xref: /openbmc/linux/net/core/rtnetlink.c (revision a1e58bbd)
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/nsproxy.h>
39 
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
42 #include <asm/string.h>
43 
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <net/arp.h>
49 #include <net/route.h>
50 #include <net/udp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 
56 struct rtnl_link
57 {
58 	rtnl_doit_func		doit;
59 	rtnl_dumpit_func	dumpit;
60 };
61 
62 static DEFINE_MUTEX(rtnl_mutex);
63 
64 void rtnl_lock(void)
65 {
66 	mutex_lock(&rtnl_mutex);
67 }
68 
69 void __rtnl_unlock(void)
70 {
71 	mutex_unlock(&rtnl_mutex);
72 }
73 
74 void rtnl_unlock(void)
75 {
76 	mutex_unlock(&rtnl_mutex);
77 	netdev_run_todo();
78 }
79 
80 int rtnl_trylock(void)
81 {
82 	return mutex_trylock(&rtnl_mutex);
83 }
84 
85 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
86 
87 static inline int rtm_msgindex(int msgtype)
88 {
89 	int msgindex = msgtype - RTM_BASE;
90 
91 	/*
92 	 * msgindex < 0 implies someone tried to register a netlink
93 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
94 	 * the message type has not been added to linux/rtnetlink.h
95 	 */
96 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
97 
98 	return msgindex;
99 }
100 
101 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
102 {
103 	struct rtnl_link *tab;
104 
105 	tab = rtnl_msg_handlers[protocol];
106 	if (tab == NULL || tab[msgindex].doit == NULL)
107 		tab = rtnl_msg_handlers[PF_UNSPEC];
108 
109 	return tab ? tab[msgindex].doit : NULL;
110 }
111 
112 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
113 {
114 	struct rtnl_link *tab;
115 
116 	tab = rtnl_msg_handlers[protocol];
117 	if (tab == NULL || tab[msgindex].dumpit == NULL)
118 		tab = rtnl_msg_handlers[PF_UNSPEC];
119 
120 	return tab ? tab[msgindex].dumpit : NULL;
121 }
122 
123 /**
124  * __rtnl_register - Register a rtnetlink message type
125  * @protocol: Protocol family or PF_UNSPEC
126  * @msgtype: rtnetlink message type
127  * @doit: Function pointer called for each request message
128  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
129  *
130  * Registers the specified function pointers (at least one of them has
131  * to be non-NULL) to be called whenever a request message for the
132  * specified protocol family and message type is received.
133  *
134  * The special protocol family PF_UNSPEC may be used to define fallback
135  * function pointers for the case when no entry for the specific protocol
136  * family exists.
137  *
138  * Returns 0 on success or a negative error code.
139  */
140 int __rtnl_register(int protocol, int msgtype,
141 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
142 {
143 	struct rtnl_link *tab;
144 	int msgindex;
145 
146 	BUG_ON(protocol < 0 || protocol >= NPROTO);
147 	msgindex = rtm_msgindex(msgtype);
148 
149 	tab = rtnl_msg_handlers[protocol];
150 	if (tab == NULL) {
151 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
152 		if (tab == NULL)
153 			return -ENOBUFS;
154 
155 		rtnl_msg_handlers[protocol] = tab;
156 	}
157 
158 	if (doit)
159 		tab[msgindex].doit = doit;
160 
161 	if (dumpit)
162 		tab[msgindex].dumpit = dumpit;
163 
164 	return 0;
165 }
166 
167 EXPORT_SYMBOL_GPL(__rtnl_register);
168 
169 /**
170  * rtnl_register - Register a rtnetlink message type
171  *
172  * Identical to __rtnl_register() but panics on failure. This is useful
173  * as failure of this function is very unlikely, it can only happen due
174  * to lack of memory when allocating the chain to store all message
175  * handlers for a protocol. Meant for use in init functions where lack
176  * of memory implies no sense in continueing.
177  */
178 void rtnl_register(int protocol, int msgtype,
179 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit)
180 {
181 	if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
182 		panic("Unable to register rtnetlink message handler, "
183 		      "protocol = %d, message type = %d\n",
184 		      protocol, msgtype);
185 }
186 
187 EXPORT_SYMBOL_GPL(rtnl_register);
188 
189 /**
190  * rtnl_unregister - Unregister a rtnetlink message type
191  * @protocol: Protocol family or PF_UNSPEC
192  * @msgtype: rtnetlink message type
193  *
194  * Returns 0 on success or a negative error code.
195  */
196 int rtnl_unregister(int protocol, int msgtype)
197 {
198 	int msgindex;
199 
200 	BUG_ON(protocol < 0 || protocol >= NPROTO);
201 	msgindex = rtm_msgindex(msgtype);
202 
203 	if (rtnl_msg_handlers[protocol] == NULL)
204 		return -ENOENT;
205 
206 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
207 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
208 
209 	return 0;
210 }
211 
212 EXPORT_SYMBOL_GPL(rtnl_unregister);
213 
214 /**
215  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
216  * @protocol : Protocol family or PF_UNSPEC
217  *
218  * Identical to calling rtnl_unregster() for all registered message types
219  * of a certain protocol family.
220  */
221 void rtnl_unregister_all(int protocol)
222 {
223 	BUG_ON(protocol < 0 || protocol >= NPROTO);
224 
225 	kfree(rtnl_msg_handlers[protocol]);
226 	rtnl_msg_handlers[protocol] = NULL;
227 }
228 
229 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
230 
231 static LIST_HEAD(link_ops);
232 
233 /**
234  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
235  * @ops: struct rtnl_link_ops * to register
236  *
237  * The caller must hold the rtnl_mutex. This function should be used
238  * by drivers that create devices during module initialization. It
239  * must be called before registering the devices.
240  *
241  * Returns 0 on success or a negative error code.
242  */
243 int __rtnl_link_register(struct rtnl_link_ops *ops)
244 {
245 	if (!ops->dellink)
246 		ops->dellink = unregister_netdevice;
247 
248 	list_add_tail(&ops->list, &link_ops);
249 	return 0;
250 }
251 
252 EXPORT_SYMBOL_GPL(__rtnl_link_register);
253 
254 /**
255  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
256  * @ops: struct rtnl_link_ops * to register
257  *
258  * Returns 0 on success or a negative error code.
259  */
260 int rtnl_link_register(struct rtnl_link_ops *ops)
261 {
262 	int err;
263 
264 	rtnl_lock();
265 	err = __rtnl_link_register(ops);
266 	rtnl_unlock();
267 	return err;
268 }
269 
270 EXPORT_SYMBOL_GPL(rtnl_link_register);
271 
272 /**
273  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
274  * @ops: struct rtnl_link_ops * to unregister
275  *
276  * The caller must hold the rtnl_mutex.
277  */
278 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
279 {
280 	struct net_device *dev, *n;
281 	struct net *net;
282 
283 	for_each_net(net) {
284 restart:
285 		for_each_netdev_safe(net, dev, n) {
286 			if (dev->rtnl_link_ops == ops) {
287 				ops->dellink(dev);
288 				goto restart;
289 			}
290 		}
291 	}
292 	list_del(&ops->list);
293 }
294 
295 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
296 
297 /**
298  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
299  * @ops: struct rtnl_link_ops * to unregister
300  */
301 void rtnl_link_unregister(struct rtnl_link_ops *ops)
302 {
303 	rtnl_lock();
304 	__rtnl_link_unregister(ops);
305 	rtnl_unlock();
306 }
307 
308 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
309 
310 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
311 {
312 	const struct rtnl_link_ops *ops;
313 
314 	list_for_each_entry(ops, &link_ops, list) {
315 		if (!strcmp(ops->kind, kind))
316 			return ops;
317 	}
318 	return NULL;
319 }
320 
321 static size_t rtnl_link_get_size(const struct net_device *dev)
322 {
323 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
324 	size_t size;
325 
326 	if (!ops)
327 		return 0;
328 
329 	size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
330 	       nlmsg_total_size(strlen(ops->kind) + 1);	 /* IFLA_INFO_KIND */
331 
332 	if (ops->get_size)
333 		/* IFLA_INFO_DATA + nested data */
334 		size += nlmsg_total_size(sizeof(struct nlattr)) +
335 			ops->get_size(dev);
336 
337 	if (ops->get_xstats_size)
338 		size += ops->get_xstats_size(dev);	/* IFLA_INFO_XSTATS */
339 
340 	return size;
341 }
342 
343 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
344 {
345 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
346 	struct nlattr *linkinfo, *data;
347 	int err = -EMSGSIZE;
348 
349 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
350 	if (linkinfo == NULL)
351 		goto out;
352 
353 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
354 		goto err_cancel_link;
355 	if (ops->fill_xstats) {
356 		err = ops->fill_xstats(skb, dev);
357 		if (err < 0)
358 			goto err_cancel_link;
359 	}
360 	if (ops->fill_info) {
361 		data = nla_nest_start(skb, IFLA_INFO_DATA);
362 		if (data == NULL)
363 			goto err_cancel_link;
364 		err = ops->fill_info(skb, dev);
365 		if (err < 0)
366 			goto err_cancel_data;
367 		nla_nest_end(skb, data);
368 	}
369 
370 	nla_nest_end(skb, linkinfo);
371 	return 0;
372 
373 err_cancel_data:
374 	nla_nest_cancel(skb, data);
375 err_cancel_link:
376 	nla_nest_cancel(skb, linkinfo);
377 out:
378 	return err;
379 }
380 
381 static const int rtm_min[RTM_NR_FAMILIES] =
382 {
383 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
384 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
385 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
386 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
387 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
388 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
389 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
390 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
391 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
392 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
393 };
394 
395 static const int rta_max[RTM_NR_FAMILIES] =
396 {
397 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
398 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
399 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
400 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
401 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
402 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
403 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
404 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
405 };
406 
407 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
408 {
409 	struct rtattr *rta;
410 	int size = RTA_LENGTH(attrlen);
411 
412 	rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
413 	rta->rta_type = attrtype;
414 	rta->rta_len = size;
415 	memcpy(RTA_DATA(rta), data, attrlen);
416 	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
417 }
418 
419 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
420 {
421 	struct sock *rtnl = net->rtnl;
422 	int err = 0;
423 
424 	NETLINK_CB(skb).dst_group = group;
425 	if (echo)
426 		atomic_inc(&skb->users);
427 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
428 	if (echo)
429 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
430 	return err;
431 }
432 
433 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
434 {
435 	struct sock *rtnl = net->rtnl;
436 
437 	return nlmsg_unicast(rtnl, skb, pid);
438 }
439 
440 int rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
441 		struct nlmsghdr *nlh, gfp_t flags)
442 {
443 	struct sock *rtnl = net->rtnl;
444 	int report = 0;
445 
446 	if (nlh)
447 		report = nlmsg_report(nlh);
448 
449 	return nlmsg_notify(rtnl, skb, pid, group, report, flags);
450 }
451 
452 void rtnl_set_sk_err(struct net *net, u32 group, int error)
453 {
454 	struct sock *rtnl = net->rtnl;
455 
456 	netlink_set_err(rtnl, 0, group, error);
457 }
458 
459 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
460 {
461 	struct nlattr *mx;
462 	int i, valid = 0;
463 
464 	mx = nla_nest_start(skb, RTA_METRICS);
465 	if (mx == NULL)
466 		return -ENOBUFS;
467 
468 	for (i = 0; i < RTAX_MAX; i++) {
469 		if (metrics[i]) {
470 			valid++;
471 			NLA_PUT_U32(skb, i+1, metrics[i]);
472 		}
473 	}
474 
475 	if (!valid) {
476 		nla_nest_cancel(skb, mx);
477 		return 0;
478 	}
479 
480 	return nla_nest_end(skb, mx);
481 
482 nla_put_failure:
483 	return nla_nest_cancel(skb, mx);
484 }
485 
486 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
487 		       u32 ts, u32 tsage, long expires, u32 error)
488 {
489 	struct rta_cacheinfo ci = {
490 		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
491 		.rta_used = dst->__use,
492 		.rta_clntref = atomic_read(&(dst->__refcnt)),
493 		.rta_error = error,
494 		.rta_id =  id,
495 		.rta_ts = ts,
496 		.rta_tsage = tsage,
497 	};
498 
499 	if (expires)
500 		ci.rta_expires = jiffies_to_clock_t(expires);
501 
502 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
503 }
504 
505 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
506 
507 static void set_operstate(struct net_device *dev, unsigned char transition)
508 {
509 	unsigned char operstate = dev->operstate;
510 
511 	switch(transition) {
512 	case IF_OPER_UP:
513 		if ((operstate == IF_OPER_DORMANT ||
514 		     operstate == IF_OPER_UNKNOWN) &&
515 		    !netif_dormant(dev))
516 			operstate = IF_OPER_UP;
517 		break;
518 
519 	case IF_OPER_DORMANT:
520 		if (operstate == IF_OPER_UP ||
521 		    operstate == IF_OPER_UNKNOWN)
522 			operstate = IF_OPER_DORMANT;
523 		break;
524 	}
525 
526 	if (dev->operstate != operstate) {
527 		write_lock_bh(&dev_base_lock);
528 		dev->operstate = operstate;
529 		write_unlock_bh(&dev_base_lock);
530 		netdev_state_change(dev);
531 	}
532 }
533 
534 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
535 				 struct net_device_stats *b)
536 {
537 	a->rx_packets = b->rx_packets;
538 	a->tx_packets = b->tx_packets;
539 	a->rx_bytes = b->rx_bytes;
540 	a->tx_bytes = b->tx_bytes;
541 	a->rx_errors = b->rx_errors;
542 	a->tx_errors = b->tx_errors;
543 	a->rx_dropped = b->rx_dropped;
544 	a->tx_dropped = b->tx_dropped;
545 
546 	a->multicast = b->multicast;
547 	a->collisions = b->collisions;
548 
549 	a->rx_length_errors = b->rx_length_errors;
550 	a->rx_over_errors = b->rx_over_errors;
551 	a->rx_crc_errors = b->rx_crc_errors;
552 	a->rx_frame_errors = b->rx_frame_errors;
553 	a->rx_fifo_errors = b->rx_fifo_errors;
554 	a->rx_missed_errors = b->rx_missed_errors;
555 
556 	a->tx_aborted_errors = b->tx_aborted_errors;
557 	a->tx_carrier_errors = b->tx_carrier_errors;
558 	a->tx_fifo_errors = b->tx_fifo_errors;
559 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
560 	a->tx_window_errors = b->tx_window_errors;
561 
562 	a->rx_compressed = b->rx_compressed;
563 	a->tx_compressed = b->tx_compressed;
564 };
565 
566 static inline size_t if_nlmsg_size(const struct net_device *dev)
567 {
568 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
569 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
570 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
571 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
572 	       + nla_total_size(sizeof(struct rtnl_link_stats))
573 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
574 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
575 	       + nla_total_size(4) /* IFLA_TXQLEN */
576 	       + nla_total_size(4) /* IFLA_WEIGHT */
577 	       + nla_total_size(4) /* IFLA_MTU */
578 	       + nla_total_size(4) /* IFLA_LINK */
579 	       + nla_total_size(4) /* IFLA_MASTER */
580 	       + nla_total_size(1) /* IFLA_OPERSTATE */
581 	       + nla_total_size(1) /* IFLA_LINKMODE */
582 	       + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
583 }
584 
585 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
586 			    int type, u32 pid, u32 seq, u32 change,
587 			    unsigned int flags)
588 {
589 	struct ifinfomsg *ifm;
590 	struct nlmsghdr *nlh;
591 
592 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
593 	if (nlh == NULL)
594 		return -EMSGSIZE;
595 
596 	ifm = nlmsg_data(nlh);
597 	ifm->ifi_family = AF_UNSPEC;
598 	ifm->__ifi_pad = 0;
599 	ifm->ifi_type = dev->type;
600 	ifm->ifi_index = dev->ifindex;
601 	ifm->ifi_flags = dev_get_flags(dev);
602 	ifm->ifi_change = change;
603 
604 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
605 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
606 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
607 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
608 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
609 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
610 
611 	if (dev->ifindex != dev->iflink)
612 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
613 
614 	if (dev->master)
615 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
616 
617 	if (dev->qdisc_sleeping)
618 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
619 
620 	if (1) {
621 		struct rtnl_link_ifmap map = {
622 			.mem_start   = dev->mem_start,
623 			.mem_end     = dev->mem_end,
624 			.base_addr   = dev->base_addr,
625 			.irq         = dev->irq,
626 			.dma         = dev->dma,
627 			.port        = dev->if_port,
628 		};
629 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
630 	}
631 
632 	if (dev->addr_len) {
633 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
634 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
635 	}
636 
637 	if (dev->get_stats) {
638 		struct net_device_stats *stats = dev->get_stats(dev);
639 		if (stats) {
640 			struct nlattr *attr;
641 
642 			attr = nla_reserve(skb, IFLA_STATS,
643 					   sizeof(struct rtnl_link_stats));
644 			if (attr == NULL)
645 				goto nla_put_failure;
646 
647 			copy_rtnl_link_stats(nla_data(attr), stats);
648 		}
649 	}
650 
651 	if (dev->rtnl_link_ops) {
652 		if (rtnl_link_fill(skb, dev) < 0)
653 			goto nla_put_failure;
654 	}
655 
656 	return nlmsg_end(skb, nlh);
657 
658 nla_put_failure:
659 	nlmsg_cancel(skb, nlh);
660 	return -EMSGSIZE;
661 }
662 
663 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
664 {
665 	struct net *net = skb->sk->sk_net;
666 	int idx;
667 	int s_idx = cb->args[0];
668 	struct net_device *dev;
669 
670 	idx = 0;
671 	for_each_netdev(net, dev) {
672 		if (idx < s_idx)
673 			goto cont;
674 		if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
675 				     NETLINK_CB(cb->skb).pid,
676 				     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
677 			break;
678 cont:
679 		idx++;
680 	}
681 	cb->args[0] = idx;
682 
683 	return skb->len;
684 }
685 
686 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
687 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
688 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
689 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
690 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
691 	[IFLA_MTU]		= { .type = NLA_U32 },
692 	[IFLA_LINK]		= { .type = NLA_U32 },
693 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
694 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
695 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
696 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
697 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
698 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
699 };
700 
701 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
702 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
703 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
704 };
705 
706 static struct net *get_net_ns_by_pid(pid_t pid)
707 {
708 	struct task_struct *tsk;
709 	struct net *net;
710 
711 	/* Lookup the network namespace */
712 	net = ERR_PTR(-ESRCH);
713 	rcu_read_lock();
714 	tsk = find_task_by_vpid(pid);
715 	if (tsk) {
716 		struct nsproxy *nsproxy;
717 		nsproxy = task_nsproxy(tsk);
718 		if (nsproxy)
719 			net = get_net(nsproxy->net_ns);
720 	}
721 	rcu_read_unlock();
722 	return net;
723 }
724 
725 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
726 {
727 	if (dev) {
728 		if (tb[IFLA_ADDRESS] &&
729 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
730 			return -EINVAL;
731 
732 		if (tb[IFLA_BROADCAST] &&
733 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
734 			return -EINVAL;
735 	}
736 
737 	return 0;
738 }
739 
740 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
741 		      struct nlattr **tb, char *ifname, int modified)
742 {
743 	int send_addr_notify = 0;
744 	int err;
745 
746 	if (tb[IFLA_NET_NS_PID]) {
747 		struct net *net;
748 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
749 		if (IS_ERR(net)) {
750 			err = PTR_ERR(net);
751 			goto errout;
752 		}
753 		err = dev_change_net_namespace(dev, net, ifname);
754 		put_net(net);
755 		if (err)
756 			goto errout;
757 		modified = 1;
758 	}
759 
760 	if (tb[IFLA_MAP]) {
761 		struct rtnl_link_ifmap *u_map;
762 		struct ifmap k_map;
763 
764 		if (!dev->set_config) {
765 			err = -EOPNOTSUPP;
766 			goto errout;
767 		}
768 
769 		if (!netif_device_present(dev)) {
770 			err = -ENODEV;
771 			goto errout;
772 		}
773 
774 		u_map = nla_data(tb[IFLA_MAP]);
775 		k_map.mem_start = (unsigned long) u_map->mem_start;
776 		k_map.mem_end = (unsigned long) u_map->mem_end;
777 		k_map.base_addr = (unsigned short) u_map->base_addr;
778 		k_map.irq = (unsigned char) u_map->irq;
779 		k_map.dma = (unsigned char) u_map->dma;
780 		k_map.port = (unsigned char) u_map->port;
781 
782 		err = dev->set_config(dev, &k_map);
783 		if (err < 0)
784 			goto errout;
785 
786 		modified = 1;
787 	}
788 
789 	if (tb[IFLA_ADDRESS]) {
790 		struct sockaddr *sa;
791 		int len;
792 
793 		if (!dev->set_mac_address) {
794 			err = -EOPNOTSUPP;
795 			goto errout;
796 		}
797 
798 		if (!netif_device_present(dev)) {
799 			err = -ENODEV;
800 			goto errout;
801 		}
802 
803 		len = sizeof(sa_family_t) + dev->addr_len;
804 		sa = kmalloc(len, GFP_KERNEL);
805 		if (!sa) {
806 			err = -ENOMEM;
807 			goto errout;
808 		}
809 		sa->sa_family = dev->type;
810 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
811 		       dev->addr_len);
812 		err = dev->set_mac_address(dev, sa);
813 		kfree(sa);
814 		if (err)
815 			goto errout;
816 		send_addr_notify = 1;
817 		modified = 1;
818 	}
819 
820 	if (tb[IFLA_MTU]) {
821 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
822 		if (err < 0)
823 			goto errout;
824 		modified = 1;
825 	}
826 
827 	/*
828 	 * Interface selected by interface index but interface
829 	 * name provided implies that a name change has been
830 	 * requested.
831 	 */
832 	if (ifm->ifi_index > 0 && ifname[0]) {
833 		err = dev_change_name(dev, ifname);
834 		if (err < 0)
835 			goto errout;
836 		modified = 1;
837 	}
838 
839 	if (tb[IFLA_BROADCAST]) {
840 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
841 		send_addr_notify = 1;
842 	}
843 
844 	if (ifm->ifi_flags || ifm->ifi_change) {
845 		unsigned int flags = ifm->ifi_flags;
846 
847 		/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
848 		if (ifm->ifi_change)
849 			flags = (flags & ifm->ifi_change) |
850 				(dev->flags & ~ifm->ifi_change);
851 		dev_change_flags(dev, flags);
852 	}
853 
854 	if (tb[IFLA_TXQLEN])
855 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
856 
857 	if (tb[IFLA_OPERSTATE])
858 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
859 
860 	if (tb[IFLA_LINKMODE]) {
861 		write_lock_bh(&dev_base_lock);
862 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
863 		write_unlock_bh(&dev_base_lock);
864 	}
865 
866 	err = 0;
867 
868 errout:
869 	if (err < 0 && modified && net_ratelimit())
870 		printk(KERN_WARNING "A link change request failed with "
871 		       "some changes comitted already. Interface %s may "
872 		       "have been left with an inconsistent configuration, "
873 		       "please check.\n", dev->name);
874 
875 	if (send_addr_notify)
876 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
877 	return err;
878 }
879 
880 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
881 {
882 	struct net *net = skb->sk->sk_net;
883 	struct ifinfomsg *ifm;
884 	struct net_device *dev;
885 	int err;
886 	struct nlattr *tb[IFLA_MAX+1];
887 	char ifname[IFNAMSIZ];
888 
889 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
890 	if (err < 0)
891 		goto errout;
892 
893 	if (tb[IFLA_IFNAME])
894 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
895 	else
896 		ifname[0] = '\0';
897 
898 	err = -EINVAL;
899 	ifm = nlmsg_data(nlh);
900 	if (ifm->ifi_index > 0)
901 		dev = dev_get_by_index(net, ifm->ifi_index);
902 	else if (tb[IFLA_IFNAME])
903 		dev = dev_get_by_name(net, ifname);
904 	else
905 		goto errout;
906 
907 	if (dev == NULL) {
908 		err = -ENODEV;
909 		goto errout;
910 	}
911 
912 	if ((err = validate_linkmsg(dev, tb)) < 0)
913 		goto errout_dev;
914 
915 	err = do_setlink(dev, ifm, tb, ifname, 0);
916 errout_dev:
917 	dev_put(dev);
918 errout:
919 	return err;
920 }
921 
922 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
923 {
924 	struct net *net = skb->sk->sk_net;
925 	const struct rtnl_link_ops *ops;
926 	struct net_device *dev;
927 	struct ifinfomsg *ifm;
928 	char ifname[IFNAMSIZ];
929 	struct nlattr *tb[IFLA_MAX+1];
930 	int err;
931 
932 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
933 	if (err < 0)
934 		return err;
935 
936 	if (tb[IFLA_IFNAME])
937 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
938 
939 	ifm = nlmsg_data(nlh);
940 	if (ifm->ifi_index > 0)
941 		dev = __dev_get_by_index(net, ifm->ifi_index);
942 	else if (tb[IFLA_IFNAME])
943 		dev = __dev_get_by_name(net, ifname);
944 	else
945 		return -EINVAL;
946 
947 	if (!dev)
948 		return -ENODEV;
949 
950 	ops = dev->rtnl_link_ops;
951 	if (!ops)
952 		return -EOPNOTSUPP;
953 
954 	ops->dellink(dev);
955 	return 0;
956 }
957 
958 struct net_device *rtnl_create_link(struct net *net, char *ifname,
959 		const struct rtnl_link_ops *ops, struct nlattr *tb[])
960 {
961 	int err;
962 	struct net_device *dev;
963 
964 	err = -ENOMEM;
965 	dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
966 	if (!dev)
967 		goto err;
968 
969 	if (strchr(dev->name, '%')) {
970 		err = dev_alloc_name(dev, dev->name);
971 		if (err < 0)
972 			goto err_free;
973 	}
974 
975 	dev->nd_net = net;
976 	dev->rtnl_link_ops = ops;
977 
978 	if (tb[IFLA_MTU])
979 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
980 	if (tb[IFLA_ADDRESS])
981 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
982 				nla_len(tb[IFLA_ADDRESS]));
983 	if (tb[IFLA_BROADCAST])
984 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
985 				nla_len(tb[IFLA_BROADCAST]));
986 	if (tb[IFLA_TXQLEN])
987 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
988 	if (tb[IFLA_OPERSTATE])
989 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
990 	if (tb[IFLA_LINKMODE])
991 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
992 
993 	return dev;
994 
995 err_free:
996 	free_netdev(dev);
997 err:
998 	return ERR_PTR(err);
999 }
1000 
1001 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1002 {
1003 	struct net *net = skb->sk->sk_net;
1004 	const struct rtnl_link_ops *ops;
1005 	struct net_device *dev;
1006 	struct ifinfomsg *ifm;
1007 	char kind[MODULE_NAME_LEN];
1008 	char ifname[IFNAMSIZ];
1009 	struct nlattr *tb[IFLA_MAX+1];
1010 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1011 	int err;
1012 
1013 #ifdef CONFIG_KMOD
1014 replay:
1015 #endif
1016 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1017 	if (err < 0)
1018 		return err;
1019 
1020 	if (tb[IFLA_IFNAME])
1021 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1022 	else
1023 		ifname[0] = '\0';
1024 
1025 	ifm = nlmsg_data(nlh);
1026 	if (ifm->ifi_index > 0)
1027 		dev = __dev_get_by_index(net, ifm->ifi_index);
1028 	else if (ifname[0])
1029 		dev = __dev_get_by_name(net, ifname);
1030 	else
1031 		dev = NULL;
1032 
1033 	if ((err = validate_linkmsg(dev, tb)) < 0)
1034 		return err;
1035 
1036 	if (tb[IFLA_LINKINFO]) {
1037 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1038 				       tb[IFLA_LINKINFO], ifla_info_policy);
1039 		if (err < 0)
1040 			return err;
1041 	} else
1042 		memset(linkinfo, 0, sizeof(linkinfo));
1043 
1044 	if (linkinfo[IFLA_INFO_KIND]) {
1045 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1046 		ops = rtnl_link_ops_get(kind);
1047 	} else {
1048 		kind[0] = '\0';
1049 		ops = NULL;
1050 	}
1051 
1052 	if (1) {
1053 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1054 
1055 		if (ops) {
1056 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1057 				err = nla_parse_nested(attr, ops->maxtype,
1058 						       linkinfo[IFLA_INFO_DATA],
1059 						       ops->policy);
1060 				if (err < 0)
1061 					return err;
1062 				data = attr;
1063 			}
1064 			if (ops->validate) {
1065 				err = ops->validate(tb, data);
1066 				if (err < 0)
1067 					return err;
1068 			}
1069 		}
1070 
1071 		if (dev) {
1072 			int modified = 0;
1073 
1074 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1075 				return -EEXIST;
1076 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1077 				return -EOPNOTSUPP;
1078 
1079 			if (linkinfo[IFLA_INFO_DATA]) {
1080 				if (!ops || ops != dev->rtnl_link_ops ||
1081 				    !ops->changelink)
1082 					return -EOPNOTSUPP;
1083 
1084 				err = ops->changelink(dev, tb, data);
1085 				if (err < 0)
1086 					return err;
1087 				modified = 1;
1088 			}
1089 
1090 			return do_setlink(dev, ifm, tb, ifname, modified);
1091 		}
1092 
1093 		if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1094 			return -ENODEV;
1095 
1096 		if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1097 			return -EOPNOTSUPP;
1098 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1099 			return -EOPNOTSUPP;
1100 
1101 		if (!ops) {
1102 #ifdef CONFIG_KMOD
1103 			if (kind[0]) {
1104 				__rtnl_unlock();
1105 				request_module("rtnl-link-%s", kind);
1106 				rtnl_lock();
1107 				ops = rtnl_link_ops_get(kind);
1108 				if (ops)
1109 					goto replay;
1110 			}
1111 #endif
1112 			return -EOPNOTSUPP;
1113 		}
1114 
1115 		if (!ifname[0])
1116 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1117 
1118 		dev = rtnl_create_link(net, ifname, ops, tb);
1119 
1120 		if (IS_ERR(dev))
1121 			err = PTR_ERR(dev);
1122 		else if (ops->newlink)
1123 			err = ops->newlink(dev, tb, data);
1124 		else
1125 			err = register_netdevice(dev);
1126 
1127 		if (err < 0 && !IS_ERR(dev))
1128 			free_netdev(dev);
1129 		return err;
1130 	}
1131 }
1132 
1133 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1134 {
1135 	struct net *net = skb->sk->sk_net;
1136 	struct ifinfomsg *ifm;
1137 	struct nlattr *tb[IFLA_MAX+1];
1138 	struct net_device *dev = NULL;
1139 	struct sk_buff *nskb;
1140 	int err;
1141 
1142 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1143 	if (err < 0)
1144 		return err;
1145 
1146 	ifm = nlmsg_data(nlh);
1147 	if (ifm->ifi_index > 0) {
1148 		dev = dev_get_by_index(net, ifm->ifi_index);
1149 		if (dev == NULL)
1150 			return -ENODEV;
1151 	} else
1152 		return -EINVAL;
1153 
1154 	nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1155 	if (nskb == NULL) {
1156 		err = -ENOBUFS;
1157 		goto errout;
1158 	}
1159 
1160 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1161 			       nlh->nlmsg_seq, 0, 0);
1162 	if (err < 0) {
1163 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1164 		WARN_ON(err == -EMSGSIZE);
1165 		kfree_skb(nskb);
1166 		goto errout;
1167 	}
1168 	err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1169 errout:
1170 	dev_put(dev);
1171 
1172 	return err;
1173 }
1174 
1175 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1176 {
1177 	int idx;
1178 	int s_idx = cb->family;
1179 
1180 	if (s_idx == 0)
1181 		s_idx = 1;
1182 	for (idx=1; idx<NPROTO; idx++) {
1183 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1184 		if (idx < s_idx || idx == PF_PACKET)
1185 			continue;
1186 		if (rtnl_msg_handlers[idx] == NULL ||
1187 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1188 			continue;
1189 		if (idx > s_idx)
1190 			memset(&cb->args[0], 0, sizeof(cb->args));
1191 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1192 			break;
1193 	}
1194 	cb->family = idx;
1195 
1196 	return skb->len;
1197 }
1198 
1199 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1200 {
1201 	struct net *net = dev->nd_net;
1202 	struct sk_buff *skb;
1203 	int err = -ENOBUFS;
1204 
1205 	skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1206 	if (skb == NULL)
1207 		goto errout;
1208 
1209 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1210 	if (err < 0) {
1211 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1212 		WARN_ON(err == -EMSGSIZE);
1213 		kfree_skb(skb);
1214 		goto errout;
1215 	}
1216 	err = rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1217 errout:
1218 	if (err < 0)
1219 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1220 }
1221 
1222 /* Protected by RTNL sempahore.  */
1223 static struct rtattr **rta_buf;
1224 static int rtattr_max;
1225 
1226 /* Process one rtnetlink message. */
1227 
1228 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1229 {
1230 	struct net *net = skb->sk->sk_net;
1231 	rtnl_doit_func doit;
1232 	int sz_idx, kind;
1233 	int min_len;
1234 	int family;
1235 	int type;
1236 	int err;
1237 
1238 	type = nlh->nlmsg_type;
1239 	if (type > RTM_MAX)
1240 		return -EOPNOTSUPP;
1241 
1242 	type -= RTM_BASE;
1243 
1244 	/* All the messages must have at least 1 byte length */
1245 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1246 		return 0;
1247 
1248 	family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1249 	if (family >= NPROTO)
1250 		return -EAFNOSUPPORT;
1251 
1252 	sz_idx = type>>2;
1253 	kind = type&3;
1254 
1255 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1256 		return -EPERM;
1257 
1258 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1259 		struct sock *rtnl;
1260 		rtnl_dumpit_func dumpit;
1261 
1262 		dumpit = rtnl_get_dumpit(family, type);
1263 		if (dumpit == NULL)
1264 			return -EOPNOTSUPP;
1265 
1266 		__rtnl_unlock();
1267 		rtnl = net->rtnl;
1268 		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1269 		rtnl_lock();
1270 		return err;
1271 	}
1272 
1273 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1274 
1275 	min_len = rtm_min[sz_idx];
1276 	if (nlh->nlmsg_len < min_len)
1277 		return -EINVAL;
1278 
1279 	if (nlh->nlmsg_len > min_len) {
1280 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1281 		struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1282 
1283 		while (RTA_OK(attr, attrlen)) {
1284 			unsigned flavor = attr->rta_type;
1285 			if (flavor) {
1286 				if (flavor > rta_max[sz_idx])
1287 					return -EINVAL;
1288 				rta_buf[flavor-1] = attr;
1289 			}
1290 			attr = RTA_NEXT(attr, attrlen);
1291 		}
1292 	}
1293 
1294 	doit = rtnl_get_doit(family, type);
1295 	if (doit == NULL)
1296 		return -EOPNOTSUPP;
1297 
1298 	return doit(skb, nlh, (void *)&rta_buf[0]);
1299 }
1300 
1301 static void rtnetlink_rcv(struct sk_buff *skb)
1302 {
1303 	rtnl_lock();
1304 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1305 	rtnl_unlock();
1306 }
1307 
1308 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1309 {
1310 	struct net_device *dev = ptr;
1311 
1312 	switch (event) {
1313 	case NETDEV_UNREGISTER:
1314 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1315 		break;
1316 	case NETDEV_REGISTER:
1317 		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1318 		break;
1319 	case NETDEV_UP:
1320 	case NETDEV_DOWN:
1321 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1322 		break;
1323 	case NETDEV_CHANGE:
1324 	case NETDEV_GOING_DOWN:
1325 		break;
1326 	default:
1327 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1328 		break;
1329 	}
1330 	return NOTIFY_DONE;
1331 }
1332 
1333 static struct notifier_block rtnetlink_dev_notifier = {
1334 	.notifier_call	= rtnetlink_event,
1335 };
1336 
1337 
1338 static int rtnetlink_net_init(struct net *net)
1339 {
1340 	struct sock *sk;
1341 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1342 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1343 	if (!sk)
1344 		return -ENOMEM;
1345 	net->rtnl = sk;
1346 	return 0;
1347 }
1348 
1349 static void rtnetlink_net_exit(struct net *net)
1350 {
1351 	netlink_kernel_release(net->rtnl);
1352 	net->rtnl = NULL;
1353 }
1354 
1355 static struct pernet_operations rtnetlink_net_ops = {
1356 	.init = rtnetlink_net_init,
1357 	.exit = rtnetlink_net_exit,
1358 };
1359 
1360 void __init rtnetlink_init(void)
1361 {
1362 	int i;
1363 
1364 	rtattr_max = 0;
1365 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1366 		if (rta_max[i] > rtattr_max)
1367 			rtattr_max = rta_max[i];
1368 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1369 	if (!rta_buf)
1370 		panic("rtnetlink_init: cannot allocate rta_buf\n");
1371 
1372 	if (register_pernet_subsys(&rtnetlink_net_ops))
1373 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
1374 
1375 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1376 	register_netdevice_notifier(&rtnetlink_dev_notifier);
1377 
1378 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1379 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1380 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1381 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1382 
1383 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1384 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1385 }
1386 
1387 EXPORT_SYMBOL(__rta_fill);
1388 EXPORT_SYMBOL(rtnetlink_put_metrics);
1389 EXPORT_SYMBOL(rtnl_lock);
1390 EXPORT_SYMBOL(rtnl_trylock);
1391 EXPORT_SYMBOL(rtnl_unlock);
1392 EXPORT_SYMBOL(rtnl_unicast);
1393 EXPORT_SYMBOL(rtnl_notify);
1394 EXPORT_SYMBOL(rtnl_set_sk_err);
1395 EXPORT_SYMBOL(rtnl_create_link);
1396 EXPORT_SYMBOL(ifla_policy);
1397