xref: /linux/net/core/rtnetlink.c (revision 9f6d3c4b76314c40c866a935d78c80fd284768bd)
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 #include <asm/system.h>
42 
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <net/ip.h>
46 #include <net/protocol.h>
47 #include <net/arp.h>
48 #include <net/route.h>
49 #include <net/udp.h>
50 #include <net/sock.h>
51 #include <net/pkt_sched.h>
52 #include <net/fib_rules.h>
53 #include <net/rtnetlink.h>
54 #include <net/net_namespace.h>
55 
56 struct rtnl_link {
57 	rtnl_doit_func		doit;
58 	rtnl_dumpit_func	dumpit;
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[NPROTO];
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 	tab = rtnl_msg_handlers[protocol];
122 	if (tab == NULL || tab[msgindex].doit == NULL)
123 		tab = rtnl_msg_handlers[PF_UNSPEC];
124 
125 	return tab ? tab[msgindex].doit : NULL;
126 }
127 
128 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
129 {
130 	struct rtnl_link *tab;
131 
132 	tab = rtnl_msg_handlers[protocol];
133 	if (tab == NULL || tab[msgindex].dumpit == NULL)
134 		tab = rtnl_msg_handlers[PF_UNSPEC];
135 
136 	return tab ? tab[msgindex].dumpit : NULL;
137 }
138 
139 /**
140  * __rtnl_register - Register a rtnetlink message type
141  * @protocol: Protocol family or PF_UNSPEC
142  * @msgtype: rtnetlink message type
143  * @doit: Function pointer called for each request message
144  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
145  *
146  * Registers the specified function pointers (at least one of them has
147  * to be non-NULL) to be called whenever a request message for the
148  * specified protocol family and message type is received.
149  *
150  * The special protocol family PF_UNSPEC may be used to define fallback
151  * function pointers for the case when no entry for the specific protocol
152  * family exists.
153  *
154  * Returns 0 on success or a negative error code.
155  */
156 int __rtnl_register(int protocol, int msgtype,
157 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
158 {
159 	struct rtnl_link *tab;
160 	int msgindex;
161 
162 	BUG_ON(protocol < 0 || protocol >= NPROTO);
163 	msgindex = rtm_msgindex(msgtype);
164 
165 	tab = rtnl_msg_handlers[protocol];
166 	if (tab == NULL) {
167 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
168 		if (tab == NULL)
169 			return -ENOBUFS;
170 
171 		rtnl_msg_handlers[protocol] = tab;
172 	}
173 
174 	if (doit)
175 		tab[msgindex].doit = doit;
176 
177 	if (dumpit)
178 		tab[msgindex].dumpit = dumpit;
179 
180 	return 0;
181 }
182 EXPORT_SYMBOL_GPL(__rtnl_register);
183 
184 /**
185  * rtnl_register - Register a rtnetlink message type
186  *
187  * Identical to __rtnl_register() but panics on failure. This is useful
188  * as failure of this function is very unlikely, it can only happen due
189  * to lack of memory when allocating the chain to store all message
190  * handlers for a protocol. Meant for use in init functions where lack
191  * of memory implies no sense in continueing.
192  */
193 void rtnl_register(int protocol, int msgtype,
194 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit)
195 {
196 	if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
197 		panic("Unable to register rtnetlink message handler, "
198 		      "protocol = %d, message type = %d\n",
199 		      protocol, msgtype);
200 }
201 EXPORT_SYMBOL_GPL(rtnl_register);
202 
203 /**
204  * rtnl_unregister - Unregister a rtnetlink message type
205  * @protocol: Protocol family or PF_UNSPEC
206  * @msgtype: rtnetlink message type
207  *
208  * Returns 0 on success or a negative error code.
209  */
210 int rtnl_unregister(int protocol, int msgtype)
211 {
212 	int msgindex;
213 
214 	BUG_ON(protocol < 0 || protocol >= NPROTO);
215 	msgindex = rtm_msgindex(msgtype);
216 
217 	if (rtnl_msg_handlers[protocol] == NULL)
218 		return -ENOENT;
219 
220 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
221 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
222 
223 	return 0;
224 }
225 EXPORT_SYMBOL_GPL(rtnl_unregister);
226 
227 /**
228  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
229  * @protocol : Protocol family or PF_UNSPEC
230  *
231  * Identical to calling rtnl_unregster() for all registered message types
232  * of a certain protocol family.
233  */
234 void rtnl_unregister_all(int protocol)
235 {
236 	BUG_ON(protocol < 0 || protocol >= NPROTO);
237 
238 	kfree(rtnl_msg_handlers[protocol]);
239 	rtnl_msg_handlers[protocol] = NULL;
240 }
241 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
242 
243 static LIST_HEAD(link_ops);
244 
245 /**
246  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
247  * @ops: struct rtnl_link_ops * to register
248  *
249  * The caller must hold the rtnl_mutex. This function should be used
250  * by drivers that create devices during module initialization. It
251  * must be called before registering the devices.
252  *
253  * Returns 0 on success or a negative error code.
254  */
255 int __rtnl_link_register(struct rtnl_link_ops *ops)
256 {
257 	if (!ops->dellink)
258 		ops->dellink = unregister_netdevice_queue;
259 
260 	list_add_tail(&ops->list, &link_ops);
261 	return 0;
262 }
263 EXPORT_SYMBOL_GPL(__rtnl_link_register);
264 
265 /**
266  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
267  * @ops: struct rtnl_link_ops * to register
268  *
269  * Returns 0 on success or a negative error code.
270  */
271 int rtnl_link_register(struct rtnl_link_ops *ops)
272 {
273 	int err;
274 
275 	rtnl_lock();
276 	err = __rtnl_link_register(ops);
277 	rtnl_unlock();
278 	return err;
279 }
280 EXPORT_SYMBOL_GPL(rtnl_link_register);
281 
282 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
283 {
284 	struct net_device *dev;
285 	LIST_HEAD(list_kill);
286 
287 	for_each_netdev(net, dev) {
288 		if (dev->rtnl_link_ops == ops)
289 			ops->dellink(dev, &list_kill);
290 	}
291 	unregister_netdevice_many(&list_kill);
292 }
293 
294 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
295 {
296 	rtnl_lock();
297 	__rtnl_kill_links(net, ops);
298 	rtnl_unlock();
299 }
300 EXPORT_SYMBOL_GPL(rtnl_kill_links);
301 
302 /**
303  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
304  * @ops: struct rtnl_link_ops * to unregister
305  *
306  * The caller must hold the rtnl_mutex.
307  */
308 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
309 {
310 	struct net *net;
311 
312 	for_each_net(net) {
313 		__rtnl_kill_links(net, ops);
314 	}
315 	list_del(&ops->list);
316 }
317 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
318 
319 /**
320  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
321  * @ops: struct rtnl_link_ops * to unregister
322  */
323 void rtnl_link_unregister(struct rtnl_link_ops *ops)
324 {
325 	rtnl_lock();
326 	__rtnl_link_unregister(ops);
327 	rtnl_unlock();
328 }
329 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
330 
331 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
332 {
333 	const struct rtnl_link_ops *ops;
334 
335 	list_for_each_entry(ops, &link_ops, list) {
336 		if (!strcmp(ops->kind, kind))
337 			return ops;
338 	}
339 	return NULL;
340 }
341 
342 static size_t rtnl_link_get_size(const struct net_device *dev)
343 {
344 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
345 	size_t size;
346 
347 	if (!ops)
348 		return 0;
349 
350 	size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
351 	       nlmsg_total_size(strlen(ops->kind) + 1);	 /* IFLA_INFO_KIND */
352 
353 	if (ops->get_size)
354 		/* IFLA_INFO_DATA + nested data */
355 		size += nlmsg_total_size(sizeof(struct nlattr)) +
356 			ops->get_size(dev);
357 
358 	if (ops->get_xstats_size)
359 		size += ops->get_xstats_size(dev);	/* IFLA_INFO_XSTATS */
360 
361 	return size;
362 }
363 
364 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
365 {
366 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
367 	struct nlattr *linkinfo, *data;
368 	int err = -EMSGSIZE;
369 
370 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
371 	if (linkinfo == NULL)
372 		goto out;
373 
374 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
375 		goto err_cancel_link;
376 	if (ops->fill_xstats) {
377 		err = ops->fill_xstats(skb, dev);
378 		if (err < 0)
379 			goto err_cancel_link;
380 	}
381 	if (ops->fill_info) {
382 		data = nla_nest_start(skb, IFLA_INFO_DATA);
383 		if (data == NULL)
384 			goto err_cancel_link;
385 		err = ops->fill_info(skb, dev);
386 		if (err < 0)
387 			goto err_cancel_data;
388 		nla_nest_end(skb, data);
389 	}
390 
391 	nla_nest_end(skb, linkinfo);
392 	return 0;
393 
394 err_cancel_data:
395 	nla_nest_cancel(skb, data);
396 err_cancel_link:
397 	nla_nest_cancel(skb, linkinfo);
398 out:
399 	return err;
400 }
401 
402 static const int rtm_min[RTM_NR_FAMILIES] =
403 {
404 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
405 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
406 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
407 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
408 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
409 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
410 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
411 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
412 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
413 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
414 };
415 
416 static const int rta_max[RTM_NR_FAMILIES] =
417 {
418 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
419 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
420 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
421 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
422 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
423 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
424 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
425 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
426 };
427 
428 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
429 {
430 	struct rtattr *rta;
431 	int size = RTA_LENGTH(attrlen);
432 
433 	rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
434 	rta->rta_type = attrtype;
435 	rta->rta_len = size;
436 	memcpy(RTA_DATA(rta), data, attrlen);
437 	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
438 }
439 EXPORT_SYMBOL(__rta_fill);
440 
441 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
442 {
443 	struct sock *rtnl = net->rtnl;
444 	int err = 0;
445 
446 	NETLINK_CB(skb).dst_group = group;
447 	if (echo)
448 		atomic_inc(&skb->users);
449 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
450 	if (echo)
451 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
452 	return err;
453 }
454 
455 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
456 {
457 	struct sock *rtnl = net->rtnl;
458 
459 	return nlmsg_unicast(rtnl, skb, pid);
460 }
461 EXPORT_SYMBOL(rtnl_unicast);
462 
463 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
464 		 struct nlmsghdr *nlh, gfp_t flags)
465 {
466 	struct sock *rtnl = net->rtnl;
467 	int report = 0;
468 
469 	if (nlh)
470 		report = nlmsg_report(nlh);
471 
472 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
473 }
474 EXPORT_SYMBOL(rtnl_notify);
475 
476 void rtnl_set_sk_err(struct net *net, u32 group, int error)
477 {
478 	struct sock *rtnl = net->rtnl;
479 
480 	netlink_set_err(rtnl, 0, group, error);
481 }
482 EXPORT_SYMBOL(rtnl_set_sk_err);
483 
484 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
485 {
486 	struct nlattr *mx;
487 	int i, valid = 0;
488 
489 	mx = nla_nest_start(skb, RTA_METRICS);
490 	if (mx == NULL)
491 		return -ENOBUFS;
492 
493 	for (i = 0; i < RTAX_MAX; i++) {
494 		if (metrics[i]) {
495 			valid++;
496 			NLA_PUT_U32(skb, i+1, metrics[i]);
497 		}
498 	}
499 
500 	if (!valid) {
501 		nla_nest_cancel(skb, mx);
502 		return 0;
503 	}
504 
505 	return nla_nest_end(skb, mx);
506 
507 nla_put_failure:
508 	nla_nest_cancel(skb, mx);
509 	return -EMSGSIZE;
510 }
511 EXPORT_SYMBOL(rtnetlink_put_metrics);
512 
513 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
514 		       u32 ts, u32 tsage, long expires, u32 error)
515 {
516 	struct rta_cacheinfo ci = {
517 		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
518 		.rta_used = dst->__use,
519 		.rta_clntref = atomic_read(&(dst->__refcnt)),
520 		.rta_error = error,
521 		.rta_id =  id,
522 		.rta_ts = ts,
523 		.rta_tsage = tsage,
524 	};
525 
526 	if (expires)
527 		ci.rta_expires = jiffies_to_clock_t(expires);
528 
529 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
530 }
531 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
532 
533 static void set_operstate(struct net_device *dev, unsigned char transition)
534 {
535 	unsigned char operstate = dev->operstate;
536 
537 	switch (transition) {
538 	case IF_OPER_UP:
539 		if ((operstate == IF_OPER_DORMANT ||
540 		     operstate == IF_OPER_UNKNOWN) &&
541 		    !netif_dormant(dev))
542 			operstate = IF_OPER_UP;
543 		break;
544 
545 	case IF_OPER_DORMANT:
546 		if (operstate == IF_OPER_UP ||
547 		    operstate == IF_OPER_UNKNOWN)
548 			operstate = IF_OPER_DORMANT;
549 		break;
550 	}
551 
552 	if (dev->operstate != operstate) {
553 		write_lock_bh(&dev_base_lock);
554 		dev->operstate = operstate;
555 		write_unlock_bh(&dev_base_lock);
556 		netdev_state_change(dev);
557 	}
558 }
559 
560 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
561 					   const struct ifinfomsg *ifm)
562 {
563 	unsigned int flags = ifm->ifi_flags;
564 
565 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
566 	if (ifm->ifi_change)
567 		flags = (flags & ifm->ifi_change) |
568 			(dev->flags & ~ifm->ifi_change);
569 
570 	return flags;
571 }
572 
573 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
574 				 const struct net_device_stats *b)
575 {
576 	a->rx_packets = b->rx_packets;
577 	a->tx_packets = b->tx_packets;
578 	a->rx_bytes = b->rx_bytes;
579 	a->tx_bytes = b->tx_bytes;
580 	a->rx_errors = b->rx_errors;
581 	a->tx_errors = b->tx_errors;
582 	a->rx_dropped = b->rx_dropped;
583 	a->tx_dropped = b->tx_dropped;
584 
585 	a->multicast = b->multicast;
586 	a->collisions = b->collisions;
587 
588 	a->rx_length_errors = b->rx_length_errors;
589 	a->rx_over_errors = b->rx_over_errors;
590 	a->rx_crc_errors = b->rx_crc_errors;
591 	a->rx_frame_errors = b->rx_frame_errors;
592 	a->rx_fifo_errors = b->rx_fifo_errors;
593 	a->rx_missed_errors = b->rx_missed_errors;
594 
595 	a->tx_aborted_errors = b->tx_aborted_errors;
596 	a->tx_carrier_errors = b->tx_carrier_errors;
597 	a->tx_fifo_errors = b->tx_fifo_errors;
598 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
599 	a->tx_window_errors = b->tx_window_errors;
600 
601 	a->rx_compressed = b->rx_compressed;
602 	a->tx_compressed = b->tx_compressed;
603 };
604 
605 /* All VF info */
606 static inline int rtnl_vfinfo_size(const struct net_device *dev)
607 {
608 	if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
609 
610 		int num_vfs = dev_num_vf(dev->dev.parent);
611 		size_t size = nlmsg_total_size(sizeof(struct nlattr));
612 		size += nlmsg_total_size(num_vfs * sizeof(struct nlattr));
613 		size += num_vfs * (sizeof(struct ifla_vf_mac) +
614 				  sizeof(struct ifla_vf_vlan) +
615 				  sizeof(struct ifla_vf_tx_rate));
616 		return size;
617 	} else
618 		return 0;
619 }
620 
621 static inline size_t if_nlmsg_size(const struct net_device *dev)
622 {
623 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
624 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
625 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
626 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
627 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
628 	       + nla_total_size(sizeof(struct rtnl_link_stats))
629 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
630 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
631 	       + nla_total_size(4) /* IFLA_TXQLEN */
632 	       + nla_total_size(4) /* IFLA_WEIGHT */
633 	       + nla_total_size(4) /* IFLA_MTU */
634 	       + nla_total_size(4) /* IFLA_LINK */
635 	       + nla_total_size(4) /* IFLA_MASTER */
636 	       + nla_total_size(1) /* IFLA_OPERSTATE */
637 	       + nla_total_size(1) /* IFLA_LINKMODE */
638 	       + nla_total_size(4) /* IFLA_NUM_VF */
639 	       + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
640 	       + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
641 }
642 
643 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
644 			    int type, u32 pid, u32 seq, u32 change,
645 			    unsigned int flags)
646 {
647 	struct ifinfomsg *ifm;
648 	struct nlmsghdr *nlh;
649 	const struct net_device_stats *stats;
650 	struct nlattr *attr;
651 
652 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
653 	if (nlh == NULL)
654 		return -EMSGSIZE;
655 
656 	ifm = nlmsg_data(nlh);
657 	ifm->ifi_family = AF_UNSPEC;
658 	ifm->__ifi_pad = 0;
659 	ifm->ifi_type = dev->type;
660 	ifm->ifi_index = dev->ifindex;
661 	ifm->ifi_flags = dev_get_flags(dev);
662 	ifm->ifi_change = change;
663 
664 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
665 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
666 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
667 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
668 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
669 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
670 
671 	if (dev->ifindex != dev->iflink)
672 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
673 
674 	if (dev->master)
675 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
676 
677 	if (dev->qdisc)
678 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
679 
680 	if (dev->ifalias)
681 		NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
682 
683 	if (1) {
684 		struct rtnl_link_ifmap map = {
685 			.mem_start   = dev->mem_start,
686 			.mem_end     = dev->mem_end,
687 			.base_addr   = dev->base_addr,
688 			.irq         = dev->irq,
689 			.dma         = dev->dma,
690 			.port        = dev->if_port,
691 		};
692 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
693 	}
694 
695 	if (dev->addr_len) {
696 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
697 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
698 	}
699 
700 	attr = nla_reserve(skb, IFLA_STATS,
701 			sizeof(struct rtnl_link_stats));
702 	if (attr == NULL)
703 		goto nla_put_failure;
704 
705 	stats = dev_get_stats(dev);
706 	copy_rtnl_link_stats(nla_data(attr), stats);
707 
708 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
709 		int i;
710 
711 		struct nlattr *vfinfo, *vf;
712 		int num_vfs = dev_num_vf(dev->dev.parent);
713 
714 		NLA_PUT_U32(skb, IFLA_NUM_VF, num_vfs);
715 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
716 		if (!vfinfo)
717 			goto nla_put_failure;
718 		for (i = 0; i < num_vfs; i++) {
719 			struct ifla_vf_info ivi;
720 			struct ifla_vf_mac vf_mac;
721 			struct ifla_vf_vlan vf_vlan;
722 			struct ifla_vf_tx_rate vf_tx_rate;
723 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
724 				break;
725 			vf_mac.vf = vf_vlan.vf = vf_tx_rate.vf = ivi.vf;
726 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
727 			vf_vlan.vlan = ivi.vlan;
728 			vf_vlan.qos = ivi.qos;
729 			vf_tx_rate.rate = ivi.tx_rate;
730 			vf = nla_nest_start(skb, IFLA_VF_INFO);
731 			if (!vf) {
732 				nla_nest_cancel(skb, vfinfo);
733 				goto nla_put_failure;
734 			}
735 			NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
736 			NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
737 			NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), &vf_tx_rate);
738 			nla_nest_end(skb, vf);
739 		}
740 		nla_nest_end(skb, vfinfo);
741 	}
742 	if (dev->rtnl_link_ops) {
743 		if (rtnl_link_fill(skb, dev) < 0)
744 			goto nla_put_failure;
745 	}
746 
747 	return nlmsg_end(skb, nlh);
748 
749 nla_put_failure:
750 	nlmsg_cancel(skb, nlh);
751 	return -EMSGSIZE;
752 }
753 
754 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
755 {
756 	struct net *net = sock_net(skb->sk);
757 	int h, s_h;
758 	int idx = 0, s_idx;
759 	struct net_device *dev;
760 	struct hlist_head *head;
761 	struct hlist_node *node;
762 
763 	s_h = cb->args[0];
764 	s_idx = cb->args[1];
765 
766 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
767 		idx = 0;
768 		head = &net->dev_index_head[h];
769 		hlist_for_each_entry(dev, node, head, index_hlist) {
770 			if (idx < s_idx)
771 				goto cont;
772 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
773 					     NETLINK_CB(cb->skb).pid,
774 					     cb->nlh->nlmsg_seq, 0,
775 					     NLM_F_MULTI) <= 0)
776 				goto out;
777 cont:
778 			idx++;
779 		}
780 	}
781 out:
782 	cb->args[1] = idx;
783 	cb->args[0] = h;
784 
785 	return skb->len;
786 }
787 
788 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
789 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
790 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
791 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
792 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
793 	[IFLA_MTU]		= { .type = NLA_U32 },
794 	[IFLA_LINK]		= { .type = NLA_U32 },
795 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
796 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
797 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
798 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
799 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
800 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
801 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
802 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
803 };
804 EXPORT_SYMBOL(ifla_policy);
805 
806 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
807 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
808 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
809 };
810 
811 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
812 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
813 };
814 
815 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
816 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
817 				    .len = sizeof(struct ifla_vf_mac) },
818 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
819 				    .len = sizeof(struct ifla_vf_vlan) },
820 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
821 				    .len = sizeof(struct ifla_vf_tx_rate) },
822 };
823 
824 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
825 {
826 	struct net *net;
827 	/* Examine the link attributes and figure out which
828 	 * network namespace we are talking about.
829 	 */
830 	if (tb[IFLA_NET_NS_PID])
831 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
832 	else
833 		net = get_net(src_net);
834 	return net;
835 }
836 EXPORT_SYMBOL(rtnl_link_get_net);
837 
838 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
839 {
840 	if (dev) {
841 		if (tb[IFLA_ADDRESS] &&
842 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
843 			return -EINVAL;
844 
845 		if (tb[IFLA_BROADCAST] &&
846 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
847 			return -EINVAL;
848 	}
849 
850 	return 0;
851 }
852 
853 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
854 {
855 	int rem, err = -EINVAL;
856 	struct nlattr *vf;
857 	const struct net_device_ops *ops = dev->netdev_ops;
858 
859 	nla_for_each_nested(vf, attr, rem) {
860 		switch (nla_type(vf)) {
861 		case IFLA_VF_MAC: {
862 			struct ifla_vf_mac *ivm;
863 			ivm = nla_data(vf);
864 			err = -EOPNOTSUPP;
865 			if (ops->ndo_set_vf_mac)
866 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
867 							  ivm->mac);
868 			break;
869 		}
870 		case IFLA_VF_VLAN: {
871 			struct ifla_vf_vlan *ivv;
872 			ivv = nla_data(vf);
873 			err = -EOPNOTSUPP;
874 			if (ops->ndo_set_vf_vlan)
875 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
876 							   ivv->vlan,
877 							   ivv->qos);
878 			break;
879 		}
880 		case IFLA_VF_TX_RATE: {
881 			struct ifla_vf_tx_rate *ivt;
882 			ivt = nla_data(vf);
883 			err = -EOPNOTSUPP;
884 			if (ops->ndo_set_vf_tx_rate)
885 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
886 							      ivt->rate);
887 			break;
888 		}
889 		default:
890 			err = -EINVAL;
891 			break;
892 		}
893 		if (err)
894 			break;
895 	}
896 	return err;
897 }
898 
899 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
900 		      struct nlattr **tb, char *ifname, int modified)
901 {
902 	const struct net_device_ops *ops = dev->netdev_ops;
903 	int send_addr_notify = 0;
904 	int err;
905 
906 	if (tb[IFLA_NET_NS_PID]) {
907 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
908 		if (IS_ERR(net)) {
909 			err = PTR_ERR(net);
910 			goto errout;
911 		}
912 		err = dev_change_net_namespace(dev, net, ifname);
913 		put_net(net);
914 		if (err)
915 			goto errout;
916 		modified = 1;
917 	}
918 
919 	if (tb[IFLA_MAP]) {
920 		struct rtnl_link_ifmap *u_map;
921 		struct ifmap k_map;
922 
923 		if (!ops->ndo_set_config) {
924 			err = -EOPNOTSUPP;
925 			goto errout;
926 		}
927 
928 		if (!netif_device_present(dev)) {
929 			err = -ENODEV;
930 			goto errout;
931 		}
932 
933 		u_map = nla_data(tb[IFLA_MAP]);
934 		k_map.mem_start = (unsigned long) u_map->mem_start;
935 		k_map.mem_end = (unsigned long) u_map->mem_end;
936 		k_map.base_addr = (unsigned short) u_map->base_addr;
937 		k_map.irq = (unsigned char) u_map->irq;
938 		k_map.dma = (unsigned char) u_map->dma;
939 		k_map.port = (unsigned char) u_map->port;
940 
941 		err = ops->ndo_set_config(dev, &k_map);
942 		if (err < 0)
943 			goto errout;
944 
945 		modified = 1;
946 	}
947 
948 	if (tb[IFLA_ADDRESS]) {
949 		struct sockaddr *sa;
950 		int len;
951 
952 		if (!ops->ndo_set_mac_address) {
953 			err = -EOPNOTSUPP;
954 			goto errout;
955 		}
956 
957 		if (!netif_device_present(dev)) {
958 			err = -ENODEV;
959 			goto errout;
960 		}
961 
962 		len = sizeof(sa_family_t) + dev->addr_len;
963 		sa = kmalloc(len, GFP_KERNEL);
964 		if (!sa) {
965 			err = -ENOMEM;
966 			goto errout;
967 		}
968 		sa->sa_family = dev->type;
969 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
970 		       dev->addr_len);
971 		err = ops->ndo_set_mac_address(dev, sa);
972 		kfree(sa);
973 		if (err)
974 			goto errout;
975 		send_addr_notify = 1;
976 		modified = 1;
977 	}
978 
979 	if (tb[IFLA_MTU]) {
980 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
981 		if (err < 0)
982 			goto errout;
983 		modified = 1;
984 	}
985 
986 	/*
987 	 * Interface selected by interface index but interface
988 	 * name provided implies that a name change has been
989 	 * requested.
990 	 */
991 	if (ifm->ifi_index > 0 && ifname[0]) {
992 		err = dev_change_name(dev, ifname);
993 		if (err < 0)
994 			goto errout;
995 		modified = 1;
996 	}
997 
998 	if (tb[IFLA_IFALIAS]) {
999 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1000 				    nla_len(tb[IFLA_IFALIAS]));
1001 		if (err < 0)
1002 			goto errout;
1003 		modified = 1;
1004 	}
1005 
1006 	if (tb[IFLA_BROADCAST]) {
1007 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1008 		send_addr_notify = 1;
1009 	}
1010 
1011 	if (ifm->ifi_flags || ifm->ifi_change) {
1012 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1013 		if (err < 0)
1014 			goto errout;
1015 	}
1016 
1017 	if (tb[IFLA_TXQLEN])
1018 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1019 
1020 	if (tb[IFLA_OPERSTATE])
1021 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1022 
1023 	if (tb[IFLA_LINKMODE]) {
1024 		write_lock_bh(&dev_base_lock);
1025 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1026 		write_unlock_bh(&dev_base_lock);
1027 	}
1028 
1029 	if (tb[IFLA_VFINFO_LIST]) {
1030 		struct nlattr *attr;
1031 		int rem;
1032 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1033 			if (nla_type(attr) != IFLA_VF_INFO)
1034 				goto errout;
1035 			err = do_setvfinfo(dev, attr);
1036 			if (err < 0)
1037 				goto errout;
1038 			modified = 1;
1039 		}
1040 	}
1041 	err = 0;
1042 
1043 errout:
1044 	if (err < 0 && modified && net_ratelimit())
1045 		printk(KERN_WARNING "A link change request failed with "
1046 		       "some changes comitted already. Interface %s may "
1047 		       "have been left with an inconsistent configuration, "
1048 		       "please check.\n", dev->name);
1049 
1050 	if (send_addr_notify)
1051 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1052 	return err;
1053 }
1054 
1055 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1056 {
1057 	struct net *net = sock_net(skb->sk);
1058 	struct ifinfomsg *ifm;
1059 	struct net_device *dev;
1060 	int err;
1061 	struct nlattr *tb[IFLA_MAX+1];
1062 	char ifname[IFNAMSIZ];
1063 
1064 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1065 	if (err < 0)
1066 		goto errout;
1067 
1068 	if (tb[IFLA_IFNAME])
1069 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1070 	else
1071 		ifname[0] = '\0';
1072 
1073 	err = -EINVAL;
1074 	ifm = nlmsg_data(nlh);
1075 	if (ifm->ifi_index > 0)
1076 		dev = __dev_get_by_index(net, ifm->ifi_index);
1077 	else if (tb[IFLA_IFNAME])
1078 		dev = __dev_get_by_name(net, ifname);
1079 	else
1080 		goto errout;
1081 
1082 	if (dev == NULL) {
1083 		err = -ENODEV;
1084 		goto errout;
1085 	}
1086 
1087 	err = validate_linkmsg(dev, tb);
1088 	if (err < 0)
1089 		goto errout;
1090 
1091 	err = do_setlink(dev, ifm, tb, ifname, 0);
1092 errout:
1093 	return err;
1094 }
1095 
1096 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1097 {
1098 	struct net *net = sock_net(skb->sk);
1099 	const struct rtnl_link_ops *ops;
1100 	struct net_device *dev;
1101 	struct ifinfomsg *ifm;
1102 	char ifname[IFNAMSIZ];
1103 	struct nlattr *tb[IFLA_MAX+1];
1104 	int err;
1105 
1106 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1107 	if (err < 0)
1108 		return err;
1109 
1110 	if (tb[IFLA_IFNAME])
1111 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1112 
1113 	ifm = nlmsg_data(nlh);
1114 	if (ifm->ifi_index > 0)
1115 		dev = __dev_get_by_index(net, ifm->ifi_index);
1116 	else if (tb[IFLA_IFNAME])
1117 		dev = __dev_get_by_name(net, ifname);
1118 	else
1119 		return -EINVAL;
1120 
1121 	if (!dev)
1122 		return -ENODEV;
1123 
1124 	ops = dev->rtnl_link_ops;
1125 	if (!ops)
1126 		return -EOPNOTSUPP;
1127 
1128 	ops->dellink(dev, NULL);
1129 	return 0;
1130 }
1131 
1132 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1133 {
1134 	unsigned int old_flags;
1135 	int err;
1136 
1137 	old_flags = dev->flags;
1138 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1139 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1140 		if (err < 0)
1141 			return err;
1142 	}
1143 
1144 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1145 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1146 
1147 	__dev_notify_flags(dev, old_flags);
1148 	return 0;
1149 }
1150 EXPORT_SYMBOL(rtnl_configure_link);
1151 
1152 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1153 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1154 {
1155 	int err;
1156 	struct net_device *dev;
1157 	unsigned int num_queues = 1;
1158 	unsigned int real_num_queues = 1;
1159 
1160 	if (ops->get_tx_queues) {
1161 		err = ops->get_tx_queues(src_net, tb, &num_queues,
1162 					 &real_num_queues);
1163 		if (err)
1164 			goto err;
1165 	}
1166 	err = -ENOMEM;
1167 	dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1168 	if (!dev)
1169 		goto err;
1170 
1171 	dev_net_set(dev, net);
1172 	dev->rtnl_link_ops = ops;
1173 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1174 	dev->real_num_tx_queues = real_num_queues;
1175 
1176 	if (strchr(dev->name, '%')) {
1177 		err = dev_alloc_name(dev, dev->name);
1178 		if (err < 0)
1179 			goto err_free;
1180 	}
1181 
1182 	if (tb[IFLA_MTU])
1183 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1184 	if (tb[IFLA_ADDRESS])
1185 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1186 				nla_len(tb[IFLA_ADDRESS]));
1187 	if (tb[IFLA_BROADCAST])
1188 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1189 				nla_len(tb[IFLA_BROADCAST]));
1190 	if (tb[IFLA_TXQLEN])
1191 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1192 	if (tb[IFLA_OPERSTATE])
1193 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1194 	if (tb[IFLA_LINKMODE])
1195 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1196 
1197 	return dev;
1198 
1199 err_free:
1200 	free_netdev(dev);
1201 err:
1202 	return ERR_PTR(err);
1203 }
1204 EXPORT_SYMBOL(rtnl_create_link);
1205 
1206 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1207 {
1208 	struct net *net = sock_net(skb->sk);
1209 	const struct rtnl_link_ops *ops;
1210 	struct net_device *dev;
1211 	struct ifinfomsg *ifm;
1212 	char kind[MODULE_NAME_LEN];
1213 	char ifname[IFNAMSIZ];
1214 	struct nlattr *tb[IFLA_MAX+1];
1215 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1216 	int err;
1217 
1218 #ifdef CONFIG_MODULES
1219 replay:
1220 #endif
1221 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1222 	if (err < 0)
1223 		return err;
1224 
1225 	if (tb[IFLA_IFNAME])
1226 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1227 	else
1228 		ifname[0] = '\0';
1229 
1230 	ifm = nlmsg_data(nlh);
1231 	if (ifm->ifi_index > 0)
1232 		dev = __dev_get_by_index(net, ifm->ifi_index);
1233 	else if (ifname[0])
1234 		dev = __dev_get_by_name(net, ifname);
1235 	else
1236 		dev = NULL;
1237 
1238 	err = validate_linkmsg(dev, tb);
1239 	if (err < 0)
1240 		return err;
1241 
1242 	if (tb[IFLA_LINKINFO]) {
1243 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1244 				       tb[IFLA_LINKINFO], ifla_info_policy);
1245 		if (err < 0)
1246 			return err;
1247 	} else
1248 		memset(linkinfo, 0, sizeof(linkinfo));
1249 
1250 	if (linkinfo[IFLA_INFO_KIND]) {
1251 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1252 		ops = rtnl_link_ops_get(kind);
1253 	} else {
1254 		kind[0] = '\0';
1255 		ops = NULL;
1256 	}
1257 
1258 	if (1) {
1259 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1260 		struct net *dest_net;
1261 
1262 		if (ops) {
1263 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1264 				err = nla_parse_nested(attr, ops->maxtype,
1265 						       linkinfo[IFLA_INFO_DATA],
1266 						       ops->policy);
1267 				if (err < 0)
1268 					return err;
1269 				data = attr;
1270 			}
1271 			if (ops->validate) {
1272 				err = ops->validate(tb, data);
1273 				if (err < 0)
1274 					return err;
1275 			}
1276 		}
1277 
1278 		if (dev) {
1279 			int modified = 0;
1280 
1281 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1282 				return -EEXIST;
1283 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1284 				return -EOPNOTSUPP;
1285 
1286 			if (linkinfo[IFLA_INFO_DATA]) {
1287 				if (!ops || ops != dev->rtnl_link_ops ||
1288 				    !ops->changelink)
1289 					return -EOPNOTSUPP;
1290 
1291 				err = ops->changelink(dev, tb, data);
1292 				if (err < 0)
1293 					return err;
1294 				modified = 1;
1295 			}
1296 
1297 			return do_setlink(dev, ifm, tb, ifname, modified);
1298 		}
1299 
1300 		if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1301 			return -ENODEV;
1302 
1303 		if (ifm->ifi_index)
1304 			return -EOPNOTSUPP;
1305 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1306 			return -EOPNOTSUPP;
1307 
1308 		if (!ops) {
1309 #ifdef CONFIG_MODULES
1310 			if (kind[0]) {
1311 				__rtnl_unlock();
1312 				request_module("rtnl-link-%s", kind);
1313 				rtnl_lock();
1314 				ops = rtnl_link_ops_get(kind);
1315 				if (ops)
1316 					goto replay;
1317 			}
1318 #endif
1319 			return -EOPNOTSUPP;
1320 		}
1321 
1322 		if (!ifname[0])
1323 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1324 
1325 		dest_net = rtnl_link_get_net(net, tb);
1326 		dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1327 
1328 		if (IS_ERR(dev))
1329 			err = PTR_ERR(dev);
1330 		else if (ops->newlink)
1331 			err = ops->newlink(net, dev, tb, data);
1332 		else
1333 			err = register_netdevice(dev);
1334 
1335 		if (err < 0 && !IS_ERR(dev))
1336 			free_netdev(dev);
1337 		if (err < 0)
1338 			goto out;
1339 
1340 		err = rtnl_configure_link(dev, ifm);
1341 		if (err < 0)
1342 			unregister_netdevice(dev);
1343 out:
1344 		put_net(dest_net);
1345 		return err;
1346 	}
1347 }
1348 
1349 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1350 {
1351 	struct net *net = sock_net(skb->sk);
1352 	struct ifinfomsg *ifm;
1353 	char ifname[IFNAMSIZ];
1354 	struct nlattr *tb[IFLA_MAX+1];
1355 	struct net_device *dev = NULL;
1356 	struct sk_buff *nskb;
1357 	int err;
1358 
1359 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1360 	if (err < 0)
1361 		return err;
1362 
1363 	if (tb[IFLA_IFNAME])
1364 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1365 
1366 	ifm = nlmsg_data(nlh);
1367 	if (ifm->ifi_index > 0)
1368 		dev = __dev_get_by_index(net, ifm->ifi_index);
1369 	else if (tb[IFLA_IFNAME])
1370 		dev = __dev_get_by_name(net, ifname);
1371 	else
1372 		return -EINVAL;
1373 
1374 	if (dev == NULL)
1375 		return -ENODEV;
1376 
1377 	nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1378 	if (nskb == NULL)
1379 		return -ENOBUFS;
1380 
1381 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1382 			       nlh->nlmsg_seq, 0, 0);
1383 	if (err < 0) {
1384 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1385 		WARN_ON(err == -EMSGSIZE);
1386 		kfree_skb(nskb);
1387 	} else
1388 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1389 
1390 	return err;
1391 }
1392 
1393 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1394 {
1395 	int idx;
1396 	int s_idx = cb->family;
1397 
1398 	if (s_idx == 0)
1399 		s_idx = 1;
1400 	for (idx = 1; idx < NPROTO; idx++) {
1401 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1402 		if (idx < s_idx || idx == PF_PACKET)
1403 			continue;
1404 		if (rtnl_msg_handlers[idx] == NULL ||
1405 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1406 			continue;
1407 		if (idx > s_idx)
1408 			memset(&cb->args[0], 0, sizeof(cb->args));
1409 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1410 			break;
1411 	}
1412 	cb->family = idx;
1413 
1414 	return skb->len;
1415 }
1416 
1417 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1418 {
1419 	struct net *net = dev_net(dev);
1420 	struct sk_buff *skb;
1421 	int err = -ENOBUFS;
1422 
1423 	skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1424 	if (skb == NULL)
1425 		goto errout;
1426 
1427 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1428 	if (err < 0) {
1429 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1430 		WARN_ON(err == -EMSGSIZE);
1431 		kfree_skb(skb);
1432 		goto errout;
1433 	}
1434 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1435 	return;
1436 errout:
1437 	if (err < 0)
1438 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1439 }
1440 
1441 /* Protected by RTNL sempahore.  */
1442 static struct rtattr **rta_buf;
1443 static int rtattr_max;
1444 
1445 /* Process one rtnetlink message. */
1446 
1447 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1448 {
1449 	struct net *net = sock_net(skb->sk);
1450 	rtnl_doit_func doit;
1451 	int sz_idx, kind;
1452 	int min_len;
1453 	int family;
1454 	int type;
1455 	int err;
1456 
1457 	type = nlh->nlmsg_type;
1458 	if (type > RTM_MAX)
1459 		return -EOPNOTSUPP;
1460 
1461 	type -= RTM_BASE;
1462 
1463 	/* All the messages must have at least 1 byte length */
1464 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1465 		return 0;
1466 
1467 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1468 	if (family >= NPROTO)
1469 		return -EAFNOSUPPORT;
1470 
1471 	sz_idx = type>>2;
1472 	kind = type&3;
1473 
1474 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1475 		return -EPERM;
1476 
1477 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1478 		struct sock *rtnl;
1479 		rtnl_dumpit_func dumpit;
1480 
1481 		dumpit = rtnl_get_dumpit(family, type);
1482 		if (dumpit == NULL)
1483 			return -EOPNOTSUPP;
1484 
1485 		__rtnl_unlock();
1486 		rtnl = net->rtnl;
1487 		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1488 		rtnl_lock();
1489 		return err;
1490 	}
1491 
1492 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1493 
1494 	min_len = rtm_min[sz_idx];
1495 	if (nlh->nlmsg_len < min_len)
1496 		return -EINVAL;
1497 
1498 	if (nlh->nlmsg_len > min_len) {
1499 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1500 		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1501 
1502 		while (RTA_OK(attr, attrlen)) {
1503 			unsigned flavor = attr->rta_type;
1504 			if (flavor) {
1505 				if (flavor > rta_max[sz_idx])
1506 					return -EINVAL;
1507 				rta_buf[flavor-1] = attr;
1508 			}
1509 			attr = RTA_NEXT(attr, attrlen);
1510 		}
1511 	}
1512 
1513 	doit = rtnl_get_doit(family, type);
1514 	if (doit == NULL)
1515 		return -EOPNOTSUPP;
1516 
1517 	return doit(skb, nlh, (void *)&rta_buf[0]);
1518 }
1519 
1520 static void rtnetlink_rcv(struct sk_buff *skb)
1521 {
1522 	rtnl_lock();
1523 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1524 	rtnl_unlock();
1525 }
1526 
1527 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1528 {
1529 	struct net_device *dev = ptr;
1530 
1531 	switch (event) {
1532 	case NETDEV_UP:
1533 	case NETDEV_DOWN:
1534 	case NETDEV_PRE_UP:
1535 	case NETDEV_POST_INIT:
1536 	case NETDEV_REGISTER:
1537 	case NETDEV_CHANGE:
1538 	case NETDEV_GOING_DOWN:
1539 	case NETDEV_UNREGISTER:
1540 	case NETDEV_UNREGISTER_BATCH:
1541 		break;
1542 	default:
1543 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1544 		break;
1545 	}
1546 	return NOTIFY_DONE;
1547 }
1548 
1549 static struct notifier_block rtnetlink_dev_notifier = {
1550 	.notifier_call	= rtnetlink_event,
1551 };
1552 
1553 
1554 static int __net_init rtnetlink_net_init(struct net *net)
1555 {
1556 	struct sock *sk;
1557 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1558 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1559 	if (!sk)
1560 		return -ENOMEM;
1561 	net->rtnl = sk;
1562 	return 0;
1563 }
1564 
1565 static void __net_exit rtnetlink_net_exit(struct net *net)
1566 {
1567 	netlink_kernel_release(net->rtnl);
1568 	net->rtnl = NULL;
1569 }
1570 
1571 static struct pernet_operations rtnetlink_net_ops = {
1572 	.init = rtnetlink_net_init,
1573 	.exit = rtnetlink_net_exit,
1574 };
1575 
1576 void __init rtnetlink_init(void)
1577 {
1578 	int i;
1579 
1580 	rtattr_max = 0;
1581 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1582 		if (rta_max[i] > rtattr_max)
1583 			rtattr_max = rta_max[i];
1584 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1585 	if (!rta_buf)
1586 		panic("rtnetlink_init: cannot allocate rta_buf\n");
1587 
1588 	if (register_pernet_subsys(&rtnetlink_net_ops))
1589 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
1590 
1591 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1592 	register_netdevice_notifier(&rtnetlink_dev_notifier);
1593 
1594 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1595 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1596 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1597 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1598 
1599 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1600 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1601 }
1602 
1603