xref: /openbmc/linux/net/core/rtnetlink.c (revision ec28dcc6b4c00b78ad269ad5b85ebd5c2d504825)
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 static inline int rtnl_vfinfo_size(const struct net_device *dev)
606 {
607 	if (dev->dev.parent && dev_is_pci(dev->dev.parent))
608 		return dev_num_vf(dev->dev.parent) *
609 			sizeof(struct ifla_vf_info);
610 	else
611 		return 0;
612 }
613 
614 static inline size_t if_nlmsg_size(const struct net_device *dev)
615 {
616 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
617 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
618 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
619 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
620 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
621 	       + nla_total_size(sizeof(struct rtnl_link_stats))
622 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
623 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
624 	       + nla_total_size(4) /* IFLA_TXQLEN */
625 	       + nla_total_size(4) /* IFLA_WEIGHT */
626 	       + nla_total_size(4) /* IFLA_MTU */
627 	       + nla_total_size(4) /* IFLA_LINK */
628 	       + nla_total_size(4) /* IFLA_MASTER */
629 	       + nla_total_size(1) /* IFLA_OPERSTATE */
630 	       + nla_total_size(1) /* IFLA_LINKMODE */
631 	       + nla_total_size(4) /* IFLA_NUM_VF */
632 	       + nla_total_size(rtnl_vfinfo_size(dev)) /* IFLA_VFINFO */
633 	       + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
634 }
635 
636 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
637 			    int type, u32 pid, u32 seq, u32 change,
638 			    unsigned int flags)
639 {
640 	struct ifinfomsg *ifm;
641 	struct nlmsghdr *nlh;
642 	const struct net_device_stats *stats;
643 	struct nlattr *attr;
644 
645 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
646 	if (nlh == NULL)
647 		return -EMSGSIZE;
648 
649 	ifm = nlmsg_data(nlh);
650 	ifm->ifi_family = AF_UNSPEC;
651 	ifm->__ifi_pad = 0;
652 	ifm->ifi_type = dev->type;
653 	ifm->ifi_index = dev->ifindex;
654 	ifm->ifi_flags = dev_get_flags(dev);
655 	ifm->ifi_change = change;
656 
657 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
658 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
659 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
660 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
661 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
662 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
663 
664 	if (dev->ifindex != dev->iflink)
665 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
666 
667 	if (dev->master)
668 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
669 
670 	if (dev->qdisc)
671 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
672 
673 	if (dev->ifalias)
674 		NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
675 
676 	if (1) {
677 		struct rtnl_link_ifmap map = {
678 			.mem_start   = dev->mem_start,
679 			.mem_end     = dev->mem_end,
680 			.base_addr   = dev->base_addr,
681 			.irq         = dev->irq,
682 			.dma         = dev->dma,
683 			.port        = dev->if_port,
684 		};
685 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
686 	}
687 
688 	if (dev->addr_len) {
689 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
690 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
691 	}
692 
693 	attr = nla_reserve(skb, IFLA_STATS,
694 			sizeof(struct rtnl_link_stats));
695 	if (attr == NULL)
696 		goto nla_put_failure;
697 
698 	stats = dev_get_stats(dev);
699 	copy_rtnl_link_stats(nla_data(attr), stats);
700 
701 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
702 		int i;
703 		struct ifla_vf_info ivi;
704 
705 		NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
706 		for (i = 0; i < dev_num_vf(dev->dev.parent); i++) {
707 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
708 				break;
709 			NLA_PUT(skb, IFLA_VFINFO, sizeof(ivi), &ivi);
710 		}
711 	}
712 	if (dev->rtnl_link_ops) {
713 		if (rtnl_link_fill(skb, dev) < 0)
714 			goto nla_put_failure;
715 	}
716 
717 	return nlmsg_end(skb, nlh);
718 
719 nla_put_failure:
720 	nlmsg_cancel(skb, nlh);
721 	return -EMSGSIZE;
722 }
723 
724 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
725 {
726 	struct net *net = sock_net(skb->sk);
727 	int h, s_h;
728 	int idx = 0, s_idx;
729 	struct net_device *dev;
730 	struct hlist_head *head;
731 	struct hlist_node *node;
732 
733 	s_h = cb->args[0];
734 	s_idx = cb->args[1];
735 
736 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
737 		idx = 0;
738 		head = &net->dev_index_head[h];
739 		hlist_for_each_entry(dev, node, head, index_hlist) {
740 			if (idx < s_idx)
741 				goto cont;
742 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
743 					     NETLINK_CB(cb->skb).pid,
744 					     cb->nlh->nlmsg_seq, 0,
745 					     NLM_F_MULTI) <= 0)
746 				goto out;
747 cont:
748 			idx++;
749 		}
750 	}
751 out:
752 	cb->args[1] = idx;
753 	cb->args[0] = h;
754 
755 	return skb->len;
756 }
757 
758 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
759 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
760 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
761 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
762 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
763 	[IFLA_MTU]		= { .type = NLA_U32 },
764 	[IFLA_LINK]		= { .type = NLA_U32 },
765 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
766 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
767 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
768 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
769 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
770 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
771 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
772 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
773 				    .len = sizeof(struct ifla_vf_mac) },
774 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
775 				    .len = sizeof(struct ifla_vf_vlan) },
776 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
777 				    .len = sizeof(struct ifla_vf_tx_rate) },
778 };
779 EXPORT_SYMBOL(ifla_policy);
780 
781 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
782 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
783 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
784 };
785 
786 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
787 {
788 	struct net *net;
789 	/* Examine the link attributes and figure out which
790 	 * network namespace we are talking about.
791 	 */
792 	if (tb[IFLA_NET_NS_PID])
793 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
794 	else
795 		net = get_net(src_net);
796 	return net;
797 }
798 EXPORT_SYMBOL(rtnl_link_get_net);
799 
800 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
801 {
802 	if (dev) {
803 		if (tb[IFLA_ADDRESS] &&
804 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
805 			return -EINVAL;
806 
807 		if (tb[IFLA_BROADCAST] &&
808 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
809 			return -EINVAL;
810 	}
811 
812 	return 0;
813 }
814 
815 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
816 		      struct nlattr **tb, char *ifname, int modified)
817 {
818 	const struct net_device_ops *ops = dev->netdev_ops;
819 	int send_addr_notify = 0;
820 	int err;
821 
822 	if (tb[IFLA_NET_NS_PID]) {
823 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
824 		if (IS_ERR(net)) {
825 			err = PTR_ERR(net);
826 			goto errout;
827 		}
828 		err = dev_change_net_namespace(dev, net, ifname);
829 		put_net(net);
830 		if (err)
831 			goto errout;
832 		modified = 1;
833 	}
834 
835 	if (tb[IFLA_MAP]) {
836 		struct rtnl_link_ifmap *u_map;
837 		struct ifmap k_map;
838 
839 		if (!ops->ndo_set_config) {
840 			err = -EOPNOTSUPP;
841 			goto errout;
842 		}
843 
844 		if (!netif_device_present(dev)) {
845 			err = -ENODEV;
846 			goto errout;
847 		}
848 
849 		u_map = nla_data(tb[IFLA_MAP]);
850 		k_map.mem_start = (unsigned long) u_map->mem_start;
851 		k_map.mem_end = (unsigned long) u_map->mem_end;
852 		k_map.base_addr = (unsigned short) u_map->base_addr;
853 		k_map.irq = (unsigned char) u_map->irq;
854 		k_map.dma = (unsigned char) u_map->dma;
855 		k_map.port = (unsigned char) u_map->port;
856 
857 		err = ops->ndo_set_config(dev, &k_map);
858 		if (err < 0)
859 			goto errout;
860 
861 		modified = 1;
862 	}
863 
864 	if (tb[IFLA_ADDRESS]) {
865 		struct sockaddr *sa;
866 		int len;
867 
868 		if (!ops->ndo_set_mac_address) {
869 			err = -EOPNOTSUPP;
870 			goto errout;
871 		}
872 
873 		if (!netif_device_present(dev)) {
874 			err = -ENODEV;
875 			goto errout;
876 		}
877 
878 		len = sizeof(sa_family_t) + dev->addr_len;
879 		sa = kmalloc(len, GFP_KERNEL);
880 		if (!sa) {
881 			err = -ENOMEM;
882 			goto errout;
883 		}
884 		sa->sa_family = dev->type;
885 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
886 		       dev->addr_len);
887 		err = ops->ndo_set_mac_address(dev, sa);
888 		kfree(sa);
889 		if (err)
890 			goto errout;
891 		send_addr_notify = 1;
892 		modified = 1;
893 	}
894 
895 	if (tb[IFLA_MTU]) {
896 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
897 		if (err < 0)
898 			goto errout;
899 		modified = 1;
900 	}
901 
902 	/*
903 	 * Interface selected by interface index but interface
904 	 * name provided implies that a name change has been
905 	 * requested.
906 	 */
907 	if (ifm->ifi_index > 0 && ifname[0]) {
908 		err = dev_change_name(dev, ifname);
909 		if (err < 0)
910 			goto errout;
911 		modified = 1;
912 	}
913 
914 	if (tb[IFLA_IFALIAS]) {
915 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
916 				    nla_len(tb[IFLA_IFALIAS]));
917 		if (err < 0)
918 			goto errout;
919 		modified = 1;
920 	}
921 
922 	if (tb[IFLA_BROADCAST]) {
923 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
924 		send_addr_notify = 1;
925 	}
926 
927 	if (ifm->ifi_flags || ifm->ifi_change) {
928 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
929 		if (err < 0)
930 			goto errout;
931 	}
932 
933 	if (tb[IFLA_TXQLEN])
934 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
935 
936 	if (tb[IFLA_OPERSTATE])
937 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
938 
939 	if (tb[IFLA_LINKMODE]) {
940 		write_lock_bh(&dev_base_lock);
941 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
942 		write_unlock_bh(&dev_base_lock);
943 	}
944 
945 	if (tb[IFLA_VF_MAC]) {
946 		struct ifla_vf_mac *ivm;
947 		ivm = nla_data(tb[IFLA_VF_MAC]);
948 		err = -EOPNOTSUPP;
949 		if (ops->ndo_set_vf_mac)
950 			err = ops->ndo_set_vf_mac(dev, ivm->vf, ivm->mac);
951 		if (err < 0)
952 			goto errout;
953 		modified = 1;
954 	}
955 
956 	if (tb[IFLA_VF_VLAN]) {
957 		struct ifla_vf_vlan *ivv;
958 		ivv = nla_data(tb[IFLA_VF_VLAN]);
959 		err = -EOPNOTSUPP;
960 		if (ops->ndo_set_vf_vlan)
961 			err = ops->ndo_set_vf_vlan(dev, ivv->vf,
962 						   ivv->vlan,
963 						   ivv->qos);
964 		if (err < 0)
965 			goto errout;
966 		modified = 1;
967 	}
968 	err = 0;
969 
970 	if (tb[IFLA_VF_TX_RATE]) {
971 		struct ifla_vf_tx_rate *ivt;
972 		ivt = nla_data(tb[IFLA_VF_TX_RATE]);
973 		err = -EOPNOTSUPP;
974 		if (ops->ndo_set_vf_tx_rate)
975 			err = ops->ndo_set_vf_tx_rate(dev, ivt->vf, ivt->rate);
976 		if (err < 0)
977 			goto errout;
978 		modified = 1;
979 	}
980 	err = 0;
981 
982 errout:
983 	if (err < 0 && modified && net_ratelimit())
984 		printk(KERN_WARNING "A link change request failed with "
985 		       "some changes comitted already. Interface %s may "
986 		       "have been left with an inconsistent configuration, "
987 		       "please check.\n", dev->name);
988 
989 	if (send_addr_notify)
990 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
991 	return err;
992 }
993 
994 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
995 {
996 	struct net *net = sock_net(skb->sk);
997 	struct ifinfomsg *ifm;
998 	struct net_device *dev;
999 	int err;
1000 	struct nlattr *tb[IFLA_MAX+1];
1001 	char ifname[IFNAMSIZ];
1002 
1003 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1004 	if (err < 0)
1005 		goto errout;
1006 
1007 	if (tb[IFLA_IFNAME])
1008 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1009 	else
1010 		ifname[0] = '\0';
1011 
1012 	err = -EINVAL;
1013 	ifm = nlmsg_data(nlh);
1014 	if (ifm->ifi_index > 0)
1015 		dev = __dev_get_by_index(net, ifm->ifi_index);
1016 	else if (tb[IFLA_IFNAME])
1017 		dev = __dev_get_by_name(net, ifname);
1018 	else
1019 		goto errout;
1020 
1021 	if (dev == NULL) {
1022 		err = -ENODEV;
1023 		goto errout;
1024 	}
1025 
1026 	err = validate_linkmsg(dev, tb);
1027 	if (err < 0)
1028 		goto errout;
1029 
1030 	err = do_setlink(dev, ifm, tb, ifname, 0);
1031 errout:
1032 	return err;
1033 }
1034 
1035 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1036 {
1037 	struct net *net = sock_net(skb->sk);
1038 	const struct rtnl_link_ops *ops;
1039 	struct net_device *dev;
1040 	struct ifinfomsg *ifm;
1041 	char ifname[IFNAMSIZ];
1042 	struct nlattr *tb[IFLA_MAX+1];
1043 	int err;
1044 
1045 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1046 	if (err < 0)
1047 		return err;
1048 
1049 	if (tb[IFLA_IFNAME])
1050 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1051 
1052 	ifm = nlmsg_data(nlh);
1053 	if (ifm->ifi_index > 0)
1054 		dev = __dev_get_by_index(net, ifm->ifi_index);
1055 	else if (tb[IFLA_IFNAME])
1056 		dev = __dev_get_by_name(net, ifname);
1057 	else
1058 		return -EINVAL;
1059 
1060 	if (!dev)
1061 		return -ENODEV;
1062 
1063 	ops = dev->rtnl_link_ops;
1064 	if (!ops)
1065 		return -EOPNOTSUPP;
1066 
1067 	ops->dellink(dev, NULL);
1068 	return 0;
1069 }
1070 
1071 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1072 {
1073 	unsigned int old_flags;
1074 	int err;
1075 
1076 	old_flags = dev->flags;
1077 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1078 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1079 		if (err < 0)
1080 			return err;
1081 	}
1082 
1083 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1084 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1085 
1086 	__dev_notify_flags(dev, old_flags);
1087 	return 0;
1088 }
1089 EXPORT_SYMBOL(rtnl_configure_link);
1090 
1091 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1092 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1093 {
1094 	int err;
1095 	struct net_device *dev;
1096 	unsigned int num_queues = 1;
1097 	unsigned int real_num_queues = 1;
1098 
1099 	if (ops->get_tx_queues) {
1100 		err = ops->get_tx_queues(src_net, tb, &num_queues,
1101 					 &real_num_queues);
1102 		if (err)
1103 			goto err;
1104 	}
1105 	err = -ENOMEM;
1106 	dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1107 	if (!dev)
1108 		goto err;
1109 
1110 	dev_net_set(dev, net);
1111 	dev->rtnl_link_ops = ops;
1112 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1113 	dev->real_num_tx_queues = real_num_queues;
1114 
1115 	if (strchr(dev->name, '%')) {
1116 		err = dev_alloc_name(dev, dev->name);
1117 		if (err < 0)
1118 			goto err_free;
1119 	}
1120 
1121 	if (tb[IFLA_MTU])
1122 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1123 	if (tb[IFLA_ADDRESS])
1124 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1125 				nla_len(tb[IFLA_ADDRESS]));
1126 	if (tb[IFLA_BROADCAST])
1127 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1128 				nla_len(tb[IFLA_BROADCAST]));
1129 	if (tb[IFLA_TXQLEN])
1130 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1131 	if (tb[IFLA_OPERSTATE])
1132 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1133 	if (tb[IFLA_LINKMODE])
1134 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1135 
1136 	return dev;
1137 
1138 err_free:
1139 	free_netdev(dev);
1140 err:
1141 	return ERR_PTR(err);
1142 }
1143 EXPORT_SYMBOL(rtnl_create_link);
1144 
1145 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1146 {
1147 	struct net *net = sock_net(skb->sk);
1148 	const struct rtnl_link_ops *ops;
1149 	struct net_device *dev;
1150 	struct ifinfomsg *ifm;
1151 	char kind[MODULE_NAME_LEN];
1152 	char ifname[IFNAMSIZ];
1153 	struct nlattr *tb[IFLA_MAX+1];
1154 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1155 	int err;
1156 
1157 #ifdef CONFIG_MODULES
1158 replay:
1159 #endif
1160 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1161 	if (err < 0)
1162 		return err;
1163 
1164 	if (tb[IFLA_IFNAME])
1165 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1166 	else
1167 		ifname[0] = '\0';
1168 
1169 	ifm = nlmsg_data(nlh);
1170 	if (ifm->ifi_index > 0)
1171 		dev = __dev_get_by_index(net, ifm->ifi_index);
1172 	else if (ifname[0])
1173 		dev = __dev_get_by_name(net, ifname);
1174 	else
1175 		dev = NULL;
1176 
1177 	err = validate_linkmsg(dev, tb);
1178 	if (err < 0)
1179 		return err;
1180 
1181 	if (tb[IFLA_LINKINFO]) {
1182 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1183 				       tb[IFLA_LINKINFO], ifla_info_policy);
1184 		if (err < 0)
1185 			return err;
1186 	} else
1187 		memset(linkinfo, 0, sizeof(linkinfo));
1188 
1189 	if (linkinfo[IFLA_INFO_KIND]) {
1190 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1191 		ops = rtnl_link_ops_get(kind);
1192 	} else {
1193 		kind[0] = '\0';
1194 		ops = NULL;
1195 	}
1196 
1197 	if (1) {
1198 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1199 		struct net *dest_net;
1200 
1201 		if (ops) {
1202 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1203 				err = nla_parse_nested(attr, ops->maxtype,
1204 						       linkinfo[IFLA_INFO_DATA],
1205 						       ops->policy);
1206 				if (err < 0)
1207 					return err;
1208 				data = attr;
1209 			}
1210 			if (ops->validate) {
1211 				err = ops->validate(tb, data);
1212 				if (err < 0)
1213 					return err;
1214 			}
1215 		}
1216 
1217 		if (dev) {
1218 			int modified = 0;
1219 
1220 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1221 				return -EEXIST;
1222 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1223 				return -EOPNOTSUPP;
1224 
1225 			if (linkinfo[IFLA_INFO_DATA]) {
1226 				if (!ops || ops != dev->rtnl_link_ops ||
1227 				    !ops->changelink)
1228 					return -EOPNOTSUPP;
1229 
1230 				err = ops->changelink(dev, tb, data);
1231 				if (err < 0)
1232 					return err;
1233 				modified = 1;
1234 			}
1235 
1236 			return do_setlink(dev, ifm, tb, ifname, modified);
1237 		}
1238 
1239 		if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1240 			return -ENODEV;
1241 
1242 		if (ifm->ifi_index)
1243 			return -EOPNOTSUPP;
1244 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1245 			return -EOPNOTSUPP;
1246 
1247 		if (!ops) {
1248 #ifdef CONFIG_MODULES
1249 			if (kind[0]) {
1250 				__rtnl_unlock();
1251 				request_module("rtnl-link-%s", kind);
1252 				rtnl_lock();
1253 				ops = rtnl_link_ops_get(kind);
1254 				if (ops)
1255 					goto replay;
1256 			}
1257 #endif
1258 			return -EOPNOTSUPP;
1259 		}
1260 
1261 		if (!ifname[0])
1262 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1263 
1264 		dest_net = rtnl_link_get_net(net, tb);
1265 		dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1266 
1267 		if (IS_ERR(dev))
1268 			err = PTR_ERR(dev);
1269 		else if (ops->newlink)
1270 			err = ops->newlink(net, dev, tb, data);
1271 		else
1272 			err = register_netdevice(dev);
1273 		if (err < 0 && !IS_ERR(dev)) {
1274 			free_netdev(dev);
1275 			goto out;
1276 		}
1277 
1278 		err = rtnl_configure_link(dev, ifm);
1279 		if (err < 0)
1280 			unregister_netdevice(dev);
1281 out:
1282 		put_net(dest_net);
1283 		return err;
1284 	}
1285 }
1286 
1287 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1288 {
1289 	struct net *net = sock_net(skb->sk);
1290 	struct ifinfomsg *ifm;
1291 	char ifname[IFNAMSIZ];
1292 	struct nlattr *tb[IFLA_MAX+1];
1293 	struct net_device *dev = NULL;
1294 	struct sk_buff *nskb;
1295 	int err;
1296 
1297 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1298 	if (err < 0)
1299 		return err;
1300 
1301 	if (tb[IFLA_IFNAME])
1302 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1303 
1304 	ifm = nlmsg_data(nlh);
1305 	if (ifm->ifi_index > 0)
1306 		dev = __dev_get_by_index(net, ifm->ifi_index);
1307 	else if (tb[IFLA_IFNAME])
1308 		dev = __dev_get_by_name(net, ifname);
1309 	else
1310 		return -EINVAL;
1311 
1312 	if (dev == NULL)
1313 		return -ENODEV;
1314 
1315 	nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1316 	if (nskb == NULL)
1317 		return -ENOBUFS;
1318 
1319 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1320 			       nlh->nlmsg_seq, 0, 0);
1321 	if (err < 0) {
1322 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1323 		WARN_ON(err == -EMSGSIZE);
1324 		kfree_skb(nskb);
1325 	} else
1326 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1327 
1328 	return err;
1329 }
1330 
1331 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1332 {
1333 	int idx;
1334 	int s_idx = cb->family;
1335 
1336 	if (s_idx == 0)
1337 		s_idx = 1;
1338 	for (idx = 1; idx < NPROTO; idx++) {
1339 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1340 		if (idx < s_idx || idx == PF_PACKET)
1341 			continue;
1342 		if (rtnl_msg_handlers[idx] == NULL ||
1343 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1344 			continue;
1345 		if (idx > s_idx)
1346 			memset(&cb->args[0], 0, sizeof(cb->args));
1347 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1348 			break;
1349 	}
1350 	cb->family = idx;
1351 
1352 	return skb->len;
1353 }
1354 
1355 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1356 {
1357 	struct net *net = dev_net(dev);
1358 	struct sk_buff *skb;
1359 	int err = -ENOBUFS;
1360 
1361 	skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1362 	if (skb == NULL)
1363 		goto errout;
1364 
1365 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1366 	if (err < 0) {
1367 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1368 		WARN_ON(err == -EMSGSIZE);
1369 		kfree_skb(skb);
1370 		goto errout;
1371 	}
1372 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1373 	return;
1374 errout:
1375 	if (err < 0)
1376 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1377 }
1378 
1379 /* Protected by RTNL sempahore.  */
1380 static struct rtattr **rta_buf;
1381 static int rtattr_max;
1382 
1383 /* Process one rtnetlink message. */
1384 
1385 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1386 {
1387 	struct net *net = sock_net(skb->sk);
1388 	rtnl_doit_func doit;
1389 	int sz_idx, kind;
1390 	int min_len;
1391 	int family;
1392 	int type;
1393 	int err;
1394 
1395 	type = nlh->nlmsg_type;
1396 	if (type > RTM_MAX)
1397 		return -EOPNOTSUPP;
1398 
1399 	type -= RTM_BASE;
1400 
1401 	/* All the messages must have at least 1 byte length */
1402 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1403 		return 0;
1404 
1405 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1406 	if (family >= NPROTO)
1407 		return -EAFNOSUPPORT;
1408 
1409 	sz_idx = type>>2;
1410 	kind = type&3;
1411 
1412 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1413 		return -EPERM;
1414 
1415 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1416 		struct sock *rtnl;
1417 		rtnl_dumpit_func dumpit;
1418 
1419 		dumpit = rtnl_get_dumpit(family, type);
1420 		if (dumpit == NULL)
1421 			return -EOPNOTSUPP;
1422 
1423 		__rtnl_unlock();
1424 		rtnl = net->rtnl;
1425 		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1426 		rtnl_lock();
1427 		return err;
1428 	}
1429 
1430 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1431 
1432 	min_len = rtm_min[sz_idx];
1433 	if (nlh->nlmsg_len < min_len)
1434 		return -EINVAL;
1435 
1436 	if (nlh->nlmsg_len > min_len) {
1437 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1438 		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1439 
1440 		while (RTA_OK(attr, attrlen)) {
1441 			unsigned flavor = attr->rta_type;
1442 			if (flavor) {
1443 				if (flavor > rta_max[sz_idx])
1444 					return -EINVAL;
1445 				rta_buf[flavor-1] = attr;
1446 			}
1447 			attr = RTA_NEXT(attr, attrlen);
1448 		}
1449 	}
1450 
1451 	doit = rtnl_get_doit(family, type);
1452 	if (doit == NULL)
1453 		return -EOPNOTSUPP;
1454 
1455 	return doit(skb, nlh, (void *)&rta_buf[0]);
1456 }
1457 
1458 static void rtnetlink_rcv(struct sk_buff *skb)
1459 {
1460 	rtnl_lock();
1461 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1462 	rtnl_unlock();
1463 }
1464 
1465 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1466 {
1467 	struct net_device *dev = ptr;
1468 
1469 	switch (event) {
1470 	case NETDEV_UP:
1471 	case NETDEV_DOWN:
1472 	case NETDEV_PRE_UP:
1473 	case NETDEV_POST_INIT:
1474 	case NETDEV_REGISTER:
1475 	case NETDEV_CHANGE:
1476 	case NETDEV_GOING_DOWN:
1477 	case NETDEV_UNREGISTER:
1478 	case NETDEV_UNREGISTER_BATCH:
1479 		break;
1480 	default:
1481 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1482 		break;
1483 	}
1484 	return NOTIFY_DONE;
1485 }
1486 
1487 static struct notifier_block rtnetlink_dev_notifier = {
1488 	.notifier_call	= rtnetlink_event,
1489 };
1490 
1491 
1492 static int __net_init rtnetlink_net_init(struct net *net)
1493 {
1494 	struct sock *sk;
1495 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1496 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1497 	if (!sk)
1498 		return -ENOMEM;
1499 	net->rtnl = sk;
1500 	return 0;
1501 }
1502 
1503 static void __net_exit rtnetlink_net_exit(struct net *net)
1504 {
1505 	netlink_kernel_release(net->rtnl);
1506 	net->rtnl = NULL;
1507 }
1508 
1509 static struct pernet_operations rtnetlink_net_ops = {
1510 	.init = rtnetlink_net_init,
1511 	.exit = rtnetlink_net_exit,
1512 };
1513 
1514 void __init rtnetlink_init(void)
1515 {
1516 	int i;
1517 
1518 	rtattr_max = 0;
1519 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1520 		if (rta_max[i] > rtattr_max)
1521 			rtattr_max = rta_max[i];
1522 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1523 	if (!rta_buf)
1524 		panic("rtnetlink_init: cannot allocate rta_buf\n");
1525 
1526 	if (register_pernet_subsys(&rtnetlink_net_ops))
1527 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
1528 
1529 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1530 	register_netdevice_notifier(&rtnetlink_dev_notifier);
1531 
1532 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1533 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1534 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1535 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1536 
1537 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1538 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1539 }
1540 
1541