xref: /openbmc/linux/net/core/rtnetlink.c (revision 17dd0ec470f97518893a5ed7160a842a35482fb4)
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/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44 
45 #include <linux/uaccess.h>
46 
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
50 #include <net/ip.h>
51 #include <net/protocol.h>
52 #include <net/arp.h>
53 #include <net/route.h>
54 #include <net/udp.h>
55 #include <net/tcp.h>
56 #include <net/sock.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
61 
62 struct rtnl_link {
63 	rtnl_doit_func		doit;
64 	rtnl_dumpit_func	dumpit;
65 	rtnl_calcit_func 	calcit;
66 };
67 
68 static DEFINE_MUTEX(rtnl_mutex);
69 
70 void rtnl_lock(void)
71 {
72 	mutex_lock(&rtnl_mutex);
73 }
74 EXPORT_SYMBOL(rtnl_lock);
75 
76 static struct sk_buff *defer_kfree_skb_list;
77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
78 {
79 	if (head && tail) {
80 		tail->next = defer_kfree_skb_list;
81 		defer_kfree_skb_list = head;
82 	}
83 }
84 EXPORT_SYMBOL(rtnl_kfree_skbs);
85 
86 void __rtnl_unlock(void)
87 {
88 	struct sk_buff *head = defer_kfree_skb_list;
89 
90 	defer_kfree_skb_list = NULL;
91 
92 	mutex_unlock(&rtnl_mutex);
93 
94 	while (head) {
95 		struct sk_buff *next = head->next;
96 
97 		kfree_skb(head);
98 		cond_resched();
99 		head = next;
100 	}
101 }
102 
103 void rtnl_unlock(void)
104 {
105 	/* This fellow will unlock it for us. */
106 	netdev_run_todo();
107 }
108 EXPORT_SYMBOL(rtnl_unlock);
109 
110 int rtnl_trylock(void)
111 {
112 	return mutex_trylock(&rtnl_mutex);
113 }
114 EXPORT_SYMBOL(rtnl_trylock);
115 
116 int rtnl_is_locked(void)
117 {
118 	return mutex_is_locked(&rtnl_mutex);
119 }
120 EXPORT_SYMBOL(rtnl_is_locked);
121 
122 #ifdef CONFIG_PROVE_LOCKING
123 bool lockdep_rtnl_is_held(void)
124 {
125 	return lockdep_is_held(&rtnl_mutex);
126 }
127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
129 
130 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
131 
132 static inline int rtm_msgindex(int msgtype)
133 {
134 	int msgindex = msgtype - RTM_BASE;
135 
136 	/*
137 	 * msgindex < 0 implies someone tried to register a netlink
138 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
139 	 * the message type has not been added to linux/rtnetlink.h
140 	 */
141 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
142 
143 	return msgindex;
144 }
145 
146 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
147 {
148 	struct rtnl_link *tab;
149 
150 	if (protocol <= RTNL_FAMILY_MAX)
151 		tab = rtnl_msg_handlers[protocol];
152 	else
153 		tab = NULL;
154 
155 	if (tab == NULL || tab[msgindex].doit == NULL)
156 		tab = rtnl_msg_handlers[PF_UNSPEC];
157 
158 	return tab[msgindex].doit;
159 }
160 
161 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
162 {
163 	struct rtnl_link *tab;
164 
165 	if (protocol <= RTNL_FAMILY_MAX)
166 		tab = rtnl_msg_handlers[protocol];
167 	else
168 		tab = NULL;
169 
170 	if (tab == NULL || tab[msgindex].dumpit == NULL)
171 		tab = rtnl_msg_handlers[PF_UNSPEC];
172 
173 	return tab[msgindex].dumpit;
174 }
175 
176 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
177 {
178 	struct rtnl_link *tab;
179 
180 	if (protocol <= RTNL_FAMILY_MAX)
181 		tab = rtnl_msg_handlers[protocol];
182 	else
183 		tab = NULL;
184 
185 	if (tab == NULL || tab[msgindex].calcit == NULL)
186 		tab = rtnl_msg_handlers[PF_UNSPEC];
187 
188 	return tab[msgindex].calcit;
189 }
190 
191 /**
192  * __rtnl_register - Register a rtnetlink message type
193  * @protocol: Protocol family or PF_UNSPEC
194  * @msgtype: rtnetlink message type
195  * @doit: Function pointer called for each request message
196  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
197  * @calcit: Function pointer to calc size of dump message
198  *
199  * Registers the specified function pointers (at least one of them has
200  * to be non-NULL) to be called whenever a request message for the
201  * specified protocol family and message type is received.
202  *
203  * The special protocol family PF_UNSPEC may be used to define fallback
204  * function pointers for the case when no entry for the specific protocol
205  * family exists.
206  *
207  * Returns 0 on success or a negative error code.
208  */
209 int __rtnl_register(int protocol, int msgtype,
210 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
211 		    rtnl_calcit_func calcit)
212 {
213 	struct rtnl_link *tab;
214 	int msgindex;
215 
216 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
217 	msgindex = rtm_msgindex(msgtype);
218 
219 	tab = rtnl_msg_handlers[protocol];
220 	if (tab == NULL) {
221 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
222 		if (tab == NULL)
223 			return -ENOBUFS;
224 
225 		rtnl_msg_handlers[protocol] = tab;
226 	}
227 
228 	if (doit)
229 		tab[msgindex].doit = doit;
230 
231 	if (dumpit)
232 		tab[msgindex].dumpit = dumpit;
233 
234 	if (calcit)
235 		tab[msgindex].calcit = calcit;
236 
237 	return 0;
238 }
239 EXPORT_SYMBOL_GPL(__rtnl_register);
240 
241 /**
242  * rtnl_register - Register a rtnetlink message type
243  *
244  * Identical to __rtnl_register() but panics on failure. This is useful
245  * as failure of this function is very unlikely, it can only happen due
246  * to lack of memory when allocating the chain to store all message
247  * handlers for a protocol. Meant for use in init functions where lack
248  * of memory implies no sense in continuing.
249  */
250 void rtnl_register(int protocol, int msgtype,
251 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
252 		   rtnl_calcit_func calcit)
253 {
254 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
255 		panic("Unable to register rtnetlink message handler, "
256 		      "protocol = %d, message type = %d\n",
257 		      protocol, msgtype);
258 }
259 EXPORT_SYMBOL_GPL(rtnl_register);
260 
261 /**
262  * rtnl_unregister - Unregister a rtnetlink message type
263  * @protocol: Protocol family or PF_UNSPEC
264  * @msgtype: rtnetlink message type
265  *
266  * Returns 0 on success or a negative error code.
267  */
268 int rtnl_unregister(int protocol, int msgtype)
269 {
270 	int msgindex;
271 
272 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
273 	msgindex = rtm_msgindex(msgtype);
274 
275 	if (rtnl_msg_handlers[protocol] == NULL)
276 		return -ENOENT;
277 
278 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
279 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
280 	rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
281 
282 	return 0;
283 }
284 EXPORT_SYMBOL_GPL(rtnl_unregister);
285 
286 /**
287  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
288  * @protocol : Protocol family or PF_UNSPEC
289  *
290  * Identical to calling rtnl_unregster() for all registered message types
291  * of a certain protocol family.
292  */
293 void rtnl_unregister_all(int protocol)
294 {
295 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
296 
297 	kfree(rtnl_msg_handlers[protocol]);
298 	rtnl_msg_handlers[protocol] = NULL;
299 }
300 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
301 
302 static LIST_HEAD(link_ops);
303 
304 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
305 {
306 	const struct rtnl_link_ops *ops;
307 
308 	list_for_each_entry(ops, &link_ops, list) {
309 		if (!strcmp(ops->kind, kind))
310 			return ops;
311 	}
312 	return NULL;
313 }
314 
315 /**
316  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
317  * @ops: struct rtnl_link_ops * to register
318  *
319  * The caller must hold the rtnl_mutex. This function should be used
320  * by drivers that create devices during module initialization. It
321  * must be called before registering the devices.
322  *
323  * Returns 0 on success or a negative error code.
324  */
325 int __rtnl_link_register(struct rtnl_link_ops *ops)
326 {
327 	if (rtnl_link_ops_get(ops->kind))
328 		return -EEXIST;
329 
330 	/* The check for setup is here because if ops
331 	 * does not have that filled up, it is not possible
332 	 * to use the ops for creating device. So do not
333 	 * fill up dellink as well. That disables rtnl_dellink.
334 	 */
335 	if (ops->setup && !ops->dellink)
336 		ops->dellink = unregister_netdevice_queue;
337 
338 	list_add_tail(&ops->list, &link_ops);
339 	return 0;
340 }
341 EXPORT_SYMBOL_GPL(__rtnl_link_register);
342 
343 /**
344  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
345  * @ops: struct rtnl_link_ops * to register
346  *
347  * Returns 0 on success or a negative error code.
348  */
349 int rtnl_link_register(struct rtnl_link_ops *ops)
350 {
351 	int err;
352 
353 	rtnl_lock();
354 	err = __rtnl_link_register(ops);
355 	rtnl_unlock();
356 	return err;
357 }
358 EXPORT_SYMBOL_GPL(rtnl_link_register);
359 
360 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
361 {
362 	struct net_device *dev;
363 	LIST_HEAD(list_kill);
364 
365 	for_each_netdev(net, dev) {
366 		if (dev->rtnl_link_ops == ops)
367 			ops->dellink(dev, &list_kill);
368 	}
369 	unregister_netdevice_many(&list_kill);
370 }
371 
372 /**
373  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
374  * @ops: struct rtnl_link_ops * to unregister
375  *
376  * The caller must hold the rtnl_mutex.
377  */
378 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
379 {
380 	struct net *net;
381 
382 	for_each_net(net) {
383 		__rtnl_kill_links(net, ops);
384 	}
385 	list_del(&ops->list);
386 }
387 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
388 
389 /* Return with the rtnl_lock held when there are no network
390  * devices unregistering in any network namespace.
391  */
392 static void rtnl_lock_unregistering_all(void)
393 {
394 	struct net *net;
395 	bool unregistering;
396 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
397 
398 	add_wait_queue(&netdev_unregistering_wq, &wait);
399 	for (;;) {
400 		unregistering = false;
401 		rtnl_lock();
402 		for_each_net(net) {
403 			if (net->dev_unreg_count > 0) {
404 				unregistering = true;
405 				break;
406 			}
407 		}
408 		if (!unregistering)
409 			break;
410 		__rtnl_unlock();
411 
412 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
413 	}
414 	remove_wait_queue(&netdev_unregistering_wq, &wait);
415 }
416 
417 /**
418  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
419  * @ops: struct rtnl_link_ops * to unregister
420  */
421 void rtnl_link_unregister(struct rtnl_link_ops *ops)
422 {
423 	/* Close the race with cleanup_net() */
424 	mutex_lock(&net_mutex);
425 	rtnl_lock_unregistering_all();
426 	__rtnl_link_unregister(ops);
427 	rtnl_unlock();
428 	mutex_unlock(&net_mutex);
429 }
430 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
431 
432 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
433 {
434 	struct net_device *master_dev;
435 	const struct rtnl_link_ops *ops;
436 
437 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
438 	if (!master_dev)
439 		return 0;
440 	ops = master_dev->rtnl_link_ops;
441 	if (!ops || !ops->get_slave_size)
442 		return 0;
443 	/* IFLA_INFO_SLAVE_DATA + nested data */
444 	return nla_total_size(sizeof(struct nlattr)) +
445 	       ops->get_slave_size(master_dev, dev);
446 }
447 
448 static size_t rtnl_link_get_size(const struct net_device *dev)
449 {
450 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
451 	size_t size;
452 
453 	if (!ops)
454 		return 0;
455 
456 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
457 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
458 
459 	if (ops->get_size)
460 		/* IFLA_INFO_DATA + nested data */
461 		size += nla_total_size(sizeof(struct nlattr)) +
462 			ops->get_size(dev);
463 
464 	if (ops->get_xstats_size)
465 		/* IFLA_INFO_XSTATS */
466 		size += nla_total_size(ops->get_xstats_size(dev));
467 
468 	size += rtnl_link_get_slave_info_data_size(dev);
469 
470 	return size;
471 }
472 
473 static LIST_HEAD(rtnl_af_ops);
474 
475 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
476 {
477 	const struct rtnl_af_ops *ops;
478 
479 	list_for_each_entry(ops, &rtnl_af_ops, list) {
480 		if (ops->family == family)
481 			return ops;
482 	}
483 
484 	return NULL;
485 }
486 
487 /**
488  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
489  * @ops: struct rtnl_af_ops * to register
490  *
491  * Returns 0 on success or a negative error code.
492  */
493 void rtnl_af_register(struct rtnl_af_ops *ops)
494 {
495 	rtnl_lock();
496 	list_add_tail(&ops->list, &rtnl_af_ops);
497 	rtnl_unlock();
498 }
499 EXPORT_SYMBOL_GPL(rtnl_af_register);
500 
501 /**
502  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
503  * @ops: struct rtnl_af_ops * to unregister
504  *
505  * The caller must hold the rtnl_mutex.
506  */
507 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
508 {
509 	list_del(&ops->list);
510 }
511 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
512 
513 /**
514  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
515  * @ops: struct rtnl_af_ops * to unregister
516  */
517 void rtnl_af_unregister(struct rtnl_af_ops *ops)
518 {
519 	rtnl_lock();
520 	__rtnl_af_unregister(ops);
521 	rtnl_unlock();
522 }
523 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
524 
525 static size_t rtnl_link_get_af_size(const struct net_device *dev,
526 				    u32 ext_filter_mask)
527 {
528 	struct rtnl_af_ops *af_ops;
529 	size_t size;
530 
531 	/* IFLA_AF_SPEC */
532 	size = nla_total_size(sizeof(struct nlattr));
533 
534 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
535 		if (af_ops->get_link_af_size) {
536 			/* AF_* + nested data */
537 			size += nla_total_size(sizeof(struct nlattr)) +
538 				af_ops->get_link_af_size(dev, ext_filter_mask);
539 		}
540 	}
541 
542 	return size;
543 }
544 
545 static bool rtnl_have_link_slave_info(const struct net_device *dev)
546 {
547 	struct net_device *master_dev;
548 
549 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
550 	if (master_dev && master_dev->rtnl_link_ops)
551 		return true;
552 	return false;
553 }
554 
555 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
556 				     const struct net_device *dev)
557 {
558 	struct net_device *master_dev;
559 	const struct rtnl_link_ops *ops;
560 	struct nlattr *slave_data;
561 	int err;
562 
563 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
564 	if (!master_dev)
565 		return 0;
566 	ops = master_dev->rtnl_link_ops;
567 	if (!ops)
568 		return 0;
569 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
570 		return -EMSGSIZE;
571 	if (ops->fill_slave_info) {
572 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
573 		if (!slave_data)
574 			return -EMSGSIZE;
575 		err = ops->fill_slave_info(skb, master_dev, dev);
576 		if (err < 0)
577 			goto err_cancel_slave_data;
578 		nla_nest_end(skb, slave_data);
579 	}
580 	return 0;
581 
582 err_cancel_slave_data:
583 	nla_nest_cancel(skb, slave_data);
584 	return err;
585 }
586 
587 static int rtnl_link_info_fill(struct sk_buff *skb,
588 			       const struct net_device *dev)
589 {
590 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
591 	struct nlattr *data;
592 	int err;
593 
594 	if (!ops)
595 		return 0;
596 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
597 		return -EMSGSIZE;
598 	if (ops->fill_xstats) {
599 		err = ops->fill_xstats(skb, dev);
600 		if (err < 0)
601 			return err;
602 	}
603 	if (ops->fill_info) {
604 		data = nla_nest_start(skb, IFLA_INFO_DATA);
605 		if (data == NULL)
606 			return -EMSGSIZE;
607 		err = ops->fill_info(skb, dev);
608 		if (err < 0)
609 			goto err_cancel_data;
610 		nla_nest_end(skb, data);
611 	}
612 	return 0;
613 
614 err_cancel_data:
615 	nla_nest_cancel(skb, data);
616 	return err;
617 }
618 
619 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
620 {
621 	struct nlattr *linkinfo;
622 	int err = -EMSGSIZE;
623 
624 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
625 	if (linkinfo == NULL)
626 		goto out;
627 
628 	err = rtnl_link_info_fill(skb, dev);
629 	if (err < 0)
630 		goto err_cancel_link;
631 
632 	err = rtnl_link_slave_info_fill(skb, dev);
633 	if (err < 0)
634 		goto err_cancel_link;
635 
636 	nla_nest_end(skb, linkinfo);
637 	return 0;
638 
639 err_cancel_link:
640 	nla_nest_cancel(skb, linkinfo);
641 out:
642 	return err;
643 }
644 
645 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
646 {
647 	struct sock *rtnl = net->rtnl;
648 	int err = 0;
649 
650 	NETLINK_CB(skb).dst_group = group;
651 	if (echo)
652 		atomic_inc(&skb->users);
653 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
654 	if (echo)
655 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
656 	return err;
657 }
658 
659 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
660 {
661 	struct sock *rtnl = net->rtnl;
662 
663 	return nlmsg_unicast(rtnl, skb, pid);
664 }
665 EXPORT_SYMBOL(rtnl_unicast);
666 
667 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
668 		 struct nlmsghdr *nlh, gfp_t flags)
669 {
670 	struct sock *rtnl = net->rtnl;
671 	int report = 0;
672 
673 	if (nlh)
674 		report = nlmsg_report(nlh);
675 
676 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
677 }
678 EXPORT_SYMBOL(rtnl_notify);
679 
680 void rtnl_set_sk_err(struct net *net, u32 group, int error)
681 {
682 	struct sock *rtnl = net->rtnl;
683 
684 	netlink_set_err(rtnl, 0, group, error);
685 }
686 EXPORT_SYMBOL(rtnl_set_sk_err);
687 
688 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
689 {
690 	struct nlattr *mx;
691 	int i, valid = 0;
692 
693 	mx = nla_nest_start(skb, RTA_METRICS);
694 	if (mx == NULL)
695 		return -ENOBUFS;
696 
697 	for (i = 0; i < RTAX_MAX; i++) {
698 		if (metrics[i]) {
699 			if (i == RTAX_CC_ALGO - 1) {
700 				char tmp[TCP_CA_NAME_MAX], *name;
701 
702 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
703 				if (!name)
704 					continue;
705 				if (nla_put_string(skb, i + 1, name))
706 					goto nla_put_failure;
707 			} else if (i == RTAX_FEATURES - 1) {
708 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
709 
710 				if (!user_features)
711 					continue;
712 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
713 				if (nla_put_u32(skb, i + 1, user_features))
714 					goto nla_put_failure;
715 			} else {
716 				if (nla_put_u32(skb, i + 1, metrics[i]))
717 					goto nla_put_failure;
718 			}
719 			valid++;
720 		}
721 	}
722 
723 	if (!valid) {
724 		nla_nest_cancel(skb, mx);
725 		return 0;
726 	}
727 
728 	return nla_nest_end(skb, mx);
729 
730 nla_put_failure:
731 	nla_nest_cancel(skb, mx);
732 	return -EMSGSIZE;
733 }
734 EXPORT_SYMBOL(rtnetlink_put_metrics);
735 
736 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
737 		       long expires, u32 error)
738 {
739 	struct rta_cacheinfo ci = {
740 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
741 		.rta_used = dst->__use,
742 		.rta_clntref = atomic_read(&(dst->__refcnt)),
743 		.rta_error = error,
744 		.rta_id =  id,
745 	};
746 
747 	if (expires) {
748 		unsigned long clock;
749 
750 		clock = jiffies_to_clock_t(abs(expires));
751 		clock = min_t(unsigned long, clock, INT_MAX);
752 		ci.rta_expires = (expires > 0) ? clock : -clock;
753 	}
754 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
755 }
756 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
757 
758 static void set_operstate(struct net_device *dev, unsigned char transition)
759 {
760 	unsigned char operstate = dev->operstate;
761 
762 	switch (transition) {
763 	case IF_OPER_UP:
764 		if ((operstate == IF_OPER_DORMANT ||
765 		     operstate == IF_OPER_UNKNOWN) &&
766 		    !netif_dormant(dev))
767 			operstate = IF_OPER_UP;
768 		break;
769 
770 	case IF_OPER_DORMANT:
771 		if (operstate == IF_OPER_UP ||
772 		    operstate == IF_OPER_UNKNOWN)
773 			operstate = IF_OPER_DORMANT;
774 		break;
775 	}
776 
777 	if (dev->operstate != operstate) {
778 		write_lock_bh(&dev_base_lock);
779 		dev->operstate = operstate;
780 		write_unlock_bh(&dev_base_lock);
781 		netdev_state_change(dev);
782 	}
783 }
784 
785 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
786 {
787 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
788 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
789 }
790 
791 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
792 					   const struct ifinfomsg *ifm)
793 {
794 	unsigned int flags = ifm->ifi_flags;
795 
796 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
797 	if (ifm->ifi_change)
798 		flags = (flags & ifm->ifi_change) |
799 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
800 
801 	return flags;
802 }
803 
804 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
805 				 const struct rtnl_link_stats64 *b)
806 {
807 	a->rx_packets = b->rx_packets;
808 	a->tx_packets = b->tx_packets;
809 	a->rx_bytes = b->rx_bytes;
810 	a->tx_bytes = b->tx_bytes;
811 	a->rx_errors = b->rx_errors;
812 	a->tx_errors = b->tx_errors;
813 	a->rx_dropped = b->rx_dropped;
814 	a->tx_dropped = b->tx_dropped;
815 
816 	a->multicast = b->multicast;
817 	a->collisions = b->collisions;
818 
819 	a->rx_length_errors = b->rx_length_errors;
820 	a->rx_over_errors = b->rx_over_errors;
821 	a->rx_crc_errors = b->rx_crc_errors;
822 	a->rx_frame_errors = b->rx_frame_errors;
823 	a->rx_fifo_errors = b->rx_fifo_errors;
824 	a->rx_missed_errors = b->rx_missed_errors;
825 
826 	a->tx_aborted_errors = b->tx_aborted_errors;
827 	a->tx_carrier_errors = b->tx_carrier_errors;
828 	a->tx_fifo_errors = b->tx_fifo_errors;
829 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
830 	a->tx_window_errors = b->tx_window_errors;
831 
832 	a->rx_compressed = b->rx_compressed;
833 	a->tx_compressed = b->tx_compressed;
834 
835 	a->rx_nohandler = b->rx_nohandler;
836 }
837 
838 /* All VF info */
839 static inline int rtnl_vfinfo_size(const struct net_device *dev,
840 				   u32 ext_filter_mask)
841 {
842 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
843 		int num_vfs = dev_num_vf(dev->dev.parent);
844 		size_t size = nla_total_size(0);
845 		size += num_vfs *
846 			(nla_total_size(0) +
847 			 nla_total_size(sizeof(struct ifla_vf_mac)) +
848 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
849 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
850 			 nla_total_size(MAX_VLAN_LIST_LEN *
851 					sizeof(struct ifla_vf_vlan_info)) +
852 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
853 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
854 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
855 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
856 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
857 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
858 			 /* IFLA_VF_STATS_RX_PACKETS */
859 			 nla_total_size_64bit(sizeof(__u64)) +
860 			 /* IFLA_VF_STATS_TX_PACKETS */
861 			 nla_total_size_64bit(sizeof(__u64)) +
862 			 /* IFLA_VF_STATS_RX_BYTES */
863 			 nla_total_size_64bit(sizeof(__u64)) +
864 			 /* IFLA_VF_STATS_TX_BYTES */
865 			 nla_total_size_64bit(sizeof(__u64)) +
866 			 /* IFLA_VF_STATS_BROADCAST */
867 			 nla_total_size_64bit(sizeof(__u64)) +
868 			 /* IFLA_VF_STATS_MULTICAST */
869 			 nla_total_size_64bit(sizeof(__u64)) +
870 			 nla_total_size(sizeof(struct ifla_vf_trust)));
871 		return size;
872 	} else
873 		return 0;
874 }
875 
876 static size_t rtnl_port_size(const struct net_device *dev,
877 			     u32 ext_filter_mask)
878 {
879 	size_t port_size = nla_total_size(4)		/* PORT_VF */
880 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
881 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
882 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
883 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
884 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
885 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
886 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
887 		+ port_size;
888 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
889 		+ port_size;
890 
891 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
892 	    !(ext_filter_mask & RTEXT_FILTER_VF))
893 		return 0;
894 	if (dev_num_vf(dev->dev.parent))
895 		return port_self_size + vf_ports_size +
896 			vf_port_size * dev_num_vf(dev->dev.parent);
897 	else
898 		return port_self_size;
899 }
900 
901 static size_t rtnl_xdp_size(void)
902 {
903 	size_t xdp_size = nla_total_size(0) +	/* nest IFLA_XDP */
904 			  nla_total_size(1) +	/* XDP_ATTACHED */
905 			  nla_total_size(4);	/* XDP_PROG_ID */
906 
907 	return xdp_size;
908 }
909 
910 static noinline size_t if_nlmsg_size(const struct net_device *dev,
911 				     u32 ext_filter_mask)
912 {
913 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
914 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
915 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
916 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
917 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
918 	       + nla_total_size(sizeof(struct rtnl_link_stats))
919 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
920 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
921 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
922 	       + nla_total_size(4) /* IFLA_TXQLEN */
923 	       + nla_total_size(4) /* IFLA_WEIGHT */
924 	       + nla_total_size(4) /* IFLA_MTU */
925 	       + nla_total_size(4) /* IFLA_LINK */
926 	       + nla_total_size(4) /* IFLA_MASTER */
927 	       + nla_total_size(1) /* IFLA_CARRIER */
928 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
929 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
930 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
931 	       + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
932 	       + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
933 	       + nla_total_size(1) /* IFLA_OPERSTATE */
934 	       + nla_total_size(1) /* IFLA_LINKMODE */
935 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
936 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
937 	       + nla_total_size(4) /* IFLA_GROUP */
938 	       + nla_total_size(ext_filter_mask
939 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
940 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
941 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
942 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
943 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
944 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
945 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
946 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
947 	       + rtnl_xdp_size() /* IFLA_XDP */
948 	       + nla_total_size(4)  /* IFLA_EVENT */
949 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
950 
951 }
952 
953 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
954 {
955 	struct nlattr *vf_ports;
956 	struct nlattr *vf_port;
957 	int vf;
958 	int err;
959 
960 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
961 	if (!vf_ports)
962 		return -EMSGSIZE;
963 
964 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
965 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
966 		if (!vf_port)
967 			goto nla_put_failure;
968 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
969 			goto nla_put_failure;
970 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
971 		if (err == -EMSGSIZE)
972 			goto nla_put_failure;
973 		if (err) {
974 			nla_nest_cancel(skb, vf_port);
975 			continue;
976 		}
977 		nla_nest_end(skb, vf_port);
978 	}
979 
980 	nla_nest_end(skb, vf_ports);
981 
982 	return 0;
983 
984 nla_put_failure:
985 	nla_nest_cancel(skb, vf_ports);
986 	return -EMSGSIZE;
987 }
988 
989 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
990 {
991 	struct nlattr *port_self;
992 	int err;
993 
994 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
995 	if (!port_self)
996 		return -EMSGSIZE;
997 
998 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
999 	if (err) {
1000 		nla_nest_cancel(skb, port_self);
1001 		return (err == -EMSGSIZE) ? err : 0;
1002 	}
1003 
1004 	nla_nest_end(skb, port_self);
1005 
1006 	return 0;
1007 }
1008 
1009 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1010 			  u32 ext_filter_mask)
1011 {
1012 	int err;
1013 
1014 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1015 	    !(ext_filter_mask & RTEXT_FILTER_VF))
1016 		return 0;
1017 
1018 	err = rtnl_port_self_fill(skb, dev);
1019 	if (err)
1020 		return err;
1021 
1022 	if (dev_num_vf(dev->dev.parent)) {
1023 		err = rtnl_vf_ports_fill(skb, dev);
1024 		if (err)
1025 			return err;
1026 	}
1027 
1028 	return 0;
1029 }
1030 
1031 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1032 {
1033 	int err;
1034 	struct netdev_phys_item_id ppid;
1035 
1036 	err = dev_get_phys_port_id(dev, &ppid);
1037 	if (err) {
1038 		if (err == -EOPNOTSUPP)
1039 			return 0;
1040 		return err;
1041 	}
1042 
1043 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1044 		return -EMSGSIZE;
1045 
1046 	return 0;
1047 }
1048 
1049 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1050 {
1051 	char name[IFNAMSIZ];
1052 	int err;
1053 
1054 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1055 	if (err) {
1056 		if (err == -EOPNOTSUPP)
1057 			return 0;
1058 		return err;
1059 	}
1060 
1061 	if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1062 		return -EMSGSIZE;
1063 
1064 	return 0;
1065 }
1066 
1067 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1068 {
1069 	int err;
1070 	struct switchdev_attr attr = {
1071 		.orig_dev = dev,
1072 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1073 		.flags = SWITCHDEV_F_NO_RECURSE,
1074 	};
1075 
1076 	err = switchdev_port_attr_get(dev, &attr);
1077 	if (err) {
1078 		if (err == -EOPNOTSUPP)
1079 			return 0;
1080 		return err;
1081 	}
1082 
1083 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1084 		    attr.u.ppid.id))
1085 		return -EMSGSIZE;
1086 
1087 	return 0;
1088 }
1089 
1090 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1091 					      struct net_device *dev)
1092 {
1093 	struct rtnl_link_stats64 *sp;
1094 	struct nlattr *attr;
1095 
1096 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1097 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1098 	if (!attr)
1099 		return -EMSGSIZE;
1100 
1101 	sp = nla_data(attr);
1102 	dev_get_stats(dev, sp);
1103 
1104 	attr = nla_reserve(skb, IFLA_STATS,
1105 			   sizeof(struct rtnl_link_stats));
1106 	if (!attr)
1107 		return -EMSGSIZE;
1108 
1109 	copy_rtnl_link_stats(nla_data(attr), sp);
1110 
1111 	return 0;
1112 }
1113 
1114 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1115 					       struct net_device *dev,
1116 					       int vfs_num,
1117 					       struct nlattr *vfinfo)
1118 {
1119 	struct ifla_vf_rss_query_en vf_rss_query_en;
1120 	struct nlattr *vf, *vfstats, *vfvlanlist;
1121 	struct ifla_vf_link_state vf_linkstate;
1122 	struct ifla_vf_vlan_info vf_vlan_info;
1123 	struct ifla_vf_spoofchk vf_spoofchk;
1124 	struct ifla_vf_tx_rate vf_tx_rate;
1125 	struct ifla_vf_stats vf_stats;
1126 	struct ifla_vf_trust vf_trust;
1127 	struct ifla_vf_vlan vf_vlan;
1128 	struct ifla_vf_rate vf_rate;
1129 	struct ifla_vf_mac vf_mac;
1130 	struct ifla_vf_info ivi;
1131 
1132 	memset(&ivi, 0, sizeof(ivi));
1133 
1134 	/* Not all SR-IOV capable drivers support the
1135 	 * spoofcheck and "RSS query enable" query.  Preset to
1136 	 * -1 so the user space tool can detect that the driver
1137 	 * didn't report anything.
1138 	 */
1139 	ivi.spoofchk = -1;
1140 	ivi.rss_query_en = -1;
1141 	ivi.trusted = -1;
1142 	/* The default value for VF link state is "auto"
1143 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1144 	 */
1145 	ivi.linkstate = 0;
1146 	/* VLAN Protocol by default is 802.1Q */
1147 	ivi.vlan_proto = htons(ETH_P_8021Q);
1148 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1149 		return 0;
1150 
1151 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1152 
1153 	vf_mac.vf =
1154 		vf_vlan.vf =
1155 		vf_vlan_info.vf =
1156 		vf_rate.vf =
1157 		vf_tx_rate.vf =
1158 		vf_spoofchk.vf =
1159 		vf_linkstate.vf =
1160 		vf_rss_query_en.vf =
1161 		vf_trust.vf = ivi.vf;
1162 
1163 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1164 	vf_vlan.vlan = ivi.vlan;
1165 	vf_vlan.qos = ivi.qos;
1166 	vf_vlan_info.vlan = ivi.vlan;
1167 	vf_vlan_info.qos = ivi.qos;
1168 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1169 	vf_tx_rate.rate = ivi.max_tx_rate;
1170 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1171 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1172 	vf_spoofchk.setting = ivi.spoofchk;
1173 	vf_linkstate.link_state = ivi.linkstate;
1174 	vf_rss_query_en.setting = ivi.rss_query_en;
1175 	vf_trust.setting = ivi.trusted;
1176 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1177 	if (!vf)
1178 		goto nla_put_vfinfo_failure;
1179 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1180 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1181 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1182 		    &vf_rate) ||
1183 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1184 		    &vf_tx_rate) ||
1185 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1186 		    &vf_spoofchk) ||
1187 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1188 		    &vf_linkstate) ||
1189 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1190 		    sizeof(vf_rss_query_en),
1191 		    &vf_rss_query_en) ||
1192 	    nla_put(skb, IFLA_VF_TRUST,
1193 		    sizeof(vf_trust), &vf_trust))
1194 		goto nla_put_vf_failure;
1195 	vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1196 	if (!vfvlanlist)
1197 		goto nla_put_vf_failure;
1198 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1199 		    &vf_vlan_info)) {
1200 		nla_nest_cancel(skb, vfvlanlist);
1201 		goto nla_put_vf_failure;
1202 	}
1203 	nla_nest_end(skb, vfvlanlist);
1204 	memset(&vf_stats, 0, sizeof(vf_stats));
1205 	if (dev->netdev_ops->ndo_get_vf_stats)
1206 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1207 						&vf_stats);
1208 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1209 	if (!vfstats)
1210 		goto nla_put_vf_failure;
1211 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1212 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1213 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1214 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1215 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1216 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1217 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1218 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1219 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1220 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1221 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1222 			      vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1223 		nla_nest_cancel(skb, vfstats);
1224 		goto nla_put_vf_failure;
1225 	}
1226 	nla_nest_end(skb, vfstats);
1227 	nla_nest_end(skb, vf);
1228 	return 0;
1229 
1230 nla_put_vf_failure:
1231 	nla_nest_cancel(skb, vf);
1232 nla_put_vfinfo_failure:
1233 	nla_nest_cancel(skb, vfinfo);
1234 	return -EMSGSIZE;
1235 }
1236 
1237 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1238 {
1239 	struct rtnl_link_ifmap map;
1240 
1241 	memset(&map, 0, sizeof(map));
1242 	map.mem_start   = dev->mem_start;
1243 	map.mem_end     = dev->mem_end;
1244 	map.base_addr   = dev->base_addr;
1245 	map.irq         = dev->irq;
1246 	map.dma         = dev->dma;
1247 	map.port        = dev->if_port;
1248 
1249 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1250 		return -EMSGSIZE;
1251 
1252 	return 0;
1253 }
1254 
1255 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1256 {
1257 	const struct net_device_ops *ops = dev->netdev_ops;
1258 	const struct bpf_prog *generic_xdp_prog;
1259 
1260 	ASSERT_RTNL();
1261 
1262 	*prog_id = 0;
1263 	generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1264 	if (generic_xdp_prog) {
1265 		*prog_id = generic_xdp_prog->aux->id;
1266 		return XDP_ATTACHED_SKB;
1267 	}
1268 	if (!ops->ndo_xdp)
1269 		return XDP_ATTACHED_NONE;
1270 
1271 	return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1272 }
1273 
1274 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1275 {
1276 	struct nlattr *xdp;
1277 	u32 prog_id;
1278 	int err;
1279 
1280 	xdp = nla_nest_start(skb, IFLA_XDP);
1281 	if (!xdp)
1282 		return -EMSGSIZE;
1283 
1284 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1285 			 rtnl_xdp_attached_mode(dev, &prog_id));
1286 	if (err)
1287 		goto err_cancel;
1288 
1289 	if (prog_id) {
1290 		err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1291 		if (err)
1292 			goto err_cancel;
1293 	}
1294 
1295 	nla_nest_end(skb, xdp);
1296 	return 0;
1297 
1298 err_cancel:
1299 	nla_nest_cancel(skb, xdp);
1300 	return err;
1301 }
1302 
1303 static u32 rtnl_get_event(unsigned long event)
1304 {
1305 	u32 rtnl_event_type = IFLA_EVENT_NONE;
1306 
1307 	switch (event) {
1308 	case NETDEV_REBOOT:
1309 		rtnl_event_type = IFLA_EVENT_REBOOT;
1310 		break;
1311 	case NETDEV_FEAT_CHANGE:
1312 		rtnl_event_type = IFLA_EVENT_FEATURES;
1313 		break;
1314 	case NETDEV_BONDING_FAILOVER:
1315 		rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1316 		break;
1317 	case NETDEV_NOTIFY_PEERS:
1318 		rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1319 		break;
1320 	case NETDEV_RESEND_IGMP:
1321 		rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1322 		break;
1323 	case NETDEV_CHANGEINFODATA:
1324 		rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1325 		break;
1326 	default:
1327 		break;
1328 	}
1329 
1330 	return rtnl_event_type;
1331 }
1332 
1333 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1334 			    int type, u32 pid, u32 seq, u32 change,
1335 			    unsigned int flags, u32 ext_filter_mask,
1336 			    u32 event)
1337 {
1338 	struct ifinfomsg *ifm;
1339 	struct nlmsghdr *nlh;
1340 	struct nlattr *af_spec;
1341 	struct rtnl_af_ops *af_ops;
1342 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1343 
1344 	ASSERT_RTNL();
1345 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1346 	if (nlh == NULL)
1347 		return -EMSGSIZE;
1348 
1349 	ifm = nlmsg_data(nlh);
1350 	ifm->ifi_family = AF_UNSPEC;
1351 	ifm->__ifi_pad = 0;
1352 	ifm->ifi_type = dev->type;
1353 	ifm->ifi_index = dev->ifindex;
1354 	ifm->ifi_flags = dev_get_flags(dev);
1355 	ifm->ifi_change = change;
1356 
1357 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1358 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1359 	    nla_put_u8(skb, IFLA_OPERSTATE,
1360 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1361 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1362 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1363 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1364 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1365 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1366 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1367 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1368 #ifdef CONFIG_RPS
1369 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1370 #endif
1371 	    (dev->ifindex != dev_get_iflink(dev) &&
1372 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1373 	    (upper_dev &&
1374 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1375 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1376 	    (dev->qdisc &&
1377 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1378 	    (dev->ifalias &&
1379 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1380 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1381 			atomic_read(&dev->carrier_changes)) ||
1382 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1383 		goto nla_put_failure;
1384 
1385 	if (event != IFLA_EVENT_NONE) {
1386 		if (nla_put_u32(skb, IFLA_EVENT, event))
1387 			goto nla_put_failure;
1388 	}
1389 
1390 	if (rtnl_fill_link_ifmap(skb, dev))
1391 		goto nla_put_failure;
1392 
1393 	if (dev->addr_len) {
1394 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1395 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1396 			goto nla_put_failure;
1397 	}
1398 
1399 	if (rtnl_phys_port_id_fill(skb, dev))
1400 		goto nla_put_failure;
1401 
1402 	if (rtnl_phys_port_name_fill(skb, dev))
1403 		goto nla_put_failure;
1404 
1405 	if (rtnl_phys_switch_id_fill(skb, dev))
1406 		goto nla_put_failure;
1407 
1408 	if (rtnl_fill_stats(skb, dev))
1409 		goto nla_put_failure;
1410 
1411 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1412 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1413 		goto nla_put_failure;
1414 
1415 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1416 	    ext_filter_mask & RTEXT_FILTER_VF) {
1417 		int i;
1418 		struct nlattr *vfinfo;
1419 		int num_vfs = dev_num_vf(dev->dev.parent);
1420 
1421 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1422 		if (!vfinfo)
1423 			goto nla_put_failure;
1424 		for (i = 0; i < num_vfs; i++) {
1425 			if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1426 				goto nla_put_failure;
1427 		}
1428 
1429 		nla_nest_end(skb, vfinfo);
1430 	}
1431 
1432 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1433 		goto nla_put_failure;
1434 
1435 	if (rtnl_xdp_fill(skb, dev))
1436 		goto nla_put_failure;
1437 
1438 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1439 		if (rtnl_link_fill(skb, dev) < 0)
1440 			goto nla_put_failure;
1441 	}
1442 
1443 	if (dev->rtnl_link_ops &&
1444 	    dev->rtnl_link_ops->get_link_net) {
1445 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1446 
1447 		if (!net_eq(dev_net(dev), link_net)) {
1448 			int id = peernet2id_alloc(dev_net(dev), link_net);
1449 
1450 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1451 				goto nla_put_failure;
1452 		}
1453 	}
1454 
1455 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1456 		goto nla_put_failure;
1457 
1458 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1459 		if (af_ops->fill_link_af) {
1460 			struct nlattr *af;
1461 			int err;
1462 
1463 			if (!(af = nla_nest_start(skb, af_ops->family)))
1464 				goto nla_put_failure;
1465 
1466 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1467 
1468 			/*
1469 			 * Caller may return ENODATA to indicate that there
1470 			 * was no data to be dumped. This is not an error, it
1471 			 * means we should trim the attribute header and
1472 			 * continue.
1473 			 */
1474 			if (err == -ENODATA)
1475 				nla_nest_cancel(skb, af);
1476 			else if (err < 0)
1477 				goto nla_put_failure;
1478 
1479 			nla_nest_end(skb, af);
1480 		}
1481 	}
1482 
1483 	nla_nest_end(skb, af_spec);
1484 
1485 	nlmsg_end(skb, nlh);
1486 	return 0;
1487 
1488 nla_put_failure:
1489 	nlmsg_cancel(skb, nlh);
1490 	return -EMSGSIZE;
1491 }
1492 
1493 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1494 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1495 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1496 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1497 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1498 	[IFLA_MTU]		= { .type = NLA_U32 },
1499 	[IFLA_LINK]		= { .type = NLA_U32 },
1500 	[IFLA_MASTER]		= { .type = NLA_U32 },
1501 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1502 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1503 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1504 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1505 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1506 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1507 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1508 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1509 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1510 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1511 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1512 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1513 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1514 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1515 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1516 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1517 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1518 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1519 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1520 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1521 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1522 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1523 	[IFLA_XDP]		= { .type = NLA_NESTED },
1524 	[IFLA_EVENT]		= { .type = NLA_U32 },
1525 	[IFLA_GROUP]		= { .type = NLA_U32 },
1526 };
1527 
1528 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1529 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1530 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1531 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1532 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1533 };
1534 
1535 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1536 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1537 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1538 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1539 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1540 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1541 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1542 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1543 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1544 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1545 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1546 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1547 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1548 };
1549 
1550 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1551 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1552 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1553 				    .len = PORT_PROFILE_MAX },
1554 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1555 				      .len = PORT_UUID_MAX },
1556 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1557 				    .len = PORT_UUID_MAX },
1558 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1559 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1560 
1561 	/* Unused, but we need to keep it here since user space could
1562 	 * fill it. It's also broken with regard to NLA_BINARY use in
1563 	 * combination with structs.
1564 	 */
1565 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1566 				    .len = sizeof(struct ifla_port_vsi) },
1567 };
1568 
1569 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1570 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1571 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1572 	[IFLA_XDP_FLAGS]	= { .type = NLA_U32 },
1573 	[IFLA_XDP_PROG_ID]	= { .type = NLA_U32 },
1574 };
1575 
1576 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1577 {
1578 	const struct rtnl_link_ops *ops = NULL;
1579 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1580 
1581 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1582 			     ifla_info_policy, NULL) < 0)
1583 		return NULL;
1584 
1585 	if (linfo[IFLA_INFO_KIND]) {
1586 		char kind[MODULE_NAME_LEN];
1587 
1588 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1589 		ops = rtnl_link_ops_get(kind);
1590 	}
1591 
1592 	return ops;
1593 }
1594 
1595 static bool link_master_filtered(struct net_device *dev, int master_idx)
1596 {
1597 	struct net_device *master;
1598 
1599 	if (!master_idx)
1600 		return false;
1601 
1602 	master = netdev_master_upper_dev_get(dev);
1603 	if (!master || master->ifindex != master_idx)
1604 		return true;
1605 
1606 	return false;
1607 }
1608 
1609 static bool link_kind_filtered(const struct net_device *dev,
1610 			       const struct rtnl_link_ops *kind_ops)
1611 {
1612 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1613 		return true;
1614 
1615 	return false;
1616 }
1617 
1618 static bool link_dump_filtered(struct net_device *dev,
1619 			       int master_idx,
1620 			       const struct rtnl_link_ops *kind_ops)
1621 {
1622 	if (link_master_filtered(dev, master_idx) ||
1623 	    link_kind_filtered(dev, kind_ops))
1624 		return true;
1625 
1626 	return false;
1627 }
1628 
1629 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1630 {
1631 	struct net *net = sock_net(skb->sk);
1632 	int h, s_h;
1633 	int idx = 0, s_idx;
1634 	struct net_device *dev;
1635 	struct hlist_head *head;
1636 	struct nlattr *tb[IFLA_MAX+1];
1637 	u32 ext_filter_mask = 0;
1638 	const struct rtnl_link_ops *kind_ops = NULL;
1639 	unsigned int flags = NLM_F_MULTI;
1640 	int master_idx = 0;
1641 	int err;
1642 	int hdrlen;
1643 
1644 	s_h = cb->args[0];
1645 	s_idx = cb->args[1];
1646 
1647 	cb->seq = net->dev_base_seq;
1648 
1649 	/* A hack to preserve kernel<->userspace interface.
1650 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1651 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1652 	 * what iproute2 < v3.9.0 used.
1653 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1654 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1655 	 */
1656 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1657 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1658 
1659 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1660 			ifla_policy, NULL) >= 0) {
1661 		if (tb[IFLA_EXT_MASK])
1662 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1663 
1664 		if (tb[IFLA_MASTER])
1665 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1666 
1667 		if (tb[IFLA_LINKINFO])
1668 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1669 
1670 		if (master_idx || kind_ops)
1671 			flags |= NLM_F_DUMP_FILTERED;
1672 	}
1673 
1674 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1675 		idx = 0;
1676 		head = &net->dev_index_head[h];
1677 		hlist_for_each_entry(dev, head, index_hlist) {
1678 			if (link_dump_filtered(dev, master_idx, kind_ops))
1679 				goto cont;
1680 			if (idx < s_idx)
1681 				goto cont;
1682 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1683 					       NETLINK_CB(cb->skb).portid,
1684 					       cb->nlh->nlmsg_seq, 0,
1685 					       flags,
1686 					       ext_filter_mask, 0);
1687 
1688 			if (err < 0) {
1689 				if (likely(skb->len))
1690 					goto out;
1691 
1692 				goto out_err;
1693 			}
1694 
1695 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1696 cont:
1697 			idx++;
1698 		}
1699 	}
1700 out:
1701 	err = skb->len;
1702 out_err:
1703 	cb->args[1] = idx;
1704 	cb->args[0] = h;
1705 
1706 	return err;
1707 }
1708 
1709 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1710 			struct netlink_ext_ack *exterr)
1711 {
1712 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1713 }
1714 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1715 
1716 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1717 {
1718 	struct net *net;
1719 	/* Examine the link attributes and figure out which
1720 	 * network namespace we are talking about.
1721 	 */
1722 	if (tb[IFLA_NET_NS_PID])
1723 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1724 	else if (tb[IFLA_NET_NS_FD])
1725 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1726 	else
1727 		net = get_net(src_net);
1728 	return net;
1729 }
1730 EXPORT_SYMBOL(rtnl_link_get_net);
1731 
1732 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1733 {
1734 	if (dev) {
1735 		if (tb[IFLA_ADDRESS] &&
1736 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1737 			return -EINVAL;
1738 
1739 		if (tb[IFLA_BROADCAST] &&
1740 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1741 			return -EINVAL;
1742 	}
1743 
1744 	if (tb[IFLA_AF_SPEC]) {
1745 		struct nlattr *af;
1746 		int rem, err;
1747 
1748 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1749 			const struct rtnl_af_ops *af_ops;
1750 
1751 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1752 				return -EAFNOSUPPORT;
1753 
1754 			if (!af_ops->set_link_af)
1755 				return -EOPNOTSUPP;
1756 
1757 			if (af_ops->validate_link_af) {
1758 				err = af_ops->validate_link_af(dev, af);
1759 				if (err < 0)
1760 					return err;
1761 			}
1762 		}
1763 	}
1764 
1765 	return 0;
1766 }
1767 
1768 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1769 				  int guid_type)
1770 {
1771 	const struct net_device_ops *ops = dev->netdev_ops;
1772 
1773 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1774 }
1775 
1776 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1777 {
1778 	if (dev->type != ARPHRD_INFINIBAND)
1779 		return -EOPNOTSUPP;
1780 
1781 	return handle_infiniband_guid(dev, ivt, guid_type);
1782 }
1783 
1784 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1785 {
1786 	const struct net_device_ops *ops = dev->netdev_ops;
1787 	int err = -EINVAL;
1788 
1789 	if (tb[IFLA_VF_MAC]) {
1790 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1791 
1792 		err = -EOPNOTSUPP;
1793 		if (ops->ndo_set_vf_mac)
1794 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1795 						  ivm->mac);
1796 		if (err < 0)
1797 			return err;
1798 	}
1799 
1800 	if (tb[IFLA_VF_VLAN]) {
1801 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1802 
1803 		err = -EOPNOTSUPP;
1804 		if (ops->ndo_set_vf_vlan)
1805 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1806 						   ivv->qos,
1807 						   htons(ETH_P_8021Q));
1808 		if (err < 0)
1809 			return err;
1810 	}
1811 
1812 	if (tb[IFLA_VF_VLAN_LIST]) {
1813 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1814 		struct nlattr *attr;
1815 		int rem, len = 0;
1816 
1817 		err = -EOPNOTSUPP;
1818 		if (!ops->ndo_set_vf_vlan)
1819 			return err;
1820 
1821 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1822 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1823 			    nla_len(attr) < NLA_HDRLEN) {
1824 				return -EINVAL;
1825 			}
1826 			if (len >= MAX_VLAN_LIST_LEN)
1827 				return -EOPNOTSUPP;
1828 			ivvl[len] = nla_data(attr);
1829 
1830 			len++;
1831 		}
1832 		if (len == 0)
1833 			return -EINVAL;
1834 
1835 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1836 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
1837 		if (err < 0)
1838 			return err;
1839 	}
1840 
1841 	if (tb[IFLA_VF_TX_RATE]) {
1842 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1843 		struct ifla_vf_info ivf;
1844 
1845 		err = -EOPNOTSUPP;
1846 		if (ops->ndo_get_vf_config)
1847 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1848 		if (err < 0)
1849 			return err;
1850 
1851 		err = -EOPNOTSUPP;
1852 		if (ops->ndo_set_vf_rate)
1853 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1854 						   ivf.min_tx_rate,
1855 						   ivt->rate);
1856 		if (err < 0)
1857 			return err;
1858 	}
1859 
1860 	if (tb[IFLA_VF_RATE]) {
1861 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1862 
1863 		err = -EOPNOTSUPP;
1864 		if (ops->ndo_set_vf_rate)
1865 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1866 						   ivt->min_tx_rate,
1867 						   ivt->max_tx_rate);
1868 		if (err < 0)
1869 			return err;
1870 	}
1871 
1872 	if (tb[IFLA_VF_SPOOFCHK]) {
1873 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1874 
1875 		err = -EOPNOTSUPP;
1876 		if (ops->ndo_set_vf_spoofchk)
1877 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1878 						       ivs->setting);
1879 		if (err < 0)
1880 			return err;
1881 	}
1882 
1883 	if (tb[IFLA_VF_LINK_STATE]) {
1884 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1885 
1886 		err = -EOPNOTSUPP;
1887 		if (ops->ndo_set_vf_link_state)
1888 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1889 							 ivl->link_state);
1890 		if (err < 0)
1891 			return err;
1892 	}
1893 
1894 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1895 		struct ifla_vf_rss_query_en *ivrssq_en;
1896 
1897 		err = -EOPNOTSUPP;
1898 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1899 		if (ops->ndo_set_vf_rss_query_en)
1900 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1901 							   ivrssq_en->setting);
1902 		if (err < 0)
1903 			return err;
1904 	}
1905 
1906 	if (tb[IFLA_VF_TRUST]) {
1907 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1908 
1909 		err = -EOPNOTSUPP;
1910 		if (ops->ndo_set_vf_trust)
1911 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1912 		if (err < 0)
1913 			return err;
1914 	}
1915 
1916 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1917 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1918 
1919 		if (!ops->ndo_set_vf_guid)
1920 			return -EOPNOTSUPP;
1921 
1922 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1923 	}
1924 
1925 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1926 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1927 
1928 		if (!ops->ndo_set_vf_guid)
1929 			return -EOPNOTSUPP;
1930 
1931 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1932 	}
1933 
1934 	return err;
1935 }
1936 
1937 static int do_set_master(struct net_device *dev, int ifindex)
1938 {
1939 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1940 	const struct net_device_ops *ops;
1941 	int err;
1942 
1943 	if (upper_dev) {
1944 		if (upper_dev->ifindex == ifindex)
1945 			return 0;
1946 		ops = upper_dev->netdev_ops;
1947 		if (ops->ndo_del_slave) {
1948 			err = ops->ndo_del_slave(upper_dev, dev);
1949 			if (err)
1950 				return err;
1951 		} else {
1952 			return -EOPNOTSUPP;
1953 		}
1954 	}
1955 
1956 	if (ifindex) {
1957 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1958 		if (!upper_dev)
1959 			return -EINVAL;
1960 		ops = upper_dev->netdev_ops;
1961 		if (ops->ndo_add_slave) {
1962 			err = ops->ndo_add_slave(upper_dev, dev);
1963 			if (err)
1964 				return err;
1965 		} else {
1966 			return -EOPNOTSUPP;
1967 		}
1968 	}
1969 	return 0;
1970 }
1971 
1972 #define DO_SETLINK_MODIFIED	0x01
1973 /* notify flag means notify + modified. */
1974 #define DO_SETLINK_NOTIFY	0x03
1975 static int do_setlink(const struct sk_buff *skb,
1976 		      struct net_device *dev, struct ifinfomsg *ifm,
1977 		      struct netlink_ext_ack *extack,
1978 		      struct nlattr **tb, char *ifname, int status)
1979 {
1980 	const struct net_device_ops *ops = dev->netdev_ops;
1981 	int err;
1982 
1983 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1984 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1985 		if (IS_ERR(net)) {
1986 			err = PTR_ERR(net);
1987 			goto errout;
1988 		}
1989 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1990 			put_net(net);
1991 			err = -EPERM;
1992 			goto errout;
1993 		}
1994 		err = dev_change_net_namespace(dev, net, ifname);
1995 		put_net(net);
1996 		if (err)
1997 			goto errout;
1998 		status |= DO_SETLINK_MODIFIED;
1999 	}
2000 
2001 	if (tb[IFLA_MAP]) {
2002 		struct rtnl_link_ifmap *u_map;
2003 		struct ifmap k_map;
2004 
2005 		if (!ops->ndo_set_config) {
2006 			err = -EOPNOTSUPP;
2007 			goto errout;
2008 		}
2009 
2010 		if (!netif_device_present(dev)) {
2011 			err = -ENODEV;
2012 			goto errout;
2013 		}
2014 
2015 		u_map = nla_data(tb[IFLA_MAP]);
2016 		k_map.mem_start = (unsigned long) u_map->mem_start;
2017 		k_map.mem_end = (unsigned long) u_map->mem_end;
2018 		k_map.base_addr = (unsigned short) u_map->base_addr;
2019 		k_map.irq = (unsigned char) u_map->irq;
2020 		k_map.dma = (unsigned char) u_map->dma;
2021 		k_map.port = (unsigned char) u_map->port;
2022 
2023 		err = ops->ndo_set_config(dev, &k_map);
2024 		if (err < 0)
2025 			goto errout;
2026 
2027 		status |= DO_SETLINK_NOTIFY;
2028 	}
2029 
2030 	if (tb[IFLA_ADDRESS]) {
2031 		struct sockaddr *sa;
2032 		int len;
2033 
2034 		len = sizeof(sa_family_t) + dev->addr_len;
2035 		sa = kmalloc(len, GFP_KERNEL);
2036 		if (!sa) {
2037 			err = -ENOMEM;
2038 			goto errout;
2039 		}
2040 		sa->sa_family = dev->type;
2041 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2042 		       dev->addr_len);
2043 		err = dev_set_mac_address(dev, sa);
2044 		kfree(sa);
2045 		if (err)
2046 			goto errout;
2047 		status |= DO_SETLINK_MODIFIED;
2048 	}
2049 
2050 	if (tb[IFLA_MTU]) {
2051 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2052 		if (err < 0)
2053 			goto errout;
2054 		status |= DO_SETLINK_MODIFIED;
2055 	}
2056 
2057 	if (tb[IFLA_GROUP]) {
2058 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2059 		status |= DO_SETLINK_NOTIFY;
2060 	}
2061 
2062 	/*
2063 	 * Interface selected by interface index but interface
2064 	 * name provided implies that a name change has been
2065 	 * requested.
2066 	 */
2067 	if (ifm->ifi_index > 0 && ifname[0]) {
2068 		err = dev_change_name(dev, ifname);
2069 		if (err < 0)
2070 			goto errout;
2071 		status |= DO_SETLINK_MODIFIED;
2072 	}
2073 
2074 	if (tb[IFLA_IFALIAS]) {
2075 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2076 				    nla_len(tb[IFLA_IFALIAS]));
2077 		if (err < 0)
2078 			goto errout;
2079 		status |= DO_SETLINK_NOTIFY;
2080 	}
2081 
2082 	if (tb[IFLA_BROADCAST]) {
2083 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2084 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2085 	}
2086 
2087 	if (ifm->ifi_flags || ifm->ifi_change) {
2088 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2089 		if (err < 0)
2090 			goto errout;
2091 	}
2092 
2093 	if (tb[IFLA_MASTER]) {
2094 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2095 		if (err)
2096 			goto errout;
2097 		status |= DO_SETLINK_MODIFIED;
2098 	}
2099 
2100 	if (tb[IFLA_CARRIER]) {
2101 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2102 		if (err)
2103 			goto errout;
2104 		status |= DO_SETLINK_MODIFIED;
2105 	}
2106 
2107 	if (tb[IFLA_TXQLEN]) {
2108 		unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2109 		unsigned int orig_len = dev->tx_queue_len;
2110 
2111 		if (dev->tx_queue_len ^ value) {
2112 			dev->tx_queue_len = value;
2113 			err = call_netdevice_notifiers(
2114 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2115 			err = notifier_to_errno(err);
2116 			if (err) {
2117 				dev->tx_queue_len = orig_len;
2118 				goto errout;
2119 			}
2120 			status |= DO_SETLINK_NOTIFY;
2121 		}
2122 	}
2123 
2124 	if (tb[IFLA_OPERSTATE])
2125 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2126 
2127 	if (tb[IFLA_LINKMODE]) {
2128 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2129 
2130 		write_lock_bh(&dev_base_lock);
2131 		if (dev->link_mode ^ value)
2132 			status |= DO_SETLINK_NOTIFY;
2133 		dev->link_mode = value;
2134 		write_unlock_bh(&dev_base_lock);
2135 	}
2136 
2137 	if (tb[IFLA_VFINFO_LIST]) {
2138 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2139 		struct nlattr *attr;
2140 		int rem;
2141 
2142 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2143 			if (nla_type(attr) != IFLA_VF_INFO ||
2144 			    nla_len(attr) < NLA_HDRLEN) {
2145 				err = -EINVAL;
2146 				goto errout;
2147 			}
2148 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2149 					       ifla_vf_policy, NULL);
2150 			if (err < 0)
2151 				goto errout;
2152 			err = do_setvfinfo(dev, vfinfo);
2153 			if (err < 0)
2154 				goto errout;
2155 			status |= DO_SETLINK_NOTIFY;
2156 		}
2157 	}
2158 	err = 0;
2159 
2160 	if (tb[IFLA_VF_PORTS]) {
2161 		struct nlattr *port[IFLA_PORT_MAX+1];
2162 		struct nlattr *attr;
2163 		int vf;
2164 		int rem;
2165 
2166 		err = -EOPNOTSUPP;
2167 		if (!ops->ndo_set_vf_port)
2168 			goto errout;
2169 
2170 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2171 			if (nla_type(attr) != IFLA_VF_PORT ||
2172 			    nla_len(attr) < NLA_HDRLEN) {
2173 				err = -EINVAL;
2174 				goto errout;
2175 			}
2176 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2177 					       ifla_port_policy, NULL);
2178 			if (err < 0)
2179 				goto errout;
2180 			if (!port[IFLA_PORT_VF]) {
2181 				err = -EOPNOTSUPP;
2182 				goto errout;
2183 			}
2184 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2185 			err = ops->ndo_set_vf_port(dev, vf, port);
2186 			if (err < 0)
2187 				goto errout;
2188 			status |= DO_SETLINK_NOTIFY;
2189 		}
2190 	}
2191 	err = 0;
2192 
2193 	if (tb[IFLA_PORT_SELF]) {
2194 		struct nlattr *port[IFLA_PORT_MAX+1];
2195 
2196 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2197 				       tb[IFLA_PORT_SELF], ifla_port_policy,
2198 				       NULL);
2199 		if (err < 0)
2200 			goto errout;
2201 
2202 		err = -EOPNOTSUPP;
2203 		if (ops->ndo_set_vf_port)
2204 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2205 		if (err < 0)
2206 			goto errout;
2207 		status |= DO_SETLINK_NOTIFY;
2208 	}
2209 
2210 	if (tb[IFLA_AF_SPEC]) {
2211 		struct nlattr *af;
2212 		int rem;
2213 
2214 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2215 			const struct rtnl_af_ops *af_ops;
2216 
2217 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2218 				BUG();
2219 
2220 			err = af_ops->set_link_af(dev, af);
2221 			if (err < 0)
2222 				goto errout;
2223 
2224 			status |= DO_SETLINK_NOTIFY;
2225 		}
2226 	}
2227 	err = 0;
2228 
2229 	if (tb[IFLA_PROTO_DOWN]) {
2230 		err = dev_change_proto_down(dev,
2231 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2232 		if (err)
2233 			goto errout;
2234 		status |= DO_SETLINK_NOTIFY;
2235 	}
2236 
2237 	if (tb[IFLA_XDP]) {
2238 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2239 		u32 xdp_flags = 0;
2240 
2241 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2242 				       ifla_xdp_policy, NULL);
2243 		if (err < 0)
2244 			goto errout;
2245 
2246 		if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2247 			err = -EINVAL;
2248 			goto errout;
2249 		}
2250 
2251 		if (xdp[IFLA_XDP_FLAGS]) {
2252 			xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2253 			if (xdp_flags & ~XDP_FLAGS_MASK) {
2254 				err = -EINVAL;
2255 				goto errout;
2256 			}
2257 			if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2258 				err = -EINVAL;
2259 				goto errout;
2260 			}
2261 		}
2262 
2263 		if (xdp[IFLA_XDP_FD]) {
2264 			err = dev_change_xdp_fd(dev, extack,
2265 						nla_get_s32(xdp[IFLA_XDP_FD]),
2266 						xdp_flags);
2267 			if (err)
2268 				goto errout;
2269 			status |= DO_SETLINK_NOTIFY;
2270 		}
2271 	}
2272 
2273 errout:
2274 	if (status & DO_SETLINK_MODIFIED) {
2275 		if (status & DO_SETLINK_NOTIFY)
2276 			netdev_state_change(dev);
2277 
2278 		if (err < 0)
2279 			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",
2280 					     dev->name);
2281 	}
2282 
2283 	return err;
2284 }
2285 
2286 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2287 			struct netlink_ext_ack *extack)
2288 {
2289 	struct net *net = sock_net(skb->sk);
2290 	struct ifinfomsg *ifm;
2291 	struct net_device *dev;
2292 	int err;
2293 	struct nlattr *tb[IFLA_MAX+1];
2294 	char ifname[IFNAMSIZ];
2295 
2296 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2297 			  extack);
2298 	if (err < 0)
2299 		goto errout;
2300 
2301 	if (tb[IFLA_IFNAME])
2302 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2303 	else
2304 		ifname[0] = '\0';
2305 
2306 	err = -EINVAL;
2307 	ifm = nlmsg_data(nlh);
2308 	if (ifm->ifi_index > 0)
2309 		dev = __dev_get_by_index(net, ifm->ifi_index);
2310 	else if (tb[IFLA_IFNAME])
2311 		dev = __dev_get_by_name(net, ifname);
2312 	else
2313 		goto errout;
2314 
2315 	if (dev == NULL) {
2316 		err = -ENODEV;
2317 		goto errout;
2318 	}
2319 
2320 	err = validate_linkmsg(dev, tb);
2321 	if (err < 0)
2322 		goto errout;
2323 
2324 	err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2325 errout:
2326 	return err;
2327 }
2328 
2329 static int rtnl_group_dellink(const struct net *net, int group)
2330 {
2331 	struct net_device *dev, *aux;
2332 	LIST_HEAD(list_kill);
2333 	bool found = false;
2334 
2335 	if (!group)
2336 		return -EPERM;
2337 
2338 	for_each_netdev(net, dev) {
2339 		if (dev->group == group) {
2340 			const struct rtnl_link_ops *ops;
2341 
2342 			found = true;
2343 			ops = dev->rtnl_link_ops;
2344 			if (!ops || !ops->dellink)
2345 				return -EOPNOTSUPP;
2346 		}
2347 	}
2348 
2349 	if (!found)
2350 		return -ENODEV;
2351 
2352 	for_each_netdev_safe(net, dev, aux) {
2353 		if (dev->group == group) {
2354 			const struct rtnl_link_ops *ops;
2355 
2356 			ops = dev->rtnl_link_ops;
2357 			ops->dellink(dev, &list_kill);
2358 		}
2359 	}
2360 	unregister_netdevice_many(&list_kill);
2361 
2362 	return 0;
2363 }
2364 
2365 int rtnl_delete_link(struct net_device *dev)
2366 {
2367 	const struct rtnl_link_ops *ops;
2368 	LIST_HEAD(list_kill);
2369 
2370 	ops = dev->rtnl_link_ops;
2371 	if (!ops || !ops->dellink)
2372 		return -EOPNOTSUPP;
2373 
2374 	ops->dellink(dev, &list_kill);
2375 	unregister_netdevice_many(&list_kill);
2376 
2377 	return 0;
2378 }
2379 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2380 
2381 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2382 			struct netlink_ext_ack *extack)
2383 {
2384 	struct net *net = sock_net(skb->sk);
2385 	struct net_device *dev;
2386 	struct ifinfomsg *ifm;
2387 	char ifname[IFNAMSIZ];
2388 	struct nlattr *tb[IFLA_MAX+1];
2389 	int err;
2390 
2391 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2392 	if (err < 0)
2393 		return err;
2394 
2395 	if (tb[IFLA_IFNAME])
2396 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2397 
2398 	ifm = nlmsg_data(nlh);
2399 	if (ifm->ifi_index > 0)
2400 		dev = __dev_get_by_index(net, ifm->ifi_index);
2401 	else if (tb[IFLA_IFNAME])
2402 		dev = __dev_get_by_name(net, ifname);
2403 	else if (tb[IFLA_GROUP])
2404 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2405 	else
2406 		return -EINVAL;
2407 
2408 	if (!dev)
2409 		return -ENODEV;
2410 
2411 	return rtnl_delete_link(dev);
2412 }
2413 
2414 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2415 {
2416 	unsigned int old_flags;
2417 	int err;
2418 
2419 	old_flags = dev->flags;
2420 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2421 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2422 		if (err < 0)
2423 			return err;
2424 	}
2425 
2426 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2427 
2428 	__dev_notify_flags(dev, old_flags, ~0U);
2429 	return 0;
2430 }
2431 EXPORT_SYMBOL(rtnl_configure_link);
2432 
2433 struct net_device *rtnl_create_link(struct net *net,
2434 	const char *ifname, unsigned char name_assign_type,
2435 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2436 {
2437 	struct net_device *dev;
2438 	unsigned int num_tx_queues = 1;
2439 	unsigned int num_rx_queues = 1;
2440 
2441 	if (tb[IFLA_NUM_TX_QUEUES])
2442 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2443 	else if (ops->get_num_tx_queues)
2444 		num_tx_queues = ops->get_num_tx_queues();
2445 
2446 	if (tb[IFLA_NUM_RX_QUEUES])
2447 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2448 	else if (ops->get_num_rx_queues)
2449 		num_rx_queues = ops->get_num_rx_queues();
2450 
2451 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2452 			       ops->setup, num_tx_queues, num_rx_queues);
2453 	if (!dev)
2454 		return ERR_PTR(-ENOMEM);
2455 
2456 	dev_net_set(dev, net);
2457 	dev->rtnl_link_ops = ops;
2458 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2459 
2460 	if (tb[IFLA_MTU])
2461 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2462 	if (tb[IFLA_ADDRESS]) {
2463 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2464 				nla_len(tb[IFLA_ADDRESS]));
2465 		dev->addr_assign_type = NET_ADDR_SET;
2466 	}
2467 	if (tb[IFLA_BROADCAST])
2468 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2469 				nla_len(tb[IFLA_BROADCAST]));
2470 	if (tb[IFLA_TXQLEN])
2471 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2472 	if (tb[IFLA_OPERSTATE])
2473 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2474 	if (tb[IFLA_LINKMODE])
2475 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2476 	if (tb[IFLA_GROUP])
2477 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2478 
2479 	return dev;
2480 }
2481 EXPORT_SYMBOL(rtnl_create_link);
2482 
2483 static int rtnl_group_changelink(const struct sk_buff *skb,
2484 		struct net *net, int group,
2485 		struct ifinfomsg *ifm,
2486 		struct netlink_ext_ack *extack,
2487 		struct nlattr **tb)
2488 {
2489 	struct net_device *dev, *aux;
2490 	int err;
2491 
2492 	for_each_netdev_safe(net, dev, aux) {
2493 		if (dev->group == group) {
2494 			err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2495 			if (err < 0)
2496 				return err;
2497 		}
2498 	}
2499 
2500 	return 0;
2501 }
2502 
2503 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2504 			struct netlink_ext_ack *extack)
2505 {
2506 	struct net *net = sock_net(skb->sk);
2507 	const struct rtnl_link_ops *ops;
2508 	const struct rtnl_link_ops *m_ops = NULL;
2509 	struct net_device *dev;
2510 	struct net_device *master_dev = NULL;
2511 	struct ifinfomsg *ifm;
2512 	char kind[MODULE_NAME_LEN];
2513 	char ifname[IFNAMSIZ];
2514 	struct nlattr *tb[IFLA_MAX+1];
2515 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2516 	unsigned char name_assign_type = NET_NAME_USER;
2517 	int err;
2518 
2519 #ifdef CONFIG_MODULES
2520 replay:
2521 #endif
2522 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2523 	if (err < 0)
2524 		return err;
2525 
2526 	if (tb[IFLA_IFNAME])
2527 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2528 	else
2529 		ifname[0] = '\0';
2530 
2531 	ifm = nlmsg_data(nlh);
2532 	if (ifm->ifi_index > 0)
2533 		dev = __dev_get_by_index(net, ifm->ifi_index);
2534 	else {
2535 		if (ifname[0])
2536 			dev = __dev_get_by_name(net, ifname);
2537 		else
2538 			dev = NULL;
2539 	}
2540 
2541 	if (dev) {
2542 		master_dev = netdev_master_upper_dev_get(dev);
2543 		if (master_dev)
2544 			m_ops = master_dev->rtnl_link_ops;
2545 	}
2546 
2547 	err = validate_linkmsg(dev, tb);
2548 	if (err < 0)
2549 		return err;
2550 
2551 	if (tb[IFLA_LINKINFO]) {
2552 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2553 				       tb[IFLA_LINKINFO], ifla_info_policy,
2554 				       NULL);
2555 		if (err < 0)
2556 			return err;
2557 	} else
2558 		memset(linkinfo, 0, sizeof(linkinfo));
2559 
2560 	if (linkinfo[IFLA_INFO_KIND]) {
2561 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2562 		ops = rtnl_link_ops_get(kind);
2563 	} else {
2564 		kind[0] = '\0';
2565 		ops = NULL;
2566 	}
2567 
2568 	if (1) {
2569 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2570 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2571 		struct nlattr **data = NULL;
2572 		struct nlattr **slave_data = NULL;
2573 		struct net *dest_net, *link_net = NULL;
2574 
2575 		if (ops) {
2576 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2577 				err = nla_parse_nested(attr, ops->maxtype,
2578 						       linkinfo[IFLA_INFO_DATA],
2579 						       ops->policy, NULL);
2580 				if (err < 0)
2581 					return err;
2582 				data = attr;
2583 			}
2584 			if (ops->validate) {
2585 				err = ops->validate(tb, data, extack);
2586 				if (err < 0)
2587 					return err;
2588 			}
2589 		}
2590 
2591 		if (m_ops) {
2592 			if (m_ops->slave_maxtype &&
2593 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2594 				err = nla_parse_nested(slave_attr,
2595 						       m_ops->slave_maxtype,
2596 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2597 						       m_ops->slave_policy,
2598 						       NULL);
2599 				if (err < 0)
2600 					return err;
2601 				slave_data = slave_attr;
2602 			}
2603 			if (m_ops->slave_validate) {
2604 				err = m_ops->slave_validate(tb, slave_data);
2605 				if (err < 0)
2606 					return err;
2607 			}
2608 		}
2609 
2610 		if (dev) {
2611 			int status = 0;
2612 
2613 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2614 				return -EEXIST;
2615 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2616 				return -EOPNOTSUPP;
2617 
2618 			if (linkinfo[IFLA_INFO_DATA]) {
2619 				if (!ops || ops != dev->rtnl_link_ops ||
2620 				    !ops->changelink)
2621 					return -EOPNOTSUPP;
2622 
2623 				err = ops->changelink(dev, tb, data, extack);
2624 				if (err < 0)
2625 					return err;
2626 				status |= DO_SETLINK_NOTIFY;
2627 			}
2628 
2629 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2630 				if (!m_ops || !m_ops->slave_changelink)
2631 					return -EOPNOTSUPP;
2632 
2633 				err = m_ops->slave_changelink(master_dev, dev,
2634 							      tb, slave_data,
2635 							      extack);
2636 				if (err < 0)
2637 					return err;
2638 				status |= DO_SETLINK_NOTIFY;
2639 			}
2640 
2641 			return do_setlink(skb, dev, ifm, extack, tb, ifname,
2642 					  status);
2643 		}
2644 
2645 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2646 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2647 				return rtnl_group_changelink(skb, net,
2648 						nla_get_u32(tb[IFLA_GROUP]),
2649 						ifm, extack, tb);
2650 			return -ENODEV;
2651 		}
2652 
2653 		if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2654 			return -EOPNOTSUPP;
2655 
2656 		if (!ops) {
2657 #ifdef CONFIG_MODULES
2658 			if (kind[0]) {
2659 				__rtnl_unlock();
2660 				request_module("rtnl-link-%s", kind);
2661 				rtnl_lock();
2662 				ops = rtnl_link_ops_get(kind);
2663 				if (ops)
2664 					goto replay;
2665 			}
2666 #endif
2667 			return -EOPNOTSUPP;
2668 		}
2669 
2670 		if (!ops->setup)
2671 			return -EOPNOTSUPP;
2672 
2673 		if (!ifname[0]) {
2674 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2675 			name_assign_type = NET_NAME_ENUM;
2676 		}
2677 
2678 		dest_net = rtnl_link_get_net(net, tb);
2679 		if (IS_ERR(dest_net))
2680 			return PTR_ERR(dest_net);
2681 
2682 		err = -EPERM;
2683 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2684 			goto out;
2685 
2686 		if (tb[IFLA_LINK_NETNSID]) {
2687 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2688 
2689 			link_net = get_net_ns_by_id(dest_net, id);
2690 			if (!link_net) {
2691 				err =  -EINVAL;
2692 				goto out;
2693 			}
2694 			err = -EPERM;
2695 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2696 				goto out;
2697 		}
2698 
2699 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2700 				       name_assign_type, ops, tb);
2701 		if (IS_ERR(dev)) {
2702 			err = PTR_ERR(dev);
2703 			goto out;
2704 		}
2705 
2706 		dev->ifindex = ifm->ifi_index;
2707 
2708 		if (ops->newlink) {
2709 			err = ops->newlink(link_net ? : net, dev, tb, data,
2710 					   extack);
2711 			/* Drivers should call free_netdev() in ->destructor
2712 			 * and unregister it on failure after registration
2713 			 * so that device could be finally freed in rtnl_unlock.
2714 			 */
2715 			if (err < 0) {
2716 				/* If device is not registered at all, free it now */
2717 				if (dev->reg_state == NETREG_UNINITIALIZED)
2718 					free_netdev(dev);
2719 				goto out;
2720 			}
2721 		} else {
2722 			err = register_netdevice(dev);
2723 			if (err < 0) {
2724 				free_netdev(dev);
2725 				goto out;
2726 			}
2727 		}
2728 		err = rtnl_configure_link(dev, ifm);
2729 		if (err < 0)
2730 			goto out_unregister;
2731 		if (link_net) {
2732 			err = dev_change_net_namespace(dev, dest_net, ifname);
2733 			if (err < 0)
2734 				goto out_unregister;
2735 		}
2736 		if (tb[IFLA_MASTER]) {
2737 			err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2738 			if (err)
2739 				goto out_unregister;
2740 		}
2741 out:
2742 		if (link_net)
2743 			put_net(link_net);
2744 		put_net(dest_net);
2745 		return err;
2746 out_unregister:
2747 		if (ops->newlink) {
2748 			LIST_HEAD(list_kill);
2749 
2750 			ops->dellink(dev, &list_kill);
2751 			unregister_netdevice_many(&list_kill);
2752 		} else {
2753 			unregister_netdevice(dev);
2754 		}
2755 		goto out;
2756 	}
2757 }
2758 
2759 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2760 			struct netlink_ext_ack *extack)
2761 {
2762 	struct net *net = sock_net(skb->sk);
2763 	struct ifinfomsg *ifm;
2764 	char ifname[IFNAMSIZ];
2765 	struct nlattr *tb[IFLA_MAX+1];
2766 	struct net_device *dev = NULL;
2767 	struct sk_buff *nskb;
2768 	int err;
2769 	u32 ext_filter_mask = 0;
2770 
2771 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2772 	if (err < 0)
2773 		return err;
2774 
2775 	if (tb[IFLA_IFNAME])
2776 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2777 
2778 	if (tb[IFLA_EXT_MASK])
2779 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2780 
2781 	ifm = nlmsg_data(nlh);
2782 	if (ifm->ifi_index > 0)
2783 		dev = __dev_get_by_index(net, ifm->ifi_index);
2784 	else if (tb[IFLA_IFNAME])
2785 		dev = __dev_get_by_name(net, ifname);
2786 	else
2787 		return -EINVAL;
2788 
2789 	if (dev == NULL)
2790 		return -ENODEV;
2791 
2792 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2793 	if (nskb == NULL)
2794 		return -ENOBUFS;
2795 
2796 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2797 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0);
2798 	if (err < 0) {
2799 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2800 		WARN_ON(err == -EMSGSIZE);
2801 		kfree_skb(nskb);
2802 	} else
2803 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2804 
2805 	return err;
2806 }
2807 
2808 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2809 {
2810 	struct net *net = sock_net(skb->sk);
2811 	struct net_device *dev;
2812 	struct nlattr *tb[IFLA_MAX+1];
2813 	u32 ext_filter_mask = 0;
2814 	u16 min_ifinfo_dump_size = 0;
2815 	int hdrlen;
2816 
2817 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2818 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2819 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2820 
2821 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2822 		if (tb[IFLA_EXT_MASK])
2823 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2824 	}
2825 
2826 	if (!ext_filter_mask)
2827 		return NLMSG_GOODSIZE;
2828 	/*
2829 	 * traverse the list of net devices and compute the minimum
2830 	 * buffer size based upon the filter mask.
2831 	 */
2832 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2833 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2834 					     if_nlmsg_size(dev,
2835 						           ext_filter_mask));
2836 	}
2837 
2838 	return nlmsg_total_size(min_ifinfo_dump_size);
2839 }
2840 
2841 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2842 {
2843 	int idx;
2844 	int s_idx = cb->family;
2845 
2846 	if (s_idx == 0)
2847 		s_idx = 1;
2848 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2849 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2850 		if (idx < s_idx || idx == PF_PACKET)
2851 			continue;
2852 		if (rtnl_msg_handlers[idx] == NULL ||
2853 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2854 			continue;
2855 		if (idx > s_idx) {
2856 			memset(&cb->args[0], 0, sizeof(cb->args));
2857 			cb->prev_seq = 0;
2858 			cb->seq = 0;
2859 		}
2860 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2861 			break;
2862 	}
2863 	cb->family = idx;
2864 
2865 	return skb->len;
2866 }
2867 
2868 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2869 				       unsigned int change,
2870 				       u32 event, gfp_t flags)
2871 {
2872 	struct net *net = dev_net(dev);
2873 	struct sk_buff *skb;
2874 	int err = -ENOBUFS;
2875 	size_t if_info_size;
2876 
2877 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2878 	if (skb == NULL)
2879 		goto errout;
2880 
2881 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event);
2882 	if (err < 0) {
2883 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2884 		WARN_ON(err == -EMSGSIZE);
2885 		kfree_skb(skb);
2886 		goto errout;
2887 	}
2888 	return skb;
2889 errout:
2890 	if (err < 0)
2891 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2892 	return NULL;
2893 }
2894 
2895 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2896 {
2897 	struct net *net = dev_net(dev);
2898 
2899 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2900 }
2901 
2902 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2903 			       unsigned int change, u32 event,
2904 			       gfp_t flags)
2905 {
2906 	struct sk_buff *skb;
2907 
2908 	if (dev->reg_state != NETREG_REGISTERED)
2909 		return;
2910 
2911 	skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags);
2912 	if (skb)
2913 		rtmsg_ifinfo_send(skb, dev, flags);
2914 }
2915 
2916 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2917 		  gfp_t flags)
2918 {
2919 	rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags);
2920 }
2921 EXPORT_SYMBOL(rtmsg_ifinfo);
2922 
2923 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2924 				   struct net_device *dev,
2925 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2926 				   int type, unsigned int flags,
2927 				   int nlflags, u16 ndm_state)
2928 {
2929 	struct nlmsghdr *nlh;
2930 	struct ndmsg *ndm;
2931 
2932 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2933 	if (!nlh)
2934 		return -EMSGSIZE;
2935 
2936 	ndm = nlmsg_data(nlh);
2937 	ndm->ndm_family  = AF_BRIDGE;
2938 	ndm->ndm_pad1	 = 0;
2939 	ndm->ndm_pad2    = 0;
2940 	ndm->ndm_flags	 = flags;
2941 	ndm->ndm_type	 = 0;
2942 	ndm->ndm_ifindex = dev->ifindex;
2943 	ndm->ndm_state   = ndm_state;
2944 
2945 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2946 		goto nla_put_failure;
2947 	if (vid)
2948 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2949 			goto nla_put_failure;
2950 
2951 	nlmsg_end(skb, nlh);
2952 	return 0;
2953 
2954 nla_put_failure:
2955 	nlmsg_cancel(skb, nlh);
2956 	return -EMSGSIZE;
2957 }
2958 
2959 static inline size_t rtnl_fdb_nlmsg_size(void)
2960 {
2961 	return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2962 	       nla_total_size(ETH_ALEN) +	/* NDA_LLADDR */
2963 	       nla_total_size(sizeof(u16)) +	/* NDA_VLAN */
2964 	       0;
2965 }
2966 
2967 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2968 			    u16 ndm_state)
2969 {
2970 	struct net *net = dev_net(dev);
2971 	struct sk_buff *skb;
2972 	int err = -ENOBUFS;
2973 
2974 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2975 	if (!skb)
2976 		goto errout;
2977 
2978 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2979 				      0, 0, type, NTF_SELF, 0, ndm_state);
2980 	if (err < 0) {
2981 		kfree_skb(skb);
2982 		goto errout;
2983 	}
2984 
2985 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2986 	return;
2987 errout:
2988 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2989 }
2990 
2991 /**
2992  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2993  */
2994 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2995 		     struct nlattr *tb[],
2996 		     struct net_device *dev,
2997 		     const unsigned char *addr, u16 vid,
2998 		     u16 flags)
2999 {
3000 	int err = -EINVAL;
3001 
3002 	/* If aging addresses are supported device will need to
3003 	 * implement its own handler for this.
3004 	 */
3005 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3006 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3007 		return err;
3008 	}
3009 
3010 	if (vid) {
3011 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3012 		return err;
3013 	}
3014 
3015 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3016 		err = dev_uc_add_excl(dev, addr);
3017 	else if (is_multicast_ether_addr(addr))
3018 		err = dev_mc_add_excl(dev, addr);
3019 
3020 	/* Only return duplicate errors if NLM_F_EXCL is set */
3021 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
3022 		err = 0;
3023 
3024 	return err;
3025 }
3026 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3027 
3028 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
3029 {
3030 	u16 vid = 0;
3031 
3032 	if (vlan_attr) {
3033 		if (nla_len(vlan_attr) != sizeof(u16)) {
3034 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
3035 			return -EINVAL;
3036 		}
3037 
3038 		vid = nla_get_u16(vlan_attr);
3039 
3040 		if (!vid || vid >= VLAN_VID_MASK) {
3041 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
3042 				vid);
3043 			return -EINVAL;
3044 		}
3045 	}
3046 	*p_vid = vid;
3047 	return 0;
3048 }
3049 
3050 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3051 			struct netlink_ext_ack *extack)
3052 {
3053 	struct net *net = sock_net(skb->sk);
3054 	struct ndmsg *ndm;
3055 	struct nlattr *tb[NDA_MAX+1];
3056 	struct net_device *dev;
3057 	u8 *addr;
3058 	u16 vid;
3059 	int err;
3060 
3061 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3062 	if (err < 0)
3063 		return err;
3064 
3065 	ndm = nlmsg_data(nlh);
3066 	if (ndm->ndm_ifindex == 0) {
3067 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3068 		return -EINVAL;
3069 	}
3070 
3071 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3072 	if (dev == NULL) {
3073 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3074 		return -ENODEV;
3075 	}
3076 
3077 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3078 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3079 		return -EINVAL;
3080 	}
3081 
3082 	addr = nla_data(tb[NDA_LLADDR]);
3083 
3084 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3085 	if (err)
3086 		return err;
3087 
3088 	err = -EOPNOTSUPP;
3089 
3090 	/* Support fdb on master device the net/bridge default case */
3091 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3092 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3093 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3094 		const struct net_device_ops *ops = br_dev->netdev_ops;
3095 
3096 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3097 				       nlh->nlmsg_flags);
3098 		if (err)
3099 			goto out;
3100 		else
3101 			ndm->ndm_flags &= ~NTF_MASTER;
3102 	}
3103 
3104 	/* Embedded bridge, macvlan, and any other device support */
3105 	if ((ndm->ndm_flags & NTF_SELF)) {
3106 		if (dev->netdev_ops->ndo_fdb_add)
3107 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3108 							   vid,
3109 							   nlh->nlmsg_flags);
3110 		else
3111 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3112 					       nlh->nlmsg_flags);
3113 
3114 		if (!err) {
3115 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3116 					ndm->ndm_state);
3117 			ndm->ndm_flags &= ~NTF_SELF;
3118 		}
3119 	}
3120 out:
3121 	return err;
3122 }
3123 
3124 /**
3125  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3126  */
3127 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3128 		     struct nlattr *tb[],
3129 		     struct net_device *dev,
3130 		     const unsigned char *addr, u16 vid)
3131 {
3132 	int err = -EINVAL;
3133 
3134 	/* If aging addresses are supported device will need to
3135 	 * implement its own handler for this.
3136 	 */
3137 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3138 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3139 		return err;
3140 	}
3141 
3142 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3143 		err = dev_uc_del(dev, addr);
3144 	else if (is_multicast_ether_addr(addr))
3145 		err = dev_mc_del(dev, addr);
3146 
3147 	return err;
3148 }
3149 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3150 
3151 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3152 			struct netlink_ext_ack *extack)
3153 {
3154 	struct net *net = sock_net(skb->sk);
3155 	struct ndmsg *ndm;
3156 	struct nlattr *tb[NDA_MAX+1];
3157 	struct net_device *dev;
3158 	int err = -EINVAL;
3159 	__u8 *addr;
3160 	u16 vid;
3161 
3162 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3163 		return -EPERM;
3164 
3165 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3166 	if (err < 0)
3167 		return err;
3168 
3169 	ndm = nlmsg_data(nlh);
3170 	if (ndm->ndm_ifindex == 0) {
3171 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3172 		return -EINVAL;
3173 	}
3174 
3175 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3176 	if (dev == NULL) {
3177 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3178 		return -ENODEV;
3179 	}
3180 
3181 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3182 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3183 		return -EINVAL;
3184 	}
3185 
3186 	addr = nla_data(tb[NDA_LLADDR]);
3187 
3188 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3189 	if (err)
3190 		return err;
3191 
3192 	err = -EOPNOTSUPP;
3193 
3194 	/* Support fdb on master device the net/bridge default case */
3195 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3196 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3197 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3198 		const struct net_device_ops *ops = br_dev->netdev_ops;
3199 
3200 		if (ops->ndo_fdb_del)
3201 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3202 
3203 		if (err)
3204 			goto out;
3205 		else
3206 			ndm->ndm_flags &= ~NTF_MASTER;
3207 	}
3208 
3209 	/* Embedded bridge, macvlan, and any other device support */
3210 	if (ndm->ndm_flags & NTF_SELF) {
3211 		if (dev->netdev_ops->ndo_fdb_del)
3212 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3213 							   vid);
3214 		else
3215 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3216 
3217 		if (!err) {
3218 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3219 					ndm->ndm_state);
3220 			ndm->ndm_flags &= ~NTF_SELF;
3221 		}
3222 	}
3223 out:
3224 	return err;
3225 }
3226 
3227 static int nlmsg_populate_fdb(struct sk_buff *skb,
3228 			      struct netlink_callback *cb,
3229 			      struct net_device *dev,
3230 			      int *idx,
3231 			      struct netdev_hw_addr_list *list)
3232 {
3233 	struct netdev_hw_addr *ha;
3234 	int err;
3235 	u32 portid, seq;
3236 
3237 	portid = NETLINK_CB(cb->skb).portid;
3238 	seq = cb->nlh->nlmsg_seq;
3239 
3240 	list_for_each_entry(ha, &list->list, list) {
3241 		if (*idx < cb->args[2])
3242 			goto skip;
3243 
3244 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3245 					      portid, seq,
3246 					      RTM_NEWNEIGH, NTF_SELF,
3247 					      NLM_F_MULTI, NUD_PERMANENT);
3248 		if (err < 0)
3249 			return err;
3250 skip:
3251 		*idx += 1;
3252 	}
3253 	return 0;
3254 }
3255 
3256 /**
3257  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3258  * @nlh: netlink message header
3259  * @dev: netdevice
3260  *
3261  * Default netdevice operation to dump the existing unicast address list.
3262  * Returns number of addresses from list put in skb.
3263  */
3264 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3265 		      struct netlink_callback *cb,
3266 		      struct net_device *dev,
3267 		      struct net_device *filter_dev,
3268 		      int *idx)
3269 {
3270 	int err;
3271 
3272 	netif_addr_lock_bh(dev);
3273 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3274 	if (err)
3275 		goto out;
3276 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3277 out:
3278 	netif_addr_unlock_bh(dev);
3279 	return err;
3280 }
3281 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3282 
3283 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3284 {
3285 	struct net_device *dev;
3286 	struct nlattr *tb[IFLA_MAX+1];
3287 	struct net_device *br_dev = NULL;
3288 	const struct net_device_ops *ops = NULL;
3289 	const struct net_device_ops *cops = NULL;
3290 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3291 	struct net *net = sock_net(skb->sk);
3292 	struct hlist_head *head;
3293 	int brport_idx = 0;
3294 	int br_idx = 0;
3295 	int h, s_h;
3296 	int idx = 0, s_idx;
3297 	int err = 0;
3298 	int fidx = 0;
3299 
3300 	err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3301 			  IFLA_MAX, ifla_policy, NULL);
3302 	if (err < 0) {
3303 		return -EINVAL;
3304 	} else if (err == 0) {
3305 		if (tb[IFLA_MASTER])
3306 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3307 	}
3308 
3309 	brport_idx = ifm->ifi_index;
3310 
3311 	if (br_idx) {
3312 		br_dev = __dev_get_by_index(net, br_idx);
3313 		if (!br_dev)
3314 			return -ENODEV;
3315 
3316 		ops = br_dev->netdev_ops;
3317 	}
3318 
3319 	s_h = cb->args[0];
3320 	s_idx = cb->args[1];
3321 
3322 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3323 		idx = 0;
3324 		head = &net->dev_index_head[h];
3325 		hlist_for_each_entry(dev, head, index_hlist) {
3326 
3327 			if (brport_idx && (dev->ifindex != brport_idx))
3328 				continue;
3329 
3330 			if (!br_idx) { /* user did not specify a specific bridge */
3331 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3332 					br_dev = netdev_master_upper_dev_get(dev);
3333 					cops = br_dev->netdev_ops;
3334 				}
3335 			} else {
3336 				if (dev != br_dev &&
3337 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3338 					continue;
3339 
3340 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3341 				    !(dev->priv_flags & IFF_EBRIDGE))
3342 					continue;
3343 				cops = ops;
3344 			}
3345 
3346 			if (idx < s_idx)
3347 				goto cont;
3348 
3349 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3350 				if (cops && cops->ndo_fdb_dump) {
3351 					err = cops->ndo_fdb_dump(skb, cb,
3352 								br_dev, dev,
3353 								&fidx);
3354 					if (err == -EMSGSIZE)
3355 						goto out;
3356 				}
3357 			}
3358 
3359 			if (dev->netdev_ops->ndo_fdb_dump)
3360 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3361 								    dev, NULL,
3362 								    &fidx);
3363 			else
3364 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3365 							&fidx);
3366 			if (err == -EMSGSIZE)
3367 				goto out;
3368 
3369 			cops = NULL;
3370 
3371 			/* reset fdb offset to 0 for rest of the interfaces */
3372 			cb->args[2] = 0;
3373 			fidx = 0;
3374 cont:
3375 			idx++;
3376 		}
3377 	}
3378 
3379 out:
3380 	cb->args[0] = h;
3381 	cb->args[1] = idx;
3382 	cb->args[2] = fidx;
3383 
3384 	return skb->len;
3385 }
3386 
3387 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3388 			       unsigned int attrnum, unsigned int flag)
3389 {
3390 	if (mask & flag)
3391 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3392 	return 0;
3393 }
3394 
3395 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3396 			    struct net_device *dev, u16 mode,
3397 			    u32 flags, u32 mask, int nlflags,
3398 			    u32 filter_mask,
3399 			    int (*vlan_fill)(struct sk_buff *skb,
3400 					     struct net_device *dev,
3401 					     u32 filter_mask))
3402 {
3403 	struct nlmsghdr *nlh;
3404 	struct ifinfomsg *ifm;
3405 	struct nlattr *br_afspec;
3406 	struct nlattr *protinfo;
3407 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3408 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3409 	int err = 0;
3410 
3411 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3412 	if (nlh == NULL)
3413 		return -EMSGSIZE;
3414 
3415 	ifm = nlmsg_data(nlh);
3416 	ifm->ifi_family = AF_BRIDGE;
3417 	ifm->__ifi_pad = 0;
3418 	ifm->ifi_type = dev->type;
3419 	ifm->ifi_index = dev->ifindex;
3420 	ifm->ifi_flags = dev_get_flags(dev);
3421 	ifm->ifi_change = 0;
3422 
3423 
3424 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3425 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3426 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3427 	    (br_dev &&
3428 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3429 	    (dev->addr_len &&
3430 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3431 	    (dev->ifindex != dev_get_iflink(dev) &&
3432 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3433 		goto nla_put_failure;
3434 
3435 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3436 	if (!br_afspec)
3437 		goto nla_put_failure;
3438 
3439 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3440 		nla_nest_cancel(skb, br_afspec);
3441 		goto nla_put_failure;
3442 	}
3443 
3444 	if (mode != BRIDGE_MODE_UNDEF) {
3445 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3446 			nla_nest_cancel(skb, br_afspec);
3447 			goto nla_put_failure;
3448 		}
3449 	}
3450 	if (vlan_fill) {
3451 		err = vlan_fill(skb, dev, filter_mask);
3452 		if (err) {
3453 			nla_nest_cancel(skb, br_afspec);
3454 			goto nla_put_failure;
3455 		}
3456 	}
3457 	nla_nest_end(skb, br_afspec);
3458 
3459 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3460 	if (!protinfo)
3461 		goto nla_put_failure;
3462 
3463 	if (brport_nla_put_flag(skb, flags, mask,
3464 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3465 	    brport_nla_put_flag(skb, flags, mask,
3466 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3467 	    brport_nla_put_flag(skb, flags, mask,
3468 				IFLA_BRPORT_FAST_LEAVE,
3469 				BR_MULTICAST_FAST_LEAVE) ||
3470 	    brport_nla_put_flag(skb, flags, mask,
3471 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3472 	    brport_nla_put_flag(skb, flags, mask,
3473 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3474 	    brport_nla_put_flag(skb, flags, mask,
3475 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3476 	    brport_nla_put_flag(skb, flags, mask,
3477 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3478 	    brport_nla_put_flag(skb, flags, mask,
3479 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3480 		nla_nest_cancel(skb, protinfo);
3481 		goto nla_put_failure;
3482 	}
3483 
3484 	nla_nest_end(skb, protinfo);
3485 
3486 	nlmsg_end(skb, nlh);
3487 	return 0;
3488 nla_put_failure:
3489 	nlmsg_cancel(skb, nlh);
3490 	return err ? err : -EMSGSIZE;
3491 }
3492 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3493 
3494 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3495 {
3496 	struct net *net = sock_net(skb->sk);
3497 	struct net_device *dev;
3498 	int idx = 0;
3499 	u32 portid = NETLINK_CB(cb->skb).portid;
3500 	u32 seq = cb->nlh->nlmsg_seq;
3501 	u32 filter_mask = 0;
3502 	int err;
3503 
3504 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3505 		struct nlattr *extfilt;
3506 
3507 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3508 					  IFLA_EXT_MASK);
3509 		if (extfilt) {
3510 			if (nla_len(extfilt) < sizeof(filter_mask))
3511 				return -EINVAL;
3512 
3513 			filter_mask = nla_get_u32(extfilt);
3514 		}
3515 	}
3516 
3517 	rcu_read_lock();
3518 	for_each_netdev_rcu(net, dev) {
3519 		const struct net_device_ops *ops = dev->netdev_ops;
3520 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3521 
3522 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3523 			if (idx >= cb->args[0]) {
3524 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3525 						skb, portid, seq, dev,
3526 						filter_mask, NLM_F_MULTI);
3527 				if (err < 0 && err != -EOPNOTSUPP) {
3528 					if (likely(skb->len))
3529 						break;
3530 
3531 					goto out_err;
3532 				}
3533 			}
3534 			idx++;
3535 		}
3536 
3537 		if (ops->ndo_bridge_getlink) {
3538 			if (idx >= cb->args[0]) {
3539 				err = ops->ndo_bridge_getlink(skb, portid,
3540 							      seq, dev,
3541 							      filter_mask,
3542 							      NLM_F_MULTI);
3543 				if (err < 0 && err != -EOPNOTSUPP) {
3544 					if (likely(skb->len))
3545 						break;
3546 
3547 					goto out_err;
3548 				}
3549 			}
3550 			idx++;
3551 		}
3552 	}
3553 	err = skb->len;
3554 out_err:
3555 	rcu_read_unlock();
3556 	cb->args[0] = idx;
3557 
3558 	return err;
3559 }
3560 
3561 static inline size_t bridge_nlmsg_size(void)
3562 {
3563 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3564 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3565 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3566 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3567 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3568 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3569 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3570 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3571 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3572 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3573 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3574 }
3575 
3576 static int rtnl_bridge_notify(struct net_device *dev)
3577 {
3578 	struct net *net = dev_net(dev);
3579 	struct sk_buff *skb;
3580 	int err = -EOPNOTSUPP;
3581 
3582 	if (!dev->netdev_ops->ndo_bridge_getlink)
3583 		return 0;
3584 
3585 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3586 	if (!skb) {
3587 		err = -ENOMEM;
3588 		goto errout;
3589 	}
3590 
3591 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3592 	if (err < 0)
3593 		goto errout;
3594 
3595 	if (!skb->len)
3596 		goto errout;
3597 
3598 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3599 	return 0;
3600 errout:
3601 	WARN_ON(err == -EMSGSIZE);
3602 	kfree_skb(skb);
3603 	if (err)
3604 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3605 	return err;
3606 }
3607 
3608 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3609 			       struct netlink_ext_ack *extack)
3610 {
3611 	struct net *net = sock_net(skb->sk);
3612 	struct ifinfomsg *ifm;
3613 	struct net_device *dev;
3614 	struct nlattr *br_spec, *attr = NULL;
3615 	int rem, err = -EOPNOTSUPP;
3616 	u16 flags = 0;
3617 	bool have_flags = false;
3618 
3619 	if (nlmsg_len(nlh) < sizeof(*ifm))
3620 		return -EINVAL;
3621 
3622 	ifm = nlmsg_data(nlh);
3623 	if (ifm->ifi_family != AF_BRIDGE)
3624 		return -EPFNOSUPPORT;
3625 
3626 	dev = __dev_get_by_index(net, ifm->ifi_index);
3627 	if (!dev) {
3628 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3629 		return -ENODEV;
3630 	}
3631 
3632 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3633 	if (br_spec) {
3634 		nla_for_each_nested(attr, br_spec, rem) {
3635 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3636 				if (nla_len(attr) < sizeof(flags))
3637 					return -EINVAL;
3638 
3639 				have_flags = true;
3640 				flags = nla_get_u16(attr);
3641 				break;
3642 			}
3643 		}
3644 	}
3645 
3646 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3647 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3648 
3649 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3650 			err = -EOPNOTSUPP;
3651 			goto out;
3652 		}
3653 
3654 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3655 		if (err)
3656 			goto out;
3657 
3658 		flags &= ~BRIDGE_FLAGS_MASTER;
3659 	}
3660 
3661 	if ((flags & BRIDGE_FLAGS_SELF)) {
3662 		if (!dev->netdev_ops->ndo_bridge_setlink)
3663 			err = -EOPNOTSUPP;
3664 		else
3665 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3666 								  flags);
3667 		if (!err) {
3668 			flags &= ~BRIDGE_FLAGS_SELF;
3669 
3670 			/* Generate event to notify upper layer of bridge
3671 			 * change
3672 			 */
3673 			err = rtnl_bridge_notify(dev);
3674 		}
3675 	}
3676 
3677 	if (have_flags)
3678 		memcpy(nla_data(attr), &flags, sizeof(flags));
3679 out:
3680 	return err;
3681 }
3682 
3683 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3684 			       struct netlink_ext_ack *extack)
3685 {
3686 	struct net *net = sock_net(skb->sk);
3687 	struct ifinfomsg *ifm;
3688 	struct net_device *dev;
3689 	struct nlattr *br_spec, *attr = NULL;
3690 	int rem, err = -EOPNOTSUPP;
3691 	u16 flags = 0;
3692 	bool have_flags = false;
3693 
3694 	if (nlmsg_len(nlh) < sizeof(*ifm))
3695 		return -EINVAL;
3696 
3697 	ifm = nlmsg_data(nlh);
3698 	if (ifm->ifi_family != AF_BRIDGE)
3699 		return -EPFNOSUPPORT;
3700 
3701 	dev = __dev_get_by_index(net, ifm->ifi_index);
3702 	if (!dev) {
3703 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3704 		return -ENODEV;
3705 	}
3706 
3707 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3708 	if (br_spec) {
3709 		nla_for_each_nested(attr, br_spec, rem) {
3710 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3711 				if (nla_len(attr) < sizeof(flags))
3712 					return -EINVAL;
3713 
3714 				have_flags = true;
3715 				flags = nla_get_u16(attr);
3716 				break;
3717 			}
3718 		}
3719 	}
3720 
3721 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3722 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3723 
3724 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3725 			err = -EOPNOTSUPP;
3726 			goto out;
3727 		}
3728 
3729 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3730 		if (err)
3731 			goto out;
3732 
3733 		flags &= ~BRIDGE_FLAGS_MASTER;
3734 	}
3735 
3736 	if ((flags & BRIDGE_FLAGS_SELF)) {
3737 		if (!dev->netdev_ops->ndo_bridge_dellink)
3738 			err = -EOPNOTSUPP;
3739 		else
3740 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3741 								  flags);
3742 
3743 		if (!err) {
3744 			flags &= ~BRIDGE_FLAGS_SELF;
3745 
3746 			/* Generate event to notify upper layer of bridge
3747 			 * change
3748 			 */
3749 			err = rtnl_bridge_notify(dev);
3750 		}
3751 	}
3752 
3753 	if (have_flags)
3754 		memcpy(nla_data(attr), &flags, sizeof(flags));
3755 out:
3756 	return err;
3757 }
3758 
3759 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3760 {
3761 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3762 	       (!idxattr || idxattr == attrid);
3763 }
3764 
3765 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3766 static int rtnl_get_offload_stats_attr_size(int attr_id)
3767 {
3768 	switch (attr_id) {
3769 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3770 		return sizeof(struct rtnl_link_stats64);
3771 	}
3772 
3773 	return 0;
3774 }
3775 
3776 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3777 				  int *prividx)
3778 {
3779 	struct nlattr *attr = NULL;
3780 	int attr_id, size;
3781 	void *attr_data;
3782 	int err;
3783 
3784 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3785 	      dev->netdev_ops->ndo_get_offload_stats))
3786 		return -ENODATA;
3787 
3788 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3789 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3790 		if (attr_id < *prividx)
3791 			continue;
3792 
3793 		size = rtnl_get_offload_stats_attr_size(attr_id);
3794 		if (!size)
3795 			continue;
3796 
3797 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3798 			continue;
3799 
3800 		attr = nla_reserve_64bit(skb, attr_id, size,
3801 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
3802 		if (!attr)
3803 			goto nla_put_failure;
3804 
3805 		attr_data = nla_data(attr);
3806 		memset(attr_data, 0, size);
3807 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3808 							     attr_data);
3809 		if (err)
3810 			goto get_offload_stats_failure;
3811 	}
3812 
3813 	if (!attr)
3814 		return -ENODATA;
3815 
3816 	*prividx = 0;
3817 	return 0;
3818 
3819 nla_put_failure:
3820 	err = -EMSGSIZE;
3821 get_offload_stats_failure:
3822 	*prividx = attr_id;
3823 	return err;
3824 }
3825 
3826 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3827 {
3828 	int nla_size = 0;
3829 	int attr_id;
3830 	int size;
3831 
3832 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3833 	      dev->netdev_ops->ndo_get_offload_stats))
3834 		return 0;
3835 
3836 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3837 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3838 		if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3839 			continue;
3840 		size = rtnl_get_offload_stats_attr_size(attr_id);
3841 		nla_size += nla_total_size_64bit(size);
3842 	}
3843 
3844 	if (nla_size != 0)
3845 		nla_size += nla_total_size(0);
3846 
3847 	return nla_size;
3848 }
3849 
3850 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3851 			       int type, u32 pid, u32 seq, u32 change,
3852 			       unsigned int flags, unsigned int filter_mask,
3853 			       int *idxattr, int *prividx)
3854 {
3855 	struct if_stats_msg *ifsm;
3856 	struct nlmsghdr *nlh;
3857 	struct nlattr *attr;
3858 	int s_prividx = *prividx;
3859 	int err;
3860 
3861 	ASSERT_RTNL();
3862 
3863 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3864 	if (!nlh)
3865 		return -EMSGSIZE;
3866 
3867 	ifsm = nlmsg_data(nlh);
3868 	ifsm->ifindex = dev->ifindex;
3869 	ifsm->filter_mask = filter_mask;
3870 
3871 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3872 		struct rtnl_link_stats64 *sp;
3873 
3874 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3875 					 sizeof(struct rtnl_link_stats64),
3876 					 IFLA_STATS_UNSPEC);
3877 		if (!attr)
3878 			goto nla_put_failure;
3879 
3880 		sp = nla_data(attr);
3881 		dev_get_stats(dev, sp);
3882 	}
3883 
3884 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3885 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3886 
3887 		if (ops && ops->fill_linkxstats) {
3888 			*idxattr = IFLA_STATS_LINK_XSTATS;
3889 			attr = nla_nest_start(skb,
3890 					      IFLA_STATS_LINK_XSTATS);
3891 			if (!attr)
3892 				goto nla_put_failure;
3893 
3894 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3895 			nla_nest_end(skb, attr);
3896 			if (err)
3897 				goto nla_put_failure;
3898 			*idxattr = 0;
3899 		}
3900 	}
3901 
3902 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3903 			     *idxattr)) {
3904 		const struct rtnl_link_ops *ops = NULL;
3905 		const struct net_device *master;
3906 
3907 		master = netdev_master_upper_dev_get(dev);
3908 		if (master)
3909 			ops = master->rtnl_link_ops;
3910 		if (ops && ops->fill_linkxstats) {
3911 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3912 			attr = nla_nest_start(skb,
3913 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3914 			if (!attr)
3915 				goto nla_put_failure;
3916 
3917 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3918 			nla_nest_end(skb, attr);
3919 			if (err)
3920 				goto nla_put_failure;
3921 			*idxattr = 0;
3922 		}
3923 	}
3924 
3925 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3926 			     *idxattr)) {
3927 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3928 		attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3929 		if (!attr)
3930 			goto nla_put_failure;
3931 
3932 		err = rtnl_get_offload_stats(skb, dev, prividx);
3933 		if (err == -ENODATA)
3934 			nla_nest_cancel(skb, attr);
3935 		else
3936 			nla_nest_end(skb, attr);
3937 
3938 		if (err && err != -ENODATA)
3939 			goto nla_put_failure;
3940 		*idxattr = 0;
3941 	}
3942 
3943 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3944 		struct rtnl_af_ops *af_ops;
3945 
3946 		*idxattr = IFLA_STATS_AF_SPEC;
3947 		attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3948 		if (!attr)
3949 			goto nla_put_failure;
3950 
3951 		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3952 			if (af_ops->fill_stats_af) {
3953 				struct nlattr *af;
3954 				int err;
3955 
3956 				af = nla_nest_start(skb, af_ops->family);
3957 				if (!af)
3958 					goto nla_put_failure;
3959 
3960 				err = af_ops->fill_stats_af(skb, dev);
3961 
3962 				if (err == -ENODATA)
3963 					nla_nest_cancel(skb, af);
3964 				else if (err < 0)
3965 					goto nla_put_failure;
3966 
3967 				nla_nest_end(skb, af);
3968 			}
3969 		}
3970 
3971 		nla_nest_end(skb, attr);
3972 
3973 		*idxattr = 0;
3974 	}
3975 
3976 	nlmsg_end(skb, nlh);
3977 
3978 	return 0;
3979 
3980 nla_put_failure:
3981 	/* not a multi message or no progress mean a real error */
3982 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3983 		nlmsg_cancel(skb, nlh);
3984 	else
3985 		nlmsg_end(skb, nlh);
3986 
3987 	return -EMSGSIZE;
3988 }
3989 
3990 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3991 				  u32 filter_mask)
3992 {
3993 	size_t size = 0;
3994 
3995 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3996 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3997 
3998 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3999 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4000 		int attr = IFLA_STATS_LINK_XSTATS;
4001 
4002 		if (ops && ops->get_linkxstats_size) {
4003 			size += nla_total_size(ops->get_linkxstats_size(dev,
4004 									attr));
4005 			/* for IFLA_STATS_LINK_XSTATS */
4006 			size += nla_total_size(0);
4007 		}
4008 	}
4009 
4010 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4011 		struct net_device *_dev = (struct net_device *)dev;
4012 		const struct rtnl_link_ops *ops = NULL;
4013 		const struct net_device *master;
4014 
4015 		/* netdev_master_upper_dev_get can't take const */
4016 		master = netdev_master_upper_dev_get(_dev);
4017 		if (master)
4018 			ops = master->rtnl_link_ops;
4019 		if (ops && ops->get_linkxstats_size) {
4020 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4021 
4022 			size += nla_total_size(ops->get_linkxstats_size(dev,
4023 									attr));
4024 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
4025 			size += nla_total_size(0);
4026 		}
4027 	}
4028 
4029 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4030 		size += rtnl_get_offload_stats_size(dev);
4031 
4032 	if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4033 		struct rtnl_af_ops *af_ops;
4034 
4035 		/* for IFLA_STATS_AF_SPEC */
4036 		size += nla_total_size(0);
4037 
4038 		list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4039 			if (af_ops->get_stats_af_size) {
4040 				size += nla_total_size(
4041 					af_ops->get_stats_af_size(dev));
4042 
4043 				/* for AF_* */
4044 				size += nla_total_size(0);
4045 			}
4046 		}
4047 	}
4048 
4049 	return size;
4050 }
4051 
4052 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4053 			  struct netlink_ext_ack *extack)
4054 {
4055 	struct net *net = sock_net(skb->sk);
4056 	struct net_device *dev = NULL;
4057 	int idxattr = 0, prividx = 0;
4058 	struct if_stats_msg *ifsm;
4059 	struct sk_buff *nskb;
4060 	u32 filter_mask;
4061 	int err;
4062 
4063 	if (nlmsg_len(nlh) < sizeof(*ifsm))
4064 		return -EINVAL;
4065 
4066 	ifsm = nlmsg_data(nlh);
4067 	if (ifsm->ifindex > 0)
4068 		dev = __dev_get_by_index(net, ifsm->ifindex);
4069 	else
4070 		return -EINVAL;
4071 
4072 	if (!dev)
4073 		return -ENODEV;
4074 
4075 	filter_mask = ifsm->filter_mask;
4076 	if (!filter_mask)
4077 		return -EINVAL;
4078 
4079 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4080 	if (!nskb)
4081 		return -ENOBUFS;
4082 
4083 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4084 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4085 				  0, filter_mask, &idxattr, &prividx);
4086 	if (err < 0) {
4087 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4088 		WARN_ON(err == -EMSGSIZE);
4089 		kfree_skb(nskb);
4090 	} else {
4091 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4092 	}
4093 
4094 	return err;
4095 }
4096 
4097 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4098 {
4099 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
4100 	struct net *net = sock_net(skb->sk);
4101 	unsigned int flags = NLM_F_MULTI;
4102 	struct if_stats_msg *ifsm;
4103 	struct hlist_head *head;
4104 	struct net_device *dev;
4105 	u32 filter_mask = 0;
4106 	int idx = 0;
4107 
4108 	s_h = cb->args[0];
4109 	s_idx = cb->args[1];
4110 	s_idxattr = cb->args[2];
4111 	s_prividx = cb->args[3];
4112 
4113 	cb->seq = net->dev_base_seq;
4114 
4115 	if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4116 		return -EINVAL;
4117 
4118 	ifsm = nlmsg_data(cb->nlh);
4119 	filter_mask = ifsm->filter_mask;
4120 	if (!filter_mask)
4121 		return -EINVAL;
4122 
4123 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4124 		idx = 0;
4125 		head = &net->dev_index_head[h];
4126 		hlist_for_each_entry(dev, head, index_hlist) {
4127 			if (idx < s_idx)
4128 				goto cont;
4129 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4130 						  NETLINK_CB(cb->skb).portid,
4131 						  cb->nlh->nlmsg_seq, 0,
4132 						  flags, filter_mask,
4133 						  &s_idxattr, &s_prividx);
4134 			/* If we ran out of room on the first message,
4135 			 * we're in trouble
4136 			 */
4137 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4138 
4139 			if (err < 0)
4140 				goto out;
4141 			s_prividx = 0;
4142 			s_idxattr = 0;
4143 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4144 cont:
4145 			idx++;
4146 		}
4147 	}
4148 out:
4149 	cb->args[3] = s_prividx;
4150 	cb->args[2] = s_idxattr;
4151 	cb->args[1] = idx;
4152 	cb->args[0] = h;
4153 
4154 	return skb->len;
4155 }
4156 
4157 /* Process one rtnetlink message. */
4158 
4159 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4160 			     struct netlink_ext_ack *extack)
4161 {
4162 	struct net *net = sock_net(skb->sk);
4163 	rtnl_doit_func doit;
4164 	int kind;
4165 	int family;
4166 	int type;
4167 	int err;
4168 
4169 	type = nlh->nlmsg_type;
4170 	if (type > RTM_MAX)
4171 		return -EOPNOTSUPP;
4172 
4173 	type -= RTM_BASE;
4174 
4175 	/* All the messages must have at least 1 byte length */
4176 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4177 		return 0;
4178 
4179 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4180 	kind = type&3;
4181 
4182 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4183 		return -EPERM;
4184 
4185 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4186 		struct sock *rtnl;
4187 		rtnl_dumpit_func dumpit;
4188 		rtnl_calcit_func calcit;
4189 		u16 min_dump_alloc = 0;
4190 
4191 		dumpit = rtnl_get_dumpit(family, type);
4192 		if (dumpit == NULL)
4193 			return -EOPNOTSUPP;
4194 		calcit = rtnl_get_calcit(family, type);
4195 		if (calcit)
4196 			min_dump_alloc = calcit(skb, nlh);
4197 
4198 		__rtnl_unlock();
4199 		rtnl = net->rtnl;
4200 		{
4201 			struct netlink_dump_control c = {
4202 				.dump		= dumpit,
4203 				.min_dump_alloc	= min_dump_alloc,
4204 			};
4205 			err = netlink_dump_start(rtnl, skb, nlh, &c);
4206 		}
4207 		rtnl_lock();
4208 		return err;
4209 	}
4210 
4211 	doit = rtnl_get_doit(family, type);
4212 	if (doit == NULL)
4213 		return -EOPNOTSUPP;
4214 
4215 	return doit(skb, nlh, extack);
4216 }
4217 
4218 static void rtnetlink_rcv(struct sk_buff *skb)
4219 {
4220 	rtnl_lock();
4221 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4222 	rtnl_unlock();
4223 }
4224 
4225 static int rtnetlink_bind(struct net *net, int group)
4226 {
4227 	switch (group) {
4228 	case RTNLGRP_IPV4_MROUTE_R:
4229 	case RTNLGRP_IPV6_MROUTE_R:
4230 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4231 			return -EPERM;
4232 		break;
4233 	}
4234 	return 0;
4235 }
4236 
4237 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4238 {
4239 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4240 
4241 	switch (event) {
4242 	case NETDEV_REBOOT:
4243 	case NETDEV_CHANGENAME:
4244 	case NETDEV_FEAT_CHANGE:
4245 	case NETDEV_BONDING_FAILOVER:
4246 	case NETDEV_NOTIFY_PEERS:
4247 	case NETDEV_RESEND_IGMP:
4248 	case NETDEV_CHANGEINFODATA:
4249 		rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4250 				   GFP_KERNEL);
4251 		break;
4252 	default:
4253 		break;
4254 	}
4255 	return NOTIFY_DONE;
4256 }
4257 
4258 static struct notifier_block rtnetlink_dev_notifier = {
4259 	.notifier_call	= rtnetlink_event,
4260 };
4261 
4262 
4263 static int __net_init rtnetlink_net_init(struct net *net)
4264 {
4265 	struct sock *sk;
4266 	struct netlink_kernel_cfg cfg = {
4267 		.groups		= RTNLGRP_MAX,
4268 		.input		= rtnetlink_rcv,
4269 		.cb_mutex	= &rtnl_mutex,
4270 		.flags		= NL_CFG_F_NONROOT_RECV,
4271 		.bind		= rtnetlink_bind,
4272 	};
4273 
4274 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4275 	if (!sk)
4276 		return -ENOMEM;
4277 	net->rtnl = sk;
4278 	return 0;
4279 }
4280 
4281 static void __net_exit rtnetlink_net_exit(struct net *net)
4282 {
4283 	netlink_kernel_release(net->rtnl);
4284 	net->rtnl = NULL;
4285 }
4286 
4287 static struct pernet_operations rtnetlink_net_ops = {
4288 	.init = rtnetlink_net_init,
4289 	.exit = rtnetlink_net_exit,
4290 };
4291 
4292 void __init rtnetlink_init(void)
4293 {
4294 	if (register_pernet_subsys(&rtnetlink_net_ops))
4295 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
4296 
4297 	register_netdevice_notifier(&rtnetlink_dev_notifier);
4298 
4299 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4300 		      rtnl_dump_ifinfo, rtnl_calcit);
4301 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4302 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4303 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4304 
4305 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4306 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4307 	rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4308 
4309 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4310 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4311 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4312 
4313 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4314 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4315 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4316 
4317 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4318 		      NULL);
4319 }
4320