xref: /linux/net/core/rtnetlink.c (revision c75c5ab575af7db707689cdbb5a5c458e9a034bb)
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/pci.h>
40 #include <linux/etherdevice.h>
41 
42 #include <asm/uaccess.h>
43 
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <net/arp.h>
49 #include <net/route.h>
50 #include <net/udp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 
57 struct rtnl_link {
58 	rtnl_doit_func		doit;
59 	rtnl_dumpit_func	dumpit;
60 	rtnl_calcit_func 	calcit;
61 };
62 
63 static DEFINE_MUTEX(rtnl_mutex);
64 
65 void rtnl_lock(void)
66 {
67 	mutex_lock(&rtnl_mutex);
68 }
69 EXPORT_SYMBOL(rtnl_lock);
70 
71 void __rtnl_unlock(void)
72 {
73 	mutex_unlock(&rtnl_mutex);
74 }
75 
76 void rtnl_unlock(void)
77 {
78 	/* This fellow will unlock it for us. */
79 	netdev_run_todo();
80 }
81 EXPORT_SYMBOL(rtnl_unlock);
82 
83 int rtnl_trylock(void)
84 {
85 	return mutex_trylock(&rtnl_mutex);
86 }
87 EXPORT_SYMBOL(rtnl_trylock);
88 
89 int rtnl_is_locked(void)
90 {
91 	return mutex_is_locked(&rtnl_mutex);
92 }
93 EXPORT_SYMBOL(rtnl_is_locked);
94 
95 #ifdef CONFIG_PROVE_LOCKING
96 int lockdep_rtnl_is_held(void)
97 {
98 	return lockdep_is_held(&rtnl_mutex);
99 }
100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
102 
103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
104 
105 static inline int rtm_msgindex(int msgtype)
106 {
107 	int msgindex = msgtype - RTM_BASE;
108 
109 	/*
110 	 * msgindex < 0 implies someone tried to register a netlink
111 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 	 * the message type has not been added to linux/rtnetlink.h
113 	 */
114 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
115 
116 	return msgindex;
117 }
118 
119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
120 {
121 	struct rtnl_link *tab;
122 
123 	if (protocol <= RTNL_FAMILY_MAX)
124 		tab = rtnl_msg_handlers[protocol];
125 	else
126 		tab = NULL;
127 
128 	if (tab == NULL || tab[msgindex].doit == NULL)
129 		tab = rtnl_msg_handlers[PF_UNSPEC];
130 
131 	return tab[msgindex].doit;
132 }
133 
134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
135 {
136 	struct rtnl_link *tab;
137 
138 	if (protocol <= RTNL_FAMILY_MAX)
139 		tab = rtnl_msg_handlers[protocol];
140 	else
141 		tab = NULL;
142 
143 	if (tab == NULL || tab[msgindex].dumpit == NULL)
144 		tab = rtnl_msg_handlers[PF_UNSPEC];
145 
146 	return tab[msgindex].dumpit;
147 }
148 
149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
150 {
151 	struct rtnl_link *tab;
152 
153 	if (protocol <= RTNL_FAMILY_MAX)
154 		tab = rtnl_msg_handlers[protocol];
155 	else
156 		tab = NULL;
157 
158 	if (tab == NULL || tab[msgindex].calcit == NULL)
159 		tab = rtnl_msg_handlers[PF_UNSPEC];
160 
161 	return tab[msgindex].calcit;
162 }
163 
164 /**
165  * __rtnl_register - Register a rtnetlink message type
166  * @protocol: Protocol family or PF_UNSPEC
167  * @msgtype: rtnetlink message type
168  * @doit: Function pointer called for each request message
169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170  * @calcit: Function pointer to calc size of dump message
171  *
172  * Registers the specified function pointers (at least one of them has
173  * to be non-NULL) to be called whenever a request message for the
174  * specified protocol family and message type is received.
175  *
176  * The special protocol family PF_UNSPEC may be used to define fallback
177  * function pointers for the case when no entry for the specific protocol
178  * family exists.
179  *
180  * Returns 0 on success or a negative error code.
181  */
182 int __rtnl_register(int protocol, int msgtype,
183 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
184 		    rtnl_calcit_func calcit)
185 {
186 	struct rtnl_link *tab;
187 	int msgindex;
188 
189 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
190 	msgindex = rtm_msgindex(msgtype);
191 
192 	tab = rtnl_msg_handlers[protocol];
193 	if (tab == NULL) {
194 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
195 		if (tab == NULL)
196 			return -ENOBUFS;
197 
198 		rtnl_msg_handlers[protocol] = tab;
199 	}
200 
201 	if (doit)
202 		tab[msgindex].doit = doit;
203 
204 	if (dumpit)
205 		tab[msgindex].dumpit = dumpit;
206 
207 	if (calcit)
208 		tab[msgindex].calcit = calcit;
209 
210 	return 0;
211 }
212 EXPORT_SYMBOL_GPL(__rtnl_register);
213 
214 /**
215  * rtnl_register - Register a rtnetlink message type
216  *
217  * Identical to __rtnl_register() but panics on failure. This is useful
218  * as failure of this function is very unlikely, it can only happen due
219  * to lack of memory when allocating the chain to store all message
220  * handlers for a protocol. Meant for use in init functions where lack
221  * of memory implies no sense in continuing.
222  */
223 void rtnl_register(int protocol, int msgtype,
224 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
225 		   rtnl_calcit_func calcit)
226 {
227 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
228 		panic("Unable to register rtnetlink message handler, "
229 		      "protocol = %d, message type = %d\n",
230 		      protocol, msgtype);
231 }
232 EXPORT_SYMBOL_GPL(rtnl_register);
233 
234 /**
235  * rtnl_unregister - Unregister a rtnetlink message type
236  * @protocol: Protocol family or PF_UNSPEC
237  * @msgtype: rtnetlink message type
238  *
239  * Returns 0 on success or a negative error code.
240  */
241 int rtnl_unregister(int protocol, int msgtype)
242 {
243 	int msgindex;
244 
245 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
246 	msgindex = rtm_msgindex(msgtype);
247 
248 	if (rtnl_msg_handlers[protocol] == NULL)
249 		return -ENOENT;
250 
251 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
252 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
253 
254 	return 0;
255 }
256 EXPORT_SYMBOL_GPL(rtnl_unregister);
257 
258 /**
259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260  * @protocol : Protocol family or PF_UNSPEC
261  *
262  * Identical to calling rtnl_unregster() for all registered message types
263  * of a certain protocol family.
264  */
265 void rtnl_unregister_all(int protocol)
266 {
267 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
268 
269 	kfree(rtnl_msg_handlers[protocol]);
270 	rtnl_msg_handlers[protocol] = NULL;
271 }
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
273 
274 static LIST_HEAD(link_ops);
275 
276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
277 {
278 	const struct rtnl_link_ops *ops;
279 
280 	list_for_each_entry(ops, &link_ops, list) {
281 		if (!strcmp(ops->kind, kind))
282 			return ops;
283 	}
284 	return NULL;
285 }
286 
287 /**
288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289  * @ops: struct rtnl_link_ops * to register
290  *
291  * The caller must hold the rtnl_mutex. This function should be used
292  * by drivers that create devices during module initialization. It
293  * must be called before registering the devices.
294  *
295  * Returns 0 on success or a negative error code.
296  */
297 int __rtnl_link_register(struct rtnl_link_ops *ops)
298 {
299 	if (rtnl_link_ops_get(ops->kind))
300 		return -EEXIST;
301 
302 	if (!ops->dellink)
303 		ops->dellink = unregister_netdevice_queue;
304 
305 	list_add_tail(&ops->list, &link_ops);
306 	return 0;
307 }
308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
309 
310 /**
311  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312  * @ops: struct rtnl_link_ops * to register
313  *
314  * Returns 0 on success or a negative error code.
315  */
316 int rtnl_link_register(struct rtnl_link_ops *ops)
317 {
318 	int err;
319 
320 	rtnl_lock();
321 	err = __rtnl_link_register(ops);
322 	rtnl_unlock();
323 	return err;
324 }
325 EXPORT_SYMBOL_GPL(rtnl_link_register);
326 
327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
328 {
329 	struct net_device *dev;
330 	LIST_HEAD(list_kill);
331 
332 	for_each_netdev(net, dev) {
333 		if (dev->rtnl_link_ops == ops)
334 			ops->dellink(dev, &list_kill);
335 	}
336 	unregister_netdevice_many(&list_kill);
337 }
338 
339 /**
340  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341  * @ops: struct rtnl_link_ops * to unregister
342  *
343  * The caller must hold the rtnl_mutex.
344  */
345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
346 {
347 	struct net *net;
348 
349 	for_each_net(net) {
350 		__rtnl_kill_links(net, ops);
351 	}
352 	list_del(&ops->list);
353 }
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
355 
356 /**
357  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
358  * @ops: struct rtnl_link_ops * to unregister
359  */
360 void rtnl_link_unregister(struct rtnl_link_ops *ops)
361 {
362 	rtnl_lock();
363 	__rtnl_link_unregister(ops);
364 	rtnl_unlock();
365 }
366 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
367 
368 static size_t rtnl_link_get_size(const struct net_device *dev)
369 {
370 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
371 	size_t size;
372 
373 	if (!ops)
374 		return 0;
375 
376 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
377 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
378 
379 	if (ops->get_size)
380 		/* IFLA_INFO_DATA + nested data */
381 		size += nla_total_size(sizeof(struct nlattr)) +
382 			ops->get_size(dev);
383 
384 	if (ops->get_xstats_size)
385 		/* IFLA_INFO_XSTATS */
386 		size += nla_total_size(ops->get_xstats_size(dev));
387 
388 	return size;
389 }
390 
391 static LIST_HEAD(rtnl_af_ops);
392 
393 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
394 {
395 	const struct rtnl_af_ops *ops;
396 
397 	list_for_each_entry(ops, &rtnl_af_ops, list) {
398 		if (ops->family == family)
399 			return ops;
400 	}
401 
402 	return NULL;
403 }
404 
405 /**
406  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
407  * @ops: struct rtnl_af_ops * to register
408  *
409  * The caller must hold the rtnl_mutex.
410  *
411  * Returns 0 on success or a negative error code.
412  */
413 int __rtnl_af_register(struct rtnl_af_ops *ops)
414 {
415 	list_add_tail(&ops->list, &rtnl_af_ops);
416 	return 0;
417 }
418 EXPORT_SYMBOL_GPL(__rtnl_af_register);
419 
420 /**
421  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
422  * @ops: struct rtnl_af_ops * to register
423  *
424  * Returns 0 on success or a negative error code.
425  */
426 int rtnl_af_register(struct rtnl_af_ops *ops)
427 {
428 	int err;
429 
430 	rtnl_lock();
431 	err = __rtnl_af_register(ops);
432 	rtnl_unlock();
433 	return err;
434 }
435 EXPORT_SYMBOL_GPL(rtnl_af_register);
436 
437 /**
438  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
439  * @ops: struct rtnl_af_ops * to unregister
440  *
441  * The caller must hold the rtnl_mutex.
442  */
443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
444 {
445 	list_del(&ops->list);
446 }
447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
448 
449 /**
450  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
451  * @ops: struct rtnl_af_ops * to unregister
452  */
453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
454 {
455 	rtnl_lock();
456 	__rtnl_af_unregister(ops);
457 	rtnl_unlock();
458 }
459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
460 
461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
462 {
463 	struct rtnl_af_ops *af_ops;
464 	size_t size;
465 
466 	/* IFLA_AF_SPEC */
467 	size = nla_total_size(sizeof(struct nlattr));
468 
469 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
470 		if (af_ops->get_link_af_size) {
471 			/* AF_* + nested data */
472 			size += nla_total_size(sizeof(struct nlattr)) +
473 				af_ops->get_link_af_size(dev);
474 		}
475 	}
476 
477 	return size;
478 }
479 
480 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
481 {
482 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
483 	struct nlattr *linkinfo, *data;
484 	int err = -EMSGSIZE;
485 
486 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
487 	if (linkinfo == NULL)
488 		goto out;
489 
490 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
491 		goto err_cancel_link;
492 	if (ops->fill_xstats) {
493 		err = ops->fill_xstats(skb, dev);
494 		if (err < 0)
495 			goto err_cancel_link;
496 	}
497 	if (ops->fill_info) {
498 		data = nla_nest_start(skb, IFLA_INFO_DATA);
499 		if (data == NULL) {
500 			err = -EMSGSIZE;
501 			goto err_cancel_link;
502 		}
503 		err = ops->fill_info(skb, dev);
504 		if (err < 0)
505 			goto err_cancel_data;
506 		nla_nest_end(skb, data);
507 	}
508 
509 	nla_nest_end(skb, linkinfo);
510 	return 0;
511 
512 err_cancel_data:
513 	nla_nest_cancel(skb, data);
514 err_cancel_link:
515 	nla_nest_cancel(skb, linkinfo);
516 out:
517 	return err;
518 }
519 
520 static const int rtm_min[RTM_NR_FAMILIES] =
521 {
522 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
523 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
524 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
525 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
526 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
527 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
528 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
529 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
530 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
531 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
532 };
533 
534 static const int rta_max[RTM_NR_FAMILIES] =
535 {
536 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
537 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
538 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
539 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
540 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
541 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
542 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
543 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
544 };
545 
546 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
547 {
548 	struct sock *rtnl = net->rtnl;
549 	int err = 0;
550 
551 	NETLINK_CB(skb).dst_group = group;
552 	if (echo)
553 		atomic_inc(&skb->users);
554 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
555 	if (echo)
556 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
557 	return err;
558 }
559 
560 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
561 {
562 	struct sock *rtnl = net->rtnl;
563 
564 	return nlmsg_unicast(rtnl, skb, pid);
565 }
566 EXPORT_SYMBOL(rtnl_unicast);
567 
568 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
569 		 struct nlmsghdr *nlh, gfp_t flags)
570 {
571 	struct sock *rtnl = net->rtnl;
572 	int report = 0;
573 
574 	if (nlh)
575 		report = nlmsg_report(nlh);
576 
577 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
578 }
579 EXPORT_SYMBOL(rtnl_notify);
580 
581 void rtnl_set_sk_err(struct net *net, u32 group, int error)
582 {
583 	struct sock *rtnl = net->rtnl;
584 
585 	netlink_set_err(rtnl, 0, group, error);
586 }
587 EXPORT_SYMBOL(rtnl_set_sk_err);
588 
589 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
590 {
591 	struct nlattr *mx;
592 	int i, valid = 0;
593 
594 	mx = nla_nest_start(skb, RTA_METRICS);
595 	if (mx == NULL)
596 		return -ENOBUFS;
597 
598 	for (i = 0; i < RTAX_MAX; i++) {
599 		if (metrics[i]) {
600 			valid++;
601 			if (nla_put_u32(skb, i+1, metrics[i]))
602 				goto nla_put_failure;
603 		}
604 	}
605 
606 	if (!valid) {
607 		nla_nest_cancel(skb, mx);
608 		return 0;
609 	}
610 
611 	return nla_nest_end(skb, mx);
612 
613 nla_put_failure:
614 	nla_nest_cancel(skb, mx);
615 	return -EMSGSIZE;
616 }
617 EXPORT_SYMBOL(rtnetlink_put_metrics);
618 
619 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
620 		       long expires, u32 error)
621 {
622 	struct rta_cacheinfo ci = {
623 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
624 		.rta_used = dst->__use,
625 		.rta_clntref = atomic_read(&(dst->__refcnt)),
626 		.rta_error = error,
627 		.rta_id =  id,
628 	};
629 
630 	if (expires) {
631 		unsigned long clock;
632 
633 		clock = jiffies_to_clock_t(abs(expires));
634 		clock = min_t(unsigned long, clock, INT_MAX);
635 		ci.rta_expires = (expires > 0) ? clock : -clock;
636 	}
637 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
638 }
639 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
640 
641 static void set_operstate(struct net_device *dev, unsigned char transition)
642 {
643 	unsigned char operstate = dev->operstate;
644 
645 	switch (transition) {
646 	case IF_OPER_UP:
647 		if ((operstate == IF_OPER_DORMANT ||
648 		     operstate == IF_OPER_UNKNOWN) &&
649 		    !netif_dormant(dev))
650 			operstate = IF_OPER_UP;
651 		break;
652 
653 	case IF_OPER_DORMANT:
654 		if (operstate == IF_OPER_UP ||
655 		    operstate == IF_OPER_UNKNOWN)
656 			operstate = IF_OPER_DORMANT;
657 		break;
658 	}
659 
660 	if (dev->operstate != operstate) {
661 		write_lock_bh(&dev_base_lock);
662 		dev->operstate = operstate;
663 		write_unlock_bh(&dev_base_lock);
664 		netdev_state_change(dev);
665 	}
666 }
667 
668 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
669 {
670 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
671 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
672 }
673 
674 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
675 					   const struct ifinfomsg *ifm)
676 {
677 	unsigned int flags = ifm->ifi_flags;
678 
679 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
680 	if (ifm->ifi_change)
681 		flags = (flags & ifm->ifi_change) |
682 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
683 
684 	return flags;
685 }
686 
687 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
688 				 const struct rtnl_link_stats64 *b)
689 {
690 	a->rx_packets = b->rx_packets;
691 	a->tx_packets = b->tx_packets;
692 	a->rx_bytes = b->rx_bytes;
693 	a->tx_bytes = b->tx_bytes;
694 	a->rx_errors = b->rx_errors;
695 	a->tx_errors = b->tx_errors;
696 	a->rx_dropped = b->rx_dropped;
697 	a->tx_dropped = b->tx_dropped;
698 
699 	a->multicast = b->multicast;
700 	a->collisions = b->collisions;
701 
702 	a->rx_length_errors = b->rx_length_errors;
703 	a->rx_over_errors = b->rx_over_errors;
704 	a->rx_crc_errors = b->rx_crc_errors;
705 	a->rx_frame_errors = b->rx_frame_errors;
706 	a->rx_fifo_errors = b->rx_fifo_errors;
707 	a->rx_missed_errors = b->rx_missed_errors;
708 
709 	a->tx_aborted_errors = b->tx_aborted_errors;
710 	a->tx_carrier_errors = b->tx_carrier_errors;
711 	a->tx_fifo_errors = b->tx_fifo_errors;
712 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
713 	a->tx_window_errors = b->tx_window_errors;
714 
715 	a->rx_compressed = b->rx_compressed;
716 	a->tx_compressed = b->tx_compressed;
717 }
718 
719 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
720 {
721 	memcpy(v, b, sizeof(*b));
722 }
723 
724 /* All VF info */
725 static inline int rtnl_vfinfo_size(const struct net_device *dev,
726 				   u32 ext_filter_mask)
727 {
728 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
729 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
730 		int num_vfs = dev_num_vf(dev->dev.parent);
731 		size_t size = nla_total_size(sizeof(struct nlattr));
732 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
733 		size += num_vfs *
734 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
735 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
736 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
737 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
738 		return size;
739 	} else
740 		return 0;
741 }
742 
743 static size_t rtnl_port_size(const struct net_device *dev)
744 {
745 	size_t port_size = nla_total_size(4)		/* PORT_VF */
746 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
747 		+ nla_total_size(sizeof(struct ifla_port_vsi))
748 							/* PORT_VSI_TYPE */
749 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
750 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
751 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
752 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
753 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
754 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
755 		+ port_size;
756 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
757 		+ port_size;
758 
759 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
760 		return 0;
761 	if (dev_num_vf(dev->dev.parent))
762 		return port_self_size + vf_ports_size +
763 			vf_port_size * dev_num_vf(dev->dev.parent);
764 	else
765 		return port_self_size;
766 }
767 
768 static noinline size_t if_nlmsg_size(const struct net_device *dev,
769 				     u32 ext_filter_mask)
770 {
771 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
772 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
773 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
774 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
775 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
776 	       + nla_total_size(sizeof(struct rtnl_link_stats))
777 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
778 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
779 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
780 	       + nla_total_size(4) /* IFLA_TXQLEN */
781 	       + nla_total_size(4) /* IFLA_WEIGHT */
782 	       + nla_total_size(4) /* IFLA_MTU */
783 	       + nla_total_size(4) /* IFLA_LINK */
784 	       + nla_total_size(4) /* IFLA_MASTER */
785 	       + nla_total_size(1) /* IFLA_CARRIER */
786 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
787 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
788 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
789 	       + nla_total_size(1) /* IFLA_OPERSTATE */
790 	       + nla_total_size(1) /* IFLA_LINKMODE */
791 	       + nla_total_size(ext_filter_mask
792 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
793 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
794 	       + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
795 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
796 	       + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
797 }
798 
799 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
800 {
801 	struct nlattr *vf_ports;
802 	struct nlattr *vf_port;
803 	int vf;
804 	int err;
805 
806 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
807 	if (!vf_ports)
808 		return -EMSGSIZE;
809 
810 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
811 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
812 		if (!vf_port)
813 			goto nla_put_failure;
814 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
815 			goto nla_put_failure;
816 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
817 		if (err == -EMSGSIZE)
818 			goto nla_put_failure;
819 		if (err) {
820 			nla_nest_cancel(skb, vf_port);
821 			continue;
822 		}
823 		nla_nest_end(skb, vf_port);
824 	}
825 
826 	nla_nest_end(skb, vf_ports);
827 
828 	return 0;
829 
830 nla_put_failure:
831 	nla_nest_cancel(skb, vf_ports);
832 	return -EMSGSIZE;
833 }
834 
835 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
836 {
837 	struct nlattr *port_self;
838 	int err;
839 
840 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
841 	if (!port_self)
842 		return -EMSGSIZE;
843 
844 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
845 	if (err) {
846 		nla_nest_cancel(skb, port_self);
847 		return (err == -EMSGSIZE) ? err : 0;
848 	}
849 
850 	nla_nest_end(skb, port_self);
851 
852 	return 0;
853 }
854 
855 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
856 {
857 	int err;
858 
859 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
860 		return 0;
861 
862 	err = rtnl_port_self_fill(skb, dev);
863 	if (err)
864 		return err;
865 
866 	if (dev_num_vf(dev->dev.parent)) {
867 		err = rtnl_vf_ports_fill(skb, dev);
868 		if (err)
869 			return err;
870 	}
871 
872 	return 0;
873 }
874 
875 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
876 			    int type, u32 pid, u32 seq, u32 change,
877 			    unsigned int flags, u32 ext_filter_mask)
878 {
879 	struct ifinfomsg *ifm;
880 	struct nlmsghdr *nlh;
881 	struct rtnl_link_stats64 temp;
882 	const struct rtnl_link_stats64 *stats;
883 	struct nlattr *attr, *af_spec;
884 	struct rtnl_af_ops *af_ops;
885 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
886 
887 	ASSERT_RTNL();
888 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
889 	if (nlh == NULL)
890 		return -EMSGSIZE;
891 
892 	ifm = nlmsg_data(nlh);
893 	ifm->ifi_family = AF_UNSPEC;
894 	ifm->__ifi_pad = 0;
895 	ifm->ifi_type = dev->type;
896 	ifm->ifi_index = dev->ifindex;
897 	ifm->ifi_flags = dev_get_flags(dev);
898 	ifm->ifi_change = change;
899 
900 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
901 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
902 	    nla_put_u8(skb, IFLA_OPERSTATE,
903 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
904 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
905 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
906 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
907 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
908 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
909 #ifdef CONFIG_RPS
910 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
911 #endif
912 	    (dev->ifindex != dev->iflink &&
913 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
914 	    (upper_dev &&
915 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
916 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
917 	    (dev->qdisc &&
918 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
919 	    (dev->ifalias &&
920 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
921 		goto nla_put_failure;
922 
923 	if (1) {
924 		struct rtnl_link_ifmap map = {
925 			.mem_start   = dev->mem_start,
926 			.mem_end     = dev->mem_end,
927 			.base_addr   = dev->base_addr,
928 			.irq         = dev->irq,
929 			.dma         = dev->dma,
930 			.port        = dev->if_port,
931 		};
932 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
933 			goto nla_put_failure;
934 	}
935 
936 	if (dev->addr_len) {
937 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
938 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
939 			goto nla_put_failure;
940 	}
941 
942 	attr = nla_reserve(skb, IFLA_STATS,
943 			sizeof(struct rtnl_link_stats));
944 	if (attr == NULL)
945 		goto nla_put_failure;
946 
947 	stats = dev_get_stats(dev, &temp);
948 	copy_rtnl_link_stats(nla_data(attr), stats);
949 
950 	attr = nla_reserve(skb, IFLA_STATS64,
951 			sizeof(struct rtnl_link_stats64));
952 	if (attr == NULL)
953 		goto nla_put_failure;
954 	copy_rtnl_link_stats64(nla_data(attr), stats);
955 
956 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
957 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
958 		goto nla_put_failure;
959 
960 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
961 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
962 		int i;
963 
964 		struct nlattr *vfinfo, *vf;
965 		int num_vfs = dev_num_vf(dev->dev.parent);
966 
967 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
968 		if (!vfinfo)
969 			goto nla_put_failure;
970 		for (i = 0; i < num_vfs; i++) {
971 			struct ifla_vf_info ivi;
972 			struct ifla_vf_mac vf_mac;
973 			struct ifla_vf_vlan vf_vlan;
974 			struct ifla_vf_tx_rate vf_tx_rate;
975 			struct ifla_vf_spoofchk vf_spoofchk;
976 
977 			/*
978 			 * Not all SR-IOV capable drivers support the
979 			 * spoofcheck query.  Preset to -1 so the user
980 			 * space tool can detect that the driver didn't
981 			 * report anything.
982 			 */
983 			ivi.spoofchk = -1;
984 			memset(ivi.mac, 0, sizeof(ivi.mac));
985 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
986 				break;
987 			vf_mac.vf =
988 				vf_vlan.vf =
989 				vf_tx_rate.vf =
990 				vf_spoofchk.vf = ivi.vf;
991 
992 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
993 			vf_vlan.vlan = ivi.vlan;
994 			vf_vlan.qos = ivi.qos;
995 			vf_tx_rate.rate = ivi.tx_rate;
996 			vf_spoofchk.setting = ivi.spoofchk;
997 			vf = nla_nest_start(skb, IFLA_VF_INFO);
998 			if (!vf) {
999 				nla_nest_cancel(skb, vfinfo);
1000 				goto nla_put_failure;
1001 			}
1002 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1003 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1004 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1005 				    &vf_tx_rate) ||
1006 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1007 				    &vf_spoofchk))
1008 				goto nla_put_failure;
1009 			nla_nest_end(skb, vf);
1010 		}
1011 		nla_nest_end(skb, vfinfo);
1012 	}
1013 
1014 	if (rtnl_port_fill(skb, dev))
1015 		goto nla_put_failure;
1016 
1017 	if (dev->rtnl_link_ops) {
1018 		if (rtnl_link_fill(skb, dev) < 0)
1019 			goto nla_put_failure;
1020 	}
1021 
1022 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1023 		goto nla_put_failure;
1024 
1025 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1026 		if (af_ops->fill_link_af) {
1027 			struct nlattr *af;
1028 			int err;
1029 
1030 			if (!(af = nla_nest_start(skb, af_ops->family)))
1031 				goto nla_put_failure;
1032 
1033 			err = af_ops->fill_link_af(skb, dev);
1034 
1035 			/*
1036 			 * Caller may return ENODATA to indicate that there
1037 			 * was no data to be dumped. This is not an error, it
1038 			 * means we should trim the attribute header and
1039 			 * continue.
1040 			 */
1041 			if (err == -ENODATA)
1042 				nla_nest_cancel(skb, af);
1043 			else if (err < 0)
1044 				goto nla_put_failure;
1045 
1046 			nla_nest_end(skb, af);
1047 		}
1048 	}
1049 
1050 	nla_nest_end(skb, af_spec);
1051 
1052 	return nlmsg_end(skb, nlh);
1053 
1054 nla_put_failure:
1055 	nlmsg_cancel(skb, nlh);
1056 	return -EMSGSIZE;
1057 }
1058 
1059 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1060 {
1061 	struct net *net = sock_net(skb->sk);
1062 	int h, s_h;
1063 	int idx = 0, s_idx;
1064 	struct net_device *dev;
1065 	struct hlist_head *head;
1066 	struct nlattr *tb[IFLA_MAX+1];
1067 	u32 ext_filter_mask = 0;
1068 
1069 	s_h = cb->args[0];
1070 	s_idx = cb->args[1];
1071 
1072 	rcu_read_lock();
1073 	cb->seq = net->dev_base_seq;
1074 
1075 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1076 			ifla_policy) >= 0) {
1077 
1078 		if (tb[IFLA_EXT_MASK])
1079 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1080 	}
1081 
1082 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1083 		idx = 0;
1084 		head = &net->dev_index_head[h];
1085 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
1086 			if (idx < s_idx)
1087 				goto cont;
1088 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1089 					     NETLINK_CB(cb->skb).portid,
1090 					     cb->nlh->nlmsg_seq, 0,
1091 					     NLM_F_MULTI,
1092 					     ext_filter_mask) <= 0)
1093 				goto out;
1094 
1095 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1096 cont:
1097 			idx++;
1098 		}
1099 	}
1100 out:
1101 	rcu_read_unlock();
1102 	cb->args[1] = idx;
1103 	cb->args[0] = h;
1104 
1105 	return skb->len;
1106 }
1107 
1108 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1109 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1110 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1111 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1112 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1113 	[IFLA_MTU]		= { .type = NLA_U32 },
1114 	[IFLA_LINK]		= { .type = NLA_U32 },
1115 	[IFLA_MASTER]		= { .type = NLA_U32 },
1116 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1117 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1118 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1119 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1120 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1121 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1122 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1123 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1124 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1125 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1126 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1127 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1128 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1129 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1130 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1131 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1132 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1133 };
1134 EXPORT_SYMBOL(ifla_policy);
1135 
1136 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1137 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1138 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1139 };
1140 
1141 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1142 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1143 };
1144 
1145 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1146 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1147 				    .len = sizeof(struct ifla_vf_mac) },
1148 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1149 				    .len = sizeof(struct ifla_vf_vlan) },
1150 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1151 				    .len = sizeof(struct ifla_vf_tx_rate) },
1152 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1153 				    .len = sizeof(struct ifla_vf_spoofchk) },
1154 };
1155 
1156 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1157 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1158 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1159 				    .len = PORT_PROFILE_MAX },
1160 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1161 				    .len = sizeof(struct ifla_port_vsi)},
1162 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1163 				      .len = PORT_UUID_MAX },
1164 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1165 				    .len = PORT_UUID_MAX },
1166 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1167 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1168 };
1169 
1170 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1171 {
1172 	struct net *net;
1173 	/* Examine the link attributes and figure out which
1174 	 * network namespace we are talking about.
1175 	 */
1176 	if (tb[IFLA_NET_NS_PID])
1177 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1178 	else if (tb[IFLA_NET_NS_FD])
1179 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1180 	else
1181 		net = get_net(src_net);
1182 	return net;
1183 }
1184 EXPORT_SYMBOL(rtnl_link_get_net);
1185 
1186 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1187 {
1188 	if (dev) {
1189 		if (tb[IFLA_ADDRESS] &&
1190 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1191 			return -EINVAL;
1192 
1193 		if (tb[IFLA_BROADCAST] &&
1194 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1195 			return -EINVAL;
1196 	}
1197 
1198 	if (tb[IFLA_AF_SPEC]) {
1199 		struct nlattr *af;
1200 		int rem, err;
1201 
1202 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1203 			const struct rtnl_af_ops *af_ops;
1204 
1205 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1206 				return -EAFNOSUPPORT;
1207 
1208 			if (!af_ops->set_link_af)
1209 				return -EOPNOTSUPP;
1210 
1211 			if (af_ops->validate_link_af) {
1212 				err = af_ops->validate_link_af(dev, af);
1213 				if (err < 0)
1214 					return err;
1215 			}
1216 		}
1217 	}
1218 
1219 	return 0;
1220 }
1221 
1222 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1223 {
1224 	int rem, err = -EINVAL;
1225 	struct nlattr *vf;
1226 	const struct net_device_ops *ops = dev->netdev_ops;
1227 
1228 	nla_for_each_nested(vf, attr, rem) {
1229 		switch (nla_type(vf)) {
1230 		case IFLA_VF_MAC: {
1231 			struct ifla_vf_mac *ivm;
1232 			ivm = nla_data(vf);
1233 			err = -EOPNOTSUPP;
1234 			if (ops->ndo_set_vf_mac)
1235 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1236 							  ivm->mac);
1237 			break;
1238 		}
1239 		case IFLA_VF_VLAN: {
1240 			struct ifla_vf_vlan *ivv;
1241 			ivv = nla_data(vf);
1242 			err = -EOPNOTSUPP;
1243 			if (ops->ndo_set_vf_vlan)
1244 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1245 							   ivv->vlan,
1246 							   ivv->qos);
1247 			break;
1248 		}
1249 		case IFLA_VF_TX_RATE: {
1250 			struct ifla_vf_tx_rate *ivt;
1251 			ivt = nla_data(vf);
1252 			err = -EOPNOTSUPP;
1253 			if (ops->ndo_set_vf_tx_rate)
1254 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1255 							      ivt->rate);
1256 			break;
1257 		}
1258 		case IFLA_VF_SPOOFCHK: {
1259 			struct ifla_vf_spoofchk *ivs;
1260 			ivs = nla_data(vf);
1261 			err = -EOPNOTSUPP;
1262 			if (ops->ndo_set_vf_spoofchk)
1263 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1264 							       ivs->setting);
1265 			break;
1266 		}
1267 		default:
1268 			err = -EINVAL;
1269 			break;
1270 		}
1271 		if (err)
1272 			break;
1273 	}
1274 	return err;
1275 }
1276 
1277 static int do_set_master(struct net_device *dev, int ifindex)
1278 {
1279 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1280 	const struct net_device_ops *ops;
1281 	int err;
1282 
1283 	if (upper_dev) {
1284 		if (upper_dev->ifindex == ifindex)
1285 			return 0;
1286 		ops = upper_dev->netdev_ops;
1287 		if (ops->ndo_del_slave) {
1288 			err = ops->ndo_del_slave(upper_dev, dev);
1289 			if (err)
1290 				return err;
1291 		} else {
1292 			return -EOPNOTSUPP;
1293 		}
1294 	}
1295 
1296 	if (ifindex) {
1297 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1298 		if (!upper_dev)
1299 			return -EINVAL;
1300 		ops = upper_dev->netdev_ops;
1301 		if (ops->ndo_add_slave) {
1302 			err = ops->ndo_add_slave(upper_dev, dev);
1303 			if (err)
1304 				return err;
1305 		} else {
1306 			return -EOPNOTSUPP;
1307 		}
1308 	}
1309 	return 0;
1310 }
1311 
1312 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1313 		      struct nlattr **tb, char *ifname, int modified)
1314 {
1315 	const struct net_device_ops *ops = dev->netdev_ops;
1316 	int err;
1317 
1318 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1319 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1320 		if (IS_ERR(net)) {
1321 			err = PTR_ERR(net);
1322 			goto errout;
1323 		}
1324 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1325 			err = -EPERM;
1326 			goto errout;
1327 		}
1328 		err = dev_change_net_namespace(dev, net, ifname);
1329 		put_net(net);
1330 		if (err)
1331 			goto errout;
1332 		modified = 1;
1333 	}
1334 
1335 	if (tb[IFLA_MAP]) {
1336 		struct rtnl_link_ifmap *u_map;
1337 		struct ifmap k_map;
1338 
1339 		if (!ops->ndo_set_config) {
1340 			err = -EOPNOTSUPP;
1341 			goto errout;
1342 		}
1343 
1344 		if (!netif_device_present(dev)) {
1345 			err = -ENODEV;
1346 			goto errout;
1347 		}
1348 
1349 		u_map = nla_data(tb[IFLA_MAP]);
1350 		k_map.mem_start = (unsigned long) u_map->mem_start;
1351 		k_map.mem_end = (unsigned long) u_map->mem_end;
1352 		k_map.base_addr = (unsigned short) u_map->base_addr;
1353 		k_map.irq = (unsigned char) u_map->irq;
1354 		k_map.dma = (unsigned char) u_map->dma;
1355 		k_map.port = (unsigned char) u_map->port;
1356 
1357 		err = ops->ndo_set_config(dev, &k_map);
1358 		if (err < 0)
1359 			goto errout;
1360 
1361 		modified = 1;
1362 	}
1363 
1364 	if (tb[IFLA_ADDRESS]) {
1365 		struct sockaddr *sa;
1366 		int len;
1367 
1368 		len = sizeof(sa_family_t) + dev->addr_len;
1369 		sa = kmalloc(len, GFP_KERNEL);
1370 		if (!sa) {
1371 			err = -ENOMEM;
1372 			goto errout;
1373 		}
1374 		sa->sa_family = dev->type;
1375 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1376 		       dev->addr_len);
1377 		err = dev_set_mac_address(dev, sa);
1378 		kfree(sa);
1379 		if (err)
1380 			goto errout;
1381 		modified = 1;
1382 	}
1383 
1384 	if (tb[IFLA_MTU]) {
1385 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1386 		if (err < 0)
1387 			goto errout;
1388 		modified = 1;
1389 	}
1390 
1391 	if (tb[IFLA_GROUP]) {
1392 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1393 		modified = 1;
1394 	}
1395 
1396 	/*
1397 	 * Interface selected by interface index but interface
1398 	 * name provided implies that a name change has been
1399 	 * requested.
1400 	 */
1401 	if (ifm->ifi_index > 0 && ifname[0]) {
1402 		err = dev_change_name(dev, ifname);
1403 		if (err < 0)
1404 			goto errout;
1405 		modified = 1;
1406 	}
1407 
1408 	if (tb[IFLA_IFALIAS]) {
1409 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1410 				    nla_len(tb[IFLA_IFALIAS]));
1411 		if (err < 0)
1412 			goto errout;
1413 		modified = 1;
1414 	}
1415 
1416 	if (tb[IFLA_BROADCAST]) {
1417 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1418 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1419 	}
1420 
1421 	if (ifm->ifi_flags || ifm->ifi_change) {
1422 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1423 		if (err < 0)
1424 			goto errout;
1425 	}
1426 
1427 	if (tb[IFLA_MASTER]) {
1428 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1429 		if (err)
1430 			goto errout;
1431 		modified = 1;
1432 	}
1433 
1434 	if (tb[IFLA_CARRIER]) {
1435 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1436 		if (err)
1437 			goto errout;
1438 		modified = 1;
1439 	}
1440 
1441 	if (tb[IFLA_TXQLEN])
1442 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1443 
1444 	if (tb[IFLA_OPERSTATE])
1445 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1446 
1447 	if (tb[IFLA_LINKMODE]) {
1448 		write_lock_bh(&dev_base_lock);
1449 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1450 		write_unlock_bh(&dev_base_lock);
1451 	}
1452 
1453 	if (tb[IFLA_VFINFO_LIST]) {
1454 		struct nlattr *attr;
1455 		int rem;
1456 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1457 			if (nla_type(attr) != IFLA_VF_INFO) {
1458 				err = -EINVAL;
1459 				goto errout;
1460 			}
1461 			err = do_setvfinfo(dev, attr);
1462 			if (err < 0)
1463 				goto errout;
1464 			modified = 1;
1465 		}
1466 	}
1467 	err = 0;
1468 
1469 	if (tb[IFLA_VF_PORTS]) {
1470 		struct nlattr *port[IFLA_PORT_MAX+1];
1471 		struct nlattr *attr;
1472 		int vf;
1473 		int rem;
1474 
1475 		err = -EOPNOTSUPP;
1476 		if (!ops->ndo_set_vf_port)
1477 			goto errout;
1478 
1479 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1480 			if (nla_type(attr) != IFLA_VF_PORT)
1481 				continue;
1482 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1483 				attr, ifla_port_policy);
1484 			if (err < 0)
1485 				goto errout;
1486 			if (!port[IFLA_PORT_VF]) {
1487 				err = -EOPNOTSUPP;
1488 				goto errout;
1489 			}
1490 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1491 			err = ops->ndo_set_vf_port(dev, vf, port);
1492 			if (err < 0)
1493 				goto errout;
1494 			modified = 1;
1495 		}
1496 	}
1497 	err = 0;
1498 
1499 	if (tb[IFLA_PORT_SELF]) {
1500 		struct nlattr *port[IFLA_PORT_MAX+1];
1501 
1502 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1503 			tb[IFLA_PORT_SELF], ifla_port_policy);
1504 		if (err < 0)
1505 			goto errout;
1506 
1507 		err = -EOPNOTSUPP;
1508 		if (ops->ndo_set_vf_port)
1509 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1510 		if (err < 0)
1511 			goto errout;
1512 		modified = 1;
1513 	}
1514 
1515 	if (tb[IFLA_AF_SPEC]) {
1516 		struct nlattr *af;
1517 		int rem;
1518 
1519 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1520 			const struct rtnl_af_ops *af_ops;
1521 
1522 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1523 				BUG();
1524 
1525 			err = af_ops->set_link_af(dev, af);
1526 			if (err < 0)
1527 				goto errout;
1528 
1529 			modified = 1;
1530 		}
1531 	}
1532 	err = 0;
1533 
1534 errout:
1535 	if (err < 0 && modified)
1536 		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",
1537 				     dev->name);
1538 
1539 	return err;
1540 }
1541 
1542 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1543 {
1544 	struct net *net = sock_net(skb->sk);
1545 	struct ifinfomsg *ifm;
1546 	struct net_device *dev;
1547 	int err;
1548 	struct nlattr *tb[IFLA_MAX+1];
1549 	char ifname[IFNAMSIZ];
1550 
1551 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1552 	if (err < 0)
1553 		goto errout;
1554 
1555 	if (tb[IFLA_IFNAME])
1556 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1557 	else
1558 		ifname[0] = '\0';
1559 
1560 	err = -EINVAL;
1561 	ifm = nlmsg_data(nlh);
1562 	if (ifm->ifi_index > 0)
1563 		dev = __dev_get_by_index(net, ifm->ifi_index);
1564 	else if (tb[IFLA_IFNAME])
1565 		dev = __dev_get_by_name(net, ifname);
1566 	else
1567 		goto errout;
1568 
1569 	if (dev == NULL) {
1570 		err = -ENODEV;
1571 		goto errout;
1572 	}
1573 
1574 	err = validate_linkmsg(dev, tb);
1575 	if (err < 0)
1576 		goto errout;
1577 
1578 	err = do_setlink(dev, ifm, tb, ifname, 0);
1579 errout:
1580 	return err;
1581 }
1582 
1583 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1584 {
1585 	struct net *net = sock_net(skb->sk);
1586 	const struct rtnl_link_ops *ops;
1587 	struct net_device *dev;
1588 	struct ifinfomsg *ifm;
1589 	char ifname[IFNAMSIZ];
1590 	struct nlattr *tb[IFLA_MAX+1];
1591 	int err;
1592 	LIST_HEAD(list_kill);
1593 
1594 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1595 	if (err < 0)
1596 		return err;
1597 
1598 	if (tb[IFLA_IFNAME])
1599 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1600 
1601 	ifm = nlmsg_data(nlh);
1602 	if (ifm->ifi_index > 0)
1603 		dev = __dev_get_by_index(net, ifm->ifi_index);
1604 	else if (tb[IFLA_IFNAME])
1605 		dev = __dev_get_by_name(net, ifname);
1606 	else
1607 		return -EINVAL;
1608 
1609 	if (!dev)
1610 		return -ENODEV;
1611 
1612 	ops = dev->rtnl_link_ops;
1613 	if (!ops)
1614 		return -EOPNOTSUPP;
1615 
1616 	ops->dellink(dev, &list_kill);
1617 	unregister_netdevice_many(&list_kill);
1618 	list_del(&list_kill);
1619 	return 0;
1620 }
1621 
1622 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1623 {
1624 	unsigned int old_flags;
1625 	int err;
1626 
1627 	old_flags = dev->flags;
1628 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1629 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1630 		if (err < 0)
1631 			return err;
1632 	}
1633 
1634 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1635 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1636 
1637 	__dev_notify_flags(dev, old_flags);
1638 	return 0;
1639 }
1640 EXPORT_SYMBOL(rtnl_configure_link);
1641 
1642 struct net_device *rtnl_create_link(struct net *net,
1643 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1644 {
1645 	int err;
1646 	struct net_device *dev;
1647 	unsigned int num_tx_queues = 1;
1648 	unsigned int num_rx_queues = 1;
1649 
1650 	if (tb[IFLA_NUM_TX_QUEUES])
1651 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1652 	else if (ops->get_num_tx_queues)
1653 		num_tx_queues = ops->get_num_tx_queues();
1654 
1655 	if (tb[IFLA_NUM_RX_QUEUES])
1656 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1657 	else if (ops->get_num_rx_queues)
1658 		num_rx_queues = ops->get_num_rx_queues();
1659 
1660 	err = -ENOMEM;
1661 	dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1662 			       num_tx_queues, num_rx_queues);
1663 	if (!dev)
1664 		goto err;
1665 
1666 	dev_net_set(dev, net);
1667 	dev->rtnl_link_ops = ops;
1668 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1669 
1670 	if (tb[IFLA_MTU])
1671 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1672 	if (tb[IFLA_ADDRESS]) {
1673 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1674 				nla_len(tb[IFLA_ADDRESS]));
1675 		dev->addr_assign_type = NET_ADDR_SET;
1676 	}
1677 	if (tb[IFLA_BROADCAST])
1678 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1679 				nla_len(tb[IFLA_BROADCAST]));
1680 	if (tb[IFLA_TXQLEN])
1681 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1682 	if (tb[IFLA_OPERSTATE])
1683 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1684 	if (tb[IFLA_LINKMODE])
1685 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1686 	if (tb[IFLA_GROUP])
1687 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1688 
1689 	return dev;
1690 
1691 err:
1692 	return ERR_PTR(err);
1693 }
1694 EXPORT_SYMBOL(rtnl_create_link);
1695 
1696 static int rtnl_group_changelink(struct net *net, int group,
1697 		struct ifinfomsg *ifm,
1698 		struct nlattr **tb)
1699 {
1700 	struct net_device *dev;
1701 	int err;
1702 
1703 	for_each_netdev(net, dev) {
1704 		if (dev->group == group) {
1705 			err = do_setlink(dev, ifm, tb, NULL, 0);
1706 			if (err < 0)
1707 				return err;
1708 		}
1709 	}
1710 
1711 	return 0;
1712 }
1713 
1714 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1715 {
1716 	struct net *net = sock_net(skb->sk);
1717 	const struct rtnl_link_ops *ops;
1718 	struct net_device *dev;
1719 	struct ifinfomsg *ifm;
1720 	char kind[MODULE_NAME_LEN];
1721 	char ifname[IFNAMSIZ];
1722 	struct nlattr *tb[IFLA_MAX+1];
1723 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1724 	int err;
1725 
1726 #ifdef CONFIG_MODULES
1727 replay:
1728 #endif
1729 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1730 	if (err < 0)
1731 		return err;
1732 
1733 	if (tb[IFLA_IFNAME])
1734 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1735 	else
1736 		ifname[0] = '\0';
1737 
1738 	ifm = nlmsg_data(nlh);
1739 	if (ifm->ifi_index > 0)
1740 		dev = __dev_get_by_index(net, ifm->ifi_index);
1741 	else {
1742 		if (ifname[0])
1743 			dev = __dev_get_by_name(net, ifname);
1744 		else
1745 			dev = NULL;
1746 	}
1747 
1748 	err = validate_linkmsg(dev, tb);
1749 	if (err < 0)
1750 		return err;
1751 
1752 	if (tb[IFLA_LINKINFO]) {
1753 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1754 				       tb[IFLA_LINKINFO], ifla_info_policy);
1755 		if (err < 0)
1756 			return err;
1757 	} else
1758 		memset(linkinfo, 0, sizeof(linkinfo));
1759 
1760 	if (linkinfo[IFLA_INFO_KIND]) {
1761 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1762 		ops = rtnl_link_ops_get(kind);
1763 	} else {
1764 		kind[0] = '\0';
1765 		ops = NULL;
1766 	}
1767 
1768 	if (1) {
1769 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1770 		struct net *dest_net;
1771 
1772 		if (ops) {
1773 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1774 				err = nla_parse_nested(attr, ops->maxtype,
1775 						       linkinfo[IFLA_INFO_DATA],
1776 						       ops->policy);
1777 				if (err < 0)
1778 					return err;
1779 				data = attr;
1780 			}
1781 			if (ops->validate) {
1782 				err = ops->validate(tb, data);
1783 				if (err < 0)
1784 					return err;
1785 			}
1786 		}
1787 
1788 		if (dev) {
1789 			int modified = 0;
1790 
1791 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1792 				return -EEXIST;
1793 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1794 				return -EOPNOTSUPP;
1795 
1796 			if (linkinfo[IFLA_INFO_DATA]) {
1797 				if (!ops || ops != dev->rtnl_link_ops ||
1798 				    !ops->changelink)
1799 					return -EOPNOTSUPP;
1800 
1801 				err = ops->changelink(dev, tb, data);
1802 				if (err < 0)
1803 					return err;
1804 				modified = 1;
1805 			}
1806 
1807 			return do_setlink(dev, ifm, tb, ifname, modified);
1808 		}
1809 
1810 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1811 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1812 				return rtnl_group_changelink(net,
1813 						nla_get_u32(tb[IFLA_GROUP]),
1814 						ifm, tb);
1815 			return -ENODEV;
1816 		}
1817 
1818 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1819 			return -EOPNOTSUPP;
1820 
1821 		if (!ops) {
1822 #ifdef CONFIG_MODULES
1823 			if (kind[0]) {
1824 				__rtnl_unlock();
1825 				request_module("rtnl-link-%s", kind);
1826 				rtnl_lock();
1827 				ops = rtnl_link_ops_get(kind);
1828 				if (ops)
1829 					goto replay;
1830 			}
1831 #endif
1832 			return -EOPNOTSUPP;
1833 		}
1834 
1835 		if (!ifname[0])
1836 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1837 
1838 		dest_net = rtnl_link_get_net(net, tb);
1839 		if (IS_ERR(dest_net))
1840 			return PTR_ERR(dest_net);
1841 
1842 		dev = rtnl_create_link(dest_net, ifname, ops, tb);
1843 		if (IS_ERR(dev)) {
1844 			err = PTR_ERR(dev);
1845 			goto out;
1846 		}
1847 
1848 		dev->ifindex = ifm->ifi_index;
1849 
1850 		if (ops->newlink)
1851 			err = ops->newlink(net, dev, tb, data);
1852 		else
1853 			err = register_netdevice(dev);
1854 
1855 		if (err < 0 && !IS_ERR(dev))
1856 			free_netdev(dev);
1857 		if (err < 0)
1858 			goto out;
1859 
1860 		err = rtnl_configure_link(dev, ifm);
1861 		if (err < 0)
1862 			unregister_netdevice(dev);
1863 out:
1864 		put_net(dest_net);
1865 		return err;
1866 	}
1867 }
1868 
1869 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1870 {
1871 	struct net *net = sock_net(skb->sk);
1872 	struct ifinfomsg *ifm;
1873 	char ifname[IFNAMSIZ];
1874 	struct nlattr *tb[IFLA_MAX+1];
1875 	struct net_device *dev = NULL;
1876 	struct sk_buff *nskb;
1877 	int err;
1878 	u32 ext_filter_mask = 0;
1879 
1880 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1881 	if (err < 0)
1882 		return err;
1883 
1884 	if (tb[IFLA_IFNAME])
1885 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1886 
1887 	if (tb[IFLA_EXT_MASK])
1888 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1889 
1890 	ifm = nlmsg_data(nlh);
1891 	if (ifm->ifi_index > 0)
1892 		dev = __dev_get_by_index(net, ifm->ifi_index);
1893 	else if (tb[IFLA_IFNAME])
1894 		dev = __dev_get_by_name(net, ifname);
1895 	else
1896 		return -EINVAL;
1897 
1898 	if (dev == NULL)
1899 		return -ENODEV;
1900 
1901 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1902 	if (nskb == NULL)
1903 		return -ENOBUFS;
1904 
1905 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1906 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1907 	if (err < 0) {
1908 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1909 		WARN_ON(err == -EMSGSIZE);
1910 		kfree_skb(nskb);
1911 	} else
1912 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1913 
1914 	return err;
1915 }
1916 
1917 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1918 {
1919 	struct net *net = sock_net(skb->sk);
1920 	struct net_device *dev;
1921 	struct nlattr *tb[IFLA_MAX+1];
1922 	u32 ext_filter_mask = 0;
1923 	u16 min_ifinfo_dump_size = 0;
1924 
1925 	if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1926 			ifla_policy) >= 0) {
1927 		if (tb[IFLA_EXT_MASK])
1928 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1929 	}
1930 
1931 	if (!ext_filter_mask)
1932 		return NLMSG_GOODSIZE;
1933 	/*
1934 	 * traverse the list of net devices and compute the minimum
1935 	 * buffer size based upon the filter mask.
1936 	 */
1937 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1938 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1939 					     if_nlmsg_size(dev,
1940 						           ext_filter_mask));
1941 	}
1942 
1943 	return min_ifinfo_dump_size;
1944 }
1945 
1946 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1947 {
1948 	int idx;
1949 	int s_idx = cb->family;
1950 
1951 	if (s_idx == 0)
1952 		s_idx = 1;
1953 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1954 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1955 		if (idx < s_idx || idx == PF_PACKET)
1956 			continue;
1957 		if (rtnl_msg_handlers[idx] == NULL ||
1958 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1959 			continue;
1960 		if (idx > s_idx)
1961 			memset(&cb->args[0], 0, sizeof(cb->args));
1962 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1963 			break;
1964 	}
1965 	cb->family = idx;
1966 
1967 	return skb->len;
1968 }
1969 
1970 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1971 {
1972 	struct net *net = dev_net(dev);
1973 	struct sk_buff *skb;
1974 	int err = -ENOBUFS;
1975 	size_t if_info_size;
1976 
1977 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1978 	if (skb == NULL)
1979 		goto errout;
1980 
1981 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1982 	if (err < 0) {
1983 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1984 		WARN_ON(err == -EMSGSIZE);
1985 		kfree_skb(skb);
1986 		goto errout;
1987 	}
1988 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1989 	return;
1990 errout:
1991 	if (err < 0)
1992 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1993 }
1994 EXPORT_SYMBOL(rtmsg_ifinfo);
1995 
1996 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1997 				   struct net_device *dev,
1998 				   u8 *addr, u32 pid, u32 seq,
1999 				   int type, unsigned int flags)
2000 {
2001 	struct nlmsghdr *nlh;
2002 	struct ndmsg *ndm;
2003 
2004 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
2005 	if (!nlh)
2006 		return -EMSGSIZE;
2007 
2008 	ndm = nlmsg_data(nlh);
2009 	ndm->ndm_family  = AF_BRIDGE;
2010 	ndm->ndm_pad1	 = 0;
2011 	ndm->ndm_pad2    = 0;
2012 	ndm->ndm_flags	 = flags;
2013 	ndm->ndm_type	 = 0;
2014 	ndm->ndm_ifindex = dev->ifindex;
2015 	ndm->ndm_state   = NUD_PERMANENT;
2016 
2017 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2018 		goto nla_put_failure;
2019 
2020 	return nlmsg_end(skb, nlh);
2021 
2022 nla_put_failure:
2023 	nlmsg_cancel(skb, nlh);
2024 	return -EMSGSIZE;
2025 }
2026 
2027 static inline size_t rtnl_fdb_nlmsg_size(void)
2028 {
2029 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2030 }
2031 
2032 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2033 {
2034 	struct net *net = dev_net(dev);
2035 	struct sk_buff *skb;
2036 	int err = -ENOBUFS;
2037 
2038 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2039 	if (!skb)
2040 		goto errout;
2041 
2042 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2043 	if (err < 0) {
2044 		kfree_skb(skb);
2045 		goto errout;
2046 	}
2047 
2048 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2049 	return;
2050 errout:
2051 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2052 }
2053 
2054 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2055 {
2056 	struct net *net = sock_net(skb->sk);
2057 	struct ndmsg *ndm;
2058 	struct nlattr *tb[NDA_MAX+1];
2059 	struct net_device *dev;
2060 	u8 *addr;
2061 	int err;
2062 
2063 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2064 	if (err < 0)
2065 		return err;
2066 
2067 	ndm = nlmsg_data(nlh);
2068 	if (ndm->ndm_ifindex == 0) {
2069 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2070 		return -EINVAL;
2071 	}
2072 
2073 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2074 	if (dev == NULL) {
2075 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2076 		return -ENODEV;
2077 	}
2078 
2079 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2080 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2081 		return -EINVAL;
2082 	}
2083 
2084 	addr = nla_data(tb[NDA_LLADDR]);
2085 	if (!is_valid_ether_addr(addr)) {
2086 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2087 		return -EINVAL;
2088 	}
2089 
2090 	err = -EOPNOTSUPP;
2091 
2092 	/* Support fdb on master device the net/bridge default case */
2093 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2094 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2095 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2096 		const struct net_device_ops *ops = br_dev->netdev_ops;
2097 
2098 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2099 		if (err)
2100 			goto out;
2101 		else
2102 			ndm->ndm_flags &= ~NTF_MASTER;
2103 	}
2104 
2105 	/* Embedded bridge, macvlan, and any other device support */
2106 	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2107 		err = dev->netdev_ops->ndo_fdb_add(ndm, tb,
2108 						   dev, addr,
2109 						   nlh->nlmsg_flags);
2110 
2111 		if (!err) {
2112 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2113 			ndm->ndm_flags &= ~NTF_SELF;
2114 		}
2115 	}
2116 out:
2117 	return err;
2118 }
2119 
2120 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2121 {
2122 	struct net *net = sock_net(skb->sk);
2123 	struct ndmsg *ndm;
2124 	struct nlattr *tb[NDA_MAX+1];
2125 	struct net_device *dev;
2126 	int err = -EINVAL;
2127 	__u8 *addr;
2128 
2129 	if (!capable(CAP_NET_ADMIN))
2130 		return -EPERM;
2131 
2132 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2133 	if (err < 0)
2134 		return err;
2135 
2136 	ndm = nlmsg_data(nlh);
2137 	if (ndm->ndm_ifindex == 0) {
2138 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2139 		return -EINVAL;
2140 	}
2141 
2142 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2143 	if (dev == NULL) {
2144 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2145 		return -ENODEV;
2146 	}
2147 
2148 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2149 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2150 		return -EINVAL;
2151 	}
2152 
2153 	addr = nla_data(tb[NDA_LLADDR]);
2154 	if (!is_valid_ether_addr(addr)) {
2155 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2156 		return -EINVAL;
2157 	}
2158 
2159 	err = -EOPNOTSUPP;
2160 
2161 	/* Support fdb on master device the net/bridge default case */
2162 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2163 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2164 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2165 		const struct net_device_ops *ops = br_dev->netdev_ops;
2166 
2167 		if (ops->ndo_fdb_del)
2168 			err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2169 
2170 		if (err)
2171 			goto out;
2172 		else
2173 			ndm->ndm_flags &= ~NTF_MASTER;
2174 	}
2175 
2176 	/* Embedded bridge, macvlan, and any other device support */
2177 	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2178 		err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2179 
2180 		if (!err) {
2181 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2182 			ndm->ndm_flags &= ~NTF_SELF;
2183 		}
2184 	}
2185 out:
2186 	return err;
2187 }
2188 
2189 static int nlmsg_populate_fdb(struct sk_buff *skb,
2190 			      struct netlink_callback *cb,
2191 			      struct net_device *dev,
2192 			      int *idx,
2193 			      struct netdev_hw_addr_list *list)
2194 {
2195 	struct netdev_hw_addr *ha;
2196 	int err;
2197 	u32 portid, seq;
2198 
2199 	portid = NETLINK_CB(cb->skb).portid;
2200 	seq = cb->nlh->nlmsg_seq;
2201 
2202 	list_for_each_entry(ha, &list->list, list) {
2203 		if (*idx < cb->args[0])
2204 			goto skip;
2205 
2206 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2207 					      portid, seq,
2208 					      RTM_NEWNEIGH, NTF_SELF);
2209 		if (err < 0)
2210 			return err;
2211 skip:
2212 		*idx += 1;
2213 	}
2214 	return 0;
2215 }
2216 
2217 /**
2218  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2219  * @nlh: netlink message header
2220  * @dev: netdevice
2221  *
2222  * Default netdevice operation to dump the existing unicast address list.
2223  * Returns zero on success.
2224  */
2225 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2226 		      struct netlink_callback *cb,
2227 		      struct net_device *dev,
2228 		      int idx)
2229 {
2230 	int err;
2231 
2232 	netif_addr_lock_bh(dev);
2233 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2234 	if (err)
2235 		goto out;
2236 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2237 out:
2238 	netif_addr_unlock_bh(dev);
2239 	return idx;
2240 }
2241 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2242 
2243 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2244 {
2245 	int idx = 0;
2246 	struct net *net = sock_net(skb->sk);
2247 	struct net_device *dev;
2248 
2249 	rcu_read_lock();
2250 	for_each_netdev_rcu(net, dev) {
2251 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2252 			struct net_device *br_dev;
2253 			const struct net_device_ops *ops;
2254 
2255 			br_dev = netdev_master_upper_dev_get(dev);
2256 			ops = br_dev->netdev_ops;
2257 			if (ops->ndo_fdb_dump)
2258 				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2259 		}
2260 
2261 		if (dev->netdev_ops->ndo_fdb_dump)
2262 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2263 	}
2264 	rcu_read_unlock();
2265 
2266 	cb->args[0] = idx;
2267 	return skb->len;
2268 }
2269 
2270 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2271 			    struct net_device *dev, u16 mode)
2272 {
2273 	struct nlmsghdr *nlh;
2274 	struct ifinfomsg *ifm;
2275 	struct nlattr *br_afspec;
2276 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2277 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2278 
2279 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2280 	if (nlh == NULL)
2281 		return -EMSGSIZE;
2282 
2283 	ifm = nlmsg_data(nlh);
2284 	ifm->ifi_family = AF_BRIDGE;
2285 	ifm->__ifi_pad = 0;
2286 	ifm->ifi_type = dev->type;
2287 	ifm->ifi_index = dev->ifindex;
2288 	ifm->ifi_flags = dev_get_flags(dev);
2289 	ifm->ifi_change = 0;
2290 
2291 
2292 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2293 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2294 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2295 	    (br_dev &&
2296 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2297 	    (dev->addr_len &&
2298 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2299 	    (dev->ifindex != dev->iflink &&
2300 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2301 		goto nla_put_failure;
2302 
2303 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2304 	if (!br_afspec)
2305 		goto nla_put_failure;
2306 
2307 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2308 	    nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2309 		nla_nest_cancel(skb, br_afspec);
2310 		goto nla_put_failure;
2311 	}
2312 	nla_nest_end(skb, br_afspec);
2313 
2314 	return nlmsg_end(skb, nlh);
2315 nla_put_failure:
2316 	nlmsg_cancel(skb, nlh);
2317 	return -EMSGSIZE;
2318 }
2319 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2320 
2321 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2322 {
2323 	struct net *net = sock_net(skb->sk);
2324 	struct net_device *dev;
2325 	int idx = 0;
2326 	u32 portid = NETLINK_CB(cb->skb).portid;
2327 	u32 seq = cb->nlh->nlmsg_seq;
2328 	struct nlattr *extfilt;
2329 	u32 filter_mask = 0;
2330 
2331 	extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct rtgenmsg),
2332 				  IFLA_EXT_MASK);
2333 	if (extfilt)
2334 		filter_mask = nla_get_u32(extfilt);
2335 
2336 	rcu_read_lock();
2337 	for_each_netdev_rcu(net, dev) {
2338 		const struct net_device_ops *ops = dev->netdev_ops;
2339 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2340 
2341 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2342 			if (idx >= cb->args[0] &&
2343 			    br_dev->netdev_ops->ndo_bridge_getlink(
2344 				    skb, portid, seq, dev, filter_mask) < 0)
2345 				break;
2346 			idx++;
2347 		}
2348 
2349 		if (ops->ndo_bridge_getlink) {
2350 			if (idx >= cb->args[0] &&
2351 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2352 						    filter_mask) < 0)
2353 				break;
2354 			idx++;
2355 		}
2356 	}
2357 	rcu_read_unlock();
2358 	cb->args[0] = idx;
2359 
2360 	return skb->len;
2361 }
2362 
2363 static inline size_t bridge_nlmsg_size(void)
2364 {
2365 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2366 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2367 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2368 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2369 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2370 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2371 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2372 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2373 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2374 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2375 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2376 }
2377 
2378 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2379 {
2380 	struct net *net = dev_net(dev);
2381 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2382 	struct sk_buff *skb;
2383 	int err = -EOPNOTSUPP;
2384 
2385 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2386 	if (!skb) {
2387 		err = -ENOMEM;
2388 		goto errout;
2389 	}
2390 
2391 	if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2392 	    br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2393 		err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2394 		if (err < 0)
2395 			goto errout;
2396 	}
2397 
2398 	if ((flags & BRIDGE_FLAGS_SELF) &&
2399 	    dev->netdev_ops->ndo_bridge_getlink) {
2400 		err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2401 		if (err < 0)
2402 			goto errout;
2403 	}
2404 
2405 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2406 	return 0;
2407 errout:
2408 	WARN_ON(err == -EMSGSIZE);
2409 	kfree_skb(skb);
2410 	rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2411 	return err;
2412 }
2413 
2414 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2415 			       void *arg)
2416 {
2417 	struct net *net = sock_net(skb->sk);
2418 	struct ifinfomsg *ifm;
2419 	struct net_device *dev;
2420 	struct nlattr *br_spec, *attr = NULL;
2421 	int rem, err = -EOPNOTSUPP;
2422 	u16 oflags, flags = 0;
2423 	bool have_flags = false;
2424 
2425 	if (nlmsg_len(nlh) < sizeof(*ifm))
2426 		return -EINVAL;
2427 
2428 	ifm = nlmsg_data(nlh);
2429 	if (ifm->ifi_family != AF_BRIDGE)
2430 		return -EPFNOSUPPORT;
2431 
2432 	dev = __dev_get_by_index(net, ifm->ifi_index);
2433 	if (!dev) {
2434 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2435 		return -ENODEV;
2436 	}
2437 
2438 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2439 	if (br_spec) {
2440 		nla_for_each_nested(attr, br_spec, rem) {
2441 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2442 				have_flags = true;
2443 				flags = nla_get_u16(attr);
2444 				break;
2445 			}
2446 		}
2447 	}
2448 
2449 	oflags = flags;
2450 
2451 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2452 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2453 
2454 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2455 			err = -EOPNOTSUPP;
2456 			goto out;
2457 		}
2458 
2459 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2460 		if (err)
2461 			goto out;
2462 
2463 		flags &= ~BRIDGE_FLAGS_MASTER;
2464 	}
2465 
2466 	if ((flags & BRIDGE_FLAGS_SELF)) {
2467 		if (!dev->netdev_ops->ndo_bridge_setlink)
2468 			err = -EOPNOTSUPP;
2469 		else
2470 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2471 
2472 		if (!err)
2473 			flags &= ~BRIDGE_FLAGS_SELF;
2474 	}
2475 
2476 	if (have_flags)
2477 		memcpy(nla_data(attr), &flags, sizeof(flags));
2478 	/* Generate event to notify upper layer of bridge change */
2479 	if (!err)
2480 		err = rtnl_bridge_notify(dev, oflags);
2481 out:
2482 	return err;
2483 }
2484 
2485 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2486 			       void *arg)
2487 {
2488 	struct net *net = sock_net(skb->sk);
2489 	struct ifinfomsg *ifm;
2490 	struct net_device *dev;
2491 	struct nlattr *br_spec, *attr = NULL;
2492 	int rem, err = -EOPNOTSUPP;
2493 	u16 oflags, flags = 0;
2494 	bool have_flags = false;
2495 
2496 	if (nlmsg_len(nlh) < sizeof(*ifm))
2497 		return -EINVAL;
2498 
2499 	ifm = nlmsg_data(nlh);
2500 	if (ifm->ifi_family != AF_BRIDGE)
2501 		return -EPFNOSUPPORT;
2502 
2503 	dev = __dev_get_by_index(net, ifm->ifi_index);
2504 	if (!dev) {
2505 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2506 		return -ENODEV;
2507 	}
2508 
2509 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2510 	if (br_spec) {
2511 		nla_for_each_nested(attr, br_spec, rem) {
2512 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2513 				have_flags = true;
2514 				flags = nla_get_u16(attr);
2515 				break;
2516 			}
2517 		}
2518 	}
2519 
2520 	oflags = flags;
2521 
2522 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2523 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2524 
2525 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2526 			err = -EOPNOTSUPP;
2527 			goto out;
2528 		}
2529 
2530 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2531 		if (err)
2532 			goto out;
2533 
2534 		flags &= ~BRIDGE_FLAGS_MASTER;
2535 	}
2536 
2537 	if ((flags & BRIDGE_FLAGS_SELF)) {
2538 		if (!dev->netdev_ops->ndo_bridge_dellink)
2539 			err = -EOPNOTSUPP;
2540 		else
2541 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2542 
2543 		if (!err)
2544 			flags &= ~BRIDGE_FLAGS_SELF;
2545 	}
2546 
2547 	if (have_flags)
2548 		memcpy(nla_data(attr), &flags, sizeof(flags));
2549 	/* Generate event to notify upper layer of bridge change */
2550 	if (!err)
2551 		err = rtnl_bridge_notify(dev, oflags);
2552 out:
2553 	return err;
2554 }
2555 
2556 /* Protected by RTNL sempahore.  */
2557 static struct rtattr **rta_buf;
2558 static int rtattr_max;
2559 
2560 /* Process one rtnetlink message. */
2561 
2562 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2563 {
2564 	struct net *net = sock_net(skb->sk);
2565 	rtnl_doit_func doit;
2566 	int sz_idx, kind;
2567 	int min_len;
2568 	int family;
2569 	int type;
2570 	int err;
2571 
2572 	type = nlh->nlmsg_type;
2573 	if (type > RTM_MAX)
2574 		return -EOPNOTSUPP;
2575 
2576 	type -= RTM_BASE;
2577 
2578 	/* All the messages must have at least 1 byte length */
2579 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2580 		return 0;
2581 
2582 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2583 	sz_idx = type>>2;
2584 	kind = type&3;
2585 
2586 	if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2587 		return -EPERM;
2588 
2589 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2590 		struct sock *rtnl;
2591 		rtnl_dumpit_func dumpit;
2592 		rtnl_calcit_func calcit;
2593 		u16 min_dump_alloc = 0;
2594 
2595 		dumpit = rtnl_get_dumpit(family, type);
2596 		if (dumpit == NULL)
2597 			return -EOPNOTSUPP;
2598 		calcit = rtnl_get_calcit(family, type);
2599 		if (calcit)
2600 			min_dump_alloc = calcit(skb, nlh);
2601 
2602 		__rtnl_unlock();
2603 		rtnl = net->rtnl;
2604 		{
2605 			struct netlink_dump_control c = {
2606 				.dump		= dumpit,
2607 				.min_dump_alloc	= min_dump_alloc,
2608 			};
2609 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2610 		}
2611 		rtnl_lock();
2612 		return err;
2613 	}
2614 
2615 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2616 
2617 	min_len = rtm_min[sz_idx];
2618 	if (nlh->nlmsg_len < min_len)
2619 		return -EINVAL;
2620 
2621 	if (nlh->nlmsg_len > min_len) {
2622 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2623 		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2624 
2625 		while (RTA_OK(attr, attrlen)) {
2626 			unsigned int flavor = attr->rta_type & NLA_TYPE_MASK;
2627 			if (flavor) {
2628 				if (flavor > rta_max[sz_idx])
2629 					return -EINVAL;
2630 				rta_buf[flavor-1] = attr;
2631 			}
2632 			attr = RTA_NEXT(attr, attrlen);
2633 		}
2634 	}
2635 
2636 	doit = rtnl_get_doit(family, type);
2637 	if (doit == NULL)
2638 		return -EOPNOTSUPP;
2639 
2640 	return doit(skb, nlh, (void *)&rta_buf[0]);
2641 }
2642 
2643 static void rtnetlink_rcv(struct sk_buff *skb)
2644 {
2645 	rtnl_lock();
2646 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2647 	rtnl_unlock();
2648 }
2649 
2650 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2651 {
2652 	struct net_device *dev = ptr;
2653 
2654 	switch (event) {
2655 	case NETDEV_UP:
2656 	case NETDEV_DOWN:
2657 	case NETDEV_PRE_UP:
2658 	case NETDEV_POST_INIT:
2659 	case NETDEV_REGISTER:
2660 	case NETDEV_CHANGE:
2661 	case NETDEV_PRE_TYPE_CHANGE:
2662 	case NETDEV_GOING_DOWN:
2663 	case NETDEV_UNREGISTER:
2664 	case NETDEV_UNREGISTER_FINAL:
2665 	case NETDEV_RELEASE:
2666 	case NETDEV_JOIN:
2667 		break;
2668 	default:
2669 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2670 		break;
2671 	}
2672 	return NOTIFY_DONE;
2673 }
2674 
2675 static struct notifier_block rtnetlink_dev_notifier = {
2676 	.notifier_call	= rtnetlink_event,
2677 };
2678 
2679 
2680 static int __net_init rtnetlink_net_init(struct net *net)
2681 {
2682 	struct sock *sk;
2683 	struct netlink_kernel_cfg cfg = {
2684 		.groups		= RTNLGRP_MAX,
2685 		.input		= rtnetlink_rcv,
2686 		.cb_mutex	= &rtnl_mutex,
2687 		.flags		= NL_CFG_F_NONROOT_RECV,
2688 	};
2689 
2690 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2691 	if (!sk)
2692 		return -ENOMEM;
2693 	net->rtnl = sk;
2694 	return 0;
2695 }
2696 
2697 static void __net_exit rtnetlink_net_exit(struct net *net)
2698 {
2699 	netlink_kernel_release(net->rtnl);
2700 	net->rtnl = NULL;
2701 }
2702 
2703 static struct pernet_operations rtnetlink_net_ops = {
2704 	.init = rtnetlink_net_init,
2705 	.exit = rtnetlink_net_exit,
2706 };
2707 
2708 void __init rtnetlink_init(void)
2709 {
2710 	int i;
2711 
2712 	rtattr_max = 0;
2713 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2714 		if (rta_max[i] > rtattr_max)
2715 			rtattr_max = rta_max[i];
2716 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2717 	if (!rta_buf)
2718 		panic("rtnetlink_init: cannot allocate rta_buf\n");
2719 
2720 	if (register_pernet_subsys(&rtnetlink_net_ops))
2721 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2722 
2723 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2724 
2725 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2726 		      rtnl_dump_ifinfo, rtnl_calcit);
2727 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2728 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2729 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2730 
2731 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2732 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2733 
2734 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2735 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2736 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2737 
2738 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2739 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2740 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2741 }
2742 
2743