xref: /openbmc/linux/net/core/rtnetlink.c (revision 7e75f74a171a8146cc3ee92d5562878b40c25fb5)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  *
15  *	Fixes:
16  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17  */
18 
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
29 #include <linux/mm.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
42 
43 #include <asm/uaccess.h>
44 
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
48 #include <net/ip.h>
49 #include <net/protocol.h>
50 #include <net/arp.h>
51 #include <net/route.h>
52 #include <net/udp.h>
53 #include <net/tcp.h>
54 #include <net/sock.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
59 
60 struct rtnl_link {
61 	rtnl_doit_func		doit;
62 	rtnl_dumpit_func	dumpit;
63 	rtnl_calcit_func 	calcit;
64 };
65 
66 static DEFINE_MUTEX(rtnl_mutex);
67 
68 void rtnl_lock(void)
69 {
70 	mutex_lock(&rtnl_mutex);
71 }
72 EXPORT_SYMBOL(rtnl_lock);
73 
74 static struct sk_buff *defer_kfree_skb_list;
75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
76 {
77 	if (head && tail) {
78 		tail->next = defer_kfree_skb_list;
79 		defer_kfree_skb_list = head;
80 	}
81 }
82 EXPORT_SYMBOL(rtnl_kfree_skbs);
83 
84 void __rtnl_unlock(void)
85 {
86 	struct sk_buff *head = defer_kfree_skb_list;
87 
88 	defer_kfree_skb_list = NULL;
89 
90 	mutex_unlock(&rtnl_mutex);
91 
92 	while (head) {
93 		struct sk_buff *next = head->next;
94 
95 		kfree_skb(head);
96 		cond_resched();
97 		head = next;
98 	}
99 }
100 
101 void rtnl_unlock(void)
102 {
103 	/* This fellow will unlock it for us. */
104 	netdev_run_todo();
105 }
106 EXPORT_SYMBOL(rtnl_unlock);
107 
108 int rtnl_trylock(void)
109 {
110 	return mutex_trylock(&rtnl_mutex);
111 }
112 EXPORT_SYMBOL(rtnl_trylock);
113 
114 int rtnl_is_locked(void)
115 {
116 	return mutex_is_locked(&rtnl_mutex);
117 }
118 EXPORT_SYMBOL(rtnl_is_locked);
119 
120 #ifdef CONFIG_PROVE_LOCKING
121 bool lockdep_rtnl_is_held(void)
122 {
123 	return lockdep_is_held(&rtnl_mutex);
124 }
125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
127 
128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
129 
130 static inline int rtm_msgindex(int msgtype)
131 {
132 	int msgindex = msgtype - RTM_BASE;
133 
134 	/*
135 	 * msgindex < 0 implies someone tried to register a netlink
136 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
137 	 * the message type has not been added to linux/rtnetlink.h
138 	 */
139 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
140 
141 	return msgindex;
142 }
143 
144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
145 {
146 	struct rtnl_link *tab;
147 
148 	if (protocol <= RTNL_FAMILY_MAX)
149 		tab = rtnl_msg_handlers[protocol];
150 	else
151 		tab = NULL;
152 
153 	if (tab == NULL || tab[msgindex].doit == NULL)
154 		tab = rtnl_msg_handlers[PF_UNSPEC];
155 
156 	return tab[msgindex].doit;
157 }
158 
159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
160 {
161 	struct rtnl_link *tab;
162 
163 	if (protocol <= RTNL_FAMILY_MAX)
164 		tab = rtnl_msg_handlers[protocol];
165 	else
166 		tab = NULL;
167 
168 	if (tab == NULL || tab[msgindex].dumpit == NULL)
169 		tab = rtnl_msg_handlers[PF_UNSPEC];
170 
171 	return tab[msgindex].dumpit;
172 }
173 
174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
175 {
176 	struct rtnl_link *tab;
177 
178 	if (protocol <= RTNL_FAMILY_MAX)
179 		tab = rtnl_msg_handlers[protocol];
180 	else
181 		tab = NULL;
182 
183 	if (tab == NULL || tab[msgindex].calcit == NULL)
184 		tab = rtnl_msg_handlers[PF_UNSPEC];
185 
186 	return tab[msgindex].calcit;
187 }
188 
189 /**
190  * __rtnl_register - Register a rtnetlink message type
191  * @protocol: Protocol family or PF_UNSPEC
192  * @msgtype: rtnetlink message type
193  * @doit: Function pointer called for each request message
194  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
195  * @calcit: Function pointer to calc size of dump message
196  *
197  * Registers the specified function pointers (at least one of them has
198  * to be non-NULL) to be called whenever a request message for the
199  * specified protocol family and message type is received.
200  *
201  * The special protocol family PF_UNSPEC may be used to define fallback
202  * function pointers for the case when no entry for the specific protocol
203  * family exists.
204  *
205  * Returns 0 on success or a negative error code.
206  */
207 int __rtnl_register(int protocol, int msgtype,
208 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
209 		    rtnl_calcit_func calcit)
210 {
211 	struct rtnl_link *tab;
212 	int msgindex;
213 
214 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
215 	msgindex = rtm_msgindex(msgtype);
216 
217 	tab = rtnl_msg_handlers[protocol];
218 	if (tab == NULL) {
219 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
220 		if (tab == NULL)
221 			return -ENOBUFS;
222 
223 		rtnl_msg_handlers[protocol] = tab;
224 	}
225 
226 	if (doit)
227 		tab[msgindex].doit = doit;
228 
229 	if (dumpit)
230 		tab[msgindex].dumpit = dumpit;
231 
232 	if (calcit)
233 		tab[msgindex].calcit = calcit;
234 
235 	return 0;
236 }
237 EXPORT_SYMBOL_GPL(__rtnl_register);
238 
239 /**
240  * rtnl_register - Register a rtnetlink message type
241  *
242  * Identical to __rtnl_register() but panics on failure. This is useful
243  * as failure of this function is very unlikely, it can only happen due
244  * to lack of memory when allocating the chain to store all message
245  * handlers for a protocol. Meant for use in init functions where lack
246  * of memory implies no sense in continuing.
247  */
248 void rtnl_register(int protocol, int msgtype,
249 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
250 		   rtnl_calcit_func calcit)
251 {
252 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
253 		panic("Unable to register rtnetlink message handler, "
254 		      "protocol = %d, message type = %d\n",
255 		      protocol, msgtype);
256 }
257 EXPORT_SYMBOL_GPL(rtnl_register);
258 
259 /**
260  * rtnl_unregister - Unregister a rtnetlink message type
261  * @protocol: Protocol family or PF_UNSPEC
262  * @msgtype: rtnetlink message type
263  *
264  * Returns 0 on success or a negative error code.
265  */
266 int rtnl_unregister(int protocol, int msgtype)
267 {
268 	int msgindex;
269 
270 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 	msgindex = rtm_msgindex(msgtype);
272 
273 	if (rtnl_msg_handlers[protocol] == NULL)
274 		return -ENOENT;
275 
276 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
277 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
278 	rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
279 
280 	return 0;
281 }
282 EXPORT_SYMBOL_GPL(rtnl_unregister);
283 
284 /**
285  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
286  * @protocol : Protocol family or PF_UNSPEC
287  *
288  * Identical to calling rtnl_unregster() for all registered message types
289  * of a certain protocol family.
290  */
291 void rtnl_unregister_all(int protocol)
292 {
293 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
294 
295 	kfree(rtnl_msg_handlers[protocol]);
296 	rtnl_msg_handlers[protocol] = NULL;
297 }
298 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
299 
300 static LIST_HEAD(link_ops);
301 
302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
303 {
304 	const struct rtnl_link_ops *ops;
305 
306 	list_for_each_entry(ops, &link_ops, list) {
307 		if (!strcmp(ops->kind, kind))
308 			return ops;
309 	}
310 	return NULL;
311 }
312 
313 /**
314  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
315  * @ops: struct rtnl_link_ops * to register
316  *
317  * The caller must hold the rtnl_mutex. This function should be used
318  * by drivers that create devices during module initialization. It
319  * must be called before registering the devices.
320  *
321  * Returns 0 on success or a negative error code.
322  */
323 int __rtnl_link_register(struct rtnl_link_ops *ops)
324 {
325 	if (rtnl_link_ops_get(ops->kind))
326 		return -EEXIST;
327 
328 	/* The check for setup is here because if ops
329 	 * does not have that filled up, it is not possible
330 	 * to use the ops for creating device. So do not
331 	 * fill up dellink as well. That disables rtnl_dellink.
332 	 */
333 	if (ops->setup && !ops->dellink)
334 		ops->dellink = unregister_netdevice_queue;
335 
336 	list_add_tail(&ops->list, &link_ops);
337 	return 0;
338 }
339 EXPORT_SYMBOL_GPL(__rtnl_link_register);
340 
341 /**
342  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
343  * @ops: struct rtnl_link_ops * to register
344  *
345  * Returns 0 on success or a negative error code.
346  */
347 int rtnl_link_register(struct rtnl_link_ops *ops)
348 {
349 	int err;
350 
351 	rtnl_lock();
352 	err = __rtnl_link_register(ops);
353 	rtnl_unlock();
354 	return err;
355 }
356 EXPORT_SYMBOL_GPL(rtnl_link_register);
357 
358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
359 {
360 	struct net_device *dev;
361 	LIST_HEAD(list_kill);
362 
363 	for_each_netdev(net, dev) {
364 		if (dev->rtnl_link_ops == ops)
365 			ops->dellink(dev, &list_kill);
366 	}
367 	unregister_netdevice_many(&list_kill);
368 }
369 
370 /**
371  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
372  * @ops: struct rtnl_link_ops * to unregister
373  *
374  * The caller must hold the rtnl_mutex.
375  */
376 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
377 {
378 	struct net *net;
379 
380 	for_each_net(net) {
381 		__rtnl_kill_links(net, ops);
382 	}
383 	list_del(&ops->list);
384 }
385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
386 
387 /* Return with the rtnl_lock held when there are no network
388  * devices unregistering in any network namespace.
389  */
390 static void rtnl_lock_unregistering_all(void)
391 {
392 	struct net *net;
393 	bool unregistering;
394 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
395 
396 	add_wait_queue(&netdev_unregistering_wq, &wait);
397 	for (;;) {
398 		unregistering = false;
399 		rtnl_lock();
400 		for_each_net(net) {
401 			if (net->dev_unreg_count > 0) {
402 				unregistering = true;
403 				break;
404 			}
405 		}
406 		if (!unregistering)
407 			break;
408 		__rtnl_unlock();
409 
410 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
411 	}
412 	remove_wait_queue(&netdev_unregistering_wq, &wait);
413 }
414 
415 /**
416  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
417  * @ops: struct rtnl_link_ops * to unregister
418  */
419 void rtnl_link_unregister(struct rtnl_link_ops *ops)
420 {
421 	/* Close the race with cleanup_net() */
422 	mutex_lock(&net_mutex);
423 	rtnl_lock_unregistering_all();
424 	__rtnl_link_unregister(ops);
425 	rtnl_unlock();
426 	mutex_unlock(&net_mutex);
427 }
428 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
429 
430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
431 {
432 	struct net_device *master_dev;
433 	const struct rtnl_link_ops *ops;
434 
435 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
436 	if (!master_dev)
437 		return 0;
438 	ops = master_dev->rtnl_link_ops;
439 	if (!ops || !ops->get_slave_size)
440 		return 0;
441 	/* IFLA_INFO_SLAVE_DATA + nested data */
442 	return nla_total_size(sizeof(struct nlattr)) +
443 	       ops->get_slave_size(master_dev, dev);
444 }
445 
446 static size_t rtnl_link_get_size(const struct net_device *dev)
447 {
448 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
449 	size_t size;
450 
451 	if (!ops)
452 		return 0;
453 
454 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
455 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
456 
457 	if (ops->get_size)
458 		/* IFLA_INFO_DATA + nested data */
459 		size += nla_total_size(sizeof(struct nlattr)) +
460 			ops->get_size(dev);
461 
462 	if (ops->get_xstats_size)
463 		/* IFLA_INFO_XSTATS */
464 		size += nla_total_size(ops->get_xstats_size(dev));
465 
466 	size += rtnl_link_get_slave_info_data_size(dev);
467 
468 	return size;
469 }
470 
471 static LIST_HEAD(rtnl_af_ops);
472 
473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
474 {
475 	const struct rtnl_af_ops *ops;
476 
477 	list_for_each_entry(ops, &rtnl_af_ops, list) {
478 		if (ops->family == family)
479 			return ops;
480 	}
481 
482 	return NULL;
483 }
484 
485 /**
486  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
487  * @ops: struct rtnl_af_ops * to register
488  *
489  * Returns 0 on success or a negative error code.
490  */
491 void rtnl_af_register(struct rtnl_af_ops *ops)
492 {
493 	rtnl_lock();
494 	list_add_tail(&ops->list, &rtnl_af_ops);
495 	rtnl_unlock();
496 }
497 EXPORT_SYMBOL_GPL(rtnl_af_register);
498 
499 /**
500  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
501  * @ops: struct rtnl_af_ops * to unregister
502  *
503  * The caller must hold the rtnl_mutex.
504  */
505 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
506 {
507 	list_del(&ops->list);
508 }
509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
510 
511 /**
512  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
513  * @ops: struct rtnl_af_ops * to unregister
514  */
515 void rtnl_af_unregister(struct rtnl_af_ops *ops)
516 {
517 	rtnl_lock();
518 	__rtnl_af_unregister(ops);
519 	rtnl_unlock();
520 }
521 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
522 
523 static size_t rtnl_link_get_af_size(const struct net_device *dev,
524 				    u32 ext_filter_mask)
525 {
526 	struct rtnl_af_ops *af_ops;
527 	size_t size;
528 
529 	/* IFLA_AF_SPEC */
530 	size = nla_total_size(sizeof(struct nlattr));
531 
532 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
533 		if (af_ops->get_link_af_size) {
534 			/* AF_* + nested data */
535 			size += nla_total_size(sizeof(struct nlattr)) +
536 				af_ops->get_link_af_size(dev, ext_filter_mask);
537 		}
538 	}
539 
540 	return size;
541 }
542 
543 static bool rtnl_have_link_slave_info(const struct net_device *dev)
544 {
545 	struct net_device *master_dev;
546 
547 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
548 	if (master_dev && master_dev->rtnl_link_ops)
549 		return true;
550 	return false;
551 }
552 
553 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
554 				     const struct net_device *dev)
555 {
556 	struct net_device *master_dev;
557 	const struct rtnl_link_ops *ops;
558 	struct nlattr *slave_data;
559 	int err;
560 
561 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
562 	if (!master_dev)
563 		return 0;
564 	ops = master_dev->rtnl_link_ops;
565 	if (!ops)
566 		return 0;
567 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
568 		return -EMSGSIZE;
569 	if (ops->fill_slave_info) {
570 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
571 		if (!slave_data)
572 			return -EMSGSIZE;
573 		err = ops->fill_slave_info(skb, master_dev, dev);
574 		if (err < 0)
575 			goto err_cancel_slave_data;
576 		nla_nest_end(skb, slave_data);
577 	}
578 	return 0;
579 
580 err_cancel_slave_data:
581 	nla_nest_cancel(skb, slave_data);
582 	return err;
583 }
584 
585 static int rtnl_link_info_fill(struct sk_buff *skb,
586 			       const struct net_device *dev)
587 {
588 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
589 	struct nlattr *data;
590 	int err;
591 
592 	if (!ops)
593 		return 0;
594 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
595 		return -EMSGSIZE;
596 	if (ops->fill_xstats) {
597 		err = ops->fill_xstats(skb, dev);
598 		if (err < 0)
599 			return err;
600 	}
601 	if (ops->fill_info) {
602 		data = nla_nest_start(skb, IFLA_INFO_DATA);
603 		if (data == NULL)
604 			return -EMSGSIZE;
605 		err = ops->fill_info(skb, dev);
606 		if (err < 0)
607 			goto err_cancel_data;
608 		nla_nest_end(skb, data);
609 	}
610 	return 0;
611 
612 err_cancel_data:
613 	nla_nest_cancel(skb, data);
614 	return err;
615 }
616 
617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
618 {
619 	struct nlattr *linkinfo;
620 	int err = -EMSGSIZE;
621 
622 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
623 	if (linkinfo == NULL)
624 		goto out;
625 
626 	err = rtnl_link_info_fill(skb, dev);
627 	if (err < 0)
628 		goto err_cancel_link;
629 
630 	err = rtnl_link_slave_info_fill(skb, dev);
631 	if (err < 0)
632 		goto err_cancel_link;
633 
634 	nla_nest_end(skb, linkinfo);
635 	return 0;
636 
637 err_cancel_link:
638 	nla_nest_cancel(skb, linkinfo);
639 out:
640 	return err;
641 }
642 
643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
644 {
645 	struct sock *rtnl = net->rtnl;
646 	int err = 0;
647 
648 	NETLINK_CB(skb).dst_group = group;
649 	if (echo)
650 		atomic_inc(&skb->users);
651 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
652 	if (echo)
653 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
654 	return err;
655 }
656 
657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
658 {
659 	struct sock *rtnl = net->rtnl;
660 
661 	return nlmsg_unicast(rtnl, skb, pid);
662 }
663 EXPORT_SYMBOL(rtnl_unicast);
664 
665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
666 		 struct nlmsghdr *nlh, gfp_t flags)
667 {
668 	struct sock *rtnl = net->rtnl;
669 	int report = 0;
670 
671 	if (nlh)
672 		report = nlmsg_report(nlh);
673 
674 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
675 }
676 EXPORT_SYMBOL(rtnl_notify);
677 
678 void rtnl_set_sk_err(struct net *net, u32 group, int error)
679 {
680 	struct sock *rtnl = net->rtnl;
681 
682 	netlink_set_err(rtnl, 0, group, error);
683 }
684 EXPORT_SYMBOL(rtnl_set_sk_err);
685 
686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
687 {
688 	struct nlattr *mx;
689 	int i, valid = 0;
690 
691 	mx = nla_nest_start(skb, RTA_METRICS);
692 	if (mx == NULL)
693 		return -ENOBUFS;
694 
695 	for (i = 0; i < RTAX_MAX; i++) {
696 		if (metrics[i]) {
697 			if (i == RTAX_CC_ALGO - 1) {
698 				char tmp[TCP_CA_NAME_MAX], *name;
699 
700 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
701 				if (!name)
702 					continue;
703 				if (nla_put_string(skb, i + 1, name))
704 					goto nla_put_failure;
705 			} else if (i == RTAX_FEATURES - 1) {
706 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
707 
708 				if (!user_features)
709 					continue;
710 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
711 				if (nla_put_u32(skb, i + 1, user_features))
712 					goto nla_put_failure;
713 			} else {
714 				if (nla_put_u32(skb, i + 1, metrics[i]))
715 					goto nla_put_failure;
716 			}
717 			valid++;
718 		}
719 	}
720 
721 	if (!valid) {
722 		nla_nest_cancel(skb, mx);
723 		return 0;
724 	}
725 
726 	return nla_nest_end(skb, mx);
727 
728 nla_put_failure:
729 	nla_nest_cancel(skb, mx);
730 	return -EMSGSIZE;
731 }
732 EXPORT_SYMBOL(rtnetlink_put_metrics);
733 
734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
735 		       long expires, u32 error)
736 {
737 	struct rta_cacheinfo ci = {
738 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
739 		.rta_used = dst->__use,
740 		.rta_clntref = atomic_read(&(dst->__refcnt)),
741 		.rta_error = error,
742 		.rta_id =  id,
743 	};
744 
745 	if (expires) {
746 		unsigned long clock;
747 
748 		clock = jiffies_to_clock_t(abs(expires));
749 		clock = min_t(unsigned long, clock, INT_MAX);
750 		ci.rta_expires = (expires > 0) ? clock : -clock;
751 	}
752 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
753 }
754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
755 
756 static void set_operstate(struct net_device *dev, unsigned char transition)
757 {
758 	unsigned char operstate = dev->operstate;
759 
760 	switch (transition) {
761 	case IF_OPER_UP:
762 		if ((operstate == IF_OPER_DORMANT ||
763 		     operstate == IF_OPER_UNKNOWN) &&
764 		    !netif_dormant(dev))
765 			operstate = IF_OPER_UP;
766 		break;
767 
768 	case IF_OPER_DORMANT:
769 		if (operstate == IF_OPER_UP ||
770 		    operstate == IF_OPER_UNKNOWN)
771 			operstate = IF_OPER_DORMANT;
772 		break;
773 	}
774 
775 	if (dev->operstate != operstate) {
776 		write_lock_bh(&dev_base_lock);
777 		dev->operstate = operstate;
778 		write_unlock_bh(&dev_base_lock);
779 		netdev_state_change(dev);
780 	}
781 }
782 
783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
784 {
785 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
786 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
787 }
788 
789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
790 					   const struct ifinfomsg *ifm)
791 {
792 	unsigned int flags = ifm->ifi_flags;
793 
794 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
795 	if (ifm->ifi_change)
796 		flags = (flags & ifm->ifi_change) |
797 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
798 
799 	return flags;
800 }
801 
802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
803 				 const struct rtnl_link_stats64 *b)
804 {
805 	a->rx_packets = b->rx_packets;
806 	a->tx_packets = b->tx_packets;
807 	a->rx_bytes = b->rx_bytes;
808 	a->tx_bytes = b->tx_bytes;
809 	a->rx_errors = b->rx_errors;
810 	a->tx_errors = b->tx_errors;
811 	a->rx_dropped = b->rx_dropped;
812 	a->tx_dropped = b->tx_dropped;
813 
814 	a->multicast = b->multicast;
815 	a->collisions = b->collisions;
816 
817 	a->rx_length_errors = b->rx_length_errors;
818 	a->rx_over_errors = b->rx_over_errors;
819 	a->rx_crc_errors = b->rx_crc_errors;
820 	a->rx_frame_errors = b->rx_frame_errors;
821 	a->rx_fifo_errors = b->rx_fifo_errors;
822 	a->rx_missed_errors = b->rx_missed_errors;
823 
824 	a->tx_aborted_errors = b->tx_aborted_errors;
825 	a->tx_carrier_errors = b->tx_carrier_errors;
826 	a->tx_fifo_errors = b->tx_fifo_errors;
827 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
828 	a->tx_window_errors = b->tx_window_errors;
829 
830 	a->rx_compressed = b->rx_compressed;
831 	a->tx_compressed = b->tx_compressed;
832 
833 	a->rx_nohandler = b->rx_nohandler;
834 }
835 
836 /* All VF info */
837 static inline int rtnl_vfinfo_size(const struct net_device *dev,
838 				   u32 ext_filter_mask)
839 {
840 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
841 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
842 		int num_vfs = dev_num_vf(dev->dev.parent);
843 		size_t size = nla_total_size(0);
844 		size += num_vfs *
845 			(nla_total_size(0) +
846 			 nla_total_size(sizeof(struct ifla_vf_mac)) +
847 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
848 			 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
849 			 nla_total_size(MAX_VLAN_LIST_LEN *
850 					sizeof(struct ifla_vf_vlan_info)) +
851 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
852 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
853 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
854 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
855 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
856 			 nla_total_size(0) + /* nest IFLA_VF_STATS */
857 			 /* IFLA_VF_STATS_RX_PACKETS */
858 			 nla_total_size_64bit(sizeof(__u64)) +
859 			 /* IFLA_VF_STATS_TX_PACKETS */
860 			 nla_total_size_64bit(sizeof(__u64)) +
861 			 /* IFLA_VF_STATS_RX_BYTES */
862 			 nla_total_size_64bit(sizeof(__u64)) +
863 			 /* IFLA_VF_STATS_TX_BYTES */
864 			 nla_total_size_64bit(sizeof(__u64)) +
865 			 /* IFLA_VF_STATS_BROADCAST */
866 			 nla_total_size_64bit(sizeof(__u64)) +
867 			 /* IFLA_VF_STATS_MULTICAST */
868 			 nla_total_size_64bit(sizeof(__u64)) +
869 			 nla_total_size(sizeof(struct ifla_vf_trust)));
870 		return size;
871 	} else
872 		return 0;
873 }
874 
875 static size_t rtnl_port_size(const struct net_device *dev,
876 			     u32 ext_filter_mask)
877 {
878 	size_t port_size = nla_total_size(4)		/* PORT_VF */
879 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
880 		+ nla_total_size(sizeof(struct ifla_port_vsi))
881 							/* PORT_VSI_TYPE */
882 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
883 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
884 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
885 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
886 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
887 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
888 		+ port_size;
889 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
890 		+ port_size;
891 
892 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
893 	    !(ext_filter_mask & RTEXT_FILTER_VF))
894 		return 0;
895 	if (dev_num_vf(dev->dev.parent))
896 		return port_self_size + vf_ports_size +
897 			vf_port_size * dev_num_vf(dev->dev.parent);
898 	else
899 		return port_self_size;
900 }
901 
902 static size_t rtnl_xdp_size(const struct net_device *dev)
903 {
904 	size_t xdp_size = nla_total_size(1);	/* XDP_ATTACHED */
905 
906 	if (!dev->netdev_ops->ndo_xdp)
907 		return 0;
908 	else
909 		return xdp_size;
910 }
911 
912 static noinline size_t if_nlmsg_size(const struct net_device *dev,
913 				     u32 ext_filter_mask)
914 {
915 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
916 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
917 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
918 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
919 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
920 	       + nla_total_size(sizeof(struct rtnl_link_stats))
921 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
922 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
923 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
924 	       + nla_total_size(4) /* IFLA_TXQLEN */
925 	       + nla_total_size(4) /* IFLA_WEIGHT */
926 	       + nla_total_size(4) /* IFLA_MTU */
927 	       + nla_total_size(4) /* IFLA_LINK */
928 	       + nla_total_size(4) /* IFLA_MASTER */
929 	       + nla_total_size(1) /* IFLA_CARRIER */
930 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
931 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
932 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
933 	       + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
934 	       + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
935 	       + nla_total_size(1) /* IFLA_OPERSTATE */
936 	       + nla_total_size(1) /* IFLA_LINKMODE */
937 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
938 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
939 	       + nla_total_size(ext_filter_mask
940 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
941 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
942 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
943 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
944 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
945 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
946 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
947 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
948 	       + rtnl_xdp_size(dev) /* IFLA_XDP */
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(skb, IFLA_PHYS_PORT_NAME, strlen(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 	/* Not all SR-IOV capable drivers support the
1133 	 * spoofcheck and "RSS query enable" query.  Preset to
1134 	 * -1 so the user space tool can detect that the driver
1135 	 * didn't report anything.
1136 	 */
1137 	ivi.spoofchk = -1;
1138 	ivi.rss_query_en = -1;
1139 	ivi.trusted = -1;
1140 	memset(ivi.mac, 0, sizeof(ivi.mac));
1141 	/* The default value for VF link state is "auto"
1142 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1143 	 */
1144 	ivi.linkstate = 0;
1145 	/* VLAN Protocol by default is 802.1Q */
1146 	ivi.vlan_proto = htons(ETH_P_8021Q);
1147 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1148 		return 0;
1149 
1150 	memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1151 
1152 	vf_mac.vf =
1153 		vf_vlan.vf =
1154 		vf_vlan_info.vf =
1155 		vf_rate.vf =
1156 		vf_tx_rate.vf =
1157 		vf_spoofchk.vf =
1158 		vf_linkstate.vf =
1159 		vf_rss_query_en.vf =
1160 		vf_trust.vf = ivi.vf;
1161 
1162 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1163 	vf_vlan.vlan = ivi.vlan;
1164 	vf_vlan.qos = ivi.qos;
1165 	vf_vlan_info.vlan = ivi.vlan;
1166 	vf_vlan_info.qos = ivi.qos;
1167 	vf_vlan_info.vlan_proto = ivi.vlan_proto;
1168 	vf_tx_rate.rate = ivi.max_tx_rate;
1169 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1170 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1171 	vf_spoofchk.setting = ivi.spoofchk;
1172 	vf_linkstate.link_state = ivi.linkstate;
1173 	vf_rss_query_en.setting = ivi.rss_query_en;
1174 	vf_trust.setting = ivi.trusted;
1175 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1176 	if (!vf)
1177 		goto nla_put_vfinfo_failure;
1178 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1179 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1180 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1181 		    &vf_rate) ||
1182 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1183 		    &vf_tx_rate) ||
1184 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1185 		    &vf_spoofchk) ||
1186 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1187 		    &vf_linkstate) ||
1188 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1189 		    sizeof(vf_rss_query_en),
1190 		    &vf_rss_query_en) ||
1191 	    nla_put(skb, IFLA_VF_TRUST,
1192 		    sizeof(vf_trust), &vf_trust))
1193 		goto nla_put_vf_failure;
1194 	vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1195 	if (!vfvlanlist)
1196 		goto nla_put_vf_failure;
1197 	if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1198 		    &vf_vlan_info)) {
1199 		nla_nest_cancel(skb, vfvlanlist);
1200 		goto nla_put_vf_failure;
1201 	}
1202 	nla_nest_end(skb, vfvlanlist);
1203 	memset(&vf_stats, 0, sizeof(vf_stats));
1204 	if (dev->netdev_ops->ndo_get_vf_stats)
1205 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1206 						&vf_stats);
1207 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1208 	if (!vfstats)
1209 		goto nla_put_vf_failure;
1210 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1211 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1212 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1213 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1214 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1215 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1216 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1217 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1218 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1219 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1220 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1221 			      vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1222 		nla_nest_cancel(skb, vfstats);
1223 		goto nla_put_vf_failure;
1224 	}
1225 	nla_nest_end(skb, vfstats);
1226 	nla_nest_end(skb, vf);
1227 	return 0;
1228 
1229 nla_put_vf_failure:
1230 	nla_nest_cancel(skb, vf);
1231 nla_put_vfinfo_failure:
1232 	nla_nest_cancel(skb, vfinfo);
1233 	return -EMSGSIZE;
1234 }
1235 
1236 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1237 {
1238 	struct rtnl_link_ifmap map;
1239 
1240 	memset(&map, 0, sizeof(map));
1241 	map.mem_start   = dev->mem_start;
1242 	map.mem_end     = dev->mem_end;
1243 	map.base_addr   = dev->base_addr;
1244 	map.irq         = dev->irq;
1245 	map.dma         = dev->dma;
1246 	map.port        = dev->if_port;
1247 
1248 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1249 		return -EMSGSIZE;
1250 
1251 	return 0;
1252 }
1253 
1254 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1255 {
1256 	struct netdev_xdp xdp_op = {};
1257 	struct nlattr *xdp;
1258 	int err;
1259 
1260 	if (!dev->netdev_ops->ndo_xdp)
1261 		return 0;
1262 	xdp = nla_nest_start(skb, IFLA_XDP);
1263 	if (!xdp)
1264 		return -EMSGSIZE;
1265 	xdp_op.command = XDP_QUERY_PROG;
1266 	err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1267 	if (err)
1268 		goto err_cancel;
1269 	err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1270 	if (err)
1271 		goto err_cancel;
1272 
1273 	nla_nest_end(skb, xdp);
1274 	return 0;
1275 
1276 err_cancel:
1277 	nla_nest_cancel(skb, xdp);
1278 	return err;
1279 }
1280 
1281 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1282 			    int type, u32 pid, u32 seq, u32 change,
1283 			    unsigned int flags, u32 ext_filter_mask)
1284 {
1285 	struct ifinfomsg *ifm;
1286 	struct nlmsghdr *nlh;
1287 	struct nlattr *af_spec;
1288 	struct rtnl_af_ops *af_ops;
1289 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1290 
1291 	ASSERT_RTNL();
1292 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1293 	if (nlh == NULL)
1294 		return -EMSGSIZE;
1295 
1296 	ifm = nlmsg_data(nlh);
1297 	ifm->ifi_family = AF_UNSPEC;
1298 	ifm->__ifi_pad = 0;
1299 	ifm->ifi_type = dev->type;
1300 	ifm->ifi_index = dev->ifindex;
1301 	ifm->ifi_flags = dev_get_flags(dev);
1302 	ifm->ifi_change = change;
1303 
1304 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1305 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1306 	    nla_put_u8(skb, IFLA_OPERSTATE,
1307 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1308 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1309 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1310 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1311 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1312 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1313 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1314 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1315 #ifdef CONFIG_RPS
1316 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1317 #endif
1318 	    (dev->ifindex != dev_get_iflink(dev) &&
1319 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1320 	    (upper_dev &&
1321 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1322 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1323 	    (dev->qdisc &&
1324 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1325 	    (dev->ifalias &&
1326 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1327 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1328 			atomic_read(&dev->carrier_changes)) ||
1329 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1330 		goto nla_put_failure;
1331 
1332 	if (rtnl_fill_link_ifmap(skb, dev))
1333 		goto nla_put_failure;
1334 
1335 	if (dev->addr_len) {
1336 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1337 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1338 			goto nla_put_failure;
1339 	}
1340 
1341 	if (rtnl_phys_port_id_fill(skb, dev))
1342 		goto nla_put_failure;
1343 
1344 	if (rtnl_phys_port_name_fill(skb, dev))
1345 		goto nla_put_failure;
1346 
1347 	if (rtnl_phys_switch_id_fill(skb, dev))
1348 		goto nla_put_failure;
1349 
1350 	if (rtnl_fill_stats(skb, dev))
1351 		goto nla_put_failure;
1352 
1353 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1354 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1355 		goto nla_put_failure;
1356 
1357 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1358 	    ext_filter_mask & RTEXT_FILTER_VF) {
1359 		int i;
1360 		struct nlattr *vfinfo;
1361 		int num_vfs = dev_num_vf(dev->dev.parent);
1362 
1363 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1364 		if (!vfinfo)
1365 			goto nla_put_failure;
1366 		for (i = 0; i < num_vfs; i++) {
1367 			if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1368 				goto nla_put_failure;
1369 		}
1370 
1371 		nla_nest_end(skb, vfinfo);
1372 	}
1373 
1374 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1375 		goto nla_put_failure;
1376 
1377 	if (rtnl_xdp_fill(skb, dev))
1378 		goto nla_put_failure;
1379 
1380 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1381 		if (rtnl_link_fill(skb, dev) < 0)
1382 			goto nla_put_failure;
1383 	}
1384 
1385 	if (dev->rtnl_link_ops &&
1386 	    dev->rtnl_link_ops->get_link_net) {
1387 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1388 
1389 		if (!net_eq(dev_net(dev), link_net)) {
1390 			int id = peernet2id_alloc(dev_net(dev), link_net);
1391 
1392 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1393 				goto nla_put_failure;
1394 		}
1395 	}
1396 
1397 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1398 		goto nla_put_failure;
1399 
1400 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1401 		if (af_ops->fill_link_af) {
1402 			struct nlattr *af;
1403 			int err;
1404 
1405 			if (!(af = nla_nest_start(skb, af_ops->family)))
1406 				goto nla_put_failure;
1407 
1408 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1409 
1410 			/*
1411 			 * Caller may return ENODATA to indicate that there
1412 			 * was no data to be dumped. This is not an error, it
1413 			 * means we should trim the attribute header and
1414 			 * continue.
1415 			 */
1416 			if (err == -ENODATA)
1417 				nla_nest_cancel(skb, af);
1418 			else if (err < 0)
1419 				goto nla_put_failure;
1420 
1421 			nla_nest_end(skb, af);
1422 		}
1423 	}
1424 
1425 	nla_nest_end(skb, af_spec);
1426 
1427 	nlmsg_end(skb, nlh);
1428 	return 0;
1429 
1430 nla_put_failure:
1431 	nlmsg_cancel(skb, nlh);
1432 	return -EMSGSIZE;
1433 }
1434 
1435 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1436 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1437 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1438 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1439 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1440 	[IFLA_MTU]		= { .type = NLA_U32 },
1441 	[IFLA_LINK]		= { .type = NLA_U32 },
1442 	[IFLA_MASTER]		= { .type = NLA_U32 },
1443 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1444 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1445 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1446 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1447 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1448 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1449 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1450 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1451 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1452 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1453 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1454 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1455 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1456 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1457 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1458 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1459 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1460 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1461 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1462 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1463 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1464 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1465 	[IFLA_XDP]		= { .type = NLA_NESTED },
1466 };
1467 
1468 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1469 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1470 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1471 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1472 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1473 };
1474 
1475 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1476 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1477 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1478 	[IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1479 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1480 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1481 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1482 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1483 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1484 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1485 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1486 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1487 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1488 };
1489 
1490 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1491 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1492 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1493 				    .len = PORT_PROFILE_MAX },
1494 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1495 				    .len = sizeof(struct ifla_port_vsi)},
1496 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1497 				      .len = PORT_UUID_MAX },
1498 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1499 				    .len = PORT_UUID_MAX },
1500 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1501 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1502 };
1503 
1504 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1505 	[IFLA_XDP_FD]		= { .type = NLA_S32 },
1506 	[IFLA_XDP_ATTACHED]	= { .type = NLA_U8 },
1507 };
1508 
1509 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1510 {
1511 	const struct rtnl_link_ops *ops = NULL;
1512 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1513 
1514 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1515 		return NULL;
1516 
1517 	if (linfo[IFLA_INFO_KIND]) {
1518 		char kind[MODULE_NAME_LEN];
1519 
1520 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1521 		ops = rtnl_link_ops_get(kind);
1522 	}
1523 
1524 	return ops;
1525 }
1526 
1527 static bool link_master_filtered(struct net_device *dev, int master_idx)
1528 {
1529 	struct net_device *master;
1530 
1531 	if (!master_idx)
1532 		return false;
1533 
1534 	master = netdev_master_upper_dev_get(dev);
1535 	if (!master || master->ifindex != master_idx)
1536 		return true;
1537 
1538 	return false;
1539 }
1540 
1541 static bool link_kind_filtered(const struct net_device *dev,
1542 			       const struct rtnl_link_ops *kind_ops)
1543 {
1544 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1545 		return true;
1546 
1547 	return false;
1548 }
1549 
1550 static bool link_dump_filtered(struct net_device *dev,
1551 			       int master_idx,
1552 			       const struct rtnl_link_ops *kind_ops)
1553 {
1554 	if (link_master_filtered(dev, master_idx) ||
1555 	    link_kind_filtered(dev, kind_ops))
1556 		return true;
1557 
1558 	return false;
1559 }
1560 
1561 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1562 {
1563 	struct net *net = sock_net(skb->sk);
1564 	int h, s_h;
1565 	int idx = 0, s_idx;
1566 	struct net_device *dev;
1567 	struct hlist_head *head;
1568 	struct nlattr *tb[IFLA_MAX+1];
1569 	u32 ext_filter_mask = 0;
1570 	const struct rtnl_link_ops *kind_ops = NULL;
1571 	unsigned int flags = NLM_F_MULTI;
1572 	int master_idx = 0;
1573 	int err;
1574 	int hdrlen;
1575 
1576 	s_h = cb->args[0];
1577 	s_idx = cb->args[1];
1578 
1579 	cb->seq = net->dev_base_seq;
1580 
1581 	/* A hack to preserve kernel<->userspace interface.
1582 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1583 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1584 	 * what iproute2 < v3.9.0 used.
1585 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1586 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1587 	 */
1588 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1589 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1590 
1591 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1592 
1593 		if (tb[IFLA_EXT_MASK])
1594 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1595 
1596 		if (tb[IFLA_MASTER])
1597 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1598 
1599 		if (tb[IFLA_LINKINFO])
1600 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1601 
1602 		if (master_idx || kind_ops)
1603 			flags |= NLM_F_DUMP_FILTERED;
1604 	}
1605 
1606 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1607 		idx = 0;
1608 		head = &net->dev_index_head[h];
1609 		hlist_for_each_entry(dev, head, index_hlist) {
1610 			if (link_dump_filtered(dev, master_idx, kind_ops))
1611 				continue;
1612 			if (idx < s_idx)
1613 				goto cont;
1614 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1615 					       NETLINK_CB(cb->skb).portid,
1616 					       cb->nlh->nlmsg_seq, 0,
1617 					       flags,
1618 					       ext_filter_mask);
1619 			/* If we ran out of room on the first message,
1620 			 * we're in trouble
1621 			 */
1622 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1623 
1624 			if (err < 0)
1625 				goto out;
1626 
1627 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1628 cont:
1629 			idx++;
1630 		}
1631 	}
1632 out:
1633 	cb->args[1] = idx;
1634 	cb->args[0] = h;
1635 
1636 	return skb->len;
1637 }
1638 
1639 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1640 {
1641 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1642 }
1643 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1644 
1645 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1646 {
1647 	struct net *net;
1648 	/* Examine the link attributes and figure out which
1649 	 * network namespace we are talking about.
1650 	 */
1651 	if (tb[IFLA_NET_NS_PID])
1652 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1653 	else if (tb[IFLA_NET_NS_FD])
1654 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1655 	else
1656 		net = get_net(src_net);
1657 	return net;
1658 }
1659 EXPORT_SYMBOL(rtnl_link_get_net);
1660 
1661 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1662 {
1663 	if (dev) {
1664 		if (tb[IFLA_ADDRESS] &&
1665 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1666 			return -EINVAL;
1667 
1668 		if (tb[IFLA_BROADCAST] &&
1669 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1670 			return -EINVAL;
1671 	}
1672 
1673 	if (tb[IFLA_AF_SPEC]) {
1674 		struct nlattr *af;
1675 		int rem, err;
1676 
1677 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1678 			const struct rtnl_af_ops *af_ops;
1679 
1680 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1681 				return -EAFNOSUPPORT;
1682 
1683 			if (!af_ops->set_link_af)
1684 				return -EOPNOTSUPP;
1685 
1686 			if (af_ops->validate_link_af) {
1687 				err = af_ops->validate_link_af(dev, af);
1688 				if (err < 0)
1689 					return err;
1690 			}
1691 		}
1692 	}
1693 
1694 	return 0;
1695 }
1696 
1697 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1698 				  int guid_type)
1699 {
1700 	const struct net_device_ops *ops = dev->netdev_ops;
1701 
1702 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1703 }
1704 
1705 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1706 {
1707 	if (dev->type != ARPHRD_INFINIBAND)
1708 		return -EOPNOTSUPP;
1709 
1710 	return handle_infiniband_guid(dev, ivt, guid_type);
1711 }
1712 
1713 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1714 {
1715 	const struct net_device_ops *ops = dev->netdev_ops;
1716 	int err = -EINVAL;
1717 
1718 	if (tb[IFLA_VF_MAC]) {
1719 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1720 
1721 		err = -EOPNOTSUPP;
1722 		if (ops->ndo_set_vf_mac)
1723 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1724 						  ivm->mac);
1725 		if (err < 0)
1726 			return err;
1727 	}
1728 
1729 	if (tb[IFLA_VF_VLAN]) {
1730 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1731 
1732 		err = -EOPNOTSUPP;
1733 		if (ops->ndo_set_vf_vlan)
1734 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1735 						   ivv->qos,
1736 						   htons(ETH_P_8021Q));
1737 		if (err < 0)
1738 			return err;
1739 	}
1740 
1741 	if (tb[IFLA_VF_VLAN_LIST]) {
1742 		struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1743 		struct nlattr *attr;
1744 		int rem, len = 0;
1745 
1746 		err = -EOPNOTSUPP;
1747 		if (!ops->ndo_set_vf_vlan)
1748 			return err;
1749 
1750 		nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1751 			if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1752 			    nla_len(attr) < NLA_HDRLEN) {
1753 				return -EINVAL;
1754 			}
1755 			if (len >= MAX_VLAN_LIST_LEN)
1756 				return -EOPNOTSUPP;
1757 			ivvl[len] = nla_data(attr);
1758 
1759 			len++;
1760 		}
1761 		if (len == 0)
1762 			return -EINVAL;
1763 
1764 		err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1765 					   ivvl[0]->qos, ivvl[0]->vlan_proto);
1766 		if (err < 0)
1767 			return err;
1768 	}
1769 
1770 	if (tb[IFLA_VF_TX_RATE]) {
1771 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1772 		struct ifla_vf_info ivf;
1773 
1774 		err = -EOPNOTSUPP;
1775 		if (ops->ndo_get_vf_config)
1776 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1777 		if (err < 0)
1778 			return err;
1779 
1780 		err = -EOPNOTSUPP;
1781 		if (ops->ndo_set_vf_rate)
1782 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1783 						   ivf.min_tx_rate,
1784 						   ivt->rate);
1785 		if (err < 0)
1786 			return err;
1787 	}
1788 
1789 	if (tb[IFLA_VF_RATE]) {
1790 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1791 
1792 		err = -EOPNOTSUPP;
1793 		if (ops->ndo_set_vf_rate)
1794 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1795 						   ivt->min_tx_rate,
1796 						   ivt->max_tx_rate);
1797 		if (err < 0)
1798 			return err;
1799 	}
1800 
1801 	if (tb[IFLA_VF_SPOOFCHK]) {
1802 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1803 
1804 		err = -EOPNOTSUPP;
1805 		if (ops->ndo_set_vf_spoofchk)
1806 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1807 						       ivs->setting);
1808 		if (err < 0)
1809 			return err;
1810 	}
1811 
1812 	if (tb[IFLA_VF_LINK_STATE]) {
1813 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1814 
1815 		err = -EOPNOTSUPP;
1816 		if (ops->ndo_set_vf_link_state)
1817 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1818 							 ivl->link_state);
1819 		if (err < 0)
1820 			return err;
1821 	}
1822 
1823 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1824 		struct ifla_vf_rss_query_en *ivrssq_en;
1825 
1826 		err = -EOPNOTSUPP;
1827 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1828 		if (ops->ndo_set_vf_rss_query_en)
1829 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1830 							   ivrssq_en->setting);
1831 		if (err < 0)
1832 			return err;
1833 	}
1834 
1835 	if (tb[IFLA_VF_TRUST]) {
1836 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1837 
1838 		err = -EOPNOTSUPP;
1839 		if (ops->ndo_set_vf_trust)
1840 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1841 		if (err < 0)
1842 			return err;
1843 	}
1844 
1845 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1846 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1847 
1848 		if (!ops->ndo_set_vf_guid)
1849 			return -EOPNOTSUPP;
1850 
1851 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1852 	}
1853 
1854 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1855 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1856 
1857 		if (!ops->ndo_set_vf_guid)
1858 			return -EOPNOTSUPP;
1859 
1860 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1861 	}
1862 
1863 	return err;
1864 }
1865 
1866 static int do_set_master(struct net_device *dev, int ifindex)
1867 {
1868 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1869 	const struct net_device_ops *ops;
1870 	int err;
1871 
1872 	if (upper_dev) {
1873 		if (upper_dev->ifindex == ifindex)
1874 			return 0;
1875 		ops = upper_dev->netdev_ops;
1876 		if (ops->ndo_del_slave) {
1877 			err = ops->ndo_del_slave(upper_dev, dev);
1878 			if (err)
1879 				return err;
1880 		} else {
1881 			return -EOPNOTSUPP;
1882 		}
1883 	}
1884 
1885 	if (ifindex) {
1886 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1887 		if (!upper_dev)
1888 			return -EINVAL;
1889 		ops = upper_dev->netdev_ops;
1890 		if (ops->ndo_add_slave) {
1891 			err = ops->ndo_add_slave(upper_dev, dev);
1892 			if (err)
1893 				return err;
1894 		} else {
1895 			return -EOPNOTSUPP;
1896 		}
1897 	}
1898 	return 0;
1899 }
1900 
1901 #define DO_SETLINK_MODIFIED	0x01
1902 /* notify flag means notify + modified. */
1903 #define DO_SETLINK_NOTIFY	0x03
1904 static int do_setlink(const struct sk_buff *skb,
1905 		      struct net_device *dev, struct ifinfomsg *ifm,
1906 		      struct nlattr **tb, char *ifname, int status)
1907 {
1908 	const struct net_device_ops *ops = dev->netdev_ops;
1909 	int err;
1910 
1911 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1912 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1913 		if (IS_ERR(net)) {
1914 			err = PTR_ERR(net);
1915 			goto errout;
1916 		}
1917 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1918 			put_net(net);
1919 			err = -EPERM;
1920 			goto errout;
1921 		}
1922 		err = dev_change_net_namespace(dev, net, ifname);
1923 		put_net(net);
1924 		if (err)
1925 			goto errout;
1926 		status |= DO_SETLINK_MODIFIED;
1927 	}
1928 
1929 	if (tb[IFLA_MAP]) {
1930 		struct rtnl_link_ifmap *u_map;
1931 		struct ifmap k_map;
1932 
1933 		if (!ops->ndo_set_config) {
1934 			err = -EOPNOTSUPP;
1935 			goto errout;
1936 		}
1937 
1938 		if (!netif_device_present(dev)) {
1939 			err = -ENODEV;
1940 			goto errout;
1941 		}
1942 
1943 		u_map = nla_data(tb[IFLA_MAP]);
1944 		k_map.mem_start = (unsigned long) u_map->mem_start;
1945 		k_map.mem_end = (unsigned long) u_map->mem_end;
1946 		k_map.base_addr = (unsigned short) u_map->base_addr;
1947 		k_map.irq = (unsigned char) u_map->irq;
1948 		k_map.dma = (unsigned char) u_map->dma;
1949 		k_map.port = (unsigned char) u_map->port;
1950 
1951 		err = ops->ndo_set_config(dev, &k_map);
1952 		if (err < 0)
1953 			goto errout;
1954 
1955 		status |= DO_SETLINK_NOTIFY;
1956 	}
1957 
1958 	if (tb[IFLA_ADDRESS]) {
1959 		struct sockaddr *sa;
1960 		int len;
1961 
1962 		len = sizeof(sa_family_t) + dev->addr_len;
1963 		sa = kmalloc(len, GFP_KERNEL);
1964 		if (!sa) {
1965 			err = -ENOMEM;
1966 			goto errout;
1967 		}
1968 		sa->sa_family = dev->type;
1969 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1970 		       dev->addr_len);
1971 		err = dev_set_mac_address(dev, sa);
1972 		kfree(sa);
1973 		if (err)
1974 			goto errout;
1975 		status |= DO_SETLINK_MODIFIED;
1976 	}
1977 
1978 	if (tb[IFLA_MTU]) {
1979 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1980 		if (err < 0)
1981 			goto errout;
1982 		status |= DO_SETLINK_MODIFIED;
1983 	}
1984 
1985 	if (tb[IFLA_GROUP]) {
1986 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1987 		status |= DO_SETLINK_NOTIFY;
1988 	}
1989 
1990 	/*
1991 	 * Interface selected by interface index but interface
1992 	 * name provided implies that a name change has been
1993 	 * requested.
1994 	 */
1995 	if (ifm->ifi_index > 0 && ifname[0]) {
1996 		err = dev_change_name(dev, ifname);
1997 		if (err < 0)
1998 			goto errout;
1999 		status |= DO_SETLINK_MODIFIED;
2000 	}
2001 
2002 	if (tb[IFLA_IFALIAS]) {
2003 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2004 				    nla_len(tb[IFLA_IFALIAS]));
2005 		if (err < 0)
2006 			goto errout;
2007 		status |= DO_SETLINK_NOTIFY;
2008 	}
2009 
2010 	if (tb[IFLA_BROADCAST]) {
2011 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2012 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2013 	}
2014 
2015 	if (ifm->ifi_flags || ifm->ifi_change) {
2016 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2017 		if (err < 0)
2018 			goto errout;
2019 	}
2020 
2021 	if (tb[IFLA_MASTER]) {
2022 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2023 		if (err)
2024 			goto errout;
2025 		status |= DO_SETLINK_MODIFIED;
2026 	}
2027 
2028 	if (tb[IFLA_CARRIER]) {
2029 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2030 		if (err)
2031 			goto errout;
2032 		status |= DO_SETLINK_MODIFIED;
2033 	}
2034 
2035 	if (tb[IFLA_TXQLEN]) {
2036 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2037 		unsigned long orig_len = dev->tx_queue_len;
2038 
2039 		if (dev->tx_queue_len ^ value) {
2040 			dev->tx_queue_len = value;
2041 			err = call_netdevice_notifiers(
2042 			      NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2043 			err = notifier_to_errno(err);
2044 			if (err) {
2045 				dev->tx_queue_len = orig_len;
2046 				goto errout;
2047 			}
2048 			status |= DO_SETLINK_NOTIFY;
2049 		}
2050 	}
2051 
2052 	if (tb[IFLA_OPERSTATE])
2053 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2054 
2055 	if (tb[IFLA_LINKMODE]) {
2056 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2057 
2058 		write_lock_bh(&dev_base_lock);
2059 		if (dev->link_mode ^ value)
2060 			status |= DO_SETLINK_NOTIFY;
2061 		dev->link_mode = value;
2062 		write_unlock_bh(&dev_base_lock);
2063 	}
2064 
2065 	if (tb[IFLA_VFINFO_LIST]) {
2066 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2067 		struct nlattr *attr;
2068 		int rem;
2069 
2070 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2071 			if (nla_type(attr) != IFLA_VF_INFO ||
2072 			    nla_len(attr) < NLA_HDRLEN) {
2073 				err = -EINVAL;
2074 				goto errout;
2075 			}
2076 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2077 					       ifla_vf_policy);
2078 			if (err < 0)
2079 				goto errout;
2080 			err = do_setvfinfo(dev, vfinfo);
2081 			if (err < 0)
2082 				goto errout;
2083 			status |= DO_SETLINK_NOTIFY;
2084 		}
2085 	}
2086 	err = 0;
2087 
2088 	if (tb[IFLA_VF_PORTS]) {
2089 		struct nlattr *port[IFLA_PORT_MAX+1];
2090 		struct nlattr *attr;
2091 		int vf;
2092 		int rem;
2093 
2094 		err = -EOPNOTSUPP;
2095 		if (!ops->ndo_set_vf_port)
2096 			goto errout;
2097 
2098 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2099 			if (nla_type(attr) != IFLA_VF_PORT ||
2100 			    nla_len(attr) < NLA_HDRLEN) {
2101 				err = -EINVAL;
2102 				goto errout;
2103 			}
2104 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2105 					       ifla_port_policy);
2106 			if (err < 0)
2107 				goto errout;
2108 			if (!port[IFLA_PORT_VF]) {
2109 				err = -EOPNOTSUPP;
2110 				goto errout;
2111 			}
2112 			vf = nla_get_u32(port[IFLA_PORT_VF]);
2113 			err = ops->ndo_set_vf_port(dev, vf, port);
2114 			if (err < 0)
2115 				goto errout;
2116 			status |= DO_SETLINK_NOTIFY;
2117 		}
2118 	}
2119 	err = 0;
2120 
2121 	if (tb[IFLA_PORT_SELF]) {
2122 		struct nlattr *port[IFLA_PORT_MAX+1];
2123 
2124 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2125 			tb[IFLA_PORT_SELF], ifla_port_policy);
2126 		if (err < 0)
2127 			goto errout;
2128 
2129 		err = -EOPNOTSUPP;
2130 		if (ops->ndo_set_vf_port)
2131 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2132 		if (err < 0)
2133 			goto errout;
2134 		status |= DO_SETLINK_NOTIFY;
2135 	}
2136 
2137 	if (tb[IFLA_AF_SPEC]) {
2138 		struct nlattr *af;
2139 		int rem;
2140 
2141 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2142 			const struct rtnl_af_ops *af_ops;
2143 
2144 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2145 				BUG();
2146 
2147 			err = af_ops->set_link_af(dev, af);
2148 			if (err < 0)
2149 				goto errout;
2150 
2151 			status |= DO_SETLINK_NOTIFY;
2152 		}
2153 	}
2154 	err = 0;
2155 
2156 	if (tb[IFLA_PROTO_DOWN]) {
2157 		err = dev_change_proto_down(dev,
2158 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2159 		if (err)
2160 			goto errout;
2161 		status |= DO_SETLINK_NOTIFY;
2162 	}
2163 
2164 	if (tb[IFLA_XDP]) {
2165 		struct nlattr *xdp[IFLA_XDP_MAX + 1];
2166 
2167 		err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2168 				       ifla_xdp_policy);
2169 		if (err < 0)
2170 			goto errout;
2171 
2172 		if (xdp[IFLA_XDP_ATTACHED]) {
2173 			err = -EINVAL;
2174 			goto errout;
2175 		}
2176 		if (xdp[IFLA_XDP_FD]) {
2177 			err = dev_change_xdp_fd(dev,
2178 						nla_get_s32(xdp[IFLA_XDP_FD]));
2179 			if (err)
2180 				goto errout;
2181 			status |= DO_SETLINK_NOTIFY;
2182 		}
2183 	}
2184 
2185 errout:
2186 	if (status & DO_SETLINK_MODIFIED) {
2187 		if (status & DO_SETLINK_NOTIFY)
2188 			netdev_state_change(dev);
2189 
2190 		if (err < 0)
2191 			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",
2192 					     dev->name);
2193 	}
2194 
2195 	return err;
2196 }
2197 
2198 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2199 {
2200 	struct net *net = sock_net(skb->sk);
2201 	struct ifinfomsg *ifm;
2202 	struct net_device *dev;
2203 	int err;
2204 	struct nlattr *tb[IFLA_MAX+1];
2205 	char ifname[IFNAMSIZ];
2206 
2207 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2208 	if (err < 0)
2209 		goto errout;
2210 
2211 	if (tb[IFLA_IFNAME])
2212 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2213 	else
2214 		ifname[0] = '\0';
2215 
2216 	err = -EINVAL;
2217 	ifm = nlmsg_data(nlh);
2218 	if (ifm->ifi_index > 0)
2219 		dev = __dev_get_by_index(net, ifm->ifi_index);
2220 	else if (tb[IFLA_IFNAME])
2221 		dev = __dev_get_by_name(net, ifname);
2222 	else
2223 		goto errout;
2224 
2225 	if (dev == NULL) {
2226 		err = -ENODEV;
2227 		goto errout;
2228 	}
2229 
2230 	err = validate_linkmsg(dev, tb);
2231 	if (err < 0)
2232 		goto errout;
2233 
2234 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2235 errout:
2236 	return err;
2237 }
2238 
2239 static int rtnl_group_dellink(const struct net *net, int group)
2240 {
2241 	struct net_device *dev, *aux;
2242 	LIST_HEAD(list_kill);
2243 	bool found = false;
2244 
2245 	if (!group)
2246 		return -EPERM;
2247 
2248 	for_each_netdev(net, dev) {
2249 		if (dev->group == group) {
2250 			const struct rtnl_link_ops *ops;
2251 
2252 			found = true;
2253 			ops = dev->rtnl_link_ops;
2254 			if (!ops || !ops->dellink)
2255 				return -EOPNOTSUPP;
2256 		}
2257 	}
2258 
2259 	if (!found)
2260 		return -ENODEV;
2261 
2262 	for_each_netdev_safe(net, dev, aux) {
2263 		if (dev->group == group) {
2264 			const struct rtnl_link_ops *ops;
2265 
2266 			ops = dev->rtnl_link_ops;
2267 			ops->dellink(dev, &list_kill);
2268 		}
2269 	}
2270 	unregister_netdevice_many(&list_kill);
2271 
2272 	return 0;
2273 }
2274 
2275 int rtnl_delete_link(struct net_device *dev)
2276 {
2277 	const struct rtnl_link_ops *ops;
2278 	LIST_HEAD(list_kill);
2279 
2280 	ops = dev->rtnl_link_ops;
2281 	if (!ops || !ops->dellink)
2282 		return -EOPNOTSUPP;
2283 
2284 	ops->dellink(dev, &list_kill);
2285 	unregister_netdevice_many(&list_kill);
2286 
2287 	return 0;
2288 }
2289 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2290 
2291 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2292 {
2293 	struct net *net = sock_net(skb->sk);
2294 	struct net_device *dev;
2295 	struct ifinfomsg *ifm;
2296 	char ifname[IFNAMSIZ];
2297 	struct nlattr *tb[IFLA_MAX+1];
2298 	int err;
2299 
2300 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2301 	if (err < 0)
2302 		return err;
2303 
2304 	if (tb[IFLA_IFNAME])
2305 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2306 
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 if (tb[IFLA_GROUP])
2313 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2314 	else
2315 		return -EINVAL;
2316 
2317 	if (!dev)
2318 		return -ENODEV;
2319 
2320 	return rtnl_delete_link(dev);
2321 }
2322 
2323 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2324 {
2325 	unsigned int old_flags;
2326 	int err;
2327 
2328 	old_flags = dev->flags;
2329 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2330 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2331 		if (err < 0)
2332 			return err;
2333 	}
2334 
2335 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2336 
2337 	__dev_notify_flags(dev, old_flags, ~0U);
2338 	return 0;
2339 }
2340 EXPORT_SYMBOL(rtnl_configure_link);
2341 
2342 struct net_device *rtnl_create_link(struct net *net,
2343 	const char *ifname, unsigned char name_assign_type,
2344 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2345 {
2346 	int err;
2347 	struct net_device *dev;
2348 	unsigned int num_tx_queues = 1;
2349 	unsigned int num_rx_queues = 1;
2350 
2351 	if (tb[IFLA_NUM_TX_QUEUES])
2352 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2353 	else if (ops->get_num_tx_queues)
2354 		num_tx_queues = ops->get_num_tx_queues();
2355 
2356 	if (tb[IFLA_NUM_RX_QUEUES])
2357 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2358 	else if (ops->get_num_rx_queues)
2359 		num_rx_queues = ops->get_num_rx_queues();
2360 
2361 	err = -ENOMEM;
2362 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2363 			       ops->setup, num_tx_queues, num_rx_queues);
2364 	if (!dev)
2365 		goto err;
2366 
2367 	dev_net_set(dev, net);
2368 	dev->rtnl_link_ops = ops;
2369 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2370 
2371 	if (tb[IFLA_MTU])
2372 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2373 	if (tb[IFLA_ADDRESS]) {
2374 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2375 				nla_len(tb[IFLA_ADDRESS]));
2376 		dev->addr_assign_type = NET_ADDR_SET;
2377 	}
2378 	if (tb[IFLA_BROADCAST])
2379 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2380 				nla_len(tb[IFLA_BROADCAST]));
2381 	if (tb[IFLA_TXQLEN])
2382 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2383 	if (tb[IFLA_OPERSTATE])
2384 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2385 	if (tb[IFLA_LINKMODE])
2386 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2387 	if (tb[IFLA_GROUP])
2388 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2389 
2390 	return dev;
2391 
2392 err:
2393 	return ERR_PTR(err);
2394 }
2395 EXPORT_SYMBOL(rtnl_create_link);
2396 
2397 static int rtnl_group_changelink(const struct sk_buff *skb,
2398 		struct net *net, int group,
2399 		struct ifinfomsg *ifm,
2400 		struct nlattr **tb)
2401 {
2402 	struct net_device *dev, *aux;
2403 	int err;
2404 
2405 	for_each_netdev_safe(net, dev, aux) {
2406 		if (dev->group == group) {
2407 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2408 			if (err < 0)
2409 				return err;
2410 		}
2411 	}
2412 
2413 	return 0;
2414 }
2415 
2416 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2417 {
2418 	struct net *net = sock_net(skb->sk);
2419 	const struct rtnl_link_ops *ops;
2420 	const struct rtnl_link_ops *m_ops = NULL;
2421 	struct net_device *dev;
2422 	struct net_device *master_dev = NULL;
2423 	struct ifinfomsg *ifm;
2424 	char kind[MODULE_NAME_LEN];
2425 	char ifname[IFNAMSIZ];
2426 	struct nlattr *tb[IFLA_MAX+1];
2427 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2428 	unsigned char name_assign_type = NET_NAME_USER;
2429 	int err;
2430 
2431 #ifdef CONFIG_MODULES
2432 replay:
2433 #endif
2434 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2435 	if (err < 0)
2436 		return err;
2437 
2438 	if (tb[IFLA_IFNAME])
2439 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2440 	else
2441 		ifname[0] = '\0';
2442 
2443 	ifm = nlmsg_data(nlh);
2444 	if (ifm->ifi_index > 0)
2445 		dev = __dev_get_by_index(net, ifm->ifi_index);
2446 	else {
2447 		if (ifname[0])
2448 			dev = __dev_get_by_name(net, ifname);
2449 		else
2450 			dev = NULL;
2451 	}
2452 
2453 	if (dev) {
2454 		master_dev = netdev_master_upper_dev_get(dev);
2455 		if (master_dev)
2456 			m_ops = master_dev->rtnl_link_ops;
2457 	}
2458 
2459 	err = validate_linkmsg(dev, tb);
2460 	if (err < 0)
2461 		return err;
2462 
2463 	if (tb[IFLA_LINKINFO]) {
2464 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2465 				       tb[IFLA_LINKINFO], ifla_info_policy);
2466 		if (err < 0)
2467 			return err;
2468 	} else
2469 		memset(linkinfo, 0, sizeof(linkinfo));
2470 
2471 	if (linkinfo[IFLA_INFO_KIND]) {
2472 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2473 		ops = rtnl_link_ops_get(kind);
2474 	} else {
2475 		kind[0] = '\0';
2476 		ops = NULL;
2477 	}
2478 
2479 	if (1) {
2480 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2481 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2482 		struct nlattr **data = NULL;
2483 		struct nlattr **slave_data = NULL;
2484 		struct net *dest_net, *link_net = NULL;
2485 
2486 		if (ops) {
2487 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2488 				err = nla_parse_nested(attr, ops->maxtype,
2489 						       linkinfo[IFLA_INFO_DATA],
2490 						       ops->policy);
2491 				if (err < 0)
2492 					return err;
2493 				data = attr;
2494 			}
2495 			if (ops->validate) {
2496 				err = ops->validate(tb, data);
2497 				if (err < 0)
2498 					return err;
2499 			}
2500 		}
2501 
2502 		if (m_ops) {
2503 			if (m_ops->slave_maxtype &&
2504 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2505 				err = nla_parse_nested(slave_attr,
2506 						       m_ops->slave_maxtype,
2507 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2508 						       m_ops->slave_policy);
2509 				if (err < 0)
2510 					return err;
2511 				slave_data = slave_attr;
2512 			}
2513 			if (m_ops->slave_validate) {
2514 				err = m_ops->slave_validate(tb, slave_data);
2515 				if (err < 0)
2516 					return err;
2517 			}
2518 		}
2519 
2520 		if (dev) {
2521 			int status = 0;
2522 
2523 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2524 				return -EEXIST;
2525 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2526 				return -EOPNOTSUPP;
2527 
2528 			if (linkinfo[IFLA_INFO_DATA]) {
2529 				if (!ops || ops != dev->rtnl_link_ops ||
2530 				    !ops->changelink)
2531 					return -EOPNOTSUPP;
2532 
2533 				err = ops->changelink(dev, tb, data);
2534 				if (err < 0)
2535 					return err;
2536 				status |= DO_SETLINK_NOTIFY;
2537 			}
2538 
2539 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2540 				if (!m_ops || !m_ops->slave_changelink)
2541 					return -EOPNOTSUPP;
2542 
2543 				err = m_ops->slave_changelink(master_dev, dev,
2544 							      tb, slave_data);
2545 				if (err < 0)
2546 					return err;
2547 				status |= DO_SETLINK_NOTIFY;
2548 			}
2549 
2550 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2551 		}
2552 
2553 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2554 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2555 				return rtnl_group_changelink(skb, net,
2556 						nla_get_u32(tb[IFLA_GROUP]),
2557 						ifm, tb);
2558 			return -ENODEV;
2559 		}
2560 
2561 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2562 			return -EOPNOTSUPP;
2563 
2564 		if (!ops) {
2565 #ifdef CONFIG_MODULES
2566 			if (kind[0]) {
2567 				__rtnl_unlock();
2568 				request_module("rtnl-link-%s", kind);
2569 				rtnl_lock();
2570 				ops = rtnl_link_ops_get(kind);
2571 				if (ops)
2572 					goto replay;
2573 			}
2574 #endif
2575 			return -EOPNOTSUPP;
2576 		}
2577 
2578 		if (!ops->setup)
2579 			return -EOPNOTSUPP;
2580 
2581 		if (!ifname[0]) {
2582 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2583 			name_assign_type = NET_NAME_ENUM;
2584 		}
2585 
2586 		dest_net = rtnl_link_get_net(net, tb);
2587 		if (IS_ERR(dest_net))
2588 			return PTR_ERR(dest_net);
2589 
2590 		err = -EPERM;
2591 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2592 			goto out;
2593 
2594 		if (tb[IFLA_LINK_NETNSID]) {
2595 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2596 
2597 			link_net = get_net_ns_by_id(dest_net, id);
2598 			if (!link_net) {
2599 				err =  -EINVAL;
2600 				goto out;
2601 			}
2602 			err = -EPERM;
2603 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2604 				goto out;
2605 		}
2606 
2607 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2608 				       name_assign_type, ops, tb);
2609 		if (IS_ERR(dev)) {
2610 			err = PTR_ERR(dev);
2611 			goto out;
2612 		}
2613 
2614 		dev->ifindex = ifm->ifi_index;
2615 
2616 		if (ops->newlink) {
2617 			err = ops->newlink(link_net ? : net, dev, tb, data);
2618 			/* Drivers should call free_netdev() in ->destructor
2619 			 * and unregister it on failure after registration
2620 			 * so that device could be finally freed in rtnl_unlock.
2621 			 */
2622 			if (err < 0) {
2623 				/* If device is not registered at all, free it now */
2624 				if (dev->reg_state == NETREG_UNINITIALIZED)
2625 					free_netdev(dev);
2626 				goto out;
2627 			}
2628 		} else {
2629 			err = register_netdevice(dev);
2630 			if (err < 0) {
2631 				free_netdev(dev);
2632 				goto out;
2633 			}
2634 		}
2635 		err = rtnl_configure_link(dev, ifm);
2636 		if (err < 0)
2637 			goto out_unregister;
2638 		if (link_net) {
2639 			err = dev_change_net_namespace(dev, dest_net, ifname);
2640 			if (err < 0)
2641 				goto out_unregister;
2642 		}
2643 out:
2644 		if (link_net)
2645 			put_net(link_net);
2646 		put_net(dest_net);
2647 		return err;
2648 out_unregister:
2649 		if (ops->newlink) {
2650 			LIST_HEAD(list_kill);
2651 
2652 			ops->dellink(dev, &list_kill);
2653 			unregister_netdevice_many(&list_kill);
2654 		} else {
2655 			unregister_netdevice(dev);
2656 		}
2657 		goto out;
2658 	}
2659 }
2660 
2661 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2662 {
2663 	struct net *net = sock_net(skb->sk);
2664 	struct ifinfomsg *ifm;
2665 	char ifname[IFNAMSIZ];
2666 	struct nlattr *tb[IFLA_MAX+1];
2667 	struct net_device *dev = NULL;
2668 	struct sk_buff *nskb;
2669 	int err;
2670 	u32 ext_filter_mask = 0;
2671 
2672 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2673 	if (err < 0)
2674 		return err;
2675 
2676 	if (tb[IFLA_IFNAME])
2677 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2678 
2679 	if (tb[IFLA_EXT_MASK])
2680 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2681 
2682 	ifm = nlmsg_data(nlh);
2683 	if (ifm->ifi_index > 0)
2684 		dev = __dev_get_by_index(net, ifm->ifi_index);
2685 	else if (tb[IFLA_IFNAME])
2686 		dev = __dev_get_by_name(net, ifname);
2687 	else
2688 		return -EINVAL;
2689 
2690 	if (dev == NULL)
2691 		return -ENODEV;
2692 
2693 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2694 	if (nskb == NULL)
2695 		return -ENOBUFS;
2696 
2697 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2698 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2699 	if (err < 0) {
2700 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2701 		WARN_ON(err == -EMSGSIZE);
2702 		kfree_skb(nskb);
2703 	} else
2704 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2705 
2706 	return err;
2707 }
2708 
2709 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2710 {
2711 	struct net *net = sock_net(skb->sk);
2712 	struct net_device *dev;
2713 	struct nlattr *tb[IFLA_MAX+1];
2714 	u32 ext_filter_mask = 0;
2715 	u16 min_ifinfo_dump_size = 0;
2716 	int hdrlen;
2717 
2718 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2719 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2720 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2721 
2722 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2723 		if (tb[IFLA_EXT_MASK])
2724 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2725 	}
2726 
2727 	if (!ext_filter_mask)
2728 		return NLMSG_GOODSIZE;
2729 	/*
2730 	 * traverse the list of net devices and compute the minimum
2731 	 * buffer size based upon the filter mask.
2732 	 */
2733 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2734 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2735 					     if_nlmsg_size(dev,
2736 						           ext_filter_mask));
2737 	}
2738 
2739 	return min_ifinfo_dump_size;
2740 }
2741 
2742 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2743 {
2744 	int idx;
2745 	int s_idx = cb->family;
2746 
2747 	if (s_idx == 0)
2748 		s_idx = 1;
2749 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2750 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2751 		if (idx < s_idx || idx == PF_PACKET)
2752 			continue;
2753 		if (rtnl_msg_handlers[idx] == NULL ||
2754 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2755 			continue;
2756 		if (idx > s_idx) {
2757 			memset(&cb->args[0], 0, sizeof(cb->args));
2758 			cb->prev_seq = 0;
2759 			cb->seq = 0;
2760 		}
2761 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2762 			break;
2763 	}
2764 	cb->family = idx;
2765 
2766 	return skb->len;
2767 }
2768 
2769 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2770 				       unsigned int change, gfp_t flags)
2771 {
2772 	struct net *net = dev_net(dev);
2773 	struct sk_buff *skb;
2774 	int err = -ENOBUFS;
2775 	size_t if_info_size;
2776 
2777 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2778 	if (skb == NULL)
2779 		goto errout;
2780 
2781 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2782 	if (err < 0) {
2783 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2784 		WARN_ON(err == -EMSGSIZE);
2785 		kfree_skb(skb);
2786 		goto errout;
2787 	}
2788 	return skb;
2789 errout:
2790 	if (err < 0)
2791 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2792 	return NULL;
2793 }
2794 
2795 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2796 {
2797 	struct net *net = dev_net(dev);
2798 
2799 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2800 }
2801 
2802 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2803 		  gfp_t flags)
2804 {
2805 	struct sk_buff *skb;
2806 
2807 	if (dev->reg_state != NETREG_REGISTERED)
2808 		return;
2809 
2810 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2811 	if (skb)
2812 		rtmsg_ifinfo_send(skb, dev, flags);
2813 }
2814 EXPORT_SYMBOL(rtmsg_ifinfo);
2815 
2816 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2817 				   struct net_device *dev,
2818 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2819 				   int type, unsigned int flags,
2820 				   int nlflags, u16 ndm_state)
2821 {
2822 	struct nlmsghdr *nlh;
2823 	struct ndmsg *ndm;
2824 
2825 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2826 	if (!nlh)
2827 		return -EMSGSIZE;
2828 
2829 	ndm = nlmsg_data(nlh);
2830 	ndm->ndm_family  = AF_BRIDGE;
2831 	ndm->ndm_pad1	 = 0;
2832 	ndm->ndm_pad2    = 0;
2833 	ndm->ndm_flags	 = flags;
2834 	ndm->ndm_type	 = 0;
2835 	ndm->ndm_ifindex = dev->ifindex;
2836 	ndm->ndm_state   = ndm_state;
2837 
2838 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2839 		goto nla_put_failure;
2840 	if (vid)
2841 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2842 			goto nla_put_failure;
2843 
2844 	nlmsg_end(skb, nlh);
2845 	return 0;
2846 
2847 nla_put_failure:
2848 	nlmsg_cancel(skb, nlh);
2849 	return -EMSGSIZE;
2850 }
2851 
2852 static inline size_t rtnl_fdb_nlmsg_size(void)
2853 {
2854 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2855 }
2856 
2857 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2858 			    u16 ndm_state)
2859 {
2860 	struct net *net = dev_net(dev);
2861 	struct sk_buff *skb;
2862 	int err = -ENOBUFS;
2863 
2864 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2865 	if (!skb)
2866 		goto errout;
2867 
2868 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2869 				      0, 0, type, NTF_SELF, 0, ndm_state);
2870 	if (err < 0) {
2871 		kfree_skb(skb);
2872 		goto errout;
2873 	}
2874 
2875 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2876 	return;
2877 errout:
2878 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2879 }
2880 
2881 /**
2882  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2883  */
2884 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2885 		     struct nlattr *tb[],
2886 		     struct net_device *dev,
2887 		     const unsigned char *addr, u16 vid,
2888 		     u16 flags)
2889 {
2890 	int err = -EINVAL;
2891 
2892 	/* If aging addresses are supported device will need to
2893 	 * implement its own handler for this.
2894 	 */
2895 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2896 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2897 		return err;
2898 	}
2899 
2900 	if (vid) {
2901 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2902 		return err;
2903 	}
2904 
2905 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2906 		err = dev_uc_add_excl(dev, addr);
2907 	else if (is_multicast_ether_addr(addr))
2908 		err = dev_mc_add_excl(dev, addr);
2909 
2910 	/* Only return duplicate errors if NLM_F_EXCL is set */
2911 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2912 		err = 0;
2913 
2914 	return err;
2915 }
2916 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2917 
2918 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2919 {
2920 	u16 vid = 0;
2921 
2922 	if (vlan_attr) {
2923 		if (nla_len(vlan_attr) != sizeof(u16)) {
2924 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2925 			return -EINVAL;
2926 		}
2927 
2928 		vid = nla_get_u16(vlan_attr);
2929 
2930 		if (!vid || vid >= VLAN_VID_MASK) {
2931 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2932 				vid);
2933 			return -EINVAL;
2934 		}
2935 	}
2936 	*p_vid = vid;
2937 	return 0;
2938 }
2939 
2940 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2941 {
2942 	struct net *net = sock_net(skb->sk);
2943 	struct ndmsg *ndm;
2944 	struct nlattr *tb[NDA_MAX+1];
2945 	struct net_device *dev;
2946 	u8 *addr;
2947 	u16 vid;
2948 	int err;
2949 
2950 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2951 	if (err < 0)
2952 		return err;
2953 
2954 	ndm = nlmsg_data(nlh);
2955 	if (ndm->ndm_ifindex == 0) {
2956 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2957 		return -EINVAL;
2958 	}
2959 
2960 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2961 	if (dev == NULL) {
2962 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2963 		return -ENODEV;
2964 	}
2965 
2966 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2967 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2968 		return -EINVAL;
2969 	}
2970 
2971 	addr = nla_data(tb[NDA_LLADDR]);
2972 
2973 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2974 	if (err)
2975 		return err;
2976 
2977 	err = -EOPNOTSUPP;
2978 
2979 	/* Support fdb on master device the net/bridge default case */
2980 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2981 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2982 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2983 		const struct net_device_ops *ops = br_dev->netdev_ops;
2984 
2985 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2986 				       nlh->nlmsg_flags);
2987 		if (err)
2988 			goto out;
2989 		else
2990 			ndm->ndm_flags &= ~NTF_MASTER;
2991 	}
2992 
2993 	/* Embedded bridge, macvlan, and any other device support */
2994 	if ((ndm->ndm_flags & NTF_SELF)) {
2995 		if (dev->netdev_ops->ndo_fdb_add)
2996 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2997 							   vid,
2998 							   nlh->nlmsg_flags);
2999 		else
3000 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3001 					       nlh->nlmsg_flags);
3002 
3003 		if (!err) {
3004 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3005 					ndm->ndm_state);
3006 			ndm->ndm_flags &= ~NTF_SELF;
3007 		}
3008 	}
3009 out:
3010 	return err;
3011 }
3012 
3013 /**
3014  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3015  */
3016 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3017 		     struct nlattr *tb[],
3018 		     struct net_device *dev,
3019 		     const unsigned char *addr, u16 vid)
3020 {
3021 	int err = -EINVAL;
3022 
3023 	/* If aging addresses are supported device will need to
3024 	 * implement its own handler for this.
3025 	 */
3026 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
3027 		pr_info("%s: FDB only supports static addresses\n", dev->name);
3028 		return err;
3029 	}
3030 
3031 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3032 		err = dev_uc_del(dev, addr);
3033 	else if (is_multicast_ether_addr(addr))
3034 		err = dev_mc_del(dev, addr);
3035 
3036 	return err;
3037 }
3038 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3039 
3040 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3041 {
3042 	struct net *net = sock_net(skb->sk);
3043 	struct ndmsg *ndm;
3044 	struct nlattr *tb[NDA_MAX+1];
3045 	struct net_device *dev;
3046 	int err = -EINVAL;
3047 	__u8 *addr;
3048 	u16 vid;
3049 
3050 	if (!netlink_capable(skb, CAP_NET_ADMIN))
3051 		return -EPERM;
3052 
3053 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3054 	if (err < 0)
3055 		return err;
3056 
3057 	ndm = nlmsg_data(nlh);
3058 	if (ndm->ndm_ifindex == 0) {
3059 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3060 		return -EINVAL;
3061 	}
3062 
3063 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3064 	if (dev == NULL) {
3065 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3066 		return -ENODEV;
3067 	}
3068 
3069 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3070 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3071 		return -EINVAL;
3072 	}
3073 
3074 	addr = nla_data(tb[NDA_LLADDR]);
3075 
3076 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3077 	if (err)
3078 		return err;
3079 
3080 	err = -EOPNOTSUPP;
3081 
3082 	/* Support fdb on master device the net/bridge default case */
3083 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3084 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
3085 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3086 		const struct net_device_ops *ops = br_dev->netdev_ops;
3087 
3088 		if (ops->ndo_fdb_del)
3089 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3090 
3091 		if (err)
3092 			goto out;
3093 		else
3094 			ndm->ndm_flags &= ~NTF_MASTER;
3095 	}
3096 
3097 	/* Embedded bridge, macvlan, and any other device support */
3098 	if (ndm->ndm_flags & NTF_SELF) {
3099 		if (dev->netdev_ops->ndo_fdb_del)
3100 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3101 							   vid);
3102 		else
3103 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3104 
3105 		if (!err) {
3106 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3107 					ndm->ndm_state);
3108 			ndm->ndm_flags &= ~NTF_SELF;
3109 		}
3110 	}
3111 out:
3112 	return err;
3113 }
3114 
3115 static int nlmsg_populate_fdb(struct sk_buff *skb,
3116 			      struct netlink_callback *cb,
3117 			      struct net_device *dev,
3118 			      int *idx,
3119 			      struct netdev_hw_addr_list *list)
3120 {
3121 	struct netdev_hw_addr *ha;
3122 	int err;
3123 	u32 portid, seq;
3124 
3125 	portid = NETLINK_CB(cb->skb).portid;
3126 	seq = cb->nlh->nlmsg_seq;
3127 
3128 	list_for_each_entry(ha, &list->list, list) {
3129 		if (*idx < cb->args[2])
3130 			goto skip;
3131 
3132 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3133 					      portid, seq,
3134 					      RTM_NEWNEIGH, NTF_SELF,
3135 					      NLM_F_MULTI, NUD_PERMANENT);
3136 		if (err < 0)
3137 			return err;
3138 skip:
3139 		*idx += 1;
3140 	}
3141 	return 0;
3142 }
3143 
3144 /**
3145  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3146  * @nlh: netlink message header
3147  * @dev: netdevice
3148  *
3149  * Default netdevice operation to dump the existing unicast address list.
3150  * Returns number of addresses from list put in skb.
3151  */
3152 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3153 		      struct netlink_callback *cb,
3154 		      struct net_device *dev,
3155 		      struct net_device *filter_dev,
3156 		      int *idx)
3157 {
3158 	int err;
3159 
3160 	netif_addr_lock_bh(dev);
3161 	err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3162 	if (err)
3163 		goto out;
3164 	nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3165 out:
3166 	netif_addr_unlock_bh(dev);
3167 	return err;
3168 }
3169 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3170 
3171 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3172 {
3173 	struct net_device *dev;
3174 	struct nlattr *tb[IFLA_MAX+1];
3175 	struct net_device *br_dev = NULL;
3176 	const struct net_device_ops *ops = NULL;
3177 	const struct net_device_ops *cops = NULL;
3178 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3179 	struct net *net = sock_net(skb->sk);
3180 	struct hlist_head *head;
3181 	int brport_idx = 0;
3182 	int br_idx = 0;
3183 	int h, s_h;
3184 	int idx = 0, s_idx;
3185 	int err = 0;
3186 	int fidx = 0;
3187 
3188 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3189 			ifla_policy) == 0) {
3190 		if (tb[IFLA_MASTER])
3191 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3192 	}
3193 
3194 	brport_idx = ifm->ifi_index;
3195 
3196 	if (br_idx) {
3197 		br_dev = __dev_get_by_index(net, br_idx);
3198 		if (!br_dev)
3199 			return -ENODEV;
3200 
3201 		ops = br_dev->netdev_ops;
3202 	}
3203 
3204 	s_h = cb->args[0];
3205 	s_idx = cb->args[1];
3206 
3207 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3208 		idx = 0;
3209 		head = &net->dev_index_head[h];
3210 		hlist_for_each_entry(dev, head, index_hlist) {
3211 
3212 			if (brport_idx && (dev->ifindex != brport_idx))
3213 				continue;
3214 
3215 			if (!br_idx) { /* user did not specify a specific bridge */
3216 				if (dev->priv_flags & IFF_BRIDGE_PORT) {
3217 					br_dev = netdev_master_upper_dev_get(dev);
3218 					cops = br_dev->netdev_ops;
3219 				}
3220 			} else {
3221 				if (dev != br_dev &&
3222 				    !(dev->priv_flags & IFF_BRIDGE_PORT))
3223 					continue;
3224 
3225 				if (br_dev != netdev_master_upper_dev_get(dev) &&
3226 				    !(dev->priv_flags & IFF_EBRIDGE))
3227 					continue;
3228 				cops = ops;
3229 			}
3230 
3231 			if (idx < s_idx)
3232 				goto cont;
3233 
3234 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3235 				if (cops && cops->ndo_fdb_dump) {
3236 					err = cops->ndo_fdb_dump(skb, cb,
3237 								br_dev, dev,
3238 								&fidx);
3239 					if (err == -EMSGSIZE)
3240 						goto out;
3241 				}
3242 			}
3243 
3244 			if (dev->netdev_ops->ndo_fdb_dump)
3245 				err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3246 								    dev, NULL,
3247 								    &fidx);
3248 			else
3249 				err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3250 							&fidx);
3251 			if (err == -EMSGSIZE)
3252 				goto out;
3253 
3254 			cops = NULL;
3255 
3256 			/* reset fdb offset to 0 for rest of the interfaces */
3257 			cb->args[2] = 0;
3258 			fidx = 0;
3259 cont:
3260 			idx++;
3261 		}
3262 	}
3263 
3264 out:
3265 	cb->args[0] = h;
3266 	cb->args[1] = idx;
3267 	cb->args[2] = fidx;
3268 
3269 	return skb->len;
3270 }
3271 
3272 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3273 			       unsigned int attrnum, unsigned int flag)
3274 {
3275 	if (mask & flag)
3276 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3277 	return 0;
3278 }
3279 
3280 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3281 			    struct net_device *dev, u16 mode,
3282 			    u32 flags, u32 mask, int nlflags,
3283 			    u32 filter_mask,
3284 			    int (*vlan_fill)(struct sk_buff *skb,
3285 					     struct net_device *dev,
3286 					     u32 filter_mask))
3287 {
3288 	struct nlmsghdr *nlh;
3289 	struct ifinfomsg *ifm;
3290 	struct nlattr *br_afspec;
3291 	struct nlattr *protinfo;
3292 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3293 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3294 	int err = 0;
3295 
3296 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3297 	if (nlh == NULL)
3298 		return -EMSGSIZE;
3299 
3300 	ifm = nlmsg_data(nlh);
3301 	ifm->ifi_family = AF_BRIDGE;
3302 	ifm->__ifi_pad = 0;
3303 	ifm->ifi_type = dev->type;
3304 	ifm->ifi_index = dev->ifindex;
3305 	ifm->ifi_flags = dev_get_flags(dev);
3306 	ifm->ifi_change = 0;
3307 
3308 
3309 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3310 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3311 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3312 	    (br_dev &&
3313 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3314 	    (dev->addr_len &&
3315 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3316 	    (dev->ifindex != dev_get_iflink(dev) &&
3317 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3318 		goto nla_put_failure;
3319 
3320 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3321 	if (!br_afspec)
3322 		goto nla_put_failure;
3323 
3324 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3325 		nla_nest_cancel(skb, br_afspec);
3326 		goto nla_put_failure;
3327 	}
3328 
3329 	if (mode != BRIDGE_MODE_UNDEF) {
3330 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3331 			nla_nest_cancel(skb, br_afspec);
3332 			goto nla_put_failure;
3333 		}
3334 	}
3335 	if (vlan_fill) {
3336 		err = vlan_fill(skb, dev, filter_mask);
3337 		if (err) {
3338 			nla_nest_cancel(skb, br_afspec);
3339 			goto nla_put_failure;
3340 		}
3341 	}
3342 	nla_nest_end(skb, br_afspec);
3343 
3344 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3345 	if (!protinfo)
3346 		goto nla_put_failure;
3347 
3348 	if (brport_nla_put_flag(skb, flags, mask,
3349 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3350 	    brport_nla_put_flag(skb, flags, mask,
3351 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3352 	    brport_nla_put_flag(skb, flags, mask,
3353 				IFLA_BRPORT_FAST_LEAVE,
3354 				BR_MULTICAST_FAST_LEAVE) ||
3355 	    brport_nla_put_flag(skb, flags, mask,
3356 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3357 	    brport_nla_put_flag(skb, flags, mask,
3358 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3359 	    brport_nla_put_flag(skb, flags, mask,
3360 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3361 	    brport_nla_put_flag(skb, flags, mask,
3362 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3363 	    brport_nla_put_flag(skb, flags, mask,
3364 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3365 		nla_nest_cancel(skb, protinfo);
3366 		goto nla_put_failure;
3367 	}
3368 
3369 	nla_nest_end(skb, protinfo);
3370 
3371 	nlmsg_end(skb, nlh);
3372 	return 0;
3373 nla_put_failure:
3374 	nlmsg_cancel(skb, nlh);
3375 	return err ? err : -EMSGSIZE;
3376 }
3377 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3378 
3379 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3380 {
3381 	struct net *net = sock_net(skb->sk);
3382 	struct net_device *dev;
3383 	int idx = 0;
3384 	u32 portid = NETLINK_CB(cb->skb).portid;
3385 	u32 seq = cb->nlh->nlmsg_seq;
3386 	u32 filter_mask = 0;
3387 	int err;
3388 
3389 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3390 		struct nlattr *extfilt;
3391 
3392 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3393 					  IFLA_EXT_MASK);
3394 		if (extfilt) {
3395 			if (nla_len(extfilt) < sizeof(filter_mask))
3396 				return -EINVAL;
3397 
3398 			filter_mask = nla_get_u32(extfilt);
3399 		}
3400 	}
3401 
3402 	rcu_read_lock();
3403 	for_each_netdev_rcu(net, dev) {
3404 		const struct net_device_ops *ops = dev->netdev_ops;
3405 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3406 
3407 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3408 			if (idx >= cb->args[0]) {
3409 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3410 						skb, portid, seq, dev,
3411 						filter_mask, NLM_F_MULTI);
3412 				if (err < 0 && err != -EOPNOTSUPP)
3413 					break;
3414 			}
3415 			idx++;
3416 		}
3417 
3418 		if (ops->ndo_bridge_getlink) {
3419 			if (idx >= cb->args[0]) {
3420 				err = ops->ndo_bridge_getlink(skb, portid,
3421 							      seq, dev,
3422 							      filter_mask,
3423 							      NLM_F_MULTI);
3424 				if (err < 0 && err != -EOPNOTSUPP)
3425 					break;
3426 			}
3427 			idx++;
3428 		}
3429 	}
3430 	rcu_read_unlock();
3431 	cb->args[0] = idx;
3432 
3433 	return skb->len;
3434 }
3435 
3436 static inline size_t bridge_nlmsg_size(void)
3437 {
3438 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3439 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3440 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3441 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3442 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3443 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3444 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3445 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3446 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3447 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3448 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3449 }
3450 
3451 static int rtnl_bridge_notify(struct net_device *dev)
3452 {
3453 	struct net *net = dev_net(dev);
3454 	struct sk_buff *skb;
3455 	int err = -EOPNOTSUPP;
3456 
3457 	if (!dev->netdev_ops->ndo_bridge_getlink)
3458 		return 0;
3459 
3460 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3461 	if (!skb) {
3462 		err = -ENOMEM;
3463 		goto errout;
3464 	}
3465 
3466 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3467 	if (err < 0)
3468 		goto errout;
3469 
3470 	if (!skb->len)
3471 		goto errout;
3472 
3473 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3474 	return 0;
3475 errout:
3476 	WARN_ON(err == -EMSGSIZE);
3477 	kfree_skb(skb);
3478 	if (err)
3479 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3480 	return err;
3481 }
3482 
3483 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3484 {
3485 	struct net *net = sock_net(skb->sk);
3486 	struct ifinfomsg *ifm;
3487 	struct net_device *dev;
3488 	struct nlattr *br_spec, *attr = NULL;
3489 	int rem, err = -EOPNOTSUPP;
3490 	u16 flags = 0;
3491 	bool have_flags = false;
3492 
3493 	if (nlmsg_len(nlh) < sizeof(*ifm))
3494 		return -EINVAL;
3495 
3496 	ifm = nlmsg_data(nlh);
3497 	if (ifm->ifi_family != AF_BRIDGE)
3498 		return -EPFNOSUPPORT;
3499 
3500 	dev = __dev_get_by_index(net, ifm->ifi_index);
3501 	if (!dev) {
3502 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3503 		return -ENODEV;
3504 	}
3505 
3506 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3507 	if (br_spec) {
3508 		nla_for_each_nested(attr, br_spec, rem) {
3509 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3510 				if (nla_len(attr) < sizeof(flags))
3511 					return -EINVAL;
3512 
3513 				have_flags = true;
3514 				flags = nla_get_u16(attr);
3515 				break;
3516 			}
3517 		}
3518 	}
3519 
3520 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3521 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3522 
3523 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3524 			err = -EOPNOTSUPP;
3525 			goto out;
3526 		}
3527 
3528 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3529 		if (err)
3530 			goto out;
3531 
3532 		flags &= ~BRIDGE_FLAGS_MASTER;
3533 	}
3534 
3535 	if ((flags & BRIDGE_FLAGS_SELF)) {
3536 		if (!dev->netdev_ops->ndo_bridge_setlink)
3537 			err = -EOPNOTSUPP;
3538 		else
3539 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3540 								  flags);
3541 		if (!err) {
3542 			flags &= ~BRIDGE_FLAGS_SELF;
3543 
3544 			/* Generate event to notify upper layer of bridge
3545 			 * change
3546 			 */
3547 			err = rtnl_bridge_notify(dev);
3548 		}
3549 	}
3550 
3551 	if (have_flags)
3552 		memcpy(nla_data(attr), &flags, sizeof(flags));
3553 out:
3554 	return err;
3555 }
3556 
3557 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3558 {
3559 	struct net *net = sock_net(skb->sk);
3560 	struct ifinfomsg *ifm;
3561 	struct net_device *dev;
3562 	struct nlattr *br_spec, *attr = NULL;
3563 	int rem, err = -EOPNOTSUPP;
3564 	u16 flags = 0;
3565 	bool have_flags = false;
3566 
3567 	if (nlmsg_len(nlh) < sizeof(*ifm))
3568 		return -EINVAL;
3569 
3570 	ifm = nlmsg_data(nlh);
3571 	if (ifm->ifi_family != AF_BRIDGE)
3572 		return -EPFNOSUPPORT;
3573 
3574 	dev = __dev_get_by_index(net, ifm->ifi_index);
3575 	if (!dev) {
3576 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3577 		return -ENODEV;
3578 	}
3579 
3580 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3581 	if (br_spec) {
3582 		nla_for_each_nested(attr, br_spec, rem) {
3583 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3584 				if (nla_len(attr) < sizeof(flags))
3585 					return -EINVAL;
3586 
3587 				have_flags = true;
3588 				flags = nla_get_u16(attr);
3589 				break;
3590 			}
3591 		}
3592 	}
3593 
3594 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3595 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3596 
3597 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3598 			err = -EOPNOTSUPP;
3599 			goto out;
3600 		}
3601 
3602 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3603 		if (err)
3604 			goto out;
3605 
3606 		flags &= ~BRIDGE_FLAGS_MASTER;
3607 	}
3608 
3609 	if ((flags & BRIDGE_FLAGS_SELF)) {
3610 		if (!dev->netdev_ops->ndo_bridge_dellink)
3611 			err = -EOPNOTSUPP;
3612 		else
3613 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3614 								  flags);
3615 
3616 		if (!err) {
3617 			flags &= ~BRIDGE_FLAGS_SELF;
3618 
3619 			/* Generate event to notify upper layer of bridge
3620 			 * change
3621 			 */
3622 			err = rtnl_bridge_notify(dev);
3623 		}
3624 	}
3625 
3626 	if (have_flags)
3627 		memcpy(nla_data(attr), &flags, sizeof(flags));
3628 out:
3629 	return err;
3630 }
3631 
3632 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3633 {
3634 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3635 	       (!idxattr || idxattr == attrid);
3636 }
3637 
3638 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3639 static int rtnl_get_offload_stats_attr_size(int attr_id)
3640 {
3641 	switch (attr_id) {
3642 	case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3643 		return sizeof(struct rtnl_link_stats64);
3644 	}
3645 
3646 	return 0;
3647 }
3648 
3649 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3650 				  int *prividx)
3651 {
3652 	struct nlattr *attr = NULL;
3653 	int attr_id, size;
3654 	void *attr_data;
3655 	int err;
3656 
3657 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3658 	      dev->netdev_ops->ndo_get_offload_stats))
3659 		return -ENODATA;
3660 
3661 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3662 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3663 		if (attr_id < *prividx)
3664 			continue;
3665 
3666 		size = rtnl_get_offload_stats_attr_size(attr_id);
3667 		if (!size)
3668 			continue;
3669 
3670 		if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3671 			continue;
3672 
3673 		attr = nla_reserve_64bit(skb, attr_id, size,
3674 					 IFLA_OFFLOAD_XSTATS_UNSPEC);
3675 		if (!attr)
3676 			goto nla_put_failure;
3677 
3678 		attr_data = nla_data(attr);
3679 		memset(attr_data, 0, size);
3680 		err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3681 							     attr_data);
3682 		if (err)
3683 			goto get_offload_stats_failure;
3684 	}
3685 
3686 	if (!attr)
3687 		return -ENODATA;
3688 
3689 	*prividx = 0;
3690 	return 0;
3691 
3692 nla_put_failure:
3693 	err = -EMSGSIZE;
3694 get_offload_stats_failure:
3695 	*prividx = attr_id;
3696 	return err;
3697 }
3698 
3699 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3700 {
3701 	int nla_size = 0;
3702 	int attr_id;
3703 	int size;
3704 
3705 	if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3706 	      dev->netdev_ops->ndo_get_offload_stats))
3707 		return 0;
3708 
3709 	for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3710 	     attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3711 		if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3712 			continue;
3713 		size = rtnl_get_offload_stats_attr_size(attr_id);
3714 		nla_size += nla_total_size_64bit(size);
3715 	}
3716 
3717 	if (nla_size != 0)
3718 		nla_size += nla_total_size(0);
3719 
3720 	return nla_size;
3721 }
3722 
3723 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3724 			       int type, u32 pid, u32 seq, u32 change,
3725 			       unsigned int flags, unsigned int filter_mask,
3726 			       int *idxattr, int *prividx)
3727 {
3728 	struct if_stats_msg *ifsm;
3729 	struct nlmsghdr *nlh;
3730 	struct nlattr *attr;
3731 	int s_prividx = *prividx;
3732 	int err;
3733 
3734 	ASSERT_RTNL();
3735 
3736 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3737 	if (!nlh)
3738 		return -EMSGSIZE;
3739 
3740 	ifsm = nlmsg_data(nlh);
3741 	ifsm->ifindex = dev->ifindex;
3742 	ifsm->filter_mask = filter_mask;
3743 
3744 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3745 		struct rtnl_link_stats64 *sp;
3746 
3747 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3748 					 sizeof(struct rtnl_link_stats64),
3749 					 IFLA_STATS_UNSPEC);
3750 		if (!attr)
3751 			goto nla_put_failure;
3752 
3753 		sp = nla_data(attr);
3754 		dev_get_stats(dev, sp);
3755 	}
3756 
3757 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3758 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3759 
3760 		if (ops && ops->fill_linkxstats) {
3761 			*idxattr = IFLA_STATS_LINK_XSTATS;
3762 			attr = nla_nest_start(skb,
3763 					      IFLA_STATS_LINK_XSTATS);
3764 			if (!attr)
3765 				goto nla_put_failure;
3766 
3767 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3768 			nla_nest_end(skb, attr);
3769 			if (err)
3770 				goto nla_put_failure;
3771 			*idxattr = 0;
3772 		}
3773 	}
3774 
3775 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3776 			     *idxattr)) {
3777 		const struct rtnl_link_ops *ops = NULL;
3778 		const struct net_device *master;
3779 
3780 		master = netdev_master_upper_dev_get(dev);
3781 		if (master)
3782 			ops = master->rtnl_link_ops;
3783 		if (ops && ops->fill_linkxstats) {
3784 			*idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3785 			attr = nla_nest_start(skb,
3786 					      IFLA_STATS_LINK_XSTATS_SLAVE);
3787 			if (!attr)
3788 				goto nla_put_failure;
3789 
3790 			err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3791 			nla_nest_end(skb, attr);
3792 			if (err)
3793 				goto nla_put_failure;
3794 			*idxattr = 0;
3795 		}
3796 	}
3797 
3798 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3799 			     *idxattr)) {
3800 		*idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3801 		attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3802 		if (!attr)
3803 			goto nla_put_failure;
3804 
3805 		err = rtnl_get_offload_stats(skb, dev, prividx);
3806 		if (err == -ENODATA)
3807 			nla_nest_cancel(skb, attr);
3808 		else
3809 			nla_nest_end(skb, attr);
3810 
3811 		if (err && err != -ENODATA)
3812 			goto nla_put_failure;
3813 		*idxattr = 0;
3814 	}
3815 
3816 	nlmsg_end(skb, nlh);
3817 
3818 	return 0;
3819 
3820 nla_put_failure:
3821 	/* not a multi message or no progress mean a real error */
3822 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3823 		nlmsg_cancel(skb, nlh);
3824 	else
3825 		nlmsg_end(skb, nlh);
3826 
3827 	return -EMSGSIZE;
3828 }
3829 
3830 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3831 				  u32 filter_mask)
3832 {
3833 	size_t size = 0;
3834 
3835 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3836 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3837 
3838 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3839 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3840 		int attr = IFLA_STATS_LINK_XSTATS;
3841 
3842 		if (ops && ops->get_linkxstats_size) {
3843 			size += nla_total_size(ops->get_linkxstats_size(dev,
3844 									attr));
3845 			/* for IFLA_STATS_LINK_XSTATS */
3846 			size += nla_total_size(0);
3847 		}
3848 	}
3849 
3850 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3851 		struct net_device *_dev = (struct net_device *)dev;
3852 		const struct rtnl_link_ops *ops = NULL;
3853 		const struct net_device *master;
3854 
3855 		/* netdev_master_upper_dev_get can't take const */
3856 		master = netdev_master_upper_dev_get(_dev);
3857 		if (master)
3858 			ops = master->rtnl_link_ops;
3859 		if (ops && ops->get_linkxstats_size) {
3860 			int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3861 
3862 			size += nla_total_size(ops->get_linkxstats_size(dev,
3863 									attr));
3864 			/* for IFLA_STATS_LINK_XSTATS_SLAVE */
3865 			size += nla_total_size(0);
3866 		}
3867 	}
3868 
3869 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3870 		size += rtnl_get_offload_stats_size(dev);
3871 
3872 	return size;
3873 }
3874 
3875 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3876 {
3877 	struct net *net = sock_net(skb->sk);
3878 	struct net_device *dev = NULL;
3879 	int idxattr = 0, prividx = 0;
3880 	struct if_stats_msg *ifsm;
3881 	struct sk_buff *nskb;
3882 	u32 filter_mask;
3883 	int err;
3884 
3885 	ifsm = nlmsg_data(nlh);
3886 	if (ifsm->ifindex > 0)
3887 		dev = __dev_get_by_index(net, ifsm->ifindex);
3888 	else
3889 		return -EINVAL;
3890 
3891 	if (!dev)
3892 		return -ENODEV;
3893 
3894 	filter_mask = ifsm->filter_mask;
3895 	if (!filter_mask)
3896 		return -EINVAL;
3897 
3898 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3899 	if (!nskb)
3900 		return -ENOBUFS;
3901 
3902 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3903 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3904 				  0, filter_mask, &idxattr, &prividx);
3905 	if (err < 0) {
3906 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3907 		WARN_ON(err == -EMSGSIZE);
3908 		kfree_skb(nskb);
3909 	} else {
3910 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3911 	}
3912 
3913 	return err;
3914 }
3915 
3916 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3917 {
3918 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
3919 	struct net *net = sock_net(skb->sk);
3920 	unsigned int flags = NLM_F_MULTI;
3921 	struct if_stats_msg *ifsm;
3922 	struct hlist_head *head;
3923 	struct net_device *dev;
3924 	u32 filter_mask = 0;
3925 	int idx = 0;
3926 
3927 	s_h = cb->args[0];
3928 	s_idx = cb->args[1];
3929 	s_idxattr = cb->args[2];
3930 	s_prividx = cb->args[3];
3931 
3932 	cb->seq = net->dev_base_seq;
3933 
3934 	ifsm = nlmsg_data(cb->nlh);
3935 	filter_mask = ifsm->filter_mask;
3936 	if (!filter_mask)
3937 		return -EINVAL;
3938 
3939 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3940 		idx = 0;
3941 		head = &net->dev_index_head[h];
3942 		hlist_for_each_entry(dev, head, index_hlist) {
3943 			if (idx < s_idx)
3944 				goto cont;
3945 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3946 						  NETLINK_CB(cb->skb).portid,
3947 						  cb->nlh->nlmsg_seq, 0,
3948 						  flags, filter_mask,
3949 						  &s_idxattr, &s_prividx);
3950 			/* If we ran out of room on the first message,
3951 			 * we're in trouble
3952 			 */
3953 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3954 
3955 			if (err < 0)
3956 				goto out;
3957 			s_prividx = 0;
3958 			s_idxattr = 0;
3959 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3960 cont:
3961 			idx++;
3962 		}
3963 	}
3964 out:
3965 	cb->args[3] = s_prividx;
3966 	cb->args[2] = s_idxattr;
3967 	cb->args[1] = idx;
3968 	cb->args[0] = h;
3969 
3970 	return skb->len;
3971 }
3972 
3973 /* Process one rtnetlink message. */
3974 
3975 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3976 {
3977 	struct net *net = sock_net(skb->sk);
3978 	rtnl_doit_func doit;
3979 	int kind;
3980 	int family;
3981 	int type;
3982 	int err;
3983 
3984 	type = nlh->nlmsg_type;
3985 	if (type > RTM_MAX)
3986 		return -EOPNOTSUPP;
3987 
3988 	type -= RTM_BASE;
3989 
3990 	/* All the messages must have at least 1 byte length */
3991 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3992 		return 0;
3993 
3994 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3995 	kind = type&3;
3996 
3997 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3998 		return -EPERM;
3999 
4000 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4001 		struct sock *rtnl;
4002 		rtnl_dumpit_func dumpit;
4003 		rtnl_calcit_func calcit;
4004 		u16 min_dump_alloc = 0;
4005 
4006 		dumpit = rtnl_get_dumpit(family, type);
4007 		if (dumpit == NULL)
4008 			return -EOPNOTSUPP;
4009 		calcit = rtnl_get_calcit(family, type);
4010 		if (calcit)
4011 			min_dump_alloc = calcit(skb, nlh);
4012 
4013 		__rtnl_unlock();
4014 		rtnl = net->rtnl;
4015 		{
4016 			struct netlink_dump_control c = {
4017 				.dump		= dumpit,
4018 				.min_dump_alloc	= min_dump_alloc,
4019 			};
4020 			err = netlink_dump_start(rtnl, skb, nlh, &c);
4021 		}
4022 		rtnl_lock();
4023 		return err;
4024 	}
4025 
4026 	doit = rtnl_get_doit(family, type);
4027 	if (doit == NULL)
4028 		return -EOPNOTSUPP;
4029 
4030 	return doit(skb, nlh);
4031 }
4032 
4033 static void rtnetlink_rcv(struct sk_buff *skb)
4034 {
4035 	rtnl_lock();
4036 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4037 	rtnl_unlock();
4038 }
4039 
4040 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4041 {
4042 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4043 
4044 	switch (event) {
4045 	case NETDEV_UP:
4046 	case NETDEV_DOWN:
4047 	case NETDEV_PRE_UP:
4048 	case NETDEV_POST_INIT:
4049 	case NETDEV_REGISTER:
4050 	case NETDEV_CHANGE:
4051 	case NETDEV_PRE_TYPE_CHANGE:
4052 	case NETDEV_GOING_DOWN:
4053 	case NETDEV_UNREGISTER:
4054 	case NETDEV_UNREGISTER_FINAL:
4055 	case NETDEV_RELEASE:
4056 	case NETDEV_JOIN:
4057 	case NETDEV_BONDING_INFO:
4058 		break;
4059 	default:
4060 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4061 		break;
4062 	}
4063 	return NOTIFY_DONE;
4064 }
4065 
4066 static struct notifier_block rtnetlink_dev_notifier = {
4067 	.notifier_call	= rtnetlink_event,
4068 };
4069 
4070 
4071 static int __net_init rtnetlink_net_init(struct net *net)
4072 {
4073 	struct sock *sk;
4074 	struct netlink_kernel_cfg cfg = {
4075 		.groups		= RTNLGRP_MAX,
4076 		.input		= rtnetlink_rcv,
4077 		.cb_mutex	= &rtnl_mutex,
4078 		.flags		= NL_CFG_F_NONROOT_RECV,
4079 	};
4080 
4081 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4082 	if (!sk)
4083 		return -ENOMEM;
4084 	net->rtnl = sk;
4085 	return 0;
4086 }
4087 
4088 static void __net_exit rtnetlink_net_exit(struct net *net)
4089 {
4090 	netlink_kernel_release(net->rtnl);
4091 	net->rtnl = NULL;
4092 }
4093 
4094 static struct pernet_operations rtnetlink_net_ops = {
4095 	.init = rtnetlink_net_init,
4096 	.exit = rtnetlink_net_exit,
4097 };
4098 
4099 void __init rtnetlink_init(void)
4100 {
4101 	if (register_pernet_subsys(&rtnetlink_net_ops))
4102 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
4103 
4104 	register_netdevice_notifier(&rtnetlink_dev_notifier);
4105 
4106 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4107 		      rtnl_dump_ifinfo, rtnl_calcit);
4108 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4109 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4110 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4111 
4112 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4113 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4114 
4115 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4116 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4117 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4118 
4119 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4120 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4121 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4122 
4123 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4124 		      NULL);
4125 }
4126