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