xref: /openbmc/linux/net/core/rtnetlink.c (revision b6dcefde)
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 
39 #include <asm/uaccess.h>
40 #include <asm/system.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 };
59 
60 static DEFINE_MUTEX(rtnl_mutex);
61 
62 void rtnl_lock(void)
63 {
64 	mutex_lock(&rtnl_mutex);
65 }
66 EXPORT_SYMBOL(rtnl_lock);
67 
68 void __rtnl_unlock(void)
69 {
70 	mutex_unlock(&rtnl_mutex);
71 }
72 
73 void rtnl_unlock(void)
74 {
75 	/* This fellow will unlock it for us. */
76 	netdev_run_todo();
77 }
78 EXPORT_SYMBOL(rtnl_unlock);
79 
80 int rtnl_trylock(void)
81 {
82 	return mutex_trylock(&rtnl_mutex);
83 }
84 EXPORT_SYMBOL(rtnl_trylock);
85 
86 int rtnl_is_locked(void)
87 {
88 	return mutex_is_locked(&rtnl_mutex);
89 }
90 EXPORT_SYMBOL(rtnl_is_locked);
91 
92 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
93 
94 static inline int rtm_msgindex(int msgtype)
95 {
96 	int msgindex = msgtype - RTM_BASE;
97 
98 	/*
99 	 * msgindex < 0 implies someone tried to register a netlink
100 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
101 	 * the message type has not been added to linux/rtnetlink.h
102 	 */
103 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
104 
105 	return msgindex;
106 }
107 
108 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
109 {
110 	struct rtnl_link *tab;
111 
112 	tab = rtnl_msg_handlers[protocol];
113 	if (tab == NULL || tab[msgindex].doit == NULL)
114 		tab = rtnl_msg_handlers[PF_UNSPEC];
115 
116 	return tab ? tab[msgindex].doit : NULL;
117 }
118 
119 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
120 {
121 	struct rtnl_link *tab;
122 
123 	tab = rtnl_msg_handlers[protocol];
124 	if (tab == NULL || tab[msgindex].dumpit == NULL)
125 		tab = rtnl_msg_handlers[PF_UNSPEC];
126 
127 	return tab ? tab[msgindex].dumpit : NULL;
128 }
129 
130 /**
131  * __rtnl_register - Register a rtnetlink message type
132  * @protocol: Protocol family or PF_UNSPEC
133  * @msgtype: rtnetlink message type
134  * @doit: Function pointer called for each request message
135  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
136  *
137  * Registers the specified function pointers (at least one of them has
138  * to be non-NULL) to be called whenever a request message for the
139  * specified protocol family and message type is received.
140  *
141  * The special protocol family PF_UNSPEC may be used to define fallback
142  * function pointers for the case when no entry for the specific protocol
143  * family exists.
144  *
145  * Returns 0 on success or a negative error code.
146  */
147 int __rtnl_register(int protocol, int msgtype,
148 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
149 {
150 	struct rtnl_link *tab;
151 	int msgindex;
152 
153 	BUG_ON(protocol < 0 || protocol >= NPROTO);
154 	msgindex = rtm_msgindex(msgtype);
155 
156 	tab = rtnl_msg_handlers[protocol];
157 	if (tab == NULL) {
158 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
159 		if (tab == NULL)
160 			return -ENOBUFS;
161 
162 		rtnl_msg_handlers[protocol] = tab;
163 	}
164 
165 	if (doit)
166 		tab[msgindex].doit = doit;
167 
168 	if (dumpit)
169 		tab[msgindex].dumpit = dumpit;
170 
171 	return 0;
172 }
173 EXPORT_SYMBOL_GPL(__rtnl_register);
174 
175 /**
176  * rtnl_register - Register a rtnetlink message type
177  *
178  * Identical to __rtnl_register() but panics on failure. This is useful
179  * as failure of this function is very unlikely, it can only happen due
180  * to lack of memory when allocating the chain to store all message
181  * handlers for a protocol. Meant for use in init functions where lack
182  * of memory implies no sense in continueing.
183  */
184 void rtnl_register(int protocol, int msgtype,
185 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit)
186 {
187 	if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
188 		panic("Unable to register rtnetlink message handler, "
189 		      "protocol = %d, message type = %d\n",
190 		      protocol, msgtype);
191 }
192 EXPORT_SYMBOL_GPL(rtnl_register);
193 
194 /**
195  * rtnl_unregister - Unregister a rtnetlink message type
196  * @protocol: Protocol family or PF_UNSPEC
197  * @msgtype: rtnetlink message type
198  *
199  * Returns 0 on success or a negative error code.
200  */
201 int rtnl_unregister(int protocol, int msgtype)
202 {
203 	int msgindex;
204 
205 	BUG_ON(protocol < 0 || protocol >= NPROTO);
206 	msgindex = rtm_msgindex(msgtype);
207 
208 	if (rtnl_msg_handlers[protocol] == NULL)
209 		return -ENOENT;
210 
211 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
212 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
213 
214 	return 0;
215 }
216 EXPORT_SYMBOL_GPL(rtnl_unregister);
217 
218 /**
219  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
220  * @protocol : Protocol family or PF_UNSPEC
221  *
222  * Identical to calling rtnl_unregster() for all registered message types
223  * of a certain protocol family.
224  */
225 void rtnl_unregister_all(int protocol)
226 {
227 	BUG_ON(protocol < 0 || protocol >= NPROTO);
228 
229 	kfree(rtnl_msg_handlers[protocol]);
230 	rtnl_msg_handlers[protocol] = NULL;
231 }
232 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
233 
234 static LIST_HEAD(link_ops);
235 
236 /**
237  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
238  * @ops: struct rtnl_link_ops * to register
239  *
240  * The caller must hold the rtnl_mutex. This function should be used
241  * by drivers that create devices during module initialization. It
242  * must be called before registering the devices.
243  *
244  * Returns 0 on success or a negative error code.
245  */
246 int __rtnl_link_register(struct rtnl_link_ops *ops)
247 {
248 	if (!ops->dellink)
249 		ops->dellink = unregister_netdevice_queue;
250 
251 	list_add_tail(&ops->list, &link_ops);
252 	return 0;
253 }
254 EXPORT_SYMBOL_GPL(__rtnl_link_register);
255 
256 /**
257  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
258  * @ops: struct rtnl_link_ops * to register
259  *
260  * Returns 0 on success or a negative error code.
261  */
262 int rtnl_link_register(struct rtnl_link_ops *ops)
263 {
264 	int err;
265 
266 	rtnl_lock();
267 	err = __rtnl_link_register(ops);
268 	rtnl_unlock();
269 	return err;
270 }
271 EXPORT_SYMBOL_GPL(rtnl_link_register);
272 
273 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
274 {
275 	struct net_device *dev;
276 	LIST_HEAD(list_kill);
277 
278 	for_each_netdev(net, dev) {
279 		if (dev->rtnl_link_ops == ops)
280 			ops->dellink(dev, &list_kill);
281 	}
282 	unregister_netdevice_many(&list_kill);
283 }
284 
285 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
286 {
287 	rtnl_lock();
288 	__rtnl_kill_links(net, ops);
289 	rtnl_unlock();
290 }
291 EXPORT_SYMBOL_GPL(rtnl_kill_links);
292 
293 /**
294  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
295  * @ops: struct rtnl_link_ops * to unregister
296  *
297  * The caller must hold the rtnl_mutex.
298  */
299 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
300 {
301 	struct net *net;
302 
303 	for_each_net(net) {
304 		__rtnl_kill_links(net, ops);
305 	}
306 	list_del(&ops->list);
307 }
308 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
309 
310 /**
311  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
312  * @ops: struct rtnl_link_ops * to unregister
313  */
314 void rtnl_link_unregister(struct rtnl_link_ops *ops)
315 {
316 	rtnl_lock();
317 	__rtnl_link_unregister(ops);
318 	rtnl_unlock();
319 }
320 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
321 
322 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
323 {
324 	const struct rtnl_link_ops *ops;
325 
326 	list_for_each_entry(ops, &link_ops, list) {
327 		if (!strcmp(ops->kind, kind))
328 			return ops;
329 	}
330 	return NULL;
331 }
332 
333 static size_t rtnl_link_get_size(const struct net_device *dev)
334 {
335 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
336 	size_t size;
337 
338 	if (!ops)
339 		return 0;
340 
341 	size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
342 	       nlmsg_total_size(strlen(ops->kind) + 1);	 /* IFLA_INFO_KIND */
343 
344 	if (ops->get_size)
345 		/* IFLA_INFO_DATA + nested data */
346 		size += nlmsg_total_size(sizeof(struct nlattr)) +
347 			ops->get_size(dev);
348 
349 	if (ops->get_xstats_size)
350 		size += ops->get_xstats_size(dev);	/* IFLA_INFO_XSTATS */
351 
352 	return size;
353 }
354 
355 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
356 {
357 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
358 	struct nlattr *linkinfo, *data;
359 	int err = -EMSGSIZE;
360 
361 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
362 	if (linkinfo == NULL)
363 		goto out;
364 
365 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
366 		goto err_cancel_link;
367 	if (ops->fill_xstats) {
368 		err = ops->fill_xstats(skb, dev);
369 		if (err < 0)
370 			goto err_cancel_link;
371 	}
372 	if (ops->fill_info) {
373 		data = nla_nest_start(skb, IFLA_INFO_DATA);
374 		if (data == NULL)
375 			goto err_cancel_link;
376 		err = ops->fill_info(skb, dev);
377 		if (err < 0)
378 			goto err_cancel_data;
379 		nla_nest_end(skb, data);
380 	}
381 
382 	nla_nest_end(skb, linkinfo);
383 	return 0;
384 
385 err_cancel_data:
386 	nla_nest_cancel(skb, data);
387 err_cancel_link:
388 	nla_nest_cancel(skb, linkinfo);
389 out:
390 	return err;
391 }
392 
393 static const int rtm_min[RTM_NR_FAMILIES] =
394 {
395 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
396 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
397 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
398 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
399 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
400 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
401 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
402 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
403 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
404 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
405 };
406 
407 static const int rta_max[RTM_NR_FAMILIES] =
408 {
409 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
410 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
411 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
412 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
413 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
414 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
415 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
416 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
417 };
418 
419 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
420 {
421 	struct rtattr *rta;
422 	int size = RTA_LENGTH(attrlen);
423 
424 	rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
425 	rta->rta_type = attrtype;
426 	rta->rta_len = size;
427 	memcpy(RTA_DATA(rta), data, attrlen);
428 	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
429 }
430 EXPORT_SYMBOL(__rta_fill);
431 
432 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
433 {
434 	struct sock *rtnl = net->rtnl;
435 	int err = 0;
436 
437 	NETLINK_CB(skb).dst_group = group;
438 	if (echo)
439 		atomic_inc(&skb->users);
440 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
441 	if (echo)
442 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
443 	return err;
444 }
445 
446 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
447 {
448 	struct sock *rtnl = net->rtnl;
449 
450 	return nlmsg_unicast(rtnl, skb, pid);
451 }
452 EXPORT_SYMBOL(rtnl_unicast);
453 
454 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
455 		 struct nlmsghdr *nlh, gfp_t flags)
456 {
457 	struct sock *rtnl = net->rtnl;
458 	int report = 0;
459 
460 	if (nlh)
461 		report = nlmsg_report(nlh);
462 
463 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
464 }
465 EXPORT_SYMBOL(rtnl_notify);
466 
467 void rtnl_set_sk_err(struct net *net, u32 group, int error)
468 {
469 	struct sock *rtnl = net->rtnl;
470 
471 	netlink_set_err(rtnl, 0, group, error);
472 }
473 EXPORT_SYMBOL(rtnl_set_sk_err);
474 
475 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
476 {
477 	struct nlattr *mx;
478 	int i, valid = 0;
479 
480 	mx = nla_nest_start(skb, RTA_METRICS);
481 	if (mx == NULL)
482 		return -ENOBUFS;
483 
484 	for (i = 0; i < RTAX_MAX; i++) {
485 		if (metrics[i]) {
486 			valid++;
487 			NLA_PUT_U32(skb, i+1, metrics[i]);
488 		}
489 	}
490 
491 	if (!valid) {
492 		nla_nest_cancel(skb, mx);
493 		return 0;
494 	}
495 
496 	return nla_nest_end(skb, mx);
497 
498 nla_put_failure:
499 	nla_nest_cancel(skb, mx);
500 	return -EMSGSIZE;
501 }
502 EXPORT_SYMBOL(rtnetlink_put_metrics);
503 
504 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
505 		       u32 ts, u32 tsage, long expires, u32 error)
506 {
507 	struct rta_cacheinfo ci = {
508 		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
509 		.rta_used = dst->__use,
510 		.rta_clntref = atomic_read(&(dst->__refcnt)),
511 		.rta_error = error,
512 		.rta_id =  id,
513 		.rta_ts = ts,
514 		.rta_tsage = tsage,
515 	};
516 
517 	if (expires)
518 		ci.rta_expires = jiffies_to_clock_t(expires);
519 
520 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
521 }
522 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
523 
524 static void set_operstate(struct net_device *dev, unsigned char transition)
525 {
526 	unsigned char operstate = dev->operstate;
527 
528 	switch (transition) {
529 	case IF_OPER_UP:
530 		if ((operstate == IF_OPER_DORMANT ||
531 		     operstate == IF_OPER_UNKNOWN) &&
532 		    !netif_dormant(dev))
533 			operstate = IF_OPER_UP;
534 		break;
535 
536 	case IF_OPER_DORMANT:
537 		if (operstate == IF_OPER_UP ||
538 		    operstate == IF_OPER_UNKNOWN)
539 			operstate = IF_OPER_DORMANT;
540 		break;
541 	}
542 
543 	if (dev->operstate != operstate) {
544 		write_lock_bh(&dev_base_lock);
545 		dev->operstate = operstate;
546 		write_unlock_bh(&dev_base_lock);
547 		netdev_state_change(dev);
548 	}
549 }
550 
551 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
552 				 const struct net_device_stats *b)
553 {
554 	a->rx_packets = b->rx_packets;
555 	a->tx_packets = b->tx_packets;
556 	a->rx_bytes = b->rx_bytes;
557 	a->tx_bytes = b->tx_bytes;
558 	a->rx_errors = b->rx_errors;
559 	a->tx_errors = b->tx_errors;
560 	a->rx_dropped = b->rx_dropped;
561 	a->tx_dropped = b->tx_dropped;
562 
563 	a->multicast = b->multicast;
564 	a->collisions = b->collisions;
565 
566 	a->rx_length_errors = b->rx_length_errors;
567 	a->rx_over_errors = b->rx_over_errors;
568 	a->rx_crc_errors = b->rx_crc_errors;
569 	a->rx_frame_errors = b->rx_frame_errors;
570 	a->rx_fifo_errors = b->rx_fifo_errors;
571 	a->rx_missed_errors = b->rx_missed_errors;
572 
573 	a->tx_aborted_errors = b->tx_aborted_errors;
574 	a->tx_carrier_errors = b->tx_carrier_errors;
575 	a->tx_fifo_errors = b->tx_fifo_errors;
576 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
577 	a->tx_window_errors = b->tx_window_errors;
578 
579 	a->rx_compressed = b->rx_compressed;
580 	a->tx_compressed = b->tx_compressed;
581 };
582 
583 static inline size_t if_nlmsg_size(const struct net_device *dev)
584 {
585 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
586 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
587 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
588 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
589 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
590 	       + nla_total_size(sizeof(struct rtnl_link_stats))
591 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
592 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
593 	       + nla_total_size(4) /* IFLA_TXQLEN */
594 	       + nla_total_size(4) /* IFLA_WEIGHT */
595 	       + nla_total_size(4) /* IFLA_MTU */
596 	       + nla_total_size(4) /* IFLA_LINK */
597 	       + nla_total_size(4) /* IFLA_MASTER */
598 	       + nla_total_size(1) /* IFLA_OPERSTATE */
599 	       + nla_total_size(1) /* IFLA_LINKMODE */
600 	       + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
601 }
602 
603 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
604 			    int type, u32 pid, u32 seq, u32 change,
605 			    unsigned int flags)
606 {
607 	struct ifinfomsg *ifm;
608 	struct nlmsghdr *nlh;
609 	const struct net_device_stats *stats;
610 	struct nlattr *attr;
611 
612 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
613 	if (nlh == NULL)
614 		return -EMSGSIZE;
615 
616 	ifm = nlmsg_data(nlh);
617 	ifm->ifi_family = AF_UNSPEC;
618 	ifm->__ifi_pad = 0;
619 	ifm->ifi_type = dev->type;
620 	ifm->ifi_index = dev->ifindex;
621 	ifm->ifi_flags = dev_get_flags(dev);
622 	ifm->ifi_change = change;
623 
624 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
625 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
626 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
627 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
628 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
629 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
630 
631 	if (dev->ifindex != dev->iflink)
632 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
633 
634 	if (dev->master)
635 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
636 
637 	if (dev->qdisc)
638 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
639 
640 	if (dev->ifalias)
641 		NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
642 
643 	if (1) {
644 		struct rtnl_link_ifmap map = {
645 			.mem_start   = dev->mem_start,
646 			.mem_end     = dev->mem_end,
647 			.base_addr   = dev->base_addr,
648 			.irq         = dev->irq,
649 			.dma         = dev->dma,
650 			.port        = dev->if_port,
651 		};
652 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
653 	}
654 
655 	if (dev->addr_len) {
656 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
657 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
658 	}
659 
660 	attr = nla_reserve(skb, IFLA_STATS,
661 			sizeof(struct rtnl_link_stats));
662 	if (attr == NULL)
663 		goto nla_put_failure;
664 
665 	stats = dev_get_stats(dev);
666 	copy_rtnl_link_stats(nla_data(attr), stats);
667 
668 	if (dev->rtnl_link_ops) {
669 		if (rtnl_link_fill(skb, dev) < 0)
670 			goto nla_put_failure;
671 	}
672 
673 	return nlmsg_end(skb, nlh);
674 
675 nla_put_failure:
676 	nlmsg_cancel(skb, nlh);
677 	return -EMSGSIZE;
678 }
679 
680 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
681 {
682 	struct net *net = sock_net(skb->sk);
683 	int h, s_h;
684 	int idx = 0, s_idx;
685 	struct net_device *dev;
686 	struct hlist_head *head;
687 	struct hlist_node *node;
688 
689 	s_h = cb->args[0];
690 	s_idx = cb->args[1];
691 
692 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
693 		idx = 0;
694 		head = &net->dev_index_head[h];
695 		hlist_for_each_entry(dev, node, head, index_hlist) {
696 			if (idx < s_idx)
697 				goto cont;
698 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
699 					     NETLINK_CB(cb->skb).pid,
700 					     cb->nlh->nlmsg_seq, 0,
701 					     NLM_F_MULTI) <= 0)
702 				goto out;
703 cont:
704 			idx++;
705 		}
706 	}
707 out:
708 	cb->args[1] = idx;
709 	cb->args[0] = h;
710 
711 	return skb->len;
712 }
713 
714 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
715 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
716 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
717 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
718 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
719 	[IFLA_MTU]		= { .type = NLA_U32 },
720 	[IFLA_LINK]		= { .type = NLA_U32 },
721 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
722 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
723 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
724 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
725 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
726 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
727 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
728 };
729 EXPORT_SYMBOL(ifla_policy);
730 
731 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
732 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
733 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
734 };
735 
736 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
737 {
738 	struct net *net;
739 	/* Examine the link attributes and figure out which
740 	 * network namespace we are talking about.
741 	 */
742 	if (tb[IFLA_NET_NS_PID])
743 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
744 	else
745 		net = get_net(src_net);
746 	return net;
747 }
748 EXPORT_SYMBOL(rtnl_link_get_net);
749 
750 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
751 {
752 	if (dev) {
753 		if (tb[IFLA_ADDRESS] &&
754 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
755 			return -EINVAL;
756 
757 		if (tb[IFLA_BROADCAST] &&
758 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
759 			return -EINVAL;
760 	}
761 
762 	return 0;
763 }
764 
765 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
766 		      struct nlattr **tb, char *ifname, int modified)
767 {
768 	const struct net_device_ops *ops = dev->netdev_ops;
769 	int send_addr_notify = 0;
770 	int err;
771 
772 	if (tb[IFLA_NET_NS_PID]) {
773 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
774 		if (IS_ERR(net)) {
775 			err = PTR_ERR(net);
776 			goto errout;
777 		}
778 		err = dev_change_net_namespace(dev, net, ifname);
779 		put_net(net);
780 		if (err)
781 			goto errout;
782 		modified = 1;
783 	}
784 
785 	if (tb[IFLA_MAP]) {
786 		struct rtnl_link_ifmap *u_map;
787 		struct ifmap k_map;
788 
789 		if (!ops->ndo_set_config) {
790 			err = -EOPNOTSUPP;
791 			goto errout;
792 		}
793 
794 		if (!netif_device_present(dev)) {
795 			err = -ENODEV;
796 			goto errout;
797 		}
798 
799 		u_map = nla_data(tb[IFLA_MAP]);
800 		k_map.mem_start = (unsigned long) u_map->mem_start;
801 		k_map.mem_end = (unsigned long) u_map->mem_end;
802 		k_map.base_addr = (unsigned short) u_map->base_addr;
803 		k_map.irq = (unsigned char) u_map->irq;
804 		k_map.dma = (unsigned char) u_map->dma;
805 		k_map.port = (unsigned char) u_map->port;
806 
807 		err = ops->ndo_set_config(dev, &k_map);
808 		if (err < 0)
809 			goto errout;
810 
811 		modified = 1;
812 	}
813 
814 	if (tb[IFLA_ADDRESS]) {
815 		struct sockaddr *sa;
816 		int len;
817 
818 		if (!ops->ndo_set_mac_address) {
819 			err = -EOPNOTSUPP;
820 			goto errout;
821 		}
822 
823 		if (!netif_device_present(dev)) {
824 			err = -ENODEV;
825 			goto errout;
826 		}
827 
828 		len = sizeof(sa_family_t) + dev->addr_len;
829 		sa = kmalloc(len, GFP_KERNEL);
830 		if (!sa) {
831 			err = -ENOMEM;
832 			goto errout;
833 		}
834 		sa->sa_family = dev->type;
835 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
836 		       dev->addr_len);
837 		err = ops->ndo_set_mac_address(dev, sa);
838 		kfree(sa);
839 		if (err)
840 			goto errout;
841 		send_addr_notify = 1;
842 		modified = 1;
843 	}
844 
845 	if (tb[IFLA_MTU]) {
846 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
847 		if (err < 0)
848 			goto errout;
849 		modified = 1;
850 	}
851 
852 	/*
853 	 * Interface selected by interface index but interface
854 	 * name provided implies that a name change has been
855 	 * requested.
856 	 */
857 	if (ifm->ifi_index > 0 && ifname[0]) {
858 		err = dev_change_name(dev, ifname);
859 		if (err < 0)
860 			goto errout;
861 		modified = 1;
862 	}
863 
864 	if (tb[IFLA_IFALIAS]) {
865 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
866 				    nla_len(tb[IFLA_IFALIAS]));
867 		if (err < 0)
868 			goto errout;
869 		modified = 1;
870 	}
871 
872 	if (tb[IFLA_BROADCAST]) {
873 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
874 		send_addr_notify = 1;
875 	}
876 
877 	if (ifm->ifi_flags || ifm->ifi_change) {
878 		unsigned int flags = ifm->ifi_flags;
879 
880 		/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
881 		if (ifm->ifi_change)
882 			flags = (flags & ifm->ifi_change) |
883 				(dev->flags & ~ifm->ifi_change);
884 		err = dev_change_flags(dev, flags);
885 		if (err < 0)
886 			goto errout;
887 	}
888 
889 	if (tb[IFLA_TXQLEN])
890 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
891 
892 	if (tb[IFLA_OPERSTATE])
893 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
894 
895 	if (tb[IFLA_LINKMODE]) {
896 		write_lock_bh(&dev_base_lock);
897 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
898 		write_unlock_bh(&dev_base_lock);
899 	}
900 
901 	err = 0;
902 
903 errout:
904 	if (err < 0 && modified && net_ratelimit())
905 		printk(KERN_WARNING "A link change request failed with "
906 		       "some changes comitted already. Interface %s may "
907 		       "have been left with an inconsistent configuration, "
908 		       "please check.\n", dev->name);
909 
910 	if (send_addr_notify)
911 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
912 	return err;
913 }
914 
915 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
916 {
917 	struct net *net = sock_net(skb->sk);
918 	struct ifinfomsg *ifm;
919 	struct net_device *dev;
920 	int err;
921 	struct nlattr *tb[IFLA_MAX+1];
922 	char ifname[IFNAMSIZ];
923 
924 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
925 	if (err < 0)
926 		goto errout;
927 
928 	if (tb[IFLA_IFNAME])
929 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
930 	else
931 		ifname[0] = '\0';
932 
933 	err = -EINVAL;
934 	ifm = nlmsg_data(nlh);
935 	if (ifm->ifi_index > 0)
936 		dev = __dev_get_by_index(net, ifm->ifi_index);
937 	else if (tb[IFLA_IFNAME])
938 		dev = __dev_get_by_name(net, ifname);
939 	else
940 		goto errout;
941 
942 	if (dev == NULL) {
943 		err = -ENODEV;
944 		goto errout;
945 	}
946 
947 	err = validate_linkmsg(dev, tb);
948 	if (err < 0)
949 		goto errout;
950 
951 	err = do_setlink(dev, ifm, tb, ifname, 0);
952 errout:
953 	return err;
954 }
955 
956 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
957 {
958 	struct net *net = sock_net(skb->sk);
959 	const struct rtnl_link_ops *ops;
960 	struct net_device *dev;
961 	struct ifinfomsg *ifm;
962 	char ifname[IFNAMSIZ];
963 	struct nlattr *tb[IFLA_MAX+1];
964 	int err;
965 
966 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
967 	if (err < 0)
968 		return err;
969 
970 	if (tb[IFLA_IFNAME])
971 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
972 
973 	ifm = nlmsg_data(nlh);
974 	if (ifm->ifi_index > 0)
975 		dev = __dev_get_by_index(net, ifm->ifi_index);
976 	else if (tb[IFLA_IFNAME])
977 		dev = __dev_get_by_name(net, ifname);
978 	else
979 		return -EINVAL;
980 
981 	if (!dev)
982 		return -ENODEV;
983 
984 	ops = dev->rtnl_link_ops;
985 	if (!ops)
986 		return -EOPNOTSUPP;
987 
988 	ops->dellink(dev, NULL);
989 	return 0;
990 }
991 
992 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
993 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
994 {
995 	int err;
996 	struct net_device *dev;
997 	unsigned int num_queues = 1;
998 	unsigned int real_num_queues = 1;
999 
1000 	if (ops->get_tx_queues) {
1001 		err = ops->get_tx_queues(src_net, tb, &num_queues,
1002 					 &real_num_queues);
1003 		if (err)
1004 			goto err;
1005 	}
1006 	err = -ENOMEM;
1007 	dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1008 	if (!dev)
1009 		goto err;
1010 
1011 	dev_net_set(dev, net);
1012 	dev->rtnl_link_ops = ops;
1013 	dev->real_num_tx_queues = real_num_queues;
1014 
1015 	if (strchr(dev->name, '%')) {
1016 		err = dev_alloc_name(dev, dev->name);
1017 		if (err < 0)
1018 			goto err_free;
1019 	}
1020 
1021 	if (tb[IFLA_MTU])
1022 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1023 	if (tb[IFLA_ADDRESS])
1024 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1025 				nla_len(tb[IFLA_ADDRESS]));
1026 	if (tb[IFLA_BROADCAST])
1027 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1028 				nla_len(tb[IFLA_BROADCAST]));
1029 	if (tb[IFLA_TXQLEN])
1030 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1031 	if (tb[IFLA_OPERSTATE])
1032 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1033 	if (tb[IFLA_LINKMODE])
1034 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1035 
1036 	return dev;
1037 
1038 err_free:
1039 	free_netdev(dev);
1040 err:
1041 	return ERR_PTR(err);
1042 }
1043 EXPORT_SYMBOL(rtnl_create_link);
1044 
1045 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1046 {
1047 	struct net *net = sock_net(skb->sk);
1048 	const struct rtnl_link_ops *ops;
1049 	struct net_device *dev;
1050 	struct ifinfomsg *ifm;
1051 	char kind[MODULE_NAME_LEN];
1052 	char ifname[IFNAMSIZ];
1053 	struct nlattr *tb[IFLA_MAX+1];
1054 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1055 	int err;
1056 
1057 #ifdef CONFIG_MODULES
1058 replay:
1059 #endif
1060 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1061 	if (err < 0)
1062 		return err;
1063 
1064 	if (tb[IFLA_IFNAME])
1065 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1066 	else
1067 		ifname[0] = '\0';
1068 
1069 	ifm = nlmsg_data(nlh);
1070 	if (ifm->ifi_index > 0)
1071 		dev = __dev_get_by_index(net, ifm->ifi_index);
1072 	else if (ifname[0])
1073 		dev = __dev_get_by_name(net, ifname);
1074 	else
1075 		dev = NULL;
1076 
1077 	err = validate_linkmsg(dev, tb);
1078 	if (err < 0)
1079 		return err;
1080 
1081 	if (tb[IFLA_LINKINFO]) {
1082 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1083 				       tb[IFLA_LINKINFO], ifla_info_policy);
1084 		if (err < 0)
1085 			return err;
1086 	} else
1087 		memset(linkinfo, 0, sizeof(linkinfo));
1088 
1089 	if (linkinfo[IFLA_INFO_KIND]) {
1090 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1091 		ops = rtnl_link_ops_get(kind);
1092 	} else {
1093 		kind[0] = '\0';
1094 		ops = NULL;
1095 	}
1096 
1097 	if (1) {
1098 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1099 		struct net *dest_net;
1100 
1101 		if (ops) {
1102 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1103 				err = nla_parse_nested(attr, ops->maxtype,
1104 						       linkinfo[IFLA_INFO_DATA],
1105 						       ops->policy);
1106 				if (err < 0)
1107 					return err;
1108 				data = attr;
1109 			}
1110 			if (ops->validate) {
1111 				err = ops->validate(tb, data);
1112 				if (err < 0)
1113 					return err;
1114 			}
1115 		}
1116 
1117 		if (dev) {
1118 			int modified = 0;
1119 
1120 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1121 				return -EEXIST;
1122 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1123 				return -EOPNOTSUPP;
1124 
1125 			if (linkinfo[IFLA_INFO_DATA]) {
1126 				if (!ops || ops != dev->rtnl_link_ops ||
1127 				    !ops->changelink)
1128 					return -EOPNOTSUPP;
1129 
1130 				err = ops->changelink(dev, tb, data);
1131 				if (err < 0)
1132 					return err;
1133 				modified = 1;
1134 			}
1135 
1136 			return do_setlink(dev, ifm, tb, ifname, modified);
1137 		}
1138 
1139 		if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1140 			return -ENODEV;
1141 
1142 		if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1143 			return -EOPNOTSUPP;
1144 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1145 			return -EOPNOTSUPP;
1146 
1147 		if (!ops) {
1148 #ifdef CONFIG_MODULES
1149 			if (kind[0]) {
1150 				__rtnl_unlock();
1151 				request_module("rtnl-link-%s", kind);
1152 				rtnl_lock();
1153 				ops = rtnl_link_ops_get(kind);
1154 				if (ops)
1155 					goto replay;
1156 			}
1157 #endif
1158 			return -EOPNOTSUPP;
1159 		}
1160 
1161 		if (!ifname[0])
1162 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1163 
1164 		dest_net = rtnl_link_get_net(net, tb);
1165 		dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1166 
1167 		if (IS_ERR(dev))
1168 			err = PTR_ERR(dev);
1169 		else if (ops->newlink)
1170 			err = ops->newlink(net, dev, tb, data);
1171 		else
1172 			err = register_netdevice(dev);
1173 		if (err < 0 && !IS_ERR(dev))
1174 			free_netdev(dev);
1175 
1176 		put_net(dest_net);
1177 		return err;
1178 	}
1179 }
1180 
1181 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1182 {
1183 	struct net *net = sock_net(skb->sk);
1184 	struct ifinfomsg *ifm;
1185 	char ifname[IFNAMSIZ];
1186 	struct nlattr *tb[IFLA_MAX+1];
1187 	struct net_device *dev = NULL;
1188 	struct sk_buff *nskb;
1189 	int err;
1190 
1191 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1192 	if (err < 0)
1193 		return err;
1194 
1195 	if (tb[IFLA_IFNAME])
1196 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1197 
1198 	ifm = nlmsg_data(nlh);
1199 	if (ifm->ifi_index > 0)
1200 		dev = __dev_get_by_index(net, ifm->ifi_index);
1201 	else if (tb[IFLA_IFNAME])
1202 		dev = __dev_get_by_name(net, ifname);
1203 	else
1204 		return -EINVAL;
1205 
1206 	if (dev == NULL)
1207 		return -ENODEV;
1208 
1209 	nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1210 	if (nskb == NULL)
1211 		return -ENOBUFS;
1212 
1213 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1214 			       nlh->nlmsg_seq, 0, 0);
1215 	if (err < 0) {
1216 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1217 		WARN_ON(err == -EMSGSIZE);
1218 		kfree_skb(nskb);
1219 	} else
1220 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1221 
1222 	return err;
1223 }
1224 
1225 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1226 {
1227 	int idx;
1228 	int s_idx = cb->family;
1229 
1230 	if (s_idx == 0)
1231 		s_idx = 1;
1232 	for (idx = 1; idx < NPROTO; idx++) {
1233 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1234 		if (idx < s_idx || idx == PF_PACKET)
1235 			continue;
1236 		if (rtnl_msg_handlers[idx] == NULL ||
1237 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1238 			continue;
1239 		if (idx > s_idx)
1240 			memset(&cb->args[0], 0, sizeof(cb->args));
1241 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1242 			break;
1243 	}
1244 	cb->family = idx;
1245 
1246 	return skb->len;
1247 }
1248 
1249 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1250 {
1251 	struct net *net = dev_net(dev);
1252 	struct sk_buff *skb;
1253 	int err = -ENOBUFS;
1254 
1255 	skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1256 	if (skb == NULL)
1257 		goto errout;
1258 
1259 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1260 	if (err < 0) {
1261 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1262 		WARN_ON(err == -EMSGSIZE);
1263 		kfree_skb(skb);
1264 		goto errout;
1265 	}
1266 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1267 	return;
1268 errout:
1269 	if (err < 0)
1270 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1271 }
1272 
1273 /* Protected by RTNL sempahore.  */
1274 static struct rtattr **rta_buf;
1275 static int rtattr_max;
1276 
1277 /* Process one rtnetlink message. */
1278 
1279 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1280 {
1281 	struct net *net = sock_net(skb->sk);
1282 	rtnl_doit_func doit;
1283 	int sz_idx, kind;
1284 	int min_len;
1285 	int family;
1286 	int type;
1287 	int err;
1288 
1289 	type = nlh->nlmsg_type;
1290 	if (type > RTM_MAX)
1291 		return -EOPNOTSUPP;
1292 
1293 	type -= RTM_BASE;
1294 
1295 	/* All the messages must have at least 1 byte length */
1296 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1297 		return 0;
1298 
1299 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1300 	if (family >= NPROTO)
1301 		return -EAFNOSUPPORT;
1302 
1303 	sz_idx = type>>2;
1304 	kind = type&3;
1305 
1306 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1307 		return -EPERM;
1308 
1309 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1310 		struct sock *rtnl;
1311 		rtnl_dumpit_func dumpit;
1312 
1313 		dumpit = rtnl_get_dumpit(family, type);
1314 		if (dumpit == NULL)
1315 			return -EOPNOTSUPP;
1316 
1317 		__rtnl_unlock();
1318 		rtnl = net->rtnl;
1319 		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1320 		rtnl_lock();
1321 		return err;
1322 	}
1323 
1324 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1325 
1326 	min_len = rtm_min[sz_idx];
1327 	if (nlh->nlmsg_len < min_len)
1328 		return -EINVAL;
1329 
1330 	if (nlh->nlmsg_len > min_len) {
1331 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1332 		struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1333 
1334 		while (RTA_OK(attr, attrlen)) {
1335 			unsigned flavor = attr->rta_type;
1336 			if (flavor) {
1337 				if (flavor > rta_max[sz_idx])
1338 					return -EINVAL;
1339 				rta_buf[flavor-1] = attr;
1340 			}
1341 			attr = RTA_NEXT(attr, attrlen);
1342 		}
1343 	}
1344 
1345 	doit = rtnl_get_doit(family, type);
1346 	if (doit == NULL)
1347 		return -EOPNOTSUPP;
1348 
1349 	return doit(skb, nlh, (void *)&rta_buf[0]);
1350 }
1351 
1352 static void rtnetlink_rcv(struct sk_buff *skb)
1353 {
1354 	rtnl_lock();
1355 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1356 	rtnl_unlock();
1357 }
1358 
1359 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1360 {
1361 	struct net_device *dev = ptr;
1362 
1363 	switch (event) {
1364 	case NETDEV_UNREGISTER:
1365 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1366 		break;
1367 	case NETDEV_UP:
1368 	case NETDEV_DOWN:
1369 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1370 		break;
1371 	case NETDEV_POST_INIT:
1372 	case NETDEV_REGISTER:
1373 	case NETDEV_CHANGE:
1374 	case NETDEV_GOING_DOWN:
1375 	case NETDEV_UNREGISTER_BATCH:
1376 		break;
1377 	default:
1378 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1379 		break;
1380 	}
1381 	return NOTIFY_DONE;
1382 }
1383 
1384 static struct notifier_block rtnetlink_dev_notifier = {
1385 	.notifier_call	= rtnetlink_event,
1386 };
1387 
1388 
1389 static int rtnetlink_net_init(struct net *net)
1390 {
1391 	struct sock *sk;
1392 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1393 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1394 	if (!sk)
1395 		return -ENOMEM;
1396 	net->rtnl = sk;
1397 	return 0;
1398 }
1399 
1400 static void rtnetlink_net_exit(struct net *net)
1401 {
1402 	netlink_kernel_release(net->rtnl);
1403 	net->rtnl = NULL;
1404 }
1405 
1406 static struct pernet_operations rtnetlink_net_ops = {
1407 	.init = rtnetlink_net_init,
1408 	.exit = rtnetlink_net_exit,
1409 };
1410 
1411 void __init rtnetlink_init(void)
1412 {
1413 	int i;
1414 
1415 	rtattr_max = 0;
1416 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1417 		if (rta_max[i] > rtattr_max)
1418 			rtattr_max = rta_max[i];
1419 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1420 	if (!rta_buf)
1421 		panic("rtnetlink_init: cannot allocate rta_buf\n");
1422 
1423 	if (register_pernet_subsys(&rtnetlink_net_ops))
1424 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
1425 
1426 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1427 	register_netdevice_notifier(&rtnetlink_dev_notifier);
1428 
1429 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1430 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1431 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1432 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1433 
1434 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1435 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1436 }
1437 
1438