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