xref: /openbmc/linux/net/core/rtnetlink.c (revision 0da85d1e)
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 void __rtnl_unlock(void)
75 {
76 	mutex_unlock(&rtnl_mutex);
77 }
78 
79 void rtnl_unlock(void)
80 {
81 	/* This fellow will unlock it for us. */
82 	netdev_run_todo();
83 }
84 EXPORT_SYMBOL(rtnl_unlock);
85 
86 int rtnl_trylock(void)
87 {
88 	return mutex_trylock(&rtnl_mutex);
89 }
90 EXPORT_SYMBOL(rtnl_trylock);
91 
92 int rtnl_is_locked(void)
93 {
94 	return mutex_is_locked(&rtnl_mutex);
95 }
96 EXPORT_SYMBOL(rtnl_is_locked);
97 
98 #ifdef CONFIG_PROVE_LOCKING
99 int lockdep_rtnl_is_held(void)
100 {
101 	return lockdep_is_held(&rtnl_mutex);
102 }
103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
105 
106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
107 
108 static inline int rtm_msgindex(int msgtype)
109 {
110 	int msgindex = msgtype - RTM_BASE;
111 
112 	/*
113 	 * msgindex < 0 implies someone tried to register a netlink
114 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
115 	 * the message type has not been added to linux/rtnetlink.h
116 	 */
117 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
118 
119 	return msgindex;
120 }
121 
122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
123 {
124 	struct rtnl_link *tab;
125 
126 	if (protocol <= RTNL_FAMILY_MAX)
127 		tab = rtnl_msg_handlers[protocol];
128 	else
129 		tab = NULL;
130 
131 	if (tab == NULL || tab[msgindex].doit == NULL)
132 		tab = rtnl_msg_handlers[PF_UNSPEC];
133 
134 	return tab[msgindex].doit;
135 }
136 
137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
138 {
139 	struct rtnl_link *tab;
140 
141 	if (protocol <= RTNL_FAMILY_MAX)
142 		tab = rtnl_msg_handlers[protocol];
143 	else
144 		tab = NULL;
145 
146 	if (tab == NULL || tab[msgindex].dumpit == NULL)
147 		tab = rtnl_msg_handlers[PF_UNSPEC];
148 
149 	return tab[msgindex].dumpit;
150 }
151 
152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
153 {
154 	struct rtnl_link *tab;
155 
156 	if (protocol <= RTNL_FAMILY_MAX)
157 		tab = rtnl_msg_handlers[protocol];
158 	else
159 		tab = NULL;
160 
161 	if (tab == NULL || tab[msgindex].calcit == NULL)
162 		tab = rtnl_msg_handlers[PF_UNSPEC];
163 
164 	return tab[msgindex].calcit;
165 }
166 
167 /**
168  * __rtnl_register - Register a rtnetlink message type
169  * @protocol: Protocol family or PF_UNSPEC
170  * @msgtype: rtnetlink message type
171  * @doit: Function pointer called for each request message
172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
173  * @calcit: Function pointer to calc size of dump message
174  *
175  * Registers the specified function pointers (at least one of them has
176  * to be non-NULL) to be called whenever a request message for the
177  * specified protocol family and message type is received.
178  *
179  * The special protocol family PF_UNSPEC may be used to define fallback
180  * function pointers for the case when no entry for the specific protocol
181  * family exists.
182  *
183  * Returns 0 on success or a negative error code.
184  */
185 int __rtnl_register(int protocol, int msgtype,
186 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
187 		    rtnl_calcit_func calcit)
188 {
189 	struct rtnl_link *tab;
190 	int msgindex;
191 
192 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
193 	msgindex = rtm_msgindex(msgtype);
194 
195 	tab = rtnl_msg_handlers[protocol];
196 	if (tab == NULL) {
197 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
198 		if (tab == NULL)
199 			return -ENOBUFS;
200 
201 		rtnl_msg_handlers[protocol] = tab;
202 	}
203 
204 	if (doit)
205 		tab[msgindex].doit = doit;
206 
207 	if (dumpit)
208 		tab[msgindex].dumpit = dumpit;
209 
210 	if (calcit)
211 		tab[msgindex].calcit = calcit;
212 
213 	return 0;
214 }
215 EXPORT_SYMBOL_GPL(__rtnl_register);
216 
217 /**
218  * rtnl_register - Register a rtnetlink message type
219  *
220  * Identical to __rtnl_register() but panics on failure. This is useful
221  * as failure of this function is very unlikely, it can only happen due
222  * to lack of memory when allocating the chain to store all message
223  * handlers for a protocol. Meant for use in init functions where lack
224  * of memory implies no sense in continuing.
225  */
226 void rtnl_register(int protocol, int msgtype,
227 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
228 		   rtnl_calcit_func calcit)
229 {
230 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
231 		panic("Unable to register rtnetlink message handler, "
232 		      "protocol = %d, message type = %d\n",
233 		      protocol, msgtype);
234 }
235 EXPORT_SYMBOL_GPL(rtnl_register);
236 
237 /**
238  * rtnl_unregister - Unregister a rtnetlink message type
239  * @protocol: Protocol family or PF_UNSPEC
240  * @msgtype: rtnetlink message type
241  *
242  * Returns 0 on success or a negative error code.
243  */
244 int rtnl_unregister(int protocol, int msgtype)
245 {
246 	int msgindex;
247 
248 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
249 	msgindex = rtm_msgindex(msgtype);
250 
251 	if (rtnl_msg_handlers[protocol] == NULL)
252 		return -ENOENT;
253 
254 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
255 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
256 
257 	return 0;
258 }
259 EXPORT_SYMBOL_GPL(rtnl_unregister);
260 
261 /**
262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
263  * @protocol : Protocol family or PF_UNSPEC
264  *
265  * Identical to calling rtnl_unregster() for all registered message types
266  * of a certain protocol family.
267  */
268 void rtnl_unregister_all(int protocol)
269 {
270 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 
272 	kfree(rtnl_msg_handlers[protocol]);
273 	rtnl_msg_handlers[protocol] = NULL;
274 }
275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
276 
277 static LIST_HEAD(link_ops);
278 
279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
280 {
281 	const struct rtnl_link_ops *ops;
282 
283 	list_for_each_entry(ops, &link_ops, list) {
284 		if (!strcmp(ops->kind, kind))
285 			return ops;
286 	}
287 	return NULL;
288 }
289 
290 /**
291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
292  * @ops: struct rtnl_link_ops * to register
293  *
294  * The caller must hold the rtnl_mutex. This function should be used
295  * by drivers that create devices during module initialization. It
296  * must be called before registering the devices.
297  *
298  * Returns 0 on success or a negative error code.
299  */
300 int __rtnl_link_register(struct rtnl_link_ops *ops)
301 {
302 	if (rtnl_link_ops_get(ops->kind))
303 		return -EEXIST;
304 
305 	/* The check for setup is here because if ops
306 	 * does not have that filled up, it is not possible
307 	 * to use the ops for creating device. So do not
308 	 * fill up dellink as well. That disables rtnl_dellink.
309 	 */
310 	if (ops->setup && !ops->dellink)
311 		ops->dellink = unregister_netdevice_queue;
312 
313 	list_add_tail(&ops->list, &link_ops);
314 	return 0;
315 }
316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
317 
318 /**
319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
320  * @ops: struct rtnl_link_ops * to register
321  *
322  * Returns 0 on success or a negative error code.
323  */
324 int rtnl_link_register(struct rtnl_link_ops *ops)
325 {
326 	int err;
327 
328 	rtnl_lock();
329 	err = __rtnl_link_register(ops);
330 	rtnl_unlock();
331 	return err;
332 }
333 EXPORT_SYMBOL_GPL(rtnl_link_register);
334 
335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
336 {
337 	struct net_device *dev;
338 	LIST_HEAD(list_kill);
339 
340 	for_each_netdev(net, dev) {
341 		if (dev->rtnl_link_ops == ops)
342 			ops->dellink(dev, &list_kill);
343 	}
344 	unregister_netdevice_many(&list_kill);
345 }
346 
347 /**
348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
349  * @ops: struct rtnl_link_ops * to unregister
350  *
351  * The caller must hold the rtnl_mutex.
352  */
353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
354 {
355 	struct net *net;
356 
357 	for_each_net(net) {
358 		__rtnl_kill_links(net, ops);
359 	}
360 	list_del(&ops->list);
361 }
362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
363 
364 /* Return with the rtnl_lock held when there are no network
365  * devices unregistering in any network namespace.
366  */
367 static void rtnl_lock_unregistering_all(void)
368 {
369 	struct net *net;
370 	bool unregistering;
371 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
372 
373 	add_wait_queue(&netdev_unregistering_wq, &wait);
374 	for (;;) {
375 		unregistering = false;
376 		rtnl_lock();
377 		for_each_net(net) {
378 			if (net->dev_unreg_count > 0) {
379 				unregistering = true;
380 				break;
381 			}
382 		}
383 		if (!unregistering)
384 			break;
385 		__rtnl_unlock();
386 
387 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
388 	}
389 	remove_wait_queue(&netdev_unregistering_wq, &wait);
390 }
391 
392 /**
393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
394  * @ops: struct rtnl_link_ops * to unregister
395  */
396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
397 {
398 	/* Close the race with cleanup_net() */
399 	mutex_lock(&net_mutex);
400 	rtnl_lock_unregistering_all();
401 	__rtnl_link_unregister(ops);
402 	rtnl_unlock();
403 	mutex_unlock(&net_mutex);
404 }
405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
406 
407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
408 {
409 	struct net_device *master_dev;
410 	const struct rtnl_link_ops *ops;
411 
412 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
413 	if (!master_dev)
414 		return 0;
415 	ops = master_dev->rtnl_link_ops;
416 	if (!ops || !ops->get_slave_size)
417 		return 0;
418 	/* IFLA_INFO_SLAVE_DATA + nested data */
419 	return nla_total_size(sizeof(struct nlattr)) +
420 	       ops->get_slave_size(master_dev, dev);
421 }
422 
423 static size_t rtnl_link_get_size(const struct net_device *dev)
424 {
425 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
426 	size_t size;
427 
428 	if (!ops)
429 		return 0;
430 
431 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
432 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
433 
434 	if (ops->get_size)
435 		/* IFLA_INFO_DATA + nested data */
436 		size += nla_total_size(sizeof(struct nlattr)) +
437 			ops->get_size(dev);
438 
439 	if (ops->get_xstats_size)
440 		/* IFLA_INFO_XSTATS */
441 		size += nla_total_size(ops->get_xstats_size(dev));
442 
443 	size += rtnl_link_get_slave_info_data_size(dev);
444 
445 	return size;
446 }
447 
448 static LIST_HEAD(rtnl_af_ops);
449 
450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
451 {
452 	const struct rtnl_af_ops *ops;
453 
454 	list_for_each_entry(ops, &rtnl_af_ops, list) {
455 		if (ops->family == family)
456 			return ops;
457 	}
458 
459 	return NULL;
460 }
461 
462 /**
463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
464  * @ops: struct rtnl_af_ops * to register
465  *
466  * Returns 0 on success or a negative error code.
467  */
468 void rtnl_af_register(struct rtnl_af_ops *ops)
469 {
470 	rtnl_lock();
471 	list_add_tail(&ops->list, &rtnl_af_ops);
472 	rtnl_unlock();
473 }
474 EXPORT_SYMBOL_GPL(rtnl_af_register);
475 
476 /**
477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
478  * @ops: struct rtnl_af_ops * to unregister
479  *
480  * The caller must hold the rtnl_mutex.
481  */
482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
483 {
484 	list_del(&ops->list);
485 }
486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
487 
488 /**
489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
490  * @ops: struct rtnl_af_ops * to unregister
491  */
492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
493 {
494 	rtnl_lock();
495 	__rtnl_af_unregister(ops);
496 	rtnl_unlock();
497 }
498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
499 
500 static size_t rtnl_link_get_af_size(const struct net_device *dev)
501 {
502 	struct rtnl_af_ops *af_ops;
503 	size_t size;
504 
505 	/* IFLA_AF_SPEC */
506 	size = nla_total_size(sizeof(struct nlattr));
507 
508 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
509 		if (af_ops->get_link_af_size) {
510 			/* AF_* + nested data */
511 			size += nla_total_size(sizeof(struct nlattr)) +
512 				af_ops->get_link_af_size(dev);
513 		}
514 	}
515 
516 	return size;
517 }
518 
519 static bool rtnl_have_link_slave_info(const struct net_device *dev)
520 {
521 	struct net_device *master_dev;
522 
523 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
524 	if (master_dev && master_dev->rtnl_link_ops)
525 		return true;
526 	return false;
527 }
528 
529 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
530 				     const struct net_device *dev)
531 {
532 	struct net_device *master_dev;
533 	const struct rtnl_link_ops *ops;
534 	struct nlattr *slave_data;
535 	int err;
536 
537 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
538 	if (!master_dev)
539 		return 0;
540 	ops = master_dev->rtnl_link_ops;
541 	if (!ops)
542 		return 0;
543 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
544 		return -EMSGSIZE;
545 	if (ops->fill_slave_info) {
546 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
547 		if (!slave_data)
548 			return -EMSGSIZE;
549 		err = ops->fill_slave_info(skb, master_dev, dev);
550 		if (err < 0)
551 			goto err_cancel_slave_data;
552 		nla_nest_end(skb, slave_data);
553 	}
554 	return 0;
555 
556 err_cancel_slave_data:
557 	nla_nest_cancel(skb, slave_data);
558 	return err;
559 }
560 
561 static int rtnl_link_info_fill(struct sk_buff *skb,
562 			       const struct net_device *dev)
563 {
564 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
565 	struct nlattr *data;
566 	int err;
567 
568 	if (!ops)
569 		return 0;
570 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
571 		return -EMSGSIZE;
572 	if (ops->fill_xstats) {
573 		err = ops->fill_xstats(skb, dev);
574 		if (err < 0)
575 			return err;
576 	}
577 	if (ops->fill_info) {
578 		data = nla_nest_start(skb, IFLA_INFO_DATA);
579 		if (data == NULL)
580 			return -EMSGSIZE;
581 		err = ops->fill_info(skb, dev);
582 		if (err < 0)
583 			goto err_cancel_data;
584 		nla_nest_end(skb, data);
585 	}
586 	return 0;
587 
588 err_cancel_data:
589 	nla_nest_cancel(skb, data);
590 	return err;
591 }
592 
593 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
594 {
595 	struct nlattr *linkinfo;
596 	int err = -EMSGSIZE;
597 
598 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
599 	if (linkinfo == NULL)
600 		goto out;
601 
602 	err = rtnl_link_info_fill(skb, dev);
603 	if (err < 0)
604 		goto err_cancel_link;
605 
606 	err = rtnl_link_slave_info_fill(skb, dev);
607 	if (err < 0)
608 		goto err_cancel_link;
609 
610 	nla_nest_end(skb, linkinfo);
611 	return 0;
612 
613 err_cancel_link:
614 	nla_nest_cancel(skb, linkinfo);
615 out:
616 	return err;
617 }
618 
619 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
620 {
621 	struct sock *rtnl = net->rtnl;
622 	int err = 0;
623 
624 	NETLINK_CB(skb).dst_group = group;
625 	if (echo)
626 		atomic_inc(&skb->users);
627 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
628 	if (echo)
629 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
630 	return err;
631 }
632 
633 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
634 {
635 	struct sock *rtnl = net->rtnl;
636 
637 	return nlmsg_unicast(rtnl, skb, pid);
638 }
639 EXPORT_SYMBOL(rtnl_unicast);
640 
641 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
642 		 struct nlmsghdr *nlh, gfp_t flags)
643 {
644 	struct sock *rtnl = net->rtnl;
645 	int report = 0;
646 
647 	if (nlh)
648 		report = nlmsg_report(nlh);
649 
650 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
651 }
652 EXPORT_SYMBOL(rtnl_notify);
653 
654 void rtnl_set_sk_err(struct net *net, u32 group, int error)
655 {
656 	struct sock *rtnl = net->rtnl;
657 
658 	netlink_set_err(rtnl, 0, group, error);
659 }
660 EXPORT_SYMBOL(rtnl_set_sk_err);
661 
662 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
663 {
664 	struct nlattr *mx;
665 	int i, valid = 0;
666 
667 	mx = nla_nest_start(skb, RTA_METRICS);
668 	if (mx == NULL)
669 		return -ENOBUFS;
670 
671 	for (i = 0; i < RTAX_MAX; i++) {
672 		if (metrics[i]) {
673 			if (i == RTAX_CC_ALGO - 1) {
674 				char tmp[TCP_CA_NAME_MAX], *name;
675 
676 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
677 				if (!name)
678 					continue;
679 				if (nla_put_string(skb, i + 1, name))
680 					goto nla_put_failure;
681 			} else {
682 				if (nla_put_u32(skb, i + 1, metrics[i]))
683 					goto nla_put_failure;
684 			}
685 			valid++;
686 		}
687 	}
688 
689 	if (!valid) {
690 		nla_nest_cancel(skb, mx);
691 		return 0;
692 	}
693 
694 	return nla_nest_end(skb, mx);
695 
696 nla_put_failure:
697 	nla_nest_cancel(skb, mx);
698 	return -EMSGSIZE;
699 }
700 EXPORT_SYMBOL(rtnetlink_put_metrics);
701 
702 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
703 		       long expires, u32 error)
704 {
705 	struct rta_cacheinfo ci = {
706 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
707 		.rta_used = dst->__use,
708 		.rta_clntref = atomic_read(&(dst->__refcnt)),
709 		.rta_error = error,
710 		.rta_id =  id,
711 	};
712 
713 	if (expires) {
714 		unsigned long clock;
715 
716 		clock = jiffies_to_clock_t(abs(expires));
717 		clock = min_t(unsigned long, clock, INT_MAX);
718 		ci.rta_expires = (expires > 0) ? clock : -clock;
719 	}
720 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
721 }
722 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
723 
724 static void set_operstate(struct net_device *dev, unsigned char transition)
725 {
726 	unsigned char operstate = dev->operstate;
727 
728 	switch (transition) {
729 	case IF_OPER_UP:
730 		if ((operstate == IF_OPER_DORMANT ||
731 		     operstate == IF_OPER_UNKNOWN) &&
732 		    !netif_dormant(dev))
733 			operstate = IF_OPER_UP;
734 		break;
735 
736 	case IF_OPER_DORMANT:
737 		if (operstate == IF_OPER_UP ||
738 		    operstate == IF_OPER_UNKNOWN)
739 			operstate = IF_OPER_DORMANT;
740 		break;
741 	}
742 
743 	if (dev->operstate != operstate) {
744 		write_lock_bh(&dev_base_lock);
745 		dev->operstate = operstate;
746 		write_unlock_bh(&dev_base_lock);
747 		netdev_state_change(dev);
748 	}
749 }
750 
751 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
752 {
753 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
754 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
755 }
756 
757 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
758 					   const struct ifinfomsg *ifm)
759 {
760 	unsigned int flags = ifm->ifi_flags;
761 
762 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
763 	if (ifm->ifi_change)
764 		flags = (flags & ifm->ifi_change) |
765 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
766 
767 	return flags;
768 }
769 
770 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
771 				 const struct rtnl_link_stats64 *b)
772 {
773 	a->rx_packets = b->rx_packets;
774 	a->tx_packets = b->tx_packets;
775 	a->rx_bytes = b->rx_bytes;
776 	a->tx_bytes = b->tx_bytes;
777 	a->rx_errors = b->rx_errors;
778 	a->tx_errors = b->tx_errors;
779 	a->rx_dropped = b->rx_dropped;
780 	a->tx_dropped = b->tx_dropped;
781 
782 	a->multicast = b->multicast;
783 	a->collisions = b->collisions;
784 
785 	a->rx_length_errors = b->rx_length_errors;
786 	a->rx_over_errors = b->rx_over_errors;
787 	a->rx_crc_errors = b->rx_crc_errors;
788 	a->rx_frame_errors = b->rx_frame_errors;
789 	a->rx_fifo_errors = b->rx_fifo_errors;
790 	a->rx_missed_errors = b->rx_missed_errors;
791 
792 	a->tx_aborted_errors = b->tx_aborted_errors;
793 	a->tx_carrier_errors = b->tx_carrier_errors;
794 	a->tx_fifo_errors = b->tx_fifo_errors;
795 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
796 	a->tx_window_errors = b->tx_window_errors;
797 
798 	a->rx_compressed = b->rx_compressed;
799 	a->tx_compressed = b->tx_compressed;
800 }
801 
802 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
803 {
804 	memcpy(v, b, sizeof(*b));
805 }
806 
807 /* All VF info */
808 static inline int rtnl_vfinfo_size(const struct net_device *dev,
809 				   u32 ext_filter_mask)
810 {
811 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
812 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
813 		int num_vfs = dev_num_vf(dev->dev.parent);
814 		size_t size = nla_total_size(sizeof(struct nlattr));
815 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
816 		size += num_vfs *
817 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
818 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
819 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
820 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
821 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
822 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)));
823 		return size;
824 	} else
825 		return 0;
826 }
827 
828 static size_t rtnl_port_size(const struct net_device *dev,
829 			     u32 ext_filter_mask)
830 {
831 	size_t port_size = nla_total_size(4)		/* PORT_VF */
832 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
833 		+ nla_total_size(sizeof(struct ifla_port_vsi))
834 							/* PORT_VSI_TYPE */
835 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
836 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
837 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
838 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
839 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
840 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
841 		+ port_size;
842 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
843 		+ port_size;
844 
845 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
846 	    !(ext_filter_mask & RTEXT_FILTER_VF))
847 		return 0;
848 	if (dev_num_vf(dev->dev.parent))
849 		return port_self_size + vf_ports_size +
850 			vf_port_size * dev_num_vf(dev->dev.parent);
851 	else
852 		return port_self_size;
853 }
854 
855 static noinline size_t if_nlmsg_size(const struct net_device *dev,
856 				     u32 ext_filter_mask)
857 {
858 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
859 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
860 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
861 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
862 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
863 	       + nla_total_size(sizeof(struct rtnl_link_stats))
864 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
865 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
866 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
867 	       + nla_total_size(4) /* IFLA_TXQLEN */
868 	       + nla_total_size(4) /* IFLA_WEIGHT */
869 	       + nla_total_size(4) /* IFLA_MTU */
870 	       + nla_total_size(4) /* IFLA_LINK */
871 	       + nla_total_size(4) /* IFLA_MASTER */
872 	       + nla_total_size(1) /* IFLA_CARRIER */
873 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
874 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
875 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
876 	       + nla_total_size(1) /* IFLA_OPERSTATE */
877 	       + nla_total_size(1) /* IFLA_LINKMODE */
878 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
879 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
880 	       + nla_total_size(ext_filter_mask
881 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
882 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
883 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
884 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
885 	       + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
886 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
887 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
888 }
889 
890 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
891 {
892 	struct nlattr *vf_ports;
893 	struct nlattr *vf_port;
894 	int vf;
895 	int err;
896 
897 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
898 	if (!vf_ports)
899 		return -EMSGSIZE;
900 
901 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
902 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
903 		if (!vf_port)
904 			goto nla_put_failure;
905 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
906 			goto nla_put_failure;
907 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
908 		if (err == -EMSGSIZE)
909 			goto nla_put_failure;
910 		if (err) {
911 			nla_nest_cancel(skb, vf_port);
912 			continue;
913 		}
914 		nla_nest_end(skb, vf_port);
915 	}
916 
917 	nla_nest_end(skb, vf_ports);
918 
919 	return 0;
920 
921 nla_put_failure:
922 	nla_nest_cancel(skb, vf_ports);
923 	return -EMSGSIZE;
924 }
925 
926 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
927 {
928 	struct nlattr *port_self;
929 	int err;
930 
931 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
932 	if (!port_self)
933 		return -EMSGSIZE;
934 
935 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
936 	if (err) {
937 		nla_nest_cancel(skb, port_self);
938 		return (err == -EMSGSIZE) ? err : 0;
939 	}
940 
941 	nla_nest_end(skb, port_self);
942 
943 	return 0;
944 }
945 
946 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
947 			  u32 ext_filter_mask)
948 {
949 	int err;
950 
951 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
952 	    !(ext_filter_mask & RTEXT_FILTER_VF))
953 		return 0;
954 
955 	err = rtnl_port_self_fill(skb, dev);
956 	if (err)
957 		return err;
958 
959 	if (dev_num_vf(dev->dev.parent)) {
960 		err = rtnl_vf_ports_fill(skb, dev);
961 		if (err)
962 			return err;
963 	}
964 
965 	return 0;
966 }
967 
968 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
969 {
970 	int err;
971 	struct netdev_phys_item_id ppid;
972 
973 	err = dev_get_phys_port_id(dev, &ppid);
974 	if (err) {
975 		if (err == -EOPNOTSUPP)
976 			return 0;
977 		return err;
978 	}
979 
980 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
981 		return -EMSGSIZE;
982 
983 	return 0;
984 }
985 
986 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
987 {
988 	char name[IFNAMSIZ];
989 	int err;
990 
991 	err = dev_get_phys_port_name(dev, name, sizeof(name));
992 	if (err) {
993 		if (err == -EOPNOTSUPP)
994 			return 0;
995 		return err;
996 	}
997 
998 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
999 		return -EMSGSIZE;
1000 
1001 	return 0;
1002 }
1003 
1004 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1005 {
1006 	int err;
1007 	struct netdev_phys_item_id psid;
1008 
1009 	err = netdev_switch_parent_id_get(dev, &psid);
1010 	if (err) {
1011 		if (err == -EOPNOTSUPP)
1012 			return 0;
1013 		return err;
1014 	}
1015 
1016 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
1017 		return -EMSGSIZE;
1018 
1019 	return 0;
1020 }
1021 
1022 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1023 			    int type, u32 pid, u32 seq, u32 change,
1024 			    unsigned int flags, u32 ext_filter_mask)
1025 {
1026 	struct ifinfomsg *ifm;
1027 	struct nlmsghdr *nlh;
1028 	struct rtnl_link_stats64 temp;
1029 	const struct rtnl_link_stats64 *stats;
1030 	struct nlattr *attr, *af_spec;
1031 	struct rtnl_af_ops *af_ops;
1032 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1033 
1034 	ASSERT_RTNL();
1035 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1036 	if (nlh == NULL)
1037 		return -EMSGSIZE;
1038 
1039 	ifm = nlmsg_data(nlh);
1040 	ifm->ifi_family = AF_UNSPEC;
1041 	ifm->__ifi_pad = 0;
1042 	ifm->ifi_type = dev->type;
1043 	ifm->ifi_index = dev->ifindex;
1044 	ifm->ifi_flags = dev_get_flags(dev);
1045 	ifm->ifi_change = change;
1046 
1047 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1048 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1049 	    nla_put_u8(skb, IFLA_OPERSTATE,
1050 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1051 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1052 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1053 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1054 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1055 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1056 #ifdef CONFIG_RPS
1057 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1058 #endif
1059 	    (dev->ifindex != dev_get_iflink(dev) &&
1060 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1061 	    (upper_dev &&
1062 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1063 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1064 	    (dev->qdisc &&
1065 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1066 	    (dev->ifalias &&
1067 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1068 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1069 			atomic_read(&dev->carrier_changes)))
1070 		goto nla_put_failure;
1071 
1072 	if (1) {
1073 		struct rtnl_link_ifmap map = {
1074 			.mem_start   = dev->mem_start,
1075 			.mem_end     = dev->mem_end,
1076 			.base_addr   = dev->base_addr,
1077 			.irq         = dev->irq,
1078 			.dma         = dev->dma,
1079 			.port        = dev->if_port,
1080 		};
1081 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1082 			goto nla_put_failure;
1083 	}
1084 
1085 	if (dev->addr_len) {
1086 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1087 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1088 			goto nla_put_failure;
1089 	}
1090 
1091 	if (rtnl_phys_port_id_fill(skb, dev))
1092 		goto nla_put_failure;
1093 
1094 	if (rtnl_phys_port_name_fill(skb, dev))
1095 		goto nla_put_failure;
1096 
1097 	if (rtnl_phys_switch_id_fill(skb, dev))
1098 		goto nla_put_failure;
1099 
1100 	attr = nla_reserve(skb, IFLA_STATS,
1101 			sizeof(struct rtnl_link_stats));
1102 	if (attr == NULL)
1103 		goto nla_put_failure;
1104 
1105 	stats = dev_get_stats(dev, &temp);
1106 	copy_rtnl_link_stats(nla_data(attr), stats);
1107 
1108 	attr = nla_reserve(skb, IFLA_STATS64,
1109 			sizeof(struct rtnl_link_stats64));
1110 	if (attr == NULL)
1111 		goto nla_put_failure;
1112 	copy_rtnl_link_stats64(nla_data(attr), stats);
1113 
1114 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1115 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1116 		goto nla_put_failure;
1117 
1118 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1119 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
1120 		int i;
1121 
1122 		struct nlattr *vfinfo, *vf;
1123 		int num_vfs = dev_num_vf(dev->dev.parent);
1124 
1125 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1126 		if (!vfinfo)
1127 			goto nla_put_failure;
1128 		for (i = 0; i < num_vfs; i++) {
1129 			struct ifla_vf_info ivi;
1130 			struct ifla_vf_mac vf_mac;
1131 			struct ifla_vf_vlan vf_vlan;
1132 			struct ifla_vf_rate vf_rate;
1133 			struct ifla_vf_tx_rate vf_tx_rate;
1134 			struct ifla_vf_spoofchk vf_spoofchk;
1135 			struct ifla_vf_link_state vf_linkstate;
1136 			struct ifla_vf_rss_query_en vf_rss_query_en;
1137 
1138 			/*
1139 			 * Not all SR-IOV capable drivers support the
1140 			 * spoofcheck and "RSS query enable" query.  Preset to
1141 			 * -1 so the user space tool can detect that the driver
1142 			 * didn't report anything.
1143 			 */
1144 			ivi.spoofchk = -1;
1145 			ivi.rss_query_en = -1;
1146 			memset(ivi.mac, 0, sizeof(ivi.mac));
1147 			/* The default value for VF link state is "auto"
1148 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
1149 			 */
1150 			ivi.linkstate = 0;
1151 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1152 				break;
1153 			vf_mac.vf =
1154 				vf_vlan.vf =
1155 				vf_rate.vf =
1156 				vf_tx_rate.vf =
1157 				vf_spoofchk.vf =
1158 				vf_linkstate.vf =
1159 				vf_rss_query_en.vf = ivi.vf;
1160 
1161 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1162 			vf_vlan.vlan = ivi.vlan;
1163 			vf_vlan.qos = ivi.qos;
1164 			vf_tx_rate.rate = ivi.max_tx_rate;
1165 			vf_rate.min_tx_rate = ivi.min_tx_rate;
1166 			vf_rate.max_tx_rate = ivi.max_tx_rate;
1167 			vf_spoofchk.setting = ivi.spoofchk;
1168 			vf_linkstate.link_state = ivi.linkstate;
1169 			vf_rss_query_en.setting = ivi.rss_query_en;
1170 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1171 			if (!vf) {
1172 				nla_nest_cancel(skb, vfinfo);
1173 				goto nla_put_failure;
1174 			}
1175 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1176 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1177 			    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1178 				    &vf_rate) ||
1179 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1180 				    &vf_tx_rate) ||
1181 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1182 				    &vf_spoofchk) ||
1183 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1184 				    &vf_linkstate) ||
1185 			    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1186 				    sizeof(vf_rss_query_en),
1187 				    &vf_rss_query_en))
1188 				goto nla_put_failure;
1189 			nla_nest_end(skb, vf);
1190 		}
1191 		nla_nest_end(skb, vfinfo);
1192 	}
1193 
1194 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1195 		goto nla_put_failure;
1196 
1197 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1198 		if (rtnl_link_fill(skb, dev) < 0)
1199 			goto nla_put_failure;
1200 	}
1201 
1202 	if (dev->rtnl_link_ops &&
1203 	    dev->rtnl_link_ops->get_link_net) {
1204 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1205 
1206 		if (!net_eq(dev_net(dev), link_net)) {
1207 			int id = peernet2id(dev_net(dev), link_net);
1208 
1209 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1210 				goto nla_put_failure;
1211 		}
1212 	}
1213 
1214 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1215 		goto nla_put_failure;
1216 
1217 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1218 		if (af_ops->fill_link_af) {
1219 			struct nlattr *af;
1220 			int err;
1221 
1222 			if (!(af = nla_nest_start(skb, af_ops->family)))
1223 				goto nla_put_failure;
1224 
1225 			err = af_ops->fill_link_af(skb, dev);
1226 
1227 			/*
1228 			 * Caller may return ENODATA to indicate that there
1229 			 * was no data to be dumped. This is not an error, it
1230 			 * means we should trim the attribute header and
1231 			 * continue.
1232 			 */
1233 			if (err == -ENODATA)
1234 				nla_nest_cancel(skb, af);
1235 			else if (err < 0)
1236 				goto nla_put_failure;
1237 
1238 			nla_nest_end(skb, af);
1239 		}
1240 	}
1241 
1242 	nla_nest_end(skb, af_spec);
1243 
1244 	nlmsg_end(skb, nlh);
1245 	return 0;
1246 
1247 nla_put_failure:
1248 	nlmsg_cancel(skb, nlh);
1249 	return -EMSGSIZE;
1250 }
1251 
1252 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1253 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1254 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1255 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1256 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1257 	[IFLA_MTU]		= { .type = NLA_U32 },
1258 	[IFLA_LINK]		= { .type = NLA_U32 },
1259 	[IFLA_MASTER]		= { .type = NLA_U32 },
1260 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1261 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1262 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1263 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1264 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1265 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1266 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1267 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1268 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1269 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1270 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1271 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1272 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1273 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1274 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1275 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1276 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1277 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1278 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1279 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1280 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1281 };
1282 
1283 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1284 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1285 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1286 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1287 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1288 };
1289 
1290 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1291 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1292 };
1293 
1294 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1295 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1296 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1297 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1298 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1299 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1300 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1301 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1302 };
1303 
1304 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1305 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1306 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1307 				    .len = PORT_PROFILE_MAX },
1308 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1309 				    .len = sizeof(struct ifla_port_vsi)},
1310 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1311 				      .len = PORT_UUID_MAX },
1312 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1313 				    .len = PORT_UUID_MAX },
1314 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1315 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1316 };
1317 
1318 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1319 {
1320 	struct net *net = sock_net(skb->sk);
1321 	int h, s_h;
1322 	int idx = 0, s_idx;
1323 	struct net_device *dev;
1324 	struct hlist_head *head;
1325 	struct nlattr *tb[IFLA_MAX+1];
1326 	u32 ext_filter_mask = 0;
1327 	int err;
1328 	int hdrlen;
1329 
1330 	s_h = cb->args[0];
1331 	s_idx = cb->args[1];
1332 
1333 	cb->seq = net->dev_base_seq;
1334 
1335 	/* A hack to preserve kernel<->userspace interface.
1336 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1337 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1338 	 * what iproute2 < v3.9.0 used.
1339 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1340 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1341 	 */
1342 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1343 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1344 
1345 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1346 
1347 		if (tb[IFLA_EXT_MASK])
1348 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1349 	}
1350 
1351 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1352 		idx = 0;
1353 		head = &net->dev_index_head[h];
1354 		hlist_for_each_entry(dev, head, index_hlist) {
1355 			if (idx < s_idx)
1356 				goto cont;
1357 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1358 					       NETLINK_CB(cb->skb).portid,
1359 					       cb->nlh->nlmsg_seq, 0,
1360 					       NLM_F_MULTI,
1361 					       ext_filter_mask);
1362 			/* If we ran out of room on the first message,
1363 			 * we're in trouble
1364 			 */
1365 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1366 
1367 			if (err < 0)
1368 				goto out;
1369 
1370 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1371 cont:
1372 			idx++;
1373 		}
1374 	}
1375 out:
1376 	cb->args[1] = idx;
1377 	cb->args[0] = h;
1378 
1379 	return skb->len;
1380 }
1381 
1382 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1383 {
1384 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1385 }
1386 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1387 
1388 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1389 {
1390 	struct net *net;
1391 	/* Examine the link attributes and figure out which
1392 	 * network namespace we are talking about.
1393 	 */
1394 	if (tb[IFLA_NET_NS_PID])
1395 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1396 	else if (tb[IFLA_NET_NS_FD])
1397 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1398 	else
1399 		net = get_net(src_net);
1400 	return net;
1401 }
1402 EXPORT_SYMBOL(rtnl_link_get_net);
1403 
1404 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1405 {
1406 	if (dev) {
1407 		if (tb[IFLA_ADDRESS] &&
1408 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1409 			return -EINVAL;
1410 
1411 		if (tb[IFLA_BROADCAST] &&
1412 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1413 			return -EINVAL;
1414 	}
1415 
1416 	if (tb[IFLA_AF_SPEC]) {
1417 		struct nlattr *af;
1418 		int rem, err;
1419 
1420 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1421 			const struct rtnl_af_ops *af_ops;
1422 
1423 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1424 				return -EAFNOSUPPORT;
1425 
1426 			if (!af_ops->set_link_af)
1427 				return -EOPNOTSUPP;
1428 
1429 			if (af_ops->validate_link_af) {
1430 				err = af_ops->validate_link_af(dev, af);
1431 				if (err < 0)
1432 					return err;
1433 			}
1434 		}
1435 	}
1436 
1437 	return 0;
1438 }
1439 
1440 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1441 {
1442 	int rem, err = -EINVAL;
1443 	struct nlattr *vf;
1444 	const struct net_device_ops *ops = dev->netdev_ops;
1445 
1446 	nla_for_each_nested(vf, attr, rem) {
1447 		switch (nla_type(vf)) {
1448 		case IFLA_VF_MAC: {
1449 			struct ifla_vf_mac *ivm;
1450 			ivm = nla_data(vf);
1451 			err = -EOPNOTSUPP;
1452 			if (ops->ndo_set_vf_mac)
1453 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1454 							  ivm->mac);
1455 			break;
1456 		}
1457 		case IFLA_VF_VLAN: {
1458 			struct ifla_vf_vlan *ivv;
1459 			ivv = nla_data(vf);
1460 			err = -EOPNOTSUPP;
1461 			if (ops->ndo_set_vf_vlan)
1462 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1463 							   ivv->vlan,
1464 							   ivv->qos);
1465 			break;
1466 		}
1467 		case IFLA_VF_TX_RATE: {
1468 			struct ifla_vf_tx_rate *ivt;
1469 			struct ifla_vf_info ivf;
1470 			ivt = nla_data(vf);
1471 			err = -EOPNOTSUPP;
1472 			if (ops->ndo_get_vf_config)
1473 				err = ops->ndo_get_vf_config(dev, ivt->vf,
1474 							     &ivf);
1475 			if (err)
1476 				break;
1477 			err = -EOPNOTSUPP;
1478 			if (ops->ndo_set_vf_rate)
1479 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1480 							   ivf.min_tx_rate,
1481 							   ivt->rate);
1482 			break;
1483 		}
1484 		case IFLA_VF_RATE: {
1485 			struct ifla_vf_rate *ivt;
1486 			ivt = nla_data(vf);
1487 			err = -EOPNOTSUPP;
1488 			if (ops->ndo_set_vf_rate)
1489 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1490 							   ivt->min_tx_rate,
1491 							   ivt->max_tx_rate);
1492 			break;
1493 		}
1494 		case IFLA_VF_SPOOFCHK: {
1495 			struct ifla_vf_spoofchk *ivs;
1496 			ivs = nla_data(vf);
1497 			err = -EOPNOTSUPP;
1498 			if (ops->ndo_set_vf_spoofchk)
1499 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1500 							       ivs->setting);
1501 			break;
1502 		}
1503 		case IFLA_VF_LINK_STATE: {
1504 			struct ifla_vf_link_state *ivl;
1505 			ivl = nla_data(vf);
1506 			err = -EOPNOTSUPP;
1507 			if (ops->ndo_set_vf_link_state)
1508 				err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1509 								 ivl->link_state);
1510 			break;
1511 		}
1512 		case IFLA_VF_RSS_QUERY_EN: {
1513 			struct ifla_vf_rss_query_en *ivrssq_en;
1514 
1515 			ivrssq_en = nla_data(vf);
1516 			err = -EOPNOTSUPP;
1517 			if (ops->ndo_set_vf_rss_query_en)
1518 				err = ops->ndo_set_vf_rss_query_en(dev,
1519 							    ivrssq_en->vf,
1520 							    ivrssq_en->setting);
1521 			break;
1522 		}
1523 		default:
1524 			err = -EINVAL;
1525 			break;
1526 		}
1527 		if (err)
1528 			break;
1529 	}
1530 	return err;
1531 }
1532 
1533 static int do_set_master(struct net_device *dev, int ifindex)
1534 {
1535 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1536 	const struct net_device_ops *ops;
1537 	int err;
1538 
1539 	if (upper_dev) {
1540 		if (upper_dev->ifindex == ifindex)
1541 			return 0;
1542 		ops = upper_dev->netdev_ops;
1543 		if (ops->ndo_del_slave) {
1544 			err = ops->ndo_del_slave(upper_dev, dev);
1545 			if (err)
1546 				return err;
1547 		} else {
1548 			return -EOPNOTSUPP;
1549 		}
1550 	}
1551 
1552 	if (ifindex) {
1553 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1554 		if (!upper_dev)
1555 			return -EINVAL;
1556 		ops = upper_dev->netdev_ops;
1557 		if (ops->ndo_add_slave) {
1558 			err = ops->ndo_add_slave(upper_dev, dev);
1559 			if (err)
1560 				return err;
1561 		} else {
1562 			return -EOPNOTSUPP;
1563 		}
1564 	}
1565 	return 0;
1566 }
1567 
1568 #define DO_SETLINK_MODIFIED	0x01
1569 /* notify flag means notify + modified. */
1570 #define DO_SETLINK_NOTIFY	0x03
1571 static int do_setlink(const struct sk_buff *skb,
1572 		      struct net_device *dev, struct ifinfomsg *ifm,
1573 		      struct nlattr **tb, char *ifname, int status)
1574 {
1575 	const struct net_device_ops *ops = dev->netdev_ops;
1576 	int err;
1577 
1578 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1579 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1580 		if (IS_ERR(net)) {
1581 			err = PTR_ERR(net);
1582 			goto errout;
1583 		}
1584 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1585 			put_net(net);
1586 			err = -EPERM;
1587 			goto errout;
1588 		}
1589 		err = dev_change_net_namespace(dev, net, ifname);
1590 		put_net(net);
1591 		if (err)
1592 			goto errout;
1593 		status |= DO_SETLINK_MODIFIED;
1594 	}
1595 
1596 	if (tb[IFLA_MAP]) {
1597 		struct rtnl_link_ifmap *u_map;
1598 		struct ifmap k_map;
1599 
1600 		if (!ops->ndo_set_config) {
1601 			err = -EOPNOTSUPP;
1602 			goto errout;
1603 		}
1604 
1605 		if (!netif_device_present(dev)) {
1606 			err = -ENODEV;
1607 			goto errout;
1608 		}
1609 
1610 		u_map = nla_data(tb[IFLA_MAP]);
1611 		k_map.mem_start = (unsigned long) u_map->mem_start;
1612 		k_map.mem_end = (unsigned long) u_map->mem_end;
1613 		k_map.base_addr = (unsigned short) u_map->base_addr;
1614 		k_map.irq = (unsigned char) u_map->irq;
1615 		k_map.dma = (unsigned char) u_map->dma;
1616 		k_map.port = (unsigned char) u_map->port;
1617 
1618 		err = ops->ndo_set_config(dev, &k_map);
1619 		if (err < 0)
1620 			goto errout;
1621 
1622 		status |= DO_SETLINK_NOTIFY;
1623 	}
1624 
1625 	if (tb[IFLA_ADDRESS]) {
1626 		struct sockaddr *sa;
1627 		int len;
1628 
1629 		len = sizeof(sa_family_t) + dev->addr_len;
1630 		sa = kmalloc(len, GFP_KERNEL);
1631 		if (!sa) {
1632 			err = -ENOMEM;
1633 			goto errout;
1634 		}
1635 		sa->sa_family = dev->type;
1636 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1637 		       dev->addr_len);
1638 		err = dev_set_mac_address(dev, sa);
1639 		kfree(sa);
1640 		if (err)
1641 			goto errout;
1642 		status |= DO_SETLINK_MODIFIED;
1643 	}
1644 
1645 	if (tb[IFLA_MTU]) {
1646 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1647 		if (err < 0)
1648 			goto errout;
1649 		status |= DO_SETLINK_MODIFIED;
1650 	}
1651 
1652 	if (tb[IFLA_GROUP]) {
1653 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1654 		status |= DO_SETLINK_NOTIFY;
1655 	}
1656 
1657 	/*
1658 	 * Interface selected by interface index but interface
1659 	 * name provided implies that a name change has been
1660 	 * requested.
1661 	 */
1662 	if (ifm->ifi_index > 0 && ifname[0]) {
1663 		err = dev_change_name(dev, ifname);
1664 		if (err < 0)
1665 			goto errout;
1666 		status |= DO_SETLINK_MODIFIED;
1667 	}
1668 
1669 	if (tb[IFLA_IFALIAS]) {
1670 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1671 				    nla_len(tb[IFLA_IFALIAS]));
1672 		if (err < 0)
1673 			goto errout;
1674 		status |= DO_SETLINK_NOTIFY;
1675 	}
1676 
1677 	if (tb[IFLA_BROADCAST]) {
1678 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1679 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1680 	}
1681 
1682 	if (ifm->ifi_flags || ifm->ifi_change) {
1683 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1684 		if (err < 0)
1685 			goto errout;
1686 	}
1687 
1688 	if (tb[IFLA_MASTER]) {
1689 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1690 		if (err)
1691 			goto errout;
1692 		status |= DO_SETLINK_MODIFIED;
1693 	}
1694 
1695 	if (tb[IFLA_CARRIER]) {
1696 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1697 		if (err)
1698 			goto errout;
1699 		status |= DO_SETLINK_MODIFIED;
1700 	}
1701 
1702 	if (tb[IFLA_TXQLEN]) {
1703 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1704 
1705 		if (dev->tx_queue_len ^ value)
1706 			status |= DO_SETLINK_NOTIFY;
1707 
1708 		dev->tx_queue_len = value;
1709 	}
1710 
1711 	if (tb[IFLA_OPERSTATE])
1712 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1713 
1714 	if (tb[IFLA_LINKMODE]) {
1715 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1716 
1717 		write_lock_bh(&dev_base_lock);
1718 		if (dev->link_mode ^ value)
1719 			status |= DO_SETLINK_NOTIFY;
1720 		dev->link_mode = value;
1721 		write_unlock_bh(&dev_base_lock);
1722 	}
1723 
1724 	if (tb[IFLA_VFINFO_LIST]) {
1725 		struct nlattr *attr;
1726 		int rem;
1727 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1728 			if (nla_type(attr) != IFLA_VF_INFO) {
1729 				err = -EINVAL;
1730 				goto errout;
1731 			}
1732 			err = do_setvfinfo(dev, attr);
1733 			if (err < 0)
1734 				goto errout;
1735 			status |= DO_SETLINK_NOTIFY;
1736 		}
1737 	}
1738 	err = 0;
1739 
1740 	if (tb[IFLA_VF_PORTS]) {
1741 		struct nlattr *port[IFLA_PORT_MAX+1];
1742 		struct nlattr *attr;
1743 		int vf;
1744 		int rem;
1745 
1746 		err = -EOPNOTSUPP;
1747 		if (!ops->ndo_set_vf_port)
1748 			goto errout;
1749 
1750 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1751 			if (nla_type(attr) != IFLA_VF_PORT)
1752 				continue;
1753 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1754 				attr, ifla_port_policy);
1755 			if (err < 0)
1756 				goto errout;
1757 			if (!port[IFLA_PORT_VF]) {
1758 				err = -EOPNOTSUPP;
1759 				goto errout;
1760 			}
1761 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1762 			err = ops->ndo_set_vf_port(dev, vf, port);
1763 			if (err < 0)
1764 				goto errout;
1765 			status |= DO_SETLINK_NOTIFY;
1766 		}
1767 	}
1768 	err = 0;
1769 
1770 	if (tb[IFLA_PORT_SELF]) {
1771 		struct nlattr *port[IFLA_PORT_MAX+1];
1772 
1773 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1774 			tb[IFLA_PORT_SELF], ifla_port_policy);
1775 		if (err < 0)
1776 			goto errout;
1777 
1778 		err = -EOPNOTSUPP;
1779 		if (ops->ndo_set_vf_port)
1780 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1781 		if (err < 0)
1782 			goto errout;
1783 		status |= DO_SETLINK_NOTIFY;
1784 	}
1785 
1786 	if (tb[IFLA_AF_SPEC]) {
1787 		struct nlattr *af;
1788 		int rem;
1789 
1790 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1791 			const struct rtnl_af_ops *af_ops;
1792 
1793 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1794 				BUG();
1795 
1796 			err = af_ops->set_link_af(dev, af);
1797 			if (err < 0)
1798 				goto errout;
1799 
1800 			status |= DO_SETLINK_NOTIFY;
1801 		}
1802 	}
1803 	err = 0;
1804 
1805 errout:
1806 	if (status & DO_SETLINK_MODIFIED) {
1807 		if (status & DO_SETLINK_NOTIFY)
1808 			netdev_state_change(dev);
1809 
1810 		if (err < 0)
1811 			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",
1812 					     dev->name);
1813 	}
1814 
1815 	return err;
1816 }
1817 
1818 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1819 {
1820 	struct net *net = sock_net(skb->sk);
1821 	struct ifinfomsg *ifm;
1822 	struct net_device *dev;
1823 	int err;
1824 	struct nlattr *tb[IFLA_MAX+1];
1825 	char ifname[IFNAMSIZ];
1826 
1827 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1828 	if (err < 0)
1829 		goto errout;
1830 
1831 	if (tb[IFLA_IFNAME])
1832 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1833 	else
1834 		ifname[0] = '\0';
1835 
1836 	err = -EINVAL;
1837 	ifm = nlmsg_data(nlh);
1838 	if (ifm->ifi_index > 0)
1839 		dev = __dev_get_by_index(net, ifm->ifi_index);
1840 	else if (tb[IFLA_IFNAME])
1841 		dev = __dev_get_by_name(net, ifname);
1842 	else
1843 		goto errout;
1844 
1845 	if (dev == NULL) {
1846 		err = -ENODEV;
1847 		goto errout;
1848 	}
1849 
1850 	err = validate_linkmsg(dev, tb);
1851 	if (err < 0)
1852 		goto errout;
1853 
1854 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1855 errout:
1856 	return err;
1857 }
1858 
1859 static int rtnl_group_dellink(const struct net *net, int group)
1860 {
1861 	struct net_device *dev, *aux;
1862 	LIST_HEAD(list_kill);
1863 	bool found = false;
1864 
1865 	if (!group)
1866 		return -EPERM;
1867 
1868 	for_each_netdev(net, dev) {
1869 		if (dev->group == group) {
1870 			const struct rtnl_link_ops *ops;
1871 
1872 			found = true;
1873 			ops = dev->rtnl_link_ops;
1874 			if (!ops || !ops->dellink)
1875 				return -EOPNOTSUPP;
1876 		}
1877 	}
1878 
1879 	if (!found)
1880 		return -ENODEV;
1881 
1882 	for_each_netdev_safe(net, dev, aux) {
1883 		if (dev->group == group) {
1884 			const struct rtnl_link_ops *ops;
1885 
1886 			ops = dev->rtnl_link_ops;
1887 			ops->dellink(dev, &list_kill);
1888 		}
1889 	}
1890 	unregister_netdevice_many(&list_kill);
1891 
1892 	return 0;
1893 }
1894 
1895 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1896 {
1897 	struct net *net = sock_net(skb->sk);
1898 	const struct rtnl_link_ops *ops;
1899 	struct net_device *dev;
1900 	struct ifinfomsg *ifm;
1901 	char ifname[IFNAMSIZ];
1902 	struct nlattr *tb[IFLA_MAX+1];
1903 	int err;
1904 	LIST_HEAD(list_kill);
1905 
1906 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1907 	if (err < 0)
1908 		return err;
1909 
1910 	if (tb[IFLA_IFNAME])
1911 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1912 
1913 	ifm = nlmsg_data(nlh);
1914 	if (ifm->ifi_index > 0)
1915 		dev = __dev_get_by_index(net, ifm->ifi_index);
1916 	else if (tb[IFLA_IFNAME])
1917 		dev = __dev_get_by_name(net, ifname);
1918 	else if (tb[IFLA_GROUP])
1919 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
1920 	else
1921 		return -EINVAL;
1922 
1923 	if (!dev)
1924 		return -ENODEV;
1925 
1926 	ops = dev->rtnl_link_ops;
1927 	if (!ops || !ops->dellink)
1928 		return -EOPNOTSUPP;
1929 
1930 	ops->dellink(dev, &list_kill);
1931 	unregister_netdevice_many(&list_kill);
1932 	return 0;
1933 }
1934 
1935 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1936 {
1937 	unsigned int old_flags;
1938 	int err;
1939 
1940 	old_flags = dev->flags;
1941 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1942 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1943 		if (err < 0)
1944 			return err;
1945 	}
1946 
1947 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1948 
1949 	__dev_notify_flags(dev, old_flags, ~0U);
1950 	return 0;
1951 }
1952 EXPORT_SYMBOL(rtnl_configure_link);
1953 
1954 struct net_device *rtnl_create_link(struct net *net,
1955 	const char *ifname, unsigned char name_assign_type,
1956 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
1957 {
1958 	int err;
1959 	struct net_device *dev;
1960 	unsigned int num_tx_queues = 1;
1961 	unsigned int num_rx_queues = 1;
1962 
1963 	if (tb[IFLA_NUM_TX_QUEUES])
1964 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1965 	else if (ops->get_num_tx_queues)
1966 		num_tx_queues = ops->get_num_tx_queues();
1967 
1968 	if (tb[IFLA_NUM_RX_QUEUES])
1969 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1970 	else if (ops->get_num_rx_queues)
1971 		num_rx_queues = ops->get_num_rx_queues();
1972 
1973 	err = -ENOMEM;
1974 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1975 			       ops->setup, num_tx_queues, num_rx_queues);
1976 	if (!dev)
1977 		goto err;
1978 
1979 	dev_net_set(dev, net);
1980 	dev->rtnl_link_ops = ops;
1981 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1982 
1983 	if (tb[IFLA_MTU])
1984 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1985 	if (tb[IFLA_ADDRESS]) {
1986 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1987 				nla_len(tb[IFLA_ADDRESS]));
1988 		dev->addr_assign_type = NET_ADDR_SET;
1989 	}
1990 	if (tb[IFLA_BROADCAST])
1991 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1992 				nla_len(tb[IFLA_BROADCAST]));
1993 	if (tb[IFLA_TXQLEN])
1994 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1995 	if (tb[IFLA_OPERSTATE])
1996 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1997 	if (tb[IFLA_LINKMODE])
1998 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1999 	if (tb[IFLA_GROUP])
2000 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2001 
2002 	return dev;
2003 
2004 err:
2005 	return ERR_PTR(err);
2006 }
2007 EXPORT_SYMBOL(rtnl_create_link);
2008 
2009 static int rtnl_group_changelink(const struct sk_buff *skb,
2010 		struct net *net, int group,
2011 		struct ifinfomsg *ifm,
2012 		struct nlattr **tb)
2013 {
2014 	struct net_device *dev, *aux;
2015 	int err;
2016 
2017 	for_each_netdev_safe(net, dev, aux) {
2018 		if (dev->group == group) {
2019 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2020 			if (err < 0)
2021 				return err;
2022 		}
2023 	}
2024 
2025 	return 0;
2026 }
2027 
2028 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2029 {
2030 	struct net *net = sock_net(skb->sk);
2031 	const struct rtnl_link_ops *ops;
2032 	const struct rtnl_link_ops *m_ops = NULL;
2033 	struct net_device *dev;
2034 	struct net_device *master_dev = NULL;
2035 	struct ifinfomsg *ifm;
2036 	char kind[MODULE_NAME_LEN];
2037 	char ifname[IFNAMSIZ];
2038 	struct nlattr *tb[IFLA_MAX+1];
2039 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2040 	unsigned char name_assign_type = NET_NAME_USER;
2041 	int err;
2042 
2043 #ifdef CONFIG_MODULES
2044 replay:
2045 #endif
2046 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2047 	if (err < 0)
2048 		return err;
2049 
2050 	if (tb[IFLA_IFNAME])
2051 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2052 	else
2053 		ifname[0] = '\0';
2054 
2055 	ifm = nlmsg_data(nlh);
2056 	if (ifm->ifi_index > 0)
2057 		dev = __dev_get_by_index(net, ifm->ifi_index);
2058 	else {
2059 		if (ifname[0])
2060 			dev = __dev_get_by_name(net, ifname);
2061 		else
2062 			dev = NULL;
2063 	}
2064 
2065 	if (dev) {
2066 		master_dev = netdev_master_upper_dev_get(dev);
2067 		if (master_dev)
2068 			m_ops = master_dev->rtnl_link_ops;
2069 	}
2070 
2071 	err = validate_linkmsg(dev, tb);
2072 	if (err < 0)
2073 		return err;
2074 
2075 	if (tb[IFLA_LINKINFO]) {
2076 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2077 				       tb[IFLA_LINKINFO], ifla_info_policy);
2078 		if (err < 0)
2079 			return err;
2080 	} else
2081 		memset(linkinfo, 0, sizeof(linkinfo));
2082 
2083 	if (linkinfo[IFLA_INFO_KIND]) {
2084 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2085 		ops = rtnl_link_ops_get(kind);
2086 	} else {
2087 		kind[0] = '\0';
2088 		ops = NULL;
2089 	}
2090 
2091 	if (1) {
2092 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2093 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2094 		struct nlattr **data = NULL;
2095 		struct nlattr **slave_data = NULL;
2096 		struct net *dest_net, *link_net = NULL;
2097 
2098 		if (ops) {
2099 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2100 				err = nla_parse_nested(attr, ops->maxtype,
2101 						       linkinfo[IFLA_INFO_DATA],
2102 						       ops->policy);
2103 				if (err < 0)
2104 					return err;
2105 				data = attr;
2106 			}
2107 			if (ops->validate) {
2108 				err = ops->validate(tb, data);
2109 				if (err < 0)
2110 					return err;
2111 			}
2112 		}
2113 
2114 		if (m_ops) {
2115 			if (m_ops->slave_maxtype &&
2116 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2117 				err = nla_parse_nested(slave_attr,
2118 						       m_ops->slave_maxtype,
2119 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2120 						       m_ops->slave_policy);
2121 				if (err < 0)
2122 					return err;
2123 				slave_data = slave_attr;
2124 			}
2125 			if (m_ops->slave_validate) {
2126 				err = m_ops->slave_validate(tb, slave_data);
2127 				if (err < 0)
2128 					return err;
2129 			}
2130 		}
2131 
2132 		if (dev) {
2133 			int status = 0;
2134 
2135 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2136 				return -EEXIST;
2137 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2138 				return -EOPNOTSUPP;
2139 
2140 			if (linkinfo[IFLA_INFO_DATA]) {
2141 				if (!ops || ops != dev->rtnl_link_ops ||
2142 				    !ops->changelink)
2143 					return -EOPNOTSUPP;
2144 
2145 				err = ops->changelink(dev, tb, data);
2146 				if (err < 0)
2147 					return err;
2148 				status |= DO_SETLINK_NOTIFY;
2149 			}
2150 
2151 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2152 				if (!m_ops || !m_ops->slave_changelink)
2153 					return -EOPNOTSUPP;
2154 
2155 				err = m_ops->slave_changelink(master_dev, dev,
2156 							      tb, slave_data);
2157 				if (err < 0)
2158 					return err;
2159 				status |= DO_SETLINK_NOTIFY;
2160 			}
2161 
2162 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2163 		}
2164 
2165 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2166 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2167 				return rtnl_group_changelink(skb, net,
2168 						nla_get_u32(tb[IFLA_GROUP]),
2169 						ifm, tb);
2170 			return -ENODEV;
2171 		}
2172 
2173 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2174 			return -EOPNOTSUPP;
2175 
2176 		if (!ops) {
2177 #ifdef CONFIG_MODULES
2178 			if (kind[0]) {
2179 				__rtnl_unlock();
2180 				request_module("rtnl-link-%s", kind);
2181 				rtnl_lock();
2182 				ops = rtnl_link_ops_get(kind);
2183 				if (ops)
2184 					goto replay;
2185 			}
2186 #endif
2187 			return -EOPNOTSUPP;
2188 		}
2189 
2190 		if (!ops->setup)
2191 			return -EOPNOTSUPP;
2192 
2193 		if (!ifname[0]) {
2194 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2195 			name_assign_type = NET_NAME_ENUM;
2196 		}
2197 
2198 		dest_net = rtnl_link_get_net(net, tb);
2199 		if (IS_ERR(dest_net))
2200 			return PTR_ERR(dest_net);
2201 
2202 		err = -EPERM;
2203 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2204 			goto out;
2205 
2206 		if (tb[IFLA_LINK_NETNSID]) {
2207 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2208 
2209 			link_net = get_net_ns_by_id(dest_net, id);
2210 			if (!link_net) {
2211 				err =  -EINVAL;
2212 				goto out;
2213 			}
2214 			err = -EPERM;
2215 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2216 				goto out;
2217 		}
2218 
2219 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2220 				       name_assign_type, ops, tb);
2221 		if (IS_ERR(dev)) {
2222 			err = PTR_ERR(dev);
2223 			goto out;
2224 		}
2225 
2226 		dev->ifindex = ifm->ifi_index;
2227 
2228 		if (ops->newlink) {
2229 			err = ops->newlink(link_net ? : net, dev, tb, data);
2230 			/* Drivers should call free_netdev() in ->destructor
2231 			 * and unregister it on failure after registration
2232 			 * so that device could be finally freed in rtnl_unlock.
2233 			 */
2234 			if (err < 0) {
2235 				/* If device is not registered at all, free it now */
2236 				if (dev->reg_state == NETREG_UNINITIALIZED)
2237 					free_netdev(dev);
2238 				goto out;
2239 			}
2240 		} else {
2241 			err = register_netdevice(dev);
2242 			if (err < 0) {
2243 				free_netdev(dev);
2244 				goto out;
2245 			}
2246 		}
2247 		err = rtnl_configure_link(dev, ifm);
2248 		if (err < 0)
2249 			goto out_unregister;
2250 		if (link_net) {
2251 			err = dev_change_net_namespace(dev, dest_net, ifname);
2252 			if (err < 0)
2253 				goto out_unregister;
2254 		}
2255 out:
2256 		if (link_net)
2257 			put_net(link_net);
2258 		put_net(dest_net);
2259 		return err;
2260 out_unregister:
2261 		if (ops->newlink) {
2262 			LIST_HEAD(list_kill);
2263 
2264 			ops->dellink(dev, &list_kill);
2265 			unregister_netdevice_many(&list_kill);
2266 		} else {
2267 			unregister_netdevice(dev);
2268 		}
2269 		goto out;
2270 	}
2271 }
2272 
2273 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2274 {
2275 	struct net *net = sock_net(skb->sk);
2276 	struct ifinfomsg *ifm;
2277 	char ifname[IFNAMSIZ];
2278 	struct nlattr *tb[IFLA_MAX+1];
2279 	struct net_device *dev = NULL;
2280 	struct sk_buff *nskb;
2281 	int err;
2282 	u32 ext_filter_mask = 0;
2283 
2284 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2285 	if (err < 0)
2286 		return err;
2287 
2288 	if (tb[IFLA_IFNAME])
2289 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2290 
2291 	if (tb[IFLA_EXT_MASK])
2292 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2293 
2294 	ifm = nlmsg_data(nlh);
2295 	if (ifm->ifi_index > 0)
2296 		dev = __dev_get_by_index(net, ifm->ifi_index);
2297 	else if (tb[IFLA_IFNAME])
2298 		dev = __dev_get_by_name(net, ifname);
2299 	else
2300 		return -EINVAL;
2301 
2302 	if (dev == NULL)
2303 		return -ENODEV;
2304 
2305 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2306 	if (nskb == NULL)
2307 		return -ENOBUFS;
2308 
2309 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2310 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2311 	if (err < 0) {
2312 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2313 		WARN_ON(err == -EMSGSIZE);
2314 		kfree_skb(nskb);
2315 	} else
2316 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2317 
2318 	return err;
2319 }
2320 
2321 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2322 {
2323 	struct net *net = sock_net(skb->sk);
2324 	struct net_device *dev;
2325 	struct nlattr *tb[IFLA_MAX+1];
2326 	u32 ext_filter_mask = 0;
2327 	u16 min_ifinfo_dump_size = 0;
2328 	int hdrlen;
2329 
2330 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2331 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2332 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2333 
2334 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2335 		if (tb[IFLA_EXT_MASK])
2336 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2337 	}
2338 
2339 	if (!ext_filter_mask)
2340 		return NLMSG_GOODSIZE;
2341 	/*
2342 	 * traverse the list of net devices and compute the minimum
2343 	 * buffer size based upon the filter mask.
2344 	 */
2345 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2346 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2347 					     if_nlmsg_size(dev,
2348 						           ext_filter_mask));
2349 	}
2350 
2351 	return min_ifinfo_dump_size;
2352 }
2353 
2354 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2355 {
2356 	int idx;
2357 	int s_idx = cb->family;
2358 
2359 	if (s_idx == 0)
2360 		s_idx = 1;
2361 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2362 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2363 		if (idx < s_idx || idx == PF_PACKET)
2364 			continue;
2365 		if (rtnl_msg_handlers[idx] == NULL ||
2366 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2367 			continue;
2368 		if (idx > s_idx) {
2369 			memset(&cb->args[0], 0, sizeof(cb->args));
2370 			cb->prev_seq = 0;
2371 			cb->seq = 0;
2372 		}
2373 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2374 			break;
2375 	}
2376 	cb->family = idx;
2377 
2378 	return skb->len;
2379 }
2380 
2381 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2382 				       unsigned int change, gfp_t flags)
2383 {
2384 	struct net *net = dev_net(dev);
2385 	struct sk_buff *skb;
2386 	int err = -ENOBUFS;
2387 	size_t if_info_size;
2388 
2389 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2390 	if (skb == NULL)
2391 		goto errout;
2392 
2393 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2394 	if (err < 0) {
2395 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2396 		WARN_ON(err == -EMSGSIZE);
2397 		kfree_skb(skb);
2398 		goto errout;
2399 	}
2400 	return skb;
2401 errout:
2402 	if (err < 0)
2403 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2404 	return NULL;
2405 }
2406 
2407 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2408 {
2409 	struct net *net = dev_net(dev);
2410 
2411 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2412 }
2413 
2414 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2415 		  gfp_t flags)
2416 {
2417 	struct sk_buff *skb;
2418 
2419 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2420 	if (skb)
2421 		rtmsg_ifinfo_send(skb, dev, flags);
2422 }
2423 EXPORT_SYMBOL(rtmsg_ifinfo);
2424 
2425 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2426 				   struct net_device *dev,
2427 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2428 				   int type, unsigned int flags,
2429 				   int nlflags)
2430 {
2431 	struct nlmsghdr *nlh;
2432 	struct ndmsg *ndm;
2433 
2434 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2435 	if (!nlh)
2436 		return -EMSGSIZE;
2437 
2438 	ndm = nlmsg_data(nlh);
2439 	ndm->ndm_family  = AF_BRIDGE;
2440 	ndm->ndm_pad1	 = 0;
2441 	ndm->ndm_pad2    = 0;
2442 	ndm->ndm_flags	 = flags;
2443 	ndm->ndm_type	 = 0;
2444 	ndm->ndm_ifindex = dev->ifindex;
2445 	ndm->ndm_state   = NUD_PERMANENT;
2446 
2447 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2448 		goto nla_put_failure;
2449 	if (vid)
2450 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2451 			goto nla_put_failure;
2452 
2453 	nlmsg_end(skb, nlh);
2454 	return 0;
2455 
2456 nla_put_failure:
2457 	nlmsg_cancel(skb, nlh);
2458 	return -EMSGSIZE;
2459 }
2460 
2461 static inline size_t rtnl_fdb_nlmsg_size(void)
2462 {
2463 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2464 }
2465 
2466 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2467 {
2468 	struct net *net = dev_net(dev);
2469 	struct sk_buff *skb;
2470 	int err = -ENOBUFS;
2471 
2472 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2473 	if (!skb)
2474 		goto errout;
2475 
2476 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2477 				      0, 0, type, NTF_SELF, 0);
2478 	if (err < 0) {
2479 		kfree_skb(skb);
2480 		goto errout;
2481 	}
2482 
2483 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2484 	return;
2485 errout:
2486 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2487 }
2488 
2489 /**
2490  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2491  */
2492 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2493 		     struct nlattr *tb[],
2494 		     struct net_device *dev,
2495 		     const unsigned char *addr, u16 vid,
2496 		     u16 flags)
2497 {
2498 	int err = -EINVAL;
2499 
2500 	/* If aging addresses are supported device will need to
2501 	 * implement its own handler for this.
2502 	 */
2503 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2504 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2505 		return err;
2506 	}
2507 
2508 	if (vid) {
2509 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2510 		return err;
2511 	}
2512 
2513 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2514 		err = dev_uc_add_excl(dev, addr);
2515 	else if (is_multicast_ether_addr(addr))
2516 		err = dev_mc_add_excl(dev, addr);
2517 
2518 	/* Only return duplicate errors if NLM_F_EXCL is set */
2519 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2520 		err = 0;
2521 
2522 	return err;
2523 }
2524 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2525 
2526 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2527 {
2528 	u16 vid = 0;
2529 
2530 	if (vlan_attr) {
2531 		if (nla_len(vlan_attr) != sizeof(u16)) {
2532 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2533 			return -EINVAL;
2534 		}
2535 
2536 		vid = nla_get_u16(vlan_attr);
2537 
2538 		if (!vid || vid >= VLAN_VID_MASK) {
2539 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2540 				vid);
2541 			return -EINVAL;
2542 		}
2543 	}
2544 	*p_vid = vid;
2545 	return 0;
2546 }
2547 
2548 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2549 {
2550 	struct net *net = sock_net(skb->sk);
2551 	struct ndmsg *ndm;
2552 	struct nlattr *tb[NDA_MAX+1];
2553 	struct net_device *dev;
2554 	u8 *addr;
2555 	u16 vid;
2556 	int err;
2557 
2558 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2559 	if (err < 0)
2560 		return err;
2561 
2562 	ndm = nlmsg_data(nlh);
2563 	if (ndm->ndm_ifindex == 0) {
2564 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2565 		return -EINVAL;
2566 	}
2567 
2568 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2569 	if (dev == NULL) {
2570 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2571 		return -ENODEV;
2572 	}
2573 
2574 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2575 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2576 		return -EINVAL;
2577 	}
2578 
2579 	addr = nla_data(tb[NDA_LLADDR]);
2580 
2581 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2582 	if (err)
2583 		return err;
2584 
2585 	err = -EOPNOTSUPP;
2586 
2587 	/* Support fdb on master device the net/bridge default case */
2588 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2589 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2590 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2591 		const struct net_device_ops *ops = br_dev->netdev_ops;
2592 
2593 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2594 				       nlh->nlmsg_flags);
2595 		if (err)
2596 			goto out;
2597 		else
2598 			ndm->ndm_flags &= ~NTF_MASTER;
2599 	}
2600 
2601 	/* Embedded bridge, macvlan, and any other device support */
2602 	if ((ndm->ndm_flags & NTF_SELF)) {
2603 		if (dev->netdev_ops->ndo_fdb_add)
2604 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2605 							   vid,
2606 							   nlh->nlmsg_flags);
2607 		else
2608 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2609 					       nlh->nlmsg_flags);
2610 
2611 		if (!err) {
2612 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2613 			ndm->ndm_flags &= ~NTF_SELF;
2614 		}
2615 	}
2616 out:
2617 	return err;
2618 }
2619 
2620 /**
2621  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2622  */
2623 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2624 		     struct nlattr *tb[],
2625 		     struct net_device *dev,
2626 		     const unsigned char *addr, u16 vid)
2627 {
2628 	int err = -EINVAL;
2629 
2630 	/* If aging addresses are supported device will need to
2631 	 * implement its own handler for this.
2632 	 */
2633 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2634 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2635 		return err;
2636 	}
2637 
2638 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2639 		err = dev_uc_del(dev, addr);
2640 	else if (is_multicast_ether_addr(addr))
2641 		err = dev_mc_del(dev, addr);
2642 
2643 	return err;
2644 }
2645 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2646 
2647 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2648 {
2649 	struct net *net = sock_net(skb->sk);
2650 	struct ndmsg *ndm;
2651 	struct nlattr *tb[NDA_MAX+1];
2652 	struct net_device *dev;
2653 	int err = -EINVAL;
2654 	__u8 *addr;
2655 	u16 vid;
2656 
2657 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2658 		return -EPERM;
2659 
2660 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2661 	if (err < 0)
2662 		return err;
2663 
2664 	ndm = nlmsg_data(nlh);
2665 	if (ndm->ndm_ifindex == 0) {
2666 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2667 		return -EINVAL;
2668 	}
2669 
2670 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2671 	if (dev == NULL) {
2672 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2673 		return -ENODEV;
2674 	}
2675 
2676 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2677 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2678 		return -EINVAL;
2679 	}
2680 
2681 	addr = nla_data(tb[NDA_LLADDR]);
2682 
2683 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2684 	if (err)
2685 		return err;
2686 
2687 	err = -EOPNOTSUPP;
2688 
2689 	/* Support fdb on master device the net/bridge default case */
2690 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2691 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2692 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2693 		const struct net_device_ops *ops = br_dev->netdev_ops;
2694 
2695 		if (ops->ndo_fdb_del)
2696 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2697 
2698 		if (err)
2699 			goto out;
2700 		else
2701 			ndm->ndm_flags &= ~NTF_MASTER;
2702 	}
2703 
2704 	/* Embedded bridge, macvlan, and any other device support */
2705 	if (ndm->ndm_flags & NTF_SELF) {
2706 		if (dev->netdev_ops->ndo_fdb_del)
2707 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2708 							   vid);
2709 		else
2710 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2711 
2712 		if (!err) {
2713 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2714 			ndm->ndm_flags &= ~NTF_SELF;
2715 		}
2716 	}
2717 out:
2718 	return err;
2719 }
2720 
2721 static int nlmsg_populate_fdb(struct sk_buff *skb,
2722 			      struct netlink_callback *cb,
2723 			      struct net_device *dev,
2724 			      int *idx,
2725 			      struct netdev_hw_addr_list *list)
2726 {
2727 	struct netdev_hw_addr *ha;
2728 	int err;
2729 	u32 portid, seq;
2730 
2731 	portid = NETLINK_CB(cb->skb).portid;
2732 	seq = cb->nlh->nlmsg_seq;
2733 
2734 	list_for_each_entry(ha, &list->list, list) {
2735 		if (*idx < cb->args[0])
2736 			goto skip;
2737 
2738 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2739 					      portid, seq,
2740 					      RTM_NEWNEIGH, NTF_SELF,
2741 					      NLM_F_MULTI);
2742 		if (err < 0)
2743 			return err;
2744 skip:
2745 		*idx += 1;
2746 	}
2747 	return 0;
2748 }
2749 
2750 /**
2751  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2752  * @nlh: netlink message header
2753  * @dev: netdevice
2754  *
2755  * Default netdevice operation to dump the existing unicast address list.
2756  * Returns number of addresses from list put in skb.
2757  */
2758 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2759 		      struct netlink_callback *cb,
2760 		      struct net_device *dev,
2761 		      struct net_device *filter_dev,
2762 		      int idx)
2763 {
2764 	int err;
2765 
2766 	netif_addr_lock_bh(dev);
2767 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2768 	if (err)
2769 		goto out;
2770 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2771 out:
2772 	netif_addr_unlock_bh(dev);
2773 	return idx;
2774 }
2775 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2776 
2777 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2778 {
2779 	struct net_device *dev;
2780 	struct nlattr *tb[IFLA_MAX+1];
2781 	struct net_device *br_dev = NULL;
2782 	const struct net_device_ops *ops = NULL;
2783 	const struct net_device_ops *cops = NULL;
2784 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2785 	struct net *net = sock_net(skb->sk);
2786 	int brport_idx = 0;
2787 	int br_idx = 0;
2788 	int idx = 0;
2789 
2790 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2791 			ifla_policy) == 0) {
2792 		if (tb[IFLA_MASTER])
2793 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
2794 	}
2795 
2796 	brport_idx = ifm->ifi_index;
2797 
2798 	if (br_idx) {
2799 		br_dev = __dev_get_by_index(net, br_idx);
2800 		if (!br_dev)
2801 			return -ENODEV;
2802 
2803 		ops = br_dev->netdev_ops;
2804 	}
2805 
2806 	for_each_netdev(net, dev) {
2807 		if (brport_idx && (dev->ifindex != brport_idx))
2808 			continue;
2809 
2810 		if (!br_idx) { /* user did not specify a specific bridge */
2811 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
2812 				br_dev = netdev_master_upper_dev_get(dev);
2813 				cops = br_dev->netdev_ops;
2814 			}
2815 
2816 		} else {
2817 			if (dev != br_dev &&
2818 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
2819 				continue;
2820 
2821 			if (br_dev != netdev_master_upper_dev_get(dev) &&
2822 			    !(dev->priv_flags & IFF_EBRIDGE))
2823 				continue;
2824 
2825 			cops = ops;
2826 		}
2827 
2828 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2829 			if (cops && cops->ndo_fdb_dump)
2830 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2831 							 idx);
2832 		}
2833 
2834 		if (dev->netdev_ops->ndo_fdb_dump)
2835 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2836 							    idx);
2837 		else
2838 			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2839 
2840 		cops = NULL;
2841 	}
2842 
2843 	cb->args[0] = idx;
2844 	return skb->len;
2845 }
2846 
2847 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2848 			       unsigned int attrnum, unsigned int flag)
2849 {
2850 	if (mask & flag)
2851 		return nla_put_u8(skb, attrnum, !!(flags & flag));
2852 	return 0;
2853 }
2854 
2855 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2856 			    struct net_device *dev, u16 mode,
2857 			    u32 flags, u32 mask, int nlflags)
2858 {
2859 	struct nlmsghdr *nlh;
2860 	struct ifinfomsg *ifm;
2861 	struct nlattr *br_afspec;
2862 	struct nlattr *protinfo;
2863 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2864 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2865 
2866 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2867 	if (nlh == NULL)
2868 		return -EMSGSIZE;
2869 
2870 	ifm = nlmsg_data(nlh);
2871 	ifm->ifi_family = AF_BRIDGE;
2872 	ifm->__ifi_pad = 0;
2873 	ifm->ifi_type = dev->type;
2874 	ifm->ifi_index = dev->ifindex;
2875 	ifm->ifi_flags = dev_get_flags(dev);
2876 	ifm->ifi_change = 0;
2877 
2878 
2879 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2880 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2881 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2882 	    (br_dev &&
2883 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2884 	    (dev->addr_len &&
2885 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2886 	    (dev->ifindex != dev_get_iflink(dev) &&
2887 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
2888 		goto nla_put_failure;
2889 
2890 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2891 	if (!br_afspec)
2892 		goto nla_put_failure;
2893 
2894 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2895 		nla_nest_cancel(skb, br_afspec);
2896 		goto nla_put_failure;
2897 	}
2898 
2899 	if (mode != BRIDGE_MODE_UNDEF) {
2900 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2901 			nla_nest_cancel(skb, br_afspec);
2902 			goto nla_put_failure;
2903 		}
2904 	}
2905 	nla_nest_end(skb, br_afspec);
2906 
2907 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2908 	if (!protinfo)
2909 		goto nla_put_failure;
2910 
2911 	if (brport_nla_put_flag(skb, flags, mask,
2912 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2913 	    brport_nla_put_flag(skb, flags, mask,
2914 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2915 	    brport_nla_put_flag(skb, flags, mask,
2916 				IFLA_BRPORT_FAST_LEAVE,
2917 				BR_MULTICAST_FAST_LEAVE) ||
2918 	    brport_nla_put_flag(skb, flags, mask,
2919 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2920 	    brport_nla_put_flag(skb, flags, mask,
2921 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2922 	    brport_nla_put_flag(skb, flags, mask,
2923 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2924 	    brport_nla_put_flag(skb, flags, mask,
2925 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2926 	    brport_nla_put_flag(skb, flags, mask,
2927 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2928 		nla_nest_cancel(skb, protinfo);
2929 		goto nla_put_failure;
2930 	}
2931 
2932 	nla_nest_end(skb, protinfo);
2933 
2934 	nlmsg_end(skb, nlh);
2935 	return 0;
2936 nla_put_failure:
2937 	nlmsg_cancel(skb, nlh);
2938 	return -EMSGSIZE;
2939 }
2940 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2941 
2942 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2943 {
2944 	struct net *net = sock_net(skb->sk);
2945 	struct net_device *dev;
2946 	int idx = 0;
2947 	u32 portid = NETLINK_CB(cb->skb).portid;
2948 	u32 seq = cb->nlh->nlmsg_seq;
2949 	u32 filter_mask = 0;
2950 
2951 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2952 		struct nlattr *extfilt;
2953 
2954 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2955 					  IFLA_EXT_MASK);
2956 		if (extfilt) {
2957 			if (nla_len(extfilt) < sizeof(filter_mask))
2958 				return -EINVAL;
2959 
2960 			filter_mask = nla_get_u32(extfilt);
2961 		}
2962 	}
2963 
2964 	rcu_read_lock();
2965 	for_each_netdev_rcu(net, dev) {
2966 		const struct net_device_ops *ops = dev->netdev_ops;
2967 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2968 
2969 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2970 			if (idx >= cb->args[0] &&
2971 			    br_dev->netdev_ops->ndo_bridge_getlink(
2972 				    skb, portid, seq, dev, filter_mask,
2973 				    NLM_F_MULTI) < 0)
2974 				break;
2975 			idx++;
2976 		}
2977 
2978 		if (ops->ndo_bridge_getlink) {
2979 			if (idx >= cb->args[0] &&
2980 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2981 						    filter_mask,
2982 						    NLM_F_MULTI) < 0)
2983 				break;
2984 			idx++;
2985 		}
2986 	}
2987 	rcu_read_unlock();
2988 	cb->args[0] = idx;
2989 
2990 	return skb->len;
2991 }
2992 
2993 static inline size_t bridge_nlmsg_size(void)
2994 {
2995 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2996 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2997 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2998 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2999 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3000 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3001 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3002 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3003 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3004 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3005 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3006 }
3007 
3008 static int rtnl_bridge_notify(struct net_device *dev)
3009 {
3010 	struct net *net = dev_net(dev);
3011 	struct sk_buff *skb;
3012 	int err = -EOPNOTSUPP;
3013 
3014 	if (!dev->netdev_ops->ndo_bridge_getlink)
3015 		return 0;
3016 
3017 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3018 	if (!skb) {
3019 		err = -ENOMEM;
3020 		goto errout;
3021 	}
3022 
3023 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3024 	if (err < 0)
3025 		goto errout;
3026 
3027 	if (!skb->len)
3028 		goto errout;
3029 
3030 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3031 	return 0;
3032 errout:
3033 	WARN_ON(err == -EMSGSIZE);
3034 	kfree_skb(skb);
3035 	if (err)
3036 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3037 	return err;
3038 }
3039 
3040 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3041 {
3042 	struct net *net = sock_net(skb->sk);
3043 	struct ifinfomsg *ifm;
3044 	struct net_device *dev;
3045 	struct nlattr *br_spec, *attr = NULL;
3046 	int rem, err = -EOPNOTSUPP;
3047 	u16 flags = 0;
3048 	bool have_flags = false;
3049 
3050 	if (nlmsg_len(nlh) < sizeof(*ifm))
3051 		return -EINVAL;
3052 
3053 	ifm = nlmsg_data(nlh);
3054 	if (ifm->ifi_family != AF_BRIDGE)
3055 		return -EPFNOSUPPORT;
3056 
3057 	dev = __dev_get_by_index(net, ifm->ifi_index);
3058 	if (!dev) {
3059 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3060 		return -ENODEV;
3061 	}
3062 
3063 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3064 	if (br_spec) {
3065 		nla_for_each_nested(attr, br_spec, rem) {
3066 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3067 				if (nla_len(attr) < sizeof(flags))
3068 					return -EINVAL;
3069 
3070 				have_flags = true;
3071 				flags = nla_get_u16(attr);
3072 				break;
3073 			}
3074 		}
3075 	}
3076 
3077 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3078 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3079 
3080 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3081 			err = -EOPNOTSUPP;
3082 			goto out;
3083 		}
3084 
3085 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3086 		if (err)
3087 			goto out;
3088 
3089 		flags &= ~BRIDGE_FLAGS_MASTER;
3090 	}
3091 
3092 	if ((flags & BRIDGE_FLAGS_SELF)) {
3093 		if (!dev->netdev_ops->ndo_bridge_setlink)
3094 			err = -EOPNOTSUPP;
3095 		else
3096 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3097 								  flags);
3098 		if (!err) {
3099 			flags &= ~BRIDGE_FLAGS_SELF;
3100 
3101 			/* Generate event to notify upper layer of bridge
3102 			 * change
3103 			 */
3104 			err = rtnl_bridge_notify(dev);
3105 		}
3106 	}
3107 
3108 	if (have_flags)
3109 		memcpy(nla_data(attr), &flags, sizeof(flags));
3110 out:
3111 	return err;
3112 }
3113 
3114 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3115 {
3116 	struct net *net = sock_net(skb->sk);
3117 	struct ifinfomsg *ifm;
3118 	struct net_device *dev;
3119 	struct nlattr *br_spec, *attr = NULL;
3120 	int rem, err = -EOPNOTSUPP;
3121 	u16 flags = 0;
3122 	bool have_flags = false;
3123 
3124 	if (nlmsg_len(nlh) < sizeof(*ifm))
3125 		return -EINVAL;
3126 
3127 	ifm = nlmsg_data(nlh);
3128 	if (ifm->ifi_family != AF_BRIDGE)
3129 		return -EPFNOSUPPORT;
3130 
3131 	dev = __dev_get_by_index(net, ifm->ifi_index);
3132 	if (!dev) {
3133 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3134 		return -ENODEV;
3135 	}
3136 
3137 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3138 	if (br_spec) {
3139 		nla_for_each_nested(attr, br_spec, rem) {
3140 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3141 				if (nla_len(attr) < sizeof(flags))
3142 					return -EINVAL;
3143 
3144 				have_flags = true;
3145 				flags = nla_get_u16(attr);
3146 				break;
3147 			}
3148 		}
3149 	}
3150 
3151 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3152 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3153 
3154 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3155 			err = -EOPNOTSUPP;
3156 			goto out;
3157 		}
3158 
3159 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3160 		if (err)
3161 			goto out;
3162 
3163 		flags &= ~BRIDGE_FLAGS_MASTER;
3164 	}
3165 
3166 	if ((flags & BRIDGE_FLAGS_SELF)) {
3167 		if (!dev->netdev_ops->ndo_bridge_dellink)
3168 			err = -EOPNOTSUPP;
3169 		else
3170 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3171 								  flags);
3172 
3173 		if (!err) {
3174 			flags &= ~BRIDGE_FLAGS_SELF;
3175 
3176 			/* Generate event to notify upper layer of bridge
3177 			 * change
3178 			 */
3179 			err = rtnl_bridge_notify(dev);
3180 		}
3181 	}
3182 
3183 	if (have_flags)
3184 		memcpy(nla_data(attr), &flags, sizeof(flags));
3185 out:
3186 	return err;
3187 }
3188 
3189 /* Process one rtnetlink message. */
3190 
3191 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3192 {
3193 	struct net *net = sock_net(skb->sk);
3194 	rtnl_doit_func doit;
3195 	int sz_idx, kind;
3196 	int family;
3197 	int type;
3198 	int err;
3199 
3200 	type = nlh->nlmsg_type;
3201 	if (type > RTM_MAX)
3202 		return -EOPNOTSUPP;
3203 
3204 	type -= RTM_BASE;
3205 
3206 	/* All the messages must have at least 1 byte length */
3207 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3208 		return 0;
3209 
3210 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3211 	sz_idx = type>>2;
3212 	kind = type&3;
3213 
3214 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3215 		return -EPERM;
3216 
3217 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3218 		struct sock *rtnl;
3219 		rtnl_dumpit_func dumpit;
3220 		rtnl_calcit_func calcit;
3221 		u16 min_dump_alloc = 0;
3222 
3223 		dumpit = rtnl_get_dumpit(family, type);
3224 		if (dumpit == NULL)
3225 			return -EOPNOTSUPP;
3226 		calcit = rtnl_get_calcit(family, type);
3227 		if (calcit)
3228 			min_dump_alloc = calcit(skb, nlh);
3229 
3230 		__rtnl_unlock();
3231 		rtnl = net->rtnl;
3232 		{
3233 			struct netlink_dump_control c = {
3234 				.dump		= dumpit,
3235 				.min_dump_alloc	= min_dump_alloc,
3236 			};
3237 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3238 		}
3239 		rtnl_lock();
3240 		return err;
3241 	}
3242 
3243 	doit = rtnl_get_doit(family, type);
3244 	if (doit == NULL)
3245 		return -EOPNOTSUPP;
3246 
3247 	return doit(skb, nlh);
3248 }
3249 
3250 static void rtnetlink_rcv(struct sk_buff *skb)
3251 {
3252 	rtnl_lock();
3253 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3254 	rtnl_unlock();
3255 }
3256 
3257 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3258 {
3259 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3260 
3261 	switch (event) {
3262 	case NETDEV_UP:
3263 	case NETDEV_DOWN:
3264 	case NETDEV_PRE_UP:
3265 	case NETDEV_POST_INIT:
3266 	case NETDEV_REGISTER:
3267 	case NETDEV_CHANGE:
3268 	case NETDEV_PRE_TYPE_CHANGE:
3269 	case NETDEV_GOING_DOWN:
3270 	case NETDEV_UNREGISTER:
3271 	case NETDEV_UNREGISTER_FINAL:
3272 	case NETDEV_RELEASE:
3273 	case NETDEV_JOIN:
3274 	case NETDEV_BONDING_INFO:
3275 		break;
3276 	default:
3277 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3278 		break;
3279 	}
3280 	return NOTIFY_DONE;
3281 }
3282 
3283 static struct notifier_block rtnetlink_dev_notifier = {
3284 	.notifier_call	= rtnetlink_event,
3285 };
3286 
3287 
3288 static int __net_init rtnetlink_net_init(struct net *net)
3289 {
3290 	struct sock *sk;
3291 	struct netlink_kernel_cfg cfg = {
3292 		.groups		= RTNLGRP_MAX,
3293 		.input		= rtnetlink_rcv,
3294 		.cb_mutex	= &rtnl_mutex,
3295 		.flags		= NL_CFG_F_NONROOT_RECV,
3296 	};
3297 
3298 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3299 	if (!sk)
3300 		return -ENOMEM;
3301 	net->rtnl = sk;
3302 	return 0;
3303 }
3304 
3305 static void __net_exit rtnetlink_net_exit(struct net *net)
3306 {
3307 	netlink_kernel_release(net->rtnl);
3308 	net->rtnl = NULL;
3309 }
3310 
3311 static struct pernet_operations rtnetlink_net_ops = {
3312 	.init = rtnetlink_net_init,
3313 	.exit = rtnetlink_net_exit,
3314 };
3315 
3316 void __init rtnetlink_init(void)
3317 {
3318 	if (register_pernet_subsys(&rtnetlink_net_ops))
3319 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3320 
3321 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3322 
3323 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3324 		      rtnl_dump_ifinfo, rtnl_calcit);
3325 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3326 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3327 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3328 
3329 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3330 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3331 
3332 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3333 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3334 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3335 
3336 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3337 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3338 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3339 }
3340 
3341