xref: /openbmc/linux/net/core/rtnetlink.c (revision bb598c1b8c9bf56981927dcb8c0dc34b8ff95342)
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 	rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
279 
280 	return 0;
281 }
282 EXPORT_SYMBOL_GPL(rtnl_unregister);
283 
284 /**
285  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
286  * @protocol : Protocol family or PF_UNSPEC
287  *
288  * Identical to calling rtnl_unregster() for all registered message types
289  * of a certain protocol family.
290  */
291 void rtnl_unregister_all(int protocol)
292 {
293 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
294 
295 	kfree(rtnl_msg_handlers[protocol]);
296 	rtnl_msg_handlers[protocol] = NULL;
297 }
298 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
299 
300 static LIST_HEAD(link_ops);
301 
302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
303 {
304 	const struct rtnl_link_ops *ops;
305 
306 	list_for_each_entry(ops, &link_ops, list) {
307 		if (!strcmp(ops->kind, kind))
308 			return ops;
309 	}
310 	return NULL;
311 }
312 
313 /**
314  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
315  * @ops: struct rtnl_link_ops * to register
316  *
317  * The caller must hold the rtnl_mutex. This function should be used
318  * by drivers that create devices during module initialization. It
319  * must be called before registering the devices.
320  *
321  * Returns 0 on success or a negative error code.
322  */
323 int __rtnl_link_register(struct rtnl_link_ops *ops)
324 {
325 	if (rtnl_link_ops_get(ops->kind))
326 		return -EEXIST;
327 
328 	/* The check for setup is here because if ops
329 	 * does not have that filled up, it is not possible
330 	 * to use the ops for creating device. So do not
331 	 * fill up dellink as well. That disables rtnl_dellink.
332 	 */
333 	if (ops->setup && !ops->dellink)
334 		ops->dellink = unregister_netdevice_queue;
335 
336 	list_add_tail(&ops->list, &link_ops);
337 	return 0;
338 }
339 EXPORT_SYMBOL_GPL(__rtnl_link_register);
340 
341 /**
342  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
343  * @ops: struct rtnl_link_ops * to register
344  *
345  * Returns 0 on success or a negative error code.
346  */
347 int rtnl_link_register(struct rtnl_link_ops *ops)
348 {
349 	int err;
350 
351 	rtnl_lock();
352 	err = __rtnl_link_register(ops);
353 	rtnl_unlock();
354 	return err;
355 }
356 EXPORT_SYMBOL_GPL(rtnl_link_register);
357 
358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
359 {
360 	struct net_device *dev;
361 	LIST_HEAD(list_kill);
362 
363 	for_each_netdev(net, dev) {
364 		if (dev->rtnl_link_ops == ops)
365 			ops->dellink(dev, &list_kill);
366 	}
367 	unregister_netdevice_many(&list_kill);
368 }
369 
370 /**
371  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
372  * @ops: struct rtnl_link_ops * to unregister
373  *
374  * The caller must hold the rtnl_mutex.
375  */
376 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
377 {
378 	struct net *net;
379 
380 	for_each_net(net) {
381 		__rtnl_kill_links(net, ops);
382 	}
383 	list_del(&ops->list);
384 }
385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
386 
387 /* Return with the rtnl_lock held when there are no network
388  * devices unregistering in any network namespace.
389  */
390 static void rtnl_lock_unregistering_all(void)
391 {
392 	struct net *net;
393 	bool unregistering;
394 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
395 
396 	add_wait_queue(&netdev_unregistering_wq, &wait);
397 	for (;;) {
398 		unregistering = false;
399 		rtnl_lock();
400 		for_each_net(net) {
401 			if (net->dev_unreg_count > 0) {
402 				unregistering = true;
403 				break;
404 			}
405 		}
406 		if (!unregistering)
407 			break;
408 		__rtnl_unlock();
409 
410 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
411 	}
412 	remove_wait_queue(&netdev_unregistering_wq, &wait);
413 }
414 
415 /**
416  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
417  * @ops: struct rtnl_link_ops * to unregister
418  */
419 void rtnl_link_unregister(struct rtnl_link_ops *ops)
420 {
421 	/* Close the race with cleanup_net() */
422 	mutex_lock(&net_mutex);
423 	rtnl_lock_unregistering_all();
424 	__rtnl_link_unregister(ops);
425 	rtnl_unlock();
426 	mutex_unlock(&net_mutex);
427 }
428 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
429 
430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
431 {
432 	struct net_device *master_dev;
433 	const struct rtnl_link_ops *ops;
434 
435 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
436 	if (!master_dev)
437 		return 0;
438 	ops = master_dev->rtnl_link_ops;
439 	if (!ops || !ops->get_slave_size)
440 		return 0;
441 	/* IFLA_INFO_SLAVE_DATA + nested data */
442 	return nla_total_size(sizeof(struct nlattr)) +
443 	       ops->get_slave_size(master_dev, dev);
444 }
445 
446 static size_t rtnl_link_get_size(const struct net_device *dev)
447 {
448 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
449 	size_t size;
450 
451 	if (!ops)
452 		return 0;
453 
454 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
455 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
456 
457 	if (ops->get_size)
458 		/* IFLA_INFO_DATA + nested data */
459 		size += nla_total_size(sizeof(struct nlattr)) +
460 			ops->get_size(dev);
461 
462 	if (ops->get_xstats_size)
463 		/* IFLA_INFO_XSTATS */
464 		size += nla_total_size(ops->get_xstats_size(dev));
465 
466 	size += rtnl_link_get_slave_info_data_size(dev);
467 
468 	return size;
469 }
470 
471 static LIST_HEAD(rtnl_af_ops);
472 
473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
474 {
475 	const struct rtnl_af_ops *ops;
476 
477 	list_for_each_entry(ops, &rtnl_af_ops, list) {
478 		if (ops->family == family)
479 			return ops;
480 	}
481 
482 	return NULL;
483 }
484 
485 /**
486  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
487  * @ops: struct rtnl_af_ops * to register
488  *
489  * Returns 0 on success or a negative error code.
490  */
491 void rtnl_af_register(struct rtnl_af_ops *ops)
492 {
493 	rtnl_lock();
494 	list_add_tail(&ops->list, &rtnl_af_ops);
495 	rtnl_unlock();
496 }
497 EXPORT_SYMBOL_GPL(rtnl_af_register);
498 
499 /**
500  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
501  * @ops: struct rtnl_af_ops * to unregister
502  *
503  * The caller must hold the rtnl_mutex.
504  */
505 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
506 {
507 	list_del(&ops->list);
508 }
509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
510 
511 /**
512  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
513  * @ops: struct rtnl_af_ops * to unregister
514  */
515 void rtnl_af_unregister(struct rtnl_af_ops *ops)
516 {
517 	rtnl_lock();
518 	__rtnl_af_unregister(ops);
519 	rtnl_unlock();
520 }
521 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
522 
523 static size_t rtnl_link_get_af_size(const struct net_device *dev,
524 				    u32 ext_filter_mask)
525 {
526 	struct rtnl_af_ops *af_ops;
527 	size_t size;
528 
529 	/* IFLA_AF_SPEC */
530 	size = nla_total_size(sizeof(struct nlattr));
531 
532 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
533 		if (af_ops->get_link_af_size) {
534 			/* AF_* + nested data */
535 			size += nla_total_size(sizeof(struct nlattr)) +
536 				af_ops->get_link_af_size(dev, ext_filter_mask);
537 		}
538 	}
539 
540 	return size;
541 }
542 
543 static bool rtnl_have_link_slave_info(const struct net_device *dev)
544 {
545 	struct net_device *master_dev;
546 
547 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
548 	if (master_dev && master_dev->rtnl_link_ops)
549 		return true;
550 	return false;
551 }
552 
553 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
554 				     const struct net_device *dev)
555 {
556 	struct net_device *master_dev;
557 	const struct rtnl_link_ops *ops;
558 	struct nlattr *slave_data;
559 	int err;
560 
561 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
562 	if (!master_dev)
563 		return 0;
564 	ops = master_dev->rtnl_link_ops;
565 	if (!ops)
566 		return 0;
567 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
568 		return -EMSGSIZE;
569 	if (ops->fill_slave_info) {
570 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
571 		if (!slave_data)
572 			return -EMSGSIZE;
573 		err = ops->fill_slave_info(skb, master_dev, dev);
574 		if (err < 0)
575 			goto err_cancel_slave_data;
576 		nla_nest_end(skb, slave_data);
577 	}
578 	return 0;
579 
580 err_cancel_slave_data:
581 	nla_nest_cancel(skb, slave_data);
582 	return err;
583 }
584 
585 static int rtnl_link_info_fill(struct sk_buff *skb,
586 			       const struct net_device *dev)
587 {
588 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
589 	struct nlattr *data;
590 	int err;
591 
592 	if (!ops)
593 		return 0;
594 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
595 		return -EMSGSIZE;
596 	if (ops->fill_xstats) {
597 		err = ops->fill_xstats(skb, dev);
598 		if (err < 0)
599 			return err;
600 	}
601 	if (ops->fill_info) {
602 		data = nla_nest_start(skb, IFLA_INFO_DATA);
603 		if (data == NULL)
604 			return -EMSGSIZE;
605 		err = ops->fill_info(skb, dev);
606 		if (err < 0)
607 			goto err_cancel_data;
608 		nla_nest_end(skb, data);
609 	}
610 	return 0;
611 
612 err_cancel_data:
613 	nla_nest_cancel(skb, data);
614 	return err;
615 }
616 
617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
618 {
619 	struct nlattr *linkinfo;
620 	int err = -EMSGSIZE;
621 
622 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
623 	if (linkinfo == NULL)
624 		goto out;
625 
626 	err = rtnl_link_info_fill(skb, dev);
627 	if (err < 0)
628 		goto err_cancel_link;
629 
630 	err = rtnl_link_slave_info_fill(skb, dev);
631 	if (err < 0)
632 		goto err_cancel_link;
633 
634 	nla_nest_end(skb, linkinfo);
635 	return 0;
636 
637 err_cancel_link:
638 	nla_nest_cancel(skb, linkinfo);
639 out:
640 	return err;
641 }
642 
643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
644 {
645 	struct sock *rtnl = net->rtnl;
646 	int err = 0;
647 
648 	NETLINK_CB(skb).dst_group = group;
649 	if (echo)
650 		atomic_inc(&skb->users);
651 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
652 	if (echo)
653 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
654 	return err;
655 }
656 
657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
658 {
659 	struct sock *rtnl = net->rtnl;
660 
661 	return nlmsg_unicast(rtnl, skb, pid);
662 }
663 EXPORT_SYMBOL(rtnl_unicast);
664 
665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
666 		 struct nlmsghdr *nlh, gfp_t flags)
667 {
668 	struct sock *rtnl = net->rtnl;
669 	int report = 0;
670 
671 	if (nlh)
672 		report = nlmsg_report(nlh);
673 
674 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
675 }
676 EXPORT_SYMBOL(rtnl_notify);
677 
678 void rtnl_set_sk_err(struct net *net, u32 group, int error)
679 {
680 	struct sock *rtnl = net->rtnl;
681 
682 	netlink_set_err(rtnl, 0, group, error);
683 }
684 EXPORT_SYMBOL(rtnl_set_sk_err);
685 
686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
687 {
688 	struct nlattr *mx;
689 	int i, valid = 0;
690 
691 	mx = nla_nest_start(skb, RTA_METRICS);
692 	if (mx == NULL)
693 		return -ENOBUFS;
694 
695 	for (i = 0; i < RTAX_MAX; i++) {
696 		if (metrics[i]) {
697 			if (i == RTAX_CC_ALGO - 1) {
698 				char tmp[TCP_CA_NAME_MAX], *name;
699 
700 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
701 				if (!name)
702 					continue;
703 				if (nla_put_string(skb, i + 1, name))
704 					goto nla_put_failure;
705 			} else if (i == RTAX_FEATURES - 1) {
706 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
707 
708 				if (!user_features)
709 					continue;
710 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
711 				if (nla_put_u32(skb, i + 1, user_features))
712 					goto nla_put_failure;
713 			} else {
714 				if (nla_put_u32(skb, i + 1, metrics[i]))
715 					goto nla_put_failure;
716 			}
717 			valid++;
718 		}
719 	}
720 
721 	if (!valid) {
722 		nla_nest_cancel(skb, mx);
723 		return 0;
724 	}
725 
726 	return nla_nest_end(skb, mx);
727 
728 nla_put_failure:
729 	nla_nest_cancel(skb, mx);
730 	return -EMSGSIZE;
731 }
732 EXPORT_SYMBOL(rtnetlink_put_metrics);
733 
734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
735 		       long expires, u32 error)
736 {
737 	struct rta_cacheinfo ci = {
738 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
739 		.rta_used = dst->__use,
740 		.rta_clntref = atomic_read(&(dst->__refcnt)),
741 		.rta_error = error,
742 		.rta_id =  id,
743 	};
744 
745 	if (expires) {
746 		unsigned long clock;
747 
748 		clock = jiffies_to_clock_t(abs(expires));
749 		clock = min_t(unsigned long, clock, INT_MAX);
750 		ci.rta_expires = (expires > 0) ? clock : -clock;
751 	}
752 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
753 }
754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
755 
756 static void set_operstate(struct net_device *dev, unsigned char transition)
757 {
758 	unsigned char operstate = dev->operstate;
759 
760 	switch (transition) {
761 	case IF_OPER_UP:
762 		if ((operstate == IF_OPER_DORMANT ||
763 		     operstate == IF_OPER_UNKNOWN) &&
764 		    !netif_dormant(dev))
765 			operstate = IF_OPER_UP;
766 		break;
767 
768 	case IF_OPER_DORMANT:
769 		if (operstate == IF_OPER_UP ||
770 		    operstate == IF_OPER_UNKNOWN)
771 			operstate = IF_OPER_DORMANT;
772 		break;
773 	}
774 
775 	if (dev->operstate != operstate) {
776 		write_lock_bh(&dev_base_lock);
777 		dev->operstate = operstate;
778 		write_unlock_bh(&dev_base_lock);
779 		netdev_state_change(dev);
780 	}
781 }
782 
783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
784 {
785 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
786 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
787 }
788 
789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
790 					   const struct ifinfomsg *ifm)
791 {
792 	unsigned int flags = ifm->ifi_flags;
793 
794 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
795 	if (ifm->ifi_change)
796 		flags = (flags & ifm->ifi_change) |
797 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
798 
799 	return flags;
800 }
801 
802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
803 				 const struct rtnl_link_stats64 *b)
804 {
805 	a->rx_packets = b->rx_packets;
806 	a->tx_packets = b->tx_packets;
807 	a->rx_bytes = b->rx_bytes;
808 	a->tx_bytes = b->tx_bytes;
809 	a->rx_errors = b->rx_errors;
810 	a->tx_errors = b->tx_errors;
811 	a->rx_dropped = b->rx_dropped;
812 	a->tx_dropped = b->tx_dropped;
813 
814 	a->multicast = b->multicast;
815 	a->collisions = b->collisions;
816 
817 	a->rx_length_errors = b->rx_length_errors;
818 	a->rx_over_errors = b->rx_over_errors;
819 	a->rx_crc_errors = b->rx_crc_errors;
820 	a->rx_frame_errors = b->rx_frame_errors;
821 	a->rx_fifo_errors = b->rx_fifo_errors;
822 	a->rx_missed_errors = b->rx_missed_errors;
823 
824 	a->tx_aborted_errors = b->tx_aborted_errors;
825 	a->tx_carrier_errors = b->tx_carrier_errors;
826 	a->tx_fifo_errors = b->tx_fifo_errors;
827 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
828 	a->tx_window_errors = b->tx_window_errors;
829 
830 	a->rx_compressed = b->rx_compressed;
831 	a->tx_compressed = b->tx_compressed;
832 
833 	a->rx_nohandler = b->rx_nohandler;
834 }
835 
836 /* All VF info */
837 static inline int rtnl_vfinfo_size(const struct net_device *dev,
838 				   u32 ext_filter_mask)
839 {
840 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
841 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
842 		int num_vfs = dev_num_vf(dev->dev.parent);
843 		size_t size = nla_total_size(sizeof(struct nlattr));
844 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
845 		size += num_vfs *
846 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
847 			 nla_total_size(MAX_VLAN_LIST_LEN *
848 					sizeof(struct nlattr)) +
849 			 nla_total_size(MAX_VLAN_LIST_LEN *
850 					sizeof(struct ifla_vf_vlan_info)) +
851 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
852 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
853 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
854 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
855 			 /* IFLA_VF_STATS_RX_PACKETS */
856 			 nla_total_size_64bit(sizeof(__u64)) +
857 			 /* IFLA_VF_STATS_TX_PACKETS */
858 			 nla_total_size_64bit(sizeof(__u64)) +
859 			 /* IFLA_VF_STATS_RX_BYTES */
860 			 nla_total_size_64bit(sizeof(__u64)) +
861 			 /* IFLA_VF_STATS_TX_BYTES */
862 			 nla_total_size_64bit(sizeof(__u64)) +
863 			 /* IFLA_VF_STATS_BROADCAST */
864 			 nla_total_size_64bit(sizeof(__u64)) +
865 			 /* IFLA_VF_STATS_MULTICAST */
866 			 nla_total_size_64bit(sizeof(__u64)) +
867 			 nla_total_size(sizeof(struct ifla_vf_trust)));
868 		return size;
869 	} else
870 		return 0;
871 }
872 
873 static size_t rtnl_port_size(const struct net_device *dev,
874 			     u32 ext_filter_mask)
875 {
876 	size_t port_size = nla_total_size(4)		/* PORT_VF */
877 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
878 		+ nla_total_size(sizeof(struct ifla_port_vsi))
879 							/* PORT_VSI_TYPE */
880 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
881 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
882 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
883 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
884 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
885 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
886 		+ port_size;
887 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
888 		+ port_size;
889 
890 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
891 	    !(ext_filter_mask & RTEXT_FILTER_VF))
892 		return 0;
893 	if (dev_num_vf(dev->dev.parent))
894 		return port_self_size + vf_ports_size +
895 			vf_port_size * dev_num_vf(dev->dev.parent);
896 	else
897 		return port_self_size;
898 }
899 
900 static size_t rtnl_xdp_size(const struct net_device *dev)
901 {
902 	size_t xdp_size = nla_total_size(1);	/* XDP_ATTACHED */
903 
904 	if (!dev->netdev_ops->ndo_xdp)
905 		return 0;
906 	else
907 		return xdp_size;
908 }
909 
910 static noinline size_t if_nlmsg_size(const struct net_device *dev,
911 				     u32 ext_filter_mask)
912 {
913 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
914 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
915 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
916 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
917 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
918 	       + nla_total_size(sizeof(struct rtnl_link_stats))
919 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
920 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
921 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
922 	       + nla_total_size(4) /* IFLA_TXQLEN */
923 	       + nla_total_size(4) /* IFLA_WEIGHT */
924 	       + nla_total_size(4) /* IFLA_MTU */
925 	       + nla_total_size(4) /* IFLA_LINK */
926 	       + nla_total_size(4) /* IFLA_MASTER */
927 	       + nla_total_size(1) /* IFLA_CARRIER */
928 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
929 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
930 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
931 	       + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
932 	       + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
933 	       + nla_total_size(1) /* IFLA_OPERSTATE */
934 	       + nla_total_size(1) /* IFLA_LINKMODE */
935 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
936 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
937 	       + nla_total_size(ext_filter_mask
938 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
939 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
940 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
941 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
942 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
943 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
944 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
945 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
946 	       + rtnl_xdp_size(dev) /* IFLA_XDP */
947 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
948 
949 }
950 
951 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
952 {
953 	struct nlattr *vf_ports;
954 	struct nlattr *vf_port;
955 	int vf;
956 	int err;
957 
958 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
959 	if (!vf_ports)
960 		return -EMSGSIZE;
961 
962 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
963 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
964 		if (!vf_port)
965 			goto nla_put_failure;
966 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
967 			goto nla_put_failure;
968 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
969 		if (err == -EMSGSIZE)
970 			goto nla_put_failure;
971 		if (err) {
972 			nla_nest_cancel(skb, vf_port);
973 			continue;
974 		}
975 		nla_nest_end(skb, vf_port);
976 	}
977 
978 	nla_nest_end(skb, vf_ports);
979 
980 	return 0;
981 
982 nla_put_failure:
983 	nla_nest_cancel(skb, vf_ports);
984 	return -EMSGSIZE;
985 }
986 
987 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
988 {
989 	struct nlattr *port_self;
990 	int err;
991 
992 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
993 	if (!port_self)
994 		return -EMSGSIZE;
995 
996 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
997 	if (err) {
998 		nla_nest_cancel(skb, port_self);
999 		return (err == -EMSGSIZE) ? err : 0;
1000 	}
1001 
1002 	nla_nest_end(skb, port_self);
1003 
1004 	return 0;
1005 }
1006 
1007 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1008 			  u32 ext_filter_mask)
1009 {
1010 	int err;
1011 
1012 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1013 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1014 		return 0;
1015 
1016 	err = rtnl_port_self_fill(skb, dev);
1017 	if (err)
1018 		return err;
1019 
1020 	if (dev_num_vf(dev->dev.parent)) {
1021 		err = rtnl_vf_ports_fill(skb, dev);
1022 		if (err)
1023 			return err;
1024 	}
1025 
1026 	return 0;
1027 }
1028 
1029 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1030 {
1031 	int err;
1032 	struct netdev_phys_item_id ppid;
1033 
1034 	err = dev_get_phys_port_id(dev, &ppid);
1035 	if (err) {
1036 		if (err == -EOPNOTSUPP)
1037 			return 0;
1038 		return err;
1039 	}
1040 
1041 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1042 		return -EMSGSIZE;
1043 
1044 	return 0;
1045 }
1046 
1047 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1048 {
1049 	char name[IFNAMSIZ];
1050 	int err;
1051 
1052 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1053 	if (err) {
1054 		if (err == -EOPNOTSUPP)
1055 			return 0;
1056 		return err;
1057 	}
1058 
1059 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1060 		return -EMSGSIZE;
1061 
1062 	return 0;
1063 }
1064 
1065 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1066 {
1067 	int err;
1068 	struct switchdev_attr attr = {
1069 		.orig_dev = dev,
1070 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1071 		.flags = SWITCHDEV_F_NO_RECURSE,
1072 	};
1073 
1074 	err = switchdev_port_attr_get(dev, &attr);
1075 	if (err) {
1076 		if (err == -EOPNOTSUPP)
1077 			return 0;
1078 		return err;
1079 	}
1080 
1081 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1082 		    attr.u.ppid.id))
1083 		return -EMSGSIZE;
1084 
1085 	return 0;
1086 }
1087 
1088 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1089 					      struct net_device *dev)
1090 {
1091 	struct rtnl_link_stats64 *sp;
1092 	struct nlattr *attr;
1093 
1094 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1095 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1096 	if (!attr)
1097 		return -EMSGSIZE;
1098 
1099 	sp = nla_data(attr);
1100 	dev_get_stats(dev, sp);
1101 
1102 	attr = nla_reserve(skb, IFLA_STATS,
1103 			   sizeof(struct rtnl_link_stats));
1104 	if (!attr)
1105 		return -EMSGSIZE;
1106 
1107 	copy_rtnl_link_stats(nla_data(attr), sp);
1108 
1109 	return 0;
1110 }
1111 
1112 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1113 					       struct net_device *dev,
1114 					       int vfs_num,
1115 					       struct nlattr *vfinfo)
1116 {
1117 	struct ifla_vf_rss_query_en vf_rss_query_en;
1118 	struct nlattr *vf, *vfstats, *vfvlanlist;
1119 	struct ifla_vf_link_state vf_linkstate;
1120 	struct ifla_vf_vlan_info vf_vlan_info;
1121 	struct ifla_vf_spoofchk vf_spoofchk;
1122 	struct ifla_vf_tx_rate vf_tx_rate;
1123 	struct ifla_vf_stats vf_stats;
1124 	struct ifla_vf_trust vf_trust;
1125 	struct ifla_vf_vlan vf_vlan;
1126 	struct ifla_vf_rate vf_rate;
1127 	struct ifla_vf_mac vf_mac;
1128 	struct ifla_vf_info ivi;
1129 
1130 	/* Not all SR-IOV capable drivers support the
1131 	 * spoofcheck and "RSS query enable" query.  Preset to
1132 	 * -1 so the user space tool can detect that the driver
1133 	 * didn't report anything.
1134 	 */
1135 	ivi.spoofchk = -1;
1136 	ivi.rss_query_en = -1;
1137 	ivi.trusted = -1;
1138 	memset(ivi.mac, 0, sizeof(ivi.mac));
1139 	/* The default value for VF link state is "auto"
1140 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1141 	 */
1142 	ivi.linkstate = 0;
1143 	/* VLAN Protocol by default is 802.1Q */
1144 	ivi.vlan_proto = htons(ETH_P_8021Q);
1145 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1146 		return 0;
1147 
1148 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1149 
1150 	vf_mac.vf =
1151 		vf_vlan.vf =
1152 		vf_vlan_info.vf =
1153 		vf_rate.vf =
1154 		vf_tx_rate.vf =
1155 		vf_spoofchk.vf =
1156 		vf_linkstate.vf =
1157 		vf_rss_query_en.vf =
1158 		vf_trust.vf = ivi.vf;
1159 
1160 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1161 	vf_vlan.vlan = ivi.vlan;
1162 	vf_vlan.qos = ivi.qos;
1163 	vf_vlan_info.vlan = ivi.vlan;
1164 	vf_vlan_info.qos = ivi.qos;
1165 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1166 	vf_tx_rate.rate = ivi.max_tx_rate;
1167 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1168 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1169 	vf_spoofchk.setting = ivi.spoofchk;
1170 	vf_linkstate.link_state = ivi.linkstate;
1171 	vf_rss_query_en.setting = ivi.rss_query_en;
1172 	vf_trust.setting = ivi.trusted;
1173 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1174 	if (!vf)
1175 		goto nla_put_vfinfo_failure;
1176 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1177 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1178 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1179 		    &vf_rate) ||
1180 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1181 		    &vf_tx_rate) ||
1182 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1183 		    &vf_spoofchk) ||
1184 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1185 		    &vf_linkstate) ||
1186 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1187 		    sizeof(vf_rss_query_en),
1188 		    &vf_rss_query_en) ||
1189 	    nla_put(skb, IFLA_VF_TRUST,
1190 		    sizeof(vf_trust), &vf_trust))
1191 		goto nla_put_vf_failure;
1192 	vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1193 	if (!vfvlanlist)
1194 		goto nla_put_vf_failure;
1195 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1196 		    &vf_vlan_info)) {
1197 		nla_nest_cancel(skb, vfvlanlist);
1198 		goto nla_put_vf_failure;
1199 	}
1200 	nla_nest_end(skb, vfvlanlist);
1201 	memset(&vf_stats, 0, sizeof(vf_stats));
1202 	if (dev->netdev_ops->ndo_get_vf_stats)
1203 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1204 						&vf_stats);
1205 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1206 	if (!vfstats)
1207 		goto nla_put_vf_failure;
1208 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1209 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1210 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1211 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1212 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1213 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1214 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1215 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1216 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1217 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1218 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1219 			      vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1220 		nla_nest_cancel(skb, vfstats);
1221 		goto nla_put_vf_failure;
1222 	}
1223 	nla_nest_end(skb, vfstats);
1224 	nla_nest_end(skb, vf);
1225 	return 0;
1226 
1227 nla_put_vf_failure:
1228 	nla_nest_cancel(skb, vf);
1229 nla_put_vfinfo_failure:
1230 	nla_nest_cancel(skb, vfinfo);
1231 	return -EMSGSIZE;
1232 }
1233 
1234 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1235 {
1236 	struct rtnl_link_ifmap map;
1237 
1238 	memset(&map, 0, sizeof(map));
1239 	map.mem_start   = dev->mem_start;
1240 	map.mem_end     = dev->mem_end;
1241 	map.base_addr   = dev->base_addr;
1242 	map.irq         = dev->irq;
1243 	map.dma         = dev->dma;
1244 	map.port        = dev->if_port;
1245 
1246 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1247 		return -EMSGSIZE;
1248 
1249 	return 0;
1250 }
1251 
1252 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1253 {
1254 	struct netdev_xdp xdp_op = {};
1255 	struct nlattr *xdp;
1256 	int err;
1257 
1258 	if (!dev->netdev_ops->ndo_xdp)
1259 		return 0;
1260 	xdp = nla_nest_start(skb, IFLA_XDP);
1261 	if (!xdp)
1262 		return -EMSGSIZE;
1263 	xdp_op.command = XDP_QUERY_PROG;
1264 	err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1265 	if (err)
1266 		goto err_cancel;
1267 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1268 	if (err)
1269 		goto err_cancel;
1270 
1271 	nla_nest_end(skb, xdp);
1272 	return 0;
1273 
1274 err_cancel:
1275 	nla_nest_cancel(skb, xdp);
1276 	return err;
1277 }
1278 
1279 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1280 			    int type, u32 pid, u32 seq, u32 change,
1281 			    unsigned int flags, u32 ext_filter_mask)
1282 {
1283 	struct ifinfomsg *ifm;
1284 	struct nlmsghdr *nlh;
1285 	struct nlattr *af_spec;
1286 	struct rtnl_af_ops *af_ops;
1287 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1288 
1289 	ASSERT_RTNL();
1290 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1291 	if (nlh == NULL)
1292 		return -EMSGSIZE;
1293 
1294 	ifm = nlmsg_data(nlh);
1295 	ifm->ifi_family = AF_UNSPEC;
1296 	ifm->__ifi_pad = 0;
1297 	ifm->ifi_type = dev->type;
1298 	ifm->ifi_index = dev->ifindex;
1299 	ifm->ifi_flags = dev_get_flags(dev);
1300 	ifm->ifi_change = change;
1301 
1302 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1303 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1304 	    nla_put_u8(skb, IFLA_OPERSTATE,
1305 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1306 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1307 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1308 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1309 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1310 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1311 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1312 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1313 #ifdef CONFIG_RPS
1314 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1315 #endif
1316 	    (dev->ifindex != dev_get_iflink(dev) &&
1317 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1318 	    (upper_dev &&
1319 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1320 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1321 	    (dev->qdisc &&
1322 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1323 	    (dev->ifalias &&
1324 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1325 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1326 			atomic_read(&dev->carrier_changes)) ||
1327 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1328 		goto nla_put_failure;
1329 
1330 	if (rtnl_fill_link_ifmap(skb, dev))
1331 		goto nla_put_failure;
1332 
1333 	if (dev->addr_len) {
1334 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1335 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1336 			goto nla_put_failure;
1337 	}
1338 
1339 	if (rtnl_phys_port_id_fill(skb, dev))
1340 		goto nla_put_failure;
1341 
1342 	if (rtnl_phys_port_name_fill(skb, dev))
1343 		goto nla_put_failure;
1344 
1345 	if (rtnl_phys_switch_id_fill(skb, dev))
1346 		goto nla_put_failure;
1347 
1348 	if (rtnl_fill_stats(skb, dev))
1349 		goto nla_put_failure;
1350 
1351 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1352 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1353 		goto nla_put_failure;
1354 
1355 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1356 	    ext_filter_mask & RTEXT_FILTER_VF) {
1357 		int i;
1358 		struct nlattr *vfinfo;
1359 		int num_vfs = dev_num_vf(dev->dev.parent);
1360 
1361 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1362 		if (!vfinfo)
1363 			goto nla_put_failure;
1364 		for (i = 0; i < num_vfs; i++) {
1365 			if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1366 				goto nla_put_failure;
1367 		}
1368 
1369 		nla_nest_end(skb, vfinfo);
1370 	}
1371 
1372 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1373 		goto nla_put_failure;
1374 
1375 	if (rtnl_xdp_fill(skb, dev))
1376 		goto nla_put_failure;
1377 
1378 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1379 		if (rtnl_link_fill(skb, dev) < 0)
1380 			goto nla_put_failure;
1381 	}
1382 
1383 	if (dev->rtnl_link_ops &&
1384 	    dev->rtnl_link_ops->get_link_net) {
1385 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1386 
1387 		if (!net_eq(dev_net(dev), link_net)) {
1388 			int id = peernet2id_alloc(dev_net(dev), link_net);
1389 
1390 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1391 				goto nla_put_failure;
1392 		}
1393 	}
1394 
1395 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1396 		goto nla_put_failure;
1397 
1398 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1399 		if (af_ops->fill_link_af) {
1400 			struct nlattr *af;
1401 			int err;
1402 
1403 			if (!(af = nla_nest_start(skb, af_ops->family)))
1404 				goto nla_put_failure;
1405 
1406 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1407 
1408 			/*
1409 			 * Caller may return ENODATA to indicate that there
1410 			 * was no data to be dumped. This is not an error, it
1411 			 * means we should trim the attribute header and
1412 			 * continue.
1413 			 */
1414 			if (err == -ENODATA)
1415 				nla_nest_cancel(skb, af);
1416 			else if (err < 0)
1417 				goto nla_put_failure;
1418 
1419 			nla_nest_end(skb, af);
1420 		}
1421 	}
1422 
1423 	nla_nest_end(skb, af_spec);
1424 
1425 	nlmsg_end(skb, nlh);
1426 	return 0;
1427 
1428 nla_put_failure:
1429 	nlmsg_cancel(skb, nlh);
1430 	return -EMSGSIZE;
1431 }
1432 
1433 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1434 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1435 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1436 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1437 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1438 	[IFLA_MTU]		= { .type = NLA_U32 },
1439 	[IFLA_LINK]		= { .type = NLA_U32 },
1440 	[IFLA_MASTER]		= { .type = NLA_U32 },
1441 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1442 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1443 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1444 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1445 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1446 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1447 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1448 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1449 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1450 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1451 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1452 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1453 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1454 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1455 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1456 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1457 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1458 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1459 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1460 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1461 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1462 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1463 	[IFLA_XDP]		= { .type = NLA_NESTED },
1464 };
1465 
1466 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1467 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1468 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1469 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1470 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1471 };
1472 
1473 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1474 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1475 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1476 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1477 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1478 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1479 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1480 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1481 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1482 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1483 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1484 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1485 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1486 };
1487 
1488 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1489 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1490 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1491 				    .len = PORT_PROFILE_MAX },
1492 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1493 				    .len = sizeof(struct ifla_port_vsi)},
1494 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1495 				      .len = PORT_UUID_MAX },
1496 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1497 				    .len = PORT_UUID_MAX },
1498 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1499 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1500 };
1501 
1502 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1503 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1504 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1505 };
1506 
1507 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1508 {
1509 	const struct rtnl_link_ops *ops = NULL;
1510 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1511 
1512 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1513 		return NULL;
1514 
1515 	if (linfo[IFLA_INFO_KIND]) {
1516 		char kind[MODULE_NAME_LEN];
1517 
1518 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1519 		ops = rtnl_link_ops_get(kind);
1520 	}
1521 
1522 	return ops;
1523 }
1524 
1525 static bool link_master_filtered(struct net_device *dev, int master_idx)
1526 {
1527 	struct net_device *master;
1528 
1529 	if (!master_idx)
1530 		return false;
1531 
1532 	master = netdev_master_upper_dev_get(dev);
1533 	if (!master || master->ifindex != master_idx)
1534 		return true;
1535 
1536 	return false;
1537 }
1538 
1539 static bool link_kind_filtered(const struct net_device *dev,
1540 			       const struct rtnl_link_ops *kind_ops)
1541 {
1542 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1543 		return true;
1544 
1545 	return false;
1546 }
1547 
1548 static bool link_dump_filtered(struct net_device *dev,
1549 			       int master_idx,
1550 			       const struct rtnl_link_ops *kind_ops)
1551 {
1552 	if (link_master_filtered(dev, master_idx) ||
1553 	    link_kind_filtered(dev, kind_ops))
1554 		return true;
1555 
1556 	return false;
1557 }
1558 
1559 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1560 {
1561 	struct net *net = sock_net(skb->sk);
1562 	int h, s_h;
1563 	int idx = 0, s_idx;
1564 	struct net_device *dev;
1565 	struct hlist_head *head;
1566 	struct nlattr *tb[IFLA_MAX+1];
1567 	u32 ext_filter_mask = 0;
1568 	const struct rtnl_link_ops *kind_ops = NULL;
1569 	unsigned int flags = NLM_F_MULTI;
1570 	int master_idx = 0;
1571 	int err;
1572 	int hdrlen;
1573 
1574 	s_h = cb->args[0];
1575 	s_idx = cb->args[1];
1576 
1577 	cb->seq = net->dev_base_seq;
1578 
1579 	/* A hack to preserve kernel<->userspace interface.
1580 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1581 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1582 	 * what iproute2 < v3.9.0 used.
1583 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1584 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1585 	 */
1586 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1587 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1588 
1589 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1590 
1591 		if (tb[IFLA_EXT_MASK])
1592 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1593 
1594 		if (tb[IFLA_MASTER])
1595 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1596 
1597 		if (tb[IFLA_LINKINFO])
1598 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1599 
1600 		if (master_idx || kind_ops)
1601 			flags |= NLM_F_DUMP_FILTERED;
1602 	}
1603 
1604 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1605 		idx = 0;
1606 		head = &net->dev_index_head[h];
1607 		hlist_for_each_entry(dev, head, index_hlist) {
1608 			if (link_dump_filtered(dev, master_idx, kind_ops))
1609 				continue;
1610 			if (idx < s_idx)
1611 				goto cont;
1612 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1613 					       NETLINK_CB(cb->skb).portid,
1614 					       cb->nlh->nlmsg_seq, 0,
1615 					       flags,
1616 					       ext_filter_mask);
1617 			/* If we ran out of room on the first message,
1618 			 * we're in trouble
1619 			 */
1620 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1621 
1622 			if (err < 0)
1623 				goto out;
1624 
1625 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1626 cont:
1627 			idx++;
1628 		}
1629 	}
1630 out:
1631 	cb->args[1] = idx;
1632 	cb->args[0] = h;
1633 
1634 	return skb->len;
1635 }
1636 
1637 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1638 {
1639 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1640 }
1641 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1642 
1643 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1644 {
1645 	struct net *net;
1646 	/* Examine the link attributes and figure out which
1647 	 * network namespace we are talking about.
1648 	 */
1649 	if (tb[IFLA_NET_NS_PID])
1650 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1651 	else if (tb[IFLA_NET_NS_FD])
1652 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1653 	else
1654 		net = get_net(src_net);
1655 	return net;
1656 }
1657 EXPORT_SYMBOL(rtnl_link_get_net);
1658 
1659 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1660 {
1661 	if (dev) {
1662 		if (tb[IFLA_ADDRESS] &&
1663 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1664 			return -EINVAL;
1665 
1666 		if (tb[IFLA_BROADCAST] &&
1667 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1668 			return -EINVAL;
1669 	}
1670 
1671 	if (tb[IFLA_AF_SPEC]) {
1672 		struct nlattr *af;
1673 		int rem, err;
1674 
1675 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1676 			const struct rtnl_af_ops *af_ops;
1677 
1678 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1679 				return -EAFNOSUPPORT;
1680 
1681 			if (!af_ops->set_link_af)
1682 				return -EOPNOTSUPP;
1683 
1684 			if (af_ops->validate_link_af) {
1685 				err = af_ops->validate_link_af(dev, af);
1686 				if (err < 0)
1687 					return err;
1688 			}
1689 		}
1690 	}
1691 
1692 	return 0;
1693 }
1694 
1695 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1696 				  int guid_type)
1697 {
1698 	const struct net_device_ops *ops = dev->netdev_ops;
1699 
1700 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1701 }
1702 
1703 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1704 {
1705 	if (dev->type != ARPHRD_INFINIBAND)
1706 		return -EOPNOTSUPP;
1707 
1708 	return handle_infiniband_guid(dev, ivt, guid_type);
1709 }
1710 
1711 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1712 {
1713 	const struct net_device_ops *ops = dev->netdev_ops;
1714 	int err = -EINVAL;
1715 
1716 	if (tb[IFLA_VF_MAC]) {
1717 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1718 
1719 		err = -EOPNOTSUPP;
1720 		if (ops->ndo_set_vf_mac)
1721 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1722 						  ivm->mac);
1723 		if (err < 0)
1724 			return err;
1725 	}
1726 
1727 	if (tb[IFLA_VF_VLAN]) {
1728 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1729 
1730 		err = -EOPNOTSUPP;
1731 		if (ops->ndo_set_vf_vlan)
1732 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1733 						   ivv->qos,
1734 						   htons(ETH_P_8021Q));
1735 		if (err < 0)
1736 			return err;
1737 	}
1738 
1739 	if (tb[IFLA_VF_VLAN_LIST]) {
1740 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1741 		struct nlattr *attr;
1742 		int rem, len = 0;
1743 
1744 		err = -EOPNOTSUPP;
1745 		if (!ops->ndo_set_vf_vlan)
1746 			return err;
1747 
1748 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1749 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1750 			    nla_len(attr) < NLA_HDRLEN) {
1751 				return -EINVAL;
1752 			}
1753 			if (len >= MAX_VLAN_LIST_LEN)
1754 				return -EOPNOTSUPP;
1755 			ivvl[len] = nla_data(attr);
1756 
1757 			len++;
1758 		}
1759 		if (len == 0)
1760 			return -EINVAL;
1761 
1762 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1763 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
1764 		if (err < 0)
1765 			return err;
1766 	}
1767 
1768 	if (tb[IFLA_VF_TX_RATE]) {
1769 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1770 		struct ifla_vf_info ivf;
1771 
1772 		err = -EOPNOTSUPP;
1773 		if (ops->ndo_get_vf_config)
1774 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1775 		if (err < 0)
1776 			return err;
1777 
1778 		err = -EOPNOTSUPP;
1779 		if (ops->ndo_set_vf_rate)
1780 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1781 						   ivf.min_tx_rate,
1782 						   ivt->rate);
1783 		if (err < 0)
1784 			return err;
1785 	}
1786 
1787 	if (tb[IFLA_VF_RATE]) {
1788 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1789 
1790 		err = -EOPNOTSUPP;
1791 		if (ops->ndo_set_vf_rate)
1792 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1793 						   ivt->min_tx_rate,
1794 						   ivt->max_tx_rate);
1795 		if (err < 0)
1796 			return err;
1797 	}
1798 
1799 	if (tb[IFLA_VF_SPOOFCHK]) {
1800 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1801 
1802 		err = -EOPNOTSUPP;
1803 		if (ops->ndo_set_vf_spoofchk)
1804 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1805 						       ivs->setting);
1806 		if (err < 0)
1807 			return err;
1808 	}
1809 
1810 	if (tb[IFLA_VF_LINK_STATE]) {
1811 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1812 
1813 		err = -EOPNOTSUPP;
1814 		if (ops->ndo_set_vf_link_state)
1815 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1816 							 ivl->link_state);
1817 		if (err < 0)
1818 			return err;
1819 	}
1820 
1821 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1822 		struct ifla_vf_rss_query_en *ivrssq_en;
1823 
1824 		err = -EOPNOTSUPP;
1825 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1826 		if (ops->ndo_set_vf_rss_query_en)
1827 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1828 							   ivrssq_en->setting);
1829 		if (err < 0)
1830 			return err;
1831 	}
1832 
1833 	if (tb[IFLA_VF_TRUST]) {
1834 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1835 
1836 		err = -EOPNOTSUPP;
1837 		if (ops->ndo_set_vf_trust)
1838 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1839 		if (err < 0)
1840 			return err;
1841 	}
1842 
1843 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1844 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1845 
1846 		if (!ops->ndo_set_vf_guid)
1847 			return -EOPNOTSUPP;
1848 
1849 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1850 	}
1851 
1852 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1853 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1854 
1855 		if (!ops->ndo_set_vf_guid)
1856 			return -EOPNOTSUPP;
1857 
1858 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1859 	}
1860 
1861 	return err;
1862 }
1863 
1864 static int do_set_master(struct net_device *dev, int ifindex)
1865 {
1866 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1867 	const struct net_device_ops *ops;
1868 	int err;
1869 
1870 	if (upper_dev) {
1871 		if (upper_dev->ifindex == ifindex)
1872 			return 0;
1873 		ops = upper_dev->netdev_ops;
1874 		if (ops->ndo_del_slave) {
1875 			err = ops->ndo_del_slave(upper_dev, dev);
1876 			if (err)
1877 				return err;
1878 		} else {
1879 			return -EOPNOTSUPP;
1880 		}
1881 	}
1882 
1883 	if (ifindex) {
1884 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1885 		if (!upper_dev)
1886 			return -EINVAL;
1887 		ops = upper_dev->netdev_ops;
1888 		if (ops->ndo_add_slave) {
1889 			err = ops->ndo_add_slave(upper_dev, dev);
1890 			if (err)
1891 				return err;
1892 		} else {
1893 			return -EOPNOTSUPP;
1894 		}
1895 	}
1896 	return 0;
1897 }
1898 
1899 #define DO_SETLINK_MODIFIED	0x01
1900 /* notify flag means notify + modified. */
1901 #define DO_SETLINK_NOTIFY	0x03
1902 static int do_setlink(const struct sk_buff *skb,
1903 		      struct net_device *dev, struct ifinfomsg *ifm,
1904 		      struct nlattr **tb, char *ifname, int status)
1905 {
1906 	const struct net_device_ops *ops = dev->netdev_ops;
1907 	int err;
1908 
1909 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1910 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1911 		if (IS_ERR(net)) {
1912 			err = PTR_ERR(net);
1913 			goto errout;
1914 		}
1915 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1916 			put_net(net);
1917 			err = -EPERM;
1918 			goto errout;
1919 		}
1920 		err = dev_change_net_namespace(dev, net, ifname);
1921 		put_net(net);
1922 		if (err)
1923 			goto errout;
1924 		status |= DO_SETLINK_MODIFIED;
1925 	}
1926 
1927 	if (tb[IFLA_MAP]) {
1928 		struct rtnl_link_ifmap *u_map;
1929 		struct ifmap k_map;
1930 
1931 		if (!ops->ndo_set_config) {
1932 			err = -EOPNOTSUPP;
1933 			goto errout;
1934 		}
1935 
1936 		if (!netif_device_present(dev)) {
1937 			err = -ENODEV;
1938 			goto errout;
1939 		}
1940 
1941 		u_map = nla_data(tb[IFLA_MAP]);
1942 		k_map.mem_start = (unsigned long) u_map->mem_start;
1943 		k_map.mem_end = (unsigned long) u_map->mem_end;
1944 		k_map.base_addr = (unsigned short) u_map->base_addr;
1945 		k_map.irq = (unsigned char) u_map->irq;
1946 		k_map.dma = (unsigned char) u_map->dma;
1947 		k_map.port = (unsigned char) u_map->port;
1948 
1949 		err = ops->ndo_set_config(dev, &k_map);
1950 		if (err < 0)
1951 			goto errout;
1952 
1953 		status |= DO_SETLINK_NOTIFY;
1954 	}
1955 
1956 	if (tb[IFLA_ADDRESS]) {
1957 		struct sockaddr *sa;
1958 		int len;
1959 
1960 		len = sizeof(sa_family_t) + dev->addr_len;
1961 		sa = kmalloc(len, GFP_KERNEL);
1962 		if (!sa) {
1963 			err = -ENOMEM;
1964 			goto errout;
1965 		}
1966 		sa->sa_family = dev->type;
1967 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1968 		       dev->addr_len);
1969 		err = dev_set_mac_address(dev, sa);
1970 		kfree(sa);
1971 		if (err)
1972 			goto errout;
1973 		status |= DO_SETLINK_MODIFIED;
1974 	}
1975 
1976 	if (tb[IFLA_MTU]) {
1977 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1978 		if (err < 0)
1979 			goto errout;
1980 		status |= DO_SETLINK_MODIFIED;
1981 	}
1982 
1983 	if (tb[IFLA_GROUP]) {
1984 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1985 		status |= DO_SETLINK_NOTIFY;
1986 	}
1987 
1988 	/*
1989 	 * Interface selected by interface index but interface
1990 	 * name provided implies that a name change has been
1991 	 * requested.
1992 	 */
1993 	if (ifm->ifi_index > 0 && ifname[0]) {
1994 		err = dev_change_name(dev, ifname);
1995 		if (err < 0)
1996 			goto errout;
1997 		status |= DO_SETLINK_MODIFIED;
1998 	}
1999 
2000 	if (tb[IFLA_IFALIAS]) {
2001 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2002 				    nla_len(tb[IFLA_IFALIAS]));
2003 		if (err < 0)
2004 			goto errout;
2005 		status |= DO_SETLINK_NOTIFY;
2006 	}
2007 
2008 	if (tb[IFLA_BROADCAST]) {
2009 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2010 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2011 	}
2012 
2013 	if (ifm->ifi_flags || ifm->ifi_change) {
2014 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2015 		if (err < 0)
2016 			goto errout;
2017 	}
2018 
2019 	if (tb[IFLA_MASTER]) {
2020 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2021 		if (err)
2022 			goto errout;
2023 		status |= DO_SETLINK_MODIFIED;
2024 	}
2025 
2026 	if (tb[IFLA_CARRIER]) {
2027 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2028 		if (err)
2029 			goto errout;
2030 		status |= DO_SETLINK_MODIFIED;
2031 	}
2032 
2033 	if (tb[IFLA_TXQLEN]) {
2034 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2035 		unsigned long orig_len = dev->tx_queue_len;
2036 
2037 		if (dev->tx_queue_len ^ value) {
2038 			dev->tx_queue_len = value;
2039 			err = call_netdevice_notifiers(
2040 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2041 			err = notifier_to_errno(err);
2042 			if (err) {
2043 				dev->tx_queue_len = orig_len;
2044 				goto errout;
2045 			}
2046 			status |= DO_SETLINK_NOTIFY;
2047 		}
2048 	}
2049 
2050 	if (tb[IFLA_OPERSTATE])
2051 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2052 
2053 	if (tb[IFLA_LINKMODE]) {
2054 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2055 
2056 		write_lock_bh(&dev_base_lock);
2057 		if (dev->link_mode ^ value)
2058 			status |= DO_SETLINK_NOTIFY;
2059 		dev->link_mode = value;
2060 		write_unlock_bh(&dev_base_lock);
2061 	}
2062 
2063 	if (tb[IFLA_VFINFO_LIST]) {
2064 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2065 		struct nlattr *attr;
2066 		int rem;
2067 
2068 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2069 			if (nla_type(attr) != IFLA_VF_INFO ||
2070 			    nla_len(attr) < NLA_HDRLEN) {
2071 				err = -EINVAL;
2072 				goto errout;
2073 			}
2074 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2075 					       ifla_vf_policy);
2076 			if (err < 0)
2077 				goto errout;
2078 			err = do_setvfinfo(dev, vfinfo);
2079 			if (err < 0)
2080 				goto errout;
2081 			status |= DO_SETLINK_NOTIFY;
2082 		}
2083 	}
2084 	err = 0;
2085 
2086 	if (tb[IFLA_VF_PORTS]) {
2087 		struct nlattr *port[IFLA_PORT_MAX+1];
2088 		struct nlattr *attr;
2089 		int vf;
2090 		int rem;
2091 
2092 		err = -EOPNOTSUPP;
2093 		if (!ops->ndo_set_vf_port)
2094 			goto errout;
2095 
2096 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2097 			if (nla_type(attr) != IFLA_VF_PORT ||
2098 			    nla_len(attr) < NLA_HDRLEN) {
2099 				err = -EINVAL;
2100 				goto errout;
2101 			}
2102 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2103 					       ifla_port_policy);
2104 			if (err < 0)
2105 				goto errout;
2106 			if (!port[IFLA_PORT_VF]) {
2107 				err = -EOPNOTSUPP;
2108 				goto errout;
2109 			}
2110 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2111 			err = ops->ndo_set_vf_port(dev, vf, port);
2112 			if (err < 0)
2113 				goto errout;
2114 			status |= DO_SETLINK_NOTIFY;
2115 		}
2116 	}
2117 	err = 0;
2118 
2119 	if (tb[IFLA_PORT_SELF]) {
2120 		struct nlattr *port[IFLA_PORT_MAX+1];
2121 
2122 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2123 			tb[IFLA_PORT_SELF], ifla_port_policy);
2124 		if (err < 0)
2125 			goto errout;
2126 
2127 		err = -EOPNOTSUPP;
2128 		if (ops->ndo_set_vf_port)
2129 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2130 		if (err < 0)
2131 			goto errout;
2132 		status |= DO_SETLINK_NOTIFY;
2133 	}
2134 
2135 	if (tb[IFLA_AF_SPEC]) {
2136 		struct nlattr *af;
2137 		int rem;
2138 
2139 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2140 			const struct rtnl_af_ops *af_ops;
2141 
2142 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2143 				BUG();
2144 
2145 			err = af_ops->set_link_af(dev, af);
2146 			if (err < 0)
2147 				goto errout;
2148 
2149 			status |= DO_SETLINK_NOTIFY;
2150 		}
2151 	}
2152 	err = 0;
2153 
2154 	if (tb[IFLA_PROTO_DOWN]) {
2155 		err = dev_change_proto_down(dev,
2156 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2157 		if (err)
2158 			goto errout;
2159 		status |= DO_SETLINK_NOTIFY;
2160 	}
2161 
2162 	if (tb[IFLA_XDP]) {
2163 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2164 
2165 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2166 				       ifla_xdp_policy);
2167 		if (err < 0)
2168 			goto errout;
2169 
2170 		if (xdp[IFLA_XDP_ATTACHED]) {
2171 			err = -EINVAL;
2172 			goto errout;
2173 		}
2174 		if (xdp[IFLA_XDP_FD]) {
2175 			err = dev_change_xdp_fd(dev,
2176 						nla_get_s32(xdp[IFLA_XDP_FD]));
2177 			if (err)
2178 				goto errout;
2179 			status |= DO_SETLINK_NOTIFY;
2180 		}
2181 	}
2182 
2183 errout:
2184 	if (status & DO_SETLINK_MODIFIED) {
2185 		if (status & DO_SETLINK_NOTIFY)
2186 			netdev_state_change(dev);
2187 
2188 		if (err < 0)
2189 			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",
2190 					     dev->name);
2191 	}
2192 
2193 	return err;
2194 }
2195 
2196 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2197 {
2198 	struct net *net = sock_net(skb->sk);
2199 	struct ifinfomsg *ifm;
2200 	struct net_device *dev;
2201 	int err;
2202 	struct nlattr *tb[IFLA_MAX+1];
2203 	char ifname[IFNAMSIZ];
2204 
2205 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2206 	if (err < 0)
2207 		goto errout;
2208 
2209 	if (tb[IFLA_IFNAME])
2210 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2211 	else
2212 		ifname[0] = '\0';
2213 
2214 	err = -EINVAL;
2215 	ifm = nlmsg_data(nlh);
2216 	if (ifm->ifi_index > 0)
2217 		dev = __dev_get_by_index(net, ifm->ifi_index);
2218 	else if (tb[IFLA_IFNAME])
2219 		dev = __dev_get_by_name(net, ifname);
2220 	else
2221 		goto errout;
2222 
2223 	if (dev == NULL) {
2224 		err = -ENODEV;
2225 		goto errout;
2226 	}
2227 
2228 	err = validate_linkmsg(dev, tb);
2229 	if (err < 0)
2230 		goto errout;
2231 
2232 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2233 errout:
2234 	return err;
2235 }
2236 
2237 static int rtnl_group_dellink(const struct net *net, int group)
2238 {
2239 	struct net_device *dev, *aux;
2240 	LIST_HEAD(list_kill);
2241 	bool found = false;
2242 
2243 	if (!group)
2244 		return -EPERM;
2245 
2246 	for_each_netdev(net, dev) {
2247 		if (dev->group == group) {
2248 			const struct rtnl_link_ops *ops;
2249 
2250 			found = true;
2251 			ops = dev->rtnl_link_ops;
2252 			if (!ops || !ops->dellink)
2253 				return -EOPNOTSUPP;
2254 		}
2255 	}
2256 
2257 	if (!found)
2258 		return -ENODEV;
2259 
2260 	for_each_netdev_safe(net, dev, aux) {
2261 		if (dev->group == group) {
2262 			const struct rtnl_link_ops *ops;
2263 
2264 			ops = dev->rtnl_link_ops;
2265 			ops->dellink(dev, &list_kill);
2266 		}
2267 	}
2268 	unregister_netdevice_many(&list_kill);
2269 
2270 	return 0;
2271 }
2272 
2273 int rtnl_delete_link(struct net_device *dev)
2274 {
2275 	const struct rtnl_link_ops *ops;
2276 	LIST_HEAD(list_kill);
2277 
2278 	ops = dev->rtnl_link_ops;
2279 	if (!ops || !ops->dellink)
2280 		return -EOPNOTSUPP;
2281 
2282 	ops->dellink(dev, &list_kill);
2283 	unregister_netdevice_many(&list_kill);
2284 
2285 	return 0;
2286 }
2287 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2288 
2289 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2290 {
2291 	struct net *net = sock_net(skb->sk);
2292 	struct net_device *dev;
2293 	struct ifinfomsg *ifm;
2294 	char ifname[IFNAMSIZ];
2295 	struct nlattr *tb[IFLA_MAX+1];
2296 	int err;
2297 
2298 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2299 	if (err < 0)
2300 		return err;
2301 
2302 	if (tb[IFLA_IFNAME])
2303 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2304 
2305 	ifm = nlmsg_data(nlh);
2306 	if (ifm->ifi_index > 0)
2307 		dev = __dev_get_by_index(net, ifm->ifi_index);
2308 	else if (tb[IFLA_IFNAME])
2309 		dev = __dev_get_by_name(net, ifname);
2310 	else if (tb[IFLA_GROUP])
2311 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2312 	else
2313 		return -EINVAL;
2314 
2315 	if (!dev)
2316 		return -ENODEV;
2317 
2318 	return rtnl_delete_link(dev);
2319 }
2320 
2321 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2322 {
2323 	unsigned int old_flags;
2324 	int err;
2325 
2326 	old_flags = dev->flags;
2327 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2328 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2329 		if (err < 0)
2330 			return err;
2331 	}
2332 
2333 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2334 
2335 	__dev_notify_flags(dev, old_flags, ~0U);
2336 	return 0;
2337 }
2338 EXPORT_SYMBOL(rtnl_configure_link);
2339 
2340 struct net_device *rtnl_create_link(struct net *net,
2341 	const char *ifname, unsigned char name_assign_type,
2342 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2343 {
2344 	int err;
2345 	struct net_device *dev;
2346 	unsigned int num_tx_queues = 1;
2347 	unsigned int num_rx_queues = 1;
2348 
2349 	if (tb[IFLA_NUM_TX_QUEUES])
2350 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2351 	else if (ops->get_num_tx_queues)
2352 		num_tx_queues = ops->get_num_tx_queues();
2353 
2354 	if (tb[IFLA_NUM_RX_QUEUES])
2355 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2356 	else if (ops->get_num_rx_queues)
2357 		num_rx_queues = ops->get_num_rx_queues();
2358 
2359 	err = -ENOMEM;
2360 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2361 			       ops->setup, num_tx_queues, num_rx_queues);
2362 	if (!dev)
2363 		goto err;
2364 
2365 	dev_net_set(dev, net);
2366 	dev->rtnl_link_ops = ops;
2367 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2368 
2369 	if (tb[IFLA_MTU])
2370 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2371 	if (tb[IFLA_ADDRESS]) {
2372 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2373 				nla_len(tb[IFLA_ADDRESS]));
2374 		dev->addr_assign_type = NET_ADDR_SET;
2375 	}
2376 	if (tb[IFLA_BROADCAST])
2377 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2378 				nla_len(tb[IFLA_BROADCAST]));
2379 	if (tb[IFLA_TXQLEN])
2380 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2381 	if (tb[IFLA_OPERSTATE])
2382 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2383 	if (tb[IFLA_LINKMODE])
2384 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2385 	if (tb[IFLA_GROUP])
2386 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2387 
2388 	return dev;
2389 
2390 err:
2391 	return ERR_PTR(err);
2392 }
2393 EXPORT_SYMBOL(rtnl_create_link);
2394 
2395 static int rtnl_group_changelink(const struct sk_buff *skb,
2396 		struct net *net, int group,
2397 		struct ifinfomsg *ifm,
2398 		struct nlattr **tb)
2399 {
2400 	struct net_device *dev, *aux;
2401 	int err;
2402 
2403 	for_each_netdev_safe(net, dev, aux) {
2404 		if (dev->group == group) {
2405 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2406 			if (err < 0)
2407 				return err;
2408 		}
2409 	}
2410 
2411 	return 0;
2412 }
2413 
2414 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2415 {
2416 	struct net *net = sock_net(skb->sk);
2417 	const struct rtnl_link_ops *ops;
2418 	const struct rtnl_link_ops *m_ops = NULL;
2419 	struct net_device *dev;
2420 	struct net_device *master_dev = NULL;
2421 	struct ifinfomsg *ifm;
2422 	char kind[MODULE_NAME_LEN];
2423 	char ifname[IFNAMSIZ];
2424 	struct nlattr *tb[IFLA_MAX+1];
2425 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2426 	unsigned char name_assign_type = NET_NAME_USER;
2427 	int err;
2428 
2429 #ifdef CONFIG_MODULES
2430 replay:
2431 #endif
2432 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2433 	if (err < 0)
2434 		return err;
2435 
2436 	if (tb[IFLA_IFNAME])
2437 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2438 	else
2439 		ifname[0] = '\0';
2440 
2441 	ifm = nlmsg_data(nlh);
2442 	if (ifm->ifi_index > 0)
2443 		dev = __dev_get_by_index(net, ifm->ifi_index);
2444 	else {
2445 		if (ifname[0])
2446 			dev = __dev_get_by_name(net, ifname);
2447 		else
2448 			dev = NULL;
2449 	}
2450 
2451 	if (dev) {
2452 		master_dev = netdev_master_upper_dev_get(dev);
2453 		if (master_dev)
2454 			m_ops = master_dev->rtnl_link_ops;
2455 	}
2456 
2457 	err = validate_linkmsg(dev, tb);
2458 	if (err < 0)
2459 		return err;
2460 
2461 	if (tb[IFLA_LINKINFO]) {
2462 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2463 				       tb[IFLA_LINKINFO], ifla_info_policy);
2464 		if (err < 0)
2465 			return err;
2466 	} else
2467 		memset(linkinfo, 0, sizeof(linkinfo));
2468 
2469 	if (linkinfo[IFLA_INFO_KIND]) {
2470 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2471 		ops = rtnl_link_ops_get(kind);
2472 	} else {
2473 		kind[0] = '\0';
2474 		ops = NULL;
2475 	}
2476 
2477 	if (1) {
2478 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2479 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2480 		struct nlattr **data = NULL;
2481 		struct nlattr **slave_data = NULL;
2482 		struct net *dest_net, *link_net = NULL;
2483 
2484 		if (ops) {
2485 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2486 				err = nla_parse_nested(attr, ops->maxtype,
2487 						       linkinfo[IFLA_INFO_DATA],
2488 						       ops->policy);
2489 				if (err < 0)
2490 					return err;
2491 				data = attr;
2492 			}
2493 			if (ops->validate) {
2494 				err = ops->validate(tb, data);
2495 				if (err < 0)
2496 					return err;
2497 			}
2498 		}
2499 
2500 		if (m_ops) {
2501 			if (m_ops->slave_maxtype &&
2502 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2503 				err = nla_parse_nested(slave_attr,
2504 						       m_ops->slave_maxtype,
2505 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2506 						       m_ops->slave_policy);
2507 				if (err < 0)
2508 					return err;
2509 				slave_data = slave_attr;
2510 			}
2511 			if (m_ops->slave_validate) {
2512 				err = m_ops->slave_validate(tb, slave_data);
2513 				if (err < 0)
2514 					return err;
2515 			}
2516 		}
2517 
2518 		if (dev) {
2519 			int status = 0;
2520 
2521 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2522 				return -EEXIST;
2523 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2524 				return -EOPNOTSUPP;
2525 
2526 			if (linkinfo[IFLA_INFO_DATA]) {
2527 				if (!ops || ops != dev->rtnl_link_ops ||
2528 				    !ops->changelink)
2529 					return -EOPNOTSUPP;
2530 
2531 				err = ops->changelink(dev, tb, data);
2532 				if (err < 0)
2533 					return err;
2534 				status |= DO_SETLINK_NOTIFY;
2535 			}
2536 
2537 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2538 				if (!m_ops || !m_ops->slave_changelink)
2539 					return -EOPNOTSUPP;
2540 
2541 				err = m_ops->slave_changelink(master_dev, dev,
2542 							      tb, slave_data);
2543 				if (err < 0)
2544 					return err;
2545 				status |= DO_SETLINK_NOTIFY;
2546 			}
2547 
2548 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2549 		}
2550 
2551 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2552 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2553 				return rtnl_group_changelink(skb, net,
2554 						nla_get_u32(tb[IFLA_GROUP]),
2555 						ifm, tb);
2556 			return -ENODEV;
2557 		}
2558 
2559 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2560 			return -EOPNOTSUPP;
2561 
2562 		if (!ops) {
2563 #ifdef CONFIG_MODULES
2564 			if (kind[0]) {
2565 				__rtnl_unlock();
2566 				request_module("rtnl-link-%s", kind);
2567 				rtnl_lock();
2568 				ops = rtnl_link_ops_get(kind);
2569 				if (ops)
2570 					goto replay;
2571 			}
2572 #endif
2573 			return -EOPNOTSUPP;
2574 		}
2575 
2576 		if (!ops->setup)
2577 			return -EOPNOTSUPP;
2578 
2579 		if (!ifname[0]) {
2580 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2581 			name_assign_type = NET_NAME_ENUM;
2582 		}
2583 
2584 		dest_net = rtnl_link_get_net(net, tb);
2585 		if (IS_ERR(dest_net))
2586 			return PTR_ERR(dest_net);
2587 
2588 		err = -EPERM;
2589 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2590 			goto out;
2591 
2592 		if (tb[IFLA_LINK_NETNSID]) {
2593 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2594 
2595 			link_net = get_net_ns_by_id(dest_net, id);
2596 			if (!link_net) {
2597 				err =  -EINVAL;
2598 				goto out;
2599 			}
2600 			err = -EPERM;
2601 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2602 				goto out;
2603 		}
2604 
2605 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2606 				       name_assign_type, ops, tb);
2607 		if (IS_ERR(dev)) {
2608 			err = PTR_ERR(dev);
2609 			goto out;
2610 		}
2611 
2612 		dev->ifindex = ifm->ifi_index;
2613 
2614 		if (ops->newlink) {
2615 			err = ops->newlink(link_net ? : net, dev, tb, data);
2616 			/* Drivers should call free_netdev() in ->destructor
2617 			 * and unregister it on failure after registration
2618 			 * so that device could be finally freed in rtnl_unlock.
2619 			 */
2620 			if (err < 0) {
2621 				/* If device is not registered at all, free it now */
2622 				if (dev->reg_state == NETREG_UNINITIALIZED)
2623 					free_netdev(dev);
2624 				goto out;
2625 			}
2626 		} else {
2627 			err = register_netdevice(dev);
2628 			if (err < 0) {
2629 				free_netdev(dev);
2630 				goto out;
2631 			}
2632 		}
2633 		err = rtnl_configure_link(dev, ifm);
2634 		if (err < 0)
2635 			goto out_unregister;
2636 		if (link_net) {
2637 			err = dev_change_net_namespace(dev, dest_net, ifname);
2638 			if (err < 0)
2639 				goto out_unregister;
2640 		}
2641 out:
2642 		if (link_net)
2643 			put_net(link_net);
2644 		put_net(dest_net);
2645 		return err;
2646 out_unregister:
2647 		if (ops->newlink) {
2648 			LIST_HEAD(list_kill);
2649 
2650 			ops->dellink(dev, &list_kill);
2651 			unregister_netdevice_many(&list_kill);
2652 		} else {
2653 			unregister_netdevice(dev);
2654 		}
2655 		goto out;
2656 	}
2657 }
2658 
2659 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2660 {
2661 	struct net *net = sock_net(skb->sk);
2662 	struct ifinfomsg *ifm;
2663 	char ifname[IFNAMSIZ];
2664 	struct nlattr *tb[IFLA_MAX+1];
2665 	struct net_device *dev = NULL;
2666 	struct sk_buff *nskb;
2667 	int err;
2668 	u32 ext_filter_mask = 0;
2669 
2670 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2671 	if (err < 0)
2672 		return err;
2673 
2674 	if (tb[IFLA_IFNAME])
2675 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2676 
2677 	if (tb[IFLA_EXT_MASK])
2678 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2679 
2680 	ifm = nlmsg_data(nlh);
2681 	if (ifm->ifi_index > 0)
2682 		dev = __dev_get_by_index(net, ifm->ifi_index);
2683 	else if (tb[IFLA_IFNAME])
2684 		dev = __dev_get_by_name(net, ifname);
2685 	else
2686 		return -EINVAL;
2687 
2688 	if (dev == NULL)
2689 		return -ENODEV;
2690 
2691 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2692 	if (nskb == NULL)
2693 		return -ENOBUFS;
2694 
2695 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2696 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2697 	if (err < 0) {
2698 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2699 		WARN_ON(err == -EMSGSIZE);
2700 		kfree_skb(nskb);
2701 	} else
2702 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2703 
2704 	return err;
2705 }
2706 
2707 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2708 {
2709 	struct net *net = sock_net(skb->sk);
2710 	struct net_device *dev;
2711 	struct nlattr *tb[IFLA_MAX+1];
2712 	u32 ext_filter_mask = 0;
2713 	u16 min_ifinfo_dump_size = 0;
2714 	int hdrlen;
2715 
2716 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2717 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2718 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2719 
2720 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2721 		if (tb[IFLA_EXT_MASK])
2722 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2723 	}
2724 
2725 	if (!ext_filter_mask)
2726 		return NLMSG_GOODSIZE;
2727 	/*
2728 	 * traverse the list of net devices and compute the minimum
2729 	 * buffer size based upon the filter mask.
2730 	 */
2731 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2732 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2733 					     if_nlmsg_size(dev,
2734 						           ext_filter_mask));
2735 	}
2736 
2737 	return min_ifinfo_dump_size;
2738 }
2739 
2740 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2741 {
2742 	int idx;
2743 	int s_idx = cb->family;
2744 
2745 	if (s_idx == 0)
2746 		s_idx = 1;
2747 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2748 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2749 		if (idx < s_idx || idx == PF_PACKET)
2750 			continue;
2751 		if (rtnl_msg_handlers[idx] == NULL ||
2752 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2753 			continue;
2754 		if (idx > s_idx) {
2755 			memset(&cb->args[0], 0, sizeof(cb->args));
2756 			cb->prev_seq = 0;
2757 			cb->seq = 0;
2758 		}
2759 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2760 			break;
2761 	}
2762 	cb->family = idx;
2763 
2764 	return skb->len;
2765 }
2766 
2767 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2768 				       unsigned int change, gfp_t flags)
2769 {
2770 	struct net *net = dev_net(dev);
2771 	struct sk_buff *skb;
2772 	int err = -ENOBUFS;
2773 	size_t if_info_size;
2774 
2775 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2776 	if (skb == NULL)
2777 		goto errout;
2778 
2779 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2780 	if (err < 0) {
2781 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2782 		WARN_ON(err == -EMSGSIZE);
2783 		kfree_skb(skb);
2784 		goto errout;
2785 	}
2786 	return skb;
2787 errout:
2788 	if (err < 0)
2789 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2790 	return NULL;
2791 }
2792 
2793 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2794 {
2795 	struct net *net = dev_net(dev);
2796 
2797 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2798 }
2799 
2800 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2801 		  gfp_t flags)
2802 {
2803 	struct sk_buff *skb;
2804 
2805 	if (dev->reg_state != NETREG_REGISTERED)
2806 		return;
2807 
2808 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2809 	if (skb)
2810 		rtmsg_ifinfo_send(skb, dev, flags);
2811 }
2812 EXPORT_SYMBOL(rtmsg_ifinfo);
2813 
2814 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2815 				   struct net_device *dev,
2816 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2817 				   int type, unsigned int flags,
2818 				   int nlflags, u16 ndm_state)
2819 {
2820 	struct nlmsghdr *nlh;
2821 	struct ndmsg *ndm;
2822 
2823 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2824 	if (!nlh)
2825 		return -EMSGSIZE;
2826 
2827 	ndm = nlmsg_data(nlh);
2828 	ndm->ndm_family  = AF_BRIDGE;
2829 	ndm->ndm_pad1	 = 0;
2830 	ndm->ndm_pad2    = 0;
2831 	ndm->ndm_flags	 = flags;
2832 	ndm->ndm_type	 = 0;
2833 	ndm->ndm_ifindex = dev->ifindex;
2834 	ndm->ndm_state   = ndm_state;
2835 
2836 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2837 		goto nla_put_failure;
2838 	if (vid)
2839 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2840 			goto nla_put_failure;
2841 
2842 	nlmsg_end(skb, nlh);
2843 	return 0;
2844 
2845 nla_put_failure:
2846 	nlmsg_cancel(skb, nlh);
2847 	return -EMSGSIZE;
2848 }
2849 
2850 static inline size_t rtnl_fdb_nlmsg_size(void)
2851 {
2852 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2853 }
2854 
2855 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2856 			    u16 ndm_state)
2857 {
2858 	struct net *net = dev_net(dev);
2859 	struct sk_buff *skb;
2860 	int err = -ENOBUFS;
2861 
2862 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2863 	if (!skb)
2864 		goto errout;
2865 
2866 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2867 				      0, 0, type, NTF_SELF, 0, ndm_state);
2868 	if (err < 0) {
2869 		kfree_skb(skb);
2870 		goto errout;
2871 	}
2872 
2873 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2874 	return;
2875 errout:
2876 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2877 }
2878 
2879 /**
2880  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2881  */
2882 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2883 		     struct nlattr *tb[],
2884 		     struct net_device *dev,
2885 		     const unsigned char *addr, u16 vid,
2886 		     u16 flags)
2887 {
2888 	int err = -EINVAL;
2889 
2890 	/* If aging addresses are supported device will need to
2891 	 * implement its own handler for this.
2892 	 */
2893 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2894 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2895 		return err;
2896 	}
2897 
2898 	if (vid) {
2899 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2900 		return err;
2901 	}
2902 
2903 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2904 		err = dev_uc_add_excl(dev, addr);
2905 	else if (is_multicast_ether_addr(addr))
2906 		err = dev_mc_add_excl(dev, addr);
2907 
2908 	/* Only return duplicate errors if NLM_F_EXCL is set */
2909 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2910 		err = 0;
2911 
2912 	return err;
2913 }
2914 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2915 
2916 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2917 {
2918 	u16 vid = 0;
2919 
2920 	if (vlan_attr) {
2921 		if (nla_len(vlan_attr) != sizeof(u16)) {
2922 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2923 			return -EINVAL;
2924 		}
2925 
2926 		vid = nla_get_u16(vlan_attr);
2927 
2928 		if (!vid || vid >= VLAN_VID_MASK) {
2929 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2930 				vid);
2931 			return -EINVAL;
2932 		}
2933 	}
2934 	*p_vid = vid;
2935 	return 0;
2936 }
2937 
2938 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2939 {
2940 	struct net *net = sock_net(skb->sk);
2941 	struct ndmsg *ndm;
2942 	struct nlattr *tb[NDA_MAX+1];
2943 	struct net_device *dev;
2944 	u8 *addr;
2945 	u16 vid;
2946 	int err;
2947 
2948 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2949 	if (err < 0)
2950 		return err;
2951 
2952 	ndm = nlmsg_data(nlh);
2953 	if (ndm->ndm_ifindex == 0) {
2954 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2955 		return -EINVAL;
2956 	}
2957 
2958 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2959 	if (dev == NULL) {
2960 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2961 		return -ENODEV;
2962 	}
2963 
2964 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2965 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2966 		return -EINVAL;
2967 	}
2968 
2969 	addr = nla_data(tb[NDA_LLADDR]);
2970 
2971 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2972 	if (err)
2973 		return err;
2974 
2975 	err = -EOPNOTSUPP;
2976 
2977 	/* Support fdb on master device the net/bridge default case */
2978 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2979 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2980 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2981 		const struct net_device_ops *ops = br_dev->netdev_ops;
2982 
2983 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2984 				       nlh->nlmsg_flags);
2985 		if (err)
2986 			goto out;
2987 		else
2988 			ndm->ndm_flags &= ~NTF_MASTER;
2989 	}
2990 
2991 	/* Embedded bridge, macvlan, and any other device support */
2992 	if ((ndm->ndm_flags & NTF_SELF)) {
2993 		if (dev->netdev_ops->ndo_fdb_add)
2994 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2995 							   vid,
2996 							   nlh->nlmsg_flags);
2997 		else
2998 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2999 					       nlh->nlmsg_flags);
3000 
3001 		if (!err) {
3002 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3003 					ndm->ndm_state);
3004 			ndm->ndm_flags &= ~NTF_SELF;
3005 		}
3006 	}
3007 out:
3008 	return err;
3009 }
3010 
3011 /**
3012  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3013  */
3014 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3015 		     struct nlattr *tb[],
3016 		     struct net_device *dev,
3017 		     const unsigned char *addr, u16 vid)
3018 {
3019 	int err = -EINVAL;
3020 
3021 	/* If aging addresses are supported device will need to
3022 	 * implement its own handler for this.
3023 	 */
3024 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3025 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3026 		return err;
3027 	}
3028 
3029 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3030 		err = dev_uc_del(dev, addr);
3031 	else if (is_multicast_ether_addr(addr))
3032 		err = dev_mc_del(dev, addr);
3033 
3034 	return err;
3035 }
3036 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3037 
3038 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3039 {
3040 	struct net *net = sock_net(skb->sk);
3041 	struct ndmsg *ndm;
3042 	struct nlattr *tb[NDA_MAX+1];
3043 	struct net_device *dev;
3044 	int err = -EINVAL;
3045 	__u8 *addr;
3046 	u16 vid;
3047 
3048 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3049 		return -EPERM;
3050 
3051 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3052 	if (err < 0)
3053 		return err;
3054 
3055 	ndm = nlmsg_data(nlh);
3056 	if (ndm->ndm_ifindex == 0) {
3057 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3058 		return -EINVAL;
3059 	}
3060 
3061 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3062 	if (dev == NULL) {
3063 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3064 		return -ENODEV;
3065 	}
3066 
3067 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3068 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3069 		return -EINVAL;
3070 	}
3071 
3072 	addr = nla_data(tb[NDA_LLADDR]);
3073 
3074 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3075 	if (err)
3076 		return err;
3077 
3078 	err = -EOPNOTSUPP;
3079 
3080 	/* Support fdb on master device the net/bridge default case */
3081 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3082 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3083 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3084 		const struct net_device_ops *ops = br_dev->netdev_ops;
3085 
3086 		if (ops->ndo_fdb_del)
3087 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3088 
3089 		if (err)
3090 			goto out;
3091 		else
3092 			ndm->ndm_flags &= ~NTF_MASTER;
3093 	}
3094 
3095 	/* Embedded bridge, macvlan, and any other device support */
3096 	if (ndm->ndm_flags & NTF_SELF) {
3097 		if (dev->netdev_ops->ndo_fdb_del)
3098 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3099 							   vid);
3100 		else
3101 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3102 
3103 		if (!err) {
3104 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3105 					ndm->ndm_state);
3106 			ndm->ndm_flags &= ~NTF_SELF;
3107 		}
3108 	}
3109 out:
3110 	return err;
3111 }
3112 
3113 static int nlmsg_populate_fdb(struct sk_buff *skb,
3114 			      struct netlink_callback *cb,
3115 			      struct net_device *dev,
3116 			      int *idx,
3117 			      struct netdev_hw_addr_list *list)
3118 {
3119 	struct netdev_hw_addr *ha;
3120 	int err;
3121 	u32 portid, seq;
3122 
3123 	portid = NETLINK_CB(cb->skb).portid;
3124 	seq = cb->nlh->nlmsg_seq;
3125 
3126 	list_for_each_entry(ha, &list->list, list) {
3127 		if (*idx < cb->args[2])
3128 			goto skip;
3129 
3130 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3131 					      portid, seq,
3132 					      RTM_NEWNEIGH, NTF_SELF,
3133 					      NLM_F_MULTI, NUD_PERMANENT);
3134 		if (err < 0)
3135 			return err;
3136 skip:
3137 		*idx += 1;
3138 	}
3139 	return 0;
3140 }
3141 
3142 /**
3143  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3144  * @nlh: netlink message header
3145  * @dev: netdevice
3146  *
3147  * Default netdevice operation to dump the existing unicast address list.
3148  * Returns number of addresses from list put in skb.
3149  */
3150 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3151 		      struct netlink_callback *cb,
3152 		      struct net_device *dev,
3153 		      struct net_device *filter_dev,
3154 		      int *idx)
3155 {
3156 	int err;
3157 
3158 	netif_addr_lock_bh(dev);
3159 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3160 	if (err)
3161 		goto out;
3162 	nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3163 out:
3164 	netif_addr_unlock_bh(dev);
3165 	return err;
3166 }
3167 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3168 
3169 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3170 {
3171 	struct net_device *dev;
3172 	struct nlattr *tb[IFLA_MAX+1];
3173 	struct net_device *br_dev = NULL;
3174 	const struct net_device_ops *ops = NULL;
3175 	const struct net_device_ops *cops = NULL;
3176 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3177 	struct net *net = sock_net(skb->sk);
3178 	struct hlist_head *head;
3179 	int brport_idx = 0;
3180 	int br_idx = 0;
3181 	int h, s_h;
3182 	int idx = 0, s_idx;
3183 	int err = 0;
3184 	int fidx = 0;
3185 
3186 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3187 			ifla_policy) == 0) {
3188 		if (tb[IFLA_MASTER])
3189 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3190 	}
3191 
3192 	brport_idx = ifm->ifi_index;
3193 
3194 	if (br_idx) {
3195 		br_dev = __dev_get_by_index(net, br_idx);
3196 		if (!br_dev)
3197 			return -ENODEV;
3198 
3199 		ops = br_dev->netdev_ops;
3200 	}
3201 
3202 	s_h = cb->args[0];
3203 	s_idx = cb->args[1];
3204 
3205 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3206 		idx = 0;
3207 		head = &net->dev_index_head[h];
3208 		hlist_for_each_entry(dev, head, index_hlist) {
3209 
3210 			if (brport_idx && (dev->ifindex != brport_idx))
3211 				continue;
3212 
3213 			if (!br_idx) { /* user did not specify a specific bridge */
3214 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3215 					br_dev = netdev_master_upper_dev_get(dev);
3216 					cops = br_dev->netdev_ops;
3217 				}
3218 			} else {
3219 				if (dev != br_dev &&
3220 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3221 					continue;
3222 
3223 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3224 				    !(dev->priv_flags & IFF_EBRIDGE))
3225 					continue;
3226 				cops = ops;
3227 			}
3228 
3229 			if (idx < s_idx)
3230 				goto cont;
3231 
3232 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3233 				if (cops && cops->ndo_fdb_dump) {
3234 					err = cops->ndo_fdb_dump(skb, cb,
3235 								br_dev, dev,
3236 								&fidx);
3237 					if (err == -EMSGSIZE)
3238 						goto out;
3239 				}
3240 			}
3241 
3242 			if (dev->netdev_ops->ndo_fdb_dump)
3243 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3244 								    dev, NULL,
3245 								    &fidx);
3246 			else
3247 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3248 							&fidx);
3249 			if (err == -EMSGSIZE)
3250 				goto out;
3251 
3252 			cops = NULL;
3253 
3254 			/* reset fdb offset to 0 for rest of the interfaces */
3255 			cb->args[2] = 0;
3256 			fidx = 0;
3257 cont:
3258 			idx++;
3259 		}
3260 	}
3261 
3262 out:
3263 	cb->args[0] = h;
3264 	cb->args[1] = idx;
3265 	cb->args[2] = fidx;
3266 
3267 	return skb->len;
3268 }
3269 
3270 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3271 			       unsigned int attrnum, unsigned int flag)
3272 {
3273 	if (mask & flag)
3274 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3275 	return 0;
3276 }
3277 
3278 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3279 			    struct net_device *dev, u16 mode,
3280 			    u32 flags, u32 mask, int nlflags,
3281 			    u32 filter_mask,
3282 			    int (*vlan_fill)(struct sk_buff *skb,
3283 					     struct net_device *dev,
3284 					     u32 filter_mask))
3285 {
3286 	struct nlmsghdr *nlh;
3287 	struct ifinfomsg *ifm;
3288 	struct nlattr *br_afspec;
3289 	struct nlattr *protinfo;
3290 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3291 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3292 	int err = 0;
3293 
3294 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3295 	if (nlh == NULL)
3296 		return -EMSGSIZE;
3297 
3298 	ifm = nlmsg_data(nlh);
3299 	ifm->ifi_family = AF_BRIDGE;
3300 	ifm->__ifi_pad = 0;
3301 	ifm->ifi_type = dev->type;
3302 	ifm->ifi_index = dev->ifindex;
3303 	ifm->ifi_flags = dev_get_flags(dev);
3304 	ifm->ifi_change = 0;
3305 
3306 
3307 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3308 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3309 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3310 	    (br_dev &&
3311 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3312 	    (dev->addr_len &&
3313 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3314 	    (dev->ifindex != dev_get_iflink(dev) &&
3315 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3316 		goto nla_put_failure;
3317 
3318 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3319 	if (!br_afspec)
3320 		goto nla_put_failure;
3321 
3322 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3323 		nla_nest_cancel(skb, br_afspec);
3324 		goto nla_put_failure;
3325 	}
3326 
3327 	if (mode != BRIDGE_MODE_UNDEF) {
3328 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3329 			nla_nest_cancel(skb, br_afspec);
3330 			goto nla_put_failure;
3331 		}
3332 	}
3333 	if (vlan_fill) {
3334 		err = vlan_fill(skb, dev, filter_mask);
3335 		if (err) {
3336 			nla_nest_cancel(skb, br_afspec);
3337 			goto nla_put_failure;
3338 		}
3339 	}
3340 	nla_nest_end(skb, br_afspec);
3341 
3342 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3343 	if (!protinfo)
3344 		goto nla_put_failure;
3345 
3346 	if (brport_nla_put_flag(skb, flags, mask,
3347 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3348 	    brport_nla_put_flag(skb, flags, mask,
3349 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3350 	    brport_nla_put_flag(skb, flags, mask,
3351 				IFLA_BRPORT_FAST_LEAVE,
3352 				BR_MULTICAST_FAST_LEAVE) ||
3353 	    brport_nla_put_flag(skb, flags, mask,
3354 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3355 	    brport_nla_put_flag(skb, flags, mask,
3356 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3357 	    brport_nla_put_flag(skb, flags, mask,
3358 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3359 	    brport_nla_put_flag(skb, flags, mask,
3360 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3361 	    brport_nla_put_flag(skb, flags, mask,
3362 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3363 		nla_nest_cancel(skb, protinfo);
3364 		goto nla_put_failure;
3365 	}
3366 
3367 	nla_nest_end(skb, protinfo);
3368 
3369 	nlmsg_end(skb, nlh);
3370 	return 0;
3371 nla_put_failure:
3372 	nlmsg_cancel(skb, nlh);
3373 	return err ? err : -EMSGSIZE;
3374 }
3375 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3376 
3377 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3378 {
3379 	struct net *net = sock_net(skb->sk);
3380 	struct net_device *dev;
3381 	int idx = 0;
3382 	u32 portid = NETLINK_CB(cb->skb).portid;
3383 	u32 seq = cb->nlh->nlmsg_seq;
3384 	u32 filter_mask = 0;
3385 	int err;
3386 
3387 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3388 		struct nlattr *extfilt;
3389 
3390 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3391 					  IFLA_EXT_MASK);
3392 		if (extfilt) {
3393 			if (nla_len(extfilt) < sizeof(filter_mask))
3394 				return -EINVAL;
3395 
3396 			filter_mask = nla_get_u32(extfilt);
3397 		}
3398 	}
3399 
3400 	rcu_read_lock();
3401 	for_each_netdev_rcu(net, dev) {
3402 		const struct net_device_ops *ops = dev->netdev_ops;
3403 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3404 
3405 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3406 			if (idx >= cb->args[0]) {
3407 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3408 						skb, portid, seq, dev,
3409 						filter_mask, NLM_F_MULTI);
3410 				if (err < 0 && err != -EOPNOTSUPP)
3411 					break;
3412 			}
3413 			idx++;
3414 		}
3415 
3416 		if (ops->ndo_bridge_getlink) {
3417 			if (idx >= cb->args[0]) {
3418 				err = ops->ndo_bridge_getlink(skb, portid,
3419 							      seq, dev,
3420 							      filter_mask,
3421 							      NLM_F_MULTI);
3422 				if (err < 0 && err != -EOPNOTSUPP)
3423 					break;
3424 			}
3425 			idx++;
3426 		}
3427 	}
3428 	rcu_read_unlock();
3429 	cb->args[0] = idx;
3430 
3431 	return skb->len;
3432 }
3433 
3434 static inline size_t bridge_nlmsg_size(void)
3435 {
3436 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3437 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3438 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3439 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3440 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3441 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3442 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3443 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3444 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3445 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3446 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3447 }
3448 
3449 static int rtnl_bridge_notify(struct net_device *dev)
3450 {
3451 	struct net *net = dev_net(dev);
3452 	struct sk_buff *skb;
3453 	int err = -EOPNOTSUPP;
3454 
3455 	if (!dev->netdev_ops->ndo_bridge_getlink)
3456 		return 0;
3457 
3458 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3459 	if (!skb) {
3460 		err = -ENOMEM;
3461 		goto errout;
3462 	}
3463 
3464 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3465 	if (err < 0)
3466 		goto errout;
3467 
3468 	if (!skb->len)
3469 		goto errout;
3470 
3471 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3472 	return 0;
3473 errout:
3474 	WARN_ON(err == -EMSGSIZE);
3475 	kfree_skb(skb);
3476 	if (err)
3477 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3478 	return err;
3479 }
3480 
3481 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3482 {
3483 	struct net *net = sock_net(skb->sk);
3484 	struct ifinfomsg *ifm;
3485 	struct net_device *dev;
3486 	struct nlattr *br_spec, *attr = NULL;
3487 	int rem, err = -EOPNOTSUPP;
3488 	u16 flags = 0;
3489 	bool have_flags = false;
3490 
3491 	if (nlmsg_len(nlh) < sizeof(*ifm))
3492 		return -EINVAL;
3493 
3494 	ifm = nlmsg_data(nlh);
3495 	if (ifm->ifi_family != AF_BRIDGE)
3496 		return -EPFNOSUPPORT;
3497 
3498 	dev = __dev_get_by_index(net, ifm->ifi_index);
3499 	if (!dev) {
3500 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3501 		return -ENODEV;
3502 	}
3503 
3504 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3505 	if (br_spec) {
3506 		nla_for_each_nested(attr, br_spec, rem) {
3507 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3508 				if (nla_len(attr) < sizeof(flags))
3509 					return -EINVAL;
3510 
3511 				have_flags = true;
3512 				flags = nla_get_u16(attr);
3513 				break;
3514 			}
3515 		}
3516 	}
3517 
3518 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3519 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3520 
3521 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3522 			err = -EOPNOTSUPP;
3523 			goto out;
3524 		}
3525 
3526 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3527 		if (err)
3528 			goto out;
3529 
3530 		flags &= ~BRIDGE_FLAGS_MASTER;
3531 	}
3532 
3533 	if ((flags & BRIDGE_FLAGS_SELF)) {
3534 		if (!dev->netdev_ops->ndo_bridge_setlink)
3535 			err = -EOPNOTSUPP;
3536 		else
3537 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3538 								  flags);
3539 		if (!err) {
3540 			flags &= ~BRIDGE_FLAGS_SELF;
3541 
3542 			/* Generate event to notify upper layer of bridge
3543 			 * change
3544 			 */
3545 			err = rtnl_bridge_notify(dev);
3546 		}
3547 	}
3548 
3549 	if (have_flags)
3550 		memcpy(nla_data(attr), &flags, sizeof(flags));
3551 out:
3552 	return err;
3553 }
3554 
3555 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3556 {
3557 	struct net *net = sock_net(skb->sk);
3558 	struct ifinfomsg *ifm;
3559 	struct net_device *dev;
3560 	struct nlattr *br_spec, *attr = NULL;
3561 	int rem, err = -EOPNOTSUPP;
3562 	u16 flags = 0;
3563 	bool have_flags = false;
3564 
3565 	if (nlmsg_len(nlh) < sizeof(*ifm))
3566 		return -EINVAL;
3567 
3568 	ifm = nlmsg_data(nlh);
3569 	if (ifm->ifi_family != AF_BRIDGE)
3570 		return -EPFNOSUPPORT;
3571 
3572 	dev = __dev_get_by_index(net, ifm->ifi_index);
3573 	if (!dev) {
3574 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3575 		return -ENODEV;
3576 	}
3577 
3578 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3579 	if (br_spec) {
3580 		nla_for_each_nested(attr, br_spec, rem) {
3581 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3582 				if (nla_len(attr) < sizeof(flags))
3583 					return -EINVAL;
3584 
3585 				have_flags = true;
3586 				flags = nla_get_u16(attr);
3587 				break;
3588 			}
3589 		}
3590 	}
3591 
3592 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3593 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3594 
3595 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3596 			err = -EOPNOTSUPP;
3597 			goto out;
3598 		}
3599 
3600 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3601 		if (err)
3602 			goto out;
3603 
3604 		flags &= ~BRIDGE_FLAGS_MASTER;
3605 	}
3606 
3607 	if ((flags & BRIDGE_FLAGS_SELF)) {
3608 		if (!dev->netdev_ops->ndo_bridge_dellink)
3609 			err = -EOPNOTSUPP;
3610 		else
3611 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3612 								  flags);
3613 
3614 		if (!err) {
3615 			flags &= ~BRIDGE_FLAGS_SELF;
3616 
3617 			/* Generate event to notify upper layer of bridge
3618 			 * change
3619 			 */
3620 			err = rtnl_bridge_notify(dev);
3621 		}
3622 	}
3623 
3624 	if (have_flags)
3625 		memcpy(nla_data(attr), &flags, sizeof(flags));
3626 out:
3627 	return err;
3628 }
3629 
3630 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3631 {
3632 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3633 	       (!idxattr || idxattr == attrid);
3634 }
3635 
3636 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3637 static int rtnl_get_offload_stats_attr_size(int attr_id)
3638 {
3639 	switch (attr_id) {
3640 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3641 		return sizeof(struct rtnl_link_stats64);
3642 	}
3643 
3644 	return 0;
3645 }
3646 
3647 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3648 				  int *prividx)
3649 {
3650 	struct nlattr *attr = NULL;
3651 	int attr_id, size;
3652 	void *attr_data;
3653 	int err;
3654 
3655 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3656 	      dev->netdev_ops->ndo_get_offload_stats))
3657 		return -ENODATA;
3658 
3659 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3660 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3661 		if (attr_id < *prividx)
3662 			continue;
3663 
3664 		size = rtnl_get_offload_stats_attr_size(attr_id);
3665 		if (!size)
3666 			continue;
3667 
3668 		if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3669 			continue;
3670 
3671 		attr = nla_reserve_64bit(skb, attr_id, size,
3672 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
3673 		if (!attr)
3674 			goto nla_put_failure;
3675 
3676 		attr_data = nla_data(attr);
3677 		memset(attr_data, 0, size);
3678 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3679 							     attr_data);
3680 		if (err)
3681 			goto get_offload_stats_failure;
3682 	}
3683 
3684 	if (!attr)
3685 		return -ENODATA;
3686 
3687 	*prividx = 0;
3688 	return 0;
3689 
3690 nla_put_failure:
3691 	err = -EMSGSIZE;
3692 get_offload_stats_failure:
3693 	*prividx = attr_id;
3694 	return err;
3695 }
3696 
3697 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3698 {
3699 	int nla_size = 0;
3700 	int attr_id;
3701 	int size;
3702 
3703 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3704 	      dev->netdev_ops->ndo_get_offload_stats))
3705 		return 0;
3706 
3707 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3708 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3709 		if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3710 			continue;
3711 		size = rtnl_get_offload_stats_attr_size(attr_id);
3712 		nla_size += nla_total_size_64bit(size);
3713 	}
3714 
3715 	if (nla_size != 0)
3716 		nla_size += nla_total_size(0);
3717 
3718 	return nla_size;
3719 }
3720 
3721 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3722 			       int type, u32 pid, u32 seq, u32 change,
3723 			       unsigned int flags, unsigned int filter_mask,
3724 			       int *idxattr, int *prividx)
3725 {
3726 	struct if_stats_msg *ifsm;
3727 	struct nlmsghdr *nlh;
3728 	struct nlattr *attr;
3729 	int s_prividx = *prividx;
3730 	int err;
3731 
3732 	ASSERT_RTNL();
3733 
3734 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3735 	if (!nlh)
3736 		return -EMSGSIZE;
3737 
3738 	ifsm = nlmsg_data(nlh);
3739 	ifsm->ifindex = dev->ifindex;
3740 	ifsm->filter_mask = filter_mask;
3741 
3742 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3743 		struct rtnl_link_stats64 *sp;
3744 
3745 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3746 					 sizeof(struct rtnl_link_stats64),
3747 					 IFLA_STATS_UNSPEC);
3748 		if (!attr)
3749 			goto nla_put_failure;
3750 
3751 		sp = nla_data(attr);
3752 		dev_get_stats(dev, sp);
3753 	}
3754 
3755 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3756 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3757 
3758 		if (ops && ops->fill_linkxstats) {
3759 			*idxattr = IFLA_STATS_LINK_XSTATS;
3760 			attr = nla_nest_start(skb,
3761 					      IFLA_STATS_LINK_XSTATS);
3762 			if (!attr)
3763 				goto nla_put_failure;
3764 
3765 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3766 			nla_nest_end(skb, attr);
3767 			if (err)
3768 				goto nla_put_failure;
3769 			*idxattr = 0;
3770 		}
3771 	}
3772 
3773 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3774 			     *idxattr)) {
3775 		const struct rtnl_link_ops *ops = NULL;
3776 		const struct net_device *master;
3777 
3778 		master = netdev_master_upper_dev_get(dev);
3779 		if (master)
3780 			ops = master->rtnl_link_ops;
3781 		if (ops && ops->fill_linkxstats) {
3782 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3783 			attr = nla_nest_start(skb,
3784 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3785 			if (!attr)
3786 				goto nla_put_failure;
3787 
3788 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3789 			nla_nest_end(skb, attr);
3790 			if (err)
3791 				goto nla_put_failure;
3792 			*idxattr = 0;
3793 		}
3794 	}
3795 
3796 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3797 			     *idxattr)) {
3798 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3799 		attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3800 		if (!attr)
3801 			goto nla_put_failure;
3802 
3803 		err = rtnl_get_offload_stats(skb, dev, prividx);
3804 		if (err == -ENODATA)
3805 			nla_nest_cancel(skb, attr);
3806 		else
3807 			nla_nest_end(skb, attr);
3808 
3809 		if (err && err != -ENODATA)
3810 			goto nla_put_failure;
3811 		*idxattr = 0;
3812 	}
3813 
3814 	nlmsg_end(skb, nlh);
3815 
3816 	return 0;
3817 
3818 nla_put_failure:
3819 	/* not a multi message or no progress mean a real error */
3820 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3821 		nlmsg_cancel(skb, nlh);
3822 	else
3823 		nlmsg_end(skb, nlh);
3824 
3825 	return -EMSGSIZE;
3826 }
3827 
3828 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3829 				  u32 filter_mask)
3830 {
3831 	size_t size = 0;
3832 
3833 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3834 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3835 
3836 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3837 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3838 		int attr = IFLA_STATS_LINK_XSTATS;
3839 
3840 		if (ops && ops->get_linkxstats_size) {
3841 			size += nla_total_size(ops->get_linkxstats_size(dev,
3842 									attr));
3843 			/* for IFLA_STATS_LINK_XSTATS */
3844 			size += nla_total_size(0);
3845 		}
3846 	}
3847 
3848 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3849 		struct net_device *_dev = (struct net_device *)dev;
3850 		const struct rtnl_link_ops *ops = NULL;
3851 		const struct net_device *master;
3852 
3853 		/* netdev_master_upper_dev_get can't take const */
3854 		master = netdev_master_upper_dev_get(_dev);
3855 		if (master)
3856 			ops = master->rtnl_link_ops;
3857 		if (ops && ops->get_linkxstats_size) {
3858 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3859 
3860 			size += nla_total_size(ops->get_linkxstats_size(dev,
3861 									attr));
3862 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
3863 			size += nla_total_size(0);
3864 		}
3865 	}
3866 
3867 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3868 		size += rtnl_get_offload_stats_size(dev);
3869 
3870 	return size;
3871 }
3872 
3873 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3874 {
3875 	struct net *net = sock_net(skb->sk);
3876 	struct net_device *dev = NULL;
3877 	int idxattr = 0, prividx = 0;
3878 	struct if_stats_msg *ifsm;
3879 	struct sk_buff *nskb;
3880 	u32 filter_mask;
3881 	int err;
3882 
3883 	ifsm = nlmsg_data(nlh);
3884 	if (ifsm->ifindex > 0)
3885 		dev = __dev_get_by_index(net, ifsm->ifindex);
3886 	else
3887 		return -EINVAL;
3888 
3889 	if (!dev)
3890 		return -ENODEV;
3891 
3892 	filter_mask = ifsm->filter_mask;
3893 	if (!filter_mask)
3894 		return -EINVAL;
3895 
3896 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3897 	if (!nskb)
3898 		return -ENOBUFS;
3899 
3900 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3901 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3902 				  0, filter_mask, &idxattr, &prividx);
3903 	if (err < 0) {
3904 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3905 		WARN_ON(err == -EMSGSIZE);
3906 		kfree_skb(nskb);
3907 	} else {
3908 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3909 	}
3910 
3911 	return err;
3912 }
3913 
3914 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3915 {
3916 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
3917 	struct net *net = sock_net(skb->sk);
3918 	unsigned int flags = NLM_F_MULTI;
3919 	struct if_stats_msg *ifsm;
3920 	struct hlist_head *head;
3921 	struct net_device *dev;
3922 	u32 filter_mask = 0;
3923 	int idx = 0;
3924 
3925 	s_h = cb->args[0];
3926 	s_idx = cb->args[1];
3927 	s_idxattr = cb->args[2];
3928 	s_prividx = cb->args[3];
3929 
3930 	cb->seq = net->dev_base_seq;
3931 
3932 	ifsm = nlmsg_data(cb->nlh);
3933 	filter_mask = ifsm->filter_mask;
3934 	if (!filter_mask)
3935 		return -EINVAL;
3936 
3937 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3938 		idx = 0;
3939 		head = &net->dev_index_head[h];
3940 		hlist_for_each_entry(dev, head, index_hlist) {
3941 			if (idx < s_idx)
3942 				goto cont;
3943 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3944 						  NETLINK_CB(cb->skb).portid,
3945 						  cb->nlh->nlmsg_seq, 0,
3946 						  flags, filter_mask,
3947 						  &s_idxattr, &s_prividx);
3948 			/* If we ran out of room on the first message,
3949 			 * we're in trouble
3950 			 */
3951 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3952 
3953 			if (err < 0)
3954 				goto out;
3955 			s_prividx = 0;
3956 			s_idxattr = 0;
3957 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3958 cont:
3959 			idx++;
3960 		}
3961 	}
3962 out:
3963 	cb->args[3] = s_prividx;
3964 	cb->args[2] = s_idxattr;
3965 	cb->args[1] = idx;
3966 	cb->args[0] = h;
3967 
3968 	return skb->len;
3969 }
3970 
3971 /* Process one rtnetlink message. */
3972 
3973 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3974 {
3975 	struct net *net = sock_net(skb->sk);
3976 	rtnl_doit_func doit;
3977 	int kind;
3978 	int family;
3979 	int type;
3980 	int err;
3981 
3982 	type = nlh->nlmsg_type;
3983 	if (type > RTM_MAX)
3984 		return -EOPNOTSUPP;
3985 
3986 	type -= RTM_BASE;
3987 
3988 	/* All the messages must have at least 1 byte length */
3989 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3990 		return 0;
3991 
3992 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3993 	kind = type&3;
3994 
3995 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3996 		return -EPERM;
3997 
3998 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3999 		struct sock *rtnl;
4000 		rtnl_dumpit_func dumpit;
4001 		rtnl_calcit_func calcit;
4002 		u16 min_dump_alloc = 0;
4003 
4004 		dumpit = rtnl_get_dumpit(family, type);
4005 		if (dumpit == NULL)
4006 			return -EOPNOTSUPP;
4007 		calcit = rtnl_get_calcit(family, type);
4008 		if (calcit)
4009 			min_dump_alloc = calcit(skb, nlh);
4010 
4011 		__rtnl_unlock();
4012 		rtnl = net->rtnl;
4013 		{
4014 			struct netlink_dump_control c = {
4015 				.dump		= dumpit,
4016 				.min_dump_alloc	= min_dump_alloc,
4017 			};
4018 			err = netlink_dump_start(rtnl, skb, nlh, &c);
4019 		}
4020 		rtnl_lock();
4021 		return err;
4022 	}
4023 
4024 	doit = rtnl_get_doit(family, type);
4025 	if (doit == NULL)
4026 		return -EOPNOTSUPP;
4027 
4028 	return doit(skb, nlh);
4029 }
4030 
4031 static void rtnetlink_rcv(struct sk_buff *skb)
4032 {
4033 	rtnl_lock();
4034 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4035 	rtnl_unlock();
4036 }
4037 
4038 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4039 {
4040 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4041 
4042 	switch (event) {
4043 	case NETDEV_UP:
4044 	case NETDEV_DOWN:
4045 	case NETDEV_PRE_UP:
4046 	case NETDEV_POST_INIT:
4047 	case NETDEV_REGISTER:
4048 	case NETDEV_CHANGE:
4049 	case NETDEV_PRE_TYPE_CHANGE:
4050 	case NETDEV_GOING_DOWN:
4051 	case NETDEV_UNREGISTER:
4052 	case NETDEV_UNREGISTER_FINAL:
4053 	case NETDEV_RELEASE:
4054 	case NETDEV_JOIN:
4055 	case NETDEV_BONDING_INFO:
4056 		break;
4057 	default:
4058 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4059 		break;
4060 	}
4061 	return NOTIFY_DONE;
4062 }
4063 
4064 static struct notifier_block rtnetlink_dev_notifier = {
4065 	.notifier_call	= rtnetlink_event,
4066 };
4067 
4068 
4069 static int __net_init rtnetlink_net_init(struct net *net)
4070 {
4071 	struct sock *sk;
4072 	struct netlink_kernel_cfg cfg = {
4073 		.groups		= RTNLGRP_MAX,
4074 		.input		= rtnetlink_rcv,
4075 		.cb_mutex	= &rtnl_mutex,
4076 		.flags		= NL_CFG_F_NONROOT_RECV,
4077 	};
4078 
4079 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4080 	if (!sk)
4081 		return -ENOMEM;
4082 	net->rtnl = sk;
4083 	return 0;
4084 }
4085 
4086 static void __net_exit rtnetlink_net_exit(struct net *net)
4087 {
4088 	netlink_kernel_release(net->rtnl);
4089 	net->rtnl = NULL;
4090 }
4091 
4092 static struct pernet_operations rtnetlink_net_ops = {
4093 	.init = rtnetlink_net_init,
4094 	.exit = rtnetlink_net_exit,
4095 };
4096 
4097 void __init rtnetlink_init(void)
4098 {
4099 	if (register_pernet_subsys(&rtnetlink_net_ops))
4100 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
4101 
4102 	register_netdevice_notifier(&rtnetlink_dev_notifier);
4103 
4104 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4105 		      rtnl_dump_ifinfo, rtnl_calcit);
4106 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4107 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4108 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4109 
4110 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4111 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4112 
4113 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4114 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4115 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4116 
4117 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4118 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4119 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4120 
4121 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4122 		      NULL);
4123 }
4124