xref: /openbmc/linux/net/core/rtnetlink.c (revision 732a675a)
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 int rtnl_is_locked(void)
86 {
87 	return mutex_is_locked(&rtnl_mutex);
88 }
89 
90 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
91 
92 static inline int rtm_msgindex(int msgtype)
93 {
94 	int msgindex = msgtype - RTM_BASE;
95 
96 	/*
97 	 * msgindex < 0 implies someone tried to register a netlink
98 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
99 	 * the message type has not been added to linux/rtnetlink.h
100 	 */
101 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
102 
103 	return msgindex;
104 }
105 
106 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
107 {
108 	struct rtnl_link *tab;
109 
110 	tab = rtnl_msg_handlers[protocol];
111 	if (tab == NULL || tab[msgindex].doit == NULL)
112 		tab = rtnl_msg_handlers[PF_UNSPEC];
113 
114 	return tab ? tab[msgindex].doit : NULL;
115 }
116 
117 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
118 {
119 	struct rtnl_link *tab;
120 
121 	tab = rtnl_msg_handlers[protocol];
122 	if (tab == NULL || tab[msgindex].dumpit == NULL)
123 		tab = rtnl_msg_handlers[PF_UNSPEC];
124 
125 	return tab ? tab[msgindex].dumpit : NULL;
126 }
127 
128 /**
129  * __rtnl_register - Register a rtnetlink message type
130  * @protocol: Protocol family or PF_UNSPEC
131  * @msgtype: rtnetlink message type
132  * @doit: Function pointer called for each request message
133  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
134  *
135  * Registers the specified function pointers (at least one of them has
136  * to be non-NULL) to be called whenever a request message for the
137  * specified protocol family and message type is received.
138  *
139  * The special protocol family PF_UNSPEC may be used to define fallback
140  * function pointers for the case when no entry for the specific protocol
141  * family exists.
142  *
143  * Returns 0 on success or a negative error code.
144  */
145 int __rtnl_register(int protocol, int msgtype,
146 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
147 {
148 	struct rtnl_link *tab;
149 	int msgindex;
150 
151 	BUG_ON(protocol < 0 || protocol >= NPROTO);
152 	msgindex = rtm_msgindex(msgtype);
153 
154 	tab = rtnl_msg_handlers[protocol];
155 	if (tab == NULL) {
156 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
157 		if (tab == NULL)
158 			return -ENOBUFS;
159 
160 		rtnl_msg_handlers[protocol] = tab;
161 	}
162 
163 	if (doit)
164 		tab[msgindex].doit = doit;
165 
166 	if (dumpit)
167 		tab[msgindex].dumpit = dumpit;
168 
169 	return 0;
170 }
171 
172 EXPORT_SYMBOL_GPL(__rtnl_register);
173 
174 /**
175  * rtnl_register - Register a rtnetlink message type
176  *
177  * Identical to __rtnl_register() but panics on failure. This is useful
178  * as failure of this function is very unlikely, it can only happen due
179  * to lack of memory when allocating the chain to store all message
180  * handlers for a protocol. Meant for use in init functions where lack
181  * of memory implies no sense in continueing.
182  */
183 void rtnl_register(int protocol, int msgtype,
184 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit)
185 {
186 	if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
187 		panic("Unable to register rtnetlink message handler, "
188 		      "protocol = %d, message type = %d\n",
189 		      protocol, msgtype);
190 }
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 
217 EXPORT_SYMBOL_GPL(rtnl_unregister);
218 
219 /**
220  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
221  * @protocol : Protocol family or PF_UNSPEC
222  *
223  * Identical to calling rtnl_unregster() for all registered message types
224  * of a certain protocol family.
225  */
226 void rtnl_unregister_all(int protocol)
227 {
228 	BUG_ON(protocol < 0 || protocol >= NPROTO);
229 
230 	kfree(rtnl_msg_handlers[protocol]);
231 	rtnl_msg_handlers[protocol] = NULL;
232 }
233 
234 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
235 
236 static LIST_HEAD(link_ops);
237 
238 /**
239  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
240  * @ops: struct rtnl_link_ops * to register
241  *
242  * The caller must hold the rtnl_mutex. This function should be used
243  * by drivers that create devices during module initialization. It
244  * must be called before registering the devices.
245  *
246  * Returns 0 on success or a negative error code.
247  */
248 int __rtnl_link_register(struct rtnl_link_ops *ops)
249 {
250 	if (!ops->dellink)
251 		ops->dellink = unregister_netdevice;
252 
253 	list_add_tail(&ops->list, &link_ops);
254 	return 0;
255 }
256 
257 EXPORT_SYMBOL_GPL(__rtnl_link_register);
258 
259 /**
260  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
261  * @ops: struct rtnl_link_ops * to register
262  *
263  * Returns 0 on success or a negative error code.
264  */
265 int rtnl_link_register(struct rtnl_link_ops *ops)
266 {
267 	int err;
268 
269 	rtnl_lock();
270 	err = __rtnl_link_register(ops);
271 	rtnl_unlock();
272 	return err;
273 }
274 
275 EXPORT_SYMBOL_GPL(rtnl_link_register);
276 
277 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
278 {
279 	struct net_device *dev;
280 restart:
281 	for_each_netdev(net, dev) {
282 		if (dev->rtnl_link_ops == ops) {
283 			ops->dellink(dev);
284 			goto restart;
285 		}
286 	}
287 }
288 
289 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
290 {
291 	rtnl_lock();
292 	__rtnl_kill_links(net, ops);
293 	rtnl_unlock();
294 }
295 EXPORT_SYMBOL_GPL(rtnl_kill_links);
296 
297 /**
298  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
299  * @ops: struct rtnl_link_ops * to unregister
300  *
301  * The caller must hold the rtnl_mutex.
302  */
303 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
304 {
305 	struct net *net;
306 
307 	for_each_net(net) {
308 		__rtnl_kill_links(net, ops);
309 	}
310 	list_del(&ops->list);
311 }
312 
313 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
314 
315 /**
316  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
317  * @ops: struct rtnl_link_ops * to unregister
318  */
319 void rtnl_link_unregister(struct rtnl_link_ops *ops)
320 {
321 	rtnl_lock();
322 	__rtnl_link_unregister(ops);
323 	rtnl_unlock();
324 }
325 
326 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
327 
328 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
329 {
330 	const struct rtnl_link_ops *ops;
331 
332 	list_for_each_entry(ops, &link_ops, list) {
333 		if (!strcmp(ops->kind, kind))
334 			return ops;
335 	}
336 	return NULL;
337 }
338 
339 static size_t rtnl_link_get_size(const struct net_device *dev)
340 {
341 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
342 	size_t size;
343 
344 	if (!ops)
345 		return 0;
346 
347 	size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
348 	       nlmsg_total_size(strlen(ops->kind) + 1);	 /* IFLA_INFO_KIND */
349 
350 	if (ops->get_size)
351 		/* IFLA_INFO_DATA + nested data */
352 		size += nlmsg_total_size(sizeof(struct nlattr)) +
353 			ops->get_size(dev);
354 
355 	if (ops->get_xstats_size)
356 		size += ops->get_xstats_size(dev);	/* IFLA_INFO_XSTATS */
357 
358 	return size;
359 }
360 
361 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
362 {
363 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
364 	struct nlattr *linkinfo, *data;
365 	int err = -EMSGSIZE;
366 
367 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
368 	if (linkinfo == NULL)
369 		goto out;
370 
371 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
372 		goto err_cancel_link;
373 	if (ops->fill_xstats) {
374 		err = ops->fill_xstats(skb, dev);
375 		if (err < 0)
376 			goto err_cancel_link;
377 	}
378 	if (ops->fill_info) {
379 		data = nla_nest_start(skb, IFLA_INFO_DATA);
380 		if (data == NULL)
381 			goto err_cancel_link;
382 		err = ops->fill_info(skb, dev);
383 		if (err < 0)
384 			goto err_cancel_data;
385 		nla_nest_end(skb, data);
386 	}
387 
388 	nla_nest_end(skb, linkinfo);
389 	return 0;
390 
391 err_cancel_data:
392 	nla_nest_cancel(skb, data);
393 err_cancel_link:
394 	nla_nest_cancel(skb, linkinfo);
395 out:
396 	return err;
397 }
398 
399 static const int rtm_min[RTM_NR_FAMILIES] =
400 {
401 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
402 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
403 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
404 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
405 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
406 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
407 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
408 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
409 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
410 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
411 };
412 
413 static const int rta_max[RTM_NR_FAMILIES] =
414 {
415 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
416 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
417 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
418 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
419 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
420 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
421 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
422 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
423 };
424 
425 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
426 {
427 	struct rtattr *rta;
428 	int size = RTA_LENGTH(attrlen);
429 
430 	rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
431 	rta->rta_type = attrtype;
432 	rta->rta_len = size;
433 	memcpy(RTA_DATA(rta), data, attrlen);
434 	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
435 }
436 
437 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
438 {
439 	struct sock *rtnl = net->rtnl;
440 	int err = 0;
441 
442 	NETLINK_CB(skb).dst_group = group;
443 	if (echo)
444 		atomic_inc(&skb->users);
445 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
446 	if (echo)
447 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
448 	return err;
449 }
450 
451 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
452 {
453 	struct sock *rtnl = net->rtnl;
454 
455 	return nlmsg_unicast(rtnl, skb, pid);
456 }
457 
458 int rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
459 		struct nlmsghdr *nlh, gfp_t flags)
460 {
461 	struct sock *rtnl = net->rtnl;
462 	int report = 0;
463 
464 	if (nlh)
465 		report = nlmsg_report(nlh);
466 
467 	return nlmsg_notify(rtnl, skb, pid, group, report, flags);
468 }
469 
470 void rtnl_set_sk_err(struct net *net, u32 group, int error)
471 {
472 	struct sock *rtnl = net->rtnl;
473 
474 	netlink_set_err(rtnl, 0, group, error);
475 }
476 
477 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
478 {
479 	struct nlattr *mx;
480 	int i, valid = 0;
481 
482 	mx = nla_nest_start(skb, RTA_METRICS);
483 	if (mx == NULL)
484 		return -ENOBUFS;
485 
486 	for (i = 0; i < RTAX_MAX; i++) {
487 		if (metrics[i]) {
488 			valid++;
489 			NLA_PUT_U32(skb, i+1, metrics[i]);
490 		}
491 	}
492 
493 	if (!valid) {
494 		nla_nest_cancel(skb, mx);
495 		return 0;
496 	}
497 
498 	return nla_nest_end(skb, mx);
499 
500 nla_put_failure:
501 	nla_nest_cancel(skb, mx);
502 	return -EMSGSIZE;
503 }
504 
505 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
506 		       u32 ts, u32 tsage, long expires, u32 error)
507 {
508 	struct rta_cacheinfo ci = {
509 		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
510 		.rta_used = dst->__use,
511 		.rta_clntref = atomic_read(&(dst->__refcnt)),
512 		.rta_error = error,
513 		.rta_id =  id,
514 		.rta_ts = ts,
515 		.rta_tsage = tsage,
516 	};
517 
518 	if (expires)
519 		ci.rta_expires = jiffies_to_clock_t(expires);
520 
521 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
522 }
523 
524 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
525 
526 static void set_operstate(struct net_device *dev, unsigned char transition)
527 {
528 	unsigned char operstate = dev->operstate;
529 
530 	switch(transition) {
531 	case IF_OPER_UP:
532 		if ((operstate == IF_OPER_DORMANT ||
533 		     operstate == IF_OPER_UNKNOWN) &&
534 		    !netif_dormant(dev))
535 			operstate = IF_OPER_UP;
536 		break;
537 
538 	case IF_OPER_DORMANT:
539 		if (operstate == IF_OPER_UP ||
540 		    operstate == IF_OPER_UNKNOWN)
541 			operstate = IF_OPER_DORMANT;
542 		break;
543 	}
544 
545 	if (dev->operstate != operstate) {
546 		write_lock_bh(&dev_base_lock);
547 		dev->operstate = operstate;
548 		write_unlock_bh(&dev_base_lock);
549 		netdev_state_change(dev);
550 	}
551 }
552 
553 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
554 				 struct net_device_stats *b)
555 {
556 	a->rx_packets = b->rx_packets;
557 	a->tx_packets = b->tx_packets;
558 	a->rx_bytes = b->rx_bytes;
559 	a->tx_bytes = b->tx_bytes;
560 	a->rx_errors = b->rx_errors;
561 	a->tx_errors = b->tx_errors;
562 	a->rx_dropped = b->rx_dropped;
563 	a->tx_dropped = b->tx_dropped;
564 
565 	a->multicast = b->multicast;
566 	a->collisions = b->collisions;
567 
568 	a->rx_length_errors = b->rx_length_errors;
569 	a->rx_over_errors = b->rx_over_errors;
570 	a->rx_crc_errors = b->rx_crc_errors;
571 	a->rx_frame_errors = b->rx_frame_errors;
572 	a->rx_fifo_errors = b->rx_fifo_errors;
573 	a->rx_missed_errors = b->rx_missed_errors;
574 
575 	a->tx_aborted_errors = b->tx_aborted_errors;
576 	a->tx_carrier_errors = b->tx_carrier_errors;
577 	a->tx_fifo_errors = b->tx_fifo_errors;
578 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
579 	a->tx_window_errors = b->tx_window_errors;
580 
581 	a->rx_compressed = b->rx_compressed;
582 	a->tx_compressed = b->tx_compressed;
583 };
584 
585 static inline size_t if_nlmsg_size(const struct net_device *dev)
586 {
587 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
588 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
589 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
590 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
591 	       + nla_total_size(sizeof(struct rtnl_link_stats))
592 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
593 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
594 	       + nla_total_size(4) /* IFLA_TXQLEN */
595 	       + nla_total_size(4) /* IFLA_WEIGHT */
596 	       + nla_total_size(4) /* IFLA_MTU */
597 	       + nla_total_size(4) /* IFLA_LINK */
598 	       + nla_total_size(4) /* IFLA_MASTER */
599 	       + nla_total_size(1) /* IFLA_OPERSTATE */
600 	       + nla_total_size(1) /* IFLA_LINKMODE */
601 	       + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
602 }
603 
604 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
605 			    int type, u32 pid, u32 seq, u32 change,
606 			    unsigned int flags)
607 {
608 	struct ifinfomsg *ifm;
609 	struct nlmsghdr *nlh;
610 
611 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
612 	if (nlh == NULL)
613 		return -EMSGSIZE;
614 
615 	ifm = nlmsg_data(nlh);
616 	ifm->ifi_family = AF_UNSPEC;
617 	ifm->__ifi_pad = 0;
618 	ifm->ifi_type = dev->type;
619 	ifm->ifi_index = dev->ifindex;
620 	ifm->ifi_flags = dev_get_flags(dev);
621 	ifm->ifi_change = change;
622 
623 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
624 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
625 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
626 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
627 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
628 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
629 
630 	if (dev->ifindex != dev->iflink)
631 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
632 
633 	if (dev->master)
634 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
635 
636 	if (dev->qdisc_sleeping)
637 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
638 
639 	if (1) {
640 		struct rtnl_link_ifmap map = {
641 			.mem_start   = dev->mem_start,
642 			.mem_end     = dev->mem_end,
643 			.base_addr   = dev->base_addr,
644 			.irq         = dev->irq,
645 			.dma         = dev->dma,
646 			.port        = dev->if_port,
647 		};
648 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
649 	}
650 
651 	if (dev->addr_len) {
652 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
653 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
654 	}
655 
656 	if (dev->get_stats) {
657 		struct net_device_stats *stats = dev->get_stats(dev);
658 		if (stats) {
659 			struct nlattr *attr;
660 
661 			attr = nla_reserve(skb, IFLA_STATS,
662 					   sizeof(struct rtnl_link_stats));
663 			if (attr == NULL)
664 				goto nla_put_failure;
665 
666 			copy_rtnl_link_stats(nla_data(attr), stats);
667 		}
668 	}
669 
670 	if (dev->rtnl_link_ops) {
671 		if (rtnl_link_fill(skb, dev) < 0)
672 			goto nla_put_failure;
673 	}
674 
675 	return nlmsg_end(skb, nlh);
676 
677 nla_put_failure:
678 	nlmsg_cancel(skb, nlh);
679 	return -EMSGSIZE;
680 }
681 
682 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
683 {
684 	struct net *net = sock_net(skb->sk);
685 	int idx;
686 	int s_idx = cb->args[0];
687 	struct net_device *dev;
688 
689 	idx = 0;
690 	for_each_netdev(net, dev) {
691 		if (idx < s_idx)
692 			goto cont;
693 		if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
694 				     NETLINK_CB(cb->skb).pid,
695 				     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
696 			break;
697 cont:
698 		idx++;
699 	}
700 	cb->args[0] = idx;
701 
702 	return skb->len;
703 }
704 
705 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
706 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
707 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
708 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
709 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
710 	[IFLA_MTU]		= { .type = NLA_U32 },
711 	[IFLA_LINK]		= { .type = NLA_U32 },
712 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
713 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
714 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
715 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
716 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
717 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
718 };
719 
720 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
721 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
722 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
723 };
724 
725 static struct net *get_net_ns_by_pid(pid_t pid)
726 {
727 	struct task_struct *tsk;
728 	struct net *net;
729 
730 	/* Lookup the network namespace */
731 	net = ERR_PTR(-ESRCH);
732 	rcu_read_lock();
733 	tsk = find_task_by_vpid(pid);
734 	if (tsk) {
735 		struct nsproxy *nsproxy;
736 		nsproxy = task_nsproxy(tsk);
737 		if (nsproxy)
738 			net = get_net(nsproxy->net_ns);
739 	}
740 	rcu_read_unlock();
741 	return net;
742 }
743 
744 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
745 {
746 	if (dev) {
747 		if (tb[IFLA_ADDRESS] &&
748 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
749 			return -EINVAL;
750 
751 		if (tb[IFLA_BROADCAST] &&
752 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
753 			return -EINVAL;
754 	}
755 
756 	return 0;
757 }
758 
759 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
760 		      struct nlattr **tb, char *ifname, int modified)
761 {
762 	int send_addr_notify = 0;
763 	int err;
764 
765 	if (tb[IFLA_NET_NS_PID]) {
766 		struct net *net;
767 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
768 		if (IS_ERR(net)) {
769 			err = PTR_ERR(net);
770 			goto errout;
771 		}
772 		err = dev_change_net_namespace(dev, net, ifname);
773 		put_net(net);
774 		if (err)
775 			goto errout;
776 		modified = 1;
777 	}
778 
779 	if (tb[IFLA_MAP]) {
780 		struct rtnl_link_ifmap *u_map;
781 		struct ifmap k_map;
782 
783 		if (!dev->set_config) {
784 			err = -EOPNOTSUPP;
785 			goto errout;
786 		}
787 
788 		if (!netif_device_present(dev)) {
789 			err = -ENODEV;
790 			goto errout;
791 		}
792 
793 		u_map = nla_data(tb[IFLA_MAP]);
794 		k_map.mem_start = (unsigned long) u_map->mem_start;
795 		k_map.mem_end = (unsigned long) u_map->mem_end;
796 		k_map.base_addr = (unsigned short) u_map->base_addr;
797 		k_map.irq = (unsigned char) u_map->irq;
798 		k_map.dma = (unsigned char) u_map->dma;
799 		k_map.port = (unsigned char) u_map->port;
800 
801 		err = dev->set_config(dev, &k_map);
802 		if (err < 0)
803 			goto errout;
804 
805 		modified = 1;
806 	}
807 
808 	if (tb[IFLA_ADDRESS]) {
809 		struct sockaddr *sa;
810 		int len;
811 
812 		if (!dev->set_mac_address) {
813 			err = -EOPNOTSUPP;
814 			goto errout;
815 		}
816 
817 		if (!netif_device_present(dev)) {
818 			err = -ENODEV;
819 			goto errout;
820 		}
821 
822 		len = sizeof(sa_family_t) + dev->addr_len;
823 		sa = kmalloc(len, GFP_KERNEL);
824 		if (!sa) {
825 			err = -ENOMEM;
826 			goto errout;
827 		}
828 		sa->sa_family = dev->type;
829 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
830 		       dev->addr_len);
831 		err = dev->set_mac_address(dev, sa);
832 		kfree(sa);
833 		if (err)
834 			goto errout;
835 		send_addr_notify = 1;
836 		modified = 1;
837 	}
838 
839 	if (tb[IFLA_MTU]) {
840 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
841 		if (err < 0)
842 			goto errout;
843 		modified = 1;
844 	}
845 
846 	/*
847 	 * Interface selected by interface index but interface
848 	 * name provided implies that a name change has been
849 	 * requested.
850 	 */
851 	if (ifm->ifi_index > 0 && ifname[0]) {
852 		err = dev_change_name(dev, ifname);
853 		if (err < 0)
854 			goto errout;
855 		modified = 1;
856 	}
857 
858 	if (tb[IFLA_BROADCAST]) {
859 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
860 		send_addr_notify = 1;
861 	}
862 
863 	if (ifm->ifi_flags || ifm->ifi_change) {
864 		unsigned int flags = ifm->ifi_flags;
865 
866 		/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
867 		if (ifm->ifi_change)
868 			flags = (flags & ifm->ifi_change) |
869 				(dev->flags & ~ifm->ifi_change);
870 		dev_change_flags(dev, flags);
871 	}
872 
873 	if (tb[IFLA_TXQLEN])
874 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
875 
876 	if (tb[IFLA_OPERSTATE])
877 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
878 
879 	if (tb[IFLA_LINKMODE]) {
880 		write_lock_bh(&dev_base_lock);
881 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
882 		write_unlock_bh(&dev_base_lock);
883 	}
884 
885 	err = 0;
886 
887 errout:
888 	if (err < 0 && modified && net_ratelimit())
889 		printk(KERN_WARNING "A link change request failed with "
890 		       "some changes comitted already. Interface %s may "
891 		       "have been left with an inconsistent configuration, "
892 		       "please check.\n", dev->name);
893 
894 	if (send_addr_notify)
895 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
896 	return err;
897 }
898 
899 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
900 {
901 	struct net *net = sock_net(skb->sk);
902 	struct ifinfomsg *ifm;
903 	struct net_device *dev;
904 	int err;
905 	struct nlattr *tb[IFLA_MAX+1];
906 	char ifname[IFNAMSIZ];
907 
908 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
909 	if (err < 0)
910 		goto errout;
911 
912 	if (tb[IFLA_IFNAME])
913 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
914 	else
915 		ifname[0] = '\0';
916 
917 	err = -EINVAL;
918 	ifm = nlmsg_data(nlh);
919 	if (ifm->ifi_index > 0)
920 		dev = dev_get_by_index(net, ifm->ifi_index);
921 	else if (tb[IFLA_IFNAME])
922 		dev = dev_get_by_name(net, ifname);
923 	else
924 		goto errout;
925 
926 	if (dev == NULL) {
927 		err = -ENODEV;
928 		goto errout;
929 	}
930 
931 	if ((err = validate_linkmsg(dev, tb)) < 0)
932 		goto errout_dev;
933 
934 	err = do_setlink(dev, ifm, tb, ifname, 0);
935 errout_dev:
936 	dev_put(dev);
937 errout:
938 	return err;
939 }
940 
941 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
942 {
943 	struct net *net = sock_net(skb->sk);
944 	const struct rtnl_link_ops *ops;
945 	struct net_device *dev;
946 	struct ifinfomsg *ifm;
947 	char ifname[IFNAMSIZ];
948 	struct nlattr *tb[IFLA_MAX+1];
949 	int err;
950 
951 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
952 	if (err < 0)
953 		return err;
954 
955 	if (tb[IFLA_IFNAME])
956 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
957 
958 	ifm = nlmsg_data(nlh);
959 	if (ifm->ifi_index > 0)
960 		dev = __dev_get_by_index(net, ifm->ifi_index);
961 	else if (tb[IFLA_IFNAME])
962 		dev = __dev_get_by_name(net, ifname);
963 	else
964 		return -EINVAL;
965 
966 	if (!dev)
967 		return -ENODEV;
968 
969 	ops = dev->rtnl_link_ops;
970 	if (!ops)
971 		return -EOPNOTSUPP;
972 
973 	ops->dellink(dev);
974 	return 0;
975 }
976 
977 struct net_device *rtnl_create_link(struct net *net, char *ifname,
978 		const struct rtnl_link_ops *ops, struct nlattr *tb[])
979 {
980 	int err;
981 	struct net_device *dev;
982 
983 	err = -ENOMEM;
984 	dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
985 	if (!dev)
986 		goto err;
987 
988 	if (strchr(dev->name, '%')) {
989 		err = dev_alloc_name(dev, dev->name);
990 		if (err < 0)
991 			goto err_free;
992 	}
993 
994 	dev_net_set(dev, net);
995 	dev->rtnl_link_ops = ops;
996 
997 	if (tb[IFLA_MTU])
998 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
999 	if (tb[IFLA_ADDRESS])
1000 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1001 				nla_len(tb[IFLA_ADDRESS]));
1002 	if (tb[IFLA_BROADCAST])
1003 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1004 				nla_len(tb[IFLA_BROADCAST]));
1005 	if (tb[IFLA_TXQLEN])
1006 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1007 	if (tb[IFLA_OPERSTATE])
1008 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1009 	if (tb[IFLA_LINKMODE])
1010 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1011 
1012 	return dev;
1013 
1014 err_free:
1015 	free_netdev(dev);
1016 err:
1017 	return ERR_PTR(err);
1018 }
1019 
1020 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1021 {
1022 	struct net *net = sock_net(skb->sk);
1023 	const struct rtnl_link_ops *ops;
1024 	struct net_device *dev;
1025 	struct ifinfomsg *ifm;
1026 	char kind[MODULE_NAME_LEN];
1027 	char ifname[IFNAMSIZ];
1028 	struct nlattr *tb[IFLA_MAX+1];
1029 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1030 	int err;
1031 
1032 #ifdef CONFIG_KMOD
1033 replay:
1034 #endif
1035 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1036 	if (err < 0)
1037 		return err;
1038 
1039 	if (tb[IFLA_IFNAME])
1040 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1041 	else
1042 		ifname[0] = '\0';
1043 
1044 	ifm = nlmsg_data(nlh);
1045 	if (ifm->ifi_index > 0)
1046 		dev = __dev_get_by_index(net, ifm->ifi_index);
1047 	else if (ifname[0])
1048 		dev = __dev_get_by_name(net, ifname);
1049 	else
1050 		dev = NULL;
1051 
1052 	if ((err = validate_linkmsg(dev, tb)) < 0)
1053 		return err;
1054 
1055 	if (tb[IFLA_LINKINFO]) {
1056 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1057 				       tb[IFLA_LINKINFO], ifla_info_policy);
1058 		if (err < 0)
1059 			return err;
1060 	} else
1061 		memset(linkinfo, 0, sizeof(linkinfo));
1062 
1063 	if (linkinfo[IFLA_INFO_KIND]) {
1064 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1065 		ops = rtnl_link_ops_get(kind);
1066 	} else {
1067 		kind[0] = '\0';
1068 		ops = NULL;
1069 	}
1070 
1071 	if (1) {
1072 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1073 
1074 		if (ops) {
1075 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1076 				err = nla_parse_nested(attr, ops->maxtype,
1077 						       linkinfo[IFLA_INFO_DATA],
1078 						       ops->policy);
1079 				if (err < 0)
1080 					return err;
1081 				data = attr;
1082 			}
1083 			if (ops->validate) {
1084 				err = ops->validate(tb, data);
1085 				if (err < 0)
1086 					return err;
1087 			}
1088 		}
1089 
1090 		if (dev) {
1091 			int modified = 0;
1092 
1093 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1094 				return -EEXIST;
1095 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1096 				return -EOPNOTSUPP;
1097 
1098 			if (linkinfo[IFLA_INFO_DATA]) {
1099 				if (!ops || ops != dev->rtnl_link_ops ||
1100 				    !ops->changelink)
1101 					return -EOPNOTSUPP;
1102 
1103 				err = ops->changelink(dev, tb, data);
1104 				if (err < 0)
1105 					return err;
1106 				modified = 1;
1107 			}
1108 
1109 			return do_setlink(dev, ifm, tb, ifname, modified);
1110 		}
1111 
1112 		if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1113 			return -ENODEV;
1114 
1115 		if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1116 			return -EOPNOTSUPP;
1117 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1118 			return -EOPNOTSUPP;
1119 
1120 		if (!ops) {
1121 #ifdef CONFIG_KMOD
1122 			if (kind[0]) {
1123 				__rtnl_unlock();
1124 				request_module("rtnl-link-%s", kind);
1125 				rtnl_lock();
1126 				ops = rtnl_link_ops_get(kind);
1127 				if (ops)
1128 					goto replay;
1129 			}
1130 #endif
1131 			return -EOPNOTSUPP;
1132 		}
1133 
1134 		if (!ifname[0])
1135 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1136 
1137 		dev = rtnl_create_link(net, ifname, ops, tb);
1138 
1139 		if (IS_ERR(dev))
1140 			err = PTR_ERR(dev);
1141 		else if (ops->newlink)
1142 			err = ops->newlink(dev, tb, data);
1143 		else
1144 			err = register_netdevice(dev);
1145 
1146 		if (err < 0 && !IS_ERR(dev))
1147 			free_netdev(dev);
1148 		return err;
1149 	}
1150 }
1151 
1152 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1153 {
1154 	struct net *net = sock_net(skb->sk);
1155 	struct ifinfomsg *ifm;
1156 	struct nlattr *tb[IFLA_MAX+1];
1157 	struct net_device *dev = NULL;
1158 	struct sk_buff *nskb;
1159 	int err;
1160 
1161 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1162 	if (err < 0)
1163 		return err;
1164 
1165 	ifm = nlmsg_data(nlh);
1166 	if (ifm->ifi_index > 0) {
1167 		dev = dev_get_by_index(net, ifm->ifi_index);
1168 		if (dev == NULL)
1169 			return -ENODEV;
1170 	} else
1171 		return -EINVAL;
1172 
1173 	nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1174 	if (nskb == NULL) {
1175 		err = -ENOBUFS;
1176 		goto errout;
1177 	}
1178 
1179 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1180 			       nlh->nlmsg_seq, 0, 0);
1181 	if (err < 0) {
1182 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1183 		WARN_ON(err == -EMSGSIZE);
1184 		kfree_skb(nskb);
1185 		goto errout;
1186 	}
1187 	err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1188 errout:
1189 	dev_put(dev);
1190 
1191 	return err;
1192 }
1193 
1194 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1195 {
1196 	int idx;
1197 	int s_idx = cb->family;
1198 
1199 	if (s_idx == 0)
1200 		s_idx = 1;
1201 	for (idx=1; idx<NPROTO; idx++) {
1202 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1203 		if (idx < s_idx || idx == PF_PACKET)
1204 			continue;
1205 		if (rtnl_msg_handlers[idx] == NULL ||
1206 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1207 			continue;
1208 		if (idx > s_idx)
1209 			memset(&cb->args[0], 0, sizeof(cb->args));
1210 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1211 			break;
1212 	}
1213 	cb->family = idx;
1214 
1215 	return skb->len;
1216 }
1217 
1218 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1219 {
1220 	struct net *net = dev_net(dev);
1221 	struct sk_buff *skb;
1222 	int err = -ENOBUFS;
1223 
1224 	skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1225 	if (skb == NULL)
1226 		goto errout;
1227 
1228 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1229 	if (err < 0) {
1230 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1231 		WARN_ON(err == -EMSGSIZE);
1232 		kfree_skb(skb);
1233 		goto errout;
1234 	}
1235 	err = rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1236 errout:
1237 	if (err < 0)
1238 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1239 }
1240 
1241 /* Protected by RTNL sempahore.  */
1242 static struct rtattr **rta_buf;
1243 static int rtattr_max;
1244 
1245 /* Process one rtnetlink message. */
1246 
1247 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1248 {
1249 	struct net *net = sock_net(skb->sk);
1250 	rtnl_doit_func doit;
1251 	int sz_idx, kind;
1252 	int min_len;
1253 	int family;
1254 	int type;
1255 	int err;
1256 
1257 	type = nlh->nlmsg_type;
1258 	if (type > RTM_MAX)
1259 		return -EOPNOTSUPP;
1260 
1261 	type -= RTM_BASE;
1262 
1263 	/* All the messages must have at least 1 byte length */
1264 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1265 		return 0;
1266 
1267 	family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1268 	if (family >= NPROTO)
1269 		return -EAFNOSUPPORT;
1270 
1271 	sz_idx = type>>2;
1272 	kind = type&3;
1273 
1274 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1275 		return -EPERM;
1276 
1277 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1278 		struct sock *rtnl;
1279 		rtnl_dumpit_func dumpit;
1280 
1281 		dumpit = rtnl_get_dumpit(family, type);
1282 		if (dumpit == NULL)
1283 			return -EOPNOTSUPP;
1284 
1285 		__rtnl_unlock();
1286 		rtnl = net->rtnl;
1287 		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1288 		rtnl_lock();
1289 		return err;
1290 	}
1291 
1292 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1293 
1294 	min_len = rtm_min[sz_idx];
1295 	if (nlh->nlmsg_len < min_len)
1296 		return -EINVAL;
1297 
1298 	if (nlh->nlmsg_len > min_len) {
1299 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1300 		struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1301 
1302 		while (RTA_OK(attr, attrlen)) {
1303 			unsigned flavor = attr->rta_type;
1304 			if (flavor) {
1305 				if (flavor > rta_max[sz_idx])
1306 					return -EINVAL;
1307 				rta_buf[flavor-1] = attr;
1308 			}
1309 			attr = RTA_NEXT(attr, attrlen);
1310 		}
1311 	}
1312 
1313 	doit = rtnl_get_doit(family, type);
1314 	if (doit == NULL)
1315 		return -EOPNOTSUPP;
1316 
1317 	return doit(skb, nlh, (void *)&rta_buf[0]);
1318 }
1319 
1320 static void rtnetlink_rcv(struct sk_buff *skb)
1321 {
1322 	rtnl_lock();
1323 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1324 	rtnl_unlock();
1325 }
1326 
1327 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1328 {
1329 	struct net_device *dev = ptr;
1330 
1331 	switch (event) {
1332 	case NETDEV_UNREGISTER:
1333 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1334 		break;
1335 	case NETDEV_REGISTER:
1336 		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1337 		break;
1338 	case NETDEV_UP:
1339 	case NETDEV_DOWN:
1340 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1341 		break;
1342 	case NETDEV_CHANGE:
1343 	case NETDEV_GOING_DOWN:
1344 		break;
1345 	default:
1346 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1347 		break;
1348 	}
1349 	return NOTIFY_DONE;
1350 }
1351 
1352 static struct notifier_block rtnetlink_dev_notifier = {
1353 	.notifier_call	= rtnetlink_event,
1354 };
1355 
1356 
1357 static int rtnetlink_net_init(struct net *net)
1358 {
1359 	struct sock *sk;
1360 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1361 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1362 	if (!sk)
1363 		return -ENOMEM;
1364 	net->rtnl = sk;
1365 	return 0;
1366 }
1367 
1368 static void rtnetlink_net_exit(struct net *net)
1369 {
1370 	netlink_kernel_release(net->rtnl);
1371 	net->rtnl = NULL;
1372 }
1373 
1374 static struct pernet_operations rtnetlink_net_ops = {
1375 	.init = rtnetlink_net_init,
1376 	.exit = rtnetlink_net_exit,
1377 };
1378 
1379 void __init rtnetlink_init(void)
1380 {
1381 	int i;
1382 
1383 	rtattr_max = 0;
1384 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1385 		if (rta_max[i] > rtattr_max)
1386 			rtattr_max = rta_max[i];
1387 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1388 	if (!rta_buf)
1389 		panic("rtnetlink_init: cannot allocate rta_buf\n");
1390 
1391 	if (register_pernet_subsys(&rtnetlink_net_ops))
1392 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
1393 
1394 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1395 	register_netdevice_notifier(&rtnetlink_dev_notifier);
1396 
1397 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1398 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1399 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1400 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1401 
1402 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1403 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1404 }
1405 
1406 EXPORT_SYMBOL(__rta_fill);
1407 EXPORT_SYMBOL(rtnetlink_put_metrics);
1408 EXPORT_SYMBOL(rtnl_lock);
1409 EXPORT_SYMBOL(rtnl_trylock);
1410 EXPORT_SYMBOL(rtnl_unlock);
1411 EXPORT_SYMBOL(rtnl_is_locked);
1412 EXPORT_SYMBOL(rtnl_unicast);
1413 EXPORT_SYMBOL(rtnl_notify);
1414 EXPORT_SYMBOL(rtnl_set_sk_err);
1415 EXPORT_SYMBOL(rtnl_create_link);
1416 EXPORT_SYMBOL(ifla_policy);
1417