?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
printf: Remove unused 'bprintf'
[drm/drm-misc.git] / net / core / rtnetlink.c
blob58df76fe408a4677db52a3c1f94674119dc4c925
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 EXPORT_SYMBOL(rtnl_lock_killable);
88
89 static struct sk_buff *defer_kfree_skb_list;
90 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
91 {
92 if (head && tail) {
93 tail->next = defer_kfree_skb_list;
94 defer_kfree_skb_list = head;
95 }
96 }
97 EXPORT_SYMBOL(rtnl_kfree_skbs);
98
99 void __rtnl_unlock(void)
100 {
101 struct sk_buff *head = defer_kfree_skb_list;
102
103 defer_kfree_skb_list = NULL;
104
105 /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
106 * is used. In some places, e.g. in cfg80211, we have code that will do
107 * something like
108 * rtnl_lock()
109 * wiphy_lock()
110 * ...
111 * rtnl_unlock()
112 *
113 * and because netdev_run_todo() acquires the RTNL for items on the list
114 * we could cause a situation such as this:
115 * Thread 1 Thread 2
116 * rtnl_lock()
117 * unregister_netdevice()
118 * __rtnl_unlock()
119 * rtnl_lock()
120 * wiphy_lock()
121 * rtnl_unlock()
122 * netdev_run_todo()
123 * __rtnl_unlock()
124 *
125 * // list not empty now
126 * // because of thread 2
127 * rtnl_lock()
128 * while (!list_empty(...))
129 * rtnl_lock()
130 * wiphy_lock()
131 * **** DEADLOCK ****
132 *
133 * However, usage of __rtnl_unlock() is rare, and so we can ensure that
134 * it's not used in cases where something is added to do the list.
135 */
136 WARN_ON(!list_empty(&net_todo_list));
137
138 mutex_unlock(&rtnl_mutex);
139
140 while (head) {
141 struct sk_buff *next = head->next;
142
143 kfree_skb(head);
144 cond_resched();
145 head = next;
146 }
147 }
148
149 void rtnl_unlock(void)
150 {
151 /* This fellow will unlock it for us. */
152 netdev_run_todo();
153 }
154 EXPORT_SYMBOL(rtnl_unlock);
155
156 int rtnl_trylock(void)
157 {
158 return mutex_trylock(&rtnl_mutex);
159 }
160 EXPORT_SYMBOL(rtnl_trylock);
161
162 int rtnl_is_locked(void)
163 {
164 return mutex_is_locked(&rtnl_mutex);
165 }
166 EXPORT_SYMBOL(rtnl_is_locked);
167
168 bool refcount_dec_and_rtnl_lock(refcount_t *r)
169 {
170 return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
171 }
172 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
173
174 #ifdef CONFIG_PROVE_LOCKING
175 bool lockdep_rtnl_is_held(void)
176 {
177 return lockdep_is_held(&rtnl_mutex);
178 }
179 EXPORT_SYMBOL(lockdep_rtnl_is_held);
180 #endif /* #ifdef CONFIG_PROVE_LOCKING */
181
182 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL
183 void __rtnl_net_lock(struct net *net)
184 {
185 ASSERT_RTNL();
186
187 mutex_lock(&net->rtnl_mutex);
188 }
189 EXPORT_SYMBOL(__rtnl_net_lock);
190
191 void __rtnl_net_unlock(struct net *net)
192 {
193 ASSERT_RTNL();
194
195 mutex_unlock(&net->rtnl_mutex);
196 }
197 EXPORT_SYMBOL(__rtnl_net_unlock);
198
199 void rtnl_net_lock(struct net *net)
200 {
201 rtnl_lock();
202 __rtnl_net_lock(net);
203 }
204 EXPORT_SYMBOL(rtnl_net_lock);
205
206 void rtnl_net_unlock(struct net *net)
207 {
208 __rtnl_net_unlock(net);
209 rtnl_unlock();
210 }
211 EXPORT_SYMBOL(rtnl_net_unlock);
212
213 int rtnl_net_trylock(struct net *net)
214 {
215 int ret = rtnl_trylock();
216
217 if (ret)
218 __rtnl_net_lock(net);
219
220 return ret;
221 }
222 EXPORT_SYMBOL(rtnl_net_trylock);
223
224 static int rtnl_net_cmp_locks(const struct net *net_a, const struct net *net_b)
225 {
226 if (net_eq(net_a, net_b))
227 return 0;
228
229 /* always init_net first */
230 if (net_eq(net_a, &init_net))
231 return -1;
232
233 if (net_eq(net_b, &init_net))
234 return 1;
235
236 /* otherwise lock in ascending order */
237 return net_a < net_b ? -1 : 1;
238 }
239
240 int rtnl_net_lock_cmp_fn(const struct lockdep_map *a, const struct lockdep_map *b)
241 {
242 const struct net *net_a, *net_b;
243
244 net_a = container_of(a, struct net, rtnl_mutex.dep_map);
245 net_b = container_of(b, struct net, rtnl_mutex.dep_map);
246
247 return rtnl_net_cmp_locks(net_a, net_b);
248 }
249
250 bool rtnl_net_is_locked(struct net *net)
251 {
252 return rtnl_is_locked() && mutex_is_locked(&net->rtnl_mutex);
253 }
254 EXPORT_SYMBOL(rtnl_net_is_locked);
255
256 bool lockdep_rtnl_net_is_held(struct net *net)
257 {
258 return lockdep_rtnl_is_held() && lockdep_is_held(&net->rtnl_mutex);
259 }
260 EXPORT_SYMBOL(lockdep_rtnl_net_is_held);
261 #else
262 static int rtnl_net_cmp_locks(const struct net *net_a, const struct net *net_b)
263 {
264 /* No need to swap */
265 return -1;
266 }
267 #endif
268
269 struct rtnl_nets {
270 /* ->newlink() needs to freeze 3 netns at most;
271 * 2 for the new device, 1 for its peer.
272 */
273 struct net *net[3];
274 unsigned char len;
275 };
276
277 static void rtnl_nets_init(struct rtnl_nets *rtnl_nets)
278 {
279 memset(rtnl_nets, 0, sizeof(*rtnl_nets));
280 }
281
282 static void rtnl_nets_destroy(struct rtnl_nets *rtnl_nets)
283 {
284 int i;
285
286 for (i = 0; i < rtnl_nets->len; i++) {
287 put_net(rtnl_nets->net[i]);
288 rtnl_nets->net[i] = NULL;
289 }
290
291 rtnl_nets->len = 0;
292 }
293
294 /**
295 * rtnl_nets_add - Add netns to be locked before ->newlink().
296 *
297 * @rtnl_nets: rtnl_nets pointer passed to ->get_peer_net().
298 * @net: netns pointer with an extra refcnt held.
299 *
300 * The extra refcnt is released in rtnl_nets_destroy().
301 */
302 static void rtnl_nets_add(struct rtnl_nets *rtnl_nets, struct net *net)
303 {
304 int i;
305
306 DEBUG_NET_WARN_ON_ONCE(rtnl_nets->len == ARRAY_SIZE(rtnl_nets->net));
307
308 for (i = 0; i < rtnl_nets->len; i++) {
309 switch (rtnl_net_cmp_locks(rtnl_nets->net[i], net)) {
310 case 0:
311 put_net(net);
312 return;
313 case 1:
314 swap(rtnl_nets->net[i], net);
315 }
316 }
317
318 rtnl_nets->net[i] = net;
319 rtnl_nets->len++;
320 }
321
322 static void rtnl_nets_lock(struct rtnl_nets *rtnl_nets)
323 {
324 int i;
325
326 rtnl_lock();
327
328 for (i = 0; i < rtnl_nets->len; i++)
329 __rtnl_net_lock(rtnl_nets->net[i]);
330 }
331
332 static void rtnl_nets_unlock(struct rtnl_nets *rtnl_nets)
333 {
334 int i;
335
336 for (i = 0; i < rtnl_nets->len; i++)
337 __rtnl_net_unlock(rtnl_nets->net[i]);
338
339 rtnl_unlock();
340 }
341
342 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
343
344 static inline int rtm_msgindex(int msgtype)
345 {
346 int msgindex = msgtype - RTM_BASE;
347
348 /*
349 * msgindex < 0 implies someone tried to register a netlink
350 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
351 * the message type has not been added to linux/rtnetlink.h
352 */
353 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
354
355 return msgindex;
356 }
357
358 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
359 {
360 struct rtnl_link __rcu **tab;
361
362 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
363 protocol = PF_UNSPEC;
364
365 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
366 if (!tab)
367 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
368
369 return rcu_dereference_rtnl(tab[msgtype]);
370 }
371
372 static int rtnl_register_internal(struct module *owner,
373 int protocol, int msgtype,
374 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
375 unsigned int flags)
376 {
377 struct rtnl_link *link, *old;
378 struct rtnl_link __rcu **tab;
379 int msgindex;
380 int ret = -ENOBUFS;
381
382 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
383 msgindex = rtm_msgindex(msgtype);
384
385 rtnl_lock();
386 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
387 if (tab == NULL) {
388 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
389 if (!tab)
390 goto unlock;
391
392 /* ensures we see the 0 stores */
393 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
394 }
395
396 old = rtnl_dereference(tab[msgindex]);
397 if (old) {
398 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
399 if (!link)
400 goto unlock;
401 } else {
402 link = kzalloc(sizeof(*link), GFP_KERNEL);
403 if (!link)
404 goto unlock;
405 }
406
407 WARN_ON(link->owner && link->owner != owner);
408 link->owner = owner;
409
410 WARN_ON(doit && link->doit && link->doit != doit);
411 if (doit)
412 link->doit = doit;
413 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
414 if (dumpit)
415 link->dumpit = dumpit;
416
417 WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
418 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
419 link->flags |= flags;
420
421 /* publish protocol:msgtype */
422 rcu_assign_pointer(tab[msgindex], link);
423 ret = 0;
424 if (old)
425 kfree_rcu(old, rcu);
426 unlock:
427 rtnl_unlock();
428 return ret;
429 }
430
431 /**
432 * rtnl_unregister - Unregister a rtnetlink message type
433 * @protocol: Protocol family or PF_UNSPEC
434 * @msgtype: rtnetlink message type
435 *
436 * Returns 0 on success or a negative error code.
437 */
438 static int rtnl_unregister(int protocol, int msgtype)
439 {
440 struct rtnl_link __rcu **tab;
441 struct rtnl_link *link;
442 int msgindex;
443
444 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
445 msgindex = rtm_msgindex(msgtype);
446
447 rtnl_lock();
448 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
449 if (!tab) {
450 rtnl_unlock();
451 return -ENOENT;
452 }
453
454 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
455 rtnl_unlock();
456
457 kfree_rcu(link, rcu);
458
459 return 0;
460 }
461
462 /**
463 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
464 * @protocol : Protocol family or PF_UNSPEC
465 *
466 * Identical to calling rtnl_unregster() for all registered message types
467 * of a certain protocol family.
468 */
469 void rtnl_unregister_all(int protocol)
470 {
471 struct rtnl_link __rcu **tab;
472 struct rtnl_link *link;
473 int msgindex;
474
475 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
476
477 rtnl_lock();
478 tab = rcu_replace_pointer_rtnl(rtnl_msg_handlers[protocol], NULL);
479 if (!tab) {
480 rtnl_unlock();
481 return;
482 }
483 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
484 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
485 kfree_rcu(link, rcu);
486 }
487 rtnl_unlock();
488
489 synchronize_net();
490
491 kfree(tab);
492 }
493 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
494
495 /**
496 * __rtnl_register_many - Register rtnetlink message types
497 * @handlers: Array of struct rtnl_msg_handlers
498 * @n: The length of @handlers
499 *
500 * Registers the specified function pointers (at least one of them has
501 * to be non-NULL) to be called whenever a request message for the
502 * specified protocol family and message type is received.
503 *
504 * The special protocol family PF_UNSPEC may be used to define fallback
505 * function pointers for the case when no entry for the specific protocol
506 * family exists.
507 *
508 * When one element of @handlers fails to register,
509 * 1) built-in: panics.
510 * 2) modules : the previous successful registrations are unwinded
511 * and an error is returned.
512 *
513 * Use rtnl_register_many().
514 */
515 int __rtnl_register_many(const struct rtnl_msg_handler *handlers, int n)
516 {
517 const struct rtnl_msg_handler *handler;
518 int i, err;
519
520 for (i = 0, handler = handlers; i < n; i++, handler++) {
521 err = rtnl_register_internal(handler->owner, handler->protocol,
522 handler->msgtype, handler->doit,
523 handler->dumpit, handler->flags);
524 if (err) {
525 if (!handler->owner)
526 panic("Unable to register rtnetlink message "
527 "handlers, %pS\n", handlers);
528
529 __rtnl_unregister_many(handlers, i);
530 break;
531 }
532 }
533
534 return err;
535 }
536 EXPORT_SYMBOL_GPL(__rtnl_register_many);
537
538 void __rtnl_unregister_many(const struct rtnl_msg_handler *handlers, int n)
539 {
540 const struct rtnl_msg_handler *handler;
541 int i;
542
543 for (i = n - 1, handler = handlers + n - 1; i >= 0; i--, handler--)
544 rtnl_unregister(handler->protocol, handler->msgtype);
545 }
546 EXPORT_SYMBOL_GPL(__rtnl_unregister_many);
547
548 static DEFINE_MUTEX(link_ops_mutex);
549 static LIST_HEAD(link_ops);
550
551 static struct rtnl_link_ops *rtnl_link_ops_get(const char *kind, int *srcu_index)
552 {
553 struct rtnl_link_ops *ops;
554
555 rcu_read_lock();
556
557 list_for_each_entry_rcu(ops, &link_ops, list) {
558 if (!strcmp(ops->kind, kind)) {
559 *srcu_index = srcu_read_lock(&ops->srcu);
560 goto unlock;
561 }
562 }
563
564 ops = NULL;
565 unlock:
566 rcu_read_unlock();
567
568 return ops;
569 }
570
571 static void rtnl_link_ops_put(struct rtnl_link_ops *ops, int srcu_index)
572 {
573 srcu_read_unlock(&ops->srcu, srcu_index);
574 }
575
576 /**
577 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
578 * @ops: struct rtnl_link_ops * to register
579 *
580 * Returns 0 on success or a negative error code.
581 */
582 int rtnl_link_register(struct rtnl_link_ops *ops)
583 {
584 struct rtnl_link_ops *tmp;
585 int err;
586
587 /* Sanity-check max sizes to avoid stack buffer overflow. */
588 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
589 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
590 return -EINVAL;
591
592 /* The check for alloc/setup is here because if ops
593 * does not have that filled up, it is not possible
594 * to use the ops for creating device. So do not
595 * fill up dellink as well. That disables rtnl_dellink.
596 */
597 if ((ops->alloc || ops->setup) && !ops->dellink)
598 ops->dellink = unregister_netdevice_queue;
599
600 err = init_srcu_struct(&ops->srcu);
601 if (err)
602 return err;
603
604 mutex_lock(&link_ops_mutex);
605
606 list_for_each_entry(tmp, &link_ops, list) {
607 if (!strcmp(ops->kind, tmp->kind)) {
608 err = -EEXIST;
609 goto unlock;
610 }
611 }
612
613 list_add_tail_rcu(&ops->list, &link_ops);
614 unlock:
615 mutex_unlock(&link_ops_mutex);
616
617 return err;
618 }
619 EXPORT_SYMBOL_GPL(rtnl_link_register);
620
621 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
622 {
623 struct net_device *dev;
624 LIST_HEAD(list_kill);
625
626 for_each_netdev(net, dev) {
627 if (dev->rtnl_link_ops == ops)
628 ops->dellink(dev, &list_kill);
629 }
630 unregister_netdevice_many(&list_kill);
631 }
632
633 /* Return with the rtnl_lock held when there are no network
634 * devices unregistering in any network namespace.
635 */
636 static void rtnl_lock_unregistering_all(void)
637 {
638 DEFINE_WAIT_FUNC(wait, woken_wake_function);
639
640 add_wait_queue(&netdev_unregistering_wq, &wait);
641 for (;;) {
642 rtnl_lock();
643 /* We held write locked pernet_ops_rwsem, and parallel
644 * setup_net() and cleanup_net() are not possible.
645 */
646 if (!atomic_read(&dev_unreg_count))
647 break;
648 __rtnl_unlock();
649
650 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
651 }
652 remove_wait_queue(&netdev_unregistering_wq, &wait);
653 }
654
655 /**
656 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
657 * @ops: struct rtnl_link_ops * to unregister
658 */
659 void rtnl_link_unregister(struct rtnl_link_ops *ops)
660 {
661 struct net *net;
662
663 mutex_lock(&link_ops_mutex);
664 list_del_rcu(&ops->list);
665 mutex_unlock(&link_ops_mutex);
666
667 synchronize_srcu(&ops->srcu);
668 cleanup_srcu_struct(&ops->srcu);
669
670 /* Close the race with setup_net() and cleanup_net() */
671 down_write(&pernet_ops_rwsem);
672 rtnl_lock_unregistering_all();
673
674 for_each_net(net)
675 __rtnl_kill_links(net, ops);
676
677 rtnl_unlock();
678 up_write(&pernet_ops_rwsem);
679 }
680 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
681
682 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
683 {
684 struct net_device *master_dev;
685 const struct rtnl_link_ops *ops;
686 size_t size = 0;
687
688 rcu_read_lock();
689
690 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
691 if (!master_dev)
692 goto out;
693
694 ops = master_dev->rtnl_link_ops;
695 if (!ops || !ops->get_slave_size)
696 goto out;
697 /* IFLA_INFO_SLAVE_DATA + nested data */
698 size = nla_total_size(sizeof(struct nlattr)) +
699 ops->get_slave_size(master_dev, dev);
700
701 out:
702 rcu_read_unlock();
703 return size;
704 }
705
706 static size_t rtnl_link_get_size(const struct net_device *dev)
707 {
708 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
709 size_t size;
710
711 if (!ops)
712 return 0;
713
714 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
715 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
716
717 if (ops->get_size)
718 /* IFLA_INFO_DATA + nested data */
719 size += nla_total_size(sizeof(struct nlattr)) +
720 ops->get_size(dev);
721
722 if (ops->get_xstats_size)
723 /* IFLA_INFO_XSTATS */
724 size += nla_total_size(ops->get_xstats_size(dev));
725
726 size += rtnl_link_get_slave_info_data_size(dev);
727
728 return size;
729 }
730
731 static LIST_HEAD(rtnl_af_ops);
732
733 static struct rtnl_af_ops *rtnl_af_lookup(const int family, int *srcu_index)
734 {
735 struct rtnl_af_ops *ops;
736
737 ASSERT_RTNL();
738
739 rcu_read_lock();
740
741 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
742 if (ops->family == family) {
743 *srcu_index = srcu_read_lock(&ops->srcu);
744 goto unlock;
745 }
746 }
747
748 ops = NULL;
749 unlock:
750 rcu_read_unlock();
751
752 return ops;
753 }
754
755 static void rtnl_af_put(struct rtnl_af_ops *ops, int srcu_index)
756 {
757 srcu_read_unlock(&ops->srcu, srcu_index);
758 }
759
760 /**
761 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
762 * @ops: struct rtnl_af_ops * to register
763 *
764 * Return: 0 on success or a negative error code.
765 */
766 int rtnl_af_register(struct rtnl_af_ops *ops)
767 {
768 int err = init_srcu_struct(&ops->srcu);
769
770 if (err)
771 return err;
772
773 rtnl_lock();
774 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
775 rtnl_unlock();
776
777 return 0;
778 }
779 EXPORT_SYMBOL_GPL(rtnl_af_register);
780
781 /**
782 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
783 * @ops: struct rtnl_af_ops * to unregister
784 */
785 void rtnl_af_unregister(struct rtnl_af_ops *ops)
786 {
787 rtnl_lock();
788 list_del_rcu(&ops->list);
789 rtnl_unlock();
790
791 synchronize_rcu();
792 synchronize_srcu(&ops->srcu);
793 cleanup_srcu_struct(&ops->srcu);
794 }
795 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
796
797 static size_t rtnl_link_get_af_size(const struct net_device *dev,
798 u32 ext_filter_mask)
799 {
800 struct rtnl_af_ops *af_ops;
801 size_t size;
802
803 /* IFLA_AF_SPEC */
804 size = nla_total_size(sizeof(struct nlattr));
805
806 rcu_read_lock();
807 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
808 if (af_ops->get_link_af_size) {
809 /* AF_* + nested data */
810 size += nla_total_size(sizeof(struct nlattr)) +
811 af_ops->get_link_af_size(dev, ext_filter_mask);
812 }
813 }
814 rcu_read_unlock();
815
816 return size;
817 }
818
819 static bool rtnl_have_link_slave_info(const struct net_device *dev)
820 {
821 struct net_device *master_dev;
822 bool ret = false;
823
824 rcu_read_lock();
825
826 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
827 if (master_dev && master_dev->rtnl_link_ops)
828 ret = true;
829 rcu_read_unlock();
830 return ret;
831 }
832
833 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
834 const struct net_device *dev)
835 {
836 struct net_device *master_dev;
837 const struct rtnl_link_ops *ops;
838 struct nlattr *slave_data;
839 int err;
840
841 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
842 if (!master_dev)
843 return 0;
844 ops = master_dev->rtnl_link_ops;
845 if (!ops)
846 return 0;
847 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
848 return -EMSGSIZE;
849 if (ops->fill_slave_info) {
850 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
851 if (!slave_data)
852 return -EMSGSIZE;
853 err = ops->fill_slave_info(skb, master_dev, dev);
854 if (err < 0)
855 goto err_cancel_slave_data;
856 nla_nest_end(skb, slave_data);
857 }
858 return 0;
859
860 err_cancel_slave_data:
861 nla_nest_cancel(skb, slave_data);
862 return err;
863 }
864
865 static int rtnl_link_info_fill(struct sk_buff *skb,
866 const struct net_device *dev)
867 {
868 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
869 struct nlattr *data;
870 int err;
871
872 if (!ops)
873 return 0;
874 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
875 return -EMSGSIZE;
876 if (ops->fill_xstats) {
877 err = ops->fill_xstats(skb, dev);
878 if (err < 0)
879 return err;
880 }
881 if (ops->fill_info) {
882 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
883 if (data == NULL)
884 return -EMSGSIZE;
885 err = ops->fill_info(skb, dev);
886 if (err < 0)
887 goto err_cancel_data;
888 nla_nest_end(skb, data);
889 }
890 return 0;
891
892 err_cancel_data:
893 nla_nest_cancel(skb, data);
894 return err;
895 }
896
897 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
898 {
899 struct nlattr *linkinfo;
900 int err = -EMSGSIZE;
901
902 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
903 if (linkinfo == NULL)
904 goto out;
905
906 err = rtnl_link_info_fill(skb, dev);
907 if (err < 0)
908 goto err_cancel_link;
909
910 err = rtnl_link_slave_info_fill(skb, dev);
911 if (err < 0)
912 goto err_cancel_link;
913
914 nla_nest_end(skb, linkinfo);
915 return 0;
916
917 err_cancel_link:
918 nla_nest_cancel(skb, linkinfo);
919 out:
920 return err;
921 }
922
923 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
924 {
925 struct sock *rtnl = net->rtnl;
926
927 return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
928 }
929
930 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
931 {
932 struct sock *rtnl = net->rtnl;
933
934 return nlmsg_unicast(rtnl, skb, pid);
935 }
936 EXPORT_SYMBOL(rtnl_unicast);
937
938 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
939 const struct nlmsghdr *nlh, gfp_t flags)
940 {
941 struct sock *rtnl = net->rtnl;
942
943 nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
944 }
945 EXPORT_SYMBOL(rtnl_notify);
946
947 void rtnl_set_sk_err(struct net *net, u32 group, int error)
948 {
949 struct sock *rtnl = net->rtnl;
950
951 netlink_set_err(rtnl, 0, group, error);
952 }
953 EXPORT_SYMBOL(rtnl_set_sk_err);
954
955 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
956 {
957 struct nlattr *mx;
958 int i, valid = 0;
959
960 /* nothing is dumped for dst_default_metrics, so just skip the loop */
961 if (metrics == dst_default_metrics.metrics)
962 return 0;
963
964 mx = nla_nest_start_noflag(skb, RTA_METRICS);
965 if (mx == NULL)
966 return -ENOBUFS;
967
968 for (i = 0; i < RTAX_MAX; i++) {
969 if (metrics[i]) {
970 if (i == RTAX_CC_ALGO - 1) {
971 char tmp[TCP_CA_NAME_MAX], *name;
972
973 name = tcp_ca_get_name_by_key(metrics[i], tmp);
974 if (!name)
975 continue;
976 if (nla_put_string(skb, i + 1, name))
977 goto nla_put_failure;
978 } else if (i == RTAX_FEATURES - 1) {
979 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
980
981 if (!user_features)
982 continue;
983 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
984 if (nla_put_u32(skb, i + 1, user_features))
985 goto nla_put_failure;
986 } else {
987 if (nla_put_u32(skb, i + 1, metrics[i]))
988 goto nla_put_failure;
989 }
990 valid++;
991 }
992 }
993
994 if (!valid) {
995 nla_nest_cancel(skb, mx);
996 return 0;
997 }
998
999 return nla_nest_end(skb, mx);
1000
1001 nla_put_failure:
1002 nla_nest_cancel(skb, mx);
1003 return -EMSGSIZE;
1004 }
1005 EXPORT_SYMBOL(rtnetlink_put_metrics);
1006
1007 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
1008 long expires, u32 error)
1009 {
1010 struct rta_cacheinfo ci = {
1011 .rta_error = error,
1012 .rta_id = id,
1013 };
1014
1015 if (dst) {
1016 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
1017 ci.rta_used = dst->__use;
1018 ci.rta_clntref = rcuref_read(&dst->__rcuref);
1019 }
1020 if (expires) {
1021 unsigned long clock;
1022
1023 clock = jiffies_to_clock_t(abs(expires));
1024 clock = min_t(unsigned long, clock, INT_MAX);
1025 ci.rta_expires = (expires > 0) ? clock : -clock;
1026 }
1027 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
1028 }
1029 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
1030
1031 void netdev_set_operstate(struct net_device *dev, int newstate)
1032 {
1033 unsigned int old = READ_ONCE(dev->operstate);
1034
1035 do {
1036 if (old == newstate)
1037 return;
1038 } while (!try_cmpxchg(&dev->operstate, &old, newstate));
1039
1040 netdev_state_change(dev);
1041 }
1042 EXPORT_SYMBOL(netdev_set_operstate);
1043
1044 static void set_operstate(struct net_device *dev, unsigned char transition)
1045 {
1046 unsigned char operstate = READ_ONCE(dev->operstate);
1047
1048 switch (transition) {
1049 case IF_OPER_UP:
1050 if ((operstate == IF_OPER_DORMANT ||
1051 operstate == IF_OPER_TESTING ||
1052 operstate == IF_OPER_UNKNOWN) &&
1053 !netif_dormant(dev) && !netif_testing(dev))
1054 operstate = IF_OPER_UP;
1055 break;
1056
1057 case IF_OPER_TESTING:
1058 if (netif_oper_up(dev))
1059 operstate = IF_OPER_TESTING;
1060 break;
1061
1062 case IF_OPER_DORMANT:
1063 if (netif_oper_up(dev))
1064 operstate = IF_OPER_DORMANT;
1065 break;
1066 }
1067
1068 netdev_set_operstate(dev, operstate);
1069 }
1070
1071 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
1072 {
1073 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
1074 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
1075 }
1076
1077 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
1078 const struct ifinfomsg *ifm)
1079 {
1080 unsigned int flags = ifm->ifi_flags;
1081
1082 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
1083 if (ifm->ifi_change)
1084 flags = (flags & ifm->ifi_change) |
1085 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
1086
1087 return flags;
1088 }
1089
1090 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
1091 const struct rtnl_link_stats64 *b)
1092 {
1093 a->rx_packets = b->rx_packets;
1094 a->tx_packets = b->tx_packets;
1095 a->rx_bytes = b->rx_bytes;
1096 a->tx_bytes = b->tx_bytes;
1097 a->rx_errors = b->rx_errors;
1098 a->tx_errors = b->tx_errors;
1099 a->rx_dropped = b->rx_dropped;
1100 a->tx_dropped = b->tx_dropped;
1101
1102 a->multicast = b->multicast;
1103 a->collisions = b->collisions;
1104
1105 a->rx_length_errors = b->rx_length_errors;
1106 a->rx_over_errors = b->rx_over_errors;
1107 a->rx_crc_errors = b->rx_crc_errors;
1108 a->rx_frame_errors = b->rx_frame_errors;
1109 a->rx_fifo_errors = b->rx_fifo_errors;
1110 a->rx_missed_errors = b->rx_missed_errors;
1111
1112 a->tx_aborted_errors = b->tx_aborted_errors;
1113 a->tx_carrier_errors = b->tx_carrier_errors;
1114 a->tx_fifo_errors = b->tx_fifo_errors;
1115 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
1116 a->tx_window_errors = b->tx_window_errors;
1117
1118 a->rx_compressed = b->rx_compressed;
1119 a->tx_compressed = b->tx_compressed;
1120
1121 a->rx_nohandler = b->rx_nohandler;
1122 }
1123
1124 /* All VF info */
1125 static inline int rtnl_vfinfo_size(const struct net_device *dev,
1126 u32 ext_filter_mask)
1127 {
1128 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
1129 int num_vfs = dev_num_vf(dev->dev.parent);
1130 size_t size = nla_total_size(0);
1131 size += num_vfs *
1132 (nla_total_size(0) +
1133 nla_total_size(sizeof(struct ifla_vf_mac)) +
1134 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
1135 nla_total_size(sizeof(struct ifla_vf_vlan)) +
1136 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
1137 nla_total_size(MAX_VLAN_LIST_LEN *
1138 sizeof(struct ifla_vf_vlan_info)) +
1139 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
1140 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
1141 nla_total_size(sizeof(struct ifla_vf_rate)) +
1142 nla_total_size(sizeof(struct ifla_vf_link_state)) +
1143 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
1144 nla_total_size(sizeof(struct ifla_vf_trust)));
1145 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1146 size += num_vfs *
1147 (nla_total_size(0) + /* nest IFLA_VF_STATS */
1148 /* IFLA_VF_STATS_RX_PACKETS */
1149 nla_total_size_64bit(sizeof(__u64)) +
1150 /* IFLA_VF_STATS_TX_PACKETS */
1151 nla_total_size_64bit(sizeof(__u64)) +
1152 /* IFLA_VF_STATS_RX_BYTES */
1153 nla_total_size_64bit(sizeof(__u64)) +
1154 /* IFLA_VF_STATS_TX_BYTES */
1155 nla_total_size_64bit(sizeof(__u64)) +
1156 /* IFLA_VF_STATS_BROADCAST */
1157 nla_total_size_64bit(sizeof(__u64)) +
1158 /* IFLA_VF_STATS_MULTICAST */
1159 nla_total_size_64bit(sizeof(__u64)) +
1160 /* IFLA_VF_STATS_RX_DROPPED */
1161 nla_total_size_64bit(sizeof(__u64)) +
1162 /* IFLA_VF_STATS_TX_DROPPED */
1163 nla_total_size_64bit(sizeof(__u64)));
1164 }
1165 return size;
1166 } else
1167 return 0;
1168 }
1169
1170 static size_t rtnl_port_size(const struct net_device *dev,
1171 u32 ext_filter_mask)
1172 {
1173 size_t port_size = nla_total_size(4) /* PORT_VF */
1174 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
1175 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
1176 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
1177 + nla_total_size(1) /* PROT_VDP_REQUEST */
1178 + nla_total_size(2); /* PORT_VDP_RESPONSE */
1179 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
1180 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
1181 + port_size;
1182 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
1183 + port_size;
1184
1185 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1186 !(ext_filter_mask & RTEXT_FILTER_VF))
1187 return 0;
1188 if (dev_num_vf(dev->dev.parent))
1189 return port_self_size + vf_ports_size +
1190 vf_port_size * dev_num_vf(dev->dev.parent);
1191 else
1192 return port_self_size;
1193 }
1194
1195 static size_t rtnl_xdp_size(void)
1196 {
1197 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
1198 nla_total_size(1) + /* XDP_ATTACHED */
1199 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
1200 nla_total_size(4); /* XDP_<mode>_PROG_ID */
1201
1202 return xdp_size;
1203 }
1204
1205 static size_t rtnl_prop_list_size(const struct net_device *dev)
1206 {
1207 struct netdev_name_node *name_node;
1208 unsigned int cnt = 0;
1209
1210 rcu_read_lock();
1211 list_for_each_entry_rcu(name_node, &dev->name_node->list, list)
1212 cnt++;
1213 rcu_read_unlock();
1214
1215 if (!cnt)
1216 return 0;
1217
1218 return nla_total_size(0) + cnt * nla_total_size(ALTIFNAMSIZ);
1219 }
1220
1221 static size_t rtnl_proto_down_size(const struct net_device *dev)
1222 {
1223 size_t size = nla_total_size(1);
1224
1225 /* Assume dev->proto_down_reason is not zero. */
1226 size += nla_total_size(0) + nla_total_size(4);
1227
1228 return size;
1229 }
1230
1231 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1232 {
1233 size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1234
1235 if (dev->devlink_port)
1236 size += devlink_nl_port_handle_size(dev->devlink_port);
1237
1238 return size;
1239 }
1240
1241 static size_t rtnl_dpll_pin_size(const struct net_device *dev)
1242 {
1243 size_t size = nla_total_size(0); /* nest IFLA_DPLL_PIN */
1244
1245 size += dpll_netdev_pin_handle_size(dev);
1246
1247 return size;
1248 }
1249
1250 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1251 u32 ext_filter_mask)
1252 {
1253 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1254 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1255 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1256 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1257 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1258 + nla_total_size(sizeof(struct rtnl_link_stats))
1259 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1260 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1261 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1262 + nla_total_size(4) /* IFLA_TXQLEN */
1263 + nla_total_size(4) /* IFLA_WEIGHT */
1264 + nla_total_size(4) /* IFLA_MTU */
1265 + nla_total_size(4) /* IFLA_LINK */
1266 + nla_total_size(4) /* IFLA_MASTER */
1267 + nla_total_size(1) /* IFLA_CARRIER */
1268 + nla_total_size(4) /* IFLA_PROMISCUITY */
1269 + nla_total_size(4) /* IFLA_ALLMULTI */
1270 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1271 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1272 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1273 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1274 + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1275 + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1276 + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1277 + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1278 + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1279 + nla_total_size(1) /* IFLA_OPERSTATE */
1280 + nla_total_size(1) /* IFLA_LINKMODE */
1281 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1282 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1283 + nla_total_size(4) /* IFLA_GROUP */
1284 + nla_total_size(ext_filter_mask
1285 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1286 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1287 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1288 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1289 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1290 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1291 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1292 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1293 + rtnl_xdp_size() /* IFLA_XDP */
1294 + nla_total_size(4) /* IFLA_EVENT */
1295 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1296 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1297 + rtnl_proto_down_size(dev) /* proto down */
1298 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1299 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1300 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1301 + nla_total_size(4) /* IFLA_MIN_MTU */
1302 + nla_total_size(4) /* IFLA_MAX_MTU */
1303 + rtnl_prop_list_size(dev)
1304 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1305 + rtnl_devlink_port_size(dev)
1306 + rtnl_dpll_pin_size(dev)
1307 + nla_total_size(8) /* IFLA_MAX_PACING_OFFLOAD_HORIZON */
1308 + 0;
1309 }
1310
1311 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1312 {
1313 struct nlattr *vf_ports;
1314 struct nlattr *vf_port;
1315 int vf;
1316 int err;
1317
1318 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1319 if (!vf_ports)
1320 return -EMSGSIZE;
1321
1322 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1323 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1324 if (!vf_port)
1325 goto nla_put_failure;
1326 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1327 goto nla_put_failure;
1328 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1329 if (err == -EMSGSIZE)
1330 goto nla_put_failure;
1331 if (err) {
1332 nla_nest_cancel(skb, vf_port);
1333 continue;
1334 }
1335 nla_nest_end(skb, vf_port);
1336 }
1337
1338 nla_nest_end(skb, vf_ports);
1339
1340 return 0;
1341
1342 nla_put_failure:
1343 nla_nest_cancel(skb, vf_ports);
1344 return -EMSGSIZE;
1345 }
1346
1347 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1348 {
1349 struct nlattr *port_self;
1350 int err;
1351
1352 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1353 if (!port_self)
1354 return -EMSGSIZE;
1355
1356 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1357 if (err) {
1358 nla_nest_cancel(skb, port_self);
1359 return (err == -EMSGSIZE) ? err : 0;
1360 }
1361
1362 nla_nest_end(skb, port_self);
1363
1364 return 0;
1365 }
1366
1367 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1368 u32 ext_filter_mask)
1369 {
1370 int err;
1371
1372 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1373 !(ext_filter_mask & RTEXT_FILTER_VF))
1374 return 0;
1375
1376 err = rtnl_port_self_fill(skb, dev);
1377 if (err)
1378 return err;
1379
1380 if (dev_num_vf(dev->dev.parent)) {
1381 err = rtnl_vf_ports_fill(skb, dev);
1382 if (err)
1383 return err;
1384 }
1385
1386 return 0;
1387 }
1388
1389 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1390 {
1391 int err;
1392 struct netdev_phys_item_id ppid;
1393
1394 err = dev_get_phys_port_id(dev, &ppid);
1395 if (err) {
1396 if (err == -EOPNOTSUPP)
1397 return 0;
1398 return err;
1399 }
1400
1401 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1402 return -EMSGSIZE;
1403
1404 return 0;
1405 }
1406
1407 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1408 {
1409 char name[IFNAMSIZ];
1410 int err;
1411
1412 err = dev_get_phys_port_name(dev, name, sizeof(name));
1413 if (err) {
1414 if (err == -EOPNOTSUPP)
1415 return 0;
1416 return err;
1417 }
1418
1419 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1420 return -EMSGSIZE;
1421
1422 return 0;
1423 }
1424
1425 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1426 {
1427 struct netdev_phys_item_id ppid = { };
1428 int err;
1429
1430 err = dev_get_port_parent_id(dev, &ppid, false);
1431 if (err) {
1432 if (err == -EOPNOTSUPP)
1433 return 0;
1434 return err;
1435 }
1436
1437 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1438 return -EMSGSIZE;
1439
1440 return 0;
1441 }
1442
1443 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1444 struct net_device *dev)
1445 {
1446 struct rtnl_link_stats64 *sp;
1447 struct nlattr *attr;
1448
1449 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1450 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1451 if (!attr)
1452 return -EMSGSIZE;
1453
1454 sp = nla_data(attr);
1455 dev_get_stats(dev, sp);
1456
1457 attr = nla_reserve(skb, IFLA_STATS,
1458 sizeof(struct rtnl_link_stats));
1459 if (!attr)
1460 return -EMSGSIZE;
1461
1462 copy_rtnl_link_stats(nla_data(attr), sp);
1463
1464 return 0;
1465 }
1466
1467 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1468 struct net_device *dev,
1469 int vfs_num,
1470 u32 ext_filter_mask)
1471 {
1472 struct ifla_vf_rss_query_en vf_rss_query_en;
1473 struct nlattr *vf, *vfstats, *vfvlanlist;
1474 struct ifla_vf_link_state vf_linkstate;
1475 struct ifla_vf_vlan_info vf_vlan_info;
1476 struct ifla_vf_spoofchk vf_spoofchk;
1477 struct ifla_vf_tx_rate vf_tx_rate;
1478 struct ifla_vf_stats vf_stats;
1479 struct ifla_vf_trust vf_trust;
1480 struct ifla_vf_vlan vf_vlan;
1481 struct ifla_vf_rate vf_rate;
1482 struct ifla_vf_mac vf_mac;
1483 struct ifla_vf_broadcast vf_broadcast;
1484 struct ifla_vf_info ivi;
1485 struct ifla_vf_guid node_guid;
1486 struct ifla_vf_guid port_guid;
1487
1488 memset(&ivi, 0, sizeof(ivi));
1489
1490 /* Not all SR-IOV capable drivers support the
1491 * spoofcheck and "RSS query enable" query. Preset to
1492 * -1 so the user space tool can detect that the driver
1493 * didn't report anything.
1494 */
1495 ivi.spoofchk = -1;
1496 ivi.rss_query_en = -1;
1497 ivi.trusted = -1;
1498 /* The default value for VF link state is "auto"
1499 * IFLA_VF_LINK_STATE_AUTO which equals zero
1500 */
1501 ivi.linkstate = 0;
1502 /* VLAN Protocol by default is 802.1Q */
1503 ivi.vlan_proto = htons(ETH_P_8021Q);
1504 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1505 return 0;
1506
1507 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1508 memset(&node_guid, 0, sizeof(node_guid));
1509 memset(&port_guid, 0, sizeof(port_guid));
1510
1511 vf_mac.vf =
1512 vf_vlan.vf =
1513 vf_vlan_info.vf =
1514 vf_rate.vf =
1515 vf_tx_rate.vf =
1516 vf_spoofchk.vf =
1517 vf_linkstate.vf =
1518 vf_rss_query_en.vf =
1519 vf_trust.vf =
1520 node_guid.vf =
1521 port_guid.vf = ivi.vf;
1522
1523 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1524 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1525 vf_vlan.vlan = ivi.vlan;
1526 vf_vlan.qos = ivi.qos;
1527 vf_vlan_info.vlan = ivi.vlan;
1528 vf_vlan_info.qos = ivi.qos;
1529 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1530 vf_tx_rate.rate = ivi.max_tx_rate;
1531 vf_rate.min_tx_rate = ivi.min_tx_rate;
1532 vf_rate.max_tx_rate = ivi.max_tx_rate;
1533 vf_spoofchk.setting = ivi.spoofchk;
1534 vf_linkstate.link_state = ivi.linkstate;
1535 vf_rss_query_en.setting = ivi.rss_query_en;
1536 vf_trust.setting = ivi.trusted;
1537 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1538 if (!vf)
1539 return -EMSGSIZE;
1540 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1541 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1542 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1543 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1544 &vf_rate) ||
1545 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1546 &vf_tx_rate) ||
1547 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1548 &vf_spoofchk) ||
1549 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1550 &vf_linkstate) ||
1551 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1552 sizeof(vf_rss_query_en),
1553 &vf_rss_query_en) ||
1554 nla_put(skb, IFLA_VF_TRUST,
1555 sizeof(vf_trust), &vf_trust))
1556 goto nla_put_vf_failure;
1557
1558 if (dev->netdev_ops->ndo_get_vf_guid &&
1559 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1560 &port_guid)) {
1561 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1562 &node_guid) ||
1563 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1564 &port_guid))
1565 goto nla_put_vf_failure;
1566 }
1567 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1568 if (!vfvlanlist)
1569 goto nla_put_vf_failure;
1570 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1571 &vf_vlan_info)) {
1572 nla_nest_cancel(skb, vfvlanlist);
1573 goto nla_put_vf_failure;
1574 }
1575 nla_nest_end(skb, vfvlanlist);
1576 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1577 memset(&vf_stats, 0, sizeof(vf_stats));
1578 if (dev->netdev_ops->ndo_get_vf_stats)
1579 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1580 &vf_stats);
1581 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1582 if (!vfstats)
1583 goto nla_put_vf_failure;
1584 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1585 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1586 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1587 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1588 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1589 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1590 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1591 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1592 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1593 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1594 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1595 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1596 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1597 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1598 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1599 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1600 nla_nest_cancel(skb, vfstats);
1601 goto nla_put_vf_failure;
1602 }
1603 nla_nest_end(skb, vfstats);
1604 }
1605 nla_nest_end(skb, vf);
1606 return 0;
1607
1608 nla_put_vf_failure:
1609 nla_nest_cancel(skb, vf);
1610 return -EMSGSIZE;
1611 }
1612
1613 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1614 struct net_device *dev,
1615 u32 ext_filter_mask)
1616 {
1617 struct nlattr *vfinfo;
1618 int i, num_vfs;
1619
1620 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1621 return 0;
1622
1623 num_vfs = dev_num_vf(dev->dev.parent);
1624 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1625 return -EMSGSIZE;
1626
1627 if (!dev->netdev_ops->ndo_get_vf_config)
1628 return 0;
1629
1630 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1631 if (!vfinfo)
1632 return -EMSGSIZE;
1633
1634 for (i = 0; i < num_vfs; i++) {
1635 if (rtnl_fill_vfinfo(skb, dev, i, ext_filter_mask)) {
1636 nla_nest_cancel(skb, vfinfo);
1637 return -EMSGSIZE;
1638 }
1639 }
1640
1641 nla_nest_end(skb, vfinfo);
1642 return 0;
1643 }
1644
1645 static int rtnl_fill_link_ifmap(struct sk_buff *skb,
1646 const struct net_device *dev)
1647 {
1648 struct rtnl_link_ifmap map;
1649
1650 memset(&map, 0, sizeof(map));
1651 map.mem_start = READ_ONCE(dev->mem_start);
1652 map.mem_end = READ_ONCE(dev->mem_end);
1653 map.base_addr = READ_ONCE(dev->base_addr);
1654 map.irq = READ_ONCE(dev->irq);
1655 map.dma = READ_ONCE(dev->dma);
1656 map.port = READ_ONCE(dev->if_port);
1657
1658 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1659 return -EMSGSIZE;
1660
1661 return 0;
1662 }
1663
1664 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1665 {
1666 const struct bpf_prog *generic_xdp_prog;
1667 u32 res = 0;
1668
1669 rcu_read_lock();
1670 generic_xdp_prog = rcu_dereference(dev->xdp_prog);
1671 if (generic_xdp_prog)
1672 res = generic_xdp_prog->aux->id;
1673 rcu_read_unlock();
1674
1675 return res;
1676 }
1677
1678 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1679 {
1680 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1681 }
1682
1683 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1684 {
1685 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1686 }
1687
1688 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1689 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1690 u32 (*get_prog_id)(struct net_device *dev))
1691 {
1692 u32 curr_id;
1693 int err;
1694
1695 curr_id = get_prog_id(dev);
1696 if (!curr_id)
1697 return 0;
1698
1699 *prog_id = curr_id;
1700 err = nla_put_u32(skb, attr, curr_id);
1701 if (err)
1702 return err;
1703
1704 if (*mode != XDP_ATTACHED_NONE)
1705 *mode = XDP_ATTACHED_MULTI;
1706 else
1707 *mode = tgt_mode;
1708
1709 return 0;
1710 }
1711
1712 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1713 {
1714 struct nlattr *xdp;
1715 u32 prog_id;
1716 int err;
1717 u8 mode;
1718
1719 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1720 if (!xdp)
1721 return -EMSGSIZE;
1722
1723 prog_id = 0;
1724 mode = XDP_ATTACHED_NONE;
1725 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1726 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1727 if (err)
1728 goto err_cancel;
1729 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1730 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1731 if (err)
1732 goto err_cancel;
1733 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1734 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1735 if (err)
1736 goto err_cancel;
1737
1738 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1739 if (err)
1740 goto err_cancel;
1741
1742 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1743 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1744 if (err)
1745 goto err_cancel;
1746 }
1747
1748 nla_nest_end(skb, xdp);
1749 return 0;
1750
1751 err_cancel:
1752 nla_nest_cancel(skb, xdp);
1753 return err;
1754 }
1755
1756 static u32 rtnl_get_event(unsigned long event)
1757 {
1758 u32 rtnl_event_type = IFLA_EVENT_NONE;
1759
1760 switch (event) {
1761 case NETDEV_REBOOT:
1762 rtnl_event_type = IFLA_EVENT_REBOOT;
1763 break;
1764 case NETDEV_FEAT_CHANGE:
1765 rtnl_event_type = IFLA_EVENT_FEATURES;
1766 break;
1767 case NETDEV_BONDING_FAILOVER:
1768 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1769 break;
1770 case NETDEV_NOTIFY_PEERS:
1771 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1772 break;
1773 case NETDEV_RESEND_IGMP:
1774 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1775 break;
1776 case NETDEV_CHANGEINFODATA:
1777 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1778 break;
1779 default:
1780 break;
1781 }
1782
1783 return rtnl_event_type;
1784 }
1785
1786 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1787 {
1788 const struct net_device *upper_dev;
1789 int ret = 0;
1790
1791 rcu_read_lock();
1792
1793 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1794 if (upper_dev)
1795 ret = nla_put_u32(skb, IFLA_MASTER,
1796 READ_ONCE(upper_dev->ifindex));
1797
1798 rcu_read_unlock();
1799 return ret;
1800 }
1801
1802 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1803 bool force)
1804 {
1805 int iflink = dev_get_iflink(dev);
1806
1807 if (force || READ_ONCE(dev->ifindex) != iflink)
1808 return nla_put_u32(skb, IFLA_LINK, iflink);
1809
1810 return 0;
1811 }
1812
1813 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1814 struct net_device *dev)
1815 {
1816 char buf[IFALIASZ];
1817 int ret;
1818
1819 ret = dev_get_alias(dev, buf, sizeof(buf));
1820 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1821 }
1822
1823 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1824 const struct net_device *dev,
1825 struct net *src_net, gfp_t gfp)
1826 {
1827 bool put_iflink = false;
1828
1829 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1830 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1831
1832 if (!net_eq(dev_net(dev), link_net)) {
1833 int id = peernet2id_alloc(src_net, link_net, gfp);
1834
1835 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1836 return -EMSGSIZE;
1837
1838 put_iflink = true;
1839 }
1840 }
1841
1842 return nla_put_iflink(skb, dev, put_iflink);
1843 }
1844
1845 static int rtnl_fill_link_af(struct sk_buff *skb,
1846 const struct net_device *dev,
1847 u32 ext_filter_mask)
1848 {
1849 const struct rtnl_af_ops *af_ops;
1850 struct nlattr *af_spec;
1851
1852 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1853 if (!af_spec)
1854 return -EMSGSIZE;
1855
1856 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1857 struct nlattr *af;
1858 int err;
1859
1860 if (!af_ops->fill_link_af)
1861 continue;
1862
1863 af = nla_nest_start_noflag(skb, af_ops->family);
1864 if (!af)
1865 return -EMSGSIZE;
1866
1867 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1868 /*
1869 * Caller may return ENODATA to indicate that there
1870 * was no data to be dumped. This is not an error, it
1871 * means we should trim the attribute header and
1872 * continue.
1873 */
1874 if (err == -ENODATA)
1875 nla_nest_cancel(skb, af);
1876 else if (err < 0)
1877 return -EMSGSIZE;
1878
1879 nla_nest_end(skb, af);
1880 }
1881
1882 nla_nest_end(skb, af_spec);
1883 return 0;
1884 }
1885
1886 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1887 const struct net_device *dev)
1888 {
1889 struct netdev_name_node *name_node;
1890 int count = 0;
1891
1892 list_for_each_entry_rcu(name_node, &dev->name_node->list, list) {
1893 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1894 return -EMSGSIZE;
1895 count++;
1896 }
1897 return count;
1898 }
1899
1900 /* RCU protected. */
1901 static int rtnl_fill_prop_list(struct sk_buff *skb,
1902 const struct net_device *dev)
1903 {
1904 struct nlattr *prop_list;
1905 int ret;
1906
1907 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1908 if (!prop_list)
1909 return -EMSGSIZE;
1910
1911 ret = rtnl_fill_alt_ifnames(skb, dev);
1912 if (ret <= 0)
1913 goto nest_cancel;
1914
1915 nla_nest_end(skb, prop_list);
1916 return 0;
1917
1918 nest_cancel:
1919 nla_nest_cancel(skb, prop_list);
1920 return ret;
1921 }
1922
1923 static int rtnl_fill_proto_down(struct sk_buff *skb,
1924 const struct net_device *dev)
1925 {
1926 struct nlattr *pr;
1927 u32 preason;
1928
1929 if (nla_put_u8(skb, IFLA_PROTO_DOWN, READ_ONCE(dev->proto_down)))
1930 goto nla_put_failure;
1931
1932 preason = READ_ONCE(dev->proto_down_reason);
1933 if (!preason)
1934 return 0;
1935
1936 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1937 if (!pr)
1938 return -EMSGSIZE;
1939
1940 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1941 nla_nest_cancel(skb, pr);
1942 goto nla_put_failure;
1943 }
1944
1945 nla_nest_end(skb, pr);
1946 return 0;
1947
1948 nla_put_failure:
1949 return -EMSGSIZE;
1950 }
1951
1952 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1953 const struct net_device *dev)
1954 {
1955 struct nlattr *devlink_port_nest;
1956 int ret;
1957
1958 devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1959 if (!devlink_port_nest)
1960 return -EMSGSIZE;
1961
1962 if (dev->devlink_port) {
1963 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1964 if (ret < 0)
1965 goto nest_cancel;
1966 }
1967
1968 nla_nest_end(skb, devlink_port_nest);
1969 return 0;
1970
1971 nest_cancel:
1972 nla_nest_cancel(skb, devlink_port_nest);
1973 return ret;
1974 }
1975
1976 static int rtnl_fill_dpll_pin(struct sk_buff *skb,
1977 const struct net_device *dev)
1978 {
1979 struct nlattr *dpll_pin_nest;
1980 int ret;
1981
1982 dpll_pin_nest = nla_nest_start(skb, IFLA_DPLL_PIN);
1983 if (!dpll_pin_nest)
1984 return -EMSGSIZE;
1985
1986 ret = dpll_netdev_add_pin_handle(skb, dev);
1987 if (ret < 0)
1988 goto nest_cancel;
1989
1990 nla_nest_end(skb, dpll_pin_nest);
1991 return 0;
1992
1993 nest_cancel:
1994 nla_nest_cancel(skb, dpll_pin_nest);
1995 return ret;
1996 }
1997
1998 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1999 struct net_device *dev, struct net *src_net,
2000 int type, u32 pid, u32 seq, u32 change,
2001 unsigned int flags, u32 ext_filter_mask,
2002 u32 event, int *new_nsid, int new_ifindex,
2003 int tgt_netnsid, gfp_t gfp)
2004 {
2005 char devname[IFNAMSIZ];
2006 struct ifinfomsg *ifm;
2007 struct nlmsghdr *nlh;
2008 struct Qdisc *qdisc;
2009
2010 ASSERT_RTNL();
2011 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
2012 if (nlh == NULL)
2013 return -EMSGSIZE;
2014
2015 ifm = nlmsg_data(nlh);
2016 ifm->ifi_family = AF_UNSPEC;
2017 ifm->__ifi_pad = 0;
2018 ifm->ifi_type = READ_ONCE(dev->type);
2019 ifm->ifi_index = READ_ONCE(dev->ifindex);
2020 ifm->ifi_flags = dev_get_flags(dev);
2021 ifm->ifi_change = change;
2022
2023 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
2024 goto nla_put_failure;
2025
2026 netdev_copy_name(dev, devname);
2027 if (nla_put_string(skb, IFLA_IFNAME, devname))
2028 goto nla_put_failure;
2029
2030 if (nla_put_u32(skb, IFLA_TXQLEN, READ_ONCE(dev->tx_queue_len)) ||
2031 nla_put_u8(skb, IFLA_OPERSTATE,
2032 netif_running(dev) ? READ_ONCE(dev->operstate) :
2033 IF_OPER_DOWN) ||
2034 nla_put_u8(skb, IFLA_LINKMODE, READ_ONCE(dev->link_mode)) ||
2035 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
2036 nla_put_u32(skb, IFLA_MIN_MTU, READ_ONCE(dev->min_mtu)) ||
2037 nla_put_u32(skb, IFLA_MAX_MTU, READ_ONCE(dev->max_mtu)) ||
2038 nla_put_u32(skb, IFLA_GROUP, READ_ONCE(dev->group)) ||
2039 nla_put_u32(skb, IFLA_PROMISCUITY, READ_ONCE(dev->promiscuity)) ||
2040 nla_put_u32(skb, IFLA_ALLMULTI, READ_ONCE(dev->allmulti)) ||
2041 nla_put_u32(skb, IFLA_NUM_TX_QUEUES,
2042 READ_ONCE(dev->num_tx_queues)) ||
2043 nla_put_u32(skb, IFLA_GSO_MAX_SEGS,
2044 READ_ONCE(dev->gso_max_segs)) ||
2045 nla_put_u32(skb, IFLA_GSO_MAX_SIZE,
2046 READ_ONCE(dev->gso_max_size)) ||
2047 nla_put_u32(skb, IFLA_GRO_MAX_SIZE,
2048 READ_ONCE(dev->gro_max_size)) ||
2049 nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE,
2050 READ_ONCE(dev->gso_ipv4_max_size)) ||
2051 nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE,
2052 READ_ONCE(dev->gro_ipv4_max_size)) ||
2053 nla_put_u32(skb, IFLA_TSO_MAX_SIZE,
2054 READ_ONCE(dev->tso_max_size)) ||
2055 nla_put_u32(skb, IFLA_TSO_MAX_SEGS,
2056 READ_ONCE(dev->tso_max_segs)) ||
2057 nla_put_uint(skb, IFLA_MAX_PACING_OFFLOAD_HORIZON,
2058 READ_ONCE(dev->max_pacing_offload_horizon)) ||
2059 #ifdef CONFIG_RPS
2060 nla_put_u32(skb, IFLA_NUM_RX_QUEUES,
2061 READ_ONCE(dev->num_rx_queues)) ||
2062 #endif
2063 put_master_ifindex(skb, dev) ||
2064 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
2065 nla_put_ifalias(skb, dev) ||
2066 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
2067 atomic_read(&dev->carrier_up_count) +
2068 atomic_read(&dev->carrier_down_count)) ||
2069 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
2070 atomic_read(&dev->carrier_up_count)) ||
2071 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
2072 atomic_read(&dev->carrier_down_count)))
2073 goto nla_put_failure;
2074
2075 if (rtnl_fill_proto_down(skb, dev))
2076 goto nla_put_failure;
2077
2078 if (event != IFLA_EVENT_NONE) {
2079 if (nla_put_u32(skb, IFLA_EVENT, event))
2080 goto nla_put_failure;
2081 }
2082
2083 if (dev->addr_len) {
2084 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
2085 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
2086 goto nla_put_failure;
2087 }
2088
2089 if (rtnl_phys_port_id_fill(skb, dev))
2090 goto nla_put_failure;
2091
2092 if (rtnl_phys_port_name_fill(skb, dev))
2093 goto nla_put_failure;
2094
2095 if (rtnl_phys_switch_id_fill(skb, dev))
2096 goto nla_put_failure;
2097
2098 if (rtnl_fill_stats(skb, dev))
2099 goto nla_put_failure;
2100
2101 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
2102 goto nla_put_failure;
2103
2104 if (rtnl_port_fill(skb, dev, ext_filter_mask))
2105 goto nla_put_failure;
2106
2107 if (rtnl_xdp_fill(skb, dev))
2108 goto nla_put_failure;
2109
2110 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
2111 if (rtnl_link_fill(skb, dev) < 0)
2112 goto nla_put_failure;
2113 }
2114
2115 if (new_nsid &&
2116 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
2117 goto nla_put_failure;
2118 if (new_ifindex &&
2119 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
2120 goto nla_put_failure;
2121
2122 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
2123 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
2124 goto nla_put_failure;
2125
2126 rcu_read_lock();
2127 if (rtnl_fill_link_netnsid(skb, dev, src_net, GFP_ATOMIC))
2128 goto nla_put_failure_rcu;
2129 qdisc = rcu_dereference(dev->qdisc);
2130 if (qdisc && nla_put_string(skb, IFLA_QDISC, qdisc->ops->id))
2131 goto nla_put_failure_rcu;
2132 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
2133 goto nla_put_failure_rcu;
2134 if (rtnl_fill_link_ifmap(skb, dev))
2135 goto nla_put_failure_rcu;
2136 if (rtnl_fill_prop_list(skb, dev))
2137 goto nla_put_failure_rcu;
2138 rcu_read_unlock();
2139
2140 if (dev->dev.parent &&
2141 nla_put_string(skb, IFLA_PARENT_DEV_NAME,
2142 dev_name(dev->dev.parent)))
2143 goto nla_put_failure;
2144
2145 if (dev->dev.parent && dev->dev.parent->bus &&
2146 nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
2147 dev->dev.parent->bus->name))
2148 goto nla_put_failure;
2149
2150 if (rtnl_fill_devlink_port(skb, dev))
2151 goto nla_put_failure;
2152
2153 if (rtnl_fill_dpll_pin(skb, dev))
2154 goto nla_put_failure;
2155
2156 nlmsg_end(skb, nlh);
2157 return 0;
2158
2159 nla_put_failure_rcu:
2160 rcu_read_unlock();
2161 nla_put_failure:
2162 nlmsg_cancel(skb, nlh);
2163 return -EMSGSIZE;
2164 }
2165
2166 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
2167 [IFLA_UNSPEC] = { .strict_start_type = IFLA_DPLL_PIN },
2168 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
2169 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
2170 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
2171 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
2172 [IFLA_MTU] = { .type = NLA_U32 },
2173 [IFLA_LINK] = { .type = NLA_U32 },
2174 [IFLA_MASTER] = { .type = NLA_U32 },
2175 [IFLA_CARRIER] = { .type = NLA_U8 },
2176 [IFLA_TXQLEN] = { .type = NLA_U32 },
2177 [IFLA_WEIGHT] = { .type = NLA_U32 },
2178 [IFLA_OPERSTATE] = { .type = NLA_U8 },
2179 [IFLA_LINKMODE] = { .type = NLA_U8 },
2180 [IFLA_LINKINFO] = { .type = NLA_NESTED },
2181 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
2182 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
2183 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
2184 * allow 0-length string (needed to remove an alias).
2185 */
2186 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
2187 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
2188 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
2189 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
2190 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
2191 [IFLA_EXT_MASK] = { .type = NLA_U32 },
2192 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
2193 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
2194 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
2195 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
2196 [IFLA_GSO_MAX_SIZE] = NLA_POLICY_MIN(NLA_U32, MAX_TCP_HEADER + 1),
2197 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2198 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
2199 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2200 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
2201 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
2202 [IFLA_XDP] = { .type = NLA_NESTED },
2203 [IFLA_EVENT] = { .type = NLA_U32 },
2204 [IFLA_GROUP] = { .type = NLA_U32 },
2205 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
2206 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
2207 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
2208 [IFLA_MIN_MTU] = { .type = NLA_U32 },
2209 [IFLA_MAX_MTU] = { .type = NLA_U32 },
2210 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
2211 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
2212 .len = ALTIFNAMSIZ - 1 },
2213 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
2214 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
2215 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
2216 [IFLA_PARENT_DEV_NAME] = { .type = NLA_NUL_STRING },
2217 [IFLA_GRO_MAX_SIZE] = { .type = NLA_U32 },
2218 [IFLA_TSO_MAX_SIZE] = { .type = NLA_REJECT },
2219 [IFLA_TSO_MAX_SEGS] = { .type = NLA_REJECT },
2220 [IFLA_ALLMULTI] = { .type = NLA_REJECT },
2221 [IFLA_GSO_IPV4_MAX_SIZE] = NLA_POLICY_MIN(NLA_U32, MAX_TCP_HEADER + 1),
2222 [IFLA_GRO_IPV4_MAX_SIZE] = { .type = NLA_U32 },
2223 };
2224
2225 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
2226 [IFLA_INFO_KIND] = { .type = NLA_STRING },
2227 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
2228 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
2229 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
2230 };
2231
2232 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
2233 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
2234 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
2235 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
2236 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
2237 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
2238 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
2239 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
2240 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
2241 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
2242 [IFLA_VF_STATS] = { .type = NLA_NESTED },
2243 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
2244 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
2245 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
2246 };
2247
2248 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2249 [IFLA_PORT_VF] = { .type = NLA_U32 },
2250 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
2251 .len = PORT_PROFILE_MAX },
2252 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2253 .len = PORT_UUID_MAX },
2254 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
2255 .len = PORT_UUID_MAX },
2256 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
2257 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
2258
2259 /* Unused, but we need to keep it here since user space could
2260 * fill it. It's also broken with regard to NLA_BINARY use in
2261 * combination with structs.
2262 */
2263 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
2264 .len = sizeof(struct ifla_port_vsi) },
2265 };
2266
2267 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2268 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2269 [IFLA_XDP_FD] = { .type = NLA_S32 },
2270 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
2271 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
2272 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
2273 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
2274 };
2275
2276 static struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla,
2277 int *ops_srcu_index)
2278 {
2279 struct nlattr *linfo[IFLA_INFO_MAX + 1];
2280 struct rtnl_link_ops *ops = NULL;
2281
2282 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2283 return NULL;
2284
2285 if (linfo[IFLA_INFO_KIND]) {
2286 char kind[MODULE_NAME_LEN];
2287
2288 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2289 ops = rtnl_link_ops_get(kind, ops_srcu_index);
2290 }
2291
2292 return ops;
2293 }
2294
2295 static bool link_master_filtered(struct net_device *dev, int master_idx)
2296 {
2297 struct net_device *master;
2298
2299 if (!master_idx)
2300 return false;
2301
2302 master = netdev_master_upper_dev_get(dev);
2303
2304 /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2305 * another invalid value for ifindex to denote "no master".
2306 */
2307 if (master_idx == -1)
2308 return !!master;
2309
2310 if (!master || master->ifindex != master_idx)
2311 return true;
2312
2313 return false;
2314 }
2315
2316 static bool link_kind_filtered(const struct net_device *dev,
2317 const struct rtnl_link_ops *kind_ops)
2318 {
2319 if (kind_ops && dev->rtnl_link_ops != kind_ops)
2320 return true;
2321
2322 return false;
2323 }
2324
2325 static bool link_dump_filtered(struct net_device *dev,
2326 int master_idx,
2327 const struct rtnl_link_ops *kind_ops)
2328 {
2329 if (link_master_filtered(dev, master_idx) ||
2330 link_kind_filtered(dev, kind_ops))
2331 return true;
2332
2333 return false;
2334 }
2335
2336 /**
2337 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2338 * @sk: netlink socket
2339 * @netnsid: network namespace identifier
2340 *
2341 * Returns the network namespace identified by netnsid on success or an error
2342 * pointer on failure.
2343 */
2344 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2345 {
2346 struct net *net;
2347
2348 net = get_net_ns_by_id(sock_net(sk), netnsid);
2349 if (!net)
2350 return ERR_PTR(-EINVAL);
2351
2352 /* For now, the caller is required to have CAP_NET_ADMIN in
2353 * the user namespace owning the target net ns.
2354 */
2355 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2356 put_net(net);
2357 return ERR_PTR(-EACCES);
2358 }
2359 return net;
2360 }
2361 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2362
2363 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2364 bool strict_check, struct nlattr **tb,
2365 struct netlink_ext_ack *extack)
2366 {
2367 int hdrlen;
2368
2369 if (strict_check) {
2370 struct ifinfomsg *ifm;
2371
2372 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2373 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2374 return -EINVAL;
2375 }
2376
2377 ifm = nlmsg_data(nlh);
2378 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2379 ifm->ifi_change) {
2380 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2381 return -EINVAL;
2382 }
2383 if (ifm->ifi_index) {
2384 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2385 return -EINVAL;
2386 }
2387
2388 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2389 IFLA_MAX, ifla_policy,
2390 extack);
2391 }
2392
2393 /* A hack to preserve kernel<->userspace interface.
2394 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2395 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2396 * what iproute2 < v3.9.0 used.
2397 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2398 * attribute, its netlink message is shorter than struct ifinfomsg.
2399 */
2400 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2401 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2402
2403 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2404 extack);
2405 }
2406
2407 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2408 {
2409 struct netlink_ext_ack *extack = cb->extack;
2410 struct rtnl_link_ops *kind_ops = NULL;
2411 const struct nlmsghdr *nlh = cb->nlh;
2412 struct net *net = sock_net(skb->sk);
2413 unsigned int flags = NLM_F_MULTI;
2414 struct nlattr *tb[IFLA_MAX+1];
2415 struct {
2416 unsigned long ifindex;
2417 } *ctx = (void *)cb->ctx;
2418 struct net *tgt_net = net;
2419 u32 ext_filter_mask = 0;
2420 struct net_device *dev;
2421 int ops_srcu_index;
2422 int master_idx = 0;
2423 int netnsid = -1;
2424 int err, i;
2425
2426 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2427 if (err < 0) {
2428 if (cb->strict_check)
2429 return err;
2430
2431 goto walk_entries;
2432 }
2433
2434 for (i = 0; i <= IFLA_MAX; ++i) {
2435 if (!tb[i])
2436 continue;
2437
2438 /* new attributes should only be added with strict checking */
2439 switch (i) {
2440 case IFLA_TARGET_NETNSID:
2441 netnsid = nla_get_s32(tb[i]);
2442 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2443 if (IS_ERR(tgt_net)) {
2444 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2445 err = PTR_ERR(tgt_net);
2446 netnsid = -1;
2447 goto out;
2448 }
2449 break;
2450 case IFLA_EXT_MASK:
2451 ext_filter_mask = nla_get_u32(tb[i]);
2452 break;
2453 case IFLA_MASTER:
2454 master_idx = nla_get_u32(tb[i]);
2455 break;
2456 case IFLA_LINKINFO:
2457 kind_ops = linkinfo_to_kind_ops(tb[i], &ops_srcu_index);
2458 break;
2459 default:
2460 if (cb->strict_check) {
2461 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2462 err = -EINVAL;
2463 goto out;
2464 }
2465 }
2466 }
2467
2468 if (master_idx || kind_ops)
2469 flags |= NLM_F_DUMP_FILTERED;
2470
2471 walk_entries:
2472 err = 0;
2473 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
2474 if (link_dump_filtered(dev, master_idx, kind_ops))
2475 continue;
2476 err = rtnl_fill_ifinfo(skb, dev, net, RTM_NEWLINK,
2477 NETLINK_CB(cb->skb).portid,
2478 nlh->nlmsg_seq, 0, flags,
2479 ext_filter_mask, 0, NULL, 0,
2480 netnsid, GFP_KERNEL);
2481 if (err < 0)
2482 break;
2483 }
2484
2485
2486 cb->seq = tgt_net->dev_base_seq;
2487 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2488
2489 out:
2490
2491 if (kind_ops)
2492 rtnl_link_ops_put(kind_ops, ops_srcu_index);
2493 if (netnsid >= 0)
2494 put_net(tgt_net);
2495
2496 return err;
2497 }
2498
2499 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2500 struct netlink_ext_ack *exterr)
2501 {
2502 const struct ifinfomsg *ifmp;
2503 const struct nlattr *attrs;
2504 size_t len;
2505
2506 ifmp = nla_data(nla_peer);
2507 attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2508 len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2509
2510 if (ifmp->ifi_index < 0) {
2511 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2512 "ifindex can't be negative");
2513 return -EINVAL;
2514 }
2515
2516 return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2517 exterr);
2518 }
2519 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2520
2521 static struct net *rtnl_link_get_net_ifla(struct nlattr *tb[])
2522 {
2523 struct net *net = NULL;
2524
2525 /* Examine the link attributes and figure out which
2526 * network namespace we are talking about.
2527 */
2528 if (tb[IFLA_NET_NS_PID])
2529 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2530 else if (tb[IFLA_NET_NS_FD])
2531 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2532
2533 return net;
2534 }
2535
2536 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2537 {
2538 struct net *net = rtnl_link_get_net_ifla(tb);
2539
2540 if (!net)
2541 net = get_net(src_net);
2542
2543 return net;
2544 }
2545 EXPORT_SYMBOL(rtnl_link_get_net);
2546
2547 /* Figure out which network namespace we are talking about by
2548 * examining the link attributes in the following order:
2549 *
2550 * 1. IFLA_NET_NS_PID
2551 * 2. IFLA_NET_NS_FD
2552 * 3. IFLA_TARGET_NETNSID
2553 */
2554 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2555 struct nlattr *tb[])
2556 {
2557 struct net *net;
2558
2559 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2560 return rtnl_link_get_net(src_net, tb);
2561
2562 if (!tb[IFLA_TARGET_NETNSID])
2563 return get_net(src_net);
2564
2565 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2566 if (!net)
2567 return ERR_PTR(-EINVAL);
2568
2569 return net;
2570 }
2571
2572 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2573 struct net *src_net,
2574 struct nlattr *tb[], int cap)
2575 {
2576 struct net *net;
2577
2578 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2579 if (IS_ERR(net))
2580 return net;
2581
2582 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2583 put_net(net);
2584 return ERR_PTR(-EPERM);
2585 }
2586
2587 return net;
2588 }
2589
2590 /* Verify that rtnetlink requests do not pass additional properties
2591 * potentially referring to different network namespaces.
2592 */
2593 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2594 struct netlink_ext_ack *extack,
2595 bool netns_id_only)
2596 {
2597
2598 if (netns_id_only) {
2599 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2600 return 0;
2601
2602 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2603 return -EOPNOTSUPP;
2604 }
2605
2606 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2607 goto invalid_attr;
2608
2609 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2610 goto invalid_attr;
2611
2612 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2613 goto invalid_attr;
2614
2615 return 0;
2616
2617 invalid_attr:
2618 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2619 return -EINVAL;
2620 }
2621
2622 static int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2623 int max_tx_rate)
2624 {
2625 const struct net_device_ops *ops = dev->netdev_ops;
2626
2627 if (!ops->ndo_set_vf_rate)
2628 return -EOPNOTSUPP;
2629 if (max_tx_rate && max_tx_rate < min_tx_rate)
2630 return -EINVAL;
2631
2632 return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2633 }
2634
2635 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2636 struct netlink_ext_ack *extack)
2637 {
2638 if (tb[IFLA_ADDRESS] &&
2639 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2640 return -EINVAL;
2641
2642 if (tb[IFLA_BROADCAST] &&
2643 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2644 return -EINVAL;
2645
2646 if (tb[IFLA_GSO_MAX_SIZE] &&
2647 nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2648 NL_SET_ERR_MSG(extack, "too big gso_max_size");
2649 return -EINVAL;
2650 }
2651
2652 if (tb[IFLA_GSO_MAX_SEGS] &&
2653 (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2654 nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2655 NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2656 return -EINVAL;
2657 }
2658
2659 if (tb[IFLA_GRO_MAX_SIZE] &&
2660 nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2661 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2662 return -EINVAL;
2663 }
2664
2665 if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2666 nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2667 NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2668 return -EINVAL;
2669 }
2670
2671 if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2672 nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2673 NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2674 return -EINVAL;
2675 }
2676
2677 if (tb[IFLA_AF_SPEC]) {
2678 struct nlattr *af;
2679 int rem, err;
2680
2681 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2682 struct rtnl_af_ops *af_ops;
2683 int af_ops_srcu_index;
2684
2685 af_ops = rtnl_af_lookup(nla_type(af), &af_ops_srcu_index);
2686 if (!af_ops)
2687 return -EAFNOSUPPORT;
2688
2689 if (!af_ops->set_link_af)
2690 err = -EOPNOTSUPP;
2691 else if (af_ops->validate_link_af)
2692 err = af_ops->validate_link_af(dev, af, extack);
2693 else
2694 err = 0;
2695
2696 rtnl_af_put(af_ops, af_ops_srcu_index);
2697
2698 if (err < 0)
2699 return err;
2700 }
2701 }
2702
2703 return 0;
2704 }
2705
2706 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2707 int guid_type)
2708 {
2709 const struct net_device_ops *ops = dev->netdev_ops;
2710
2711 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2712 }
2713
2714 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2715 {
2716 if (dev->type != ARPHRD_INFINIBAND)
2717 return -EOPNOTSUPP;
2718
2719 return handle_infiniband_guid(dev, ivt, guid_type);
2720 }
2721
2722 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2723 {
2724 const struct net_device_ops *ops = dev->netdev_ops;
2725 int err = -EINVAL;
2726
2727 if (tb[IFLA_VF_MAC]) {
2728 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2729
2730 if (ivm->vf >= INT_MAX)
2731 return -EINVAL;
2732 err = -EOPNOTSUPP;
2733 if (ops->ndo_set_vf_mac)
2734 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2735 ivm->mac);
2736 if (err < 0)
2737 return err;
2738 }
2739
2740 if (tb[IFLA_VF_VLAN]) {
2741 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2742
2743 if (ivv->vf >= INT_MAX)
2744 return -EINVAL;
2745 err = -EOPNOTSUPP;
2746 if (ops->ndo_set_vf_vlan)
2747 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2748 ivv->qos,
2749 htons(ETH_P_8021Q));
2750 if (err < 0)
2751 return err;
2752 }
2753
2754 if (tb[IFLA_VF_VLAN_LIST]) {
2755 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2756 struct nlattr *attr;
2757 int rem, len = 0;
2758
2759 err = -EOPNOTSUPP;
2760 if (!ops->ndo_set_vf_vlan)
2761 return err;
2762
2763 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2764 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2765 nla_len(attr) < sizeof(struct ifla_vf_vlan_info)) {
2766 return -EINVAL;
2767 }
2768 if (len >= MAX_VLAN_LIST_LEN)
2769 return -EOPNOTSUPP;
2770 ivvl[len] = nla_data(attr);
2771
2772 len++;
2773 }
2774 if (len == 0)
2775 return -EINVAL;
2776
2777 if (ivvl[0]->vf >= INT_MAX)
2778 return -EINVAL;
2779 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2780 ivvl[0]->qos, ivvl[0]->vlan_proto);
2781 if (err < 0)
2782 return err;
2783 }
2784
2785 if (tb[IFLA_VF_TX_RATE]) {
2786 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2787 struct ifla_vf_info ivf;
2788
2789 if (ivt->vf >= INT_MAX)
2790 return -EINVAL;
2791 err = -EOPNOTSUPP;
2792 if (ops->ndo_get_vf_config)
2793 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2794 if (err < 0)
2795 return err;
2796
2797 err = rtnl_set_vf_rate(dev, ivt->vf,
2798 ivf.min_tx_rate, ivt->rate);
2799 if (err < 0)
2800 return err;
2801 }
2802
2803 if (tb[IFLA_VF_RATE]) {
2804 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2805
2806 if (ivt->vf >= INT_MAX)
2807 return -EINVAL;
2808
2809 err = rtnl_set_vf_rate(dev, ivt->vf,
2810 ivt->min_tx_rate, ivt->max_tx_rate);
2811 if (err < 0)
2812 return err;
2813 }
2814
2815 if (tb[IFLA_VF_SPOOFCHK]) {
2816 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2817
2818 if (ivs->vf >= INT_MAX)
2819 return -EINVAL;
2820 err = -EOPNOTSUPP;
2821 if (ops->ndo_set_vf_spoofchk)
2822 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2823 ivs->setting);
2824 if (err < 0)
2825 return err;
2826 }
2827
2828 if (tb[IFLA_VF_LINK_STATE]) {
2829 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2830
2831 if (ivl->vf >= INT_MAX)
2832 return -EINVAL;
2833 err = -EOPNOTSUPP;
2834 if (ops->ndo_set_vf_link_state)
2835 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2836 ivl->link_state);
2837 if (err < 0)
2838 return err;
2839 }
2840
2841 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2842 struct ifla_vf_rss_query_en *ivrssq_en;
2843
2844 err = -EOPNOTSUPP;
2845 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2846 if (ivrssq_en->vf >= INT_MAX)
2847 return -EINVAL;
2848 if (ops->ndo_set_vf_rss_query_en)
2849 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2850 ivrssq_en->setting);
2851 if (err < 0)
2852 return err;
2853 }
2854
2855 if (tb[IFLA_VF_TRUST]) {
2856 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2857
2858 if (ivt->vf >= INT_MAX)
2859 return -EINVAL;
2860 err = -EOPNOTSUPP;
2861 if (ops->ndo_set_vf_trust)
2862 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2863 if (err < 0)
2864 return err;
2865 }
2866
2867 if (tb[IFLA_VF_IB_NODE_GUID]) {
2868 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2869
2870 if (ivt->vf >= INT_MAX)
2871 return -EINVAL;
2872 if (!ops->ndo_set_vf_guid)
2873 return -EOPNOTSUPP;
2874 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2875 }
2876
2877 if (tb[IFLA_VF_IB_PORT_GUID]) {
2878 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2879
2880 if (ivt->vf >= INT_MAX)
2881 return -EINVAL;
2882 if (!ops->ndo_set_vf_guid)
2883 return -EOPNOTSUPP;
2884
2885 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2886 }
2887
2888 return err;
2889 }
2890
2891 static int do_set_master(struct net_device *dev, int ifindex,
2892 struct netlink_ext_ack *extack)
2893 {
2894 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2895 const struct net_device_ops *ops;
2896 int err;
2897
2898 if (upper_dev) {
2899 if (upper_dev->ifindex == ifindex)
2900 return 0;
2901 ops = upper_dev->netdev_ops;
2902 if (ops->ndo_del_slave) {
2903 err = ops->ndo_del_slave(upper_dev, dev);
2904 if (err)
2905 return err;
2906 } else {
2907 return -EOPNOTSUPP;
2908 }
2909 }
2910
2911 if (ifindex) {
2912 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2913 if (!upper_dev)
2914 return -EINVAL;
2915 ops = upper_dev->netdev_ops;
2916 if (ops->ndo_add_slave) {
2917 err = ops->ndo_add_slave(upper_dev, dev, extack);
2918 if (err)
2919 return err;
2920 } else {
2921 return -EOPNOTSUPP;
2922 }
2923 }
2924 return 0;
2925 }
2926
2927 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2928 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2929 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2930 };
2931
2932 static int do_set_proto_down(struct net_device *dev,
2933 struct nlattr *nl_proto_down,
2934 struct nlattr *nl_proto_down_reason,
2935 struct netlink_ext_ack *extack)
2936 {
2937 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2938 unsigned long mask = 0;
2939 u32 value;
2940 bool proto_down;
2941 int err;
2942
2943 if (!dev->change_proto_down) {
2944 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2945 return -EOPNOTSUPP;
2946 }
2947
2948 if (nl_proto_down_reason) {
2949 err = nla_parse_nested_deprecated(pdreason,
2950 IFLA_PROTO_DOWN_REASON_MAX,
2951 nl_proto_down_reason,
2952 ifla_proto_down_reason_policy,
2953 NULL);
2954 if (err < 0)
2955 return err;
2956
2957 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2958 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2959 return -EINVAL;
2960 }
2961
2962 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2963
2964 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2965 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2966
2967 dev_change_proto_down_reason(dev, mask, value);
2968 }
2969
2970 if (nl_proto_down) {
2971 proto_down = nla_get_u8(nl_proto_down);
2972
2973 /* Don't turn off protodown if there are active reasons */
2974 if (!proto_down && dev->proto_down_reason) {
2975 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2976 return -EBUSY;
2977 }
2978 err = dev_change_proto_down(dev,
2979 proto_down);
2980 if (err)
2981 return err;
2982 }
2983
2984 return 0;
2985 }
2986
2987 #define DO_SETLINK_MODIFIED 0x01
2988 /* notify flag means notify + modified. */
2989 #define DO_SETLINK_NOTIFY 0x03
2990 static int do_setlink(const struct sk_buff *skb, struct net_device *dev,
2991 struct net *tgt_net, struct ifinfomsg *ifm,
2992 struct netlink_ext_ack *extack,
2993 struct nlattr **tb, int status)
2994 {
2995 const struct net_device_ops *ops = dev->netdev_ops;
2996 char ifname[IFNAMSIZ];
2997 int err;
2998
2999 err = validate_linkmsg(dev, tb, extack);
3000 if (err < 0)
3001 goto errout;
3002
3003 if (tb[IFLA_IFNAME])
3004 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3005 else
3006 ifname[0] = '\0';
3007
3008 if (!net_eq(tgt_net, dev_net(dev))) {
3009 const char *pat = ifname[0] ? ifname : NULL;
3010 int new_ifindex;
3011
3012 new_ifindex = nla_get_s32_default(tb[IFLA_NEW_IFINDEX], 0);
3013
3014 err = __dev_change_net_namespace(dev, tgt_net, pat, new_ifindex);
3015 if (err)
3016 goto errout;
3017
3018 status |= DO_SETLINK_MODIFIED;
3019 }
3020
3021 if (tb[IFLA_MAP]) {
3022 struct rtnl_link_ifmap *u_map;
3023 struct ifmap k_map;
3024
3025 if (!ops->ndo_set_config) {
3026 err = -EOPNOTSUPP;
3027 goto errout;
3028 }
3029
3030 if (!netif_device_present(dev)) {
3031 err = -ENODEV;
3032 goto errout;
3033 }
3034
3035 u_map = nla_data(tb[IFLA_MAP]);
3036 k_map.mem_start = (unsigned long) u_map->mem_start;
3037 k_map.mem_end = (unsigned long) u_map->mem_end;
3038 k_map.base_addr = (unsigned short) u_map->base_addr;
3039 k_map.irq = (unsigned char) u_map->irq;
3040 k_map.dma = (unsigned char) u_map->dma;
3041 k_map.port = (unsigned char) u_map->port;
3042
3043 err = ops->ndo_set_config(dev, &k_map);
3044 if (err < 0)
3045 goto errout;
3046
3047 status |= DO_SETLINK_NOTIFY;
3048 }
3049
3050 if (tb[IFLA_ADDRESS]) {
3051 struct sockaddr *sa;
3052 int len;
3053
3054 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
3055 sizeof(*sa));
3056 sa = kmalloc(len, GFP_KERNEL);
3057 if (!sa) {
3058 err = -ENOMEM;
3059 goto errout;
3060 }
3061 sa->sa_family = dev->type;
3062 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
3063 dev->addr_len);
3064 err = dev_set_mac_address_user(dev, sa, extack);
3065 kfree(sa);
3066 if (err)
3067 goto errout;
3068 status |= DO_SETLINK_MODIFIED;
3069 }
3070
3071 if (tb[IFLA_MTU]) {
3072 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
3073 if (err < 0)
3074 goto errout;
3075 status |= DO_SETLINK_MODIFIED;
3076 }
3077
3078 if (tb[IFLA_GROUP]) {
3079 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3080 status |= DO_SETLINK_NOTIFY;
3081 }
3082
3083 /*
3084 * Interface selected by interface index but interface
3085 * name provided implies that a name change has been
3086 * requested.
3087 */
3088 if (ifm->ifi_index > 0 && ifname[0]) {
3089 err = dev_change_name(dev, ifname);
3090 if (err < 0)
3091 goto errout;
3092 status |= DO_SETLINK_MODIFIED;
3093 }
3094
3095 if (tb[IFLA_IFALIAS]) {
3096 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
3097 nla_len(tb[IFLA_IFALIAS]));
3098 if (err < 0)
3099 goto errout;
3100 status |= DO_SETLINK_NOTIFY;
3101 }
3102
3103 if (tb[IFLA_BROADCAST]) {
3104 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
3105 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
3106 }
3107
3108 if (ifm->ifi_flags || ifm->ifi_change) {
3109 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3110 extack);
3111 if (err < 0)
3112 goto errout;
3113 }
3114
3115 if (tb[IFLA_MASTER]) {
3116 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3117 if (err)
3118 goto errout;
3119 status |= DO_SETLINK_MODIFIED;
3120 }
3121
3122 if (tb[IFLA_CARRIER]) {
3123 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
3124 if (err)
3125 goto errout;
3126 status |= DO_SETLINK_MODIFIED;
3127 }
3128
3129 if (tb[IFLA_TXQLEN]) {
3130 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
3131
3132 err = dev_change_tx_queue_len(dev, value);
3133 if (err)
3134 goto errout;
3135 status |= DO_SETLINK_MODIFIED;
3136 }
3137
3138 if (tb[IFLA_GSO_MAX_SIZE]) {
3139 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
3140
3141 if (dev->gso_max_size ^ max_size) {
3142 netif_set_gso_max_size(dev, max_size);
3143 status |= DO_SETLINK_MODIFIED;
3144 }
3145 }
3146
3147 if (tb[IFLA_GSO_MAX_SEGS]) {
3148 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3149
3150 if (dev->gso_max_segs ^ max_segs) {
3151 netif_set_gso_max_segs(dev, max_segs);
3152 status |= DO_SETLINK_MODIFIED;
3153 }
3154 }
3155
3156 if (tb[IFLA_GRO_MAX_SIZE]) {
3157 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
3158
3159 if (dev->gro_max_size ^ gro_max_size) {
3160 netif_set_gro_max_size(dev, gro_max_size);
3161 status |= DO_SETLINK_MODIFIED;
3162 }
3163 }
3164
3165 if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
3166 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
3167
3168 if (dev->gso_ipv4_max_size ^ max_size) {
3169 netif_set_gso_ipv4_max_size(dev, max_size);
3170 status |= DO_SETLINK_MODIFIED;
3171 }
3172 }
3173
3174 if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
3175 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
3176
3177 if (dev->gro_ipv4_max_size ^ gro_max_size) {
3178 netif_set_gro_ipv4_max_size(dev, gro_max_size);
3179 status |= DO_SETLINK_MODIFIED;
3180 }
3181 }
3182
3183 if (tb[IFLA_OPERSTATE])
3184 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3185
3186 if (tb[IFLA_LINKMODE]) {
3187 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
3188
3189 if (dev->link_mode ^ value)
3190 status |= DO_SETLINK_NOTIFY;
3191 WRITE_ONCE(dev->link_mode, value);
3192 }
3193
3194 if (tb[IFLA_VFINFO_LIST]) {
3195 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
3196 struct nlattr *attr;
3197 int rem;
3198
3199 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
3200 if (nla_type(attr) != IFLA_VF_INFO ||
3201 nla_len(attr) < NLA_HDRLEN) {
3202 err = -EINVAL;
3203 goto errout;
3204 }
3205 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
3206 attr,
3207 ifla_vf_policy,
3208 NULL);
3209 if (err < 0)
3210 goto errout;
3211 err = do_setvfinfo(dev, vfinfo);
3212 if (err < 0)
3213 goto errout;
3214 status |= DO_SETLINK_NOTIFY;
3215 }
3216 }
3217 err = 0;
3218
3219 if (tb[IFLA_VF_PORTS]) {
3220 struct nlattr *port[IFLA_PORT_MAX+1];
3221 struct nlattr *attr;
3222 int vf;
3223 int rem;
3224
3225 err = -EOPNOTSUPP;
3226 if (!ops->ndo_set_vf_port)
3227 goto errout;
3228
3229 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
3230 if (nla_type(attr) != IFLA_VF_PORT ||
3231 nla_len(attr) < NLA_HDRLEN) {
3232 err = -EINVAL;
3233 goto errout;
3234 }
3235 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3236 attr,
3237 ifla_port_policy,
3238 NULL);
3239 if (err < 0)
3240 goto errout;
3241 if (!port[IFLA_PORT_VF]) {
3242 err = -EOPNOTSUPP;
3243 goto errout;
3244 }
3245 vf = nla_get_u32(port[IFLA_PORT_VF]);
3246 err = ops->ndo_set_vf_port(dev, vf, port);
3247 if (err < 0)
3248 goto errout;
3249 status |= DO_SETLINK_NOTIFY;
3250 }
3251 }
3252 err = 0;
3253
3254 if (tb[IFLA_PORT_SELF]) {
3255 struct nlattr *port[IFLA_PORT_MAX+1];
3256
3257 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3258 tb[IFLA_PORT_SELF],
3259 ifla_port_policy, NULL);
3260 if (err < 0)
3261 goto errout;
3262
3263 err = -EOPNOTSUPP;
3264 if (ops->ndo_set_vf_port)
3265 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3266 if (err < 0)
3267 goto errout;
3268 status |= DO_SETLINK_NOTIFY;
3269 }
3270
3271 if (tb[IFLA_AF_SPEC]) {
3272 struct nlattr *af;
3273 int rem;
3274
3275 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3276 struct rtnl_af_ops *af_ops;
3277 int af_ops_srcu_index;
3278
3279 af_ops = rtnl_af_lookup(nla_type(af), &af_ops_srcu_index);
3280 if (!af_ops) {
3281 err = -EAFNOSUPPORT;
3282 goto errout;
3283 }
3284
3285 err = af_ops->set_link_af(dev, af, extack);
3286 rtnl_af_put(af_ops, af_ops_srcu_index);
3287
3288 if (err < 0)
3289 goto errout;
3290
3291 status |= DO_SETLINK_NOTIFY;
3292 }
3293 }
3294 err = 0;
3295
3296 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3297 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3298 tb[IFLA_PROTO_DOWN_REASON], extack);
3299 if (err)
3300 goto errout;
3301 status |= DO_SETLINK_NOTIFY;
3302 }
3303
3304 if (tb[IFLA_XDP]) {
3305 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3306 u32 xdp_flags = 0;
3307
3308 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3309 tb[IFLA_XDP],
3310 ifla_xdp_policy, NULL);
3311 if (err < 0)
3312 goto errout;
3313
3314 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3315 err = -EINVAL;
3316 goto errout;
3317 }
3318
3319 if (xdp[IFLA_XDP_FLAGS]) {
3320 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3321 if (xdp_flags & ~XDP_FLAGS_MASK) {
3322 err = -EINVAL;
3323 goto errout;
3324 }
3325 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3326 err = -EINVAL;
3327 goto errout;
3328 }
3329 }
3330
3331 if (xdp[IFLA_XDP_FD]) {
3332 int expected_fd = -1;
3333
3334 if (xdp_flags & XDP_FLAGS_REPLACE) {
3335 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3336 err = -EINVAL;
3337 goto errout;
3338 }
3339 expected_fd =
3340 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3341 }
3342
3343 err = dev_change_xdp_fd(dev, extack,
3344 nla_get_s32(xdp[IFLA_XDP_FD]),
3345 expected_fd,
3346 xdp_flags);
3347 if (err)
3348 goto errout;
3349 status |= DO_SETLINK_NOTIFY;
3350 }
3351 }
3352
3353 errout:
3354 if (status & DO_SETLINK_MODIFIED) {
3355 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3356 netdev_state_change(dev);
3357
3358 if (err < 0)
3359 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",
3360 dev->name);
3361 }
3362
3363 return err;
3364 }
3365
3366 static struct net_device *rtnl_dev_get(struct net *net,
3367 struct nlattr *tb[])
3368 {
3369 char ifname[ALTIFNAMSIZ];
3370
3371 if (tb[IFLA_IFNAME])
3372 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3373 else if (tb[IFLA_ALT_IFNAME])
3374 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3375 else
3376 return NULL;
3377
3378 return __dev_get_by_name(net, ifname);
3379 }
3380
3381 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3382 struct netlink_ext_ack *extack)
3383 {
3384 struct ifinfomsg *ifm = nlmsg_data(nlh);
3385 struct net *net = sock_net(skb->sk);
3386 struct nlattr *tb[IFLA_MAX+1];
3387 struct net_device *dev = NULL;
3388 struct rtnl_nets rtnl_nets;
3389 struct net *tgt_net;
3390 int err;
3391
3392 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3393 ifla_policy, extack);
3394 if (err < 0)
3395 goto errout;
3396
3397 err = rtnl_ensure_unique_netns(tb, extack, false);
3398 if (err < 0)
3399 goto errout;
3400
3401 tgt_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3402 if (IS_ERR(tgt_net)) {
3403 err = PTR_ERR(tgt_net);
3404 goto errout;
3405 }
3406
3407 rtnl_nets_init(&rtnl_nets);
3408 rtnl_nets_add(&rtnl_nets, get_net(net));
3409 rtnl_nets_add(&rtnl_nets, tgt_net);
3410
3411 rtnl_nets_lock(&rtnl_nets);
3412
3413 if (ifm->ifi_index > 0)
3414 dev = __dev_get_by_index(net, ifm->ifi_index);
3415 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3416 dev = rtnl_dev_get(net, tb);
3417 else
3418 err = -EINVAL;
3419
3420 if (dev)
3421 err = do_setlink(skb, dev, tgt_net, ifm, extack, tb, 0);
3422 else if (!err)
3423 err = -ENODEV;
3424
3425 rtnl_nets_unlock(&rtnl_nets);
3426 errout:
3427 return err;
3428 }
3429
3430 static int rtnl_group_dellink(const struct net *net, int group)
3431 {
3432 struct net_device *dev, *aux;
3433 LIST_HEAD(list_kill);
3434 bool found = false;
3435
3436 if (!group)
3437 return -EPERM;
3438
3439 for_each_netdev(net, dev) {
3440 if (dev->group == group) {
3441 const struct rtnl_link_ops *ops;
3442
3443 found = true;
3444 ops = dev->rtnl_link_ops;
3445 if (!ops || !ops->dellink)
3446 return -EOPNOTSUPP;
3447 }
3448 }
3449
3450 if (!found)
3451 return -ENODEV;
3452
3453 for_each_netdev_safe(net, dev, aux) {
3454 if (dev->group == group) {
3455 const struct rtnl_link_ops *ops;
3456
3457 ops = dev->rtnl_link_ops;
3458 ops->dellink(dev, &list_kill);
3459 }
3460 }
3461 unregister_netdevice_many(&list_kill);
3462
3463 return 0;
3464 }
3465
3466 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3467 {
3468 const struct rtnl_link_ops *ops;
3469 LIST_HEAD(list_kill);
3470
3471 ops = dev->rtnl_link_ops;
3472 if (!ops || !ops->dellink)
3473 return -EOPNOTSUPP;
3474
3475 ops->dellink(dev, &list_kill);
3476 unregister_netdevice_many_notify(&list_kill, portid, nlh);
3477
3478 return 0;
3479 }
3480 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3481
3482 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3483 struct netlink_ext_ack *extack)
3484 {
3485 struct ifinfomsg *ifm = nlmsg_data(nlh);
3486 struct net *net = sock_net(skb->sk);
3487 u32 portid = NETLINK_CB(skb).portid;
3488 struct nlattr *tb[IFLA_MAX+1];
3489 struct net_device *dev = NULL;
3490 struct net *tgt_net = net;
3491 int netnsid = -1;
3492 int err;
3493
3494 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3495 ifla_policy, extack);
3496 if (err < 0)
3497 return err;
3498
3499 err = rtnl_ensure_unique_netns(tb, extack, true);
3500 if (err < 0)
3501 return err;
3502
3503 if (tb[IFLA_TARGET_NETNSID]) {
3504 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3505 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3506 if (IS_ERR(tgt_net))
3507 return PTR_ERR(tgt_net);
3508 }
3509
3510 rtnl_net_lock(tgt_net);
3511
3512 if (ifm->ifi_index > 0)
3513 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3514 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3515 dev = rtnl_dev_get(tgt_net, tb);
3516
3517 if (dev)
3518 err = rtnl_delete_link(dev, portid, nlh);
3519 else if (ifm->ifi_index > 0 || tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3520 err = -ENODEV;
3521 else if (tb[IFLA_GROUP])
3522 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3523 else
3524 err = -EINVAL;
3525
3526 rtnl_net_unlock(tgt_net);
3527
3528 if (netnsid >= 0)
3529 put_net(tgt_net);
3530
3531 return err;
3532 }
3533
3534 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3535 u32 portid, const struct nlmsghdr *nlh)
3536 {
3537 unsigned int old_flags;
3538 int err;
3539
3540 old_flags = dev->flags;
3541 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3542 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3543 NULL);
3544 if (err < 0)
3545 return err;
3546 }
3547
3548 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3549 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3550 } else {
3551 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3552 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3553 }
3554 return 0;
3555 }
3556 EXPORT_SYMBOL(rtnl_configure_link);
3557
3558 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3559 unsigned char name_assign_type,
3560 const struct rtnl_link_ops *ops,
3561 struct nlattr *tb[],
3562 struct netlink_ext_ack *extack)
3563 {
3564 struct net_device *dev;
3565 unsigned int num_tx_queues = 1;
3566 unsigned int num_rx_queues = 1;
3567 int err;
3568
3569 if (tb[IFLA_NUM_TX_QUEUES])
3570 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3571 else if (ops->get_num_tx_queues)
3572 num_tx_queues = ops->get_num_tx_queues();
3573
3574 if (tb[IFLA_NUM_RX_QUEUES])
3575 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3576 else if (ops->get_num_rx_queues)
3577 num_rx_queues = ops->get_num_rx_queues();
3578
3579 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3580 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3581 return ERR_PTR(-EINVAL);
3582 }
3583
3584 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3585 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3586 return ERR_PTR(-EINVAL);
3587 }
3588
3589 if (ops->alloc) {
3590 dev = ops->alloc(tb, ifname, name_assign_type,
3591 num_tx_queues, num_rx_queues);
3592 if (IS_ERR(dev))
3593 return dev;
3594 } else {
3595 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3596 name_assign_type, ops->setup,
3597 num_tx_queues, num_rx_queues);
3598 }
3599
3600 if (!dev)
3601 return ERR_PTR(-ENOMEM);
3602
3603 err = validate_linkmsg(dev, tb, extack);
3604 if (err < 0) {
3605 free_netdev(dev);
3606 return ERR_PTR(err);
3607 }
3608
3609 dev_net_set(dev, net);
3610 dev->rtnl_link_ops = ops;
3611 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3612
3613 if (tb[IFLA_MTU]) {
3614 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3615
3616 err = dev_validate_mtu(dev, mtu, extack);
3617 if (err) {
3618 free_netdev(dev);
3619 return ERR_PTR(err);
3620 }
3621 dev->mtu = mtu;
3622 }
3623 if (tb[IFLA_ADDRESS]) {
3624 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3625 nla_len(tb[IFLA_ADDRESS]));
3626 dev->addr_assign_type = NET_ADDR_SET;
3627 }
3628 if (tb[IFLA_BROADCAST])
3629 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3630 nla_len(tb[IFLA_BROADCAST]));
3631 if (tb[IFLA_TXQLEN])
3632 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3633 if (tb[IFLA_OPERSTATE])
3634 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3635 if (tb[IFLA_LINKMODE])
3636 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3637 if (tb[IFLA_GROUP])
3638 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3639 if (tb[IFLA_GSO_MAX_SIZE])
3640 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3641 if (tb[IFLA_GSO_MAX_SEGS])
3642 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3643 if (tb[IFLA_GRO_MAX_SIZE])
3644 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3645 if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3646 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3647 if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3648 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3649
3650 return dev;
3651 }
3652 EXPORT_SYMBOL(rtnl_create_link);
3653
3654 struct rtnl_newlink_tbs {
3655 struct nlattr *tb[IFLA_MAX + 1];
3656 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3657 struct nlattr *attr[RTNL_MAX_TYPE + 1];
3658 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3659 };
3660
3661 static int rtnl_changelink(const struct sk_buff *skb, struct nlmsghdr *nlh,
3662 const struct rtnl_link_ops *ops,
3663 struct net_device *dev, struct net *tgt_net,
3664 struct rtnl_newlink_tbs *tbs,
3665 struct nlattr **data,
3666 struct netlink_ext_ack *extack)
3667 {
3668 struct nlattr ** const linkinfo = tbs->linkinfo;
3669 struct nlattr ** const tb = tbs->tb;
3670 int status = 0;
3671 int err;
3672
3673 if (nlh->nlmsg_flags & NLM_F_EXCL)
3674 return -EEXIST;
3675
3676 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3677 return -EOPNOTSUPP;
3678
3679 if (linkinfo[IFLA_INFO_DATA]) {
3680 if (!ops || ops != dev->rtnl_link_ops || !ops->changelink)
3681 return -EOPNOTSUPP;
3682
3683 err = ops->changelink(dev, tb, data, extack);
3684 if (err < 0)
3685 return err;
3686
3687 status |= DO_SETLINK_NOTIFY;
3688 }
3689
3690 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3691 const struct rtnl_link_ops *m_ops = NULL;
3692 struct nlattr **slave_data = NULL;
3693 struct net_device *master_dev;
3694
3695 master_dev = netdev_master_upper_dev_get(dev);
3696 if (master_dev)
3697 m_ops = master_dev->rtnl_link_ops;
3698
3699 if (!m_ops || !m_ops->slave_changelink)
3700 return -EOPNOTSUPP;
3701
3702 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3703 return -EINVAL;
3704
3705 if (m_ops->slave_maxtype) {
3706 err = nla_parse_nested_deprecated(tbs->slave_attr,
3707 m_ops->slave_maxtype,
3708 linkinfo[IFLA_INFO_SLAVE_DATA],
3709 m_ops->slave_policy, extack);
3710 if (err < 0)
3711 return err;
3712
3713 slave_data = tbs->slave_attr;
3714 }
3715
3716 err = m_ops->slave_changelink(master_dev, dev, tb, slave_data, extack);
3717 if (err < 0)
3718 return err;
3719
3720 status |= DO_SETLINK_NOTIFY;
3721 }
3722
3723 return do_setlink(skb, dev, tgt_net, nlmsg_data(nlh), extack, tb, status);
3724 }
3725
3726 static int rtnl_group_changelink(const struct sk_buff *skb,
3727 struct net *net, struct net *tgt_net,
3728 int group, struct ifinfomsg *ifm,
3729 struct netlink_ext_ack *extack,
3730 struct nlattr **tb)
3731 {
3732 struct net_device *dev, *aux;
3733 int err;
3734
3735 for_each_netdev_safe(net, dev, aux) {
3736 if (dev->group == group) {
3737 err = do_setlink(skb, dev, tgt_net, ifm, extack, tb, 0);
3738 if (err < 0)
3739 return err;
3740 }
3741 }
3742
3743 return 0;
3744 }
3745
3746 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3747 const struct rtnl_link_ops *ops,
3748 struct net *tgt_net, struct net *link_net,
3749 const struct nlmsghdr *nlh,
3750 struct nlattr **tb, struct nlattr **data,
3751 struct netlink_ext_ack *extack)
3752 {
3753 unsigned char name_assign_type = NET_NAME_USER;
3754 struct net *net = sock_net(skb->sk);
3755 u32 portid = NETLINK_CB(skb).portid;
3756 struct net_device *dev;
3757 char ifname[IFNAMSIZ];
3758 int err;
3759
3760 if (!ops->alloc && !ops->setup)
3761 return -EOPNOTSUPP;
3762
3763 if (tb[IFLA_IFNAME]) {
3764 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3765 } else {
3766 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3767 name_assign_type = NET_NAME_ENUM;
3768 }
3769
3770 dev = rtnl_create_link(link_net ? : tgt_net, ifname,
3771 name_assign_type, ops, tb, extack);
3772 if (IS_ERR(dev)) {
3773 err = PTR_ERR(dev);
3774 goto out;
3775 }
3776
3777 dev->ifindex = ifm->ifi_index;
3778
3779 if (ops->newlink)
3780 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3781 else
3782 err = register_netdevice(dev);
3783 if (err < 0) {
3784 free_netdev(dev);
3785 goto out;
3786 }
3787
3788 err = rtnl_configure_link(dev, ifm, portid, nlh);
3789 if (err < 0)
3790 goto out_unregister;
3791 if (link_net) {
3792 err = dev_change_net_namespace(dev, tgt_net, ifname);
3793 if (err < 0)
3794 goto out_unregister;
3795 }
3796 if (tb[IFLA_MASTER]) {
3797 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3798 if (err)
3799 goto out_unregister;
3800 }
3801 out:
3802 return err;
3803 out_unregister:
3804 if (ops->newlink) {
3805 LIST_HEAD(list_kill);
3806
3807 ops->dellink(dev, &list_kill);
3808 unregister_netdevice_many(&list_kill);
3809 } else {
3810 unregister_netdevice(dev);
3811 }
3812 goto out;
3813 }
3814
3815 static int rtnl_add_peer_net(struct rtnl_nets *rtnl_nets,
3816 const struct rtnl_link_ops *ops,
3817 struct nlattr *data[],
3818 struct netlink_ext_ack *extack)
3819 {
3820 struct nlattr *tb[IFLA_MAX + 1];
3821 struct net *net;
3822 int err;
3823
3824 if (!data || !data[ops->peer_type])
3825 return 0;
3826
3827 err = rtnl_nla_parse_ifinfomsg(tb, data[ops->peer_type], extack);
3828 if (err < 0)
3829 return err;
3830
3831 if (ops->validate) {
3832 err = ops->validate(tb, NULL, extack);
3833 if (err < 0)
3834 return err;
3835 }
3836
3837 net = rtnl_link_get_net_ifla(tb);
3838 if (IS_ERR(net))
3839 return PTR_ERR(net);
3840 if (net)
3841 rtnl_nets_add(rtnl_nets, net);
3842
3843 return 0;
3844 }
3845
3846 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3847 const struct rtnl_link_ops *ops,
3848 struct net *tgt_net, struct net *link_net,
3849 struct rtnl_newlink_tbs *tbs,
3850 struct nlattr **data,
3851 struct netlink_ext_ack *extack)
3852 {
3853 struct nlattr ** const tb = tbs->tb;
3854 struct net *net = sock_net(skb->sk);
3855 struct net_device *dev;
3856 struct ifinfomsg *ifm;
3857 bool link_specified;
3858
3859 ifm = nlmsg_data(nlh);
3860 if (ifm->ifi_index > 0) {
3861 link_specified = true;
3862 dev = __dev_get_by_index(net, ifm->ifi_index);
3863 } else if (ifm->ifi_index < 0) {
3864 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3865 return -EINVAL;
3866 } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3867 link_specified = true;
3868 dev = rtnl_dev_get(net, tb);
3869 } else {
3870 link_specified = false;
3871 dev = NULL;
3872 }
3873
3874 if (dev)
3875 return rtnl_changelink(skb, nlh, ops, dev, tgt_net, tbs, data, extack);
3876
3877 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3878 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3879 * or it's for a group
3880 */
3881 if (link_specified || !tb[IFLA_GROUP])
3882 return -ENODEV;
3883
3884 return rtnl_group_changelink(skb, net, tgt_net,
3885 nla_get_u32(tb[IFLA_GROUP]),
3886 ifm, extack, tb);
3887 }
3888
3889 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3890 return -EOPNOTSUPP;
3891
3892 if (!ops) {
3893 NL_SET_ERR_MSG(extack, "Unknown device type");
3894 return -EOPNOTSUPP;
3895 }
3896
3897 return rtnl_newlink_create(skb, ifm, ops, tgt_net, link_net, nlh, tb, data, extack);
3898 }
3899
3900 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3901 struct netlink_ext_ack *extack)
3902 {
3903 struct nlattr **tb, **linkinfo, **data = NULL;
3904 struct net *tgt_net, *link_net = NULL;
3905 struct rtnl_link_ops *ops = NULL;
3906 struct rtnl_newlink_tbs *tbs;
3907 struct rtnl_nets rtnl_nets;
3908 int ops_srcu_index;
3909 int ret;
3910
3911 tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3912 if (!tbs)
3913 return -ENOMEM;
3914
3915 tb = tbs->tb;
3916 ret = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg), tb,
3917 IFLA_MAX, ifla_policy, extack);
3918 if (ret < 0)
3919 goto free;
3920
3921 ret = rtnl_ensure_unique_netns(tb, extack, false);
3922 if (ret < 0)
3923 goto free;
3924
3925 linkinfo = tbs->linkinfo;
3926 if (tb[IFLA_LINKINFO]) {
3927 ret = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3928 tb[IFLA_LINKINFO],
3929 ifla_info_policy, NULL);
3930 if (ret < 0)
3931 goto free;
3932 } else {
3933 memset(linkinfo, 0, sizeof(tbs->linkinfo));
3934 }
3935
3936 if (linkinfo[IFLA_INFO_KIND]) {
3937 char kind[MODULE_NAME_LEN];
3938
3939 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3940 ops = rtnl_link_ops_get(kind, &ops_srcu_index);
3941 #ifdef CONFIG_MODULES
3942 if (!ops) {
3943 request_module("rtnl-link-%s", kind);
3944 ops = rtnl_link_ops_get(kind, &ops_srcu_index);
3945 }
3946 #endif
3947 }
3948
3949 rtnl_nets_init(&rtnl_nets);
3950
3951 if (ops) {
3952 if (ops->maxtype > RTNL_MAX_TYPE) {
3953 ret = -EINVAL;
3954 goto put_ops;
3955 }
3956
3957 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3958 ret = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3959 linkinfo[IFLA_INFO_DATA],
3960 ops->policy, extack);
3961 if (ret < 0)
3962 goto put_ops;
3963
3964 data = tbs->attr;
3965 }
3966
3967 if (ops->validate) {
3968 ret = ops->validate(tb, data, extack);
3969 if (ret < 0)
3970 goto put_ops;
3971 }
3972
3973 if (ops->peer_type) {
3974 ret = rtnl_add_peer_net(&rtnl_nets, ops, data, extack);
3975 if (ret < 0)
3976 goto put_ops;
3977 }
3978 }
3979
3980 tgt_net = rtnl_link_get_net_capable(skb, sock_net(skb->sk), tb, CAP_NET_ADMIN);
3981 if (IS_ERR(tgt_net)) {
3982 ret = PTR_ERR(tgt_net);
3983 goto put_net;
3984 }
3985
3986 rtnl_nets_add(&rtnl_nets, tgt_net);
3987
3988 if (tb[IFLA_LINK_NETNSID]) {
3989 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3990
3991 link_net = get_net_ns_by_id(tgt_net, id);
3992 if (!link_net) {
3993 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3994 ret = -EINVAL;
3995 goto put_net;
3996 }
3997
3998 rtnl_nets_add(&rtnl_nets, link_net);
3999
4000 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) {
4001 ret = -EPERM;
4002 goto put_net;
4003 }
4004 }
4005
4006 rtnl_nets_lock(&rtnl_nets);
4007 ret = __rtnl_newlink(skb, nlh, ops, tgt_net, link_net, tbs, data, extack);
4008 rtnl_nets_unlock(&rtnl_nets);
4009
4010 put_net:
4011 rtnl_nets_destroy(&rtnl_nets);
4012 put_ops:
4013 if (ops)
4014 rtnl_link_ops_put(ops, ops_srcu_index);
4015 free:
4016 kfree(tbs);
4017 return ret;
4018 }
4019
4020 static int rtnl_valid_getlink_req(struct sk_buff *skb,
4021 const struct nlmsghdr *nlh,
4022 struct nlattr **tb,
4023 struct netlink_ext_ack *extack)
4024 {
4025 struct ifinfomsg *ifm;
4026 int i, err;
4027
4028 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4029 NL_SET_ERR_MSG(extack, "Invalid header for get link");
4030 return -EINVAL;
4031 }
4032
4033 if (!netlink_strict_get_check(skb))
4034 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
4035 ifla_policy, extack);
4036
4037 ifm = nlmsg_data(nlh);
4038 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4039 ifm->ifi_change) {
4040 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
4041 return -EINVAL;
4042 }
4043
4044 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
4045 ifla_policy, extack);
4046 if (err)
4047 return err;
4048
4049 for (i = 0; i <= IFLA_MAX; i++) {
4050 if (!tb[i])
4051 continue;
4052
4053 switch (i) {
4054 case IFLA_IFNAME:
4055 case IFLA_ALT_IFNAME:
4056 case IFLA_EXT_MASK:
4057 case IFLA_TARGET_NETNSID:
4058 break;
4059 default:
4060 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
4061 return -EINVAL;
4062 }
4063 }
4064
4065 return 0;
4066 }
4067
4068 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4069 struct netlink_ext_ack *extack)
4070 {
4071 struct net *net = sock_net(skb->sk);
4072 struct net *tgt_net = net;
4073 struct ifinfomsg *ifm;
4074 struct nlattr *tb[IFLA_MAX+1];
4075 struct net_device *dev = NULL;
4076 struct sk_buff *nskb;
4077 int netnsid = -1;
4078 int err;
4079 u32 ext_filter_mask = 0;
4080
4081 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
4082 if (err < 0)
4083 return err;
4084
4085 err = rtnl_ensure_unique_netns(tb, extack, true);
4086 if (err < 0)
4087 return err;
4088
4089 if (tb[IFLA_TARGET_NETNSID]) {
4090 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
4091 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
4092 if (IS_ERR(tgt_net))
4093 return PTR_ERR(tgt_net);
4094 }
4095
4096 if (tb[IFLA_EXT_MASK])
4097 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
4098
4099 err = -EINVAL;
4100 ifm = nlmsg_data(nlh);
4101 if (ifm->ifi_index > 0)
4102 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
4103 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
4104 dev = rtnl_dev_get(tgt_net, tb);
4105 else
4106 goto out;
4107
4108 err = -ENODEV;
4109 if (dev == NULL)
4110 goto out;
4111
4112 err = -ENOBUFS;
4113 nskb = nlmsg_new_large(if_nlmsg_size(dev, ext_filter_mask));
4114 if (nskb == NULL)
4115 goto out;
4116
4117 /* Synchronize the carrier state so we don't report a state
4118 * that we're not actually going to honour immediately; if
4119 * the driver just did a carrier off->on transition, we can
4120 * only TX if link watch work has run, but without this we'd
4121 * already report carrier on, even if it doesn't work yet.
4122 */
4123 linkwatch_sync_dev(dev);
4124
4125 err = rtnl_fill_ifinfo(nskb, dev, net,
4126 RTM_NEWLINK, NETLINK_CB(skb).portid,
4127 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
4128 0, NULL, 0, netnsid, GFP_KERNEL);
4129 if (err < 0) {
4130 /* -EMSGSIZE implies BUG in if_nlmsg_size */
4131 WARN_ON(err == -EMSGSIZE);
4132 kfree_skb(nskb);
4133 } else
4134 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4135 out:
4136 if (netnsid >= 0)
4137 put_net(tgt_net);
4138
4139 return err;
4140 }
4141
4142 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
4143 bool *changed, struct netlink_ext_ack *extack)
4144 {
4145 char *alt_ifname;
4146 size_t size;
4147 int err;
4148
4149 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
4150 if (err)
4151 return err;
4152
4153 if (cmd == RTM_NEWLINKPROP) {
4154 size = rtnl_prop_list_size(dev);
4155 size += nla_total_size(ALTIFNAMSIZ);
4156 if (size >= U16_MAX) {
4157 NL_SET_ERR_MSG(extack,
4158 "effective property list too long");
4159 return -EINVAL;
4160 }
4161 }
4162
4163 alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
4164 if (!alt_ifname)
4165 return -ENOMEM;
4166
4167 if (cmd == RTM_NEWLINKPROP) {
4168 err = netdev_name_node_alt_create(dev, alt_ifname);
4169 if (!err)
4170 alt_ifname = NULL;
4171 } else if (cmd == RTM_DELLINKPROP) {
4172 err = netdev_name_node_alt_destroy(dev, alt_ifname);
4173 } else {
4174 WARN_ON_ONCE(1);
4175 err = -EINVAL;
4176 }
4177
4178 kfree(alt_ifname);
4179 if (!err)
4180 *changed = true;
4181 return err;
4182 }
4183
4184 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
4185 struct netlink_ext_ack *extack)
4186 {
4187 struct net *net = sock_net(skb->sk);
4188 struct nlattr *tb[IFLA_MAX + 1];
4189 struct net_device *dev;
4190 struct ifinfomsg *ifm;
4191 bool changed = false;
4192 struct nlattr *attr;
4193 int err, rem;
4194
4195 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
4196 if (err)
4197 return err;
4198
4199 err = rtnl_ensure_unique_netns(tb, extack, true);
4200 if (err)
4201 return err;
4202
4203 ifm = nlmsg_data(nlh);
4204 if (ifm->ifi_index > 0)
4205 dev = __dev_get_by_index(net, ifm->ifi_index);
4206 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
4207 dev = rtnl_dev_get(net, tb);
4208 else
4209 return -EINVAL;
4210
4211 if (!dev)
4212 return -ENODEV;
4213
4214 if (!tb[IFLA_PROP_LIST])
4215 return 0;
4216
4217 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
4218 switch (nla_type(attr)) {
4219 case IFLA_ALT_IFNAME:
4220 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
4221 if (err)
4222 return err;
4223 break;
4224 }
4225 }
4226
4227 if (changed)
4228 netdev_state_change(dev);
4229 return 0;
4230 }
4231
4232 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
4233 struct netlink_ext_ack *extack)
4234 {
4235 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
4236 }
4237
4238 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
4239 struct netlink_ext_ack *extack)
4240 {
4241 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
4242 }
4243
4244 static noinline_for_stack u32 rtnl_calcit(struct sk_buff *skb,
4245 struct nlmsghdr *nlh)
4246 {
4247 struct net *net = sock_net(skb->sk);
4248 size_t min_ifinfo_dump_size = 0;
4249 u32 ext_filter_mask = 0;
4250 struct net_device *dev;
4251 struct nlattr *nla;
4252 int hdrlen, rem;
4253
4254 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
4255 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
4256 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
4257
4258 if (nlh->nlmsg_len < nlmsg_msg_size(hdrlen))
4259 return NLMSG_GOODSIZE;
4260
4261 nla_for_each_attr_type(nla, IFLA_EXT_MASK,
4262 nlmsg_attrdata(nlh, hdrlen),
4263 nlmsg_attrlen(nlh, hdrlen), rem) {
4264 if (nla_len(nla) == sizeof(u32))
4265 ext_filter_mask = nla_get_u32(nla);
4266 }
4267
4268 if (!ext_filter_mask)
4269 return NLMSG_GOODSIZE;
4270 /*
4271 * traverse the list of net devices and compute the minimum
4272 * buffer size based upon the filter mask.
4273 */
4274 rcu_read_lock();
4275 for_each_netdev_rcu(net, dev) {
4276 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
4277 if_nlmsg_size(dev, ext_filter_mask));
4278 }
4279 rcu_read_unlock();
4280
4281 return nlmsg_total_size(min_ifinfo_dump_size);
4282 }
4283
4284 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
4285 {
4286 int idx;
4287 int s_idx = cb->family;
4288 int type = cb->nlh->nlmsg_type - RTM_BASE;
4289 int ret = 0;
4290
4291 if (s_idx == 0)
4292 s_idx = 1;
4293
4294 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
4295 struct rtnl_link __rcu **tab;
4296 struct rtnl_link *link;
4297 rtnl_dumpit_func dumpit;
4298
4299 if (idx < s_idx || idx == PF_PACKET)
4300 continue;
4301
4302 if (type < 0 || type >= RTM_NR_MSGTYPES)
4303 continue;
4304
4305 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
4306 if (!tab)
4307 continue;
4308
4309 link = rcu_dereference_rtnl(tab[type]);
4310 if (!link)
4311 continue;
4312
4313 dumpit = link->dumpit;
4314 if (!dumpit)
4315 continue;
4316
4317 if (idx > s_idx) {
4318 memset(&cb->args[0], 0, sizeof(cb->args));
4319 cb->prev_seq = 0;
4320 cb->seq = 0;
4321 }
4322 ret = dumpit(skb, cb);
4323 if (ret)
4324 break;
4325 }
4326 cb->family = idx;
4327
4328 return skb->len ? : ret;
4329 }
4330
4331 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
4332 unsigned int change,
4333 u32 event, gfp_t flags, int *new_nsid,
4334 int new_ifindex, u32 portid,
4335 const struct nlmsghdr *nlh)
4336 {
4337 struct net *net = dev_net(dev);
4338 struct sk_buff *skb;
4339 int err = -ENOBUFS;
4340 u32 seq = 0;
4341
4342 skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4343 if (skb == NULL)
4344 goto errout;
4345
4346 if (nlmsg_report(nlh))
4347 seq = nlmsg_seq(nlh);
4348 else
4349 portid = 0;
4350
4351 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4352 type, portid, seq, change, 0, 0, event,
4353 new_nsid, new_ifindex, -1, flags);
4354 if (err < 0) {
4355 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4356 WARN_ON(err == -EMSGSIZE);
4357 kfree_skb(skb);
4358 goto errout;
4359 }
4360 return skb;
4361 errout:
4362 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4363 return NULL;
4364 }
4365
4366 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4367 u32 portid, const struct nlmsghdr *nlh)
4368 {
4369 struct net *net = dev_net(dev);
4370
4371 rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4372 }
4373
4374 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4375 unsigned int change, u32 event,
4376 gfp_t flags, int *new_nsid, int new_ifindex,
4377 u32 portid, const struct nlmsghdr *nlh)
4378 {
4379 struct sk_buff *skb;
4380
4381 if (dev->reg_state != NETREG_REGISTERED)
4382 return;
4383
4384 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4385 new_ifindex, portid, nlh);
4386 if (skb)
4387 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4388 }
4389
4390 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4391 gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4392 {
4393 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4394 NULL, 0, portid, nlh);
4395 }
4396
4397 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4398 gfp_t flags, int *new_nsid, int new_ifindex)
4399 {
4400 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4401 new_nsid, new_ifindex, 0, NULL);
4402 }
4403
4404 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4405 struct net_device *dev,
4406 u8 *addr, u16 vid, u32 pid, u32 seq,
4407 int type, unsigned int flags,
4408 int nlflags, u16 ndm_state)
4409 {
4410 struct nlmsghdr *nlh;
4411 struct ndmsg *ndm;
4412
4413 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4414 if (!nlh)
4415 return -EMSGSIZE;
4416
4417 ndm = nlmsg_data(nlh);
4418 ndm->ndm_family = AF_BRIDGE;
4419 ndm->ndm_pad1 = 0;
4420 ndm->ndm_pad2 = 0;
4421 ndm->ndm_flags = flags;
4422 ndm->ndm_type = 0;
4423 ndm->ndm_ifindex = dev->ifindex;
4424 ndm->ndm_state = ndm_state;
4425
4426 if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
4427 goto nla_put_failure;
4428 if (vid)
4429 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4430 goto nla_put_failure;
4431
4432 nlmsg_end(skb, nlh);
4433 return 0;
4434
4435 nla_put_failure:
4436 nlmsg_cancel(skb, nlh);
4437 return -EMSGSIZE;
4438 }
4439
4440 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
4441 {
4442 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4443 nla_total_size(dev->addr_len) + /* NDA_LLADDR */
4444 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
4445 0;
4446 }
4447
4448 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4449 u16 ndm_state)
4450 {
4451 struct net *net = dev_net(dev);
4452 struct sk_buff *skb;
4453 int err = -ENOBUFS;
4454
4455 skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
4456 if (!skb)
4457 goto errout;
4458
4459 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4460 0, 0, type, NTF_SELF, 0, ndm_state);
4461 if (err < 0) {
4462 kfree_skb(skb);
4463 goto errout;
4464 }
4465
4466 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4467 return;
4468 errout:
4469 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4470 }
4471
4472 /*
4473 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4474 */
4475 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4476 struct nlattr *tb[],
4477 struct net_device *dev,
4478 const unsigned char *addr, u16 vid,
4479 u16 flags)
4480 {
4481 int err = -EINVAL;
4482
4483 /* If aging addresses are supported device will need to
4484 * implement its own handler for this.
4485 */
4486 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4487 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4488 return err;
4489 }
4490
4491 if (tb[NDA_FLAGS_EXT]) {
4492 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4493 return err;
4494 }
4495
4496 if (vid) {
4497 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4498 return err;
4499 }
4500
4501 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4502 err = dev_uc_add_excl(dev, addr);
4503 else if (is_multicast_ether_addr(addr))
4504 err = dev_mc_add_excl(dev, addr);
4505
4506 /* Only return duplicate errors if NLM_F_EXCL is set */
4507 if (err == -EEXIST && !(flags & NLM_F_EXCL))
4508 err = 0;
4509
4510 return err;
4511 }
4512 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4513
4514 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4515 struct netlink_ext_ack *extack)
4516 {
4517 u16 vid = 0;
4518
4519 if (vlan_attr) {
4520 if (nla_len(vlan_attr) != sizeof(u16)) {
4521 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4522 return -EINVAL;
4523 }
4524
4525 vid = nla_get_u16(vlan_attr);
4526
4527 if (!vid || vid >= VLAN_VID_MASK) {
4528 NL_SET_ERR_MSG(extack, "invalid vlan id");
4529 return -EINVAL;
4530 }
4531 }
4532 *p_vid = vid;
4533 return 0;
4534 }
4535
4536 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4537 struct netlink_ext_ack *extack)
4538 {
4539 struct net *net = sock_net(skb->sk);
4540 struct ndmsg *ndm;
4541 struct nlattr *tb[NDA_MAX+1];
4542 struct net_device *dev;
4543 u8 *addr;
4544 u16 vid;
4545 int err;
4546
4547 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4548 extack);
4549 if (err < 0)
4550 return err;
4551
4552 ndm = nlmsg_data(nlh);
4553 if (ndm->ndm_ifindex == 0) {
4554 NL_SET_ERR_MSG(extack, "invalid ifindex");
4555 return -EINVAL;
4556 }
4557
4558 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4559 if (dev == NULL) {
4560 NL_SET_ERR_MSG(extack, "unknown ifindex");
4561 return -ENODEV;
4562 }
4563
4564 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4565 NL_SET_ERR_MSG(extack, "invalid address");
4566 return -EINVAL;
4567 }
4568
4569 if (dev->type != ARPHRD_ETHER) {
4570 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4571 return -EINVAL;
4572 }
4573
4574 addr = nla_data(tb[NDA_LLADDR]);
4575
4576 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4577 if (err)
4578 return err;
4579
4580 err = -EOPNOTSUPP;
4581
4582 /* Support fdb on master device the net/bridge default case */
4583 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4584 netif_is_bridge_port(dev)) {
4585 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4586 const struct net_device_ops *ops = br_dev->netdev_ops;
4587 bool notified = false;
4588
4589 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4590 nlh->nlmsg_flags, &notified, extack);
4591 if (err)
4592 goto out;
4593 else
4594 ndm->ndm_flags &= ~NTF_MASTER;
4595 }
4596
4597 /* Embedded bridge, macvlan, and any other device support */
4598 if ((ndm->ndm_flags & NTF_SELF)) {
4599 bool notified = false;
4600
4601 if (dev->netdev_ops->ndo_fdb_add)
4602 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4603 vid,
4604 nlh->nlmsg_flags,
4605 &notified, extack);
4606 else
4607 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4608 nlh->nlmsg_flags);
4609
4610 if (!err && !notified) {
4611 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4612 ndm->ndm_state);
4613 ndm->ndm_flags &= ~NTF_SELF;
4614 }
4615 }
4616 out:
4617 return err;
4618 }
4619
4620 /*
4621 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4622 */
4623 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4624 struct nlattr *tb[],
4625 struct net_device *dev,
4626 const unsigned char *addr, u16 vid)
4627 {
4628 int err = -EINVAL;
4629
4630 /* If aging addresses are supported device will need to
4631 * implement its own handler for this.
4632 */
4633 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4634 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4635 return err;
4636 }
4637
4638 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4639 err = dev_uc_del(dev, addr);
4640 else if (is_multicast_ether_addr(addr))
4641 err = dev_mc_del(dev, addr);
4642
4643 return err;
4644 }
4645 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4646
4647 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4648 struct netlink_ext_ack *extack)
4649 {
4650 bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4651 struct net *net = sock_net(skb->sk);
4652 const struct net_device_ops *ops;
4653 struct ndmsg *ndm;
4654 struct nlattr *tb[NDA_MAX+1];
4655 struct net_device *dev;
4656 __u8 *addr = NULL;
4657 int err;
4658 u16 vid;
4659
4660 if (!netlink_capable(skb, CAP_NET_ADMIN))
4661 return -EPERM;
4662
4663 if (!del_bulk) {
4664 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4665 NULL, extack);
4666 } else {
4667 /* For bulk delete, the drivers will parse the message with
4668 * policy.
4669 */
4670 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
4671 }
4672 if (err < 0)
4673 return err;
4674
4675 ndm = nlmsg_data(nlh);
4676 if (ndm->ndm_ifindex == 0) {
4677 NL_SET_ERR_MSG(extack, "invalid ifindex");
4678 return -EINVAL;
4679 }
4680
4681 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4682 if (dev == NULL) {
4683 NL_SET_ERR_MSG(extack, "unknown ifindex");
4684 return -ENODEV;
4685 }
4686
4687 if (!del_bulk) {
4688 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4689 NL_SET_ERR_MSG(extack, "invalid address");
4690 return -EINVAL;
4691 }
4692 addr = nla_data(tb[NDA_LLADDR]);
4693
4694 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4695 if (err)
4696 return err;
4697 }
4698
4699 if (dev->type != ARPHRD_ETHER) {
4700 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4701 return -EINVAL;
4702 }
4703
4704 err = -EOPNOTSUPP;
4705
4706 /* Support fdb on master device the net/bridge default case */
4707 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4708 netif_is_bridge_port(dev)) {
4709 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4710 bool notified = false;
4711
4712 ops = br_dev->netdev_ops;
4713 if (!del_bulk) {
4714 if (ops->ndo_fdb_del)
4715 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid,
4716 &notified, extack);
4717 } else {
4718 if (ops->ndo_fdb_del_bulk)
4719 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4720 }
4721
4722 if (err)
4723 goto out;
4724 else
4725 ndm->ndm_flags &= ~NTF_MASTER;
4726 }
4727
4728 /* Embedded bridge, macvlan, and any other device support */
4729 if (ndm->ndm_flags & NTF_SELF) {
4730 bool notified = false;
4731
4732 ops = dev->netdev_ops;
4733 if (!del_bulk) {
4734 if (ops->ndo_fdb_del)
4735 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid,
4736 &notified, extack);
4737 else
4738 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4739 } else {
4740 /* in case err was cleared by NTF_MASTER call */
4741 err = -EOPNOTSUPP;
4742 if (ops->ndo_fdb_del_bulk)
4743 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4744 }
4745
4746 if (!err) {
4747 if (!del_bulk && !notified)
4748 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4749 ndm->ndm_state);
4750 ndm->ndm_flags &= ~NTF_SELF;
4751 }
4752 }
4753 out:
4754 return err;
4755 }
4756
4757 static int nlmsg_populate_fdb(struct sk_buff *skb,
4758 struct netlink_callback *cb,
4759 struct net_device *dev,
4760 int *idx,
4761 struct netdev_hw_addr_list *list)
4762 {
4763 struct netdev_hw_addr *ha;
4764 int err;
4765 u32 portid, seq;
4766
4767 portid = NETLINK_CB(cb->skb).portid;
4768 seq = cb->nlh->nlmsg_seq;
4769
4770 list_for_each_entry(ha, &list->list, list) {
4771 if (*idx < cb->args[2])
4772 goto skip;
4773
4774 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4775 portid, seq,
4776 RTM_NEWNEIGH, NTF_SELF,
4777 NLM_F_MULTI, NUD_PERMANENT);
4778 if (err < 0)
4779 return err;
4780 skip:
4781 *idx += 1;
4782 }
4783 return 0;
4784 }
4785
4786 /**
4787 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4788 * @skb: socket buffer to store message in
4789 * @cb: netlink callback
4790 * @dev: netdevice
4791 * @filter_dev: ignored
4792 * @idx: the number of FDB table entries dumped is added to *@idx
4793 *
4794 * Default netdevice operation to dump the existing unicast address list.
4795 * Returns number of addresses from list put in skb.
4796 */
4797 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4798 struct netlink_callback *cb,
4799 struct net_device *dev,
4800 struct net_device *filter_dev,
4801 int *idx)
4802 {
4803 int err;
4804
4805 if (dev->type != ARPHRD_ETHER)
4806 return -EINVAL;
4807
4808 netif_addr_lock_bh(dev);
4809 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4810 if (err)
4811 goto out;
4812 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4813 out:
4814 netif_addr_unlock_bh(dev);
4815 return err;
4816 }
4817 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4818
4819 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4820 int *br_idx, int *brport_idx,
4821 struct netlink_ext_ack *extack)
4822 {
4823 struct nlattr *tb[NDA_MAX + 1];
4824 struct ndmsg *ndm;
4825 int err, i;
4826
4827 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4828 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4829 return -EINVAL;
4830 }
4831
4832 ndm = nlmsg_data(nlh);
4833 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4834 ndm->ndm_flags || ndm->ndm_type) {
4835 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4836 return -EINVAL;
4837 }
4838
4839 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4840 NDA_MAX, NULL, extack);
4841 if (err < 0)
4842 return err;
4843
4844 *brport_idx = ndm->ndm_ifindex;
4845 for (i = 0; i <= NDA_MAX; ++i) {
4846 if (!tb[i])
4847 continue;
4848
4849 switch (i) {
4850 case NDA_IFINDEX:
4851 if (nla_len(tb[i]) != sizeof(u32)) {
4852 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4853 return -EINVAL;
4854 }
4855 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4856 break;
4857 case NDA_MASTER:
4858 if (nla_len(tb[i]) != sizeof(u32)) {
4859 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4860 return -EINVAL;
4861 }
4862 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4863 break;
4864 default:
4865 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4866 return -EINVAL;
4867 }
4868 }
4869
4870 return 0;
4871 }
4872
4873 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4874 int *br_idx, int *brport_idx,
4875 struct netlink_ext_ack *extack)
4876 {
4877 struct nlattr *tb[IFLA_MAX+1];
4878 int err;
4879
4880 /* A hack to preserve kernel<->userspace interface.
4881 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4882 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4883 * So, check for ndmsg with an optional u32 attribute (not used here).
4884 * Fortunately these sizes don't conflict with the size of ifinfomsg
4885 * with an optional attribute.
4886 */
4887 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4888 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4889 nla_attr_size(sizeof(u32)))) {
4890 struct ifinfomsg *ifm;
4891
4892 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4893 tb, IFLA_MAX, ifla_policy,
4894 extack);
4895 if (err < 0) {
4896 return -EINVAL;
4897 } else if (err == 0) {
4898 if (tb[IFLA_MASTER])
4899 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4900 }
4901
4902 ifm = nlmsg_data(nlh);
4903 *brport_idx = ifm->ifi_index;
4904 }
4905 return 0;
4906 }
4907
4908 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4909 {
4910 struct net_device *dev;
4911 struct net_device *br_dev = NULL;
4912 const struct net_device_ops *ops = NULL;
4913 const struct net_device_ops *cops = NULL;
4914 struct net *net = sock_net(skb->sk);
4915 struct hlist_head *head;
4916 int brport_idx = 0;
4917 int br_idx = 0;
4918 int h, s_h;
4919 int idx = 0, s_idx;
4920 int err = 0;
4921 int fidx = 0;
4922
4923 if (cb->strict_check)
4924 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4925 cb->extack);
4926 else
4927 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4928 cb->extack);
4929 if (err < 0)
4930 return err;
4931
4932 if (br_idx) {
4933 br_dev = __dev_get_by_index(net, br_idx);
4934 if (!br_dev)
4935 return -ENODEV;
4936
4937 ops = br_dev->netdev_ops;
4938 }
4939
4940 s_h = cb->args[0];
4941 s_idx = cb->args[1];
4942
4943 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4944 idx = 0;
4945 head = &net->dev_index_head[h];
4946 hlist_for_each_entry(dev, head, index_hlist) {
4947
4948 if (brport_idx && (dev->ifindex != brport_idx))
4949 continue;
4950
4951 if (!br_idx) { /* user did not specify a specific bridge */
4952 if (netif_is_bridge_port(dev)) {
4953 br_dev = netdev_master_upper_dev_get(dev);
4954 cops = br_dev->netdev_ops;
4955 }
4956 } else {
4957 if (dev != br_dev &&
4958 !netif_is_bridge_port(dev))
4959 continue;
4960
4961 if (br_dev != netdev_master_upper_dev_get(dev) &&
4962 !netif_is_bridge_master(dev))
4963 continue;
4964 cops = ops;
4965 }
4966
4967 if (idx < s_idx)
4968 goto cont;
4969
4970 if (netif_is_bridge_port(dev)) {
4971 if (cops && cops->ndo_fdb_dump) {
4972 err = cops->ndo_fdb_dump(skb, cb,
4973 br_dev, dev,
4974 &fidx);
4975 if (err == -EMSGSIZE)
4976 goto out;
4977 }
4978 }
4979
4980 if (dev->netdev_ops->ndo_fdb_dump)
4981 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4982 dev, NULL,
4983 &fidx);
4984 else
4985 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4986 &fidx);
4987 if (err == -EMSGSIZE)
4988 goto out;
4989
4990 cops = NULL;
4991
4992 /* reset fdb offset to 0 for rest of the interfaces */
4993 cb->args[2] = 0;
4994 fidx = 0;
4995 cont:
4996 idx++;
4997 }
4998 }
4999
5000 out:
5001 cb->args[0] = h;
5002 cb->args[1] = idx;
5003 cb->args[2] = fidx;
5004
5005 return skb->len;
5006 }
5007
5008 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
5009 struct nlattr **tb, u8 *ndm_flags,
5010 int *br_idx, int *brport_idx, u8 **addr,
5011 u16 *vid, struct netlink_ext_ack *extack)
5012 {
5013 struct ndmsg *ndm;
5014 int err, i;
5015
5016 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
5017 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
5018 return -EINVAL;
5019 }
5020
5021 ndm = nlmsg_data(nlh);
5022 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
5023 ndm->ndm_type) {
5024 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
5025 return -EINVAL;
5026 }
5027
5028 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
5029 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
5030 return -EINVAL;
5031 }
5032
5033 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
5034 NDA_MAX, nda_policy, extack);
5035 if (err < 0)
5036 return err;
5037
5038 *ndm_flags = ndm->ndm_flags;
5039 *brport_idx = ndm->ndm_ifindex;
5040 for (i = 0; i <= NDA_MAX; ++i) {
5041 if (!tb[i])
5042 continue;
5043
5044 switch (i) {
5045 case NDA_MASTER:
5046 *br_idx = nla_get_u32(tb[i]);
5047 break;
5048 case NDA_LLADDR:
5049 if (nla_len(tb[i]) != ETH_ALEN) {
5050 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
5051 return -EINVAL;
5052 }
5053 *addr = nla_data(tb[i]);
5054 break;
5055 case NDA_VLAN:
5056 err = fdb_vid_parse(tb[i], vid, extack);
5057 if (err)
5058 return err;
5059 break;
5060 case NDA_VNI:
5061 break;
5062 default:
5063 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
5064 return -EINVAL;
5065 }
5066 }
5067
5068 return 0;
5069 }
5070
5071 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5072 struct netlink_ext_ack *extack)
5073 {
5074 struct net_device *dev = NULL, *br_dev = NULL;
5075 const struct net_device_ops *ops = NULL;
5076 struct net *net = sock_net(in_skb->sk);
5077 struct nlattr *tb[NDA_MAX + 1];
5078 struct sk_buff *skb;
5079 int brport_idx = 0;
5080 u8 ndm_flags = 0;
5081 int br_idx = 0;
5082 u8 *addr = NULL;
5083 u16 vid = 0;
5084 int err;
5085
5086 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
5087 &brport_idx, &addr, &vid, extack);
5088 if (err < 0)
5089 return err;
5090
5091 if (!addr) {
5092 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
5093 return -EINVAL;
5094 }
5095
5096 if (brport_idx) {
5097 dev = __dev_get_by_index(net, brport_idx);
5098 if (!dev) {
5099 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
5100 return -ENODEV;
5101 }
5102 }
5103
5104 if (br_idx) {
5105 if (dev) {
5106 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
5107 return -EINVAL;
5108 }
5109
5110 br_dev = __dev_get_by_index(net, br_idx);
5111 if (!br_dev) {
5112 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
5113 return -EINVAL;
5114 }
5115 ops = br_dev->netdev_ops;
5116 }
5117
5118 if (dev) {
5119 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
5120 if (!netif_is_bridge_port(dev)) {
5121 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
5122 return -EINVAL;
5123 }
5124 br_dev = netdev_master_upper_dev_get(dev);
5125 if (!br_dev) {
5126 NL_SET_ERR_MSG(extack, "Master of device not found");
5127 return -EINVAL;
5128 }
5129 ops = br_dev->netdev_ops;
5130 } else {
5131 if (!(ndm_flags & NTF_SELF)) {
5132 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
5133 return -EINVAL;
5134 }
5135 ops = dev->netdev_ops;
5136 }
5137 }
5138
5139 if (!br_dev && !dev) {
5140 NL_SET_ERR_MSG(extack, "No device specified");
5141 return -ENODEV;
5142 }
5143
5144 if (!ops || !ops->ndo_fdb_get) {
5145 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
5146 return -EOPNOTSUPP;
5147 }
5148
5149 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
5150 if (!skb)
5151 return -ENOBUFS;
5152
5153 if (br_dev)
5154 dev = br_dev;
5155 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
5156 NETLINK_CB(in_skb).portid,
5157 nlh->nlmsg_seq, extack);
5158 if (err)
5159 goto out;
5160
5161 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5162 out:
5163 kfree_skb(skb);
5164 return err;
5165 }
5166
5167 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
5168 unsigned int attrnum, unsigned int flag)
5169 {
5170 if (mask & flag)
5171 return nla_put_u8(skb, attrnum, !!(flags & flag));
5172 return 0;
5173 }
5174
5175 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
5176 struct net_device *dev, u16 mode,
5177 u32 flags, u32 mask, int nlflags,
5178 u32 filter_mask,
5179 int (*vlan_fill)(struct sk_buff *skb,
5180 struct net_device *dev,
5181 u32 filter_mask))
5182 {
5183 struct nlmsghdr *nlh;
5184 struct ifinfomsg *ifm;
5185 struct nlattr *br_afspec;
5186 struct nlattr *protinfo;
5187 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
5188 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5189 int err = 0;
5190
5191 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
5192 if (nlh == NULL)
5193 return -EMSGSIZE;
5194
5195 ifm = nlmsg_data(nlh);
5196 ifm->ifi_family = AF_BRIDGE;
5197 ifm->__ifi_pad = 0;
5198 ifm->ifi_type = dev->type;
5199 ifm->ifi_index = dev->ifindex;
5200 ifm->ifi_flags = dev_get_flags(dev);
5201 ifm->ifi_change = 0;
5202
5203
5204 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5205 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5206 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
5207 (br_dev &&
5208 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
5209 (dev->addr_len &&
5210 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5211 (dev->ifindex != dev_get_iflink(dev) &&
5212 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
5213 goto nla_put_failure;
5214
5215 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
5216 if (!br_afspec)
5217 goto nla_put_failure;
5218
5219 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
5220 nla_nest_cancel(skb, br_afspec);
5221 goto nla_put_failure;
5222 }
5223
5224 if (mode != BRIDGE_MODE_UNDEF) {
5225 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
5226 nla_nest_cancel(skb, br_afspec);
5227 goto nla_put_failure;
5228 }
5229 }
5230 if (vlan_fill) {
5231 err = vlan_fill(skb, dev, filter_mask);
5232 if (err) {
5233 nla_nest_cancel(skb, br_afspec);
5234 goto nla_put_failure;
5235 }
5236 }
5237 nla_nest_end(skb, br_afspec);
5238
5239 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
5240 if (!protinfo)
5241 goto nla_put_failure;
5242
5243 if (brport_nla_put_flag(skb, flags, mask,
5244 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
5245 brport_nla_put_flag(skb, flags, mask,
5246 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
5247 brport_nla_put_flag(skb, flags, mask,
5248 IFLA_BRPORT_FAST_LEAVE,
5249 BR_MULTICAST_FAST_LEAVE) ||
5250 brport_nla_put_flag(skb, flags, mask,
5251 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
5252 brport_nla_put_flag(skb, flags, mask,
5253 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
5254 brport_nla_put_flag(skb, flags, mask,
5255 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
5256 brport_nla_put_flag(skb, flags, mask,
5257 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
5258 brport_nla_put_flag(skb, flags, mask,
5259 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
5260 brport_nla_put_flag(skb, flags, mask,
5261 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
5262 brport_nla_put_flag(skb, flags, mask,
5263 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
5264 nla_nest_cancel(skb, protinfo);
5265 goto nla_put_failure;
5266 }
5267
5268 nla_nest_end(skb, protinfo);
5269
5270 nlmsg_end(skb, nlh);
5271 return 0;
5272 nla_put_failure:
5273 nlmsg_cancel(skb, nlh);
5274 return err ? err : -EMSGSIZE;
5275 }
5276 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
5277
5278 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
5279 bool strict_check, u32 *filter_mask,
5280 struct netlink_ext_ack *extack)
5281 {
5282 struct nlattr *tb[IFLA_MAX+1];
5283 int err, i;
5284
5285 if (strict_check) {
5286 struct ifinfomsg *ifm;
5287
5288 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5289 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
5290 return -EINVAL;
5291 }
5292
5293 ifm = nlmsg_data(nlh);
5294 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5295 ifm->ifi_change || ifm->ifi_index) {
5296 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
5297 return -EINVAL;
5298 }
5299
5300 err = nlmsg_parse_deprecated_strict(nlh,
5301 sizeof(struct ifinfomsg),
5302 tb, IFLA_MAX, ifla_policy,
5303 extack);
5304 } else {
5305 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
5306 tb, IFLA_MAX, ifla_policy,
5307 extack);
5308 }
5309 if (err < 0)
5310 return err;
5311
5312 /* new attributes should only be added with strict checking */
5313 for (i = 0; i <= IFLA_MAX; ++i) {
5314 if (!tb[i])
5315 continue;
5316
5317 switch (i) {
5318 case IFLA_EXT_MASK:
5319 *filter_mask = nla_get_u32(tb[i]);
5320 break;
5321 default:
5322 if (strict_check) {
5323 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
5324 return -EINVAL;
5325 }
5326 }
5327 }
5328
5329 return 0;
5330 }
5331
5332 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
5333 {
5334 const struct nlmsghdr *nlh = cb->nlh;
5335 struct net *net = sock_net(skb->sk);
5336 struct net_device *dev;
5337 int idx = 0;
5338 u32 portid = NETLINK_CB(cb->skb).portid;
5339 u32 seq = nlh->nlmsg_seq;
5340 u32 filter_mask = 0;
5341 int err;
5342
5343 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5344 cb->extack);
5345 if (err < 0 && cb->strict_check)
5346 return err;
5347
5348 rcu_read_lock();
5349 for_each_netdev_rcu(net, dev) {
5350 const struct net_device_ops *ops = dev->netdev_ops;
5351 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5352
5353 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5354 if (idx >= cb->args[0]) {
5355 err = br_dev->netdev_ops->ndo_bridge_getlink(
5356 skb, portid, seq, dev,
5357 filter_mask, NLM_F_MULTI);
5358 if (err < 0 && err != -EOPNOTSUPP) {
5359 if (likely(skb->len))
5360 break;
5361
5362 goto out_err;
5363 }
5364 }
5365 idx++;
5366 }
5367
5368 if (ops->ndo_bridge_getlink) {
5369 if (idx >= cb->args[0]) {
5370 err = ops->ndo_bridge_getlink(skb, portid,
5371 seq, dev,
5372 filter_mask,
5373 NLM_F_MULTI);
5374 if (err < 0 && err != -EOPNOTSUPP) {
5375 if (likely(skb->len))
5376 break;
5377
5378 goto out_err;
5379 }
5380 }
5381 idx++;
5382 }
5383 }
5384 err = skb->len;
5385 out_err:
5386 rcu_read_unlock();
5387 cb->args[0] = idx;
5388
5389 return err;
5390 }
5391
5392 static inline size_t bridge_nlmsg_size(void)
5393 {
5394 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5395 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5396 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5397 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
5398 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
5399 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
5400 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
5401 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
5402 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5403 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
5404 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
5405 }
5406
5407 static int rtnl_bridge_notify(struct net_device *dev)
5408 {
5409 struct net *net = dev_net(dev);
5410 struct sk_buff *skb;
5411 int err = -EOPNOTSUPP;
5412
5413 if (!dev->netdev_ops->ndo_bridge_getlink)
5414 return 0;
5415
5416 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5417 if (!skb) {
5418 err = -ENOMEM;
5419 goto errout;
5420 }
5421
5422 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5423 if (err < 0)
5424 goto errout;
5425
5426 /* Notification info is only filled for bridge ports, not the bridge
5427 * device itself. Therefore, a zero notification length is valid and
5428 * should not result in an error.
5429 */
5430 if (!skb->len)
5431 goto errout;
5432
5433 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5434 return 0;
5435 errout:
5436 WARN_ON(err == -EMSGSIZE);
5437 kfree_skb(skb);
5438 if (err)
5439 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5440 return err;
5441 }
5442
5443 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5444 struct netlink_ext_ack *extack)
5445 {
5446 struct net *net = sock_net(skb->sk);
5447 struct ifinfomsg *ifm;
5448 struct net_device *dev;
5449 struct nlattr *br_spec, *attr, *br_flags_attr = NULL;
5450 int rem, err = -EOPNOTSUPP;
5451 u16 flags = 0;
5452
5453 if (nlmsg_len(nlh) < sizeof(*ifm))
5454 return -EINVAL;
5455
5456 ifm = nlmsg_data(nlh);
5457 if (ifm->ifi_family != AF_BRIDGE)
5458 return -EPFNOSUPPORT;
5459
5460 dev = __dev_get_by_index(net, ifm->ifi_index);
5461 if (!dev) {
5462 NL_SET_ERR_MSG(extack, "unknown ifindex");
5463 return -ENODEV;
5464 }
5465
5466 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5467 if (br_spec) {
5468 nla_for_each_nested(attr, br_spec, rem) {
5469 if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !br_flags_attr) {
5470 if (nla_len(attr) < sizeof(flags))
5471 return -EINVAL;
5472
5473 br_flags_attr = attr;
5474 flags = nla_get_u16(attr);
5475 }
5476
5477 if (nla_type(attr) == IFLA_BRIDGE_MODE) {
5478 if (nla_len(attr) < sizeof(u16))
5479 return -EINVAL;
5480 }
5481 }
5482 }
5483
5484 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5485 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5486
5487 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5488 err = -EOPNOTSUPP;
5489 goto out;
5490 }
5491
5492 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5493 extack);
5494 if (err)
5495 goto out;
5496
5497 flags &= ~BRIDGE_FLAGS_MASTER;
5498 }
5499
5500 if ((flags & BRIDGE_FLAGS_SELF)) {
5501 if (!dev->netdev_ops->ndo_bridge_setlink)
5502 err = -EOPNOTSUPP;
5503 else
5504 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5505 flags,
5506 extack);
5507 if (!err) {
5508 flags &= ~BRIDGE_FLAGS_SELF;
5509
5510 /* Generate event to notify upper layer of bridge
5511 * change
5512 */
5513 err = rtnl_bridge_notify(dev);
5514 }
5515 }
5516
5517 if (br_flags_attr)
5518 memcpy(nla_data(br_flags_attr), &flags, sizeof(flags));
5519 out:
5520 return err;
5521 }
5522
5523 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5524 struct netlink_ext_ack *extack)
5525 {
5526 struct net *net = sock_net(skb->sk);
5527 struct ifinfomsg *ifm;
5528 struct net_device *dev;
5529 struct nlattr *br_spec, *attr = NULL;
5530 int rem, err = -EOPNOTSUPP;
5531 u16 flags = 0;
5532 bool have_flags = false;
5533
5534 if (nlmsg_len(nlh) < sizeof(*ifm))
5535 return -EINVAL;
5536
5537 ifm = nlmsg_data(nlh);
5538 if (ifm->ifi_family != AF_BRIDGE)
5539 return -EPFNOSUPPORT;
5540
5541 dev = __dev_get_by_index(net, ifm->ifi_index);
5542 if (!dev) {
5543 NL_SET_ERR_MSG(extack, "unknown ifindex");
5544 return -ENODEV;
5545 }
5546
5547 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5548 if (br_spec) {
5549 nla_for_each_nested_type(attr, IFLA_BRIDGE_FLAGS, br_spec,
5550 rem) {
5551 if (nla_len(attr) < sizeof(flags))
5552 return -EINVAL;
5553
5554 have_flags = true;
5555 flags = nla_get_u16(attr);
5556 break;
5557 }
5558 }
5559
5560 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5561 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5562
5563 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5564 err = -EOPNOTSUPP;
5565 goto out;
5566 }
5567
5568 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5569 if (err)
5570 goto out;
5571
5572 flags &= ~BRIDGE_FLAGS_MASTER;
5573 }
5574
5575 if ((flags & BRIDGE_FLAGS_SELF)) {
5576 if (!dev->netdev_ops->ndo_bridge_dellink)
5577 err = -EOPNOTSUPP;
5578 else
5579 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5580 flags);
5581
5582 if (!err) {
5583 flags &= ~BRIDGE_FLAGS_SELF;
5584
5585 /* Generate event to notify upper layer of bridge
5586 * change
5587 */
5588 err = rtnl_bridge_notify(dev);
5589 }
5590 }
5591
5592 if (have_flags)
5593 memcpy(nla_data(attr), &flags, sizeof(flags));
5594 out:
5595 return err;
5596 }
5597
5598 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5599 {
5600 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5601 (!idxattr || idxattr == attrid);
5602 }
5603
5604 static bool
5605 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5606 {
5607 return dev->netdev_ops &&
5608 dev->netdev_ops->ndo_has_offload_stats &&
5609 dev->netdev_ops->ndo_get_offload_stats &&
5610 dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5611 }
5612
5613 static unsigned int
5614 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5615 {
5616 return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5617 sizeof(struct rtnl_link_stats64) : 0;
5618 }
5619
5620 static int
5621 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5622 struct sk_buff *skb)
5623 {
5624 unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5625 struct nlattr *attr = NULL;
5626 void *attr_data;
5627 int err;
5628
5629 if (!size)
5630 return -ENODATA;
5631
5632 attr = nla_reserve_64bit(skb, attr_id, size,
5633 IFLA_OFFLOAD_XSTATS_UNSPEC);
5634 if (!attr)
5635 return -EMSGSIZE;
5636
5637 attr_data = nla_data(attr);
5638 memset(attr_data, 0, size);
5639
5640 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5641 if (err)
5642 return err;
5643
5644 return 0;
5645 }
5646
5647 static unsigned int
5648 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5649 enum netdev_offload_xstats_type type)
5650 {
5651 bool enabled = netdev_offload_xstats_enabled(dev, type);
5652
5653 return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5654 }
5655
5656 struct rtnl_offload_xstats_request_used {
5657 bool request;
5658 bool used;
5659 };
5660
5661 static int
5662 rtnl_offload_xstats_get_stats(struct net_device *dev,
5663 enum netdev_offload_xstats_type type,
5664 struct rtnl_offload_xstats_request_used *ru,
5665 struct rtnl_hw_stats64 *stats,
5666 struct netlink_ext_ack *extack)
5667 {
5668 bool request;
5669 bool used;
5670 int err;
5671
5672 request = netdev_offload_xstats_enabled(dev, type);
5673 if (!request) {
5674 used = false;
5675 goto out;
5676 }
5677
5678 err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5679 if (err)
5680 return err;
5681
5682 out:
5683 if (ru) {
5684 ru->request = request;
5685 ru->used = used;
5686 }
5687 return 0;
5688 }
5689
5690 static int
5691 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5692 struct rtnl_offload_xstats_request_used *ru)
5693 {
5694 struct nlattr *nest;
5695
5696 nest = nla_nest_start(skb, attr_id);
5697 if (!nest)
5698 return -EMSGSIZE;
5699
5700 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5701 goto nla_put_failure;
5702
5703 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5704 goto nla_put_failure;
5705
5706 nla_nest_end(skb, nest);
5707 return 0;
5708
5709 nla_put_failure:
5710 nla_nest_cancel(skb, nest);
5711 return -EMSGSIZE;
5712 }
5713
5714 static int
5715 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5716 struct netlink_ext_ack *extack)
5717 {
5718 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5719 struct rtnl_offload_xstats_request_used ru_l3;
5720 struct nlattr *nest;
5721 int err;
5722
5723 err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5724 if (err)
5725 return err;
5726
5727 nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5728 if (!nest)
5729 return -EMSGSIZE;
5730
5731 if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5732 IFLA_OFFLOAD_XSTATS_L3_STATS,
5733 &ru_l3))
5734 goto nla_put_failure;
5735
5736 nla_nest_end(skb, nest);
5737 return 0;
5738
5739 nla_put_failure:
5740 nla_nest_cancel(skb, nest);
5741 return -EMSGSIZE;
5742 }
5743
5744 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5745 int *prividx, u32 off_filter_mask,
5746 struct netlink_ext_ack *extack)
5747 {
5748 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5749 int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5750 int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5751 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5752 bool have_data = false;
5753 int err;
5754
5755 if (*prividx <= attr_id_cpu_hit &&
5756 (off_filter_mask &
5757 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5758 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5759 if (!err) {
5760 have_data = true;
5761 } else if (err != -ENODATA) {
5762 *prividx = attr_id_cpu_hit;
5763 return err;
5764 }
5765 }
5766
5767 if (*prividx <= attr_id_hw_s_info &&
5768 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5769 *prividx = attr_id_hw_s_info;
5770
5771 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5772 if (err)
5773 return err;
5774
5775 have_data = true;
5776 *prividx = 0;
5777 }
5778
5779 if (*prividx <= attr_id_l3_stats &&
5780 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5781 unsigned int size_l3;
5782 struct nlattr *attr;
5783
5784 *prividx = attr_id_l3_stats;
5785
5786 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5787 if (!size_l3)
5788 goto skip_l3_stats;
5789 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5790 IFLA_OFFLOAD_XSTATS_UNSPEC);
5791 if (!attr)
5792 return -EMSGSIZE;
5793
5794 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5795 nla_data(attr), extack);
5796 if (err)
5797 return err;
5798
5799 have_data = true;
5800 skip_l3_stats:
5801 *prividx = 0;
5802 }
5803
5804 if (!have_data)
5805 return -ENODATA;
5806
5807 *prividx = 0;
5808 return 0;
5809 }
5810
5811 static unsigned int
5812 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5813 enum netdev_offload_xstats_type type)
5814 {
5815 return nla_total_size(0) +
5816 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5817 nla_total_size(sizeof(u8)) +
5818 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5819 nla_total_size(sizeof(u8)) +
5820 0;
5821 }
5822
5823 static unsigned int
5824 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5825 {
5826 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5827
5828 return nla_total_size(0) +
5829 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5830 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5831 0;
5832 }
5833
5834 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5835 u32 off_filter_mask)
5836 {
5837 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5838 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5839 int nla_size = 0;
5840 int size;
5841
5842 if (off_filter_mask &
5843 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5844 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5845 nla_size += nla_total_size_64bit(size);
5846 }
5847
5848 if (off_filter_mask &
5849 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5850 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5851
5852 if (off_filter_mask &
5853 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5854 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5855 nla_size += nla_total_size_64bit(size);
5856 }
5857
5858 if (nla_size != 0)
5859 nla_size += nla_total_size(0);
5860
5861 return nla_size;
5862 }
5863
5864 struct rtnl_stats_dump_filters {
5865 /* mask[0] filters outer attributes. Then individual nests have their
5866 * filtering mask at the index of the nested attribute.
5867 */
5868 u32 mask[IFLA_STATS_MAX + 1];
5869 };
5870
5871 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5872 int type, u32 pid, u32 seq, u32 change,
5873 unsigned int flags,
5874 const struct rtnl_stats_dump_filters *filters,
5875 int *idxattr, int *prividx,
5876 struct netlink_ext_ack *extack)
5877 {
5878 unsigned int filter_mask = filters->mask[0];
5879 struct if_stats_msg *ifsm;
5880 struct nlmsghdr *nlh;
5881 struct nlattr *attr;
5882 int s_prividx = *prividx;
5883 int err;
5884
5885 ASSERT_RTNL();
5886
5887 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5888 if (!nlh)
5889 return -EMSGSIZE;
5890
5891 ifsm = nlmsg_data(nlh);
5892 ifsm->family = PF_UNSPEC;
5893 ifsm->pad1 = 0;
5894 ifsm->pad2 = 0;
5895 ifsm->ifindex = dev->ifindex;
5896 ifsm->filter_mask = filter_mask;
5897
5898 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5899 struct rtnl_link_stats64 *sp;
5900
5901 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5902 sizeof(struct rtnl_link_stats64),
5903 IFLA_STATS_UNSPEC);
5904 if (!attr) {
5905 err = -EMSGSIZE;
5906 goto nla_put_failure;
5907 }
5908
5909 sp = nla_data(attr);
5910 dev_get_stats(dev, sp);
5911 }
5912
5913 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5914 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5915
5916 if (ops && ops->fill_linkxstats) {
5917 *idxattr = IFLA_STATS_LINK_XSTATS;
5918 attr = nla_nest_start_noflag(skb,
5919 IFLA_STATS_LINK_XSTATS);
5920 if (!attr) {
5921 err = -EMSGSIZE;
5922 goto nla_put_failure;
5923 }
5924
5925 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5926 nla_nest_end(skb, attr);
5927 if (err)
5928 goto nla_put_failure;
5929 *idxattr = 0;
5930 }
5931 }
5932
5933 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5934 *idxattr)) {
5935 const struct rtnl_link_ops *ops = NULL;
5936 const struct net_device *master;
5937
5938 master = netdev_master_upper_dev_get(dev);
5939 if (master)
5940 ops = master->rtnl_link_ops;
5941 if (ops && ops->fill_linkxstats) {
5942 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5943 attr = nla_nest_start_noflag(skb,
5944 IFLA_STATS_LINK_XSTATS_SLAVE);
5945 if (!attr) {
5946 err = -EMSGSIZE;
5947 goto nla_put_failure;
5948 }
5949
5950 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5951 nla_nest_end(skb, attr);
5952 if (err)
5953 goto nla_put_failure;
5954 *idxattr = 0;
5955 }
5956 }
5957
5958 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5959 *idxattr)) {
5960 u32 off_filter_mask;
5961
5962 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5963 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5964 attr = nla_nest_start_noflag(skb,
5965 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5966 if (!attr) {
5967 err = -EMSGSIZE;
5968 goto nla_put_failure;
5969 }
5970
5971 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5972 off_filter_mask, extack);
5973 if (err == -ENODATA)
5974 nla_nest_cancel(skb, attr);
5975 else
5976 nla_nest_end(skb, attr);
5977
5978 if (err && err != -ENODATA)
5979 goto nla_put_failure;
5980 *idxattr = 0;
5981 }
5982
5983 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5984 struct rtnl_af_ops *af_ops;
5985
5986 *idxattr = IFLA_STATS_AF_SPEC;
5987 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5988 if (!attr) {
5989 err = -EMSGSIZE;
5990 goto nla_put_failure;
5991 }
5992
5993 rcu_read_lock();
5994 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5995 if (af_ops->fill_stats_af) {
5996 struct nlattr *af;
5997
5998 af = nla_nest_start_noflag(skb,
5999 af_ops->family);
6000 if (!af) {
6001 rcu_read_unlock();
6002 err = -EMSGSIZE;
6003 goto nla_put_failure;
6004 }
6005 err = af_ops->fill_stats_af(skb, dev);
6006
6007 if (err == -ENODATA) {
6008 nla_nest_cancel(skb, af);
6009 } else if (err < 0) {
6010 rcu_read_unlock();
6011 goto nla_put_failure;
6012 }
6013
6014 nla_nest_end(skb, af);
6015 }
6016 }
6017 rcu_read_unlock();
6018
6019 nla_nest_end(skb, attr);
6020
6021 *idxattr = 0;
6022 }
6023
6024 nlmsg_end(skb, nlh);
6025
6026 return 0;
6027
6028 nla_put_failure:
6029 /* not a multi message or no progress mean a real error */
6030 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
6031 nlmsg_cancel(skb, nlh);
6032 else
6033 nlmsg_end(skb, nlh);
6034
6035 return err;
6036 }
6037
6038 static size_t if_nlmsg_stats_size(const struct net_device *dev,
6039 const struct rtnl_stats_dump_filters *filters)
6040 {
6041 size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
6042 unsigned int filter_mask = filters->mask[0];
6043
6044 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
6045 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
6046
6047 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
6048 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
6049 int attr = IFLA_STATS_LINK_XSTATS;
6050
6051 if (ops && ops->get_linkxstats_size) {
6052 size += nla_total_size(ops->get_linkxstats_size(dev,
6053 attr));
6054 /* for IFLA_STATS_LINK_XSTATS */
6055 size += nla_total_size(0);
6056 }
6057 }
6058
6059 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
6060 struct net_device *_dev = (struct net_device *)dev;
6061 const struct rtnl_link_ops *ops = NULL;
6062 const struct net_device *master;
6063
6064 /* netdev_master_upper_dev_get can't take const */
6065 master = netdev_master_upper_dev_get(_dev);
6066 if (master)
6067 ops = master->rtnl_link_ops;
6068 if (ops && ops->get_linkxstats_size) {
6069 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
6070
6071 size += nla_total_size(ops->get_linkxstats_size(dev,
6072 attr));
6073 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
6074 size += nla_total_size(0);
6075 }
6076 }
6077
6078 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
6079 u32 off_filter_mask;
6080
6081 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
6082 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
6083 }
6084
6085 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
6086 struct rtnl_af_ops *af_ops;
6087
6088 /* for IFLA_STATS_AF_SPEC */
6089 size += nla_total_size(0);
6090
6091 rcu_read_lock();
6092 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
6093 if (af_ops->get_stats_af_size) {
6094 size += nla_total_size(
6095 af_ops->get_stats_af_size(dev));
6096
6097 /* for AF_* */
6098 size += nla_total_size(0);
6099 }
6100 }
6101 rcu_read_unlock();
6102 }
6103
6104 return size;
6105 }
6106
6107 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
6108
6109 static const struct nla_policy
6110 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
6111 [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
6112 NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
6113 };
6114
6115 static const struct nla_policy
6116 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
6117 [IFLA_STATS_GET_FILTERS] =
6118 NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
6119 };
6120
6121 static const struct nla_policy
6122 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
6123 [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
6124 };
6125
6126 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
6127 struct rtnl_stats_dump_filters *filters,
6128 struct netlink_ext_ack *extack)
6129 {
6130 struct nlattr *tb[IFLA_STATS_MAX + 1];
6131 int err;
6132 int at;
6133
6134 err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
6135 rtnl_stats_get_policy_filters, extack);
6136 if (err < 0)
6137 return err;
6138
6139 for (at = 1; at <= IFLA_STATS_MAX; at++) {
6140 if (tb[at]) {
6141 if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
6142 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
6143 return -EINVAL;
6144 }
6145 filters->mask[at] = nla_get_u32(tb[at]);
6146 }
6147 }
6148
6149 return 0;
6150 }
6151
6152 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
6153 u32 filter_mask,
6154 struct rtnl_stats_dump_filters *filters,
6155 struct netlink_ext_ack *extack)
6156 {
6157 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6158 int err;
6159 int i;
6160
6161 filters->mask[0] = filter_mask;
6162 for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
6163 filters->mask[i] = -1U;
6164
6165 err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
6166 IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
6167 if (err < 0)
6168 return err;
6169
6170 if (tb[IFLA_STATS_GET_FILTERS]) {
6171 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
6172 filters, extack);
6173 if (err)
6174 return err;
6175 }
6176
6177 return 0;
6178 }
6179
6180 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
6181 bool is_dump, struct netlink_ext_ack *extack)
6182 {
6183 struct if_stats_msg *ifsm;
6184
6185 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
6186 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
6187 return -EINVAL;
6188 }
6189
6190 if (!strict_check)
6191 return 0;
6192
6193 ifsm = nlmsg_data(nlh);
6194
6195 /* only requests using strict checks can pass data to influence
6196 * the dump. The legacy exception is filter_mask.
6197 */
6198 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
6199 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
6200 return -EINVAL;
6201 }
6202 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
6203 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
6204 return -EINVAL;
6205 }
6206
6207 return 0;
6208 }
6209
6210 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
6211 struct netlink_ext_ack *extack)
6212 {
6213 struct rtnl_stats_dump_filters filters;
6214 struct net *net = sock_net(skb->sk);
6215 struct net_device *dev = NULL;
6216 int idxattr = 0, prividx = 0;
6217 struct if_stats_msg *ifsm;
6218 struct sk_buff *nskb;
6219 int err;
6220
6221 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6222 false, extack);
6223 if (err)
6224 return err;
6225
6226 ifsm = nlmsg_data(nlh);
6227 if (ifsm->ifindex > 0)
6228 dev = __dev_get_by_index(net, ifsm->ifindex);
6229 else
6230 return -EINVAL;
6231
6232 if (!dev)
6233 return -ENODEV;
6234
6235 if (!ifsm->filter_mask) {
6236 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
6237 return -EINVAL;
6238 }
6239
6240 err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
6241 if (err)
6242 return err;
6243
6244 nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
6245 if (!nskb)
6246 return -ENOBUFS;
6247
6248 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
6249 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
6250 0, &filters, &idxattr, &prividx, extack);
6251 if (err < 0) {
6252 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
6253 WARN_ON(err == -EMSGSIZE);
6254 kfree_skb(nskb);
6255 } else {
6256 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
6257 }
6258
6259 return err;
6260 }
6261
6262 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
6263 {
6264 struct netlink_ext_ack *extack = cb->extack;
6265 struct rtnl_stats_dump_filters filters;
6266 struct net *net = sock_net(skb->sk);
6267 unsigned int flags = NLM_F_MULTI;
6268 struct if_stats_msg *ifsm;
6269 struct {
6270 unsigned long ifindex;
6271 int idxattr;
6272 int prividx;
6273 } *ctx = (void *)cb->ctx;
6274 struct net_device *dev;
6275 int err;
6276
6277 cb->seq = net->dev_base_seq;
6278
6279 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
6280 if (err)
6281 return err;
6282
6283 ifsm = nlmsg_data(cb->nlh);
6284 if (!ifsm->filter_mask) {
6285 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
6286 return -EINVAL;
6287 }
6288
6289 err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
6290 extack);
6291 if (err)
6292 return err;
6293
6294 for_each_netdev_dump(net, dev, ctx->ifindex) {
6295 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
6296 NETLINK_CB(cb->skb).portid,
6297 cb->nlh->nlmsg_seq, 0,
6298 flags, &filters,
6299 &ctx->idxattr, &ctx->prividx,
6300 extack);
6301 /* If we ran out of room on the first message,
6302 * we're in trouble.
6303 */
6304 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
6305
6306 if (err < 0)
6307 break;
6308 ctx->prividx = 0;
6309 ctx->idxattr = 0;
6310 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
6311 }
6312
6313 return err;
6314 }
6315
6316 void rtnl_offload_xstats_notify(struct net_device *dev)
6317 {
6318 struct rtnl_stats_dump_filters response_filters = {};
6319 struct net *net = dev_net(dev);
6320 int idxattr = 0, prividx = 0;
6321 struct sk_buff *skb;
6322 int err = -ENOBUFS;
6323
6324 ASSERT_RTNL();
6325
6326 response_filters.mask[0] |=
6327 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6328 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6329 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6330
6331 skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6332 GFP_KERNEL);
6333 if (!skb)
6334 goto errout;
6335
6336 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6337 &response_filters, &idxattr, &prividx, NULL);
6338 if (err < 0) {
6339 kfree_skb(skb);
6340 goto errout;
6341 }
6342
6343 rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6344 return;
6345
6346 errout:
6347 rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6348 }
6349 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6350
6351 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6352 struct netlink_ext_ack *extack)
6353 {
6354 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6355 struct rtnl_stats_dump_filters response_filters = {};
6356 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6357 struct net *net = sock_net(skb->sk);
6358 struct net_device *dev = NULL;
6359 struct if_stats_msg *ifsm;
6360 bool notify = false;
6361 int err;
6362
6363 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6364 false, extack);
6365 if (err)
6366 return err;
6367
6368 ifsm = nlmsg_data(nlh);
6369 if (ifsm->family != AF_UNSPEC) {
6370 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6371 return -EINVAL;
6372 }
6373
6374 if (ifsm->ifindex > 0)
6375 dev = __dev_get_by_index(net, ifsm->ifindex);
6376 else
6377 return -EINVAL;
6378
6379 if (!dev)
6380 return -ENODEV;
6381
6382 if (ifsm->filter_mask) {
6383 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6384 return -EINVAL;
6385 }
6386
6387 err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6388 ifla_stats_set_policy, extack);
6389 if (err < 0)
6390 return err;
6391
6392 if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6393 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6394
6395 if (req)
6396 err = netdev_offload_xstats_enable(dev, t_l3, extack);
6397 else
6398 err = netdev_offload_xstats_disable(dev, t_l3);
6399
6400 if (!err)
6401 notify = true;
6402 else if (err != -EALREADY)
6403 return err;
6404
6405 response_filters.mask[0] |=
6406 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6407 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6408 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6409 }
6410
6411 if (notify)
6412 rtnl_offload_xstats_notify(dev);
6413
6414 return 0;
6415 }
6416
6417 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6418 struct netlink_ext_ack *extack)
6419 {
6420 struct br_port_msg *bpm;
6421
6422 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6423 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6424 return -EINVAL;
6425 }
6426
6427 bpm = nlmsg_data(nlh);
6428 if (bpm->ifindex) {
6429 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6430 return -EINVAL;
6431 }
6432 if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6433 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6434 return -EINVAL;
6435 }
6436
6437 return 0;
6438 }
6439
6440 struct rtnl_mdb_dump_ctx {
6441 long idx;
6442 };
6443
6444 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6445 {
6446 struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6447 struct net *net = sock_net(skb->sk);
6448 struct net_device *dev;
6449 int idx, s_idx;
6450 int err;
6451
6452 NL_ASSERT_CTX_FITS(struct rtnl_mdb_dump_ctx);
6453
6454 if (cb->strict_check) {
6455 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6456 if (err)
6457 return err;
6458 }
6459
6460 s_idx = ctx->idx;
6461 idx = 0;
6462
6463 for_each_netdev(net, dev) {
6464 if (idx < s_idx)
6465 goto skip;
6466 if (!dev->netdev_ops->ndo_mdb_dump)
6467 goto skip;
6468
6469 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6470 if (err == -EMSGSIZE)
6471 goto out;
6472 /* Moving on to next device, reset markers and sequence
6473 * counters since they are all maintained per-device.
6474 */
6475 memset(cb->ctx, 0, sizeof(cb->ctx));
6476 cb->prev_seq = 0;
6477 cb->seq = 0;
6478 skip:
6479 idx++;
6480 }
6481
6482 out:
6483 ctx->idx = idx;
6484 return skb->len;
6485 }
6486
6487 static int rtnl_validate_mdb_entry_get(const struct nlattr *attr,
6488 struct netlink_ext_ack *extack)
6489 {
6490 struct br_mdb_entry *entry = nla_data(attr);
6491
6492 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6493 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6494 return -EINVAL;
6495 }
6496
6497 if (entry->ifindex) {
6498 NL_SET_ERR_MSG(extack, "Entry ifindex cannot be specified");
6499 return -EINVAL;
6500 }
6501
6502 if (entry->state) {
6503 NL_SET_ERR_MSG(extack, "Entry state cannot be specified");
6504 return -EINVAL;
6505 }
6506
6507 if (entry->flags) {
6508 NL_SET_ERR_MSG(extack, "Entry flags cannot be specified");
6509 return -EINVAL;
6510 }
6511
6512 if (entry->vid >= VLAN_VID_MASK) {
6513 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6514 return -EINVAL;
6515 }
6516
6517 if (entry->addr.proto != htons(ETH_P_IP) &&
6518 entry->addr.proto != htons(ETH_P_IPV6) &&
6519 entry->addr.proto != 0) {
6520 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6521 return -EINVAL;
6522 }
6523
6524 return 0;
6525 }
6526
6527 static const struct nla_policy mdba_get_policy[MDBA_GET_ENTRY_MAX + 1] = {
6528 [MDBA_GET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6529 rtnl_validate_mdb_entry_get,
6530 sizeof(struct br_mdb_entry)),
6531 [MDBA_GET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6532 };
6533
6534 static int rtnl_mdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6535 struct netlink_ext_ack *extack)
6536 {
6537 struct nlattr *tb[MDBA_GET_ENTRY_MAX + 1];
6538 struct net *net = sock_net(in_skb->sk);
6539 struct br_port_msg *bpm;
6540 struct net_device *dev;
6541 int err;
6542
6543 err = nlmsg_parse(nlh, sizeof(struct br_port_msg), tb,
6544 MDBA_GET_ENTRY_MAX, mdba_get_policy, extack);
6545 if (err)
6546 return err;
6547
6548 bpm = nlmsg_data(nlh);
6549 if (!bpm->ifindex) {
6550 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6551 return -EINVAL;
6552 }
6553
6554 dev = __dev_get_by_index(net, bpm->ifindex);
6555 if (!dev) {
6556 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6557 return -ENODEV;
6558 }
6559
6560 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_GET_ENTRY)) {
6561 NL_SET_ERR_MSG(extack, "Missing MDBA_GET_ENTRY attribute");
6562 return -EINVAL;
6563 }
6564
6565 if (!dev->netdev_ops->ndo_mdb_get) {
6566 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6567 return -EOPNOTSUPP;
6568 }
6569
6570 return dev->netdev_ops->ndo_mdb_get(dev, tb, NETLINK_CB(in_skb).portid,
6571 nlh->nlmsg_seq, extack);
6572 }
6573
6574 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6575 struct netlink_ext_ack *extack)
6576 {
6577 struct br_mdb_entry *entry = nla_data(attr);
6578
6579 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6580 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6581 return -EINVAL;
6582 }
6583
6584 if (entry->ifindex == 0) {
6585 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6586 return -EINVAL;
6587 }
6588
6589 if (entry->addr.proto == htons(ETH_P_IP)) {
6590 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6591 !ipv4_is_zeronet(entry->addr.u.ip4)) {
6592 NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6593 return -EINVAL;
6594 }
6595 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6596 NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6597 return -EINVAL;
6598 }
6599 #if IS_ENABLED(CONFIG_IPV6)
6600 } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6601 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6602 NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6603 return -EINVAL;
6604 }
6605 #endif
6606 } else if (entry->addr.proto == 0) {
6607 /* L2 mdb */
6608 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6609 NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6610 return -EINVAL;
6611 }
6612 } else {
6613 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6614 return -EINVAL;
6615 }
6616
6617 if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6618 NL_SET_ERR_MSG(extack, "Unknown entry state");
6619 return -EINVAL;
6620 }
6621 if (entry->vid >= VLAN_VID_MASK) {
6622 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6623 return -EINVAL;
6624 }
6625
6626 return 0;
6627 }
6628
6629 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6630 [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6631 [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6632 rtnl_validate_mdb_entry,
6633 sizeof(struct br_mdb_entry)),
6634 [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6635 };
6636
6637 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6638 struct netlink_ext_ack *extack)
6639 {
6640 struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6641 struct net *net = sock_net(skb->sk);
6642 struct br_port_msg *bpm;
6643 struct net_device *dev;
6644 int err;
6645
6646 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6647 MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6648 if (err)
6649 return err;
6650
6651 bpm = nlmsg_data(nlh);
6652 if (!bpm->ifindex) {
6653 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6654 return -EINVAL;
6655 }
6656
6657 dev = __dev_get_by_index(net, bpm->ifindex);
6658 if (!dev) {
6659 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6660 return -ENODEV;
6661 }
6662
6663 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6664 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6665 return -EINVAL;
6666 }
6667
6668 if (!dev->netdev_ops->ndo_mdb_add) {
6669 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6670 return -EOPNOTSUPP;
6671 }
6672
6673 return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6674 }
6675
6676 static int rtnl_validate_mdb_entry_del_bulk(const struct nlattr *attr,
6677 struct netlink_ext_ack *extack)
6678 {
6679 struct br_mdb_entry *entry = nla_data(attr);
6680 struct br_mdb_entry zero_entry = {};
6681
6682 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6683 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6684 return -EINVAL;
6685 }
6686
6687 if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6688 NL_SET_ERR_MSG(extack, "Unknown entry state");
6689 return -EINVAL;
6690 }
6691
6692 if (entry->flags) {
6693 NL_SET_ERR_MSG(extack, "Entry flags cannot be set");
6694 return -EINVAL;
6695 }
6696
6697 if (entry->vid >= VLAN_N_VID - 1) {
6698 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6699 return -EINVAL;
6700 }
6701
6702 if (memcmp(&entry->addr, &zero_entry.addr, sizeof(entry->addr))) {
6703 NL_SET_ERR_MSG(extack, "Entry address cannot be set");
6704 return -EINVAL;
6705 }
6706
6707 return 0;
6708 }
6709
6710 static const struct nla_policy mdba_del_bulk_policy[MDBA_SET_ENTRY_MAX + 1] = {
6711 [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6712 rtnl_validate_mdb_entry_del_bulk,
6713 sizeof(struct br_mdb_entry)),
6714 [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6715 };
6716
6717 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6718 struct netlink_ext_ack *extack)
6719 {
6720 bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
6721 struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6722 struct net *net = sock_net(skb->sk);
6723 struct br_port_msg *bpm;
6724 struct net_device *dev;
6725 int err;
6726
6727 if (!del_bulk)
6728 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6729 MDBA_SET_ENTRY_MAX, mdba_policy,
6730 extack);
6731 else
6732 err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY_MAX,
6733 mdba_del_bulk_policy, extack);
6734 if (err)
6735 return err;
6736
6737 bpm = nlmsg_data(nlh);
6738 if (!bpm->ifindex) {
6739 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6740 return -EINVAL;
6741 }
6742
6743 dev = __dev_get_by_index(net, bpm->ifindex);
6744 if (!dev) {
6745 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6746 return -ENODEV;
6747 }
6748
6749 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6750 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6751 return -EINVAL;
6752 }
6753
6754 if (del_bulk) {
6755 if (!dev->netdev_ops->ndo_mdb_del_bulk) {
6756 NL_SET_ERR_MSG(extack, "Device does not support MDB bulk deletion");
6757 return -EOPNOTSUPP;
6758 }
6759 return dev->netdev_ops->ndo_mdb_del_bulk(dev, tb, extack);
6760 }
6761
6762 if (!dev->netdev_ops->ndo_mdb_del) {
6763 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6764 return -EOPNOTSUPP;
6765 }
6766
6767 return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6768 }
6769
6770 /* Process one rtnetlink message. */
6771
6772 static int rtnl_dumpit(struct sk_buff *skb, struct netlink_callback *cb)
6773 {
6774 const bool needs_lock = !(cb->flags & RTNL_FLAG_DUMP_UNLOCKED);
6775 rtnl_dumpit_func dumpit = cb->data;
6776 int err;
6777
6778 /* Previous iteration have already finished, avoid calling->dumpit()
6779 * again, it may not expect to be called after it reached the end.
6780 */
6781 if (!dumpit)
6782 return 0;
6783
6784 if (needs_lock)
6785 rtnl_lock();
6786 err = dumpit(skb, cb);
6787 if (needs_lock)
6788 rtnl_unlock();
6789
6790 /* Old dump handlers used to send NLM_DONE as in a separate recvmsg().
6791 * Some applications which parse netlink manually depend on this.
6792 */
6793 if (cb->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE) {
6794 if (err < 0 && err != -EMSGSIZE)
6795 return err;
6796 if (!err)
6797 cb->data = NULL;
6798
6799 return skb->len;
6800 }
6801 return err;
6802 }
6803
6804 static int rtnetlink_dump_start(struct sock *ssk, struct sk_buff *skb,
6805 const struct nlmsghdr *nlh,
6806 struct netlink_dump_control *control)
6807 {
6808 if (control->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE ||
6809 !(control->flags & RTNL_FLAG_DUMP_UNLOCKED)) {
6810 WARN_ON(control->data);
6811 control->data = control->dump;
6812 control->dump = rtnl_dumpit;
6813 }
6814
6815 return netlink_dump_start(ssk, skb, nlh, control);
6816 }
6817
6818 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6819 struct netlink_ext_ack *extack)
6820 {
6821 struct net *net = sock_net(skb->sk);
6822 struct rtnl_link *link;
6823 enum rtnl_kinds kind;
6824 struct module *owner;
6825 int err = -EOPNOTSUPP;
6826 rtnl_doit_func doit;
6827 unsigned int flags;
6828 int family;
6829 int type;
6830
6831 type = nlh->nlmsg_type;
6832 if (type > RTM_MAX)
6833 return -EOPNOTSUPP;
6834
6835 type -= RTM_BASE;
6836
6837 /* All the messages must have at least 1 byte length */
6838 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6839 return 0;
6840
6841 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6842 kind = rtnl_msgtype_kind(type);
6843
6844 if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6845 return -EPERM;
6846
6847 rcu_read_lock();
6848 if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6849 struct sock *rtnl;
6850 rtnl_dumpit_func dumpit;
6851 u32 min_dump_alloc = 0;
6852
6853 link = rtnl_get_link(family, type);
6854 if (!link || !link->dumpit) {
6855 family = PF_UNSPEC;
6856 link = rtnl_get_link(family, type);
6857 if (!link || !link->dumpit)
6858 goto err_unlock;
6859 }
6860 owner = link->owner;
6861 dumpit = link->dumpit;
6862 flags = link->flags;
6863
6864 if (type == RTM_GETLINK - RTM_BASE)
6865 min_dump_alloc = rtnl_calcit(skb, nlh);
6866
6867 err = 0;
6868 /* need to do this before rcu_read_unlock() */
6869 if (!try_module_get(owner))
6870 err = -EPROTONOSUPPORT;
6871
6872 rcu_read_unlock();
6873
6874 rtnl = net->rtnl;
6875 if (err == 0) {
6876 struct netlink_dump_control c = {
6877 .dump = dumpit,
6878 .min_dump_alloc = min_dump_alloc,
6879 .module = owner,
6880 .flags = flags,
6881 };
6882 err = rtnetlink_dump_start(rtnl, skb, nlh, &c);
6883 /* netlink_dump_start() will keep a reference on
6884 * module if dump is still in progress.
6885 */
6886 module_put(owner);
6887 }
6888 return err;
6889 }
6890
6891 link = rtnl_get_link(family, type);
6892 if (!link || !link->doit) {
6893 family = PF_UNSPEC;
6894 link = rtnl_get_link(PF_UNSPEC, type);
6895 if (!link || !link->doit)
6896 goto out_unlock;
6897 }
6898
6899 owner = link->owner;
6900 if (!try_module_get(owner)) {
6901 err = -EPROTONOSUPPORT;
6902 goto out_unlock;
6903 }
6904
6905 flags = link->flags;
6906 if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6907 !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6908 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6909 module_put(owner);
6910 goto err_unlock;
6911 }
6912
6913 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6914 doit = link->doit;
6915 rcu_read_unlock();
6916 if (doit)
6917 err = doit(skb, nlh, extack);
6918 module_put(owner);
6919 return err;
6920 }
6921 rcu_read_unlock();
6922
6923 rtnl_lock();
6924 link = rtnl_get_link(family, type);
6925 if (link && link->doit)
6926 err = link->doit(skb, nlh, extack);
6927 rtnl_unlock();
6928
6929 module_put(owner);
6930
6931 return err;
6932
6933 out_unlock:
6934 rcu_read_unlock();
6935 return err;
6936
6937 err_unlock:
6938 rcu_read_unlock();
6939 return -EOPNOTSUPP;
6940 }
6941
6942 static void rtnetlink_rcv(struct sk_buff *skb)
6943 {
6944 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6945 }
6946
6947 static int rtnetlink_bind(struct net *net, int group)
6948 {
6949 switch (group) {
6950 case RTNLGRP_IPV4_MROUTE_R:
6951 case RTNLGRP_IPV6_MROUTE_R:
6952 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6953 return -EPERM;
6954 break;
6955 }
6956 return 0;
6957 }
6958
6959 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6960 {
6961 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6962
6963 switch (event) {
6964 case NETDEV_REBOOT:
6965 case NETDEV_CHANGEMTU:
6966 case NETDEV_CHANGEADDR:
6967 case NETDEV_CHANGENAME:
6968 case NETDEV_FEAT_CHANGE:
6969 case NETDEV_BONDING_FAILOVER:
6970 case NETDEV_POST_TYPE_CHANGE:
6971 case NETDEV_NOTIFY_PEERS:
6972 case NETDEV_CHANGEUPPER:
6973 case NETDEV_RESEND_IGMP:
6974 case NETDEV_CHANGEINFODATA:
6975 case NETDEV_CHANGELOWERSTATE:
6976 case NETDEV_CHANGE_TX_QUEUE_LEN:
6977 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6978 GFP_KERNEL, NULL, 0, 0, NULL);
6979 break;
6980 default:
6981 break;
6982 }
6983 return NOTIFY_DONE;
6984 }
6985
6986 static struct notifier_block rtnetlink_dev_notifier = {
6987 .notifier_call = rtnetlink_event,
6988 };
6989
6990
6991 static int __net_init rtnetlink_net_init(struct net *net)
6992 {
6993 struct sock *sk;
6994 struct netlink_kernel_cfg cfg = {
6995 .groups = RTNLGRP_MAX,
6996 .input = rtnetlink_rcv,
6997 .flags = NL_CFG_F_NONROOT_RECV,
6998 .bind = rtnetlink_bind,
6999 };
7000
7001 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
7002 if (!sk)
7003 return -ENOMEM;
7004 net->rtnl = sk;
7005 return 0;
7006 }
7007
7008 static void __net_exit rtnetlink_net_exit(struct net *net)
7009 {
7010 netlink_kernel_release(net->rtnl);
7011 net->rtnl = NULL;
7012 }
7013
7014 static struct pernet_operations rtnetlink_net_ops = {
7015 .init = rtnetlink_net_init,
7016 .exit = rtnetlink_net_exit,
7017 };
7018
7019 static const struct rtnl_msg_handler rtnetlink_rtnl_msg_handlers[] __initconst = {
7020 {.msgtype = RTM_NEWLINK, .doit = rtnl_newlink,
7021 .flags = RTNL_FLAG_DOIT_PERNET},
7022 {.msgtype = RTM_DELLINK, .doit = rtnl_dellink,
7023 .flags = RTNL_FLAG_DOIT_PERNET_WIP},
7024 {.msgtype = RTM_GETLINK, .doit = rtnl_getlink,
7025 .dumpit = rtnl_dump_ifinfo, .flags = RTNL_FLAG_DUMP_SPLIT_NLM_DONE},
7026 {.msgtype = RTM_SETLINK, .doit = rtnl_setlink,
7027 .flags = RTNL_FLAG_DOIT_PERNET_WIP},
7028 {.msgtype = RTM_GETADDR, .dumpit = rtnl_dump_all},
7029 {.msgtype = RTM_GETROUTE, .dumpit = rtnl_dump_all},
7030 {.msgtype = RTM_GETNETCONF, .dumpit = rtnl_dump_all},
7031 {.msgtype = RTM_GETSTATS, .doit = rtnl_stats_get,
7032 .dumpit = rtnl_stats_dump},
7033 {.msgtype = RTM_SETSTATS, .doit = rtnl_stats_set},
7034 {.msgtype = RTM_NEWLINKPROP, .doit = rtnl_newlinkprop},
7035 {.msgtype = RTM_DELLINKPROP, .doit = rtnl_dellinkprop},
7036 {.protocol = PF_BRIDGE, .msgtype = RTM_GETLINK,
7037 .dumpit = rtnl_bridge_getlink},
7038 {.protocol = PF_BRIDGE, .msgtype = RTM_DELLINK,
7039 .doit = rtnl_bridge_dellink},
7040 {.protocol = PF_BRIDGE, .msgtype = RTM_SETLINK,
7041 .doit = rtnl_bridge_setlink},
7042 {.protocol = PF_BRIDGE, .msgtype = RTM_NEWNEIGH, .doit = rtnl_fdb_add},
7043 {.protocol = PF_BRIDGE, .msgtype = RTM_DELNEIGH, .doit = rtnl_fdb_del,
7044 .flags = RTNL_FLAG_BULK_DEL_SUPPORTED},
7045 {.protocol = PF_BRIDGE, .msgtype = RTM_GETNEIGH, .doit = rtnl_fdb_get,
7046 .dumpit = rtnl_fdb_dump},
7047 {.protocol = PF_BRIDGE, .msgtype = RTM_NEWMDB, .doit = rtnl_mdb_add},
7048 {.protocol = PF_BRIDGE, .msgtype = RTM_DELMDB, .doit = rtnl_mdb_del,
7049 .flags = RTNL_FLAG_BULK_DEL_SUPPORTED},
7050 {.protocol = PF_BRIDGE, .msgtype = RTM_GETMDB, .doit = rtnl_mdb_get,
7051 .dumpit = rtnl_mdb_dump},
7052 };
7053
7054 void __init rtnetlink_init(void)
7055 {
7056 if (register_pernet_subsys(&rtnetlink_net_ops))
7057 panic("rtnetlink_init: cannot initialize rtnetlink\n");
7058
7059 register_netdevice_notifier(&rtnetlink_dev_notifier);
7060
7061 rtnl_register_many(rtnetlink_rtnl_msg_handlers);
7062 }
Kernel DRM miscellaneous fixes and cross-tree changes