xref: /openbmc/linux/net/core/rtnetlink.c (revision 2942e90050569525628a9f34e0daaa9b661b49cc)
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 rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
175 {
176 	struct rtattr *mx = (struct rtattr*)skb->tail;
177 	int i;
178 
179 	RTA_PUT(skb, RTA_METRICS, 0, NULL);
180 	for (i=0; i<RTAX_MAX; i++) {
181 		if (metrics[i])
182 			RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
183 	}
184 	mx->rta_len = skb->tail - (u8*)mx;
185 	if (mx->rta_len == RTA_LENGTH(0))
186 		skb_trim(skb, (u8*)mx - skb->data);
187 	return 0;
188 
189 rtattr_failure:
190 	skb_trim(skb, (u8*)mx - skb->data);
191 	return -1;
192 }
193 
194 
195 static void set_operstate(struct net_device *dev, unsigned char transition)
196 {
197 	unsigned char operstate = dev->operstate;
198 
199 	switch(transition) {
200 	case IF_OPER_UP:
201 		if ((operstate == IF_OPER_DORMANT ||
202 		     operstate == IF_OPER_UNKNOWN) &&
203 		    !netif_dormant(dev))
204 			operstate = IF_OPER_UP;
205 		break;
206 
207 	case IF_OPER_DORMANT:
208 		if (operstate == IF_OPER_UP ||
209 		    operstate == IF_OPER_UNKNOWN)
210 			operstate = IF_OPER_DORMANT;
211 		break;
212 	};
213 
214 	if (dev->operstate != operstate) {
215 		write_lock_bh(&dev_base_lock);
216 		dev->operstate = operstate;
217 		write_unlock_bh(&dev_base_lock);
218 		netdev_state_change(dev);
219 	}
220 }
221 
222 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
223 				 struct net_device_stats *b)
224 {
225 	a->rx_packets = b->rx_packets;
226 	a->tx_packets = b->tx_packets;
227 	a->rx_bytes = b->rx_bytes;
228 	a->tx_bytes = b->tx_bytes;
229 	a->rx_errors = b->rx_errors;
230 	a->tx_errors = b->tx_errors;
231 	a->rx_dropped = b->rx_dropped;
232 	a->tx_dropped = b->tx_dropped;
233 
234 	a->multicast = b->multicast;
235 	a->collisions = b->collisions;
236 
237 	a->rx_length_errors = b->rx_length_errors;
238 	a->rx_over_errors = b->rx_over_errors;
239 	a->rx_crc_errors = b->rx_crc_errors;
240 	a->rx_frame_errors = b->rx_frame_errors;
241 	a->rx_fifo_errors = b->rx_fifo_errors;
242 	a->rx_missed_errors = b->rx_missed_errors;
243 
244 	a->tx_aborted_errors = b->tx_aborted_errors;
245 	a->tx_carrier_errors = b->tx_carrier_errors;
246 	a->tx_fifo_errors = b->tx_fifo_errors;
247 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
248 	a->tx_window_errors = b->tx_window_errors;
249 
250 	a->rx_compressed = b->rx_compressed;
251 	a->tx_compressed = b->tx_compressed;
252 };
253 
254 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
255 			    void *iwbuf, int iwbuflen, int type, u32 pid,
256 			    u32 seq, u32 change, unsigned int flags)
257 {
258 	struct ifinfomsg *ifm;
259 	struct nlmsghdr *nlh;
260 
261 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
262 	if (nlh == NULL)
263 		return -ENOBUFS;
264 
265 	ifm = nlmsg_data(nlh);
266 	ifm->ifi_family = AF_UNSPEC;
267 	ifm->__ifi_pad = 0;
268 	ifm->ifi_type = dev->type;
269 	ifm->ifi_index = dev->ifindex;
270 	ifm->ifi_flags = dev_get_flags(dev);
271 	ifm->ifi_change = change;
272 
273 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
274 	NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
275 	NLA_PUT_U32(skb, IFLA_WEIGHT, dev->weight);
276 	NLA_PUT_U8(skb, IFLA_OPERSTATE,
277 		   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
278 	NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
279 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
280 
281 	if (dev->ifindex != dev->iflink)
282 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
283 
284 	if (dev->master)
285 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
286 
287 	if (dev->qdisc_sleeping)
288 		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
289 
290 	if (1) {
291 		struct rtnl_link_ifmap map = {
292 			.mem_start   = dev->mem_start,
293 			.mem_end     = dev->mem_end,
294 			.base_addr   = dev->base_addr,
295 			.irq         = dev->irq,
296 			.dma         = dev->dma,
297 			.port        = dev->if_port,
298 		};
299 		NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
300 	}
301 
302 	if (dev->addr_len) {
303 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
304 		NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
305 	}
306 
307 	if (dev->get_stats) {
308 		struct net_device_stats *stats = dev->get_stats(dev);
309 		if (stats) {
310 			struct nlattr *attr;
311 
312 			attr = nla_reserve(skb, IFLA_STATS,
313 					   sizeof(struct rtnl_link_stats));
314 			if (attr == NULL)
315 				goto nla_put_failure;
316 
317 			copy_rtnl_link_stats(nla_data(attr), stats);
318 		}
319 	}
320 
321 	if (iwbuf)
322 		NLA_PUT(skb, IFLA_WIRELESS, iwbuflen, iwbuf);
323 
324 	return nlmsg_end(skb, nlh);
325 
326 nla_put_failure:
327 	return nlmsg_cancel(skb, nlh);
328 }
329 
330 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
331 {
332 	int idx;
333 	int s_idx = cb->args[0];
334 	struct net_device *dev;
335 
336 	read_lock(&dev_base_lock);
337 	for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
338 		if (idx < s_idx)
339 			continue;
340 		if (rtnl_fill_ifinfo(skb, dev, NULL, 0, RTM_NEWLINK,
341 				     NETLINK_CB(cb->skb).pid,
342 				     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
343 			break;
344 	}
345 	read_unlock(&dev_base_lock);
346 	cb->args[0] = idx;
347 
348 	return skb->len;
349 }
350 
351 static struct nla_policy ifla_policy[IFLA_MAX+1] __read_mostly = {
352 	[IFLA_IFNAME]		= { .type = NLA_STRING },
353 	[IFLA_MAP]		= { .minlen = sizeof(struct rtnl_link_ifmap) },
354 	[IFLA_MTU]		= { .type = NLA_U32 },
355 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
356 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
357 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
358 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
359 };
360 
361 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
362 {
363 	struct ifinfomsg *ifm;
364 	struct net_device *dev;
365 	int err, send_addr_notify = 0, modified = 0;
366 	struct nlattr *tb[IFLA_MAX+1];
367 	char ifname[IFNAMSIZ];
368 
369 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
370 	if (err < 0)
371 		goto errout;
372 
373 	if (tb[IFLA_IFNAME] &&
374 	    nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ) >= IFNAMSIZ)
375 		return -EINVAL;
376 
377 	err = -EINVAL;
378 	ifm = nlmsg_data(nlh);
379 	if (ifm->ifi_index >= 0)
380 		dev = dev_get_by_index(ifm->ifi_index);
381 	else if (tb[IFLA_IFNAME])
382 		dev = dev_get_by_name(ifname);
383 	else
384 		goto errout;
385 
386 	if (dev == NULL) {
387 		err = -ENODEV;
388 		goto errout;
389 	}
390 
391 	if (tb[IFLA_ADDRESS] &&
392 	    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
393 		goto errout_dev;
394 
395 	if (tb[IFLA_BROADCAST] &&
396 	    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
397 		goto errout_dev;
398 
399 	if (tb[IFLA_MAP]) {
400 		struct rtnl_link_ifmap *u_map;
401 		struct ifmap k_map;
402 
403 		if (!dev->set_config) {
404 			err = -EOPNOTSUPP;
405 			goto errout_dev;
406 		}
407 
408 		if (!netif_device_present(dev)) {
409 			err = -ENODEV;
410 			goto errout_dev;
411 		}
412 
413 		u_map = nla_data(tb[IFLA_MAP]);
414 		k_map.mem_start = (unsigned long) u_map->mem_start;
415 		k_map.mem_end = (unsigned long) u_map->mem_end;
416 		k_map.base_addr = (unsigned short) u_map->base_addr;
417 		k_map.irq = (unsigned char) u_map->irq;
418 		k_map.dma = (unsigned char) u_map->dma;
419 		k_map.port = (unsigned char) u_map->port;
420 
421 		err = dev->set_config(dev, &k_map);
422 		if (err < 0)
423 			goto errout_dev;
424 
425 		modified = 1;
426 	}
427 
428 	if (tb[IFLA_ADDRESS]) {
429 		struct sockaddr *sa;
430 		int len;
431 
432 		if (!dev->set_mac_address) {
433 			err = -EOPNOTSUPP;
434 			goto errout_dev;
435 		}
436 
437 		if (!netif_device_present(dev)) {
438 			err = -ENODEV;
439 			goto errout_dev;
440 		}
441 
442 		len = sizeof(sa_family_t) + dev->addr_len;
443 		sa = kmalloc(len, GFP_KERNEL);
444 		if (!sa) {
445 			err = -ENOMEM;
446 			goto errout_dev;
447 		}
448 		sa->sa_family = dev->type;
449 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
450 		       dev->addr_len);
451 		err = dev->set_mac_address(dev, sa);
452 		kfree(sa);
453 		if (err)
454 			goto errout_dev;
455 		send_addr_notify = 1;
456 		modified = 1;
457 	}
458 
459 	if (tb[IFLA_MTU]) {
460 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
461 		if (err < 0)
462 			goto errout_dev;
463 		modified = 1;
464 	}
465 
466 	/*
467 	 * Interface selected by interface index but interface
468 	 * name provided implies that a name change has been
469 	 * requested.
470 	 */
471 	if (ifm->ifi_index >= 0 && ifname[0]) {
472 		err = dev_change_name(dev, ifname);
473 		if (err < 0)
474 			goto errout_dev;
475 		modified = 1;
476 	}
477 
478 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
479 	if (tb[IFLA_WIRELESS]) {
480 		/* Call Wireless Extensions.
481 		 * Various stuff checked in there... */
482 		err = wireless_rtnetlink_set(dev, nla_data(tb[IFLA_WIRELESS]),
483 					     nla_len(tb[IFLA_WIRELESS]));
484 		if (err < 0)
485 			goto errout_dev;
486 	}
487 #endif	/* CONFIG_NET_WIRELESS_RTNETLINK */
488 
489 	if (tb[IFLA_BROADCAST]) {
490 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
491 		send_addr_notify = 1;
492 	}
493 
494 
495 	if (ifm->ifi_flags)
496 		dev_change_flags(dev, ifm->ifi_flags);
497 
498 	if (tb[IFLA_TXQLEN])
499 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
500 
501 	if (tb[IFLA_WEIGHT])
502 		dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
503 
504 	if (tb[IFLA_OPERSTATE])
505 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
506 
507 	if (tb[IFLA_LINKMODE]) {
508 		write_lock_bh(&dev_base_lock);
509 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
510 		write_unlock_bh(&dev_base_lock);
511 	}
512 
513 	err = 0;
514 
515 errout_dev:
516 	if (err < 0 && modified && net_ratelimit())
517 		printk(KERN_WARNING "A link change request failed with "
518 		       "some changes comitted already. Interface %s may "
519 		       "have been left with an inconsistent configuration, "
520 		       "please check.\n", dev->name);
521 
522 	if (send_addr_notify)
523 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
524 
525 	dev_put(dev);
526 errout:
527 	return err;
528 }
529 
530 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
531 {
532 	struct ifinfomsg *ifm;
533 	struct nlattr *tb[IFLA_MAX+1];
534 	struct net_device *dev = NULL;
535 	struct sk_buff *nskb;
536 	char *iw_buf = NULL, *iw = NULL;
537 	int iw_buf_len = 0;
538 	int err, payload;
539 
540 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
541 	if (err < 0)
542 		goto errout;
543 
544 	ifm = nlmsg_data(nlh);
545 	if (ifm->ifi_index >= 0) {
546 		dev = dev_get_by_index(ifm->ifi_index);
547 		if (dev == NULL)
548 			return -ENODEV;
549 	} else
550 		return -EINVAL;
551 
552 
553 #ifdef CONFIG_NET_WIRELESS_RTNETLINK
554 	if (tb[IFLA_WIRELESS]) {
555 		/* Call Wireless Extensions. We need to know the size before
556 		 * we can alloc. Various stuff checked in there... */
557 		err = wireless_rtnetlink_get(dev, nla_data(tb[IFLA_WIRELESS]),
558 					     nla_len(tb[IFLA_WIRELESS]),
559 					     &iw_buf, &iw_buf_len);
560 		if (err < 0)
561 			goto errout;
562 
563 		iw += IW_EV_POINT_OFF;
564 	}
565 #endif	/* CONFIG_NET_WIRELESS_RTNETLINK */
566 
567 	payload = NLMSG_ALIGN(sizeof(struct ifinfomsg) +
568 			      nla_total_size(iw_buf_len));
569 	nskb = nlmsg_new(nlmsg_total_size(payload), GFP_KERNEL);
570 	if (nskb == NULL) {
571 		err = -ENOBUFS;
572 		goto errout;
573 	}
574 
575 	err = rtnl_fill_ifinfo(nskb, dev, iw, iw_buf_len, RTM_NEWLINK,
576 			       NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 0);
577 	if (err <= 0) {
578 		kfree_skb(skb);
579 		goto errout;
580 	}
581 
582 	err = rtnl_unicast(skb, NETLINK_CB(skb).pid);
583 errout:
584 	kfree(iw_buf);
585 	dev_put(dev);
586 
587 	return err;
588 }
589 
590 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
591 {
592 	int idx;
593 	int s_idx = cb->family;
594 
595 	if (s_idx == 0)
596 		s_idx = 1;
597 	for (idx=1; idx<NPROTO; idx++) {
598 		int type = cb->nlh->nlmsg_type-RTM_BASE;
599 		if (idx < s_idx || idx == PF_PACKET)
600 			continue;
601 		if (rtnetlink_links[idx] == NULL ||
602 		    rtnetlink_links[idx][type].dumpit == NULL)
603 			continue;
604 		if (idx > s_idx)
605 			memset(&cb->args[0], 0, sizeof(cb->args));
606 		if (rtnetlink_links[idx][type].dumpit(skb, cb))
607 			break;
608 	}
609 	cb->family = idx;
610 
611 	return skb->len;
612 }
613 
614 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
615 {
616 	struct sk_buff *skb;
617 	int size = NLMSG_SPACE(sizeof(struct ifinfomsg) +
618 			       sizeof(struct rtnl_link_ifmap) +
619 			       sizeof(struct rtnl_link_stats) + 128);
620 
621 	skb = nlmsg_new(size, GFP_KERNEL);
622 	if (!skb)
623 		return;
624 
625 	if (rtnl_fill_ifinfo(skb, dev, NULL, 0, type, 0, 0, change, 0) < 0) {
626 		kfree_skb(skb);
627 		return;
628 	}
629 	NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
630 	netlink_broadcast(rtnl, skb, 0, RTNLGRP_LINK, GFP_KERNEL);
631 }
632 
633 /* Protected by RTNL sempahore.  */
634 static struct rtattr **rta_buf;
635 static int rtattr_max;
636 
637 /* Process one rtnetlink message. */
638 
639 static __inline__ int
640 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
641 {
642 	struct rtnetlink_link *link;
643 	struct rtnetlink_link *link_tab;
644 	int sz_idx, kind;
645 	int min_len;
646 	int family;
647 	int type;
648 	int err;
649 
650 	/* Only requests are handled by kernel now */
651 	if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
652 		return 0;
653 
654 	type = nlh->nlmsg_type;
655 
656 	/* A control message: ignore them */
657 	if (type < RTM_BASE)
658 		return 0;
659 
660 	/* Unknown message: reply with EINVAL */
661 	if (type > RTM_MAX)
662 		goto err_inval;
663 
664 	type -= RTM_BASE;
665 
666 	/* All the messages must have at least 1 byte length */
667 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
668 		return 0;
669 
670 	family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
671 	if (family >= NPROTO) {
672 		*errp = -EAFNOSUPPORT;
673 		return -1;
674 	}
675 
676 	link_tab = rtnetlink_links[family];
677 	if (link_tab == NULL)
678 		link_tab = rtnetlink_links[PF_UNSPEC];
679 	link = &link_tab[type];
680 
681 	sz_idx = type>>2;
682 	kind = type&3;
683 
684 	if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN)) {
685 		*errp = -EPERM;
686 		return -1;
687 	}
688 
689 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
690 		if (link->dumpit == NULL)
691 			link = &(rtnetlink_links[PF_UNSPEC][type]);
692 
693 		if (link->dumpit == NULL)
694 			goto err_inval;
695 
696 		if ((*errp = netlink_dump_start(rtnl, skb, nlh,
697 						link->dumpit, NULL)) != 0) {
698 			return -1;
699 		}
700 
701 		netlink_queue_skip(nlh, skb);
702 		return -1;
703 	}
704 
705 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
706 
707 	min_len = rtm_min[sz_idx];
708 	if (nlh->nlmsg_len < min_len)
709 		goto err_inval;
710 
711 	if (nlh->nlmsg_len > min_len) {
712 		int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
713 		struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
714 
715 		while (RTA_OK(attr, attrlen)) {
716 			unsigned flavor = attr->rta_type;
717 			if (flavor) {
718 				if (flavor > rta_max[sz_idx])
719 					goto err_inval;
720 				rta_buf[flavor-1] = attr;
721 			}
722 			attr = RTA_NEXT(attr, attrlen);
723 		}
724 	}
725 
726 	if (link->doit == NULL)
727 		link = &(rtnetlink_links[PF_UNSPEC][type]);
728 	if (link->doit == NULL)
729 		goto err_inval;
730 	err = link->doit(skb, nlh, (void *)&rta_buf[0]);
731 
732 	*errp = err;
733 	return err;
734 
735 err_inval:
736 	*errp = -EINVAL;
737 	return -1;
738 }
739 
740 static void rtnetlink_rcv(struct sock *sk, int len)
741 {
742 	unsigned int qlen = 0;
743 
744 	do {
745 		mutex_lock(&rtnl_mutex);
746 		netlink_run_queue(sk, &qlen, &rtnetlink_rcv_msg);
747 		mutex_unlock(&rtnl_mutex);
748 
749 		netdev_run_todo();
750 	} while (qlen);
751 }
752 
753 static struct rtnetlink_link link_rtnetlink_table[RTM_NR_MSGTYPES] =
754 {
755 	[RTM_GETLINK     - RTM_BASE] = { .doit   = rtnl_getlink,
756 					 .dumpit = rtnl_dump_ifinfo	 },
757 	[RTM_SETLINK     - RTM_BASE] = { .doit   = rtnl_setlink		 },
758 	[RTM_GETADDR     - RTM_BASE] = { .dumpit = rtnl_dump_all	 },
759 	[RTM_GETROUTE    - RTM_BASE] = { .dumpit = rtnl_dump_all	 },
760 	[RTM_NEWNEIGH    - RTM_BASE] = { .doit   = neigh_add		 },
761 	[RTM_DELNEIGH    - RTM_BASE] = { .doit   = neigh_delete		 },
762 	[RTM_GETNEIGH    - RTM_BASE] = { .dumpit = neigh_dump_info	 },
763 #ifdef CONFIG_FIB_RULES
764 	[RTM_NEWRULE     - RTM_BASE] = { .doit   = fib_nl_newrule	 },
765 	[RTM_DELRULE     - RTM_BASE] = { .doit   = fib_nl_delrule	 },
766 #endif
767 	[RTM_GETRULE     - RTM_BASE] = { .dumpit = rtnl_dump_all	 },
768 	[RTM_GETNEIGHTBL - RTM_BASE] = { .dumpit = neightbl_dump_info	 },
769 	[RTM_SETNEIGHTBL - RTM_BASE] = { .doit   = neightbl_set		 },
770 };
771 
772 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
773 {
774 	struct net_device *dev = ptr;
775 	switch (event) {
776 	case NETDEV_UNREGISTER:
777 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
778 		break;
779 	case NETDEV_REGISTER:
780 		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
781 		break;
782 	case NETDEV_UP:
783 	case NETDEV_DOWN:
784 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
785 		break;
786 	case NETDEV_CHANGE:
787 	case NETDEV_GOING_DOWN:
788 		break;
789 	default:
790 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
791 		break;
792 	}
793 	return NOTIFY_DONE;
794 }
795 
796 static struct notifier_block rtnetlink_dev_notifier = {
797 	.notifier_call	= rtnetlink_event,
798 };
799 
800 void __init rtnetlink_init(void)
801 {
802 	int i;
803 
804 	rtattr_max = 0;
805 	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
806 		if (rta_max[i] > rtattr_max)
807 			rtattr_max = rta_max[i];
808 	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
809 	if (!rta_buf)
810 		panic("rtnetlink_init: cannot allocate rta_buf\n");
811 
812 	rtnl = netlink_kernel_create(NETLINK_ROUTE, RTNLGRP_MAX, rtnetlink_rcv,
813 	                             THIS_MODULE);
814 	if (rtnl == NULL)
815 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
816 	netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
817 	register_netdevice_notifier(&rtnetlink_dev_notifier);
818 	rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
819 	rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
820 }
821 
822 EXPORT_SYMBOL(__rta_fill);
823 EXPORT_SYMBOL(rtattr_strlcpy);
824 EXPORT_SYMBOL(rtattr_parse);
825 EXPORT_SYMBOL(rtnetlink_links);
826 EXPORT_SYMBOL(rtnetlink_put_metrics);
827 EXPORT_SYMBOL(rtnl);
828 EXPORT_SYMBOL(rtnl_lock);
829 EXPORT_SYMBOL(rtnl_trylock);
830 EXPORT_SYMBOL(rtnl_unlock);
831 EXPORT_SYMBOL(rtnl_unicast);
832