]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/core/rtnetlink.c
netfilter: nft_ct: prepare for key-dependent error unwind
[mirror_ubuntu-artful-kernel.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/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
42
43 #include <linux/uaccess.h>
44
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
48 #include <net/ip.h>
49 #include <net/protocol.h>
50 #include <net/arp.h>
51 #include <net/route.h>
52 #include <net/udp.h>
53 #include <net/tcp.h>
54 #include <net/sock.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
59
60 struct rtnl_link {
61 rtnl_doit_func doit;
62 rtnl_dumpit_func dumpit;
63 rtnl_calcit_func calcit;
64 };
65
66 static DEFINE_MUTEX(rtnl_mutex);
67
68 void rtnl_lock(void)
69 {
70 mutex_lock(&rtnl_mutex);
71 }
72 EXPORT_SYMBOL(rtnl_lock);
73
74 static struct sk_buff *defer_kfree_skb_list;
75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
76 {
77 if (head && tail) {
78 tail->next = defer_kfree_skb_list;
79 defer_kfree_skb_list = head;
80 }
81 }
82 EXPORT_SYMBOL(rtnl_kfree_skbs);
83
84 void __rtnl_unlock(void)
85 {
86 struct sk_buff *head = defer_kfree_skb_list;
87
88 defer_kfree_skb_list = NULL;
89
90 mutex_unlock(&rtnl_mutex);
91
92 while (head) {
93 struct sk_buff *next = head->next;
94
95 kfree_skb(head);
96 cond_resched();
97 head = next;
98 }
99 }
100
101 void rtnl_unlock(void)
102 {
103 /* This fellow will unlock it for us. */
104 netdev_run_todo();
105 }
106 EXPORT_SYMBOL(rtnl_unlock);
107
108 int rtnl_trylock(void)
109 {
110 return mutex_trylock(&rtnl_mutex);
111 }
112 EXPORT_SYMBOL(rtnl_trylock);
113
114 int rtnl_is_locked(void)
115 {
116 return mutex_is_locked(&rtnl_mutex);
117 }
118 EXPORT_SYMBOL(rtnl_is_locked);
119
120 #ifdef CONFIG_PROVE_LOCKING
121 bool lockdep_rtnl_is_held(void)
122 {
123 return lockdep_is_held(&rtnl_mutex);
124 }
125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
127
128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
129
130 static inline int rtm_msgindex(int msgtype)
131 {
132 int msgindex = msgtype - RTM_BASE;
133
134 /*
135 * msgindex < 0 implies someone tried to register a netlink
136 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
137 * the message type has not been added to linux/rtnetlink.h
138 */
139 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
140
141 return msgindex;
142 }
143
144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
145 {
146 struct rtnl_link *tab;
147
148 if (protocol <= RTNL_FAMILY_MAX)
149 tab = rtnl_msg_handlers[protocol];
150 else
151 tab = NULL;
152
153 if (tab == NULL || tab[msgindex].doit == NULL)
154 tab = rtnl_msg_handlers[PF_UNSPEC];
155
156 return tab[msgindex].doit;
157 }
158
159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
160 {
161 struct rtnl_link *tab;
162
163 if (protocol <= RTNL_FAMILY_MAX)
164 tab = rtnl_msg_handlers[protocol];
165 else
166 tab = NULL;
167
168 if (tab == NULL || tab[msgindex].dumpit == NULL)
169 tab = rtnl_msg_handlers[PF_UNSPEC];
170
171 return tab[msgindex].dumpit;
172 }
173
174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
175 {
176 struct rtnl_link *tab;
177
178 if (protocol <= RTNL_FAMILY_MAX)
179 tab = rtnl_msg_handlers[protocol];
180 else
181 tab = NULL;
182
183 if (tab == NULL || tab[msgindex].calcit == NULL)
184 tab = rtnl_msg_handlers[PF_UNSPEC];
185
186 return tab[msgindex].calcit;
187 }
188
189 /**
190 * __rtnl_register - Register a rtnetlink message type
191 * @protocol: Protocol family or PF_UNSPEC
192 * @msgtype: rtnetlink message type
193 * @doit: Function pointer called for each request message
194 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
195 * @calcit: Function pointer to calc size of dump message
196 *
197 * Registers the specified function pointers (at least one of them has
198 * to be non-NULL) to be called whenever a request message for the
199 * specified protocol family and message type is received.
200 *
201 * The special protocol family PF_UNSPEC may be used to define fallback
202 * function pointers for the case when no entry for the specific protocol
203 * family exists.
204 *
205 * Returns 0 on success or a negative error code.
206 */
207 int __rtnl_register(int protocol, int msgtype,
208 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
209 rtnl_calcit_func calcit)
210 {
211 struct rtnl_link *tab;
212 int msgindex;
213
214 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
215 msgindex = rtm_msgindex(msgtype);
216
217 tab = rtnl_msg_handlers[protocol];
218 if (tab == NULL) {
219 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
220 if (tab == NULL)
221 return -ENOBUFS;
222
223 rtnl_msg_handlers[protocol] = tab;
224 }
225
226 if (doit)
227 tab[msgindex].doit = doit;
228
229 if (dumpit)
230 tab[msgindex].dumpit = dumpit;
231
232 if (calcit)
233 tab[msgindex].calcit = calcit;
234
235 return 0;
236 }
237 EXPORT_SYMBOL_GPL(__rtnl_register);
238
239 /**
240 * rtnl_register - Register a rtnetlink message type
241 *
242 * Identical to __rtnl_register() but panics on failure. This is useful
243 * as failure of this function is very unlikely, it can only happen due
244 * to lack of memory when allocating the chain to store all message
245 * handlers for a protocol. Meant for use in init functions where lack
246 * of memory implies no sense in continuing.
247 */
248 void rtnl_register(int protocol, int msgtype,
249 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
250 rtnl_calcit_func calcit)
251 {
252 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
253 panic("Unable to register rtnetlink message handler, "
254 "protocol = %d, message type = %d\n",
255 protocol, msgtype);
256 }
257 EXPORT_SYMBOL_GPL(rtnl_register);
258
259 /**
260 * rtnl_unregister - Unregister a rtnetlink message type
261 * @protocol: Protocol family or PF_UNSPEC
262 * @msgtype: rtnetlink message type
263 *
264 * Returns 0 on success or a negative error code.
265 */
266 int rtnl_unregister(int protocol, int msgtype)
267 {
268 int msgindex;
269
270 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 msgindex = rtm_msgindex(msgtype);
272
273 if (rtnl_msg_handlers[protocol] == NULL)
274 return -ENOENT;
275
276 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
277 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
278 rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
279
280 return 0;
281 }
282 EXPORT_SYMBOL_GPL(rtnl_unregister);
283
284 /**
285 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
286 * @protocol : Protocol family or PF_UNSPEC
287 *
288 * Identical to calling rtnl_unregster() for all registered message types
289 * of a certain protocol family.
290 */
291 void rtnl_unregister_all(int protocol)
292 {
293 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
294
295 kfree(rtnl_msg_handlers[protocol]);
296 rtnl_msg_handlers[protocol] = NULL;
297 }
298 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
299
300 static LIST_HEAD(link_ops);
301
302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
303 {
304 const struct rtnl_link_ops *ops;
305
306 list_for_each_entry(ops, &link_ops, list) {
307 if (!strcmp(ops->kind, kind))
308 return ops;
309 }
310 return NULL;
311 }
312
313 /**
314 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
315 * @ops: struct rtnl_link_ops * to register
316 *
317 * The caller must hold the rtnl_mutex. This function should be used
318 * by drivers that create devices during module initialization. It
319 * must be called before registering the devices.
320 *
321 * Returns 0 on success or a negative error code.
322 */
323 int __rtnl_link_register(struct rtnl_link_ops *ops)
324 {
325 if (rtnl_link_ops_get(ops->kind))
326 return -EEXIST;
327
328 /* The check for setup is here because if ops
329 * does not have that filled up, it is not possible
330 * to use the ops for creating device. So do not
331 * fill up dellink as well. That disables rtnl_dellink.
332 */
333 if (ops->setup && !ops->dellink)
334 ops->dellink = unregister_netdevice_queue;
335
336 list_add_tail(&ops->list, &link_ops);
337 return 0;
338 }
339 EXPORT_SYMBOL_GPL(__rtnl_link_register);
340
341 /**
342 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
343 * @ops: struct rtnl_link_ops * to register
344 *
345 * Returns 0 on success or a negative error code.
346 */
347 int rtnl_link_register(struct rtnl_link_ops *ops)
348 {
349 int err;
350
351 rtnl_lock();
352 err = __rtnl_link_register(ops);
353 rtnl_unlock();
354 return err;
355 }
356 EXPORT_SYMBOL_GPL(rtnl_link_register);
357
358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
359 {
360 struct net_device *dev;
361 LIST_HEAD(list_kill);
362
363 for_each_netdev(net, dev) {
364 if (dev->rtnl_link_ops == ops)
365 ops->dellink(dev, &list_kill);
366 }
367 unregister_netdevice_many(&list_kill);
368 }
369
370 /**
371 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
372 * @ops: struct rtnl_link_ops * to unregister
373 *
374 * The caller must hold the rtnl_mutex.
375 */
376 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
377 {
378 struct net *net;
379
380 for_each_net(net) {
381 __rtnl_kill_links(net, ops);
382 }
383 list_del(&ops->list);
384 }
385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
386
387 /* Return with the rtnl_lock held when there are no network
388 * devices unregistering in any network namespace.
389 */
390 static void rtnl_lock_unregistering_all(void)
391 {
392 struct net *net;
393 bool unregistering;
394 DEFINE_WAIT_FUNC(wait, woken_wake_function);
395
396 add_wait_queue(&netdev_unregistering_wq, &wait);
397 for (;;) {
398 unregistering = false;
399 rtnl_lock();
400 for_each_net(net) {
401 if (net->dev_unreg_count > 0) {
402 unregistering = true;
403 break;
404 }
405 }
406 if (!unregistering)
407 break;
408 __rtnl_unlock();
409
410 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
411 }
412 remove_wait_queue(&netdev_unregistering_wq, &wait);
413 }
414
415 /**
416 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
417 * @ops: struct rtnl_link_ops * to unregister
418 */
419 void rtnl_link_unregister(struct rtnl_link_ops *ops)
420 {
421 /* Close the race with cleanup_net() */
422 mutex_lock(&net_mutex);
423 rtnl_lock_unregistering_all();
424 __rtnl_link_unregister(ops);
425 rtnl_unlock();
426 mutex_unlock(&net_mutex);
427 }
428 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
429
430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
431 {
432 struct net_device *master_dev;
433 const struct rtnl_link_ops *ops;
434
435 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
436 if (!master_dev)
437 return 0;
438 ops = master_dev->rtnl_link_ops;
439 if (!ops || !ops->get_slave_size)
440 return 0;
441 /* IFLA_INFO_SLAVE_DATA + nested data */
442 return nla_total_size(sizeof(struct nlattr)) +
443 ops->get_slave_size(master_dev, dev);
444 }
445
446 static size_t rtnl_link_get_size(const struct net_device *dev)
447 {
448 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
449 size_t size;
450
451 if (!ops)
452 return 0;
453
454 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
455 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
456
457 if (ops->get_size)
458 /* IFLA_INFO_DATA + nested data */
459 size += nla_total_size(sizeof(struct nlattr)) +
460 ops->get_size(dev);
461
462 if (ops->get_xstats_size)
463 /* IFLA_INFO_XSTATS */
464 size += nla_total_size(ops->get_xstats_size(dev));
465
466 size += rtnl_link_get_slave_info_data_size(dev);
467
468 return size;
469 }
470
471 static LIST_HEAD(rtnl_af_ops);
472
473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
474 {
475 const struct rtnl_af_ops *ops;
476
477 list_for_each_entry(ops, &rtnl_af_ops, list) {
478 if (ops->family == family)
479 return ops;
480 }
481
482 return NULL;
483 }
484
485 /**
486 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
487 * @ops: struct rtnl_af_ops * to register
488 *
489 * Returns 0 on success or a negative error code.
490 */
491 void rtnl_af_register(struct rtnl_af_ops *ops)
492 {
493 rtnl_lock();
494 list_add_tail(&ops->list, &rtnl_af_ops);
495 rtnl_unlock();
496 }
497 EXPORT_SYMBOL_GPL(rtnl_af_register);
498
499 /**
500 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
501 * @ops: struct rtnl_af_ops * to unregister
502 *
503 * The caller must hold the rtnl_mutex.
504 */
505 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
506 {
507 list_del(&ops->list);
508 }
509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
510
511 /**
512 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
513 * @ops: struct rtnl_af_ops * to unregister
514 */
515 void rtnl_af_unregister(struct rtnl_af_ops *ops)
516 {
517 rtnl_lock();
518 __rtnl_af_unregister(ops);
519 rtnl_unlock();
520 }
521 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
522
523 static size_t rtnl_link_get_af_size(const struct net_device *dev,
524 u32 ext_filter_mask)
525 {
526 struct rtnl_af_ops *af_ops;
527 size_t size;
528
529 /* IFLA_AF_SPEC */
530 size = nla_total_size(sizeof(struct nlattr));
531
532 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
533 if (af_ops->get_link_af_size) {
534 /* AF_* + nested data */
535 size += nla_total_size(sizeof(struct nlattr)) +
536 af_ops->get_link_af_size(dev, ext_filter_mask);
537 }
538 }
539
540 return size;
541 }
542
543 static bool rtnl_have_link_slave_info(const struct net_device *dev)
544 {
545 struct net_device *master_dev;
546
547 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
548 if (master_dev && master_dev->rtnl_link_ops)
549 return true;
550 return false;
551 }
552
553 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
554 const struct net_device *dev)
555 {
556 struct net_device *master_dev;
557 const struct rtnl_link_ops *ops;
558 struct nlattr *slave_data;
559 int err;
560
561 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
562 if (!master_dev)
563 return 0;
564 ops = master_dev->rtnl_link_ops;
565 if (!ops)
566 return 0;
567 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
568 return -EMSGSIZE;
569 if (ops->fill_slave_info) {
570 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
571 if (!slave_data)
572 return -EMSGSIZE;
573 err = ops->fill_slave_info(skb, master_dev, dev);
574 if (err < 0)
575 goto err_cancel_slave_data;
576 nla_nest_end(skb, slave_data);
577 }
578 return 0;
579
580 err_cancel_slave_data:
581 nla_nest_cancel(skb, slave_data);
582 return err;
583 }
584
585 static int rtnl_link_info_fill(struct sk_buff *skb,
586 const struct net_device *dev)
587 {
588 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
589 struct nlattr *data;
590 int err;
591
592 if (!ops)
593 return 0;
594 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
595 return -EMSGSIZE;
596 if (ops->fill_xstats) {
597 err = ops->fill_xstats(skb, dev);
598 if (err < 0)
599 return err;
600 }
601 if (ops->fill_info) {
602 data = nla_nest_start(skb, IFLA_INFO_DATA);
603 if (data == NULL)
604 return -EMSGSIZE;
605 err = ops->fill_info(skb, dev);
606 if (err < 0)
607 goto err_cancel_data;
608 nla_nest_end(skb, data);
609 }
610 return 0;
611
612 err_cancel_data:
613 nla_nest_cancel(skb, data);
614 return err;
615 }
616
617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
618 {
619 struct nlattr *linkinfo;
620 int err = -EMSGSIZE;
621
622 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
623 if (linkinfo == NULL)
624 goto out;
625
626 err = rtnl_link_info_fill(skb, dev);
627 if (err < 0)
628 goto err_cancel_link;
629
630 err = rtnl_link_slave_info_fill(skb, dev);
631 if (err < 0)
632 goto err_cancel_link;
633
634 nla_nest_end(skb, linkinfo);
635 return 0;
636
637 err_cancel_link:
638 nla_nest_cancel(skb, linkinfo);
639 out:
640 return err;
641 }
642
643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
644 {
645 struct sock *rtnl = net->rtnl;
646 int err = 0;
647
648 NETLINK_CB(skb).dst_group = group;
649 if (echo)
650 atomic_inc(&skb->users);
651 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
652 if (echo)
653 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
654 return err;
655 }
656
657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
658 {
659 struct sock *rtnl = net->rtnl;
660
661 return nlmsg_unicast(rtnl, skb, pid);
662 }
663 EXPORT_SYMBOL(rtnl_unicast);
664
665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
666 struct nlmsghdr *nlh, gfp_t flags)
667 {
668 struct sock *rtnl = net->rtnl;
669 int report = 0;
670
671 if (nlh)
672 report = nlmsg_report(nlh);
673
674 nlmsg_notify(rtnl, skb, pid, group, report, flags);
675 }
676 EXPORT_SYMBOL(rtnl_notify);
677
678 void rtnl_set_sk_err(struct net *net, u32 group, int error)
679 {
680 struct sock *rtnl = net->rtnl;
681
682 netlink_set_err(rtnl, 0, group, error);
683 }
684 EXPORT_SYMBOL(rtnl_set_sk_err);
685
686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
687 {
688 struct nlattr *mx;
689 int i, valid = 0;
690
691 mx = nla_nest_start(skb, RTA_METRICS);
692 if (mx == NULL)
693 return -ENOBUFS;
694
695 for (i = 0; i < RTAX_MAX; i++) {
696 if (metrics[i]) {
697 if (i == RTAX_CC_ALGO - 1) {
698 char tmp[TCP_CA_NAME_MAX], *name;
699
700 name = tcp_ca_get_name_by_key(metrics[i], tmp);
701 if (!name)
702 continue;
703 if (nla_put_string(skb, i + 1, name))
704 goto nla_put_failure;
705 } else if (i == RTAX_FEATURES - 1) {
706 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
707
708 if (!user_features)
709 continue;
710 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
711 if (nla_put_u32(skb, i + 1, user_features))
712 goto nla_put_failure;
713 } else {
714 if (nla_put_u32(skb, i + 1, metrics[i]))
715 goto nla_put_failure;
716 }
717 valid++;
718 }
719 }
720
721 if (!valid) {
722 nla_nest_cancel(skb, mx);
723 return 0;
724 }
725
726 return nla_nest_end(skb, mx);
727
728 nla_put_failure:
729 nla_nest_cancel(skb, mx);
730 return -EMSGSIZE;
731 }
732 EXPORT_SYMBOL(rtnetlink_put_metrics);
733
734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
735 long expires, u32 error)
736 {
737 struct rta_cacheinfo ci = {
738 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
739 .rta_used = dst->__use,
740 .rta_clntref = atomic_read(&(dst->__refcnt)),
741 .rta_error = error,
742 .rta_id = id,
743 };
744
745 if (expires) {
746 unsigned long clock;
747
748 clock = jiffies_to_clock_t(abs(expires));
749 clock = min_t(unsigned long, clock, INT_MAX);
750 ci.rta_expires = (expires > 0) ? clock : -clock;
751 }
752 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
753 }
754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
755
756 static void set_operstate(struct net_device *dev, unsigned char transition)
757 {
758 unsigned char operstate = dev->operstate;
759
760 switch (transition) {
761 case IF_OPER_UP:
762 if ((operstate == IF_OPER_DORMANT ||
763 operstate == IF_OPER_UNKNOWN) &&
764 !netif_dormant(dev))
765 operstate = IF_OPER_UP;
766 break;
767
768 case IF_OPER_DORMANT:
769 if (operstate == IF_OPER_UP ||
770 operstate == IF_OPER_UNKNOWN)
771 operstate = IF_OPER_DORMANT;
772 break;
773 }
774
775 if (dev->operstate != operstate) {
776 write_lock_bh(&dev_base_lock);
777 dev->operstate = operstate;
778 write_unlock_bh(&dev_base_lock);
779 netdev_state_change(dev);
780 }
781 }
782
783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
784 {
785 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
786 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
787 }
788
789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
790 const struct ifinfomsg *ifm)
791 {
792 unsigned int flags = ifm->ifi_flags;
793
794 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
795 if (ifm->ifi_change)
796 flags = (flags & ifm->ifi_change) |
797 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
798
799 return flags;
800 }
801
802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
803 const struct rtnl_link_stats64 *b)
804 {
805 a->rx_packets = b->rx_packets;
806 a->tx_packets = b->tx_packets;
807 a->rx_bytes = b->rx_bytes;
808 a->tx_bytes = b->tx_bytes;
809 a->rx_errors = b->rx_errors;
810 a->tx_errors = b->tx_errors;
811 a->rx_dropped = b->rx_dropped;
812 a->tx_dropped = b->tx_dropped;
813
814 a->multicast = b->multicast;
815 a->collisions = b->collisions;
816
817 a->rx_length_errors = b->rx_length_errors;
818 a->rx_over_errors = b->rx_over_errors;
819 a->rx_crc_errors = b->rx_crc_errors;
820 a->rx_frame_errors = b->rx_frame_errors;
821 a->rx_fifo_errors = b->rx_fifo_errors;
822 a->rx_missed_errors = b->rx_missed_errors;
823
824 a->tx_aborted_errors = b->tx_aborted_errors;
825 a->tx_carrier_errors = b->tx_carrier_errors;
826 a->tx_fifo_errors = b->tx_fifo_errors;
827 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
828 a->tx_window_errors = b->tx_window_errors;
829
830 a->rx_compressed = b->rx_compressed;
831 a->tx_compressed = b->tx_compressed;
832
833 a->rx_nohandler = b->rx_nohandler;
834 }
835
836 /* All VF info */
837 static inline int rtnl_vfinfo_size(const struct net_device *dev,
838 u32 ext_filter_mask)
839 {
840 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
841 int num_vfs = dev_num_vf(dev->dev.parent);
842 size_t size = nla_total_size(0);
843 size += num_vfs *
844 (nla_total_size(0) +
845 nla_total_size(sizeof(struct ifla_vf_mac)) +
846 nla_total_size(sizeof(struct ifla_vf_vlan)) +
847 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
848 nla_total_size(MAX_VLAN_LIST_LEN *
849 sizeof(struct ifla_vf_vlan_info)) +
850 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
851 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
852 nla_total_size(sizeof(struct ifla_vf_rate)) +
853 nla_total_size(sizeof(struct ifla_vf_link_state)) +
854 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
855 nla_total_size(0) + /* nest IFLA_VF_STATS */
856 /* IFLA_VF_STATS_RX_PACKETS */
857 nla_total_size_64bit(sizeof(__u64)) +
858 /* IFLA_VF_STATS_TX_PACKETS */
859 nla_total_size_64bit(sizeof(__u64)) +
860 /* IFLA_VF_STATS_RX_BYTES */
861 nla_total_size_64bit(sizeof(__u64)) +
862 /* IFLA_VF_STATS_TX_BYTES */
863 nla_total_size_64bit(sizeof(__u64)) +
864 /* IFLA_VF_STATS_BROADCAST */
865 nla_total_size_64bit(sizeof(__u64)) +
866 /* IFLA_VF_STATS_MULTICAST */
867 nla_total_size_64bit(sizeof(__u64)) +
868 nla_total_size(sizeof(struct ifla_vf_trust)));
869 return size;
870 } else
871 return 0;
872 }
873
874 static size_t rtnl_port_size(const struct net_device *dev,
875 u32 ext_filter_mask)
876 {
877 size_t port_size = nla_total_size(4) /* PORT_VF */
878 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
879 + nla_total_size(sizeof(struct ifla_port_vsi))
880 /* PORT_VSI_TYPE */
881 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
882 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
883 + nla_total_size(1) /* PROT_VDP_REQUEST */
884 + nla_total_size(2); /* PORT_VDP_RESPONSE */
885 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
886 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
887 + port_size;
888 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
889 + port_size;
890
891 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
892 !(ext_filter_mask & RTEXT_FILTER_VF))
893 return 0;
894 if (dev_num_vf(dev->dev.parent))
895 return port_self_size + vf_ports_size +
896 vf_port_size * dev_num_vf(dev->dev.parent);
897 else
898 return port_self_size;
899 }
900
901 static size_t rtnl_xdp_size(const struct net_device *dev)
902 {
903 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
904 nla_total_size(1); /* XDP_ATTACHED */
905
906 if (!dev->netdev_ops->ndo_xdp)
907 return 0;
908 else
909 return xdp_size;
910 }
911
912 static noinline size_t if_nlmsg_size(const struct net_device *dev,
913 u32 ext_filter_mask)
914 {
915 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
916 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
917 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
918 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
919 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
920 + nla_total_size(sizeof(struct rtnl_link_stats))
921 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
922 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
923 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
924 + nla_total_size(4) /* IFLA_TXQLEN */
925 + nla_total_size(4) /* IFLA_WEIGHT */
926 + nla_total_size(4) /* IFLA_MTU */
927 + nla_total_size(4) /* IFLA_LINK */
928 + nla_total_size(4) /* IFLA_MASTER */
929 + nla_total_size(1) /* IFLA_CARRIER */
930 + nla_total_size(4) /* IFLA_PROMISCUITY */
931 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
932 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
933 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
934 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
935 + nla_total_size(1) /* IFLA_OPERSTATE */
936 + nla_total_size(1) /* IFLA_LINKMODE */
937 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
938 + nla_total_size(4) /* IFLA_LINK_NETNSID */
939 + nla_total_size(ext_filter_mask
940 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
941 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
942 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
943 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
944 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
945 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
946 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
947 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
948 + rtnl_xdp_size(dev) /* IFLA_XDP */
949 + nla_total_size(1); /* IFLA_PROTO_DOWN */
950
951 }
952
953 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
954 {
955 struct nlattr *vf_ports;
956 struct nlattr *vf_port;
957 int vf;
958 int err;
959
960 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
961 if (!vf_ports)
962 return -EMSGSIZE;
963
964 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
965 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
966 if (!vf_port)
967 goto nla_put_failure;
968 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
969 goto nla_put_failure;
970 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
971 if (err == -EMSGSIZE)
972 goto nla_put_failure;
973 if (err) {
974 nla_nest_cancel(skb, vf_port);
975 continue;
976 }
977 nla_nest_end(skb, vf_port);
978 }
979
980 nla_nest_end(skb, vf_ports);
981
982 return 0;
983
984 nla_put_failure:
985 nla_nest_cancel(skb, vf_ports);
986 return -EMSGSIZE;
987 }
988
989 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
990 {
991 struct nlattr *port_self;
992 int err;
993
994 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
995 if (!port_self)
996 return -EMSGSIZE;
997
998 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
999 if (err) {
1000 nla_nest_cancel(skb, port_self);
1001 return (err == -EMSGSIZE) ? err : 0;
1002 }
1003
1004 nla_nest_end(skb, port_self);
1005
1006 return 0;
1007 }
1008
1009 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1010 u32 ext_filter_mask)
1011 {
1012 int err;
1013
1014 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1015 !(ext_filter_mask & RTEXT_FILTER_VF))
1016 return 0;
1017
1018 err = rtnl_port_self_fill(skb, dev);
1019 if (err)
1020 return err;
1021
1022 if (dev_num_vf(dev->dev.parent)) {
1023 err = rtnl_vf_ports_fill(skb, dev);
1024 if (err)
1025 return err;
1026 }
1027
1028 return 0;
1029 }
1030
1031 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1032 {
1033 int err;
1034 struct netdev_phys_item_id ppid;
1035
1036 err = dev_get_phys_port_id(dev, &ppid);
1037 if (err) {
1038 if (err == -EOPNOTSUPP)
1039 return 0;
1040 return err;
1041 }
1042
1043 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1044 return -EMSGSIZE;
1045
1046 return 0;
1047 }
1048
1049 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1050 {
1051 char name[IFNAMSIZ];
1052 int err;
1053
1054 err = dev_get_phys_port_name(dev, name, sizeof(name));
1055 if (err) {
1056 if (err == -EOPNOTSUPP)
1057 return 0;
1058 return err;
1059 }
1060
1061 if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1062 return -EMSGSIZE;
1063
1064 return 0;
1065 }
1066
1067 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1068 {
1069 int err;
1070 struct switchdev_attr attr = {
1071 .orig_dev = dev,
1072 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1073 .flags = SWITCHDEV_F_NO_RECURSE,
1074 };
1075
1076 err = switchdev_port_attr_get(dev, &attr);
1077 if (err) {
1078 if (err == -EOPNOTSUPP)
1079 return 0;
1080 return err;
1081 }
1082
1083 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1084 attr.u.ppid.id))
1085 return -EMSGSIZE;
1086
1087 return 0;
1088 }
1089
1090 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1091 struct net_device *dev)
1092 {
1093 struct rtnl_link_stats64 *sp;
1094 struct nlattr *attr;
1095
1096 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1097 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1098 if (!attr)
1099 return -EMSGSIZE;
1100
1101 sp = nla_data(attr);
1102 dev_get_stats(dev, sp);
1103
1104 attr = nla_reserve(skb, IFLA_STATS,
1105 sizeof(struct rtnl_link_stats));
1106 if (!attr)
1107 return -EMSGSIZE;
1108
1109 copy_rtnl_link_stats(nla_data(attr), sp);
1110
1111 return 0;
1112 }
1113
1114 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1115 struct net_device *dev,
1116 int vfs_num,
1117 struct nlattr *vfinfo)
1118 {
1119 struct ifla_vf_rss_query_en vf_rss_query_en;
1120 struct nlattr *vf, *vfstats, *vfvlanlist;
1121 struct ifla_vf_link_state vf_linkstate;
1122 struct ifla_vf_vlan_info vf_vlan_info;
1123 struct ifla_vf_spoofchk vf_spoofchk;
1124 struct ifla_vf_tx_rate vf_tx_rate;
1125 struct ifla_vf_stats vf_stats;
1126 struct ifla_vf_trust vf_trust;
1127 struct ifla_vf_vlan vf_vlan;
1128 struct ifla_vf_rate vf_rate;
1129 struct ifla_vf_mac vf_mac;
1130 struct ifla_vf_info ivi;
1131
1132 /* Not all SR-IOV capable drivers support the
1133 * spoofcheck and "RSS query enable" query. Preset to
1134 * -1 so the user space tool can detect that the driver
1135 * didn't report anything.
1136 */
1137 ivi.spoofchk = -1;
1138 ivi.rss_query_en = -1;
1139 ivi.trusted = -1;
1140 memset(ivi.mac, 0, sizeof(ivi.mac));
1141 /* The default value for VF link state is "auto"
1142 * IFLA_VF_LINK_STATE_AUTO which equals zero
1143 */
1144 ivi.linkstate = 0;
1145 /* VLAN Protocol by default is 802.1Q */
1146 ivi.vlan_proto = htons(ETH_P_8021Q);
1147 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1148 return 0;
1149
1150 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1151
1152 vf_mac.vf =
1153 vf_vlan.vf =
1154 vf_vlan_info.vf =
1155 vf_rate.vf =
1156 vf_tx_rate.vf =
1157 vf_spoofchk.vf =
1158 vf_linkstate.vf =
1159 vf_rss_query_en.vf =
1160 vf_trust.vf = ivi.vf;
1161
1162 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1163 vf_vlan.vlan = ivi.vlan;
1164 vf_vlan.qos = ivi.qos;
1165 vf_vlan_info.vlan = ivi.vlan;
1166 vf_vlan_info.qos = ivi.qos;
1167 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1168 vf_tx_rate.rate = ivi.max_tx_rate;
1169 vf_rate.min_tx_rate = ivi.min_tx_rate;
1170 vf_rate.max_tx_rate = ivi.max_tx_rate;
1171 vf_spoofchk.setting = ivi.spoofchk;
1172 vf_linkstate.link_state = ivi.linkstate;
1173 vf_rss_query_en.setting = ivi.rss_query_en;
1174 vf_trust.setting = ivi.trusted;
1175 vf = nla_nest_start(skb, IFLA_VF_INFO);
1176 if (!vf)
1177 goto nla_put_vfinfo_failure;
1178 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1179 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1180 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1181 &vf_rate) ||
1182 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1183 &vf_tx_rate) ||
1184 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1185 &vf_spoofchk) ||
1186 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1187 &vf_linkstate) ||
1188 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1189 sizeof(vf_rss_query_en),
1190 &vf_rss_query_en) ||
1191 nla_put(skb, IFLA_VF_TRUST,
1192 sizeof(vf_trust), &vf_trust))
1193 goto nla_put_vf_failure;
1194 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1195 if (!vfvlanlist)
1196 goto nla_put_vf_failure;
1197 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1198 &vf_vlan_info)) {
1199 nla_nest_cancel(skb, vfvlanlist);
1200 goto nla_put_vf_failure;
1201 }
1202 nla_nest_end(skb, vfvlanlist);
1203 memset(&vf_stats, 0, sizeof(vf_stats));
1204 if (dev->netdev_ops->ndo_get_vf_stats)
1205 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1206 &vf_stats);
1207 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1208 if (!vfstats)
1209 goto nla_put_vf_failure;
1210 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1211 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1212 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1213 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1214 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1215 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1216 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1217 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1218 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1219 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1220 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1221 vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1222 nla_nest_cancel(skb, vfstats);
1223 goto nla_put_vf_failure;
1224 }
1225 nla_nest_end(skb, vfstats);
1226 nla_nest_end(skb, vf);
1227 return 0;
1228
1229 nla_put_vf_failure:
1230 nla_nest_cancel(skb, vf);
1231 nla_put_vfinfo_failure:
1232 nla_nest_cancel(skb, vfinfo);
1233 return -EMSGSIZE;
1234 }
1235
1236 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1237 {
1238 struct rtnl_link_ifmap map;
1239
1240 memset(&map, 0, sizeof(map));
1241 map.mem_start = dev->mem_start;
1242 map.mem_end = dev->mem_end;
1243 map.base_addr = dev->base_addr;
1244 map.irq = dev->irq;
1245 map.dma = dev->dma;
1246 map.port = dev->if_port;
1247
1248 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1249 return -EMSGSIZE;
1250
1251 return 0;
1252 }
1253
1254 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1255 {
1256 struct netdev_xdp xdp_op = {};
1257 struct nlattr *xdp;
1258 int err;
1259
1260 if (!dev->netdev_ops->ndo_xdp)
1261 return 0;
1262 xdp = nla_nest_start(skb, IFLA_XDP);
1263 if (!xdp)
1264 return -EMSGSIZE;
1265 xdp_op.command = XDP_QUERY_PROG;
1266 err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1267 if (err)
1268 goto err_cancel;
1269 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1270 if (err)
1271 goto err_cancel;
1272
1273 nla_nest_end(skb, xdp);
1274 return 0;
1275
1276 err_cancel:
1277 nla_nest_cancel(skb, xdp);
1278 return err;
1279 }
1280
1281 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1282 int type, u32 pid, u32 seq, u32 change,
1283 unsigned int flags, u32 ext_filter_mask)
1284 {
1285 struct ifinfomsg *ifm;
1286 struct nlmsghdr *nlh;
1287 struct nlattr *af_spec;
1288 struct rtnl_af_ops *af_ops;
1289 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1290
1291 ASSERT_RTNL();
1292 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1293 if (nlh == NULL)
1294 return -EMSGSIZE;
1295
1296 ifm = nlmsg_data(nlh);
1297 ifm->ifi_family = AF_UNSPEC;
1298 ifm->__ifi_pad = 0;
1299 ifm->ifi_type = dev->type;
1300 ifm->ifi_index = dev->ifindex;
1301 ifm->ifi_flags = dev_get_flags(dev);
1302 ifm->ifi_change = change;
1303
1304 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1305 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1306 nla_put_u8(skb, IFLA_OPERSTATE,
1307 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1308 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1309 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1310 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1311 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1312 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1313 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1314 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1315 #ifdef CONFIG_RPS
1316 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1317 #endif
1318 (dev->ifindex != dev_get_iflink(dev) &&
1319 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1320 (upper_dev &&
1321 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1322 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1323 (dev->qdisc &&
1324 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1325 (dev->ifalias &&
1326 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1327 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1328 atomic_read(&dev->carrier_changes)) ||
1329 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1330 goto nla_put_failure;
1331
1332 if (rtnl_fill_link_ifmap(skb, dev))
1333 goto nla_put_failure;
1334
1335 if (dev->addr_len) {
1336 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1337 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1338 goto nla_put_failure;
1339 }
1340
1341 if (rtnl_phys_port_id_fill(skb, dev))
1342 goto nla_put_failure;
1343
1344 if (rtnl_phys_port_name_fill(skb, dev))
1345 goto nla_put_failure;
1346
1347 if (rtnl_phys_switch_id_fill(skb, dev))
1348 goto nla_put_failure;
1349
1350 if (rtnl_fill_stats(skb, dev))
1351 goto nla_put_failure;
1352
1353 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1354 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1355 goto nla_put_failure;
1356
1357 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1358 ext_filter_mask & RTEXT_FILTER_VF) {
1359 int i;
1360 struct nlattr *vfinfo;
1361 int num_vfs = dev_num_vf(dev->dev.parent);
1362
1363 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1364 if (!vfinfo)
1365 goto nla_put_failure;
1366 for (i = 0; i < num_vfs; i++) {
1367 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1368 goto nla_put_failure;
1369 }
1370
1371 nla_nest_end(skb, vfinfo);
1372 }
1373
1374 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1375 goto nla_put_failure;
1376
1377 if (rtnl_xdp_fill(skb, dev))
1378 goto nla_put_failure;
1379
1380 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1381 if (rtnl_link_fill(skb, dev) < 0)
1382 goto nla_put_failure;
1383 }
1384
1385 if (dev->rtnl_link_ops &&
1386 dev->rtnl_link_ops->get_link_net) {
1387 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1388
1389 if (!net_eq(dev_net(dev), link_net)) {
1390 int id = peernet2id_alloc(dev_net(dev), link_net);
1391
1392 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1393 goto nla_put_failure;
1394 }
1395 }
1396
1397 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1398 goto nla_put_failure;
1399
1400 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1401 if (af_ops->fill_link_af) {
1402 struct nlattr *af;
1403 int err;
1404
1405 if (!(af = nla_nest_start(skb, af_ops->family)))
1406 goto nla_put_failure;
1407
1408 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1409
1410 /*
1411 * Caller may return ENODATA to indicate that there
1412 * was no data to be dumped. This is not an error, it
1413 * means we should trim the attribute header and
1414 * continue.
1415 */
1416 if (err == -ENODATA)
1417 nla_nest_cancel(skb, af);
1418 else if (err < 0)
1419 goto nla_put_failure;
1420
1421 nla_nest_end(skb, af);
1422 }
1423 }
1424
1425 nla_nest_end(skb, af_spec);
1426
1427 nlmsg_end(skb, nlh);
1428 return 0;
1429
1430 nla_put_failure:
1431 nlmsg_cancel(skb, nlh);
1432 return -EMSGSIZE;
1433 }
1434
1435 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1436 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1437 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1438 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1439 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1440 [IFLA_MTU] = { .type = NLA_U32 },
1441 [IFLA_LINK] = { .type = NLA_U32 },
1442 [IFLA_MASTER] = { .type = NLA_U32 },
1443 [IFLA_CARRIER] = { .type = NLA_U8 },
1444 [IFLA_TXQLEN] = { .type = NLA_U32 },
1445 [IFLA_WEIGHT] = { .type = NLA_U32 },
1446 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1447 [IFLA_LINKMODE] = { .type = NLA_U8 },
1448 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1449 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1450 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1451 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1452 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1453 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1454 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1455 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1456 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1457 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1458 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1459 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1460 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1461 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1462 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1463 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1464 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1465 [IFLA_XDP] = { .type = NLA_NESTED },
1466 };
1467
1468 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1469 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1470 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1471 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1472 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1473 };
1474
1475 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1476 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1477 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1478 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1479 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1480 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1481 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1482 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1483 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1484 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1485 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1486 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1487 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1488 };
1489
1490 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1491 [IFLA_PORT_VF] = { .type = NLA_U32 },
1492 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1493 .len = PORT_PROFILE_MAX },
1494 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1495 .len = sizeof(struct ifla_port_vsi)},
1496 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1497 .len = PORT_UUID_MAX },
1498 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1499 .len = PORT_UUID_MAX },
1500 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1501 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1502 };
1503
1504 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1505 [IFLA_XDP_FD] = { .type = NLA_S32 },
1506 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1507 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1508 };
1509
1510 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1511 {
1512 const struct rtnl_link_ops *ops = NULL;
1513 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1514
1515 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1516 return NULL;
1517
1518 if (linfo[IFLA_INFO_KIND]) {
1519 char kind[MODULE_NAME_LEN];
1520
1521 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1522 ops = rtnl_link_ops_get(kind);
1523 }
1524
1525 return ops;
1526 }
1527
1528 static bool link_master_filtered(struct net_device *dev, int master_idx)
1529 {
1530 struct net_device *master;
1531
1532 if (!master_idx)
1533 return false;
1534
1535 master = netdev_master_upper_dev_get(dev);
1536 if (!master || master->ifindex != master_idx)
1537 return true;
1538
1539 return false;
1540 }
1541
1542 static bool link_kind_filtered(const struct net_device *dev,
1543 const struct rtnl_link_ops *kind_ops)
1544 {
1545 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1546 return true;
1547
1548 return false;
1549 }
1550
1551 static bool link_dump_filtered(struct net_device *dev,
1552 int master_idx,
1553 const struct rtnl_link_ops *kind_ops)
1554 {
1555 if (link_master_filtered(dev, master_idx) ||
1556 link_kind_filtered(dev, kind_ops))
1557 return true;
1558
1559 return false;
1560 }
1561
1562 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1563 {
1564 struct net *net = sock_net(skb->sk);
1565 int h, s_h;
1566 int idx = 0, s_idx;
1567 struct net_device *dev;
1568 struct hlist_head *head;
1569 struct nlattr *tb[IFLA_MAX+1];
1570 u32 ext_filter_mask = 0;
1571 const struct rtnl_link_ops *kind_ops = NULL;
1572 unsigned int flags = NLM_F_MULTI;
1573 int master_idx = 0;
1574 int err;
1575 int hdrlen;
1576
1577 s_h = cb->args[0];
1578 s_idx = cb->args[1];
1579
1580 cb->seq = net->dev_base_seq;
1581
1582 /* A hack to preserve kernel<->userspace interface.
1583 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1584 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1585 * what iproute2 < v3.9.0 used.
1586 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1587 * attribute, its netlink message is shorter than struct ifinfomsg.
1588 */
1589 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1590 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1591
1592 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1593
1594 if (tb[IFLA_EXT_MASK])
1595 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1596
1597 if (tb[IFLA_MASTER])
1598 master_idx = nla_get_u32(tb[IFLA_MASTER]);
1599
1600 if (tb[IFLA_LINKINFO])
1601 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1602
1603 if (master_idx || kind_ops)
1604 flags |= NLM_F_DUMP_FILTERED;
1605 }
1606
1607 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1608 idx = 0;
1609 head = &net->dev_index_head[h];
1610 hlist_for_each_entry(dev, head, index_hlist) {
1611 if (link_dump_filtered(dev, master_idx, kind_ops))
1612 goto cont;
1613 if (idx < s_idx)
1614 goto cont;
1615 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1616 NETLINK_CB(cb->skb).portid,
1617 cb->nlh->nlmsg_seq, 0,
1618 flags,
1619 ext_filter_mask);
1620 /* If we ran out of room on the first message,
1621 * we're in trouble
1622 */
1623 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1624
1625 if (err < 0)
1626 goto out;
1627
1628 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1629 cont:
1630 idx++;
1631 }
1632 }
1633 out:
1634 cb->args[1] = idx;
1635 cb->args[0] = h;
1636
1637 return skb->len;
1638 }
1639
1640 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1641 {
1642 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1643 }
1644 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1645
1646 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1647 {
1648 struct net *net;
1649 /* Examine the link attributes and figure out which
1650 * network namespace we are talking about.
1651 */
1652 if (tb[IFLA_NET_NS_PID])
1653 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1654 else if (tb[IFLA_NET_NS_FD])
1655 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1656 else
1657 net = get_net(src_net);
1658 return net;
1659 }
1660 EXPORT_SYMBOL(rtnl_link_get_net);
1661
1662 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1663 {
1664 if (dev) {
1665 if (tb[IFLA_ADDRESS] &&
1666 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1667 return -EINVAL;
1668
1669 if (tb[IFLA_BROADCAST] &&
1670 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1671 return -EINVAL;
1672 }
1673
1674 if (tb[IFLA_AF_SPEC]) {
1675 struct nlattr *af;
1676 int rem, err;
1677
1678 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1679 const struct rtnl_af_ops *af_ops;
1680
1681 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1682 return -EAFNOSUPPORT;
1683
1684 if (!af_ops->set_link_af)
1685 return -EOPNOTSUPP;
1686
1687 if (af_ops->validate_link_af) {
1688 err = af_ops->validate_link_af(dev, af);
1689 if (err < 0)
1690 return err;
1691 }
1692 }
1693 }
1694
1695 return 0;
1696 }
1697
1698 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1699 int guid_type)
1700 {
1701 const struct net_device_ops *ops = dev->netdev_ops;
1702
1703 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1704 }
1705
1706 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1707 {
1708 if (dev->type != ARPHRD_INFINIBAND)
1709 return -EOPNOTSUPP;
1710
1711 return handle_infiniband_guid(dev, ivt, guid_type);
1712 }
1713
1714 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1715 {
1716 const struct net_device_ops *ops = dev->netdev_ops;
1717 int err = -EINVAL;
1718
1719 if (tb[IFLA_VF_MAC]) {
1720 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1721
1722 err = -EOPNOTSUPP;
1723 if (ops->ndo_set_vf_mac)
1724 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1725 ivm->mac);
1726 if (err < 0)
1727 return err;
1728 }
1729
1730 if (tb[IFLA_VF_VLAN]) {
1731 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1732
1733 err = -EOPNOTSUPP;
1734 if (ops->ndo_set_vf_vlan)
1735 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1736 ivv->qos,
1737 htons(ETH_P_8021Q));
1738 if (err < 0)
1739 return err;
1740 }
1741
1742 if (tb[IFLA_VF_VLAN_LIST]) {
1743 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1744 struct nlattr *attr;
1745 int rem, len = 0;
1746
1747 err = -EOPNOTSUPP;
1748 if (!ops->ndo_set_vf_vlan)
1749 return err;
1750
1751 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1752 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1753 nla_len(attr) < NLA_HDRLEN) {
1754 return -EINVAL;
1755 }
1756 if (len >= MAX_VLAN_LIST_LEN)
1757 return -EOPNOTSUPP;
1758 ivvl[len] = nla_data(attr);
1759
1760 len++;
1761 }
1762 if (len == 0)
1763 return -EINVAL;
1764
1765 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1766 ivvl[0]->qos, ivvl[0]->vlan_proto);
1767 if (err < 0)
1768 return err;
1769 }
1770
1771 if (tb[IFLA_VF_TX_RATE]) {
1772 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1773 struct ifla_vf_info ivf;
1774
1775 err = -EOPNOTSUPP;
1776 if (ops->ndo_get_vf_config)
1777 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1778 if (err < 0)
1779 return err;
1780
1781 err = -EOPNOTSUPP;
1782 if (ops->ndo_set_vf_rate)
1783 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1784 ivf.min_tx_rate,
1785 ivt->rate);
1786 if (err < 0)
1787 return err;
1788 }
1789
1790 if (tb[IFLA_VF_RATE]) {
1791 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1792
1793 err = -EOPNOTSUPP;
1794 if (ops->ndo_set_vf_rate)
1795 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1796 ivt->min_tx_rate,
1797 ivt->max_tx_rate);
1798 if (err < 0)
1799 return err;
1800 }
1801
1802 if (tb[IFLA_VF_SPOOFCHK]) {
1803 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1804
1805 err = -EOPNOTSUPP;
1806 if (ops->ndo_set_vf_spoofchk)
1807 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1808 ivs->setting);
1809 if (err < 0)
1810 return err;
1811 }
1812
1813 if (tb[IFLA_VF_LINK_STATE]) {
1814 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1815
1816 err = -EOPNOTSUPP;
1817 if (ops->ndo_set_vf_link_state)
1818 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1819 ivl->link_state);
1820 if (err < 0)
1821 return err;
1822 }
1823
1824 if (tb[IFLA_VF_RSS_QUERY_EN]) {
1825 struct ifla_vf_rss_query_en *ivrssq_en;
1826
1827 err = -EOPNOTSUPP;
1828 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1829 if (ops->ndo_set_vf_rss_query_en)
1830 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1831 ivrssq_en->setting);
1832 if (err < 0)
1833 return err;
1834 }
1835
1836 if (tb[IFLA_VF_TRUST]) {
1837 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1838
1839 err = -EOPNOTSUPP;
1840 if (ops->ndo_set_vf_trust)
1841 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1842 if (err < 0)
1843 return err;
1844 }
1845
1846 if (tb[IFLA_VF_IB_NODE_GUID]) {
1847 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1848
1849 if (!ops->ndo_set_vf_guid)
1850 return -EOPNOTSUPP;
1851
1852 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1853 }
1854
1855 if (tb[IFLA_VF_IB_PORT_GUID]) {
1856 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1857
1858 if (!ops->ndo_set_vf_guid)
1859 return -EOPNOTSUPP;
1860
1861 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1862 }
1863
1864 return err;
1865 }
1866
1867 static int do_set_master(struct net_device *dev, int ifindex)
1868 {
1869 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1870 const struct net_device_ops *ops;
1871 int err;
1872
1873 if (upper_dev) {
1874 if (upper_dev->ifindex == ifindex)
1875 return 0;
1876 ops = upper_dev->netdev_ops;
1877 if (ops->ndo_del_slave) {
1878 err = ops->ndo_del_slave(upper_dev, dev);
1879 if (err)
1880 return err;
1881 } else {
1882 return -EOPNOTSUPP;
1883 }
1884 }
1885
1886 if (ifindex) {
1887 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1888 if (!upper_dev)
1889 return -EINVAL;
1890 ops = upper_dev->netdev_ops;
1891 if (ops->ndo_add_slave) {
1892 err = ops->ndo_add_slave(upper_dev, dev);
1893 if (err)
1894 return err;
1895 } else {
1896 return -EOPNOTSUPP;
1897 }
1898 }
1899 return 0;
1900 }
1901
1902 #define DO_SETLINK_MODIFIED 0x01
1903 /* notify flag means notify + modified. */
1904 #define DO_SETLINK_NOTIFY 0x03
1905 static int do_setlink(const struct sk_buff *skb,
1906 struct net_device *dev, struct ifinfomsg *ifm,
1907 struct nlattr **tb, char *ifname, int status)
1908 {
1909 const struct net_device_ops *ops = dev->netdev_ops;
1910 int err;
1911
1912 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1913 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1914 if (IS_ERR(net)) {
1915 err = PTR_ERR(net);
1916 goto errout;
1917 }
1918 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1919 put_net(net);
1920 err = -EPERM;
1921 goto errout;
1922 }
1923 err = dev_change_net_namespace(dev, net, ifname);
1924 put_net(net);
1925 if (err)
1926 goto errout;
1927 status |= DO_SETLINK_MODIFIED;
1928 }
1929
1930 if (tb[IFLA_MAP]) {
1931 struct rtnl_link_ifmap *u_map;
1932 struct ifmap k_map;
1933
1934 if (!ops->ndo_set_config) {
1935 err = -EOPNOTSUPP;
1936 goto errout;
1937 }
1938
1939 if (!netif_device_present(dev)) {
1940 err = -ENODEV;
1941 goto errout;
1942 }
1943
1944 u_map = nla_data(tb[IFLA_MAP]);
1945 k_map.mem_start = (unsigned long) u_map->mem_start;
1946 k_map.mem_end = (unsigned long) u_map->mem_end;
1947 k_map.base_addr = (unsigned short) u_map->base_addr;
1948 k_map.irq = (unsigned char) u_map->irq;
1949 k_map.dma = (unsigned char) u_map->dma;
1950 k_map.port = (unsigned char) u_map->port;
1951
1952 err = ops->ndo_set_config(dev, &k_map);
1953 if (err < 0)
1954 goto errout;
1955
1956 status |= DO_SETLINK_NOTIFY;
1957 }
1958
1959 if (tb[IFLA_ADDRESS]) {
1960 struct sockaddr *sa;
1961 int len;
1962
1963 len = sizeof(sa_family_t) + dev->addr_len;
1964 sa = kmalloc(len, GFP_KERNEL);
1965 if (!sa) {
1966 err = -ENOMEM;
1967 goto errout;
1968 }
1969 sa->sa_family = dev->type;
1970 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1971 dev->addr_len);
1972 err = dev_set_mac_address(dev, sa);
1973 kfree(sa);
1974 if (err)
1975 goto errout;
1976 status |= DO_SETLINK_MODIFIED;
1977 }
1978
1979 if (tb[IFLA_MTU]) {
1980 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1981 if (err < 0)
1982 goto errout;
1983 status |= DO_SETLINK_MODIFIED;
1984 }
1985
1986 if (tb[IFLA_GROUP]) {
1987 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1988 status |= DO_SETLINK_NOTIFY;
1989 }
1990
1991 /*
1992 * Interface selected by interface index but interface
1993 * name provided implies that a name change has been
1994 * requested.
1995 */
1996 if (ifm->ifi_index > 0 && ifname[0]) {
1997 err = dev_change_name(dev, ifname);
1998 if (err < 0)
1999 goto errout;
2000 status |= DO_SETLINK_MODIFIED;
2001 }
2002
2003 if (tb[IFLA_IFALIAS]) {
2004 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2005 nla_len(tb[IFLA_IFALIAS]));
2006 if (err < 0)
2007 goto errout;
2008 status |= DO_SETLINK_NOTIFY;
2009 }
2010
2011 if (tb[IFLA_BROADCAST]) {
2012 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2013 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2014 }
2015
2016 if (ifm->ifi_flags || ifm->ifi_change) {
2017 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2018 if (err < 0)
2019 goto errout;
2020 }
2021
2022 if (tb[IFLA_MASTER]) {
2023 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2024 if (err)
2025 goto errout;
2026 status |= DO_SETLINK_MODIFIED;
2027 }
2028
2029 if (tb[IFLA_CARRIER]) {
2030 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2031 if (err)
2032 goto errout;
2033 status |= DO_SETLINK_MODIFIED;
2034 }
2035
2036 if (tb[IFLA_TXQLEN]) {
2037 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2038 unsigned long orig_len = dev->tx_queue_len;
2039
2040 if (dev->tx_queue_len ^ value) {
2041 dev->tx_queue_len = value;
2042 err = call_netdevice_notifiers(
2043 NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2044 err = notifier_to_errno(err);
2045 if (err) {
2046 dev->tx_queue_len = orig_len;
2047 goto errout;
2048 }
2049 status |= DO_SETLINK_NOTIFY;
2050 }
2051 }
2052
2053 if (tb[IFLA_OPERSTATE])
2054 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2055
2056 if (tb[IFLA_LINKMODE]) {
2057 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2058
2059 write_lock_bh(&dev_base_lock);
2060 if (dev->link_mode ^ value)
2061 status |= DO_SETLINK_NOTIFY;
2062 dev->link_mode = value;
2063 write_unlock_bh(&dev_base_lock);
2064 }
2065
2066 if (tb[IFLA_VFINFO_LIST]) {
2067 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2068 struct nlattr *attr;
2069 int rem;
2070
2071 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2072 if (nla_type(attr) != IFLA_VF_INFO ||
2073 nla_len(attr) < NLA_HDRLEN) {
2074 err = -EINVAL;
2075 goto errout;
2076 }
2077 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2078 ifla_vf_policy);
2079 if (err < 0)
2080 goto errout;
2081 err = do_setvfinfo(dev, vfinfo);
2082 if (err < 0)
2083 goto errout;
2084 status |= DO_SETLINK_NOTIFY;
2085 }
2086 }
2087 err = 0;
2088
2089 if (tb[IFLA_VF_PORTS]) {
2090 struct nlattr *port[IFLA_PORT_MAX+1];
2091 struct nlattr *attr;
2092 int vf;
2093 int rem;
2094
2095 err = -EOPNOTSUPP;
2096 if (!ops->ndo_set_vf_port)
2097 goto errout;
2098
2099 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2100 if (nla_type(attr) != IFLA_VF_PORT ||
2101 nla_len(attr) < NLA_HDRLEN) {
2102 err = -EINVAL;
2103 goto errout;
2104 }
2105 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2106 ifla_port_policy);
2107 if (err < 0)
2108 goto errout;
2109 if (!port[IFLA_PORT_VF]) {
2110 err = -EOPNOTSUPP;
2111 goto errout;
2112 }
2113 vf = nla_get_u32(port[IFLA_PORT_VF]);
2114 err = ops->ndo_set_vf_port(dev, vf, port);
2115 if (err < 0)
2116 goto errout;
2117 status |= DO_SETLINK_NOTIFY;
2118 }
2119 }
2120 err = 0;
2121
2122 if (tb[IFLA_PORT_SELF]) {
2123 struct nlattr *port[IFLA_PORT_MAX+1];
2124
2125 err = nla_parse_nested(port, IFLA_PORT_MAX,
2126 tb[IFLA_PORT_SELF], ifla_port_policy);
2127 if (err < 0)
2128 goto errout;
2129
2130 err = -EOPNOTSUPP;
2131 if (ops->ndo_set_vf_port)
2132 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2133 if (err < 0)
2134 goto errout;
2135 status |= DO_SETLINK_NOTIFY;
2136 }
2137
2138 if (tb[IFLA_AF_SPEC]) {
2139 struct nlattr *af;
2140 int rem;
2141
2142 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2143 const struct rtnl_af_ops *af_ops;
2144
2145 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2146 BUG();
2147
2148 err = af_ops->set_link_af(dev, af);
2149 if (err < 0)
2150 goto errout;
2151
2152 status |= DO_SETLINK_NOTIFY;
2153 }
2154 }
2155 err = 0;
2156
2157 if (tb[IFLA_PROTO_DOWN]) {
2158 err = dev_change_proto_down(dev,
2159 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2160 if (err)
2161 goto errout;
2162 status |= DO_SETLINK_NOTIFY;
2163 }
2164
2165 if (tb[IFLA_XDP]) {
2166 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2167 u32 xdp_flags = 0;
2168
2169 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2170 ifla_xdp_policy);
2171 if (err < 0)
2172 goto errout;
2173
2174 if (xdp[IFLA_XDP_ATTACHED]) {
2175 err = -EINVAL;
2176 goto errout;
2177 }
2178
2179 if (xdp[IFLA_XDP_FLAGS]) {
2180 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2181 if (xdp_flags & ~XDP_FLAGS_MASK) {
2182 err = -EINVAL;
2183 goto errout;
2184 }
2185 }
2186
2187 if (xdp[IFLA_XDP_FD]) {
2188 err = dev_change_xdp_fd(dev,
2189 nla_get_s32(xdp[IFLA_XDP_FD]),
2190 xdp_flags);
2191 if (err)
2192 goto errout;
2193 status |= DO_SETLINK_NOTIFY;
2194 }
2195 }
2196
2197 errout:
2198 if (status & DO_SETLINK_MODIFIED) {
2199 if (status & DO_SETLINK_NOTIFY)
2200 netdev_state_change(dev);
2201
2202 if (err < 0)
2203 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",
2204 dev->name);
2205 }
2206
2207 return err;
2208 }
2209
2210 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2211 {
2212 struct net *net = sock_net(skb->sk);
2213 struct ifinfomsg *ifm;
2214 struct net_device *dev;
2215 int err;
2216 struct nlattr *tb[IFLA_MAX+1];
2217 char ifname[IFNAMSIZ];
2218
2219 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2220 if (err < 0)
2221 goto errout;
2222
2223 if (tb[IFLA_IFNAME])
2224 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2225 else
2226 ifname[0] = '\0';
2227
2228 err = -EINVAL;
2229 ifm = nlmsg_data(nlh);
2230 if (ifm->ifi_index > 0)
2231 dev = __dev_get_by_index(net, ifm->ifi_index);
2232 else if (tb[IFLA_IFNAME])
2233 dev = __dev_get_by_name(net, ifname);
2234 else
2235 goto errout;
2236
2237 if (dev == NULL) {
2238 err = -ENODEV;
2239 goto errout;
2240 }
2241
2242 err = validate_linkmsg(dev, tb);
2243 if (err < 0)
2244 goto errout;
2245
2246 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2247 errout:
2248 return err;
2249 }
2250
2251 static int rtnl_group_dellink(const struct net *net, int group)
2252 {
2253 struct net_device *dev, *aux;
2254 LIST_HEAD(list_kill);
2255 bool found = false;
2256
2257 if (!group)
2258 return -EPERM;
2259
2260 for_each_netdev(net, dev) {
2261 if (dev->group == group) {
2262 const struct rtnl_link_ops *ops;
2263
2264 found = true;
2265 ops = dev->rtnl_link_ops;
2266 if (!ops || !ops->dellink)
2267 return -EOPNOTSUPP;
2268 }
2269 }
2270
2271 if (!found)
2272 return -ENODEV;
2273
2274 for_each_netdev_safe(net, dev, aux) {
2275 if (dev->group == group) {
2276 const struct rtnl_link_ops *ops;
2277
2278 ops = dev->rtnl_link_ops;
2279 ops->dellink(dev, &list_kill);
2280 }
2281 }
2282 unregister_netdevice_many(&list_kill);
2283
2284 return 0;
2285 }
2286
2287 int rtnl_delete_link(struct net_device *dev)
2288 {
2289 const struct rtnl_link_ops *ops;
2290 LIST_HEAD(list_kill);
2291
2292 ops = dev->rtnl_link_ops;
2293 if (!ops || !ops->dellink)
2294 return -EOPNOTSUPP;
2295
2296 ops->dellink(dev, &list_kill);
2297 unregister_netdevice_many(&list_kill);
2298
2299 return 0;
2300 }
2301 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2302
2303 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2304 {
2305 struct net *net = sock_net(skb->sk);
2306 struct net_device *dev;
2307 struct ifinfomsg *ifm;
2308 char ifname[IFNAMSIZ];
2309 struct nlattr *tb[IFLA_MAX+1];
2310 int err;
2311
2312 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2313 if (err < 0)
2314 return err;
2315
2316 if (tb[IFLA_IFNAME])
2317 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2318
2319 ifm = nlmsg_data(nlh);
2320 if (ifm->ifi_index > 0)
2321 dev = __dev_get_by_index(net, ifm->ifi_index);
2322 else if (tb[IFLA_IFNAME])
2323 dev = __dev_get_by_name(net, ifname);
2324 else if (tb[IFLA_GROUP])
2325 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2326 else
2327 return -EINVAL;
2328
2329 if (!dev)
2330 return -ENODEV;
2331
2332 return rtnl_delete_link(dev);
2333 }
2334
2335 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2336 {
2337 unsigned int old_flags;
2338 int err;
2339
2340 old_flags = dev->flags;
2341 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2342 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2343 if (err < 0)
2344 return err;
2345 }
2346
2347 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2348
2349 __dev_notify_flags(dev, old_flags, ~0U);
2350 return 0;
2351 }
2352 EXPORT_SYMBOL(rtnl_configure_link);
2353
2354 struct net_device *rtnl_create_link(struct net *net,
2355 const char *ifname, unsigned char name_assign_type,
2356 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2357 {
2358 int err;
2359 struct net_device *dev;
2360 unsigned int num_tx_queues = 1;
2361 unsigned int num_rx_queues = 1;
2362
2363 if (tb[IFLA_NUM_TX_QUEUES])
2364 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2365 else if (ops->get_num_tx_queues)
2366 num_tx_queues = ops->get_num_tx_queues();
2367
2368 if (tb[IFLA_NUM_RX_QUEUES])
2369 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2370 else if (ops->get_num_rx_queues)
2371 num_rx_queues = ops->get_num_rx_queues();
2372
2373 err = -ENOMEM;
2374 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2375 ops->setup, num_tx_queues, num_rx_queues);
2376 if (!dev)
2377 goto err;
2378
2379 dev_net_set(dev, net);
2380 dev->rtnl_link_ops = ops;
2381 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2382
2383 if (tb[IFLA_MTU])
2384 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2385 if (tb[IFLA_ADDRESS]) {
2386 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2387 nla_len(tb[IFLA_ADDRESS]));
2388 dev->addr_assign_type = NET_ADDR_SET;
2389 }
2390 if (tb[IFLA_BROADCAST])
2391 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2392 nla_len(tb[IFLA_BROADCAST]));
2393 if (tb[IFLA_TXQLEN])
2394 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2395 if (tb[IFLA_OPERSTATE])
2396 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2397 if (tb[IFLA_LINKMODE])
2398 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2399 if (tb[IFLA_GROUP])
2400 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2401
2402 return dev;
2403
2404 err:
2405 return ERR_PTR(err);
2406 }
2407 EXPORT_SYMBOL(rtnl_create_link);
2408
2409 static int rtnl_group_changelink(const struct sk_buff *skb,
2410 struct net *net, int group,
2411 struct ifinfomsg *ifm,
2412 struct nlattr **tb)
2413 {
2414 struct net_device *dev, *aux;
2415 int err;
2416
2417 for_each_netdev_safe(net, dev, aux) {
2418 if (dev->group == group) {
2419 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2420 if (err < 0)
2421 return err;
2422 }
2423 }
2424
2425 return 0;
2426 }
2427
2428 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2429 {
2430 struct net *net = sock_net(skb->sk);
2431 const struct rtnl_link_ops *ops;
2432 const struct rtnl_link_ops *m_ops = NULL;
2433 struct net_device *dev;
2434 struct net_device *master_dev = NULL;
2435 struct ifinfomsg *ifm;
2436 char kind[MODULE_NAME_LEN];
2437 char ifname[IFNAMSIZ];
2438 struct nlattr *tb[IFLA_MAX+1];
2439 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2440 unsigned char name_assign_type = NET_NAME_USER;
2441 int err;
2442
2443 #ifdef CONFIG_MODULES
2444 replay:
2445 #endif
2446 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2447 if (err < 0)
2448 return err;
2449
2450 if (tb[IFLA_IFNAME])
2451 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2452 else
2453 ifname[0] = '\0';
2454
2455 ifm = nlmsg_data(nlh);
2456 if (ifm->ifi_index > 0)
2457 dev = __dev_get_by_index(net, ifm->ifi_index);
2458 else {
2459 if (ifname[0])
2460 dev = __dev_get_by_name(net, ifname);
2461 else
2462 dev = NULL;
2463 }
2464
2465 if (dev) {
2466 master_dev = netdev_master_upper_dev_get(dev);
2467 if (master_dev)
2468 m_ops = master_dev->rtnl_link_ops;
2469 }
2470
2471 err = validate_linkmsg(dev, tb);
2472 if (err < 0)
2473 return err;
2474
2475 if (tb[IFLA_LINKINFO]) {
2476 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2477 tb[IFLA_LINKINFO], ifla_info_policy);
2478 if (err < 0)
2479 return err;
2480 } else
2481 memset(linkinfo, 0, sizeof(linkinfo));
2482
2483 if (linkinfo[IFLA_INFO_KIND]) {
2484 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2485 ops = rtnl_link_ops_get(kind);
2486 } else {
2487 kind[0] = '\0';
2488 ops = NULL;
2489 }
2490
2491 if (1) {
2492 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2493 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2494 struct nlattr **data = NULL;
2495 struct nlattr **slave_data = NULL;
2496 struct net *dest_net, *link_net = NULL;
2497
2498 if (ops) {
2499 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2500 err = nla_parse_nested(attr, ops->maxtype,
2501 linkinfo[IFLA_INFO_DATA],
2502 ops->policy);
2503 if (err < 0)
2504 return err;
2505 data = attr;
2506 }
2507 if (ops->validate) {
2508 err = ops->validate(tb, data);
2509 if (err < 0)
2510 return err;
2511 }
2512 }
2513
2514 if (m_ops) {
2515 if (m_ops->slave_maxtype &&
2516 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2517 err = nla_parse_nested(slave_attr,
2518 m_ops->slave_maxtype,
2519 linkinfo[IFLA_INFO_SLAVE_DATA],
2520 m_ops->slave_policy);
2521 if (err < 0)
2522 return err;
2523 slave_data = slave_attr;
2524 }
2525 if (m_ops->slave_validate) {
2526 err = m_ops->slave_validate(tb, slave_data);
2527 if (err < 0)
2528 return err;
2529 }
2530 }
2531
2532 if (dev) {
2533 int status = 0;
2534
2535 if (nlh->nlmsg_flags & NLM_F_EXCL)
2536 return -EEXIST;
2537 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2538 return -EOPNOTSUPP;
2539
2540 if (linkinfo[IFLA_INFO_DATA]) {
2541 if (!ops || ops != dev->rtnl_link_ops ||
2542 !ops->changelink)
2543 return -EOPNOTSUPP;
2544
2545 err = ops->changelink(dev, tb, data);
2546 if (err < 0)
2547 return err;
2548 status |= DO_SETLINK_NOTIFY;
2549 }
2550
2551 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2552 if (!m_ops || !m_ops->slave_changelink)
2553 return -EOPNOTSUPP;
2554
2555 err = m_ops->slave_changelink(master_dev, dev,
2556 tb, slave_data);
2557 if (err < 0)
2558 return err;
2559 status |= DO_SETLINK_NOTIFY;
2560 }
2561
2562 return do_setlink(skb, dev, ifm, tb, ifname, status);
2563 }
2564
2565 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2566 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2567 return rtnl_group_changelink(skb, net,
2568 nla_get_u32(tb[IFLA_GROUP]),
2569 ifm, tb);
2570 return -ENODEV;
2571 }
2572
2573 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2574 return -EOPNOTSUPP;
2575
2576 if (!ops) {
2577 #ifdef CONFIG_MODULES
2578 if (kind[0]) {
2579 __rtnl_unlock();
2580 request_module("rtnl-link-%s", kind);
2581 rtnl_lock();
2582 ops = rtnl_link_ops_get(kind);
2583 if (ops)
2584 goto replay;
2585 }
2586 #endif
2587 return -EOPNOTSUPP;
2588 }
2589
2590 if (!ops->setup)
2591 return -EOPNOTSUPP;
2592
2593 if (!ifname[0]) {
2594 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2595 name_assign_type = NET_NAME_ENUM;
2596 }
2597
2598 dest_net = rtnl_link_get_net(net, tb);
2599 if (IS_ERR(dest_net))
2600 return PTR_ERR(dest_net);
2601
2602 err = -EPERM;
2603 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2604 goto out;
2605
2606 if (tb[IFLA_LINK_NETNSID]) {
2607 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2608
2609 link_net = get_net_ns_by_id(dest_net, id);
2610 if (!link_net) {
2611 err = -EINVAL;
2612 goto out;
2613 }
2614 err = -EPERM;
2615 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2616 goto out;
2617 }
2618
2619 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2620 name_assign_type, ops, tb);
2621 if (IS_ERR(dev)) {
2622 err = PTR_ERR(dev);
2623 goto out;
2624 }
2625
2626 dev->ifindex = ifm->ifi_index;
2627
2628 if (ops->newlink) {
2629 err = ops->newlink(link_net ? : net, dev, tb, data);
2630 /* Drivers should call free_netdev() in ->destructor
2631 * and unregister it on failure after registration
2632 * so that device could be finally freed in rtnl_unlock.
2633 */
2634 if (err < 0) {
2635 /* If device is not registered at all, free it now */
2636 if (dev->reg_state == NETREG_UNINITIALIZED)
2637 free_netdev(dev);
2638 goto out;
2639 }
2640 } else {
2641 err = register_netdevice(dev);
2642 if (err < 0) {
2643 free_netdev(dev);
2644 goto out;
2645 }
2646 }
2647 err = rtnl_configure_link(dev, ifm);
2648 if (err < 0)
2649 goto out_unregister;
2650 if (link_net) {
2651 err = dev_change_net_namespace(dev, dest_net, ifname);
2652 if (err < 0)
2653 goto out_unregister;
2654 }
2655 if (tb[IFLA_MASTER]) {
2656 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2657 if (err)
2658 goto out_unregister;
2659 }
2660 out:
2661 if (link_net)
2662 put_net(link_net);
2663 put_net(dest_net);
2664 return err;
2665 out_unregister:
2666 if (ops->newlink) {
2667 LIST_HEAD(list_kill);
2668
2669 ops->dellink(dev, &list_kill);
2670 unregister_netdevice_many(&list_kill);
2671 } else {
2672 unregister_netdevice(dev);
2673 }
2674 goto out;
2675 }
2676 }
2677
2678 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2679 {
2680 struct net *net = sock_net(skb->sk);
2681 struct ifinfomsg *ifm;
2682 char ifname[IFNAMSIZ];
2683 struct nlattr *tb[IFLA_MAX+1];
2684 struct net_device *dev = NULL;
2685 struct sk_buff *nskb;
2686 int err;
2687 u32 ext_filter_mask = 0;
2688
2689 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2690 if (err < 0)
2691 return err;
2692
2693 if (tb[IFLA_IFNAME])
2694 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2695
2696 if (tb[IFLA_EXT_MASK])
2697 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2698
2699 ifm = nlmsg_data(nlh);
2700 if (ifm->ifi_index > 0)
2701 dev = __dev_get_by_index(net, ifm->ifi_index);
2702 else if (tb[IFLA_IFNAME])
2703 dev = __dev_get_by_name(net, ifname);
2704 else
2705 return -EINVAL;
2706
2707 if (dev == NULL)
2708 return -ENODEV;
2709
2710 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2711 if (nskb == NULL)
2712 return -ENOBUFS;
2713
2714 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2715 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2716 if (err < 0) {
2717 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2718 WARN_ON(err == -EMSGSIZE);
2719 kfree_skb(nskb);
2720 } else
2721 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2722
2723 return err;
2724 }
2725
2726 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2727 {
2728 struct net *net = sock_net(skb->sk);
2729 struct net_device *dev;
2730 struct nlattr *tb[IFLA_MAX+1];
2731 u32 ext_filter_mask = 0;
2732 u16 min_ifinfo_dump_size = 0;
2733 int hdrlen;
2734
2735 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2736 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2737 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2738
2739 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2740 if (tb[IFLA_EXT_MASK])
2741 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2742 }
2743
2744 if (!ext_filter_mask)
2745 return NLMSG_GOODSIZE;
2746 /*
2747 * traverse the list of net devices and compute the minimum
2748 * buffer size based upon the filter mask.
2749 */
2750 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2751 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2752 if_nlmsg_size(dev,
2753 ext_filter_mask));
2754 }
2755
2756 return nlmsg_total_size(min_ifinfo_dump_size);
2757 }
2758
2759 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2760 {
2761 int idx;
2762 int s_idx = cb->family;
2763
2764 if (s_idx == 0)
2765 s_idx = 1;
2766 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2767 int type = cb->nlh->nlmsg_type-RTM_BASE;
2768 if (idx < s_idx || idx == PF_PACKET)
2769 continue;
2770 if (rtnl_msg_handlers[idx] == NULL ||
2771 rtnl_msg_handlers[idx][type].dumpit == NULL)
2772 continue;
2773 if (idx > s_idx) {
2774 memset(&cb->args[0], 0, sizeof(cb->args));
2775 cb->prev_seq = 0;
2776 cb->seq = 0;
2777 }
2778 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2779 break;
2780 }
2781 cb->family = idx;
2782
2783 return skb->len;
2784 }
2785
2786 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2787 unsigned int change, gfp_t flags)
2788 {
2789 struct net *net = dev_net(dev);
2790 struct sk_buff *skb;
2791 int err = -ENOBUFS;
2792 size_t if_info_size;
2793
2794 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2795 if (skb == NULL)
2796 goto errout;
2797
2798 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2799 if (err < 0) {
2800 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2801 WARN_ON(err == -EMSGSIZE);
2802 kfree_skb(skb);
2803 goto errout;
2804 }
2805 return skb;
2806 errout:
2807 if (err < 0)
2808 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2809 return NULL;
2810 }
2811
2812 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2813 {
2814 struct net *net = dev_net(dev);
2815
2816 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2817 }
2818
2819 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2820 gfp_t flags)
2821 {
2822 struct sk_buff *skb;
2823
2824 if (dev->reg_state != NETREG_REGISTERED)
2825 return;
2826
2827 skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2828 if (skb)
2829 rtmsg_ifinfo_send(skb, dev, flags);
2830 }
2831 EXPORT_SYMBOL(rtmsg_ifinfo);
2832
2833 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2834 struct net_device *dev,
2835 u8 *addr, u16 vid, u32 pid, u32 seq,
2836 int type, unsigned int flags,
2837 int nlflags, u16 ndm_state)
2838 {
2839 struct nlmsghdr *nlh;
2840 struct ndmsg *ndm;
2841
2842 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2843 if (!nlh)
2844 return -EMSGSIZE;
2845
2846 ndm = nlmsg_data(nlh);
2847 ndm->ndm_family = AF_BRIDGE;
2848 ndm->ndm_pad1 = 0;
2849 ndm->ndm_pad2 = 0;
2850 ndm->ndm_flags = flags;
2851 ndm->ndm_type = 0;
2852 ndm->ndm_ifindex = dev->ifindex;
2853 ndm->ndm_state = ndm_state;
2854
2855 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2856 goto nla_put_failure;
2857 if (vid)
2858 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2859 goto nla_put_failure;
2860
2861 nlmsg_end(skb, nlh);
2862 return 0;
2863
2864 nla_put_failure:
2865 nlmsg_cancel(skb, nlh);
2866 return -EMSGSIZE;
2867 }
2868
2869 static inline size_t rtnl_fdb_nlmsg_size(void)
2870 {
2871 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2872 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
2873 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
2874 0;
2875 }
2876
2877 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2878 u16 ndm_state)
2879 {
2880 struct net *net = dev_net(dev);
2881 struct sk_buff *skb;
2882 int err = -ENOBUFS;
2883
2884 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2885 if (!skb)
2886 goto errout;
2887
2888 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2889 0, 0, type, NTF_SELF, 0, ndm_state);
2890 if (err < 0) {
2891 kfree_skb(skb);
2892 goto errout;
2893 }
2894
2895 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2896 return;
2897 errout:
2898 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2899 }
2900
2901 /**
2902 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2903 */
2904 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2905 struct nlattr *tb[],
2906 struct net_device *dev,
2907 const unsigned char *addr, u16 vid,
2908 u16 flags)
2909 {
2910 int err = -EINVAL;
2911
2912 /* If aging addresses are supported device will need to
2913 * implement its own handler for this.
2914 */
2915 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2916 pr_info("%s: FDB only supports static addresses\n", dev->name);
2917 return err;
2918 }
2919
2920 if (vid) {
2921 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2922 return err;
2923 }
2924
2925 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2926 err = dev_uc_add_excl(dev, addr);
2927 else if (is_multicast_ether_addr(addr))
2928 err = dev_mc_add_excl(dev, addr);
2929
2930 /* Only return duplicate errors if NLM_F_EXCL is set */
2931 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2932 err = 0;
2933
2934 return err;
2935 }
2936 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2937
2938 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2939 {
2940 u16 vid = 0;
2941
2942 if (vlan_attr) {
2943 if (nla_len(vlan_attr) != sizeof(u16)) {
2944 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2945 return -EINVAL;
2946 }
2947
2948 vid = nla_get_u16(vlan_attr);
2949
2950 if (!vid || vid >= VLAN_VID_MASK) {
2951 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2952 vid);
2953 return -EINVAL;
2954 }
2955 }
2956 *p_vid = vid;
2957 return 0;
2958 }
2959
2960 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2961 {
2962 struct net *net = sock_net(skb->sk);
2963 struct ndmsg *ndm;
2964 struct nlattr *tb[NDA_MAX+1];
2965 struct net_device *dev;
2966 u8 *addr;
2967 u16 vid;
2968 int err;
2969
2970 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2971 if (err < 0)
2972 return err;
2973
2974 ndm = nlmsg_data(nlh);
2975 if (ndm->ndm_ifindex == 0) {
2976 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2977 return -EINVAL;
2978 }
2979
2980 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2981 if (dev == NULL) {
2982 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2983 return -ENODEV;
2984 }
2985
2986 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2987 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2988 return -EINVAL;
2989 }
2990
2991 addr = nla_data(tb[NDA_LLADDR]);
2992
2993 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2994 if (err)
2995 return err;
2996
2997 err = -EOPNOTSUPP;
2998
2999 /* Support fdb on master device the net/bridge default case */
3000 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3001 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3002 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3003 const struct net_device_ops *ops = br_dev->netdev_ops;
3004
3005 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3006 nlh->nlmsg_flags);
3007 if (err)
3008 goto out;
3009 else
3010 ndm->ndm_flags &= ~NTF_MASTER;
3011 }
3012
3013 /* Embedded bridge, macvlan, and any other device support */
3014 if ((ndm->ndm_flags & NTF_SELF)) {
3015 if (dev->netdev_ops->ndo_fdb_add)
3016 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3017 vid,
3018 nlh->nlmsg_flags);
3019 else
3020 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3021 nlh->nlmsg_flags);
3022
3023 if (!err) {
3024 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3025 ndm->ndm_state);
3026 ndm->ndm_flags &= ~NTF_SELF;
3027 }
3028 }
3029 out:
3030 return err;
3031 }
3032
3033 /**
3034 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3035 */
3036 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3037 struct nlattr *tb[],
3038 struct net_device *dev,
3039 const unsigned char *addr, u16 vid)
3040 {
3041 int err = -EINVAL;
3042
3043 /* If aging addresses are supported device will need to
3044 * implement its own handler for this.
3045 */
3046 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3047 pr_info("%s: FDB only supports static addresses\n", dev->name);
3048 return err;
3049 }
3050
3051 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3052 err = dev_uc_del(dev, addr);
3053 else if (is_multicast_ether_addr(addr))
3054 err = dev_mc_del(dev, addr);
3055
3056 return err;
3057 }
3058 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3059
3060 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3061 {
3062 struct net *net = sock_net(skb->sk);
3063 struct ndmsg *ndm;
3064 struct nlattr *tb[NDA_MAX+1];
3065 struct net_device *dev;
3066 int err = -EINVAL;
3067 __u8 *addr;
3068 u16 vid;
3069
3070 if (!netlink_capable(skb, CAP_NET_ADMIN))
3071 return -EPERM;
3072
3073 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3074 if (err < 0)
3075 return err;
3076
3077 ndm = nlmsg_data(nlh);
3078 if (ndm->ndm_ifindex == 0) {
3079 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3080 return -EINVAL;
3081 }
3082
3083 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3084 if (dev == NULL) {
3085 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3086 return -ENODEV;
3087 }
3088
3089 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3090 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3091 return -EINVAL;
3092 }
3093
3094 addr = nla_data(tb[NDA_LLADDR]);
3095
3096 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3097 if (err)
3098 return err;
3099
3100 err = -EOPNOTSUPP;
3101
3102 /* Support fdb on master device the net/bridge default case */
3103 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3104 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3105 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3106 const struct net_device_ops *ops = br_dev->netdev_ops;
3107
3108 if (ops->ndo_fdb_del)
3109 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3110
3111 if (err)
3112 goto out;
3113 else
3114 ndm->ndm_flags &= ~NTF_MASTER;
3115 }
3116
3117 /* Embedded bridge, macvlan, and any other device support */
3118 if (ndm->ndm_flags & NTF_SELF) {
3119 if (dev->netdev_ops->ndo_fdb_del)
3120 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3121 vid);
3122 else
3123 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3124
3125 if (!err) {
3126 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3127 ndm->ndm_state);
3128 ndm->ndm_flags &= ~NTF_SELF;
3129 }
3130 }
3131 out:
3132 return err;
3133 }
3134
3135 static int nlmsg_populate_fdb(struct sk_buff *skb,
3136 struct netlink_callback *cb,
3137 struct net_device *dev,
3138 int *idx,
3139 struct netdev_hw_addr_list *list)
3140 {
3141 struct netdev_hw_addr *ha;
3142 int err;
3143 u32 portid, seq;
3144
3145 portid = NETLINK_CB(cb->skb).portid;
3146 seq = cb->nlh->nlmsg_seq;
3147
3148 list_for_each_entry(ha, &list->list, list) {
3149 if (*idx < cb->args[2])
3150 goto skip;
3151
3152 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3153 portid, seq,
3154 RTM_NEWNEIGH, NTF_SELF,
3155 NLM_F_MULTI, NUD_PERMANENT);
3156 if (err < 0)
3157 return err;
3158 skip:
3159 *idx += 1;
3160 }
3161 return 0;
3162 }
3163
3164 /**
3165 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3166 * @nlh: netlink message header
3167 * @dev: netdevice
3168 *
3169 * Default netdevice operation to dump the existing unicast address list.
3170 * Returns number of addresses from list put in skb.
3171 */
3172 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3173 struct netlink_callback *cb,
3174 struct net_device *dev,
3175 struct net_device *filter_dev,
3176 int *idx)
3177 {
3178 int err;
3179
3180 netif_addr_lock_bh(dev);
3181 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3182 if (err)
3183 goto out;
3184 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3185 out:
3186 netif_addr_unlock_bh(dev);
3187 return err;
3188 }
3189 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3190
3191 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3192 {
3193 struct net_device *dev;
3194 struct nlattr *tb[IFLA_MAX+1];
3195 struct net_device *br_dev = NULL;
3196 const struct net_device_ops *ops = NULL;
3197 const struct net_device_ops *cops = NULL;
3198 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3199 struct net *net = sock_net(skb->sk);
3200 struct hlist_head *head;
3201 int brport_idx = 0;
3202 int br_idx = 0;
3203 int h, s_h;
3204 int idx = 0, s_idx;
3205 int err = 0;
3206 int fidx = 0;
3207
3208 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3209 ifla_policy) == 0) {
3210 if (tb[IFLA_MASTER])
3211 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3212 }
3213
3214 brport_idx = ifm->ifi_index;
3215
3216 if (br_idx) {
3217 br_dev = __dev_get_by_index(net, br_idx);
3218 if (!br_dev)
3219 return -ENODEV;
3220
3221 ops = br_dev->netdev_ops;
3222 }
3223
3224 s_h = cb->args[0];
3225 s_idx = cb->args[1];
3226
3227 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3228 idx = 0;
3229 head = &net->dev_index_head[h];
3230 hlist_for_each_entry(dev, head, index_hlist) {
3231
3232 if (brport_idx && (dev->ifindex != brport_idx))
3233 continue;
3234
3235 if (!br_idx) { /* user did not specify a specific bridge */
3236 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3237 br_dev = netdev_master_upper_dev_get(dev);
3238 cops = br_dev->netdev_ops;
3239 }
3240 } else {
3241 if (dev != br_dev &&
3242 !(dev->priv_flags & IFF_BRIDGE_PORT))
3243 continue;
3244
3245 if (br_dev != netdev_master_upper_dev_get(dev) &&
3246 !(dev->priv_flags & IFF_EBRIDGE))
3247 continue;
3248 cops = ops;
3249 }
3250
3251 if (idx < s_idx)
3252 goto cont;
3253
3254 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3255 if (cops && cops->ndo_fdb_dump) {
3256 err = cops->ndo_fdb_dump(skb, cb,
3257 br_dev, dev,
3258 &fidx);
3259 if (err == -EMSGSIZE)
3260 goto out;
3261 }
3262 }
3263
3264 if (dev->netdev_ops->ndo_fdb_dump)
3265 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3266 dev, NULL,
3267 &fidx);
3268 else
3269 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3270 &fidx);
3271 if (err == -EMSGSIZE)
3272 goto out;
3273
3274 cops = NULL;
3275
3276 /* reset fdb offset to 0 for rest of the interfaces */
3277 cb->args[2] = 0;
3278 fidx = 0;
3279 cont:
3280 idx++;
3281 }
3282 }
3283
3284 out:
3285 cb->args[0] = h;
3286 cb->args[1] = idx;
3287 cb->args[2] = fidx;
3288
3289 return skb->len;
3290 }
3291
3292 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3293 unsigned int attrnum, unsigned int flag)
3294 {
3295 if (mask & flag)
3296 return nla_put_u8(skb, attrnum, !!(flags & flag));
3297 return 0;
3298 }
3299
3300 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3301 struct net_device *dev, u16 mode,
3302 u32 flags, u32 mask, int nlflags,
3303 u32 filter_mask,
3304 int (*vlan_fill)(struct sk_buff *skb,
3305 struct net_device *dev,
3306 u32 filter_mask))
3307 {
3308 struct nlmsghdr *nlh;
3309 struct ifinfomsg *ifm;
3310 struct nlattr *br_afspec;
3311 struct nlattr *protinfo;
3312 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3313 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3314 int err = 0;
3315
3316 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3317 if (nlh == NULL)
3318 return -EMSGSIZE;
3319
3320 ifm = nlmsg_data(nlh);
3321 ifm->ifi_family = AF_BRIDGE;
3322 ifm->__ifi_pad = 0;
3323 ifm->ifi_type = dev->type;
3324 ifm->ifi_index = dev->ifindex;
3325 ifm->ifi_flags = dev_get_flags(dev);
3326 ifm->ifi_change = 0;
3327
3328
3329 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3330 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3331 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3332 (br_dev &&
3333 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3334 (dev->addr_len &&
3335 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3336 (dev->ifindex != dev_get_iflink(dev) &&
3337 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3338 goto nla_put_failure;
3339
3340 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3341 if (!br_afspec)
3342 goto nla_put_failure;
3343
3344 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3345 nla_nest_cancel(skb, br_afspec);
3346 goto nla_put_failure;
3347 }
3348
3349 if (mode != BRIDGE_MODE_UNDEF) {
3350 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3351 nla_nest_cancel(skb, br_afspec);
3352 goto nla_put_failure;
3353 }
3354 }
3355 if (vlan_fill) {
3356 err = vlan_fill(skb, dev, filter_mask);
3357 if (err) {
3358 nla_nest_cancel(skb, br_afspec);
3359 goto nla_put_failure;
3360 }
3361 }
3362 nla_nest_end(skb, br_afspec);
3363
3364 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3365 if (!protinfo)
3366 goto nla_put_failure;
3367
3368 if (brport_nla_put_flag(skb, flags, mask,
3369 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3370 brport_nla_put_flag(skb, flags, mask,
3371 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3372 brport_nla_put_flag(skb, flags, mask,
3373 IFLA_BRPORT_FAST_LEAVE,
3374 BR_MULTICAST_FAST_LEAVE) ||
3375 brport_nla_put_flag(skb, flags, mask,
3376 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3377 brport_nla_put_flag(skb, flags, mask,
3378 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3379 brport_nla_put_flag(skb, flags, mask,
3380 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3381 brport_nla_put_flag(skb, flags, mask,
3382 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3383 brport_nla_put_flag(skb, flags, mask,
3384 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3385 nla_nest_cancel(skb, protinfo);
3386 goto nla_put_failure;
3387 }
3388
3389 nla_nest_end(skb, protinfo);
3390
3391 nlmsg_end(skb, nlh);
3392 return 0;
3393 nla_put_failure:
3394 nlmsg_cancel(skb, nlh);
3395 return err ? err : -EMSGSIZE;
3396 }
3397 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3398
3399 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3400 {
3401 struct net *net = sock_net(skb->sk);
3402 struct net_device *dev;
3403 int idx = 0;
3404 u32 portid = NETLINK_CB(cb->skb).portid;
3405 u32 seq = cb->nlh->nlmsg_seq;
3406 u32 filter_mask = 0;
3407 int err;
3408
3409 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3410 struct nlattr *extfilt;
3411
3412 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3413 IFLA_EXT_MASK);
3414 if (extfilt) {
3415 if (nla_len(extfilt) < sizeof(filter_mask))
3416 return -EINVAL;
3417
3418 filter_mask = nla_get_u32(extfilt);
3419 }
3420 }
3421
3422 rcu_read_lock();
3423 for_each_netdev_rcu(net, dev) {
3424 const struct net_device_ops *ops = dev->netdev_ops;
3425 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3426
3427 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3428 if (idx >= cb->args[0]) {
3429 err = br_dev->netdev_ops->ndo_bridge_getlink(
3430 skb, portid, seq, dev,
3431 filter_mask, NLM_F_MULTI);
3432 if (err < 0 && err != -EOPNOTSUPP)
3433 break;
3434 }
3435 idx++;
3436 }
3437
3438 if (ops->ndo_bridge_getlink) {
3439 if (idx >= cb->args[0]) {
3440 err = ops->ndo_bridge_getlink(skb, portid,
3441 seq, dev,
3442 filter_mask,
3443 NLM_F_MULTI);
3444 if (err < 0 && err != -EOPNOTSUPP)
3445 break;
3446 }
3447 idx++;
3448 }
3449 }
3450 rcu_read_unlock();
3451 cb->args[0] = idx;
3452
3453 return skb->len;
3454 }
3455
3456 static inline size_t bridge_nlmsg_size(void)
3457 {
3458 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3459 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3460 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3461 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3462 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3463 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3464 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3465 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3466 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3467 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3468 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3469 }
3470
3471 static int rtnl_bridge_notify(struct net_device *dev)
3472 {
3473 struct net *net = dev_net(dev);
3474 struct sk_buff *skb;
3475 int err = -EOPNOTSUPP;
3476
3477 if (!dev->netdev_ops->ndo_bridge_getlink)
3478 return 0;
3479
3480 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3481 if (!skb) {
3482 err = -ENOMEM;
3483 goto errout;
3484 }
3485
3486 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3487 if (err < 0)
3488 goto errout;
3489
3490 if (!skb->len)
3491 goto errout;
3492
3493 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3494 return 0;
3495 errout:
3496 WARN_ON(err == -EMSGSIZE);
3497 kfree_skb(skb);
3498 if (err)
3499 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3500 return err;
3501 }
3502
3503 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3504 {
3505 struct net *net = sock_net(skb->sk);
3506 struct ifinfomsg *ifm;
3507 struct net_device *dev;
3508 struct nlattr *br_spec, *attr = NULL;
3509 int rem, err = -EOPNOTSUPP;
3510 u16 flags = 0;
3511 bool have_flags = false;
3512
3513 if (nlmsg_len(nlh) < sizeof(*ifm))
3514 return -EINVAL;
3515
3516 ifm = nlmsg_data(nlh);
3517 if (ifm->ifi_family != AF_BRIDGE)
3518 return -EPFNOSUPPORT;
3519
3520 dev = __dev_get_by_index(net, ifm->ifi_index);
3521 if (!dev) {
3522 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3523 return -ENODEV;
3524 }
3525
3526 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3527 if (br_spec) {
3528 nla_for_each_nested(attr, br_spec, rem) {
3529 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3530 if (nla_len(attr) < sizeof(flags))
3531 return -EINVAL;
3532
3533 have_flags = true;
3534 flags = nla_get_u16(attr);
3535 break;
3536 }
3537 }
3538 }
3539
3540 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3541 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3542
3543 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3544 err = -EOPNOTSUPP;
3545 goto out;
3546 }
3547
3548 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3549 if (err)
3550 goto out;
3551
3552 flags &= ~BRIDGE_FLAGS_MASTER;
3553 }
3554
3555 if ((flags & BRIDGE_FLAGS_SELF)) {
3556 if (!dev->netdev_ops->ndo_bridge_setlink)
3557 err = -EOPNOTSUPP;
3558 else
3559 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3560 flags);
3561 if (!err) {
3562 flags &= ~BRIDGE_FLAGS_SELF;
3563
3564 /* Generate event to notify upper layer of bridge
3565 * change
3566 */
3567 err = rtnl_bridge_notify(dev);
3568 }
3569 }
3570
3571 if (have_flags)
3572 memcpy(nla_data(attr), &flags, sizeof(flags));
3573 out:
3574 return err;
3575 }
3576
3577 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3578 {
3579 struct net *net = sock_net(skb->sk);
3580 struct ifinfomsg *ifm;
3581 struct net_device *dev;
3582 struct nlattr *br_spec, *attr = NULL;
3583 int rem, err = -EOPNOTSUPP;
3584 u16 flags = 0;
3585 bool have_flags = false;
3586
3587 if (nlmsg_len(nlh) < sizeof(*ifm))
3588 return -EINVAL;
3589
3590 ifm = nlmsg_data(nlh);
3591 if (ifm->ifi_family != AF_BRIDGE)
3592 return -EPFNOSUPPORT;
3593
3594 dev = __dev_get_by_index(net, ifm->ifi_index);
3595 if (!dev) {
3596 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3597 return -ENODEV;
3598 }
3599
3600 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3601 if (br_spec) {
3602 nla_for_each_nested(attr, br_spec, rem) {
3603 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3604 if (nla_len(attr) < sizeof(flags))
3605 return -EINVAL;
3606
3607 have_flags = true;
3608 flags = nla_get_u16(attr);
3609 break;
3610 }
3611 }
3612 }
3613
3614 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3615 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3616
3617 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3618 err = -EOPNOTSUPP;
3619 goto out;
3620 }
3621
3622 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3623 if (err)
3624 goto out;
3625
3626 flags &= ~BRIDGE_FLAGS_MASTER;
3627 }
3628
3629 if ((flags & BRIDGE_FLAGS_SELF)) {
3630 if (!dev->netdev_ops->ndo_bridge_dellink)
3631 err = -EOPNOTSUPP;
3632 else
3633 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3634 flags);
3635
3636 if (!err) {
3637 flags &= ~BRIDGE_FLAGS_SELF;
3638
3639 /* Generate event to notify upper layer of bridge
3640 * change
3641 */
3642 err = rtnl_bridge_notify(dev);
3643 }
3644 }
3645
3646 if (have_flags)
3647 memcpy(nla_data(attr), &flags, sizeof(flags));
3648 out:
3649 return err;
3650 }
3651
3652 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3653 {
3654 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3655 (!idxattr || idxattr == attrid);
3656 }
3657
3658 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3659 static int rtnl_get_offload_stats_attr_size(int attr_id)
3660 {
3661 switch (attr_id) {
3662 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3663 return sizeof(struct rtnl_link_stats64);
3664 }
3665
3666 return 0;
3667 }
3668
3669 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3670 int *prividx)
3671 {
3672 struct nlattr *attr = NULL;
3673 int attr_id, size;
3674 void *attr_data;
3675 int err;
3676
3677 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3678 dev->netdev_ops->ndo_get_offload_stats))
3679 return -ENODATA;
3680
3681 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3682 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3683 if (attr_id < *prividx)
3684 continue;
3685
3686 size = rtnl_get_offload_stats_attr_size(attr_id);
3687 if (!size)
3688 continue;
3689
3690 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3691 continue;
3692
3693 attr = nla_reserve_64bit(skb, attr_id, size,
3694 IFLA_OFFLOAD_XSTATS_UNSPEC);
3695 if (!attr)
3696 goto nla_put_failure;
3697
3698 attr_data = nla_data(attr);
3699 memset(attr_data, 0, size);
3700 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3701 attr_data);
3702 if (err)
3703 goto get_offload_stats_failure;
3704 }
3705
3706 if (!attr)
3707 return -ENODATA;
3708
3709 *prividx = 0;
3710 return 0;
3711
3712 nla_put_failure:
3713 err = -EMSGSIZE;
3714 get_offload_stats_failure:
3715 *prividx = attr_id;
3716 return err;
3717 }
3718
3719 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3720 {
3721 int nla_size = 0;
3722 int attr_id;
3723 int size;
3724
3725 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3726 dev->netdev_ops->ndo_get_offload_stats))
3727 return 0;
3728
3729 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3730 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3731 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3732 continue;
3733 size = rtnl_get_offload_stats_attr_size(attr_id);
3734 nla_size += nla_total_size_64bit(size);
3735 }
3736
3737 if (nla_size != 0)
3738 nla_size += nla_total_size(0);
3739
3740 return nla_size;
3741 }
3742
3743 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3744 int type, u32 pid, u32 seq, u32 change,
3745 unsigned int flags, unsigned int filter_mask,
3746 int *idxattr, int *prividx)
3747 {
3748 struct if_stats_msg *ifsm;
3749 struct nlmsghdr *nlh;
3750 struct nlattr *attr;
3751 int s_prividx = *prividx;
3752 int err;
3753
3754 ASSERT_RTNL();
3755
3756 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3757 if (!nlh)
3758 return -EMSGSIZE;
3759
3760 ifsm = nlmsg_data(nlh);
3761 ifsm->ifindex = dev->ifindex;
3762 ifsm->filter_mask = filter_mask;
3763
3764 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3765 struct rtnl_link_stats64 *sp;
3766
3767 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3768 sizeof(struct rtnl_link_stats64),
3769 IFLA_STATS_UNSPEC);
3770 if (!attr)
3771 goto nla_put_failure;
3772
3773 sp = nla_data(attr);
3774 dev_get_stats(dev, sp);
3775 }
3776
3777 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3778 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3779
3780 if (ops && ops->fill_linkxstats) {
3781 *idxattr = IFLA_STATS_LINK_XSTATS;
3782 attr = nla_nest_start(skb,
3783 IFLA_STATS_LINK_XSTATS);
3784 if (!attr)
3785 goto nla_put_failure;
3786
3787 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3788 nla_nest_end(skb, attr);
3789 if (err)
3790 goto nla_put_failure;
3791 *idxattr = 0;
3792 }
3793 }
3794
3795 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3796 *idxattr)) {
3797 const struct rtnl_link_ops *ops = NULL;
3798 const struct net_device *master;
3799
3800 master = netdev_master_upper_dev_get(dev);
3801 if (master)
3802 ops = master->rtnl_link_ops;
3803 if (ops && ops->fill_linkxstats) {
3804 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3805 attr = nla_nest_start(skb,
3806 IFLA_STATS_LINK_XSTATS_SLAVE);
3807 if (!attr)
3808 goto nla_put_failure;
3809
3810 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3811 nla_nest_end(skb, attr);
3812 if (err)
3813 goto nla_put_failure;
3814 *idxattr = 0;
3815 }
3816 }
3817
3818 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3819 *idxattr)) {
3820 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3821 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3822 if (!attr)
3823 goto nla_put_failure;
3824
3825 err = rtnl_get_offload_stats(skb, dev, prividx);
3826 if (err == -ENODATA)
3827 nla_nest_cancel(skb, attr);
3828 else
3829 nla_nest_end(skb, attr);
3830
3831 if (err && err != -ENODATA)
3832 goto nla_put_failure;
3833 *idxattr = 0;
3834 }
3835
3836 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3837 struct rtnl_af_ops *af_ops;
3838
3839 *idxattr = IFLA_STATS_AF_SPEC;
3840 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3841 if (!attr)
3842 goto nla_put_failure;
3843
3844 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3845 if (af_ops->fill_stats_af) {
3846 struct nlattr *af;
3847 int err;
3848
3849 af = nla_nest_start(skb, af_ops->family);
3850 if (!af)
3851 goto nla_put_failure;
3852
3853 err = af_ops->fill_stats_af(skb, dev);
3854
3855 if (err == -ENODATA)
3856 nla_nest_cancel(skb, af);
3857 else if (err < 0)
3858 goto nla_put_failure;
3859
3860 nla_nest_end(skb, af);
3861 }
3862 }
3863
3864 nla_nest_end(skb, attr);
3865
3866 *idxattr = 0;
3867 }
3868
3869 nlmsg_end(skb, nlh);
3870
3871 return 0;
3872
3873 nla_put_failure:
3874 /* not a multi message or no progress mean a real error */
3875 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3876 nlmsg_cancel(skb, nlh);
3877 else
3878 nlmsg_end(skb, nlh);
3879
3880 return -EMSGSIZE;
3881 }
3882
3883 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3884 u32 filter_mask)
3885 {
3886 size_t size = 0;
3887
3888 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3889 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3890
3891 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3892 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3893 int attr = IFLA_STATS_LINK_XSTATS;
3894
3895 if (ops && ops->get_linkxstats_size) {
3896 size += nla_total_size(ops->get_linkxstats_size(dev,
3897 attr));
3898 /* for IFLA_STATS_LINK_XSTATS */
3899 size += nla_total_size(0);
3900 }
3901 }
3902
3903 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3904 struct net_device *_dev = (struct net_device *)dev;
3905 const struct rtnl_link_ops *ops = NULL;
3906 const struct net_device *master;
3907
3908 /* netdev_master_upper_dev_get can't take const */
3909 master = netdev_master_upper_dev_get(_dev);
3910 if (master)
3911 ops = master->rtnl_link_ops;
3912 if (ops && ops->get_linkxstats_size) {
3913 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3914
3915 size += nla_total_size(ops->get_linkxstats_size(dev,
3916 attr));
3917 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3918 size += nla_total_size(0);
3919 }
3920 }
3921
3922 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3923 size += rtnl_get_offload_stats_size(dev);
3924
3925 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
3926 struct rtnl_af_ops *af_ops;
3927
3928 /* for IFLA_STATS_AF_SPEC */
3929 size += nla_total_size(0);
3930
3931 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3932 if (af_ops->get_stats_af_size) {
3933 size += nla_total_size(
3934 af_ops->get_stats_af_size(dev));
3935
3936 /* for AF_* */
3937 size += nla_total_size(0);
3938 }
3939 }
3940 }
3941
3942 return size;
3943 }
3944
3945 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3946 {
3947 struct net *net = sock_net(skb->sk);
3948 struct net_device *dev = NULL;
3949 int idxattr = 0, prividx = 0;
3950 struct if_stats_msg *ifsm;
3951 struct sk_buff *nskb;
3952 u32 filter_mask;
3953 int err;
3954
3955 if (nlmsg_len(nlh) < sizeof(*ifsm))
3956 return -EINVAL;
3957
3958 ifsm = nlmsg_data(nlh);
3959 if (ifsm->ifindex > 0)
3960 dev = __dev_get_by_index(net, ifsm->ifindex);
3961 else
3962 return -EINVAL;
3963
3964 if (!dev)
3965 return -ENODEV;
3966
3967 filter_mask = ifsm->filter_mask;
3968 if (!filter_mask)
3969 return -EINVAL;
3970
3971 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3972 if (!nskb)
3973 return -ENOBUFS;
3974
3975 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3976 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3977 0, filter_mask, &idxattr, &prividx);
3978 if (err < 0) {
3979 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3980 WARN_ON(err == -EMSGSIZE);
3981 kfree_skb(nskb);
3982 } else {
3983 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3984 }
3985
3986 return err;
3987 }
3988
3989 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3990 {
3991 int h, s_h, err, s_idx, s_idxattr, s_prividx;
3992 struct net *net = sock_net(skb->sk);
3993 unsigned int flags = NLM_F_MULTI;
3994 struct if_stats_msg *ifsm;
3995 struct hlist_head *head;
3996 struct net_device *dev;
3997 u32 filter_mask = 0;
3998 int idx = 0;
3999
4000 s_h = cb->args[0];
4001 s_idx = cb->args[1];
4002 s_idxattr = cb->args[2];
4003 s_prividx = cb->args[3];
4004
4005 cb->seq = net->dev_base_seq;
4006
4007 if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4008 return -EINVAL;
4009
4010 ifsm = nlmsg_data(cb->nlh);
4011 filter_mask = ifsm->filter_mask;
4012 if (!filter_mask)
4013 return -EINVAL;
4014
4015 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4016 idx = 0;
4017 head = &net->dev_index_head[h];
4018 hlist_for_each_entry(dev, head, index_hlist) {
4019 if (idx < s_idx)
4020 goto cont;
4021 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4022 NETLINK_CB(cb->skb).portid,
4023 cb->nlh->nlmsg_seq, 0,
4024 flags, filter_mask,
4025 &s_idxattr, &s_prividx);
4026 /* If we ran out of room on the first message,
4027 * we're in trouble
4028 */
4029 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4030
4031 if (err < 0)
4032 goto out;
4033 s_prividx = 0;
4034 s_idxattr = 0;
4035 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4036 cont:
4037 idx++;
4038 }
4039 }
4040 out:
4041 cb->args[3] = s_prividx;
4042 cb->args[2] = s_idxattr;
4043 cb->args[1] = idx;
4044 cb->args[0] = h;
4045
4046 return skb->len;
4047 }
4048
4049 /* Process one rtnetlink message. */
4050
4051 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
4052 {
4053 struct net *net = sock_net(skb->sk);
4054 rtnl_doit_func doit;
4055 int kind;
4056 int family;
4057 int type;
4058 int err;
4059
4060 type = nlh->nlmsg_type;
4061 if (type > RTM_MAX)
4062 return -EOPNOTSUPP;
4063
4064 type -= RTM_BASE;
4065
4066 /* All the messages must have at least 1 byte length */
4067 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4068 return 0;
4069
4070 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4071 kind = type&3;
4072
4073 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4074 return -EPERM;
4075
4076 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4077 struct sock *rtnl;
4078 rtnl_dumpit_func dumpit;
4079 rtnl_calcit_func calcit;
4080 u16 min_dump_alloc = 0;
4081
4082 dumpit = rtnl_get_dumpit(family, type);
4083 if (dumpit == NULL)
4084 return -EOPNOTSUPP;
4085 calcit = rtnl_get_calcit(family, type);
4086 if (calcit)
4087 min_dump_alloc = calcit(skb, nlh);
4088
4089 __rtnl_unlock();
4090 rtnl = net->rtnl;
4091 {
4092 struct netlink_dump_control c = {
4093 .dump = dumpit,
4094 .min_dump_alloc = min_dump_alloc,
4095 };
4096 err = netlink_dump_start(rtnl, skb, nlh, &c);
4097 }
4098 rtnl_lock();
4099 return err;
4100 }
4101
4102 doit = rtnl_get_doit(family, type);
4103 if (doit == NULL)
4104 return -EOPNOTSUPP;
4105
4106 return doit(skb, nlh);
4107 }
4108
4109 static void rtnetlink_rcv(struct sk_buff *skb)
4110 {
4111 rtnl_lock();
4112 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4113 rtnl_unlock();
4114 }
4115
4116 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4117 {
4118 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4119
4120 switch (event) {
4121 case NETDEV_UP:
4122 case NETDEV_DOWN:
4123 case NETDEV_PRE_UP:
4124 case NETDEV_POST_INIT:
4125 case NETDEV_REGISTER:
4126 case NETDEV_CHANGE:
4127 case NETDEV_PRE_TYPE_CHANGE:
4128 case NETDEV_GOING_DOWN:
4129 case NETDEV_UNREGISTER:
4130 case NETDEV_UNREGISTER_FINAL:
4131 case NETDEV_RELEASE:
4132 case NETDEV_JOIN:
4133 case NETDEV_BONDING_INFO:
4134 break;
4135 default:
4136 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4137 break;
4138 }
4139 return NOTIFY_DONE;
4140 }
4141
4142 static struct notifier_block rtnetlink_dev_notifier = {
4143 .notifier_call = rtnetlink_event,
4144 };
4145
4146
4147 static int __net_init rtnetlink_net_init(struct net *net)
4148 {
4149 struct sock *sk;
4150 struct netlink_kernel_cfg cfg = {
4151 .groups = RTNLGRP_MAX,
4152 .input = rtnetlink_rcv,
4153 .cb_mutex = &rtnl_mutex,
4154 .flags = NL_CFG_F_NONROOT_RECV,
4155 };
4156
4157 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4158 if (!sk)
4159 return -ENOMEM;
4160 net->rtnl = sk;
4161 return 0;
4162 }
4163
4164 static void __net_exit rtnetlink_net_exit(struct net *net)
4165 {
4166 netlink_kernel_release(net->rtnl);
4167 net->rtnl = NULL;
4168 }
4169
4170 static struct pernet_operations rtnetlink_net_ops = {
4171 .init = rtnetlink_net_init,
4172 .exit = rtnetlink_net_exit,
4173 };
4174
4175 void __init rtnetlink_init(void)
4176 {
4177 if (register_pernet_subsys(&rtnetlink_net_ops))
4178 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4179
4180 register_netdevice_notifier(&rtnetlink_dev_notifier);
4181
4182 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4183 rtnl_dump_ifinfo, rtnl_calcit);
4184 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4185 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4186 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4187
4188 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4189 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4190
4191 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4192 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4193 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4194
4195 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4196 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4197 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4198
4199 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4200 NULL);
4201 }