xref: /openbmc/linux/net/core/rtnetlink.c (revision d37512a277dfb2cef8a578e25a3246f61399a55a)
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 		return size;
823 	} else
824 		return 0;
825 }
826 
827 static size_t rtnl_port_size(const struct net_device *dev,
828 			     u32 ext_filter_mask)
829 {
830 	size_t port_size = nla_total_size(4)		/* PORT_VF */
831 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
832 		+ nla_total_size(sizeof(struct ifla_port_vsi))
833 							/* PORT_VSI_TYPE */
834 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
835 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
836 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
837 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
838 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
839 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
840 		+ port_size;
841 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
842 		+ port_size;
843 
844 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
845 	    !(ext_filter_mask & RTEXT_FILTER_VF))
846 		return 0;
847 	if (dev_num_vf(dev->dev.parent))
848 		return port_self_size + vf_ports_size +
849 			vf_port_size * dev_num_vf(dev->dev.parent);
850 	else
851 		return port_self_size;
852 }
853 
854 static noinline size_t if_nlmsg_size(const struct net_device *dev,
855 				     u32 ext_filter_mask)
856 {
857 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
858 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
859 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
860 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
861 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
862 	       + nla_total_size(sizeof(struct rtnl_link_stats))
863 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
864 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
865 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
866 	       + nla_total_size(4) /* IFLA_TXQLEN */
867 	       + nla_total_size(4) /* IFLA_WEIGHT */
868 	       + nla_total_size(4) /* IFLA_MTU */
869 	       + nla_total_size(4) /* IFLA_LINK */
870 	       + nla_total_size(4) /* IFLA_MASTER */
871 	       + nla_total_size(1) /* IFLA_CARRIER */
872 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
873 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
874 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
875 	       + nla_total_size(1) /* IFLA_OPERSTATE */
876 	       + nla_total_size(1) /* IFLA_LINKMODE */
877 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
878 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
879 	       + nla_total_size(ext_filter_mask
880 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
881 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
882 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
883 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
884 	       + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
885 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
886 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
887 }
888 
889 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
890 {
891 	struct nlattr *vf_ports;
892 	struct nlattr *vf_port;
893 	int vf;
894 	int err;
895 
896 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
897 	if (!vf_ports)
898 		return -EMSGSIZE;
899 
900 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
901 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
902 		if (!vf_port)
903 			goto nla_put_failure;
904 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
905 			goto nla_put_failure;
906 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
907 		if (err == -EMSGSIZE)
908 			goto nla_put_failure;
909 		if (err) {
910 			nla_nest_cancel(skb, vf_port);
911 			continue;
912 		}
913 		nla_nest_end(skb, vf_port);
914 	}
915 
916 	nla_nest_end(skb, vf_ports);
917 
918 	return 0;
919 
920 nla_put_failure:
921 	nla_nest_cancel(skb, vf_ports);
922 	return -EMSGSIZE;
923 }
924 
925 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
926 {
927 	struct nlattr *port_self;
928 	int err;
929 
930 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
931 	if (!port_self)
932 		return -EMSGSIZE;
933 
934 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
935 	if (err) {
936 		nla_nest_cancel(skb, port_self);
937 		return (err == -EMSGSIZE) ? err : 0;
938 	}
939 
940 	nla_nest_end(skb, port_self);
941 
942 	return 0;
943 }
944 
945 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
946 			  u32 ext_filter_mask)
947 {
948 	int err;
949 
950 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
951 	    !(ext_filter_mask & RTEXT_FILTER_VF))
952 		return 0;
953 
954 	err = rtnl_port_self_fill(skb, dev);
955 	if (err)
956 		return err;
957 
958 	if (dev_num_vf(dev->dev.parent)) {
959 		err = rtnl_vf_ports_fill(skb, dev);
960 		if (err)
961 			return err;
962 	}
963 
964 	return 0;
965 }
966 
967 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
968 {
969 	int err;
970 	struct netdev_phys_item_id ppid;
971 
972 	err = dev_get_phys_port_id(dev, &ppid);
973 	if (err) {
974 		if (err == -EOPNOTSUPP)
975 			return 0;
976 		return err;
977 	}
978 
979 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
980 		return -EMSGSIZE;
981 
982 	return 0;
983 }
984 
985 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
986 {
987 	int err;
988 	struct netdev_phys_item_id psid;
989 
990 	err = netdev_switch_parent_id_get(dev, &psid);
991 	if (err) {
992 		if (err == -EOPNOTSUPP)
993 			return 0;
994 		return err;
995 	}
996 
997 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
998 		return -EMSGSIZE;
999 
1000 	return 0;
1001 }
1002 
1003 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1004 			    int type, u32 pid, u32 seq, u32 change,
1005 			    unsigned int flags, u32 ext_filter_mask)
1006 {
1007 	struct ifinfomsg *ifm;
1008 	struct nlmsghdr *nlh;
1009 	struct rtnl_link_stats64 temp;
1010 	const struct rtnl_link_stats64 *stats;
1011 	struct nlattr *attr, *af_spec;
1012 	struct rtnl_af_ops *af_ops;
1013 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1014 
1015 	ASSERT_RTNL();
1016 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1017 	if (nlh == NULL)
1018 		return -EMSGSIZE;
1019 
1020 	ifm = nlmsg_data(nlh);
1021 	ifm->ifi_family = AF_UNSPEC;
1022 	ifm->__ifi_pad = 0;
1023 	ifm->ifi_type = dev->type;
1024 	ifm->ifi_index = dev->ifindex;
1025 	ifm->ifi_flags = dev_get_flags(dev);
1026 	ifm->ifi_change = change;
1027 
1028 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1029 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1030 	    nla_put_u8(skb, IFLA_OPERSTATE,
1031 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1032 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1033 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1034 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1035 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1036 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1037 #ifdef CONFIG_RPS
1038 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1039 #endif
1040 	    (dev->ifindex != dev->iflink &&
1041 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1042 	    (upper_dev &&
1043 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1044 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1045 	    (dev->qdisc &&
1046 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1047 	    (dev->ifalias &&
1048 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1049 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1050 			atomic_read(&dev->carrier_changes)))
1051 		goto nla_put_failure;
1052 
1053 	if (1) {
1054 		struct rtnl_link_ifmap map = {
1055 			.mem_start   = dev->mem_start,
1056 			.mem_end     = dev->mem_end,
1057 			.base_addr   = dev->base_addr,
1058 			.irq         = dev->irq,
1059 			.dma         = dev->dma,
1060 			.port        = dev->if_port,
1061 		};
1062 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1063 			goto nla_put_failure;
1064 	}
1065 
1066 	if (dev->addr_len) {
1067 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1068 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1069 			goto nla_put_failure;
1070 	}
1071 
1072 	if (rtnl_phys_port_id_fill(skb, dev))
1073 		goto nla_put_failure;
1074 
1075 	if (rtnl_phys_switch_id_fill(skb, dev))
1076 		goto nla_put_failure;
1077 
1078 	attr = nla_reserve(skb, IFLA_STATS,
1079 			sizeof(struct rtnl_link_stats));
1080 	if (attr == NULL)
1081 		goto nla_put_failure;
1082 
1083 	stats = dev_get_stats(dev, &temp);
1084 	copy_rtnl_link_stats(nla_data(attr), stats);
1085 
1086 	attr = nla_reserve(skb, IFLA_STATS64,
1087 			sizeof(struct rtnl_link_stats64));
1088 	if (attr == NULL)
1089 		goto nla_put_failure;
1090 	copy_rtnl_link_stats64(nla_data(attr), stats);
1091 
1092 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1093 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1094 		goto nla_put_failure;
1095 
1096 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1097 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
1098 		int i;
1099 
1100 		struct nlattr *vfinfo, *vf;
1101 		int num_vfs = dev_num_vf(dev->dev.parent);
1102 
1103 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1104 		if (!vfinfo)
1105 			goto nla_put_failure;
1106 		for (i = 0; i < num_vfs; i++) {
1107 			struct ifla_vf_info ivi;
1108 			struct ifla_vf_mac vf_mac;
1109 			struct ifla_vf_vlan vf_vlan;
1110 			struct ifla_vf_rate vf_rate;
1111 			struct ifla_vf_tx_rate vf_tx_rate;
1112 			struct ifla_vf_spoofchk vf_spoofchk;
1113 			struct ifla_vf_link_state vf_linkstate;
1114 
1115 			/*
1116 			 * Not all SR-IOV capable drivers support the
1117 			 * spoofcheck query.  Preset to -1 so the user
1118 			 * space tool can detect that the driver didn't
1119 			 * report anything.
1120 			 */
1121 			ivi.spoofchk = -1;
1122 			memset(ivi.mac, 0, sizeof(ivi.mac));
1123 			/* The default value for VF link state is "auto"
1124 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
1125 			 */
1126 			ivi.linkstate = 0;
1127 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1128 				break;
1129 			vf_mac.vf =
1130 				vf_vlan.vf =
1131 				vf_rate.vf =
1132 				vf_tx_rate.vf =
1133 				vf_spoofchk.vf =
1134 				vf_linkstate.vf = ivi.vf;
1135 
1136 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1137 			vf_vlan.vlan = ivi.vlan;
1138 			vf_vlan.qos = ivi.qos;
1139 			vf_tx_rate.rate = ivi.max_tx_rate;
1140 			vf_rate.min_tx_rate = ivi.min_tx_rate;
1141 			vf_rate.max_tx_rate = ivi.max_tx_rate;
1142 			vf_spoofchk.setting = ivi.spoofchk;
1143 			vf_linkstate.link_state = ivi.linkstate;
1144 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1145 			if (!vf) {
1146 				nla_nest_cancel(skb, vfinfo);
1147 				goto nla_put_failure;
1148 			}
1149 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1150 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1151 			    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1152 				    &vf_rate) ||
1153 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1154 				    &vf_tx_rate) ||
1155 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1156 				    &vf_spoofchk) ||
1157 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1158 				    &vf_linkstate))
1159 				goto nla_put_failure;
1160 			nla_nest_end(skb, vf);
1161 		}
1162 		nla_nest_end(skb, vfinfo);
1163 	}
1164 
1165 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1166 		goto nla_put_failure;
1167 
1168 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1169 		if (rtnl_link_fill(skb, dev) < 0)
1170 			goto nla_put_failure;
1171 	}
1172 
1173 	if (dev->rtnl_link_ops &&
1174 	    dev->rtnl_link_ops->get_link_net) {
1175 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1176 
1177 		if (!net_eq(dev_net(dev), link_net)) {
1178 			int id = peernet2id(dev_net(dev), link_net);
1179 
1180 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1181 				goto nla_put_failure;
1182 		}
1183 	}
1184 
1185 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1186 		goto nla_put_failure;
1187 
1188 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1189 		if (af_ops->fill_link_af) {
1190 			struct nlattr *af;
1191 			int err;
1192 
1193 			if (!(af = nla_nest_start(skb, af_ops->family)))
1194 				goto nla_put_failure;
1195 
1196 			err = af_ops->fill_link_af(skb, dev);
1197 
1198 			/*
1199 			 * Caller may return ENODATA to indicate that there
1200 			 * was no data to be dumped. This is not an error, it
1201 			 * means we should trim the attribute header and
1202 			 * continue.
1203 			 */
1204 			if (err == -ENODATA)
1205 				nla_nest_cancel(skb, af);
1206 			else if (err < 0)
1207 				goto nla_put_failure;
1208 
1209 			nla_nest_end(skb, af);
1210 		}
1211 	}
1212 
1213 	nla_nest_end(skb, af_spec);
1214 
1215 	nlmsg_end(skb, nlh);
1216 	return 0;
1217 
1218 nla_put_failure:
1219 	nlmsg_cancel(skb, nlh);
1220 	return -EMSGSIZE;
1221 }
1222 
1223 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1224 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1225 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1226 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1227 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1228 	[IFLA_MTU]		= { .type = NLA_U32 },
1229 	[IFLA_LINK]		= { .type = NLA_U32 },
1230 	[IFLA_MASTER]		= { .type = NLA_U32 },
1231 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1232 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1233 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1234 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1235 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1236 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1237 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1238 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1239 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1240 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1241 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1242 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1243 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1244 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1245 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1246 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1247 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1248 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1249 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1250 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1251 };
1252 
1253 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1254 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1255 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1256 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1257 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1258 };
1259 
1260 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1261 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1262 };
1263 
1264 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1265 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1266 				    .len = sizeof(struct ifla_vf_mac) },
1267 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1268 				    .len = sizeof(struct ifla_vf_vlan) },
1269 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1270 				    .len = sizeof(struct ifla_vf_tx_rate) },
1271 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1272 				    .len = sizeof(struct ifla_vf_spoofchk) },
1273 	[IFLA_VF_RATE]		= { .type = NLA_BINARY,
1274 				    .len = sizeof(struct ifla_vf_rate) },
1275 	[IFLA_VF_LINK_STATE]	= { .type = NLA_BINARY,
1276 				    .len = sizeof(struct ifla_vf_link_state) },
1277 };
1278 
1279 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1280 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1281 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1282 				    .len = PORT_PROFILE_MAX },
1283 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1284 				    .len = sizeof(struct ifla_port_vsi)},
1285 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1286 				      .len = PORT_UUID_MAX },
1287 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1288 				    .len = PORT_UUID_MAX },
1289 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1290 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1291 };
1292 
1293 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1294 {
1295 	struct net *net = sock_net(skb->sk);
1296 	int h, s_h;
1297 	int idx = 0, s_idx;
1298 	struct net_device *dev;
1299 	struct hlist_head *head;
1300 	struct nlattr *tb[IFLA_MAX+1];
1301 	u32 ext_filter_mask = 0;
1302 	int err;
1303 	int hdrlen;
1304 
1305 	s_h = cb->args[0];
1306 	s_idx = cb->args[1];
1307 
1308 	rcu_read_lock();
1309 	cb->seq = net->dev_base_seq;
1310 
1311 	/* A hack to preserve kernel<->userspace interface.
1312 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1313 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1314 	 * what iproute2 < v3.9.0 used.
1315 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1316 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1317 	 */
1318 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1319 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1320 
1321 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1322 
1323 		if (tb[IFLA_EXT_MASK])
1324 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1325 	}
1326 
1327 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1328 		idx = 0;
1329 		head = &net->dev_index_head[h];
1330 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
1331 			if (idx < s_idx)
1332 				goto cont;
1333 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1334 					       NETLINK_CB(cb->skb).portid,
1335 					       cb->nlh->nlmsg_seq, 0,
1336 					       NLM_F_MULTI,
1337 					       ext_filter_mask);
1338 			/* If we ran out of room on the first message,
1339 			 * we're in trouble
1340 			 */
1341 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1342 
1343 			if (err < 0)
1344 				goto out;
1345 
1346 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1347 cont:
1348 			idx++;
1349 		}
1350 	}
1351 out:
1352 	rcu_read_unlock();
1353 	cb->args[1] = idx;
1354 	cb->args[0] = h;
1355 
1356 	return skb->len;
1357 }
1358 
1359 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1360 {
1361 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1362 }
1363 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1364 
1365 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1366 {
1367 	struct net *net;
1368 	/* Examine the link attributes and figure out which
1369 	 * network namespace we are talking about.
1370 	 */
1371 	if (tb[IFLA_NET_NS_PID])
1372 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1373 	else if (tb[IFLA_NET_NS_FD])
1374 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1375 	else
1376 		net = get_net(src_net);
1377 	return net;
1378 }
1379 EXPORT_SYMBOL(rtnl_link_get_net);
1380 
1381 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1382 {
1383 	if (dev) {
1384 		if (tb[IFLA_ADDRESS] &&
1385 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1386 			return -EINVAL;
1387 
1388 		if (tb[IFLA_BROADCAST] &&
1389 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1390 			return -EINVAL;
1391 	}
1392 
1393 	if (tb[IFLA_AF_SPEC]) {
1394 		struct nlattr *af;
1395 		int rem, err;
1396 
1397 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1398 			const struct rtnl_af_ops *af_ops;
1399 
1400 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1401 				return -EAFNOSUPPORT;
1402 
1403 			if (!af_ops->set_link_af)
1404 				return -EOPNOTSUPP;
1405 
1406 			if (af_ops->validate_link_af) {
1407 				err = af_ops->validate_link_af(dev, af);
1408 				if (err < 0)
1409 					return err;
1410 			}
1411 		}
1412 	}
1413 
1414 	return 0;
1415 }
1416 
1417 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1418 {
1419 	int rem, err = -EINVAL;
1420 	struct nlattr *vf;
1421 	const struct net_device_ops *ops = dev->netdev_ops;
1422 
1423 	nla_for_each_nested(vf, attr, rem) {
1424 		switch (nla_type(vf)) {
1425 		case IFLA_VF_MAC: {
1426 			struct ifla_vf_mac *ivm;
1427 			ivm = nla_data(vf);
1428 			err = -EOPNOTSUPP;
1429 			if (ops->ndo_set_vf_mac)
1430 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1431 							  ivm->mac);
1432 			break;
1433 		}
1434 		case IFLA_VF_VLAN: {
1435 			struct ifla_vf_vlan *ivv;
1436 			ivv = nla_data(vf);
1437 			err = -EOPNOTSUPP;
1438 			if (ops->ndo_set_vf_vlan)
1439 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1440 							   ivv->vlan,
1441 							   ivv->qos);
1442 			break;
1443 		}
1444 		case IFLA_VF_TX_RATE: {
1445 			struct ifla_vf_tx_rate *ivt;
1446 			struct ifla_vf_info ivf;
1447 			ivt = nla_data(vf);
1448 			err = -EOPNOTSUPP;
1449 			if (ops->ndo_get_vf_config)
1450 				err = ops->ndo_get_vf_config(dev, ivt->vf,
1451 							     &ivf);
1452 			if (err)
1453 				break;
1454 			err = -EOPNOTSUPP;
1455 			if (ops->ndo_set_vf_rate)
1456 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1457 							   ivf.min_tx_rate,
1458 							   ivt->rate);
1459 			break;
1460 		}
1461 		case IFLA_VF_RATE: {
1462 			struct ifla_vf_rate *ivt;
1463 			ivt = nla_data(vf);
1464 			err = -EOPNOTSUPP;
1465 			if (ops->ndo_set_vf_rate)
1466 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1467 							   ivt->min_tx_rate,
1468 							   ivt->max_tx_rate);
1469 			break;
1470 		}
1471 		case IFLA_VF_SPOOFCHK: {
1472 			struct ifla_vf_spoofchk *ivs;
1473 			ivs = nla_data(vf);
1474 			err = -EOPNOTSUPP;
1475 			if (ops->ndo_set_vf_spoofchk)
1476 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1477 							       ivs->setting);
1478 			break;
1479 		}
1480 		case IFLA_VF_LINK_STATE: {
1481 			struct ifla_vf_link_state *ivl;
1482 			ivl = nla_data(vf);
1483 			err = -EOPNOTSUPP;
1484 			if (ops->ndo_set_vf_link_state)
1485 				err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1486 								 ivl->link_state);
1487 			break;
1488 		}
1489 		default:
1490 			err = -EINVAL;
1491 			break;
1492 		}
1493 		if (err)
1494 			break;
1495 	}
1496 	return err;
1497 }
1498 
1499 static int do_set_master(struct net_device *dev, int ifindex)
1500 {
1501 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1502 	const struct net_device_ops *ops;
1503 	int err;
1504 
1505 	if (upper_dev) {
1506 		if (upper_dev->ifindex == ifindex)
1507 			return 0;
1508 		ops = upper_dev->netdev_ops;
1509 		if (ops->ndo_del_slave) {
1510 			err = ops->ndo_del_slave(upper_dev, dev);
1511 			if (err)
1512 				return err;
1513 		} else {
1514 			return -EOPNOTSUPP;
1515 		}
1516 	}
1517 
1518 	if (ifindex) {
1519 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1520 		if (!upper_dev)
1521 			return -EINVAL;
1522 		ops = upper_dev->netdev_ops;
1523 		if (ops->ndo_add_slave) {
1524 			err = ops->ndo_add_slave(upper_dev, dev);
1525 			if (err)
1526 				return err;
1527 		} else {
1528 			return -EOPNOTSUPP;
1529 		}
1530 	}
1531 	return 0;
1532 }
1533 
1534 #define DO_SETLINK_MODIFIED	0x01
1535 /* notify flag means notify + modified. */
1536 #define DO_SETLINK_NOTIFY	0x03
1537 static int do_setlink(const struct sk_buff *skb,
1538 		      struct net_device *dev, struct ifinfomsg *ifm,
1539 		      struct nlattr **tb, char *ifname, int status)
1540 {
1541 	const struct net_device_ops *ops = dev->netdev_ops;
1542 	int err;
1543 
1544 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1545 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1546 		if (IS_ERR(net)) {
1547 			err = PTR_ERR(net);
1548 			goto errout;
1549 		}
1550 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1551 			put_net(net);
1552 			err = -EPERM;
1553 			goto errout;
1554 		}
1555 		err = dev_change_net_namespace(dev, net, ifname);
1556 		put_net(net);
1557 		if (err)
1558 			goto errout;
1559 		status |= DO_SETLINK_MODIFIED;
1560 	}
1561 
1562 	if (tb[IFLA_MAP]) {
1563 		struct rtnl_link_ifmap *u_map;
1564 		struct ifmap k_map;
1565 
1566 		if (!ops->ndo_set_config) {
1567 			err = -EOPNOTSUPP;
1568 			goto errout;
1569 		}
1570 
1571 		if (!netif_device_present(dev)) {
1572 			err = -ENODEV;
1573 			goto errout;
1574 		}
1575 
1576 		u_map = nla_data(tb[IFLA_MAP]);
1577 		k_map.mem_start = (unsigned long) u_map->mem_start;
1578 		k_map.mem_end = (unsigned long) u_map->mem_end;
1579 		k_map.base_addr = (unsigned short) u_map->base_addr;
1580 		k_map.irq = (unsigned char) u_map->irq;
1581 		k_map.dma = (unsigned char) u_map->dma;
1582 		k_map.port = (unsigned char) u_map->port;
1583 
1584 		err = ops->ndo_set_config(dev, &k_map);
1585 		if (err < 0)
1586 			goto errout;
1587 
1588 		status |= DO_SETLINK_NOTIFY;
1589 	}
1590 
1591 	if (tb[IFLA_ADDRESS]) {
1592 		struct sockaddr *sa;
1593 		int len;
1594 
1595 		len = sizeof(sa_family_t) + dev->addr_len;
1596 		sa = kmalloc(len, GFP_KERNEL);
1597 		if (!sa) {
1598 			err = -ENOMEM;
1599 			goto errout;
1600 		}
1601 		sa->sa_family = dev->type;
1602 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1603 		       dev->addr_len);
1604 		err = dev_set_mac_address(dev, sa);
1605 		kfree(sa);
1606 		if (err)
1607 			goto errout;
1608 		status |= DO_SETLINK_MODIFIED;
1609 	}
1610 
1611 	if (tb[IFLA_MTU]) {
1612 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1613 		if (err < 0)
1614 			goto errout;
1615 		status |= DO_SETLINK_MODIFIED;
1616 	}
1617 
1618 	if (tb[IFLA_GROUP]) {
1619 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1620 		status |= DO_SETLINK_NOTIFY;
1621 	}
1622 
1623 	/*
1624 	 * Interface selected by interface index but interface
1625 	 * name provided implies that a name change has been
1626 	 * requested.
1627 	 */
1628 	if (ifm->ifi_index > 0 && ifname[0]) {
1629 		err = dev_change_name(dev, ifname);
1630 		if (err < 0)
1631 			goto errout;
1632 		status |= DO_SETLINK_MODIFIED;
1633 	}
1634 
1635 	if (tb[IFLA_IFALIAS]) {
1636 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1637 				    nla_len(tb[IFLA_IFALIAS]));
1638 		if (err < 0)
1639 			goto errout;
1640 		status |= DO_SETLINK_NOTIFY;
1641 	}
1642 
1643 	if (tb[IFLA_BROADCAST]) {
1644 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1645 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1646 	}
1647 
1648 	if (ifm->ifi_flags || ifm->ifi_change) {
1649 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1650 		if (err < 0)
1651 			goto errout;
1652 	}
1653 
1654 	if (tb[IFLA_MASTER]) {
1655 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1656 		if (err)
1657 			goto errout;
1658 		status |= DO_SETLINK_MODIFIED;
1659 	}
1660 
1661 	if (tb[IFLA_CARRIER]) {
1662 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1663 		if (err)
1664 			goto errout;
1665 		status |= DO_SETLINK_MODIFIED;
1666 	}
1667 
1668 	if (tb[IFLA_TXQLEN]) {
1669 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1670 
1671 		if (dev->tx_queue_len ^ value)
1672 			status |= DO_SETLINK_NOTIFY;
1673 
1674 		dev->tx_queue_len = value;
1675 	}
1676 
1677 	if (tb[IFLA_OPERSTATE])
1678 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1679 
1680 	if (tb[IFLA_LINKMODE]) {
1681 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1682 
1683 		write_lock_bh(&dev_base_lock);
1684 		if (dev->link_mode ^ value)
1685 			status |= DO_SETLINK_NOTIFY;
1686 		dev->link_mode = value;
1687 		write_unlock_bh(&dev_base_lock);
1688 	}
1689 
1690 	if (tb[IFLA_VFINFO_LIST]) {
1691 		struct nlattr *attr;
1692 		int rem;
1693 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1694 			if (nla_type(attr) != IFLA_VF_INFO) {
1695 				err = -EINVAL;
1696 				goto errout;
1697 			}
1698 			err = do_setvfinfo(dev, attr);
1699 			if (err < 0)
1700 				goto errout;
1701 			status |= DO_SETLINK_NOTIFY;
1702 		}
1703 	}
1704 	err = 0;
1705 
1706 	if (tb[IFLA_VF_PORTS]) {
1707 		struct nlattr *port[IFLA_PORT_MAX+1];
1708 		struct nlattr *attr;
1709 		int vf;
1710 		int rem;
1711 
1712 		err = -EOPNOTSUPP;
1713 		if (!ops->ndo_set_vf_port)
1714 			goto errout;
1715 
1716 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1717 			if (nla_type(attr) != IFLA_VF_PORT)
1718 				continue;
1719 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1720 				attr, ifla_port_policy);
1721 			if (err < 0)
1722 				goto errout;
1723 			if (!port[IFLA_PORT_VF]) {
1724 				err = -EOPNOTSUPP;
1725 				goto errout;
1726 			}
1727 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1728 			err = ops->ndo_set_vf_port(dev, vf, port);
1729 			if (err < 0)
1730 				goto errout;
1731 			status |= DO_SETLINK_NOTIFY;
1732 		}
1733 	}
1734 	err = 0;
1735 
1736 	if (tb[IFLA_PORT_SELF]) {
1737 		struct nlattr *port[IFLA_PORT_MAX+1];
1738 
1739 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1740 			tb[IFLA_PORT_SELF], ifla_port_policy);
1741 		if (err < 0)
1742 			goto errout;
1743 
1744 		err = -EOPNOTSUPP;
1745 		if (ops->ndo_set_vf_port)
1746 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1747 		if (err < 0)
1748 			goto errout;
1749 		status |= DO_SETLINK_NOTIFY;
1750 	}
1751 
1752 	if (tb[IFLA_AF_SPEC]) {
1753 		struct nlattr *af;
1754 		int rem;
1755 
1756 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1757 			const struct rtnl_af_ops *af_ops;
1758 
1759 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1760 				BUG();
1761 
1762 			err = af_ops->set_link_af(dev, af);
1763 			if (err < 0)
1764 				goto errout;
1765 
1766 			status |= DO_SETLINK_NOTIFY;
1767 		}
1768 	}
1769 	err = 0;
1770 
1771 errout:
1772 	if (status & DO_SETLINK_MODIFIED) {
1773 		if (status & DO_SETLINK_NOTIFY)
1774 			netdev_state_change(dev);
1775 
1776 		if (err < 0)
1777 			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",
1778 					     dev->name);
1779 	}
1780 
1781 	return err;
1782 }
1783 
1784 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1785 {
1786 	struct net *net = sock_net(skb->sk);
1787 	struct ifinfomsg *ifm;
1788 	struct net_device *dev;
1789 	int err;
1790 	struct nlattr *tb[IFLA_MAX+1];
1791 	char ifname[IFNAMSIZ];
1792 
1793 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1794 	if (err < 0)
1795 		goto errout;
1796 
1797 	if (tb[IFLA_IFNAME])
1798 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1799 	else
1800 		ifname[0] = '\0';
1801 
1802 	err = -EINVAL;
1803 	ifm = nlmsg_data(nlh);
1804 	if (ifm->ifi_index > 0)
1805 		dev = __dev_get_by_index(net, ifm->ifi_index);
1806 	else if (tb[IFLA_IFNAME])
1807 		dev = __dev_get_by_name(net, ifname);
1808 	else
1809 		goto errout;
1810 
1811 	if (dev == NULL) {
1812 		err = -ENODEV;
1813 		goto errout;
1814 	}
1815 
1816 	err = validate_linkmsg(dev, tb);
1817 	if (err < 0)
1818 		goto errout;
1819 
1820 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1821 errout:
1822 	return err;
1823 }
1824 
1825 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1826 {
1827 	struct net *net = sock_net(skb->sk);
1828 	const struct rtnl_link_ops *ops;
1829 	struct net_device *dev;
1830 	struct ifinfomsg *ifm;
1831 	char ifname[IFNAMSIZ];
1832 	struct nlattr *tb[IFLA_MAX+1];
1833 	int err;
1834 	LIST_HEAD(list_kill);
1835 
1836 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1837 	if (err < 0)
1838 		return err;
1839 
1840 	if (tb[IFLA_IFNAME])
1841 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1842 
1843 	ifm = nlmsg_data(nlh);
1844 	if (ifm->ifi_index > 0)
1845 		dev = __dev_get_by_index(net, ifm->ifi_index);
1846 	else if (tb[IFLA_IFNAME])
1847 		dev = __dev_get_by_name(net, ifname);
1848 	else
1849 		return -EINVAL;
1850 
1851 	if (!dev)
1852 		return -ENODEV;
1853 
1854 	ops = dev->rtnl_link_ops;
1855 	if (!ops || !ops->dellink)
1856 		return -EOPNOTSUPP;
1857 
1858 	ops->dellink(dev, &list_kill);
1859 	unregister_netdevice_many(&list_kill);
1860 	return 0;
1861 }
1862 
1863 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1864 {
1865 	unsigned int old_flags;
1866 	int err;
1867 
1868 	old_flags = dev->flags;
1869 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1870 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1871 		if (err < 0)
1872 			return err;
1873 	}
1874 
1875 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1876 
1877 	__dev_notify_flags(dev, old_flags, ~0U);
1878 	return 0;
1879 }
1880 EXPORT_SYMBOL(rtnl_configure_link);
1881 
1882 struct net_device *rtnl_create_link(struct net *net,
1883 	char *ifname, unsigned char name_assign_type,
1884 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
1885 {
1886 	int err;
1887 	struct net_device *dev;
1888 	unsigned int num_tx_queues = 1;
1889 	unsigned int num_rx_queues = 1;
1890 
1891 	if (tb[IFLA_NUM_TX_QUEUES])
1892 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1893 	else if (ops->get_num_tx_queues)
1894 		num_tx_queues = ops->get_num_tx_queues();
1895 
1896 	if (tb[IFLA_NUM_RX_QUEUES])
1897 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1898 	else if (ops->get_num_rx_queues)
1899 		num_rx_queues = ops->get_num_rx_queues();
1900 
1901 	err = -ENOMEM;
1902 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1903 			       ops->setup, num_tx_queues, num_rx_queues);
1904 	if (!dev)
1905 		goto err;
1906 
1907 	dev_net_set(dev, net);
1908 	dev->rtnl_link_ops = ops;
1909 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1910 
1911 	if (tb[IFLA_MTU])
1912 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1913 	if (tb[IFLA_ADDRESS]) {
1914 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1915 				nla_len(tb[IFLA_ADDRESS]));
1916 		dev->addr_assign_type = NET_ADDR_SET;
1917 	}
1918 	if (tb[IFLA_BROADCAST])
1919 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1920 				nla_len(tb[IFLA_BROADCAST]));
1921 	if (tb[IFLA_TXQLEN])
1922 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1923 	if (tb[IFLA_OPERSTATE])
1924 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1925 	if (tb[IFLA_LINKMODE])
1926 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1927 	if (tb[IFLA_GROUP])
1928 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1929 
1930 	return dev;
1931 
1932 err:
1933 	return ERR_PTR(err);
1934 }
1935 EXPORT_SYMBOL(rtnl_create_link);
1936 
1937 static int rtnl_group_changelink(const struct sk_buff *skb,
1938 		struct net *net, int group,
1939 		struct ifinfomsg *ifm,
1940 		struct nlattr **tb)
1941 {
1942 	struct net_device *dev;
1943 	int err;
1944 
1945 	for_each_netdev(net, dev) {
1946 		if (dev->group == group) {
1947 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1948 			if (err < 0)
1949 				return err;
1950 		}
1951 	}
1952 
1953 	return 0;
1954 }
1955 
1956 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1957 {
1958 	struct net *net = sock_net(skb->sk);
1959 	const struct rtnl_link_ops *ops;
1960 	const struct rtnl_link_ops *m_ops = NULL;
1961 	struct net_device *dev;
1962 	struct net_device *master_dev = NULL;
1963 	struct ifinfomsg *ifm;
1964 	char kind[MODULE_NAME_LEN];
1965 	char ifname[IFNAMSIZ];
1966 	struct nlattr *tb[IFLA_MAX+1];
1967 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1968 	unsigned char name_assign_type = NET_NAME_USER;
1969 	int err;
1970 
1971 #ifdef CONFIG_MODULES
1972 replay:
1973 #endif
1974 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1975 	if (err < 0)
1976 		return err;
1977 
1978 	if (tb[IFLA_IFNAME])
1979 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1980 	else
1981 		ifname[0] = '\0';
1982 
1983 	ifm = nlmsg_data(nlh);
1984 	if (ifm->ifi_index > 0)
1985 		dev = __dev_get_by_index(net, ifm->ifi_index);
1986 	else {
1987 		if (ifname[0])
1988 			dev = __dev_get_by_name(net, ifname);
1989 		else
1990 			dev = NULL;
1991 	}
1992 
1993 	if (dev) {
1994 		master_dev = netdev_master_upper_dev_get(dev);
1995 		if (master_dev)
1996 			m_ops = master_dev->rtnl_link_ops;
1997 	}
1998 
1999 	err = validate_linkmsg(dev, tb);
2000 	if (err < 0)
2001 		return err;
2002 
2003 	if (tb[IFLA_LINKINFO]) {
2004 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2005 				       tb[IFLA_LINKINFO], ifla_info_policy);
2006 		if (err < 0)
2007 			return err;
2008 	} else
2009 		memset(linkinfo, 0, sizeof(linkinfo));
2010 
2011 	if (linkinfo[IFLA_INFO_KIND]) {
2012 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2013 		ops = rtnl_link_ops_get(kind);
2014 	} else {
2015 		kind[0] = '\0';
2016 		ops = NULL;
2017 	}
2018 
2019 	if (1) {
2020 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
2021 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
2022 		struct nlattr **data = NULL;
2023 		struct nlattr **slave_data = NULL;
2024 		struct net *dest_net;
2025 
2026 		if (ops) {
2027 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2028 				err = nla_parse_nested(attr, ops->maxtype,
2029 						       linkinfo[IFLA_INFO_DATA],
2030 						       ops->policy);
2031 				if (err < 0)
2032 					return err;
2033 				data = attr;
2034 			}
2035 			if (ops->validate) {
2036 				err = ops->validate(tb, data);
2037 				if (err < 0)
2038 					return err;
2039 			}
2040 		}
2041 
2042 		if (m_ops) {
2043 			if (m_ops->slave_maxtype &&
2044 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2045 				err = nla_parse_nested(slave_attr,
2046 						       m_ops->slave_maxtype,
2047 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2048 						       m_ops->slave_policy);
2049 				if (err < 0)
2050 					return err;
2051 				slave_data = slave_attr;
2052 			}
2053 			if (m_ops->slave_validate) {
2054 				err = m_ops->slave_validate(tb, slave_data);
2055 				if (err < 0)
2056 					return err;
2057 			}
2058 		}
2059 
2060 		if (dev) {
2061 			int status = 0;
2062 
2063 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2064 				return -EEXIST;
2065 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2066 				return -EOPNOTSUPP;
2067 
2068 			if (linkinfo[IFLA_INFO_DATA]) {
2069 				if (!ops || ops != dev->rtnl_link_ops ||
2070 				    !ops->changelink)
2071 					return -EOPNOTSUPP;
2072 
2073 				err = ops->changelink(dev, tb, data);
2074 				if (err < 0)
2075 					return err;
2076 				status |= DO_SETLINK_NOTIFY;
2077 			}
2078 
2079 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2080 				if (!m_ops || !m_ops->slave_changelink)
2081 					return -EOPNOTSUPP;
2082 
2083 				err = m_ops->slave_changelink(master_dev, dev,
2084 							      tb, slave_data);
2085 				if (err < 0)
2086 					return err;
2087 				status |= DO_SETLINK_NOTIFY;
2088 			}
2089 
2090 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2091 		}
2092 
2093 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2094 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2095 				return rtnl_group_changelink(skb, net,
2096 						nla_get_u32(tb[IFLA_GROUP]),
2097 						ifm, tb);
2098 			return -ENODEV;
2099 		}
2100 
2101 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2102 			return -EOPNOTSUPP;
2103 
2104 		if (!ops) {
2105 #ifdef CONFIG_MODULES
2106 			if (kind[0]) {
2107 				__rtnl_unlock();
2108 				request_module("rtnl-link-%s", kind);
2109 				rtnl_lock();
2110 				ops = rtnl_link_ops_get(kind);
2111 				if (ops)
2112 					goto replay;
2113 			}
2114 #endif
2115 			return -EOPNOTSUPP;
2116 		}
2117 
2118 		if (!ops->setup)
2119 			return -EOPNOTSUPP;
2120 
2121 		if (!ifname[0]) {
2122 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2123 			name_assign_type = NET_NAME_ENUM;
2124 		}
2125 
2126 		dest_net = rtnl_link_get_net(net, tb);
2127 		if (IS_ERR(dest_net))
2128 			return PTR_ERR(dest_net);
2129 
2130 		dev = rtnl_create_link(dest_net, ifname, name_assign_type, ops, tb);
2131 		if (IS_ERR(dev)) {
2132 			err = PTR_ERR(dev);
2133 			goto out;
2134 		}
2135 
2136 		dev->ifindex = ifm->ifi_index;
2137 
2138 		if (ops->newlink) {
2139 			err = ops->newlink(net, dev, tb, data);
2140 			/* Drivers should call free_netdev() in ->destructor
2141 			 * and unregister it on failure after registration
2142 			 * so that device could be finally freed in rtnl_unlock.
2143 			 */
2144 			if (err < 0) {
2145 				/* If device is not registered at all, free it now */
2146 				if (dev->reg_state == NETREG_UNINITIALIZED)
2147 					free_netdev(dev);
2148 				goto out;
2149 			}
2150 		} else {
2151 			err = register_netdevice(dev);
2152 			if (err < 0) {
2153 				free_netdev(dev);
2154 				goto out;
2155 			}
2156 		}
2157 		err = rtnl_configure_link(dev, ifm);
2158 		if (err < 0)
2159 			unregister_netdevice(dev);
2160 out:
2161 		put_net(dest_net);
2162 		return err;
2163 	}
2164 }
2165 
2166 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2167 {
2168 	struct net *net = sock_net(skb->sk);
2169 	struct ifinfomsg *ifm;
2170 	char ifname[IFNAMSIZ];
2171 	struct nlattr *tb[IFLA_MAX+1];
2172 	struct net_device *dev = NULL;
2173 	struct sk_buff *nskb;
2174 	int err;
2175 	u32 ext_filter_mask = 0;
2176 
2177 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2178 	if (err < 0)
2179 		return err;
2180 
2181 	if (tb[IFLA_IFNAME])
2182 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2183 
2184 	if (tb[IFLA_EXT_MASK])
2185 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2186 
2187 	ifm = nlmsg_data(nlh);
2188 	if (ifm->ifi_index > 0)
2189 		dev = __dev_get_by_index(net, ifm->ifi_index);
2190 	else if (tb[IFLA_IFNAME])
2191 		dev = __dev_get_by_name(net, ifname);
2192 	else
2193 		return -EINVAL;
2194 
2195 	if (dev == NULL)
2196 		return -ENODEV;
2197 
2198 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2199 	if (nskb == NULL)
2200 		return -ENOBUFS;
2201 
2202 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2203 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2204 	if (err < 0) {
2205 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2206 		WARN_ON(err == -EMSGSIZE);
2207 		kfree_skb(nskb);
2208 	} else
2209 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2210 
2211 	return err;
2212 }
2213 
2214 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2215 {
2216 	struct net *net = sock_net(skb->sk);
2217 	struct net_device *dev;
2218 	struct nlattr *tb[IFLA_MAX+1];
2219 	u32 ext_filter_mask = 0;
2220 	u16 min_ifinfo_dump_size = 0;
2221 	int hdrlen;
2222 
2223 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2224 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2225 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2226 
2227 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2228 		if (tb[IFLA_EXT_MASK])
2229 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2230 	}
2231 
2232 	if (!ext_filter_mask)
2233 		return NLMSG_GOODSIZE;
2234 	/*
2235 	 * traverse the list of net devices and compute the minimum
2236 	 * buffer size based upon the filter mask.
2237 	 */
2238 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2239 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2240 					     if_nlmsg_size(dev,
2241 						           ext_filter_mask));
2242 	}
2243 
2244 	return min_ifinfo_dump_size;
2245 }
2246 
2247 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2248 {
2249 	int idx;
2250 	int s_idx = cb->family;
2251 
2252 	if (s_idx == 0)
2253 		s_idx = 1;
2254 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2255 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2256 		if (idx < s_idx || idx == PF_PACKET)
2257 			continue;
2258 		if (rtnl_msg_handlers[idx] == NULL ||
2259 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2260 			continue;
2261 		if (idx > s_idx) {
2262 			memset(&cb->args[0], 0, sizeof(cb->args));
2263 			cb->prev_seq = 0;
2264 			cb->seq = 0;
2265 		}
2266 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2267 			break;
2268 	}
2269 	cb->family = idx;
2270 
2271 	return skb->len;
2272 }
2273 
2274 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2275 				       unsigned int change, gfp_t flags)
2276 {
2277 	struct net *net = dev_net(dev);
2278 	struct sk_buff *skb;
2279 	int err = -ENOBUFS;
2280 	size_t if_info_size;
2281 
2282 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2283 	if (skb == NULL)
2284 		goto errout;
2285 
2286 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2287 	if (err < 0) {
2288 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2289 		WARN_ON(err == -EMSGSIZE);
2290 		kfree_skb(skb);
2291 		goto errout;
2292 	}
2293 	return skb;
2294 errout:
2295 	if (err < 0)
2296 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2297 	return NULL;
2298 }
2299 
2300 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2301 {
2302 	struct net *net = dev_net(dev);
2303 
2304 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2305 }
2306 
2307 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2308 		  gfp_t flags)
2309 {
2310 	struct sk_buff *skb;
2311 
2312 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2313 	if (skb)
2314 		rtmsg_ifinfo_send(skb, dev, flags);
2315 }
2316 EXPORT_SYMBOL(rtmsg_ifinfo);
2317 
2318 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2319 				   struct net_device *dev,
2320 				   u8 *addr, u32 pid, u32 seq,
2321 				   int type, unsigned int flags,
2322 				   int nlflags)
2323 {
2324 	struct nlmsghdr *nlh;
2325 	struct ndmsg *ndm;
2326 
2327 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2328 	if (!nlh)
2329 		return -EMSGSIZE;
2330 
2331 	ndm = nlmsg_data(nlh);
2332 	ndm->ndm_family  = AF_BRIDGE;
2333 	ndm->ndm_pad1	 = 0;
2334 	ndm->ndm_pad2    = 0;
2335 	ndm->ndm_flags	 = flags;
2336 	ndm->ndm_type	 = 0;
2337 	ndm->ndm_ifindex = dev->ifindex;
2338 	ndm->ndm_state   = NUD_PERMANENT;
2339 
2340 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2341 		goto nla_put_failure;
2342 
2343 	nlmsg_end(skb, nlh);
2344 	return 0;
2345 
2346 nla_put_failure:
2347 	nlmsg_cancel(skb, nlh);
2348 	return -EMSGSIZE;
2349 }
2350 
2351 static inline size_t rtnl_fdb_nlmsg_size(void)
2352 {
2353 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2354 }
2355 
2356 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2357 {
2358 	struct net *net = dev_net(dev);
2359 	struct sk_buff *skb;
2360 	int err = -ENOBUFS;
2361 
2362 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2363 	if (!skb)
2364 		goto errout;
2365 
2366 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2367 	if (err < 0) {
2368 		kfree_skb(skb);
2369 		goto errout;
2370 	}
2371 
2372 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2373 	return;
2374 errout:
2375 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2376 }
2377 
2378 /**
2379  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2380  */
2381 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2382 		     struct nlattr *tb[],
2383 		     struct net_device *dev,
2384 		     const unsigned char *addr, u16 vid,
2385 		     u16 flags)
2386 {
2387 	int err = -EINVAL;
2388 
2389 	/* If aging addresses are supported device will need to
2390 	 * implement its own handler for this.
2391 	 */
2392 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2393 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2394 		return err;
2395 	}
2396 
2397 	if (vid) {
2398 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2399 		return err;
2400 	}
2401 
2402 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2403 		err = dev_uc_add_excl(dev, addr);
2404 	else if (is_multicast_ether_addr(addr))
2405 		err = dev_mc_add_excl(dev, addr);
2406 
2407 	/* Only return duplicate errors if NLM_F_EXCL is set */
2408 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2409 		err = 0;
2410 
2411 	return err;
2412 }
2413 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2414 
2415 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2416 {
2417 	u16 vid = 0;
2418 
2419 	if (vlan_attr) {
2420 		if (nla_len(vlan_attr) != sizeof(u16)) {
2421 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2422 			return -EINVAL;
2423 		}
2424 
2425 		vid = nla_get_u16(vlan_attr);
2426 
2427 		if (!vid || vid >= VLAN_VID_MASK) {
2428 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2429 				vid);
2430 			return -EINVAL;
2431 		}
2432 	}
2433 	*p_vid = vid;
2434 	return 0;
2435 }
2436 
2437 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2438 {
2439 	struct net *net = sock_net(skb->sk);
2440 	struct ndmsg *ndm;
2441 	struct nlattr *tb[NDA_MAX+1];
2442 	struct net_device *dev;
2443 	u8 *addr;
2444 	u16 vid;
2445 	int err;
2446 
2447 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2448 	if (err < 0)
2449 		return err;
2450 
2451 	ndm = nlmsg_data(nlh);
2452 	if (ndm->ndm_ifindex == 0) {
2453 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2454 		return -EINVAL;
2455 	}
2456 
2457 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2458 	if (dev == NULL) {
2459 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2460 		return -ENODEV;
2461 	}
2462 
2463 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2464 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2465 		return -EINVAL;
2466 	}
2467 
2468 	addr = nla_data(tb[NDA_LLADDR]);
2469 
2470 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2471 	if (err)
2472 		return err;
2473 
2474 	err = -EOPNOTSUPP;
2475 
2476 	/* Support fdb on master device the net/bridge default case */
2477 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2478 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2479 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2480 		const struct net_device_ops *ops = br_dev->netdev_ops;
2481 
2482 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2483 				       nlh->nlmsg_flags);
2484 		if (err)
2485 			goto out;
2486 		else
2487 			ndm->ndm_flags &= ~NTF_MASTER;
2488 	}
2489 
2490 	/* Embedded bridge, macvlan, and any other device support */
2491 	if ((ndm->ndm_flags & NTF_SELF)) {
2492 		if (dev->netdev_ops->ndo_fdb_add)
2493 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2494 							   vid,
2495 							   nlh->nlmsg_flags);
2496 		else
2497 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2498 					       nlh->nlmsg_flags);
2499 
2500 		if (!err) {
2501 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2502 			ndm->ndm_flags &= ~NTF_SELF;
2503 		}
2504 	}
2505 out:
2506 	return err;
2507 }
2508 
2509 /**
2510  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2511  */
2512 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2513 		     struct nlattr *tb[],
2514 		     struct net_device *dev,
2515 		     const unsigned char *addr, u16 vid)
2516 {
2517 	int err = -EINVAL;
2518 
2519 	/* If aging addresses are supported device will need to
2520 	 * implement its own handler for this.
2521 	 */
2522 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2523 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2524 		return err;
2525 	}
2526 
2527 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2528 		err = dev_uc_del(dev, addr);
2529 	else if (is_multicast_ether_addr(addr))
2530 		err = dev_mc_del(dev, addr);
2531 
2532 	return err;
2533 }
2534 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2535 
2536 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2537 {
2538 	struct net *net = sock_net(skb->sk);
2539 	struct ndmsg *ndm;
2540 	struct nlattr *tb[NDA_MAX+1];
2541 	struct net_device *dev;
2542 	int err = -EINVAL;
2543 	__u8 *addr;
2544 	u16 vid;
2545 
2546 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2547 		return -EPERM;
2548 
2549 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2550 	if (err < 0)
2551 		return err;
2552 
2553 	ndm = nlmsg_data(nlh);
2554 	if (ndm->ndm_ifindex == 0) {
2555 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2556 		return -EINVAL;
2557 	}
2558 
2559 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2560 	if (dev == NULL) {
2561 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2562 		return -ENODEV;
2563 	}
2564 
2565 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2566 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2567 		return -EINVAL;
2568 	}
2569 
2570 	addr = nla_data(tb[NDA_LLADDR]);
2571 
2572 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2573 	if (err)
2574 		return err;
2575 
2576 	err = -EOPNOTSUPP;
2577 
2578 	/* Support fdb on master device the net/bridge default case */
2579 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2580 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2581 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2582 		const struct net_device_ops *ops = br_dev->netdev_ops;
2583 
2584 		if (ops->ndo_fdb_del)
2585 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2586 
2587 		if (err)
2588 			goto out;
2589 		else
2590 			ndm->ndm_flags &= ~NTF_MASTER;
2591 	}
2592 
2593 	/* Embedded bridge, macvlan, and any other device support */
2594 	if (ndm->ndm_flags & NTF_SELF) {
2595 		if (dev->netdev_ops->ndo_fdb_del)
2596 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2597 							   vid);
2598 		else
2599 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2600 
2601 		if (!err) {
2602 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2603 			ndm->ndm_flags &= ~NTF_SELF;
2604 		}
2605 	}
2606 out:
2607 	return err;
2608 }
2609 
2610 static int nlmsg_populate_fdb(struct sk_buff *skb,
2611 			      struct netlink_callback *cb,
2612 			      struct net_device *dev,
2613 			      int *idx,
2614 			      struct netdev_hw_addr_list *list)
2615 {
2616 	struct netdev_hw_addr *ha;
2617 	int err;
2618 	u32 portid, seq;
2619 
2620 	portid = NETLINK_CB(cb->skb).portid;
2621 	seq = cb->nlh->nlmsg_seq;
2622 
2623 	list_for_each_entry(ha, &list->list, list) {
2624 		if (*idx < cb->args[0])
2625 			goto skip;
2626 
2627 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2628 					      portid, seq,
2629 					      RTM_NEWNEIGH, NTF_SELF,
2630 					      NLM_F_MULTI);
2631 		if (err < 0)
2632 			return err;
2633 skip:
2634 		*idx += 1;
2635 	}
2636 	return 0;
2637 }
2638 
2639 /**
2640  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2641  * @nlh: netlink message header
2642  * @dev: netdevice
2643  *
2644  * Default netdevice operation to dump the existing unicast address list.
2645  * Returns number of addresses from list put in skb.
2646  */
2647 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2648 		      struct netlink_callback *cb,
2649 		      struct net_device *dev,
2650 		      struct net_device *filter_dev,
2651 		      int idx)
2652 {
2653 	int err;
2654 
2655 	netif_addr_lock_bh(dev);
2656 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2657 	if (err)
2658 		goto out;
2659 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2660 out:
2661 	netif_addr_unlock_bh(dev);
2662 	return idx;
2663 }
2664 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2665 
2666 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2667 {
2668 	struct net_device *dev;
2669 	struct nlattr *tb[IFLA_MAX+1];
2670 	struct net_device *bdev = NULL;
2671 	struct net_device *br_dev = NULL;
2672 	const struct net_device_ops *ops = NULL;
2673 	const struct net_device_ops *cops = NULL;
2674 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2675 	struct net *net = sock_net(skb->sk);
2676 	int brport_idx = 0;
2677 	int br_idx = 0;
2678 	int idx = 0;
2679 
2680 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2681 			ifla_policy) == 0) {
2682 		if (tb[IFLA_MASTER])
2683 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
2684 	}
2685 
2686 	brport_idx = ifm->ifi_index;
2687 
2688 	if (br_idx) {
2689 		br_dev = __dev_get_by_index(net, br_idx);
2690 		if (!br_dev)
2691 			return -ENODEV;
2692 
2693 		ops = br_dev->netdev_ops;
2694 		bdev = br_dev;
2695 	}
2696 
2697 	for_each_netdev(net, dev) {
2698 		if (brport_idx && (dev->ifindex != brport_idx))
2699 			continue;
2700 
2701 		if (!br_idx) { /* user did not specify a specific bridge */
2702 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
2703 				br_dev = netdev_master_upper_dev_get(dev);
2704 				cops = br_dev->netdev_ops;
2705 			}
2706 
2707 			bdev = dev;
2708 		} else {
2709 			if (dev != br_dev &&
2710 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
2711 				continue;
2712 
2713 			if (br_dev != netdev_master_upper_dev_get(dev) &&
2714 			    !(dev->priv_flags & IFF_EBRIDGE))
2715 				continue;
2716 
2717 			bdev = br_dev;
2718 			cops = ops;
2719 		}
2720 
2721 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2722 			if (cops && cops->ndo_fdb_dump)
2723 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2724 							 idx);
2725 		}
2726 
2727 		if (dev->netdev_ops->ndo_fdb_dump)
2728 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2729 							    idx);
2730 		else
2731 			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2732 
2733 		cops = NULL;
2734 	}
2735 
2736 	cb->args[0] = idx;
2737 	return skb->len;
2738 }
2739 
2740 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2741 			       unsigned int attrnum, unsigned int flag)
2742 {
2743 	if (mask & flag)
2744 		return nla_put_u8(skb, attrnum, !!(flags & flag));
2745 	return 0;
2746 }
2747 
2748 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2749 			    struct net_device *dev, u16 mode,
2750 			    u32 flags, u32 mask)
2751 {
2752 	struct nlmsghdr *nlh;
2753 	struct ifinfomsg *ifm;
2754 	struct nlattr *br_afspec;
2755 	struct nlattr *protinfo;
2756 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2757 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2758 
2759 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2760 	if (nlh == NULL)
2761 		return -EMSGSIZE;
2762 
2763 	ifm = nlmsg_data(nlh);
2764 	ifm->ifi_family = AF_BRIDGE;
2765 	ifm->__ifi_pad = 0;
2766 	ifm->ifi_type = dev->type;
2767 	ifm->ifi_index = dev->ifindex;
2768 	ifm->ifi_flags = dev_get_flags(dev);
2769 	ifm->ifi_change = 0;
2770 
2771 
2772 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2773 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2774 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2775 	    (br_dev &&
2776 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2777 	    (dev->addr_len &&
2778 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2779 	    (dev->ifindex != dev->iflink &&
2780 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2781 		goto nla_put_failure;
2782 
2783 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2784 	if (!br_afspec)
2785 		goto nla_put_failure;
2786 
2787 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2788 		nla_nest_cancel(skb, br_afspec);
2789 		goto nla_put_failure;
2790 	}
2791 
2792 	if (mode != BRIDGE_MODE_UNDEF) {
2793 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2794 			nla_nest_cancel(skb, br_afspec);
2795 			goto nla_put_failure;
2796 		}
2797 	}
2798 	nla_nest_end(skb, br_afspec);
2799 
2800 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2801 	if (!protinfo)
2802 		goto nla_put_failure;
2803 
2804 	if (brport_nla_put_flag(skb, flags, mask,
2805 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2806 	    brport_nla_put_flag(skb, flags, mask,
2807 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2808 	    brport_nla_put_flag(skb, flags, mask,
2809 				IFLA_BRPORT_FAST_LEAVE,
2810 				BR_MULTICAST_FAST_LEAVE) ||
2811 	    brport_nla_put_flag(skb, flags, mask,
2812 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2813 	    brport_nla_put_flag(skb, flags, mask,
2814 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2815 	    brport_nla_put_flag(skb, flags, mask,
2816 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2817 	    brport_nla_put_flag(skb, flags, mask,
2818 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2819 	    brport_nla_put_flag(skb, flags, mask,
2820 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2821 		nla_nest_cancel(skb, protinfo);
2822 		goto nla_put_failure;
2823 	}
2824 
2825 	nla_nest_end(skb, protinfo);
2826 
2827 	nlmsg_end(skb, nlh);
2828 	return 0;
2829 nla_put_failure:
2830 	nlmsg_cancel(skb, nlh);
2831 	return -EMSGSIZE;
2832 }
2833 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2834 
2835 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2836 {
2837 	struct net *net = sock_net(skb->sk);
2838 	struct net_device *dev;
2839 	int idx = 0;
2840 	u32 portid = NETLINK_CB(cb->skb).portid;
2841 	u32 seq = cb->nlh->nlmsg_seq;
2842 	u32 filter_mask = 0;
2843 
2844 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2845 		struct nlattr *extfilt;
2846 
2847 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2848 					  IFLA_EXT_MASK);
2849 		if (extfilt) {
2850 			if (nla_len(extfilt) < sizeof(filter_mask))
2851 				return -EINVAL;
2852 
2853 			filter_mask = nla_get_u32(extfilt);
2854 		}
2855 	}
2856 
2857 	rcu_read_lock();
2858 	for_each_netdev_rcu(net, dev) {
2859 		const struct net_device_ops *ops = dev->netdev_ops;
2860 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2861 
2862 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2863 			if (idx >= cb->args[0] &&
2864 			    br_dev->netdev_ops->ndo_bridge_getlink(
2865 				    skb, portid, seq, dev, filter_mask) < 0)
2866 				break;
2867 			idx++;
2868 		}
2869 
2870 		if (ops->ndo_bridge_getlink) {
2871 			if (idx >= cb->args[0] &&
2872 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2873 						    filter_mask) < 0)
2874 				break;
2875 			idx++;
2876 		}
2877 	}
2878 	rcu_read_unlock();
2879 	cb->args[0] = idx;
2880 
2881 	return skb->len;
2882 }
2883 
2884 static inline size_t bridge_nlmsg_size(void)
2885 {
2886 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2887 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2888 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2889 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2890 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2891 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2892 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2893 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2894 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2895 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2896 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2897 }
2898 
2899 static int rtnl_bridge_notify(struct net_device *dev)
2900 {
2901 	struct net *net = dev_net(dev);
2902 	struct sk_buff *skb;
2903 	int err = -EOPNOTSUPP;
2904 
2905 	if (!dev->netdev_ops->ndo_bridge_getlink)
2906 		return 0;
2907 
2908 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2909 	if (!skb) {
2910 		err = -ENOMEM;
2911 		goto errout;
2912 	}
2913 
2914 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2915 	if (err < 0)
2916 		goto errout;
2917 
2918 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2919 	return 0;
2920 errout:
2921 	WARN_ON(err == -EMSGSIZE);
2922 	kfree_skb(skb);
2923 	rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2924 	return err;
2925 }
2926 
2927 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2928 {
2929 	struct net *net = sock_net(skb->sk);
2930 	struct ifinfomsg *ifm;
2931 	struct net_device *dev;
2932 	struct nlattr *br_spec, *attr = NULL;
2933 	int rem, err = -EOPNOTSUPP;
2934 	u16 flags = 0;
2935 	bool have_flags = false;
2936 
2937 	if (nlmsg_len(nlh) < sizeof(*ifm))
2938 		return -EINVAL;
2939 
2940 	ifm = nlmsg_data(nlh);
2941 	if (ifm->ifi_family != AF_BRIDGE)
2942 		return -EPFNOSUPPORT;
2943 
2944 	dev = __dev_get_by_index(net, ifm->ifi_index);
2945 	if (!dev) {
2946 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2947 		return -ENODEV;
2948 	}
2949 
2950 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2951 	if (br_spec) {
2952 		nla_for_each_nested(attr, br_spec, rem) {
2953 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2954 				if (nla_len(attr) < sizeof(flags))
2955 					return -EINVAL;
2956 
2957 				have_flags = true;
2958 				flags = nla_get_u16(attr);
2959 				break;
2960 			}
2961 		}
2962 	}
2963 
2964 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2965 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2966 
2967 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2968 			err = -EOPNOTSUPP;
2969 			goto out;
2970 		}
2971 
2972 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2973 		if (err)
2974 			goto out;
2975 
2976 		flags &= ~BRIDGE_FLAGS_MASTER;
2977 	}
2978 
2979 	if ((flags & BRIDGE_FLAGS_SELF)) {
2980 		if (!dev->netdev_ops->ndo_bridge_setlink)
2981 			err = -EOPNOTSUPP;
2982 		else
2983 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2984 		if (!err) {
2985 			flags &= ~BRIDGE_FLAGS_SELF;
2986 
2987 			/* Generate event to notify upper layer of bridge
2988 			 * change
2989 			 */
2990 			err = rtnl_bridge_notify(dev);
2991 		}
2992 	}
2993 
2994 	if (have_flags)
2995 		memcpy(nla_data(attr), &flags, sizeof(flags));
2996 out:
2997 	return err;
2998 }
2999 
3000 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3001 {
3002 	struct net *net = sock_net(skb->sk);
3003 	struct ifinfomsg *ifm;
3004 	struct net_device *dev;
3005 	struct nlattr *br_spec, *attr = NULL;
3006 	int rem, err = -EOPNOTSUPP;
3007 	u16 flags = 0;
3008 	bool have_flags = false;
3009 
3010 	if (nlmsg_len(nlh) < sizeof(*ifm))
3011 		return -EINVAL;
3012 
3013 	ifm = nlmsg_data(nlh);
3014 	if (ifm->ifi_family != AF_BRIDGE)
3015 		return -EPFNOSUPPORT;
3016 
3017 	dev = __dev_get_by_index(net, ifm->ifi_index);
3018 	if (!dev) {
3019 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3020 		return -ENODEV;
3021 	}
3022 
3023 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3024 	if (br_spec) {
3025 		nla_for_each_nested(attr, br_spec, rem) {
3026 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3027 				if (nla_len(attr) < sizeof(flags))
3028 					return -EINVAL;
3029 
3030 				have_flags = true;
3031 				flags = nla_get_u16(attr);
3032 				break;
3033 			}
3034 		}
3035 	}
3036 
3037 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3038 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3039 
3040 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3041 			err = -EOPNOTSUPP;
3042 			goto out;
3043 		}
3044 
3045 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
3046 		if (err)
3047 			goto out;
3048 
3049 		flags &= ~BRIDGE_FLAGS_MASTER;
3050 	}
3051 
3052 	if ((flags & BRIDGE_FLAGS_SELF)) {
3053 		if (!dev->netdev_ops->ndo_bridge_dellink)
3054 			err = -EOPNOTSUPP;
3055 		else
3056 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
3057 
3058 		if (!err) {
3059 			flags &= ~BRIDGE_FLAGS_SELF;
3060 
3061 			/* Generate event to notify upper layer of bridge
3062 			 * change
3063 			 */
3064 			err = rtnl_bridge_notify(dev);
3065 		}
3066 	}
3067 
3068 	if (have_flags)
3069 		memcpy(nla_data(attr), &flags, sizeof(flags));
3070 out:
3071 	return err;
3072 }
3073 
3074 /* Process one rtnetlink message. */
3075 
3076 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3077 {
3078 	struct net *net = sock_net(skb->sk);
3079 	rtnl_doit_func doit;
3080 	int sz_idx, kind;
3081 	int family;
3082 	int type;
3083 	int err;
3084 
3085 	type = nlh->nlmsg_type;
3086 	if (type > RTM_MAX)
3087 		return -EOPNOTSUPP;
3088 
3089 	type -= RTM_BASE;
3090 
3091 	/* All the messages must have at least 1 byte length */
3092 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3093 		return 0;
3094 
3095 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3096 	sz_idx = type>>2;
3097 	kind = type&3;
3098 
3099 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3100 		return -EPERM;
3101 
3102 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3103 		struct sock *rtnl;
3104 		rtnl_dumpit_func dumpit;
3105 		rtnl_calcit_func calcit;
3106 		u16 min_dump_alloc = 0;
3107 
3108 		dumpit = rtnl_get_dumpit(family, type);
3109 		if (dumpit == NULL)
3110 			return -EOPNOTSUPP;
3111 		calcit = rtnl_get_calcit(family, type);
3112 		if (calcit)
3113 			min_dump_alloc = calcit(skb, nlh);
3114 
3115 		__rtnl_unlock();
3116 		rtnl = net->rtnl;
3117 		{
3118 			struct netlink_dump_control c = {
3119 				.dump		= dumpit,
3120 				.min_dump_alloc	= min_dump_alloc,
3121 			};
3122 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3123 		}
3124 		rtnl_lock();
3125 		return err;
3126 	}
3127 
3128 	doit = rtnl_get_doit(family, type);
3129 	if (doit == NULL)
3130 		return -EOPNOTSUPP;
3131 
3132 	return doit(skb, nlh);
3133 }
3134 
3135 static void rtnetlink_rcv(struct sk_buff *skb)
3136 {
3137 	rtnl_lock();
3138 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3139 	rtnl_unlock();
3140 }
3141 
3142 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3143 {
3144 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3145 
3146 	switch (event) {
3147 	case NETDEV_UP:
3148 	case NETDEV_DOWN:
3149 	case NETDEV_PRE_UP:
3150 	case NETDEV_POST_INIT:
3151 	case NETDEV_REGISTER:
3152 	case NETDEV_CHANGE:
3153 	case NETDEV_PRE_TYPE_CHANGE:
3154 	case NETDEV_GOING_DOWN:
3155 	case NETDEV_UNREGISTER:
3156 	case NETDEV_UNREGISTER_FINAL:
3157 	case NETDEV_RELEASE:
3158 	case NETDEV_JOIN:
3159 		break;
3160 	default:
3161 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3162 		break;
3163 	}
3164 	return NOTIFY_DONE;
3165 }
3166 
3167 static struct notifier_block rtnetlink_dev_notifier = {
3168 	.notifier_call	= rtnetlink_event,
3169 };
3170 
3171 
3172 static int __net_init rtnetlink_net_init(struct net *net)
3173 {
3174 	struct sock *sk;
3175 	struct netlink_kernel_cfg cfg = {
3176 		.groups		= RTNLGRP_MAX,
3177 		.input		= rtnetlink_rcv,
3178 		.cb_mutex	= &rtnl_mutex,
3179 		.flags		= NL_CFG_F_NONROOT_RECV,
3180 	};
3181 
3182 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3183 	if (!sk)
3184 		return -ENOMEM;
3185 	net->rtnl = sk;
3186 	return 0;
3187 }
3188 
3189 static void __net_exit rtnetlink_net_exit(struct net *net)
3190 {
3191 	netlink_kernel_release(net->rtnl);
3192 	net->rtnl = NULL;
3193 }
3194 
3195 static struct pernet_operations rtnetlink_net_ops = {
3196 	.init = rtnetlink_net_init,
3197 	.exit = rtnetlink_net_exit,
3198 };
3199 
3200 void __init rtnetlink_init(void)
3201 {
3202 	if (register_pernet_subsys(&rtnetlink_net_ops))
3203 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3204 
3205 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3206 
3207 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3208 		      rtnl_dump_ifinfo, rtnl_calcit);
3209 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3210 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3211 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3212 
3213 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3214 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3215 
3216 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3217 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3218 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3219 
3220 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3221 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3222 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3223 }
3224 
3225