xref: /linux/net/core/rtnetlink.c (revision b889fcf63cb62e7fdb7816565e28f44dbe4a76a5)
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 			goto err_cancel_link;
501 		err = ops->fill_info(skb, dev);
502 		if (err < 0)
503 			goto err_cancel_data;
504 		nla_nest_end(skb, data);
505 	}
506 
507 	nla_nest_end(skb, linkinfo);
508 	return 0;
509 
510 err_cancel_data:
511 	nla_nest_cancel(skb, data);
512 err_cancel_link:
513 	nla_nest_cancel(skb, linkinfo);
514 out:
515 	return err;
516 }
517 
518 static const int rtm_min[RTM_NR_FAMILIES] =
519 {
520 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
521 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
522 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
523 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
524 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
525 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
526 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
527 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
528 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
529 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
530 };
531 
532 static const int rta_max[RTM_NR_FAMILIES] =
533 {
534 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
535 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
536 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
537 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
538 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
539 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
540 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
541 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
542 };
543 
544 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
545 {
546 	struct sock *rtnl = net->rtnl;
547 	int err = 0;
548 
549 	NETLINK_CB(skb).dst_group = group;
550 	if (echo)
551 		atomic_inc(&skb->users);
552 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
553 	if (echo)
554 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
555 	return err;
556 }
557 
558 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
559 {
560 	struct sock *rtnl = net->rtnl;
561 
562 	return nlmsg_unicast(rtnl, skb, pid);
563 }
564 EXPORT_SYMBOL(rtnl_unicast);
565 
566 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
567 		 struct nlmsghdr *nlh, gfp_t flags)
568 {
569 	struct sock *rtnl = net->rtnl;
570 	int report = 0;
571 
572 	if (nlh)
573 		report = nlmsg_report(nlh);
574 
575 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
576 }
577 EXPORT_SYMBOL(rtnl_notify);
578 
579 void rtnl_set_sk_err(struct net *net, u32 group, int error)
580 {
581 	struct sock *rtnl = net->rtnl;
582 
583 	netlink_set_err(rtnl, 0, group, error);
584 }
585 EXPORT_SYMBOL(rtnl_set_sk_err);
586 
587 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
588 {
589 	struct nlattr *mx;
590 	int i, valid = 0;
591 
592 	mx = nla_nest_start(skb, RTA_METRICS);
593 	if (mx == NULL)
594 		return -ENOBUFS;
595 
596 	for (i = 0; i < RTAX_MAX; i++) {
597 		if (metrics[i]) {
598 			valid++;
599 			if (nla_put_u32(skb, i+1, metrics[i]))
600 				goto nla_put_failure;
601 		}
602 	}
603 
604 	if (!valid) {
605 		nla_nest_cancel(skb, mx);
606 		return 0;
607 	}
608 
609 	return nla_nest_end(skb, mx);
610 
611 nla_put_failure:
612 	nla_nest_cancel(skb, mx);
613 	return -EMSGSIZE;
614 }
615 EXPORT_SYMBOL(rtnetlink_put_metrics);
616 
617 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
618 		       long expires, u32 error)
619 {
620 	struct rta_cacheinfo ci = {
621 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
622 		.rta_used = dst->__use,
623 		.rta_clntref = atomic_read(&(dst->__refcnt)),
624 		.rta_error = error,
625 		.rta_id =  id,
626 	};
627 
628 	if (expires) {
629 		unsigned long clock;
630 
631 		clock = jiffies_to_clock_t(abs(expires));
632 		clock = min_t(unsigned long, clock, INT_MAX);
633 		ci.rta_expires = (expires > 0) ? clock : -clock;
634 	}
635 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
636 }
637 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
638 
639 static void set_operstate(struct net_device *dev, unsigned char transition)
640 {
641 	unsigned char operstate = dev->operstate;
642 
643 	switch (transition) {
644 	case IF_OPER_UP:
645 		if ((operstate == IF_OPER_DORMANT ||
646 		     operstate == IF_OPER_UNKNOWN) &&
647 		    !netif_dormant(dev))
648 			operstate = IF_OPER_UP;
649 		break;
650 
651 	case IF_OPER_DORMANT:
652 		if (operstate == IF_OPER_UP ||
653 		    operstate == IF_OPER_UNKNOWN)
654 			operstate = IF_OPER_DORMANT;
655 		break;
656 	}
657 
658 	if (dev->operstate != operstate) {
659 		write_lock_bh(&dev_base_lock);
660 		dev->operstate = operstate;
661 		write_unlock_bh(&dev_base_lock);
662 		netdev_state_change(dev);
663 	}
664 }
665 
666 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
667 {
668 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
669 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
670 }
671 
672 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
673 					   const struct ifinfomsg *ifm)
674 {
675 	unsigned int flags = ifm->ifi_flags;
676 
677 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
678 	if (ifm->ifi_change)
679 		flags = (flags & ifm->ifi_change) |
680 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
681 
682 	return flags;
683 }
684 
685 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
686 				 const struct rtnl_link_stats64 *b)
687 {
688 	a->rx_packets = b->rx_packets;
689 	a->tx_packets = b->tx_packets;
690 	a->rx_bytes = b->rx_bytes;
691 	a->tx_bytes = b->tx_bytes;
692 	a->rx_errors = b->rx_errors;
693 	a->tx_errors = b->tx_errors;
694 	a->rx_dropped = b->rx_dropped;
695 	a->tx_dropped = b->tx_dropped;
696 
697 	a->multicast = b->multicast;
698 	a->collisions = b->collisions;
699 
700 	a->rx_length_errors = b->rx_length_errors;
701 	a->rx_over_errors = b->rx_over_errors;
702 	a->rx_crc_errors = b->rx_crc_errors;
703 	a->rx_frame_errors = b->rx_frame_errors;
704 	a->rx_fifo_errors = b->rx_fifo_errors;
705 	a->rx_missed_errors = b->rx_missed_errors;
706 
707 	a->tx_aborted_errors = b->tx_aborted_errors;
708 	a->tx_carrier_errors = b->tx_carrier_errors;
709 	a->tx_fifo_errors = b->tx_fifo_errors;
710 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
711 	a->tx_window_errors = b->tx_window_errors;
712 
713 	a->rx_compressed = b->rx_compressed;
714 	a->tx_compressed = b->tx_compressed;
715 }
716 
717 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
718 {
719 	memcpy(v, b, sizeof(*b));
720 }
721 
722 /* All VF info */
723 static inline int rtnl_vfinfo_size(const struct net_device *dev,
724 				   u32 ext_filter_mask)
725 {
726 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
727 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
728 		int num_vfs = dev_num_vf(dev->dev.parent);
729 		size_t size = nla_total_size(sizeof(struct nlattr));
730 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
731 		size += num_vfs *
732 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
733 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
734 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
735 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
736 		return size;
737 	} else
738 		return 0;
739 }
740 
741 static size_t rtnl_port_size(const struct net_device *dev)
742 {
743 	size_t port_size = nla_total_size(4)		/* PORT_VF */
744 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
745 		+ nla_total_size(sizeof(struct ifla_port_vsi))
746 							/* PORT_VSI_TYPE */
747 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
748 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
749 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
750 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
751 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
752 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
753 		+ port_size;
754 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
755 		+ port_size;
756 
757 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
758 		return 0;
759 	if (dev_num_vf(dev->dev.parent))
760 		return port_self_size + vf_ports_size +
761 			vf_port_size * dev_num_vf(dev->dev.parent);
762 	else
763 		return port_self_size;
764 }
765 
766 static noinline size_t if_nlmsg_size(const struct net_device *dev,
767 				     u32 ext_filter_mask)
768 {
769 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
770 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
771 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
772 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
773 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
774 	       + nla_total_size(sizeof(struct rtnl_link_stats))
775 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
776 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
777 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
778 	       + nla_total_size(4) /* IFLA_TXQLEN */
779 	       + nla_total_size(4) /* IFLA_WEIGHT */
780 	       + nla_total_size(4) /* IFLA_MTU */
781 	       + nla_total_size(4) /* IFLA_LINK */
782 	       + nla_total_size(4) /* IFLA_MASTER */
783 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
784 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
785 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
786 	       + nla_total_size(1) /* IFLA_OPERSTATE */
787 	       + nla_total_size(1) /* IFLA_LINKMODE */
788 	       + nla_total_size(ext_filter_mask
789 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
790 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
791 	       + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
792 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
793 	       + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
794 }
795 
796 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
797 {
798 	struct nlattr *vf_ports;
799 	struct nlattr *vf_port;
800 	int vf;
801 	int err;
802 
803 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
804 	if (!vf_ports)
805 		return -EMSGSIZE;
806 
807 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
808 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
809 		if (!vf_port)
810 			goto nla_put_failure;
811 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
812 			goto nla_put_failure;
813 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
814 		if (err == -EMSGSIZE)
815 			goto nla_put_failure;
816 		if (err) {
817 			nla_nest_cancel(skb, vf_port);
818 			continue;
819 		}
820 		nla_nest_end(skb, vf_port);
821 	}
822 
823 	nla_nest_end(skb, vf_ports);
824 
825 	return 0;
826 
827 nla_put_failure:
828 	nla_nest_cancel(skb, vf_ports);
829 	return -EMSGSIZE;
830 }
831 
832 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
833 {
834 	struct nlattr *port_self;
835 	int err;
836 
837 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
838 	if (!port_self)
839 		return -EMSGSIZE;
840 
841 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
842 	if (err) {
843 		nla_nest_cancel(skb, port_self);
844 		return (err == -EMSGSIZE) ? err : 0;
845 	}
846 
847 	nla_nest_end(skb, port_self);
848 
849 	return 0;
850 }
851 
852 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
853 {
854 	int err;
855 
856 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
857 		return 0;
858 
859 	err = rtnl_port_self_fill(skb, dev);
860 	if (err)
861 		return err;
862 
863 	if (dev_num_vf(dev->dev.parent)) {
864 		err = rtnl_vf_ports_fill(skb, dev);
865 		if (err)
866 			return err;
867 	}
868 
869 	return 0;
870 }
871 
872 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
873 			    int type, u32 pid, u32 seq, u32 change,
874 			    unsigned int flags, u32 ext_filter_mask)
875 {
876 	struct ifinfomsg *ifm;
877 	struct nlmsghdr *nlh;
878 	struct rtnl_link_stats64 temp;
879 	const struct rtnl_link_stats64 *stats;
880 	struct nlattr *attr, *af_spec;
881 	struct rtnl_af_ops *af_ops;
882 
883 	ASSERT_RTNL();
884 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
885 	if (nlh == NULL)
886 		return -EMSGSIZE;
887 
888 	ifm = nlmsg_data(nlh);
889 	ifm->ifi_family = AF_UNSPEC;
890 	ifm->__ifi_pad = 0;
891 	ifm->ifi_type = dev->type;
892 	ifm->ifi_index = dev->ifindex;
893 	ifm->ifi_flags = dev_get_flags(dev);
894 	ifm->ifi_change = change;
895 
896 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
897 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
898 	    nla_put_u8(skb, IFLA_OPERSTATE,
899 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
900 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
901 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
902 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
903 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
904 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
905 #ifdef CONFIG_RPS
906 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
907 #endif
908 	    (dev->ifindex != dev->iflink &&
909 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
910 	    (dev->master &&
911 	     nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
912 	    (dev->qdisc &&
913 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
914 	    (dev->ifalias &&
915 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
916 		goto nla_put_failure;
917 
918 	if (1) {
919 		struct rtnl_link_ifmap map = {
920 			.mem_start   = dev->mem_start,
921 			.mem_end     = dev->mem_end,
922 			.base_addr   = dev->base_addr,
923 			.irq         = dev->irq,
924 			.dma         = dev->dma,
925 			.port        = dev->if_port,
926 		};
927 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
928 			goto nla_put_failure;
929 	}
930 
931 	if (dev->addr_len) {
932 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
933 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
934 			goto nla_put_failure;
935 	}
936 
937 	attr = nla_reserve(skb, IFLA_STATS,
938 			sizeof(struct rtnl_link_stats));
939 	if (attr == NULL)
940 		goto nla_put_failure;
941 
942 	stats = dev_get_stats(dev, &temp);
943 	copy_rtnl_link_stats(nla_data(attr), stats);
944 
945 	attr = nla_reserve(skb, IFLA_STATS64,
946 			sizeof(struct rtnl_link_stats64));
947 	if (attr == NULL)
948 		goto nla_put_failure;
949 	copy_rtnl_link_stats64(nla_data(attr), stats);
950 
951 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
952 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
953 		goto nla_put_failure;
954 
955 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
956 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
957 		int i;
958 
959 		struct nlattr *vfinfo, *vf;
960 		int num_vfs = dev_num_vf(dev->dev.parent);
961 
962 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
963 		if (!vfinfo)
964 			goto nla_put_failure;
965 		for (i = 0; i < num_vfs; i++) {
966 			struct ifla_vf_info ivi;
967 			struct ifla_vf_mac vf_mac;
968 			struct ifla_vf_vlan vf_vlan;
969 			struct ifla_vf_tx_rate vf_tx_rate;
970 			struct ifla_vf_spoofchk vf_spoofchk;
971 
972 			/*
973 			 * Not all SR-IOV capable drivers support the
974 			 * spoofcheck query.  Preset to -1 so the user
975 			 * space tool can detect that the driver didn't
976 			 * report anything.
977 			 */
978 			ivi.spoofchk = -1;
979 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
980 				break;
981 			vf_mac.vf =
982 				vf_vlan.vf =
983 				vf_tx_rate.vf =
984 				vf_spoofchk.vf = ivi.vf;
985 
986 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
987 			vf_vlan.vlan = ivi.vlan;
988 			vf_vlan.qos = ivi.qos;
989 			vf_tx_rate.rate = ivi.tx_rate;
990 			vf_spoofchk.setting = ivi.spoofchk;
991 			vf = nla_nest_start(skb, IFLA_VF_INFO);
992 			if (!vf) {
993 				nla_nest_cancel(skb, vfinfo);
994 				goto nla_put_failure;
995 			}
996 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
997 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
998 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
999 				    &vf_tx_rate) ||
1000 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1001 				    &vf_spoofchk))
1002 				goto nla_put_failure;
1003 			nla_nest_end(skb, vf);
1004 		}
1005 		nla_nest_end(skb, vfinfo);
1006 	}
1007 
1008 	if (rtnl_port_fill(skb, dev))
1009 		goto nla_put_failure;
1010 
1011 	if (dev->rtnl_link_ops) {
1012 		if (rtnl_link_fill(skb, dev) < 0)
1013 			goto nla_put_failure;
1014 	}
1015 
1016 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1017 		goto nla_put_failure;
1018 
1019 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1020 		if (af_ops->fill_link_af) {
1021 			struct nlattr *af;
1022 			int err;
1023 
1024 			if (!(af = nla_nest_start(skb, af_ops->family)))
1025 				goto nla_put_failure;
1026 
1027 			err = af_ops->fill_link_af(skb, dev);
1028 
1029 			/*
1030 			 * Caller may return ENODATA to indicate that there
1031 			 * was no data to be dumped. This is not an error, it
1032 			 * means we should trim the attribute header and
1033 			 * continue.
1034 			 */
1035 			if (err == -ENODATA)
1036 				nla_nest_cancel(skb, af);
1037 			else if (err < 0)
1038 				goto nla_put_failure;
1039 
1040 			nla_nest_end(skb, af);
1041 		}
1042 	}
1043 
1044 	nla_nest_end(skb, af_spec);
1045 
1046 	return nlmsg_end(skb, nlh);
1047 
1048 nla_put_failure:
1049 	nlmsg_cancel(skb, nlh);
1050 	return -EMSGSIZE;
1051 }
1052 
1053 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1054 {
1055 	struct net *net = sock_net(skb->sk);
1056 	int h, s_h;
1057 	int idx = 0, s_idx;
1058 	struct net_device *dev;
1059 	struct hlist_head *head;
1060 	struct hlist_node *node;
1061 	struct nlattr *tb[IFLA_MAX+1];
1062 	u32 ext_filter_mask = 0;
1063 
1064 	s_h = cb->args[0];
1065 	s_idx = cb->args[1];
1066 
1067 	rcu_read_lock();
1068 	cb->seq = net->dev_base_seq;
1069 
1070 	if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1071 			ifla_policy) >= 0) {
1072 
1073 		if (tb[IFLA_EXT_MASK])
1074 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1075 	}
1076 
1077 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1078 		idx = 0;
1079 		head = &net->dev_index_head[h];
1080 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1081 			if (idx < s_idx)
1082 				goto cont;
1083 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1084 					     NETLINK_CB(cb->skb).portid,
1085 					     cb->nlh->nlmsg_seq, 0,
1086 					     NLM_F_MULTI,
1087 					     ext_filter_mask) <= 0)
1088 				goto out;
1089 
1090 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1091 cont:
1092 			idx++;
1093 		}
1094 	}
1095 out:
1096 	rcu_read_unlock();
1097 	cb->args[1] = idx;
1098 	cb->args[0] = h;
1099 
1100 	return skb->len;
1101 }
1102 
1103 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1104 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1105 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1106 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1107 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1108 	[IFLA_MTU]		= { .type = NLA_U32 },
1109 	[IFLA_LINK]		= { .type = NLA_U32 },
1110 	[IFLA_MASTER]		= { .type = NLA_U32 },
1111 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1112 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1113 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1114 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1115 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1116 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1117 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1118 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1119 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1120 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1121 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1122 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1123 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1124 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1125 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1126 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1127 };
1128 EXPORT_SYMBOL(ifla_policy);
1129 
1130 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1131 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1132 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1133 };
1134 
1135 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1136 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1137 };
1138 
1139 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1140 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1141 				    .len = sizeof(struct ifla_vf_mac) },
1142 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1143 				    .len = sizeof(struct ifla_vf_vlan) },
1144 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1145 				    .len = sizeof(struct ifla_vf_tx_rate) },
1146 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1147 				    .len = sizeof(struct ifla_vf_spoofchk) },
1148 };
1149 
1150 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1151 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1152 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1153 				    .len = PORT_PROFILE_MAX },
1154 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1155 				    .len = sizeof(struct ifla_port_vsi)},
1156 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1157 				      .len = PORT_UUID_MAX },
1158 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1159 				    .len = PORT_UUID_MAX },
1160 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1161 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1162 };
1163 
1164 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1165 {
1166 	struct net *net;
1167 	/* Examine the link attributes and figure out which
1168 	 * network namespace we are talking about.
1169 	 */
1170 	if (tb[IFLA_NET_NS_PID])
1171 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1172 	else if (tb[IFLA_NET_NS_FD])
1173 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1174 	else
1175 		net = get_net(src_net);
1176 	return net;
1177 }
1178 EXPORT_SYMBOL(rtnl_link_get_net);
1179 
1180 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1181 {
1182 	if (dev) {
1183 		if (tb[IFLA_ADDRESS] &&
1184 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1185 			return -EINVAL;
1186 
1187 		if (tb[IFLA_BROADCAST] &&
1188 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1189 			return -EINVAL;
1190 	}
1191 
1192 	if (tb[IFLA_AF_SPEC]) {
1193 		struct nlattr *af;
1194 		int rem, err;
1195 
1196 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1197 			const struct rtnl_af_ops *af_ops;
1198 
1199 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1200 				return -EAFNOSUPPORT;
1201 
1202 			if (!af_ops->set_link_af)
1203 				return -EOPNOTSUPP;
1204 
1205 			if (af_ops->validate_link_af) {
1206 				err = af_ops->validate_link_af(dev, af);
1207 				if (err < 0)
1208 					return err;
1209 			}
1210 		}
1211 	}
1212 
1213 	return 0;
1214 }
1215 
1216 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1217 {
1218 	int rem, err = -EINVAL;
1219 	struct nlattr *vf;
1220 	const struct net_device_ops *ops = dev->netdev_ops;
1221 
1222 	nla_for_each_nested(vf, attr, rem) {
1223 		switch (nla_type(vf)) {
1224 		case IFLA_VF_MAC: {
1225 			struct ifla_vf_mac *ivm;
1226 			ivm = nla_data(vf);
1227 			err = -EOPNOTSUPP;
1228 			if (ops->ndo_set_vf_mac)
1229 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1230 							  ivm->mac);
1231 			break;
1232 		}
1233 		case IFLA_VF_VLAN: {
1234 			struct ifla_vf_vlan *ivv;
1235 			ivv = nla_data(vf);
1236 			err = -EOPNOTSUPP;
1237 			if (ops->ndo_set_vf_vlan)
1238 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1239 							   ivv->vlan,
1240 							   ivv->qos);
1241 			break;
1242 		}
1243 		case IFLA_VF_TX_RATE: {
1244 			struct ifla_vf_tx_rate *ivt;
1245 			ivt = nla_data(vf);
1246 			err = -EOPNOTSUPP;
1247 			if (ops->ndo_set_vf_tx_rate)
1248 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1249 							      ivt->rate);
1250 			break;
1251 		}
1252 		case IFLA_VF_SPOOFCHK: {
1253 			struct ifla_vf_spoofchk *ivs;
1254 			ivs = nla_data(vf);
1255 			err = -EOPNOTSUPP;
1256 			if (ops->ndo_set_vf_spoofchk)
1257 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1258 							       ivs->setting);
1259 			break;
1260 		}
1261 		default:
1262 			err = -EINVAL;
1263 			break;
1264 		}
1265 		if (err)
1266 			break;
1267 	}
1268 	return err;
1269 }
1270 
1271 static int do_set_master(struct net_device *dev, int ifindex)
1272 {
1273 	struct net_device *master_dev;
1274 	const struct net_device_ops *ops;
1275 	int err;
1276 
1277 	if (dev->master) {
1278 		if (dev->master->ifindex == ifindex)
1279 			return 0;
1280 		ops = dev->master->netdev_ops;
1281 		if (ops->ndo_del_slave) {
1282 			err = ops->ndo_del_slave(dev->master, dev);
1283 			if (err)
1284 				return err;
1285 		} else {
1286 			return -EOPNOTSUPP;
1287 		}
1288 	}
1289 
1290 	if (ifindex) {
1291 		master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1292 		if (!master_dev)
1293 			return -EINVAL;
1294 		ops = master_dev->netdev_ops;
1295 		if (ops->ndo_add_slave) {
1296 			err = ops->ndo_add_slave(master_dev, dev);
1297 			if (err)
1298 				return err;
1299 		} else {
1300 			return -EOPNOTSUPP;
1301 		}
1302 	}
1303 	return 0;
1304 }
1305 
1306 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1307 		      struct nlattr **tb, char *ifname, int modified)
1308 {
1309 	const struct net_device_ops *ops = dev->netdev_ops;
1310 	int send_addr_notify = 0;
1311 	int err;
1312 
1313 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1314 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1315 		if (IS_ERR(net)) {
1316 			err = PTR_ERR(net);
1317 			goto errout;
1318 		}
1319 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1320 			err = -EPERM;
1321 			goto errout;
1322 		}
1323 		err = dev_change_net_namespace(dev, net, ifname);
1324 		put_net(net);
1325 		if (err)
1326 			goto errout;
1327 		modified = 1;
1328 	}
1329 
1330 	if (tb[IFLA_MAP]) {
1331 		struct rtnl_link_ifmap *u_map;
1332 		struct ifmap k_map;
1333 
1334 		if (!ops->ndo_set_config) {
1335 			err = -EOPNOTSUPP;
1336 			goto errout;
1337 		}
1338 
1339 		if (!netif_device_present(dev)) {
1340 			err = -ENODEV;
1341 			goto errout;
1342 		}
1343 
1344 		u_map = nla_data(tb[IFLA_MAP]);
1345 		k_map.mem_start = (unsigned long) u_map->mem_start;
1346 		k_map.mem_end = (unsigned long) u_map->mem_end;
1347 		k_map.base_addr = (unsigned short) u_map->base_addr;
1348 		k_map.irq = (unsigned char) u_map->irq;
1349 		k_map.dma = (unsigned char) u_map->dma;
1350 		k_map.port = (unsigned char) u_map->port;
1351 
1352 		err = ops->ndo_set_config(dev, &k_map);
1353 		if (err < 0)
1354 			goto errout;
1355 
1356 		modified = 1;
1357 	}
1358 
1359 	if (tb[IFLA_ADDRESS]) {
1360 		struct sockaddr *sa;
1361 		int len;
1362 
1363 		if (!ops->ndo_set_mac_address) {
1364 			err = -EOPNOTSUPP;
1365 			goto errout;
1366 		}
1367 
1368 		if (!netif_device_present(dev)) {
1369 			err = -ENODEV;
1370 			goto errout;
1371 		}
1372 
1373 		len = sizeof(sa_family_t) + dev->addr_len;
1374 		sa = kmalloc(len, GFP_KERNEL);
1375 		if (!sa) {
1376 			err = -ENOMEM;
1377 			goto errout;
1378 		}
1379 		sa->sa_family = dev->type;
1380 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1381 		       dev->addr_len);
1382 		err = ops->ndo_set_mac_address(dev, sa);
1383 		kfree(sa);
1384 		if (err)
1385 			goto errout;
1386 		send_addr_notify = 1;
1387 		modified = 1;
1388 		add_device_randomness(dev->dev_addr, dev->addr_len);
1389 	}
1390 
1391 	if (tb[IFLA_MTU]) {
1392 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1393 		if (err < 0)
1394 			goto errout;
1395 		modified = 1;
1396 	}
1397 
1398 	if (tb[IFLA_GROUP]) {
1399 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1400 		modified = 1;
1401 	}
1402 
1403 	/*
1404 	 * Interface selected by interface index but interface
1405 	 * name provided implies that a name change has been
1406 	 * requested.
1407 	 */
1408 	if (ifm->ifi_index > 0 && ifname[0]) {
1409 		err = dev_change_name(dev, ifname);
1410 		if (err < 0)
1411 			goto errout;
1412 		modified = 1;
1413 	}
1414 
1415 	if (tb[IFLA_IFALIAS]) {
1416 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1417 				    nla_len(tb[IFLA_IFALIAS]));
1418 		if (err < 0)
1419 			goto errout;
1420 		modified = 1;
1421 	}
1422 
1423 	if (tb[IFLA_BROADCAST]) {
1424 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1425 		send_addr_notify = 1;
1426 	}
1427 
1428 	if (ifm->ifi_flags || ifm->ifi_change) {
1429 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1430 		if (err < 0)
1431 			goto errout;
1432 	}
1433 
1434 	if (tb[IFLA_MASTER]) {
1435 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
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 	if (send_addr_notify)
1540 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1541 
1542 	return err;
1543 }
1544 
1545 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1546 {
1547 	struct net *net = sock_net(skb->sk);
1548 	struct ifinfomsg *ifm;
1549 	struct net_device *dev;
1550 	int err;
1551 	struct nlattr *tb[IFLA_MAX+1];
1552 	char ifname[IFNAMSIZ];
1553 
1554 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1555 	if (err < 0)
1556 		goto errout;
1557 
1558 	if (tb[IFLA_IFNAME])
1559 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1560 	else
1561 		ifname[0] = '\0';
1562 
1563 	err = -EINVAL;
1564 	ifm = nlmsg_data(nlh);
1565 	if (ifm->ifi_index > 0)
1566 		dev = __dev_get_by_index(net, ifm->ifi_index);
1567 	else if (tb[IFLA_IFNAME])
1568 		dev = __dev_get_by_name(net, ifname);
1569 	else
1570 		goto errout;
1571 
1572 	if (dev == NULL) {
1573 		err = -ENODEV;
1574 		goto errout;
1575 	}
1576 
1577 	err = validate_linkmsg(dev, tb);
1578 	if (err < 0)
1579 		goto errout;
1580 
1581 	err = do_setlink(dev, ifm, tb, ifname, 0);
1582 errout:
1583 	return err;
1584 }
1585 
1586 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1587 {
1588 	struct net *net = sock_net(skb->sk);
1589 	const struct rtnl_link_ops *ops;
1590 	struct net_device *dev;
1591 	struct ifinfomsg *ifm;
1592 	char ifname[IFNAMSIZ];
1593 	struct nlattr *tb[IFLA_MAX+1];
1594 	int err;
1595 	LIST_HEAD(list_kill);
1596 
1597 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1598 	if (err < 0)
1599 		return err;
1600 
1601 	if (tb[IFLA_IFNAME])
1602 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1603 
1604 	ifm = nlmsg_data(nlh);
1605 	if (ifm->ifi_index > 0)
1606 		dev = __dev_get_by_index(net, ifm->ifi_index);
1607 	else if (tb[IFLA_IFNAME])
1608 		dev = __dev_get_by_name(net, ifname);
1609 	else
1610 		return -EINVAL;
1611 
1612 	if (!dev)
1613 		return -ENODEV;
1614 
1615 	ops = dev->rtnl_link_ops;
1616 	if (!ops)
1617 		return -EOPNOTSUPP;
1618 
1619 	ops->dellink(dev, &list_kill);
1620 	unregister_netdevice_many(&list_kill);
1621 	list_del(&list_kill);
1622 	return 0;
1623 }
1624 
1625 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1626 {
1627 	unsigned int old_flags;
1628 	int err;
1629 
1630 	old_flags = dev->flags;
1631 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1632 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1633 		if (err < 0)
1634 			return err;
1635 	}
1636 
1637 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1638 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1639 
1640 	__dev_notify_flags(dev, old_flags);
1641 	return 0;
1642 }
1643 EXPORT_SYMBOL(rtnl_configure_link);
1644 
1645 struct net_device *rtnl_create_link(struct net *net,
1646 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1647 {
1648 	int err;
1649 	struct net_device *dev;
1650 	unsigned int num_tx_queues = 1;
1651 	unsigned int num_rx_queues = 1;
1652 
1653 	if (tb[IFLA_NUM_TX_QUEUES])
1654 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1655 	else if (ops->get_num_tx_queues)
1656 		num_tx_queues = ops->get_num_tx_queues();
1657 
1658 	if (tb[IFLA_NUM_RX_QUEUES])
1659 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1660 	else if (ops->get_num_rx_queues)
1661 		num_rx_queues = ops->get_num_rx_queues();
1662 
1663 	err = -ENOMEM;
1664 	dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1665 			       num_tx_queues, num_rx_queues);
1666 	if (!dev)
1667 		goto err;
1668 
1669 	dev_net_set(dev, net);
1670 	dev->rtnl_link_ops = ops;
1671 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1672 
1673 	if (tb[IFLA_MTU])
1674 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1675 	if (tb[IFLA_ADDRESS])
1676 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1677 				nla_len(tb[IFLA_ADDRESS]));
1678 	if (tb[IFLA_BROADCAST])
1679 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1680 				nla_len(tb[IFLA_BROADCAST]));
1681 	if (tb[IFLA_TXQLEN])
1682 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1683 	if (tb[IFLA_OPERSTATE])
1684 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1685 	if (tb[IFLA_LINKMODE])
1686 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1687 	if (tb[IFLA_GROUP])
1688 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1689 
1690 	return dev;
1691 
1692 err:
1693 	return ERR_PTR(err);
1694 }
1695 EXPORT_SYMBOL(rtnl_create_link);
1696 
1697 static int rtnl_group_changelink(struct net *net, int group,
1698 		struct ifinfomsg *ifm,
1699 		struct nlattr **tb)
1700 {
1701 	struct net_device *dev;
1702 	int err;
1703 
1704 	for_each_netdev(net, dev) {
1705 		if (dev->group == group) {
1706 			err = do_setlink(dev, ifm, tb, NULL, 0);
1707 			if (err < 0)
1708 				return err;
1709 		}
1710 	}
1711 
1712 	return 0;
1713 }
1714 
1715 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1716 {
1717 	struct net *net = sock_net(skb->sk);
1718 	const struct rtnl_link_ops *ops;
1719 	struct net_device *dev;
1720 	struct ifinfomsg *ifm;
1721 	char kind[MODULE_NAME_LEN];
1722 	char ifname[IFNAMSIZ];
1723 	struct nlattr *tb[IFLA_MAX+1];
1724 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1725 	int err;
1726 
1727 #ifdef CONFIG_MODULES
1728 replay:
1729 #endif
1730 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1731 	if (err < 0)
1732 		return err;
1733 
1734 	if (tb[IFLA_IFNAME])
1735 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1736 	else
1737 		ifname[0] = '\0';
1738 
1739 	ifm = nlmsg_data(nlh);
1740 	if (ifm->ifi_index > 0)
1741 		dev = __dev_get_by_index(net, ifm->ifi_index);
1742 	else {
1743 		if (ifname[0])
1744 			dev = __dev_get_by_name(net, ifname);
1745 		else
1746 			dev = NULL;
1747 	}
1748 
1749 	err = validate_linkmsg(dev, tb);
1750 	if (err < 0)
1751 		return err;
1752 
1753 	if (tb[IFLA_LINKINFO]) {
1754 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1755 				       tb[IFLA_LINKINFO], ifla_info_policy);
1756 		if (err < 0)
1757 			return err;
1758 	} else
1759 		memset(linkinfo, 0, sizeof(linkinfo));
1760 
1761 	if (linkinfo[IFLA_INFO_KIND]) {
1762 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1763 		ops = rtnl_link_ops_get(kind);
1764 	} else {
1765 		kind[0] = '\0';
1766 		ops = NULL;
1767 	}
1768 
1769 	if (1) {
1770 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1771 		struct net *dest_net;
1772 
1773 		if (ops) {
1774 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1775 				err = nla_parse_nested(attr, ops->maxtype,
1776 						       linkinfo[IFLA_INFO_DATA],
1777 						       ops->policy);
1778 				if (err < 0)
1779 					return err;
1780 				data = attr;
1781 			}
1782 			if (ops->validate) {
1783 				err = ops->validate(tb, data);
1784 				if (err < 0)
1785 					return err;
1786 			}
1787 		}
1788 
1789 		if (dev) {
1790 			int modified = 0;
1791 
1792 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1793 				return -EEXIST;
1794 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1795 				return -EOPNOTSUPP;
1796 
1797 			if (linkinfo[IFLA_INFO_DATA]) {
1798 				if (!ops || ops != dev->rtnl_link_ops ||
1799 				    !ops->changelink)
1800 					return -EOPNOTSUPP;
1801 
1802 				err = ops->changelink(dev, tb, data);
1803 				if (err < 0)
1804 					return err;
1805 				modified = 1;
1806 			}
1807 
1808 			return do_setlink(dev, ifm, tb, ifname, modified);
1809 		}
1810 
1811 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1812 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1813 				return rtnl_group_changelink(net,
1814 						nla_get_u32(tb[IFLA_GROUP]),
1815 						ifm, tb);
1816 			return -ENODEV;
1817 		}
1818 
1819 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1820 			return -EOPNOTSUPP;
1821 
1822 		if (!ops) {
1823 #ifdef CONFIG_MODULES
1824 			if (kind[0]) {
1825 				__rtnl_unlock();
1826 				request_module("rtnl-link-%s", kind);
1827 				rtnl_lock();
1828 				ops = rtnl_link_ops_get(kind);
1829 				if (ops)
1830 					goto replay;
1831 			}
1832 #endif
1833 			return -EOPNOTSUPP;
1834 		}
1835 
1836 		if (!ifname[0])
1837 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1838 
1839 		dest_net = rtnl_link_get_net(net, tb);
1840 		if (IS_ERR(dest_net))
1841 			return PTR_ERR(dest_net);
1842 
1843 		dev = rtnl_create_link(dest_net, ifname, ops, tb);
1844 		if (IS_ERR(dev)) {
1845 			err = PTR_ERR(dev);
1846 			goto out;
1847 		}
1848 
1849 		dev->ifindex = ifm->ifi_index;
1850 
1851 		if (ops->newlink)
1852 			err = ops->newlink(net, dev, tb, data);
1853 		else
1854 			err = register_netdevice(dev);
1855 
1856 		if (err < 0 && !IS_ERR(dev))
1857 			free_netdev(dev);
1858 		if (err < 0)
1859 			goto out;
1860 
1861 		err = rtnl_configure_link(dev, ifm);
1862 		if (err < 0)
1863 			unregister_netdevice(dev);
1864 out:
1865 		put_net(dest_net);
1866 		return err;
1867 	}
1868 }
1869 
1870 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1871 {
1872 	struct net *net = sock_net(skb->sk);
1873 	struct ifinfomsg *ifm;
1874 	char ifname[IFNAMSIZ];
1875 	struct nlattr *tb[IFLA_MAX+1];
1876 	struct net_device *dev = NULL;
1877 	struct sk_buff *nskb;
1878 	int err;
1879 	u32 ext_filter_mask = 0;
1880 
1881 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1882 	if (err < 0)
1883 		return err;
1884 
1885 	if (tb[IFLA_IFNAME])
1886 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1887 
1888 	if (tb[IFLA_EXT_MASK])
1889 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1890 
1891 	ifm = nlmsg_data(nlh);
1892 	if (ifm->ifi_index > 0)
1893 		dev = __dev_get_by_index(net, ifm->ifi_index);
1894 	else if (tb[IFLA_IFNAME])
1895 		dev = __dev_get_by_name(net, ifname);
1896 	else
1897 		return -EINVAL;
1898 
1899 	if (dev == NULL)
1900 		return -ENODEV;
1901 
1902 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1903 	if (nskb == NULL)
1904 		return -ENOBUFS;
1905 
1906 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1907 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1908 	if (err < 0) {
1909 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1910 		WARN_ON(err == -EMSGSIZE);
1911 		kfree_skb(nskb);
1912 	} else
1913 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1914 
1915 	return err;
1916 }
1917 
1918 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1919 {
1920 	struct net *net = sock_net(skb->sk);
1921 	struct net_device *dev;
1922 	struct nlattr *tb[IFLA_MAX+1];
1923 	u32 ext_filter_mask = 0;
1924 	u16 min_ifinfo_dump_size = 0;
1925 
1926 	if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1927 			ifla_policy) >= 0) {
1928 		if (tb[IFLA_EXT_MASK])
1929 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1930 	}
1931 
1932 	if (!ext_filter_mask)
1933 		return NLMSG_GOODSIZE;
1934 	/*
1935 	 * traverse the list of net devices and compute the minimum
1936 	 * buffer size based upon the filter mask.
1937 	 */
1938 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1939 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1940 					     if_nlmsg_size(dev,
1941 						           ext_filter_mask));
1942 	}
1943 
1944 	return min_ifinfo_dump_size;
1945 }
1946 
1947 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1948 {
1949 	int idx;
1950 	int s_idx = cb->family;
1951 
1952 	if (s_idx == 0)
1953 		s_idx = 1;
1954 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1955 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1956 		if (idx < s_idx || idx == PF_PACKET)
1957 			continue;
1958 		if (rtnl_msg_handlers[idx] == NULL ||
1959 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1960 			continue;
1961 		if (idx > s_idx)
1962 			memset(&cb->args[0], 0, sizeof(cb->args));
1963 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1964 			break;
1965 	}
1966 	cb->family = idx;
1967 
1968 	return skb->len;
1969 }
1970 
1971 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1972 {
1973 	struct net *net = dev_net(dev);
1974 	struct sk_buff *skb;
1975 	int err = -ENOBUFS;
1976 	size_t if_info_size;
1977 
1978 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1979 	if (skb == NULL)
1980 		goto errout;
1981 
1982 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1983 	if (err < 0) {
1984 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1985 		WARN_ON(err == -EMSGSIZE);
1986 		kfree_skb(skb);
1987 		goto errout;
1988 	}
1989 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1990 	return;
1991 errout:
1992 	if (err < 0)
1993 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1994 }
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 net_device *master = NULL;
2058 	struct ndmsg *ndm;
2059 	struct nlattr *tb[NDA_MAX+1];
2060 	struct net_device *dev;
2061 	u8 *addr;
2062 	int err;
2063 
2064 	if (!capable(CAP_NET_ADMIN))
2065 		return -EPERM;
2066 
2067 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2068 	if (err < 0)
2069 		return err;
2070 
2071 	ndm = nlmsg_data(nlh);
2072 	if (ndm->ndm_ifindex == 0) {
2073 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2074 		return -EINVAL;
2075 	}
2076 
2077 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2078 	if (dev == NULL) {
2079 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2080 		return -ENODEV;
2081 	}
2082 
2083 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2084 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2085 		return -EINVAL;
2086 	}
2087 
2088 	addr = nla_data(tb[NDA_LLADDR]);
2089 	if (!is_valid_ether_addr(addr)) {
2090 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2091 		return -EINVAL;
2092 	}
2093 
2094 	err = -EOPNOTSUPP;
2095 
2096 	/* Support fdb on master device the net/bridge default case */
2097 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2098 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2099 		master = dev->master;
2100 		err = master->netdev_ops->ndo_fdb_add(ndm, tb,
2101 						      dev, addr,
2102 						      nlh->nlmsg_flags);
2103 		if (err)
2104 			goto out;
2105 		else
2106 			ndm->ndm_flags &= ~NTF_MASTER;
2107 	}
2108 
2109 	/* Embedded bridge, macvlan, and any other device support */
2110 	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2111 		err = dev->netdev_ops->ndo_fdb_add(ndm, tb,
2112 						   dev, addr,
2113 						   nlh->nlmsg_flags);
2114 
2115 		if (!err) {
2116 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2117 			ndm->ndm_flags &= ~NTF_SELF;
2118 		}
2119 	}
2120 out:
2121 	return err;
2122 }
2123 
2124 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2125 {
2126 	struct net *net = sock_net(skb->sk);
2127 	struct ndmsg *ndm;
2128 	struct nlattr *llattr;
2129 	struct net_device *dev;
2130 	int err = -EINVAL;
2131 	__u8 *addr;
2132 
2133 	if (!capable(CAP_NET_ADMIN))
2134 		return -EPERM;
2135 
2136 	if (nlmsg_len(nlh) < sizeof(*ndm))
2137 		return -EINVAL;
2138 
2139 	ndm = nlmsg_data(nlh);
2140 	if (ndm->ndm_ifindex == 0) {
2141 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2142 		return -EINVAL;
2143 	}
2144 
2145 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2146 	if (dev == NULL) {
2147 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2148 		return -ENODEV;
2149 	}
2150 
2151 	llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2152 	if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2153 		pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2154 		return -EINVAL;
2155 	}
2156 
2157 	addr = nla_data(llattr);
2158 	err = -EOPNOTSUPP;
2159 
2160 	/* Support fdb on master device the net/bridge default case */
2161 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2162 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2163 		struct net_device *master = dev->master;
2164 
2165 		if (master->netdev_ops->ndo_fdb_del)
2166 			err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2167 
2168 		if (err)
2169 			goto out;
2170 		else
2171 			ndm->ndm_flags &= ~NTF_MASTER;
2172 	}
2173 
2174 	/* Embedded bridge, macvlan, and any other device support */
2175 	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2176 		err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2177 
2178 		if (!err) {
2179 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2180 			ndm->ndm_flags &= ~NTF_SELF;
2181 		}
2182 	}
2183 out:
2184 	return err;
2185 }
2186 
2187 static int nlmsg_populate_fdb(struct sk_buff *skb,
2188 			      struct netlink_callback *cb,
2189 			      struct net_device *dev,
2190 			      int *idx,
2191 			      struct netdev_hw_addr_list *list)
2192 {
2193 	struct netdev_hw_addr *ha;
2194 	int err;
2195 	u32 portid, seq;
2196 
2197 	portid = NETLINK_CB(cb->skb).portid;
2198 	seq = cb->nlh->nlmsg_seq;
2199 
2200 	list_for_each_entry(ha, &list->list, list) {
2201 		if (*idx < cb->args[0])
2202 			goto skip;
2203 
2204 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2205 					      portid, seq,
2206 					      RTM_NEWNEIGH, NTF_SELF);
2207 		if (err < 0)
2208 			return err;
2209 skip:
2210 		*idx += 1;
2211 	}
2212 	return 0;
2213 }
2214 
2215 /**
2216  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2217  * @nlh: netlink message header
2218  * @dev: netdevice
2219  *
2220  * Default netdevice operation to dump the existing unicast address list.
2221  * Returns zero on success.
2222  */
2223 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2224 		      struct netlink_callback *cb,
2225 		      struct net_device *dev,
2226 		      int idx)
2227 {
2228 	int err;
2229 
2230 	netif_addr_lock_bh(dev);
2231 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2232 	if (err)
2233 		goto out;
2234 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2235 out:
2236 	netif_addr_unlock_bh(dev);
2237 	return idx;
2238 }
2239 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2240 
2241 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2242 {
2243 	int idx = 0;
2244 	struct net *net = sock_net(skb->sk);
2245 	struct net_device *dev;
2246 
2247 	rcu_read_lock();
2248 	for_each_netdev_rcu(net, dev) {
2249 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2250 			struct net_device *master = dev->master;
2251 			const struct net_device_ops *ops = master->netdev_ops;
2252 
2253 			if (ops->ndo_fdb_dump)
2254 				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2255 		}
2256 
2257 		if (dev->netdev_ops->ndo_fdb_dump)
2258 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2259 	}
2260 	rcu_read_unlock();
2261 
2262 	cb->args[0] = idx;
2263 	return skb->len;
2264 }
2265 
2266 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2267 			    struct net_device *dev, u16 mode)
2268 {
2269 	struct nlmsghdr *nlh;
2270 	struct ifinfomsg *ifm;
2271 	struct nlattr *br_afspec;
2272 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2273 
2274 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2275 	if (nlh == NULL)
2276 		return -EMSGSIZE;
2277 
2278 	ifm = nlmsg_data(nlh);
2279 	ifm->ifi_family = AF_BRIDGE;
2280 	ifm->__ifi_pad = 0;
2281 	ifm->ifi_type = dev->type;
2282 	ifm->ifi_index = dev->ifindex;
2283 	ifm->ifi_flags = dev_get_flags(dev);
2284 	ifm->ifi_change = 0;
2285 
2286 
2287 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2288 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2289 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2290 	    (dev->master &&
2291 	     nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
2292 	    (dev->addr_len &&
2293 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2294 	    (dev->ifindex != dev->iflink &&
2295 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2296 		goto nla_put_failure;
2297 
2298 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2299 	if (!br_afspec)
2300 		goto nla_put_failure;
2301 
2302 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2303 	    nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2304 		nla_nest_cancel(skb, br_afspec);
2305 		goto nla_put_failure;
2306 	}
2307 	nla_nest_end(skb, br_afspec);
2308 
2309 	return nlmsg_end(skb, nlh);
2310 nla_put_failure:
2311 	nlmsg_cancel(skb, nlh);
2312 	return -EMSGSIZE;
2313 }
2314 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2315 
2316 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2317 {
2318 	struct net *net = sock_net(skb->sk);
2319 	struct net_device *dev;
2320 	int idx = 0;
2321 	u32 portid = NETLINK_CB(cb->skb).portid;
2322 	u32 seq = cb->nlh->nlmsg_seq;
2323 
2324 	rcu_read_lock();
2325 	for_each_netdev_rcu(net, dev) {
2326 		const struct net_device_ops *ops = dev->netdev_ops;
2327 		struct net_device *master = dev->master;
2328 
2329 		if (master && master->netdev_ops->ndo_bridge_getlink) {
2330 			if (idx >= cb->args[0] &&
2331 			    master->netdev_ops->ndo_bridge_getlink(
2332 				    skb, portid, seq, dev) < 0)
2333 				break;
2334 			idx++;
2335 		}
2336 
2337 		if (ops->ndo_bridge_getlink) {
2338 			if (idx >= cb->args[0] &&
2339 			    ops->ndo_bridge_getlink(skb, portid, seq, dev) < 0)
2340 				break;
2341 			idx++;
2342 		}
2343 	}
2344 	rcu_read_unlock();
2345 	cb->args[0] = idx;
2346 
2347 	return skb->len;
2348 }
2349 
2350 static inline size_t bridge_nlmsg_size(void)
2351 {
2352 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2353 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2354 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2355 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2356 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2357 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2358 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2359 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2360 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2361 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2362 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2363 }
2364 
2365 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2366 {
2367 	struct net *net = dev_net(dev);
2368 	struct net_device *master = dev->master;
2369 	struct sk_buff *skb;
2370 	int err = -EOPNOTSUPP;
2371 
2372 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2373 	if (!skb) {
2374 		err = -ENOMEM;
2375 		goto errout;
2376 	}
2377 
2378 	if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2379 	    master && master->netdev_ops->ndo_bridge_getlink) {
2380 		err = master->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev);
2381 		if (err < 0)
2382 			goto errout;
2383 	}
2384 
2385 	if ((flags & BRIDGE_FLAGS_SELF) &&
2386 	    dev->netdev_ops->ndo_bridge_getlink) {
2387 		err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev);
2388 		if (err < 0)
2389 			goto errout;
2390 	}
2391 
2392 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2393 	return 0;
2394 errout:
2395 	WARN_ON(err == -EMSGSIZE);
2396 	kfree_skb(skb);
2397 	rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2398 	return err;
2399 }
2400 
2401 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2402 			       void *arg)
2403 {
2404 	struct net *net = sock_net(skb->sk);
2405 	struct ifinfomsg *ifm;
2406 	struct net_device *dev;
2407 	struct nlattr *br_spec, *attr = NULL;
2408 	int rem, err = -EOPNOTSUPP;
2409 	u16 oflags, flags = 0;
2410 	bool have_flags = false;
2411 
2412 	if (nlmsg_len(nlh) < sizeof(*ifm))
2413 		return -EINVAL;
2414 
2415 	ifm = nlmsg_data(nlh);
2416 	if (ifm->ifi_family != AF_BRIDGE)
2417 		return -EPFNOSUPPORT;
2418 
2419 	dev = __dev_get_by_index(net, ifm->ifi_index);
2420 	if (!dev) {
2421 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2422 		return -ENODEV;
2423 	}
2424 
2425 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2426 	if (br_spec) {
2427 		nla_for_each_nested(attr, br_spec, rem) {
2428 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2429 				have_flags = true;
2430 				flags = nla_get_u16(attr);
2431 				break;
2432 			}
2433 		}
2434 	}
2435 
2436 	oflags = flags;
2437 
2438 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2439 		if (!dev->master ||
2440 		    !dev->master->netdev_ops->ndo_bridge_setlink) {
2441 			err = -EOPNOTSUPP;
2442 			goto out;
2443 		}
2444 
2445 		err = dev->master->netdev_ops->ndo_bridge_setlink(dev, nlh);
2446 		if (err)
2447 			goto out;
2448 
2449 		flags &= ~BRIDGE_FLAGS_MASTER;
2450 	}
2451 
2452 	if ((flags & BRIDGE_FLAGS_SELF)) {
2453 		if (!dev->netdev_ops->ndo_bridge_setlink)
2454 			err = -EOPNOTSUPP;
2455 		else
2456 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2457 
2458 		if (!err)
2459 			flags &= ~BRIDGE_FLAGS_SELF;
2460 	}
2461 
2462 	if (have_flags)
2463 		memcpy(nla_data(attr), &flags, sizeof(flags));
2464 	/* Generate event to notify upper layer of bridge change */
2465 	if (!err)
2466 		err = rtnl_bridge_notify(dev, oflags);
2467 out:
2468 	return err;
2469 }
2470 
2471 /* Protected by RTNL sempahore.  */
2472 static struct rtattr **rta_buf;
2473 static int rtattr_max;
2474 
2475 /* Process one rtnetlink message. */
2476 
2477 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2478 {
2479 	struct net *net = sock_net(skb->sk);
2480 	rtnl_doit_func doit;
2481 	int sz_idx, kind;
2482 	int min_len;
2483 	int family;
2484 	int type;
2485 	int err;
2486 
2487 	type = nlh->nlmsg_type;
2488 	if (type > RTM_MAX)
2489 		return -EOPNOTSUPP;
2490 
2491 	type -= RTM_BASE;
2492 
2493 	/* All the messages must have at least 1 byte length */
2494 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2495 		return 0;
2496 
2497 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2498 	sz_idx = type>>2;
2499 	kind = type&3;
2500 
2501 	if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2502 		return -EPERM;
2503 
2504 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2505 		struct sock *rtnl;
2506 		rtnl_dumpit_func dumpit;
2507 		rtnl_calcit_func calcit;
2508 		u16 min_dump_alloc = 0;
2509 
2510 		dumpit = rtnl_get_dumpit(family, type);
2511 		if (dumpit == NULL)
2512 			return -EOPNOTSUPP;
2513 		calcit = rtnl_get_calcit(family, type);
2514 		if (calcit)
2515 			min_dump_alloc = calcit(skb, nlh);
2516 
2517 		__rtnl_unlock();
2518 		rtnl = net->rtnl;
2519 		{
2520 			struct netlink_dump_control c = {
2521 				.dump		= dumpit,
2522 				.min_dump_alloc	= min_dump_alloc,
2523 			};
2524 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2525 		}
2526 		rtnl_lock();
2527 		return err;
2528 	}
2529 
2530 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2531 
2532 	min_len = rtm_min[sz_idx];
2533 	if (nlh->nlmsg_len < min_len)
2534 		return -EINVAL;
2535 
2536 	if (nlh->nlmsg_len > min_len) {
2537 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2538 		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2539 
2540 		while (RTA_OK(attr, attrlen)) {
2541 			unsigned int flavor = attr->rta_type;
2542 			if (flavor) {
2543 				if (flavor > rta_max[sz_idx])
2544 					return -EINVAL;
2545 				rta_buf[flavor-1] = attr;
2546 			}
2547 			attr = RTA_NEXT(attr, attrlen);
2548 		}
2549 	}
2550 
2551 	doit = rtnl_get_doit(family, type);
2552 	if (doit == NULL)
2553 		return -EOPNOTSUPP;
2554 
2555 	return doit(skb, nlh, (void *)&rta_buf[0]);
2556 }
2557 
2558 static void rtnetlink_rcv(struct sk_buff *skb)
2559 {
2560 	rtnl_lock();
2561 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2562 	rtnl_unlock();
2563 }
2564 
2565 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2566 {
2567 	struct net_device *dev = ptr;
2568 
2569 	switch (event) {
2570 	case NETDEV_UP:
2571 	case NETDEV_DOWN:
2572 	case NETDEV_PRE_UP:
2573 	case NETDEV_POST_INIT:
2574 	case NETDEV_REGISTER:
2575 	case NETDEV_CHANGE:
2576 	case NETDEV_PRE_TYPE_CHANGE:
2577 	case NETDEV_GOING_DOWN:
2578 	case NETDEV_UNREGISTER:
2579 	case NETDEV_UNREGISTER_FINAL:
2580 	case NETDEV_RELEASE:
2581 	case NETDEV_JOIN:
2582 		break;
2583 	default:
2584 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2585 		break;
2586 	}
2587 	return NOTIFY_DONE;
2588 }
2589 
2590 static struct notifier_block rtnetlink_dev_notifier = {
2591 	.notifier_call	= rtnetlink_event,
2592 };
2593 
2594 
2595 static int __net_init rtnetlink_net_init(struct net *net)
2596 {
2597 	struct sock *sk;
2598 	struct netlink_kernel_cfg cfg = {
2599 		.groups		= RTNLGRP_MAX,
2600 		.input		= rtnetlink_rcv,
2601 		.cb_mutex	= &rtnl_mutex,
2602 		.flags		= NL_CFG_F_NONROOT_RECV,
2603 	};
2604 
2605 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2606 	if (!sk)
2607 		return -ENOMEM;
2608 	net->rtnl = sk;
2609 	return 0;
2610 }
2611 
2612 static void __net_exit rtnetlink_net_exit(struct net *net)
2613 {
2614 	netlink_kernel_release(net->rtnl);
2615 	net->rtnl = NULL;
2616 }
2617 
2618 static struct pernet_operations rtnetlink_net_ops = {
2619 	.init = rtnetlink_net_init,
2620 	.exit = rtnetlink_net_exit,
2621 };
2622 
2623 void __init rtnetlink_init(void)
2624 {
2625 	int i;
2626 
2627 	rtattr_max = 0;
2628 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2629 		if (rta_max[i] > rtattr_max)
2630 			rtattr_max = rta_max[i];
2631 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2632 	if (!rta_buf)
2633 		panic("rtnetlink_init: cannot allocate rta_buf\n");
2634 
2635 	if (register_pernet_subsys(&rtnetlink_net_ops))
2636 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2637 
2638 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2639 
2640 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2641 		      rtnl_dump_ifinfo, rtnl_calcit);
2642 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2643 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2644 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2645 
2646 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2647 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2648 
2649 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2650 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2651 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2652 
2653 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2654 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2655 }
2656 
2657