xref: /openbmc/linux/net/core/rtnetlink.c (revision 08283d30)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Routing netlink socket interface: protocol independent part.
8  *
9  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10  *
11  *	Fixes:
12  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
13  */
14 
15 #include <linux/bitops.h>
16 #include <linux/errno.h>
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/kernel.h>
21 #include <linux/timer.h>
22 #include <linux/string.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/fcntl.h>
26 #include <linux/mm.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/capability.h>
30 #include <linux/skbuff.h>
31 #include <linux/init.h>
32 #include <linux/security.h>
33 #include <linux/mutex.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_vlan.h>
37 #include <linux/pci.h>
38 #include <linux/etherdevice.h>
39 #include <linux/bpf.h>
40 
41 #include <linux/uaccess.h>
42 
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <net/ip.h>
46 #include <net/protocol.h>
47 #include <net/arp.h>
48 #include <net/route.h>
49 #include <net/udp.h>
50 #include <net/tcp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 
57 #define RTNL_MAX_TYPE		50
58 #define RTNL_SLAVE_MAX_TYPE	36
59 
60 struct rtnl_link {
61 	rtnl_doit_func		doit;
62 	rtnl_dumpit_func	dumpit;
63 	struct module		*owner;
64 	unsigned int		flags;
65 	struct rcu_head		rcu;
66 };
67 
68 static DEFINE_MUTEX(rtnl_mutex);
69 
70 void rtnl_lock(void)
71 {
72 	mutex_lock(&rtnl_mutex);
73 }
74 EXPORT_SYMBOL(rtnl_lock);
75 
76 int rtnl_lock_killable(void)
77 {
78 	return mutex_lock_killable(&rtnl_mutex);
79 }
80 EXPORT_SYMBOL(rtnl_lock_killable);
81 
82 static struct sk_buff *defer_kfree_skb_list;
83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
84 {
85 	if (head && tail) {
86 		tail->next = defer_kfree_skb_list;
87 		defer_kfree_skb_list = head;
88 	}
89 }
90 EXPORT_SYMBOL(rtnl_kfree_skbs);
91 
92 void __rtnl_unlock(void)
93 {
94 	struct sk_buff *head = defer_kfree_skb_list;
95 
96 	defer_kfree_skb_list = NULL;
97 
98 	mutex_unlock(&rtnl_mutex);
99 
100 	while (head) {
101 		struct sk_buff *next = head->next;
102 
103 		kfree_skb(head);
104 		cond_resched();
105 		head = next;
106 	}
107 }
108 
109 void rtnl_unlock(void)
110 {
111 	/* This fellow will unlock it for us. */
112 	netdev_run_todo();
113 }
114 EXPORT_SYMBOL(rtnl_unlock);
115 
116 int rtnl_trylock(void)
117 {
118 	return mutex_trylock(&rtnl_mutex);
119 }
120 EXPORT_SYMBOL(rtnl_trylock);
121 
122 int rtnl_is_locked(void)
123 {
124 	return mutex_is_locked(&rtnl_mutex);
125 }
126 EXPORT_SYMBOL(rtnl_is_locked);
127 
128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
129 {
130 	return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
131 }
132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
133 
134 #ifdef CONFIG_PROVE_LOCKING
135 bool lockdep_rtnl_is_held(void)
136 {
137 	return lockdep_is_held(&rtnl_mutex);
138 }
139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
141 
142 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
143 
144 static inline int rtm_msgindex(int msgtype)
145 {
146 	int msgindex = msgtype - RTM_BASE;
147 
148 	/*
149 	 * msgindex < 0 implies someone tried to register a netlink
150 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
151 	 * the message type has not been added to linux/rtnetlink.h
152 	 */
153 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
154 
155 	return msgindex;
156 }
157 
158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
159 {
160 	struct rtnl_link **tab;
161 
162 	if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
163 		protocol = PF_UNSPEC;
164 
165 	tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
166 	if (!tab)
167 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
168 
169 	return tab[msgtype];
170 }
171 
172 static int rtnl_register_internal(struct module *owner,
173 				  int protocol, int msgtype,
174 				  rtnl_doit_func doit, rtnl_dumpit_func dumpit,
175 				  unsigned int flags)
176 {
177 	struct rtnl_link *link, *old;
178 	struct rtnl_link __rcu **tab;
179 	int msgindex;
180 	int ret = -ENOBUFS;
181 
182 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
183 	msgindex = rtm_msgindex(msgtype);
184 
185 	rtnl_lock();
186 	tab = rtnl_msg_handlers[protocol];
187 	if (tab == NULL) {
188 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
189 		if (!tab)
190 			goto unlock;
191 
192 		/* ensures we see the 0 stores */
193 		rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
194 	}
195 
196 	old = rtnl_dereference(tab[msgindex]);
197 	if (old) {
198 		link = kmemdup(old, sizeof(*old), GFP_KERNEL);
199 		if (!link)
200 			goto unlock;
201 	} else {
202 		link = kzalloc(sizeof(*link), GFP_KERNEL);
203 		if (!link)
204 			goto unlock;
205 	}
206 
207 	WARN_ON(link->owner && link->owner != owner);
208 	link->owner = owner;
209 
210 	WARN_ON(doit && link->doit && link->doit != doit);
211 	if (doit)
212 		link->doit = doit;
213 	WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
214 	if (dumpit)
215 		link->dumpit = dumpit;
216 
217 	link->flags |= flags;
218 
219 	/* publish protocol:msgtype */
220 	rcu_assign_pointer(tab[msgindex], link);
221 	ret = 0;
222 	if (old)
223 		kfree_rcu(old, rcu);
224 unlock:
225 	rtnl_unlock();
226 	return ret;
227 }
228 
229 /**
230  * rtnl_register_module - Register a rtnetlink message type
231  *
232  * @owner: module registering the hook (THIS_MODULE)
233  * @protocol: Protocol family or PF_UNSPEC
234  * @msgtype: rtnetlink message type
235  * @doit: Function pointer called for each request message
236  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
237  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
238  *
239  * Like rtnl_register, but for use by removable modules.
240  */
241 int rtnl_register_module(struct module *owner,
242 			 int protocol, int msgtype,
243 			 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
244 			 unsigned int flags)
245 {
246 	return rtnl_register_internal(owner, protocol, msgtype,
247 				      doit, dumpit, flags);
248 }
249 EXPORT_SYMBOL_GPL(rtnl_register_module);
250 
251 /**
252  * rtnl_register - Register a rtnetlink message type
253  * @protocol: Protocol family or PF_UNSPEC
254  * @msgtype: rtnetlink message type
255  * @doit: Function pointer called for each request message
256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
257  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
258  *
259  * Registers the specified function pointers (at least one of them has
260  * to be non-NULL) to be called whenever a request message for the
261  * specified protocol family and message type is received.
262  *
263  * The special protocol family PF_UNSPEC may be used to define fallback
264  * function pointers for the case when no entry for the specific protocol
265  * family exists.
266  */
267 void rtnl_register(int protocol, int msgtype,
268 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
269 		   unsigned int flags)
270 {
271 	int err;
272 
273 	err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
274 				     flags);
275 	if (err)
276 		pr_err("Unable to register rtnetlink message handler, "
277 		       "protocol = %d, message type = %d\n", protocol, msgtype);
278 }
279 
280 /**
281  * rtnl_unregister - Unregister a rtnetlink message type
282  * @protocol: Protocol family or PF_UNSPEC
283  * @msgtype: rtnetlink message type
284  *
285  * Returns 0 on success or a negative error code.
286  */
287 int rtnl_unregister(int protocol, int msgtype)
288 {
289 	struct rtnl_link **tab, *link;
290 	int msgindex;
291 
292 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
293 	msgindex = rtm_msgindex(msgtype);
294 
295 	rtnl_lock();
296 	tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
297 	if (!tab) {
298 		rtnl_unlock();
299 		return -ENOENT;
300 	}
301 
302 	link = tab[msgindex];
303 	rcu_assign_pointer(tab[msgindex], NULL);
304 	rtnl_unlock();
305 
306 	kfree_rcu(link, rcu);
307 
308 	return 0;
309 }
310 EXPORT_SYMBOL_GPL(rtnl_unregister);
311 
312 /**
313  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
314  * @protocol : Protocol family or PF_UNSPEC
315  *
316  * Identical to calling rtnl_unregster() for all registered message types
317  * of a certain protocol family.
318  */
319 void rtnl_unregister_all(int protocol)
320 {
321 	struct rtnl_link **tab, *link;
322 	int msgindex;
323 
324 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
325 
326 	rtnl_lock();
327 	tab = rtnl_msg_handlers[protocol];
328 	if (!tab) {
329 		rtnl_unlock();
330 		return;
331 	}
332 	RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
333 	for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
334 		link = tab[msgindex];
335 		if (!link)
336 			continue;
337 
338 		rcu_assign_pointer(tab[msgindex], NULL);
339 		kfree_rcu(link, rcu);
340 	}
341 	rtnl_unlock();
342 
343 	synchronize_net();
344 
345 	kfree(tab);
346 }
347 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
348 
349 static LIST_HEAD(link_ops);
350 
351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
352 {
353 	const struct rtnl_link_ops *ops;
354 
355 	list_for_each_entry(ops, &link_ops, list) {
356 		if (!strcmp(ops->kind, kind))
357 			return ops;
358 	}
359 	return NULL;
360 }
361 
362 /**
363  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
364  * @ops: struct rtnl_link_ops * to register
365  *
366  * The caller must hold the rtnl_mutex. This function should be used
367  * by drivers that create devices during module initialization. It
368  * must be called before registering the devices.
369  *
370  * Returns 0 on success or a negative error code.
371  */
372 int __rtnl_link_register(struct rtnl_link_ops *ops)
373 {
374 	if (rtnl_link_ops_get(ops->kind))
375 		return -EEXIST;
376 
377 	/* The check for setup is here because if ops
378 	 * does not have that filled up, it is not possible
379 	 * to use the ops for creating device. So do not
380 	 * fill up dellink as well. That disables rtnl_dellink.
381 	 */
382 	if (ops->setup && !ops->dellink)
383 		ops->dellink = unregister_netdevice_queue;
384 
385 	list_add_tail(&ops->list, &link_ops);
386 	return 0;
387 }
388 EXPORT_SYMBOL_GPL(__rtnl_link_register);
389 
390 /**
391  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
392  * @ops: struct rtnl_link_ops * to register
393  *
394  * Returns 0 on success or a negative error code.
395  */
396 int rtnl_link_register(struct rtnl_link_ops *ops)
397 {
398 	int err;
399 
400 	/* Sanity-check max sizes to avoid stack buffer overflow. */
401 	if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
402 		    ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
403 		return -EINVAL;
404 
405 	rtnl_lock();
406 	err = __rtnl_link_register(ops);
407 	rtnl_unlock();
408 	return err;
409 }
410 EXPORT_SYMBOL_GPL(rtnl_link_register);
411 
412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
413 {
414 	struct net_device *dev;
415 	LIST_HEAD(list_kill);
416 
417 	for_each_netdev(net, dev) {
418 		if (dev->rtnl_link_ops == ops)
419 			ops->dellink(dev, &list_kill);
420 	}
421 	unregister_netdevice_many(&list_kill);
422 }
423 
424 /**
425  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
426  * @ops: struct rtnl_link_ops * to unregister
427  *
428  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
429  * integrity (hold pernet_ops_rwsem for writing to close the race
430  * with setup_net() and cleanup_net()).
431  */
432 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
433 {
434 	struct net *net;
435 
436 	for_each_net(net) {
437 		__rtnl_kill_links(net, ops);
438 	}
439 	list_del(&ops->list);
440 }
441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
442 
443 /* Return with the rtnl_lock held when there are no network
444  * devices unregistering in any network namespace.
445  */
446 static void rtnl_lock_unregistering_all(void)
447 {
448 	struct net *net;
449 	bool unregistering;
450 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
451 
452 	add_wait_queue(&netdev_unregistering_wq, &wait);
453 	for (;;) {
454 		unregistering = false;
455 		rtnl_lock();
456 		/* We held write locked pernet_ops_rwsem, and parallel
457 		 * setup_net() and cleanup_net() are not possible.
458 		 */
459 		for_each_net(net) {
460 			if (net->dev_unreg_count > 0) {
461 				unregistering = true;
462 				break;
463 			}
464 		}
465 		if (!unregistering)
466 			break;
467 		__rtnl_unlock();
468 
469 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
470 	}
471 	remove_wait_queue(&netdev_unregistering_wq, &wait);
472 }
473 
474 /**
475  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
476  * @ops: struct rtnl_link_ops * to unregister
477  */
478 void rtnl_link_unregister(struct rtnl_link_ops *ops)
479 {
480 	/* Close the race with setup_net() and cleanup_net() */
481 	down_write(&pernet_ops_rwsem);
482 	rtnl_lock_unregistering_all();
483 	__rtnl_link_unregister(ops);
484 	rtnl_unlock();
485 	up_write(&pernet_ops_rwsem);
486 }
487 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
488 
489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
490 {
491 	struct net_device *master_dev;
492 	const struct rtnl_link_ops *ops;
493 	size_t size = 0;
494 
495 	rcu_read_lock();
496 
497 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
498 	if (!master_dev)
499 		goto out;
500 
501 	ops = master_dev->rtnl_link_ops;
502 	if (!ops || !ops->get_slave_size)
503 		goto out;
504 	/* IFLA_INFO_SLAVE_DATA + nested data */
505 	size = nla_total_size(sizeof(struct nlattr)) +
506 	       ops->get_slave_size(master_dev, dev);
507 
508 out:
509 	rcu_read_unlock();
510 	return size;
511 }
512 
513 static size_t rtnl_link_get_size(const struct net_device *dev)
514 {
515 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
516 	size_t size;
517 
518 	if (!ops)
519 		return 0;
520 
521 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
522 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
523 
524 	if (ops->get_size)
525 		/* IFLA_INFO_DATA + nested data */
526 		size += nla_total_size(sizeof(struct nlattr)) +
527 			ops->get_size(dev);
528 
529 	if (ops->get_xstats_size)
530 		/* IFLA_INFO_XSTATS */
531 		size += nla_total_size(ops->get_xstats_size(dev));
532 
533 	size += rtnl_link_get_slave_info_data_size(dev);
534 
535 	return size;
536 }
537 
538 static LIST_HEAD(rtnl_af_ops);
539 
540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
541 {
542 	const struct rtnl_af_ops *ops;
543 
544 	list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
545 		if (ops->family == family)
546 			return ops;
547 	}
548 
549 	return NULL;
550 }
551 
552 /**
553  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
554  * @ops: struct rtnl_af_ops * to register
555  *
556  * Returns 0 on success or a negative error code.
557  */
558 void rtnl_af_register(struct rtnl_af_ops *ops)
559 {
560 	rtnl_lock();
561 	list_add_tail_rcu(&ops->list, &rtnl_af_ops);
562 	rtnl_unlock();
563 }
564 EXPORT_SYMBOL_GPL(rtnl_af_register);
565 
566 /**
567  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
568  * @ops: struct rtnl_af_ops * to unregister
569  */
570 void rtnl_af_unregister(struct rtnl_af_ops *ops)
571 {
572 	rtnl_lock();
573 	list_del_rcu(&ops->list);
574 	rtnl_unlock();
575 
576 	synchronize_rcu();
577 }
578 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
579 
580 static size_t rtnl_link_get_af_size(const struct net_device *dev,
581 				    u32 ext_filter_mask)
582 {
583 	struct rtnl_af_ops *af_ops;
584 	size_t size;
585 
586 	/* IFLA_AF_SPEC */
587 	size = nla_total_size(sizeof(struct nlattr));
588 
589 	rcu_read_lock();
590 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
591 		if (af_ops->get_link_af_size) {
592 			/* AF_* + nested data */
593 			size += nla_total_size(sizeof(struct nlattr)) +
594 				af_ops->get_link_af_size(dev, ext_filter_mask);
595 		}
596 	}
597 	rcu_read_unlock();
598 
599 	return size;
600 }
601 
602 static bool rtnl_have_link_slave_info(const struct net_device *dev)
603 {
604 	struct net_device *master_dev;
605 	bool ret = false;
606 
607 	rcu_read_lock();
608 
609 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
610 	if (master_dev && master_dev->rtnl_link_ops)
611 		ret = true;
612 	rcu_read_unlock();
613 	return ret;
614 }
615 
616 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
617 				     const struct net_device *dev)
618 {
619 	struct net_device *master_dev;
620 	const struct rtnl_link_ops *ops;
621 	struct nlattr *slave_data;
622 	int err;
623 
624 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
625 	if (!master_dev)
626 		return 0;
627 	ops = master_dev->rtnl_link_ops;
628 	if (!ops)
629 		return 0;
630 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
631 		return -EMSGSIZE;
632 	if (ops->fill_slave_info) {
633 		slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
634 		if (!slave_data)
635 			return -EMSGSIZE;
636 		err = ops->fill_slave_info(skb, master_dev, dev);
637 		if (err < 0)
638 			goto err_cancel_slave_data;
639 		nla_nest_end(skb, slave_data);
640 	}
641 	return 0;
642 
643 err_cancel_slave_data:
644 	nla_nest_cancel(skb, slave_data);
645 	return err;
646 }
647 
648 static int rtnl_link_info_fill(struct sk_buff *skb,
649 			       const struct net_device *dev)
650 {
651 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
652 	struct nlattr *data;
653 	int err;
654 
655 	if (!ops)
656 		return 0;
657 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
658 		return -EMSGSIZE;
659 	if (ops->fill_xstats) {
660 		err = ops->fill_xstats(skb, dev);
661 		if (err < 0)
662 			return err;
663 	}
664 	if (ops->fill_info) {
665 		data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
666 		if (data == NULL)
667 			return -EMSGSIZE;
668 		err = ops->fill_info(skb, dev);
669 		if (err < 0)
670 			goto err_cancel_data;
671 		nla_nest_end(skb, data);
672 	}
673 	return 0;
674 
675 err_cancel_data:
676 	nla_nest_cancel(skb, data);
677 	return err;
678 }
679 
680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
681 {
682 	struct nlattr *linkinfo;
683 	int err = -EMSGSIZE;
684 
685 	linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
686 	if (linkinfo == NULL)
687 		goto out;
688 
689 	err = rtnl_link_info_fill(skb, dev);
690 	if (err < 0)
691 		goto err_cancel_link;
692 
693 	err = rtnl_link_slave_info_fill(skb, dev);
694 	if (err < 0)
695 		goto err_cancel_link;
696 
697 	nla_nest_end(skb, linkinfo);
698 	return 0;
699 
700 err_cancel_link:
701 	nla_nest_cancel(skb, linkinfo);
702 out:
703 	return err;
704 }
705 
706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
707 {
708 	struct sock *rtnl = net->rtnl;
709 	int err = 0;
710 
711 	NETLINK_CB(skb).dst_group = group;
712 	if (echo)
713 		refcount_inc(&skb->users);
714 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
715 	if (echo)
716 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
717 	return err;
718 }
719 
720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
721 {
722 	struct sock *rtnl = net->rtnl;
723 
724 	return nlmsg_unicast(rtnl, skb, pid);
725 }
726 EXPORT_SYMBOL(rtnl_unicast);
727 
728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
729 		 struct nlmsghdr *nlh, gfp_t flags)
730 {
731 	struct sock *rtnl = net->rtnl;
732 	int report = 0;
733 
734 	if (nlh)
735 		report = nlmsg_report(nlh);
736 
737 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
738 }
739 EXPORT_SYMBOL(rtnl_notify);
740 
741 void rtnl_set_sk_err(struct net *net, u32 group, int error)
742 {
743 	struct sock *rtnl = net->rtnl;
744 
745 	netlink_set_err(rtnl, 0, group, error);
746 }
747 EXPORT_SYMBOL(rtnl_set_sk_err);
748 
749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
750 {
751 	struct nlattr *mx;
752 	int i, valid = 0;
753 
754 	/* nothing is dumped for dst_default_metrics, so just skip the loop */
755 	if (metrics == dst_default_metrics.metrics)
756 		return 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_broadcast)) +
916 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
917 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
918 			 nla_total_size(MAX_VLAN_LIST_LEN *
919 					sizeof(struct ifla_vf_vlan_info)) +
920 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
921 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
922 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
923 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
924 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
925 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
926 			 /* IFLA_VF_STATS_RX_PACKETS */
927 			 nla_total_size_64bit(sizeof(__u64)) +
928 			 /* IFLA_VF_STATS_TX_PACKETS */
929 			 nla_total_size_64bit(sizeof(__u64)) +
930 			 /* IFLA_VF_STATS_RX_BYTES */
931 			 nla_total_size_64bit(sizeof(__u64)) +
932 			 /* IFLA_VF_STATS_TX_BYTES */
933 			 nla_total_size_64bit(sizeof(__u64)) +
934 			 /* IFLA_VF_STATS_BROADCAST */
935 			 nla_total_size_64bit(sizeof(__u64)) +
936 			 /* IFLA_VF_STATS_MULTICAST */
937 			 nla_total_size_64bit(sizeof(__u64)) +
938 			 /* IFLA_VF_STATS_RX_DROPPED */
939 			 nla_total_size_64bit(sizeof(__u64)) +
940 			 /* IFLA_VF_STATS_TX_DROPPED */
941 			 nla_total_size_64bit(sizeof(__u64)) +
942 			 nla_total_size(sizeof(struct ifla_vf_trust)));
943 		return size;
944 	} else
945 		return 0;
946 }
947 
948 static size_t rtnl_port_size(const struct net_device *dev,
949 			     u32 ext_filter_mask)
950 {
951 	size_t port_size = nla_total_size(4)		/* PORT_VF */
952 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
953 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
954 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
955 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
956 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
957 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
958 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
959 		+ port_size;
960 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
961 		+ port_size;
962 
963 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
964 	    !(ext_filter_mask & RTEXT_FILTER_VF))
965 		return 0;
966 	if (dev_num_vf(dev->dev.parent))
967 		return port_self_size + vf_ports_size +
968 			vf_port_size * dev_num_vf(dev->dev.parent);
969 	else
970 		return port_self_size;
971 }
972 
973 static size_t rtnl_xdp_size(void)
974 {
975 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
976 			  nla_total_size(1) +	/* XDP_ATTACHED */
977 			  nla_total_size(4) +	/* XDP_PROG_ID (or 1st mode) */
978 			  nla_total_size(4);	/* XDP_<mode>_PROG_ID */
979 
980 	return xdp_size;
981 }
982 
983 static noinline size_t if_nlmsg_size(const struct net_device *dev,
984 				     u32 ext_filter_mask)
985 {
986 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
987 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
988 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
989 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
990 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
991 	       + nla_total_size(sizeof(struct rtnl_link_stats))
992 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
993 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
994 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
995 	       + nla_total_size(4) /* IFLA_TXQLEN */
996 	       + nla_total_size(4) /* IFLA_WEIGHT */
997 	       + nla_total_size(4) /* IFLA_MTU */
998 	       + nla_total_size(4) /* IFLA_LINK */
999 	       + nla_total_size(4) /* IFLA_MASTER */
1000 	       + nla_total_size(1) /* IFLA_CARRIER */
1001 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
1002 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1003 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1004 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1005 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1006 	       + nla_total_size(1) /* IFLA_OPERSTATE */
1007 	       + nla_total_size(1) /* IFLA_LINKMODE */
1008 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1009 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
1010 	       + nla_total_size(4) /* IFLA_GROUP */
1011 	       + nla_total_size(ext_filter_mask
1012 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1013 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1014 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1015 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1016 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1017 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1018 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1019 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1020 	       + rtnl_xdp_size() /* IFLA_XDP */
1021 	       + nla_total_size(4)  /* IFLA_EVENT */
1022 	       + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1023 	       + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1024 	       + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1025 	       + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1026 	       + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1027 	       + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1028 	       + nla_total_size(4)  /* IFLA_MIN_MTU */
1029 	       + nla_total_size(4)  /* IFLA_MAX_MTU */
1030 	       + 0;
1031 }
1032 
1033 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1034 {
1035 	struct nlattr *vf_ports;
1036 	struct nlattr *vf_port;
1037 	int vf;
1038 	int err;
1039 
1040 	vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1041 	if (!vf_ports)
1042 		return -EMSGSIZE;
1043 
1044 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1045 		vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1046 		if (!vf_port)
1047 			goto nla_put_failure;
1048 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1049 			goto nla_put_failure;
1050 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1051 		if (err == -EMSGSIZE)
1052 			goto nla_put_failure;
1053 		if (err) {
1054 			nla_nest_cancel(skb, vf_port);
1055 			continue;
1056 		}
1057 		nla_nest_end(skb, vf_port);
1058 	}
1059 
1060 	nla_nest_end(skb, vf_ports);
1061 
1062 	return 0;
1063 
1064 nla_put_failure:
1065 	nla_nest_cancel(skb, vf_ports);
1066 	return -EMSGSIZE;
1067 }
1068 
1069 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1070 {
1071 	struct nlattr *port_self;
1072 	int err;
1073 
1074 	port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1075 	if (!port_self)
1076 		return -EMSGSIZE;
1077 
1078 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1079 	if (err) {
1080 		nla_nest_cancel(skb, port_self);
1081 		return (err == -EMSGSIZE) ? err : 0;
1082 	}
1083 
1084 	nla_nest_end(skb, port_self);
1085 
1086 	return 0;
1087 }
1088 
1089 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1090 			  u32 ext_filter_mask)
1091 {
1092 	int err;
1093 
1094 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1095 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1096 		return 0;
1097 
1098 	err = rtnl_port_self_fill(skb, dev);
1099 	if (err)
1100 		return err;
1101 
1102 	if (dev_num_vf(dev->dev.parent)) {
1103 		err = rtnl_vf_ports_fill(skb, dev);
1104 		if (err)
1105 			return err;
1106 	}
1107 
1108 	return 0;
1109 }
1110 
1111 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1112 {
1113 	int err;
1114 	struct netdev_phys_item_id ppid;
1115 
1116 	err = dev_get_phys_port_id(dev, &ppid);
1117 	if (err) {
1118 		if (err == -EOPNOTSUPP)
1119 			return 0;
1120 		return err;
1121 	}
1122 
1123 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1124 		return -EMSGSIZE;
1125 
1126 	return 0;
1127 }
1128 
1129 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1130 {
1131 	char name[IFNAMSIZ];
1132 	int err;
1133 
1134 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1135 	if (err) {
1136 		if (err == -EOPNOTSUPP)
1137 			return 0;
1138 		return err;
1139 	}
1140 
1141 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1142 		return -EMSGSIZE;
1143 
1144 	return 0;
1145 }
1146 
1147 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1148 {
1149 	struct netdev_phys_item_id ppid = { };
1150 	int err;
1151 
1152 	err = dev_get_port_parent_id(dev, &ppid, false);
1153 	if (err) {
1154 		if (err == -EOPNOTSUPP)
1155 			return 0;
1156 		return err;
1157 	}
1158 
1159 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1160 		return -EMSGSIZE;
1161 
1162 	return 0;
1163 }
1164 
1165 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1166 					      struct net_device *dev)
1167 {
1168 	struct rtnl_link_stats64 *sp;
1169 	struct nlattr *attr;
1170 
1171 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1172 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1173 	if (!attr)
1174 		return -EMSGSIZE;
1175 
1176 	sp = nla_data(attr);
1177 	dev_get_stats(dev, sp);
1178 
1179 	attr = nla_reserve(skb, IFLA_STATS,
1180 			   sizeof(struct rtnl_link_stats));
1181 	if (!attr)
1182 		return -EMSGSIZE;
1183 
1184 	copy_rtnl_link_stats(nla_data(attr), sp);
1185 
1186 	return 0;
1187 }
1188 
1189 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1190 					       struct net_device *dev,
1191 					       int vfs_num,
1192 					       struct nlattr *vfinfo)
1193 {
1194 	struct ifla_vf_rss_query_en vf_rss_query_en;
1195 	struct nlattr *vf, *vfstats, *vfvlanlist;
1196 	struct ifla_vf_link_state vf_linkstate;
1197 	struct ifla_vf_vlan_info vf_vlan_info;
1198 	struct ifla_vf_spoofchk vf_spoofchk;
1199 	struct ifla_vf_tx_rate vf_tx_rate;
1200 	struct ifla_vf_stats vf_stats;
1201 	struct ifla_vf_trust vf_trust;
1202 	struct ifla_vf_vlan vf_vlan;
1203 	struct ifla_vf_rate vf_rate;
1204 	struct ifla_vf_mac vf_mac;
1205 	struct ifla_vf_broadcast vf_broadcast;
1206 	struct ifla_vf_info ivi;
1207 
1208 	memset(&ivi, 0, sizeof(ivi));
1209 
1210 	/* Not all SR-IOV capable drivers support the
1211 	 * spoofcheck and "RSS query enable" query.  Preset to
1212 	 * -1 so the user space tool can detect that the driver
1213 	 * didn't report anything.
1214 	 */
1215 	ivi.spoofchk = -1;
1216 	ivi.rss_query_en = -1;
1217 	ivi.trusted = -1;
1218 	/* The default value for VF link state is "auto"
1219 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1220 	 */
1221 	ivi.linkstate = 0;
1222 	/* VLAN Protocol by default is 802.1Q */
1223 	ivi.vlan_proto = htons(ETH_P_8021Q);
1224 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1225 		return 0;
1226 
1227 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1228 
1229 	vf_mac.vf =
1230 		vf_vlan.vf =
1231 		vf_vlan_info.vf =
1232 		vf_rate.vf =
1233 		vf_tx_rate.vf =
1234 		vf_spoofchk.vf =
1235 		vf_linkstate.vf =
1236 		vf_rss_query_en.vf =
1237 		vf_trust.vf = ivi.vf;
1238 
1239 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1240 	memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1241 	vf_vlan.vlan = ivi.vlan;
1242 	vf_vlan.qos = ivi.qos;
1243 	vf_vlan_info.vlan = ivi.vlan;
1244 	vf_vlan_info.qos = ivi.qos;
1245 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1246 	vf_tx_rate.rate = ivi.max_tx_rate;
1247 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1248 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1249 	vf_spoofchk.setting = ivi.spoofchk;
1250 	vf_linkstate.link_state = ivi.linkstate;
1251 	vf_rss_query_en.setting = ivi.rss_query_en;
1252 	vf_trust.setting = ivi.trusted;
1253 	vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1254 	if (!vf)
1255 		goto nla_put_vfinfo_failure;
1256 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1257 	    nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1258 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1259 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1260 		    &vf_rate) ||
1261 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1262 		    &vf_tx_rate) ||
1263 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1264 		    &vf_spoofchk) ||
1265 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1266 		    &vf_linkstate) ||
1267 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1268 		    sizeof(vf_rss_query_en),
1269 		    &vf_rss_query_en) ||
1270 	    nla_put(skb, IFLA_VF_TRUST,
1271 		    sizeof(vf_trust), &vf_trust))
1272 		goto nla_put_vf_failure;
1273 	vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1274 	if (!vfvlanlist)
1275 		goto nla_put_vf_failure;
1276 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1277 		    &vf_vlan_info)) {
1278 		nla_nest_cancel(skb, vfvlanlist);
1279 		goto nla_put_vf_failure;
1280 	}
1281 	nla_nest_end(skb, vfvlanlist);
1282 	memset(&vf_stats, 0, sizeof(vf_stats));
1283 	if (dev->netdev_ops->ndo_get_vf_stats)
1284 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1285 						&vf_stats);
1286 	vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1287 	if (!vfstats)
1288 		goto nla_put_vf_failure;
1289 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1290 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1291 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1292 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1293 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1294 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1295 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1296 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1297 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1298 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1299 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1300 			      vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1301 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1302 			      vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1303 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1304 			      vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1305 		nla_nest_cancel(skb, vfstats);
1306 		goto nla_put_vf_failure;
1307 	}
1308 	nla_nest_end(skb, vfstats);
1309 	nla_nest_end(skb, vf);
1310 	return 0;
1311 
1312 nla_put_vf_failure:
1313 	nla_nest_cancel(skb, vf);
1314 nla_put_vfinfo_failure:
1315 	nla_nest_cancel(skb, vfinfo);
1316 	return -EMSGSIZE;
1317 }
1318 
1319 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1320 					   struct net_device *dev,
1321 					   u32 ext_filter_mask)
1322 {
1323 	struct nlattr *vfinfo;
1324 	int i, num_vfs;
1325 
1326 	if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1327 		return 0;
1328 
1329 	num_vfs = dev_num_vf(dev->dev.parent);
1330 	if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1331 		return -EMSGSIZE;
1332 
1333 	if (!dev->netdev_ops->ndo_get_vf_config)
1334 		return 0;
1335 
1336 	vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1337 	if (!vfinfo)
1338 		return -EMSGSIZE;
1339 
1340 	for (i = 0; i < num_vfs; i++) {
1341 		if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1342 			return -EMSGSIZE;
1343 	}
1344 
1345 	nla_nest_end(skb, vfinfo);
1346 	return 0;
1347 }
1348 
1349 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1350 {
1351 	struct rtnl_link_ifmap map;
1352 
1353 	memset(&map, 0, sizeof(map));
1354 	map.mem_start   = dev->mem_start;
1355 	map.mem_end     = dev->mem_end;
1356 	map.base_addr   = dev->base_addr;
1357 	map.irq         = dev->irq;
1358 	map.dma         = dev->dma;
1359 	map.port        = dev->if_port;
1360 
1361 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1362 		return -EMSGSIZE;
1363 
1364 	return 0;
1365 }
1366 
1367 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1368 {
1369 	const struct bpf_prog *generic_xdp_prog;
1370 
1371 	ASSERT_RTNL();
1372 
1373 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1374 	if (!generic_xdp_prog)
1375 		return 0;
1376 	return generic_xdp_prog->aux->id;
1377 }
1378 
1379 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1380 {
1381 	return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG);
1382 }
1383 
1384 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1385 {
1386 	return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf,
1387 			       XDP_QUERY_PROG_HW);
1388 }
1389 
1390 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1391 			       u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1392 			       u32 (*get_prog_id)(struct net_device *dev))
1393 {
1394 	u32 curr_id;
1395 	int err;
1396 
1397 	curr_id = get_prog_id(dev);
1398 	if (!curr_id)
1399 		return 0;
1400 
1401 	*prog_id = curr_id;
1402 	err = nla_put_u32(skb, attr, curr_id);
1403 	if (err)
1404 		return err;
1405 
1406 	if (*mode != XDP_ATTACHED_NONE)
1407 		*mode = XDP_ATTACHED_MULTI;
1408 	else
1409 		*mode = tgt_mode;
1410 
1411 	return 0;
1412 }
1413 
1414 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1415 {
1416 	struct nlattr *xdp;
1417 	u32 prog_id;
1418 	int err;
1419 	u8 mode;
1420 
1421 	xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1422 	if (!xdp)
1423 		return -EMSGSIZE;
1424 
1425 	prog_id = 0;
1426 	mode = XDP_ATTACHED_NONE;
1427 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1428 				  IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1429 	if (err)
1430 		goto err_cancel;
1431 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1432 				  IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1433 	if (err)
1434 		goto err_cancel;
1435 	err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1436 				  IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1437 	if (err)
1438 		goto err_cancel;
1439 
1440 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1441 	if (err)
1442 		goto err_cancel;
1443 
1444 	if (prog_id && mode != XDP_ATTACHED_MULTI) {
1445 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1446 		if (err)
1447 			goto err_cancel;
1448 	}
1449 
1450 	nla_nest_end(skb, xdp);
1451 	return 0;
1452 
1453 err_cancel:
1454 	nla_nest_cancel(skb, xdp);
1455 	return err;
1456 }
1457 
1458 static u32 rtnl_get_event(unsigned long event)
1459 {
1460 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1461 
1462 	switch (event) {
1463 	case NETDEV_REBOOT:
1464 		rtnl_event_type = IFLA_EVENT_REBOOT;
1465 		break;
1466 	case NETDEV_FEAT_CHANGE:
1467 		rtnl_event_type = IFLA_EVENT_FEATURES;
1468 		break;
1469 	case NETDEV_BONDING_FAILOVER:
1470 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1471 		break;
1472 	case NETDEV_NOTIFY_PEERS:
1473 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1474 		break;
1475 	case NETDEV_RESEND_IGMP:
1476 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1477 		break;
1478 	case NETDEV_CHANGEINFODATA:
1479 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1480 		break;
1481 	default:
1482 		break;
1483 	}
1484 
1485 	return rtnl_event_type;
1486 }
1487 
1488 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1489 {
1490 	const struct net_device *upper_dev;
1491 	int ret = 0;
1492 
1493 	rcu_read_lock();
1494 
1495 	upper_dev = netdev_master_upper_dev_get_rcu(dev);
1496 	if (upper_dev)
1497 		ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1498 
1499 	rcu_read_unlock();
1500 	return ret;
1501 }
1502 
1503 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1504 			  bool force)
1505 {
1506 	int ifindex = dev_get_iflink(dev);
1507 
1508 	if (force || dev->ifindex != ifindex)
1509 		return nla_put_u32(skb, IFLA_LINK, ifindex);
1510 
1511 	return 0;
1512 }
1513 
1514 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1515 					      struct net_device *dev)
1516 {
1517 	char buf[IFALIASZ];
1518 	int ret;
1519 
1520 	ret = dev_get_alias(dev, buf, sizeof(buf));
1521 	return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1522 }
1523 
1524 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1525 				  const struct net_device *dev,
1526 				  struct net *src_net, gfp_t gfp)
1527 {
1528 	bool put_iflink = false;
1529 
1530 	if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1531 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1532 
1533 		if (!net_eq(dev_net(dev), link_net)) {
1534 			int id = peernet2id_alloc(src_net, link_net, gfp);
1535 
1536 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1537 				return -EMSGSIZE;
1538 
1539 			put_iflink = true;
1540 		}
1541 	}
1542 
1543 	return nla_put_iflink(skb, dev, put_iflink);
1544 }
1545 
1546 static int rtnl_fill_link_af(struct sk_buff *skb,
1547 			     const struct net_device *dev,
1548 			     u32 ext_filter_mask)
1549 {
1550 	const struct rtnl_af_ops *af_ops;
1551 	struct nlattr *af_spec;
1552 
1553 	af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1554 	if (!af_spec)
1555 		return -EMSGSIZE;
1556 
1557 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1558 		struct nlattr *af;
1559 		int err;
1560 
1561 		if (!af_ops->fill_link_af)
1562 			continue;
1563 
1564 		af = nla_nest_start_noflag(skb, af_ops->family);
1565 		if (!af)
1566 			return -EMSGSIZE;
1567 
1568 		err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1569 		/*
1570 		 * Caller may return ENODATA to indicate that there
1571 		 * was no data to be dumped. This is not an error, it
1572 		 * means we should trim the attribute header and
1573 		 * continue.
1574 		 */
1575 		if (err == -ENODATA)
1576 			nla_nest_cancel(skb, af);
1577 		else if (err < 0)
1578 			return -EMSGSIZE;
1579 
1580 		nla_nest_end(skb, af);
1581 	}
1582 
1583 	nla_nest_end(skb, af_spec);
1584 	return 0;
1585 }
1586 
1587 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1588 			    struct net_device *dev, struct net *src_net,
1589 			    int type, u32 pid, u32 seq, u32 change,
1590 			    unsigned int flags, u32 ext_filter_mask,
1591 			    u32 event, int *new_nsid, int new_ifindex,
1592 			    int tgt_netnsid, gfp_t gfp)
1593 {
1594 	struct ifinfomsg *ifm;
1595 	struct nlmsghdr *nlh;
1596 
1597 	ASSERT_RTNL();
1598 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1599 	if (nlh == NULL)
1600 		return -EMSGSIZE;
1601 
1602 	ifm = nlmsg_data(nlh);
1603 	ifm->ifi_family = AF_UNSPEC;
1604 	ifm->__ifi_pad = 0;
1605 	ifm->ifi_type = dev->type;
1606 	ifm->ifi_index = dev->ifindex;
1607 	ifm->ifi_flags = dev_get_flags(dev);
1608 	ifm->ifi_change = change;
1609 
1610 	if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1611 		goto nla_put_failure;
1612 
1613 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1614 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1615 	    nla_put_u8(skb, IFLA_OPERSTATE,
1616 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1617 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1618 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1619 	    nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1620 	    nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1621 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1622 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1623 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1624 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1625 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1626 #ifdef CONFIG_RPS
1627 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1628 #endif
1629 	    put_master_ifindex(skb, dev) ||
1630 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1631 	    (dev->qdisc &&
1632 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1633 	    nla_put_ifalias(skb, dev) ||
1634 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1635 			atomic_read(&dev->carrier_up_count) +
1636 			atomic_read(&dev->carrier_down_count)) ||
1637 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1638 	    nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1639 			atomic_read(&dev->carrier_up_count)) ||
1640 	    nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1641 			atomic_read(&dev->carrier_down_count)))
1642 		goto nla_put_failure;
1643 
1644 	if (event != IFLA_EVENT_NONE) {
1645 		if (nla_put_u32(skb, IFLA_EVENT, event))
1646 			goto nla_put_failure;
1647 	}
1648 
1649 	if (rtnl_fill_link_ifmap(skb, dev))
1650 		goto nla_put_failure;
1651 
1652 	if (dev->addr_len) {
1653 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1654 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1655 			goto nla_put_failure;
1656 	}
1657 
1658 	if (rtnl_phys_port_id_fill(skb, dev))
1659 		goto nla_put_failure;
1660 
1661 	if (rtnl_phys_port_name_fill(skb, dev))
1662 		goto nla_put_failure;
1663 
1664 	if (rtnl_phys_switch_id_fill(skb, dev))
1665 		goto nla_put_failure;
1666 
1667 	if (rtnl_fill_stats(skb, dev))
1668 		goto nla_put_failure;
1669 
1670 	if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1671 		goto nla_put_failure;
1672 
1673 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1674 		goto nla_put_failure;
1675 
1676 	if (rtnl_xdp_fill(skb, dev))
1677 		goto nla_put_failure;
1678 
1679 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1680 		if (rtnl_link_fill(skb, dev) < 0)
1681 			goto nla_put_failure;
1682 	}
1683 
1684 	if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1685 		goto nla_put_failure;
1686 
1687 	if (new_nsid &&
1688 	    nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1689 		goto nla_put_failure;
1690 	if (new_ifindex &&
1691 	    nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1692 		goto nla_put_failure;
1693 
1694 
1695 	rcu_read_lock();
1696 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1697 		goto nla_put_failure_rcu;
1698 	rcu_read_unlock();
1699 
1700 	nlmsg_end(skb, nlh);
1701 	return 0;
1702 
1703 nla_put_failure_rcu:
1704 	rcu_read_unlock();
1705 nla_put_failure:
1706 	nlmsg_cancel(skb, nlh);
1707 	return -EMSGSIZE;
1708 }
1709 
1710 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1711 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1712 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1713 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1714 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1715 	[IFLA_MTU]		= { .type = NLA_U32 },
1716 	[IFLA_LINK]		= { .type = NLA_U32 },
1717 	[IFLA_MASTER]		= { .type = NLA_U32 },
1718 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1719 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1720 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1721 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1722 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1723 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1724 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1725 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1726 	/* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1727 	 * allow 0-length string (needed to remove an alias).
1728 	 */
1729 	[IFLA_IFALIAS]	        = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1730 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1731 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1732 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1733 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1734 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1735 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1736 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1737 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1738 	[IFLA_GSO_MAX_SEGS]	= { .type = NLA_U32 },
1739 	[IFLA_GSO_MAX_SIZE]	= { .type = NLA_U32 },
1740 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1741 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1742 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1743 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1744 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1745 	[IFLA_XDP]		= { .type = NLA_NESTED },
1746 	[IFLA_EVENT]		= { .type = NLA_U32 },
1747 	[IFLA_GROUP]		= { .type = NLA_U32 },
1748 	[IFLA_TARGET_NETNSID]	= { .type = NLA_S32 },
1749 	[IFLA_CARRIER_UP_COUNT]	= { .type = NLA_U32 },
1750 	[IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1751 	[IFLA_MIN_MTU]		= { .type = NLA_U32 },
1752 	[IFLA_MAX_MTU]		= { .type = NLA_U32 },
1753 };
1754 
1755 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1756 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1757 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1758 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1759 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1760 };
1761 
1762 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1763 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1764 	[IFLA_VF_BROADCAST]	= { .type = NLA_REJECT },
1765 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1766 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1767 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1768 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1769 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1770 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1771 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1772 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1773 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1774 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1775 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1776 };
1777 
1778 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1779 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1780 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1781 				    .len = PORT_PROFILE_MAX },
1782 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1783 				      .len = PORT_UUID_MAX },
1784 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1785 				    .len = PORT_UUID_MAX },
1786 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1787 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1788 
1789 	/* Unused, but we need to keep it here since user space could
1790 	 * fill it. It's also broken with regard to NLA_BINARY use in
1791 	 * combination with structs.
1792 	 */
1793 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1794 				    .len = sizeof(struct ifla_port_vsi) },
1795 };
1796 
1797 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1798 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1799 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1800 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1801 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1802 };
1803 
1804 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1805 {
1806 	const struct rtnl_link_ops *ops = NULL;
1807 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1808 
1809 	if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1810 		return NULL;
1811 
1812 	if (linfo[IFLA_INFO_KIND]) {
1813 		char kind[MODULE_NAME_LEN];
1814 
1815 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1816 		ops = rtnl_link_ops_get(kind);
1817 	}
1818 
1819 	return ops;
1820 }
1821 
1822 static bool link_master_filtered(struct net_device *dev, int master_idx)
1823 {
1824 	struct net_device *master;
1825 
1826 	if (!master_idx)
1827 		return false;
1828 
1829 	master = netdev_master_upper_dev_get(dev);
1830 	if (!master || master->ifindex != master_idx)
1831 		return true;
1832 
1833 	return false;
1834 }
1835 
1836 static bool link_kind_filtered(const struct net_device *dev,
1837 			       const struct rtnl_link_ops *kind_ops)
1838 {
1839 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1840 		return true;
1841 
1842 	return false;
1843 }
1844 
1845 static bool link_dump_filtered(struct net_device *dev,
1846 			       int master_idx,
1847 			       const struct rtnl_link_ops *kind_ops)
1848 {
1849 	if (link_master_filtered(dev, master_idx) ||
1850 	    link_kind_filtered(dev, kind_ops))
1851 		return true;
1852 
1853 	return false;
1854 }
1855 
1856 /**
1857  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1858  * @sk: netlink socket
1859  * @netnsid: network namespace identifier
1860  *
1861  * Returns the network namespace identified by netnsid on success or an error
1862  * pointer on failure.
1863  */
1864 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1865 {
1866 	struct net *net;
1867 
1868 	net = get_net_ns_by_id(sock_net(sk), netnsid);
1869 	if (!net)
1870 		return ERR_PTR(-EINVAL);
1871 
1872 	/* For now, the caller is required to have CAP_NET_ADMIN in
1873 	 * the user namespace owning the target net ns.
1874 	 */
1875 	if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1876 		put_net(net);
1877 		return ERR_PTR(-EACCES);
1878 	}
1879 	return net;
1880 }
1881 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
1882 
1883 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
1884 				      bool strict_check, struct nlattr **tb,
1885 				      struct netlink_ext_ack *extack)
1886 {
1887 	int hdrlen;
1888 
1889 	if (strict_check) {
1890 		struct ifinfomsg *ifm;
1891 
1892 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1893 			NL_SET_ERR_MSG(extack, "Invalid header for link dump");
1894 			return -EINVAL;
1895 		}
1896 
1897 		ifm = nlmsg_data(nlh);
1898 		if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
1899 		    ifm->ifi_change) {
1900 			NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
1901 			return -EINVAL;
1902 		}
1903 		if (ifm->ifi_index) {
1904 			NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
1905 			return -EINVAL;
1906 		}
1907 
1908 		return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
1909 						     IFLA_MAX, ifla_policy,
1910 						     extack);
1911 	}
1912 
1913 	/* A hack to preserve kernel<->userspace interface.
1914 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1915 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1916 	 * what iproute2 < v3.9.0 used.
1917 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1918 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1919 	 */
1920 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1921 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1922 
1923 	return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
1924 				      extack);
1925 }
1926 
1927 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1928 {
1929 	struct netlink_ext_ack *extack = cb->extack;
1930 	const struct nlmsghdr *nlh = cb->nlh;
1931 	struct net *net = sock_net(skb->sk);
1932 	struct net *tgt_net = net;
1933 	int h, s_h;
1934 	int idx = 0, s_idx;
1935 	struct net_device *dev;
1936 	struct hlist_head *head;
1937 	struct nlattr *tb[IFLA_MAX+1];
1938 	u32 ext_filter_mask = 0;
1939 	const struct rtnl_link_ops *kind_ops = NULL;
1940 	unsigned int flags = NLM_F_MULTI;
1941 	int master_idx = 0;
1942 	int netnsid = -1;
1943 	int err, i;
1944 
1945 	s_h = cb->args[0];
1946 	s_idx = cb->args[1];
1947 
1948 	err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
1949 	if (err < 0) {
1950 		if (cb->strict_check)
1951 			return err;
1952 
1953 		goto walk_entries;
1954 	}
1955 
1956 	for (i = 0; i <= IFLA_MAX; ++i) {
1957 		if (!tb[i])
1958 			continue;
1959 
1960 		/* new attributes should only be added with strict checking */
1961 		switch (i) {
1962 		case IFLA_TARGET_NETNSID:
1963 			netnsid = nla_get_s32(tb[i]);
1964 			tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
1965 			if (IS_ERR(tgt_net)) {
1966 				NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
1967 				return PTR_ERR(tgt_net);
1968 			}
1969 			break;
1970 		case IFLA_EXT_MASK:
1971 			ext_filter_mask = nla_get_u32(tb[i]);
1972 			break;
1973 		case IFLA_MASTER:
1974 			master_idx = nla_get_u32(tb[i]);
1975 			break;
1976 		case IFLA_LINKINFO:
1977 			kind_ops = linkinfo_to_kind_ops(tb[i]);
1978 			break;
1979 		default:
1980 			if (cb->strict_check) {
1981 				NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
1982 				return -EINVAL;
1983 			}
1984 		}
1985 	}
1986 
1987 	if (master_idx || kind_ops)
1988 		flags |= NLM_F_DUMP_FILTERED;
1989 
1990 walk_entries:
1991 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1992 		idx = 0;
1993 		head = &tgt_net->dev_index_head[h];
1994 		hlist_for_each_entry(dev, head, index_hlist) {
1995 			if (link_dump_filtered(dev, master_idx, kind_ops))
1996 				goto cont;
1997 			if (idx < s_idx)
1998 				goto cont;
1999 			err = rtnl_fill_ifinfo(skb, dev, net,
2000 					       RTM_NEWLINK,
2001 					       NETLINK_CB(cb->skb).portid,
2002 					       nlh->nlmsg_seq, 0, flags,
2003 					       ext_filter_mask, 0, NULL, 0,
2004 					       netnsid, GFP_KERNEL);
2005 
2006 			if (err < 0) {
2007 				if (likely(skb->len))
2008 					goto out;
2009 
2010 				goto out_err;
2011 			}
2012 cont:
2013 			idx++;
2014 		}
2015 	}
2016 out:
2017 	err = skb->len;
2018 out_err:
2019 	cb->args[1] = idx;
2020 	cb->args[0] = h;
2021 	cb->seq = net->dev_base_seq;
2022 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2023 	if (netnsid >= 0)
2024 		put_net(tgt_net);
2025 
2026 	return err;
2027 }
2028 
2029 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2030 			struct netlink_ext_ack *exterr)
2031 {
2032 	return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2033 				    exterr);
2034 }
2035 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2036 
2037 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2038 {
2039 	struct net *net;
2040 	/* Examine the link attributes and figure out which
2041 	 * network namespace we are talking about.
2042 	 */
2043 	if (tb[IFLA_NET_NS_PID])
2044 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2045 	else if (tb[IFLA_NET_NS_FD])
2046 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2047 	else
2048 		net = get_net(src_net);
2049 	return net;
2050 }
2051 EXPORT_SYMBOL(rtnl_link_get_net);
2052 
2053 /* Figure out which network namespace we are talking about by
2054  * examining the link attributes in the following order:
2055  *
2056  * 1. IFLA_NET_NS_PID
2057  * 2. IFLA_NET_NS_FD
2058  * 3. IFLA_TARGET_NETNSID
2059  */
2060 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2061 					       struct nlattr *tb[])
2062 {
2063 	struct net *net;
2064 
2065 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2066 		return rtnl_link_get_net(src_net, tb);
2067 
2068 	if (!tb[IFLA_TARGET_NETNSID])
2069 		return get_net(src_net);
2070 
2071 	net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2072 	if (!net)
2073 		return ERR_PTR(-EINVAL);
2074 
2075 	return net;
2076 }
2077 
2078 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2079 					     struct net *src_net,
2080 					     struct nlattr *tb[], int cap)
2081 {
2082 	struct net *net;
2083 
2084 	net = rtnl_link_get_net_by_nlattr(src_net, tb);
2085 	if (IS_ERR(net))
2086 		return net;
2087 
2088 	if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2089 		put_net(net);
2090 		return ERR_PTR(-EPERM);
2091 	}
2092 
2093 	return net;
2094 }
2095 
2096 /* Verify that rtnetlink requests do not pass additional properties
2097  * potentially referring to different network namespaces.
2098  */
2099 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2100 				    struct netlink_ext_ack *extack,
2101 				    bool netns_id_only)
2102 {
2103 
2104 	if (netns_id_only) {
2105 		if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2106 			return 0;
2107 
2108 		NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2109 		return -EOPNOTSUPP;
2110 	}
2111 
2112 	if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2113 		goto invalid_attr;
2114 
2115 	if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2116 		goto invalid_attr;
2117 
2118 	if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2119 		goto invalid_attr;
2120 
2121 	return 0;
2122 
2123 invalid_attr:
2124 	NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2125 	return -EINVAL;
2126 }
2127 
2128 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2129 {
2130 	if (dev) {
2131 		if (tb[IFLA_ADDRESS] &&
2132 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2133 			return -EINVAL;
2134 
2135 		if (tb[IFLA_BROADCAST] &&
2136 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2137 			return -EINVAL;
2138 	}
2139 
2140 	if (tb[IFLA_AF_SPEC]) {
2141 		struct nlattr *af;
2142 		int rem, err;
2143 
2144 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2145 			const struct rtnl_af_ops *af_ops;
2146 
2147 			rcu_read_lock();
2148 			af_ops = rtnl_af_lookup(nla_type(af));
2149 			if (!af_ops) {
2150 				rcu_read_unlock();
2151 				return -EAFNOSUPPORT;
2152 			}
2153 
2154 			if (!af_ops->set_link_af) {
2155 				rcu_read_unlock();
2156 				return -EOPNOTSUPP;
2157 			}
2158 
2159 			if (af_ops->validate_link_af) {
2160 				err = af_ops->validate_link_af(dev, af);
2161 				if (err < 0) {
2162 					rcu_read_unlock();
2163 					return err;
2164 				}
2165 			}
2166 
2167 			rcu_read_unlock();
2168 		}
2169 	}
2170 
2171 	return 0;
2172 }
2173 
2174 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2175 				  int guid_type)
2176 {
2177 	const struct net_device_ops *ops = dev->netdev_ops;
2178 
2179 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2180 }
2181 
2182 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2183 {
2184 	if (dev->type != ARPHRD_INFINIBAND)
2185 		return -EOPNOTSUPP;
2186 
2187 	return handle_infiniband_guid(dev, ivt, guid_type);
2188 }
2189 
2190 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2191 {
2192 	const struct net_device_ops *ops = dev->netdev_ops;
2193 	int err = -EINVAL;
2194 
2195 	if (tb[IFLA_VF_MAC]) {
2196 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2197 
2198 		err = -EOPNOTSUPP;
2199 		if (ops->ndo_set_vf_mac)
2200 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
2201 						  ivm->mac);
2202 		if (err < 0)
2203 			return err;
2204 	}
2205 
2206 	if (tb[IFLA_VF_VLAN]) {
2207 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2208 
2209 		err = -EOPNOTSUPP;
2210 		if (ops->ndo_set_vf_vlan)
2211 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2212 						   ivv->qos,
2213 						   htons(ETH_P_8021Q));
2214 		if (err < 0)
2215 			return err;
2216 	}
2217 
2218 	if (tb[IFLA_VF_VLAN_LIST]) {
2219 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2220 		struct nlattr *attr;
2221 		int rem, len = 0;
2222 
2223 		err = -EOPNOTSUPP;
2224 		if (!ops->ndo_set_vf_vlan)
2225 			return err;
2226 
2227 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2228 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2229 			    nla_len(attr) < NLA_HDRLEN) {
2230 				return -EINVAL;
2231 			}
2232 			if (len >= MAX_VLAN_LIST_LEN)
2233 				return -EOPNOTSUPP;
2234 			ivvl[len] = nla_data(attr);
2235 
2236 			len++;
2237 		}
2238 		if (len == 0)
2239 			return -EINVAL;
2240 
2241 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2242 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
2243 		if (err < 0)
2244 			return err;
2245 	}
2246 
2247 	if (tb[IFLA_VF_TX_RATE]) {
2248 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2249 		struct ifla_vf_info ivf;
2250 
2251 		err = -EOPNOTSUPP;
2252 		if (ops->ndo_get_vf_config)
2253 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2254 		if (err < 0)
2255 			return err;
2256 
2257 		err = -EOPNOTSUPP;
2258 		if (ops->ndo_set_vf_rate)
2259 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2260 						   ivf.min_tx_rate,
2261 						   ivt->rate);
2262 		if (err < 0)
2263 			return err;
2264 	}
2265 
2266 	if (tb[IFLA_VF_RATE]) {
2267 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2268 
2269 		err = -EOPNOTSUPP;
2270 		if (ops->ndo_set_vf_rate)
2271 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
2272 						   ivt->min_tx_rate,
2273 						   ivt->max_tx_rate);
2274 		if (err < 0)
2275 			return err;
2276 	}
2277 
2278 	if (tb[IFLA_VF_SPOOFCHK]) {
2279 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2280 
2281 		err = -EOPNOTSUPP;
2282 		if (ops->ndo_set_vf_spoofchk)
2283 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2284 						       ivs->setting);
2285 		if (err < 0)
2286 			return err;
2287 	}
2288 
2289 	if (tb[IFLA_VF_LINK_STATE]) {
2290 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2291 
2292 		err = -EOPNOTSUPP;
2293 		if (ops->ndo_set_vf_link_state)
2294 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2295 							 ivl->link_state);
2296 		if (err < 0)
2297 			return err;
2298 	}
2299 
2300 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
2301 		struct ifla_vf_rss_query_en *ivrssq_en;
2302 
2303 		err = -EOPNOTSUPP;
2304 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2305 		if (ops->ndo_set_vf_rss_query_en)
2306 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2307 							   ivrssq_en->setting);
2308 		if (err < 0)
2309 			return err;
2310 	}
2311 
2312 	if (tb[IFLA_VF_TRUST]) {
2313 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2314 
2315 		err = -EOPNOTSUPP;
2316 		if (ops->ndo_set_vf_trust)
2317 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2318 		if (err < 0)
2319 			return err;
2320 	}
2321 
2322 	if (tb[IFLA_VF_IB_NODE_GUID]) {
2323 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2324 
2325 		if (!ops->ndo_set_vf_guid)
2326 			return -EOPNOTSUPP;
2327 
2328 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2329 	}
2330 
2331 	if (tb[IFLA_VF_IB_PORT_GUID]) {
2332 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2333 
2334 		if (!ops->ndo_set_vf_guid)
2335 			return -EOPNOTSUPP;
2336 
2337 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2338 	}
2339 
2340 	return err;
2341 }
2342 
2343 static int do_set_master(struct net_device *dev, int ifindex,
2344 			 struct netlink_ext_ack *extack)
2345 {
2346 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2347 	const struct net_device_ops *ops;
2348 	int err;
2349 
2350 	if (upper_dev) {
2351 		if (upper_dev->ifindex == ifindex)
2352 			return 0;
2353 		ops = upper_dev->netdev_ops;
2354 		if (ops->ndo_del_slave) {
2355 			err = ops->ndo_del_slave(upper_dev, dev);
2356 			if (err)
2357 				return err;
2358 			netdev_update_lockdep_key(dev);
2359 		} else {
2360 			return -EOPNOTSUPP;
2361 		}
2362 	}
2363 
2364 	if (ifindex) {
2365 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2366 		if (!upper_dev)
2367 			return -EINVAL;
2368 		ops = upper_dev->netdev_ops;
2369 		if (ops->ndo_add_slave) {
2370 			err = ops->ndo_add_slave(upper_dev, dev, extack);
2371 			if (err)
2372 				return err;
2373 		} else {
2374 			return -EOPNOTSUPP;
2375 		}
2376 	}
2377 	return 0;
2378 }
2379 
2380 #define DO_SETLINK_MODIFIED	0x01
2381 /* notify flag means notify + modified. */
2382 #define DO_SETLINK_NOTIFY	0x03
2383 static int do_setlink(const struct sk_buff *skb,
2384 		      struct net_device *dev, struct ifinfomsg *ifm,
2385 		      struct netlink_ext_ack *extack,
2386 		      struct nlattr **tb, char *ifname, int status)
2387 {
2388 	const struct net_device_ops *ops = dev->netdev_ops;
2389 	int err;
2390 
2391 	err = validate_linkmsg(dev, tb);
2392 	if (err < 0)
2393 		return err;
2394 
2395 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2396 		struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2397 							    tb, CAP_NET_ADMIN);
2398 		if (IS_ERR(net)) {
2399 			err = PTR_ERR(net);
2400 			goto errout;
2401 		}
2402 
2403 		err = dev_change_net_namespace(dev, net, ifname);
2404 		put_net(net);
2405 		if (err)
2406 			goto errout;
2407 		status |= DO_SETLINK_MODIFIED;
2408 	}
2409 
2410 	if (tb[IFLA_MAP]) {
2411 		struct rtnl_link_ifmap *u_map;
2412 		struct ifmap k_map;
2413 
2414 		if (!ops->ndo_set_config) {
2415 			err = -EOPNOTSUPP;
2416 			goto errout;
2417 		}
2418 
2419 		if (!netif_device_present(dev)) {
2420 			err = -ENODEV;
2421 			goto errout;
2422 		}
2423 
2424 		u_map = nla_data(tb[IFLA_MAP]);
2425 		k_map.mem_start = (unsigned long) u_map->mem_start;
2426 		k_map.mem_end = (unsigned long) u_map->mem_end;
2427 		k_map.base_addr = (unsigned short) u_map->base_addr;
2428 		k_map.irq = (unsigned char) u_map->irq;
2429 		k_map.dma = (unsigned char) u_map->dma;
2430 		k_map.port = (unsigned char) u_map->port;
2431 
2432 		err = ops->ndo_set_config(dev, &k_map);
2433 		if (err < 0)
2434 			goto errout;
2435 
2436 		status |= DO_SETLINK_NOTIFY;
2437 	}
2438 
2439 	if (tb[IFLA_ADDRESS]) {
2440 		struct sockaddr *sa;
2441 		int len;
2442 
2443 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2444 						  sizeof(*sa));
2445 		sa = kmalloc(len, GFP_KERNEL);
2446 		if (!sa) {
2447 			err = -ENOMEM;
2448 			goto errout;
2449 		}
2450 		sa->sa_family = dev->type;
2451 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2452 		       dev->addr_len);
2453 		err = dev_set_mac_address(dev, sa, extack);
2454 		kfree(sa);
2455 		if (err)
2456 			goto errout;
2457 		status |= DO_SETLINK_MODIFIED;
2458 	}
2459 
2460 	if (tb[IFLA_MTU]) {
2461 		err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2462 		if (err < 0)
2463 			goto errout;
2464 		status |= DO_SETLINK_MODIFIED;
2465 	}
2466 
2467 	if (tb[IFLA_GROUP]) {
2468 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2469 		status |= DO_SETLINK_NOTIFY;
2470 	}
2471 
2472 	/*
2473 	 * Interface selected by interface index but interface
2474 	 * name provided implies that a name change has been
2475 	 * requested.
2476 	 */
2477 	if (ifm->ifi_index > 0 && ifname[0]) {
2478 		err = dev_change_name(dev, ifname);
2479 		if (err < 0)
2480 			goto errout;
2481 		status |= DO_SETLINK_MODIFIED;
2482 	}
2483 
2484 	if (tb[IFLA_IFALIAS]) {
2485 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2486 				    nla_len(tb[IFLA_IFALIAS]));
2487 		if (err < 0)
2488 			goto errout;
2489 		status |= DO_SETLINK_NOTIFY;
2490 	}
2491 
2492 	if (tb[IFLA_BROADCAST]) {
2493 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2494 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2495 	}
2496 
2497 	if (ifm->ifi_flags || ifm->ifi_change) {
2498 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2499 				       extack);
2500 		if (err < 0)
2501 			goto errout;
2502 	}
2503 
2504 	if (tb[IFLA_MASTER]) {
2505 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2506 		if (err)
2507 			goto errout;
2508 		status |= DO_SETLINK_MODIFIED;
2509 	}
2510 
2511 	if (tb[IFLA_CARRIER]) {
2512 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2513 		if (err)
2514 			goto errout;
2515 		status |= DO_SETLINK_MODIFIED;
2516 	}
2517 
2518 	if (tb[IFLA_TXQLEN]) {
2519 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2520 
2521 		err = dev_change_tx_queue_len(dev, value);
2522 		if (err)
2523 			goto errout;
2524 		status |= DO_SETLINK_MODIFIED;
2525 	}
2526 
2527 	if (tb[IFLA_GSO_MAX_SIZE]) {
2528 		u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2529 
2530 		if (max_size > GSO_MAX_SIZE) {
2531 			err = -EINVAL;
2532 			goto errout;
2533 		}
2534 
2535 		if (dev->gso_max_size ^ max_size) {
2536 			netif_set_gso_max_size(dev, max_size);
2537 			status |= DO_SETLINK_MODIFIED;
2538 		}
2539 	}
2540 
2541 	if (tb[IFLA_GSO_MAX_SEGS]) {
2542 		u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2543 
2544 		if (max_segs > GSO_MAX_SEGS) {
2545 			err = -EINVAL;
2546 			goto errout;
2547 		}
2548 
2549 		if (dev->gso_max_segs ^ max_segs) {
2550 			dev->gso_max_segs = max_segs;
2551 			status |= DO_SETLINK_MODIFIED;
2552 		}
2553 	}
2554 
2555 	if (tb[IFLA_OPERSTATE])
2556 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2557 
2558 	if (tb[IFLA_LINKMODE]) {
2559 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2560 
2561 		write_lock_bh(&dev_base_lock);
2562 		if (dev->link_mode ^ value)
2563 			status |= DO_SETLINK_NOTIFY;
2564 		dev->link_mode = value;
2565 		write_unlock_bh(&dev_base_lock);
2566 	}
2567 
2568 	if (tb[IFLA_VFINFO_LIST]) {
2569 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2570 		struct nlattr *attr;
2571 		int rem;
2572 
2573 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2574 			if (nla_type(attr) != IFLA_VF_INFO ||
2575 			    nla_len(attr) < NLA_HDRLEN) {
2576 				err = -EINVAL;
2577 				goto errout;
2578 			}
2579 			err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2580 							  attr,
2581 							  ifla_vf_policy,
2582 							  NULL);
2583 			if (err < 0)
2584 				goto errout;
2585 			err = do_setvfinfo(dev, vfinfo);
2586 			if (err < 0)
2587 				goto errout;
2588 			status |= DO_SETLINK_NOTIFY;
2589 		}
2590 	}
2591 	err = 0;
2592 
2593 	if (tb[IFLA_VF_PORTS]) {
2594 		struct nlattr *port[IFLA_PORT_MAX+1];
2595 		struct nlattr *attr;
2596 		int vf;
2597 		int rem;
2598 
2599 		err = -EOPNOTSUPP;
2600 		if (!ops->ndo_set_vf_port)
2601 			goto errout;
2602 
2603 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2604 			if (nla_type(attr) != IFLA_VF_PORT ||
2605 			    nla_len(attr) < NLA_HDRLEN) {
2606 				err = -EINVAL;
2607 				goto errout;
2608 			}
2609 			err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2610 							  attr,
2611 							  ifla_port_policy,
2612 							  NULL);
2613 			if (err < 0)
2614 				goto errout;
2615 			if (!port[IFLA_PORT_VF]) {
2616 				err = -EOPNOTSUPP;
2617 				goto errout;
2618 			}
2619 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2620 			err = ops->ndo_set_vf_port(dev, vf, port);
2621 			if (err < 0)
2622 				goto errout;
2623 			status |= DO_SETLINK_NOTIFY;
2624 		}
2625 	}
2626 	err = 0;
2627 
2628 	if (tb[IFLA_PORT_SELF]) {
2629 		struct nlattr *port[IFLA_PORT_MAX+1];
2630 
2631 		err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2632 						  tb[IFLA_PORT_SELF],
2633 						  ifla_port_policy, NULL);
2634 		if (err < 0)
2635 			goto errout;
2636 
2637 		err = -EOPNOTSUPP;
2638 		if (ops->ndo_set_vf_port)
2639 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2640 		if (err < 0)
2641 			goto errout;
2642 		status |= DO_SETLINK_NOTIFY;
2643 	}
2644 
2645 	if (tb[IFLA_AF_SPEC]) {
2646 		struct nlattr *af;
2647 		int rem;
2648 
2649 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2650 			const struct rtnl_af_ops *af_ops;
2651 
2652 			rcu_read_lock();
2653 
2654 			BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2655 
2656 			err = af_ops->set_link_af(dev, af);
2657 			if (err < 0) {
2658 				rcu_read_unlock();
2659 				goto errout;
2660 			}
2661 
2662 			rcu_read_unlock();
2663 			status |= DO_SETLINK_NOTIFY;
2664 		}
2665 	}
2666 	err = 0;
2667 
2668 	if (tb[IFLA_PROTO_DOWN]) {
2669 		err = dev_change_proto_down(dev,
2670 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2671 		if (err)
2672 			goto errout;
2673 		status |= DO_SETLINK_NOTIFY;
2674 	}
2675 
2676 	if (tb[IFLA_XDP]) {
2677 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2678 		u32 xdp_flags = 0;
2679 
2680 		err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2681 						  tb[IFLA_XDP],
2682 						  ifla_xdp_policy, NULL);
2683 		if (err < 0)
2684 			goto errout;
2685 
2686 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2687 			err = -EINVAL;
2688 			goto errout;
2689 		}
2690 
2691 		if (xdp[IFLA_XDP_FLAGS]) {
2692 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2693 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2694 				err = -EINVAL;
2695 				goto errout;
2696 			}
2697 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2698 				err = -EINVAL;
2699 				goto errout;
2700 			}
2701 		}
2702 
2703 		if (xdp[IFLA_XDP_FD]) {
2704 			err = dev_change_xdp_fd(dev, extack,
2705 						nla_get_s32(xdp[IFLA_XDP_FD]),
2706 						xdp_flags);
2707 			if (err)
2708 				goto errout;
2709 			status |= DO_SETLINK_NOTIFY;
2710 		}
2711 	}
2712 
2713 errout:
2714 	if (status & DO_SETLINK_MODIFIED) {
2715 		if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2716 			netdev_state_change(dev);
2717 
2718 		if (err < 0)
2719 			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",
2720 					     dev->name);
2721 	}
2722 
2723 	return err;
2724 }
2725 
2726 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2727 			struct netlink_ext_ack *extack)
2728 {
2729 	struct net *net = sock_net(skb->sk);
2730 	struct ifinfomsg *ifm;
2731 	struct net_device *dev;
2732 	int err;
2733 	struct nlattr *tb[IFLA_MAX+1];
2734 	char ifname[IFNAMSIZ];
2735 
2736 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2737 				     ifla_policy, extack);
2738 	if (err < 0)
2739 		goto errout;
2740 
2741 	err = rtnl_ensure_unique_netns(tb, extack, false);
2742 	if (err < 0)
2743 		goto errout;
2744 
2745 	if (tb[IFLA_IFNAME])
2746 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2747 	else
2748 		ifname[0] = '\0';
2749 
2750 	err = -EINVAL;
2751 	ifm = nlmsg_data(nlh);
2752 	if (ifm->ifi_index > 0)
2753 		dev = __dev_get_by_index(net, ifm->ifi_index);
2754 	else if (tb[IFLA_IFNAME])
2755 		dev = __dev_get_by_name(net, ifname);
2756 	else
2757 		goto errout;
2758 
2759 	if (dev == NULL) {
2760 		err = -ENODEV;
2761 		goto errout;
2762 	}
2763 
2764 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2765 errout:
2766 	return err;
2767 }
2768 
2769 static int rtnl_group_dellink(const struct net *net, int group)
2770 {
2771 	struct net_device *dev, *aux;
2772 	LIST_HEAD(list_kill);
2773 	bool found = false;
2774 
2775 	if (!group)
2776 		return -EPERM;
2777 
2778 	for_each_netdev(net, dev) {
2779 		if (dev->group == group) {
2780 			const struct rtnl_link_ops *ops;
2781 
2782 			found = true;
2783 			ops = dev->rtnl_link_ops;
2784 			if (!ops || !ops->dellink)
2785 				return -EOPNOTSUPP;
2786 		}
2787 	}
2788 
2789 	if (!found)
2790 		return -ENODEV;
2791 
2792 	for_each_netdev_safe(net, dev, aux) {
2793 		if (dev->group == group) {
2794 			const struct rtnl_link_ops *ops;
2795 
2796 			ops = dev->rtnl_link_ops;
2797 			ops->dellink(dev, &list_kill);
2798 		}
2799 	}
2800 	unregister_netdevice_many(&list_kill);
2801 
2802 	return 0;
2803 }
2804 
2805 int rtnl_delete_link(struct net_device *dev)
2806 {
2807 	const struct rtnl_link_ops *ops;
2808 	LIST_HEAD(list_kill);
2809 
2810 	ops = dev->rtnl_link_ops;
2811 	if (!ops || !ops->dellink)
2812 		return -EOPNOTSUPP;
2813 
2814 	ops->dellink(dev, &list_kill);
2815 	unregister_netdevice_many(&list_kill);
2816 
2817 	return 0;
2818 }
2819 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2820 
2821 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2822 			struct netlink_ext_ack *extack)
2823 {
2824 	struct net *net = sock_net(skb->sk);
2825 	struct net *tgt_net = net;
2826 	struct net_device *dev = NULL;
2827 	struct ifinfomsg *ifm;
2828 	char ifname[IFNAMSIZ];
2829 	struct nlattr *tb[IFLA_MAX+1];
2830 	int err;
2831 	int netnsid = -1;
2832 
2833 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2834 				     ifla_policy, extack);
2835 	if (err < 0)
2836 		return err;
2837 
2838 	err = rtnl_ensure_unique_netns(tb, extack, true);
2839 	if (err < 0)
2840 		return err;
2841 
2842 	if (tb[IFLA_IFNAME])
2843 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2844 
2845 	if (tb[IFLA_TARGET_NETNSID]) {
2846 		netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
2847 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
2848 		if (IS_ERR(tgt_net))
2849 			return PTR_ERR(tgt_net);
2850 	}
2851 
2852 	err = -EINVAL;
2853 	ifm = nlmsg_data(nlh);
2854 	if (ifm->ifi_index > 0)
2855 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2856 	else if (tb[IFLA_IFNAME])
2857 		dev = __dev_get_by_name(tgt_net, ifname);
2858 	else if (tb[IFLA_GROUP])
2859 		err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2860 	else
2861 		goto out;
2862 
2863 	if (!dev) {
2864 		if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2865 			err = -ENODEV;
2866 
2867 		goto out;
2868 	}
2869 
2870 	err = rtnl_delete_link(dev);
2871 
2872 out:
2873 	if (netnsid >= 0)
2874 		put_net(tgt_net);
2875 
2876 	return err;
2877 }
2878 
2879 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2880 {
2881 	unsigned int old_flags;
2882 	int err;
2883 
2884 	old_flags = dev->flags;
2885 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2886 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2887 					 NULL);
2888 		if (err < 0)
2889 			return err;
2890 	}
2891 
2892 	if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2893 		__dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2894 	} else {
2895 		dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2896 		__dev_notify_flags(dev, old_flags, ~0U);
2897 	}
2898 	return 0;
2899 }
2900 EXPORT_SYMBOL(rtnl_configure_link);
2901 
2902 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
2903 				    unsigned char name_assign_type,
2904 				    const struct rtnl_link_ops *ops,
2905 				    struct nlattr *tb[],
2906 				    struct netlink_ext_ack *extack)
2907 {
2908 	struct net_device *dev;
2909 	unsigned int num_tx_queues = 1;
2910 	unsigned int num_rx_queues = 1;
2911 
2912 	if (tb[IFLA_NUM_TX_QUEUES])
2913 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2914 	else if (ops->get_num_tx_queues)
2915 		num_tx_queues = ops->get_num_tx_queues();
2916 
2917 	if (tb[IFLA_NUM_RX_QUEUES])
2918 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2919 	else if (ops->get_num_rx_queues)
2920 		num_rx_queues = ops->get_num_rx_queues();
2921 
2922 	if (num_tx_queues < 1 || num_tx_queues > 4096) {
2923 		NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
2924 		return ERR_PTR(-EINVAL);
2925 	}
2926 
2927 	if (num_rx_queues < 1 || num_rx_queues > 4096) {
2928 		NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
2929 		return ERR_PTR(-EINVAL);
2930 	}
2931 
2932 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2933 			       ops->setup, num_tx_queues, num_rx_queues);
2934 	if (!dev)
2935 		return ERR_PTR(-ENOMEM);
2936 
2937 	dev_net_set(dev, net);
2938 	dev->rtnl_link_ops = ops;
2939 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2940 
2941 	if (tb[IFLA_MTU])
2942 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2943 	if (tb[IFLA_ADDRESS]) {
2944 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2945 				nla_len(tb[IFLA_ADDRESS]));
2946 		dev->addr_assign_type = NET_ADDR_SET;
2947 	}
2948 	if (tb[IFLA_BROADCAST])
2949 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2950 				nla_len(tb[IFLA_BROADCAST]));
2951 	if (tb[IFLA_TXQLEN])
2952 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2953 	if (tb[IFLA_OPERSTATE])
2954 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2955 	if (tb[IFLA_LINKMODE])
2956 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2957 	if (tb[IFLA_GROUP])
2958 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2959 	if (tb[IFLA_GSO_MAX_SIZE])
2960 		netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2961 	if (tb[IFLA_GSO_MAX_SEGS])
2962 		dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2963 
2964 	return dev;
2965 }
2966 EXPORT_SYMBOL(rtnl_create_link);
2967 
2968 static int rtnl_group_changelink(const struct sk_buff *skb,
2969 		struct net *net, int group,
2970 		struct ifinfomsg *ifm,
2971 		struct netlink_ext_ack *extack,
2972 		struct nlattr **tb)
2973 {
2974 	struct net_device *dev, *aux;
2975 	int err;
2976 
2977 	for_each_netdev_safe(net, dev, aux) {
2978 		if (dev->group == group) {
2979 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2980 			if (err < 0)
2981 				return err;
2982 		}
2983 	}
2984 
2985 	return 0;
2986 }
2987 
2988 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2989 			  struct nlattr **attr, struct netlink_ext_ack *extack)
2990 {
2991 	struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
2992 	unsigned char name_assign_type = NET_NAME_USER;
2993 	struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
2994 	const struct rtnl_link_ops *m_ops = NULL;
2995 	struct net_device *master_dev = NULL;
2996 	struct net *net = sock_net(skb->sk);
2997 	const struct rtnl_link_ops *ops;
2998 	struct nlattr *tb[IFLA_MAX + 1];
2999 	struct net *dest_net, *link_net;
3000 	struct nlattr **slave_data;
3001 	char kind[MODULE_NAME_LEN];
3002 	struct net_device *dev;
3003 	struct ifinfomsg *ifm;
3004 	char ifname[IFNAMSIZ];
3005 	struct nlattr **data;
3006 	int err;
3007 
3008 #ifdef CONFIG_MODULES
3009 replay:
3010 #endif
3011 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3012 				     ifla_policy, extack);
3013 	if (err < 0)
3014 		return err;
3015 
3016 	err = rtnl_ensure_unique_netns(tb, extack, false);
3017 	if (err < 0)
3018 		return err;
3019 
3020 	if (tb[IFLA_IFNAME])
3021 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3022 	else
3023 		ifname[0] = '\0';
3024 
3025 	ifm = nlmsg_data(nlh);
3026 	if (ifm->ifi_index > 0)
3027 		dev = __dev_get_by_index(net, ifm->ifi_index);
3028 	else {
3029 		if (ifname[0])
3030 			dev = __dev_get_by_name(net, ifname);
3031 		else
3032 			dev = NULL;
3033 	}
3034 
3035 	if (dev) {
3036 		master_dev = netdev_master_upper_dev_get(dev);
3037 		if (master_dev)
3038 			m_ops = master_dev->rtnl_link_ops;
3039 	}
3040 
3041 	err = validate_linkmsg(dev, tb);
3042 	if (err < 0)
3043 		return err;
3044 
3045 	if (tb[IFLA_LINKINFO]) {
3046 		err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3047 						  tb[IFLA_LINKINFO],
3048 						  ifla_info_policy, NULL);
3049 		if (err < 0)
3050 			return err;
3051 	} else
3052 		memset(linkinfo, 0, sizeof(linkinfo));
3053 
3054 	if (linkinfo[IFLA_INFO_KIND]) {
3055 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3056 		ops = rtnl_link_ops_get(kind);
3057 	} else {
3058 		kind[0] = '\0';
3059 		ops = NULL;
3060 	}
3061 
3062 	data = NULL;
3063 	if (ops) {
3064 		if (ops->maxtype > RTNL_MAX_TYPE)
3065 			return -EINVAL;
3066 
3067 		if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3068 			err = nla_parse_nested_deprecated(attr, ops->maxtype,
3069 							  linkinfo[IFLA_INFO_DATA],
3070 							  ops->policy, extack);
3071 			if (err < 0)
3072 				return err;
3073 			data = attr;
3074 		}
3075 		if (ops->validate) {
3076 			err = ops->validate(tb, data, extack);
3077 			if (err < 0)
3078 				return err;
3079 		}
3080 	}
3081 
3082 	slave_data = NULL;
3083 	if (m_ops) {
3084 		if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3085 			return -EINVAL;
3086 
3087 		if (m_ops->slave_maxtype &&
3088 		    linkinfo[IFLA_INFO_SLAVE_DATA]) {
3089 			err = nla_parse_nested_deprecated(slave_attr,
3090 							  m_ops->slave_maxtype,
3091 							  linkinfo[IFLA_INFO_SLAVE_DATA],
3092 							  m_ops->slave_policy,
3093 							  extack);
3094 			if (err < 0)
3095 				return err;
3096 			slave_data = slave_attr;
3097 		}
3098 	}
3099 
3100 	if (dev) {
3101 		int status = 0;
3102 
3103 		if (nlh->nlmsg_flags & NLM_F_EXCL)
3104 			return -EEXIST;
3105 		if (nlh->nlmsg_flags & NLM_F_REPLACE)
3106 			return -EOPNOTSUPP;
3107 
3108 		if (linkinfo[IFLA_INFO_DATA]) {
3109 			if (!ops || ops != dev->rtnl_link_ops ||
3110 			    !ops->changelink)
3111 				return -EOPNOTSUPP;
3112 
3113 			err = ops->changelink(dev, tb, data, extack);
3114 			if (err < 0)
3115 				return err;
3116 			status |= DO_SETLINK_NOTIFY;
3117 		}
3118 
3119 		if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3120 			if (!m_ops || !m_ops->slave_changelink)
3121 				return -EOPNOTSUPP;
3122 
3123 			err = m_ops->slave_changelink(master_dev, dev, tb,
3124 						      slave_data, extack);
3125 			if (err < 0)
3126 				return err;
3127 			status |= DO_SETLINK_NOTIFY;
3128 		}
3129 
3130 		return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3131 	}
3132 
3133 	if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3134 		if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3135 			return rtnl_group_changelink(skb, net,
3136 						nla_get_u32(tb[IFLA_GROUP]),
3137 						ifm, extack, tb);
3138 		return -ENODEV;
3139 	}
3140 
3141 	if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3142 		return -EOPNOTSUPP;
3143 
3144 	if (!ops) {
3145 #ifdef CONFIG_MODULES
3146 		if (kind[0]) {
3147 			__rtnl_unlock();
3148 			request_module("rtnl-link-%s", kind);
3149 			rtnl_lock();
3150 			ops = rtnl_link_ops_get(kind);
3151 			if (ops)
3152 				goto replay;
3153 		}
3154 #endif
3155 		NL_SET_ERR_MSG(extack, "Unknown device type");
3156 		return -EOPNOTSUPP;
3157 	}
3158 
3159 	if (!ops->setup)
3160 		return -EOPNOTSUPP;
3161 
3162 	if (!ifname[0]) {
3163 		snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3164 		name_assign_type = NET_NAME_ENUM;
3165 	}
3166 
3167 	dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3168 	if (IS_ERR(dest_net))
3169 		return PTR_ERR(dest_net);
3170 
3171 	if (tb[IFLA_LINK_NETNSID]) {
3172 		int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3173 
3174 		link_net = get_net_ns_by_id(dest_net, id);
3175 		if (!link_net) {
3176 			NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3177 			err =  -EINVAL;
3178 			goto out;
3179 		}
3180 		err = -EPERM;
3181 		if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3182 			goto out;
3183 	} else {
3184 		link_net = NULL;
3185 	}
3186 
3187 	dev = rtnl_create_link(link_net ? : dest_net, ifname,
3188 			       name_assign_type, ops, tb, extack);
3189 	if (IS_ERR(dev)) {
3190 		err = PTR_ERR(dev);
3191 		goto out;
3192 	}
3193 
3194 	dev->ifindex = ifm->ifi_index;
3195 
3196 	if (ops->newlink) {
3197 		err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3198 		/* Drivers should call free_netdev() in ->destructor
3199 		 * and unregister it on failure after registration
3200 		 * so that device could be finally freed in rtnl_unlock.
3201 		 */
3202 		if (err < 0) {
3203 			/* If device is not registered at all, free it now */
3204 			if (dev->reg_state == NETREG_UNINITIALIZED)
3205 				free_netdev(dev);
3206 			goto out;
3207 		}
3208 	} else {
3209 		err = register_netdevice(dev);
3210 		if (err < 0) {
3211 			free_netdev(dev);
3212 			goto out;
3213 		}
3214 	}
3215 	err = rtnl_configure_link(dev, ifm);
3216 	if (err < 0)
3217 		goto out_unregister;
3218 	if (link_net) {
3219 		err = dev_change_net_namespace(dev, dest_net, ifname);
3220 		if (err < 0)
3221 			goto out_unregister;
3222 	}
3223 	if (tb[IFLA_MASTER]) {
3224 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3225 		if (err)
3226 			goto out_unregister;
3227 	}
3228 out:
3229 	if (link_net)
3230 		put_net(link_net);
3231 	put_net(dest_net);
3232 	return err;
3233 out_unregister:
3234 	if (ops->newlink) {
3235 		LIST_HEAD(list_kill);
3236 
3237 		ops->dellink(dev, &list_kill);
3238 		unregister_netdevice_many(&list_kill);
3239 	} else {
3240 		unregister_netdevice(dev);
3241 	}
3242 	goto out;
3243 }
3244 
3245 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3246 			struct netlink_ext_ack *extack)
3247 {
3248 	struct nlattr **attr;
3249 	int ret;
3250 
3251 	attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3252 	if (!attr)
3253 		return -ENOMEM;
3254 
3255 	ret = __rtnl_newlink(skb, nlh, attr, extack);
3256 	kfree(attr);
3257 	return ret;
3258 }
3259 
3260 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3261 				  const struct nlmsghdr *nlh,
3262 				  struct nlattr **tb,
3263 				  struct netlink_ext_ack *extack)
3264 {
3265 	struct ifinfomsg *ifm;
3266 	int i, err;
3267 
3268 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3269 		NL_SET_ERR_MSG(extack, "Invalid header for get link");
3270 		return -EINVAL;
3271 	}
3272 
3273 	if (!netlink_strict_get_check(skb))
3274 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3275 					      ifla_policy, extack);
3276 
3277 	ifm = nlmsg_data(nlh);
3278 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3279 	    ifm->ifi_change) {
3280 		NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3281 		return -EINVAL;
3282 	}
3283 
3284 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3285 					    ifla_policy, extack);
3286 	if (err)
3287 		return err;
3288 
3289 	for (i = 0; i <= IFLA_MAX; i++) {
3290 		if (!tb[i])
3291 			continue;
3292 
3293 		switch (i) {
3294 		case IFLA_IFNAME:
3295 		case IFLA_EXT_MASK:
3296 		case IFLA_TARGET_NETNSID:
3297 			break;
3298 		default:
3299 			NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3300 			return -EINVAL;
3301 		}
3302 	}
3303 
3304 	return 0;
3305 }
3306 
3307 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3308 			struct netlink_ext_ack *extack)
3309 {
3310 	struct net *net = sock_net(skb->sk);
3311 	struct net *tgt_net = net;
3312 	struct ifinfomsg *ifm;
3313 	char ifname[IFNAMSIZ];
3314 	struct nlattr *tb[IFLA_MAX+1];
3315 	struct net_device *dev = NULL;
3316 	struct sk_buff *nskb;
3317 	int netnsid = -1;
3318 	int err;
3319 	u32 ext_filter_mask = 0;
3320 
3321 	err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3322 	if (err < 0)
3323 		return err;
3324 
3325 	err = rtnl_ensure_unique_netns(tb, extack, true);
3326 	if (err < 0)
3327 		return err;
3328 
3329 	if (tb[IFLA_TARGET_NETNSID]) {
3330 		netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3331 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3332 		if (IS_ERR(tgt_net))
3333 			return PTR_ERR(tgt_net);
3334 	}
3335 
3336 	if (tb[IFLA_IFNAME])
3337 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3338 
3339 	if (tb[IFLA_EXT_MASK])
3340 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3341 
3342 	err = -EINVAL;
3343 	ifm = nlmsg_data(nlh);
3344 	if (ifm->ifi_index > 0)
3345 		dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3346 	else if (tb[IFLA_IFNAME])
3347 		dev = __dev_get_by_name(tgt_net, ifname);
3348 	else
3349 		goto out;
3350 
3351 	err = -ENODEV;
3352 	if (dev == NULL)
3353 		goto out;
3354 
3355 	err = -ENOBUFS;
3356 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3357 	if (nskb == NULL)
3358 		goto out;
3359 
3360 	err = rtnl_fill_ifinfo(nskb, dev, net,
3361 			       RTM_NEWLINK, NETLINK_CB(skb).portid,
3362 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3363 			       0, NULL, 0, netnsid, GFP_KERNEL);
3364 	if (err < 0) {
3365 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
3366 		WARN_ON(err == -EMSGSIZE);
3367 		kfree_skb(nskb);
3368 	} else
3369 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3370 out:
3371 	if (netnsid >= 0)
3372 		put_net(tgt_net);
3373 
3374 	return err;
3375 }
3376 
3377 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3378 {
3379 	struct net *net = sock_net(skb->sk);
3380 	struct net_device *dev;
3381 	struct nlattr *tb[IFLA_MAX+1];
3382 	u32 ext_filter_mask = 0;
3383 	u16 min_ifinfo_dump_size = 0;
3384 	int hdrlen;
3385 
3386 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3387 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3388 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3389 
3390 	if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3391 		if (tb[IFLA_EXT_MASK])
3392 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3393 	}
3394 
3395 	if (!ext_filter_mask)
3396 		return NLMSG_GOODSIZE;
3397 	/*
3398 	 * traverse the list of net devices and compute the minimum
3399 	 * buffer size based upon the filter mask.
3400 	 */
3401 	rcu_read_lock();
3402 	for_each_netdev_rcu(net, dev) {
3403 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3404 					     if_nlmsg_size(dev,
3405 						           ext_filter_mask));
3406 	}
3407 	rcu_read_unlock();
3408 
3409 	return nlmsg_total_size(min_ifinfo_dump_size);
3410 }
3411 
3412 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3413 {
3414 	int idx;
3415 	int s_idx = cb->family;
3416 	int type = cb->nlh->nlmsg_type - RTM_BASE;
3417 	int ret = 0;
3418 
3419 	if (s_idx == 0)
3420 		s_idx = 1;
3421 
3422 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3423 		struct rtnl_link **tab;
3424 		struct rtnl_link *link;
3425 		rtnl_dumpit_func dumpit;
3426 
3427 		if (idx < s_idx || idx == PF_PACKET)
3428 			continue;
3429 
3430 		if (type < 0 || type >= RTM_NR_MSGTYPES)
3431 			continue;
3432 
3433 		tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3434 		if (!tab)
3435 			continue;
3436 
3437 		link = tab[type];
3438 		if (!link)
3439 			continue;
3440 
3441 		dumpit = link->dumpit;
3442 		if (!dumpit)
3443 			continue;
3444 
3445 		if (idx > s_idx) {
3446 			memset(&cb->args[0], 0, sizeof(cb->args));
3447 			cb->prev_seq = 0;
3448 			cb->seq = 0;
3449 		}
3450 		ret = dumpit(skb, cb);
3451 		if (ret)
3452 			break;
3453 	}
3454 	cb->family = idx;
3455 
3456 	return skb->len ? : ret;
3457 }
3458 
3459 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3460 				       unsigned int change,
3461 				       u32 event, gfp_t flags, int *new_nsid,
3462 				       int new_ifindex)
3463 {
3464 	struct net *net = dev_net(dev);
3465 	struct sk_buff *skb;
3466 	int err = -ENOBUFS;
3467 	size_t if_info_size;
3468 
3469 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3470 	if (skb == NULL)
3471 		goto errout;
3472 
3473 	err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3474 			       type, 0, 0, change, 0, 0, event,
3475 			       new_nsid, new_ifindex, -1, flags);
3476 	if (err < 0) {
3477 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
3478 		WARN_ON(err == -EMSGSIZE);
3479 		kfree_skb(skb);
3480 		goto errout;
3481 	}
3482 	return skb;
3483 errout:
3484 	if (err < 0)
3485 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3486 	return NULL;
3487 }
3488 
3489 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3490 {
3491 	struct net *net = dev_net(dev);
3492 
3493 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3494 }
3495 
3496 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3497 			       unsigned int change, u32 event,
3498 			       gfp_t flags, int *new_nsid, int new_ifindex)
3499 {
3500 	struct sk_buff *skb;
3501 
3502 	if (dev->reg_state != NETREG_REGISTERED)
3503 		return;
3504 
3505 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3506 				     new_ifindex);
3507 	if (skb)
3508 		rtmsg_ifinfo_send(skb, dev, flags);
3509 }
3510 
3511 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3512 		  gfp_t flags)
3513 {
3514 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3515 			   NULL, 0);
3516 }
3517 
3518 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3519 			 gfp_t flags, int *new_nsid, int new_ifindex)
3520 {
3521 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3522 			   new_nsid, new_ifindex);
3523 }
3524 
3525 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3526 				   struct net_device *dev,
3527 				   u8 *addr, u16 vid, u32 pid, u32 seq,
3528 				   int type, unsigned int flags,
3529 				   int nlflags, u16 ndm_state)
3530 {
3531 	struct nlmsghdr *nlh;
3532 	struct ndmsg *ndm;
3533 
3534 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3535 	if (!nlh)
3536 		return -EMSGSIZE;
3537 
3538 	ndm = nlmsg_data(nlh);
3539 	ndm->ndm_family  = AF_BRIDGE;
3540 	ndm->ndm_pad1	 = 0;
3541 	ndm->ndm_pad2    = 0;
3542 	ndm->ndm_flags	 = flags;
3543 	ndm->ndm_type	 = 0;
3544 	ndm->ndm_ifindex = dev->ifindex;
3545 	ndm->ndm_state   = ndm_state;
3546 
3547 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3548 		goto nla_put_failure;
3549 	if (vid)
3550 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3551 			goto nla_put_failure;
3552 
3553 	nlmsg_end(skb, nlh);
3554 	return 0;
3555 
3556 nla_put_failure:
3557 	nlmsg_cancel(skb, nlh);
3558 	return -EMSGSIZE;
3559 }
3560 
3561 static inline size_t rtnl_fdb_nlmsg_size(void)
3562 {
3563 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3564 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
3565 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
3566 	       0;
3567 }
3568 
3569 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3570 			    u16 ndm_state)
3571 {
3572 	struct net *net = dev_net(dev);
3573 	struct sk_buff *skb;
3574 	int err = -ENOBUFS;
3575 
3576 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3577 	if (!skb)
3578 		goto errout;
3579 
3580 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3581 				      0, 0, type, NTF_SELF, 0, ndm_state);
3582 	if (err < 0) {
3583 		kfree_skb(skb);
3584 		goto errout;
3585 	}
3586 
3587 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3588 	return;
3589 errout:
3590 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3591 }
3592 
3593 /*
3594  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3595  */
3596 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3597 		     struct nlattr *tb[],
3598 		     struct net_device *dev,
3599 		     const unsigned char *addr, u16 vid,
3600 		     u16 flags)
3601 {
3602 	int err = -EINVAL;
3603 
3604 	/* If aging addresses are supported device will need to
3605 	 * implement its own handler for this.
3606 	 */
3607 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3608 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3609 		return err;
3610 	}
3611 
3612 	if (vid) {
3613 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3614 		return err;
3615 	}
3616 
3617 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3618 		err = dev_uc_add_excl(dev, addr);
3619 	else if (is_multicast_ether_addr(addr))
3620 		err = dev_mc_add_excl(dev, addr);
3621 
3622 	/* Only return duplicate errors if NLM_F_EXCL is set */
3623 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3624 		err = 0;
3625 
3626 	return err;
3627 }
3628 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3629 
3630 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3631 			 struct netlink_ext_ack *extack)
3632 {
3633 	u16 vid = 0;
3634 
3635 	if (vlan_attr) {
3636 		if (nla_len(vlan_attr) != sizeof(u16)) {
3637 			NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3638 			return -EINVAL;
3639 		}
3640 
3641 		vid = nla_get_u16(vlan_attr);
3642 
3643 		if (!vid || vid >= VLAN_VID_MASK) {
3644 			NL_SET_ERR_MSG(extack, "invalid vlan id");
3645 			return -EINVAL;
3646 		}
3647 	}
3648 	*p_vid = vid;
3649 	return 0;
3650 }
3651 
3652 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3653 			struct netlink_ext_ack *extack)
3654 {
3655 	struct net *net = sock_net(skb->sk);
3656 	struct ndmsg *ndm;
3657 	struct nlattr *tb[NDA_MAX+1];
3658 	struct net_device *dev;
3659 	u8 *addr;
3660 	u16 vid;
3661 	int err;
3662 
3663 	err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3664 				     extack);
3665 	if (err < 0)
3666 		return err;
3667 
3668 	ndm = nlmsg_data(nlh);
3669 	if (ndm->ndm_ifindex == 0) {
3670 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3671 		return -EINVAL;
3672 	}
3673 
3674 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3675 	if (dev == NULL) {
3676 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3677 		return -ENODEV;
3678 	}
3679 
3680 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3681 		NL_SET_ERR_MSG(extack, "invalid address");
3682 		return -EINVAL;
3683 	}
3684 
3685 	if (dev->type != ARPHRD_ETHER) {
3686 		NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3687 		return -EINVAL;
3688 	}
3689 
3690 	addr = nla_data(tb[NDA_LLADDR]);
3691 
3692 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3693 	if (err)
3694 		return err;
3695 
3696 	err = -EOPNOTSUPP;
3697 
3698 	/* Support fdb on master device the net/bridge default case */
3699 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3700 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3701 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3702 		const struct net_device_ops *ops = br_dev->netdev_ops;
3703 
3704 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3705 				       nlh->nlmsg_flags, extack);
3706 		if (err)
3707 			goto out;
3708 		else
3709 			ndm->ndm_flags &= ~NTF_MASTER;
3710 	}
3711 
3712 	/* Embedded bridge, macvlan, and any other device support */
3713 	if ((ndm->ndm_flags & NTF_SELF)) {
3714 		if (dev->netdev_ops->ndo_fdb_add)
3715 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3716 							   vid,
3717 							   nlh->nlmsg_flags,
3718 							   extack);
3719 		else
3720 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3721 					       nlh->nlmsg_flags);
3722 
3723 		if (!err) {
3724 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3725 					ndm->ndm_state);
3726 			ndm->ndm_flags &= ~NTF_SELF;
3727 		}
3728 	}
3729 out:
3730 	return err;
3731 }
3732 
3733 /*
3734  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3735  */
3736 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3737 		     struct nlattr *tb[],
3738 		     struct net_device *dev,
3739 		     const unsigned char *addr, u16 vid)
3740 {
3741 	int err = -EINVAL;
3742 
3743 	/* If aging addresses are supported device will need to
3744 	 * implement its own handler for this.
3745 	 */
3746 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3747 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3748 		return err;
3749 	}
3750 
3751 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3752 		err = dev_uc_del(dev, addr);
3753 	else if (is_multicast_ether_addr(addr))
3754 		err = dev_mc_del(dev, addr);
3755 
3756 	return err;
3757 }
3758 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3759 
3760 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3761 			struct netlink_ext_ack *extack)
3762 {
3763 	struct net *net = sock_net(skb->sk);
3764 	struct ndmsg *ndm;
3765 	struct nlattr *tb[NDA_MAX+1];
3766 	struct net_device *dev;
3767 	int err = -EINVAL;
3768 	__u8 *addr;
3769 	u16 vid;
3770 
3771 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3772 		return -EPERM;
3773 
3774 	err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3775 				     extack);
3776 	if (err < 0)
3777 		return err;
3778 
3779 	ndm = nlmsg_data(nlh);
3780 	if (ndm->ndm_ifindex == 0) {
3781 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3782 		return -EINVAL;
3783 	}
3784 
3785 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3786 	if (dev == NULL) {
3787 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3788 		return -ENODEV;
3789 	}
3790 
3791 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3792 		NL_SET_ERR_MSG(extack, "invalid address");
3793 		return -EINVAL;
3794 	}
3795 
3796 	if (dev->type != ARPHRD_ETHER) {
3797 		NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
3798 		return -EINVAL;
3799 	}
3800 
3801 	addr = nla_data(tb[NDA_LLADDR]);
3802 
3803 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3804 	if (err)
3805 		return err;
3806 
3807 	err = -EOPNOTSUPP;
3808 
3809 	/* Support fdb on master device the net/bridge default case */
3810 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3811 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3812 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3813 		const struct net_device_ops *ops = br_dev->netdev_ops;
3814 
3815 		if (ops->ndo_fdb_del)
3816 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3817 
3818 		if (err)
3819 			goto out;
3820 		else
3821 			ndm->ndm_flags &= ~NTF_MASTER;
3822 	}
3823 
3824 	/* Embedded bridge, macvlan, and any other device support */
3825 	if (ndm->ndm_flags & NTF_SELF) {
3826 		if (dev->netdev_ops->ndo_fdb_del)
3827 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3828 							   vid);
3829 		else
3830 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3831 
3832 		if (!err) {
3833 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3834 					ndm->ndm_state);
3835 			ndm->ndm_flags &= ~NTF_SELF;
3836 		}
3837 	}
3838 out:
3839 	return err;
3840 }
3841 
3842 static int nlmsg_populate_fdb(struct sk_buff *skb,
3843 			      struct netlink_callback *cb,
3844 			      struct net_device *dev,
3845 			      int *idx,
3846 			      struct netdev_hw_addr_list *list)
3847 {
3848 	struct netdev_hw_addr *ha;
3849 	int err;
3850 	u32 portid, seq;
3851 
3852 	portid = NETLINK_CB(cb->skb).portid;
3853 	seq = cb->nlh->nlmsg_seq;
3854 
3855 	list_for_each_entry(ha, &list->list, list) {
3856 		if (*idx < cb->args[2])
3857 			goto skip;
3858 
3859 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3860 					      portid, seq,
3861 					      RTM_NEWNEIGH, NTF_SELF,
3862 					      NLM_F_MULTI, NUD_PERMANENT);
3863 		if (err < 0)
3864 			return err;
3865 skip:
3866 		*idx += 1;
3867 	}
3868 	return 0;
3869 }
3870 
3871 /**
3872  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3873  * @skb: socket buffer to store message in
3874  * @cb: netlink callback
3875  * @dev: netdevice
3876  * @filter_dev: ignored
3877  * @idx: the number of FDB table entries dumped is added to *@idx
3878  *
3879  * Default netdevice operation to dump the existing unicast address list.
3880  * Returns number of addresses from list put in skb.
3881  */
3882 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3883 		      struct netlink_callback *cb,
3884 		      struct net_device *dev,
3885 		      struct net_device *filter_dev,
3886 		      int *idx)
3887 {
3888 	int err;
3889 
3890 	if (dev->type != ARPHRD_ETHER)
3891 		return -EINVAL;
3892 
3893 	netif_addr_lock_bh(dev);
3894 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3895 	if (err)
3896 		goto out;
3897 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3898 out:
3899 	netif_addr_unlock_bh(dev);
3900 	return err;
3901 }
3902 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3903 
3904 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
3905 				 int *br_idx, int *brport_idx,
3906 				 struct netlink_ext_ack *extack)
3907 {
3908 	struct nlattr *tb[NDA_MAX + 1];
3909 	struct ndmsg *ndm;
3910 	int err, i;
3911 
3912 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
3913 		NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
3914 		return -EINVAL;
3915 	}
3916 
3917 	ndm = nlmsg_data(nlh);
3918 	if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
3919 	    ndm->ndm_flags || ndm->ndm_type) {
3920 		NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
3921 		return -EINVAL;
3922 	}
3923 
3924 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
3925 					    NDA_MAX, NULL, extack);
3926 	if (err < 0)
3927 		return err;
3928 
3929 	*brport_idx = ndm->ndm_ifindex;
3930 	for (i = 0; i <= NDA_MAX; ++i) {
3931 		if (!tb[i])
3932 			continue;
3933 
3934 		switch (i) {
3935 		case NDA_IFINDEX:
3936 			if (nla_len(tb[i]) != sizeof(u32)) {
3937 				NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
3938 				return -EINVAL;
3939 			}
3940 			*brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
3941 			break;
3942 		case NDA_MASTER:
3943 			if (nla_len(tb[i]) != sizeof(u32)) {
3944 				NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
3945 				return -EINVAL;
3946 			}
3947 			*br_idx = nla_get_u32(tb[NDA_MASTER]);
3948 			break;
3949 		default:
3950 			NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
3951 			return -EINVAL;
3952 		}
3953 	}
3954 
3955 	return 0;
3956 }
3957 
3958 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
3959 				 int *br_idx, int *brport_idx,
3960 				 struct netlink_ext_ack *extack)
3961 {
3962 	struct nlattr *tb[IFLA_MAX+1];
3963 	int err;
3964 
3965 	/* A hack to preserve kernel<->userspace interface.
3966 	 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
3967 	 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
3968 	 * So, check for ndmsg with an optional u32 attribute (not used here).
3969 	 * Fortunately these sizes don't conflict with the size of ifinfomsg
3970 	 * with an optional attribute.
3971 	 */
3972 	if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
3973 	    (nlmsg_len(nlh) != sizeof(struct ndmsg) +
3974 	     nla_attr_size(sizeof(u32)))) {
3975 		struct ifinfomsg *ifm;
3976 
3977 		err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
3978 					     tb, IFLA_MAX, ifla_policy,
3979 					     extack);
3980 		if (err < 0) {
3981 			return -EINVAL;
3982 		} else if (err == 0) {
3983 			if (tb[IFLA_MASTER])
3984 				*br_idx = nla_get_u32(tb[IFLA_MASTER]);
3985 		}
3986 
3987 		ifm = nlmsg_data(nlh);
3988 		*brport_idx = ifm->ifi_index;
3989 	}
3990 	return 0;
3991 }
3992 
3993 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3994 {
3995 	struct net_device *dev;
3996 	struct net_device *br_dev = NULL;
3997 	const struct net_device_ops *ops = NULL;
3998 	const struct net_device_ops *cops = NULL;
3999 	struct net *net = sock_net(skb->sk);
4000 	struct hlist_head *head;
4001 	int brport_idx = 0;
4002 	int br_idx = 0;
4003 	int h, s_h;
4004 	int idx = 0, s_idx;
4005 	int err = 0;
4006 	int fidx = 0;
4007 
4008 	if (cb->strict_check)
4009 		err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4010 					    cb->extack);
4011 	else
4012 		err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4013 					    cb->extack);
4014 	if (err < 0)
4015 		return err;
4016 
4017 	if (br_idx) {
4018 		br_dev = __dev_get_by_index(net, br_idx);
4019 		if (!br_dev)
4020 			return -ENODEV;
4021 
4022 		ops = br_dev->netdev_ops;
4023 	}
4024 
4025 	s_h = cb->args[0];
4026 	s_idx = cb->args[1];
4027 
4028 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4029 		idx = 0;
4030 		head = &net->dev_index_head[h];
4031 		hlist_for_each_entry(dev, head, index_hlist) {
4032 
4033 			if (brport_idx && (dev->ifindex != brport_idx))
4034 				continue;
4035 
4036 			if (!br_idx) { /* user did not specify a specific bridge */
4037 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
4038 					br_dev = netdev_master_upper_dev_get(dev);
4039 					cops = br_dev->netdev_ops;
4040 				}
4041 			} else {
4042 				if (dev != br_dev &&
4043 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
4044 					continue;
4045 
4046 				if (br_dev != netdev_master_upper_dev_get(dev) &&
4047 				    !(dev->priv_flags & IFF_EBRIDGE))
4048 					continue;
4049 				cops = ops;
4050 			}
4051 
4052 			if (idx < s_idx)
4053 				goto cont;
4054 
4055 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
4056 				if (cops && cops->ndo_fdb_dump) {
4057 					err = cops->ndo_fdb_dump(skb, cb,
4058 								br_dev, dev,
4059 								&fidx);
4060 					if (err == -EMSGSIZE)
4061 						goto out;
4062 				}
4063 			}
4064 
4065 			if (dev->netdev_ops->ndo_fdb_dump)
4066 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4067 								    dev, NULL,
4068 								    &fidx);
4069 			else
4070 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4071 							&fidx);
4072 			if (err == -EMSGSIZE)
4073 				goto out;
4074 
4075 			cops = NULL;
4076 
4077 			/* reset fdb offset to 0 for rest of the interfaces */
4078 			cb->args[2] = 0;
4079 			fidx = 0;
4080 cont:
4081 			idx++;
4082 		}
4083 	}
4084 
4085 out:
4086 	cb->args[0] = h;
4087 	cb->args[1] = idx;
4088 	cb->args[2] = fidx;
4089 
4090 	return skb->len;
4091 }
4092 
4093 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4094 				struct nlattr **tb, u8 *ndm_flags,
4095 				int *br_idx, int *brport_idx, u8 **addr,
4096 				u16 *vid, struct netlink_ext_ack *extack)
4097 {
4098 	struct ndmsg *ndm;
4099 	int err, i;
4100 
4101 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4102 		NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4103 		return -EINVAL;
4104 	}
4105 
4106 	ndm = nlmsg_data(nlh);
4107 	if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4108 	    ndm->ndm_type) {
4109 		NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4110 		return -EINVAL;
4111 	}
4112 
4113 	if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4114 		NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4115 		return -EINVAL;
4116 	}
4117 
4118 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4119 					    NDA_MAX, nda_policy, extack);
4120 	if (err < 0)
4121 		return err;
4122 
4123 	*ndm_flags = ndm->ndm_flags;
4124 	*brport_idx = ndm->ndm_ifindex;
4125 	for (i = 0; i <= NDA_MAX; ++i) {
4126 		if (!tb[i])
4127 			continue;
4128 
4129 		switch (i) {
4130 		case NDA_MASTER:
4131 			*br_idx = nla_get_u32(tb[i]);
4132 			break;
4133 		case NDA_LLADDR:
4134 			if (nla_len(tb[i]) != ETH_ALEN) {
4135 				NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4136 				return -EINVAL;
4137 			}
4138 			*addr = nla_data(tb[i]);
4139 			break;
4140 		case NDA_VLAN:
4141 			err = fdb_vid_parse(tb[i], vid, extack);
4142 			if (err)
4143 				return err;
4144 			break;
4145 		case NDA_VNI:
4146 			break;
4147 		default:
4148 			NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4149 			return -EINVAL;
4150 		}
4151 	}
4152 
4153 	return 0;
4154 }
4155 
4156 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4157 			struct netlink_ext_ack *extack)
4158 {
4159 	struct net_device *dev = NULL, *br_dev = NULL;
4160 	const struct net_device_ops *ops = NULL;
4161 	struct net *net = sock_net(in_skb->sk);
4162 	struct nlattr *tb[NDA_MAX + 1];
4163 	struct sk_buff *skb;
4164 	int brport_idx = 0;
4165 	u8 ndm_flags = 0;
4166 	int br_idx = 0;
4167 	u8 *addr = NULL;
4168 	u16 vid = 0;
4169 	int err;
4170 
4171 	err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4172 				   &brport_idx, &addr, &vid, extack);
4173 	if (err < 0)
4174 		return err;
4175 
4176 	if (!addr) {
4177 		NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4178 		return -EINVAL;
4179 	}
4180 
4181 	if (brport_idx) {
4182 		dev = __dev_get_by_index(net, brport_idx);
4183 		if (!dev) {
4184 			NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4185 			return -ENODEV;
4186 		}
4187 	}
4188 
4189 	if (br_idx) {
4190 		if (dev) {
4191 			NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4192 			return -EINVAL;
4193 		}
4194 
4195 		br_dev = __dev_get_by_index(net, br_idx);
4196 		if (!br_dev) {
4197 			NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4198 			return -EINVAL;
4199 		}
4200 		ops = br_dev->netdev_ops;
4201 	}
4202 
4203 	if (dev) {
4204 		if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4205 			if (!(dev->priv_flags & IFF_BRIDGE_PORT)) {
4206 				NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4207 				return -EINVAL;
4208 			}
4209 			br_dev = netdev_master_upper_dev_get(dev);
4210 			if (!br_dev) {
4211 				NL_SET_ERR_MSG(extack, "Master of device not found");
4212 				return -EINVAL;
4213 			}
4214 			ops = br_dev->netdev_ops;
4215 		} else {
4216 			if (!(ndm_flags & NTF_SELF)) {
4217 				NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4218 				return -EINVAL;
4219 			}
4220 			ops = dev->netdev_ops;
4221 		}
4222 	}
4223 
4224 	if (!br_dev && !dev) {
4225 		NL_SET_ERR_MSG(extack, "No device specified");
4226 		return -ENODEV;
4227 	}
4228 
4229 	if (!ops || !ops->ndo_fdb_get) {
4230 		NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4231 		return -EOPNOTSUPP;
4232 	}
4233 
4234 	skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4235 	if (!skb)
4236 		return -ENOBUFS;
4237 
4238 	if (br_dev)
4239 		dev = br_dev;
4240 	err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4241 			       NETLINK_CB(in_skb).portid,
4242 			       nlh->nlmsg_seq, extack);
4243 	if (err)
4244 		goto out;
4245 
4246 	return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4247 out:
4248 	kfree_skb(skb);
4249 	return err;
4250 }
4251 
4252 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4253 			       unsigned int attrnum, unsigned int flag)
4254 {
4255 	if (mask & flag)
4256 		return nla_put_u8(skb, attrnum, !!(flags & flag));
4257 	return 0;
4258 }
4259 
4260 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4261 			    struct net_device *dev, u16 mode,
4262 			    u32 flags, u32 mask, int nlflags,
4263 			    u32 filter_mask,
4264 			    int (*vlan_fill)(struct sk_buff *skb,
4265 					     struct net_device *dev,
4266 					     u32 filter_mask))
4267 {
4268 	struct nlmsghdr *nlh;
4269 	struct ifinfomsg *ifm;
4270 	struct nlattr *br_afspec;
4271 	struct nlattr *protinfo;
4272 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4273 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4274 	int err = 0;
4275 
4276 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4277 	if (nlh == NULL)
4278 		return -EMSGSIZE;
4279 
4280 	ifm = nlmsg_data(nlh);
4281 	ifm->ifi_family = AF_BRIDGE;
4282 	ifm->__ifi_pad = 0;
4283 	ifm->ifi_type = dev->type;
4284 	ifm->ifi_index = dev->ifindex;
4285 	ifm->ifi_flags = dev_get_flags(dev);
4286 	ifm->ifi_change = 0;
4287 
4288 
4289 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4290 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4291 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4292 	    (br_dev &&
4293 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4294 	    (dev->addr_len &&
4295 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4296 	    (dev->ifindex != dev_get_iflink(dev) &&
4297 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4298 		goto nla_put_failure;
4299 
4300 	br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4301 	if (!br_afspec)
4302 		goto nla_put_failure;
4303 
4304 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4305 		nla_nest_cancel(skb, br_afspec);
4306 		goto nla_put_failure;
4307 	}
4308 
4309 	if (mode != BRIDGE_MODE_UNDEF) {
4310 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4311 			nla_nest_cancel(skb, br_afspec);
4312 			goto nla_put_failure;
4313 		}
4314 	}
4315 	if (vlan_fill) {
4316 		err = vlan_fill(skb, dev, filter_mask);
4317 		if (err) {
4318 			nla_nest_cancel(skb, br_afspec);
4319 			goto nla_put_failure;
4320 		}
4321 	}
4322 	nla_nest_end(skb, br_afspec);
4323 
4324 	protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4325 	if (!protinfo)
4326 		goto nla_put_failure;
4327 
4328 	if (brport_nla_put_flag(skb, flags, mask,
4329 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4330 	    brport_nla_put_flag(skb, flags, mask,
4331 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4332 	    brport_nla_put_flag(skb, flags, mask,
4333 				IFLA_BRPORT_FAST_LEAVE,
4334 				BR_MULTICAST_FAST_LEAVE) ||
4335 	    brport_nla_put_flag(skb, flags, mask,
4336 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4337 	    brport_nla_put_flag(skb, flags, mask,
4338 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4339 	    brport_nla_put_flag(skb, flags, mask,
4340 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4341 	    brport_nla_put_flag(skb, flags, mask,
4342 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4343 	    brport_nla_put_flag(skb, flags, mask,
4344 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
4345 		nla_nest_cancel(skb, protinfo);
4346 		goto nla_put_failure;
4347 	}
4348 
4349 	nla_nest_end(skb, protinfo);
4350 
4351 	nlmsg_end(skb, nlh);
4352 	return 0;
4353 nla_put_failure:
4354 	nlmsg_cancel(skb, nlh);
4355 	return err ? err : -EMSGSIZE;
4356 }
4357 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4358 
4359 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4360 				    bool strict_check, u32 *filter_mask,
4361 				    struct netlink_ext_ack *extack)
4362 {
4363 	struct nlattr *tb[IFLA_MAX+1];
4364 	int err, i;
4365 
4366 	if (strict_check) {
4367 		struct ifinfomsg *ifm;
4368 
4369 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4370 			NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4371 			return -EINVAL;
4372 		}
4373 
4374 		ifm = nlmsg_data(nlh);
4375 		if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4376 		    ifm->ifi_change || ifm->ifi_index) {
4377 			NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4378 			return -EINVAL;
4379 		}
4380 
4381 		err = nlmsg_parse_deprecated_strict(nlh,
4382 						    sizeof(struct ifinfomsg),
4383 						    tb, IFLA_MAX, ifla_policy,
4384 						    extack);
4385 	} else {
4386 		err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4387 					     tb, IFLA_MAX, ifla_policy,
4388 					     extack);
4389 	}
4390 	if (err < 0)
4391 		return err;
4392 
4393 	/* new attributes should only be added with strict checking */
4394 	for (i = 0; i <= IFLA_MAX; ++i) {
4395 		if (!tb[i])
4396 			continue;
4397 
4398 		switch (i) {
4399 		case IFLA_EXT_MASK:
4400 			*filter_mask = nla_get_u32(tb[i]);
4401 			break;
4402 		default:
4403 			if (strict_check) {
4404 				NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4405 				return -EINVAL;
4406 			}
4407 		}
4408 	}
4409 
4410 	return 0;
4411 }
4412 
4413 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4414 {
4415 	const struct nlmsghdr *nlh = cb->nlh;
4416 	struct net *net = sock_net(skb->sk);
4417 	struct net_device *dev;
4418 	int idx = 0;
4419 	u32 portid = NETLINK_CB(cb->skb).portid;
4420 	u32 seq = nlh->nlmsg_seq;
4421 	u32 filter_mask = 0;
4422 	int err;
4423 
4424 	err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4425 				       cb->extack);
4426 	if (err < 0 && cb->strict_check)
4427 		return err;
4428 
4429 	rcu_read_lock();
4430 	for_each_netdev_rcu(net, dev) {
4431 		const struct net_device_ops *ops = dev->netdev_ops;
4432 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4433 
4434 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4435 			if (idx >= cb->args[0]) {
4436 				err = br_dev->netdev_ops->ndo_bridge_getlink(
4437 						skb, portid, seq, dev,
4438 						filter_mask, NLM_F_MULTI);
4439 				if (err < 0 && err != -EOPNOTSUPP) {
4440 					if (likely(skb->len))
4441 						break;
4442 
4443 					goto out_err;
4444 				}
4445 			}
4446 			idx++;
4447 		}
4448 
4449 		if (ops->ndo_bridge_getlink) {
4450 			if (idx >= cb->args[0]) {
4451 				err = ops->ndo_bridge_getlink(skb, portid,
4452 							      seq, dev,
4453 							      filter_mask,
4454 							      NLM_F_MULTI);
4455 				if (err < 0 && err != -EOPNOTSUPP) {
4456 					if (likely(skb->len))
4457 						break;
4458 
4459 					goto out_err;
4460 				}
4461 			}
4462 			idx++;
4463 		}
4464 	}
4465 	err = skb->len;
4466 out_err:
4467 	rcu_read_unlock();
4468 	cb->args[0] = idx;
4469 
4470 	return err;
4471 }
4472 
4473 static inline size_t bridge_nlmsg_size(void)
4474 {
4475 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4476 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
4477 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
4478 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
4479 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
4480 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
4481 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
4482 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
4483 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
4484 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
4485 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
4486 }
4487 
4488 static int rtnl_bridge_notify(struct net_device *dev)
4489 {
4490 	struct net *net = dev_net(dev);
4491 	struct sk_buff *skb;
4492 	int err = -EOPNOTSUPP;
4493 
4494 	if (!dev->netdev_ops->ndo_bridge_getlink)
4495 		return 0;
4496 
4497 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4498 	if (!skb) {
4499 		err = -ENOMEM;
4500 		goto errout;
4501 	}
4502 
4503 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4504 	if (err < 0)
4505 		goto errout;
4506 
4507 	if (!skb->len)
4508 		goto errout;
4509 
4510 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4511 	return 0;
4512 errout:
4513 	WARN_ON(err == -EMSGSIZE);
4514 	kfree_skb(skb);
4515 	if (err)
4516 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4517 	return err;
4518 }
4519 
4520 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4521 			       struct netlink_ext_ack *extack)
4522 {
4523 	struct net *net = sock_net(skb->sk);
4524 	struct ifinfomsg *ifm;
4525 	struct net_device *dev;
4526 	struct nlattr *br_spec, *attr = NULL;
4527 	int rem, err = -EOPNOTSUPP;
4528 	u16 flags = 0;
4529 	bool have_flags = false;
4530 
4531 	if (nlmsg_len(nlh) < sizeof(*ifm))
4532 		return -EINVAL;
4533 
4534 	ifm = nlmsg_data(nlh);
4535 	if (ifm->ifi_family != AF_BRIDGE)
4536 		return -EPFNOSUPPORT;
4537 
4538 	dev = __dev_get_by_index(net, ifm->ifi_index);
4539 	if (!dev) {
4540 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4541 		return -ENODEV;
4542 	}
4543 
4544 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4545 	if (br_spec) {
4546 		nla_for_each_nested(attr, br_spec, rem) {
4547 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4548 				if (nla_len(attr) < sizeof(flags))
4549 					return -EINVAL;
4550 
4551 				have_flags = true;
4552 				flags = nla_get_u16(attr);
4553 				break;
4554 			}
4555 		}
4556 	}
4557 
4558 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4559 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4560 
4561 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4562 			err = -EOPNOTSUPP;
4563 			goto out;
4564 		}
4565 
4566 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4567 							     extack);
4568 		if (err)
4569 			goto out;
4570 
4571 		flags &= ~BRIDGE_FLAGS_MASTER;
4572 	}
4573 
4574 	if ((flags & BRIDGE_FLAGS_SELF)) {
4575 		if (!dev->netdev_ops->ndo_bridge_setlink)
4576 			err = -EOPNOTSUPP;
4577 		else
4578 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4579 								  flags,
4580 								  extack);
4581 		if (!err) {
4582 			flags &= ~BRIDGE_FLAGS_SELF;
4583 
4584 			/* Generate event to notify upper layer of bridge
4585 			 * change
4586 			 */
4587 			err = rtnl_bridge_notify(dev);
4588 		}
4589 	}
4590 
4591 	if (have_flags)
4592 		memcpy(nla_data(attr), &flags, sizeof(flags));
4593 out:
4594 	return err;
4595 }
4596 
4597 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4598 			       struct netlink_ext_ack *extack)
4599 {
4600 	struct net *net = sock_net(skb->sk);
4601 	struct ifinfomsg *ifm;
4602 	struct net_device *dev;
4603 	struct nlattr *br_spec, *attr = NULL;
4604 	int rem, err = -EOPNOTSUPP;
4605 	u16 flags = 0;
4606 	bool have_flags = false;
4607 
4608 	if (nlmsg_len(nlh) < sizeof(*ifm))
4609 		return -EINVAL;
4610 
4611 	ifm = nlmsg_data(nlh);
4612 	if (ifm->ifi_family != AF_BRIDGE)
4613 		return -EPFNOSUPPORT;
4614 
4615 	dev = __dev_get_by_index(net, ifm->ifi_index);
4616 	if (!dev) {
4617 		NL_SET_ERR_MSG(extack, "unknown ifindex");
4618 		return -ENODEV;
4619 	}
4620 
4621 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4622 	if (br_spec) {
4623 		nla_for_each_nested(attr, br_spec, rem) {
4624 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4625 				if (nla_len(attr) < sizeof(flags))
4626 					return -EINVAL;
4627 
4628 				have_flags = true;
4629 				flags = nla_get_u16(attr);
4630 				break;
4631 			}
4632 		}
4633 	}
4634 
4635 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4636 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4637 
4638 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4639 			err = -EOPNOTSUPP;
4640 			goto out;
4641 		}
4642 
4643 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4644 		if (err)
4645 			goto out;
4646 
4647 		flags &= ~BRIDGE_FLAGS_MASTER;
4648 	}
4649 
4650 	if ((flags & BRIDGE_FLAGS_SELF)) {
4651 		if (!dev->netdev_ops->ndo_bridge_dellink)
4652 			err = -EOPNOTSUPP;
4653 		else
4654 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4655 								  flags);
4656 
4657 		if (!err) {
4658 			flags &= ~BRIDGE_FLAGS_SELF;
4659 
4660 			/* Generate event to notify upper layer of bridge
4661 			 * change
4662 			 */
4663 			err = rtnl_bridge_notify(dev);
4664 		}
4665 	}
4666 
4667 	if (have_flags)
4668 		memcpy(nla_data(attr), &flags, sizeof(flags));
4669 out:
4670 	return err;
4671 }
4672 
4673 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4674 {
4675 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4676 	       (!idxattr || idxattr == attrid);
4677 }
4678 
4679 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4680 static int rtnl_get_offload_stats_attr_size(int attr_id)
4681 {
4682 	switch (attr_id) {
4683 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4684 		return sizeof(struct rtnl_link_stats64);
4685 	}
4686 
4687 	return 0;
4688 }
4689 
4690 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4691 				  int *prividx)
4692 {
4693 	struct nlattr *attr = NULL;
4694 	int attr_id, size;
4695 	void *attr_data;
4696 	int err;
4697 
4698 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4699 	      dev->netdev_ops->ndo_get_offload_stats))
4700 		return -ENODATA;
4701 
4702 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4703 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4704 		if (attr_id < *prividx)
4705 			continue;
4706 
4707 		size = rtnl_get_offload_stats_attr_size(attr_id);
4708 		if (!size)
4709 			continue;
4710 
4711 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4712 			continue;
4713 
4714 		attr = nla_reserve_64bit(skb, attr_id, size,
4715 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
4716 		if (!attr)
4717 			goto nla_put_failure;
4718 
4719 		attr_data = nla_data(attr);
4720 		memset(attr_data, 0, size);
4721 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4722 							     attr_data);
4723 		if (err)
4724 			goto get_offload_stats_failure;
4725 	}
4726 
4727 	if (!attr)
4728 		return -ENODATA;
4729 
4730 	*prividx = 0;
4731 	return 0;
4732 
4733 nla_put_failure:
4734 	err = -EMSGSIZE;
4735 get_offload_stats_failure:
4736 	*prividx = attr_id;
4737 	return err;
4738 }
4739 
4740 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4741 {
4742 	int nla_size = 0;
4743 	int attr_id;
4744 	int size;
4745 
4746 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4747 	      dev->netdev_ops->ndo_get_offload_stats))
4748 		return 0;
4749 
4750 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4751 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4752 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4753 			continue;
4754 		size = rtnl_get_offload_stats_attr_size(attr_id);
4755 		nla_size += nla_total_size_64bit(size);
4756 	}
4757 
4758 	if (nla_size != 0)
4759 		nla_size += nla_total_size(0);
4760 
4761 	return nla_size;
4762 }
4763 
4764 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4765 			       int type, u32 pid, u32 seq, u32 change,
4766 			       unsigned int flags, unsigned int filter_mask,
4767 			       int *idxattr, int *prividx)
4768 {
4769 	struct if_stats_msg *ifsm;
4770 	struct nlmsghdr *nlh;
4771 	struct nlattr *attr;
4772 	int s_prividx = *prividx;
4773 	int err;
4774 
4775 	ASSERT_RTNL();
4776 
4777 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4778 	if (!nlh)
4779 		return -EMSGSIZE;
4780 
4781 	ifsm = nlmsg_data(nlh);
4782 	ifsm->family = PF_UNSPEC;
4783 	ifsm->pad1 = 0;
4784 	ifsm->pad2 = 0;
4785 	ifsm->ifindex = dev->ifindex;
4786 	ifsm->filter_mask = filter_mask;
4787 
4788 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4789 		struct rtnl_link_stats64 *sp;
4790 
4791 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4792 					 sizeof(struct rtnl_link_stats64),
4793 					 IFLA_STATS_UNSPEC);
4794 		if (!attr)
4795 			goto nla_put_failure;
4796 
4797 		sp = nla_data(attr);
4798 		dev_get_stats(dev, sp);
4799 	}
4800 
4801 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4802 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4803 
4804 		if (ops && ops->fill_linkxstats) {
4805 			*idxattr = IFLA_STATS_LINK_XSTATS;
4806 			attr = nla_nest_start_noflag(skb,
4807 						     IFLA_STATS_LINK_XSTATS);
4808 			if (!attr)
4809 				goto nla_put_failure;
4810 
4811 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4812 			nla_nest_end(skb, attr);
4813 			if (err)
4814 				goto nla_put_failure;
4815 			*idxattr = 0;
4816 		}
4817 	}
4818 
4819 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4820 			     *idxattr)) {
4821 		const struct rtnl_link_ops *ops = NULL;
4822 		const struct net_device *master;
4823 
4824 		master = netdev_master_upper_dev_get(dev);
4825 		if (master)
4826 			ops = master->rtnl_link_ops;
4827 		if (ops && ops->fill_linkxstats) {
4828 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4829 			attr = nla_nest_start_noflag(skb,
4830 						     IFLA_STATS_LINK_XSTATS_SLAVE);
4831 			if (!attr)
4832 				goto nla_put_failure;
4833 
4834 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4835 			nla_nest_end(skb, attr);
4836 			if (err)
4837 				goto nla_put_failure;
4838 			*idxattr = 0;
4839 		}
4840 	}
4841 
4842 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4843 			     *idxattr)) {
4844 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4845 		attr = nla_nest_start_noflag(skb,
4846 					     IFLA_STATS_LINK_OFFLOAD_XSTATS);
4847 		if (!attr)
4848 			goto nla_put_failure;
4849 
4850 		err = rtnl_get_offload_stats(skb, dev, prividx);
4851 		if (err == -ENODATA)
4852 			nla_nest_cancel(skb, attr);
4853 		else
4854 			nla_nest_end(skb, attr);
4855 
4856 		if (err && err != -ENODATA)
4857 			goto nla_put_failure;
4858 		*idxattr = 0;
4859 	}
4860 
4861 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4862 		struct rtnl_af_ops *af_ops;
4863 
4864 		*idxattr = IFLA_STATS_AF_SPEC;
4865 		attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
4866 		if (!attr)
4867 			goto nla_put_failure;
4868 
4869 		rcu_read_lock();
4870 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4871 			if (af_ops->fill_stats_af) {
4872 				struct nlattr *af;
4873 				int err;
4874 
4875 				af = nla_nest_start_noflag(skb,
4876 							   af_ops->family);
4877 				if (!af) {
4878 					rcu_read_unlock();
4879 					goto nla_put_failure;
4880 				}
4881 				err = af_ops->fill_stats_af(skb, dev);
4882 
4883 				if (err == -ENODATA) {
4884 					nla_nest_cancel(skb, af);
4885 				} else if (err < 0) {
4886 					rcu_read_unlock();
4887 					goto nla_put_failure;
4888 				}
4889 
4890 				nla_nest_end(skb, af);
4891 			}
4892 		}
4893 		rcu_read_unlock();
4894 
4895 		nla_nest_end(skb, attr);
4896 
4897 		*idxattr = 0;
4898 	}
4899 
4900 	nlmsg_end(skb, nlh);
4901 
4902 	return 0;
4903 
4904 nla_put_failure:
4905 	/* not a multi message or no progress mean a real error */
4906 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4907 		nlmsg_cancel(skb, nlh);
4908 	else
4909 		nlmsg_end(skb, nlh);
4910 
4911 	return -EMSGSIZE;
4912 }
4913 
4914 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4915 				  u32 filter_mask)
4916 {
4917 	size_t size = 0;
4918 
4919 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4920 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4921 
4922 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4923 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4924 		int attr = IFLA_STATS_LINK_XSTATS;
4925 
4926 		if (ops && ops->get_linkxstats_size) {
4927 			size += nla_total_size(ops->get_linkxstats_size(dev,
4928 									attr));
4929 			/* for IFLA_STATS_LINK_XSTATS */
4930 			size += nla_total_size(0);
4931 		}
4932 	}
4933 
4934 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4935 		struct net_device *_dev = (struct net_device *)dev;
4936 		const struct rtnl_link_ops *ops = NULL;
4937 		const struct net_device *master;
4938 
4939 		/* netdev_master_upper_dev_get can't take const */
4940 		master = netdev_master_upper_dev_get(_dev);
4941 		if (master)
4942 			ops = master->rtnl_link_ops;
4943 		if (ops && ops->get_linkxstats_size) {
4944 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4945 
4946 			size += nla_total_size(ops->get_linkxstats_size(dev,
4947 									attr));
4948 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
4949 			size += nla_total_size(0);
4950 		}
4951 	}
4952 
4953 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4954 		size += rtnl_get_offload_stats_size(dev);
4955 
4956 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4957 		struct rtnl_af_ops *af_ops;
4958 
4959 		/* for IFLA_STATS_AF_SPEC */
4960 		size += nla_total_size(0);
4961 
4962 		rcu_read_lock();
4963 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4964 			if (af_ops->get_stats_af_size) {
4965 				size += nla_total_size(
4966 					af_ops->get_stats_af_size(dev));
4967 
4968 				/* for AF_* */
4969 				size += nla_total_size(0);
4970 			}
4971 		}
4972 		rcu_read_unlock();
4973 	}
4974 
4975 	return size;
4976 }
4977 
4978 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
4979 				bool is_dump, struct netlink_ext_ack *extack)
4980 {
4981 	struct if_stats_msg *ifsm;
4982 
4983 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
4984 		NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
4985 		return -EINVAL;
4986 	}
4987 
4988 	if (!strict_check)
4989 		return 0;
4990 
4991 	ifsm = nlmsg_data(nlh);
4992 
4993 	/* only requests using strict checks can pass data to influence
4994 	 * the dump. The legacy exception is filter_mask.
4995 	 */
4996 	if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
4997 		NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
4998 		return -EINVAL;
4999 	}
5000 	if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5001 		NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5002 		return -EINVAL;
5003 	}
5004 	if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5005 		NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5006 		return -EINVAL;
5007 	}
5008 
5009 	return 0;
5010 }
5011 
5012 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5013 			  struct netlink_ext_ack *extack)
5014 {
5015 	struct net *net = sock_net(skb->sk);
5016 	struct net_device *dev = NULL;
5017 	int idxattr = 0, prividx = 0;
5018 	struct if_stats_msg *ifsm;
5019 	struct sk_buff *nskb;
5020 	u32 filter_mask;
5021 	int err;
5022 
5023 	err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5024 				   false, extack);
5025 	if (err)
5026 		return err;
5027 
5028 	ifsm = nlmsg_data(nlh);
5029 	if (ifsm->ifindex > 0)
5030 		dev = __dev_get_by_index(net, ifsm->ifindex);
5031 	else
5032 		return -EINVAL;
5033 
5034 	if (!dev)
5035 		return -ENODEV;
5036 
5037 	filter_mask = ifsm->filter_mask;
5038 	if (!filter_mask)
5039 		return -EINVAL;
5040 
5041 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5042 	if (!nskb)
5043 		return -ENOBUFS;
5044 
5045 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5046 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5047 				  0, filter_mask, &idxattr, &prividx);
5048 	if (err < 0) {
5049 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5050 		WARN_ON(err == -EMSGSIZE);
5051 		kfree_skb(nskb);
5052 	} else {
5053 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5054 	}
5055 
5056 	return err;
5057 }
5058 
5059 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5060 {
5061 	struct netlink_ext_ack *extack = cb->extack;
5062 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
5063 	struct net *net = sock_net(skb->sk);
5064 	unsigned int flags = NLM_F_MULTI;
5065 	struct if_stats_msg *ifsm;
5066 	struct hlist_head *head;
5067 	struct net_device *dev;
5068 	u32 filter_mask = 0;
5069 	int idx = 0;
5070 
5071 	s_h = cb->args[0];
5072 	s_idx = cb->args[1];
5073 	s_idxattr = cb->args[2];
5074 	s_prividx = cb->args[3];
5075 
5076 	cb->seq = net->dev_base_seq;
5077 
5078 	err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5079 	if (err)
5080 		return err;
5081 
5082 	ifsm = nlmsg_data(cb->nlh);
5083 	filter_mask = ifsm->filter_mask;
5084 	if (!filter_mask) {
5085 		NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5086 		return -EINVAL;
5087 	}
5088 
5089 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5090 		idx = 0;
5091 		head = &net->dev_index_head[h];
5092 		hlist_for_each_entry(dev, head, index_hlist) {
5093 			if (idx < s_idx)
5094 				goto cont;
5095 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5096 						  NETLINK_CB(cb->skb).portid,
5097 						  cb->nlh->nlmsg_seq, 0,
5098 						  flags, filter_mask,
5099 						  &s_idxattr, &s_prividx);
5100 			/* If we ran out of room on the first message,
5101 			 * we're in trouble
5102 			 */
5103 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5104 
5105 			if (err < 0)
5106 				goto out;
5107 			s_prividx = 0;
5108 			s_idxattr = 0;
5109 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5110 cont:
5111 			idx++;
5112 		}
5113 	}
5114 out:
5115 	cb->args[3] = s_prividx;
5116 	cb->args[2] = s_idxattr;
5117 	cb->args[1] = idx;
5118 	cb->args[0] = h;
5119 
5120 	return skb->len;
5121 }
5122 
5123 /* Process one rtnetlink message. */
5124 
5125 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5126 			     struct netlink_ext_ack *extack)
5127 {
5128 	struct net *net = sock_net(skb->sk);
5129 	struct rtnl_link *link;
5130 	struct module *owner;
5131 	int err = -EOPNOTSUPP;
5132 	rtnl_doit_func doit;
5133 	unsigned int flags;
5134 	int kind;
5135 	int family;
5136 	int type;
5137 
5138 	type = nlh->nlmsg_type;
5139 	if (type > RTM_MAX)
5140 		return -EOPNOTSUPP;
5141 
5142 	type -= RTM_BASE;
5143 
5144 	/* All the messages must have at least 1 byte length */
5145 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5146 		return 0;
5147 
5148 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5149 	kind = type&3;
5150 
5151 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5152 		return -EPERM;
5153 
5154 	rcu_read_lock();
5155 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5156 		struct sock *rtnl;
5157 		rtnl_dumpit_func dumpit;
5158 		u16 min_dump_alloc = 0;
5159 
5160 		link = rtnl_get_link(family, type);
5161 		if (!link || !link->dumpit) {
5162 			family = PF_UNSPEC;
5163 			link = rtnl_get_link(family, type);
5164 			if (!link || !link->dumpit)
5165 				goto err_unlock;
5166 		}
5167 		owner = link->owner;
5168 		dumpit = link->dumpit;
5169 
5170 		if (type == RTM_GETLINK - RTM_BASE)
5171 			min_dump_alloc = rtnl_calcit(skb, nlh);
5172 
5173 		err = 0;
5174 		/* need to do this before rcu_read_unlock() */
5175 		if (!try_module_get(owner))
5176 			err = -EPROTONOSUPPORT;
5177 
5178 		rcu_read_unlock();
5179 
5180 		rtnl = net->rtnl;
5181 		if (err == 0) {
5182 			struct netlink_dump_control c = {
5183 				.dump		= dumpit,
5184 				.min_dump_alloc	= min_dump_alloc,
5185 				.module		= owner,
5186 			};
5187 			err = netlink_dump_start(rtnl, skb, nlh, &c);
5188 			/* netlink_dump_start() will keep a reference on
5189 			 * module if dump is still in progress.
5190 			 */
5191 			module_put(owner);
5192 		}
5193 		return err;
5194 	}
5195 
5196 	link = rtnl_get_link(family, type);
5197 	if (!link || !link->doit) {
5198 		family = PF_UNSPEC;
5199 		link = rtnl_get_link(PF_UNSPEC, type);
5200 		if (!link || !link->doit)
5201 			goto out_unlock;
5202 	}
5203 
5204 	owner = link->owner;
5205 	if (!try_module_get(owner)) {
5206 		err = -EPROTONOSUPPORT;
5207 		goto out_unlock;
5208 	}
5209 
5210 	flags = link->flags;
5211 	if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5212 		doit = link->doit;
5213 		rcu_read_unlock();
5214 		if (doit)
5215 			err = doit(skb, nlh, extack);
5216 		module_put(owner);
5217 		return err;
5218 	}
5219 	rcu_read_unlock();
5220 
5221 	rtnl_lock();
5222 	link = rtnl_get_link(family, type);
5223 	if (link && link->doit)
5224 		err = link->doit(skb, nlh, extack);
5225 	rtnl_unlock();
5226 
5227 	module_put(owner);
5228 
5229 	return err;
5230 
5231 out_unlock:
5232 	rcu_read_unlock();
5233 	return err;
5234 
5235 err_unlock:
5236 	rcu_read_unlock();
5237 	return -EOPNOTSUPP;
5238 }
5239 
5240 static void rtnetlink_rcv(struct sk_buff *skb)
5241 {
5242 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5243 }
5244 
5245 static int rtnetlink_bind(struct net *net, int group)
5246 {
5247 	switch (group) {
5248 	case RTNLGRP_IPV4_MROUTE_R:
5249 	case RTNLGRP_IPV6_MROUTE_R:
5250 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5251 			return -EPERM;
5252 		break;
5253 	}
5254 	return 0;
5255 }
5256 
5257 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5258 {
5259 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5260 
5261 	switch (event) {
5262 	case NETDEV_REBOOT:
5263 	case NETDEV_CHANGEMTU:
5264 	case NETDEV_CHANGEADDR:
5265 	case NETDEV_CHANGENAME:
5266 	case NETDEV_FEAT_CHANGE:
5267 	case NETDEV_BONDING_FAILOVER:
5268 	case NETDEV_POST_TYPE_CHANGE:
5269 	case NETDEV_NOTIFY_PEERS:
5270 	case NETDEV_CHANGEUPPER:
5271 	case NETDEV_RESEND_IGMP:
5272 	case NETDEV_CHANGEINFODATA:
5273 	case NETDEV_CHANGELOWERSTATE:
5274 	case NETDEV_CHANGE_TX_QUEUE_LEN:
5275 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5276 				   GFP_KERNEL, NULL, 0);
5277 		break;
5278 	default:
5279 		break;
5280 	}
5281 	return NOTIFY_DONE;
5282 }
5283 
5284 static struct notifier_block rtnetlink_dev_notifier = {
5285 	.notifier_call	= rtnetlink_event,
5286 };
5287 
5288 
5289 static int __net_init rtnetlink_net_init(struct net *net)
5290 {
5291 	struct sock *sk;
5292 	struct netlink_kernel_cfg cfg = {
5293 		.groups		= RTNLGRP_MAX,
5294 		.input		= rtnetlink_rcv,
5295 		.cb_mutex	= &rtnl_mutex,
5296 		.flags		= NL_CFG_F_NONROOT_RECV,
5297 		.bind		= rtnetlink_bind,
5298 	};
5299 
5300 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5301 	if (!sk)
5302 		return -ENOMEM;
5303 	net->rtnl = sk;
5304 	return 0;
5305 }
5306 
5307 static void __net_exit rtnetlink_net_exit(struct net *net)
5308 {
5309 	netlink_kernel_release(net->rtnl);
5310 	net->rtnl = NULL;
5311 }
5312 
5313 static struct pernet_operations rtnetlink_net_ops = {
5314 	.init = rtnetlink_net_init,
5315 	.exit = rtnetlink_net_exit,
5316 };
5317 
5318 void __init rtnetlink_init(void)
5319 {
5320 	if (register_pernet_subsys(&rtnetlink_net_ops))
5321 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
5322 
5323 	register_netdevice_notifier(&rtnetlink_dev_notifier);
5324 
5325 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5326 		      rtnl_dump_ifinfo, 0);
5327 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5328 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5329 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5330 
5331 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5332 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5333 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5334 
5335 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5336 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5337 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5338 
5339 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5340 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5341 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5342 
5343 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5344 		      0);
5345 }
5346