xref: /openbmc/linux/net/core/rtnetlink.c (revision 6f52b16c5b29b89d92c0e7236f4655dc8491ad70)
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(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(&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  * The caller must hold the rtnl_mutex.
483  */
484 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
485 {
486 	list_del(&ops->list);
487 }
488 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
489 
490 /**
491  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
492  * @ops: struct rtnl_af_ops * to unregister
493  */
494 void rtnl_af_unregister(struct rtnl_af_ops *ops)
495 {
496 	rtnl_lock();
497 	__rtnl_af_unregister(ops);
498 	rtnl_unlock();
499 }
500 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
501 
502 static size_t rtnl_link_get_af_size(const struct net_device *dev,
503 				    u32 ext_filter_mask)
504 {
505 	struct rtnl_af_ops *af_ops;
506 	size_t size;
507 
508 	/* IFLA_AF_SPEC */
509 	size = nla_total_size(sizeof(struct nlattr));
510 
511 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
512 		if (af_ops->get_link_af_size) {
513 			/* AF_* + nested data */
514 			size += nla_total_size(sizeof(struct nlattr)) +
515 				af_ops->get_link_af_size(dev, ext_filter_mask);
516 		}
517 	}
518 
519 	return size;
520 }
521 
522 static bool rtnl_have_link_slave_info(const struct net_device *dev)
523 {
524 	struct net_device *master_dev;
525 
526 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
527 	if (master_dev && master_dev->rtnl_link_ops)
528 		return true;
529 	return false;
530 }
531 
532 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
533 				     const struct net_device *dev)
534 {
535 	struct net_device *master_dev;
536 	const struct rtnl_link_ops *ops;
537 	struct nlattr *slave_data;
538 	int err;
539 
540 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
541 	if (!master_dev)
542 		return 0;
543 	ops = master_dev->rtnl_link_ops;
544 	if (!ops)
545 		return 0;
546 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
547 		return -EMSGSIZE;
548 	if (ops->fill_slave_info) {
549 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
550 		if (!slave_data)
551 			return -EMSGSIZE;
552 		err = ops->fill_slave_info(skb, master_dev, dev);
553 		if (err < 0)
554 			goto err_cancel_slave_data;
555 		nla_nest_end(skb, slave_data);
556 	}
557 	return 0;
558 
559 err_cancel_slave_data:
560 	nla_nest_cancel(skb, slave_data);
561 	return err;
562 }
563 
564 static int rtnl_link_info_fill(struct sk_buff *skb,
565 			       const struct net_device *dev)
566 {
567 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
568 	struct nlattr *data;
569 	int err;
570 
571 	if (!ops)
572 		return 0;
573 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
574 		return -EMSGSIZE;
575 	if (ops->fill_xstats) {
576 		err = ops->fill_xstats(skb, dev);
577 		if (err < 0)
578 			return err;
579 	}
580 	if (ops->fill_info) {
581 		data = nla_nest_start(skb, IFLA_INFO_DATA);
582 		if (data == NULL)
583 			return -EMSGSIZE;
584 		err = ops->fill_info(skb, dev);
585 		if (err < 0)
586 			goto err_cancel_data;
587 		nla_nest_end(skb, data);
588 	}
589 	return 0;
590 
591 err_cancel_data:
592 	nla_nest_cancel(skb, data);
593 	return err;
594 }
595 
596 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
597 {
598 	struct nlattr *linkinfo;
599 	int err = -EMSGSIZE;
600 
601 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
602 	if (linkinfo == NULL)
603 		goto out;
604 
605 	err = rtnl_link_info_fill(skb, dev);
606 	if (err < 0)
607 		goto err_cancel_link;
608 
609 	err = rtnl_link_slave_info_fill(skb, dev);
610 	if (err < 0)
611 		goto err_cancel_link;
612 
613 	nla_nest_end(skb, linkinfo);
614 	return 0;
615 
616 err_cancel_link:
617 	nla_nest_cancel(skb, linkinfo);
618 out:
619 	return err;
620 }
621 
622 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
623 {
624 	struct sock *rtnl = net->rtnl;
625 	int err = 0;
626 
627 	NETLINK_CB(skb).dst_group = group;
628 	if (echo)
629 		refcount_inc(&skb->users);
630 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
631 	if (echo)
632 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
633 	return err;
634 }
635 
636 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
637 {
638 	struct sock *rtnl = net->rtnl;
639 
640 	return nlmsg_unicast(rtnl, skb, pid);
641 }
642 EXPORT_SYMBOL(rtnl_unicast);
643 
644 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
645 		 struct nlmsghdr *nlh, gfp_t flags)
646 {
647 	struct sock *rtnl = net->rtnl;
648 	int report = 0;
649 
650 	if (nlh)
651 		report = nlmsg_report(nlh);
652 
653 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
654 }
655 EXPORT_SYMBOL(rtnl_notify);
656 
657 void rtnl_set_sk_err(struct net *net, u32 group, int error)
658 {
659 	struct sock *rtnl = net->rtnl;
660 
661 	netlink_set_err(rtnl, 0, group, error);
662 }
663 EXPORT_SYMBOL(rtnl_set_sk_err);
664 
665 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
666 {
667 	struct nlattr *mx;
668 	int i, valid = 0;
669 
670 	mx = nla_nest_start(skb, RTA_METRICS);
671 	if (mx == NULL)
672 		return -ENOBUFS;
673 
674 	for (i = 0; i < RTAX_MAX; i++) {
675 		if (metrics[i]) {
676 			if (i == RTAX_CC_ALGO - 1) {
677 				char tmp[TCP_CA_NAME_MAX], *name;
678 
679 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
680 				if (!name)
681 					continue;
682 				if (nla_put_string(skb, i + 1, name))
683 					goto nla_put_failure;
684 			} else if (i == RTAX_FEATURES - 1) {
685 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
686 
687 				if (!user_features)
688 					continue;
689 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
690 				if (nla_put_u32(skb, i + 1, user_features))
691 					goto nla_put_failure;
692 			} else {
693 				if (nla_put_u32(skb, i + 1, metrics[i]))
694 					goto nla_put_failure;
695 			}
696 			valid++;
697 		}
698 	}
699 
700 	if (!valid) {
701 		nla_nest_cancel(skb, mx);
702 		return 0;
703 	}
704 
705 	return nla_nest_end(skb, mx);
706 
707 nla_put_failure:
708 	nla_nest_cancel(skb, mx);
709 	return -EMSGSIZE;
710 }
711 EXPORT_SYMBOL(rtnetlink_put_metrics);
712 
713 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
714 		       long expires, u32 error)
715 {
716 	struct rta_cacheinfo ci = {
717 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
718 		.rta_used = dst->__use,
719 		.rta_clntref = atomic_read(&(dst->__refcnt)),
720 		.rta_error = error,
721 		.rta_id =  id,
722 	};
723 
724 	if (expires) {
725 		unsigned long clock;
726 
727 		clock = jiffies_to_clock_t(abs(expires));
728 		clock = min_t(unsigned long, clock, INT_MAX);
729 		ci.rta_expires = (expires > 0) ? clock : -clock;
730 	}
731 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
732 }
733 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
734 
735 static void set_operstate(struct net_device *dev, unsigned char transition)
736 {
737 	unsigned char operstate = dev->operstate;
738 
739 	switch (transition) {
740 	case IF_OPER_UP:
741 		if ((operstate == IF_OPER_DORMANT ||
742 		     operstate == IF_OPER_UNKNOWN) &&
743 		    !netif_dormant(dev))
744 			operstate = IF_OPER_UP;
745 		break;
746 
747 	case IF_OPER_DORMANT:
748 		if (operstate == IF_OPER_UP ||
749 		    operstate == IF_OPER_UNKNOWN)
750 			operstate = IF_OPER_DORMANT;
751 		break;
752 	}
753 
754 	if (dev->operstate != operstate) {
755 		write_lock_bh(&dev_base_lock);
756 		dev->operstate = operstate;
757 		write_unlock_bh(&dev_base_lock);
758 		netdev_state_change(dev);
759 	}
760 }
761 
762 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
763 {
764 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
765 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
766 }
767 
768 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
769 					   const struct ifinfomsg *ifm)
770 {
771 	unsigned int flags = ifm->ifi_flags;
772 
773 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
774 	if (ifm->ifi_change)
775 		flags = (flags & ifm->ifi_change) |
776 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
777 
778 	return flags;
779 }
780 
781 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
782 				 const struct rtnl_link_stats64 *b)
783 {
784 	a->rx_packets = b->rx_packets;
785 	a->tx_packets = b->tx_packets;
786 	a->rx_bytes = b->rx_bytes;
787 	a->tx_bytes = b->tx_bytes;
788 	a->rx_errors = b->rx_errors;
789 	a->tx_errors = b->tx_errors;
790 	a->rx_dropped = b->rx_dropped;
791 	a->tx_dropped = b->tx_dropped;
792 
793 	a->multicast = b->multicast;
794 	a->collisions = b->collisions;
795 
796 	a->rx_length_errors = b->rx_length_errors;
797 	a->rx_over_errors = b->rx_over_errors;
798 	a->rx_crc_errors = b->rx_crc_errors;
799 	a->rx_frame_errors = b->rx_frame_errors;
800 	a->rx_fifo_errors = b->rx_fifo_errors;
801 	a->rx_missed_errors = b->rx_missed_errors;
802 
803 	a->tx_aborted_errors = b->tx_aborted_errors;
804 	a->tx_carrier_errors = b->tx_carrier_errors;
805 	a->tx_fifo_errors = b->tx_fifo_errors;
806 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
807 	a->tx_window_errors = b->tx_window_errors;
808 
809 	a->rx_compressed = b->rx_compressed;
810 	a->tx_compressed = b->tx_compressed;
811 
812 	a->rx_nohandler = b->rx_nohandler;
813 }
814 
815 /* All VF info */
816 static inline int rtnl_vfinfo_size(const struct net_device *dev,
817 				   u32 ext_filter_mask)
818 {
819 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
820 		int num_vfs = dev_num_vf(dev->dev.parent);
821 		size_t size = nla_total_size(0);
822 		size += num_vfs *
823 			(nla_total_size(0) +
824 			 nla_total_size(sizeof(struct ifla_vf_mac)) +
825 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
826 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
827 			 nla_total_size(MAX_VLAN_LIST_LEN *
828 					sizeof(struct ifla_vf_vlan_info)) +
829 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
830 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
831 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
832 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
833 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
834 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
835 			 /* IFLA_VF_STATS_RX_PACKETS */
836 			 nla_total_size_64bit(sizeof(__u64)) +
837 			 /* IFLA_VF_STATS_TX_PACKETS */
838 			 nla_total_size_64bit(sizeof(__u64)) +
839 			 /* IFLA_VF_STATS_RX_BYTES */
840 			 nla_total_size_64bit(sizeof(__u64)) +
841 			 /* IFLA_VF_STATS_TX_BYTES */
842 			 nla_total_size_64bit(sizeof(__u64)) +
843 			 /* IFLA_VF_STATS_BROADCAST */
844 			 nla_total_size_64bit(sizeof(__u64)) +
845 			 /* IFLA_VF_STATS_MULTICAST */
846 			 nla_total_size_64bit(sizeof(__u64)) +
847 			 nla_total_size(sizeof(struct ifla_vf_trust)));
848 		return size;
849 	} else
850 		return 0;
851 }
852 
853 static size_t rtnl_port_size(const struct net_device *dev,
854 			     u32 ext_filter_mask)
855 {
856 	size_t port_size = nla_total_size(4)		/* PORT_VF */
857 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
858 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
859 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
860 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
861 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
862 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
863 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
864 		+ port_size;
865 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
866 		+ port_size;
867 
868 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
869 	    !(ext_filter_mask & RTEXT_FILTER_VF))
870 		return 0;
871 	if (dev_num_vf(dev->dev.parent))
872 		return port_self_size + vf_ports_size +
873 			vf_port_size * dev_num_vf(dev->dev.parent);
874 	else
875 		return port_self_size;
876 }
877 
878 static size_t rtnl_xdp_size(void)
879 {
880 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
881 			  nla_total_size(1) +	/* XDP_ATTACHED */
882 			  nla_total_size(4);	/* XDP_PROG_ID */
883 
884 	return xdp_size;
885 }
886 
887 static noinline size_t if_nlmsg_size(const struct net_device *dev,
888 				     u32 ext_filter_mask)
889 {
890 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
891 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
892 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
893 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
894 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
895 	       + nla_total_size(sizeof(struct rtnl_link_stats))
896 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
897 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
898 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
899 	       + nla_total_size(4) /* IFLA_TXQLEN */
900 	       + nla_total_size(4) /* IFLA_WEIGHT */
901 	       + nla_total_size(4) /* IFLA_MTU */
902 	       + nla_total_size(4) /* IFLA_LINK */
903 	       + nla_total_size(4) /* IFLA_MASTER */
904 	       + nla_total_size(1) /* IFLA_CARRIER */
905 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
906 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
907 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
908 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
909 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
910 	       + nla_total_size(1) /* IFLA_OPERSTATE */
911 	       + nla_total_size(1) /* IFLA_LINKMODE */
912 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
913 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
914 	       + nla_total_size(4) /* IFLA_GROUP */
915 	       + nla_total_size(ext_filter_mask
916 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
917 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
918 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
919 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
920 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
921 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
922 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
923 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
924 	       + rtnl_xdp_size() /* IFLA_XDP */
925 	       + nla_total_size(4)  /* IFLA_EVENT */
926 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
927 
928 }
929 
930 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
931 {
932 	struct nlattr *vf_ports;
933 	struct nlattr *vf_port;
934 	int vf;
935 	int err;
936 
937 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
938 	if (!vf_ports)
939 		return -EMSGSIZE;
940 
941 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
942 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
943 		if (!vf_port)
944 			goto nla_put_failure;
945 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
946 			goto nla_put_failure;
947 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
948 		if (err == -EMSGSIZE)
949 			goto nla_put_failure;
950 		if (err) {
951 			nla_nest_cancel(skb, vf_port);
952 			continue;
953 		}
954 		nla_nest_end(skb, vf_port);
955 	}
956 
957 	nla_nest_end(skb, vf_ports);
958 
959 	return 0;
960 
961 nla_put_failure:
962 	nla_nest_cancel(skb, vf_ports);
963 	return -EMSGSIZE;
964 }
965 
966 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
967 {
968 	struct nlattr *port_self;
969 	int err;
970 
971 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
972 	if (!port_self)
973 		return -EMSGSIZE;
974 
975 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
976 	if (err) {
977 		nla_nest_cancel(skb, port_self);
978 		return (err == -EMSGSIZE) ? err : 0;
979 	}
980 
981 	nla_nest_end(skb, port_self);
982 
983 	return 0;
984 }
985 
986 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
987 			  u32 ext_filter_mask)
988 {
989 	int err;
990 
991 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
992 	    !(ext_filter_mask & RTEXT_FILTER_VF))
993 		return 0;
994 
995 	err = rtnl_port_self_fill(skb, dev);
996 	if (err)
997 		return err;
998 
999 	if (dev_num_vf(dev->dev.parent)) {
1000 		err = rtnl_vf_ports_fill(skb, dev);
1001 		if (err)
1002 			return err;
1003 	}
1004 
1005 	return 0;
1006 }
1007 
1008 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1009 {
1010 	int err;
1011 	struct netdev_phys_item_id ppid;
1012 
1013 	err = dev_get_phys_port_id(dev, &ppid);
1014 	if (err) {
1015 		if (err == -EOPNOTSUPP)
1016 			return 0;
1017 		return err;
1018 	}
1019 
1020 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1021 		return -EMSGSIZE;
1022 
1023 	return 0;
1024 }
1025 
1026 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1027 {
1028 	char name[IFNAMSIZ];
1029 	int err;
1030 
1031 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1032 	if (err) {
1033 		if (err == -EOPNOTSUPP)
1034 			return 0;
1035 		return err;
1036 	}
1037 
1038 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1039 		return -EMSGSIZE;
1040 
1041 	return 0;
1042 }
1043 
1044 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1045 {
1046 	int err;
1047 	struct switchdev_attr attr = {
1048 		.orig_dev = dev,
1049 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1050 		.flags = SWITCHDEV_F_NO_RECURSE,
1051 	};
1052 
1053 	err = switchdev_port_attr_get(dev, &attr);
1054 	if (err) {
1055 		if (err == -EOPNOTSUPP)
1056 			return 0;
1057 		return err;
1058 	}
1059 
1060 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1061 		    attr.u.ppid.id))
1062 		return -EMSGSIZE;
1063 
1064 	return 0;
1065 }
1066 
1067 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1068 					      struct net_device *dev)
1069 {
1070 	struct rtnl_link_stats64 *sp;
1071 	struct nlattr *attr;
1072 
1073 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1074 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1075 	if (!attr)
1076 		return -EMSGSIZE;
1077 
1078 	sp = nla_data(attr);
1079 	dev_get_stats(dev, sp);
1080 
1081 	attr = nla_reserve(skb, IFLA_STATS,
1082 			   sizeof(struct rtnl_link_stats));
1083 	if (!attr)
1084 		return -EMSGSIZE;
1085 
1086 	copy_rtnl_link_stats(nla_data(attr), sp);
1087 
1088 	return 0;
1089 }
1090 
1091 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1092 					       struct net_device *dev,
1093 					       int vfs_num,
1094 					       struct nlattr *vfinfo)
1095 {
1096 	struct ifla_vf_rss_query_en vf_rss_query_en;
1097 	struct nlattr *vf, *vfstats, *vfvlanlist;
1098 	struct ifla_vf_link_state vf_linkstate;
1099 	struct ifla_vf_vlan_info vf_vlan_info;
1100 	struct ifla_vf_spoofchk vf_spoofchk;
1101 	struct ifla_vf_tx_rate vf_tx_rate;
1102 	struct ifla_vf_stats vf_stats;
1103 	struct ifla_vf_trust vf_trust;
1104 	struct ifla_vf_vlan vf_vlan;
1105 	struct ifla_vf_rate vf_rate;
1106 	struct ifla_vf_mac vf_mac;
1107 	struct ifla_vf_info ivi;
1108 
1109 	memset(&ivi, 0, sizeof(ivi));
1110 
1111 	/* Not all SR-IOV capable drivers support the
1112 	 * spoofcheck and "RSS query enable" query.  Preset to
1113 	 * -1 so the user space tool can detect that the driver
1114 	 * didn't report anything.
1115 	 */
1116 	ivi.spoofchk = -1;
1117 	ivi.rss_query_en = -1;
1118 	ivi.trusted = -1;
1119 	/* The default value for VF link state is "auto"
1120 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1121 	 */
1122 	ivi.linkstate = 0;
1123 	/* VLAN Protocol by default is 802.1Q */
1124 	ivi.vlan_proto = htons(ETH_P_8021Q);
1125 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1126 		return 0;
1127 
1128 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1129 
1130 	vf_mac.vf =
1131 		vf_vlan.vf =
1132 		vf_vlan_info.vf =
1133 		vf_rate.vf =
1134 		vf_tx_rate.vf =
1135 		vf_spoofchk.vf =
1136 		vf_linkstate.vf =
1137 		vf_rss_query_en.vf =
1138 		vf_trust.vf = ivi.vf;
1139 
1140 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1141 	vf_vlan.vlan = ivi.vlan;
1142 	vf_vlan.qos = ivi.qos;
1143 	vf_vlan_info.vlan = ivi.vlan;
1144 	vf_vlan_info.qos = ivi.qos;
1145 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1146 	vf_tx_rate.rate = ivi.max_tx_rate;
1147 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1148 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1149 	vf_spoofchk.setting = ivi.spoofchk;
1150 	vf_linkstate.link_state = ivi.linkstate;
1151 	vf_rss_query_en.setting = ivi.rss_query_en;
1152 	vf_trust.setting = ivi.trusted;
1153 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1154 	if (!vf)
1155 		goto nla_put_vfinfo_failure;
1156 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1157 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1158 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1159 		    &vf_rate) ||
1160 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1161 		    &vf_tx_rate) ||
1162 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1163 		    &vf_spoofchk) ||
1164 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1165 		    &vf_linkstate) ||
1166 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1167 		    sizeof(vf_rss_query_en),
1168 		    &vf_rss_query_en) ||
1169 	    nla_put(skb, IFLA_VF_TRUST,
1170 		    sizeof(vf_trust), &vf_trust))
1171 		goto nla_put_vf_failure;
1172 	vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1173 	if (!vfvlanlist)
1174 		goto nla_put_vf_failure;
1175 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1176 		    &vf_vlan_info)) {
1177 		nla_nest_cancel(skb, vfvlanlist);
1178 		goto nla_put_vf_failure;
1179 	}
1180 	nla_nest_end(skb, vfvlanlist);
1181 	memset(&vf_stats, 0, sizeof(vf_stats));
1182 	if (dev->netdev_ops->ndo_get_vf_stats)
1183 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1184 						&vf_stats);
1185 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1186 	if (!vfstats)
1187 		goto nla_put_vf_failure;
1188 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1189 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1190 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1191 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1192 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1193 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1194 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1195 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1196 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1197 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1198 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1199 			      vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1200 		nla_nest_cancel(skb, vfstats);
1201 		goto nla_put_vf_failure;
1202 	}
1203 	nla_nest_end(skb, vfstats);
1204 	nla_nest_end(skb, vf);
1205 	return 0;
1206 
1207 nla_put_vf_failure:
1208 	nla_nest_cancel(skb, vf);
1209 nla_put_vfinfo_failure:
1210 	nla_nest_cancel(skb, vfinfo);
1211 	return -EMSGSIZE;
1212 }
1213 
1214 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1215 {
1216 	struct rtnl_link_ifmap map;
1217 
1218 	memset(&map, 0, sizeof(map));
1219 	map.mem_start   = dev->mem_start;
1220 	map.mem_end     = dev->mem_end;
1221 	map.base_addr   = dev->base_addr;
1222 	map.irq         = dev->irq;
1223 	map.dma         = dev->dma;
1224 	map.port        = dev->if_port;
1225 
1226 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1227 		return -EMSGSIZE;
1228 
1229 	return 0;
1230 }
1231 
1232 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1233 {
1234 	const struct net_device_ops *ops = dev->netdev_ops;
1235 	const struct bpf_prog *generic_xdp_prog;
1236 
1237 	ASSERT_RTNL();
1238 
1239 	*prog_id = 0;
1240 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1241 	if (generic_xdp_prog) {
1242 		*prog_id = generic_xdp_prog->aux->id;
1243 		return XDP_ATTACHED_SKB;
1244 	}
1245 	if (!ops->ndo_xdp)
1246 		return XDP_ATTACHED_NONE;
1247 
1248 	return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1249 }
1250 
1251 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1252 {
1253 	struct nlattr *xdp;
1254 	u32 prog_id;
1255 	int err;
1256 
1257 	xdp = nla_nest_start(skb, IFLA_XDP);
1258 	if (!xdp)
1259 		return -EMSGSIZE;
1260 
1261 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1262 			 rtnl_xdp_attached_mode(dev, &prog_id));
1263 	if (err)
1264 		goto err_cancel;
1265 
1266 	if (prog_id) {
1267 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1268 		if (err)
1269 			goto err_cancel;
1270 	}
1271 
1272 	nla_nest_end(skb, xdp);
1273 	return 0;
1274 
1275 err_cancel:
1276 	nla_nest_cancel(skb, xdp);
1277 	return err;
1278 }
1279 
1280 static u32 rtnl_get_event(unsigned long event)
1281 {
1282 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1283 
1284 	switch (event) {
1285 	case NETDEV_REBOOT:
1286 		rtnl_event_type = IFLA_EVENT_REBOOT;
1287 		break;
1288 	case NETDEV_FEAT_CHANGE:
1289 		rtnl_event_type = IFLA_EVENT_FEATURES;
1290 		break;
1291 	case NETDEV_BONDING_FAILOVER:
1292 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1293 		break;
1294 	case NETDEV_NOTIFY_PEERS:
1295 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1296 		break;
1297 	case NETDEV_RESEND_IGMP:
1298 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1299 		break;
1300 	case NETDEV_CHANGEINFODATA:
1301 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1302 		break;
1303 	default:
1304 		break;
1305 	}
1306 
1307 	return rtnl_event_type;
1308 }
1309 
1310 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1311 			    int type, u32 pid, u32 seq, u32 change,
1312 			    unsigned int flags, u32 ext_filter_mask,
1313 			    u32 event)
1314 {
1315 	struct ifinfomsg *ifm;
1316 	struct nlmsghdr *nlh;
1317 	struct nlattr *af_spec;
1318 	struct rtnl_af_ops *af_ops;
1319 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1320 
1321 	ASSERT_RTNL();
1322 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1323 	if (nlh == NULL)
1324 		return -EMSGSIZE;
1325 
1326 	ifm = nlmsg_data(nlh);
1327 	ifm->ifi_family = AF_UNSPEC;
1328 	ifm->__ifi_pad = 0;
1329 	ifm->ifi_type = dev->type;
1330 	ifm->ifi_index = dev->ifindex;
1331 	ifm->ifi_flags = dev_get_flags(dev);
1332 	ifm->ifi_change = change;
1333 
1334 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1335 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1336 	    nla_put_u8(skb, IFLA_OPERSTATE,
1337 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1338 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1339 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1340 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1341 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1342 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1343 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1344 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1345 #ifdef CONFIG_RPS
1346 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1347 #endif
1348 	    (dev->ifindex != dev_get_iflink(dev) &&
1349 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1350 	    (upper_dev &&
1351 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1352 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1353 	    (dev->qdisc &&
1354 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1355 	    (dev->ifalias &&
1356 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1357 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1358 			atomic_read(&dev->carrier_changes)) ||
1359 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1360 		goto nla_put_failure;
1361 
1362 	if (event != IFLA_EVENT_NONE) {
1363 		if (nla_put_u32(skb, IFLA_EVENT, event))
1364 			goto nla_put_failure;
1365 	}
1366 
1367 	if (rtnl_fill_link_ifmap(skb, dev))
1368 		goto nla_put_failure;
1369 
1370 	if (dev->addr_len) {
1371 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1372 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1373 			goto nla_put_failure;
1374 	}
1375 
1376 	if (rtnl_phys_port_id_fill(skb, dev))
1377 		goto nla_put_failure;
1378 
1379 	if (rtnl_phys_port_name_fill(skb, dev))
1380 		goto nla_put_failure;
1381 
1382 	if (rtnl_phys_switch_id_fill(skb, dev))
1383 		goto nla_put_failure;
1384 
1385 	if (rtnl_fill_stats(skb, dev))
1386 		goto nla_put_failure;
1387 
1388 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1389 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1390 		goto nla_put_failure;
1391 
1392 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1393 	    ext_filter_mask & RTEXT_FILTER_VF) {
1394 		int i;
1395 		struct nlattr *vfinfo;
1396 		int num_vfs = dev_num_vf(dev->dev.parent);
1397 
1398 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1399 		if (!vfinfo)
1400 			goto nla_put_failure;
1401 		for (i = 0; i < num_vfs; i++) {
1402 			if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1403 				goto nla_put_failure;
1404 		}
1405 
1406 		nla_nest_end(skb, vfinfo);
1407 	}
1408 
1409 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1410 		goto nla_put_failure;
1411 
1412 	if (rtnl_xdp_fill(skb, dev))
1413 		goto nla_put_failure;
1414 
1415 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1416 		if (rtnl_link_fill(skb, dev) < 0)
1417 			goto nla_put_failure;
1418 	}
1419 
1420 	if (dev->rtnl_link_ops &&
1421 	    dev->rtnl_link_ops->get_link_net) {
1422 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1423 
1424 		if (!net_eq(dev_net(dev), link_net)) {
1425 			int id = peernet2id_alloc(dev_net(dev), link_net);
1426 
1427 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1428 				goto nla_put_failure;
1429 		}
1430 	}
1431 
1432 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1433 		goto nla_put_failure;
1434 
1435 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1436 		if (af_ops->fill_link_af) {
1437 			struct nlattr *af;
1438 			int err;
1439 
1440 			if (!(af = nla_nest_start(skb, af_ops->family)))
1441 				goto nla_put_failure;
1442 
1443 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1444 
1445 			/*
1446 			 * Caller may return ENODATA to indicate that there
1447 			 * was no data to be dumped. This is not an error, it
1448 			 * means we should trim the attribute header and
1449 			 * continue.
1450 			 */
1451 			if (err == -ENODATA)
1452 				nla_nest_cancel(skb, af);
1453 			else if (err < 0)
1454 				goto nla_put_failure;
1455 
1456 			nla_nest_end(skb, af);
1457 		}
1458 	}
1459 
1460 	nla_nest_end(skb, af_spec);
1461 
1462 	nlmsg_end(skb, nlh);
1463 	return 0;
1464 
1465 nla_put_failure:
1466 	nlmsg_cancel(skb, nlh);
1467 	return -EMSGSIZE;
1468 }
1469 
1470 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1471 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1472 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1473 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1474 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1475 	[IFLA_MTU]		= { .type = NLA_U32 },
1476 	[IFLA_LINK]		= { .type = NLA_U32 },
1477 	[IFLA_MASTER]		= { .type = NLA_U32 },
1478 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1479 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1480 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1481 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1482 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1483 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1484 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1485 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1486 	/* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1487 	 * allow 0-length string (needed to remove an alias).
1488 	 */
1489 	[IFLA_IFALIAS]	        = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1490 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1491 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1492 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1493 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1494 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1495 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1496 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1497 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1498 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1499 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1500 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1501 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1502 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1503 	[IFLA_XDP]		= { .type = NLA_NESTED },
1504 	[IFLA_EVENT]		= { .type = NLA_U32 },
1505 	[IFLA_GROUP]		= { .type = NLA_U32 },
1506 };
1507 
1508 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1509 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1510 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1511 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1512 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1513 };
1514 
1515 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1516 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1517 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1518 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1519 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1520 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1521 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1522 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1523 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1524 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1525 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1526 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1527 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1528 };
1529 
1530 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1531 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1532 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1533 				    .len = PORT_PROFILE_MAX },
1534 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1535 				      .len = PORT_UUID_MAX },
1536 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1537 				    .len = PORT_UUID_MAX },
1538 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1539 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1540 
1541 	/* Unused, but we need to keep it here since user space could
1542 	 * fill it. It's also broken with regard to NLA_BINARY use in
1543 	 * combination with structs.
1544 	 */
1545 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1546 				    .len = sizeof(struct ifla_port_vsi) },
1547 };
1548 
1549 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1550 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1551 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1552 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1553 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1554 };
1555 
1556 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1557 {
1558 	const struct rtnl_link_ops *ops = NULL;
1559 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1560 
1561 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1562 			     ifla_info_policy, NULL) < 0)
1563 		return NULL;
1564 
1565 	if (linfo[IFLA_INFO_KIND]) {
1566 		char kind[MODULE_NAME_LEN];
1567 
1568 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1569 		ops = rtnl_link_ops_get(kind);
1570 	}
1571 
1572 	return ops;
1573 }
1574 
1575 static bool link_master_filtered(struct net_device *dev, int master_idx)
1576 {
1577 	struct net_device *master;
1578 
1579 	if (!master_idx)
1580 		return false;
1581 
1582 	master = netdev_master_upper_dev_get(dev);
1583 	if (!master || master->ifindex != master_idx)
1584 		return true;
1585 
1586 	return false;
1587 }
1588 
1589 static bool link_kind_filtered(const struct net_device *dev,
1590 			       const struct rtnl_link_ops *kind_ops)
1591 {
1592 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1593 		return true;
1594 
1595 	return false;
1596 }
1597 
1598 static bool link_dump_filtered(struct net_device *dev,
1599 			       int master_idx,
1600 			       const struct rtnl_link_ops *kind_ops)
1601 {
1602 	if (link_master_filtered(dev, master_idx) ||
1603 	    link_kind_filtered(dev, kind_ops))
1604 		return true;
1605 
1606 	return false;
1607 }
1608 
1609 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1610 {
1611 	struct net *net = sock_net(skb->sk);
1612 	int h, s_h;
1613 	int idx = 0, s_idx;
1614 	struct net_device *dev;
1615 	struct hlist_head *head;
1616 	struct nlattr *tb[IFLA_MAX+1];
1617 	u32 ext_filter_mask = 0;
1618 	const struct rtnl_link_ops *kind_ops = NULL;
1619 	unsigned int flags = NLM_F_MULTI;
1620 	int master_idx = 0;
1621 	int err;
1622 	int hdrlen;
1623 
1624 	s_h = cb->args[0];
1625 	s_idx = cb->args[1];
1626 
1627 	/* A hack to preserve kernel<->userspace interface.
1628 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1629 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1630 	 * what iproute2 < v3.9.0 used.
1631 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1632 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1633 	 */
1634 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1635 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1636 
1637 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1638 			ifla_policy, NULL) >= 0) {
1639 		if (tb[IFLA_EXT_MASK])
1640 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1641 
1642 		if (tb[IFLA_MASTER])
1643 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1644 
1645 		if (tb[IFLA_LINKINFO])
1646 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1647 
1648 		if (master_idx || kind_ops)
1649 			flags |= NLM_F_DUMP_FILTERED;
1650 	}
1651 
1652 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1653 		idx = 0;
1654 		head = &net->dev_index_head[h];
1655 		hlist_for_each_entry(dev, head, index_hlist) {
1656 			if (link_dump_filtered(dev, master_idx, kind_ops))
1657 				goto cont;
1658 			if (idx < s_idx)
1659 				goto cont;
1660 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1661 					       NETLINK_CB(cb->skb).portid,
1662 					       cb->nlh->nlmsg_seq, 0,
1663 					       flags,
1664 					       ext_filter_mask, 0);
1665 
1666 			if (err < 0) {
1667 				if (likely(skb->len))
1668 					goto out;
1669 
1670 				goto out_err;
1671 			}
1672 cont:
1673 			idx++;
1674 		}
1675 	}
1676 out:
1677 	err = skb->len;
1678 out_err:
1679 	cb->args[1] = idx;
1680 	cb->args[0] = h;
1681 	cb->seq = net->dev_base_seq;
1682 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1683 
1684 	return err;
1685 }
1686 
1687 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1688 			struct netlink_ext_ack *exterr)
1689 {
1690 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1691 }
1692 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1693 
1694 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1695 {
1696 	struct net *net;
1697 	/* Examine the link attributes and figure out which
1698 	 * network namespace we are talking about.
1699 	 */
1700 	if (tb[IFLA_NET_NS_PID])
1701 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1702 	else if (tb[IFLA_NET_NS_FD])
1703 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1704 	else
1705 		net = get_net(src_net);
1706 	return net;
1707 }
1708 EXPORT_SYMBOL(rtnl_link_get_net);
1709 
1710 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1711 {
1712 	if (dev) {
1713 		if (tb[IFLA_ADDRESS] &&
1714 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1715 			return -EINVAL;
1716 
1717 		if (tb[IFLA_BROADCAST] &&
1718 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1719 			return -EINVAL;
1720 	}
1721 
1722 	if (tb[IFLA_AF_SPEC]) {
1723 		struct nlattr *af;
1724 		int rem, err;
1725 
1726 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1727 			const struct rtnl_af_ops *af_ops;
1728 
1729 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1730 				return -EAFNOSUPPORT;
1731 
1732 			if (!af_ops->set_link_af)
1733 				return -EOPNOTSUPP;
1734 
1735 			if (af_ops->validate_link_af) {
1736 				err = af_ops->validate_link_af(dev, af);
1737 				if (err < 0)
1738 					return err;
1739 			}
1740 		}
1741 	}
1742 
1743 	return 0;
1744 }
1745 
1746 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1747 				  int guid_type)
1748 {
1749 	const struct net_device_ops *ops = dev->netdev_ops;
1750 
1751 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1752 }
1753 
1754 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1755 {
1756 	if (dev->type != ARPHRD_INFINIBAND)
1757 		return -EOPNOTSUPP;
1758 
1759 	return handle_infiniband_guid(dev, ivt, guid_type);
1760 }
1761 
1762 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1763 {
1764 	const struct net_device_ops *ops = dev->netdev_ops;
1765 	int err = -EINVAL;
1766 
1767 	if (tb[IFLA_VF_MAC]) {
1768 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1769 
1770 		err = -EOPNOTSUPP;
1771 		if (ops->ndo_set_vf_mac)
1772 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1773 						  ivm->mac);
1774 		if (err < 0)
1775 			return err;
1776 	}
1777 
1778 	if (tb[IFLA_VF_VLAN]) {
1779 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1780 
1781 		err = -EOPNOTSUPP;
1782 		if (ops->ndo_set_vf_vlan)
1783 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1784 						   ivv->qos,
1785 						   htons(ETH_P_8021Q));
1786 		if (err < 0)
1787 			return err;
1788 	}
1789 
1790 	if (tb[IFLA_VF_VLAN_LIST]) {
1791 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1792 		struct nlattr *attr;
1793 		int rem, len = 0;
1794 
1795 		err = -EOPNOTSUPP;
1796 		if (!ops->ndo_set_vf_vlan)
1797 			return err;
1798 
1799 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1800 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1801 			    nla_len(attr) < NLA_HDRLEN) {
1802 				return -EINVAL;
1803 			}
1804 			if (len >= MAX_VLAN_LIST_LEN)
1805 				return -EOPNOTSUPP;
1806 			ivvl[len] = nla_data(attr);
1807 
1808 			len++;
1809 		}
1810 		if (len == 0)
1811 			return -EINVAL;
1812 
1813 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1814 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
1815 		if (err < 0)
1816 			return err;
1817 	}
1818 
1819 	if (tb[IFLA_VF_TX_RATE]) {
1820 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1821 		struct ifla_vf_info ivf;
1822 
1823 		err = -EOPNOTSUPP;
1824 		if (ops->ndo_get_vf_config)
1825 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1826 		if (err < 0)
1827 			return err;
1828 
1829 		err = -EOPNOTSUPP;
1830 		if (ops->ndo_set_vf_rate)
1831 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1832 						   ivf.min_tx_rate,
1833 						   ivt->rate);
1834 		if (err < 0)
1835 			return err;
1836 	}
1837 
1838 	if (tb[IFLA_VF_RATE]) {
1839 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1840 
1841 		err = -EOPNOTSUPP;
1842 		if (ops->ndo_set_vf_rate)
1843 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1844 						   ivt->min_tx_rate,
1845 						   ivt->max_tx_rate);
1846 		if (err < 0)
1847 			return err;
1848 	}
1849 
1850 	if (tb[IFLA_VF_SPOOFCHK]) {
1851 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1852 
1853 		err = -EOPNOTSUPP;
1854 		if (ops->ndo_set_vf_spoofchk)
1855 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1856 						       ivs->setting);
1857 		if (err < 0)
1858 			return err;
1859 	}
1860 
1861 	if (tb[IFLA_VF_LINK_STATE]) {
1862 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1863 
1864 		err = -EOPNOTSUPP;
1865 		if (ops->ndo_set_vf_link_state)
1866 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1867 							 ivl->link_state);
1868 		if (err < 0)
1869 			return err;
1870 	}
1871 
1872 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1873 		struct ifla_vf_rss_query_en *ivrssq_en;
1874 
1875 		err = -EOPNOTSUPP;
1876 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1877 		if (ops->ndo_set_vf_rss_query_en)
1878 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1879 							   ivrssq_en->setting);
1880 		if (err < 0)
1881 			return err;
1882 	}
1883 
1884 	if (tb[IFLA_VF_TRUST]) {
1885 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1886 
1887 		err = -EOPNOTSUPP;
1888 		if (ops->ndo_set_vf_trust)
1889 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1890 		if (err < 0)
1891 			return err;
1892 	}
1893 
1894 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1895 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1896 
1897 		if (!ops->ndo_set_vf_guid)
1898 			return -EOPNOTSUPP;
1899 
1900 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1901 	}
1902 
1903 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1904 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1905 
1906 		if (!ops->ndo_set_vf_guid)
1907 			return -EOPNOTSUPP;
1908 
1909 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1910 	}
1911 
1912 	return err;
1913 }
1914 
1915 static int do_set_master(struct net_device *dev, int ifindex)
1916 {
1917 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1918 	const struct net_device_ops *ops;
1919 	int err;
1920 
1921 	if (upper_dev) {
1922 		if (upper_dev->ifindex == ifindex)
1923 			return 0;
1924 		ops = upper_dev->netdev_ops;
1925 		if (ops->ndo_del_slave) {
1926 			err = ops->ndo_del_slave(upper_dev, dev);
1927 			if (err)
1928 				return err;
1929 		} else {
1930 			return -EOPNOTSUPP;
1931 		}
1932 	}
1933 
1934 	if (ifindex) {
1935 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1936 		if (!upper_dev)
1937 			return -EINVAL;
1938 		ops = upper_dev->netdev_ops;
1939 		if (ops->ndo_add_slave) {
1940 			err = ops->ndo_add_slave(upper_dev, dev);
1941 			if (err)
1942 				return err;
1943 		} else {
1944 			return -EOPNOTSUPP;
1945 		}
1946 	}
1947 	return 0;
1948 }
1949 
1950 #define DO_SETLINK_MODIFIED	0x01
1951 /* notify flag means notify + modified. */
1952 #define DO_SETLINK_NOTIFY	0x03
1953 static int do_setlink(const struct sk_buff *skb,
1954 		      struct net_device *dev, struct ifinfomsg *ifm,
1955 		      struct netlink_ext_ack *extack,
1956 		      struct nlattr **tb, char *ifname, int status)
1957 {
1958 	const struct net_device_ops *ops = dev->netdev_ops;
1959 	int err;
1960 
1961 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1962 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1963 		if (IS_ERR(net)) {
1964 			err = PTR_ERR(net);
1965 			goto errout;
1966 		}
1967 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1968 			put_net(net);
1969 			err = -EPERM;
1970 			goto errout;
1971 		}
1972 		err = dev_change_net_namespace(dev, net, ifname);
1973 		put_net(net);
1974 		if (err)
1975 			goto errout;
1976 		status |= DO_SETLINK_MODIFIED;
1977 	}
1978 
1979 	if (tb[IFLA_MAP]) {
1980 		struct rtnl_link_ifmap *u_map;
1981 		struct ifmap k_map;
1982 
1983 		if (!ops->ndo_set_config) {
1984 			err = -EOPNOTSUPP;
1985 			goto errout;
1986 		}
1987 
1988 		if (!netif_device_present(dev)) {
1989 			err = -ENODEV;
1990 			goto errout;
1991 		}
1992 
1993 		u_map = nla_data(tb[IFLA_MAP]);
1994 		k_map.mem_start = (unsigned long) u_map->mem_start;
1995 		k_map.mem_end = (unsigned long) u_map->mem_end;
1996 		k_map.base_addr = (unsigned short) u_map->base_addr;
1997 		k_map.irq = (unsigned char) u_map->irq;
1998 		k_map.dma = (unsigned char) u_map->dma;
1999 		k_map.port = (unsigned char) u_map->port;
2000 
2001 		err = ops->ndo_set_config(dev, &k_map);
2002 		if (err < 0)
2003 			goto errout;
2004 
2005 		status |= DO_SETLINK_NOTIFY;
2006 	}
2007 
2008 	if (tb[IFLA_ADDRESS]) {
2009 		struct sockaddr *sa;
2010 		int len;
2011 
2012 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2013 						  sizeof(*sa));
2014 		sa = kmalloc(len, GFP_KERNEL);
2015 		if (!sa) {
2016 			err = -ENOMEM;
2017 			goto errout;
2018 		}
2019 		sa->sa_family = dev->type;
2020 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2021 		       dev->addr_len);
2022 		err = dev_set_mac_address(dev, sa);
2023 		kfree(sa);
2024 		if (err)
2025 			goto errout;
2026 		status |= DO_SETLINK_MODIFIED;
2027 	}
2028 
2029 	if (tb[IFLA_MTU]) {
2030 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2031 		if (err < 0)
2032 			goto errout;
2033 		status |= DO_SETLINK_MODIFIED;
2034 	}
2035 
2036 	if (tb[IFLA_GROUP]) {
2037 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2038 		status |= DO_SETLINK_NOTIFY;
2039 	}
2040 
2041 	/*
2042 	 * Interface selected by interface index but interface
2043 	 * name provided implies that a name change has been
2044 	 * requested.
2045 	 */
2046 	if (ifm->ifi_index > 0 && ifname[0]) {
2047 		err = dev_change_name(dev, ifname);
2048 		if (err < 0)
2049 			goto errout;
2050 		status |= DO_SETLINK_MODIFIED;
2051 	}
2052 
2053 	if (tb[IFLA_IFALIAS]) {
2054 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2055 				    nla_len(tb[IFLA_IFALIAS]));
2056 		if (err < 0)
2057 			goto errout;
2058 		status |= DO_SETLINK_NOTIFY;
2059 	}
2060 
2061 	if (tb[IFLA_BROADCAST]) {
2062 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2063 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2064 	}
2065 
2066 	if (ifm->ifi_flags || ifm->ifi_change) {
2067 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2068 		if (err < 0)
2069 			goto errout;
2070 	}
2071 
2072 	if (tb[IFLA_MASTER]) {
2073 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2074 		if (err)
2075 			goto errout;
2076 		status |= DO_SETLINK_MODIFIED;
2077 	}
2078 
2079 	if (tb[IFLA_CARRIER]) {
2080 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2081 		if (err)
2082 			goto errout;
2083 		status |= DO_SETLINK_MODIFIED;
2084 	}
2085 
2086 	if (tb[IFLA_TXQLEN]) {
2087 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2088 		unsigned int orig_len = dev->tx_queue_len;
2089 
2090 		if (dev->tx_queue_len ^ value) {
2091 			dev->tx_queue_len = value;
2092 			err = call_netdevice_notifiers(
2093 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2094 			err = notifier_to_errno(err);
2095 			if (err) {
2096 				dev->tx_queue_len = orig_len;
2097 				goto errout;
2098 			}
2099 			status |= DO_SETLINK_MODIFIED;
2100 		}
2101 	}
2102 
2103 	if (tb[IFLA_OPERSTATE])
2104 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2105 
2106 	if (tb[IFLA_LINKMODE]) {
2107 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2108 
2109 		write_lock_bh(&dev_base_lock);
2110 		if (dev->link_mode ^ value)
2111 			status |= DO_SETLINK_NOTIFY;
2112 		dev->link_mode = value;
2113 		write_unlock_bh(&dev_base_lock);
2114 	}
2115 
2116 	if (tb[IFLA_VFINFO_LIST]) {
2117 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2118 		struct nlattr *attr;
2119 		int rem;
2120 
2121 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2122 			if (nla_type(attr) != IFLA_VF_INFO ||
2123 			    nla_len(attr) < NLA_HDRLEN) {
2124 				err = -EINVAL;
2125 				goto errout;
2126 			}
2127 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2128 					       ifla_vf_policy, NULL);
2129 			if (err < 0)
2130 				goto errout;
2131 			err = do_setvfinfo(dev, vfinfo);
2132 			if (err < 0)
2133 				goto errout;
2134 			status |= DO_SETLINK_NOTIFY;
2135 		}
2136 	}
2137 	err = 0;
2138 
2139 	if (tb[IFLA_VF_PORTS]) {
2140 		struct nlattr *port[IFLA_PORT_MAX+1];
2141 		struct nlattr *attr;
2142 		int vf;
2143 		int rem;
2144 
2145 		err = -EOPNOTSUPP;
2146 		if (!ops->ndo_set_vf_port)
2147 			goto errout;
2148 
2149 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2150 			if (nla_type(attr) != IFLA_VF_PORT ||
2151 			    nla_len(attr) < NLA_HDRLEN) {
2152 				err = -EINVAL;
2153 				goto errout;
2154 			}
2155 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2156 					       ifla_port_policy, NULL);
2157 			if (err < 0)
2158 				goto errout;
2159 			if (!port[IFLA_PORT_VF]) {
2160 				err = -EOPNOTSUPP;
2161 				goto errout;
2162 			}
2163 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2164 			err = ops->ndo_set_vf_port(dev, vf, port);
2165 			if (err < 0)
2166 				goto errout;
2167 			status |= DO_SETLINK_NOTIFY;
2168 		}
2169 	}
2170 	err = 0;
2171 
2172 	if (tb[IFLA_PORT_SELF]) {
2173 		struct nlattr *port[IFLA_PORT_MAX+1];
2174 
2175 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2176 				       tb[IFLA_PORT_SELF], ifla_port_policy,
2177 				       NULL);
2178 		if (err < 0)
2179 			goto errout;
2180 
2181 		err = -EOPNOTSUPP;
2182 		if (ops->ndo_set_vf_port)
2183 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2184 		if (err < 0)
2185 			goto errout;
2186 		status |= DO_SETLINK_NOTIFY;
2187 	}
2188 
2189 	if (tb[IFLA_AF_SPEC]) {
2190 		struct nlattr *af;
2191 		int rem;
2192 
2193 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2194 			const struct rtnl_af_ops *af_ops;
2195 
2196 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2197 				BUG();
2198 
2199 			err = af_ops->set_link_af(dev, af);
2200 			if (err < 0)
2201 				goto errout;
2202 
2203 			status |= DO_SETLINK_NOTIFY;
2204 		}
2205 	}
2206 	err = 0;
2207 
2208 	if (tb[IFLA_PROTO_DOWN]) {
2209 		err = dev_change_proto_down(dev,
2210 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2211 		if (err)
2212 			goto errout;
2213 		status |= DO_SETLINK_NOTIFY;
2214 	}
2215 
2216 	if (tb[IFLA_XDP]) {
2217 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2218 		u32 xdp_flags = 0;
2219 
2220 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2221 				       ifla_xdp_policy, NULL);
2222 		if (err < 0)
2223 			goto errout;
2224 
2225 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2226 			err = -EINVAL;
2227 			goto errout;
2228 		}
2229 
2230 		if (xdp[IFLA_XDP_FLAGS]) {
2231 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2232 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2233 				err = -EINVAL;
2234 				goto errout;
2235 			}
2236 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2237 				err = -EINVAL;
2238 				goto errout;
2239 			}
2240 		}
2241 
2242 		if (xdp[IFLA_XDP_FD]) {
2243 			err = dev_change_xdp_fd(dev, extack,
2244 						nla_get_s32(xdp[IFLA_XDP_FD]),
2245 						xdp_flags);
2246 			if (err)
2247 				goto errout;
2248 			status |= DO_SETLINK_NOTIFY;
2249 		}
2250 	}
2251 
2252 errout:
2253 	if (status & DO_SETLINK_MODIFIED) {
2254 		if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2255 			netdev_state_change(dev);
2256 
2257 		if (err < 0)
2258 			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",
2259 					     dev->name);
2260 	}
2261 
2262 	return err;
2263 }
2264 
2265 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2266 			struct netlink_ext_ack *extack)
2267 {
2268 	struct net *net = sock_net(skb->sk);
2269 	struct ifinfomsg *ifm;
2270 	struct net_device *dev;
2271 	int err;
2272 	struct nlattr *tb[IFLA_MAX+1];
2273 	char ifname[IFNAMSIZ];
2274 
2275 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2276 			  extack);
2277 	if (err < 0)
2278 		goto errout;
2279 
2280 	if (tb[IFLA_IFNAME])
2281 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2282 	else
2283 		ifname[0] = '\0';
2284 
2285 	err = -EINVAL;
2286 	ifm = nlmsg_data(nlh);
2287 	if (ifm->ifi_index > 0)
2288 		dev = __dev_get_by_index(net, ifm->ifi_index);
2289 	else if (tb[IFLA_IFNAME])
2290 		dev = __dev_get_by_name(net, ifname);
2291 	else
2292 		goto errout;
2293 
2294 	if (dev == NULL) {
2295 		err = -ENODEV;
2296 		goto errout;
2297 	}
2298 
2299 	err = validate_linkmsg(dev, tb);
2300 	if (err < 0)
2301 		goto errout;
2302 
2303 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2304 errout:
2305 	return err;
2306 }
2307 
2308 static int rtnl_group_dellink(const struct net *net, int group)
2309 {
2310 	struct net_device *dev, *aux;
2311 	LIST_HEAD(list_kill);
2312 	bool found = false;
2313 
2314 	if (!group)
2315 		return -EPERM;
2316 
2317 	for_each_netdev(net, dev) {
2318 		if (dev->group == group) {
2319 			const struct rtnl_link_ops *ops;
2320 
2321 			found = true;
2322 			ops = dev->rtnl_link_ops;
2323 			if (!ops || !ops->dellink)
2324 				return -EOPNOTSUPP;
2325 		}
2326 	}
2327 
2328 	if (!found)
2329 		return -ENODEV;
2330 
2331 	for_each_netdev_safe(net, dev, aux) {
2332 		if (dev->group == group) {
2333 			const struct rtnl_link_ops *ops;
2334 
2335 			ops = dev->rtnl_link_ops;
2336 			ops->dellink(dev, &list_kill);
2337 		}
2338 	}
2339 	unregister_netdevice_many(&list_kill);
2340 
2341 	return 0;
2342 }
2343 
2344 int rtnl_delete_link(struct net_device *dev)
2345 {
2346 	const struct rtnl_link_ops *ops;
2347 	LIST_HEAD(list_kill);
2348 
2349 	ops = dev->rtnl_link_ops;
2350 	if (!ops || !ops->dellink)
2351 		return -EOPNOTSUPP;
2352 
2353 	ops->dellink(dev, &list_kill);
2354 	unregister_netdevice_many(&list_kill);
2355 
2356 	return 0;
2357 }
2358 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2359 
2360 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2361 			struct netlink_ext_ack *extack)
2362 {
2363 	struct net *net = sock_net(skb->sk);
2364 	struct net_device *dev;
2365 	struct ifinfomsg *ifm;
2366 	char ifname[IFNAMSIZ];
2367 	struct nlattr *tb[IFLA_MAX+1];
2368 	int err;
2369 
2370 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2371 	if (err < 0)
2372 		return err;
2373 
2374 	if (tb[IFLA_IFNAME])
2375 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2376 
2377 	ifm = nlmsg_data(nlh);
2378 	if (ifm->ifi_index > 0)
2379 		dev = __dev_get_by_index(net, ifm->ifi_index);
2380 	else if (tb[IFLA_IFNAME])
2381 		dev = __dev_get_by_name(net, ifname);
2382 	else if (tb[IFLA_GROUP])
2383 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2384 	else
2385 		return -EINVAL;
2386 
2387 	if (!dev)
2388 		return -ENODEV;
2389 
2390 	return rtnl_delete_link(dev);
2391 }
2392 
2393 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2394 {
2395 	unsigned int old_flags;
2396 	int err;
2397 
2398 	old_flags = dev->flags;
2399 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2400 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2401 		if (err < 0)
2402 			return err;
2403 	}
2404 
2405 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2406 
2407 	__dev_notify_flags(dev, old_flags, ~0U);
2408 	return 0;
2409 }
2410 EXPORT_SYMBOL(rtnl_configure_link);
2411 
2412 struct net_device *rtnl_create_link(struct net *net,
2413 	const char *ifname, unsigned char name_assign_type,
2414 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2415 {
2416 	struct net_device *dev;
2417 	unsigned int num_tx_queues = 1;
2418 	unsigned int num_rx_queues = 1;
2419 
2420 	if (tb[IFLA_NUM_TX_QUEUES])
2421 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2422 	else if (ops->get_num_tx_queues)
2423 		num_tx_queues = ops->get_num_tx_queues();
2424 
2425 	if (tb[IFLA_NUM_RX_QUEUES])
2426 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2427 	else if (ops->get_num_rx_queues)
2428 		num_rx_queues = ops->get_num_rx_queues();
2429 
2430 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2431 			       ops->setup, num_tx_queues, num_rx_queues);
2432 	if (!dev)
2433 		return ERR_PTR(-ENOMEM);
2434 
2435 	dev_net_set(dev, net);
2436 	dev->rtnl_link_ops = ops;
2437 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2438 
2439 	if (tb[IFLA_MTU])
2440 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2441 	if (tb[IFLA_ADDRESS]) {
2442 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2443 				nla_len(tb[IFLA_ADDRESS]));
2444 		dev->addr_assign_type = NET_ADDR_SET;
2445 	}
2446 	if (tb[IFLA_BROADCAST])
2447 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2448 				nla_len(tb[IFLA_BROADCAST]));
2449 	if (tb[IFLA_TXQLEN])
2450 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2451 	if (tb[IFLA_OPERSTATE])
2452 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2453 	if (tb[IFLA_LINKMODE])
2454 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2455 	if (tb[IFLA_GROUP])
2456 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2457 
2458 	return dev;
2459 }
2460 EXPORT_SYMBOL(rtnl_create_link);
2461 
2462 static int rtnl_group_changelink(const struct sk_buff *skb,
2463 		struct net *net, int group,
2464 		struct ifinfomsg *ifm,
2465 		struct netlink_ext_ack *extack,
2466 		struct nlattr **tb)
2467 {
2468 	struct net_device *dev, *aux;
2469 	int err;
2470 
2471 	for_each_netdev_safe(net, dev, aux) {
2472 		if (dev->group == group) {
2473 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2474 			if (err < 0)
2475 				return err;
2476 		}
2477 	}
2478 
2479 	return 0;
2480 }
2481 
2482 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2483 			struct netlink_ext_ack *extack)
2484 {
2485 	struct net *net = sock_net(skb->sk);
2486 	const struct rtnl_link_ops *ops;
2487 	const struct rtnl_link_ops *m_ops = NULL;
2488 	struct net_device *dev;
2489 	struct net_device *master_dev = NULL;
2490 	struct ifinfomsg *ifm;
2491 	char kind[MODULE_NAME_LEN];
2492 	char ifname[IFNAMSIZ];
2493 	struct nlattr *tb[IFLA_MAX+1];
2494 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2495 	unsigned char name_assign_type = NET_NAME_USER;
2496 	int err;
2497 
2498 #ifdef CONFIG_MODULES
2499 replay:
2500 #endif
2501 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2502 	if (err < 0)
2503 		return err;
2504 
2505 	if (tb[IFLA_IFNAME])
2506 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2507 	else
2508 		ifname[0] = '\0';
2509 
2510 	ifm = nlmsg_data(nlh);
2511 	if (ifm->ifi_index > 0)
2512 		dev = __dev_get_by_index(net, ifm->ifi_index);
2513 	else {
2514 		if (ifname[0])
2515 			dev = __dev_get_by_name(net, ifname);
2516 		else
2517 			dev = NULL;
2518 	}
2519 
2520 	if (dev) {
2521 		master_dev = netdev_master_upper_dev_get(dev);
2522 		if (master_dev)
2523 			m_ops = master_dev->rtnl_link_ops;
2524 	}
2525 
2526 	err = validate_linkmsg(dev, tb);
2527 	if (err < 0)
2528 		return err;
2529 
2530 	if (tb[IFLA_LINKINFO]) {
2531 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2532 				       tb[IFLA_LINKINFO], ifla_info_policy,
2533 				       NULL);
2534 		if (err < 0)
2535 			return err;
2536 	} else
2537 		memset(linkinfo, 0, sizeof(linkinfo));
2538 
2539 	if (linkinfo[IFLA_INFO_KIND]) {
2540 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2541 		ops = rtnl_link_ops_get(kind);
2542 	} else {
2543 		kind[0] = '\0';
2544 		ops = NULL;
2545 	}
2546 
2547 	if (1) {
2548 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2549 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2550 		struct nlattr **data = NULL;
2551 		struct nlattr **slave_data = NULL;
2552 		struct net *dest_net, *link_net = NULL;
2553 
2554 		if (ops) {
2555 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2556 				err = nla_parse_nested(attr, ops->maxtype,
2557 						       linkinfo[IFLA_INFO_DATA],
2558 						       ops->policy, NULL);
2559 				if (err < 0)
2560 					return err;
2561 				data = attr;
2562 			}
2563 			if (ops->validate) {
2564 				err = ops->validate(tb, data, extack);
2565 				if (err < 0)
2566 					return err;
2567 			}
2568 		}
2569 
2570 		if (m_ops) {
2571 			if (m_ops->slave_maxtype &&
2572 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2573 				err = nla_parse_nested(slave_attr,
2574 						       m_ops->slave_maxtype,
2575 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2576 						       m_ops->slave_policy,
2577 						       NULL);
2578 				if (err < 0)
2579 					return err;
2580 				slave_data = slave_attr;
2581 			}
2582 			if (m_ops->slave_validate) {
2583 				err = m_ops->slave_validate(tb, slave_data,
2584 							    extack);
2585 				if (err < 0)
2586 					return err;
2587 			}
2588 		}
2589 
2590 		if (dev) {
2591 			int status = 0;
2592 
2593 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2594 				return -EEXIST;
2595 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2596 				return -EOPNOTSUPP;
2597 
2598 			if (linkinfo[IFLA_INFO_DATA]) {
2599 				if (!ops || ops != dev->rtnl_link_ops ||
2600 				    !ops->changelink)
2601 					return -EOPNOTSUPP;
2602 
2603 				err = ops->changelink(dev, tb, data, extack);
2604 				if (err < 0)
2605 					return err;
2606 				status |= DO_SETLINK_NOTIFY;
2607 			}
2608 
2609 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2610 				if (!m_ops || !m_ops->slave_changelink)
2611 					return -EOPNOTSUPP;
2612 
2613 				err = m_ops->slave_changelink(master_dev, dev,
2614 							      tb, slave_data,
2615 							      extack);
2616 				if (err < 0)
2617 					return err;
2618 				status |= DO_SETLINK_NOTIFY;
2619 			}
2620 
2621 			return do_setlink(skb, dev, ifm, extack, tb, ifname,
2622 					  status);
2623 		}
2624 
2625 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2626 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2627 				return rtnl_group_changelink(skb, net,
2628 						nla_get_u32(tb[IFLA_GROUP]),
2629 						ifm, extack, tb);
2630 			return -ENODEV;
2631 		}
2632 
2633 		if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2634 			return -EOPNOTSUPP;
2635 
2636 		if (!ops) {
2637 #ifdef CONFIG_MODULES
2638 			if (kind[0]) {
2639 				__rtnl_unlock();
2640 				request_module("rtnl-link-%s", kind);
2641 				rtnl_lock();
2642 				ops = rtnl_link_ops_get(kind);
2643 				if (ops)
2644 					goto replay;
2645 			}
2646 #endif
2647 			return -EOPNOTSUPP;
2648 		}
2649 
2650 		if (!ops->setup)
2651 			return -EOPNOTSUPP;
2652 
2653 		if (!ifname[0]) {
2654 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2655 			name_assign_type = NET_NAME_ENUM;
2656 		}
2657 
2658 		dest_net = rtnl_link_get_net(net, tb);
2659 		if (IS_ERR(dest_net))
2660 			return PTR_ERR(dest_net);
2661 
2662 		err = -EPERM;
2663 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2664 			goto out;
2665 
2666 		if (tb[IFLA_LINK_NETNSID]) {
2667 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2668 
2669 			link_net = get_net_ns_by_id(dest_net, id);
2670 			if (!link_net) {
2671 				err =  -EINVAL;
2672 				goto out;
2673 			}
2674 			err = -EPERM;
2675 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2676 				goto out;
2677 		}
2678 
2679 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2680 				       name_assign_type, ops, tb);
2681 		if (IS_ERR(dev)) {
2682 			err = PTR_ERR(dev);
2683 			goto out;
2684 		}
2685 
2686 		dev->ifindex = ifm->ifi_index;
2687 
2688 		if (ops->newlink) {
2689 			err = ops->newlink(link_net ? : net, dev, tb, data,
2690 					   extack);
2691 			/* Drivers should call free_netdev() in ->destructor
2692 			 * and unregister it on failure after registration
2693 			 * so that device could be finally freed in rtnl_unlock.
2694 			 */
2695 			if (err < 0) {
2696 				/* If device is not registered at all, free it now */
2697 				if (dev->reg_state == NETREG_UNINITIALIZED)
2698 					free_netdev(dev);
2699 				goto out;
2700 			}
2701 		} else {
2702 			err = register_netdevice(dev);
2703 			if (err < 0) {
2704 				free_netdev(dev);
2705 				goto out;
2706 			}
2707 		}
2708 		err = rtnl_configure_link(dev, ifm);
2709 		if (err < 0)
2710 			goto out_unregister;
2711 		if (link_net) {
2712 			err = dev_change_net_namespace(dev, dest_net, ifname);
2713 			if (err < 0)
2714 				goto out_unregister;
2715 		}
2716 		if (tb[IFLA_MASTER]) {
2717 			err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2718 			if (err)
2719 				goto out_unregister;
2720 		}
2721 out:
2722 		if (link_net)
2723 			put_net(link_net);
2724 		put_net(dest_net);
2725 		return err;
2726 out_unregister:
2727 		if (ops->newlink) {
2728 			LIST_HEAD(list_kill);
2729 
2730 			ops->dellink(dev, &list_kill);
2731 			unregister_netdevice_many(&list_kill);
2732 		} else {
2733 			unregister_netdevice(dev);
2734 		}
2735 		goto out;
2736 	}
2737 }
2738 
2739 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2740 			struct netlink_ext_ack *extack)
2741 {
2742 	struct net *net = sock_net(skb->sk);
2743 	struct ifinfomsg *ifm;
2744 	char ifname[IFNAMSIZ];
2745 	struct nlattr *tb[IFLA_MAX+1];
2746 	struct net_device *dev = NULL;
2747 	struct sk_buff *nskb;
2748 	int err;
2749 	u32 ext_filter_mask = 0;
2750 
2751 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2752 	if (err < 0)
2753 		return err;
2754 
2755 	if (tb[IFLA_IFNAME])
2756 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2757 
2758 	if (tb[IFLA_EXT_MASK])
2759 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2760 
2761 	ifm = nlmsg_data(nlh);
2762 	if (ifm->ifi_index > 0)
2763 		dev = __dev_get_by_index(net, ifm->ifi_index);
2764 	else if (tb[IFLA_IFNAME])
2765 		dev = __dev_get_by_name(net, ifname);
2766 	else
2767 		return -EINVAL;
2768 
2769 	if (dev == NULL)
2770 		return -ENODEV;
2771 
2772 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2773 	if (nskb == NULL)
2774 		return -ENOBUFS;
2775 
2776 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2777 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0);
2778 	if (err < 0) {
2779 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2780 		WARN_ON(err == -EMSGSIZE);
2781 		kfree_skb(nskb);
2782 	} else
2783 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2784 
2785 	return err;
2786 }
2787 
2788 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2789 {
2790 	struct net *net = sock_net(skb->sk);
2791 	struct net_device *dev;
2792 	struct nlattr *tb[IFLA_MAX+1];
2793 	u32 ext_filter_mask = 0;
2794 	u16 min_ifinfo_dump_size = 0;
2795 	int hdrlen;
2796 
2797 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2798 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2799 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2800 
2801 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2802 		if (tb[IFLA_EXT_MASK])
2803 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2804 	}
2805 
2806 	if (!ext_filter_mask)
2807 		return NLMSG_GOODSIZE;
2808 	/*
2809 	 * traverse the list of net devices and compute the minimum
2810 	 * buffer size based upon the filter mask.
2811 	 */
2812 	rcu_read_lock();
2813 	for_each_netdev_rcu(net, dev) {
2814 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2815 					     if_nlmsg_size(dev,
2816 						           ext_filter_mask));
2817 	}
2818 	rcu_read_unlock();
2819 
2820 	return nlmsg_total_size(min_ifinfo_dump_size);
2821 }
2822 
2823 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2824 {
2825 	int idx;
2826 	int s_idx = cb->family;
2827 
2828 	if (s_idx == 0)
2829 		s_idx = 1;
2830 
2831 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2832 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2833 		struct rtnl_link *handlers;
2834 		rtnl_dumpit_func dumpit;
2835 
2836 		if (idx < s_idx || idx == PF_PACKET)
2837 			continue;
2838 
2839 		handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
2840 		if (!handlers)
2841 			continue;
2842 
2843 		dumpit = READ_ONCE(handlers[type].dumpit);
2844 		if (!dumpit)
2845 			continue;
2846 
2847 		if (idx > s_idx) {
2848 			memset(&cb->args[0], 0, sizeof(cb->args));
2849 			cb->prev_seq = 0;
2850 			cb->seq = 0;
2851 		}
2852 		if (dumpit(skb, cb))
2853 			break;
2854 	}
2855 	cb->family = idx;
2856 
2857 	return skb->len;
2858 }
2859 
2860 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2861 				       unsigned int change,
2862 				       u32 event, gfp_t flags)
2863 {
2864 	struct net *net = dev_net(dev);
2865 	struct sk_buff *skb;
2866 	int err = -ENOBUFS;
2867 	size_t if_info_size;
2868 
2869 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2870 	if (skb == NULL)
2871 		goto errout;
2872 
2873 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event);
2874 	if (err < 0) {
2875 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2876 		WARN_ON(err == -EMSGSIZE);
2877 		kfree_skb(skb);
2878 		goto errout;
2879 	}
2880 	return skb;
2881 errout:
2882 	if (err < 0)
2883 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2884 	return NULL;
2885 }
2886 
2887 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2888 {
2889 	struct net *net = dev_net(dev);
2890 
2891 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2892 }
2893 
2894 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2895 			       unsigned int change, u32 event,
2896 			       gfp_t flags)
2897 {
2898 	struct sk_buff *skb;
2899 
2900 	if (dev->reg_state != NETREG_REGISTERED)
2901 		return;
2902 
2903 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags);
2904 	if (skb)
2905 		rtmsg_ifinfo_send(skb, dev, flags);
2906 }
2907 
2908 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2909 		  gfp_t flags)
2910 {
2911 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags);
2912 }
2913 EXPORT_SYMBOL(rtmsg_ifinfo);
2914 
2915 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2916 				   struct net_device *dev,
2917 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2918 				   int type, unsigned int flags,
2919 				   int nlflags, u16 ndm_state)
2920 {
2921 	struct nlmsghdr *nlh;
2922 	struct ndmsg *ndm;
2923 
2924 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2925 	if (!nlh)
2926 		return -EMSGSIZE;
2927 
2928 	ndm = nlmsg_data(nlh);
2929 	ndm->ndm_family  = AF_BRIDGE;
2930 	ndm->ndm_pad1	 = 0;
2931 	ndm->ndm_pad2    = 0;
2932 	ndm->ndm_flags	 = flags;
2933 	ndm->ndm_type	 = 0;
2934 	ndm->ndm_ifindex = dev->ifindex;
2935 	ndm->ndm_state   = ndm_state;
2936 
2937 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2938 		goto nla_put_failure;
2939 	if (vid)
2940 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2941 			goto nla_put_failure;
2942 
2943 	nlmsg_end(skb, nlh);
2944 	return 0;
2945 
2946 nla_put_failure:
2947 	nlmsg_cancel(skb, nlh);
2948 	return -EMSGSIZE;
2949 }
2950 
2951 static inline size_t rtnl_fdb_nlmsg_size(void)
2952 {
2953 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2954 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
2955 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
2956 	       0;
2957 }
2958 
2959 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2960 			    u16 ndm_state)
2961 {
2962 	struct net *net = dev_net(dev);
2963 	struct sk_buff *skb;
2964 	int err = -ENOBUFS;
2965 
2966 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2967 	if (!skb)
2968 		goto errout;
2969 
2970 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2971 				      0, 0, type, NTF_SELF, 0, ndm_state);
2972 	if (err < 0) {
2973 		kfree_skb(skb);
2974 		goto errout;
2975 	}
2976 
2977 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2978 	return;
2979 errout:
2980 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2981 }
2982 
2983 /**
2984  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2985  */
2986 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2987 		     struct nlattr *tb[],
2988 		     struct net_device *dev,
2989 		     const unsigned char *addr, u16 vid,
2990 		     u16 flags)
2991 {
2992 	int err = -EINVAL;
2993 
2994 	/* If aging addresses are supported device will need to
2995 	 * implement its own handler for this.
2996 	 */
2997 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2998 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2999 		return err;
3000 	}
3001 
3002 	if (vid) {
3003 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3004 		return err;
3005 	}
3006 
3007 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3008 		err = dev_uc_add_excl(dev, addr);
3009 	else if (is_multicast_ether_addr(addr))
3010 		err = dev_mc_add_excl(dev, addr);
3011 
3012 	/* Only return duplicate errors if NLM_F_EXCL is set */
3013 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3014 		err = 0;
3015 
3016 	return err;
3017 }
3018 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3019 
3020 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
3021 {
3022 	u16 vid = 0;
3023 
3024 	if (vlan_attr) {
3025 		if (nla_len(vlan_attr) != sizeof(u16)) {
3026 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
3027 			return -EINVAL;
3028 		}
3029 
3030 		vid = nla_get_u16(vlan_attr);
3031 
3032 		if (!vid || vid >= VLAN_VID_MASK) {
3033 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
3034 				vid);
3035 			return -EINVAL;
3036 		}
3037 	}
3038 	*p_vid = vid;
3039 	return 0;
3040 }
3041 
3042 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3043 			struct netlink_ext_ack *extack)
3044 {
3045 	struct net *net = sock_net(skb->sk);
3046 	struct ndmsg *ndm;
3047 	struct nlattr *tb[NDA_MAX+1];
3048 	struct net_device *dev;
3049 	u8 *addr;
3050 	u16 vid;
3051 	int err;
3052 
3053 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3054 	if (err < 0)
3055 		return err;
3056 
3057 	ndm = nlmsg_data(nlh);
3058 	if (ndm->ndm_ifindex == 0) {
3059 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3060 		return -EINVAL;
3061 	}
3062 
3063 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3064 	if (dev == NULL) {
3065 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3066 		return -ENODEV;
3067 	}
3068 
3069 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3070 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3071 		return -EINVAL;
3072 	}
3073 
3074 	addr = nla_data(tb[NDA_LLADDR]);
3075 
3076 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3077 	if (err)
3078 		return err;
3079 
3080 	err = -EOPNOTSUPP;
3081 
3082 	/* Support fdb on master device the net/bridge default case */
3083 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3084 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3085 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3086 		const struct net_device_ops *ops = br_dev->netdev_ops;
3087 
3088 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3089 				       nlh->nlmsg_flags);
3090 		if (err)
3091 			goto out;
3092 		else
3093 			ndm->ndm_flags &= ~NTF_MASTER;
3094 	}
3095 
3096 	/* Embedded bridge, macvlan, and any other device support */
3097 	if ((ndm->ndm_flags & NTF_SELF)) {
3098 		if (dev->netdev_ops->ndo_fdb_add)
3099 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3100 							   vid,
3101 							   nlh->nlmsg_flags);
3102 		else
3103 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3104 					       nlh->nlmsg_flags);
3105 
3106 		if (!err) {
3107 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3108 					ndm->ndm_state);
3109 			ndm->ndm_flags &= ~NTF_SELF;
3110 		}
3111 	}
3112 out:
3113 	return err;
3114 }
3115 
3116 /**
3117  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3118  */
3119 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3120 		     struct nlattr *tb[],
3121 		     struct net_device *dev,
3122 		     const unsigned char *addr, u16 vid)
3123 {
3124 	int err = -EINVAL;
3125 
3126 	/* If aging addresses are supported device will need to
3127 	 * implement its own handler for this.
3128 	 */
3129 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3130 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3131 		return err;
3132 	}
3133 
3134 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3135 		err = dev_uc_del(dev, addr);
3136 	else if (is_multicast_ether_addr(addr))
3137 		err = dev_mc_del(dev, addr);
3138 
3139 	return err;
3140 }
3141 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3142 
3143 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3144 			struct netlink_ext_ack *extack)
3145 {
3146 	struct net *net = sock_net(skb->sk);
3147 	struct ndmsg *ndm;
3148 	struct nlattr *tb[NDA_MAX+1];
3149 	struct net_device *dev;
3150 	int err = -EINVAL;
3151 	__u8 *addr;
3152 	u16 vid;
3153 
3154 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3155 		return -EPERM;
3156 
3157 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3158 	if (err < 0)
3159 		return err;
3160 
3161 	ndm = nlmsg_data(nlh);
3162 	if (ndm->ndm_ifindex == 0) {
3163 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3164 		return -EINVAL;
3165 	}
3166 
3167 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3168 	if (dev == NULL) {
3169 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3170 		return -ENODEV;
3171 	}
3172 
3173 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3174 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3175 		return -EINVAL;
3176 	}
3177 
3178 	addr = nla_data(tb[NDA_LLADDR]);
3179 
3180 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3181 	if (err)
3182 		return err;
3183 
3184 	err = -EOPNOTSUPP;
3185 
3186 	/* Support fdb on master device the net/bridge default case */
3187 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3188 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3189 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3190 		const struct net_device_ops *ops = br_dev->netdev_ops;
3191 
3192 		if (ops->ndo_fdb_del)
3193 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3194 
3195 		if (err)
3196 			goto out;
3197 		else
3198 			ndm->ndm_flags &= ~NTF_MASTER;
3199 	}
3200 
3201 	/* Embedded bridge, macvlan, and any other device support */
3202 	if (ndm->ndm_flags & NTF_SELF) {
3203 		if (dev->netdev_ops->ndo_fdb_del)
3204 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3205 							   vid);
3206 		else
3207 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3208 
3209 		if (!err) {
3210 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3211 					ndm->ndm_state);
3212 			ndm->ndm_flags &= ~NTF_SELF;
3213 		}
3214 	}
3215 out:
3216 	return err;
3217 }
3218 
3219 static int nlmsg_populate_fdb(struct sk_buff *skb,
3220 			      struct netlink_callback *cb,
3221 			      struct net_device *dev,
3222 			      int *idx,
3223 			      struct netdev_hw_addr_list *list)
3224 {
3225 	struct netdev_hw_addr *ha;
3226 	int err;
3227 	u32 portid, seq;
3228 
3229 	portid = NETLINK_CB(cb->skb).portid;
3230 	seq = cb->nlh->nlmsg_seq;
3231 
3232 	list_for_each_entry(ha, &list->list, list) {
3233 		if (*idx < cb->args[2])
3234 			goto skip;
3235 
3236 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3237 					      portid, seq,
3238 					      RTM_NEWNEIGH, NTF_SELF,
3239 					      NLM_F_MULTI, NUD_PERMANENT);
3240 		if (err < 0)
3241 			return err;
3242 skip:
3243 		*idx += 1;
3244 	}
3245 	return 0;
3246 }
3247 
3248 /**
3249  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3250  * @nlh: netlink message header
3251  * @dev: netdevice
3252  *
3253  * Default netdevice operation to dump the existing unicast address list.
3254  * Returns number of addresses from list put in skb.
3255  */
3256 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3257 		      struct netlink_callback *cb,
3258 		      struct net_device *dev,
3259 		      struct net_device *filter_dev,
3260 		      int *idx)
3261 {
3262 	int err;
3263 
3264 	netif_addr_lock_bh(dev);
3265 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3266 	if (err)
3267 		goto out;
3268 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3269 out:
3270 	netif_addr_unlock_bh(dev);
3271 	return err;
3272 }
3273 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3274 
3275 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3276 {
3277 	struct net_device *dev;
3278 	struct nlattr *tb[IFLA_MAX+1];
3279 	struct net_device *br_dev = NULL;
3280 	const struct net_device_ops *ops = NULL;
3281 	const struct net_device_ops *cops = NULL;
3282 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3283 	struct net *net = sock_net(skb->sk);
3284 	struct hlist_head *head;
3285 	int brport_idx = 0;
3286 	int br_idx = 0;
3287 	int h, s_h;
3288 	int idx = 0, s_idx;
3289 	int err = 0;
3290 	int fidx = 0;
3291 
3292 	err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3293 			  IFLA_MAX, ifla_policy, NULL);
3294 	if (err < 0) {
3295 		return -EINVAL;
3296 	} else if (err == 0) {
3297 		if (tb[IFLA_MASTER])
3298 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3299 	}
3300 
3301 	brport_idx = ifm->ifi_index;
3302 
3303 	if (br_idx) {
3304 		br_dev = __dev_get_by_index(net, br_idx);
3305 		if (!br_dev)
3306 			return -ENODEV;
3307 
3308 		ops = br_dev->netdev_ops;
3309 	}
3310 
3311 	s_h = cb->args[0];
3312 	s_idx = cb->args[1];
3313 
3314 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3315 		idx = 0;
3316 		head = &net->dev_index_head[h];
3317 		hlist_for_each_entry(dev, head, index_hlist) {
3318 
3319 			if (brport_idx && (dev->ifindex != brport_idx))
3320 				continue;
3321 
3322 			if (!br_idx) { /* user did not specify a specific bridge */
3323 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3324 					br_dev = netdev_master_upper_dev_get(dev);
3325 					cops = br_dev->netdev_ops;
3326 				}
3327 			} else {
3328 				if (dev != br_dev &&
3329 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3330 					continue;
3331 
3332 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3333 				    !(dev->priv_flags & IFF_EBRIDGE))
3334 					continue;
3335 				cops = ops;
3336 			}
3337 
3338 			if (idx < s_idx)
3339 				goto cont;
3340 
3341 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3342 				if (cops && cops->ndo_fdb_dump) {
3343 					err = cops->ndo_fdb_dump(skb, cb,
3344 								br_dev, dev,
3345 								&fidx);
3346 					if (err == -EMSGSIZE)
3347 						goto out;
3348 				}
3349 			}
3350 
3351 			if (dev->netdev_ops->ndo_fdb_dump)
3352 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3353 								    dev, NULL,
3354 								    &fidx);
3355 			else
3356 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3357 							&fidx);
3358 			if (err == -EMSGSIZE)
3359 				goto out;
3360 
3361 			cops = NULL;
3362 
3363 			/* reset fdb offset to 0 for rest of the interfaces */
3364 			cb->args[2] = 0;
3365 			fidx = 0;
3366 cont:
3367 			idx++;
3368 		}
3369 	}
3370 
3371 out:
3372 	cb->args[0] = h;
3373 	cb->args[1] = idx;
3374 	cb->args[2] = fidx;
3375 
3376 	return skb->len;
3377 }
3378 
3379 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3380 			       unsigned int attrnum, unsigned int flag)
3381 {
3382 	if (mask & flag)
3383 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3384 	return 0;
3385 }
3386 
3387 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3388 			    struct net_device *dev, u16 mode,
3389 			    u32 flags, u32 mask, int nlflags,
3390 			    u32 filter_mask,
3391 			    int (*vlan_fill)(struct sk_buff *skb,
3392 					     struct net_device *dev,
3393 					     u32 filter_mask))
3394 {
3395 	struct nlmsghdr *nlh;
3396 	struct ifinfomsg *ifm;
3397 	struct nlattr *br_afspec;
3398 	struct nlattr *protinfo;
3399 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3400 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3401 	int err = 0;
3402 
3403 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3404 	if (nlh == NULL)
3405 		return -EMSGSIZE;
3406 
3407 	ifm = nlmsg_data(nlh);
3408 	ifm->ifi_family = AF_BRIDGE;
3409 	ifm->__ifi_pad = 0;
3410 	ifm->ifi_type = dev->type;
3411 	ifm->ifi_index = dev->ifindex;
3412 	ifm->ifi_flags = dev_get_flags(dev);
3413 	ifm->ifi_change = 0;
3414 
3415 
3416 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3417 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3418 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3419 	    (br_dev &&
3420 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3421 	    (dev->addr_len &&
3422 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3423 	    (dev->ifindex != dev_get_iflink(dev) &&
3424 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3425 		goto nla_put_failure;
3426 
3427 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3428 	if (!br_afspec)
3429 		goto nla_put_failure;
3430 
3431 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3432 		nla_nest_cancel(skb, br_afspec);
3433 		goto nla_put_failure;
3434 	}
3435 
3436 	if (mode != BRIDGE_MODE_UNDEF) {
3437 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3438 			nla_nest_cancel(skb, br_afspec);
3439 			goto nla_put_failure;
3440 		}
3441 	}
3442 	if (vlan_fill) {
3443 		err = vlan_fill(skb, dev, filter_mask);
3444 		if (err) {
3445 			nla_nest_cancel(skb, br_afspec);
3446 			goto nla_put_failure;
3447 		}
3448 	}
3449 	nla_nest_end(skb, br_afspec);
3450 
3451 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3452 	if (!protinfo)
3453 		goto nla_put_failure;
3454 
3455 	if (brport_nla_put_flag(skb, flags, mask,
3456 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3457 	    brport_nla_put_flag(skb, flags, mask,
3458 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3459 	    brport_nla_put_flag(skb, flags, mask,
3460 				IFLA_BRPORT_FAST_LEAVE,
3461 				BR_MULTICAST_FAST_LEAVE) ||
3462 	    brport_nla_put_flag(skb, flags, mask,
3463 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3464 	    brport_nla_put_flag(skb, flags, mask,
3465 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3466 	    brport_nla_put_flag(skb, flags, mask,
3467 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3468 	    brport_nla_put_flag(skb, flags, mask,
3469 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3470 	    brport_nla_put_flag(skb, flags, mask,
3471 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3472 		nla_nest_cancel(skb, protinfo);
3473 		goto nla_put_failure;
3474 	}
3475 
3476 	nla_nest_end(skb, protinfo);
3477 
3478 	nlmsg_end(skb, nlh);
3479 	return 0;
3480 nla_put_failure:
3481 	nlmsg_cancel(skb, nlh);
3482 	return err ? err : -EMSGSIZE;
3483 }
3484 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3485 
3486 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3487 {
3488 	struct net *net = sock_net(skb->sk);
3489 	struct net_device *dev;
3490 	int idx = 0;
3491 	u32 portid = NETLINK_CB(cb->skb).portid;
3492 	u32 seq = cb->nlh->nlmsg_seq;
3493 	u32 filter_mask = 0;
3494 	int err;
3495 
3496 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3497 		struct nlattr *extfilt;
3498 
3499 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3500 					  IFLA_EXT_MASK);
3501 		if (extfilt) {
3502 			if (nla_len(extfilt) < sizeof(filter_mask))
3503 				return -EINVAL;
3504 
3505 			filter_mask = nla_get_u32(extfilt);
3506 		}
3507 	}
3508 
3509 	rcu_read_lock();
3510 	for_each_netdev_rcu(net, dev) {
3511 		const struct net_device_ops *ops = dev->netdev_ops;
3512 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3513 
3514 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3515 			if (idx >= cb->args[0]) {
3516 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3517 						skb, portid, seq, dev,
3518 						filter_mask, NLM_F_MULTI);
3519 				if (err < 0 && err != -EOPNOTSUPP) {
3520 					if (likely(skb->len))
3521 						break;
3522 
3523 					goto out_err;
3524 				}
3525 			}
3526 			idx++;
3527 		}
3528 
3529 		if (ops->ndo_bridge_getlink) {
3530 			if (idx >= cb->args[0]) {
3531 				err = ops->ndo_bridge_getlink(skb, portid,
3532 							      seq, dev,
3533 							      filter_mask,
3534 							      NLM_F_MULTI);
3535 				if (err < 0 && err != -EOPNOTSUPP) {
3536 					if (likely(skb->len))
3537 						break;
3538 
3539 					goto out_err;
3540 				}
3541 			}
3542 			idx++;
3543 		}
3544 	}
3545 	err = skb->len;
3546 out_err:
3547 	rcu_read_unlock();
3548 	cb->args[0] = idx;
3549 
3550 	return err;
3551 }
3552 
3553 static inline size_t bridge_nlmsg_size(void)
3554 {
3555 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3556 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3557 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3558 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3559 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3560 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3561 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3562 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3563 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3564 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3565 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3566 }
3567 
3568 static int rtnl_bridge_notify(struct net_device *dev)
3569 {
3570 	struct net *net = dev_net(dev);
3571 	struct sk_buff *skb;
3572 	int err = -EOPNOTSUPP;
3573 
3574 	if (!dev->netdev_ops->ndo_bridge_getlink)
3575 		return 0;
3576 
3577 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3578 	if (!skb) {
3579 		err = -ENOMEM;
3580 		goto errout;
3581 	}
3582 
3583 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3584 	if (err < 0)
3585 		goto errout;
3586 
3587 	if (!skb->len)
3588 		goto errout;
3589 
3590 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3591 	return 0;
3592 errout:
3593 	WARN_ON(err == -EMSGSIZE);
3594 	kfree_skb(skb);
3595 	if (err)
3596 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3597 	return err;
3598 }
3599 
3600 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3601 			       struct netlink_ext_ack *extack)
3602 {
3603 	struct net *net = sock_net(skb->sk);
3604 	struct ifinfomsg *ifm;
3605 	struct net_device *dev;
3606 	struct nlattr *br_spec, *attr = NULL;
3607 	int rem, err = -EOPNOTSUPP;
3608 	u16 flags = 0;
3609 	bool have_flags = false;
3610 
3611 	if (nlmsg_len(nlh) < sizeof(*ifm))
3612 		return -EINVAL;
3613 
3614 	ifm = nlmsg_data(nlh);
3615 	if (ifm->ifi_family != AF_BRIDGE)
3616 		return -EPFNOSUPPORT;
3617 
3618 	dev = __dev_get_by_index(net, ifm->ifi_index);
3619 	if (!dev) {
3620 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3621 		return -ENODEV;
3622 	}
3623 
3624 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3625 	if (br_spec) {
3626 		nla_for_each_nested(attr, br_spec, rem) {
3627 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3628 				if (nla_len(attr) < sizeof(flags))
3629 					return -EINVAL;
3630 
3631 				have_flags = true;
3632 				flags = nla_get_u16(attr);
3633 				break;
3634 			}
3635 		}
3636 	}
3637 
3638 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3639 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3640 
3641 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3642 			err = -EOPNOTSUPP;
3643 			goto out;
3644 		}
3645 
3646 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3647 		if (err)
3648 			goto out;
3649 
3650 		flags &= ~BRIDGE_FLAGS_MASTER;
3651 	}
3652 
3653 	if ((flags & BRIDGE_FLAGS_SELF)) {
3654 		if (!dev->netdev_ops->ndo_bridge_setlink)
3655 			err = -EOPNOTSUPP;
3656 		else
3657 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3658 								  flags);
3659 		if (!err) {
3660 			flags &= ~BRIDGE_FLAGS_SELF;
3661 
3662 			/* Generate event to notify upper layer of bridge
3663 			 * change
3664 			 */
3665 			err = rtnl_bridge_notify(dev);
3666 		}
3667 	}
3668 
3669 	if (have_flags)
3670 		memcpy(nla_data(attr), &flags, sizeof(flags));
3671 out:
3672 	return err;
3673 }
3674 
3675 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3676 			       struct netlink_ext_ack *extack)
3677 {
3678 	struct net *net = sock_net(skb->sk);
3679 	struct ifinfomsg *ifm;
3680 	struct net_device *dev;
3681 	struct nlattr *br_spec, *attr = NULL;
3682 	int rem, err = -EOPNOTSUPP;
3683 	u16 flags = 0;
3684 	bool have_flags = false;
3685 
3686 	if (nlmsg_len(nlh) < sizeof(*ifm))
3687 		return -EINVAL;
3688 
3689 	ifm = nlmsg_data(nlh);
3690 	if (ifm->ifi_family != AF_BRIDGE)
3691 		return -EPFNOSUPPORT;
3692 
3693 	dev = __dev_get_by_index(net, ifm->ifi_index);
3694 	if (!dev) {
3695 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3696 		return -ENODEV;
3697 	}
3698 
3699 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3700 	if (br_spec) {
3701 		nla_for_each_nested(attr, br_spec, rem) {
3702 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3703 				if (nla_len(attr) < sizeof(flags))
3704 					return -EINVAL;
3705 
3706 				have_flags = true;
3707 				flags = nla_get_u16(attr);
3708 				break;
3709 			}
3710 		}
3711 	}
3712 
3713 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3714 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3715 
3716 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3717 			err = -EOPNOTSUPP;
3718 			goto out;
3719 		}
3720 
3721 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3722 		if (err)
3723 			goto out;
3724 
3725 		flags &= ~BRIDGE_FLAGS_MASTER;
3726 	}
3727 
3728 	if ((flags & BRIDGE_FLAGS_SELF)) {
3729 		if (!dev->netdev_ops->ndo_bridge_dellink)
3730 			err = -EOPNOTSUPP;
3731 		else
3732 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3733 								  flags);
3734 
3735 		if (!err) {
3736 			flags &= ~BRIDGE_FLAGS_SELF;
3737 
3738 			/* Generate event to notify upper layer of bridge
3739 			 * change
3740 			 */
3741 			err = rtnl_bridge_notify(dev);
3742 		}
3743 	}
3744 
3745 	if (have_flags)
3746 		memcpy(nla_data(attr), &flags, sizeof(flags));
3747 out:
3748 	return err;
3749 }
3750 
3751 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3752 {
3753 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3754 	       (!idxattr || idxattr == attrid);
3755 }
3756 
3757 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3758 static int rtnl_get_offload_stats_attr_size(int attr_id)
3759 {
3760 	switch (attr_id) {
3761 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3762 		return sizeof(struct rtnl_link_stats64);
3763 	}
3764 
3765 	return 0;
3766 }
3767 
3768 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3769 				  int *prividx)
3770 {
3771 	struct nlattr *attr = NULL;
3772 	int attr_id, size;
3773 	void *attr_data;
3774 	int err;
3775 
3776 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3777 	      dev->netdev_ops->ndo_get_offload_stats))
3778 		return -ENODATA;
3779 
3780 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3781 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3782 		if (attr_id < *prividx)
3783 			continue;
3784 
3785 		size = rtnl_get_offload_stats_attr_size(attr_id);
3786 		if (!size)
3787 			continue;
3788 
3789 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3790 			continue;
3791 
3792 		attr = nla_reserve_64bit(skb, attr_id, size,
3793 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
3794 		if (!attr)
3795 			goto nla_put_failure;
3796 
3797 		attr_data = nla_data(attr);
3798 		memset(attr_data, 0, size);
3799 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3800 							     attr_data);
3801 		if (err)
3802 			goto get_offload_stats_failure;
3803 	}
3804 
3805 	if (!attr)
3806 		return -ENODATA;
3807 
3808 	*prividx = 0;
3809 	return 0;
3810 
3811 nla_put_failure:
3812 	err = -EMSGSIZE;
3813 get_offload_stats_failure:
3814 	*prividx = attr_id;
3815 	return err;
3816 }
3817 
3818 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3819 {
3820 	int nla_size = 0;
3821 	int attr_id;
3822 	int size;
3823 
3824 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3825 	      dev->netdev_ops->ndo_get_offload_stats))
3826 		return 0;
3827 
3828 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3829 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3830 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3831 			continue;
3832 		size = rtnl_get_offload_stats_attr_size(attr_id);
3833 		nla_size += nla_total_size_64bit(size);
3834 	}
3835 
3836 	if (nla_size != 0)
3837 		nla_size += nla_total_size(0);
3838 
3839 	return nla_size;
3840 }
3841 
3842 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3843 			       int type, u32 pid, u32 seq, u32 change,
3844 			       unsigned int flags, unsigned int filter_mask,
3845 			       int *idxattr, int *prividx)
3846 {
3847 	struct if_stats_msg *ifsm;
3848 	struct nlmsghdr *nlh;
3849 	struct nlattr *attr;
3850 	int s_prividx = *prividx;
3851 	int err;
3852 
3853 	ASSERT_RTNL();
3854 
3855 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3856 	if (!nlh)
3857 		return -EMSGSIZE;
3858 
3859 	ifsm = nlmsg_data(nlh);
3860 	ifsm->family = PF_UNSPEC;
3861 	ifsm->pad1 = 0;
3862 	ifsm->pad2 = 0;
3863 	ifsm->ifindex = dev->ifindex;
3864 	ifsm->filter_mask = filter_mask;
3865 
3866 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3867 		struct rtnl_link_stats64 *sp;
3868 
3869 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3870 					 sizeof(struct rtnl_link_stats64),
3871 					 IFLA_STATS_UNSPEC);
3872 		if (!attr)
3873 			goto nla_put_failure;
3874 
3875 		sp = nla_data(attr);
3876 		dev_get_stats(dev, sp);
3877 	}
3878 
3879 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3880 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3881 
3882 		if (ops && ops->fill_linkxstats) {
3883 			*idxattr = IFLA_STATS_LINK_XSTATS;
3884 			attr = nla_nest_start(skb,
3885 					      IFLA_STATS_LINK_XSTATS);
3886 			if (!attr)
3887 				goto nla_put_failure;
3888 
3889 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3890 			nla_nest_end(skb, attr);
3891 			if (err)
3892 				goto nla_put_failure;
3893 			*idxattr = 0;
3894 		}
3895 	}
3896 
3897 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3898 			     *idxattr)) {
3899 		const struct rtnl_link_ops *ops = NULL;
3900 		const struct net_device *master;
3901 
3902 		master = netdev_master_upper_dev_get(dev);
3903 		if (master)
3904 			ops = master->rtnl_link_ops;
3905 		if (ops && ops->fill_linkxstats) {
3906 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3907 			attr = nla_nest_start(skb,
3908 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3909 			if (!attr)
3910 				goto nla_put_failure;
3911 
3912 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3913 			nla_nest_end(skb, attr);
3914 			if (err)
3915 				goto nla_put_failure;
3916 			*idxattr = 0;
3917 		}
3918 	}
3919 
3920 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3921 			     *idxattr)) {
3922 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3923 		attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3924 		if (!attr)
3925 			goto nla_put_failure;
3926 
3927 		err = rtnl_get_offload_stats(skb, dev, prividx);
3928 		if (err == -ENODATA)
3929 			nla_nest_cancel(skb, attr);
3930 		else
3931 			nla_nest_end(skb, attr);
3932 
3933 		if (err && err != -ENODATA)
3934 			goto nla_put_failure;
3935 		*idxattr = 0;
3936 	}
3937 
3938 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3939 		struct rtnl_af_ops *af_ops;
3940 
3941 		*idxattr = IFLA_STATS_AF_SPEC;
3942 		attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3943 		if (!attr)
3944 			goto nla_put_failure;
3945 
3946 		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3947 			if (af_ops->fill_stats_af) {
3948 				struct nlattr *af;
3949 				int err;
3950 
3951 				af = nla_nest_start(skb, af_ops->family);
3952 				if (!af)
3953 					goto nla_put_failure;
3954 
3955 				err = af_ops->fill_stats_af(skb, dev);
3956 
3957 				if (err == -ENODATA)
3958 					nla_nest_cancel(skb, af);
3959 				else if (err < 0)
3960 					goto nla_put_failure;
3961 
3962 				nla_nest_end(skb, af);
3963 			}
3964 		}
3965 
3966 		nla_nest_end(skb, attr);
3967 
3968 		*idxattr = 0;
3969 	}
3970 
3971 	nlmsg_end(skb, nlh);
3972 
3973 	return 0;
3974 
3975 nla_put_failure:
3976 	/* not a multi message or no progress mean a real error */
3977 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3978 		nlmsg_cancel(skb, nlh);
3979 	else
3980 		nlmsg_end(skb, nlh);
3981 
3982 	return -EMSGSIZE;
3983 }
3984 
3985 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3986 				  u32 filter_mask)
3987 {
3988 	size_t size = 0;
3989 
3990 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3991 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3992 
3993 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3994 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3995 		int attr = IFLA_STATS_LINK_XSTATS;
3996 
3997 		if (ops && ops->get_linkxstats_size) {
3998 			size += nla_total_size(ops->get_linkxstats_size(dev,
3999 									attr));
4000 			/* for IFLA_STATS_LINK_XSTATS */
4001 			size += nla_total_size(0);
4002 		}
4003 	}
4004 
4005 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4006 		struct net_device *_dev = (struct net_device *)dev;
4007 		const struct rtnl_link_ops *ops = NULL;
4008 		const struct net_device *master;
4009 
4010 		/* netdev_master_upper_dev_get can't take const */
4011 		master = netdev_master_upper_dev_get(_dev);
4012 		if (master)
4013 			ops = master->rtnl_link_ops;
4014 		if (ops && ops->get_linkxstats_size) {
4015 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4016 
4017 			size += nla_total_size(ops->get_linkxstats_size(dev,
4018 									attr));
4019 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
4020 			size += nla_total_size(0);
4021 		}
4022 	}
4023 
4024 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4025 		size += rtnl_get_offload_stats_size(dev);
4026 
4027 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4028 		struct rtnl_af_ops *af_ops;
4029 
4030 		/* for IFLA_STATS_AF_SPEC */
4031 		size += nla_total_size(0);
4032 
4033 		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4034 			if (af_ops->get_stats_af_size) {
4035 				size += nla_total_size(
4036 					af_ops->get_stats_af_size(dev));
4037 
4038 				/* for AF_* */
4039 				size += nla_total_size(0);
4040 			}
4041 		}
4042 	}
4043 
4044 	return size;
4045 }
4046 
4047 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4048 			  struct netlink_ext_ack *extack)
4049 {
4050 	struct net *net = sock_net(skb->sk);
4051 	struct net_device *dev = NULL;
4052 	int idxattr = 0, prividx = 0;
4053 	struct if_stats_msg *ifsm;
4054 	struct sk_buff *nskb;
4055 	u32 filter_mask;
4056 	int err;
4057 
4058 	if (nlmsg_len(nlh) < sizeof(*ifsm))
4059 		return -EINVAL;
4060 
4061 	ifsm = nlmsg_data(nlh);
4062 	if (ifsm->ifindex > 0)
4063 		dev = __dev_get_by_index(net, ifsm->ifindex);
4064 	else
4065 		return -EINVAL;
4066 
4067 	if (!dev)
4068 		return -ENODEV;
4069 
4070 	filter_mask = ifsm->filter_mask;
4071 	if (!filter_mask)
4072 		return -EINVAL;
4073 
4074 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4075 	if (!nskb)
4076 		return -ENOBUFS;
4077 
4078 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4079 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4080 				  0, filter_mask, &idxattr, &prividx);
4081 	if (err < 0) {
4082 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4083 		WARN_ON(err == -EMSGSIZE);
4084 		kfree_skb(nskb);
4085 	} else {
4086 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4087 	}
4088 
4089 	return err;
4090 }
4091 
4092 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4093 {
4094 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
4095 	struct net *net = sock_net(skb->sk);
4096 	unsigned int flags = NLM_F_MULTI;
4097 	struct if_stats_msg *ifsm;
4098 	struct hlist_head *head;
4099 	struct net_device *dev;
4100 	u32 filter_mask = 0;
4101 	int idx = 0;
4102 
4103 	s_h = cb->args[0];
4104 	s_idx = cb->args[1];
4105 	s_idxattr = cb->args[2];
4106 	s_prividx = cb->args[3];
4107 
4108 	cb->seq = net->dev_base_seq;
4109 
4110 	if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4111 		return -EINVAL;
4112 
4113 	ifsm = nlmsg_data(cb->nlh);
4114 	filter_mask = ifsm->filter_mask;
4115 	if (!filter_mask)
4116 		return -EINVAL;
4117 
4118 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4119 		idx = 0;
4120 		head = &net->dev_index_head[h];
4121 		hlist_for_each_entry(dev, head, index_hlist) {
4122 			if (idx < s_idx)
4123 				goto cont;
4124 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4125 						  NETLINK_CB(cb->skb).portid,
4126 						  cb->nlh->nlmsg_seq, 0,
4127 						  flags, filter_mask,
4128 						  &s_idxattr, &s_prividx);
4129 			/* If we ran out of room on the first message,
4130 			 * we're in trouble
4131 			 */
4132 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4133 
4134 			if (err < 0)
4135 				goto out;
4136 			s_prividx = 0;
4137 			s_idxattr = 0;
4138 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4139 cont:
4140 			idx++;
4141 		}
4142 	}
4143 out:
4144 	cb->args[3] = s_prividx;
4145 	cb->args[2] = s_idxattr;
4146 	cb->args[1] = idx;
4147 	cb->args[0] = h;
4148 
4149 	return skb->len;
4150 }
4151 
4152 /* Process one rtnetlink message. */
4153 
4154 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4155 			     struct netlink_ext_ack *extack)
4156 {
4157 	struct net *net = sock_net(skb->sk);
4158 	struct rtnl_link *handlers;
4159 	int err = -EOPNOTSUPP;
4160 	rtnl_doit_func doit;
4161 	unsigned int flags;
4162 	int kind;
4163 	int family;
4164 	int type;
4165 
4166 	type = nlh->nlmsg_type;
4167 	if (type > RTM_MAX)
4168 		return -EOPNOTSUPP;
4169 
4170 	type -= RTM_BASE;
4171 
4172 	/* All the messages must have at least 1 byte length */
4173 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4174 		return 0;
4175 
4176 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4177 	kind = type&3;
4178 
4179 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4180 		return -EPERM;
4181 
4182 	if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4183 		family = PF_UNSPEC;
4184 
4185 	rcu_read_lock();
4186 	handlers = rcu_dereference(rtnl_msg_handlers[family]);
4187 	if (!handlers) {
4188 		family = PF_UNSPEC;
4189 		handlers = rcu_dereference(rtnl_msg_handlers[family]);
4190 	}
4191 
4192 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4193 		struct sock *rtnl;
4194 		rtnl_dumpit_func dumpit;
4195 		u16 min_dump_alloc = 0;
4196 
4197 		dumpit = READ_ONCE(handlers[type].dumpit);
4198 		if (!dumpit) {
4199 			family = PF_UNSPEC;
4200 			handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4201 			if (!handlers)
4202 				goto err_unlock;
4203 
4204 			dumpit = READ_ONCE(handlers[type].dumpit);
4205 			if (!dumpit)
4206 				goto err_unlock;
4207 		}
4208 
4209 		refcount_inc(&rtnl_msg_handlers_ref[family]);
4210 
4211 		if (type == RTM_GETLINK - RTM_BASE)
4212 			min_dump_alloc = rtnl_calcit(skb, nlh);
4213 
4214 		rcu_read_unlock();
4215 
4216 		rtnl = net->rtnl;
4217 		{
4218 			struct netlink_dump_control c = {
4219 				.dump		= dumpit,
4220 				.min_dump_alloc	= min_dump_alloc,
4221 			};
4222 			err = netlink_dump_start(rtnl, skb, nlh, &c);
4223 		}
4224 		refcount_dec(&rtnl_msg_handlers_ref[family]);
4225 		return err;
4226 	}
4227 
4228 	doit = READ_ONCE(handlers[type].doit);
4229 	if (!doit) {
4230 		family = PF_UNSPEC;
4231 		handlers = rcu_dereference(rtnl_msg_handlers[family]);
4232 	}
4233 
4234 	flags = READ_ONCE(handlers[type].flags);
4235 	if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4236 		refcount_inc(&rtnl_msg_handlers_ref[family]);
4237 		doit = READ_ONCE(handlers[type].doit);
4238 		rcu_read_unlock();
4239 		if (doit)
4240 			err = doit(skb, nlh, extack);
4241 		refcount_dec(&rtnl_msg_handlers_ref[family]);
4242 		return err;
4243 	}
4244 
4245 	rcu_read_unlock();
4246 
4247 	rtnl_lock();
4248 	handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4249 	if (handlers) {
4250 		doit = READ_ONCE(handlers[type].doit);
4251 		if (doit)
4252 			err = doit(skb, nlh, extack);
4253 	}
4254 	rtnl_unlock();
4255 	return err;
4256 
4257 err_unlock:
4258 	rcu_read_unlock();
4259 	return -EOPNOTSUPP;
4260 }
4261 
4262 static void rtnetlink_rcv(struct sk_buff *skb)
4263 {
4264 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4265 }
4266 
4267 static int rtnetlink_bind(struct net *net, int group)
4268 {
4269 	switch (group) {
4270 	case RTNLGRP_IPV4_MROUTE_R:
4271 	case RTNLGRP_IPV6_MROUTE_R:
4272 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4273 			return -EPERM;
4274 		break;
4275 	}
4276 	return 0;
4277 }
4278 
4279 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4280 {
4281 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4282 
4283 	switch (event) {
4284 	case NETDEV_REBOOT:
4285 	case NETDEV_CHANGEMTU:
4286 	case NETDEV_CHANGEADDR:
4287 	case NETDEV_CHANGENAME:
4288 	case NETDEV_FEAT_CHANGE:
4289 	case NETDEV_BONDING_FAILOVER:
4290 	case NETDEV_POST_TYPE_CHANGE:
4291 	case NETDEV_NOTIFY_PEERS:
4292 	case NETDEV_CHANGEUPPER:
4293 	case NETDEV_RESEND_IGMP:
4294 	case NETDEV_CHANGEINFODATA:
4295 	case NETDEV_CHANGE_TX_QUEUE_LEN:
4296 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4297 				   GFP_KERNEL);
4298 		break;
4299 	default:
4300 		break;
4301 	}
4302 	return NOTIFY_DONE;
4303 }
4304 
4305 static struct notifier_block rtnetlink_dev_notifier = {
4306 	.notifier_call	= rtnetlink_event,
4307 };
4308 
4309 
4310 static int __net_init rtnetlink_net_init(struct net *net)
4311 {
4312 	struct sock *sk;
4313 	struct netlink_kernel_cfg cfg = {
4314 		.groups		= RTNLGRP_MAX,
4315 		.input		= rtnetlink_rcv,
4316 		.cb_mutex	= &rtnl_mutex,
4317 		.flags		= NL_CFG_F_NONROOT_RECV,
4318 		.bind		= rtnetlink_bind,
4319 	};
4320 
4321 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4322 	if (!sk)
4323 		return -ENOMEM;
4324 	net->rtnl = sk;
4325 	return 0;
4326 }
4327 
4328 static void __net_exit rtnetlink_net_exit(struct net *net)
4329 {
4330 	netlink_kernel_release(net->rtnl);
4331 	net->rtnl = NULL;
4332 }
4333 
4334 static struct pernet_operations rtnetlink_net_ops = {
4335 	.init = rtnetlink_net_init,
4336 	.exit = rtnetlink_net_exit,
4337 };
4338 
4339 void __init rtnetlink_init(void)
4340 {
4341 	int i;
4342 
4343 	for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4344 		refcount_set(&rtnl_msg_handlers_ref[i], 1);
4345 
4346 	if (register_pernet_subsys(&rtnetlink_net_ops))
4347 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
4348 
4349 	register_netdevice_notifier(&rtnetlink_dev_notifier);
4350 
4351 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4352 		      rtnl_dump_ifinfo, 0);
4353 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4354 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4355 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4356 
4357 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4358 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4359 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4360 
4361 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4362 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4363 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4364 
4365 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4366 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4367 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4368 
4369 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4370 		      0);
4371 }
4372