xref: /openbmc/linux/net/core/rtnetlink.c (revision 77d84ff8)
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 	       + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
772 }
773 
774 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
775 {
776 	struct nlattr *vf_ports;
777 	struct nlattr *vf_port;
778 	int vf;
779 	int err;
780 
781 	vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
782 	if (!vf_ports)
783 		return -EMSGSIZE;
784 
785 	for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
786 		vf_port = nla_nest_start(skb, IFLA_VF_PORT);
787 		if (!vf_port)
788 			goto nla_put_failure;
789 		if (nla_put_u32(skb, IFLA_PORT_VF, vf))
790 			goto nla_put_failure;
791 		err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
792 		if (err == -EMSGSIZE)
793 			goto nla_put_failure;
794 		if (err) {
795 			nla_nest_cancel(skb, vf_port);
796 			continue;
797 		}
798 		nla_nest_end(skb, vf_port);
799 	}
800 
801 	nla_nest_end(skb, vf_ports);
802 
803 	return 0;
804 
805 nla_put_failure:
806 	nla_nest_cancel(skb, vf_ports);
807 	return -EMSGSIZE;
808 }
809 
810 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
811 {
812 	struct nlattr *port_self;
813 	int err;
814 
815 	port_self = nla_nest_start(skb, IFLA_PORT_SELF);
816 	if (!port_self)
817 		return -EMSGSIZE;
818 
819 	err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
820 	if (err) {
821 		nla_nest_cancel(skb, port_self);
822 		return (err == -EMSGSIZE) ? err : 0;
823 	}
824 
825 	nla_nest_end(skb, port_self);
826 
827 	return 0;
828 }
829 
830 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
831 {
832 	int err;
833 
834 	if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
835 		return 0;
836 
837 	err = rtnl_port_self_fill(skb, dev);
838 	if (err)
839 		return err;
840 
841 	if (dev_num_vf(dev->dev.parent)) {
842 		err = rtnl_vf_ports_fill(skb, dev);
843 		if (err)
844 			return err;
845 	}
846 
847 	return 0;
848 }
849 
850 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
851 {
852 	int err;
853 	struct netdev_phys_port_id ppid;
854 
855 	err = dev_get_phys_port_id(dev, &ppid);
856 	if (err) {
857 		if (err == -EOPNOTSUPP)
858 			return 0;
859 		return err;
860 	}
861 
862 	if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
863 		return -EMSGSIZE;
864 
865 	return 0;
866 }
867 
868 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
869 			    int type, u32 pid, u32 seq, u32 change,
870 			    unsigned int flags, u32 ext_filter_mask)
871 {
872 	struct ifinfomsg *ifm;
873 	struct nlmsghdr *nlh;
874 	struct rtnl_link_stats64 temp;
875 	const struct rtnl_link_stats64 *stats;
876 	struct nlattr *attr, *af_spec;
877 	struct rtnl_af_ops *af_ops;
878 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
879 
880 	ASSERT_RTNL();
881 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
882 	if (nlh == NULL)
883 		return -EMSGSIZE;
884 
885 	ifm = nlmsg_data(nlh);
886 	ifm->ifi_family = AF_UNSPEC;
887 	ifm->__ifi_pad = 0;
888 	ifm->ifi_type = dev->type;
889 	ifm->ifi_index = dev->ifindex;
890 	ifm->ifi_flags = dev_get_flags(dev);
891 	ifm->ifi_change = change;
892 
893 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
894 	    nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
895 	    nla_put_u8(skb, IFLA_OPERSTATE,
896 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
897 	    nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
898 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
899 	    nla_put_u32(skb, IFLA_GROUP, dev->group) ||
900 	    nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
901 	    nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
902 #ifdef CONFIG_RPS
903 	    nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
904 #endif
905 	    (dev->ifindex != dev->iflink &&
906 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
907 	    (upper_dev &&
908 	     nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
909 	    nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
910 	    (dev->qdisc &&
911 	     nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
912 	    (dev->ifalias &&
913 	     nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
914 		goto nla_put_failure;
915 
916 	if (1) {
917 		struct rtnl_link_ifmap map = {
918 			.mem_start   = dev->mem_start,
919 			.mem_end     = dev->mem_end,
920 			.base_addr   = dev->base_addr,
921 			.irq         = dev->irq,
922 			.dma         = dev->dma,
923 			.port        = dev->if_port,
924 		};
925 		if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
926 			goto nla_put_failure;
927 	}
928 
929 	if (dev->addr_len) {
930 		if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
931 		    nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
932 			goto nla_put_failure;
933 	}
934 
935 	if (rtnl_phys_port_id_fill(skb, dev))
936 		goto nla_put_failure;
937 
938 	attr = nla_reserve(skb, IFLA_STATS,
939 			sizeof(struct rtnl_link_stats));
940 	if (attr == NULL)
941 		goto nla_put_failure;
942 
943 	stats = dev_get_stats(dev, &temp);
944 	copy_rtnl_link_stats(nla_data(attr), stats);
945 
946 	attr = nla_reserve(skb, IFLA_STATS64,
947 			sizeof(struct rtnl_link_stats64));
948 	if (attr == NULL)
949 		goto nla_put_failure;
950 	copy_rtnl_link_stats64(nla_data(attr), stats);
951 
952 	if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
953 	    nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
954 		goto nla_put_failure;
955 
956 	if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
957 	    && (ext_filter_mask & RTEXT_FILTER_VF)) {
958 		int i;
959 
960 		struct nlattr *vfinfo, *vf;
961 		int num_vfs = dev_num_vf(dev->dev.parent);
962 
963 		vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
964 		if (!vfinfo)
965 			goto nla_put_failure;
966 		for (i = 0; i < num_vfs; i++) {
967 			struct ifla_vf_info ivi;
968 			struct ifla_vf_mac vf_mac;
969 			struct ifla_vf_vlan vf_vlan;
970 			struct ifla_vf_tx_rate vf_tx_rate;
971 			struct ifla_vf_spoofchk vf_spoofchk;
972 			struct ifla_vf_link_state vf_linkstate;
973 
974 			/*
975 			 * Not all SR-IOV capable drivers support the
976 			 * spoofcheck query.  Preset to -1 so the user
977 			 * space tool can detect that the driver didn't
978 			 * report anything.
979 			 */
980 			ivi.spoofchk = -1;
981 			memset(ivi.mac, 0, sizeof(ivi.mac));
982 			/* The default value for VF link state is "auto"
983 			 * IFLA_VF_LINK_STATE_AUTO which equals zero
984 			 */
985 			ivi.linkstate = 0;
986 			if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
987 				break;
988 			vf_mac.vf =
989 				vf_vlan.vf =
990 				vf_tx_rate.vf =
991 				vf_spoofchk.vf =
992 				vf_linkstate.vf = ivi.vf;
993 
994 			memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
995 			vf_vlan.vlan = ivi.vlan;
996 			vf_vlan.qos = ivi.qos;
997 			vf_tx_rate.rate = ivi.tx_rate;
998 			vf_spoofchk.setting = ivi.spoofchk;
999 			vf_linkstate.link_state = ivi.linkstate;
1000 			vf = nla_nest_start(skb, IFLA_VF_INFO);
1001 			if (!vf) {
1002 				nla_nest_cancel(skb, vfinfo);
1003 				goto nla_put_failure;
1004 			}
1005 			if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1006 			    nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1007 			    nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1008 				    &vf_tx_rate) ||
1009 			    nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1010 				    &vf_spoofchk) ||
1011 			    nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1012 				    &vf_linkstate))
1013 				goto nla_put_failure;
1014 			nla_nest_end(skb, vf);
1015 		}
1016 		nla_nest_end(skb, vfinfo);
1017 	}
1018 
1019 	if (rtnl_port_fill(skb, dev))
1020 		goto nla_put_failure;
1021 
1022 	if (dev->rtnl_link_ops) {
1023 		if (rtnl_link_fill(skb, dev) < 0)
1024 			goto nla_put_failure;
1025 	}
1026 
1027 	if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1028 		goto nla_put_failure;
1029 
1030 	list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1031 		if (af_ops->fill_link_af) {
1032 			struct nlattr *af;
1033 			int err;
1034 
1035 			if (!(af = nla_nest_start(skb, af_ops->family)))
1036 				goto nla_put_failure;
1037 
1038 			err = af_ops->fill_link_af(skb, dev);
1039 
1040 			/*
1041 			 * Caller may return ENODATA to indicate that there
1042 			 * was no data to be dumped. This is not an error, it
1043 			 * means we should trim the attribute header and
1044 			 * continue.
1045 			 */
1046 			if (err == -ENODATA)
1047 				nla_nest_cancel(skb, af);
1048 			else if (err < 0)
1049 				goto nla_put_failure;
1050 
1051 			nla_nest_end(skb, af);
1052 		}
1053 	}
1054 
1055 	nla_nest_end(skb, af_spec);
1056 
1057 	return nlmsg_end(skb, nlh);
1058 
1059 nla_put_failure:
1060 	nlmsg_cancel(skb, nlh);
1061 	return -EMSGSIZE;
1062 }
1063 
1064 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1065 {
1066 	struct net *net = sock_net(skb->sk);
1067 	int h, s_h;
1068 	int idx = 0, s_idx;
1069 	struct net_device *dev;
1070 	struct hlist_head *head;
1071 	struct nlattr *tb[IFLA_MAX+1];
1072 	u32 ext_filter_mask = 0;
1073 
1074 	s_h = cb->args[0];
1075 	s_idx = cb->args[1];
1076 
1077 	rcu_read_lock();
1078 	cb->seq = net->dev_base_seq;
1079 
1080 	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1081 			ifla_policy) >= 0) {
1082 
1083 		if (tb[IFLA_EXT_MASK])
1084 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1085 	}
1086 
1087 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1088 		idx = 0;
1089 		head = &net->dev_index_head[h];
1090 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
1091 			if (idx < s_idx)
1092 				goto cont;
1093 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1094 					     NETLINK_CB(cb->skb).portid,
1095 					     cb->nlh->nlmsg_seq, 0,
1096 					     NLM_F_MULTI,
1097 					     ext_filter_mask) <= 0)
1098 				goto out;
1099 
1100 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1101 cont:
1102 			idx++;
1103 		}
1104 	}
1105 out:
1106 	rcu_read_unlock();
1107 	cb->args[1] = idx;
1108 	cb->args[0] = h;
1109 
1110 	return skb->len;
1111 }
1112 
1113 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1114 	[IFLA_IFNAME]		= { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1115 	[IFLA_ADDRESS]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1116 	[IFLA_BROADCAST]	= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1117 	[IFLA_MAP]		= { .len = sizeof(struct rtnl_link_ifmap) },
1118 	[IFLA_MTU]		= { .type = NLA_U32 },
1119 	[IFLA_LINK]		= { .type = NLA_U32 },
1120 	[IFLA_MASTER]		= { .type = NLA_U32 },
1121 	[IFLA_CARRIER]		= { .type = NLA_U8 },
1122 	[IFLA_TXQLEN]		= { .type = NLA_U32 },
1123 	[IFLA_WEIGHT]		= { .type = NLA_U32 },
1124 	[IFLA_OPERSTATE]	= { .type = NLA_U8 },
1125 	[IFLA_LINKMODE]		= { .type = NLA_U8 },
1126 	[IFLA_LINKINFO]		= { .type = NLA_NESTED },
1127 	[IFLA_NET_NS_PID]	= { .type = NLA_U32 },
1128 	[IFLA_NET_NS_FD]	= { .type = NLA_U32 },
1129 	[IFLA_IFALIAS]	        = { .type = NLA_STRING, .len = IFALIASZ-1 },
1130 	[IFLA_VFINFO_LIST]	= {. type = NLA_NESTED },
1131 	[IFLA_VF_PORTS]		= { .type = NLA_NESTED },
1132 	[IFLA_PORT_SELF]	= { .type = NLA_NESTED },
1133 	[IFLA_AF_SPEC]		= { .type = NLA_NESTED },
1134 	[IFLA_EXT_MASK]		= { .type = NLA_U32 },
1135 	[IFLA_PROMISCUITY]	= { .type = NLA_U32 },
1136 	[IFLA_NUM_TX_QUEUES]	= { .type = NLA_U32 },
1137 	[IFLA_NUM_RX_QUEUES]	= { .type = NLA_U32 },
1138 	[IFLA_PHYS_PORT_ID]	= { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1139 };
1140 EXPORT_SYMBOL(ifla_policy);
1141 
1142 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1143 	[IFLA_INFO_KIND]	= { .type = NLA_STRING },
1144 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
1145 };
1146 
1147 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1148 	[IFLA_VF_INFO]		= { .type = NLA_NESTED },
1149 };
1150 
1151 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1152 	[IFLA_VF_MAC]		= { .type = NLA_BINARY,
1153 				    .len = sizeof(struct ifla_vf_mac) },
1154 	[IFLA_VF_VLAN]		= { .type = NLA_BINARY,
1155 				    .len = sizeof(struct ifla_vf_vlan) },
1156 	[IFLA_VF_TX_RATE]	= { .type = NLA_BINARY,
1157 				    .len = sizeof(struct ifla_vf_tx_rate) },
1158 	[IFLA_VF_SPOOFCHK]	= { .type = NLA_BINARY,
1159 				    .len = sizeof(struct ifla_vf_spoofchk) },
1160 };
1161 
1162 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1163 	[IFLA_PORT_VF]		= { .type = NLA_U32 },
1164 	[IFLA_PORT_PROFILE]	= { .type = NLA_STRING,
1165 				    .len = PORT_PROFILE_MAX },
1166 	[IFLA_PORT_VSI_TYPE]	= { .type = NLA_BINARY,
1167 				    .len = sizeof(struct ifla_port_vsi)},
1168 	[IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1169 				      .len = PORT_UUID_MAX },
1170 	[IFLA_PORT_HOST_UUID]	= { .type = NLA_STRING,
1171 				    .len = PORT_UUID_MAX },
1172 	[IFLA_PORT_REQUEST]	= { .type = NLA_U8, },
1173 	[IFLA_PORT_RESPONSE]	= { .type = NLA_U16, },
1174 };
1175 
1176 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1177 {
1178 	struct net *net;
1179 	/* Examine the link attributes and figure out which
1180 	 * network namespace we are talking about.
1181 	 */
1182 	if (tb[IFLA_NET_NS_PID])
1183 		net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1184 	else if (tb[IFLA_NET_NS_FD])
1185 		net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1186 	else
1187 		net = get_net(src_net);
1188 	return net;
1189 }
1190 EXPORT_SYMBOL(rtnl_link_get_net);
1191 
1192 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1193 {
1194 	if (dev) {
1195 		if (tb[IFLA_ADDRESS] &&
1196 		    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1197 			return -EINVAL;
1198 
1199 		if (tb[IFLA_BROADCAST] &&
1200 		    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1201 			return -EINVAL;
1202 	}
1203 
1204 	if (tb[IFLA_AF_SPEC]) {
1205 		struct nlattr *af;
1206 		int rem, err;
1207 
1208 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1209 			const struct rtnl_af_ops *af_ops;
1210 
1211 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1212 				return -EAFNOSUPPORT;
1213 
1214 			if (!af_ops->set_link_af)
1215 				return -EOPNOTSUPP;
1216 
1217 			if (af_ops->validate_link_af) {
1218 				err = af_ops->validate_link_af(dev, af);
1219 				if (err < 0)
1220 					return err;
1221 			}
1222 		}
1223 	}
1224 
1225 	return 0;
1226 }
1227 
1228 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1229 {
1230 	int rem, err = -EINVAL;
1231 	struct nlattr *vf;
1232 	const struct net_device_ops *ops = dev->netdev_ops;
1233 
1234 	nla_for_each_nested(vf, attr, rem) {
1235 		switch (nla_type(vf)) {
1236 		case IFLA_VF_MAC: {
1237 			struct ifla_vf_mac *ivm;
1238 			ivm = nla_data(vf);
1239 			err = -EOPNOTSUPP;
1240 			if (ops->ndo_set_vf_mac)
1241 				err = ops->ndo_set_vf_mac(dev, ivm->vf,
1242 							  ivm->mac);
1243 			break;
1244 		}
1245 		case IFLA_VF_VLAN: {
1246 			struct ifla_vf_vlan *ivv;
1247 			ivv = nla_data(vf);
1248 			err = -EOPNOTSUPP;
1249 			if (ops->ndo_set_vf_vlan)
1250 				err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1251 							   ivv->vlan,
1252 							   ivv->qos);
1253 			break;
1254 		}
1255 		case IFLA_VF_TX_RATE: {
1256 			struct ifla_vf_tx_rate *ivt;
1257 			ivt = nla_data(vf);
1258 			err = -EOPNOTSUPP;
1259 			if (ops->ndo_set_vf_tx_rate)
1260 				err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1261 							      ivt->rate);
1262 			break;
1263 		}
1264 		case IFLA_VF_SPOOFCHK: {
1265 			struct ifla_vf_spoofchk *ivs;
1266 			ivs = nla_data(vf);
1267 			err = -EOPNOTSUPP;
1268 			if (ops->ndo_set_vf_spoofchk)
1269 				err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1270 							       ivs->setting);
1271 			break;
1272 		}
1273 		case IFLA_VF_LINK_STATE: {
1274 			struct ifla_vf_link_state *ivl;
1275 			ivl = nla_data(vf);
1276 			err = -EOPNOTSUPP;
1277 			if (ops->ndo_set_vf_link_state)
1278 				err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1279 								 ivl->link_state);
1280 			break;
1281 		}
1282 		default:
1283 			err = -EINVAL;
1284 			break;
1285 		}
1286 		if (err)
1287 			break;
1288 	}
1289 	return err;
1290 }
1291 
1292 static int do_set_master(struct net_device *dev, int ifindex)
1293 {
1294 	struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1295 	const struct net_device_ops *ops;
1296 	int err;
1297 
1298 	if (upper_dev) {
1299 		if (upper_dev->ifindex == ifindex)
1300 			return 0;
1301 		ops = upper_dev->netdev_ops;
1302 		if (ops->ndo_del_slave) {
1303 			err = ops->ndo_del_slave(upper_dev, dev);
1304 			if (err)
1305 				return err;
1306 		} else {
1307 			return -EOPNOTSUPP;
1308 		}
1309 	}
1310 
1311 	if (ifindex) {
1312 		upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1313 		if (!upper_dev)
1314 			return -EINVAL;
1315 		ops = upper_dev->netdev_ops;
1316 		if (ops->ndo_add_slave) {
1317 			err = ops->ndo_add_slave(upper_dev, dev);
1318 			if (err)
1319 				return err;
1320 		} else {
1321 			return -EOPNOTSUPP;
1322 		}
1323 	}
1324 	return 0;
1325 }
1326 
1327 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1328 		      struct nlattr **tb, char *ifname, int modified)
1329 {
1330 	const struct net_device_ops *ops = dev->netdev_ops;
1331 	int err;
1332 
1333 	if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1334 		struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1335 		if (IS_ERR(net)) {
1336 			err = PTR_ERR(net);
1337 			goto errout;
1338 		}
1339 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1340 			err = -EPERM;
1341 			goto errout;
1342 		}
1343 		err = dev_change_net_namespace(dev, net, ifname);
1344 		put_net(net);
1345 		if (err)
1346 			goto errout;
1347 		modified = 1;
1348 	}
1349 
1350 	if (tb[IFLA_MAP]) {
1351 		struct rtnl_link_ifmap *u_map;
1352 		struct ifmap k_map;
1353 
1354 		if (!ops->ndo_set_config) {
1355 			err = -EOPNOTSUPP;
1356 			goto errout;
1357 		}
1358 
1359 		if (!netif_device_present(dev)) {
1360 			err = -ENODEV;
1361 			goto errout;
1362 		}
1363 
1364 		u_map = nla_data(tb[IFLA_MAP]);
1365 		k_map.mem_start = (unsigned long) u_map->mem_start;
1366 		k_map.mem_end = (unsigned long) u_map->mem_end;
1367 		k_map.base_addr = (unsigned short) u_map->base_addr;
1368 		k_map.irq = (unsigned char) u_map->irq;
1369 		k_map.dma = (unsigned char) u_map->dma;
1370 		k_map.port = (unsigned char) u_map->port;
1371 
1372 		err = ops->ndo_set_config(dev, &k_map);
1373 		if (err < 0)
1374 			goto errout;
1375 
1376 		modified = 1;
1377 	}
1378 
1379 	if (tb[IFLA_ADDRESS]) {
1380 		struct sockaddr *sa;
1381 		int len;
1382 
1383 		len = sizeof(sa_family_t) + dev->addr_len;
1384 		sa = kmalloc(len, GFP_KERNEL);
1385 		if (!sa) {
1386 			err = -ENOMEM;
1387 			goto errout;
1388 		}
1389 		sa->sa_family = dev->type;
1390 		memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1391 		       dev->addr_len);
1392 		err = dev_set_mac_address(dev, sa);
1393 		kfree(sa);
1394 		if (err)
1395 			goto errout;
1396 		modified = 1;
1397 	}
1398 
1399 	if (tb[IFLA_MTU]) {
1400 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1401 		if (err < 0)
1402 			goto errout;
1403 		modified = 1;
1404 	}
1405 
1406 	if (tb[IFLA_GROUP]) {
1407 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1408 		modified = 1;
1409 	}
1410 
1411 	/*
1412 	 * Interface selected by interface index but interface
1413 	 * name provided implies that a name change has been
1414 	 * requested.
1415 	 */
1416 	if (ifm->ifi_index > 0 && ifname[0]) {
1417 		err = dev_change_name(dev, ifname);
1418 		if (err < 0)
1419 			goto errout;
1420 		modified = 1;
1421 	}
1422 
1423 	if (tb[IFLA_IFALIAS]) {
1424 		err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1425 				    nla_len(tb[IFLA_IFALIAS]));
1426 		if (err < 0)
1427 			goto errout;
1428 		modified = 1;
1429 	}
1430 
1431 	if (tb[IFLA_BROADCAST]) {
1432 		nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1433 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1434 	}
1435 
1436 	if (ifm->ifi_flags || ifm->ifi_change) {
1437 		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1438 		if (err < 0)
1439 			goto errout;
1440 	}
1441 
1442 	if (tb[IFLA_MASTER]) {
1443 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1444 		if (err)
1445 			goto errout;
1446 		modified = 1;
1447 	}
1448 
1449 	if (tb[IFLA_CARRIER]) {
1450 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1451 		if (err)
1452 			goto errout;
1453 		modified = 1;
1454 	}
1455 
1456 	if (tb[IFLA_TXQLEN])
1457 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1458 
1459 	if (tb[IFLA_OPERSTATE])
1460 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1461 
1462 	if (tb[IFLA_LINKMODE]) {
1463 		write_lock_bh(&dev_base_lock);
1464 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1465 		write_unlock_bh(&dev_base_lock);
1466 	}
1467 
1468 	if (tb[IFLA_VFINFO_LIST]) {
1469 		struct nlattr *attr;
1470 		int rem;
1471 		nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1472 			if (nla_type(attr) != IFLA_VF_INFO) {
1473 				err = -EINVAL;
1474 				goto errout;
1475 			}
1476 			err = do_setvfinfo(dev, attr);
1477 			if (err < 0)
1478 				goto errout;
1479 			modified = 1;
1480 		}
1481 	}
1482 	err = 0;
1483 
1484 	if (tb[IFLA_VF_PORTS]) {
1485 		struct nlattr *port[IFLA_PORT_MAX+1];
1486 		struct nlattr *attr;
1487 		int vf;
1488 		int rem;
1489 
1490 		err = -EOPNOTSUPP;
1491 		if (!ops->ndo_set_vf_port)
1492 			goto errout;
1493 
1494 		nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1495 			if (nla_type(attr) != IFLA_VF_PORT)
1496 				continue;
1497 			err = nla_parse_nested(port, IFLA_PORT_MAX,
1498 				attr, ifla_port_policy);
1499 			if (err < 0)
1500 				goto errout;
1501 			if (!port[IFLA_PORT_VF]) {
1502 				err = -EOPNOTSUPP;
1503 				goto errout;
1504 			}
1505 			vf = nla_get_u32(port[IFLA_PORT_VF]);
1506 			err = ops->ndo_set_vf_port(dev, vf, port);
1507 			if (err < 0)
1508 				goto errout;
1509 			modified = 1;
1510 		}
1511 	}
1512 	err = 0;
1513 
1514 	if (tb[IFLA_PORT_SELF]) {
1515 		struct nlattr *port[IFLA_PORT_MAX+1];
1516 
1517 		err = nla_parse_nested(port, IFLA_PORT_MAX,
1518 			tb[IFLA_PORT_SELF], ifla_port_policy);
1519 		if (err < 0)
1520 			goto errout;
1521 
1522 		err = -EOPNOTSUPP;
1523 		if (ops->ndo_set_vf_port)
1524 			err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1525 		if (err < 0)
1526 			goto errout;
1527 		modified = 1;
1528 	}
1529 
1530 	if (tb[IFLA_AF_SPEC]) {
1531 		struct nlattr *af;
1532 		int rem;
1533 
1534 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1535 			const struct rtnl_af_ops *af_ops;
1536 
1537 			if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1538 				BUG();
1539 
1540 			err = af_ops->set_link_af(dev, af);
1541 			if (err < 0)
1542 				goto errout;
1543 
1544 			modified = 1;
1545 		}
1546 	}
1547 	err = 0;
1548 
1549 errout:
1550 	if (err < 0 && modified)
1551 		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",
1552 				     dev->name);
1553 
1554 	return err;
1555 }
1556 
1557 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1558 {
1559 	struct net *net = sock_net(skb->sk);
1560 	struct ifinfomsg *ifm;
1561 	struct net_device *dev;
1562 	int err;
1563 	struct nlattr *tb[IFLA_MAX+1];
1564 	char ifname[IFNAMSIZ];
1565 
1566 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1567 	if (err < 0)
1568 		goto errout;
1569 
1570 	if (tb[IFLA_IFNAME])
1571 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1572 	else
1573 		ifname[0] = '\0';
1574 
1575 	err = -EINVAL;
1576 	ifm = nlmsg_data(nlh);
1577 	if (ifm->ifi_index > 0)
1578 		dev = __dev_get_by_index(net, ifm->ifi_index);
1579 	else if (tb[IFLA_IFNAME])
1580 		dev = __dev_get_by_name(net, ifname);
1581 	else
1582 		goto errout;
1583 
1584 	if (dev == NULL) {
1585 		err = -ENODEV;
1586 		goto errout;
1587 	}
1588 
1589 	err = validate_linkmsg(dev, tb);
1590 	if (err < 0)
1591 		goto errout;
1592 
1593 	err = do_setlink(dev, ifm, tb, ifname, 0);
1594 errout:
1595 	return err;
1596 }
1597 
1598 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1599 {
1600 	struct net *net = sock_net(skb->sk);
1601 	const struct rtnl_link_ops *ops;
1602 	struct net_device *dev;
1603 	struct ifinfomsg *ifm;
1604 	char ifname[IFNAMSIZ];
1605 	struct nlattr *tb[IFLA_MAX+1];
1606 	int err;
1607 	LIST_HEAD(list_kill);
1608 
1609 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1610 	if (err < 0)
1611 		return err;
1612 
1613 	if (tb[IFLA_IFNAME])
1614 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1615 
1616 	ifm = nlmsg_data(nlh);
1617 	if (ifm->ifi_index > 0)
1618 		dev = __dev_get_by_index(net, ifm->ifi_index);
1619 	else if (tb[IFLA_IFNAME])
1620 		dev = __dev_get_by_name(net, ifname);
1621 	else
1622 		return -EINVAL;
1623 
1624 	if (!dev)
1625 		return -ENODEV;
1626 
1627 	ops = dev->rtnl_link_ops;
1628 	if (!ops)
1629 		return -EOPNOTSUPP;
1630 
1631 	ops->dellink(dev, &list_kill);
1632 	unregister_netdevice_many(&list_kill);
1633 	list_del(&list_kill);
1634 	return 0;
1635 }
1636 
1637 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1638 {
1639 	unsigned int old_flags;
1640 	int err;
1641 
1642 	old_flags = dev->flags;
1643 	if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1644 		err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1645 		if (err < 0)
1646 			return err;
1647 	}
1648 
1649 	dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1650 
1651 	__dev_notify_flags(dev, old_flags, ~0U);
1652 	return 0;
1653 }
1654 EXPORT_SYMBOL(rtnl_configure_link);
1655 
1656 struct net_device *rtnl_create_link(struct net *net,
1657 	char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1658 {
1659 	int err;
1660 	struct net_device *dev;
1661 	unsigned int num_tx_queues = 1;
1662 	unsigned int num_rx_queues = 1;
1663 
1664 	if (tb[IFLA_NUM_TX_QUEUES])
1665 		num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1666 	else if (ops->get_num_tx_queues)
1667 		num_tx_queues = ops->get_num_tx_queues();
1668 
1669 	if (tb[IFLA_NUM_RX_QUEUES])
1670 		num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1671 	else if (ops->get_num_rx_queues)
1672 		num_rx_queues = ops->get_num_rx_queues();
1673 
1674 	err = -ENOMEM;
1675 	dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1676 			       num_tx_queues, num_rx_queues);
1677 	if (!dev)
1678 		goto err;
1679 
1680 	dev_net_set(dev, net);
1681 	dev->rtnl_link_ops = ops;
1682 	dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1683 
1684 	if (tb[IFLA_MTU])
1685 		dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1686 	if (tb[IFLA_ADDRESS]) {
1687 		memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1688 				nla_len(tb[IFLA_ADDRESS]));
1689 		dev->addr_assign_type = NET_ADDR_SET;
1690 	}
1691 	if (tb[IFLA_BROADCAST])
1692 		memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1693 				nla_len(tb[IFLA_BROADCAST]));
1694 	if (tb[IFLA_TXQLEN])
1695 		dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1696 	if (tb[IFLA_OPERSTATE])
1697 		set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1698 	if (tb[IFLA_LINKMODE])
1699 		dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1700 	if (tb[IFLA_GROUP])
1701 		dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1702 
1703 	return dev;
1704 
1705 err:
1706 	return ERR_PTR(err);
1707 }
1708 EXPORT_SYMBOL(rtnl_create_link);
1709 
1710 static int rtnl_group_changelink(struct net *net, int group,
1711 		struct ifinfomsg *ifm,
1712 		struct nlattr **tb)
1713 {
1714 	struct net_device *dev;
1715 	int err;
1716 
1717 	for_each_netdev(net, dev) {
1718 		if (dev->group == group) {
1719 			err = do_setlink(dev, ifm, tb, NULL, 0);
1720 			if (err < 0)
1721 				return err;
1722 		}
1723 	}
1724 
1725 	return 0;
1726 }
1727 
1728 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1729 {
1730 	struct net *net = sock_net(skb->sk);
1731 	const struct rtnl_link_ops *ops;
1732 	struct net_device *dev;
1733 	struct ifinfomsg *ifm;
1734 	char kind[MODULE_NAME_LEN];
1735 	char ifname[IFNAMSIZ];
1736 	struct nlattr *tb[IFLA_MAX+1];
1737 	struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1738 	int err;
1739 
1740 #ifdef CONFIG_MODULES
1741 replay:
1742 #endif
1743 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1744 	if (err < 0)
1745 		return err;
1746 
1747 	if (tb[IFLA_IFNAME])
1748 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1749 	else
1750 		ifname[0] = '\0';
1751 
1752 	ifm = nlmsg_data(nlh);
1753 	if (ifm->ifi_index > 0)
1754 		dev = __dev_get_by_index(net, ifm->ifi_index);
1755 	else {
1756 		if (ifname[0])
1757 			dev = __dev_get_by_name(net, ifname);
1758 		else
1759 			dev = NULL;
1760 	}
1761 
1762 	err = validate_linkmsg(dev, tb);
1763 	if (err < 0)
1764 		return err;
1765 
1766 	if (tb[IFLA_LINKINFO]) {
1767 		err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1768 				       tb[IFLA_LINKINFO], ifla_info_policy);
1769 		if (err < 0)
1770 			return err;
1771 	} else
1772 		memset(linkinfo, 0, sizeof(linkinfo));
1773 
1774 	if (linkinfo[IFLA_INFO_KIND]) {
1775 		nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1776 		ops = rtnl_link_ops_get(kind);
1777 	} else {
1778 		kind[0] = '\0';
1779 		ops = NULL;
1780 	}
1781 
1782 	if (1) {
1783 		struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1784 		struct net *dest_net;
1785 
1786 		if (ops) {
1787 			if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1788 				err = nla_parse_nested(attr, ops->maxtype,
1789 						       linkinfo[IFLA_INFO_DATA],
1790 						       ops->policy);
1791 				if (err < 0)
1792 					return err;
1793 				data = attr;
1794 			}
1795 			if (ops->validate) {
1796 				err = ops->validate(tb, data);
1797 				if (err < 0)
1798 					return err;
1799 			}
1800 		}
1801 
1802 		if (dev) {
1803 			int modified = 0;
1804 
1805 			if (nlh->nlmsg_flags & NLM_F_EXCL)
1806 				return -EEXIST;
1807 			if (nlh->nlmsg_flags & NLM_F_REPLACE)
1808 				return -EOPNOTSUPP;
1809 
1810 			if (linkinfo[IFLA_INFO_DATA]) {
1811 				if (!ops || ops != dev->rtnl_link_ops ||
1812 				    !ops->changelink)
1813 					return -EOPNOTSUPP;
1814 
1815 				err = ops->changelink(dev, tb, data);
1816 				if (err < 0)
1817 					return err;
1818 				modified = 1;
1819 			}
1820 
1821 			return do_setlink(dev, ifm, tb, ifname, modified);
1822 		}
1823 
1824 		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1825 			if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1826 				return rtnl_group_changelink(net,
1827 						nla_get_u32(tb[IFLA_GROUP]),
1828 						ifm, tb);
1829 			return -ENODEV;
1830 		}
1831 
1832 		if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1833 			return -EOPNOTSUPP;
1834 
1835 		if (!ops) {
1836 #ifdef CONFIG_MODULES
1837 			if (kind[0]) {
1838 				__rtnl_unlock();
1839 				request_module("rtnl-link-%s", kind);
1840 				rtnl_lock();
1841 				ops = rtnl_link_ops_get(kind);
1842 				if (ops)
1843 					goto replay;
1844 			}
1845 #endif
1846 			return -EOPNOTSUPP;
1847 		}
1848 
1849 		if (!ifname[0])
1850 			snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1851 
1852 		dest_net = rtnl_link_get_net(net, tb);
1853 		if (IS_ERR(dest_net))
1854 			return PTR_ERR(dest_net);
1855 
1856 		dev = rtnl_create_link(dest_net, ifname, ops, tb);
1857 		if (IS_ERR(dev)) {
1858 			err = PTR_ERR(dev);
1859 			goto out;
1860 		}
1861 
1862 		dev->ifindex = ifm->ifi_index;
1863 
1864 		if (ops->newlink)
1865 			err = ops->newlink(net, dev, tb, data);
1866 		else
1867 			err = register_netdevice(dev);
1868 
1869 		if (err < 0) {
1870 			free_netdev(dev);
1871 			goto out;
1872 		}
1873 
1874 		err = rtnl_configure_link(dev, ifm);
1875 		if (err < 0)
1876 			unregister_netdevice(dev);
1877 out:
1878 		put_net(dest_net);
1879 		return err;
1880 	}
1881 }
1882 
1883 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1884 {
1885 	struct net *net = sock_net(skb->sk);
1886 	struct ifinfomsg *ifm;
1887 	char ifname[IFNAMSIZ];
1888 	struct nlattr *tb[IFLA_MAX+1];
1889 	struct net_device *dev = NULL;
1890 	struct sk_buff *nskb;
1891 	int err;
1892 	u32 ext_filter_mask = 0;
1893 
1894 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1895 	if (err < 0)
1896 		return err;
1897 
1898 	if (tb[IFLA_IFNAME])
1899 		nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1900 
1901 	if (tb[IFLA_EXT_MASK])
1902 		ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1903 
1904 	ifm = nlmsg_data(nlh);
1905 	if (ifm->ifi_index > 0)
1906 		dev = __dev_get_by_index(net, ifm->ifi_index);
1907 	else if (tb[IFLA_IFNAME])
1908 		dev = __dev_get_by_name(net, ifname);
1909 	else
1910 		return -EINVAL;
1911 
1912 	if (dev == NULL)
1913 		return -ENODEV;
1914 
1915 	nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1916 	if (nskb == NULL)
1917 		return -ENOBUFS;
1918 
1919 	err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1920 			       nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1921 	if (err < 0) {
1922 		/* -EMSGSIZE implies BUG in if_nlmsg_size */
1923 		WARN_ON(err == -EMSGSIZE);
1924 		kfree_skb(nskb);
1925 	} else
1926 		err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1927 
1928 	return err;
1929 }
1930 
1931 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1932 {
1933 	struct net *net = sock_net(skb->sk);
1934 	struct net_device *dev;
1935 	struct nlattr *tb[IFLA_MAX+1];
1936 	u32 ext_filter_mask = 0;
1937 	u16 min_ifinfo_dump_size = 0;
1938 
1939 	if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1940 			ifla_policy) >= 0) {
1941 		if (tb[IFLA_EXT_MASK])
1942 			ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1943 	}
1944 
1945 	if (!ext_filter_mask)
1946 		return NLMSG_GOODSIZE;
1947 	/*
1948 	 * traverse the list of net devices and compute the minimum
1949 	 * buffer size based upon the filter mask.
1950 	 */
1951 	list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1952 		min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1953 					     if_nlmsg_size(dev,
1954 						           ext_filter_mask));
1955 	}
1956 
1957 	return min_ifinfo_dump_size;
1958 }
1959 
1960 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1961 {
1962 	int idx;
1963 	int s_idx = cb->family;
1964 
1965 	if (s_idx == 0)
1966 		s_idx = 1;
1967 	for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1968 		int type = cb->nlh->nlmsg_type-RTM_BASE;
1969 		if (idx < s_idx || idx == PF_PACKET)
1970 			continue;
1971 		if (rtnl_msg_handlers[idx] == NULL ||
1972 		    rtnl_msg_handlers[idx][type].dumpit == NULL)
1973 			continue;
1974 		if (idx > s_idx) {
1975 			memset(&cb->args[0], 0, sizeof(cb->args));
1976 			cb->prev_seq = 0;
1977 			cb->seq = 0;
1978 		}
1979 		if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1980 			break;
1981 	}
1982 	cb->family = idx;
1983 
1984 	return skb->len;
1985 }
1986 
1987 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
1988 		  gfp_t flags)
1989 {
1990 	struct net *net = dev_net(dev);
1991 	struct sk_buff *skb;
1992 	int err = -ENOBUFS;
1993 	size_t if_info_size;
1994 
1995 	skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
1996 	if (skb == NULL)
1997 		goto errout;
1998 
1999 	err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2000 	if (err < 0) {
2001 		/* -EMSGSIZE implies BUG in if_nlmsg_size() */
2002 		WARN_ON(err == -EMSGSIZE);
2003 		kfree_skb(skb);
2004 		goto errout;
2005 	}
2006 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2007 	return;
2008 errout:
2009 	if (err < 0)
2010 		rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2011 }
2012 EXPORT_SYMBOL(rtmsg_ifinfo);
2013 
2014 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2015 				   struct net_device *dev,
2016 				   u8 *addr, u32 pid, u32 seq,
2017 				   int type, unsigned int flags)
2018 {
2019 	struct nlmsghdr *nlh;
2020 	struct ndmsg *ndm;
2021 
2022 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
2023 	if (!nlh)
2024 		return -EMSGSIZE;
2025 
2026 	ndm = nlmsg_data(nlh);
2027 	ndm->ndm_family  = AF_BRIDGE;
2028 	ndm->ndm_pad1	 = 0;
2029 	ndm->ndm_pad2    = 0;
2030 	ndm->ndm_flags	 = flags;
2031 	ndm->ndm_type	 = 0;
2032 	ndm->ndm_ifindex = dev->ifindex;
2033 	ndm->ndm_state   = NUD_PERMANENT;
2034 
2035 	if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2036 		goto nla_put_failure;
2037 
2038 	return nlmsg_end(skb, nlh);
2039 
2040 nla_put_failure:
2041 	nlmsg_cancel(skb, nlh);
2042 	return -EMSGSIZE;
2043 }
2044 
2045 static inline size_t rtnl_fdb_nlmsg_size(void)
2046 {
2047 	return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2048 }
2049 
2050 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2051 {
2052 	struct net *net = dev_net(dev);
2053 	struct sk_buff *skb;
2054 	int err = -ENOBUFS;
2055 
2056 	skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2057 	if (!skb)
2058 		goto errout;
2059 
2060 	err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2061 	if (err < 0) {
2062 		kfree_skb(skb);
2063 		goto errout;
2064 	}
2065 
2066 	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2067 	return;
2068 errout:
2069 	rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2070 }
2071 
2072 /**
2073  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2074  */
2075 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2076 		     struct nlattr *tb[],
2077 		     struct net_device *dev,
2078 		     const unsigned char *addr,
2079 		     u16 flags)
2080 {
2081 	int err = -EINVAL;
2082 
2083 	/* If aging addresses are supported device will need to
2084 	 * implement its own handler for this.
2085 	 */
2086 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2087 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2088 		return err;
2089 	}
2090 
2091 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2092 		err = dev_uc_add_excl(dev, addr);
2093 	else if (is_multicast_ether_addr(addr))
2094 		err = dev_mc_add_excl(dev, addr);
2095 
2096 	/* Only return duplicate errors if NLM_F_EXCL is set */
2097 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
2098 		err = 0;
2099 
2100 	return err;
2101 }
2102 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2103 
2104 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2105 {
2106 	struct net *net = sock_net(skb->sk);
2107 	struct ndmsg *ndm;
2108 	struct nlattr *tb[NDA_MAX+1];
2109 	struct net_device *dev;
2110 	u8 *addr;
2111 	int err;
2112 
2113 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2114 	if (err < 0)
2115 		return err;
2116 
2117 	ndm = nlmsg_data(nlh);
2118 	if (ndm->ndm_ifindex == 0) {
2119 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2120 		return -EINVAL;
2121 	}
2122 
2123 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2124 	if (dev == NULL) {
2125 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2126 		return -ENODEV;
2127 	}
2128 
2129 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2130 		pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2131 		return -EINVAL;
2132 	}
2133 
2134 	addr = nla_data(tb[NDA_LLADDR]);
2135 
2136 	err = -EOPNOTSUPP;
2137 
2138 	/* Support fdb on master device the net/bridge default case */
2139 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2140 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2141 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2142 		const struct net_device_ops *ops = br_dev->netdev_ops;
2143 
2144 		err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2145 		if (err)
2146 			goto out;
2147 		else
2148 			ndm->ndm_flags &= ~NTF_MASTER;
2149 	}
2150 
2151 	/* Embedded bridge, macvlan, and any other device support */
2152 	if ((ndm->ndm_flags & NTF_SELF)) {
2153 		if (dev->netdev_ops->ndo_fdb_add)
2154 			err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2155 							   nlh->nlmsg_flags);
2156 		else
2157 			err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2158 					       nlh->nlmsg_flags);
2159 
2160 		if (!err) {
2161 			rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2162 			ndm->ndm_flags &= ~NTF_SELF;
2163 		}
2164 	}
2165 out:
2166 	return err;
2167 }
2168 
2169 /**
2170  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2171  */
2172 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2173 		     struct nlattr *tb[],
2174 		     struct net_device *dev,
2175 		     const unsigned char *addr)
2176 {
2177 	int err = -EOPNOTSUPP;
2178 
2179 	/* If aging addresses are supported device will need to
2180 	 * implement its own handler for this.
2181 	 */
2182 	if (!(ndm->ndm_state & NUD_PERMANENT)) {
2183 		pr_info("%s: FDB only supports static addresses\n", dev->name);
2184 		return -EINVAL;
2185 	}
2186 
2187 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2188 		err = dev_uc_del(dev, addr);
2189 	else if (is_multicast_ether_addr(addr))
2190 		err = dev_mc_del(dev, addr);
2191 	else
2192 		err = -EINVAL;
2193 
2194 	return err;
2195 }
2196 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2197 
2198 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2199 {
2200 	struct net *net = sock_net(skb->sk);
2201 	struct ndmsg *ndm;
2202 	struct nlattr *tb[NDA_MAX+1];
2203 	struct net_device *dev;
2204 	int err = -EINVAL;
2205 	__u8 *addr;
2206 
2207 	if (!capable(CAP_NET_ADMIN))
2208 		return -EPERM;
2209 
2210 	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2211 	if (err < 0)
2212 		return err;
2213 
2214 	ndm = nlmsg_data(nlh);
2215 	if (ndm->ndm_ifindex == 0) {
2216 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2217 		return -EINVAL;
2218 	}
2219 
2220 	dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2221 	if (dev == NULL) {
2222 		pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2223 		return -ENODEV;
2224 	}
2225 
2226 	if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2227 		pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2228 		return -EINVAL;
2229 	}
2230 
2231 	addr = nla_data(tb[NDA_LLADDR]);
2232 
2233 	err = -EOPNOTSUPP;
2234 
2235 	/* Support fdb on master device the net/bridge default case */
2236 	if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2237 	    (dev->priv_flags & IFF_BRIDGE_PORT)) {
2238 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2239 		const struct net_device_ops *ops = br_dev->netdev_ops;
2240 
2241 		if (ops->ndo_fdb_del)
2242 			err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2243 
2244 		if (err)
2245 			goto out;
2246 		else
2247 			ndm->ndm_flags &= ~NTF_MASTER;
2248 	}
2249 
2250 	/* Embedded bridge, macvlan, and any other device support */
2251 	if (ndm->ndm_flags & NTF_SELF) {
2252 		if (dev->netdev_ops->ndo_fdb_del)
2253 			err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2254 		else
2255 			err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2256 
2257 		if (!err) {
2258 			rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2259 			ndm->ndm_flags &= ~NTF_SELF;
2260 		}
2261 	}
2262 out:
2263 	return err;
2264 }
2265 
2266 static int nlmsg_populate_fdb(struct sk_buff *skb,
2267 			      struct netlink_callback *cb,
2268 			      struct net_device *dev,
2269 			      int *idx,
2270 			      struct netdev_hw_addr_list *list)
2271 {
2272 	struct netdev_hw_addr *ha;
2273 	int err;
2274 	u32 portid, seq;
2275 
2276 	portid = NETLINK_CB(cb->skb).portid;
2277 	seq = cb->nlh->nlmsg_seq;
2278 
2279 	list_for_each_entry(ha, &list->list, list) {
2280 		if (*idx < cb->args[0])
2281 			goto skip;
2282 
2283 		err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2284 					      portid, seq,
2285 					      RTM_NEWNEIGH, NTF_SELF);
2286 		if (err < 0)
2287 			return err;
2288 skip:
2289 		*idx += 1;
2290 	}
2291 	return 0;
2292 }
2293 
2294 /**
2295  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2296  * @nlh: netlink message header
2297  * @dev: netdevice
2298  *
2299  * Default netdevice operation to dump the existing unicast address list.
2300  * Returns number of addresses from list put in skb.
2301  */
2302 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2303 		      struct netlink_callback *cb,
2304 		      struct net_device *dev,
2305 		      int idx)
2306 {
2307 	int err;
2308 
2309 	netif_addr_lock_bh(dev);
2310 	err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2311 	if (err)
2312 		goto out;
2313 	nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2314 out:
2315 	netif_addr_unlock_bh(dev);
2316 	return idx;
2317 }
2318 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2319 
2320 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2321 {
2322 	int idx = 0;
2323 	struct net *net = sock_net(skb->sk);
2324 	struct net_device *dev;
2325 
2326 	rcu_read_lock();
2327 	for_each_netdev_rcu(net, dev) {
2328 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
2329 			struct net_device *br_dev;
2330 			const struct net_device_ops *ops;
2331 
2332 			br_dev = netdev_master_upper_dev_get(dev);
2333 			ops = br_dev->netdev_ops;
2334 			if (ops->ndo_fdb_dump)
2335 				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2336 		}
2337 
2338 		if (dev->netdev_ops->ndo_fdb_dump)
2339 			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2340 		else
2341 			idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2342 	}
2343 	rcu_read_unlock();
2344 
2345 	cb->args[0] = idx;
2346 	return skb->len;
2347 }
2348 
2349 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2350 			    struct net_device *dev, u16 mode)
2351 {
2352 	struct nlmsghdr *nlh;
2353 	struct ifinfomsg *ifm;
2354 	struct nlattr *br_afspec;
2355 	u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2356 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2357 
2358 	nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2359 	if (nlh == NULL)
2360 		return -EMSGSIZE;
2361 
2362 	ifm = nlmsg_data(nlh);
2363 	ifm->ifi_family = AF_BRIDGE;
2364 	ifm->__ifi_pad = 0;
2365 	ifm->ifi_type = dev->type;
2366 	ifm->ifi_index = dev->ifindex;
2367 	ifm->ifi_flags = dev_get_flags(dev);
2368 	ifm->ifi_change = 0;
2369 
2370 
2371 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2372 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2373 	    nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2374 	    (br_dev &&
2375 	     nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2376 	    (dev->addr_len &&
2377 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2378 	    (dev->ifindex != dev->iflink &&
2379 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2380 		goto nla_put_failure;
2381 
2382 	br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2383 	if (!br_afspec)
2384 		goto nla_put_failure;
2385 
2386 	if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2387 	    nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2388 		nla_nest_cancel(skb, br_afspec);
2389 		goto nla_put_failure;
2390 	}
2391 	nla_nest_end(skb, br_afspec);
2392 
2393 	return nlmsg_end(skb, nlh);
2394 nla_put_failure:
2395 	nlmsg_cancel(skb, nlh);
2396 	return -EMSGSIZE;
2397 }
2398 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2399 
2400 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2401 {
2402 	struct net *net = sock_net(skb->sk);
2403 	struct net_device *dev;
2404 	int idx = 0;
2405 	u32 portid = NETLINK_CB(cb->skb).portid;
2406 	u32 seq = cb->nlh->nlmsg_seq;
2407 	struct nlattr *extfilt;
2408 	u32 filter_mask = 0;
2409 
2410 	extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2411 				  IFLA_EXT_MASK);
2412 	if (extfilt)
2413 		filter_mask = nla_get_u32(extfilt);
2414 
2415 	rcu_read_lock();
2416 	for_each_netdev_rcu(net, dev) {
2417 		const struct net_device_ops *ops = dev->netdev_ops;
2418 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2419 
2420 		if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2421 			if (idx >= cb->args[0] &&
2422 			    br_dev->netdev_ops->ndo_bridge_getlink(
2423 				    skb, portid, seq, dev, filter_mask) < 0)
2424 				break;
2425 			idx++;
2426 		}
2427 
2428 		if (ops->ndo_bridge_getlink) {
2429 			if (idx >= cb->args[0] &&
2430 			    ops->ndo_bridge_getlink(skb, portid, seq, dev,
2431 						    filter_mask) < 0)
2432 				break;
2433 			idx++;
2434 		}
2435 	}
2436 	rcu_read_unlock();
2437 	cb->args[0] = idx;
2438 
2439 	return skb->len;
2440 }
2441 
2442 static inline size_t bridge_nlmsg_size(void)
2443 {
2444 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2445 		+ nla_total_size(IFNAMSIZ)	/* IFLA_IFNAME */
2446 		+ nla_total_size(MAX_ADDR_LEN)	/* IFLA_ADDRESS */
2447 		+ nla_total_size(sizeof(u32))	/* IFLA_MASTER */
2448 		+ nla_total_size(sizeof(u32))	/* IFLA_MTU */
2449 		+ nla_total_size(sizeof(u32))	/* IFLA_LINK */
2450 		+ nla_total_size(sizeof(u32))	/* IFLA_OPERSTATE */
2451 		+ nla_total_size(sizeof(u8))	/* IFLA_PROTINFO */
2452 		+ nla_total_size(sizeof(struct nlattr))	/* IFLA_AF_SPEC */
2453 		+ nla_total_size(sizeof(u16))	/* IFLA_BRIDGE_FLAGS */
2454 		+ nla_total_size(sizeof(u16));	/* IFLA_BRIDGE_MODE */
2455 }
2456 
2457 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2458 {
2459 	struct net *net = dev_net(dev);
2460 	struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2461 	struct sk_buff *skb;
2462 	int err = -EOPNOTSUPP;
2463 
2464 	skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2465 	if (!skb) {
2466 		err = -ENOMEM;
2467 		goto errout;
2468 	}
2469 
2470 	if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2471 	    br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2472 		err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2473 		if (err < 0)
2474 			goto errout;
2475 	}
2476 
2477 	if ((flags & BRIDGE_FLAGS_SELF) &&
2478 	    dev->netdev_ops->ndo_bridge_getlink) {
2479 		err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2480 		if (err < 0)
2481 			goto errout;
2482 	}
2483 
2484 	rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2485 	return 0;
2486 errout:
2487 	WARN_ON(err == -EMSGSIZE);
2488 	kfree_skb(skb);
2489 	rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2490 	return err;
2491 }
2492 
2493 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2494 {
2495 	struct net *net = sock_net(skb->sk);
2496 	struct ifinfomsg *ifm;
2497 	struct net_device *dev;
2498 	struct nlattr *br_spec, *attr = NULL;
2499 	int rem, err = -EOPNOTSUPP;
2500 	u16 oflags, flags = 0;
2501 	bool have_flags = false;
2502 
2503 	if (nlmsg_len(nlh) < sizeof(*ifm))
2504 		return -EINVAL;
2505 
2506 	ifm = nlmsg_data(nlh);
2507 	if (ifm->ifi_family != AF_BRIDGE)
2508 		return -EPFNOSUPPORT;
2509 
2510 	dev = __dev_get_by_index(net, ifm->ifi_index);
2511 	if (!dev) {
2512 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2513 		return -ENODEV;
2514 	}
2515 
2516 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2517 	if (br_spec) {
2518 		nla_for_each_nested(attr, br_spec, rem) {
2519 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2520 				have_flags = true;
2521 				flags = nla_get_u16(attr);
2522 				break;
2523 			}
2524 		}
2525 	}
2526 
2527 	oflags = flags;
2528 
2529 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2530 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2531 
2532 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2533 			err = -EOPNOTSUPP;
2534 			goto out;
2535 		}
2536 
2537 		err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2538 		if (err)
2539 			goto out;
2540 
2541 		flags &= ~BRIDGE_FLAGS_MASTER;
2542 	}
2543 
2544 	if ((flags & BRIDGE_FLAGS_SELF)) {
2545 		if (!dev->netdev_ops->ndo_bridge_setlink)
2546 			err = -EOPNOTSUPP;
2547 		else
2548 			err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2549 
2550 		if (!err)
2551 			flags &= ~BRIDGE_FLAGS_SELF;
2552 	}
2553 
2554 	if (have_flags)
2555 		memcpy(nla_data(attr), &flags, sizeof(flags));
2556 	/* Generate event to notify upper layer of bridge change */
2557 	if (!err)
2558 		err = rtnl_bridge_notify(dev, oflags);
2559 out:
2560 	return err;
2561 }
2562 
2563 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2564 {
2565 	struct net *net = sock_net(skb->sk);
2566 	struct ifinfomsg *ifm;
2567 	struct net_device *dev;
2568 	struct nlattr *br_spec, *attr = NULL;
2569 	int rem, err = -EOPNOTSUPP;
2570 	u16 oflags, flags = 0;
2571 	bool have_flags = false;
2572 
2573 	if (nlmsg_len(nlh) < sizeof(*ifm))
2574 		return -EINVAL;
2575 
2576 	ifm = nlmsg_data(nlh);
2577 	if (ifm->ifi_family != AF_BRIDGE)
2578 		return -EPFNOSUPPORT;
2579 
2580 	dev = __dev_get_by_index(net, ifm->ifi_index);
2581 	if (!dev) {
2582 		pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2583 		return -ENODEV;
2584 	}
2585 
2586 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2587 	if (br_spec) {
2588 		nla_for_each_nested(attr, br_spec, rem) {
2589 			if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2590 				have_flags = true;
2591 				flags = nla_get_u16(attr);
2592 				break;
2593 			}
2594 		}
2595 	}
2596 
2597 	oflags = flags;
2598 
2599 	if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2600 		struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2601 
2602 		if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2603 			err = -EOPNOTSUPP;
2604 			goto out;
2605 		}
2606 
2607 		err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2608 		if (err)
2609 			goto out;
2610 
2611 		flags &= ~BRIDGE_FLAGS_MASTER;
2612 	}
2613 
2614 	if ((flags & BRIDGE_FLAGS_SELF)) {
2615 		if (!dev->netdev_ops->ndo_bridge_dellink)
2616 			err = -EOPNOTSUPP;
2617 		else
2618 			err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2619 
2620 		if (!err)
2621 			flags &= ~BRIDGE_FLAGS_SELF;
2622 	}
2623 
2624 	if (have_flags)
2625 		memcpy(nla_data(attr), &flags, sizeof(flags));
2626 	/* Generate event to notify upper layer of bridge change */
2627 	if (!err)
2628 		err = rtnl_bridge_notify(dev, oflags);
2629 out:
2630 	return err;
2631 }
2632 
2633 /* Process one rtnetlink message. */
2634 
2635 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2636 {
2637 	struct net *net = sock_net(skb->sk);
2638 	rtnl_doit_func doit;
2639 	int sz_idx, kind;
2640 	int family;
2641 	int type;
2642 	int err;
2643 
2644 	type = nlh->nlmsg_type;
2645 	if (type > RTM_MAX)
2646 		return -EOPNOTSUPP;
2647 
2648 	type -= RTM_BASE;
2649 
2650 	/* All the messages must have at least 1 byte length */
2651 	if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2652 		return 0;
2653 
2654 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2655 	sz_idx = type>>2;
2656 	kind = type&3;
2657 
2658 	if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2659 		return -EPERM;
2660 
2661 	if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2662 		struct sock *rtnl;
2663 		rtnl_dumpit_func dumpit;
2664 		rtnl_calcit_func calcit;
2665 		u16 min_dump_alloc = 0;
2666 
2667 		dumpit = rtnl_get_dumpit(family, type);
2668 		if (dumpit == NULL)
2669 			return -EOPNOTSUPP;
2670 		calcit = rtnl_get_calcit(family, type);
2671 		if (calcit)
2672 			min_dump_alloc = calcit(skb, nlh);
2673 
2674 		__rtnl_unlock();
2675 		rtnl = net->rtnl;
2676 		{
2677 			struct netlink_dump_control c = {
2678 				.dump		= dumpit,
2679 				.min_dump_alloc	= min_dump_alloc,
2680 			};
2681 			err = netlink_dump_start(rtnl, skb, nlh, &c);
2682 		}
2683 		rtnl_lock();
2684 		return err;
2685 	}
2686 
2687 	doit = rtnl_get_doit(family, type);
2688 	if (doit == NULL)
2689 		return -EOPNOTSUPP;
2690 
2691 	return doit(skb, nlh);
2692 }
2693 
2694 static void rtnetlink_rcv(struct sk_buff *skb)
2695 {
2696 	rtnl_lock();
2697 	netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2698 	rtnl_unlock();
2699 }
2700 
2701 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2702 {
2703 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2704 
2705 	switch (event) {
2706 	case NETDEV_UP:
2707 	case NETDEV_DOWN:
2708 	case NETDEV_PRE_UP:
2709 	case NETDEV_POST_INIT:
2710 	case NETDEV_REGISTER:
2711 	case NETDEV_CHANGE:
2712 	case NETDEV_PRE_TYPE_CHANGE:
2713 	case NETDEV_GOING_DOWN:
2714 	case NETDEV_UNREGISTER:
2715 	case NETDEV_UNREGISTER_FINAL:
2716 	case NETDEV_RELEASE:
2717 	case NETDEV_JOIN:
2718 		break;
2719 	default:
2720 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2721 		break;
2722 	}
2723 	return NOTIFY_DONE;
2724 }
2725 
2726 static struct notifier_block rtnetlink_dev_notifier = {
2727 	.notifier_call	= rtnetlink_event,
2728 };
2729 
2730 
2731 static int __net_init rtnetlink_net_init(struct net *net)
2732 {
2733 	struct sock *sk;
2734 	struct netlink_kernel_cfg cfg = {
2735 		.groups		= RTNLGRP_MAX,
2736 		.input		= rtnetlink_rcv,
2737 		.cb_mutex	= &rtnl_mutex,
2738 		.flags		= NL_CFG_F_NONROOT_RECV,
2739 	};
2740 
2741 	sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2742 	if (!sk)
2743 		return -ENOMEM;
2744 	net->rtnl = sk;
2745 	return 0;
2746 }
2747 
2748 static void __net_exit rtnetlink_net_exit(struct net *net)
2749 {
2750 	netlink_kernel_release(net->rtnl);
2751 	net->rtnl = NULL;
2752 }
2753 
2754 static struct pernet_operations rtnetlink_net_ops = {
2755 	.init = rtnetlink_net_init,
2756 	.exit = rtnetlink_net_exit,
2757 };
2758 
2759 void __init rtnetlink_init(void)
2760 {
2761 	if (register_pernet_subsys(&rtnetlink_net_ops))
2762 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
2763 
2764 	register_netdevice_notifier(&rtnetlink_dev_notifier);
2765 
2766 	rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2767 		      rtnl_dump_ifinfo, rtnl_calcit);
2768 	rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2769 	rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2770 	rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2771 
2772 	rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2773 	rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2774 
2775 	rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2776 	rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2777 	rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2778 
2779 	rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2780 	rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2781 	rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2782 }
2783 
2784