xref: /openbmc/linux/net/core/rtnetlink.c (revision 0a3abcf75bf391fec4e32356ab5ddb8f5d2e6b41)
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 int set_operstate(struct net_device *dev, unsigned char transition, bool send_notification)
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 
531 		if (send_notification)
532 			netdev_state_change(dev);
533 		return 1;
534 	} else
535 		return 0;
536 }
537 
538 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
539 				 struct net_device_stats *b)
540 {
541 	a->rx_packets = b->rx_packets;
542 	a->tx_packets = b->tx_packets;
543 	a->rx_bytes = b->rx_bytes;
544 	a->tx_bytes = b->tx_bytes;
545 	a->rx_errors = b->rx_errors;
546 	a->tx_errors = b->tx_errors;
547 	a->rx_dropped = b->rx_dropped;
548 	a->tx_dropped = b->tx_dropped;
549 
550 	a->multicast = b->multicast;
551 	a->collisions = b->collisions;
552 
553 	a->rx_length_errors = b->rx_length_errors;
554 	a->rx_over_errors = b->rx_over_errors;
555 	a->rx_crc_errors = b->rx_crc_errors;
556 	a->rx_frame_errors = b->rx_frame_errors;
557 	a->rx_fifo_errors = b->rx_fifo_errors;
558 	a->rx_missed_errors = b->rx_missed_errors;
559 
560 	a->tx_aborted_errors = b->tx_aborted_errors;
561 	a->tx_carrier_errors = b->tx_carrier_errors;
562 	a->tx_fifo_errors = b->tx_fifo_errors;
563 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
564 	a->tx_window_errors = b->tx_window_errors;
565 
566 	a->rx_compressed = b->rx_compressed;
567 	a->tx_compressed = b->tx_compressed;
568 };
569 
570 static inline size_t if_nlmsg_size(const struct net_device *dev)
571 {
572 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
573 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
574 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
575 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
576 	       + nla_total_size(sizeof(struct rtnl_link_stats))
577 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
578 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
579 	       + nla_total_size(4) /* IFLA_TXQLEN */
580 	       + nla_total_size(4) /* IFLA_WEIGHT */
581 	       + nla_total_size(4) /* IFLA_MTU */
582 	       + nla_total_size(4) /* IFLA_LINK */
583 	       + nla_total_size(4) /* IFLA_MASTER */
584 	       + nla_total_size(1) /* IFLA_OPERSTATE */
585 	       + nla_total_size(1) /* IFLA_LINKMODE */
586 	       + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
587 }
588 
589 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
590 			    int type, u32 pid, u32 seq, u32 change,
591 			    unsigned int flags)
592 {
593 	struct ifinfomsg *ifm;
594 	struct nlmsghdr *nlh;
595 
596 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
597 	if (nlh == NULL)
598 		return -EMSGSIZE;
599 
600 	ifm = nlmsg_data(nlh);
601 	ifm->ifi_family = AF_UNSPEC;
602 	ifm->__ifi_pad = 0;
603 	ifm->ifi_type = dev->type;
604 	ifm->ifi_index = dev->ifindex;
605 	ifm->ifi_flags = dev_get_flags(dev);
606 	ifm->ifi_change = change;
607 
608 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
609 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
610 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
611 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
612 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
613 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
614 
615 	if (dev->ifindex != dev->iflink)
616 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
617 
618 	if (dev->master)
619 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
620 
621 	if (dev->qdisc_sleeping)
622 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
623 
624 	if (1) {
625 		struct rtnl_link_ifmap map = {
626 			.mem_start   = dev->mem_start,
627 			.mem_end     = dev->mem_end,
628 			.base_addr   = dev->base_addr,
629 			.irq         = dev->irq,
630 			.dma         = dev->dma,
631 			.port        = dev->if_port,
632 		};
633 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
634 	}
635 
636 	if (dev->addr_len) {
637 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
638 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
639 	}
640 
641 	if (dev->get_stats) {
642 		struct net_device_stats *stats = dev->get_stats(dev);
643 		if (stats) {
644 			struct nlattr *attr;
645 
646 			attr = nla_reserve(skb, IFLA_STATS,
647 					   sizeof(struct rtnl_link_stats));
648 			if (attr == NULL)
649 				goto nla_put_failure;
650 
651 			copy_rtnl_link_stats(nla_data(attr), stats);
652 		}
653 	}
654 
655 	if (dev->rtnl_link_ops) {
656 		if (rtnl_link_fill(skb, dev) < 0)
657 			goto nla_put_failure;
658 	}
659 
660 	return nlmsg_end(skb, nlh);
661 
662 nla_put_failure:
663 	nlmsg_cancel(skb, nlh);
664 	return -EMSGSIZE;
665 }
666 
667 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
668 {
669 	struct net *net = skb->sk->sk_net;
670 	int idx;
671 	int s_idx = cb->args[0];
672 	struct net_device *dev;
673 
674 	idx = 0;
675 	for_each_netdev(net, dev) {
676 		if (idx < s_idx)
677 			goto cont;
678 		if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
679 				     NETLINK_CB(cb->skb).pid,
680 				     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
681 			break;
682 cont:
683 		idx++;
684 	}
685 	cb->args[0] = idx;
686 
687 	return skb->len;
688 }
689 
690 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
691 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
692 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
693 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
694 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
695 	[IFLA_MTU]		= { .type = NLA_U32 },
696 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
697 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
698 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
699 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
700 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
701 };
702 
703 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
704 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
705 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
706 };
707 
708 static struct net *get_net_ns_by_pid(pid_t pid)
709 {
710 	struct task_struct *tsk;
711 	struct net *net;
712 
713 	/* Lookup the network namespace */
714 	net = ERR_PTR(-ESRCH);
715 	rcu_read_lock();
716 	tsk = find_task_by_vpid(pid);
717 	if (tsk) {
718 		struct nsproxy *nsproxy;
719 		nsproxy = task_nsproxy(tsk);
720 		if (nsproxy)
721 			net = get_net(nsproxy->net_ns);
722 	}
723 	rcu_read_unlock();
724 	return net;
725 }
726 
727 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
728 		      struct nlattr **tb, char *ifname, int modified)
729 {
730 	int send_addr_notify = 0;
731 	int err;
732 
733 	if (tb[IFLA_NET_NS_PID]) {
734 		struct net *net;
735 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
736 		if (IS_ERR(net)) {
737 			err = PTR_ERR(net);
738 			goto errout;
739 		}
740 		err = dev_change_net_namespace(dev, net, ifname);
741 		put_net(net);
742 		if (err)
743 			goto errout;
744 		modified = 1;
745 	}
746 
747 	if (tb[IFLA_MAP]) {
748 		struct rtnl_link_ifmap *u_map;
749 		struct ifmap k_map;
750 
751 		if (!dev->set_config) {
752 			err = -EOPNOTSUPP;
753 			goto errout;
754 		}
755 
756 		if (!netif_device_present(dev)) {
757 			err = -ENODEV;
758 			goto errout;
759 		}
760 
761 		u_map = nla_data(tb[IFLA_MAP]);
762 		k_map.mem_start = (unsigned long) u_map->mem_start;
763 		k_map.mem_end = (unsigned long) u_map->mem_end;
764 		k_map.base_addr = (unsigned short) u_map->base_addr;
765 		k_map.irq = (unsigned char) u_map->irq;
766 		k_map.dma = (unsigned char) u_map->dma;
767 		k_map.port = (unsigned char) u_map->port;
768 
769 		err = dev->set_config(dev, &k_map);
770 		if (err < 0)
771 			goto errout;
772 
773 		modified = 1;
774 	}
775 
776 	if (tb[IFLA_ADDRESS]) {
777 		struct sockaddr *sa;
778 		int len;
779 
780 		if (!dev->set_mac_address) {
781 			err = -EOPNOTSUPP;
782 			goto errout;
783 		}
784 
785 		if (!netif_device_present(dev)) {
786 			err = -ENODEV;
787 			goto errout;
788 		}
789 
790 		len = sizeof(sa_family_t) + dev->addr_len;
791 		sa = kmalloc(len, GFP_KERNEL);
792 		if (!sa) {
793 			err = -ENOMEM;
794 			goto errout;
795 		}
796 		sa->sa_family = dev->type;
797 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
798 		       dev->addr_len);
799 		err = dev->set_mac_address(dev, sa);
800 		kfree(sa);
801 		if (err)
802 			goto errout;
803 		send_addr_notify = 1;
804 		modified = 1;
805 	}
806 
807 	if (tb[IFLA_MTU]) {
808 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
809 		if (err < 0)
810 			goto errout;
811 		modified = 1;
812 	}
813 
814 	/*
815 	 * Interface selected by interface index but interface
816 	 * name provided implies that a name change has been
817 	 * requested.
818 	 */
819 	if (ifm->ifi_index > 0 && ifname[0]) {
820 		err = dev_change_name(dev, ifname);
821 		if (err < 0)
822 			goto errout;
823 		modified = 1;
824 	}
825 
826 	if (tb[IFLA_BROADCAST]) {
827 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
828 		send_addr_notify = 1;
829 		modified = 1;
830 	}
831 
832 	if (ifm->ifi_flags || ifm->ifi_change) {
833 		unsigned int flags = ifm->ifi_flags;
834 
835 		/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
836 		if (ifm->ifi_change)
837 			flags = (flags & ifm->ifi_change) |
838 				(dev->flags & ~ifm->ifi_change);
839 		dev_change_flags(dev, flags);
840 	}
841 
842 	if (tb[IFLA_TXQLEN]) {
843 		if (dev->tx_queue_len != nla_get_u32(tb[IFLA_TXQLEN])) {
844 			dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
845 			modified = 1;
846 		}
847 	}
848 
849 	if (tb[IFLA_OPERSTATE])
850 		modified |= set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]), false);
851 
852 	if (tb[IFLA_LINKMODE]) {
853 		if (dev->link_mode != nla_get_u8(tb[IFLA_LINKMODE])) {
854 			write_lock_bh(&dev_base_lock);
855 			dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
856 			write_lock_bh(&dev_base_lock);
857 			modified = 1;
858 		}
859 	}
860 
861 	err = 0;
862 
863 errout:
864 	if (err < 0 && modified && net_ratelimit())
865 		printk(KERN_WARNING "A link change request failed with "
866 		       "some changes comitted already. Interface %s may "
867 		       "have been left with an inconsistent configuration, "
868 		       "please check.\n", dev->name);
869 
870 	if (send_addr_notify)
871 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
872 
873 	if (modified)
874 		netdev_state_change(dev);
875 
876 	return err;
877 }
878 
879 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
880 {
881 	struct net *net = skb->sk->sk_net;
882 	struct ifinfomsg *ifm;
883 	struct net_device *dev;
884 	int err;
885 	struct nlattr *tb[IFLA_MAX+1];
886 	char ifname[IFNAMSIZ];
887 
888 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
889 	if (err < 0)
890 		goto errout;
891 
892 	if (tb[IFLA_IFNAME])
893 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
894 	else
895 		ifname[0] = '\0';
896 
897 	err = -EINVAL;
898 	ifm = nlmsg_data(nlh);
899 	if (ifm->ifi_index > 0)
900 		dev = dev_get_by_index(net, ifm->ifi_index);
901 	else if (tb[IFLA_IFNAME])
902 		dev = dev_get_by_name(net, ifname);
903 	else
904 		goto errout;
905 
906 	if (dev == NULL) {
907 		err = -ENODEV;
908 		goto errout;
909 	}
910 
911 	if (tb[IFLA_ADDRESS] &&
912 	    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
913 		goto errout_dev;
914 
915 	if (tb[IFLA_BROADCAST] &&
916 	    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
917 		goto errout_dev;
918 
919 	err = do_setlink(dev, ifm, tb, ifname, 0);
920 errout_dev:
921 	dev_put(dev);
922 errout:
923 	return err;
924 }
925 
926 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
927 {
928 	struct net *net = skb->sk->sk_net;
929 	const struct rtnl_link_ops *ops;
930 	struct net_device *dev;
931 	struct ifinfomsg *ifm;
932 	char ifname[IFNAMSIZ];
933 	struct nlattr *tb[IFLA_MAX+1];
934 	int err;
935 
936 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
937 	if (err < 0)
938 		return err;
939 
940 	if (tb[IFLA_IFNAME])
941 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
942 
943 	ifm = nlmsg_data(nlh);
944 	if (ifm->ifi_index > 0)
945 		dev = __dev_get_by_index(net, ifm->ifi_index);
946 	else if (tb[IFLA_IFNAME])
947 		dev = __dev_get_by_name(net, ifname);
948 	else
949 		return -EINVAL;
950 
951 	if (!dev)
952 		return -ENODEV;
953 
954 	ops = dev->rtnl_link_ops;
955 	if (!ops)
956 		return -EOPNOTSUPP;
957 
958 	ops->dellink(dev);
959 	return 0;
960 }
961 
962 struct net_device *rtnl_create_link(struct net *net, char *ifname,
963 		const struct rtnl_link_ops *ops, struct nlattr *tb[])
964 {
965 	int err;
966 	struct net_device *dev;
967 
968 	err = -ENOMEM;
969 	dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
970 	if (!dev)
971 		goto err;
972 
973 	if (strchr(dev->name, '%')) {
974 		err = dev_alloc_name(dev, dev->name);
975 		if (err < 0)
976 			goto err_free;
977 	}
978 
979 	dev->nd_net = net;
980 	dev->rtnl_link_ops = ops;
981 
982 	if (tb[IFLA_MTU])
983 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
984 	if (tb[IFLA_ADDRESS])
985 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
986 				nla_len(tb[IFLA_ADDRESS]));
987 	if (tb[IFLA_BROADCAST])
988 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
989 				nla_len(tb[IFLA_BROADCAST]));
990 	if (tb[IFLA_TXQLEN])
991 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
992 	if (tb[IFLA_OPERSTATE])
993 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]), true);
994 	if (tb[IFLA_LINKMODE])
995 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
996 
997 	return dev;
998 
999 err_free:
1000 	free_netdev(dev);
1001 err:
1002 	return ERR_PTR(err);
1003 }
1004 
1005 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1006 {
1007 	struct net *net = skb->sk->sk_net;
1008 	const struct rtnl_link_ops *ops;
1009 	struct net_device *dev;
1010 	struct ifinfomsg *ifm;
1011 	char kind[MODULE_NAME_LEN];
1012 	char ifname[IFNAMSIZ];
1013 	struct nlattr *tb[IFLA_MAX+1];
1014 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1015 	int err;
1016 
1017 #ifdef CONFIG_KMOD
1018 replay:
1019 #endif
1020 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1021 	if (err < 0)
1022 		return err;
1023 
1024 	if (tb[IFLA_IFNAME])
1025 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1026 	else
1027 		ifname[0] = '\0';
1028 
1029 	ifm = nlmsg_data(nlh);
1030 	if (ifm->ifi_index > 0)
1031 		dev = __dev_get_by_index(net, ifm->ifi_index);
1032 	else if (ifname[0])
1033 		dev = __dev_get_by_name(net, ifname);
1034 	else
1035 		dev = NULL;
1036 
1037 	if (tb[IFLA_LINKINFO]) {
1038 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1039 				       tb[IFLA_LINKINFO], ifla_info_policy);
1040 		if (err < 0)
1041 			return err;
1042 	} else
1043 		memset(linkinfo, 0, sizeof(linkinfo));
1044 
1045 	if (linkinfo[IFLA_INFO_KIND]) {
1046 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1047 		ops = rtnl_link_ops_get(kind);
1048 	} else {
1049 		kind[0] = '\0';
1050 		ops = NULL;
1051 	}
1052 
1053 	if (1) {
1054 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1055 
1056 		if (ops) {
1057 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1058 				err = nla_parse_nested(attr, ops->maxtype,
1059 						       linkinfo[IFLA_INFO_DATA],
1060 						       ops->policy);
1061 				if (err < 0)
1062 					return err;
1063 				data = attr;
1064 			}
1065 			if (ops->validate) {
1066 				err = ops->validate(tb, data);
1067 				if (err < 0)
1068 					return err;
1069 			}
1070 		}
1071 
1072 		if (dev) {
1073 			int modified = 0;
1074 
1075 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1076 				return -EEXIST;
1077 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1078 				return -EOPNOTSUPP;
1079 
1080 			if (linkinfo[IFLA_INFO_DATA]) {
1081 				if (!ops || ops != dev->rtnl_link_ops ||
1082 				    !ops->changelink)
1083 					return -EOPNOTSUPP;
1084 
1085 				err = ops->changelink(dev, tb, data);
1086 				if (err < 0)
1087 					return err;
1088 				modified = 1;
1089 			}
1090 
1091 			return do_setlink(dev, ifm, tb, ifname, modified);
1092 		}
1093 
1094 		if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1095 			return -ENODEV;
1096 
1097 		if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1098 			return -EOPNOTSUPP;
1099 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1100 			return -EOPNOTSUPP;
1101 
1102 		if (!ops) {
1103 #ifdef CONFIG_KMOD
1104 			if (kind[0]) {
1105 				__rtnl_unlock();
1106 				request_module("rtnl-link-%s", kind);
1107 				rtnl_lock();
1108 				ops = rtnl_link_ops_get(kind);
1109 				if (ops)
1110 					goto replay;
1111 			}
1112 #endif
1113 			return -EOPNOTSUPP;
1114 		}
1115 
1116 		if (!ifname[0])
1117 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1118 
1119 		dev = rtnl_create_link(net, ifname, ops, tb);
1120 
1121 		if (IS_ERR(dev))
1122 			err = PTR_ERR(dev);
1123 		else if (ops->newlink)
1124 			err = ops->newlink(dev, tb, data);
1125 		else
1126 			err = register_netdevice(dev);
1127 
1128 		if (err < 0 && !IS_ERR(dev))
1129 			free_netdev(dev);
1130 		return err;
1131 	}
1132 }
1133 
1134 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1135 {
1136 	struct net *net = skb->sk->sk_net;
1137 	struct ifinfomsg *ifm;
1138 	struct nlattr *tb[IFLA_MAX+1];
1139 	struct net_device *dev = NULL;
1140 	struct sk_buff *nskb;
1141 	int err;
1142 
1143 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1144 	if (err < 0)
1145 		return err;
1146 
1147 	ifm = nlmsg_data(nlh);
1148 	if (ifm->ifi_index > 0) {
1149 		dev = dev_get_by_index(net, ifm->ifi_index);
1150 		if (dev == NULL)
1151 			return -ENODEV;
1152 	} else
1153 		return -EINVAL;
1154 
1155 	nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1156 	if (nskb == NULL) {
1157 		err = -ENOBUFS;
1158 		goto errout;
1159 	}
1160 
1161 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1162 			       nlh->nlmsg_seq, 0, 0);
1163 	if (err < 0) {
1164 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1165 		WARN_ON(err == -EMSGSIZE);
1166 		kfree_skb(nskb);
1167 		goto errout;
1168 	}
1169 	err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1170 errout:
1171 	dev_put(dev);
1172 
1173 	return err;
1174 }
1175 
1176 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1177 {
1178 	int idx;
1179 	int s_idx = cb->family;
1180 
1181 	if (s_idx == 0)
1182 		s_idx = 1;
1183 	for (idx=1; idx<NPROTO; idx++) {
1184 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1185 		if (idx < s_idx || idx == PF_PACKET)
1186 			continue;
1187 		if (rtnl_msg_handlers[idx] == NULL ||
1188 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1189 			continue;
1190 		if (idx > s_idx)
1191 			memset(&cb->args[0], 0, sizeof(cb->args));
1192 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1193 			break;
1194 	}
1195 	cb->family = idx;
1196 
1197 	return skb->len;
1198 }
1199 
1200 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1201 {
1202 	struct net *net = dev->nd_net;
1203 	struct sk_buff *skb;
1204 	int err = -ENOBUFS;
1205 
1206 	skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1207 	if (skb == NULL)
1208 		goto errout;
1209 
1210 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1211 	if (err < 0) {
1212 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1213 		WARN_ON(err == -EMSGSIZE);
1214 		kfree_skb(skb);
1215 		goto errout;
1216 	}
1217 	err = rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1218 errout:
1219 	if (err < 0)
1220 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1221 }
1222 
1223 /* Protected by RTNL sempahore.  */
1224 static struct rtattr **rta_buf;
1225 static int rtattr_max;
1226 
1227 /* Process one rtnetlink message. */
1228 
1229 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1230 {
1231 	struct net *net = skb->sk->sk_net;
1232 	rtnl_doit_func doit;
1233 	int sz_idx, kind;
1234 	int min_len;
1235 	int family;
1236 	int type;
1237 	int err;
1238 
1239 	type = nlh->nlmsg_type;
1240 	if (type > RTM_MAX)
1241 		return -EOPNOTSUPP;
1242 
1243 	type -= RTM_BASE;
1244 
1245 	/* All the messages must have at least 1 byte length */
1246 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1247 		return 0;
1248 
1249 	family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1250 	if (family >= NPROTO)
1251 		return -EAFNOSUPPORT;
1252 
1253 	sz_idx = type>>2;
1254 	kind = type&3;
1255 
1256 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1257 		return -EPERM;
1258 
1259 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1260 		struct sock *rtnl;
1261 		rtnl_dumpit_func dumpit;
1262 
1263 		dumpit = rtnl_get_dumpit(family, type);
1264 		if (dumpit == NULL)
1265 			return -EOPNOTSUPP;
1266 
1267 		__rtnl_unlock();
1268 		rtnl = net->rtnl;
1269 		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1270 		rtnl_lock();
1271 		return err;
1272 	}
1273 
1274 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1275 
1276 	min_len = rtm_min[sz_idx];
1277 	if (nlh->nlmsg_len < min_len)
1278 		return -EINVAL;
1279 
1280 	if (nlh->nlmsg_len > min_len) {
1281 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1282 		struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1283 
1284 		while (RTA_OK(attr, attrlen)) {
1285 			unsigned flavor = attr->rta_type;
1286 			if (flavor) {
1287 				if (flavor > rta_max[sz_idx])
1288 					return -EINVAL;
1289 				rta_buf[flavor-1] = attr;
1290 			}
1291 			attr = RTA_NEXT(attr, attrlen);
1292 		}
1293 	}
1294 
1295 	doit = rtnl_get_doit(family, type);
1296 	if (doit == NULL)
1297 		return -EOPNOTSUPP;
1298 
1299 	return doit(skb, nlh, (void *)&rta_buf[0]);
1300 }
1301 
1302 static void rtnetlink_rcv(struct sk_buff *skb)
1303 {
1304 	rtnl_lock();
1305 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1306 	rtnl_unlock();
1307 }
1308 
1309 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1310 {
1311 	struct net_device *dev = ptr;
1312 
1313 	switch (event) {
1314 	case NETDEV_UNREGISTER:
1315 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1316 		break;
1317 	case NETDEV_REGISTER:
1318 		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1319 		break;
1320 	case NETDEV_UP:
1321 	case NETDEV_DOWN:
1322 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1323 		break;
1324 	case NETDEV_CHANGE:
1325 	case NETDEV_GOING_DOWN:
1326 		break;
1327 	default:
1328 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1329 		break;
1330 	}
1331 	return NOTIFY_DONE;
1332 }
1333 
1334 static struct notifier_block rtnetlink_dev_notifier = {
1335 	.notifier_call	= rtnetlink_event,
1336 };
1337 
1338 
1339 static int rtnetlink_net_init(struct net *net)
1340 {
1341 	struct sock *sk;
1342 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1343 				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1344 	if (!sk)
1345 		return -ENOMEM;
1346 	net->rtnl = sk;
1347 	return 0;
1348 }
1349 
1350 static void rtnetlink_net_exit(struct net *net)
1351 {
1352 	netlink_kernel_release(net->rtnl);
1353 	net->rtnl = NULL;
1354 }
1355 
1356 static struct pernet_operations rtnetlink_net_ops = {
1357 	.init = rtnetlink_net_init,
1358 	.exit = rtnetlink_net_exit,
1359 };
1360 
1361 void __init rtnetlink_init(void)
1362 {
1363 	int i;
1364 
1365 	rtattr_max = 0;
1366 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1367 		if (rta_max[i] > rtattr_max)
1368 			rtattr_max = rta_max[i];
1369 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1370 	if (!rta_buf)
1371 		panic("rtnetlink_init: cannot allocate rta_buf\n");
1372 
1373 	if (register_pernet_subsys(&rtnetlink_net_ops))
1374 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
1375 
1376 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1377 	register_netdevice_notifier(&rtnetlink_dev_notifier);
1378 
1379 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1380 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1381 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1382 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1383 
1384 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1385 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1386 }
1387 
1388 EXPORT_SYMBOL(__rta_fill);
1389 EXPORT_SYMBOL(rtnetlink_put_metrics);
1390 EXPORT_SYMBOL(rtnl_lock);
1391 EXPORT_SYMBOL(rtnl_trylock);
1392 EXPORT_SYMBOL(rtnl_unlock);
1393 EXPORT_SYMBOL(rtnl_unicast);
1394 EXPORT_SYMBOL(rtnl_notify);
1395 EXPORT_SYMBOL(rtnl_set_sk_err);
1396 EXPORT_SYMBOL(rtnl_create_link);
1397 EXPORT_SYMBOL(ifla_policy);
1398