xref: /linux/net/core/rtnetlink.c (revision 995231c820e3bd3633cb38bf4ea6f2541e1da331)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  *
15  *	Fixes:
16  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17  */
18 
19 #include <linux/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44 
45 #include <linux/uaccess.h>
46 
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
50 #include <net/ip.h>
51 #include <net/protocol.h>
52 #include <net/arp.h>
53 #include <net/route.h>
54 #include <net/udp.h>
55 #include <net/tcp.h>
56 #include <net/sock.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
61 
62 struct rtnl_link {
63 	rtnl_doit_func		doit;
64 	rtnl_dumpit_func	dumpit;
65 	unsigned int		flags;
66 };
67 
68 static DEFINE_MUTEX(rtnl_mutex);
69 
70 void rtnl_lock(void)
71 {
72 	mutex_lock(&rtnl_mutex);
73 }
74 EXPORT_SYMBOL(rtnl_lock);
75 
76 static struct sk_buff *defer_kfree_skb_list;
77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
78 {
79 	if (head && tail) {
80 		tail->next = defer_kfree_skb_list;
81 		defer_kfree_skb_list = head;
82 	}
83 }
84 EXPORT_SYMBOL(rtnl_kfree_skbs);
85 
86 void __rtnl_unlock(void)
87 {
88 	struct sk_buff *head = defer_kfree_skb_list;
89 
90 	defer_kfree_skb_list = NULL;
91 
92 	mutex_unlock(&rtnl_mutex);
93 
94 	while (head) {
95 		struct sk_buff *next = head->next;
96 
97 		kfree_skb(head);
98 		cond_resched();
99 		head = next;
100 	}
101 }
102 
103 void rtnl_unlock(void)
104 {
105 	/* This fellow will unlock it for us. */
106 	netdev_run_todo();
107 }
108 EXPORT_SYMBOL(rtnl_unlock);
109 
110 int rtnl_trylock(void)
111 {
112 	return mutex_trylock(&rtnl_mutex);
113 }
114 EXPORT_SYMBOL(rtnl_trylock);
115 
116 int rtnl_is_locked(void)
117 {
118 	return mutex_is_locked(&rtnl_mutex);
119 }
120 EXPORT_SYMBOL(rtnl_is_locked);
121 
122 #ifdef CONFIG_PROVE_LOCKING
123 bool lockdep_rtnl_is_held(void)
124 {
125 	return lockdep_is_held(&rtnl_mutex);
126 }
127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
129 
130 static struct rtnl_link __rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
131 static refcount_t rtnl_msg_handlers_ref[RTNL_FAMILY_MAX + 1];
132 
133 static inline int rtm_msgindex(int msgtype)
134 {
135 	int msgindex = msgtype - RTM_BASE;
136 
137 	/*
138 	 * msgindex < 0 implies someone tried to register a netlink
139 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
140 	 * the message type has not been added to linux/rtnetlink.h
141 	 */
142 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
143 
144 	return msgindex;
145 }
146 
147 /**
148  * __rtnl_register - Register a rtnetlink message type
149  * @protocol: Protocol family or PF_UNSPEC
150  * @msgtype: rtnetlink message type
151  * @doit: Function pointer called for each request message
152  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
153  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
154  *
155  * Registers the specified function pointers (at least one of them has
156  * to be non-NULL) to be called whenever a request message for the
157  * specified protocol family and message type is received.
158  *
159  * The special protocol family PF_UNSPEC may be used to define fallback
160  * function pointers for the case when no entry for the specific protocol
161  * family exists.
162  *
163  * Returns 0 on success or a negative error code.
164  */
165 int __rtnl_register(int protocol, int msgtype,
166 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
167 		    unsigned int flags)
168 {
169 	struct rtnl_link *tab;
170 	int msgindex;
171 
172 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
173 	msgindex = rtm_msgindex(msgtype);
174 
175 	tab = rcu_dereference_raw(rtnl_msg_handlers[protocol]);
176 	if (tab == NULL) {
177 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
178 		if (tab == NULL)
179 			return -ENOBUFS;
180 
181 		rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
182 	}
183 
184 	if (doit)
185 		tab[msgindex].doit = doit;
186 	if (dumpit)
187 		tab[msgindex].dumpit = dumpit;
188 	tab[msgindex].flags |= flags;
189 
190 	return 0;
191 }
192 EXPORT_SYMBOL_GPL(__rtnl_register);
193 
194 /**
195  * rtnl_register - Register a rtnetlink message type
196  *
197  * Identical to __rtnl_register() but panics on failure. This is useful
198  * as failure of this function is very unlikely, it can only happen due
199  * to lack of memory when allocating the chain to store all message
200  * handlers for a protocol. Meant for use in init functions where lack
201  * of memory implies no sense in continuing.
202  */
203 void rtnl_register(int protocol, int msgtype,
204 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
205 		   unsigned int flags)
206 {
207 	if (__rtnl_register(protocol, msgtype, doit, dumpit, flags) < 0)
208 		panic("Unable to register rtnetlink message handler, "
209 		      "protocol = %d, message type = %d\n",
210 		      protocol, msgtype);
211 }
212 EXPORT_SYMBOL_GPL(rtnl_register);
213 
214 /**
215  * rtnl_unregister - Unregister a rtnetlink message type
216  * @protocol: Protocol family or PF_UNSPEC
217  * @msgtype: rtnetlink message type
218  *
219  * Returns 0 on success or a negative error code.
220  */
221 int rtnl_unregister(int protocol, int msgtype)
222 {
223 	struct rtnl_link *handlers;
224 	int msgindex;
225 
226 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
227 	msgindex = rtm_msgindex(msgtype);
228 
229 	rtnl_lock();
230 	handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
231 	if (!handlers) {
232 		rtnl_unlock();
233 		return -ENOENT;
234 	}
235 
236 	handlers[msgindex].doit = NULL;
237 	handlers[msgindex].dumpit = NULL;
238 	handlers[msgindex].flags = 0;
239 	rtnl_unlock();
240 
241 	return 0;
242 }
243 EXPORT_SYMBOL_GPL(rtnl_unregister);
244 
245 /**
246  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
247  * @protocol : Protocol family or PF_UNSPEC
248  *
249  * Identical to calling rtnl_unregster() for all registered message types
250  * of a certain protocol family.
251  */
252 void rtnl_unregister_all(int protocol)
253 {
254 	struct rtnl_link *handlers;
255 
256 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
257 
258 	rtnl_lock();
259 	handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
260 	RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
261 	rtnl_unlock();
262 
263 	synchronize_net();
264 
265 	while (refcount_read(&rtnl_msg_handlers_ref[protocol]) > 1)
266 		schedule();
267 	kfree(handlers);
268 }
269 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
270 
271 static LIST_HEAD(link_ops);
272 
273 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
274 {
275 	const struct rtnl_link_ops *ops;
276 
277 	list_for_each_entry(ops, &link_ops, list) {
278 		if (!strcmp(ops->kind, kind))
279 			return ops;
280 	}
281 	return NULL;
282 }
283 
284 /**
285  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
286  * @ops: struct rtnl_link_ops * to register
287  *
288  * The caller must hold the rtnl_mutex. This function should be used
289  * by drivers that create devices during module initialization. It
290  * must be called before registering the devices.
291  *
292  * Returns 0 on success or a negative error code.
293  */
294 int __rtnl_link_register(struct rtnl_link_ops *ops)
295 {
296 	if (rtnl_link_ops_get(ops->kind))
297 		return -EEXIST;
298 
299 	/* The check for setup is here because if ops
300 	 * does not have that filled up, it is not possible
301 	 * to use the ops for creating device. So do not
302 	 * fill up dellink as well. That disables rtnl_dellink.
303 	 */
304 	if (ops->setup && !ops->dellink)
305 		ops->dellink = unregister_netdevice_queue;
306 
307 	list_add_tail(&ops->list, &link_ops);
308 	return 0;
309 }
310 EXPORT_SYMBOL_GPL(__rtnl_link_register);
311 
312 /**
313  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
314  * @ops: struct rtnl_link_ops * to register
315  *
316  * Returns 0 on success or a negative error code.
317  */
318 int rtnl_link_register(struct rtnl_link_ops *ops)
319 {
320 	int err;
321 
322 	rtnl_lock();
323 	err = __rtnl_link_register(ops);
324 	rtnl_unlock();
325 	return err;
326 }
327 EXPORT_SYMBOL_GPL(rtnl_link_register);
328 
329 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
330 {
331 	struct net_device *dev;
332 	LIST_HEAD(list_kill);
333 
334 	for_each_netdev(net, dev) {
335 		if (dev->rtnl_link_ops == ops)
336 			ops->dellink(dev, &list_kill);
337 	}
338 	unregister_netdevice_many(&list_kill);
339 }
340 
341 /**
342  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
343  * @ops: struct rtnl_link_ops * to unregister
344  *
345  * The caller must hold the rtnl_mutex.
346  */
347 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
348 {
349 	struct net *net;
350 
351 	for_each_net(net) {
352 		__rtnl_kill_links(net, ops);
353 	}
354 	list_del(&ops->list);
355 }
356 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
357 
358 /* Return with the rtnl_lock held when there are no network
359  * devices unregistering in any network namespace.
360  */
361 static void rtnl_lock_unregistering_all(void)
362 {
363 	struct net *net;
364 	bool unregistering;
365 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
366 
367 	add_wait_queue(&netdev_unregistering_wq, &wait);
368 	for (;;) {
369 		unregistering = false;
370 		rtnl_lock();
371 		for_each_net(net) {
372 			if (net->dev_unreg_count > 0) {
373 				unregistering = true;
374 				break;
375 			}
376 		}
377 		if (!unregistering)
378 			break;
379 		__rtnl_unlock();
380 
381 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
382 	}
383 	remove_wait_queue(&netdev_unregistering_wq, &wait);
384 }
385 
386 /**
387  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
388  * @ops: struct rtnl_link_ops * to unregister
389  */
390 void rtnl_link_unregister(struct rtnl_link_ops *ops)
391 {
392 	/* Close the race with cleanup_net() */
393 	mutex_lock(&net_mutex);
394 	rtnl_lock_unregistering_all();
395 	__rtnl_link_unregister(ops);
396 	rtnl_unlock();
397 	mutex_unlock(&net_mutex);
398 }
399 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
400 
401 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
402 {
403 	struct net_device *master_dev;
404 	const struct rtnl_link_ops *ops;
405 	size_t size = 0;
406 
407 	rcu_read_lock();
408 
409 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
410 	if (!master_dev)
411 		goto out;
412 
413 	ops = master_dev->rtnl_link_ops;
414 	if (!ops || !ops->get_slave_size)
415 		goto out;
416 	/* IFLA_INFO_SLAVE_DATA + nested data */
417 	size = nla_total_size(sizeof(struct nlattr)) +
418 	       ops->get_slave_size(master_dev, dev);
419 
420 out:
421 	rcu_read_unlock();
422 	return size;
423 }
424 
425 static size_t rtnl_link_get_size(const struct net_device *dev)
426 {
427 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
428 	size_t size;
429 
430 	if (!ops)
431 		return 0;
432 
433 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
434 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
435 
436 	if (ops->get_size)
437 		/* IFLA_INFO_DATA + nested data */
438 		size += nla_total_size(sizeof(struct nlattr)) +
439 			ops->get_size(dev);
440 
441 	if (ops->get_xstats_size)
442 		/* IFLA_INFO_XSTATS */
443 		size += nla_total_size(ops->get_xstats_size(dev));
444 
445 	size += rtnl_link_get_slave_info_data_size(dev);
446 
447 	return size;
448 }
449 
450 static LIST_HEAD(rtnl_af_ops);
451 
452 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
453 {
454 	const struct rtnl_af_ops *ops;
455 
456 	list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
457 		if (ops->family == family)
458 			return ops;
459 	}
460 
461 	return NULL;
462 }
463 
464 /**
465  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
466  * @ops: struct rtnl_af_ops * to register
467  *
468  * Returns 0 on success or a negative error code.
469  */
470 void rtnl_af_register(struct rtnl_af_ops *ops)
471 {
472 	rtnl_lock();
473 	list_add_tail_rcu(&ops->list, &rtnl_af_ops);
474 	rtnl_unlock();
475 }
476 EXPORT_SYMBOL_GPL(rtnl_af_register);
477 
478 /**
479  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
480  * @ops: struct rtnl_af_ops * to unregister
481  */
482 void rtnl_af_unregister(struct rtnl_af_ops *ops)
483 {
484 	rtnl_lock();
485 	list_del_rcu(&ops->list);
486 	rtnl_unlock();
487 
488 	synchronize_rcu();
489 }
490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
491 
492 static size_t rtnl_link_get_af_size(const struct net_device *dev,
493 				    u32 ext_filter_mask)
494 {
495 	struct rtnl_af_ops *af_ops;
496 	size_t size;
497 
498 	/* IFLA_AF_SPEC */
499 	size = nla_total_size(sizeof(struct nlattr));
500 
501 	rcu_read_lock();
502 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
503 		if (af_ops->get_link_af_size) {
504 			/* AF_* + nested data */
505 			size += nla_total_size(sizeof(struct nlattr)) +
506 				af_ops->get_link_af_size(dev, ext_filter_mask);
507 		}
508 	}
509 	rcu_read_unlock();
510 
511 	return size;
512 }
513 
514 static bool rtnl_have_link_slave_info(const struct net_device *dev)
515 {
516 	struct net_device *master_dev;
517 	bool ret = false;
518 
519 	rcu_read_lock();
520 
521 	master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
522 	if (master_dev && master_dev->rtnl_link_ops)
523 		ret = true;
524 	rcu_read_unlock();
525 	return ret;
526 }
527 
528 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
529 				     const struct net_device *dev)
530 {
531 	struct net_device *master_dev;
532 	const struct rtnl_link_ops *ops;
533 	struct nlattr *slave_data;
534 	int err;
535 
536 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
537 	if (!master_dev)
538 		return 0;
539 	ops = master_dev->rtnl_link_ops;
540 	if (!ops)
541 		return 0;
542 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
543 		return -EMSGSIZE;
544 	if (ops->fill_slave_info) {
545 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
546 		if (!slave_data)
547 			return -EMSGSIZE;
548 		err = ops->fill_slave_info(skb, master_dev, dev);
549 		if (err < 0)
550 			goto err_cancel_slave_data;
551 		nla_nest_end(skb, slave_data);
552 	}
553 	return 0;
554 
555 err_cancel_slave_data:
556 	nla_nest_cancel(skb, slave_data);
557 	return err;
558 }
559 
560 static int rtnl_link_info_fill(struct sk_buff *skb,
561 			       const struct net_device *dev)
562 {
563 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
564 	struct nlattr *data;
565 	int err;
566 
567 	if (!ops)
568 		return 0;
569 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
570 		return -EMSGSIZE;
571 	if (ops->fill_xstats) {
572 		err = ops->fill_xstats(skb, dev);
573 		if (err < 0)
574 			return err;
575 	}
576 	if (ops->fill_info) {
577 		data = nla_nest_start(skb, IFLA_INFO_DATA);
578 		if (data == NULL)
579 			return -EMSGSIZE;
580 		err = ops->fill_info(skb, dev);
581 		if (err < 0)
582 			goto err_cancel_data;
583 		nla_nest_end(skb, data);
584 	}
585 	return 0;
586 
587 err_cancel_data:
588 	nla_nest_cancel(skb, data);
589 	return err;
590 }
591 
592 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
593 {
594 	struct nlattr *linkinfo;
595 	int err = -EMSGSIZE;
596 
597 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
598 	if (linkinfo == NULL)
599 		goto out;
600 
601 	err = rtnl_link_info_fill(skb, dev);
602 	if (err < 0)
603 		goto err_cancel_link;
604 
605 	err = rtnl_link_slave_info_fill(skb, dev);
606 	if (err < 0)
607 		goto err_cancel_link;
608 
609 	nla_nest_end(skb, linkinfo);
610 	return 0;
611 
612 err_cancel_link:
613 	nla_nest_cancel(skb, linkinfo);
614 out:
615 	return err;
616 }
617 
618 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
619 {
620 	struct sock *rtnl = net->rtnl;
621 	int err = 0;
622 
623 	NETLINK_CB(skb).dst_group = group;
624 	if (echo)
625 		refcount_inc(&skb->users);
626 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
627 	if (echo)
628 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
629 	return err;
630 }
631 
632 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
633 {
634 	struct sock *rtnl = net->rtnl;
635 
636 	return nlmsg_unicast(rtnl, skb, pid);
637 }
638 EXPORT_SYMBOL(rtnl_unicast);
639 
640 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
641 		 struct nlmsghdr *nlh, gfp_t flags)
642 {
643 	struct sock *rtnl = net->rtnl;
644 	int report = 0;
645 
646 	if (nlh)
647 		report = nlmsg_report(nlh);
648 
649 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
650 }
651 EXPORT_SYMBOL(rtnl_notify);
652 
653 void rtnl_set_sk_err(struct net *net, u32 group, int error)
654 {
655 	struct sock *rtnl = net->rtnl;
656 
657 	netlink_set_err(rtnl, 0, group, error);
658 }
659 EXPORT_SYMBOL(rtnl_set_sk_err);
660 
661 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
662 {
663 	struct nlattr *mx;
664 	int i, valid = 0;
665 
666 	mx = nla_nest_start(skb, RTA_METRICS);
667 	if (mx == NULL)
668 		return -ENOBUFS;
669 
670 	for (i = 0; i < RTAX_MAX; i++) {
671 		if (metrics[i]) {
672 			if (i == RTAX_CC_ALGO - 1) {
673 				char tmp[TCP_CA_NAME_MAX], *name;
674 
675 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
676 				if (!name)
677 					continue;
678 				if (nla_put_string(skb, i + 1, name))
679 					goto nla_put_failure;
680 			} else if (i == RTAX_FEATURES - 1) {
681 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
682 
683 				if (!user_features)
684 					continue;
685 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
686 				if (nla_put_u32(skb, i + 1, user_features))
687 					goto nla_put_failure;
688 			} else {
689 				if (nla_put_u32(skb, i + 1, metrics[i]))
690 					goto nla_put_failure;
691 			}
692 			valid++;
693 		}
694 	}
695 
696 	if (!valid) {
697 		nla_nest_cancel(skb, mx);
698 		return 0;
699 	}
700 
701 	return nla_nest_end(skb, mx);
702 
703 nla_put_failure:
704 	nla_nest_cancel(skb, mx);
705 	return -EMSGSIZE;
706 }
707 EXPORT_SYMBOL(rtnetlink_put_metrics);
708 
709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
710 		       long expires, u32 error)
711 {
712 	struct rta_cacheinfo ci = {
713 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
714 		.rta_used = dst->__use,
715 		.rta_clntref = atomic_read(&(dst->__refcnt)),
716 		.rta_error = error,
717 		.rta_id =  id,
718 	};
719 
720 	if (expires) {
721 		unsigned long clock;
722 
723 		clock = jiffies_to_clock_t(abs(expires));
724 		clock = min_t(unsigned long, clock, INT_MAX);
725 		ci.rta_expires = (expires > 0) ? clock : -clock;
726 	}
727 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
728 }
729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
730 
731 static void set_operstate(struct net_device *dev, unsigned char transition)
732 {
733 	unsigned char operstate = dev->operstate;
734 
735 	switch (transition) {
736 	case IF_OPER_UP:
737 		if ((operstate == IF_OPER_DORMANT ||
738 		     operstate == IF_OPER_UNKNOWN) &&
739 		    !netif_dormant(dev))
740 			operstate = IF_OPER_UP;
741 		break;
742 
743 	case IF_OPER_DORMANT:
744 		if (operstate == IF_OPER_UP ||
745 		    operstate == IF_OPER_UNKNOWN)
746 			operstate = IF_OPER_DORMANT;
747 		break;
748 	}
749 
750 	if (dev->operstate != operstate) {
751 		write_lock_bh(&dev_base_lock);
752 		dev->operstate = operstate;
753 		write_unlock_bh(&dev_base_lock);
754 		netdev_state_change(dev);
755 	}
756 }
757 
758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
759 {
760 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
761 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
762 }
763 
764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
765 					   const struct ifinfomsg *ifm)
766 {
767 	unsigned int flags = ifm->ifi_flags;
768 
769 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
770 	if (ifm->ifi_change)
771 		flags = (flags & ifm->ifi_change) |
772 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
773 
774 	return flags;
775 }
776 
777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
778 				 const struct rtnl_link_stats64 *b)
779 {
780 	a->rx_packets = b->rx_packets;
781 	a->tx_packets = b->tx_packets;
782 	a->rx_bytes = b->rx_bytes;
783 	a->tx_bytes = b->tx_bytes;
784 	a->rx_errors = b->rx_errors;
785 	a->tx_errors = b->tx_errors;
786 	a->rx_dropped = b->rx_dropped;
787 	a->tx_dropped = b->tx_dropped;
788 
789 	a->multicast = b->multicast;
790 	a->collisions = b->collisions;
791 
792 	a->rx_length_errors = b->rx_length_errors;
793 	a->rx_over_errors = b->rx_over_errors;
794 	a->rx_crc_errors = b->rx_crc_errors;
795 	a->rx_frame_errors = b->rx_frame_errors;
796 	a->rx_fifo_errors = b->rx_fifo_errors;
797 	a->rx_missed_errors = b->rx_missed_errors;
798 
799 	a->tx_aborted_errors = b->tx_aborted_errors;
800 	a->tx_carrier_errors = b->tx_carrier_errors;
801 	a->tx_fifo_errors = b->tx_fifo_errors;
802 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
803 	a->tx_window_errors = b->tx_window_errors;
804 
805 	a->rx_compressed = b->rx_compressed;
806 	a->tx_compressed = b->tx_compressed;
807 
808 	a->rx_nohandler = b->rx_nohandler;
809 }
810 
811 /* All VF info */
812 static inline int rtnl_vfinfo_size(const struct net_device *dev,
813 				   u32 ext_filter_mask)
814 {
815 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
816 		int num_vfs = dev_num_vf(dev->dev.parent);
817 		size_t size = nla_total_size(0);
818 		size += num_vfs *
819 			(nla_total_size(0) +
820 			 nla_total_size(sizeof(struct ifla_vf_mac)) +
821 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
822 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
823 			 nla_total_size(MAX_VLAN_LIST_LEN *
824 					sizeof(struct ifla_vf_vlan_info)) +
825 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
826 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
827 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
828 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
829 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
830 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
831 			 /* IFLA_VF_STATS_RX_PACKETS */
832 			 nla_total_size_64bit(sizeof(__u64)) +
833 			 /* IFLA_VF_STATS_TX_PACKETS */
834 			 nla_total_size_64bit(sizeof(__u64)) +
835 			 /* IFLA_VF_STATS_RX_BYTES */
836 			 nla_total_size_64bit(sizeof(__u64)) +
837 			 /* IFLA_VF_STATS_TX_BYTES */
838 			 nla_total_size_64bit(sizeof(__u64)) +
839 			 /* IFLA_VF_STATS_BROADCAST */
840 			 nla_total_size_64bit(sizeof(__u64)) +
841 			 /* IFLA_VF_STATS_MULTICAST */
842 			 nla_total_size_64bit(sizeof(__u64)) +
843 			 nla_total_size(sizeof(struct ifla_vf_trust)));
844 		return size;
845 	} else
846 		return 0;
847 }
848 
849 static size_t rtnl_port_size(const struct net_device *dev,
850 			     u32 ext_filter_mask)
851 {
852 	size_t port_size = nla_total_size(4)		/* PORT_VF */
853 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
854 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
855 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
856 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
857 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
858 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
859 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
860 		+ port_size;
861 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
862 		+ port_size;
863 
864 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
865 	    !(ext_filter_mask & RTEXT_FILTER_VF))
866 		return 0;
867 	if (dev_num_vf(dev->dev.parent))
868 		return port_self_size + vf_ports_size +
869 			vf_port_size * dev_num_vf(dev->dev.parent);
870 	else
871 		return port_self_size;
872 }
873 
874 static size_t rtnl_xdp_size(void)
875 {
876 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
877 			  nla_total_size(1) +	/* XDP_ATTACHED */
878 			  nla_total_size(4);	/* XDP_PROG_ID */
879 
880 	return xdp_size;
881 }
882 
883 static noinline size_t if_nlmsg_size(const struct net_device *dev,
884 				     u32 ext_filter_mask)
885 {
886 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
887 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
888 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
889 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
890 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
891 	       + nla_total_size(sizeof(struct rtnl_link_stats))
892 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
893 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
894 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
895 	       + nla_total_size(4) /* IFLA_TXQLEN */
896 	       + nla_total_size(4) /* IFLA_WEIGHT */
897 	       + nla_total_size(4) /* IFLA_MTU */
898 	       + nla_total_size(4) /* IFLA_LINK */
899 	       + nla_total_size(4) /* IFLA_MASTER */
900 	       + nla_total_size(1) /* IFLA_CARRIER */
901 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
902 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
903 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
904 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
905 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
906 	       + nla_total_size(1) /* IFLA_OPERSTATE */
907 	       + nla_total_size(1) /* IFLA_LINKMODE */
908 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
909 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
910 	       + nla_total_size(4) /* IFLA_GROUP */
911 	       + nla_total_size(ext_filter_mask
912 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
913 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
914 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
915 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
916 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
917 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
918 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
919 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
920 	       + rtnl_xdp_size() /* IFLA_XDP */
921 	       + nla_total_size(4)  /* IFLA_EVENT */
922 	       + nla_total_size(4)  /* IFLA_NEW_NETNSID */
923 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
924 
925 }
926 
927 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
928 {
929 	struct nlattr *vf_ports;
930 	struct nlattr *vf_port;
931 	int vf;
932 	int err;
933 
934 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
935 	if (!vf_ports)
936 		return -EMSGSIZE;
937 
938 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
939 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
940 		if (!vf_port)
941 			goto nla_put_failure;
942 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
943 			goto nla_put_failure;
944 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
945 		if (err == -EMSGSIZE)
946 			goto nla_put_failure;
947 		if (err) {
948 			nla_nest_cancel(skb, vf_port);
949 			continue;
950 		}
951 		nla_nest_end(skb, vf_port);
952 	}
953 
954 	nla_nest_end(skb, vf_ports);
955 
956 	return 0;
957 
958 nla_put_failure:
959 	nla_nest_cancel(skb, vf_ports);
960 	return -EMSGSIZE;
961 }
962 
963 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
964 {
965 	struct nlattr *port_self;
966 	int err;
967 
968 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
969 	if (!port_self)
970 		return -EMSGSIZE;
971 
972 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
973 	if (err) {
974 		nla_nest_cancel(skb, port_self);
975 		return (err == -EMSGSIZE) ? err : 0;
976 	}
977 
978 	nla_nest_end(skb, port_self);
979 
980 	return 0;
981 }
982 
983 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
984 			  u32 ext_filter_mask)
985 {
986 	int err;
987 
988 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
989 	    !(ext_filter_mask & RTEXT_FILTER_VF))
990 		return 0;
991 
992 	err = rtnl_port_self_fill(skb, dev);
993 	if (err)
994 		return err;
995 
996 	if (dev_num_vf(dev->dev.parent)) {
997 		err = rtnl_vf_ports_fill(skb, dev);
998 		if (err)
999 			return err;
1000 	}
1001 
1002 	return 0;
1003 }
1004 
1005 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1006 {
1007 	int err;
1008 	struct netdev_phys_item_id ppid;
1009 
1010 	err = dev_get_phys_port_id(dev, &ppid);
1011 	if (err) {
1012 		if (err == -EOPNOTSUPP)
1013 			return 0;
1014 		return err;
1015 	}
1016 
1017 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1018 		return -EMSGSIZE;
1019 
1020 	return 0;
1021 }
1022 
1023 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1024 {
1025 	char name[IFNAMSIZ];
1026 	int err;
1027 
1028 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1029 	if (err) {
1030 		if (err == -EOPNOTSUPP)
1031 			return 0;
1032 		return err;
1033 	}
1034 
1035 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1036 		return -EMSGSIZE;
1037 
1038 	return 0;
1039 }
1040 
1041 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1042 {
1043 	int err;
1044 	struct switchdev_attr attr = {
1045 		.orig_dev = dev,
1046 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1047 		.flags = SWITCHDEV_F_NO_RECURSE,
1048 	};
1049 
1050 	err = switchdev_port_attr_get(dev, &attr);
1051 	if (err) {
1052 		if (err == -EOPNOTSUPP)
1053 			return 0;
1054 		return err;
1055 	}
1056 
1057 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1058 		    attr.u.ppid.id))
1059 		return -EMSGSIZE;
1060 
1061 	return 0;
1062 }
1063 
1064 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1065 					      struct net_device *dev)
1066 {
1067 	struct rtnl_link_stats64 *sp;
1068 	struct nlattr *attr;
1069 
1070 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1071 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1072 	if (!attr)
1073 		return -EMSGSIZE;
1074 
1075 	sp = nla_data(attr);
1076 	dev_get_stats(dev, sp);
1077 
1078 	attr = nla_reserve(skb, IFLA_STATS,
1079 			   sizeof(struct rtnl_link_stats));
1080 	if (!attr)
1081 		return -EMSGSIZE;
1082 
1083 	copy_rtnl_link_stats(nla_data(attr), sp);
1084 
1085 	return 0;
1086 }
1087 
1088 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1089 					       struct net_device *dev,
1090 					       int vfs_num,
1091 					       struct nlattr *vfinfo)
1092 {
1093 	struct ifla_vf_rss_query_en vf_rss_query_en;
1094 	struct nlattr *vf, *vfstats, *vfvlanlist;
1095 	struct ifla_vf_link_state vf_linkstate;
1096 	struct ifla_vf_vlan_info vf_vlan_info;
1097 	struct ifla_vf_spoofchk vf_spoofchk;
1098 	struct ifla_vf_tx_rate vf_tx_rate;
1099 	struct ifla_vf_stats vf_stats;
1100 	struct ifla_vf_trust vf_trust;
1101 	struct ifla_vf_vlan vf_vlan;
1102 	struct ifla_vf_rate vf_rate;
1103 	struct ifla_vf_mac vf_mac;
1104 	struct ifla_vf_info ivi;
1105 
1106 	memset(&ivi, 0, sizeof(ivi));
1107 
1108 	/* Not all SR-IOV capable drivers support the
1109 	 * spoofcheck and "RSS query enable" query.  Preset to
1110 	 * -1 so the user space tool can detect that the driver
1111 	 * didn't report anything.
1112 	 */
1113 	ivi.spoofchk = -1;
1114 	ivi.rss_query_en = -1;
1115 	ivi.trusted = -1;
1116 	/* The default value for VF link state is "auto"
1117 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1118 	 */
1119 	ivi.linkstate = 0;
1120 	/* VLAN Protocol by default is 802.1Q */
1121 	ivi.vlan_proto = htons(ETH_P_8021Q);
1122 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1123 		return 0;
1124 
1125 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1126 
1127 	vf_mac.vf =
1128 		vf_vlan.vf =
1129 		vf_vlan_info.vf =
1130 		vf_rate.vf =
1131 		vf_tx_rate.vf =
1132 		vf_spoofchk.vf =
1133 		vf_linkstate.vf =
1134 		vf_rss_query_en.vf =
1135 		vf_trust.vf = ivi.vf;
1136 
1137 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1138 	vf_vlan.vlan = ivi.vlan;
1139 	vf_vlan.qos = ivi.qos;
1140 	vf_vlan_info.vlan = ivi.vlan;
1141 	vf_vlan_info.qos = ivi.qos;
1142 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1143 	vf_tx_rate.rate = ivi.max_tx_rate;
1144 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1145 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1146 	vf_spoofchk.setting = ivi.spoofchk;
1147 	vf_linkstate.link_state = ivi.linkstate;
1148 	vf_rss_query_en.setting = ivi.rss_query_en;
1149 	vf_trust.setting = ivi.trusted;
1150 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1151 	if (!vf)
1152 		goto nla_put_vfinfo_failure;
1153 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1154 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1155 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1156 		    &vf_rate) ||
1157 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1158 		    &vf_tx_rate) ||
1159 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1160 		    &vf_spoofchk) ||
1161 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1162 		    &vf_linkstate) ||
1163 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1164 		    sizeof(vf_rss_query_en),
1165 		    &vf_rss_query_en) ||
1166 	    nla_put(skb, IFLA_VF_TRUST,
1167 		    sizeof(vf_trust), &vf_trust))
1168 		goto nla_put_vf_failure;
1169 	vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1170 	if (!vfvlanlist)
1171 		goto nla_put_vf_failure;
1172 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1173 		    &vf_vlan_info)) {
1174 		nla_nest_cancel(skb, vfvlanlist);
1175 		goto nla_put_vf_failure;
1176 	}
1177 	nla_nest_end(skb, vfvlanlist);
1178 	memset(&vf_stats, 0, sizeof(vf_stats));
1179 	if (dev->netdev_ops->ndo_get_vf_stats)
1180 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1181 						&vf_stats);
1182 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1183 	if (!vfstats)
1184 		goto nla_put_vf_failure;
1185 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1186 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1187 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1188 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1189 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1190 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1191 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1192 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1193 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1194 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1195 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1196 			      vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1197 		nla_nest_cancel(skb, vfstats);
1198 		goto nla_put_vf_failure;
1199 	}
1200 	nla_nest_end(skb, vfstats);
1201 	nla_nest_end(skb, vf);
1202 	return 0;
1203 
1204 nla_put_vf_failure:
1205 	nla_nest_cancel(skb, vf);
1206 nla_put_vfinfo_failure:
1207 	nla_nest_cancel(skb, vfinfo);
1208 	return -EMSGSIZE;
1209 }
1210 
1211 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1212 					   struct net_device *dev,
1213 					   u32 ext_filter_mask)
1214 {
1215 	struct nlattr *vfinfo;
1216 	int i, num_vfs;
1217 
1218 	if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1219 		return 0;
1220 
1221 	num_vfs = dev_num_vf(dev->dev.parent);
1222 	if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1223 		return -EMSGSIZE;
1224 
1225 	if (!dev->netdev_ops->ndo_get_vf_config)
1226 		return 0;
1227 
1228 	vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1229 	if (!vfinfo)
1230 		return -EMSGSIZE;
1231 
1232 	for (i = 0; i < num_vfs; i++) {
1233 		if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1234 			return -EMSGSIZE;
1235 	}
1236 
1237 	nla_nest_end(skb, vfinfo);
1238 	return 0;
1239 }
1240 
1241 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1242 {
1243 	struct rtnl_link_ifmap map;
1244 
1245 	memset(&map, 0, sizeof(map));
1246 	map.mem_start   = dev->mem_start;
1247 	map.mem_end     = dev->mem_end;
1248 	map.base_addr   = dev->base_addr;
1249 	map.irq         = dev->irq;
1250 	map.dma         = dev->dma;
1251 	map.port        = dev->if_port;
1252 
1253 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1254 		return -EMSGSIZE;
1255 
1256 	return 0;
1257 }
1258 
1259 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1260 {
1261 	const struct net_device_ops *ops = dev->netdev_ops;
1262 	const struct bpf_prog *generic_xdp_prog;
1263 
1264 	ASSERT_RTNL();
1265 
1266 	*prog_id = 0;
1267 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1268 	if (generic_xdp_prog) {
1269 		*prog_id = generic_xdp_prog->aux->id;
1270 		return XDP_ATTACHED_SKB;
1271 	}
1272 	if (!ops->ndo_xdp)
1273 		return XDP_ATTACHED_NONE;
1274 
1275 	return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1276 }
1277 
1278 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1279 {
1280 	struct nlattr *xdp;
1281 	u32 prog_id;
1282 	int err;
1283 
1284 	xdp = nla_nest_start(skb, IFLA_XDP);
1285 	if (!xdp)
1286 		return -EMSGSIZE;
1287 
1288 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1289 			 rtnl_xdp_attached_mode(dev, &prog_id));
1290 	if (err)
1291 		goto err_cancel;
1292 
1293 	if (prog_id) {
1294 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1295 		if (err)
1296 			goto err_cancel;
1297 	}
1298 
1299 	nla_nest_end(skb, xdp);
1300 	return 0;
1301 
1302 err_cancel:
1303 	nla_nest_cancel(skb, xdp);
1304 	return err;
1305 }
1306 
1307 static u32 rtnl_get_event(unsigned long event)
1308 {
1309 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1310 
1311 	switch (event) {
1312 	case NETDEV_REBOOT:
1313 		rtnl_event_type = IFLA_EVENT_REBOOT;
1314 		break;
1315 	case NETDEV_FEAT_CHANGE:
1316 		rtnl_event_type = IFLA_EVENT_FEATURES;
1317 		break;
1318 	case NETDEV_BONDING_FAILOVER:
1319 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1320 		break;
1321 	case NETDEV_NOTIFY_PEERS:
1322 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1323 		break;
1324 	case NETDEV_RESEND_IGMP:
1325 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1326 		break;
1327 	case NETDEV_CHANGEINFODATA:
1328 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1329 		break;
1330 	default:
1331 		break;
1332 	}
1333 
1334 	return rtnl_event_type;
1335 }
1336 
1337 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1338 {
1339 	const struct net_device *upper_dev;
1340 	int ret = 0;
1341 
1342 	rcu_read_lock();
1343 
1344 	upper_dev = netdev_master_upper_dev_get_rcu(dev);
1345 	if (upper_dev)
1346 		ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1347 
1348 	rcu_read_unlock();
1349 	return ret;
1350 }
1351 
1352 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1353 {
1354 	int ifindex = dev_get_iflink(dev);
1355 
1356 	if (dev->ifindex == ifindex)
1357 		return 0;
1358 
1359 	return nla_put_u32(skb, IFLA_LINK, ifindex);
1360 }
1361 
1362 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1363 					      struct net_device *dev)
1364 {
1365 	char buf[IFALIASZ];
1366 	int ret;
1367 
1368 	ret = dev_get_alias(dev, buf, sizeof(buf));
1369 	return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1370 }
1371 
1372 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1373 				  const struct net_device *dev)
1374 {
1375 	if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1376 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1377 
1378 		if (!net_eq(dev_net(dev), link_net)) {
1379 			int id = peernet2id_alloc(dev_net(dev), link_net);
1380 
1381 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1382 				return -EMSGSIZE;
1383 		}
1384 	}
1385 
1386 	return 0;
1387 }
1388 
1389 static int rtnl_fill_link_af(struct sk_buff *skb,
1390 			     const struct net_device *dev,
1391 			     u32 ext_filter_mask)
1392 {
1393 	const struct rtnl_af_ops *af_ops;
1394 	struct nlattr *af_spec;
1395 
1396 	af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1397 	if (!af_spec)
1398 		return -EMSGSIZE;
1399 
1400 	list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1401 		struct nlattr *af;
1402 		int err;
1403 
1404 		if (!af_ops->fill_link_af)
1405 			continue;
1406 
1407 		af = nla_nest_start(skb, af_ops->family);
1408 		if (!af)
1409 			return -EMSGSIZE;
1410 
1411 		err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1412 		/*
1413 		 * Caller may return ENODATA to indicate that there
1414 		 * was no data to be dumped. This is not an error, it
1415 		 * means we should trim the attribute header and
1416 		 * continue.
1417 		 */
1418 		if (err == -ENODATA)
1419 			nla_nest_cancel(skb, af);
1420 		else if (err < 0)
1421 			return -EMSGSIZE;
1422 
1423 		nla_nest_end(skb, af);
1424 	}
1425 
1426 	nla_nest_end(skb, af_spec);
1427 	return 0;
1428 }
1429 
1430 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1431 			    int type, u32 pid, u32 seq, u32 change,
1432 			    unsigned int flags, u32 ext_filter_mask,
1433 			    u32 event, int *new_nsid)
1434 {
1435 	struct ifinfomsg *ifm;
1436 	struct nlmsghdr *nlh;
1437 
1438 	ASSERT_RTNL();
1439 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1440 	if (nlh == NULL)
1441 		return -EMSGSIZE;
1442 
1443 	ifm = nlmsg_data(nlh);
1444 	ifm->ifi_family = AF_UNSPEC;
1445 	ifm->__ifi_pad = 0;
1446 	ifm->ifi_type = dev->type;
1447 	ifm->ifi_index = dev->ifindex;
1448 	ifm->ifi_flags = dev_get_flags(dev);
1449 	ifm->ifi_change = change;
1450 
1451 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1452 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1453 	    nla_put_u8(skb, IFLA_OPERSTATE,
1454 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1455 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1456 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1457 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1458 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1459 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1460 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1461 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1462 #ifdef CONFIG_RPS
1463 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1464 #endif
1465 	    nla_put_iflink(skb, dev) ||
1466 	    put_master_ifindex(skb, dev) ||
1467 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1468 	    (dev->qdisc &&
1469 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1470 	    nla_put_ifalias(skb, dev) ||
1471 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1472 			atomic_read(&dev->carrier_changes)) ||
1473 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1474 		goto nla_put_failure;
1475 
1476 	if (event != IFLA_EVENT_NONE) {
1477 		if (nla_put_u32(skb, IFLA_EVENT, event))
1478 			goto nla_put_failure;
1479 	}
1480 
1481 	if (rtnl_fill_link_ifmap(skb, dev))
1482 		goto nla_put_failure;
1483 
1484 	if (dev->addr_len) {
1485 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1486 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1487 			goto nla_put_failure;
1488 	}
1489 
1490 	if (rtnl_phys_port_id_fill(skb, dev))
1491 		goto nla_put_failure;
1492 
1493 	if (rtnl_phys_port_name_fill(skb, dev))
1494 		goto nla_put_failure;
1495 
1496 	if (rtnl_phys_switch_id_fill(skb, dev))
1497 		goto nla_put_failure;
1498 
1499 	if (rtnl_fill_stats(skb, dev))
1500 		goto nla_put_failure;
1501 
1502 	if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1503 		goto nla_put_failure;
1504 
1505 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1506 		goto nla_put_failure;
1507 
1508 	if (rtnl_xdp_fill(skb, dev))
1509 		goto nla_put_failure;
1510 
1511 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1512 		if (rtnl_link_fill(skb, dev) < 0)
1513 			goto nla_put_failure;
1514 	}
1515 
1516 	if (rtnl_fill_link_netnsid(skb, dev))
1517 		goto nla_put_failure;
1518 
1519 	if (new_nsid &&
1520 	    nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1521 		goto nla_put_failure;
1522 
1523 	rcu_read_lock();
1524 	if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1525 		goto nla_put_failure_rcu;
1526 	rcu_read_unlock();
1527 
1528 	nlmsg_end(skb, nlh);
1529 	return 0;
1530 
1531 nla_put_failure_rcu:
1532 	rcu_read_unlock();
1533 nla_put_failure:
1534 	nlmsg_cancel(skb, nlh);
1535 	return -EMSGSIZE;
1536 }
1537 
1538 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1539 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1540 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1541 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1542 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1543 	[IFLA_MTU]		= { .type = NLA_U32 },
1544 	[IFLA_LINK]		= { .type = NLA_U32 },
1545 	[IFLA_MASTER]		= { .type = NLA_U32 },
1546 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1547 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1548 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1549 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1550 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1551 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1552 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1553 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1554 	/* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1555 	 * allow 0-length string (needed to remove an alias).
1556 	 */
1557 	[IFLA_IFALIAS]	        = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1558 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1559 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1560 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1561 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1562 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1563 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1564 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1565 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1566 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1567 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1568 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1569 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1570 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1571 	[IFLA_XDP]		= { .type = NLA_NESTED },
1572 	[IFLA_EVENT]		= { .type = NLA_U32 },
1573 	[IFLA_GROUP]		= { .type = NLA_U32 },
1574 };
1575 
1576 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1577 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1578 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1579 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1580 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1581 };
1582 
1583 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1584 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1585 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1586 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1587 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1588 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1589 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1590 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1591 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1592 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1593 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1594 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1595 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1596 };
1597 
1598 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1599 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1600 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1601 				    .len = PORT_PROFILE_MAX },
1602 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1603 				      .len = PORT_UUID_MAX },
1604 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1605 				    .len = PORT_UUID_MAX },
1606 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1607 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1608 
1609 	/* Unused, but we need to keep it here since user space could
1610 	 * fill it. It's also broken with regard to NLA_BINARY use in
1611 	 * combination with structs.
1612 	 */
1613 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1614 				    .len = sizeof(struct ifla_port_vsi) },
1615 };
1616 
1617 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1618 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1619 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1620 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1621 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1622 };
1623 
1624 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1625 {
1626 	const struct rtnl_link_ops *ops = NULL;
1627 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1628 
1629 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1630 			     ifla_info_policy, NULL) < 0)
1631 		return NULL;
1632 
1633 	if (linfo[IFLA_INFO_KIND]) {
1634 		char kind[MODULE_NAME_LEN];
1635 
1636 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1637 		ops = rtnl_link_ops_get(kind);
1638 	}
1639 
1640 	return ops;
1641 }
1642 
1643 static bool link_master_filtered(struct net_device *dev, int master_idx)
1644 {
1645 	struct net_device *master;
1646 
1647 	if (!master_idx)
1648 		return false;
1649 
1650 	master = netdev_master_upper_dev_get(dev);
1651 	if (!master || master->ifindex != master_idx)
1652 		return true;
1653 
1654 	return false;
1655 }
1656 
1657 static bool link_kind_filtered(const struct net_device *dev,
1658 			       const struct rtnl_link_ops *kind_ops)
1659 {
1660 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1661 		return true;
1662 
1663 	return false;
1664 }
1665 
1666 static bool link_dump_filtered(struct net_device *dev,
1667 			       int master_idx,
1668 			       const struct rtnl_link_ops *kind_ops)
1669 {
1670 	if (link_master_filtered(dev, master_idx) ||
1671 	    link_kind_filtered(dev, kind_ops))
1672 		return true;
1673 
1674 	return false;
1675 }
1676 
1677 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1678 {
1679 	struct net *net = sock_net(skb->sk);
1680 	int h, s_h;
1681 	int idx = 0, s_idx;
1682 	struct net_device *dev;
1683 	struct hlist_head *head;
1684 	struct nlattr *tb[IFLA_MAX+1];
1685 	u32 ext_filter_mask = 0;
1686 	const struct rtnl_link_ops *kind_ops = NULL;
1687 	unsigned int flags = NLM_F_MULTI;
1688 	int master_idx = 0;
1689 	int err;
1690 	int hdrlen;
1691 
1692 	s_h = cb->args[0];
1693 	s_idx = cb->args[1];
1694 
1695 	/* A hack to preserve kernel<->userspace interface.
1696 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1697 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1698 	 * what iproute2 < v3.9.0 used.
1699 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1700 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1701 	 */
1702 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1703 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1704 
1705 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1706 			ifla_policy, NULL) >= 0) {
1707 		if (tb[IFLA_EXT_MASK])
1708 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1709 
1710 		if (tb[IFLA_MASTER])
1711 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1712 
1713 		if (tb[IFLA_LINKINFO])
1714 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1715 
1716 		if (master_idx || kind_ops)
1717 			flags |= NLM_F_DUMP_FILTERED;
1718 	}
1719 
1720 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1721 		idx = 0;
1722 		head = &net->dev_index_head[h];
1723 		hlist_for_each_entry(dev, head, index_hlist) {
1724 			if (link_dump_filtered(dev, master_idx, kind_ops))
1725 				goto cont;
1726 			if (idx < s_idx)
1727 				goto cont;
1728 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1729 					       NETLINK_CB(cb->skb).portid,
1730 					       cb->nlh->nlmsg_seq, 0,
1731 					       flags,
1732 					       ext_filter_mask, 0, NULL);
1733 
1734 			if (err < 0) {
1735 				if (likely(skb->len))
1736 					goto out;
1737 
1738 				goto out_err;
1739 			}
1740 cont:
1741 			idx++;
1742 		}
1743 	}
1744 out:
1745 	err = skb->len;
1746 out_err:
1747 	cb->args[1] = idx;
1748 	cb->args[0] = h;
1749 	cb->seq = net->dev_base_seq;
1750 	nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1751 
1752 	return err;
1753 }
1754 
1755 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1756 			struct netlink_ext_ack *exterr)
1757 {
1758 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1759 }
1760 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1761 
1762 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1763 {
1764 	struct net *net;
1765 	/* Examine the link attributes and figure out which
1766 	 * network namespace we are talking about.
1767 	 */
1768 	if (tb[IFLA_NET_NS_PID])
1769 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1770 	else if (tb[IFLA_NET_NS_FD])
1771 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1772 	else
1773 		net = get_net(src_net);
1774 	return net;
1775 }
1776 EXPORT_SYMBOL(rtnl_link_get_net);
1777 
1778 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1779 {
1780 	if (dev) {
1781 		if (tb[IFLA_ADDRESS] &&
1782 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1783 			return -EINVAL;
1784 
1785 		if (tb[IFLA_BROADCAST] &&
1786 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1787 			return -EINVAL;
1788 	}
1789 
1790 	if (tb[IFLA_AF_SPEC]) {
1791 		struct nlattr *af;
1792 		int rem, err;
1793 
1794 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1795 			const struct rtnl_af_ops *af_ops;
1796 
1797 			rcu_read_lock();
1798 			af_ops = rtnl_af_lookup(nla_type(af));
1799 			if (!af_ops) {
1800 				rcu_read_unlock();
1801 				return -EAFNOSUPPORT;
1802 			}
1803 
1804 			if (!af_ops->set_link_af) {
1805 				rcu_read_unlock();
1806 				return -EOPNOTSUPP;
1807 			}
1808 
1809 			if (af_ops->validate_link_af) {
1810 				err = af_ops->validate_link_af(dev, af);
1811 				if (err < 0) {
1812 					rcu_read_unlock();
1813 					return err;
1814 				}
1815 			}
1816 
1817 			rcu_read_unlock();
1818 		}
1819 	}
1820 
1821 	return 0;
1822 }
1823 
1824 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1825 				  int guid_type)
1826 {
1827 	const struct net_device_ops *ops = dev->netdev_ops;
1828 
1829 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1830 }
1831 
1832 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1833 {
1834 	if (dev->type != ARPHRD_INFINIBAND)
1835 		return -EOPNOTSUPP;
1836 
1837 	return handle_infiniband_guid(dev, ivt, guid_type);
1838 }
1839 
1840 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1841 {
1842 	const struct net_device_ops *ops = dev->netdev_ops;
1843 	int err = -EINVAL;
1844 
1845 	if (tb[IFLA_VF_MAC]) {
1846 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1847 
1848 		err = -EOPNOTSUPP;
1849 		if (ops->ndo_set_vf_mac)
1850 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1851 						  ivm->mac);
1852 		if (err < 0)
1853 			return err;
1854 	}
1855 
1856 	if (tb[IFLA_VF_VLAN]) {
1857 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1858 
1859 		err = -EOPNOTSUPP;
1860 		if (ops->ndo_set_vf_vlan)
1861 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1862 						   ivv->qos,
1863 						   htons(ETH_P_8021Q));
1864 		if (err < 0)
1865 			return err;
1866 	}
1867 
1868 	if (tb[IFLA_VF_VLAN_LIST]) {
1869 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1870 		struct nlattr *attr;
1871 		int rem, len = 0;
1872 
1873 		err = -EOPNOTSUPP;
1874 		if (!ops->ndo_set_vf_vlan)
1875 			return err;
1876 
1877 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1878 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1879 			    nla_len(attr) < NLA_HDRLEN) {
1880 				return -EINVAL;
1881 			}
1882 			if (len >= MAX_VLAN_LIST_LEN)
1883 				return -EOPNOTSUPP;
1884 			ivvl[len] = nla_data(attr);
1885 
1886 			len++;
1887 		}
1888 		if (len == 0)
1889 			return -EINVAL;
1890 
1891 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1892 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
1893 		if (err < 0)
1894 			return err;
1895 	}
1896 
1897 	if (tb[IFLA_VF_TX_RATE]) {
1898 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1899 		struct ifla_vf_info ivf;
1900 
1901 		err = -EOPNOTSUPP;
1902 		if (ops->ndo_get_vf_config)
1903 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1904 		if (err < 0)
1905 			return err;
1906 
1907 		err = -EOPNOTSUPP;
1908 		if (ops->ndo_set_vf_rate)
1909 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1910 						   ivf.min_tx_rate,
1911 						   ivt->rate);
1912 		if (err < 0)
1913 			return err;
1914 	}
1915 
1916 	if (tb[IFLA_VF_RATE]) {
1917 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1918 
1919 		err = -EOPNOTSUPP;
1920 		if (ops->ndo_set_vf_rate)
1921 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1922 						   ivt->min_tx_rate,
1923 						   ivt->max_tx_rate);
1924 		if (err < 0)
1925 			return err;
1926 	}
1927 
1928 	if (tb[IFLA_VF_SPOOFCHK]) {
1929 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1930 
1931 		err = -EOPNOTSUPP;
1932 		if (ops->ndo_set_vf_spoofchk)
1933 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1934 						       ivs->setting);
1935 		if (err < 0)
1936 			return err;
1937 	}
1938 
1939 	if (tb[IFLA_VF_LINK_STATE]) {
1940 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1941 
1942 		err = -EOPNOTSUPP;
1943 		if (ops->ndo_set_vf_link_state)
1944 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1945 							 ivl->link_state);
1946 		if (err < 0)
1947 			return err;
1948 	}
1949 
1950 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1951 		struct ifla_vf_rss_query_en *ivrssq_en;
1952 
1953 		err = -EOPNOTSUPP;
1954 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1955 		if (ops->ndo_set_vf_rss_query_en)
1956 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1957 							   ivrssq_en->setting);
1958 		if (err < 0)
1959 			return err;
1960 	}
1961 
1962 	if (tb[IFLA_VF_TRUST]) {
1963 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1964 
1965 		err = -EOPNOTSUPP;
1966 		if (ops->ndo_set_vf_trust)
1967 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1968 		if (err < 0)
1969 			return err;
1970 	}
1971 
1972 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1973 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1974 
1975 		if (!ops->ndo_set_vf_guid)
1976 			return -EOPNOTSUPP;
1977 
1978 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1979 	}
1980 
1981 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1982 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1983 
1984 		if (!ops->ndo_set_vf_guid)
1985 			return -EOPNOTSUPP;
1986 
1987 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1988 	}
1989 
1990 	return err;
1991 }
1992 
1993 static int do_set_master(struct net_device *dev, int ifindex,
1994 			 struct netlink_ext_ack *extack)
1995 {
1996 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1997 	const struct net_device_ops *ops;
1998 	int err;
1999 
2000 	if (upper_dev) {
2001 		if (upper_dev->ifindex == ifindex)
2002 			return 0;
2003 		ops = upper_dev->netdev_ops;
2004 		if (ops->ndo_del_slave) {
2005 			err = ops->ndo_del_slave(upper_dev, dev);
2006 			if (err)
2007 				return err;
2008 		} else {
2009 			return -EOPNOTSUPP;
2010 		}
2011 	}
2012 
2013 	if (ifindex) {
2014 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2015 		if (!upper_dev)
2016 			return -EINVAL;
2017 		ops = upper_dev->netdev_ops;
2018 		if (ops->ndo_add_slave) {
2019 			err = ops->ndo_add_slave(upper_dev, dev, extack);
2020 			if (err)
2021 				return err;
2022 		} else {
2023 			return -EOPNOTSUPP;
2024 		}
2025 	}
2026 	return 0;
2027 }
2028 
2029 #define DO_SETLINK_MODIFIED	0x01
2030 /* notify flag means notify + modified. */
2031 #define DO_SETLINK_NOTIFY	0x03
2032 static int do_setlink(const struct sk_buff *skb,
2033 		      struct net_device *dev, struct ifinfomsg *ifm,
2034 		      struct netlink_ext_ack *extack,
2035 		      struct nlattr **tb, char *ifname, int status)
2036 {
2037 	const struct net_device_ops *ops = dev->netdev_ops;
2038 	int err;
2039 
2040 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
2041 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
2042 		if (IS_ERR(net)) {
2043 			err = PTR_ERR(net);
2044 			goto errout;
2045 		}
2046 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
2047 			put_net(net);
2048 			err = -EPERM;
2049 			goto errout;
2050 		}
2051 		err = dev_change_net_namespace(dev, net, ifname);
2052 		put_net(net);
2053 		if (err)
2054 			goto errout;
2055 		status |= DO_SETLINK_MODIFIED;
2056 	}
2057 
2058 	if (tb[IFLA_MAP]) {
2059 		struct rtnl_link_ifmap *u_map;
2060 		struct ifmap k_map;
2061 
2062 		if (!ops->ndo_set_config) {
2063 			err = -EOPNOTSUPP;
2064 			goto errout;
2065 		}
2066 
2067 		if (!netif_device_present(dev)) {
2068 			err = -ENODEV;
2069 			goto errout;
2070 		}
2071 
2072 		u_map = nla_data(tb[IFLA_MAP]);
2073 		k_map.mem_start = (unsigned long) u_map->mem_start;
2074 		k_map.mem_end = (unsigned long) u_map->mem_end;
2075 		k_map.base_addr = (unsigned short) u_map->base_addr;
2076 		k_map.irq = (unsigned char) u_map->irq;
2077 		k_map.dma = (unsigned char) u_map->dma;
2078 		k_map.port = (unsigned char) u_map->port;
2079 
2080 		err = ops->ndo_set_config(dev, &k_map);
2081 		if (err < 0)
2082 			goto errout;
2083 
2084 		status |= DO_SETLINK_NOTIFY;
2085 	}
2086 
2087 	if (tb[IFLA_ADDRESS]) {
2088 		struct sockaddr *sa;
2089 		int len;
2090 
2091 		len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2092 						  sizeof(*sa));
2093 		sa = kmalloc(len, GFP_KERNEL);
2094 		if (!sa) {
2095 			err = -ENOMEM;
2096 			goto errout;
2097 		}
2098 		sa->sa_family = dev->type;
2099 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2100 		       dev->addr_len);
2101 		err = dev_set_mac_address(dev, sa);
2102 		kfree(sa);
2103 		if (err)
2104 			goto errout;
2105 		status |= DO_SETLINK_MODIFIED;
2106 	}
2107 
2108 	if (tb[IFLA_MTU]) {
2109 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2110 		if (err < 0)
2111 			goto errout;
2112 		status |= DO_SETLINK_MODIFIED;
2113 	}
2114 
2115 	if (tb[IFLA_GROUP]) {
2116 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2117 		status |= DO_SETLINK_NOTIFY;
2118 	}
2119 
2120 	/*
2121 	 * Interface selected by interface index but interface
2122 	 * name provided implies that a name change has been
2123 	 * requested.
2124 	 */
2125 	if (ifm->ifi_index > 0 && ifname[0]) {
2126 		err = dev_change_name(dev, ifname);
2127 		if (err < 0)
2128 			goto errout;
2129 		status |= DO_SETLINK_MODIFIED;
2130 	}
2131 
2132 	if (tb[IFLA_IFALIAS]) {
2133 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2134 				    nla_len(tb[IFLA_IFALIAS]));
2135 		if (err < 0)
2136 			goto errout;
2137 		status |= DO_SETLINK_NOTIFY;
2138 	}
2139 
2140 	if (tb[IFLA_BROADCAST]) {
2141 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2142 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2143 	}
2144 
2145 	if (ifm->ifi_flags || ifm->ifi_change) {
2146 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2147 		if (err < 0)
2148 			goto errout;
2149 	}
2150 
2151 	if (tb[IFLA_MASTER]) {
2152 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2153 		if (err)
2154 			goto errout;
2155 		status |= DO_SETLINK_MODIFIED;
2156 	}
2157 
2158 	if (tb[IFLA_CARRIER]) {
2159 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2160 		if (err)
2161 			goto errout;
2162 		status |= DO_SETLINK_MODIFIED;
2163 	}
2164 
2165 	if (tb[IFLA_TXQLEN]) {
2166 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2167 		unsigned int orig_len = dev->tx_queue_len;
2168 
2169 		if (dev->tx_queue_len ^ value) {
2170 			dev->tx_queue_len = value;
2171 			err = call_netdevice_notifiers(
2172 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2173 			err = notifier_to_errno(err);
2174 			if (err) {
2175 				dev->tx_queue_len = orig_len;
2176 				goto errout;
2177 			}
2178 			status |= DO_SETLINK_MODIFIED;
2179 		}
2180 	}
2181 
2182 	if (tb[IFLA_OPERSTATE])
2183 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2184 
2185 	if (tb[IFLA_LINKMODE]) {
2186 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2187 
2188 		write_lock_bh(&dev_base_lock);
2189 		if (dev->link_mode ^ value)
2190 			status |= DO_SETLINK_NOTIFY;
2191 		dev->link_mode = value;
2192 		write_unlock_bh(&dev_base_lock);
2193 	}
2194 
2195 	if (tb[IFLA_VFINFO_LIST]) {
2196 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2197 		struct nlattr *attr;
2198 		int rem;
2199 
2200 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2201 			if (nla_type(attr) != IFLA_VF_INFO ||
2202 			    nla_len(attr) < NLA_HDRLEN) {
2203 				err = -EINVAL;
2204 				goto errout;
2205 			}
2206 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2207 					       ifla_vf_policy, NULL);
2208 			if (err < 0)
2209 				goto errout;
2210 			err = do_setvfinfo(dev, vfinfo);
2211 			if (err < 0)
2212 				goto errout;
2213 			status |= DO_SETLINK_NOTIFY;
2214 		}
2215 	}
2216 	err = 0;
2217 
2218 	if (tb[IFLA_VF_PORTS]) {
2219 		struct nlattr *port[IFLA_PORT_MAX+1];
2220 		struct nlattr *attr;
2221 		int vf;
2222 		int rem;
2223 
2224 		err = -EOPNOTSUPP;
2225 		if (!ops->ndo_set_vf_port)
2226 			goto errout;
2227 
2228 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2229 			if (nla_type(attr) != IFLA_VF_PORT ||
2230 			    nla_len(attr) < NLA_HDRLEN) {
2231 				err = -EINVAL;
2232 				goto errout;
2233 			}
2234 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2235 					       ifla_port_policy, NULL);
2236 			if (err < 0)
2237 				goto errout;
2238 			if (!port[IFLA_PORT_VF]) {
2239 				err = -EOPNOTSUPP;
2240 				goto errout;
2241 			}
2242 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2243 			err = ops->ndo_set_vf_port(dev, vf, port);
2244 			if (err < 0)
2245 				goto errout;
2246 			status |= DO_SETLINK_NOTIFY;
2247 		}
2248 	}
2249 	err = 0;
2250 
2251 	if (tb[IFLA_PORT_SELF]) {
2252 		struct nlattr *port[IFLA_PORT_MAX+1];
2253 
2254 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2255 				       tb[IFLA_PORT_SELF], ifla_port_policy,
2256 				       NULL);
2257 		if (err < 0)
2258 			goto errout;
2259 
2260 		err = -EOPNOTSUPP;
2261 		if (ops->ndo_set_vf_port)
2262 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2263 		if (err < 0)
2264 			goto errout;
2265 		status |= DO_SETLINK_NOTIFY;
2266 	}
2267 
2268 	if (tb[IFLA_AF_SPEC]) {
2269 		struct nlattr *af;
2270 		int rem;
2271 
2272 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2273 			const struct rtnl_af_ops *af_ops;
2274 
2275 			rcu_read_lock();
2276 
2277 			BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2278 
2279 			err = af_ops->set_link_af(dev, af);
2280 			if (err < 0) {
2281 				rcu_read_unlock();
2282 				goto errout;
2283 			}
2284 
2285 			rcu_read_unlock();
2286 			status |= DO_SETLINK_NOTIFY;
2287 		}
2288 	}
2289 	err = 0;
2290 
2291 	if (tb[IFLA_PROTO_DOWN]) {
2292 		err = dev_change_proto_down(dev,
2293 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2294 		if (err)
2295 			goto errout;
2296 		status |= DO_SETLINK_NOTIFY;
2297 	}
2298 
2299 	if (tb[IFLA_XDP]) {
2300 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2301 		u32 xdp_flags = 0;
2302 
2303 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2304 				       ifla_xdp_policy, NULL);
2305 		if (err < 0)
2306 			goto errout;
2307 
2308 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2309 			err = -EINVAL;
2310 			goto errout;
2311 		}
2312 
2313 		if (xdp[IFLA_XDP_FLAGS]) {
2314 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2315 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2316 				err = -EINVAL;
2317 				goto errout;
2318 			}
2319 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2320 				err = -EINVAL;
2321 				goto errout;
2322 			}
2323 		}
2324 
2325 		if (xdp[IFLA_XDP_FD]) {
2326 			err = dev_change_xdp_fd(dev, extack,
2327 						nla_get_s32(xdp[IFLA_XDP_FD]),
2328 						xdp_flags);
2329 			if (err)
2330 				goto errout;
2331 			status |= DO_SETLINK_NOTIFY;
2332 		}
2333 	}
2334 
2335 errout:
2336 	if (status & DO_SETLINK_MODIFIED) {
2337 		if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2338 			netdev_state_change(dev);
2339 
2340 		if (err < 0)
2341 			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",
2342 					     dev->name);
2343 	}
2344 
2345 	return err;
2346 }
2347 
2348 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2349 			struct netlink_ext_ack *extack)
2350 {
2351 	struct net *net = sock_net(skb->sk);
2352 	struct ifinfomsg *ifm;
2353 	struct net_device *dev;
2354 	int err;
2355 	struct nlattr *tb[IFLA_MAX+1];
2356 	char ifname[IFNAMSIZ];
2357 
2358 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2359 			  extack);
2360 	if (err < 0)
2361 		goto errout;
2362 
2363 	if (tb[IFLA_IFNAME])
2364 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2365 	else
2366 		ifname[0] = '\0';
2367 
2368 	err = -EINVAL;
2369 	ifm = nlmsg_data(nlh);
2370 	if (ifm->ifi_index > 0)
2371 		dev = __dev_get_by_index(net, ifm->ifi_index);
2372 	else if (tb[IFLA_IFNAME])
2373 		dev = __dev_get_by_name(net, ifname);
2374 	else
2375 		goto errout;
2376 
2377 	if (dev == NULL) {
2378 		err = -ENODEV;
2379 		goto errout;
2380 	}
2381 
2382 	err = validate_linkmsg(dev, tb);
2383 	if (err < 0)
2384 		goto errout;
2385 
2386 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2387 errout:
2388 	return err;
2389 }
2390 
2391 static int rtnl_group_dellink(const struct net *net, int group)
2392 {
2393 	struct net_device *dev, *aux;
2394 	LIST_HEAD(list_kill);
2395 	bool found = false;
2396 
2397 	if (!group)
2398 		return -EPERM;
2399 
2400 	for_each_netdev(net, dev) {
2401 		if (dev->group == group) {
2402 			const struct rtnl_link_ops *ops;
2403 
2404 			found = true;
2405 			ops = dev->rtnl_link_ops;
2406 			if (!ops || !ops->dellink)
2407 				return -EOPNOTSUPP;
2408 		}
2409 	}
2410 
2411 	if (!found)
2412 		return -ENODEV;
2413 
2414 	for_each_netdev_safe(net, dev, aux) {
2415 		if (dev->group == group) {
2416 			const struct rtnl_link_ops *ops;
2417 
2418 			ops = dev->rtnl_link_ops;
2419 			ops->dellink(dev, &list_kill);
2420 		}
2421 	}
2422 	unregister_netdevice_many(&list_kill);
2423 
2424 	return 0;
2425 }
2426 
2427 int rtnl_delete_link(struct net_device *dev)
2428 {
2429 	const struct rtnl_link_ops *ops;
2430 	LIST_HEAD(list_kill);
2431 
2432 	ops = dev->rtnl_link_ops;
2433 	if (!ops || !ops->dellink)
2434 		return -EOPNOTSUPP;
2435 
2436 	ops->dellink(dev, &list_kill);
2437 	unregister_netdevice_many(&list_kill);
2438 
2439 	return 0;
2440 }
2441 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2442 
2443 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2444 			struct netlink_ext_ack *extack)
2445 {
2446 	struct net *net = sock_net(skb->sk);
2447 	struct net_device *dev;
2448 	struct ifinfomsg *ifm;
2449 	char ifname[IFNAMSIZ];
2450 	struct nlattr *tb[IFLA_MAX+1];
2451 	int err;
2452 
2453 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2454 	if (err < 0)
2455 		return err;
2456 
2457 	if (tb[IFLA_IFNAME])
2458 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2459 
2460 	ifm = nlmsg_data(nlh);
2461 	if (ifm->ifi_index > 0)
2462 		dev = __dev_get_by_index(net, ifm->ifi_index);
2463 	else if (tb[IFLA_IFNAME])
2464 		dev = __dev_get_by_name(net, ifname);
2465 	else if (tb[IFLA_GROUP])
2466 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2467 	else
2468 		return -EINVAL;
2469 
2470 	if (!dev)
2471 		return -ENODEV;
2472 
2473 	return rtnl_delete_link(dev);
2474 }
2475 
2476 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2477 {
2478 	unsigned int old_flags;
2479 	int err;
2480 
2481 	old_flags = dev->flags;
2482 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2483 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2484 		if (err < 0)
2485 			return err;
2486 	}
2487 
2488 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2489 
2490 	__dev_notify_flags(dev, old_flags, ~0U);
2491 	return 0;
2492 }
2493 EXPORT_SYMBOL(rtnl_configure_link);
2494 
2495 struct net_device *rtnl_create_link(struct net *net,
2496 	const char *ifname, unsigned char name_assign_type,
2497 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2498 {
2499 	struct net_device *dev;
2500 	unsigned int num_tx_queues = 1;
2501 	unsigned int num_rx_queues = 1;
2502 
2503 	if (tb[IFLA_NUM_TX_QUEUES])
2504 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2505 	else if (ops->get_num_tx_queues)
2506 		num_tx_queues = ops->get_num_tx_queues();
2507 
2508 	if (tb[IFLA_NUM_RX_QUEUES])
2509 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2510 	else if (ops->get_num_rx_queues)
2511 		num_rx_queues = ops->get_num_rx_queues();
2512 
2513 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2514 			       ops->setup, num_tx_queues, num_rx_queues);
2515 	if (!dev)
2516 		return ERR_PTR(-ENOMEM);
2517 
2518 	dev_net_set(dev, net);
2519 	dev->rtnl_link_ops = ops;
2520 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2521 
2522 	if (tb[IFLA_MTU])
2523 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2524 	if (tb[IFLA_ADDRESS]) {
2525 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2526 				nla_len(tb[IFLA_ADDRESS]));
2527 		dev->addr_assign_type = NET_ADDR_SET;
2528 	}
2529 	if (tb[IFLA_BROADCAST])
2530 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2531 				nla_len(tb[IFLA_BROADCAST]));
2532 	if (tb[IFLA_TXQLEN])
2533 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2534 	if (tb[IFLA_OPERSTATE])
2535 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2536 	if (tb[IFLA_LINKMODE])
2537 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2538 	if (tb[IFLA_GROUP])
2539 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2540 
2541 	return dev;
2542 }
2543 EXPORT_SYMBOL(rtnl_create_link);
2544 
2545 static int rtnl_group_changelink(const struct sk_buff *skb,
2546 		struct net *net, int group,
2547 		struct ifinfomsg *ifm,
2548 		struct netlink_ext_ack *extack,
2549 		struct nlattr **tb)
2550 {
2551 	struct net_device *dev, *aux;
2552 	int err;
2553 
2554 	for_each_netdev_safe(net, dev, aux) {
2555 		if (dev->group == group) {
2556 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2557 			if (err < 0)
2558 				return err;
2559 		}
2560 	}
2561 
2562 	return 0;
2563 }
2564 
2565 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2566 			struct netlink_ext_ack *extack)
2567 {
2568 	struct net *net = sock_net(skb->sk);
2569 	const struct rtnl_link_ops *ops;
2570 	const struct rtnl_link_ops *m_ops = NULL;
2571 	struct net_device *dev;
2572 	struct net_device *master_dev = NULL;
2573 	struct ifinfomsg *ifm;
2574 	char kind[MODULE_NAME_LEN];
2575 	char ifname[IFNAMSIZ];
2576 	struct nlattr *tb[IFLA_MAX+1];
2577 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2578 	unsigned char name_assign_type = NET_NAME_USER;
2579 	int err;
2580 
2581 #ifdef CONFIG_MODULES
2582 replay:
2583 #endif
2584 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2585 	if (err < 0)
2586 		return err;
2587 
2588 	if (tb[IFLA_IFNAME])
2589 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2590 	else
2591 		ifname[0] = '\0';
2592 
2593 	ifm = nlmsg_data(nlh);
2594 	if (ifm->ifi_index > 0)
2595 		dev = __dev_get_by_index(net, ifm->ifi_index);
2596 	else {
2597 		if (ifname[0])
2598 			dev = __dev_get_by_name(net, ifname);
2599 		else
2600 			dev = NULL;
2601 	}
2602 
2603 	if (dev) {
2604 		master_dev = netdev_master_upper_dev_get(dev);
2605 		if (master_dev)
2606 			m_ops = master_dev->rtnl_link_ops;
2607 	}
2608 
2609 	err = validate_linkmsg(dev, tb);
2610 	if (err < 0)
2611 		return err;
2612 
2613 	if (tb[IFLA_LINKINFO]) {
2614 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2615 				       tb[IFLA_LINKINFO], ifla_info_policy,
2616 				       NULL);
2617 		if (err < 0)
2618 			return err;
2619 	} else
2620 		memset(linkinfo, 0, sizeof(linkinfo));
2621 
2622 	if (linkinfo[IFLA_INFO_KIND]) {
2623 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2624 		ops = rtnl_link_ops_get(kind);
2625 	} else {
2626 		kind[0] = '\0';
2627 		ops = NULL;
2628 	}
2629 
2630 	if (1) {
2631 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2632 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2633 		struct nlattr **data = NULL;
2634 		struct nlattr **slave_data = NULL;
2635 		struct net *dest_net, *link_net = NULL;
2636 
2637 		if (ops) {
2638 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2639 				err = nla_parse_nested(attr, ops->maxtype,
2640 						       linkinfo[IFLA_INFO_DATA],
2641 						       ops->policy, NULL);
2642 				if (err < 0)
2643 					return err;
2644 				data = attr;
2645 			}
2646 			if (ops->validate) {
2647 				err = ops->validate(tb, data, extack);
2648 				if (err < 0)
2649 					return err;
2650 			}
2651 		}
2652 
2653 		if (m_ops) {
2654 			if (m_ops->slave_maxtype &&
2655 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2656 				err = nla_parse_nested(slave_attr,
2657 						       m_ops->slave_maxtype,
2658 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2659 						       m_ops->slave_policy,
2660 						       NULL);
2661 				if (err < 0)
2662 					return err;
2663 				slave_data = slave_attr;
2664 			}
2665 		}
2666 
2667 		if (dev) {
2668 			int status = 0;
2669 
2670 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2671 				return -EEXIST;
2672 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2673 				return -EOPNOTSUPP;
2674 
2675 			if (linkinfo[IFLA_INFO_DATA]) {
2676 				if (!ops || ops != dev->rtnl_link_ops ||
2677 				    !ops->changelink)
2678 					return -EOPNOTSUPP;
2679 
2680 				err = ops->changelink(dev, tb, data, extack);
2681 				if (err < 0)
2682 					return err;
2683 				status |= DO_SETLINK_NOTIFY;
2684 			}
2685 
2686 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2687 				if (!m_ops || !m_ops->slave_changelink)
2688 					return -EOPNOTSUPP;
2689 
2690 				err = m_ops->slave_changelink(master_dev, dev,
2691 							      tb, slave_data,
2692 							      extack);
2693 				if (err < 0)
2694 					return err;
2695 				status |= DO_SETLINK_NOTIFY;
2696 			}
2697 
2698 			return do_setlink(skb, dev, ifm, extack, tb, ifname,
2699 					  status);
2700 		}
2701 
2702 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2703 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2704 				return rtnl_group_changelink(skb, net,
2705 						nla_get_u32(tb[IFLA_GROUP]),
2706 						ifm, extack, tb);
2707 			return -ENODEV;
2708 		}
2709 
2710 		if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2711 			return -EOPNOTSUPP;
2712 
2713 		if (!ops) {
2714 #ifdef CONFIG_MODULES
2715 			if (kind[0]) {
2716 				__rtnl_unlock();
2717 				request_module("rtnl-link-%s", kind);
2718 				rtnl_lock();
2719 				ops = rtnl_link_ops_get(kind);
2720 				if (ops)
2721 					goto replay;
2722 			}
2723 #endif
2724 			return -EOPNOTSUPP;
2725 		}
2726 
2727 		if (!ops->setup)
2728 			return -EOPNOTSUPP;
2729 
2730 		if (!ifname[0]) {
2731 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2732 			name_assign_type = NET_NAME_ENUM;
2733 		}
2734 
2735 		dest_net = rtnl_link_get_net(net, tb);
2736 		if (IS_ERR(dest_net))
2737 			return PTR_ERR(dest_net);
2738 
2739 		err = -EPERM;
2740 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2741 			goto out;
2742 
2743 		if (tb[IFLA_LINK_NETNSID]) {
2744 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2745 
2746 			link_net = get_net_ns_by_id(dest_net, id);
2747 			if (!link_net) {
2748 				err =  -EINVAL;
2749 				goto out;
2750 			}
2751 			err = -EPERM;
2752 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2753 				goto out;
2754 		}
2755 
2756 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2757 				       name_assign_type, ops, tb);
2758 		if (IS_ERR(dev)) {
2759 			err = PTR_ERR(dev);
2760 			goto out;
2761 		}
2762 
2763 		dev->ifindex = ifm->ifi_index;
2764 
2765 		if (ops->newlink) {
2766 			err = ops->newlink(link_net ? : net, dev, tb, data,
2767 					   extack);
2768 			/* Drivers should call free_netdev() in ->destructor
2769 			 * and unregister it on failure after registration
2770 			 * so that device could be finally freed in rtnl_unlock.
2771 			 */
2772 			if (err < 0) {
2773 				/* If device is not registered at all, free it now */
2774 				if (dev->reg_state == NETREG_UNINITIALIZED)
2775 					free_netdev(dev);
2776 				goto out;
2777 			}
2778 		} else {
2779 			err = register_netdevice(dev);
2780 			if (err < 0) {
2781 				free_netdev(dev);
2782 				goto out;
2783 			}
2784 		}
2785 		err = rtnl_configure_link(dev, ifm);
2786 		if (err < 0)
2787 			goto out_unregister;
2788 		if (link_net) {
2789 			err = dev_change_net_namespace(dev, dest_net, ifname);
2790 			if (err < 0)
2791 				goto out_unregister;
2792 		}
2793 		if (tb[IFLA_MASTER]) {
2794 			err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
2795 					    extack);
2796 			if (err)
2797 				goto out_unregister;
2798 		}
2799 out:
2800 		if (link_net)
2801 			put_net(link_net);
2802 		put_net(dest_net);
2803 		return err;
2804 out_unregister:
2805 		if (ops->newlink) {
2806 			LIST_HEAD(list_kill);
2807 
2808 			ops->dellink(dev, &list_kill);
2809 			unregister_netdevice_many(&list_kill);
2810 		} else {
2811 			unregister_netdevice(dev);
2812 		}
2813 		goto out;
2814 	}
2815 }
2816 
2817 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2818 			struct netlink_ext_ack *extack)
2819 {
2820 	struct net *net = sock_net(skb->sk);
2821 	struct ifinfomsg *ifm;
2822 	char ifname[IFNAMSIZ];
2823 	struct nlattr *tb[IFLA_MAX+1];
2824 	struct net_device *dev = NULL;
2825 	struct sk_buff *nskb;
2826 	int err;
2827 	u32 ext_filter_mask = 0;
2828 
2829 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2830 	if (err < 0)
2831 		return err;
2832 
2833 	if (tb[IFLA_IFNAME])
2834 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2835 
2836 	if (tb[IFLA_EXT_MASK])
2837 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2838 
2839 	ifm = nlmsg_data(nlh);
2840 	if (ifm->ifi_index > 0)
2841 		dev = __dev_get_by_index(net, ifm->ifi_index);
2842 	else if (tb[IFLA_IFNAME])
2843 		dev = __dev_get_by_name(net, ifname);
2844 	else
2845 		return -EINVAL;
2846 
2847 	if (dev == NULL)
2848 		return -ENODEV;
2849 
2850 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2851 	if (nskb == NULL)
2852 		return -ENOBUFS;
2853 
2854 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2855 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0, NULL);
2856 	if (err < 0) {
2857 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2858 		WARN_ON(err == -EMSGSIZE);
2859 		kfree_skb(nskb);
2860 	} else
2861 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2862 
2863 	return err;
2864 }
2865 
2866 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2867 {
2868 	struct net *net = sock_net(skb->sk);
2869 	struct net_device *dev;
2870 	struct nlattr *tb[IFLA_MAX+1];
2871 	u32 ext_filter_mask = 0;
2872 	u16 min_ifinfo_dump_size = 0;
2873 	int hdrlen;
2874 
2875 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2876 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2877 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2878 
2879 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2880 		if (tb[IFLA_EXT_MASK])
2881 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2882 	}
2883 
2884 	if (!ext_filter_mask)
2885 		return NLMSG_GOODSIZE;
2886 	/*
2887 	 * traverse the list of net devices and compute the minimum
2888 	 * buffer size based upon the filter mask.
2889 	 */
2890 	rcu_read_lock();
2891 	for_each_netdev_rcu(net, dev) {
2892 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2893 					     if_nlmsg_size(dev,
2894 						           ext_filter_mask));
2895 	}
2896 	rcu_read_unlock();
2897 
2898 	return nlmsg_total_size(min_ifinfo_dump_size);
2899 }
2900 
2901 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2902 {
2903 	int idx;
2904 	int s_idx = cb->family;
2905 
2906 	if (s_idx == 0)
2907 		s_idx = 1;
2908 
2909 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2910 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2911 		struct rtnl_link *handlers;
2912 		rtnl_dumpit_func dumpit;
2913 
2914 		if (idx < s_idx || idx == PF_PACKET)
2915 			continue;
2916 
2917 		handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
2918 		if (!handlers)
2919 			continue;
2920 
2921 		dumpit = READ_ONCE(handlers[type].dumpit);
2922 		if (!dumpit)
2923 			continue;
2924 
2925 		if (idx > s_idx) {
2926 			memset(&cb->args[0], 0, sizeof(cb->args));
2927 			cb->prev_seq = 0;
2928 			cb->seq = 0;
2929 		}
2930 		if (dumpit(skb, cb))
2931 			break;
2932 	}
2933 	cb->family = idx;
2934 
2935 	return skb->len;
2936 }
2937 
2938 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2939 				       unsigned int change,
2940 				       u32 event, gfp_t flags, int *new_nsid)
2941 {
2942 	struct net *net = dev_net(dev);
2943 	struct sk_buff *skb;
2944 	int err = -ENOBUFS;
2945 	size_t if_info_size;
2946 
2947 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2948 	if (skb == NULL)
2949 		goto errout;
2950 
2951 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event,
2952 			       new_nsid);
2953 	if (err < 0) {
2954 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2955 		WARN_ON(err == -EMSGSIZE);
2956 		kfree_skb(skb);
2957 		goto errout;
2958 	}
2959 	return skb;
2960 errout:
2961 	if (err < 0)
2962 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2963 	return NULL;
2964 }
2965 
2966 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2967 {
2968 	struct net *net = dev_net(dev);
2969 
2970 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2971 }
2972 
2973 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2974 			       unsigned int change, u32 event,
2975 			       gfp_t flags, int *new_nsid)
2976 {
2977 	struct sk_buff *skb;
2978 
2979 	if (dev->reg_state != NETREG_REGISTERED)
2980 		return;
2981 
2982 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid);
2983 	if (skb)
2984 		rtmsg_ifinfo_send(skb, dev, flags);
2985 }
2986 
2987 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2988 		  gfp_t flags)
2989 {
2990 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags, NULL);
2991 }
2992 
2993 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
2994 			 gfp_t flags, int *new_nsid)
2995 {
2996 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
2997 			   new_nsid);
2998 }
2999 
3000 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3001 				   struct net_device *dev,
3002 				   u8 *addr, u16 vid, u32 pid, u32 seq,
3003 				   int type, unsigned int flags,
3004 				   int nlflags, u16 ndm_state)
3005 {
3006 	struct nlmsghdr *nlh;
3007 	struct ndmsg *ndm;
3008 
3009 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3010 	if (!nlh)
3011 		return -EMSGSIZE;
3012 
3013 	ndm = nlmsg_data(nlh);
3014 	ndm->ndm_family  = AF_BRIDGE;
3015 	ndm->ndm_pad1	 = 0;
3016 	ndm->ndm_pad2    = 0;
3017 	ndm->ndm_flags	 = flags;
3018 	ndm->ndm_type	 = 0;
3019 	ndm->ndm_ifindex = dev->ifindex;
3020 	ndm->ndm_state   = ndm_state;
3021 
3022 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3023 		goto nla_put_failure;
3024 	if (vid)
3025 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3026 			goto nla_put_failure;
3027 
3028 	nlmsg_end(skb, nlh);
3029 	return 0;
3030 
3031 nla_put_failure:
3032 	nlmsg_cancel(skb, nlh);
3033 	return -EMSGSIZE;
3034 }
3035 
3036 static inline size_t rtnl_fdb_nlmsg_size(void)
3037 {
3038 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3039 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
3040 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
3041 	       0;
3042 }
3043 
3044 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3045 			    u16 ndm_state)
3046 {
3047 	struct net *net = dev_net(dev);
3048 	struct sk_buff *skb;
3049 	int err = -ENOBUFS;
3050 
3051 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3052 	if (!skb)
3053 		goto errout;
3054 
3055 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3056 				      0, 0, type, NTF_SELF, 0, ndm_state);
3057 	if (err < 0) {
3058 		kfree_skb(skb);
3059 		goto errout;
3060 	}
3061 
3062 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3063 	return;
3064 errout:
3065 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3066 }
3067 
3068 /**
3069  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3070  */
3071 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3072 		     struct nlattr *tb[],
3073 		     struct net_device *dev,
3074 		     const unsigned char *addr, u16 vid,
3075 		     u16 flags)
3076 {
3077 	int err = -EINVAL;
3078 
3079 	/* If aging addresses are supported device will need to
3080 	 * implement its own handler for this.
3081 	 */
3082 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3083 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3084 		return err;
3085 	}
3086 
3087 	if (vid) {
3088 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3089 		return err;
3090 	}
3091 
3092 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3093 		err = dev_uc_add_excl(dev, addr);
3094 	else if (is_multicast_ether_addr(addr))
3095 		err = dev_mc_add_excl(dev, addr);
3096 
3097 	/* Only return duplicate errors if NLM_F_EXCL is set */
3098 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3099 		err = 0;
3100 
3101 	return err;
3102 }
3103 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3104 
3105 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3106 			 struct netlink_ext_ack *extack)
3107 {
3108 	u16 vid = 0;
3109 
3110 	if (vlan_attr) {
3111 		if (nla_len(vlan_attr) != sizeof(u16)) {
3112 			NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3113 			return -EINVAL;
3114 		}
3115 
3116 		vid = nla_get_u16(vlan_attr);
3117 
3118 		if (!vid || vid >= VLAN_VID_MASK) {
3119 			NL_SET_ERR_MSG(extack, "invalid vlan id");
3120 			return -EINVAL;
3121 		}
3122 	}
3123 	*p_vid = vid;
3124 	return 0;
3125 }
3126 
3127 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3128 			struct netlink_ext_ack *extack)
3129 {
3130 	struct net *net = sock_net(skb->sk);
3131 	struct ndmsg *ndm;
3132 	struct nlattr *tb[NDA_MAX+1];
3133 	struct net_device *dev;
3134 	u8 *addr;
3135 	u16 vid;
3136 	int err;
3137 
3138 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3139 	if (err < 0)
3140 		return err;
3141 
3142 	ndm = nlmsg_data(nlh);
3143 	if (ndm->ndm_ifindex == 0) {
3144 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3145 		return -EINVAL;
3146 	}
3147 
3148 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3149 	if (dev == NULL) {
3150 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3151 		return -ENODEV;
3152 	}
3153 
3154 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3155 		NL_SET_ERR_MSG(extack, "invalid address");
3156 		return -EINVAL;
3157 	}
3158 
3159 	addr = nla_data(tb[NDA_LLADDR]);
3160 
3161 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3162 	if (err)
3163 		return err;
3164 
3165 	err = -EOPNOTSUPP;
3166 
3167 	/* Support fdb on master device the net/bridge default case */
3168 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3169 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3170 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3171 		const struct net_device_ops *ops = br_dev->netdev_ops;
3172 
3173 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3174 				       nlh->nlmsg_flags);
3175 		if (err)
3176 			goto out;
3177 		else
3178 			ndm->ndm_flags &= ~NTF_MASTER;
3179 	}
3180 
3181 	/* Embedded bridge, macvlan, and any other device support */
3182 	if ((ndm->ndm_flags & NTF_SELF)) {
3183 		if (dev->netdev_ops->ndo_fdb_add)
3184 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3185 							   vid,
3186 							   nlh->nlmsg_flags);
3187 		else
3188 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3189 					       nlh->nlmsg_flags);
3190 
3191 		if (!err) {
3192 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3193 					ndm->ndm_state);
3194 			ndm->ndm_flags &= ~NTF_SELF;
3195 		}
3196 	}
3197 out:
3198 	return err;
3199 }
3200 
3201 /**
3202  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3203  */
3204 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3205 		     struct nlattr *tb[],
3206 		     struct net_device *dev,
3207 		     const unsigned char *addr, u16 vid)
3208 {
3209 	int err = -EINVAL;
3210 
3211 	/* If aging addresses are supported device will need to
3212 	 * implement its own handler for this.
3213 	 */
3214 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3215 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3216 		return err;
3217 	}
3218 
3219 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3220 		err = dev_uc_del(dev, addr);
3221 	else if (is_multicast_ether_addr(addr))
3222 		err = dev_mc_del(dev, addr);
3223 
3224 	return err;
3225 }
3226 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3227 
3228 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3229 			struct netlink_ext_ack *extack)
3230 {
3231 	struct net *net = sock_net(skb->sk);
3232 	struct ndmsg *ndm;
3233 	struct nlattr *tb[NDA_MAX+1];
3234 	struct net_device *dev;
3235 	int err = -EINVAL;
3236 	__u8 *addr;
3237 	u16 vid;
3238 
3239 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3240 		return -EPERM;
3241 
3242 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3243 	if (err < 0)
3244 		return err;
3245 
3246 	ndm = nlmsg_data(nlh);
3247 	if (ndm->ndm_ifindex == 0) {
3248 		NL_SET_ERR_MSG(extack, "invalid ifindex");
3249 		return -EINVAL;
3250 	}
3251 
3252 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3253 	if (dev == NULL) {
3254 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3255 		return -ENODEV;
3256 	}
3257 
3258 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3259 		NL_SET_ERR_MSG(extack, "invalid address");
3260 		return -EINVAL;
3261 	}
3262 
3263 	addr = nla_data(tb[NDA_LLADDR]);
3264 
3265 	err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3266 	if (err)
3267 		return err;
3268 
3269 	err = -EOPNOTSUPP;
3270 
3271 	/* Support fdb on master device the net/bridge default case */
3272 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3273 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3274 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3275 		const struct net_device_ops *ops = br_dev->netdev_ops;
3276 
3277 		if (ops->ndo_fdb_del)
3278 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3279 
3280 		if (err)
3281 			goto out;
3282 		else
3283 			ndm->ndm_flags &= ~NTF_MASTER;
3284 	}
3285 
3286 	/* Embedded bridge, macvlan, and any other device support */
3287 	if (ndm->ndm_flags & NTF_SELF) {
3288 		if (dev->netdev_ops->ndo_fdb_del)
3289 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3290 							   vid);
3291 		else
3292 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3293 
3294 		if (!err) {
3295 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3296 					ndm->ndm_state);
3297 			ndm->ndm_flags &= ~NTF_SELF;
3298 		}
3299 	}
3300 out:
3301 	return err;
3302 }
3303 
3304 static int nlmsg_populate_fdb(struct sk_buff *skb,
3305 			      struct netlink_callback *cb,
3306 			      struct net_device *dev,
3307 			      int *idx,
3308 			      struct netdev_hw_addr_list *list)
3309 {
3310 	struct netdev_hw_addr *ha;
3311 	int err;
3312 	u32 portid, seq;
3313 
3314 	portid = NETLINK_CB(cb->skb).portid;
3315 	seq = cb->nlh->nlmsg_seq;
3316 
3317 	list_for_each_entry(ha, &list->list, list) {
3318 		if (*idx < cb->args[2])
3319 			goto skip;
3320 
3321 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3322 					      portid, seq,
3323 					      RTM_NEWNEIGH, NTF_SELF,
3324 					      NLM_F_MULTI, NUD_PERMANENT);
3325 		if (err < 0)
3326 			return err;
3327 skip:
3328 		*idx += 1;
3329 	}
3330 	return 0;
3331 }
3332 
3333 /**
3334  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3335  * @nlh: netlink message header
3336  * @dev: netdevice
3337  *
3338  * Default netdevice operation to dump the existing unicast address list.
3339  * Returns number of addresses from list put in skb.
3340  */
3341 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3342 		      struct netlink_callback *cb,
3343 		      struct net_device *dev,
3344 		      struct net_device *filter_dev,
3345 		      int *idx)
3346 {
3347 	int err;
3348 
3349 	netif_addr_lock_bh(dev);
3350 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3351 	if (err)
3352 		goto out;
3353 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3354 out:
3355 	netif_addr_unlock_bh(dev);
3356 	return err;
3357 }
3358 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3359 
3360 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3361 {
3362 	struct net_device *dev;
3363 	struct nlattr *tb[IFLA_MAX+1];
3364 	struct net_device *br_dev = NULL;
3365 	const struct net_device_ops *ops = NULL;
3366 	const struct net_device_ops *cops = NULL;
3367 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3368 	struct net *net = sock_net(skb->sk);
3369 	struct hlist_head *head;
3370 	int brport_idx = 0;
3371 	int br_idx = 0;
3372 	int h, s_h;
3373 	int idx = 0, s_idx;
3374 	int err = 0;
3375 	int fidx = 0;
3376 
3377 	err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3378 			  IFLA_MAX, ifla_policy, NULL);
3379 	if (err < 0) {
3380 		return -EINVAL;
3381 	} else if (err == 0) {
3382 		if (tb[IFLA_MASTER])
3383 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3384 	}
3385 
3386 	brport_idx = ifm->ifi_index;
3387 
3388 	if (br_idx) {
3389 		br_dev = __dev_get_by_index(net, br_idx);
3390 		if (!br_dev)
3391 			return -ENODEV;
3392 
3393 		ops = br_dev->netdev_ops;
3394 	}
3395 
3396 	s_h = cb->args[0];
3397 	s_idx = cb->args[1];
3398 
3399 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3400 		idx = 0;
3401 		head = &net->dev_index_head[h];
3402 		hlist_for_each_entry(dev, head, index_hlist) {
3403 
3404 			if (brport_idx && (dev->ifindex != brport_idx))
3405 				continue;
3406 
3407 			if (!br_idx) { /* user did not specify a specific bridge */
3408 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3409 					br_dev = netdev_master_upper_dev_get(dev);
3410 					cops = br_dev->netdev_ops;
3411 				}
3412 			} else {
3413 				if (dev != br_dev &&
3414 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3415 					continue;
3416 
3417 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3418 				    !(dev->priv_flags & IFF_EBRIDGE))
3419 					continue;
3420 				cops = ops;
3421 			}
3422 
3423 			if (idx < s_idx)
3424 				goto cont;
3425 
3426 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3427 				if (cops && cops->ndo_fdb_dump) {
3428 					err = cops->ndo_fdb_dump(skb, cb,
3429 								br_dev, dev,
3430 								&fidx);
3431 					if (err == -EMSGSIZE)
3432 						goto out;
3433 				}
3434 			}
3435 
3436 			if (dev->netdev_ops->ndo_fdb_dump)
3437 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3438 								    dev, NULL,
3439 								    &fidx);
3440 			else
3441 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3442 							&fidx);
3443 			if (err == -EMSGSIZE)
3444 				goto out;
3445 
3446 			cops = NULL;
3447 
3448 			/* reset fdb offset to 0 for rest of the interfaces */
3449 			cb->args[2] = 0;
3450 			fidx = 0;
3451 cont:
3452 			idx++;
3453 		}
3454 	}
3455 
3456 out:
3457 	cb->args[0] = h;
3458 	cb->args[1] = idx;
3459 	cb->args[2] = fidx;
3460 
3461 	return skb->len;
3462 }
3463 
3464 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3465 			       unsigned int attrnum, unsigned int flag)
3466 {
3467 	if (mask & flag)
3468 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3469 	return 0;
3470 }
3471 
3472 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3473 			    struct net_device *dev, u16 mode,
3474 			    u32 flags, u32 mask, int nlflags,
3475 			    u32 filter_mask,
3476 			    int (*vlan_fill)(struct sk_buff *skb,
3477 					     struct net_device *dev,
3478 					     u32 filter_mask))
3479 {
3480 	struct nlmsghdr *nlh;
3481 	struct ifinfomsg *ifm;
3482 	struct nlattr *br_afspec;
3483 	struct nlattr *protinfo;
3484 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3485 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3486 	int err = 0;
3487 
3488 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3489 	if (nlh == NULL)
3490 		return -EMSGSIZE;
3491 
3492 	ifm = nlmsg_data(nlh);
3493 	ifm->ifi_family = AF_BRIDGE;
3494 	ifm->__ifi_pad = 0;
3495 	ifm->ifi_type = dev->type;
3496 	ifm->ifi_index = dev->ifindex;
3497 	ifm->ifi_flags = dev_get_flags(dev);
3498 	ifm->ifi_change = 0;
3499 
3500 
3501 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3502 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3503 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3504 	    (br_dev &&
3505 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3506 	    (dev->addr_len &&
3507 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3508 	    (dev->ifindex != dev_get_iflink(dev) &&
3509 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3510 		goto nla_put_failure;
3511 
3512 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3513 	if (!br_afspec)
3514 		goto nla_put_failure;
3515 
3516 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3517 		nla_nest_cancel(skb, br_afspec);
3518 		goto nla_put_failure;
3519 	}
3520 
3521 	if (mode != BRIDGE_MODE_UNDEF) {
3522 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3523 			nla_nest_cancel(skb, br_afspec);
3524 			goto nla_put_failure;
3525 		}
3526 	}
3527 	if (vlan_fill) {
3528 		err = vlan_fill(skb, dev, filter_mask);
3529 		if (err) {
3530 			nla_nest_cancel(skb, br_afspec);
3531 			goto nla_put_failure;
3532 		}
3533 	}
3534 	nla_nest_end(skb, br_afspec);
3535 
3536 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3537 	if (!protinfo)
3538 		goto nla_put_failure;
3539 
3540 	if (brport_nla_put_flag(skb, flags, mask,
3541 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3542 	    brport_nla_put_flag(skb, flags, mask,
3543 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3544 	    brport_nla_put_flag(skb, flags, mask,
3545 				IFLA_BRPORT_FAST_LEAVE,
3546 				BR_MULTICAST_FAST_LEAVE) ||
3547 	    brport_nla_put_flag(skb, flags, mask,
3548 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3549 	    brport_nla_put_flag(skb, flags, mask,
3550 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3551 	    brport_nla_put_flag(skb, flags, mask,
3552 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3553 	    brport_nla_put_flag(skb, flags, mask,
3554 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3555 	    brport_nla_put_flag(skb, flags, mask,
3556 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3557 		nla_nest_cancel(skb, protinfo);
3558 		goto nla_put_failure;
3559 	}
3560 
3561 	nla_nest_end(skb, protinfo);
3562 
3563 	nlmsg_end(skb, nlh);
3564 	return 0;
3565 nla_put_failure:
3566 	nlmsg_cancel(skb, nlh);
3567 	return err ? err : -EMSGSIZE;
3568 }
3569 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3570 
3571 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3572 {
3573 	struct net *net = sock_net(skb->sk);
3574 	struct net_device *dev;
3575 	int idx = 0;
3576 	u32 portid = NETLINK_CB(cb->skb).portid;
3577 	u32 seq = cb->nlh->nlmsg_seq;
3578 	u32 filter_mask = 0;
3579 	int err;
3580 
3581 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3582 		struct nlattr *extfilt;
3583 
3584 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3585 					  IFLA_EXT_MASK);
3586 		if (extfilt) {
3587 			if (nla_len(extfilt) < sizeof(filter_mask))
3588 				return -EINVAL;
3589 
3590 			filter_mask = nla_get_u32(extfilt);
3591 		}
3592 	}
3593 
3594 	rcu_read_lock();
3595 	for_each_netdev_rcu(net, dev) {
3596 		const struct net_device_ops *ops = dev->netdev_ops;
3597 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3598 
3599 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3600 			if (idx >= cb->args[0]) {
3601 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3602 						skb, portid, seq, dev,
3603 						filter_mask, NLM_F_MULTI);
3604 				if (err < 0 && err != -EOPNOTSUPP) {
3605 					if (likely(skb->len))
3606 						break;
3607 
3608 					goto out_err;
3609 				}
3610 			}
3611 			idx++;
3612 		}
3613 
3614 		if (ops->ndo_bridge_getlink) {
3615 			if (idx >= cb->args[0]) {
3616 				err = ops->ndo_bridge_getlink(skb, portid,
3617 							      seq, dev,
3618 							      filter_mask,
3619 							      NLM_F_MULTI);
3620 				if (err < 0 && err != -EOPNOTSUPP) {
3621 					if (likely(skb->len))
3622 						break;
3623 
3624 					goto out_err;
3625 				}
3626 			}
3627 			idx++;
3628 		}
3629 	}
3630 	err = skb->len;
3631 out_err:
3632 	rcu_read_unlock();
3633 	cb->args[0] = idx;
3634 
3635 	return err;
3636 }
3637 
3638 static inline size_t bridge_nlmsg_size(void)
3639 {
3640 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3641 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3642 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3643 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3644 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3645 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3646 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3647 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3648 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3649 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3650 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3651 }
3652 
3653 static int rtnl_bridge_notify(struct net_device *dev)
3654 {
3655 	struct net *net = dev_net(dev);
3656 	struct sk_buff *skb;
3657 	int err = -EOPNOTSUPP;
3658 
3659 	if (!dev->netdev_ops->ndo_bridge_getlink)
3660 		return 0;
3661 
3662 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3663 	if (!skb) {
3664 		err = -ENOMEM;
3665 		goto errout;
3666 	}
3667 
3668 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3669 	if (err < 0)
3670 		goto errout;
3671 
3672 	if (!skb->len)
3673 		goto errout;
3674 
3675 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3676 	return 0;
3677 errout:
3678 	WARN_ON(err == -EMSGSIZE);
3679 	kfree_skb(skb);
3680 	if (err)
3681 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3682 	return err;
3683 }
3684 
3685 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3686 			       struct netlink_ext_ack *extack)
3687 {
3688 	struct net *net = sock_net(skb->sk);
3689 	struct ifinfomsg *ifm;
3690 	struct net_device *dev;
3691 	struct nlattr *br_spec, *attr = NULL;
3692 	int rem, err = -EOPNOTSUPP;
3693 	u16 flags = 0;
3694 	bool have_flags = false;
3695 
3696 	if (nlmsg_len(nlh) < sizeof(*ifm))
3697 		return -EINVAL;
3698 
3699 	ifm = nlmsg_data(nlh);
3700 	if (ifm->ifi_family != AF_BRIDGE)
3701 		return -EPFNOSUPPORT;
3702 
3703 	dev = __dev_get_by_index(net, ifm->ifi_index);
3704 	if (!dev) {
3705 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3706 		return -ENODEV;
3707 	}
3708 
3709 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3710 	if (br_spec) {
3711 		nla_for_each_nested(attr, br_spec, rem) {
3712 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3713 				if (nla_len(attr) < sizeof(flags))
3714 					return -EINVAL;
3715 
3716 				have_flags = true;
3717 				flags = nla_get_u16(attr);
3718 				break;
3719 			}
3720 		}
3721 	}
3722 
3723 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3724 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3725 
3726 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3727 			err = -EOPNOTSUPP;
3728 			goto out;
3729 		}
3730 
3731 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3732 		if (err)
3733 			goto out;
3734 
3735 		flags &= ~BRIDGE_FLAGS_MASTER;
3736 	}
3737 
3738 	if ((flags & BRIDGE_FLAGS_SELF)) {
3739 		if (!dev->netdev_ops->ndo_bridge_setlink)
3740 			err = -EOPNOTSUPP;
3741 		else
3742 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3743 								  flags);
3744 		if (!err) {
3745 			flags &= ~BRIDGE_FLAGS_SELF;
3746 
3747 			/* Generate event to notify upper layer of bridge
3748 			 * change
3749 			 */
3750 			err = rtnl_bridge_notify(dev);
3751 		}
3752 	}
3753 
3754 	if (have_flags)
3755 		memcpy(nla_data(attr), &flags, sizeof(flags));
3756 out:
3757 	return err;
3758 }
3759 
3760 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3761 			       struct netlink_ext_ack *extack)
3762 {
3763 	struct net *net = sock_net(skb->sk);
3764 	struct ifinfomsg *ifm;
3765 	struct net_device *dev;
3766 	struct nlattr *br_spec, *attr = NULL;
3767 	int rem, err = -EOPNOTSUPP;
3768 	u16 flags = 0;
3769 	bool have_flags = false;
3770 
3771 	if (nlmsg_len(nlh) < sizeof(*ifm))
3772 		return -EINVAL;
3773 
3774 	ifm = nlmsg_data(nlh);
3775 	if (ifm->ifi_family != AF_BRIDGE)
3776 		return -EPFNOSUPPORT;
3777 
3778 	dev = __dev_get_by_index(net, ifm->ifi_index);
3779 	if (!dev) {
3780 		NL_SET_ERR_MSG(extack, "unknown ifindex");
3781 		return -ENODEV;
3782 	}
3783 
3784 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3785 	if (br_spec) {
3786 		nla_for_each_nested(attr, br_spec, rem) {
3787 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3788 				if (nla_len(attr) < sizeof(flags))
3789 					return -EINVAL;
3790 
3791 				have_flags = true;
3792 				flags = nla_get_u16(attr);
3793 				break;
3794 			}
3795 		}
3796 	}
3797 
3798 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3799 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3800 
3801 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3802 			err = -EOPNOTSUPP;
3803 			goto out;
3804 		}
3805 
3806 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3807 		if (err)
3808 			goto out;
3809 
3810 		flags &= ~BRIDGE_FLAGS_MASTER;
3811 	}
3812 
3813 	if ((flags & BRIDGE_FLAGS_SELF)) {
3814 		if (!dev->netdev_ops->ndo_bridge_dellink)
3815 			err = -EOPNOTSUPP;
3816 		else
3817 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3818 								  flags);
3819 
3820 		if (!err) {
3821 			flags &= ~BRIDGE_FLAGS_SELF;
3822 
3823 			/* Generate event to notify upper layer of bridge
3824 			 * change
3825 			 */
3826 			err = rtnl_bridge_notify(dev);
3827 		}
3828 	}
3829 
3830 	if (have_flags)
3831 		memcpy(nla_data(attr), &flags, sizeof(flags));
3832 out:
3833 	return err;
3834 }
3835 
3836 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3837 {
3838 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3839 	       (!idxattr || idxattr == attrid);
3840 }
3841 
3842 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3843 static int rtnl_get_offload_stats_attr_size(int attr_id)
3844 {
3845 	switch (attr_id) {
3846 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3847 		return sizeof(struct rtnl_link_stats64);
3848 	}
3849 
3850 	return 0;
3851 }
3852 
3853 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3854 				  int *prividx)
3855 {
3856 	struct nlattr *attr = NULL;
3857 	int attr_id, size;
3858 	void *attr_data;
3859 	int err;
3860 
3861 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3862 	      dev->netdev_ops->ndo_get_offload_stats))
3863 		return -ENODATA;
3864 
3865 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3866 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3867 		if (attr_id < *prividx)
3868 			continue;
3869 
3870 		size = rtnl_get_offload_stats_attr_size(attr_id);
3871 		if (!size)
3872 			continue;
3873 
3874 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3875 			continue;
3876 
3877 		attr = nla_reserve_64bit(skb, attr_id, size,
3878 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
3879 		if (!attr)
3880 			goto nla_put_failure;
3881 
3882 		attr_data = nla_data(attr);
3883 		memset(attr_data, 0, size);
3884 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3885 							     attr_data);
3886 		if (err)
3887 			goto get_offload_stats_failure;
3888 	}
3889 
3890 	if (!attr)
3891 		return -ENODATA;
3892 
3893 	*prividx = 0;
3894 	return 0;
3895 
3896 nla_put_failure:
3897 	err = -EMSGSIZE;
3898 get_offload_stats_failure:
3899 	*prividx = attr_id;
3900 	return err;
3901 }
3902 
3903 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3904 {
3905 	int nla_size = 0;
3906 	int attr_id;
3907 	int size;
3908 
3909 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3910 	      dev->netdev_ops->ndo_get_offload_stats))
3911 		return 0;
3912 
3913 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3914 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3915 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3916 			continue;
3917 		size = rtnl_get_offload_stats_attr_size(attr_id);
3918 		nla_size += nla_total_size_64bit(size);
3919 	}
3920 
3921 	if (nla_size != 0)
3922 		nla_size += nla_total_size(0);
3923 
3924 	return nla_size;
3925 }
3926 
3927 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3928 			       int type, u32 pid, u32 seq, u32 change,
3929 			       unsigned int flags, unsigned int filter_mask,
3930 			       int *idxattr, int *prividx)
3931 {
3932 	struct if_stats_msg *ifsm;
3933 	struct nlmsghdr *nlh;
3934 	struct nlattr *attr;
3935 	int s_prividx = *prividx;
3936 	int err;
3937 
3938 	ASSERT_RTNL();
3939 
3940 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3941 	if (!nlh)
3942 		return -EMSGSIZE;
3943 
3944 	ifsm = nlmsg_data(nlh);
3945 	ifsm->family = PF_UNSPEC;
3946 	ifsm->pad1 = 0;
3947 	ifsm->pad2 = 0;
3948 	ifsm->ifindex = dev->ifindex;
3949 	ifsm->filter_mask = filter_mask;
3950 
3951 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3952 		struct rtnl_link_stats64 *sp;
3953 
3954 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3955 					 sizeof(struct rtnl_link_stats64),
3956 					 IFLA_STATS_UNSPEC);
3957 		if (!attr)
3958 			goto nla_put_failure;
3959 
3960 		sp = nla_data(attr);
3961 		dev_get_stats(dev, sp);
3962 	}
3963 
3964 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3965 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3966 
3967 		if (ops && ops->fill_linkxstats) {
3968 			*idxattr = IFLA_STATS_LINK_XSTATS;
3969 			attr = nla_nest_start(skb,
3970 					      IFLA_STATS_LINK_XSTATS);
3971 			if (!attr)
3972 				goto nla_put_failure;
3973 
3974 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3975 			nla_nest_end(skb, attr);
3976 			if (err)
3977 				goto nla_put_failure;
3978 			*idxattr = 0;
3979 		}
3980 	}
3981 
3982 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3983 			     *idxattr)) {
3984 		const struct rtnl_link_ops *ops = NULL;
3985 		const struct net_device *master;
3986 
3987 		master = netdev_master_upper_dev_get(dev);
3988 		if (master)
3989 			ops = master->rtnl_link_ops;
3990 		if (ops && ops->fill_linkxstats) {
3991 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3992 			attr = nla_nest_start(skb,
3993 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3994 			if (!attr)
3995 				goto nla_put_failure;
3996 
3997 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3998 			nla_nest_end(skb, attr);
3999 			if (err)
4000 				goto nla_put_failure;
4001 			*idxattr = 0;
4002 		}
4003 	}
4004 
4005 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4006 			     *idxattr)) {
4007 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4008 		attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4009 		if (!attr)
4010 			goto nla_put_failure;
4011 
4012 		err = rtnl_get_offload_stats(skb, dev, prividx);
4013 		if (err == -ENODATA)
4014 			nla_nest_cancel(skb, attr);
4015 		else
4016 			nla_nest_end(skb, attr);
4017 
4018 		if (err && err != -ENODATA)
4019 			goto nla_put_failure;
4020 		*idxattr = 0;
4021 	}
4022 
4023 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4024 		struct rtnl_af_ops *af_ops;
4025 
4026 		*idxattr = IFLA_STATS_AF_SPEC;
4027 		attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4028 		if (!attr)
4029 			goto nla_put_failure;
4030 
4031 		rcu_read_lock();
4032 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4033 			if (af_ops->fill_stats_af) {
4034 				struct nlattr *af;
4035 				int err;
4036 
4037 				af = nla_nest_start(skb, af_ops->family);
4038 				if (!af) {
4039 					rcu_read_unlock();
4040 					goto nla_put_failure;
4041 				}
4042 				err = af_ops->fill_stats_af(skb, dev);
4043 
4044 				if (err == -ENODATA) {
4045 					nla_nest_cancel(skb, af);
4046 				} else if (err < 0) {
4047 					rcu_read_unlock();
4048 					goto nla_put_failure;
4049 				}
4050 
4051 				nla_nest_end(skb, af);
4052 			}
4053 		}
4054 		rcu_read_unlock();
4055 
4056 		nla_nest_end(skb, attr);
4057 
4058 		*idxattr = 0;
4059 	}
4060 
4061 	nlmsg_end(skb, nlh);
4062 
4063 	return 0;
4064 
4065 nla_put_failure:
4066 	/* not a multi message or no progress mean a real error */
4067 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4068 		nlmsg_cancel(skb, nlh);
4069 	else
4070 		nlmsg_end(skb, nlh);
4071 
4072 	return -EMSGSIZE;
4073 }
4074 
4075 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4076 				  u32 filter_mask)
4077 {
4078 	size_t size = 0;
4079 
4080 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4081 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4082 
4083 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4084 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4085 		int attr = IFLA_STATS_LINK_XSTATS;
4086 
4087 		if (ops && ops->get_linkxstats_size) {
4088 			size += nla_total_size(ops->get_linkxstats_size(dev,
4089 									attr));
4090 			/* for IFLA_STATS_LINK_XSTATS */
4091 			size += nla_total_size(0);
4092 		}
4093 	}
4094 
4095 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4096 		struct net_device *_dev = (struct net_device *)dev;
4097 		const struct rtnl_link_ops *ops = NULL;
4098 		const struct net_device *master;
4099 
4100 		/* netdev_master_upper_dev_get can't take const */
4101 		master = netdev_master_upper_dev_get(_dev);
4102 		if (master)
4103 			ops = master->rtnl_link_ops;
4104 		if (ops && ops->get_linkxstats_size) {
4105 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4106 
4107 			size += nla_total_size(ops->get_linkxstats_size(dev,
4108 									attr));
4109 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
4110 			size += nla_total_size(0);
4111 		}
4112 	}
4113 
4114 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4115 		size += rtnl_get_offload_stats_size(dev);
4116 
4117 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4118 		struct rtnl_af_ops *af_ops;
4119 
4120 		/* for IFLA_STATS_AF_SPEC */
4121 		size += nla_total_size(0);
4122 
4123 		rcu_read_lock();
4124 		list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4125 			if (af_ops->get_stats_af_size) {
4126 				size += nla_total_size(
4127 					af_ops->get_stats_af_size(dev));
4128 
4129 				/* for AF_* */
4130 				size += nla_total_size(0);
4131 			}
4132 		}
4133 		rcu_read_unlock();
4134 	}
4135 
4136 	return size;
4137 }
4138 
4139 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4140 			  struct netlink_ext_ack *extack)
4141 {
4142 	struct net *net = sock_net(skb->sk);
4143 	struct net_device *dev = NULL;
4144 	int idxattr = 0, prividx = 0;
4145 	struct if_stats_msg *ifsm;
4146 	struct sk_buff *nskb;
4147 	u32 filter_mask;
4148 	int err;
4149 
4150 	if (nlmsg_len(nlh) < sizeof(*ifsm))
4151 		return -EINVAL;
4152 
4153 	ifsm = nlmsg_data(nlh);
4154 	if (ifsm->ifindex > 0)
4155 		dev = __dev_get_by_index(net, ifsm->ifindex);
4156 	else
4157 		return -EINVAL;
4158 
4159 	if (!dev)
4160 		return -ENODEV;
4161 
4162 	filter_mask = ifsm->filter_mask;
4163 	if (!filter_mask)
4164 		return -EINVAL;
4165 
4166 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4167 	if (!nskb)
4168 		return -ENOBUFS;
4169 
4170 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4171 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4172 				  0, filter_mask, &idxattr, &prividx);
4173 	if (err < 0) {
4174 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4175 		WARN_ON(err == -EMSGSIZE);
4176 		kfree_skb(nskb);
4177 	} else {
4178 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4179 	}
4180 
4181 	return err;
4182 }
4183 
4184 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4185 {
4186 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
4187 	struct net *net = sock_net(skb->sk);
4188 	unsigned int flags = NLM_F_MULTI;
4189 	struct if_stats_msg *ifsm;
4190 	struct hlist_head *head;
4191 	struct net_device *dev;
4192 	u32 filter_mask = 0;
4193 	int idx = 0;
4194 
4195 	s_h = cb->args[0];
4196 	s_idx = cb->args[1];
4197 	s_idxattr = cb->args[2];
4198 	s_prividx = cb->args[3];
4199 
4200 	cb->seq = net->dev_base_seq;
4201 
4202 	if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4203 		return -EINVAL;
4204 
4205 	ifsm = nlmsg_data(cb->nlh);
4206 	filter_mask = ifsm->filter_mask;
4207 	if (!filter_mask)
4208 		return -EINVAL;
4209 
4210 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4211 		idx = 0;
4212 		head = &net->dev_index_head[h];
4213 		hlist_for_each_entry(dev, head, index_hlist) {
4214 			if (idx < s_idx)
4215 				goto cont;
4216 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4217 						  NETLINK_CB(cb->skb).portid,
4218 						  cb->nlh->nlmsg_seq, 0,
4219 						  flags, filter_mask,
4220 						  &s_idxattr, &s_prividx);
4221 			/* If we ran out of room on the first message,
4222 			 * we're in trouble
4223 			 */
4224 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4225 
4226 			if (err < 0)
4227 				goto out;
4228 			s_prividx = 0;
4229 			s_idxattr = 0;
4230 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4231 cont:
4232 			idx++;
4233 		}
4234 	}
4235 out:
4236 	cb->args[3] = s_prividx;
4237 	cb->args[2] = s_idxattr;
4238 	cb->args[1] = idx;
4239 	cb->args[0] = h;
4240 
4241 	return skb->len;
4242 }
4243 
4244 /* Process one rtnetlink message. */
4245 
4246 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4247 			     struct netlink_ext_ack *extack)
4248 {
4249 	struct net *net = sock_net(skb->sk);
4250 	struct rtnl_link *handlers;
4251 	int err = -EOPNOTSUPP;
4252 	rtnl_doit_func doit;
4253 	unsigned int flags;
4254 	int kind;
4255 	int family;
4256 	int type;
4257 
4258 	type = nlh->nlmsg_type;
4259 	if (type > RTM_MAX)
4260 		return -EOPNOTSUPP;
4261 
4262 	type -= RTM_BASE;
4263 
4264 	/* All the messages must have at least 1 byte length */
4265 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4266 		return 0;
4267 
4268 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4269 	kind = type&3;
4270 
4271 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4272 		return -EPERM;
4273 
4274 	if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4275 		family = PF_UNSPEC;
4276 
4277 	rcu_read_lock();
4278 	handlers = rcu_dereference(rtnl_msg_handlers[family]);
4279 	if (!handlers) {
4280 		family = PF_UNSPEC;
4281 		handlers = rcu_dereference(rtnl_msg_handlers[family]);
4282 	}
4283 
4284 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4285 		struct sock *rtnl;
4286 		rtnl_dumpit_func dumpit;
4287 		u16 min_dump_alloc = 0;
4288 
4289 		dumpit = READ_ONCE(handlers[type].dumpit);
4290 		if (!dumpit) {
4291 			family = PF_UNSPEC;
4292 			handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4293 			if (!handlers)
4294 				goto err_unlock;
4295 
4296 			dumpit = READ_ONCE(handlers[type].dumpit);
4297 			if (!dumpit)
4298 				goto err_unlock;
4299 		}
4300 
4301 		refcount_inc(&rtnl_msg_handlers_ref[family]);
4302 
4303 		if (type == RTM_GETLINK - RTM_BASE)
4304 			min_dump_alloc = rtnl_calcit(skb, nlh);
4305 
4306 		rcu_read_unlock();
4307 
4308 		rtnl = net->rtnl;
4309 		{
4310 			struct netlink_dump_control c = {
4311 				.dump		= dumpit,
4312 				.min_dump_alloc	= min_dump_alloc,
4313 			};
4314 			err = netlink_dump_start(rtnl, skb, nlh, &c);
4315 		}
4316 		refcount_dec(&rtnl_msg_handlers_ref[family]);
4317 		return err;
4318 	}
4319 
4320 	doit = READ_ONCE(handlers[type].doit);
4321 	if (!doit) {
4322 		family = PF_UNSPEC;
4323 		handlers = rcu_dereference(rtnl_msg_handlers[family]);
4324 	}
4325 
4326 	flags = READ_ONCE(handlers[type].flags);
4327 	if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4328 		refcount_inc(&rtnl_msg_handlers_ref[family]);
4329 		doit = READ_ONCE(handlers[type].doit);
4330 		rcu_read_unlock();
4331 		if (doit)
4332 			err = doit(skb, nlh, extack);
4333 		refcount_dec(&rtnl_msg_handlers_ref[family]);
4334 		return err;
4335 	}
4336 
4337 	rcu_read_unlock();
4338 
4339 	rtnl_lock();
4340 	handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4341 	if (handlers) {
4342 		doit = READ_ONCE(handlers[type].doit);
4343 		if (doit)
4344 			err = doit(skb, nlh, extack);
4345 	}
4346 	rtnl_unlock();
4347 	return err;
4348 
4349 err_unlock:
4350 	rcu_read_unlock();
4351 	return -EOPNOTSUPP;
4352 }
4353 
4354 static void rtnetlink_rcv(struct sk_buff *skb)
4355 {
4356 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4357 }
4358 
4359 static int rtnetlink_bind(struct net *net, int group)
4360 {
4361 	switch (group) {
4362 	case RTNLGRP_IPV4_MROUTE_R:
4363 	case RTNLGRP_IPV6_MROUTE_R:
4364 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4365 			return -EPERM;
4366 		break;
4367 	}
4368 	return 0;
4369 }
4370 
4371 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4372 {
4373 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4374 
4375 	switch (event) {
4376 	case NETDEV_REBOOT:
4377 	case NETDEV_CHANGEMTU:
4378 	case NETDEV_CHANGEADDR:
4379 	case NETDEV_CHANGENAME:
4380 	case NETDEV_FEAT_CHANGE:
4381 	case NETDEV_BONDING_FAILOVER:
4382 	case NETDEV_POST_TYPE_CHANGE:
4383 	case NETDEV_NOTIFY_PEERS:
4384 	case NETDEV_CHANGEUPPER:
4385 	case NETDEV_RESEND_IGMP:
4386 	case NETDEV_CHANGEINFODATA:
4387 	case NETDEV_CHANGELOWERSTATE:
4388 	case NETDEV_CHANGE_TX_QUEUE_LEN:
4389 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4390 				   GFP_KERNEL, NULL);
4391 		break;
4392 	default:
4393 		break;
4394 	}
4395 	return NOTIFY_DONE;
4396 }
4397 
4398 static struct notifier_block rtnetlink_dev_notifier = {
4399 	.notifier_call	= rtnetlink_event,
4400 };
4401 
4402 
4403 static int __net_init rtnetlink_net_init(struct net *net)
4404 {
4405 	struct sock *sk;
4406 	struct netlink_kernel_cfg cfg = {
4407 		.groups		= RTNLGRP_MAX,
4408 		.input		= rtnetlink_rcv,
4409 		.cb_mutex	= &rtnl_mutex,
4410 		.flags		= NL_CFG_F_NONROOT_RECV,
4411 		.bind		= rtnetlink_bind,
4412 	};
4413 
4414 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4415 	if (!sk)
4416 		return -ENOMEM;
4417 	net->rtnl = sk;
4418 	return 0;
4419 }
4420 
4421 static void __net_exit rtnetlink_net_exit(struct net *net)
4422 {
4423 	netlink_kernel_release(net->rtnl);
4424 	net->rtnl = NULL;
4425 }
4426 
4427 static struct pernet_operations rtnetlink_net_ops = {
4428 	.init = rtnetlink_net_init,
4429 	.exit = rtnetlink_net_exit,
4430 };
4431 
4432 void __init rtnetlink_init(void)
4433 {
4434 	int i;
4435 
4436 	for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4437 		refcount_set(&rtnl_msg_handlers_ref[i], 1);
4438 
4439 	if (register_pernet_subsys(&rtnetlink_net_ops))
4440 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
4441 
4442 	register_netdevice_notifier(&rtnetlink_dev_notifier);
4443 
4444 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4445 		      rtnl_dump_ifinfo, 0);
4446 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4447 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4448 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4449 
4450 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4451 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4452 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4453 
4454 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4455 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4456 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4457 
4458 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4459 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4460 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4461 
4462 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4463 		      0);
4464 }
4465