xref: /linux/net/core/rtnetlink.c (revision 9ee0034b8f49aaaa7e7c2da8db1038915db99c19)
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/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
29 #include <linux/mm.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
42 
43 #include <asm/uaccess.h>
44 
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
48 #include <net/ip.h>
49 #include <net/protocol.h>
50 #include <net/arp.h>
51 #include <net/route.h>
52 #include <net/udp.h>
53 #include <net/tcp.h>
54 #include <net/sock.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
59 
60 struct rtnl_link {
61 	rtnl_doit_func		doit;
62 	rtnl_dumpit_func	dumpit;
63 	rtnl_calcit_func 	calcit;
64 };
65 
66 static DEFINE_MUTEX(rtnl_mutex);
67 
68 void rtnl_lock(void)
69 {
70 	mutex_lock(&rtnl_mutex);
71 }
72 EXPORT_SYMBOL(rtnl_lock);
73 
74 static struct sk_buff *defer_kfree_skb_list;
75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
76 {
77 	if (head && tail) {
78 		tail->next = defer_kfree_skb_list;
79 		defer_kfree_skb_list = head;
80 	}
81 }
82 EXPORT_SYMBOL(rtnl_kfree_skbs);
83 
84 void __rtnl_unlock(void)
85 {
86 	struct sk_buff *head = defer_kfree_skb_list;
87 
88 	defer_kfree_skb_list = NULL;
89 
90 	mutex_unlock(&rtnl_mutex);
91 
92 	while (head) {
93 		struct sk_buff *next = head->next;
94 
95 		kfree_skb(head);
96 		cond_resched();
97 		head = next;
98 	}
99 }
100 
101 void rtnl_unlock(void)
102 {
103 	/* This fellow will unlock it for us. */
104 	netdev_run_todo();
105 }
106 EXPORT_SYMBOL(rtnl_unlock);
107 
108 int rtnl_trylock(void)
109 {
110 	return mutex_trylock(&rtnl_mutex);
111 }
112 EXPORT_SYMBOL(rtnl_trylock);
113 
114 int rtnl_is_locked(void)
115 {
116 	return mutex_is_locked(&rtnl_mutex);
117 }
118 EXPORT_SYMBOL(rtnl_is_locked);
119 
120 #ifdef CONFIG_PROVE_LOCKING
121 bool lockdep_rtnl_is_held(void)
122 {
123 	return lockdep_is_held(&rtnl_mutex);
124 }
125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
127 
128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
129 
130 static inline int rtm_msgindex(int msgtype)
131 {
132 	int msgindex = msgtype - RTM_BASE;
133 
134 	/*
135 	 * msgindex < 0 implies someone tried to register a netlink
136 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
137 	 * the message type has not been added to linux/rtnetlink.h
138 	 */
139 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
140 
141 	return msgindex;
142 }
143 
144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
145 {
146 	struct rtnl_link *tab;
147 
148 	if (protocol <= RTNL_FAMILY_MAX)
149 		tab = rtnl_msg_handlers[protocol];
150 	else
151 		tab = NULL;
152 
153 	if (tab == NULL || tab[msgindex].doit == NULL)
154 		tab = rtnl_msg_handlers[PF_UNSPEC];
155 
156 	return tab[msgindex].doit;
157 }
158 
159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
160 {
161 	struct rtnl_link *tab;
162 
163 	if (protocol <= RTNL_FAMILY_MAX)
164 		tab = rtnl_msg_handlers[protocol];
165 	else
166 		tab = NULL;
167 
168 	if (tab == NULL || tab[msgindex].dumpit == NULL)
169 		tab = rtnl_msg_handlers[PF_UNSPEC];
170 
171 	return tab[msgindex].dumpit;
172 }
173 
174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
175 {
176 	struct rtnl_link *tab;
177 
178 	if (protocol <= RTNL_FAMILY_MAX)
179 		tab = rtnl_msg_handlers[protocol];
180 	else
181 		tab = NULL;
182 
183 	if (tab == NULL || tab[msgindex].calcit == NULL)
184 		tab = rtnl_msg_handlers[PF_UNSPEC];
185 
186 	return tab[msgindex].calcit;
187 }
188 
189 /**
190  * __rtnl_register - Register a rtnetlink message type
191  * @protocol: Protocol family or PF_UNSPEC
192  * @msgtype: rtnetlink message type
193  * @doit: Function pointer called for each request message
194  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
195  * @calcit: Function pointer to calc size of dump message
196  *
197  * Registers the specified function pointers (at least one of them has
198  * to be non-NULL) to be called whenever a request message for the
199  * specified protocol family and message type is received.
200  *
201  * The special protocol family PF_UNSPEC may be used to define fallback
202  * function pointers for the case when no entry for the specific protocol
203  * family exists.
204  *
205  * Returns 0 on success or a negative error code.
206  */
207 int __rtnl_register(int protocol, int msgtype,
208 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
209 		    rtnl_calcit_func calcit)
210 {
211 	struct rtnl_link *tab;
212 	int msgindex;
213 
214 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
215 	msgindex = rtm_msgindex(msgtype);
216 
217 	tab = rtnl_msg_handlers[protocol];
218 	if (tab == NULL) {
219 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
220 		if (tab == NULL)
221 			return -ENOBUFS;
222 
223 		rtnl_msg_handlers[protocol] = tab;
224 	}
225 
226 	if (doit)
227 		tab[msgindex].doit = doit;
228 
229 	if (dumpit)
230 		tab[msgindex].dumpit = dumpit;
231 
232 	if (calcit)
233 		tab[msgindex].calcit = calcit;
234 
235 	return 0;
236 }
237 EXPORT_SYMBOL_GPL(__rtnl_register);
238 
239 /**
240  * rtnl_register - Register a rtnetlink message type
241  *
242  * Identical to __rtnl_register() but panics on failure. This is useful
243  * as failure of this function is very unlikely, it can only happen due
244  * to lack of memory when allocating the chain to store all message
245  * handlers for a protocol. Meant for use in init functions where lack
246  * of memory implies no sense in continuing.
247  */
248 void rtnl_register(int protocol, int msgtype,
249 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
250 		   rtnl_calcit_func calcit)
251 {
252 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
253 		panic("Unable to register rtnetlink message handler, "
254 		      "protocol = %d, message type = %d\n",
255 		      protocol, msgtype);
256 }
257 EXPORT_SYMBOL_GPL(rtnl_register);
258 
259 /**
260  * rtnl_unregister - Unregister a rtnetlink message type
261  * @protocol: Protocol family or PF_UNSPEC
262  * @msgtype: rtnetlink message type
263  *
264  * Returns 0 on success or a negative error code.
265  */
266 int rtnl_unregister(int protocol, int msgtype)
267 {
268 	int msgindex;
269 
270 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 	msgindex = rtm_msgindex(msgtype);
272 
273 	if (rtnl_msg_handlers[protocol] == NULL)
274 		return -ENOENT;
275 
276 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
277 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
278 
279 	return 0;
280 }
281 EXPORT_SYMBOL_GPL(rtnl_unregister);
282 
283 /**
284  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
285  * @protocol : Protocol family or PF_UNSPEC
286  *
287  * Identical to calling rtnl_unregster() for all registered message types
288  * of a certain protocol family.
289  */
290 void rtnl_unregister_all(int protocol)
291 {
292 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
293 
294 	kfree(rtnl_msg_handlers[protocol]);
295 	rtnl_msg_handlers[protocol] = NULL;
296 }
297 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
298 
299 static LIST_HEAD(link_ops);
300 
301 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
302 {
303 	const struct rtnl_link_ops *ops;
304 
305 	list_for_each_entry(ops, &link_ops, list) {
306 		if (!strcmp(ops->kind, kind))
307 			return ops;
308 	}
309 	return NULL;
310 }
311 
312 /**
313  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
314  * @ops: struct rtnl_link_ops * to register
315  *
316  * The caller must hold the rtnl_mutex. This function should be used
317  * by drivers that create devices during module initialization. It
318  * must be called before registering the devices.
319  *
320  * Returns 0 on success or a negative error code.
321  */
322 int __rtnl_link_register(struct rtnl_link_ops *ops)
323 {
324 	if (rtnl_link_ops_get(ops->kind))
325 		return -EEXIST;
326 
327 	/* The check for setup is here because if ops
328 	 * does not have that filled up, it is not possible
329 	 * to use the ops for creating device. So do not
330 	 * fill up dellink as well. That disables rtnl_dellink.
331 	 */
332 	if (ops->setup && !ops->dellink)
333 		ops->dellink = unregister_netdevice_queue;
334 
335 	list_add_tail(&ops->list, &link_ops);
336 	return 0;
337 }
338 EXPORT_SYMBOL_GPL(__rtnl_link_register);
339 
340 /**
341  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
342  * @ops: struct rtnl_link_ops * to register
343  *
344  * Returns 0 on success or a negative error code.
345  */
346 int rtnl_link_register(struct rtnl_link_ops *ops)
347 {
348 	int err;
349 
350 	rtnl_lock();
351 	err = __rtnl_link_register(ops);
352 	rtnl_unlock();
353 	return err;
354 }
355 EXPORT_SYMBOL_GPL(rtnl_link_register);
356 
357 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
358 {
359 	struct net_device *dev;
360 	LIST_HEAD(list_kill);
361 
362 	for_each_netdev(net, dev) {
363 		if (dev->rtnl_link_ops == ops)
364 			ops->dellink(dev, &list_kill);
365 	}
366 	unregister_netdevice_many(&list_kill);
367 }
368 
369 /**
370  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
371  * @ops: struct rtnl_link_ops * to unregister
372  *
373  * The caller must hold the rtnl_mutex.
374  */
375 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
376 {
377 	struct net *net;
378 
379 	for_each_net(net) {
380 		__rtnl_kill_links(net, ops);
381 	}
382 	list_del(&ops->list);
383 }
384 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
385 
386 /* Return with the rtnl_lock held when there are no network
387  * devices unregistering in any network namespace.
388  */
389 static void rtnl_lock_unregistering_all(void)
390 {
391 	struct net *net;
392 	bool unregistering;
393 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
394 
395 	add_wait_queue(&netdev_unregistering_wq, &wait);
396 	for (;;) {
397 		unregistering = false;
398 		rtnl_lock();
399 		for_each_net(net) {
400 			if (net->dev_unreg_count > 0) {
401 				unregistering = true;
402 				break;
403 			}
404 		}
405 		if (!unregistering)
406 			break;
407 		__rtnl_unlock();
408 
409 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
410 	}
411 	remove_wait_queue(&netdev_unregistering_wq, &wait);
412 }
413 
414 /**
415  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
416  * @ops: struct rtnl_link_ops * to unregister
417  */
418 void rtnl_link_unregister(struct rtnl_link_ops *ops)
419 {
420 	/* Close the race with cleanup_net() */
421 	mutex_lock(&net_mutex);
422 	rtnl_lock_unregistering_all();
423 	__rtnl_link_unregister(ops);
424 	rtnl_unlock();
425 	mutex_unlock(&net_mutex);
426 }
427 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
428 
429 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
430 {
431 	struct net_device *master_dev;
432 	const struct rtnl_link_ops *ops;
433 
434 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
435 	if (!master_dev)
436 		return 0;
437 	ops = master_dev->rtnl_link_ops;
438 	if (!ops || !ops->get_slave_size)
439 		return 0;
440 	/* IFLA_INFO_SLAVE_DATA + nested data */
441 	return nla_total_size(sizeof(struct nlattr)) +
442 	       ops->get_slave_size(master_dev, dev);
443 }
444 
445 static size_t rtnl_link_get_size(const struct net_device *dev)
446 {
447 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
448 	size_t size;
449 
450 	if (!ops)
451 		return 0;
452 
453 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
454 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
455 
456 	if (ops->get_size)
457 		/* IFLA_INFO_DATA + nested data */
458 		size += nla_total_size(sizeof(struct nlattr)) +
459 			ops->get_size(dev);
460 
461 	if (ops->get_xstats_size)
462 		/* IFLA_INFO_XSTATS */
463 		size += nla_total_size(ops->get_xstats_size(dev));
464 
465 	size += rtnl_link_get_slave_info_data_size(dev);
466 
467 	return size;
468 }
469 
470 static LIST_HEAD(rtnl_af_ops);
471 
472 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
473 {
474 	const struct rtnl_af_ops *ops;
475 
476 	list_for_each_entry(ops, &rtnl_af_ops, list) {
477 		if (ops->family == family)
478 			return ops;
479 	}
480 
481 	return NULL;
482 }
483 
484 /**
485  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
486  * @ops: struct rtnl_af_ops * to register
487  *
488  * Returns 0 on success or a negative error code.
489  */
490 void rtnl_af_register(struct rtnl_af_ops *ops)
491 {
492 	rtnl_lock();
493 	list_add_tail(&ops->list, &rtnl_af_ops);
494 	rtnl_unlock();
495 }
496 EXPORT_SYMBOL_GPL(rtnl_af_register);
497 
498 /**
499  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
500  * @ops: struct rtnl_af_ops * to unregister
501  *
502  * The caller must hold the rtnl_mutex.
503  */
504 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
505 {
506 	list_del(&ops->list);
507 }
508 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
509 
510 /**
511  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
512  * @ops: struct rtnl_af_ops * to unregister
513  */
514 void rtnl_af_unregister(struct rtnl_af_ops *ops)
515 {
516 	rtnl_lock();
517 	__rtnl_af_unregister(ops);
518 	rtnl_unlock();
519 }
520 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
521 
522 static size_t rtnl_link_get_af_size(const struct net_device *dev,
523 				    u32 ext_filter_mask)
524 {
525 	struct rtnl_af_ops *af_ops;
526 	size_t size;
527 
528 	/* IFLA_AF_SPEC */
529 	size = nla_total_size(sizeof(struct nlattr));
530 
531 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
532 		if (af_ops->get_link_af_size) {
533 			/* AF_* + nested data */
534 			size += nla_total_size(sizeof(struct nlattr)) +
535 				af_ops->get_link_af_size(dev, ext_filter_mask);
536 		}
537 	}
538 
539 	return size;
540 }
541 
542 static bool rtnl_have_link_slave_info(const struct net_device *dev)
543 {
544 	struct net_device *master_dev;
545 
546 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
547 	if (master_dev && master_dev->rtnl_link_ops)
548 		return true;
549 	return false;
550 }
551 
552 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
553 				     const struct net_device *dev)
554 {
555 	struct net_device *master_dev;
556 	const struct rtnl_link_ops *ops;
557 	struct nlattr *slave_data;
558 	int err;
559 
560 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
561 	if (!master_dev)
562 		return 0;
563 	ops = master_dev->rtnl_link_ops;
564 	if (!ops)
565 		return 0;
566 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
567 		return -EMSGSIZE;
568 	if (ops->fill_slave_info) {
569 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
570 		if (!slave_data)
571 			return -EMSGSIZE;
572 		err = ops->fill_slave_info(skb, master_dev, dev);
573 		if (err < 0)
574 			goto err_cancel_slave_data;
575 		nla_nest_end(skb, slave_data);
576 	}
577 	return 0;
578 
579 err_cancel_slave_data:
580 	nla_nest_cancel(skb, slave_data);
581 	return err;
582 }
583 
584 static int rtnl_link_info_fill(struct sk_buff *skb,
585 			       const struct net_device *dev)
586 {
587 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
588 	struct nlattr *data;
589 	int err;
590 
591 	if (!ops)
592 		return 0;
593 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
594 		return -EMSGSIZE;
595 	if (ops->fill_xstats) {
596 		err = ops->fill_xstats(skb, dev);
597 		if (err < 0)
598 			return err;
599 	}
600 	if (ops->fill_info) {
601 		data = nla_nest_start(skb, IFLA_INFO_DATA);
602 		if (data == NULL)
603 			return -EMSGSIZE;
604 		err = ops->fill_info(skb, dev);
605 		if (err < 0)
606 			goto err_cancel_data;
607 		nla_nest_end(skb, data);
608 	}
609 	return 0;
610 
611 err_cancel_data:
612 	nla_nest_cancel(skb, data);
613 	return err;
614 }
615 
616 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
617 {
618 	struct nlattr *linkinfo;
619 	int err = -EMSGSIZE;
620 
621 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
622 	if (linkinfo == NULL)
623 		goto out;
624 
625 	err = rtnl_link_info_fill(skb, dev);
626 	if (err < 0)
627 		goto err_cancel_link;
628 
629 	err = rtnl_link_slave_info_fill(skb, dev);
630 	if (err < 0)
631 		goto err_cancel_link;
632 
633 	nla_nest_end(skb, linkinfo);
634 	return 0;
635 
636 err_cancel_link:
637 	nla_nest_cancel(skb, linkinfo);
638 out:
639 	return err;
640 }
641 
642 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
643 {
644 	struct sock *rtnl = net->rtnl;
645 	int err = 0;
646 
647 	NETLINK_CB(skb).dst_group = group;
648 	if (echo)
649 		atomic_inc(&skb->users);
650 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
651 	if (echo)
652 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
653 	return err;
654 }
655 
656 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
657 {
658 	struct sock *rtnl = net->rtnl;
659 
660 	return nlmsg_unicast(rtnl, skb, pid);
661 }
662 EXPORT_SYMBOL(rtnl_unicast);
663 
664 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
665 		 struct nlmsghdr *nlh, gfp_t flags)
666 {
667 	struct sock *rtnl = net->rtnl;
668 	int report = 0;
669 
670 	if (nlh)
671 		report = nlmsg_report(nlh);
672 
673 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
674 }
675 EXPORT_SYMBOL(rtnl_notify);
676 
677 void rtnl_set_sk_err(struct net *net, u32 group, int error)
678 {
679 	struct sock *rtnl = net->rtnl;
680 
681 	netlink_set_err(rtnl, 0, group, error);
682 }
683 EXPORT_SYMBOL(rtnl_set_sk_err);
684 
685 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
686 {
687 	struct nlattr *mx;
688 	int i, valid = 0;
689 
690 	mx = nla_nest_start(skb, RTA_METRICS);
691 	if (mx == NULL)
692 		return -ENOBUFS;
693 
694 	for (i = 0; i < RTAX_MAX; i++) {
695 		if (metrics[i]) {
696 			if (i == RTAX_CC_ALGO - 1) {
697 				char tmp[TCP_CA_NAME_MAX], *name;
698 
699 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
700 				if (!name)
701 					continue;
702 				if (nla_put_string(skb, i + 1, name))
703 					goto nla_put_failure;
704 			} else if (i == RTAX_FEATURES - 1) {
705 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
706 
707 				if (!user_features)
708 					continue;
709 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
710 				if (nla_put_u32(skb, i + 1, user_features))
711 					goto nla_put_failure;
712 			} else {
713 				if (nla_put_u32(skb, i + 1, metrics[i]))
714 					goto nla_put_failure;
715 			}
716 			valid++;
717 		}
718 	}
719 
720 	if (!valid) {
721 		nla_nest_cancel(skb, mx);
722 		return 0;
723 	}
724 
725 	return nla_nest_end(skb, mx);
726 
727 nla_put_failure:
728 	nla_nest_cancel(skb, mx);
729 	return -EMSGSIZE;
730 }
731 EXPORT_SYMBOL(rtnetlink_put_metrics);
732 
733 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
734 		       long expires, u32 error)
735 {
736 	struct rta_cacheinfo ci = {
737 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
738 		.rta_used = dst->__use,
739 		.rta_clntref = atomic_read(&(dst->__refcnt)),
740 		.rta_error = error,
741 		.rta_id =  id,
742 	};
743 
744 	if (expires) {
745 		unsigned long clock;
746 
747 		clock = jiffies_to_clock_t(abs(expires));
748 		clock = min_t(unsigned long, clock, INT_MAX);
749 		ci.rta_expires = (expires > 0) ? clock : -clock;
750 	}
751 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
752 }
753 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
754 
755 static void set_operstate(struct net_device *dev, unsigned char transition)
756 {
757 	unsigned char operstate = dev->operstate;
758 
759 	switch (transition) {
760 	case IF_OPER_UP:
761 		if ((operstate == IF_OPER_DORMANT ||
762 		     operstate == IF_OPER_UNKNOWN) &&
763 		    !netif_dormant(dev))
764 			operstate = IF_OPER_UP;
765 		break;
766 
767 	case IF_OPER_DORMANT:
768 		if (operstate == IF_OPER_UP ||
769 		    operstate == IF_OPER_UNKNOWN)
770 			operstate = IF_OPER_DORMANT;
771 		break;
772 	}
773 
774 	if (dev->operstate != operstate) {
775 		write_lock_bh(&dev_base_lock);
776 		dev->operstate = operstate;
777 		write_unlock_bh(&dev_base_lock);
778 		netdev_state_change(dev);
779 	}
780 }
781 
782 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
783 {
784 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
785 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
786 }
787 
788 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
789 					   const struct ifinfomsg *ifm)
790 {
791 	unsigned int flags = ifm->ifi_flags;
792 
793 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
794 	if (ifm->ifi_change)
795 		flags = (flags & ifm->ifi_change) |
796 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
797 
798 	return flags;
799 }
800 
801 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
802 				 const struct rtnl_link_stats64 *b)
803 {
804 	a->rx_packets = b->rx_packets;
805 	a->tx_packets = b->tx_packets;
806 	a->rx_bytes = b->rx_bytes;
807 	a->tx_bytes = b->tx_bytes;
808 	a->rx_errors = b->rx_errors;
809 	a->tx_errors = b->tx_errors;
810 	a->rx_dropped = b->rx_dropped;
811 	a->tx_dropped = b->tx_dropped;
812 
813 	a->multicast = b->multicast;
814 	a->collisions = b->collisions;
815 
816 	a->rx_length_errors = b->rx_length_errors;
817 	a->rx_over_errors = b->rx_over_errors;
818 	a->rx_crc_errors = b->rx_crc_errors;
819 	a->rx_frame_errors = b->rx_frame_errors;
820 	a->rx_fifo_errors = b->rx_fifo_errors;
821 	a->rx_missed_errors = b->rx_missed_errors;
822 
823 	a->tx_aborted_errors = b->tx_aborted_errors;
824 	a->tx_carrier_errors = b->tx_carrier_errors;
825 	a->tx_fifo_errors = b->tx_fifo_errors;
826 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
827 	a->tx_window_errors = b->tx_window_errors;
828 
829 	a->rx_compressed = b->rx_compressed;
830 	a->tx_compressed = b->tx_compressed;
831 
832 	a->rx_nohandler = b->rx_nohandler;
833 }
834 
835 /* All VF info */
836 static inline int rtnl_vfinfo_size(const struct net_device *dev,
837 				   u32 ext_filter_mask)
838 {
839 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
840 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
841 		int num_vfs = dev_num_vf(dev->dev.parent);
842 		size_t size = nla_total_size(sizeof(struct nlattr));
843 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
844 		size += num_vfs *
845 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
846 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
847 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
848 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
849 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
850 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
851 			 /* IFLA_VF_STATS_RX_PACKETS */
852 			 nla_total_size_64bit(sizeof(__u64)) +
853 			 /* IFLA_VF_STATS_TX_PACKETS */
854 			 nla_total_size_64bit(sizeof(__u64)) +
855 			 /* IFLA_VF_STATS_RX_BYTES */
856 			 nla_total_size_64bit(sizeof(__u64)) +
857 			 /* IFLA_VF_STATS_TX_BYTES */
858 			 nla_total_size_64bit(sizeof(__u64)) +
859 			 /* IFLA_VF_STATS_BROADCAST */
860 			 nla_total_size_64bit(sizeof(__u64)) +
861 			 /* IFLA_VF_STATS_MULTICAST */
862 			 nla_total_size_64bit(sizeof(__u64)) +
863 			 nla_total_size(sizeof(struct ifla_vf_trust)));
864 		return size;
865 	} else
866 		return 0;
867 }
868 
869 static size_t rtnl_port_size(const struct net_device *dev,
870 			     u32 ext_filter_mask)
871 {
872 	size_t port_size = nla_total_size(4)		/* PORT_VF */
873 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
874 		+ nla_total_size(sizeof(struct ifla_port_vsi))
875 							/* PORT_VSI_TYPE */
876 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
877 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
878 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
879 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
880 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
881 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
882 		+ port_size;
883 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
884 		+ port_size;
885 
886 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
887 	    !(ext_filter_mask & RTEXT_FILTER_VF))
888 		return 0;
889 	if (dev_num_vf(dev->dev.parent))
890 		return port_self_size + vf_ports_size +
891 			vf_port_size * dev_num_vf(dev->dev.parent);
892 	else
893 		return port_self_size;
894 }
895 
896 static size_t rtnl_xdp_size(const struct net_device *dev)
897 {
898 	size_t xdp_size = nla_total_size(1);	/* XDP_ATTACHED */
899 
900 	if (!dev->netdev_ops->ndo_xdp)
901 		return 0;
902 	else
903 		return xdp_size;
904 }
905 
906 static noinline size_t if_nlmsg_size(const struct net_device *dev,
907 				     u32 ext_filter_mask)
908 {
909 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
910 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
911 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
912 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
913 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
914 	       + nla_total_size(sizeof(struct rtnl_link_stats))
915 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
916 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
917 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
918 	       + nla_total_size(4) /* IFLA_TXQLEN */
919 	       + nla_total_size(4) /* IFLA_WEIGHT */
920 	       + nla_total_size(4) /* IFLA_MTU */
921 	       + nla_total_size(4) /* IFLA_LINK */
922 	       + nla_total_size(4) /* IFLA_MASTER */
923 	       + nla_total_size(1) /* IFLA_CARRIER */
924 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
925 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
926 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
927 	       + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
928 	       + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
929 	       + nla_total_size(1) /* IFLA_OPERSTATE */
930 	       + nla_total_size(1) /* IFLA_LINKMODE */
931 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
932 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
933 	       + nla_total_size(ext_filter_mask
934 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
935 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
936 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
937 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
938 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
939 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
940 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
941 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
942 	       + rtnl_xdp_size(dev) /* IFLA_XDP */
943 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
944 
945 }
946 
947 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
948 {
949 	struct nlattr *vf_ports;
950 	struct nlattr *vf_port;
951 	int vf;
952 	int err;
953 
954 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
955 	if (!vf_ports)
956 		return -EMSGSIZE;
957 
958 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
959 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
960 		if (!vf_port)
961 			goto nla_put_failure;
962 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
963 			goto nla_put_failure;
964 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
965 		if (err == -EMSGSIZE)
966 			goto nla_put_failure;
967 		if (err) {
968 			nla_nest_cancel(skb, vf_port);
969 			continue;
970 		}
971 		nla_nest_end(skb, vf_port);
972 	}
973 
974 	nla_nest_end(skb, vf_ports);
975 
976 	return 0;
977 
978 nla_put_failure:
979 	nla_nest_cancel(skb, vf_ports);
980 	return -EMSGSIZE;
981 }
982 
983 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
984 {
985 	struct nlattr *port_self;
986 	int err;
987 
988 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
989 	if (!port_self)
990 		return -EMSGSIZE;
991 
992 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
993 	if (err) {
994 		nla_nest_cancel(skb, port_self);
995 		return (err == -EMSGSIZE) ? err : 0;
996 	}
997 
998 	nla_nest_end(skb, port_self);
999 
1000 	return 0;
1001 }
1002 
1003 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1004 			  u32 ext_filter_mask)
1005 {
1006 	int err;
1007 
1008 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1009 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1010 		return 0;
1011 
1012 	err = rtnl_port_self_fill(skb, dev);
1013 	if (err)
1014 		return err;
1015 
1016 	if (dev_num_vf(dev->dev.parent)) {
1017 		err = rtnl_vf_ports_fill(skb, dev);
1018 		if (err)
1019 			return err;
1020 	}
1021 
1022 	return 0;
1023 }
1024 
1025 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1026 {
1027 	int err;
1028 	struct netdev_phys_item_id ppid;
1029 
1030 	err = dev_get_phys_port_id(dev, &ppid);
1031 	if (err) {
1032 		if (err == -EOPNOTSUPP)
1033 			return 0;
1034 		return err;
1035 	}
1036 
1037 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1038 		return -EMSGSIZE;
1039 
1040 	return 0;
1041 }
1042 
1043 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1044 {
1045 	char name[IFNAMSIZ];
1046 	int err;
1047 
1048 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1049 	if (err) {
1050 		if (err == -EOPNOTSUPP)
1051 			return 0;
1052 		return err;
1053 	}
1054 
1055 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1056 		return -EMSGSIZE;
1057 
1058 	return 0;
1059 }
1060 
1061 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1062 {
1063 	int err;
1064 	struct switchdev_attr attr = {
1065 		.orig_dev = dev,
1066 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1067 		.flags = SWITCHDEV_F_NO_RECURSE,
1068 	};
1069 
1070 	err = switchdev_port_attr_get(dev, &attr);
1071 	if (err) {
1072 		if (err == -EOPNOTSUPP)
1073 			return 0;
1074 		return err;
1075 	}
1076 
1077 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1078 		    attr.u.ppid.id))
1079 		return -EMSGSIZE;
1080 
1081 	return 0;
1082 }
1083 
1084 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1085 					      struct net_device *dev)
1086 {
1087 	struct rtnl_link_stats64 *sp;
1088 	struct nlattr *attr;
1089 
1090 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1091 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1092 	if (!attr)
1093 		return -EMSGSIZE;
1094 
1095 	sp = nla_data(attr);
1096 	dev_get_stats(dev, sp);
1097 
1098 	attr = nla_reserve(skb, IFLA_STATS,
1099 			   sizeof(struct rtnl_link_stats));
1100 	if (!attr)
1101 		return -EMSGSIZE;
1102 
1103 	copy_rtnl_link_stats(nla_data(attr), sp);
1104 
1105 	return 0;
1106 }
1107 
1108 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1109 					       struct net_device *dev,
1110 					       int vfs_num,
1111 					       struct nlattr *vfinfo)
1112 {
1113 	struct ifla_vf_rss_query_en vf_rss_query_en;
1114 	struct ifla_vf_link_state vf_linkstate;
1115 	struct ifla_vf_spoofchk vf_spoofchk;
1116 	struct ifla_vf_tx_rate vf_tx_rate;
1117 	struct ifla_vf_stats vf_stats;
1118 	struct ifla_vf_trust vf_trust;
1119 	struct ifla_vf_vlan vf_vlan;
1120 	struct ifla_vf_rate vf_rate;
1121 	struct nlattr *vf, *vfstats;
1122 	struct ifla_vf_mac vf_mac;
1123 	struct ifla_vf_info ivi;
1124 
1125 	/* Not all SR-IOV capable drivers support the
1126 	 * spoofcheck and "RSS query enable" query.  Preset to
1127 	 * -1 so the user space tool can detect that the driver
1128 	 * didn't report anything.
1129 	 */
1130 	ivi.spoofchk = -1;
1131 	ivi.rss_query_en = -1;
1132 	ivi.trusted = -1;
1133 	memset(ivi.mac, 0, sizeof(ivi.mac));
1134 	/* The default value for VF link state is "auto"
1135 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1136 	 */
1137 	ivi.linkstate = 0;
1138 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1139 		return 0;
1140 
1141 	vf_mac.vf =
1142 		vf_vlan.vf =
1143 		vf_rate.vf =
1144 		vf_tx_rate.vf =
1145 		vf_spoofchk.vf =
1146 		vf_linkstate.vf =
1147 		vf_rss_query_en.vf =
1148 		vf_trust.vf = ivi.vf;
1149 
1150 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1151 	vf_vlan.vlan = ivi.vlan;
1152 	vf_vlan.qos = ivi.qos;
1153 	vf_tx_rate.rate = ivi.max_tx_rate;
1154 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1155 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1156 	vf_spoofchk.setting = ivi.spoofchk;
1157 	vf_linkstate.link_state = ivi.linkstate;
1158 	vf_rss_query_en.setting = ivi.rss_query_en;
1159 	vf_trust.setting = ivi.trusted;
1160 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1161 	if (!vf) {
1162 		nla_nest_cancel(skb, vfinfo);
1163 		return -EMSGSIZE;
1164 	}
1165 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1166 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1167 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1168 		    &vf_rate) ||
1169 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1170 		    &vf_tx_rate) ||
1171 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1172 		    &vf_spoofchk) ||
1173 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1174 		    &vf_linkstate) ||
1175 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1176 		    sizeof(vf_rss_query_en),
1177 		    &vf_rss_query_en) ||
1178 	    nla_put(skb, IFLA_VF_TRUST,
1179 		    sizeof(vf_trust), &vf_trust))
1180 		return -EMSGSIZE;
1181 	memset(&vf_stats, 0, sizeof(vf_stats));
1182 	if (dev->netdev_ops->ndo_get_vf_stats)
1183 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1184 						&vf_stats);
1185 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1186 	if (!vfstats) {
1187 		nla_nest_cancel(skb, vf);
1188 		nla_nest_cancel(skb, vfinfo);
1189 		return -EMSGSIZE;
1190 	}
1191 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1192 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1193 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1194 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1195 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1196 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1197 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1198 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1199 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1200 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1201 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1202 			      vf_stats.multicast, IFLA_VF_STATS_PAD))
1203 		return -EMSGSIZE;
1204 	nla_nest_end(skb, vfstats);
1205 	nla_nest_end(skb, vf);
1206 	return 0;
1207 }
1208 
1209 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1210 {
1211 	struct rtnl_link_ifmap map;
1212 
1213 	memset(&map, 0, sizeof(map));
1214 	map.mem_start   = dev->mem_start;
1215 	map.mem_end     = dev->mem_end;
1216 	map.base_addr   = dev->base_addr;
1217 	map.irq         = dev->irq;
1218 	map.dma         = dev->dma;
1219 	map.port        = dev->if_port;
1220 
1221 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1222 		return -EMSGSIZE;
1223 
1224 	return 0;
1225 }
1226 
1227 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1228 {
1229 	struct netdev_xdp xdp_op = {};
1230 	struct nlattr *xdp;
1231 	int err;
1232 
1233 	if (!dev->netdev_ops->ndo_xdp)
1234 		return 0;
1235 	xdp = nla_nest_start(skb, IFLA_XDP);
1236 	if (!xdp)
1237 		return -EMSGSIZE;
1238 	xdp_op.command = XDP_QUERY_PROG;
1239 	err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1240 	if (err)
1241 		goto err_cancel;
1242 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1243 	if (err)
1244 		goto err_cancel;
1245 
1246 	nla_nest_end(skb, xdp);
1247 	return 0;
1248 
1249 err_cancel:
1250 	nla_nest_cancel(skb, xdp);
1251 	return err;
1252 }
1253 
1254 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1255 			    int type, u32 pid, u32 seq, u32 change,
1256 			    unsigned int flags, u32 ext_filter_mask)
1257 {
1258 	struct ifinfomsg *ifm;
1259 	struct nlmsghdr *nlh;
1260 	struct nlattr *af_spec;
1261 	struct rtnl_af_ops *af_ops;
1262 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1263 
1264 	ASSERT_RTNL();
1265 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1266 	if (nlh == NULL)
1267 		return -EMSGSIZE;
1268 
1269 	ifm = nlmsg_data(nlh);
1270 	ifm->ifi_family = AF_UNSPEC;
1271 	ifm->__ifi_pad = 0;
1272 	ifm->ifi_type = dev->type;
1273 	ifm->ifi_index = dev->ifindex;
1274 	ifm->ifi_flags = dev_get_flags(dev);
1275 	ifm->ifi_change = change;
1276 
1277 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1278 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1279 	    nla_put_u8(skb, IFLA_OPERSTATE,
1280 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1281 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1282 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1283 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1284 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1285 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1286 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1287 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1288 #ifdef CONFIG_RPS
1289 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1290 #endif
1291 	    (dev->ifindex != dev_get_iflink(dev) &&
1292 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1293 	    (upper_dev &&
1294 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1295 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1296 	    (dev->qdisc &&
1297 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1298 	    (dev->ifalias &&
1299 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1300 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1301 			atomic_read(&dev->carrier_changes)) ||
1302 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1303 		goto nla_put_failure;
1304 
1305 	if (rtnl_fill_link_ifmap(skb, dev))
1306 		goto nla_put_failure;
1307 
1308 	if (dev->addr_len) {
1309 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1310 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1311 			goto nla_put_failure;
1312 	}
1313 
1314 	if (rtnl_phys_port_id_fill(skb, dev))
1315 		goto nla_put_failure;
1316 
1317 	if (rtnl_phys_port_name_fill(skb, dev))
1318 		goto nla_put_failure;
1319 
1320 	if (rtnl_phys_switch_id_fill(skb, dev))
1321 		goto nla_put_failure;
1322 
1323 	if (rtnl_fill_stats(skb, dev))
1324 		goto nla_put_failure;
1325 
1326 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1327 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1328 		goto nla_put_failure;
1329 
1330 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1331 	    ext_filter_mask & RTEXT_FILTER_VF) {
1332 		int i;
1333 		struct nlattr *vfinfo;
1334 		int num_vfs = dev_num_vf(dev->dev.parent);
1335 
1336 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1337 		if (!vfinfo)
1338 			goto nla_put_failure;
1339 		for (i = 0; i < num_vfs; i++) {
1340 			if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1341 				goto nla_put_failure;
1342 		}
1343 
1344 		nla_nest_end(skb, vfinfo);
1345 	}
1346 
1347 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1348 		goto nla_put_failure;
1349 
1350 	if (rtnl_xdp_fill(skb, dev))
1351 		goto nla_put_failure;
1352 
1353 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1354 		if (rtnl_link_fill(skb, dev) < 0)
1355 			goto nla_put_failure;
1356 	}
1357 
1358 	if (dev->rtnl_link_ops &&
1359 	    dev->rtnl_link_ops->get_link_net) {
1360 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1361 
1362 		if (!net_eq(dev_net(dev), link_net)) {
1363 			int id = peernet2id_alloc(dev_net(dev), link_net);
1364 
1365 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1366 				goto nla_put_failure;
1367 		}
1368 	}
1369 
1370 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1371 		goto nla_put_failure;
1372 
1373 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1374 		if (af_ops->fill_link_af) {
1375 			struct nlattr *af;
1376 			int err;
1377 
1378 			if (!(af = nla_nest_start(skb, af_ops->family)))
1379 				goto nla_put_failure;
1380 
1381 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1382 
1383 			/*
1384 			 * Caller may return ENODATA to indicate that there
1385 			 * was no data to be dumped. This is not an error, it
1386 			 * means we should trim the attribute header and
1387 			 * continue.
1388 			 */
1389 			if (err == -ENODATA)
1390 				nla_nest_cancel(skb, af);
1391 			else if (err < 0)
1392 				goto nla_put_failure;
1393 
1394 			nla_nest_end(skb, af);
1395 		}
1396 	}
1397 
1398 	nla_nest_end(skb, af_spec);
1399 
1400 	nlmsg_end(skb, nlh);
1401 	return 0;
1402 
1403 nla_put_failure:
1404 	nlmsg_cancel(skb, nlh);
1405 	return -EMSGSIZE;
1406 }
1407 
1408 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1409 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1410 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1411 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1412 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1413 	[IFLA_MTU]		= { .type = NLA_U32 },
1414 	[IFLA_LINK]		= { .type = NLA_U32 },
1415 	[IFLA_MASTER]		= { .type = NLA_U32 },
1416 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1417 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1418 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1419 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1420 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1421 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1422 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1423 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1424 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1425 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1426 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1427 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1428 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1429 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1430 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1431 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1432 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1433 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1434 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1435 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1436 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1437 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1438 	[IFLA_XDP]		= { .type = NLA_NESTED },
1439 };
1440 
1441 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1442 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1443 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1444 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1445 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1446 };
1447 
1448 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1449 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1450 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1451 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1452 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1453 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1454 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1455 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1456 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1457 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1458 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1459 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1460 };
1461 
1462 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1463 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1464 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1465 				    .len = PORT_PROFILE_MAX },
1466 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1467 				    .len = sizeof(struct ifla_port_vsi)},
1468 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1469 				      .len = PORT_UUID_MAX },
1470 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1471 				    .len = PORT_UUID_MAX },
1472 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1473 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1474 };
1475 
1476 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1477 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1478 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1479 };
1480 
1481 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1482 {
1483 	const struct rtnl_link_ops *ops = NULL;
1484 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1485 
1486 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1487 		return NULL;
1488 
1489 	if (linfo[IFLA_INFO_KIND]) {
1490 		char kind[MODULE_NAME_LEN];
1491 
1492 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1493 		ops = rtnl_link_ops_get(kind);
1494 	}
1495 
1496 	return ops;
1497 }
1498 
1499 static bool link_master_filtered(struct net_device *dev, int master_idx)
1500 {
1501 	struct net_device *master;
1502 
1503 	if (!master_idx)
1504 		return false;
1505 
1506 	master = netdev_master_upper_dev_get(dev);
1507 	if (!master || master->ifindex != master_idx)
1508 		return true;
1509 
1510 	return false;
1511 }
1512 
1513 static bool link_kind_filtered(const struct net_device *dev,
1514 			       const struct rtnl_link_ops *kind_ops)
1515 {
1516 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1517 		return true;
1518 
1519 	return false;
1520 }
1521 
1522 static bool link_dump_filtered(struct net_device *dev,
1523 			       int master_idx,
1524 			       const struct rtnl_link_ops *kind_ops)
1525 {
1526 	if (link_master_filtered(dev, master_idx) ||
1527 	    link_kind_filtered(dev, kind_ops))
1528 		return true;
1529 
1530 	return false;
1531 }
1532 
1533 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1534 {
1535 	struct net *net = sock_net(skb->sk);
1536 	int h, s_h;
1537 	int idx = 0, s_idx;
1538 	struct net_device *dev;
1539 	struct hlist_head *head;
1540 	struct nlattr *tb[IFLA_MAX+1];
1541 	u32 ext_filter_mask = 0;
1542 	const struct rtnl_link_ops *kind_ops = NULL;
1543 	unsigned int flags = NLM_F_MULTI;
1544 	int master_idx = 0;
1545 	int err;
1546 	int hdrlen;
1547 
1548 	s_h = cb->args[0];
1549 	s_idx = cb->args[1];
1550 
1551 	cb->seq = net->dev_base_seq;
1552 
1553 	/* A hack to preserve kernel<->userspace interface.
1554 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1555 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1556 	 * what iproute2 < v3.9.0 used.
1557 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1558 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1559 	 */
1560 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1561 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1562 
1563 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1564 
1565 		if (tb[IFLA_EXT_MASK])
1566 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1567 
1568 		if (tb[IFLA_MASTER])
1569 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1570 
1571 		if (tb[IFLA_LINKINFO])
1572 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1573 
1574 		if (master_idx || kind_ops)
1575 			flags |= NLM_F_DUMP_FILTERED;
1576 	}
1577 
1578 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1579 		idx = 0;
1580 		head = &net->dev_index_head[h];
1581 		hlist_for_each_entry(dev, head, index_hlist) {
1582 			if (link_dump_filtered(dev, master_idx, kind_ops))
1583 				continue;
1584 			if (idx < s_idx)
1585 				goto cont;
1586 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1587 					       NETLINK_CB(cb->skb).portid,
1588 					       cb->nlh->nlmsg_seq, 0,
1589 					       flags,
1590 					       ext_filter_mask);
1591 			/* If we ran out of room on the first message,
1592 			 * we're in trouble
1593 			 */
1594 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1595 
1596 			if (err < 0)
1597 				goto out;
1598 
1599 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1600 cont:
1601 			idx++;
1602 		}
1603 	}
1604 out:
1605 	cb->args[1] = idx;
1606 	cb->args[0] = h;
1607 
1608 	return skb->len;
1609 }
1610 
1611 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1612 {
1613 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1614 }
1615 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1616 
1617 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1618 {
1619 	struct net *net;
1620 	/* Examine the link attributes and figure out which
1621 	 * network namespace we are talking about.
1622 	 */
1623 	if (tb[IFLA_NET_NS_PID])
1624 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1625 	else if (tb[IFLA_NET_NS_FD])
1626 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1627 	else
1628 		net = get_net(src_net);
1629 	return net;
1630 }
1631 EXPORT_SYMBOL(rtnl_link_get_net);
1632 
1633 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1634 {
1635 	if (dev) {
1636 		if (tb[IFLA_ADDRESS] &&
1637 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1638 			return -EINVAL;
1639 
1640 		if (tb[IFLA_BROADCAST] &&
1641 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1642 			return -EINVAL;
1643 	}
1644 
1645 	if (tb[IFLA_AF_SPEC]) {
1646 		struct nlattr *af;
1647 		int rem, err;
1648 
1649 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1650 			const struct rtnl_af_ops *af_ops;
1651 
1652 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1653 				return -EAFNOSUPPORT;
1654 
1655 			if (!af_ops->set_link_af)
1656 				return -EOPNOTSUPP;
1657 
1658 			if (af_ops->validate_link_af) {
1659 				err = af_ops->validate_link_af(dev, af);
1660 				if (err < 0)
1661 					return err;
1662 			}
1663 		}
1664 	}
1665 
1666 	return 0;
1667 }
1668 
1669 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1670 				  int guid_type)
1671 {
1672 	const struct net_device_ops *ops = dev->netdev_ops;
1673 
1674 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1675 }
1676 
1677 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1678 {
1679 	if (dev->type != ARPHRD_INFINIBAND)
1680 		return -EOPNOTSUPP;
1681 
1682 	return handle_infiniband_guid(dev, ivt, guid_type);
1683 }
1684 
1685 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1686 {
1687 	const struct net_device_ops *ops = dev->netdev_ops;
1688 	int err = -EINVAL;
1689 
1690 	if (tb[IFLA_VF_MAC]) {
1691 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1692 
1693 		err = -EOPNOTSUPP;
1694 		if (ops->ndo_set_vf_mac)
1695 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1696 						  ivm->mac);
1697 		if (err < 0)
1698 			return err;
1699 	}
1700 
1701 	if (tb[IFLA_VF_VLAN]) {
1702 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1703 
1704 		err = -EOPNOTSUPP;
1705 		if (ops->ndo_set_vf_vlan)
1706 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1707 						   ivv->qos);
1708 		if (err < 0)
1709 			return err;
1710 	}
1711 
1712 	if (tb[IFLA_VF_TX_RATE]) {
1713 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1714 		struct ifla_vf_info ivf;
1715 
1716 		err = -EOPNOTSUPP;
1717 		if (ops->ndo_get_vf_config)
1718 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1719 		if (err < 0)
1720 			return err;
1721 
1722 		err = -EOPNOTSUPP;
1723 		if (ops->ndo_set_vf_rate)
1724 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1725 						   ivf.min_tx_rate,
1726 						   ivt->rate);
1727 		if (err < 0)
1728 			return err;
1729 	}
1730 
1731 	if (tb[IFLA_VF_RATE]) {
1732 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1733 
1734 		err = -EOPNOTSUPP;
1735 		if (ops->ndo_set_vf_rate)
1736 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1737 						   ivt->min_tx_rate,
1738 						   ivt->max_tx_rate);
1739 		if (err < 0)
1740 			return err;
1741 	}
1742 
1743 	if (tb[IFLA_VF_SPOOFCHK]) {
1744 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1745 
1746 		err = -EOPNOTSUPP;
1747 		if (ops->ndo_set_vf_spoofchk)
1748 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1749 						       ivs->setting);
1750 		if (err < 0)
1751 			return err;
1752 	}
1753 
1754 	if (tb[IFLA_VF_LINK_STATE]) {
1755 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1756 
1757 		err = -EOPNOTSUPP;
1758 		if (ops->ndo_set_vf_link_state)
1759 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1760 							 ivl->link_state);
1761 		if (err < 0)
1762 			return err;
1763 	}
1764 
1765 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1766 		struct ifla_vf_rss_query_en *ivrssq_en;
1767 
1768 		err = -EOPNOTSUPP;
1769 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1770 		if (ops->ndo_set_vf_rss_query_en)
1771 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1772 							   ivrssq_en->setting);
1773 		if (err < 0)
1774 			return err;
1775 	}
1776 
1777 	if (tb[IFLA_VF_TRUST]) {
1778 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1779 
1780 		err = -EOPNOTSUPP;
1781 		if (ops->ndo_set_vf_trust)
1782 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1783 		if (err < 0)
1784 			return err;
1785 	}
1786 
1787 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1788 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1789 
1790 		if (!ops->ndo_set_vf_guid)
1791 			return -EOPNOTSUPP;
1792 
1793 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1794 	}
1795 
1796 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1797 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1798 
1799 		if (!ops->ndo_set_vf_guid)
1800 			return -EOPNOTSUPP;
1801 
1802 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1803 	}
1804 
1805 	return err;
1806 }
1807 
1808 static int do_set_master(struct net_device *dev, int ifindex)
1809 {
1810 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1811 	const struct net_device_ops *ops;
1812 	int err;
1813 
1814 	if (upper_dev) {
1815 		if (upper_dev->ifindex == ifindex)
1816 			return 0;
1817 		ops = upper_dev->netdev_ops;
1818 		if (ops->ndo_del_slave) {
1819 			err = ops->ndo_del_slave(upper_dev, dev);
1820 			if (err)
1821 				return err;
1822 		} else {
1823 			return -EOPNOTSUPP;
1824 		}
1825 	}
1826 
1827 	if (ifindex) {
1828 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1829 		if (!upper_dev)
1830 			return -EINVAL;
1831 		ops = upper_dev->netdev_ops;
1832 		if (ops->ndo_add_slave) {
1833 			err = ops->ndo_add_slave(upper_dev, dev);
1834 			if (err)
1835 				return err;
1836 		} else {
1837 			return -EOPNOTSUPP;
1838 		}
1839 	}
1840 	return 0;
1841 }
1842 
1843 #define DO_SETLINK_MODIFIED	0x01
1844 /* notify flag means notify + modified. */
1845 #define DO_SETLINK_NOTIFY	0x03
1846 static int do_setlink(const struct sk_buff *skb,
1847 		      struct net_device *dev, struct ifinfomsg *ifm,
1848 		      struct nlattr **tb, char *ifname, int status)
1849 {
1850 	const struct net_device_ops *ops = dev->netdev_ops;
1851 	int err;
1852 
1853 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1854 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1855 		if (IS_ERR(net)) {
1856 			err = PTR_ERR(net);
1857 			goto errout;
1858 		}
1859 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1860 			put_net(net);
1861 			err = -EPERM;
1862 			goto errout;
1863 		}
1864 		err = dev_change_net_namespace(dev, net, ifname);
1865 		put_net(net);
1866 		if (err)
1867 			goto errout;
1868 		status |= DO_SETLINK_MODIFIED;
1869 	}
1870 
1871 	if (tb[IFLA_MAP]) {
1872 		struct rtnl_link_ifmap *u_map;
1873 		struct ifmap k_map;
1874 
1875 		if (!ops->ndo_set_config) {
1876 			err = -EOPNOTSUPP;
1877 			goto errout;
1878 		}
1879 
1880 		if (!netif_device_present(dev)) {
1881 			err = -ENODEV;
1882 			goto errout;
1883 		}
1884 
1885 		u_map = nla_data(tb[IFLA_MAP]);
1886 		k_map.mem_start = (unsigned long) u_map->mem_start;
1887 		k_map.mem_end = (unsigned long) u_map->mem_end;
1888 		k_map.base_addr = (unsigned short) u_map->base_addr;
1889 		k_map.irq = (unsigned char) u_map->irq;
1890 		k_map.dma = (unsigned char) u_map->dma;
1891 		k_map.port = (unsigned char) u_map->port;
1892 
1893 		err = ops->ndo_set_config(dev, &k_map);
1894 		if (err < 0)
1895 			goto errout;
1896 
1897 		status |= DO_SETLINK_NOTIFY;
1898 	}
1899 
1900 	if (tb[IFLA_ADDRESS]) {
1901 		struct sockaddr *sa;
1902 		int len;
1903 
1904 		len = sizeof(sa_family_t) + dev->addr_len;
1905 		sa = kmalloc(len, GFP_KERNEL);
1906 		if (!sa) {
1907 			err = -ENOMEM;
1908 			goto errout;
1909 		}
1910 		sa->sa_family = dev->type;
1911 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1912 		       dev->addr_len);
1913 		err = dev_set_mac_address(dev, sa);
1914 		kfree(sa);
1915 		if (err)
1916 			goto errout;
1917 		status |= DO_SETLINK_MODIFIED;
1918 	}
1919 
1920 	if (tb[IFLA_MTU]) {
1921 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1922 		if (err < 0)
1923 			goto errout;
1924 		status |= DO_SETLINK_MODIFIED;
1925 	}
1926 
1927 	if (tb[IFLA_GROUP]) {
1928 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1929 		status |= DO_SETLINK_NOTIFY;
1930 	}
1931 
1932 	/*
1933 	 * Interface selected by interface index but interface
1934 	 * name provided implies that a name change has been
1935 	 * requested.
1936 	 */
1937 	if (ifm->ifi_index > 0 && ifname[0]) {
1938 		err = dev_change_name(dev, ifname);
1939 		if (err < 0)
1940 			goto errout;
1941 		status |= DO_SETLINK_MODIFIED;
1942 	}
1943 
1944 	if (tb[IFLA_IFALIAS]) {
1945 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1946 				    nla_len(tb[IFLA_IFALIAS]));
1947 		if (err < 0)
1948 			goto errout;
1949 		status |= DO_SETLINK_NOTIFY;
1950 	}
1951 
1952 	if (tb[IFLA_BROADCAST]) {
1953 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1954 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1955 	}
1956 
1957 	if (ifm->ifi_flags || ifm->ifi_change) {
1958 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1959 		if (err < 0)
1960 			goto errout;
1961 	}
1962 
1963 	if (tb[IFLA_MASTER]) {
1964 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1965 		if (err)
1966 			goto errout;
1967 		status |= DO_SETLINK_MODIFIED;
1968 	}
1969 
1970 	if (tb[IFLA_CARRIER]) {
1971 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1972 		if (err)
1973 			goto errout;
1974 		status |= DO_SETLINK_MODIFIED;
1975 	}
1976 
1977 	if (tb[IFLA_TXQLEN]) {
1978 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1979 		unsigned long orig_len = dev->tx_queue_len;
1980 
1981 		if (dev->tx_queue_len ^ value) {
1982 			dev->tx_queue_len = value;
1983 			err = call_netdevice_notifiers(
1984 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
1985 			err = notifier_to_errno(err);
1986 			if (err) {
1987 				dev->tx_queue_len = orig_len;
1988 				goto errout;
1989 			}
1990 			status |= DO_SETLINK_NOTIFY;
1991 		}
1992 	}
1993 
1994 	if (tb[IFLA_OPERSTATE])
1995 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1996 
1997 	if (tb[IFLA_LINKMODE]) {
1998 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1999 
2000 		write_lock_bh(&dev_base_lock);
2001 		if (dev->link_mode ^ value)
2002 			status |= DO_SETLINK_NOTIFY;
2003 		dev->link_mode = value;
2004 		write_unlock_bh(&dev_base_lock);
2005 	}
2006 
2007 	if (tb[IFLA_VFINFO_LIST]) {
2008 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2009 		struct nlattr *attr;
2010 		int rem;
2011 
2012 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2013 			if (nla_type(attr) != IFLA_VF_INFO ||
2014 			    nla_len(attr) < NLA_HDRLEN) {
2015 				err = -EINVAL;
2016 				goto errout;
2017 			}
2018 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2019 					       ifla_vf_policy);
2020 			if (err < 0)
2021 				goto errout;
2022 			err = do_setvfinfo(dev, vfinfo);
2023 			if (err < 0)
2024 				goto errout;
2025 			status |= DO_SETLINK_NOTIFY;
2026 		}
2027 	}
2028 	err = 0;
2029 
2030 	if (tb[IFLA_VF_PORTS]) {
2031 		struct nlattr *port[IFLA_PORT_MAX+1];
2032 		struct nlattr *attr;
2033 		int vf;
2034 		int rem;
2035 
2036 		err = -EOPNOTSUPP;
2037 		if (!ops->ndo_set_vf_port)
2038 			goto errout;
2039 
2040 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2041 			if (nla_type(attr) != IFLA_VF_PORT ||
2042 			    nla_len(attr) < NLA_HDRLEN) {
2043 				err = -EINVAL;
2044 				goto errout;
2045 			}
2046 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2047 					       ifla_port_policy);
2048 			if (err < 0)
2049 				goto errout;
2050 			if (!port[IFLA_PORT_VF]) {
2051 				err = -EOPNOTSUPP;
2052 				goto errout;
2053 			}
2054 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2055 			err = ops->ndo_set_vf_port(dev, vf, port);
2056 			if (err < 0)
2057 				goto errout;
2058 			status |= DO_SETLINK_NOTIFY;
2059 		}
2060 	}
2061 	err = 0;
2062 
2063 	if (tb[IFLA_PORT_SELF]) {
2064 		struct nlattr *port[IFLA_PORT_MAX+1];
2065 
2066 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2067 			tb[IFLA_PORT_SELF], ifla_port_policy);
2068 		if (err < 0)
2069 			goto errout;
2070 
2071 		err = -EOPNOTSUPP;
2072 		if (ops->ndo_set_vf_port)
2073 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2074 		if (err < 0)
2075 			goto errout;
2076 		status |= DO_SETLINK_NOTIFY;
2077 	}
2078 
2079 	if (tb[IFLA_AF_SPEC]) {
2080 		struct nlattr *af;
2081 		int rem;
2082 
2083 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2084 			const struct rtnl_af_ops *af_ops;
2085 
2086 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2087 				BUG();
2088 
2089 			err = af_ops->set_link_af(dev, af);
2090 			if (err < 0)
2091 				goto errout;
2092 
2093 			status |= DO_SETLINK_NOTIFY;
2094 		}
2095 	}
2096 	err = 0;
2097 
2098 	if (tb[IFLA_PROTO_DOWN]) {
2099 		err = dev_change_proto_down(dev,
2100 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2101 		if (err)
2102 			goto errout;
2103 		status |= DO_SETLINK_NOTIFY;
2104 	}
2105 
2106 	if (tb[IFLA_XDP]) {
2107 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2108 
2109 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2110 				       ifla_xdp_policy);
2111 		if (err < 0)
2112 			goto errout;
2113 
2114 		if (xdp[IFLA_XDP_ATTACHED]) {
2115 			err = -EINVAL;
2116 			goto errout;
2117 		}
2118 		if (xdp[IFLA_XDP_FD]) {
2119 			err = dev_change_xdp_fd(dev,
2120 						nla_get_s32(xdp[IFLA_XDP_FD]));
2121 			if (err)
2122 				goto errout;
2123 			status |= DO_SETLINK_NOTIFY;
2124 		}
2125 	}
2126 
2127 errout:
2128 	if (status & DO_SETLINK_MODIFIED) {
2129 		if (status & DO_SETLINK_NOTIFY)
2130 			netdev_state_change(dev);
2131 
2132 		if (err < 0)
2133 			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",
2134 					     dev->name);
2135 	}
2136 
2137 	return err;
2138 }
2139 
2140 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2141 {
2142 	struct net *net = sock_net(skb->sk);
2143 	struct ifinfomsg *ifm;
2144 	struct net_device *dev;
2145 	int err;
2146 	struct nlattr *tb[IFLA_MAX+1];
2147 	char ifname[IFNAMSIZ];
2148 
2149 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2150 	if (err < 0)
2151 		goto errout;
2152 
2153 	if (tb[IFLA_IFNAME])
2154 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2155 	else
2156 		ifname[0] = '\0';
2157 
2158 	err = -EINVAL;
2159 	ifm = nlmsg_data(nlh);
2160 	if (ifm->ifi_index > 0)
2161 		dev = __dev_get_by_index(net, ifm->ifi_index);
2162 	else if (tb[IFLA_IFNAME])
2163 		dev = __dev_get_by_name(net, ifname);
2164 	else
2165 		goto errout;
2166 
2167 	if (dev == NULL) {
2168 		err = -ENODEV;
2169 		goto errout;
2170 	}
2171 
2172 	err = validate_linkmsg(dev, tb);
2173 	if (err < 0)
2174 		goto errout;
2175 
2176 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2177 errout:
2178 	return err;
2179 }
2180 
2181 static int rtnl_group_dellink(const struct net *net, int group)
2182 {
2183 	struct net_device *dev, *aux;
2184 	LIST_HEAD(list_kill);
2185 	bool found = false;
2186 
2187 	if (!group)
2188 		return -EPERM;
2189 
2190 	for_each_netdev(net, dev) {
2191 		if (dev->group == group) {
2192 			const struct rtnl_link_ops *ops;
2193 
2194 			found = true;
2195 			ops = dev->rtnl_link_ops;
2196 			if (!ops || !ops->dellink)
2197 				return -EOPNOTSUPP;
2198 		}
2199 	}
2200 
2201 	if (!found)
2202 		return -ENODEV;
2203 
2204 	for_each_netdev_safe(net, dev, aux) {
2205 		if (dev->group == group) {
2206 			const struct rtnl_link_ops *ops;
2207 
2208 			ops = dev->rtnl_link_ops;
2209 			ops->dellink(dev, &list_kill);
2210 		}
2211 	}
2212 	unregister_netdevice_many(&list_kill);
2213 
2214 	return 0;
2215 }
2216 
2217 int rtnl_delete_link(struct net_device *dev)
2218 {
2219 	const struct rtnl_link_ops *ops;
2220 	LIST_HEAD(list_kill);
2221 
2222 	ops = dev->rtnl_link_ops;
2223 	if (!ops || !ops->dellink)
2224 		return -EOPNOTSUPP;
2225 
2226 	ops->dellink(dev, &list_kill);
2227 	unregister_netdevice_many(&list_kill);
2228 
2229 	return 0;
2230 }
2231 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2232 
2233 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2234 {
2235 	struct net *net = sock_net(skb->sk);
2236 	struct net_device *dev;
2237 	struct ifinfomsg *ifm;
2238 	char ifname[IFNAMSIZ];
2239 	struct nlattr *tb[IFLA_MAX+1];
2240 	int err;
2241 
2242 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2243 	if (err < 0)
2244 		return err;
2245 
2246 	if (tb[IFLA_IFNAME])
2247 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2248 
2249 	ifm = nlmsg_data(nlh);
2250 	if (ifm->ifi_index > 0)
2251 		dev = __dev_get_by_index(net, ifm->ifi_index);
2252 	else if (tb[IFLA_IFNAME])
2253 		dev = __dev_get_by_name(net, ifname);
2254 	else if (tb[IFLA_GROUP])
2255 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2256 	else
2257 		return -EINVAL;
2258 
2259 	if (!dev)
2260 		return -ENODEV;
2261 
2262 	return rtnl_delete_link(dev);
2263 }
2264 
2265 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2266 {
2267 	unsigned int old_flags;
2268 	int err;
2269 
2270 	old_flags = dev->flags;
2271 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2272 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2273 		if (err < 0)
2274 			return err;
2275 	}
2276 
2277 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2278 
2279 	__dev_notify_flags(dev, old_flags, ~0U);
2280 	return 0;
2281 }
2282 EXPORT_SYMBOL(rtnl_configure_link);
2283 
2284 struct net_device *rtnl_create_link(struct net *net,
2285 	const char *ifname, unsigned char name_assign_type,
2286 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2287 {
2288 	int err;
2289 	struct net_device *dev;
2290 	unsigned int num_tx_queues = 1;
2291 	unsigned int num_rx_queues = 1;
2292 
2293 	if (tb[IFLA_NUM_TX_QUEUES])
2294 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2295 	else if (ops->get_num_tx_queues)
2296 		num_tx_queues = ops->get_num_tx_queues();
2297 
2298 	if (tb[IFLA_NUM_RX_QUEUES])
2299 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2300 	else if (ops->get_num_rx_queues)
2301 		num_rx_queues = ops->get_num_rx_queues();
2302 
2303 	err = -ENOMEM;
2304 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2305 			       ops->setup, num_tx_queues, num_rx_queues);
2306 	if (!dev)
2307 		goto err;
2308 
2309 	dev_net_set(dev, net);
2310 	dev->rtnl_link_ops = ops;
2311 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2312 
2313 	if (tb[IFLA_MTU])
2314 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2315 	if (tb[IFLA_ADDRESS]) {
2316 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2317 				nla_len(tb[IFLA_ADDRESS]));
2318 		dev->addr_assign_type = NET_ADDR_SET;
2319 	}
2320 	if (tb[IFLA_BROADCAST])
2321 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2322 				nla_len(tb[IFLA_BROADCAST]));
2323 	if (tb[IFLA_TXQLEN])
2324 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2325 	if (tb[IFLA_OPERSTATE])
2326 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2327 	if (tb[IFLA_LINKMODE])
2328 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2329 	if (tb[IFLA_GROUP])
2330 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2331 
2332 	return dev;
2333 
2334 err:
2335 	return ERR_PTR(err);
2336 }
2337 EXPORT_SYMBOL(rtnl_create_link);
2338 
2339 static int rtnl_group_changelink(const struct sk_buff *skb,
2340 		struct net *net, int group,
2341 		struct ifinfomsg *ifm,
2342 		struct nlattr **tb)
2343 {
2344 	struct net_device *dev, *aux;
2345 	int err;
2346 
2347 	for_each_netdev_safe(net, dev, aux) {
2348 		if (dev->group == group) {
2349 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2350 			if (err < 0)
2351 				return err;
2352 		}
2353 	}
2354 
2355 	return 0;
2356 }
2357 
2358 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2359 {
2360 	struct net *net = sock_net(skb->sk);
2361 	const struct rtnl_link_ops *ops;
2362 	const struct rtnl_link_ops *m_ops = NULL;
2363 	struct net_device *dev;
2364 	struct net_device *master_dev = NULL;
2365 	struct ifinfomsg *ifm;
2366 	char kind[MODULE_NAME_LEN];
2367 	char ifname[IFNAMSIZ];
2368 	struct nlattr *tb[IFLA_MAX+1];
2369 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2370 	unsigned char name_assign_type = NET_NAME_USER;
2371 	int err;
2372 
2373 #ifdef CONFIG_MODULES
2374 replay:
2375 #endif
2376 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2377 	if (err < 0)
2378 		return err;
2379 
2380 	if (tb[IFLA_IFNAME])
2381 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2382 	else
2383 		ifname[0] = '\0';
2384 
2385 	ifm = nlmsg_data(nlh);
2386 	if (ifm->ifi_index > 0)
2387 		dev = __dev_get_by_index(net, ifm->ifi_index);
2388 	else {
2389 		if (ifname[0])
2390 			dev = __dev_get_by_name(net, ifname);
2391 		else
2392 			dev = NULL;
2393 	}
2394 
2395 	if (dev) {
2396 		master_dev = netdev_master_upper_dev_get(dev);
2397 		if (master_dev)
2398 			m_ops = master_dev->rtnl_link_ops;
2399 	}
2400 
2401 	err = validate_linkmsg(dev, tb);
2402 	if (err < 0)
2403 		return err;
2404 
2405 	if (tb[IFLA_LINKINFO]) {
2406 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2407 				       tb[IFLA_LINKINFO], ifla_info_policy);
2408 		if (err < 0)
2409 			return err;
2410 	} else
2411 		memset(linkinfo, 0, sizeof(linkinfo));
2412 
2413 	if (linkinfo[IFLA_INFO_KIND]) {
2414 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2415 		ops = rtnl_link_ops_get(kind);
2416 	} else {
2417 		kind[0] = '\0';
2418 		ops = NULL;
2419 	}
2420 
2421 	if (1) {
2422 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2423 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2424 		struct nlattr **data = NULL;
2425 		struct nlattr **slave_data = NULL;
2426 		struct net *dest_net, *link_net = NULL;
2427 
2428 		if (ops) {
2429 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2430 				err = nla_parse_nested(attr, ops->maxtype,
2431 						       linkinfo[IFLA_INFO_DATA],
2432 						       ops->policy);
2433 				if (err < 0)
2434 					return err;
2435 				data = attr;
2436 			}
2437 			if (ops->validate) {
2438 				err = ops->validate(tb, data);
2439 				if (err < 0)
2440 					return err;
2441 			}
2442 		}
2443 
2444 		if (m_ops) {
2445 			if (m_ops->slave_maxtype &&
2446 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2447 				err = nla_parse_nested(slave_attr,
2448 						       m_ops->slave_maxtype,
2449 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2450 						       m_ops->slave_policy);
2451 				if (err < 0)
2452 					return err;
2453 				slave_data = slave_attr;
2454 			}
2455 			if (m_ops->slave_validate) {
2456 				err = m_ops->slave_validate(tb, slave_data);
2457 				if (err < 0)
2458 					return err;
2459 			}
2460 		}
2461 
2462 		if (dev) {
2463 			int status = 0;
2464 
2465 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2466 				return -EEXIST;
2467 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2468 				return -EOPNOTSUPP;
2469 
2470 			if (linkinfo[IFLA_INFO_DATA]) {
2471 				if (!ops || ops != dev->rtnl_link_ops ||
2472 				    !ops->changelink)
2473 					return -EOPNOTSUPP;
2474 
2475 				err = ops->changelink(dev, tb, data);
2476 				if (err < 0)
2477 					return err;
2478 				status |= DO_SETLINK_NOTIFY;
2479 			}
2480 
2481 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2482 				if (!m_ops || !m_ops->slave_changelink)
2483 					return -EOPNOTSUPP;
2484 
2485 				err = m_ops->slave_changelink(master_dev, dev,
2486 							      tb, slave_data);
2487 				if (err < 0)
2488 					return err;
2489 				status |= DO_SETLINK_NOTIFY;
2490 			}
2491 
2492 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2493 		}
2494 
2495 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2496 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2497 				return rtnl_group_changelink(skb, net,
2498 						nla_get_u32(tb[IFLA_GROUP]),
2499 						ifm, tb);
2500 			return -ENODEV;
2501 		}
2502 
2503 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2504 			return -EOPNOTSUPP;
2505 
2506 		if (!ops) {
2507 #ifdef CONFIG_MODULES
2508 			if (kind[0]) {
2509 				__rtnl_unlock();
2510 				request_module("rtnl-link-%s", kind);
2511 				rtnl_lock();
2512 				ops = rtnl_link_ops_get(kind);
2513 				if (ops)
2514 					goto replay;
2515 			}
2516 #endif
2517 			return -EOPNOTSUPP;
2518 		}
2519 
2520 		if (!ops->setup)
2521 			return -EOPNOTSUPP;
2522 
2523 		if (!ifname[0]) {
2524 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2525 			name_assign_type = NET_NAME_ENUM;
2526 		}
2527 
2528 		dest_net = rtnl_link_get_net(net, tb);
2529 		if (IS_ERR(dest_net))
2530 			return PTR_ERR(dest_net);
2531 
2532 		err = -EPERM;
2533 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2534 			goto out;
2535 
2536 		if (tb[IFLA_LINK_NETNSID]) {
2537 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2538 
2539 			link_net = get_net_ns_by_id(dest_net, id);
2540 			if (!link_net) {
2541 				err =  -EINVAL;
2542 				goto out;
2543 			}
2544 			err = -EPERM;
2545 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2546 				goto out;
2547 		}
2548 
2549 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2550 				       name_assign_type, ops, tb);
2551 		if (IS_ERR(dev)) {
2552 			err = PTR_ERR(dev);
2553 			goto out;
2554 		}
2555 
2556 		dev->ifindex = ifm->ifi_index;
2557 
2558 		if (ops->newlink) {
2559 			err = ops->newlink(link_net ? : net, dev, tb, data);
2560 			/* Drivers should call free_netdev() in ->destructor
2561 			 * and unregister it on failure after registration
2562 			 * so that device could be finally freed in rtnl_unlock.
2563 			 */
2564 			if (err < 0) {
2565 				/* If device is not registered at all, free it now */
2566 				if (dev->reg_state == NETREG_UNINITIALIZED)
2567 					free_netdev(dev);
2568 				goto out;
2569 			}
2570 		} else {
2571 			err = register_netdevice(dev);
2572 			if (err < 0) {
2573 				free_netdev(dev);
2574 				goto out;
2575 			}
2576 		}
2577 		err = rtnl_configure_link(dev, ifm);
2578 		if (err < 0)
2579 			goto out_unregister;
2580 		if (link_net) {
2581 			err = dev_change_net_namespace(dev, dest_net, ifname);
2582 			if (err < 0)
2583 				goto out_unregister;
2584 		}
2585 out:
2586 		if (link_net)
2587 			put_net(link_net);
2588 		put_net(dest_net);
2589 		return err;
2590 out_unregister:
2591 		if (ops->newlink) {
2592 			LIST_HEAD(list_kill);
2593 
2594 			ops->dellink(dev, &list_kill);
2595 			unregister_netdevice_many(&list_kill);
2596 		} else {
2597 			unregister_netdevice(dev);
2598 		}
2599 		goto out;
2600 	}
2601 }
2602 
2603 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2604 {
2605 	struct net *net = sock_net(skb->sk);
2606 	struct ifinfomsg *ifm;
2607 	char ifname[IFNAMSIZ];
2608 	struct nlattr *tb[IFLA_MAX+1];
2609 	struct net_device *dev = NULL;
2610 	struct sk_buff *nskb;
2611 	int err;
2612 	u32 ext_filter_mask = 0;
2613 
2614 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2615 	if (err < 0)
2616 		return err;
2617 
2618 	if (tb[IFLA_IFNAME])
2619 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2620 
2621 	if (tb[IFLA_EXT_MASK])
2622 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2623 
2624 	ifm = nlmsg_data(nlh);
2625 	if (ifm->ifi_index > 0)
2626 		dev = __dev_get_by_index(net, ifm->ifi_index);
2627 	else if (tb[IFLA_IFNAME])
2628 		dev = __dev_get_by_name(net, ifname);
2629 	else
2630 		return -EINVAL;
2631 
2632 	if (dev == NULL)
2633 		return -ENODEV;
2634 
2635 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2636 	if (nskb == NULL)
2637 		return -ENOBUFS;
2638 
2639 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2640 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2641 	if (err < 0) {
2642 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2643 		WARN_ON(err == -EMSGSIZE);
2644 		kfree_skb(nskb);
2645 	} else
2646 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2647 
2648 	return err;
2649 }
2650 
2651 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2652 {
2653 	struct net *net = sock_net(skb->sk);
2654 	struct net_device *dev;
2655 	struct nlattr *tb[IFLA_MAX+1];
2656 	u32 ext_filter_mask = 0;
2657 	u16 min_ifinfo_dump_size = 0;
2658 	int hdrlen;
2659 
2660 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2661 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2662 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2663 
2664 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2665 		if (tb[IFLA_EXT_MASK])
2666 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2667 	}
2668 
2669 	if (!ext_filter_mask)
2670 		return NLMSG_GOODSIZE;
2671 	/*
2672 	 * traverse the list of net devices and compute the minimum
2673 	 * buffer size based upon the filter mask.
2674 	 */
2675 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2676 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2677 					     if_nlmsg_size(dev,
2678 						           ext_filter_mask));
2679 	}
2680 
2681 	return min_ifinfo_dump_size;
2682 }
2683 
2684 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2685 {
2686 	int idx;
2687 	int s_idx = cb->family;
2688 
2689 	if (s_idx == 0)
2690 		s_idx = 1;
2691 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2692 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2693 		if (idx < s_idx || idx == PF_PACKET)
2694 			continue;
2695 		if (rtnl_msg_handlers[idx] == NULL ||
2696 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2697 			continue;
2698 		if (idx > s_idx) {
2699 			memset(&cb->args[0], 0, sizeof(cb->args));
2700 			cb->prev_seq = 0;
2701 			cb->seq = 0;
2702 		}
2703 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2704 			break;
2705 	}
2706 	cb->family = idx;
2707 
2708 	return skb->len;
2709 }
2710 
2711 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2712 				       unsigned int change, gfp_t flags)
2713 {
2714 	struct net *net = dev_net(dev);
2715 	struct sk_buff *skb;
2716 	int err = -ENOBUFS;
2717 	size_t if_info_size;
2718 
2719 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2720 	if (skb == NULL)
2721 		goto errout;
2722 
2723 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2724 	if (err < 0) {
2725 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2726 		WARN_ON(err == -EMSGSIZE);
2727 		kfree_skb(skb);
2728 		goto errout;
2729 	}
2730 	return skb;
2731 errout:
2732 	if (err < 0)
2733 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2734 	return NULL;
2735 }
2736 
2737 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2738 {
2739 	struct net *net = dev_net(dev);
2740 
2741 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2742 }
2743 
2744 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2745 		  gfp_t flags)
2746 {
2747 	struct sk_buff *skb;
2748 
2749 	if (dev->reg_state != NETREG_REGISTERED)
2750 		return;
2751 
2752 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2753 	if (skb)
2754 		rtmsg_ifinfo_send(skb, dev, flags);
2755 }
2756 EXPORT_SYMBOL(rtmsg_ifinfo);
2757 
2758 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2759 				   struct net_device *dev,
2760 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2761 				   int type, unsigned int flags,
2762 				   int nlflags, u16 ndm_state)
2763 {
2764 	struct nlmsghdr *nlh;
2765 	struct ndmsg *ndm;
2766 
2767 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2768 	if (!nlh)
2769 		return -EMSGSIZE;
2770 
2771 	ndm = nlmsg_data(nlh);
2772 	ndm->ndm_family  = AF_BRIDGE;
2773 	ndm->ndm_pad1	 = 0;
2774 	ndm->ndm_pad2    = 0;
2775 	ndm->ndm_flags	 = flags;
2776 	ndm->ndm_type	 = 0;
2777 	ndm->ndm_ifindex = dev->ifindex;
2778 	ndm->ndm_state   = ndm_state;
2779 
2780 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2781 		goto nla_put_failure;
2782 	if (vid)
2783 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2784 			goto nla_put_failure;
2785 
2786 	nlmsg_end(skb, nlh);
2787 	return 0;
2788 
2789 nla_put_failure:
2790 	nlmsg_cancel(skb, nlh);
2791 	return -EMSGSIZE;
2792 }
2793 
2794 static inline size_t rtnl_fdb_nlmsg_size(void)
2795 {
2796 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2797 }
2798 
2799 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2800 			    u16 ndm_state)
2801 {
2802 	struct net *net = dev_net(dev);
2803 	struct sk_buff *skb;
2804 	int err = -ENOBUFS;
2805 
2806 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2807 	if (!skb)
2808 		goto errout;
2809 
2810 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2811 				      0, 0, type, NTF_SELF, 0, ndm_state);
2812 	if (err < 0) {
2813 		kfree_skb(skb);
2814 		goto errout;
2815 	}
2816 
2817 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2818 	return;
2819 errout:
2820 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2821 }
2822 
2823 /**
2824  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2825  */
2826 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2827 		     struct nlattr *tb[],
2828 		     struct net_device *dev,
2829 		     const unsigned char *addr, u16 vid,
2830 		     u16 flags)
2831 {
2832 	int err = -EINVAL;
2833 
2834 	/* If aging addresses are supported device will need to
2835 	 * implement its own handler for this.
2836 	 */
2837 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2838 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2839 		return err;
2840 	}
2841 
2842 	if (vid) {
2843 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2844 		return err;
2845 	}
2846 
2847 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2848 		err = dev_uc_add_excl(dev, addr);
2849 	else if (is_multicast_ether_addr(addr))
2850 		err = dev_mc_add_excl(dev, addr);
2851 
2852 	/* Only return duplicate errors if NLM_F_EXCL is set */
2853 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2854 		err = 0;
2855 
2856 	return err;
2857 }
2858 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2859 
2860 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2861 {
2862 	u16 vid = 0;
2863 
2864 	if (vlan_attr) {
2865 		if (nla_len(vlan_attr) != sizeof(u16)) {
2866 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2867 			return -EINVAL;
2868 		}
2869 
2870 		vid = nla_get_u16(vlan_attr);
2871 
2872 		if (!vid || vid >= VLAN_VID_MASK) {
2873 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2874 				vid);
2875 			return -EINVAL;
2876 		}
2877 	}
2878 	*p_vid = vid;
2879 	return 0;
2880 }
2881 
2882 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2883 {
2884 	struct net *net = sock_net(skb->sk);
2885 	struct ndmsg *ndm;
2886 	struct nlattr *tb[NDA_MAX+1];
2887 	struct net_device *dev;
2888 	u8 *addr;
2889 	u16 vid;
2890 	int err;
2891 
2892 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2893 	if (err < 0)
2894 		return err;
2895 
2896 	ndm = nlmsg_data(nlh);
2897 	if (ndm->ndm_ifindex == 0) {
2898 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2899 		return -EINVAL;
2900 	}
2901 
2902 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2903 	if (dev == NULL) {
2904 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2905 		return -ENODEV;
2906 	}
2907 
2908 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2909 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2910 		return -EINVAL;
2911 	}
2912 
2913 	addr = nla_data(tb[NDA_LLADDR]);
2914 
2915 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2916 	if (err)
2917 		return err;
2918 
2919 	err = -EOPNOTSUPP;
2920 
2921 	/* Support fdb on master device the net/bridge default case */
2922 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2923 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2924 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2925 		const struct net_device_ops *ops = br_dev->netdev_ops;
2926 
2927 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2928 				       nlh->nlmsg_flags);
2929 		if (err)
2930 			goto out;
2931 		else
2932 			ndm->ndm_flags &= ~NTF_MASTER;
2933 	}
2934 
2935 	/* Embedded bridge, macvlan, and any other device support */
2936 	if ((ndm->ndm_flags & NTF_SELF)) {
2937 		if (dev->netdev_ops->ndo_fdb_add)
2938 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2939 							   vid,
2940 							   nlh->nlmsg_flags);
2941 		else
2942 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2943 					       nlh->nlmsg_flags);
2944 
2945 		if (!err) {
2946 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2947 					ndm->ndm_state);
2948 			ndm->ndm_flags &= ~NTF_SELF;
2949 		}
2950 	}
2951 out:
2952 	return err;
2953 }
2954 
2955 /**
2956  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2957  */
2958 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2959 		     struct nlattr *tb[],
2960 		     struct net_device *dev,
2961 		     const unsigned char *addr, u16 vid)
2962 {
2963 	int err = -EINVAL;
2964 
2965 	/* If aging addresses are supported device will need to
2966 	 * implement its own handler for this.
2967 	 */
2968 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2969 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2970 		return err;
2971 	}
2972 
2973 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2974 		err = dev_uc_del(dev, addr);
2975 	else if (is_multicast_ether_addr(addr))
2976 		err = dev_mc_del(dev, addr);
2977 
2978 	return err;
2979 }
2980 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2981 
2982 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2983 {
2984 	struct net *net = sock_net(skb->sk);
2985 	struct ndmsg *ndm;
2986 	struct nlattr *tb[NDA_MAX+1];
2987 	struct net_device *dev;
2988 	int err = -EINVAL;
2989 	__u8 *addr;
2990 	u16 vid;
2991 
2992 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2993 		return -EPERM;
2994 
2995 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2996 	if (err < 0)
2997 		return err;
2998 
2999 	ndm = nlmsg_data(nlh);
3000 	if (ndm->ndm_ifindex == 0) {
3001 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3002 		return -EINVAL;
3003 	}
3004 
3005 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3006 	if (dev == NULL) {
3007 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3008 		return -ENODEV;
3009 	}
3010 
3011 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3012 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3013 		return -EINVAL;
3014 	}
3015 
3016 	addr = nla_data(tb[NDA_LLADDR]);
3017 
3018 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3019 	if (err)
3020 		return err;
3021 
3022 	err = -EOPNOTSUPP;
3023 
3024 	/* Support fdb on master device the net/bridge default case */
3025 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3026 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3027 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3028 		const struct net_device_ops *ops = br_dev->netdev_ops;
3029 
3030 		if (ops->ndo_fdb_del)
3031 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3032 
3033 		if (err)
3034 			goto out;
3035 		else
3036 			ndm->ndm_flags &= ~NTF_MASTER;
3037 	}
3038 
3039 	/* Embedded bridge, macvlan, and any other device support */
3040 	if (ndm->ndm_flags & NTF_SELF) {
3041 		if (dev->netdev_ops->ndo_fdb_del)
3042 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3043 							   vid);
3044 		else
3045 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3046 
3047 		if (!err) {
3048 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3049 					ndm->ndm_state);
3050 			ndm->ndm_flags &= ~NTF_SELF;
3051 		}
3052 	}
3053 out:
3054 	return err;
3055 }
3056 
3057 static int nlmsg_populate_fdb(struct sk_buff *skb,
3058 			      struct netlink_callback *cb,
3059 			      struct net_device *dev,
3060 			      int *idx,
3061 			      struct netdev_hw_addr_list *list)
3062 {
3063 	struct netdev_hw_addr *ha;
3064 	int err;
3065 	u32 portid, seq;
3066 
3067 	portid = NETLINK_CB(cb->skb).portid;
3068 	seq = cb->nlh->nlmsg_seq;
3069 
3070 	list_for_each_entry(ha, &list->list, list) {
3071 		if (*idx < cb->args[2])
3072 			goto skip;
3073 
3074 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3075 					      portid, seq,
3076 					      RTM_NEWNEIGH, NTF_SELF,
3077 					      NLM_F_MULTI, NUD_PERMANENT);
3078 		if (err < 0)
3079 			return err;
3080 skip:
3081 		*idx += 1;
3082 	}
3083 	return 0;
3084 }
3085 
3086 /**
3087  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3088  * @nlh: netlink message header
3089  * @dev: netdevice
3090  *
3091  * Default netdevice operation to dump the existing unicast address list.
3092  * Returns number of addresses from list put in skb.
3093  */
3094 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3095 		      struct netlink_callback *cb,
3096 		      struct net_device *dev,
3097 		      struct net_device *filter_dev,
3098 		      int *idx)
3099 {
3100 	int err;
3101 
3102 	netif_addr_lock_bh(dev);
3103 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3104 	if (err)
3105 		goto out;
3106 	nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3107 out:
3108 	netif_addr_unlock_bh(dev);
3109 	return err;
3110 }
3111 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3112 
3113 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3114 {
3115 	struct net_device *dev;
3116 	struct nlattr *tb[IFLA_MAX+1];
3117 	struct net_device *br_dev = NULL;
3118 	const struct net_device_ops *ops = NULL;
3119 	const struct net_device_ops *cops = NULL;
3120 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3121 	struct net *net = sock_net(skb->sk);
3122 	struct hlist_head *head;
3123 	int brport_idx = 0;
3124 	int br_idx = 0;
3125 	int h, s_h;
3126 	int idx = 0, s_idx;
3127 	int err = 0;
3128 	int fidx = 0;
3129 
3130 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3131 			ifla_policy) == 0) {
3132 		if (tb[IFLA_MASTER])
3133 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3134 	}
3135 
3136 	brport_idx = ifm->ifi_index;
3137 
3138 	if (br_idx) {
3139 		br_dev = __dev_get_by_index(net, br_idx);
3140 		if (!br_dev)
3141 			return -ENODEV;
3142 
3143 		ops = br_dev->netdev_ops;
3144 	}
3145 
3146 	s_h = cb->args[0];
3147 	s_idx = cb->args[1];
3148 
3149 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3150 		idx = 0;
3151 		head = &net->dev_index_head[h];
3152 		hlist_for_each_entry(dev, head, index_hlist) {
3153 
3154 			if (brport_idx && (dev->ifindex != brport_idx))
3155 				continue;
3156 
3157 			if (!br_idx) { /* user did not specify a specific bridge */
3158 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3159 					br_dev = netdev_master_upper_dev_get(dev);
3160 					cops = br_dev->netdev_ops;
3161 				}
3162 			} else {
3163 				if (dev != br_dev &&
3164 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3165 					continue;
3166 
3167 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3168 				    !(dev->priv_flags & IFF_EBRIDGE))
3169 					continue;
3170 				cops = ops;
3171 			}
3172 
3173 			if (idx < s_idx)
3174 				goto cont;
3175 
3176 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3177 				if (cops && cops->ndo_fdb_dump) {
3178 					err = cops->ndo_fdb_dump(skb, cb,
3179 								br_dev, dev,
3180 								&fidx);
3181 					if (err == -EMSGSIZE)
3182 						goto out;
3183 				}
3184 			}
3185 
3186 			if (dev->netdev_ops->ndo_fdb_dump)
3187 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3188 								    dev, NULL,
3189 								    &fidx);
3190 			else
3191 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3192 							&fidx);
3193 			if (err == -EMSGSIZE)
3194 				goto out;
3195 
3196 			cops = NULL;
3197 
3198 			/* reset fdb offset to 0 for rest of the interfaces */
3199 			cb->args[2] = 0;
3200 			fidx = 0;
3201 cont:
3202 			idx++;
3203 		}
3204 	}
3205 
3206 out:
3207 	cb->args[0] = h;
3208 	cb->args[1] = idx;
3209 	cb->args[2] = fidx;
3210 
3211 	return skb->len;
3212 }
3213 
3214 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3215 			       unsigned int attrnum, unsigned int flag)
3216 {
3217 	if (mask & flag)
3218 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3219 	return 0;
3220 }
3221 
3222 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3223 			    struct net_device *dev, u16 mode,
3224 			    u32 flags, u32 mask, int nlflags,
3225 			    u32 filter_mask,
3226 			    int (*vlan_fill)(struct sk_buff *skb,
3227 					     struct net_device *dev,
3228 					     u32 filter_mask))
3229 {
3230 	struct nlmsghdr *nlh;
3231 	struct ifinfomsg *ifm;
3232 	struct nlattr *br_afspec;
3233 	struct nlattr *protinfo;
3234 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3235 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3236 	int err = 0;
3237 
3238 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3239 	if (nlh == NULL)
3240 		return -EMSGSIZE;
3241 
3242 	ifm = nlmsg_data(nlh);
3243 	ifm->ifi_family = AF_BRIDGE;
3244 	ifm->__ifi_pad = 0;
3245 	ifm->ifi_type = dev->type;
3246 	ifm->ifi_index = dev->ifindex;
3247 	ifm->ifi_flags = dev_get_flags(dev);
3248 	ifm->ifi_change = 0;
3249 
3250 
3251 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3252 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3253 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3254 	    (br_dev &&
3255 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3256 	    (dev->addr_len &&
3257 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3258 	    (dev->ifindex != dev_get_iflink(dev) &&
3259 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3260 		goto nla_put_failure;
3261 
3262 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3263 	if (!br_afspec)
3264 		goto nla_put_failure;
3265 
3266 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3267 		nla_nest_cancel(skb, br_afspec);
3268 		goto nla_put_failure;
3269 	}
3270 
3271 	if (mode != BRIDGE_MODE_UNDEF) {
3272 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3273 			nla_nest_cancel(skb, br_afspec);
3274 			goto nla_put_failure;
3275 		}
3276 	}
3277 	if (vlan_fill) {
3278 		err = vlan_fill(skb, dev, filter_mask);
3279 		if (err) {
3280 			nla_nest_cancel(skb, br_afspec);
3281 			goto nla_put_failure;
3282 		}
3283 	}
3284 	nla_nest_end(skb, br_afspec);
3285 
3286 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3287 	if (!protinfo)
3288 		goto nla_put_failure;
3289 
3290 	if (brport_nla_put_flag(skb, flags, mask,
3291 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3292 	    brport_nla_put_flag(skb, flags, mask,
3293 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3294 	    brport_nla_put_flag(skb, flags, mask,
3295 				IFLA_BRPORT_FAST_LEAVE,
3296 				BR_MULTICAST_FAST_LEAVE) ||
3297 	    brport_nla_put_flag(skb, flags, mask,
3298 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3299 	    brport_nla_put_flag(skb, flags, mask,
3300 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3301 	    brport_nla_put_flag(skb, flags, mask,
3302 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3303 	    brport_nla_put_flag(skb, flags, mask,
3304 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3305 	    brport_nla_put_flag(skb, flags, mask,
3306 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3307 		nla_nest_cancel(skb, protinfo);
3308 		goto nla_put_failure;
3309 	}
3310 
3311 	nla_nest_end(skb, protinfo);
3312 
3313 	nlmsg_end(skb, nlh);
3314 	return 0;
3315 nla_put_failure:
3316 	nlmsg_cancel(skb, nlh);
3317 	return err ? err : -EMSGSIZE;
3318 }
3319 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3320 
3321 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3322 {
3323 	struct net *net = sock_net(skb->sk);
3324 	struct net_device *dev;
3325 	int idx = 0;
3326 	u32 portid = NETLINK_CB(cb->skb).portid;
3327 	u32 seq = cb->nlh->nlmsg_seq;
3328 	u32 filter_mask = 0;
3329 	int err;
3330 
3331 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3332 		struct nlattr *extfilt;
3333 
3334 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3335 					  IFLA_EXT_MASK);
3336 		if (extfilt) {
3337 			if (nla_len(extfilt) < sizeof(filter_mask))
3338 				return -EINVAL;
3339 
3340 			filter_mask = nla_get_u32(extfilt);
3341 		}
3342 	}
3343 
3344 	rcu_read_lock();
3345 	for_each_netdev_rcu(net, dev) {
3346 		const struct net_device_ops *ops = dev->netdev_ops;
3347 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3348 
3349 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3350 			if (idx >= cb->args[0]) {
3351 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3352 						skb, portid, seq, dev,
3353 						filter_mask, NLM_F_MULTI);
3354 				if (err < 0 && err != -EOPNOTSUPP)
3355 					break;
3356 			}
3357 			idx++;
3358 		}
3359 
3360 		if (ops->ndo_bridge_getlink) {
3361 			if (idx >= cb->args[0]) {
3362 				err = ops->ndo_bridge_getlink(skb, portid,
3363 							      seq, dev,
3364 							      filter_mask,
3365 							      NLM_F_MULTI);
3366 				if (err < 0 && err != -EOPNOTSUPP)
3367 					break;
3368 			}
3369 			idx++;
3370 		}
3371 	}
3372 	rcu_read_unlock();
3373 	cb->args[0] = idx;
3374 
3375 	return skb->len;
3376 }
3377 
3378 static inline size_t bridge_nlmsg_size(void)
3379 {
3380 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3381 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3382 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3383 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3384 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3385 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3386 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3387 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3388 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3389 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3390 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3391 }
3392 
3393 static int rtnl_bridge_notify(struct net_device *dev)
3394 {
3395 	struct net *net = dev_net(dev);
3396 	struct sk_buff *skb;
3397 	int err = -EOPNOTSUPP;
3398 
3399 	if (!dev->netdev_ops->ndo_bridge_getlink)
3400 		return 0;
3401 
3402 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3403 	if (!skb) {
3404 		err = -ENOMEM;
3405 		goto errout;
3406 	}
3407 
3408 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3409 	if (err < 0)
3410 		goto errout;
3411 
3412 	if (!skb->len)
3413 		goto errout;
3414 
3415 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3416 	return 0;
3417 errout:
3418 	WARN_ON(err == -EMSGSIZE);
3419 	kfree_skb(skb);
3420 	if (err)
3421 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3422 	return err;
3423 }
3424 
3425 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3426 {
3427 	struct net *net = sock_net(skb->sk);
3428 	struct ifinfomsg *ifm;
3429 	struct net_device *dev;
3430 	struct nlattr *br_spec, *attr = NULL;
3431 	int rem, err = -EOPNOTSUPP;
3432 	u16 flags = 0;
3433 	bool have_flags = false;
3434 
3435 	if (nlmsg_len(nlh) < sizeof(*ifm))
3436 		return -EINVAL;
3437 
3438 	ifm = nlmsg_data(nlh);
3439 	if (ifm->ifi_family != AF_BRIDGE)
3440 		return -EPFNOSUPPORT;
3441 
3442 	dev = __dev_get_by_index(net, ifm->ifi_index);
3443 	if (!dev) {
3444 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3445 		return -ENODEV;
3446 	}
3447 
3448 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3449 	if (br_spec) {
3450 		nla_for_each_nested(attr, br_spec, rem) {
3451 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3452 				if (nla_len(attr) < sizeof(flags))
3453 					return -EINVAL;
3454 
3455 				have_flags = true;
3456 				flags = nla_get_u16(attr);
3457 				break;
3458 			}
3459 		}
3460 	}
3461 
3462 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3463 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3464 
3465 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3466 			err = -EOPNOTSUPP;
3467 			goto out;
3468 		}
3469 
3470 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3471 		if (err)
3472 			goto out;
3473 
3474 		flags &= ~BRIDGE_FLAGS_MASTER;
3475 	}
3476 
3477 	if ((flags & BRIDGE_FLAGS_SELF)) {
3478 		if (!dev->netdev_ops->ndo_bridge_setlink)
3479 			err = -EOPNOTSUPP;
3480 		else
3481 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3482 								  flags);
3483 		if (!err) {
3484 			flags &= ~BRIDGE_FLAGS_SELF;
3485 
3486 			/* Generate event to notify upper layer of bridge
3487 			 * change
3488 			 */
3489 			err = rtnl_bridge_notify(dev);
3490 		}
3491 	}
3492 
3493 	if (have_flags)
3494 		memcpy(nla_data(attr), &flags, sizeof(flags));
3495 out:
3496 	return err;
3497 }
3498 
3499 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3500 {
3501 	struct net *net = sock_net(skb->sk);
3502 	struct ifinfomsg *ifm;
3503 	struct net_device *dev;
3504 	struct nlattr *br_spec, *attr = NULL;
3505 	int rem, err = -EOPNOTSUPP;
3506 	u16 flags = 0;
3507 	bool have_flags = false;
3508 
3509 	if (nlmsg_len(nlh) < sizeof(*ifm))
3510 		return -EINVAL;
3511 
3512 	ifm = nlmsg_data(nlh);
3513 	if (ifm->ifi_family != AF_BRIDGE)
3514 		return -EPFNOSUPPORT;
3515 
3516 	dev = __dev_get_by_index(net, ifm->ifi_index);
3517 	if (!dev) {
3518 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3519 		return -ENODEV;
3520 	}
3521 
3522 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3523 	if (br_spec) {
3524 		nla_for_each_nested(attr, br_spec, rem) {
3525 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3526 				if (nla_len(attr) < sizeof(flags))
3527 					return -EINVAL;
3528 
3529 				have_flags = true;
3530 				flags = nla_get_u16(attr);
3531 				break;
3532 			}
3533 		}
3534 	}
3535 
3536 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3537 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3538 
3539 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3540 			err = -EOPNOTSUPP;
3541 			goto out;
3542 		}
3543 
3544 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3545 		if (err)
3546 			goto out;
3547 
3548 		flags &= ~BRIDGE_FLAGS_MASTER;
3549 	}
3550 
3551 	if ((flags & BRIDGE_FLAGS_SELF)) {
3552 		if (!dev->netdev_ops->ndo_bridge_dellink)
3553 			err = -EOPNOTSUPP;
3554 		else
3555 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3556 								  flags);
3557 
3558 		if (!err) {
3559 			flags &= ~BRIDGE_FLAGS_SELF;
3560 
3561 			/* Generate event to notify upper layer of bridge
3562 			 * change
3563 			 */
3564 			err = rtnl_bridge_notify(dev);
3565 		}
3566 	}
3567 
3568 	if (have_flags)
3569 		memcpy(nla_data(attr), &flags, sizeof(flags));
3570 out:
3571 	return err;
3572 }
3573 
3574 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3575 {
3576 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3577 	       (!idxattr || idxattr == attrid);
3578 }
3579 
3580 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3581 			       int type, u32 pid, u32 seq, u32 change,
3582 			       unsigned int flags, unsigned int filter_mask,
3583 			       int *idxattr, int *prividx)
3584 {
3585 	struct if_stats_msg *ifsm;
3586 	struct nlmsghdr *nlh;
3587 	struct nlattr *attr;
3588 	int s_prividx = *prividx;
3589 
3590 	ASSERT_RTNL();
3591 
3592 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3593 	if (!nlh)
3594 		return -EMSGSIZE;
3595 
3596 	ifsm = nlmsg_data(nlh);
3597 	ifsm->ifindex = dev->ifindex;
3598 	ifsm->filter_mask = filter_mask;
3599 
3600 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3601 		struct rtnl_link_stats64 *sp;
3602 
3603 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3604 					 sizeof(struct rtnl_link_stats64),
3605 					 IFLA_STATS_UNSPEC);
3606 		if (!attr)
3607 			goto nla_put_failure;
3608 
3609 		sp = nla_data(attr);
3610 		dev_get_stats(dev, sp);
3611 	}
3612 
3613 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3614 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3615 
3616 		if (ops && ops->fill_linkxstats) {
3617 			int err;
3618 
3619 			*idxattr = IFLA_STATS_LINK_XSTATS;
3620 			attr = nla_nest_start(skb,
3621 					      IFLA_STATS_LINK_XSTATS);
3622 			if (!attr)
3623 				goto nla_put_failure;
3624 
3625 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3626 			nla_nest_end(skb, attr);
3627 			if (err)
3628 				goto nla_put_failure;
3629 			*idxattr = 0;
3630 		}
3631 	}
3632 
3633 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3634 			     *idxattr)) {
3635 		const struct rtnl_link_ops *ops = NULL;
3636 		const struct net_device *master;
3637 
3638 		master = netdev_master_upper_dev_get(dev);
3639 		if (master)
3640 			ops = master->rtnl_link_ops;
3641 		if (ops && ops->fill_linkxstats) {
3642 			int err;
3643 
3644 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3645 			attr = nla_nest_start(skb,
3646 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3647 			if (!attr)
3648 				goto nla_put_failure;
3649 
3650 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3651 			nla_nest_end(skb, attr);
3652 			if (err)
3653 				goto nla_put_failure;
3654 			*idxattr = 0;
3655 		}
3656 	}
3657 
3658 	nlmsg_end(skb, nlh);
3659 
3660 	return 0;
3661 
3662 nla_put_failure:
3663 	/* not a multi message or no progress mean a real error */
3664 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3665 		nlmsg_cancel(skb, nlh);
3666 	else
3667 		nlmsg_end(skb, nlh);
3668 
3669 	return -EMSGSIZE;
3670 }
3671 
3672 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3673 				  u32 filter_mask)
3674 {
3675 	size_t size = 0;
3676 
3677 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3678 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3679 
3680 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3681 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3682 		int attr = IFLA_STATS_LINK_XSTATS;
3683 
3684 		if (ops && ops->get_linkxstats_size) {
3685 			size += nla_total_size(ops->get_linkxstats_size(dev,
3686 									attr));
3687 			/* for IFLA_STATS_LINK_XSTATS */
3688 			size += nla_total_size(0);
3689 		}
3690 	}
3691 
3692 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3693 		struct net_device *_dev = (struct net_device *)dev;
3694 		const struct rtnl_link_ops *ops = NULL;
3695 		const struct net_device *master;
3696 
3697 		/* netdev_master_upper_dev_get can't take const */
3698 		master = netdev_master_upper_dev_get(_dev);
3699 		if (master)
3700 			ops = master->rtnl_link_ops;
3701 		if (ops && ops->get_linkxstats_size) {
3702 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3703 
3704 			size += nla_total_size(ops->get_linkxstats_size(dev,
3705 									attr));
3706 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
3707 			size += nla_total_size(0);
3708 		}
3709 	}
3710 
3711 	return size;
3712 }
3713 
3714 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3715 {
3716 	struct net *net = sock_net(skb->sk);
3717 	struct net_device *dev = NULL;
3718 	int idxattr = 0, prividx = 0;
3719 	struct if_stats_msg *ifsm;
3720 	struct sk_buff *nskb;
3721 	u32 filter_mask;
3722 	int err;
3723 
3724 	ifsm = nlmsg_data(nlh);
3725 	if (ifsm->ifindex > 0)
3726 		dev = __dev_get_by_index(net, ifsm->ifindex);
3727 	else
3728 		return -EINVAL;
3729 
3730 	if (!dev)
3731 		return -ENODEV;
3732 
3733 	filter_mask = ifsm->filter_mask;
3734 	if (!filter_mask)
3735 		return -EINVAL;
3736 
3737 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3738 	if (!nskb)
3739 		return -ENOBUFS;
3740 
3741 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3742 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3743 				  0, filter_mask, &idxattr, &prividx);
3744 	if (err < 0) {
3745 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3746 		WARN_ON(err == -EMSGSIZE);
3747 		kfree_skb(nskb);
3748 	} else {
3749 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3750 	}
3751 
3752 	return err;
3753 }
3754 
3755 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3756 {
3757 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
3758 	struct net *net = sock_net(skb->sk);
3759 	unsigned int flags = NLM_F_MULTI;
3760 	struct if_stats_msg *ifsm;
3761 	struct hlist_head *head;
3762 	struct net_device *dev;
3763 	u32 filter_mask = 0;
3764 	int idx = 0;
3765 
3766 	s_h = cb->args[0];
3767 	s_idx = cb->args[1];
3768 	s_idxattr = cb->args[2];
3769 	s_prividx = cb->args[3];
3770 
3771 	cb->seq = net->dev_base_seq;
3772 
3773 	ifsm = nlmsg_data(cb->nlh);
3774 	filter_mask = ifsm->filter_mask;
3775 	if (!filter_mask)
3776 		return -EINVAL;
3777 
3778 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3779 		idx = 0;
3780 		head = &net->dev_index_head[h];
3781 		hlist_for_each_entry(dev, head, index_hlist) {
3782 			if (idx < s_idx)
3783 				goto cont;
3784 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3785 						  NETLINK_CB(cb->skb).portid,
3786 						  cb->nlh->nlmsg_seq, 0,
3787 						  flags, filter_mask,
3788 						  &s_idxattr, &s_prividx);
3789 			/* If we ran out of room on the first message,
3790 			 * we're in trouble
3791 			 */
3792 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3793 
3794 			if (err < 0)
3795 				goto out;
3796 			s_prividx = 0;
3797 			s_idxattr = 0;
3798 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3799 cont:
3800 			idx++;
3801 		}
3802 	}
3803 out:
3804 	cb->args[3] = s_prividx;
3805 	cb->args[2] = s_idxattr;
3806 	cb->args[1] = idx;
3807 	cb->args[0] = h;
3808 
3809 	return skb->len;
3810 }
3811 
3812 /* Process one rtnetlink message. */
3813 
3814 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3815 {
3816 	struct net *net = sock_net(skb->sk);
3817 	rtnl_doit_func doit;
3818 	int kind;
3819 	int family;
3820 	int type;
3821 	int err;
3822 
3823 	type = nlh->nlmsg_type;
3824 	if (type > RTM_MAX)
3825 		return -EOPNOTSUPP;
3826 
3827 	type -= RTM_BASE;
3828 
3829 	/* All the messages must have at least 1 byte length */
3830 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3831 		return 0;
3832 
3833 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3834 	kind = type&3;
3835 
3836 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3837 		return -EPERM;
3838 
3839 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3840 		struct sock *rtnl;
3841 		rtnl_dumpit_func dumpit;
3842 		rtnl_calcit_func calcit;
3843 		u16 min_dump_alloc = 0;
3844 
3845 		dumpit = rtnl_get_dumpit(family, type);
3846 		if (dumpit == NULL)
3847 			return -EOPNOTSUPP;
3848 		calcit = rtnl_get_calcit(family, type);
3849 		if (calcit)
3850 			min_dump_alloc = calcit(skb, nlh);
3851 
3852 		__rtnl_unlock();
3853 		rtnl = net->rtnl;
3854 		{
3855 			struct netlink_dump_control c = {
3856 				.dump		= dumpit,
3857 				.min_dump_alloc	= min_dump_alloc,
3858 			};
3859 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3860 		}
3861 		rtnl_lock();
3862 		return err;
3863 	}
3864 
3865 	doit = rtnl_get_doit(family, type);
3866 	if (doit == NULL)
3867 		return -EOPNOTSUPP;
3868 
3869 	return doit(skb, nlh);
3870 }
3871 
3872 static void rtnetlink_rcv(struct sk_buff *skb)
3873 {
3874 	rtnl_lock();
3875 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3876 	rtnl_unlock();
3877 }
3878 
3879 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3880 {
3881 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3882 
3883 	switch (event) {
3884 	case NETDEV_UP:
3885 	case NETDEV_DOWN:
3886 	case NETDEV_PRE_UP:
3887 	case NETDEV_POST_INIT:
3888 	case NETDEV_REGISTER:
3889 	case NETDEV_CHANGE:
3890 	case NETDEV_PRE_TYPE_CHANGE:
3891 	case NETDEV_GOING_DOWN:
3892 	case NETDEV_UNREGISTER:
3893 	case NETDEV_UNREGISTER_FINAL:
3894 	case NETDEV_RELEASE:
3895 	case NETDEV_JOIN:
3896 	case NETDEV_BONDING_INFO:
3897 		break;
3898 	default:
3899 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3900 		break;
3901 	}
3902 	return NOTIFY_DONE;
3903 }
3904 
3905 static struct notifier_block rtnetlink_dev_notifier = {
3906 	.notifier_call	= rtnetlink_event,
3907 };
3908 
3909 
3910 static int __net_init rtnetlink_net_init(struct net *net)
3911 {
3912 	struct sock *sk;
3913 	struct netlink_kernel_cfg cfg = {
3914 		.groups		= RTNLGRP_MAX,
3915 		.input		= rtnetlink_rcv,
3916 		.cb_mutex	= &rtnl_mutex,
3917 		.flags		= NL_CFG_F_NONROOT_RECV,
3918 	};
3919 
3920 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3921 	if (!sk)
3922 		return -ENOMEM;
3923 	net->rtnl = sk;
3924 	return 0;
3925 }
3926 
3927 static void __net_exit rtnetlink_net_exit(struct net *net)
3928 {
3929 	netlink_kernel_release(net->rtnl);
3930 	net->rtnl = NULL;
3931 }
3932 
3933 static struct pernet_operations rtnetlink_net_ops = {
3934 	.init = rtnetlink_net_init,
3935 	.exit = rtnetlink_net_exit,
3936 };
3937 
3938 void __init rtnetlink_init(void)
3939 {
3940 	if (register_pernet_subsys(&rtnetlink_net_ops))
3941 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3942 
3943 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3944 
3945 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3946 		      rtnl_dump_ifinfo, rtnl_calcit);
3947 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3948 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3949 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3950 
3951 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3952 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3953 
3954 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3955 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3956 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3957 
3958 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3959 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3960 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3961 
3962 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
3963 		      NULL);
3964 }
3965