?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / net / core / rtnetlink.c
blobbc290413a49d5e3c7e115c5d900e087882735397
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/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44
45 #include <linux/uaccess.h>
46
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
50 #include <net/ip.h>
51 #include <net/protocol.h>
52 #include <net/arp.h>
53 #include <net/route.h>
54 #include <net/udp.h>
55 #include <net/tcp.h>
56 #include <net/sock.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
61
62 struct rtnl_link {
63 rtnl_doit_func doit;
64 rtnl_dumpit_func dumpit;
65 struct module *owner;
66 unsigned int flags;
67 struct rcu_head rcu;
68 };
69
70 static DEFINE_MUTEX(rtnl_mutex);
71
72 void rtnl_lock(void)
73 {
74 mutex_lock(&rtnl_mutex);
75 }
76 EXPORT_SYMBOL(rtnl_lock);
77
78 static struct sk_buff *defer_kfree_skb_list;
79 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
80 {
81 if (head && tail) {
82 tail->next = defer_kfree_skb_list;
83 defer_kfree_skb_list = head;
84 }
85 }
86 EXPORT_SYMBOL(rtnl_kfree_skbs);
87
88 void __rtnl_unlock(void)
89 {
90 struct sk_buff *head = defer_kfree_skb_list;
91
92 defer_kfree_skb_list = NULL;
93
94 mutex_unlock(&rtnl_mutex);
95
96 while (head) {
97 struct sk_buff *next = head->next;
98
99 kfree_skb(head);
100 cond_resched();
101 head = next;
102 }
103 }
104
105 void rtnl_unlock(void)
106 {
107 /* This fellow will unlock it for us. */
108 netdev_run_todo();
109 }
110 EXPORT_SYMBOL(rtnl_unlock);
111
112 int rtnl_trylock(void)
113 {
114 return mutex_trylock(&rtnl_mutex);
115 }
116 EXPORT_SYMBOL(rtnl_trylock);
117
118 int rtnl_is_locked(void)
119 {
120 return mutex_is_locked(&rtnl_mutex);
121 }
122 EXPORT_SYMBOL(rtnl_is_locked);
123
124 #ifdef CONFIG_PROVE_LOCKING
125 bool lockdep_rtnl_is_held(void)
126 {
127 return lockdep_is_held(&rtnl_mutex);
128 }
129 EXPORT_SYMBOL(lockdep_rtnl_is_held);
130 #endif /* #ifdef CONFIG_PROVE_LOCKING */
131
132 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
133
134 static inline int rtm_msgindex(int msgtype)
135 {
136 int msgindex = msgtype - RTM_BASE;
137
138 /*
139 * msgindex < 0 implies someone tried to register a netlink
140 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
141 * the message type has not been added to linux/rtnetlink.h
142 */
143 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
144
145 return msgindex;
146 }
147
148 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
149 {
150 struct rtnl_link **tab;
151
152 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
153 protocol = PF_UNSPEC;
154
155 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
156 if (!tab)
157 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
158
159 return tab[msgtype];
160 }
161
162 static int rtnl_register_internal(struct module *owner,
163 int protocol, int msgtype,
164 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
165 unsigned int flags)
166 {
167 struct rtnl_link *link, *old;
168 struct rtnl_link __rcu **tab;
169 int msgindex;
170 int ret = -ENOBUFS;
171
172 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
173 msgindex = rtm_msgindex(msgtype);
174
175 rtnl_lock();
176 tab = rtnl_msg_handlers[protocol];
177 if (tab == NULL) {
178 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
179 if (!tab)
180 goto unlock;
181
182 /* ensures we see the 0 stores */
183 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
184 }
185
186 old = rtnl_dereference(tab[msgindex]);
187 if (old) {
188 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
189 if (!link)
190 goto unlock;
191 } else {
192 link = kzalloc(sizeof(*link), GFP_KERNEL);
193 if (!link)
194 goto unlock;
195 }
196
197 WARN_ON(link->owner && link->owner != owner);
198 link->owner = owner;
199
200 WARN_ON(doit && link->doit && link->doit != doit);
201 if (doit)
202 link->doit = doit;
203 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
204 if (dumpit)
205 link->dumpit = dumpit;
206
207 link->flags |= flags;
208
209 /* publish protocol:msgtype */
210 rcu_assign_pointer(tab[msgindex], link);
211 ret = 0;
212 if (old)
213 kfree_rcu(old, rcu);
214 unlock:
215 rtnl_unlock();
216 return ret;
217 }
218
219 /**
220 * rtnl_register_module - Register a rtnetlink message type
221 *
222 * @owner: module registering the hook (THIS_MODULE)
223 * @protocol: Protocol family or PF_UNSPEC
224 * @msgtype: rtnetlink message type
225 * @doit: Function pointer called for each request message
226 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
227 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
228 *
229 * Like rtnl_register, but for use by removable modules.
230 */
231 int rtnl_register_module(struct module *owner,
232 int protocol, int msgtype,
233 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
234 unsigned int flags)
235 {
236 return rtnl_register_internal(owner, protocol, msgtype,
237 doit, dumpit, flags);
238 }
239 EXPORT_SYMBOL_GPL(rtnl_register_module);
240
241 /**
242 * rtnl_register - Register a rtnetlink message type
243 * @protocol: Protocol family or PF_UNSPEC
244 * @msgtype: rtnetlink message type
245 * @doit: Function pointer called for each request message
246 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
247 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
248 *
249 * Registers the specified function pointers (at least one of them has
250 * to be non-NULL) to be called whenever a request message for the
251 * specified protocol family and message type is received.
252 *
253 * The special protocol family PF_UNSPEC may be used to define fallback
254 * function pointers for the case when no entry for the specific protocol
255 * family exists.
256 */
257 void rtnl_register(int protocol, int msgtype,
258 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
259 unsigned int flags)
260 {
261 int err;
262
263 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
264 flags);
265 if (err)
266 pr_err("Unable to register rtnetlink message handler, "
267 "protocol = %d, message type = %d\n", protocol, msgtype);
268 }
269
270 /**
271 * rtnl_unregister - Unregister a rtnetlink message type
272 * @protocol: Protocol family or PF_UNSPEC
273 * @msgtype: rtnetlink message type
274 *
275 * Returns 0 on success or a negative error code.
276 */
277 int rtnl_unregister(int protocol, int msgtype)
278 {
279 struct rtnl_link **tab, *link;
280 int msgindex;
281
282 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
283 msgindex = rtm_msgindex(msgtype);
284
285 rtnl_lock();
286 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
287 if (!tab) {
288 rtnl_unlock();
289 return -ENOENT;
290 }
291
292 link = tab[msgindex];
293 rcu_assign_pointer(tab[msgindex], NULL);
294 rtnl_unlock();
295
296 kfree_rcu(link, rcu);
297
298 return 0;
299 }
300 EXPORT_SYMBOL_GPL(rtnl_unregister);
301
302 /**
303 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
304 * @protocol : Protocol family or PF_UNSPEC
305 *
306 * Identical to calling rtnl_unregster() for all registered message types
307 * of a certain protocol family.
308 */
309 void rtnl_unregister_all(int protocol)
310 {
311 struct rtnl_link **tab, *link;
312 int msgindex;
313
314 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
315
316 rtnl_lock();
317 tab = rtnl_msg_handlers[protocol];
318 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
319 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
320 link = tab[msgindex];
321 if (!link)
322 continue;
323
324 rcu_assign_pointer(tab[msgindex], NULL);
325 kfree_rcu(link, rcu);
326 }
327 rtnl_unlock();
328
329 synchronize_net();
330
331 kfree(tab);
332 }
333 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
334
335 static LIST_HEAD(link_ops);
336
337 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
338 {
339 const struct rtnl_link_ops *ops;
340
341 list_for_each_entry(ops, &link_ops, list) {
342 if (!strcmp(ops->kind, kind))
343 return ops;
344 }
345 return NULL;
346 }
347
348 /**
349 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
350 * @ops: struct rtnl_link_ops * to register
351 *
352 * The caller must hold the rtnl_mutex. This function should be used
353 * by drivers that create devices during module initialization. It
354 * must be called before registering the devices.
355 *
356 * Returns 0 on success or a negative error code.
357 */
358 int __rtnl_link_register(struct rtnl_link_ops *ops)
359 {
360 if (rtnl_link_ops_get(ops->kind))
361 return -EEXIST;
362
363 /* The check for setup is here because if ops
364 * does not have that filled up, it is not possible
365 * to use the ops for creating device. So do not
366 * fill up dellink as well. That disables rtnl_dellink.
367 */
368 if (ops->setup && !ops->dellink)
369 ops->dellink = unregister_netdevice_queue;
370
371 list_add_tail(&ops->list, &link_ops);
372 return 0;
373 }
374 EXPORT_SYMBOL_GPL(__rtnl_link_register);
375
376 /**
377 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
378 * @ops: struct rtnl_link_ops * to register
379 *
380 * Returns 0 on success or a negative error code.
381 */
382 int rtnl_link_register(struct rtnl_link_ops *ops)
383 {
384 int err;
385
386 rtnl_lock();
387 err = __rtnl_link_register(ops);
388 rtnl_unlock();
389 return err;
390 }
391 EXPORT_SYMBOL_GPL(rtnl_link_register);
392
393 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
394 {
395 struct net_device *dev;
396 LIST_HEAD(list_kill);
397
398 for_each_netdev(net, dev) {
399 if (dev->rtnl_link_ops == ops)
400 ops->dellink(dev, &list_kill);
401 }
402 unregister_netdevice_many(&list_kill);
403 }
404
405 /**
406 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
407 * @ops: struct rtnl_link_ops * to unregister
408 *
409 * The caller must hold the rtnl_mutex.
410 */
411 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
412 {
413 struct net *net;
414
415 for_each_net(net) {
416 __rtnl_kill_links(net, ops);
417 }
418 list_del(&ops->list);
419 }
420 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
421
422 /* Return with the rtnl_lock held when there are no network
423 * devices unregistering in any network namespace.
424 */
425 static void rtnl_lock_unregistering_all(void)
426 {
427 struct net *net;
428 bool unregistering;
429 DEFINE_WAIT_FUNC(wait, woken_wake_function);
430
431 add_wait_queue(&netdev_unregistering_wq, &wait);
432 for (;;) {
433 unregistering = false;
434 rtnl_lock();
435 for_each_net(net) {
436 if (net->dev_unreg_count > 0) {
437 unregistering = true;
438 break;
439 }
440 }
441 if (!unregistering)
442 break;
443 __rtnl_unlock();
444
445 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
446 }
447 remove_wait_queue(&netdev_unregistering_wq, &wait);
448 }
449
450 /**
451 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
452 * @ops: struct rtnl_link_ops * to unregister
453 */
454 void rtnl_link_unregister(struct rtnl_link_ops *ops)
455 {
456 /* Close the race with cleanup_net() */
457 mutex_lock(&net_mutex);
458 rtnl_lock_unregistering_all();
459 __rtnl_link_unregister(ops);
460 rtnl_unlock();
461 mutex_unlock(&net_mutex);
462 }
463 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
464
465 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
466 {
467 struct net_device *master_dev;
468 const struct rtnl_link_ops *ops;
469 size_t size = 0;
470
471 rcu_read_lock();
472
473 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
474 if (!master_dev)
475 goto out;
476
477 ops = master_dev->rtnl_link_ops;
478 if (!ops || !ops->get_slave_size)
479 goto out;
480 /* IFLA_INFO_SLAVE_DATA + nested data */
481 size = nla_total_size(sizeof(struct nlattr)) +
482 ops->get_slave_size(master_dev, dev);
483
484 out:
485 rcu_read_unlock();
486 return size;
487 }
488
489 static size_t rtnl_link_get_size(const struct net_device *dev)
490 {
491 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
492 size_t size;
493
494 if (!ops)
495 return 0;
496
497 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
498 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
499
500 if (ops->get_size)
501 /* IFLA_INFO_DATA + nested data */
502 size += nla_total_size(sizeof(struct nlattr)) +
503 ops->get_size(dev);
504
505 if (ops->get_xstats_size)
506 /* IFLA_INFO_XSTATS */
507 size += nla_total_size(ops->get_xstats_size(dev));
508
509 size += rtnl_link_get_slave_info_data_size(dev);
510
511 return size;
512 }
513
514 static LIST_HEAD(rtnl_af_ops);
515
516 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
517 {
518 const struct rtnl_af_ops *ops;
519
520 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
521 if (ops->family == family)
522 return ops;
523 }
524
525 return NULL;
526 }
527
528 /**
529 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
530 * @ops: struct rtnl_af_ops * to register
531 *
532 * Returns 0 on success or a negative error code.
533 */
534 void rtnl_af_register(struct rtnl_af_ops *ops)
535 {
536 rtnl_lock();
537 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
538 rtnl_unlock();
539 }
540 EXPORT_SYMBOL_GPL(rtnl_af_register);
541
542 /**
543 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
544 * @ops: struct rtnl_af_ops * to unregister
545 */
546 void rtnl_af_unregister(struct rtnl_af_ops *ops)
547 {
548 rtnl_lock();
549 list_del_rcu(&ops->list);
550 rtnl_unlock();
551
552 synchronize_rcu();
553 }
554 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
555
556 static size_t rtnl_link_get_af_size(const struct net_device *dev,
557 u32 ext_filter_mask)
558 {
559 struct rtnl_af_ops *af_ops;
560 size_t size;
561
562 /* IFLA_AF_SPEC */
563 size = nla_total_size(sizeof(struct nlattr));
564
565 rcu_read_lock();
566 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
567 if (af_ops->get_link_af_size) {
568 /* AF_* + nested data */
569 size += nla_total_size(sizeof(struct nlattr)) +
570 af_ops->get_link_af_size(dev, ext_filter_mask);
571 }
572 }
573 rcu_read_unlock();
574
575 return size;
576 }
577
578 static bool rtnl_have_link_slave_info(const struct net_device *dev)
579 {
580 struct net_device *master_dev;
581 bool ret = false;
582
583 rcu_read_lock();
584
585 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
586 if (master_dev && master_dev->rtnl_link_ops)
587 ret = true;
588 rcu_read_unlock();
589 return ret;
590 }
591
592 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
593 const struct net_device *dev)
594 {
595 struct net_device *master_dev;
596 const struct rtnl_link_ops *ops;
597 struct nlattr *slave_data;
598 int err;
599
600 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
601 if (!master_dev)
602 return 0;
603 ops = master_dev->rtnl_link_ops;
604 if (!ops)
605 return 0;
606 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
607 return -EMSGSIZE;
608 if (ops->fill_slave_info) {
609 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
610 if (!slave_data)
611 return -EMSGSIZE;
612 err = ops->fill_slave_info(skb, master_dev, dev);
613 if (err < 0)
614 goto err_cancel_slave_data;
615 nla_nest_end(skb, slave_data);
616 }
617 return 0;
618
619 err_cancel_slave_data:
620 nla_nest_cancel(skb, slave_data);
621 return err;
622 }
623
624 static int rtnl_link_info_fill(struct sk_buff *skb,
625 const struct net_device *dev)
626 {
627 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
628 struct nlattr *data;
629 int err;
630
631 if (!ops)
632 return 0;
633 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
634 return -EMSGSIZE;
635 if (ops->fill_xstats) {
636 err = ops->fill_xstats(skb, dev);
637 if (err < 0)
638 return err;
639 }
640 if (ops->fill_info) {
641 data = nla_nest_start(skb, IFLA_INFO_DATA);
642 if (data == NULL)
643 return -EMSGSIZE;
644 err = ops->fill_info(skb, dev);
645 if (err < 0)
646 goto err_cancel_data;
647 nla_nest_end(skb, data);
648 }
649 return 0;
650
651 err_cancel_data:
652 nla_nest_cancel(skb, data);
653 return err;
654 }
655
656 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
657 {
658 struct nlattr *linkinfo;
659 int err = -EMSGSIZE;
660
661 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
662 if (linkinfo == NULL)
663 goto out;
664
665 err = rtnl_link_info_fill(skb, dev);
666 if (err < 0)
667 goto err_cancel_link;
668
669 err = rtnl_link_slave_info_fill(skb, dev);
670 if (err < 0)
671 goto err_cancel_link;
672
673 nla_nest_end(skb, linkinfo);
674 return 0;
675
676 err_cancel_link:
677 nla_nest_cancel(skb, linkinfo);
678 out:
679 return err;
680 }
681
682 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
683 {
684 struct sock *rtnl = net->rtnl;
685 int err = 0;
686
687 NETLINK_CB(skb).dst_group = group;
688 if (echo)
689 refcount_inc(&skb->users);
690 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
691 if (echo)
692 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
693 return err;
694 }
695
696 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
697 {
698 struct sock *rtnl = net->rtnl;
699
700 return nlmsg_unicast(rtnl, skb, pid);
701 }
702 EXPORT_SYMBOL(rtnl_unicast);
703
704 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
705 struct nlmsghdr *nlh, gfp_t flags)
706 {
707 struct sock *rtnl = net->rtnl;
708 int report = 0;
709
710 if (nlh)
711 report = nlmsg_report(nlh);
712
713 nlmsg_notify(rtnl, skb, pid, group, report, flags);
714 }
715 EXPORT_SYMBOL(rtnl_notify);
716
717 void rtnl_set_sk_err(struct net *net, u32 group, int error)
718 {
719 struct sock *rtnl = net->rtnl;
720
721 netlink_set_err(rtnl, 0, group, error);
722 }
723 EXPORT_SYMBOL(rtnl_set_sk_err);
724
725 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
726 {
727 struct nlattr *mx;
728 int i, valid = 0;
729
730 mx = nla_nest_start(skb, RTA_METRICS);
731 if (mx == NULL)
732 return -ENOBUFS;
733
734 for (i = 0; i < RTAX_MAX; i++) {
735 if (metrics[i]) {
736 if (i == RTAX_CC_ALGO - 1) {
737 char tmp[TCP_CA_NAME_MAX], *name;
738
739 name = tcp_ca_get_name_by_key(metrics[i], tmp);
740 if (!name)
741 continue;
742 if (nla_put_string(skb, i + 1, name))
743 goto nla_put_failure;
744 } else if (i == RTAX_FEATURES - 1) {
745 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
746
747 if (!user_features)
748 continue;
749 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
750 if (nla_put_u32(skb, i + 1, user_features))
751 goto nla_put_failure;
752 } else {
753 if (nla_put_u32(skb, i + 1, metrics[i]))
754 goto nla_put_failure;
755 }
756 valid++;
757 }
758 }
759
760 if (!valid) {
761 nla_nest_cancel(skb, mx);
762 return 0;
763 }
764
765 return nla_nest_end(skb, mx);
766
767 nla_put_failure:
768 nla_nest_cancel(skb, mx);
769 return -EMSGSIZE;
770 }
771 EXPORT_SYMBOL(rtnetlink_put_metrics);
772
773 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
774 long expires, u32 error)
775 {
776 struct rta_cacheinfo ci = {
777 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
778 .rta_used = dst->__use,
779 .rta_clntref = atomic_read(&(dst->__refcnt)),
780 .rta_error = error,
781 .rta_id = id,
782 };
783
784 if (expires) {
785 unsigned long clock;
786
787 clock = jiffies_to_clock_t(abs(expires));
788 clock = min_t(unsigned long, clock, INT_MAX);
789 ci.rta_expires = (expires > 0) ? clock : -clock;
790 }
791 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
792 }
793 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
794
795 static void set_operstate(struct net_device *dev, unsigned char transition)
796 {
797 unsigned char operstate = dev->operstate;
798
799 switch (transition) {
800 case IF_OPER_UP:
801 if ((operstate == IF_OPER_DORMANT ||
802 operstate == IF_OPER_UNKNOWN) &&
803 !netif_dormant(dev))
804 operstate = IF_OPER_UP;
805 break;
806
807 case IF_OPER_DORMANT:
808 if (operstate == IF_OPER_UP ||
809 operstate == IF_OPER_UNKNOWN)
810 operstate = IF_OPER_DORMANT;
811 break;
812 }
813
814 if (dev->operstate != operstate) {
815 write_lock_bh(&dev_base_lock);
816 dev->operstate = operstate;
817 write_unlock_bh(&dev_base_lock);
818 netdev_state_change(dev);
819 }
820 }
821
822 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
823 {
824 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
825 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
826 }
827
828 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
829 const struct ifinfomsg *ifm)
830 {
831 unsigned int flags = ifm->ifi_flags;
832
833 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
834 if (ifm->ifi_change)
835 flags = (flags & ifm->ifi_change) |
836 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
837
838 return flags;
839 }
840
841 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
842 const struct rtnl_link_stats64 *b)
843 {
844 a->rx_packets = b->rx_packets;
845 a->tx_packets = b->tx_packets;
846 a->rx_bytes = b->rx_bytes;
847 a->tx_bytes = b->tx_bytes;
848 a->rx_errors = b->rx_errors;
849 a->tx_errors = b->tx_errors;
850 a->rx_dropped = b->rx_dropped;
851 a->tx_dropped = b->tx_dropped;
852
853 a->multicast = b->multicast;
854 a->collisions = b->collisions;
855
856 a->rx_length_errors = b->rx_length_errors;
857 a->rx_over_errors = b->rx_over_errors;
858 a->rx_crc_errors = b->rx_crc_errors;
859 a->rx_frame_errors = b->rx_frame_errors;
860 a->rx_fifo_errors = b->rx_fifo_errors;
861 a->rx_missed_errors = b->rx_missed_errors;
862
863 a->tx_aborted_errors = b->tx_aborted_errors;
864 a->tx_carrier_errors = b->tx_carrier_errors;
865 a->tx_fifo_errors = b->tx_fifo_errors;
866 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
867 a->tx_window_errors = b->tx_window_errors;
868
869 a->rx_compressed = b->rx_compressed;
870 a->tx_compressed = b->tx_compressed;
871
872 a->rx_nohandler = b->rx_nohandler;
873 }
874
875 /* All VF info */
876 static inline int rtnl_vfinfo_size(const struct net_device *dev,
877 u32 ext_filter_mask)
878 {
879 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
880 int num_vfs = dev_num_vf(dev->dev.parent);
881 size_t size = nla_total_size(0);
882 size += num_vfs *
883 (nla_total_size(0) +
884 nla_total_size(sizeof(struct ifla_vf_mac)) +
885 nla_total_size(sizeof(struct ifla_vf_vlan)) +
886 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
887 nla_total_size(MAX_VLAN_LIST_LEN *
888 sizeof(struct ifla_vf_vlan_info)) +
889 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
890 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
891 nla_total_size(sizeof(struct ifla_vf_rate)) +
892 nla_total_size(sizeof(struct ifla_vf_link_state)) +
893 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
894 nla_total_size(0) + /* nest IFLA_VF_STATS */
895 /* IFLA_VF_STATS_RX_PACKETS */
896 nla_total_size_64bit(sizeof(__u64)) +
897 /* IFLA_VF_STATS_TX_PACKETS */
898 nla_total_size_64bit(sizeof(__u64)) +
899 /* IFLA_VF_STATS_RX_BYTES */
900 nla_total_size_64bit(sizeof(__u64)) +
901 /* IFLA_VF_STATS_TX_BYTES */
902 nla_total_size_64bit(sizeof(__u64)) +
903 /* IFLA_VF_STATS_BROADCAST */
904 nla_total_size_64bit(sizeof(__u64)) +
905 /* IFLA_VF_STATS_MULTICAST */
906 nla_total_size_64bit(sizeof(__u64)) +
907 /* IFLA_VF_STATS_RX_DROPPED */
908 nla_total_size_64bit(sizeof(__u64)) +
909 /* IFLA_VF_STATS_TX_DROPPED */
910 nla_total_size_64bit(sizeof(__u64)) +
911 nla_total_size(sizeof(struct ifla_vf_trust)));
912 return size;
913 } else
914 return 0;
915 }
916
917 static size_t rtnl_port_size(const struct net_device *dev,
918 u32 ext_filter_mask)
919 {
920 size_t port_size = nla_total_size(4) /* PORT_VF */
921 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
922 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
923 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
924 + nla_total_size(1) /* PROT_VDP_REQUEST */
925 + nla_total_size(2); /* PORT_VDP_RESPONSE */
926 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
927 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
928 + port_size;
929 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
930 + port_size;
931
932 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
933 !(ext_filter_mask & RTEXT_FILTER_VF))
934 return 0;
935 if (dev_num_vf(dev->dev.parent))
936 return port_self_size + vf_ports_size +
937 vf_port_size * dev_num_vf(dev->dev.parent);
938 else
939 return port_self_size;
940 }
941
942 static size_t rtnl_xdp_size(void)
943 {
944 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
945 nla_total_size(1) + /* XDP_ATTACHED */
946 nla_total_size(4); /* XDP_PROG_ID */
947
948 return xdp_size;
949 }
950
951 static noinline size_t if_nlmsg_size(const struct net_device *dev,
952 u32 ext_filter_mask)
953 {
954 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
955 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
956 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
957 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
958 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
959 + nla_total_size(sizeof(struct rtnl_link_stats))
960 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
961 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
962 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
963 + nla_total_size(4) /* IFLA_TXQLEN */
964 + nla_total_size(4) /* IFLA_WEIGHT */
965 + nla_total_size(4) /* IFLA_MTU */
966 + nla_total_size(4) /* IFLA_LINK */
967 + nla_total_size(4) /* IFLA_MASTER */
968 + nla_total_size(1) /* IFLA_CARRIER */
969 + nla_total_size(4) /* IFLA_PROMISCUITY */
970 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
971 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
972 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
973 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
974 + nla_total_size(1) /* IFLA_OPERSTATE */
975 + nla_total_size(1) /* IFLA_LINKMODE */
976 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
977 + nla_total_size(4) /* IFLA_LINK_NETNSID */
978 + nla_total_size(4) /* IFLA_GROUP */
979 + nla_total_size(ext_filter_mask
980 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
981 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
982 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
983 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
984 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
985 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
986 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
987 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
988 + rtnl_xdp_size() /* IFLA_XDP */
989 + nla_total_size(4) /* IFLA_EVENT */
990 + nla_total_size(4) /* IFLA_NEW_NETNSID */
991 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
992 + nla_total_size(1) /* IFLA_PROTO_DOWN */
993 + nla_total_size(4) /* IFLA_IF_NETNSID */
994 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
995 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
996 + 0;
997 }
998
999 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1000 {
1001 struct nlattr *vf_ports;
1002 struct nlattr *vf_port;
1003 int vf;
1004 int err;
1005
1006 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
1007 if (!vf_ports)
1008 return -EMSGSIZE;
1009
1010 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1011 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
1012 if (!vf_port)
1013 goto nla_put_failure;
1014 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1015 goto nla_put_failure;
1016 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1017 if (err == -EMSGSIZE)
1018 goto nla_put_failure;
1019 if (err) {
1020 nla_nest_cancel(skb, vf_port);
1021 continue;
1022 }
1023 nla_nest_end(skb, vf_port);
1024 }
1025
1026 nla_nest_end(skb, vf_ports);
1027
1028 return 0;
1029
1030 nla_put_failure:
1031 nla_nest_cancel(skb, vf_ports);
1032 return -EMSGSIZE;
1033 }
1034
1035 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1036 {
1037 struct nlattr *port_self;
1038 int err;
1039
1040 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
1041 if (!port_self)
1042 return -EMSGSIZE;
1043
1044 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1045 if (err) {
1046 nla_nest_cancel(skb, port_self);
1047 return (err == -EMSGSIZE) ? err : 0;
1048 }
1049
1050 nla_nest_end(skb, port_self);
1051
1052 return 0;
1053 }
1054
1055 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1056 u32 ext_filter_mask)
1057 {
1058 int err;
1059
1060 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1061 !(ext_filter_mask & RTEXT_FILTER_VF))
1062 return 0;
1063
1064 err = rtnl_port_self_fill(skb, dev);
1065 if (err)
1066 return err;
1067
1068 if (dev_num_vf(dev->dev.parent)) {
1069 err = rtnl_vf_ports_fill(skb, dev);
1070 if (err)
1071 return err;
1072 }
1073
1074 return 0;
1075 }
1076
1077 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1078 {
1079 int err;
1080 struct netdev_phys_item_id ppid;
1081
1082 err = dev_get_phys_port_id(dev, &ppid);
1083 if (err) {
1084 if (err == -EOPNOTSUPP)
1085 return 0;
1086 return err;
1087 }
1088
1089 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1090 return -EMSGSIZE;
1091
1092 return 0;
1093 }
1094
1095 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1096 {
1097 char name[IFNAMSIZ];
1098 int err;
1099
1100 err = dev_get_phys_port_name(dev, name, sizeof(name));
1101 if (err) {
1102 if (err == -EOPNOTSUPP)
1103 return 0;
1104 return err;
1105 }
1106
1107 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1108 return -EMSGSIZE;
1109
1110 return 0;
1111 }
1112
1113 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1114 {
1115 int err;
1116 struct switchdev_attr attr = {
1117 .orig_dev = dev,
1118 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1119 .flags = SWITCHDEV_F_NO_RECURSE,
1120 };
1121
1122 err = switchdev_port_attr_get(dev, &attr);
1123 if (err) {
1124 if (err == -EOPNOTSUPP)
1125 return 0;
1126 return err;
1127 }
1128
1129 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1130 attr.u.ppid.id))
1131 return -EMSGSIZE;
1132
1133 return 0;
1134 }
1135
1136 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1137 struct net_device *dev)
1138 {
1139 struct rtnl_link_stats64 *sp;
1140 struct nlattr *attr;
1141
1142 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1143 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1144 if (!attr)
1145 return -EMSGSIZE;
1146
1147 sp = nla_data(attr);
1148 dev_get_stats(dev, sp);
1149
1150 attr = nla_reserve(skb, IFLA_STATS,
1151 sizeof(struct rtnl_link_stats));
1152 if (!attr)
1153 return -EMSGSIZE;
1154
1155 copy_rtnl_link_stats(nla_data(attr), sp);
1156
1157 return 0;
1158 }
1159
1160 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1161 struct net_device *dev,
1162 int vfs_num,
1163 struct nlattr *vfinfo)
1164 {
1165 struct ifla_vf_rss_query_en vf_rss_query_en;
1166 struct nlattr *vf, *vfstats, *vfvlanlist;
1167 struct ifla_vf_link_state vf_linkstate;
1168 struct ifla_vf_vlan_info vf_vlan_info;
1169 struct ifla_vf_spoofchk vf_spoofchk;
1170 struct ifla_vf_tx_rate vf_tx_rate;
1171 struct ifla_vf_stats vf_stats;
1172 struct ifla_vf_trust vf_trust;
1173 struct ifla_vf_vlan vf_vlan;
1174 struct ifla_vf_rate vf_rate;
1175 struct ifla_vf_mac vf_mac;
1176 struct ifla_vf_info ivi;
1177
1178 memset(&ivi, 0, sizeof(ivi));
1179
1180 /* Not all SR-IOV capable drivers support the
1181 * spoofcheck and "RSS query enable" query. Preset to
1182 * -1 so the user space tool can detect that the driver
1183 * didn't report anything.
1184 */
1185 ivi.spoofchk = -1;
1186 ivi.rss_query_en = -1;
1187 ivi.trusted = -1;
1188 /* The default value for VF link state is "auto"
1189 * IFLA_VF_LINK_STATE_AUTO which equals zero
1190 */
1191 ivi.linkstate = 0;
1192 /* VLAN Protocol by default is 802.1Q */
1193 ivi.vlan_proto = htons(ETH_P_8021Q);
1194 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1195 return 0;
1196
1197 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1198
1199 vf_mac.vf =
1200 vf_vlan.vf =
1201 vf_vlan_info.vf =
1202 vf_rate.vf =
1203 vf_tx_rate.vf =
1204 vf_spoofchk.vf =
1205 vf_linkstate.vf =
1206 vf_rss_query_en.vf =
1207 vf_trust.vf = ivi.vf;
1208
1209 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1210 vf_vlan.vlan = ivi.vlan;
1211 vf_vlan.qos = ivi.qos;
1212 vf_vlan_info.vlan = ivi.vlan;
1213 vf_vlan_info.qos = ivi.qos;
1214 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1215 vf_tx_rate.rate = ivi.max_tx_rate;
1216 vf_rate.min_tx_rate = ivi.min_tx_rate;
1217 vf_rate.max_tx_rate = ivi.max_tx_rate;
1218 vf_spoofchk.setting = ivi.spoofchk;
1219 vf_linkstate.link_state = ivi.linkstate;
1220 vf_rss_query_en.setting = ivi.rss_query_en;
1221 vf_trust.setting = ivi.trusted;
1222 vf = nla_nest_start(skb, IFLA_VF_INFO);
1223 if (!vf)
1224 goto nla_put_vfinfo_failure;
1225 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1226 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1227 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1228 &vf_rate) ||
1229 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1230 &vf_tx_rate) ||
1231 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1232 &vf_spoofchk) ||
1233 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1234 &vf_linkstate) ||
1235 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1236 sizeof(vf_rss_query_en),
1237 &vf_rss_query_en) ||
1238 nla_put(skb, IFLA_VF_TRUST,
1239 sizeof(vf_trust), &vf_trust))
1240 goto nla_put_vf_failure;
1241 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1242 if (!vfvlanlist)
1243 goto nla_put_vf_failure;
1244 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1245 &vf_vlan_info)) {
1246 nla_nest_cancel(skb, vfvlanlist);
1247 goto nla_put_vf_failure;
1248 }
1249 nla_nest_end(skb, vfvlanlist);
1250 memset(&vf_stats, 0, sizeof(vf_stats));
1251 if (dev->netdev_ops->ndo_get_vf_stats)
1252 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1253 &vf_stats);
1254 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1255 if (!vfstats)
1256 goto nla_put_vf_failure;
1257 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1258 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1259 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1260 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1261 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1262 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1263 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1264 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1265 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1266 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1267 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1268 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1269 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1270 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1271 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1272 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1273 nla_nest_cancel(skb, vfstats);
1274 goto nla_put_vf_failure;
1275 }
1276 nla_nest_end(skb, vfstats);
1277 nla_nest_end(skb, vf);
1278 return 0;
1279
1280 nla_put_vf_failure:
1281 nla_nest_cancel(skb, vf);
1282 nla_put_vfinfo_failure:
1283 nla_nest_cancel(skb, vfinfo);
1284 return -EMSGSIZE;
1285 }
1286
1287 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1288 struct net_device *dev,
1289 u32 ext_filter_mask)
1290 {
1291 struct nlattr *vfinfo;
1292 int i, num_vfs;
1293
1294 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1295 return 0;
1296
1297 num_vfs = dev_num_vf(dev->dev.parent);
1298 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1299 return -EMSGSIZE;
1300
1301 if (!dev->netdev_ops->ndo_get_vf_config)
1302 return 0;
1303
1304 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1305 if (!vfinfo)
1306 return -EMSGSIZE;
1307
1308 for (i = 0; i < num_vfs; i++) {
1309 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1310 return -EMSGSIZE;
1311 }
1312
1313 nla_nest_end(skb, vfinfo);
1314 return 0;
1315 }
1316
1317 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1318 {
1319 struct rtnl_link_ifmap map;
1320
1321 memset(&map, 0, sizeof(map));
1322 map.mem_start = dev->mem_start;
1323 map.mem_end = dev->mem_end;
1324 map.base_addr = dev->base_addr;
1325 map.irq = dev->irq;
1326 map.dma = dev->dma;
1327 map.port = dev->if_port;
1328
1329 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1330 return -EMSGSIZE;
1331
1332 return 0;
1333 }
1334
1335 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1336 {
1337 const struct net_device_ops *ops = dev->netdev_ops;
1338 const struct bpf_prog *generic_xdp_prog;
1339 struct netdev_bpf xdp;
1340
1341 ASSERT_RTNL();
1342
1343 *prog_id = 0;
1344 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1345 if (generic_xdp_prog) {
1346 *prog_id = generic_xdp_prog->aux->id;
1347 return XDP_ATTACHED_SKB;
1348 }
1349 if (!ops->ndo_bpf)
1350 return XDP_ATTACHED_NONE;
1351
1352 __dev_xdp_query(dev, ops->ndo_bpf, &xdp);
1353 *prog_id = xdp.prog_id;
1354
1355 return xdp.prog_attached;
1356 }
1357
1358 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1359 {
1360 struct nlattr *xdp;
1361 u32 prog_id;
1362 int err;
1363
1364 xdp = nla_nest_start(skb, IFLA_XDP);
1365 if (!xdp)
1366 return -EMSGSIZE;
1367
1368 err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1369 rtnl_xdp_attached_mode(dev, &prog_id));
1370 if (err)
1371 goto err_cancel;
1372
1373 if (prog_id) {
1374 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1375 if (err)
1376 goto err_cancel;
1377 }
1378
1379 nla_nest_end(skb, xdp);
1380 return 0;
1381
1382 err_cancel:
1383 nla_nest_cancel(skb, xdp);
1384 return err;
1385 }
1386
1387 static u32 rtnl_get_event(unsigned long event)
1388 {
1389 u32 rtnl_event_type = IFLA_EVENT_NONE;
1390
1391 switch (event) {
1392 case NETDEV_REBOOT:
1393 rtnl_event_type = IFLA_EVENT_REBOOT;
1394 break;
1395 case NETDEV_FEAT_CHANGE:
1396 rtnl_event_type = IFLA_EVENT_FEATURES;
1397 break;
1398 case NETDEV_BONDING_FAILOVER:
1399 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1400 break;
1401 case NETDEV_NOTIFY_PEERS:
1402 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1403 break;
1404 case NETDEV_RESEND_IGMP:
1405 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1406 break;
1407 case NETDEV_CHANGEINFODATA:
1408 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1409 break;
1410 default:
1411 break;
1412 }
1413
1414 return rtnl_event_type;
1415 }
1416
1417 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1418 {
1419 const struct net_device *upper_dev;
1420 int ret = 0;
1421
1422 rcu_read_lock();
1423
1424 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1425 if (upper_dev)
1426 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1427
1428 rcu_read_unlock();
1429 return ret;
1430 }
1431
1432 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1433 {
1434 int ifindex = dev_get_iflink(dev);
1435
1436 if (dev->ifindex == ifindex)
1437 return 0;
1438
1439 return nla_put_u32(skb, IFLA_LINK, ifindex);
1440 }
1441
1442 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1443 struct net_device *dev)
1444 {
1445 char buf[IFALIASZ];
1446 int ret;
1447
1448 ret = dev_get_alias(dev, buf, sizeof(buf));
1449 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1450 }
1451
1452 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1453 const struct net_device *dev,
1454 struct net *src_net)
1455 {
1456 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1457 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1458
1459 if (!net_eq(dev_net(dev), link_net)) {
1460 int id = peernet2id_alloc(src_net, link_net);
1461
1462 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1463 return -EMSGSIZE;
1464 }
1465 }
1466
1467 return 0;
1468 }
1469
1470 static int rtnl_fill_link_af(struct sk_buff *skb,
1471 const struct net_device *dev,
1472 u32 ext_filter_mask)
1473 {
1474 const struct rtnl_af_ops *af_ops;
1475 struct nlattr *af_spec;
1476
1477 af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1478 if (!af_spec)
1479 return -EMSGSIZE;
1480
1481 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1482 struct nlattr *af;
1483 int err;
1484
1485 if (!af_ops->fill_link_af)
1486 continue;
1487
1488 af = nla_nest_start(skb, af_ops->family);
1489 if (!af)
1490 return -EMSGSIZE;
1491
1492 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1493 /*
1494 * Caller may return ENODATA to indicate that there
1495 * was no data to be dumped. This is not an error, it
1496 * means we should trim the attribute header and
1497 * continue.
1498 */
1499 if (err == -ENODATA)
1500 nla_nest_cancel(skb, af);
1501 else if (err < 0)
1502 return -EMSGSIZE;
1503
1504 nla_nest_end(skb, af);
1505 }
1506
1507 nla_nest_end(skb, af_spec);
1508 return 0;
1509 }
1510
1511 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1512 struct net_device *dev, struct net *src_net,
1513 int type, u32 pid, u32 seq, u32 change,
1514 unsigned int flags, u32 ext_filter_mask,
1515 u32 event, int *new_nsid, int new_ifindex,
1516 int tgt_netnsid)
1517 {
1518 struct ifinfomsg *ifm;
1519 struct nlmsghdr *nlh;
1520
1521 ASSERT_RTNL();
1522 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1523 if (nlh == NULL)
1524 return -EMSGSIZE;
1525
1526 ifm = nlmsg_data(nlh);
1527 ifm->ifi_family = AF_UNSPEC;
1528 ifm->__ifi_pad = 0;
1529 ifm->ifi_type = dev->type;
1530 ifm->ifi_index = dev->ifindex;
1531 ifm->ifi_flags = dev_get_flags(dev);
1532 ifm->ifi_change = change;
1533
1534 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid))
1535 goto nla_put_failure;
1536
1537 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1538 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1539 nla_put_u8(skb, IFLA_OPERSTATE,
1540 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1541 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1542 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1543 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1544 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1545 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1546 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1547 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1548 #ifdef CONFIG_RPS
1549 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1550 #endif
1551 nla_put_iflink(skb, dev) ||
1552 put_master_ifindex(skb, dev) ||
1553 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1554 (dev->qdisc &&
1555 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1556 nla_put_ifalias(skb, dev) ||
1557 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1558 atomic_read(&dev->carrier_up_count) +
1559 atomic_read(&dev->carrier_down_count)) ||
1560 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1561 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1562 atomic_read(&dev->carrier_up_count)) ||
1563 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1564 atomic_read(&dev->carrier_down_count)))
1565 goto nla_put_failure;
1566
1567 if (event != IFLA_EVENT_NONE) {
1568 if (nla_put_u32(skb, IFLA_EVENT, event))
1569 goto nla_put_failure;
1570 }
1571
1572 if (rtnl_fill_link_ifmap(skb, dev))
1573 goto nla_put_failure;
1574
1575 if (dev->addr_len) {
1576 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1577 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1578 goto nla_put_failure;
1579 }
1580
1581 if (rtnl_phys_port_id_fill(skb, dev))
1582 goto nla_put_failure;
1583
1584 if (rtnl_phys_port_name_fill(skb, dev))
1585 goto nla_put_failure;
1586
1587 if (rtnl_phys_switch_id_fill(skb, dev))
1588 goto nla_put_failure;
1589
1590 if (rtnl_fill_stats(skb, dev))
1591 goto nla_put_failure;
1592
1593 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1594 goto nla_put_failure;
1595
1596 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1597 goto nla_put_failure;
1598
1599 if (rtnl_xdp_fill(skb, dev))
1600 goto nla_put_failure;
1601
1602 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1603 if (rtnl_link_fill(skb, dev) < 0)
1604 goto nla_put_failure;
1605 }
1606
1607 if (rtnl_fill_link_netnsid(skb, dev, src_net))
1608 goto nla_put_failure;
1609
1610 if (new_nsid &&
1611 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1612 goto nla_put_failure;
1613 if (new_ifindex &&
1614 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1615 goto nla_put_failure;
1616
1617
1618 rcu_read_lock();
1619 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1620 goto nla_put_failure_rcu;
1621 rcu_read_unlock();
1622
1623 nlmsg_end(skb, nlh);
1624 return 0;
1625
1626 nla_put_failure_rcu:
1627 rcu_read_unlock();
1628 nla_put_failure:
1629 nlmsg_cancel(skb, nlh);
1630 return -EMSGSIZE;
1631 }
1632
1633 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1634 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1635 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1636 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1637 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1638 [IFLA_MTU] = { .type = NLA_U32 },
1639 [IFLA_LINK] = { .type = NLA_U32 },
1640 [IFLA_MASTER] = { .type = NLA_U32 },
1641 [IFLA_CARRIER] = { .type = NLA_U8 },
1642 [IFLA_TXQLEN] = { .type = NLA_U32 },
1643 [IFLA_WEIGHT] = { .type = NLA_U32 },
1644 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1645 [IFLA_LINKMODE] = { .type = NLA_U8 },
1646 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1647 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1648 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1649 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1650 * allow 0-length string (needed to remove an alias).
1651 */
1652 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1653 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1654 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1655 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1656 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1657 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1658 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1659 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1660 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1661 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1662 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1663 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1664 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1665 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1666 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1667 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1668 [IFLA_XDP] = { .type = NLA_NESTED },
1669 [IFLA_EVENT] = { .type = NLA_U32 },
1670 [IFLA_GROUP] = { .type = NLA_U32 },
1671 [IFLA_IF_NETNSID] = { .type = NLA_S32 },
1672 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1673 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1674 };
1675
1676 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1677 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1678 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1679 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1680 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1681 };
1682
1683 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1684 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1685 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1686 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1687 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1688 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1689 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1690 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1691 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1692 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1693 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1694 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1695 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1696 };
1697
1698 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1699 [IFLA_PORT_VF] = { .type = NLA_U32 },
1700 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1701 .len = PORT_PROFILE_MAX },
1702 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1703 .len = PORT_UUID_MAX },
1704 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1705 .len = PORT_UUID_MAX },
1706 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1707 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1708
1709 /* Unused, but we need to keep it here since user space could
1710 * fill it. It's also broken with regard to NLA_BINARY use in
1711 * combination with structs.
1712 */
1713 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1714 .len = sizeof(struct ifla_port_vsi) },
1715 };
1716
1717 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1718 [IFLA_XDP_FD] = { .type = NLA_S32 },
1719 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1720 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1721 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1722 };
1723
1724 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1725 {
1726 const struct rtnl_link_ops *ops = NULL;
1727 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1728
1729 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1730 ifla_info_policy, NULL) < 0)
1731 return NULL;
1732
1733 if (linfo[IFLA_INFO_KIND]) {
1734 char kind[MODULE_NAME_LEN];
1735
1736 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1737 ops = rtnl_link_ops_get(kind);
1738 }
1739
1740 return ops;
1741 }
1742
1743 static bool link_master_filtered(struct net_device *dev, int master_idx)
1744 {
1745 struct net_device *master;
1746
1747 if (!master_idx)
1748 return false;
1749
1750 master = netdev_master_upper_dev_get(dev);
1751 if (!master || master->ifindex != master_idx)
1752 return true;
1753
1754 return false;
1755 }
1756
1757 static bool link_kind_filtered(const struct net_device *dev,
1758 const struct rtnl_link_ops *kind_ops)
1759 {
1760 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1761 return true;
1762
1763 return false;
1764 }
1765
1766 static bool link_dump_filtered(struct net_device *dev,
1767 int master_idx,
1768 const struct rtnl_link_ops *kind_ops)
1769 {
1770 if (link_master_filtered(dev, master_idx) ||
1771 link_kind_filtered(dev, kind_ops))
1772 return true;
1773
1774 return false;
1775 }
1776
1777 static struct net *get_target_net(struct sock *sk, int netnsid)
1778 {
1779 struct net *net;
1780
1781 net = get_net_ns_by_id(sock_net(sk), netnsid);
1782 if (!net)
1783 return ERR_PTR(-EINVAL);
1784
1785 /* For now, the caller is required to have CAP_NET_ADMIN in
1786 * the user namespace owning the target net ns.
1787 */
1788 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1789 put_net(net);
1790 return ERR_PTR(-EACCES);
1791 }
1792 return net;
1793 }
1794
1795 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1796 {
1797 struct net *net = sock_net(skb->sk);
1798 struct net *tgt_net = net;
1799 int h, s_h;
1800 int idx = 0, s_idx;
1801 struct net_device *dev;
1802 struct hlist_head *head;
1803 struct nlattr *tb[IFLA_MAX+1];
1804 u32 ext_filter_mask = 0;
1805 const struct rtnl_link_ops *kind_ops = NULL;
1806 unsigned int flags = NLM_F_MULTI;
1807 int master_idx = 0;
1808 int netnsid = -1;
1809 int err;
1810 int hdrlen;
1811
1812 s_h = cb->args[0];
1813 s_idx = cb->args[1];
1814
1815 /* A hack to preserve kernel<->userspace interface.
1816 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1817 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1818 * what iproute2 < v3.9.0 used.
1819 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1820 * attribute, its netlink message is shorter than struct ifinfomsg.
1821 */
1822 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1823 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1824
1825 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1826 ifla_policy, NULL) >= 0) {
1827 if (tb[IFLA_IF_NETNSID]) {
1828 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
1829 tgt_net = get_target_net(skb->sk, netnsid);
1830 if (IS_ERR(tgt_net)) {
1831 tgt_net = net;
1832 netnsid = -1;
1833 }
1834 }
1835
1836 if (tb[IFLA_EXT_MASK])
1837 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1838
1839 if (tb[IFLA_MASTER])
1840 master_idx = nla_get_u32(tb[IFLA_MASTER]);
1841
1842 if (tb[IFLA_LINKINFO])
1843 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1844
1845 if (master_idx || kind_ops)
1846 flags |= NLM_F_DUMP_FILTERED;
1847 }
1848
1849 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1850 idx = 0;
1851 head = &tgt_net->dev_index_head[h];
1852 hlist_for_each_entry(dev, head, index_hlist) {
1853 if (link_dump_filtered(dev, master_idx, kind_ops))
1854 goto cont;
1855 if (idx < s_idx)
1856 goto cont;
1857 err = rtnl_fill_ifinfo(skb, dev, net,
1858 RTM_NEWLINK,
1859 NETLINK_CB(cb->skb).portid,
1860 cb->nlh->nlmsg_seq, 0,
1861 flags,
1862 ext_filter_mask, 0, NULL, 0,
1863 netnsid);
1864
1865 if (err < 0) {
1866 if (likely(skb->len))
1867 goto out;
1868
1869 goto out_err;
1870 }
1871 cont:
1872 idx++;
1873 }
1874 }
1875 out:
1876 err = skb->len;
1877 out_err:
1878 cb->args[1] = idx;
1879 cb->args[0] = h;
1880 cb->seq = net->dev_base_seq;
1881 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1882 if (netnsid >= 0)
1883 put_net(tgt_net);
1884
1885 return err;
1886 }
1887
1888 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1889 struct netlink_ext_ack *exterr)
1890 {
1891 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1892 }
1893 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1894
1895 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1896 {
1897 struct net *net;
1898 /* Examine the link attributes and figure out which
1899 * network namespace we are talking about.
1900 */
1901 if (tb[IFLA_NET_NS_PID])
1902 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1903 else if (tb[IFLA_NET_NS_FD])
1904 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1905 else
1906 net = get_net(src_net);
1907 return net;
1908 }
1909 EXPORT_SYMBOL(rtnl_link_get_net);
1910
1911 /* Figure out which network namespace we are talking about by
1912 * examining the link attributes in the following order:
1913 *
1914 * 1. IFLA_NET_NS_PID
1915 * 2. IFLA_NET_NS_FD
1916 * 3. IFLA_IF_NETNSID
1917 */
1918 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
1919 struct nlattr *tb[])
1920 {
1921 struct net *net;
1922
1923 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
1924 return rtnl_link_get_net(src_net, tb);
1925
1926 if (!tb[IFLA_IF_NETNSID])
1927 return get_net(src_net);
1928
1929 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID]));
1930 if (!net)
1931 return ERR_PTR(-EINVAL);
1932
1933 return net;
1934 }
1935
1936 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
1937 struct net *src_net,
1938 struct nlattr *tb[], int cap)
1939 {
1940 struct net *net;
1941
1942 net = rtnl_link_get_net_by_nlattr(src_net, tb);
1943 if (IS_ERR(net))
1944 return net;
1945
1946 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
1947 put_net(net);
1948 return ERR_PTR(-EPERM);
1949 }
1950
1951 return net;
1952 }
1953
1954 /* Verify that rtnetlink requests do not pass additional properties
1955 * potentially referring to different network namespaces.
1956 */
1957 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
1958 struct netlink_ext_ack *extack,
1959 bool netns_id_only)
1960 {
1961
1962 if (netns_id_only) {
1963 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
1964 return 0;
1965
1966 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
1967 return -EOPNOTSUPP;
1968 }
1969
1970 if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
1971 goto invalid_attr;
1972
1973 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD]))
1974 goto invalid_attr;
1975
1976 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID]))
1977 goto invalid_attr;
1978
1979 return 0;
1980
1981 invalid_attr:
1982 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
1983 return -EINVAL;
1984 }
1985
1986 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1987 {
1988 if (dev) {
1989 if (tb[IFLA_ADDRESS] &&
1990 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1991 return -EINVAL;
1992
1993 if (tb[IFLA_BROADCAST] &&
1994 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1995 return -EINVAL;
1996 }
1997
1998 if (tb[IFLA_AF_SPEC]) {
1999 struct nlattr *af;
2000 int rem, err;
2001
2002 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2003 const struct rtnl_af_ops *af_ops;
2004
2005 rcu_read_lock();
2006 af_ops = rtnl_af_lookup(nla_type(af));
2007 if (!af_ops) {
2008 rcu_read_unlock();
2009 return -EAFNOSUPPORT;
2010 }
2011
2012 if (!af_ops->set_link_af) {
2013 rcu_read_unlock();
2014 return -EOPNOTSUPP;
2015 }
2016
2017 if (af_ops->validate_link_af) {
2018 err = af_ops->validate_link_af(dev, af);
2019 if (err < 0) {
2020 rcu_read_unlock();
2021 return err;
2022 }
2023 }
2024
2025 rcu_read_unlock();
2026 }
2027 }
2028
2029 return 0;
2030 }
2031
2032 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2033 int guid_type)
2034 {
2035 const struct net_device_ops *ops = dev->netdev_ops;
2036
2037 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2038 }
2039
2040 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2041 {
2042 if (dev->type != ARPHRD_INFINIBAND)
2043 return -EOPNOTSUPP;
2044
2045 return handle_infiniband_guid(dev, ivt, guid_type);
2046 }
2047
2048 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2049 {
2050 const struct net_device_ops *ops = dev->netdev_ops;
2051 int err = -EINVAL;
2052
2053 if (tb[IFLA_VF_MAC]) {
2054 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2055
2056 err = -EOPNOTSUPP;
2057 if (ops->ndo_set_vf_mac)
2058 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2059 ivm->mac);
2060 if (err < 0)
2061 return err;
2062 }
2063
2064 if (tb[IFLA_VF_VLAN]) {
2065 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2066
2067 err = -EOPNOTSUPP;
2068 if (ops->ndo_set_vf_vlan)
2069 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2070 ivv->qos,
2071 htons(ETH_P_8021Q));
2072 if (err < 0)
2073 return err;
2074 }
2075
2076 if (tb[IFLA_VF_VLAN_LIST]) {
2077 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2078 struct nlattr *attr;
2079 int rem, len = 0;
2080
2081 err = -EOPNOTSUPP;
2082 if (!ops->ndo_set_vf_vlan)
2083 return err;
2084
2085 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2086 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2087 nla_len(attr) < NLA_HDRLEN) {
2088 return -EINVAL;
2089 }
2090 if (len >= MAX_VLAN_LIST_LEN)
2091 return -EOPNOTSUPP;
2092 ivvl[len] = nla_data(attr);
2093
2094 len++;
2095 }
2096 if (len == 0)
2097 return -EINVAL;
2098
2099 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2100 ivvl[0]->qos, ivvl[0]->vlan_proto);
2101 if (err < 0)
2102 return err;
2103 }
2104
2105 if (tb[IFLA_VF_TX_RATE]) {
2106 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2107 struct ifla_vf_info ivf;
2108
2109 err = -EOPNOTSUPP;
2110 if (ops->ndo_get_vf_config)
2111 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2112 if (err < 0)
2113 return err;
2114
2115 err = -EOPNOTSUPP;
2116 if (ops->ndo_set_vf_rate)
2117 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2118 ivf.min_tx_rate,
2119 ivt->rate);
2120 if (err < 0)
2121 return err;
2122 }
2123
2124 if (tb[IFLA_VF_RATE]) {
2125 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2126
2127 err = -EOPNOTSUPP;
2128 if (ops->ndo_set_vf_rate)
2129 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2130 ivt->min_tx_rate,
2131 ivt->max_tx_rate);
2132 if (err < 0)
2133 return err;
2134 }
2135
2136 if (tb[IFLA_VF_SPOOFCHK]) {
2137 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2138
2139 err = -EOPNOTSUPP;
2140 if (ops->ndo_set_vf_spoofchk)
2141 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2142 ivs->setting);
2143 if (err < 0)
2144 return err;
2145 }
2146
2147 if (tb[IFLA_VF_LINK_STATE]) {
2148 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2149
2150 err = -EOPNOTSUPP;
2151 if (ops->ndo_set_vf_link_state)
2152 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2153 ivl->link_state);
2154 if (err < 0)
2155 return err;
2156 }
2157
2158 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2159 struct ifla_vf_rss_query_en *ivrssq_en;
2160
2161 err = -EOPNOTSUPP;
2162 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2163 if (ops->ndo_set_vf_rss_query_en)
2164 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2165 ivrssq_en->setting);
2166 if (err < 0)
2167 return err;
2168 }
2169
2170 if (tb[IFLA_VF_TRUST]) {
2171 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2172
2173 err = -EOPNOTSUPP;
2174 if (ops->ndo_set_vf_trust)
2175 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2176 if (err < 0)
2177 return err;
2178 }
2179
2180 if (tb[IFLA_VF_IB_NODE_GUID]) {
2181 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2182
2183 if (!ops->ndo_set_vf_guid)
2184 return -EOPNOTSUPP;
2185
2186 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2187 }
2188
2189 if (tb[IFLA_VF_IB_PORT_GUID]) {
2190 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2191
2192 if (!ops->ndo_set_vf_guid)
2193 return -EOPNOTSUPP;
2194
2195 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2196 }
2197
2198 return err;
2199 }
2200
2201 static int do_set_master(struct net_device *dev, int ifindex,
2202 struct netlink_ext_ack *extack)
2203 {
2204 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2205 const struct net_device_ops *ops;
2206 int err;
2207
2208 if (upper_dev) {
2209 if (upper_dev->ifindex == ifindex)
2210 return 0;
2211 ops = upper_dev->netdev_ops;
2212 if (ops->ndo_del_slave) {
2213 err = ops->ndo_del_slave(upper_dev, dev);
2214 if (err)
2215 return err;
2216 } else {
2217 return -EOPNOTSUPP;
2218 }
2219 }
2220
2221 if (ifindex) {
2222 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2223 if (!upper_dev)
2224 return -EINVAL;
2225 ops = upper_dev->netdev_ops;
2226 if (ops->ndo_add_slave) {
2227 err = ops->ndo_add_slave(upper_dev, dev, extack);
2228 if (err)
2229 return err;
2230 } else {
2231 return -EOPNOTSUPP;
2232 }
2233 }
2234 return 0;
2235 }
2236
2237 #define DO_SETLINK_MODIFIED 0x01
2238 /* notify flag means notify + modified. */
2239 #define DO_SETLINK_NOTIFY 0x03
2240 static int do_setlink(const struct sk_buff *skb,
2241 struct net_device *dev, struct ifinfomsg *ifm,
2242 struct netlink_ext_ack *extack,
2243 struct nlattr **tb, char *ifname, int status)
2244 {
2245 const struct net_device_ops *ops = dev->netdev_ops;
2246 int err;
2247
2248 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) {
2249 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2250 tb, CAP_NET_ADMIN);
2251 if (IS_ERR(net)) {
2252 err = PTR_ERR(net);
2253 goto errout;
2254 }
2255
2256 err = dev_change_net_namespace(dev, net, ifname);
2257 put_net(net);
2258 if (err)
2259 goto errout;
2260 status |= DO_SETLINK_MODIFIED;
2261 }
2262
2263 if (tb[IFLA_MAP]) {
2264 struct rtnl_link_ifmap *u_map;
2265 struct ifmap k_map;
2266
2267 if (!ops->ndo_set_config) {
2268 err = -EOPNOTSUPP;
2269 goto errout;
2270 }
2271
2272 if (!netif_device_present(dev)) {
2273 err = -ENODEV;
2274 goto errout;
2275 }
2276
2277 u_map = nla_data(tb[IFLA_MAP]);
2278 k_map.mem_start = (unsigned long) u_map->mem_start;
2279 k_map.mem_end = (unsigned long) u_map->mem_end;
2280 k_map.base_addr = (unsigned short) u_map->base_addr;
2281 k_map.irq = (unsigned char) u_map->irq;
2282 k_map.dma = (unsigned char) u_map->dma;
2283 k_map.port = (unsigned char) u_map->port;
2284
2285 err = ops->ndo_set_config(dev, &k_map);
2286 if (err < 0)
2287 goto errout;
2288
2289 status |= DO_SETLINK_NOTIFY;
2290 }
2291
2292 if (tb[IFLA_ADDRESS]) {
2293 struct sockaddr *sa;
2294 int len;
2295
2296 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2297 sizeof(*sa));
2298 sa = kmalloc(len, GFP_KERNEL);
2299 if (!sa) {
2300 err = -ENOMEM;
2301 goto errout;
2302 }
2303 sa->sa_family = dev->type;
2304 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2305 dev->addr_len);
2306 err = dev_set_mac_address(dev, sa);
2307 kfree(sa);
2308 if (err)
2309 goto errout;
2310 status |= DO_SETLINK_MODIFIED;
2311 }
2312
2313 if (tb[IFLA_MTU]) {
2314 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2315 if (err < 0)
2316 goto errout;
2317 status |= DO_SETLINK_MODIFIED;
2318 }
2319
2320 if (tb[IFLA_GROUP]) {
2321 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2322 status |= DO_SETLINK_NOTIFY;
2323 }
2324
2325 /*
2326 * Interface selected by interface index but interface
2327 * name provided implies that a name change has been
2328 * requested.
2329 */
2330 if (ifm->ifi_index > 0 && ifname[0]) {
2331 err = dev_change_name(dev, ifname);
2332 if (err < 0)
2333 goto errout;
2334 status |= DO_SETLINK_MODIFIED;
2335 }
2336
2337 if (tb[IFLA_IFALIAS]) {
2338 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2339 nla_len(tb[IFLA_IFALIAS]));
2340 if (err < 0)
2341 goto errout;
2342 status |= DO_SETLINK_NOTIFY;
2343 }
2344
2345 if (tb[IFLA_BROADCAST]) {
2346 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2347 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2348 }
2349
2350 if (ifm->ifi_flags || ifm->ifi_change) {
2351 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2352 if (err < 0)
2353 goto errout;
2354 }
2355
2356 if (tb[IFLA_MASTER]) {
2357 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2358 if (err)
2359 goto errout;
2360 status |= DO_SETLINK_MODIFIED;
2361 }
2362
2363 if (tb[IFLA_CARRIER]) {
2364 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2365 if (err)
2366 goto errout;
2367 status |= DO_SETLINK_MODIFIED;
2368 }
2369
2370 if (tb[IFLA_TXQLEN]) {
2371 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2372
2373 err = dev_change_tx_queue_len(dev, value);
2374 if (err)
2375 goto errout;
2376 status |= DO_SETLINK_MODIFIED;
2377 }
2378
2379 if (tb[IFLA_GSO_MAX_SIZE]) {
2380 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2381
2382 if (max_size > GSO_MAX_SIZE) {
2383 err = -EINVAL;
2384 goto errout;
2385 }
2386
2387 if (dev->gso_max_size ^ max_size) {
2388 netif_set_gso_max_size(dev, max_size);
2389 status |= DO_SETLINK_MODIFIED;
2390 }
2391 }
2392
2393 if (tb[IFLA_GSO_MAX_SEGS]) {
2394 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2395
2396 if (max_segs > GSO_MAX_SEGS) {
2397 err = -EINVAL;
2398 goto errout;
2399 }
2400
2401 if (dev->gso_max_segs ^ max_segs) {
2402 dev->gso_max_segs = max_segs;
2403 status |= DO_SETLINK_MODIFIED;
2404 }
2405 }
2406
2407 if (tb[IFLA_OPERSTATE])
2408 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2409
2410 if (tb[IFLA_LINKMODE]) {
2411 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2412
2413 write_lock_bh(&dev_base_lock);
2414 if (dev->link_mode ^ value)
2415 status |= DO_SETLINK_NOTIFY;
2416 dev->link_mode = value;
2417 write_unlock_bh(&dev_base_lock);
2418 }
2419
2420 if (tb[IFLA_VFINFO_LIST]) {
2421 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2422 struct nlattr *attr;
2423 int rem;
2424
2425 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2426 if (nla_type(attr) != IFLA_VF_INFO ||
2427 nla_len(attr) < NLA_HDRLEN) {
2428 err = -EINVAL;
2429 goto errout;
2430 }
2431 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2432 ifla_vf_policy, NULL);
2433 if (err < 0)
2434 goto errout;
2435 err = do_setvfinfo(dev, vfinfo);
2436 if (err < 0)
2437 goto errout;
2438 status |= DO_SETLINK_NOTIFY;
2439 }
2440 }
2441 err = 0;
2442
2443 if (tb[IFLA_VF_PORTS]) {
2444 struct nlattr *port[IFLA_PORT_MAX+1];
2445 struct nlattr *attr;
2446 int vf;
2447 int rem;
2448
2449 err = -EOPNOTSUPP;
2450 if (!ops->ndo_set_vf_port)
2451 goto errout;
2452
2453 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2454 if (nla_type(attr) != IFLA_VF_PORT ||
2455 nla_len(attr) < NLA_HDRLEN) {
2456 err = -EINVAL;
2457 goto errout;
2458 }
2459 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2460 ifla_port_policy, NULL);
2461 if (err < 0)
2462 goto errout;
2463 if (!port[IFLA_PORT_VF]) {
2464 err = -EOPNOTSUPP;
2465 goto errout;
2466 }
2467 vf = nla_get_u32(port[IFLA_PORT_VF]);
2468 err = ops->ndo_set_vf_port(dev, vf, port);
2469 if (err < 0)
2470 goto errout;
2471 status |= DO_SETLINK_NOTIFY;
2472 }
2473 }
2474 err = 0;
2475
2476 if (tb[IFLA_PORT_SELF]) {
2477 struct nlattr *port[IFLA_PORT_MAX+1];
2478
2479 err = nla_parse_nested(port, IFLA_PORT_MAX,
2480 tb[IFLA_PORT_SELF], ifla_port_policy,
2481 NULL);
2482 if (err < 0)
2483 goto errout;
2484
2485 err = -EOPNOTSUPP;
2486 if (ops->ndo_set_vf_port)
2487 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2488 if (err < 0)
2489 goto errout;
2490 status |= DO_SETLINK_NOTIFY;
2491 }
2492
2493 if (tb[IFLA_AF_SPEC]) {
2494 struct nlattr *af;
2495 int rem;
2496
2497 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2498 const struct rtnl_af_ops *af_ops;
2499
2500 rcu_read_lock();
2501
2502 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2503
2504 err = af_ops->set_link_af(dev, af);
2505 if (err < 0) {
2506 rcu_read_unlock();
2507 goto errout;
2508 }
2509
2510 rcu_read_unlock();
2511 status |= DO_SETLINK_NOTIFY;
2512 }
2513 }
2514 err = 0;
2515
2516 if (tb[IFLA_PROTO_DOWN]) {
2517 err = dev_change_proto_down(dev,
2518 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2519 if (err)
2520 goto errout;
2521 status |= DO_SETLINK_NOTIFY;
2522 }
2523
2524 if (tb[IFLA_XDP]) {
2525 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2526 u32 xdp_flags = 0;
2527
2528 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2529 ifla_xdp_policy, NULL);
2530 if (err < 0)
2531 goto errout;
2532
2533 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2534 err = -EINVAL;
2535 goto errout;
2536 }
2537
2538 if (xdp[IFLA_XDP_FLAGS]) {
2539 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2540 if (xdp_flags & ~XDP_FLAGS_MASK) {
2541 err = -EINVAL;
2542 goto errout;
2543 }
2544 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2545 err = -EINVAL;
2546 goto errout;
2547 }
2548 }
2549
2550 if (xdp[IFLA_XDP_FD]) {
2551 err = dev_change_xdp_fd(dev, extack,
2552 nla_get_s32(xdp[IFLA_XDP_FD]),
2553 xdp_flags);
2554 if (err)
2555 goto errout;
2556 status |= DO_SETLINK_NOTIFY;
2557 }
2558 }
2559
2560 errout:
2561 if (status & DO_SETLINK_MODIFIED) {
2562 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2563 netdev_state_change(dev);
2564
2565 if (err < 0)
2566 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",
2567 dev->name);
2568 }
2569
2570 return err;
2571 }
2572
2573 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2574 struct netlink_ext_ack *extack)
2575 {
2576 struct net *net = sock_net(skb->sk);
2577 struct ifinfomsg *ifm;
2578 struct net_device *dev;
2579 int err;
2580 struct nlattr *tb[IFLA_MAX+1];
2581 char ifname[IFNAMSIZ];
2582
2583 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2584 extack);
2585 if (err < 0)
2586 goto errout;
2587
2588 err = rtnl_ensure_unique_netns(tb, extack, false);
2589 if (err < 0)
2590 goto errout;
2591
2592 if (tb[IFLA_IFNAME])
2593 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2594 else
2595 ifname[0] = '\0';
2596
2597 err = -EINVAL;
2598 ifm = nlmsg_data(nlh);
2599 if (ifm->ifi_index > 0)
2600 dev = __dev_get_by_index(net, ifm->ifi_index);
2601 else if (tb[IFLA_IFNAME])
2602 dev = __dev_get_by_name(net, ifname);
2603 else
2604 goto errout;
2605
2606 if (dev == NULL) {
2607 err = -ENODEV;
2608 goto errout;
2609 }
2610
2611 err = validate_linkmsg(dev, tb);
2612 if (err < 0)
2613 goto errout;
2614
2615 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2616 errout:
2617 return err;
2618 }
2619
2620 static int rtnl_group_dellink(const struct net *net, int group)
2621 {
2622 struct net_device *dev, *aux;
2623 LIST_HEAD(list_kill);
2624 bool found = false;
2625
2626 if (!group)
2627 return -EPERM;
2628
2629 for_each_netdev(net, dev) {
2630 if (dev->group == group) {
2631 const struct rtnl_link_ops *ops;
2632
2633 found = true;
2634 ops = dev->rtnl_link_ops;
2635 if (!ops || !ops->dellink)
2636 return -EOPNOTSUPP;
2637 }
2638 }
2639
2640 if (!found)
2641 return -ENODEV;
2642
2643 for_each_netdev_safe(net, dev, aux) {
2644 if (dev->group == group) {
2645 const struct rtnl_link_ops *ops;
2646
2647 ops = dev->rtnl_link_ops;
2648 ops->dellink(dev, &list_kill);
2649 }
2650 }
2651 unregister_netdevice_many(&list_kill);
2652
2653 return 0;
2654 }
2655
2656 int rtnl_delete_link(struct net_device *dev)
2657 {
2658 const struct rtnl_link_ops *ops;
2659 LIST_HEAD(list_kill);
2660
2661 ops = dev->rtnl_link_ops;
2662 if (!ops || !ops->dellink)
2663 return -EOPNOTSUPP;
2664
2665 ops->dellink(dev, &list_kill);
2666 unregister_netdevice_many(&list_kill);
2667
2668 return 0;
2669 }
2670 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2671
2672 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2673 struct netlink_ext_ack *extack)
2674 {
2675 struct net *net = sock_net(skb->sk);
2676 struct net *tgt_net = net;
2677 struct net_device *dev = NULL;
2678 struct ifinfomsg *ifm;
2679 char ifname[IFNAMSIZ];
2680 struct nlattr *tb[IFLA_MAX+1];
2681 int err;
2682 int netnsid = -1;
2683
2684 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2685 if (err < 0)
2686 return err;
2687
2688 err = rtnl_ensure_unique_netns(tb, extack, true);
2689 if (err < 0)
2690 return err;
2691
2692 if (tb[IFLA_IFNAME])
2693 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2694
2695 if (tb[IFLA_IF_NETNSID]) {
2696 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2697 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2698 if (IS_ERR(tgt_net))
2699 return PTR_ERR(tgt_net);
2700 }
2701
2702 err = -EINVAL;
2703 ifm = nlmsg_data(nlh);
2704 if (ifm->ifi_index > 0)
2705 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2706 else if (tb[IFLA_IFNAME])
2707 dev = __dev_get_by_name(tgt_net, ifname);
2708 else if (tb[IFLA_GROUP])
2709 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2710 else
2711 goto out;
2712
2713 if (!dev) {
2714 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2715 err = -ENODEV;
2716
2717 goto out;
2718 }
2719
2720 err = rtnl_delete_link(dev);
2721
2722 out:
2723 if (netnsid >= 0)
2724 put_net(tgt_net);
2725
2726 return err;
2727 }
2728
2729 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2730 {
2731 unsigned int old_flags;
2732 int err;
2733
2734 old_flags = dev->flags;
2735 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2736 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2737 if (err < 0)
2738 return err;
2739 }
2740
2741 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2742
2743 __dev_notify_flags(dev, old_flags, ~0U);
2744 return 0;
2745 }
2746 EXPORT_SYMBOL(rtnl_configure_link);
2747
2748 struct net_device *rtnl_create_link(struct net *net,
2749 const char *ifname, unsigned char name_assign_type,
2750 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2751 {
2752 struct net_device *dev;
2753 unsigned int num_tx_queues = 1;
2754 unsigned int num_rx_queues = 1;
2755
2756 if (tb[IFLA_NUM_TX_QUEUES])
2757 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2758 else if (ops->get_num_tx_queues)
2759 num_tx_queues = ops->get_num_tx_queues();
2760
2761 if (tb[IFLA_NUM_RX_QUEUES])
2762 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2763 else if (ops->get_num_rx_queues)
2764 num_rx_queues = ops->get_num_rx_queues();
2765
2766 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2767 ops->setup, num_tx_queues, num_rx_queues);
2768 if (!dev)
2769 return ERR_PTR(-ENOMEM);
2770
2771 dev_net_set(dev, net);
2772 dev->rtnl_link_ops = ops;
2773 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2774
2775 if (tb[IFLA_MTU])
2776 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2777 if (tb[IFLA_ADDRESS]) {
2778 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2779 nla_len(tb[IFLA_ADDRESS]));
2780 dev->addr_assign_type = NET_ADDR_SET;
2781 }
2782 if (tb[IFLA_BROADCAST])
2783 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2784 nla_len(tb[IFLA_BROADCAST]));
2785 if (tb[IFLA_TXQLEN])
2786 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2787 if (tb[IFLA_OPERSTATE])
2788 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2789 if (tb[IFLA_LINKMODE])
2790 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2791 if (tb[IFLA_GROUP])
2792 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2793 if (tb[IFLA_GSO_MAX_SIZE])
2794 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2795 if (tb[IFLA_GSO_MAX_SEGS])
2796 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2797
2798 return dev;
2799 }
2800 EXPORT_SYMBOL(rtnl_create_link);
2801
2802 static int rtnl_group_changelink(const struct sk_buff *skb,
2803 struct net *net, int group,
2804 struct ifinfomsg *ifm,
2805 struct netlink_ext_ack *extack,
2806 struct nlattr **tb)
2807 {
2808 struct net_device *dev, *aux;
2809 int err;
2810
2811 for_each_netdev_safe(net, dev, aux) {
2812 if (dev->group == group) {
2813 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2814 if (err < 0)
2815 return err;
2816 }
2817 }
2818
2819 return 0;
2820 }
2821
2822 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2823 struct netlink_ext_ack *extack)
2824 {
2825 struct net *net = sock_net(skb->sk);
2826 const struct rtnl_link_ops *ops;
2827 const struct rtnl_link_ops *m_ops = NULL;
2828 struct net_device *dev;
2829 struct net_device *master_dev = NULL;
2830 struct ifinfomsg *ifm;
2831 char kind[MODULE_NAME_LEN];
2832 char ifname[IFNAMSIZ];
2833 struct nlattr *tb[IFLA_MAX+1];
2834 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2835 unsigned char name_assign_type = NET_NAME_USER;
2836 int err;
2837
2838 #ifdef CONFIG_MODULES
2839 replay:
2840 #endif
2841 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2842 if (err < 0)
2843 return err;
2844
2845 err = rtnl_ensure_unique_netns(tb, extack, false);
2846 if (err < 0)
2847 return err;
2848
2849 if (tb[IFLA_IFNAME])
2850 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2851 else
2852 ifname[0] = '\0';
2853
2854 ifm = nlmsg_data(nlh);
2855 if (ifm->ifi_index > 0)
2856 dev = __dev_get_by_index(net, ifm->ifi_index);
2857 else {
2858 if (ifname[0])
2859 dev = __dev_get_by_name(net, ifname);
2860 else
2861 dev = NULL;
2862 }
2863
2864 if (dev) {
2865 master_dev = netdev_master_upper_dev_get(dev);
2866 if (master_dev)
2867 m_ops = master_dev->rtnl_link_ops;
2868 }
2869
2870 err = validate_linkmsg(dev, tb);
2871 if (err < 0)
2872 return err;
2873
2874 if (tb[IFLA_LINKINFO]) {
2875 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2876 tb[IFLA_LINKINFO], ifla_info_policy,
2877 NULL);
2878 if (err < 0)
2879 return err;
2880 } else
2881 memset(linkinfo, 0, sizeof(linkinfo));
2882
2883 if (linkinfo[IFLA_INFO_KIND]) {
2884 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2885 ops = rtnl_link_ops_get(kind);
2886 } else {
2887 kind[0] = '\0';
2888 ops = NULL;
2889 }
2890
2891 if (1) {
2892 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2893 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2894 struct nlattr **data = NULL;
2895 struct nlattr **slave_data = NULL;
2896 struct net *dest_net, *link_net = NULL;
2897
2898 if (ops) {
2899 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2900 err = nla_parse_nested(attr, ops->maxtype,
2901 linkinfo[IFLA_INFO_DATA],
2902 ops->policy, NULL);
2903 if (err < 0)
2904 return err;
2905 data = attr;
2906 }
2907 if (ops->validate) {
2908 err = ops->validate(tb, data, extack);
2909 if (err < 0)
2910 return err;
2911 }
2912 }
2913
2914 if (m_ops) {
2915 if (m_ops->slave_maxtype &&
2916 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2917 err = nla_parse_nested(slave_attr,
2918 m_ops->slave_maxtype,
2919 linkinfo[IFLA_INFO_SLAVE_DATA],
2920 m_ops->slave_policy,
2921 NULL);
2922 if (err < 0)
2923 return err;
2924 slave_data = slave_attr;
2925 }
2926 }
2927
2928 if (dev) {
2929 int status = 0;
2930
2931 if (nlh->nlmsg_flags & NLM_F_EXCL)
2932 return -EEXIST;
2933 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2934 return -EOPNOTSUPP;
2935
2936 if (linkinfo[IFLA_INFO_DATA]) {
2937 if (!ops || ops != dev->rtnl_link_ops ||
2938 !ops->changelink)
2939 return -EOPNOTSUPP;
2940
2941 err = ops->changelink(dev, tb, data, extack);
2942 if (err < 0)
2943 return err;
2944 status |= DO_SETLINK_NOTIFY;
2945 }
2946
2947 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2948 if (!m_ops || !m_ops->slave_changelink)
2949 return -EOPNOTSUPP;
2950
2951 err = m_ops->slave_changelink(master_dev, dev,
2952 tb, slave_data,
2953 extack);
2954 if (err < 0)
2955 return err;
2956 status |= DO_SETLINK_NOTIFY;
2957 }
2958
2959 return do_setlink(skb, dev, ifm, extack, tb, ifname,
2960 status);
2961 }
2962
2963 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2964 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2965 return rtnl_group_changelink(skb, net,
2966 nla_get_u32(tb[IFLA_GROUP]),
2967 ifm, extack, tb);
2968 return -ENODEV;
2969 }
2970
2971 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2972 return -EOPNOTSUPP;
2973
2974 if (!ops) {
2975 #ifdef CONFIG_MODULES
2976 if (kind[0]) {
2977 __rtnl_unlock();
2978 request_module("rtnl-link-%s", kind);
2979 rtnl_lock();
2980 ops = rtnl_link_ops_get(kind);
2981 if (ops)
2982 goto replay;
2983 }
2984 #endif
2985 return -EOPNOTSUPP;
2986 }
2987
2988 if (!ops->setup)
2989 return -EOPNOTSUPP;
2990
2991 if (!ifname[0]) {
2992 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2993 name_assign_type = NET_NAME_ENUM;
2994 }
2995
2996 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
2997 if (IS_ERR(dest_net))
2998 return PTR_ERR(dest_net);
2999
3000 if (tb[IFLA_LINK_NETNSID]) {
3001 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3002
3003 link_net = get_net_ns_by_id(dest_net, id);
3004 if (!link_net) {
3005 err = -EINVAL;
3006 goto out;
3007 }
3008 err = -EPERM;
3009 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3010 goto out;
3011 }
3012
3013 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3014 name_assign_type, ops, tb);
3015 if (IS_ERR(dev)) {
3016 err = PTR_ERR(dev);
3017 goto out;
3018 }
3019
3020 dev->ifindex = ifm->ifi_index;
3021
3022 if (ops->newlink) {
3023 err = ops->newlink(link_net ? : net, dev, tb, data,
3024 extack);
3025 /* Drivers should call free_netdev() in ->destructor
3026 * and unregister it on failure after registration
3027 * so that device could be finally freed in rtnl_unlock.
3028 */
3029 if (err < 0) {
3030 /* If device is not registered at all, free it now */
3031 if (dev->reg_state == NETREG_UNINITIALIZED)
3032 free_netdev(dev);
3033 goto out;
3034 }
3035 } else {
3036 err = register_netdevice(dev);
3037 if (err < 0) {
3038 free_netdev(dev);
3039 goto out;
3040 }
3041 }
3042 err = rtnl_configure_link(dev, ifm);
3043 if (err < 0)
3044 goto out_unregister;
3045 if (link_net) {
3046 err = dev_change_net_namespace(dev, dest_net, ifname);
3047 if (err < 0)
3048 goto out_unregister;
3049 }
3050 if (tb[IFLA_MASTER]) {
3051 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
3052 extack);
3053 if (err)
3054 goto out_unregister;
3055 }
3056 out:
3057 if (link_net)
3058 put_net(link_net);
3059 put_net(dest_net);
3060 return err;
3061 out_unregister:
3062 if (ops->newlink) {
3063 LIST_HEAD(list_kill);
3064
3065 ops->dellink(dev, &list_kill);
3066 unregister_netdevice_many(&list_kill);
3067 } else {
3068 unregister_netdevice(dev);
3069 }
3070 goto out;
3071 }
3072 }
3073
3074 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3075 struct netlink_ext_ack *extack)
3076 {
3077 struct net *net = sock_net(skb->sk);
3078 struct net *tgt_net = net;
3079 struct ifinfomsg *ifm;
3080 char ifname[IFNAMSIZ];
3081 struct nlattr *tb[IFLA_MAX+1];
3082 struct net_device *dev = NULL;
3083 struct sk_buff *nskb;
3084 int netnsid = -1;
3085 int err;
3086 u32 ext_filter_mask = 0;
3087
3088 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3089 if (err < 0)
3090 return err;
3091
3092 err = rtnl_ensure_unique_netns(tb, extack, true);
3093 if (err < 0)
3094 return err;
3095
3096 if (tb[IFLA_IF_NETNSID]) {
3097 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
3098 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
3099 if (IS_ERR(tgt_net))
3100 return PTR_ERR(tgt_net);
3101 }
3102
3103 if (tb[IFLA_IFNAME])
3104 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3105
3106 if (tb[IFLA_EXT_MASK])
3107 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3108
3109 err = -EINVAL;
3110 ifm = nlmsg_data(nlh);
3111 if (ifm->ifi_index > 0)
3112 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3113 else if (tb[IFLA_IFNAME])
3114 dev = __dev_get_by_name(tgt_net, ifname);
3115 else
3116 goto out;
3117
3118 err = -ENODEV;
3119 if (dev == NULL)
3120 goto out;
3121
3122 err = -ENOBUFS;
3123 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3124 if (nskb == NULL)
3125 goto out;
3126
3127 err = rtnl_fill_ifinfo(nskb, dev, net,
3128 RTM_NEWLINK, NETLINK_CB(skb).portid,
3129 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3130 0, NULL, 0, netnsid);
3131 if (err < 0) {
3132 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3133 WARN_ON(err == -EMSGSIZE);
3134 kfree_skb(nskb);
3135 } else
3136 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3137 out:
3138 if (netnsid >= 0)
3139 put_net(tgt_net);
3140
3141 return err;
3142 }
3143
3144 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3145 {
3146 struct net *net = sock_net(skb->sk);
3147 struct net_device *dev;
3148 struct nlattr *tb[IFLA_MAX+1];
3149 u32 ext_filter_mask = 0;
3150 u16 min_ifinfo_dump_size = 0;
3151 int hdrlen;
3152
3153 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3154 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3155 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3156
3157 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3158 if (tb[IFLA_EXT_MASK])
3159 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3160 }
3161
3162 if (!ext_filter_mask)
3163 return NLMSG_GOODSIZE;
3164 /*
3165 * traverse the list of net devices and compute the minimum
3166 * buffer size based upon the filter mask.
3167 */
3168 rcu_read_lock();
3169 for_each_netdev_rcu(net, dev) {
3170 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3171 if_nlmsg_size(dev,
3172 ext_filter_mask));
3173 }
3174 rcu_read_unlock();
3175
3176 return nlmsg_total_size(min_ifinfo_dump_size);
3177 }
3178
3179 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3180 {
3181 int idx;
3182 int s_idx = cb->family;
3183
3184 if (s_idx == 0)
3185 s_idx = 1;
3186
3187 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3188 struct rtnl_link **tab;
3189 int type = cb->nlh->nlmsg_type-RTM_BASE;
3190 struct rtnl_link *link;
3191 rtnl_dumpit_func dumpit;
3192
3193 if (idx < s_idx || idx == PF_PACKET)
3194 continue;
3195
3196 if (type < 0 || type >= RTM_NR_MSGTYPES)
3197 continue;
3198
3199 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3200 if (!tab)
3201 continue;
3202
3203 link = tab[type];
3204 if (!link)
3205 continue;
3206
3207 dumpit = link->dumpit;
3208 if (!dumpit)
3209 continue;
3210
3211 if (idx > s_idx) {
3212 memset(&cb->args[0], 0, sizeof(cb->args));
3213 cb->prev_seq = 0;
3214 cb->seq = 0;
3215 }
3216 if (dumpit(skb, cb))
3217 break;
3218 }
3219 cb->family = idx;
3220
3221 return skb->len;
3222 }
3223
3224 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3225 unsigned int change,
3226 u32 event, gfp_t flags, int *new_nsid,
3227 int new_ifindex)
3228 {
3229 struct net *net = dev_net(dev);
3230 struct sk_buff *skb;
3231 int err = -ENOBUFS;
3232 size_t if_info_size;
3233
3234 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3235 if (skb == NULL)
3236 goto errout;
3237
3238 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3239 type, 0, 0, change, 0, 0, event,
3240 new_nsid, new_ifindex, -1);
3241 if (err < 0) {
3242 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3243 WARN_ON(err == -EMSGSIZE);
3244 kfree_skb(skb);
3245 goto errout;
3246 }
3247 return skb;
3248 errout:
3249 if (err < 0)
3250 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3251 return NULL;
3252 }
3253
3254 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3255 {
3256 struct net *net = dev_net(dev);
3257
3258 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3259 }
3260
3261 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3262 unsigned int change, u32 event,
3263 gfp_t flags, int *new_nsid, int new_ifindex)
3264 {
3265 struct sk_buff *skb;
3266
3267 if (dev->reg_state != NETREG_REGISTERED)
3268 return;
3269
3270 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3271 new_ifindex);
3272 if (skb)
3273 rtmsg_ifinfo_send(skb, dev, flags);
3274 }
3275
3276 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3277 gfp_t flags)
3278 {
3279 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3280 NULL, 0);
3281 }
3282
3283 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3284 gfp_t flags, int *new_nsid, int new_ifindex)
3285 {
3286 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3287 new_nsid, new_ifindex);
3288 }
3289
3290 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3291 struct net_device *dev,
3292 u8 *addr, u16 vid, u32 pid, u32 seq,
3293 int type, unsigned int flags,
3294 int nlflags, u16 ndm_state)
3295 {
3296 struct nlmsghdr *nlh;
3297 struct ndmsg *ndm;
3298
3299 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3300 if (!nlh)
3301 return -EMSGSIZE;
3302
3303 ndm = nlmsg_data(nlh);
3304 ndm->ndm_family = AF_BRIDGE;
3305 ndm->ndm_pad1 = 0;
3306 ndm->ndm_pad2 = 0;
3307 ndm->ndm_flags = flags;
3308 ndm->ndm_type = 0;
3309 ndm->ndm_ifindex = dev->ifindex;
3310 ndm->ndm_state = ndm_state;
3311
3312 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3313 goto nla_put_failure;
3314 if (vid)
3315 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3316 goto nla_put_failure;
3317
3318 nlmsg_end(skb, nlh);
3319 return 0;
3320
3321 nla_put_failure:
3322 nlmsg_cancel(skb, nlh);
3323 return -EMSGSIZE;
3324 }
3325
3326 static inline size_t rtnl_fdb_nlmsg_size(void)
3327 {
3328 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3329 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3330 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3331 0;
3332 }
3333
3334 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3335 u16 ndm_state)
3336 {
3337 struct net *net = dev_net(dev);
3338 struct sk_buff *skb;
3339 int err = -ENOBUFS;
3340
3341 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3342 if (!skb)
3343 goto errout;
3344
3345 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3346 0, 0, type, NTF_SELF, 0, ndm_state);
3347 if (err < 0) {
3348 kfree_skb(skb);
3349 goto errout;
3350 }
3351
3352 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3353 return;
3354 errout:
3355 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3356 }
3357
3358 /**
3359 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3360 */
3361 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3362 struct nlattr *tb[],
3363 struct net_device *dev,
3364 const unsigned char *addr, u16 vid,
3365 u16 flags)
3366 {
3367 int err = -EINVAL;
3368
3369 /* If aging addresses are supported device will need to
3370 * implement its own handler for this.
3371 */
3372 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3373 pr_info("%s: FDB only supports static addresses\n", dev->name);
3374 return err;
3375 }
3376
3377 if (vid) {
3378 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3379 return err;
3380 }
3381
3382 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3383 err = dev_uc_add_excl(dev, addr);
3384 else if (is_multicast_ether_addr(addr))
3385 err = dev_mc_add_excl(dev, addr);
3386
3387 /* Only return duplicate errors if NLM_F_EXCL is set */
3388 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3389 err = 0;
3390
3391 return err;
3392 }
3393 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3394
3395 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3396 struct netlink_ext_ack *extack)
3397 {
3398 u16 vid = 0;
3399
3400 if (vlan_attr) {
3401 if (nla_len(vlan_attr) != sizeof(u16)) {
3402 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3403 return -EINVAL;
3404 }
3405
3406 vid = nla_get_u16(vlan_attr);
3407
3408 if (!vid || vid >= VLAN_VID_MASK) {
3409 NL_SET_ERR_MSG(extack, "invalid vlan id");
3410 return -EINVAL;
3411 }
3412 }
3413 *p_vid = vid;
3414 return 0;
3415 }
3416
3417 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3418 struct netlink_ext_ack *extack)
3419 {
3420 struct net *net = sock_net(skb->sk);
3421 struct ndmsg *ndm;
3422 struct nlattr *tb[NDA_MAX+1];
3423 struct net_device *dev;
3424 u8 *addr;
3425 u16 vid;
3426 int err;
3427
3428 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3429 if (err < 0)
3430 return err;
3431
3432 ndm = nlmsg_data(nlh);
3433 if (ndm->ndm_ifindex == 0) {
3434 NL_SET_ERR_MSG(extack, "invalid ifindex");
3435 return -EINVAL;
3436 }
3437
3438 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3439 if (dev == NULL) {
3440 NL_SET_ERR_MSG(extack, "unknown ifindex");
3441 return -ENODEV;
3442 }
3443
3444 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3445 NL_SET_ERR_MSG(extack, "invalid address");
3446 return -EINVAL;
3447 }
3448
3449 addr = nla_data(tb[NDA_LLADDR]);
3450
3451 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3452 if (err)
3453 return err;
3454
3455 err = -EOPNOTSUPP;
3456
3457 /* Support fdb on master device the net/bridge default case */
3458 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3459 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3460 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3461 const struct net_device_ops *ops = br_dev->netdev_ops;
3462
3463 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3464 nlh->nlmsg_flags);
3465 if (err)
3466 goto out;
3467 else
3468 ndm->ndm_flags &= ~NTF_MASTER;
3469 }
3470
3471 /* Embedded bridge, macvlan, and any other device support */
3472 if ((ndm->ndm_flags & NTF_SELF)) {
3473 if (dev->netdev_ops->ndo_fdb_add)
3474 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3475 vid,
3476 nlh->nlmsg_flags);
3477 else
3478 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3479 nlh->nlmsg_flags);
3480
3481 if (!err) {
3482 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3483 ndm->ndm_state);
3484 ndm->ndm_flags &= ~NTF_SELF;
3485 }
3486 }
3487 out:
3488 return err;
3489 }
3490
3491 /**
3492 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3493 */
3494 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3495 struct nlattr *tb[],
3496 struct net_device *dev,
3497 const unsigned char *addr, u16 vid)
3498 {
3499 int err = -EINVAL;
3500
3501 /* If aging addresses are supported device will need to
3502 * implement its own handler for this.
3503 */
3504 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3505 pr_info("%s: FDB only supports static addresses\n", dev->name);
3506 return err;
3507 }
3508
3509 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3510 err = dev_uc_del(dev, addr);
3511 else if (is_multicast_ether_addr(addr))
3512 err = dev_mc_del(dev, addr);
3513
3514 return err;
3515 }
3516 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3517
3518 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3519 struct netlink_ext_ack *extack)
3520 {
3521 struct net *net = sock_net(skb->sk);
3522 struct ndmsg *ndm;
3523 struct nlattr *tb[NDA_MAX+1];
3524 struct net_device *dev;
3525 int err = -EINVAL;
3526 __u8 *addr;
3527 u16 vid;
3528
3529 if (!netlink_capable(skb, CAP_NET_ADMIN))
3530 return -EPERM;
3531
3532 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3533 if (err < 0)
3534 return err;
3535
3536 ndm = nlmsg_data(nlh);
3537 if (ndm->ndm_ifindex == 0) {
3538 NL_SET_ERR_MSG(extack, "invalid ifindex");
3539 return -EINVAL;
3540 }
3541
3542 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3543 if (dev == NULL) {
3544 NL_SET_ERR_MSG(extack, "unknown ifindex");
3545 return -ENODEV;
3546 }
3547
3548 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3549 NL_SET_ERR_MSG(extack, "invalid address");
3550 return -EINVAL;
3551 }
3552
3553 addr = nla_data(tb[NDA_LLADDR]);
3554
3555 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3556 if (err)
3557 return err;
3558
3559 err = -EOPNOTSUPP;
3560
3561 /* Support fdb on master device the net/bridge default case */
3562 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3563 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3564 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3565 const struct net_device_ops *ops = br_dev->netdev_ops;
3566
3567 if (ops->ndo_fdb_del)
3568 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3569
3570 if (err)
3571 goto out;
3572 else
3573 ndm->ndm_flags &= ~NTF_MASTER;
3574 }
3575
3576 /* Embedded bridge, macvlan, and any other device support */
3577 if (ndm->ndm_flags & NTF_SELF) {
3578 if (dev->netdev_ops->ndo_fdb_del)
3579 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3580 vid);
3581 else
3582 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3583
3584 if (!err) {
3585 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3586 ndm->ndm_state);
3587 ndm->ndm_flags &= ~NTF_SELF;
3588 }
3589 }
3590 out:
3591 return err;
3592 }
3593
3594 static int nlmsg_populate_fdb(struct sk_buff *skb,
3595 struct netlink_callback *cb,
3596 struct net_device *dev,
3597 int *idx,
3598 struct netdev_hw_addr_list *list)
3599 {
3600 struct netdev_hw_addr *ha;
3601 int err;
3602 u32 portid, seq;
3603
3604 portid = NETLINK_CB(cb->skb).portid;
3605 seq = cb->nlh->nlmsg_seq;
3606
3607 list_for_each_entry(ha, &list->list, list) {
3608 if (*idx < cb->args[2])
3609 goto skip;
3610
3611 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3612 portid, seq,
3613 RTM_NEWNEIGH, NTF_SELF,
3614 NLM_F_MULTI, NUD_PERMANENT);
3615 if (err < 0)
3616 return err;
3617 skip:
3618 *idx += 1;
3619 }
3620 return 0;
3621 }
3622
3623 /**
3624 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3625 * @nlh: netlink message header
3626 * @dev: netdevice
3627 *
3628 * Default netdevice operation to dump the existing unicast address list.
3629 * Returns number of addresses from list put in skb.
3630 */
3631 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3632 struct netlink_callback *cb,
3633 struct net_device *dev,
3634 struct net_device *filter_dev,
3635 int *idx)
3636 {
3637 int err;
3638
3639 netif_addr_lock_bh(dev);
3640 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3641 if (err)
3642 goto out;
3643 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3644 out:
3645 netif_addr_unlock_bh(dev);
3646 return err;
3647 }
3648 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3649
3650 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3651 {
3652 struct net_device *dev;
3653 struct nlattr *tb[IFLA_MAX+1];
3654 struct net_device *br_dev = NULL;
3655 const struct net_device_ops *ops = NULL;
3656 const struct net_device_ops *cops = NULL;
3657 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3658 struct net *net = sock_net(skb->sk);
3659 struct hlist_head *head;
3660 int brport_idx = 0;
3661 int br_idx = 0;
3662 int h, s_h;
3663 int idx = 0, s_idx;
3664 int err = 0;
3665 int fidx = 0;
3666
3667 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3668 IFLA_MAX, ifla_policy, NULL);
3669 if (err < 0) {
3670 return -EINVAL;
3671 } else if (err == 0) {
3672 if (tb[IFLA_MASTER])
3673 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3674 }
3675
3676 brport_idx = ifm->ifi_index;
3677
3678 if (br_idx) {
3679 br_dev = __dev_get_by_index(net, br_idx);
3680 if (!br_dev)
3681 return -ENODEV;
3682
3683 ops = br_dev->netdev_ops;
3684 }
3685
3686 s_h = cb->args[0];
3687 s_idx = cb->args[1];
3688
3689 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3690 idx = 0;
3691 head = &net->dev_index_head[h];
3692 hlist_for_each_entry(dev, head, index_hlist) {
3693
3694 if (brport_idx && (dev->ifindex != brport_idx))
3695 continue;
3696
3697 if (!br_idx) { /* user did not specify a specific bridge */
3698 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3699 br_dev = netdev_master_upper_dev_get(dev);
3700 cops = br_dev->netdev_ops;
3701 }
3702 } else {
3703 if (dev != br_dev &&
3704 !(dev->priv_flags & IFF_BRIDGE_PORT))
3705 continue;
3706
3707 if (br_dev != netdev_master_upper_dev_get(dev) &&
3708 !(dev->priv_flags & IFF_EBRIDGE))
3709 continue;
3710 cops = ops;
3711 }
3712
3713 if (idx < s_idx)
3714 goto cont;
3715
3716 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3717 if (cops && cops->ndo_fdb_dump) {
3718 err = cops->ndo_fdb_dump(skb, cb,
3719 br_dev, dev,
3720 &fidx);
3721 if (err == -EMSGSIZE)
3722 goto out;
3723 }
3724 }
3725
3726 if (dev->netdev_ops->ndo_fdb_dump)
3727 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3728 dev, NULL,
3729 &fidx);
3730 else
3731 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3732 &fidx);
3733 if (err == -EMSGSIZE)
3734 goto out;
3735
3736 cops = NULL;
3737
3738 /* reset fdb offset to 0 for rest of the interfaces */
3739 cb->args[2] = 0;
3740 fidx = 0;
3741 cont:
3742 idx++;
3743 }
3744 }
3745
3746 out:
3747 cb->args[0] = h;
3748 cb->args[1] = idx;
3749 cb->args[2] = fidx;
3750
3751 return skb->len;
3752 }
3753
3754 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3755 unsigned int attrnum, unsigned int flag)
3756 {
3757 if (mask & flag)
3758 return nla_put_u8(skb, attrnum, !!(flags & flag));
3759 return 0;
3760 }
3761
3762 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3763 struct net_device *dev, u16 mode,
3764 u32 flags, u32 mask, int nlflags,
3765 u32 filter_mask,
3766 int (*vlan_fill)(struct sk_buff *skb,
3767 struct net_device *dev,
3768 u32 filter_mask))
3769 {
3770 struct nlmsghdr *nlh;
3771 struct ifinfomsg *ifm;
3772 struct nlattr *br_afspec;
3773 struct nlattr *protinfo;
3774 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3775 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3776 int err = 0;
3777
3778 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3779 if (nlh == NULL)
3780 return -EMSGSIZE;
3781
3782 ifm = nlmsg_data(nlh);
3783 ifm->ifi_family = AF_BRIDGE;
3784 ifm->__ifi_pad = 0;
3785 ifm->ifi_type = dev->type;
3786 ifm->ifi_index = dev->ifindex;
3787 ifm->ifi_flags = dev_get_flags(dev);
3788 ifm->ifi_change = 0;
3789
3790
3791 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3792 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3793 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3794 (br_dev &&
3795 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3796 (dev->addr_len &&
3797 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3798 (dev->ifindex != dev_get_iflink(dev) &&
3799 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3800 goto nla_put_failure;
3801
3802 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3803 if (!br_afspec)
3804 goto nla_put_failure;
3805
3806 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3807 nla_nest_cancel(skb, br_afspec);
3808 goto nla_put_failure;
3809 }
3810
3811 if (mode != BRIDGE_MODE_UNDEF) {
3812 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3813 nla_nest_cancel(skb, br_afspec);
3814 goto nla_put_failure;
3815 }
3816 }
3817 if (vlan_fill) {
3818 err = vlan_fill(skb, dev, filter_mask);
3819 if (err) {
3820 nla_nest_cancel(skb, br_afspec);
3821 goto nla_put_failure;
3822 }
3823 }
3824 nla_nest_end(skb, br_afspec);
3825
3826 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3827 if (!protinfo)
3828 goto nla_put_failure;
3829
3830 if (brport_nla_put_flag(skb, flags, mask,
3831 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3832 brport_nla_put_flag(skb, flags, mask,
3833 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3834 brport_nla_put_flag(skb, flags, mask,
3835 IFLA_BRPORT_FAST_LEAVE,
3836 BR_MULTICAST_FAST_LEAVE) ||
3837 brport_nla_put_flag(skb, flags, mask,
3838 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3839 brport_nla_put_flag(skb, flags, mask,
3840 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3841 brport_nla_put_flag(skb, flags, mask,
3842 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3843 brport_nla_put_flag(skb, flags, mask,
3844 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3845 brport_nla_put_flag(skb, flags, mask,
3846 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3847 nla_nest_cancel(skb, protinfo);
3848 goto nla_put_failure;
3849 }
3850
3851 nla_nest_end(skb, protinfo);
3852
3853 nlmsg_end(skb, nlh);
3854 return 0;
3855 nla_put_failure:
3856 nlmsg_cancel(skb, nlh);
3857 return err ? err : -EMSGSIZE;
3858 }
3859 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3860
3861 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3862 {
3863 struct net *net = sock_net(skb->sk);
3864 struct net_device *dev;
3865 int idx = 0;
3866 u32 portid = NETLINK_CB(cb->skb).portid;
3867 u32 seq = cb->nlh->nlmsg_seq;
3868 u32 filter_mask = 0;
3869 int err;
3870
3871 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3872 struct nlattr *extfilt;
3873
3874 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3875 IFLA_EXT_MASK);
3876 if (extfilt) {
3877 if (nla_len(extfilt) < sizeof(filter_mask))
3878 return -EINVAL;
3879
3880 filter_mask = nla_get_u32(extfilt);
3881 }
3882 }
3883
3884 rcu_read_lock();
3885 for_each_netdev_rcu(net, dev) {
3886 const struct net_device_ops *ops = dev->netdev_ops;
3887 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3888
3889 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3890 if (idx >= cb->args[0]) {
3891 err = br_dev->netdev_ops->ndo_bridge_getlink(
3892 skb, portid, seq, dev,
3893 filter_mask, NLM_F_MULTI);
3894 if (err < 0 && err != -EOPNOTSUPP) {
3895 if (likely(skb->len))
3896 break;
3897
3898 goto out_err;
3899 }
3900 }
3901 idx++;
3902 }
3903
3904 if (ops->ndo_bridge_getlink) {
3905 if (idx >= cb->args[0]) {
3906 err = ops->ndo_bridge_getlink(skb, portid,
3907 seq, dev,
3908 filter_mask,
3909 NLM_F_MULTI);
3910 if (err < 0 && err != -EOPNOTSUPP) {
3911 if (likely(skb->len))
3912 break;
3913
3914 goto out_err;
3915 }
3916 }
3917 idx++;
3918 }
3919 }
3920 err = skb->len;
3921 out_err:
3922 rcu_read_unlock();
3923 cb->args[0] = idx;
3924
3925 return err;
3926 }
3927
3928 static inline size_t bridge_nlmsg_size(void)
3929 {
3930 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3931 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3932 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3933 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3934 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3935 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3936 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3937 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3938 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3939 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3940 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3941 }
3942
3943 static int rtnl_bridge_notify(struct net_device *dev)
3944 {
3945 struct net *net = dev_net(dev);
3946 struct sk_buff *skb;
3947 int err = -EOPNOTSUPP;
3948
3949 if (!dev->netdev_ops->ndo_bridge_getlink)
3950 return 0;
3951
3952 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3953 if (!skb) {
3954 err = -ENOMEM;
3955 goto errout;
3956 }
3957
3958 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3959 if (err < 0)
3960 goto errout;
3961
3962 if (!skb->len)
3963 goto errout;
3964
3965 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3966 return 0;
3967 errout:
3968 WARN_ON(err == -EMSGSIZE);
3969 kfree_skb(skb);
3970 if (err)
3971 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3972 return err;
3973 }
3974
3975 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3976 struct netlink_ext_ack *extack)
3977 {
3978 struct net *net = sock_net(skb->sk);
3979 struct ifinfomsg *ifm;
3980 struct net_device *dev;
3981 struct nlattr *br_spec, *attr = NULL;
3982 int rem, err = -EOPNOTSUPP;
3983 u16 flags = 0;
3984 bool have_flags = false;
3985
3986 if (nlmsg_len(nlh) < sizeof(*ifm))
3987 return -EINVAL;
3988
3989 ifm = nlmsg_data(nlh);
3990 if (ifm->ifi_family != AF_BRIDGE)
3991 return -EPFNOSUPPORT;
3992
3993 dev = __dev_get_by_index(net, ifm->ifi_index);
3994 if (!dev) {
3995 NL_SET_ERR_MSG(extack, "unknown ifindex");
3996 return -ENODEV;
3997 }
3998
3999 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4000 if (br_spec) {
4001 nla_for_each_nested(attr, br_spec, rem) {
4002 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4003 if (nla_len(attr) < sizeof(flags))
4004 return -EINVAL;
4005
4006 have_flags = true;
4007 flags = nla_get_u16(attr);
4008 break;
4009 }
4010 }
4011 }
4012
4013 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4014 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4015
4016 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4017 err = -EOPNOTSUPP;
4018 goto out;
4019 }
4020
4021 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
4022 if (err)
4023 goto out;
4024
4025 flags &= ~BRIDGE_FLAGS_MASTER;
4026 }
4027
4028 if ((flags & BRIDGE_FLAGS_SELF)) {
4029 if (!dev->netdev_ops->ndo_bridge_setlink)
4030 err = -EOPNOTSUPP;
4031 else
4032 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4033 flags);
4034 if (!err) {
4035 flags &= ~BRIDGE_FLAGS_SELF;
4036
4037 /* Generate event to notify upper layer of bridge
4038 * change
4039 */
4040 err = rtnl_bridge_notify(dev);
4041 }
4042 }
4043
4044 if (have_flags)
4045 memcpy(nla_data(attr), &flags, sizeof(flags));
4046 out:
4047 return err;
4048 }
4049
4050 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4051 struct netlink_ext_ack *extack)
4052 {
4053 struct net *net = sock_net(skb->sk);
4054 struct ifinfomsg *ifm;
4055 struct net_device *dev;
4056 struct nlattr *br_spec, *attr = NULL;
4057 int rem, err = -EOPNOTSUPP;
4058 u16 flags = 0;
4059 bool have_flags = false;
4060
4061 if (nlmsg_len(nlh) < sizeof(*ifm))
4062 return -EINVAL;
4063
4064 ifm = nlmsg_data(nlh);
4065 if (ifm->ifi_family != AF_BRIDGE)
4066 return -EPFNOSUPPORT;
4067
4068 dev = __dev_get_by_index(net, ifm->ifi_index);
4069 if (!dev) {
4070 NL_SET_ERR_MSG(extack, "unknown ifindex");
4071 return -ENODEV;
4072 }
4073
4074 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4075 if (br_spec) {
4076 nla_for_each_nested(attr, br_spec, rem) {
4077 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4078 if (nla_len(attr) < sizeof(flags))
4079 return -EINVAL;
4080
4081 have_flags = true;
4082 flags = nla_get_u16(attr);
4083 break;
4084 }
4085 }
4086 }
4087
4088 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4089 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4090
4091 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4092 err = -EOPNOTSUPP;
4093 goto out;
4094 }
4095
4096 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4097 if (err)
4098 goto out;
4099
4100 flags &= ~BRIDGE_FLAGS_MASTER;
4101 }
4102
4103 if ((flags & BRIDGE_FLAGS_SELF)) {
4104 if (!dev->netdev_ops->ndo_bridge_dellink)
4105 err = -EOPNOTSUPP;
4106 else
4107 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4108 flags);
4109
4110 if (!err) {
4111 flags &= ~BRIDGE_FLAGS_SELF;
4112
4113 /* Generate event to notify upper layer of bridge
4114 * change
4115 */
4116 err = rtnl_bridge_notify(dev);
4117 }
4118 }
4119
4120 if (have_flags)
4121 memcpy(nla_data(attr), &flags, sizeof(flags));
4122 out:
4123 return err;
4124 }
4125
4126 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4127 {
4128 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4129 (!idxattr || idxattr == attrid);
4130 }
4131
4132 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4133 static int rtnl_get_offload_stats_attr_size(int attr_id)
4134 {
4135 switch (attr_id) {
4136 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4137 return sizeof(struct rtnl_link_stats64);
4138 }
4139
4140 return 0;
4141 }
4142
4143 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4144 int *prividx)
4145 {
4146 struct nlattr *attr = NULL;
4147 int attr_id, size;
4148 void *attr_data;
4149 int err;
4150
4151 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4152 dev->netdev_ops->ndo_get_offload_stats))
4153 return -ENODATA;
4154
4155 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4156 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4157 if (attr_id < *prividx)
4158 continue;
4159
4160 size = rtnl_get_offload_stats_attr_size(attr_id);
4161 if (!size)
4162 continue;
4163
4164 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4165 continue;
4166
4167 attr = nla_reserve_64bit(skb, attr_id, size,
4168 IFLA_OFFLOAD_XSTATS_UNSPEC);
4169 if (!attr)
4170 goto nla_put_failure;
4171
4172 attr_data = nla_data(attr);
4173 memset(attr_data, 0, size);
4174 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4175 attr_data);
4176 if (err)
4177 goto get_offload_stats_failure;
4178 }
4179
4180 if (!attr)
4181 return -ENODATA;
4182
4183 *prividx = 0;
4184 return 0;
4185
4186 nla_put_failure:
4187 err = -EMSGSIZE;
4188 get_offload_stats_failure:
4189 *prividx = attr_id;
4190 return err;
4191 }
4192
4193 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4194 {
4195 int nla_size = 0;
4196 int attr_id;
4197 int size;
4198
4199 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4200 dev->netdev_ops->ndo_get_offload_stats))
4201 return 0;
4202
4203 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4204 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4205 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4206 continue;
4207 size = rtnl_get_offload_stats_attr_size(attr_id);
4208 nla_size += nla_total_size_64bit(size);
4209 }
4210
4211 if (nla_size != 0)
4212 nla_size += nla_total_size(0);
4213
4214 return nla_size;
4215 }
4216
4217 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4218 int type, u32 pid, u32 seq, u32 change,
4219 unsigned int flags, unsigned int filter_mask,
4220 int *idxattr, int *prividx)
4221 {
4222 struct if_stats_msg *ifsm;
4223 struct nlmsghdr *nlh;
4224 struct nlattr *attr;
4225 int s_prividx = *prividx;
4226 int err;
4227
4228 ASSERT_RTNL();
4229
4230 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4231 if (!nlh)
4232 return -EMSGSIZE;
4233
4234 ifsm = nlmsg_data(nlh);
4235 ifsm->family = PF_UNSPEC;
4236 ifsm->pad1 = 0;
4237 ifsm->pad2 = 0;
4238 ifsm->ifindex = dev->ifindex;
4239 ifsm->filter_mask = filter_mask;
4240
4241 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4242 struct rtnl_link_stats64 *sp;
4243
4244 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4245 sizeof(struct rtnl_link_stats64),
4246 IFLA_STATS_UNSPEC);
4247 if (!attr)
4248 goto nla_put_failure;
4249
4250 sp = nla_data(attr);
4251 dev_get_stats(dev, sp);
4252 }
4253
4254 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4255 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4256
4257 if (ops && ops->fill_linkxstats) {
4258 *idxattr = IFLA_STATS_LINK_XSTATS;
4259 attr = nla_nest_start(skb,
4260 IFLA_STATS_LINK_XSTATS);
4261 if (!attr)
4262 goto nla_put_failure;
4263
4264 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4265 nla_nest_end(skb, attr);
4266 if (err)
4267 goto nla_put_failure;
4268 *idxattr = 0;
4269 }
4270 }
4271
4272 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4273 *idxattr)) {
4274 const struct rtnl_link_ops *ops = NULL;
4275 const struct net_device *master;
4276
4277 master = netdev_master_upper_dev_get(dev);
4278 if (master)
4279 ops = master->rtnl_link_ops;
4280 if (ops && ops->fill_linkxstats) {
4281 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4282 attr = nla_nest_start(skb,
4283 IFLA_STATS_LINK_XSTATS_SLAVE);
4284 if (!attr)
4285 goto nla_put_failure;
4286
4287 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4288 nla_nest_end(skb, attr);
4289 if (err)
4290 goto nla_put_failure;
4291 *idxattr = 0;
4292 }
4293 }
4294
4295 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4296 *idxattr)) {
4297 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4298 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4299 if (!attr)
4300 goto nla_put_failure;
4301
4302 err = rtnl_get_offload_stats(skb, dev, prividx);
4303 if (err == -ENODATA)
4304 nla_nest_cancel(skb, attr);
4305 else
4306 nla_nest_end(skb, attr);
4307
4308 if (err && err != -ENODATA)
4309 goto nla_put_failure;
4310 *idxattr = 0;
4311 }
4312
4313 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4314 struct rtnl_af_ops *af_ops;
4315
4316 *idxattr = IFLA_STATS_AF_SPEC;
4317 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4318 if (!attr)
4319 goto nla_put_failure;
4320
4321 rcu_read_lock();
4322 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4323 if (af_ops->fill_stats_af) {
4324 struct nlattr *af;
4325 int err;
4326
4327 af = nla_nest_start(skb, af_ops->family);
4328 if (!af) {
4329 rcu_read_unlock();
4330 goto nla_put_failure;
4331 }
4332 err = af_ops->fill_stats_af(skb, dev);
4333
4334 if (err == -ENODATA) {
4335 nla_nest_cancel(skb, af);
4336 } else if (err < 0) {
4337 rcu_read_unlock();
4338 goto nla_put_failure;
4339 }
4340
4341 nla_nest_end(skb, af);
4342 }
4343 }
4344 rcu_read_unlock();
4345
4346 nla_nest_end(skb, attr);
4347
4348 *idxattr = 0;
4349 }
4350
4351 nlmsg_end(skb, nlh);
4352
4353 return 0;
4354
4355 nla_put_failure:
4356 /* not a multi message or no progress mean a real error */
4357 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4358 nlmsg_cancel(skb, nlh);
4359 else
4360 nlmsg_end(skb, nlh);
4361
4362 return -EMSGSIZE;
4363 }
4364
4365 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4366 u32 filter_mask)
4367 {
4368 size_t size = 0;
4369
4370 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4371 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4372
4373 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4374 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4375 int attr = IFLA_STATS_LINK_XSTATS;
4376
4377 if (ops && ops->get_linkxstats_size) {
4378 size += nla_total_size(ops->get_linkxstats_size(dev,
4379 attr));
4380 /* for IFLA_STATS_LINK_XSTATS */
4381 size += nla_total_size(0);
4382 }
4383 }
4384
4385 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4386 struct net_device *_dev = (struct net_device *)dev;
4387 const struct rtnl_link_ops *ops = NULL;
4388 const struct net_device *master;
4389
4390 /* netdev_master_upper_dev_get can't take const */
4391 master = netdev_master_upper_dev_get(_dev);
4392 if (master)
4393 ops = master->rtnl_link_ops;
4394 if (ops && ops->get_linkxstats_size) {
4395 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4396
4397 size += nla_total_size(ops->get_linkxstats_size(dev,
4398 attr));
4399 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4400 size += nla_total_size(0);
4401 }
4402 }
4403
4404 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4405 size += rtnl_get_offload_stats_size(dev);
4406
4407 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4408 struct rtnl_af_ops *af_ops;
4409
4410 /* for IFLA_STATS_AF_SPEC */
4411 size += nla_total_size(0);
4412
4413 rcu_read_lock();
4414 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4415 if (af_ops->get_stats_af_size) {
4416 size += nla_total_size(
4417 af_ops->get_stats_af_size(dev));
4418
4419 /* for AF_* */
4420 size += nla_total_size(0);
4421 }
4422 }
4423 rcu_read_unlock();
4424 }
4425
4426 return size;
4427 }
4428
4429 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4430 struct netlink_ext_ack *extack)
4431 {
4432 struct net *net = sock_net(skb->sk);
4433 struct net_device *dev = NULL;
4434 int idxattr = 0, prividx = 0;
4435 struct if_stats_msg *ifsm;
4436 struct sk_buff *nskb;
4437 u32 filter_mask;
4438 int err;
4439
4440 if (nlmsg_len(nlh) < sizeof(*ifsm))
4441 return -EINVAL;
4442
4443 ifsm = nlmsg_data(nlh);
4444 if (ifsm->ifindex > 0)
4445 dev = __dev_get_by_index(net, ifsm->ifindex);
4446 else
4447 return -EINVAL;
4448
4449 if (!dev)
4450 return -ENODEV;
4451
4452 filter_mask = ifsm->filter_mask;
4453 if (!filter_mask)
4454 return -EINVAL;
4455
4456 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4457 if (!nskb)
4458 return -ENOBUFS;
4459
4460 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4461 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4462 0, filter_mask, &idxattr, &prividx);
4463 if (err < 0) {
4464 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4465 WARN_ON(err == -EMSGSIZE);
4466 kfree_skb(nskb);
4467 } else {
4468 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4469 }
4470
4471 return err;
4472 }
4473
4474 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4475 {
4476 int h, s_h, err, s_idx, s_idxattr, s_prividx;
4477 struct net *net = sock_net(skb->sk);
4478 unsigned int flags = NLM_F_MULTI;
4479 struct if_stats_msg *ifsm;
4480 struct hlist_head *head;
4481 struct net_device *dev;
4482 u32 filter_mask = 0;
4483 int idx = 0;
4484
4485 s_h = cb->args[0];
4486 s_idx = cb->args[1];
4487 s_idxattr = cb->args[2];
4488 s_prividx = cb->args[3];
4489
4490 cb->seq = net->dev_base_seq;
4491
4492 if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4493 return -EINVAL;
4494
4495 ifsm = nlmsg_data(cb->nlh);
4496 filter_mask = ifsm->filter_mask;
4497 if (!filter_mask)
4498 return -EINVAL;
4499
4500 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4501 idx = 0;
4502 head = &net->dev_index_head[h];
4503 hlist_for_each_entry(dev, head, index_hlist) {
4504 if (idx < s_idx)
4505 goto cont;
4506 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4507 NETLINK_CB(cb->skb).portid,
4508 cb->nlh->nlmsg_seq, 0,
4509 flags, filter_mask,
4510 &s_idxattr, &s_prividx);
4511 /* If we ran out of room on the first message,
4512 * we're in trouble
4513 */
4514 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4515
4516 if (err < 0)
4517 goto out;
4518 s_prividx = 0;
4519 s_idxattr = 0;
4520 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4521 cont:
4522 idx++;
4523 }
4524 }
4525 out:
4526 cb->args[3] = s_prividx;
4527 cb->args[2] = s_idxattr;
4528 cb->args[1] = idx;
4529 cb->args[0] = h;
4530
4531 return skb->len;
4532 }
4533
4534 /* Process one rtnetlink message. */
4535
4536 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4537 struct netlink_ext_ack *extack)
4538 {
4539 struct net *net = sock_net(skb->sk);
4540 struct rtnl_link *link;
4541 struct module *owner;
4542 int err = -EOPNOTSUPP;
4543 rtnl_doit_func doit;
4544 unsigned int flags;
4545 int kind;
4546 int family;
4547 int type;
4548
4549 type = nlh->nlmsg_type;
4550 if (type > RTM_MAX)
4551 return -EOPNOTSUPP;
4552
4553 type -= RTM_BASE;
4554
4555 /* All the messages must have at least 1 byte length */
4556 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4557 return 0;
4558
4559 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4560 kind = type&3;
4561
4562 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4563 return -EPERM;
4564
4565 rcu_read_lock();
4566 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4567 struct sock *rtnl;
4568 rtnl_dumpit_func dumpit;
4569 u16 min_dump_alloc = 0;
4570
4571 link = rtnl_get_link(family, type);
4572 if (!link || !link->dumpit) {
4573 family = PF_UNSPEC;
4574 link = rtnl_get_link(family, type);
4575 if (!link || !link->dumpit)
4576 goto err_unlock;
4577 }
4578 owner = link->owner;
4579 dumpit = link->dumpit;
4580
4581 if (type == RTM_GETLINK - RTM_BASE)
4582 min_dump_alloc = rtnl_calcit(skb, nlh);
4583
4584 err = 0;
4585 /* need to do this before rcu_read_unlock() */
4586 if (!try_module_get(owner))
4587 err = -EPROTONOSUPPORT;
4588
4589 rcu_read_unlock();
4590
4591 rtnl = net->rtnl;
4592 if (err == 0) {
4593 struct netlink_dump_control c = {
4594 .dump = dumpit,
4595 .min_dump_alloc = min_dump_alloc,
4596 .module = owner,
4597 };
4598 err = netlink_dump_start(rtnl, skb, nlh, &c);
4599 /* netlink_dump_start() will keep a reference on
4600 * module if dump is still in progress.
4601 */
4602 module_put(owner);
4603 }
4604 return err;
4605 }
4606
4607 link = rtnl_get_link(family, type);
4608 if (!link || !link->doit) {
4609 family = PF_UNSPEC;
4610 link = rtnl_get_link(PF_UNSPEC, type);
4611 if (!link || !link->doit)
4612 goto out_unlock;
4613 }
4614
4615 owner = link->owner;
4616 if (!try_module_get(owner)) {
4617 err = -EPROTONOSUPPORT;
4618 goto out_unlock;
4619 }
4620
4621 flags = link->flags;
4622 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4623 doit = link->doit;
4624 rcu_read_unlock();
4625 if (doit)
4626 err = doit(skb, nlh, extack);
4627 module_put(owner);
4628 return err;
4629 }
4630 rcu_read_unlock();
4631
4632 rtnl_lock();
4633 link = rtnl_get_link(family, type);
4634 if (link && link->doit)
4635 err = link->doit(skb, nlh, extack);
4636 rtnl_unlock();
4637
4638 module_put(owner);
4639
4640 return err;
4641
4642 out_unlock:
4643 rcu_read_unlock();
4644 return err;
4645
4646 err_unlock:
4647 rcu_read_unlock();
4648 return -EOPNOTSUPP;
4649 }
4650
4651 static void rtnetlink_rcv(struct sk_buff *skb)
4652 {
4653 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4654 }
4655
4656 static int rtnetlink_bind(struct net *net, int group)
4657 {
4658 switch (group) {
4659 case RTNLGRP_IPV4_MROUTE_R:
4660 case RTNLGRP_IPV6_MROUTE_R:
4661 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4662 return -EPERM;
4663 break;
4664 }
4665 return 0;
4666 }
4667
4668 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4669 {
4670 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4671
4672 switch (event) {
4673 case NETDEV_REBOOT:
4674 case NETDEV_CHANGEMTU:
4675 case NETDEV_CHANGEADDR:
4676 case NETDEV_CHANGENAME:
4677 case NETDEV_FEAT_CHANGE:
4678 case NETDEV_BONDING_FAILOVER:
4679 case NETDEV_POST_TYPE_CHANGE:
4680 case NETDEV_NOTIFY_PEERS:
4681 case NETDEV_CHANGEUPPER:
4682 case NETDEV_RESEND_IGMP:
4683 case NETDEV_CHANGEINFODATA:
4684 case NETDEV_CHANGELOWERSTATE:
4685 case NETDEV_CHANGE_TX_QUEUE_LEN:
4686 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4687 GFP_KERNEL, NULL, 0);
4688 break;
4689 default:
4690 break;
4691 }
4692 return NOTIFY_DONE;
4693 }
4694
4695 static struct notifier_block rtnetlink_dev_notifier = {
4696 .notifier_call = rtnetlink_event,
4697 };
4698
4699
4700 static int __net_init rtnetlink_net_init(struct net *net)
4701 {
4702 struct sock *sk;
4703 struct netlink_kernel_cfg cfg = {
4704 .groups = RTNLGRP_MAX,
4705 .input = rtnetlink_rcv,
4706 .cb_mutex = &rtnl_mutex,
4707 .flags = NL_CFG_F_NONROOT_RECV,
4708 .bind = rtnetlink_bind,
4709 };
4710
4711 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4712 if (!sk)
4713 return -ENOMEM;
4714 net->rtnl = sk;
4715 return 0;
4716 }
4717
4718 static void __net_exit rtnetlink_net_exit(struct net *net)
4719 {
4720 netlink_kernel_release(net->rtnl);
4721 net->rtnl = NULL;
4722 }
4723
4724 static struct pernet_operations rtnetlink_net_ops = {
4725 .init = rtnetlink_net_init,
4726 .exit = rtnetlink_net_exit,
4727 };
4728
4729 void __init rtnetlink_init(void)
4730 {
4731 if (register_pernet_subsys(&rtnetlink_net_ops))
4732 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4733
4734 register_netdevice_notifier(&rtnetlink_dev_notifier);
4735
4736 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4737 rtnl_dump_ifinfo, 0);
4738 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4739 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4740 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4741
4742 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4743 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4744 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4745
4746 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4747 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4748 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4749
4750 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4751 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4752 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4753
4754 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4755 0);
4756 }
CRIS linux kernel tree