xref: /openbmc/linux/net/core/rtnetlink.c (revision 97676b6b5538b3e059d33b8338e7d5cc41c5f1f1)
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/sched.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.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/netlink.h>
55 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
56 #include <linux/wireless.h>
57 #include <net/iw_handler.h>
58 #endif	/* CONFIG_NET_WIRELESS_RTNETLINK */
59 
60 static DEFINE_MUTEX(rtnl_mutex);
61 
62 void rtnl_lock(void)
63 {
64 	mutex_lock(&rtnl_mutex);
65 }
66 
67 void __rtnl_unlock(void)
68 {
69 	mutex_unlock(&rtnl_mutex);
70 }
71 
72 void rtnl_unlock(void)
73 {
74 	mutex_unlock(&rtnl_mutex);
75 	if (rtnl && rtnl->sk_receive_queue.qlen)
76 		rtnl->sk_data_ready(rtnl, 0);
77 	netdev_run_todo();
78 }
79 
80 int rtnl_trylock(void)
81 {
82 	return mutex_trylock(&rtnl_mutex);
83 }
84 
85 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
86 {
87 	memset(tb, 0, sizeof(struct rtattr*)*maxattr);
88 
89 	while (RTA_OK(rta, len)) {
90 		unsigned flavor = rta->rta_type;
91 		if (flavor && flavor <= maxattr)
92 			tb[flavor-1] = rta;
93 		rta = RTA_NEXT(rta, len);
94 	}
95 	return 0;
96 }
97 
98 struct sock *rtnl;
99 
100 struct rtnetlink_link * rtnetlink_links[NPROTO];
101 
102 static const int rtm_min[RTM_NR_FAMILIES] =
103 {
104 	[RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
105 	[RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
106 	[RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
107 	[RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
108 	[RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
109 	[RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
110 	[RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
111 	[RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
112 	[RTM_FAM(RTM_NEWPREFIX)]    = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
113 	[RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
114 	[RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
115 };
116 
117 static const int rta_max[RTM_NR_FAMILIES] =
118 {
119 	[RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
120 	[RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
121 	[RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
122 	[RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
123 	[RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
124 	[RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
125 	[RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
126 	[RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
127 };
128 
129 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
130 {
131 	struct rtattr *rta;
132 	int size = RTA_LENGTH(attrlen);
133 
134 	rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
135 	rta->rta_type = attrtype;
136 	rta->rta_len = size;
137 	memcpy(RTA_DATA(rta), data, attrlen);
138 	memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
139 }
140 
141 size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
142 {
143 	size_t ret = RTA_PAYLOAD(rta);
144 	char *src = RTA_DATA(rta);
145 
146 	if (ret > 0 && src[ret - 1] == '\0')
147 		ret--;
148 	if (size > 0) {
149 		size_t len = (ret >= size) ? size - 1 : ret;
150 		memset(dest, 0, size);
151 		memcpy(dest, src, len);
152 	}
153 	return ret;
154 }
155 
156 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
157 {
158 	int err = 0;
159 
160 	NETLINK_CB(skb).dst_group = group;
161 	if (echo)
162 		atomic_inc(&skb->users);
163 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
164 	if (echo)
165 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
166 	return err;
167 }
168 
169 int rtnl_unicast(struct sk_buff *skb, u32 pid)
170 {
171 	return nlmsg_unicast(rtnl, skb, pid);
172 }
173 
174 int rtnl_notify(struct sk_buff *skb, u32 pid, u32 group,
175 		struct nlmsghdr *nlh, gfp_t flags)
176 {
177 	int report = 0;
178 
179 	if (nlh)
180 		report = nlmsg_report(nlh);
181 
182 	return nlmsg_notify(rtnl, skb, pid, group, report, flags);
183 }
184 
185 void rtnl_set_sk_err(u32 group, int error)
186 {
187 	netlink_set_err(rtnl, 0, group, error);
188 }
189 
190 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
191 {
192 	struct rtattr *mx = (struct rtattr*)skb->tail;
193 	int i;
194 
195 	RTA_PUT(skb, RTA_METRICS, 0, NULL);
196 	for (i=0; i<RTAX_MAX; i++) {
197 		if (metrics[i])
198 			RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
199 	}
200 	mx->rta_len = skb->tail - (u8*)mx;
201 	if (mx->rta_len == RTA_LENGTH(0))
202 		skb_trim(skb, (u8*)mx - skb->data);
203 	return 0;
204 
205 rtattr_failure:
206 	skb_trim(skb, (u8*)mx - skb->data);
207 	return -1;
208 }
209 
210 
211 static void set_operstate(struct net_device *dev, unsigned char transition)
212 {
213 	unsigned char operstate = dev->operstate;
214 
215 	switch(transition) {
216 	case IF_OPER_UP:
217 		if ((operstate == IF_OPER_DORMANT ||
218 		     operstate == IF_OPER_UNKNOWN) &&
219 		    !netif_dormant(dev))
220 			operstate = IF_OPER_UP;
221 		break;
222 
223 	case IF_OPER_DORMANT:
224 		if (operstate == IF_OPER_UP ||
225 		    operstate == IF_OPER_UNKNOWN)
226 			operstate = IF_OPER_DORMANT;
227 		break;
228 	};
229 
230 	if (dev->operstate != operstate) {
231 		write_lock_bh(&dev_base_lock);
232 		dev->operstate = operstate;
233 		write_unlock_bh(&dev_base_lock);
234 		netdev_state_change(dev);
235 	}
236 }
237 
238 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
239 				 struct net_device_stats *b)
240 {
241 	a->rx_packets = b->rx_packets;
242 	a->tx_packets = b->tx_packets;
243 	a->rx_bytes = b->rx_bytes;
244 	a->tx_bytes = b->tx_bytes;
245 	a->rx_errors = b->rx_errors;
246 	a->tx_errors = b->tx_errors;
247 	a->rx_dropped = b->rx_dropped;
248 	a->tx_dropped = b->tx_dropped;
249 
250 	a->multicast = b->multicast;
251 	a->collisions = b->collisions;
252 
253 	a->rx_length_errors = b->rx_length_errors;
254 	a->rx_over_errors = b->rx_over_errors;
255 	a->rx_crc_errors = b->rx_crc_errors;
256 	a->rx_frame_errors = b->rx_frame_errors;
257 	a->rx_fifo_errors = b->rx_fifo_errors;
258 	a->rx_missed_errors = b->rx_missed_errors;
259 
260 	a->tx_aborted_errors = b->tx_aborted_errors;
261 	a->tx_carrier_errors = b->tx_carrier_errors;
262 	a->tx_fifo_errors = b->tx_fifo_errors;
263 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
264 	a->tx_window_errors = b->tx_window_errors;
265 
266 	a->rx_compressed = b->rx_compressed;
267 	a->tx_compressed = b->tx_compressed;
268 };
269 
270 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
271 			    void *iwbuf, int iwbuflen, int type, u32 pid,
272 			    u32 seq, u32 change, unsigned int flags)
273 {
274 	struct ifinfomsg *ifm;
275 	struct nlmsghdr *nlh;
276 
277 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
278 	if (nlh == NULL)
279 		return -ENOBUFS;
280 
281 	ifm = nlmsg_data(nlh);
282 	ifm->ifi_family = AF_UNSPEC;
283 	ifm->__ifi_pad = 0;
284 	ifm->ifi_type = dev->type;
285 	ifm->ifi_index = dev->ifindex;
286 	ifm->ifi_flags = dev_get_flags(dev);
287 	ifm->ifi_change = change;
288 
289 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
290 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
291 	NLA_PUT_U32(skb, IFLA_WEIGHT, dev->weight);
292 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
293 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
294 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
295 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
296 
297 	if (dev->ifindex != dev->iflink)
298 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
299 
300 	if (dev->master)
301 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
302 
303 	if (dev->qdisc_sleeping)
304 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
305 
306 	if (1) {
307 		struct rtnl_link_ifmap map = {
308 			.mem_start   = dev->mem_start,
309 			.mem_end     = dev->mem_end,
310 			.base_addr   = dev->base_addr,
311 			.irq         = dev->irq,
312 			.dma         = dev->dma,
313 			.port        = dev->if_port,
314 		};
315 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
316 	}
317 
318 	if (dev->addr_len) {
319 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
320 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
321 	}
322 
323 	if (dev->get_stats) {
324 		struct net_device_stats *stats = dev->get_stats(dev);
325 		if (stats) {
326 			struct nlattr *attr;
327 
328 			attr = nla_reserve(skb, IFLA_STATS,
329 					   sizeof(struct rtnl_link_stats));
330 			if (attr == NULL)
331 				goto nla_put_failure;
332 
333 			copy_rtnl_link_stats(nla_data(attr), stats);
334 		}
335 	}
336 
337 	if (iwbuf)
338 		NLA_PUT(skb, IFLA_WIRELESS, iwbuflen, iwbuf);
339 
340 	return nlmsg_end(skb, nlh);
341 
342 nla_put_failure:
343 	return nlmsg_cancel(skb, nlh);
344 }
345 
346 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
347 {
348 	int idx;
349 	int s_idx = cb->args[0];
350 	struct net_device *dev;
351 
352 	read_lock(&dev_base_lock);
353 	for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
354 		if (idx < s_idx)
355 			continue;
356 		if (rtnl_fill_ifinfo(skb, dev, NULL, 0, RTM_NEWLINK,
357 				     NETLINK_CB(cb->skb).pid,
358 				     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
359 			break;
360 	}
361 	read_unlock(&dev_base_lock);
362 	cb->args[0] = idx;
363 
364 	return skb->len;
365 }
366 
367 static struct nla_policy ifla_policy[IFLA_MAX+1] __read_mostly = {
368 	[IFLA_IFNAME]		= { .type = NLA_STRING },
369 	[IFLA_MAP]		= { .minlen = sizeof(struct rtnl_link_ifmap) },
370 	[IFLA_MTU]		= { .type = NLA_U32 },
371 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
372 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
373 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
374 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
375 };
376 
377 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
378 {
379 	struct ifinfomsg *ifm;
380 	struct net_device *dev;
381 	int err, send_addr_notify = 0, modified = 0;
382 	struct nlattr *tb[IFLA_MAX+1];
383 	char ifname[IFNAMSIZ];
384 
385 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
386 	if (err < 0)
387 		goto errout;
388 
389 	if (tb[IFLA_IFNAME] &&
390 	    nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ) >= IFNAMSIZ)
391 		return -EINVAL;
392 
393 	err = -EINVAL;
394 	ifm = nlmsg_data(nlh);
395 	if (ifm->ifi_index >= 0)
396 		dev = dev_get_by_index(ifm->ifi_index);
397 	else if (tb[IFLA_IFNAME])
398 		dev = dev_get_by_name(ifname);
399 	else
400 		goto errout;
401 
402 	if (dev == NULL) {
403 		err = -ENODEV;
404 		goto errout;
405 	}
406 
407 	if (tb[IFLA_ADDRESS] &&
408 	    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
409 		goto errout_dev;
410 
411 	if (tb[IFLA_BROADCAST] &&
412 	    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
413 		goto errout_dev;
414 
415 	if (tb[IFLA_MAP]) {
416 		struct rtnl_link_ifmap *u_map;
417 		struct ifmap k_map;
418 
419 		if (!dev->set_config) {
420 			err = -EOPNOTSUPP;
421 			goto errout_dev;
422 		}
423 
424 		if (!netif_device_present(dev)) {
425 			err = -ENODEV;
426 			goto errout_dev;
427 		}
428 
429 		u_map = nla_data(tb[IFLA_MAP]);
430 		k_map.mem_start = (unsigned long) u_map->mem_start;
431 		k_map.mem_end = (unsigned long) u_map->mem_end;
432 		k_map.base_addr = (unsigned short) u_map->base_addr;
433 		k_map.irq = (unsigned char) u_map->irq;
434 		k_map.dma = (unsigned char) u_map->dma;
435 		k_map.port = (unsigned char) u_map->port;
436 
437 		err = dev->set_config(dev, &k_map);
438 		if (err < 0)
439 			goto errout_dev;
440 
441 		modified = 1;
442 	}
443 
444 	if (tb[IFLA_ADDRESS]) {
445 		struct sockaddr *sa;
446 		int len;
447 
448 		if (!dev->set_mac_address) {
449 			err = -EOPNOTSUPP;
450 			goto errout_dev;
451 		}
452 
453 		if (!netif_device_present(dev)) {
454 			err = -ENODEV;
455 			goto errout_dev;
456 		}
457 
458 		len = sizeof(sa_family_t) + dev->addr_len;
459 		sa = kmalloc(len, GFP_KERNEL);
460 		if (!sa) {
461 			err = -ENOMEM;
462 			goto errout_dev;
463 		}
464 		sa->sa_family = dev->type;
465 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
466 		       dev->addr_len);
467 		err = dev->set_mac_address(dev, sa);
468 		kfree(sa);
469 		if (err)
470 			goto errout_dev;
471 		send_addr_notify = 1;
472 		modified = 1;
473 	}
474 
475 	if (tb[IFLA_MTU]) {
476 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
477 		if (err < 0)
478 			goto errout_dev;
479 		modified = 1;
480 	}
481 
482 	/*
483 	 * Interface selected by interface index but interface
484 	 * name provided implies that a name change has been
485 	 * requested.
486 	 */
487 	if (ifm->ifi_index >= 0 && ifname[0]) {
488 		err = dev_change_name(dev, ifname);
489 		if (err < 0)
490 			goto errout_dev;
491 		modified = 1;
492 	}
493 
494 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
495 	if (tb[IFLA_WIRELESS]) {
496 		/* Call Wireless Extensions.
497 		 * Various stuff checked in there... */
498 		err = wireless_rtnetlink_set(dev, nla_data(tb[IFLA_WIRELESS]),
499 					     nla_len(tb[IFLA_WIRELESS]));
500 		if (err < 0)
501 			goto errout_dev;
502 	}
503 #endif	/* CONFIG_NET_WIRELESS_RTNETLINK */
504 
505 	if (tb[IFLA_BROADCAST]) {
506 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
507 		send_addr_notify = 1;
508 	}
509 
510 
511 	if (ifm->ifi_flags)
512 		dev_change_flags(dev, ifm->ifi_flags);
513 
514 	if (tb[IFLA_TXQLEN])
515 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
516 
517 	if (tb[IFLA_WEIGHT])
518 		dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
519 
520 	if (tb[IFLA_OPERSTATE])
521 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
522 
523 	if (tb[IFLA_LINKMODE]) {
524 		write_lock_bh(&dev_base_lock);
525 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
526 		write_unlock_bh(&dev_base_lock);
527 	}
528 
529 	err = 0;
530 
531 errout_dev:
532 	if (err < 0 && modified && net_ratelimit())
533 		printk(KERN_WARNING "A link change request failed with "
534 		       "some changes comitted already. Interface %s may "
535 		       "have been left with an inconsistent configuration, "
536 		       "please check.\n", dev->name);
537 
538 	if (send_addr_notify)
539 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
540 
541 	dev_put(dev);
542 errout:
543 	return err;
544 }
545 
546 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
547 {
548 	struct ifinfomsg *ifm;
549 	struct nlattr *tb[IFLA_MAX+1];
550 	struct net_device *dev = NULL;
551 	struct sk_buff *nskb;
552 	char *iw_buf = NULL, *iw = NULL;
553 	int iw_buf_len = 0;
554 	int err, payload;
555 
556 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
557 	if (err < 0)
558 		goto errout;
559 
560 	ifm = nlmsg_data(nlh);
561 	if (ifm->ifi_index >= 0) {
562 		dev = dev_get_by_index(ifm->ifi_index);
563 		if (dev == NULL)
564 			return -ENODEV;
565 	} else
566 		return -EINVAL;
567 
568 
569 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
570 	if (tb[IFLA_WIRELESS]) {
571 		/* Call Wireless Extensions. We need to know the size before
572 		 * we can alloc. Various stuff checked in there... */
573 		err = wireless_rtnetlink_get(dev, nla_data(tb[IFLA_WIRELESS]),
574 					     nla_len(tb[IFLA_WIRELESS]),
575 					     &iw_buf, &iw_buf_len);
576 		if (err < 0)
577 			goto errout;
578 
579 		iw += IW_EV_POINT_OFF;
580 	}
581 #endif	/* CONFIG_NET_WIRELESS_RTNETLINK */
582 
583 	payload = NLMSG_ALIGN(sizeof(struct ifinfomsg) +
584 			      nla_total_size(iw_buf_len));
585 	nskb = nlmsg_new(nlmsg_total_size(payload), GFP_KERNEL);
586 	if (nskb == NULL) {
587 		err = -ENOBUFS;
588 		goto errout;
589 	}
590 
591 	err = rtnl_fill_ifinfo(nskb, dev, iw, iw_buf_len, RTM_NEWLINK,
592 			       NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 0);
593 	if (err <= 0) {
594 		kfree_skb(skb);
595 		goto errout;
596 	}
597 
598 	err = rtnl_unicast(skb, NETLINK_CB(skb).pid);
599 errout:
600 	kfree(iw_buf);
601 	dev_put(dev);
602 
603 	return err;
604 }
605 
606 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
607 {
608 	int idx;
609 	int s_idx = cb->family;
610 
611 	if (s_idx == 0)
612 		s_idx = 1;
613 	for (idx=1; idx<NPROTO; idx++) {
614 		int type = cb->nlh->nlmsg_type-RTM_BASE;
615 		if (idx < s_idx || idx == PF_PACKET)
616 			continue;
617 		if (rtnetlink_links[idx] == NULL ||
618 		    rtnetlink_links[idx][type].dumpit == NULL)
619 			continue;
620 		if (idx > s_idx)
621 			memset(&cb->args[0], 0, sizeof(cb->args));
622 		if (rtnetlink_links[idx][type].dumpit(skb, cb))
623 			break;
624 	}
625 	cb->family = idx;
626 
627 	return skb->len;
628 }
629 
630 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
631 {
632 	struct sk_buff *skb;
633 	int size = NLMSG_SPACE(sizeof(struct ifinfomsg) +
634 			       sizeof(struct rtnl_link_ifmap) +
635 			       sizeof(struct rtnl_link_stats) + 128);
636 
637 	skb = nlmsg_new(size, GFP_KERNEL);
638 	if (!skb)
639 		return;
640 
641 	if (rtnl_fill_ifinfo(skb, dev, NULL, 0, type, 0, 0, change, 0) < 0) {
642 		kfree_skb(skb);
643 		return;
644 	}
645 	NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
646 	netlink_broadcast(rtnl, skb, 0, RTNLGRP_LINK, GFP_KERNEL);
647 }
648 
649 /* Protected by RTNL sempahore.  */
650 static struct rtattr **rta_buf;
651 static int rtattr_max;
652 
653 /* Process one rtnetlink message. */
654 
655 static __inline__ int
656 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
657 {
658 	struct rtnetlink_link *link;
659 	struct rtnetlink_link *link_tab;
660 	int sz_idx, kind;
661 	int min_len;
662 	int family;
663 	int type;
664 	int err;
665 
666 	/* Only requests are handled by kernel now */
667 	if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
668 		return 0;
669 
670 	type = nlh->nlmsg_type;
671 
672 	/* A control message: ignore them */
673 	if (type < RTM_BASE)
674 		return 0;
675 
676 	/* Unknown message: reply with EINVAL */
677 	if (type > RTM_MAX)
678 		goto err_inval;
679 
680 	type -= RTM_BASE;
681 
682 	/* All the messages must have at least 1 byte length */
683 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
684 		return 0;
685 
686 	family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
687 	if (family >= NPROTO) {
688 		*errp = -EAFNOSUPPORT;
689 		return -1;
690 	}
691 
692 	link_tab = rtnetlink_links[family];
693 	if (link_tab == NULL)
694 		link_tab = rtnetlink_links[PF_UNSPEC];
695 	link = &link_tab[type];
696 
697 	sz_idx = type>>2;
698 	kind = type&3;
699 
700 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN)) {
701 		*errp = -EPERM;
702 		return -1;
703 	}
704 
705 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
706 		if (link->dumpit == NULL)
707 			link = &(rtnetlink_links[PF_UNSPEC][type]);
708 
709 		if (link->dumpit == NULL)
710 			goto err_inval;
711 
712 		if ((*errp = netlink_dump_start(rtnl, skb, nlh,
713 						link->dumpit, NULL)) != 0) {
714 			return -1;
715 		}
716 
717 		netlink_queue_skip(nlh, skb);
718 		return -1;
719 	}
720 
721 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
722 
723 	min_len = rtm_min[sz_idx];
724 	if (nlh->nlmsg_len < min_len)
725 		goto err_inval;
726 
727 	if (nlh->nlmsg_len > min_len) {
728 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
729 		struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
730 
731 		while (RTA_OK(attr, attrlen)) {
732 			unsigned flavor = attr->rta_type;
733 			if (flavor) {
734 				if (flavor > rta_max[sz_idx])
735 					goto err_inval;
736 				rta_buf[flavor-1] = attr;
737 			}
738 			attr = RTA_NEXT(attr, attrlen);
739 		}
740 	}
741 
742 	if (link->doit == NULL)
743 		link = &(rtnetlink_links[PF_UNSPEC][type]);
744 	if (link->doit == NULL)
745 		goto err_inval;
746 	err = link->doit(skb, nlh, (void *)&rta_buf[0]);
747 
748 	*errp = err;
749 	return err;
750 
751 err_inval:
752 	*errp = -EINVAL;
753 	return -1;
754 }
755 
756 static void rtnetlink_rcv(struct sock *sk, int len)
757 {
758 	unsigned int qlen = 0;
759 
760 	do {
761 		mutex_lock(&rtnl_mutex);
762 		netlink_run_queue(sk, &qlen, &rtnetlink_rcv_msg);
763 		mutex_unlock(&rtnl_mutex);
764 
765 		netdev_run_todo();
766 	} while (qlen);
767 }
768 
769 static struct rtnetlink_link link_rtnetlink_table[RTM_NR_MSGTYPES] =
770 {
771 	[RTM_GETLINK     - RTM_BASE] = { .doit   = rtnl_getlink,
772 					 .dumpit = rtnl_dump_ifinfo	 },
773 	[RTM_SETLINK     - RTM_BASE] = { .doit   = rtnl_setlink		 },
774 	[RTM_GETADDR     - RTM_BASE] = { .dumpit = rtnl_dump_all	 },
775 	[RTM_GETROUTE    - RTM_BASE] = { .dumpit = rtnl_dump_all	 },
776 	[RTM_NEWNEIGH    - RTM_BASE] = { .doit   = neigh_add		 },
777 	[RTM_DELNEIGH    - RTM_BASE] = { .doit   = neigh_delete		 },
778 	[RTM_GETNEIGH    - RTM_BASE] = { .dumpit = neigh_dump_info	 },
779 #ifdef CONFIG_FIB_RULES
780 	[RTM_NEWRULE     - RTM_BASE] = { .doit   = fib_nl_newrule	 },
781 	[RTM_DELRULE     - RTM_BASE] = { .doit   = fib_nl_delrule	 },
782 #endif
783 	[RTM_GETRULE     - RTM_BASE] = { .dumpit = rtnl_dump_all	 },
784 	[RTM_GETNEIGHTBL - RTM_BASE] = { .dumpit = neightbl_dump_info	 },
785 	[RTM_SETNEIGHTBL - RTM_BASE] = { .doit   = neightbl_set		 },
786 };
787 
788 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
789 {
790 	struct net_device *dev = ptr;
791 	switch (event) {
792 	case NETDEV_UNREGISTER:
793 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
794 		break;
795 	case NETDEV_REGISTER:
796 		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
797 		break;
798 	case NETDEV_UP:
799 	case NETDEV_DOWN:
800 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
801 		break;
802 	case NETDEV_CHANGE:
803 	case NETDEV_GOING_DOWN:
804 		break;
805 	default:
806 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
807 		break;
808 	}
809 	return NOTIFY_DONE;
810 }
811 
812 static struct notifier_block rtnetlink_dev_notifier = {
813 	.notifier_call	= rtnetlink_event,
814 };
815 
816 void __init rtnetlink_init(void)
817 {
818 	int i;
819 
820 	rtattr_max = 0;
821 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
822 		if (rta_max[i] > rtattr_max)
823 			rtattr_max = rta_max[i];
824 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
825 	if (!rta_buf)
826 		panic("rtnetlink_init: cannot allocate rta_buf\n");
827 
828 	rtnl = netlink_kernel_create(NETLINK_ROUTE, RTNLGRP_MAX, rtnetlink_rcv,
829 	                             THIS_MODULE);
830 	if (rtnl == NULL)
831 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
832 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
833 	register_netdevice_notifier(&rtnetlink_dev_notifier);
834 	rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
835 	rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
836 }
837 
838 EXPORT_SYMBOL(__rta_fill);
839 EXPORT_SYMBOL(rtattr_strlcpy);
840 EXPORT_SYMBOL(rtattr_parse);
841 EXPORT_SYMBOL(rtnetlink_links);
842 EXPORT_SYMBOL(rtnetlink_put_metrics);
843 EXPORT_SYMBOL(rtnl);
844 EXPORT_SYMBOL(rtnl_lock);
845 EXPORT_SYMBOL(rtnl_trylock);
846 EXPORT_SYMBOL(rtnl_unlock);
847 EXPORT_SYMBOL(rtnl_unicast);
848 EXPORT_SYMBOL(rtnl_notify);
849 EXPORT_SYMBOL(rtnl_set_sk_err);
850