?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 6.14-rc1
[linux.git] / net / core / rtnetlink.c
blob1f4d4b5570ab896593a6b3216678b90c7b0696de
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Routing netlink socket interface: protocol independent part.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 *
11 * Fixes:
12 * Vitaly E. Lavrov RTA_OK arithmetic was wrong.
13 */
14
15 #include <linux/bitops.h>
16 #include <linux/errno.h>
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/kernel.h>
21 #include <linux/timer.h>
22 #include <linux/string.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/fcntl.h>
26 #include <linux/mm.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/capability.h>
30 #include <linux/skbuff.h>
31 #include <linux/init.h>
32 #include <linux/security.h>
33 #include <linux/mutex.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_vlan.h>
37 #include <linux/pci.h>
38 #include <linux/etherdevice.h>
39 #include <linux/bpf.h>
40
41 #include <linux/uaccess.h>
42
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <net/ip.h>
46 #include <net/protocol.h>
47 #include <net/arp.h>
48 #include <net/route.h>
49 #include <net/udp.h>
50 #include <net/tcp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 #include <net/devlink.h>
57 #if IS_ENABLED(CONFIG_IPV6)
58 #include <net/addrconf.h>
59 #endif
60 #include <linux/dpll.h>
61
62 #include "dev.h"
63
64 #define RTNL_MAX_TYPE 50
65 #define RTNL_SLAVE_MAX_TYPE 44
66
67 struct rtnl_link {
68 rtnl_doit_func doit;
69 rtnl_dumpit_func dumpit;
70 struct module *owner;
71 unsigned int flags;
72 struct rcu_head rcu;
73 };
74
75 static DEFINE_MUTEX(rtnl_mutex);
76
77 void rtnl_lock(void)
78 {
79 mutex_lock(&rtnl_mutex);
80 }
81 EXPORT_SYMBOL(rtnl_lock);
82
83 int rtnl_lock_killable(void)
84 {
85 return mutex_lock_killable(&rtnl_mutex);
86 }
87
88 static struct sk_buff *defer_kfree_skb_list;
89 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
90 {
91 if (head && tail) {
92 tail->next = defer_kfree_skb_list;
93 defer_kfree_skb_list = head;
94 }
95 }
96 EXPORT_SYMBOL(rtnl_kfree_skbs);
97
98 void __rtnl_unlock(void)
99 {
100 struct sk_buff *head = defer_kfree_skb_list;
101
102 defer_kfree_skb_list = NULL;
103
104 /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
105 * is used. In some places, e.g. in cfg80211, we have code that will do
106 * something like
107 * rtnl_lock()
108 * wiphy_lock()
109 * ...
110 * rtnl_unlock()
111 *
112 * and because netdev_run_todo() acquires the RTNL for items on the list
113 * we could cause a situation such as this:
114 * Thread 1 Thread 2
115 * rtnl_lock()
116 * unregister_netdevice()
117 * __rtnl_unlock()
118 * rtnl_lock()
119 * wiphy_lock()
120 * rtnl_unlock()
121 * netdev_run_todo()
122 * __rtnl_unlock()
123 *
124 * // list not empty now
125 * // because of thread 2
126 * rtnl_lock()
127 * while (!list_empty(...))
128 * rtnl_lock()
129 * wiphy_lock()
130 * **** DEADLOCK ****
131 *
132 * However, usage of __rtnl_unlock() is rare, and so we can ensure that
133 * it's not used in cases where something is added to do the list.
134 */
135 WARN_ON(!list_empty(&net_todo_list));
136
137 mutex_unlock(&rtnl_mutex);
138
139 while (head) {
140 struct sk_buff *next = head->next;
141
142 kfree_skb(head);
143 cond_resched();
144 head = next;
145 }
146 }
147
148 void rtnl_unlock(void)
149 {
150 /* This fellow will unlock it for us. */
151 netdev_run_todo();
152 }
153 EXPORT_SYMBOL(rtnl_unlock);
154
155 int rtnl_trylock(void)
156 {
157 return mutex_trylock(&rtnl_mutex);
158 }
159 EXPORT_SYMBOL(rtnl_trylock);
160
161 int rtnl_is_locked(void)
162 {
163 return mutex_is_locked(&rtnl_mutex);
164 }
165 EXPORT_SYMBOL(rtnl_is_locked);
166
167 bool refcount_dec_and_rtnl_lock(refcount_t *r)
168 {
169 return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
170 }
171 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
172
173 #ifdef CONFIG_PROVE_LOCKING
174 bool lockdep_rtnl_is_held(void)
175 {
176 return lockdep_is_held(&rtnl_mutex);
177 }
178 EXPORT_SYMBOL(lockdep_rtnl_is_held);
179 #endif /* #ifdef CONFIG_PROVE_LOCKING */
180
181 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL
182 void __rtnl_net_lock(struct net *net)
183 {
184 ASSERT_RTNL();
185
186 mutex_lock(&net->rtnl_mutex);
187 }
188 EXPORT_SYMBOL(__rtnl_net_lock);
189
190 void __rtnl_net_unlock(struct net *net)
191 {
192 ASSERT_RTNL();
193
194 mutex_unlock(&net->rtnl_mutex);
195 }
196 EXPORT_SYMBOL(__rtnl_net_unlock);
197
198 void rtnl_net_lock(struct net *net)
199 {
200 rtnl_lock();
201 __rtnl_net_lock(net);
202 }
203 EXPORT_SYMBOL(rtnl_net_lock);
204
205 void rtnl_net_unlock(struct net *net)
206 {
207 __rtnl_net_unlock(net);
208 rtnl_unlock();
209 }
210 EXPORT_SYMBOL(rtnl_net_unlock);
211
212 int rtnl_net_trylock(struct net *net)
213 {
214 int ret = rtnl_trylock();
215
216 if (ret)
217 __rtnl_net_lock(net);
218
219 return ret;
220 }
221 EXPORT_SYMBOL(rtnl_net_trylock);
222
223 int rtnl_net_lock_killable(struct net *net)
224 {
225 int ret = rtnl_lock_killable();
226
227 if (!ret)
228 __rtnl_net_lock(net);
229
230 return ret;
231 }
232
233 static int rtnl_net_cmp_locks(const struct net *net_a, const struct net *net_b)
234 {
235 if (net_eq(net_a, net_b))
236 return 0;
237
238 /* always init_net first */
239 if (net_eq(net_a, &init_net))
240 return -1;
241
242 if (net_eq(net_b, &init_net))
243 return 1;
244
245 /* otherwise lock in ascending order */
246 return net_a < net_b ? -1 : 1;
247 }
248
249 int rtnl_net_lock_cmp_fn(const struct lockdep_map *a, const struct lockdep_map *b)
250 {
251 const struct net *net_a, *net_b;
252
253 net_a = container_of(a, struct net, rtnl_mutex.dep_map);
254 net_b = container_of(b, struct net, rtnl_mutex.dep_map);
255
256 return rtnl_net_cmp_locks(net_a, net_b);
257 }
258
259 bool rtnl_net_is_locked(struct net *net)
260 {
261 return rtnl_is_locked() && mutex_is_locked(&net->rtnl_mutex);
262 }
263 EXPORT_SYMBOL(rtnl_net_is_locked);
264
265 bool lockdep_rtnl_net_is_held(struct net *net)
266 {
267 return lockdep_rtnl_is_held() && lockdep_is_held(&net->rtnl_mutex);
268 }
269 EXPORT_SYMBOL(lockdep_rtnl_net_is_held);
270 #else
271 static int rtnl_net_cmp_locks(const struct net *net_a, const struct net *net_b)
272 {
273 /* No need to swap */
274 return -1;
275 }
276 #endif
277
278 struct rtnl_nets {
279 /* ->newlink() needs to freeze 3 netns at most;
280 * 2 for the new device, 1 for its peer.
281 */
282 struct net *net[3];
283 unsigned char len;
284 };
285
286 static void rtnl_nets_init(struct rtnl_nets *rtnl_nets)
287 {
288 memset(rtnl_nets, 0, sizeof(*rtnl_nets));
289 }
290
291 static void rtnl_nets_destroy(struct rtnl_nets *rtnl_nets)
292 {
293 int i;
294
295 for (i = 0; i < rtnl_nets->len; i++) {
296 put_net(rtnl_nets->net[i]);
297 rtnl_nets->net[i] = NULL;
298 }
299
300 rtnl_nets->len = 0;
301 }
302
303 /**
304 * rtnl_nets_add - Add netns to be locked before ->newlink().
305 *
306 * @rtnl_nets: rtnl_nets pointer passed to ->get_peer_net().
307 * @net: netns pointer with an extra refcnt held.
308 *
309 * The extra refcnt is released in rtnl_nets_destroy().
310 */
311 static void rtnl_nets_add(struct rtnl_nets *rtnl_nets, struct net *net)
312 {
313 int i;
314
315 DEBUG_NET_WARN_ON_ONCE(rtnl_nets->len == ARRAY_SIZE(rtnl_nets->net));
316
317 for (i = 0; i < rtnl_nets->len; i++) {
318 switch (rtnl_net_cmp_locks(rtnl_nets->net[i], net)) {
319 case 0:
320 put_net(net);
321 return;
322 case 1:
323 swap(rtnl_nets->net[i], net);
324 }
325 }
326
327 rtnl_nets->net[i] = net;
328 rtnl_nets->len++;
329 }
330
331 static void rtnl_nets_lock(struct rtnl_nets *rtnl_nets)
332 {
333 int i;
334
335 rtnl_lock();
336
337 for (i = 0; i < rtnl_nets->len; i++)
338 __rtnl_net_lock(rtnl_nets->net[i]);
339 }
340
341 static void rtnl_nets_unlock(struct rtnl_nets *rtnl_nets)
342 {
343 int i;
344
345 for (i = 0; i < rtnl_nets->len; i++)
346 __rtnl_net_unlock(rtnl_nets->net[i]);
347
348 rtnl_unlock();
349 }
350
351 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
352
353 static inline int rtm_msgindex(int msgtype)
354 {
355 int msgindex = msgtype - RTM_BASE;
356
357 /*
358 * msgindex < 0 implies someone tried to register a netlink
359 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
360 * the message type has not been added to linux/rtnetlink.h
361 */
362 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
363
364 return msgindex;
365 }
366
367 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
368 {
369 struct rtnl_link __rcu **tab;
370
371 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
372 protocol = PF_UNSPEC;
373
374 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
375 if (!tab)
376 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
377
378 return rcu_dereference_rtnl(tab[msgtype]);
379 }
380
381 static int rtnl_register_internal(struct module *owner,
382 int protocol, int msgtype,
383 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
384 unsigned int flags)
385 {
386 struct rtnl_link *link, *old;
387 struct rtnl_link __rcu **tab;
388 int msgindex;
389 int ret = -ENOBUFS;
390
391 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
392 msgindex = rtm_msgindex(msgtype);
393
394 rtnl_lock();
395 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
396 if (tab == NULL) {
397 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
398 if (!tab)
399 goto unlock;
400
401 /* ensures we see the 0 stores */
402 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
403 }
404
405 old = rtnl_dereference(tab[msgindex]);
406 if (old) {
407 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
408 if (!link)
409 goto unlock;
410 } else {
411 link = kzalloc(sizeof(*link), GFP_KERNEL);
412 if (!link)
413 goto unlock;
414 }
415
416 WARN_ON(link->owner && link->owner != owner);
417 link->owner = owner;
418
419 WARN_ON(doit && link->doit && link->doit != doit);
420 if (doit)
421 link->doit = doit;
422 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
423 if (dumpit)
424 link->dumpit = dumpit;
425
426 WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
427 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
428 link->flags |= flags;
429
430 /* publish protocol:msgtype */
431 rcu_assign_pointer(tab[msgindex], link);
432 ret = 0;
433 if (old)
434 kfree_rcu(old, rcu);
435 unlock:
436 rtnl_unlock();
437 return ret;
438 }
439
440 /**
441 * rtnl_unregister - Unregister a rtnetlink message type
442 * @protocol: Protocol family or PF_UNSPEC
443 * @msgtype: rtnetlink message type
444 *
445 * Returns 0 on success or a negative error code.
446 */
447 static int rtnl_unregister(int protocol, int msgtype)
448 {
449 struct rtnl_link __rcu **tab;
450 struct rtnl_link *link;
451 int msgindex;
452
453 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
454 msgindex = rtm_msgindex(msgtype);
455
456 rtnl_lock();
457 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
458 if (!tab) {
459 rtnl_unlock();
460 return -ENOENT;
461 }
462
463 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
464 rtnl_unlock();
465
466 kfree_rcu(link, rcu);
467
468 return 0;
469 }
470
471 /**
472 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
473 * @protocol : Protocol family or PF_UNSPEC
474 *
475 * Identical to calling rtnl_unregster() for all registered message types
476 * of a certain protocol family.
477 */
478 void rtnl_unregister_all(int protocol)
479 {
480 struct rtnl_link __rcu **tab;
481 struct rtnl_link *link;
482 int msgindex;
483
484 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
485
486 rtnl_lock();
487 tab = rcu_replace_pointer_rtnl(rtnl_msg_handlers[protocol], NULL);
488 if (!tab) {
489 rtnl_unlock();
490 return;
491 }
492 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
493 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
494 kfree_rcu(link, rcu);
495 }
496 rtnl_unlock();
497
498 synchronize_net();
499
500 kfree(tab);
501 }
502 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
503
504 /**
505 * __rtnl_register_many - Register rtnetlink message types
506 * @handlers: Array of struct rtnl_msg_handlers
507 * @n: The length of @handlers
508 *
509 * Registers the specified function pointers (at least one of them has
510 * to be non-NULL) to be called whenever a request message for the
511 * specified protocol family and message type is received.
512 *
513 * The special protocol family PF_UNSPEC may be used to define fallback
514 * function pointers for the case when no entry for the specific protocol
515 * family exists.
516 *
517 * When one element of @handlers fails to register,
518 * 1) built-in: panics.
519 * 2) modules : the previous successful registrations are unwinded
520 * and an error is returned.
521 *
522 * Use rtnl_register_many().
523 */
524 int __rtnl_register_many(const struct rtnl_msg_handler *handlers, int n)
525 {
526 const struct rtnl_msg_handler *handler;
527 int i, err;
528
529 for (i = 0, handler = handlers; i < n; i++, handler++) {
530 err = rtnl_register_internal(handler->owner, handler->protocol,
531 handler->msgtype, handler->doit,
532 handler->dumpit, handler->flags);
533 if (err) {
534 if (!handler->owner)
535 panic("Unable to register rtnetlink message "
536 "handlers, %pS\n", handlers);
537
538 __rtnl_unregister_many(handlers, i);
539 break;
540 }
541 }
542
543 return err;
544 }
545 EXPORT_SYMBOL_GPL(__rtnl_register_many);
546
547 void __rtnl_unregister_many(const struct rtnl_msg_handler *handlers, int n)
548 {
549 const struct rtnl_msg_handler *handler;
550 int i;
551
552 for (i = n - 1, handler = handlers + n - 1; i >= 0; i--, handler--)
553 rtnl_unregister(handler->protocol, handler->msgtype);
554 }
555 EXPORT_SYMBOL_GPL(__rtnl_unregister_many);
556
557 static DEFINE_MUTEX(link_ops_mutex);
558 static LIST_HEAD(link_ops);
559
560 static struct rtnl_link_ops *rtnl_link_ops_get(const char *kind, int *srcu_index)
561 {
562 struct rtnl_link_ops *ops;
563
564 rcu_read_lock();
565
566 list_for_each_entry_rcu(ops, &link_ops, list) {
567 if (!strcmp(ops->kind, kind)) {
568 *srcu_index = srcu_read_lock(&ops->srcu);
569 goto unlock;
570 }
571 }
572
573 ops = NULL;
574 unlock:
575 rcu_read_unlock();
576
577 return ops;
578 }
579
580 static void rtnl_link_ops_put(struct rtnl_link_ops *ops, int srcu_index)
581 {
582 srcu_read_unlock(&ops->srcu, srcu_index);
583 }
584
585 /**
586 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
587 * @ops: struct rtnl_link_ops * to register
588 *
589 * Returns 0 on success or a negative error code.
590 */
591 int rtnl_link_register(struct rtnl_link_ops *ops)
592 {
593 struct rtnl_link_ops *tmp;
594 int err;
595
596 /* Sanity-check max sizes to avoid stack buffer overflow. */
597 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
598 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
599 return -EINVAL;
600
601 /* The check for alloc/setup is here because if ops
602 * does not have that filled up, it is not possible
603 * to use the ops for creating device. So do not
604 * fill up dellink as well. That disables rtnl_dellink.
605 */
606 if ((ops->alloc || ops->setup) && !ops->dellink)
607 ops->dellink = unregister_netdevice_queue;
608
609 err = init_srcu_struct(&ops->srcu);
610 if (err)
611 return err;
612
613 mutex_lock(&link_ops_mutex);
614
615 list_for_each_entry(tmp, &link_ops, list) {
616 if (!strcmp(ops->kind, tmp->kind)) {
617 err = -EEXIST;
618 goto unlock;
619 }
620 }
621
622 list_add_tail_rcu(&ops->list, &link_ops);
623 unlock:
624 mutex_unlock(&link_ops_mutex);
625
626 return err;
627 }
628 EXPORT_SYMBOL_GPL(rtnl_link_register);
629
630 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
631 {
632 struct net_device *dev;
633 LIST_HEAD(list_kill);
634
635 for_each_netdev(net, dev) {
636 if (dev->rtnl_link_ops == ops)
637 ops->dellink(dev, &list_kill);
638 }
639 unregister_netdevice_many(&list_kill);
640 }
641
642 /* Return with the rtnl_lock held when there are no network
643 * devices unregistering in any network namespace.
644 */
645 static void rtnl_lock_unregistering_all(void)
646 {
647 DEFINE_WAIT_FUNC(wait, woken_wake_function);
648
649 add_wait_queue(&netdev_unregistering_wq, &wait);
650 for (;;) {
651 rtnl_lock();
652 /* We held write locked pernet_ops_rwsem, and parallel
653 * setup_net() and cleanup_net() are not possible.
654 */
655 if (!atomic_read(&dev_unreg_count))
656 break;
657 __rtnl_unlock();
658
659 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
660 }
661 remove_wait_queue(&netdev_unregistering_wq, &wait);
662 }
663
664 /**
665 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
666 * @ops: struct rtnl_link_ops * to unregister
667 */
668 void rtnl_link_unregister(struct rtnl_link_ops *ops)
669 {
670 struct net *net;
671
672 mutex_lock(&link_ops_mutex);
673 list_del_rcu(&ops->list);
674 mutex_unlock(&link_ops_mutex);
675
676 synchronize_srcu(&ops->srcu);
677 cleanup_srcu_struct(&ops->srcu);
678
679 /* Close the race with setup_net() and cleanup_net() */
680 down_write(&pernet_ops_rwsem);
681 rtnl_lock_unregistering_all();
682
683 for_each_net(net)
684 __rtnl_kill_links(net, ops);
685
686 rtnl_unlock();
687 up_write(&pernet_ops_rwsem);
688 }
689 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
690
691 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
692 {
693 struct net_device *master_dev;
694 const struct rtnl_link_ops *ops;
695 size_t size = 0;
696
697 rcu_read_lock();
698
699 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
700 if (!master_dev)
701 goto out;
702
703 ops = master_dev->rtnl_link_ops;
704 if (!ops || !ops->get_slave_size)
705 goto out;
706 /* IFLA_INFO_SLAVE_DATA + nested data */
707 size = nla_total_size(sizeof(struct nlattr)) +
708 ops->get_slave_size(master_dev, dev);
709
710 out:
711 rcu_read_unlock();
712 return size;
713 }
714
715 static size_t rtnl_link_get_size(const struct net_device *dev)
716 {
717 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
718 size_t size;
719
720 if (!ops)
721 return 0;
722
723 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
724 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
725
726 if (ops->get_size)
727 /* IFLA_INFO_DATA + nested data */
728 size += nla_total_size(sizeof(struct nlattr)) +
729 ops->get_size(dev);
730
731 if (ops->get_xstats_size)
732 /* IFLA_INFO_XSTATS */
733 size += nla_total_size(ops->get_xstats_size(dev));
734
735 size += rtnl_link_get_slave_info_data_size(dev);
736
737 return size;
738 }
739
740 static LIST_HEAD(rtnl_af_ops);
741
742 static struct rtnl_af_ops *rtnl_af_lookup(const int family, int *srcu_index)
743 {
744 struct rtnl_af_ops *ops;
745
746 ASSERT_RTNL();
747
748 rcu_read_lock();
749
750 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
751 if (ops->family == family) {
752 *srcu_index = srcu_read_lock(&ops->srcu);
753 goto unlock;
754 }
755 }
756
757 ops = NULL;
758 unlock:
759 rcu_read_unlock();
760
761 return ops;
762 }
763
764 static void rtnl_af_put(struct rtnl_af_ops *ops, int srcu_index)
765 {
766 srcu_read_unlock(&ops->srcu, srcu_index);
767 }
768
769 /**
770 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
771 * @ops: struct rtnl_af_ops * to register
772 *
773 * Return: 0 on success or a negative error code.
774 */
775 int rtnl_af_register(struct rtnl_af_ops *ops)
776 {
777 int err = init_srcu_struct(&ops->srcu);
778
779 if (err)
780 return err;
781
782 rtnl_lock();
783 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
784 rtnl_unlock();
785
786 return 0;
787 }
788 EXPORT_SYMBOL_GPL(rtnl_af_register);
789
790 /**
791 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
792 * @ops: struct rtnl_af_ops * to unregister
793 */
794 void rtnl_af_unregister(struct rtnl_af_ops *ops)
795 {
796 rtnl_lock();
797 list_del_rcu(&ops->list);
798 rtnl_unlock();
799
800 synchronize_rcu();
801 synchronize_srcu(&ops->srcu);
802 cleanup_srcu_struct(&ops->srcu);
803 }
804 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
805
806 static size_t rtnl_link_get_af_size(const struct net_device *dev,
807 u32 ext_filter_mask)
808 {
809 struct rtnl_af_ops *af_ops;
810 size_t size;
811
812 /* IFLA_AF_SPEC */
813 size = nla_total_size(sizeof(struct nlattr));
814
815 rcu_read_lock();
816 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
817 if (af_ops->get_link_af_size) {
818 /* AF_* + nested data */
819 size += nla_total_size(sizeof(struct nlattr)) +
820 af_ops->get_link_af_size(dev, ext_filter_mask);
821 }
822 }
823 rcu_read_unlock();
824
825 return size;
826 }
827
828 static bool rtnl_have_link_slave_info(const struct net_device *dev)
829 {
830 struct net_device *master_dev;
831 bool ret = false;
832
833 rcu_read_lock();
834
835 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
836 if (master_dev && master_dev->rtnl_link_ops)
837 ret = true;
838 rcu_read_unlock();
839 return ret;
840 }
841
842 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
843 const struct net_device *dev)
844 {
845 struct net_device *master_dev;
846 const struct rtnl_link_ops *ops;
847 struct nlattr *slave_data;
848 int err;
849
850 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
851 if (!master_dev)
852 return 0;
853 ops = master_dev->rtnl_link_ops;
854 if (!ops)
855 return 0;
856 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
857 return -EMSGSIZE;
858 if (ops->fill_slave_info) {
859 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
860 if (!slave_data)
861 return -EMSGSIZE;
862 err = ops->fill_slave_info(skb, master_dev, dev);
863 if (err < 0)
864 goto err_cancel_slave_data;
865 nla_nest_end(skb, slave_data);
866 }
867 return 0;
868
869 err_cancel_slave_data:
870 nla_nest_cancel(skb, slave_data);
871 return err;
872 }
873
874 static int rtnl_link_info_fill(struct sk_buff *skb,
875 const struct net_device *dev)
876 {
877 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
878 struct nlattr *data;
879 int err;
880
881 if (!ops)
882 return 0;
883 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
884 return -EMSGSIZE;
885 if (ops->fill_xstats) {
886 err = ops->fill_xstats(skb, dev);
887 if (err < 0)
888 return err;
889 }
890 if (ops->fill_info) {
891 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
892 if (data == NULL)
893 return -EMSGSIZE;
894 err = ops->fill_info(skb, dev);
895 if (err < 0)
896 goto err_cancel_data;
897 nla_nest_end(skb, data);
898 }
899 return 0;
900
901 err_cancel_data:
902 nla_nest_cancel(skb, data);
903 return err;
904 }
905
906 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
907 {
908 struct nlattr *linkinfo;
909 int err = -EMSGSIZE;
910
911 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
912 if (linkinfo == NULL)
913 goto out;
914
915 err = rtnl_link_info_fill(skb, dev);
916 if (err < 0)
917 goto err_cancel_link;
918
919 err = rtnl_link_slave_info_fill(skb, dev);
920 if (err < 0)
921 goto err_cancel_link;
922
923 nla_nest_end(skb, linkinfo);
924 return 0;
925
926 err_cancel_link:
927 nla_nest_cancel(skb, linkinfo);
928 out:
929 return err;
930 }
931
932 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
933 {
934 struct sock *rtnl = net->rtnl;
935
936 return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
937 }
938
939 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
940 {
941 struct sock *rtnl = net->rtnl;
942
943 return nlmsg_unicast(rtnl, skb, pid);
944 }
945 EXPORT_SYMBOL(rtnl_unicast);
946
947 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
948 const struct nlmsghdr *nlh, gfp_t flags)
949 {
950 struct sock *rtnl = net->rtnl;
951
952 nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
953 }
954 EXPORT_SYMBOL(rtnl_notify);
955
956 void rtnl_set_sk_err(struct net *net, u32 group, int error)
957 {
958 struct sock *rtnl = net->rtnl;
959
960 netlink_set_err(rtnl, 0, group, error);
961 }
962 EXPORT_SYMBOL(rtnl_set_sk_err);
963
964 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
965 {
966 struct nlattr *mx;
967 int i, valid = 0;
968
969 /* nothing is dumped for dst_default_metrics, so just skip the loop */
970 if (metrics == dst_default_metrics.metrics)
971 return 0;
972
973 mx = nla_nest_start_noflag(skb, RTA_METRICS);
974 if (mx == NULL)
975 return -ENOBUFS;
976
977 for (i = 0; i < RTAX_MAX; i++) {
978 if (metrics[i]) {
979 if (i == RTAX_CC_ALGO - 1) {
980 char tmp[TCP_CA_NAME_MAX], *name;
981
982 name = tcp_ca_get_name_by_key(metrics[i], tmp);
983 if (!name)
984 continue;
985 if (nla_put_string(skb, i + 1, name))
986 goto nla_put_failure;
987 } else if (i == RTAX_FEATURES - 1) {
988 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
989
990 if (!user_features)
991 continue;
992 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
993 if (nla_put_u32(skb, i + 1, user_features))
994 goto nla_put_failure;
995 } else {
996 if (nla_put_u32(skb, i + 1, metrics[i]))
997 goto nla_put_failure;
998 }
999 valid++;
1000 }
1001 }
1002
1003 if (!valid) {
1004 nla_nest_cancel(skb, mx);
1005 return 0;
1006 }
1007
1008 return nla_nest_end(skb, mx);
1009
1010 nla_put_failure:
1011 nla_nest_cancel(skb, mx);
1012 return -EMSGSIZE;
1013 }
1014 EXPORT_SYMBOL(rtnetlink_put_metrics);
1015
1016 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
1017 long expires, u32 error)
1018 {
1019 struct rta_cacheinfo ci = {
1020 .rta_error = error,
1021 .rta_id = id,
1022 };
1023
1024 if (dst) {
1025 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
1026 ci.rta_used = dst->__use;
1027 ci.rta_clntref = rcuref_read(&dst->__rcuref);
1028 }
1029 if (expires) {
1030 unsigned long clock;
1031
1032 clock = jiffies_to_clock_t(abs(expires));
1033 clock = min_t(unsigned long, clock, INT_MAX);
1034 ci.rta_expires = (expires > 0) ? clock : -clock;
1035 }
1036 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
1037 }
1038 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
1039
1040 void netdev_set_operstate(struct net_device *dev, int newstate)
1041 {
1042 unsigned int old = READ_ONCE(dev->operstate);
1043
1044 do {
1045 if (old == newstate)
1046 return;
1047 } while (!try_cmpxchg(&dev->operstate, &old, newstate));
1048
1049 netdev_state_change(dev);
1050 }
1051 EXPORT_SYMBOL(netdev_set_operstate);
1052
1053 static void set_operstate(struct net_device *dev, unsigned char transition)
1054 {
1055 unsigned char operstate = READ_ONCE(dev->operstate);
1056
1057 switch (transition) {
1058 case IF_OPER_UP:
1059 if ((operstate == IF_OPER_DORMANT ||
1060 operstate == IF_OPER_TESTING ||
1061 operstate == IF_OPER_UNKNOWN) &&
1062 !netif_dormant(dev) && !netif_testing(dev))
1063 operstate = IF_OPER_UP;
1064 break;
1065
1066 case IF_OPER_TESTING:
1067 if (netif_oper_up(dev))
1068 operstate = IF_OPER_TESTING;
1069 break;
1070
1071 case IF_OPER_DORMANT:
1072 if (netif_oper_up(dev))
1073 operstate = IF_OPER_DORMANT;
1074 break;
1075 }
1076
1077 netdev_set_operstate(dev, operstate);
1078 }
1079
1080 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
1081 {
1082 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
1083 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
1084 }
1085
1086 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
1087 const struct ifinfomsg *ifm)
1088 {
1089 unsigned int flags = ifm->ifi_flags;
1090
1091 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
1092 if (ifm->ifi_change)
1093 flags = (flags & ifm->ifi_change) |
1094 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
1095
1096 return flags;
1097 }
1098
1099 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
1100 const struct rtnl_link_stats64 *b)
1101 {
1102 a->rx_packets = b->rx_packets;
1103 a->tx_packets = b->tx_packets;
1104 a->rx_bytes = b->rx_bytes;
1105 a->tx_bytes = b->tx_bytes;
1106 a->rx_errors = b->rx_errors;
1107 a->tx_errors = b->tx_errors;
1108 a->rx_dropped = b->rx_dropped;
1109 a->tx_dropped = b->tx_dropped;
1110
1111 a->multicast = b->multicast;
1112 a->collisions = b->collisions;
1113
1114 a->rx_length_errors = b->rx_length_errors;
1115 a->rx_over_errors = b->rx_over_errors;
1116 a->rx_crc_errors = b->rx_crc_errors;
1117 a->rx_frame_errors = b->rx_frame_errors;
1118 a->rx_fifo_errors = b->rx_fifo_errors;
1119 a->rx_missed_errors = b->rx_missed_errors;
1120
1121 a->tx_aborted_errors = b->tx_aborted_errors;
1122 a->tx_carrier_errors = b->tx_carrier_errors;
1123 a->tx_fifo_errors = b->tx_fifo_errors;
1124 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
1125 a->tx_window_errors = b->tx_window_errors;
1126
1127 a->rx_compressed = b->rx_compressed;
1128 a->tx_compressed = b->tx_compressed;
1129
1130 a->rx_nohandler = b->rx_nohandler;
1131 }
1132
1133 /* All VF info */
1134 static inline int rtnl_vfinfo_size(const struct net_device *dev,
1135 u32 ext_filter_mask)
1136 {
1137 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
1138 int num_vfs = dev_num_vf(dev->dev.parent);
1139 size_t size = nla_total_size(0);
1140 size += num_vfs *
1141 (nla_total_size(0) +
1142 nla_total_size(sizeof(struct ifla_vf_mac)) +
1143 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
1144 nla_total_size(sizeof(struct ifla_vf_vlan)) +
1145 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
1146 nla_total_size(MAX_VLAN_LIST_LEN *
1147 sizeof(struct ifla_vf_vlan_info)) +
1148 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
1149 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
1150 nla_total_size(sizeof(struct ifla_vf_rate)) +
1151 nla_total_size(sizeof(struct ifla_vf_link_state)) +
1152 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
1153 nla_total_size(sizeof(struct ifla_vf_trust)));
1154 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1155 size += num_vfs *
1156 (nla_total_size(0) + /* nest IFLA_VF_STATS */
1157 /* IFLA_VF_STATS_RX_PACKETS */
1158 nla_total_size_64bit(sizeof(__u64)) +
1159 /* IFLA_VF_STATS_TX_PACKETS */
1160 nla_total_size_64bit(sizeof(__u64)) +
1161 /* IFLA_VF_STATS_RX_BYTES */
1162 nla_total_size_64bit(sizeof(__u64)) +
1163 /* IFLA_VF_STATS_TX_BYTES */
1164 nla_total_size_64bit(sizeof(__u64)) +
1165 /* IFLA_VF_STATS_BROADCAST */
1166 nla_total_size_64bit(sizeof(__u64)) +
1167 /* IFLA_VF_STATS_MULTICAST */
1168 nla_total_size_64bit(sizeof(__u64)) +
1169 /* IFLA_VF_STATS_RX_DROPPED */
1170 nla_total_size_64bit(sizeof(__u64)) +
1171 /* IFLA_VF_STATS_TX_DROPPED */
1172 nla_total_size_64bit(sizeof(__u64)));
1173 }
1174 return size;
1175 } else
1176 return 0;
1177 }
1178
1179 static size_t rtnl_port_size(const struct net_device *dev,
1180 u32 ext_filter_mask)
1181 {
1182 size_t port_size = nla_total_size(4) /* PORT_VF */
1183 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
1184 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
1185 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
1186 + nla_total_size(1) /* PROT_VDP_REQUEST */
1187 + nla_total_size(2); /* PORT_VDP_RESPONSE */
1188 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
1189 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
1190 + port_size;
1191 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
1192 + port_size;
1193
1194 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1195 !(ext_filter_mask & RTEXT_FILTER_VF))
1196 return 0;
1197 if (dev_num_vf(dev->dev.parent))
1198 return port_self_size + vf_ports_size +
1199 vf_port_size * dev_num_vf(dev->dev.parent);
1200 else
1201 return port_self_size;
1202 }
1203
1204 static size_t rtnl_xdp_size(void)
1205 {
1206 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
1207 nla_total_size(1) + /* XDP_ATTACHED */
1208 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
1209 nla_total_size(4); /* XDP_<mode>_PROG_ID */
1210
1211 return xdp_size;
1212 }
1213
1214 static size_t rtnl_prop_list_size(const struct net_device *dev)
1215 {
1216 struct netdev_name_node *name_node;
1217 unsigned int cnt = 0;
1218
1219 rcu_read_lock();
1220 list_for_each_entry_rcu(name_node, &dev->name_node->list, list)
1221 cnt++;
1222 rcu_read_unlock();
1223
1224 if (!cnt)
1225 return 0;
1226
1227 return nla_total_size(0) + cnt * nla_total_size(ALTIFNAMSIZ);
1228 }
1229
1230 static size_t rtnl_proto_down_size(const struct net_device *dev)
1231 {
1232 size_t size = nla_total_size(1);
1233
1234 /* Assume dev->proto_down_reason is not zero. */
1235 size += nla_total_size(0) + nla_total_size(4);
1236
1237 return size;
1238 }
1239
1240 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1241 {
1242 size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1243
1244 if (dev->devlink_port)
1245 size += devlink_nl_port_handle_size(dev->devlink_port);
1246
1247 return size;
1248 }
1249
1250 static size_t rtnl_dpll_pin_size(const struct net_device *dev)
1251 {
1252 size_t size = nla_total_size(0); /* nest IFLA_DPLL_PIN */
1253
1254 size += dpll_netdev_pin_handle_size(dev);
1255
1256 return size;
1257 }
1258
1259 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1260 u32 ext_filter_mask)
1261 {
1262 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1263 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1264 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1265 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1266 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1267 + nla_total_size(sizeof(struct rtnl_link_stats))
1268 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1269 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1270 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1271 + nla_total_size(4) /* IFLA_TXQLEN */
1272 + nla_total_size(4) /* IFLA_WEIGHT */
1273 + nla_total_size(4) /* IFLA_MTU */
1274 + nla_total_size(4) /* IFLA_LINK */
1275 + nla_total_size(4) /* IFLA_MASTER */
1276 + nla_total_size(1) /* IFLA_CARRIER */
1277 + nla_total_size(4) /* IFLA_PROMISCUITY */
1278 + nla_total_size(4) /* IFLA_ALLMULTI */
1279 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1280 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1281 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1282 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1283 + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1284 + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1285 + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1286 + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1287 + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1288 + nla_total_size(1) /* IFLA_OPERSTATE */
1289 + nla_total_size(1) /* IFLA_LINKMODE */
1290 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1291 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1292 + nla_total_size(4) /* IFLA_GROUP */
1293 + nla_total_size(ext_filter_mask
1294 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1295 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1296 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1297 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1298 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1299 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1300 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1301 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1302 + rtnl_xdp_size() /* IFLA_XDP */
1303 + nla_total_size(4) /* IFLA_EVENT */
1304 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1305 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1306 + rtnl_proto_down_size(dev) /* proto down */
1307 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1308 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1309 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1310 + nla_total_size(4) /* IFLA_MIN_MTU */
1311 + nla_total_size(4) /* IFLA_MAX_MTU */
1312 + rtnl_prop_list_size(dev)
1313 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1314 + rtnl_devlink_port_size(dev)
1315 + rtnl_dpll_pin_size(dev)
1316 + nla_total_size(8) /* IFLA_MAX_PACING_OFFLOAD_HORIZON */
1317 + 0;
1318 }
1319
1320 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1321 {
1322 struct nlattr *vf_ports;
1323 struct nlattr *vf_port;
1324 int vf;
1325 int err;
1326
1327 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1328 if (!vf_ports)
1329 return -EMSGSIZE;
1330
1331 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1332 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1333 if (!vf_port)
1334 goto nla_put_failure;
1335 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1336 goto nla_put_failure;
1337 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1338 if (err == -EMSGSIZE)
1339 goto nla_put_failure;
1340 if (err) {
1341 nla_nest_cancel(skb, vf_port);
1342 continue;
1343 }
1344 nla_nest_end(skb, vf_port);
1345 }
1346
1347 nla_nest_end(skb, vf_ports);
1348
1349 return 0;
1350
1351 nla_put_failure:
1352 nla_nest_cancel(skb, vf_ports);
1353 return -EMSGSIZE;
1354 }
1355
1356 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1357 {
1358 struct nlattr *port_self;
1359 int err;
1360
1361 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1362 if (!port_self)
1363 return -EMSGSIZE;
1364
1365 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1366 if (err) {
1367 nla_nest_cancel(skb, port_self);
1368 return (err == -EMSGSIZE) ? err : 0;
1369 }
1370
1371 nla_nest_end(skb, port_self);
1372
1373 return 0;
1374 }
1375
1376 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1377 u32 ext_filter_mask)
1378 {
1379 int err;
1380
1381 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1382 !(ext_filter_mask & RTEXT_FILTER_VF))
1383 return 0;
1384
1385 err = rtnl_port_self_fill(skb, dev);
1386 if (err)
1387 return err;
1388
1389 if (dev_num_vf(dev->dev.parent)) {
1390 err = rtnl_vf_ports_fill(skb, dev);
1391 if (err)
1392 return err;
1393 }
1394
1395 return 0;
1396 }
1397
1398 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1399 {
1400 int err;
1401 struct netdev_phys_item_id ppid;
1402
1403 err = dev_get_phys_port_id(dev, &ppid);
1404 if (err) {
1405 if (err == -EOPNOTSUPP)
1406 return 0;
1407 return err;
1408 }
1409
1410 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1411 return -EMSGSIZE;
1412
1413 return 0;
1414 }
1415
1416 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1417 {
1418 char name[IFNAMSIZ];
1419 int err;
1420
1421 err = dev_get_phys_port_name(dev, name, sizeof(name));
1422 if (err) {
1423 if (err == -EOPNOTSUPP)
1424 return 0;
1425 return err;
1426 }
1427
1428 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1429 return -EMSGSIZE;
1430
1431 return 0;
1432 }
1433
1434 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1435 {
1436 struct netdev_phys_item_id ppid = { };
1437 int err;
1438
1439 err = dev_get_port_parent_id(dev, &ppid, false);
1440 if (err) {
1441 if (err == -EOPNOTSUPP)
1442 return 0;
1443 return err;
1444 }
1445
1446 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1447 return -EMSGSIZE;
1448
1449 return 0;
1450 }
1451
1452 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1453 struct net_device *dev)
1454 {
1455 struct rtnl_link_stats64 *sp;
1456 struct nlattr *attr;
1457
1458 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1459 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1460 if (!attr)
1461 return -EMSGSIZE;
1462
1463 sp = nla_data(attr);
1464 dev_get_stats(dev, sp);
1465
1466 attr = nla_reserve(skb, IFLA_STATS,
1467 sizeof(struct rtnl_link_stats));
1468 if (!attr)
1469 return -EMSGSIZE;
1470
1471 copy_rtnl_link_stats(nla_data(attr), sp);
1472
1473 return 0;
1474 }
1475
1476 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1477 struct net_device *dev,
1478 int vfs_num,
1479 u32 ext_filter_mask)
1480 {
1481 struct ifla_vf_rss_query_en vf_rss_query_en;
1482 struct nlattr *vf, *vfstats, *vfvlanlist;
1483 struct ifla_vf_link_state vf_linkstate;
1484 struct ifla_vf_vlan_info vf_vlan_info;
1485 struct ifla_vf_spoofchk vf_spoofchk;
1486 struct ifla_vf_tx_rate vf_tx_rate;
1487 struct ifla_vf_stats vf_stats;
1488 struct ifla_vf_trust vf_trust;
1489 struct ifla_vf_vlan vf_vlan;
1490 struct ifla_vf_rate vf_rate;
1491 struct ifla_vf_mac vf_mac;
1492 struct ifla_vf_broadcast vf_broadcast;
1493 struct ifla_vf_info ivi;
1494 struct ifla_vf_guid node_guid;
1495 struct ifla_vf_guid port_guid;
1496
1497 memset(&ivi, 0, sizeof(ivi));
1498
1499 /* Not all SR-IOV capable drivers support the
1500 * spoofcheck and "RSS query enable" query. Preset to
1501 * -1 so the user space tool can detect that the driver
1502 * didn't report anything.
1503 */
1504 ivi.spoofchk = -1;
1505 ivi.rss_query_en = -1;
1506 ivi.trusted = -1;
1507 /* The default value for VF link state is "auto"
1508 * IFLA_VF_LINK_STATE_AUTO which equals zero
1509 */
1510 ivi.linkstate = 0;
1511 /* VLAN Protocol by default is 802.1Q */
1512 ivi.vlan_proto = htons(ETH_P_8021Q);
1513 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1514 return 0;
1515
1516 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1517 memset(&node_guid, 0, sizeof(node_guid));
1518 memset(&port_guid, 0, sizeof(port_guid));
1519
1520 vf_mac.vf =
1521 vf_vlan.vf =
1522 vf_vlan_info.vf =
1523 vf_rate.vf =
1524 vf_tx_rate.vf =
1525 vf_spoofchk.vf =
1526 vf_linkstate.vf =
1527 vf_rss_query_en.vf =
1528 vf_trust.vf =
1529 node_guid.vf =
1530 port_guid.vf = ivi.vf;
1531
1532 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1533 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1534 vf_vlan.vlan = ivi.vlan;
1535 vf_vlan.qos = ivi.qos;
1536 vf_vlan_info.vlan = ivi.vlan;
1537 vf_vlan_info.qos = ivi.qos;
1538 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1539 vf_tx_rate.rate = ivi.max_tx_rate;
1540 vf_rate.min_tx_rate = ivi.min_tx_rate;
1541 vf_rate.max_tx_rate = ivi.max_tx_rate;
1542 vf_spoofchk.setting = ivi.spoofchk;
1543 vf_linkstate.link_state = ivi.linkstate;
1544 vf_rss_query_en.setting = ivi.rss_query_en;
1545 vf_trust.setting = ivi.trusted;
1546 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1547 if (!vf)
1548 return -EMSGSIZE;
1549 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1550 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1551 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1552 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1553 &vf_rate) ||
1554 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1555 &vf_tx_rate) ||
1556 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1557 &vf_spoofchk) ||
1558 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1559 &vf_linkstate) ||
1560 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1561 sizeof(vf_rss_query_en),
1562 &vf_rss_query_en) ||
1563 nla_put(skb, IFLA_VF_TRUST,
1564 sizeof(vf_trust), &vf_trust))
1565 goto nla_put_vf_failure;
1566
1567 if (dev->netdev_ops->ndo_get_vf_guid &&
1568 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1569 &port_guid)) {
1570 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1571 &node_guid) ||
1572 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1573 &port_guid))
1574 goto nla_put_vf_failure;
1575 }
1576 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1577 if (!vfvlanlist)
1578 goto nla_put_vf_failure;
1579 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1580 &vf_vlan_info)) {
1581 nla_nest_cancel(skb, vfvlanlist);
1582 goto nla_put_vf_failure;
1583 }
1584 nla_nest_end(skb, vfvlanlist);
1585 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1586 memset(&vf_stats, 0, sizeof(vf_stats));
1587 if (dev->netdev_ops->ndo_get_vf_stats)
1588 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1589 &vf_stats);
1590 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1591 if (!vfstats)
1592 goto nla_put_vf_failure;
1593 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1594 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1595 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1596 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1597 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1598 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1599 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1600 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1601 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1602 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1603 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1604 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1605 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1606 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1607 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1608 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1609 nla_nest_cancel(skb, vfstats);
1610 goto nla_put_vf_failure;
1611 }
1612 nla_nest_end(skb, vfstats);
1613 }
1614 nla_nest_end(skb, vf);
1615 return 0;
1616
1617 nla_put_vf_failure:
1618 nla_nest_cancel(skb, vf);
1619 return -EMSGSIZE;
1620 }
1621
1622 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1623 struct net_device *dev,
1624 u32 ext_filter_mask)
1625 {
1626 struct nlattr *vfinfo;
1627 int i, num_vfs;
1628
1629 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1630 return 0;
1631
1632 num_vfs = dev_num_vf(dev->dev.parent);
1633 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1634 return -EMSGSIZE;
1635
1636 if (!dev->netdev_ops->ndo_get_vf_config)
1637 return 0;
1638
1639 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1640 if (!vfinfo)
1641 return -EMSGSIZE;
1642
1643 for (i = 0; i < num_vfs; i++) {
1644 if (rtnl_fill_vfinfo(skb, dev, i, ext_filter_mask)) {
1645 nla_nest_cancel(skb, vfinfo);
1646 return -EMSGSIZE;
1647 }
1648 }
1649
1650 nla_nest_end(skb, vfinfo);
1651 return 0;
1652 }
1653
1654 static int rtnl_fill_link_ifmap(struct sk_buff *skb,
1655 const struct net_device *dev)
1656 {
1657 struct rtnl_link_ifmap map;
1658
1659 memset(&map, 0, sizeof(map));
1660 map.mem_start = READ_ONCE(dev->mem_start);
1661 map.mem_end = READ_ONCE(dev->mem_end);
1662 map.base_addr = READ_ONCE(dev->base_addr);
1663 map.irq = READ_ONCE(dev->irq);
1664 map.dma = READ_ONCE(dev->dma);
1665 map.port = READ_ONCE(dev->if_port);
1666
1667 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1668 return -EMSGSIZE;
1669
1670 return 0;
1671 }
1672
1673 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1674 {
1675 const struct bpf_prog *generic_xdp_prog;
1676 u32 res = 0;
1677
1678 rcu_read_lock();
1679 generic_xdp_prog = rcu_dereference(dev->xdp_prog);
1680 if (generic_xdp_prog)
1681 res = generic_xdp_prog->aux->id;
1682 rcu_read_unlock();
1683
1684 return res;
1685 }
1686
1687 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1688 {
1689 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1690 }
1691
1692 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1693 {
1694 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1695 }
1696
1697 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1698 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1699 u32 (*get_prog_id)(struct net_device *dev))
1700 {
1701 u32 curr_id;
1702 int err;
1703
1704 curr_id = get_prog_id(dev);
1705 if (!curr_id)
1706 return 0;
1707
1708 *prog_id = curr_id;
1709 err = nla_put_u32(skb, attr, curr_id);
1710 if (err)
1711 return err;
1712
1713 if (*mode != XDP_ATTACHED_NONE)
1714 *mode = XDP_ATTACHED_MULTI;
1715 else
1716 *mode = tgt_mode;
1717
1718 return 0;
1719 }
1720
1721 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1722 {
1723 struct nlattr *xdp;
1724 u32 prog_id;
1725 int err;
1726 u8 mode;
1727
1728 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1729 if (!xdp)
1730 return -EMSGSIZE;
1731
1732 prog_id = 0;
1733 mode = XDP_ATTACHED_NONE;
1734 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1735 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1736 if (err)
1737 goto err_cancel;
1738 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1739 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1740 if (err)
1741 goto err_cancel;
1742 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1743 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1744 if (err)
1745 goto err_cancel;
1746
1747 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1748 if (err)
1749 goto err_cancel;
1750
1751 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1752 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1753 if (err)
1754 goto err_cancel;
1755 }
1756
1757 nla_nest_end(skb, xdp);
1758 return 0;
1759
1760 err_cancel:
1761 nla_nest_cancel(skb, xdp);
1762 return err;
1763 }
1764
1765 static u32 rtnl_get_event(unsigned long event)
1766 {
1767 u32 rtnl_event_type = IFLA_EVENT_NONE;
1768
1769 switch (event) {
1770 case NETDEV_REBOOT:
1771 rtnl_event_type = IFLA_EVENT_REBOOT;
1772 break;
1773 case NETDEV_FEAT_CHANGE:
1774 rtnl_event_type = IFLA_EVENT_FEATURES;
1775 break;
1776 case NETDEV_BONDING_FAILOVER:
1777 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1778 break;
1779 case NETDEV_NOTIFY_PEERS:
1780 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1781 break;
1782 case NETDEV_RESEND_IGMP:
1783 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1784 break;
1785 case NETDEV_CHANGEINFODATA:
1786 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1787 break;
1788 default:
1789 break;
1790 }
1791
1792 return rtnl_event_type;
1793 }
1794
1795 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1796 {
1797 const struct net_device *upper_dev;
1798 int ret = 0;
1799
1800 rcu_read_lock();
1801
1802 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1803 if (upper_dev)
1804 ret = nla_put_u32(skb, IFLA_MASTER,
1805 READ_ONCE(upper_dev->ifindex));
1806
1807 rcu_read_unlock();
1808 return ret;
1809 }
1810
1811 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1812 bool force)
1813 {
1814 int iflink = dev_get_iflink(dev);
1815
1816 if (force || READ_ONCE(dev->ifindex) != iflink)
1817 return nla_put_u32(skb, IFLA_LINK, iflink);
1818
1819 return 0;
1820 }
1821
1822 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1823 struct net_device *dev)
1824 {
1825 char buf[IFALIASZ];
1826 int ret;
1827
1828 ret = dev_get_alias(dev, buf, sizeof(buf));
1829 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1830 }
1831
1832 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1833 const struct net_device *dev,
1834 struct net *src_net, gfp_t gfp)
1835 {
1836 bool put_iflink = false;
1837
1838 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1839 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1840
1841 if (!net_eq(dev_net(dev), link_net)) {
1842 int id = peernet2id_alloc(src_net, link_net, gfp);
1843
1844 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1845 return -EMSGSIZE;
1846
1847 put_iflink = true;
1848 }
1849 }
1850
1851 return nla_put_iflink(skb, dev, put_iflink);
1852 }
1853
1854 static int rtnl_fill_link_af(struct sk_buff *skb,
1855 const struct net_device *dev,
1856 u32 ext_filter_mask)
1857 {
1858 const struct rtnl_af_ops *af_ops;
1859 struct nlattr *af_spec;
1860
1861 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1862 if (!af_spec)
1863 return -EMSGSIZE;
1864
1865 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1866 struct nlattr *af;
1867 int err;
1868
1869 if (!af_ops->fill_link_af)
1870 continue;
1871
1872 af = nla_nest_start_noflag(skb, af_ops->family);
1873 if (!af)
1874 return -EMSGSIZE;
1875
1876 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1877 /*
1878 * Caller may return ENODATA to indicate that there
1879 * was no data to be dumped. This is not an error, it
1880 * means we should trim the attribute header and
1881 * continue.
1882 */
1883 if (err == -ENODATA)
1884 nla_nest_cancel(skb, af);
1885 else if (err < 0)
1886 return -EMSGSIZE;
1887
1888 nla_nest_end(skb, af);
1889 }
1890
1891 nla_nest_end(skb, af_spec);
1892 return 0;
1893 }
1894
1895 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1896 const struct net_device *dev)
1897 {
1898 struct netdev_name_node *name_node;
1899 int count = 0;
1900
1901 list_for_each_entry_rcu(name_node, &dev->name_node->list, list) {
1902 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1903 return -EMSGSIZE;
1904 count++;
1905 }
1906 return count;
1907 }
1908
1909 /* RCU protected. */
1910 static int rtnl_fill_prop_list(struct sk_buff *skb,
1911 const struct net_device *dev)
1912 {
1913 struct nlattr *prop_list;
1914 int ret;
1915
1916 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1917 if (!prop_list)
1918 return -EMSGSIZE;
1919
1920 ret = rtnl_fill_alt_ifnames(skb, dev);
1921 if (ret <= 0)
1922 goto nest_cancel;
1923
1924 nla_nest_end(skb, prop_list);
1925 return 0;
1926
1927 nest_cancel:
1928 nla_nest_cancel(skb, prop_list);
1929 return ret;
1930 }
1931
1932 static int rtnl_fill_proto_down(struct sk_buff *skb,
1933 const struct net_device *dev)
1934 {
1935 struct nlattr *pr;
1936 u32 preason;
1937
1938 if (nla_put_u8(skb, IFLA_PROTO_DOWN, READ_ONCE(dev->proto_down)))
1939 goto nla_put_failure;
1940
1941 preason = READ_ONCE(dev->proto_down_reason);
1942 if (!preason)
1943 return 0;
1944
1945 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1946 if (!pr)
1947 return -EMSGSIZE;
1948
1949 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1950 nla_nest_cancel(skb, pr);
1951 goto nla_put_failure;
1952 }
1953
1954 nla_nest_end(skb, pr);
1955 return 0;
1956
1957 nla_put_failure:
1958 return -EMSGSIZE;
1959 }
1960
1961 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1962 const struct net_device *dev)
1963 {
1964 struct nlattr *devlink_port_nest;
1965 int ret;
1966
1967 devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1968 if (!devlink_port_nest)
1969 return -EMSGSIZE;
1970
1971 if (dev->devlink_port) {
1972 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1973 if (ret < 0)
1974 goto nest_cancel;
1975 }
1976
1977 nla_nest_end(skb, devlink_port_nest);
1978 return 0;
1979
1980 nest_cancel:
1981 nla_nest_cancel(skb, devlink_port_nest);
1982 return ret;
1983 }
1984
1985 static int rtnl_fill_dpll_pin(struct sk_buff *skb,
1986 const struct net_device *dev)
1987 {
1988 struct nlattr *dpll_pin_nest;
1989 int ret;
1990
1991 dpll_pin_nest = nla_nest_start(skb, IFLA_DPLL_PIN);
1992 if (!dpll_pin_nest)
1993 return -EMSGSIZE;
1994
1995 ret = dpll_netdev_add_pin_handle(skb, dev);
1996 if (ret < 0)
1997 goto nest_cancel;
1998
1999 nla_nest_end(skb, dpll_pin_nest);
2000 return 0;
2001
2002 nest_cancel:
2003 nla_nest_cancel(skb, dpll_pin_nest);
2004 return ret;
2005 }
2006
2007 static int rtnl_fill_ifinfo(struct sk_buff *skb,
2008 struct net_device *dev, struct net *src_net,
2009 int type, u32 pid, u32 seq, u32 change,
2010 unsigned int flags, u32 ext_filter_mask,
2011 u32 event, int *new_nsid, int new_ifindex,
2012 int tgt_netnsid, gfp_t gfp)
2013 {
2014 char devname[IFNAMSIZ];
2015 struct ifinfomsg *ifm;
2016 struct nlmsghdr *nlh;
2017 struct Qdisc *qdisc;
2018
2019 ASSERT_RTNL();
2020 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
2021 if (nlh == NULL)
2022 return -EMSGSIZE;
2023
2024 ifm = nlmsg_data(nlh);
2025 ifm->ifi_family = AF_UNSPEC;
2026 ifm->__ifi_pad = 0;
2027 ifm->ifi_type = READ_ONCE(dev->type);
2028 ifm->ifi_index = READ_ONCE(dev->ifindex);
2029 ifm->ifi_flags = dev_get_flags(dev);
2030 ifm->ifi_change = change;
2031
2032 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
2033 goto nla_put_failure;
2034
2035 netdev_copy_name(dev, devname);
2036 if (nla_put_string(skb, IFLA_IFNAME, devname))
2037 goto nla_put_failure;
2038
2039 if (nla_put_u32(skb, IFLA_TXQLEN, READ_ONCE(dev->tx_queue_len)) ||
2040 nla_put_u8(skb, IFLA_OPERSTATE,
2041 netif_running(dev) ? READ_ONCE(dev->operstate) :
2042 IF_OPER_DOWN) ||
2043 nla_put_u8(skb, IFLA_LINKMODE, READ_ONCE(dev->link_mode)) ||
2044 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
2045 nla_put_u32(skb, IFLA_MIN_MTU, READ_ONCE(dev->min_mtu)) ||
2046 nla_put_u32(skb, IFLA_MAX_MTU, READ_ONCE(dev->max_mtu)) ||
2047 nla_put_u32(skb, IFLA_GROUP, READ_ONCE(dev->group)) ||
2048 nla_put_u32(skb, IFLA_PROMISCUITY, READ_ONCE(dev->promiscuity)) ||
2049 nla_put_u32(skb, IFLA_ALLMULTI, READ_ONCE(dev->allmulti)) ||
2050 nla_put_u32(skb, IFLA_NUM_TX_QUEUES,
2051 READ_ONCE(dev->num_tx_queues)) ||
2052 nla_put_u32(skb, IFLA_GSO_MAX_SEGS,
2053 READ_ONCE(dev->gso_max_segs)) ||
2054 nla_put_u32(skb, IFLA_GSO_MAX_SIZE,
2055 READ_ONCE(dev->gso_max_size)) ||
2056 nla_put_u32(skb, IFLA_GRO_MAX_SIZE,
2057 READ_ONCE(dev->gro_max_size)) ||
2058 nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE,
2059 READ_ONCE(dev->gso_ipv4_max_size)) ||
2060 nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE,
2061 READ_ONCE(dev->gro_ipv4_max_size)) ||
2062 nla_put_u32(skb, IFLA_TSO_MAX_SIZE,
2063 READ_ONCE(dev->tso_max_size)) ||
2064 nla_put_u32(skb, IFLA_TSO_MAX_SEGS,
2065 READ_ONCE(dev->tso_max_segs)) ||
2066 nla_put_uint(skb, IFLA_MAX_PACING_OFFLOAD_HORIZON,
2067 READ_ONCE(dev->max_pacing_offload_horizon)) ||
2068 #ifdef CONFIG_RPS
2069 nla_put_u32(skb, IFLA_NUM_RX_QUEUES,
2070 READ_ONCE(dev->num_rx_queues)) ||
2071 #endif
2072 put_master_ifindex(skb, dev) ||
2073 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
2074 nla_put_ifalias(skb, dev) ||
2075 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
2076 atomic_read(&dev->carrier_up_count) +
2077 atomic_read(&dev->carrier_down_count)) ||
2078 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
2079 atomic_read(&dev->carrier_up_count)) ||
2080 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
2081 atomic_read(&dev->carrier_down_count)))
2082 goto nla_put_failure;
2083
2084 if (rtnl_fill_proto_down(skb, dev))
2085 goto nla_put_failure;
2086
2087 if (event != IFLA_EVENT_NONE) {
2088 if (nla_put_u32(skb, IFLA_EVENT, event))
2089 goto nla_put_failure;
2090 }
2091
2092 if (dev->addr_len) {
2093 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
2094 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
2095 goto nla_put_failure;
2096 }
2097
2098 if (rtnl_phys_port_id_fill(skb, dev))
2099 goto nla_put_failure;
2100
2101 if (rtnl_phys_port_name_fill(skb, dev))
2102 goto nla_put_failure;
2103
2104 if (rtnl_phys_switch_id_fill(skb, dev))
2105 goto nla_put_failure;
2106
2107 if (rtnl_fill_stats(skb, dev))
2108 goto nla_put_failure;
2109
2110 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
2111 goto nla_put_failure;
2112
2113 if (rtnl_port_fill(skb, dev, ext_filter_mask))
2114 goto nla_put_failure;
2115
2116 if (rtnl_xdp_fill(skb, dev))
2117 goto nla_put_failure;
2118
2119 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
2120 if (rtnl_link_fill(skb, dev) < 0)
2121 goto nla_put_failure;
2122 }
2123
2124 if (new_nsid &&
2125 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
2126 goto nla_put_failure;
2127 if (new_ifindex &&
2128 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
2129 goto nla_put_failure;
2130
2131 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
2132 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
2133 goto nla_put_failure;
2134
2135 rcu_read_lock();
2136 if (rtnl_fill_link_netnsid(skb, dev, src_net, GFP_ATOMIC))
2137 goto nla_put_failure_rcu;
2138 qdisc = rcu_dereference(dev->qdisc);
2139 if (qdisc && nla_put_string(skb, IFLA_QDISC, qdisc->ops->id))
2140 goto nla_put_failure_rcu;
2141 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
2142 goto nla_put_failure_rcu;
2143 if (rtnl_fill_link_ifmap(skb, dev))
2144 goto nla_put_failure_rcu;
2145 if (rtnl_fill_prop_list(skb, dev))
2146 goto nla_put_failure_rcu;
2147 rcu_read_unlock();
2148
2149 if (dev->dev.parent &&
2150 nla_put_string(skb, IFLA_PARENT_DEV_NAME,
2151 dev_name(dev->dev.parent)))
2152 goto nla_put_failure;
2153
2154 if (dev->dev.parent && dev->dev.parent->bus &&
2155 nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
2156 dev->dev.parent->bus->name))
2157 goto nla_put_failure;
2158
2159 if (rtnl_fill_devlink_port(skb, dev))
2160 goto nla_put_failure;
2161
2162 if (rtnl_fill_dpll_pin(skb, dev))
2163 goto nla_put_failure;
2164
2165 nlmsg_end(skb, nlh);
2166 return 0;
2167
2168 nla_put_failure_rcu:
2169 rcu_read_unlock();
2170 nla_put_failure:
2171 nlmsg_cancel(skb, nlh);
2172 return -EMSGSIZE;
2173 }
2174
2175 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
2176 [IFLA_UNSPEC] = { .strict_start_type = IFLA_DPLL_PIN },
2177 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
2178 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
2179 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
2180 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
2181 [IFLA_MTU] = { .type = NLA_U32 },
2182 [IFLA_LINK] = { .type = NLA_U32 },
2183 [IFLA_MASTER] = { .type = NLA_U32 },
2184 [IFLA_CARRIER] = { .type = NLA_U8 },
2185 [IFLA_TXQLEN] = { .type = NLA_U32 },
2186 [IFLA_WEIGHT] = { .type = NLA_U32 },
2187 [IFLA_OPERSTATE] = { .type = NLA_U8 },
2188 [IFLA_LINKMODE] = { .type = NLA_U8 },
2189 [IFLA_LINKINFO] = { .type = NLA_NESTED },
2190 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
2191 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
2192 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
2193 * allow 0-length string (needed to remove an alias).
2194 */
2195 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
2196 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
2197 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
2198 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
2199 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
2200 [IFLA_EXT_MASK] = { .type = NLA_U32 },
2201 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
2202 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
2203 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
2204 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
2205 [IFLA_GSO_MAX_SIZE] = NLA_POLICY_MIN(NLA_U32, MAX_TCP_HEADER + 1),
2206 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2207 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
2208 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2209 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
2210 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
2211 [IFLA_XDP] = { .type = NLA_NESTED },
2212 [IFLA_EVENT] = { .type = NLA_U32 },
2213 [IFLA_GROUP] = { .type = NLA_U32 },
2214 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
2215 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
2216 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
2217 [IFLA_MIN_MTU] = { .type = NLA_U32 },
2218 [IFLA_MAX_MTU] = { .type = NLA_U32 },
2219 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
2220 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
2221 .len = ALTIFNAMSIZ - 1 },
2222 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
2223 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
2224 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
2225 [IFLA_PARENT_DEV_NAME] = { .type = NLA_NUL_STRING },
2226 [IFLA_GRO_MAX_SIZE] = { .type = NLA_U32 },
2227 [IFLA_TSO_MAX_SIZE] = { .type = NLA_REJECT },
2228 [IFLA_TSO_MAX_SEGS] = { .type = NLA_REJECT },
2229 [IFLA_ALLMULTI] = { .type = NLA_REJECT },
2230 [IFLA_GSO_IPV4_MAX_SIZE] = NLA_POLICY_MIN(NLA_U32, MAX_TCP_HEADER + 1),
2231 [IFLA_GRO_IPV4_MAX_SIZE] = { .type = NLA_U32 },
2232 };
2233
2234 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
2235 [IFLA_INFO_KIND] = { .type = NLA_STRING },
2236 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
2237 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
2238 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
2239 };
2240
2241 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
2242 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
2243 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
2244 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
2245 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
2246 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
2247 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
2248 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
2249 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
2250 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
2251 [IFLA_VF_STATS] = { .type = NLA_NESTED },
2252 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
2253 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
2254 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
2255 };
2256
2257 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2258 [IFLA_PORT_VF] = { .type = NLA_U32 },
2259 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
2260 .len = PORT_PROFILE_MAX },
2261 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2262 .len = PORT_UUID_MAX },
2263 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
2264 .len = PORT_UUID_MAX },
2265 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
2266 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
2267
2268 /* Unused, but we need to keep it here since user space could
2269 * fill it. It's also broken with regard to NLA_BINARY use in
2270 * combination with structs.
2271 */
2272 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
2273 .len = sizeof(struct ifla_port_vsi) },
2274 };
2275
2276 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2277 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2278 [IFLA_XDP_FD] = { .type = NLA_S32 },
2279 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
2280 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
2281 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
2282 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
2283 };
2284
2285 static struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla,
2286 int *ops_srcu_index)
2287 {
2288 struct nlattr *linfo[IFLA_INFO_MAX + 1];
2289 struct rtnl_link_ops *ops = NULL;
2290
2291 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2292 return NULL;
2293
2294 if (linfo[IFLA_INFO_KIND]) {
2295 char kind[MODULE_NAME_LEN];
2296
2297 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2298 ops = rtnl_link_ops_get(kind, ops_srcu_index);
2299 }
2300
2301 return ops;
2302 }
2303
2304 static bool link_master_filtered(struct net_device *dev, int master_idx)
2305 {
2306 struct net_device *master;
2307
2308 if (!master_idx)
2309 return false;
2310
2311 master = netdev_master_upper_dev_get(dev);
2312
2313 /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2314 * another invalid value for ifindex to denote "no master".
2315 */
2316 if (master_idx == -1)
2317 return !!master;
2318
2319 if (!master || master->ifindex != master_idx)
2320 return true;
2321
2322 return false;
2323 }
2324
2325 static bool link_kind_filtered(const struct net_device *dev,
2326 const struct rtnl_link_ops *kind_ops)
2327 {
2328 if (kind_ops && dev->rtnl_link_ops != kind_ops)
2329 return true;
2330
2331 return false;
2332 }
2333
2334 static bool link_dump_filtered(struct net_device *dev,
2335 int master_idx,
2336 const struct rtnl_link_ops *kind_ops)
2337 {
2338 if (link_master_filtered(dev, master_idx) ||
2339 link_kind_filtered(dev, kind_ops))
2340 return true;
2341
2342 return false;
2343 }
2344
2345 /**
2346 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2347 * @sk: netlink socket
2348 * @netnsid: network namespace identifier
2349 *
2350 * Returns the network namespace identified by netnsid on success or an error
2351 * pointer on failure.
2352 */
2353 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2354 {
2355 struct net *net;
2356
2357 net = get_net_ns_by_id(sock_net(sk), netnsid);
2358 if (!net)
2359 return ERR_PTR(-EINVAL);
2360
2361 /* For now, the caller is required to have CAP_NET_ADMIN in
2362 * the user namespace owning the target net ns.
2363 */
2364 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2365 put_net(net);
2366 return ERR_PTR(-EACCES);
2367 }
2368 return net;
2369 }
2370 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2371
2372 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2373 bool strict_check, struct nlattr **tb,
2374 struct netlink_ext_ack *extack)
2375 {
2376 int hdrlen;
2377
2378 if (strict_check) {
2379 struct ifinfomsg *ifm;
2380
2381 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2382 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2383 return -EINVAL;
2384 }
2385
2386 ifm = nlmsg_data(nlh);
2387 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2388 ifm->ifi_change) {
2389 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2390 return -EINVAL;
2391 }
2392 if (ifm->ifi_index) {
2393 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2394 return -EINVAL;
2395 }
2396
2397 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2398 IFLA_MAX, ifla_policy,
2399 extack);
2400 }
2401
2402 /* A hack to preserve kernel<->userspace interface.
2403 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2404 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2405 * what iproute2 < v3.9.0 used.
2406 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2407 * attribute, its netlink message is shorter than struct ifinfomsg.
2408 */
2409 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2410 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2411
2412 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2413 extack);
2414 }
2415
2416 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2417 {
2418 struct netlink_ext_ack *extack = cb->extack;
2419 struct rtnl_link_ops *kind_ops = NULL;
2420 const struct nlmsghdr *nlh = cb->nlh;
2421 struct net *net = sock_net(skb->sk);
2422 unsigned int flags = NLM_F_MULTI;
2423 struct nlattr *tb[IFLA_MAX+1];
2424 struct {
2425 unsigned long ifindex;
2426 } *ctx = (void *)cb->ctx;
2427 struct net *tgt_net = net;
2428 u32 ext_filter_mask = 0;
2429 struct net_device *dev;
2430 int ops_srcu_index;
2431 int master_idx = 0;
2432 int netnsid = -1;
2433 int err, i;
2434
2435 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2436 if (err < 0) {
2437 if (cb->strict_check)
2438 return err;
2439
2440 goto walk_entries;
2441 }
2442
2443 for (i = 0; i <= IFLA_MAX; ++i) {
2444 if (!tb[i])
2445 continue;
2446
2447 /* new attributes should only be added with strict checking */
2448 switch (i) {
2449 case IFLA_TARGET_NETNSID:
2450 netnsid = nla_get_s32(tb[i]);
2451 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2452 if (IS_ERR(tgt_net)) {
2453 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2454 err = PTR_ERR(tgt_net);
2455 netnsid = -1;
2456 goto out;
2457 }
2458 break;
2459 case IFLA_EXT_MASK:
2460 ext_filter_mask = nla_get_u32(tb[i]);
2461 break;
2462 case IFLA_MASTER:
2463 master_idx = nla_get_u32(tb[i]);
2464 break;
2465 case IFLA_LINKINFO:
2466 kind_ops = linkinfo_to_kind_ops(tb[i], &ops_srcu_index);
2467 break;
2468 default:
2469 if (cb->strict_check) {
2470 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2471 err = -EINVAL;
2472 goto out;
2473 }
2474 }
2475 }
2476
2477 if (master_idx || kind_ops)
2478 flags |= NLM_F_DUMP_FILTERED;
2479
2480 walk_entries:
2481 err = 0;
2482 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
2483 if (link_dump_filtered(dev, master_idx, kind_ops))
2484 continue;
2485 err = rtnl_fill_ifinfo(skb, dev, net, RTM_NEWLINK,
2486 NETLINK_CB(cb->skb).portid,
2487 nlh->nlmsg_seq, 0, flags,
2488 ext_filter_mask, 0, NULL, 0,
2489 netnsid, GFP_KERNEL);
2490 if (err < 0)
2491 break;
2492 }
2493
2494
2495 cb->seq = tgt_net->dev_base_seq;
2496 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2497
2498 out:
2499
2500 if (kind_ops)
2501 rtnl_link_ops_put(kind_ops, ops_srcu_index);
2502 if (netnsid >= 0)
2503 put_net(tgt_net);
2504
2505 return err;
2506 }
2507
2508 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2509 struct netlink_ext_ack *exterr)
2510 {
2511 const struct ifinfomsg *ifmp;
2512 const struct nlattr *attrs;
2513 size_t len;
2514
2515 ifmp = nla_data(nla_peer);
2516 attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2517 len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2518
2519 if (ifmp->ifi_index < 0) {
2520 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2521 "ifindex can't be negative");
2522 return -EINVAL;
2523 }
2524
2525 return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2526 exterr);
2527 }
2528 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2529
2530 static struct net *rtnl_link_get_net_ifla(struct nlattr *tb[])
2531 {
2532 struct net *net = NULL;
2533
2534 /* Examine the link attributes and figure out which
2535 * network namespace we are talking about.
2536 */
2537 if (tb[IFLA_NET_NS_PID])
2538 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2539 else if (tb[IFLA_NET_NS_FD])
2540 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2541
2542 return net;
2543 }
2544
2545 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2546 {
2547 struct net *net = rtnl_link_get_net_ifla(tb);
2548
2549 if (!net)
2550 net = get_net(src_net);
2551
2552 return net;
2553 }
2554 EXPORT_SYMBOL(rtnl_link_get_net);
2555
2556 /* Figure out which network namespace we are talking about by
2557 * examining the link attributes in the following order:
2558 *
2559 * 1. IFLA_NET_NS_PID
2560 * 2. IFLA_NET_NS_FD
2561 * 3. IFLA_TARGET_NETNSID
2562 */
2563 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2564 struct nlattr *tb[])
2565 {
2566 struct net *net;
2567
2568 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2569 return rtnl_link_get_net(src_net, tb);
2570
2571 if (!tb[IFLA_TARGET_NETNSID])
2572 return get_net(src_net);
2573
2574 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2575 if (!net)
2576 return ERR_PTR(-EINVAL);
2577
2578 return net;
2579 }
2580
2581 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2582 struct net *src_net,
2583 struct nlattr *tb[], int cap)
2584 {
2585 struct net *net;
2586
2587 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2588 if (IS_ERR(net))
2589 return net;
2590
2591 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2592 put_net(net);
2593 return ERR_PTR(-EPERM);
2594 }
2595
2596 return net;
2597 }
2598
2599 /* Verify that rtnetlink requests do not pass additional properties
2600 * potentially referring to different network namespaces.
2601 */
2602 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2603 struct netlink_ext_ack *extack,
2604 bool netns_id_only)
2605 {
2606
2607 if (netns_id_only) {
2608 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2609 return 0;
2610
2611 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2612 return -EOPNOTSUPP;
2613 }
2614
2615 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2616 goto invalid_attr;
2617
2618 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2619 goto invalid_attr;
2620
2621 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2622 goto invalid_attr;
2623
2624 return 0;
2625
2626 invalid_attr:
2627 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2628 return -EINVAL;
2629 }
2630
2631 static int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2632 int max_tx_rate)
2633 {
2634 const struct net_device_ops *ops = dev->netdev_ops;
2635
2636 if (!ops->ndo_set_vf_rate)
2637 return -EOPNOTSUPP;
2638 if (max_tx_rate && max_tx_rate < min_tx_rate)
2639 return -EINVAL;
2640
2641 return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2642 }
2643
2644 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2645 struct netlink_ext_ack *extack)
2646 {
2647 if (tb[IFLA_ADDRESS] &&
2648 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2649 return -EINVAL;
2650
2651 if (tb[IFLA_BROADCAST] &&
2652 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2653 return -EINVAL;
2654
2655 if (tb[IFLA_GSO_MAX_SIZE] &&
2656 nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2657 NL_SET_ERR_MSG(extack, "too big gso_max_size");
2658 return -EINVAL;
2659 }
2660
2661 if (tb[IFLA_GSO_MAX_SEGS] &&
2662 (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2663 nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2664 NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2665 return -EINVAL;
2666 }
2667
2668 if (tb[IFLA_GRO_MAX_SIZE] &&
2669 nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2670 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2671 return -EINVAL;
2672 }
2673
2674 if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2675 nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2676 NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2677 return -EINVAL;
2678 }
2679
2680 if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2681 nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2682 NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2683 return -EINVAL;
2684 }
2685
2686 if (tb[IFLA_AF_SPEC]) {
2687 struct nlattr *af;
2688 int rem, err;
2689
2690 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2691 struct rtnl_af_ops *af_ops;
2692 int af_ops_srcu_index;
2693
2694 af_ops = rtnl_af_lookup(nla_type(af), &af_ops_srcu_index);
2695 if (!af_ops)
2696 return -EAFNOSUPPORT;
2697
2698 if (!af_ops->set_link_af)
2699 err = -EOPNOTSUPP;
2700 else if (af_ops->validate_link_af)
2701 err = af_ops->validate_link_af(dev, af, extack);
2702 else
2703 err = 0;
2704
2705 rtnl_af_put(af_ops, af_ops_srcu_index);
2706
2707 if (err < 0)
2708 return err;
2709 }
2710 }
2711
2712 return 0;
2713 }
2714
2715 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2716 int guid_type)
2717 {
2718 const struct net_device_ops *ops = dev->netdev_ops;
2719
2720 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2721 }
2722
2723 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2724 {
2725 if (dev->type != ARPHRD_INFINIBAND)
2726 return -EOPNOTSUPP;
2727
2728 return handle_infiniband_guid(dev, ivt, guid_type);
2729 }
2730
2731 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2732 {
2733 const struct net_device_ops *ops = dev->netdev_ops;
2734 int err = -EINVAL;
2735
2736 if (tb[IFLA_VF_MAC]) {
2737 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2738
2739 if (ivm->vf >= INT_MAX)
2740 return -EINVAL;
2741 err = -EOPNOTSUPP;
2742 if (ops->ndo_set_vf_mac)
2743 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2744 ivm->mac);
2745 if (err < 0)
2746 return err;
2747 }
2748
2749 if (tb[IFLA_VF_VLAN]) {
2750 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2751
2752 if (ivv->vf >= INT_MAX)
2753 return -EINVAL;
2754 err = -EOPNOTSUPP;
2755 if (ops->ndo_set_vf_vlan)
2756 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2757 ivv->qos,
2758 htons(ETH_P_8021Q));
2759 if (err < 0)
2760 return err;
2761 }
2762
2763 if (tb[IFLA_VF_VLAN_LIST]) {
2764 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2765 struct nlattr *attr;
2766 int rem, len = 0;
2767
2768 err = -EOPNOTSUPP;
2769 if (!ops->ndo_set_vf_vlan)
2770 return err;
2771
2772 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2773 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2774 nla_len(attr) < sizeof(struct ifla_vf_vlan_info)) {
2775 return -EINVAL;
2776 }
2777 if (len >= MAX_VLAN_LIST_LEN)
2778 return -EOPNOTSUPP;
2779 ivvl[len] = nla_data(attr);
2780
2781 len++;
2782 }
2783 if (len == 0)
2784 return -EINVAL;
2785
2786 if (ivvl[0]->vf >= INT_MAX)
2787 return -EINVAL;
2788 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2789 ivvl[0]->qos, ivvl[0]->vlan_proto);
2790 if (err < 0)
2791 return err;
2792 }
2793
2794 if (tb[IFLA_VF_TX_RATE]) {
2795 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2796 struct ifla_vf_info ivf;
2797
2798 if (ivt->vf >= INT_MAX)
2799 return -EINVAL;
2800 err = -EOPNOTSUPP;
2801 if (ops->ndo_get_vf_config)
2802 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2803 if (err < 0)
2804 return err;
2805
2806 err = rtnl_set_vf_rate(dev, ivt->vf,
2807 ivf.min_tx_rate, ivt->rate);
2808 if (err < 0)
2809 return err;
2810 }
2811
2812 if (tb[IFLA_VF_RATE]) {
2813 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2814
2815 if (ivt->vf >= INT_MAX)
2816 return -EINVAL;
2817
2818 err = rtnl_set_vf_rate(dev, ivt->vf,
2819 ivt->min_tx_rate, ivt->max_tx_rate);
2820 if (err < 0)
2821 return err;
2822 }
2823
2824 if (tb[IFLA_VF_SPOOFCHK]) {
2825 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2826
2827 if (ivs->vf >= INT_MAX)
2828 return -EINVAL;
2829 err = -EOPNOTSUPP;
2830 if (ops->ndo_set_vf_spoofchk)
2831 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2832 ivs->setting);
2833 if (err < 0)
2834 return err;
2835 }
2836
2837 if (tb[IFLA_VF_LINK_STATE]) {
2838 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2839
2840 if (ivl->vf >= INT_MAX)
2841 return -EINVAL;
2842 err = -EOPNOTSUPP;
2843 if (ops->ndo_set_vf_link_state)
2844 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2845 ivl->link_state);
2846 if (err < 0)
2847 return err;
2848 }
2849
2850 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2851 struct ifla_vf_rss_query_en *ivrssq_en;
2852
2853 err = -EOPNOTSUPP;
2854 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2855 if (ivrssq_en->vf >= INT_MAX)
2856 return -EINVAL;
2857 if (ops->ndo_set_vf_rss_query_en)
2858 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2859 ivrssq_en->setting);
2860 if (err < 0)
2861 return err;
2862 }
2863
2864 if (tb[IFLA_VF_TRUST]) {
2865 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2866
2867 if (ivt->vf >= INT_MAX)
2868 return -EINVAL;
2869 err = -EOPNOTSUPP;
2870 if (ops->ndo_set_vf_trust)
2871 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2872 if (err < 0)
2873 return err;
2874 }
2875
2876 if (tb[IFLA_VF_IB_NODE_GUID]) {
2877 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2878
2879 if (ivt->vf >= INT_MAX)
2880 return -EINVAL;
2881 if (!ops->ndo_set_vf_guid)
2882 return -EOPNOTSUPP;
2883 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2884 }
2885
2886 if (tb[IFLA_VF_IB_PORT_GUID]) {
2887 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2888
2889 if (ivt->vf >= INT_MAX)
2890 return -EINVAL;
2891 if (!ops->ndo_set_vf_guid)
2892 return -EOPNOTSUPP;
2893
2894 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2895 }
2896
2897 return err;
2898 }
2899
2900 static int do_set_master(struct net_device *dev, int ifindex,
2901 struct netlink_ext_ack *extack)
2902 {
2903 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2904 const struct net_device_ops *ops;
2905 int err;
2906
2907 if (upper_dev) {
2908 if (upper_dev->ifindex == ifindex)
2909 return 0;
2910 ops = upper_dev->netdev_ops;
2911 if (ops->ndo_del_slave) {
2912 err = ops->ndo_del_slave(upper_dev, dev);
2913 if (err)
2914 return err;
2915 } else {
2916 return -EOPNOTSUPP;
2917 }
2918 }
2919
2920 if (ifindex) {
2921 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2922 if (!upper_dev)
2923 return -EINVAL;
2924 ops = upper_dev->netdev_ops;
2925 if (ops->ndo_add_slave) {
2926 err = ops->ndo_add_slave(upper_dev, dev, extack);
2927 if (err)
2928 return err;
2929 } else {
2930 return -EOPNOTSUPP;
2931 }
2932 }
2933 return 0;
2934 }
2935
2936 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2937 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2938 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2939 };
2940
2941 static int do_set_proto_down(struct net_device *dev,
2942 struct nlattr *nl_proto_down,
2943 struct nlattr *nl_proto_down_reason,
2944 struct netlink_ext_ack *extack)
2945 {
2946 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2947 unsigned long mask = 0;
2948 u32 value;
2949 bool proto_down;
2950 int err;
2951
2952 if (!dev->change_proto_down) {
2953 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2954 return -EOPNOTSUPP;
2955 }
2956
2957 if (nl_proto_down_reason) {
2958 err = nla_parse_nested_deprecated(pdreason,
2959 IFLA_PROTO_DOWN_REASON_MAX,
2960 nl_proto_down_reason,
2961 ifla_proto_down_reason_policy,
2962 NULL);
2963 if (err < 0)
2964 return err;
2965
2966 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2967 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2968 return -EINVAL;
2969 }
2970
2971 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2972
2973 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2974 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2975
2976 dev_change_proto_down_reason(dev, mask, value);
2977 }
2978
2979 if (nl_proto_down) {
2980 proto_down = nla_get_u8(nl_proto_down);
2981
2982 /* Don't turn off protodown if there are active reasons */
2983 if (!proto_down && dev->proto_down_reason) {
2984 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2985 return -EBUSY;
2986 }
2987 err = dev_change_proto_down(dev,
2988 proto_down);
2989 if (err)
2990 return err;
2991 }
2992
2993 return 0;
2994 }
2995
2996 #define DO_SETLINK_MODIFIED 0x01
2997 /* notify flag means notify + modified. */
2998 #define DO_SETLINK_NOTIFY 0x03
2999 static int do_setlink(const struct sk_buff *skb, struct net_device *dev,
3000 struct net *tgt_net, struct ifinfomsg *ifm,
3001 struct netlink_ext_ack *extack,
3002 struct nlattr **tb, int status)
3003 {
3004 const struct net_device_ops *ops = dev->netdev_ops;
3005 char ifname[IFNAMSIZ];
3006 int err;
3007
3008 err = validate_linkmsg(dev, tb, extack);
3009 if (err < 0)
3010 goto errout;
3011
3012 if (tb[IFLA_IFNAME])
3013 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3014 else
3015 ifname[0] = '\0';
3016
3017 if (!net_eq(tgt_net, dev_net(dev))) {
3018 const char *pat = ifname[0] ? ifname : NULL;
3019 int new_ifindex;
3020
3021 new_ifindex = nla_get_s32_default(tb[IFLA_NEW_IFINDEX], 0);
3022
3023 err = __dev_change_net_namespace(dev, tgt_net, pat, new_ifindex);
3024 if (err)
3025 goto errout;
3026
3027 status |= DO_SETLINK_MODIFIED;
3028 }
3029
3030 if (tb[IFLA_MAP]) {
3031 struct rtnl_link_ifmap *u_map;
3032 struct ifmap k_map;
3033
3034 if (!ops->ndo_set_config) {
3035 err = -EOPNOTSUPP;
3036 goto errout;
3037 }
3038
3039 if (!netif_device_present(dev)) {
3040 err = -ENODEV;
3041 goto errout;
3042 }
3043
3044 u_map = nla_data(tb[IFLA_MAP]);
3045 k_map.mem_start = (unsigned long) u_map->mem_start;
3046 k_map.mem_end = (unsigned long) u_map->mem_end;
3047 k_map.base_addr = (unsigned short) u_map->base_addr;
3048 k_map.irq = (unsigned char) u_map->irq;
3049 k_map.dma = (unsigned char) u_map->dma;
3050 k_map.port = (unsigned char) u_map->port;
3051
3052 err = ops->ndo_set_config(dev, &k_map);
3053 if (err < 0)
3054 goto errout;
3055
3056 status |= DO_SETLINK_NOTIFY;
3057 }
3058
3059 if (tb[IFLA_ADDRESS]) {
3060 struct sockaddr *sa;
3061 int len;
3062
3063 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
3064 sizeof(*sa));
3065 sa = kmalloc(len, GFP_KERNEL);
3066 if (!sa) {
3067 err = -ENOMEM;
3068 goto errout;
3069 }
3070 sa->sa_family = dev->type;
3071 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
3072 dev->addr_len);
3073 err = dev_set_mac_address_user(dev, sa, extack);
3074 kfree(sa);
3075 if (err)
3076 goto errout;
3077 status |= DO_SETLINK_MODIFIED;
3078 }
3079
3080 if (tb[IFLA_MTU]) {
3081 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
3082 if (err < 0)
3083 goto errout;
3084 status |= DO_SETLINK_MODIFIED;
3085 }
3086
3087 if (tb[IFLA_GROUP]) {
3088 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3089 status |= DO_SETLINK_NOTIFY;
3090 }
3091
3092 /*
3093 * Interface selected by interface index but interface
3094 * name provided implies that a name change has been
3095 * requested.
3096 */
3097 if (ifm->ifi_index > 0 && ifname[0]) {
3098 err = dev_change_name(dev, ifname);
3099 if (err < 0)
3100 goto errout;
3101 status |= DO_SETLINK_MODIFIED;
3102 }
3103
3104 if (tb[IFLA_IFALIAS]) {
3105 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
3106 nla_len(tb[IFLA_IFALIAS]));
3107 if (err < 0)
3108 goto errout;
3109 status |= DO_SETLINK_NOTIFY;
3110 }
3111
3112 if (tb[IFLA_BROADCAST]) {
3113 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
3114 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
3115 }
3116
3117 if (ifm->ifi_flags || ifm->ifi_change) {
3118 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3119 extack);
3120 if (err < 0)
3121 goto errout;
3122 }
3123
3124 if (tb[IFLA_MASTER]) {
3125 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3126 if (err)
3127 goto errout;
3128 status |= DO_SETLINK_MODIFIED;
3129 }
3130
3131 if (tb[IFLA_CARRIER]) {
3132 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
3133 if (err)
3134 goto errout;
3135 status |= DO_SETLINK_MODIFIED;
3136 }
3137
3138 if (tb[IFLA_TXQLEN]) {
3139 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
3140
3141 err = dev_change_tx_queue_len(dev, value);
3142 if (err)
3143 goto errout;
3144 status |= DO_SETLINK_MODIFIED;
3145 }
3146
3147 if (tb[IFLA_GSO_MAX_SIZE]) {
3148 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
3149
3150 if (dev->gso_max_size ^ max_size) {
3151 netif_set_gso_max_size(dev, max_size);
3152 status |= DO_SETLINK_MODIFIED;
3153 }
3154 }
3155
3156 if (tb[IFLA_GSO_MAX_SEGS]) {
3157 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3158
3159 if (dev->gso_max_segs ^ max_segs) {
3160 netif_set_gso_max_segs(dev, max_segs);
3161 status |= DO_SETLINK_MODIFIED;
3162 }
3163 }
3164
3165 if (tb[IFLA_GRO_MAX_SIZE]) {
3166 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
3167
3168 if (dev->gro_max_size ^ gro_max_size) {
3169 netif_set_gro_max_size(dev, gro_max_size);
3170 status |= DO_SETLINK_MODIFIED;
3171 }
3172 }
3173
3174 if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
3175 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
3176
3177 if (dev->gso_ipv4_max_size ^ max_size) {
3178 netif_set_gso_ipv4_max_size(dev, max_size);
3179 status |= DO_SETLINK_MODIFIED;
3180 }
3181 }
3182
3183 if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
3184 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
3185
3186 if (dev->gro_ipv4_max_size ^ gro_max_size) {
3187 netif_set_gro_ipv4_max_size(dev, gro_max_size);
3188 status |= DO_SETLINK_MODIFIED;
3189 }
3190 }
3191
3192 if (tb[IFLA_OPERSTATE])
3193 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3194
3195 if (tb[IFLA_LINKMODE]) {
3196 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
3197
3198 if (dev->link_mode ^ value)
3199 status |= DO_SETLINK_NOTIFY;
3200 WRITE_ONCE(dev->link_mode, value);
3201 }
3202
3203 if (tb[IFLA_VFINFO_LIST]) {
3204 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
3205 struct nlattr *attr;
3206 int rem;
3207
3208 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
3209 if (nla_type(attr) != IFLA_VF_INFO ||
3210 nla_len(attr) < NLA_HDRLEN) {
3211 err = -EINVAL;
3212 goto errout;
3213 }
3214 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
3215 attr,
3216 ifla_vf_policy,
3217 NULL);
3218 if (err < 0)
3219 goto errout;
3220 err = do_setvfinfo(dev, vfinfo);
3221 if (err < 0)
3222 goto errout;
3223 status |= DO_SETLINK_NOTIFY;
3224 }
3225 }
3226 err = 0;
3227
3228 if (tb[IFLA_VF_PORTS]) {
3229 struct nlattr *port[IFLA_PORT_MAX+1];
3230 struct nlattr *attr;
3231 int vf;
3232 int rem;
3233
3234 err = -EOPNOTSUPP;
3235 if (!ops->ndo_set_vf_port)
3236 goto errout;
3237
3238 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
3239 if (nla_type(attr) != IFLA_VF_PORT ||
3240 nla_len(attr) < NLA_HDRLEN) {
3241 err = -EINVAL;
3242 goto errout;
3243 }
3244 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3245 attr,
3246 ifla_port_policy,
3247 NULL);
3248 if (err < 0)
3249 goto errout;
3250 if (!port[IFLA_PORT_VF]) {
3251 err = -EOPNOTSUPP;
3252 goto errout;
3253 }
3254 vf = nla_get_u32(port[IFLA_PORT_VF]);
3255 err = ops->ndo_set_vf_port(dev, vf, port);
3256 if (err < 0)
3257 goto errout;
3258 status |= DO_SETLINK_NOTIFY;
3259 }
3260 }
3261 err = 0;
3262
3263 if (tb[IFLA_PORT_SELF]) {
3264 struct nlattr *port[IFLA_PORT_MAX+1];
3265
3266 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3267 tb[IFLA_PORT_SELF],
3268 ifla_port_policy, NULL);
3269 if (err < 0)
3270 goto errout;
3271
3272 err = -EOPNOTSUPP;
3273 if (ops->ndo_set_vf_port)
3274 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3275 if (err < 0)
3276 goto errout;
3277 status |= DO_SETLINK_NOTIFY;
3278 }
3279
3280 if (tb[IFLA_AF_SPEC]) {
3281 struct nlattr *af;
3282 int rem;
3283
3284 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3285 struct rtnl_af_ops *af_ops;
3286 int af_ops_srcu_index;
3287
3288 af_ops = rtnl_af_lookup(nla_type(af), &af_ops_srcu_index);
3289 if (!af_ops) {
3290 err = -EAFNOSUPPORT;
3291 goto errout;
3292 }
3293
3294 err = af_ops->set_link_af(dev, af, extack);
3295 rtnl_af_put(af_ops, af_ops_srcu_index);
3296
3297 if (err < 0)
3298 goto errout;
3299
3300 status |= DO_SETLINK_NOTIFY;
3301 }
3302 }
3303 err = 0;
3304
3305 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3306 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3307 tb[IFLA_PROTO_DOWN_REASON], extack);
3308 if (err)
3309 goto errout;
3310 status |= DO_SETLINK_NOTIFY;
3311 }
3312
3313 if (tb[IFLA_XDP]) {
3314 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3315 u32 xdp_flags = 0;
3316
3317 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3318 tb[IFLA_XDP],
3319 ifla_xdp_policy, NULL);
3320 if (err < 0)
3321 goto errout;
3322
3323 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3324 err = -EINVAL;
3325 goto errout;
3326 }
3327
3328 if (xdp[IFLA_XDP_FLAGS]) {
3329 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3330 if (xdp_flags & ~XDP_FLAGS_MASK) {
3331 err = -EINVAL;
3332 goto errout;
3333 }
3334 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3335 err = -EINVAL;
3336 goto errout;
3337 }
3338 }
3339
3340 if (xdp[IFLA_XDP_FD]) {
3341 int expected_fd = -1;
3342
3343 if (xdp_flags & XDP_FLAGS_REPLACE) {
3344 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3345 err = -EINVAL;
3346 goto errout;
3347 }
3348 expected_fd =
3349 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3350 }
3351
3352 err = dev_change_xdp_fd(dev, extack,
3353 nla_get_s32(xdp[IFLA_XDP_FD]),
3354 expected_fd,
3355 xdp_flags);
3356 if (err)
3357 goto errout;
3358 status |= DO_SETLINK_NOTIFY;
3359 }
3360 }
3361
3362 errout:
3363 if (status & DO_SETLINK_MODIFIED) {
3364 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3365 netdev_state_change(dev);
3366
3367 if (err < 0)
3368 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",
3369 dev->name);
3370 }
3371
3372 return err;
3373 }
3374
3375 static struct net_device *rtnl_dev_get(struct net *net,
3376 struct nlattr *tb[])
3377 {
3378 char ifname[ALTIFNAMSIZ];
3379
3380 if (tb[IFLA_IFNAME])
3381 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3382 else if (tb[IFLA_ALT_IFNAME])
3383 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3384 else
3385 return NULL;
3386
3387 return __dev_get_by_name(net, ifname);
3388 }
3389
3390 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3391 struct netlink_ext_ack *extack)
3392 {
3393 struct ifinfomsg *ifm = nlmsg_data(nlh);
3394 struct net *net = sock_net(skb->sk);
3395 struct nlattr *tb[IFLA_MAX+1];
3396 struct net_device *dev = NULL;
3397 struct rtnl_nets rtnl_nets;
3398 struct net *tgt_net;
3399 int err;
3400
3401 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3402 ifla_policy, extack);
3403 if (err < 0)
3404 goto errout;
3405
3406 err = rtnl_ensure_unique_netns(tb, extack, false);
3407 if (err < 0)
3408 goto errout;
3409
3410 tgt_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3411 if (IS_ERR(tgt_net)) {
3412 err = PTR_ERR(tgt_net);
3413 goto errout;
3414 }
3415
3416 rtnl_nets_init(&rtnl_nets);
3417 rtnl_nets_add(&rtnl_nets, get_net(net));
3418 rtnl_nets_add(&rtnl_nets, tgt_net);
3419
3420 rtnl_nets_lock(&rtnl_nets);
3421
3422 if (ifm->ifi_index > 0)
3423 dev = __dev_get_by_index(net, ifm->ifi_index);
3424 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3425 dev = rtnl_dev_get(net, tb);
3426 else
3427 err = -EINVAL;
3428
3429 if (dev)
3430 err = do_setlink(skb, dev, tgt_net, ifm, extack, tb, 0);
3431 else if (!err)
3432 err = -ENODEV;
3433
3434 rtnl_nets_unlock(&rtnl_nets);
3435 errout:
3436 return err;
3437 }
3438
3439 static int rtnl_group_dellink(const struct net *net, int group)
3440 {
3441 struct net_device *dev, *aux;
3442 LIST_HEAD(list_kill);
3443 bool found = false;
3444
3445 if (!group)
3446 return -EPERM;
3447
3448 for_each_netdev(net, dev) {
3449 if (dev->group == group) {
3450 const struct rtnl_link_ops *ops;
3451
3452 found = true;
3453 ops = dev->rtnl_link_ops;
3454 if (!ops || !ops->dellink)
3455 return -EOPNOTSUPP;
3456 }
3457 }
3458
3459 if (!found)
3460 return -ENODEV;
3461
3462 for_each_netdev_safe(net, dev, aux) {
3463 if (dev->group == group) {
3464 const struct rtnl_link_ops *ops;
3465
3466 ops = dev->rtnl_link_ops;
3467 ops->dellink(dev, &list_kill);
3468 }
3469 }
3470 unregister_netdevice_many(&list_kill);
3471
3472 return 0;
3473 }
3474
3475 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3476 {
3477 const struct rtnl_link_ops *ops;
3478 LIST_HEAD(list_kill);
3479
3480 ops = dev->rtnl_link_ops;
3481 if (!ops || !ops->dellink)
3482 return -EOPNOTSUPP;
3483
3484 ops->dellink(dev, &list_kill);
3485 unregister_netdevice_many_notify(&list_kill, portid, nlh);
3486
3487 return 0;
3488 }
3489 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3490
3491 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3492 struct netlink_ext_ack *extack)
3493 {
3494 struct ifinfomsg *ifm = nlmsg_data(nlh);
3495 struct net *net = sock_net(skb->sk);
3496 u32 portid = NETLINK_CB(skb).portid;
3497 struct nlattr *tb[IFLA_MAX+1];
3498 struct net_device *dev = NULL;
3499 struct net *tgt_net = net;
3500 int netnsid = -1;
3501 int err;
3502
3503 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3504 ifla_policy, extack);
3505 if (err < 0)
3506 return err;
3507
3508 err = rtnl_ensure_unique_netns(tb, extack, true);
3509 if (err < 0)
3510 return err;
3511
3512 if (tb[IFLA_TARGET_NETNSID]) {
3513 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3514 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3515 if (IS_ERR(tgt_net))
3516 return PTR_ERR(tgt_net);
3517 }
3518
3519 rtnl_net_lock(tgt_net);
3520
3521 if (ifm->ifi_index > 0)
3522 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3523 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3524 dev = rtnl_dev_get(tgt_net, tb);
3525
3526 if (dev)
3527 err = rtnl_delete_link(dev, portid, nlh);
3528 else if (ifm->ifi_index > 0 || tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3529 err = -ENODEV;
3530 else if (tb[IFLA_GROUP])
3531 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3532 else
3533 err = -EINVAL;
3534
3535 rtnl_net_unlock(tgt_net);
3536
3537 if (netnsid >= 0)
3538 put_net(tgt_net);
3539
3540 return err;
3541 }
3542
3543 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3544 u32 portid, const struct nlmsghdr *nlh)
3545 {
3546 unsigned int old_flags;
3547 int err;
3548
3549 old_flags = dev->flags;
3550 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3551 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3552 NULL);
3553 if (err < 0)
3554 return err;
3555 }
3556
3557 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3558 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3559 } else {
3560 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3561 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3562 }
3563 return 0;
3564 }
3565 EXPORT_SYMBOL(rtnl_configure_link);
3566
3567 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3568 unsigned char name_assign_type,
3569 const struct rtnl_link_ops *ops,
3570 struct nlattr *tb[],
3571 struct netlink_ext_ack *extack)
3572 {
3573 struct net_device *dev;
3574 unsigned int num_tx_queues = 1;
3575 unsigned int num_rx_queues = 1;
3576 int err;
3577
3578 if (tb[IFLA_NUM_TX_QUEUES])
3579 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3580 else if (ops->get_num_tx_queues)
3581 num_tx_queues = ops->get_num_tx_queues();
3582
3583 if (tb[IFLA_NUM_RX_QUEUES])
3584 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3585 else if (ops->get_num_rx_queues)
3586 num_rx_queues = ops->get_num_rx_queues();
3587
3588 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3589 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3590 return ERR_PTR(-EINVAL);
3591 }
3592
3593 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3594 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3595 return ERR_PTR(-EINVAL);
3596 }
3597
3598 if (ops->alloc) {
3599 dev = ops->alloc(tb, ifname, name_assign_type,
3600 num_tx_queues, num_rx_queues);
3601 if (IS_ERR(dev))
3602 return dev;
3603 } else {
3604 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3605 name_assign_type, ops->setup,
3606 num_tx_queues, num_rx_queues);
3607 }
3608
3609 if (!dev)
3610 return ERR_PTR(-ENOMEM);
3611
3612 err = validate_linkmsg(dev, tb, extack);
3613 if (err < 0) {
3614 free_netdev(dev);
3615 return ERR_PTR(err);
3616 }
3617
3618 dev_net_set(dev, net);
3619 dev->rtnl_link_ops = ops;
3620 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3621
3622 if (tb[IFLA_MTU]) {
3623 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3624
3625 err = dev_validate_mtu(dev, mtu, extack);
3626 if (err) {
3627 free_netdev(dev);
3628 return ERR_PTR(err);
3629 }
3630 dev->mtu = mtu;
3631 }
3632 if (tb[IFLA_ADDRESS]) {
3633 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3634 nla_len(tb[IFLA_ADDRESS]));
3635 dev->addr_assign_type = NET_ADDR_SET;
3636 }
3637 if (tb[IFLA_BROADCAST])
3638 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3639 nla_len(tb[IFLA_BROADCAST]));
3640 if (tb[IFLA_TXQLEN])
3641 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3642 if (tb[IFLA_OPERSTATE])
3643 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3644 if (tb[IFLA_LINKMODE])
3645 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3646 if (tb[IFLA_GROUP])
3647 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3648 if (tb[IFLA_GSO_MAX_SIZE])
3649 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3650 if (tb[IFLA_GSO_MAX_SEGS])
3651 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3652 if (tb[IFLA_GRO_MAX_SIZE])
3653 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3654 if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3655 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3656 if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3657 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3658
3659 return dev;
3660 }
3661 EXPORT_SYMBOL(rtnl_create_link);
3662
3663 struct rtnl_newlink_tbs {
3664 struct nlattr *tb[IFLA_MAX + 1];
3665 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3666 struct nlattr *attr[RTNL_MAX_TYPE + 1];
3667 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3668 };
3669
3670 static int rtnl_changelink(const struct sk_buff *skb, struct nlmsghdr *nlh,
3671 const struct rtnl_link_ops *ops,
3672 struct net_device *dev, struct net *tgt_net,
3673 struct rtnl_newlink_tbs *tbs,
3674 struct nlattr **data,
3675 struct netlink_ext_ack *extack)
3676 {
3677 struct nlattr ** const linkinfo = tbs->linkinfo;
3678 struct nlattr ** const tb = tbs->tb;
3679 int status = 0;
3680 int err;
3681
3682 if (nlh->nlmsg_flags & NLM_F_EXCL)
3683 return -EEXIST;
3684
3685 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3686 return -EOPNOTSUPP;
3687
3688 if (linkinfo[IFLA_INFO_DATA]) {
3689 if (!ops || ops != dev->rtnl_link_ops || !ops->changelink)
3690 return -EOPNOTSUPP;
3691
3692 err = ops->changelink(dev, tb, data, extack);
3693 if (err < 0)
3694 return err;
3695
3696 status |= DO_SETLINK_NOTIFY;
3697 }
3698
3699 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3700 const struct rtnl_link_ops *m_ops = NULL;
3701 struct nlattr **slave_data = NULL;
3702 struct net_device *master_dev;
3703
3704 master_dev = netdev_master_upper_dev_get(dev);
3705 if (master_dev)
3706 m_ops = master_dev->rtnl_link_ops;
3707
3708 if (!m_ops || !m_ops->slave_changelink)
3709 return -EOPNOTSUPP;
3710
3711 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3712 return -EINVAL;
3713
3714 if (m_ops->slave_maxtype) {
3715 err = nla_parse_nested_deprecated(tbs->slave_attr,
3716 m_ops->slave_maxtype,
3717 linkinfo[IFLA_INFO_SLAVE_DATA],
3718 m_ops->slave_policy, extack);
3719 if (err < 0)
3720 return err;
3721
3722 slave_data = tbs->slave_attr;
3723 }
3724
3725 err = m_ops->slave_changelink(master_dev, dev, tb, slave_data, extack);
3726 if (err < 0)
3727 return err;
3728
3729 status |= DO_SETLINK_NOTIFY;
3730 }
3731
3732 return do_setlink(skb, dev, tgt_net, nlmsg_data(nlh), extack, tb, status);
3733 }
3734
3735 static int rtnl_group_changelink(const struct sk_buff *skb,
3736 struct net *net, struct net *tgt_net,
3737 int group, struct ifinfomsg *ifm,
3738 struct netlink_ext_ack *extack,
3739 struct nlattr **tb)
3740 {
3741 struct net_device *dev, *aux;
3742 int err;
3743
3744 for_each_netdev_safe(net, dev, aux) {
3745 if (dev->group == group) {
3746 err = do_setlink(skb, dev, tgt_net, ifm, extack, tb, 0);
3747 if (err < 0)
3748 return err;
3749 }
3750 }
3751
3752 return 0;
3753 }
3754
3755 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3756 const struct rtnl_link_ops *ops,
3757 struct net *tgt_net, struct net *link_net,
3758 struct net *peer_net,
3759 const struct nlmsghdr *nlh,
3760 struct nlattr **tb, struct nlattr **data,
3761 struct netlink_ext_ack *extack)
3762 {
3763 unsigned char name_assign_type = NET_NAME_USER;
3764 struct net *net = sock_net(skb->sk);
3765 u32 portid = NETLINK_CB(skb).portid;
3766 struct net_device *dev;
3767 char ifname[IFNAMSIZ];
3768 int err;
3769
3770 if (!ops->alloc && !ops->setup)
3771 return -EOPNOTSUPP;
3772
3773 if (tb[IFLA_IFNAME]) {
3774 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3775 } else {
3776 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3777 name_assign_type = NET_NAME_ENUM;
3778 }
3779
3780 dev = rtnl_create_link(link_net ? : tgt_net, ifname,
3781 name_assign_type, ops, tb, extack);
3782 if (IS_ERR(dev)) {
3783 err = PTR_ERR(dev);
3784 goto out;
3785 }
3786
3787 dev->ifindex = ifm->ifi_index;
3788
3789 if (link_net)
3790 net = link_net;
3791 if (peer_net)
3792 net = peer_net;
3793
3794 if (ops->newlink)
3795 err = ops->newlink(net, dev, tb, data, extack);
3796 else
3797 err = register_netdevice(dev);
3798 if (err < 0) {
3799 free_netdev(dev);
3800 goto out;
3801 }
3802
3803 err = rtnl_configure_link(dev, ifm, portid, nlh);
3804 if (err < 0)
3805 goto out_unregister;
3806 if (link_net) {
3807 err = dev_change_net_namespace(dev, tgt_net, ifname);
3808 if (err < 0)
3809 goto out_unregister;
3810 }
3811 if (tb[IFLA_MASTER]) {
3812 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3813 if (err)
3814 goto out_unregister;
3815 }
3816 out:
3817 return err;
3818 out_unregister:
3819 if (ops->newlink) {
3820 LIST_HEAD(list_kill);
3821
3822 ops->dellink(dev, &list_kill);
3823 unregister_netdevice_many(&list_kill);
3824 } else {
3825 unregister_netdevice(dev);
3826 }
3827 goto out;
3828 }
3829
3830 static struct net *rtnl_get_peer_net(const struct rtnl_link_ops *ops,
3831 struct nlattr *tbp[],
3832 struct nlattr *data[],
3833 struct netlink_ext_ack *extack)
3834 {
3835 struct nlattr *tb[IFLA_MAX + 1];
3836 int err;
3837
3838 if (!data || !data[ops->peer_type])
3839 return rtnl_link_get_net_ifla(tbp);
3840
3841 err = rtnl_nla_parse_ifinfomsg(tb, data[ops->peer_type], extack);
3842 if (err < 0)
3843 return ERR_PTR(err);
3844
3845 if (ops->validate) {
3846 err = ops->validate(tb, NULL, extack);
3847 if (err < 0)
3848 return ERR_PTR(err);
3849 }
3850
3851 return rtnl_link_get_net_ifla(tb);
3852 }
3853
3854 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3855 const struct rtnl_link_ops *ops,
3856 struct net *tgt_net, struct net *link_net,
3857 struct net *peer_net,
3858 struct rtnl_newlink_tbs *tbs,
3859 struct nlattr **data,
3860 struct netlink_ext_ack *extack)
3861 {
3862 struct nlattr ** const tb = tbs->tb;
3863 struct net *net = sock_net(skb->sk);
3864 struct net_device *dev;
3865 struct ifinfomsg *ifm;
3866 bool link_specified;
3867
3868 ifm = nlmsg_data(nlh);
3869 if (ifm->ifi_index > 0) {
3870 link_specified = true;
3871 dev = __dev_get_by_index(net, ifm->ifi_index);
3872 } else if (ifm->ifi_index < 0) {
3873 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3874 return -EINVAL;
3875 } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3876 link_specified = true;
3877 dev = rtnl_dev_get(net, tb);
3878 } else {
3879 link_specified = false;
3880 dev = NULL;
3881 }
3882
3883 if (dev)
3884 return rtnl_changelink(skb, nlh, ops, dev, tgt_net, tbs, data, extack);
3885
3886 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3887 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3888 * or it's for a group
3889 */
3890 if (link_specified || !tb[IFLA_GROUP])
3891 return -ENODEV;
3892
3893 return rtnl_group_changelink(skb, net, tgt_net,
3894 nla_get_u32(tb[IFLA_GROUP]),
3895 ifm, extack, tb);
3896 }
3897
3898 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3899 return -EOPNOTSUPP;
3900
3901 if (!ops) {
3902 NL_SET_ERR_MSG(extack, "Unknown device type");
3903 return -EOPNOTSUPP;
3904 }
3905
3906 return rtnl_newlink_create(skb, ifm, ops, tgt_net, link_net, peer_net, nlh,
3907 tb, data, extack);
3908 }
3909
3910 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3911 struct netlink_ext_ack *extack)
3912 {
3913 struct net *tgt_net, *link_net = NULL, *peer_net = NULL;
3914 struct nlattr **tb, **linkinfo, **data = NULL;
3915 struct rtnl_link_ops *ops = NULL;
3916 struct rtnl_newlink_tbs *tbs;
3917 struct rtnl_nets rtnl_nets;
3918 int ops_srcu_index;
3919 int ret;
3920
3921 tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3922 if (!tbs)
3923 return -ENOMEM;
3924
3925 tb = tbs->tb;
3926 ret = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg), tb,
3927 IFLA_MAX, ifla_policy, extack);
3928 if (ret < 0)
3929 goto free;
3930
3931 ret = rtnl_ensure_unique_netns(tb, extack, false);
3932 if (ret < 0)
3933 goto free;
3934
3935 linkinfo = tbs->linkinfo;
3936 if (tb[IFLA_LINKINFO]) {
3937 ret = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3938 tb[IFLA_LINKINFO],
3939 ifla_info_policy, NULL);
3940 if (ret < 0)
3941 goto free;
3942 } else {
3943 memset(linkinfo, 0, sizeof(tbs->linkinfo));
3944 }
3945
3946 if (linkinfo[IFLA_INFO_KIND]) {
3947 char kind[MODULE_NAME_LEN];
3948
3949 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3950 ops = rtnl_link_ops_get(kind, &ops_srcu_index);
3951 #ifdef CONFIG_MODULES
3952 if (!ops) {
3953 request_module("rtnl-link-%s", kind);
3954 ops = rtnl_link_ops_get(kind, &ops_srcu_index);
3955 }
3956 #endif
3957 }
3958
3959 rtnl_nets_init(&rtnl_nets);
3960
3961 if (ops) {
3962 if (ops->maxtype > RTNL_MAX_TYPE) {
3963 ret = -EINVAL;
3964 goto put_ops;
3965 }
3966
3967 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3968 ret = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3969 linkinfo[IFLA_INFO_DATA],
3970 ops->policy, extack);
3971 if (ret < 0)
3972 goto put_ops;
3973
3974 data = tbs->attr;
3975 }
3976
3977 if (ops->validate) {
3978 ret = ops->validate(tb, data, extack);
3979 if (ret < 0)
3980 goto put_ops;
3981 }
3982
3983 if (ops->peer_type) {
3984 peer_net = rtnl_get_peer_net(ops, tb, data, extack);
3985 if (IS_ERR(peer_net)) {
3986 ret = PTR_ERR(peer_net);
3987 goto put_ops;
3988 }
3989 if (peer_net)
3990 rtnl_nets_add(&rtnl_nets, peer_net);
3991 }
3992 }
3993
3994 tgt_net = rtnl_link_get_net_capable(skb, sock_net(skb->sk), tb, CAP_NET_ADMIN);
3995 if (IS_ERR(tgt_net)) {
3996 ret = PTR_ERR(tgt_net);
3997 goto put_net;
3998 }
3999
4000 rtnl_nets_add(&rtnl_nets, tgt_net);
4001
4002 if (tb[IFLA_LINK_NETNSID]) {
4003 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
4004
4005 link_net = get_net_ns_by_id(tgt_net, id);
4006 if (!link_net) {
4007 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
4008 ret = -EINVAL;
4009 goto put_net;
4010 }
4011
4012 rtnl_nets_add(&rtnl_nets, link_net);
4013
4014 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) {
4015 ret = -EPERM;
4016 goto put_net;
4017 }
4018 }
4019
4020 rtnl_nets_lock(&rtnl_nets);
4021 ret = __rtnl_newlink(skb, nlh, ops, tgt_net, link_net, peer_net, tbs, data, extack);
4022 rtnl_nets_unlock(&rtnl_nets);
4023
4024 put_net:
4025 rtnl_nets_destroy(&rtnl_nets);
4026 put_ops:
4027 if (ops)
4028 rtnl_link_ops_put(ops, ops_srcu_index);
4029 free:
4030 kfree(tbs);
4031 return ret;
4032 }
4033
4034 static int rtnl_valid_getlink_req(struct sk_buff *skb,
4035 const struct nlmsghdr *nlh,
4036 struct nlattr **tb,
4037 struct netlink_ext_ack *extack)
4038 {
4039 struct ifinfomsg *ifm;
4040 int i, err;
4041
4042 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4043 NL_SET_ERR_MSG(extack, "Invalid header for get link");
4044 return -EINVAL;
4045 }
4046
4047 if (!netlink_strict_get_check(skb))
4048 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
4049 ifla_policy, extack);
4050
4051 ifm = nlmsg_data(nlh);
4052 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4053 ifm->ifi_change) {
4054 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
4055 return -EINVAL;
4056 }
4057
4058 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
4059 ifla_policy, extack);
4060 if (err)
4061 return err;
4062
4063 for (i = 0; i <= IFLA_MAX; i++) {
4064 if (!tb[i])
4065 continue;
4066
4067 switch (i) {
4068 case IFLA_IFNAME:
4069 case IFLA_ALT_IFNAME:
4070 case IFLA_EXT_MASK:
4071 case IFLA_TARGET_NETNSID:
4072 break;
4073 default:
4074 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
4075 return -EINVAL;
4076 }
4077 }
4078
4079 return 0;
4080 }
4081
4082 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4083 struct netlink_ext_ack *extack)
4084 {
4085 struct net *net = sock_net(skb->sk);
4086 struct net *tgt_net = net;
4087 struct ifinfomsg *ifm;
4088 struct nlattr *tb[IFLA_MAX+1];
4089 struct net_device *dev = NULL;
4090 struct sk_buff *nskb;
4091 int netnsid = -1;
4092 int err;
4093 u32 ext_filter_mask = 0;
4094
4095 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
4096 if (err < 0)
4097 return err;
4098
4099 err = rtnl_ensure_unique_netns(tb, extack, true);
4100 if (err < 0)
4101 return err;
4102
4103 if (tb[IFLA_TARGET_NETNSID]) {
4104 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
4105 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
4106 if (IS_ERR(tgt_net))
4107 return PTR_ERR(tgt_net);
4108 }
4109
4110 if (tb[IFLA_EXT_MASK])
4111 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
4112
4113 err = -EINVAL;
4114 ifm = nlmsg_data(nlh);
4115 if (ifm->ifi_index > 0)
4116 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
4117 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
4118 dev = rtnl_dev_get(tgt_net, tb);
4119 else
4120 goto out;
4121
4122 err = -ENODEV;
4123 if (dev == NULL)
4124 goto out;
4125
4126 err = -ENOBUFS;
4127 nskb = nlmsg_new_large(if_nlmsg_size(dev, ext_filter_mask));
4128 if (nskb == NULL)
4129 goto out;
4130
4131 /* Synchronize the carrier state so we don't report a state
4132 * that we're not actually going to honour immediately; if
4133 * the driver just did a carrier off->on transition, we can
4134 * only TX if link watch work has run, but without this we'd
4135 * already report carrier on, even if it doesn't work yet.
4136 */
4137 linkwatch_sync_dev(dev);
4138
4139 err = rtnl_fill_ifinfo(nskb, dev, net,
4140 RTM_NEWLINK, NETLINK_CB(skb).portid,
4141 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
4142 0, NULL, 0, netnsid, GFP_KERNEL);
4143 if (err < 0) {
4144 /* -EMSGSIZE implies BUG in if_nlmsg_size */
4145 WARN_ON(err == -EMSGSIZE);
4146 kfree_skb(nskb);
4147 } else
4148 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4149 out:
4150 if (netnsid >= 0)
4151 put_net(tgt_net);
4152
4153 return err;
4154 }
4155
4156 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
4157 bool *changed, struct netlink_ext_ack *extack)
4158 {
4159 char *alt_ifname;
4160 size_t size;
4161 int err;
4162
4163 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
4164 if (err)
4165 return err;
4166
4167 if (cmd == RTM_NEWLINKPROP) {
4168 size = rtnl_prop_list_size(dev);
4169 size += nla_total_size(ALTIFNAMSIZ);
4170 if (size >= U16_MAX) {
4171 NL_SET_ERR_MSG(extack,
4172 "effective property list too long");
4173 return -EINVAL;
4174 }
4175 }
4176
4177 alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
4178 if (!alt_ifname)
4179 return -ENOMEM;
4180
4181 if (cmd == RTM_NEWLINKPROP) {
4182 err = netdev_name_node_alt_create(dev, alt_ifname);
4183 if (!err)
4184 alt_ifname = NULL;
4185 } else if (cmd == RTM_DELLINKPROP) {
4186 err = netdev_name_node_alt_destroy(dev, alt_ifname);
4187 } else {
4188 WARN_ON_ONCE(1);
4189 err = -EINVAL;
4190 }
4191
4192 kfree(alt_ifname);
4193 if (!err)
4194 *changed = true;
4195 return err;
4196 }
4197
4198 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
4199 struct netlink_ext_ack *extack)
4200 {
4201 struct net *net = sock_net(skb->sk);
4202 struct nlattr *tb[IFLA_MAX + 1];
4203 struct net_device *dev;
4204 struct ifinfomsg *ifm;
4205 bool changed = false;
4206 struct nlattr *attr;
4207 int err, rem;
4208
4209 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
4210 if (err)
4211 return err;
4212
4213 err = rtnl_ensure_unique_netns(tb, extack, true);
4214 if (err)
4215 return err;
4216
4217 ifm = nlmsg_data(nlh);
4218 if (ifm->ifi_index > 0)
4219 dev = __dev_get_by_index(net, ifm->ifi_index);
4220 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
4221 dev = rtnl_dev_get(net, tb);
4222 else
4223 return -EINVAL;
4224
4225 if (!dev)
4226 return -ENODEV;
4227
4228 if (!tb[IFLA_PROP_LIST])
4229 return 0;
4230
4231 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
4232 switch (nla_type(attr)) {
4233 case IFLA_ALT_IFNAME:
4234 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
4235 if (err)
4236 return err;
4237 break;
4238 }
4239 }
4240
4241 if (changed)
4242 netdev_state_change(dev);
4243 return 0;
4244 }
4245
4246 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
4247 struct netlink_ext_ack *extack)
4248 {
4249 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
4250 }
4251
4252 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
4253 struct netlink_ext_ack *extack)
4254 {
4255 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
4256 }
4257
4258 static noinline_for_stack u32 rtnl_calcit(struct sk_buff *skb,
4259 struct nlmsghdr *nlh)
4260 {
4261 struct net *net = sock_net(skb->sk);
4262 size_t min_ifinfo_dump_size = 0;
4263 u32 ext_filter_mask = 0;
4264 struct net_device *dev;
4265 struct nlattr *nla;
4266 int hdrlen, rem;
4267
4268 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
4269 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
4270 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
4271
4272 if (nlh->nlmsg_len < nlmsg_msg_size(hdrlen))
4273 return NLMSG_GOODSIZE;
4274
4275 nla_for_each_attr_type(nla, IFLA_EXT_MASK,
4276 nlmsg_attrdata(nlh, hdrlen),
4277 nlmsg_attrlen(nlh, hdrlen), rem) {
4278 if (nla_len(nla) == sizeof(u32))
4279 ext_filter_mask = nla_get_u32(nla);
4280 }
4281
4282 if (!ext_filter_mask)
4283 return NLMSG_GOODSIZE;
4284 /*
4285 * traverse the list of net devices and compute the minimum
4286 * buffer size based upon the filter mask.
4287 */
4288 rcu_read_lock();
4289 for_each_netdev_rcu(net, dev) {
4290 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
4291 if_nlmsg_size(dev, ext_filter_mask));
4292 }
4293 rcu_read_unlock();
4294
4295 return nlmsg_total_size(min_ifinfo_dump_size);
4296 }
4297
4298 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
4299 {
4300 int idx;
4301 int s_idx = cb->family;
4302 int type = cb->nlh->nlmsg_type - RTM_BASE;
4303 int ret = 0;
4304
4305 if (s_idx == 0)
4306 s_idx = 1;
4307
4308 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
4309 struct rtnl_link __rcu **tab;
4310 struct rtnl_link *link;
4311 rtnl_dumpit_func dumpit;
4312
4313 if (idx < s_idx || idx == PF_PACKET)
4314 continue;
4315
4316 if (type < 0 || type >= RTM_NR_MSGTYPES)
4317 continue;
4318
4319 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
4320 if (!tab)
4321 continue;
4322
4323 link = rcu_dereference_rtnl(tab[type]);
4324 if (!link)
4325 continue;
4326
4327 dumpit = link->dumpit;
4328 if (!dumpit)
4329 continue;
4330
4331 if (idx > s_idx) {
4332 memset(&cb->args[0], 0, sizeof(cb->args));
4333 cb->prev_seq = 0;
4334 cb->seq = 0;
4335 }
4336 ret = dumpit(skb, cb);
4337 if (ret)
4338 break;
4339 }
4340 cb->family = idx;
4341
4342 return skb->len ? : ret;
4343 }
4344
4345 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
4346 unsigned int change,
4347 u32 event, gfp_t flags, int *new_nsid,
4348 int new_ifindex, u32 portid,
4349 const struct nlmsghdr *nlh)
4350 {
4351 struct net *net = dev_net(dev);
4352 struct sk_buff *skb;
4353 int err = -ENOBUFS;
4354 u32 seq = 0;
4355
4356 skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4357 if (skb == NULL)
4358 goto errout;
4359
4360 if (nlmsg_report(nlh))
4361 seq = nlmsg_seq(nlh);
4362 else
4363 portid = 0;
4364
4365 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4366 type, portid, seq, change, 0, 0, event,
4367 new_nsid, new_ifindex, -1, flags);
4368 if (err < 0) {
4369 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4370 WARN_ON(err == -EMSGSIZE);
4371 kfree_skb(skb);
4372 goto errout;
4373 }
4374 return skb;
4375 errout:
4376 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4377 return NULL;
4378 }
4379
4380 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4381 u32 portid, const struct nlmsghdr *nlh)
4382 {
4383 struct net *net = dev_net(dev);
4384
4385 rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4386 }
4387
4388 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4389 unsigned int change, u32 event,
4390 gfp_t flags, int *new_nsid, int new_ifindex,
4391 u32 portid, const struct nlmsghdr *nlh)
4392 {
4393 struct sk_buff *skb;
4394
4395 if (dev->reg_state != NETREG_REGISTERED)
4396 return;
4397
4398 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4399 new_ifindex, portid, nlh);
4400 if (skb)
4401 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4402 }
4403
4404 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4405 gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4406 {
4407 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4408 NULL, 0, portid, nlh);
4409 }
4410
4411 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4412 gfp_t flags, int *new_nsid, int new_ifindex)
4413 {
4414 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4415 new_nsid, new_ifindex, 0, NULL);
4416 }
4417
4418 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4419 struct net_device *dev,
4420 u8 *addr, u16 vid, u32 pid, u32 seq,
4421 int type, unsigned int flags,
4422 int nlflags, u16 ndm_state)
4423 {
4424 struct nlmsghdr *nlh;
4425 struct ndmsg *ndm;
4426
4427 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4428 if (!nlh)
4429 return -EMSGSIZE;
4430
4431 ndm = nlmsg_data(nlh);
4432 ndm->ndm_family = AF_BRIDGE;
4433 ndm->ndm_pad1 = 0;
4434 ndm->ndm_pad2 = 0;
4435 ndm->ndm_flags = flags;
4436 ndm->ndm_type = 0;
4437 ndm->ndm_ifindex = dev->ifindex;
4438 ndm->ndm_state = ndm_state;
4439
4440 if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
4441 goto nla_put_failure;
4442 if (vid)
4443 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4444 goto nla_put_failure;
4445
4446 nlmsg_end(skb, nlh);
4447 return 0;
4448
4449 nla_put_failure:
4450 nlmsg_cancel(skb, nlh);
4451 return -EMSGSIZE;
4452 }
4453
4454 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
4455 {
4456 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4457 nla_total_size(dev->addr_len) + /* NDA_LLADDR */
4458 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
4459 0;
4460 }
4461
4462 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4463 u16 ndm_state)
4464 {
4465 struct net *net = dev_net(dev);
4466 struct sk_buff *skb;
4467 int err = -ENOBUFS;
4468
4469 skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
4470 if (!skb)
4471 goto errout;
4472
4473 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4474 0, 0, type, NTF_SELF, 0, ndm_state);
4475 if (err < 0) {
4476 kfree_skb(skb);
4477 goto errout;
4478 }
4479
4480 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4481 return;
4482 errout:
4483 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4484 }
4485
4486 /*
4487 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4488 */
4489 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4490 struct nlattr *tb[],
4491 struct net_device *dev,
4492 const unsigned char *addr, u16 vid,
4493 u16 flags)
4494 {
4495 int err = -EINVAL;
4496
4497 /* If aging addresses are supported device will need to
4498 * implement its own handler for this.
4499 */
4500 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4501 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4502 return err;
4503 }
4504
4505 if (tb[NDA_FLAGS_EXT]) {
4506 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4507 return err;
4508 }
4509
4510 if (vid) {
4511 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4512 return err;
4513 }
4514
4515 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4516 err = dev_uc_add_excl(dev, addr);
4517 else if (is_multicast_ether_addr(addr))
4518 err = dev_mc_add_excl(dev, addr);
4519
4520 /* Only return duplicate errors if NLM_F_EXCL is set */
4521 if (err == -EEXIST && !(flags & NLM_F_EXCL))
4522 err = 0;
4523
4524 return err;
4525 }
4526 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4527
4528 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4529 struct netlink_ext_ack *extack)
4530 {
4531 u16 vid = 0;
4532
4533 if (vlan_attr) {
4534 if (nla_len(vlan_attr) != sizeof(u16)) {
4535 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4536 return -EINVAL;
4537 }
4538
4539 vid = nla_get_u16(vlan_attr);
4540
4541 if (!vid || vid >= VLAN_VID_MASK) {
4542 NL_SET_ERR_MSG(extack, "invalid vlan id");
4543 return -EINVAL;
4544 }
4545 }
4546 *p_vid = vid;
4547 return 0;
4548 }
4549
4550 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4551 struct netlink_ext_ack *extack)
4552 {
4553 struct net *net = sock_net(skb->sk);
4554 struct ndmsg *ndm;
4555 struct nlattr *tb[NDA_MAX+1];
4556 struct net_device *dev;
4557 u8 *addr;
4558 u16 vid;
4559 int err;
4560
4561 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4562 extack);
4563 if (err < 0)
4564 return err;
4565
4566 ndm = nlmsg_data(nlh);
4567 if (ndm->ndm_ifindex == 0) {
4568 NL_SET_ERR_MSG(extack, "invalid ifindex");
4569 return -EINVAL;
4570 }
4571
4572 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4573 if (dev == NULL) {
4574 NL_SET_ERR_MSG(extack, "unknown ifindex");
4575 return -ENODEV;
4576 }
4577
4578 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4579 NL_SET_ERR_MSG(extack, "invalid address");
4580 return -EINVAL;
4581 }
4582
4583 if (dev->type != ARPHRD_ETHER) {
4584 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4585 return -EINVAL;
4586 }
4587
4588 addr = nla_data(tb[NDA_LLADDR]);
4589
4590 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4591 if (err)
4592 return err;
4593
4594 err = -EOPNOTSUPP;
4595
4596 /* Support fdb on master device the net/bridge default case */
4597 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4598 netif_is_bridge_port(dev)) {
4599 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4600 const struct net_device_ops *ops = br_dev->netdev_ops;
4601 bool notified = false;
4602
4603 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4604 nlh->nlmsg_flags, &notified, extack);
4605 if (err)
4606 goto out;
4607 else
4608 ndm->ndm_flags &= ~NTF_MASTER;
4609 }
4610
4611 /* Embedded bridge, macvlan, and any other device support */
4612 if ((ndm->ndm_flags & NTF_SELF)) {
4613 bool notified = false;
4614
4615 if (dev->netdev_ops->ndo_fdb_add)
4616 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4617 vid,
4618 nlh->nlmsg_flags,
4619 &notified, extack);
4620 else
4621 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4622 nlh->nlmsg_flags);
4623
4624 if (!err && !notified) {
4625 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4626 ndm->ndm_state);
4627 ndm->ndm_flags &= ~NTF_SELF;
4628 }
4629 }
4630 out:
4631 return err;
4632 }
4633
4634 /*
4635 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4636 */
4637 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4638 struct nlattr *tb[],
4639 struct net_device *dev,
4640 const unsigned char *addr, u16 vid)
4641 {
4642 int err = -EINVAL;
4643
4644 /* If aging addresses are supported device will need to
4645 * implement its own handler for this.
4646 */
4647 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4648 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4649 return err;
4650 }
4651
4652 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4653 err = dev_uc_del(dev, addr);
4654 else if (is_multicast_ether_addr(addr))
4655 err = dev_mc_del(dev, addr);
4656
4657 return err;
4658 }
4659 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4660
4661 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4662 struct netlink_ext_ack *extack)
4663 {
4664 bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4665 struct net *net = sock_net(skb->sk);
4666 const struct net_device_ops *ops;
4667 struct ndmsg *ndm;
4668 struct nlattr *tb[NDA_MAX+1];
4669 struct net_device *dev;
4670 __u8 *addr = NULL;
4671 int err;
4672 u16 vid;
4673
4674 if (!netlink_capable(skb, CAP_NET_ADMIN))
4675 return -EPERM;
4676
4677 if (!del_bulk) {
4678 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4679 NULL, extack);
4680 } else {
4681 /* For bulk delete, the drivers will parse the message with
4682 * policy.
4683 */
4684 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
4685 }
4686 if (err < 0)
4687 return err;
4688
4689 ndm = nlmsg_data(nlh);
4690 if (ndm->ndm_ifindex == 0) {
4691 NL_SET_ERR_MSG(extack, "invalid ifindex");
4692 return -EINVAL;
4693 }
4694
4695 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4696 if (dev == NULL) {
4697 NL_SET_ERR_MSG(extack, "unknown ifindex");
4698 return -ENODEV;
4699 }
4700
4701 if (!del_bulk) {
4702 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4703 NL_SET_ERR_MSG(extack, "invalid address");
4704 return -EINVAL;
4705 }
4706 addr = nla_data(tb[NDA_LLADDR]);
4707
4708 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4709 if (err)
4710 return err;
4711 }
4712
4713 if (dev->type != ARPHRD_ETHER) {
4714 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4715 return -EINVAL;
4716 }
4717
4718 err = -EOPNOTSUPP;
4719
4720 /* Support fdb on master device the net/bridge default case */
4721 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4722 netif_is_bridge_port(dev)) {
4723 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4724 bool notified = false;
4725
4726 ops = br_dev->netdev_ops;
4727 if (!del_bulk) {
4728 if (ops->ndo_fdb_del)
4729 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid,
4730 &notified, extack);
4731 } else {
4732 if (ops->ndo_fdb_del_bulk)
4733 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4734 }
4735
4736 if (err)
4737 goto out;
4738 else
4739 ndm->ndm_flags &= ~NTF_MASTER;
4740 }
4741
4742 /* Embedded bridge, macvlan, and any other device support */
4743 if (ndm->ndm_flags & NTF_SELF) {
4744 bool notified = false;
4745
4746 ops = dev->netdev_ops;
4747 if (!del_bulk) {
4748 if (ops->ndo_fdb_del)
4749 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid,
4750 &notified, extack);
4751 else
4752 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4753 } else {
4754 /* in case err was cleared by NTF_MASTER call */
4755 err = -EOPNOTSUPP;
4756 if (ops->ndo_fdb_del_bulk)
4757 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4758 }
4759
4760 if (!err) {
4761 if (!del_bulk && !notified)
4762 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4763 ndm->ndm_state);
4764 ndm->ndm_flags &= ~NTF_SELF;
4765 }
4766 }
4767 out:
4768 return err;
4769 }
4770
4771 static int nlmsg_populate_fdb(struct sk_buff *skb,
4772 struct netlink_callback *cb,
4773 struct net_device *dev,
4774 int *idx,
4775 struct netdev_hw_addr_list *list)
4776 {
4777 struct ndo_fdb_dump_context *ctx = (void *)cb->ctx;
4778 struct netdev_hw_addr *ha;
4779 u32 portid, seq;
4780 int err;
4781
4782 portid = NETLINK_CB(cb->skb).portid;
4783 seq = cb->nlh->nlmsg_seq;
4784
4785 list_for_each_entry(ha, &list->list, list) {
4786 if (*idx < ctx->fdb_idx)
4787 goto skip;
4788
4789 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4790 portid, seq,
4791 RTM_NEWNEIGH, NTF_SELF,
4792 NLM_F_MULTI, NUD_PERMANENT);
4793 if (err < 0)
4794 return err;
4795 skip:
4796 *idx += 1;
4797 }
4798 return 0;
4799 }
4800
4801 /**
4802 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4803 * @skb: socket buffer to store message in
4804 * @cb: netlink callback
4805 * @dev: netdevice
4806 * @filter_dev: ignored
4807 * @idx: the number of FDB table entries dumped is added to *@idx
4808 *
4809 * Default netdevice operation to dump the existing unicast address list.
4810 * Returns number of addresses from list put in skb.
4811 */
4812 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4813 struct netlink_callback *cb,
4814 struct net_device *dev,
4815 struct net_device *filter_dev,
4816 int *idx)
4817 {
4818 int err;
4819
4820 if (dev->type != ARPHRD_ETHER)
4821 return -EINVAL;
4822
4823 netif_addr_lock_bh(dev);
4824 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4825 if (err)
4826 goto out;
4827 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4828 out:
4829 netif_addr_unlock_bh(dev);
4830 return err;
4831 }
4832 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4833
4834 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4835 int *br_idx, int *brport_idx,
4836 struct netlink_ext_ack *extack)
4837 {
4838 struct nlattr *tb[NDA_MAX + 1];
4839 struct ndmsg *ndm;
4840 int err, i;
4841
4842 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4843 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4844 return -EINVAL;
4845 }
4846
4847 ndm = nlmsg_data(nlh);
4848 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4849 ndm->ndm_flags || ndm->ndm_type) {
4850 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4851 return -EINVAL;
4852 }
4853
4854 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4855 NDA_MAX, NULL, extack);
4856 if (err < 0)
4857 return err;
4858
4859 *brport_idx = ndm->ndm_ifindex;
4860 for (i = 0; i <= NDA_MAX; ++i) {
4861 if (!tb[i])
4862 continue;
4863
4864 switch (i) {
4865 case NDA_IFINDEX:
4866 if (nla_len(tb[i]) != sizeof(u32)) {
4867 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4868 return -EINVAL;
4869 }
4870 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4871 break;
4872 case NDA_MASTER:
4873 if (nla_len(tb[i]) != sizeof(u32)) {
4874 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4875 return -EINVAL;
4876 }
4877 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4878 break;
4879 default:
4880 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4881 return -EINVAL;
4882 }
4883 }
4884
4885 return 0;
4886 }
4887
4888 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4889 int *br_idx, int *brport_idx,
4890 struct netlink_ext_ack *extack)
4891 {
4892 struct nlattr *tb[IFLA_MAX+1];
4893 int err;
4894
4895 /* A hack to preserve kernel<->userspace interface.
4896 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4897 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4898 * So, check for ndmsg with an optional u32 attribute (not used here).
4899 * Fortunately these sizes don't conflict with the size of ifinfomsg
4900 * with an optional attribute.
4901 */
4902 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4903 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4904 nla_attr_size(sizeof(u32)))) {
4905 struct ifinfomsg *ifm;
4906
4907 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4908 tb, IFLA_MAX, ifla_policy,
4909 extack);
4910 if (err < 0) {
4911 return -EINVAL;
4912 } else if (err == 0) {
4913 if (tb[IFLA_MASTER])
4914 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4915 }
4916
4917 ifm = nlmsg_data(nlh);
4918 *brport_idx = ifm->ifi_index;
4919 }
4920 return 0;
4921 }
4922
4923 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4924 {
4925 const struct net_device_ops *ops = NULL, *cops = NULL;
4926 struct ndo_fdb_dump_context *ctx = (void *)cb->ctx;
4927 struct net_device *dev, *br_dev = NULL;
4928 struct net *net = sock_net(skb->sk);
4929 int brport_idx = 0;
4930 int br_idx = 0;
4931 int fidx = 0;
4932 int err;
4933
4934 NL_ASSERT_CTX_FITS(struct ndo_fdb_dump_context);
4935
4936 if (cb->strict_check)
4937 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4938 cb->extack);
4939 else
4940 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4941 cb->extack);
4942 if (err < 0)
4943 return err;
4944
4945 if (br_idx) {
4946 br_dev = __dev_get_by_index(net, br_idx);
4947 if (!br_dev)
4948 return -ENODEV;
4949
4950 ops = br_dev->netdev_ops;
4951 }
4952
4953 for_each_netdev_dump(net, dev, ctx->ifindex) {
4954 if (brport_idx && (dev->ifindex != brport_idx))
4955 continue;
4956
4957 if (!br_idx) { /* user did not specify a specific bridge */
4958 if (netif_is_bridge_port(dev)) {
4959 br_dev = netdev_master_upper_dev_get(dev);
4960 cops = br_dev->netdev_ops;
4961 }
4962 } else {
4963 if (dev != br_dev &&
4964 !netif_is_bridge_port(dev))
4965 continue;
4966
4967 if (br_dev != netdev_master_upper_dev_get(dev) &&
4968 !netif_is_bridge_master(dev))
4969 continue;
4970 cops = ops;
4971 }
4972
4973 if (netif_is_bridge_port(dev)) {
4974 if (cops && cops->ndo_fdb_dump) {
4975 err = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
4976 &fidx);
4977 if (err == -EMSGSIZE)
4978 break;
4979 }
4980 }
4981
4982 if (dev->netdev_ops->ndo_fdb_dump)
4983 err = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
4984 &fidx);
4985 else
4986 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL, &fidx);
4987 if (err == -EMSGSIZE)
4988 break;
4989
4990 cops = NULL;
4991
4992 /* reset fdb offset to 0 for rest of the interfaces */
4993 ctx->fdb_idx = 0;
4994 fidx = 0;
4995 }
4996
4997 ctx->fdb_idx = fidx;
4998
4999 return skb->len;
5000 }
5001
5002 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
5003 struct nlattr **tb, u8 *ndm_flags,
5004 int *br_idx, int *brport_idx, u8 **addr,
5005 u16 *vid, struct netlink_ext_ack *extack)
5006 {
5007 struct ndmsg *ndm;
5008 int err, i;
5009
5010 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
5011 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
5012 return -EINVAL;
5013 }
5014
5015 ndm = nlmsg_data(nlh);
5016 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
5017 ndm->ndm_type) {
5018 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
5019 return -EINVAL;
5020 }
5021
5022 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
5023 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
5024 return -EINVAL;
5025 }
5026
5027 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
5028 NDA_MAX, nda_policy, extack);
5029 if (err < 0)
5030 return err;
5031
5032 *ndm_flags = ndm->ndm_flags;
5033 *brport_idx = ndm->ndm_ifindex;
5034 for (i = 0; i <= NDA_MAX; ++i) {
5035 if (!tb[i])
5036 continue;
5037
5038 switch (i) {
5039 case NDA_MASTER:
5040 *br_idx = nla_get_u32(tb[i]);
5041 break;
5042 case NDA_LLADDR:
5043 if (nla_len(tb[i]) != ETH_ALEN) {
5044 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
5045 return -EINVAL;
5046 }
5047 *addr = nla_data(tb[i]);
5048 break;
5049 case NDA_VLAN:
5050 err = fdb_vid_parse(tb[i], vid, extack);
5051 if (err)
5052 return err;
5053 break;
5054 case NDA_VNI:
5055 break;
5056 default:
5057 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
5058 return -EINVAL;
5059 }
5060 }
5061
5062 return 0;
5063 }
5064
5065 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5066 struct netlink_ext_ack *extack)
5067 {
5068 struct net_device *dev = NULL, *br_dev = NULL;
5069 const struct net_device_ops *ops = NULL;
5070 struct net *net = sock_net(in_skb->sk);
5071 struct nlattr *tb[NDA_MAX + 1];
5072 struct sk_buff *skb;
5073 int brport_idx = 0;
5074 u8 ndm_flags = 0;
5075 int br_idx = 0;
5076 u8 *addr = NULL;
5077 u16 vid = 0;
5078 int err;
5079
5080 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
5081 &brport_idx, &addr, &vid, extack);
5082 if (err < 0)
5083 return err;
5084
5085 if (!addr) {
5086 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
5087 return -EINVAL;
5088 }
5089
5090 if (brport_idx) {
5091 dev = __dev_get_by_index(net, brport_idx);
5092 if (!dev) {
5093 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
5094 return -ENODEV;
5095 }
5096 }
5097
5098 if (br_idx) {
5099 if (dev) {
5100 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
5101 return -EINVAL;
5102 }
5103
5104 br_dev = __dev_get_by_index(net, br_idx);
5105 if (!br_dev) {
5106 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
5107 return -EINVAL;
5108 }
5109 ops = br_dev->netdev_ops;
5110 }
5111
5112 if (dev) {
5113 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
5114 if (!netif_is_bridge_port(dev)) {
5115 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
5116 return -EINVAL;
5117 }
5118 br_dev = netdev_master_upper_dev_get(dev);
5119 if (!br_dev) {
5120 NL_SET_ERR_MSG(extack, "Master of device not found");
5121 return -EINVAL;
5122 }
5123 ops = br_dev->netdev_ops;
5124 } else {
5125 if (!(ndm_flags & NTF_SELF)) {
5126 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
5127 return -EINVAL;
5128 }
5129 ops = dev->netdev_ops;
5130 }
5131 }
5132
5133 if (!br_dev && !dev) {
5134 NL_SET_ERR_MSG(extack, "No device specified");
5135 return -ENODEV;
5136 }
5137
5138 if (!ops || !ops->ndo_fdb_get) {
5139 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
5140 return -EOPNOTSUPP;
5141 }
5142
5143 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
5144 if (!skb)
5145 return -ENOBUFS;
5146
5147 if (br_dev)
5148 dev = br_dev;
5149 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
5150 NETLINK_CB(in_skb).portid,
5151 nlh->nlmsg_seq, extack);
5152 if (err)
5153 goto out;
5154
5155 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5156 out:
5157 kfree_skb(skb);
5158 return err;
5159 }
5160
5161 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
5162 unsigned int attrnum, unsigned int flag)
5163 {
5164 if (mask & flag)
5165 return nla_put_u8(skb, attrnum, !!(flags & flag));
5166 return 0;
5167 }
5168
5169 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
5170 struct net_device *dev, u16 mode,
5171 u32 flags, u32 mask, int nlflags,
5172 u32 filter_mask,
5173 int (*vlan_fill)(struct sk_buff *skb,
5174 struct net_device *dev,
5175 u32 filter_mask))
5176 {
5177 struct nlmsghdr *nlh;
5178 struct ifinfomsg *ifm;
5179 struct nlattr *br_afspec;
5180 struct nlattr *protinfo;
5181 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
5182 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5183 int err = 0;
5184
5185 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
5186 if (nlh == NULL)
5187 return -EMSGSIZE;
5188
5189 ifm = nlmsg_data(nlh);
5190 ifm->ifi_family = AF_BRIDGE;
5191 ifm->__ifi_pad = 0;
5192 ifm->ifi_type = dev->type;
5193 ifm->ifi_index = dev->ifindex;
5194 ifm->ifi_flags = dev_get_flags(dev);
5195 ifm->ifi_change = 0;
5196
5197
5198 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5199 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5200 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
5201 (br_dev &&
5202 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
5203 (dev->addr_len &&
5204 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5205 (dev->ifindex != dev_get_iflink(dev) &&
5206 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
5207 goto nla_put_failure;
5208
5209 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
5210 if (!br_afspec)
5211 goto nla_put_failure;
5212
5213 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
5214 nla_nest_cancel(skb, br_afspec);
5215 goto nla_put_failure;
5216 }
5217
5218 if (mode != BRIDGE_MODE_UNDEF) {
5219 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
5220 nla_nest_cancel(skb, br_afspec);
5221 goto nla_put_failure;
5222 }
5223 }
5224 if (vlan_fill) {
5225 err = vlan_fill(skb, dev, filter_mask);
5226 if (err) {
5227 nla_nest_cancel(skb, br_afspec);
5228 goto nla_put_failure;
5229 }
5230 }
5231 nla_nest_end(skb, br_afspec);
5232
5233 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
5234 if (!protinfo)
5235 goto nla_put_failure;
5236
5237 if (brport_nla_put_flag(skb, flags, mask,
5238 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
5239 brport_nla_put_flag(skb, flags, mask,
5240 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
5241 brport_nla_put_flag(skb, flags, mask,
5242 IFLA_BRPORT_FAST_LEAVE,
5243 BR_MULTICAST_FAST_LEAVE) ||
5244 brport_nla_put_flag(skb, flags, mask,
5245 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
5246 brport_nla_put_flag(skb, flags, mask,
5247 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
5248 brport_nla_put_flag(skb, flags, mask,
5249 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
5250 brport_nla_put_flag(skb, flags, mask,
5251 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
5252 brport_nla_put_flag(skb, flags, mask,
5253 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
5254 brport_nla_put_flag(skb, flags, mask,
5255 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
5256 brport_nla_put_flag(skb, flags, mask,
5257 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
5258 nla_nest_cancel(skb, protinfo);
5259 goto nla_put_failure;
5260 }
5261
5262 nla_nest_end(skb, protinfo);
5263
5264 nlmsg_end(skb, nlh);
5265 return 0;
5266 nla_put_failure:
5267 nlmsg_cancel(skb, nlh);
5268 return err ? err : -EMSGSIZE;
5269 }
5270 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
5271
5272 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
5273 bool strict_check, u32 *filter_mask,
5274 struct netlink_ext_ack *extack)
5275 {
5276 struct nlattr *tb[IFLA_MAX+1];
5277 int err, i;
5278
5279 if (strict_check) {
5280 struct ifinfomsg *ifm;
5281
5282 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5283 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
5284 return -EINVAL;
5285 }
5286
5287 ifm = nlmsg_data(nlh);
5288 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5289 ifm->ifi_change || ifm->ifi_index) {
5290 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
5291 return -EINVAL;
5292 }
5293
5294 err = nlmsg_parse_deprecated_strict(nlh,
5295 sizeof(struct ifinfomsg),
5296 tb, IFLA_MAX, ifla_policy,
5297 extack);
5298 } else {
5299 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
5300 tb, IFLA_MAX, ifla_policy,
5301 extack);
5302 }
5303 if (err < 0)
5304 return err;
5305
5306 /* new attributes should only be added with strict checking */
5307 for (i = 0; i <= IFLA_MAX; ++i) {
5308 if (!tb[i])
5309 continue;
5310
5311 switch (i) {
5312 case IFLA_EXT_MASK:
5313 *filter_mask = nla_get_u32(tb[i]);
5314 break;
5315 default:
5316 if (strict_check) {
5317 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
5318 return -EINVAL;
5319 }
5320 }
5321 }
5322
5323 return 0;
5324 }
5325
5326 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
5327 {
5328 const struct nlmsghdr *nlh = cb->nlh;
5329 struct net *net = sock_net(skb->sk);
5330 struct net_device *dev;
5331 int idx = 0;
5332 u32 portid = NETLINK_CB(cb->skb).portid;
5333 u32 seq = nlh->nlmsg_seq;
5334 u32 filter_mask = 0;
5335 int err;
5336
5337 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5338 cb->extack);
5339 if (err < 0 && cb->strict_check)
5340 return err;
5341
5342 rcu_read_lock();
5343 for_each_netdev_rcu(net, dev) {
5344 const struct net_device_ops *ops = dev->netdev_ops;
5345 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5346
5347 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5348 if (idx >= cb->args[0]) {
5349 err = br_dev->netdev_ops->ndo_bridge_getlink(
5350 skb, portid, seq, dev,
5351 filter_mask, NLM_F_MULTI);
5352 if (err < 0 && err != -EOPNOTSUPP) {
5353 if (likely(skb->len))
5354 break;
5355
5356 goto out_err;
5357 }
5358 }
5359 idx++;
5360 }
5361
5362 if (ops->ndo_bridge_getlink) {
5363 if (idx >= cb->args[0]) {
5364 err = ops->ndo_bridge_getlink(skb, portid,
5365 seq, dev,
5366 filter_mask,
5367 NLM_F_MULTI);
5368 if (err < 0 && err != -EOPNOTSUPP) {
5369 if (likely(skb->len))
5370 break;
5371
5372 goto out_err;
5373 }
5374 }
5375 idx++;
5376 }
5377 }
5378 err = skb->len;
5379 out_err:
5380 rcu_read_unlock();
5381 cb->args[0] = idx;
5382
5383 return err;
5384 }
5385
5386 static inline size_t bridge_nlmsg_size(void)
5387 {
5388 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5389 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5390 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5391 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
5392 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
5393 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
5394 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
5395 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
5396 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5397 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
5398 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
5399 }
5400
5401 static int rtnl_bridge_notify(struct net_device *dev)
5402 {
5403 struct net *net = dev_net(dev);
5404 struct sk_buff *skb;
5405 int err = -EOPNOTSUPP;
5406
5407 if (!dev->netdev_ops->ndo_bridge_getlink)
5408 return 0;
5409
5410 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5411 if (!skb) {
5412 err = -ENOMEM;
5413 goto errout;
5414 }
5415
5416 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5417 if (err < 0)
5418 goto errout;
5419
5420 /* Notification info is only filled for bridge ports, not the bridge
5421 * device itself. Therefore, a zero notification length is valid and
5422 * should not result in an error.
5423 */
5424 if (!skb->len)
5425 goto errout;
5426
5427 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5428 return 0;
5429 errout:
5430 WARN_ON(err == -EMSGSIZE);
5431 kfree_skb(skb);
5432 if (err)
5433 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5434 return err;
5435 }
5436
5437 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5438 struct netlink_ext_ack *extack)
5439 {
5440 struct net *net = sock_net(skb->sk);
5441 struct ifinfomsg *ifm;
5442 struct net_device *dev;
5443 struct nlattr *br_spec, *attr, *br_flags_attr = NULL;
5444 int rem, err = -EOPNOTSUPP;
5445 u16 flags = 0;
5446
5447 if (nlmsg_len(nlh) < sizeof(*ifm))
5448 return -EINVAL;
5449
5450 ifm = nlmsg_data(nlh);
5451 if (ifm->ifi_family != AF_BRIDGE)
5452 return -EPFNOSUPPORT;
5453
5454 dev = __dev_get_by_index(net, ifm->ifi_index);
5455 if (!dev) {
5456 NL_SET_ERR_MSG(extack, "unknown ifindex");
5457 return -ENODEV;
5458 }
5459
5460 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5461 if (br_spec) {
5462 nla_for_each_nested(attr, br_spec, rem) {
5463 if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !br_flags_attr) {
5464 if (nla_len(attr) < sizeof(flags))
5465 return -EINVAL;
5466
5467 br_flags_attr = attr;
5468 flags = nla_get_u16(attr);
5469 }
5470
5471 if (nla_type(attr) == IFLA_BRIDGE_MODE) {
5472 if (nla_len(attr) < sizeof(u16))
5473 return -EINVAL;
5474 }
5475 }
5476 }
5477
5478 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5479 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5480
5481 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5482 err = -EOPNOTSUPP;
5483 goto out;
5484 }
5485
5486 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5487 extack);
5488 if (err)
5489 goto out;
5490
5491 flags &= ~BRIDGE_FLAGS_MASTER;
5492 }
5493
5494 if ((flags & BRIDGE_FLAGS_SELF)) {
5495 if (!dev->netdev_ops->ndo_bridge_setlink)
5496 err = -EOPNOTSUPP;
5497 else
5498 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5499 flags,
5500 extack);
5501 if (!err) {
5502 flags &= ~BRIDGE_FLAGS_SELF;
5503
5504 /* Generate event to notify upper layer of bridge
5505 * change
5506 */
5507 err = rtnl_bridge_notify(dev);
5508 }
5509 }
5510
5511 if (br_flags_attr)
5512 memcpy(nla_data(br_flags_attr), &flags, sizeof(flags));
5513 out:
5514 return err;
5515 }
5516
5517 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5518 struct netlink_ext_ack *extack)
5519 {
5520 struct net *net = sock_net(skb->sk);
5521 struct ifinfomsg *ifm;
5522 struct net_device *dev;
5523 struct nlattr *br_spec, *attr = NULL;
5524 int rem, err = -EOPNOTSUPP;
5525 u16 flags = 0;
5526 bool have_flags = false;
5527
5528 if (nlmsg_len(nlh) < sizeof(*ifm))
5529 return -EINVAL;
5530
5531 ifm = nlmsg_data(nlh);
5532 if (ifm->ifi_family != AF_BRIDGE)
5533 return -EPFNOSUPPORT;
5534
5535 dev = __dev_get_by_index(net, ifm->ifi_index);
5536 if (!dev) {
5537 NL_SET_ERR_MSG(extack, "unknown ifindex");
5538 return -ENODEV;
5539 }
5540
5541 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5542 if (br_spec) {
5543 nla_for_each_nested_type(attr, IFLA_BRIDGE_FLAGS, br_spec,
5544 rem) {
5545 if (nla_len(attr) < sizeof(flags))
5546 return -EINVAL;
5547
5548 have_flags = true;
5549 flags = nla_get_u16(attr);
5550 break;
5551 }
5552 }
5553
5554 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5555 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5556
5557 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5558 err = -EOPNOTSUPP;
5559 goto out;
5560 }
5561
5562 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5563 if (err)
5564 goto out;
5565
5566 flags &= ~BRIDGE_FLAGS_MASTER;
5567 }
5568
5569 if ((flags & BRIDGE_FLAGS_SELF)) {
5570 if (!dev->netdev_ops->ndo_bridge_dellink)
5571 err = -EOPNOTSUPP;
5572 else
5573 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5574 flags);
5575
5576 if (!err) {
5577 flags &= ~BRIDGE_FLAGS_SELF;
5578
5579 /* Generate event to notify upper layer of bridge
5580 * change
5581 */
5582 err = rtnl_bridge_notify(dev);
5583 }
5584 }
5585
5586 if (have_flags)
5587 memcpy(nla_data(attr), &flags, sizeof(flags));
5588 out:
5589 return err;
5590 }
5591
5592 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5593 {
5594 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5595 (!idxattr || idxattr == attrid);
5596 }
5597
5598 static bool
5599 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5600 {
5601 return dev->netdev_ops &&
5602 dev->netdev_ops->ndo_has_offload_stats &&
5603 dev->netdev_ops->ndo_get_offload_stats &&
5604 dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5605 }
5606
5607 static unsigned int
5608 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5609 {
5610 return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5611 sizeof(struct rtnl_link_stats64) : 0;
5612 }
5613
5614 static int
5615 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5616 struct sk_buff *skb)
5617 {
5618 unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5619 struct nlattr *attr = NULL;
5620 void *attr_data;
5621 int err;
5622
5623 if (!size)
5624 return -ENODATA;
5625
5626 attr = nla_reserve_64bit(skb, attr_id, size,
5627 IFLA_OFFLOAD_XSTATS_UNSPEC);
5628 if (!attr)
5629 return -EMSGSIZE;
5630
5631 attr_data = nla_data(attr);
5632 memset(attr_data, 0, size);
5633
5634 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5635 if (err)
5636 return err;
5637
5638 return 0;
5639 }
5640
5641 static unsigned int
5642 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5643 enum netdev_offload_xstats_type type)
5644 {
5645 bool enabled = netdev_offload_xstats_enabled(dev, type);
5646
5647 return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5648 }
5649
5650 struct rtnl_offload_xstats_request_used {
5651 bool request;
5652 bool used;
5653 };
5654
5655 static int
5656 rtnl_offload_xstats_get_stats(struct net_device *dev,
5657 enum netdev_offload_xstats_type type,
5658 struct rtnl_offload_xstats_request_used *ru,
5659 struct rtnl_hw_stats64 *stats,
5660 struct netlink_ext_ack *extack)
5661 {
5662 bool request;
5663 bool used;
5664 int err;
5665
5666 request = netdev_offload_xstats_enabled(dev, type);
5667 if (!request) {
5668 used = false;
5669 goto out;
5670 }
5671
5672 err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5673 if (err)
5674 return err;
5675
5676 out:
5677 if (ru) {
5678 ru->request = request;
5679 ru->used = used;
5680 }
5681 return 0;
5682 }
5683
5684 static int
5685 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5686 struct rtnl_offload_xstats_request_used *ru)
5687 {
5688 struct nlattr *nest;
5689
5690 nest = nla_nest_start(skb, attr_id);
5691 if (!nest)
5692 return -EMSGSIZE;
5693
5694 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5695 goto nla_put_failure;
5696
5697 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5698 goto nla_put_failure;
5699
5700 nla_nest_end(skb, nest);
5701 return 0;
5702
5703 nla_put_failure:
5704 nla_nest_cancel(skb, nest);
5705 return -EMSGSIZE;
5706 }
5707
5708 static int
5709 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5710 struct netlink_ext_ack *extack)
5711 {
5712 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5713 struct rtnl_offload_xstats_request_used ru_l3;
5714 struct nlattr *nest;
5715 int err;
5716
5717 err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5718 if (err)
5719 return err;
5720
5721 nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5722 if (!nest)
5723 return -EMSGSIZE;
5724
5725 if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5726 IFLA_OFFLOAD_XSTATS_L3_STATS,
5727 &ru_l3))
5728 goto nla_put_failure;
5729
5730 nla_nest_end(skb, nest);
5731 return 0;
5732
5733 nla_put_failure:
5734 nla_nest_cancel(skb, nest);
5735 return -EMSGSIZE;
5736 }
5737
5738 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5739 int *prividx, u32 off_filter_mask,
5740 struct netlink_ext_ack *extack)
5741 {
5742 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5743 int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5744 int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5745 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5746 bool have_data = false;
5747 int err;
5748
5749 if (*prividx <= attr_id_cpu_hit &&
5750 (off_filter_mask &
5751 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5752 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5753 if (!err) {
5754 have_data = true;
5755 } else if (err != -ENODATA) {
5756 *prividx = attr_id_cpu_hit;
5757 return err;
5758 }
5759 }
5760
5761 if (*prividx <= attr_id_hw_s_info &&
5762 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5763 *prividx = attr_id_hw_s_info;
5764
5765 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5766 if (err)
5767 return err;
5768
5769 have_data = true;
5770 *prividx = 0;
5771 }
5772
5773 if (*prividx <= attr_id_l3_stats &&
5774 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5775 unsigned int size_l3;
5776 struct nlattr *attr;
5777
5778 *prividx = attr_id_l3_stats;
5779
5780 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5781 if (!size_l3)
5782 goto skip_l3_stats;
5783 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5784 IFLA_OFFLOAD_XSTATS_UNSPEC);
5785 if (!attr)
5786 return -EMSGSIZE;
5787
5788 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5789 nla_data(attr), extack);
5790 if (err)
5791 return err;
5792
5793 have_data = true;
5794 skip_l3_stats:
5795 *prividx = 0;
5796 }
5797
5798 if (!have_data)
5799 return -ENODATA;
5800
5801 *prividx = 0;
5802 return 0;
5803 }
5804
5805 static unsigned int
5806 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5807 enum netdev_offload_xstats_type type)
5808 {
5809 return nla_total_size(0) +
5810 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5811 nla_total_size(sizeof(u8)) +
5812 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5813 nla_total_size(sizeof(u8)) +
5814 0;
5815 }
5816
5817 static unsigned int
5818 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5819 {
5820 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5821
5822 return nla_total_size(0) +
5823 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5824 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5825 0;
5826 }
5827
5828 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5829 u32 off_filter_mask)
5830 {
5831 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5832 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5833 int nla_size = 0;
5834 int size;
5835
5836 if (off_filter_mask &
5837 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5838 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5839 nla_size += nla_total_size_64bit(size);
5840 }
5841
5842 if (off_filter_mask &
5843 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5844 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5845
5846 if (off_filter_mask &
5847 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5848 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5849 nla_size += nla_total_size_64bit(size);
5850 }
5851
5852 if (nla_size != 0)
5853 nla_size += nla_total_size(0);
5854
5855 return nla_size;
5856 }
5857
5858 struct rtnl_stats_dump_filters {
5859 /* mask[0] filters outer attributes. Then individual nests have their
5860 * filtering mask at the index of the nested attribute.
5861 */
5862 u32 mask[IFLA_STATS_MAX + 1];
5863 };
5864
5865 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5866 int type, u32 pid, u32 seq, u32 change,
5867 unsigned int flags,
5868 const struct rtnl_stats_dump_filters *filters,
5869 int *idxattr, int *prividx,
5870 struct netlink_ext_ack *extack)
5871 {
5872 unsigned int filter_mask = filters->mask[0];
5873 struct if_stats_msg *ifsm;
5874 struct nlmsghdr *nlh;
5875 struct nlattr *attr;
5876 int s_prividx = *prividx;
5877 int err;
5878
5879 ASSERT_RTNL();
5880
5881 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5882 if (!nlh)
5883 return -EMSGSIZE;
5884
5885 ifsm = nlmsg_data(nlh);
5886 ifsm->family = PF_UNSPEC;
5887 ifsm->pad1 = 0;
5888 ifsm->pad2 = 0;
5889 ifsm->ifindex = dev->ifindex;
5890 ifsm->filter_mask = filter_mask;
5891
5892 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5893 struct rtnl_link_stats64 *sp;
5894
5895 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5896 sizeof(struct rtnl_link_stats64),
5897 IFLA_STATS_UNSPEC);
5898 if (!attr) {
5899 err = -EMSGSIZE;
5900 goto nla_put_failure;
5901 }
5902
5903 sp = nla_data(attr);
5904 dev_get_stats(dev, sp);
5905 }
5906
5907 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5908 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5909
5910 if (ops && ops->fill_linkxstats) {
5911 *idxattr = IFLA_STATS_LINK_XSTATS;
5912 attr = nla_nest_start_noflag(skb,
5913 IFLA_STATS_LINK_XSTATS);
5914 if (!attr) {
5915 err = -EMSGSIZE;
5916 goto nla_put_failure;
5917 }
5918
5919 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5920 nla_nest_end(skb, attr);
5921 if (err)
5922 goto nla_put_failure;
5923 *idxattr = 0;
5924 }
5925 }
5926
5927 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5928 *idxattr)) {
5929 const struct rtnl_link_ops *ops = NULL;
5930 const struct net_device *master;
5931
5932 master = netdev_master_upper_dev_get(dev);
5933 if (master)
5934 ops = master->rtnl_link_ops;
5935 if (ops && ops->fill_linkxstats) {
5936 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5937 attr = nla_nest_start_noflag(skb,
5938 IFLA_STATS_LINK_XSTATS_SLAVE);
5939 if (!attr) {
5940 err = -EMSGSIZE;
5941 goto nla_put_failure;
5942 }
5943
5944 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5945 nla_nest_end(skb, attr);
5946 if (err)
5947 goto nla_put_failure;
5948 *idxattr = 0;
5949 }
5950 }
5951
5952 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5953 *idxattr)) {
5954 u32 off_filter_mask;
5955
5956 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5957 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5958 attr = nla_nest_start_noflag(skb,
5959 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5960 if (!attr) {
5961 err = -EMSGSIZE;
5962 goto nla_put_failure;
5963 }
5964
5965 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5966 off_filter_mask, extack);
5967 if (err == -ENODATA)
5968 nla_nest_cancel(skb, attr);
5969 else
5970 nla_nest_end(skb, attr);
5971
5972 if (err && err != -ENODATA)
5973 goto nla_put_failure;
5974 *idxattr = 0;
5975 }
5976
5977 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5978 struct rtnl_af_ops *af_ops;
5979
5980 *idxattr = IFLA_STATS_AF_SPEC;
5981 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5982 if (!attr) {
5983 err = -EMSGSIZE;
5984 goto nla_put_failure;
5985 }
5986
5987 rcu_read_lock();
5988 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5989 if (af_ops->fill_stats_af) {
5990 struct nlattr *af;
5991
5992 af = nla_nest_start_noflag(skb,
5993 af_ops->family);
5994 if (!af) {
5995 rcu_read_unlock();
5996 err = -EMSGSIZE;
5997 goto nla_put_failure;
5998 }
5999 err = af_ops->fill_stats_af(skb, dev);
6000
6001 if (err == -ENODATA) {
6002 nla_nest_cancel(skb, af);
6003 } else if (err < 0) {
6004 rcu_read_unlock();
6005 goto nla_put_failure;
6006 }
6007
6008 nla_nest_end(skb, af);
6009 }
6010 }
6011 rcu_read_unlock();
6012
6013 nla_nest_end(skb, attr);
6014
6015 *idxattr = 0;
6016 }
6017
6018 nlmsg_end(skb, nlh);
6019
6020 return 0;
6021
6022 nla_put_failure:
6023 /* not a multi message or no progress mean a real error */
6024 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
6025 nlmsg_cancel(skb, nlh);
6026 else
6027 nlmsg_end(skb, nlh);
6028
6029 return err;
6030 }
6031
6032 static size_t if_nlmsg_stats_size(const struct net_device *dev,
6033 const struct rtnl_stats_dump_filters *filters)
6034 {
6035 size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
6036 unsigned int filter_mask = filters->mask[0];
6037
6038 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
6039 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
6040
6041 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
6042 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
6043 int attr = IFLA_STATS_LINK_XSTATS;
6044
6045 if (ops && ops->get_linkxstats_size) {
6046 size += nla_total_size(ops->get_linkxstats_size(dev,
6047 attr));
6048 /* for IFLA_STATS_LINK_XSTATS */
6049 size += nla_total_size(0);
6050 }
6051 }
6052
6053 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
6054 struct net_device *_dev = (struct net_device *)dev;
6055 const struct rtnl_link_ops *ops = NULL;
6056 const struct net_device *master;
6057
6058 /* netdev_master_upper_dev_get can't take const */
6059 master = netdev_master_upper_dev_get(_dev);
6060 if (master)
6061 ops = master->rtnl_link_ops;
6062 if (ops && ops->get_linkxstats_size) {
6063 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
6064
6065 size += nla_total_size(ops->get_linkxstats_size(dev,
6066 attr));
6067 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
6068 size += nla_total_size(0);
6069 }
6070 }
6071
6072 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
6073 u32 off_filter_mask;
6074
6075 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
6076 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
6077 }
6078
6079 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
6080 struct rtnl_af_ops *af_ops;
6081
6082 /* for IFLA_STATS_AF_SPEC */
6083 size += nla_total_size(0);
6084
6085 rcu_read_lock();
6086 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
6087 if (af_ops->get_stats_af_size) {
6088 size += nla_total_size(
6089 af_ops->get_stats_af_size(dev));
6090
6091 /* for AF_* */
6092 size += nla_total_size(0);
6093 }
6094 }
6095 rcu_read_unlock();
6096 }
6097
6098 return size;
6099 }
6100
6101 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
6102
6103 static const struct nla_policy
6104 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
6105 [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
6106 NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
6107 };
6108
6109 static const struct nla_policy
6110 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
6111 [IFLA_STATS_GET_FILTERS] =
6112 NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
6113 };
6114
6115 static const struct nla_policy
6116 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
6117 [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
6118 };
6119
6120 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
6121 struct rtnl_stats_dump_filters *filters,
6122 struct netlink_ext_ack *extack)
6123 {
6124 struct nlattr *tb[IFLA_STATS_MAX + 1];
6125 int err;
6126 int at;
6127
6128 err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
6129 rtnl_stats_get_policy_filters, extack);
6130 if (err < 0)
6131 return err;
6132
6133 for (at = 1; at <= IFLA_STATS_MAX; at++) {
6134 if (tb[at]) {
6135 if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
6136 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
6137 return -EINVAL;
6138 }
6139 filters->mask[at] = nla_get_u32(tb[at]);
6140 }
6141 }
6142
6143 return 0;
6144 }
6145
6146 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
6147 u32 filter_mask,
6148 struct rtnl_stats_dump_filters *filters,
6149 struct netlink_ext_ack *extack)
6150 {
6151 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6152 int err;
6153 int i;
6154
6155 filters->mask[0] = filter_mask;
6156 for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
6157 filters->mask[i] = -1U;
6158
6159 err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
6160 IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
6161 if (err < 0)
6162 return err;
6163
6164 if (tb[IFLA_STATS_GET_FILTERS]) {
6165 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
6166 filters, extack);
6167 if (err)
6168 return err;
6169 }
6170
6171 return 0;
6172 }
6173
6174 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
6175 bool is_dump, struct netlink_ext_ack *extack)
6176 {
6177 struct if_stats_msg *ifsm;
6178
6179 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
6180 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
6181 return -EINVAL;
6182 }
6183
6184 if (!strict_check)
6185 return 0;
6186
6187 ifsm = nlmsg_data(nlh);
6188
6189 /* only requests using strict checks can pass data to influence
6190 * the dump. The legacy exception is filter_mask.
6191 */
6192 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
6193 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
6194 return -EINVAL;
6195 }
6196 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
6197 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
6198 return -EINVAL;
6199 }
6200
6201 return 0;
6202 }
6203
6204 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
6205 struct netlink_ext_ack *extack)
6206 {
6207 struct rtnl_stats_dump_filters filters;
6208 struct net *net = sock_net(skb->sk);
6209 struct net_device *dev = NULL;
6210 int idxattr = 0, prividx = 0;
6211 struct if_stats_msg *ifsm;
6212 struct sk_buff *nskb;
6213 int err;
6214
6215 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6216 false, extack);
6217 if (err)
6218 return err;
6219
6220 ifsm = nlmsg_data(nlh);
6221 if (ifsm->ifindex > 0)
6222 dev = __dev_get_by_index(net, ifsm->ifindex);
6223 else
6224 return -EINVAL;
6225
6226 if (!dev)
6227 return -ENODEV;
6228
6229 if (!ifsm->filter_mask) {
6230 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
6231 return -EINVAL;
6232 }
6233
6234 err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
6235 if (err)
6236 return err;
6237
6238 nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
6239 if (!nskb)
6240 return -ENOBUFS;
6241
6242 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
6243 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
6244 0, &filters, &idxattr, &prividx, extack);
6245 if (err < 0) {
6246 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
6247 WARN_ON(err == -EMSGSIZE);
6248 kfree_skb(nskb);
6249 } else {
6250 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
6251 }
6252
6253 return err;
6254 }
6255
6256 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
6257 {
6258 struct netlink_ext_ack *extack = cb->extack;
6259 struct rtnl_stats_dump_filters filters;
6260 struct net *net = sock_net(skb->sk);
6261 unsigned int flags = NLM_F_MULTI;
6262 struct if_stats_msg *ifsm;
6263 struct {
6264 unsigned long ifindex;
6265 int idxattr;
6266 int prividx;
6267 } *ctx = (void *)cb->ctx;
6268 struct net_device *dev;
6269 int err;
6270
6271 cb->seq = net->dev_base_seq;
6272
6273 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
6274 if (err)
6275 return err;
6276
6277 ifsm = nlmsg_data(cb->nlh);
6278 if (!ifsm->filter_mask) {
6279 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
6280 return -EINVAL;
6281 }
6282
6283 err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
6284 extack);
6285 if (err)
6286 return err;
6287
6288 for_each_netdev_dump(net, dev, ctx->ifindex) {
6289 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
6290 NETLINK_CB(cb->skb).portid,
6291 cb->nlh->nlmsg_seq, 0,
6292 flags, &filters,
6293 &ctx->idxattr, &ctx->prividx,
6294 extack);
6295 /* If we ran out of room on the first message,
6296 * we're in trouble.
6297 */
6298 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
6299
6300 if (err < 0)
6301 break;
6302 ctx->prividx = 0;
6303 ctx->idxattr = 0;
6304 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
6305 }
6306
6307 return err;
6308 }
6309
6310 void rtnl_offload_xstats_notify(struct net_device *dev)
6311 {
6312 struct rtnl_stats_dump_filters response_filters = {};
6313 struct net *net = dev_net(dev);
6314 int idxattr = 0, prividx = 0;
6315 struct sk_buff *skb;
6316 int err = -ENOBUFS;
6317
6318 ASSERT_RTNL();
6319
6320 response_filters.mask[0] |=
6321 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6322 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6323 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6324
6325 skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6326 GFP_KERNEL);
6327 if (!skb)
6328 goto errout;
6329
6330 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6331 &response_filters, &idxattr, &prividx, NULL);
6332 if (err < 0) {
6333 kfree_skb(skb);
6334 goto errout;
6335 }
6336
6337 rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6338 return;
6339
6340 errout:
6341 rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6342 }
6343 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6344
6345 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6346 struct netlink_ext_ack *extack)
6347 {
6348 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6349 struct rtnl_stats_dump_filters response_filters = {};
6350 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6351 struct net *net = sock_net(skb->sk);
6352 struct net_device *dev = NULL;
6353 struct if_stats_msg *ifsm;
6354 bool notify = false;
6355 int err;
6356
6357 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6358 false, extack);
6359 if (err)
6360 return err;
6361
6362 ifsm = nlmsg_data(nlh);
6363 if (ifsm->family != AF_UNSPEC) {
6364 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6365 return -EINVAL;
6366 }
6367
6368 if (ifsm->ifindex > 0)
6369 dev = __dev_get_by_index(net, ifsm->ifindex);
6370 else
6371 return -EINVAL;
6372
6373 if (!dev)
6374 return -ENODEV;
6375
6376 if (ifsm->filter_mask) {
6377 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6378 return -EINVAL;
6379 }
6380
6381 err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6382 ifla_stats_set_policy, extack);
6383 if (err < 0)
6384 return err;
6385
6386 if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6387 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6388
6389 if (req)
6390 err = netdev_offload_xstats_enable(dev, t_l3, extack);
6391 else
6392 err = netdev_offload_xstats_disable(dev, t_l3);
6393
6394 if (!err)
6395 notify = true;
6396 else if (err != -EALREADY)
6397 return err;
6398
6399 response_filters.mask[0] |=
6400 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6401 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6402 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6403 }
6404
6405 if (notify)
6406 rtnl_offload_xstats_notify(dev);
6407
6408 return 0;
6409 }
6410
6411 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6412 struct netlink_ext_ack *extack)
6413 {
6414 struct br_port_msg *bpm;
6415
6416 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6417 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6418 return -EINVAL;
6419 }
6420
6421 bpm = nlmsg_data(nlh);
6422 if (bpm->ifindex) {
6423 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6424 return -EINVAL;
6425 }
6426 if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6427 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6428 return -EINVAL;
6429 }
6430
6431 return 0;
6432 }
6433
6434 struct rtnl_mdb_dump_ctx {
6435 long idx;
6436 };
6437
6438 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6439 {
6440 struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6441 struct net *net = sock_net(skb->sk);
6442 struct net_device *dev;
6443 int idx, s_idx;
6444 int err;
6445
6446 NL_ASSERT_CTX_FITS(struct rtnl_mdb_dump_ctx);
6447
6448 if (cb->strict_check) {
6449 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6450 if (err)
6451 return err;
6452 }
6453
6454 s_idx = ctx->idx;
6455 idx = 0;
6456
6457 for_each_netdev(net, dev) {
6458 if (idx < s_idx)
6459 goto skip;
6460 if (!dev->netdev_ops->ndo_mdb_dump)
6461 goto skip;
6462
6463 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6464 if (err == -EMSGSIZE)
6465 goto out;
6466 /* Moving on to next device, reset markers and sequence
6467 * counters since they are all maintained per-device.
6468 */
6469 memset(cb->ctx, 0, sizeof(cb->ctx));
6470 cb->prev_seq = 0;
6471 cb->seq = 0;
6472 skip:
6473 idx++;
6474 }
6475
6476 out:
6477 ctx->idx = idx;
6478 return skb->len;
6479 }
6480
6481 static int rtnl_validate_mdb_entry_get(const struct nlattr *attr,
6482 struct netlink_ext_ack *extack)
6483 {
6484 struct br_mdb_entry *entry = nla_data(attr);
6485
6486 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6487 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6488 return -EINVAL;
6489 }
6490
6491 if (entry->ifindex) {
6492 NL_SET_ERR_MSG(extack, "Entry ifindex cannot be specified");
6493 return -EINVAL;
6494 }
6495
6496 if (entry->state) {
6497 NL_SET_ERR_MSG(extack, "Entry state cannot be specified");
6498 return -EINVAL;
6499 }
6500
6501 if (entry->flags) {
6502 NL_SET_ERR_MSG(extack, "Entry flags cannot be specified");
6503 return -EINVAL;
6504 }
6505
6506 if (entry->vid >= VLAN_VID_MASK) {
6507 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6508 return -EINVAL;
6509 }
6510
6511 if (entry->addr.proto != htons(ETH_P_IP) &&
6512 entry->addr.proto != htons(ETH_P_IPV6) &&
6513 entry->addr.proto != 0) {
6514 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6515 return -EINVAL;
6516 }
6517
6518 return 0;
6519 }
6520
6521 static const struct nla_policy mdba_get_policy[MDBA_GET_ENTRY_MAX + 1] = {
6522 [MDBA_GET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6523 rtnl_validate_mdb_entry_get,
6524 sizeof(struct br_mdb_entry)),
6525 [MDBA_GET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6526 };
6527
6528 static int rtnl_mdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6529 struct netlink_ext_ack *extack)
6530 {
6531 struct nlattr *tb[MDBA_GET_ENTRY_MAX + 1];
6532 struct net *net = sock_net(in_skb->sk);
6533 struct br_port_msg *bpm;
6534 struct net_device *dev;
6535 int err;
6536
6537 err = nlmsg_parse(nlh, sizeof(struct br_port_msg), tb,
6538 MDBA_GET_ENTRY_MAX, mdba_get_policy, extack);
6539 if (err)
6540 return err;
6541
6542 bpm = nlmsg_data(nlh);
6543 if (!bpm->ifindex) {
6544 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6545 return -EINVAL;
6546 }
6547
6548 dev = __dev_get_by_index(net, bpm->ifindex);
6549 if (!dev) {
6550 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6551 return -ENODEV;
6552 }
6553
6554 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_GET_ENTRY)) {
6555 NL_SET_ERR_MSG(extack, "Missing MDBA_GET_ENTRY attribute");
6556 return -EINVAL;
6557 }
6558
6559 if (!dev->netdev_ops->ndo_mdb_get) {
6560 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6561 return -EOPNOTSUPP;
6562 }
6563
6564 return dev->netdev_ops->ndo_mdb_get(dev, tb, NETLINK_CB(in_skb).portid,
6565 nlh->nlmsg_seq, extack);
6566 }
6567
6568 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6569 struct netlink_ext_ack *extack)
6570 {
6571 struct br_mdb_entry *entry = nla_data(attr);
6572
6573 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6574 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6575 return -EINVAL;
6576 }
6577
6578 if (entry->ifindex == 0) {
6579 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6580 return -EINVAL;
6581 }
6582
6583 if (entry->addr.proto == htons(ETH_P_IP)) {
6584 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6585 !ipv4_is_zeronet(entry->addr.u.ip4)) {
6586 NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6587 return -EINVAL;
6588 }
6589 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6590 NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6591 return -EINVAL;
6592 }
6593 #if IS_ENABLED(CONFIG_IPV6)
6594 } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6595 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6596 NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6597 return -EINVAL;
6598 }
6599 #endif
6600 } else if (entry->addr.proto == 0) {
6601 /* L2 mdb */
6602 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6603 NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6604 return -EINVAL;
6605 }
6606 } else {
6607 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6608 return -EINVAL;
6609 }
6610
6611 if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6612 NL_SET_ERR_MSG(extack, "Unknown entry state");
6613 return -EINVAL;
6614 }
6615 if (entry->vid >= VLAN_VID_MASK) {
6616 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6617 return -EINVAL;
6618 }
6619
6620 return 0;
6621 }
6622
6623 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6624 [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6625 [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6626 rtnl_validate_mdb_entry,
6627 sizeof(struct br_mdb_entry)),
6628 [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6629 };
6630
6631 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6632 struct netlink_ext_ack *extack)
6633 {
6634 struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6635 struct net *net = sock_net(skb->sk);
6636 struct br_port_msg *bpm;
6637 struct net_device *dev;
6638 int err;
6639
6640 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6641 MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6642 if (err)
6643 return err;
6644
6645 bpm = nlmsg_data(nlh);
6646 if (!bpm->ifindex) {
6647 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6648 return -EINVAL;
6649 }
6650
6651 dev = __dev_get_by_index(net, bpm->ifindex);
6652 if (!dev) {
6653 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6654 return -ENODEV;
6655 }
6656
6657 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6658 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6659 return -EINVAL;
6660 }
6661
6662 if (!dev->netdev_ops->ndo_mdb_add) {
6663 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6664 return -EOPNOTSUPP;
6665 }
6666
6667 return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6668 }
6669
6670 static int rtnl_validate_mdb_entry_del_bulk(const struct nlattr *attr,
6671 struct netlink_ext_ack *extack)
6672 {
6673 struct br_mdb_entry *entry = nla_data(attr);
6674 struct br_mdb_entry zero_entry = {};
6675
6676 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6677 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6678 return -EINVAL;
6679 }
6680
6681 if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6682 NL_SET_ERR_MSG(extack, "Unknown entry state");
6683 return -EINVAL;
6684 }
6685
6686 if (entry->flags) {
6687 NL_SET_ERR_MSG(extack, "Entry flags cannot be set");
6688 return -EINVAL;
6689 }
6690
6691 if (entry->vid >= VLAN_N_VID - 1) {
6692 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6693 return -EINVAL;
6694 }
6695
6696 if (memcmp(&entry->addr, &zero_entry.addr, sizeof(entry->addr))) {
6697 NL_SET_ERR_MSG(extack, "Entry address cannot be set");
6698 return -EINVAL;
6699 }
6700
6701 return 0;
6702 }
6703
6704 static const struct nla_policy mdba_del_bulk_policy[MDBA_SET_ENTRY_MAX + 1] = {
6705 [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6706 rtnl_validate_mdb_entry_del_bulk,
6707 sizeof(struct br_mdb_entry)),
6708 [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6709 };
6710
6711 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6712 struct netlink_ext_ack *extack)
6713 {
6714 bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
6715 struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6716 struct net *net = sock_net(skb->sk);
6717 struct br_port_msg *bpm;
6718 struct net_device *dev;
6719 int err;
6720
6721 if (!del_bulk)
6722 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6723 MDBA_SET_ENTRY_MAX, mdba_policy,
6724 extack);
6725 else
6726 err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY_MAX,
6727 mdba_del_bulk_policy, extack);
6728 if (err)
6729 return err;
6730
6731 bpm = nlmsg_data(nlh);
6732 if (!bpm->ifindex) {
6733 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6734 return -EINVAL;
6735 }
6736
6737 dev = __dev_get_by_index(net, bpm->ifindex);
6738 if (!dev) {
6739 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6740 return -ENODEV;
6741 }
6742
6743 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6744 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6745 return -EINVAL;
6746 }
6747
6748 if (del_bulk) {
6749 if (!dev->netdev_ops->ndo_mdb_del_bulk) {
6750 NL_SET_ERR_MSG(extack, "Device does not support MDB bulk deletion");
6751 return -EOPNOTSUPP;
6752 }
6753 return dev->netdev_ops->ndo_mdb_del_bulk(dev, tb, extack);
6754 }
6755
6756 if (!dev->netdev_ops->ndo_mdb_del) {
6757 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6758 return -EOPNOTSUPP;
6759 }
6760
6761 return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6762 }
6763
6764 /* Process one rtnetlink message. */
6765
6766 static int rtnl_dumpit(struct sk_buff *skb, struct netlink_callback *cb)
6767 {
6768 const bool needs_lock = !(cb->flags & RTNL_FLAG_DUMP_UNLOCKED);
6769 rtnl_dumpit_func dumpit = cb->data;
6770 int err;
6771
6772 /* Previous iteration have already finished, avoid calling->dumpit()
6773 * again, it may not expect to be called after it reached the end.
6774 */
6775 if (!dumpit)
6776 return 0;
6777
6778 if (needs_lock)
6779 rtnl_lock();
6780 err = dumpit(skb, cb);
6781 if (needs_lock)
6782 rtnl_unlock();
6783
6784 /* Old dump handlers used to send NLM_DONE as in a separate recvmsg().
6785 * Some applications which parse netlink manually depend on this.
6786 */
6787 if (cb->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE) {
6788 if (err < 0 && err != -EMSGSIZE)
6789 return err;
6790 if (!err)
6791 cb->data = NULL;
6792
6793 return skb->len;
6794 }
6795 return err;
6796 }
6797
6798 static int rtnetlink_dump_start(struct sock *ssk, struct sk_buff *skb,
6799 const struct nlmsghdr *nlh,
6800 struct netlink_dump_control *control)
6801 {
6802 if (control->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE ||
6803 !(control->flags & RTNL_FLAG_DUMP_UNLOCKED)) {
6804 WARN_ON(control->data);
6805 control->data = control->dump;
6806 control->dump = rtnl_dumpit;
6807 }
6808
6809 return netlink_dump_start(ssk, skb, nlh, control);
6810 }
6811
6812 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6813 struct netlink_ext_ack *extack)
6814 {
6815 struct net *net = sock_net(skb->sk);
6816 struct rtnl_link *link;
6817 enum rtnl_kinds kind;
6818 struct module *owner;
6819 int err = -EOPNOTSUPP;
6820 rtnl_doit_func doit;
6821 unsigned int flags;
6822 int family;
6823 int type;
6824
6825 type = nlh->nlmsg_type;
6826 if (type > RTM_MAX)
6827 return -EOPNOTSUPP;
6828
6829 type -= RTM_BASE;
6830
6831 /* All the messages must have at least 1 byte length */
6832 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6833 return 0;
6834
6835 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6836 kind = rtnl_msgtype_kind(type);
6837
6838 if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6839 return -EPERM;
6840
6841 rcu_read_lock();
6842 if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6843 struct sock *rtnl;
6844 rtnl_dumpit_func dumpit;
6845 u32 min_dump_alloc = 0;
6846
6847 link = rtnl_get_link(family, type);
6848 if (!link || !link->dumpit) {
6849 family = PF_UNSPEC;
6850 link = rtnl_get_link(family, type);
6851 if (!link || !link->dumpit)
6852 goto err_unlock;
6853 }
6854 owner = link->owner;
6855 dumpit = link->dumpit;
6856 flags = link->flags;
6857
6858 if (type == RTM_GETLINK - RTM_BASE)
6859 min_dump_alloc = rtnl_calcit(skb, nlh);
6860
6861 err = 0;
6862 /* need to do this before rcu_read_unlock() */
6863 if (!try_module_get(owner))
6864 err = -EPROTONOSUPPORT;
6865
6866 rcu_read_unlock();
6867
6868 rtnl = net->rtnl;
6869 if (err == 0) {
6870 struct netlink_dump_control c = {
6871 .dump = dumpit,
6872 .min_dump_alloc = min_dump_alloc,
6873 .module = owner,
6874 .flags = flags,
6875 };
6876 err = rtnetlink_dump_start(rtnl, skb, nlh, &c);
6877 /* netlink_dump_start() will keep a reference on
6878 * module if dump is still in progress.
6879 */
6880 module_put(owner);
6881 }
6882 return err;
6883 }
6884
6885 link = rtnl_get_link(family, type);
6886 if (!link || !link->doit) {
6887 family = PF_UNSPEC;
6888 link = rtnl_get_link(PF_UNSPEC, type);
6889 if (!link || !link->doit)
6890 goto out_unlock;
6891 }
6892
6893 owner = link->owner;
6894 if (!try_module_get(owner)) {
6895 err = -EPROTONOSUPPORT;
6896 goto out_unlock;
6897 }
6898
6899 flags = link->flags;
6900 if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6901 !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6902 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6903 module_put(owner);
6904 goto err_unlock;
6905 }
6906
6907 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6908 doit = link->doit;
6909 rcu_read_unlock();
6910 if (doit)
6911 err = doit(skb, nlh, extack);
6912 module_put(owner);
6913 return err;
6914 }
6915 rcu_read_unlock();
6916
6917 rtnl_lock();
6918 link = rtnl_get_link(family, type);
6919 if (link && link->doit)
6920 err = link->doit(skb, nlh, extack);
6921 rtnl_unlock();
6922
6923 module_put(owner);
6924
6925 return err;
6926
6927 out_unlock:
6928 rcu_read_unlock();
6929 return err;
6930
6931 err_unlock:
6932 rcu_read_unlock();
6933 return -EOPNOTSUPP;
6934 }
6935
6936 static void rtnetlink_rcv(struct sk_buff *skb)
6937 {
6938 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6939 }
6940
6941 static int rtnetlink_bind(struct net *net, int group)
6942 {
6943 switch (group) {
6944 case RTNLGRP_IPV4_MROUTE_R:
6945 case RTNLGRP_IPV6_MROUTE_R:
6946 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6947 return -EPERM;
6948 break;
6949 }
6950 return 0;
6951 }
6952
6953 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6954 {
6955 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6956
6957 switch (event) {
6958 case NETDEV_REBOOT:
6959 case NETDEV_CHANGEMTU:
6960 case NETDEV_CHANGEADDR:
6961 case NETDEV_CHANGENAME:
6962 case NETDEV_FEAT_CHANGE:
6963 case NETDEV_BONDING_FAILOVER:
6964 case NETDEV_POST_TYPE_CHANGE:
6965 case NETDEV_NOTIFY_PEERS:
6966 case NETDEV_CHANGEUPPER:
6967 case NETDEV_RESEND_IGMP:
6968 case NETDEV_CHANGEINFODATA:
6969 case NETDEV_CHANGELOWERSTATE:
6970 case NETDEV_CHANGE_TX_QUEUE_LEN:
6971 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6972 GFP_KERNEL, NULL, 0, 0, NULL);
6973 break;
6974 default:
6975 break;
6976 }
6977 return NOTIFY_DONE;
6978 }
6979
6980 static struct notifier_block rtnetlink_dev_notifier = {
6981 .notifier_call = rtnetlink_event,
6982 };
6983
6984
6985 static int __net_init rtnetlink_net_init(struct net *net)
6986 {
6987 struct sock *sk;
6988 struct netlink_kernel_cfg cfg = {
6989 .groups = RTNLGRP_MAX,
6990 .input = rtnetlink_rcv,
6991 .flags = NL_CFG_F_NONROOT_RECV,
6992 .bind = rtnetlink_bind,
6993 };
6994
6995 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6996 if (!sk)
6997 return -ENOMEM;
6998 net->rtnl = sk;
6999 return 0;
7000 }
7001
7002 static void __net_exit rtnetlink_net_exit(struct net *net)
7003 {
7004 netlink_kernel_release(net->rtnl);
7005 net->rtnl = NULL;
7006 }
7007
7008 static struct pernet_operations rtnetlink_net_ops = {
7009 .init = rtnetlink_net_init,
7010 .exit = rtnetlink_net_exit,
7011 };
7012
7013 static const struct rtnl_msg_handler rtnetlink_rtnl_msg_handlers[] __initconst = {
7014 {.msgtype = RTM_NEWLINK, .doit = rtnl_newlink,
7015 .flags = RTNL_FLAG_DOIT_PERNET},
7016 {.msgtype = RTM_DELLINK, .doit = rtnl_dellink,
7017 .flags = RTNL_FLAG_DOIT_PERNET_WIP},
7018 {.msgtype = RTM_GETLINK, .doit = rtnl_getlink,
7019 .dumpit = rtnl_dump_ifinfo, .flags = RTNL_FLAG_DUMP_SPLIT_NLM_DONE},
7020 {.msgtype = RTM_SETLINK, .doit = rtnl_setlink,
7021 .flags = RTNL_FLAG_DOIT_PERNET_WIP},
7022 {.msgtype = RTM_GETADDR, .dumpit = rtnl_dump_all},
7023 {.msgtype = RTM_GETROUTE, .dumpit = rtnl_dump_all},
7024 {.msgtype = RTM_GETNETCONF, .dumpit = rtnl_dump_all},
7025 {.msgtype = RTM_GETSTATS, .doit = rtnl_stats_get,
7026 .dumpit = rtnl_stats_dump},
7027 {.msgtype = RTM_SETSTATS, .doit = rtnl_stats_set},
7028 {.msgtype = RTM_NEWLINKPROP, .doit = rtnl_newlinkprop},
7029 {.msgtype = RTM_DELLINKPROP, .doit = rtnl_dellinkprop},
7030 {.protocol = PF_BRIDGE, .msgtype = RTM_GETLINK,
7031 .dumpit = rtnl_bridge_getlink},
7032 {.protocol = PF_BRIDGE, .msgtype = RTM_DELLINK,
7033 .doit = rtnl_bridge_dellink},
7034 {.protocol = PF_BRIDGE, .msgtype = RTM_SETLINK,
7035 .doit = rtnl_bridge_setlink},
7036 {.protocol = PF_BRIDGE, .msgtype = RTM_NEWNEIGH, .doit = rtnl_fdb_add},
7037 {.protocol = PF_BRIDGE, .msgtype = RTM_DELNEIGH, .doit = rtnl_fdb_del,
7038 .flags = RTNL_FLAG_BULK_DEL_SUPPORTED},
7039 {.protocol = PF_BRIDGE, .msgtype = RTM_GETNEIGH, .doit = rtnl_fdb_get,
7040 .dumpit = rtnl_fdb_dump},
7041 {.protocol = PF_BRIDGE, .msgtype = RTM_NEWMDB, .doit = rtnl_mdb_add},
7042 {.protocol = PF_BRIDGE, .msgtype = RTM_DELMDB, .doit = rtnl_mdb_del,
7043 .flags = RTNL_FLAG_BULK_DEL_SUPPORTED},
7044 {.protocol = PF_BRIDGE, .msgtype = RTM_GETMDB, .doit = rtnl_mdb_get,
7045 .dumpit = rtnl_mdb_dump},
7046 };
7047
7048 void __init rtnetlink_init(void)
7049 {
7050 if (register_pernet_subsys(&rtnetlink_net_ops))
7051 panic("rtnetlink_init: cannot initialize rtnetlink\n");
7052
7053 register_netdevice_notifier(&rtnetlink_dev_notifier);
7054
7055 rtnl_register_many(rtnetlink_rtnl_msg_handlers);
7056 }
Linux Kernel