xref: /openbmc/linux/net/core/rtnetlink.c (revision 63dc02bd)
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/pci.h>
39 
40 #include <asm/uaccess.h>
41 
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <net/ip.h>
45 #include <net/protocol.h>
46 #include <net/arp.h>
47 #include <net/route.h>
48 #include <net/udp.h>
49 #include <net/sock.h>
50 #include <net/pkt_sched.h>
51 #include <net/fib_rules.h>
52 #include <net/rtnetlink.h>
53 #include <net/net_namespace.h>
54 
55 struct rtnl_link {
56 	rtnl_doit_func		doit;
57 	rtnl_dumpit_func	dumpit;
58 	rtnl_calcit_func 	calcit;
59 };
60 
61 static DEFINE_MUTEX(rtnl_mutex);
62 
63 void rtnl_lock(void)
64 {
65 	mutex_lock(&rtnl_mutex);
66 }
67 EXPORT_SYMBOL(rtnl_lock);
68 
69 void __rtnl_unlock(void)
70 {
71 	mutex_unlock(&rtnl_mutex);
72 }
73 
74 void rtnl_unlock(void)
75 {
76 	/* This fellow will unlock it for us. */
77 	netdev_run_todo();
78 }
79 EXPORT_SYMBOL(rtnl_unlock);
80 
81 int rtnl_trylock(void)
82 {
83 	return mutex_trylock(&rtnl_mutex);
84 }
85 EXPORT_SYMBOL(rtnl_trylock);
86 
87 int rtnl_is_locked(void)
88 {
89 	return mutex_is_locked(&rtnl_mutex);
90 }
91 EXPORT_SYMBOL(rtnl_is_locked);
92 
93 #ifdef CONFIG_PROVE_LOCKING
94 int lockdep_rtnl_is_held(void)
95 {
96 	return lockdep_is_held(&rtnl_mutex);
97 }
98 EXPORT_SYMBOL(lockdep_rtnl_is_held);
99 #endif /* #ifdef CONFIG_PROVE_LOCKING */
100 
101 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
102 
103 static inline int rtm_msgindex(int msgtype)
104 {
105 	int msgindex = msgtype - RTM_BASE;
106 
107 	/*
108 	 * msgindex < 0 implies someone tried to register a netlink
109 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
110 	 * the message type has not been added to linux/rtnetlink.h
111 	 */
112 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
113 
114 	return msgindex;
115 }
116 
117 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
118 {
119 	struct rtnl_link *tab;
120 
121 	if (protocol <= RTNL_FAMILY_MAX)
122 		tab = rtnl_msg_handlers[protocol];
123 	else
124 		tab = NULL;
125 
126 	if (tab == NULL || tab[msgindex].doit == NULL)
127 		tab = rtnl_msg_handlers[PF_UNSPEC];
128 
129 	return tab ? tab[msgindex].doit : NULL;
130 }
131 
132 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
133 {
134 	struct rtnl_link *tab;
135 
136 	if (protocol <= RTNL_FAMILY_MAX)
137 		tab = rtnl_msg_handlers[protocol];
138 	else
139 		tab = NULL;
140 
141 	if (tab == NULL || tab[msgindex].dumpit == NULL)
142 		tab = rtnl_msg_handlers[PF_UNSPEC];
143 
144 	return tab ? tab[msgindex].dumpit : NULL;
145 }
146 
147 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
148 {
149 	struct rtnl_link *tab;
150 
151 	if (protocol <= RTNL_FAMILY_MAX)
152 		tab = rtnl_msg_handlers[protocol];
153 	else
154 		tab = NULL;
155 
156 	if (tab == NULL || tab[msgindex].calcit == NULL)
157 		tab = rtnl_msg_handlers[PF_UNSPEC];
158 
159 	return tab ? tab[msgindex].calcit : NULL;
160 }
161 
162 /**
163  * __rtnl_register - Register a rtnetlink message type
164  * @protocol: Protocol family or PF_UNSPEC
165  * @msgtype: rtnetlink message type
166  * @doit: Function pointer called for each request message
167  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
168  * @calcit: Function pointer to calc size of dump message
169  *
170  * Registers the specified function pointers (at least one of them has
171  * to be non-NULL) to be called whenever a request message for the
172  * specified protocol family and message type is received.
173  *
174  * The special protocol family PF_UNSPEC may be used to define fallback
175  * function pointers for the case when no entry for the specific protocol
176  * family exists.
177  *
178  * Returns 0 on success or a negative error code.
179  */
180 int __rtnl_register(int protocol, int msgtype,
181 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
182 		    rtnl_calcit_func calcit)
183 {
184 	struct rtnl_link *tab;
185 	int msgindex;
186 
187 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
188 	msgindex = rtm_msgindex(msgtype);
189 
190 	tab = rtnl_msg_handlers[protocol];
191 	if (tab == NULL) {
192 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
193 		if (tab == NULL)
194 			return -ENOBUFS;
195 
196 		rtnl_msg_handlers[protocol] = tab;
197 	}
198 
199 	if (doit)
200 		tab[msgindex].doit = doit;
201 
202 	if (dumpit)
203 		tab[msgindex].dumpit = dumpit;
204 
205 	if (calcit)
206 		tab[msgindex].calcit = calcit;
207 
208 	return 0;
209 }
210 EXPORT_SYMBOL_GPL(__rtnl_register);
211 
212 /**
213  * rtnl_register - Register a rtnetlink message type
214  *
215  * Identical to __rtnl_register() but panics on failure. This is useful
216  * as failure of this function is very unlikely, it can only happen due
217  * to lack of memory when allocating the chain to store all message
218  * handlers for a protocol. Meant for use in init functions where lack
219  * of memory implies no sense in continuing.
220  */
221 void rtnl_register(int protocol, int msgtype,
222 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
223 		   rtnl_calcit_func calcit)
224 {
225 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
226 		panic("Unable to register rtnetlink message handler, "
227 		      "protocol = %d, message type = %d\n",
228 		      protocol, msgtype);
229 }
230 EXPORT_SYMBOL_GPL(rtnl_register);
231 
232 /**
233  * rtnl_unregister - Unregister a rtnetlink message type
234  * @protocol: Protocol family or PF_UNSPEC
235  * @msgtype: rtnetlink message type
236  *
237  * Returns 0 on success or a negative error code.
238  */
239 int rtnl_unregister(int protocol, int msgtype)
240 {
241 	int msgindex;
242 
243 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
244 	msgindex = rtm_msgindex(msgtype);
245 
246 	if (rtnl_msg_handlers[protocol] == NULL)
247 		return -ENOENT;
248 
249 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
250 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
251 
252 	return 0;
253 }
254 EXPORT_SYMBOL_GPL(rtnl_unregister);
255 
256 /**
257  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
258  * @protocol : Protocol family or PF_UNSPEC
259  *
260  * Identical to calling rtnl_unregster() for all registered message types
261  * of a certain protocol family.
262  */
263 void rtnl_unregister_all(int protocol)
264 {
265 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
266 
267 	kfree(rtnl_msg_handlers[protocol]);
268 	rtnl_msg_handlers[protocol] = NULL;
269 }
270 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
271 
272 static LIST_HEAD(link_ops);
273 
274 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
275 {
276 	const struct rtnl_link_ops *ops;
277 
278 	list_for_each_entry(ops, &link_ops, list) {
279 		if (!strcmp(ops->kind, kind))
280 			return ops;
281 	}
282 	return NULL;
283 }
284 
285 /**
286  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
287  * @ops: struct rtnl_link_ops * to register
288  *
289  * The caller must hold the rtnl_mutex. This function should be used
290  * by drivers that create devices during module initialization. It
291  * must be called before registering the devices.
292  *
293  * Returns 0 on success or a negative error code.
294  */
295 int __rtnl_link_register(struct rtnl_link_ops *ops)
296 {
297 	if (rtnl_link_ops_get(ops->kind))
298 		return -EEXIST;
299 
300 	if (!ops->dellink)
301 		ops->dellink = unregister_netdevice_queue;
302 
303 	list_add_tail(&ops->list, &link_ops);
304 	return 0;
305 }
306 EXPORT_SYMBOL_GPL(__rtnl_link_register);
307 
308 /**
309  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
310  * @ops: struct rtnl_link_ops * to register
311  *
312  * Returns 0 on success or a negative error code.
313  */
314 int rtnl_link_register(struct rtnl_link_ops *ops)
315 {
316 	int err;
317 
318 	rtnl_lock();
319 	err = __rtnl_link_register(ops);
320 	rtnl_unlock();
321 	return err;
322 }
323 EXPORT_SYMBOL_GPL(rtnl_link_register);
324 
325 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
326 {
327 	struct net_device *dev;
328 	LIST_HEAD(list_kill);
329 
330 	for_each_netdev(net, dev) {
331 		if (dev->rtnl_link_ops == ops)
332 			ops->dellink(dev, &list_kill);
333 	}
334 	unregister_netdevice_many(&list_kill);
335 }
336 
337 /**
338  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
339  * @ops: struct rtnl_link_ops * to unregister
340  *
341  * The caller must hold the rtnl_mutex.
342  */
343 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
344 {
345 	struct net *net;
346 
347 	for_each_net(net) {
348 		__rtnl_kill_links(net, ops);
349 	}
350 	list_del(&ops->list);
351 }
352 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
353 
354 /**
355  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
356  * @ops: struct rtnl_link_ops * to unregister
357  */
358 void rtnl_link_unregister(struct rtnl_link_ops *ops)
359 {
360 	rtnl_lock();
361 	__rtnl_link_unregister(ops);
362 	rtnl_unlock();
363 }
364 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
365 
366 static size_t rtnl_link_get_size(const struct net_device *dev)
367 {
368 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
369 	size_t size;
370 
371 	if (!ops)
372 		return 0;
373 
374 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
375 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
376 
377 	if (ops->get_size)
378 		/* IFLA_INFO_DATA + nested data */
379 		size += nla_total_size(sizeof(struct nlattr)) +
380 			ops->get_size(dev);
381 
382 	if (ops->get_xstats_size)
383 		/* IFLA_INFO_XSTATS */
384 		size += nla_total_size(ops->get_xstats_size(dev));
385 
386 	return size;
387 }
388 
389 static LIST_HEAD(rtnl_af_ops);
390 
391 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
392 {
393 	const struct rtnl_af_ops *ops;
394 
395 	list_for_each_entry(ops, &rtnl_af_ops, list) {
396 		if (ops->family == family)
397 			return ops;
398 	}
399 
400 	return NULL;
401 }
402 
403 /**
404  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
405  * @ops: struct rtnl_af_ops * to register
406  *
407  * The caller must hold the rtnl_mutex.
408  *
409  * Returns 0 on success or a negative error code.
410  */
411 int __rtnl_af_register(struct rtnl_af_ops *ops)
412 {
413 	list_add_tail(&ops->list, &rtnl_af_ops);
414 	return 0;
415 }
416 EXPORT_SYMBOL_GPL(__rtnl_af_register);
417 
418 /**
419  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
420  * @ops: struct rtnl_af_ops * to register
421  *
422  * Returns 0 on success or a negative error code.
423  */
424 int rtnl_af_register(struct rtnl_af_ops *ops)
425 {
426 	int err;
427 
428 	rtnl_lock();
429 	err = __rtnl_af_register(ops);
430 	rtnl_unlock();
431 	return err;
432 }
433 EXPORT_SYMBOL_GPL(rtnl_af_register);
434 
435 /**
436  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
437  * @ops: struct rtnl_af_ops * to unregister
438  *
439  * The caller must hold the rtnl_mutex.
440  */
441 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
442 {
443 	list_del(&ops->list);
444 }
445 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
446 
447 /**
448  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
449  * @ops: struct rtnl_af_ops * to unregister
450  */
451 void rtnl_af_unregister(struct rtnl_af_ops *ops)
452 {
453 	rtnl_lock();
454 	__rtnl_af_unregister(ops);
455 	rtnl_unlock();
456 }
457 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
458 
459 static size_t rtnl_link_get_af_size(const struct net_device *dev)
460 {
461 	struct rtnl_af_ops *af_ops;
462 	size_t size;
463 
464 	/* IFLA_AF_SPEC */
465 	size = nla_total_size(sizeof(struct nlattr));
466 
467 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
468 		if (af_ops->get_link_af_size) {
469 			/* AF_* + nested data */
470 			size += nla_total_size(sizeof(struct nlattr)) +
471 				af_ops->get_link_af_size(dev);
472 		}
473 	}
474 
475 	return size;
476 }
477 
478 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
479 {
480 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
481 	struct nlattr *linkinfo, *data;
482 	int err = -EMSGSIZE;
483 
484 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
485 	if (linkinfo == NULL)
486 		goto out;
487 
488 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
489 		goto err_cancel_link;
490 	if (ops->fill_xstats) {
491 		err = ops->fill_xstats(skb, dev);
492 		if (err < 0)
493 			goto err_cancel_link;
494 	}
495 	if (ops->fill_info) {
496 		data = nla_nest_start(skb, IFLA_INFO_DATA);
497 		if (data == NULL)
498 			goto err_cancel_link;
499 		err = ops->fill_info(skb, dev);
500 		if (err < 0)
501 			goto err_cancel_data;
502 		nla_nest_end(skb, data);
503 	}
504 
505 	nla_nest_end(skb, linkinfo);
506 	return 0;
507 
508 err_cancel_data:
509 	nla_nest_cancel(skb, data);
510 err_cancel_link:
511 	nla_nest_cancel(skb, linkinfo);
512 out:
513 	return err;
514 }
515 
516 static const int rtm_min[RTM_NR_FAMILIES] =
517 {
518 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
519 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
520 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
521 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
522 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
523 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
524 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
525 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
526 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
527 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
528 };
529 
530 static const int rta_max[RTM_NR_FAMILIES] =
531 {
532 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
533 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
534 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
535 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
536 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
537 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
538 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
539 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
540 };
541 
542 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
543 {
544 	struct rtattr *rta;
545 	int size = RTA_LENGTH(attrlen);
546 
547 	rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
548 	rta->rta_type = attrtype;
549 	rta->rta_len = size;
550 	memcpy(RTA_DATA(rta), data, attrlen);
551 	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
552 }
553 EXPORT_SYMBOL(__rta_fill);
554 
555 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
556 {
557 	struct sock *rtnl = net->rtnl;
558 	int err = 0;
559 
560 	NETLINK_CB(skb).dst_group = group;
561 	if (echo)
562 		atomic_inc(&skb->users);
563 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
564 	if (echo)
565 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
566 	return err;
567 }
568 
569 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
570 {
571 	struct sock *rtnl = net->rtnl;
572 
573 	return nlmsg_unicast(rtnl, skb, pid);
574 }
575 EXPORT_SYMBOL(rtnl_unicast);
576 
577 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
578 		 struct nlmsghdr *nlh, gfp_t flags)
579 {
580 	struct sock *rtnl = net->rtnl;
581 	int report = 0;
582 
583 	if (nlh)
584 		report = nlmsg_report(nlh);
585 
586 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
587 }
588 EXPORT_SYMBOL(rtnl_notify);
589 
590 void rtnl_set_sk_err(struct net *net, u32 group, int error)
591 {
592 	struct sock *rtnl = net->rtnl;
593 
594 	netlink_set_err(rtnl, 0, group, error);
595 }
596 EXPORT_SYMBOL(rtnl_set_sk_err);
597 
598 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
599 {
600 	struct nlattr *mx;
601 	int i, valid = 0;
602 
603 	mx = nla_nest_start(skb, RTA_METRICS);
604 	if (mx == NULL)
605 		return -ENOBUFS;
606 
607 	for (i = 0; i < RTAX_MAX; i++) {
608 		if (metrics[i]) {
609 			valid++;
610 			NLA_PUT_U32(skb, i+1, metrics[i]);
611 		}
612 	}
613 
614 	if (!valid) {
615 		nla_nest_cancel(skb, mx);
616 		return 0;
617 	}
618 
619 	return nla_nest_end(skb, mx);
620 
621 nla_put_failure:
622 	nla_nest_cancel(skb, mx);
623 	return -EMSGSIZE;
624 }
625 EXPORT_SYMBOL(rtnetlink_put_metrics);
626 
627 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
628 		       u32 ts, u32 tsage, long expires, u32 error)
629 {
630 	struct rta_cacheinfo ci = {
631 		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
632 		.rta_used = dst->__use,
633 		.rta_clntref = atomic_read(&(dst->__refcnt)),
634 		.rta_error = error,
635 		.rta_id =  id,
636 		.rta_ts = ts,
637 		.rta_tsage = tsage,
638 	};
639 
640 	if (expires)
641 		ci.rta_expires = jiffies_to_clock_t(expires);
642 
643 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
644 }
645 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
646 
647 static void set_operstate(struct net_device *dev, unsigned char transition)
648 {
649 	unsigned char operstate = dev->operstate;
650 
651 	switch (transition) {
652 	case IF_OPER_UP:
653 		if ((operstate == IF_OPER_DORMANT ||
654 		     operstate == IF_OPER_UNKNOWN) &&
655 		    !netif_dormant(dev))
656 			operstate = IF_OPER_UP;
657 		break;
658 
659 	case IF_OPER_DORMANT:
660 		if (operstate == IF_OPER_UP ||
661 		    operstate == IF_OPER_UNKNOWN)
662 			operstate = IF_OPER_DORMANT;
663 		break;
664 	}
665 
666 	if (dev->operstate != operstate) {
667 		write_lock_bh(&dev_base_lock);
668 		dev->operstate = operstate;
669 		write_unlock_bh(&dev_base_lock);
670 		netdev_state_change(dev);
671 	}
672 }
673 
674 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
675 					   const struct ifinfomsg *ifm)
676 {
677 	unsigned int flags = ifm->ifi_flags;
678 
679 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
680 	if (ifm->ifi_change)
681 		flags = (flags & ifm->ifi_change) |
682 			(dev->flags & ~ifm->ifi_change);
683 
684 	return flags;
685 }
686 
687 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
688 				 const struct rtnl_link_stats64 *b)
689 {
690 	a->rx_packets = b->rx_packets;
691 	a->tx_packets = b->tx_packets;
692 	a->rx_bytes = b->rx_bytes;
693 	a->tx_bytes = b->tx_bytes;
694 	a->rx_errors = b->rx_errors;
695 	a->tx_errors = b->tx_errors;
696 	a->rx_dropped = b->rx_dropped;
697 	a->tx_dropped = b->tx_dropped;
698 
699 	a->multicast = b->multicast;
700 	a->collisions = b->collisions;
701 
702 	a->rx_length_errors = b->rx_length_errors;
703 	a->rx_over_errors = b->rx_over_errors;
704 	a->rx_crc_errors = b->rx_crc_errors;
705 	a->rx_frame_errors = b->rx_frame_errors;
706 	a->rx_fifo_errors = b->rx_fifo_errors;
707 	a->rx_missed_errors = b->rx_missed_errors;
708 
709 	a->tx_aborted_errors = b->tx_aborted_errors;
710 	a->tx_carrier_errors = b->tx_carrier_errors;
711 	a->tx_fifo_errors = b->tx_fifo_errors;
712 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
713 	a->tx_window_errors = b->tx_window_errors;
714 
715 	a->rx_compressed = b->rx_compressed;
716 	a->tx_compressed = b->tx_compressed;
717 }
718 
719 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
720 {
721 	memcpy(v, b, sizeof(*b));
722 }
723 
724 /* All VF info */
725 static inline int rtnl_vfinfo_size(const struct net_device *dev,
726 				   u32 ext_filter_mask)
727 {
728 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
729 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
730 		int num_vfs = dev_num_vf(dev->dev.parent);
731 		size_t size = nla_total_size(sizeof(struct nlattr));
732 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
733 		size += num_vfs *
734 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
735 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
736 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
737 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
738 		return size;
739 	} else
740 		return 0;
741 }
742 
743 static size_t rtnl_port_size(const struct net_device *dev)
744 {
745 	size_t port_size = nla_total_size(4)		/* PORT_VF */
746 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
747 		+ nla_total_size(sizeof(struct ifla_port_vsi))
748 							/* PORT_VSI_TYPE */
749 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
750 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
751 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
752 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
753 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
754 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
755 		+ port_size;
756 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
757 		+ port_size;
758 
759 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
760 		return 0;
761 	if (dev_num_vf(dev->dev.parent))
762 		return port_self_size + vf_ports_size +
763 			vf_port_size * dev_num_vf(dev->dev.parent);
764 	else
765 		return port_self_size;
766 }
767 
768 static noinline size_t if_nlmsg_size(const struct net_device *dev,
769 				     u32 ext_filter_mask)
770 {
771 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
772 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
773 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
774 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
775 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
776 	       + nla_total_size(sizeof(struct rtnl_link_stats))
777 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
778 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
779 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
780 	       + nla_total_size(4) /* IFLA_TXQLEN */
781 	       + nla_total_size(4) /* IFLA_WEIGHT */
782 	       + nla_total_size(4) /* IFLA_MTU */
783 	       + nla_total_size(4) /* IFLA_LINK */
784 	       + nla_total_size(4) /* IFLA_MASTER */
785 	       + nla_total_size(1) /* IFLA_OPERSTATE */
786 	       + nla_total_size(1) /* IFLA_LINKMODE */
787 	       + nla_total_size(ext_filter_mask
788 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
789 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
790 	       + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
791 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
792 	       + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
793 }
794 
795 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
796 {
797 	struct nlattr *vf_ports;
798 	struct nlattr *vf_port;
799 	int vf;
800 	int err;
801 
802 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
803 	if (!vf_ports)
804 		return -EMSGSIZE;
805 
806 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
807 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
808 		if (!vf_port)
809 			goto nla_put_failure;
810 		NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
811 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
812 		if (err == -EMSGSIZE)
813 			goto nla_put_failure;
814 		if (err) {
815 			nla_nest_cancel(skb, vf_port);
816 			continue;
817 		}
818 		nla_nest_end(skb, vf_port);
819 	}
820 
821 	nla_nest_end(skb, vf_ports);
822 
823 	return 0;
824 
825 nla_put_failure:
826 	nla_nest_cancel(skb, vf_ports);
827 	return -EMSGSIZE;
828 }
829 
830 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
831 {
832 	struct nlattr *port_self;
833 	int err;
834 
835 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
836 	if (!port_self)
837 		return -EMSGSIZE;
838 
839 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
840 	if (err) {
841 		nla_nest_cancel(skb, port_self);
842 		return (err == -EMSGSIZE) ? err : 0;
843 	}
844 
845 	nla_nest_end(skb, port_self);
846 
847 	return 0;
848 }
849 
850 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
851 {
852 	int err;
853 
854 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
855 		return 0;
856 
857 	err = rtnl_port_self_fill(skb, dev);
858 	if (err)
859 		return err;
860 
861 	if (dev_num_vf(dev->dev.parent)) {
862 		err = rtnl_vf_ports_fill(skb, dev);
863 		if (err)
864 			return err;
865 	}
866 
867 	return 0;
868 }
869 
870 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
871 			    int type, u32 pid, u32 seq, u32 change,
872 			    unsigned int flags, u32 ext_filter_mask)
873 {
874 	struct ifinfomsg *ifm;
875 	struct nlmsghdr *nlh;
876 	struct rtnl_link_stats64 temp;
877 	const struct rtnl_link_stats64 *stats;
878 	struct nlattr *attr, *af_spec;
879 	struct rtnl_af_ops *af_ops;
880 
881 	ASSERT_RTNL();
882 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
883 	if (nlh == NULL)
884 		return -EMSGSIZE;
885 
886 	ifm = nlmsg_data(nlh);
887 	ifm->ifi_family = AF_UNSPEC;
888 	ifm->__ifi_pad = 0;
889 	ifm->ifi_type = dev->type;
890 	ifm->ifi_index = dev->ifindex;
891 	ifm->ifi_flags = dev_get_flags(dev);
892 	ifm->ifi_change = change;
893 
894 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
895 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
896 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
897 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
898 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
899 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
900 	NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
901 
902 	if (dev->ifindex != dev->iflink)
903 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
904 
905 	if (dev->master)
906 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
907 
908 	if (dev->qdisc)
909 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
910 
911 	if (dev->ifalias)
912 		NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
913 
914 	if (1) {
915 		struct rtnl_link_ifmap map = {
916 			.mem_start   = dev->mem_start,
917 			.mem_end     = dev->mem_end,
918 			.base_addr   = dev->base_addr,
919 			.irq         = dev->irq,
920 			.dma         = dev->dma,
921 			.port        = dev->if_port,
922 		};
923 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
924 	}
925 
926 	if (dev->addr_len) {
927 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
928 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
929 	}
930 
931 	attr = nla_reserve(skb, IFLA_STATS,
932 			sizeof(struct rtnl_link_stats));
933 	if (attr == NULL)
934 		goto nla_put_failure;
935 
936 	stats = dev_get_stats(dev, &temp);
937 	copy_rtnl_link_stats(nla_data(attr), stats);
938 
939 	attr = nla_reserve(skb, IFLA_STATS64,
940 			sizeof(struct rtnl_link_stats64));
941 	if (attr == NULL)
942 		goto nla_put_failure;
943 	copy_rtnl_link_stats64(nla_data(attr), stats);
944 
945 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF))
946 		NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
947 
948 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
949 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
950 		int i;
951 
952 		struct nlattr *vfinfo, *vf;
953 		int num_vfs = dev_num_vf(dev->dev.parent);
954 
955 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
956 		if (!vfinfo)
957 			goto nla_put_failure;
958 		for (i = 0; i < num_vfs; i++) {
959 			struct ifla_vf_info ivi;
960 			struct ifla_vf_mac vf_mac;
961 			struct ifla_vf_vlan vf_vlan;
962 			struct ifla_vf_tx_rate vf_tx_rate;
963 			struct ifla_vf_spoofchk vf_spoofchk;
964 
965 			/*
966 			 * Not all SR-IOV capable drivers support the
967 			 * spoofcheck query.  Preset to -1 so the user
968 			 * space tool can detect that the driver didn't
969 			 * report anything.
970 			 */
971 			ivi.spoofchk = -1;
972 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
973 				break;
974 			vf_mac.vf =
975 				vf_vlan.vf =
976 				vf_tx_rate.vf =
977 				vf_spoofchk.vf = ivi.vf;
978 
979 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
980 			vf_vlan.vlan = ivi.vlan;
981 			vf_vlan.qos = ivi.qos;
982 			vf_tx_rate.rate = ivi.tx_rate;
983 			vf_spoofchk.setting = ivi.spoofchk;
984 			vf = nla_nest_start(skb, IFLA_VF_INFO);
985 			if (!vf) {
986 				nla_nest_cancel(skb, vfinfo);
987 				goto nla_put_failure;
988 			}
989 			NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
990 			NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
991 			NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
992 				&vf_tx_rate);
993 			NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
994 				&vf_spoofchk);
995 			nla_nest_end(skb, vf);
996 		}
997 		nla_nest_end(skb, vfinfo);
998 	}
999 
1000 	if (rtnl_port_fill(skb, dev))
1001 		goto nla_put_failure;
1002 
1003 	if (dev->rtnl_link_ops) {
1004 		if (rtnl_link_fill(skb, dev) < 0)
1005 			goto nla_put_failure;
1006 	}
1007 
1008 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1009 		goto nla_put_failure;
1010 
1011 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1012 		if (af_ops->fill_link_af) {
1013 			struct nlattr *af;
1014 			int err;
1015 
1016 			if (!(af = nla_nest_start(skb, af_ops->family)))
1017 				goto nla_put_failure;
1018 
1019 			err = af_ops->fill_link_af(skb, dev);
1020 
1021 			/*
1022 			 * Caller may return ENODATA to indicate that there
1023 			 * was no data to be dumped. This is not an error, it
1024 			 * means we should trim the attribute header and
1025 			 * continue.
1026 			 */
1027 			if (err == -ENODATA)
1028 				nla_nest_cancel(skb, af);
1029 			else if (err < 0)
1030 				goto nla_put_failure;
1031 
1032 			nla_nest_end(skb, af);
1033 		}
1034 	}
1035 
1036 	nla_nest_end(skb, af_spec);
1037 
1038 	return nlmsg_end(skb, nlh);
1039 
1040 nla_put_failure:
1041 	nlmsg_cancel(skb, nlh);
1042 	return -EMSGSIZE;
1043 }
1044 
1045 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1046 {
1047 	struct net *net = sock_net(skb->sk);
1048 	int h, s_h;
1049 	int idx = 0, s_idx;
1050 	struct net_device *dev;
1051 	struct hlist_head *head;
1052 	struct hlist_node *node;
1053 	struct nlattr *tb[IFLA_MAX+1];
1054 	u32 ext_filter_mask = 0;
1055 
1056 	s_h = cb->args[0];
1057 	s_idx = cb->args[1];
1058 
1059 	rcu_read_lock();
1060 	cb->seq = net->dev_base_seq;
1061 
1062 	if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1063 			ifla_policy) >= 0) {
1064 
1065 		if (tb[IFLA_EXT_MASK])
1066 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1067 	}
1068 
1069 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1070 		idx = 0;
1071 		head = &net->dev_index_head[h];
1072 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1073 			if (idx < s_idx)
1074 				goto cont;
1075 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1076 					     NETLINK_CB(cb->skb).pid,
1077 					     cb->nlh->nlmsg_seq, 0,
1078 					     NLM_F_MULTI,
1079 					     ext_filter_mask) <= 0)
1080 				goto out;
1081 
1082 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1083 cont:
1084 			idx++;
1085 		}
1086 	}
1087 out:
1088 	rcu_read_unlock();
1089 	cb->args[1] = idx;
1090 	cb->args[0] = h;
1091 
1092 	return skb->len;
1093 }
1094 
1095 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1096 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1097 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1098 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1099 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1100 	[IFLA_MTU]		= { .type = NLA_U32 },
1101 	[IFLA_LINK]		= { .type = NLA_U32 },
1102 	[IFLA_MASTER]		= { .type = NLA_U32 },
1103 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1104 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1105 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1106 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1107 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1108 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1109 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1110 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1111 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1112 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1113 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1114 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1115 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1116 };
1117 EXPORT_SYMBOL(ifla_policy);
1118 
1119 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1120 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1121 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1122 };
1123 
1124 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1125 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1126 };
1127 
1128 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1129 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1130 				    .len = sizeof(struct ifla_vf_mac) },
1131 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1132 				    .len = sizeof(struct ifla_vf_vlan) },
1133 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1134 				    .len = sizeof(struct ifla_vf_tx_rate) },
1135 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1136 				    .len = sizeof(struct ifla_vf_spoofchk) },
1137 };
1138 
1139 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1140 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1141 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1142 				    .len = PORT_PROFILE_MAX },
1143 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1144 				    .len = sizeof(struct ifla_port_vsi)},
1145 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1146 				      .len = PORT_UUID_MAX },
1147 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1148 				    .len = PORT_UUID_MAX },
1149 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1150 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1151 };
1152 
1153 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1154 {
1155 	struct net *net;
1156 	/* Examine the link attributes and figure out which
1157 	 * network namespace we are talking about.
1158 	 */
1159 	if (tb[IFLA_NET_NS_PID])
1160 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1161 	else if (tb[IFLA_NET_NS_FD])
1162 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1163 	else
1164 		net = get_net(src_net);
1165 	return net;
1166 }
1167 EXPORT_SYMBOL(rtnl_link_get_net);
1168 
1169 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1170 {
1171 	if (dev) {
1172 		if (tb[IFLA_ADDRESS] &&
1173 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1174 			return -EINVAL;
1175 
1176 		if (tb[IFLA_BROADCAST] &&
1177 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1178 			return -EINVAL;
1179 	}
1180 
1181 	if (tb[IFLA_AF_SPEC]) {
1182 		struct nlattr *af;
1183 		int rem, err;
1184 
1185 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1186 			const struct rtnl_af_ops *af_ops;
1187 
1188 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1189 				return -EAFNOSUPPORT;
1190 
1191 			if (!af_ops->set_link_af)
1192 				return -EOPNOTSUPP;
1193 
1194 			if (af_ops->validate_link_af) {
1195 				err = af_ops->validate_link_af(dev, af);
1196 				if (err < 0)
1197 					return err;
1198 			}
1199 		}
1200 	}
1201 
1202 	return 0;
1203 }
1204 
1205 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1206 {
1207 	int rem, err = -EINVAL;
1208 	struct nlattr *vf;
1209 	const struct net_device_ops *ops = dev->netdev_ops;
1210 
1211 	nla_for_each_nested(vf, attr, rem) {
1212 		switch (nla_type(vf)) {
1213 		case IFLA_VF_MAC: {
1214 			struct ifla_vf_mac *ivm;
1215 			ivm = nla_data(vf);
1216 			err = -EOPNOTSUPP;
1217 			if (ops->ndo_set_vf_mac)
1218 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1219 							  ivm->mac);
1220 			break;
1221 		}
1222 		case IFLA_VF_VLAN: {
1223 			struct ifla_vf_vlan *ivv;
1224 			ivv = nla_data(vf);
1225 			err = -EOPNOTSUPP;
1226 			if (ops->ndo_set_vf_vlan)
1227 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1228 							   ivv->vlan,
1229 							   ivv->qos);
1230 			break;
1231 		}
1232 		case IFLA_VF_TX_RATE: {
1233 			struct ifla_vf_tx_rate *ivt;
1234 			ivt = nla_data(vf);
1235 			err = -EOPNOTSUPP;
1236 			if (ops->ndo_set_vf_tx_rate)
1237 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1238 							      ivt->rate);
1239 			break;
1240 		}
1241 		case IFLA_VF_SPOOFCHK: {
1242 			struct ifla_vf_spoofchk *ivs;
1243 			ivs = nla_data(vf);
1244 			err = -EOPNOTSUPP;
1245 			if (ops->ndo_set_vf_spoofchk)
1246 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1247 							       ivs->setting);
1248 			break;
1249 		}
1250 		default:
1251 			err = -EINVAL;
1252 			break;
1253 		}
1254 		if (err)
1255 			break;
1256 	}
1257 	return err;
1258 }
1259 
1260 static int do_set_master(struct net_device *dev, int ifindex)
1261 {
1262 	struct net_device *master_dev;
1263 	const struct net_device_ops *ops;
1264 	int err;
1265 
1266 	if (dev->master) {
1267 		if (dev->master->ifindex == ifindex)
1268 			return 0;
1269 		ops = dev->master->netdev_ops;
1270 		if (ops->ndo_del_slave) {
1271 			err = ops->ndo_del_slave(dev->master, dev);
1272 			if (err)
1273 				return err;
1274 		} else {
1275 			return -EOPNOTSUPP;
1276 		}
1277 	}
1278 
1279 	if (ifindex) {
1280 		master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1281 		if (!master_dev)
1282 			return -EINVAL;
1283 		ops = master_dev->netdev_ops;
1284 		if (ops->ndo_add_slave) {
1285 			err = ops->ndo_add_slave(master_dev, dev);
1286 			if (err)
1287 				return err;
1288 		} else {
1289 			return -EOPNOTSUPP;
1290 		}
1291 	}
1292 	return 0;
1293 }
1294 
1295 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1296 		      struct nlattr **tb, char *ifname, int modified)
1297 {
1298 	const struct net_device_ops *ops = dev->netdev_ops;
1299 	int send_addr_notify = 0;
1300 	int err;
1301 
1302 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1303 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1304 		if (IS_ERR(net)) {
1305 			err = PTR_ERR(net);
1306 			goto errout;
1307 		}
1308 		err = dev_change_net_namespace(dev, net, ifname);
1309 		put_net(net);
1310 		if (err)
1311 			goto errout;
1312 		modified = 1;
1313 	}
1314 
1315 	if (tb[IFLA_MAP]) {
1316 		struct rtnl_link_ifmap *u_map;
1317 		struct ifmap k_map;
1318 
1319 		if (!ops->ndo_set_config) {
1320 			err = -EOPNOTSUPP;
1321 			goto errout;
1322 		}
1323 
1324 		if (!netif_device_present(dev)) {
1325 			err = -ENODEV;
1326 			goto errout;
1327 		}
1328 
1329 		u_map = nla_data(tb[IFLA_MAP]);
1330 		k_map.mem_start = (unsigned long) u_map->mem_start;
1331 		k_map.mem_end = (unsigned long) u_map->mem_end;
1332 		k_map.base_addr = (unsigned short) u_map->base_addr;
1333 		k_map.irq = (unsigned char) u_map->irq;
1334 		k_map.dma = (unsigned char) u_map->dma;
1335 		k_map.port = (unsigned char) u_map->port;
1336 
1337 		err = ops->ndo_set_config(dev, &k_map);
1338 		if (err < 0)
1339 			goto errout;
1340 
1341 		modified = 1;
1342 	}
1343 
1344 	if (tb[IFLA_ADDRESS]) {
1345 		struct sockaddr *sa;
1346 		int len;
1347 
1348 		if (!ops->ndo_set_mac_address) {
1349 			err = -EOPNOTSUPP;
1350 			goto errout;
1351 		}
1352 
1353 		if (!netif_device_present(dev)) {
1354 			err = -ENODEV;
1355 			goto errout;
1356 		}
1357 
1358 		len = sizeof(sa_family_t) + dev->addr_len;
1359 		sa = kmalloc(len, GFP_KERNEL);
1360 		if (!sa) {
1361 			err = -ENOMEM;
1362 			goto errout;
1363 		}
1364 		sa->sa_family = dev->type;
1365 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1366 		       dev->addr_len);
1367 		err = ops->ndo_set_mac_address(dev, sa);
1368 		kfree(sa);
1369 		if (err)
1370 			goto errout;
1371 		send_addr_notify = 1;
1372 		modified = 1;
1373 	}
1374 
1375 	if (tb[IFLA_MTU]) {
1376 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1377 		if (err < 0)
1378 			goto errout;
1379 		modified = 1;
1380 	}
1381 
1382 	if (tb[IFLA_GROUP]) {
1383 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1384 		modified = 1;
1385 	}
1386 
1387 	/*
1388 	 * Interface selected by interface index but interface
1389 	 * name provided implies that a name change has been
1390 	 * requested.
1391 	 */
1392 	if (ifm->ifi_index > 0 && ifname[0]) {
1393 		err = dev_change_name(dev, ifname);
1394 		if (err < 0)
1395 			goto errout;
1396 		modified = 1;
1397 	}
1398 
1399 	if (tb[IFLA_IFALIAS]) {
1400 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1401 				    nla_len(tb[IFLA_IFALIAS]));
1402 		if (err < 0)
1403 			goto errout;
1404 		modified = 1;
1405 	}
1406 
1407 	if (tb[IFLA_BROADCAST]) {
1408 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1409 		send_addr_notify = 1;
1410 	}
1411 
1412 	if (ifm->ifi_flags || ifm->ifi_change) {
1413 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1414 		if (err < 0)
1415 			goto errout;
1416 	}
1417 
1418 	if (tb[IFLA_MASTER]) {
1419 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1420 		if (err)
1421 			goto errout;
1422 		modified = 1;
1423 	}
1424 
1425 	if (tb[IFLA_TXQLEN])
1426 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1427 
1428 	if (tb[IFLA_OPERSTATE])
1429 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1430 
1431 	if (tb[IFLA_LINKMODE]) {
1432 		write_lock_bh(&dev_base_lock);
1433 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1434 		write_unlock_bh(&dev_base_lock);
1435 	}
1436 
1437 	if (tb[IFLA_VFINFO_LIST]) {
1438 		struct nlattr *attr;
1439 		int rem;
1440 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1441 			if (nla_type(attr) != IFLA_VF_INFO) {
1442 				err = -EINVAL;
1443 				goto errout;
1444 			}
1445 			err = do_setvfinfo(dev, attr);
1446 			if (err < 0)
1447 				goto errout;
1448 			modified = 1;
1449 		}
1450 	}
1451 	err = 0;
1452 
1453 	if (tb[IFLA_VF_PORTS]) {
1454 		struct nlattr *port[IFLA_PORT_MAX+1];
1455 		struct nlattr *attr;
1456 		int vf;
1457 		int rem;
1458 
1459 		err = -EOPNOTSUPP;
1460 		if (!ops->ndo_set_vf_port)
1461 			goto errout;
1462 
1463 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1464 			if (nla_type(attr) != IFLA_VF_PORT)
1465 				continue;
1466 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1467 				attr, ifla_port_policy);
1468 			if (err < 0)
1469 				goto errout;
1470 			if (!port[IFLA_PORT_VF]) {
1471 				err = -EOPNOTSUPP;
1472 				goto errout;
1473 			}
1474 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1475 			err = ops->ndo_set_vf_port(dev, vf, port);
1476 			if (err < 0)
1477 				goto errout;
1478 			modified = 1;
1479 		}
1480 	}
1481 	err = 0;
1482 
1483 	if (tb[IFLA_PORT_SELF]) {
1484 		struct nlattr *port[IFLA_PORT_MAX+1];
1485 
1486 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1487 			tb[IFLA_PORT_SELF], ifla_port_policy);
1488 		if (err < 0)
1489 			goto errout;
1490 
1491 		err = -EOPNOTSUPP;
1492 		if (ops->ndo_set_vf_port)
1493 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1494 		if (err < 0)
1495 			goto errout;
1496 		modified = 1;
1497 	}
1498 
1499 	if (tb[IFLA_AF_SPEC]) {
1500 		struct nlattr *af;
1501 		int rem;
1502 
1503 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1504 			const struct rtnl_af_ops *af_ops;
1505 
1506 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1507 				BUG();
1508 
1509 			err = af_ops->set_link_af(dev, af);
1510 			if (err < 0)
1511 				goto errout;
1512 
1513 			modified = 1;
1514 		}
1515 	}
1516 	err = 0;
1517 
1518 errout:
1519 	if (err < 0 && modified && net_ratelimit())
1520 		printk(KERN_WARNING "A link change request failed with "
1521 		       "some changes committed already. Interface %s may "
1522 		       "have been left with an inconsistent configuration, "
1523 		       "please check.\n", dev->name);
1524 
1525 	if (send_addr_notify)
1526 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1527 
1528 	return err;
1529 }
1530 
1531 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1532 {
1533 	struct net *net = sock_net(skb->sk);
1534 	struct ifinfomsg *ifm;
1535 	struct net_device *dev;
1536 	int err;
1537 	struct nlattr *tb[IFLA_MAX+1];
1538 	char ifname[IFNAMSIZ];
1539 
1540 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1541 	if (err < 0)
1542 		goto errout;
1543 
1544 	if (tb[IFLA_IFNAME])
1545 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1546 	else
1547 		ifname[0] = '\0';
1548 
1549 	err = -EINVAL;
1550 	ifm = nlmsg_data(nlh);
1551 	if (ifm->ifi_index > 0)
1552 		dev = __dev_get_by_index(net, ifm->ifi_index);
1553 	else if (tb[IFLA_IFNAME])
1554 		dev = __dev_get_by_name(net, ifname);
1555 	else
1556 		goto errout;
1557 
1558 	if (dev == NULL) {
1559 		err = -ENODEV;
1560 		goto errout;
1561 	}
1562 
1563 	err = validate_linkmsg(dev, tb);
1564 	if (err < 0)
1565 		goto errout;
1566 
1567 	err = do_setlink(dev, ifm, tb, ifname, 0);
1568 errout:
1569 	return err;
1570 }
1571 
1572 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1573 {
1574 	struct net *net = sock_net(skb->sk);
1575 	const struct rtnl_link_ops *ops;
1576 	struct net_device *dev;
1577 	struct ifinfomsg *ifm;
1578 	char ifname[IFNAMSIZ];
1579 	struct nlattr *tb[IFLA_MAX+1];
1580 	int err;
1581 	LIST_HEAD(list_kill);
1582 
1583 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1584 	if (err < 0)
1585 		return err;
1586 
1587 	if (tb[IFLA_IFNAME])
1588 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1589 
1590 	ifm = nlmsg_data(nlh);
1591 	if (ifm->ifi_index > 0)
1592 		dev = __dev_get_by_index(net, ifm->ifi_index);
1593 	else if (tb[IFLA_IFNAME])
1594 		dev = __dev_get_by_name(net, ifname);
1595 	else
1596 		return -EINVAL;
1597 
1598 	if (!dev)
1599 		return -ENODEV;
1600 
1601 	ops = dev->rtnl_link_ops;
1602 	if (!ops)
1603 		return -EOPNOTSUPP;
1604 
1605 	ops->dellink(dev, &list_kill);
1606 	unregister_netdevice_many(&list_kill);
1607 	list_del(&list_kill);
1608 	return 0;
1609 }
1610 
1611 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1612 {
1613 	unsigned int old_flags;
1614 	int err;
1615 
1616 	old_flags = dev->flags;
1617 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1618 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1619 		if (err < 0)
1620 			return err;
1621 	}
1622 
1623 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1624 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1625 
1626 	__dev_notify_flags(dev, old_flags);
1627 	return 0;
1628 }
1629 EXPORT_SYMBOL(rtnl_configure_link);
1630 
1631 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1632 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1633 {
1634 	int err;
1635 	struct net_device *dev;
1636 	unsigned int num_queues = 1;
1637 	unsigned int real_num_queues = 1;
1638 
1639 	if (ops->get_tx_queues) {
1640 		err = ops->get_tx_queues(src_net, tb, &num_queues,
1641 					 &real_num_queues);
1642 		if (err)
1643 			goto err;
1644 	}
1645 	err = -ENOMEM;
1646 	dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1647 	if (!dev)
1648 		goto err;
1649 
1650 	dev_net_set(dev, net);
1651 	dev->rtnl_link_ops = ops;
1652 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1653 
1654 	if (tb[IFLA_MTU])
1655 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1656 	if (tb[IFLA_ADDRESS])
1657 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1658 				nla_len(tb[IFLA_ADDRESS]));
1659 	if (tb[IFLA_BROADCAST])
1660 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1661 				nla_len(tb[IFLA_BROADCAST]));
1662 	if (tb[IFLA_TXQLEN])
1663 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1664 	if (tb[IFLA_OPERSTATE])
1665 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1666 	if (tb[IFLA_LINKMODE])
1667 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1668 	if (tb[IFLA_GROUP])
1669 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1670 
1671 	return dev;
1672 
1673 err:
1674 	return ERR_PTR(err);
1675 }
1676 EXPORT_SYMBOL(rtnl_create_link);
1677 
1678 static int rtnl_group_changelink(struct net *net, int group,
1679 		struct ifinfomsg *ifm,
1680 		struct nlattr **tb)
1681 {
1682 	struct net_device *dev;
1683 	int err;
1684 
1685 	for_each_netdev(net, dev) {
1686 		if (dev->group == group) {
1687 			err = do_setlink(dev, ifm, tb, NULL, 0);
1688 			if (err < 0)
1689 				return err;
1690 		}
1691 	}
1692 
1693 	return 0;
1694 }
1695 
1696 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1697 {
1698 	struct net *net = sock_net(skb->sk);
1699 	const struct rtnl_link_ops *ops;
1700 	struct net_device *dev;
1701 	struct ifinfomsg *ifm;
1702 	char kind[MODULE_NAME_LEN];
1703 	char ifname[IFNAMSIZ];
1704 	struct nlattr *tb[IFLA_MAX+1];
1705 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1706 	int err;
1707 
1708 #ifdef CONFIG_MODULES
1709 replay:
1710 #endif
1711 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1712 	if (err < 0)
1713 		return err;
1714 
1715 	if (tb[IFLA_IFNAME])
1716 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1717 	else
1718 		ifname[0] = '\0';
1719 
1720 	ifm = nlmsg_data(nlh);
1721 	if (ifm->ifi_index > 0)
1722 		dev = __dev_get_by_index(net, ifm->ifi_index);
1723 	else {
1724 		if (ifname[0])
1725 			dev = __dev_get_by_name(net, ifname);
1726 		else
1727 			dev = NULL;
1728 	}
1729 
1730 	err = validate_linkmsg(dev, tb);
1731 	if (err < 0)
1732 		return err;
1733 
1734 	if (tb[IFLA_LINKINFO]) {
1735 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1736 				       tb[IFLA_LINKINFO], ifla_info_policy);
1737 		if (err < 0)
1738 			return err;
1739 	} else
1740 		memset(linkinfo, 0, sizeof(linkinfo));
1741 
1742 	if (linkinfo[IFLA_INFO_KIND]) {
1743 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1744 		ops = rtnl_link_ops_get(kind);
1745 	} else {
1746 		kind[0] = '\0';
1747 		ops = NULL;
1748 	}
1749 
1750 	if (1) {
1751 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1752 		struct net *dest_net;
1753 
1754 		if (ops) {
1755 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1756 				err = nla_parse_nested(attr, ops->maxtype,
1757 						       linkinfo[IFLA_INFO_DATA],
1758 						       ops->policy);
1759 				if (err < 0)
1760 					return err;
1761 				data = attr;
1762 			}
1763 			if (ops->validate) {
1764 				err = ops->validate(tb, data);
1765 				if (err < 0)
1766 					return err;
1767 			}
1768 		}
1769 
1770 		if (dev) {
1771 			int modified = 0;
1772 
1773 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1774 				return -EEXIST;
1775 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1776 				return -EOPNOTSUPP;
1777 
1778 			if (linkinfo[IFLA_INFO_DATA]) {
1779 				if (!ops || ops != dev->rtnl_link_ops ||
1780 				    !ops->changelink)
1781 					return -EOPNOTSUPP;
1782 
1783 				err = ops->changelink(dev, tb, data);
1784 				if (err < 0)
1785 					return err;
1786 				modified = 1;
1787 			}
1788 
1789 			return do_setlink(dev, ifm, tb, ifname, modified);
1790 		}
1791 
1792 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1793 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1794 				return rtnl_group_changelink(net,
1795 						nla_get_u32(tb[IFLA_GROUP]),
1796 						ifm, tb);
1797 			return -ENODEV;
1798 		}
1799 
1800 		if (ifm->ifi_index)
1801 			return -EOPNOTSUPP;
1802 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1803 			return -EOPNOTSUPP;
1804 
1805 		if (!ops) {
1806 #ifdef CONFIG_MODULES
1807 			if (kind[0]) {
1808 				__rtnl_unlock();
1809 				request_module("rtnl-link-%s", kind);
1810 				rtnl_lock();
1811 				ops = rtnl_link_ops_get(kind);
1812 				if (ops)
1813 					goto replay;
1814 			}
1815 #endif
1816 			return -EOPNOTSUPP;
1817 		}
1818 
1819 		if (!ifname[0])
1820 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1821 
1822 		dest_net = rtnl_link_get_net(net, tb);
1823 		if (IS_ERR(dest_net))
1824 			return PTR_ERR(dest_net);
1825 
1826 		dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1827 
1828 		if (IS_ERR(dev))
1829 			err = PTR_ERR(dev);
1830 		else if (ops->newlink)
1831 			err = ops->newlink(net, dev, tb, data);
1832 		else
1833 			err = register_netdevice(dev);
1834 
1835 		if (err < 0 && !IS_ERR(dev))
1836 			free_netdev(dev);
1837 		if (err < 0)
1838 			goto out;
1839 
1840 		err = rtnl_configure_link(dev, ifm);
1841 		if (err < 0)
1842 			unregister_netdevice(dev);
1843 out:
1844 		put_net(dest_net);
1845 		return err;
1846 	}
1847 }
1848 
1849 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1850 {
1851 	struct net *net = sock_net(skb->sk);
1852 	struct ifinfomsg *ifm;
1853 	char ifname[IFNAMSIZ];
1854 	struct nlattr *tb[IFLA_MAX+1];
1855 	struct net_device *dev = NULL;
1856 	struct sk_buff *nskb;
1857 	int err;
1858 	u32 ext_filter_mask = 0;
1859 
1860 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1861 	if (err < 0)
1862 		return err;
1863 
1864 	if (tb[IFLA_IFNAME])
1865 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1866 
1867 	if (tb[IFLA_EXT_MASK])
1868 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1869 
1870 	ifm = nlmsg_data(nlh);
1871 	if (ifm->ifi_index > 0)
1872 		dev = __dev_get_by_index(net, ifm->ifi_index);
1873 	else if (tb[IFLA_IFNAME])
1874 		dev = __dev_get_by_name(net, ifname);
1875 	else
1876 		return -EINVAL;
1877 
1878 	if (dev == NULL)
1879 		return -ENODEV;
1880 
1881 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1882 	if (nskb == NULL)
1883 		return -ENOBUFS;
1884 
1885 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1886 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1887 	if (err < 0) {
1888 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1889 		WARN_ON(err == -EMSGSIZE);
1890 		kfree_skb(nskb);
1891 	} else
1892 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1893 
1894 	return err;
1895 }
1896 
1897 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1898 {
1899 	struct net *net = sock_net(skb->sk);
1900 	struct net_device *dev;
1901 	struct nlattr *tb[IFLA_MAX+1];
1902 	u32 ext_filter_mask = 0;
1903 	u16 min_ifinfo_dump_size = 0;
1904 
1905 	if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1906 			ifla_policy) >= 0) {
1907 		if (tb[IFLA_EXT_MASK])
1908 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1909 	}
1910 
1911 	if (!ext_filter_mask)
1912 		return NLMSG_GOODSIZE;
1913 	/*
1914 	 * traverse the list of net devices and compute the minimum
1915 	 * buffer size based upon the filter mask.
1916 	 */
1917 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1918 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1919 					     if_nlmsg_size(dev,
1920 						           ext_filter_mask));
1921 	}
1922 
1923 	return min_ifinfo_dump_size;
1924 }
1925 
1926 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1927 {
1928 	int idx;
1929 	int s_idx = cb->family;
1930 
1931 	if (s_idx == 0)
1932 		s_idx = 1;
1933 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1934 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1935 		if (idx < s_idx || idx == PF_PACKET)
1936 			continue;
1937 		if (rtnl_msg_handlers[idx] == NULL ||
1938 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1939 			continue;
1940 		if (idx > s_idx)
1941 			memset(&cb->args[0], 0, sizeof(cb->args));
1942 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1943 			break;
1944 	}
1945 	cb->family = idx;
1946 
1947 	return skb->len;
1948 }
1949 
1950 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1951 {
1952 	struct net *net = dev_net(dev);
1953 	struct sk_buff *skb;
1954 	int err = -ENOBUFS;
1955 	size_t if_info_size;
1956 
1957 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1958 	if (skb == NULL)
1959 		goto errout;
1960 
1961 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1962 	if (err < 0) {
1963 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1964 		WARN_ON(err == -EMSGSIZE);
1965 		kfree_skb(skb);
1966 		goto errout;
1967 	}
1968 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1969 	return;
1970 errout:
1971 	if (err < 0)
1972 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1973 }
1974 
1975 /* Protected by RTNL sempahore.  */
1976 static struct rtattr **rta_buf;
1977 static int rtattr_max;
1978 
1979 /* Process one rtnetlink message. */
1980 
1981 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1982 {
1983 	struct net *net = sock_net(skb->sk);
1984 	rtnl_doit_func doit;
1985 	int sz_idx, kind;
1986 	int min_len;
1987 	int family;
1988 	int type;
1989 	int err;
1990 
1991 	type = nlh->nlmsg_type;
1992 	if (type > RTM_MAX)
1993 		return -EOPNOTSUPP;
1994 
1995 	type -= RTM_BASE;
1996 
1997 	/* All the messages must have at least 1 byte length */
1998 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1999 		return 0;
2000 
2001 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2002 	sz_idx = type>>2;
2003 	kind = type&3;
2004 
2005 	if (kind != 2 && !capable(CAP_NET_ADMIN))
2006 		return -EPERM;
2007 
2008 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2009 		struct sock *rtnl;
2010 		rtnl_dumpit_func dumpit;
2011 		rtnl_calcit_func calcit;
2012 		u16 min_dump_alloc = 0;
2013 
2014 		dumpit = rtnl_get_dumpit(family, type);
2015 		if (dumpit == NULL)
2016 			return -EOPNOTSUPP;
2017 		calcit = rtnl_get_calcit(family, type);
2018 		if (calcit)
2019 			min_dump_alloc = calcit(skb, nlh);
2020 
2021 		__rtnl_unlock();
2022 		rtnl = net->rtnl;
2023 		{
2024 			struct netlink_dump_control c = {
2025 				.dump		= dumpit,
2026 				.min_dump_alloc	= min_dump_alloc,
2027 			};
2028 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2029 		}
2030 		rtnl_lock();
2031 		return err;
2032 	}
2033 
2034 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2035 
2036 	min_len = rtm_min[sz_idx];
2037 	if (nlh->nlmsg_len < min_len)
2038 		return -EINVAL;
2039 
2040 	if (nlh->nlmsg_len > min_len) {
2041 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2042 		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2043 
2044 		while (RTA_OK(attr, attrlen)) {
2045 			unsigned flavor = attr->rta_type;
2046 			if (flavor) {
2047 				if (flavor > rta_max[sz_idx])
2048 					return -EINVAL;
2049 				rta_buf[flavor-1] = attr;
2050 			}
2051 			attr = RTA_NEXT(attr, attrlen);
2052 		}
2053 	}
2054 
2055 	doit = rtnl_get_doit(family, type);
2056 	if (doit == NULL)
2057 		return -EOPNOTSUPP;
2058 
2059 	return doit(skb, nlh, (void *)&rta_buf[0]);
2060 }
2061 
2062 static void rtnetlink_rcv(struct sk_buff *skb)
2063 {
2064 	rtnl_lock();
2065 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2066 	rtnl_unlock();
2067 }
2068 
2069 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2070 {
2071 	struct net_device *dev = ptr;
2072 
2073 	switch (event) {
2074 	case NETDEV_UP:
2075 	case NETDEV_DOWN:
2076 	case NETDEV_PRE_UP:
2077 	case NETDEV_POST_INIT:
2078 	case NETDEV_REGISTER:
2079 	case NETDEV_CHANGE:
2080 	case NETDEV_PRE_TYPE_CHANGE:
2081 	case NETDEV_GOING_DOWN:
2082 	case NETDEV_UNREGISTER:
2083 	case NETDEV_UNREGISTER_BATCH:
2084 	case NETDEV_RELEASE:
2085 	case NETDEV_JOIN:
2086 		break;
2087 	default:
2088 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2089 		break;
2090 	}
2091 	return NOTIFY_DONE;
2092 }
2093 
2094 static struct notifier_block rtnetlink_dev_notifier = {
2095 	.notifier_call	= rtnetlink_event,
2096 };
2097 
2098 
2099 static int __net_init rtnetlink_net_init(struct net *net)
2100 {
2101 	struct sock *sk;
2102 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2103 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2104 	if (!sk)
2105 		return -ENOMEM;
2106 	net->rtnl = sk;
2107 	return 0;
2108 }
2109 
2110 static void __net_exit rtnetlink_net_exit(struct net *net)
2111 {
2112 	netlink_kernel_release(net->rtnl);
2113 	net->rtnl = NULL;
2114 }
2115 
2116 static struct pernet_operations rtnetlink_net_ops = {
2117 	.init = rtnetlink_net_init,
2118 	.exit = rtnetlink_net_exit,
2119 };
2120 
2121 void __init rtnetlink_init(void)
2122 {
2123 	int i;
2124 
2125 	rtattr_max = 0;
2126 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2127 		if (rta_max[i] > rtattr_max)
2128 			rtattr_max = rta_max[i];
2129 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2130 	if (!rta_buf)
2131 		panic("rtnetlink_init: cannot allocate rta_buf\n");
2132 
2133 	if (register_pernet_subsys(&rtnetlink_net_ops))
2134 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2135 
2136 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2137 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2138 
2139 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2140 		      rtnl_dump_ifinfo, rtnl_calcit);
2141 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2142 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2143 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2144 
2145 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2146 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2147 }
2148 
2149