (vlan_fill) {
4310 		err = vlan_fill(skb, dev, filter_mask);
4311 		if (err) {
4312 			nla_nest_cancel(skb, br_afspec);
4313 			goto nla_put_failure;
4314 		}
4315 	}
4316 	nla_nest_end(skb, br_afspec);
4317 
4318 	protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4319 	if (!protinfo)
4320 		goto nla_put_failure;
4321 
4322 	if (brport_nla_put_flag(skb, flags, mask,
4323 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4324 	    brport_nla_put_flag(skb, flags, mask,
4325 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4326 	    brport_nla_put_flag(skb, flags, mask,
4327 				IFLA_BRPORT_FAST_LEAVE,
4328 				BR_MULTICAST_FAST_LEAVE) ||
4329 	    brport_nla_put_flag(skb, flags, mask,
4330 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4331 	    brport_nla_put_flag(skb, flags, mask,
4332 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4333 	    brport_nla_put_flag(skb, flags, mask,
4334 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4335 	    brport_nla_put_flag(skb, flags, mask,
4336 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4337 	    brport_nla_put_flag(skb, flags, mask,
4338 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
4339 		nla_nest_cancel(skb, protinfo);
4340 		goto nla_put_failure;
4341 	}
4342 
4343 	nla_nest_end(skb, protinfo);
4344 
4345 	nlmsg_end(skb, nlh);
4346 	return 0;
4347 nla_put_failure:
4348 	nlmsg_cancel(skb, nlh);
4349 	return err ? err : -EMSGSIZE;
4350 }
4351 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4352 
4353 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4354 				    bool strict_check, u32 *filter_mask,
4355 				    struct netlink_ext_ack *extack)
4356 {
4357 	struct nlattr *tb[IFLA_MAX+1];
4358 	int err, i;
4359 
4360 	if (strict_check) {
4361 		struct ifinfomsg *ifm;
4362 
4363 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4364 			NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4365 			return -EINVAL;
4366 		}
4367 
4368 		ifm = nlmsg_data(nlh);
4369 		if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4370 		    ifm->ifi_change || ifm->ifi_index) {
4371 			NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4372 			return -EINVAL;
4373 		}
4374 
4375 		err = nlmsg_parse_deprecated_strict(nlh,
4376 						    sizeof(struct ifinfomsg),
4377 						    tb, IFLA_MAX, ifla_policy,
4378 						    extack);
4379 	} else {
4380 		err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4381 					     tb, IFLA_MAX, ifla_policy,
4382 					     extack);
4383 	}
4384 	if (err < 0)
4385 		return err;
4386 
4387 	/* new attributes should only be added with strict checking */
4388 	for (i = 0; i <= IFLA_MAX; ++i) {
4389 		if (!tb[i])
4390 			continue;
4391 
4392 		switch (i) {
4393 		case IFLA_EXT_MASK:
4394 			*filter_mask = nla_get_u32(tb[i]);
4395 			break;
4396 		default:
4397 			if (strict_check) {
4398 				NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4399 				return -EINVAL;
4400 			}
4401 		}
4402 	}
4403 
4404 	return 0;
4405 }
4406 
4407 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4408 {
4409 	const struct nlmsghdr *nlh = cb->nlh;
4410 	struct net *net = sock_net(skb->sk);
4411 	struct net_device *dev;
4412 	int idx = 0;
4413 	u32 portid = NETLINK_CB(cb->skb).portid;
4414 	u32 seq = nlh->nlmsg_seq;
4415 	u32 filter_mask = 0;
4416 	int err;
4417 
4418 	err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4419 				       cb->extack);
4420 	if (err < 0 && cb->strict_check)
4421 		return err;
4422 
4423 	rcu_read_lock();
4424 	for_each_netdev_rcu(net, dev) {
4425 		const struct net_device_ops *ops = dev->netdev_ops;
4426 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4427 
4428 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4429 			if (idx >= cb->args[0]) {
4430 				err = br_dev->netdev_ops->ndo_bridge_getlink(
4431 						skb, portid, seq, dev,
4432 						filter_mask, NLM_F_MULTI);
4433 				if (err < 0 && err != -EOPNOTSUPP) {
4434 					if (likely(skb->len))
4435 						break;
4436 
4437 					goto out_err;
4438 				}
4439 			}
4440 			idx++;
4441 		}
4442 
4443 		if (ops->ndo_bridge_getlink) {
4444 			if (idx >= cb->args[0]) {
4445 				err = ops->ndo_bridge_getlink(skb, portid,
4446 							      seq, dev,
4447 							      filter_mask,
4448 							      NLM_F_MULTI);
4449 				if (err < 0 && err != -EOPNOTSUPP) {
4450 					if (likely(skb->len))
4451 						break;
4452 
4453 					goto out_err;
4454 				}
4455 			}
4456 			idx++;
4457 		}
4458 	}
4459 	err = skb->len;
4460 out_err:
4461 	rcu_read_unlock();
4462 	cb->args[0] = idx;
4463 
4464 	return err;
4465 }
4466 
4467 static inline size_t bridge_nlmsg_size(void)
4468 {
4469 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4470 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
4471 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
4472 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
4473 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
4474 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
4475 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
4476 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
4477 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
4478 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
4479 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
4480 }
4481 
4482 static int rtnl_bridge_notify(struct net_device *dev)
4483 {
4484 	struct net *net = dev_net(dev);
4485 	struct sk_buff *skb;
4486 	int err = -EOPNOTSUPP;
4487 
4488 	if (!dev->netdev_ops->ndo_bridge_getlink)
4489 		return 0;
4490 
4491 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4492 	if (!skb) {
4493 		err = -ENOMEM;
4494 		goto errout;
4495 	}
4496 
4497 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4498 	if (err < 0)
4499 		goto errout;
4500 
4501 	if (!skb->len)
4502 		goto errout;
4503 
4504 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4505 	return 0;
4506 errout:
4507 	WARN_ON(err == -EMSGSIZE);
4508 	kfree_skb(skb);
4509 	if (err)
4510 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4511 	return err;
4512 }
4513 
4514 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4515 			       struct netlink_ext_ack *extack)
4516 {
4517 	struct net *net = sock_net(skb->sk);
4518 	struct ifinfomsg *ifm;
4519 	struct net_device *dev;
4520 	struct nlattr *br_spec, *attr = NULL;
4521 	int rem, err = -EOPNOTSUPP;
4522 	u16 flags = 0;
4523 	bool have_flags = false;
4524 
4525 	if (nlmsg_len(nlh) < sizeof(*ifm))
4526 		return -EINVAL;
4527 
4528 	ifm = nlmsg_data(nlh);
4529 	if (ifm->ifi_family != AF_BRIDGE)
4530 		return -EPFNOSUPPORT;
4531 
4532 	dev = __dev_get_by_index(net, ifm->ifi_index);
4533 	if (!dev) {
4534 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4535 		return -ENODEV;
4536 	}
4537 
4538 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4539 	if (br_spec) {
4540 		nla_for_each_nested(attr, br_spec, rem) {
4541 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4542 				if (nla_len(attr) < sizeof(flags))
4543 					return -EINVAL;
4544 
4545 				have_flags = true;
4546 				flags = nla_get_u16(attr);
4547 				break;
4548 			}
4549 		}
4550 	}
4551 
4552 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4553 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4554 
4555 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4556 			err = -EOPNOTSUPP;
4557 			goto out;
4558 		}
4559 
4560 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4561 							     extack);
4562 		if (err)
4563 			goto out;
4564 
4565 		flags &= ~BRIDGE_FLAGS_MASTER;
4566 	}
4567 
4568 	if ((flags & BRIDGE_FLAGS_SELF)) {
4569 		if (!dev->netdev_ops->ndo_bridge_setlink)
4570 			err = -EOPNOTSUPP;
4571 		else
4572 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4573 								  flags,
4574 								  extack);
4575 		if (!err) {
4576 			flags &= ~BRIDGE_FLAGS_SELF;
4577 
4578 			/* Generate event to notify upper layer of bridge
4579 			 * change
4580 			 */
4581 			err = rtnl_bridge_notify(dev);
4582 		}
4583 	}
4584 
4585 	if (have_flags)
4586 		memcpy(nla_data(attr), &flags, sizeof(flags));
4587 out:
4588 	return err;
4589 }
4590 
4591 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4592 			       struct netlink_ext_ack *extack)
4593 {
4594 	struct net *net = sock_net(skb->sk);
4595 	struct ifinfomsg *ifm;
4596 	struct net_device *dev;
4597 	struct nlattr *br_spec, *attr = NULL;
4598 	int rem, err = -EOPNOTSUPP;
4599 	u16 flags = 0;
4600 	bool have_flags = false;
4601 
4602 	if (nlmsg_len(nlh) < sizeof(*ifm))
4603 		return -EINVAL;
4604 
4605 	ifm = nlmsg_data(nlh);
4606 	if (ifm->ifi_family != AF_BRIDGE)
4607 		return -EPFNOSUPPORT;
4608 
4609 	dev = __dev_get_by_index(net, ifm->ifi_index);
4610 	if (!dev) {
4611 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4612 		return -ENODEV;
4613 	}
4614 
4615 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4616 	if (br_spec) {
4617 		nla_for_each_nested(attr, br_spec, rem) {
4618 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4619 				if (nla_len(attr) < sizeof(flags))
4620 					return -EINVAL;
4621 
4622 				have_flags = true;
4623 				flags = nla_get_u16(attr);
4624 				break;
4625 			}
4626 		}
4627 	}
4628 
4629 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4630 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4631 
4632 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4633 			err = -EOPNOTSUPP;
4634 			goto out;
4635 		}
4636 
4637 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4638 		if (err)
4639 			goto out;
4640 
4641 		flags &= ~BRIDGE_FLAGS_MASTER;
4642 	}
4643 
4644 	if ((flags & BRIDGE_FLAGS_SELF)) {
4645 		if (!dev->netdev_ops->ndo_bridge_dellink)
4646 			err = -EOPNOTSUPP;
4647 		else
4648 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4649 								  flags);
4650 
4651 		if (!err) {
4652 			flags &= ~BRIDGE_FLAGS_SELF;
4653 
4654 			/* Generate event to notify upper layer of bridge
4655 			 * change
4656 			 */
4657 			err = rtnl_bridge_notify(dev);
4658 		}
4659 	}
4660 
4661 	if (have_flags)
4662 		memcpy(nla_data(attr), &flags, sizeof(flags));
4663 out:
4664 	return err;
4665 }
4666 
4667 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4668 {
4669 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4670 	       (!idxattr || idxattr == attrid);
4671 }
4672 
4673 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4674 static int rtnl_get_offload_stats_attr_size(int attr_id)
4675 {
4676 	switch (attr_id) {
4677 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4678 		return sizeof(struct rtnl_link_stats64);
4679 	}
4680 
4681 	return 0;
4682 }
4683 
4684 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4685 				  int *prividx)
4686 {
4687 	struct nlattr *attr = NULL;
4688 	int attr_id, size;
4689 	void *attr_data;
4690 	int err;
4691 
4692 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4693 	      dev->netdev_ops->ndo_get_offload_stats))
4694 		return -ENODATA;
4695 
4696 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4697 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4698 		if (attr_id < *prividx)
4699 			continue;
4700 
4701 		size = rtnl_get_offload_stats_attr_size(attr_id);
4702 		if (!size)
4703 			continue;
4704 
4705 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4706 			continue;
4707 
4708 		attr = nla_reserve_64bit(skb, attr_id, size,
4709 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
4710 		if (!attr)
4711 			goto nla_put_failure;
4712 
4713 		attr_data = nla_data(attr);
4714 		memset(attr_data, 0, size);
4715 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4716 							     attr_data);
4717 		if (err)
4718 			goto get_offload_stats_failure;
4719 	}
4720 
4721 	if (!attr)
4722 		return -ENODATA;
4723 
4724 	*prividx = 0;
4725 	return 0;
4726 
4727 nla_put_failure:
4728 	err = -EMSGSIZE;
4729 get_offload_stats_failure:
4730 	*prividx = attr_id;
4731 	return err;
4732 }
4733 
4734 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4735 {
4736 	int nla_size = 0;
4737 	int attr_id;
4738 	int size;
4739 
4740 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4741 	      dev->netdev_ops->ndo_get_offload_stats))
4742 		return 0;
4743 
4744 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4745 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4746 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4747 			continue;
4748 		size = rtnl_get_offload_stats_attr_size(attr_id);
4749 		nla_size += nla_total_size_64bit(size);
4750 	}
4751 
4752 	if (nla_size != 0)
4753 		nla_size += nla_total_size(0);
4754 
4755 	return nla_size;
4756 }
4757 
4758 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4759 			       int type, u32 pid, u32 seq, u32 change,
4760 			       unsigned int flags, unsigned int filter_mask,
4761 			       int *idxattr, int *prividx)
4762 {
4763 	struct if_stats_msg *ifsm;
4764 	struct nlmsghdr *nlh;
4765 	struct nlattr *attr;
4766 	int s_prividx = *prividx;
4767 	int err;
4768 
4769 	ASSERT_RTNL();
4770 
4771 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4772 	if (!nlh)
4773 		return -EMSGSIZE;
4774 
4775 	ifsm = nlmsg_data(nlh);
4776 	ifsm->family = PF_UNSPEC;
4777 	ifsm->pad1 = 0;
4778 	ifsm->pad2 = 0;
4779 	ifsm->ifindex = dev->ifindex;
4780 	ifsm->filter_mask = filter_mask;
4781 
4782 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4783 		struct rtnl_link_stats64 *sp;
4784 
4785 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4786 					 sizeof(struct rtnl_link_stats64),
4787 					 IFLA_STATS_UNSPEC);
4788 		if (!attr)
4789 			goto nla_put_failure;
4790 
4791 		sp = nla_data(attr);
4792 		dev_get_stats(dev, sp);
4793 	}
4794 
4795 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4796 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4797 
4798 		if (ops && ops->fill_linkxstats) {
4799 			*idxattr = IFLA_STATS_LINK_XSTATS;
4800 			attr = nla_nest_start_noflag(skb,
4801 						     IFLA_STATS_LINK_XSTATS);
4802 			if (!attr)
4803 				goto nla_put_failure;
4804 
4805 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4806 			nla_nest_end(skb, attr);
4807 			if (err)
4808 				goto nla_put_failure;
4809 			*idxattr = 0;
4810 		}
4811 	}
4812 
4813 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4814 			     *idxattr)) {
4815 		const struct rtnl_link_ops *ops = NULL;
4816 		const struct net_device *master;
4817 
4818 		master = netdev_master_upper_dev_get(dev);
4819 		if (master)
4820 			ops = master->rtnl_link_ops;
4821 		if (ops && ops->fill_linkxstats) {
4822 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4823 			attr = nla_nest_start_noflag(skb,
4824 						     IFLA_STATS_LINK_XSTATS_SLAVE);
4825 			if (!attr)
4826 				goto nla_put_failure;
4827 
4828 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4829 			nla_nest_end(skb, attr);
4830 			if (err)
4831 				goto nla_put_failure;
4832 			*idxattr = 0;
4833 		}
4834 	}
4835 
4836 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4837 			     *idxattr)) {
4838 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4839 		attr = nla_nest_start_noflag(skb,
4840 					     IFLA_STATS_LINK_OFFLOAD_XSTATS);
4841 		if (!attr)
4842 			goto nla_put_failure;
4843 
4844 		err = rtnl_get_offload_stats(skb, dev, prividx);
4845 		if (err == -ENODATA)
4846 			nla_nest_cancel(skb, attr);
4847 		else
4848 			nla_nest_end(skb, attr);
4849 
4850 		if (err && err != -ENODATA)
4851 			goto nla_put_failure;
4852 		*idxattr = 0;
4853 	}
4854 
4855 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4856 		struct rtnl_af_ops *af_ops;
4857 
4858 		*idxattr = IFLA_STATS_AF_SPEC;
4859 		attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
4860 		if (!attr)
4861 			goto nla_put_failure;
4862 
4863 		rcu_read_lock();
4864 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4865 			if (af_ops->fill_stats_af) {
4866 				struct nlattr *af;
4867 				int err;
4868 
4869 				af = nla_nest_start_noflag(skb,
4870 							   af_ops->family);
4871 				if (!af) {
4872 					rcu_read_unlock();
4873 					goto nla_put_failure;
4874 				}
4875 				err = af_ops->fill_stats_af(skb, dev);
4876 
4877 				if (err == -ENODATA) {
4878 					nla_nest_cancel(skb, af);
4879 				} else if (err < 0) {
4880 					rcu_read_unlock();
4881 					goto nla_put_failure;
4882 				}
4883 
4884 				nla_nest_end(skb, af);
4885 			}
4886 		}
4887 		rcu_read_unlock();
4888 
4889 		nla_nest_end(skb, attr);
4890 
4891 		*idxattr = 0;
4892 	}
4893 
4894 	nlmsg_end(skb, nlh);
4895 
4896 	return 0;
4897 
4898 nla_put_failure:
4899 	/* not a multi message or no progress mean a real error */
4900 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4901 		nlmsg_cancel(skb, nlh);
4902 	else
4903 		nlmsg_end(skb, nlh);
4904 
4905 	return -EMSGSIZE;
4906 }
4907 
4908 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4909 				  u32 filter_mask)
4910 {
4911 	size_t size = 0;
4912 
4913 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4914 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4915 
4916 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4917 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4918 		int attr = IFLA_STATS_LINK_XSTATS;
4919 
4920 		if (ops && ops->get_linkxstats_size) {
4921 			size += nla_total_size(ops->get_linkxstats_size(dev,
4922 									attr));
4923 			/* for IFLA_STATS_LINK_XSTATS */
4924 			size += nla_total_size(0);
4925 		}
4926 	}
4927 
4928 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4929 		struct net_device *_dev = (struct net_device *)dev;
4930 		const struct rtnl_link_ops *ops = NULL;
4931 		const struct net_device *master;
4932 
4933 		/* netdev_master_upper_dev_get can't take const */
4934 		master = netdev_master_upper_dev_get(_dev);
4935 		if (master)
4936 			ops = master->rtnl_link_ops;
4937 		if (ops && ops->get_linkxstats_size) {
4938 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4939 
4940 			size += nla_total_size(ops->get_linkxstats_size(dev,
4941 									attr));
4942 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
4943 			size += nla_total_size(0);
4944 		}
4945 	}
4946 
4947 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4948 		size += rtnl_get_offload_stats_size(dev);
4949 
4950 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4951 		struct rtnl_af_ops *af_ops;
4952 
4953 		/* for IFLA_STATS_AF_SPEC */
4954 		size += nla_total_size(0);
4955 
4956 		rcu_read_lock();
4957 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4958 			if (af_ops->get_stats_af_size) {
4959 				size += nla_total_size(
4960 					af_ops->get_stats_af_size(dev));
4961 
4962 				/* for AF_* */
4963 				size += nla_total_size(0);
4964 			}
4965 		}
4966 		rcu_read_unlock();
4967 	}
4968 
4969 	return size;
4970 }
4971 
4972 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
4973 				bool is_dump, struct netlink_ext_ack *extack)
4974 {
4975 	struct if_stats_msg *ifsm;
4976 
4977 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
4978 		NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
4979 		return -EINVAL;
4980 	}
4981 
4982 	if (!strict_check)
4983 		return 0;
4984 
4985 	ifsm = nlmsg_data(nlh);
4986 
4987 	/* only requests using strict checks can pass data to influence
4988 	 * the dump. The legacy exception is filter_mask.
4989 	 */
4990 	if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
4991 		NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
4992 		return -EINVAL;
4993 	}
4994 	if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
4995 		NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
4996 		return -EINVAL;
4997 	}
4998 	if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
4999 		NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5000 		return -EINVAL;
5001 	}
5002 
5003 	return 0;
5004 }
5005 
5006 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5007 			  struct netlink_ext_ack *extack)
5008 {
5009 	struct net *net = sock_net(skb->sk);
5010 	struct net_device *dev = NULL;
5011 	int idxattr = 0, prividx = 0;
5012 	struct if_stats_msg *ifsm;
5013 	struct sk_buff *nskb;
5014 	u32 filter_mask;
5015 	int err;
5016 
5017 	err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5018 				   false, extack);
5019 	if (err)
5020 		return err;
5021 
5022 	ifsm = nlmsg_data(nlh);
5023 	if (ifsm->ifindex > 0)
5024 		dev = __dev_get_by_index(net, ifsm->ifindex);
5025 	else
5026 		return -EINVAL;
5027 
5028 	if (!dev)
5029 		return -ENODEV;
5030 
5031 	filter_mask = ifsm->filter_mask;
5032 	if (!filter_mask)
5033 		return -EINVAL;
5034 
5035 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5036 	if (!nskb)
5037 		return -ENOBUFS;
5038 
5039 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5040 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5041 				  0, filter_mask, &idxattr, &prividx);
5042 	if (err < 0) {
5043 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5044 		WARN_ON(err == -EMSGSIZE);
5045 		kfree_skb(nskb);
5046 	} else {
5047 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5048 	}
5049 
5050 	return err;
5051 }
5052 
5053 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5054 {
5055 	struct netlink_ext_ack *extack = cb->extack;
5056 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
5057 	struct net *net = sock_net(skb->sk);
5058 	unsigned int flags = NLM_F_MULTI;
5059 	struct if_stats_msg *ifsm;
5060 	struct hlist_head *head;
5061 	struct net_device *dev;
5062 	u32 filter_mask = 0;
5063 	int idx = 0;
5064 
5065 	s_h = cb->args[0];
5066 	s_idx = cb->args[1];
5067 	s_idxattr = cb->args[2];
5068 	s_prividx = cb->args[3];
5069 
5070 	cb->seq = net->dev_base_seq;
5071 
5072 	err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5073 	if (err)
5074 		return err;
5075 
5076 	ifsm = nlmsg_data(cb->nlh);
5077 	filter_mask = ifsm->filter_mask;
5078 	if (!filter_mask) {
5079 		NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5080 		return -EINVAL;
5081 	}
5082 
5083 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5084 		idx = 0;
5085 		head = &net->dev_index_head[h];
5086 		hlist_for_each_entry(dev, head, index_hlist) {
5087 			if (idx < s_idx)
5088 				goto cont;
5089 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5090 						  NETLINK_CB(cb->skb).portid,
5091 						  cb->nlh->nlmsg_seq, 0,
5092 						  flags, filter_mask,
5093 						  &s_idxattr, &s_prividx);
5094 			/* If we ran out of room on the first message,
5095 			 * we're in trouble
5096 			 */
5097 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5098 
5099 			if (err < 0)
5100 				goto out;
5101 			s_prividx = 0;
5102 			s_idxattr = 0;
5103 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5104 cont:
5105 			idx++;
5106 		}
5107 	}
5108 out:
5109 	cb->args[3] = s_prividx;
5110 	cb->args[2] = s_idxattr;
5111 	cb->args[1] = idx;
5112 	cb->args[0] = h;
5113 
5114 	return skb->len;
5115 }
5116 
5117 /* Process one rtnetlink message. */
5118 
5119 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5120 			     struct netlink_ext_ack *extack)
5121 {
5122 	struct net *net = sock_net(skb->sk);
5123 	struct rtnl_link *link;
5124 	struct module *owner;
5125 	int err = -EOPNOTSUPP;
5126 	rtnl_doit_func doit;
5127 	unsigned int flags;
5128 	int kind;
5129 	int family;
5130 	int type;
5131 
5132 	type = nlh->nlmsg_type;
5133 	if (type > RTM_MAX)
5134 		return -EOPNOTSUPP;
5135 
5136 	type -= RTM_BASE;
5137 
5138 	/* All the messages must have at least 1 byte length */
5139 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5140 		return 0;
5141 
5142 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5143 	kind = type&3;
5144 
5145 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5146 		return -EPERM;
5147 
5148 	rcu_read_lock();
5149 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5150 		struct sock *rtnl;
5151 		rtnl_dumpit_func dumpit;
5152 		u16 min_dump_alloc = 0;
5153 
5154 		link = rtnl_get_link(family, type);
5155 		if (!link || !link->dumpit) {
5156 			family = PF_UNSPEC;
5157 			link = rtnl_get_link(family, type);
5158 			if (!link || !link->dumpit)
5159 				goto err_unlock;
5160 		}
5161 		owner = link->owner;
5162 		dumpit = link->dumpit;
5163 
5164 		if (type == RTM_GETLINK - RTM_BASE)
5165 			min_dump_alloc = rtnl_calcit(skb, nlh);
5166 
5167 		err = 0;
5168 		/* need to do this before rcu_read_unlock() */
5169 		if (!try_module_get(owner))
5170 			err = -EPROTONOSUPPORT;
5171 
5172 		rcu_read_unlock();
5173 
5174 		rtnl = net->rtnl;
5175 		if (err == 0) {
5176 			struct netlink_dump_control c = {
5177 				.dump		= dumpit,
5178 				.min_dump_alloc	= min_dump_alloc,
5179 				.module		= owner,
5180 			};
5181 			err = netlink_dump_start(rtnl, skb, nlh, &c);
5182 			/* netlink_dump_start() will keep a reference on
5183 			 * module if dump is still in progress.
5184 			 */
5185 			module_put(owner);
5186 		}
5187 		return err;
5188 	}
5189 
5190 	link = rtnl_get_link(family, type);
5191 	if (!link || !link->doit) {
5192 		family = PF_UNSPEC;
5193 		link = rtnl_get_link(PF_UNSPEC, type);
5194 		if (!link || !link->doit)
5195 			goto out_unlock;
5196 	}
5197 
5198 	owner = link->owner;
5199 	if (!try_module_get(owner)) {
5200 		err = -EPROTONOSUPPORT;
5201 		goto out_unlock;
5202 	}
5203 
5204 	flags = link->flags;
5205 	if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5206 		doit = link->doit;
5207 		rcu_read_unlock();
5208 		if (doit)
5209 			err = doit(skb, nlh, extack);
5210 		module_put(owner);
5211 		return err;
5212 	}
5213 	rcu_read_unlock();
5214 
5215 	rtnl_lock();
5216 	link = rtnl_get_link(family, type);
5217 	if (link && link->doit)
5218 		err = link->doit(skb, nlh, extack);
5219 	rtnl_unlock();
5220 
5221 	module_put(owner);
5222 
5223 	return err;
5224 
5225 out_unlock:
5226 	rcu_read_unlock();
5227 	return err;
5228 
5229 err_unlock:
5230 	rcu_read_unlock();
5231 	return -EOPNOTSUPP;
5232 }
5233 
5234 static void rtnetlink_rcv(struct sk_buff *skb)
5235 {
5236 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5237 }
5238 
5239 static int rtnetlink_bind(struct net *net, int group)
5240 {
5241 	switch (group) {
5242 	case RTNLGRP_IPV4_MROUTE_R:
5243 	case RTNLGRP_IPV6_MROUTE_R:
5244 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5245 			return -EPERM;
5246 		break;
5247 	}
5248 	return 0;
5249 }
5250 
5251 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5252 {
5253 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5254 
5255 	switch (event) {
5256 	case NETDEV_REBOOT:
5257 	case NETDEV_CHANGEMTU:
5258 	case NETDEV_CHANGEADDR:
5259 	case NETDEV_CHANGENAME:
5260 	case NETDEV_FEAT_CHANGE:
5261 	case NETDEV_BONDING_FAILOVER:
5262 	case NETDEV_POST_TYPE_CHANGE:
5263 	case NETDEV_NOTIFY_PEERS:
5264 	case NETDEV_CHANGEUPPER:
5265 	case NETDEV_RESEND_IGMP:
5266 	case NETDEV_CHANGEINFODATA:
5267 	case NETDEV_CHANGELOWERSTATE:
5268 	case NETDEV_CHANGE_TX_QUEUE_LEN:
5269 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5270 				   GFP_KERNEL, NULL, 0);
5271 		break;
5272 	default:
5273 		break;
5274 	}
5275 	return NOTIFY_DONE;
5276 }
5277 
5278 static struct notifier_block rtnetlink_dev_notifier = {
5279 	.notifier_call	= rtnetlink_event,
5280 };
5281 
5282 
5283 static int __net_init rtnetlink_net_init(struct net *net)
5284 {
5285 	struct sock *sk;
5286 	struct netlink_kernel_cfg cfg = {
5287 		.groups		= RTNLGRP_MAX,
5288 		.input		= rtnetlink_rcv,
5289 		.cb_mutex	= &rtnl_mutex,
5290 		.flags		= NL_CFG_F_NONROOT_RECV,
5291 		.bind		= rtnetlink_bind,
5292 	};
5293 
5294 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5295 	if (!sk)
5296 		return -ENOMEM;
5297 	net->rtnl = sk;
5298 	return 0;
5299 }
5300 
5301 static void __net_exit rtnetlink_net_exit(struct net *net)
5302 {
5303 	netlink_kernel_release(net->rtnl);
5304 	net->rtnl = NULL;
5305 }
5306 
5307 static struct pernet_operations rtnetlink_net_ops = {
5308 	.init = rtnetlink_net_init,
5309 	.exit = rtnetlink_net_exit,
5310 };
5311 
5312 void __init rtnetlink_init(void)
5313 {
5314 	if (register_pernet_subsys(&rtnetlink_net_ops))
5315 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
5316 
5317 	register_netdevice_notifier(&rtnetlink_dev_notifier);
5318 
5319 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5320 		      rtnl_dump_ifinfo, 0);
5321 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5322 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5323 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5324 
5325 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5326 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5327 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5328 
5329 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5330 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5331 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5332 
5333 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5334 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5335 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5336 
5337 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5338 		      0);
5339 }
5340