xref: /openbmc/linux/net/core/rtnetlink.c (revision 1b5c5493e3e68181be344cb51bf9df192d05ffc2)
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 
279 	return 0;
280 }
281 EXPORT_SYMBOL_GPL(rtnl_unregister);
282 
283 /**
284  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
285  * @protocol : Protocol family or PF_UNSPEC
286  *
287  * Identical to calling rtnl_unregster() for all registered message types
288  * of a certain protocol family.
289  */
290 void rtnl_unregister_all(int protocol)
291 {
292 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
293 
294 	kfree(rtnl_msg_handlers[protocol]);
295 	rtnl_msg_handlers[protocol] = NULL;
296 }
297 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
298 
299 static LIST_HEAD(link_ops);
300 
301 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
302 {
303 	const struct rtnl_link_ops *ops;
304 
305 	list_for_each_entry(ops, &link_ops, list) {
306 		if (!strcmp(ops->kind, kind))
307 			return ops;
308 	}
309 	return NULL;
310 }
311 
312 /**
313  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
314  * @ops: struct rtnl_link_ops * to register
315  *
316  * The caller must hold the rtnl_mutex. This function should be used
317  * by drivers that create devices during module initialization. It
318  * must be called before registering the devices.
319  *
320  * Returns 0 on success or a negative error code.
321  */
322 int __rtnl_link_register(struct rtnl_link_ops *ops)
323 {
324 	if (rtnl_link_ops_get(ops->kind))
325 		return -EEXIST;
326 
327 	/* The check for setup is here because if ops
328 	 * does not have that filled up, it is not possible
329 	 * to use the ops for creating device. So do not
330 	 * fill up dellink as well. That disables rtnl_dellink.
331 	 */
332 	if (ops->setup && !ops->dellink)
333 		ops->dellink = unregister_netdevice_queue;
334 
335 	list_add_tail(&ops->list, &link_ops);
336 	return 0;
337 }
338 EXPORT_SYMBOL_GPL(__rtnl_link_register);
339 
340 /**
341  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
342  * @ops: struct rtnl_link_ops * to register
343  *
344  * Returns 0 on success or a negative error code.
345  */
346 int rtnl_link_register(struct rtnl_link_ops *ops)
347 {
348 	int err;
349 
350 	rtnl_lock();
351 	err = __rtnl_link_register(ops);
352 	rtnl_unlock();
353 	return err;
354 }
355 EXPORT_SYMBOL_GPL(rtnl_link_register);
356 
357 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
358 {
359 	struct net_device *dev;
360 	LIST_HEAD(list_kill);
361 
362 	for_each_netdev(net, dev) {
363 		if (dev->rtnl_link_ops == ops)
364 			ops->dellink(dev, &list_kill);
365 	}
366 	unregister_netdevice_many(&list_kill);
367 }
368 
369 /**
370  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
371  * @ops: struct rtnl_link_ops * to unregister
372  *
373  * The caller must hold the rtnl_mutex.
374  */
375 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
376 {
377 	struct net *net;
378 
379 	for_each_net(net) {
380 		__rtnl_kill_links(net, ops);
381 	}
382 	list_del(&ops->list);
383 }
384 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
385 
386 /* Return with the rtnl_lock held when there are no network
387  * devices unregistering in any network namespace.
388  */
389 static void rtnl_lock_unregistering_all(void)
390 {
391 	struct net *net;
392 	bool unregistering;
393 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
394 
395 	add_wait_queue(&netdev_unregistering_wq, &wait);
396 	for (;;) {
397 		unregistering = false;
398 		rtnl_lock();
399 		for_each_net(net) {
400 			if (net->dev_unreg_count > 0) {
401 				unregistering = true;
402 				break;
403 			}
404 		}
405 		if (!unregistering)
406 			break;
407 		__rtnl_unlock();
408 
409 		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
410 	}
411 	remove_wait_queue(&netdev_unregistering_wq, &wait);
412 }
413 
414 /**
415  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
416  * @ops: struct rtnl_link_ops * to unregister
417  */
418 void rtnl_link_unregister(struct rtnl_link_ops *ops)
419 {
420 	/* Close the race with cleanup_net() */
421 	mutex_lock(&net_mutex);
422 	rtnl_lock_unregistering_all();
423 	__rtnl_link_unregister(ops);
424 	rtnl_unlock();
425 	mutex_unlock(&net_mutex);
426 }
427 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
428 
429 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
430 {
431 	struct net_device *master_dev;
432 	const struct rtnl_link_ops *ops;
433 
434 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
435 	if (!master_dev)
436 		return 0;
437 	ops = master_dev->rtnl_link_ops;
438 	if (!ops || !ops->get_slave_size)
439 		return 0;
440 	/* IFLA_INFO_SLAVE_DATA + nested data */
441 	return nla_total_size(sizeof(struct nlattr)) +
442 	       ops->get_slave_size(master_dev, dev);
443 }
444 
445 static size_t rtnl_link_get_size(const struct net_device *dev)
446 {
447 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
448 	size_t size;
449 
450 	if (!ops)
451 		return 0;
452 
453 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
454 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
455 
456 	if (ops->get_size)
457 		/* IFLA_INFO_DATA + nested data */
458 		size += nla_total_size(sizeof(struct nlattr)) +
459 			ops->get_size(dev);
460 
461 	if (ops->get_xstats_size)
462 		/* IFLA_INFO_XSTATS */
463 		size += nla_total_size(ops->get_xstats_size(dev));
464 
465 	size += rtnl_link_get_slave_info_data_size(dev);
466 
467 	return size;
468 }
469 
470 static LIST_HEAD(rtnl_af_ops);
471 
472 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
473 {
474 	const struct rtnl_af_ops *ops;
475 
476 	list_for_each_entry(ops, &rtnl_af_ops, list) {
477 		if (ops->family == family)
478 			return ops;
479 	}
480 
481 	return NULL;
482 }
483 
484 /**
485  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
486  * @ops: struct rtnl_af_ops * to register
487  *
488  * Returns 0 on success or a negative error code.
489  */
490 void rtnl_af_register(struct rtnl_af_ops *ops)
491 {
492 	rtnl_lock();
493 	list_add_tail(&ops->list, &rtnl_af_ops);
494 	rtnl_unlock();
495 }
496 EXPORT_SYMBOL_GPL(rtnl_af_register);
497 
498 /**
499  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
500  * @ops: struct rtnl_af_ops * to unregister
501  *
502  * The caller must hold the rtnl_mutex.
503  */
504 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
505 {
506 	list_del(&ops->list);
507 }
508 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
509 
510 /**
511  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
512  * @ops: struct rtnl_af_ops * to unregister
513  */
514 void rtnl_af_unregister(struct rtnl_af_ops *ops)
515 {
516 	rtnl_lock();
517 	__rtnl_af_unregister(ops);
518 	rtnl_unlock();
519 }
520 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
521 
522 static size_t rtnl_link_get_af_size(const struct net_device *dev,
523 				    u32 ext_filter_mask)
524 {
525 	struct rtnl_af_ops *af_ops;
526 	size_t size;
527 
528 	/* IFLA_AF_SPEC */
529 	size = nla_total_size(sizeof(struct nlattr));
530 
531 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
532 		if (af_ops->get_link_af_size) {
533 			/* AF_* + nested data */
534 			size += nla_total_size(sizeof(struct nlattr)) +
535 				af_ops->get_link_af_size(dev, ext_filter_mask);
536 		}
537 	}
538 
539 	return size;
540 }
541 
542 static bool rtnl_have_link_slave_info(const struct net_device *dev)
543 {
544 	struct net_device *master_dev;
545 
546 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
547 	if (master_dev && master_dev->rtnl_link_ops)
548 		return true;
549 	return false;
550 }
551 
552 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
553 				     const struct net_device *dev)
554 {
555 	struct net_device *master_dev;
556 	const struct rtnl_link_ops *ops;
557 	struct nlattr *slave_data;
558 	int err;
559 
560 	master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
561 	if (!master_dev)
562 		return 0;
563 	ops = master_dev->rtnl_link_ops;
564 	if (!ops)
565 		return 0;
566 	if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
567 		return -EMSGSIZE;
568 	if (ops->fill_slave_info) {
569 		slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
570 		if (!slave_data)
571 			return -EMSGSIZE;
572 		err = ops->fill_slave_info(skb, master_dev, dev);
573 		if (err < 0)
574 			goto err_cancel_slave_data;
575 		nla_nest_end(skb, slave_data);
576 	}
577 	return 0;
578 
579 err_cancel_slave_data:
580 	nla_nest_cancel(skb, slave_data);
581 	return err;
582 }
583 
584 static int rtnl_link_info_fill(struct sk_buff *skb,
585 			       const struct net_device *dev)
586 {
587 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
588 	struct nlattr *data;
589 	int err;
590 
591 	if (!ops)
592 		return 0;
593 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
594 		return -EMSGSIZE;
595 	if (ops->fill_xstats) {
596 		err = ops->fill_xstats(skb, dev);
597 		if (err < 0)
598 			return err;
599 	}
600 	if (ops->fill_info) {
601 		data = nla_nest_start(skb, IFLA_INFO_DATA);
602 		if (data == NULL)
603 			return -EMSGSIZE;
604 		err = ops->fill_info(skb, dev);
605 		if (err < 0)
606 			goto err_cancel_data;
607 		nla_nest_end(skb, data);
608 	}
609 	return 0;
610 
611 err_cancel_data:
612 	nla_nest_cancel(skb, data);
613 	return err;
614 }
615 
616 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
617 {
618 	struct nlattr *linkinfo;
619 	int err = -EMSGSIZE;
620 
621 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
622 	if (linkinfo == NULL)
623 		goto out;
624 
625 	err = rtnl_link_info_fill(skb, dev);
626 	if (err < 0)
627 		goto err_cancel_link;
628 
629 	err = rtnl_link_slave_info_fill(skb, dev);
630 	if (err < 0)
631 		goto err_cancel_link;
632 
633 	nla_nest_end(skb, linkinfo);
634 	return 0;
635 
636 err_cancel_link:
637 	nla_nest_cancel(skb, linkinfo);
638 out:
639 	return err;
640 }
641 
642 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
643 {
644 	struct sock *rtnl = net->rtnl;
645 	int err = 0;
646 
647 	NETLINK_CB(skb).dst_group = group;
648 	if (echo)
649 		atomic_inc(&skb->users);
650 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
651 	if (echo)
652 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
653 	return err;
654 }
655 
656 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
657 {
658 	struct sock *rtnl = net->rtnl;
659 
660 	return nlmsg_unicast(rtnl, skb, pid);
661 }
662 EXPORT_SYMBOL(rtnl_unicast);
663 
664 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
665 		 struct nlmsghdr *nlh, gfp_t flags)
666 {
667 	struct sock *rtnl = net->rtnl;
668 	int report = 0;
669 
670 	if (nlh)
671 		report = nlmsg_report(nlh);
672 
673 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
674 }
675 EXPORT_SYMBOL(rtnl_notify);
676 
677 void rtnl_set_sk_err(struct net *net, u32 group, int error)
678 {
679 	struct sock *rtnl = net->rtnl;
680 
681 	netlink_set_err(rtnl, 0, group, error);
682 }
683 EXPORT_SYMBOL(rtnl_set_sk_err);
684 
685 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
686 {
687 	struct nlattr *mx;
688 	int i, valid = 0;
689 
690 	mx = nla_nest_start(skb, RTA_METRICS);
691 	if (mx == NULL)
692 		return -ENOBUFS;
693 
694 	for (i = 0; i < RTAX_MAX; i++) {
695 		if (metrics[i]) {
696 			if (i == RTAX_CC_ALGO - 1) {
697 				char tmp[TCP_CA_NAME_MAX], *name;
698 
699 				name = tcp_ca_get_name_by_key(metrics[i], tmp);
700 				if (!name)
701 					continue;
702 				if (nla_put_string(skb, i + 1, name))
703 					goto nla_put_failure;
704 			} else if (i == RTAX_FEATURES - 1) {
705 				u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
706 
707 				BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
708 				if (nla_put_u32(skb, i + 1, user_features))
709 					goto nla_put_failure;
710 			} else {
711 				if (nla_put_u32(skb, i + 1, metrics[i]))
712 					goto nla_put_failure;
713 			}
714 			valid++;
715 		}
716 	}
717 
718 	if (!valid) {
719 		nla_nest_cancel(skb, mx);
720 		return 0;
721 	}
722 
723 	return nla_nest_end(skb, mx);
724 
725 nla_put_failure:
726 	nla_nest_cancel(skb, mx);
727 	return -EMSGSIZE;
728 }
729 EXPORT_SYMBOL(rtnetlink_put_metrics);
730 
731 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
732 		       long expires, u32 error)
733 {
734 	struct rta_cacheinfo ci = {
735 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
736 		.rta_used = dst->__use,
737 		.rta_clntref = atomic_read(&(dst->__refcnt)),
738 		.rta_error = error,
739 		.rta_id =  id,
740 	};
741 
742 	if (expires) {
743 		unsigned long clock;
744 
745 		clock = jiffies_to_clock_t(abs(expires));
746 		clock = min_t(unsigned long, clock, INT_MAX);
747 		ci.rta_expires = (expires > 0) ? clock : -clock;
748 	}
749 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
750 }
751 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
752 
753 static void set_operstate(struct net_device *dev, unsigned char transition)
754 {
755 	unsigned char operstate = dev->operstate;
756 
757 	switch (transition) {
758 	case IF_OPER_UP:
759 		if ((operstate == IF_OPER_DORMANT ||
760 		     operstate == IF_OPER_UNKNOWN) &&
761 		    !netif_dormant(dev))
762 			operstate = IF_OPER_UP;
763 		break;
764 
765 	case IF_OPER_DORMANT:
766 		if (operstate == IF_OPER_UP ||
767 		    operstate == IF_OPER_UNKNOWN)
768 			operstate = IF_OPER_DORMANT;
769 		break;
770 	}
771 
772 	if (dev->operstate != operstate) {
773 		write_lock_bh(&dev_base_lock);
774 		dev->operstate = operstate;
775 		write_unlock_bh(&dev_base_lock);
776 		netdev_state_change(dev);
777 	}
778 }
779 
780 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
781 {
782 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
783 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
784 }
785 
786 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
787 					   const struct ifinfomsg *ifm)
788 {
789 	unsigned int flags = ifm->ifi_flags;
790 
791 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
792 	if (ifm->ifi_change)
793 		flags = (flags & ifm->ifi_change) |
794 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
795 
796 	return flags;
797 }
798 
799 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
800 				 const struct rtnl_link_stats64 *b)
801 {
802 	a->rx_packets = b->rx_packets;
803 	a->tx_packets = b->tx_packets;
804 	a->rx_bytes = b->rx_bytes;
805 	a->tx_bytes = b->tx_bytes;
806 	a->rx_errors = b->rx_errors;
807 	a->tx_errors = b->tx_errors;
808 	a->rx_dropped = b->rx_dropped;
809 	a->tx_dropped = b->tx_dropped;
810 
811 	a->multicast = b->multicast;
812 	a->collisions = b->collisions;
813 
814 	a->rx_length_errors = b->rx_length_errors;
815 	a->rx_over_errors = b->rx_over_errors;
816 	a->rx_crc_errors = b->rx_crc_errors;
817 	a->rx_frame_errors = b->rx_frame_errors;
818 	a->rx_fifo_errors = b->rx_fifo_errors;
819 	a->rx_missed_errors = b->rx_missed_errors;
820 
821 	a->tx_aborted_errors = b->tx_aborted_errors;
822 	a->tx_carrier_errors = b->tx_carrier_errors;
823 	a->tx_fifo_errors = b->tx_fifo_errors;
824 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
825 	a->tx_window_errors = b->tx_window_errors;
826 
827 	a->rx_compressed = b->rx_compressed;
828 	a->tx_compressed = b->tx_compressed;
829 
830 	a->rx_nohandler = b->rx_nohandler;
831 }
832 
833 /* All VF info */
834 static inline int rtnl_vfinfo_size(const struct net_device *dev,
835 				   u32 ext_filter_mask)
836 {
837 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
838 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
839 		int num_vfs = dev_num_vf(dev->dev.parent);
840 		size_t size = nla_total_size(sizeof(struct nlattr));
841 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
842 		size += num_vfs *
843 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
844 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
845 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
846 			 nla_total_size(sizeof(struct ifla_vf_rate)) +
847 			 nla_total_size(sizeof(struct ifla_vf_link_state)) +
848 			 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
849 			 /* IFLA_VF_STATS_RX_PACKETS */
850 			 nla_total_size_64bit(sizeof(__u64)) +
851 			 /* IFLA_VF_STATS_TX_PACKETS */
852 			 nla_total_size_64bit(sizeof(__u64)) +
853 			 /* IFLA_VF_STATS_RX_BYTES */
854 			 nla_total_size_64bit(sizeof(__u64)) +
855 			 /* IFLA_VF_STATS_TX_BYTES */
856 			 nla_total_size_64bit(sizeof(__u64)) +
857 			 /* IFLA_VF_STATS_BROADCAST */
858 			 nla_total_size_64bit(sizeof(__u64)) +
859 			 /* IFLA_VF_STATS_MULTICAST */
860 			 nla_total_size_64bit(sizeof(__u64)) +
861 			 nla_total_size(sizeof(struct ifla_vf_trust)));
862 		return size;
863 	} else
864 		return 0;
865 }
866 
867 static size_t rtnl_port_size(const struct net_device *dev,
868 			     u32 ext_filter_mask)
869 {
870 	size_t port_size = nla_total_size(4)		/* PORT_VF */
871 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
872 		+ nla_total_size(sizeof(struct ifla_port_vsi))
873 							/* PORT_VSI_TYPE */
874 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
875 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
876 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
877 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
878 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
879 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
880 		+ port_size;
881 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
882 		+ port_size;
883 
884 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
885 	    !(ext_filter_mask & RTEXT_FILTER_VF))
886 		return 0;
887 	if (dev_num_vf(dev->dev.parent))
888 		return port_self_size + vf_ports_size +
889 			vf_port_size * dev_num_vf(dev->dev.parent);
890 	else
891 		return port_self_size;
892 }
893 
894 static noinline size_t if_nlmsg_size(const struct net_device *dev,
895 				     u32 ext_filter_mask)
896 {
897 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
898 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
899 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
900 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
901 	       + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
902 	       + nla_total_size(sizeof(struct rtnl_link_stats))
903 	       + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
904 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
905 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
906 	       + nla_total_size(4) /* IFLA_TXQLEN */
907 	       + nla_total_size(4) /* IFLA_WEIGHT */
908 	       + nla_total_size(4) /* IFLA_MTU */
909 	       + nla_total_size(4) /* IFLA_LINK */
910 	       + nla_total_size(4) /* IFLA_MASTER */
911 	       + nla_total_size(1) /* IFLA_CARRIER */
912 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
913 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
914 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
915 	       + nla_total_size(4) /* IFLA_MAX_GSO_SEGS */
916 	       + nla_total_size(4) /* IFLA_MAX_GSO_SIZE */
917 	       + nla_total_size(1) /* IFLA_OPERSTATE */
918 	       + nla_total_size(1) /* IFLA_LINKMODE */
919 	       + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
920 	       + nla_total_size(4) /* IFLA_LINK_NETNSID */
921 	       + nla_total_size(ext_filter_mask
922 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
923 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
924 	       + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
925 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
926 	       + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
927 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
928 	       + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
929 	       + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
930 	       + nla_total_size(1); /* IFLA_PROTO_DOWN */
931 
932 }
933 
934 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
935 {
936 	struct nlattr *vf_ports;
937 	struct nlattr *vf_port;
938 	int vf;
939 	int err;
940 
941 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
942 	if (!vf_ports)
943 		return -EMSGSIZE;
944 
945 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
946 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
947 		if (!vf_port)
948 			goto nla_put_failure;
949 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
950 			goto nla_put_failure;
951 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
952 		if (err == -EMSGSIZE)
953 			goto nla_put_failure;
954 		if (err) {
955 			nla_nest_cancel(skb, vf_port);
956 			continue;
957 		}
958 		nla_nest_end(skb, vf_port);
959 	}
960 
961 	nla_nest_end(skb, vf_ports);
962 
963 	return 0;
964 
965 nla_put_failure:
966 	nla_nest_cancel(skb, vf_ports);
967 	return -EMSGSIZE;
968 }
969 
970 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
971 {
972 	struct nlattr *port_self;
973 	int err;
974 
975 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
976 	if (!port_self)
977 		return -EMSGSIZE;
978 
979 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
980 	if (err) {
981 		nla_nest_cancel(skb, port_self);
982 		return (err == -EMSGSIZE) ? err : 0;
983 	}
984 
985 	nla_nest_end(skb, port_self);
986 
987 	return 0;
988 }
989 
990 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
991 			  u32 ext_filter_mask)
992 {
993 	int err;
994 
995 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
996 	    !(ext_filter_mask & RTEXT_FILTER_VF))
997 		return 0;
998 
999 	err = rtnl_port_self_fill(skb, dev);
1000 	if (err)
1001 		return err;
1002 
1003 	if (dev_num_vf(dev->dev.parent)) {
1004 		err = rtnl_vf_ports_fill(skb, dev);
1005 		if (err)
1006 			return err;
1007 	}
1008 
1009 	return 0;
1010 }
1011 
1012 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1013 {
1014 	int err;
1015 	struct netdev_phys_item_id ppid;
1016 
1017 	err = dev_get_phys_port_id(dev, &ppid);
1018 	if (err) {
1019 		if (err == -EOPNOTSUPP)
1020 			return 0;
1021 		return err;
1022 	}
1023 
1024 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1025 		return -EMSGSIZE;
1026 
1027 	return 0;
1028 }
1029 
1030 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1031 {
1032 	char name[IFNAMSIZ];
1033 	int err;
1034 
1035 	err = dev_get_phys_port_name(dev, name, sizeof(name));
1036 	if (err) {
1037 		if (err == -EOPNOTSUPP)
1038 			return 0;
1039 		return err;
1040 	}
1041 
1042 	if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1043 		return -EMSGSIZE;
1044 
1045 	return 0;
1046 }
1047 
1048 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1049 {
1050 	int err;
1051 	struct switchdev_attr attr = {
1052 		.orig_dev = dev,
1053 		.id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1054 		.flags = SWITCHDEV_F_NO_RECURSE,
1055 	};
1056 
1057 	err = switchdev_port_attr_get(dev, &attr);
1058 	if (err) {
1059 		if (err == -EOPNOTSUPP)
1060 			return 0;
1061 		return err;
1062 	}
1063 
1064 	if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1065 		    attr.u.ppid.id))
1066 		return -EMSGSIZE;
1067 
1068 	return 0;
1069 }
1070 
1071 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1072 					      struct net_device *dev)
1073 {
1074 	struct rtnl_link_stats64 *sp;
1075 	struct nlattr *attr;
1076 
1077 	attr = nla_reserve_64bit(skb, IFLA_STATS64,
1078 				 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1079 	if (!attr)
1080 		return -EMSGSIZE;
1081 
1082 	sp = nla_data(attr);
1083 	dev_get_stats(dev, sp);
1084 
1085 	attr = nla_reserve(skb, IFLA_STATS,
1086 			   sizeof(struct rtnl_link_stats));
1087 	if (!attr)
1088 		return -EMSGSIZE;
1089 
1090 	copy_rtnl_link_stats(nla_data(attr), sp);
1091 
1092 	return 0;
1093 }
1094 
1095 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1096 					       struct net_device *dev,
1097 					       int vfs_num,
1098 					       struct nlattr *vfinfo)
1099 {
1100 	struct ifla_vf_rss_query_en vf_rss_query_en;
1101 	struct ifla_vf_link_state vf_linkstate;
1102 	struct ifla_vf_spoofchk vf_spoofchk;
1103 	struct ifla_vf_tx_rate vf_tx_rate;
1104 	struct ifla_vf_stats vf_stats;
1105 	struct ifla_vf_trust vf_trust;
1106 	struct ifla_vf_vlan vf_vlan;
1107 	struct ifla_vf_rate vf_rate;
1108 	struct nlattr *vf, *vfstats;
1109 	struct ifla_vf_mac vf_mac;
1110 	struct ifla_vf_info ivi;
1111 
1112 	/* Not all SR-IOV capable drivers support the
1113 	 * spoofcheck and "RSS query enable" query.  Preset to
1114 	 * -1 so the user space tool can detect that the driver
1115 	 * didn't report anything.
1116 	 */
1117 	ivi.spoofchk = -1;
1118 	ivi.rss_query_en = -1;
1119 	ivi.trusted = -1;
1120 	memset(ivi.mac, 0, sizeof(ivi.mac));
1121 	/* The default value for VF link state is "auto"
1122 	 * IFLA_VF_LINK_STATE_AUTO which equals zero
1123 	 */
1124 	ivi.linkstate = 0;
1125 	if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1126 		return 0;
1127 
1128 	vf_mac.vf =
1129 		vf_vlan.vf =
1130 		vf_rate.vf =
1131 		vf_tx_rate.vf =
1132 		vf_spoofchk.vf =
1133 		vf_linkstate.vf =
1134 		vf_rss_query_en.vf =
1135 		vf_trust.vf = ivi.vf;
1136 
1137 	memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1138 	vf_vlan.vlan = ivi.vlan;
1139 	vf_vlan.qos = ivi.qos;
1140 	vf_tx_rate.rate = ivi.max_tx_rate;
1141 	vf_rate.min_tx_rate = ivi.min_tx_rate;
1142 	vf_rate.max_tx_rate = ivi.max_tx_rate;
1143 	vf_spoofchk.setting = ivi.spoofchk;
1144 	vf_linkstate.link_state = ivi.linkstate;
1145 	vf_rss_query_en.setting = ivi.rss_query_en;
1146 	vf_trust.setting = ivi.trusted;
1147 	vf = nla_nest_start(skb, IFLA_VF_INFO);
1148 	if (!vf) {
1149 		nla_nest_cancel(skb, vfinfo);
1150 		return -EMSGSIZE;
1151 	}
1152 	if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1153 	    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1154 	    nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1155 		    &vf_rate) ||
1156 	    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1157 		    &vf_tx_rate) ||
1158 	    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1159 		    &vf_spoofchk) ||
1160 	    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1161 		    &vf_linkstate) ||
1162 	    nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1163 		    sizeof(vf_rss_query_en),
1164 		    &vf_rss_query_en) ||
1165 	    nla_put(skb, IFLA_VF_TRUST,
1166 		    sizeof(vf_trust), &vf_trust))
1167 		return -EMSGSIZE;
1168 	memset(&vf_stats, 0, sizeof(vf_stats));
1169 	if (dev->netdev_ops->ndo_get_vf_stats)
1170 		dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1171 						&vf_stats);
1172 	vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1173 	if (!vfstats) {
1174 		nla_nest_cancel(skb, vf);
1175 		nla_nest_cancel(skb, vfinfo);
1176 		return -EMSGSIZE;
1177 	}
1178 	if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1179 			      vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1180 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1181 			      vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1182 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1183 			      vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1184 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1185 			      vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1186 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1187 			      vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1188 	    nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1189 			      vf_stats.multicast, IFLA_VF_STATS_PAD))
1190 		return -EMSGSIZE;
1191 	nla_nest_end(skb, vfstats);
1192 	nla_nest_end(skb, vf);
1193 	return 0;
1194 }
1195 
1196 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1197 {
1198 	struct rtnl_link_ifmap map;
1199 
1200 	memset(&map, 0, sizeof(map));
1201 	map.mem_start   = dev->mem_start;
1202 	map.mem_end     = dev->mem_end;
1203 	map.base_addr   = dev->base_addr;
1204 	map.irq         = dev->irq;
1205 	map.dma         = dev->dma;
1206 	map.port        = dev->if_port;
1207 
1208 	if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1209 		return -EMSGSIZE;
1210 
1211 	return 0;
1212 }
1213 
1214 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1215 			    int type, u32 pid, u32 seq, u32 change,
1216 			    unsigned int flags, u32 ext_filter_mask)
1217 {
1218 	struct ifinfomsg *ifm;
1219 	struct nlmsghdr *nlh;
1220 	struct nlattr *af_spec;
1221 	struct rtnl_af_ops *af_ops;
1222 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1223 
1224 	ASSERT_RTNL();
1225 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1226 	if (nlh == NULL)
1227 		return -EMSGSIZE;
1228 
1229 	ifm = nlmsg_data(nlh);
1230 	ifm->ifi_family = AF_UNSPEC;
1231 	ifm->__ifi_pad = 0;
1232 	ifm->ifi_type = dev->type;
1233 	ifm->ifi_index = dev->ifindex;
1234 	ifm->ifi_flags = dev_get_flags(dev);
1235 	ifm->ifi_change = change;
1236 
1237 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1238 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1239 	    nla_put_u8(skb, IFLA_OPERSTATE,
1240 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1241 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1242 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1243 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1244 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1245 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1246 	    nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1247 	    nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1248 #ifdef CONFIG_RPS
1249 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1250 #endif
1251 	    (dev->ifindex != dev_get_iflink(dev) &&
1252 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1253 	    (upper_dev &&
1254 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1255 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1256 	    (dev->qdisc &&
1257 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1258 	    (dev->ifalias &&
1259 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1260 	    nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1261 			atomic_read(&dev->carrier_changes)) ||
1262 	    nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1263 		goto nla_put_failure;
1264 
1265 	if (rtnl_fill_link_ifmap(skb, dev))
1266 		goto nla_put_failure;
1267 
1268 	if (dev->addr_len) {
1269 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1270 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1271 			goto nla_put_failure;
1272 	}
1273 
1274 	if (rtnl_phys_port_id_fill(skb, dev))
1275 		goto nla_put_failure;
1276 
1277 	if (rtnl_phys_port_name_fill(skb, dev))
1278 		goto nla_put_failure;
1279 
1280 	if (rtnl_phys_switch_id_fill(skb, dev))
1281 		goto nla_put_failure;
1282 
1283 	if (rtnl_fill_stats(skb, dev))
1284 		goto nla_put_failure;
1285 
1286 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1287 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1288 		goto nla_put_failure;
1289 
1290 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1291 	    ext_filter_mask & RTEXT_FILTER_VF) {
1292 		int i;
1293 		struct nlattr *vfinfo;
1294 		int num_vfs = dev_num_vf(dev->dev.parent);
1295 
1296 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1297 		if (!vfinfo)
1298 			goto nla_put_failure;
1299 		for (i = 0; i < num_vfs; i++) {
1300 			if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1301 				goto nla_put_failure;
1302 		}
1303 
1304 		nla_nest_end(skb, vfinfo);
1305 	}
1306 
1307 	if (rtnl_port_fill(skb, dev, ext_filter_mask))
1308 		goto nla_put_failure;
1309 
1310 	if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1311 		if (rtnl_link_fill(skb, dev) < 0)
1312 			goto nla_put_failure;
1313 	}
1314 
1315 	if (dev->rtnl_link_ops &&
1316 	    dev->rtnl_link_ops->get_link_net) {
1317 		struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1318 
1319 		if (!net_eq(dev_net(dev), link_net)) {
1320 			int id = peernet2id_alloc(dev_net(dev), link_net);
1321 
1322 			if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1323 				goto nla_put_failure;
1324 		}
1325 	}
1326 
1327 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1328 		goto nla_put_failure;
1329 
1330 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1331 		if (af_ops->fill_link_af) {
1332 			struct nlattr *af;
1333 			int err;
1334 
1335 			if (!(af = nla_nest_start(skb, af_ops->family)))
1336 				goto nla_put_failure;
1337 
1338 			err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1339 
1340 			/*
1341 			 * Caller may return ENODATA to indicate that there
1342 			 * was no data to be dumped. This is not an error, it
1343 			 * means we should trim the attribute header and
1344 			 * continue.
1345 			 */
1346 			if (err == -ENODATA)
1347 				nla_nest_cancel(skb, af);
1348 			else if (err < 0)
1349 				goto nla_put_failure;
1350 
1351 			nla_nest_end(skb, af);
1352 		}
1353 	}
1354 
1355 	nla_nest_end(skb, af_spec);
1356 
1357 	nlmsg_end(skb, nlh);
1358 	return 0;
1359 
1360 nla_put_failure:
1361 	nlmsg_cancel(skb, nlh);
1362 	return -EMSGSIZE;
1363 }
1364 
1365 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1366 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1367 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1368 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1369 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1370 	[IFLA_MTU]		= { .type = NLA_U32 },
1371 	[IFLA_LINK]		= { .type = NLA_U32 },
1372 	[IFLA_MASTER]		= { .type = NLA_U32 },
1373 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1374 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1375 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1376 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1377 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1378 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1379 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1380 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1381 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1382 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1383 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1384 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1385 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1386 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1387 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1388 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1389 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1390 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1391 	[IFLA_CARRIER_CHANGES]	= { .type = NLA_U32 },  /* ignored */
1392 	[IFLA_PHYS_SWITCH_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1393 	[IFLA_LINK_NETNSID]	= { .type = NLA_S32 },
1394 	[IFLA_PROTO_DOWN]	= { .type = NLA_U8 },
1395 };
1396 
1397 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1398 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1399 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1400 	[IFLA_INFO_SLAVE_KIND]	= { .type = NLA_STRING },
1401 	[IFLA_INFO_SLAVE_DATA]	= { .type = NLA_NESTED },
1402 };
1403 
1404 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1405 	[IFLA_VF_MAC]		= { .len = sizeof(struct ifla_vf_mac) },
1406 	[IFLA_VF_VLAN]		= { .len = sizeof(struct ifla_vf_vlan) },
1407 	[IFLA_VF_TX_RATE]	= { .len = sizeof(struct ifla_vf_tx_rate) },
1408 	[IFLA_VF_SPOOFCHK]	= { .len = sizeof(struct ifla_vf_spoofchk) },
1409 	[IFLA_VF_RATE]		= { .len = sizeof(struct ifla_vf_rate) },
1410 	[IFLA_VF_LINK_STATE]	= { .len = sizeof(struct ifla_vf_link_state) },
1411 	[IFLA_VF_RSS_QUERY_EN]	= { .len = sizeof(struct ifla_vf_rss_query_en) },
1412 	[IFLA_VF_STATS]		= { .type = NLA_NESTED },
1413 	[IFLA_VF_TRUST]		= { .len = sizeof(struct ifla_vf_trust) },
1414 	[IFLA_VF_IB_NODE_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1415 	[IFLA_VF_IB_PORT_GUID]	= { .len = sizeof(struct ifla_vf_guid) },
1416 };
1417 
1418 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1419 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1420 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1421 				    .len = PORT_PROFILE_MAX },
1422 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1423 				    .len = sizeof(struct ifla_port_vsi)},
1424 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1425 				      .len = PORT_UUID_MAX },
1426 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1427 				    .len = PORT_UUID_MAX },
1428 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1429 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1430 };
1431 
1432 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1433 {
1434 	const struct rtnl_link_ops *ops = NULL;
1435 	struct nlattr *linfo[IFLA_INFO_MAX + 1];
1436 
1437 	if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1438 		return NULL;
1439 
1440 	if (linfo[IFLA_INFO_KIND]) {
1441 		char kind[MODULE_NAME_LEN];
1442 
1443 		nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1444 		ops = rtnl_link_ops_get(kind);
1445 	}
1446 
1447 	return ops;
1448 }
1449 
1450 static bool link_master_filtered(struct net_device *dev, int master_idx)
1451 {
1452 	struct net_device *master;
1453 
1454 	if (!master_idx)
1455 		return false;
1456 
1457 	master = netdev_master_upper_dev_get(dev);
1458 	if (!master || master->ifindex != master_idx)
1459 		return true;
1460 
1461 	return false;
1462 }
1463 
1464 static bool link_kind_filtered(const struct net_device *dev,
1465 			       const struct rtnl_link_ops *kind_ops)
1466 {
1467 	if (kind_ops && dev->rtnl_link_ops != kind_ops)
1468 		return true;
1469 
1470 	return false;
1471 }
1472 
1473 static bool link_dump_filtered(struct net_device *dev,
1474 			       int master_idx,
1475 			       const struct rtnl_link_ops *kind_ops)
1476 {
1477 	if (link_master_filtered(dev, master_idx) ||
1478 	    link_kind_filtered(dev, kind_ops))
1479 		return true;
1480 
1481 	return false;
1482 }
1483 
1484 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1485 {
1486 	struct net *net = sock_net(skb->sk);
1487 	int h, s_h;
1488 	int idx = 0, s_idx;
1489 	struct net_device *dev;
1490 	struct hlist_head *head;
1491 	struct nlattr *tb[IFLA_MAX+1];
1492 	u32 ext_filter_mask = 0;
1493 	const struct rtnl_link_ops *kind_ops = NULL;
1494 	unsigned int flags = NLM_F_MULTI;
1495 	int master_idx = 0;
1496 	int err;
1497 	int hdrlen;
1498 
1499 	s_h = cb->args[0];
1500 	s_idx = cb->args[1];
1501 
1502 	cb->seq = net->dev_base_seq;
1503 
1504 	/* A hack to preserve kernel<->userspace interface.
1505 	 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1506 	 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1507 	 * what iproute2 < v3.9.0 used.
1508 	 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1509 	 * attribute, its netlink message is shorter than struct ifinfomsg.
1510 	 */
1511 	hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1512 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1513 
1514 	if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1515 
1516 		if (tb[IFLA_EXT_MASK])
1517 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1518 
1519 		if (tb[IFLA_MASTER])
1520 			master_idx = nla_get_u32(tb[IFLA_MASTER]);
1521 
1522 		if (tb[IFLA_LINKINFO])
1523 			kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1524 
1525 		if (master_idx || kind_ops)
1526 			flags |= NLM_F_DUMP_FILTERED;
1527 	}
1528 
1529 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1530 		idx = 0;
1531 		head = &net->dev_index_head[h];
1532 		hlist_for_each_entry(dev, head, index_hlist) {
1533 			if (link_dump_filtered(dev, master_idx, kind_ops))
1534 				continue;
1535 			if (idx < s_idx)
1536 				goto cont;
1537 			err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1538 					       NETLINK_CB(cb->skb).portid,
1539 					       cb->nlh->nlmsg_seq, 0,
1540 					       flags,
1541 					       ext_filter_mask);
1542 			/* If we ran out of room on the first message,
1543 			 * we're in trouble
1544 			 */
1545 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1546 
1547 			if (err < 0)
1548 				goto out;
1549 
1550 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1551 cont:
1552 			idx++;
1553 		}
1554 	}
1555 out:
1556 	cb->args[1] = idx;
1557 	cb->args[0] = h;
1558 
1559 	return skb->len;
1560 }
1561 
1562 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1563 {
1564 	return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1565 }
1566 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1567 
1568 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1569 {
1570 	struct net *net;
1571 	/* Examine the link attributes and figure out which
1572 	 * network namespace we are talking about.
1573 	 */
1574 	if (tb[IFLA_NET_NS_PID])
1575 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1576 	else if (tb[IFLA_NET_NS_FD])
1577 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1578 	else
1579 		net = get_net(src_net);
1580 	return net;
1581 }
1582 EXPORT_SYMBOL(rtnl_link_get_net);
1583 
1584 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1585 {
1586 	if (dev) {
1587 		if (tb[IFLA_ADDRESS] &&
1588 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1589 			return -EINVAL;
1590 
1591 		if (tb[IFLA_BROADCAST] &&
1592 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1593 			return -EINVAL;
1594 	}
1595 
1596 	if (tb[IFLA_AF_SPEC]) {
1597 		struct nlattr *af;
1598 		int rem, err;
1599 
1600 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1601 			const struct rtnl_af_ops *af_ops;
1602 
1603 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1604 				return -EAFNOSUPPORT;
1605 
1606 			if (!af_ops->set_link_af)
1607 				return -EOPNOTSUPP;
1608 
1609 			if (af_ops->validate_link_af) {
1610 				err = af_ops->validate_link_af(dev, af);
1611 				if (err < 0)
1612 					return err;
1613 			}
1614 		}
1615 	}
1616 
1617 	return 0;
1618 }
1619 
1620 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1621 				  int guid_type)
1622 {
1623 	const struct net_device_ops *ops = dev->netdev_ops;
1624 
1625 	return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1626 }
1627 
1628 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1629 {
1630 	if (dev->type != ARPHRD_INFINIBAND)
1631 		return -EOPNOTSUPP;
1632 
1633 	return handle_infiniband_guid(dev, ivt, guid_type);
1634 }
1635 
1636 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1637 {
1638 	const struct net_device_ops *ops = dev->netdev_ops;
1639 	int err = -EINVAL;
1640 
1641 	if (tb[IFLA_VF_MAC]) {
1642 		struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1643 
1644 		err = -EOPNOTSUPP;
1645 		if (ops->ndo_set_vf_mac)
1646 			err = ops->ndo_set_vf_mac(dev, ivm->vf,
1647 						  ivm->mac);
1648 		if (err < 0)
1649 			return err;
1650 	}
1651 
1652 	if (tb[IFLA_VF_VLAN]) {
1653 		struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1654 
1655 		err = -EOPNOTSUPP;
1656 		if (ops->ndo_set_vf_vlan)
1657 			err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1658 						   ivv->qos);
1659 		if (err < 0)
1660 			return err;
1661 	}
1662 
1663 	if (tb[IFLA_VF_TX_RATE]) {
1664 		struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1665 		struct ifla_vf_info ivf;
1666 
1667 		err = -EOPNOTSUPP;
1668 		if (ops->ndo_get_vf_config)
1669 			err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1670 		if (err < 0)
1671 			return err;
1672 
1673 		err = -EOPNOTSUPP;
1674 		if (ops->ndo_set_vf_rate)
1675 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1676 						   ivf.min_tx_rate,
1677 						   ivt->rate);
1678 		if (err < 0)
1679 			return err;
1680 	}
1681 
1682 	if (tb[IFLA_VF_RATE]) {
1683 		struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1684 
1685 		err = -EOPNOTSUPP;
1686 		if (ops->ndo_set_vf_rate)
1687 			err = ops->ndo_set_vf_rate(dev, ivt->vf,
1688 						   ivt->min_tx_rate,
1689 						   ivt->max_tx_rate);
1690 		if (err < 0)
1691 			return err;
1692 	}
1693 
1694 	if (tb[IFLA_VF_SPOOFCHK]) {
1695 		struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1696 
1697 		err = -EOPNOTSUPP;
1698 		if (ops->ndo_set_vf_spoofchk)
1699 			err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1700 						       ivs->setting);
1701 		if (err < 0)
1702 			return err;
1703 	}
1704 
1705 	if (tb[IFLA_VF_LINK_STATE]) {
1706 		struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1707 
1708 		err = -EOPNOTSUPP;
1709 		if (ops->ndo_set_vf_link_state)
1710 			err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1711 							 ivl->link_state);
1712 		if (err < 0)
1713 			return err;
1714 	}
1715 
1716 	if (tb[IFLA_VF_RSS_QUERY_EN]) {
1717 		struct ifla_vf_rss_query_en *ivrssq_en;
1718 
1719 		err = -EOPNOTSUPP;
1720 		ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1721 		if (ops->ndo_set_vf_rss_query_en)
1722 			err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1723 							   ivrssq_en->setting);
1724 		if (err < 0)
1725 			return err;
1726 	}
1727 
1728 	if (tb[IFLA_VF_TRUST]) {
1729 		struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1730 
1731 		err = -EOPNOTSUPP;
1732 		if (ops->ndo_set_vf_trust)
1733 			err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1734 		if (err < 0)
1735 			return err;
1736 	}
1737 
1738 	if (tb[IFLA_VF_IB_NODE_GUID]) {
1739 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1740 
1741 		if (!ops->ndo_set_vf_guid)
1742 			return -EOPNOTSUPP;
1743 
1744 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1745 	}
1746 
1747 	if (tb[IFLA_VF_IB_PORT_GUID]) {
1748 		struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1749 
1750 		if (!ops->ndo_set_vf_guid)
1751 			return -EOPNOTSUPP;
1752 
1753 		return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1754 	}
1755 
1756 	return err;
1757 }
1758 
1759 static int do_set_master(struct net_device *dev, int ifindex)
1760 {
1761 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1762 	const struct net_device_ops *ops;
1763 	int err;
1764 
1765 	if (upper_dev) {
1766 		if (upper_dev->ifindex == ifindex)
1767 			return 0;
1768 		ops = upper_dev->netdev_ops;
1769 		if (ops->ndo_del_slave) {
1770 			err = ops->ndo_del_slave(upper_dev, dev);
1771 			if (err)
1772 				return err;
1773 		} else {
1774 			return -EOPNOTSUPP;
1775 		}
1776 	}
1777 
1778 	if (ifindex) {
1779 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1780 		if (!upper_dev)
1781 			return -EINVAL;
1782 		ops = upper_dev->netdev_ops;
1783 		if (ops->ndo_add_slave) {
1784 			err = ops->ndo_add_slave(upper_dev, dev);
1785 			if (err)
1786 				return err;
1787 		} else {
1788 			return -EOPNOTSUPP;
1789 		}
1790 	}
1791 	return 0;
1792 }
1793 
1794 #define DO_SETLINK_MODIFIED	0x01
1795 /* notify flag means notify + modified. */
1796 #define DO_SETLINK_NOTIFY	0x03
1797 static int do_setlink(const struct sk_buff *skb,
1798 		      struct net_device *dev, struct ifinfomsg *ifm,
1799 		      struct nlattr **tb, char *ifname, int status)
1800 {
1801 	const struct net_device_ops *ops = dev->netdev_ops;
1802 	int err;
1803 
1804 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1805 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1806 		if (IS_ERR(net)) {
1807 			err = PTR_ERR(net);
1808 			goto errout;
1809 		}
1810 		if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1811 			put_net(net);
1812 			err = -EPERM;
1813 			goto errout;
1814 		}
1815 		err = dev_change_net_namespace(dev, net, ifname);
1816 		put_net(net);
1817 		if (err)
1818 			goto errout;
1819 		status |= DO_SETLINK_MODIFIED;
1820 	}
1821 
1822 	if (tb[IFLA_MAP]) {
1823 		struct rtnl_link_ifmap *u_map;
1824 		struct ifmap k_map;
1825 
1826 		if (!ops->ndo_set_config) {
1827 			err = -EOPNOTSUPP;
1828 			goto errout;
1829 		}
1830 
1831 		if (!netif_device_present(dev)) {
1832 			err = -ENODEV;
1833 			goto errout;
1834 		}
1835 
1836 		u_map = nla_data(tb[IFLA_MAP]);
1837 		k_map.mem_start = (unsigned long) u_map->mem_start;
1838 		k_map.mem_end = (unsigned long) u_map->mem_end;
1839 		k_map.base_addr = (unsigned short) u_map->base_addr;
1840 		k_map.irq = (unsigned char) u_map->irq;
1841 		k_map.dma = (unsigned char) u_map->dma;
1842 		k_map.port = (unsigned char) u_map->port;
1843 
1844 		err = ops->ndo_set_config(dev, &k_map);
1845 		if (err < 0)
1846 			goto errout;
1847 
1848 		status |= DO_SETLINK_NOTIFY;
1849 	}
1850 
1851 	if (tb[IFLA_ADDRESS]) {
1852 		struct sockaddr *sa;
1853 		int len;
1854 
1855 		len = sizeof(sa_family_t) + dev->addr_len;
1856 		sa = kmalloc(len, GFP_KERNEL);
1857 		if (!sa) {
1858 			err = -ENOMEM;
1859 			goto errout;
1860 		}
1861 		sa->sa_family = dev->type;
1862 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1863 		       dev->addr_len);
1864 		err = dev_set_mac_address(dev, sa);
1865 		kfree(sa);
1866 		if (err)
1867 			goto errout;
1868 		status |= DO_SETLINK_MODIFIED;
1869 	}
1870 
1871 	if (tb[IFLA_MTU]) {
1872 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1873 		if (err < 0)
1874 			goto errout;
1875 		status |= DO_SETLINK_MODIFIED;
1876 	}
1877 
1878 	if (tb[IFLA_GROUP]) {
1879 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1880 		status |= DO_SETLINK_NOTIFY;
1881 	}
1882 
1883 	/*
1884 	 * Interface selected by interface index but interface
1885 	 * name provided implies that a name change has been
1886 	 * requested.
1887 	 */
1888 	if (ifm->ifi_index > 0 && ifname[0]) {
1889 		err = dev_change_name(dev, ifname);
1890 		if (err < 0)
1891 			goto errout;
1892 		status |= DO_SETLINK_MODIFIED;
1893 	}
1894 
1895 	if (tb[IFLA_IFALIAS]) {
1896 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1897 				    nla_len(tb[IFLA_IFALIAS]));
1898 		if (err < 0)
1899 			goto errout;
1900 		status |= DO_SETLINK_NOTIFY;
1901 	}
1902 
1903 	if (tb[IFLA_BROADCAST]) {
1904 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1905 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1906 	}
1907 
1908 	if (ifm->ifi_flags || ifm->ifi_change) {
1909 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1910 		if (err < 0)
1911 			goto errout;
1912 	}
1913 
1914 	if (tb[IFLA_MASTER]) {
1915 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1916 		if (err)
1917 			goto errout;
1918 		status |= DO_SETLINK_MODIFIED;
1919 	}
1920 
1921 	if (tb[IFLA_CARRIER]) {
1922 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1923 		if (err)
1924 			goto errout;
1925 		status |= DO_SETLINK_MODIFIED;
1926 	}
1927 
1928 	if (tb[IFLA_TXQLEN]) {
1929 		unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1930 
1931 		if (dev->tx_queue_len ^ value)
1932 			status |= DO_SETLINK_NOTIFY;
1933 
1934 		dev->tx_queue_len = value;
1935 	}
1936 
1937 	if (tb[IFLA_OPERSTATE])
1938 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1939 
1940 	if (tb[IFLA_LINKMODE]) {
1941 		unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1942 
1943 		write_lock_bh(&dev_base_lock);
1944 		if (dev->link_mode ^ value)
1945 			status |= DO_SETLINK_NOTIFY;
1946 		dev->link_mode = value;
1947 		write_unlock_bh(&dev_base_lock);
1948 	}
1949 
1950 	if (tb[IFLA_VFINFO_LIST]) {
1951 		struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1952 		struct nlattr *attr;
1953 		int rem;
1954 
1955 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1956 			if (nla_type(attr) != IFLA_VF_INFO ||
1957 			    nla_len(attr) < NLA_HDRLEN) {
1958 				err = -EINVAL;
1959 				goto errout;
1960 			}
1961 			err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1962 					       ifla_vf_policy);
1963 			if (err < 0)
1964 				goto errout;
1965 			err = do_setvfinfo(dev, vfinfo);
1966 			if (err < 0)
1967 				goto errout;
1968 			status |= DO_SETLINK_NOTIFY;
1969 		}
1970 	}
1971 	err = 0;
1972 
1973 	if (tb[IFLA_VF_PORTS]) {
1974 		struct nlattr *port[IFLA_PORT_MAX+1];
1975 		struct nlattr *attr;
1976 		int vf;
1977 		int rem;
1978 
1979 		err = -EOPNOTSUPP;
1980 		if (!ops->ndo_set_vf_port)
1981 			goto errout;
1982 
1983 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1984 			if (nla_type(attr) != IFLA_VF_PORT ||
1985 			    nla_len(attr) < NLA_HDRLEN) {
1986 				err = -EINVAL;
1987 				goto errout;
1988 			}
1989 			err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1990 					       ifla_port_policy);
1991 			if (err < 0)
1992 				goto errout;
1993 			if (!port[IFLA_PORT_VF]) {
1994 				err = -EOPNOTSUPP;
1995 				goto errout;
1996 			}
1997 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1998 			err = ops->ndo_set_vf_port(dev, vf, port);
1999 			if (err < 0)
2000 				goto errout;
2001 			status |= DO_SETLINK_NOTIFY;
2002 		}
2003 	}
2004 	err = 0;
2005 
2006 	if (tb[IFLA_PORT_SELF]) {
2007 		struct nlattr *port[IFLA_PORT_MAX+1];
2008 
2009 		err = nla_parse_nested(port, IFLA_PORT_MAX,
2010 			tb[IFLA_PORT_SELF], ifla_port_policy);
2011 		if (err < 0)
2012 			goto errout;
2013 
2014 		err = -EOPNOTSUPP;
2015 		if (ops->ndo_set_vf_port)
2016 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2017 		if (err < 0)
2018 			goto errout;
2019 		status |= DO_SETLINK_NOTIFY;
2020 	}
2021 
2022 	if (tb[IFLA_AF_SPEC]) {
2023 		struct nlattr *af;
2024 		int rem;
2025 
2026 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2027 			const struct rtnl_af_ops *af_ops;
2028 
2029 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2030 				BUG();
2031 
2032 			err = af_ops->set_link_af(dev, af);
2033 			if (err < 0)
2034 				goto errout;
2035 
2036 			status |= DO_SETLINK_NOTIFY;
2037 		}
2038 	}
2039 	err = 0;
2040 
2041 	if (tb[IFLA_PROTO_DOWN]) {
2042 		err = dev_change_proto_down(dev,
2043 					    nla_get_u8(tb[IFLA_PROTO_DOWN]));
2044 		if (err)
2045 			goto errout;
2046 		status |= DO_SETLINK_NOTIFY;
2047 	}
2048 
2049 errout:
2050 	if (status & DO_SETLINK_MODIFIED) {
2051 		if (status & DO_SETLINK_NOTIFY)
2052 			netdev_state_change(dev);
2053 
2054 		if (err < 0)
2055 			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",
2056 					     dev->name);
2057 	}
2058 
2059 	return err;
2060 }
2061 
2062 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2063 {
2064 	struct net *net = sock_net(skb->sk);
2065 	struct ifinfomsg *ifm;
2066 	struct net_device *dev;
2067 	int err;
2068 	struct nlattr *tb[IFLA_MAX+1];
2069 	char ifname[IFNAMSIZ];
2070 
2071 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2072 	if (err < 0)
2073 		goto errout;
2074 
2075 	if (tb[IFLA_IFNAME])
2076 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2077 	else
2078 		ifname[0] = '\0';
2079 
2080 	err = -EINVAL;
2081 	ifm = nlmsg_data(nlh);
2082 	if (ifm->ifi_index > 0)
2083 		dev = __dev_get_by_index(net, ifm->ifi_index);
2084 	else if (tb[IFLA_IFNAME])
2085 		dev = __dev_get_by_name(net, ifname);
2086 	else
2087 		goto errout;
2088 
2089 	if (dev == NULL) {
2090 		err = -ENODEV;
2091 		goto errout;
2092 	}
2093 
2094 	err = validate_linkmsg(dev, tb);
2095 	if (err < 0)
2096 		goto errout;
2097 
2098 	err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2099 errout:
2100 	return err;
2101 }
2102 
2103 static int rtnl_group_dellink(const struct net *net, int group)
2104 {
2105 	struct net_device *dev, *aux;
2106 	LIST_HEAD(list_kill);
2107 	bool found = false;
2108 
2109 	if (!group)
2110 		return -EPERM;
2111 
2112 	for_each_netdev(net, dev) {
2113 		if (dev->group == group) {
2114 			const struct rtnl_link_ops *ops;
2115 
2116 			found = true;
2117 			ops = dev->rtnl_link_ops;
2118 			if (!ops || !ops->dellink)
2119 				return -EOPNOTSUPP;
2120 		}
2121 	}
2122 
2123 	if (!found)
2124 		return -ENODEV;
2125 
2126 	for_each_netdev_safe(net, dev, aux) {
2127 		if (dev->group == group) {
2128 			const struct rtnl_link_ops *ops;
2129 
2130 			ops = dev->rtnl_link_ops;
2131 			ops->dellink(dev, &list_kill);
2132 		}
2133 	}
2134 	unregister_netdevice_many(&list_kill);
2135 
2136 	return 0;
2137 }
2138 
2139 int rtnl_delete_link(struct net_device *dev)
2140 {
2141 	const struct rtnl_link_ops *ops;
2142 	LIST_HEAD(list_kill);
2143 
2144 	ops = dev->rtnl_link_ops;
2145 	if (!ops || !ops->dellink)
2146 		return -EOPNOTSUPP;
2147 
2148 	ops->dellink(dev, &list_kill);
2149 	unregister_netdevice_many(&list_kill);
2150 
2151 	return 0;
2152 }
2153 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2154 
2155 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2156 {
2157 	struct net *net = sock_net(skb->sk);
2158 	struct net_device *dev;
2159 	struct ifinfomsg *ifm;
2160 	char ifname[IFNAMSIZ];
2161 	struct nlattr *tb[IFLA_MAX+1];
2162 	int err;
2163 
2164 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2165 	if (err < 0)
2166 		return err;
2167 
2168 	if (tb[IFLA_IFNAME])
2169 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2170 
2171 	ifm = nlmsg_data(nlh);
2172 	if (ifm->ifi_index > 0)
2173 		dev = __dev_get_by_index(net, ifm->ifi_index);
2174 	else if (tb[IFLA_IFNAME])
2175 		dev = __dev_get_by_name(net, ifname);
2176 	else if (tb[IFLA_GROUP])
2177 		return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2178 	else
2179 		return -EINVAL;
2180 
2181 	if (!dev)
2182 		return -ENODEV;
2183 
2184 	return rtnl_delete_link(dev);
2185 }
2186 
2187 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2188 {
2189 	unsigned int old_flags;
2190 	int err;
2191 
2192 	old_flags = dev->flags;
2193 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2194 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2195 		if (err < 0)
2196 			return err;
2197 	}
2198 
2199 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2200 
2201 	__dev_notify_flags(dev, old_flags, ~0U);
2202 	return 0;
2203 }
2204 EXPORT_SYMBOL(rtnl_configure_link);
2205 
2206 struct net_device *rtnl_create_link(struct net *net,
2207 	const char *ifname, unsigned char name_assign_type,
2208 	const struct rtnl_link_ops *ops, struct nlattr *tb[])
2209 {
2210 	int err;
2211 	struct net_device *dev;
2212 	unsigned int num_tx_queues = 1;
2213 	unsigned int num_rx_queues = 1;
2214 
2215 	if (tb[IFLA_NUM_TX_QUEUES])
2216 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2217 	else if (ops->get_num_tx_queues)
2218 		num_tx_queues = ops->get_num_tx_queues();
2219 
2220 	if (tb[IFLA_NUM_RX_QUEUES])
2221 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2222 	else if (ops->get_num_rx_queues)
2223 		num_rx_queues = ops->get_num_rx_queues();
2224 
2225 	err = -ENOMEM;
2226 	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2227 			       ops->setup, num_tx_queues, num_rx_queues);
2228 	if (!dev)
2229 		goto err;
2230 
2231 	dev_net_set(dev, net);
2232 	dev->rtnl_link_ops = ops;
2233 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2234 
2235 	if (tb[IFLA_MTU])
2236 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2237 	if (tb[IFLA_ADDRESS]) {
2238 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2239 				nla_len(tb[IFLA_ADDRESS]));
2240 		dev->addr_assign_type = NET_ADDR_SET;
2241 	}
2242 	if (tb[IFLA_BROADCAST])
2243 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2244 				nla_len(tb[IFLA_BROADCAST]));
2245 	if (tb[IFLA_TXQLEN])
2246 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2247 	if (tb[IFLA_OPERSTATE])
2248 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2249 	if (tb[IFLA_LINKMODE])
2250 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2251 	if (tb[IFLA_GROUP])
2252 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2253 
2254 	return dev;
2255 
2256 err:
2257 	return ERR_PTR(err);
2258 }
2259 EXPORT_SYMBOL(rtnl_create_link);
2260 
2261 static int rtnl_group_changelink(const struct sk_buff *skb,
2262 		struct net *net, int group,
2263 		struct ifinfomsg *ifm,
2264 		struct nlattr **tb)
2265 {
2266 	struct net_device *dev, *aux;
2267 	int err;
2268 
2269 	for_each_netdev_safe(net, dev, aux) {
2270 		if (dev->group == group) {
2271 			err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2272 			if (err < 0)
2273 				return err;
2274 		}
2275 	}
2276 
2277 	return 0;
2278 }
2279 
2280 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2281 {
2282 	struct net *net = sock_net(skb->sk);
2283 	const struct rtnl_link_ops *ops;
2284 	const struct rtnl_link_ops *m_ops = NULL;
2285 	struct net_device *dev;
2286 	struct net_device *master_dev = NULL;
2287 	struct ifinfomsg *ifm;
2288 	char kind[MODULE_NAME_LEN];
2289 	char ifname[IFNAMSIZ];
2290 	struct nlattr *tb[IFLA_MAX+1];
2291 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2292 	unsigned char name_assign_type = NET_NAME_USER;
2293 	int err;
2294 
2295 #ifdef CONFIG_MODULES
2296 replay:
2297 #endif
2298 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2299 	if (err < 0)
2300 		return err;
2301 
2302 	if (tb[IFLA_IFNAME])
2303 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2304 	else
2305 		ifname[0] = '\0';
2306 
2307 	ifm = nlmsg_data(nlh);
2308 	if (ifm->ifi_index > 0)
2309 		dev = __dev_get_by_index(net, ifm->ifi_index);
2310 	else {
2311 		if (ifname[0])
2312 			dev = __dev_get_by_name(net, ifname);
2313 		else
2314 			dev = NULL;
2315 	}
2316 
2317 	if (dev) {
2318 		master_dev = netdev_master_upper_dev_get(dev);
2319 		if (master_dev)
2320 			m_ops = master_dev->rtnl_link_ops;
2321 	}
2322 
2323 	err = validate_linkmsg(dev, tb);
2324 	if (err < 0)
2325 		return err;
2326 
2327 	if (tb[IFLA_LINKINFO]) {
2328 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2329 				       tb[IFLA_LINKINFO], ifla_info_policy);
2330 		if (err < 0)
2331 			return err;
2332 	} else
2333 		memset(linkinfo, 0, sizeof(linkinfo));
2334 
2335 	if (linkinfo[IFLA_INFO_KIND]) {
2336 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2337 		ops = rtnl_link_ops_get(kind);
2338 	} else {
2339 		kind[0] = '\0';
2340 		ops = NULL;
2341 	}
2342 
2343 	if (1) {
2344 		struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2345 		struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2346 		struct nlattr **data = NULL;
2347 		struct nlattr **slave_data = NULL;
2348 		struct net *dest_net, *link_net = NULL;
2349 
2350 		if (ops) {
2351 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2352 				err = nla_parse_nested(attr, ops->maxtype,
2353 						       linkinfo[IFLA_INFO_DATA],
2354 						       ops->policy);
2355 				if (err < 0)
2356 					return err;
2357 				data = attr;
2358 			}
2359 			if (ops->validate) {
2360 				err = ops->validate(tb, data);
2361 				if (err < 0)
2362 					return err;
2363 			}
2364 		}
2365 
2366 		if (m_ops) {
2367 			if (m_ops->slave_maxtype &&
2368 			    linkinfo[IFLA_INFO_SLAVE_DATA]) {
2369 				err = nla_parse_nested(slave_attr,
2370 						       m_ops->slave_maxtype,
2371 						       linkinfo[IFLA_INFO_SLAVE_DATA],
2372 						       m_ops->slave_policy);
2373 				if (err < 0)
2374 					return err;
2375 				slave_data = slave_attr;
2376 			}
2377 			if (m_ops->slave_validate) {
2378 				err = m_ops->slave_validate(tb, slave_data);
2379 				if (err < 0)
2380 					return err;
2381 			}
2382 		}
2383 
2384 		if (dev) {
2385 			int status = 0;
2386 
2387 			if (nlh->nlmsg_flags & NLM_F_EXCL)
2388 				return -EEXIST;
2389 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
2390 				return -EOPNOTSUPP;
2391 
2392 			if (linkinfo[IFLA_INFO_DATA]) {
2393 				if (!ops || ops != dev->rtnl_link_ops ||
2394 				    !ops->changelink)
2395 					return -EOPNOTSUPP;
2396 
2397 				err = ops->changelink(dev, tb, data);
2398 				if (err < 0)
2399 					return err;
2400 				status |= DO_SETLINK_NOTIFY;
2401 			}
2402 
2403 			if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2404 				if (!m_ops || !m_ops->slave_changelink)
2405 					return -EOPNOTSUPP;
2406 
2407 				err = m_ops->slave_changelink(master_dev, dev,
2408 							      tb, slave_data);
2409 				if (err < 0)
2410 					return err;
2411 				status |= DO_SETLINK_NOTIFY;
2412 			}
2413 
2414 			return do_setlink(skb, dev, ifm, tb, ifname, status);
2415 		}
2416 
2417 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2418 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2419 				return rtnl_group_changelink(skb, net,
2420 						nla_get_u32(tb[IFLA_GROUP]),
2421 						ifm, tb);
2422 			return -ENODEV;
2423 		}
2424 
2425 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2426 			return -EOPNOTSUPP;
2427 
2428 		if (!ops) {
2429 #ifdef CONFIG_MODULES
2430 			if (kind[0]) {
2431 				__rtnl_unlock();
2432 				request_module("rtnl-link-%s", kind);
2433 				rtnl_lock();
2434 				ops = rtnl_link_ops_get(kind);
2435 				if (ops)
2436 					goto replay;
2437 			}
2438 #endif
2439 			return -EOPNOTSUPP;
2440 		}
2441 
2442 		if (!ops->setup)
2443 			return -EOPNOTSUPP;
2444 
2445 		if (!ifname[0]) {
2446 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2447 			name_assign_type = NET_NAME_ENUM;
2448 		}
2449 
2450 		dest_net = rtnl_link_get_net(net, tb);
2451 		if (IS_ERR(dest_net))
2452 			return PTR_ERR(dest_net);
2453 
2454 		err = -EPERM;
2455 		if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2456 			goto out;
2457 
2458 		if (tb[IFLA_LINK_NETNSID]) {
2459 			int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2460 
2461 			link_net = get_net_ns_by_id(dest_net, id);
2462 			if (!link_net) {
2463 				err =  -EINVAL;
2464 				goto out;
2465 			}
2466 			err = -EPERM;
2467 			if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2468 				goto out;
2469 		}
2470 
2471 		dev = rtnl_create_link(link_net ? : dest_net, ifname,
2472 				       name_assign_type, ops, tb);
2473 		if (IS_ERR(dev)) {
2474 			err = PTR_ERR(dev);
2475 			goto out;
2476 		}
2477 
2478 		dev->ifindex = ifm->ifi_index;
2479 
2480 		if (ops->newlink) {
2481 			err = ops->newlink(link_net ? : net, dev, tb, data);
2482 			/* Drivers should call free_netdev() in ->destructor
2483 			 * and unregister it on failure after registration
2484 			 * so that device could be finally freed in rtnl_unlock.
2485 			 */
2486 			if (err < 0) {
2487 				/* If device is not registered at all, free it now */
2488 				if (dev->reg_state == NETREG_UNINITIALIZED)
2489 					free_netdev(dev);
2490 				goto out;
2491 			}
2492 		} else {
2493 			err = register_netdevice(dev);
2494 			if (err < 0) {
2495 				free_netdev(dev);
2496 				goto out;
2497 			}
2498 		}
2499 		err = rtnl_configure_link(dev, ifm);
2500 		if (err < 0)
2501 			goto out_unregister;
2502 		if (link_net) {
2503 			err = dev_change_net_namespace(dev, dest_net, ifname);
2504 			if (err < 0)
2505 				goto out_unregister;
2506 		}
2507 out:
2508 		if (link_net)
2509 			put_net(link_net);
2510 		put_net(dest_net);
2511 		return err;
2512 out_unregister:
2513 		if (ops->newlink) {
2514 			LIST_HEAD(list_kill);
2515 
2516 			ops->dellink(dev, &list_kill);
2517 			unregister_netdevice_many(&list_kill);
2518 		} else {
2519 			unregister_netdevice(dev);
2520 		}
2521 		goto out;
2522 	}
2523 }
2524 
2525 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2526 {
2527 	struct net *net = sock_net(skb->sk);
2528 	struct ifinfomsg *ifm;
2529 	char ifname[IFNAMSIZ];
2530 	struct nlattr *tb[IFLA_MAX+1];
2531 	struct net_device *dev = NULL;
2532 	struct sk_buff *nskb;
2533 	int err;
2534 	u32 ext_filter_mask = 0;
2535 
2536 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2537 	if (err < 0)
2538 		return err;
2539 
2540 	if (tb[IFLA_IFNAME])
2541 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2542 
2543 	if (tb[IFLA_EXT_MASK])
2544 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2545 
2546 	ifm = nlmsg_data(nlh);
2547 	if (ifm->ifi_index > 0)
2548 		dev = __dev_get_by_index(net, ifm->ifi_index);
2549 	else if (tb[IFLA_IFNAME])
2550 		dev = __dev_get_by_name(net, ifname);
2551 	else
2552 		return -EINVAL;
2553 
2554 	if (dev == NULL)
2555 		return -ENODEV;
2556 
2557 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2558 	if (nskb == NULL)
2559 		return -ENOBUFS;
2560 
2561 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2562 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2563 	if (err < 0) {
2564 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
2565 		WARN_ON(err == -EMSGSIZE);
2566 		kfree_skb(nskb);
2567 	} else
2568 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2569 
2570 	return err;
2571 }
2572 
2573 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2574 {
2575 	struct net *net = sock_net(skb->sk);
2576 	struct net_device *dev;
2577 	struct nlattr *tb[IFLA_MAX+1];
2578 	u32 ext_filter_mask = 0;
2579 	u16 min_ifinfo_dump_size = 0;
2580 	int hdrlen;
2581 
2582 	/* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2583 	hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2584 		 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2585 
2586 	if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2587 		if (tb[IFLA_EXT_MASK])
2588 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2589 	}
2590 
2591 	if (!ext_filter_mask)
2592 		return NLMSG_GOODSIZE;
2593 	/*
2594 	 * traverse the list of net devices and compute the minimum
2595 	 * buffer size based upon the filter mask.
2596 	 */
2597 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2598 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2599 					     if_nlmsg_size(dev,
2600 						           ext_filter_mask));
2601 	}
2602 
2603 	return min_ifinfo_dump_size;
2604 }
2605 
2606 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2607 {
2608 	int idx;
2609 	int s_idx = cb->family;
2610 
2611 	if (s_idx == 0)
2612 		s_idx = 1;
2613 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2614 		int type = cb->nlh->nlmsg_type-RTM_BASE;
2615 		if (idx < s_idx || idx == PF_PACKET)
2616 			continue;
2617 		if (rtnl_msg_handlers[idx] == NULL ||
2618 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
2619 			continue;
2620 		if (idx > s_idx) {
2621 			memset(&cb->args[0], 0, sizeof(cb->args));
2622 			cb->prev_seq = 0;
2623 			cb->seq = 0;
2624 		}
2625 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2626 			break;
2627 	}
2628 	cb->family = idx;
2629 
2630 	return skb->len;
2631 }
2632 
2633 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2634 				       unsigned int change, gfp_t flags)
2635 {
2636 	struct net *net = dev_net(dev);
2637 	struct sk_buff *skb;
2638 	int err = -ENOBUFS;
2639 	size_t if_info_size;
2640 
2641 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2642 	if (skb == NULL)
2643 		goto errout;
2644 
2645 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2646 	if (err < 0) {
2647 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2648 		WARN_ON(err == -EMSGSIZE);
2649 		kfree_skb(skb);
2650 		goto errout;
2651 	}
2652 	return skb;
2653 errout:
2654 	if (err < 0)
2655 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2656 	return NULL;
2657 }
2658 
2659 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2660 {
2661 	struct net *net = dev_net(dev);
2662 
2663 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2664 }
2665 
2666 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2667 		  gfp_t flags)
2668 {
2669 	struct sk_buff *skb;
2670 
2671 	if (dev->reg_state != NETREG_REGISTERED)
2672 		return;
2673 
2674 	skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2675 	if (skb)
2676 		rtmsg_ifinfo_send(skb, dev, flags);
2677 }
2678 EXPORT_SYMBOL(rtmsg_ifinfo);
2679 
2680 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2681 				   struct net_device *dev,
2682 				   u8 *addr, u16 vid, u32 pid, u32 seq,
2683 				   int type, unsigned int flags,
2684 				   int nlflags, u16 ndm_state)
2685 {
2686 	struct nlmsghdr *nlh;
2687 	struct ndmsg *ndm;
2688 
2689 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2690 	if (!nlh)
2691 		return -EMSGSIZE;
2692 
2693 	ndm = nlmsg_data(nlh);
2694 	ndm->ndm_family  = AF_BRIDGE;
2695 	ndm->ndm_pad1	 = 0;
2696 	ndm->ndm_pad2    = 0;
2697 	ndm->ndm_flags	 = flags;
2698 	ndm->ndm_type	 = 0;
2699 	ndm->ndm_ifindex = dev->ifindex;
2700 	ndm->ndm_state   = ndm_state;
2701 
2702 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2703 		goto nla_put_failure;
2704 	if (vid)
2705 		if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2706 			goto nla_put_failure;
2707 
2708 	nlmsg_end(skb, nlh);
2709 	return 0;
2710 
2711 nla_put_failure:
2712 	nlmsg_cancel(skb, nlh);
2713 	return -EMSGSIZE;
2714 }
2715 
2716 static inline size_t rtnl_fdb_nlmsg_size(void)
2717 {
2718 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2719 }
2720 
2721 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2722 			    u16 ndm_state)
2723 {
2724 	struct net *net = dev_net(dev);
2725 	struct sk_buff *skb;
2726 	int err = -ENOBUFS;
2727 
2728 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2729 	if (!skb)
2730 		goto errout;
2731 
2732 	err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2733 				      0, 0, type, NTF_SELF, 0, ndm_state);
2734 	if (err < 0) {
2735 		kfree_skb(skb);
2736 		goto errout;
2737 	}
2738 
2739 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2740 	return;
2741 errout:
2742 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2743 }
2744 
2745 /**
2746  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2747  */
2748 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2749 		     struct nlattr *tb[],
2750 		     struct net_device *dev,
2751 		     const unsigned char *addr, u16 vid,
2752 		     u16 flags)
2753 {
2754 	int err = -EINVAL;
2755 
2756 	/* If aging addresses are supported device will need to
2757 	 * implement its own handler for this.
2758 	 */
2759 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2760 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2761 		return err;
2762 	}
2763 
2764 	if (vid) {
2765 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2766 		return err;
2767 	}
2768 
2769 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2770 		err = dev_uc_add_excl(dev, addr);
2771 	else if (is_multicast_ether_addr(addr))
2772 		err = dev_mc_add_excl(dev, addr);
2773 
2774 	/* Only return duplicate errors if NLM_F_EXCL is set */
2775 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2776 		err = 0;
2777 
2778 	return err;
2779 }
2780 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2781 
2782 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2783 {
2784 	u16 vid = 0;
2785 
2786 	if (vlan_attr) {
2787 		if (nla_len(vlan_attr) != sizeof(u16)) {
2788 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2789 			return -EINVAL;
2790 		}
2791 
2792 		vid = nla_get_u16(vlan_attr);
2793 
2794 		if (!vid || vid >= VLAN_VID_MASK) {
2795 			pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2796 				vid);
2797 			return -EINVAL;
2798 		}
2799 	}
2800 	*p_vid = vid;
2801 	return 0;
2802 }
2803 
2804 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2805 {
2806 	struct net *net = sock_net(skb->sk);
2807 	struct ndmsg *ndm;
2808 	struct nlattr *tb[NDA_MAX+1];
2809 	struct net_device *dev;
2810 	u8 *addr;
2811 	u16 vid;
2812 	int err;
2813 
2814 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2815 	if (err < 0)
2816 		return err;
2817 
2818 	ndm = nlmsg_data(nlh);
2819 	if (ndm->ndm_ifindex == 0) {
2820 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2821 		return -EINVAL;
2822 	}
2823 
2824 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2825 	if (dev == NULL) {
2826 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2827 		return -ENODEV;
2828 	}
2829 
2830 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2831 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2832 		return -EINVAL;
2833 	}
2834 
2835 	addr = nla_data(tb[NDA_LLADDR]);
2836 
2837 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2838 	if (err)
2839 		return err;
2840 
2841 	err = -EOPNOTSUPP;
2842 
2843 	/* Support fdb on master device the net/bridge default case */
2844 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2845 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2846 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2847 		const struct net_device_ops *ops = br_dev->netdev_ops;
2848 
2849 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2850 				       nlh->nlmsg_flags);
2851 		if (err)
2852 			goto out;
2853 		else
2854 			ndm->ndm_flags &= ~NTF_MASTER;
2855 	}
2856 
2857 	/* Embedded bridge, macvlan, and any other device support */
2858 	if ((ndm->ndm_flags & NTF_SELF)) {
2859 		if (dev->netdev_ops->ndo_fdb_add)
2860 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2861 							   vid,
2862 							   nlh->nlmsg_flags);
2863 		else
2864 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2865 					       nlh->nlmsg_flags);
2866 
2867 		if (!err) {
2868 			rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
2869 					ndm->ndm_state);
2870 			ndm->ndm_flags &= ~NTF_SELF;
2871 		}
2872 	}
2873 out:
2874 	return err;
2875 }
2876 
2877 /**
2878  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2879  */
2880 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2881 		     struct nlattr *tb[],
2882 		     struct net_device *dev,
2883 		     const unsigned char *addr, u16 vid)
2884 {
2885 	int err = -EINVAL;
2886 
2887 	/* If aging addresses are supported device will need to
2888 	 * implement its own handler for this.
2889 	 */
2890 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2891 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2892 		return err;
2893 	}
2894 
2895 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2896 		err = dev_uc_del(dev, addr);
2897 	else if (is_multicast_ether_addr(addr))
2898 		err = dev_mc_del(dev, addr);
2899 
2900 	return err;
2901 }
2902 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2903 
2904 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2905 {
2906 	struct net *net = sock_net(skb->sk);
2907 	struct ndmsg *ndm;
2908 	struct nlattr *tb[NDA_MAX+1];
2909 	struct net_device *dev;
2910 	int err = -EINVAL;
2911 	__u8 *addr;
2912 	u16 vid;
2913 
2914 	if (!netlink_capable(skb, CAP_NET_ADMIN))
2915 		return -EPERM;
2916 
2917 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2918 	if (err < 0)
2919 		return err;
2920 
2921 	ndm = nlmsg_data(nlh);
2922 	if (ndm->ndm_ifindex == 0) {
2923 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2924 		return -EINVAL;
2925 	}
2926 
2927 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2928 	if (dev == NULL) {
2929 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2930 		return -ENODEV;
2931 	}
2932 
2933 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2934 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2935 		return -EINVAL;
2936 	}
2937 
2938 	addr = nla_data(tb[NDA_LLADDR]);
2939 
2940 	err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2941 	if (err)
2942 		return err;
2943 
2944 	err = -EOPNOTSUPP;
2945 
2946 	/* Support fdb on master device the net/bridge default case */
2947 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2948 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2949 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2950 		const struct net_device_ops *ops = br_dev->netdev_ops;
2951 
2952 		if (ops->ndo_fdb_del)
2953 			err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2954 
2955 		if (err)
2956 			goto out;
2957 		else
2958 			ndm->ndm_flags &= ~NTF_MASTER;
2959 	}
2960 
2961 	/* Embedded bridge, macvlan, and any other device support */
2962 	if (ndm->ndm_flags & NTF_SELF) {
2963 		if (dev->netdev_ops->ndo_fdb_del)
2964 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2965 							   vid);
2966 		else
2967 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2968 
2969 		if (!err) {
2970 			rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
2971 					ndm->ndm_state);
2972 			ndm->ndm_flags &= ~NTF_SELF;
2973 		}
2974 	}
2975 out:
2976 	return err;
2977 }
2978 
2979 static int nlmsg_populate_fdb(struct sk_buff *skb,
2980 			      struct netlink_callback *cb,
2981 			      struct net_device *dev,
2982 			      int *idx,
2983 			      struct netdev_hw_addr_list *list)
2984 {
2985 	struct netdev_hw_addr *ha;
2986 	int err;
2987 	u32 portid, seq;
2988 
2989 	portid = NETLINK_CB(cb->skb).portid;
2990 	seq = cb->nlh->nlmsg_seq;
2991 
2992 	list_for_each_entry(ha, &list->list, list) {
2993 		if (*idx < cb->args[0])
2994 			goto skip;
2995 
2996 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2997 					      portid, seq,
2998 					      RTM_NEWNEIGH, NTF_SELF,
2999 					      NLM_F_MULTI, NUD_PERMANENT);
3000 		if (err < 0)
3001 			return err;
3002 skip:
3003 		*idx += 1;
3004 	}
3005 	return 0;
3006 }
3007 
3008 /**
3009  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3010  * @nlh: netlink message header
3011  * @dev: netdevice
3012  *
3013  * Default netdevice operation to dump the existing unicast address list.
3014  * Returns number of addresses from list put in skb.
3015  */
3016 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3017 		      struct netlink_callback *cb,
3018 		      struct net_device *dev,
3019 		      struct net_device *filter_dev,
3020 		      int idx)
3021 {
3022 	int err;
3023 
3024 	netif_addr_lock_bh(dev);
3025 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
3026 	if (err)
3027 		goto out;
3028 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
3029 out:
3030 	netif_addr_unlock_bh(dev);
3031 	cb->args[1] = err;
3032 	return idx;
3033 }
3034 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3035 
3036 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3037 {
3038 	struct net_device *dev;
3039 	struct nlattr *tb[IFLA_MAX+1];
3040 	struct net_device *br_dev = NULL;
3041 	const struct net_device_ops *ops = NULL;
3042 	const struct net_device_ops *cops = NULL;
3043 	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3044 	struct net *net = sock_net(skb->sk);
3045 	int brport_idx = 0;
3046 	int br_idx = 0;
3047 	int idx = 0;
3048 
3049 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3050 			ifla_policy) == 0) {
3051 		if (tb[IFLA_MASTER])
3052 			br_idx = nla_get_u32(tb[IFLA_MASTER]);
3053 	}
3054 
3055 	brport_idx = ifm->ifi_index;
3056 
3057 	if (br_idx) {
3058 		br_dev = __dev_get_by_index(net, br_idx);
3059 		if (!br_dev)
3060 			return -ENODEV;
3061 
3062 		ops = br_dev->netdev_ops;
3063 	}
3064 
3065 	cb->args[1] = 0;
3066 	for_each_netdev(net, dev) {
3067 		if (brport_idx && (dev->ifindex != brport_idx))
3068 			continue;
3069 
3070 		if (!br_idx) { /* user did not specify a specific bridge */
3071 			if (dev->priv_flags & IFF_BRIDGE_PORT) {
3072 				br_dev = netdev_master_upper_dev_get(dev);
3073 				cops = br_dev->netdev_ops;
3074 			}
3075 
3076 		} else {
3077 			if (dev != br_dev &&
3078 			    !(dev->priv_flags & IFF_BRIDGE_PORT))
3079 				continue;
3080 
3081 			if (br_dev != netdev_master_upper_dev_get(dev) &&
3082 			    !(dev->priv_flags & IFF_EBRIDGE))
3083 				continue;
3084 
3085 			cops = ops;
3086 		}
3087 
3088 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
3089 			if (cops && cops->ndo_fdb_dump)
3090 				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3091 							 idx);
3092 		}
3093 		if (cb->args[1] == -EMSGSIZE)
3094 			break;
3095 
3096 		if (dev->netdev_ops->ndo_fdb_dump)
3097 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3098 							    idx);
3099 		else
3100 			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3101 		if (cb->args[1] == -EMSGSIZE)
3102 			break;
3103 
3104 		cops = NULL;
3105 	}
3106 
3107 	cb->args[0] = idx;
3108 	return skb->len;
3109 }
3110 
3111 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3112 			       unsigned int attrnum, unsigned int flag)
3113 {
3114 	if (mask & flag)
3115 		return nla_put_u8(skb, attrnum, !!(flags & flag));
3116 	return 0;
3117 }
3118 
3119 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3120 			    struct net_device *dev, u16 mode,
3121 			    u32 flags, u32 mask, int nlflags,
3122 			    u32 filter_mask,
3123 			    int (*vlan_fill)(struct sk_buff *skb,
3124 					     struct net_device *dev,
3125 					     u32 filter_mask))
3126 {
3127 	struct nlmsghdr *nlh;
3128 	struct ifinfomsg *ifm;
3129 	struct nlattr *br_afspec;
3130 	struct nlattr *protinfo;
3131 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3132 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3133 	int err = 0;
3134 
3135 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3136 	if (nlh == NULL)
3137 		return -EMSGSIZE;
3138 
3139 	ifm = nlmsg_data(nlh);
3140 	ifm->ifi_family = AF_BRIDGE;
3141 	ifm->__ifi_pad = 0;
3142 	ifm->ifi_type = dev->type;
3143 	ifm->ifi_index = dev->ifindex;
3144 	ifm->ifi_flags = dev_get_flags(dev);
3145 	ifm->ifi_change = 0;
3146 
3147 
3148 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3149 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3150 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3151 	    (br_dev &&
3152 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3153 	    (dev->addr_len &&
3154 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3155 	    (dev->ifindex != dev_get_iflink(dev) &&
3156 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3157 		goto nla_put_failure;
3158 
3159 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3160 	if (!br_afspec)
3161 		goto nla_put_failure;
3162 
3163 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3164 		nla_nest_cancel(skb, br_afspec);
3165 		goto nla_put_failure;
3166 	}
3167 
3168 	if (mode != BRIDGE_MODE_UNDEF) {
3169 		if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3170 			nla_nest_cancel(skb, br_afspec);
3171 			goto nla_put_failure;
3172 		}
3173 	}
3174 	if (vlan_fill) {
3175 		err = vlan_fill(skb, dev, filter_mask);
3176 		if (err) {
3177 			nla_nest_cancel(skb, br_afspec);
3178 			goto nla_put_failure;
3179 		}
3180 	}
3181 	nla_nest_end(skb, br_afspec);
3182 
3183 	protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3184 	if (!protinfo)
3185 		goto nla_put_failure;
3186 
3187 	if (brport_nla_put_flag(skb, flags, mask,
3188 				IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3189 	    brport_nla_put_flag(skb, flags, mask,
3190 				IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3191 	    brport_nla_put_flag(skb, flags, mask,
3192 				IFLA_BRPORT_FAST_LEAVE,
3193 				BR_MULTICAST_FAST_LEAVE) ||
3194 	    brport_nla_put_flag(skb, flags, mask,
3195 				IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3196 	    brport_nla_put_flag(skb, flags, mask,
3197 				IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3198 	    brport_nla_put_flag(skb, flags, mask,
3199 				IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3200 	    brport_nla_put_flag(skb, flags, mask,
3201 				IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3202 	    brport_nla_put_flag(skb, flags, mask,
3203 				IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3204 		nla_nest_cancel(skb, protinfo);
3205 		goto nla_put_failure;
3206 	}
3207 
3208 	nla_nest_end(skb, protinfo);
3209 
3210 	nlmsg_end(skb, nlh);
3211 	return 0;
3212 nla_put_failure:
3213 	nlmsg_cancel(skb, nlh);
3214 	return err ? err : -EMSGSIZE;
3215 }
3216 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3217 
3218 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3219 {
3220 	struct net *net = sock_net(skb->sk);
3221 	struct net_device *dev;
3222 	int idx = 0;
3223 	u32 portid = NETLINK_CB(cb->skb).portid;
3224 	u32 seq = cb->nlh->nlmsg_seq;
3225 	u32 filter_mask = 0;
3226 	int err;
3227 
3228 	if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3229 		struct nlattr *extfilt;
3230 
3231 		extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3232 					  IFLA_EXT_MASK);
3233 		if (extfilt) {
3234 			if (nla_len(extfilt) < sizeof(filter_mask))
3235 				return -EINVAL;
3236 
3237 			filter_mask = nla_get_u32(extfilt);
3238 		}
3239 	}
3240 
3241 	rcu_read_lock();
3242 	for_each_netdev_rcu(net, dev) {
3243 		const struct net_device_ops *ops = dev->netdev_ops;
3244 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3245 
3246 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3247 			if (idx >= cb->args[0]) {
3248 				err = br_dev->netdev_ops->ndo_bridge_getlink(
3249 						skb, portid, seq, dev,
3250 						filter_mask, NLM_F_MULTI);
3251 				if (err < 0 && err != -EOPNOTSUPP)
3252 					break;
3253 			}
3254 			idx++;
3255 		}
3256 
3257 		if (ops->ndo_bridge_getlink) {
3258 			if (idx >= cb->args[0]) {
3259 				err = ops->ndo_bridge_getlink(skb, portid,
3260 							      seq, dev,
3261 							      filter_mask,
3262 							      NLM_F_MULTI);
3263 				if (err < 0 && err != -EOPNOTSUPP)
3264 					break;
3265 			}
3266 			idx++;
3267 		}
3268 	}
3269 	rcu_read_unlock();
3270 	cb->args[0] = idx;
3271 
3272 	return skb->len;
3273 }
3274 
3275 static inline size_t bridge_nlmsg_size(void)
3276 {
3277 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3278 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
3279 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
3280 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
3281 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
3282 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
3283 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
3284 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
3285 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
3286 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
3287 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
3288 }
3289 
3290 static int rtnl_bridge_notify(struct net_device *dev)
3291 {
3292 	struct net *net = dev_net(dev);
3293 	struct sk_buff *skb;
3294 	int err = -EOPNOTSUPP;
3295 
3296 	if (!dev->netdev_ops->ndo_bridge_getlink)
3297 		return 0;
3298 
3299 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3300 	if (!skb) {
3301 		err = -ENOMEM;
3302 		goto errout;
3303 	}
3304 
3305 	err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3306 	if (err < 0)
3307 		goto errout;
3308 
3309 	if (!skb->len)
3310 		goto errout;
3311 
3312 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3313 	return 0;
3314 errout:
3315 	WARN_ON(err == -EMSGSIZE);
3316 	kfree_skb(skb);
3317 	if (err)
3318 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3319 	return err;
3320 }
3321 
3322 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3323 {
3324 	struct net *net = sock_net(skb->sk);
3325 	struct ifinfomsg *ifm;
3326 	struct net_device *dev;
3327 	struct nlattr *br_spec, *attr = NULL;
3328 	int rem, err = -EOPNOTSUPP;
3329 	u16 flags = 0;
3330 	bool have_flags = false;
3331 
3332 	if (nlmsg_len(nlh) < sizeof(*ifm))
3333 		return -EINVAL;
3334 
3335 	ifm = nlmsg_data(nlh);
3336 	if (ifm->ifi_family != AF_BRIDGE)
3337 		return -EPFNOSUPPORT;
3338 
3339 	dev = __dev_get_by_index(net, ifm->ifi_index);
3340 	if (!dev) {
3341 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3342 		return -ENODEV;
3343 	}
3344 
3345 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3346 	if (br_spec) {
3347 		nla_for_each_nested(attr, br_spec, rem) {
3348 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3349 				if (nla_len(attr) < sizeof(flags))
3350 					return -EINVAL;
3351 
3352 				have_flags = true;
3353 				flags = nla_get_u16(attr);
3354 				break;
3355 			}
3356 		}
3357 	}
3358 
3359 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3360 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3361 
3362 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3363 			err = -EOPNOTSUPP;
3364 			goto out;
3365 		}
3366 
3367 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3368 		if (err)
3369 			goto out;
3370 
3371 		flags &= ~BRIDGE_FLAGS_MASTER;
3372 	}
3373 
3374 	if ((flags & BRIDGE_FLAGS_SELF)) {
3375 		if (!dev->netdev_ops->ndo_bridge_setlink)
3376 			err = -EOPNOTSUPP;
3377 		else
3378 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3379 								  flags);
3380 		if (!err) {
3381 			flags &= ~BRIDGE_FLAGS_SELF;
3382 
3383 			/* Generate event to notify upper layer of bridge
3384 			 * change
3385 			 */
3386 			err = rtnl_bridge_notify(dev);
3387 		}
3388 	}
3389 
3390 	if (have_flags)
3391 		memcpy(nla_data(attr), &flags, sizeof(flags));
3392 out:
3393 	return err;
3394 }
3395 
3396 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3397 {
3398 	struct net *net = sock_net(skb->sk);
3399 	struct ifinfomsg *ifm;
3400 	struct net_device *dev;
3401 	struct nlattr *br_spec, *attr = NULL;
3402 	int rem, err = -EOPNOTSUPP;
3403 	u16 flags = 0;
3404 	bool have_flags = false;
3405 
3406 	if (nlmsg_len(nlh) < sizeof(*ifm))
3407 		return -EINVAL;
3408 
3409 	ifm = nlmsg_data(nlh);
3410 	if (ifm->ifi_family != AF_BRIDGE)
3411 		return -EPFNOSUPPORT;
3412 
3413 	dev = __dev_get_by_index(net, ifm->ifi_index);
3414 	if (!dev) {
3415 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3416 		return -ENODEV;
3417 	}
3418 
3419 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3420 	if (br_spec) {
3421 		nla_for_each_nested(attr, br_spec, rem) {
3422 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3423 				if (nla_len(attr) < sizeof(flags))
3424 					return -EINVAL;
3425 
3426 				have_flags = true;
3427 				flags = nla_get_u16(attr);
3428 				break;
3429 			}
3430 		}
3431 	}
3432 
3433 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3434 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3435 
3436 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3437 			err = -EOPNOTSUPP;
3438 			goto out;
3439 		}
3440 
3441 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3442 		if (err)
3443 			goto out;
3444 
3445 		flags &= ~BRIDGE_FLAGS_MASTER;
3446 	}
3447 
3448 	if ((flags & BRIDGE_FLAGS_SELF)) {
3449 		if (!dev->netdev_ops->ndo_bridge_dellink)
3450 			err = -EOPNOTSUPP;
3451 		else
3452 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3453 								  flags);
3454 
3455 		if (!err) {
3456 			flags &= ~BRIDGE_FLAGS_SELF;
3457 
3458 			/* Generate event to notify upper layer of bridge
3459 			 * change
3460 			 */
3461 			err = rtnl_bridge_notify(dev);
3462 		}
3463 	}
3464 
3465 	if (have_flags)
3466 		memcpy(nla_data(attr), &flags, sizeof(flags));
3467 out:
3468 	return err;
3469 }
3470 
3471 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3472 {
3473 	return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3474 	       (!idxattr || idxattr == attrid);
3475 }
3476 
3477 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3478 			       int type, u32 pid, u32 seq, u32 change,
3479 			       unsigned int flags, unsigned int filter_mask,
3480 			       int *idxattr, int *prividx)
3481 {
3482 	struct if_stats_msg *ifsm;
3483 	struct nlmsghdr *nlh;
3484 	struct nlattr *attr;
3485 	int s_prividx = *prividx;
3486 
3487 	ASSERT_RTNL();
3488 
3489 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3490 	if (!nlh)
3491 		return -EMSGSIZE;
3492 
3493 	ifsm = nlmsg_data(nlh);
3494 	ifsm->ifindex = dev->ifindex;
3495 	ifsm->filter_mask = filter_mask;
3496 
3497 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3498 		struct rtnl_link_stats64 *sp;
3499 
3500 		attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3501 					 sizeof(struct rtnl_link_stats64),
3502 					 IFLA_STATS_UNSPEC);
3503 		if (!attr)
3504 			goto nla_put_failure;
3505 
3506 		sp = nla_data(attr);
3507 		dev_get_stats(dev, sp);
3508 	}
3509 
3510 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3511 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3512 
3513 		if (ops && ops->fill_linkxstats) {
3514 			int err;
3515 
3516 			*idxattr = IFLA_STATS_LINK_XSTATS;
3517 			attr = nla_nest_start(skb,
3518 					      IFLA_STATS_LINK_XSTATS);
3519 			if (!attr)
3520 				goto nla_put_failure;
3521 
3522 			err = ops->fill_linkxstats(skb, dev, prividx);
3523 			nla_nest_end(skb, attr);
3524 			if (err)
3525 				goto nla_put_failure;
3526 			*idxattr = 0;
3527 		}
3528 	}
3529 
3530 	nlmsg_end(skb, nlh);
3531 
3532 	return 0;
3533 
3534 nla_put_failure:
3535 	/* not a multi message or no progress mean a real error */
3536 	if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3537 		nlmsg_cancel(skb, nlh);
3538 	else
3539 		nlmsg_end(skb, nlh);
3540 
3541 	return -EMSGSIZE;
3542 }
3543 
3544 static const struct nla_policy ifla_stats_policy[IFLA_STATS_MAX + 1] = {
3545 	[IFLA_STATS_LINK_64]	= { .len = sizeof(struct rtnl_link_stats64) },
3546 };
3547 
3548 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3549 				  u32 filter_mask)
3550 {
3551 	size_t size = 0;
3552 
3553 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3554 		size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3555 
3556 	if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3557 		const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3558 
3559 		if (ops && ops->get_linkxstats_size) {
3560 			size += nla_total_size(ops->get_linkxstats_size(dev));
3561 			/* for IFLA_STATS_LINK_XSTATS */
3562 			size += nla_total_size(0);
3563 		}
3564 	}
3565 
3566 	return size;
3567 }
3568 
3569 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3570 {
3571 	struct net *net = sock_net(skb->sk);
3572 	struct net_device *dev = NULL;
3573 	int idxattr = 0, prividx = 0;
3574 	struct if_stats_msg *ifsm;
3575 	struct sk_buff *nskb;
3576 	u32 filter_mask;
3577 	int err;
3578 
3579 	ifsm = nlmsg_data(nlh);
3580 	if (ifsm->ifindex > 0)
3581 		dev = __dev_get_by_index(net, ifsm->ifindex);
3582 	else
3583 		return -EINVAL;
3584 
3585 	if (!dev)
3586 		return -ENODEV;
3587 
3588 	filter_mask = ifsm->filter_mask;
3589 	if (!filter_mask)
3590 		return -EINVAL;
3591 
3592 	nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3593 	if (!nskb)
3594 		return -ENOBUFS;
3595 
3596 	err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3597 				  NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3598 				  0, filter_mask, &idxattr, &prividx);
3599 	if (err < 0) {
3600 		/* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3601 		WARN_ON(err == -EMSGSIZE);
3602 		kfree_skb(nskb);
3603 	} else {
3604 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3605 	}
3606 
3607 	return err;
3608 }
3609 
3610 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3611 {
3612 	int h, s_h, err, s_idx, s_idxattr, s_prividx;
3613 	struct net *net = sock_net(skb->sk);
3614 	unsigned int flags = NLM_F_MULTI;
3615 	struct if_stats_msg *ifsm;
3616 	struct hlist_head *head;
3617 	struct net_device *dev;
3618 	u32 filter_mask = 0;
3619 	int idx = 0;
3620 
3621 	s_h = cb->args[0];
3622 	s_idx = cb->args[1];
3623 	s_idxattr = cb->args[2];
3624 	s_prividx = cb->args[3];
3625 
3626 	cb->seq = net->dev_base_seq;
3627 
3628 	ifsm = nlmsg_data(cb->nlh);
3629 	filter_mask = ifsm->filter_mask;
3630 	if (!filter_mask)
3631 		return -EINVAL;
3632 
3633 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3634 		idx = 0;
3635 		head = &net->dev_index_head[h];
3636 		hlist_for_each_entry(dev, head, index_hlist) {
3637 			if (idx < s_idx)
3638 				goto cont;
3639 			err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
3640 						  NETLINK_CB(cb->skb).portid,
3641 						  cb->nlh->nlmsg_seq, 0,
3642 						  flags, filter_mask,
3643 						  &s_idxattr, &s_prividx);
3644 			/* If we ran out of room on the first message,
3645 			 * we're in trouble
3646 			 */
3647 			WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
3648 
3649 			if (err < 0)
3650 				goto out;
3651 			s_prividx = 0;
3652 			s_idxattr = 0;
3653 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3654 cont:
3655 			idx++;
3656 		}
3657 	}
3658 out:
3659 	cb->args[3] = s_prividx;
3660 	cb->args[2] = s_idxattr;
3661 	cb->args[1] = idx;
3662 	cb->args[0] = h;
3663 
3664 	return skb->len;
3665 }
3666 
3667 /* Process one rtnetlink message. */
3668 
3669 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3670 {
3671 	struct net *net = sock_net(skb->sk);
3672 	rtnl_doit_func doit;
3673 	int kind;
3674 	int family;
3675 	int type;
3676 	int err;
3677 
3678 	type = nlh->nlmsg_type;
3679 	if (type > RTM_MAX)
3680 		return -EOPNOTSUPP;
3681 
3682 	type -= RTM_BASE;
3683 
3684 	/* All the messages must have at least 1 byte length */
3685 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3686 		return 0;
3687 
3688 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3689 	kind = type&3;
3690 
3691 	if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3692 		return -EPERM;
3693 
3694 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3695 		struct sock *rtnl;
3696 		rtnl_dumpit_func dumpit;
3697 		rtnl_calcit_func calcit;
3698 		u16 min_dump_alloc = 0;
3699 
3700 		dumpit = rtnl_get_dumpit(family, type);
3701 		if (dumpit == NULL)
3702 			return -EOPNOTSUPP;
3703 		calcit = rtnl_get_calcit(family, type);
3704 		if (calcit)
3705 			min_dump_alloc = calcit(skb, nlh);
3706 
3707 		__rtnl_unlock();
3708 		rtnl = net->rtnl;
3709 		{
3710 			struct netlink_dump_control c = {
3711 				.dump		= dumpit,
3712 				.min_dump_alloc	= min_dump_alloc,
3713 			};
3714 			err = netlink_dump_start(rtnl, skb, nlh, &c);
3715 		}
3716 		rtnl_lock();
3717 		return err;
3718 	}
3719 
3720 	doit = rtnl_get_doit(family, type);
3721 	if (doit == NULL)
3722 		return -EOPNOTSUPP;
3723 
3724 	return doit(skb, nlh);
3725 }
3726 
3727 static void rtnetlink_rcv(struct sk_buff *skb)
3728 {
3729 	rtnl_lock();
3730 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3731 	rtnl_unlock();
3732 }
3733 
3734 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3735 {
3736 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3737 
3738 	switch (event) {
3739 	case NETDEV_UP:
3740 	case NETDEV_DOWN:
3741 	case NETDEV_PRE_UP:
3742 	case NETDEV_POST_INIT:
3743 	case NETDEV_REGISTER:
3744 	case NETDEV_CHANGE:
3745 	case NETDEV_PRE_TYPE_CHANGE:
3746 	case NETDEV_GOING_DOWN:
3747 	case NETDEV_UNREGISTER:
3748 	case NETDEV_UNREGISTER_FINAL:
3749 	case NETDEV_RELEASE:
3750 	case NETDEV_JOIN:
3751 	case NETDEV_BONDING_INFO:
3752 		break;
3753 	default:
3754 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3755 		break;
3756 	}
3757 	return NOTIFY_DONE;
3758 }
3759 
3760 static struct notifier_block rtnetlink_dev_notifier = {
3761 	.notifier_call	= rtnetlink_event,
3762 };
3763 
3764 
3765 static int __net_init rtnetlink_net_init(struct net *net)
3766 {
3767 	struct sock *sk;
3768 	struct netlink_kernel_cfg cfg = {
3769 		.groups		= RTNLGRP_MAX,
3770 		.input		= rtnetlink_rcv,
3771 		.cb_mutex	= &rtnl_mutex,
3772 		.flags		= NL_CFG_F_NONROOT_RECV,
3773 	};
3774 
3775 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3776 	if (!sk)
3777 		return -ENOMEM;
3778 	net->rtnl = sk;
3779 	return 0;
3780 }
3781 
3782 static void __net_exit rtnetlink_net_exit(struct net *net)
3783 {
3784 	netlink_kernel_release(net->rtnl);
3785 	net->rtnl = NULL;
3786 }
3787 
3788 static struct pernet_operations rtnetlink_net_ops = {
3789 	.init = rtnetlink_net_init,
3790 	.exit = rtnetlink_net_exit,
3791 };
3792 
3793 void __init rtnetlink_init(void)
3794 {
3795 	if (register_pernet_subsys(&rtnetlink_net_ops))
3796 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
3797 
3798 	register_netdevice_notifier(&rtnetlink_dev_notifier);
3799 
3800 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3801 		      rtnl_dump_ifinfo, rtnl_calcit);
3802 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3803 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3804 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3805 
3806 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3807 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3808 
3809 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3810 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3811 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3812 
3813 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3814 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3815 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3816 
3817 	rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
3818 		      NULL);
3819 }
3820