xref: /openbmc/linux/net/core/rtnetlink.c (revision d1fdd9138682e0f272beee0cb08b6328c5478b26)
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 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
708 				if (nla_put_u32(skb, i + 1, user_features))
709 					goto nla_put_failure;
710 			} else {
711 				if (nla_put_u32(skb, i + 1, metrics[i]))
712 					goto nla_put_failure;
713 			}
714 			valid++;
715 		}
716 	}
717 
718 	if (!valid) {
719 		nla_nest_cancel(skb, mx);
720 		return 0;
721 	}
722 
723 	return nla_nest_end(skb, mx);
724 
725 nla_put_failure:
726 	nla_nest_cancel(skb, mx);
727 	return -EMSGSIZE;
728 }
729 EXPORT_SYMBOL(rtnetlink_put_metrics);
730 
731 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
732 		       long expires, u32 error)
733 {
734 	struct rta_cacheinfo ci = {
735 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
736 		.rta_used = dst->__use,
737 		.rta_clntref = atomic_read(&(dst->__refcnt)),
738 		.rta_error = error,
739 		.rta_id =  id,
740 	};
741 
742 	if (expires) {
743 		unsigned long clock;
744 
745 		clock = jiffies_to_clock_t(abs(expires));
746 		clock = min_t(unsigned long, clock, INT_MAX);
747 		ci.rta_expires = (expires > 0) ? clock : -clock;
748 	}
749 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
750 }
751 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
752 
753 static void set_operstate(struct net_device *dev, unsigned char transition)
754 {
755 	unsigned char operstate = dev->operstate;
756 
757 	switch (transition) {
758 	case IF_OPER_UP:
759 		if ((operstate == IF_OPER_DORMANT ||
760 		     operstate == IF_OPER_UNKNOWN) &&
761 		    !netif_dormant(dev))
762 			operstate = IF_OPER_UP;
763 		break;
764 
765 	case IF_OPER_DORMANT:
766 		if (operstate == IF_OPER_UP ||
767 		    operstate == IF_OPER_UNKNOWN)
768 			operstate = IF_OPER_DORMANT;
769 		break;
770 	}
771 
772 	if (dev->operstate != operstate) {
773 		write_lock_bh(&dev_base_lock);
774 		dev->operstate = operstate;
775 		write_unlock_bh(&dev_base_lock);
776 		netdev_state_change(dev);
777 	}
778 }
779 
780 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
781 {
782 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
783 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
784 }
785 
786 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
787 					   const struct ifinfomsg *ifm)
788 {
789 	unsigned int flags = ifm->ifi_flags;
790 
791 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
792 	if (ifm->ifi_change)
793 		flags = (flags & ifm->ifi_change) |
794 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
795 
796 	return flags;
797 }
798 
799 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
800 				 const struct rtnl_link_stats64 *b)
801 {
802 	a->rx_packets = b->rx_packets;
803 	a->tx_packets = b->tx_packets;
804 	a->rx_bytes = b->rx_bytes;
805 	a->tx_bytes = b->tx_bytes;
806 	a->rx_errors = b->rx_errors;
807 	a->tx_errors = b->tx_errors;
808 	a->rx_dropped = b->rx_dropped;
809 	a->tx_dropped = b->tx_dropped;
810 
811 	a->multicast = b->multicast;
812 	a->collisions = b->collisions;
813 
814 	a->rx_length_errors = b->rx_length_errors;
815 	a->rx_over_errors = b->rx_over_errors;
816 	a->rx_crc_errors = b->rx_crc_errors;
817 	a->rx_frame_errors = b->rx_frame_errors;
818 	a->rx_fifo_errors = b->rx_fifo_errors;
819 	a->rx_missed_errors = b->rx_missed_errors;
820 
821 	a->tx_aborted_errors = b->tx_aborted_errors;
822 	a->tx_carrier_errors = b->tx_carrier_errors;
823 	a->tx_fifo_errors = b->tx_fifo_errors;
824 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
825 	a->tx_window_errors = b->tx_window_errors;
826 
827 	a->rx_compressed = b->rx_compressed;
828 	a->tx_compressed = b->tx_compressed;
829 
830 	a->rx_nohandler = b->rx_nohandler;
831 }
832 
833 /* All VF info */
834 static inline int rtnl_vfinfo_size(const struct net_device *dev,
835 				   u32 ext_filter_mask)
836 {
837 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
838 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
839 		int num_vfs = dev_num_vf(dev->dev.parent);
840 		size_t size = nla_total_size(sizeof(struct nlattr));
841 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
842 		size += num_vfs *
843 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
844 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
845 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
846 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
847 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
848 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
849 			 /* IFLA_VF_STATS_RX_PACKETS */
850 			 nla_total_size_64bit(sizeof(__u64)) +
851 			 /* IFLA_VF_STATS_TX_PACKETS */
852 			 nla_total_size_64bit(sizeof(__u64)) +
853 			 /* IFLA_VF_STATS_RX_BYTES */
854 			 nla_total_size_64bit(sizeof(__u64)) +
855 			 /* IFLA_VF_STATS_TX_BYTES */
856 			 nla_total_size_64bit(sizeof(__u64)) +
857 			 /* IFLA_VF_STATS_BROADCAST */
858 			 nla_total_size_64bit(sizeof(__u64)) +
859 			 /* IFLA_VF_STATS_MULTICAST */
860 			 nla_total_size_64bit(sizeof(__u64)) +
861 			 nla_total_size(sizeof(struct ifla_vf_trust)));
862 		return size;
863 	} else
864 		return 0;
865 }
866 
867 static size_t rtnl_port_size(const struct net_device *dev,
868 			     u32 ext_filter_mask)
869 {
870 	size_t port_size = nla_total_size(4)		/* PORT_VF */
871 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
872 		+ nla_total_size(sizeof(struct ifla_port_vsi))
873 							/* PORT_VSI_TYPE */
874 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
875 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
876 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
877 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
878 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
879 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
880 		+ port_size;
881 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
882 		+ port_size;
883 
884 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
885 	    !(ext_filter_mask & RTEXT_FILTER_VF))
886 		return 0;
887 	if (dev_num_vf(dev->dev.parent))
888 		return port_self_size + vf_ports_size +
889 			vf_port_size * dev_num_vf(dev->dev.parent);
890 	else
891 		return port_self_size;
892 }
893 
894 static size_t rtnl_xdp_size(const struct net_device *dev)
895 {
896 	size_t xdp_size = nla_total_size(1);	/* XDP_ATTACHED */
897 
898 	if (!dev->netdev_ops->ndo_xdp)
899 		return 0;
900 	else
901 		return xdp_size;
902 }
903 
904 static noinline size_t if_nlmsg_size(const struct net_device *dev,
905 				     u32 ext_filter_mask)
906 {
907 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
908 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
909 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
910 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
911 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
912 	       + nla_total_size(sizeof(struct rtnl_link_stats))
913 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
914 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
915 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
916 	       + nla_total_size(4) /* IFLA_TXQLEN */
917 	       + nla_total_size(4) /* IFLA_WEIGHT */
918 	       + nla_total_size(4) /* IFLA_MTU */
919 	       + nla_total_size(4) /* IFLA_LINK */
920 	       + nla_total_size(4) /* IFLA_MASTER */
921 	       + nla_total_size(1) /* IFLA_CARRIER */
922 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
923 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
924 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
925 	       + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
926 	       + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
927 	       + nla_total_size(1) /* IFLA_OPERSTATE */
928 	       + nla_total_size(1) /* IFLA_LINKMODE */
929 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
930 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
931 	       + nla_total_size(ext_filter_mask
932 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
933 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
934 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
935 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
936 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
937 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
938 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
939 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
940 	       + rtnl_xdp_size(dev) /* IFLA_XDP */
941 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
942 
943 }
944 
945 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
946 {
947 	struct nlattr *vf_ports;
948 	struct nlattr *vf_port;
949 	int vf;
950 	int err;
951 
952 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
953 	if (!vf_ports)
954 		return -EMSGSIZE;
955 
956 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
957 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
958 		if (!vf_port)
959 			goto nla_put_failure;
960 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
961 			goto nla_put_failure;
962 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
963 		if (err == -EMSGSIZE)
964 			goto nla_put_failure;
965 		if (err) {
966 			nla_nest_cancel(skb, vf_port);
967 			continue;
968 		}
969 		nla_nest_end(skb, vf_port);
970 	}
971 
972 	nla_nest_end(skb, vf_ports);
973 
974 	return 0;
975 
976 nla_put_failure:
977 	nla_nest_cancel(skb, vf_ports);
978 	return -EMSGSIZE;
979 }
980 
981 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
982 {
983 	struct nlattr *port_self;
984 	int err;
985 
986 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
987 	if (!port_self)
988 		return -EMSGSIZE;
989 
990 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
991 	if (err) {
992 		nla_nest_cancel(skb, port_self);
993 		return (err == -EMSGSIZE) ? err : 0;
994 	}
995 
996 	nla_nest_end(skb, port_self);
997 
998 	return 0;
999 }
1000 
1001 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1002 			  u32 ext_filter_mask)
1003 {
1004 	int err;
1005 
1006 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1007 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1008 		return 0;
1009 
1010 	err = rtnl_port_self_fill(skb, dev);
1011 	if (err)
1012 		return err;
1013 
1014 	if (dev_num_vf(dev->dev.parent)) {
1015 		err = rtnl_vf_ports_fill(skb, dev);
1016 		if (err)
1017 			return err;
1018 	}
1019 
1020 	return 0;
1021 }
1022 
1023 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1024 {
1025 	int err;
1026 	struct netdev_phys_item_id ppid;
1027 
1028 	err = dev_get_phys_port_id(dev, &ppid);
1029 	if (err) {
1030 		if (err == -EOPNOTSUPP)
1031 			return 0;
1032 		return err;
1033 	}
1034 
1035 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1036 		return -EMSGSIZE;
1037 
1038 	return 0;
1039 }
1040 
1041 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1042 {
1043 	char name[IFNAMSIZ];
1044 	int err;
1045 
1046 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1047 	if (err) {
1048 		if (err == -EOPNOTSUPP)
1049 			return 0;
1050 		return err;
1051 	}
1052 
1053 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1054 		return -EMSGSIZE;
1055 
1056 	return 0;
1057 }
1058 
1059 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1060 {
1061 	int err;
1062 	struct switchdev_attr attr = {
1063 		.orig_dev = dev,
1064 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1065 		.flags = SWITCHDEV_F_NO_RECURSE,
1066 	};
1067 
1068 	err = switchdev_port_attr_get(dev, &attr);
1069 	if (err) {
1070 		if (err == -EOPNOTSUPP)
1071 			return 0;
1072 		return err;
1073 	}
1074 
1075 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1076 		    attr.u.ppid.id))
1077 		return -EMSGSIZE;
1078 
1079 	return 0;
1080 }
1081 
1082 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1083 					      struct net_device *dev)
1084 {
1085 	struct rtnl_link_stats64 *sp;
1086 	struct nlattr *attr;
1087 
1088 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1089 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1090 	if (!attr)
1091 		return -EMSGSIZE;
1092 
1093 	sp = nla_data(attr);
1094 	dev_get_stats(dev, sp);
1095 
1096 	attr = nla_reserve(skb, IFLA_STATS,
1097 			   sizeof(struct rtnl_link_stats));
1098 	if (!attr)
1099 		return -EMSGSIZE;
1100 
1101 	copy_rtnl_link_stats(nla_data(attr), sp);
1102 
1103 	return 0;
1104 }
1105 
1106 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1107 					       struct net_device *dev,
1108 					       int vfs_num,
1109 					       struct nlattr *vfinfo)
1110 {
1111 	struct ifla_vf_rss_query_en vf_rss_query_en;
1112 	struct ifla_vf_link_state vf_linkstate;
1113 	struct ifla_vf_spoofchk vf_spoofchk;
1114 	struct ifla_vf_tx_rate vf_tx_rate;
1115 	struct ifla_vf_stats vf_stats;
1116 	struct ifla_vf_trust vf_trust;
1117 	struct ifla_vf_vlan vf_vlan;
1118 	struct ifla_vf_rate vf_rate;
1119 	struct nlattr *vf, *vfstats;
1120 	struct ifla_vf_mac vf_mac;
1121 	struct ifla_vf_info ivi;
1122 
1123 	/* Not all SR-IOV capable drivers support the
1124 	 * spoofcheck and "RSS query enable" query.  Preset to
1125 	 * -1 so the user space tool can detect that the driver
1126 	 * didn't report anything.
1127 	 */
1128 	ivi.spoofchk = -1;
1129 	ivi.rss_query_en = -1;
1130 	ivi.trusted = -1;
1131 	memset(ivi.mac, 0, sizeof(ivi.mac));
1132 	/* The default value for VF link state is "auto"
1133 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1134 	 */
1135 	ivi.linkstate = 0;
1136 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1137 		return 0;
1138 
1139 	vf_mac.vf =
1140 		vf_vlan.vf =
1141 		vf_rate.vf =
1142 		vf_tx_rate.vf =
1143 		vf_spoofchk.vf =
1144 		vf_linkstate.vf =
1145 		vf_rss_query_en.vf =
1146 		vf_trust.vf = ivi.vf;
1147 
1148 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1149 	vf_vlan.vlan = ivi.vlan;
1150 	vf_vlan.qos = ivi.qos;
1151 	vf_tx_rate.rate = ivi.max_tx_rate;
1152 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1153 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1154 	vf_spoofchk.setting = ivi.spoofchk;
1155 	vf_linkstate.link_state = ivi.linkstate;
1156 	vf_rss_query_en.setting = ivi.rss_query_en;
1157 	vf_trust.setting = ivi.trusted;
1158 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1159 	if (!vf) {
1160 		nla_nest_cancel(skb, vfinfo);
1161 		return -EMSGSIZE;
1162 	}
1163 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1164 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1165 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1166 		    &vf_rate) ||
1167 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1168 		    &vf_tx_rate) ||
1169 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1170 		    &vf_spoofchk) ||
1171 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1172 		    &vf_linkstate) ||
1173 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1174 		    sizeof(vf_rss_query_en),
1175 		    &vf_rss_query_en) ||
1176 	    nla_put(skb, IFLA_VF_TRUST,
1177 		    sizeof(vf_trust), &vf_trust))
1178 		return -EMSGSIZE;
1179 	memset(&vf_stats, 0, sizeof(vf_stats));
1180 	if (dev->netdev_ops->ndo_get_vf_stats)
1181 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1182 						&vf_stats);
1183 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1184 	if (!vfstats) {
1185 		nla_nest_cancel(skb, vf);
1186 		nla_nest_cancel(skb, vfinfo);
1187 		return -EMSGSIZE;
1188 	}
1189 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1190 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1191 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1192 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1193 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1194 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1195 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1196 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1197 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1198 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1199 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1200 			      vf_stats.multicast, IFLA_VF_STATS_PAD))
1201 		return -EMSGSIZE;
1202 	nla_nest_end(skb, vfstats);
1203 	nla_nest_end(skb, vf);
1204 	return 0;
1205 }
1206 
1207 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1208 {
1209 	struct rtnl_link_ifmap map;
1210 
1211 	memset(&map, 0, sizeof(map));
1212 	map.mem_start   = dev->mem_start;
1213 	map.mem_end     = dev->mem_end;
1214 	map.base_addr   = dev->base_addr;
1215 	map.irq         = dev->irq;
1216 	map.dma         = dev->dma;
1217 	map.port        = dev->if_port;
1218 
1219 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1220 		return -EMSGSIZE;
1221 
1222 	return 0;
1223 }
1224 
1225 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1226 {
1227 	struct netdev_xdp xdp_op = {};
1228 	struct nlattr *xdp;
1229 	int err;
1230 
1231 	if (!dev->netdev_ops->ndo_xdp)
1232 		return 0;
1233 	xdp = nla_nest_start(skb, IFLA_XDP);
1234 	if (!xdp)
1235 		return -EMSGSIZE;
1236 	xdp_op.command = XDP_QUERY_PROG;
1237 	err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1238 	if (err)
1239 		goto err_cancel;
1240 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1241 	if (err)
1242 		goto err_cancel;
1243 
1244 	nla_nest_end(skb, xdp);
1245 	return 0;
1246 
1247 err_cancel:
1248 	nla_nest_cancel(skb, xdp);
1249 	return err;
1250 }
1251 
1252 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1253 			    int type, u32 pid, u32 seq, u32 change,
1254 			    unsigned int flags, u32 ext_filter_mask)
1255 {
1256 	struct ifinfomsg *ifm;
1257 	struct nlmsghdr *nlh;
1258 	struct nlattr *af_spec;
1259 	struct rtnl_af_ops *af_ops;
1260 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1261 
1262 	ASSERT_RTNL();
1263 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1264 	if (nlh == NULL)
1265 		return -EMSGSIZE;
1266 
1267 	ifm = nlmsg_data(nlh);
1268 	ifm->ifi_family = AF_UNSPEC;
1269 	ifm->__ifi_pad = 0;
1270 	ifm->ifi_type = dev->type;
1271 	ifm->ifi_index = dev->ifindex;
1272 	ifm->ifi_flags = dev_get_flags(dev);
1273 	ifm->ifi_change = change;
1274 
1275 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1276 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1277 	    nla_put_u8(skb, IFLA_OPERSTATE,
1278 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1279 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1280 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1281 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1282 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1283 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1284 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1285 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1286 #ifdef CONFIG_RPS
1287 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1288 #endif
1289 	    (dev->ifindex != dev_get_iflink(dev) &&
1290 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1291 	    (upper_dev &&
1292 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1293 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1294 	    (dev->qdisc &&
1295 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1296 	    (dev->ifalias &&
1297 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1298 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1299 			atomic_read(&dev->carrier_changes)) ||
1300 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1301 		goto nla_put_failure;
1302 
1303 	if (rtnl_fill_link_ifmap(skb, dev))
1304 		goto nla_put_failure;
1305 
1306 	if (dev->addr_len) {
1307 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1308 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1309 			goto nla_put_failure;
1310 	}
1311 
1312 	if (rtnl_phys_port_id_fill(skb, dev))
1313 		goto nla_put_failure;
1314 
1315 	if (rtnl_phys_port_name_fill(skb, dev))
1316 		goto nla_put_failure;
1317 
1318 	if (rtnl_phys_switch_id_fill(skb, dev))
1319 		goto nla_put_failure;
1320 
1321 	if (rtnl_fill_stats(skb, dev))
1322 		goto nla_put_failure;
1323 
1324 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1325 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1326 		goto nla_put_failure;
1327 
1328 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1329 	    ext_filter_mask & RTEXT_FILTER_VF) {
1330 		int i;
1331 		struct nlattr *vfinfo;
1332 		int num_vfs = dev_num_vf(dev->dev.parent);
1333 
1334 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1335 		if (!vfinfo)
1336 			goto nla_put_failure;
1337 		for (i = 0; i < num_vfs; i++) {
1338 			if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1339 				goto nla_put_failure;
1340 		}
1341 
1342 		nla_nest_end(skb, vfinfo);
1343 	}
1344 
1345 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1346 		goto nla_put_failure;
1347 
1348 	if (rtnl_xdp_fill(skb, dev))
1349 		goto nla_put_failure;
1350 
1351 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1352 		if (rtnl_link_fill(skb, dev) < 0)
1353 			goto nla_put_failure;
1354 	}
1355 
1356 	if (dev->rtnl_link_ops &&
1357 	    dev->rtnl_link_ops->get_link_net) {
1358 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1359 
1360 		if (!net_eq(dev_net(dev), link_net)) {
1361 			int id = peernet2id_alloc(dev_net(dev), link_net);
1362 
1363 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1364 				goto nla_put_failure;
1365 		}
1366 	}
1367 
1368 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1369 		goto nla_put_failure;
1370 
1371 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1372 		if (af_ops->fill_link_af) {
1373 			struct nlattr *af;
1374 			int err;
1375 
1376 			if (!(af = nla_nest_start(skb, af_ops->family)))
1377 				goto nla_put_failure;
1378 
1379 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1380 
1381 			/*
1382 			 * Caller may return ENODATA to indicate that there
1383 			 * was no data to be dumped. This is not an error, it
1384 			 * means we should trim the attribute header and
1385 			 * continue.
1386 			 */
1387 			if (err == -ENODATA)
1388 				nla_nest_cancel(skb, af);
1389 			else if (err < 0)
1390 				goto nla_put_failure;
1391 
1392 			nla_nest_end(skb, af);
1393 		}
1394 	}
1395 
1396 	nla_nest_end(skb, af_spec);
1397 
1398 	nlmsg_end(skb, nlh);
1399 	return 0;
1400 
1401 nla_put_failure:
1402 	nlmsg_cancel(skb, nlh);
1403 	return -EMSGSIZE;
1404 }
1405 
1406 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1407 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1408 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1409 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1410 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1411 	[IFLA_MTU]		= { .type = NLA_U32 },
1412 	[IFLA_LINK]		= { .type = NLA_U32 },
1413 	[IFLA_MASTER]		= { .type = NLA_U32 },
1414 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1415 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1416 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1417 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1418 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1419 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1420 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1421 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1422 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1423 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1424 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1425 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1426 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1427 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1428 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1429 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1430 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1431 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1432 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1433 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1434 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1435 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1436 	[IFLA_XDP]		= { .type = NLA_NESTED },
1437 };
1438 
1439 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1440 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1441 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1442 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1443 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1444 };
1445 
1446 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1447 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1448 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1449 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1450 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1451 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1452 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1453 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1454 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1455 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1456 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1457 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1458 };
1459 
1460 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1461 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1462 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1463 				    .len = PORT_PROFILE_MAX },
1464 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1465 				    .len = sizeof(struct ifla_port_vsi)},
1466 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1467 				      .len = PORT_UUID_MAX },
1468 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1469 				    .len = PORT_UUID_MAX },
1470 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1471 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1472 };
1473 
1474 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1475 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1476 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1477 };
1478 
1479 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1480 {
1481 	const struct rtnl_link_ops *ops = NULL;
1482 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1483 
1484 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1485 		return NULL;
1486 
1487 	if (linfo[IFLA_INFO_KIND]) {
1488 		char kind[MODULE_NAME_LEN];
1489 
1490 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1491 		ops = rtnl_link_ops_get(kind);
1492 	}
1493 
1494 	return ops;
1495 }
1496 
1497 static bool link_master_filtered(struct net_device *dev, int master_idx)
1498 {
1499 	struct net_device *master;
1500 
1501 	if (!master_idx)
1502 		return false;
1503 
1504 	master = netdev_master_upper_dev_get(dev);
1505 	if (!master || master->ifindex != master_idx)
1506 		return true;
1507 
1508 	return false;
1509 }
1510 
1511 static bool link_kind_filtered(const struct net_device *dev,
1512 			       const struct rtnl_link_ops *kind_ops)
1513 {
1514 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1515 		return true;
1516 
1517 	return false;
1518 }
1519 
1520 static bool link_dump_filtered(struct net_device *dev,
1521 			       int master_idx,
1522 			       const struct rtnl_link_ops *kind_ops)
1523 {
1524 	if (link_master_filtered(dev, master_idx) ||
1525 	    link_kind_filtered(dev, kind_ops))
1526 		return true;
1527 
1528 	return false;
1529 }
1530 
1531 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1532 {
1533 	struct net *net = sock_net(skb->sk);
1534 	int h, s_h;
1535 	int idx = 0, s_idx;
1536 	struct net_device *dev;
1537 	struct hlist_head *head;
1538 	struct nlattr *tb[IFLA_MAX+1];
1539 	u32 ext_filter_mask = 0;
1540 	const struct rtnl_link_ops *kind_ops = NULL;
1541 	unsigned int flags = NLM_F_MULTI;
1542 	int master_idx = 0;
1543 	int err;
1544 	int hdrlen;
1545 
1546 	s_h = cb->args[0];
1547 	s_idx = cb->args[1];
1548 
1549 	cb->seq = net->dev_base_seq;
1550 
1551 	/* A hack to preserve kernel<->userspace interface.
1552 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1553 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1554 	 * what iproute2 < v3.9.0 used.
1555 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1556 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1557 	 */
1558 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1559 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1560 
1561 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1562 
1563 		if (tb[IFLA_EXT_MASK])
1564 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1565 
1566 		if (tb[IFLA_MASTER])
1567 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1568 
1569 		if (tb[IFLA_LINKINFO])
1570 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1571 
1572 		if (master_idx || kind_ops)
1573 			flags |= NLM_F_DUMP_FILTERED;
1574 	}
1575 
1576 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1577 		idx = 0;
1578 		head = &net->dev_index_head[h];
1579 		hlist_for_each_entry(dev, head, index_hlist) {
1580 			if (link_dump_filtered(dev, master_idx, kind_ops))
1581 				continue;
1582 			if (idx < s_idx)
1583 				goto cont;
1584 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1585 					       NETLINK_CB(cb->skb).portid,
1586 					       cb->nlh->nlmsg_seq, 0,
1587 					       flags,
1588 					       ext_filter_mask);
1589 			/* If we ran out of room on the first message,
1590 			 * we're in trouble
1591 			 */
1592 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1593 
1594 			if (err < 0)
1595 				goto out;
1596 
1597 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1598 cont:
1599 			idx++;
1600 		}
1601 	}
1602 out:
1603 	cb->args[1] = idx;
1604 	cb->args[0] = h;
1605 
1606 	return skb->len;
1607 }
1608 
1609 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1610 {
1611 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1612 }
1613 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1614 
1615 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1616 {
1617 	struct net *net;
1618 	/* Examine the link attributes and figure out which
1619 	 * network namespace we are talking about.
1620 	 */
1621 	if (tb[IFLA_NET_NS_PID])
1622 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1623 	else if (tb[IFLA_NET_NS_FD])
1624 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1625 	else
1626 		net = get_net(src_net);
1627 	return net;
1628 }
1629 EXPORT_SYMBOL(rtnl_link_get_net);
1630 
1631 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1632 {
1633 	if (dev) {
1634 		if (tb[IFLA_ADDRESS] &&
1635 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1636 			return -EINVAL;
1637 
1638 		if (tb[IFLA_BROADCAST] &&
1639 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1640 			return -EINVAL;
1641 	}
1642 
1643 	if (tb[IFLA_AF_SPEC]) {
1644 		struct nlattr *af;
1645 		int rem, err;
1646 
1647 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1648 			const struct rtnl_af_ops *af_ops;
1649 
1650 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1651 				return -EAFNOSUPPORT;
1652 
1653 			if (!af_ops->set_link_af)
1654 				return -EOPNOTSUPP;
1655 
1656 			if (af_ops->validate_link_af) {
1657 				err = af_ops->validate_link_af(dev, af);
1658 				if (err < 0)
1659 					return err;
1660 			}
1661 		}
1662 	}
1663 
1664 	return 0;
1665 }
1666 
1667 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1668 				  int guid_type)
1669 {
1670 	const struct net_device_ops *ops = dev->netdev_ops;
1671 
1672 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1673 }
1674 
1675 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1676 {
1677 	if (dev->type != ARPHRD_INFINIBAND)
1678 		return -EOPNOTSUPP;
1679 
1680 	return handle_infiniband_guid(dev, ivt, guid_type);
1681 }
1682 
1683 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1684 {
1685 	const struct net_device_ops *ops = dev->netdev_ops;
1686 	int err = -EINVAL;
1687 
1688 	if (tb[IFLA_VF_MAC]) {
1689 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1690 
1691 		err = -EOPNOTSUPP;
1692 		if (ops->ndo_set_vf_mac)
1693 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1694 						  ivm->mac);
1695 		if (err < 0)
1696 			return err;
1697 	}
1698 
1699 	if (tb[IFLA_VF_VLAN]) {
1700 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1701 
1702 		err = -EOPNOTSUPP;
1703 		if (ops->ndo_set_vf_vlan)
1704 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1705 						   ivv->qos);
1706 		if (err < 0)
1707 			return err;
1708 	}
1709 
1710 	if (tb[IFLA_VF_TX_RATE]) {
1711 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1712 		struct ifla_vf_info ivf;
1713 
1714 		err = -EOPNOTSUPP;
1715 		if (ops->ndo_get_vf_config)
1716 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1717 		if (err < 0)
1718 			return err;
1719 
1720 		err = -EOPNOTSUPP;
1721 		if (ops->ndo_set_vf_rate)
1722 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1723 						   ivf.min_tx_rate,
1724 						   ivt->rate);
1725 		if (err < 0)
1726 			return err;
1727 	}
1728 
1729 	if (tb[IFLA_VF_RATE]) {
1730 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1731 
1732 		err = -EOPNOTSUPP;
1733 		if (ops->ndo_set_vf_rate)
1734 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1735 						   ivt->min_tx_rate,
1736 						   ivt->max_tx_rate);
1737 		if (err < 0)
1738 			return err;
1739 	}
1740 
1741 	if (tb[IFLA_VF_SPOOFCHK]) {
1742 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1743 
1744 		err = -EOPNOTSUPP;
1745 		if (ops->ndo_set_vf_spoofchk)
1746 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1747 						       ivs->setting);
1748 		if (err < 0)
1749 			return err;
1750 	}
1751 
1752 	if (tb[IFLA_VF_LINK_STATE]) {
1753 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1754 
1755 		err = -EOPNOTSUPP;
1756 		if (ops->ndo_set_vf_link_state)
1757 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1758 							 ivl->link_state);
1759 		if (err < 0)
1760 			return err;
1761 	}
1762 
1763 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1764 		struct ifla_vf_rss_query_en *ivrssq_en;
1765 
1766 		err = -EOPNOTSUPP;
1767 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1768 		if (ops->ndo_set_vf_rss_query_en)
1769 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1770 							   ivrssq_en->setting);
1771 		if (err < 0)
1772 			return err;
1773 	}
1774 
1775 	if (tb[IFLA_VF_TRUST]) {
1776 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1777 
1778 		err = -EOPNOTSUPP;
1779 		if (ops->ndo_set_vf_trust)
1780 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1781 		if (err < 0)
1782 			return err;
1783 	}
1784 
1785 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1786 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1787 
1788 		if (!ops->ndo_set_vf_guid)
1789 			return -EOPNOTSUPP;
1790 
1791 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1792 	}
1793 
1794 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1795 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1796 
1797 		if (!ops->ndo_set_vf_guid)
1798 			return -EOPNOTSUPP;
1799 
1800 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1801 	}
1802 
1803 	return err;
1804 }
1805 
1806 static int do_set_master(struct net_device *dev, int ifindex)
1807 {
1808 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1809 	const struct net_device_ops *ops;
1810 	int err;
1811 
1812 	if (upper_dev) {
1813 		if (upper_dev->ifindex == ifindex)
1814 			return 0;
1815 		ops = upper_dev->netdev_ops;
1816 		if (ops->ndo_del_slave) {
1817 			err = ops->ndo_del_slave(upper_dev, dev);
1818 			if (err)
1819 				return err;
1820 		} else {
1821 			return -EOPNOTSUPP;
1822 		}
1823 	}
1824 
1825 	if (ifindex) {
1826 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1827 		if (!upper_dev)
1828 			return -EINVAL;
1829 		ops = upper_dev->netdev_ops;
1830 		if (ops->ndo_add_slave) {
1831 			err = ops->ndo_add_slave(upper_dev, dev);
1832 			if (err)
1833 				return err;
1834 		} else {
1835 			return -EOPNOTSUPP;
1836 		}
1837 	}
1838 	return 0;
1839 }
1840 
1841 #define DO_SETLINK_MODIFIED	0x01
1842 /* notify flag means notify + modified. */
1843 #define DO_SETLINK_NOTIFY	0x03
1844 static int do_setlink(const struct sk_buff *skb,
1845 		      struct net_device *dev, struct ifinfomsg *ifm,
1846 		      struct nlattr **tb, char *ifname, int status)
1847 {
1848 	const struct net_device_ops *ops = dev->netdev_ops;
1849 	int err;
1850 
1851 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1852 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1853 		if (IS_ERR(net)) {
1854 			err = PTR_ERR(net);
1855 			goto errout;
1856 		}
1857 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1858 			put_net(net);
1859 			err = -EPERM;
1860 			goto errout;
1861 		}
1862 		err = dev_change_net_namespace(dev, net, ifname);
1863 		put_net(net);
1864 		if (err)
1865 			goto errout;
1866 		status |= DO_SETLINK_MODIFIED;
1867 	}
1868 
1869 	if (tb[IFLA_MAP]) {
1870 		struct rtnl_link_ifmap *u_map;
1871 		struct ifmap k_map;
1872 
1873 		if (!ops->ndo_set_config) {
1874 			err = -EOPNOTSUPP;
1875 			goto errout;
1876 		}
1877 
1878 		if (!netif_device_present(dev)) {
1879 			err = -ENODEV;
1880 			goto errout;
1881 		}
1882 
1883 		u_map = nla_data(tb[IFLA_MAP]);
1884 		k_map.mem_start = (unsigned long) u_map->mem_start;
1885 		k_map.mem_end = (unsigned long) u_map->mem_end;
1886 		k_map.base_addr = (unsigned short) u_map->base_addr;
1887 		k_map.irq = (unsigned char) u_map->irq;
1888 		k_map.dma = (unsigned char) u_map->dma;
1889 		k_map.port = (unsigned char) u_map->port;
1890 
1891 		err = ops->ndo_set_config(dev, &k_map);
1892 		if (err < 0)
1893 			goto errout;
1894 
1895 		status |= DO_SETLINK_NOTIFY;
1896 	}
1897 
1898 	if (tb[IFLA_ADDRESS]) {
1899 		struct sockaddr *sa;
1900 		int len;
1901 
1902 		len = sizeof(sa_family_t) + dev->addr_len;
1903 		sa = kmalloc(len, GFP_KERNEL);
1904 		if (!sa) {
1905 			err = -ENOMEM;
1906 			goto errout;
1907 		}
1908 		sa->sa_family = dev->type;
1909 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1910 		       dev->addr_len);
1911 		err = dev_set_mac_address(dev, sa);
1912 		kfree(sa);
1913 		if (err)
1914 			goto errout;
1915 		status |= DO_SETLINK_MODIFIED;
1916 	}
1917 
1918 	if (tb[IFLA_MTU]) {
1919 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1920 		if (err < 0)
1921 			goto errout;
1922 		status |= DO_SETLINK_MODIFIED;
1923 	}
1924 
1925 	if (tb[IFLA_GROUP]) {
1926 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1927 		status |= DO_SETLINK_NOTIFY;
1928 	}
1929 
1930 	/*
1931 	 * Interface selected by interface index but interface
1932 	 * name provided implies that a name change has been
1933 	 * requested.
1934 	 */
1935 	if (ifm->ifi_index > 0 && ifname[0]) {
1936 		err = dev_change_name(dev, ifname);
1937 		if (err < 0)
1938 			goto errout;
1939 		status |= DO_SETLINK_MODIFIED;
1940 	}
1941 
1942 	if (tb[IFLA_IFALIAS]) {
1943 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1944 				    nla_len(tb[IFLA_IFALIAS]));
1945 		if (err < 0)
1946 			goto errout;
1947 		status |= DO_SETLINK_NOTIFY;
1948 	}
1949 
1950 	if (tb[IFLA_BROADCAST]) {
1951 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1952 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1953 	}
1954 
1955 	if (ifm->ifi_flags || ifm->ifi_change) {
1956 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1957 		if (err < 0)
1958 			goto errout;
1959 	}
1960 
1961 	if (tb[IFLA_MASTER]) {
1962 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1963 		if (err)
1964 			goto errout;
1965 		status |= DO_SETLINK_MODIFIED;
1966 	}
1967 
1968 	if (tb[IFLA_CARRIER]) {
1969 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1970 		if (err)
1971 			goto errout;
1972 		status |= DO_SETLINK_MODIFIED;
1973 	}
1974 
1975 	if (tb[IFLA_TXQLEN]) {
1976 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1977 		unsigned long orig_len = dev->tx_queue_len;
1978 
1979 		if (dev->tx_queue_len ^ value) {
1980 			dev->tx_queue_len = value;
1981 			err = call_netdevice_notifiers(
1982 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
1983 			err = notifier_to_errno(err);
1984 			if (err) {
1985 				dev->tx_queue_len = orig_len;
1986 				goto errout;
1987 			}
1988 			status |= DO_SETLINK_NOTIFY;
1989 		}
1990 	}
1991 
1992 	if (tb[IFLA_OPERSTATE])
1993 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1994 
1995 	if (tb[IFLA_LINKMODE]) {
1996 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1997 
1998 		write_lock_bh(&dev_base_lock);
1999 		if (dev->link_mode ^ value)
2000 			status |= DO_SETLINK_NOTIFY;
2001 		dev->link_mode = value;
2002 		write_unlock_bh(&dev_base_lock);
2003 	}
2004 
2005 	if (tb[IFLA_VFINFO_LIST]) {
2006 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2007 		struct nlattr *attr;
2008 		int rem;
2009 
2010 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2011 			if (nla_type(attr) != IFLA_VF_INFO ||
2012 			    nla_len(attr) < NLA_HDRLEN) {
2013 				err = -EINVAL;
2014 				goto errout;
2015 			}
2016 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2017 					       ifla_vf_policy);
2018 			if (err < 0)
2019 				goto errout;
2020 			err = do_setvfinfo(dev, vfinfo);
2021 			if (err < 0)
2022 				goto errout;
2023 			status |= DO_SETLINK_NOTIFY;
2024 		}
2025 	}
2026 	err = 0;
2027 
2028 	if (tb[IFLA_VF_PORTS]) {
2029 		struct nlattr *port[IFLA_PORT_MAX+1];
2030 		struct nlattr *attr;
2031 		int vf;
2032 		int rem;
2033 
2034 		err = -EOPNOTSUPP;
2035 		if (!ops->ndo_set_vf_port)
2036 			goto errout;
2037 
2038 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2039 			if (nla_type(attr) != IFLA_VF_PORT ||
2040 			    nla_len(attr) < NLA_HDRLEN) {
2041 				err = -EINVAL;
2042 				goto errout;
2043 			}
2044 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2045 					       ifla_port_policy);
2046 			if (err < 0)
2047 				goto errout;
2048 			if (!port[IFLA_PORT_VF]) {
2049 				err = -EOPNOTSUPP;
2050 				goto errout;
2051 			}
2052 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2053 			err = ops->ndo_set_vf_port(dev, vf, port);
2054 			if (err < 0)
2055 				goto errout;
2056 			status |= DO_SETLINK_NOTIFY;
2057 		}
2058 	}
2059 	err = 0;
2060 
2061 	if (tb[IFLA_PORT_SELF]) {
2062 		struct nlattr *port[IFLA_PORT_MAX+1];
2063 
2064 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2065 			tb[IFLA_PORT_SELF], ifla_port_policy);
2066 		if (err < 0)
2067 			goto errout;
2068 
2069 		err = -EOPNOTSUPP;
2070 		if (ops->ndo_set_vf_port)
2071 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2072 		if (err < 0)
2073 			goto errout;
2074 		status |= DO_SETLINK_NOTIFY;
2075 	}
2076 
2077 	if (tb[IFLA_AF_SPEC]) {
2078 		struct nlattr *af;
2079 		int rem;
2080 
2081 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2082 			const struct rtnl_af_ops *af_ops;
2083 
2084 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2085 				BUG();
2086 
2087 			err = af_ops->set_link_af(dev, af);
2088 			if (err < 0)
2089 				goto errout;
2090 
2091 			status |= DO_SETLINK_NOTIFY;
2092 		}
2093 	}
2094 	err = 0;
2095 
2096 	if (tb[IFLA_PROTO_DOWN]) {
2097 		err = dev_change_proto_down(dev,
2098 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2099 		if (err)
2100 			goto errout;
2101 		status |= DO_SETLINK_NOTIFY;
2102 	}
2103 
2104 	if (tb[IFLA_XDP]) {
2105 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2106 
2107 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2108 				       ifla_xdp_policy);
2109 		if (err < 0)
2110 			goto errout;
2111 
2112 		if (xdp[IFLA_XDP_FD]) {
2113 			err = dev_change_xdp_fd(dev,
2114 						nla_get_s32(xdp[IFLA_XDP_FD]));
2115 			if (err)
2116 				goto errout;
2117 			status |= DO_SETLINK_NOTIFY;
2118 		}
2119 	}
2120 
2121 errout:
2122 	if (status & DO_SETLINK_MODIFIED) {
2123 		if (status & DO_SETLINK_NOTIFY)
2124 			netdev_state_change(dev);
2125 
2126 		if (err < 0)
2127 			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",
2128 					     dev->name);
2129 	}
2130 
2131 	return err;
2132 }
2133 
2134 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2135 {
2136 	struct net *net = sock_net(skb->sk);
2137 	struct ifinfomsg *ifm;
2138 	struct net_device *dev;
2139 	int err;
2140 	struct nlattr *tb[IFLA_MAX+1];
2141 	char ifname[IFNAMSIZ];
2142 
2143 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2144 	if (err < 0)
2145 		goto errout;
2146 
2147 	if (tb[IFLA_IFNAME])
2148 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2149 	else
2150 		ifname[0] = '\0';
2151 
2152 	err = -EINVAL;
2153 	ifm = nlmsg_data(nlh);
2154 	if (ifm->ifi_index > 0)
2155 		dev = __dev_get_by_index(net, ifm->ifi_index);
2156 	else if (tb[IFLA_IFNAME])
2157 		dev = __dev_get_by_name(net, ifname);
2158 	else
2159 		goto errout;
2160 
2161 	if (dev == NULL) {
2162 		err = -ENODEV;
2163 		goto errout;
2164 	}
2165 
2166 	err = validate_linkmsg(dev, tb);
2167 	if (err < 0)
2168 		goto errout;
2169 
2170 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2171 errout:
2172 	return err;
2173 }
2174 
2175 static int rtnl_group_dellink(const struct net *net, int group)
2176 {
2177 	struct net_device *dev, *aux;
2178 	LIST_HEAD(list_kill);
2179 	bool found = false;
2180 
2181 	if (!group)
2182 		return -EPERM;
2183 
2184 	for_each_netdev(net, dev) {
2185 		if (dev->group == group) {
2186 			const struct rtnl_link_ops *ops;
2187 
2188 			found = true;
2189 			ops = dev->rtnl_link_ops;
2190 			if (!ops || !ops->dellink)
2191 				return -EOPNOTSUPP;
2192 		}
2193 	}
2194 
2195 	if (!found)
2196 		return -ENODEV;
2197 
2198 	for_each_netdev_safe(net, dev, aux) {
2199 		if (dev->group == group) {
2200 			const struct rtnl_link_ops *ops;
2201 
2202 			ops = dev->rtnl_link_ops;
2203 			ops->dellink(dev, &list_kill);
2204 		}
2205 	}
2206 	unregister_netdevice_many(&list_kill);
2207 
2208 	return 0;
2209 }
2210 
2211 int rtnl_delete_link(struct net_device *dev)
2212 {
2213 	const struct rtnl_link_ops *ops;
2214 	LIST_HEAD(list_kill);
2215 
2216 	ops = dev->rtnl_link_ops;
2217 	if (!ops || !ops->dellink)
2218 		return -EOPNOTSUPP;
2219 
2220 	ops->dellink(dev, &list_kill);
2221 	unregister_netdevice_many(&list_kill);
2222 
2223 	return 0;
2224 }
2225 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2226 
2227 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2228 {
2229 	struct net *net = sock_net(skb->sk);
2230 	struct net_device *dev;
2231 	struct ifinfomsg *ifm;
2232 	char ifname[IFNAMSIZ];
2233 	struct nlattr *tb[IFLA_MAX+1];
2234 	int err;
2235 
2236 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2237 	if (err < 0)
2238 		return err;
2239 
2240 	if (tb[IFLA_IFNAME])
2241 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2242 
2243 	ifm = nlmsg_data(nlh);
2244 	if (ifm->ifi_index > 0)
2245 		dev = __dev_get_by_index(net, ifm->ifi_index);
2246 	else if (tb[IFLA_IFNAME])
2247 		dev = __dev_get_by_name(net, ifname);
2248 	else if (tb[IFLA_GROUP])
2249 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2250 	else
2251 		return -EINVAL;
2252 
2253 	if (!dev)
2254 		return -ENODEV;
2255 
2256 	return rtnl_delete_link(dev);
2257 }
2258 
2259 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2260 {
2261 	unsigned int old_flags;
2262 	int err;
2263 
2264 	old_flags = dev->flags;
2265 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2266 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2267 		if (err < 0)
2268 			return err;
2269 	}
2270 
2271 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2272 
2273 	__dev_notify_flags(dev, old_flags, ~0U);
2274 	return 0;
2275 }
2276 EXPORT_SYMBOL(rtnl_configure_link);
2277 
2278 struct net_device *rtnl_create_link(struct net *net,
2279 	const char *ifname, unsigned char name_assign_type,
2280 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2281 {
2282 	int err;
2283 	struct net_device *dev;
2284 	unsigned int num_tx_queues = 1;
2285 	unsigned int num_rx_queues = 1;
2286 
2287 	if (tb[IFLA_NUM_TX_QUEUES])
2288 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2289 	else if (ops->get_num_tx_queues)
2290 		num_tx_queues = ops->get_num_tx_queues();
2291 
2292 	if (tb[IFLA_NUM_RX_QUEUES])
2293 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2294 	else if (ops->get_num_rx_queues)
2295 		num_rx_queues = ops->get_num_rx_queues();
2296 
2297 	err = -ENOMEM;
2298 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2299 			       ops->setup, num_tx_queues, num_rx_queues);
2300 	if (!dev)
2301 		goto err;
2302 
2303 	dev_net_set(dev, net);
2304 	dev->rtnl_link_ops = ops;
2305 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2306 
2307 	if (tb[IFLA_MTU])
2308 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2309 	if (tb[IFLA_ADDRESS]) {
2310 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2311 				nla_len(tb[IFLA_ADDRESS]));
2312 		dev->addr_assign_type = NET_ADDR_SET;
2313 	}
2314 	if (tb[IFLA_BROADCAST])
2315 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2316 				nla_len(tb[IFLA_BROADCAST]));
2317 	if (tb[IFLA_TXQLEN])
2318 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2319 	if (tb[IFLA_OPERSTATE])
2320 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2321 	if (tb[IFLA_LINKMODE])
2322 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2323 	if (tb[IFLA_GROUP])
2324 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2325 
2326 	return dev;
2327 
2328 err:
2329 	return ERR_PTR(err);
2330 }
2331 EXPORT_SYMBOL(rtnl_create_link);
2332 
2333 static int rtnl_group_changelink(const struct sk_buff *skb,
2334 		struct net *net, int group,
2335 		struct ifinfomsg *ifm,
2336 		struct nlattr **tb)
2337 {
2338 	struct net_device *dev, *aux;
2339 	int err;
2340 
2341 	for_each_netdev_safe(net, dev, aux) {
2342 		if (dev->group == group) {
2343 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2344 			if (err < 0)
2345 				return err;
2346 		}
2347 	}
2348 
2349 	return 0;
2350 }
2351 
2352 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2353 {
2354 	struct net *net = sock_net(skb->sk);
2355 	const struct rtnl_link_ops *ops;
2356 	const struct rtnl_link_ops *m_ops = NULL;
2357 	struct net_device *dev;
2358 	struct net_device *master_dev = NULL;
2359 	struct ifinfomsg *ifm;
2360 	char kind[MODULE_NAME_LEN];
2361 	char ifname[IFNAMSIZ];
2362 	struct nlattr *tb[IFLA_MAX+1];
2363 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2364 	unsigned char name_assign_type = NET_NAME_USER;
2365 	int err;
2366 
2367 #ifdef CONFIG_MODULES
2368 replay:
2369 #endif
2370 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2371 	if (err < 0)
2372 		return err;
2373 
2374 	if (tb[IFLA_IFNAME])
2375 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2376 	else
2377 		ifname[0] = '\0';
2378 
2379 	ifm = nlmsg_data(nlh);
2380 	if (ifm->ifi_index > 0)
2381 		dev = __dev_get_by_index(net, ifm->ifi_index);
2382 	else {
2383 		if (ifname[0])
2384 			dev = __dev_get_by_name(net, ifname);
2385 		else
2386 			dev = NULL;
2387 	}
2388 
2389 	if (dev) {
2390 		master_dev = netdev_master_upper_dev_get(dev);
2391 		if (master_dev)
2392 			m_ops = master_dev->rtnl_link_ops;
2393 	}
2394 
2395 	err = validate_linkmsg(dev, tb);
2396 	if (err < 0)
2397 		return err;
2398 
2399 	if (tb[IFLA_LINKINFO]) {
2400 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2401 				       tb[IFLA_LINKINFO], ifla_info_policy);
2402 		if (err < 0)
2403 			return err;
2404 	} else
2405 		memset(linkinfo, 0, sizeof(linkinfo));
2406 
2407 	if (linkinfo[IFLA_INFO_KIND]) {
2408 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2409 		ops = rtnl_link_ops_get(kind);
2410 	} else {
2411 		kind[0] = '\0';
2412 		ops = NULL;
2413 	}
2414 
2415 	if (1) {
2416 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2417 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2418 		struct nlattr **data = NULL;
2419 		struct nlattr **slave_data = NULL;
2420 		struct net *dest_net, *link_net = NULL;
2421 
2422 		if (ops) {
2423 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2424 				err = nla_parse_nested(attr, ops->maxtype,
2425 						       linkinfo[IFLA_INFO_DATA],
2426 						       ops->policy);
2427 				if (err < 0)
2428 					return err;
2429 				data = attr;
2430 			}
2431 			if (ops->validate) {
2432 				err = ops->validate(tb, data);
2433 				if (err < 0)
2434 					return err;
2435 			}
2436 		}
2437 
2438 		if (m_ops) {
2439 			if (m_ops->slave_maxtype &&
2440 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2441 				err = nla_parse_nested(slave_attr,
2442 						       m_ops->slave_maxtype,
2443 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2444 						       m_ops->slave_policy);
2445 				if (err < 0)
2446 					return err;
2447 				slave_data = slave_attr;
2448 			}
2449 			if (m_ops->slave_validate) {
2450 				err = m_ops->slave_validate(tb, slave_data);
2451 				if (err < 0)
2452 					return err;
2453 			}
2454 		}
2455 
2456 		if (dev) {
2457 			int status = 0;
2458 
2459 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2460 				return -EEXIST;
2461 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2462 				return -EOPNOTSUPP;
2463 
2464 			if (linkinfo[IFLA_INFO_DATA]) {
2465 				if (!ops || ops != dev->rtnl_link_ops ||
2466 				    !ops->changelink)
2467 					return -EOPNOTSUPP;
2468 
2469 				err = ops->changelink(dev, tb, data);
2470 				if (err < 0)
2471 					return err;
2472 				status |= DO_SETLINK_NOTIFY;
2473 			}
2474 
2475 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2476 				if (!m_ops || !m_ops->slave_changelink)
2477 					return -EOPNOTSUPP;
2478 
2479 				err = m_ops->slave_changelink(master_dev, dev,
2480 							      tb, slave_data);
2481 				if (err < 0)
2482 					return err;
2483 				status |= DO_SETLINK_NOTIFY;
2484 			}
2485 
2486 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2487 		}
2488 
2489 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2490 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2491 				return rtnl_group_changelink(skb, net,
2492 						nla_get_u32(tb[IFLA_GROUP]),
2493 						ifm, tb);
2494 			return -ENODEV;
2495 		}
2496 
2497 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2498 			return -EOPNOTSUPP;
2499 
2500 		if (!ops) {
2501 #ifdef CONFIG_MODULES
2502 			if (kind[0]) {
2503 				__rtnl_unlock();
2504 				request_module("rtnl-link-%s", kind);
2505 				rtnl_lock();
2506 				ops = rtnl_link_ops_get(kind);
2507 				if (ops)
2508 					goto replay;
2509 			}
2510 #endif
2511 			return -EOPNOTSUPP;
2512 		}
2513 
2514 		if (!ops->setup)
2515 			return -EOPNOTSUPP;
2516 
2517 		if (!ifname[0]) {
2518 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2519 			name_assign_type = NET_NAME_ENUM;
2520 		}
2521 
2522 		dest_net = rtnl_link_get_net(net, tb);
2523 		if (IS_ERR(dest_net))
2524 			return PTR_ERR(dest_net);
2525 
2526 		err = -EPERM;
2527 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2528 			goto out;
2529 
2530 		if (tb[IFLA_LINK_NETNSID]) {
2531 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2532 
2533 			link_net = get_net_ns_by_id(dest_net, id);
2534 			if (!link_net) {
2535 				err =  -EINVAL;
2536 				goto out;
2537 			}
2538 			err = -EPERM;
2539 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2540 				goto out;
2541 		}
2542 
2543 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2544 				       name_assign_type, ops, tb);
2545 		if (IS_ERR(dev)) {
2546 			err = PTR_ERR(dev);
2547 			goto out;
2548 		}
2549 
2550 		dev->ifindex = ifm->ifi_index;
2551 
2552 		if (ops->newlink) {
2553 			err = ops->newlink(link_net ? : net, dev, tb, data);
2554 			/* Drivers should call free_netdev() in ->destructor
2555 			 * and unregister it on failure after registration
2556 			 * so that device could be finally freed in rtnl_unlock.
2557 			 */
2558 			if (err < 0) {
2559 				/* If device is not registered at all, free it now */
2560 				if (dev->reg_state == NETREG_UNINITIALIZED)
2561 					free_netdev(dev);
2562 				goto out;
2563 			}
2564 		} else {
2565 			err = register_netdevice(dev);
2566 			if (err < 0) {
2567 				free_netdev(dev);
2568 				goto out;
2569 			}
2570 		}
2571 		err = rtnl_configure_link(dev, ifm);
2572 		if (err < 0)
2573 			goto out_unregister;
2574 		if (link_net) {
2575 			err = dev_change_net_namespace(dev, dest_net, ifname);
2576 			if (err < 0)
2577 				goto out_unregister;
2578 		}
2579 out:
2580 		if (link_net)
2581 			put_net(link_net);
2582 		put_net(dest_net);
2583 		return err;
2584 out_unregister:
2585 		if (ops->newlink) {
2586 			LIST_HEAD(list_kill);
2587 
2588 			ops->dellink(dev, &list_kill);
2589 			unregister_netdevice_many(&list_kill);
2590 		} else {
2591 			unregister_netdevice(dev);
2592 		}
2593 		goto out;
2594 	}
2595 }
2596 
2597 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2598 {
2599 	struct net *net = sock_net(skb->sk);
2600 	struct ifinfomsg *ifm;
2601 	char ifname[IFNAMSIZ];
2602 	struct nlattr *tb[IFLA_MAX+1];
2603 	struct net_device *dev = NULL;
2604 	struct sk_buff *nskb;
2605 	int err;
2606 	u32 ext_filter_mask = 0;
2607 
2608 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2609 	if (err < 0)
2610 		return err;
2611 
2612 	if (tb[IFLA_IFNAME])
2613 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2614 
2615 	if (tb[IFLA_EXT_MASK])
2616 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2617 
2618 	ifm = nlmsg_data(nlh);
2619 	if (ifm->ifi_index > 0)
2620 		dev = __dev_get_by_index(net, ifm->ifi_index);
2621 	else if (tb[IFLA_IFNAME])
2622 		dev = __dev_get_by_name(net, ifname);
2623 	else
2624 		return -EINVAL;
2625 
2626 	if (dev == NULL)
2627 		return -ENODEV;
2628 
2629 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2630 	if (nskb == NULL)
2631 		return -ENOBUFS;
2632 
2633 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2634 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2635 	if (err < 0) {
2636 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2637 		WARN_ON(err == -EMSGSIZE);
2638 		kfree_skb(nskb);
2639 	} else
2640 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2641 
2642 	return err;
2643 }
2644 
2645 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2646 {
2647 	struct net *net = sock_net(skb->sk);
2648 	struct net_device *dev;
2649 	struct nlattr *tb[IFLA_MAX+1];
2650 	u32 ext_filter_mask = 0;
2651 	u16 min_ifinfo_dump_size = 0;
2652 	int hdrlen;
2653 
2654 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2655 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2656 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2657 
2658 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2659 		if (tb[IFLA_EXT_MASK])
2660 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2661 	}
2662 
2663 	if (!ext_filter_mask)
2664 		return NLMSG_GOODSIZE;
2665 	/*
2666 	 * traverse the list of net devices and compute the minimum
2667 	 * buffer size based upon the filter mask.
2668 	 */
2669 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2670 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2671 					     if_nlmsg_size(dev,
2672 						           ext_filter_mask));
2673 	}
2674 
2675 	return min_ifinfo_dump_size;
2676 }
2677 
2678 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2679 {
2680 	int idx;
2681 	int s_idx = cb->family;
2682 
2683 	if (s_idx == 0)
2684 		s_idx = 1;
2685 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2686 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2687 		if (idx < s_idx || idx == PF_PACKET)
2688 			continue;
2689 		if (rtnl_msg_handlers[idx] == NULL ||
2690 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2691 			continue;
2692 		if (idx > s_idx) {
2693 			memset(&cb->args[0], 0, sizeof(cb->args));
2694 			cb->prev_seq = 0;
2695 			cb->seq = 0;
2696 		}
2697 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2698 			break;
2699 	}
2700 	cb->family = idx;
2701 
2702 	return skb->len;
2703 }
2704 
2705 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2706 				       unsigned int change, gfp_t flags)
2707 {
2708 	struct net *net = dev_net(dev);
2709 	struct sk_buff *skb;
2710 	int err = -ENOBUFS;
2711 	size_t if_info_size;
2712 
2713 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2714 	if (skb == NULL)
2715 		goto errout;
2716 
2717 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2718 	if (err < 0) {
2719 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2720 		WARN_ON(err == -EMSGSIZE);
2721 		kfree_skb(skb);
2722 		goto errout;
2723 	}
2724 	return skb;
2725 errout:
2726 	if (err < 0)
2727 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2728 	return NULL;
2729 }
2730 
2731 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2732 {
2733 	struct net *net = dev_net(dev);
2734 
2735 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2736 }
2737 
2738 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2739 		  gfp_t flags)
2740 {
2741 	struct sk_buff *skb;
2742 
2743 	if (dev->reg_state != NETREG_REGISTERED)
2744 		return;
2745 
2746 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2747 	if (skb)
2748 		rtmsg_ifinfo_send(skb, dev, flags);
2749 }
2750 EXPORT_SYMBOL(rtmsg_ifinfo);
2751 
2752 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2753 				   struct net_device *dev,
2754 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2755 				   int type, unsigned int flags,
2756 				   int nlflags, u16 ndm_state)
2757 {
2758 	struct nlmsghdr *nlh;
2759 	struct ndmsg *ndm;
2760 
2761 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2762 	if (!nlh)
2763 		return -EMSGSIZE;
2764 
2765 	ndm = nlmsg_data(nlh);
2766 	ndm->ndm_family  = AF_BRIDGE;
2767 	ndm->ndm_pad1	 = 0;
2768 	ndm->ndm_pad2    = 0;
2769 	ndm->ndm_flags	 = flags;
2770 	ndm->ndm_type	 = 0;
2771 	ndm->ndm_ifindex = dev->ifindex;
2772 	ndm->ndm_state   = ndm_state;
2773 
2774 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2775 		goto nla_put_failure;
2776 	if (vid)
2777 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2778 			goto nla_put_failure;
2779 
2780 	nlmsg_end(skb, nlh);
2781 	return 0;
2782 
2783 nla_put_failure:
2784 	nlmsg_cancel(skb, nlh);
2785 	return -EMSGSIZE;
2786 }
2787 
2788 static inline size_t rtnl_fdb_nlmsg_size(void)
2789 {
2790 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2791 }
2792 
2793 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2794 			    u16 ndm_state)
2795 {
2796 	struct net *net = dev_net(dev);
2797 	struct sk_buff *skb;
2798 	int err = -ENOBUFS;
2799 
2800 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2801 	if (!skb)
2802 		goto errout;
2803 
2804 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2805 				      0, 0, type, NTF_SELF, 0, ndm_state);
2806 	if (err < 0) {
2807 		kfree_skb(skb);
2808 		goto errout;
2809 	}
2810 
2811 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2812 	return;
2813 errout:
2814 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2815 }
2816 
2817 /**
2818  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2819  */
2820 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2821 		     struct nlattr *tb[],
2822 		     struct net_device *dev,
2823 		     const unsigned char *addr, u16 vid,
2824 		     u16 flags)
2825 {
2826 	int err = -EINVAL;
2827 
2828 	/* If aging addresses are supported device will need to
2829 	 * implement its own handler for this.
2830 	 */
2831 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2832 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2833 		return err;
2834 	}
2835 
2836 	if (vid) {
2837 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2838 		return err;
2839 	}
2840 
2841 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2842 		err = dev_uc_add_excl(dev, addr);
2843 	else if (is_multicast_ether_addr(addr))
2844 		err = dev_mc_add_excl(dev, addr);
2845 
2846 	/* Only return duplicate errors if NLM_F_EXCL is set */
2847 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2848 		err = 0;
2849 
2850 	return err;
2851 }
2852 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2853 
2854 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2855 {
2856 	u16 vid = 0;
2857 
2858 	if (vlan_attr) {
2859 		if (nla_len(vlan_attr) != sizeof(u16)) {
2860 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2861 			return -EINVAL;
2862 		}
2863 
2864 		vid = nla_get_u16(vlan_attr);
2865 
2866 		if (!vid || vid >= VLAN_VID_MASK) {
2867 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2868 				vid);
2869 			return -EINVAL;
2870 		}
2871 	}
2872 	*p_vid = vid;
2873 	return 0;
2874 }
2875 
2876 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2877 {
2878 	struct net *net = sock_net(skb->sk);
2879 	struct ndmsg *ndm;
2880 	struct nlattr *tb[NDA_MAX+1];
2881 	struct net_device *dev;
2882 	u8 *addr;
2883 	u16 vid;
2884 	int err;
2885 
2886 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2887 	if (err < 0)
2888 		return err;
2889 
2890 	ndm = nlmsg_data(nlh);
2891 	if (ndm->ndm_ifindex == 0) {
2892 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2893 		return -EINVAL;
2894 	}
2895 
2896 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2897 	if (dev == NULL) {
2898 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2899 		return -ENODEV;
2900 	}
2901 
2902 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2903 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2904 		return -EINVAL;
2905 	}
2906 
2907 	addr = nla_data(tb[NDA_LLADDR]);
2908 
2909 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2910 	if (err)
2911 		return err;
2912 
2913 	err = -EOPNOTSUPP;
2914 
2915 	/* Support fdb on master device the net/bridge default case */
2916 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2917 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2918 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2919 		const struct net_device_ops *ops = br_dev->netdev_ops;
2920 
2921 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2922 				       nlh->nlmsg_flags);
2923 		if (err)
2924 			goto out;
2925 		else
2926 			ndm->ndm_flags &= ~NTF_MASTER;
2927 	}
2928 
2929 	/* Embedded bridge, macvlan, and any other device support */
2930 	if ((ndm->ndm_flags & NTF_SELF)) {
2931 		if (dev->netdev_ops->ndo_fdb_add)
2932 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2933 							   vid,
2934 							   nlh->nlmsg_flags);
2935 		else
2936 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2937 					       nlh->nlmsg_flags);
2938 
2939 		if (!err) {
2940 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2941 					ndm->ndm_state);
2942 			ndm->ndm_flags &= ~NTF_SELF;
2943 		}
2944 	}
2945 out:
2946 	return err;
2947 }
2948 
2949 /**
2950  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2951  */
2952 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2953 		     struct nlattr *tb[],
2954 		     struct net_device *dev,
2955 		     const unsigned char *addr, u16 vid)
2956 {
2957 	int err = -EINVAL;
2958 
2959 	/* If aging addresses are supported device will need to
2960 	 * implement its own handler for this.
2961 	 */
2962 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2963 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2964 		return err;
2965 	}
2966 
2967 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2968 		err = dev_uc_del(dev, addr);
2969 	else if (is_multicast_ether_addr(addr))
2970 		err = dev_mc_del(dev, addr);
2971 
2972 	return err;
2973 }
2974 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2975 
2976 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2977 {
2978 	struct net *net = sock_net(skb->sk);
2979 	struct ndmsg *ndm;
2980 	struct nlattr *tb[NDA_MAX+1];
2981 	struct net_device *dev;
2982 	int err = -EINVAL;
2983 	__u8 *addr;
2984 	u16 vid;
2985 
2986 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2987 		return -EPERM;
2988 
2989 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2990 	if (err < 0)
2991 		return err;
2992 
2993 	ndm = nlmsg_data(nlh);
2994 	if (ndm->ndm_ifindex == 0) {
2995 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2996 		return -EINVAL;
2997 	}
2998 
2999 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3000 	if (dev == NULL) {
3001 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3002 		return -ENODEV;
3003 	}
3004 
3005 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3006 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3007 		return -EINVAL;
3008 	}
3009 
3010 	addr = nla_data(tb[NDA_LLADDR]);
3011 
3012 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3013 	if (err)
3014 		return err;
3015 
3016 	err = -EOPNOTSUPP;
3017 
3018 	/* Support fdb on master device the net/bridge default case */
3019 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3020 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3021 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3022 		const struct net_device_ops *ops = br_dev->netdev_ops;
3023 
3024 		if (ops->ndo_fdb_del)
3025 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3026 
3027 		if (err)
3028 			goto out;
3029 		else
3030 			ndm->ndm_flags &= ~NTF_MASTER;
3031 	}
3032 
3033 	/* Embedded bridge, macvlan, and any other device support */
3034 	if (ndm->ndm_flags & NTF_SELF) {
3035 		if (dev->netdev_ops->ndo_fdb_del)
3036 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3037 							   vid);
3038 		else
3039 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3040 
3041 		if (!err) {
3042 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3043 					ndm->ndm_state);
3044 			ndm->ndm_flags &= ~NTF_SELF;
3045 		}
3046 	}
3047 out:
3048 	return err;
3049 }
3050 
3051 static int nlmsg_populate_fdb(struct sk_buff *skb,
3052 			      struct netlink_callback *cb,
3053 			      struct net_device *dev,
3054 			      int *idx,
3055 			      struct netdev_hw_addr_list *list)
3056 {
3057 	struct netdev_hw_addr *ha;
3058 	int err;
3059 	u32 portid, seq;
3060 
3061 	portid = NETLINK_CB(cb->skb).portid;
3062 	seq = cb->nlh->nlmsg_seq;
3063 
3064 	list_for_each_entry(ha, &list->list, list) {
3065 		if (*idx < cb->args[0])
3066 			goto skip;
3067 
3068 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3069 					      portid, seq,
3070 					      RTM_NEWNEIGH, NTF_SELF,
3071 					      NLM_F_MULTI, NUD_PERMANENT);
3072 		if (err < 0)
3073 			return err;
3074 skip:
3075 		*idx += 1;
3076 	}
3077 	return 0;
3078 }
3079 
3080 /**
3081  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3082  * @nlh: netlink message header
3083  * @dev: netdevice
3084  *
3085  * Default netdevice operation to dump the existing unicast address list.
3086  * Returns number of addresses from list put in skb.
3087  */
3088 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3089 		      struct netlink_callback *cb,
3090 		      struct net_device *dev,
3091 		      struct net_device *filter_dev,
3092 		      int idx)
3093 {
3094 	int err;
3095 
3096 	netif_addr_lock_bh(dev);
3097 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
3098 	if (err)
3099 		goto out;
3100 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
3101 out:
3102 	netif_addr_unlock_bh(dev);
3103 	cb->args[1] = err;
3104 	return idx;
3105 }
3106 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3107 
3108 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3109 {
3110 	struct net_device *dev;
3111 	struct nlattr *tb[IFLA_MAX+1];
3112 	struct net_device *br_dev = NULL;
3113 	const struct net_device_ops *ops = NULL;
3114 	const struct net_device_ops *cops = NULL;
3115 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3116 	struct net *net = sock_net(skb->sk);
3117 	int brport_idx = 0;
3118 	int br_idx = 0;
3119 	int idx = 0;
3120 
3121 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3122 			ifla_policy) == 0) {
3123 		if (tb[IFLA_MASTER])
3124 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3125 	}
3126 
3127 	brport_idx = ifm->ifi_index;
3128 
3129 	if (br_idx) {
3130 		br_dev = __dev_get_by_index(net, br_idx);
3131 		if (!br_dev)
3132 			return -ENODEV;
3133 
3134 		ops = br_dev->netdev_ops;
3135 	}
3136 
3137 	cb->args[1] = 0;
3138 	for_each_netdev(net, dev) {
3139 		if (brport_idx && (dev->ifindex != brport_idx))
3140 			continue;
3141 
3142 		if (!br_idx) { /* user did not specify a specific bridge */
3143 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3144 				br_dev = netdev_master_upper_dev_get(dev);
3145 				cops = br_dev->netdev_ops;
3146 			}
3147 
3148 		} else {
3149 			if (dev != br_dev &&
3150 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
3151 				continue;
3152 
3153 			if (br_dev != netdev_master_upper_dev_get(dev) &&
3154 			    !(dev->priv_flags & IFF_EBRIDGE))
3155 				continue;
3156 
3157 			cops = ops;
3158 		}
3159 
3160 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
3161 			if (cops && cops->ndo_fdb_dump)
3162 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3163 							 idx);
3164 		}
3165 		if (cb->args[1] == -EMSGSIZE)
3166 			break;
3167 
3168 		if (dev->netdev_ops->ndo_fdb_dump)
3169 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3170 							    idx);
3171 		else
3172 			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3173 		if (cb->args[1] == -EMSGSIZE)
3174 			break;
3175 
3176 		cops = NULL;
3177 	}
3178 
3179 	cb->args[0] = idx;
3180 	return skb->len;
3181 }
3182 
3183 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3184 			       unsigned int attrnum, unsigned int flag)
3185 {
3186 	if (mask & flag)
3187 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3188 	return 0;
3189 }
3190 
3191 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3192 			    struct net_device *dev, u16 mode,
3193 			    u32 flags, u32 mask, int nlflags,
3194 			    u32 filter_mask,
3195 			    int (*vlan_fill)(struct sk_buff *skb,
3196 					     struct net_device *dev,
3197 					     u32 filter_mask))
3198 {
3199 	struct nlmsghdr *nlh;
3200 	struct ifinfomsg *ifm;
3201 	struct nlattr *br_afspec;
3202 	struct nlattr *protinfo;
3203 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3204 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3205 	int err = 0;
3206 
3207 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3208 	if (nlh == NULL)
3209 		return -EMSGSIZE;
3210 
3211 	ifm = nlmsg_data(nlh);
3212 	ifm->ifi_family = AF_BRIDGE;
3213 	ifm->__ifi_pad = 0;
3214 	ifm->ifi_type = dev->type;
3215 	ifm->ifi_index = dev->ifindex;
3216 	ifm->ifi_flags = dev_get_flags(dev);
3217 	ifm->ifi_change = 0;
3218 
3219 
3220 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3221 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3222 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3223 	    (br_dev &&
3224 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3225 	    (dev->addr_len &&
3226 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3227 	    (dev->ifindex != dev_get_iflink(dev) &&
3228 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3229 		goto nla_put_failure;
3230 
3231 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3232 	if (!br_afspec)
3233 		goto nla_put_failure;
3234 
3235 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3236 		nla_nest_cancel(skb, br_afspec);
3237 		goto nla_put_failure;
3238 	}
3239 
3240 	if (mode != BRIDGE_MODE_UNDEF) {
3241 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3242 			nla_nest_cancel(skb, br_afspec);
3243 			goto nla_put_failure;
3244 		}
3245 	}
3246 	if (vlan_fill) {
3247 		err = vlan_fill(skb, dev, filter_mask);
3248 		if (err) {
3249 			nla_nest_cancel(skb, br_afspec);
3250 			goto nla_put_failure;
3251 		}
3252 	}
3253 	nla_nest_end(skb, br_afspec);
3254 
3255 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3256 	if (!protinfo)
3257 		goto nla_put_failure;
3258 
3259 	if (brport_nla_put_flag(skb, flags, mask,
3260 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3261 	    brport_nla_put_flag(skb, flags, mask,
3262 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3263 	    brport_nla_put_flag(skb, flags, mask,
3264 				IFLA_BRPORT_FAST_LEAVE,
3265 				BR_MULTICAST_FAST_LEAVE) ||
3266 	    brport_nla_put_flag(skb, flags, mask,
3267 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3268 	    brport_nla_put_flag(skb, flags, mask,
3269 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3270 	    brport_nla_put_flag(skb, flags, mask,
3271 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3272 	    brport_nla_put_flag(skb, flags, mask,
3273 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3274 	    brport_nla_put_flag(skb, flags, mask,
3275 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3276 		nla_nest_cancel(skb, protinfo);
3277 		goto nla_put_failure;
3278 	}
3279 
3280 	nla_nest_end(skb, protinfo);
3281 
3282 	nlmsg_end(skb, nlh);
3283 	return 0;
3284 nla_put_failure:
3285 	nlmsg_cancel(skb, nlh);
3286 	return err ? err : -EMSGSIZE;
3287 }
3288 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3289 
3290 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3291 {
3292 	struct net *net = sock_net(skb->sk);
3293 	struct net_device *dev;
3294 	int idx = 0;
3295 	u32 portid = NETLINK_CB(cb->skb).portid;
3296 	u32 seq = cb->nlh->nlmsg_seq;
3297 	u32 filter_mask = 0;
3298 	int err;
3299 
3300 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3301 		struct nlattr *extfilt;
3302 
3303 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3304 					  IFLA_EXT_MASK);
3305 		if (extfilt) {
3306 			if (nla_len(extfilt) < sizeof(filter_mask))
3307 				return -EINVAL;
3308 
3309 			filter_mask = nla_get_u32(extfilt);
3310 		}
3311 	}
3312 
3313 	rcu_read_lock();
3314 	for_each_netdev_rcu(net, dev) {
3315 		const struct net_device_ops *ops = dev->netdev_ops;
3316 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3317 
3318 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3319 			if (idx >= cb->args[0]) {
3320 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3321 						skb, portid, seq, dev,
3322 						filter_mask, NLM_F_MULTI);
3323 				if (err < 0 && err != -EOPNOTSUPP)
3324 					break;
3325 			}
3326 			idx++;
3327 		}
3328 
3329 		if (ops->ndo_bridge_getlink) {
3330 			if (idx >= cb->args[0]) {
3331 				err = ops->ndo_bridge_getlink(skb, portid,
3332 							      seq, dev,
3333 							      filter_mask,
3334 							      NLM_F_MULTI);
3335 				if (err < 0 && err != -EOPNOTSUPP)
3336 					break;
3337 			}
3338 			idx++;
3339 		}
3340 	}
3341 	rcu_read_unlock();
3342 	cb->args[0] = idx;
3343 
3344 	return skb->len;
3345 }
3346 
3347 static inline size_t bridge_nlmsg_size(void)
3348 {
3349 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3350 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3351 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3352 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3353 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3354 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3355 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3356 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3357 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3358 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3359 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3360 }
3361 
3362 static int rtnl_bridge_notify(struct net_device *dev)
3363 {
3364 	struct net *net = dev_net(dev);
3365 	struct sk_buff *skb;
3366 	int err = -EOPNOTSUPP;
3367 
3368 	if (!dev->netdev_ops->ndo_bridge_getlink)
3369 		return 0;
3370 
3371 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3372 	if (!skb) {
3373 		err = -ENOMEM;
3374 		goto errout;
3375 	}
3376 
3377 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3378 	if (err < 0)
3379 		goto errout;
3380 
3381 	if (!skb->len)
3382 		goto errout;
3383 
3384 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3385 	return 0;
3386 errout:
3387 	WARN_ON(err == -EMSGSIZE);
3388 	kfree_skb(skb);
3389 	if (err)
3390 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3391 	return err;
3392 }
3393 
3394 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3395 {
3396 	struct net *net = sock_net(skb->sk);
3397 	struct ifinfomsg *ifm;
3398 	struct net_device *dev;
3399 	struct nlattr *br_spec, *attr = NULL;
3400 	int rem, err = -EOPNOTSUPP;
3401 	u16 flags = 0;
3402 	bool have_flags = false;
3403 
3404 	if (nlmsg_len(nlh) < sizeof(*ifm))
3405 		return -EINVAL;
3406 
3407 	ifm = nlmsg_data(nlh);
3408 	if (ifm->ifi_family != AF_BRIDGE)
3409 		return -EPFNOSUPPORT;
3410 
3411 	dev = __dev_get_by_index(net, ifm->ifi_index);
3412 	if (!dev) {
3413 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3414 		return -ENODEV;
3415 	}
3416 
3417 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3418 	if (br_spec) {
3419 		nla_for_each_nested(attr, br_spec, rem) {
3420 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3421 				if (nla_len(attr) < sizeof(flags))
3422 					return -EINVAL;
3423 
3424 				have_flags = true;
3425 				flags = nla_get_u16(attr);
3426 				break;
3427 			}
3428 		}
3429 	}
3430 
3431 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3432 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3433 
3434 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3435 			err = -EOPNOTSUPP;
3436 			goto out;
3437 		}
3438 
3439 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3440 		if (err)
3441 			goto out;
3442 
3443 		flags &= ~BRIDGE_FLAGS_MASTER;
3444 	}
3445 
3446 	if ((flags & BRIDGE_FLAGS_SELF)) {
3447 		if (!dev->netdev_ops->ndo_bridge_setlink)
3448 			err = -EOPNOTSUPP;
3449 		else
3450 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3451 								  flags);
3452 		if (!err) {
3453 			flags &= ~BRIDGE_FLAGS_SELF;
3454 
3455 			/* Generate event to notify upper layer of bridge
3456 			 * change
3457 			 */
3458 			err = rtnl_bridge_notify(dev);
3459 		}
3460 	}
3461 
3462 	if (have_flags)
3463 		memcpy(nla_data(attr), &flags, sizeof(flags));
3464 out:
3465 	return err;
3466 }
3467 
3468 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3469 {
3470 	struct net *net = sock_net(skb->sk);
3471 	struct ifinfomsg *ifm;
3472 	struct net_device *dev;
3473 	struct nlattr *br_spec, *attr = NULL;
3474 	int rem, err = -EOPNOTSUPP;
3475 	u16 flags = 0;
3476 	bool have_flags = false;
3477 
3478 	if (nlmsg_len(nlh) < sizeof(*ifm))
3479 		return -EINVAL;
3480 
3481 	ifm = nlmsg_data(nlh);
3482 	if (ifm->ifi_family != AF_BRIDGE)
3483 		return -EPFNOSUPPORT;
3484 
3485 	dev = __dev_get_by_index(net, ifm->ifi_index);
3486 	if (!dev) {
3487 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3488 		return -ENODEV;
3489 	}
3490 
3491 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3492 	if (br_spec) {
3493 		nla_for_each_nested(attr, br_spec, rem) {
3494 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3495 				if (nla_len(attr) < sizeof(flags))
3496 					return -EINVAL;
3497 
3498 				have_flags = true;
3499 				flags = nla_get_u16(attr);
3500 				break;
3501 			}
3502 		}
3503 	}
3504 
3505 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3506 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3507 
3508 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3509 			err = -EOPNOTSUPP;
3510 			goto out;
3511 		}
3512 
3513 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3514 		if (err)
3515 			goto out;
3516 
3517 		flags &= ~BRIDGE_FLAGS_MASTER;
3518 	}
3519 
3520 	if ((flags & BRIDGE_FLAGS_SELF)) {
3521 		if (!dev->netdev_ops->ndo_bridge_dellink)
3522 			err = -EOPNOTSUPP;
3523 		else
3524 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3525 								  flags);
3526 
3527 		if (!err) {
3528 			flags &= ~BRIDGE_FLAGS_SELF;
3529 
3530 			/* Generate event to notify upper layer of bridge
3531 			 * change
3532 			 */
3533 			err = rtnl_bridge_notify(dev);
3534 		}
3535 	}
3536 
3537 	if (have_flags)
3538 		memcpy(nla_data(attr), &flags, sizeof(flags));
3539 out:
3540 	return err;
3541 }
3542 
3543 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3544 {
3545 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3546 	       (!idxattr || idxattr == attrid);
3547 }
3548 
3549 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3550 			       int type, u32 pid, u32 seq, u32 change,
3551 			       unsigned int flags, unsigned int filter_mask,
3552 			       int *idxattr, int *prividx)
3553 {
3554 	struct if_stats_msg *ifsm;
3555 	struct nlmsghdr *nlh;
3556 	struct nlattr *attr;
3557 	int s_prividx = *prividx;
3558 
3559 	ASSERT_RTNL();
3560 
3561 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3562 	if (!nlh)
3563 		return -EMSGSIZE;
3564 
3565 	ifsm = nlmsg_data(nlh);
3566 	ifsm->ifindex = dev->ifindex;
3567 	ifsm->filter_mask = filter_mask;
3568 
3569 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3570 		struct rtnl_link_stats64 *sp;
3571 
3572 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3573 					 sizeof(struct rtnl_link_stats64),
3574 					 IFLA_STATS_UNSPEC);
3575 		if (!attr)
3576 			goto nla_put_failure;
3577 
3578 		sp = nla_data(attr);
3579 		dev_get_stats(dev, sp);
3580 	}
3581 
3582 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3583 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3584 
3585 		if (ops && ops->fill_linkxstats) {
3586 			int err;
3587 
3588 			*idxattr = IFLA_STATS_LINK_XSTATS;
3589 			attr = nla_nest_start(skb,
3590 					      IFLA_STATS_LINK_XSTATS);
3591 			if (!attr)
3592 				goto nla_put_failure;
3593 
3594 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3595 			nla_nest_end(skb, attr);
3596 			if (err)
3597 				goto nla_put_failure;
3598 			*idxattr = 0;
3599 		}
3600 	}
3601 
3602 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3603 			     *idxattr)) {
3604 		const struct rtnl_link_ops *ops = NULL;
3605 		const struct net_device *master;
3606 
3607 		master = netdev_master_upper_dev_get(dev);
3608 		if (master)
3609 			ops = master->rtnl_link_ops;
3610 		if (ops && ops->fill_linkxstats) {
3611 			int err;
3612 
3613 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3614 			attr = nla_nest_start(skb,
3615 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3616 			if (!attr)
3617 				goto nla_put_failure;
3618 
3619 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3620 			nla_nest_end(skb, attr);
3621 			if (err)
3622 				goto nla_put_failure;
3623 			*idxattr = 0;
3624 		}
3625 	}
3626 
3627 	nlmsg_end(skb, nlh);
3628 
3629 	return 0;
3630 
3631 nla_put_failure:
3632 	/* not a multi message or no progress mean a real error */
3633 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3634 		nlmsg_cancel(skb, nlh);
3635 	else
3636 		nlmsg_end(skb, nlh);
3637 
3638 	return -EMSGSIZE;
3639 }
3640 
3641 static const struct nla_policy ifla_stats_policy[IFLA_STATS_MAX + 1] = {
3642 	[IFLA_STATS_LINK_64]	= { .len = sizeof(struct rtnl_link_stats64) },
3643 };
3644 
3645 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3646 				  u32 filter_mask)
3647 {
3648 	size_t size = 0;
3649 
3650 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3651 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3652 
3653 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3654 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3655 		int attr = IFLA_STATS_LINK_XSTATS;
3656 
3657 		if (ops && ops->get_linkxstats_size) {
3658 			size += nla_total_size(ops->get_linkxstats_size(dev,
3659 									attr));
3660 			/* for IFLA_STATS_LINK_XSTATS */
3661 			size += nla_total_size(0);
3662 		}
3663 	}
3664 
3665 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3666 		struct net_device *_dev = (struct net_device *)dev;
3667 		const struct rtnl_link_ops *ops = NULL;
3668 		const struct net_device *master;
3669 
3670 		/* netdev_master_upper_dev_get can't take const */
3671 		master = netdev_master_upper_dev_get(_dev);
3672 		if (master)
3673 			ops = master->rtnl_link_ops;
3674 		if (ops && ops->get_linkxstats_size) {
3675 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3676 
3677 			size += nla_total_size(ops->get_linkxstats_size(dev,
3678 									attr));
3679 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
3680 			size += nla_total_size(0);
3681 		}
3682 	}
3683 
3684 	return size;
3685 }
3686 
3687 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3688 {
3689 	struct net *net = sock_net(skb->sk);
3690 	struct net_device *dev = NULL;
3691 	int idxattr = 0, prividx = 0;
3692 	struct if_stats_msg *ifsm;
3693 	struct sk_buff *nskb;
3694 	u32 filter_mask;
3695 	int err;
3696 
3697 	ifsm = nlmsg_data(nlh);
3698 	if (ifsm->ifindex > 0)
3699 		dev = __dev_get_by_index(net, ifsm->ifindex);
3700 	else
3701 		return -EINVAL;
3702 
3703 	if (!dev)
3704 		return -ENODEV;
3705 
3706 	filter_mask = ifsm->filter_mask;
3707 	if (!filter_mask)
3708 		return -EINVAL;
3709 
3710 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3711 	if (!nskb)
3712 		return -ENOBUFS;
3713 
3714 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3715 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3716 				  0, filter_mask, &idxattr, &prividx);
3717 	if (err < 0) {
3718 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3719 		WARN_ON(err == -EMSGSIZE);
3720 		kfree_skb(nskb);
3721 	} else {
3722 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3723 	}
3724 
3725 	return err;
3726 }
3727 
3728 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3729 {
3730 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
3731 	struct net *net = sock_net(skb->sk);
3732 	unsigned int flags = NLM_F_MULTI;
3733 	struct if_stats_msg *ifsm;
3734 	struct hlist_head *head;
3735 	struct net_device *dev;
3736 	u32 filter_mask = 0;
3737 	int idx = 0;
3738 
3739 	s_h = cb->args[0];
3740 	s_idx = cb->args[1];
3741 	s_idxattr = cb->args[2];
3742 	s_prividx = cb->args[3];
3743 
3744 	cb->seq = net->dev_base_seq;
3745 
3746 	ifsm = nlmsg_data(cb->nlh);
3747 	filter_mask = ifsm->filter_mask;
3748 	if (!filter_mask)
3749 		return -EINVAL;
3750 
3751 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3752 		idx = 0;
3753 		head = &net->dev_index_head[h];
3754 		hlist_for_each_entry(dev, head, index_hlist) {
3755 			if (idx < s_idx)
3756 				goto cont;
3757 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3758 						  NETLINK_CB(cb->skb).portid,
3759 						  cb->nlh->nlmsg_seq, 0,
3760 						  flags, filter_mask,
3761 						  &s_idxattr, &s_prividx);
3762 			/* If we ran out of room on the first message,
3763 			 * we're in trouble
3764 			 */
3765 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3766 
3767 			if (err < 0)
3768 				goto out;
3769 			s_prividx = 0;
3770 			s_idxattr = 0;
3771 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3772 cont:
3773 			idx++;
3774 		}
3775 	}
3776 out:
3777 	cb->args[3] = s_prividx;
3778 	cb->args[2] = s_idxattr;
3779 	cb->args[1] = idx;
3780 	cb->args[0] = h;
3781 
3782 	return skb->len;
3783 }
3784 
3785 /* Process one rtnetlink message. */
3786 
3787 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3788 {
3789 	struct net *net = sock_net(skb->sk);
3790 	rtnl_doit_func doit;
3791 	int kind;
3792 	int family;
3793 	int type;
3794 	int err;
3795 
3796 	type = nlh->nlmsg_type;
3797 	if (type > RTM_MAX)
3798 		return -EOPNOTSUPP;
3799 
3800 	type -= RTM_BASE;
3801 
3802 	/* All the messages must have at least 1 byte length */
3803 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3804 		return 0;
3805 
3806 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3807 	kind = type&3;
3808 
3809 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3810 		return -EPERM;
3811 
3812 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3813 		struct sock *rtnl;
3814 		rtnl_dumpit_func dumpit;
3815 		rtnl_calcit_func calcit;
3816 		u16 min_dump_alloc = 0;
3817 
3818 		dumpit = rtnl_get_dumpit(family, type);
3819 		if (dumpit == NULL)
3820 			return -EOPNOTSUPP;
3821 		calcit = rtnl_get_calcit(family, type);
3822 		if (calcit)
3823 			min_dump_alloc = calcit(skb, nlh);
3824 
3825 		__rtnl_unlock();
3826 		rtnl = net->rtnl;
3827 		{
3828 			struct netlink_dump_control c = {
3829 				.dump		= dumpit,
3830 				.min_dump_alloc	= min_dump_alloc,
3831 			};
3832 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3833 		}
3834 		rtnl_lock();
3835 		return err;
3836 	}
3837 
3838 	doit = rtnl_get_doit(family, type);
3839 	if (doit == NULL)
3840 		return -EOPNOTSUPP;
3841 
3842 	return doit(skb, nlh);
3843 }
3844 
3845 static void rtnetlink_rcv(struct sk_buff *skb)
3846 {
3847 	rtnl_lock();
3848 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3849 	rtnl_unlock();
3850 }
3851 
3852 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3853 {
3854 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3855 
3856 	switch (event) {
3857 	case NETDEV_UP:
3858 	case NETDEV_DOWN:
3859 	case NETDEV_PRE_UP:
3860 	case NETDEV_POST_INIT:
3861 	case NETDEV_REGISTER:
3862 	case NETDEV_CHANGE:
3863 	case NETDEV_PRE_TYPE_CHANGE:
3864 	case NETDEV_GOING_DOWN:
3865 	case NETDEV_UNREGISTER:
3866 	case NETDEV_UNREGISTER_FINAL:
3867 	case NETDEV_RELEASE:
3868 	case NETDEV_JOIN:
3869 	case NETDEV_BONDING_INFO:
3870 		break;
3871 	default:
3872 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3873 		break;
3874 	}
3875 	return NOTIFY_DONE;
3876 }
3877 
3878 static struct notifier_block rtnetlink_dev_notifier = {
3879 	.notifier_call	= rtnetlink_event,
3880 };
3881 
3882 
3883 static int __net_init rtnetlink_net_init(struct net *net)
3884 {
3885 	struct sock *sk;
3886 	struct netlink_kernel_cfg cfg = {
3887 		.groups		= RTNLGRP_MAX,
3888 		.input		= rtnetlink_rcv,
3889 		.cb_mutex	= &rtnl_mutex,
3890 		.flags		= NL_CFG_F_NONROOT_RECV,
3891 	};
3892 
3893 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3894 	if (!sk)
3895 		return -ENOMEM;
3896 	net->rtnl = sk;
3897 	return 0;
3898 }
3899 
3900 static void __net_exit rtnetlink_net_exit(struct net *net)
3901 {
3902 	netlink_kernel_release(net->rtnl);
3903 	net->rtnl = NULL;
3904 }
3905 
3906 static struct pernet_operations rtnetlink_net_ops = {
3907 	.init = rtnetlink_net_init,
3908 	.exit = rtnetlink_net_exit,
3909 };
3910 
3911 void __init rtnetlink_init(void)
3912 {
3913 	if (register_pernet_subsys(&rtnetlink_net_ops))
3914 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3915 
3916 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3917 
3918 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3919 		      rtnl_dump_ifinfo, rtnl_calcit);
3920 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3921 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3922 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3923 
3924 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3925 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3926 
3927 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3928 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3929 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3930 
3931 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3932 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3933 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3934 
3935 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
3936 		      NULL);
3937 }
3938