1/*
2 * INET		An implementation of the TCP/IP protocol suite for the LINUX
3 *		operating system.  INET is implemented using the  BSD Socket
4 *		interface as the means of communication with the user level.
5 *
6 *		Routing netlink socket interface: protocol independent part.
7 *
8 * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9 *
10 *		This program is free software; you can redistribute it and/or
11 *		modify it under the terms of the GNU General Public License
12 *		as published by the Free Software Foundation; either version
13 *		2 of the License, or (at your option) any later version.
14 *
15 *	Fixes:
16 *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17 */
18
19#include <linux/errno.h>
20#include <linux/module.h>
21#include <linux/types.h>
22#include <linux/socket.h>
23#include <linux/kernel.h>
24#include <linux/timer.h>
25#include <linux/string.h>
26#include <linux/sockios.h>
27#include <linux/net.h>
28#include <linux/fcntl.h>
29#include <linux/mm.h>
30#include <linux/slab.h>
31#include <linux/interrupt.h>
32#include <linux/capability.h>
33#include <linux/skbuff.h>
34#include <linux/init.h>
35#include <linux/security.h>
36#include <linux/mutex.h>
37#include <linux/if_addr.h>
38
39#include <asm/uaccess.h>
40#include <asm/system.h>
41#include <asm/string.h>
42
43#include <linux/inet.h>
44#include <linux/netdevice.h>
45#include <net/ip.h>
46#include <net/protocol.h>
47#include <net/arp.h>
48#include <net/route.h>
49#include <net/udp.h>
50#include <net/sock.h>
51#include <net/pkt_sched.h>
52#include <net/fib_rules.h>
53#include <net/rtnetlink.h>
54
55struct rtnl_link
56{
57	rtnl_doit_func		doit;
58	rtnl_dumpit_func	dumpit;
59};
60
61static DEFINE_MUTEX(rtnl_mutex);
62static struct sock *rtnl;
63
64void rtnl_lock(void)
65{
66	mutex_lock(&rtnl_mutex);
67}
68
69void __rtnl_unlock(void)
70{
71	mutex_unlock(&rtnl_mutex);
72}
73
74void rtnl_unlock(void)
75{
76	mutex_unlock(&rtnl_mutex);
77	if (rtnl && rtnl->sk_receive_queue.qlen)
78		rtnl->sk_data_ready(rtnl, 0);
79	netdev_run_todo();
80}
81
82int rtnl_trylock(void)
83{
84	return mutex_trylock(&rtnl_mutex);
85}
86
87int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
88{
89	memset(tb, 0, sizeof(struct rtattr*)*maxattr);
90
91	while (RTA_OK(rta, len)) {
92		unsigned flavor = rta->rta_type;
93		if (flavor && flavor <= maxattr)
94			tb[flavor-1] = rta;
95		rta = RTA_NEXT(rta, len);
96	}
97	return 0;
98}
99
100static struct rtnl_link *rtnl_msg_handlers[NPROTO];
101
102static inline int rtm_msgindex(int msgtype)
103{
104	int msgindex = msgtype - RTM_BASE;
105
106	/*
107	 * msgindex < 0 implies someone tried to register a netlink
108	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
109	 * the message type has not been added to linux/rtnetlink.h
110	 */
111	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
112
113	return msgindex;
114}
115
116static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
117{
118	struct rtnl_link *tab;
119
120	tab = rtnl_msg_handlers[protocol];
121	if (tab == NULL || tab[msgindex].doit == NULL)
122		tab = rtnl_msg_handlers[PF_UNSPEC];
123
124	return tab ? tab[msgindex].doit : NULL;
125}
126
127static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
128{
129	struct rtnl_link *tab;
130
131	tab = rtnl_msg_handlers[protocol];
132	if (tab == NULL || tab[msgindex].dumpit == NULL)
133		tab = rtnl_msg_handlers[PF_UNSPEC];
134
135	return tab ? tab[msgindex].dumpit : NULL;
136}
137
138/**
139 * __rtnl_register - Register a rtnetlink message type
140 * @protocol: Protocol family or PF_UNSPEC
141 * @msgtype: rtnetlink message type
142 * @doit: Function pointer called for each request message
143 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
144 *
145 * Registers the specified function pointers (at least one of them has
146 * to be non-NULL) to be called whenever a request message for the
147 * specified protocol family and message type is received.
148 *
149 * The special protocol family PF_UNSPEC may be used to define fallback
150 * function pointers for the case when no entry for the specific protocol
151 * family exists.
152 *
153 * Returns 0 on success or a negative error code.
154 */
155int __rtnl_register(int protocol, int msgtype,
156		    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
157{
158	struct rtnl_link *tab;
159	int msgindex;
160
161	BUG_ON(protocol < 0 || protocol >= NPROTO);
162	msgindex = rtm_msgindex(msgtype);
163
164	tab = rtnl_msg_handlers[protocol];
165	if (tab == NULL) {
166		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
167		if (tab == NULL)
168			return -ENOBUFS;
169
170		rtnl_msg_handlers[protocol] = tab;
171	}
172
173	if (doit)
174		tab[msgindex].doit = doit;
175
176	if (dumpit)
177		tab[msgindex].dumpit = dumpit;
178
179	return 0;
180}
181
182EXPORT_SYMBOL_GPL(__rtnl_register);
183
184/**
185 * rtnl_register - Register a rtnetlink message type
186 *
187 * Identical to __rtnl_register() but panics on failure. This is useful
188 * as failure of this function is very unlikely, it can only happen due
189 * to lack of memory when allocating the chain to store all message
190 * handlers for a protocol. Meant for use in init functions where lack
191 * of memory implies no sense in continueing.
192 */
193void rtnl_register(int protocol, int msgtype,
194		   rtnl_doit_func doit, rtnl_dumpit_func dumpit)
195{
196	if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
197		panic("Unable to register rtnetlink message handler, "
198		      "protocol = %d, message type = %d\n",
199		      protocol, msgtype);
200}
201
202EXPORT_SYMBOL_GPL(rtnl_register);
203
204/**
205 * rtnl_unregister - Unregister a rtnetlink message type
206 * @protocol: Protocol family or PF_UNSPEC
207 * @msgtype: rtnetlink message type
208 *
209 * Returns 0 on success or a negative error code.
210 */
211int rtnl_unregister(int protocol, int msgtype)
212{
213	int msgindex;
214
215	BUG_ON(protocol < 0 || protocol >= NPROTO);
216	msgindex = rtm_msgindex(msgtype);
217
218	if (rtnl_msg_handlers[protocol] == NULL)
219		return -ENOENT;
220
221	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
222	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
223
224	return 0;
225}
226
227EXPORT_SYMBOL_GPL(rtnl_unregister);
228
229/**
230 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
231 * @protocol : Protocol family or PF_UNSPEC
232 *
233 * Identical to calling rtnl_unregster() for all registered message types
234 * of a certain protocol family.
235 */
236void rtnl_unregister_all(int protocol)
237{
238	BUG_ON(protocol < 0 || protocol >= NPROTO);
239
240	kfree(rtnl_msg_handlers[protocol]);
241	rtnl_msg_handlers[protocol] = NULL;
242}
243
244EXPORT_SYMBOL_GPL(rtnl_unregister_all);
245
246static const int rtm_min[RTM_NR_FAMILIES] =
247{
248	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
249	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
250	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
251	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
252	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
253	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
254	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
255	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
256	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
257	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
258};
259
260static const int rta_max[RTM_NR_FAMILIES] =
261{
262	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
263	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
264	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
265	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
266	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
267	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
268	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
269	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
270};
271
272void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
273{
274	struct rtattr *rta;
275	int size = RTA_LENGTH(attrlen);
276
277	rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
278	rta->rta_type = attrtype;
279	rta->rta_len = size;
280	memcpy(RTA_DATA(rta), data, attrlen);
281	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
282}
283
284size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
285{
286	size_t ret = RTA_PAYLOAD(rta);
287	char *src = RTA_DATA(rta);
288
289	if (ret > 0 && src[ret - 1] == '\0')
290		ret--;
291	if (size > 0) {
292		size_t len = (ret >= size) ? size - 1 : ret;
293		memset(dest, 0, size);
294		memcpy(dest, src, len);
295	}
296	return ret;
297}
298
299int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
300{
301	int err = 0;
302
303	NETLINK_CB(skb).dst_group = group;
304	if (echo)
305		atomic_inc(&skb->users);
306	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
307	if (echo)
308		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
309	return err;
310}
311
312int rtnl_unicast(struct sk_buff *skb, u32 pid)
313{
314	return nlmsg_unicast(rtnl, skb, pid);
315}
316
317int rtnl_notify(struct sk_buff *skb, u32 pid, u32 group,
318		struct nlmsghdr *nlh, gfp_t flags)
319{
320	int report = 0;
321
322	if (nlh)
323		report = nlmsg_report(nlh);
324
325	return nlmsg_notify(rtnl, skb, pid, group, report, flags);
326}
327
328void rtnl_set_sk_err(u32 group, int error)
329{
330	netlink_set_err(rtnl, 0, group, error);
331}
332
333int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
334{
335	struct nlattr *mx;
336	int i, valid = 0;
337
338	mx = nla_nest_start(skb, RTA_METRICS);
339	if (mx == NULL)
340		return -ENOBUFS;
341
342	for (i = 0; i < RTAX_MAX; i++) {
343		if (metrics[i]) {
344			valid++;
345			NLA_PUT_U32(skb, i+1, metrics[i]);
346		}
347	}
348
349	if (!valid) {
350		nla_nest_cancel(skb, mx);
351		return 0;
352	}
353
354	return nla_nest_end(skb, mx);
355
356nla_put_failure:
357	return nla_nest_cancel(skb, mx);
358}
359
360int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
361		       u32 ts, u32 tsage, long expires, u32 error)
362{
363	struct rta_cacheinfo ci = {
364		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
365		.rta_used = dst->__use,
366		.rta_clntref = atomic_read(&(dst->__refcnt)),
367		.rta_error = error,
368		.rta_id =  id,
369		.rta_ts = ts,
370		.rta_tsage = tsage,
371	};
372
373	if (expires)
374		ci.rta_expires = jiffies_to_clock_t(expires);
375
376	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
377}
378
379EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
380
381static void set_operstate(struct net_device *dev, unsigned char transition)
382{
383	unsigned char operstate = dev->operstate;
384
385	switch(transition) {
386	case IF_OPER_UP:
387		if ((operstate == IF_OPER_DORMANT ||
388		     operstate == IF_OPER_UNKNOWN) &&
389		    !netif_dormant(dev))
390			operstate = IF_OPER_UP;
391		break;
392
393	case IF_OPER_DORMANT:
394		if (operstate == IF_OPER_UP ||
395		    operstate == IF_OPER_UNKNOWN)
396			operstate = IF_OPER_DORMANT;
397		break;
398	}
399
400	if (dev->operstate != operstate) {
401		write_lock_bh(&dev_base_lock);
402		dev->operstate = operstate;
403		write_unlock_bh(&dev_base_lock);
404		netdev_state_change(dev);
405	}
406}
407
408static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
409				 struct net_device_stats *b)
410{
411	a->rx_packets = b->rx_packets;
412	a->tx_packets = b->tx_packets;
413	a->rx_bytes = b->rx_bytes;
414	a->tx_bytes = b->tx_bytes;
415	a->rx_errors = b->rx_errors;
416	a->tx_errors = b->tx_errors;
417	a->rx_dropped = b->rx_dropped;
418	a->tx_dropped = b->tx_dropped;
419
420	a->multicast = b->multicast;
421	a->collisions = b->collisions;
422
423	a->rx_length_errors = b->rx_length_errors;
424	a->rx_over_errors = b->rx_over_errors;
425	a->rx_crc_errors = b->rx_crc_errors;
426	a->rx_frame_errors = b->rx_frame_errors;
427	a->rx_fifo_errors = b->rx_fifo_errors;
428	a->rx_missed_errors = b->rx_missed_errors;
429
430	a->tx_aborted_errors = b->tx_aborted_errors;
431	a->tx_carrier_errors = b->tx_carrier_errors;
432	a->tx_fifo_errors = b->tx_fifo_errors;
433	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
434	a->tx_window_errors = b->tx_window_errors;
435
436	a->rx_compressed = b->rx_compressed;
437	a->tx_compressed = b->tx_compressed;
438};
439
440static inline size_t if_nlmsg_size(void)
441{
442	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
443	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
444	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
445	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
446	       + nla_total_size(sizeof(struct rtnl_link_stats))
447	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
448	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
449	       + nla_total_size(4) /* IFLA_TXQLEN */
450	       + nla_total_size(4) /* IFLA_WEIGHT */
451	       + nla_total_size(4) /* IFLA_MTU */
452	       + nla_total_size(4) /* IFLA_LINK */
453	       + nla_total_size(4) /* IFLA_MASTER */
454	       + nla_total_size(1) /* IFLA_OPERSTATE */
455	       + nla_total_size(1); /* IFLA_LINKMODE */
456}
457
458static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
459			    int type, u32 pid, u32 seq, u32 change,
460			    unsigned int flags)
461{
462	struct ifinfomsg *ifm;
463	struct nlmsghdr *nlh;
464
465	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
466	if (nlh == NULL)
467		return -EMSGSIZE;
468
469	ifm = nlmsg_data(nlh);
470	ifm->ifi_family = AF_UNSPEC;
471	ifm->__ifi_pad = 0;
472	ifm->ifi_type = dev->type;
473	ifm->ifi_index = dev->ifindex;
474	ifm->ifi_flags = dev_get_flags(dev);
475	ifm->ifi_change = change;
476
477	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
478	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
479	NLA_PUT_U32(skb, IFLA_WEIGHT, dev->weight);
480	NLA_PUT_U8(skb, IFLA_OPERSTATE,
481		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
482	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
483	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
484
485	if (dev->ifindex != dev->iflink)
486		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
487
488	if (dev->master)
489		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
490
491	if (dev->qdisc_sleeping)
492		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
493
494	if (1) {
495		struct rtnl_link_ifmap map = {
496			.mem_start   = dev->mem_start,
497			.mem_end     = dev->mem_end,
498			.base_addr   = dev->base_addr,
499			.irq         = dev->irq,
500			.dma         = dev->dma,
501			.port        = dev->if_port,
502		};
503		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
504	}
505
506	if (dev->addr_len) {
507		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
508		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
509	}
510
511	if (dev->get_stats) {
512		struct net_device_stats *stats = dev->get_stats(dev);
513		if (stats) {
514			struct nlattr *attr;
515
516			attr = nla_reserve(skb, IFLA_STATS,
517					   sizeof(struct rtnl_link_stats));
518			if (attr == NULL)
519				goto nla_put_failure;
520
521			copy_rtnl_link_stats(nla_data(attr), stats);
522		}
523	}
524
525	return nlmsg_end(skb, nlh);
526
527nla_put_failure:
528	nlmsg_cancel(skb, nlh);
529	return -EMSGSIZE;
530}
531
532static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
533{
534	int idx;
535	int s_idx = cb->args[0];
536	struct net_device *dev;
537
538	idx = 0;
539	for_each_netdev(dev) {
540		if (idx < s_idx)
541			goto cont;
542		if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
543				     NETLINK_CB(cb->skb).pid,
544				     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
545			break;
546cont:
547		idx++;
548	}
549	cb->args[0] = idx;
550
551	return skb->len;
552}
553
554static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
555	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
556	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
557	[IFLA_MTU]		= { .type = NLA_U32 },
558	[IFLA_TXQLEN]		= { .type = NLA_U32 },
559	[IFLA_WEIGHT]		= { .type = NLA_U32 },
560	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
561	[IFLA_LINKMODE]		= { .type = NLA_U8 },
562};
563
564static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
565{
566	struct ifinfomsg *ifm;
567	struct net_device *dev;
568	int err, send_addr_notify = 0, modified = 0;
569	struct nlattr *tb[IFLA_MAX+1];
570	char ifname[IFNAMSIZ];
571
572	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
573	if (err < 0)
574		goto errout;
575
576	if (tb[IFLA_IFNAME])
577		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
578	else
579		ifname[0] = '\0';
580
581	err = -EINVAL;
582	ifm = nlmsg_data(nlh);
583	if (ifm->ifi_index > 0)
584		dev = dev_get_by_index(ifm->ifi_index);
585	else if (tb[IFLA_IFNAME])
586		dev = dev_get_by_name(ifname);
587	else
588		goto errout;
589
590	if (dev == NULL) {
591		err = -ENODEV;
592		goto errout;
593	}
594
595	if (tb[IFLA_ADDRESS] &&
596	    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
597		goto errout_dev;
598
599	if (tb[IFLA_BROADCAST] &&
600	    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
601		goto errout_dev;
602
603	if (tb[IFLA_MAP]) {
604		struct rtnl_link_ifmap *u_map;
605		struct ifmap k_map;
606
607		if (!dev->set_config) {
608			err = -EOPNOTSUPP;
609			goto errout_dev;
610		}
611
612		if (!netif_device_present(dev)) {
613			err = -ENODEV;
614			goto errout_dev;
615		}
616
617		u_map = nla_data(tb[IFLA_MAP]);
618		k_map.mem_start = (unsigned long) u_map->mem_start;
619		k_map.mem_end = (unsigned long) u_map->mem_end;
620		k_map.base_addr = (unsigned short) u_map->base_addr;
621		k_map.irq = (unsigned char) u_map->irq;
622		k_map.dma = (unsigned char) u_map->dma;
623		k_map.port = (unsigned char) u_map->port;
624
625		err = dev->set_config(dev, &k_map);
626		if (err < 0)
627			goto errout_dev;
628
629		modified = 1;
630	}
631
632	if (tb[IFLA_ADDRESS]) {
633		struct sockaddr *sa;
634		int len;
635
636		if (!dev->set_mac_address) {
637			err = -EOPNOTSUPP;
638			goto errout_dev;
639		}
640
641		if (!netif_device_present(dev)) {
642			err = -ENODEV;
643			goto errout_dev;
644		}
645
646		len = sizeof(sa_family_t) + dev->addr_len;
647		sa = kmalloc(len, GFP_KERNEL);
648		if (!sa) {
649			err = -ENOMEM;
650			goto errout_dev;
651		}
652		sa->sa_family = dev->type;
653		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
654		       dev->addr_len);
655		err = dev->set_mac_address(dev, sa);
656		kfree(sa);
657		if (err)
658			goto errout_dev;
659		send_addr_notify = 1;
660		modified = 1;
661	}
662
663	if (tb[IFLA_MTU]) {
664		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
665		if (err < 0)
666			goto errout_dev;
667		modified = 1;
668	}
669
670	/*
671	 * Interface selected by interface index but interface
672	 * name provided implies that a name change has been
673	 * requested.
674	 */
675	if (ifm->ifi_index > 0 && ifname[0]) {
676		err = dev_change_name(dev, ifname);
677		if (err < 0)
678			goto errout_dev;
679		modified = 1;
680	}
681
682	if (tb[IFLA_BROADCAST]) {
683		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
684		send_addr_notify = 1;
685	}
686
687
688	if (ifm->ifi_flags || ifm->ifi_change) {
689		unsigned int flags = ifm->ifi_flags;
690
691		/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
692		if (ifm->ifi_change)
693			flags = (flags & ifm->ifi_change) |
694				(dev->flags & ~ifm->ifi_change);
695		dev_change_flags(dev, flags);
696	}
697
698	if (tb[IFLA_TXQLEN])
699		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
700
701	if (tb[IFLA_WEIGHT])
702		dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
703
704	if (tb[IFLA_OPERSTATE])
705		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
706
707	if (tb[IFLA_LINKMODE]) {
708		write_lock_bh(&dev_base_lock);
709		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
710		write_unlock_bh(&dev_base_lock);
711	}
712
713	err = 0;
714
715errout_dev:
716	if (err < 0 && modified && net_ratelimit())
717		printk(KERN_WARNING "A link change request failed with "
718		       "some changes comitted already. Interface %s may "
719		       "have been left with an inconsistent configuration, "
720		       "please check.\n", dev->name);
721
722	if (send_addr_notify)
723		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
724
725	dev_put(dev);
726errout:
727	return err;
728}
729
730static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
731{
732	struct ifinfomsg *ifm;
733	struct nlattr *tb[IFLA_MAX+1];
734	struct net_device *dev = NULL;
735	struct sk_buff *nskb;
736	int err;
737
738	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
739	if (err < 0)
740		return err;
741
742	ifm = nlmsg_data(nlh);
743	if (ifm->ifi_index > 0) {
744		dev = dev_get_by_index(ifm->ifi_index);
745		if (dev == NULL)
746			return -ENODEV;
747	} else
748		return -EINVAL;
749
750	nskb = nlmsg_new(if_nlmsg_size(), GFP_KERNEL);
751	if (nskb == NULL) {
752		err = -ENOBUFS;
753		goto errout;
754	}
755
756	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
757			       nlh->nlmsg_seq, 0, 0);
758	if (err < 0) {
759		/* -EMSGSIZE implies BUG in if_nlmsg_size */
760		WARN_ON(err == -EMSGSIZE);
761		kfree_skb(nskb);
762		goto errout;
763	}
764	err = rtnl_unicast(nskb, NETLINK_CB(skb).pid);
765errout:
766	dev_put(dev);
767
768	return err;
769}
770
771static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
772{
773	int idx;
774	int s_idx = cb->family;
775
776	if (s_idx == 0)
777		s_idx = 1;
778	for (idx=1; idx<NPROTO; idx++) {
779		int type = cb->nlh->nlmsg_type-RTM_BASE;
780		if (idx < s_idx || idx == PF_PACKET)
781			continue;
782		if (rtnl_msg_handlers[idx] == NULL ||
783		    rtnl_msg_handlers[idx][type].dumpit == NULL)
784			continue;
785		if (idx > s_idx)
786			memset(&cb->args[0], 0, sizeof(cb->args));
787		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
788			break;
789	}
790	cb->family = idx;
791
792	return skb->len;
793}
794
795void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
796{
797	struct sk_buff *skb;
798	int err = -ENOBUFS;
799
800	skb = nlmsg_new(if_nlmsg_size(), GFP_KERNEL);
801	if (skb == NULL)
802		goto errout;
803
804	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
805	if (err < 0) {
806		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
807		WARN_ON(err == -EMSGSIZE);
808		kfree_skb(skb);
809		goto errout;
810	}
811	err = rtnl_notify(skb, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
812errout:
813	if (err < 0)
814		rtnl_set_sk_err(RTNLGRP_LINK, err);
815}
816
817/* Protected by RTNL sempahore.  */
818static struct rtattr **rta_buf;
819static int rtattr_max;
820
821/* Process one rtnetlink message. */
822
823static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
824{
825	rtnl_doit_func doit;
826	int sz_idx, kind;
827	int min_len;
828	int family;
829	int type;
830	int err;
831
832	type = nlh->nlmsg_type;
833	if (type > RTM_MAX)
834		return -EOPNOTSUPP;
835
836	type -= RTM_BASE;
837
838	/* All the messages must have at least 1 byte length */
839	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
840		return 0;
841
842	family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
843	if (family >= NPROTO)
844		return -EAFNOSUPPORT;
845
846	sz_idx = type>>2;
847	kind = type&3;
848
849	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
850		return -EPERM;
851
852	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
853		rtnl_dumpit_func dumpit;
854
855		dumpit = rtnl_get_dumpit(family, type);
856		if (dumpit == NULL)
857			return -EOPNOTSUPP;
858
859		__rtnl_unlock();
860		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
861		rtnl_lock();
862		return err;
863	}
864
865	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
866
867	min_len = rtm_min[sz_idx];
868	if (nlh->nlmsg_len < min_len)
869		return -EINVAL;
870
871	if (nlh->nlmsg_len > min_len) {
872		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
873		struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
874
875		while (RTA_OK(attr, attrlen)) {
876			unsigned flavor = attr->rta_type;
877			if (flavor) {
878				if (flavor > rta_max[sz_idx])
879					return -EINVAL;
880				rta_buf[flavor-1] = attr;
881			}
882			attr = RTA_NEXT(attr, attrlen);
883		}
884	}
885
886	doit = rtnl_get_doit(family, type);
887	if (doit == NULL)
888		return -EOPNOTSUPP;
889
890	return doit(skb, nlh, (void *)&rta_buf[0]);
891}
892
893static void rtnetlink_rcv(struct sock *sk, int len)
894{
895	unsigned int qlen = 0;
896
897	do {
898		mutex_lock(&rtnl_mutex);
899		netlink_run_queue(sk, &qlen, &rtnetlink_rcv_msg);
900		mutex_unlock(&rtnl_mutex);
901
902		netdev_run_todo();
903	} while (qlen);
904}
905
906static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
907{
908	struct net_device *dev = ptr;
909	switch (event) {
910	case NETDEV_UNREGISTER:
911		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
912		break;
913	case NETDEV_REGISTER:
914		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
915		break;
916	case NETDEV_UP:
917	case NETDEV_DOWN:
918		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
919		break;
920	case NETDEV_CHANGE:
921	case NETDEV_GOING_DOWN:
922		break;
923	default:
924		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
925		break;
926	}
927	return NOTIFY_DONE;
928}
929
930static struct notifier_block rtnetlink_dev_notifier = {
931	.notifier_call	= rtnetlink_event,
932};
933
934void __init rtnetlink_init(void)
935{
936	int i;
937
938	rtattr_max = 0;
939	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
940		if (rta_max[i] > rtattr_max)
941			rtattr_max = rta_max[i];
942	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
943	if (!rta_buf)
944		panic("rtnetlink_init: cannot allocate rta_buf\n");
945
946	rtnl = netlink_kernel_create(NETLINK_ROUTE, RTNLGRP_MAX, rtnetlink_rcv,
947				     &rtnl_mutex, THIS_MODULE);
948	if (rtnl == NULL)
949		panic("rtnetlink_init: cannot initialize rtnetlink\n");
950	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
951	register_netdevice_notifier(&rtnetlink_dev_notifier);
952
953	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
954	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
955
956	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
957	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
958}
959
960EXPORT_SYMBOL(__rta_fill);
961EXPORT_SYMBOL(rtattr_strlcpy);
962EXPORT_SYMBOL(rtattr_parse);
963EXPORT_SYMBOL(rtnetlink_put_metrics);
964EXPORT_SYMBOL(rtnl_lock);
965EXPORT_SYMBOL(rtnl_trylock);
966EXPORT_SYMBOL(rtnl_unlock);
967EXPORT_SYMBOL(rtnl_unicast);
968EXPORT_SYMBOL(rtnl_notify);
969EXPORT_SYMBOL(rtnl_set_sk_err);
970