xref: /openbmc/linux/net/core/rtnetlink.c (revision 66400d543072d6ff6f8df036e3f12f5c154c19cc)
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_port_name_fill(struct sk_buff *skb, struct net_device *dev)
986 {
987 	char name[IFNAMSIZ];
988 	int err;
989 
990 	err = dev_get_phys_port_name(dev, name, sizeof(name));
991 	if (err) {
992 		if (err == -EOPNOTSUPP)
993 			return 0;
994 		return err;
995 	}
996 
997 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
998 		return -EMSGSIZE;
999 
1000 	return 0;
1001 }
1002 
1003 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1004 {
1005 	int err;
1006 	struct netdev_phys_item_id psid;
1007 
1008 	err = netdev_switch_parent_id_get(dev, &psid);
1009 	if (err) {
1010 		if (err == -EOPNOTSUPP)
1011 			return 0;
1012 		return err;
1013 	}
1014 
1015 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
1016 		return -EMSGSIZE;
1017 
1018 	return 0;
1019 }
1020 
1021 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1022 			    int type, u32 pid, u32 seq, u32 change,
1023 			    unsigned int flags, u32 ext_filter_mask)
1024 {
1025 	struct ifinfomsg *ifm;
1026 	struct nlmsghdr *nlh;
1027 	struct rtnl_link_stats64 temp;
1028 	const struct rtnl_link_stats64 *stats;
1029 	struct nlattr *attr, *af_spec;
1030 	struct rtnl_af_ops *af_ops;
1031 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1032 
1033 	ASSERT_RTNL();
1034 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1035 	if (nlh == NULL)
1036 		return -EMSGSIZE;
1037 
1038 	ifm = nlmsg_data(nlh);
1039 	ifm->ifi_family = AF_UNSPEC;
1040 	ifm->__ifi_pad = 0;
1041 	ifm->ifi_type = dev->type;
1042 	ifm->ifi_index = dev->ifindex;
1043 	ifm->ifi_flags = dev_get_flags(dev);
1044 	ifm->ifi_change = change;
1045 
1046 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1047 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1048 	    nla_put_u8(skb, IFLA_OPERSTATE,
1049 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1050 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1051 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1052 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1053 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1054 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1055 #ifdef CONFIG_RPS
1056 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1057 #endif
1058 	    (dev->ifindex != dev->iflink &&
1059 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1060 	    (upper_dev &&
1061 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1062 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1063 	    (dev->qdisc &&
1064 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1065 	    (dev->ifalias &&
1066 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1067 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1068 			atomic_read(&dev->carrier_changes)))
1069 		goto nla_put_failure;
1070 
1071 	if (1) {
1072 		struct rtnl_link_ifmap map = {
1073 			.mem_start   = dev->mem_start,
1074 			.mem_end     = dev->mem_end,
1075 			.base_addr   = dev->base_addr,
1076 			.irq         = dev->irq,
1077 			.dma         = dev->dma,
1078 			.port        = dev->if_port,
1079 		};
1080 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1081 			goto nla_put_failure;
1082 	}
1083 
1084 	if (dev->addr_len) {
1085 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1086 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1087 			goto nla_put_failure;
1088 	}
1089 
1090 	if (rtnl_phys_port_id_fill(skb, dev))
1091 		goto nla_put_failure;
1092 
1093 	if (rtnl_phys_port_name_fill(skb, dev))
1094 		goto nla_put_failure;
1095 
1096 	if (rtnl_phys_switch_id_fill(skb, dev))
1097 		goto nla_put_failure;
1098 
1099 	attr = nla_reserve(skb, IFLA_STATS,
1100 			sizeof(struct rtnl_link_stats));
1101 	if (attr == NULL)
1102 		goto nla_put_failure;
1103 
1104 	stats = dev_get_stats(dev, &temp);
1105 	copy_rtnl_link_stats(nla_data(attr), stats);
1106 
1107 	attr = nla_reserve(skb, IFLA_STATS64,
1108 			sizeof(struct rtnl_link_stats64));
1109 	if (attr == NULL)
1110 		goto nla_put_failure;
1111 	copy_rtnl_link_stats64(nla_data(attr), stats);
1112 
1113 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1114 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1115 		goto nla_put_failure;
1116 
1117 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1118 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
1119 		int i;
1120 
1121 		struct nlattr *vfinfo, *vf;
1122 		int num_vfs = dev_num_vf(dev->dev.parent);
1123 
1124 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1125 		if (!vfinfo)
1126 			goto nla_put_failure;
1127 		for (i = 0; i < num_vfs; i++) {
1128 			struct ifla_vf_info ivi;
1129 			struct ifla_vf_mac vf_mac;
1130 			struct ifla_vf_vlan vf_vlan;
1131 			struct ifla_vf_rate vf_rate;
1132 			struct ifla_vf_tx_rate vf_tx_rate;
1133 			struct ifla_vf_spoofchk vf_spoofchk;
1134 			struct ifla_vf_link_state vf_linkstate;
1135 
1136 			/*
1137 			 * Not all SR-IOV capable drivers support the
1138 			 * spoofcheck query.  Preset to -1 so the user
1139 			 * space tool can detect that the driver didn't
1140 			 * report anything.
1141 			 */
1142 			ivi.spoofchk = -1;
1143 			memset(ivi.mac, 0, sizeof(ivi.mac));
1144 			/* The default value for VF link state is "auto"
1145 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
1146 			 */
1147 			ivi.linkstate = 0;
1148 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1149 				break;
1150 			vf_mac.vf =
1151 				vf_vlan.vf =
1152 				vf_rate.vf =
1153 				vf_tx_rate.vf =
1154 				vf_spoofchk.vf =
1155 				vf_linkstate.vf = ivi.vf;
1156 
1157 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1158 			vf_vlan.vlan = ivi.vlan;
1159 			vf_vlan.qos = ivi.qos;
1160 			vf_tx_rate.rate = ivi.max_tx_rate;
1161 			vf_rate.min_tx_rate = ivi.min_tx_rate;
1162 			vf_rate.max_tx_rate = ivi.max_tx_rate;
1163 			vf_spoofchk.setting = ivi.spoofchk;
1164 			vf_linkstate.link_state = ivi.linkstate;
1165 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1166 			if (!vf) {
1167 				nla_nest_cancel(skb, vfinfo);
1168 				goto nla_put_failure;
1169 			}
1170 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1171 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1172 			    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1173 				    &vf_rate) ||
1174 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1175 				    &vf_tx_rate) ||
1176 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1177 				    &vf_spoofchk) ||
1178 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1179 				    &vf_linkstate))
1180 				goto nla_put_failure;
1181 			nla_nest_end(skb, vf);
1182 		}
1183 		nla_nest_end(skb, vfinfo);
1184 	}
1185 
1186 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1187 		goto nla_put_failure;
1188 
1189 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1190 		if (rtnl_link_fill(skb, dev) < 0)
1191 			goto nla_put_failure;
1192 	}
1193 
1194 	if (dev->rtnl_link_ops &&
1195 	    dev->rtnl_link_ops->get_link_net) {
1196 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1197 
1198 		if (!net_eq(dev_net(dev), link_net)) {
1199 			int id = peernet2id(dev_net(dev), link_net);
1200 
1201 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1202 				goto nla_put_failure;
1203 		}
1204 	}
1205 
1206 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1207 		goto nla_put_failure;
1208 
1209 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1210 		if (af_ops->fill_link_af) {
1211 			struct nlattr *af;
1212 			int err;
1213 
1214 			if (!(af = nla_nest_start(skb, af_ops->family)))
1215 				goto nla_put_failure;
1216 
1217 			err = af_ops->fill_link_af(skb, dev);
1218 
1219 			/*
1220 			 * Caller may return ENODATA to indicate that there
1221 			 * was no data to be dumped. This is not an error, it
1222 			 * means we should trim the attribute header and
1223 			 * continue.
1224 			 */
1225 			if (err == -ENODATA)
1226 				nla_nest_cancel(skb, af);
1227 			else if (err < 0)
1228 				goto nla_put_failure;
1229 
1230 			nla_nest_end(skb, af);
1231 		}
1232 	}
1233 
1234 	nla_nest_end(skb, af_spec);
1235 
1236 	nlmsg_end(skb, nlh);
1237 	return 0;
1238 
1239 nla_put_failure:
1240 	nlmsg_cancel(skb, nlh);
1241 	return -EMSGSIZE;
1242 }
1243 
1244 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1245 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1246 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1247 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1248 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1249 	[IFLA_MTU]		= { .type = NLA_U32 },
1250 	[IFLA_LINK]		= { .type = NLA_U32 },
1251 	[IFLA_MASTER]		= { .type = NLA_U32 },
1252 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1253 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1254 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1255 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1256 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1257 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1258 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1259 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1260 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1261 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1262 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1263 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1264 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1265 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1266 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1267 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1268 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1269 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1270 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1271 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1272 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1273 };
1274 
1275 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1276 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1277 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1278 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1279 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1280 };
1281 
1282 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1283 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1284 };
1285 
1286 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1287 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1288 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1289 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1290 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1291 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1292 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1293 };
1294 
1295 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1296 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1297 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1298 				    .len = PORT_PROFILE_MAX },
1299 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1300 				    .len = sizeof(struct ifla_port_vsi)},
1301 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1302 				      .len = PORT_UUID_MAX },
1303 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1304 				    .len = PORT_UUID_MAX },
1305 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1306 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1307 };
1308 
1309 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1310 {
1311 	struct net *net = sock_net(skb->sk);
1312 	int h, s_h;
1313 	int idx = 0, s_idx;
1314 	struct net_device *dev;
1315 	struct hlist_head *head;
1316 	struct nlattr *tb[IFLA_MAX+1];
1317 	u32 ext_filter_mask = 0;
1318 	int err;
1319 	int hdrlen;
1320 
1321 	s_h = cb->args[0];
1322 	s_idx = cb->args[1];
1323 
1324 	cb->seq = net->dev_base_seq;
1325 
1326 	/* A hack to preserve kernel<->userspace interface.
1327 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1328 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1329 	 * what iproute2 < v3.9.0 used.
1330 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1331 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1332 	 */
1333 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1334 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1335 
1336 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1337 
1338 		if (tb[IFLA_EXT_MASK])
1339 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1340 	}
1341 
1342 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1343 		idx = 0;
1344 		head = &net->dev_index_head[h];
1345 		hlist_for_each_entry(dev, head, index_hlist) {
1346 			if (idx < s_idx)
1347 				goto cont;
1348 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1349 					       NETLINK_CB(cb->skb).portid,
1350 					       cb->nlh->nlmsg_seq, 0,
1351 					       NLM_F_MULTI,
1352 					       ext_filter_mask);
1353 			/* If we ran out of room on the first message,
1354 			 * we're in trouble
1355 			 */
1356 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1357 
1358 			if (err < 0)
1359 				goto out;
1360 
1361 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1362 cont:
1363 			idx++;
1364 		}
1365 	}
1366 out:
1367 	cb->args[1] = idx;
1368 	cb->args[0] = h;
1369 
1370 	return skb->len;
1371 }
1372 
1373 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1374 {
1375 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1376 }
1377 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1378 
1379 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1380 {
1381 	struct net *net;
1382 	/* Examine the link attributes and figure out which
1383 	 * network namespace we are talking about.
1384 	 */
1385 	if (tb[IFLA_NET_NS_PID])
1386 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1387 	else if (tb[IFLA_NET_NS_FD])
1388 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1389 	else
1390 		net = get_net(src_net);
1391 	return net;
1392 }
1393 EXPORT_SYMBOL(rtnl_link_get_net);
1394 
1395 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1396 {
1397 	if (dev) {
1398 		if (tb[IFLA_ADDRESS] &&
1399 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1400 			return -EINVAL;
1401 
1402 		if (tb[IFLA_BROADCAST] &&
1403 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1404 			return -EINVAL;
1405 	}
1406 
1407 	if (tb[IFLA_AF_SPEC]) {
1408 		struct nlattr *af;
1409 		int rem, err;
1410 
1411 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1412 			const struct rtnl_af_ops *af_ops;
1413 
1414 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1415 				return -EAFNOSUPPORT;
1416 
1417 			if (!af_ops->set_link_af)
1418 				return -EOPNOTSUPP;
1419 
1420 			if (af_ops->validate_link_af) {
1421 				err = af_ops->validate_link_af(dev, af);
1422 				if (err < 0)
1423 					return err;
1424 			}
1425 		}
1426 	}
1427 
1428 	return 0;
1429 }
1430 
1431 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1432 {
1433 	int rem, err = -EINVAL;
1434 	struct nlattr *vf;
1435 	const struct net_device_ops *ops = dev->netdev_ops;
1436 
1437 	nla_for_each_nested(vf, attr, rem) {
1438 		switch (nla_type(vf)) {
1439 		case IFLA_VF_MAC: {
1440 			struct ifla_vf_mac *ivm;
1441 			ivm = nla_data(vf);
1442 			err = -EOPNOTSUPP;
1443 			if (ops->ndo_set_vf_mac)
1444 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1445 							  ivm->mac);
1446 			break;
1447 		}
1448 		case IFLA_VF_VLAN: {
1449 			struct ifla_vf_vlan *ivv;
1450 			ivv = nla_data(vf);
1451 			err = -EOPNOTSUPP;
1452 			if (ops->ndo_set_vf_vlan)
1453 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1454 							   ivv->vlan,
1455 							   ivv->qos);
1456 			break;
1457 		}
1458 		case IFLA_VF_TX_RATE: {
1459 			struct ifla_vf_tx_rate *ivt;
1460 			struct ifla_vf_info ivf;
1461 			ivt = nla_data(vf);
1462 			err = -EOPNOTSUPP;
1463 			if (ops->ndo_get_vf_config)
1464 				err = ops->ndo_get_vf_config(dev, ivt->vf,
1465 							     &ivf);
1466 			if (err)
1467 				break;
1468 			err = -EOPNOTSUPP;
1469 			if (ops->ndo_set_vf_rate)
1470 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1471 							   ivf.min_tx_rate,
1472 							   ivt->rate);
1473 			break;
1474 		}
1475 		case IFLA_VF_RATE: {
1476 			struct ifla_vf_rate *ivt;
1477 			ivt = nla_data(vf);
1478 			err = -EOPNOTSUPP;
1479 			if (ops->ndo_set_vf_rate)
1480 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1481 							   ivt->min_tx_rate,
1482 							   ivt->max_tx_rate);
1483 			break;
1484 		}
1485 		case IFLA_VF_SPOOFCHK: {
1486 			struct ifla_vf_spoofchk *ivs;
1487 			ivs = nla_data(vf);
1488 			err = -EOPNOTSUPP;
1489 			if (ops->ndo_set_vf_spoofchk)
1490 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1491 							       ivs->setting);
1492 			break;
1493 		}
1494 		case IFLA_VF_LINK_STATE: {
1495 			struct ifla_vf_link_state *ivl;
1496 			ivl = nla_data(vf);
1497 			err = -EOPNOTSUPP;
1498 			if (ops->ndo_set_vf_link_state)
1499 				err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1500 								 ivl->link_state);
1501 			break;
1502 		}
1503 		default:
1504 			err = -EINVAL;
1505 			break;
1506 		}
1507 		if (err)
1508 			break;
1509 	}
1510 	return err;
1511 }
1512 
1513 static int do_set_master(struct net_device *dev, int ifindex)
1514 {
1515 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1516 	const struct net_device_ops *ops;
1517 	int err;
1518 
1519 	if (upper_dev) {
1520 		if (upper_dev->ifindex == ifindex)
1521 			return 0;
1522 		ops = upper_dev->netdev_ops;
1523 		if (ops->ndo_del_slave) {
1524 			err = ops->ndo_del_slave(upper_dev, dev);
1525 			if (err)
1526 				return err;
1527 		} else {
1528 			return -EOPNOTSUPP;
1529 		}
1530 	}
1531 
1532 	if (ifindex) {
1533 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1534 		if (!upper_dev)
1535 			return -EINVAL;
1536 		ops = upper_dev->netdev_ops;
1537 		if (ops->ndo_add_slave) {
1538 			err = ops->ndo_add_slave(upper_dev, dev);
1539 			if (err)
1540 				return err;
1541 		} else {
1542 			return -EOPNOTSUPP;
1543 		}
1544 	}
1545 	return 0;
1546 }
1547 
1548 #define DO_SETLINK_MODIFIED	0x01
1549 /* notify flag means notify + modified. */
1550 #define DO_SETLINK_NOTIFY	0x03
1551 static int do_setlink(const struct sk_buff *skb,
1552 		      struct net_device *dev, struct ifinfomsg *ifm,
1553 		      struct nlattr **tb, char *ifname, int status)
1554 {
1555 	const struct net_device_ops *ops = dev->netdev_ops;
1556 	int err;
1557 
1558 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1559 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1560 		if (IS_ERR(net)) {
1561 			err = PTR_ERR(net);
1562 			goto errout;
1563 		}
1564 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1565 			put_net(net);
1566 			err = -EPERM;
1567 			goto errout;
1568 		}
1569 		err = dev_change_net_namespace(dev, net, ifname);
1570 		put_net(net);
1571 		if (err)
1572 			goto errout;
1573 		status |= DO_SETLINK_MODIFIED;
1574 	}
1575 
1576 	if (tb[IFLA_MAP]) {
1577 		struct rtnl_link_ifmap *u_map;
1578 		struct ifmap k_map;
1579 
1580 		if (!ops->ndo_set_config) {
1581 			err = -EOPNOTSUPP;
1582 			goto errout;
1583 		}
1584 
1585 		if (!netif_device_present(dev)) {
1586 			err = -ENODEV;
1587 			goto errout;
1588 		}
1589 
1590 		u_map = nla_data(tb[IFLA_MAP]);
1591 		k_map.mem_start = (unsigned long) u_map->mem_start;
1592 		k_map.mem_end = (unsigned long) u_map->mem_end;
1593 		k_map.base_addr = (unsigned short) u_map->base_addr;
1594 		k_map.irq = (unsigned char) u_map->irq;
1595 		k_map.dma = (unsigned char) u_map->dma;
1596 		k_map.port = (unsigned char) u_map->port;
1597 
1598 		err = ops->ndo_set_config(dev, &k_map);
1599 		if (err < 0)
1600 			goto errout;
1601 
1602 		status |= DO_SETLINK_NOTIFY;
1603 	}
1604 
1605 	if (tb[IFLA_ADDRESS]) {
1606 		struct sockaddr *sa;
1607 		int len;
1608 
1609 		len = sizeof(sa_family_t) + dev->addr_len;
1610 		sa = kmalloc(len, GFP_KERNEL);
1611 		if (!sa) {
1612 			err = -ENOMEM;
1613 			goto errout;
1614 		}
1615 		sa->sa_family = dev->type;
1616 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1617 		       dev->addr_len);
1618 		err = dev_set_mac_address(dev, sa);
1619 		kfree(sa);
1620 		if (err)
1621 			goto errout;
1622 		status |= DO_SETLINK_MODIFIED;
1623 	}
1624 
1625 	if (tb[IFLA_MTU]) {
1626 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1627 		if (err < 0)
1628 			goto errout;
1629 		status |= DO_SETLINK_MODIFIED;
1630 	}
1631 
1632 	if (tb[IFLA_GROUP]) {
1633 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1634 		status |= DO_SETLINK_NOTIFY;
1635 	}
1636 
1637 	/*
1638 	 * Interface selected by interface index but interface
1639 	 * name provided implies that a name change has been
1640 	 * requested.
1641 	 */
1642 	if (ifm->ifi_index > 0 && ifname[0]) {
1643 		err = dev_change_name(dev, ifname);
1644 		if (err < 0)
1645 			goto errout;
1646 		status |= DO_SETLINK_MODIFIED;
1647 	}
1648 
1649 	if (tb[IFLA_IFALIAS]) {
1650 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1651 				    nla_len(tb[IFLA_IFALIAS]));
1652 		if (err < 0)
1653 			goto errout;
1654 		status |= DO_SETLINK_NOTIFY;
1655 	}
1656 
1657 	if (tb[IFLA_BROADCAST]) {
1658 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1659 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1660 	}
1661 
1662 	if (ifm->ifi_flags || ifm->ifi_change) {
1663 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1664 		if (err < 0)
1665 			goto errout;
1666 	}
1667 
1668 	if (tb[IFLA_MASTER]) {
1669 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1670 		if (err)
1671 			goto errout;
1672 		status |= DO_SETLINK_MODIFIED;
1673 	}
1674 
1675 	if (tb[IFLA_CARRIER]) {
1676 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1677 		if (err)
1678 			goto errout;
1679 		status |= DO_SETLINK_MODIFIED;
1680 	}
1681 
1682 	if (tb[IFLA_TXQLEN]) {
1683 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1684 
1685 		if (dev->tx_queue_len ^ value)
1686 			status |= DO_SETLINK_NOTIFY;
1687 
1688 		dev->tx_queue_len = value;
1689 	}
1690 
1691 	if (tb[IFLA_OPERSTATE])
1692 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1693 
1694 	if (tb[IFLA_LINKMODE]) {
1695 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1696 
1697 		write_lock_bh(&dev_base_lock);
1698 		if (dev->link_mode ^ value)
1699 			status |= DO_SETLINK_NOTIFY;
1700 		dev->link_mode = value;
1701 		write_unlock_bh(&dev_base_lock);
1702 	}
1703 
1704 	if (tb[IFLA_VFINFO_LIST]) {
1705 		struct nlattr *attr;
1706 		int rem;
1707 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1708 			if (nla_type(attr) != IFLA_VF_INFO) {
1709 				err = -EINVAL;
1710 				goto errout;
1711 			}
1712 			err = do_setvfinfo(dev, attr);
1713 			if (err < 0)
1714 				goto errout;
1715 			status |= DO_SETLINK_NOTIFY;
1716 		}
1717 	}
1718 	err = 0;
1719 
1720 	if (tb[IFLA_VF_PORTS]) {
1721 		struct nlattr *port[IFLA_PORT_MAX+1];
1722 		struct nlattr *attr;
1723 		int vf;
1724 		int rem;
1725 
1726 		err = -EOPNOTSUPP;
1727 		if (!ops->ndo_set_vf_port)
1728 			goto errout;
1729 
1730 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1731 			if (nla_type(attr) != IFLA_VF_PORT)
1732 				continue;
1733 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1734 				attr, ifla_port_policy);
1735 			if (err < 0)
1736 				goto errout;
1737 			if (!port[IFLA_PORT_VF]) {
1738 				err = -EOPNOTSUPP;
1739 				goto errout;
1740 			}
1741 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1742 			err = ops->ndo_set_vf_port(dev, vf, port);
1743 			if (err < 0)
1744 				goto errout;
1745 			status |= DO_SETLINK_NOTIFY;
1746 		}
1747 	}
1748 	err = 0;
1749 
1750 	if (tb[IFLA_PORT_SELF]) {
1751 		struct nlattr *port[IFLA_PORT_MAX+1];
1752 
1753 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1754 			tb[IFLA_PORT_SELF], ifla_port_policy);
1755 		if (err < 0)
1756 			goto errout;
1757 
1758 		err = -EOPNOTSUPP;
1759 		if (ops->ndo_set_vf_port)
1760 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1761 		if (err < 0)
1762 			goto errout;
1763 		status |= DO_SETLINK_NOTIFY;
1764 	}
1765 
1766 	if (tb[IFLA_AF_SPEC]) {
1767 		struct nlattr *af;
1768 		int rem;
1769 
1770 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1771 			const struct rtnl_af_ops *af_ops;
1772 
1773 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1774 				BUG();
1775 
1776 			err = af_ops->set_link_af(dev, af);
1777 			if (err < 0)
1778 				goto errout;
1779 
1780 			status |= DO_SETLINK_NOTIFY;
1781 		}
1782 	}
1783 	err = 0;
1784 
1785 errout:
1786 	if (status & DO_SETLINK_MODIFIED) {
1787 		if (status & DO_SETLINK_NOTIFY)
1788 			netdev_state_change(dev);
1789 
1790 		if (err < 0)
1791 			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",
1792 					     dev->name);
1793 	}
1794 
1795 	return err;
1796 }
1797 
1798 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1799 {
1800 	struct net *net = sock_net(skb->sk);
1801 	struct ifinfomsg *ifm;
1802 	struct net_device *dev;
1803 	int err;
1804 	struct nlattr *tb[IFLA_MAX+1];
1805 	char ifname[IFNAMSIZ];
1806 
1807 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1808 	if (err < 0)
1809 		goto errout;
1810 
1811 	if (tb[IFLA_IFNAME])
1812 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1813 	else
1814 		ifname[0] = '\0';
1815 
1816 	err = -EINVAL;
1817 	ifm = nlmsg_data(nlh);
1818 	if (ifm->ifi_index > 0)
1819 		dev = __dev_get_by_index(net, ifm->ifi_index);
1820 	else if (tb[IFLA_IFNAME])
1821 		dev = __dev_get_by_name(net, ifname);
1822 	else
1823 		goto errout;
1824 
1825 	if (dev == NULL) {
1826 		err = -ENODEV;
1827 		goto errout;
1828 	}
1829 
1830 	err = validate_linkmsg(dev, tb);
1831 	if (err < 0)
1832 		goto errout;
1833 
1834 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1835 errout:
1836 	return err;
1837 }
1838 
1839 static int rtnl_group_dellink(const struct net *net, int group)
1840 {
1841 	struct net_device *dev, *aux;
1842 	LIST_HEAD(list_kill);
1843 	bool found = false;
1844 
1845 	if (!group)
1846 		return -EPERM;
1847 
1848 	for_each_netdev(net, dev) {
1849 		if (dev->group == group) {
1850 			const struct rtnl_link_ops *ops;
1851 
1852 			found = true;
1853 			ops = dev->rtnl_link_ops;
1854 			if (!ops || !ops->dellink)
1855 				return -EOPNOTSUPP;
1856 		}
1857 	}
1858 
1859 	if (!found)
1860 		return -ENODEV;
1861 
1862 	for_each_netdev_safe(net, dev, aux) {
1863 		if (dev->group == group) {
1864 			const struct rtnl_link_ops *ops;
1865 
1866 			ops = dev->rtnl_link_ops;
1867 			ops->dellink(dev, &list_kill);
1868 		}
1869 	}
1870 	unregister_netdevice_many(&list_kill);
1871 
1872 	return 0;
1873 }
1874 
1875 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1876 {
1877 	struct net *net = sock_net(skb->sk);
1878 	const struct rtnl_link_ops *ops;
1879 	struct net_device *dev;
1880 	struct ifinfomsg *ifm;
1881 	char ifname[IFNAMSIZ];
1882 	struct nlattr *tb[IFLA_MAX+1];
1883 	int err;
1884 	LIST_HEAD(list_kill);
1885 
1886 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1887 	if (err < 0)
1888 		return err;
1889 
1890 	if (tb[IFLA_IFNAME])
1891 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1892 
1893 	ifm = nlmsg_data(nlh);
1894 	if (ifm->ifi_index > 0)
1895 		dev = __dev_get_by_index(net, ifm->ifi_index);
1896 	else if (tb[IFLA_IFNAME])
1897 		dev = __dev_get_by_name(net, ifname);
1898 	else if (tb[IFLA_GROUP])
1899 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
1900 	else
1901 		return -EINVAL;
1902 
1903 	if (!dev)
1904 		return -ENODEV;
1905 
1906 	ops = dev->rtnl_link_ops;
1907 	if (!ops || !ops->dellink)
1908 		return -EOPNOTSUPP;
1909 
1910 	ops->dellink(dev, &list_kill);
1911 	unregister_netdevice_many(&list_kill);
1912 	return 0;
1913 }
1914 
1915 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1916 {
1917 	unsigned int old_flags;
1918 	int err;
1919 
1920 	old_flags = dev->flags;
1921 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1922 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1923 		if (err < 0)
1924 			return err;
1925 	}
1926 
1927 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1928 
1929 	__dev_notify_flags(dev, old_flags, ~0U);
1930 	return 0;
1931 }
1932 EXPORT_SYMBOL(rtnl_configure_link);
1933 
1934 struct net_device *rtnl_create_link(struct net *net,
1935 	char *ifname, unsigned char name_assign_type,
1936 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
1937 {
1938 	int err;
1939 	struct net_device *dev;
1940 	unsigned int num_tx_queues = 1;
1941 	unsigned int num_rx_queues = 1;
1942 
1943 	if (tb[IFLA_NUM_TX_QUEUES])
1944 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1945 	else if (ops->get_num_tx_queues)
1946 		num_tx_queues = ops->get_num_tx_queues();
1947 
1948 	if (tb[IFLA_NUM_RX_QUEUES])
1949 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1950 	else if (ops->get_num_rx_queues)
1951 		num_rx_queues = ops->get_num_rx_queues();
1952 
1953 	err = -ENOMEM;
1954 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1955 			       ops->setup, num_tx_queues, num_rx_queues);
1956 	if (!dev)
1957 		goto err;
1958 
1959 	dev_net_set(dev, net);
1960 	dev->rtnl_link_ops = ops;
1961 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1962 
1963 	if (tb[IFLA_MTU])
1964 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1965 	if (tb[IFLA_ADDRESS]) {
1966 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1967 				nla_len(tb[IFLA_ADDRESS]));
1968 		dev->addr_assign_type = NET_ADDR_SET;
1969 	}
1970 	if (tb[IFLA_BROADCAST])
1971 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1972 				nla_len(tb[IFLA_BROADCAST]));
1973 	if (tb[IFLA_TXQLEN])
1974 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1975 	if (tb[IFLA_OPERSTATE])
1976 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1977 	if (tb[IFLA_LINKMODE])
1978 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1979 	if (tb[IFLA_GROUP])
1980 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1981 
1982 	return dev;
1983 
1984 err:
1985 	return ERR_PTR(err);
1986 }
1987 EXPORT_SYMBOL(rtnl_create_link);
1988 
1989 static int rtnl_group_changelink(const struct sk_buff *skb,
1990 		struct net *net, int group,
1991 		struct ifinfomsg *ifm,
1992 		struct nlattr **tb)
1993 {
1994 	struct net_device *dev;
1995 	int err;
1996 
1997 	for_each_netdev(net, dev) {
1998 		if (dev->group == group) {
1999 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2000 			if (err < 0)
2001 				return err;
2002 		}
2003 	}
2004 
2005 	return 0;
2006 }
2007 
2008 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2009 {
2010 	struct net *net = sock_net(skb->sk);
2011 	const struct rtnl_link_ops *ops;
2012 	const struct rtnl_link_ops *m_ops = NULL;
2013 	struct net_device *dev;
2014 	struct net_device *master_dev = NULL;
2015 	struct ifinfomsg *ifm;
2016 	char kind[MODULE_NAME_LEN];
2017 	char ifname[IFNAMSIZ];
2018 	struct nlattr *tb[IFLA_MAX+1];
2019 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2020 	unsigned char name_assign_type = NET_NAME_USER;
2021 	int err;
2022 
2023 #ifdef CONFIG_MODULES
2024 replay:
2025 #endif
2026 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2027 	if (err < 0)
2028 		return err;
2029 
2030 	if (tb[IFLA_IFNAME])
2031 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2032 	else
2033 		ifname[0] = '\0';
2034 
2035 	ifm = nlmsg_data(nlh);
2036 	if (ifm->ifi_index > 0)
2037 		dev = __dev_get_by_index(net, ifm->ifi_index);
2038 	else {
2039 		if (ifname[0])
2040 			dev = __dev_get_by_name(net, ifname);
2041 		else
2042 			dev = NULL;
2043 	}
2044 
2045 	if (dev) {
2046 		master_dev = netdev_master_upper_dev_get(dev);
2047 		if (master_dev)
2048 			m_ops = master_dev->rtnl_link_ops;
2049 	}
2050 
2051 	err = validate_linkmsg(dev, tb);
2052 	if (err < 0)
2053 		return err;
2054 
2055 	if (tb[IFLA_LINKINFO]) {
2056 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2057 				       tb[IFLA_LINKINFO], ifla_info_policy);
2058 		if (err < 0)
2059 			return err;
2060 	} else
2061 		memset(linkinfo, 0, sizeof(linkinfo));
2062 
2063 	if (linkinfo[IFLA_INFO_KIND]) {
2064 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2065 		ops = rtnl_link_ops_get(kind);
2066 	} else {
2067 		kind[0] = '\0';
2068 		ops = NULL;
2069 	}
2070 
2071 	if (1) {
2072 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2073 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2074 		struct nlattr **data = NULL;
2075 		struct nlattr **slave_data = NULL;
2076 		struct net *dest_net, *link_net = NULL;
2077 
2078 		if (ops) {
2079 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2080 				err = nla_parse_nested(attr, ops->maxtype,
2081 						       linkinfo[IFLA_INFO_DATA],
2082 						       ops->policy);
2083 				if (err < 0)
2084 					return err;
2085 				data = attr;
2086 			}
2087 			if (ops->validate) {
2088 				err = ops->validate(tb, data);
2089 				if (err < 0)
2090 					return err;
2091 			}
2092 		}
2093 
2094 		if (m_ops) {
2095 			if (m_ops->slave_maxtype &&
2096 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2097 				err = nla_parse_nested(slave_attr,
2098 						       m_ops->slave_maxtype,
2099 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2100 						       m_ops->slave_policy);
2101 				if (err < 0)
2102 					return err;
2103 				slave_data = slave_attr;
2104 			}
2105 			if (m_ops->slave_validate) {
2106 				err = m_ops->slave_validate(tb, slave_data);
2107 				if (err < 0)
2108 					return err;
2109 			}
2110 		}
2111 
2112 		if (dev) {
2113 			int status = 0;
2114 
2115 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2116 				return -EEXIST;
2117 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2118 				return -EOPNOTSUPP;
2119 
2120 			if (linkinfo[IFLA_INFO_DATA]) {
2121 				if (!ops || ops != dev->rtnl_link_ops ||
2122 				    !ops->changelink)
2123 					return -EOPNOTSUPP;
2124 
2125 				err = ops->changelink(dev, tb, data);
2126 				if (err < 0)
2127 					return err;
2128 				status |= DO_SETLINK_NOTIFY;
2129 			}
2130 
2131 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2132 				if (!m_ops || !m_ops->slave_changelink)
2133 					return -EOPNOTSUPP;
2134 
2135 				err = m_ops->slave_changelink(master_dev, dev,
2136 							      tb, slave_data);
2137 				if (err < 0)
2138 					return err;
2139 				status |= DO_SETLINK_NOTIFY;
2140 			}
2141 
2142 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2143 		}
2144 
2145 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2146 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2147 				return rtnl_group_changelink(skb, net,
2148 						nla_get_u32(tb[IFLA_GROUP]),
2149 						ifm, tb);
2150 			return -ENODEV;
2151 		}
2152 
2153 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2154 			return -EOPNOTSUPP;
2155 
2156 		if (!ops) {
2157 #ifdef CONFIG_MODULES
2158 			if (kind[0]) {
2159 				__rtnl_unlock();
2160 				request_module("rtnl-link-%s", kind);
2161 				rtnl_lock();
2162 				ops = rtnl_link_ops_get(kind);
2163 				if (ops)
2164 					goto replay;
2165 			}
2166 #endif
2167 			return -EOPNOTSUPP;
2168 		}
2169 
2170 		if (!ops->setup)
2171 			return -EOPNOTSUPP;
2172 
2173 		if (!ifname[0]) {
2174 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2175 			name_assign_type = NET_NAME_ENUM;
2176 		}
2177 
2178 		dest_net = rtnl_link_get_net(net, tb);
2179 		if (IS_ERR(dest_net))
2180 			return PTR_ERR(dest_net);
2181 
2182 		err = -EPERM;
2183 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2184 			goto out;
2185 
2186 		if (tb[IFLA_LINK_NETNSID]) {
2187 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2188 
2189 			link_net = get_net_ns_by_id(dest_net, id);
2190 			if (!link_net) {
2191 				err =  -EINVAL;
2192 				goto out;
2193 			}
2194 			err = -EPERM;
2195 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2196 				goto out;
2197 		}
2198 
2199 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2200 				       name_assign_type, ops, tb);
2201 		if (IS_ERR(dev)) {
2202 			err = PTR_ERR(dev);
2203 			goto out;
2204 		}
2205 
2206 		dev->ifindex = ifm->ifi_index;
2207 
2208 		if (ops->newlink) {
2209 			err = ops->newlink(link_net ? : net, dev, tb, data);
2210 			/* Drivers should call free_netdev() in ->destructor
2211 			 * and unregister it on failure after registration
2212 			 * so that device could be finally freed in rtnl_unlock.
2213 			 */
2214 			if (err < 0) {
2215 				/* If device is not registered at all, free it now */
2216 				if (dev->reg_state == NETREG_UNINITIALIZED)
2217 					free_netdev(dev);
2218 				goto out;
2219 			}
2220 		} else {
2221 			err = register_netdevice(dev);
2222 			if (err < 0) {
2223 				free_netdev(dev);
2224 				goto out;
2225 			}
2226 		}
2227 		err = rtnl_configure_link(dev, ifm);
2228 		if (err < 0)
2229 			goto out_unregister;
2230 		if (link_net) {
2231 			err = dev_change_net_namespace(dev, dest_net, ifname);
2232 			if (err < 0)
2233 				goto out_unregister;
2234 		}
2235 out:
2236 		if (link_net)
2237 			put_net(link_net);
2238 		put_net(dest_net);
2239 		return err;
2240 out_unregister:
2241 		if (ops->newlink) {
2242 			LIST_HEAD(list_kill);
2243 
2244 			ops->dellink(dev, &list_kill);
2245 			unregister_netdevice_many(&list_kill);
2246 		} else {
2247 			unregister_netdevice(dev);
2248 		}
2249 		goto out;
2250 	}
2251 }
2252 
2253 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2254 {
2255 	struct net *net = sock_net(skb->sk);
2256 	struct ifinfomsg *ifm;
2257 	char ifname[IFNAMSIZ];
2258 	struct nlattr *tb[IFLA_MAX+1];
2259 	struct net_device *dev = NULL;
2260 	struct sk_buff *nskb;
2261 	int err;
2262 	u32 ext_filter_mask = 0;
2263 
2264 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2265 	if (err < 0)
2266 		return err;
2267 
2268 	if (tb[IFLA_IFNAME])
2269 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2270 
2271 	if (tb[IFLA_EXT_MASK])
2272 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2273 
2274 	ifm = nlmsg_data(nlh);
2275 	if (ifm->ifi_index > 0)
2276 		dev = __dev_get_by_index(net, ifm->ifi_index);
2277 	else if (tb[IFLA_IFNAME])
2278 		dev = __dev_get_by_name(net, ifname);
2279 	else
2280 		return -EINVAL;
2281 
2282 	if (dev == NULL)
2283 		return -ENODEV;
2284 
2285 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2286 	if (nskb == NULL)
2287 		return -ENOBUFS;
2288 
2289 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2290 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2291 	if (err < 0) {
2292 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2293 		WARN_ON(err == -EMSGSIZE);
2294 		kfree_skb(nskb);
2295 	} else
2296 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2297 
2298 	return err;
2299 }
2300 
2301 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2302 {
2303 	struct net *net = sock_net(skb->sk);
2304 	struct net_device *dev;
2305 	struct nlattr *tb[IFLA_MAX+1];
2306 	u32 ext_filter_mask = 0;
2307 	u16 min_ifinfo_dump_size = 0;
2308 	int hdrlen;
2309 
2310 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2311 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2312 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2313 
2314 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2315 		if (tb[IFLA_EXT_MASK])
2316 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2317 	}
2318 
2319 	if (!ext_filter_mask)
2320 		return NLMSG_GOODSIZE;
2321 	/*
2322 	 * traverse the list of net devices and compute the minimum
2323 	 * buffer size based upon the filter mask.
2324 	 */
2325 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2326 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2327 					     if_nlmsg_size(dev,
2328 						           ext_filter_mask));
2329 	}
2330 
2331 	return min_ifinfo_dump_size;
2332 }
2333 
2334 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2335 {
2336 	int idx;
2337 	int s_idx = cb->family;
2338 
2339 	if (s_idx == 0)
2340 		s_idx = 1;
2341 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2342 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2343 		if (idx < s_idx || idx == PF_PACKET)
2344 			continue;
2345 		if (rtnl_msg_handlers[idx] == NULL ||
2346 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2347 			continue;
2348 		if (idx > s_idx) {
2349 			memset(&cb->args[0], 0, sizeof(cb->args));
2350 			cb->prev_seq = 0;
2351 			cb->seq = 0;
2352 		}
2353 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2354 			break;
2355 	}
2356 	cb->family = idx;
2357 
2358 	return skb->len;
2359 }
2360 
2361 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2362 				       unsigned int change, gfp_t flags)
2363 {
2364 	struct net *net = dev_net(dev);
2365 	struct sk_buff *skb;
2366 	int err = -ENOBUFS;
2367 	size_t if_info_size;
2368 
2369 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2370 	if (skb == NULL)
2371 		goto errout;
2372 
2373 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2374 	if (err < 0) {
2375 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2376 		WARN_ON(err == -EMSGSIZE);
2377 		kfree_skb(skb);
2378 		goto errout;
2379 	}
2380 	return skb;
2381 errout:
2382 	if (err < 0)
2383 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2384 	return NULL;
2385 }
2386 
2387 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2388 {
2389 	struct net *net = dev_net(dev);
2390 
2391 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2392 }
2393 
2394 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2395 		  gfp_t flags)
2396 {
2397 	struct sk_buff *skb;
2398 
2399 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2400 	if (skb)
2401 		rtmsg_ifinfo_send(skb, dev, flags);
2402 }
2403 EXPORT_SYMBOL(rtmsg_ifinfo);
2404 
2405 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2406 				   struct net_device *dev,
2407 				   u8 *addr, u32 pid, u32 seq,
2408 				   int type, unsigned int flags,
2409 				   int nlflags)
2410 {
2411 	struct nlmsghdr *nlh;
2412 	struct ndmsg *ndm;
2413 
2414 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2415 	if (!nlh)
2416 		return -EMSGSIZE;
2417 
2418 	ndm = nlmsg_data(nlh);
2419 	ndm->ndm_family  = AF_BRIDGE;
2420 	ndm->ndm_pad1	 = 0;
2421 	ndm->ndm_pad2    = 0;
2422 	ndm->ndm_flags	 = flags;
2423 	ndm->ndm_type	 = 0;
2424 	ndm->ndm_ifindex = dev->ifindex;
2425 	ndm->ndm_state   = NUD_PERMANENT;
2426 
2427 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2428 		goto nla_put_failure;
2429 
2430 	nlmsg_end(skb, nlh);
2431 	return 0;
2432 
2433 nla_put_failure:
2434 	nlmsg_cancel(skb, nlh);
2435 	return -EMSGSIZE;
2436 }
2437 
2438 static inline size_t rtnl_fdb_nlmsg_size(void)
2439 {
2440 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2441 }
2442 
2443 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2444 {
2445 	struct net *net = dev_net(dev);
2446 	struct sk_buff *skb;
2447 	int err = -ENOBUFS;
2448 
2449 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2450 	if (!skb)
2451 		goto errout;
2452 
2453 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2454 	if (err < 0) {
2455 		kfree_skb(skb);
2456 		goto errout;
2457 	}
2458 
2459 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2460 	return;
2461 errout:
2462 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2463 }
2464 
2465 /**
2466  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2467  */
2468 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2469 		     struct nlattr *tb[],
2470 		     struct net_device *dev,
2471 		     const unsigned char *addr, u16 vid,
2472 		     u16 flags)
2473 {
2474 	int err = -EINVAL;
2475 
2476 	/* If aging addresses are supported device will need to
2477 	 * implement its own handler for this.
2478 	 */
2479 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2480 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2481 		return err;
2482 	}
2483 
2484 	if (vid) {
2485 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2486 		return err;
2487 	}
2488 
2489 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2490 		err = dev_uc_add_excl(dev, addr);
2491 	else if (is_multicast_ether_addr(addr))
2492 		err = dev_mc_add_excl(dev, addr);
2493 
2494 	/* Only return duplicate errors if NLM_F_EXCL is set */
2495 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2496 		err = 0;
2497 
2498 	return err;
2499 }
2500 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2501 
2502 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2503 {
2504 	u16 vid = 0;
2505 
2506 	if (vlan_attr) {
2507 		if (nla_len(vlan_attr) != sizeof(u16)) {
2508 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2509 			return -EINVAL;
2510 		}
2511 
2512 		vid = nla_get_u16(vlan_attr);
2513 
2514 		if (!vid || vid >= VLAN_VID_MASK) {
2515 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2516 				vid);
2517 			return -EINVAL;
2518 		}
2519 	}
2520 	*p_vid = vid;
2521 	return 0;
2522 }
2523 
2524 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2525 {
2526 	struct net *net = sock_net(skb->sk);
2527 	struct ndmsg *ndm;
2528 	struct nlattr *tb[NDA_MAX+1];
2529 	struct net_device *dev;
2530 	u8 *addr;
2531 	u16 vid;
2532 	int err;
2533 
2534 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2535 	if (err < 0)
2536 		return err;
2537 
2538 	ndm = nlmsg_data(nlh);
2539 	if (ndm->ndm_ifindex == 0) {
2540 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2541 		return -EINVAL;
2542 	}
2543 
2544 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2545 	if (dev == NULL) {
2546 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2547 		return -ENODEV;
2548 	}
2549 
2550 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2551 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2552 		return -EINVAL;
2553 	}
2554 
2555 	addr = nla_data(tb[NDA_LLADDR]);
2556 
2557 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2558 	if (err)
2559 		return err;
2560 
2561 	err = -EOPNOTSUPP;
2562 
2563 	/* Support fdb on master device the net/bridge default case */
2564 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2565 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2566 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2567 		const struct net_device_ops *ops = br_dev->netdev_ops;
2568 
2569 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2570 				       nlh->nlmsg_flags);
2571 		if (err)
2572 			goto out;
2573 		else
2574 			ndm->ndm_flags &= ~NTF_MASTER;
2575 	}
2576 
2577 	/* Embedded bridge, macvlan, and any other device support */
2578 	if ((ndm->ndm_flags & NTF_SELF)) {
2579 		if (dev->netdev_ops->ndo_fdb_add)
2580 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2581 							   vid,
2582 							   nlh->nlmsg_flags);
2583 		else
2584 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2585 					       nlh->nlmsg_flags);
2586 
2587 		if (!err) {
2588 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2589 			ndm->ndm_flags &= ~NTF_SELF;
2590 		}
2591 	}
2592 out:
2593 	return err;
2594 }
2595 
2596 /**
2597  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2598  */
2599 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2600 		     struct nlattr *tb[],
2601 		     struct net_device *dev,
2602 		     const unsigned char *addr, u16 vid)
2603 {
2604 	int err = -EINVAL;
2605 
2606 	/* If aging addresses are supported device will need to
2607 	 * implement its own handler for this.
2608 	 */
2609 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2610 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2611 		return err;
2612 	}
2613 
2614 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2615 		err = dev_uc_del(dev, addr);
2616 	else if (is_multicast_ether_addr(addr))
2617 		err = dev_mc_del(dev, addr);
2618 
2619 	return err;
2620 }
2621 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2622 
2623 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2624 {
2625 	struct net *net = sock_net(skb->sk);
2626 	struct ndmsg *ndm;
2627 	struct nlattr *tb[NDA_MAX+1];
2628 	struct net_device *dev;
2629 	int err = -EINVAL;
2630 	__u8 *addr;
2631 	u16 vid;
2632 
2633 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2634 		return -EPERM;
2635 
2636 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2637 	if (err < 0)
2638 		return err;
2639 
2640 	ndm = nlmsg_data(nlh);
2641 	if (ndm->ndm_ifindex == 0) {
2642 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2643 		return -EINVAL;
2644 	}
2645 
2646 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2647 	if (dev == NULL) {
2648 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2649 		return -ENODEV;
2650 	}
2651 
2652 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2653 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2654 		return -EINVAL;
2655 	}
2656 
2657 	addr = nla_data(tb[NDA_LLADDR]);
2658 
2659 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2660 	if (err)
2661 		return err;
2662 
2663 	err = -EOPNOTSUPP;
2664 
2665 	/* Support fdb on master device the net/bridge default case */
2666 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2667 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2668 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2669 		const struct net_device_ops *ops = br_dev->netdev_ops;
2670 
2671 		if (ops->ndo_fdb_del)
2672 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2673 
2674 		if (err)
2675 			goto out;
2676 		else
2677 			ndm->ndm_flags &= ~NTF_MASTER;
2678 	}
2679 
2680 	/* Embedded bridge, macvlan, and any other device support */
2681 	if (ndm->ndm_flags & NTF_SELF) {
2682 		if (dev->netdev_ops->ndo_fdb_del)
2683 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2684 							   vid);
2685 		else
2686 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2687 
2688 		if (!err) {
2689 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2690 			ndm->ndm_flags &= ~NTF_SELF;
2691 		}
2692 	}
2693 out:
2694 	return err;
2695 }
2696 
2697 static int nlmsg_populate_fdb(struct sk_buff *skb,
2698 			      struct netlink_callback *cb,
2699 			      struct net_device *dev,
2700 			      int *idx,
2701 			      struct netdev_hw_addr_list *list)
2702 {
2703 	struct netdev_hw_addr *ha;
2704 	int err;
2705 	u32 portid, seq;
2706 
2707 	portid = NETLINK_CB(cb->skb).portid;
2708 	seq = cb->nlh->nlmsg_seq;
2709 
2710 	list_for_each_entry(ha, &list->list, list) {
2711 		if (*idx < cb->args[0])
2712 			goto skip;
2713 
2714 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2715 					      portid, seq,
2716 					      RTM_NEWNEIGH, NTF_SELF,
2717 					      NLM_F_MULTI);
2718 		if (err < 0)
2719 			return err;
2720 skip:
2721 		*idx += 1;
2722 	}
2723 	return 0;
2724 }
2725 
2726 /**
2727  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2728  * @nlh: netlink message header
2729  * @dev: netdevice
2730  *
2731  * Default netdevice operation to dump the existing unicast address list.
2732  * Returns number of addresses from list put in skb.
2733  */
2734 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2735 		      struct netlink_callback *cb,
2736 		      struct net_device *dev,
2737 		      struct net_device *filter_dev,
2738 		      int idx)
2739 {
2740 	int err;
2741 
2742 	netif_addr_lock_bh(dev);
2743 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2744 	if (err)
2745 		goto out;
2746 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2747 out:
2748 	netif_addr_unlock_bh(dev);
2749 	return idx;
2750 }
2751 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2752 
2753 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2754 {
2755 	struct net_device *dev;
2756 	struct nlattr *tb[IFLA_MAX+1];
2757 	struct net_device *bdev = NULL;
2758 	struct net_device *br_dev = NULL;
2759 	const struct net_device_ops *ops = NULL;
2760 	const struct net_device_ops *cops = NULL;
2761 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2762 	struct net *net = sock_net(skb->sk);
2763 	int brport_idx = 0;
2764 	int br_idx = 0;
2765 	int idx = 0;
2766 
2767 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2768 			ifla_policy) == 0) {
2769 		if (tb[IFLA_MASTER])
2770 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
2771 	}
2772 
2773 	brport_idx = ifm->ifi_index;
2774 
2775 	if (br_idx) {
2776 		br_dev = __dev_get_by_index(net, br_idx);
2777 		if (!br_dev)
2778 			return -ENODEV;
2779 
2780 		ops = br_dev->netdev_ops;
2781 		bdev = br_dev;
2782 	}
2783 
2784 	for_each_netdev(net, dev) {
2785 		if (brport_idx && (dev->ifindex != brport_idx))
2786 			continue;
2787 
2788 		if (!br_idx) { /* user did not specify a specific bridge */
2789 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
2790 				br_dev = netdev_master_upper_dev_get(dev);
2791 				cops = br_dev->netdev_ops;
2792 			}
2793 
2794 			bdev = dev;
2795 		} else {
2796 			if (dev != br_dev &&
2797 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
2798 				continue;
2799 
2800 			if (br_dev != netdev_master_upper_dev_get(dev) &&
2801 			    !(dev->priv_flags & IFF_EBRIDGE))
2802 				continue;
2803 
2804 			bdev = br_dev;
2805 			cops = ops;
2806 		}
2807 
2808 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2809 			if (cops && cops->ndo_fdb_dump)
2810 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2811 							 idx);
2812 		}
2813 
2814 		if (dev->netdev_ops->ndo_fdb_dump)
2815 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2816 							    idx);
2817 		else
2818 			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2819 
2820 		cops = NULL;
2821 	}
2822 
2823 	cb->args[0] = idx;
2824 	return skb->len;
2825 }
2826 
2827 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2828 			       unsigned int attrnum, unsigned int flag)
2829 {
2830 	if (mask & flag)
2831 		return nla_put_u8(skb, attrnum, !!(flags & flag));
2832 	return 0;
2833 }
2834 
2835 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2836 			    struct net_device *dev, u16 mode,
2837 			    u32 flags, u32 mask)
2838 {
2839 	struct nlmsghdr *nlh;
2840 	struct ifinfomsg *ifm;
2841 	struct nlattr *br_afspec;
2842 	struct nlattr *protinfo;
2843 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2844 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2845 
2846 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2847 	if (nlh == NULL)
2848 		return -EMSGSIZE;
2849 
2850 	ifm = nlmsg_data(nlh);
2851 	ifm->ifi_family = AF_BRIDGE;
2852 	ifm->__ifi_pad = 0;
2853 	ifm->ifi_type = dev->type;
2854 	ifm->ifi_index = dev->ifindex;
2855 	ifm->ifi_flags = dev_get_flags(dev);
2856 	ifm->ifi_change = 0;
2857 
2858 
2859 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2860 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2861 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2862 	    (br_dev &&
2863 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2864 	    (dev->addr_len &&
2865 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2866 	    (dev->ifindex != dev->iflink &&
2867 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2868 		goto nla_put_failure;
2869 
2870 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2871 	if (!br_afspec)
2872 		goto nla_put_failure;
2873 
2874 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2875 		nla_nest_cancel(skb, br_afspec);
2876 		goto nla_put_failure;
2877 	}
2878 
2879 	if (mode != BRIDGE_MODE_UNDEF) {
2880 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2881 			nla_nest_cancel(skb, br_afspec);
2882 			goto nla_put_failure;
2883 		}
2884 	}
2885 	nla_nest_end(skb, br_afspec);
2886 
2887 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2888 	if (!protinfo)
2889 		goto nla_put_failure;
2890 
2891 	if (brport_nla_put_flag(skb, flags, mask,
2892 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2893 	    brport_nla_put_flag(skb, flags, mask,
2894 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2895 	    brport_nla_put_flag(skb, flags, mask,
2896 				IFLA_BRPORT_FAST_LEAVE,
2897 				BR_MULTICAST_FAST_LEAVE) ||
2898 	    brport_nla_put_flag(skb, flags, mask,
2899 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2900 	    brport_nla_put_flag(skb, flags, mask,
2901 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2902 	    brport_nla_put_flag(skb, flags, mask,
2903 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2904 	    brport_nla_put_flag(skb, flags, mask,
2905 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2906 	    brport_nla_put_flag(skb, flags, mask,
2907 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2908 		nla_nest_cancel(skb, protinfo);
2909 		goto nla_put_failure;
2910 	}
2911 
2912 	nla_nest_end(skb, protinfo);
2913 
2914 	nlmsg_end(skb, nlh);
2915 	return 0;
2916 nla_put_failure:
2917 	nlmsg_cancel(skb, nlh);
2918 	return -EMSGSIZE;
2919 }
2920 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2921 
2922 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2923 {
2924 	struct net *net = sock_net(skb->sk);
2925 	struct net_device *dev;
2926 	int idx = 0;
2927 	u32 portid = NETLINK_CB(cb->skb).portid;
2928 	u32 seq = cb->nlh->nlmsg_seq;
2929 	u32 filter_mask = 0;
2930 
2931 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2932 		struct nlattr *extfilt;
2933 
2934 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2935 					  IFLA_EXT_MASK);
2936 		if (extfilt) {
2937 			if (nla_len(extfilt) < sizeof(filter_mask))
2938 				return -EINVAL;
2939 
2940 			filter_mask = nla_get_u32(extfilt);
2941 		}
2942 	}
2943 
2944 	rcu_read_lock();
2945 	for_each_netdev_rcu(net, dev) {
2946 		const struct net_device_ops *ops = dev->netdev_ops;
2947 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2948 
2949 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2950 			if (idx >= cb->args[0] &&
2951 			    br_dev->netdev_ops->ndo_bridge_getlink(
2952 				    skb, portid, seq, dev, filter_mask) < 0)
2953 				break;
2954 			idx++;
2955 		}
2956 
2957 		if (ops->ndo_bridge_getlink) {
2958 			if (idx >= cb->args[0] &&
2959 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2960 						    filter_mask) < 0)
2961 				break;
2962 			idx++;
2963 		}
2964 	}
2965 	rcu_read_unlock();
2966 	cb->args[0] = idx;
2967 
2968 	return skb->len;
2969 }
2970 
2971 static inline size_t bridge_nlmsg_size(void)
2972 {
2973 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2974 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2975 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2976 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2977 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2978 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2979 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2980 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2981 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2982 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2983 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2984 }
2985 
2986 static int rtnl_bridge_notify(struct net_device *dev)
2987 {
2988 	struct net *net = dev_net(dev);
2989 	struct sk_buff *skb;
2990 	int err = -EOPNOTSUPP;
2991 
2992 	if (!dev->netdev_ops->ndo_bridge_getlink)
2993 		return 0;
2994 
2995 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2996 	if (!skb) {
2997 		err = -ENOMEM;
2998 		goto errout;
2999 	}
3000 
3001 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
3002 	if (err < 0)
3003 		goto errout;
3004 
3005 	if (!skb->len)
3006 		goto errout;
3007 
3008 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3009 	return 0;
3010 errout:
3011 	WARN_ON(err == -EMSGSIZE);
3012 	kfree_skb(skb);
3013 	if (err)
3014 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3015 	return err;
3016 }
3017 
3018 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3019 {
3020 	struct net *net = sock_net(skb->sk);
3021 	struct ifinfomsg *ifm;
3022 	struct net_device *dev;
3023 	struct nlattr *br_spec, *attr = NULL;
3024 	int rem, err = -EOPNOTSUPP;
3025 	u16 flags = 0;
3026 	bool have_flags = false;
3027 
3028 	if (nlmsg_len(nlh) < sizeof(*ifm))
3029 		return -EINVAL;
3030 
3031 	ifm = nlmsg_data(nlh);
3032 	if (ifm->ifi_family != AF_BRIDGE)
3033 		return -EPFNOSUPPORT;
3034 
3035 	dev = __dev_get_by_index(net, ifm->ifi_index);
3036 	if (!dev) {
3037 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3038 		return -ENODEV;
3039 	}
3040 
3041 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3042 	if (br_spec) {
3043 		nla_for_each_nested(attr, br_spec, rem) {
3044 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3045 				if (nla_len(attr) < sizeof(flags))
3046 					return -EINVAL;
3047 
3048 				have_flags = true;
3049 				flags = nla_get_u16(attr);
3050 				break;
3051 			}
3052 		}
3053 	}
3054 
3055 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3056 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3057 
3058 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3059 			err = -EOPNOTSUPP;
3060 			goto out;
3061 		}
3062 
3063 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3064 		if (err)
3065 			goto out;
3066 
3067 		flags &= ~BRIDGE_FLAGS_MASTER;
3068 	}
3069 
3070 	if ((flags & BRIDGE_FLAGS_SELF)) {
3071 		if (!dev->netdev_ops->ndo_bridge_setlink)
3072 			err = -EOPNOTSUPP;
3073 		else
3074 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3075 								  flags);
3076 		if (!err) {
3077 			flags &= ~BRIDGE_FLAGS_SELF;
3078 
3079 			/* Generate event to notify upper layer of bridge
3080 			 * change
3081 			 */
3082 			err = rtnl_bridge_notify(dev);
3083 		}
3084 	}
3085 
3086 	if (have_flags)
3087 		memcpy(nla_data(attr), &flags, sizeof(flags));
3088 out:
3089 	return err;
3090 }
3091 
3092 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3093 {
3094 	struct net *net = sock_net(skb->sk);
3095 	struct ifinfomsg *ifm;
3096 	struct net_device *dev;
3097 	struct nlattr *br_spec, *attr = NULL;
3098 	int rem, err = -EOPNOTSUPP;
3099 	u16 flags = 0;
3100 	bool have_flags = false;
3101 
3102 	if (nlmsg_len(nlh) < sizeof(*ifm))
3103 		return -EINVAL;
3104 
3105 	ifm = nlmsg_data(nlh);
3106 	if (ifm->ifi_family != AF_BRIDGE)
3107 		return -EPFNOSUPPORT;
3108 
3109 	dev = __dev_get_by_index(net, ifm->ifi_index);
3110 	if (!dev) {
3111 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3112 		return -ENODEV;
3113 	}
3114 
3115 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3116 	if (br_spec) {
3117 		nla_for_each_nested(attr, br_spec, rem) {
3118 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3119 				if (nla_len(attr) < sizeof(flags))
3120 					return -EINVAL;
3121 
3122 				have_flags = true;
3123 				flags = nla_get_u16(attr);
3124 				break;
3125 			}
3126 		}
3127 	}
3128 
3129 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3130 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3131 
3132 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3133 			err = -EOPNOTSUPP;
3134 			goto out;
3135 		}
3136 
3137 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3138 		if (err)
3139 			goto out;
3140 
3141 		flags &= ~BRIDGE_FLAGS_MASTER;
3142 	}
3143 
3144 	if ((flags & BRIDGE_FLAGS_SELF)) {
3145 		if (!dev->netdev_ops->ndo_bridge_dellink)
3146 			err = -EOPNOTSUPP;
3147 		else
3148 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3149 								  flags);
3150 
3151 		if (!err) {
3152 			flags &= ~BRIDGE_FLAGS_SELF;
3153 
3154 			/* Generate event to notify upper layer of bridge
3155 			 * change
3156 			 */
3157 			err = rtnl_bridge_notify(dev);
3158 		}
3159 	}
3160 
3161 	if (have_flags)
3162 		memcpy(nla_data(attr), &flags, sizeof(flags));
3163 out:
3164 	return err;
3165 }
3166 
3167 /* Process one rtnetlink message. */
3168 
3169 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3170 {
3171 	struct net *net = sock_net(skb->sk);
3172 	rtnl_doit_func doit;
3173 	int sz_idx, kind;
3174 	int family;
3175 	int type;
3176 	int err;
3177 
3178 	type = nlh->nlmsg_type;
3179 	if (type > RTM_MAX)
3180 		return -EOPNOTSUPP;
3181 
3182 	type -= RTM_BASE;
3183 
3184 	/* All the messages must have at least 1 byte length */
3185 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3186 		return 0;
3187 
3188 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3189 	sz_idx = type>>2;
3190 	kind = type&3;
3191 
3192 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3193 		return -EPERM;
3194 
3195 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3196 		struct sock *rtnl;
3197 		rtnl_dumpit_func dumpit;
3198 		rtnl_calcit_func calcit;
3199 		u16 min_dump_alloc = 0;
3200 
3201 		dumpit = rtnl_get_dumpit(family, type);
3202 		if (dumpit == NULL)
3203 			return -EOPNOTSUPP;
3204 		calcit = rtnl_get_calcit(family, type);
3205 		if (calcit)
3206 			min_dump_alloc = calcit(skb, nlh);
3207 
3208 		__rtnl_unlock();
3209 		rtnl = net->rtnl;
3210 		{
3211 			struct netlink_dump_control c = {
3212 				.dump		= dumpit,
3213 				.min_dump_alloc	= min_dump_alloc,
3214 			};
3215 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3216 		}
3217 		rtnl_lock();
3218 		return err;
3219 	}
3220 
3221 	doit = rtnl_get_doit(family, type);
3222 	if (doit == NULL)
3223 		return -EOPNOTSUPP;
3224 
3225 	return doit(skb, nlh);
3226 }
3227 
3228 static void rtnetlink_rcv(struct sk_buff *skb)
3229 {
3230 	rtnl_lock();
3231 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3232 	rtnl_unlock();
3233 }
3234 
3235 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3236 {
3237 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3238 
3239 	switch (event) {
3240 	case NETDEV_UP:
3241 	case NETDEV_DOWN:
3242 	case NETDEV_PRE_UP:
3243 	case NETDEV_POST_INIT:
3244 	case NETDEV_REGISTER:
3245 	case NETDEV_CHANGE:
3246 	case NETDEV_PRE_TYPE_CHANGE:
3247 	case NETDEV_GOING_DOWN:
3248 	case NETDEV_UNREGISTER:
3249 	case NETDEV_UNREGISTER_FINAL:
3250 	case NETDEV_RELEASE:
3251 	case NETDEV_JOIN:
3252 	case NETDEV_BONDING_INFO:
3253 		break;
3254 	default:
3255 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3256 		break;
3257 	}
3258 	return NOTIFY_DONE;
3259 }
3260 
3261 static struct notifier_block rtnetlink_dev_notifier = {
3262 	.notifier_call	= rtnetlink_event,
3263 };
3264 
3265 
3266 static int __net_init rtnetlink_net_init(struct net *net)
3267 {
3268 	struct sock *sk;
3269 	struct netlink_kernel_cfg cfg = {
3270 		.groups		= RTNLGRP_MAX,
3271 		.input		= rtnetlink_rcv,
3272 		.cb_mutex	= &rtnl_mutex,
3273 		.flags		= NL_CFG_F_NONROOT_RECV,
3274 	};
3275 
3276 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3277 	if (!sk)
3278 		return -ENOMEM;
3279 	net->rtnl = sk;
3280 	return 0;
3281 }
3282 
3283 static void __net_exit rtnetlink_net_exit(struct net *net)
3284 {
3285 	netlink_kernel_release(net->rtnl);
3286 	net->rtnl = NULL;
3287 }
3288 
3289 static struct pernet_operations rtnetlink_net_ops = {
3290 	.init = rtnetlink_net_init,
3291 	.exit = rtnetlink_net_exit,
3292 };
3293 
3294 void __init rtnetlink_init(void)
3295 {
3296 	if (register_pernet_subsys(&rtnetlink_net_ops))
3297 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3298 
3299 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3300 
3301 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3302 		      rtnl_dump_ifinfo, rtnl_calcit);
3303 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3304 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3305 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3306 
3307 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3308 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3309 
3310 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3311 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3312 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3313 
3314 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3315 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3316 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3317 }
3318 
3319