xref: /linux/net/core/rtnetlink.c (revision 83a37b3292f4aca799b355179ad6fbdd78a08e10)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  *
15  *	Fixes:
16  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17  */
18 
19 #include <linux/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44 
45 #include <linux/uaccess.h>
46 
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
50 #include <net/ip.h>
51 #include <net/protocol.h>
52 #include <net/arp.h>
53 #include <net/route.h>
54 #include <net/udp.h>
55 #include <net/tcp.h>
56 #include <net/sock.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
61 
62 struct rtnl_link {
63 	rtnl_doit_func		doit;
64 	rtnl_dumpit_func	dumpit;
65 	unsigned int		flags;
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 static struct sk_buff *defer_kfree_skb_list;
77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
78 {
79 	if (head && tail) {
80 		tail->next = defer_kfree_skb_list;
81 		defer_kfree_skb_list = head;
82 	}
83 }
84 EXPORT_SYMBOL(rtnl_kfree_skbs);
85 
86 void __rtnl_unlock(void)
87 {
88 	struct sk_buff *head = defer_kfree_skb_list;
89 
90 	defer_kfree_skb_list = NULL;
91 
92 	mutex_unlock(&rtnl_mutex);
93 
94 	while (head) {
95 		struct sk_buff *next = head->next;
96 
97 		kfree_skb(head);
98 		cond_resched();
99 		head = next;
100 	}
101 }
102 
103 void rtnl_unlock(void)
104 {
105 	/* This fellow will unlock it for us. */
106 	netdev_run_todo();
107 }
108 EXPORT_SYMBOL(rtnl_unlock);
109 
110 int rtnl_trylock(void)
111 {
112 	return mutex_trylock(&rtnl_mutex);
113 }
114 EXPORT_SYMBOL(rtnl_trylock);
115 
116 int rtnl_is_locked(void)
117 {
118 	return mutex_is_locked(&rtnl_mutex);
119 }
120 EXPORT_SYMBOL(rtnl_is_locked);
121 
122 #ifdef CONFIG_PROVE_LOCKING
123 bool lockdep_rtnl_is_held(void)
124 {
125 	return lockdep_is_held(&rtnl_mutex);
126 }
127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
129 
130 static struct rtnl_link __rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
131 static refcount_t rtnl_msg_handlers_ref[RTNL_FAMILY_MAX + 1];
132 
133 static inline int rtm_msgindex(int msgtype)
134 {
135 	int msgindex = msgtype - RTM_BASE;
136 
137 	/*
138 	 * msgindex < 0 implies someone tried to register a netlink
139 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
140 	 * the message type has not been added to linux/rtnetlink.h
141 	 */
142 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
143 
144 	return msgindex;
145 }
146 
147 /**
148  * __rtnl_register - Register a rtnetlink message type
149  * @protocol: Protocol family or PF_UNSPEC
150  * @msgtype: rtnetlink message type
151  * @doit: Function pointer called for each request message
152  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
153  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
154  *
155  * Registers the specified function pointers (at least one of them has
156  * to be non-NULL) to be called whenever a request message for the
157  * specified protocol family and message type is received.
158  *
159  * The special protocol family PF_UNSPEC may be used to define fallback
160  * function pointers for the case when no entry for the specific protocol
161  * family exists.
162  *
163  * Returns 0 on success or a negative error code.
164  */
165 int __rtnl_register(int protocol, int msgtype,
166 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
167 		    unsigned int flags)
168 {
169 	struct rtnl_link *tab;
170 	int msgindex;
171 
172 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
173 	msgindex = rtm_msgindex(msgtype);
174 
175 	tab = rcu_dereference_raw(rtnl_msg_handlers[protocol]);
176 	if (tab == NULL) {
177 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
178 		if (tab == NULL)
179 			return -ENOBUFS;
180 
181 		rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
182 	}
183 
184 	if (doit)
185 		tab[msgindex].doit = doit;
186 	if (dumpit)
187 		tab[msgindex].dumpit = dumpit;
188 	tab[msgindex].flags |= flags;
189 
190 	return 0;
191 }
192 EXPORT_SYMBOL_GPL(__rtnl_register);
193 
194 /**
195  * rtnl_register - Register a rtnetlink message type
196  *
197  * Identical to __rtnl_register() but panics on failure. This is useful
198  * as failure of this function is very unlikely, it can only happen due
199  * to lack of memory when allocating the chain to store all message
200  * handlers for a protocol. Meant for use in init functions where lack
201  * of memory implies no sense in continuing.
202  */
203 void rtnl_register(int protocol, int msgtype,
204 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
205 		   unsigned int flags)
206 {
207 	if (__rtnl_register(protocol, msgtype, doit, dumpit, flags) < 0)
208 		panic("Unable to register rtnetlink message handler, "
209 		      "protocol = %d, message type = %d\n",
210 		      protocol, msgtype);
211 }
212 EXPORT_SYMBOL_GPL(rtnl_register);
213 
214 /**
215  * rtnl_unregister - Unregister a rtnetlink message type
216  * @protocol: Protocol family or PF_UNSPEC
217  * @msgtype: rtnetlink message type
218  *
219  * Returns 0 on success or a negative error code.
220  */
221 int rtnl_unregister(int protocol, int msgtype)
222 {
223 	struct rtnl_link *handlers;
224 	int msgindex;
225 
226 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
227 	msgindex = rtm_msgindex(msgtype);
228 
229 	rtnl_lock();
230 	handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
231 	if (!handlers) {
232 		rtnl_unlock();
233 		return -ENOENT;
234 	}
235 
236 	handlers[msgindex].doit = NULL;
237 	handlers[msgindex].dumpit = NULL;
238 	handlers[msgindex].flags = 0;
239 	rtnl_unlock();
240 
241 	return 0;
242 }
243 EXPORT_SYMBOL_GPL(rtnl_unregister);
244 
245 /**
246  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
247  * @protocol : Protocol family or PF_UNSPEC
248  *
249  * Identical to calling rtnl_unregster() for all registered message types
250  * of a certain protocol family.
251  */
252 void rtnl_unregister_all(int protocol)
253 {
254 	struct rtnl_link *handlers;
255 
256 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
257 
258 	rtnl_lock();
259 	handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
260 	RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
261 	rtnl_unlock();
262 
263 	synchronize_net();
264 
265 	while (refcount_read(&rtnl_msg_handlers_ref[protocol]) > 1)
266 		schedule();
267 	kfree(handlers);
268 }
269 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
270 
271 static LIST_HEAD(link_ops);
272 
273 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
274 {
275 	const struct rtnl_link_ops *ops;
276 
277 	list_for_each_entry(ops, &link_ops, list) {
278 		if (!strcmp(ops->kind, kind))
279 			return ops;
280 	}
281 	return NULL;
282 }
283 
284 /**
285  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
286  * @ops: struct rtnl_link_ops * to register
287  *
288  * The caller must hold the rtnl_mutex. This function should be used
289  * by drivers that create devices during module initialization. It
290  * must be called before registering the devices.
291  *
292  * Returns 0 on success or a negative error code.
293  */
294 int __rtnl_link_register(struct rtnl_link_ops *ops)
295 {
296 	if (rtnl_link_ops_get(ops->kind))
297 		return -EEXIST;
298 
299 	/* The check for setup is here because if ops
300 	 * does not have that filled up, it is not possible
301 	 * to use the ops for creating device. So do not
302 	 * fill up dellink as well. That disables rtnl_dellink.
303 	 */
304 	if (ops->setup && !ops->dellink)
305 		ops->dellink = unregister_netdevice_queue;
306 
307 	list_add_tail(&ops->list, &link_ops);
308 	return 0;
309 }
310 EXPORT_SYMBOL_GPL(__rtnl_link_register);
311 
312 /**
313  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
314  * @ops: struct rtnl_link_ops * to register
315  *
316  * Returns 0 on success or a negative error code.
317  */
318 int rtnl_link_register(struct rtnl_link_ops *ops)
319 {
320 	int err;
321 
322 	rtnl_lock();
323 	err = __rtnl_link_register(ops);
324 	rtnl_unlock();
325 	return err;
326 }
327 EXPORT_SYMBOL_GPL(rtnl_link_register);
328 
329 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
330 {
331 	struct net_device *dev;
332 	LIST_HEAD(list_kill);
333 
334 	for_each_netdev(net, dev) {
335 		if (dev->rtnl_link_ops == ops)
336 			ops->dellink(dev, &list_kill);
337 	}
338 	unregister_netdevice_many(&list_kill);
339 }
340 
341 /**
342  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
343  * @ops: struct rtnl_link_ops * to unregister
344  *
345  * The caller must hold the rtnl_mutex.
346  */
347 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
348 {
349 	struct net *net;
350 
351 	for_each_net(net) {
352 		__rtnl_kill_links(net, ops);
353 	}
354 	list_del(&ops->list);
355 }
356 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
357 
358 /* Return with the rtnl_lock held when there are no network
359  * devices unregistering in any network namespace.
360  */
361 static void rtnl_lock_unregistering_all(void)
362 {
363 	struct net *net;
364 	bool unregistering;
365 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
366 
367 	add_wait_queue(&netdev_unregistering_wq, &wait);
368 	for (;;) {
369 		unregistering = false;
370 		rtnl_lock();
371 		for_each_net(net) {
372 			if (net->dev_unreg_count > 0) {
373 				unregistering = true;
374 				break;
375 			}
376 		}
377 		if (!unregistering)
378 			break;
379 		__rtnl_unlock();
380 
381 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
382 	}
383 	remove_wait_queue(&netdev_unregistering_wq, &wait);
384 }
385 
386 /**
387  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
388  * @ops: struct rtnl_link_ops * to unregister
389  */
390 void rtnl_link_unregister(struct rtnl_link_ops *ops)
391 {
392 	/* Close the race with cleanup_net() */
393 	mutex_lock(&net_mutex);
394 	rtnl_lock_unregistering_all();
395 	__rtnl_link_unregister(ops);
396 	rtnl_unlock();
397 	mutex_unlock(&net_mutex);
398 }
399 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
400 
401 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
402 {
403 	struct net_device *master_dev;
404 	const struct rtnl_link_ops *ops;
405 	size_t size = 0;
406 
407 	rcu_read_lock();
408 
409 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
410 	if (!master_dev)
411 		goto out;
412 
413 	ops = master_dev->rtnl_link_ops;
414 	if (!ops || !ops->get_slave_size)
415 		goto out;
416 	/* IFLA_INFO_SLAVE_DATA + nested data */
417 	size = nla_total_size(sizeof(struct nlattr)) +
418 	       ops->get_slave_size(master_dev, dev);
419 
420 out:
421 	rcu_read_unlock();
422 	return size;
423 }
424 
425 static size_t rtnl_link_get_size(const struct net_device *dev)
426 {
427 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
428 	size_t size;
429 
430 	if (!ops)
431 		return 0;
432 
433 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
434 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
435 
436 	if (ops->get_size)
437 		/* IFLA_INFO_DATA + nested data */
438 		size += nla_total_size(sizeof(struct nlattr)) +
439 			ops->get_size(dev);
440 
441 	if (ops->get_xstats_size)
442 		/* IFLA_INFO_XSTATS */
443 		size += nla_total_size(ops->get_xstats_size(dev));
444 
445 	size += rtnl_link_get_slave_info_data_size(dev);
446 
447 	return size;
448 }
449 
450 static LIST_HEAD(rtnl_af_ops);
451 
452 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
453 {
454 	const struct rtnl_af_ops *ops;
455 
456 	list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
457 		if (ops->family == family)
458 			return ops;
459 	}
460 
461 	return NULL;
462 }
463 
464 /**
465  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
466  * @ops: struct rtnl_af_ops * to register
467  *
468  * Returns 0 on success or a negative error code.
469  */
470 void rtnl_af_register(struct rtnl_af_ops *ops)
471 {
472 	rtnl_lock();
473 	list_add_tail_rcu(&ops->list, &rtnl_af_ops);
474 	rtnl_unlock();
475 }
476 EXPORT_SYMBOL_GPL(rtnl_af_register);
477 
478 /**
479  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
480  * @ops: struct rtnl_af_ops * to unregister
481  */
482 void rtnl_af_unregister(struct rtnl_af_ops *ops)
483 {
484 	rtnl_lock();
485 	list_del_rcu(&ops->list);
486 	rtnl_unlock();
487 
488 	synchronize_rcu();
489 }
490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
491 
492 static size_t rtnl_link_get_af_size(const struct net_device *dev,
493 				    u32 ext_filter_mask)
494 {
495 	struct rtnl_af_ops *af_ops;
496 	size_t size;
497 
498 	/* IFLA_AF_SPEC */
499 	size = nla_total_size(sizeof(struct nlattr));
500 
501 	rcu_read_lock();
502 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
503 		if (af_ops->get_link_af_size) {
504 			/* AF_* + nested data */
505 			size += nla_total_size(sizeof(struct nlattr)) +
506 				af_ops->get_link_af_size(dev, ext_filter_mask);
507 		}
508 	}
509 	rcu_read_unlock();
510 
511 	return size;
512 }
513 
514 static bool rtnl_have_link_slave_info(const struct net_device *dev)
515 {
516 	struct net_device *master_dev;
517 	bool ret = false;
518 
519 	rcu_read_lock();
520 
521 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
522 	if (master_dev && master_dev->rtnl_link_ops)
523 		ret = true;
524 	rcu_read_unlock();
525 	return ret;
526 }
527 
528 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
529 				     const struct net_device *dev)
530 {
531 	struct net_device *master_dev;
532 	const struct rtnl_link_ops *ops;
533 	struct nlattr *slave_data;
534 	int err;
535 
536 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
537 	if (!master_dev)
538 		return 0;
539 	ops = master_dev->rtnl_link_ops;
540 	if (!ops)
541 		return 0;
542 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
543 		return -EMSGSIZE;
544 	if (ops->fill_slave_info) {
545 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
546 		if (!slave_data)
547 			return -EMSGSIZE;
548 		err = ops->fill_slave_info(skb, master_dev, dev);
549 		if (err < 0)
550 			goto err_cancel_slave_data;
551 		nla_nest_end(skb, slave_data);
552 	}
553 	return 0;
554 
555 err_cancel_slave_data:
556 	nla_nest_cancel(skb, slave_data);
557 	return err;
558 }
559 
560 static int rtnl_link_info_fill(struct sk_buff *skb,
561 			       const struct net_device *dev)
562 {
563 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
564 	struct nlattr *data;
565 	int err;
566 
567 	if (!ops)
568 		return 0;
569 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
570 		return -EMSGSIZE;
571 	if (ops->fill_xstats) {
572 		err = ops->fill_xstats(skb, dev);
573 		if (err < 0)
574 			return err;
575 	}
576 	if (ops->fill_info) {
577 		data = nla_nest_start(skb, IFLA_INFO_DATA);
578 		if (data == NULL)
579 			return -EMSGSIZE;
580 		err = ops->fill_info(skb, dev);
581 		if (err < 0)
582 			goto err_cancel_data;
583 		nla_nest_end(skb, data);
584 	}
585 	return 0;
586 
587 err_cancel_data:
588 	nla_nest_cancel(skb, data);
589 	return err;
590 }
591 
592 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
593 {
594 	struct nlattr *linkinfo;
595 	int err = -EMSGSIZE;
596 
597 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
598 	if (linkinfo == NULL)
599 		goto out;
600 
601 	err = rtnl_link_info_fill(skb, dev);
602 	if (err < 0)
603 		goto err_cancel_link;
604 
605 	err = rtnl_link_slave_info_fill(skb, dev);
606 	if (err < 0)
607 		goto err_cancel_link;
608 
609 	nla_nest_end(skb, linkinfo);
610 	return 0;
611 
612 err_cancel_link:
613 	nla_nest_cancel(skb, linkinfo);
614 out:
615 	return err;
616 }
617 
618 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
619 {
620 	struct sock *rtnl = net->rtnl;
621 	int err = 0;
622 
623 	NETLINK_CB(skb).dst_group = group;
624 	if (echo)
625 		refcount_inc(&skb->users);
626 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
627 	if (echo)
628 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
629 	return err;
630 }
631 
632 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
633 {
634 	struct sock *rtnl = net->rtnl;
635 
636 	return nlmsg_unicast(rtnl, skb, pid);
637 }
638 EXPORT_SYMBOL(rtnl_unicast);
639 
640 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
641 		 struct nlmsghdr *nlh, gfp_t flags)
642 {
643 	struct sock *rtnl = net->rtnl;
644 	int report = 0;
645 
646 	if (nlh)
647 		report = nlmsg_report(nlh);
648 
649 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
650 }
651 EXPORT_SYMBOL(rtnl_notify);
652 
653 void rtnl_set_sk_err(struct net *net, u32 group, int error)
654 {
655 	struct sock *rtnl = net->rtnl;
656 
657 	netlink_set_err(rtnl, 0, group, error);
658 }
659 EXPORT_SYMBOL(rtnl_set_sk_err);
660 
661 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
662 {
663 	struct nlattr *mx;
664 	int i, valid = 0;
665 
666 	mx = nla_nest_start(skb, RTA_METRICS);
667 	if (mx == NULL)
668 		return -ENOBUFS;
669 
670 	for (i = 0; i < RTAX_MAX; i++) {
671 		if (metrics[i]) {
672 			if (i == RTAX_CC_ALGO - 1) {
673 				char tmp[TCP_CA_NAME_MAX], *name;
674 
675 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
676 				if (!name)
677 					continue;
678 				if (nla_put_string(skb, i + 1, name))
679 					goto nla_put_failure;
680 			} else if (i == RTAX_FEATURES - 1) {
681 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
682 
683 				if (!user_features)
684 					continue;
685 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
686 				if (nla_put_u32(skb, i + 1, user_features))
687 					goto nla_put_failure;
688 			} else {
689 				if (nla_put_u32(skb, i + 1, metrics[i]))
690 					goto nla_put_failure;
691 			}
692 			valid++;
693 		}
694 	}
695 
696 	if (!valid) {
697 		nla_nest_cancel(skb, mx);
698 		return 0;
699 	}
700 
701 	return nla_nest_end(skb, mx);
702 
703 nla_put_failure:
704 	nla_nest_cancel(skb, mx);
705 	return -EMSGSIZE;
706 }
707 EXPORT_SYMBOL(rtnetlink_put_metrics);
708 
709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
710 		       long expires, u32 error)
711 {
712 	struct rta_cacheinfo ci = {
713 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
714 		.rta_used = dst->__use,
715 		.rta_clntref = atomic_read(&(dst->__refcnt)),
716 		.rta_error = error,
717 		.rta_id =  id,
718 	};
719 
720 	if (expires) {
721 		unsigned long clock;
722 
723 		clock = jiffies_to_clock_t(abs(expires));
724 		clock = min_t(unsigned long, clock, INT_MAX);
725 		ci.rta_expires = (expires > 0) ? clock : -clock;
726 	}
727 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
728 }
729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
730 
731 static void set_operstate(struct net_device *dev, unsigned char transition)
732 {
733 	unsigned char operstate = dev->operstate;
734 
735 	switch (transition) {
736 	case IF_OPER_UP:
737 		if ((operstate == IF_OPER_DORMANT ||
738 		     operstate == IF_OPER_UNKNOWN) &&
739 		    !netif_dormant(dev))
740 			operstate = IF_OPER_UP;
741 		break;
742 
743 	case IF_OPER_DORMANT:
744 		if (operstate == IF_OPER_UP ||
745 		    operstate == IF_OPER_UNKNOWN)
746 			operstate = IF_OPER_DORMANT;
747 		break;
748 	}
749 
750 	if (dev->operstate != operstate) {
751 		write_lock_bh(&dev_base_lock);
752 		dev->operstate = operstate;
753 		write_unlock_bh(&dev_base_lock);
754 		netdev_state_change(dev);
755 	}
756 }
757 
758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
759 {
760 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
761 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
762 }
763 
764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
765 					   const struct ifinfomsg *ifm)
766 {
767 	unsigned int flags = ifm->ifi_flags;
768 
769 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
770 	if (ifm->ifi_change)
771 		flags = (flags & ifm->ifi_change) |
772 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
773 
774 	return flags;
775 }
776 
777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
778 				 const struct rtnl_link_stats64 *b)
779 {
780 	a->rx_packets = b->rx_packets;
781 	a->tx_packets = b->tx_packets;
782 	a->rx_bytes = b->rx_bytes;
783 	a->tx_bytes = b->tx_bytes;
784 	a->rx_errors = b->rx_errors;
785 	a->tx_errors = b->tx_errors;
786 	a->rx_dropped = b->rx_dropped;
787 	a->tx_dropped = b->tx_dropped;
788 
789 	a->multicast = b->multicast;
790 	a->collisions = b->collisions;
791 
792 	a->rx_length_errors = b->rx_length_errors;
793 	a->rx_over_errors = b->rx_over_errors;
794 	a->rx_crc_errors = b->rx_crc_errors;
795 	a->rx_frame_errors = b->rx_frame_errors;
796 	a->rx_fifo_errors = b->rx_fifo_errors;
797 	a->rx_missed_errors = b->rx_missed_errors;
798 
799 	a->tx_aborted_errors = b->tx_aborted_errors;
800 	a->tx_carrier_errors = b->tx_carrier_errors;
801 	a->tx_fifo_errors = b->tx_fifo_errors;
802 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
803 	a->tx_window_errors = b->tx_window_errors;
804 
805 	a->rx_compressed = b->rx_compressed;
806 	a->tx_compressed = b->tx_compressed;
807 
808 	a->rx_nohandler = b->rx_nohandler;
809 }
810 
811 /* All VF info */
812 static inline int rtnl_vfinfo_size(const struct net_device *dev,
813 				   u32 ext_filter_mask)
814 {
815 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
816 		int num_vfs = dev_num_vf(dev->dev.parent);
817 		size_t size = nla_total_size(0);
818 		size += num_vfs *
819 			(nla_total_size(0) +
820 			 nla_total_size(sizeof(struct ifla_vf_mac)) +
821 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
822 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
823 			 nla_total_size(MAX_VLAN_LIST_LEN *
824 					sizeof(struct ifla_vf_vlan_info)) +
825 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
826 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
827 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
828 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
829 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
830 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
831 			 /* IFLA_VF_STATS_RX_PACKETS */
832 			 nla_total_size_64bit(sizeof(__u64)) +
833 			 /* IFLA_VF_STATS_TX_PACKETS */
834 			 nla_total_size_64bit(sizeof(__u64)) +
835 			 /* IFLA_VF_STATS_RX_BYTES */
836 			 nla_total_size_64bit(sizeof(__u64)) +
837 			 /* IFLA_VF_STATS_TX_BYTES */
838 			 nla_total_size_64bit(sizeof(__u64)) +
839 			 /* IFLA_VF_STATS_BROADCAST */
840 			 nla_total_size_64bit(sizeof(__u64)) +
841 			 /* IFLA_VF_STATS_MULTICAST */
842 			 nla_total_size_64bit(sizeof(__u64)) +
843 			 nla_total_size(sizeof(struct ifla_vf_trust)));
844 		return size;
845 	} else
846 		return 0;
847 }
848 
849 static size_t rtnl_port_size(const struct net_device *dev,
850 			     u32 ext_filter_mask)
851 {
852 	size_t port_size = nla_total_size(4)		/* PORT_VF */
853 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
854 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
855 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
856 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
857 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
858 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
859 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
860 		+ port_size;
861 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
862 		+ port_size;
863 
864 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
865 	    !(ext_filter_mask & RTEXT_FILTER_VF))
866 		return 0;
867 	if (dev_num_vf(dev->dev.parent))
868 		return port_self_size + vf_ports_size +
869 			vf_port_size * dev_num_vf(dev->dev.parent);
870 	else
871 		return port_self_size;
872 }
873 
874 static size_t rtnl_xdp_size(void)
875 {
876 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
877 			  nla_total_size(1) +	/* XDP_ATTACHED */
878 			  nla_total_size(4);	/* XDP_PROG_ID */
879 
880 	return xdp_size;
881 }
882 
883 static noinline size_t if_nlmsg_size(const struct net_device *dev,
884 				     u32 ext_filter_mask)
885 {
886 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
887 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
888 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
889 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
890 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
891 	       + nla_total_size(sizeof(struct rtnl_link_stats))
892 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
893 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
894 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
895 	       + nla_total_size(4) /* IFLA_TXQLEN */
896 	       + nla_total_size(4) /* IFLA_WEIGHT */
897 	       + nla_total_size(4) /* IFLA_MTU */
898 	       + nla_total_size(4) /* IFLA_LINK */
899 	       + nla_total_size(4) /* IFLA_MASTER */
900 	       + nla_total_size(1) /* IFLA_CARRIER */
901 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
902 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
903 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
904 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
905 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
906 	       + nla_total_size(1) /* IFLA_OPERSTATE */
907 	       + nla_total_size(1) /* IFLA_LINKMODE */
908 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
909 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
910 	       + nla_total_size(4) /* IFLA_GROUP */
911 	       + nla_total_size(ext_filter_mask
912 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
913 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
914 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
915 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
916 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
917 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
918 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
919 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
920 	       + rtnl_xdp_size() /* IFLA_XDP */
921 	       + nla_total_size(4)  /* IFLA_EVENT */
922 	       + nla_total_size(4)  /* IFLA_NEW_NETNSID */
923 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
924 
925 }
926 
927 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
928 {
929 	struct nlattr *vf_ports;
930 	struct nlattr *vf_port;
931 	int vf;
932 	int err;
933 
934 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
935 	if (!vf_ports)
936 		return -EMSGSIZE;
937 
938 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
939 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
940 		if (!vf_port)
941 			goto nla_put_failure;
942 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
943 			goto nla_put_failure;
944 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
945 		if (err == -EMSGSIZE)
946 			goto nla_put_failure;
947 		if (err) {
948 			nla_nest_cancel(skb, vf_port);
949 			continue;
950 		}
951 		nla_nest_end(skb, vf_port);
952 	}
953 
954 	nla_nest_end(skb, vf_ports);
955 
956 	return 0;
957 
958 nla_put_failure:
959 	nla_nest_cancel(skb, vf_ports);
960 	return -EMSGSIZE;
961 }
962 
963 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
964 {
965 	struct nlattr *port_self;
966 	int err;
967 
968 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
969 	if (!port_self)
970 		return -EMSGSIZE;
971 
972 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
973 	if (err) {
974 		nla_nest_cancel(skb, port_self);
975 		return (err == -EMSGSIZE) ? err : 0;
976 	}
977 
978 	nla_nest_end(skb, port_self);
979 
980 	return 0;
981 }
982 
983 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
984 			  u32 ext_filter_mask)
985 {
986 	int err;
987 
988 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
989 	    !(ext_filter_mask & RTEXT_FILTER_VF))
990 		return 0;
991 
992 	err = rtnl_port_self_fill(skb, dev);
993 	if (err)
994 		return err;
995 
996 	if (dev_num_vf(dev->dev.parent)) {
997 		err = rtnl_vf_ports_fill(skb, dev);
998 		if (err)
999 			return err;
1000 	}
1001 
1002 	return 0;
1003 }
1004 
1005 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1006 {
1007 	int err;
1008 	struct netdev_phys_item_id ppid;
1009 
1010 	err = dev_get_phys_port_id(dev, &ppid);
1011 	if (err) {
1012 		if (err == -EOPNOTSUPP)
1013 			return 0;
1014 		return err;
1015 	}
1016 
1017 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1018 		return -EMSGSIZE;
1019 
1020 	return 0;
1021 }
1022 
1023 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1024 {
1025 	char name[IFNAMSIZ];
1026 	int err;
1027 
1028 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1029 	if (err) {
1030 		if (err == -EOPNOTSUPP)
1031 			return 0;
1032 		return err;
1033 	}
1034 
1035 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1036 		return -EMSGSIZE;
1037 
1038 	return 0;
1039 }
1040 
1041 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1042 {
1043 	int err;
1044 	struct switchdev_attr attr = {
1045 		.orig_dev = dev,
1046 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1047 		.flags = SWITCHDEV_F_NO_RECURSE,
1048 	};
1049 
1050 	err = switchdev_port_attr_get(dev, &attr);
1051 	if (err) {
1052 		if (err == -EOPNOTSUPP)
1053 			return 0;
1054 		return err;
1055 	}
1056 
1057 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1058 		    attr.u.ppid.id))
1059 		return -EMSGSIZE;
1060 
1061 	return 0;
1062 }
1063 
1064 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1065 					      struct net_device *dev)
1066 {
1067 	struct rtnl_link_stats64 *sp;
1068 	struct nlattr *attr;
1069 
1070 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1071 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1072 	if (!attr)
1073 		return -EMSGSIZE;
1074 
1075 	sp = nla_data(attr);
1076 	dev_get_stats(dev, sp);
1077 
1078 	attr = nla_reserve(skb, IFLA_STATS,
1079 			   sizeof(struct rtnl_link_stats));
1080 	if (!attr)
1081 		return -EMSGSIZE;
1082 
1083 	copy_rtnl_link_stats(nla_data(attr), sp);
1084 
1085 	return 0;
1086 }
1087 
1088 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1089 					       struct net_device *dev,
1090 					       int vfs_num,
1091 					       struct nlattr *vfinfo)
1092 {
1093 	struct ifla_vf_rss_query_en vf_rss_query_en;
1094 	struct nlattr *vf, *vfstats, *vfvlanlist;
1095 	struct ifla_vf_link_state vf_linkstate;
1096 	struct ifla_vf_vlan_info vf_vlan_info;
1097 	struct ifla_vf_spoofchk vf_spoofchk;
1098 	struct ifla_vf_tx_rate vf_tx_rate;
1099 	struct ifla_vf_stats vf_stats;
1100 	struct ifla_vf_trust vf_trust;
1101 	struct ifla_vf_vlan vf_vlan;
1102 	struct ifla_vf_rate vf_rate;
1103 	struct ifla_vf_mac vf_mac;
1104 	struct ifla_vf_info ivi;
1105 
1106 	memset(&ivi, 0, sizeof(ivi));
1107 
1108 	/* Not all SR-IOV capable drivers support the
1109 	 * spoofcheck and "RSS query enable" query.  Preset to
1110 	 * -1 so the user space tool can detect that the driver
1111 	 * didn't report anything.
1112 	 */
1113 	ivi.spoofchk = -1;
1114 	ivi.rss_query_en = -1;
1115 	ivi.trusted = -1;
1116 	/* The default value for VF link state is "auto"
1117 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1118 	 */
1119 	ivi.linkstate = 0;
1120 	/* VLAN Protocol by default is 802.1Q */
1121 	ivi.vlan_proto = htons(ETH_P_8021Q);
1122 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1123 		return 0;
1124 
1125 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1126 
1127 	vf_mac.vf =
1128 		vf_vlan.vf =
1129 		vf_vlan_info.vf =
1130 		vf_rate.vf =
1131 		vf_tx_rate.vf =
1132 		vf_spoofchk.vf =
1133 		vf_linkstate.vf =
1134 		vf_rss_query_en.vf =
1135 		vf_trust.vf = ivi.vf;
1136 
1137 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1138 	vf_vlan.vlan = ivi.vlan;
1139 	vf_vlan.qos = ivi.qos;
1140 	vf_vlan_info.vlan = ivi.vlan;
1141 	vf_vlan_info.qos = ivi.qos;
1142 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1143 	vf_tx_rate.rate = ivi.max_tx_rate;
1144 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1145 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1146 	vf_spoofchk.setting = ivi.spoofchk;
1147 	vf_linkstate.link_state = ivi.linkstate;
1148 	vf_rss_query_en.setting = ivi.rss_query_en;
1149 	vf_trust.setting = ivi.trusted;
1150 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1151 	if (!vf)
1152 		goto nla_put_vfinfo_failure;
1153 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1154 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1155 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1156 		    &vf_rate) ||
1157 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1158 		    &vf_tx_rate) ||
1159 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1160 		    &vf_spoofchk) ||
1161 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1162 		    &vf_linkstate) ||
1163 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1164 		    sizeof(vf_rss_query_en),
1165 		    &vf_rss_query_en) ||
1166 	    nla_put(skb, IFLA_VF_TRUST,
1167 		    sizeof(vf_trust), &vf_trust))
1168 		goto nla_put_vf_failure;
1169 	vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1170 	if (!vfvlanlist)
1171 		goto nla_put_vf_failure;
1172 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1173 		    &vf_vlan_info)) {
1174 		nla_nest_cancel(skb, vfvlanlist);
1175 		goto nla_put_vf_failure;
1176 	}
1177 	nla_nest_end(skb, vfvlanlist);
1178 	memset(&vf_stats, 0, sizeof(vf_stats));
1179 	if (dev->netdev_ops->ndo_get_vf_stats)
1180 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1181 						&vf_stats);
1182 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1183 	if (!vfstats)
1184 		goto nla_put_vf_failure;
1185 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1186 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1187 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1188 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1189 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1190 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1191 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1192 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1193 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1194 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1195 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1196 			      vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1197 		nla_nest_cancel(skb, vfstats);
1198 		goto nla_put_vf_failure;
1199 	}
1200 	nla_nest_end(skb, vfstats);
1201 	nla_nest_end(skb, vf);
1202 	return 0;
1203 
1204 nla_put_vf_failure:
1205 	nla_nest_cancel(skb, vf);
1206 nla_put_vfinfo_failure:
1207 	nla_nest_cancel(skb, vfinfo);
1208 	return -EMSGSIZE;
1209 }
1210 
1211 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1212 					   struct net_device *dev,
1213 					   u32 ext_filter_mask)
1214 {
1215 	struct nlattr *vfinfo;
1216 	int i, num_vfs;
1217 
1218 	if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1219 		return 0;
1220 
1221 	num_vfs = dev_num_vf(dev->dev.parent);
1222 	if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1223 		return -EMSGSIZE;
1224 
1225 	if (!dev->netdev_ops->ndo_get_vf_config)
1226 		return 0;
1227 
1228 	vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1229 	if (!vfinfo)
1230 		return -EMSGSIZE;
1231 
1232 	for (i = 0; i < num_vfs; i++) {
1233 		if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1234 			return -EMSGSIZE;
1235 	}
1236 
1237 	nla_nest_end(skb, vfinfo);
1238 	return 0;
1239 }
1240 
1241 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1242 {
1243 	struct rtnl_link_ifmap map;
1244 
1245 	memset(&map, 0, sizeof(map));
1246 	map.mem_start   = dev->mem_start;
1247 	map.mem_end     = dev->mem_end;
1248 	map.base_addr   = dev->base_addr;
1249 	map.irq         = dev->irq;
1250 	map.dma         = dev->dma;
1251 	map.port        = dev->if_port;
1252 
1253 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1254 		return -EMSGSIZE;
1255 
1256 	return 0;
1257 }
1258 
1259 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1260 {
1261 	const struct net_device_ops *ops = dev->netdev_ops;
1262 	const struct bpf_prog *generic_xdp_prog;
1263 
1264 	ASSERT_RTNL();
1265 
1266 	*prog_id = 0;
1267 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1268 	if (generic_xdp_prog) {
1269 		*prog_id = generic_xdp_prog->aux->id;
1270 		return XDP_ATTACHED_SKB;
1271 	}
1272 	if (!ops->ndo_xdp)
1273 		return XDP_ATTACHED_NONE;
1274 
1275 	return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1276 }
1277 
1278 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1279 {
1280 	struct nlattr *xdp;
1281 	u32 prog_id;
1282 	int err;
1283 
1284 	xdp = nla_nest_start(skb, IFLA_XDP);
1285 	if (!xdp)
1286 		return -EMSGSIZE;
1287 
1288 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1289 			 rtnl_xdp_attached_mode(dev, &prog_id));
1290 	if (err)
1291 		goto err_cancel;
1292 
1293 	if (prog_id) {
1294 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1295 		if (err)
1296 			goto err_cancel;
1297 	}
1298 
1299 	nla_nest_end(skb, xdp);
1300 	return 0;
1301 
1302 err_cancel:
1303 	nla_nest_cancel(skb, xdp);
1304 	return err;
1305 }
1306 
1307 static u32 rtnl_get_event(unsigned long event)
1308 {
1309 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1310 
1311 	switch (event) {
1312 	case NETDEV_REBOOT:
1313 		rtnl_event_type = IFLA_EVENT_REBOOT;
1314 		break;
1315 	case NETDEV_FEAT_CHANGE:
1316 		rtnl_event_type = IFLA_EVENT_FEATURES;
1317 		break;
1318 	case NETDEV_BONDING_FAILOVER:
1319 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1320 		break;
1321 	case NETDEV_NOTIFY_PEERS:
1322 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1323 		break;
1324 	case NETDEV_RESEND_IGMP:
1325 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1326 		break;
1327 	case NETDEV_CHANGEINFODATA:
1328 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1329 		break;
1330 	default:
1331 		break;
1332 	}
1333 
1334 	return rtnl_event_type;
1335 }
1336 
1337 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1338 {
1339 	const struct net_device *upper_dev;
1340 	int ret = 0;
1341 
1342 	rcu_read_lock();
1343 
1344 	upper_dev = netdev_master_upper_dev_get_rcu(dev);
1345 	if (upper_dev)
1346 		ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1347 
1348 	rcu_read_unlock();
1349 	return ret;
1350 }
1351 
1352 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1353 {
1354 	int ifindex = dev_get_iflink(dev);
1355 
1356 	if (dev->ifindex == ifindex)
1357 		return 0;
1358 
1359 	return nla_put_u32(skb, IFLA_LINK, ifindex);
1360 }
1361 
1362 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1363 					      struct net_device *dev)
1364 {
1365 	char buf[IFALIASZ];
1366 	int ret;
1367 
1368 	ret = dev_get_alias(dev, buf, sizeof(buf));
1369 	return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1370 }
1371 
1372 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1373 				  const struct net_device *dev)
1374 {
1375 	if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1376 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1377 
1378 		if (!net_eq(dev_net(dev), link_net)) {
1379 			int id = peernet2id_alloc(dev_net(dev), link_net);
1380 
1381 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1382 				return -EMSGSIZE;
1383 		}
1384 	}
1385 
1386 	return 0;
1387 }
1388 
1389 static int rtnl_fill_link_af(struct sk_buff *skb,
1390 			     const struct net_device *dev,
1391 			     u32 ext_filter_mask)
1392 {
1393 	const struct rtnl_af_ops *af_ops;
1394 	struct nlattr *af_spec;
1395 
1396 	af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1397 	if (!af_spec)
1398 		return -EMSGSIZE;
1399 
1400 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1401 		struct nlattr *af;
1402 		int err;
1403 
1404 		if (!af_ops->fill_link_af)
1405 			continue;
1406 
1407 		af = nla_nest_start(skb, af_ops->family);
1408 		if (!af)
1409 			return -EMSGSIZE;
1410 
1411 		err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1412 		/*
1413 		 * Caller may return ENODATA to indicate that there
1414 		 * was no data to be dumped. This is not an error, it
1415 		 * means we should trim the attribute header and
1416 		 * continue.
1417 		 */
1418 		if (err == -ENODATA)
1419 			nla_nest_cancel(skb, af);
1420 		else if (err < 0)
1421 			return -EMSGSIZE;
1422 
1423 		nla_nest_end(skb, af);
1424 	}
1425 
1426 	nla_nest_end(skb, af_spec);
1427 	return 0;
1428 }
1429 
1430 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1431 			    int type, u32 pid, u32 seq, u32 change,
1432 			    unsigned int flags, u32 ext_filter_mask,
1433 			    u32 event, int *new_nsid)
1434 {
1435 	struct ifinfomsg *ifm;
1436 	struct nlmsghdr *nlh;
1437 
1438 	ASSERT_RTNL();
1439 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1440 	if (nlh == NULL)
1441 		return -EMSGSIZE;
1442 
1443 	ifm = nlmsg_data(nlh);
1444 	ifm->ifi_family = AF_UNSPEC;
1445 	ifm->__ifi_pad = 0;
1446 	ifm->ifi_type = dev->type;
1447 	ifm->ifi_index = dev->ifindex;
1448 	ifm->ifi_flags = dev_get_flags(dev);
1449 	ifm->ifi_change = change;
1450 
1451 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1452 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1453 	    nla_put_u8(skb, IFLA_OPERSTATE,
1454 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1455 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1456 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1457 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1458 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1459 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1460 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1461 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1462 #ifdef CONFIG_RPS
1463 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1464 #endif
1465 	    nla_put_iflink(skb, dev) ||
1466 	    put_master_ifindex(skb, dev) ||
1467 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1468 	    (dev->qdisc &&
1469 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1470 	    nla_put_ifalias(skb, dev) ||
1471 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1472 			atomic_read(&dev->carrier_changes)) ||
1473 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1474 		goto nla_put_failure;
1475 
1476 	if (event != IFLA_EVENT_NONE) {
1477 		if (nla_put_u32(skb, IFLA_EVENT, event))
1478 			goto nla_put_failure;
1479 	}
1480 
1481 	if (rtnl_fill_link_ifmap(skb, dev))
1482 		goto nla_put_failure;
1483 
1484 	if (dev->addr_len) {
1485 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1486 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1487 			goto nla_put_failure;
1488 	}
1489 
1490 	if (rtnl_phys_port_id_fill(skb, dev))
1491 		goto nla_put_failure;
1492 
1493 	if (rtnl_phys_port_name_fill(skb, dev))
1494 		goto nla_put_failure;
1495 
1496 	if (rtnl_phys_switch_id_fill(skb, dev))
1497 		goto nla_put_failure;
1498 
1499 	if (rtnl_fill_stats(skb, dev))
1500 		goto nla_put_failure;
1501 
1502 	if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1503 		goto nla_put_failure;
1504 
1505 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1506 		goto nla_put_failure;
1507 
1508 	if (rtnl_xdp_fill(skb, dev))
1509 		goto nla_put_failure;
1510 
1511 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1512 		if (rtnl_link_fill(skb, dev) < 0)
1513 			goto nla_put_failure;
1514 	}
1515 
1516 	if (rtnl_fill_link_netnsid(skb, dev))
1517 		goto nla_put_failure;
1518 
1519 	if (new_nsid &&
1520 	    nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1521 		goto nla_put_failure;
1522 
1523 	rcu_read_lock();
1524 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1525 		goto nla_put_failure_rcu;
1526 	rcu_read_unlock();
1527 
1528 	nlmsg_end(skb, nlh);
1529 	return 0;
1530 
1531 nla_put_failure_rcu:
1532 	rcu_read_unlock();
1533 nla_put_failure:
1534 	nlmsg_cancel(skb, nlh);
1535 	return -EMSGSIZE;
1536 }
1537 
1538 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1539 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1540 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1541 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1542 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1543 	[IFLA_MTU]		= { .type = NLA_U32 },
1544 	[IFLA_LINK]		= { .type = NLA_U32 },
1545 	[IFLA_MASTER]		= { .type = NLA_U32 },
1546 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1547 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1548 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1549 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1550 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1551 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1552 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1553 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1554 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1555 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1556 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1557 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1558 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1559 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1560 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1561 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1562 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1563 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1564 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1565 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1566 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1567 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1568 	[IFLA_XDP]		= { .type = NLA_NESTED },
1569 	[IFLA_EVENT]		= { .type = NLA_U32 },
1570 	[IFLA_GROUP]		= { .type = NLA_U32 },
1571 };
1572 
1573 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1574 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1575 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1576 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1577 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1578 };
1579 
1580 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1581 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1582 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1583 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1584 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1585 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1586 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1587 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1588 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1589 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1590 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1591 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1592 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1593 };
1594 
1595 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1596 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1597 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1598 				    .len = PORT_PROFILE_MAX },
1599 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1600 				      .len = PORT_UUID_MAX },
1601 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1602 				    .len = PORT_UUID_MAX },
1603 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1604 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1605 
1606 	/* Unused, but we need to keep it here since user space could
1607 	 * fill it. It's also broken with regard to NLA_BINARY use in
1608 	 * combination with structs.
1609 	 */
1610 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1611 				    .len = sizeof(struct ifla_port_vsi) },
1612 };
1613 
1614 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1615 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1616 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1617 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1618 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1619 };
1620 
1621 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1622 {
1623 	const struct rtnl_link_ops *ops = NULL;
1624 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1625 
1626 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1627 			     ifla_info_policy, NULL) < 0)
1628 		return NULL;
1629 
1630 	if (linfo[IFLA_INFO_KIND]) {
1631 		char kind[MODULE_NAME_LEN];
1632 
1633 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1634 		ops = rtnl_link_ops_get(kind);
1635 	}
1636 
1637 	return ops;
1638 }
1639 
1640 static bool link_master_filtered(struct net_device *dev, int master_idx)
1641 {
1642 	struct net_device *master;
1643 
1644 	if (!master_idx)
1645 		return false;
1646 
1647 	master = netdev_master_upper_dev_get(dev);
1648 	if (!master || master->ifindex != master_idx)
1649 		return true;
1650 
1651 	return false;
1652 }
1653 
1654 static bool link_kind_filtered(const struct net_device *dev,
1655 			       const struct rtnl_link_ops *kind_ops)
1656 {
1657 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1658 		return true;
1659 
1660 	return false;
1661 }
1662 
1663 static bool link_dump_filtered(struct net_device *dev,
1664 			       int master_idx,
1665 			       const struct rtnl_link_ops *kind_ops)
1666 {
1667 	if (link_master_filtered(dev, master_idx) ||
1668 	    link_kind_filtered(dev, kind_ops))
1669 		return true;
1670 
1671 	return false;
1672 }
1673 
1674 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1675 {
1676 	struct net *net = sock_net(skb->sk);
1677 	int h, s_h;
1678 	int idx = 0, s_idx;
1679 	struct net_device *dev;
1680 	struct hlist_head *head;
1681 	struct nlattr *tb[IFLA_MAX+1];
1682 	u32 ext_filter_mask = 0;
1683 	const struct rtnl_link_ops *kind_ops = NULL;
1684 	unsigned int flags = NLM_F_MULTI;
1685 	int master_idx = 0;
1686 	int err;
1687 	int hdrlen;
1688 
1689 	s_h = cb->args[0];
1690 	s_idx = cb->args[1];
1691 
1692 	/* A hack to preserve kernel<->userspace interface.
1693 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1694 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1695 	 * what iproute2 < v3.9.0 used.
1696 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1697 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1698 	 */
1699 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1700 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1701 
1702 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1703 			ifla_policy, NULL) >= 0) {
1704 		if (tb[IFLA_EXT_MASK])
1705 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1706 
1707 		if (tb[IFLA_MASTER])
1708 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1709 
1710 		if (tb[IFLA_LINKINFO])
1711 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1712 
1713 		if (master_idx || kind_ops)
1714 			flags |= NLM_F_DUMP_FILTERED;
1715 	}
1716 
1717 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1718 		idx = 0;
1719 		head = &net->dev_index_head[h];
1720 		hlist_for_each_entry(dev, head, index_hlist) {
1721 			if (link_dump_filtered(dev, master_idx, kind_ops))
1722 				goto cont;
1723 			if (idx < s_idx)
1724 				goto cont;
1725 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1726 					       NETLINK_CB(cb->skb).portid,
1727 					       cb->nlh->nlmsg_seq, 0,
1728 					       flags,
1729 					       ext_filter_mask, 0, NULL);
1730 
1731 			if (err < 0) {
1732 				if (likely(skb->len))
1733 					goto out;
1734 
1735 				goto out_err;
1736 			}
1737 cont:
1738 			idx++;
1739 		}
1740 	}
1741 out:
1742 	err = skb->len;
1743 out_err:
1744 	cb->args[1] = idx;
1745 	cb->args[0] = h;
1746 	cb->seq = net->dev_base_seq;
1747 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1748 
1749 	return err;
1750 }
1751 
1752 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1753 			struct netlink_ext_ack *exterr)
1754 {
1755 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1756 }
1757 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1758 
1759 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1760 {
1761 	struct net *net;
1762 	/* Examine the link attributes and figure out which
1763 	 * network namespace we are talking about.
1764 	 */
1765 	if (tb[IFLA_NET_NS_PID])
1766 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1767 	else if (tb[IFLA_NET_NS_FD])
1768 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1769 	else
1770 		net = get_net(src_net);
1771 	return net;
1772 }
1773 EXPORT_SYMBOL(rtnl_link_get_net);
1774 
1775 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1776 {
1777 	if (dev) {
1778 		if (tb[IFLA_ADDRESS] &&
1779 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1780 			return -EINVAL;
1781 
1782 		if (tb[IFLA_BROADCAST] &&
1783 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1784 			return -EINVAL;
1785 	}
1786 
1787 	if (tb[IFLA_AF_SPEC]) {
1788 		struct nlattr *af;
1789 		int rem, err;
1790 
1791 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1792 			const struct rtnl_af_ops *af_ops;
1793 
1794 			rcu_read_lock();
1795 			af_ops = rtnl_af_lookup(nla_type(af));
1796 			if (!af_ops) {
1797 				rcu_read_unlock();
1798 				return -EAFNOSUPPORT;
1799 			}
1800 
1801 			if (!af_ops->set_link_af) {
1802 				rcu_read_unlock();
1803 				return -EOPNOTSUPP;
1804 			}
1805 
1806 			if (af_ops->validate_link_af) {
1807 				err = af_ops->validate_link_af(dev, af);
1808 				if (err < 0) {
1809 					rcu_read_unlock();
1810 					return err;
1811 				}
1812 			}
1813 
1814 			rcu_read_unlock();
1815 		}
1816 	}
1817 
1818 	return 0;
1819 }
1820 
1821 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1822 				  int guid_type)
1823 {
1824 	const struct net_device_ops *ops = dev->netdev_ops;
1825 
1826 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1827 }
1828 
1829 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1830 {
1831 	if (dev->type != ARPHRD_INFINIBAND)
1832 		return -EOPNOTSUPP;
1833 
1834 	return handle_infiniband_guid(dev, ivt, guid_type);
1835 }
1836 
1837 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1838 {
1839 	const struct net_device_ops *ops = dev->netdev_ops;
1840 	int err = -EINVAL;
1841 
1842 	if (tb[IFLA_VF_MAC]) {
1843 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1844 
1845 		err = -EOPNOTSUPP;
1846 		if (ops->ndo_set_vf_mac)
1847 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1848 						  ivm->mac);
1849 		if (err < 0)
1850 			return err;
1851 	}
1852 
1853 	if (tb[IFLA_VF_VLAN]) {
1854 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1855 
1856 		err = -EOPNOTSUPP;
1857 		if (ops->ndo_set_vf_vlan)
1858 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1859 						   ivv->qos,
1860 						   htons(ETH_P_8021Q));
1861 		if (err < 0)
1862 			return err;
1863 	}
1864 
1865 	if (tb[IFLA_VF_VLAN_LIST]) {
1866 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1867 		struct nlattr *attr;
1868 		int rem, len = 0;
1869 
1870 		err = -EOPNOTSUPP;
1871 		if (!ops->ndo_set_vf_vlan)
1872 			return err;
1873 
1874 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1875 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1876 			    nla_len(attr) < NLA_HDRLEN) {
1877 				return -EINVAL;
1878 			}
1879 			if (len >= MAX_VLAN_LIST_LEN)
1880 				return -EOPNOTSUPP;
1881 			ivvl[len] = nla_data(attr);
1882 
1883 			len++;
1884 		}
1885 		if (len == 0)
1886 			return -EINVAL;
1887 
1888 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1889 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
1890 		if (err < 0)
1891 			return err;
1892 	}
1893 
1894 	if (tb[IFLA_VF_TX_RATE]) {
1895 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1896 		struct ifla_vf_info ivf;
1897 
1898 		err = -EOPNOTSUPP;
1899 		if (ops->ndo_get_vf_config)
1900 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1901 		if (err < 0)
1902 			return err;
1903 
1904 		err = -EOPNOTSUPP;
1905 		if (ops->ndo_set_vf_rate)
1906 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1907 						   ivf.min_tx_rate,
1908 						   ivt->rate);
1909 		if (err < 0)
1910 			return err;
1911 	}
1912 
1913 	if (tb[IFLA_VF_RATE]) {
1914 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1915 
1916 		err = -EOPNOTSUPP;
1917 		if (ops->ndo_set_vf_rate)
1918 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1919 						   ivt->min_tx_rate,
1920 						   ivt->max_tx_rate);
1921 		if (err < 0)
1922 			return err;
1923 	}
1924 
1925 	if (tb[IFLA_VF_SPOOFCHK]) {
1926 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1927 
1928 		err = -EOPNOTSUPP;
1929 		if (ops->ndo_set_vf_spoofchk)
1930 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1931 						       ivs->setting);
1932 		if (err < 0)
1933 			return err;
1934 	}
1935 
1936 	if (tb[IFLA_VF_LINK_STATE]) {
1937 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1938 
1939 		err = -EOPNOTSUPP;
1940 		if (ops->ndo_set_vf_link_state)
1941 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1942 							 ivl->link_state);
1943 		if (err < 0)
1944 			return err;
1945 	}
1946 
1947 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1948 		struct ifla_vf_rss_query_en *ivrssq_en;
1949 
1950 		err = -EOPNOTSUPP;
1951 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1952 		if (ops->ndo_set_vf_rss_query_en)
1953 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1954 							   ivrssq_en->setting);
1955 		if (err < 0)
1956 			return err;
1957 	}
1958 
1959 	if (tb[IFLA_VF_TRUST]) {
1960 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1961 
1962 		err = -EOPNOTSUPP;
1963 		if (ops->ndo_set_vf_trust)
1964 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1965 		if (err < 0)
1966 			return err;
1967 	}
1968 
1969 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1970 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1971 
1972 		if (!ops->ndo_set_vf_guid)
1973 			return -EOPNOTSUPP;
1974 
1975 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1976 	}
1977 
1978 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1979 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1980 
1981 		if (!ops->ndo_set_vf_guid)
1982 			return -EOPNOTSUPP;
1983 
1984 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1985 	}
1986 
1987 	return err;
1988 }
1989 
1990 static int do_set_master(struct net_device *dev, int ifindex,
1991 			 struct netlink_ext_ack *extack)
1992 {
1993 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1994 	const struct net_device_ops *ops;
1995 	int err;
1996 
1997 	if (upper_dev) {
1998 		if (upper_dev->ifindex == ifindex)
1999 			return 0;
2000 		ops = upper_dev->netdev_ops;
2001 		if (ops->ndo_del_slave) {
2002 			err = ops->ndo_del_slave(upper_dev, dev);
2003 			if (err)
2004 				return err;
2005 		} else {
2006 			return -EOPNOTSUPP;
2007 		}
2008 	}
2009 
2010 	if (ifindex) {
2011 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2012 		if (!upper_dev)
2013 			return -EINVAL;
2014 		ops = upper_dev->netdev_ops;
2015 		if (ops->ndo_add_slave) {
2016 			err = ops->ndo_add_slave(upper_dev, dev, extack);
2017 			if (err)
2018 				return err;
2019 		} else {
2020 			return -EOPNOTSUPP;
2021 		}
2022 	}
2023 	return 0;
2024 }
2025 
2026 #define DO_SETLINK_MODIFIED	0x01
2027 /* notify flag means notify + modified. */
2028 #define DO_SETLINK_NOTIFY	0x03
2029 static int do_setlink(const struct sk_buff *skb,
2030 		      struct net_device *dev, struct ifinfomsg *ifm,
2031 		      struct netlink_ext_ack *extack,
2032 		      struct nlattr **tb, char *ifname, int status)
2033 {
2034 	const struct net_device_ops *ops = dev->netdev_ops;
2035 	int err;
2036 
2037 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
2038 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
2039 		if (IS_ERR(net)) {
2040 			err = PTR_ERR(net);
2041 			goto errout;
2042 		}
2043 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
2044 			put_net(net);
2045 			err = -EPERM;
2046 			goto errout;
2047 		}
2048 		err = dev_change_net_namespace(dev, net, ifname);
2049 		put_net(net);
2050 		if (err)
2051 			goto errout;
2052 		status |= DO_SETLINK_MODIFIED;
2053 	}
2054 
2055 	if (tb[IFLA_MAP]) {
2056 		struct rtnl_link_ifmap *u_map;
2057 		struct ifmap k_map;
2058 
2059 		if (!ops->ndo_set_config) {
2060 			err = -EOPNOTSUPP;
2061 			goto errout;
2062 		}
2063 
2064 		if (!netif_device_present(dev)) {
2065 			err = -ENODEV;
2066 			goto errout;
2067 		}
2068 
2069 		u_map = nla_data(tb[IFLA_MAP]);
2070 		k_map.mem_start = (unsigned long) u_map->mem_start;
2071 		k_map.mem_end = (unsigned long) u_map->mem_end;
2072 		k_map.base_addr = (unsigned short) u_map->base_addr;
2073 		k_map.irq = (unsigned char) u_map->irq;
2074 		k_map.dma = (unsigned char) u_map->dma;
2075 		k_map.port = (unsigned char) u_map->port;
2076 
2077 		err = ops->ndo_set_config(dev, &k_map);
2078 		if (err < 0)
2079 			goto errout;
2080 
2081 		status |= DO_SETLINK_NOTIFY;
2082 	}
2083 
2084 	if (tb[IFLA_ADDRESS]) {
2085 		struct sockaddr *sa;
2086 		int len;
2087 
2088 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2089 						  sizeof(*sa));
2090 		sa = kmalloc(len, GFP_KERNEL);
2091 		if (!sa) {
2092 			err = -ENOMEM;
2093 			goto errout;
2094 		}
2095 		sa->sa_family = dev->type;
2096 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2097 		       dev->addr_len);
2098 		err = dev_set_mac_address(dev, sa);
2099 		kfree(sa);
2100 		if (err)
2101 			goto errout;
2102 		status |= DO_SETLINK_MODIFIED;
2103 	}
2104 
2105 	if (tb[IFLA_MTU]) {
2106 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2107 		if (err < 0)
2108 			goto errout;
2109 		status |= DO_SETLINK_MODIFIED;
2110 	}
2111 
2112 	if (tb[IFLA_GROUP]) {
2113 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2114 		status |= DO_SETLINK_NOTIFY;
2115 	}
2116 
2117 	/*
2118 	 * Interface selected by interface index but interface
2119 	 * name provided implies that a name change has been
2120 	 * requested.
2121 	 */
2122 	if (ifm->ifi_index > 0 && ifname[0]) {
2123 		err = dev_change_name(dev, ifname);
2124 		if (err < 0)
2125 			goto errout;
2126 		status |= DO_SETLINK_MODIFIED;
2127 	}
2128 
2129 	if (tb[IFLA_IFALIAS]) {
2130 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2131 				    nla_len(tb[IFLA_IFALIAS]));
2132 		if (err < 0)
2133 			goto errout;
2134 		status |= DO_SETLINK_NOTIFY;
2135 	}
2136 
2137 	if (tb[IFLA_BROADCAST]) {
2138 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2139 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2140 	}
2141 
2142 	if (ifm->ifi_flags || ifm->ifi_change) {
2143 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2144 		if (err < 0)
2145 			goto errout;
2146 	}
2147 
2148 	if (tb[IFLA_MASTER]) {
2149 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2150 		if (err)
2151 			goto errout;
2152 		status |= DO_SETLINK_MODIFIED;
2153 	}
2154 
2155 	if (tb[IFLA_CARRIER]) {
2156 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2157 		if (err)
2158 			goto errout;
2159 		status |= DO_SETLINK_MODIFIED;
2160 	}
2161 
2162 	if (tb[IFLA_TXQLEN]) {
2163 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2164 		unsigned int orig_len = dev->tx_queue_len;
2165 
2166 		if (dev->tx_queue_len ^ value) {
2167 			dev->tx_queue_len = value;
2168 			err = call_netdevice_notifiers(
2169 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2170 			err = notifier_to_errno(err);
2171 			if (err) {
2172 				dev->tx_queue_len = orig_len;
2173 				goto errout;
2174 			}
2175 			status |= DO_SETLINK_NOTIFY;
2176 		}
2177 	}
2178 
2179 	if (tb[IFLA_OPERSTATE])
2180 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2181 
2182 	if (tb[IFLA_LINKMODE]) {
2183 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2184 
2185 		write_lock_bh(&dev_base_lock);
2186 		if (dev->link_mode ^ value)
2187 			status |= DO_SETLINK_NOTIFY;
2188 		dev->link_mode = value;
2189 		write_unlock_bh(&dev_base_lock);
2190 	}
2191 
2192 	if (tb[IFLA_VFINFO_LIST]) {
2193 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2194 		struct nlattr *attr;
2195 		int rem;
2196 
2197 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2198 			if (nla_type(attr) != IFLA_VF_INFO ||
2199 			    nla_len(attr) < NLA_HDRLEN) {
2200 				err = -EINVAL;
2201 				goto errout;
2202 			}
2203 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2204 					       ifla_vf_policy, NULL);
2205 			if (err < 0)
2206 				goto errout;
2207 			err = do_setvfinfo(dev, vfinfo);
2208 			if (err < 0)
2209 				goto errout;
2210 			status |= DO_SETLINK_NOTIFY;
2211 		}
2212 	}
2213 	err = 0;
2214 
2215 	if (tb[IFLA_VF_PORTS]) {
2216 		struct nlattr *port[IFLA_PORT_MAX+1];
2217 		struct nlattr *attr;
2218 		int vf;
2219 		int rem;
2220 
2221 		err = -EOPNOTSUPP;
2222 		if (!ops->ndo_set_vf_port)
2223 			goto errout;
2224 
2225 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2226 			if (nla_type(attr) != IFLA_VF_PORT ||
2227 			    nla_len(attr) < NLA_HDRLEN) {
2228 				err = -EINVAL;
2229 				goto errout;
2230 			}
2231 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2232 					       ifla_port_policy, NULL);
2233 			if (err < 0)
2234 				goto errout;
2235 			if (!port[IFLA_PORT_VF]) {
2236 				err = -EOPNOTSUPP;
2237 				goto errout;
2238 			}
2239 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2240 			err = ops->ndo_set_vf_port(dev, vf, port);
2241 			if (err < 0)
2242 				goto errout;
2243 			status |= DO_SETLINK_NOTIFY;
2244 		}
2245 	}
2246 	err = 0;
2247 
2248 	if (tb[IFLA_PORT_SELF]) {
2249 		struct nlattr *port[IFLA_PORT_MAX+1];
2250 
2251 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2252 				       tb[IFLA_PORT_SELF], ifla_port_policy,
2253 				       NULL);
2254 		if (err < 0)
2255 			goto errout;
2256 
2257 		err = -EOPNOTSUPP;
2258 		if (ops->ndo_set_vf_port)
2259 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2260 		if (err < 0)
2261 			goto errout;
2262 		status |= DO_SETLINK_NOTIFY;
2263 	}
2264 
2265 	if (tb[IFLA_AF_SPEC]) {
2266 		struct nlattr *af;
2267 		int rem;
2268 
2269 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2270 			const struct rtnl_af_ops *af_ops;
2271 
2272 			rcu_read_lock();
2273 
2274 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2275 				BUG();
2276 
2277 			err = af_ops->set_link_af(dev, af);
2278 			if (err < 0) {
2279 				rcu_read_unlock();
2280 				goto errout;
2281 			}
2282 
2283 			rcu_read_unlock();
2284 			status |= DO_SETLINK_NOTIFY;
2285 		}
2286 	}
2287 	err = 0;
2288 
2289 	if (tb[IFLA_PROTO_DOWN]) {
2290 		err = dev_change_proto_down(dev,
2291 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2292 		if (err)
2293 			goto errout;
2294 		status |= DO_SETLINK_NOTIFY;
2295 	}
2296 
2297 	if (tb[IFLA_XDP]) {
2298 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2299 		u32 xdp_flags = 0;
2300 
2301 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2302 				       ifla_xdp_policy, NULL);
2303 		if (err < 0)
2304 			goto errout;
2305 
2306 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2307 			err = -EINVAL;
2308 			goto errout;
2309 		}
2310 
2311 		if (xdp[IFLA_XDP_FLAGS]) {
2312 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2313 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2314 				err = -EINVAL;
2315 				goto errout;
2316 			}
2317 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2318 				err = -EINVAL;
2319 				goto errout;
2320 			}
2321 		}
2322 
2323 		if (xdp[IFLA_XDP_FD]) {
2324 			err = dev_change_xdp_fd(dev, extack,
2325 						nla_get_s32(xdp[IFLA_XDP_FD]),
2326 						xdp_flags);
2327 			if (err)
2328 				goto errout;
2329 			status |= DO_SETLINK_NOTIFY;
2330 		}
2331 	}
2332 
2333 errout:
2334 	if (status & DO_SETLINK_MODIFIED) {
2335 		if (status & DO_SETLINK_NOTIFY)
2336 			netdev_state_change(dev);
2337 
2338 		if (err < 0)
2339 			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",
2340 					     dev->name);
2341 	}
2342 
2343 	return err;
2344 }
2345 
2346 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2347 			struct netlink_ext_ack *extack)
2348 {
2349 	struct net *net = sock_net(skb->sk);
2350 	struct ifinfomsg *ifm;
2351 	struct net_device *dev;
2352 	int err;
2353 	struct nlattr *tb[IFLA_MAX+1];
2354 	char ifname[IFNAMSIZ];
2355 
2356 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2357 			  extack);
2358 	if (err < 0)
2359 		goto errout;
2360 
2361 	if (tb[IFLA_IFNAME])
2362 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2363 	else
2364 		ifname[0] = '\0';
2365 
2366 	err = -EINVAL;
2367 	ifm = nlmsg_data(nlh);
2368 	if (ifm->ifi_index > 0)
2369 		dev = __dev_get_by_index(net, ifm->ifi_index);
2370 	else if (tb[IFLA_IFNAME])
2371 		dev = __dev_get_by_name(net, ifname);
2372 	else
2373 		goto errout;
2374 
2375 	if (dev == NULL) {
2376 		err = -ENODEV;
2377 		goto errout;
2378 	}
2379 
2380 	err = validate_linkmsg(dev, tb);
2381 	if (err < 0)
2382 		goto errout;
2383 
2384 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2385 errout:
2386 	return err;
2387 }
2388 
2389 static int rtnl_group_dellink(const struct net *net, int group)
2390 {
2391 	struct net_device *dev, *aux;
2392 	LIST_HEAD(list_kill);
2393 	bool found = false;
2394 
2395 	if (!group)
2396 		return -EPERM;
2397 
2398 	for_each_netdev(net, dev) {
2399 		if (dev->group == group) {
2400 			const struct rtnl_link_ops *ops;
2401 
2402 			found = true;
2403 			ops = dev->rtnl_link_ops;
2404 			if (!ops || !ops->dellink)
2405 				return -EOPNOTSUPP;
2406 		}
2407 	}
2408 
2409 	if (!found)
2410 		return -ENODEV;
2411 
2412 	for_each_netdev_safe(net, dev, aux) {
2413 		if (dev->group == group) {
2414 			const struct rtnl_link_ops *ops;
2415 
2416 			ops = dev->rtnl_link_ops;
2417 			ops->dellink(dev, &list_kill);
2418 		}
2419 	}
2420 	unregister_netdevice_many(&list_kill);
2421 
2422 	return 0;
2423 }
2424 
2425 int rtnl_delete_link(struct net_device *dev)
2426 {
2427 	const struct rtnl_link_ops *ops;
2428 	LIST_HEAD(list_kill);
2429 
2430 	ops = dev->rtnl_link_ops;
2431 	if (!ops || !ops->dellink)
2432 		return -EOPNOTSUPP;
2433 
2434 	ops->dellink(dev, &list_kill);
2435 	unregister_netdevice_many(&list_kill);
2436 
2437 	return 0;
2438 }
2439 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2440 
2441 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2442 			struct netlink_ext_ack *extack)
2443 {
2444 	struct net *net = sock_net(skb->sk);
2445 	struct net_device *dev;
2446 	struct ifinfomsg *ifm;
2447 	char ifname[IFNAMSIZ];
2448 	struct nlattr *tb[IFLA_MAX+1];
2449 	int err;
2450 
2451 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2452 	if (err < 0)
2453 		return err;
2454 
2455 	if (tb[IFLA_IFNAME])
2456 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2457 
2458 	ifm = nlmsg_data(nlh);
2459 	if (ifm->ifi_index > 0)
2460 		dev = __dev_get_by_index(net, ifm->ifi_index);
2461 	else if (tb[IFLA_IFNAME])
2462 		dev = __dev_get_by_name(net, ifname);
2463 	else if (tb[IFLA_GROUP])
2464 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2465 	else
2466 		return -EINVAL;
2467 
2468 	if (!dev)
2469 		return -ENODEV;
2470 
2471 	return rtnl_delete_link(dev);
2472 }
2473 
2474 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2475 {
2476 	unsigned int old_flags;
2477 	int err;
2478 
2479 	old_flags = dev->flags;
2480 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2481 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2482 		if (err < 0)
2483 			return err;
2484 	}
2485 
2486 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2487 
2488 	__dev_notify_flags(dev, old_flags, ~0U);
2489 	return 0;
2490 }
2491 EXPORT_SYMBOL(rtnl_configure_link);
2492 
2493 struct net_device *rtnl_create_link(struct net *net,
2494 	const char *ifname, unsigned char name_assign_type,
2495 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2496 {
2497 	struct net_device *dev;
2498 	unsigned int num_tx_queues = 1;
2499 	unsigned int num_rx_queues = 1;
2500 
2501 	if (tb[IFLA_NUM_TX_QUEUES])
2502 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2503 	else if (ops->get_num_tx_queues)
2504 		num_tx_queues = ops->get_num_tx_queues();
2505 
2506 	if (tb[IFLA_NUM_RX_QUEUES])
2507 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2508 	else if (ops->get_num_rx_queues)
2509 		num_rx_queues = ops->get_num_rx_queues();
2510 
2511 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2512 			       ops->setup, num_tx_queues, num_rx_queues);
2513 	if (!dev)
2514 		return ERR_PTR(-ENOMEM);
2515 
2516 	dev_net_set(dev, net);
2517 	dev->rtnl_link_ops = ops;
2518 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2519 
2520 	if (tb[IFLA_MTU])
2521 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2522 	if (tb[IFLA_ADDRESS]) {
2523 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2524 				nla_len(tb[IFLA_ADDRESS]));
2525 		dev->addr_assign_type = NET_ADDR_SET;
2526 	}
2527 	if (tb[IFLA_BROADCAST])
2528 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2529 				nla_len(tb[IFLA_BROADCAST]));
2530 	if (tb[IFLA_TXQLEN])
2531 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2532 	if (tb[IFLA_OPERSTATE])
2533 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2534 	if (tb[IFLA_LINKMODE])
2535 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2536 	if (tb[IFLA_GROUP])
2537 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2538 
2539 	return dev;
2540 }
2541 EXPORT_SYMBOL(rtnl_create_link);
2542 
2543 static int rtnl_group_changelink(const struct sk_buff *skb,
2544 		struct net *net, int group,
2545 		struct ifinfomsg *ifm,
2546 		struct netlink_ext_ack *extack,
2547 		struct nlattr **tb)
2548 {
2549 	struct net_device *dev, *aux;
2550 	int err;
2551 
2552 	for_each_netdev_safe(net, dev, aux) {
2553 		if (dev->group == group) {
2554 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2555 			if (err < 0)
2556 				return err;
2557 		}
2558 	}
2559 
2560 	return 0;
2561 }
2562 
2563 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2564 			struct netlink_ext_ack *extack)
2565 {
2566 	struct net *net = sock_net(skb->sk);
2567 	const struct rtnl_link_ops *ops;
2568 	const struct rtnl_link_ops *m_ops = NULL;
2569 	struct net_device *dev;
2570 	struct net_device *master_dev = NULL;
2571 	struct ifinfomsg *ifm;
2572 	char kind[MODULE_NAME_LEN];
2573 	char ifname[IFNAMSIZ];
2574 	struct nlattr *tb[IFLA_MAX+1];
2575 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2576 	unsigned char name_assign_type = NET_NAME_USER;
2577 	int err;
2578 
2579 #ifdef CONFIG_MODULES
2580 replay:
2581 #endif
2582 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2583 	if (err < 0)
2584 		return err;
2585 
2586 	if (tb[IFLA_IFNAME])
2587 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2588 	else
2589 		ifname[0] = '\0';
2590 
2591 	ifm = nlmsg_data(nlh);
2592 	if (ifm->ifi_index > 0)
2593 		dev = __dev_get_by_index(net, ifm->ifi_index);
2594 	else {
2595 		if (ifname[0])
2596 			dev = __dev_get_by_name(net, ifname);
2597 		else
2598 			dev = NULL;
2599 	}
2600 
2601 	if (dev) {
2602 		master_dev = netdev_master_upper_dev_get(dev);
2603 		if (master_dev)
2604 			m_ops = master_dev->rtnl_link_ops;
2605 	}
2606 
2607 	err = validate_linkmsg(dev, tb);
2608 	if (err < 0)
2609 		return err;
2610 
2611 	if (tb[IFLA_LINKINFO]) {
2612 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2613 				       tb[IFLA_LINKINFO], ifla_info_policy,
2614 				       NULL);
2615 		if (err < 0)
2616 			return err;
2617 	} else
2618 		memset(linkinfo, 0, sizeof(linkinfo));
2619 
2620 	if (linkinfo[IFLA_INFO_KIND]) {
2621 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2622 		ops = rtnl_link_ops_get(kind);
2623 	} else {
2624 		kind[0] = '\0';
2625 		ops = NULL;
2626 	}
2627 
2628 	if (1) {
2629 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2630 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2631 		struct nlattr **data = NULL;
2632 		struct nlattr **slave_data = NULL;
2633 		struct net *dest_net, *link_net = NULL;
2634 
2635 		if (ops) {
2636 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2637 				err = nla_parse_nested(attr, ops->maxtype,
2638 						       linkinfo[IFLA_INFO_DATA],
2639 						       ops->policy, NULL);
2640 				if (err < 0)
2641 					return err;
2642 				data = attr;
2643 			}
2644 			if (ops->validate) {
2645 				err = ops->validate(tb, data, extack);
2646 				if (err < 0)
2647 					return err;
2648 			}
2649 		}
2650 
2651 		if (m_ops) {
2652 			if (m_ops->slave_maxtype &&
2653 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2654 				err = nla_parse_nested(slave_attr,
2655 						       m_ops->slave_maxtype,
2656 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2657 						       m_ops->slave_policy,
2658 						       NULL);
2659 				if (err < 0)
2660 					return err;
2661 				slave_data = slave_attr;
2662 			}
2663 		}
2664 
2665 		if (dev) {
2666 			int status = 0;
2667 
2668 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2669 				return -EEXIST;
2670 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2671 				return -EOPNOTSUPP;
2672 
2673 			if (linkinfo[IFLA_INFO_DATA]) {
2674 				if (!ops || ops != dev->rtnl_link_ops ||
2675 				    !ops->changelink)
2676 					return -EOPNOTSUPP;
2677 
2678 				err = ops->changelink(dev, tb, data, extack);
2679 				if (err < 0)
2680 					return err;
2681 				status |= DO_SETLINK_NOTIFY;
2682 			}
2683 
2684 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2685 				if (!m_ops || !m_ops->slave_changelink)
2686 					return -EOPNOTSUPP;
2687 
2688 				err = m_ops->slave_changelink(master_dev, dev,
2689 							      tb, slave_data,
2690 							      extack);
2691 				if (err < 0)
2692 					return err;
2693 				status |= DO_SETLINK_NOTIFY;
2694 			}
2695 
2696 			return do_setlink(skb, dev, ifm, extack, tb, ifname,
2697 					  status);
2698 		}
2699 
2700 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2701 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2702 				return rtnl_group_changelink(skb, net,
2703 						nla_get_u32(tb[IFLA_GROUP]),
2704 						ifm, extack, tb);
2705 			return -ENODEV;
2706 		}
2707 
2708 		if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2709 			return -EOPNOTSUPP;
2710 
2711 		if (!ops) {
2712 #ifdef CONFIG_MODULES
2713 			if (kind[0]) {
2714 				__rtnl_unlock();
2715 				request_module("rtnl-link-%s", kind);
2716 				rtnl_lock();
2717 				ops = rtnl_link_ops_get(kind);
2718 				if (ops)
2719 					goto replay;
2720 			}
2721 #endif
2722 			return -EOPNOTSUPP;
2723 		}
2724 
2725 		if (!ops->setup)
2726 			return -EOPNOTSUPP;
2727 
2728 		if (!ifname[0]) {
2729 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2730 			name_assign_type = NET_NAME_ENUM;
2731 		}
2732 
2733 		dest_net = rtnl_link_get_net(net, tb);
2734 		if (IS_ERR(dest_net))
2735 			return PTR_ERR(dest_net);
2736 
2737 		err = -EPERM;
2738 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2739 			goto out;
2740 
2741 		if (tb[IFLA_LINK_NETNSID]) {
2742 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2743 
2744 			link_net = get_net_ns_by_id(dest_net, id);
2745 			if (!link_net) {
2746 				err =  -EINVAL;
2747 				goto out;
2748 			}
2749 			err = -EPERM;
2750 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2751 				goto out;
2752 		}
2753 
2754 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2755 				       name_assign_type, ops, tb);
2756 		if (IS_ERR(dev)) {
2757 			err = PTR_ERR(dev);
2758 			goto out;
2759 		}
2760 
2761 		dev->ifindex = ifm->ifi_index;
2762 
2763 		if (ops->newlink) {
2764 			err = ops->newlink(link_net ? : net, dev, tb, data,
2765 					   extack);
2766 			/* Drivers should call free_netdev() in ->destructor
2767 			 * and unregister it on failure after registration
2768 			 * so that device could be finally freed in rtnl_unlock.
2769 			 */
2770 			if (err < 0) {
2771 				/* If device is not registered at all, free it now */
2772 				if (dev->reg_state == NETREG_UNINITIALIZED)
2773 					free_netdev(dev);
2774 				goto out;
2775 			}
2776 		} else {
2777 			err = register_netdevice(dev);
2778 			if (err < 0) {
2779 				free_netdev(dev);
2780 				goto out;
2781 			}
2782 		}
2783 		err = rtnl_configure_link(dev, ifm);
2784 		if (err < 0)
2785 			goto out_unregister;
2786 		if (link_net) {
2787 			err = dev_change_net_namespace(dev, dest_net, ifname);
2788 			if (err < 0)
2789 				goto out_unregister;
2790 		}
2791 		if (tb[IFLA_MASTER]) {
2792 			err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
2793 					    extack);
2794 			if (err)
2795 				goto out_unregister;
2796 		}
2797 out:
2798 		if (link_net)
2799 			put_net(link_net);
2800 		put_net(dest_net);
2801 		return err;
2802 out_unregister:
2803 		if (ops->newlink) {
2804 			LIST_HEAD(list_kill);
2805 
2806 			ops->dellink(dev, &list_kill);
2807 			unregister_netdevice_many(&list_kill);
2808 		} else {
2809 			unregister_netdevice(dev);
2810 		}
2811 		goto out;
2812 	}
2813 }
2814 
2815 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2816 			struct netlink_ext_ack *extack)
2817 {
2818 	struct net *net = sock_net(skb->sk);
2819 	struct ifinfomsg *ifm;
2820 	char ifname[IFNAMSIZ];
2821 	struct nlattr *tb[IFLA_MAX+1];
2822 	struct net_device *dev = NULL;
2823 	struct sk_buff *nskb;
2824 	int err;
2825 	u32 ext_filter_mask = 0;
2826 
2827 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2828 	if (err < 0)
2829 		return err;
2830 
2831 	if (tb[IFLA_IFNAME])
2832 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2833 
2834 	if (tb[IFLA_EXT_MASK])
2835 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2836 
2837 	ifm = nlmsg_data(nlh);
2838 	if (ifm->ifi_index > 0)
2839 		dev = __dev_get_by_index(net, ifm->ifi_index);
2840 	else if (tb[IFLA_IFNAME])
2841 		dev = __dev_get_by_name(net, ifname);
2842 	else
2843 		return -EINVAL;
2844 
2845 	if (dev == NULL)
2846 		return -ENODEV;
2847 
2848 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2849 	if (nskb == NULL)
2850 		return -ENOBUFS;
2851 
2852 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2853 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0, NULL);
2854 	if (err < 0) {
2855 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2856 		WARN_ON(err == -EMSGSIZE);
2857 		kfree_skb(nskb);
2858 	} else
2859 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2860 
2861 	return err;
2862 }
2863 
2864 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2865 {
2866 	struct net *net = sock_net(skb->sk);
2867 	struct net_device *dev;
2868 	struct nlattr *tb[IFLA_MAX+1];
2869 	u32 ext_filter_mask = 0;
2870 	u16 min_ifinfo_dump_size = 0;
2871 	int hdrlen;
2872 
2873 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2874 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2875 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2876 
2877 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2878 		if (tb[IFLA_EXT_MASK])
2879 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2880 	}
2881 
2882 	if (!ext_filter_mask)
2883 		return NLMSG_GOODSIZE;
2884 	/*
2885 	 * traverse the list of net devices and compute the minimum
2886 	 * buffer size based upon the filter mask.
2887 	 */
2888 	rcu_read_lock();
2889 	for_each_netdev_rcu(net, dev) {
2890 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2891 					     if_nlmsg_size(dev,
2892 						           ext_filter_mask));
2893 	}
2894 	rcu_read_unlock();
2895 
2896 	return nlmsg_total_size(min_ifinfo_dump_size);
2897 }
2898 
2899 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2900 {
2901 	int idx;
2902 	int s_idx = cb->family;
2903 
2904 	if (s_idx == 0)
2905 		s_idx = 1;
2906 
2907 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2908 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2909 		struct rtnl_link *handlers;
2910 		rtnl_dumpit_func dumpit;
2911 
2912 		if (idx < s_idx || idx == PF_PACKET)
2913 			continue;
2914 
2915 		handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
2916 		if (!handlers)
2917 			continue;
2918 
2919 		dumpit = READ_ONCE(handlers[type].dumpit);
2920 		if (!dumpit)
2921 			continue;
2922 
2923 		if (idx > s_idx) {
2924 			memset(&cb->args[0], 0, sizeof(cb->args));
2925 			cb->prev_seq = 0;
2926 			cb->seq = 0;
2927 		}
2928 		if (dumpit(skb, cb))
2929 			break;
2930 	}
2931 	cb->family = idx;
2932 
2933 	return skb->len;
2934 }
2935 
2936 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2937 				       unsigned int change,
2938 				       u32 event, gfp_t flags, int *new_nsid)
2939 {
2940 	struct net *net = dev_net(dev);
2941 	struct sk_buff *skb;
2942 	int err = -ENOBUFS;
2943 	size_t if_info_size;
2944 
2945 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2946 	if (skb == NULL)
2947 		goto errout;
2948 
2949 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event,
2950 			       new_nsid);
2951 	if (err < 0) {
2952 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2953 		WARN_ON(err == -EMSGSIZE);
2954 		kfree_skb(skb);
2955 		goto errout;
2956 	}
2957 	return skb;
2958 errout:
2959 	if (err < 0)
2960 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2961 	return NULL;
2962 }
2963 
2964 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2965 {
2966 	struct net *net = dev_net(dev);
2967 
2968 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2969 }
2970 
2971 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2972 			       unsigned int change, u32 event,
2973 			       gfp_t flags, int *new_nsid)
2974 {
2975 	struct sk_buff *skb;
2976 
2977 	if (dev->reg_state != NETREG_REGISTERED)
2978 		return;
2979 
2980 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid);
2981 	if (skb)
2982 		rtmsg_ifinfo_send(skb, dev, flags);
2983 }
2984 
2985 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2986 		  gfp_t flags)
2987 {
2988 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags, NULL);
2989 }
2990 EXPORT_SYMBOL(rtmsg_ifinfo);
2991 
2992 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
2993 			 gfp_t flags, int *new_nsid)
2994 {
2995 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
2996 			   new_nsid);
2997 }
2998 
2999 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3000 				   struct net_device *dev,
3001 				   u8 *addr, u16 vid, u32 pid, u32 seq,
3002 				   int type, unsigned int flags,
3003 				   int nlflags, u16 ndm_state)
3004 {
3005 	struct nlmsghdr *nlh;
3006 	struct ndmsg *ndm;
3007 
3008 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3009 	if (!nlh)
3010 		return -EMSGSIZE;
3011 
3012 	ndm = nlmsg_data(nlh);
3013 	ndm->ndm_family  = AF_BRIDGE;
3014 	ndm->ndm_pad1	 = 0;
3015 	ndm->ndm_pad2    = 0;
3016 	ndm->ndm_flags	 = flags;
3017 	ndm->ndm_type	 = 0;
3018 	ndm->ndm_ifindex = dev->ifindex;
3019 	ndm->ndm_state   = ndm_state;
3020 
3021 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3022 		goto nla_put_failure;
3023 	if (vid)
3024 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3025 			goto nla_put_failure;
3026 
3027 	nlmsg_end(skb, nlh);
3028 	return 0;
3029 
3030 nla_put_failure:
3031 	nlmsg_cancel(skb, nlh);
3032 	return -EMSGSIZE;
3033 }
3034 
3035 static inline size_t rtnl_fdb_nlmsg_size(void)
3036 {
3037 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3038 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
3039 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
3040 	       0;
3041 }
3042 
3043 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3044 			    u16 ndm_state)
3045 {
3046 	struct net *net = dev_net(dev);
3047 	struct sk_buff *skb;
3048 	int err = -ENOBUFS;
3049 
3050 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3051 	if (!skb)
3052 		goto errout;
3053 
3054 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3055 				      0, 0, type, NTF_SELF, 0, ndm_state);
3056 	if (err < 0) {
3057 		kfree_skb(skb);
3058 		goto errout;
3059 	}
3060 
3061 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3062 	return;
3063 errout:
3064 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3065 }
3066 
3067 /**
3068  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3069  */
3070 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3071 		     struct nlattr *tb[],
3072 		     struct net_device *dev,
3073 		     const unsigned char *addr, u16 vid,
3074 		     u16 flags)
3075 {
3076 	int err = -EINVAL;
3077 
3078 	/* If aging addresses are supported device will need to
3079 	 * implement its own handler for this.
3080 	 */
3081 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3082 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3083 		return err;
3084 	}
3085 
3086 	if (vid) {
3087 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3088 		return err;
3089 	}
3090 
3091 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3092 		err = dev_uc_add_excl(dev, addr);
3093 	else if (is_multicast_ether_addr(addr))
3094 		err = dev_mc_add_excl(dev, addr);
3095 
3096 	/* Only return duplicate errors if NLM_F_EXCL is set */
3097 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3098 		err = 0;
3099 
3100 	return err;
3101 }
3102 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3103 
3104 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3105 			 struct netlink_ext_ack *extack)
3106 {
3107 	u16 vid = 0;
3108 
3109 	if (vlan_attr) {
3110 		if (nla_len(vlan_attr) != sizeof(u16)) {
3111 			NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3112 			return -EINVAL;
3113 		}
3114 
3115 		vid = nla_get_u16(vlan_attr);
3116 
3117 		if (!vid || vid >= VLAN_VID_MASK) {
3118 			NL_SET_ERR_MSG(extack, "invalid vlan id");
3119 			return -EINVAL;
3120 		}
3121 	}
3122 	*p_vid = vid;
3123 	return 0;
3124 }
3125 
3126 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3127 			struct netlink_ext_ack *extack)
3128 {
3129 	struct net *net = sock_net(skb->sk);
3130 	struct ndmsg *ndm;
3131 	struct nlattr *tb[NDA_MAX+1];
3132 	struct net_device *dev;
3133 	u8 *addr;
3134 	u16 vid;
3135 	int err;
3136 
3137 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3138 	if (err < 0)
3139 		return err;
3140 
3141 	ndm = nlmsg_data(nlh);
3142 	if (ndm->ndm_ifindex == 0) {
3143 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3144 		return -EINVAL;
3145 	}
3146 
3147 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3148 	if (dev == NULL) {
3149 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3150 		return -ENODEV;
3151 	}
3152 
3153 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3154 		NL_SET_ERR_MSG(extack, "invalid address");
3155 		return -EINVAL;
3156 	}
3157 
3158 	addr = nla_data(tb[NDA_LLADDR]);
3159 
3160 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3161 	if (err)
3162 		return err;
3163 
3164 	err = -EOPNOTSUPP;
3165 
3166 	/* Support fdb on master device the net/bridge default case */
3167 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3168 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3169 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3170 		const struct net_device_ops *ops = br_dev->netdev_ops;
3171 
3172 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3173 				       nlh->nlmsg_flags);
3174 		if (err)
3175 			goto out;
3176 		else
3177 			ndm->ndm_flags &= ~NTF_MASTER;
3178 	}
3179 
3180 	/* Embedded bridge, macvlan, and any other device support */
3181 	if ((ndm->ndm_flags & NTF_SELF)) {
3182 		if (dev->netdev_ops->ndo_fdb_add)
3183 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3184 							   vid,
3185 							   nlh->nlmsg_flags);
3186 		else
3187 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3188 					       nlh->nlmsg_flags);
3189 
3190 		if (!err) {
3191 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3192 					ndm->ndm_state);
3193 			ndm->ndm_flags &= ~NTF_SELF;
3194 		}
3195 	}
3196 out:
3197 	return err;
3198 }
3199 
3200 /**
3201  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3202  */
3203 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3204 		     struct nlattr *tb[],
3205 		     struct net_device *dev,
3206 		     const unsigned char *addr, u16 vid)
3207 {
3208 	int err = -EINVAL;
3209 
3210 	/* If aging addresses are supported device will need to
3211 	 * implement its own handler for this.
3212 	 */
3213 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3214 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3215 		return err;
3216 	}
3217 
3218 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3219 		err = dev_uc_del(dev, addr);
3220 	else if (is_multicast_ether_addr(addr))
3221 		err = dev_mc_del(dev, addr);
3222 
3223 	return err;
3224 }
3225 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3226 
3227 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3228 			struct netlink_ext_ack *extack)
3229 {
3230 	struct net *net = sock_net(skb->sk);
3231 	struct ndmsg *ndm;
3232 	struct nlattr *tb[NDA_MAX+1];
3233 	struct net_device *dev;
3234 	int err = -EINVAL;
3235 	__u8 *addr;
3236 	u16 vid;
3237 
3238 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3239 		return -EPERM;
3240 
3241 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3242 	if (err < 0)
3243 		return err;
3244 
3245 	ndm = nlmsg_data(nlh);
3246 	if (ndm->ndm_ifindex == 0) {
3247 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3248 		return -EINVAL;
3249 	}
3250 
3251 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3252 	if (dev == NULL) {
3253 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3254 		return -ENODEV;
3255 	}
3256 
3257 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3258 		NL_SET_ERR_MSG(extack, "invalid address");
3259 		return -EINVAL;
3260 	}
3261 
3262 	addr = nla_data(tb[NDA_LLADDR]);
3263 
3264 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3265 	if (err)
3266 		return err;
3267 
3268 	err = -EOPNOTSUPP;
3269 
3270 	/* Support fdb on master device the net/bridge default case */
3271 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3272 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3273 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3274 		const struct net_device_ops *ops = br_dev->netdev_ops;
3275 
3276 		if (ops->ndo_fdb_del)
3277 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3278 
3279 		if (err)
3280 			goto out;
3281 		else
3282 			ndm->ndm_flags &= ~NTF_MASTER;
3283 	}
3284 
3285 	/* Embedded bridge, macvlan, and any other device support */
3286 	if (ndm->ndm_flags & NTF_SELF) {
3287 		if (dev->netdev_ops->ndo_fdb_del)
3288 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3289 							   vid);
3290 		else
3291 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3292 
3293 		if (!err) {
3294 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3295 					ndm->ndm_state);
3296 			ndm->ndm_flags &= ~NTF_SELF;
3297 		}
3298 	}
3299 out:
3300 	return err;
3301 }
3302 
3303 static int nlmsg_populate_fdb(struct sk_buff *skb,
3304 			      struct netlink_callback *cb,
3305 			      struct net_device *dev,
3306 			      int *idx,
3307 			      struct netdev_hw_addr_list *list)
3308 {
3309 	struct netdev_hw_addr *ha;
3310 	int err;
3311 	u32 portid, seq;
3312 
3313 	portid = NETLINK_CB(cb->skb).portid;
3314 	seq = cb->nlh->nlmsg_seq;
3315 
3316 	list_for_each_entry(ha, &list->list, list) {
3317 		if (*idx < cb->args[2])
3318 			goto skip;
3319 
3320 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3321 					      portid, seq,
3322 					      RTM_NEWNEIGH, NTF_SELF,
3323 					      NLM_F_MULTI, NUD_PERMANENT);
3324 		if (err < 0)
3325 			return err;
3326 skip:
3327 		*idx += 1;
3328 	}
3329 	return 0;
3330 }
3331 
3332 /**
3333  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3334  * @nlh: netlink message header
3335  * @dev: netdevice
3336  *
3337  * Default netdevice operation to dump the existing unicast address list.
3338  * Returns number of addresses from list put in skb.
3339  */
3340 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3341 		      struct netlink_callback *cb,
3342 		      struct net_device *dev,
3343 		      struct net_device *filter_dev,
3344 		      int *idx)
3345 {
3346 	int err;
3347 
3348 	netif_addr_lock_bh(dev);
3349 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3350 	if (err)
3351 		goto out;
3352 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3353 out:
3354 	netif_addr_unlock_bh(dev);
3355 	return err;
3356 }
3357 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3358 
3359 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3360 {
3361 	struct net_device *dev;
3362 	struct nlattr *tb[IFLA_MAX+1];
3363 	struct net_device *br_dev = NULL;
3364 	const struct net_device_ops *ops = NULL;
3365 	const struct net_device_ops *cops = NULL;
3366 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3367 	struct net *net = sock_net(skb->sk);
3368 	struct hlist_head *head;
3369 	int brport_idx = 0;
3370 	int br_idx = 0;
3371 	int h, s_h;
3372 	int idx = 0, s_idx;
3373 	int err = 0;
3374 	int fidx = 0;
3375 
3376 	err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3377 			  IFLA_MAX, ifla_policy, NULL);
3378 	if (err < 0) {
3379 		return -EINVAL;
3380 	} else if (err == 0) {
3381 		if (tb[IFLA_MASTER])
3382 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3383 	}
3384 
3385 	brport_idx = ifm->ifi_index;
3386 
3387 	if (br_idx) {
3388 		br_dev = __dev_get_by_index(net, br_idx);
3389 		if (!br_dev)
3390 			return -ENODEV;
3391 
3392 		ops = br_dev->netdev_ops;
3393 	}
3394 
3395 	s_h = cb->args[0];
3396 	s_idx = cb->args[1];
3397 
3398 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3399 		idx = 0;
3400 		head = &net->dev_index_head[h];
3401 		hlist_for_each_entry(dev, head, index_hlist) {
3402 
3403 			if (brport_idx && (dev->ifindex != brport_idx))
3404 				continue;
3405 
3406 			if (!br_idx) { /* user did not specify a specific bridge */
3407 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3408 					br_dev = netdev_master_upper_dev_get(dev);
3409 					cops = br_dev->netdev_ops;
3410 				}
3411 			} else {
3412 				if (dev != br_dev &&
3413 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3414 					continue;
3415 
3416 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3417 				    !(dev->priv_flags & IFF_EBRIDGE))
3418 					continue;
3419 				cops = ops;
3420 			}
3421 
3422 			if (idx < s_idx)
3423 				goto cont;
3424 
3425 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3426 				if (cops && cops->ndo_fdb_dump) {
3427 					err = cops->ndo_fdb_dump(skb, cb,
3428 								br_dev, dev,
3429 								&fidx);
3430 					if (err == -EMSGSIZE)
3431 						goto out;
3432 				}
3433 			}
3434 
3435 			if (dev->netdev_ops->ndo_fdb_dump)
3436 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3437 								    dev, NULL,
3438 								    &fidx);
3439 			else
3440 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3441 							&fidx);
3442 			if (err == -EMSGSIZE)
3443 				goto out;
3444 
3445 			cops = NULL;
3446 
3447 			/* reset fdb offset to 0 for rest of the interfaces */
3448 			cb->args[2] = 0;
3449 			fidx = 0;
3450 cont:
3451 			idx++;
3452 		}
3453 	}
3454 
3455 out:
3456 	cb->args[0] = h;
3457 	cb->args[1] = idx;
3458 	cb->args[2] = fidx;
3459 
3460 	return skb->len;
3461 }
3462 
3463 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3464 			       unsigned int attrnum, unsigned int flag)
3465 {
3466 	if (mask & flag)
3467 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3468 	return 0;
3469 }
3470 
3471 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3472 			    struct net_device *dev, u16 mode,
3473 			    u32 flags, u32 mask, int nlflags,
3474 			    u32 filter_mask,
3475 			    int (*vlan_fill)(struct sk_buff *skb,
3476 					     struct net_device *dev,
3477 					     u32 filter_mask))
3478 {
3479 	struct nlmsghdr *nlh;
3480 	struct ifinfomsg *ifm;
3481 	struct nlattr *br_afspec;
3482 	struct nlattr *protinfo;
3483 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3484 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3485 	int err = 0;
3486 
3487 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3488 	if (nlh == NULL)
3489 		return -EMSGSIZE;
3490 
3491 	ifm = nlmsg_data(nlh);
3492 	ifm->ifi_family = AF_BRIDGE;
3493 	ifm->__ifi_pad = 0;
3494 	ifm->ifi_type = dev->type;
3495 	ifm->ifi_index = dev->ifindex;
3496 	ifm->ifi_flags = dev_get_flags(dev);
3497 	ifm->ifi_change = 0;
3498 
3499 
3500 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3501 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3502 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3503 	    (br_dev &&
3504 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3505 	    (dev->addr_len &&
3506 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3507 	    (dev->ifindex != dev_get_iflink(dev) &&
3508 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3509 		goto nla_put_failure;
3510 
3511 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3512 	if (!br_afspec)
3513 		goto nla_put_failure;
3514 
3515 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3516 		nla_nest_cancel(skb, br_afspec);
3517 		goto nla_put_failure;
3518 	}
3519 
3520 	if (mode != BRIDGE_MODE_UNDEF) {
3521 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3522 			nla_nest_cancel(skb, br_afspec);
3523 			goto nla_put_failure;
3524 		}
3525 	}
3526 	if (vlan_fill) {
3527 		err = vlan_fill(skb, dev, filter_mask);
3528 		if (err) {
3529 			nla_nest_cancel(skb, br_afspec);
3530 			goto nla_put_failure;
3531 		}
3532 	}
3533 	nla_nest_end(skb, br_afspec);
3534 
3535 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3536 	if (!protinfo)
3537 		goto nla_put_failure;
3538 
3539 	if (brport_nla_put_flag(skb, flags, mask,
3540 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3541 	    brport_nla_put_flag(skb, flags, mask,
3542 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3543 	    brport_nla_put_flag(skb, flags, mask,
3544 				IFLA_BRPORT_FAST_LEAVE,
3545 				BR_MULTICAST_FAST_LEAVE) ||
3546 	    brport_nla_put_flag(skb, flags, mask,
3547 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3548 	    brport_nla_put_flag(skb, flags, mask,
3549 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3550 	    brport_nla_put_flag(skb, flags, mask,
3551 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3552 	    brport_nla_put_flag(skb, flags, mask,
3553 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3554 	    brport_nla_put_flag(skb, flags, mask,
3555 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3556 		nla_nest_cancel(skb, protinfo);
3557 		goto nla_put_failure;
3558 	}
3559 
3560 	nla_nest_end(skb, protinfo);
3561 
3562 	nlmsg_end(skb, nlh);
3563 	return 0;
3564 nla_put_failure:
3565 	nlmsg_cancel(skb, nlh);
3566 	return err ? err : -EMSGSIZE;
3567 }
3568 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3569 
3570 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3571 {
3572 	struct net *net = sock_net(skb->sk);
3573 	struct net_device *dev;
3574 	int idx = 0;
3575 	u32 portid = NETLINK_CB(cb->skb).portid;
3576 	u32 seq = cb->nlh->nlmsg_seq;
3577 	u32 filter_mask = 0;
3578 	int err;
3579 
3580 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3581 		struct nlattr *extfilt;
3582 
3583 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3584 					  IFLA_EXT_MASK);
3585 		if (extfilt) {
3586 			if (nla_len(extfilt) < sizeof(filter_mask))
3587 				return -EINVAL;
3588 
3589 			filter_mask = nla_get_u32(extfilt);
3590 		}
3591 	}
3592 
3593 	rcu_read_lock();
3594 	for_each_netdev_rcu(net, dev) {
3595 		const struct net_device_ops *ops = dev->netdev_ops;
3596 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3597 
3598 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3599 			if (idx >= cb->args[0]) {
3600 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3601 						skb, portid, seq, dev,
3602 						filter_mask, NLM_F_MULTI);
3603 				if (err < 0 && err != -EOPNOTSUPP) {
3604 					if (likely(skb->len))
3605 						break;
3606 
3607 					goto out_err;
3608 				}
3609 			}
3610 			idx++;
3611 		}
3612 
3613 		if (ops->ndo_bridge_getlink) {
3614 			if (idx >= cb->args[0]) {
3615 				err = ops->ndo_bridge_getlink(skb, portid,
3616 							      seq, dev,
3617 							      filter_mask,
3618 							      NLM_F_MULTI);
3619 				if (err < 0 && err != -EOPNOTSUPP) {
3620 					if (likely(skb->len))
3621 						break;
3622 
3623 					goto out_err;
3624 				}
3625 			}
3626 			idx++;
3627 		}
3628 	}
3629 	err = skb->len;
3630 out_err:
3631 	rcu_read_unlock();
3632 	cb->args[0] = idx;
3633 
3634 	return err;
3635 }
3636 
3637 static inline size_t bridge_nlmsg_size(void)
3638 {
3639 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3640 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3641 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3642 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3643 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3644 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3645 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3646 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3647 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3648 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3649 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3650 }
3651 
3652 static int rtnl_bridge_notify(struct net_device *dev)
3653 {
3654 	struct net *net = dev_net(dev);
3655 	struct sk_buff *skb;
3656 	int err = -EOPNOTSUPP;
3657 
3658 	if (!dev->netdev_ops->ndo_bridge_getlink)
3659 		return 0;
3660 
3661 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3662 	if (!skb) {
3663 		err = -ENOMEM;
3664 		goto errout;
3665 	}
3666 
3667 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3668 	if (err < 0)
3669 		goto errout;
3670 
3671 	if (!skb->len)
3672 		goto errout;
3673 
3674 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3675 	return 0;
3676 errout:
3677 	WARN_ON(err == -EMSGSIZE);
3678 	kfree_skb(skb);
3679 	if (err)
3680 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3681 	return err;
3682 }
3683 
3684 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3685 			       struct netlink_ext_ack *extack)
3686 {
3687 	struct net *net = sock_net(skb->sk);
3688 	struct ifinfomsg *ifm;
3689 	struct net_device *dev;
3690 	struct nlattr *br_spec, *attr = NULL;
3691 	int rem, err = -EOPNOTSUPP;
3692 	u16 flags = 0;
3693 	bool have_flags = false;
3694 
3695 	if (nlmsg_len(nlh) < sizeof(*ifm))
3696 		return -EINVAL;
3697 
3698 	ifm = nlmsg_data(nlh);
3699 	if (ifm->ifi_family != AF_BRIDGE)
3700 		return -EPFNOSUPPORT;
3701 
3702 	dev = __dev_get_by_index(net, ifm->ifi_index);
3703 	if (!dev) {
3704 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3705 		return -ENODEV;
3706 	}
3707 
3708 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3709 	if (br_spec) {
3710 		nla_for_each_nested(attr, br_spec, rem) {
3711 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3712 				if (nla_len(attr) < sizeof(flags))
3713 					return -EINVAL;
3714 
3715 				have_flags = true;
3716 				flags = nla_get_u16(attr);
3717 				break;
3718 			}
3719 		}
3720 	}
3721 
3722 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3723 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3724 
3725 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3726 			err = -EOPNOTSUPP;
3727 			goto out;
3728 		}
3729 
3730 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3731 		if (err)
3732 			goto out;
3733 
3734 		flags &= ~BRIDGE_FLAGS_MASTER;
3735 	}
3736 
3737 	if ((flags & BRIDGE_FLAGS_SELF)) {
3738 		if (!dev->netdev_ops->ndo_bridge_setlink)
3739 			err = -EOPNOTSUPP;
3740 		else
3741 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3742 								  flags);
3743 		if (!err) {
3744 			flags &= ~BRIDGE_FLAGS_SELF;
3745 
3746 			/* Generate event to notify upper layer of bridge
3747 			 * change
3748 			 */
3749 			err = rtnl_bridge_notify(dev);
3750 		}
3751 	}
3752 
3753 	if (have_flags)
3754 		memcpy(nla_data(attr), &flags, sizeof(flags));
3755 out:
3756 	return err;
3757 }
3758 
3759 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3760 			       struct netlink_ext_ack *extack)
3761 {
3762 	struct net *net = sock_net(skb->sk);
3763 	struct ifinfomsg *ifm;
3764 	struct net_device *dev;
3765 	struct nlattr *br_spec, *attr = NULL;
3766 	int rem, err = -EOPNOTSUPP;
3767 	u16 flags = 0;
3768 	bool have_flags = false;
3769 
3770 	if (nlmsg_len(nlh) < sizeof(*ifm))
3771 		return -EINVAL;
3772 
3773 	ifm = nlmsg_data(nlh);
3774 	if (ifm->ifi_family != AF_BRIDGE)
3775 		return -EPFNOSUPPORT;
3776 
3777 	dev = __dev_get_by_index(net, ifm->ifi_index);
3778 	if (!dev) {
3779 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3780 		return -ENODEV;
3781 	}
3782 
3783 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3784 	if (br_spec) {
3785 		nla_for_each_nested(attr, br_spec, rem) {
3786 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3787 				if (nla_len(attr) < sizeof(flags))
3788 					return -EINVAL;
3789 
3790 				have_flags = true;
3791 				flags = nla_get_u16(attr);
3792 				break;
3793 			}
3794 		}
3795 	}
3796 
3797 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3798 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3799 
3800 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3801 			err = -EOPNOTSUPP;
3802 			goto out;
3803 		}
3804 
3805 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3806 		if (err)
3807 			goto out;
3808 
3809 		flags &= ~BRIDGE_FLAGS_MASTER;
3810 	}
3811 
3812 	if ((flags & BRIDGE_FLAGS_SELF)) {
3813 		if (!dev->netdev_ops->ndo_bridge_dellink)
3814 			err = -EOPNOTSUPP;
3815 		else
3816 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3817 								  flags);
3818 
3819 		if (!err) {
3820 			flags &= ~BRIDGE_FLAGS_SELF;
3821 
3822 			/* Generate event to notify upper layer of bridge
3823 			 * change
3824 			 */
3825 			err = rtnl_bridge_notify(dev);
3826 		}
3827 	}
3828 
3829 	if (have_flags)
3830 		memcpy(nla_data(attr), &flags, sizeof(flags));
3831 out:
3832 	return err;
3833 }
3834 
3835 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3836 {
3837 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3838 	       (!idxattr || idxattr == attrid);
3839 }
3840 
3841 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3842 static int rtnl_get_offload_stats_attr_size(int attr_id)
3843 {
3844 	switch (attr_id) {
3845 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3846 		return sizeof(struct rtnl_link_stats64);
3847 	}
3848 
3849 	return 0;
3850 }
3851 
3852 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3853 				  int *prividx)
3854 {
3855 	struct nlattr *attr = NULL;
3856 	int attr_id, size;
3857 	void *attr_data;
3858 	int err;
3859 
3860 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3861 	      dev->netdev_ops->ndo_get_offload_stats))
3862 		return -ENODATA;
3863 
3864 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3865 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3866 		if (attr_id < *prividx)
3867 			continue;
3868 
3869 		size = rtnl_get_offload_stats_attr_size(attr_id);
3870 		if (!size)
3871 			continue;
3872 
3873 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3874 			continue;
3875 
3876 		attr = nla_reserve_64bit(skb, attr_id, size,
3877 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
3878 		if (!attr)
3879 			goto nla_put_failure;
3880 
3881 		attr_data = nla_data(attr);
3882 		memset(attr_data, 0, size);
3883 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3884 							     attr_data);
3885 		if (err)
3886 			goto get_offload_stats_failure;
3887 	}
3888 
3889 	if (!attr)
3890 		return -ENODATA;
3891 
3892 	*prividx = 0;
3893 	return 0;
3894 
3895 nla_put_failure:
3896 	err = -EMSGSIZE;
3897 get_offload_stats_failure:
3898 	*prividx = attr_id;
3899 	return err;
3900 }
3901 
3902 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3903 {
3904 	int nla_size = 0;
3905 	int attr_id;
3906 	int size;
3907 
3908 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3909 	      dev->netdev_ops->ndo_get_offload_stats))
3910 		return 0;
3911 
3912 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3913 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3914 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3915 			continue;
3916 		size = rtnl_get_offload_stats_attr_size(attr_id);
3917 		nla_size += nla_total_size_64bit(size);
3918 	}
3919 
3920 	if (nla_size != 0)
3921 		nla_size += nla_total_size(0);
3922 
3923 	return nla_size;
3924 }
3925 
3926 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3927 			       int type, u32 pid, u32 seq, u32 change,
3928 			       unsigned int flags, unsigned int filter_mask,
3929 			       int *idxattr, int *prividx)
3930 {
3931 	struct if_stats_msg *ifsm;
3932 	struct nlmsghdr *nlh;
3933 	struct nlattr *attr;
3934 	int s_prividx = *prividx;
3935 	int err;
3936 
3937 	ASSERT_RTNL();
3938 
3939 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3940 	if (!nlh)
3941 		return -EMSGSIZE;
3942 
3943 	ifsm = nlmsg_data(nlh);
3944 	ifsm->family = PF_UNSPEC;
3945 	ifsm->pad1 = 0;
3946 	ifsm->pad2 = 0;
3947 	ifsm->ifindex = dev->ifindex;
3948 	ifsm->filter_mask = filter_mask;
3949 
3950 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3951 		struct rtnl_link_stats64 *sp;
3952 
3953 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3954 					 sizeof(struct rtnl_link_stats64),
3955 					 IFLA_STATS_UNSPEC);
3956 		if (!attr)
3957 			goto nla_put_failure;
3958 
3959 		sp = nla_data(attr);
3960 		dev_get_stats(dev, sp);
3961 	}
3962 
3963 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3964 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3965 
3966 		if (ops && ops->fill_linkxstats) {
3967 			*idxattr = IFLA_STATS_LINK_XSTATS;
3968 			attr = nla_nest_start(skb,
3969 					      IFLA_STATS_LINK_XSTATS);
3970 			if (!attr)
3971 				goto nla_put_failure;
3972 
3973 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3974 			nla_nest_end(skb, attr);
3975 			if (err)
3976 				goto nla_put_failure;
3977 			*idxattr = 0;
3978 		}
3979 	}
3980 
3981 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3982 			     *idxattr)) {
3983 		const struct rtnl_link_ops *ops = NULL;
3984 		const struct net_device *master;
3985 
3986 		master = netdev_master_upper_dev_get(dev);
3987 		if (master)
3988 			ops = master->rtnl_link_ops;
3989 		if (ops && ops->fill_linkxstats) {
3990 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3991 			attr = nla_nest_start(skb,
3992 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3993 			if (!attr)
3994 				goto nla_put_failure;
3995 
3996 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3997 			nla_nest_end(skb, attr);
3998 			if (err)
3999 				goto nla_put_failure;
4000 			*idxattr = 0;
4001 		}
4002 	}
4003 
4004 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4005 			     *idxattr)) {
4006 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4007 		attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4008 		if (!attr)
4009 			goto nla_put_failure;
4010 
4011 		err = rtnl_get_offload_stats(skb, dev, prividx);
4012 		if (err == -ENODATA)
4013 			nla_nest_cancel(skb, attr);
4014 		else
4015 			nla_nest_end(skb, attr);
4016 
4017 		if (err && err != -ENODATA)
4018 			goto nla_put_failure;
4019 		*idxattr = 0;
4020 	}
4021 
4022 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4023 		struct rtnl_af_ops *af_ops;
4024 
4025 		*idxattr = IFLA_STATS_AF_SPEC;
4026 		attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4027 		if (!attr)
4028 			goto nla_put_failure;
4029 
4030 		rcu_read_lock();
4031 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4032 			if (af_ops->fill_stats_af) {
4033 				struct nlattr *af;
4034 				int err;
4035 
4036 				af = nla_nest_start(skb, af_ops->family);
4037 				if (!af) {
4038 					rcu_read_unlock();
4039 					goto nla_put_failure;
4040 				}
4041 				err = af_ops->fill_stats_af(skb, dev);
4042 
4043 				if (err == -ENODATA) {
4044 					nla_nest_cancel(skb, af);
4045 				} else if (err < 0) {
4046 					rcu_read_unlock();
4047 					goto nla_put_failure;
4048 				}
4049 
4050 				nla_nest_end(skb, af);
4051 			}
4052 		}
4053 		rcu_read_unlock();
4054 
4055 		nla_nest_end(skb, attr);
4056 
4057 		*idxattr = 0;
4058 	}
4059 
4060 	nlmsg_end(skb, nlh);
4061 
4062 	return 0;
4063 
4064 nla_put_failure:
4065 	/* not a multi message or no progress mean a real error */
4066 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4067 		nlmsg_cancel(skb, nlh);
4068 	else
4069 		nlmsg_end(skb, nlh);
4070 
4071 	return -EMSGSIZE;
4072 }
4073 
4074 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4075 				  u32 filter_mask)
4076 {
4077 	size_t size = 0;
4078 
4079 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4080 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4081 
4082 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4083 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4084 		int attr = IFLA_STATS_LINK_XSTATS;
4085 
4086 		if (ops && ops->get_linkxstats_size) {
4087 			size += nla_total_size(ops->get_linkxstats_size(dev,
4088 									attr));
4089 			/* for IFLA_STATS_LINK_XSTATS */
4090 			size += nla_total_size(0);
4091 		}
4092 	}
4093 
4094 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4095 		struct net_device *_dev = (struct net_device *)dev;
4096 		const struct rtnl_link_ops *ops = NULL;
4097 		const struct net_device *master;
4098 
4099 		/* netdev_master_upper_dev_get can't take const */
4100 		master = netdev_master_upper_dev_get(_dev);
4101 		if (master)
4102 			ops = master->rtnl_link_ops;
4103 		if (ops && ops->get_linkxstats_size) {
4104 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4105 
4106 			size += nla_total_size(ops->get_linkxstats_size(dev,
4107 									attr));
4108 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
4109 			size += nla_total_size(0);
4110 		}
4111 	}
4112 
4113 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4114 		size += rtnl_get_offload_stats_size(dev);
4115 
4116 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4117 		struct rtnl_af_ops *af_ops;
4118 
4119 		/* for IFLA_STATS_AF_SPEC */
4120 		size += nla_total_size(0);
4121 
4122 		rcu_read_lock();
4123 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4124 			if (af_ops->get_stats_af_size) {
4125 				size += nla_total_size(
4126 					af_ops->get_stats_af_size(dev));
4127 
4128 				/* for AF_* */
4129 				size += nla_total_size(0);
4130 			}
4131 		}
4132 		rcu_read_unlock();
4133 	}
4134 
4135 	return size;
4136 }
4137 
4138 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4139 			  struct netlink_ext_ack *extack)
4140 {
4141 	struct net *net = sock_net(skb->sk);
4142 	struct net_device *dev = NULL;
4143 	int idxattr = 0, prividx = 0;
4144 	struct if_stats_msg *ifsm;
4145 	struct sk_buff *nskb;
4146 	u32 filter_mask;
4147 	int err;
4148 
4149 	if (nlmsg_len(nlh) < sizeof(*ifsm))
4150 		return -EINVAL;
4151 
4152 	ifsm = nlmsg_data(nlh);
4153 	if (ifsm->ifindex > 0)
4154 		dev = __dev_get_by_index(net, ifsm->ifindex);
4155 	else
4156 		return -EINVAL;
4157 
4158 	if (!dev)
4159 		return -ENODEV;
4160 
4161 	filter_mask = ifsm->filter_mask;
4162 	if (!filter_mask)
4163 		return -EINVAL;
4164 
4165 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4166 	if (!nskb)
4167 		return -ENOBUFS;
4168 
4169 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4170 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4171 				  0, filter_mask, &idxattr, &prividx);
4172 	if (err < 0) {
4173 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4174 		WARN_ON(err == -EMSGSIZE);
4175 		kfree_skb(nskb);
4176 	} else {
4177 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4178 	}
4179 
4180 	return err;
4181 }
4182 
4183 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4184 {
4185 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
4186 	struct net *net = sock_net(skb->sk);
4187 	unsigned int flags = NLM_F_MULTI;
4188 	struct if_stats_msg *ifsm;
4189 	struct hlist_head *head;
4190 	struct net_device *dev;
4191 	u32 filter_mask = 0;
4192 	int idx = 0;
4193 
4194 	s_h = cb->args[0];
4195 	s_idx = cb->args[1];
4196 	s_idxattr = cb->args[2];
4197 	s_prividx = cb->args[3];
4198 
4199 	cb->seq = net->dev_base_seq;
4200 
4201 	if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4202 		return -EINVAL;
4203 
4204 	ifsm = nlmsg_data(cb->nlh);
4205 	filter_mask = ifsm->filter_mask;
4206 	if (!filter_mask)
4207 		return -EINVAL;
4208 
4209 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4210 		idx = 0;
4211 		head = &net->dev_index_head[h];
4212 		hlist_for_each_entry(dev, head, index_hlist) {
4213 			if (idx < s_idx)
4214 				goto cont;
4215 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4216 						  NETLINK_CB(cb->skb).portid,
4217 						  cb->nlh->nlmsg_seq, 0,
4218 						  flags, filter_mask,
4219 						  &s_idxattr, &s_prividx);
4220 			/* If we ran out of room on the first message,
4221 			 * we're in trouble
4222 			 */
4223 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4224 
4225 			if (err < 0)
4226 				goto out;
4227 			s_prividx = 0;
4228 			s_idxattr = 0;
4229 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4230 cont:
4231 			idx++;
4232 		}
4233 	}
4234 out:
4235 	cb->args[3] = s_prividx;
4236 	cb->args[2] = s_idxattr;
4237 	cb->args[1] = idx;
4238 	cb->args[0] = h;
4239 
4240 	return skb->len;
4241 }
4242 
4243 /* Process one rtnetlink message. */
4244 
4245 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4246 			     struct netlink_ext_ack *extack)
4247 {
4248 	struct net *net = sock_net(skb->sk);
4249 	struct rtnl_link *handlers;
4250 	int err = -EOPNOTSUPP;
4251 	rtnl_doit_func doit;
4252 	unsigned int flags;
4253 	int kind;
4254 	int family;
4255 	int type;
4256 
4257 	type = nlh->nlmsg_type;
4258 	if (type > RTM_MAX)
4259 		return -EOPNOTSUPP;
4260 
4261 	type -= RTM_BASE;
4262 
4263 	/* All the messages must have at least 1 byte length */
4264 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4265 		return 0;
4266 
4267 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4268 	kind = type&3;
4269 
4270 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4271 		return -EPERM;
4272 
4273 	if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4274 		family = PF_UNSPEC;
4275 
4276 	rcu_read_lock();
4277 	handlers = rcu_dereference(rtnl_msg_handlers[family]);
4278 	if (!handlers) {
4279 		family = PF_UNSPEC;
4280 		handlers = rcu_dereference(rtnl_msg_handlers[family]);
4281 	}
4282 
4283 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4284 		struct sock *rtnl;
4285 		rtnl_dumpit_func dumpit;
4286 		u16 min_dump_alloc = 0;
4287 
4288 		dumpit = READ_ONCE(handlers[type].dumpit);
4289 		if (!dumpit) {
4290 			family = PF_UNSPEC;
4291 			handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4292 			if (!handlers)
4293 				goto err_unlock;
4294 
4295 			dumpit = READ_ONCE(handlers[type].dumpit);
4296 			if (!dumpit)
4297 				goto err_unlock;
4298 		}
4299 
4300 		refcount_inc(&rtnl_msg_handlers_ref[family]);
4301 
4302 		if (type == RTM_GETLINK - RTM_BASE)
4303 			min_dump_alloc = rtnl_calcit(skb, nlh);
4304 
4305 		rcu_read_unlock();
4306 
4307 		rtnl = net->rtnl;
4308 		{
4309 			struct netlink_dump_control c = {
4310 				.dump		= dumpit,
4311 				.min_dump_alloc	= min_dump_alloc,
4312 			};
4313 			err = netlink_dump_start(rtnl, skb, nlh, &c);
4314 		}
4315 		refcount_dec(&rtnl_msg_handlers_ref[family]);
4316 		return err;
4317 	}
4318 
4319 	doit = READ_ONCE(handlers[type].doit);
4320 	if (!doit) {
4321 		family = PF_UNSPEC;
4322 		handlers = rcu_dereference(rtnl_msg_handlers[family]);
4323 	}
4324 
4325 	flags = READ_ONCE(handlers[type].flags);
4326 	if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4327 		refcount_inc(&rtnl_msg_handlers_ref[family]);
4328 		doit = READ_ONCE(handlers[type].doit);
4329 		rcu_read_unlock();
4330 		if (doit)
4331 			err = doit(skb, nlh, extack);
4332 		refcount_dec(&rtnl_msg_handlers_ref[family]);
4333 		return err;
4334 	}
4335 
4336 	rcu_read_unlock();
4337 
4338 	rtnl_lock();
4339 	handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4340 	if (handlers) {
4341 		doit = READ_ONCE(handlers[type].doit);
4342 		if (doit)
4343 			err = doit(skb, nlh, extack);
4344 	}
4345 	rtnl_unlock();
4346 	return err;
4347 
4348 err_unlock:
4349 	rcu_read_unlock();
4350 	return -EOPNOTSUPP;
4351 }
4352 
4353 static void rtnetlink_rcv(struct sk_buff *skb)
4354 {
4355 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4356 }
4357 
4358 static int rtnetlink_bind(struct net *net, int group)
4359 {
4360 	switch (group) {
4361 	case RTNLGRP_IPV4_MROUTE_R:
4362 	case RTNLGRP_IPV6_MROUTE_R:
4363 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4364 			return -EPERM;
4365 		break;
4366 	}
4367 	return 0;
4368 }
4369 
4370 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4371 {
4372 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4373 
4374 	switch (event) {
4375 	case NETDEV_REBOOT:
4376 	case NETDEV_CHANGEADDR:
4377 	case NETDEV_CHANGENAME:
4378 	case NETDEV_FEAT_CHANGE:
4379 	case NETDEV_BONDING_FAILOVER:
4380 	case NETDEV_NOTIFY_PEERS:
4381 	case NETDEV_RESEND_IGMP:
4382 	case NETDEV_CHANGEINFODATA:
4383 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4384 				   GFP_KERNEL, NULL);
4385 		break;
4386 	default:
4387 		break;
4388 	}
4389 	return NOTIFY_DONE;
4390 }
4391 
4392 static struct notifier_block rtnetlink_dev_notifier = {
4393 	.notifier_call	= rtnetlink_event,
4394 };
4395 
4396 
4397 static int __net_init rtnetlink_net_init(struct net *net)
4398 {
4399 	struct sock *sk;
4400 	struct netlink_kernel_cfg cfg = {
4401 		.groups		= RTNLGRP_MAX,
4402 		.input		= rtnetlink_rcv,
4403 		.cb_mutex	= &rtnl_mutex,
4404 		.flags		= NL_CFG_F_NONROOT_RECV,
4405 		.bind		= rtnetlink_bind,
4406 	};
4407 
4408 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4409 	if (!sk)
4410 		return -ENOMEM;
4411 	net->rtnl = sk;
4412 	return 0;
4413 }
4414 
4415 static void __net_exit rtnetlink_net_exit(struct net *net)
4416 {
4417 	netlink_kernel_release(net->rtnl);
4418 	net->rtnl = NULL;
4419 }
4420 
4421 static struct pernet_operations rtnetlink_net_ops = {
4422 	.init = rtnetlink_net_init,
4423 	.exit = rtnetlink_net_exit,
4424 };
4425 
4426 void __init rtnetlink_init(void)
4427 {
4428 	int i;
4429 
4430 	for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4431 		refcount_set(&rtnl_msg_handlers_ref[i], 1);
4432 
4433 	if (register_pernet_subsys(&rtnetlink_net_ops))
4434 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
4435 
4436 	register_netdevice_notifier(&rtnetlink_dev_notifier);
4437 
4438 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4439 		      rtnl_dump_ifinfo, 0);
4440 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4441 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4442 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4443 
4444 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4445 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4446 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4447 
4448 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4449 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4450 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4451 
4452 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4453 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4454 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4455 
4456 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4457 		      0);
4458 }
4459