rtnetlink.c revision afdc3238ec948531205f5c5f77d2de7bae519c71 
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
100int rtattr_parse_nested_compat(struct rtattr *tb[], int maxattr,
101			       struct rtattr *rta, void **data, int len)
102{
103	if (RTA_PAYLOAD(rta) < len)
104		return -1;
105	*data = RTA_DATA(rta);
106
107	if (RTA_PAYLOAD(rta) >= RTA_ALIGN(len) + sizeof(struct rtattr)) {
108		rta = RTA_DATA(rta) + RTA_ALIGN(len);
109		return rtattr_parse_nested(tb, maxattr, rta);
110	}
111	memset(tb, 0, sizeof(struct rtattr *) * maxattr);
112	return 0;
113}
114
115static struct rtnl_link *rtnl_msg_handlers[NPROTO];
116
117static inline int rtm_msgindex(int msgtype)
118{
119	int msgindex = msgtype - RTM_BASE;
120
121	/*
122	 * msgindex < 0 implies someone tried to register a netlink
123	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
124	 * the message type has not been added to linux/rtnetlink.h
125	 */
126	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
127
128	return msgindex;
129}
130
131static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
132{
133	struct rtnl_link *tab;
134
135	tab = rtnl_msg_handlers[protocol];
136	if (tab == NULL || tab[msgindex].doit == NULL)
137		tab = rtnl_msg_handlers[PF_UNSPEC];
138
139	return tab ? tab[msgindex].doit : NULL;
140}
141
142static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
143{
144	struct rtnl_link *tab;
145
146	tab = rtnl_msg_handlers[protocol];
147	if (tab == NULL || tab[msgindex].dumpit == NULL)
148		tab = rtnl_msg_handlers[PF_UNSPEC];
149
150	return tab ? tab[msgindex].dumpit : NULL;
151}
152
153/**
154 * __rtnl_register - Register a rtnetlink message type
155 * @protocol: Protocol family or PF_UNSPEC
156 * @msgtype: rtnetlink message type
157 * @doit: Function pointer called for each request message
158 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
159 *
160 * Registers the specified function pointers (at least one of them has
161 * to be non-NULL) to be called whenever a request message for the
162 * specified protocol family and message type is received.
163 *
164 * The special protocol family PF_UNSPEC may be used to define fallback
165 * function pointers for the case when no entry for the specific protocol
166 * family exists.
167 *
168 * Returns 0 on success or a negative error code.
169 */
170int __rtnl_register(int protocol, int msgtype,
171		    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
172{
173	struct rtnl_link *tab;
174	int msgindex;
175
176	BUG_ON(protocol < 0 || protocol >= NPROTO);
177	msgindex = rtm_msgindex(msgtype);
178
179	tab = rtnl_msg_handlers[protocol];
180	if (tab == NULL) {
181		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
182		if (tab == NULL)
183			return -ENOBUFS;
184
185		rtnl_msg_handlers[protocol] = tab;
186	}
187
188	if (doit)
189		tab[msgindex].doit = doit;
190
191	if (dumpit)
192		tab[msgindex].dumpit = dumpit;
193
194	return 0;
195}
196
197EXPORT_SYMBOL_GPL(__rtnl_register);
198
199/**
200 * rtnl_register - Register a rtnetlink message type
201 *
202 * Identical to __rtnl_register() but panics on failure. This is useful
203 * as failure of this function is very unlikely, it can only happen due
204 * to lack of memory when allocating the chain to store all message
205 * handlers for a protocol. Meant for use in init functions where lack
206 * of memory implies no sense in continueing.
207 */
208void rtnl_register(int protocol, int msgtype,
209		   rtnl_doit_func doit, rtnl_dumpit_func dumpit)
210{
211	if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
212		panic("Unable to register rtnetlink message handler, "
213		      "protocol = %d, message type = %d\n",
214		      protocol, msgtype);
215}
216
217EXPORT_SYMBOL_GPL(rtnl_register);
218
219/**
220 * rtnl_unregister - Unregister a rtnetlink message type
221 * @protocol: Protocol family or PF_UNSPEC
222 * @msgtype: rtnetlink message type
223 *
224 * Returns 0 on success or a negative error code.
225 */
226int rtnl_unregister(int protocol, int msgtype)
227{
228	int msgindex;
229
230	BUG_ON(protocol < 0 || protocol >= NPROTO);
231	msgindex = rtm_msgindex(msgtype);
232
233	if (rtnl_msg_handlers[protocol] == NULL)
234		return -ENOENT;
235
236	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
237	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
238
239	return 0;
240}
241
242EXPORT_SYMBOL_GPL(rtnl_unregister);
243
244/**
245 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
246 * @protocol : Protocol family or PF_UNSPEC
247 *
248 * Identical to calling rtnl_unregster() for all registered message types
249 * of a certain protocol family.
250 */
251void rtnl_unregister_all(int protocol)
252{
253	BUG_ON(protocol < 0 || protocol >= NPROTO);
254
255	kfree(rtnl_msg_handlers[protocol]);
256	rtnl_msg_handlers[protocol] = NULL;
257}
258
259EXPORT_SYMBOL_GPL(rtnl_unregister_all);
260
261static LIST_HEAD(link_ops);
262
263/**
264 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
265 * @ops: struct rtnl_link_ops * to register
266 *
267 * The caller must hold the rtnl_mutex. This function should be used
268 * by drivers that create devices during module initialization. It
269 * must be called before registering the devices.
270 *
271 * Returns 0 on success or a negative error code.
272 */
273int __rtnl_link_register(struct rtnl_link_ops *ops)
274{
275	list_add_tail(&ops->list, &link_ops);
276	return 0;
277}
278
279EXPORT_SYMBOL_GPL(__rtnl_link_register);
280
281/**
282 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
283 * @ops: struct rtnl_link_ops * to register
284 *
285 * Returns 0 on success or a negative error code.
286 */
287int rtnl_link_register(struct rtnl_link_ops *ops)
288{
289	int err;
290
291	rtnl_lock();
292	err = __rtnl_link_register(ops);
293	rtnl_unlock();
294	return err;
295}
296
297EXPORT_SYMBOL_GPL(rtnl_link_register);
298
299/**
300 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
301 * @ops: struct rtnl_link_ops * to unregister
302 *
303 * The caller must hold the rtnl_mutex. This function should be used
304 * by drivers that unregister devices during module unloading. It must
305 * be called after unregistering the devices.
306 */
307void __rtnl_link_unregister(struct rtnl_link_ops *ops)
308{
309	list_del(&ops->list);
310}
311
312EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
313
314/**
315 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
316 * @ops: struct rtnl_link_ops * to unregister
317 */
318void rtnl_link_unregister(struct rtnl_link_ops *ops)
319{
320	rtnl_lock();
321	__rtnl_link_unregister(ops);
322	rtnl_unlock();
323}
324
325EXPORT_SYMBOL_GPL(rtnl_link_unregister);
326
327static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
328{
329	const struct rtnl_link_ops *ops;
330
331	list_for_each_entry(ops, &link_ops, list) {
332		if (!strcmp(ops->kind, kind))
333			return ops;
334	}
335	return NULL;
336}
337
338static size_t rtnl_link_get_size(const struct net_device *dev)
339{
340	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
341	size_t size;
342
343	if (!ops)
344		return 0;
345
346	size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
347	       nlmsg_total_size(strlen(ops->kind) + 1);	 /* IFLA_INFO_KIND */
348
349	if (ops->get_size)
350		/* IFLA_INFO_DATA + nested data */
351		size += nlmsg_total_size(sizeof(struct nlattr)) +
352			ops->get_size(dev);
353
354	if (ops->get_xstats_size)
355		size += ops->get_xstats_size(dev);	/* IFLA_INFO_XSTATS */
356
357	return size;
358}
359
360static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
361{
362	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
363	struct nlattr *linkinfo, *data;
364	int err = -EMSGSIZE;
365
366	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
367	if (linkinfo == NULL)
368		goto out;
369
370	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
371		goto err_cancel_link;
372	if (ops->fill_xstats) {
373		err = ops->fill_xstats(skb, dev);
374		if (err < 0)
375			goto err_cancel_link;
376	}
377	if (ops->fill_info) {
378		data = nla_nest_start(skb, IFLA_INFO_DATA);
379		if (data == NULL)
380			goto err_cancel_link;
381		err = ops->fill_info(skb, dev);
382		if (err < 0)
383			goto err_cancel_data;
384		nla_nest_end(skb, data);
385	}
386
387	nla_nest_end(skb, linkinfo);
388	return 0;
389
390err_cancel_data:
391	nla_nest_cancel(skb, data);
392err_cancel_link:
393	nla_nest_cancel(skb, linkinfo);
394out:
395	return err;
396}
397
398static const int rtm_min[RTM_NR_FAMILIES] =
399{
400	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
401	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
402	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
403	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
404	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
405	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
406	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
407	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
408	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
409	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
410};
411
412static const int rta_max[RTM_NR_FAMILIES] =
413{
414	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
415	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
416	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
417	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
418	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
419	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
420	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
421	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
422};
423
424void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
425{
426	struct rtattr *rta;
427	int size = RTA_LENGTH(attrlen);
428
429	rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
430	rta->rta_type = attrtype;
431	rta->rta_len = size;
432	memcpy(RTA_DATA(rta), data, attrlen);
433	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
434}
435
436size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
437{
438	size_t ret = RTA_PAYLOAD(rta);
439	char *src = RTA_DATA(rta);
440
441	if (ret > 0 && src[ret - 1] == '\0')
442		ret--;
443	if (size > 0) {
444		size_t len = (ret >= size) ? size - 1 : ret;
445		memset(dest, 0, size);
446		memcpy(dest, src, len);
447	}
448	return ret;
449}
450
451int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
452{
453	int err = 0;
454
455	NETLINK_CB(skb).dst_group = group;
456	if (echo)
457		atomic_inc(&skb->users);
458	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
459	if (echo)
460		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
461	return err;
462}
463
464int rtnl_unicast(struct sk_buff *skb, u32 pid)
465{
466	return nlmsg_unicast(rtnl, skb, pid);
467}
468
469int rtnl_notify(struct sk_buff *skb, u32 pid, u32 group,
470		struct nlmsghdr *nlh, gfp_t flags)
471{
472	int report = 0;
473
474	if (nlh)
475		report = nlmsg_report(nlh);
476
477	return nlmsg_notify(rtnl, skb, pid, group, report, flags);
478}
479
480void rtnl_set_sk_err(u32 group, int error)
481{
482	netlink_set_err(rtnl, 0, group, error);
483}
484
485int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
486{
487	struct nlattr *mx;
488	int i, valid = 0;
489
490	mx = nla_nest_start(skb, RTA_METRICS);
491	if (mx == NULL)
492		return -ENOBUFS;
493
494	for (i = 0; i < RTAX_MAX; i++) {
495		if (metrics[i]) {
496			valid++;
497			NLA_PUT_U32(skb, i+1, metrics[i]);
498		}
499	}
500
501	if (!valid) {
502		nla_nest_cancel(skb, mx);
503		return 0;
504	}
505
506	return nla_nest_end(skb, mx);
507
508nla_put_failure:
509	return nla_nest_cancel(skb, mx);
510}
511
512int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
513		       u32 ts, u32 tsage, long expires, u32 error)
514{
515	struct rta_cacheinfo ci = {
516		.rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
517		.rta_used = dst->__use,
518		.rta_clntref = atomic_read(&(dst->__refcnt)),
519		.rta_error = error,
520		.rta_id =  id,
521		.rta_ts = ts,
522		.rta_tsage = tsage,
523	};
524
525	if (expires)
526		ci.rta_expires = jiffies_to_clock_t(expires);
527
528	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
529}
530
531EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
532
533static void set_operstate(struct net_device *dev, unsigned char transition)
534{
535	unsigned char operstate = dev->operstate;
536
537	switch(transition) {
538	case IF_OPER_UP:
539		if ((operstate == IF_OPER_DORMANT ||
540		     operstate == IF_OPER_UNKNOWN) &&
541		    !netif_dormant(dev))
542			operstate = IF_OPER_UP;
543		break;
544
545	case IF_OPER_DORMANT:
546		if (operstate == IF_OPER_UP ||
547		    operstate == IF_OPER_UNKNOWN)
548			operstate = IF_OPER_DORMANT;
549		break;
550	}
551
552	if (dev->operstate != operstate) {
553		write_lock_bh(&dev_base_lock);
554		dev->operstate = operstate;
555		write_unlock_bh(&dev_base_lock);
556		netdev_state_change(dev);
557	}
558}
559
560static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
561				 struct net_device_stats *b)
562{
563	a->rx_packets = b->rx_packets;
564	a->tx_packets = b->tx_packets;
565	a->rx_bytes = b->rx_bytes;
566	a->tx_bytes = b->tx_bytes;
567	a->rx_errors = b->rx_errors;
568	a->tx_errors = b->tx_errors;
569	a->rx_dropped = b->rx_dropped;
570	a->tx_dropped = b->tx_dropped;
571
572	a->multicast = b->multicast;
573	a->collisions = b->collisions;
574
575	a->rx_length_errors = b->rx_length_errors;
576	a->rx_over_errors = b->rx_over_errors;
577	a->rx_crc_errors = b->rx_crc_errors;
578	a->rx_frame_errors = b->rx_frame_errors;
579	a->rx_fifo_errors = b->rx_fifo_errors;
580	a->rx_missed_errors = b->rx_missed_errors;
581
582	a->tx_aborted_errors = b->tx_aborted_errors;
583	a->tx_carrier_errors = b->tx_carrier_errors;
584	a->tx_fifo_errors = b->tx_fifo_errors;
585	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
586	a->tx_window_errors = b->tx_window_errors;
587
588	a->rx_compressed = b->rx_compressed;
589	a->tx_compressed = b->tx_compressed;
590};
591
592static inline size_t if_nlmsg_size(const struct net_device *dev)
593{
594	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
595	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
596	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
597	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
598	       + nla_total_size(sizeof(struct rtnl_link_stats))
599	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
600	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
601	       + nla_total_size(4) /* IFLA_TXQLEN */
602	       + nla_total_size(4) /* IFLA_WEIGHT */
603	       + nla_total_size(4) /* IFLA_MTU */
604	       + nla_total_size(4) /* IFLA_LINK */
605	       + nla_total_size(4) /* IFLA_MASTER */
606	       + nla_total_size(1) /* IFLA_OPERSTATE */
607	       + nla_total_size(1) /* IFLA_LINKMODE */
608	       + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
609}
610
611static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
612			    int type, u32 pid, u32 seq, u32 change,
613			    unsigned int flags)
614{
615	struct ifinfomsg *ifm;
616	struct nlmsghdr *nlh;
617
618	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
619	if (nlh == NULL)
620		return -EMSGSIZE;
621
622	ifm = nlmsg_data(nlh);
623	ifm->ifi_family = AF_UNSPEC;
624	ifm->__ifi_pad = 0;
625	ifm->ifi_type = dev->type;
626	ifm->ifi_index = dev->ifindex;
627	ifm->ifi_flags = dev_get_flags(dev);
628	ifm->ifi_change = change;
629
630	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
631	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
632	NLA_PUT_U32(skb, IFLA_WEIGHT, dev->weight);
633	NLA_PUT_U8(skb, IFLA_OPERSTATE,
634		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
635	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
636	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
637
638	if (dev->ifindex != dev->iflink)
639		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
640
641	if (dev->master)
642		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
643
644	if (dev->qdisc_sleeping)
645		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
646
647	if (1) {
648		struct rtnl_link_ifmap map = {
649			.mem_start   = dev->mem_start,
650			.mem_end     = dev->mem_end,
651			.base_addr   = dev->base_addr,
652			.irq         = dev->irq,
653			.dma         = dev->dma,
654			.port        = dev->if_port,
655		};
656		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
657	}
658
659	if (dev->addr_len) {
660		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
661		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
662	}
663
664	if (dev->get_stats) {
665		struct net_device_stats *stats = dev->get_stats(dev);
666		if (stats) {
667			struct nlattr *attr;
668
669			attr = nla_reserve(skb, IFLA_STATS,
670					   sizeof(struct rtnl_link_stats));
671			if (attr == NULL)
672				goto nla_put_failure;
673
674			copy_rtnl_link_stats(nla_data(attr), stats);
675		}
676	}
677
678	if (dev->rtnl_link_ops) {
679		if (rtnl_link_fill(skb, dev) < 0)
680			goto nla_put_failure;
681	}
682
683	return nlmsg_end(skb, nlh);
684
685nla_put_failure:
686	nlmsg_cancel(skb, nlh);
687	return -EMSGSIZE;
688}
689
690static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
691{
692	int idx;
693	int s_idx = cb->args[0];
694	struct net_device *dev;
695
696	idx = 0;
697	for_each_netdev(dev) {
698		if (idx < s_idx)
699			goto cont;
700		if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
701				     NETLINK_CB(cb->skb).pid,
702				     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
703			break;
704cont:
705		idx++;
706	}
707	cb->args[0] = idx;
708
709	return skb->len;
710}
711
712static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
713	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
714	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
715	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
716	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
717	[IFLA_MTU]		= { .type = NLA_U32 },
718	[IFLA_TXQLEN]		= { .type = NLA_U32 },
719	[IFLA_WEIGHT]		= { .type = NLA_U32 },
720	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
721	[IFLA_LINKMODE]		= { .type = NLA_U8 },
722};
723
724static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
725	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
726	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
727};
728
729static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
730		      struct nlattr **tb, char *ifname, int modified)
731{
732	int send_addr_notify = 0;
733	int err;
734
735	if (tb[IFLA_MAP]) {
736		struct rtnl_link_ifmap *u_map;
737		struct ifmap k_map;
738
739		if (!dev->set_config) {
740			err = -EOPNOTSUPP;
741			goto errout;
742		}
743
744		if (!netif_device_present(dev)) {
745			err = -ENODEV;
746			goto errout;
747		}
748
749		u_map = nla_data(tb[IFLA_MAP]);
750		k_map.mem_start = (unsigned long) u_map->mem_start;
751		k_map.mem_end = (unsigned long) u_map->mem_end;
752		k_map.base_addr = (unsigned short) u_map->base_addr;
753		k_map.irq = (unsigned char) u_map->irq;
754		k_map.dma = (unsigned char) u_map->dma;
755		k_map.port = (unsigned char) u_map->port;
756
757		err = dev->set_config(dev, &k_map);
758		if (err < 0)
759			goto errout;
760
761		modified = 1;
762	}
763
764	if (tb[IFLA_ADDRESS]) {
765		struct sockaddr *sa;
766		int len;
767
768		if (!dev->set_mac_address) {
769			err = -EOPNOTSUPP;
770			goto errout;
771		}
772
773		if (!netif_device_present(dev)) {
774			err = -ENODEV;
775			goto errout;
776		}
777
778		len = sizeof(sa_family_t) + dev->addr_len;
779		sa = kmalloc(len, GFP_KERNEL);
780		if (!sa) {
781			err = -ENOMEM;
782			goto errout;
783		}
784		sa->sa_family = dev->type;
785		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
786		       dev->addr_len);
787		err = dev->set_mac_address(dev, sa);
788		kfree(sa);
789		if (err)
790			goto errout;
791		send_addr_notify = 1;
792		modified = 1;
793	}
794
795	if (tb[IFLA_MTU]) {
796		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
797		if (err < 0)
798			goto errout;
799		modified = 1;
800	}
801
802	/*
803	 * Interface selected by interface index but interface
804	 * name provided implies that a name change has been
805	 * requested.
806	 */
807	if (ifm->ifi_index > 0 && ifname[0]) {
808		err = dev_change_name(dev, ifname);
809		if (err < 0)
810			goto errout;
811		modified = 1;
812	}
813
814	if (tb[IFLA_BROADCAST]) {
815		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
816		send_addr_notify = 1;
817	}
818
819	if (ifm->ifi_flags || ifm->ifi_change) {
820		unsigned int flags = ifm->ifi_flags;
821
822		/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
823		if (ifm->ifi_change)
824			flags = (flags & ifm->ifi_change) |
825				(dev->flags & ~ifm->ifi_change);
826		dev_change_flags(dev, flags);
827	}
828
829	if (tb[IFLA_TXQLEN])
830		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
831
832	if (tb[IFLA_WEIGHT])
833		dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
834
835	if (tb[IFLA_OPERSTATE])
836		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
837
838	if (tb[IFLA_LINKMODE]) {
839		write_lock_bh(&dev_base_lock);
840		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
841		write_unlock_bh(&dev_base_lock);
842	}
843
844	err = 0;
845
846errout:
847	if (err < 0 && modified && net_ratelimit())
848		printk(KERN_WARNING "A link change request failed with "
849		       "some changes comitted already. Interface %s may "
850		       "have been left with an inconsistent configuration, "
851		       "please check.\n", dev->name);
852
853	if (send_addr_notify)
854		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
855	return err;
856}
857
858static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
859{
860	struct ifinfomsg *ifm;
861	struct net_device *dev;
862	int err;
863	struct nlattr *tb[IFLA_MAX+1];
864	char ifname[IFNAMSIZ];
865
866	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
867	if (err < 0)
868		goto errout;
869
870	if (tb[IFLA_IFNAME])
871		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
872	else
873		ifname[0] = '\0';
874
875	err = -EINVAL;
876	ifm = nlmsg_data(nlh);
877	if (ifm->ifi_index > 0)
878		dev = dev_get_by_index(ifm->ifi_index);
879	else if (tb[IFLA_IFNAME])
880		dev = dev_get_by_name(ifname);
881	else
882		goto errout;
883
884	if (dev == NULL) {
885		err = -ENODEV;
886		goto errout;
887	}
888
889	if (tb[IFLA_ADDRESS] &&
890	    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
891		goto errout_dev;
892
893	if (tb[IFLA_BROADCAST] &&
894	    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
895		goto errout_dev;
896
897	err = do_setlink(dev, ifm, tb, ifname, 0);
898errout_dev:
899	dev_put(dev);
900errout:
901	return err;
902}
903
904static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
905{
906	const struct rtnl_link_ops *ops;
907	struct net_device *dev;
908	struct ifinfomsg *ifm;
909	char ifname[IFNAMSIZ];
910	struct nlattr *tb[IFLA_MAX+1];
911	int err;
912
913	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
914	if (err < 0)
915		return err;
916
917	if (tb[IFLA_IFNAME])
918		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
919
920	ifm = nlmsg_data(nlh);
921	if (ifm->ifi_index > 0)
922		dev = __dev_get_by_index(ifm->ifi_index);
923	else if (tb[IFLA_IFNAME])
924		dev = __dev_get_by_name(ifname);
925	else
926		return -EINVAL;
927
928	if (!dev)
929		return -ENODEV;
930
931	ops = dev->rtnl_link_ops;
932	if (!ops)
933		return -EOPNOTSUPP;
934
935	ops->dellink(dev);
936	return 0;
937}
938
939static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
940{
941	const struct rtnl_link_ops *ops;
942	struct net_device *dev;
943	struct ifinfomsg *ifm;
944	char kind[MODULE_NAME_LEN];
945	char ifname[IFNAMSIZ];
946	struct nlattr *tb[IFLA_MAX+1];
947	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
948	int err;
949
950replay:
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	else
958		ifname[0] = '\0';
959
960	ifm = nlmsg_data(nlh);
961	if (ifm->ifi_index > 0)
962		dev = __dev_get_by_index(ifm->ifi_index);
963	else if (ifname[0])
964		dev = __dev_get_by_name(ifname);
965	else
966		dev = NULL;
967
968	if (tb[IFLA_LINKINFO]) {
969		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
970				       tb[IFLA_LINKINFO], ifla_info_policy);
971		if (err < 0)
972			return err;
973	} else
974		memset(linkinfo, 0, sizeof(linkinfo));
975
976	if (linkinfo[IFLA_INFO_KIND]) {
977		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
978		ops = rtnl_link_ops_get(kind);
979	} else {
980		kind[0] = '\0';
981		ops = NULL;
982	}
983
984	if (1) {
985		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
986
987		if (ops) {
988			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
989				err = nla_parse_nested(attr, ops->maxtype,
990						       linkinfo[IFLA_INFO_DATA],
991						       ops->policy);
992				if (err < 0)
993					return err;
994				data = attr;
995			}
996			if (ops->validate) {
997				err = ops->validate(tb, data);
998				if (err < 0)
999					return err;
1000			}
1001		}
1002
1003		if (dev) {
1004			int modified = 0;
1005
1006			if (nlh->nlmsg_flags & NLM_F_EXCL)
1007				return -EEXIST;
1008			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1009				return -EOPNOTSUPP;
1010
1011			if (linkinfo[IFLA_INFO_DATA]) {
1012				if (!ops || ops != dev->rtnl_link_ops ||
1013				    !ops->changelink)
1014					return -EOPNOTSUPP;
1015
1016				err = ops->changelink(dev, tb, data);
1017				if (err < 0)
1018					return err;
1019				modified = 1;
1020			}
1021
1022			return do_setlink(dev, ifm, tb, ifname, modified);
1023		}
1024
1025		if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1026			return -ENODEV;
1027
1028		if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1029			return -EOPNOTSUPP;
1030		if (tb[IFLA_ADDRESS] || tb[IFLA_BROADCAST] || tb[IFLA_MAP] ||
1031		    tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1032			return -EOPNOTSUPP;
1033
1034		if (!ops) {
1035#ifdef CONFIG_KMOD
1036			if (kind[0]) {
1037				__rtnl_unlock();
1038				request_module("rtnl-link-%s", kind);
1039				rtnl_lock();
1040				ops = rtnl_link_ops_get(kind);
1041				if (ops)
1042					goto replay;
1043			}
1044#endif
1045			return -EOPNOTSUPP;
1046		}
1047
1048		if (!ifname[0])
1049			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1050		dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
1051		if (!dev)
1052			return -ENOMEM;
1053
1054		if (strchr(dev->name, '%')) {
1055			err = dev_alloc_name(dev, dev->name);
1056			if (err < 0)
1057				goto err_free;
1058		}
1059		dev->rtnl_link_ops = ops;
1060
1061		if (tb[IFLA_MTU])
1062			dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1063		if (tb[IFLA_TXQLEN])
1064			dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1065		if (tb[IFLA_WEIGHT])
1066			dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
1067		if (tb[IFLA_OPERSTATE])
1068			set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1069		if (tb[IFLA_LINKMODE])
1070			dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1071
1072		err = ops->newlink(dev, tb, data);
1073err_free:
1074		if (err < 0)
1075			free_netdev(dev);
1076		return err;
1077	}
1078}
1079
1080static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1081{
1082	struct ifinfomsg *ifm;
1083	struct nlattr *tb[IFLA_MAX+1];
1084	struct net_device *dev = NULL;
1085	struct sk_buff *nskb;
1086	int err;
1087
1088	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1089	if (err < 0)
1090		return err;
1091
1092	ifm = nlmsg_data(nlh);
1093	if (ifm->ifi_index > 0) {
1094		dev = dev_get_by_index(ifm->ifi_index);
1095		if (dev == NULL)
1096			return -ENODEV;
1097	} else
1098		return -EINVAL;
1099
1100	nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1101	if (nskb == NULL) {
1102		err = -ENOBUFS;
1103		goto errout;
1104	}
1105
1106	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1107			       nlh->nlmsg_seq, 0, 0);
1108	if (err < 0) {
1109		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1110		WARN_ON(err == -EMSGSIZE);
1111		kfree_skb(nskb);
1112		goto errout;
1113	}
1114	err = rtnl_unicast(nskb, NETLINK_CB(skb).pid);
1115errout:
1116	dev_put(dev);
1117
1118	return err;
1119}
1120
1121static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1122{
1123	int idx;
1124	int s_idx = cb->family;
1125
1126	if (s_idx == 0)
1127		s_idx = 1;
1128	for (idx=1; idx<NPROTO; idx++) {
1129		int type = cb->nlh->nlmsg_type-RTM_BASE;
1130		if (idx < s_idx || idx == PF_PACKET)
1131			continue;
1132		if (rtnl_msg_handlers[idx] == NULL ||
1133		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1134			continue;
1135		if (idx > s_idx)
1136			memset(&cb->args[0], 0, sizeof(cb->args));
1137		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1138			break;
1139	}
1140	cb->family = idx;
1141
1142	return skb->len;
1143}
1144
1145void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1146{
1147	struct sk_buff *skb;
1148	int err = -ENOBUFS;
1149
1150	skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1151	if (skb == NULL)
1152		goto errout;
1153
1154	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1155	if (err < 0) {
1156		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1157		WARN_ON(err == -EMSGSIZE);
1158		kfree_skb(skb);
1159		goto errout;
1160	}
1161	err = rtnl_notify(skb, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1162errout:
1163	if (err < 0)
1164		rtnl_set_sk_err(RTNLGRP_LINK, err);
1165}
1166
1167/* Protected by RTNL sempahore.  */
1168static struct rtattr **rta_buf;
1169static int rtattr_max;
1170
1171/* Process one rtnetlink message. */
1172
1173static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1174{
1175	rtnl_doit_func doit;
1176	int sz_idx, kind;
1177	int min_len;
1178	int family;
1179	int type;
1180	int err;
1181
1182	type = nlh->nlmsg_type;
1183	if (type > RTM_MAX)
1184		return -EOPNOTSUPP;
1185
1186	type -= RTM_BASE;
1187
1188	/* All the messages must have at least 1 byte length */
1189	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1190		return 0;
1191
1192	family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1193	if (family >= NPROTO)
1194		return -EAFNOSUPPORT;
1195
1196	sz_idx = type>>2;
1197	kind = type&3;
1198
1199	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1200		return -EPERM;
1201
1202	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1203		rtnl_dumpit_func dumpit;
1204
1205		dumpit = rtnl_get_dumpit(family, type);
1206		if (dumpit == NULL)
1207			return -EOPNOTSUPP;
1208
1209		__rtnl_unlock();
1210		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1211		rtnl_lock();
1212		return err;
1213	}
1214
1215	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1216
1217	min_len = rtm_min[sz_idx];
1218	if (nlh->nlmsg_len < min_len)
1219		return -EINVAL;
1220
1221	if (nlh->nlmsg_len > min_len) {
1222		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1223		struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1224
1225		while (RTA_OK(attr, attrlen)) {
1226			unsigned flavor = attr->rta_type;
1227			if (flavor) {
1228				if (flavor > rta_max[sz_idx])
1229					return -EINVAL;
1230				rta_buf[flavor-1] = attr;
1231			}
1232			attr = RTA_NEXT(attr, attrlen);
1233		}
1234	}
1235
1236	doit = rtnl_get_doit(family, type);
1237	if (doit == NULL)
1238		return -EOPNOTSUPP;
1239
1240	return doit(skb, nlh, (void *)&rta_buf[0]);
1241}
1242
1243static void rtnetlink_rcv(struct sock *sk, int len)
1244{
1245	unsigned int qlen = 0;
1246
1247	do {
1248		mutex_lock(&rtnl_mutex);
1249		netlink_run_queue(sk, &qlen, &rtnetlink_rcv_msg);
1250		mutex_unlock(&rtnl_mutex);
1251
1252		netdev_run_todo();
1253	} while (qlen);
1254}
1255
1256static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1257{
1258	struct net_device *dev = ptr;
1259	switch (event) {
1260	case NETDEV_UNREGISTER:
1261		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1262		break;
1263	case NETDEV_REGISTER:
1264		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1265		break;
1266	case NETDEV_UP:
1267	case NETDEV_DOWN:
1268		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1269		break;
1270	case NETDEV_CHANGE:
1271	case NETDEV_GOING_DOWN:
1272		break;
1273	default:
1274		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1275		break;
1276	}
1277	return NOTIFY_DONE;
1278}
1279
1280static struct notifier_block rtnetlink_dev_notifier = {
1281	.notifier_call	= rtnetlink_event,
1282};
1283
1284void __init rtnetlink_init(void)
1285{
1286	int i;
1287
1288	rtattr_max = 0;
1289	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1290		if (rta_max[i] > rtattr_max)
1291			rtattr_max = rta_max[i];
1292	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1293	if (!rta_buf)
1294		panic("rtnetlink_init: cannot allocate rta_buf\n");
1295
1296	rtnl = netlink_kernel_create(NETLINK_ROUTE, RTNLGRP_MAX, rtnetlink_rcv,
1297				     &rtnl_mutex, THIS_MODULE);
1298	if (rtnl == NULL)
1299		panic("rtnetlink_init: cannot initialize rtnetlink\n");
1300	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1301	register_netdevice_notifier(&rtnetlink_dev_notifier);
1302
1303	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1304	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1305	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1306	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1307
1308	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1309	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1310}
1311
1312EXPORT_SYMBOL(__rta_fill);
1313EXPORT_SYMBOL(rtattr_strlcpy);
1314EXPORT_SYMBOL(rtattr_parse);
1315EXPORT_SYMBOL(rtattr_parse_nested_compat);
1316EXPORT_SYMBOL(rtnetlink_put_metrics);
1317EXPORT_SYMBOL(rtnl_lock);
1318EXPORT_SYMBOL(rtnl_trylock);
1319EXPORT_SYMBOL(rtnl_unlock);
1320EXPORT_SYMBOL(rtnl_unicast);
1321EXPORT_SYMBOL(rtnl_notify);
1322EXPORT_SYMBOL(rtnl_set_sk_err);
1323