aio: don't include aio.h in sched.h
[linux-3.10.git] / net / core / rtnetlink.c
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
951                         /*
952                          * Not all SR-IOV capable drivers support the
953                          * spoofcheck query.  Preset to -1 so the user
954                          * space tool can detect that the driver didn't
955                          * report anything.
956                          */
957                         ivi.spoofchk = -1;
958                         memset(ivi.mac, 0, sizeof(ivi.mac));
959                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
960                                 break;
961                         vf_mac.vf =
962                                 vf_vlan.vf =
963                                 vf_tx_rate.vf =
964                                 vf_spoofchk.vf = ivi.vf;
965
966                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
967                         vf_vlan.vlan = ivi.vlan;
968                         vf_vlan.qos = ivi.qos;
969                         vf_tx_rate.rate = ivi.tx_rate;
970                         vf_spoofchk.setting = ivi.spoofchk;
971                         vf = nla_nest_start(skb, IFLA_VF_INFO);
972                         if (!vf) {
973                                 nla_nest_cancel(skb, vfinfo);
974                                 goto nla_put_failure;
975                         }
976                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
977                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
978                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
979                                     &vf_tx_rate) ||
980                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
981                                     &vf_spoofchk))
982                                 goto nla_put_failure;
983                         nla_nest_end(skb, vf);
984                 }
985                 nla_nest_end(skb, vfinfo);
986         }
987
988         if (rtnl_port_fill(skb, dev))
989                 goto nla_put_failure;
990
991         if (dev->rtnl_link_ops) {
992                 if (rtnl_link_fill(skb, dev) < 0)
993                         goto nla_put_failure;
994         }
995
996         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
997                 goto nla_put_failure;
998
999         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1000                 if (af_ops->fill_link_af) {
1001                         struct nlattr *af;
1002                         int err;
1003
1004                         if (!(af = nla_nest_start(skb, af_ops->family)))
1005                                 goto nla_put_failure;
1006
1007                         err = af_ops->fill_link_af(skb, dev);
1008
1009                         /*
1010                          * Caller may return ENODATA to indicate that there
1011                          * was no data to be dumped. This is not an error, it
1012                          * means we should trim the attribute header and
1013                          * continue.
1014                          */
1015                         if (err == -ENODATA)
1016                                 nla_nest_cancel(skb, af);
1017                         else if (err < 0)
1018                                 goto nla_put_failure;
1019
1020                         nla_nest_end(skb, af);
1021                 }
1022         }
1023
1024         nla_nest_end(skb, af_spec);
1025
1026         return nlmsg_end(skb, nlh);
1027
1028 nla_put_failure:
1029         nlmsg_cancel(skb, nlh);
1030         return -EMSGSIZE;
1031 }
1032
1033 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1034 {
1035         struct net *net = sock_net(skb->sk);
1036         int h, s_h;
1037         int idx = 0, s_idx;
1038         struct net_device *dev;
1039         struct hlist_head *head;
1040         struct nlattr *tb[IFLA_MAX+1];
1041         u32 ext_filter_mask = 0;
1042
1043         s_h = cb->args[0];
1044         s_idx = cb->args[1];
1045
1046         rcu_read_lock();
1047         cb->seq = net->dev_base_seq;
1048
1049         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1050                         ifla_policy) >= 0) {
1051
1052                 if (tb[IFLA_EXT_MASK])
1053                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1054         }
1055
1056         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1057                 idx = 0;
1058                 head = &net->dev_index_head[h];
1059                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1060                         if (idx < s_idx)
1061                                 goto cont;
1062                         if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1063                                              NETLINK_CB(cb->skb).portid,
1064                                              cb->nlh->nlmsg_seq, 0,
1065                                              NLM_F_MULTI,
1066                                              ext_filter_mask) <= 0)
1067                                 goto out;
1068
1069                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1070 cont:
1071                         idx++;
1072                 }
1073         }
1074 out:
1075         rcu_read_unlock();
1076         cb->args[1] = idx;
1077         cb->args[0] = h;
1078
1079         return skb->len;
1080 }
1081
1082 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1083         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1084         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1085         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1086         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1087         [IFLA_MTU]              = { .type = NLA_U32 },
1088         [IFLA_LINK]             = { .type = NLA_U32 },
1089         [IFLA_MASTER]           = { .type = NLA_U32 },
1090         [IFLA_CARRIER]          = { .type = NLA_U8 },
1091         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1092         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1093         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1094         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1095         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1096         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1097         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1098         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1099         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1100         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1101         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1102         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1103         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1104         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1105         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1106         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1107 };
1108 EXPORT_SYMBOL(ifla_policy);
1109
1110 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1111         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1112         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1113 };
1114
1115 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1116         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1117 };
1118
1119 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1120         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1121                                     .len = sizeof(struct ifla_vf_mac) },
1122         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1123                                     .len = sizeof(struct ifla_vf_vlan) },
1124         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1125                                     .len = sizeof(struct ifla_vf_tx_rate) },
1126         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1127                                     .len = sizeof(struct ifla_vf_spoofchk) },
1128 };
1129
1130 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1131         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1132         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1133                                     .len = PORT_PROFILE_MAX },
1134         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1135                                     .len = sizeof(struct ifla_port_vsi)},
1136         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1137                                       .len = PORT_UUID_MAX },
1138         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1139                                     .len = PORT_UUID_MAX },
1140         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1141         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1142 };
1143
1144 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1145 {
1146         struct net *net;
1147         /* Examine the link attributes and figure out which
1148          * network namespace we are talking about.
1149          */
1150         if (tb[IFLA_NET_NS_PID])
1151                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1152         else if (tb[IFLA_NET_NS_FD])
1153                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1154         else
1155                 net = get_net(src_net);
1156         return net;
1157 }
1158 EXPORT_SYMBOL(rtnl_link_get_net);
1159
1160 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1161 {
1162         if (dev) {
1163                 if (tb[IFLA_ADDRESS] &&
1164                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1165                         return -EINVAL;
1166
1167                 if (tb[IFLA_BROADCAST] &&
1168                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1169                         return -EINVAL;
1170         }
1171
1172         if (tb[IFLA_AF_SPEC]) {
1173                 struct nlattr *af;
1174                 int rem, err;
1175
1176                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1177                         const struct rtnl_af_ops *af_ops;
1178
1179                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1180                                 return -EAFNOSUPPORT;
1181
1182                         if (!af_ops->set_link_af)
1183                                 return -EOPNOTSUPP;
1184
1185                         if (af_ops->validate_link_af) {
1186                                 err = af_ops->validate_link_af(dev, af);
1187                                 if (err < 0)
1188                                         return err;
1189                         }
1190                 }
1191         }
1192
1193         return 0;
1194 }
1195
1196 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1197 {
1198         int rem, err = -EINVAL;
1199         struct nlattr *vf;
1200         const struct net_device_ops *ops = dev->netdev_ops;
1201
1202         nla_for_each_nested(vf, attr, rem) {
1203                 switch (nla_type(vf)) {
1204                 case IFLA_VF_MAC: {
1205                         struct ifla_vf_mac *ivm;
1206                         ivm = nla_data(vf);
1207                         err = -EOPNOTSUPP;
1208                         if (ops->ndo_set_vf_mac)
1209                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1210                                                           ivm->mac);
1211                         break;
1212                 }
1213                 case IFLA_VF_VLAN: {
1214                         struct ifla_vf_vlan *ivv;
1215                         ivv = nla_data(vf);
1216                         err = -EOPNOTSUPP;
1217                         if (ops->ndo_set_vf_vlan)
1218                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1219                                                            ivv->vlan,
1220                                                            ivv->qos);
1221                         break;
1222                 }
1223                 case IFLA_VF_TX_RATE: {
1224                         struct ifla_vf_tx_rate *ivt;
1225                         ivt = nla_data(vf);
1226                         err = -EOPNOTSUPP;
1227                         if (ops->ndo_set_vf_tx_rate)
1228                                 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1229                                                               ivt->rate);
1230                         break;
1231                 }
1232                 case IFLA_VF_SPOOFCHK: {
1233                         struct ifla_vf_spoofchk *ivs;
1234                         ivs = nla_data(vf);
1235                         err = -EOPNOTSUPP;
1236                         if (ops->ndo_set_vf_spoofchk)
1237                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1238                                                                ivs->setting);
1239                         break;
1240                 }
1241                 default:
1242                         err = -EINVAL;
1243                         break;
1244                 }
1245                 if (err)
1246                         break;
1247         }
1248         return err;
1249 }
1250
1251 static int do_set_master(struct net_device *dev, int ifindex)
1252 {
1253         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1254         const struct net_device_ops *ops;
1255         int err;
1256
1257         if (upper_dev) {
1258                 if (upper_dev->ifindex == ifindex)
1259                         return 0;
1260                 ops = upper_dev->netdev_ops;
1261                 if (ops->ndo_del_slave) {
1262                         err = ops->ndo_del_slave(upper_dev, dev);
1263                         if (err)
1264                                 return err;
1265                 } else {
1266                         return -EOPNOTSUPP;
1267                 }
1268         }
1269
1270         if (ifindex) {
1271                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1272                 if (!upper_dev)
1273                         return -EINVAL;
1274                 ops = upper_dev->netdev_ops;
1275                 if (ops->ndo_add_slave) {
1276                         err = ops->ndo_add_slave(upper_dev, dev);
1277                         if (err)
1278                                 return err;
1279                 } else {
1280                         return -EOPNOTSUPP;
1281                 }
1282         }
1283         return 0;
1284 }
1285
1286 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1287                       struct nlattr **tb, char *ifname, int modified)
1288 {
1289         const struct net_device_ops *ops = dev->netdev_ops;
1290         int err;
1291
1292         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1293                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1294                 if (IS_ERR(net)) {
1295                         err = PTR_ERR(net);
1296                         goto errout;
1297                 }
1298                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1299                         err = -EPERM;
1300                         goto errout;
1301                 }
1302                 err = dev_change_net_namespace(dev, net, ifname);
1303                 put_net(net);
1304                 if (err)
1305                         goto errout;
1306                 modified = 1;
1307         }
1308
1309         if (tb[IFLA_MAP]) {
1310                 struct rtnl_link_ifmap *u_map;
1311                 struct ifmap k_map;
1312
1313                 if (!ops->ndo_set_config) {
1314                         err = -EOPNOTSUPP;
1315                         goto errout;
1316                 }
1317
1318                 if (!netif_device_present(dev)) {
1319                         err = -ENODEV;
1320                         goto errout;
1321                 }
1322
1323                 u_map = nla_data(tb[IFLA_MAP]);
1324                 k_map.mem_start = (unsigned long) u_map->mem_start;
1325                 k_map.mem_end = (unsigned long) u_map->mem_end;
1326                 k_map.base_addr = (unsigned short) u_map->base_addr;
1327                 k_map.irq = (unsigned char) u_map->irq;
1328                 k_map.dma = (unsigned char) u_map->dma;
1329                 k_map.port = (unsigned char) u_map->port;
1330
1331                 err = ops->ndo_set_config(dev, &k_map);
1332                 if (err < 0)
1333                         goto errout;
1334
1335                 modified = 1;
1336         }
1337
1338         if (tb[IFLA_ADDRESS]) {
1339                 struct sockaddr *sa;
1340                 int len;
1341
1342                 len = sizeof(sa_family_t) + dev->addr_len;
1343                 sa = kmalloc(len, GFP_KERNEL);
1344                 if (!sa) {
1345                         err = -ENOMEM;
1346                         goto errout;
1347                 }
1348                 sa->sa_family = dev->type;
1349                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1350                        dev->addr_len);
1351                 err = dev_set_mac_address(dev, sa);
1352                 kfree(sa);
1353                 if (err)
1354                         goto errout;
1355                 modified = 1;
1356         }
1357
1358         if (tb[IFLA_MTU]) {
1359                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1360                 if (err < 0)
1361                         goto errout;
1362                 modified = 1;
1363         }
1364
1365         if (tb[IFLA_GROUP]) {
1366                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1367                 modified = 1;
1368         }
1369
1370         /*
1371          * Interface selected by interface index but interface
1372          * name provided implies that a name change has been
1373          * requested.
1374          */
1375         if (ifm->ifi_index > 0 && ifname[0]) {
1376                 err = dev_change_name(dev, ifname);
1377                 if (err < 0)
1378                         goto errout;
1379                 modified = 1;
1380         }
1381
1382         if (tb[IFLA_IFALIAS]) {
1383                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1384                                     nla_len(tb[IFLA_IFALIAS]));
1385                 if (err < 0)
1386                         goto errout;
1387                 modified = 1;
1388         }
1389
1390         if (tb[IFLA_BROADCAST]) {
1391                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1392                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1393         }
1394
1395         if (ifm->ifi_flags || ifm->ifi_change) {
1396                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1397                 if (err < 0)
1398                         goto errout;
1399         }
1400
1401         if (tb[IFLA_MASTER]) {
1402                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1403                 if (err)
1404                         goto errout;
1405                 modified = 1;
1406         }
1407
1408         if (tb[IFLA_CARRIER]) {
1409                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1410                 if (err)
1411                         goto errout;
1412                 modified = 1;
1413         }
1414
1415         if (tb[IFLA_TXQLEN])
1416                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1417
1418         if (tb[IFLA_OPERSTATE])
1419                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1420
1421         if (tb[IFLA_LINKMODE]) {
1422                 write_lock_bh(&dev_base_lock);
1423                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1424                 write_unlock_bh(&dev_base_lock);
1425         }
1426
1427         if (tb[IFLA_VFINFO_LIST]) {
1428                 struct nlattr *attr;
1429                 int rem;
1430                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1431                         if (nla_type(attr) != IFLA_VF_INFO) {
1432                                 err = -EINVAL;
1433                                 goto errout;
1434                         }
1435                         err = do_setvfinfo(dev, attr);
1436                         if (err < 0)
1437                                 goto errout;
1438                         modified = 1;
1439                 }
1440         }
1441         err = 0;
1442
1443         if (tb[IFLA_VF_PORTS]) {
1444                 struct nlattr *port[IFLA_PORT_MAX+1];
1445                 struct nlattr *attr;
1446                 int vf;
1447                 int rem;
1448
1449                 err = -EOPNOTSUPP;
1450                 if (!ops->ndo_set_vf_port)
1451                         goto errout;
1452
1453                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1454                         if (nla_type(attr) != IFLA_VF_PORT)
1455                                 continue;
1456                         err = nla_parse_nested(port, IFLA_PORT_MAX,
1457                                 attr, ifla_port_policy);
1458                         if (err < 0)
1459                                 goto errout;
1460                         if (!port[IFLA_PORT_VF]) {
1461                                 err = -EOPNOTSUPP;
1462                                 goto errout;
1463                         }
1464                         vf = nla_get_u32(port[IFLA_PORT_VF]);
1465                         err = ops->ndo_set_vf_port(dev, vf, port);
1466                         if (err < 0)
1467                                 goto errout;
1468                         modified = 1;
1469                 }
1470         }
1471         err = 0;
1472
1473         if (tb[IFLA_PORT_SELF]) {
1474                 struct nlattr *port[IFLA_PORT_MAX+1];
1475
1476                 err = nla_parse_nested(port, IFLA_PORT_MAX,
1477                         tb[IFLA_PORT_SELF], ifla_port_policy);
1478                 if (err < 0)
1479                         goto errout;
1480
1481                 err = -EOPNOTSUPP;
1482                 if (ops->ndo_set_vf_port)
1483                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1484                 if (err < 0)
1485                         goto errout;
1486                 modified = 1;
1487         }
1488
1489         if (tb[IFLA_AF_SPEC]) {
1490                 struct nlattr *af;
1491                 int rem;
1492
1493                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1494                         const struct rtnl_af_ops *af_ops;
1495
1496                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1497                                 BUG();
1498
1499                         err = af_ops->set_link_af(dev, af);
1500                         if (err < 0)
1501                                 goto errout;
1502
1503                         modified = 1;
1504                 }
1505         }
1506         err = 0;
1507
1508 errout:
1509         if (err < 0 && modified)
1510                 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",
1511                                      dev->name);
1512
1513         return err;
1514 }
1515
1516 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1517 {
1518         struct net *net = sock_net(skb->sk);
1519         struct ifinfomsg *ifm;
1520         struct net_device *dev;
1521         int err;
1522         struct nlattr *tb[IFLA_MAX+1];
1523         char ifname[IFNAMSIZ];
1524
1525         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1526         if (err < 0)
1527                 goto errout;
1528
1529         if (tb[IFLA_IFNAME])
1530                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1531         else
1532                 ifname[0] = '\0';
1533
1534         err = -EINVAL;
1535         ifm = nlmsg_data(nlh);
1536         if (ifm->ifi_index > 0)
1537                 dev = __dev_get_by_index(net, ifm->ifi_index);
1538         else if (tb[IFLA_IFNAME])
1539                 dev = __dev_get_by_name(net, ifname);
1540         else
1541                 goto errout;
1542
1543         if (dev == NULL) {
1544                 err = -ENODEV;
1545                 goto errout;
1546         }
1547
1548         err = validate_linkmsg(dev, tb);
1549         if (err < 0)
1550                 goto errout;
1551
1552         err = do_setlink(dev, ifm, tb, ifname, 0);
1553 errout:
1554         return err;
1555 }
1556
1557 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1558 {
1559         struct net *net = sock_net(skb->sk);
1560         const struct rtnl_link_ops *ops;
1561         struct net_device *dev;
1562         struct ifinfomsg *ifm;
1563         char ifname[IFNAMSIZ];
1564         struct nlattr *tb[IFLA_MAX+1];
1565         int err;
1566         LIST_HEAD(list_kill);
1567
1568         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1569         if (err < 0)
1570                 return err;
1571
1572         if (tb[IFLA_IFNAME])
1573                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1574
1575         ifm = nlmsg_data(nlh);
1576         if (ifm->ifi_index > 0)
1577                 dev = __dev_get_by_index(net, ifm->ifi_index);
1578         else if (tb[IFLA_IFNAME])
1579                 dev = __dev_get_by_name(net, ifname);
1580         else
1581                 return -EINVAL;
1582
1583         if (!dev)
1584                 return -ENODEV;
1585
1586         ops = dev->rtnl_link_ops;
1587         if (!ops)
1588                 return -EOPNOTSUPP;
1589
1590         ops->dellink(dev, &list_kill);
1591         unregister_netdevice_many(&list_kill);
1592         list_del(&list_kill);
1593         return 0;
1594 }
1595
1596 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1597 {
1598         unsigned int old_flags;
1599         int err;
1600
1601         old_flags = dev->flags;
1602         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1603                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1604                 if (err < 0)
1605                         return err;
1606         }
1607
1608         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1609         rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1610
1611         __dev_notify_flags(dev, old_flags);
1612         return 0;
1613 }
1614 EXPORT_SYMBOL(rtnl_configure_link);
1615
1616 struct net_device *rtnl_create_link(struct net *net,
1617         char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1618 {
1619         int err;
1620         struct net_device *dev;
1621         unsigned int num_tx_queues = 1;
1622         unsigned int num_rx_queues = 1;
1623
1624         if (tb[IFLA_NUM_TX_QUEUES])
1625                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1626         else if (ops->get_num_tx_queues)
1627                 num_tx_queues = ops->get_num_tx_queues();
1628
1629         if (tb[IFLA_NUM_RX_QUEUES])
1630                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1631         else if (ops->get_num_rx_queues)
1632                 num_rx_queues = ops->get_num_rx_queues();
1633
1634         err = -ENOMEM;
1635         dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1636                                num_tx_queues, num_rx_queues);
1637         if (!dev)
1638                 goto err;
1639
1640         dev_net_set(dev, net);
1641         dev->rtnl_link_ops = ops;
1642         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1643
1644         if (tb[IFLA_MTU])
1645                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1646         if (tb[IFLA_ADDRESS]) {
1647                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1648                                 nla_len(tb[IFLA_ADDRESS]));
1649                 dev->addr_assign_type = NET_ADDR_SET;
1650         }
1651         if (tb[IFLA_BROADCAST])
1652                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1653                                 nla_len(tb[IFLA_BROADCAST]));
1654         if (tb[IFLA_TXQLEN])
1655                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1656         if (tb[IFLA_OPERSTATE])
1657                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1658         if (tb[IFLA_LINKMODE])
1659                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1660         if (tb[IFLA_GROUP])
1661                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1662
1663         return dev;
1664
1665 err:
1666         return ERR_PTR(err);
1667 }
1668 EXPORT_SYMBOL(rtnl_create_link);
1669
1670 static int rtnl_group_changelink(struct net *net, int group,
1671                 struct ifinfomsg *ifm,
1672                 struct nlattr **tb)
1673 {
1674         struct net_device *dev;
1675         int err;
1676
1677         for_each_netdev(net, dev) {
1678                 if (dev->group == group) {
1679                         err = do_setlink(dev, ifm, tb, NULL, 0);
1680                         if (err < 0)
1681                                 return err;
1682                 }
1683         }
1684
1685         return 0;
1686 }
1687
1688 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1689 {
1690         struct net *net = sock_net(skb->sk);
1691         const struct rtnl_link_ops *ops;
1692         struct net_device *dev;
1693         struct ifinfomsg *ifm;
1694         char kind[MODULE_NAME_LEN];
1695         char ifname[IFNAMSIZ];
1696         struct nlattr *tb[IFLA_MAX+1];
1697         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1698         int err;
1699
1700 #ifdef CONFIG_MODULES
1701 replay:
1702 #endif
1703         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1704         if (err < 0)
1705                 return err;
1706
1707         if (tb[IFLA_IFNAME])
1708                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1709         else
1710                 ifname[0] = '\0';
1711
1712         ifm = nlmsg_data(nlh);
1713         if (ifm->ifi_index > 0)
1714                 dev = __dev_get_by_index(net, ifm->ifi_index);
1715         else {
1716                 if (ifname[0])
1717                         dev = __dev_get_by_name(net, ifname);
1718                 else
1719                         dev = NULL;
1720         }
1721
1722         err = validate_linkmsg(dev, tb);
1723         if (err < 0)
1724                 return err;
1725
1726         if (tb[IFLA_LINKINFO]) {
1727                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1728                                        tb[IFLA_LINKINFO], ifla_info_policy);
1729                 if (err < 0)
1730                         return err;
1731         } else
1732                 memset(linkinfo, 0, sizeof(linkinfo));
1733
1734         if (linkinfo[IFLA_INFO_KIND]) {
1735                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1736                 ops = rtnl_link_ops_get(kind);
1737         } else {
1738                 kind[0] = '\0';
1739                 ops = NULL;
1740         }
1741
1742         if (1) {
1743                 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1744                 struct net *dest_net;
1745
1746                 if (ops) {
1747                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1748                                 err = nla_parse_nested(attr, ops->maxtype,
1749                                                        linkinfo[IFLA_INFO_DATA],
1750                                                        ops->policy);
1751                                 if (err < 0)
1752                                         return err;
1753                                 data = attr;
1754                         }
1755                         if (ops->validate) {
1756                                 err = ops->validate(tb, data);
1757                                 if (err < 0)
1758                                         return err;
1759                         }
1760                 }
1761
1762                 if (dev) {
1763                         int modified = 0;
1764
1765                         if (nlh->nlmsg_flags & NLM_F_EXCL)
1766                                 return -EEXIST;
1767                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
1768                                 return -EOPNOTSUPP;
1769
1770                         if (linkinfo[IFLA_INFO_DATA]) {
1771                                 if (!ops || ops != dev->rtnl_link_ops ||
1772                                     !ops->changelink)
1773                                         return -EOPNOTSUPP;
1774
1775                                 err = ops->changelink(dev, tb, data);
1776                                 if (err < 0)
1777                                         return err;
1778                                 modified = 1;
1779                         }
1780
1781                         return do_setlink(dev, ifm, tb, ifname, modified);
1782                 }
1783
1784                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1785                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1786                                 return rtnl_group_changelink(net,
1787                                                 nla_get_u32(tb[IFLA_GROUP]),
1788                                                 ifm, tb);
1789                         return -ENODEV;
1790                 }
1791
1792                 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1793                         return -EOPNOTSUPP;
1794
1795                 if (!ops) {
1796 #ifdef CONFIG_MODULES
1797                         if (kind[0]) {
1798                                 __rtnl_unlock();
1799                                 request_module("rtnl-link-%s", kind);
1800                                 rtnl_lock();
1801                                 ops = rtnl_link_ops_get(kind);
1802                                 if (ops)
1803                                         goto replay;
1804                         }
1805 #endif
1806                         return -EOPNOTSUPP;
1807                 }
1808
1809                 if (!ifname[0])
1810                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1811
1812                 dest_net = rtnl_link_get_net(net, tb);
1813                 if (IS_ERR(dest_net))
1814                         return PTR_ERR(dest_net);
1815
1816                 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1817                 if (IS_ERR(dev)) {
1818                         err = PTR_ERR(dev);
1819                         goto out;
1820                 }
1821
1822                 dev->ifindex = ifm->ifi_index;
1823
1824                 if (ops->newlink)
1825                         err = ops->newlink(net, dev, tb, data);
1826                 else
1827                         err = register_netdevice(dev);
1828
1829                 if (err < 0 && !IS_ERR(dev))
1830                         free_netdev(dev);
1831                 if (err < 0)
1832                         goto out;
1833
1834                 err = rtnl_configure_link(dev, ifm);
1835                 if (err < 0)
1836                         unregister_netdevice(dev);
1837 out:
1838                 put_net(dest_net);
1839                 return err;
1840         }
1841 }
1842
1843 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1844 {
1845         struct net *net = sock_net(skb->sk);
1846         struct ifinfomsg *ifm;
1847         char ifname[IFNAMSIZ];
1848         struct nlattr *tb[IFLA_MAX+1];
1849         struct net_device *dev = NULL;
1850         struct sk_buff *nskb;
1851         int err;
1852         u32 ext_filter_mask = 0;
1853
1854         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1855         if (err < 0)
1856                 return err;
1857
1858         if (tb[IFLA_IFNAME])
1859                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1860
1861         if (tb[IFLA_EXT_MASK])
1862                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1863
1864         ifm = nlmsg_data(nlh);
1865         if (ifm->ifi_index > 0)
1866                 dev = __dev_get_by_index(net, ifm->ifi_index);
1867         else if (tb[IFLA_IFNAME])
1868                 dev = __dev_get_by_name(net, ifname);
1869         else
1870                 return -EINVAL;
1871
1872         if (dev == NULL)
1873                 return -ENODEV;
1874
1875         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1876         if (nskb == NULL)
1877                 return -ENOBUFS;
1878
1879         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1880                                nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1881         if (err < 0) {
1882                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1883                 WARN_ON(err == -EMSGSIZE);
1884                 kfree_skb(nskb);
1885         } else
1886                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1887
1888         return err;
1889 }
1890
1891 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1892 {
1893         struct net *net = sock_net(skb->sk);
1894         struct net_device *dev;
1895         struct nlattr *tb[IFLA_MAX+1];
1896         u32 ext_filter_mask = 0;
1897         u16 min_ifinfo_dump_size = 0;
1898
1899         if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1900                         ifla_policy) >= 0) {
1901                 if (tb[IFLA_EXT_MASK])
1902                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1903         }
1904
1905         if (!ext_filter_mask)
1906                 return NLMSG_GOODSIZE;
1907         /*
1908          * traverse the list of net devices and compute the minimum
1909          * buffer size based upon the filter mask.
1910          */
1911         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1912                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1913                                              if_nlmsg_size(dev,
1914                                                            ext_filter_mask));
1915         }
1916
1917         return min_ifinfo_dump_size;
1918 }
1919
1920 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1921 {
1922         int idx;
1923         int s_idx = cb->family;
1924
1925         if (s_idx == 0)
1926                 s_idx = 1;
1927         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1928                 int type = cb->nlh->nlmsg_type-RTM_BASE;
1929                 if (idx < s_idx || idx == PF_PACKET)
1930                         continue;
1931                 if (rtnl_msg_handlers[idx] == NULL ||
1932                     rtnl_msg_handlers[idx][type].dumpit == NULL)
1933                         continue;
1934                 if (idx > s_idx) {
1935                         memset(&cb->args[0], 0, sizeof(cb->args));
1936                         cb->prev_seq = 0;
1937                         cb->seq = 0;
1938                 }
1939                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1940                         break;
1941         }
1942         cb->family = idx;
1943
1944         return skb->len;
1945 }
1946
1947 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1948 {
1949         struct net *net = dev_net(dev);
1950         struct sk_buff *skb;
1951         int err = -ENOBUFS;
1952         size_t if_info_size;
1953
1954         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1955         if (skb == NULL)
1956                 goto errout;
1957
1958         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1959         if (err < 0) {
1960                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1961                 WARN_ON(err == -EMSGSIZE);
1962                 kfree_skb(skb);
1963                 goto errout;
1964         }
1965         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1966         return;
1967 errout:
1968         if (err < 0)
1969                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1970 }
1971 EXPORT_SYMBOL(rtmsg_ifinfo);
1972
1973 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1974                                    struct net_device *dev,
1975                                    u8 *addr, u32 pid, u32 seq,
1976                                    int type, unsigned int flags)
1977 {
1978         struct nlmsghdr *nlh;
1979         struct ndmsg *ndm;
1980
1981         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1982         if (!nlh)
1983                 return -EMSGSIZE;
1984
1985         ndm = nlmsg_data(nlh);
1986         ndm->ndm_family  = AF_BRIDGE;
1987         ndm->ndm_pad1    = 0;
1988         ndm->ndm_pad2    = 0;
1989         ndm->ndm_flags   = flags;
1990         ndm->ndm_type    = 0;
1991         ndm->ndm_ifindex = dev->ifindex;
1992         ndm->ndm_state   = NUD_PERMANENT;
1993
1994         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
1995                 goto nla_put_failure;
1996
1997         return nlmsg_end(skb, nlh);
1998
1999 nla_put_failure:
2000         nlmsg_cancel(skb, nlh);
2001         return -EMSGSIZE;
2002 }
2003
2004 static inline size_t rtnl_fdb_nlmsg_size(void)
2005 {
2006         return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2007 }
2008
2009 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2010 {
2011         struct net *net = dev_net(dev);
2012         struct sk_buff *skb;
2013         int err = -ENOBUFS;
2014
2015         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2016         if (!skb)
2017                 goto errout;
2018
2019         err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2020         if (err < 0) {
2021                 kfree_skb(skb);
2022                 goto errout;
2023         }
2024
2025         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2026         return;
2027 errout:
2028         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2029 }
2030
2031 /**
2032  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2033  */
2034 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2035                      struct nlattr *tb[],
2036                      struct net_device *dev,
2037                      const unsigned char *addr,
2038                      u16 flags)
2039 {
2040         int err = -EINVAL;
2041
2042         /* If aging addresses are supported device will need to
2043          * implement its own handler for this.
2044          */
2045         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2046                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2047                 return err;
2048         }
2049
2050         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2051                 err = dev_uc_add_excl(dev, addr);
2052         else if (is_multicast_ether_addr(addr))
2053                 err = dev_mc_add_excl(dev, addr);
2054
2055         /* Only return duplicate errors if NLM_F_EXCL is set */
2056         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2057                 err = 0;
2058
2059         return err;
2060 }
2061 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2062
2063 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2064 {
2065         struct net *net = sock_net(skb->sk);
2066         struct ndmsg *ndm;
2067         struct nlattr *tb[NDA_MAX+1];
2068         struct net_device *dev;
2069         u8 *addr;
2070         int err;
2071
2072         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2073         if (err < 0)
2074                 return err;
2075
2076         ndm = nlmsg_data(nlh);
2077         if (ndm->ndm_ifindex == 0) {
2078                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2079                 return -EINVAL;
2080         }
2081
2082         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2083         if (dev == NULL) {
2084                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2085                 return -ENODEV;
2086         }
2087
2088         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2089                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2090                 return -EINVAL;
2091         }
2092
2093         addr = nla_data(tb[NDA_LLADDR]);
2094         if (is_zero_ether_addr(addr)) {
2095                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2096                 return -EINVAL;
2097         }
2098
2099         err = -EOPNOTSUPP;
2100
2101         /* Support fdb on master device the net/bridge default case */
2102         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2103             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2104                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2105                 const struct net_device_ops *ops = br_dev->netdev_ops;
2106
2107                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2108                 if (err)
2109                         goto out;
2110                 else
2111                         ndm->ndm_flags &= ~NTF_MASTER;
2112         }
2113
2114         /* Embedded bridge, macvlan, and any other device support */
2115         if ((ndm->ndm_flags & NTF_SELF)) {
2116                 if (dev->netdev_ops->ndo_fdb_add)
2117                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2118                                                            nlh->nlmsg_flags);
2119                 else
2120                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2121                                                nlh->nlmsg_flags);
2122
2123                 if (!err) {
2124                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2125                         ndm->ndm_flags &= ~NTF_SELF;
2126                 }
2127         }
2128 out:
2129         return err;
2130 }
2131
2132 /**
2133  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2134  */
2135 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2136                      struct nlattr *tb[],
2137                      struct net_device *dev,
2138                      const unsigned char *addr)
2139 {
2140         int err = -EOPNOTSUPP;
2141
2142         /* If aging addresses are supported device will need to
2143          * implement its own handler for this.
2144          */
2145         if (ndm->ndm_state & NUD_PERMANENT) {
2146                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2147                 return -EINVAL;
2148         }
2149
2150         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2151                 err = dev_uc_del(dev, addr);
2152         else if (is_multicast_ether_addr(addr))
2153                 err = dev_mc_del(dev, addr);
2154         else
2155                 err = -EINVAL;
2156
2157         return err;
2158 }
2159 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2160
2161 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2162 {
2163         struct net *net = sock_net(skb->sk);
2164         struct ndmsg *ndm;
2165         struct nlattr *tb[NDA_MAX+1];
2166         struct net_device *dev;
2167         int err = -EINVAL;
2168         __u8 *addr;
2169
2170         if (!capable(CAP_NET_ADMIN))
2171                 return -EPERM;
2172
2173         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2174         if (err < 0)
2175                 return err;
2176
2177         ndm = nlmsg_data(nlh);
2178         if (ndm->ndm_ifindex == 0) {
2179                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2180                 return -EINVAL;
2181         }
2182
2183         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2184         if (dev == NULL) {
2185                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2186                 return -ENODEV;
2187         }
2188
2189         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2190                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2191                 return -EINVAL;
2192         }
2193
2194         addr = nla_data(tb[NDA_LLADDR]);
2195         if (is_zero_ether_addr(addr)) {
2196                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2197                 return -EINVAL;
2198         }
2199
2200         err = -EOPNOTSUPP;
2201
2202         /* Support fdb on master device the net/bridge default case */
2203         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2204             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2205                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2206                 const struct net_device_ops *ops = br_dev->netdev_ops;
2207
2208                 if (ops->ndo_fdb_del)
2209                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2210
2211                 if (err)
2212                         goto out;
2213                 else
2214                         ndm->ndm_flags &= ~NTF_MASTER;
2215         }
2216
2217         /* Embedded bridge, macvlan, and any other device support */
2218         if (ndm->ndm_flags & NTF_SELF) {
2219                 if (dev->netdev_ops->ndo_fdb_del)
2220                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2221                 else
2222                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2223
2224                 if (!err) {
2225                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2226                         ndm->ndm_flags &= ~NTF_SELF;
2227                 }
2228         }
2229 out:
2230         return err;
2231 }
2232
2233 static int nlmsg_populate_fdb(struct sk_buff *skb,
2234                               struct netlink_callback *cb,
2235                               struct net_device *dev,
2236                               int *idx,
2237                               struct netdev_hw_addr_list *list)
2238 {
2239         struct netdev_hw_addr *ha;
2240         int err;
2241         u32 portid, seq;
2242
2243         portid = NETLINK_CB(cb->skb).portid;
2244         seq = cb->nlh->nlmsg_seq;
2245
2246         list_for_each_entry(ha, &list->list, list) {
2247                 if (*idx < cb->args[0])
2248                         goto skip;
2249
2250                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2251                                               portid, seq,
2252                                               RTM_NEWNEIGH, NTF_SELF);
2253                 if (err < 0)
2254                         return err;
2255 skip:
2256                 *idx += 1;
2257         }
2258         return 0;
2259 }
2260
2261 /**
2262  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2263  * @nlh: netlink message header
2264  * @dev: netdevice
2265  *
2266  * Default netdevice operation to dump the existing unicast address list.
2267  * Returns number of addresses from list put in skb.
2268  */
2269 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2270                       struct netlink_callback *cb,
2271                       struct net_device *dev,
2272                       int idx)
2273 {
2274         int err;
2275
2276         netif_addr_lock_bh(dev);
2277         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2278         if (err)
2279                 goto out;
2280         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2281 out:
2282         netif_addr_unlock_bh(dev);
2283         return idx;
2284 }
2285 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2286
2287 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2288 {
2289         int idx = 0;
2290         struct net *net = sock_net(skb->sk);
2291         struct net_device *dev;
2292
2293         rcu_read_lock();
2294         for_each_netdev_rcu(net, dev) {
2295                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2296                         struct net_device *br_dev;
2297                         const struct net_device_ops *ops;
2298
2299                         br_dev = netdev_master_upper_dev_get(dev);
2300                         ops = br_dev->netdev_ops;
2301                         if (ops->ndo_fdb_dump)
2302                                 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2303                 }
2304
2305                 if (dev->netdev_ops->ndo_fdb_dump)
2306                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2307                 else
2308                         idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2309         }
2310         rcu_read_unlock();
2311
2312         cb->args[0] = idx;
2313         return skb->len;
2314 }
2315
2316 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2317                             struct net_device *dev, u16 mode)
2318 {
2319         struct nlmsghdr *nlh;
2320         struct ifinfomsg *ifm;
2321         struct nlattr *br_afspec;
2322         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2323         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2324
2325         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2326         if (nlh == NULL)
2327                 return -EMSGSIZE;
2328
2329         ifm = nlmsg_data(nlh);
2330         ifm->ifi_family = AF_BRIDGE;
2331         ifm->__ifi_pad = 0;
2332         ifm->ifi_type = dev->type;
2333         ifm->ifi_index = dev->ifindex;
2334         ifm->ifi_flags = dev_get_flags(dev);
2335         ifm->ifi_change = 0;
2336
2337
2338         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2339             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2340             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2341             (br_dev &&
2342              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2343             (dev->addr_len &&
2344              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2345             (dev->ifindex != dev->iflink &&
2346              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2347                 goto nla_put_failure;
2348
2349         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2350         if (!br_afspec)
2351                 goto nla_put_failure;
2352
2353         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2354             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2355                 nla_nest_cancel(skb, br_afspec);
2356                 goto nla_put_failure;
2357         }
2358         nla_nest_end(skb, br_afspec);
2359
2360         return nlmsg_end(skb, nlh);
2361 nla_put_failure:
2362         nlmsg_cancel(skb, nlh);
2363         return -EMSGSIZE;
2364 }
2365 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2366
2367 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2368 {
2369         struct net *net = sock_net(skb->sk);
2370         struct net_device *dev;
2371         int idx = 0;
2372         u32 portid = NETLINK_CB(cb->skb).portid;
2373         u32 seq = cb->nlh->nlmsg_seq;
2374         struct nlattr *extfilt;
2375         u32 filter_mask = 0;
2376
2377         extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct rtgenmsg),
2378                                   IFLA_EXT_MASK);
2379         if (extfilt)
2380                 filter_mask = nla_get_u32(extfilt);
2381
2382         rcu_read_lock();
2383         for_each_netdev_rcu(net, dev) {
2384                 const struct net_device_ops *ops = dev->netdev_ops;
2385                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2386
2387                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2388                         if (idx >= cb->args[0] &&
2389                             br_dev->netdev_ops->ndo_bridge_getlink(
2390                                     skb, portid, seq, dev, filter_mask) < 0)
2391                                 break;
2392                         idx++;
2393                 }
2394
2395                 if (ops->ndo_bridge_getlink) {
2396                         if (idx >= cb->args[0] &&
2397                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2398                                                     filter_mask) < 0)
2399                                 break;
2400                         idx++;
2401                 }
2402         }
2403         rcu_read_unlock();
2404         cb->args[0] = idx;
2405
2406         return skb->len;
2407 }
2408
2409 static inline size_t bridge_nlmsg_size(void)
2410 {
2411         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2412                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2413                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2414                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2415                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2416                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2417                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2418                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2419                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2420                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2421                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2422 }
2423
2424 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2425 {
2426         struct net *net = dev_net(dev);
2427         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2428         struct sk_buff *skb;
2429         int err = -EOPNOTSUPP;
2430
2431         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2432         if (!skb) {
2433                 err = -ENOMEM;
2434                 goto errout;
2435         }
2436
2437         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2438             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2439                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2440                 if (err < 0)
2441                         goto errout;
2442         }
2443
2444         if ((flags & BRIDGE_FLAGS_SELF) &&
2445             dev->netdev_ops->ndo_bridge_getlink) {
2446                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2447                 if (err < 0)
2448                         goto errout;
2449         }
2450
2451         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2452         return 0;
2453 errout:
2454         WARN_ON(err == -EMSGSIZE);
2455         kfree_skb(skb);
2456         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2457         return err;
2458 }
2459
2460 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2461 {
2462         struct net *net = sock_net(skb->sk);
2463         struct ifinfomsg *ifm;
2464         struct net_device *dev;
2465         struct nlattr *br_spec, *attr = NULL;
2466         int rem, err = -EOPNOTSUPP;
2467         u16 oflags, flags = 0;
2468         bool have_flags = false;
2469
2470         if (nlmsg_len(nlh) < sizeof(*ifm))
2471                 return -EINVAL;
2472
2473         ifm = nlmsg_data(nlh);
2474         if (ifm->ifi_family != AF_BRIDGE)
2475                 return -EPFNOSUPPORT;
2476
2477         dev = __dev_get_by_index(net, ifm->ifi_index);
2478         if (!dev) {
2479                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2480                 return -ENODEV;
2481         }
2482
2483         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2484         if (br_spec) {
2485                 nla_for_each_nested(attr, br_spec, rem) {
2486                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2487                                 have_flags = true;
2488                                 flags = nla_get_u16(attr);
2489                                 break;
2490                         }
2491                 }
2492         }
2493
2494         oflags = flags;
2495
2496         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2497                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2498
2499                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2500                         err = -EOPNOTSUPP;
2501                         goto out;
2502                 }
2503
2504                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2505                 if (err)
2506                         goto out;
2507
2508                 flags &= ~BRIDGE_FLAGS_MASTER;
2509         }
2510
2511         if ((flags & BRIDGE_FLAGS_SELF)) {
2512                 if (!dev->netdev_ops->ndo_bridge_setlink)
2513                         err = -EOPNOTSUPP;
2514                 else
2515                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2516
2517                 if (!err)
2518                         flags &= ~BRIDGE_FLAGS_SELF;
2519         }
2520
2521         if (have_flags)
2522                 memcpy(nla_data(attr), &flags, sizeof(flags));
2523         /* Generate event to notify upper layer of bridge change */
2524         if (!err)
2525                 err = rtnl_bridge_notify(dev, oflags);
2526 out:
2527         return err;
2528 }
2529
2530 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2531 {
2532         struct net *net = sock_net(skb->sk);
2533         struct ifinfomsg *ifm;
2534         struct net_device *dev;
2535         struct nlattr *br_spec, *attr = NULL;
2536         int rem, err = -EOPNOTSUPP;
2537         u16 oflags, flags = 0;
2538         bool have_flags = false;
2539
2540         if (nlmsg_len(nlh) < sizeof(*ifm))
2541                 return -EINVAL;
2542
2543         ifm = nlmsg_data(nlh);
2544         if (ifm->ifi_family != AF_BRIDGE)
2545                 return -EPFNOSUPPORT;
2546
2547         dev = __dev_get_by_index(net, ifm->ifi_index);
2548         if (!dev) {
2549                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2550                 return -ENODEV;
2551         }
2552
2553         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2554         if (br_spec) {
2555                 nla_for_each_nested(attr, br_spec, rem) {
2556                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2557                                 have_flags = true;
2558                                 flags = nla_get_u16(attr);
2559                                 break;
2560                         }
2561                 }
2562         }
2563
2564         oflags = flags;
2565
2566         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2567                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2568
2569                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2570                         err = -EOPNOTSUPP;
2571                         goto out;
2572                 }
2573
2574                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2575                 if (err)
2576                         goto out;
2577
2578                 flags &= ~BRIDGE_FLAGS_MASTER;
2579         }
2580
2581         if ((flags & BRIDGE_FLAGS_SELF)) {
2582                 if (!dev->netdev_ops->ndo_bridge_dellink)
2583                         err = -EOPNOTSUPP;
2584                 else
2585                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2586
2587                 if (!err)
2588                         flags &= ~BRIDGE_FLAGS_SELF;
2589         }
2590
2591         if (have_flags)
2592                 memcpy(nla_data(attr), &flags, sizeof(flags));
2593         /* Generate event to notify upper layer of bridge change */
2594         if (!err)
2595                 err = rtnl_bridge_notify(dev, oflags);
2596 out:
2597         return err;
2598 }
2599
2600 /* Process one rtnetlink message. */
2601
2602 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2603 {
2604         struct net *net = sock_net(skb->sk);
2605         rtnl_doit_func doit;
2606         int sz_idx, kind;
2607         int family;
2608         int type;
2609         int err;
2610
2611         type = nlh->nlmsg_type;
2612         if (type > RTM_MAX)
2613                 return -EOPNOTSUPP;
2614
2615         type -= RTM_BASE;
2616
2617         /* All the messages must have at least 1 byte length */
2618         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2619                 return 0;
2620
2621         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2622         sz_idx = type>>2;
2623         kind = type&3;
2624
2625         if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2626                 return -EPERM;
2627
2628         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2629                 struct sock *rtnl;
2630                 rtnl_dumpit_func dumpit;
2631                 rtnl_calcit_func calcit;
2632                 u16 min_dump_alloc = 0;
2633
2634                 dumpit = rtnl_get_dumpit(family, type);
2635                 if (dumpit == NULL)
2636                         return -EOPNOTSUPP;
2637                 calcit = rtnl_get_calcit(family, type);
2638                 if (calcit)
2639                         min_dump_alloc = calcit(skb, nlh);
2640
2641                 __rtnl_unlock();
2642                 rtnl = net->rtnl;
2643                 {
2644                         struct netlink_dump_control c = {
2645                                 .dump           = dumpit,
2646                                 .min_dump_alloc = min_dump_alloc,
2647                         };
2648                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2649                 }
2650                 rtnl_lock();
2651                 return err;
2652         }
2653
2654         doit = rtnl_get_doit(family, type);
2655         if (doit == NULL)
2656                 return -EOPNOTSUPP;
2657
2658         return doit(skb, nlh);
2659 }
2660
2661 static void rtnetlink_rcv(struct sk_buff *skb)
2662 {
2663         rtnl_lock();
2664         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2665         rtnl_unlock();
2666 }
2667
2668 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2669 {
2670         struct net_device *dev = ptr;
2671
2672         switch (event) {
2673         case NETDEV_UP:
2674         case NETDEV_DOWN:
2675         case NETDEV_PRE_UP:
2676         case NETDEV_POST_INIT:
2677         case NETDEV_REGISTER:
2678         case NETDEV_CHANGE:
2679         case NETDEV_PRE_TYPE_CHANGE:
2680         case NETDEV_GOING_DOWN:
2681         case NETDEV_UNREGISTER:
2682         case NETDEV_UNREGISTER_FINAL:
2683         case NETDEV_RELEASE:
2684         case NETDEV_JOIN:
2685                 break;
2686         default:
2687                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2688                 break;
2689         }
2690         return NOTIFY_DONE;
2691 }
2692
2693 static struct notifier_block rtnetlink_dev_notifier = {
2694         .notifier_call  = rtnetlink_event,
2695 };
2696
2697
2698 static int __net_init rtnetlink_net_init(struct net *net)
2699 {
2700         struct sock *sk;
2701         struct netlink_kernel_cfg cfg = {
2702                 .groups         = RTNLGRP_MAX,
2703                 .input          = rtnetlink_rcv,
2704                 .cb_mutex       = &rtnl_mutex,
2705                 .flags          = NL_CFG_F_NONROOT_RECV,
2706         };
2707
2708         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2709         if (!sk)
2710                 return -ENOMEM;
2711         net->rtnl = sk;
2712         return 0;
2713 }
2714
2715 static void __net_exit rtnetlink_net_exit(struct net *net)
2716 {
2717         netlink_kernel_release(net->rtnl);
2718         net->rtnl = NULL;
2719 }
2720
2721 static struct pernet_operations rtnetlink_net_ops = {
2722         .init = rtnetlink_net_init,
2723         .exit = rtnetlink_net_exit,
2724 };
2725
2726 void __init rtnetlink_init(void)
2727 {
2728         if (register_pernet_subsys(&rtnetlink_net_ops))
2729                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2730
2731         register_netdevice_notifier(&rtnetlink_dev_notifier);
2732
2733         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2734                       rtnl_dump_ifinfo, rtnl_calcit);
2735         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2736         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2737         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2738
2739         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2740         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2741
2742         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2743         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2744         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2745
2746         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2747         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2748         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
2749 }
2750