xref: /openbmc/linux/net/core/rtnetlink.c (revision 5104d265)
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 			err = -EMSGSIZE;
501 			goto err_cancel_link;
502 		}
503 		err = ops->fill_info(skb, dev);
504 		if (err < 0)
505 			goto err_cancel_data;
506 		nla_nest_end(skb, data);
507 	}
508 
509 	nla_nest_end(skb, linkinfo);
510 	return 0;
511 
512 err_cancel_data:
513 	nla_nest_cancel(skb, data);
514 err_cancel_link:
515 	nla_nest_cancel(skb, linkinfo);
516 out:
517 	return err;
518 }
519 
520 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
521 {
522 	struct sock *rtnl = net->rtnl;
523 	int err = 0;
524 
525 	NETLINK_CB(skb).dst_group = group;
526 	if (echo)
527 		atomic_inc(&skb->users);
528 	netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
529 	if (echo)
530 		err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
531 	return err;
532 }
533 
534 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
535 {
536 	struct sock *rtnl = net->rtnl;
537 
538 	return nlmsg_unicast(rtnl, skb, pid);
539 }
540 EXPORT_SYMBOL(rtnl_unicast);
541 
542 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
543 		 struct nlmsghdr *nlh, gfp_t flags)
544 {
545 	struct sock *rtnl = net->rtnl;
546 	int report = 0;
547 
548 	if (nlh)
549 		report = nlmsg_report(nlh);
550 
551 	nlmsg_notify(rtnl, skb, pid, group, report, flags);
552 }
553 EXPORT_SYMBOL(rtnl_notify);
554 
555 void rtnl_set_sk_err(struct net *net, u32 group, int error)
556 {
557 	struct sock *rtnl = net->rtnl;
558 
559 	netlink_set_err(rtnl, 0, group, error);
560 }
561 EXPORT_SYMBOL(rtnl_set_sk_err);
562 
563 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
564 {
565 	struct nlattr *mx;
566 	int i, valid = 0;
567 
568 	mx = nla_nest_start(skb, RTA_METRICS);
569 	if (mx == NULL)
570 		return -ENOBUFS;
571 
572 	for (i = 0; i < RTAX_MAX; i++) {
573 		if (metrics[i]) {
574 			valid++;
575 			if (nla_put_u32(skb, i+1, metrics[i]))
576 				goto nla_put_failure;
577 		}
578 	}
579 
580 	if (!valid) {
581 		nla_nest_cancel(skb, mx);
582 		return 0;
583 	}
584 
585 	return nla_nest_end(skb, mx);
586 
587 nla_put_failure:
588 	nla_nest_cancel(skb, mx);
589 	return -EMSGSIZE;
590 }
591 EXPORT_SYMBOL(rtnetlink_put_metrics);
592 
593 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
594 		       long expires, u32 error)
595 {
596 	struct rta_cacheinfo ci = {
597 		.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
598 		.rta_used = dst->__use,
599 		.rta_clntref = atomic_read(&(dst->__refcnt)),
600 		.rta_error = error,
601 		.rta_id =  id,
602 	};
603 
604 	if (expires) {
605 		unsigned long clock;
606 
607 		clock = jiffies_to_clock_t(abs(expires));
608 		clock = min_t(unsigned long, clock, INT_MAX);
609 		ci.rta_expires = (expires > 0) ? clock : -clock;
610 	}
611 	return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
612 }
613 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
614 
615 static void set_operstate(struct net_device *dev, unsigned char transition)
616 {
617 	unsigned char operstate = dev->operstate;
618 
619 	switch (transition) {
620 	case IF_OPER_UP:
621 		if ((operstate == IF_OPER_DORMANT ||
622 		     operstate == IF_OPER_UNKNOWN) &&
623 		    !netif_dormant(dev))
624 			operstate = IF_OPER_UP;
625 		break;
626 
627 	case IF_OPER_DORMANT:
628 		if (operstate == IF_OPER_UP ||
629 		    operstate == IF_OPER_UNKNOWN)
630 			operstate = IF_OPER_DORMANT;
631 		break;
632 	}
633 
634 	if (dev->operstate != operstate) {
635 		write_lock_bh(&dev_base_lock);
636 		dev->operstate = operstate;
637 		write_unlock_bh(&dev_base_lock);
638 		netdev_state_change(dev);
639 	}
640 }
641 
642 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
643 {
644 	return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
645 	       (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
646 }
647 
648 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
649 					   const struct ifinfomsg *ifm)
650 {
651 	unsigned int flags = ifm->ifi_flags;
652 
653 	/* bugwards compatibility: ifi_change == 0 is treated as ~0 */
654 	if (ifm->ifi_change)
655 		flags = (flags & ifm->ifi_change) |
656 			(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
657 
658 	return flags;
659 }
660 
661 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
662 				 const struct rtnl_link_stats64 *b)
663 {
664 	a->rx_packets = b->rx_packets;
665 	a->tx_packets = b->tx_packets;
666 	a->rx_bytes = b->rx_bytes;
667 	a->tx_bytes = b->tx_bytes;
668 	a->rx_errors = b->rx_errors;
669 	a->tx_errors = b->tx_errors;
670 	a->rx_dropped = b->rx_dropped;
671 	a->tx_dropped = b->tx_dropped;
672 
673 	a->multicast = b->multicast;
674 	a->collisions = b->collisions;
675 
676 	a->rx_length_errors = b->rx_length_errors;
677 	a->rx_over_errors = b->rx_over_errors;
678 	a->rx_crc_errors = b->rx_crc_errors;
679 	a->rx_frame_errors = b->rx_frame_errors;
680 	a->rx_fifo_errors = b->rx_fifo_errors;
681 	a->rx_missed_errors = b->rx_missed_errors;
682 
683 	a->tx_aborted_errors = b->tx_aborted_errors;
684 	a->tx_carrier_errors = b->tx_carrier_errors;
685 	a->tx_fifo_errors = b->tx_fifo_errors;
686 	a->tx_heartbeat_errors = b->tx_heartbeat_errors;
687 	a->tx_window_errors = b->tx_window_errors;
688 
689 	a->rx_compressed = b->rx_compressed;
690 	a->tx_compressed = b->tx_compressed;
691 }
692 
693 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
694 {
695 	memcpy(v, b, sizeof(*b));
696 }
697 
698 /* All VF info */
699 static inline int rtnl_vfinfo_size(const struct net_device *dev,
700 				   u32 ext_filter_mask)
701 {
702 	if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
703 	    (ext_filter_mask & RTEXT_FILTER_VF)) {
704 		int num_vfs = dev_num_vf(dev->dev.parent);
705 		size_t size = nla_total_size(sizeof(struct nlattr));
706 		size += nla_total_size(num_vfs * sizeof(struct nlattr));
707 		size += num_vfs *
708 			(nla_total_size(sizeof(struct ifla_vf_mac)) +
709 			 nla_total_size(sizeof(struct ifla_vf_vlan)) +
710 			 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
711 			 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
712 		return size;
713 	} else
714 		return 0;
715 }
716 
717 static size_t rtnl_port_size(const struct net_device *dev)
718 {
719 	size_t port_size = nla_total_size(4)		/* PORT_VF */
720 		+ nla_total_size(PORT_PROFILE_MAX)	/* PORT_PROFILE */
721 		+ nla_total_size(sizeof(struct ifla_port_vsi))
722 							/* PORT_VSI_TYPE */
723 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_INSTANCE_UUID */
724 		+ nla_total_size(PORT_UUID_MAX)		/* PORT_HOST_UUID */
725 		+ nla_total_size(1)			/* PROT_VDP_REQUEST */
726 		+ nla_total_size(2);			/* PORT_VDP_RESPONSE */
727 	size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
728 	size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
729 		+ port_size;
730 	size_t port_self_size = nla_total_size(sizeof(struct nlattr))
731 		+ port_size;
732 
733 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
734 		return 0;
735 	if (dev_num_vf(dev->dev.parent))
736 		return port_self_size + vf_ports_size +
737 			vf_port_size * dev_num_vf(dev->dev.parent);
738 	else
739 		return port_self_size;
740 }
741 
742 static noinline size_t if_nlmsg_size(const struct net_device *dev,
743 				     u32 ext_filter_mask)
744 {
745 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
746 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
747 	       + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
748 	       + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
749 	       + nla_total_size(sizeof(struct rtnl_link_ifmap))
750 	       + nla_total_size(sizeof(struct rtnl_link_stats))
751 	       + nla_total_size(sizeof(struct rtnl_link_stats64))
752 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
753 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
754 	       + nla_total_size(4) /* IFLA_TXQLEN */
755 	       + nla_total_size(4) /* IFLA_WEIGHT */
756 	       + nla_total_size(4) /* IFLA_MTU */
757 	       + nla_total_size(4) /* IFLA_LINK */
758 	       + nla_total_size(4) /* IFLA_MASTER */
759 	       + nla_total_size(1) /* IFLA_CARRIER */
760 	       + nla_total_size(4) /* IFLA_PROMISCUITY */
761 	       + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
762 	       + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
763 	       + nla_total_size(1) /* IFLA_OPERSTATE */
764 	       + nla_total_size(1) /* IFLA_LINKMODE */
765 	       + nla_total_size(ext_filter_mask
766 			        & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
767 	       + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
768 	       + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
769 	       + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
770 	       + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
771 }
772 
773 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
774 {
775 	struct nlattr *vf_ports;
776 	struct nlattr *vf_port;
777 	int vf;
778 	int err;
779 
780 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
781 	if (!vf_ports)
782 		return -EMSGSIZE;
783 
784 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
785 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
786 		if (!vf_port)
787 			goto nla_put_failure;
788 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
789 			goto nla_put_failure;
790 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
791 		if (err == -EMSGSIZE)
792 			goto nla_put_failure;
793 		if (err) {
794 			nla_nest_cancel(skb, vf_port);
795 			continue;
796 		}
797 		nla_nest_end(skb, vf_port);
798 	}
799 
800 	nla_nest_end(skb, vf_ports);
801 
802 	return 0;
803 
804 nla_put_failure:
805 	nla_nest_cancel(skb, vf_ports);
806 	return -EMSGSIZE;
807 }
808 
809 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
810 {
811 	struct nlattr *port_self;
812 	int err;
813 
814 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
815 	if (!port_self)
816 		return -EMSGSIZE;
817 
818 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
819 	if (err) {
820 		nla_nest_cancel(skb, port_self);
821 		return (err == -EMSGSIZE) ? err : 0;
822 	}
823 
824 	nla_nest_end(skb, port_self);
825 
826 	return 0;
827 }
828 
829 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
830 {
831 	int err;
832 
833 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
834 		return 0;
835 
836 	err = rtnl_port_self_fill(skb, dev);
837 	if (err)
838 		return err;
839 
840 	if (dev_num_vf(dev->dev.parent)) {
841 		err = rtnl_vf_ports_fill(skb, dev);
842 		if (err)
843 			return err;
844 	}
845 
846 	return 0;
847 }
848 
849 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
850 			    int type, u32 pid, u32 seq, u32 change,
851 			    unsigned int flags, u32 ext_filter_mask)
852 {
853 	struct ifinfomsg *ifm;
854 	struct nlmsghdr *nlh;
855 	struct rtnl_link_stats64 temp;
856 	const struct rtnl_link_stats64 *stats;
857 	struct nlattr *attr, *af_spec;
858 	struct rtnl_af_ops *af_ops;
859 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
860 
861 	ASSERT_RTNL();
862 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
863 	if (nlh == NULL)
864 		return -EMSGSIZE;
865 
866 	ifm = nlmsg_data(nlh);
867 	ifm->ifi_family = AF_UNSPEC;
868 	ifm->__ifi_pad = 0;
869 	ifm->ifi_type = dev->type;
870 	ifm->ifi_index = dev->ifindex;
871 	ifm->ifi_flags = dev_get_flags(dev);
872 	ifm->ifi_change = change;
873 
874 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
875 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
876 	    nla_put_u8(skb, IFLA_OPERSTATE,
877 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
878 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
879 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
880 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
881 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
882 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
883 #ifdef CONFIG_RPS
884 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
885 #endif
886 	    (dev->ifindex != dev->iflink &&
887 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
888 	    (upper_dev &&
889 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
890 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
891 	    (dev->qdisc &&
892 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
893 	    (dev->ifalias &&
894 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
895 		goto nla_put_failure;
896 
897 	if (1) {
898 		struct rtnl_link_ifmap map = {
899 			.mem_start   = dev->mem_start,
900 			.mem_end     = dev->mem_end,
901 			.base_addr   = dev->base_addr,
902 			.irq         = dev->irq,
903 			.dma         = dev->dma,
904 			.port        = dev->if_port,
905 		};
906 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
907 			goto nla_put_failure;
908 	}
909 
910 	if (dev->addr_len) {
911 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
912 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
913 			goto nla_put_failure;
914 	}
915 
916 	attr = nla_reserve(skb, IFLA_STATS,
917 			sizeof(struct rtnl_link_stats));
918 	if (attr == NULL)
919 		goto nla_put_failure;
920 
921 	stats = dev_get_stats(dev, &temp);
922 	copy_rtnl_link_stats(nla_data(attr), stats);
923 
924 	attr = nla_reserve(skb, IFLA_STATS64,
925 			sizeof(struct rtnl_link_stats64));
926 	if (attr == NULL)
927 		goto nla_put_failure;
928 	copy_rtnl_link_stats64(nla_data(attr), stats);
929 
930 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
931 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
932 		goto nla_put_failure;
933 
934 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
935 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
936 		int i;
937 
938 		struct nlattr *vfinfo, *vf;
939 		int num_vfs = dev_num_vf(dev->dev.parent);
940 
941 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
942 		if (!vfinfo)
943 			goto nla_put_failure;
944 		for (i = 0; i < num_vfs; i++) {
945 			struct ifla_vf_info ivi;
946 			struct ifla_vf_mac vf_mac;
947 			struct ifla_vf_vlan vf_vlan;
948 			struct ifla_vf_tx_rate vf_tx_rate;
949 			struct ifla_vf_spoofchk vf_spoofchk;
950 			struct ifla_vf_link_state vf_linkstate;
951 
952 			/*
953 			 * Not all SR-IOV capable drivers support the
954 			 * spoofcheck query.  Preset to -1 so the user
955 			 * space tool can detect that the driver didn't
956 			 * report anything.
957 			 */
958 			ivi.spoofchk = -1;
959 			memset(ivi.mac, 0, sizeof(ivi.mac));
960 			/* The default value for VF link state is "auto"
961 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
962 			 */
963 			ivi.linkstate = 0;
964 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
965 				break;
966 			vf_mac.vf =
967 				vf_vlan.vf =
968 				vf_tx_rate.vf =
969 				vf_spoofchk.vf =
970 				vf_linkstate.vf = ivi.vf;
971 
972 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
973 			vf_vlan.vlan = ivi.vlan;
974 			vf_vlan.qos = ivi.qos;
975 			vf_tx_rate.rate = ivi.tx_rate;
976 			vf_spoofchk.setting = ivi.spoofchk;
977 			vf_linkstate.link_state = ivi.linkstate;
978 			vf = nla_nest_start(skb, IFLA_VF_INFO);
979 			if (!vf) {
980 				nla_nest_cancel(skb, vfinfo);
981 				goto nla_put_failure;
982 			}
983 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
984 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
985 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
986 				    &vf_tx_rate) ||
987 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
988 				    &vf_spoofchk) ||
989 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
990 				    &vf_linkstate))
991 				goto nla_put_failure;
992 			nla_nest_end(skb, vf);
993 		}
994 		nla_nest_end(skb, vfinfo);
995 	}
996 
997 	if (rtnl_port_fill(skb, dev))
998 		goto nla_put_failure;
999 
1000 	if (dev->rtnl_link_ops) {
1001 		if (rtnl_link_fill(skb, dev) < 0)
1002 			goto nla_put_failure;
1003 	}
1004 
1005 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1006 		goto nla_put_failure;
1007 
1008 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1009 		if (af_ops->fill_link_af) {
1010 			struct nlattr *af;
1011 			int err;
1012 
1013 			if (!(af = nla_nest_start(skb, af_ops->family)))
1014 				goto nla_put_failure;
1015 
1016 			err = af_ops->fill_link_af(skb, dev);
1017 
1018 			/*
1019 			 * Caller may return ENODATA to indicate that there
1020 			 * was no data to be dumped. This is not an error, it
1021 			 * means we should trim the attribute header and
1022 			 * continue.
1023 			 */
1024 			if (err == -ENODATA)
1025 				nla_nest_cancel(skb, af);
1026 			else if (err < 0)
1027 				goto nla_put_failure;
1028 
1029 			nla_nest_end(skb, af);
1030 		}
1031 	}
1032 
1033 	nla_nest_end(skb, af_spec);
1034 
1035 	return nlmsg_end(skb, nlh);
1036 
1037 nla_put_failure:
1038 	nlmsg_cancel(skb, nlh);
1039 	return -EMSGSIZE;
1040 }
1041 
1042 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1043 {
1044 	struct net *net = sock_net(skb->sk);
1045 	int h, s_h;
1046 	int idx = 0, s_idx;
1047 	struct net_device *dev;
1048 	struct hlist_head *head;
1049 	struct nlattr *tb[IFLA_MAX+1];
1050 	u32 ext_filter_mask = 0;
1051 
1052 	s_h = cb->args[0];
1053 	s_idx = cb->args[1];
1054 
1055 	rcu_read_lock();
1056 	cb->seq = net->dev_base_seq;
1057 
1058 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1059 			ifla_policy) >= 0) {
1060 
1061 		if (tb[IFLA_EXT_MASK])
1062 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1063 	}
1064 
1065 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1066 		idx = 0;
1067 		head = &net->dev_index_head[h];
1068 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
1069 			if (idx < s_idx)
1070 				goto cont;
1071 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1072 					     NETLINK_CB(cb->skb).portid,
1073 					     cb->nlh->nlmsg_seq, 0,
1074 					     NLM_F_MULTI,
1075 					     ext_filter_mask) <= 0)
1076 				goto out;
1077 
1078 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1079 cont:
1080 			idx++;
1081 		}
1082 	}
1083 out:
1084 	rcu_read_unlock();
1085 	cb->args[1] = idx;
1086 	cb->args[0] = h;
1087 
1088 	return skb->len;
1089 }
1090 
1091 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1092 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1093 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1094 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1095 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1096 	[IFLA_MTU]		= { .type = NLA_U32 },
1097 	[IFLA_LINK]		= { .type = NLA_U32 },
1098 	[IFLA_MASTER]		= { .type = NLA_U32 },
1099 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1100 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1101 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1102 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1103 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1104 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1105 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1106 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1107 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1108 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1109 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1110 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1111 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1112 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1113 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1114 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1115 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1116 };
1117 EXPORT_SYMBOL(ifla_policy);
1118 
1119 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1120 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1121 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1122 };
1123 
1124 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1125 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1126 };
1127 
1128 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1129 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1130 				    .len = sizeof(struct ifla_vf_mac) },
1131 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1132 				    .len = sizeof(struct ifla_vf_vlan) },
1133 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1134 				    .len = sizeof(struct ifla_vf_tx_rate) },
1135 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1136 				    .len = sizeof(struct ifla_vf_spoofchk) },
1137 };
1138 
1139 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1140 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1141 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1142 				    .len = PORT_PROFILE_MAX },
1143 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1144 				    .len = sizeof(struct ifla_port_vsi)},
1145 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1146 				      .len = PORT_UUID_MAX },
1147 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1148 				    .len = PORT_UUID_MAX },
1149 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1150 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1151 };
1152 
1153 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1154 {
1155 	struct net *net;
1156 	/* Examine the link attributes and figure out which
1157 	 * network namespace we are talking about.
1158 	 */
1159 	if (tb[IFLA_NET_NS_PID])
1160 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1161 	else if (tb[IFLA_NET_NS_FD])
1162 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1163 	else
1164 		net = get_net(src_net);
1165 	return net;
1166 }
1167 EXPORT_SYMBOL(rtnl_link_get_net);
1168 
1169 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1170 {
1171 	if (dev) {
1172 		if (tb[IFLA_ADDRESS] &&
1173 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1174 			return -EINVAL;
1175 
1176 		if (tb[IFLA_BROADCAST] &&
1177 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1178 			return -EINVAL;
1179 	}
1180 
1181 	if (tb[IFLA_AF_SPEC]) {
1182 		struct nlattr *af;
1183 		int rem, err;
1184 
1185 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1186 			const struct rtnl_af_ops *af_ops;
1187 
1188 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1189 				return -EAFNOSUPPORT;
1190 
1191 			if (!af_ops->set_link_af)
1192 				return -EOPNOTSUPP;
1193 
1194 			if (af_ops->validate_link_af) {
1195 				err = af_ops->validate_link_af(dev, af);
1196 				if (err < 0)
1197 					return err;
1198 			}
1199 		}
1200 	}
1201 
1202 	return 0;
1203 }
1204 
1205 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1206 {
1207 	int rem, err = -EINVAL;
1208 	struct nlattr *vf;
1209 	const struct net_device_ops *ops = dev->netdev_ops;
1210 
1211 	nla_for_each_nested(vf, attr, rem) {
1212 		switch (nla_type(vf)) {
1213 		case IFLA_VF_MAC: {
1214 			struct ifla_vf_mac *ivm;
1215 			ivm = nla_data(vf);
1216 			err = -EOPNOTSUPP;
1217 			if (ops->ndo_set_vf_mac)
1218 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1219 							  ivm->mac);
1220 			break;
1221 		}
1222 		case IFLA_VF_VLAN: {
1223 			struct ifla_vf_vlan *ivv;
1224 			ivv = nla_data(vf);
1225 			err = -EOPNOTSUPP;
1226 			if (ops->ndo_set_vf_vlan)
1227 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1228 							   ivv->vlan,
1229 							   ivv->qos);
1230 			break;
1231 		}
1232 		case IFLA_VF_TX_RATE: {
1233 			struct ifla_vf_tx_rate *ivt;
1234 			ivt = nla_data(vf);
1235 			err = -EOPNOTSUPP;
1236 			if (ops->ndo_set_vf_tx_rate)
1237 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1238 							      ivt->rate);
1239 			break;
1240 		}
1241 		case IFLA_VF_SPOOFCHK: {
1242 			struct ifla_vf_spoofchk *ivs;
1243 			ivs = nla_data(vf);
1244 			err = -EOPNOTSUPP;
1245 			if (ops->ndo_set_vf_spoofchk)
1246 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1247 							       ivs->setting);
1248 			break;
1249 		}
1250 		case IFLA_VF_LINK_STATE: {
1251 			struct ifla_vf_link_state *ivl;
1252 			ivl = nla_data(vf);
1253 			err = -EOPNOTSUPP;
1254 			if (ops->ndo_set_vf_link_state)
1255 				err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1256 								 ivl->link_state);
1257 			break;
1258 		}
1259 		default:
1260 			err = -EINVAL;
1261 			break;
1262 		}
1263 		if (err)
1264 			break;
1265 	}
1266 	return err;
1267 }
1268 
1269 static int do_set_master(struct net_device *dev, int ifindex)
1270 {
1271 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1272 	const struct net_device_ops *ops;
1273 	int err;
1274 
1275 	if (upper_dev) {
1276 		if (upper_dev->ifindex == ifindex)
1277 			return 0;
1278 		ops = upper_dev->netdev_ops;
1279 		if (ops->ndo_del_slave) {
1280 			err = ops->ndo_del_slave(upper_dev, dev);
1281 			if (err)
1282 				return err;
1283 		} else {
1284 			return -EOPNOTSUPP;
1285 		}
1286 	}
1287 
1288 	if (ifindex) {
1289 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1290 		if (!upper_dev)
1291 			return -EINVAL;
1292 		ops = upper_dev->netdev_ops;
1293 		if (ops->ndo_add_slave) {
1294 			err = ops->ndo_add_slave(upper_dev, dev);
1295 			if (err)
1296 				return err;
1297 		} else {
1298 			return -EOPNOTSUPP;
1299 		}
1300 	}
1301 	return 0;
1302 }
1303 
1304 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1305 		      struct nlattr **tb, char *ifname, int modified)
1306 {
1307 	const struct net_device_ops *ops = dev->netdev_ops;
1308 	int err;
1309 
1310 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1311 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1312 		if (IS_ERR(net)) {
1313 			err = PTR_ERR(net);
1314 			goto errout;
1315 		}
1316 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1317 			err = -EPERM;
1318 			goto errout;
1319 		}
1320 		err = dev_change_net_namespace(dev, net, ifname);
1321 		put_net(net);
1322 		if (err)
1323 			goto errout;
1324 		modified = 1;
1325 	}
1326 
1327 	if (tb[IFLA_MAP]) {
1328 		struct rtnl_link_ifmap *u_map;
1329 		struct ifmap k_map;
1330 
1331 		if (!ops->ndo_set_config) {
1332 			err = -EOPNOTSUPP;
1333 			goto errout;
1334 		}
1335 
1336 		if (!netif_device_present(dev)) {
1337 			err = -ENODEV;
1338 			goto errout;
1339 		}
1340 
1341 		u_map = nla_data(tb[IFLA_MAP]);
1342 		k_map.mem_start = (unsigned long) u_map->mem_start;
1343 		k_map.mem_end = (unsigned long) u_map->mem_end;
1344 		k_map.base_addr = (unsigned short) u_map->base_addr;
1345 		k_map.irq = (unsigned char) u_map->irq;
1346 		k_map.dma = (unsigned char) u_map->dma;
1347 		k_map.port = (unsigned char) u_map->port;
1348 
1349 		err = ops->ndo_set_config(dev, &k_map);
1350 		if (err < 0)
1351 			goto errout;
1352 
1353 		modified = 1;
1354 	}
1355 
1356 	if (tb[IFLA_ADDRESS]) {
1357 		struct sockaddr *sa;
1358 		int len;
1359 
1360 		len = sizeof(sa_family_t) + dev->addr_len;
1361 		sa = kmalloc(len, GFP_KERNEL);
1362 		if (!sa) {
1363 			err = -ENOMEM;
1364 			goto errout;
1365 		}
1366 		sa->sa_family = dev->type;
1367 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1368 		       dev->addr_len);
1369 		err = dev_set_mac_address(dev, sa);
1370 		kfree(sa);
1371 		if (err)
1372 			goto errout;
1373 		modified = 1;
1374 	}
1375 
1376 	if (tb[IFLA_MTU]) {
1377 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1378 		if (err < 0)
1379 			goto errout;
1380 		modified = 1;
1381 	}
1382 
1383 	if (tb[IFLA_GROUP]) {
1384 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1385 		modified = 1;
1386 	}
1387 
1388 	/*
1389 	 * Interface selected by interface index but interface
1390 	 * name provided implies that a name change has been
1391 	 * requested.
1392 	 */
1393 	if (ifm->ifi_index > 0 && ifname[0]) {
1394 		err = dev_change_name(dev, ifname);
1395 		if (err < 0)
1396 			goto errout;
1397 		modified = 1;
1398 	}
1399 
1400 	if (tb[IFLA_IFALIAS]) {
1401 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1402 				    nla_len(tb[IFLA_IFALIAS]));
1403 		if (err < 0)
1404 			goto errout;
1405 		modified = 1;
1406 	}
1407 
1408 	if (tb[IFLA_BROADCAST]) {
1409 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1410 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1411 	}
1412 
1413 	if (ifm->ifi_flags || ifm->ifi_change) {
1414 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1415 		if (err < 0)
1416 			goto errout;
1417 	}
1418 
1419 	if (tb[IFLA_MASTER]) {
1420 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1421 		if (err)
1422 			goto errout;
1423 		modified = 1;
1424 	}
1425 
1426 	if (tb[IFLA_CARRIER]) {
1427 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1428 		if (err)
1429 			goto errout;
1430 		modified = 1;
1431 	}
1432 
1433 	if (tb[IFLA_TXQLEN])
1434 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1435 
1436 	if (tb[IFLA_OPERSTATE])
1437 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1438 
1439 	if (tb[IFLA_LINKMODE]) {
1440 		write_lock_bh(&dev_base_lock);
1441 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1442 		write_unlock_bh(&dev_base_lock);
1443 	}
1444 
1445 	if (tb[IFLA_VFINFO_LIST]) {
1446 		struct nlattr *attr;
1447 		int rem;
1448 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1449 			if (nla_type(attr) != IFLA_VF_INFO) {
1450 				err = -EINVAL;
1451 				goto errout;
1452 			}
1453 			err = do_setvfinfo(dev, attr);
1454 			if (err < 0)
1455 				goto errout;
1456 			modified = 1;
1457 		}
1458 	}
1459 	err = 0;
1460 
1461 	if (tb[IFLA_VF_PORTS]) {
1462 		struct nlattr *port[IFLA_PORT_MAX+1];
1463 		struct nlattr *attr;
1464 		int vf;
1465 		int rem;
1466 
1467 		err = -EOPNOTSUPP;
1468 		if (!ops->ndo_set_vf_port)
1469 			goto errout;
1470 
1471 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1472 			if (nla_type(attr) != IFLA_VF_PORT)
1473 				continue;
1474 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1475 				attr, ifla_port_policy);
1476 			if (err < 0)
1477 				goto errout;
1478 			if (!port[IFLA_PORT_VF]) {
1479 				err = -EOPNOTSUPP;
1480 				goto errout;
1481 			}
1482 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1483 			err = ops->ndo_set_vf_port(dev, vf, port);
1484 			if (err < 0)
1485 				goto errout;
1486 			modified = 1;
1487 		}
1488 	}
1489 	err = 0;
1490 
1491 	if (tb[IFLA_PORT_SELF]) {
1492 		struct nlattr *port[IFLA_PORT_MAX+1];
1493 
1494 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1495 			tb[IFLA_PORT_SELF], ifla_port_policy);
1496 		if (err < 0)
1497 			goto errout;
1498 
1499 		err = -EOPNOTSUPP;
1500 		if (ops->ndo_set_vf_port)
1501 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1502 		if (err < 0)
1503 			goto errout;
1504 		modified = 1;
1505 	}
1506 
1507 	if (tb[IFLA_AF_SPEC]) {
1508 		struct nlattr *af;
1509 		int rem;
1510 
1511 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1512 			const struct rtnl_af_ops *af_ops;
1513 
1514 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1515 				BUG();
1516 
1517 			err = af_ops->set_link_af(dev, af);
1518 			if (err < 0)
1519 				goto errout;
1520 
1521 			modified = 1;
1522 		}
1523 	}
1524 	err = 0;
1525 
1526 errout:
1527 	if (err < 0 && modified)
1528 		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",
1529 				     dev->name);
1530 
1531 	return err;
1532 }
1533 
1534 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1535 {
1536 	struct net *net = sock_net(skb->sk);
1537 	struct ifinfomsg *ifm;
1538 	struct net_device *dev;
1539 	int err;
1540 	struct nlattr *tb[IFLA_MAX+1];
1541 	char ifname[IFNAMSIZ];
1542 
1543 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1544 	if (err < 0)
1545 		goto errout;
1546 
1547 	if (tb[IFLA_IFNAME])
1548 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1549 	else
1550 		ifname[0] = '\0';
1551 
1552 	err = -EINVAL;
1553 	ifm = nlmsg_data(nlh);
1554 	if (ifm->ifi_index > 0)
1555 		dev = __dev_get_by_index(net, ifm->ifi_index);
1556 	else if (tb[IFLA_IFNAME])
1557 		dev = __dev_get_by_name(net, ifname);
1558 	else
1559 		goto errout;
1560 
1561 	if (dev == NULL) {
1562 		err = -ENODEV;
1563 		goto errout;
1564 	}
1565 
1566 	err = validate_linkmsg(dev, tb);
1567 	if (err < 0)
1568 		goto errout;
1569 
1570 	err = do_setlink(dev, ifm, tb, ifname, 0);
1571 errout:
1572 	return err;
1573 }
1574 
1575 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1576 {
1577 	struct net *net = sock_net(skb->sk);
1578 	const struct rtnl_link_ops *ops;
1579 	struct net_device *dev;
1580 	struct ifinfomsg *ifm;
1581 	char ifname[IFNAMSIZ];
1582 	struct nlattr *tb[IFLA_MAX+1];
1583 	int err;
1584 	LIST_HEAD(list_kill);
1585 
1586 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1587 	if (err < 0)
1588 		return err;
1589 
1590 	if (tb[IFLA_IFNAME])
1591 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1592 
1593 	ifm = nlmsg_data(nlh);
1594 	if (ifm->ifi_index > 0)
1595 		dev = __dev_get_by_index(net, ifm->ifi_index);
1596 	else if (tb[IFLA_IFNAME])
1597 		dev = __dev_get_by_name(net, ifname);
1598 	else
1599 		return -EINVAL;
1600 
1601 	if (!dev)
1602 		return -ENODEV;
1603 
1604 	ops = dev->rtnl_link_ops;
1605 	if (!ops)
1606 		return -EOPNOTSUPP;
1607 
1608 	ops->dellink(dev, &list_kill);
1609 	unregister_netdevice_many(&list_kill);
1610 	list_del(&list_kill);
1611 	return 0;
1612 }
1613 
1614 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1615 {
1616 	unsigned int old_flags;
1617 	int err;
1618 
1619 	old_flags = dev->flags;
1620 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1621 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1622 		if (err < 0)
1623 			return err;
1624 	}
1625 
1626 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1627 	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1628 
1629 	__dev_notify_flags(dev, old_flags);
1630 	return 0;
1631 }
1632 EXPORT_SYMBOL(rtnl_configure_link);
1633 
1634 struct net_device *rtnl_create_link(struct net *net,
1635 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1636 {
1637 	int err;
1638 	struct net_device *dev;
1639 	unsigned int num_tx_queues = 1;
1640 	unsigned int num_rx_queues = 1;
1641 
1642 	if (tb[IFLA_NUM_TX_QUEUES])
1643 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1644 	else if (ops->get_num_tx_queues)
1645 		num_tx_queues = ops->get_num_tx_queues();
1646 
1647 	if (tb[IFLA_NUM_RX_QUEUES])
1648 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1649 	else if (ops->get_num_rx_queues)
1650 		num_rx_queues = ops->get_num_rx_queues();
1651 
1652 	err = -ENOMEM;
1653 	dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1654 			       num_tx_queues, num_rx_queues);
1655 	if (!dev)
1656 		goto err;
1657 
1658 	dev_net_set(dev, net);
1659 	dev->rtnl_link_ops = ops;
1660 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1661 
1662 	if (tb[IFLA_MTU])
1663 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1664 	if (tb[IFLA_ADDRESS]) {
1665 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1666 				nla_len(tb[IFLA_ADDRESS]));
1667 		dev->addr_assign_type = NET_ADDR_SET;
1668 	}
1669 	if (tb[IFLA_BROADCAST])
1670 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1671 				nla_len(tb[IFLA_BROADCAST]));
1672 	if (tb[IFLA_TXQLEN])
1673 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1674 	if (tb[IFLA_OPERSTATE])
1675 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1676 	if (tb[IFLA_LINKMODE])
1677 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1678 	if (tb[IFLA_GROUP])
1679 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1680 
1681 	return dev;
1682 
1683 err:
1684 	return ERR_PTR(err);
1685 }
1686 EXPORT_SYMBOL(rtnl_create_link);
1687 
1688 static int rtnl_group_changelink(struct net *net, int group,
1689 		struct ifinfomsg *ifm,
1690 		struct nlattr **tb)
1691 {
1692 	struct net_device *dev;
1693 	int err;
1694 
1695 	for_each_netdev(net, dev) {
1696 		if (dev->group == group) {
1697 			err = do_setlink(dev, ifm, tb, NULL, 0);
1698 			if (err < 0)
1699 				return err;
1700 		}
1701 	}
1702 
1703 	return 0;
1704 }
1705 
1706 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1707 {
1708 	struct net *net = sock_net(skb->sk);
1709 	const struct rtnl_link_ops *ops;
1710 	struct net_device *dev;
1711 	struct ifinfomsg *ifm;
1712 	char kind[MODULE_NAME_LEN];
1713 	char ifname[IFNAMSIZ];
1714 	struct nlattr *tb[IFLA_MAX+1];
1715 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1716 	int err;
1717 
1718 #ifdef CONFIG_MODULES
1719 replay:
1720 #endif
1721 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1722 	if (err < 0)
1723 		return err;
1724 
1725 	if (tb[IFLA_IFNAME])
1726 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1727 	else
1728 		ifname[0] = '\0';
1729 
1730 	ifm = nlmsg_data(nlh);
1731 	if (ifm->ifi_index > 0)
1732 		dev = __dev_get_by_index(net, ifm->ifi_index);
1733 	else {
1734 		if (ifname[0])
1735 			dev = __dev_get_by_name(net, ifname);
1736 		else
1737 			dev = NULL;
1738 	}
1739 
1740 	err = validate_linkmsg(dev, tb);
1741 	if (err < 0)
1742 		return err;
1743 
1744 	if (tb[IFLA_LINKINFO]) {
1745 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1746 				       tb[IFLA_LINKINFO], ifla_info_policy);
1747 		if (err < 0)
1748 			return err;
1749 	} else
1750 		memset(linkinfo, 0, sizeof(linkinfo));
1751 
1752 	if (linkinfo[IFLA_INFO_KIND]) {
1753 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1754 		ops = rtnl_link_ops_get(kind);
1755 	} else {
1756 		kind[0] = '\0';
1757 		ops = NULL;
1758 	}
1759 
1760 	if (1) {
1761 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1762 		struct net *dest_net;
1763 
1764 		if (ops) {
1765 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1766 				err = nla_parse_nested(attr, ops->maxtype,
1767 						       linkinfo[IFLA_INFO_DATA],
1768 						       ops->policy);
1769 				if (err < 0)
1770 					return err;
1771 				data = attr;
1772 			}
1773 			if (ops->validate) {
1774 				err = ops->validate(tb, data);
1775 				if (err < 0)
1776 					return err;
1777 			}
1778 		}
1779 
1780 		if (dev) {
1781 			int modified = 0;
1782 
1783 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1784 				return -EEXIST;
1785 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1786 				return -EOPNOTSUPP;
1787 
1788 			if (linkinfo[IFLA_INFO_DATA]) {
1789 				if (!ops || ops != dev->rtnl_link_ops ||
1790 				    !ops->changelink)
1791 					return -EOPNOTSUPP;
1792 
1793 				err = ops->changelink(dev, tb, data);
1794 				if (err < 0)
1795 					return err;
1796 				modified = 1;
1797 			}
1798 
1799 			return do_setlink(dev, ifm, tb, ifname, modified);
1800 		}
1801 
1802 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1803 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1804 				return rtnl_group_changelink(net,
1805 						nla_get_u32(tb[IFLA_GROUP]),
1806 						ifm, tb);
1807 			return -ENODEV;
1808 		}
1809 
1810 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1811 			return -EOPNOTSUPP;
1812 
1813 		if (!ops) {
1814 #ifdef CONFIG_MODULES
1815 			if (kind[0]) {
1816 				__rtnl_unlock();
1817 				request_module("rtnl-link-%s", kind);
1818 				rtnl_lock();
1819 				ops = rtnl_link_ops_get(kind);
1820 				if (ops)
1821 					goto replay;
1822 			}
1823 #endif
1824 			return -EOPNOTSUPP;
1825 		}
1826 
1827 		if (!ifname[0])
1828 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1829 
1830 		dest_net = rtnl_link_get_net(net, tb);
1831 		if (IS_ERR(dest_net))
1832 			return PTR_ERR(dest_net);
1833 
1834 		dev = rtnl_create_link(dest_net, ifname, ops, tb);
1835 		if (IS_ERR(dev)) {
1836 			err = PTR_ERR(dev);
1837 			goto out;
1838 		}
1839 
1840 		dev->ifindex = ifm->ifi_index;
1841 
1842 		if (ops->newlink)
1843 			err = ops->newlink(net, dev, tb, data);
1844 		else
1845 			err = register_netdevice(dev);
1846 
1847 		if (err < 0 && !IS_ERR(dev))
1848 			free_netdev(dev);
1849 		if (err < 0)
1850 			goto out;
1851 
1852 		err = rtnl_configure_link(dev, ifm);
1853 		if (err < 0)
1854 			unregister_netdevice(dev);
1855 out:
1856 		put_net(dest_net);
1857 		return err;
1858 	}
1859 }
1860 
1861 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1862 {
1863 	struct net *net = sock_net(skb->sk);
1864 	struct ifinfomsg *ifm;
1865 	char ifname[IFNAMSIZ];
1866 	struct nlattr *tb[IFLA_MAX+1];
1867 	struct net_device *dev = NULL;
1868 	struct sk_buff *nskb;
1869 	int err;
1870 	u32 ext_filter_mask = 0;
1871 
1872 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1873 	if (err < 0)
1874 		return err;
1875 
1876 	if (tb[IFLA_IFNAME])
1877 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1878 
1879 	if (tb[IFLA_EXT_MASK])
1880 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1881 
1882 	ifm = nlmsg_data(nlh);
1883 	if (ifm->ifi_index > 0)
1884 		dev = __dev_get_by_index(net, ifm->ifi_index);
1885 	else if (tb[IFLA_IFNAME])
1886 		dev = __dev_get_by_name(net, ifname);
1887 	else
1888 		return -EINVAL;
1889 
1890 	if (dev == NULL)
1891 		return -ENODEV;
1892 
1893 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1894 	if (nskb == NULL)
1895 		return -ENOBUFS;
1896 
1897 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1898 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1899 	if (err < 0) {
1900 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1901 		WARN_ON(err == -EMSGSIZE);
1902 		kfree_skb(nskb);
1903 	} else
1904 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1905 
1906 	return err;
1907 }
1908 
1909 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1910 {
1911 	struct net *net = sock_net(skb->sk);
1912 	struct net_device *dev;
1913 	struct nlattr *tb[IFLA_MAX+1];
1914 	u32 ext_filter_mask = 0;
1915 	u16 min_ifinfo_dump_size = 0;
1916 
1917 	if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1918 			ifla_policy) >= 0) {
1919 		if (tb[IFLA_EXT_MASK])
1920 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1921 	}
1922 
1923 	if (!ext_filter_mask)
1924 		return NLMSG_GOODSIZE;
1925 	/*
1926 	 * traverse the list of net devices and compute the minimum
1927 	 * buffer size based upon the filter mask.
1928 	 */
1929 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1930 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1931 					     if_nlmsg_size(dev,
1932 						           ext_filter_mask));
1933 	}
1934 
1935 	return min_ifinfo_dump_size;
1936 }
1937 
1938 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1939 {
1940 	int idx;
1941 	int s_idx = cb->family;
1942 
1943 	if (s_idx == 0)
1944 		s_idx = 1;
1945 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1946 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1947 		if (idx < s_idx || idx == PF_PACKET)
1948 			continue;
1949 		if (rtnl_msg_handlers[idx] == NULL ||
1950 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1951 			continue;
1952 		if (idx > s_idx) {
1953 			memset(&cb->args[0], 0, sizeof(cb->args));
1954 			cb->prev_seq = 0;
1955 			cb->seq = 0;
1956 		}
1957 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1958 			break;
1959 	}
1960 	cb->family = idx;
1961 
1962 	return skb->len;
1963 }
1964 
1965 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1966 {
1967 	struct net *net = dev_net(dev);
1968 	struct sk_buff *skb;
1969 	int err = -ENOBUFS;
1970 	size_t if_info_size;
1971 
1972 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1973 	if (skb == NULL)
1974 		goto errout;
1975 
1976 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1977 	if (err < 0) {
1978 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
1979 		WARN_ON(err == -EMSGSIZE);
1980 		kfree_skb(skb);
1981 		goto errout;
1982 	}
1983 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1984 	return;
1985 errout:
1986 	if (err < 0)
1987 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1988 }
1989 EXPORT_SYMBOL(rtmsg_ifinfo);
1990 
1991 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1992 				   struct net_device *dev,
1993 				   u8 *addr, u32 pid, u32 seq,
1994 				   int type, unsigned int flags)
1995 {
1996 	struct nlmsghdr *nlh;
1997 	struct ndmsg *ndm;
1998 
1999 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
2000 	if (!nlh)
2001 		return -EMSGSIZE;
2002 
2003 	ndm = nlmsg_data(nlh);
2004 	ndm->ndm_family  = AF_BRIDGE;
2005 	ndm->ndm_pad1	 = 0;
2006 	ndm->ndm_pad2    = 0;
2007 	ndm->ndm_flags	 = flags;
2008 	ndm->ndm_type	 = 0;
2009 	ndm->ndm_ifindex = dev->ifindex;
2010 	ndm->ndm_state   = NUD_PERMANENT;
2011 
2012 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2013 		goto nla_put_failure;
2014 
2015 	return nlmsg_end(skb, nlh);
2016 
2017 nla_put_failure:
2018 	nlmsg_cancel(skb, nlh);
2019 	return -EMSGSIZE;
2020 }
2021 
2022 static inline size_t rtnl_fdb_nlmsg_size(void)
2023 {
2024 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2025 }
2026 
2027 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2028 {
2029 	struct net *net = dev_net(dev);
2030 	struct sk_buff *skb;
2031 	int err = -ENOBUFS;
2032 
2033 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2034 	if (!skb)
2035 		goto errout;
2036 
2037 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2038 	if (err < 0) {
2039 		kfree_skb(skb);
2040 		goto errout;
2041 	}
2042 
2043 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2044 	return;
2045 errout:
2046 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2047 }
2048 
2049 /**
2050  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2051  */
2052 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2053 		     struct nlattr *tb[],
2054 		     struct net_device *dev,
2055 		     const unsigned char *addr,
2056 		     u16 flags)
2057 {
2058 	int err = -EINVAL;
2059 
2060 	/* If aging addresses are supported device will need to
2061 	 * implement its own handler for this.
2062 	 */
2063 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2064 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2065 		return err;
2066 	}
2067 
2068 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2069 		err = dev_uc_add_excl(dev, addr);
2070 	else if (is_multicast_ether_addr(addr))
2071 		err = dev_mc_add_excl(dev, addr);
2072 
2073 	/* Only return duplicate errors if NLM_F_EXCL is set */
2074 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2075 		err = 0;
2076 
2077 	return err;
2078 }
2079 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2080 
2081 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2082 {
2083 	struct net *net = sock_net(skb->sk);
2084 	struct ndmsg *ndm;
2085 	struct nlattr *tb[NDA_MAX+1];
2086 	struct net_device *dev;
2087 	u8 *addr;
2088 	int err;
2089 
2090 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2091 	if (err < 0)
2092 		return err;
2093 
2094 	ndm = nlmsg_data(nlh);
2095 	if (ndm->ndm_ifindex == 0) {
2096 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2097 		return -EINVAL;
2098 	}
2099 
2100 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2101 	if (dev == NULL) {
2102 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2103 		return -ENODEV;
2104 	}
2105 
2106 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2107 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2108 		return -EINVAL;
2109 	}
2110 
2111 	addr = nla_data(tb[NDA_LLADDR]);
2112 
2113 	err = -EOPNOTSUPP;
2114 
2115 	/* Support fdb on master device the net/bridge default case */
2116 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2117 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2118 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2119 		const struct net_device_ops *ops = br_dev->netdev_ops;
2120 
2121 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2122 		if (err)
2123 			goto out;
2124 		else
2125 			ndm->ndm_flags &= ~NTF_MASTER;
2126 	}
2127 
2128 	/* Embedded bridge, macvlan, and any other device support */
2129 	if ((ndm->ndm_flags & NTF_SELF)) {
2130 		if (dev->netdev_ops->ndo_fdb_add)
2131 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2132 							   nlh->nlmsg_flags);
2133 		else
2134 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2135 					       nlh->nlmsg_flags);
2136 
2137 		if (!err) {
2138 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2139 			ndm->ndm_flags &= ~NTF_SELF;
2140 		}
2141 	}
2142 out:
2143 	return err;
2144 }
2145 
2146 /**
2147  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2148  */
2149 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2150 		     struct nlattr *tb[],
2151 		     struct net_device *dev,
2152 		     const unsigned char *addr)
2153 {
2154 	int err = -EOPNOTSUPP;
2155 
2156 	/* If aging addresses are supported device will need to
2157 	 * implement its own handler for this.
2158 	 */
2159 	if (ndm->ndm_state & NUD_PERMANENT) {
2160 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2161 		return -EINVAL;
2162 	}
2163 
2164 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2165 		err = dev_uc_del(dev, addr);
2166 	else if (is_multicast_ether_addr(addr))
2167 		err = dev_mc_del(dev, addr);
2168 	else
2169 		err = -EINVAL;
2170 
2171 	return err;
2172 }
2173 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2174 
2175 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2176 {
2177 	struct net *net = sock_net(skb->sk);
2178 	struct ndmsg *ndm;
2179 	struct nlattr *tb[NDA_MAX+1];
2180 	struct net_device *dev;
2181 	int err = -EINVAL;
2182 	__u8 *addr;
2183 
2184 	if (!capable(CAP_NET_ADMIN))
2185 		return -EPERM;
2186 
2187 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2188 	if (err < 0)
2189 		return err;
2190 
2191 	ndm = nlmsg_data(nlh);
2192 	if (ndm->ndm_ifindex == 0) {
2193 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2194 		return -EINVAL;
2195 	}
2196 
2197 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2198 	if (dev == NULL) {
2199 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2200 		return -ENODEV;
2201 	}
2202 
2203 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2204 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2205 		return -EINVAL;
2206 	}
2207 
2208 	addr = nla_data(tb[NDA_LLADDR]);
2209 
2210 	err = -EOPNOTSUPP;
2211 
2212 	/* Support fdb on master device the net/bridge default case */
2213 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2214 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2215 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2216 		const struct net_device_ops *ops = br_dev->netdev_ops;
2217 
2218 		if (ops->ndo_fdb_del)
2219 			err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2220 
2221 		if (err)
2222 			goto out;
2223 		else
2224 			ndm->ndm_flags &= ~NTF_MASTER;
2225 	}
2226 
2227 	/* Embedded bridge, macvlan, and any other device support */
2228 	if (ndm->ndm_flags & NTF_SELF) {
2229 		if (dev->netdev_ops->ndo_fdb_del)
2230 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2231 		else
2232 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2233 
2234 		if (!err) {
2235 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2236 			ndm->ndm_flags &= ~NTF_SELF;
2237 		}
2238 	}
2239 out:
2240 	return err;
2241 }
2242 
2243 static int nlmsg_populate_fdb(struct sk_buff *skb,
2244 			      struct netlink_callback *cb,
2245 			      struct net_device *dev,
2246 			      int *idx,
2247 			      struct netdev_hw_addr_list *list)
2248 {
2249 	struct netdev_hw_addr *ha;
2250 	int err;
2251 	u32 portid, seq;
2252 
2253 	portid = NETLINK_CB(cb->skb).portid;
2254 	seq = cb->nlh->nlmsg_seq;
2255 
2256 	list_for_each_entry(ha, &list->list, list) {
2257 		if (*idx < cb->args[0])
2258 			goto skip;
2259 
2260 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2261 					      portid, seq,
2262 					      RTM_NEWNEIGH, NTF_SELF);
2263 		if (err < 0)
2264 			return err;
2265 skip:
2266 		*idx += 1;
2267 	}
2268 	return 0;
2269 }
2270 
2271 /**
2272  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2273  * @nlh: netlink message header
2274  * @dev: netdevice
2275  *
2276  * Default netdevice operation to dump the existing unicast address list.
2277  * Returns number of addresses from list put in skb.
2278  */
2279 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2280 		      struct netlink_callback *cb,
2281 		      struct net_device *dev,
2282 		      int idx)
2283 {
2284 	int err;
2285 
2286 	netif_addr_lock_bh(dev);
2287 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2288 	if (err)
2289 		goto out;
2290 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2291 out:
2292 	netif_addr_unlock_bh(dev);
2293 	return idx;
2294 }
2295 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2296 
2297 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2298 {
2299 	int idx = 0;
2300 	struct net *net = sock_net(skb->sk);
2301 	struct net_device *dev;
2302 
2303 	rcu_read_lock();
2304 	for_each_netdev_rcu(net, dev) {
2305 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2306 			struct net_device *br_dev;
2307 			const struct net_device_ops *ops;
2308 
2309 			br_dev = netdev_master_upper_dev_get(dev);
2310 			ops = br_dev->netdev_ops;
2311 			if (ops->ndo_fdb_dump)
2312 				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2313 		}
2314 
2315 		if (dev->netdev_ops->ndo_fdb_dump)
2316 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2317 		else
2318 			idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2319 	}
2320 	rcu_read_unlock();
2321 
2322 	cb->args[0] = idx;
2323 	return skb->len;
2324 }
2325 
2326 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2327 			    struct net_device *dev, u16 mode)
2328 {
2329 	struct nlmsghdr *nlh;
2330 	struct ifinfomsg *ifm;
2331 	struct nlattr *br_afspec;
2332 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2333 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2334 
2335 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2336 	if (nlh == NULL)
2337 		return -EMSGSIZE;
2338 
2339 	ifm = nlmsg_data(nlh);
2340 	ifm->ifi_family = AF_BRIDGE;
2341 	ifm->__ifi_pad = 0;
2342 	ifm->ifi_type = dev->type;
2343 	ifm->ifi_index = dev->ifindex;
2344 	ifm->ifi_flags = dev_get_flags(dev);
2345 	ifm->ifi_change = 0;
2346 
2347 
2348 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2349 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2350 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2351 	    (br_dev &&
2352 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2353 	    (dev->addr_len &&
2354 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2355 	    (dev->ifindex != dev->iflink &&
2356 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2357 		goto nla_put_failure;
2358 
2359 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2360 	if (!br_afspec)
2361 		goto nla_put_failure;
2362 
2363 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2364 	    nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2365 		nla_nest_cancel(skb, br_afspec);
2366 		goto nla_put_failure;
2367 	}
2368 	nla_nest_end(skb, br_afspec);
2369 
2370 	return nlmsg_end(skb, nlh);
2371 nla_put_failure:
2372 	nlmsg_cancel(skb, nlh);
2373 	return -EMSGSIZE;
2374 }
2375 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2376 
2377 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2378 {
2379 	struct net *net = sock_net(skb->sk);
2380 	struct net_device *dev;
2381 	int idx = 0;
2382 	u32 portid = NETLINK_CB(cb->skb).portid;
2383 	u32 seq = cb->nlh->nlmsg_seq;
2384 	struct nlattr *extfilt;
2385 	u32 filter_mask = 0;
2386 
2387 	extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct rtgenmsg),
2388 				  IFLA_EXT_MASK);
2389 	if (extfilt)
2390 		filter_mask = nla_get_u32(extfilt);
2391 
2392 	rcu_read_lock();
2393 	for_each_netdev_rcu(net, dev) {
2394 		const struct net_device_ops *ops = dev->netdev_ops;
2395 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2396 
2397 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2398 			if (idx >= cb->args[0] &&
2399 			    br_dev->netdev_ops->ndo_bridge_getlink(
2400 				    skb, portid, seq, dev, filter_mask) < 0)
2401 				break;
2402 			idx++;
2403 		}
2404 
2405 		if (ops->ndo_bridge_getlink) {
2406 			if (idx >= cb->args[0] &&
2407 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2408 						    filter_mask) < 0)
2409 				break;
2410 			idx++;
2411 		}
2412 	}
2413 	rcu_read_unlock();
2414 	cb->args[0] = idx;
2415 
2416 	return skb->len;
2417 }
2418 
2419 static inline size_t bridge_nlmsg_size(void)
2420 {
2421 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2422 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2423 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2424 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2425 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2426 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2427 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2428 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2429 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2430 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2431 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2432 }
2433 
2434 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2435 {
2436 	struct net *net = dev_net(dev);
2437 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2438 	struct sk_buff *skb;
2439 	int err = -EOPNOTSUPP;
2440 
2441 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2442 	if (!skb) {
2443 		err = -ENOMEM;
2444 		goto errout;
2445 	}
2446 
2447 	if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2448 	    br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2449 		err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2450 		if (err < 0)
2451 			goto errout;
2452 	}
2453 
2454 	if ((flags & BRIDGE_FLAGS_SELF) &&
2455 	    dev->netdev_ops->ndo_bridge_getlink) {
2456 		err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2457 		if (err < 0)
2458 			goto errout;
2459 	}
2460 
2461 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2462 	return 0;
2463 errout:
2464 	WARN_ON(err == -EMSGSIZE);
2465 	kfree_skb(skb);
2466 	rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2467 	return err;
2468 }
2469 
2470 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2471 {
2472 	struct net *net = sock_net(skb->sk);
2473 	struct ifinfomsg *ifm;
2474 	struct net_device *dev;
2475 	struct nlattr *br_spec, *attr = NULL;
2476 	int rem, err = -EOPNOTSUPP;
2477 	u16 oflags, flags = 0;
2478 	bool have_flags = false;
2479 
2480 	if (nlmsg_len(nlh) < sizeof(*ifm))
2481 		return -EINVAL;
2482 
2483 	ifm = nlmsg_data(nlh);
2484 	if (ifm->ifi_family != AF_BRIDGE)
2485 		return -EPFNOSUPPORT;
2486 
2487 	dev = __dev_get_by_index(net, ifm->ifi_index);
2488 	if (!dev) {
2489 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2490 		return -ENODEV;
2491 	}
2492 
2493 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2494 	if (br_spec) {
2495 		nla_for_each_nested(attr, br_spec, rem) {
2496 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2497 				have_flags = true;
2498 				flags = nla_get_u16(attr);
2499 				break;
2500 			}
2501 		}
2502 	}
2503 
2504 	oflags = flags;
2505 
2506 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2507 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2508 
2509 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2510 			err = -EOPNOTSUPP;
2511 			goto out;
2512 		}
2513 
2514 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2515 		if (err)
2516 			goto out;
2517 
2518 		flags &= ~BRIDGE_FLAGS_MASTER;
2519 	}
2520 
2521 	if ((flags & BRIDGE_FLAGS_SELF)) {
2522 		if (!dev->netdev_ops->ndo_bridge_setlink)
2523 			err = -EOPNOTSUPP;
2524 		else
2525 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2526 
2527 		if (!err)
2528 			flags &= ~BRIDGE_FLAGS_SELF;
2529 	}
2530 
2531 	if (have_flags)
2532 		memcpy(nla_data(attr), &flags, sizeof(flags));
2533 	/* Generate event to notify upper layer of bridge change */
2534 	if (!err)
2535 		err = rtnl_bridge_notify(dev, oflags);
2536 out:
2537 	return err;
2538 }
2539 
2540 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2541 {
2542 	struct net *net = sock_net(skb->sk);
2543 	struct ifinfomsg *ifm;
2544 	struct net_device *dev;
2545 	struct nlattr *br_spec, *attr = NULL;
2546 	int rem, err = -EOPNOTSUPP;
2547 	u16 oflags, flags = 0;
2548 	bool have_flags = false;
2549 
2550 	if (nlmsg_len(nlh) < sizeof(*ifm))
2551 		return -EINVAL;
2552 
2553 	ifm = nlmsg_data(nlh);
2554 	if (ifm->ifi_family != AF_BRIDGE)
2555 		return -EPFNOSUPPORT;
2556 
2557 	dev = __dev_get_by_index(net, ifm->ifi_index);
2558 	if (!dev) {
2559 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2560 		return -ENODEV;
2561 	}
2562 
2563 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2564 	if (br_spec) {
2565 		nla_for_each_nested(attr, br_spec, rem) {
2566 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2567 				have_flags = true;
2568 				flags = nla_get_u16(attr);
2569 				break;
2570 			}
2571 		}
2572 	}
2573 
2574 	oflags = flags;
2575 
2576 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2577 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2578 
2579 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2580 			err = -EOPNOTSUPP;
2581 			goto out;
2582 		}
2583 
2584 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2585 		if (err)
2586 			goto out;
2587 
2588 		flags &= ~BRIDGE_FLAGS_MASTER;
2589 	}
2590 
2591 	if ((flags & BRIDGE_FLAGS_SELF)) {
2592 		if (!dev->netdev_ops->ndo_bridge_dellink)
2593 			err = -EOPNOTSUPP;
2594 		else
2595 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2596 
2597 		if (!err)
2598 			flags &= ~BRIDGE_FLAGS_SELF;
2599 	}
2600 
2601 	if (have_flags)
2602 		memcpy(nla_data(attr), &flags, sizeof(flags));
2603 	/* Generate event to notify upper layer of bridge change */
2604 	if (!err)
2605 		err = rtnl_bridge_notify(dev, oflags);
2606 out:
2607 	return err;
2608 }
2609 
2610 /* Process one rtnetlink message. */
2611 
2612 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2613 {
2614 	struct net *net = sock_net(skb->sk);
2615 	rtnl_doit_func doit;
2616 	int sz_idx, kind;
2617 	int family;
2618 	int type;
2619 	int err;
2620 
2621 	type = nlh->nlmsg_type;
2622 	if (type > RTM_MAX)
2623 		return -EOPNOTSUPP;
2624 
2625 	type -= RTM_BASE;
2626 
2627 	/* All the messages must have at least 1 byte length */
2628 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2629 		return 0;
2630 
2631 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2632 	sz_idx = type>>2;
2633 	kind = type&3;
2634 
2635 	if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2636 		return -EPERM;
2637 
2638 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2639 		struct sock *rtnl;
2640 		rtnl_dumpit_func dumpit;
2641 		rtnl_calcit_func calcit;
2642 		u16 min_dump_alloc = 0;
2643 
2644 		dumpit = rtnl_get_dumpit(family, type);
2645 		if (dumpit == NULL)
2646 			return -EOPNOTSUPP;
2647 		calcit = rtnl_get_calcit(family, type);
2648 		if (calcit)
2649 			min_dump_alloc = calcit(skb, nlh);
2650 
2651 		__rtnl_unlock();
2652 		rtnl = net->rtnl;
2653 		{
2654 			struct netlink_dump_control c = {
2655 				.dump		= dumpit,
2656 				.min_dump_alloc	= min_dump_alloc,
2657 			};
2658 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2659 		}
2660 		rtnl_lock();
2661 		return err;
2662 	}
2663 
2664 	doit = rtnl_get_doit(family, type);
2665 	if (doit == NULL)
2666 		return -EOPNOTSUPP;
2667 
2668 	return doit(skb, nlh);
2669 }
2670 
2671 static void rtnetlink_rcv(struct sk_buff *skb)
2672 {
2673 	rtnl_lock();
2674 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2675 	rtnl_unlock();
2676 }
2677 
2678 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2679 {
2680 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2681 
2682 	switch (event) {
2683 	case NETDEV_UP:
2684 	case NETDEV_DOWN:
2685 	case NETDEV_PRE_UP:
2686 	case NETDEV_POST_INIT:
2687 	case NETDEV_REGISTER:
2688 	case NETDEV_CHANGE:
2689 	case NETDEV_PRE_TYPE_CHANGE:
2690 	case NETDEV_GOING_DOWN:
2691 	case NETDEV_UNREGISTER:
2692 	case NETDEV_UNREGISTER_FINAL:
2693 	case NETDEV_RELEASE:
2694 	case NETDEV_JOIN:
2695 		break;
2696 	default:
2697 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2698 		break;
2699 	}
2700 	return NOTIFY_DONE;
2701 }
2702 
2703 static struct notifier_block rtnetlink_dev_notifier = {
2704 	.notifier_call	= rtnetlink_event,
2705 };
2706 
2707 
2708 static int __net_init rtnetlink_net_init(struct net *net)
2709 {
2710 	struct sock *sk;
2711 	struct netlink_kernel_cfg cfg = {
2712 		.groups		= RTNLGRP_MAX,
2713 		.input		= rtnetlink_rcv,
2714 		.cb_mutex	= &rtnl_mutex,
2715 		.flags		= NL_CFG_F_NONROOT_RECV,
2716 	};
2717 
2718 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2719 	if (!sk)
2720 		return -ENOMEM;
2721 	net->rtnl = sk;
2722 	return 0;
2723 }
2724 
2725 static void __net_exit rtnetlink_net_exit(struct net *net)
2726 {
2727 	netlink_kernel_release(net->rtnl);
2728 	net->rtnl = NULL;
2729 }
2730 
2731 static struct pernet_operations rtnetlink_net_ops = {
2732 	.init = rtnetlink_net_init,
2733 	.exit = rtnetlink_net_exit,
2734 };
2735 
2736 void __init rtnetlink_init(void)
2737 {
2738 	if (register_pernet_subsys(&rtnetlink_net_ops))
2739 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2740 
2741 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2742 
2743 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2744 		      rtnl_dump_ifinfo, rtnl_calcit);
2745 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2746 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2747 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2748 
2749 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2750 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2751 
2752 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2753 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2754 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2755 
2756 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2757 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2758 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2759 }
2760 
2761