xref: /linux/net/core/rtnetlink.c (revision 26b0d14106954ae46d2f4f7eec3481828a210f7d)
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 ? tab[msgindex].doit : NULL;
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 ? tab[msgindex].dumpit : NULL;
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 ? tab[msgindex].calcit : NULL;
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 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
545 {
546 	struct rtattr *rta;
547 	int size = RTA_LENGTH(attrlen);
548 
549 	rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
550 	rta->rta_type = attrtype;
551 	rta->rta_len = size;
552 	memcpy(RTA_DATA(rta), data, attrlen);
553 	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
554 }
555 EXPORT_SYMBOL(__rta_fill);
556 
557 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
558 {
559 	struct sock *rtnl = net->rtnl;
560 	int err = 0;
561 
562 	NETLINK_CB(skb).dst_group = group;
563 	if (echo)
564 		atomic_inc(&skb->users);
565 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
566 	if (echo)
567 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
568 	return err;
569 }
570 
571 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
572 {
573 	struct sock *rtnl = net->rtnl;
574 
575 	return nlmsg_unicast(rtnl, skb, pid);
576 }
577 EXPORT_SYMBOL(rtnl_unicast);
578 
579 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
580 		 struct nlmsghdr *nlh, gfp_t flags)
581 {
582 	struct sock *rtnl = net->rtnl;
583 	int report = 0;
584 
585 	if (nlh)
586 		report = nlmsg_report(nlh);
587 
588 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
589 }
590 EXPORT_SYMBOL(rtnl_notify);
591 
592 void rtnl_set_sk_err(struct net *net, u32 group, int error)
593 {
594 	struct sock *rtnl = net->rtnl;
595 
596 	netlink_set_err(rtnl, 0, group, error);
597 }
598 EXPORT_SYMBOL(rtnl_set_sk_err);
599 
600 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
601 {
602 	struct nlattr *mx;
603 	int i, valid = 0;
604 
605 	mx = nla_nest_start(skb, RTA_METRICS);
606 	if (mx == NULL)
607 		return -ENOBUFS;
608 
609 	for (i = 0; i < RTAX_MAX; i++) {
610 		if (metrics[i]) {
611 			valid++;
612 			if (nla_put_u32(skb, i+1, metrics[i]))
613 				goto nla_put_failure;
614 		}
615 	}
616 
617 	if (!valid) {
618 		nla_nest_cancel(skb, mx);
619 		return 0;
620 	}
621 
622 	return nla_nest_end(skb, mx);
623 
624 nla_put_failure:
625 	nla_nest_cancel(skb, mx);
626 	return -EMSGSIZE;
627 }
628 EXPORT_SYMBOL(rtnetlink_put_metrics);
629 
630 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
631 		       u32 ts, u32 tsage, long expires, u32 error)
632 {
633 	struct rta_cacheinfo ci = {
634 		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
635 		.rta_used = dst->__use,
636 		.rta_clntref = atomic_read(&(dst->__refcnt)),
637 		.rta_error = error,
638 		.rta_id =  id,
639 		.rta_ts = ts,
640 		.rta_tsage = tsage,
641 	};
642 
643 	if (expires)
644 		ci.rta_expires = jiffies_to_clock_t(expires);
645 
646 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
647 }
648 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
649 
650 static void set_operstate(struct net_device *dev, unsigned char transition)
651 {
652 	unsigned char operstate = dev->operstate;
653 
654 	switch (transition) {
655 	case IF_OPER_UP:
656 		if ((operstate == IF_OPER_DORMANT ||
657 		     operstate == IF_OPER_UNKNOWN) &&
658 		    !netif_dormant(dev))
659 			operstate = IF_OPER_UP;
660 		break;
661 
662 	case IF_OPER_DORMANT:
663 		if (operstate == IF_OPER_UP ||
664 		    operstate == IF_OPER_UNKNOWN)
665 			operstate = IF_OPER_DORMANT;
666 		break;
667 	}
668 
669 	if (dev->operstate != operstate) {
670 		write_lock_bh(&dev_base_lock);
671 		dev->operstate = operstate;
672 		write_unlock_bh(&dev_base_lock);
673 		netdev_state_change(dev);
674 	}
675 }
676 
677 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
678 					   const struct ifinfomsg *ifm)
679 {
680 	unsigned int flags = ifm->ifi_flags;
681 
682 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
683 	if (ifm->ifi_change)
684 		flags = (flags & ifm->ifi_change) |
685 			(dev->flags & ~ifm->ifi_change);
686 
687 	return flags;
688 }
689 
690 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
691 				 const struct rtnl_link_stats64 *b)
692 {
693 	a->rx_packets = b->rx_packets;
694 	a->tx_packets = b->tx_packets;
695 	a->rx_bytes = b->rx_bytes;
696 	a->tx_bytes = b->tx_bytes;
697 	a->rx_errors = b->rx_errors;
698 	a->tx_errors = b->tx_errors;
699 	a->rx_dropped = b->rx_dropped;
700 	a->tx_dropped = b->tx_dropped;
701 
702 	a->multicast = b->multicast;
703 	a->collisions = b->collisions;
704 
705 	a->rx_length_errors = b->rx_length_errors;
706 	a->rx_over_errors = b->rx_over_errors;
707 	a->rx_crc_errors = b->rx_crc_errors;
708 	a->rx_frame_errors = b->rx_frame_errors;
709 	a->rx_fifo_errors = b->rx_fifo_errors;
710 	a->rx_missed_errors = b->rx_missed_errors;
711 
712 	a->tx_aborted_errors = b->tx_aborted_errors;
713 	a->tx_carrier_errors = b->tx_carrier_errors;
714 	a->tx_fifo_errors = b->tx_fifo_errors;
715 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
716 	a->tx_window_errors = b->tx_window_errors;
717 
718 	a->rx_compressed = b->rx_compressed;
719 	a->tx_compressed = b->tx_compressed;
720 }
721 
722 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
723 {
724 	memcpy(v, b, sizeof(*b));
725 }
726 
727 /* All VF info */
728 static inline int rtnl_vfinfo_size(const struct net_device *dev,
729 				   u32 ext_filter_mask)
730 {
731 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
732 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
733 		int num_vfs = dev_num_vf(dev->dev.parent);
734 		size_t size = nla_total_size(sizeof(struct nlattr));
735 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
736 		size += num_vfs *
737 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
738 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
739 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
740 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
741 		return size;
742 	} else
743 		return 0;
744 }
745 
746 static size_t rtnl_port_size(const struct net_device *dev)
747 {
748 	size_t port_size = nla_total_size(4)		/* PORT_VF */
749 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
750 		+ nla_total_size(sizeof(struct ifla_port_vsi))
751 							/* PORT_VSI_TYPE */
752 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
753 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
754 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
755 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
756 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
757 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
758 		+ port_size;
759 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
760 		+ port_size;
761 
762 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
763 		return 0;
764 	if (dev_num_vf(dev->dev.parent))
765 		return port_self_size + vf_ports_size +
766 			vf_port_size * dev_num_vf(dev->dev.parent);
767 	else
768 		return port_self_size;
769 }
770 
771 static noinline size_t if_nlmsg_size(const struct net_device *dev,
772 				     u32 ext_filter_mask)
773 {
774 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
775 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
776 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
777 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
778 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
779 	       + nla_total_size(sizeof(struct rtnl_link_stats))
780 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
781 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
782 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
783 	       + nla_total_size(4) /* IFLA_TXQLEN */
784 	       + nla_total_size(4) /* IFLA_WEIGHT */
785 	       + nla_total_size(4) /* IFLA_MTU */
786 	       + nla_total_size(4) /* IFLA_LINK */
787 	       + nla_total_size(4) /* IFLA_MASTER */
788 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
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 
886 	ASSERT_RTNL();
887 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
888 	if (nlh == NULL)
889 		return -EMSGSIZE;
890 
891 	ifm = nlmsg_data(nlh);
892 	ifm->ifi_family = AF_UNSPEC;
893 	ifm->__ifi_pad = 0;
894 	ifm->ifi_type = dev->type;
895 	ifm->ifi_index = dev->ifindex;
896 	ifm->ifi_flags = dev_get_flags(dev);
897 	ifm->ifi_change = change;
898 
899 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
900 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
901 	    nla_put_u8(skb, IFLA_OPERSTATE,
902 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
903 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
904 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
905 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
906 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
907 	    (dev->ifindex != dev->iflink &&
908 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
909 	    (dev->master &&
910 	     nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
911 	    (dev->qdisc &&
912 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
913 	    (dev->ifalias &&
914 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
915 		goto nla_put_failure;
916 
917 	if (1) {
918 		struct rtnl_link_ifmap map = {
919 			.mem_start   = dev->mem_start,
920 			.mem_end     = dev->mem_end,
921 			.base_addr   = dev->base_addr,
922 			.irq         = dev->irq,
923 			.dma         = dev->dma,
924 			.port        = dev->if_port,
925 		};
926 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
927 			goto nla_put_failure;
928 	}
929 
930 	if (dev->addr_len) {
931 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
932 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
933 			goto nla_put_failure;
934 	}
935 
936 	attr = nla_reserve(skb, IFLA_STATS,
937 			sizeof(struct rtnl_link_stats));
938 	if (attr == NULL)
939 		goto nla_put_failure;
940 
941 	stats = dev_get_stats(dev, &temp);
942 	copy_rtnl_link_stats(nla_data(attr), stats);
943 
944 	attr = nla_reserve(skb, IFLA_STATS64,
945 			sizeof(struct rtnl_link_stats64));
946 	if (attr == NULL)
947 		goto nla_put_failure;
948 	copy_rtnl_link_stats64(nla_data(attr), stats);
949 
950 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
951 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
952 		goto nla_put_failure;
953 
954 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
955 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
956 		int i;
957 
958 		struct nlattr *vfinfo, *vf;
959 		int num_vfs = dev_num_vf(dev->dev.parent);
960 
961 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
962 		if (!vfinfo)
963 			goto nla_put_failure;
964 		for (i = 0; i < num_vfs; i++) {
965 			struct ifla_vf_info ivi;
966 			struct ifla_vf_mac vf_mac;
967 			struct ifla_vf_vlan vf_vlan;
968 			struct ifla_vf_tx_rate vf_tx_rate;
969 			struct ifla_vf_spoofchk vf_spoofchk;
970 
971 			/*
972 			 * Not all SR-IOV capable drivers support the
973 			 * spoofcheck query.  Preset to -1 so the user
974 			 * space tool can detect that the driver didn't
975 			 * report anything.
976 			 */
977 			ivi.spoofchk = -1;
978 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
979 				break;
980 			vf_mac.vf =
981 				vf_vlan.vf =
982 				vf_tx_rate.vf =
983 				vf_spoofchk.vf = ivi.vf;
984 
985 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
986 			vf_vlan.vlan = ivi.vlan;
987 			vf_vlan.qos = ivi.qos;
988 			vf_tx_rate.rate = ivi.tx_rate;
989 			vf_spoofchk.setting = ivi.spoofchk;
990 			vf = nla_nest_start(skb, IFLA_VF_INFO);
991 			if (!vf) {
992 				nla_nest_cancel(skb, vfinfo);
993 				goto nla_put_failure;
994 			}
995 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
996 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
997 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
998 				    &vf_tx_rate) ||
999 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1000 				    &vf_spoofchk))
1001 				goto nla_put_failure;
1002 			nla_nest_end(skb, vf);
1003 		}
1004 		nla_nest_end(skb, vfinfo);
1005 	}
1006 
1007 	if (rtnl_port_fill(skb, dev))
1008 		goto nla_put_failure;
1009 
1010 	if (dev->rtnl_link_ops) {
1011 		if (rtnl_link_fill(skb, dev) < 0)
1012 			goto nla_put_failure;
1013 	}
1014 
1015 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1016 		goto nla_put_failure;
1017 
1018 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1019 		if (af_ops->fill_link_af) {
1020 			struct nlattr *af;
1021 			int err;
1022 
1023 			if (!(af = nla_nest_start(skb, af_ops->family)))
1024 				goto nla_put_failure;
1025 
1026 			err = af_ops->fill_link_af(skb, dev);
1027 
1028 			/*
1029 			 * Caller may return ENODATA to indicate that there
1030 			 * was no data to be dumped. This is not an error, it
1031 			 * means we should trim the attribute header and
1032 			 * continue.
1033 			 */
1034 			if (err == -ENODATA)
1035 				nla_nest_cancel(skb, af);
1036 			else if (err < 0)
1037 				goto nla_put_failure;
1038 
1039 			nla_nest_end(skb, af);
1040 		}
1041 	}
1042 
1043 	nla_nest_end(skb, af_spec);
1044 
1045 	return nlmsg_end(skb, nlh);
1046 
1047 nla_put_failure:
1048 	nlmsg_cancel(skb, nlh);
1049 	return -EMSGSIZE;
1050 }
1051 
1052 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1053 {
1054 	struct net *net = sock_net(skb->sk);
1055 	int h, s_h;
1056 	int idx = 0, s_idx;
1057 	struct net_device *dev;
1058 	struct hlist_head *head;
1059 	struct hlist_node *node;
1060 	struct nlattr *tb[IFLA_MAX+1];
1061 	u32 ext_filter_mask = 0;
1062 
1063 	s_h = cb->args[0];
1064 	s_idx = cb->args[1];
1065 
1066 	rcu_read_lock();
1067 	cb->seq = net->dev_base_seq;
1068 
1069 	if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1070 			ifla_policy) >= 0) {
1071 
1072 		if (tb[IFLA_EXT_MASK])
1073 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1074 	}
1075 
1076 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1077 		idx = 0;
1078 		head = &net->dev_index_head[h];
1079 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1080 			if (idx < s_idx)
1081 				goto cont;
1082 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1083 					     NETLINK_CB(cb->skb).pid,
1084 					     cb->nlh->nlmsg_seq, 0,
1085 					     NLM_F_MULTI,
1086 					     ext_filter_mask) <= 0)
1087 				goto out;
1088 
1089 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1090 cont:
1091 			idx++;
1092 		}
1093 	}
1094 out:
1095 	rcu_read_unlock();
1096 	cb->args[1] = idx;
1097 	cb->args[0] = h;
1098 
1099 	return skb->len;
1100 }
1101 
1102 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1103 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1104 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1105 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1106 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1107 	[IFLA_MTU]		= { .type = NLA_U32 },
1108 	[IFLA_LINK]		= { .type = NLA_U32 },
1109 	[IFLA_MASTER]		= { .type = NLA_U32 },
1110 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1111 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1112 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1113 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1114 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1115 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1116 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1117 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1118 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1119 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1120 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1121 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1122 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1123 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1124 };
1125 EXPORT_SYMBOL(ifla_policy);
1126 
1127 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1128 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1129 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1130 };
1131 
1132 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1133 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1134 };
1135 
1136 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1137 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1138 				    .len = sizeof(struct ifla_vf_mac) },
1139 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1140 				    .len = sizeof(struct ifla_vf_vlan) },
1141 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1142 				    .len = sizeof(struct ifla_vf_tx_rate) },
1143 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1144 				    .len = sizeof(struct ifla_vf_spoofchk) },
1145 };
1146 
1147 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1148 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1149 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1150 				    .len = PORT_PROFILE_MAX },
1151 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1152 				    .len = sizeof(struct ifla_port_vsi)},
1153 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1154 				      .len = PORT_UUID_MAX },
1155 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1156 				    .len = PORT_UUID_MAX },
1157 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1158 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1159 };
1160 
1161 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1162 {
1163 	struct net *net;
1164 	/* Examine the link attributes and figure out which
1165 	 * network namespace we are talking about.
1166 	 */
1167 	if (tb[IFLA_NET_NS_PID])
1168 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1169 	else if (tb[IFLA_NET_NS_FD])
1170 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1171 	else
1172 		net = get_net(src_net);
1173 	return net;
1174 }
1175 EXPORT_SYMBOL(rtnl_link_get_net);
1176 
1177 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1178 {
1179 	if (dev) {
1180 		if (tb[IFLA_ADDRESS] &&
1181 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1182 			return -EINVAL;
1183 
1184 		if (tb[IFLA_BROADCAST] &&
1185 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1186 			return -EINVAL;
1187 	}
1188 
1189 	if (tb[IFLA_AF_SPEC]) {
1190 		struct nlattr *af;
1191 		int rem, err;
1192 
1193 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1194 			const struct rtnl_af_ops *af_ops;
1195 
1196 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1197 				return -EAFNOSUPPORT;
1198 
1199 			if (!af_ops->set_link_af)
1200 				return -EOPNOTSUPP;
1201 
1202 			if (af_ops->validate_link_af) {
1203 				err = af_ops->validate_link_af(dev, af);
1204 				if (err < 0)
1205 					return err;
1206 			}
1207 		}
1208 	}
1209 
1210 	return 0;
1211 }
1212 
1213 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1214 {
1215 	int rem, err = -EINVAL;
1216 	struct nlattr *vf;
1217 	const struct net_device_ops *ops = dev->netdev_ops;
1218 
1219 	nla_for_each_nested(vf, attr, rem) {
1220 		switch (nla_type(vf)) {
1221 		case IFLA_VF_MAC: {
1222 			struct ifla_vf_mac *ivm;
1223 			ivm = nla_data(vf);
1224 			err = -EOPNOTSUPP;
1225 			if (ops->ndo_set_vf_mac)
1226 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1227 							  ivm->mac);
1228 			break;
1229 		}
1230 		case IFLA_VF_VLAN: {
1231 			struct ifla_vf_vlan *ivv;
1232 			ivv = nla_data(vf);
1233 			err = -EOPNOTSUPP;
1234 			if (ops->ndo_set_vf_vlan)
1235 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1236 							   ivv->vlan,
1237 							   ivv->qos);
1238 			break;
1239 		}
1240 		case IFLA_VF_TX_RATE: {
1241 			struct ifla_vf_tx_rate *ivt;
1242 			ivt = nla_data(vf);
1243 			err = -EOPNOTSUPP;
1244 			if (ops->ndo_set_vf_tx_rate)
1245 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1246 							      ivt->rate);
1247 			break;
1248 		}
1249 		case IFLA_VF_SPOOFCHK: {
1250 			struct ifla_vf_spoofchk *ivs;
1251 			ivs = nla_data(vf);
1252 			err = -EOPNOTSUPP;
1253 			if (ops->ndo_set_vf_spoofchk)
1254 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1255 							       ivs->setting);
1256 			break;
1257 		}
1258 		default:
1259 			err = -EINVAL;
1260 			break;
1261 		}
1262 		if (err)
1263 			break;
1264 	}
1265 	return err;
1266 }
1267 
1268 static int do_set_master(struct net_device *dev, int ifindex)
1269 {
1270 	struct net_device *master_dev;
1271 	const struct net_device_ops *ops;
1272 	int err;
1273 
1274 	if (dev->master) {
1275 		if (dev->master->ifindex == ifindex)
1276 			return 0;
1277 		ops = dev->master->netdev_ops;
1278 		if (ops->ndo_del_slave) {
1279 			err = ops->ndo_del_slave(dev->master, dev);
1280 			if (err)
1281 				return err;
1282 		} else {
1283 			return -EOPNOTSUPP;
1284 		}
1285 	}
1286 
1287 	if (ifindex) {
1288 		master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1289 		if (!master_dev)
1290 			return -EINVAL;
1291 		ops = master_dev->netdev_ops;
1292 		if (ops->ndo_add_slave) {
1293 			err = ops->ndo_add_slave(master_dev, dev);
1294 			if (err)
1295 				return err;
1296 		} else {
1297 			return -EOPNOTSUPP;
1298 		}
1299 	}
1300 	return 0;
1301 }
1302 
1303 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1304 		      struct nlattr **tb, char *ifname, int modified)
1305 {
1306 	const struct net_device_ops *ops = dev->netdev_ops;
1307 	int send_addr_notify = 0;
1308 	int err;
1309 
1310 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1311 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1312 		if (IS_ERR(net)) {
1313 			err = PTR_ERR(net);
1314 			goto errout;
1315 		}
1316 		err = dev_change_net_namespace(dev, net, ifname);
1317 		put_net(net);
1318 		if (err)
1319 			goto errout;
1320 		modified = 1;
1321 	}
1322 
1323 	if (tb[IFLA_MAP]) {
1324 		struct rtnl_link_ifmap *u_map;
1325 		struct ifmap k_map;
1326 
1327 		if (!ops->ndo_set_config) {
1328 			err = -EOPNOTSUPP;
1329 			goto errout;
1330 		}
1331 
1332 		if (!netif_device_present(dev)) {
1333 			err = -ENODEV;
1334 			goto errout;
1335 		}
1336 
1337 		u_map = nla_data(tb[IFLA_MAP]);
1338 		k_map.mem_start = (unsigned long) u_map->mem_start;
1339 		k_map.mem_end = (unsigned long) u_map->mem_end;
1340 		k_map.base_addr = (unsigned short) u_map->base_addr;
1341 		k_map.irq = (unsigned char) u_map->irq;
1342 		k_map.dma = (unsigned char) u_map->dma;
1343 		k_map.port = (unsigned char) u_map->port;
1344 
1345 		err = ops->ndo_set_config(dev, &k_map);
1346 		if (err < 0)
1347 			goto errout;
1348 
1349 		modified = 1;
1350 	}
1351 
1352 	if (tb[IFLA_ADDRESS]) {
1353 		struct sockaddr *sa;
1354 		int len;
1355 
1356 		if (!ops->ndo_set_mac_address) {
1357 			err = -EOPNOTSUPP;
1358 			goto errout;
1359 		}
1360 
1361 		if (!netif_device_present(dev)) {
1362 			err = -ENODEV;
1363 			goto errout;
1364 		}
1365 
1366 		len = sizeof(sa_family_t) + dev->addr_len;
1367 		sa = kmalloc(len, GFP_KERNEL);
1368 		if (!sa) {
1369 			err = -ENOMEM;
1370 			goto errout;
1371 		}
1372 		sa->sa_family = dev->type;
1373 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1374 		       dev->addr_len);
1375 		err = ops->ndo_set_mac_address(dev, sa);
1376 		kfree(sa);
1377 		if (err)
1378 			goto errout;
1379 		send_addr_notify = 1;
1380 		modified = 1;
1381 	}
1382 
1383 	if (tb[IFLA_MTU]) {
1384 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1385 		if (err < 0)
1386 			goto errout;
1387 		modified = 1;
1388 	}
1389 
1390 	if (tb[IFLA_GROUP]) {
1391 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1392 		modified = 1;
1393 	}
1394 
1395 	/*
1396 	 * Interface selected by interface index but interface
1397 	 * name provided implies that a name change has been
1398 	 * requested.
1399 	 */
1400 	if (ifm->ifi_index > 0 && ifname[0]) {
1401 		err = dev_change_name(dev, ifname);
1402 		if (err < 0)
1403 			goto errout;
1404 		modified = 1;
1405 	}
1406 
1407 	if (tb[IFLA_IFALIAS]) {
1408 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1409 				    nla_len(tb[IFLA_IFALIAS]));
1410 		if (err < 0)
1411 			goto errout;
1412 		modified = 1;
1413 	}
1414 
1415 	if (tb[IFLA_BROADCAST]) {
1416 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1417 		send_addr_notify = 1;
1418 	}
1419 
1420 	if (ifm->ifi_flags || ifm->ifi_change) {
1421 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1422 		if (err < 0)
1423 			goto errout;
1424 	}
1425 
1426 	if (tb[IFLA_MASTER]) {
1427 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1428 		if (err)
1429 			goto errout;
1430 		modified = 1;
1431 	}
1432 
1433 	if (tb[IFLA_TXQLEN])
1434 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1435 
1436 	if (tb[IFLA_OPERSTATE])
1437 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1438 
1439 	if (tb[IFLA_LINKMODE]) {
1440 		write_lock_bh(&dev_base_lock);
1441 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1442 		write_unlock_bh(&dev_base_lock);
1443 	}
1444 
1445 	if (tb[IFLA_VFINFO_LIST]) {
1446 		struct nlattr *attr;
1447 		int rem;
1448 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1449 			if (nla_type(attr) != IFLA_VF_INFO) {
1450 				err = -EINVAL;
1451 				goto errout;
1452 			}
1453 			err = do_setvfinfo(dev, attr);
1454 			if (err < 0)
1455 				goto errout;
1456 			modified = 1;
1457 		}
1458 	}
1459 	err = 0;
1460 
1461 	if (tb[IFLA_VF_PORTS]) {
1462 		struct nlattr *port[IFLA_PORT_MAX+1];
1463 		struct nlattr *attr;
1464 		int vf;
1465 		int rem;
1466 
1467 		err = -EOPNOTSUPP;
1468 		if (!ops->ndo_set_vf_port)
1469 			goto errout;
1470 
1471 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1472 			if (nla_type(attr) != IFLA_VF_PORT)
1473 				continue;
1474 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1475 				attr, ifla_port_policy);
1476 			if (err < 0)
1477 				goto errout;
1478 			if (!port[IFLA_PORT_VF]) {
1479 				err = -EOPNOTSUPP;
1480 				goto errout;
1481 			}
1482 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1483 			err = ops->ndo_set_vf_port(dev, vf, port);
1484 			if (err < 0)
1485 				goto errout;
1486 			modified = 1;
1487 		}
1488 	}
1489 	err = 0;
1490 
1491 	if (tb[IFLA_PORT_SELF]) {
1492 		struct nlattr *port[IFLA_PORT_MAX+1];
1493 
1494 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1495 			tb[IFLA_PORT_SELF], ifla_port_policy);
1496 		if (err < 0)
1497 			goto errout;
1498 
1499 		err = -EOPNOTSUPP;
1500 		if (ops->ndo_set_vf_port)
1501 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1502 		if (err < 0)
1503 			goto errout;
1504 		modified = 1;
1505 	}
1506 
1507 	if (tb[IFLA_AF_SPEC]) {
1508 		struct nlattr *af;
1509 		int rem;
1510 
1511 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1512 			const struct rtnl_af_ops *af_ops;
1513 
1514 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1515 				BUG();
1516 
1517 			err = af_ops->set_link_af(dev, af);
1518 			if (err < 0)
1519 				goto errout;
1520 
1521 			modified = 1;
1522 		}
1523 	}
1524 	err = 0;
1525 
1526 errout:
1527 	if (err < 0 && modified)
1528 		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",
1529 				     dev->name);
1530 
1531 	if (send_addr_notify)
1532 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1533 
1534 	return err;
1535 }
1536 
1537 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1538 {
1539 	struct net *net = sock_net(skb->sk);
1540 	struct ifinfomsg *ifm;
1541 	struct net_device *dev;
1542 	int err;
1543 	struct nlattr *tb[IFLA_MAX+1];
1544 	char ifname[IFNAMSIZ];
1545 
1546 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1547 	if (err < 0)
1548 		goto errout;
1549 
1550 	if (tb[IFLA_IFNAME])
1551 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1552 	else
1553 		ifname[0] = '\0';
1554 
1555 	err = -EINVAL;
1556 	ifm = nlmsg_data(nlh);
1557 	if (ifm->ifi_index > 0)
1558 		dev = __dev_get_by_index(net, ifm->ifi_index);
1559 	else if (tb[IFLA_IFNAME])
1560 		dev = __dev_get_by_name(net, ifname);
1561 	else
1562 		goto errout;
1563 
1564 	if (dev == NULL) {
1565 		err = -ENODEV;
1566 		goto errout;
1567 	}
1568 
1569 	err = validate_linkmsg(dev, tb);
1570 	if (err < 0)
1571 		goto errout;
1572 
1573 	err = do_setlink(dev, ifm, tb, ifname, 0);
1574 errout:
1575 	return err;
1576 }
1577 
1578 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1579 {
1580 	struct net *net = sock_net(skb->sk);
1581 	const struct rtnl_link_ops *ops;
1582 	struct net_device *dev;
1583 	struct ifinfomsg *ifm;
1584 	char ifname[IFNAMSIZ];
1585 	struct nlattr *tb[IFLA_MAX+1];
1586 	int err;
1587 	LIST_HEAD(list_kill);
1588 
1589 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1590 	if (err < 0)
1591 		return err;
1592 
1593 	if (tb[IFLA_IFNAME])
1594 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1595 
1596 	ifm = nlmsg_data(nlh);
1597 	if (ifm->ifi_index > 0)
1598 		dev = __dev_get_by_index(net, ifm->ifi_index);
1599 	else if (tb[IFLA_IFNAME])
1600 		dev = __dev_get_by_name(net, ifname);
1601 	else
1602 		return -EINVAL;
1603 
1604 	if (!dev)
1605 		return -ENODEV;
1606 
1607 	ops = dev->rtnl_link_ops;
1608 	if (!ops)
1609 		return -EOPNOTSUPP;
1610 
1611 	ops->dellink(dev, &list_kill);
1612 	unregister_netdevice_many(&list_kill);
1613 	list_del(&list_kill);
1614 	return 0;
1615 }
1616 
1617 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1618 {
1619 	unsigned int old_flags;
1620 	int err;
1621 
1622 	old_flags = dev->flags;
1623 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1624 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1625 		if (err < 0)
1626 			return err;
1627 	}
1628 
1629 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1630 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1631 
1632 	__dev_notify_flags(dev, old_flags);
1633 	return 0;
1634 }
1635 EXPORT_SYMBOL(rtnl_configure_link);
1636 
1637 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1638 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1639 {
1640 	int err;
1641 	struct net_device *dev;
1642 	unsigned int num_queues = 1;
1643 
1644 	if (ops->get_tx_queues) {
1645 		err = ops->get_tx_queues(src_net, tb);
1646 		if (err < 0)
1647 			goto err;
1648 		num_queues = err;
1649 	}
1650 
1651 	err = -ENOMEM;
1652 	dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1653 	if (!dev)
1654 		goto err;
1655 
1656 	dev_net_set(dev, net);
1657 	dev->rtnl_link_ops = ops;
1658 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1659 
1660 	if (tb[IFLA_MTU])
1661 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1662 	if (tb[IFLA_ADDRESS])
1663 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1664 				nla_len(tb[IFLA_ADDRESS]));
1665 	if (tb[IFLA_BROADCAST])
1666 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1667 				nla_len(tb[IFLA_BROADCAST]));
1668 	if (tb[IFLA_TXQLEN])
1669 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1670 	if (tb[IFLA_OPERSTATE])
1671 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1672 	if (tb[IFLA_LINKMODE])
1673 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1674 	if (tb[IFLA_GROUP])
1675 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1676 
1677 	return dev;
1678 
1679 err:
1680 	return ERR_PTR(err);
1681 }
1682 EXPORT_SYMBOL(rtnl_create_link);
1683 
1684 static int rtnl_group_changelink(struct net *net, int group,
1685 		struct ifinfomsg *ifm,
1686 		struct nlattr **tb)
1687 {
1688 	struct net_device *dev;
1689 	int err;
1690 
1691 	for_each_netdev(net, dev) {
1692 		if (dev->group == group) {
1693 			err = do_setlink(dev, ifm, tb, NULL, 0);
1694 			if (err < 0)
1695 				return err;
1696 		}
1697 	}
1698 
1699 	return 0;
1700 }
1701 
1702 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1703 {
1704 	struct net *net = sock_net(skb->sk);
1705 	const struct rtnl_link_ops *ops;
1706 	struct net_device *dev;
1707 	struct ifinfomsg *ifm;
1708 	char kind[MODULE_NAME_LEN];
1709 	char ifname[IFNAMSIZ];
1710 	struct nlattr *tb[IFLA_MAX+1];
1711 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1712 	int err;
1713 
1714 #ifdef CONFIG_MODULES
1715 replay:
1716 #endif
1717 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1718 	if (err < 0)
1719 		return err;
1720 
1721 	if (tb[IFLA_IFNAME])
1722 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1723 	else
1724 		ifname[0] = '\0';
1725 
1726 	ifm = nlmsg_data(nlh);
1727 	if (ifm->ifi_index > 0)
1728 		dev = __dev_get_by_index(net, ifm->ifi_index);
1729 	else {
1730 		if (ifname[0])
1731 			dev = __dev_get_by_name(net, ifname);
1732 		else
1733 			dev = NULL;
1734 	}
1735 
1736 	err = validate_linkmsg(dev, tb);
1737 	if (err < 0)
1738 		return err;
1739 
1740 	if (tb[IFLA_LINKINFO]) {
1741 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1742 				       tb[IFLA_LINKINFO], ifla_info_policy);
1743 		if (err < 0)
1744 			return err;
1745 	} else
1746 		memset(linkinfo, 0, sizeof(linkinfo));
1747 
1748 	if (linkinfo[IFLA_INFO_KIND]) {
1749 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1750 		ops = rtnl_link_ops_get(kind);
1751 	} else {
1752 		kind[0] = '\0';
1753 		ops = NULL;
1754 	}
1755 
1756 	if (1) {
1757 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1758 		struct net *dest_net;
1759 
1760 		if (ops) {
1761 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1762 				err = nla_parse_nested(attr, ops->maxtype,
1763 						       linkinfo[IFLA_INFO_DATA],
1764 						       ops->policy);
1765 				if (err < 0)
1766 					return err;
1767 				data = attr;
1768 			}
1769 			if (ops->validate) {
1770 				err = ops->validate(tb, data);
1771 				if (err < 0)
1772 					return err;
1773 			}
1774 		}
1775 
1776 		if (dev) {
1777 			int modified = 0;
1778 
1779 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1780 				return -EEXIST;
1781 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1782 				return -EOPNOTSUPP;
1783 
1784 			if (linkinfo[IFLA_INFO_DATA]) {
1785 				if (!ops || ops != dev->rtnl_link_ops ||
1786 				    !ops->changelink)
1787 					return -EOPNOTSUPP;
1788 
1789 				err = ops->changelink(dev, tb, data);
1790 				if (err < 0)
1791 					return err;
1792 				modified = 1;
1793 			}
1794 
1795 			return do_setlink(dev, ifm, tb, ifname, modified);
1796 		}
1797 
1798 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1799 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1800 				return rtnl_group_changelink(net,
1801 						nla_get_u32(tb[IFLA_GROUP]),
1802 						ifm, tb);
1803 			return -ENODEV;
1804 		}
1805 
1806 		if (ifm->ifi_index)
1807 			return -EOPNOTSUPP;
1808 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1809 			return -EOPNOTSUPP;
1810 
1811 		if (!ops) {
1812 #ifdef CONFIG_MODULES
1813 			if (kind[0]) {
1814 				__rtnl_unlock();
1815 				request_module("rtnl-link-%s", kind);
1816 				rtnl_lock();
1817 				ops = rtnl_link_ops_get(kind);
1818 				if (ops)
1819 					goto replay;
1820 			}
1821 #endif
1822 			return -EOPNOTSUPP;
1823 		}
1824 
1825 		if (!ifname[0])
1826 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1827 
1828 		dest_net = rtnl_link_get_net(net, tb);
1829 		if (IS_ERR(dest_net))
1830 			return PTR_ERR(dest_net);
1831 
1832 		dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1833 
1834 		if (IS_ERR(dev))
1835 			err = PTR_ERR(dev);
1836 		else if (ops->newlink)
1837 			err = ops->newlink(net, dev, tb, data);
1838 		else
1839 			err = register_netdevice(dev);
1840 
1841 		if (err < 0 && !IS_ERR(dev))
1842 			free_netdev(dev);
1843 		if (err < 0)
1844 			goto out;
1845 
1846 		err = rtnl_configure_link(dev, ifm);
1847 		if (err < 0)
1848 			unregister_netdevice(dev);
1849 out:
1850 		put_net(dest_net);
1851 		return err;
1852 	}
1853 }
1854 
1855 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1856 {
1857 	struct net *net = sock_net(skb->sk);
1858 	struct ifinfomsg *ifm;
1859 	char ifname[IFNAMSIZ];
1860 	struct nlattr *tb[IFLA_MAX+1];
1861 	struct net_device *dev = NULL;
1862 	struct sk_buff *nskb;
1863 	int err;
1864 	u32 ext_filter_mask = 0;
1865 
1866 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1867 	if (err < 0)
1868 		return err;
1869 
1870 	if (tb[IFLA_IFNAME])
1871 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1872 
1873 	if (tb[IFLA_EXT_MASK])
1874 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1875 
1876 	ifm = nlmsg_data(nlh);
1877 	if (ifm->ifi_index > 0)
1878 		dev = __dev_get_by_index(net, ifm->ifi_index);
1879 	else if (tb[IFLA_IFNAME])
1880 		dev = __dev_get_by_name(net, ifname);
1881 	else
1882 		return -EINVAL;
1883 
1884 	if (dev == NULL)
1885 		return -ENODEV;
1886 
1887 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1888 	if (nskb == NULL)
1889 		return -ENOBUFS;
1890 
1891 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1892 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1893 	if (err < 0) {
1894 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1895 		WARN_ON(err == -EMSGSIZE);
1896 		kfree_skb(nskb);
1897 	} else
1898 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1899 
1900 	return err;
1901 }
1902 
1903 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1904 {
1905 	struct net *net = sock_net(skb->sk);
1906 	struct net_device *dev;
1907 	struct nlattr *tb[IFLA_MAX+1];
1908 	u32 ext_filter_mask = 0;
1909 	u16 min_ifinfo_dump_size = 0;
1910 
1911 	if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1912 			ifla_policy) >= 0) {
1913 		if (tb[IFLA_EXT_MASK])
1914 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1915 	}
1916 
1917 	if (!ext_filter_mask)
1918 		return NLMSG_GOODSIZE;
1919 	/*
1920 	 * traverse the list of net devices and compute the minimum
1921 	 * buffer size based upon the filter mask.
1922 	 */
1923 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1924 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1925 					     if_nlmsg_size(dev,
1926 						           ext_filter_mask));
1927 	}
1928 
1929 	return min_ifinfo_dump_size;
1930 }
1931 
1932 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1933 {
1934 	int idx;
1935 	int s_idx = cb->family;
1936 
1937 	if (s_idx == 0)
1938 		s_idx = 1;
1939 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1940 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1941 		if (idx < s_idx || idx == PF_PACKET)
1942 			continue;
1943 		if (rtnl_msg_handlers[idx] == NULL ||
1944 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1945 			continue;
1946 		if (idx > s_idx)
1947 			memset(&cb->args[0], 0, sizeof(cb->args));
1948 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1949 			break;
1950 	}
1951 	cb->family = idx;
1952 
1953 	return skb->len;
1954 }
1955 
1956 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1957 {
1958 	struct net *net = dev_net(dev);
1959 	struct sk_buff *skb;
1960 	int err = -ENOBUFS;
1961 	size_t if_info_size;
1962 
1963 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1964 	if (skb == NULL)
1965 		goto errout;
1966 
1967 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1968 	if (err < 0) {
1969 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1970 		WARN_ON(err == -EMSGSIZE);
1971 		kfree_skb(skb);
1972 		goto errout;
1973 	}
1974 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1975 	return;
1976 errout:
1977 	if (err < 0)
1978 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1979 }
1980 
1981 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1982 				   struct net_device *dev,
1983 				   u8 *addr, u32 pid, u32 seq,
1984 				   int type, unsigned int flags)
1985 {
1986 	struct nlmsghdr *nlh;
1987 	struct ndmsg *ndm;
1988 
1989 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1990 	if (!nlh)
1991 		return -EMSGSIZE;
1992 
1993 	ndm = nlmsg_data(nlh);
1994 	ndm->ndm_family  = AF_BRIDGE;
1995 	ndm->ndm_pad1	 = 0;
1996 	ndm->ndm_pad2    = 0;
1997 	ndm->ndm_flags	 = flags;
1998 	ndm->ndm_type	 = 0;
1999 	ndm->ndm_ifindex = dev->ifindex;
2000 	ndm->ndm_state   = NUD_PERMANENT;
2001 
2002 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2003 		goto nla_put_failure;
2004 
2005 	return nlmsg_end(skb, nlh);
2006 
2007 nla_put_failure:
2008 	nlmsg_cancel(skb, nlh);
2009 	return -EMSGSIZE;
2010 }
2011 
2012 static inline size_t rtnl_fdb_nlmsg_size(void)
2013 {
2014 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2015 }
2016 
2017 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2018 {
2019 	struct net *net = dev_net(dev);
2020 	struct sk_buff *skb;
2021 	int err = -ENOBUFS;
2022 
2023 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2024 	if (!skb)
2025 		goto errout;
2026 
2027 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2028 	if (err < 0) {
2029 		kfree_skb(skb);
2030 		goto errout;
2031 	}
2032 
2033 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2034 	return;
2035 errout:
2036 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2037 }
2038 
2039 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2040 {
2041 	struct net *net = sock_net(skb->sk);
2042 	struct net_device *master = NULL;
2043 	struct ndmsg *ndm;
2044 	struct nlattr *tb[NDA_MAX+1];
2045 	struct net_device *dev;
2046 	u8 *addr;
2047 	int err;
2048 
2049 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2050 	if (err < 0)
2051 		return err;
2052 
2053 	ndm = nlmsg_data(nlh);
2054 	if (ndm->ndm_ifindex == 0) {
2055 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2056 		return -EINVAL;
2057 	}
2058 
2059 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2060 	if (dev == NULL) {
2061 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2062 		return -ENODEV;
2063 	}
2064 
2065 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2066 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2067 		return -EINVAL;
2068 	}
2069 
2070 	addr = nla_data(tb[NDA_LLADDR]);
2071 	if (!is_valid_ether_addr(addr)) {
2072 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2073 		return -EINVAL;
2074 	}
2075 
2076 	err = -EOPNOTSUPP;
2077 
2078 	/* Support fdb on master device the net/bridge default case */
2079 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2080 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2081 		master = dev->master;
2082 		err = master->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2083 						      nlh->nlmsg_flags);
2084 		if (err)
2085 			goto out;
2086 		else
2087 			ndm->ndm_flags &= ~NTF_MASTER;
2088 	}
2089 
2090 	/* Embedded bridge, macvlan, and any other device support */
2091 	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_add) {
2092 		err = dev->netdev_ops->ndo_fdb_add(ndm, dev, addr,
2093 						   nlh->nlmsg_flags);
2094 
2095 		if (!err) {
2096 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2097 			ndm->ndm_flags &= ~NTF_SELF;
2098 		}
2099 	}
2100 out:
2101 	return err;
2102 }
2103 
2104 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2105 {
2106 	struct net *net = sock_net(skb->sk);
2107 	struct ndmsg *ndm;
2108 	struct nlattr *llattr;
2109 	struct net_device *dev;
2110 	int err = -EINVAL;
2111 	__u8 *addr;
2112 
2113 	if (nlmsg_len(nlh) < sizeof(*ndm))
2114 		return -EINVAL;
2115 
2116 	ndm = nlmsg_data(nlh);
2117 	if (ndm->ndm_ifindex == 0) {
2118 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2119 		return -EINVAL;
2120 	}
2121 
2122 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2123 	if (dev == NULL) {
2124 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2125 		return -ENODEV;
2126 	}
2127 
2128 	llattr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_LLADDR);
2129 	if (llattr == NULL || nla_len(llattr) != ETH_ALEN) {
2130 		pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2131 		return -EINVAL;
2132 	}
2133 
2134 	addr = nla_data(llattr);
2135 	err = -EOPNOTSUPP;
2136 
2137 	/* Support fdb on master device the net/bridge default case */
2138 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2139 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2140 		struct net_device *master = dev->master;
2141 
2142 		if (master->netdev_ops->ndo_fdb_del)
2143 			err = master->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2144 
2145 		if (err)
2146 			goto out;
2147 		else
2148 			ndm->ndm_flags &= ~NTF_MASTER;
2149 	}
2150 
2151 	/* Embedded bridge, macvlan, and any other device support */
2152 	if ((ndm->ndm_flags & NTF_SELF) && dev->netdev_ops->ndo_fdb_del) {
2153 		err = dev->netdev_ops->ndo_fdb_del(ndm, dev, addr);
2154 
2155 		if (!err) {
2156 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2157 			ndm->ndm_flags &= ~NTF_SELF;
2158 		}
2159 	}
2160 out:
2161 	return err;
2162 }
2163 
2164 static int nlmsg_populate_fdb(struct sk_buff *skb,
2165 			      struct netlink_callback *cb,
2166 			      struct net_device *dev,
2167 			      int *idx,
2168 			      struct netdev_hw_addr_list *list)
2169 {
2170 	struct netdev_hw_addr *ha;
2171 	int err;
2172 	u32 pid, seq;
2173 
2174 	pid = NETLINK_CB(cb->skb).pid;
2175 	seq = cb->nlh->nlmsg_seq;
2176 
2177 	list_for_each_entry(ha, &list->list, list) {
2178 		if (*idx < cb->args[0])
2179 			goto skip;
2180 
2181 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2182 					      pid, seq, 0, NTF_SELF);
2183 		if (err < 0)
2184 			return err;
2185 skip:
2186 		*idx += 1;
2187 	}
2188 	return 0;
2189 }
2190 
2191 /**
2192  * ndo_dflt_fdb_dump: default netdevice operation to dump an FDB table.
2193  * @nlh: netlink message header
2194  * @dev: netdevice
2195  *
2196  * Default netdevice operation to dump the existing unicast address list.
2197  * Returns zero on success.
2198  */
2199 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2200 		      struct netlink_callback *cb,
2201 		      struct net_device *dev,
2202 		      int idx)
2203 {
2204 	int err;
2205 
2206 	netif_addr_lock_bh(dev);
2207 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2208 	if (err)
2209 		goto out;
2210 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2211 out:
2212 	netif_addr_unlock_bh(dev);
2213 	return idx;
2214 }
2215 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2216 
2217 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2218 {
2219 	int idx = 0;
2220 	struct net *net = sock_net(skb->sk);
2221 	struct net_device *dev;
2222 
2223 	rcu_read_lock();
2224 	for_each_netdev_rcu(net, dev) {
2225 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2226 			struct net_device *master = dev->master;
2227 			const struct net_device_ops *ops = master->netdev_ops;
2228 
2229 			if (ops->ndo_fdb_dump)
2230 				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2231 		}
2232 
2233 		if (dev->netdev_ops->ndo_fdb_dump)
2234 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2235 	}
2236 	rcu_read_unlock();
2237 
2238 	cb->args[0] = idx;
2239 	return skb->len;
2240 }
2241 
2242 /* Protected by RTNL sempahore.  */
2243 static struct rtattr **rta_buf;
2244 static int rtattr_max;
2245 
2246 /* Process one rtnetlink message. */
2247 
2248 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2249 {
2250 	struct net *net = sock_net(skb->sk);
2251 	rtnl_doit_func doit;
2252 	int sz_idx, kind;
2253 	int min_len;
2254 	int family;
2255 	int type;
2256 	int err;
2257 
2258 	type = nlh->nlmsg_type;
2259 	if (type > RTM_MAX)
2260 		return -EOPNOTSUPP;
2261 
2262 	type -= RTM_BASE;
2263 
2264 	/* All the messages must have at least 1 byte length */
2265 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2266 		return 0;
2267 
2268 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2269 	sz_idx = type>>2;
2270 	kind = type&3;
2271 
2272 	if (kind != 2 && !capable(CAP_NET_ADMIN))
2273 		return -EPERM;
2274 
2275 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2276 		struct sock *rtnl;
2277 		rtnl_dumpit_func dumpit;
2278 		rtnl_calcit_func calcit;
2279 		u16 min_dump_alloc = 0;
2280 
2281 		dumpit = rtnl_get_dumpit(family, type);
2282 		if (dumpit == NULL)
2283 			return -EOPNOTSUPP;
2284 		calcit = rtnl_get_calcit(family, type);
2285 		if (calcit)
2286 			min_dump_alloc = calcit(skb, nlh);
2287 
2288 		__rtnl_unlock();
2289 		rtnl = net->rtnl;
2290 		{
2291 			struct netlink_dump_control c = {
2292 				.dump		= dumpit,
2293 				.min_dump_alloc	= min_dump_alloc,
2294 			};
2295 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2296 		}
2297 		rtnl_lock();
2298 		return err;
2299 	}
2300 
2301 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2302 
2303 	min_len = rtm_min[sz_idx];
2304 	if (nlh->nlmsg_len < min_len)
2305 		return -EINVAL;
2306 
2307 	if (nlh->nlmsg_len > min_len) {
2308 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2309 		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2310 
2311 		while (RTA_OK(attr, attrlen)) {
2312 			unsigned int flavor = attr->rta_type;
2313 			if (flavor) {
2314 				if (flavor > rta_max[sz_idx])
2315 					return -EINVAL;
2316 				rta_buf[flavor-1] = attr;
2317 			}
2318 			attr = RTA_NEXT(attr, attrlen);
2319 		}
2320 	}
2321 
2322 	doit = rtnl_get_doit(family, type);
2323 	if (doit == NULL)
2324 		return -EOPNOTSUPP;
2325 
2326 	return doit(skb, nlh, (void *)&rta_buf[0]);
2327 }
2328 
2329 static void rtnetlink_rcv(struct sk_buff *skb)
2330 {
2331 	rtnl_lock();
2332 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2333 	rtnl_unlock();
2334 }
2335 
2336 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2337 {
2338 	struct net_device *dev = ptr;
2339 
2340 	switch (event) {
2341 	case NETDEV_UP:
2342 	case NETDEV_DOWN:
2343 	case NETDEV_PRE_UP:
2344 	case NETDEV_POST_INIT:
2345 	case NETDEV_REGISTER:
2346 	case NETDEV_CHANGE:
2347 	case NETDEV_PRE_TYPE_CHANGE:
2348 	case NETDEV_GOING_DOWN:
2349 	case NETDEV_UNREGISTER:
2350 	case NETDEV_UNREGISTER_BATCH:
2351 	case NETDEV_RELEASE:
2352 	case NETDEV_JOIN:
2353 		break;
2354 	default:
2355 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2356 		break;
2357 	}
2358 	return NOTIFY_DONE;
2359 }
2360 
2361 static struct notifier_block rtnetlink_dev_notifier = {
2362 	.notifier_call	= rtnetlink_event,
2363 };
2364 
2365 
2366 static int __net_init rtnetlink_net_init(struct net *net)
2367 {
2368 	struct sock *sk;
2369 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2370 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2371 	if (!sk)
2372 		return -ENOMEM;
2373 	net->rtnl = sk;
2374 	return 0;
2375 }
2376 
2377 static void __net_exit rtnetlink_net_exit(struct net *net)
2378 {
2379 	netlink_kernel_release(net->rtnl);
2380 	net->rtnl = NULL;
2381 }
2382 
2383 static struct pernet_operations rtnetlink_net_ops = {
2384 	.init = rtnetlink_net_init,
2385 	.exit = rtnetlink_net_exit,
2386 };
2387 
2388 void __init rtnetlink_init(void)
2389 {
2390 	int i;
2391 
2392 	rtattr_max = 0;
2393 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2394 		if (rta_max[i] > rtattr_max)
2395 			rtattr_max = rta_max[i];
2396 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2397 	if (!rta_buf)
2398 		panic("rtnetlink_init: cannot allocate rta_buf\n");
2399 
2400 	if (register_pernet_subsys(&rtnetlink_net_ops))
2401 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2402 
2403 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2404 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2405 
2406 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2407 		      rtnl_dump_ifinfo, rtnl_calcit);
2408 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2409 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2410 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2411 
2412 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2413 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2414 
2415 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2416 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2417 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2418 }
2419 
2420