xref: /linux/net/core/rtnetlink.c (revision 005438a8eef063495ac059d128eea71b58de50e5)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  *
15  *	Fixes:
16  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17  */
18 
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
29 #include <linux/mm.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
42 
43 #include <asm/uaccess.h>
44 
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
48 #include <net/ip.h>
49 #include <net/protocol.h>
50 #include <net/arp.h>
51 #include <net/route.h>
52 #include <net/udp.h>
53 #include <net/tcp.h>
54 #include <net/sock.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
59 
60 struct rtnl_link {
61 	rtnl_doit_func		doit;
62 	rtnl_dumpit_func	dumpit;
63 	rtnl_calcit_func 	calcit;
64 };
65 
66 static DEFINE_MUTEX(rtnl_mutex);
67 
68 void rtnl_lock(void)
69 {
70 	mutex_lock(&rtnl_mutex);
71 }
72 EXPORT_SYMBOL(rtnl_lock);
73 
74 void __rtnl_unlock(void)
75 {
76 	mutex_unlock(&rtnl_mutex);
77 }
78 
79 void rtnl_unlock(void)
80 {
81 	/* This fellow will unlock it for us. */
82 	netdev_run_todo();
83 }
84 EXPORT_SYMBOL(rtnl_unlock);
85 
86 int rtnl_trylock(void)
87 {
88 	return mutex_trylock(&rtnl_mutex);
89 }
90 EXPORT_SYMBOL(rtnl_trylock);
91 
92 int rtnl_is_locked(void)
93 {
94 	return mutex_is_locked(&rtnl_mutex);
95 }
96 EXPORT_SYMBOL(rtnl_is_locked);
97 
98 #ifdef CONFIG_PROVE_LOCKING
99 int lockdep_rtnl_is_held(void)
100 {
101 	return lockdep_is_held(&rtnl_mutex);
102 }
103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
105 
106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
107 
108 static inline int rtm_msgindex(int msgtype)
109 {
110 	int msgindex = msgtype - RTM_BASE;
111 
112 	/*
113 	 * msgindex < 0 implies someone tried to register a netlink
114 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
115 	 * the message type has not been added to linux/rtnetlink.h
116 	 */
117 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
118 
119 	return msgindex;
120 }
121 
122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
123 {
124 	struct rtnl_link *tab;
125 
126 	if (protocol <= RTNL_FAMILY_MAX)
127 		tab = rtnl_msg_handlers[protocol];
128 	else
129 		tab = NULL;
130 
131 	if (tab == NULL || tab[msgindex].doit == NULL)
132 		tab = rtnl_msg_handlers[PF_UNSPEC];
133 
134 	return tab[msgindex].doit;
135 }
136 
137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
138 {
139 	struct rtnl_link *tab;
140 
141 	if (protocol <= RTNL_FAMILY_MAX)
142 		tab = rtnl_msg_handlers[protocol];
143 	else
144 		tab = NULL;
145 
146 	if (tab == NULL || tab[msgindex].dumpit == NULL)
147 		tab = rtnl_msg_handlers[PF_UNSPEC];
148 
149 	return tab[msgindex].dumpit;
150 }
151 
152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
153 {
154 	struct rtnl_link *tab;
155 
156 	if (protocol <= RTNL_FAMILY_MAX)
157 		tab = rtnl_msg_handlers[protocol];
158 	else
159 		tab = NULL;
160 
161 	if (tab == NULL || tab[msgindex].calcit == NULL)
162 		tab = rtnl_msg_handlers[PF_UNSPEC];
163 
164 	return tab[msgindex].calcit;
165 }
166 
167 /**
168  * __rtnl_register - Register a rtnetlink message type
169  * @protocol: Protocol family or PF_UNSPEC
170  * @msgtype: rtnetlink message type
171  * @doit: Function pointer called for each request message
172  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
173  * @calcit: Function pointer to calc size of dump message
174  *
175  * Registers the specified function pointers (at least one of them has
176  * to be non-NULL) to be called whenever a request message for the
177  * specified protocol family and message type is received.
178  *
179  * The special protocol family PF_UNSPEC may be used to define fallback
180  * function pointers for the case when no entry for the specific protocol
181  * family exists.
182  *
183  * Returns 0 on success or a negative error code.
184  */
185 int __rtnl_register(int protocol, int msgtype,
186 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
187 		    rtnl_calcit_func calcit)
188 {
189 	struct rtnl_link *tab;
190 	int msgindex;
191 
192 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
193 	msgindex = rtm_msgindex(msgtype);
194 
195 	tab = rtnl_msg_handlers[protocol];
196 	if (tab == NULL) {
197 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
198 		if (tab == NULL)
199 			return -ENOBUFS;
200 
201 		rtnl_msg_handlers[protocol] = tab;
202 	}
203 
204 	if (doit)
205 		tab[msgindex].doit = doit;
206 
207 	if (dumpit)
208 		tab[msgindex].dumpit = dumpit;
209 
210 	if (calcit)
211 		tab[msgindex].calcit = calcit;
212 
213 	return 0;
214 }
215 EXPORT_SYMBOL_GPL(__rtnl_register);
216 
217 /**
218  * rtnl_register - Register a rtnetlink message type
219  *
220  * Identical to __rtnl_register() but panics on failure. This is useful
221  * as failure of this function is very unlikely, it can only happen due
222  * to lack of memory when allocating the chain to store all message
223  * handlers for a protocol. Meant for use in init functions where lack
224  * of memory implies no sense in continuing.
225  */
226 void rtnl_register(int protocol, int msgtype,
227 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
228 		   rtnl_calcit_func calcit)
229 {
230 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
231 		panic("Unable to register rtnetlink message handler, "
232 		      "protocol = %d, message type = %d\n",
233 		      protocol, msgtype);
234 }
235 EXPORT_SYMBOL_GPL(rtnl_register);
236 
237 /**
238  * rtnl_unregister - Unregister a rtnetlink message type
239  * @protocol: Protocol family or PF_UNSPEC
240  * @msgtype: rtnetlink message type
241  *
242  * Returns 0 on success or a negative error code.
243  */
244 int rtnl_unregister(int protocol, int msgtype)
245 {
246 	int msgindex;
247 
248 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
249 	msgindex = rtm_msgindex(msgtype);
250 
251 	if (rtnl_msg_handlers[protocol] == NULL)
252 		return -ENOENT;
253 
254 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
255 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
256 
257 	return 0;
258 }
259 EXPORT_SYMBOL_GPL(rtnl_unregister);
260 
261 /**
262  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
263  * @protocol : Protocol family or PF_UNSPEC
264  *
265  * Identical to calling rtnl_unregster() for all registered message types
266  * of a certain protocol family.
267  */
268 void rtnl_unregister_all(int protocol)
269 {
270 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 
272 	kfree(rtnl_msg_handlers[protocol]);
273 	rtnl_msg_handlers[protocol] = NULL;
274 }
275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
276 
277 static LIST_HEAD(link_ops);
278 
279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
280 {
281 	const struct rtnl_link_ops *ops;
282 
283 	list_for_each_entry(ops, &link_ops, list) {
284 		if (!strcmp(ops->kind, kind))
285 			return ops;
286 	}
287 	return NULL;
288 }
289 
290 /**
291  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
292  * @ops: struct rtnl_link_ops * to register
293  *
294  * The caller must hold the rtnl_mutex. This function should be used
295  * by drivers that create devices during module initialization. It
296  * must be called before registering the devices.
297  *
298  * Returns 0 on success or a negative error code.
299  */
300 int __rtnl_link_register(struct rtnl_link_ops *ops)
301 {
302 	if (rtnl_link_ops_get(ops->kind))
303 		return -EEXIST;
304 
305 	/* The check for setup is here because if ops
306 	 * does not have that filled up, it is not possible
307 	 * to use the ops for creating device. So do not
308 	 * fill up dellink as well. That disables rtnl_dellink.
309 	 */
310 	if (ops->setup && !ops->dellink)
311 		ops->dellink = unregister_netdevice_queue;
312 
313 	list_add_tail(&ops->list, &link_ops);
314 	return 0;
315 }
316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
317 
318 /**
319  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
320  * @ops: struct rtnl_link_ops * to register
321  *
322  * Returns 0 on success or a negative error code.
323  */
324 int rtnl_link_register(struct rtnl_link_ops *ops)
325 {
326 	int err;
327 
328 	rtnl_lock();
329 	err = __rtnl_link_register(ops);
330 	rtnl_unlock();
331 	return err;
332 }
333 EXPORT_SYMBOL_GPL(rtnl_link_register);
334 
335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
336 {
337 	struct net_device *dev;
338 	LIST_HEAD(list_kill);
339 
340 	for_each_netdev(net, dev) {
341 		if (dev->rtnl_link_ops == ops)
342 			ops->dellink(dev, &list_kill);
343 	}
344 	unregister_netdevice_many(&list_kill);
345 }
346 
347 /**
348  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
349  * @ops: struct rtnl_link_ops * to unregister
350  *
351  * The caller must hold the rtnl_mutex.
352  */
353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
354 {
355 	struct net *net;
356 
357 	for_each_net(net) {
358 		__rtnl_kill_links(net, ops);
359 	}
360 	list_del(&ops->list);
361 }
362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
363 
364 /* Return with the rtnl_lock held when there are no network
365  * devices unregistering in any network namespace.
366  */
367 static void rtnl_lock_unregistering_all(void)
368 {
369 	struct net *net;
370 	bool unregistering;
371 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
372 
373 	add_wait_queue(&netdev_unregistering_wq, &wait);
374 	for (;;) {
375 		unregistering = false;
376 		rtnl_lock();
377 		for_each_net(net) {
378 			if (net->dev_unreg_count > 0) {
379 				unregistering = true;
380 				break;
381 			}
382 		}
383 		if (!unregistering)
384 			break;
385 		__rtnl_unlock();
386 
387 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
388 	}
389 	remove_wait_queue(&netdev_unregistering_wq, &wait);
390 }
391 
392 /**
393  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
394  * @ops: struct rtnl_link_ops * to unregister
395  */
396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
397 {
398 	/* Close the race with cleanup_net() */
399 	mutex_lock(&net_mutex);
400 	rtnl_lock_unregistering_all();
401 	__rtnl_link_unregister(ops);
402 	rtnl_unlock();
403 	mutex_unlock(&net_mutex);
404 }
405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
406 
407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
408 {
409 	struct net_device *master_dev;
410 	const struct rtnl_link_ops *ops;
411 
412 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
413 	if (!master_dev)
414 		return 0;
415 	ops = master_dev->rtnl_link_ops;
416 	if (!ops || !ops->get_slave_size)
417 		return 0;
418 	/* IFLA_INFO_SLAVE_DATA + nested data */
419 	return nla_total_size(sizeof(struct nlattr)) +
420 	       ops->get_slave_size(master_dev, dev);
421 }
422 
423 static size_t rtnl_link_get_size(const struct net_device *dev)
424 {
425 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
426 	size_t size;
427 
428 	if (!ops)
429 		return 0;
430 
431 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
432 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
433 
434 	if (ops->get_size)
435 		/* IFLA_INFO_DATA + nested data */
436 		size += nla_total_size(sizeof(struct nlattr)) +
437 			ops->get_size(dev);
438 
439 	if (ops->get_xstats_size)
440 		/* IFLA_INFO_XSTATS */
441 		size += nla_total_size(ops->get_xstats_size(dev));
442 
443 	size += rtnl_link_get_slave_info_data_size(dev);
444 
445 	return size;
446 }
447 
448 static LIST_HEAD(rtnl_af_ops);
449 
450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
451 {
452 	const struct rtnl_af_ops *ops;
453 
454 	list_for_each_entry(ops, &rtnl_af_ops, list) {
455 		if (ops->family == family)
456 			return ops;
457 	}
458 
459 	return NULL;
460 }
461 
462 /**
463  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
464  * @ops: struct rtnl_af_ops * to register
465  *
466  * Returns 0 on success or a negative error code.
467  */
468 void rtnl_af_register(struct rtnl_af_ops *ops)
469 {
470 	rtnl_lock();
471 	list_add_tail(&ops->list, &rtnl_af_ops);
472 	rtnl_unlock();
473 }
474 EXPORT_SYMBOL_GPL(rtnl_af_register);
475 
476 /**
477  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
478  * @ops: struct rtnl_af_ops * to unregister
479  *
480  * The caller must hold the rtnl_mutex.
481  */
482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
483 {
484 	list_del(&ops->list);
485 }
486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
487 
488 /**
489  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
490  * @ops: struct rtnl_af_ops * to unregister
491  */
492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
493 {
494 	rtnl_lock();
495 	__rtnl_af_unregister(ops);
496 	rtnl_unlock();
497 }
498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
499 
500 static size_t rtnl_link_get_af_size(const struct net_device *dev)
501 {
502 	struct rtnl_af_ops *af_ops;
503 	size_t size;
504 
505 	/* IFLA_AF_SPEC */
506 	size = nla_total_size(sizeof(struct nlattr));
507 
508 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
509 		if (af_ops->get_link_af_size) {
510 			/* AF_* + nested data */
511 			size += nla_total_size(sizeof(struct nlattr)) +
512 				af_ops->get_link_af_size(dev);
513 		}
514 	}
515 
516 	return size;
517 }
518 
519 static bool rtnl_have_link_slave_info(const struct net_device *dev)
520 {
521 	struct net_device *master_dev;
522 
523 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
524 	if (master_dev && master_dev->rtnl_link_ops)
525 		return true;
526 	return false;
527 }
528 
529 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
530 				     const struct net_device *dev)
531 {
532 	struct net_device *master_dev;
533 	const struct rtnl_link_ops *ops;
534 	struct nlattr *slave_data;
535 	int err;
536 
537 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
538 	if (!master_dev)
539 		return 0;
540 	ops = master_dev->rtnl_link_ops;
541 	if (!ops)
542 		return 0;
543 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
544 		return -EMSGSIZE;
545 	if (ops->fill_slave_info) {
546 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
547 		if (!slave_data)
548 			return -EMSGSIZE;
549 		err = ops->fill_slave_info(skb, master_dev, dev);
550 		if (err < 0)
551 			goto err_cancel_slave_data;
552 		nla_nest_end(skb, slave_data);
553 	}
554 	return 0;
555 
556 err_cancel_slave_data:
557 	nla_nest_cancel(skb, slave_data);
558 	return err;
559 }
560 
561 static int rtnl_link_info_fill(struct sk_buff *skb,
562 			       const struct net_device *dev)
563 {
564 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
565 	struct nlattr *data;
566 	int err;
567 
568 	if (!ops)
569 		return 0;
570 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
571 		return -EMSGSIZE;
572 	if (ops->fill_xstats) {
573 		err = ops->fill_xstats(skb, dev);
574 		if (err < 0)
575 			return err;
576 	}
577 	if (ops->fill_info) {
578 		data = nla_nest_start(skb, IFLA_INFO_DATA);
579 		if (data == NULL)
580 			return -EMSGSIZE;
581 		err = ops->fill_info(skb, dev);
582 		if (err < 0)
583 			goto err_cancel_data;
584 		nla_nest_end(skb, data);
585 	}
586 	return 0;
587 
588 err_cancel_data:
589 	nla_nest_cancel(skb, data);
590 	return err;
591 }
592 
593 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
594 {
595 	struct nlattr *linkinfo;
596 	int err = -EMSGSIZE;
597 
598 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
599 	if (linkinfo == NULL)
600 		goto out;
601 
602 	err = rtnl_link_info_fill(skb, dev);
603 	if (err < 0)
604 		goto err_cancel_link;
605 
606 	err = rtnl_link_slave_info_fill(skb, dev);
607 	if (err < 0)
608 		goto err_cancel_link;
609 
610 	nla_nest_end(skb, linkinfo);
611 	return 0;
612 
613 err_cancel_link:
614 	nla_nest_cancel(skb, linkinfo);
615 out:
616 	return err;
617 }
618 
619 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
620 {
621 	struct sock *rtnl = net->rtnl;
622 	int err = 0;
623 
624 	NETLINK_CB(skb).dst_group = group;
625 	if (echo)
626 		atomic_inc(&skb->users);
627 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
628 	if (echo)
629 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
630 	return err;
631 }
632 
633 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
634 {
635 	struct sock *rtnl = net->rtnl;
636 
637 	return nlmsg_unicast(rtnl, skb, pid);
638 }
639 EXPORT_SYMBOL(rtnl_unicast);
640 
641 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
642 		 struct nlmsghdr *nlh, gfp_t flags)
643 {
644 	struct sock *rtnl = net->rtnl;
645 	int report = 0;
646 
647 	if (nlh)
648 		report = nlmsg_report(nlh);
649 
650 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
651 }
652 EXPORT_SYMBOL(rtnl_notify);
653 
654 void rtnl_set_sk_err(struct net *net, u32 group, int error)
655 {
656 	struct sock *rtnl = net->rtnl;
657 
658 	netlink_set_err(rtnl, 0, group, error);
659 }
660 EXPORT_SYMBOL(rtnl_set_sk_err);
661 
662 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
663 {
664 	struct nlattr *mx;
665 	int i, valid = 0;
666 
667 	mx = nla_nest_start(skb, RTA_METRICS);
668 	if (mx == NULL)
669 		return -ENOBUFS;
670 
671 	for (i = 0; i < RTAX_MAX; i++) {
672 		if (metrics[i]) {
673 			if (i == RTAX_CC_ALGO - 1) {
674 				char tmp[TCP_CA_NAME_MAX], *name;
675 
676 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
677 				if (!name)
678 					continue;
679 				if (nla_put_string(skb, i + 1, name))
680 					goto nla_put_failure;
681 			} else {
682 				if (nla_put_u32(skb, i + 1, metrics[i]))
683 					goto nla_put_failure;
684 			}
685 			valid++;
686 		}
687 	}
688 
689 	if (!valid) {
690 		nla_nest_cancel(skb, mx);
691 		return 0;
692 	}
693 
694 	return nla_nest_end(skb, mx);
695 
696 nla_put_failure:
697 	nla_nest_cancel(skb, mx);
698 	return -EMSGSIZE;
699 }
700 EXPORT_SYMBOL(rtnetlink_put_metrics);
701 
702 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
703 		       long expires, u32 error)
704 {
705 	struct rta_cacheinfo ci = {
706 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
707 		.rta_used = dst->__use,
708 		.rta_clntref = atomic_read(&(dst->__refcnt)),
709 		.rta_error = error,
710 		.rta_id =  id,
711 	};
712 
713 	if (expires) {
714 		unsigned long clock;
715 
716 		clock = jiffies_to_clock_t(abs(expires));
717 		clock = min_t(unsigned long, clock, INT_MAX);
718 		ci.rta_expires = (expires > 0) ? clock : -clock;
719 	}
720 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
721 }
722 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
723 
724 static void set_operstate(struct net_device *dev, unsigned char transition)
725 {
726 	unsigned char operstate = dev->operstate;
727 
728 	switch (transition) {
729 	case IF_OPER_UP:
730 		if ((operstate == IF_OPER_DORMANT ||
731 		     operstate == IF_OPER_UNKNOWN) &&
732 		    !netif_dormant(dev))
733 			operstate = IF_OPER_UP;
734 		break;
735 
736 	case IF_OPER_DORMANT:
737 		if (operstate == IF_OPER_UP ||
738 		    operstate == IF_OPER_UNKNOWN)
739 			operstate = IF_OPER_DORMANT;
740 		break;
741 	}
742 
743 	if (dev->operstate != operstate) {
744 		write_lock_bh(&dev_base_lock);
745 		dev->operstate = operstate;
746 		write_unlock_bh(&dev_base_lock);
747 		netdev_state_change(dev);
748 	}
749 }
750 
751 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
752 {
753 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
754 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
755 }
756 
757 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
758 					   const struct ifinfomsg *ifm)
759 {
760 	unsigned int flags = ifm->ifi_flags;
761 
762 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
763 	if (ifm->ifi_change)
764 		flags = (flags & ifm->ifi_change) |
765 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
766 
767 	return flags;
768 }
769 
770 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
771 				 const struct rtnl_link_stats64 *b)
772 {
773 	a->rx_packets = b->rx_packets;
774 	a->tx_packets = b->tx_packets;
775 	a->rx_bytes = b->rx_bytes;
776 	a->tx_bytes = b->tx_bytes;
777 	a->rx_errors = b->rx_errors;
778 	a->tx_errors = b->tx_errors;
779 	a->rx_dropped = b->rx_dropped;
780 	a->tx_dropped = b->tx_dropped;
781 
782 	a->multicast = b->multicast;
783 	a->collisions = b->collisions;
784 
785 	a->rx_length_errors = b->rx_length_errors;
786 	a->rx_over_errors = b->rx_over_errors;
787 	a->rx_crc_errors = b->rx_crc_errors;
788 	a->rx_frame_errors = b->rx_frame_errors;
789 	a->rx_fifo_errors = b->rx_fifo_errors;
790 	a->rx_missed_errors = b->rx_missed_errors;
791 
792 	a->tx_aborted_errors = b->tx_aborted_errors;
793 	a->tx_carrier_errors = b->tx_carrier_errors;
794 	a->tx_fifo_errors = b->tx_fifo_errors;
795 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
796 	a->tx_window_errors = b->tx_window_errors;
797 
798 	a->rx_compressed = b->rx_compressed;
799 	a->tx_compressed = b->tx_compressed;
800 }
801 
802 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
803 {
804 	memcpy(v, b, sizeof(*b));
805 }
806 
807 /* All VF info */
808 static inline int rtnl_vfinfo_size(const struct net_device *dev,
809 				   u32 ext_filter_mask)
810 {
811 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
812 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
813 		int num_vfs = dev_num_vf(dev->dev.parent);
814 		size_t size = nla_total_size(sizeof(struct nlattr));
815 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
816 		size += num_vfs *
817 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
818 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
819 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
820 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
821 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
822 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
823 			 /* IFLA_VF_STATS_RX_PACKETS */
824 			 nla_total_size(sizeof(__u64)) +
825 			 /* IFLA_VF_STATS_TX_PACKETS */
826 			 nla_total_size(sizeof(__u64)) +
827 			 /* IFLA_VF_STATS_RX_BYTES */
828 			 nla_total_size(sizeof(__u64)) +
829 			 /* IFLA_VF_STATS_TX_BYTES */
830 			 nla_total_size(sizeof(__u64)) +
831 			 /* IFLA_VF_STATS_BROADCAST */
832 			 nla_total_size(sizeof(__u64)) +
833 			 /* IFLA_VF_STATS_MULTICAST */
834 			 nla_total_size(sizeof(__u64)));
835 		return size;
836 	} else
837 		return 0;
838 }
839 
840 static size_t rtnl_port_size(const struct net_device *dev,
841 			     u32 ext_filter_mask)
842 {
843 	size_t port_size = nla_total_size(4)		/* PORT_VF */
844 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
845 		+ nla_total_size(sizeof(struct ifla_port_vsi))
846 							/* PORT_VSI_TYPE */
847 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
848 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
849 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
850 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
851 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
852 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
853 		+ port_size;
854 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
855 		+ port_size;
856 
857 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
858 	    !(ext_filter_mask & RTEXT_FILTER_VF))
859 		return 0;
860 	if (dev_num_vf(dev->dev.parent))
861 		return port_self_size + vf_ports_size +
862 			vf_port_size * dev_num_vf(dev->dev.parent);
863 	else
864 		return port_self_size;
865 }
866 
867 static noinline size_t if_nlmsg_size(const struct net_device *dev,
868 				     u32 ext_filter_mask)
869 {
870 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
871 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
872 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
873 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
874 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
875 	       + nla_total_size(sizeof(struct rtnl_link_stats))
876 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
877 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
878 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
879 	       + nla_total_size(4) /* IFLA_TXQLEN */
880 	       + nla_total_size(4) /* IFLA_WEIGHT */
881 	       + nla_total_size(4) /* IFLA_MTU */
882 	       + nla_total_size(4) /* IFLA_LINK */
883 	       + nla_total_size(4) /* IFLA_MASTER */
884 	       + nla_total_size(1) /* IFLA_CARRIER */
885 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
886 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
887 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
888 	       + nla_total_size(1) /* IFLA_OPERSTATE */
889 	       + nla_total_size(1) /* IFLA_LINKMODE */
890 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
891 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
892 	       + nla_total_size(ext_filter_mask
893 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
894 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
895 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
896 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
897 	       + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
898 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
899 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
900 }
901 
902 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
903 {
904 	struct nlattr *vf_ports;
905 	struct nlattr *vf_port;
906 	int vf;
907 	int err;
908 
909 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
910 	if (!vf_ports)
911 		return -EMSGSIZE;
912 
913 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
914 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
915 		if (!vf_port)
916 			goto nla_put_failure;
917 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
918 			goto nla_put_failure;
919 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
920 		if (err == -EMSGSIZE)
921 			goto nla_put_failure;
922 		if (err) {
923 			nla_nest_cancel(skb, vf_port);
924 			continue;
925 		}
926 		nla_nest_end(skb, vf_port);
927 	}
928 
929 	nla_nest_end(skb, vf_ports);
930 
931 	return 0;
932 
933 nla_put_failure:
934 	nla_nest_cancel(skb, vf_ports);
935 	return -EMSGSIZE;
936 }
937 
938 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
939 {
940 	struct nlattr *port_self;
941 	int err;
942 
943 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
944 	if (!port_self)
945 		return -EMSGSIZE;
946 
947 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
948 	if (err) {
949 		nla_nest_cancel(skb, port_self);
950 		return (err == -EMSGSIZE) ? err : 0;
951 	}
952 
953 	nla_nest_end(skb, port_self);
954 
955 	return 0;
956 }
957 
958 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
959 			  u32 ext_filter_mask)
960 {
961 	int err;
962 
963 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
964 	    !(ext_filter_mask & RTEXT_FILTER_VF))
965 		return 0;
966 
967 	err = rtnl_port_self_fill(skb, dev);
968 	if (err)
969 		return err;
970 
971 	if (dev_num_vf(dev->dev.parent)) {
972 		err = rtnl_vf_ports_fill(skb, dev);
973 		if (err)
974 			return err;
975 	}
976 
977 	return 0;
978 }
979 
980 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
981 {
982 	int err;
983 	struct netdev_phys_item_id ppid;
984 
985 	err = dev_get_phys_port_id(dev, &ppid);
986 	if (err) {
987 		if (err == -EOPNOTSUPP)
988 			return 0;
989 		return err;
990 	}
991 
992 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
993 		return -EMSGSIZE;
994 
995 	return 0;
996 }
997 
998 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
999 {
1000 	char name[IFNAMSIZ];
1001 	int err;
1002 
1003 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1004 	if (err) {
1005 		if (err == -EOPNOTSUPP)
1006 			return 0;
1007 		return err;
1008 	}
1009 
1010 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1011 		return -EMSGSIZE;
1012 
1013 	return 0;
1014 }
1015 
1016 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1017 {
1018 	int err;
1019 	struct switchdev_attr attr = {
1020 		.id = SWITCHDEV_ATTR_PORT_PARENT_ID,
1021 		.flags = SWITCHDEV_F_NO_RECURSE,
1022 	};
1023 
1024 	err = switchdev_port_attr_get(dev, &attr);
1025 	if (err) {
1026 		if (err == -EOPNOTSUPP)
1027 			return 0;
1028 		return err;
1029 	}
1030 
1031 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1032 		    attr.u.ppid.id))
1033 		return -EMSGSIZE;
1034 
1035 	return 0;
1036 }
1037 
1038 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1039 			    int type, u32 pid, u32 seq, u32 change,
1040 			    unsigned int flags, u32 ext_filter_mask)
1041 {
1042 	struct ifinfomsg *ifm;
1043 	struct nlmsghdr *nlh;
1044 	struct rtnl_link_stats64 temp;
1045 	const struct rtnl_link_stats64 *stats;
1046 	struct nlattr *attr, *af_spec;
1047 	struct rtnl_af_ops *af_ops;
1048 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1049 
1050 	ASSERT_RTNL();
1051 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1052 	if (nlh == NULL)
1053 		return -EMSGSIZE;
1054 
1055 	ifm = nlmsg_data(nlh);
1056 	ifm->ifi_family = AF_UNSPEC;
1057 	ifm->__ifi_pad = 0;
1058 	ifm->ifi_type = dev->type;
1059 	ifm->ifi_index = dev->ifindex;
1060 	ifm->ifi_flags = dev_get_flags(dev);
1061 	ifm->ifi_change = change;
1062 
1063 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1064 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1065 	    nla_put_u8(skb, IFLA_OPERSTATE,
1066 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1067 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1068 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1069 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1070 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1071 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1072 #ifdef CONFIG_RPS
1073 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1074 #endif
1075 	    (dev->ifindex != dev_get_iflink(dev) &&
1076 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1077 	    (upper_dev &&
1078 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1079 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1080 	    (dev->qdisc &&
1081 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1082 	    (dev->ifalias &&
1083 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1084 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1085 			atomic_read(&dev->carrier_changes)))
1086 		goto nla_put_failure;
1087 
1088 	if (1) {
1089 		struct rtnl_link_ifmap map = {
1090 			.mem_start   = dev->mem_start,
1091 			.mem_end     = dev->mem_end,
1092 			.base_addr   = dev->base_addr,
1093 			.irq         = dev->irq,
1094 			.dma         = dev->dma,
1095 			.port        = dev->if_port,
1096 		};
1097 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1098 			goto nla_put_failure;
1099 	}
1100 
1101 	if (dev->addr_len) {
1102 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1103 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1104 			goto nla_put_failure;
1105 	}
1106 
1107 	if (rtnl_phys_port_id_fill(skb, dev))
1108 		goto nla_put_failure;
1109 
1110 	if (rtnl_phys_port_name_fill(skb, dev))
1111 		goto nla_put_failure;
1112 
1113 	if (rtnl_phys_switch_id_fill(skb, dev))
1114 		goto nla_put_failure;
1115 
1116 	attr = nla_reserve(skb, IFLA_STATS,
1117 			sizeof(struct rtnl_link_stats));
1118 	if (attr == NULL)
1119 		goto nla_put_failure;
1120 
1121 	stats = dev_get_stats(dev, &temp);
1122 	copy_rtnl_link_stats(nla_data(attr), stats);
1123 
1124 	attr = nla_reserve(skb, IFLA_STATS64,
1125 			sizeof(struct rtnl_link_stats64));
1126 	if (attr == NULL)
1127 		goto nla_put_failure;
1128 	copy_rtnl_link_stats64(nla_data(attr), stats);
1129 
1130 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1131 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1132 		goto nla_put_failure;
1133 
1134 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1135 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
1136 		int i;
1137 
1138 		struct nlattr *vfinfo, *vf, *vfstats;
1139 		int num_vfs = dev_num_vf(dev->dev.parent);
1140 
1141 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1142 		if (!vfinfo)
1143 			goto nla_put_failure;
1144 		for (i = 0; i < num_vfs; i++) {
1145 			struct ifla_vf_info ivi;
1146 			struct ifla_vf_mac vf_mac;
1147 			struct ifla_vf_vlan vf_vlan;
1148 			struct ifla_vf_rate vf_rate;
1149 			struct ifla_vf_tx_rate vf_tx_rate;
1150 			struct ifla_vf_spoofchk vf_spoofchk;
1151 			struct ifla_vf_link_state vf_linkstate;
1152 			struct ifla_vf_rss_query_en vf_rss_query_en;
1153 			struct ifla_vf_stats vf_stats;
1154 
1155 			/*
1156 			 * Not all SR-IOV capable drivers support the
1157 			 * spoofcheck and "RSS query enable" query.  Preset to
1158 			 * -1 so the user space tool can detect that the driver
1159 			 * didn't report anything.
1160 			 */
1161 			ivi.spoofchk = -1;
1162 			ivi.rss_query_en = -1;
1163 			memset(ivi.mac, 0, sizeof(ivi.mac));
1164 			/* The default value for VF link state is "auto"
1165 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
1166 			 */
1167 			ivi.linkstate = 0;
1168 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1169 				break;
1170 			vf_mac.vf =
1171 				vf_vlan.vf =
1172 				vf_rate.vf =
1173 				vf_tx_rate.vf =
1174 				vf_spoofchk.vf =
1175 				vf_linkstate.vf =
1176 				vf_rss_query_en.vf = ivi.vf;
1177 
1178 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1179 			vf_vlan.vlan = ivi.vlan;
1180 			vf_vlan.qos = ivi.qos;
1181 			vf_tx_rate.rate = ivi.max_tx_rate;
1182 			vf_rate.min_tx_rate = ivi.min_tx_rate;
1183 			vf_rate.max_tx_rate = ivi.max_tx_rate;
1184 			vf_spoofchk.setting = ivi.spoofchk;
1185 			vf_linkstate.link_state = ivi.linkstate;
1186 			vf_rss_query_en.setting = ivi.rss_query_en;
1187 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1188 			if (!vf) {
1189 				nla_nest_cancel(skb, vfinfo);
1190 				goto nla_put_failure;
1191 			}
1192 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1193 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1194 			    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1195 				    &vf_rate) ||
1196 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1197 				    &vf_tx_rate) ||
1198 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1199 				    &vf_spoofchk) ||
1200 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1201 				    &vf_linkstate) ||
1202 			    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1203 				    sizeof(vf_rss_query_en),
1204 				    &vf_rss_query_en))
1205 				goto nla_put_failure;
1206 			memset(&vf_stats, 0, sizeof(vf_stats));
1207 			if (dev->netdev_ops->ndo_get_vf_stats)
1208 				dev->netdev_ops->ndo_get_vf_stats(dev, i,
1209 								  &vf_stats);
1210 			vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1211 			if (!vfstats) {
1212 				nla_nest_cancel(skb, vf);
1213 				nla_nest_cancel(skb, vfinfo);
1214 				goto nla_put_failure;
1215 			}
1216 			if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1217 					vf_stats.rx_packets) ||
1218 			    nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1219 					vf_stats.tx_packets) ||
1220 			    nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1221 					vf_stats.rx_bytes) ||
1222 			    nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1223 					vf_stats.tx_bytes) ||
1224 			    nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1225 					vf_stats.broadcast) ||
1226 			    nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1227 					vf_stats.multicast))
1228 				goto nla_put_failure;
1229 			nla_nest_end(skb, vfstats);
1230 			nla_nest_end(skb, vf);
1231 		}
1232 		nla_nest_end(skb, vfinfo);
1233 	}
1234 
1235 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1236 		goto nla_put_failure;
1237 
1238 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1239 		if (rtnl_link_fill(skb, dev) < 0)
1240 			goto nla_put_failure;
1241 	}
1242 
1243 	if (dev->rtnl_link_ops &&
1244 	    dev->rtnl_link_ops->get_link_net) {
1245 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1246 
1247 		if (!net_eq(dev_net(dev), link_net)) {
1248 			int id = peernet2id_alloc(dev_net(dev), link_net);
1249 
1250 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1251 				goto nla_put_failure;
1252 		}
1253 	}
1254 
1255 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1256 		goto nla_put_failure;
1257 
1258 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1259 		if (af_ops->fill_link_af) {
1260 			struct nlattr *af;
1261 			int err;
1262 
1263 			if (!(af = nla_nest_start(skb, af_ops->family)))
1264 				goto nla_put_failure;
1265 
1266 			err = af_ops->fill_link_af(skb, dev);
1267 
1268 			/*
1269 			 * Caller may return ENODATA to indicate that there
1270 			 * was no data to be dumped. This is not an error, it
1271 			 * means we should trim the attribute header and
1272 			 * continue.
1273 			 */
1274 			if (err == -ENODATA)
1275 				nla_nest_cancel(skb, af);
1276 			else if (err < 0)
1277 				goto nla_put_failure;
1278 
1279 			nla_nest_end(skb, af);
1280 		}
1281 	}
1282 
1283 	nla_nest_end(skb, af_spec);
1284 
1285 	nlmsg_end(skb, nlh);
1286 	return 0;
1287 
1288 nla_put_failure:
1289 	nlmsg_cancel(skb, nlh);
1290 	return -EMSGSIZE;
1291 }
1292 
1293 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1294 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1295 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1296 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1297 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1298 	[IFLA_MTU]		= { .type = NLA_U32 },
1299 	[IFLA_LINK]		= { .type = NLA_U32 },
1300 	[IFLA_MASTER]		= { .type = NLA_U32 },
1301 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1302 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1303 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1304 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1305 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1306 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1307 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1308 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1309 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1310 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1311 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1312 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1313 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1314 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1315 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1316 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1317 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1318 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1319 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1320 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1321 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1322 };
1323 
1324 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1325 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1326 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1327 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1328 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1329 };
1330 
1331 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1332 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1333 };
1334 
1335 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1336 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1337 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1338 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1339 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1340 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1341 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1342 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1343 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1344 };
1345 
1346 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1347 	[IFLA_VF_STATS_RX_PACKETS]	= { .type = NLA_U64 },
1348 	[IFLA_VF_STATS_TX_PACKETS]	= { .type = NLA_U64 },
1349 	[IFLA_VF_STATS_RX_BYTES]	= { .type = NLA_U64 },
1350 	[IFLA_VF_STATS_TX_BYTES]	= { .type = NLA_U64 },
1351 	[IFLA_VF_STATS_BROADCAST]	= { .type = NLA_U64 },
1352 	[IFLA_VF_STATS_MULTICAST]	= { .type = NLA_U64 },
1353 };
1354 
1355 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1356 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1357 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1358 				    .len = PORT_PROFILE_MAX },
1359 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1360 				    .len = sizeof(struct ifla_port_vsi)},
1361 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1362 				      .len = PORT_UUID_MAX },
1363 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1364 				    .len = PORT_UUID_MAX },
1365 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1366 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1367 };
1368 
1369 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1370 {
1371 	struct net *net = sock_net(skb->sk);
1372 	int h, s_h;
1373 	int idx = 0, s_idx;
1374 	struct net_device *dev;
1375 	struct hlist_head *head;
1376 	struct nlattr *tb[IFLA_MAX+1];
1377 	u32 ext_filter_mask = 0;
1378 	int err;
1379 	int hdrlen;
1380 
1381 	s_h = cb->args[0];
1382 	s_idx = cb->args[1];
1383 
1384 	cb->seq = net->dev_base_seq;
1385 
1386 	/* A hack to preserve kernel<->userspace interface.
1387 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1388 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1389 	 * what iproute2 < v3.9.0 used.
1390 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1391 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1392 	 */
1393 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1394 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1395 
1396 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1397 
1398 		if (tb[IFLA_EXT_MASK])
1399 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1400 	}
1401 
1402 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1403 		idx = 0;
1404 		head = &net->dev_index_head[h];
1405 		hlist_for_each_entry(dev, head, index_hlist) {
1406 			if (idx < s_idx)
1407 				goto cont;
1408 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1409 					       NETLINK_CB(cb->skb).portid,
1410 					       cb->nlh->nlmsg_seq, 0,
1411 					       NLM_F_MULTI,
1412 					       ext_filter_mask);
1413 			/* If we ran out of room on the first message,
1414 			 * we're in trouble
1415 			 */
1416 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1417 
1418 			if (err < 0)
1419 				goto out;
1420 
1421 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1422 cont:
1423 			idx++;
1424 		}
1425 	}
1426 out:
1427 	cb->args[1] = idx;
1428 	cb->args[0] = h;
1429 
1430 	return skb->len;
1431 }
1432 
1433 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1434 {
1435 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1436 }
1437 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1438 
1439 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1440 {
1441 	struct net *net;
1442 	/* Examine the link attributes and figure out which
1443 	 * network namespace we are talking about.
1444 	 */
1445 	if (tb[IFLA_NET_NS_PID])
1446 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1447 	else if (tb[IFLA_NET_NS_FD])
1448 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1449 	else
1450 		net = get_net(src_net);
1451 	return net;
1452 }
1453 EXPORT_SYMBOL(rtnl_link_get_net);
1454 
1455 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1456 {
1457 	if (dev) {
1458 		if (tb[IFLA_ADDRESS] &&
1459 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1460 			return -EINVAL;
1461 
1462 		if (tb[IFLA_BROADCAST] &&
1463 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1464 			return -EINVAL;
1465 	}
1466 
1467 	if (tb[IFLA_AF_SPEC]) {
1468 		struct nlattr *af;
1469 		int rem, err;
1470 
1471 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1472 			const struct rtnl_af_ops *af_ops;
1473 
1474 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1475 				return -EAFNOSUPPORT;
1476 
1477 			if (!af_ops->set_link_af)
1478 				return -EOPNOTSUPP;
1479 
1480 			if (af_ops->validate_link_af) {
1481 				err = af_ops->validate_link_af(dev, af);
1482 				if (err < 0)
1483 					return err;
1484 			}
1485 		}
1486 	}
1487 
1488 	return 0;
1489 }
1490 
1491 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1492 {
1493 	int rem, err = -EINVAL;
1494 	struct nlattr *vf;
1495 	const struct net_device_ops *ops = dev->netdev_ops;
1496 
1497 	nla_for_each_nested(vf, attr, rem) {
1498 		switch (nla_type(vf)) {
1499 		case IFLA_VF_MAC: {
1500 			struct ifla_vf_mac *ivm;
1501 			ivm = nla_data(vf);
1502 			err = -EOPNOTSUPP;
1503 			if (ops->ndo_set_vf_mac)
1504 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1505 							  ivm->mac);
1506 			break;
1507 		}
1508 		case IFLA_VF_VLAN: {
1509 			struct ifla_vf_vlan *ivv;
1510 			ivv = nla_data(vf);
1511 			err = -EOPNOTSUPP;
1512 			if (ops->ndo_set_vf_vlan)
1513 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1514 							   ivv->vlan,
1515 							   ivv->qos);
1516 			break;
1517 		}
1518 		case IFLA_VF_TX_RATE: {
1519 			struct ifla_vf_tx_rate *ivt;
1520 			struct ifla_vf_info ivf;
1521 			ivt = nla_data(vf);
1522 			err = -EOPNOTSUPP;
1523 			if (ops->ndo_get_vf_config)
1524 				err = ops->ndo_get_vf_config(dev, ivt->vf,
1525 							     &ivf);
1526 			if (err)
1527 				break;
1528 			err = -EOPNOTSUPP;
1529 			if (ops->ndo_set_vf_rate)
1530 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1531 							   ivf.min_tx_rate,
1532 							   ivt->rate);
1533 			break;
1534 		}
1535 		case IFLA_VF_RATE: {
1536 			struct ifla_vf_rate *ivt;
1537 			ivt = nla_data(vf);
1538 			err = -EOPNOTSUPP;
1539 			if (ops->ndo_set_vf_rate)
1540 				err = ops->ndo_set_vf_rate(dev, ivt->vf,
1541 							   ivt->min_tx_rate,
1542 							   ivt->max_tx_rate);
1543 			break;
1544 		}
1545 		case IFLA_VF_SPOOFCHK: {
1546 			struct ifla_vf_spoofchk *ivs;
1547 			ivs = nla_data(vf);
1548 			err = -EOPNOTSUPP;
1549 			if (ops->ndo_set_vf_spoofchk)
1550 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1551 							       ivs->setting);
1552 			break;
1553 		}
1554 		case IFLA_VF_LINK_STATE: {
1555 			struct ifla_vf_link_state *ivl;
1556 			ivl = nla_data(vf);
1557 			err = -EOPNOTSUPP;
1558 			if (ops->ndo_set_vf_link_state)
1559 				err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1560 								 ivl->link_state);
1561 			break;
1562 		}
1563 		case IFLA_VF_RSS_QUERY_EN: {
1564 			struct ifla_vf_rss_query_en *ivrssq_en;
1565 
1566 			ivrssq_en = nla_data(vf);
1567 			err = -EOPNOTSUPP;
1568 			if (ops->ndo_set_vf_rss_query_en)
1569 				err = ops->ndo_set_vf_rss_query_en(dev,
1570 							    ivrssq_en->vf,
1571 							    ivrssq_en->setting);
1572 			break;
1573 		}
1574 		default:
1575 			err = -EINVAL;
1576 			break;
1577 		}
1578 		if (err)
1579 			break;
1580 	}
1581 	return err;
1582 }
1583 
1584 static int do_set_master(struct net_device *dev, int ifindex)
1585 {
1586 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1587 	const struct net_device_ops *ops;
1588 	int err;
1589 
1590 	if (upper_dev) {
1591 		if (upper_dev->ifindex == ifindex)
1592 			return 0;
1593 		ops = upper_dev->netdev_ops;
1594 		if (ops->ndo_del_slave) {
1595 			err = ops->ndo_del_slave(upper_dev, dev);
1596 			if (err)
1597 				return err;
1598 		} else {
1599 			return -EOPNOTSUPP;
1600 		}
1601 	}
1602 
1603 	if (ifindex) {
1604 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1605 		if (!upper_dev)
1606 			return -EINVAL;
1607 		ops = upper_dev->netdev_ops;
1608 		if (ops->ndo_add_slave) {
1609 			err = ops->ndo_add_slave(upper_dev, dev);
1610 			if (err)
1611 				return err;
1612 		} else {
1613 			return -EOPNOTSUPP;
1614 		}
1615 	}
1616 	return 0;
1617 }
1618 
1619 #define DO_SETLINK_MODIFIED	0x01
1620 /* notify flag means notify + modified. */
1621 #define DO_SETLINK_NOTIFY	0x03
1622 static int do_setlink(const struct sk_buff *skb,
1623 		      struct net_device *dev, struct ifinfomsg *ifm,
1624 		      struct nlattr **tb, char *ifname, int status)
1625 {
1626 	const struct net_device_ops *ops = dev->netdev_ops;
1627 	int err;
1628 
1629 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1630 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1631 		if (IS_ERR(net)) {
1632 			err = PTR_ERR(net);
1633 			goto errout;
1634 		}
1635 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1636 			put_net(net);
1637 			err = -EPERM;
1638 			goto errout;
1639 		}
1640 		err = dev_change_net_namespace(dev, net, ifname);
1641 		put_net(net);
1642 		if (err)
1643 			goto errout;
1644 		status |= DO_SETLINK_MODIFIED;
1645 	}
1646 
1647 	if (tb[IFLA_MAP]) {
1648 		struct rtnl_link_ifmap *u_map;
1649 		struct ifmap k_map;
1650 
1651 		if (!ops->ndo_set_config) {
1652 			err = -EOPNOTSUPP;
1653 			goto errout;
1654 		}
1655 
1656 		if (!netif_device_present(dev)) {
1657 			err = -ENODEV;
1658 			goto errout;
1659 		}
1660 
1661 		u_map = nla_data(tb[IFLA_MAP]);
1662 		k_map.mem_start = (unsigned long) u_map->mem_start;
1663 		k_map.mem_end = (unsigned long) u_map->mem_end;
1664 		k_map.base_addr = (unsigned short) u_map->base_addr;
1665 		k_map.irq = (unsigned char) u_map->irq;
1666 		k_map.dma = (unsigned char) u_map->dma;
1667 		k_map.port = (unsigned char) u_map->port;
1668 
1669 		err = ops->ndo_set_config(dev, &k_map);
1670 		if (err < 0)
1671 			goto errout;
1672 
1673 		status |= DO_SETLINK_NOTIFY;
1674 	}
1675 
1676 	if (tb[IFLA_ADDRESS]) {
1677 		struct sockaddr *sa;
1678 		int len;
1679 
1680 		len = sizeof(sa_family_t) + dev->addr_len;
1681 		sa = kmalloc(len, GFP_KERNEL);
1682 		if (!sa) {
1683 			err = -ENOMEM;
1684 			goto errout;
1685 		}
1686 		sa->sa_family = dev->type;
1687 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1688 		       dev->addr_len);
1689 		err = dev_set_mac_address(dev, sa);
1690 		kfree(sa);
1691 		if (err)
1692 			goto errout;
1693 		status |= DO_SETLINK_MODIFIED;
1694 	}
1695 
1696 	if (tb[IFLA_MTU]) {
1697 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1698 		if (err < 0)
1699 			goto errout;
1700 		status |= DO_SETLINK_MODIFIED;
1701 	}
1702 
1703 	if (tb[IFLA_GROUP]) {
1704 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1705 		status |= DO_SETLINK_NOTIFY;
1706 	}
1707 
1708 	/*
1709 	 * Interface selected by interface index but interface
1710 	 * name provided implies that a name change has been
1711 	 * requested.
1712 	 */
1713 	if (ifm->ifi_index > 0 && ifname[0]) {
1714 		err = dev_change_name(dev, ifname);
1715 		if (err < 0)
1716 			goto errout;
1717 		status |= DO_SETLINK_MODIFIED;
1718 	}
1719 
1720 	if (tb[IFLA_IFALIAS]) {
1721 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1722 				    nla_len(tb[IFLA_IFALIAS]));
1723 		if (err < 0)
1724 			goto errout;
1725 		status |= DO_SETLINK_NOTIFY;
1726 	}
1727 
1728 	if (tb[IFLA_BROADCAST]) {
1729 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1730 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1731 	}
1732 
1733 	if (ifm->ifi_flags || ifm->ifi_change) {
1734 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1735 		if (err < 0)
1736 			goto errout;
1737 	}
1738 
1739 	if (tb[IFLA_MASTER]) {
1740 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1741 		if (err)
1742 			goto errout;
1743 		status |= DO_SETLINK_MODIFIED;
1744 	}
1745 
1746 	if (tb[IFLA_CARRIER]) {
1747 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1748 		if (err)
1749 			goto errout;
1750 		status |= DO_SETLINK_MODIFIED;
1751 	}
1752 
1753 	if (tb[IFLA_TXQLEN]) {
1754 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1755 
1756 		if (dev->tx_queue_len ^ value)
1757 			status |= DO_SETLINK_NOTIFY;
1758 
1759 		dev->tx_queue_len = value;
1760 	}
1761 
1762 	if (tb[IFLA_OPERSTATE])
1763 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1764 
1765 	if (tb[IFLA_LINKMODE]) {
1766 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1767 
1768 		write_lock_bh(&dev_base_lock);
1769 		if (dev->link_mode ^ value)
1770 			status |= DO_SETLINK_NOTIFY;
1771 		dev->link_mode = value;
1772 		write_unlock_bh(&dev_base_lock);
1773 	}
1774 
1775 	if (tb[IFLA_VFINFO_LIST]) {
1776 		struct nlattr *attr;
1777 		int rem;
1778 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1779 			if (nla_type(attr) != IFLA_VF_INFO) {
1780 				err = -EINVAL;
1781 				goto errout;
1782 			}
1783 			err = do_setvfinfo(dev, attr);
1784 			if (err < 0)
1785 				goto errout;
1786 			status |= DO_SETLINK_NOTIFY;
1787 		}
1788 	}
1789 	err = 0;
1790 
1791 	if (tb[IFLA_VF_PORTS]) {
1792 		struct nlattr *port[IFLA_PORT_MAX+1];
1793 		struct nlattr *attr;
1794 		int vf;
1795 		int rem;
1796 
1797 		err = -EOPNOTSUPP;
1798 		if (!ops->ndo_set_vf_port)
1799 			goto errout;
1800 
1801 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1802 			if (nla_type(attr) != IFLA_VF_PORT)
1803 				continue;
1804 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1805 				attr, ifla_port_policy);
1806 			if (err < 0)
1807 				goto errout;
1808 			if (!port[IFLA_PORT_VF]) {
1809 				err = -EOPNOTSUPP;
1810 				goto errout;
1811 			}
1812 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1813 			err = ops->ndo_set_vf_port(dev, vf, port);
1814 			if (err < 0)
1815 				goto errout;
1816 			status |= DO_SETLINK_NOTIFY;
1817 		}
1818 	}
1819 	err = 0;
1820 
1821 	if (tb[IFLA_PORT_SELF]) {
1822 		struct nlattr *port[IFLA_PORT_MAX+1];
1823 
1824 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1825 			tb[IFLA_PORT_SELF], ifla_port_policy);
1826 		if (err < 0)
1827 			goto errout;
1828 
1829 		err = -EOPNOTSUPP;
1830 		if (ops->ndo_set_vf_port)
1831 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1832 		if (err < 0)
1833 			goto errout;
1834 		status |= DO_SETLINK_NOTIFY;
1835 	}
1836 
1837 	if (tb[IFLA_AF_SPEC]) {
1838 		struct nlattr *af;
1839 		int rem;
1840 
1841 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1842 			const struct rtnl_af_ops *af_ops;
1843 
1844 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1845 				BUG();
1846 
1847 			err = af_ops->set_link_af(dev, af);
1848 			if (err < 0)
1849 				goto errout;
1850 
1851 			status |= DO_SETLINK_NOTIFY;
1852 		}
1853 	}
1854 	err = 0;
1855 
1856 errout:
1857 	if (status & DO_SETLINK_MODIFIED) {
1858 		if (status & DO_SETLINK_NOTIFY)
1859 			netdev_state_change(dev);
1860 
1861 		if (err < 0)
1862 			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",
1863 					     dev->name);
1864 	}
1865 
1866 	return err;
1867 }
1868 
1869 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1870 {
1871 	struct net *net = sock_net(skb->sk);
1872 	struct ifinfomsg *ifm;
1873 	struct net_device *dev;
1874 	int err;
1875 	struct nlattr *tb[IFLA_MAX+1];
1876 	char ifname[IFNAMSIZ];
1877 
1878 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1879 	if (err < 0)
1880 		goto errout;
1881 
1882 	if (tb[IFLA_IFNAME])
1883 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1884 	else
1885 		ifname[0] = '\0';
1886 
1887 	err = -EINVAL;
1888 	ifm = nlmsg_data(nlh);
1889 	if (ifm->ifi_index > 0)
1890 		dev = __dev_get_by_index(net, ifm->ifi_index);
1891 	else if (tb[IFLA_IFNAME])
1892 		dev = __dev_get_by_name(net, ifname);
1893 	else
1894 		goto errout;
1895 
1896 	if (dev == NULL) {
1897 		err = -ENODEV;
1898 		goto errout;
1899 	}
1900 
1901 	err = validate_linkmsg(dev, tb);
1902 	if (err < 0)
1903 		goto errout;
1904 
1905 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1906 errout:
1907 	return err;
1908 }
1909 
1910 static int rtnl_group_dellink(const struct net *net, int group)
1911 {
1912 	struct net_device *dev, *aux;
1913 	LIST_HEAD(list_kill);
1914 	bool found = false;
1915 
1916 	if (!group)
1917 		return -EPERM;
1918 
1919 	for_each_netdev(net, dev) {
1920 		if (dev->group == group) {
1921 			const struct rtnl_link_ops *ops;
1922 
1923 			found = true;
1924 			ops = dev->rtnl_link_ops;
1925 			if (!ops || !ops->dellink)
1926 				return -EOPNOTSUPP;
1927 		}
1928 	}
1929 
1930 	if (!found)
1931 		return -ENODEV;
1932 
1933 	for_each_netdev_safe(net, dev, aux) {
1934 		if (dev->group == group) {
1935 			const struct rtnl_link_ops *ops;
1936 
1937 			ops = dev->rtnl_link_ops;
1938 			ops->dellink(dev, &list_kill);
1939 		}
1940 	}
1941 	unregister_netdevice_many(&list_kill);
1942 
1943 	return 0;
1944 }
1945 
1946 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1947 {
1948 	struct net *net = sock_net(skb->sk);
1949 	const struct rtnl_link_ops *ops;
1950 	struct net_device *dev;
1951 	struct ifinfomsg *ifm;
1952 	char ifname[IFNAMSIZ];
1953 	struct nlattr *tb[IFLA_MAX+1];
1954 	int err;
1955 	LIST_HEAD(list_kill);
1956 
1957 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1958 	if (err < 0)
1959 		return err;
1960 
1961 	if (tb[IFLA_IFNAME])
1962 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1963 
1964 	ifm = nlmsg_data(nlh);
1965 	if (ifm->ifi_index > 0)
1966 		dev = __dev_get_by_index(net, ifm->ifi_index);
1967 	else if (tb[IFLA_IFNAME])
1968 		dev = __dev_get_by_name(net, ifname);
1969 	else if (tb[IFLA_GROUP])
1970 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
1971 	else
1972 		return -EINVAL;
1973 
1974 	if (!dev)
1975 		return -ENODEV;
1976 
1977 	ops = dev->rtnl_link_ops;
1978 	if (!ops || !ops->dellink)
1979 		return -EOPNOTSUPP;
1980 
1981 	ops->dellink(dev, &list_kill);
1982 	unregister_netdevice_many(&list_kill);
1983 	return 0;
1984 }
1985 
1986 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1987 {
1988 	unsigned int old_flags;
1989 	int err;
1990 
1991 	old_flags = dev->flags;
1992 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1993 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1994 		if (err < 0)
1995 			return err;
1996 	}
1997 
1998 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1999 
2000 	__dev_notify_flags(dev, old_flags, ~0U);
2001 	return 0;
2002 }
2003 EXPORT_SYMBOL(rtnl_configure_link);
2004 
2005 struct net_device *rtnl_create_link(struct net *net,
2006 	const char *ifname, unsigned char name_assign_type,
2007 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2008 {
2009 	int err;
2010 	struct net_device *dev;
2011 	unsigned int num_tx_queues = 1;
2012 	unsigned int num_rx_queues = 1;
2013 
2014 	if (tb[IFLA_NUM_TX_QUEUES])
2015 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2016 	else if (ops->get_num_tx_queues)
2017 		num_tx_queues = ops->get_num_tx_queues();
2018 
2019 	if (tb[IFLA_NUM_RX_QUEUES])
2020 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2021 	else if (ops->get_num_rx_queues)
2022 		num_rx_queues = ops->get_num_rx_queues();
2023 
2024 	err = -ENOMEM;
2025 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2026 			       ops->setup, num_tx_queues, num_rx_queues);
2027 	if (!dev)
2028 		goto err;
2029 
2030 	dev_net_set(dev, net);
2031 	dev->rtnl_link_ops = ops;
2032 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2033 
2034 	if (tb[IFLA_MTU])
2035 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2036 	if (tb[IFLA_ADDRESS]) {
2037 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2038 				nla_len(tb[IFLA_ADDRESS]));
2039 		dev->addr_assign_type = NET_ADDR_SET;
2040 	}
2041 	if (tb[IFLA_BROADCAST])
2042 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2043 				nla_len(tb[IFLA_BROADCAST]));
2044 	if (tb[IFLA_TXQLEN])
2045 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2046 	if (tb[IFLA_OPERSTATE])
2047 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2048 	if (tb[IFLA_LINKMODE])
2049 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2050 	if (tb[IFLA_GROUP])
2051 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2052 
2053 	return dev;
2054 
2055 err:
2056 	return ERR_PTR(err);
2057 }
2058 EXPORT_SYMBOL(rtnl_create_link);
2059 
2060 static int rtnl_group_changelink(const struct sk_buff *skb,
2061 		struct net *net, int group,
2062 		struct ifinfomsg *ifm,
2063 		struct nlattr **tb)
2064 {
2065 	struct net_device *dev, *aux;
2066 	int err;
2067 
2068 	for_each_netdev_safe(net, dev, aux) {
2069 		if (dev->group == group) {
2070 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2071 			if (err < 0)
2072 				return err;
2073 		}
2074 	}
2075 
2076 	return 0;
2077 }
2078 
2079 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2080 {
2081 	struct net *net = sock_net(skb->sk);
2082 	const struct rtnl_link_ops *ops;
2083 	const struct rtnl_link_ops *m_ops = NULL;
2084 	struct net_device *dev;
2085 	struct net_device *master_dev = NULL;
2086 	struct ifinfomsg *ifm;
2087 	char kind[MODULE_NAME_LEN];
2088 	char ifname[IFNAMSIZ];
2089 	struct nlattr *tb[IFLA_MAX+1];
2090 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2091 	unsigned char name_assign_type = NET_NAME_USER;
2092 	int err;
2093 
2094 #ifdef CONFIG_MODULES
2095 replay:
2096 #endif
2097 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2098 	if (err < 0)
2099 		return err;
2100 
2101 	if (tb[IFLA_IFNAME])
2102 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2103 	else
2104 		ifname[0] = '\0';
2105 
2106 	ifm = nlmsg_data(nlh);
2107 	if (ifm->ifi_index > 0)
2108 		dev = __dev_get_by_index(net, ifm->ifi_index);
2109 	else {
2110 		if (ifname[0])
2111 			dev = __dev_get_by_name(net, ifname);
2112 		else
2113 			dev = NULL;
2114 	}
2115 
2116 	if (dev) {
2117 		master_dev = netdev_master_upper_dev_get(dev);
2118 		if (master_dev)
2119 			m_ops = master_dev->rtnl_link_ops;
2120 	}
2121 
2122 	err = validate_linkmsg(dev, tb);
2123 	if (err < 0)
2124 		return err;
2125 
2126 	if (tb[IFLA_LINKINFO]) {
2127 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2128 				       tb[IFLA_LINKINFO], ifla_info_policy);
2129 		if (err < 0)
2130 			return err;
2131 	} else
2132 		memset(linkinfo, 0, sizeof(linkinfo));
2133 
2134 	if (linkinfo[IFLA_INFO_KIND]) {
2135 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2136 		ops = rtnl_link_ops_get(kind);
2137 	} else {
2138 		kind[0] = '\0';
2139 		ops = NULL;
2140 	}
2141 
2142 	if (1) {
2143 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2144 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2145 		struct nlattr **data = NULL;
2146 		struct nlattr **slave_data = NULL;
2147 		struct net *dest_net, *link_net = NULL;
2148 
2149 		if (ops) {
2150 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2151 				err = nla_parse_nested(attr, ops->maxtype,
2152 						       linkinfo[IFLA_INFO_DATA],
2153 						       ops->policy);
2154 				if (err < 0)
2155 					return err;
2156 				data = attr;
2157 			}
2158 			if (ops->validate) {
2159 				err = ops->validate(tb, data);
2160 				if (err < 0)
2161 					return err;
2162 			}
2163 		}
2164 
2165 		if (m_ops) {
2166 			if (m_ops->slave_maxtype &&
2167 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2168 				err = nla_parse_nested(slave_attr,
2169 						       m_ops->slave_maxtype,
2170 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2171 						       m_ops->slave_policy);
2172 				if (err < 0)
2173 					return err;
2174 				slave_data = slave_attr;
2175 			}
2176 			if (m_ops->slave_validate) {
2177 				err = m_ops->slave_validate(tb, slave_data);
2178 				if (err < 0)
2179 					return err;
2180 			}
2181 		}
2182 
2183 		if (dev) {
2184 			int status = 0;
2185 
2186 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2187 				return -EEXIST;
2188 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2189 				return -EOPNOTSUPP;
2190 
2191 			if (linkinfo[IFLA_INFO_DATA]) {
2192 				if (!ops || ops != dev->rtnl_link_ops ||
2193 				    !ops->changelink)
2194 					return -EOPNOTSUPP;
2195 
2196 				err = ops->changelink(dev, tb, data);
2197 				if (err < 0)
2198 					return err;
2199 				status |= DO_SETLINK_NOTIFY;
2200 			}
2201 
2202 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2203 				if (!m_ops || !m_ops->slave_changelink)
2204 					return -EOPNOTSUPP;
2205 
2206 				err = m_ops->slave_changelink(master_dev, dev,
2207 							      tb, slave_data);
2208 				if (err < 0)
2209 					return err;
2210 				status |= DO_SETLINK_NOTIFY;
2211 			}
2212 
2213 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2214 		}
2215 
2216 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2217 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2218 				return rtnl_group_changelink(skb, net,
2219 						nla_get_u32(tb[IFLA_GROUP]),
2220 						ifm, tb);
2221 			return -ENODEV;
2222 		}
2223 
2224 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2225 			return -EOPNOTSUPP;
2226 
2227 		if (!ops) {
2228 #ifdef CONFIG_MODULES
2229 			if (kind[0]) {
2230 				__rtnl_unlock();
2231 				request_module("rtnl-link-%s", kind);
2232 				rtnl_lock();
2233 				ops = rtnl_link_ops_get(kind);
2234 				if (ops)
2235 					goto replay;
2236 			}
2237 #endif
2238 			return -EOPNOTSUPP;
2239 		}
2240 
2241 		if (!ops->setup)
2242 			return -EOPNOTSUPP;
2243 
2244 		if (!ifname[0]) {
2245 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2246 			name_assign_type = NET_NAME_ENUM;
2247 		}
2248 
2249 		dest_net = rtnl_link_get_net(net, tb);
2250 		if (IS_ERR(dest_net))
2251 			return PTR_ERR(dest_net);
2252 
2253 		err = -EPERM;
2254 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2255 			goto out;
2256 
2257 		if (tb[IFLA_LINK_NETNSID]) {
2258 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2259 
2260 			link_net = get_net_ns_by_id(dest_net, id);
2261 			if (!link_net) {
2262 				err =  -EINVAL;
2263 				goto out;
2264 			}
2265 			err = -EPERM;
2266 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2267 				goto out;
2268 		}
2269 
2270 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2271 				       name_assign_type, ops, tb);
2272 		if (IS_ERR(dev)) {
2273 			err = PTR_ERR(dev);
2274 			goto out;
2275 		}
2276 
2277 		dev->ifindex = ifm->ifi_index;
2278 
2279 		if (ops->newlink) {
2280 			err = ops->newlink(link_net ? : net, dev, tb, data);
2281 			/* Drivers should call free_netdev() in ->destructor
2282 			 * and unregister it on failure after registration
2283 			 * so that device could be finally freed in rtnl_unlock.
2284 			 */
2285 			if (err < 0) {
2286 				/* If device is not registered at all, free it now */
2287 				if (dev->reg_state == NETREG_UNINITIALIZED)
2288 					free_netdev(dev);
2289 				goto out;
2290 			}
2291 		} else {
2292 			err = register_netdevice(dev);
2293 			if (err < 0) {
2294 				free_netdev(dev);
2295 				goto out;
2296 			}
2297 		}
2298 		err = rtnl_configure_link(dev, ifm);
2299 		if (err < 0)
2300 			goto out_unregister;
2301 		if (link_net) {
2302 			err = dev_change_net_namespace(dev, dest_net, ifname);
2303 			if (err < 0)
2304 				goto out_unregister;
2305 		}
2306 out:
2307 		if (link_net)
2308 			put_net(link_net);
2309 		put_net(dest_net);
2310 		return err;
2311 out_unregister:
2312 		if (ops->newlink) {
2313 			LIST_HEAD(list_kill);
2314 
2315 			ops->dellink(dev, &list_kill);
2316 			unregister_netdevice_many(&list_kill);
2317 		} else {
2318 			unregister_netdevice(dev);
2319 		}
2320 		goto out;
2321 	}
2322 }
2323 
2324 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2325 {
2326 	struct net *net = sock_net(skb->sk);
2327 	struct ifinfomsg *ifm;
2328 	char ifname[IFNAMSIZ];
2329 	struct nlattr *tb[IFLA_MAX+1];
2330 	struct net_device *dev = NULL;
2331 	struct sk_buff *nskb;
2332 	int err;
2333 	u32 ext_filter_mask = 0;
2334 
2335 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2336 	if (err < 0)
2337 		return err;
2338 
2339 	if (tb[IFLA_IFNAME])
2340 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2341 
2342 	if (tb[IFLA_EXT_MASK])
2343 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2344 
2345 	ifm = nlmsg_data(nlh);
2346 	if (ifm->ifi_index > 0)
2347 		dev = __dev_get_by_index(net, ifm->ifi_index);
2348 	else if (tb[IFLA_IFNAME])
2349 		dev = __dev_get_by_name(net, ifname);
2350 	else
2351 		return -EINVAL;
2352 
2353 	if (dev == NULL)
2354 		return -ENODEV;
2355 
2356 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2357 	if (nskb == NULL)
2358 		return -ENOBUFS;
2359 
2360 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2361 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2362 	if (err < 0) {
2363 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2364 		WARN_ON(err == -EMSGSIZE);
2365 		kfree_skb(nskb);
2366 	} else
2367 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2368 
2369 	return err;
2370 }
2371 
2372 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2373 {
2374 	struct net *net = sock_net(skb->sk);
2375 	struct net_device *dev;
2376 	struct nlattr *tb[IFLA_MAX+1];
2377 	u32 ext_filter_mask = 0;
2378 	u16 min_ifinfo_dump_size = 0;
2379 	int hdrlen;
2380 
2381 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2382 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2383 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2384 
2385 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2386 		if (tb[IFLA_EXT_MASK])
2387 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2388 	}
2389 
2390 	if (!ext_filter_mask)
2391 		return NLMSG_GOODSIZE;
2392 	/*
2393 	 * traverse the list of net devices and compute the minimum
2394 	 * buffer size based upon the filter mask.
2395 	 */
2396 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2397 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2398 					     if_nlmsg_size(dev,
2399 						           ext_filter_mask));
2400 	}
2401 
2402 	return min_ifinfo_dump_size;
2403 }
2404 
2405 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2406 {
2407 	int idx;
2408 	int s_idx = cb->family;
2409 
2410 	if (s_idx == 0)
2411 		s_idx = 1;
2412 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2413 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2414 		if (idx < s_idx || idx == PF_PACKET)
2415 			continue;
2416 		if (rtnl_msg_handlers[idx] == NULL ||
2417 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2418 			continue;
2419 		if (idx > s_idx) {
2420 			memset(&cb->args[0], 0, sizeof(cb->args));
2421 			cb->prev_seq = 0;
2422 			cb->seq = 0;
2423 		}
2424 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2425 			break;
2426 	}
2427 	cb->family = idx;
2428 
2429 	return skb->len;
2430 }
2431 
2432 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2433 				       unsigned int change, gfp_t flags)
2434 {
2435 	struct net *net = dev_net(dev);
2436 	struct sk_buff *skb;
2437 	int err = -ENOBUFS;
2438 	size_t if_info_size;
2439 
2440 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2441 	if (skb == NULL)
2442 		goto errout;
2443 
2444 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2445 	if (err < 0) {
2446 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2447 		WARN_ON(err == -EMSGSIZE);
2448 		kfree_skb(skb);
2449 		goto errout;
2450 	}
2451 	return skb;
2452 errout:
2453 	if (err < 0)
2454 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2455 	return NULL;
2456 }
2457 
2458 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2459 {
2460 	struct net *net = dev_net(dev);
2461 
2462 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2463 }
2464 
2465 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2466 		  gfp_t flags)
2467 {
2468 	struct sk_buff *skb;
2469 
2470 	if (dev->reg_state != NETREG_REGISTERED)
2471 		return;
2472 
2473 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2474 	if (skb)
2475 		rtmsg_ifinfo_send(skb, dev, flags);
2476 }
2477 EXPORT_SYMBOL(rtmsg_ifinfo);
2478 
2479 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2480 				   struct net_device *dev,
2481 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2482 				   int type, unsigned int flags,
2483 				   int nlflags)
2484 {
2485 	struct nlmsghdr *nlh;
2486 	struct ndmsg *ndm;
2487 
2488 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2489 	if (!nlh)
2490 		return -EMSGSIZE;
2491 
2492 	ndm = nlmsg_data(nlh);
2493 	ndm->ndm_family  = AF_BRIDGE;
2494 	ndm->ndm_pad1	 = 0;
2495 	ndm->ndm_pad2    = 0;
2496 	ndm->ndm_flags	 = flags;
2497 	ndm->ndm_type	 = 0;
2498 	ndm->ndm_ifindex = dev->ifindex;
2499 	ndm->ndm_state   = NUD_PERMANENT;
2500 
2501 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2502 		goto nla_put_failure;
2503 	if (vid)
2504 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2505 			goto nla_put_failure;
2506 
2507 	nlmsg_end(skb, nlh);
2508 	return 0;
2509 
2510 nla_put_failure:
2511 	nlmsg_cancel(skb, nlh);
2512 	return -EMSGSIZE;
2513 }
2514 
2515 static inline size_t rtnl_fdb_nlmsg_size(void)
2516 {
2517 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2518 }
2519 
2520 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2521 {
2522 	struct net *net = dev_net(dev);
2523 	struct sk_buff *skb;
2524 	int err = -ENOBUFS;
2525 
2526 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2527 	if (!skb)
2528 		goto errout;
2529 
2530 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2531 				      0, 0, type, NTF_SELF, 0);
2532 	if (err < 0) {
2533 		kfree_skb(skb);
2534 		goto errout;
2535 	}
2536 
2537 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2538 	return;
2539 errout:
2540 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2541 }
2542 
2543 /**
2544  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2545  */
2546 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2547 		     struct nlattr *tb[],
2548 		     struct net_device *dev,
2549 		     const unsigned char *addr, u16 vid,
2550 		     u16 flags)
2551 {
2552 	int err = -EINVAL;
2553 
2554 	/* If aging addresses are supported device will need to
2555 	 * implement its own handler for this.
2556 	 */
2557 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2558 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2559 		return err;
2560 	}
2561 
2562 	if (vid) {
2563 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2564 		return err;
2565 	}
2566 
2567 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2568 		err = dev_uc_add_excl(dev, addr);
2569 	else if (is_multicast_ether_addr(addr))
2570 		err = dev_mc_add_excl(dev, addr);
2571 
2572 	/* Only return duplicate errors if NLM_F_EXCL is set */
2573 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2574 		err = 0;
2575 
2576 	return err;
2577 }
2578 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2579 
2580 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2581 {
2582 	u16 vid = 0;
2583 
2584 	if (vlan_attr) {
2585 		if (nla_len(vlan_attr) != sizeof(u16)) {
2586 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2587 			return -EINVAL;
2588 		}
2589 
2590 		vid = nla_get_u16(vlan_attr);
2591 
2592 		if (!vid || vid >= VLAN_VID_MASK) {
2593 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2594 				vid);
2595 			return -EINVAL;
2596 		}
2597 	}
2598 	*p_vid = vid;
2599 	return 0;
2600 }
2601 
2602 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2603 {
2604 	struct net *net = sock_net(skb->sk);
2605 	struct ndmsg *ndm;
2606 	struct nlattr *tb[NDA_MAX+1];
2607 	struct net_device *dev;
2608 	u8 *addr;
2609 	u16 vid;
2610 	int err;
2611 
2612 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2613 	if (err < 0)
2614 		return err;
2615 
2616 	ndm = nlmsg_data(nlh);
2617 	if (ndm->ndm_ifindex == 0) {
2618 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2619 		return -EINVAL;
2620 	}
2621 
2622 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2623 	if (dev == NULL) {
2624 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2625 		return -ENODEV;
2626 	}
2627 
2628 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2629 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2630 		return -EINVAL;
2631 	}
2632 
2633 	addr = nla_data(tb[NDA_LLADDR]);
2634 
2635 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2636 	if (err)
2637 		return err;
2638 
2639 	err = -EOPNOTSUPP;
2640 
2641 	/* Support fdb on master device the net/bridge default case */
2642 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2643 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2644 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2645 		const struct net_device_ops *ops = br_dev->netdev_ops;
2646 
2647 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2648 				       nlh->nlmsg_flags);
2649 		if (err)
2650 			goto out;
2651 		else
2652 			ndm->ndm_flags &= ~NTF_MASTER;
2653 	}
2654 
2655 	/* Embedded bridge, macvlan, and any other device support */
2656 	if ((ndm->ndm_flags & NTF_SELF)) {
2657 		if (dev->netdev_ops->ndo_fdb_add)
2658 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2659 							   vid,
2660 							   nlh->nlmsg_flags);
2661 		else
2662 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2663 					       nlh->nlmsg_flags);
2664 
2665 		if (!err) {
2666 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2667 			ndm->ndm_flags &= ~NTF_SELF;
2668 		}
2669 	}
2670 out:
2671 	return err;
2672 }
2673 
2674 /**
2675  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2676  */
2677 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2678 		     struct nlattr *tb[],
2679 		     struct net_device *dev,
2680 		     const unsigned char *addr, u16 vid)
2681 {
2682 	int err = -EINVAL;
2683 
2684 	/* If aging addresses are supported device will need to
2685 	 * implement its own handler for this.
2686 	 */
2687 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2688 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2689 		return err;
2690 	}
2691 
2692 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2693 		err = dev_uc_del(dev, addr);
2694 	else if (is_multicast_ether_addr(addr))
2695 		err = dev_mc_del(dev, addr);
2696 
2697 	return err;
2698 }
2699 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2700 
2701 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2702 {
2703 	struct net *net = sock_net(skb->sk);
2704 	struct ndmsg *ndm;
2705 	struct nlattr *tb[NDA_MAX+1];
2706 	struct net_device *dev;
2707 	int err = -EINVAL;
2708 	__u8 *addr;
2709 	u16 vid;
2710 
2711 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2712 		return -EPERM;
2713 
2714 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2715 	if (err < 0)
2716 		return err;
2717 
2718 	ndm = nlmsg_data(nlh);
2719 	if (ndm->ndm_ifindex == 0) {
2720 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2721 		return -EINVAL;
2722 	}
2723 
2724 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2725 	if (dev == NULL) {
2726 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2727 		return -ENODEV;
2728 	}
2729 
2730 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2731 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2732 		return -EINVAL;
2733 	}
2734 
2735 	addr = nla_data(tb[NDA_LLADDR]);
2736 
2737 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2738 	if (err)
2739 		return err;
2740 
2741 	err = -EOPNOTSUPP;
2742 
2743 	/* Support fdb on master device the net/bridge default case */
2744 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2745 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2746 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2747 		const struct net_device_ops *ops = br_dev->netdev_ops;
2748 
2749 		if (ops->ndo_fdb_del)
2750 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2751 
2752 		if (err)
2753 			goto out;
2754 		else
2755 			ndm->ndm_flags &= ~NTF_MASTER;
2756 	}
2757 
2758 	/* Embedded bridge, macvlan, and any other device support */
2759 	if (ndm->ndm_flags & NTF_SELF) {
2760 		if (dev->netdev_ops->ndo_fdb_del)
2761 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2762 							   vid);
2763 		else
2764 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2765 
2766 		if (!err) {
2767 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2768 			ndm->ndm_flags &= ~NTF_SELF;
2769 		}
2770 	}
2771 out:
2772 	return err;
2773 }
2774 
2775 static int nlmsg_populate_fdb(struct sk_buff *skb,
2776 			      struct netlink_callback *cb,
2777 			      struct net_device *dev,
2778 			      int *idx,
2779 			      struct netdev_hw_addr_list *list)
2780 {
2781 	struct netdev_hw_addr *ha;
2782 	int err;
2783 	u32 portid, seq;
2784 
2785 	portid = NETLINK_CB(cb->skb).portid;
2786 	seq = cb->nlh->nlmsg_seq;
2787 
2788 	list_for_each_entry(ha, &list->list, list) {
2789 		if (*idx < cb->args[0])
2790 			goto skip;
2791 
2792 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2793 					      portid, seq,
2794 					      RTM_NEWNEIGH, NTF_SELF,
2795 					      NLM_F_MULTI);
2796 		if (err < 0)
2797 			return err;
2798 skip:
2799 		*idx += 1;
2800 	}
2801 	return 0;
2802 }
2803 
2804 /**
2805  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2806  * @nlh: netlink message header
2807  * @dev: netdevice
2808  *
2809  * Default netdevice operation to dump the existing unicast address list.
2810  * Returns number of addresses from list put in skb.
2811  */
2812 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2813 		      struct netlink_callback *cb,
2814 		      struct net_device *dev,
2815 		      struct net_device *filter_dev,
2816 		      int idx)
2817 {
2818 	int err;
2819 
2820 	netif_addr_lock_bh(dev);
2821 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2822 	if (err)
2823 		goto out;
2824 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2825 out:
2826 	netif_addr_unlock_bh(dev);
2827 	return idx;
2828 }
2829 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2830 
2831 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2832 {
2833 	struct net_device *dev;
2834 	struct nlattr *tb[IFLA_MAX+1];
2835 	struct net_device *br_dev = NULL;
2836 	const struct net_device_ops *ops = NULL;
2837 	const struct net_device_ops *cops = NULL;
2838 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2839 	struct net *net = sock_net(skb->sk);
2840 	int brport_idx = 0;
2841 	int br_idx = 0;
2842 	int idx = 0;
2843 
2844 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2845 			ifla_policy) == 0) {
2846 		if (tb[IFLA_MASTER])
2847 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
2848 	}
2849 
2850 	brport_idx = ifm->ifi_index;
2851 
2852 	if (br_idx) {
2853 		br_dev = __dev_get_by_index(net, br_idx);
2854 		if (!br_dev)
2855 			return -ENODEV;
2856 
2857 		ops = br_dev->netdev_ops;
2858 	}
2859 
2860 	for_each_netdev(net, dev) {
2861 		if (brport_idx && (dev->ifindex != brport_idx))
2862 			continue;
2863 
2864 		if (!br_idx) { /* user did not specify a specific bridge */
2865 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
2866 				br_dev = netdev_master_upper_dev_get(dev);
2867 				cops = br_dev->netdev_ops;
2868 			}
2869 
2870 		} else {
2871 			if (dev != br_dev &&
2872 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
2873 				continue;
2874 
2875 			if (br_dev != netdev_master_upper_dev_get(dev) &&
2876 			    !(dev->priv_flags & IFF_EBRIDGE))
2877 				continue;
2878 
2879 			cops = ops;
2880 		}
2881 
2882 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2883 			if (cops && cops->ndo_fdb_dump)
2884 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2885 							 idx);
2886 		}
2887 
2888 		if (dev->netdev_ops->ndo_fdb_dump)
2889 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2890 							    idx);
2891 		else
2892 			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2893 
2894 		cops = NULL;
2895 	}
2896 
2897 	cb->args[0] = idx;
2898 	return skb->len;
2899 }
2900 
2901 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2902 			       unsigned int attrnum, unsigned int flag)
2903 {
2904 	if (mask & flag)
2905 		return nla_put_u8(skb, attrnum, !!(flags & flag));
2906 	return 0;
2907 }
2908 
2909 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2910 			    struct net_device *dev, u16 mode,
2911 			    u32 flags, u32 mask, int nlflags,
2912 			    u32 filter_mask,
2913 			    int (*vlan_fill)(struct sk_buff *skb,
2914 					     struct net_device *dev,
2915 					     u32 filter_mask))
2916 {
2917 	struct nlmsghdr *nlh;
2918 	struct ifinfomsg *ifm;
2919 	struct nlattr *br_afspec;
2920 	struct nlattr *protinfo;
2921 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2922 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2923 	int err = 0;
2924 
2925 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
2926 	if (nlh == NULL)
2927 		return -EMSGSIZE;
2928 
2929 	ifm = nlmsg_data(nlh);
2930 	ifm->ifi_family = AF_BRIDGE;
2931 	ifm->__ifi_pad = 0;
2932 	ifm->ifi_type = dev->type;
2933 	ifm->ifi_index = dev->ifindex;
2934 	ifm->ifi_flags = dev_get_flags(dev);
2935 	ifm->ifi_change = 0;
2936 
2937 
2938 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2939 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2940 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2941 	    (br_dev &&
2942 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2943 	    (dev->addr_len &&
2944 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2945 	    (dev->ifindex != dev_get_iflink(dev) &&
2946 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
2947 		goto nla_put_failure;
2948 
2949 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2950 	if (!br_afspec)
2951 		goto nla_put_failure;
2952 
2953 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2954 		nla_nest_cancel(skb, br_afspec);
2955 		goto nla_put_failure;
2956 	}
2957 
2958 	if (mode != BRIDGE_MODE_UNDEF) {
2959 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2960 			nla_nest_cancel(skb, br_afspec);
2961 			goto nla_put_failure;
2962 		}
2963 	}
2964 	if (vlan_fill) {
2965 		err = vlan_fill(skb, dev, filter_mask);
2966 		if (err) {
2967 			nla_nest_cancel(skb, br_afspec);
2968 			goto nla_put_failure;
2969 		}
2970 	}
2971 	nla_nest_end(skb, br_afspec);
2972 
2973 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2974 	if (!protinfo)
2975 		goto nla_put_failure;
2976 
2977 	if (brport_nla_put_flag(skb, flags, mask,
2978 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2979 	    brport_nla_put_flag(skb, flags, mask,
2980 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2981 	    brport_nla_put_flag(skb, flags, mask,
2982 				IFLA_BRPORT_FAST_LEAVE,
2983 				BR_MULTICAST_FAST_LEAVE) ||
2984 	    brport_nla_put_flag(skb, flags, mask,
2985 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2986 	    brport_nla_put_flag(skb, flags, mask,
2987 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2988 	    brport_nla_put_flag(skb, flags, mask,
2989 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2990 	    brport_nla_put_flag(skb, flags, mask,
2991 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2992 	    brport_nla_put_flag(skb, flags, mask,
2993 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2994 		nla_nest_cancel(skb, protinfo);
2995 		goto nla_put_failure;
2996 	}
2997 
2998 	nla_nest_end(skb, protinfo);
2999 
3000 	nlmsg_end(skb, nlh);
3001 	return 0;
3002 nla_put_failure:
3003 	nlmsg_cancel(skb, nlh);
3004 	return err ? err : -EMSGSIZE;
3005 }
3006 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3007 
3008 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3009 {
3010 	struct net *net = sock_net(skb->sk);
3011 	struct net_device *dev;
3012 	int idx = 0;
3013 	u32 portid = NETLINK_CB(cb->skb).portid;
3014 	u32 seq = cb->nlh->nlmsg_seq;
3015 	u32 filter_mask = 0;
3016 
3017 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3018 		struct nlattr *extfilt;
3019 
3020 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3021 					  IFLA_EXT_MASK);
3022 		if (extfilt) {
3023 			if (nla_len(extfilt) < sizeof(filter_mask))
3024 				return -EINVAL;
3025 
3026 			filter_mask = nla_get_u32(extfilt);
3027 		}
3028 	}
3029 
3030 	rcu_read_lock();
3031 	for_each_netdev_rcu(net, dev) {
3032 		const struct net_device_ops *ops = dev->netdev_ops;
3033 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3034 
3035 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3036 			if (idx >= cb->args[0] &&
3037 			    br_dev->netdev_ops->ndo_bridge_getlink(
3038 				    skb, portid, seq, dev, filter_mask,
3039 				    NLM_F_MULTI) < 0)
3040 				break;
3041 			idx++;
3042 		}
3043 
3044 		if (ops->ndo_bridge_getlink) {
3045 			if (idx >= cb->args[0] &&
3046 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
3047 						    filter_mask,
3048 						    NLM_F_MULTI) < 0)
3049 				break;
3050 			idx++;
3051 		}
3052 	}
3053 	rcu_read_unlock();
3054 	cb->args[0] = idx;
3055 
3056 	return skb->len;
3057 }
3058 
3059 static inline size_t bridge_nlmsg_size(void)
3060 {
3061 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3062 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3063 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3064 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3065 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3066 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3067 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3068 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3069 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3070 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3071 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3072 }
3073 
3074 static int rtnl_bridge_notify(struct net_device *dev)
3075 {
3076 	struct net *net = dev_net(dev);
3077 	struct sk_buff *skb;
3078 	int err = -EOPNOTSUPP;
3079 
3080 	if (!dev->netdev_ops->ndo_bridge_getlink)
3081 		return 0;
3082 
3083 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3084 	if (!skb) {
3085 		err = -ENOMEM;
3086 		goto errout;
3087 	}
3088 
3089 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3090 	if (err < 0)
3091 		goto errout;
3092 
3093 	if (!skb->len)
3094 		goto errout;
3095 
3096 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3097 	return 0;
3098 errout:
3099 	WARN_ON(err == -EMSGSIZE);
3100 	kfree_skb(skb);
3101 	if (err)
3102 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3103 	return err;
3104 }
3105 
3106 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3107 {
3108 	struct net *net = sock_net(skb->sk);
3109 	struct ifinfomsg *ifm;
3110 	struct net_device *dev;
3111 	struct nlattr *br_spec, *attr = NULL;
3112 	int rem, err = -EOPNOTSUPP;
3113 	u16 flags = 0;
3114 	bool have_flags = false;
3115 
3116 	if (nlmsg_len(nlh) < sizeof(*ifm))
3117 		return -EINVAL;
3118 
3119 	ifm = nlmsg_data(nlh);
3120 	if (ifm->ifi_family != AF_BRIDGE)
3121 		return -EPFNOSUPPORT;
3122 
3123 	dev = __dev_get_by_index(net, ifm->ifi_index);
3124 	if (!dev) {
3125 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3126 		return -ENODEV;
3127 	}
3128 
3129 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3130 	if (br_spec) {
3131 		nla_for_each_nested(attr, br_spec, rem) {
3132 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3133 				if (nla_len(attr) < sizeof(flags))
3134 					return -EINVAL;
3135 
3136 				have_flags = true;
3137 				flags = nla_get_u16(attr);
3138 				break;
3139 			}
3140 		}
3141 	}
3142 
3143 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3144 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3145 
3146 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3147 			err = -EOPNOTSUPP;
3148 			goto out;
3149 		}
3150 
3151 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3152 		if (err)
3153 			goto out;
3154 
3155 		flags &= ~BRIDGE_FLAGS_MASTER;
3156 	}
3157 
3158 	if ((flags & BRIDGE_FLAGS_SELF)) {
3159 		if (!dev->netdev_ops->ndo_bridge_setlink)
3160 			err = -EOPNOTSUPP;
3161 		else
3162 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3163 								  flags);
3164 		if (!err) {
3165 			flags &= ~BRIDGE_FLAGS_SELF;
3166 
3167 			/* Generate event to notify upper layer of bridge
3168 			 * change
3169 			 */
3170 			err = rtnl_bridge_notify(dev);
3171 		}
3172 	}
3173 
3174 	if (have_flags)
3175 		memcpy(nla_data(attr), &flags, sizeof(flags));
3176 out:
3177 	return err;
3178 }
3179 
3180 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3181 {
3182 	struct net *net = sock_net(skb->sk);
3183 	struct ifinfomsg *ifm;
3184 	struct net_device *dev;
3185 	struct nlattr *br_spec, *attr = NULL;
3186 	int rem, err = -EOPNOTSUPP;
3187 	u16 flags = 0;
3188 	bool have_flags = false;
3189 
3190 	if (nlmsg_len(nlh) < sizeof(*ifm))
3191 		return -EINVAL;
3192 
3193 	ifm = nlmsg_data(nlh);
3194 	if (ifm->ifi_family != AF_BRIDGE)
3195 		return -EPFNOSUPPORT;
3196 
3197 	dev = __dev_get_by_index(net, ifm->ifi_index);
3198 	if (!dev) {
3199 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3200 		return -ENODEV;
3201 	}
3202 
3203 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3204 	if (br_spec) {
3205 		nla_for_each_nested(attr, br_spec, rem) {
3206 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3207 				if (nla_len(attr) < sizeof(flags))
3208 					return -EINVAL;
3209 
3210 				have_flags = true;
3211 				flags = nla_get_u16(attr);
3212 				break;
3213 			}
3214 		}
3215 	}
3216 
3217 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3218 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3219 
3220 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3221 			err = -EOPNOTSUPP;
3222 			goto out;
3223 		}
3224 
3225 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3226 		if (err)
3227 			goto out;
3228 
3229 		flags &= ~BRIDGE_FLAGS_MASTER;
3230 	}
3231 
3232 	if ((flags & BRIDGE_FLAGS_SELF)) {
3233 		if (!dev->netdev_ops->ndo_bridge_dellink)
3234 			err = -EOPNOTSUPP;
3235 		else
3236 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3237 								  flags);
3238 
3239 		if (!err) {
3240 			flags &= ~BRIDGE_FLAGS_SELF;
3241 
3242 			/* Generate event to notify upper layer of bridge
3243 			 * change
3244 			 */
3245 			err = rtnl_bridge_notify(dev);
3246 		}
3247 	}
3248 
3249 	if (have_flags)
3250 		memcpy(nla_data(attr), &flags, sizeof(flags));
3251 out:
3252 	return err;
3253 }
3254 
3255 /* Process one rtnetlink message. */
3256 
3257 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3258 {
3259 	struct net *net = sock_net(skb->sk);
3260 	rtnl_doit_func doit;
3261 	int sz_idx, kind;
3262 	int family;
3263 	int type;
3264 	int err;
3265 
3266 	type = nlh->nlmsg_type;
3267 	if (type > RTM_MAX)
3268 		return -EOPNOTSUPP;
3269 
3270 	type -= RTM_BASE;
3271 
3272 	/* All the messages must have at least 1 byte length */
3273 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3274 		return 0;
3275 
3276 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3277 	sz_idx = type>>2;
3278 	kind = type&3;
3279 
3280 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3281 		return -EPERM;
3282 
3283 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3284 		struct sock *rtnl;
3285 		rtnl_dumpit_func dumpit;
3286 		rtnl_calcit_func calcit;
3287 		u16 min_dump_alloc = 0;
3288 
3289 		dumpit = rtnl_get_dumpit(family, type);
3290 		if (dumpit == NULL)
3291 			return -EOPNOTSUPP;
3292 		calcit = rtnl_get_calcit(family, type);
3293 		if (calcit)
3294 			min_dump_alloc = calcit(skb, nlh);
3295 
3296 		__rtnl_unlock();
3297 		rtnl = net->rtnl;
3298 		{
3299 			struct netlink_dump_control c = {
3300 				.dump		= dumpit,
3301 				.min_dump_alloc	= min_dump_alloc,
3302 			};
3303 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3304 		}
3305 		rtnl_lock();
3306 		return err;
3307 	}
3308 
3309 	doit = rtnl_get_doit(family, type);
3310 	if (doit == NULL)
3311 		return -EOPNOTSUPP;
3312 
3313 	return doit(skb, nlh);
3314 }
3315 
3316 static void rtnetlink_rcv(struct sk_buff *skb)
3317 {
3318 	rtnl_lock();
3319 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3320 	rtnl_unlock();
3321 }
3322 
3323 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3324 {
3325 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3326 
3327 	switch (event) {
3328 	case NETDEV_UP:
3329 	case NETDEV_DOWN:
3330 	case NETDEV_PRE_UP:
3331 	case NETDEV_POST_INIT:
3332 	case NETDEV_REGISTER:
3333 	case NETDEV_CHANGE:
3334 	case NETDEV_PRE_TYPE_CHANGE:
3335 	case NETDEV_GOING_DOWN:
3336 	case NETDEV_UNREGISTER:
3337 	case NETDEV_UNREGISTER_FINAL:
3338 	case NETDEV_RELEASE:
3339 	case NETDEV_JOIN:
3340 	case NETDEV_BONDING_INFO:
3341 		break;
3342 	default:
3343 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3344 		break;
3345 	}
3346 	return NOTIFY_DONE;
3347 }
3348 
3349 static struct notifier_block rtnetlink_dev_notifier = {
3350 	.notifier_call	= rtnetlink_event,
3351 };
3352 
3353 
3354 static int __net_init rtnetlink_net_init(struct net *net)
3355 {
3356 	struct sock *sk;
3357 	struct netlink_kernel_cfg cfg = {
3358 		.groups		= RTNLGRP_MAX,
3359 		.input		= rtnetlink_rcv,
3360 		.cb_mutex	= &rtnl_mutex,
3361 		.flags		= NL_CFG_F_NONROOT_RECV,
3362 	};
3363 
3364 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3365 	if (!sk)
3366 		return -ENOMEM;
3367 	net->rtnl = sk;
3368 	return 0;
3369 }
3370 
3371 static void __net_exit rtnetlink_net_exit(struct net *net)
3372 {
3373 	netlink_kernel_release(net->rtnl);
3374 	net->rtnl = NULL;
3375 }
3376 
3377 static struct pernet_operations rtnetlink_net_ops = {
3378 	.init = rtnetlink_net_init,
3379 	.exit = rtnetlink_net_exit,
3380 };
3381 
3382 void __init rtnetlink_init(void)
3383 {
3384 	if (register_pernet_subsys(&rtnetlink_net_ops))
3385 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3386 
3387 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3388 
3389 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3390 		      rtnl_dump_ifinfo, rtnl_calcit);
3391 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3392 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3393 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3394 
3395 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3396 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3397 
3398 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3399 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3400 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3401 
3402 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3403 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3404 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3405 }
3406 
3407