xref: /openbmc/linux/net/core/rtnetlink.c (revision 661d2967b3f1b34eeaa7e212e7b9bbe8ee072b59)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Routing netlink socket interface: protocol independent part.
7  *
8  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  *
15  *	Fixes:
16  *	Vitaly E. Lavrov		RTA_OK arithmetics was wrong.
17  */
18 
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
29 #include <linux/mm.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/pci.h>
40 #include <linux/etherdevice.h>
41 
42 #include <asm/uaccess.h>
43 
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <net/arp.h>
49 #include <net/route.h>
50 #include <net/udp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 
57 struct rtnl_link {
58 	rtnl_doit_func		doit;
59 	rtnl_dumpit_func	dumpit;
60 	rtnl_calcit_func 	calcit;
61 };
62 
63 static DEFINE_MUTEX(rtnl_mutex);
64 
65 void rtnl_lock(void)
66 {
67 	mutex_lock(&rtnl_mutex);
68 }
69 EXPORT_SYMBOL(rtnl_lock);
70 
71 void __rtnl_unlock(void)
72 {
73 	mutex_unlock(&rtnl_mutex);
74 }
75 
76 void rtnl_unlock(void)
77 {
78 	/* This fellow will unlock it for us. */
79 	netdev_run_todo();
80 }
81 EXPORT_SYMBOL(rtnl_unlock);
82 
83 int rtnl_trylock(void)
84 {
85 	return mutex_trylock(&rtnl_mutex);
86 }
87 EXPORT_SYMBOL(rtnl_trylock);
88 
89 int rtnl_is_locked(void)
90 {
91 	return mutex_is_locked(&rtnl_mutex);
92 }
93 EXPORT_SYMBOL(rtnl_is_locked);
94 
95 #ifdef CONFIG_PROVE_LOCKING
96 int lockdep_rtnl_is_held(void)
97 {
98 	return lockdep_is_held(&rtnl_mutex);
99 }
100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
102 
103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
104 
105 static inline int rtm_msgindex(int msgtype)
106 {
107 	int msgindex = msgtype - RTM_BASE;
108 
109 	/*
110 	 * msgindex < 0 implies someone tried to register a netlink
111 	 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 	 * the message type has not been added to linux/rtnetlink.h
113 	 */
114 	BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
115 
116 	return msgindex;
117 }
118 
119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
120 {
121 	struct rtnl_link *tab;
122 
123 	if (protocol <= RTNL_FAMILY_MAX)
124 		tab = rtnl_msg_handlers[protocol];
125 	else
126 		tab = NULL;
127 
128 	if (tab == NULL || tab[msgindex].doit == NULL)
129 		tab = rtnl_msg_handlers[PF_UNSPEC];
130 
131 	return tab[msgindex].doit;
132 }
133 
134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
135 {
136 	struct rtnl_link *tab;
137 
138 	if (protocol <= RTNL_FAMILY_MAX)
139 		tab = rtnl_msg_handlers[protocol];
140 	else
141 		tab = NULL;
142 
143 	if (tab == NULL || tab[msgindex].dumpit == NULL)
144 		tab = rtnl_msg_handlers[PF_UNSPEC];
145 
146 	return tab[msgindex].dumpit;
147 }
148 
149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
150 {
151 	struct rtnl_link *tab;
152 
153 	if (protocol <= RTNL_FAMILY_MAX)
154 		tab = rtnl_msg_handlers[protocol];
155 	else
156 		tab = NULL;
157 
158 	if (tab == NULL || tab[msgindex].calcit == NULL)
159 		tab = rtnl_msg_handlers[PF_UNSPEC];
160 
161 	return tab[msgindex].calcit;
162 }
163 
164 /**
165  * __rtnl_register - Register a rtnetlink message type
166  * @protocol: Protocol family or PF_UNSPEC
167  * @msgtype: rtnetlink message type
168  * @doit: Function pointer called for each request message
169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170  * @calcit: Function pointer to calc size of dump message
171  *
172  * Registers the specified function pointers (at least one of them has
173  * to be non-NULL) to be called whenever a request message for the
174  * specified protocol family and message type is received.
175  *
176  * The special protocol family PF_UNSPEC may be used to define fallback
177  * function pointers for the case when no entry for the specific protocol
178  * family exists.
179  *
180  * Returns 0 on success or a negative error code.
181  */
182 int __rtnl_register(int protocol, int msgtype,
183 		    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
184 		    rtnl_calcit_func calcit)
185 {
186 	struct rtnl_link *tab;
187 	int msgindex;
188 
189 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
190 	msgindex = rtm_msgindex(msgtype);
191 
192 	tab = rtnl_msg_handlers[protocol];
193 	if (tab == NULL) {
194 		tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
195 		if (tab == NULL)
196 			return -ENOBUFS;
197 
198 		rtnl_msg_handlers[protocol] = tab;
199 	}
200 
201 	if (doit)
202 		tab[msgindex].doit = doit;
203 
204 	if (dumpit)
205 		tab[msgindex].dumpit = dumpit;
206 
207 	if (calcit)
208 		tab[msgindex].calcit = calcit;
209 
210 	return 0;
211 }
212 EXPORT_SYMBOL_GPL(__rtnl_register);
213 
214 /**
215  * rtnl_register - Register a rtnetlink message type
216  *
217  * Identical to __rtnl_register() but panics on failure. This is useful
218  * as failure of this function is very unlikely, it can only happen due
219  * to lack of memory when allocating the chain to store all message
220  * handlers for a protocol. Meant for use in init functions where lack
221  * of memory implies no sense in continuing.
222  */
223 void rtnl_register(int protocol, int msgtype,
224 		   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
225 		   rtnl_calcit_func calcit)
226 {
227 	if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
228 		panic("Unable to register rtnetlink message handler, "
229 		      "protocol = %d, message type = %d\n",
230 		      protocol, msgtype);
231 }
232 EXPORT_SYMBOL_GPL(rtnl_register);
233 
234 /**
235  * rtnl_unregister - Unregister a rtnetlink message type
236  * @protocol: Protocol family or PF_UNSPEC
237  * @msgtype: rtnetlink message type
238  *
239  * Returns 0 on success or a negative error code.
240  */
241 int rtnl_unregister(int protocol, int msgtype)
242 {
243 	int msgindex;
244 
245 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
246 	msgindex = rtm_msgindex(msgtype);
247 
248 	if (rtnl_msg_handlers[protocol] == NULL)
249 		return -ENOENT;
250 
251 	rtnl_msg_handlers[protocol][msgindex].doit = NULL;
252 	rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
253 
254 	return 0;
255 }
256 EXPORT_SYMBOL_GPL(rtnl_unregister);
257 
258 /**
259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260  * @protocol : Protocol family or PF_UNSPEC
261  *
262  * Identical to calling rtnl_unregster() for all registered message types
263  * of a certain protocol family.
264  */
265 void rtnl_unregister_all(int protocol)
266 {
267 	BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
268 
269 	kfree(rtnl_msg_handlers[protocol]);
270 	rtnl_msg_handlers[protocol] = NULL;
271 }
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
273 
274 static LIST_HEAD(link_ops);
275 
276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
277 {
278 	const struct rtnl_link_ops *ops;
279 
280 	list_for_each_entry(ops, &link_ops, list) {
281 		if (!strcmp(ops->kind, kind))
282 			return ops;
283 	}
284 	return NULL;
285 }
286 
287 /**
288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289  * @ops: struct rtnl_link_ops * to register
290  *
291  * The caller must hold the rtnl_mutex. This function should be used
292  * by drivers that create devices during module initialization. It
293  * must be called before registering the devices.
294  *
295  * Returns 0 on success or a negative error code.
296  */
297 int __rtnl_link_register(struct rtnl_link_ops *ops)
298 {
299 	if (rtnl_link_ops_get(ops->kind))
300 		return -EEXIST;
301 
302 	if (!ops->dellink)
303 		ops->dellink = unregister_netdevice_queue;
304 
305 	list_add_tail(&ops->list, &link_ops);
306 	return 0;
307 }
308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
309 
310 /**
311  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312  * @ops: struct rtnl_link_ops * to register
313  *
314  * Returns 0 on success or a negative error code.
315  */
316 int rtnl_link_register(struct rtnl_link_ops *ops)
317 {
318 	int err;
319 
320 	rtnl_lock();
321 	err = __rtnl_link_register(ops);
322 	rtnl_unlock();
323 	return err;
324 }
325 EXPORT_SYMBOL_GPL(rtnl_link_register);
326 
327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
328 {
329 	struct net_device *dev;
330 	LIST_HEAD(list_kill);
331 
332 	for_each_netdev(net, dev) {
333 		if (dev->rtnl_link_ops == ops)
334 			ops->dellink(dev, &list_kill);
335 	}
336 	unregister_netdevice_many(&list_kill);
337 }
338 
339 /**
340  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341  * @ops: struct rtnl_link_ops * to unregister
342  *
343  * The caller must hold the rtnl_mutex.
344  */
345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
346 {
347 	struct net *net;
348 
349 	for_each_net(net) {
350 		__rtnl_kill_links(net, ops);
351 	}
352 	list_del(&ops->list);
353 }
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
355 
356 /**
357  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
358  * @ops: struct rtnl_link_ops * to unregister
359  */
360 void rtnl_link_unregister(struct rtnl_link_ops *ops)
361 {
362 	rtnl_lock();
363 	__rtnl_link_unregister(ops);
364 	rtnl_unlock();
365 }
366 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
367 
368 static size_t rtnl_link_get_size(const struct net_device *dev)
369 {
370 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
371 	size_t size;
372 
373 	if (!ops)
374 		return 0;
375 
376 	size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
377 	       nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
378 
379 	if (ops->get_size)
380 		/* IFLA_INFO_DATA + nested data */
381 		size += nla_total_size(sizeof(struct nlattr)) +
382 			ops->get_size(dev);
383 
384 	if (ops->get_xstats_size)
385 		/* IFLA_INFO_XSTATS */
386 		size += nla_total_size(ops->get_xstats_size(dev));
387 
388 	return size;
389 }
390 
391 static LIST_HEAD(rtnl_af_ops);
392 
393 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
394 {
395 	const struct rtnl_af_ops *ops;
396 
397 	list_for_each_entry(ops, &rtnl_af_ops, list) {
398 		if (ops->family == family)
399 			return ops;
400 	}
401 
402 	return NULL;
403 }
404 
405 /**
406  * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
407  * @ops: struct rtnl_af_ops * to register
408  *
409  * The caller must hold the rtnl_mutex.
410  *
411  * Returns 0 on success or a negative error code.
412  */
413 int __rtnl_af_register(struct rtnl_af_ops *ops)
414 {
415 	list_add_tail(&ops->list, &rtnl_af_ops);
416 	return 0;
417 }
418 EXPORT_SYMBOL_GPL(__rtnl_af_register);
419 
420 /**
421  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
422  * @ops: struct rtnl_af_ops * to register
423  *
424  * Returns 0 on success or a negative error code.
425  */
426 int rtnl_af_register(struct rtnl_af_ops *ops)
427 {
428 	int err;
429 
430 	rtnl_lock();
431 	err = __rtnl_af_register(ops);
432 	rtnl_unlock();
433 	return err;
434 }
435 EXPORT_SYMBOL_GPL(rtnl_af_register);
436 
437 /**
438  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
439  * @ops: struct rtnl_af_ops * to unregister
440  *
441  * The caller must hold the rtnl_mutex.
442  */
443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
444 {
445 	list_del(&ops->list);
446 }
447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
448 
449 /**
450  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
451  * @ops: struct rtnl_af_ops * to unregister
452  */
453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
454 {
455 	rtnl_lock();
456 	__rtnl_af_unregister(ops);
457 	rtnl_unlock();
458 }
459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
460 
461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
462 {
463 	struct rtnl_af_ops *af_ops;
464 	size_t size;
465 
466 	/* IFLA_AF_SPEC */
467 	size = nla_total_size(sizeof(struct nlattr));
468 
469 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
470 		if (af_ops->get_link_af_size) {
471 			/* AF_* + nested data */
472 			size += nla_total_size(sizeof(struct nlattr)) +
473 				af_ops->get_link_af_size(dev);
474 		}
475 	}
476 
477 	return size;
478 }
479 
480 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
481 {
482 	const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
483 	struct nlattr *linkinfo, *data;
484 	int err = -EMSGSIZE;
485 
486 	linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
487 	if (linkinfo == NULL)
488 		goto out;
489 
490 	if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
491 		goto err_cancel_link;
492 	if (ops->fill_xstats) {
493 		err = ops->fill_xstats(skb, dev);
494 		if (err < 0)
495 			goto err_cancel_link;
496 	}
497 	if (ops->fill_info) {
498 		data = nla_nest_start(skb, IFLA_INFO_DATA);
499 		if (data == NULL)
500 			goto err_cancel_link;
501 		err = ops->fill_info(skb, dev);
502 		if (err < 0)
503 			goto err_cancel_data;
504 		nla_nest_end(skb, data);
505 	}
506 
507 	nla_nest_end(skb, linkinfo);
508 	return 0;
509 
510 err_cancel_data:
511 	nla_nest_cancel(skb, data);
512 err_cancel_link:
513 	nla_nest_cancel(skb, linkinfo);
514 out:
515 	return err;
516 }
517 
518 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
519 {
520 	struct sock *rtnl = net->rtnl;
521 	int err = 0;
522 
523 	NETLINK_CB(skb).dst_group = group;
524 	if (echo)
525 		atomic_inc(&skb->users);
526 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
527 	if (echo)
528 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
529 	return err;
530 }
531 
532 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
533 {
534 	struct sock *rtnl = net->rtnl;
535 
536 	return nlmsg_unicast(rtnl, skb, pid);
537 }
538 EXPORT_SYMBOL(rtnl_unicast);
539 
540 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
541 		 struct nlmsghdr *nlh, gfp_t flags)
542 {
543 	struct sock *rtnl = net->rtnl;
544 	int report = 0;
545 
546 	if (nlh)
547 		report = nlmsg_report(nlh);
548 
549 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
550 }
551 EXPORT_SYMBOL(rtnl_notify);
552 
553 void rtnl_set_sk_err(struct net *net, u32 group, int error)
554 {
555 	struct sock *rtnl = net->rtnl;
556 
557 	netlink_set_err(rtnl, 0, group, error);
558 }
559 EXPORT_SYMBOL(rtnl_set_sk_err);
560 
561 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
562 {
563 	struct nlattr *mx;
564 	int i, valid = 0;
565 
566 	mx = nla_nest_start(skb, RTA_METRICS);
567 	if (mx == NULL)
568 		return -ENOBUFS;
569 
570 	for (i = 0; i < RTAX_MAX; i++) {
571 		if (metrics[i]) {
572 			valid++;
573 			if (nla_put_u32(skb, i+1, metrics[i]))
574 				goto nla_put_failure;
575 		}
576 	}
577 
578 	if (!valid) {
579 		nla_nest_cancel(skb, mx);
580 		return 0;
581 	}
582 
583 	return nla_nest_end(skb, mx);
584 
585 nla_put_failure:
586 	nla_nest_cancel(skb, mx);
587 	return -EMSGSIZE;
588 }
589 EXPORT_SYMBOL(rtnetlink_put_metrics);
590 
591 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
592 		       long expires, u32 error)
593 {
594 	struct rta_cacheinfo ci = {
595 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
596 		.rta_used = dst->__use,
597 		.rta_clntref = atomic_read(&(dst->__refcnt)),
598 		.rta_error = error,
599 		.rta_id =  id,
600 	};
601 
602 	if (expires) {
603 		unsigned long clock;
604 
605 		clock = jiffies_to_clock_t(abs(expires));
606 		clock = min_t(unsigned long, clock, INT_MAX);
607 		ci.rta_expires = (expires > 0) ? clock : -clock;
608 	}
609 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
610 }
611 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
612 
613 static void set_operstate(struct net_device *dev, unsigned char transition)
614 {
615 	unsigned char operstate = dev->operstate;
616 
617 	switch (transition) {
618 	case IF_OPER_UP:
619 		if ((operstate == IF_OPER_DORMANT ||
620 		     operstate == IF_OPER_UNKNOWN) &&
621 		    !netif_dormant(dev))
622 			operstate = IF_OPER_UP;
623 		break;
624 
625 	case IF_OPER_DORMANT:
626 		if (operstate == IF_OPER_UP ||
627 		    operstate == IF_OPER_UNKNOWN)
628 			operstate = IF_OPER_DORMANT;
629 		break;
630 	}
631 
632 	if (dev->operstate != operstate) {
633 		write_lock_bh(&dev_base_lock);
634 		dev->operstate = operstate;
635 		write_unlock_bh(&dev_base_lock);
636 		netdev_state_change(dev);
637 	}
638 }
639 
640 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
641 {
642 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
643 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
644 }
645 
646 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
647 					   const struct ifinfomsg *ifm)
648 {
649 	unsigned int flags = ifm->ifi_flags;
650 
651 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
652 	if (ifm->ifi_change)
653 		flags = (flags & ifm->ifi_change) |
654 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
655 
656 	return flags;
657 }
658 
659 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
660 				 const struct rtnl_link_stats64 *b)
661 {
662 	a->rx_packets = b->rx_packets;
663 	a->tx_packets = b->tx_packets;
664 	a->rx_bytes = b->rx_bytes;
665 	a->tx_bytes = b->tx_bytes;
666 	a->rx_errors = b->rx_errors;
667 	a->tx_errors = b->tx_errors;
668 	a->rx_dropped = b->rx_dropped;
669 	a->tx_dropped = b->tx_dropped;
670 
671 	a->multicast = b->multicast;
672 	a->collisions = b->collisions;
673 
674 	a->rx_length_errors = b->rx_length_errors;
675 	a->rx_over_errors = b->rx_over_errors;
676 	a->rx_crc_errors = b->rx_crc_errors;
677 	a->rx_frame_errors = b->rx_frame_errors;
678 	a->rx_fifo_errors = b->rx_fifo_errors;
679 	a->rx_missed_errors = b->rx_missed_errors;
680 
681 	a->tx_aborted_errors = b->tx_aborted_errors;
682 	a->tx_carrier_errors = b->tx_carrier_errors;
683 	a->tx_fifo_errors = b->tx_fifo_errors;
684 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
685 	a->tx_window_errors = b->tx_window_errors;
686 
687 	a->rx_compressed = b->rx_compressed;
688 	a->tx_compressed = b->tx_compressed;
689 }
690 
691 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
692 {
693 	memcpy(v, b, sizeof(*b));
694 }
695 
696 /* All VF info */
697 static inline int rtnl_vfinfo_size(const struct net_device *dev,
698 				   u32 ext_filter_mask)
699 {
700 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
701 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
702 		int num_vfs = dev_num_vf(dev->dev.parent);
703 		size_t size = nla_total_size(sizeof(struct nlattr));
704 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
705 		size += num_vfs *
706 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
707 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
708 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
709 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
710 		return size;
711 	} else
712 		return 0;
713 }
714 
715 static size_t rtnl_port_size(const struct net_device *dev)
716 {
717 	size_t port_size = nla_total_size(4)		/* PORT_VF */
718 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
719 		+ nla_total_size(sizeof(struct ifla_port_vsi))
720 							/* PORT_VSI_TYPE */
721 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
722 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
723 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
724 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
725 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
726 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
727 		+ port_size;
728 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
729 		+ port_size;
730 
731 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
732 		return 0;
733 	if (dev_num_vf(dev->dev.parent))
734 		return port_self_size + vf_ports_size +
735 			vf_port_size * dev_num_vf(dev->dev.parent);
736 	else
737 		return port_self_size;
738 }
739 
740 static noinline size_t if_nlmsg_size(const struct net_device *dev,
741 				     u32 ext_filter_mask)
742 {
743 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
744 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
745 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
746 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
747 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
748 	       + nla_total_size(sizeof(struct rtnl_link_stats))
749 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
750 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
751 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
752 	       + nla_total_size(4) /* IFLA_TXQLEN */
753 	       + nla_total_size(4) /* IFLA_WEIGHT */
754 	       + nla_total_size(4) /* IFLA_MTU */
755 	       + nla_total_size(4) /* IFLA_LINK */
756 	       + nla_total_size(4) /* IFLA_MASTER */
757 	       + nla_total_size(1) /* IFLA_CARRIER */
758 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
759 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
760 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
761 	       + nla_total_size(1) /* IFLA_OPERSTATE */
762 	       + nla_total_size(1) /* IFLA_LINKMODE */
763 	       + nla_total_size(ext_filter_mask
764 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
765 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
766 	       + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
767 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
768 	       + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
769 }
770 
771 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
772 {
773 	struct nlattr *vf_ports;
774 	struct nlattr *vf_port;
775 	int vf;
776 	int err;
777 
778 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
779 	if (!vf_ports)
780 		return -EMSGSIZE;
781 
782 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
783 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
784 		if (!vf_port)
785 			goto nla_put_failure;
786 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
787 			goto nla_put_failure;
788 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
789 		if (err == -EMSGSIZE)
790 			goto nla_put_failure;
791 		if (err) {
792 			nla_nest_cancel(skb, vf_port);
793 			continue;
794 		}
795 		nla_nest_end(skb, vf_port);
796 	}
797 
798 	nla_nest_end(skb, vf_ports);
799 
800 	return 0;
801 
802 nla_put_failure:
803 	nla_nest_cancel(skb, vf_ports);
804 	return -EMSGSIZE;
805 }
806 
807 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
808 {
809 	struct nlattr *port_self;
810 	int err;
811 
812 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
813 	if (!port_self)
814 		return -EMSGSIZE;
815 
816 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
817 	if (err) {
818 		nla_nest_cancel(skb, port_self);
819 		return (err == -EMSGSIZE) ? err : 0;
820 	}
821 
822 	nla_nest_end(skb, port_self);
823 
824 	return 0;
825 }
826 
827 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
828 {
829 	int err;
830 
831 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
832 		return 0;
833 
834 	err = rtnl_port_self_fill(skb, dev);
835 	if (err)
836 		return err;
837 
838 	if (dev_num_vf(dev->dev.parent)) {
839 		err = rtnl_vf_ports_fill(skb, dev);
840 		if (err)
841 			return err;
842 	}
843 
844 	return 0;
845 }
846 
847 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
848 			    int type, u32 pid, u32 seq, u32 change,
849 			    unsigned int flags, u32 ext_filter_mask)
850 {
851 	struct ifinfomsg *ifm;
852 	struct nlmsghdr *nlh;
853 	struct rtnl_link_stats64 temp;
854 	const struct rtnl_link_stats64 *stats;
855 	struct nlattr *attr, *af_spec;
856 	struct rtnl_af_ops *af_ops;
857 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
858 
859 	ASSERT_RTNL();
860 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
861 	if (nlh == NULL)
862 		return -EMSGSIZE;
863 
864 	ifm = nlmsg_data(nlh);
865 	ifm->ifi_family = AF_UNSPEC;
866 	ifm->__ifi_pad = 0;
867 	ifm->ifi_type = dev->type;
868 	ifm->ifi_index = dev->ifindex;
869 	ifm->ifi_flags = dev_get_flags(dev);
870 	ifm->ifi_change = change;
871 
872 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
873 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
874 	    nla_put_u8(skb, IFLA_OPERSTATE,
875 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
876 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
877 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
878 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
879 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
880 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
881 #ifdef CONFIG_RPS
882 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
883 #endif
884 	    (dev->ifindex != dev->iflink &&
885 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
886 	    (upper_dev &&
887 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
888 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
889 	    (dev->qdisc &&
890 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
891 	    (dev->ifalias &&
892 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
893 		goto nla_put_failure;
894 
895 	if (1) {
896 		struct rtnl_link_ifmap map = {
897 			.mem_start   = dev->mem_start,
898 			.mem_end     = dev->mem_end,
899 			.base_addr   = dev->base_addr,
900 			.irq         = dev->irq,
901 			.dma         = dev->dma,
902 			.port        = dev->if_port,
903 		};
904 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
905 			goto nla_put_failure;
906 	}
907 
908 	if (dev->addr_len) {
909 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
910 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
911 			goto nla_put_failure;
912 	}
913 
914 	attr = nla_reserve(skb, IFLA_STATS,
915 			sizeof(struct rtnl_link_stats));
916 	if (attr == NULL)
917 		goto nla_put_failure;
918 
919 	stats = dev_get_stats(dev, &temp);
920 	copy_rtnl_link_stats(nla_data(attr), stats);
921 
922 	attr = nla_reserve(skb, IFLA_STATS64,
923 			sizeof(struct rtnl_link_stats64));
924 	if (attr == NULL)
925 		goto nla_put_failure;
926 	copy_rtnl_link_stats64(nla_data(attr), stats);
927 
928 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
929 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
930 		goto nla_put_failure;
931 
932 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
933 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
934 		int i;
935 
936 		struct nlattr *vfinfo, *vf;
937 		int num_vfs = dev_num_vf(dev->dev.parent);
938 
939 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
940 		if (!vfinfo)
941 			goto nla_put_failure;
942 		for (i = 0; i < num_vfs; i++) {
943 			struct ifla_vf_info ivi;
944 			struct ifla_vf_mac vf_mac;
945 			struct ifla_vf_vlan vf_vlan;
946 			struct ifla_vf_tx_rate vf_tx_rate;
947 			struct ifla_vf_spoofchk vf_spoofchk;
948 
949 			/*
950 			 * Not all SR-IOV capable drivers support the
951 			 * spoofcheck query.  Preset to -1 so the user
952 			 * space tool can detect that the driver didn't
953 			 * report anything.
954 			 */
955 			ivi.spoofchk = -1;
956 			memset(ivi.mac, 0, sizeof(ivi.mac));
957 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
958 				break;
959 			vf_mac.vf =
960 				vf_vlan.vf =
961 				vf_tx_rate.vf =
962 				vf_spoofchk.vf = ivi.vf;
963 
964 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
965 			vf_vlan.vlan = ivi.vlan;
966 			vf_vlan.qos = ivi.qos;
967 			vf_tx_rate.rate = ivi.tx_rate;
968 			vf_spoofchk.setting = ivi.spoofchk;
969 			vf = nla_nest_start(skb, IFLA_VF_INFO);
970 			if (!vf) {
971 				nla_nest_cancel(skb, vfinfo);
972 				goto nla_put_failure;
973 			}
974 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
975 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
976 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
977 				    &vf_tx_rate) ||
978 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
979 				    &vf_spoofchk))
980 				goto nla_put_failure;
981 			nla_nest_end(skb, vf);
982 		}
983 		nla_nest_end(skb, vfinfo);
984 	}
985 
986 	if (rtnl_port_fill(skb, dev))
987 		goto nla_put_failure;
988 
989 	if (dev->rtnl_link_ops) {
990 		if (rtnl_link_fill(skb, dev) < 0)
991 			goto nla_put_failure;
992 	}
993 
994 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
995 		goto nla_put_failure;
996 
997 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
998 		if (af_ops->fill_link_af) {
999 			struct nlattr *af;
1000 			int err;
1001 
1002 			if (!(af = nla_nest_start(skb, af_ops->family)))
1003 				goto nla_put_failure;
1004 
1005 			err = af_ops->fill_link_af(skb, dev);
1006 
1007 			/*
1008 			 * Caller may return ENODATA to indicate that there
1009 			 * was no data to be dumped. This is not an error, it
1010 			 * means we should trim the attribute header and
1011 			 * continue.
1012 			 */
1013 			if (err == -ENODATA)
1014 				nla_nest_cancel(skb, af);
1015 			else if (err < 0)
1016 				goto nla_put_failure;
1017 
1018 			nla_nest_end(skb, af);
1019 		}
1020 	}
1021 
1022 	nla_nest_end(skb, af_spec);
1023 
1024 	return nlmsg_end(skb, nlh);
1025 
1026 nla_put_failure:
1027 	nlmsg_cancel(skb, nlh);
1028 	return -EMSGSIZE;
1029 }
1030 
1031 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1032 {
1033 	struct net *net = sock_net(skb->sk);
1034 	int h, s_h;
1035 	int idx = 0, s_idx;
1036 	struct net_device *dev;
1037 	struct hlist_head *head;
1038 	struct nlattr *tb[IFLA_MAX+1];
1039 	u32 ext_filter_mask = 0;
1040 
1041 	s_h = cb->args[0];
1042 	s_idx = cb->args[1];
1043 
1044 	rcu_read_lock();
1045 	cb->seq = net->dev_base_seq;
1046 
1047 	if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1048 			ifla_policy) >= 0) {
1049 
1050 		if (tb[IFLA_EXT_MASK])
1051 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1052 	}
1053 
1054 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1055 		idx = 0;
1056 		head = &net->dev_index_head[h];
1057 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
1058 			if (idx < s_idx)
1059 				goto cont;
1060 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1061 					     NETLINK_CB(cb->skb).portid,
1062 					     cb->nlh->nlmsg_seq, 0,
1063 					     NLM_F_MULTI,
1064 					     ext_filter_mask) <= 0)
1065 				goto out;
1066 
1067 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1068 cont:
1069 			idx++;
1070 		}
1071 	}
1072 out:
1073 	rcu_read_unlock();
1074 	cb->args[1] = idx;
1075 	cb->args[0] = h;
1076 
1077 	return skb->len;
1078 }
1079 
1080 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1081 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1082 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1083 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1084 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1085 	[IFLA_MTU]		= { .type = NLA_U32 },
1086 	[IFLA_LINK]		= { .type = NLA_U32 },
1087 	[IFLA_MASTER]		= { .type = NLA_U32 },
1088 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1089 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1090 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1091 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1092 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1093 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1094 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1095 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1096 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1097 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1098 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1099 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1100 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1101 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1102 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1103 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1104 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1105 };
1106 EXPORT_SYMBOL(ifla_policy);
1107 
1108 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1109 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1110 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1111 };
1112 
1113 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1114 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1115 };
1116 
1117 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1118 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1119 				    .len = sizeof(struct ifla_vf_mac) },
1120 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1121 				    .len = sizeof(struct ifla_vf_vlan) },
1122 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1123 				    .len = sizeof(struct ifla_vf_tx_rate) },
1124 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1125 				    .len = sizeof(struct ifla_vf_spoofchk) },
1126 };
1127 
1128 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1129 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1130 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1131 				    .len = PORT_PROFILE_MAX },
1132 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1133 				    .len = sizeof(struct ifla_port_vsi)},
1134 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1135 				      .len = PORT_UUID_MAX },
1136 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1137 				    .len = PORT_UUID_MAX },
1138 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1139 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1140 };
1141 
1142 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1143 {
1144 	struct net *net;
1145 	/* Examine the link attributes and figure out which
1146 	 * network namespace we are talking about.
1147 	 */
1148 	if (tb[IFLA_NET_NS_PID])
1149 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1150 	else if (tb[IFLA_NET_NS_FD])
1151 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1152 	else
1153 		net = get_net(src_net);
1154 	return net;
1155 }
1156 EXPORT_SYMBOL(rtnl_link_get_net);
1157 
1158 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1159 {
1160 	if (dev) {
1161 		if (tb[IFLA_ADDRESS] &&
1162 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1163 			return -EINVAL;
1164 
1165 		if (tb[IFLA_BROADCAST] &&
1166 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1167 			return -EINVAL;
1168 	}
1169 
1170 	if (tb[IFLA_AF_SPEC]) {
1171 		struct nlattr *af;
1172 		int rem, err;
1173 
1174 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1175 			const struct rtnl_af_ops *af_ops;
1176 
1177 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1178 				return -EAFNOSUPPORT;
1179 
1180 			if (!af_ops->set_link_af)
1181 				return -EOPNOTSUPP;
1182 
1183 			if (af_ops->validate_link_af) {
1184 				err = af_ops->validate_link_af(dev, af);
1185 				if (err < 0)
1186 					return err;
1187 			}
1188 		}
1189 	}
1190 
1191 	return 0;
1192 }
1193 
1194 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1195 {
1196 	int rem, err = -EINVAL;
1197 	struct nlattr *vf;
1198 	const struct net_device_ops *ops = dev->netdev_ops;
1199 
1200 	nla_for_each_nested(vf, attr, rem) {
1201 		switch (nla_type(vf)) {
1202 		case IFLA_VF_MAC: {
1203 			struct ifla_vf_mac *ivm;
1204 			ivm = nla_data(vf);
1205 			err = -EOPNOTSUPP;
1206 			if (ops->ndo_set_vf_mac)
1207 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1208 							  ivm->mac);
1209 			break;
1210 		}
1211 		case IFLA_VF_VLAN: {
1212 			struct ifla_vf_vlan *ivv;
1213 			ivv = nla_data(vf);
1214 			err = -EOPNOTSUPP;
1215 			if (ops->ndo_set_vf_vlan)
1216 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1217 							   ivv->vlan,
1218 							   ivv->qos);
1219 			break;
1220 		}
1221 		case IFLA_VF_TX_RATE: {
1222 			struct ifla_vf_tx_rate *ivt;
1223 			ivt = nla_data(vf);
1224 			err = -EOPNOTSUPP;
1225 			if (ops->ndo_set_vf_tx_rate)
1226 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1227 							      ivt->rate);
1228 			break;
1229 		}
1230 		case IFLA_VF_SPOOFCHK: {
1231 			struct ifla_vf_spoofchk *ivs;
1232 			ivs = nla_data(vf);
1233 			err = -EOPNOTSUPP;
1234 			if (ops->ndo_set_vf_spoofchk)
1235 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1236 							       ivs->setting);
1237 			break;
1238 		}
1239 		default:
1240 			err = -EINVAL;
1241 			break;
1242 		}
1243 		if (err)
1244 			break;
1245 	}
1246 	return err;
1247 }
1248 
1249 static int do_set_master(struct net_device *dev, int ifindex)
1250 {
1251 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1252 	const struct net_device_ops *ops;
1253 	int err;
1254 
1255 	if (upper_dev) {
1256 		if (upper_dev->ifindex == ifindex)
1257 			return 0;
1258 		ops = upper_dev->netdev_ops;
1259 		if (ops->ndo_del_slave) {
1260 			err = ops->ndo_del_slave(upper_dev, dev);
1261 			if (err)
1262 				return err;
1263 		} else {
1264 			return -EOPNOTSUPP;
1265 		}
1266 	}
1267 
1268 	if (ifindex) {
1269 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1270 		if (!upper_dev)
1271 			return -EINVAL;
1272 		ops = upper_dev->netdev_ops;
1273 		if (ops->ndo_add_slave) {
1274 			err = ops->ndo_add_slave(upper_dev, dev);
1275 			if (err)
1276 				return err;
1277 		} else {
1278 			return -EOPNOTSUPP;
1279 		}
1280 	}
1281 	return 0;
1282 }
1283 
1284 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1285 		      struct nlattr **tb, char *ifname, int modified)
1286 {
1287 	const struct net_device_ops *ops = dev->netdev_ops;
1288 	int err;
1289 
1290 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1291 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1292 		if (IS_ERR(net)) {
1293 			err = PTR_ERR(net);
1294 			goto errout;
1295 		}
1296 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1297 			err = -EPERM;
1298 			goto errout;
1299 		}
1300 		err = dev_change_net_namespace(dev, net, ifname);
1301 		put_net(net);
1302 		if (err)
1303 			goto errout;
1304 		modified = 1;
1305 	}
1306 
1307 	if (tb[IFLA_MAP]) {
1308 		struct rtnl_link_ifmap *u_map;
1309 		struct ifmap k_map;
1310 
1311 		if (!ops->ndo_set_config) {
1312 			err = -EOPNOTSUPP;
1313 			goto errout;
1314 		}
1315 
1316 		if (!netif_device_present(dev)) {
1317 			err = -ENODEV;
1318 			goto errout;
1319 		}
1320 
1321 		u_map = nla_data(tb[IFLA_MAP]);
1322 		k_map.mem_start = (unsigned long) u_map->mem_start;
1323 		k_map.mem_end = (unsigned long) u_map->mem_end;
1324 		k_map.base_addr = (unsigned short) u_map->base_addr;
1325 		k_map.irq = (unsigned char) u_map->irq;
1326 		k_map.dma = (unsigned char) u_map->dma;
1327 		k_map.port = (unsigned char) u_map->port;
1328 
1329 		err = ops->ndo_set_config(dev, &k_map);
1330 		if (err < 0)
1331 			goto errout;
1332 
1333 		modified = 1;
1334 	}
1335 
1336 	if (tb[IFLA_ADDRESS]) {
1337 		struct sockaddr *sa;
1338 		int len;
1339 
1340 		len = sizeof(sa_family_t) + dev->addr_len;
1341 		sa = kmalloc(len, GFP_KERNEL);
1342 		if (!sa) {
1343 			err = -ENOMEM;
1344 			goto errout;
1345 		}
1346 		sa->sa_family = dev->type;
1347 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1348 		       dev->addr_len);
1349 		err = dev_set_mac_address(dev, sa);
1350 		kfree(sa);
1351 		if (err)
1352 			goto errout;
1353 		modified = 1;
1354 	}
1355 
1356 	if (tb[IFLA_MTU]) {
1357 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1358 		if (err < 0)
1359 			goto errout;
1360 		modified = 1;
1361 	}
1362 
1363 	if (tb[IFLA_GROUP]) {
1364 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1365 		modified = 1;
1366 	}
1367 
1368 	/*
1369 	 * Interface selected by interface index but interface
1370 	 * name provided implies that a name change has been
1371 	 * requested.
1372 	 */
1373 	if (ifm->ifi_index > 0 && ifname[0]) {
1374 		err = dev_change_name(dev, ifname);
1375 		if (err < 0)
1376 			goto errout;
1377 		modified = 1;
1378 	}
1379 
1380 	if (tb[IFLA_IFALIAS]) {
1381 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1382 				    nla_len(tb[IFLA_IFALIAS]));
1383 		if (err < 0)
1384 			goto errout;
1385 		modified = 1;
1386 	}
1387 
1388 	if (tb[IFLA_BROADCAST]) {
1389 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1390 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1391 	}
1392 
1393 	if (ifm->ifi_flags || ifm->ifi_change) {
1394 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1395 		if (err < 0)
1396 			goto errout;
1397 	}
1398 
1399 	if (tb[IFLA_MASTER]) {
1400 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1401 		if (err)
1402 			goto errout;
1403 		modified = 1;
1404 	}
1405 
1406 	if (tb[IFLA_CARRIER]) {
1407 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1408 		if (err)
1409 			goto errout;
1410 		modified = 1;
1411 	}
1412 
1413 	if (tb[IFLA_TXQLEN])
1414 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1415 
1416 	if (tb[IFLA_OPERSTATE])
1417 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1418 
1419 	if (tb[IFLA_LINKMODE]) {
1420 		write_lock_bh(&dev_base_lock);
1421 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1422 		write_unlock_bh(&dev_base_lock);
1423 	}
1424 
1425 	if (tb[IFLA_VFINFO_LIST]) {
1426 		struct nlattr *attr;
1427 		int rem;
1428 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1429 			if (nla_type(attr) != IFLA_VF_INFO) {
1430 				err = -EINVAL;
1431 				goto errout;
1432 			}
1433 			err = do_setvfinfo(dev, attr);
1434 			if (err < 0)
1435 				goto errout;
1436 			modified = 1;
1437 		}
1438 	}
1439 	err = 0;
1440 
1441 	if (tb[IFLA_VF_PORTS]) {
1442 		struct nlattr *port[IFLA_PORT_MAX+1];
1443 		struct nlattr *attr;
1444 		int vf;
1445 		int rem;
1446 
1447 		err = -EOPNOTSUPP;
1448 		if (!ops->ndo_set_vf_port)
1449 			goto errout;
1450 
1451 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1452 			if (nla_type(attr) != IFLA_VF_PORT)
1453 				continue;
1454 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1455 				attr, ifla_port_policy);
1456 			if (err < 0)
1457 				goto errout;
1458 			if (!port[IFLA_PORT_VF]) {
1459 				err = -EOPNOTSUPP;
1460 				goto errout;
1461 			}
1462 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1463 			err = ops->ndo_set_vf_port(dev, vf, port);
1464 			if (err < 0)
1465 				goto errout;
1466 			modified = 1;
1467 		}
1468 	}
1469 	err = 0;
1470 
1471 	if (tb[IFLA_PORT_SELF]) {
1472 		struct nlattr *port[IFLA_PORT_MAX+1];
1473 
1474 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1475 			tb[IFLA_PORT_SELF], ifla_port_policy);
1476 		if (err < 0)
1477 			goto errout;
1478 
1479 		err = -EOPNOTSUPP;
1480 		if (ops->ndo_set_vf_port)
1481 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1482 		if (err < 0)
1483 			goto errout;
1484 		modified = 1;
1485 	}
1486 
1487 	if (tb[IFLA_AF_SPEC]) {
1488 		struct nlattr *af;
1489 		int rem;
1490 
1491 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1492 			const struct rtnl_af_ops *af_ops;
1493 
1494 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1495 				BUG();
1496 
1497 			err = af_ops->set_link_af(dev, af);
1498 			if (err < 0)
1499 				goto errout;
1500 
1501 			modified = 1;
1502 		}
1503 	}
1504 	err = 0;
1505 
1506 errout:
1507 	if (err < 0 && modified)
1508 		net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
1509 				     dev->name);
1510 
1511 	return err;
1512 }
1513 
1514 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1515 {
1516 	struct net *net = sock_net(skb->sk);
1517 	struct ifinfomsg *ifm;
1518 	struct net_device *dev;
1519 	int err;
1520 	struct nlattr *tb[IFLA_MAX+1];
1521 	char ifname[IFNAMSIZ];
1522 
1523 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1524 	if (err < 0)
1525 		goto errout;
1526 
1527 	if (tb[IFLA_IFNAME])
1528 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1529 	else
1530 		ifname[0] = '\0';
1531 
1532 	err = -EINVAL;
1533 	ifm = nlmsg_data(nlh);
1534 	if (ifm->ifi_index > 0)
1535 		dev = __dev_get_by_index(net, ifm->ifi_index);
1536 	else if (tb[IFLA_IFNAME])
1537 		dev = __dev_get_by_name(net, ifname);
1538 	else
1539 		goto errout;
1540 
1541 	if (dev == NULL) {
1542 		err = -ENODEV;
1543 		goto errout;
1544 	}
1545 
1546 	err = validate_linkmsg(dev, tb);
1547 	if (err < 0)
1548 		goto errout;
1549 
1550 	err = do_setlink(dev, ifm, tb, ifname, 0);
1551 errout:
1552 	return err;
1553 }
1554 
1555 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1556 {
1557 	struct net *net = sock_net(skb->sk);
1558 	const struct rtnl_link_ops *ops;
1559 	struct net_device *dev;
1560 	struct ifinfomsg *ifm;
1561 	char ifname[IFNAMSIZ];
1562 	struct nlattr *tb[IFLA_MAX+1];
1563 	int err;
1564 	LIST_HEAD(list_kill);
1565 
1566 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1567 	if (err < 0)
1568 		return err;
1569 
1570 	if (tb[IFLA_IFNAME])
1571 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1572 
1573 	ifm = nlmsg_data(nlh);
1574 	if (ifm->ifi_index > 0)
1575 		dev = __dev_get_by_index(net, ifm->ifi_index);
1576 	else if (tb[IFLA_IFNAME])
1577 		dev = __dev_get_by_name(net, ifname);
1578 	else
1579 		return -EINVAL;
1580 
1581 	if (!dev)
1582 		return -ENODEV;
1583 
1584 	ops = dev->rtnl_link_ops;
1585 	if (!ops)
1586 		return -EOPNOTSUPP;
1587 
1588 	ops->dellink(dev, &list_kill);
1589 	unregister_netdevice_many(&list_kill);
1590 	list_del(&list_kill);
1591 	return 0;
1592 }
1593 
1594 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1595 {
1596 	unsigned int old_flags;
1597 	int err;
1598 
1599 	old_flags = dev->flags;
1600 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1601 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1602 		if (err < 0)
1603 			return err;
1604 	}
1605 
1606 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1607 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1608 
1609 	__dev_notify_flags(dev, old_flags);
1610 	return 0;
1611 }
1612 EXPORT_SYMBOL(rtnl_configure_link);
1613 
1614 struct net_device *rtnl_create_link(struct net *net,
1615 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1616 {
1617 	int err;
1618 	struct net_device *dev;
1619 	unsigned int num_tx_queues = 1;
1620 	unsigned int num_rx_queues = 1;
1621 
1622 	if (tb[IFLA_NUM_TX_QUEUES])
1623 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1624 	else if (ops->get_num_tx_queues)
1625 		num_tx_queues = ops->get_num_tx_queues();
1626 
1627 	if (tb[IFLA_NUM_RX_QUEUES])
1628 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1629 	else if (ops->get_num_rx_queues)
1630 		num_rx_queues = ops->get_num_rx_queues();
1631 
1632 	err = -ENOMEM;
1633 	dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1634 			       num_tx_queues, num_rx_queues);
1635 	if (!dev)
1636 		goto err;
1637 
1638 	dev_net_set(dev, net);
1639 	dev->rtnl_link_ops = ops;
1640 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1641 
1642 	if (tb[IFLA_MTU])
1643 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1644 	if (tb[IFLA_ADDRESS]) {
1645 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1646 				nla_len(tb[IFLA_ADDRESS]));
1647 		dev->addr_assign_type = NET_ADDR_SET;
1648 	}
1649 	if (tb[IFLA_BROADCAST])
1650 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1651 				nla_len(tb[IFLA_BROADCAST]));
1652 	if (tb[IFLA_TXQLEN])
1653 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1654 	if (tb[IFLA_OPERSTATE])
1655 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1656 	if (tb[IFLA_LINKMODE])
1657 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1658 	if (tb[IFLA_GROUP])
1659 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1660 
1661 	return dev;
1662 
1663 err:
1664 	return ERR_PTR(err);
1665 }
1666 EXPORT_SYMBOL(rtnl_create_link);
1667 
1668 static int rtnl_group_changelink(struct net *net, int group,
1669 		struct ifinfomsg *ifm,
1670 		struct nlattr **tb)
1671 {
1672 	struct net_device *dev;
1673 	int err;
1674 
1675 	for_each_netdev(net, dev) {
1676 		if (dev->group == group) {
1677 			err = do_setlink(dev, ifm, tb, NULL, 0);
1678 			if (err < 0)
1679 				return err;
1680 		}
1681 	}
1682 
1683 	return 0;
1684 }
1685 
1686 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1687 {
1688 	struct net *net = sock_net(skb->sk);
1689 	const struct rtnl_link_ops *ops;
1690 	struct net_device *dev;
1691 	struct ifinfomsg *ifm;
1692 	char kind[MODULE_NAME_LEN];
1693 	char ifname[IFNAMSIZ];
1694 	struct nlattr *tb[IFLA_MAX+1];
1695 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1696 	int err;
1697 
1698 #ifdef CONFIG_MODULES
1699 replay:
1700 #endif
1701 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1702 	if (err < 0)
1703 		return err;
1704 
1705 	if (tb[IFLA_IFNAME])
1706 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1707 	else
1708 		ifname[0] = '\0';
1709 
1710 	ifm = nlmsg_data(nlh);
1711 	if (ifm->ifi_index > 0)
1712 		dev = __dev_get_by_index(net, ifm->ifi_index);
1713 	else {
1714 		if (ifname[0])
1715 			dev = __dev_get_by_name(net, ifname);
1716 		else
1717 			dev = NULL;
1718 	}
1719 
1720 	err = validate_linkmsg(dev, tb);
1721 	if (err < 0)
1722 		return err;
1723 
1724 	if (tb[IFLA_LINKINFO]) {
1725 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1726 				       tb[IFLA_LINKINFO], ifla_info_policy);
1727 		if (err < 0)
1728 			return err;
1729 	} else
1730 		memset(linkinfo, 0, sizeof(linkinfo));
1731 
1732 	if (linkinfo[IFLA_INFO_KIND]) {
1733 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1734 		ops = rtnl_link_ops_get(kind);
1735 	} else {
1736 		kind[0] = '\0';
1737 		ops = NULL;
1738 	}
1739 
1740 	if (1) {
1741 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1742 		struct net *dest_net;
1743 
1744 		if (ops) {
1745 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1746 				err = nla_parse_nested(attr, ops->maxtype,
1747 						       linkinfo[IFLA_INFO_DATA],
1748 						       ops->policy);
1749 				if (err < 0)
1750 					return err;
1751 				data = attr;
1752 			}
1753 			if (ops->validate) {
1754 				err = ops->validate(tb, data);
1755 				if (err < 0)
1756 					return err;
1757 			}
1758 		}
1759 
1760 		if (dev) {
1761 			int modified = 0;
1762 
1763 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1764 				return -EEXIST;
1765 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1766 				return -EOPNOTSUPP;
1767 
1768 			if (linkinfo[IFLA_INFO_DATA]) {
1769 				if (!ops || ops != dev->rtnl_link_ops ||
1770 				    !ops->changelink)
1771 					return -EOPNOTSUPP;
1772 
1773 				err = ops->changelink(dev, tb, data);
1774 				if (err < 0)
1775 					return err;
1776 				modified = 1;
1777 			}
1778 
1779 			return do_setlink(dev, ifm, tb, ifname, modified);
1780 		}
1781 
1782 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1783 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1784 				return rtnl_group_changelink(net,
1785 						nla_get_u32(tb[IFLA_GROUP]),
1786 						ifm, tb);
1787 			return -ENODEV;
1788 		}
1789 
1790 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1791 			return -EOPNOTSUPP;
1792 
1793 		if (!ops) {
1794 #ifdef CONFIG_MODULES
1795 			if (kind[0]) {
1796 				__rtnl_unlock();
1797 				request_module("rtnl-link-%s", kind);
1798 				rtnl_lock();
1799 				ops = rtnl_link_ops_get(kind);
1800 				if (ops)
1801 					goto replay;
1802 			}
1803 #endif
1804 			return -EOPNOTSUPP;
1805 		}
1806 
1807 		if (!ifname[0])
1808 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1809 
1810 		dest_net = rtnl_link_get_net(net, tb);
1811 		if (IS_ERR(dest_net))
1812 			return PTR_ERR(dest_net);
1813 
1814 		dev = rtnl_create_link(dest_net, ifname, ops, tb);
1815 		if (IS_ERR(dev)) {
1816 			err = PTR_ERR(dev);
1817 			goto out;
1818 		}
1819 
1820 		dev->ifindex = ifm->ifi_index;
1821 
1822 		if (ops->newlink)
1823 			err = ops->newlink(net, dev, tb, data);
1824 		else
1825 			err = register_netdevice(dev);
1826 
1827 		if (err < 0 && !IS_ERR(dev))
1828 			free_netdev(dev);
1829 		if (err < 0)
1830 			goto out;
1831 
1832 		err = rtnl_configure_link(dev, ifm);
1833 		if (err < 0)
1834 			unregister_netdevice(dev);
1835 out:
1836 		put_net(dest_net);
1837 		return err;
1838 	}
1839 }
1840 
1841 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1842 {
1843 	struct net *net = sock_net(skb->sk);
1844 	struct ifinfomsg *ifm;
1845 	char ifname[IFNAMSIZ];
1846 	struct nlattr *tb[IFLA_MAX+1];
1847 	struct net_device *dev = NULL;
1848 	struct sk_buff *nskb;
1849 	int err;
1850 	u32 ext_filter_mask = 0;
1851 
1852 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1853 	if (err < 0)
1854 		return err;
1855 
1856 	if (tb[IFLA_IFNAME])
1857 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1858 
1859 	if (tb[IFLA_EXT_MASK])
1860 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1861 
1862 	ifm = nlmsg_data(nlh);
1863 	if (ifm->ifi_index > 0)
1864 		dev = __dev_get_by_index(net, ifm->ifi_index);
1865 	else if (tb[IFLA_IFNAME])
1866 		dev = __dev_get_by_name(net, ifname);
1867 	else
1868 		return -EINVAL;
1869 
1870 	if (dev == NULL)
1871 		return -ENODEV;
1872 
1873 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1874 	if (nskb == NULL)
1875 		return -ENOBUFS;
1876 
1877 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1878 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1879 	if (err < 0) {
1880 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1881 		WARN_ON(err == -EMSGSIZE);
1882 		kfree_skb(nskb);
1883 	} else
1884 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1885 
1886 	return err;
1887 }
1888 
1889 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1890 {
1891 	struct net *net = sock_net(skb->sk);
1892 	struct net_device *dev;
1893 	struct nlattr *tb[IFLA_MAX+1];
1894 	u32 ext_filter_mask = 0;
1895 	u16 min_ifinfo_dump_size = 0;
1896 
1897 	if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1898 			ifla_policy) >= 0) {
1899 		if (tb[IFLA_EXT_MASK])
1900 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1901 	}
1902 
1903 	if (!ext_filter_mask)
1904 		return NLMSG_GOODSIZE;
1905 	/*
1906 	 * traverse the list of net devices and compute the minimum
1907 	 * buffer size based upon the filter mask.
1908 	 */
1909 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1910 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1911 					     if_nlmsg_size(dev,
1912 						           ext_filter_mask));
1913 	}
1914 
1915 	return min_ifinfo_dump_size;
1916 }
1917 
1918 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1919 {
1920 	int idx;
1921 	int s_idx = cb->family;
1922 
1923 	if (s_idx == 0)
1924 		s_idx = 1;
1925 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1926 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1927 		if (idx < s_idx || idx == PF_PACKET)
1928 			continue;
1929 		if (rtnl_msg_handlers[idx] == NULL ||
1930 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1931 			continue;
1932 		if (idx > s_idx)
1933 			memset(&cb->args[0], 0, sizeof(cb->args));
1934 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1935 			break;
1936 	}
1937 	cb->family = idx;
1938 
1939 	return skb->len;
1940 }
1941 
1942 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1943 {
1944 	struct net *net = dev_net(dev);
1945 	struct sk_buff *skb;
1946 	int err = -ENOBUFS;
1947 	size_t if_info_size;
1948 
1949 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1950 	if (skb == NULL)
1951 		goto errout;
1952 
1953 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1954 	if (err < 0) {
1955 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1956 		WARN_ON(err == -EMSGSIZE);
1957 		kfree_skb(skb);
1958 		goto errout;
1959 	}
1960 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1961 	return;
1962 errout:
1963 	if (err < 0)
1964 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1965 }
1966 EXPORT_SYMBOL(rtmsg_ifinfo);
1967 
1968 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1969 				   struct net_device *dev,
1970 				   u8 *addr, u32 pid, u32 seq,
1971 				   int type, unsigned int flags)
1972 {
1973 	struct nlmsghdr *nlh;
1974 	struct ndmsg *ndm;
1975 
1976 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1977 	if (!nlh)
1978 		return -EMSGSIZE;
1979 
1980 	ndm = nlmsg_data(nlh);
1981 	ndm->ndm_family  = AF_BRIDGE;
1982 	ndm->ndm_pad1	 = 0;
1983 	ndm->ndm_pad2    = 0;
1984 	ndm->ndm_flags	 = flags;
1985 	ndm->ndm_type	 = 0;
1986 	ndm->ndm_ifindex = dev->ifindex;
1987 	ndm->ndm_state   = NUD_PERMANENT;
1988 
1989 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
1990 		goto nla_put_failure;
1991 
1992 	return nlmsg_end(skb, nlh);
1993 
1994 nla_put_failure:
1995 	nlmsg_cancel(skb, nlh);
1996 	return -EMSGSIZE;
1997 }
1998 
1999 static inline size_t rtnl_fdb_nlmsg_size(void)
2000 {
2001 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2002 }
2003 
2004 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2005 {
2006 	struct net *net = dev_net(dev);
2007 	struct sk_buff *skb;
2008 	int err = -ENOBUFS;
2009 
2010 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2011 	if (!skb)
2012 		goto errout;
2013 
2014 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2015 	if (err < 0) {
2016 		kfree_skb(skb);
2017 		goto errout;
2018 	}
2019 
2020 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2021 	return;
2022 errout:
2023 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2024 }
2025 
2026 /**
2027  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2028  */
2029 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2030 		     struct nlattr *tb[],
2031 		     struct net_device *dev,
2032 		     const unsigned char *addr,
2033 		     u16 flags)
2034 {
2035 	int err = -EINVAL;
2036 
2037 	/* If aging addresses are supported device will need to
2038 	 * implement its own handler for this.
2039 	 */
2040 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2041 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2042 		return err;
2043 	}
2044 
2045 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2046 		err = dev_uc_add_excl(dev, addr);
2047 	else if (is_multicast_ether_addr(addr))
2048 		err = dev_mc_add_excl(dev, addr);
2049 
2050 	/* Only return duplicate errors if NLM_F_EXCL is set */
2051 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2052 		err = 0;
2053 
2054 	return err;
2055 }
2056 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2057 
2058 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2059 {
2060 	struct net *net = sock_net(skb->sk);
2061 	struct ndmsg *ndm;
2062 	struct nlattr *tb[NDA_MAX+1];
2063 	struct net_device *dev;
2064 	u8 *addr;
2065 	int err;
2066 
2067 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2068 	if (err < 0)
2069 		return err;
2070 
2071 	ndm = nlmsg_data(nlh);
2072 	if (ndm->ndm_ifindex == 0) {
2073 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2074 		return -EINVAL;
2075 	}
2076 
2077 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2078 	if (dev == NULL) {
2079 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2080 		return -ENODEV;
2081 	}
2082 
2083 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2084 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2085 		return -EINVAL;
2086 	}
2087 
2088 	addr = nla_data(tb[NDA_LLADDR]);
2089 	if (is_zero_ether_addr(addr)) {
2090 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2091 		return -EINVAL;
2092 	}
2093 
2094 	err = -EOPNOTSUPP;
2095 
2096 	/* Support fdb on master device the net/bridge default case */
2097 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2098 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2099 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2100 		const struct net_device_ops *ops = br_dev->netdev_ops;
2101 
2102 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2103 		if (err)
2104 			goto out;
2105 		else
2106 			ndm->ndm_flags &= ~NTF_MASTER;
2107 	}
2108 
2109 	/* Embedded bridge, macvlan, and any other device support */
2110 	if ((ndm->ndm_flags & NTF_SELF)) {
2111 		if (dev->netdev_ops->ndo_fdb_add)
2112 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2113 							   nlh->nlmsg_flags);
2114 		else
2115 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2116 					       nlh->nlmsg_flags);
2117 
2118 		if (!err) {
2119 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2120 			ndm->ndm_flags &= ~NTF_SELF;
2121 		}
2122 	}
2123 out:
2124 	return err;
2125 }
2126 
2127 /**
2128  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2129  */
2130 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2131 		     struct nlattr *tb[],
2132 		     struct net_device *dev,
2133 		     const unsigned char *addr)
2134 {
2135 	int err = -EOPNOTSUPP;
2136 
2137 	/* If aging addresses are supported device will need to
2138 	 * implement its own handler for this.
2139 	 */
2140 	if (ndm->ndm_state & NUD_PERMANENT) {
2141 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2142 		return -EINVAL;
2143 	}
2144 
2145 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2146 		err = dev_uc_del(dev, addr);
2147 	else if (is_multicast_ether_addr(addr))
2148 		err = dev_mc_del(dev, addr);
2149 	else
2150 		err = -EINVAL;
2151 
2152 	return err;
2153 }
2154 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2155 
2156 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2157 {
2158 	struct net *net = sock_net(skb->sk);
2159 	struct ndmsg *ndm;
2160 	struct nlattr *tb[NDA_MAX+1];
2161 	struct net_device *dev;
2162 	int err = -EINVAL;
2163 	__u8 *addr;
2164 
2165 	if (!capable(CAP_NET_ADMIN))
2166 		return -EPERM;
2167 
2168 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2169 	if (err < 0)
2170 		return err;
2171 
2172 	ndm = nlmsg_data(nlh);
2173 	if (ndm->ndm_ifindex == 0) {
2174 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2175 		return -EINVAL;
2176 	}
2177 
2178 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2179 	if (dev == NULL) {
2180 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2181 		return -ENODEV;
2182 	}
2183 
2184 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2185 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2186 		return -EINVAL;
2187 	}
2188 
2189 	addr = nla_data(tb[NDA_LLADDR]);
2190 	if (!is_valid_ether_addr(addr)) {
2191 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2192 		return -EINVAL;
2193 	}
2194 
2195 	err = -EOPNOTSUPP;
2196 
2197 	/* Support fdb on master device the net/bridge default case */
2198 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2199 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2200 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2201 		const struct net_device_ops *ops = br_dev->netdev_ops;
2202 
2203 		if (ops->ndo_fdb_del)
2204 			err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2205 
2206 		if (err)
2207 			goto out;
2208 		else
2209 			ndm->ndm_flags &= ~NTF_MASTER;
2210 	}
2211 
2212 	/* Embedded bridge, macvlan, and any other device support */
2213 	if (ndm->ndm_flags & NTF_SELF) {
2214 		if (dev->netdev_ops->ndo_fdb_del)
2215 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2216 		else
2217 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2218 
2219 		if (!err) {
2220 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2221 			ndm->ndm_flags &= ~NTF_SELF;
2222 		}
2223 	}
2224 out:
2225 	return err;
2226 }
2227 
2228 static int nlmsg_populate_fdb(struct sk_buff *skb,
2229 			      struct netlink_callback *cb,
2230 			      struct net_device *dev,
2231 			      int *idx,
2232 			      struct netdev_hw_addr_list *list)
2233 {
2234 	struct netdev_hw_addr *ha;
2235 	int err;
2236 	u32 portid, seq;
2237 
2238 	portid = NETLINK_CB(cb->skb).portid;
2239 	seq = cb->nlh->nlmsg_seq;
2240 
2241 	list_for_each_entry(ha, &list->list, list) {
2242 		if (*idx < cb->args[0])
2243 			goto skip;
2244 
2245 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2246 					      portid, seq,
2247 					      RTM_NEWNEIGH, NTF_SELF);
2248 		if (err < 0)
2249 			return err;
2250 skip:
2251 		*idx += 1;
2252 	}
2253 	return 0;
2254 }
2255 
2256 /**
2257  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2258  * @nlh: netlink message header
2259  * @dev: netdevice
2260  *
2261  * Default netdevice operation to dump the existing unicast address list.
2262  * Returns zero on success.
2263  */
2264 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2265 		      struct netlink_callback *cb,
2266 		      struct net_device *dev,
2267 		      int idx)
2268 {
2269 	int err;
2270 
2271 	netif_addr_lock_bh(dev);
2272 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2273 	if (err)
2274 		goto out;
2275 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2276 out:
2277 	netif_addr_unlock_bh(dev);
2278 	return idx;
2279 }
2280 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2281 
2282 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2283 {
2284 	int idx = 0;
2285 	struct net *net = sock_net(skb->sk);
2286 	struct net_device *dev;
2287 
2288 	rcu_read_lock();
2289 	for_each_netdev_rcu(net, dev) {
2290 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2291 			struct net_device *br_dev;
2292 			const struct net_device_ops *ops;
2293 
2294 			br_dev = netdev_master_upper_dev_get(dev);
2295 			ops = br_dev->netdev_ops;
2296 			if (ops->ndo_fdb_dump)
2297 				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2298 		}
2299 
2300 		if (dev->netdev_ops->ndo_fdb_dump)
2301 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2302 		else
2303 			ndo_dflt_fdb_dump(skb, cb, dev, idx);
2304 	}
2305 	rcu_read_unlock();
2306 
2307 	cb->args[0] = idx;
2308 	return skb->len;
2309 }
2310 
2311 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2312 			    struct net_device *dev, u16 mode)
2313 {
2314 	struct nlmsghdr *nlh;
2315 	struct ifinfomsg *ifm;
2316 	struct nlattr *br_afspec;
2317 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2318 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2319 
2320 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2321 	if (nlh == NULL)
2322 		return -EMSGSIZE;
2323 
2324 	ifm = nlmsg_data(nlh);
2325 	ifm->ifi_family = AF_BRIDGE;
2326 	ifm->__ifi_pad = 0;
2327 	ifm->ifi_type = dev->type;
2328 	ifm->ifi_index = dev->ifindex;
2329 	ifm->ifi_flags = dev_get_flags(dev);
2330 	ifm->ifi_change = 0;
2331 
2332 
2333 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2334 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2335 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2336 	    (br_dev &&
2337 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2338 	    (dev->addr_len &&
2339 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2340 	    (dev->ifindex != dev->iflink &&
2341 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2342 		goto nla_put_failure;
2343 
2344 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2345 	if (!br_afspec)
2346 		goto nla_put_failure;
2347 
2348 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2349 	    nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2350 		nla_nest_cancel(skb, br_afspec);
2351 		goto nla_put_failure;
2352 	}
2353 	nla_nest_end(skb, br_afspec);
2354 
2355 	return nlmsg_end(skb, nlh);
2356 nla_put_failure:
2357 	nlmsg_cancel(skb, nlh);
2358 	return -EMSGSIZE;
2359 }
2360 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2361 
2362 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2363 {
2364 	struct net *net = sock_net(skb->sk);
2365 	struct net_device *dev;
2366 	int idx = 0;
2367 	u32 portid = NETLINK_CB(cb->skb).portid;
2368 	u32 seq = cb->nlh->nlmsg_seq;
2369 	struct nlattr *extfilt;
2370 	u32 filter_mask = 0;
2371 
2372 	extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct rtgenmsg),
2373 				  IFLA_EXT_MASK);
2374 	if (extfilt)
2375 		filter_mask = nla_get_u32(extfilt);
2376 
2377 	rcu_read_lock();
2378 	for_each_netdev_rcu(net, dev) {
2379 		const struct net_device_ops *ops = dev->netdev_ops;
2380 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2381 
2382 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2383 			if (idx >= cb->args[0] &&
2384 			    br_dev->netdev_ops->ndo_bridge_getlink(
2385 				    skb, portid, seq, dev, filter_mask) < 0)
2386 				break;
2387 			idx++;
2388 		}
2389 
2390 		if (ops->ndo_bridge_getlink) {
2391 			if (idx >= cb->args[0] &&
2392 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2393 						    filter_mask) < 0)
2394 				break;
2395 			idx++;
2396 		}
2397 	}
2398 	rcu_read_unlock();
2399 	cb->args[0] = idx;
2400 
2401 	return skb->len;
2402 }
2403 
2404 static inline size_t bridge_nlmsg_size(void)
2405 {
2406 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2407 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2408 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2409 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2410 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2411 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2412 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2413 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2414 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2415 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2416 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2417 }
2418 
2419 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2420 {
2421 	struct net *net = dev_net(dev);
2422 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2423 	struct sk_buff *skb;
2424 	int err = -EOPNOTSUPP;
2425 
2426 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2427 	if (!skb) {
2428 		err = -ENOMEM;
2429 		goto errout;
2430 	}
2431 
2432 	if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2433 	    br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2434 		err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2435 		if (err < 0)
2436 			goto errout;
2437 	}
2438 
2439 	if ((flags & BRIDGE_FLAGS_SELF) &&
2440 	    dev->netdev_ops->ndo_bridge_getlink) {
2441 		err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2442 		if (err < 0)
2443 			goto errout;
2444 	}
2445 
2446 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2447 	return 0;
2448 errout:
2449 	WARN_ON(err == -EMSGSIZE);
2450 	kfree_skb(skb);
2451 	rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2452 	return err;
2453 }
2454 
2455 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2456 {
2457 	struct net *net = sock_net(skb->sk);
2458 	struct ifinfomsg *ifm;
2459 	struct net_device *dev;
2460 	struct nlattr *br_spec, *attr = NULL;
2461 	int rem, err = -EOPNOTSUPP;
2462 	u16 oflags, flags = 0;
2463 	bool have_flags = false;
2464 
2465 	if (nlmsg_len(nlh) < sizeof(*ifm))
2466 		return -EINVAL;
2467 
2468 	ifm = nlmsg_data(nlh);
2469 	if (ifm->ifi_family != AF_BRIDGE)
2470 		return -EPFNOSUPPORT;
2471 
2472 	dev = __dev_get_by_index(net, ifm->ifi_index);
2473 	if (!dev) {
2474 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2475 		return -ENODEV;
2476 	}
2477 
2478 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2479 	if (br_spec) {
2480 		nla_for_each_nested(attr, br_spec, rem) {
2481 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2482 				have_flags = true;
2483 				flags = nla_get_u16(attr);
2484 				break;
2485 			}
2486 		}
2487 	}
2488 
2489 	oflags = flags;
2490 
2491 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2492 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2493 
2494 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2495 			err = -EOPNOTSUPP;
2496 			goto out;
2497 		}
2498 
2499 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2500 		if (err)
2501 			goto out;
2502 
2503 		flags &= ~BRIDGE_FLAGS_MASTER;
2504 	}
2505 
2506 	if ((flags & BRIDGE_FLAGS_SELF)) {
2507 		if (!dev->netdev_ops->ndo_bridge_setlink)
2508 			err = -EOPNOTSUPP;
2509 		else
2510 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2511 
2512 		if (!err)
2513 			flags &= ~BRIDGE_FLAGS_SELF;
2514 	}
2515 
2516 	if (have_flags)
2517 		memcpy(nla_data(attr), &flags, sizeof(flags));
2518 	/* Generate event to notify upper layer of bridge change */
2519 	if (!err)
2520 		err = rtnl_bridge_notify(dev, oflags);
2521 out:
2522 	return err;
2523 }
2524 
2525 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2526 {
2527 	struct net *net = sock_net(skb->sk);
2528 	struct ifinfomsg *ifm;
2529 	struct net_device *dev;
2530 	struct nlattr *br_spec, *attr = NULL;
2531 	int rem, err = -EOPNOTSUPP;
2532 	u16 oflags, flags = 0;
2533 	bool have_flags = false;
2534 
2535 	if (nlmsg_len(nlh) < sizeof(*ifm))
2536 		return -EINVAL;
2537 
2538 	ifm = nlmsg_data(nlh);
2539 	if (ifm->ifi_family != AF_BRIDGE)
2540 		return -EPFNOSUPPORT;
2541 
2542 	dev = __dev_get_by_index(net, ifm->ifi_index);
2543 	if (!dev) {
2544 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2545 		return -ENODEV;
2546 	}
2547 
2548 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2549 	if (br_spec) {
2550 		nla_for_each_nested(attr, br_spec, rem) {
2551 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2552 				have_flags = true;
2553 				flags = nla_get_u16(attr);
2554 				break;
2555 			}
2556 		}
2557 	}
2558 
2559 	oflags = flags;
2560 
2561 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2562 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2563 
2564 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2565 			err = -EOPNOTSUPP;
2566 			goto out;
2567 		}
2568 
2569 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2570 		if (err)
2571 			goto out;
2572 
2573 		flags &= ~BRIDGE_FLAGS_MASTER;
2574 	}
2575 
2576 	if ((flags & BRIDGE_FLAGS_SELF)) {
2577 		if (!dev->netdev_ops->ndo_bridge_dellink)
2578 			err = -EOPNOTSUPP;
2579 		else
2580 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2581 
2582 		if (!err)
2583 			flags &= ~BRIDGE_FLAGS_SELF;
2584 	}
2585 
2586 	if (have_flags)
2587 		memcpy(nla_data(attr), &flags, sizeof(flags));
2588 	/* Generate event to notify upper layer of bridge change */
2589 	if (!err)
2590 		err = rtnl_bridge_notify(dev, oflags);
2591 out:
2592 	return err;
2593 }
2594 
2595 /* Process one rtnetlink message. */
2596 
2597 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2598 {
2599 	struct net *net = sock_net(skb->sk);
2600 	rtnl_doit_func doit;
2601 	int sz_idx, kind;
2602 	int family;
2603 	int type;
2604 	int err;
2605 
2606 	type = nlh->nlmsg_type;
2607 	if (type > RTM_MAX)
2608 		return -EOPNOTSUPP;
2609 
2610 	type -= RTM_BASE;
2611 
2612 	/* All the messages must have at least 1 byte length */
2613 	if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2614 		return 0;
2615 
2616 	family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2617 	sz_idx = type>>2;
2618 	kind = type&3;
2619 
2620 	if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2621 		return -EPERM;
2622 
2623 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2624 		struct sock *rtnl;
2625 		rtnl_dumpit_func dumpit;
2626 		rtnl_calcit_func calcit;
2627 		u16 min_dump_alloc = 0;
2628 
2629 		dumpit = rtnl_get_dumpit(family, type);
2630 		if (dumpit == NULL)
2631 			return -EOPNOTSUPP;
2632 		calcit = rtnl_get_calcit(family, type);
2633 		if (calcit)
2634 			min_dump_alloc = calcit(skb, nlh);
2635 
2636 		__rtnl_unlock();
2637 		rtnl = net->rtnl;
2638 		{
2639 			struct netlink_dump_control c = {
2640 				.dump		= dumpit,
2641 				.min_dump_alloc	= min_dump_alloc,
2642 			};
2643 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2644 		}
2645 		rtnl_lock();
2646 		return err;
2647 	}
2648 
2649 	doit = rtnl_get_doit(family, type);
2650 	if (doit == NULL)
2651 		return -EOPNOTSUPP;
2652 
2653 	return doit(skb, nlh);
2654 }
2655 
2656 static void rtnetlink_rcv(struct sk_buff *skb)
2657 {
2658 	rtnl_lock();
2659 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2660 	rtnl_unlock();
2661 }
2662 
2663 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2664 {
2665 	struct net_device *dev = ptr;
2666 
2667 	switch (event) {
2668 	case NETDEV_UP:
2669 	case NETDEV_DOWN:
2670 	case NETDEV_PRE_UP:
2671 	case NETDEV_POST_INIT:
2672 	case NETDEV_REGISTER:
2673 	case NETDEV_CHANGE:
2674 	case NETDEV_PRE_TYPE_CHANGE:
2675 	case NETDEV_GOING_DOWN:
2676 	case NETDEV_UNREGISTER:
2677 	case NETDEV_UNREGISTER_FINAL:
2678 	case NETDEV_RELEASE:
2679 	case NETDEV_JOIN:
2680 		break;
2681 	default:
2682 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2683 		break;
2684 	}
2685 	return NOTIFY_DONE;
2686 }
2687 
2688 static struct notifier_block rtnetlink_dev_notifier = {
2689 	.notifier_call	= rtnetlink_event,
2690 };
2691 
2692 
2693 static int __net_init rtnetlink_net_init(struct net *net)
2694 {
2695 	struct sock *sk;
2696 	struct netlink_kernel_cfg cfg = {
2697 		.groups		= RTNLGRP_MAX,
2698 		.input		= rtnetlink_rcv,
2699 		.cb_mutex	= &rtnl_mutex,
2700 		.flags		= NL_CFG_F_NONROOT_RECV,
2701 	};
2702 
2703 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2704 	if (!sk)
2705 		return -ENOMEM;
2706 	net->rtnl = sk;
2707 	return 0;
2708 }
2709 
2710 static void __net_exit rtnetlink_net_exit(struct net *net)
2711 {
2712 	netlink_kernel_release(net->rtnl);
2713 	net->rtnl = NULL;
2714 }
2715 
2716 static struct pernet_operations rtnetlink_net_ops = {
2717 	.init = rtnetlink_net_init,
2718 	.exit = rtnetlink_net_exit,
2719 };
2720 
2721 void __init rtnetlink_init(void)
2722 {
2723 	if (register_pernet_subsys(&rtnetlink_net_ops))
2724 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2725 
2726 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2727 
2728 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2729 		      rtnl_dump_ifinfo, rtnl_calcit);
2730 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2731 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2732 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2733 
2734 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2735 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2736 
2737 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2738 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2739 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2740 
2741 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2742 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2743 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2744 }
2745 
2746