drm/panel-edp: Add BOE NV140FHM-NZ panel entry
[drm/drm-misc.git] / net / core / rtnetlink.c
blobab5f201bf0ab41b463175f501e8560b4d64d9b0a
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.
7 * Routing netlink socket interface: protocol independent part.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
11 * Fixes:
12 * Vitaly E. Lavrov RTA_OK arithmetic was wrong.
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>
41 #include <linux/uaccess.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <net/ip.h>
46 #include <net/protocol.h>
47 #include <net/arp.h>
48 #include <net/route.h>
49 #include <net/udp.h>
50 #include <net/tcp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 #include <net/devlink.h>
57 #if IS_ENABLED(CONFIG_IPV6)
58 #include <net/addrconf.h>
59 #endif
60 #include <linux/dpll.h>
62 #include "dev.h"
64 #define RTNL_MAX_TYPE 50
65 #define RTNL_SLAVE_MAX_TYPE 44
67 struct rtnl_link {
68 rtnl_doit_func doit;
69 rtnl_dumpit_func dumpit;
70 struct module *owner;
71 unsigned int flags;
72 struct rcu_head rcu;
75 static DEFINE_MUTEX(rtnl_mutex);
77 void rtnl_lock(void)
79 mutex_lock(&rtnl_mutex);
81 EXPORT_SYMBOL(rtnl_lock);
83 int rtnl_lock_killable(void)
85 return mutex_lock_killable(&rtnl_mutex);
87 EXPORT_SYMBOL(rtnl_lock_killable);
89 static struct sk_buff *defer_kfree_skb_list;
90 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
92 if (head && tail) {
93 tail->next = defer_kfree_skb_list;
94 defer_kfree_skb_list = head;
97 EXPORT_SYMBOL(rtnl_kfree_skbs);
99 void __rtnl_unlock(void)
101 struct sk_buff *head = defer_kfree_skb_list;
103 defer_kfree_skb_list = NULL;
105 /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
106 * is used. In some places, e.g. in cfg80211, we have code that will do
107 * something like
108 * rtnl_lock()
109 * wiphy_lock()
110 * ...
111 * rtnl_unlock()
113 * and because netdev_run_todo() acquires the RTNL for items on the list
114 * we could cause a situation such as this:
115 * Thread 1 Thread 2
116 * rtnl_lock()
117 * unregister_netdevice()
118 * __rtnl_unlock()
119 * rtnl_lock()
120 * wiphy_lock()
121 * rtnl_unlock()
122 * netdev_run_todo()
123 * __rtnl_unlock()
125 * // list not empty now
126 * // because of thread 2
127 * rtnl_lock()
128 * while (!list_empty(...))
129 * rtnl_lock()
130 * wiphy_lock()
131 * **** DEADLOCK ****
133 * However, usage of __rtnl_unlock() is rare, and so we can ensure that
134 * it's not used in cases where something is added to do the list.
136 WARN_ON(!list_empty(&net_todo_list));
138 mutex_unlock(&rtnl_mutex);
140 while (head) {
141 struct sk_buff *next = head->next;
143 kfree_skb(head);
144 cond_resched();
145 head = next;
149 void rtnl_unlock(void)
151 /* This fellow will unlock it for us. */
152 netdev_run_todo();
154 EXPORT_SYMBOL(rtnl_unlock);
156 int rtnl_trylock(void)
158 return mutex_trylock(&rtnl_mutex);
160 EXPORT_SYMBOL(rtnl_trylock);
162 int rtnl_is_locked(void)
164 return mutex_is_locked(&rtnl_mutex);
166 EXPORT_SYMBOL(rtnl_is_locked);
168 bool refcount_dec_and_rtnl_lock(refcount_t *r)
170 return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
172 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
174 #ifdef CONFIG_PROVE_LOCKING
175 bool lockdep_rtnl_is_held(void)
177 return lockdep_is_held(&rtnl_mutex);
179 EXPORT_SYMBOL(lockdep_rtnl_is_held);
180 #endif /* #ifdef CONFIG_PROVE_LOCKING */
182 #ifdef CONFIG_DEBUG_NET_SMALL_RTNL
183 void __rtnl_net_lock(struct net *net)
185 ASSERT_RTNL();
187 mutex_lock(&net->rtnl_mutex);
189 EXPORT_SYMBOL(__rtnl_net_lock);
191 void __rtnl_net_unlock(struct net *net)
193 ASSERT_RTNL();
195 mutex_unlock(&net->rtnl_mutex);
197 EXPORT_SYMBOL(__rtnl_net_unlock);
199 void rtnl_net_lock(struct net *net)
201 rtnl_lock();
202 __rtnl_net_lock(net);
204 EXPORT_SYMBOL(rtnl_net_lock);
206 void rtnl_net_unlock(struct net *net)
208 __rtnl_net_unlock(net);
209 rtnl_unlock();
211 EXPORT_SYMBOL(rtnl_net_unlock);
213 int rtnl_net_trylock(struct net *net)
215 int ret = rtnl_trylock();
217 if (ret)
218 __rtnl_net_lock(net);
220 return ret;
222 EXPORT_SYMBOL(rtnl_net_trylock);
224 static int rtnl_net_cmp_locks(const struct net *net_a, const struct net *net_b)
226 if (net_eq(net_a, net_b))
227 return 0;
229 /* always init_net first */
230 if (net_eq(net_a, &init_net))
231 return -1;
233 if (net_eq(net_b, &init_net))
234 return 1;
236 /* otherwise lock in ascending order */
237 return net_a < net_b ? -1 : 1;
240 int rtnl_net_lock_cmp_fn(const struct lockdep_map *a, const struct lockdep_map *b)
242 const struct net *net_a, *net_b;
244 net_a = container_of(a, struct net, rtnl_mutex.dep_map);
245 net_b = container_of(b, struct net, rtnl_mutex.dep_map);
247 return rtnl_net_cmp_locks(net_a, net_b);
250 bool rtnl_net_is_locked(struct net *net)
252 return rtnl_is_locked() && mutex_is_locked(&net->rtnl_mutex);
254 EXPORT_SYMBOL(rtnl_net_is_locked);
256 bool lockdep_rtnl_net_is_held(struct net *net)
258 return lockdep_rtnl_is_held() && lockdep_is_held(&net->rtnl_mutex);
260 EXPORT_SYMBOL(lockdep_rtnl_net_is_held);
261 #else
262 static int rtnl_net_cmp_locks(const struct net *net_a, const struct net *net_b)
264 /* No need to swap */
265 return -1;
267 #endif
269 struct rtnl_nets {
270 /* ->newlink() needs to freeze 3 netns at most;
271 * 2 for the new device, 1 for its peer.
273 struct net *net[3];
274 unsigned char len;
277 static void rtnl_nets_init(struct rtnl_nets *rtnl_nets)
279 memset(rtnl_nets, 0, sizeof(*rtnl_nets));
282 static void rtnl_nets_destroy(struct rtnl_nets *rtnl_nets)
284 int i;
286 for (i = 0; i < rtnl_nets->len; i++) {
287 put_net(rtnl_nets->net[i]);
288 rtnl_nets->net[i] = NULL;
291 rtnl_nets->len = 0;
295 * rtnl_nets_add - Add netns to be locked before ->newlink().
297 * @rtnl_nets: rtnl_nets pointer passed to ->get_peer_net().
298 * @net: netns pointer with an extra refcnt held.
300 * The extra refcnt is released in rtnl_nets_destroy().
302 static void rtnl_nets_add(struct rtnl_nets *rtnl_nets, struct net *net)
304 int i;
306 DEBUG_NET_WARN_ON_ONCE(rtnl_nets->len == ARRAY_SIZE(rtnl_nets->net));
308 for (i = 0; i < rtnl_nets->len; i++) {
309 switch (rtnl_net_cmp_locks(rtnl_nets->net[i], net)) {
310 case 0:
311 put_net(net);
312 return;
313 case 1:
314 swap(rtnl_nets->net[i], net);
318 rtnl_nets->net[i] = net;
319 rtnl_nets->len++;
322 static void rtnl_nets_lock(struct rtnl_nets *rtnl_nets)
324 int i;
326 rtnl_lock();
328 for (i = 0; i < rtnl_nets->len; i++)
329 __rtnl_net_lock(rtnl_nets->net[i]);
332 static void rtnl_nets_unlock(struct rtnl_nets *rtnl_nets)
334 int i;
336 for (i = 0; i < rtnl_nets->len; i++)
337 __rtnl_net_unlock(rtnl_nets->net[i]);
339 rtnl_unlock();
342 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
344 static inline int rtm_msgindex(int msgtype)
346 int msgindex = msgtype - RTM_BASE;
349 * msgindex < 0 implies someone tried to register a netlink
350 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
351 * the message type has not been added to linux/rtnetlink.h
353 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
355 return msgindex;
358 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
360 struct rtnl_link __rcu **tab;
362 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
363 protocol = PF_UNSPEC;
365 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
366 if (!tab)
367 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
369 return rcu_dereference_rtnl(tab[msgtype]);
372 static int rtnl_register_internal(struct module *owner,
373 int protocol, int msgtype,
374 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
375 unsigned int flags)
377 struct rtnl_link *link, *old;
378 struct rtnl_link __rcu **tab;
379 int msgindex;
380 int ret = -ENOBUFS;
382 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
383 msgindex = rtm_msgindex(msgtype);
385 rtnl_lock();
386 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
387 if (tab == NULL) {
388 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
389 if (!tab)
390 goto unlock;
392 /* ensures we see the 0 stores */
393 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
396 old = rtnl_dereference(tab[msgindex]);
397 if (old) {
398 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
399 if (!link)
400 goto unlock;
401 } else {
402 link = kzalloc(sizeof(*link), GFP_KERNEL);
403 if (!link)
404 goto unlock;
407 WARN_ON(link->owner && link->owner != owner);
408 link->owner = owner;
410 WARN_ON(doit && link->doit && link->doit != doit);
411 if (doit)
412 link->doit = doit;
413 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
414 if (dumpit)
415 link->dumpit = dumpit;
417 WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
418 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
419 link->flags |= flags;
421 /* publish protocol:msgtype */
422 rcu_assign_pointer(tab[msgindex], link);
423 ret = 0;
424 if (old)
425 kfree_rcu(old, rcu);
426 unlock:
427 rtnl_unlock();
428 return ret;
432 * rtnl_unregister - Unregister a rtnetlink message type
433 * @protocol: Protocol family or PF_UNSPEC
434 * @msgtype: rtnetlink message type
436 * Returns 0 on success or a negative error code.
438 static int rtnl_unregister(int protocol, int msgtype)
440 struct rtnl_link __rcu **tab;
441 struct rtnl_link *link;
442 int msgindex;
444 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
445 msgindex = rtm_msgindex(msgtype);
447 rtnl_lock();
448 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
449 if (!tab) {
450 rtnl_unlock();
451 return -ENOENT;
454 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
455 rtnl_unlock();
457 kfree_rcu(link, rcu);
459 return 0;
463 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
464 * @protocol : Protocol family or PF_UNSPEC
466 * Identical to calling rtnl_unregster() for all registered message types
467 * of a certain protocol family.
469 void rtnl_unregister_all(int protocol)
471 struct rtnl_link __rcu **tab;
472 struct rtnl_link *link;
473 int msgindex;
475 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
477 rtnl_lock();
478 tab = rcu_replace_pointer_rtnl(rtnl_msg_handlers[protocol], NULL);
479 if (!tab) {
480 rtnl_unlock();
481 return;
483 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
484 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
485 kfree_rcu(link, rcu);
487 rtnl_unlock();
489 synchronize_net();
491 kfree(tab);
493 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
496 * __rtnl_register_many - Register rtnetlink message types
497 * @handlers: Array of struct rtnl_msg_handlers
498 * @n: The length of @handlers
500 * Registers the specified function pointers (at least one of them has
501 * to be non-NULL) to be called whenever a request message for the
502 * specified protocol family and message type is received.
504 * The special protocol family PF_UNSPEC may be used to define fallback
505 * function pointers for the case when no entry for the specific protocol
506 * family exists.
508 * When one element of @handlers fails to register,
509 * 1) built-in: panics.
510 * 2) modules : the previous successful registrations are unwinded
511 * and an error is returned.
513 * Use rtnl_register_many().
515 int __rtnl_register_many(const struct rtnl_msg_handler *handlers, int n)
517 const struct rtnl_msg_handler *handler;
518 int i, err;
520 for (i = 0, handler = handlers; i < n; i++, handler++) {
521 err = rtnl_register_internal(handler->owner, handler->protocol,
522 handler->msgtype, handler->doit,
523 handler->dumpit, handler->flags);
524 if (err) {
525 if (!handler->owner)
526 panic("Unable to register rtnetlink message "
527 "handlers, %pS\n", handlers);
529 __rtnl_unregister_many(handlers, i);
530 break;
534 return err;
536 EXPORT_SYMBOL_GPL(__rtnl_register_many);
538 void __rtnl_unregister_many(const struct rtnl_msg_handler *handlers, int n)
540 const struct rtnl_msg_handler *handler;
541 int i;
543 for (i = n - 1, handler = handlers + n - 1; i >= 0; i--, handler--)
544 rtnl_unregister(handler->protocol, handler->msgtype);
546 EXPORT_SYMBOL_GPL(__rtnl_unregister_many);
548 static DEFINE_MUTEX(link_ops_mutex);
549 static LIST_HEAD(link_ops);
551 static struct rtnl_link_ops *rtnl_link_ops_get(const char *kind, int *srcu_index)
553 struct rtnl_link_ops *ops;
555 rcu_read_lock();
557 list_for_each_entry_rcu(ops, &link_ops, list) {
558 if (!strcmp(ops->kind, kind)) {
559 *srcu_index = srcu_read_lock(&ops->srcu);
560 goto unlock;
564 ops = NULL;
565 unlock:
566 rcu_read_unlock();
568 return ops;
571 static void rtnl_link_ops_put(struct rtnl_link_ops *ops, int srcu_index)
573 srcu_read_unlock(&ops->srcu, srcu_index);
577 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
578 * @ops: struct rtnl_link_ops * to register
580 * Returns 0 on success or a negative error code.
582 int rtnl_link_register(struct rtnl_link_ops *ops)
584 struct rtnl_link_ops *tmp;
585 int err;
587 /* Sanity-check max sizes to avoid stack buffer overflow. */
588 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
589 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
590 return -EINVAL;
592 /* The check for alloc/setup is here because if ops
593 * does not have that filled up, it is not possible
594 * to use the ops for creating device. So do not
595 * fill up dellink as well. That disables rtnl_dellink.
597 if ((ops->alloc || ops->setup) && !ops->dellink)
598 ops->dellink = unregister_netdevice_queue;
600 err = init_srcu_struct(&ops->srcu);
601 if (err)
602 return err;
604 mutex_lock(&link_ops_mutex);
606 list_for_each_entry(tmp, &link_ops, list) {
607 if (!strcmp(ops->kind, tmp->kind)) {
608 err = -EEXIST;
609 goto unlock;
613 list_add_tail_rcu(&ops->list, &link_ops);
614 unlock:
615 mutex_unlock(&link_ops_mutex);
617 return err;
619 EXPORT_SYMBOL_GPL(rtnl_link_register);
621 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
623 struct net_device *dev;
624 LIST_HEAD(list_kill);
626 for_each_netdev(net, dev) {
627 if (dev->rtnl_link_ops == ops)
628 ops->dellink(dev, &list_kill);
630 unregister_netdevice_many(&list_kill);
633 /* Return with the rtnl_lock held when there are no network
634 * devices unregistering in any network namespace.
636 static void rtnl_lock_unregistering_all(void)
638 DEFINE_WAIT_FUNC(wait, woken_wake_function);
640 add_wait_queue(&netdev_unregistering_wq, &wait);
641 for (;;) {
642 rtnl_lock();
643 /* We held write locked pernet_ops_rwsem, and parallel
644 * setup_net() and cleanup_net() are not possible.
646 if (!atomic_read(&dev_unreg_count))
647 break;
648 __rtnl_unlock();
650 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
652 remove_wait_queue(&netdev_unregistering_wq, &wait);
656 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
657 * @ops: struct rtnl_link_ops * to unregister
659 void rtnl_link_unregister(struct rtnl_link_ops *ops)
661 struct net *net;
663 mutex_lock(&link_ops_mutex);
664 list_del_rcu(&ops->list);
665 mutex_unlock(&link_ops_mutex);
667 synchronize_srcu(&ops->srcu);
668 cleanup_srcu_struct(&ops->srcu);
670 /* Close the race with setup_net() and cleanup_net() */
671 down_write(&pernet_ops_rwsem);
672 rtnl_lock_unregistering_all();
674 for_each_net(net)
675 __rtnl_kill_links(net, ops);
677 rtnl_unlock();
678 up_write(&pernet_ops_rwsem);
680 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
682 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
684 struct net_device *master_dev;
685 const struct rtnl_link_ops *ops;
686 size_t size = 0;
688 rcu_read_lock();
690 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
691 if (!master_dev)
692 goto out;
694 ops = master_dev->rtnl_link_ops;
695 if (!ops || !ops->get_slave_size)
696 goto out;
697 /* IFLA_INFO_SLAVE_DATA + nested data */
698 size = nla_total_size(sizeof(struct nlattr)) +
699 ops->get_slave_size(master_dev, dev);
701 out:
702 rcu_read_unlock();
703 return size;
706 static size_t rtnl_link_get_size(const struct net_device *dev)
708 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
709 size_t size;
711 if (!ops)
712 return 0;
714 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
715 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
717 if (ops->get_size)
718 /* IFLA_INFO_DATA + nested data */
719 size += nla_total_size(sizeof(struct nlattr)) +
720 ops->get_size(dev);
722 if (ops->get_xstats_size)
723 /* IFLA_INFO_XSTATS */
724 size += nla_total_size(ops->get_xstats_size(dev));
726 size += rtnl_link_get_slave_info_data_size(dev);
728 return size;
731 static LIST_HEAD(rtnl_af_ops);
733 static struct rtnl_af_ops *rtnl_af_lookup(const int family, int *srcu_index)
735 struct rtnl_af_ops *ops;
737 ASSERT_RTNL();
739 rcu_read_lock();
741 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
742 if (ops->family == family) {
743 *srcu_index = srcu_read_lock(&ops->srcu);
744 goto unlock;
748 ops = NULL;
749 unlock:
750 rcu_read_unlock();
752 return ops;
755 static void rtnl_af_put(struct rtnl_af_ops *ops, int srcu_index)
757 srcu_read_unlock(&ops->srcu, srcu_index);
761 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
762 * @ops: struct rtnl_af_ops * to register
764 * Return: 0 on success or a negative error code.
766 int rtnl_af_register(struct rtnl_af_ops *ops)
768 int err = init_srcu_struct(&ops->srcu);
770 if (err)
771 return err;
773 rtnl_lock();
774 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
775 rtnl_unlock();
777 return 0;
779 EXPORT_SYMBOL_GPL(rtnl_af_register);
782 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
783 * @ops: struct rtnl_af_ops * to unregister
785 void rtnl_af_unregister(struct rtnl_af_ops *ops)
787 rtnl_lock();
788 list_del_rcu(&ops->list);
789 rtnl_unlock();
791 synchronize_rcu();
792 synchronize_srcu(&ops->srcu);
793 cleanup_srcu_struct(&ops->srcu);
795 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
797 static size_t rtnl_link_get_af_size(const struct net_device *dev,
798 u32 ext_filter_mask)
800 struct rtnl_af_ops *af_ops;
801 size_t size;
803 /* IFLA_AF_SPEC */
804 size = nla_total_size(sizeof(struct nlattr));
806 rcu_read_lock();
807 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
808 if (af_ops->get_link_af_size) {
809 /* AF_* + nested data */
810 size += nla_total_size(sizeof(struct nlattr)) +
811 af_ops->get_link_af_size(dev, ext_filter_mask);
814 rcu_read_unlock();
816 return size;
819 static bool rtnl_have_link_slave_info(const struct net_device *dev)
821 struct net_device *master_dev;
822 bool ret = false;
824 rcu_read_lock();
826 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
827 if (master_dev && master_dev->rtnl_link_ops)
828 ret = true;
829 rcu_read_unlock();
830 return ret;
833 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
834 const struct net_device *dev)
836 struct net_device *master_dev;
837 const struct rtnl_link_ops *ops;
838 struct nlattr *slave_data;
839 int err;
841 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
842 if (!master_dev)
843 return 0;
844 ops = master_dev->rtnl_link_ops;
845 if (!ops)
846 return 0;
847 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
848 return -EMSGSIZE;
849 if (ops->fill_slave_info) {
850 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
851 if (!slave_data)
852 return -EMSGSIZE;
853 err = ops->fill_slave_info(skb, master_dev, dev);
854 if (err < 0)
855 goto err_cancel_slave_data;
856 nla_nest_end(skb, slave_data);
858 return 0;
860 err_cancel_slave_data:
861 nla_nest_cancel(skb, slave_data);
862 return err;
865 static int rtnl_link_info_fill(struct sk_buff *skb,
866 const struct net_device *dev)
868 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
869 struct nlattr *data;
870 int err;
872 if (!ops)
873 return 0;
874 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
875 return -EMSGSIZE;
876 if (ops->fill_xstats) {
877 err = ops->fill_xstats(skb, dev);
878 if (err < 0)
879 return err;
881 if (ops->fill_info) {
882 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
883 if (data == NULL)
884 return -EMSGSIZE;
885 err = ops->fill_info(skb, dev);
886 if (err < 0)
887 goto err_cancel_data;
888 nla_nest_end(skb, data);
890 return 0;
892 err_cancel_data:
893 nla_nest_cancel(skb, data);
894 return err;
897 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
899 struct nlattr *linkinfo;
900 int err = -EMSGSIZE;
902 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
903 if (linkinfo == NULL)
904 goto out;
906 err = rtnl_link_info_fill(skb, dev);
907 if (err < 0)
908 goto err_cancel_link;
910 err = rtnl_link_slave_info_fill(skb, dev);
911 if (err < 0)
912 goto err_cancel_link;
914 nla_nest_end(skb, linkinfo);
915 return 0;
917 err_cancel_link:
918 nla_nest_cancel(skb, linkinfo);
919 out:
920 return err;
923 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
925 struct sock *rtnl = net->rtnl;
927 return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
930 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
932 struct sock *rtnl = net->rtnl;
934 return nlmsg_unicast(rtnl, skb, pid);
936 EXPORT_SYMBOL(rtnl_unicast);
938 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
939 const struct nlmsghdr *nlh, gfp_t flags)
941 struct sock *rtnl = net->rtnl;
943 nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
945 EXPORT_SYMBOL(rtnl_notify);
947 void rtnl_set_sk_err(struct net *net, u32 group, int error)
949 struct sock *rtnl = net->rtnl;
951 netlink_set_err(rtnl, 0, group, error);
953 EXPORT_SYMBOL(rtnl_set_sk_err);
955 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
957 struct nlattr *mx;
958 int i, valid = 0;
960 /* nothing is dumped for dst_default_metrics, so just skip the loop */
961 if (metrics == dst_default_metrics.metrics)
962 return 0;
964 mx = nla_nest_start_noflag(skb, RTA_METRICS);
965 if (mx == NULL)
966 return -ENOBUFS;
968 for (i = 0; i < RTAX_MAX; i++) {
969 if (metrics[i]) {
970 if (i == RTAX_CC_ALGO - 1) {
971 char tmp[TCP_CA_NAME_MAX], *name;
973 name = tcp_ca_get_name_by_key(metrics[i], tmp);
974 if (!name)
975 continue;
976 if (nla_put_string(skb, i + 1, name))
977 goto nla_put_failure;
978 } else if (i == RTAX_FEATURES - 1) {
979 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
981 if (!user_features)
982 continue;
983 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
984 if (nla_put_u32(skb, i + 1, user_features))
985 goto nla_put_failure;
986 } else {
987 if (nla_put_u32(skb, i + 1, metrics[i]))
988 goto nla_put_failure;
990 valid++;
994 if (!valid) {
995 nla_nest_cancel(skb, mx);
996 return 0;
999 return nla_nest_end(skb, mx);
1001 nla_put_failure:
1002 nla_nest_cancel(skb, mx);
1003 return -EMSGSIZE;
1005 EXPORT_SYMBOL(rtnetlink_put_metrics);
1007 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
1008 long expires, u32 error)
1010 struct rta_cacheinfo ci = {
1011 .rta_error = error,
1012 .rta_id = id,
1015 if (dst) {
1016 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
1017 ci.rta_used = dst->__use;
1018 ci.rta_clntref = rcuref_read(&dst->__rcuref);
1020 if (expires) {
1021 unsigned long clock;
1023 clock = jiffies_to_clock_t(abs(expires));
1024 clock = min_t(unsigned long, clock, INT_MAX);
1025 ci.rta_expires = (expires > 0) ? clock : -clock;
1027 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
1029 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
1031 void netdev_set_operstate(struct net_device *dev, int newstate)
1033 unsigned int old = READ_ONCE(dev->operstate);
1035 do {
1036 if (old == newstate)
1037 return;
1038 } while (!try_cmpxchg(&dev->operstate, &old, newstate));
1040 netdev_state_change(dev);
1042 EXPORT_SYMBOL(netdev_set_operstate);
1044 static void set_operstate(struct net_device *dev, unsigned char transition)
1046 unsigned char operstate = READ_ONCE(dev->operstate);
1048 switch (transition) {
1049 case IF_OPER_UP:
1050 if ((operstate == IF_OPER_DORMANT ||
1051 operstate == IF_OPER_TESTING ||
1052 operstate == IF_OPER_UNKNOWN) &&
1053 !netif_dormant(dev) && !netif_testing(dev))
1054 operstate = IF_OPER_UP;
1055 break;
1057 case IF_OPER_TESTING:
1058 if (netif_oper_up(dev))
1059 operstate = IF_OPER_TESTING;
1060 break;
1062 case IF_OPER_DORMANT:
1063 if (netif_oper_up(dev))
1064 operstate = IF_OPER_DORMANT;
1065 break;
1068 netdev_set_operstate(dev, operstate);
1071 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
1073 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
1074 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
1077 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
1078 const struct ifinfomsg *ifm)
1080 unsigned int flags = ifm->ifi_flags;
1082 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
1083 if (ifm->ifi_change)
1084 flags = (flags & ifm->ifi_change) |
1085 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
1087 return flags;
1090 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
1091 const struct rtnl_link_stats64 *b)
1093 a->rx_packets = b->rx_packets;
1094 a->tx_packets = b->tx_packets;
1095 a->rx_bytes = b->rx_bytes;
1096 a->tx_bytes = b->tx_bytes;
1097 a->rx_errors = b->rx_errors;
1098 a->tx_errors = b->tx_errors;
1099 a->rx_dropped = b->rx_dropped;
1100 a->tx_dropped = b->tx_dropped;
1102 a->multicast = b->multicast;
1103 a->collisions = b->collisions;
1105 a->rx_length_errors = b->rx_length_errors;
1106 a->rx_over_errors = b->rx_over_errors;
1107 a->rx_crc_errors = b->rx_crc_errors;
1108 a->rx_frame_errors = b->rx_frame_errors;
1109 a->rx_fifo_errors = b->rx_fifo_errors;
1110 a->rx_missed_errors = b->rx_missed_errors;
1112 a->tx_aborted_errors = b->tx_aborted_errors;
1113 a->tx_carrier_errors = b->tx_carrier_errors;
1114 a->tx_fifo_errors = b->tx_fifo_errors;
1115 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
1116 a->tx_window_errors = b->tx_window_errors;
1118 a->rx_compressed = b->rx_compressed;
1119 a->tx_compressed = b->tx_compressed;
1121 a->rx_nohandler = b->rx_nohandler;
1124 /* All VF info */
1125 static inline int rtnl_vfinfo_size(const struct net_device *dev,
1126 u32 ext_filter_mask)
1128 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
1129 int num_vfs = dev_num_vf(dev->dev.parent);
1130 size_t size = nla_total_size(0);
1131 size += num_vfs *
1132 (nla_total_size(0) +
1133 nla_total_size(sizeof(struct ifla_vf_mac)) +
1134 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
1135 nla_total_size(sizeof(struct ifla_vf_vlan)) +
1136 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
1137 nla_total_size(MAX_VLAN_LIST_LEN *
1138 sizeof(struct ifla_vf_vlan_info)) +
1139 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
1140 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
1141 nla_total_size(sizeof(struct ifla_vf_rate)) +
1142 nla_total_size(sizeof(struct ifla_vf_link_state)) +
1143 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
1144 nla_total_size(sizeof(struct ifla_vf_trust)));
1145 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1146 size += num_vfs *
1147 (nla_total_size(0) + /* nest IFLA_VF_STATS */
1148 /* IFLA_VF_STATS_RX_PACKETS */
1149 nla_total_size_64bit(sizeof(__u64)) +
1150 /* IFLA_VF_STATS_TX_PACKETS */
1151 nla_total_size_64bit(sizeof(__u64)) +
1152 /* IFLA_VF_STATS_RX_BYTES */
1153 nla_total_size_64bit(sizeof(__u64)) +
1154 /* IFLA_VF_STATS_TX_BYTES */
1155 nla_total_size_64bit(sizeof(__u64)) +
1156 /* IFLA_VF_STATS_BROADCAST */
1157 nla_total_size_64bit(sizeof(__u64)) +
1158 /* IFLA_VF_STATS_MULTICAST */
1159 nla_total_size_64bit(sizeof(__u64)) +
1160 /* IFLA_VF_STATS_RX_DROPPED */
1161 nla_total_size_64bit(sizeof(__u64)) +
1162 /* IFLA_VF_STATS_TX_DROPPED */
1163 nla_total_size_64bit(sizeof(__u64)));
1165 return size;
1166 } else
1167 return 0;
1170 static size_t rtnl_port_size(const struct net_device *dev,
1171 u32 ext_filter_mask)
1173 size_t port_size = nla_total_size(4) /* PORT_VF */
1174 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
1175 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
1176 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
1177 + nla_total_size(1) /* PROT_VDP_REQUEST */
1178 + nla_total_size(2); /* PORT_VDP_RESPONSE */
1179 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
1180 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
1181 + port_size;
1182 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
1183 + port_size;
1185 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1186 !(ext_filter_mask & RTEXT_FILTER_VF))
1187 return 0;
1188 if (dev_num_vf(dev->dev.parent))
1189 return port_self_size + vf_ports_size +
1190 vf_port_size * dev_num_vf(dev->dev.parent);
1191 else
1192 return port_self_size;
1195 static size_t rtnl_xdp_size(void)
1197 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
1198 nla_total_size(1) + /* XDP_ATTACHED */
1199 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
1200 nla_total_size(4); /* XDP_<mode>_PROG_ID */
1202 return xdp_size;
1205 static size_t rtnl_prop_list_size(const struct net_device *dev)
1207 struct netdev_name_node *name_node;
1208 unsigned int cnt = 0;
1210 rcu_read_lock();
1211 list_for_each_entry_rcu(name_node, &dev->name_node->list, list)
1212 cnt++;
1213 rcu_read_unlock();
1215 if (!cnt)
1216 return 0;
1218 return nla_total_size(0) + cnt * nla_total_size(ALTIFNAMSIZ);
1221 static size_t rtnl_proto_down_size(const struct net_device *dev)
1223 size_t size = nla_total_size(1);
1225 /* Assume dev->proto_down_reason is not zero. */
1226 size += nla_total_size(0) + nla_total_size(4);
1228 return size;
1231 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1233 size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1235 if (dev->devlink_port)
1236 size += devlink_nl_port_handle_size(dev->devlink_port);
1238 return size;
1241 static size_t rtnl_dpll_pin_size(const struct net_device *dev)
1243 size_t size = nla_total_size(0); /* nest IFLA_DPLL_PIN */
1245 size += dpll_netdev_pin_handle_size(dev);
1247 return size;
1250 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1251 u32 ext_filter_mask)
1253 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1254 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1255 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1256 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1257 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1258 + nla_total_size(sizeof(struct rtnl_link_stats))
1259 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1260 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1261 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1262 + nla_total_size(4) /* IFLA_TXQLEN */
1263 + nla_total_size(4) /* IFLA_WEIGHT */
1264 + nla_total_size(4) /* IFLA_MTU */
1265 + nla_total_size(4) /* IFLA_LINK */
1266 + nla_total_size(4) /* IFLA_MASTER */
1267 + nla_total_size(1) /* IFLA_CARRIER */
1268 + nla_total_size(4) /* IFLA_PROMISCUITY */
1269 + nla_total_size(4) /* IFLA_ALLMULTI */
1270 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1271 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1272 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1273 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1274 + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1275 + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1276 + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1277 + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1278 + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1279 + nla_total_size(1) /* IFLA_OPERSTATE */
1280 + nla_total_size(1) /* IFLA_LINKMODE */
1281 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1282 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1283 + nla_total_size(4) /* IFLA_GROUP */
1284 + nla_total_size(ext_filter_mask
1285 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1286 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1287 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1288 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1289 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1290 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1291 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1292 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1293 + rtnl_xdp_size() /* IFLA_XDP */
1294 + nla_total_size(4) /* IFLA_EVENT */
1295 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1296 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1297 + rtnl_proto_down_size(dev) /* proto down */
1298 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1299 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1300 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1301 + nla_total_size(4) /* IFLA_MIN_MTU */
1302 + nla_total_size(4) /* IFLA_MAX_MTU */
1303 + rtnl_prop_list_size(dev)
1304 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1305 + rtnl_devlink_port_size(dev)
1306 + rtnl_dpll_pin_size(dev)
1307 + nla_total_size(8) /* IFLA_MAX_PACING_OFFLOAD_HORIZON */
1308 + 0;
1311 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1313 struct nlattr *vf_ports;
1314 struct nlattr *vf_port;
1315 int vf;
1316 int err;
1318 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1319 if (!vf_ports)
1320 return -EMSGSIZE;
1322 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1323 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1324 if (!vf_port)
1325 goto nla_put_failure;
1326 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1327 goto nla_put_failure;
1328 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1329 if (err == -EMSGSIZE)
1330 goto nla_put_failure;
1331 if (err) {
1332 nla_nest_cancel(skb, vf_port);
1333 continue;
1335 nla_nest_end(skb, vf_port);
1338 nla_nest_end(skb, vf_ports);
1340 return 0;
1342 nla_put_failure:
1343 nla_nest_cancel(skb, vf_ports);
1344 return -EMSGSIZE;
1347 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1349 struct nlattr *port_self;
1350 int err;
1352 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1353 if (!port_self)
1354 return -EMSGSIZE;
1356 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1357 if (err) {
1358 nla_nest_cancel(skb, port_self);
1359 return (err == -EMSGSIZE) ? err : 0;
1362 nla_nest_end(skb, port_self);
1364 return 0;
1367 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1368 u32 ext_filter_mask)
1370 int err;
1372 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1373 !(ext_filter_mask & RTEXT_FILTER_VF))
1374 return 0;
1376 err = rtnl_port_self_fill(skb, dev);
1377 if (err)
1378 return err;
1380 if (dev_num_vf(dev->dev.parent)) {
1381 err = rtnl_vf_ports_fill(skb, dev);
1382 if (err)
1383 return err;
1386 return 0;
1389 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1391 int err;
1392 struct netdev_phys_item_id ppid;
1394 err = dev_get_phys_port_id(dev, &ppid);
1395 if (err) {
1396 if (err == -EOPNOTSUPP)
1397 return 0;
1398 return err;
1401 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1402 return -EMSGSIZE;
1404 return 0;
1407 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1409 char name[IFNAMSIZ];
1410 int err;
1412 err = dev_get_phys_port_name(dev, name, sizeof(name));
1413 if (err) {
1414 if (err == -EOPNOTSUPP)
1415 return 0;
1416 return err;
1419 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1420 return -EMSGSIZE;
1422 return 0;
1425 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1427 struct netdev_phys_item_id ppid = { };
1428 int err;
1430 err = dev_get_port_parent_id(dev, &ppid, false);
1431 if (err) {
1432 if (err == -EOPNOTSUPP)
1433 return 0;
1434 return err;
1437 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1438 return -EMSGSIZE;
1440 return 0;
1443 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1444 struct net_device *dev)
1446 struct rtnl_link_stats64 *sp;
1447 struct nlattr *attr;
1449 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1450 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1451 if (!attr)
1452 return -EMSGSIZE;
1454 sp = nla_data(attr);
1455 dev_get_stats(dev, sp);
1457 attr = nla_reserve(skb, IFLA_STATS,
1458 sizeof(struct rtnl_link_stats));
1459 if (!attr)
1460 return -EMSGSIZE;
1462 copy_rtnl_link_stats(nla_data(attr), sp);
1464 return 0;
1467 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1468 struct net_device *dev,
1469 int vfs_num,
1470 u32 ext_filter_mask)
1472 struct ifla_vf_rss_query_en vf_rss_query_en;
1473 struct nlattr *vf, *vfstats, *vfvlanlist;
1474 struct ifla_vf_link_state vf_linkstate;
1475 struct ifla_vf_vlan_info vf_vlan_info;
1476 struct ifla_vf_spoofchk vf_spoofchk;
1477 struct ifla_vf_tx_rate vf_tx_rate;
1478 struct ifla_vf_stats vf_stats;
1479 struct ifla_vf_trust vf_trust;
1480 struct ifla_vf_vlan vf_vlan;
1481 struct ifla_vf_rate vf_rate;
1482 struct ifla_vf_mac vf_mac;
1483 struct ifla_vf_broadcast vf_broadcast;
1484 struct ifla_vf_info ivi;
1485 struct ifla_vf_guid node_guid;
1486 struct ifla_vf_guid port_guid;
1488 memset(&ivi, 0, sizeof(ivi));
1490 /* Not all SR-IOV capable drivers support the
1491 * spoofcheck and "RSS query enable" query. Preset to
1492 * -1 so the user space tool can detect that the driver
1493 * didn't report anything.
1495 ivi.spoofchk = -1;
1496 ivi.rss_query_en = -1;
1497 ivi.trusted = -1;
1498 /* The default value for VF link state is "auto"
1499 * IFLA_VF_LINK_STATE_AUTO which equals zero
1501 ivi.linkstate = 0;
1502 /* VLAN Protocol by default is 802.1Q */
1503 ivi.vlan_proto = htons(ETH_P_8021Q);
1504 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1505 return 0;
1507 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1508 memset(&node_guid, 0, sizeof(node_guid));
1509 memset(&port_guid, 0, sizeof(port_guid));
1511 vf_mac.vf =
1512 vf_vlan.vf =
1513 vf_vlan_info.vf =
1514 vf_rate.vf =
1515 vf_tx_rate.vf =
1516 vf_spoofchk.vf =
1517 vf_linkstate.vf =
1518 vf_rss_query_en.vf =
1519 vf_trust.vf =
1520 node_guid.vf =
1521 port_guid.vf = ivi.vf;
1523 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1524 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1525 vf_vlan.vlan = ivi.vlan;
1526 vf_vlan.qos = ivi.qos;
1527 vf_vlan_info.vlan = ivi.vlan;
1528 vf_vlan_info.qos = ivi.qos;
1529 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1530 vf_tx_rate.rate = ivi.max_tx_rate;
1531 vf_rate.min_tx_rate = ivi.min_tx_rate;
1532 vf_rate.max_tx_rate = ivi.max_tx_rate;
1533 vf_spoofchk.setting = ivi.spoofchk;
1534 vf_linkstate.link_state = ivi.linkstate;
1535 vf_rss_query_en.setting = ivi.rss_query_en;
1536 vf_trust.setting = ivi.trusted;
1537 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1538 if (!vf)
1539 return -EMSGSIZE;
1540 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1541 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1542 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1543 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1544 &vf_rate) ||
1545 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1546 &vf_tx_rate) ||
1547 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1548 &vf_spoofchk) ||
1549 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1550 &vf_linkstate) ||
1551 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1552 sizeof(vf_rss_query_en),
1553 &vf_rss_query_en) ||
1554 nla_put(skb, IFLA_VF_TRUST,
1555 sizeof(vf_trust), &vf_trust))
1556 goto nla_put_vf_failure;
1558 if (dev->netdev_ops->ndo_get_vf_guid &&
1559 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1560 &port_guid)) {
1561 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1562 &node_guid) ||
1563 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1564 &port_guid))
1565 goto nla_put_vf_failure;
1567 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1568 if (!vfvlanlist)
1569 goto nla_put_vf_failure;
1570 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1571 &vf_vlan_info)) {
1572 nla_nest_cancel(skb, vfvlanlist);
1573 goto nla_put_vf_failure;
1575 nla_nest_end(skb, vfvlanlist);
1576 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1577 memset(&vf_stats, 0, sizeof(vf_stats));
1578 if (dev->netdev_ops->ndo_get_vf_stats)
1579 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1580 &vf_stats);
1581 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1582 if (!vfstats)
1583 goto nla_put_vf_failure;
1584 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1585 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1586 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1587 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1588 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1589 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1590 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1591 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1592 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1593 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1594 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1595 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1596 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1597 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1598 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1599 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1600 nla_nest_cancel(skb, vfstats);
1601 goto nla_put_vf_failure;
1603 nla_nest_end(skb, vfstats);
1605 nla_nest_end(skb, vf);
1606 return 0;
1608 nla_put_vf_failure:
1609 nla_nest_cancel(skb, vf);
1610 return -EMSGSIZE;
1613 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1614 struct net_device *dev,
1615 u32 ext_filter_mask)
1617 struct nlattr *vfinfo;
1618 int i, num_vfs;
1620 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1621 return 0;
1623 num_vfs = dev_num_vf(dev->dev.parent);
1624 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1625 return -EMSGSIZE;
1627 if (!dev->netdev_ops->ndo_get_vf_config)
1628 return 0;
1630 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1631 if (!vfinfo)
1632 return -EMSGSIZE;
1634 for (i = 0; i < num_vfs; i++) {
1635 if (rtnl_fill_vfinfo(skb, dev, i, ext_filter_mask)) {
1636 nla_nest_cancel(skb, vfinfo);
1637 return -EMSGSIZE;
1641 nla_nest_end(skb, vfinfo);
1642 return 0;
1645 static int rtnl_fill_link_ifmap(struct sk_buff *skb,
1646 const struct net_device *dev)
1648 struct rtnl_link_ifmap map;
1650 memset(&map, 0, sizeof(map));
1651 map.mem_start = READ_ONCE(dev->mem_start);
1652 map.mem_end = READ_ONCE(dev->mem_end);
1653 map.base_addr = READ_ONCE(dev->base_addr);
1654 map.irq = READ_ONCE(dev->irq);
1655 map.dma = READ_ONCE(dev->dma);
1656 map.port = READ_ONCE(dev->if_port);
1658 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1659 return -EMSGSIZE;
1661 return 0;
1664 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1666 const struct bpf_prog *generic_xdp_prog;
1667 u32 res = 0;
1669 rcu_read_lock();
1670 generic_xdp_prog = rcu_dereference(dev->xdp_prog);
1671 if (generic_xdp_prog)
1672 res = generic_xdp_prog->aux->id;
1673 rcu_read_unlock();
1675 return res;
1678 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1680 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1683 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1685 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1688 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1689 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1690 u32 (*get_prog_id)(struct net_device *dev))
1692 u32 curr_id;
1693 int err;
1695 curr_id = get_prog_id(dev);
1696 if (!curr_id)
1697 return 0;
1699 *prog_id = curr_id;
1700 err = nla_put_u32(skb, attr, curr_id);
1701 if (err)
1702 return err;
1704 if (*mode != XDP_ATTACHED_NONE)
1705 *mode = XDP_ATTACHED_MULTI;
1706 else
1707 *mode = tgt_mode;
1709 return 0;
1712 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1714 struct nlattr *xdp;
1715 u32 prog_id;
1716 int err;
1717 u8 mode;
1719 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1720 if (!xdp)
1721 return -EMSGSIZE;
1723 prog_id = 0;
1724 mode = XDP_ATTACHED_NONE;
1725 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1726 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1727 if (err)
1728 goto err_cancel;
1729 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1730 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1731 if (err)
1732 goto err_cancel;
1733 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1734 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1735 if (err)
1736 goto err_cancel;
1738 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1739 if (err)
1740 goto err_cancel;
1742 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1743 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1744 if (err)
1745 goto err_cancel;
1748 nla_nest_end(skb, xdp);
1749 return 0;
1751 err_cancel:
1752 nla_nest_cancel(skb, xdp);
1753 return err;
1756 static u32 rtnl_get_event(unsigned long event)
1758 u32 rtnl_event_type = IFLA_EVENT_NONE;
1760 switch (event) {
1761 case NETDEV_REBOOT:
1762 rtnl_event_type = IFLA_EVENT_REBOOT;
1763 break;
1764 case NETDEV_FEAT_CHANGE:
1765 rtnl_event_type = IFLA_EVENT_FEATURES;
1766 break;
1767 case NETDEV_BONDING_FAILOVER:
1768 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1769 break;
1770 case NETDEV_NOTIFY_PEERS:
1771 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1772 break;
1773 case NETDEV_RESEND_IGMP:
1774 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1775 break;
1776 case NETDEV_CHANGEINFODATA:
1777 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1778 break;
1779 default:
1780 break;
1783 return rtnl_event_type;
1786 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1788 const struct net_device *upper_dev;
1789 int ret = 0;
1791 rcu_read_lock();
1793 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1794 if (upper_dev)
1795 ret = nla_put_u32(skb, IFLA_MASTER,
1796 READ_ONCE(upper_dev->ifindex));
1798 rcu_read_unlock();
1799 return ret;
1802 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1803 bool force)
1805 int iflink = dev_get_iflink(dev);
1807 if (force || READ_ONCE(dev->ifindex) != iflink)
1808 return nla_put_u32(skb, IFLA_LINK, iflink);
1810 return 0;
1813 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1814 struct net_device *dev)
1816 char buf[IFALIASZ];
1817 int ret;
1819 ret = dev_get_alias(dev, buf, sizeof(buf));
1820 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1823 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1824 const struct net_device *dev,
1825 struct net *src_net, gfp_t gfp)
1827 bool put_iflink = false;
1829 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1830 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1832 if (!net_eq(dev_net(dev), link_net)) {
1833 int id = peernet2id_alloc(src_net, link_net, gfp);
1835 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1836 return -EMSGSIZE;
1838 put_iflink = true;
1842 return nla_put_iflink(skb, dev, put_iflink);
1845 static int rtnl_fill_link_af(struct sk_buff *skb,
1846 const struct net_device *dev,
1847 u32 ext_filter_mask)
1849 const struct rtnl_af_ops *af_ops;
1850 struct nlattr *af_spec;
1852 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1853 if (!af_spec)
1854 return -EMSGSIZE;
1856 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1857 struct nlattr *af;
1858 int err;
1860 if (!af_ops->fill_link_af)
1861 continue;
1863 af = nla_nest_start_noflag(skb, af_ops->family);
1864 if (!af)
1865 return -EMSGSIZE;
1867 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1869 * Caller may return ENODATA to indicate that there
1870 * was no data to be dumped. This is not an error, it
1871 * means we should trim the attribute header and
1872 * continue.
1874 if (err == -ENODATA)
1875 nla_nest_cancel(skb, af);
1876 else if (err < 0)
1877 return -EMSGSIZE;
1879 nla_nest_end(skb, af);
1882 nla_nest_end(skb, af_spec);
1883 return 0;
1886 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1887 const struct net_device *dev)
1889 struct netdev_name_node *name_node;
1890 int count = 0;
1892 list_for_each_entry_rcu(name_node, &dev->name_node->list, list) {
1893 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1894 return -EMSGSIZE;
1895 count++;
1897 return count;
1900 /* RCU protected. */
1901 static int rtnl_fill_prop_list(struct sk_buff *skb,
1902 const struct net_device *dev)
1904 struct nlattr *prop_list;
1905 int ret;
1907 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1908 if (!prop_list)
1909 return -EMSGSIZE;
1911 ret = rtnl_fill_alt_ifnames(skb, dev);
1912 if (ret <= 0)
1913 goto nest_cancel;
1915 nla_nest_end(skb, prop_list);
1916 return 0;
1918 nest_cancel:
1919 nla_nest_cancel(skb, prop_list);
1920 return ret;
1923 static int rtnl_fill_proto_down(struct sk_buff *skb,
1924 const struct net_device *dev)
1926 struct nlattr *pr;
1927 u32 preason;
1929 if (nla_put_u8(skb, IFLA_PROTO_DOWN, READ_ONCE(dev->proto_down)))
1930 goto nla_put_failure;
1932 preason = READ_ONCE(dev->proto_down_reason);
1933 if (!preason)
1934 return 0;
1936 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1937 if (!pr)
1938 return -EMSGSIZE;
1940 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1941 nla_nest_cancel(skb, pr);
1942 goto nla_put_failure;
1945 nla_nest_end(skb, pr);
1946 return 0;
1948 nla_put_failure:
1949 return -EMSGSIZE;
1952 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1953 const struct net_device *dev)
1955 struct nlattr *devlink_port_nest;
1956 int ret;
1958 devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1959 if (!devlink_port_nest)
1960 return -EMSGSIZE;
1962 if (dev->devlink_port) {
1963 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1964 if (ret < 0)
1965 goto nest_cancel;
1968 nla_nest_end(skb, devlink_port_nest);
1969 return 0;
1971 nest_cancel:
1972 nla_nest_cancel(skb, devlink_port_nest);
1973 return ret;
1976 static int rtnl_fill_dpll_pin(struct sk_buff *skb,
1977 const struct net_device *dev)
1979 struct nlattr *dpll_pin_nest;
1980 int ret;
1982 dpll_pin_nest = nla_nest_start(skb, IFLA_DPLL_PIN);
1983 if (!dpll_pin_nest)
1984 return -EMSGSIZE;
1986 ret = dpll_netdev_add_pin_handle(skb, dev);
1987 if (ret < 0)
1988 goto nest_cancel;
1990 nla_nest_end(skb, dpll_pin_nest);
1991 return 0;
1993 nest_cancel:
1994 nla_nest_cancel(skb, dpll_pin_nest);
1995 return ret;
1998 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1999 struct net_device *dev, struct net *src_net,
2000 int type, u32 pid, u32 seq, u32 change,
2001 unsigned int flags, u32 ext_filter_mask,
2002 u32 event, int *new_nsid, int new_ifindex,
2003 int tgt_netnsid, gfp_t gfp)
2005 char devname[IFNAMSIZ];
2006 struct ifinfomsg *ifm;
2007 struct nlmsghdr *nlh;
2008 struct Qdisc *qdisc;
2010 ASSERT_RTNL();
2011 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
2012 if (nlh == NULL)
2013 return -EMSGSIZE;
2015 ifm = nlmsg_data(nlh);
2016 ifm->ifi_family = AF_UNSPEC;
2017 ifm->__ifi_pad = 0;
2018 ifm->ifi_type = READ_ONCE(dev->type);
2019 ifm->ifi_index = READ_ONCE(dev->ifindex);
2020 ifm->ifi_flags = dev_get_flags(dev);
2021 ifm->ifi_change = change;
2023 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
2024 goto nla_put_failure;
2026 netdev_copy_name(dev, devname);
2027 if (nla_put_string(skb, IFLA_IFNAME, devname))
2028 goto nla_put_failure;
2030 if (nla_put_u32(skb, IFLA_TXQLEN, READ_ONCE(dev->tx_queue_len)) ||
2031 nla_put_u8(skb, IFLA_OPERSTATE,
2032 netif_running(dev) ? READ_ONCE(dev->operstate) :
2033 IF_OPER_DOWN) ||
2034 nla_put_u8(skb, IFLA_LINKMODE, READ_ONCE(dev->link_mode)) ||
2035 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
2036 nla_put_u32(skb, IFLA_MIN_MTU, READ_ONCE(dev->min_mtu)) ||
2037 nla_put_u32(skb, IFLA_MAX_MTU, READ_ONCE(dev->max_mtu)) ||
2038 nla_put_u32(skb, IFLA_GROUP, READ_ONCE(dev->group)) ||
2039 nla_put_u32(skb, IFLA_PROMISCUITY, READ_ONCE(dev->promiscuity)) ||
2040 nla_put_u32(skb, IFLA_ALLMULTI, READ_ONCE(dev->allmulti)) ||
2041 nla_put_u32(skb, IFLA_NUM_TX_QUEUES,
2042 READ_ONCE(dev->num_tx_queues)) ||
2043 nla_put_u32(skb, IFLA_GSO_MAX_SEGS,
2044 READ_ONCE(dev->gso_max_segs)) ||
2045 nla_put_u32(skb, IFLA_GSO_MAX_SIZE,
2046 READ_ONCE(dev->gso_max_size)) ||
2047 nla_put_u32(skb, IFLA_GRO_MAX_SIZE,
2048 READ_ONCE(dev->gro_max_size)) ||
2049 nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE,
2050 READ_ONCE(dev->gso_ipv4_max_size)) ||
2051 nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE,
2052 READ_ONCE(dev->gro_ipv4_max_size)) ||
2053 nla_put_u32(skb, IFLA_TSO_MAX_SIZE,
2054 READ_ONCE(dev->tso_max_size)) ||
2055 nla_put_u32(skb, IFLA_TSO_MAX_SEGS,
2056 READ_ONCE(dev->tso_max_segs)) ||
2057 nla_put_uint(skb, IFLA_MAX_PACING_OFFLOAD_HORIZON,
2058 READ_ONCE(dev->max_pacing_offload_horizon)) ||
2059 #ifdef CONFIG_RPS
2060 nla_put_u32(skb, IFLA_NUM_RX_QUEUES,
2061 READ_ONCE(dev->num_rx_queues)) ||
2062 #endif
2063 put_master_ifindex(skb, dev) ||
2064 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
2065 nla_put_ifalias(skb, dev) ||
2066 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
2067 atomic_read(&dev->carrier_up_count) +
2068 atomic_read(&dev->carrier_down_count)) ||
2069 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
2070 atomic_read(&dev->carrier_up_count)) ||
2071 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
2072 atomic_read(&dev->carrier_down_count)))
2073 goto nla_put_failure;
2075 if (rtnl_fill_proto_down(skb, dev))
2076 goto nla_put_failure;
2078 if (event != IFLA_EVENT_NONE) {
2079 if (nla_put_u32(skb, IFLA_EVENT, event))
2080 goto nla_put_failure;
2083 if (dev->addr_len) {
2084 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
2085 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
2086 goto nla_put_failure;
2089 if (rtnl_phys_port_id_fill(skb, dev))
2090 goto nla_put_failure;
2092 if (rtnl_phys_port_name_fill(skb, dev))
2093 goto nla_put_failure;
2095 if (rtnl_phys_switch_id_fill(skb, dev))
2096 goto nla_put_failure;
2098 if (rtnl_fill_stats(skb, dev))
2099 goto nla_put_failure;
2101 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
2102 goto nla_put_failure;
2104 if (rtnl_port_fill(skb, dev, ext_filter_mask))
2105 goto nla_put_failure;
2107 if (rtnl_xdp_fill(skb, dev))
2108 goto nla_put_failure;
2110 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
2111 if (rtnl_link_fill(skb, dev) < 0)
2112 goto nla_put_failure;
2115 if (new_nsid &&
2116 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
2117 goto nla_put_failure;
2118 if (new_ifindex &&
2119 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
2120 goto nla_put_failure;
2122 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
2123 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
2124 goto nla_put_failure;
2126 rcu_read_lock();
2127 if (rtnl_fill_link_netnsid(skb, dev, src_net, GFP_ATOMIC))
2128 goto nla_put_failure_rcu;
2129 qdisc = rcu_dereference(dev->qdisc);
2130 if (qdisc && nla_put_string(skb, IFLA_QDISC, qdisc->ops->id))
2131 goto nla_put_failure_rcu;
2132 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
2133 goto nla_put_failure_rcu;
2134 if (rtnl_fill_link_ifmap(skb, dev))
2135 goto nla_put_failure_rcu;
2136 if (rtnl_fill_prop_list(skb, dev))
2137 goto nla_put_failure_rcu;
2138 rcu_read_unlock();
2140 if (dev->dev.parent &&
2141 nla_put_string(skb, IFLA_PARENT_DEV_NAME,
2142 dev_name(dev->dev.parent)))
2143 goto nla_put_failure;
2145 if (dev->dev.parent && dev->dev.parent->bus &&
2146 nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
2147 dev->dev.parent->bus->name))
2148 goto nla_put_failure;
2150 if (rtnl_fill_devlink_port(skb, dev))
2151 goto nla_put_failure;
2153 if (rtnl_fill_dpll_pin(skb, dev))
2154 goto nla_put_failure;
2156 nlmsg_end(skb, nlh);
2157 return 0;
2159 nla_put_failure_rcu:
2160 rcu_read_unlock();
2161 nla_put_failure:
2162 nlmsg_cancel(skb, nlh);
2163 return -EMSGSIZE;
2166 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
2167 [IFLA_UNSPEC] = { .strict_start_type = IFLA_DPLL_PIN },
2168 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
2169 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
2170 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
2171 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
2172 [IFLA_MTU] = { .type = NLA_U32 },
2173 [IFLA_LINK] = { .type = NLA_U32 },
2174 [IFLA_MASTER] = { .type = NLA_U32 },
2175 [IFLA_CARRIER] = { .type = NLA_U8 },
2176 [IFLA_TXQLEN] = { .type = NLA_U32 },
2177 [IFLA_WEIGHT] = { .type = NLA_U32 },
2178 [IFLA_OPERSTATE] = { .type = NLA_U8 },
2179 [IFLA_LINKMODE] = { .type = NLA_U8 },
2180 [IFLA_LINKINFO] = { .type = NLA_NESTED },
2181 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
2182 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
2183 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
2184 * allow 0-length string (needed to remove an alias).
2186 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
2187 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
2188 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
2189 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
2190 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
2191 [IFLA_EXT_MASK] = { .type = NLA_U32 },
2192 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
2193 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
2194 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
2195 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
2196 [IFLA_GSO_MAX_SIZE] = NLA_POLICY_MIN(NLA_U32, MAX_TCP_HEADER + 1),
2197 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2198 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
2199 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2200 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
2201 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
2202 [IFLA_XDP] = { .type = NLA_NESTED },
2203 [IFLA_EVENT] = { .type = NLA_U32 },
2204 [IFLA_GROUP] = { .type = NLA_U32 },
2205 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
2206 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
2207 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
2208 [IFLA_MIN_MTU] = { .type = NLA_U32 },
2209 [IFLA_MAX_MTU] = { .type = NLA_U32 },
2210 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
2211 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
2212 .len = ALTIFNAMSIZ - 1 },
2213 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
2214 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
2215 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
2216 [IFLA_PARENT_DEV_NAME] = { .type = NLA_NUL_STRING },
2217 [IFLA_GRO_MAX_SIZE] = { .type = NLA_U32 },
2218 [IFLA_TSO_MAX_SIZE] = { .type = NLA_REJECT },
2219 [IFLA_TSO_MAX_SEGS] = { .type = NLA_REJECT },
2220 [IFLA_ALLMULTI] = { .type = NLA_REJECT },
2221 [IFLA_GSO_IPV4_MAX_SIZE] = NLA_POLICY_MIN(NLA_U32, MAX_TCP_HEADER + 1),
2222 [IFLA_GRO_IPV4_MAX_SIZE] = { .type = NLA_U32 },
2225 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
2226 [IFLA_INFO_KIND] = { .type = NLA_STRING },
2227 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
2228 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
2229 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
2232 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
2233 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
2234 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
2235 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
2236 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
2237 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
2238 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
2239 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
2240 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
2241 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
2242 [IFLA_VF_STATS] = { .type = NLA_NESTED },
2243 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
2244 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
2245 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
2248 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2249 [IFLA_PORT_VF] = { .type = NLA_U32 },
2250 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
2251 .len = PORT_PROFILE_MAX },
2252 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2253 .len = PORT_UUID_MAX },
2254 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
2255 .len = PORT_UUID_MAX },
2256 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
2257 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
2259 /* Unused, but we need to keep it here since user space could
2260 * fill it. It's also broken with regard to NLA_BINARY use in
2261 * combination with structs.
2263 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
2264 .len = sizeof(struct ifla_port_vsi) },
2267 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2268 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2269 [IFLA_XDP_FD] = { .type = NLA_S32 },
2270 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
2271 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
2272 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
2273 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
2276 static struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla,
2277 int *ops_srcu_index)
2279 struct nlattr *linfo[IFLA_INFO_MAX + 1];
2280 struct rtnl_link_ops *ops = NULL;
2282 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2283 return NULL;
2285 if (linfo[IFLA_INFO_KIND]) {
2286 char kind[MODULE_NAME_LEN];
2288 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2289 ops = rtnl_link_ops_get(kind, ops_srcu_index);
2292 return ops;
2295 static bool link_master_filtered(struct net_device *dev, int master_idx)
2297 struct net_device *master;
2299 if (!master_idx)
2300 return false;
2302 master = netdev_master_upper_dev_get(dev);
2304 /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2305 * another invalid value for ifindex to denote "no master".
2307 if (master_idx == -1)
2308 return !!master;
2310 if (!master || master->ifindex != master_idx)
2311 return true;
2313 return false;
2316 static bool link_kind_filtered(const struct net_device *dev,
2317 const struct rtnl_link_ops *kind_ops)
2319 if (kind_ops && dev->rtnl_link_ops != kind_ops)
2320 return true;
2322 return false;
2325 static bool link_dump_filtered(struct net_device *dev,
2326 int master_idx,
2327 const struct rtnl_link_ops *kind_ops)
2329 if (link_master_filtered(dev, master_idx) ||
2330 link_kind_filtered(dev, kind_ops))
2331 return true;
2333 return false;
2337 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2338 * @sk: netlink socket
2339 * @netnsid: network namespace identifier
2341 * Returns the network namespace identified by netnsid on success or an error
2342 * pointer on failure.
2344 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2346 struct net *net;
2348 net = get_net_ns_by_id(sock_net(sk), netnsid);
2349 if (!net)
2350 return ERR_PTR(-EINVAL);
2352 /* For now, the caller is required to have CAP_NET_ADMIN in
2353 * the user namespace owning the target net ns.
2355 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2356 put_net(net);
2357 return ERR_PTR(-EACCES);
2359 return net;
2361 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2363 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2364 bool strict_check, struct nlattr **tb,
2365 struct netlink_ext_ack *extack)
2367 int hdrlen;
2369 if (strict_check) {
2370 struct ifinfomsg *ifm;
2372 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2373 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2374 return -EINVAL;
2377 ifm = nlmsg_data(nlh);
2378 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2379 ifm->ifi_change) {
2380 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2381 return -EINVAL;
2383 if (ifm->ifi_index) {
2384 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2385 return -EINVAL;
2388 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2389 IFLA_MAX, ifla_policy,
2390 extack);
2393 /* A hack to preserve kernel<->userspace interface.
2394 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2395 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2396 * what iproute2 < v3.9.0 used.
2397 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2398 * attribute, its netlink message is shorter than struct ifinfomsg.
2400 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2401 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2403 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2404 extack);
2407 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2409 struct netlink_ext_ack *extack = cb->extack;
2410 struct rtnl_link_ops *kind_ops = NULL;
2411 const struct nlmsghdr *nlh = cb->nlh;
2412 struct net *net = sock_net(skb->sk);
2413 unsigned int flags = NLM_F_MULTI;
2414 struct nlattr *tb[IFLA_MAX+1];
2415 struct {
2416 unsigned long ifindex;
2417 } *ctx = (void *)cb->ctx;
2418 struct net *tgt_net = net;
2419 u32 ext_filter_mask = 0;
2420 struct net_device *dev;
2421 int ops_srcu_index;
2422 int master_idx = 0;
2423 int netnsid = -1;
2424 int err, i;
2426 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2427 if (err < 0) {
2428 if (cb->strict_check)
2429 return err;
2431 goto walk_entries;
2434 for (i = 0; i <= IFLA_MAX; ++i) {
2435 if (!tb[i])
2436 continue;
2438 /* new attributes should only be added with strict checking */
2439 switch (i) {
2440 case IFLA_TARGET_NETNSID:
2441 netnsid = nla_get_s32(tb[i]);
2442 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2443 if (IS_ERR(tgt_net)) {
2444 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2445 err = PTR_ERR(tgt_net);
2446 netnsid = -1;
2447 goto out;
2449 break;
2450 case IFLA_EXT_MASK:
2451 ext_filter_mask = nla_get_u32(tb[i]);
2452 break;
2453 case IFLA_MASTER:
2454 master_idx = nla_get_u32(tb[i]);
2455 break;
2456 case IFLA_LINKINFO:
2457 kind_ops = linkinfo_to_kind_ops(tb[i], &ops_srcu_index);
2458 break;
2459 default:
2460 if (cb->strict_check) {
2461 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2462 err = -EINVAL;
2463 goto out;
2468 if (master_idx || kind_ops)
2469 flags |= NLM_F_DUMP_FILTERED;
2471 walk_entries:
2472 err = 0;
2473 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
2474 if (link_dump_filtered(dev, master_idx, kind_ops))
2475 continue;
2476 err = rtnl_fill_ifinfo(skb, dev, net, RTM_NEWLINK,
2477 NETLINK_CB(cb->skb).portid,
2478 nlh->nlmsg_seq, 0, flags,
2479 ext_filter_mask, 0, NULL, 0,
2480 netnsid, GFP_KERNEL);
2481 if (err < 0)
2482 break;
2486 cb->seq = tgt_net->dev_base_seq;
2487 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2489 out:
2491 if (kind_ops)
2492 rtnl_link_ops_put(kind_ops, ops_srcu_index);
2493 if (netnsid >= 0)
2494 put_net(tgt_net);
2496 return err;
2499 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2500 struct netlink_ext_ack *exterr)
2502 const struct ifinfomsg *ifmp;
2503 const struct nlattr *attrs;
2504 size_t len;
2506 ifmp = nla_data(nla_peer);
2507 attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2508 len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2510 if (ifmp->ifi_index < 0) {
2511 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2512 "ifindex can't be negative");
2513 return -EINVAL;
2516 return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2517 exterr);
2519 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2521 static struct net *rtnl_link_get_net_ifla(struct nlattr *tb[])
2523 struct net *net = NULL;
2525 /* Examine the link attributes and figure out which
2526 * network namespace we are talking about.
2528 if (tb[IFLA_NET_NS_PID])
2529 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2530 else if (tb[IFLA_NET_NS_FD])
2531 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2533 return net;
2536 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2538 struct net *net = rtnl_link_get_net_ifla(tb);
2540 if (!net)
2541 net = get_net(src_net);
2543 return net;
2545 EXPORT_SYMBOL(rtnl_link_get_net);
2547 /* Figure out which network namespace we are talking about by
2548 * examining the link attributes in the following order:
2550 * 1. IFLA_NET_NS_PID
2551 * 2. IFLA_NET_NS_FD
2552 * 3. IFLA_TARGET_NETNSID
2554 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2555 struct nlattr *tb[])
2557 struct net *net;
2559 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2560 return rtnl_link_get_net(src_net, tb);
2562 if (!tb[IFLA_TARGET_NETNSID])
2563 return get_net(src_net);
2565 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2566 if (!net)
2567 return ERR_PTR(-EINVAL);
2569 return net;
2572 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2573 struct net *src_net,
2574 struct nlattr *tb[], int cap)
2576 struct net *net;
2578 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2579 if (IS_ERR(net))
2580 return net;
2582 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2583 put_net(net);
2584 return ERR_PTR(-EPERM);
2587 return net;
2590 /* Verify that rtnetlink requests do not pass additional properties
2591 * potentially referring to different network namespaces.
2593 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2594 struct netlink_ext_ack *extack,
2595 bool netns_id_only)
2598 if (netns_id_only) {
2599 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2600 return 0;
2602 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2603 return -EOPNOTSUPP;
2606 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2607 goto invalid_attr;
2609 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2610 goto invalid_attr;
2612 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2613 goto invalid_attr;
2615 return 0;
2617 invalid_attr:
2618 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2619 return -EINVAL;
2622 static int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2623 int max_tx_rate)
2625 const struct net_device_ops *ops = dev->netdev_ops;
2627 if (!ops->ndo_set_vf_rate)
2628 return -EOPNOTSUPP;
2629 if (max_tx_rate && max_tx_rate < min_tx_rate)
2630 return -EINVAL;
2632 return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2635 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2636 struct netlink_ext_ack *extack)
2638 if (tb[IFLA_ADDRESS] &&
2639 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2640 return -EINVAL;
2642 if (tb[IFLA_BROADCAST] &&
2643 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2644 return -EINVAL;
2646 if (tb[IFLA_GSO_MAX_SIZE] &&
2647 nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2648 NL_SET_ERR_MSG(extack, "too big gso_max_size");
2649 return -EINVAL;
2652 if (tb[IFLA_GSO_MAX_SEGS] &&
2653 (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2654 nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2655 NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2656 return -EINVAL;
2659 if (tb[IFLA_GRO_MAX_SIZE] &&
2660 nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2661 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2662 return -EINVAL;
2665 if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2666 nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2667 NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2668 return -EINVAL;
2671 if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2672 nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2673 NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2674 return -EINVAL;
2677 if (tb[IFLA_AF_SPEC]) {
2678 struct nlattr *af;
2679 int rem, err;
2681 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2682 struct rtnl_af_ops *af_ops;
2683 int af_ops_srcu_index;
2685 af_ops = rtnl_af_lookup(nla_type(af), &af_ops_srcu_index);
2686 if (!af_ops)
2687 return -EAFNOSUPPORT;
2689 if (!af_ops->set_link_af)
2690 err = -EOPNOTSUPP;
2691 else if (af_ops->validate_link_af)
2692 err = af_ops->validate_link_af(dev, af, extack);
2693 else
2694 err = 0;
2696 rtnl_af_put(af_ops, af_ops_srcu_index);
2698 if (err < 0)
2699 return err;
2703 return 0;
2706 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2707 int guid_type)
2709 const struct net_device_ops *ops = dev->netdev_ops;
2711 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2714 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2716 if (dev->type != ARPHRD_INFINIBAND)
2717 return -EOPNOTSUPP;
2719 return handle_infiniband_guid(dev, ivt, guid_type);
2722 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2724 const struct net_device_ops *ops = dev->netdev_ops;
2725 int err = -EINVAL;
2727 if (tb[IFLA_VF_MAC]) {
2728 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2730 if (ivm->vf >= INT_MAX)
2731 return -EINVAL;
2732 err = -EOPNOTSUPP;
2733 if (ops->ndo_set_vf_mac)
2734 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2735 ivm->mac);
2736 if (err < 0)
2737 return err;
2740 if (tb[IFLA_VF_VLAN]) {
2741 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2743 if (ivv->vf >= INT_MAX)
2744 return -EINVAL;
2745 err = -EOPNOTSUPP;
2746 if (ops->ndo_set_vf_vlan)
2747 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2748 ivv->qos,
2749 htons(ETH_P_8021Q));
2750 if (err < 0)
2751 return err;
2754 if (tb[IFLA_VF_VLAN_LIST]) {
2755 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2756 struct nlattr *attr;
2757 int rem, len = 0;
2759 err = -EOPNOTSUPP;
2760 if (!ops->ndo_set_vf_vlan)
2761 return err;
2763 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2764 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2765 nla_len(attr) < sizeof(struct ifla_vf_vlan_info)) {
2766 return -EINVAL;
2768 if (len >= MAX_VLAN_LIST_LEN)
2769 return -EOPNOTSUPP;
2770 ivvl[len] = nla_data(attr);
2772 len++;
2774 if (len == 0)
2775 return -EINVAL;
2777 if (ivvl[0]->vf >= INT_MAX)
2778 return -EINVAL;
2779 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2780 ivvl[0]->qos, ivvl[0]->vlan_proto);
2781 if (err < 0)
2782 return err;
2785 if (tb[IFLA_VF_TX_RATE]) {
2786 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2787 struct ifla_vf_info ivf;
2789 if (ivt->vf >= INT_MAX)
2790 return -EINVAL;
2791 err = -EOPNOTSUPP;
2792 if (ops->ndo_get_vf_config)
2793 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2794 if (err < 0)
2795 return err;
2797 err = rtnl_set_vf_rate(dev, ivt->vf,
2798 ivf.min_tx_rate, ivt->rate);
2799 if (err < 0)
2800 return err;
2803 if (tb[IFLA_VF_RATE]) {
2804 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2806 if (ivt->vf >= INT_MAX)
2807 return -EINVAL;
2809 err = rtnl_set_vf_rate(dev, ivt->vf,
2810 ivt->min_tx_rate, ivt->max_tx_rate);
2811 if (err < 0)
2812 return err;
2815 if (tb[IFLA_VF_SPOOFCHK]) {
2816 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2818 if (ivs->vf >= INT_MAX)
2819 return -EINVAL;
2820 err = -EOPNOTSUPP;
2821 if (ops->ndo_set_vf_spoofchk)
2822 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2823 ivs->setting);
2824 if (err < 0)
2825 return err;
2828 if (tb[IFLA_VF_LINK_STATE]) {
2829 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2831 if (ivl->vf >= INT_MAX)
2832 return -EINVAL;
2833 err = -EOPNOTSUPP;
2834 if (ops->ndo_set_vf_link_state)
2835 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2836 ivl->link_state);
2837 if (err < 0)
2838 return err;
2841 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2842 struct ifla_vf_rss_query_en *ivrssq_en;
2844 err = -EOPNOTSUPP;
2845 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2846 if (ivrssq_en->vf >= INT_MAX)
2847 return -EINVAL;
2848 if (ops->ndo_set_vf_rss_query_en)
2849 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2850 ivrssq_en->setting);
2851 if (err < 0)
2852 return err;
2855 if (tb[IFLA_VF_TRUST]) {
2856 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2858 if (ivt->vf >= INT_MAX)
2859 return -EINVAL;
2860 err = -EOPNOTSUPP;
2861 if (ops->ndo_set_vf_trust)
2862 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2863 if (err < 0)
2864 return err;
2867 if (tb[IFLA_VF_IB_NODE_GUID]) {
2868 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2870 if (ivt->vf >= INT_MAX)
2871 return -EINVAL;
2872 if (!ops->ndo_set_vf_guid)
2873 return -EOPNOTSUPP;
2874 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2877 if (tb[IFLA_VF_IB_PORT_GUID]) {
2878 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2880 if (ivt->vf >= INT_MAX)
2881 return -EINVAL;
2882 if (!ops->ndo_set_vf_guid)
2883 return -EOPNOTSUPP;
2885 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2888 return err;
2891 static int do_set_master(struct net_device *dev, int ifindex,
2892 struct netlink_ext_ack *extack)
2894 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2895 const struct net_device_ops *ops;
2896 int err;
2898 if (upper_dev) {
2899 if (upper_dev->ifindex == ifindex)
2900 return 0;
2901 ops = upper_dev->netdev_ops;
2902 if (ops->ndo_del_slave) {
2903 err = ops->ndo_del_slave(upper_dev, dev);
2904 if (err)
2905 return err;
2906 } else {
2907 return -EOPNOTSUPP;
2911 if (ifindex) {
2912 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2913 if (!upper_dev)
2914 return -EINVAL;
2915 ops = upper_dev->netdev_ops;
2916 if (ops->ndo_add_slave) {
2917 err = ops->ndo_add_slave(upper_dev, dev, extack);
2918 if (err)
2919 return err;
2920 } else {
2921 return -EOPNOTSUPP;
2924 return 0;
2927 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2928 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2929 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2932 static int do_set_proto_down(struct net_device *dev,
2933 struct nlattr *nl_proto_down,
2934 struct nlattr *nl_proto_down_reason,
2935 struct netlink_ext_ack *extack)
2937 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2938 unsigned long mask = 0;
2939 u32 value;
2940 bool proto_down;
2941 int err;
2943 if (!dev->change_proto_down) {
2944 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2945 return -EOPNOTSUPP;
2948 if (nl_proto_down_reason) {
2949 err = nla_parse_nested_deprecated(pdreason,
2950 IFLA_PROTO_DOWN_REASON_MAX,
2951 nl_proto_down_reason,
2952 ifla_proto_down_reason_policy,
2953 NULL);
2954 if (err < 0)
2955 return err;
2957 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2958 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2959 return -EINVAL;
2962 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2964 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2965 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2967 dev_change_proto_down_reason(dev, mask, value);
2970 if (nl_proto_down) {
2971 proto_down = nla_get_u8(nl_proto_down);
2973 /* Don't turn off protodown if there are active reasons */
2974 if (!proto_down && dev->proto_down_reason) {
2975 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2976 return -EBUSY;
2978 err = dev_change_proto_down(dev,
2979 proto_down);
2980 if (err)
2981 return err;
2984 return 0;
2987 #define DO_SETLINK_MODIFIED 0x01
2988 /* notify flag means notify + modified. */
2989 #define DO_SETLINK_NOTIFY 0x03
2990 static int do_setlink(const struct sk_buff *skb, struct net_device *dev,
2991 struct net *tgt_net, struct ifinfomsg *ifm,
2992 struct netlink_ext_ack *extack,
2993 struct nlattr **tb, int status)
2995 const struct net_device_ops *ops = dev->netdev_ops;
2996 char ifname[IFNAMSIZ];
2997 int err;
2999 err = validate_linkmsg(dev, tb, extack);
3000 if (err < 0)
3001 goto errout;
3003 if (tb[IFLA_IFNAME])
3004 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3005 else
3006 ifname[0] = '\0';
3008 if (!net_eq(tgt_net, dev_net(dev))) {
3009 const char *pat = ifname[0] ? ifname : NULL;
3010 int new_ifindex;
3012 new_ifindex = nla_get_s32_default(tb[IFLA_NEW_IFINDEX], 0);
3014 err = __dev_change_net_namespace(dev, tgt_net, pat, new_ifindex);
3015 if (err)
3016 goto errout;
3018 status |= DO_SETLINK_MODIFIED;
3021 if (tb[IFLA_MAP]) {
3022 struct rtnl_link_ifmap *u_map;
3023 struct ifmap k_map;
3025 if (!ops->ndo_set_config) {
3026 err = -EOPNOTSUPP;
3027 goto errout;
3030 if (!netif_device_present(dev)) {
3031 err = -ENODEV;
3032 goto errout;
3035 u_map = nla_data(tb[IFLA_MAP]);
3036 k_map.mem_start = (unsigned long) u_map->mem_start;
3037 k_map.mem_end = (unsigned long) u_map->mem_end;
3038 k_map.base_addr = (unsigned short) u_map->base_addr;
3039 k_map.irq = (unsigned char) u_map->irq;
3040 k_map.dma = (unsigned char) u_map->dma;
3041 k_map.port = (unsigned char) u_map->port;
3043 err = ops->ndo_set_config(dev, &k_map);
3044 if (err < 0)
3045 goto errout;
3047 status |= DO_SETLINK_NOTIFY;
3050 if (tb[IFLA_ADDRESS]) {
3051 struct sockaddr *sa;
3052 int len;
3054 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
3055 sizeof(*sa));
3056 sa = kmalloc(len, GFP_KERNEL);
3057 if (!sa) {
3058 err = -ENOMEM;
3059 goto errout;
3061 sa->sa_family = dev->type;
3062 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
3063 dev->addr_len);
3064 err = dev_set_mac_address_user(dev, sa, extack);
3065 kfree(sa);
3066 if (err)
3067 goto errout;
3068 status |= DO_SETLINK_MODIFIED;
3071 if (tb[IFLA_MTU]) {
3072 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
3073 if (err < 0)
3074 goto errout;
3075 status |= DO_SETLINK_MODIFIED;
3078 if (tb[IFLA_GROUP]) {
3079 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3080 status |= DO_SETLINK_NOTIFY;
3084 * Interface selected by interface index but interface
3085 * name provided implies that a name change has been
3086 * requested.
3088 if (ifm->ifi_index > 0 && ifname[0]) {
3089 err = dev_change_name(dev, ifname);
3090 if (err < 0)
3091 goto errout;
3092 status |= DO_SETLINK_MODIFIED;
3095 if (tb[IFLA_IFALIAS]) {
3096 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
3097 nla_len(tb[IFLA_IFALIAS]));
3098 if (err < 0)
3099 goto errout;
3100 status |= DO_SETLINK_NOTIFY;
3103 if (tb[IFLA_BROADCAST]) {
3104 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
3105 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
3108 if (ifm->ifi_flags || ifm->ifi_change) {
3109 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3110 extack);
3111 if (err < 0)
3112 goto errout;
3115 if (tb[IFLA_MASTER]) {
3116 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3117 if (err)
3118 goto errout;
3119 status |= DO_SETLINK_MODIFIED;
3122 if (tb[IFLA_CARRIER]) {
3123 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
3124 if (err)
3125 goto errout;
3126 status |= DO_SETLINK_MODIFIED;
3129 if (tb[IFLA_TXQLEN]) {
3130 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
3132 err = dev_change_tx_queue_len(dev, value);
3133 if (err)
3134 goto errout;
3135 status |= DO_SETLINK_MODIFIED;
3138 if (tb[IFLA_GSO_MAX_SIZE]) {
3139 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
3141 if (dev->gso_max_size ^ max_size) {
3142 netif_set_gso_max_size(dev, max_size);
3143 status |= DO_SETLINK_MODIFIED;
3147 if (tb[IFLA_GSO_MAX_SEGS]) {
3148 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3150 if (dev->gso_max_segs ^ max_segs) {
3151 netif_set_gso_max_segs(dev, max_segs);
3152 status |= DO_SETLINK_MODIFIED;
3156 if (tb[IFLA_GRO_MAX_SIZE]) {
3157 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
3159 if (dev->gro_max_size ^ gro_max_size) {
3160 netif_set_gro_max_size(dev, gro_max_size);
3161 status |= DO_SETLINK_MODIFIED;
3165 if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
3166 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
3168 if (dev->gso_ipv4_max_size ^ max_size) {
3169 netif_set_gso_ipv4_max_size(dev, max_size);
3170 status |= DO_SETLINK_MODIFIED;
3174 if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
3175 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
3177 if (dev->gro_ipv4_max_size ^ gro_max_size) {
3178 netif_set_gro_ipv4_max_size(dev, gro_max_size);
3179 status |= DO_SETLINK_MODIFIED;
3183 if (tb[IFLA_OPERSTATE])
3184 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3186 if (tb[IFLA_LINKMODE]) {
3187 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
3189 if (dev->link_mode ^ value)
3190 status |= DO_SETLINK_NOTIFY;
3191 WRITE_ONCE(dev->link_mode, value);
3194 if (tb[IFLA_VFINFO_LIST]) {
3195 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
3196 struct nlattr *attr;
3197 int rem;
3199 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
3200 if (nla_type(attr) != IFLA_VF_INFO ||
3201 nla_len(attr) < NLA_HDRLEN) {
3202 err = -EINVAL;
3203 goto errout;
3205 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
3206 attr,
3207 ifla_vf_policy,
3208 NULL);
3209 if (err < 0)
3210 goto errout;
3211 err = do_setvfinfo(dev, vfinfo);
3212 if (err < 0)
3213 goto errout;
3214 status |= DO_SETLINK_NOTIFY;
3217 err = 0;
3219 if (tb[IFLA_VF_PORTS]) {
3220 struct nlattr *port[IFLA_PORT_MAX+1];
3221 struct nlattr *attr;
3222 int vf;
3223 int rem;
3225 err = -EOPNOTSUPP;
3226 if (!ops->ndo_set_vf_port)
3227 goto errout;
3229 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
3230 if (nla_type(attr) != IFLA_VF_PORT ||
3231 nla_len(attr) < NLA_HDRLEN) {
3232 err = -EINVAL;
3233 goto errout;
3235 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3236 attr,
3237 ifla_port_policy,
3238 NULL);
3239 if (err < 0)
3240 goto errout;
3241 if (!port[IFLA_PORT_VF]) {
3242 err = -EOPNOTSUPP;
3243 goto errout;
3245 vf = nla_get_u32(port[IFLA_PORT_VF]);
3246 err = ops->ndo_set_vf_port(dev, vf, port);
3247 if (err < 0)
3248 goto errout;
3249 status |= DO_SETLINK_NOTIFY;
3252 err = 0;
3254 if (tb[IFLA_PORT_SELF]) {
3255 struct nlattr *port[IFLA_PORT_MAX+1];
3257 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3258 tb[IFLA_PORT_SELF],
3259 ifla_port_policy, NULL);
3260 if (err < 0)
3261 goto errout;
3263 err = -EOPNOTSUPP;
3264 if (ops->ndo_set_vf_port)
3265 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3266 if (err < 0)
3267 goto errout;
3268 status |= DO_SETLINK_NOTIFY;
3271 if (tb[IFLA_AF_SPEC]) {
3272 struct nlattr *af;
3273 int rem;
3275 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3276 struct rtnl_af_ops *af_ops;
3277 int af_ops_srcu_index;
3279 af_ops = rtnl_af_lookup(nla_type(af), &af_ops_srcu_index);
3280 if (!af_ops) {
3281 err = -EAFNOSUPPORT;
3282 goto errout;
3285 err = af_ops->set_link_af(dev, af, extack);
3286 rtnl_af_put(af_ops, af_ops_srcu_index);
3288 if (err < 0)
3289 goto errout;
3291 status |= DO_SETLINK_NOTIFY;
3294 err = 0;
3296 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3297 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3298 tb[IFLA_PROTO_DOWN_REASON], extack);
3299 if (err)
3300 goto errout;
3301 status |= DO_SETLINK_NOTIFY;
3304 if (tb[IFLA_XDP]) {
3305 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3306 u32 xdp_flags = 0;
3308 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3309 tb[IFLA_XDP],
3310 ifla_xdp_policy, NULL);
3311 if (err < 0)
3312 goto errout;
3314 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3315 err = -EINVAL;
3316 goto errout;
3319 if (xdp[IFLA_XDP_FLAGS]) {
3320 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3321 if (xdp_flags & ~XDP_FLAGS_MASK) {
3322 err = -EINVAL;
3323 goto errout;
3325 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3326 err = -EINVAL;
3327 goto errout;
3331 if (xdp[IFLA_XDP_FD]) {
3332 int expected_fd = -1;
3334 if (xdp_flags & XDP_FLAGS_REPLACE) {
3335 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3336 err = -EINVAL;
3337 goto errout;
3339 expected_fd =
3340 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3343 err = dev_change_xdp_fd(dev, extack,
3344 nla_get_s32(xdp[IFLA_XDP_FD]),
3345 expected_fd,
3346 xdp_flags);
3347 if (err)
3348 goto errout;
3349 status |= DO_SETLINK_NOTIFY;
3353 errout:
3354 if (status & DO_SETLINK_MODIFIED) {
3355 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3356 netdev_state_change(dev);
3358 if (err < 0)
3359 net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
3360 dev->name);
3363 return err;
3366 static struct net_device *rtnl_dev_get(struct net *net,
3367 struct nlattr *tb[])
3369 char ifname[ALTIFNAMSIZ];
3371 if (tb[IFLA_IFNAME])
3372 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3373 else if (tb[IFLA_ALT_IFNAME])
3374 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3375 else
3376 return NULL;
3378 return __dev_get_by_name(net, ifname);
3381 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3382 struct netlink_ext_ack *extack)
3384 struct ifinfomsg *ifm = nlmsg_data(nlh);
3385 struct net *net = sock_net(skb->sk);
3386 struct nlattr *tb[IFLA_MAX+1];
3387 struct net_device *dev = NULL;
3388 struct rtnl_nets rtnl_nets;
3389 struct net *tgt_net;
3390 int err;
3392 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3393 ifla_policy, extack);
3394 if (err < 0)
3395 goto errout;
3397 err = rtnl_ensure_unique_netns(tb, extack, false);
3398 if (err < 0)
3399 goto errout;
3401 tgt_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3402 if (IS_ERR(tgt_net)) {
3403 err = PTR_ERR(tgt_net);
3404 goto errout;
3407 rtnl_nets_init(&rtnl_nets);
3408 rtnl_nets_add(&rtnl_nets, get_net(net));
3409 rtnl_nets_add(&rtnl_nets, tgt_net);
3411 rtnl_nets_lock(&rtnl_nets);
3413 if (ifm->ifi_index > 0)
3414 dev = __dev_get_by_index(net, ifm->ifi_index);
3415 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3416 dev = rtnl_dev_get(net, tb);
3417 else
3418 err = -EINVAL;
3420 if (dev)
3421 err = do_setlink(skb, dev, tgt_net, ifm, extack, tb, 0);
3422 else if (!err)
3423 err = -ENODEV;
3425 rtnl_nets_unlock(&rtnl_nets);
3426 errout:
3427 return err;
3430 static int rtnl_group_dellink(const struct net *net, int group)
3432 struct net_device *dev, *aux;
3433 LIST_HEAD(list_kill);
3434 bool found = false;
3436 if (!group)
3437 return -EPERM;
3439 for_each_netdev(net, dev) {
3440 if (dev->group == group) {
3441 const struct rtnl_link_ops *ops;
3443 found = true;
3444 ops = dev->rtnl_link_ops;
3445 if (!ops || !ops->dellink)
3446 return -EOPNOTSUPP;
3450 if (!found)
3451 return -ENODEV;
3453 for_each_netdev_safe(net, dev, aux) {
3454 if (dev->group == group) {
3455 const struct rtnl_link_ops *ops;
3457 ops = dev->rtnl_link_ops;
3458 ops->dellink(dev, &list_kill);
3461 unregister_netdevice_many(&list_kill);
3463 return 0;
3466 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3468 const struct rtnl_link_ops *ops;
3469 LIST_HEAD(list_kill);
3471 ops = dev->rtnl_link_ops;
3472 if (!ops || !ops->dellink)
3473 return -EOPNOTSUPP;
3475 ops->dellink(dev, &list_kill);
3476 unregister_netdevice_many_notify(&list_kill, portid, nlh);
3478 return 0;
3480 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3482 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3483 struct netlink_ext_ack *extack)
3485 struct ifinfomsg *ifm = nlmsg_data(nlh);
3486 struct net *net = sock_net(skb->sk);
3487 u32 portid = NETLINK_CB(skb).portid;
3488 struct nlattr *tb[IFLA_MAX+1];
3489 struct net_device *dev = NULL;
3490 struct net *tgt_net = net;
3491 int netnsid = -1;
3492 int err;
3494 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3495 ifla_policy, extack);
3496 if (err < 0)
3497 return err;
3499 err = rtnl_ensure_unique_netns(tb, extack, true);
3500 if (err < 0)
3501 return err;
3503 if (tb[IFLA_TARGET_NETNSID]) {
3504 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3505 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3506 if (IS_ERR(tgt_net))
3507 return PTR_ERR(tgt_net);
3510 rtnl_net_lock(tgt_net);
3512 if (ifm->ifi_index > 0)
3513 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3514 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3515 dev = rtnl_dev_get(tgt_net, tb);
3517 if (dev)
3518 err = rtnl_delete_link(dev, portid, nlh);
3519 else if (ifm->ifi_index > 0 || tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3520 err = -ENODEV;
3521 else if (tb[IFLA_GROUP])
3522 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3523 else
3524 err = -EINVAL;
3526 rtnl_net_unlock(tgt_net);
3528 if (netnsid >= 0)
3529 put_net(tgt_net);
3531 return err;
3534 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3535 u32 portid, const struct nlmsghdr *nlh)
3537 unsigned int old_flags;
3538 int err;
3540 old_flags = dev->flags;
3541 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3542 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3543 NULL);
3544 if (err < 0)
3545 return err;
3548 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3549 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3550 } else {
3551 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3552 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3554 return 0;
3556 EXPORT_SYMBOL(rtnl_configure_link);
3558 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3559 unsigned char name_assign_type,
3560 const struct rtnl_link_ops *ops,
3561 struct nlattr *tb[],
3562 struct netlink_ext_ack *extack)
3564 struct net_device *dev;
3565 unsigned int num_tx_queues = 1;
3566 unsigned int num_rx_queues = 1;
3567 int err;
3569 if (tb[IFLA_NUM_TX_QUEUES])
3570 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3571 else if (ops->get_num_tx_queues)
3572 num_tx_queues = ops->get_num_tx_queues();
3574 if (tb[IFLA_NUM_RX_QUEUES])
3575 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3576 else if (ops->get_num_rx_queues)
3577 num_rx_queues = ops->get_num_rx_queues();
3579 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3580 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3581 return ERR_PTR(-EINVAL);
3584 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3585 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3586 return ERR_PTR(-EINVAL);
3589 if (ops->alloc) {
3590 dev = ops->alloc(tb, ifname, name_assign_type,
3591 num_tx_queues, num_rx_queues);
3592 if (IS_ERR(dev))
3593 return dev;
3594 } else {
3595 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3596 name_assign_type, ops->setup,
3597 num_tx_queues, num_rx_queues);
3600 if (!dev)
3601 return ERR_PTR(-ENOMEM);
3603 err = validate_linkmsg(dev, tb, extack);
3604 if (err < 0) {
3605 free_netdev(dev);
3606 return ERR_PTR(err);
3609 dev_net_set(dev, net);
3610 dev->rtnl_link_ops = ops;
3611 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3613 if (tb[IFLA_MTU]) {
3614 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3616 err = dev_validate_mtu(dev, mtu, extack);
3617 if (err) {
3618 free_netdev(dev);
3619 return ERR_PTR(err);
3621 dev->mtu = mtu;
3623 if (tb[IFLA_ADDRESS]) {
3624 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3625 nla_len(tb[IFLA_ADDRESS]));
3626 dev->addr_assign_type = NET_ADDR_SET;
3628 if (tb[IFLA_BROADCAST])
3629 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3630 nla_len(tb[IFLA_BROADCAST]));
3631 if (tb[IFLA_TXQLEN])
3632 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3633 if (tb[IFLA_OPERSTATE])
3634 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3635 if (tb[IFLA_LINKMODE])
3636 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3637 if (tb[IFLA_GROUP])
3638 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3639 if (tb[IFLA_GSO_MAX_SIZE])
3640 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3641 if (tb[IFLA_GSO_MAX_SEGS])
3642 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3643 if (tb[IFLA_GRO_MAX_SIZE])
3644 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3645 if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3646 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3647 if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3648 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3650 return dev;
3652 EXPORT_SYMBOL(rtnl_create_link);
3654 struct rtnl_newlink_tbs {
3655 struct nlattr *tb[IFLA_MAX + 1];
3656 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3657 struct nlattr *attr[RTNL_MAX_TYPE + 1];
3658 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3661 static int rtnl_changelink(const struct sk_buff *skb, struct nlmsghdr *nlh,
3662 const struct rtnl_link_ops *ops,
3663 struct net_device *dev, struct net *tgt_net,
3664 struct rtnl_newlink_tbs *tbs,
3665 struct nlattr **data,
3666 struct netlink_ext_ack *extack)
3668 struct nlattr ** const linkinfo = tbs->linkinfo;
3669 struct nlattr ** const tb = tbs->tb;
3670 int status = 0;
3671 int err;
3673 if (nlh->nlmsg_flags & NLM_F_EXCL)
3674 return -EEXIST;
3676 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3677 return -EOPNOTSUPP;
3679 if (linkinfo[IFLA_INFO_DATA]) {
3680 if (!ops || ops != dev->rtnl_link_ops || !ops->changelink)
3681 return -EOPNOTSUPP;
3683 err = ops->changelink(dev, tb, data, extack);
3684 if (err < 0)
3685 return err;
3687 status |= DO_SETLINK_NOTIFY;
3690 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3691 const struct rtnl_link_ops *m_ops = NULL;
3692 struct nlattr **slave_data = NULL;
3693 struct net_device *master_dev;
3695 master_dev = netdev_master_upper_dev_get(dev);
3696 if (master_dev)
3697 m_ops = master_dev->rtnl_link_ops;
3699 if (!m_ops || !m_ops->slave_changelink)
3700 return -EOPNOTSUPP;
3702 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3703 return -EINVAL;
3705 if (m_ops->slave_maxtype) {
3706 err = nla_parse_nested_deprecated(tbs->slave_attr,
3707 m_ops->slave_maxtype,
3708 linkinfo[IFLA_INFO_SLAVE_DATA],
3709 m_ops->slave_policy, extack);
3710 if (err < 0)
3711 return err;
3713 slave_data = tbs->slave_attr;
3716 err = m_ops->slave_changelink(master_dev, dev, tb, slave_data, extack);
3717 if (err < 0)
3718 return err;
3720 status |= DO_SETLINK_NOTIFY;
3723 return do_setlink(skb, dev, tgt_net, nlmsg_data(nlh), extack, tb, status);
3726 static int rtnl_group_changelink(const struct sk_buff *skb,
3727 struct net *net, struct net *tgt_net,
3728 int group, struct ifinfomsg *ifm,
3729 struct netlink_ext_ack *extack,
3730 struct nlattr **tb)
3732 struct net_device *dev, *aux;
3733 int err;
3735 for_each_netdev_safe(net, dev, aux) {
3736 if (dev->group == group) {
3737 err = do_setlink(skb, dev, tgt_net, ifm, extack, tb, 0);
3738 if (err < 0)
3739 return err;
3743 return 0;
3746 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3747 const struct rtnl_link_ops *ops,
3748 struct net *tgt_net, struct net *link_net,
3749 struct net *peer_net,
3750 const struct nlmsghdr *nlh,
3751 struct nlattr **tb, struct nlattr **data,
3752 struct netlink_ext_ack *extack)
3754 unsigned char name_assign_type = NET_NAME_USER;
3755 struct net *net = sock_net(skb->sk);
3756 u32 portid = NETLINK_CB(skb).portid;
3757 struct net_device *dev;
3758 char ifname[IFNAMSIZ];
3759 int err;
3761 if (!ops->alloc && !ops->setup)
3762 return -EOPNOTSUPP;
3764 if (tb[IFLA_IFNAME]) {
3765 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3766 } else {
3767 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3768 name_assign_type = NET_NAME_ENUM;
3771 dev = rtnl_create_link(link_net ? : tgt_net, ifname,
3772 name_assign_type, ops, tb, extack);
3773 if (IS_ERR(dev)) {
3774 err = PTR_ERR(dev);
3775 goto out;
3778 dev->ifindex = ifm->ifi_index;
3780 if (link_net)
3781 net = link_net;
3782 if (peer_net)
3783 net = peer_net;
3785 if (ops->newlink)
3786 err = ops->newlink(net, dev, tb, data, extack);
3787 else
3788 err = register_netdevice(dev);
3789 if (err < 0) {
3790 free_netdev(dev);
3791 goto out;
3794 err = rtnl_configure_link(dev, ifm, portid, nlh);
3795 if (err < 0)
3796 goto out_unregister;
3797 if (link_net) {
3798 err = dev_change_net_namespace(dev, tgt_net, ifname);
3799 if (err < 0)
3800 goto out_unregister;
3802 if (tb[IFLA_MASTER]) {
3803 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3804 if (err)
3805 goto out_unregister;
3807 out:
3808 return err;
3809 out_unregister:
3810 if (ops->newlink) {
3811 LIST_HEAD(list_kill);
3813 ops->dellink(dev, &list_kill);
3814 unregister_netdevice_many(&list_kill);
3815 } else {
3816 unregister_netdevice(dev);
3818 goto out;
3821 static struct net *rtnl_get_peer_net(const struct rtnl_link_ops *ops,
3822 struct nlattr *data[],
3823 struct netlink_ext_ack *extack)
3825 struct nlattr *tb[IFLA_MAX + 1];
3826 int err;
3828 if (!data || !data[ops->peer_type])
3829 return NULL;
3831 err = rtnl_nla_parse_ifinfomsg(tb, data[ops->peer_type], extack);
3832 if (err < 0)
3833 return ERR_PTR(err);
3835 if (ops->validate) {
3836 err = ops->validate(tb, NULL, extack);
3837 if (err < 0)
3838 return ERR_PTR(err);
3841 return rtnl_link_get_net_ifla(tb);
3844 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3845 const struct rtnl_link_ops *ops,
3846 struct net *tgt_net, struct net *link_net,
3847 struct net *peer_net,
3848 struct rtnl_newlink_tbs *tbs,
3849 struct nlattr **data,
3850 struct netlink_ext_ack *extack)
3852 struct nlattr ** const tb = tbs->tb;
3853 struct net *net = sock_net(skb->sk);
3854 struct net_device *dev;
3855 struct ifinfomsg *ifm;
3856 bool link_specified;
3858 ifm = nlmsg_data(nlh);
3859 if (ifm->ifi_index > 0) {
3860 link_specified = true;
3861 dev = __dev_get_by_index(net, ifm->ifi_index);
3862 } else if (ifm->ifi_index < 0) {
3863 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3864 return -EINVAL;
3865 } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3866 link_specified = true;
3867 dev = rtnl_dev_get(net, tb);
3868 } else {
3869 link_specified = false;
3870 dev = NULL;
3873 if (dev)
3874 return rtnl_changelink(skb, nlh, ops, dev, tgt_net, tbs, data, extack);
3876 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3877 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3878 * or it's for a group
3880 if (link_specified || !tb[IFLA_GROUP])
3881 return -ENODEV;
3883 return rtnl_group_changelink(skb, net, tgt_net,
3884 nla_get_u32(tb[IFLA_GROUP]),
3885 ifm, extack, tb);
3888 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3889 return -EOPNOTSUPP;
3891 if (!ops) {
3892 NL_SET_ERR_MSG(extack, "Unknown device type");
3893 return -EOPNOTSUPP;
3896 return rtnl_newlink_create(skb, ifm, ops, tgt_net, link_net, peer_net, nlh,
3897 tb, data, extack);
3900 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3901 struct netlink_ext_ack *extack)
3903 struct net *tgt_net, *link_net = NULL, *peer_net = NULL;
3904 struct nlattr **tb, **linkinfo, **data = NULL;
3905 struct rtnl_link_ops *ops = NULL;
3906 struct rtnl_newlink_tbs *tbs;
3907 struct rtnl_nets rtnl_nets;
3908 int ops_srcu_index;
3909 int ret;
3911 tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3912 if (!tbs)
3913 return -ENOMEM;
3915 tb = tbs->tb;
3916 ret = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg), tb,
3917 IFLA_MAX, ifla_policy, extack);
3918 if (ret < 0)
3919 goto free;
3921 ret = rtnl_ensure_unique_netns(tb, extack, false);
3922 if (ret < 0)
3923 goto free;
3925 linkinfo = tbs->linkinfo;
3926 if (tb[IFLA_LINKINFO]) {
3927 ret = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3928 tb[IFLA_LINKINFO],
3929 ifla_info_policy, NULL);
3930 if (ret < 0)
3931 goto free;
3932 } else {
3933 memset(linkinfo, 0, sizeof(tbs->linkinfo));
3936 if (linkinfo[IFLA_INFO_KIND]) {
3937 char kind[MODULE_NAME_LEN];
3939 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3940 ops = rtnl_link_ops_get(kind, &ops_srcu_index);
3941 #ifdef CONFIG_MODULES
3942 if (!ops) {
3943 request_module("rtnl-link-%s", kind);
3944 ops = rtnl_link_ops_get(kind, &ops_srcu_index);
3946 #endif
3949 rtnl_nets_init(&rtnl_nets);
3951 if (ops) {
3952 if (ops->maxtype > RTNL_MAX_TYPE) {
3953 ret = -EINVAL;
3954 goto put_ops;
3957 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3958 ret = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3959 linkinfo[IFLA_INFO_DATA],
3960 ops->policy, extack);
3961 if (ret < 0)
3962 goto put_ops;
3964 data = tbs->attr;
3967 if (ops->validate) {
3968 ret = ops->validate(tb, data, extack);
3969 if (ret < 0)
3970 goto put_ops;
3973 if (ops->peer_type) {
3974 peer_net = rtnl_get_peer_net(ops, data, extack);
3975 if (IS_ERR(peer_net))
3976 goto put_ops;
3977 if (peer_net)
3978 rtnl_nets_add(&rtnl_nets, peer_net);
3982 tgt_net = rtnl_link_get_net_capable(skb, sock_net(skb->sk), tb, CAP_NET_ADMIN);
3983 if (IS_ERR(tgt_net)) {
3984 ret = PTR_ERR(tgt_net);
3985 goto put_net;
3988 rtnl_nets_add(&rtnl_nets, tgt_net);
3990 if (tb[IFLA_LINK_NETNSID]) {
3991 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3993 link_net = get_net_ns_by_id(tgt_net, id);
3994 if (!link_net) {
3995 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3996 ret = -EINVAL;
3997 goto put_net;
4000 rtnl_nets_add(&rtnl_nets, link_net);
4002 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) {
4003 ret = -EPERM;
4004 goto put_net;
4008 rtnl_nets_lock(&rtnl_nets);
4009 ret = __rtnl_newlink(skb, nlh, ops, tgt_net, link_net, peer_net, tbs, data, extack);
4010 rtnl_nets_unlock(&rtnl_nets);
4012 put_net:
4013 rtnl_nets_destroy(&rtnl_nets);
4014 put_ops:
4015 if (ops)
4016 rtnl_link_ops_put(ops, ops_srcu_index);
4017 free:
4018 kfree(tbs);
4019 return ret;
4022 static int rtnl_valid_getlink_req(struct sk_buff *skb,
4023 const struct nlmsghdr *nlh,
4024 struct nlattr **tb,
4025 struct netlink_ext_ack *extack)
4027 struct ifinfomsg *ifm;
4028 int i, err;
4030 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4031 NL_SET_ERR_MSG(extack, "Invalid header for get link");
4032 return -EINVAL;
4035 if (!netlink_strict_get_check(skb))
4036 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
4037 ifla_policy, extack);
4039 ifm = nlmsg_data(nlh);
4040 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4041 ifm->ifi_change) {
4042 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
4043 return -EINVAL;
4046 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
4047 ifla_policy, extack);
4048 if (err)
4049 return err;
4051 for (i = 0; i <= IFLA_MAX; i++) {
4052 if (!tb[i])
4053 continue;
4055 switch (i) {
4056 case IFLA_IFNAME:
4057 case IFLA_ALT_IFNAME:
4058 case IFLA_EXT_MASK:
4059 case IFLA_TARGET_NETNSID:
4060 break;
4061 default:
4062 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
4063 return -EINVAL;
4067 return 0;
4070 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4071 struct netlink_ext_ack *extack)
4073 struct net *net = sock_net(skb->sk);
4074 struct net *tgt_net = net;
4075 struct ifinfomsg *ifm;
4076 struct nlattr *tb[IFLA_MAX+1];
4077 struct net_device *dev = NULL;
4078 struct sk_buff *nskb;
4079 int netnsid = -1;
4080 int err;
4081 u32 ext_filter_mask = 0;
4083 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
4084 if (err < 0)
4085 return err;
4087 err = rtnl_ensure_unique_netns(tb, extack, true);
4088 if (err < 0)
4089 return err;
4091 if (tb[IFLA_TARGET_NETNSID]) {
4092 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
4093 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
4094 if (IS_ERR(tgt_net))
4095 return PTR_ERR(tgt_net);
4098 if (tb[IFLA_EXT_MASK])
4099 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
4101 err = -EINVAL;
4102 ifm = nlmsg_data(nlh);
4103 if (ifm->ifi_index > 0)
4104 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
4105 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
4106 dev = rtnl_dev_get(tgt_net, tb);
4107 else
4108 goto out;
4110 err = -ENODEV;
4111 if (dev == NULL)
4112 goto out;
4114 err = -ENOBUFS;
4115 nskb = nlmsg_new_large(if_nlmsg_size(dev, ext_filter_mask));
4116 if (nskb == NULL)
4117 goto out;
4119 /* Synchronize the carrier state so we don't report a state
4120 * that we're not actually going to honour immediately; if
4121 * the driver just did a carrier off->on transition, we can
4122 * only TX if link watch work has run, but without this we'd
4123 * already report carrier on, even if it doesn't work yet.
4125 linkwatch_sync_dev(dev);
4127 err = rtnl_fill_ifinfo(nskb, dev, net,
4128 RTM_NEWLINK, NETLINK_CB(skb).portid,
4129 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
4130 0, NULL, 0, netnsid, GFP_KERNEL);
4131 if (err < 0) {
4132 /* -EMSGSIZE implies BUG in if_nlmsg_size */
4133 WARN_ON(err == -EMSGSIZE);
4134 kfree_skb(nskb);
4135 } else
4136 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4137 out:
4138 if (netnsid >= 0)
4139 put_net(tgt_net);
4141 return err;
4144 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
4145 bool *changed, struct netlink_ext_ack *extack)
4147 char *alt_ifname;
4148 size_t size;
4149 int err;
4151 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
4152 if (err)
4153 return err;
4155 if (cmd == RTM_NEWLINKPROP) {
4156 size = rtnl_prop_list_size(dev);
4157 size += nla_total_size(ALTIFNAMSIZ);
4158 if (size >= U16_MAX) {
4159 NL_SET_ERR_MSG(extack,
4160 "effective property list too long");
4161 return -EINVAL;
4165 alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
4166 if (!alt_ifname)
4167 return -ENOMEM;
4169 if (cmd == RTM_NEWLINKPROP) {
4170 err = netdev_name_node_alt_create(dev, alt_ifname);
4171 if (!err)
4172 alt_ifname = NULL;
4173 } else if (cmd == RTM_DELLINKPROP) {
4174 err = netdev_name_node_alt_destroy(dev, alt_ifname);
4175 } else {
4176 WARN_ON_ONCE(1);
4177 err = -EINVAL;
4180 kfree(alt_ifname);
4181 if (!err)
4182 *changed = true;
4183 return err;
4186 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
4187 struct netlink_ext_ack *extack)
4189 struct net *net = sock_net(skb->sk);
4190 struct nlattr *tb[IFLA_MAX + 1];
4191 struct net_device *dev;
4192 struct ifinfomsg *ifm;
4193 bool changed = false;
4194 struct nlattr *attr;
4195 int err, rem;
4197 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
4198 if (err)
4199 return err;
4201 err = rtnl_ensure_unique_netns(tb, extack, true);
4202 if (err)
4203 return err;
4205 ifm = nlmsg_data(nlh);
4206 if (ifm->ifi_index > 0)
4207 dev = __dev_get_by_index(net, ifm->ifi_index);
4208 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
4209 dev = rtnl_dev_get(net, tb);
4210 else
4211 return -EINVAL;
4213 if (!dev)
4214 return -ENODEV;
4216 if (!tb[IFLA_PROP_LIST])
4217 return 0;
4219 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
4220 switch (nla_type(attr)) {
4221 case IFLA_ALT_IFNAME:
4222 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
4223 if (err)
4224 return err;
4225 break;
4229 if (changed)
4230 netdev_state_change(dev);
4231 return 0;
4234 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
4235 struct netlink_ext_ack *extack)
4237 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
4240 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
4241 struct netlink_ext_ack *extack)
4243 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
4246 static noinline_for_stack u32 rtnl_calcit(struct sk_buff *skb,
4247 struct nlmsghdr *nlh)
4249 struct net *net = sock_net(skb->sk);
4250 size_t min_ifinfo_dump_size = 0;
4251 u32 ext_filter_mask = 0;
4252 struct net_device *dev;
4253 struct nlattr *nla;
4254 int hdrlen, rem;
4256 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
4257 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
4258 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
4260 if (nlh->nlmsg_len < nlmsg_msg_size(hdrlen))
4261 return NLMSG_GOODSIZE;
4263 nla_for_each_attr_type(nla, IFLA_EXT_MASK,
4264 nlmsg_attrdata(nlh, hdrlen),
4265 nlmsg_attrlen(nlh, hdrlen), rem) {
4266 if (nla_len(nla) == sizeof(u32))
4267 ext_filter_mask = nla_get_u32(nla);
4270 if (!ext_filter_mask)
4271 return NLMSG_GOODSIZE;
4273 * traverse the list of net devices and compute the minimum
4274 * buffer size based upon the filter mask.
4276 rcu_read_lock();
4277 for_each_netdev_rcu(net, dev) {
4278 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
4279 if_nlmsg_size(dev, ext_filter_mask));
4281 rcu_read_unlock();
4283 return nlmsg_total_size(min_ifinfo_dump_size);
4286 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
4288 int idx;
4289 int s_idx = cb->family;
4290 int type = cb->nlh->nlmsg_type - RTM_BASE;
4291 int ret = 0;
4293 if (s_idx == 0)
4294 s_idx = 1;
4296 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
4297 struct rtnl_link __rcu **tab;
4298 struct rtnl_link *link;
4299 rtnl_dumpit_func dumpit;
4301 if (idx < s_idx || idx == PF_PACKET)
4302 continue;
4304 if (type < 0 || type >= RTM_NR_MSGTYPES)
4305 continue;
4307 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
4308 if (!tab)
4309 continue;
4311 link = rcu_dereference_rtnl(tab[type]);
4312 if (!link)
4313 continue;
4315 dumpit = link->dumpit;
4316 if (!dumpit)
4317 continue;
4319 if (idx > s_idx) {
4320 memset(&cb->args[0], 0, sizeof(cb->args));
4321 cb->prev_seq = 0;
4322 cb->seq = 0;
4324 ret = dumpit(skb, cb);
4325 if (ret)
4326 break;
4328 cb->family = idx;
4330 return skb->len ? : ret;
4333 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
4334 unsigned int change,
4335 u32 event, gfp_t flags, int *new_nsid,
4336 int new_ifindex, u32 portid,
4337 const struct nlmsghdr *nlh)
4339 struct net *net = dev_net(dev);
4340 struct sk_buff *skb;
4341 int err = -ENOBUFS;
4342 u32 seq = 0;
4344 skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4345 if (skb == NULL)
4346 goto errout;
4348 if (nlmsg_report(nlh))
4349 seq = nlmsg_seq(nlh);
4350 else
4351 portid = 0;
4353 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4354 type, portid, seq, change, 0, 0, event,
4355 new_nsid, new_ifindex, -1, flags);
4356 if (err < 0) {
4357 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4358 WARN_ON(err == -EMSGSIZE);
4359 kfree_skb(skb);
4360 goto errout;
4362 return skb;
4363 errout:
4364 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4365 return NULL;
4368 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4369 u32 portid, const struct nlmsghdr *nlh)
4371 struct net *net = dev_net(dev);
4373 rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4376 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4377 unsigned int change, u32 event,
4378 gfp_t flags, int *new_nsid, int new_ifindex,
4379 u32 portid, const struct nlmsghdr *nlh)
4381 struct sk_buff *skb;
4383 if (dev->reg_state != NETREG_REGISTERED)
4384 return;
4386 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4387 new_ifindex, portid, nlh);
4388 if (skb)
4389 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4392 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4393 gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4395 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4396 NULL, 0, portid, nlh);
4399 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4400 gfp_t flags, int *new_nsid, int new_ifindex)
4402 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4403 new_nsid, new_ifindex, 0, NULL);
4406 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4407 struct net_device *dev,
4408 u8 *addr, u16 vid, u32 pid, u32 seq,
4409 int type, unsigned int flags,
4410 int nlflags, u16 ndm_state)
4412 struct nlmsghdr *nlh;
4413 struct ndmsg *ndm;
4415 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4416 if (!nlh)
4417 return -EMSGSIZE;
4419 ndm = nlmsg_data(nlh);
4420 ndm->ndm_family = AF_BRIDGE;
4421 ndm->ndm_pad1 = 0;
4422 ndm->ndm_pad2 = 0;
4423 ndm->ndm_flags = flags;
4424 ndm->ndm_type = 0;
4425 ndm->ndm_ifindex = dev->ifindex;
4426 ndm->ndm_state = ndm_state;
4428 if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
4429 goto nla_put_failure;
4430 if (vid)
4431 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4432 goto nla_put_failure;
4434 nlmsg_end(skb, nlh);
4435 return 0;
4437 nla_put_failure:
4438 nlmsg_cancel(skb, nlh);
4439 return -EMSGSIZE;
4442 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
4444 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4445 nla_total_size(dev->addr_len) + /* NDA_LLADDR */
4446 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
4450 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4451 u16 ndm_state)
4453 struct net *net = dev_net(dev);
4454 struct sk_buff *skb;
4455 int err = -ENOBUFS;
4457 skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
4458 if (!skb)
4459 goto errout;
4461 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4462 0, 0, type, NTF_SELF, 0, ndm_state);
4463 if (err < 0) {
4464 kfree_skb(skb);
4465 goto errout;
4468 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4469 return;
4470 errout:
4471 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4475 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4477 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4478 struct nlattr *tb[],
4479 struct net_device *dev,
4480 const unsigned char *addr, u16 vid,
4481 u16 flags)
4483 int err = -EINVAL;
4485 /* If aging addresses are supported device will need to
4486 * implement its own handler for this.
4488 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4489 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4490 return err;
4493 if (tb[NDA_FLAGS_EXT]) {
4494 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4495 return err;
4498 if (vid) {
4499 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4500 return err;
4503 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4504 err = dev_uc_add_excl(dev, addr);
4505 else if (is_multicast_ether_addr(addr))
4506 err = dev_mc_add_excl(dev, addr);
4508 /* Only return duplicate errors if NLM_F_EXCL is set */
4509 if (err == -EEXIST && !(flags & NLM_F_EXCL))
4510 err = 0;
4512 return err;
4514 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4516 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4517 struct netlink_ext_ack *extack)
4519 u16 vid = 0;
4521 if (vlan_attr) {
4522 if (nla_len(vlan_attr) != sizeof(u16)) {
4523 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4524 return -EINVAL;
4527 vid = nla_get_u16(vlan_attr);
4529 if (!vid || vid >= VLAN_VID_MASK) {
4530 NL_SET_ERR_MSG(extack, "invalid vlan id");
4531 return -EINVAL;
4534 *p_vid = vid;
4535 return 0;
4538 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4539 struct netlink_ext_ack *extack)
4541 struct net *net = sock_net(skb->sk);
4542 struct ndmsg *ndm;
4543 struct nlattr *tb[NDA_MAX+1];
4544 struct net_device *dev;
4545 u8 *addr;
4546 u16 vid;
4547 int err;
4549 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4550 extack);
4551 if (err < 0)
4552 return err;
4554 ndm = nlmsg_data(nlh);
4555 if (ndm->ndm_ifindex == 0) {
4556 NL_SET_ERR_MSG(extack, "invalid ifindex");
4557 return -EINVAL;
4560 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4561 if (dev == NULL) {
4562 NL_SET_ERR_MSG(extack, "unknown ifindex");
4563 return -ENODEV;
4566 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4567 NL_SET_ERR_MSG(extack, "invalid address");
4568 return -EINVAL;
4571 if (dev->type != ARPHRD_ETHER) {
4572 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4573 return -EINVAL;
4576 addr = nla_data(tb[NDA_LLADDR]);
4578 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4579 if (err)
4580 return err;
4582 err = -EOPNOTSUPP;
4584 /* Support fdb on master device the net/bridge default case */
4585 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4586 netif_is_bridge_port(dev)) {
4587 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4588 const struct net_device_ops *ops = br_dev->netdev_ops;
4589 bool notified = false;
4591 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4592 nlh->nlmsg_flags, &notified, extack);
4593 if (err)
4594 goto out;
4595 else
4596 ndm->ndm_flags &= ~NTF_MASTER;
4599 /* Embedded bridge, macvlan, and any other device support */
4600 if ((ndm->ndm_flags & NTF_SELF)) {
4601 bool notified = false;
4603 if (dev->netdev_ops->ndo_fdb_add)
4604 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4605 vid,
4606 nlh->nlmsg_flags,
4607 &notified, extack);
4608 else
4609 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4610 nlh->nlmsg_flags);
4612 if (!err && !notified) {
4613 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4614 ndm->ndm_state);
4615 ndm->ndm_flags &= ~NTF_SELF;
4618 out:
4619 return err;
4623 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4625 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4626 struct nlattr *tb[],
4627 struct net_device *dev,
4628 const unsigned char *addr, u16 vid)
4630 int err = -EINVAL;
4632 /* If aging addresses are supported device will need to
4633 * implement its own handler for this.
4635 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4636 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4637 return err;
4640 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4641 err = dev_uc_del(dev, addr);
4642 else if (is_multicast_ether_addr(addr))
4643 err = dev_mc_del(dev, addr);
4645 return err;
4647 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4649 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4650 struct netlink_ext_ack *extack)
4652 bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4653 struct net *net = sock_net(skb->sk);
4654 const struct net_device_ops *ops;
4655 struct ndmsg *ndm;
4656 struct nlattr *tb[NDA_MAX+1];
4657 struct net_device *dev;
4658 __u8 *addr = NULL;
4659 int err;
4660 u16 vid;
4662 if (!netlink_capable(skb, CAP_NET_ADMIN))
4663 return -EPERM;
4665 if (!del_bulk) {
4666 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4667 NULL, extack);
4668 } else {
4669 /* For bulk delete, the drivers will parse the message with
4670 * policy.
4672 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
4674 if (err < 0)
4675 return err;
4677 ndm = nlmsg_data(nlh);
4678 if (ndm->ndm_ifindex == 0) {
4679 NL_SET_ERR_MSG(extack, "invalid ifindex");
4680 return -EINVAL;
4683 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4684 if (dev == NULL) {
4685 NL_SET_ERR_MSG(extack, "unknown ifindex");
4686 return -ENODEV;
4689 if (!del_bulk) {
4690 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4691 NL_SET_ERR_MSG(extack, "invalid address");
4692 return -EINVAL;
4694 addr = nla_data(tb[NDA_LLADDR]);
4696 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4697 if (err)
4698 return err;
4701 if (dev->type != ARPHRD_ETHER) {
4702 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4703 return -EINVAL;
4706 err = -EOPNOTSUPP;
4708 /* Support fdb on master device the net/bridge default case */
4709 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4710 netif_is_bridge_port(dev)) {
4711 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4712 bool notified = false;
4714 ops = br_dev->netdev_ops;
4715 if (!del_bulk) {
4716 if (ops->ndo_fdb_del)
4717 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid,
4718 &notified, extack);
4719 } else {
4720 if (ops->ndo_fdb_del_bulk)
4721 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4724 if (err)
4725 goto out;
4726 else
4727 ndm->ndm_flags &= ~NTF_MASTER;
4730 /* Embedded bridge, macvlan, and any other device support */
4731 if (ndm->ndm_flags & NTF_SELF) {
4732 bool notified = false;
4734 ops = dev->netdev_ops;
4735 if (!del_bulk) {
4736 if (ops->ndo_fdb_del)
4737 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid,
4738 &notified, extack);
4739 else
4740 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4741 } else {
4742 /* in case err was cleared by NTF_MASTER call */
4743 err = -EOPNOTSUPP;
4744 if (ops->ndo_fdb_del_bulk)
4745 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4748 if (!err) {
4749 if (!del_bulk && !notified)
4750 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4751 ndm->ndm_state);
4752 ndm->ndm_flags &= ~NTF_SELF;
4755 out:
4756 return err;
4759 static int nlmsg_populate_fdb(struct sk_buff *skb,
4760 struct netlink_callback *cb,
4761 struct net_device *dev,
4762 int *idx,
4763 struct netdev_hw_addr_list *list)
4765 struct netdev_hw_addr *ha;
4766 int err;
4767 u32 portid, seq;
4769 portid = NETLINK_CB(cb->skb).portid;
4770 seq = cb->nlh->nlmsg_seq;
4772 list_for_each_entry(ha, &list->list, list) {
4773 if (*idx < cb->args[2])
4774 goto skip;
4776 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4777 portid, seq,
4778 RTM_NEWNEIGH, NTF_SELF,
4779 NLM_F_MULTI, NUD_PERMANENT);
4780 if (err < 0)
4781 return err;
4782 skip:
4783 *idx += 1;
4785 return 0;
4789 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4790 * @skb: socket buffer to store message in
4791 * @cb: netlink callback
4792 * @dev: netdevice
4793 * @filter_dev: ignored
4794 * @idx: the number of FDB table entries dumped is added to *@idx
4796 * Default netdevice operation to dump the existing unicast address list.
4797 * Returns number of addresses from list put in skb.
4799 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4800 struct netlink_callback *cb,
4801 struct net_device *dev,
4802 struct net_device *filter_dev,
4803 int *idx)
4805 int err;
4807 if (dev->type != ARPHRD_ETHER)
4808 return -EINVAL;
4810 netif_addr_lock_bh(dev);
4811 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4812 if (err)
4813 goto out;
4814 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4815 out:
4816 netif_addr_unlock_bh(dev);
4817 return err;
4819 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4821 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4822 int *br_idx, int *brport_idx,
4823 struct netlink_ext_ack *extack)
4825 struct nlattr *tb[NDA_MAX + 1];
4826 struct ndmsg *ndm;
4827 int err, i;
4829 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4830 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4831 return -EINVAL;
4834 ndm = nlmsg_data(nlh);
4835 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4836 ndm->ndm_flags || ndm->ndm_type) {
4837 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4838 return -EINVAL;
4841 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4842 NDA_MAX, NULL, extack);
4843 if (err < 0)
4844 return err;
4846 *brport_idx = ndm->ndm_ifindex;
4847 for (i = 0; i <= NDA_MAX; ++i) {
4848 if (!tb[i])
4849 continue;
4851 switch (i) {
4852 case NDA_IFINDEX:
4853 if (nla_len(tb[i]) != sizeof(u32)) {
4854 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4855 return -EINVAL;
4857 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4858 break;
4859 case NDA_MASTER:
4860 if (nla_len(tb[i]) != sizeof(u32)) {
4861 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4862 return -EINVAL;
4864 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4865 break;
4866 default:
4867 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4868 return -EINVAL;
4872 return 0;
4875 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4876 int *br_idx, int *brport_idx,
4877 struct netlink_ext_ack *extack)
4879 struct nlattr *tb[IFLA_MAX+1];
4880 int err;
4882 /* A hack to preserve kernel<->userspace interface.
4883 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4884 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4885 * So, check for ndmsg with an optional u32 attribute (not used here).
4886 * Fortunately these sizes don't conflict with the size of ifinfomsg
4887 * with an optional attribute.
4889 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4890 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4891 nla_attr_size(sizeof(u32)))) {
4892 struct ifinfomsg *ifm;
4894 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4895 tb, IFLA_MAX, ifla_policy,
4896 extack);
4897 if (err < 0) {
4898 return -EINVAL;
4899 } else if (err == 0) {
4900 if (tb[IFLA_MASTER])
4901 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4904 ifm = nlmsg_data(nlh);
4905 *brport_idx = ifm->ifi_index;
4907 return 0;
4910 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4912 struct net_device *dev;
4913 struct net_device *br_dev = NULL;
4914 const struct net_device_ops *ops = NULL;
4915 const struct net_device_ops *cops = NULL;
4916 struct net *net = sock_net(skb->sk);
4917 struct hlist_head *head;
4918 int brport_idx = 0;
4919 int br_idx = 0;
4920 int h, s_h;
4921 int idx = 0, s_idx;
4922 int err = 0;
4923 int fidx = 0;
4925 if (cb->strict_check)
4926 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4927 cb->extack);
4928 else
4929 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4930 cb->extack);
4931 if (err < 0)
4932 return err;
4934 if (br_idx) {
4935 br_dev = __dev_get_by_index(net, br_idx);
4936 if (!br_dev)
4937 return -ENODEV;
4939 ops = br_dev->netdev_ops;
4942 s_h = cb->args[0];
4943 s_idx = cb->args[1];
4945 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4946 idx = 0;
4947 head = &net->dev_index_head[h];
4948 hlist_for_each_entry(dev, head, index_hlist) {
4950 if (brport_idx && (dev->ifindex != brport_idx))
4951 continue;
4953 if (!br_idx) { /* user did not specify a specific bridge */
4954 if (netif_is_bridge_port(dev)) {
4955 br_dev = netdev_master_upper_dev_get(dev);
4956 cops = br_dev->netdev_ops;
4958 } else {
4959 if (dev != br_dev &&
4960 !netif_is_bridge_port(dev))
4961 continue;
4963 if (br_dev != netdev_master_upper_dev_get(dev) &&
4964 !netif_is_bridge_master(dev))
4965 continue;
4966 cops = ops;
4969 if (idx < s_idx)
4970 goto cont;
4972 if (netif_is_bridge_port(dev)) {
4973 if (cops && cops->ndo_fdb_dump) {
4974 err = cops->ndo_fdb_dump(skb, cb,
4975 br_dev, dev,
4976 &fidx);
4977 if (err == -EMSGSIZE)
4978 goto out;
4982 if (dev->netdev_ops->ndo_fdb_dump)
4983 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4984 dev, NULL,
4985 &fidx);
4986 else
4987 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4988 &fidx);
4989 if (err == -EMSGSIZE)
4990 goto out;
4992 cops = NULL;
4994 /* reset fdb offset to 0 for rest of the interfaces */
4995 cb->args[2] = 0;
4996 fidx = 0;
4997 cont:
4998 idx++;
5002 out:
5003 cb->args[0] = h;
5004 cb->args[1] = idx;
5005 cb->args[2] = fidx;
5007 return skb->len;
5010 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
5011 struct nlattr **tb, u8 *ndm_flags,
5012 int *br_idx, int *brport_idx, u8 **addr,
5013 u16 *vid, struct netlink_ext_ack *extack)
5015 struct ndmsg *ndm;
5016 int err, i;
5018 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
5019 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
5020 return -EINVAL;
5023 ndm = nlmsg_data(nlh);
5024 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
5025 ndm->ndm_type) {
5026 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
5027 return -EINVAL;
5030 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
5031 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
5032 return -EINVAL;
5035 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
5036 NDA_MAX, nda_policy, extack);
5037 if (err < 0)
5038 return err;
5040 *ndm_flags = ndm->ndm_flags;
5041 *brport_idx = ndm->ndm_ifindex;
5042 for (i = 0; i <= NDA_MAX; ++i) {
5043 if (!tb[i])
5044 continue;
5046 switch (i) {
5047 case NDA_MASTER:
5048 *br_idx = nla_get_u32(tb[i]);
5049 break;
5050 case NDA_LLADDR:
5051 if (nla_len(tb[i]) != ETH_ALEN) {
5052 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
5053 return -EINVAL;
5055 *addr = nla_data(tb[i]);
5056 break;
5057 case NDA_VLAN:
5058 err = fdb_vid_parse(tb[i], vid, extack);
5059 if (err)
5060 return err;
5061 break;
5062 case NDA_VNI:
5063 break;
5064 default:
5065 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
5066 return -EINVAL;
5070 return 0;
5073 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5074 struct netlink_ext_ack *extack)
5076 struct net_device *dev = NULL, *br_dev = NULL;
5077 const struct net_device_ops *ops = NULL;
5078 struct net *net = sock_net(in_skb->sk);
5079 struct nlattr *tb[NDA_MAX + 1];
5080 struct sk_buff *skb;
5081 int brport_idx = 0;
5082 u8 ndm_flags = 0;
5083 int br_idx = 0;
5084 u8 *addr = NULL;
5085 u16 vid = 0;
5086 int err;
5088 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
5089 &brport_idx, &addr, &vid, extack);
5090 if (err < 0)
5091 return err;
5093 if (!addr) {
5094 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
5095 return -EINVAL;
5098 if (brport_idx) {
5099 dev = __dev_get_by_index(net, brport_idx);
5100 if (!dev) {
5101 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
5102 return -ENODEV;
5106 if (br_idx) {
5107 if (dev) {
5108 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
5109 return -EINVAL;
5112 br_dev = __dev_get_by_index(net, br_idx);
5113 if (!br_dev) {
5114 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
5115 return -EINVAL;
5117 ops = br_dev->netdev_ops;
5120 if (dev) {
5121 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
5122 if (!netif_is_bridge_port(dev)) {
5123 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
5124 return -EINVAL;
5126 br_dev = netdev_master_upper_dev_get(dev);
5127 if (!br_dev) {
5128 NL_SET_ERR_MSG(extack, "Master of device not found");
5129 return -EINVAL;
5131 ops = br_dev->netdev_ops;
5132 } else {
5133 if (!(ndm_flags & NTF_SELF)) {
5134 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
5135 return -EINVAL;
5137 ops = dev->netdev_ops;
5141 if (!br_dev && !dev) {
5142 NL_SET_ERR_MSG(extack, "No device specified");
5143 return -ENODEV;
5146 if (!ops || !ops->ndo_fdb_get) {
5147 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
5148 return -EOPNOTSUPP;
5151 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
5152 if (!skb)
5153 return -ENOBUFS;
5155 if (br_dev)
5156 dev = br_dev;
5157 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
5158 NETLINK_CB(in_skb).portid,
5159 nlh->nlmsg_seq, extack);
5160 if (err)
5161 goto out;
5163 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5164 out:
5165 kfree_skb(skb);
5166 return err;
5169 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
5170 unsigned int attrnum, unsigned int flag)
5172 if (mask & flag)
5173 return nla_put_u8(skb, attrnum, !!(flags & flag));
5174 return 0;
5177 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
5178 struct net_device *dev, u16 mode,
5179 u32 flags, u32 mask, int nlflags,
5180 u32 filter_mask,
5181 int (*vlan_fill)(struct sk_buff *skb,
5182 struct net_device *dev,
5183 u32 filter_mask))
5185 struct nlmsghdr *nlh;
5186 struct ifinfomsg *ifm;
5187 struct nlattr *br_afspec;
5188 struct nlattr *protinfo;
5189 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
5190 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5191 int err = 0;
5193 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
5194 if (nlh == NULL)
5195 return -EMSGSIZE;
5197 ifm = nlmsg_data(nlh);
5198 ifm->ifi_family = AF_BRIDGE;
5199 ifm->__ifi_pad = 0;
5200 ifm->ifi_type = dev->type;
5201 ifm->ifi_index = dev->ifindex;
5202 ifm->ifi_flags = dev_get_flags(dev);
5203 ifm->ifi_change = 0;
5206 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5207 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5208 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
5209 (br_dev &&
5210 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
5211 (dev->addr_len &&
5212 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5213 (dev->ifindex != dev_get_iflink(dev) &&
5214 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
5215 goto nla_put_failure;
5217 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
5218 if (!br_afspec)
5219 goto nla_put_failure;
5221 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
5222 nla_nest_cancel(skb, br_afspec);
5223 goto nla_put_failure;
5226 if (mode != BRIDGE_MODE_UNDEF) {
5227 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
5228 nla_nest_cancel(skb, br_afspec);
5229 goto nla_put_failure;
5232 if (vlan_fill) {
5233 err = vlan_fill(skb, dev, filter_mask);
5234 if (err) {
5235 nla_nest_cancel(skb, br_afspec);
5236 goto nla_put_failure;
5239 nla_nest_end(skb, br_afspec);
5241 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
5242 if (!protinfo)
5243 goto nla_put_failure;
5245 if (brport_nla_put_flag(skb, flags, mask,
5246 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
5247 brport_nla_put_flag(skb, flags, mask,
5248 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
5249 brport_nla_put_flag(skb, flags, mask,
5250 IFLA_BRPORT_FAST_LEAVE,
5251 BR_MULTICAST_FAST_LEAVE) ||
5252 brport_nla_put_flag(skb, flags, mask,
5253 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
5254 brport_nla_put_flag(skb, flags, mask,
5255 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
5256 brport_nla_put_flag(skb, flags, mask,
5257 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
5258 brport_nla_put_flag(skb, flags, mask,
5259 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
5260 brport_nla_put_flag(skb, flags, mask,
5261 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
5262 brport_nla_put_flag(skb, flags, mask,
5263 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
5264 brport_nla_put_flag(skb, flags, mask,
5265 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
5266 nla_nest_cancel(skb, protinfo);
5267 goto nla_put_failure;
5270 nla_nest_end(skb, protinfo);
5272 nlmsg_end(skb, nlh);
5273 return 0;
5274 nla_put_failure:
5275 nlmsg_cancel(skb, nlh);
5276 return err ? err : -EMSGSIZE;
5278 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
5280 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
5281 bool strict_check, u32 *filter_mask,
5282 struct netlink_ext_ack *extack)
5284 struct nlattr *tb[IFLA_MAX+1];
5285 int err, i;
5287 if (strict_check) {
5288 struct ifinfomsg *ifm;
5290 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5291 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
5292 return -EINVAL;
5295 ifm = nlmsg_data(nlh);
5296 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5297 ifm->ifi_change || ifm->ifi_index) {
5298 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
5299 return -EINVAL;
5302 err = nlmsg_parse_deprecated_strict(nlh,
5303 sizeof(struct ifinfomsg),
5304 tb, IFLA_MAX, ifla_policy,
5305 extack);
5306 } else {
5307 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
5308 tb, IFLA_MAX, ifla_policy,
5309 extack);
5311 if (err < 0)
5312 return err;
5314 /* new attributes should only be added with strict checking */
5315 for (i = 0; i <= IFLA_MAX; ++i) {
5316 if (!tb[i])
5317 continue;
5319 switch (i) {
5320 case IFLA_EXT_MASK:
5321 *filter_mask = nla_get_u32(tb[i]);
5322 break;
5323 default:
5324 if (strict_check) {
5325 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
5326 return -EINVAL;
5331 return 0;
5334 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
5336 const struct nlmsghdr *nlh = cb->nlh;
5337 struct net *net = sock_net(skb->sk);
5338 struct net_device *dev;
5339 int idx = 0;
5340 u32 portid = NETLINK_CB(cb->skb).portid;
5341 u32 seq = nlh->nlmsg_seq;
5342 u32 filter_mask = 0;
5343 int err;
5345 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5346 cb->extack);
5347 if (err < 0 && cb->strict_check)
5348 return err;
5350 rcu_read_lock();
5351 for_each_netdev_rcu(net, dev) {
5352 const struct net_device_ops *ops = dev->netdev_ops;
5353 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5355 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5356 if (idx >= cb->args[0]) {
5357 err = br_dev->netdev_ops->ndo_bridge_getlink(
5358 skb, portid, seq, dev,
5359 filter_mask, NLM_F_MULTI);
5360 if (err < 0 && err != -EOPNOTSUPP) {
5361 if (likely(skb->len))
5362 break;
5364 goto out_err;
5367 idx++;
5370 if (ops->ndo_bridge_getlink) {
5371 if (idx >= cb->args[0]) {
5372 err = ops->ndo_bridge_getlink(skb, portid,
5373 seq, dev,
5374 filter_mask,
5375 NLM_F_MULTI);
5376 if (err < 0 && err != -EOPNOTSUPP) {
5377 if (likely(skb->len))
5378 break;
5380 goto out_err;
5383 idx++;
5386 err = skb->len;
5387 out_err:
5388 rcu_read_unlock();
5389 cb->args[0] = idx;
5391 return err;
5394 static inline size_t bridge_nlmsg_size(void)
5396 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5397 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5398 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5399 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
5400 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
5401 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
5402 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
5403 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
5404 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5405 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
5406 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
5409 static int rtnl_bridge_notify(struct net_device *dev)
5411 struct net *net = dev_net(dev);
5412 struct sk_buff *skb;
5413 int err = -EOPNOTSUPP;
5415 if (!dev->netdev_ops->ndo_bridge_getlink)
5416 return 0;
5418 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5419 if (!skb) {
5420 err = -ENOMEM;
5421 goto errout;
5424 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5425 if (err < 0)
5426 goto errout;
5428 /* Notification info is only filled for bridge ports, not the bridge
5429 * device itself. Therefore, a zero notification length is valid and
5430 * should not result in an error.
5432 if (!skb->len)
5433 goto errout;
5435 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5436 return 0;
5437 errout:
5438 WARN_ON(err == -EMSGSIZE);
5439 kfree_skb(skb);
5440 if (err)
5441 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5442 return err;
5445 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5446 struct netlink_ext_ack *extack)
5448 struct net *net = sock_net(skb->sk);
5449 struct ifinfomsg *ifm;
5450 struct net_device *dev;
5451 struct nlattr *br_spec, *attr, *br_flags_attr = NULL;
5452 int rem, err = -EOPNOTSUPP;
5453 u16 flags = 0;
5455 if (nlmsg_len(nlh) < sizeof(*ifm))
5456 return -EINVAL;
5458 ifm = nlmsg_data(nlh);
5459 if (ifm->ifi_family != AF_BRIDGE)
5460 return -EPFNOSUPPORT;
5462 dev = __dev_get_by_index(net, ifm->ifi_index);
5463 if (!dev) {
5464 NL_SET_ERR_MSG(extack, "unknown ifindex");
5465 return -ENODEV;
5468 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5469 if (br_spec) {
5470 nla_for_each_nested(attr, br_spec, rem) {
5471 if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !br_flags_attr) {
5472 if (nla_len(attr) < sizeof(flags))
5473 return -EINVAL;
5475 br_flags_attr = attr;
5476 flags = nla_get_u16(attr);
5479 if (nla_type(attr) == IFLA_BRIDGE_MODE) {
5480 if (nla_len(attr) < sizeof(u16))
5481 return -EINVAL;
5486 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5487 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5489 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5490 err = -EOPNOTSUPP;
5491 goto out;
5494 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5495 extack);
5496 if (err)
5497 goto out;
5499 flags &= ~BRIDGE_FLAGS_MASTER;
5502 if ((flags & BRIDGE_FLAGS_SELF)) {
5503 if (!dev->netdev_ops->ndo_bridge_setlink)
5504 err = -EOPNOTSUPP;
5505 else
5506 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5507 flags,
5508 extack);
5509 if (!err) {
5510 flags &= ~BRIDGE_FLAGS_SELF;
5512 /* Generate event to notify upper layer of bridge
5513 * change
5515 err = rtnl_bridge_notify(dev);
5519 if (br_flags_attr)
5520 memcpy(nla_data(br_flags_attr), &flags, sizeof(flags));
5521 out:
5522 return err;
5525 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5526 struct netlink_ext_ack *extack)
5528 struct net *net = sock_net(skb->sk);
5529 struct ifinfomsg *ifm;
5530 struct net_device *dev;
5531 struct nlattr *br_spec, *attr = NULL;
5532 int rem, err = -EOPNOTSUPP;
5533 u16 flags = 0;
5534 bool have_flags = false;
5536 if (nlmsg_len(nlh) < sizeof(*ifm))
5537 return -EINVAL;
5539 ifm = nlmsg_data(nlh);
5540 if (ifm->ifi_family != AF_BRIDGE)
5541 return -EPFNOSUPPORT;
5543 dev = __dev_get_by_index(net, ifm->ifi_index);
5544 if (!dev) {
5545 NL_SET_ERR_MSG(extack, "unknown ifindex");
5546 return -ENODEV;
5549 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5550 if (br_spec) {
5551 nla_for_each_nested_type(attr, IFLA_BRIDGE_FLAGS, br_spec,
5552 rem) {
5553 if (nla_len(attr) < sizeof(flags))
5554 return -EINVAL;
5556 have_flags = true;
5557 flags = nla_get_u16(attr);
5558 break;
5562 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5563 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5565 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5566 err = -EOPNOTSUPP;
5567 goto out;
5570 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5571 if (err)
5572 goto out;
5574 flags &= ~BRIDGE_FLAGS_MASTER;
5577 if ((flags & BRIDGE_FLAGS_SELF)) {
5578 if (!dev->netdev_ops->ndo_bridge_dellink)
5579 err = -EOPNOTSUPP;
5580 else
5581 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5582 flags);
5584 if (!err) {
5585 flags &= ~BRIDGE_FLAGS_SELF;
5587 /* Generate event to notify upper layer of bridge
5588 * change
5590 err = rtnl_bridge_notify(dev);
5594 if (have_flags)
5595 memcpy(nla_data(attr), &flags, sizeof(flags));
5596 out:
5597 return err;
5600 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5602 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5603 (!idxattr || idxattr == attrid);
5606 static bool
5607 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5609 return dev->netdev_ops &&
5610 dev->netdev_ops->ndo_has_offload_stats &&
5611 dev->netdev_ops->ndo_get_offload_stats &&
5612 dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5615 static unsigned int
5616 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5618 return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5619 sizeof(struct rtnl_link_stats64) : 0;
5622 static int
5623 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5624 struct sk_buff *skb)
5626 unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5627 struct nlattr *attr = NULL;
5628 void *attr_data;
5629 int err;
5631 if (!size)
5632 return -ENODATA;
5634 attr = nla_reserve_64bit(skb, attr_id, size,
5635 IFLA_OFFLOAD_XSTATS_UNSPEC);
5636 if (!attr)
5637 return -EMSGSIZE;
5639 attr_data = nla_data(attr);
5640 memset(attr_data, 0, size);
5642 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5643 if (err)
5644 return err;
5646 return 0;
5649 static unsigned int
5650 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5651 enum netdev_offload_xstats_type type)
5653 bool enabled = netdev_offload_xstats_enabled(dev, type);
5655 return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5658 struct rtnl_offload_xstats_request_used {
5659 bool request;
5660 bool used;
5663 static int
5664 rtnl_offload_xstats_get_stats(struct net_device *dev,
5665 enum netdev_offload_xstats_type type,
5666 struct rtnl_offload_xstats_request_used *ru,
5667 struct rtnl_hw_stats64 *stats,
5668 struct netlink_ext_ack *extack)
5670 bool request;
5671 bool used;
5672 int err;
5674 request = netdev_offload_xstats_enabled(dev, type);
5675 if (!request) {
5676 used = false;
5677 goto out;
5680 err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5681 if (err)
5682 return err;
5684 out:
5685 if (ru) {
5686 ru->request = request;
5687 ru->used = used;
5689 return 0;
5692 static int
5693 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5694 struct rtnl_offload_xstats_request_used *ru)
5696 struct nlattr *nest;
5698 nest = nla_nest_start(skb, attr_id);
5699 if (!nest)
5700 return -EMSGSIZE;
5702 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5703 goto nla_put_failure;
5705 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5706 goto nla_put_failure;
5708 nla_nest_end(skb, nest);
5709 return 0;
5711 nla_put_failure:
5712 nla_nest_cancel(skb, nest);
5713 return -EMSGSIZE;
5716 static int
5717 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5718 struct netlink_ext_ack *extack)
5720 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5721 struct rtnl_offload_xstats_request_used ru_l3;
5722 struct nlattr *nest;
5723 int err;
5725 err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5726 if (err)
5727 return err;
5729 nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5730 if (!nest)
5731 return -EMSGSIZE;
5733 if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5734 IFLA_OFFLOAD_XSTATS_L3_STATS,
5735 &ru_l3))
5736 goto nla_put_failure;
5738 nla_nest_end(skb, nest);
5739 return 0;
5741 nla_put_failure:
5742 nla_nest_cancel(skb, nest);
5743 return -EMSGSIZE;
5746 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5747 int *prividx, u32 off_filter_mask,
5748 struct netlink_ext_ack *extack)
5750 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5751 int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5752 int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5753 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5754 bool have_data = false;
5755 int err;
5757 if (*prividx <= attr_id_cpu_hit &&
5758 (off_filter_mask &
5759 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5760 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5761 if (!err) {
5762 have_data = true;
5763 } else if (err != -ENODATA) {
5764 *prividx = attr_id_cpu_hit;
5765 return err;
5769 if (*prividx <= attr_id_hw_s_info &&
5770 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5771 *prividx = attr_id_hw_s_info;
5773 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5774 if (err)
5775 return err;
5777 have_data = true;
5778 *prividx = 0;
5781 if (*prividx <= attr_id_l3_stats &&
5782 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5783 unsigned int size_l3;
5784 struct nlattr *attr;
5786 *prividx = attr_id_l3_stats;
5788 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5789 if (!size_l3)
5790 goto skip_l3_stats;
5791 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5792 IFLA_OFFLOAD_XSTATS_UNSPEC);
5793 if (!attr)
5794 return -EMSGSIZE;
5796 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5797 nla_data(attr), extack);
5798 if (err)
5799 return err;
5801 have_data = true;
5802 skip_l3_stats:
5803 *prividx = 0;
5806 if (!have_data)
5807 return -ENODATA;
5809 *prividx = 0;
5810 return 0;
5813 static unsigned int
5814 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5815 enum netdev_offload_xstats_type type)
5817 return nla_total_size(0) +
5818 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5819 nla_total_size(sizeof(u8)) +
5820 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5821 nla_total_size(sizeof(u8)) +
5825 static unsigned int
5826 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5828 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5830 return nla_total_size(0) +
5831 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5832 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5836 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5837 u32 off_filter_mask)
5839 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5840 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5841 int nla_size = 0;
5842 int size;
5844 if (off_filter_mask &
5845 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5846 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5847 nla_size += nla_total_size_64bit(size);
5850 if (off_filter_mask &
5851 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5852 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5854 if (off_filter_mask &
5855 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5856 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5857 nla_size += nla_total_size_64bit(size);
5860 if (nla_size != 0)
5861 nla_size += nla_total_size(0);
5863 return nla_size;
5866 struct rtnl_stats_dump_filters {
5867 /* mask[0] filters outer attributes. Then individual nests have their
5868 * filtering mask at the index of the nested attribute.
5870 u32 mask[IFLA_STATS_MAX + 1];
5873 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5874 int type, u32 pid, u32 seq, u32 change,
5875 unsigned int flags,
5876 const struct rtnl_stats_dump_filters *filters,
5877 int *idxattr, int *prividx,
5878 struct netlink_ext_ack *extack)
5880 unsigned int filter_mask = filters->mask[0];
5881 struct if_stats_msg *ifsm;
5882 struct nlmsghdr *nlh;
5883 struct nlattr *attr;
5884 int s_prividx = *prividx;
5885 int err;
5887 ASSERT_RTNL();
5889 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5890 if (!nlh)
5891 return -EMSGSIZE;
5893 ifsm = nlmsg_data(nlh);
5894 ifsm->family = PF_UNSPEC;
5895 ifsm->pad1 = 0;
5896 ifsm->pad2 = 0;
5897 ifsm->ifindex = dev->ifindex;
5898 ifsm->filter_mask = filter_mask;
5900 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5901 struct rtnl_link_stats64 *sp;
5903 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5904 sizeof(struct rtnl_link_stats64),
5905 IFLA_STATS_UNSPEC);
5906 if (!attr) {
5907 err = -EMSGSIZE;
5908 goto nla_put_failure;
5911 sp = nla_data(attr);
5912 dev_get_stats(dev, sp);
5915 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5916 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5918 if (ops && ops->fill_linkxstats) {
5919 *idxattr = IFLA_STATS_LINK_XSTATS;
5920 attr = nla_nest_start_noflag(skb,
5921 IFLA_STATS_LINK_XSTATS);
5922 if (!attr) {
5923 err = -EMSGSIZE;
5924 goto nla_put_failure;
5927 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5928 nla_nest_end(skb, attr);
5929 if (err)
5930 goto nla_put_failure;
5931 *idxattr = 0;
5935 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5936 *idxattr)) {
5937 const struct rtnl_link_ops *ops = NULL;
5938 const struct net_device *master;
5940 master = netdev_master_upper_dev_get(dev);
5941 if (master)
5942 ops = master->rtnl_link_ops;
5943 if (ops && ops->fill_linkxstats) {
5944 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5945 attr = nla_nest_start_noflag(skb,
5946 IFLA_STATS_LINK_XSTATS_SLAVE);
5947 if (!attr) {
5948 err = -EMSGSIZE;
5949 goto nla_put_failure;
5952 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5953 nla_nest_end(skb, attr);
5954 if (err)
5955 goto nla_put_failure;
5956 *idxattr = 0;
5960 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5961 *idxattr)) {
5962 u32 off_filter_mask;
5964 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5965 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5966 attr = nla_nest_start_noflag(skb,
5967 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5968 if (!attr) {
5969 err = -EMSGSIZE;
5970 goto nla_put_failure;
5973 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5974 off_filter_mask, extack);
5975 if (err == -ENODATA)
5976 nla_nest_cancel(skb, attr);
5977 else
5978 nla_nest_end(skb, attr);
5980 if (err && err != -ENODATA)
5981 goto nla_put_failure;
5982 *idxattr = 0;
5985 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5986 struct rtnl_af_ops *af_ops;
5988 *idxattr = IFLA_STATS_AF_SPEC;
5989 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5990 if (!attr) {
5991 err = -EMSGSIZE;
5992 goto nla_put_failure;
5995 rcu_read_lock();
5996 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5997 if (af_ops->fill_stats_af) {
5998 struct nlattr *af;
6000 af = nla_nest_start_noflag(skb,
6001 af_ops->family);
6002 if (!af) {
6003 rcu_read_unlock();
6004 err = -EMSGSIZE;
6005 goto nla_put_failure;
6007 err = af_ops->fill_stats_af(skb, dev);
6009 if (err == -ENODATA) {
6010 nla_nest_cancel(skb, af);
6011 } else if (err < 0) {
6012 rcu_read_unlock();
6013 goto nla_put_failure;
6016 nla_nest_end(skb, af);
6019 rcu_read_unlock();
6021 nla_nest_end(skb, attr);
6023 *idxattr = 0;
6026 nlmsg_end(skb, nlh);
6028 return 0;
6030 nla_put_failure:
6031 /* not a multi message or no progress mean a real error */
6032 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
6033 nlmsg_cancel(skb, nlh);
6034 else
6035 nlmsg_end(skb, nlh);
6037 return err;
6040 static size_t if_nlmsg_stats_size(const struct net_device *dev,
6041 const struct rtnl_stats_dump_filters *filters)
6043 size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
6044 unsigned int filter_mask = filters->mask[0];
6046 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
6047 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
6049 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
6050 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
6051 int attr = IFLA_STATS_LINK_XSTATS;
6053 if (ops && ops->get_linkxstats_size) {
6054 size += nla_total_size(ops->get_linkxstats_size(dev,
6055 attr));
6056 /* for IFLA_STATS_LINK_XSTATS */
6057 size += nla_total_size(0);
6061 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
6062 struct net_device *_dev = (struct net_device *)dev;
6063 const struct rtnl_link_ops *ops = NULL;
6064 const struct net_device *master;
6066 /* netdev_master_upper_dev_get can't take const */
6067 master = netdev_master_upper_dev_get(_dev);
6068 if (master)
6069 ops = master->rtnl_link_ops;
6070 if (ops && ops->get_linkxstats_size) {
6071 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
6073 size += nla_total_size(ops->get_linkxstats_size(dev,
6074 attr));
6075 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
6076 size += nla_total_size(0);
6080 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
6081 u32 off_filter_mask;
6083 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
6084 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
6087 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
6088 struct rtnl_af_ops *af_ops;
6090 /* for IFLA_STATS_AF_SPEC */
6091 size += nla_total_size(0);
6093 rcu_read_lock();
6094 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
6095 if (af_ops->get_stats_af_size) {
6096 size += nla_total_size(
6097 af_ops->get_stats_af_size(dev));
6099 /* for AF_* */
6100 size += nla_total_size(0);
6103 rcu_read_unlock();
6106 return size;
6109 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
6111 static const struct nla_policy
6112 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
6113 [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
6114 NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
6117 static const struct nla_policy
6118 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
6119 [IFLA_STATS_GET_FILTERS] =
6120 NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
6123 static const struct nla_policy
6124 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
6125 [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
6128 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
6129 struct rtnl_stats_dump_filters *filters,
6130 struct netlink_ext_ack *extack)
6132 struct nlattr *tb[IFLA_STATS_MAX + 1];
6133 int err;
6134 int at;
6136 err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
6137 rtnl_stats_get_policy_filters, extack);
6138 if (err < 0)
6139 return err;
6141 for (at = 1; at <= IFLA_STATS_MAX; at++) {
6142 if (tb[at]) {
6143 if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
6144 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
6145 return -EINVAL;
6147 filters->mask[at] = nla_get_u32(tb[at]);
6151 return 0;
6154 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
6155 u32 filter_mask,
6156 struct rtnl_stats_dump_filters *filters,
6157 struct netlink_ext_ack *extack)
6159 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6160 int err;
6161 int i;
6163 filters->mask[0] = filter_mask;
6164 for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
6165 filters->mask[i] = -1U;
6167 err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
6168 IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
6169 if (err < 0)
6170 return err;
6172 if (tb[IFLA_STATS_GET_FILTERS]) {
6173 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
6174 filters, extack);
6175 if (err)
6176 return err;
6179 return 0;
6182 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
6183 bool is_dump, struct netlink_ext_ack *extack)
6185 struct if_stats_msg *ifsm;
6187 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
6188 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
6189 return -EINVAL;
6192 if (!strict_check)
6193 return 0;
6195 ifsm = nlmsg_data(nlh);
6197 /* only requests using strict checks can pass data to influence
6198 * the dump. The legacy exception is filter_mask.
6200 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
6201 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
6202 return -EINVAL;
6204 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
6205 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
6206 return -EINVAL;
6209 return 0;
6212 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
6213 struct netlink_ext_ack *extack)
6215 struct rtnl_stats_dump_filters filters;
6216 struct net *net = sock_net(skb->sk);
6217 struct net_device *dev = NULL;
6218 int idxattr = 0, prividx = 0;
6219 struct if_stats_msg *ifsm;
6220 struct sk_buff *nskb;
6221 int err;
6223 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6224 false, extack);
6225 if (err)
6226 return err;
6228 ifsm = nlmsg_data(nlh);
6229 if (ifsm->ifindex > 0)
6230 dev = __dev_get_by_index(net, ifsm->ifindex);
6231 else
6232 return -EINVAL;
6234 if (!dev)
6235 return -ENODEV;
6237 if (!ifsm->filter_mask) {
6238 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
6239 return -EINVAL;
6242 err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
6243 if (err)
6244 return err;
6246 nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
6247 if (!nskb)
6248 return -ENOBUFS;
6250 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
6251 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
6252 0, &filters, &idxattr, &prividx, extack);
6253 if (err < 0) {
6254 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
6255 WARN_ON(err == -EMSGSIZE);
6256 kfree_skb(nskb);
6257 } else {
6258 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
6261 return err;
6264 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
6266 struct netlink_ext_ack *extack = cb->extack;
6267 struct rtnl_stats_dump_filters filters;
6268 struct net *net = sock_net(skb->sk);
6269 unsigned int flags = NLM_F_MULTI;
6270 struct if_stats_msg *ifsm;
6271 struct {
6272 unsigned long ifindex;
6273 int idxattr;
6274 int prividx;
6275 } *ctx = (void *)cb->ctx;
6276 struct net_device *dev;
6277 int err;
6279 cb->seq = net->dev_base_seq;
6281 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
6282 if (err)
6283 return err;
6285 ifsm = nlmsg_data(cb->nlh);
6286 if (!ifsm->filter_mask) {
6287 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
6288 return -EINVAL;
6291 err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
6292 extack);
6293 if (err)
6294 return err;
6296 for_each_netdev_dump(net, dev, ctx->ifindex) {
6297 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
6298 NETLINK_CB(cb->skb).portid,
6299 cb->nlh->nlmsg_seq, 0,
6300 flags, &filters,
6301 &ctx->idxattr, &ctx->prividx,
6302 extack);
6303 /* If we ran out of room on the first message,
6304 * we're in trouble.
6306 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
6308 if (err < 0)
6309 break;
6310 ctx->prividx = 0;
6311 ctx->idxattr = 0;
6312 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
6315 return err;
6318 void rtnl_offload_xstats_notify(struct net_device *dev)
6320 struct rtnl_stats_dump_filters response_filters = {};
6321 struct net *net = dev_net(dev);
6322 int idxattr = 0, prividx = 0;
6323 struct sk_buff *skb;
6324 int err = -ENOBUFS;
6326 ASSERT_RTNL();
6328 response_filters.mask[0] |=
6329 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6330 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6331 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6333 skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6334 GFP_KERNEL);
6335 if (!skb)
6336 goto errout;
6338 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6339 &response_filters, &idxattr, &prividx, NULL);
6340 if (err < 0) {
6341 kfree_skb(skb);
6342 goto errout;
6345 rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6346 return;
6348 errout:
6349 rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6351 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6353 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6354 struct netlink_ext_ack *extack)
6356 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6357 struct rtnl_stats_dump_filters response_filters = {};
6358 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6359 struct net *net = sock_net(skb->sk);
6360 struct net_device *dev = NULL;
6361 struct if_stats_msg *ifsm;
6362 bool notify = false;
6363 int err;
6365 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6366 false, extack);
6367 if (err)
6368 return err;
6370 ifsm = nlmsg_data(nlh);
6371 if (ifsm->family != AF_UNSPEC) {
6372 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6373 return -EINVAL;
6376 if (ifsm->ifindex > 0)
6377 dev = __dev_get_by_index(net, ifsm->ifindex);
6378 else
6379 return -EINVAL;
6381 if (!dev)
6382 return -ENODEV;
6384 if (ifsm->filter_mask) {
6385 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6386 return -EINVAL;
6389 err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6390 ifla_stats_set_policy, extack);
6391 if (err < 0)
6392 return err;
6394 if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6395 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6397 if (req)
6398 err = netdev_offload_xstats_enable(dev, t_l3, extack);
6399 else
6400 err = netdev_offload_xstats_disable(dev, t_l3);
6402 if (!err)
6403 notify = true;
6404 else if (err != -EALREADY)
6405 return err;
6407 response_filters.mask[0] |=
6408 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6409 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6410 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6413 if (notify)
6414 rtnl_offload_xstats_notify(dev);
6416 return 0;
6419 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6420 struct netlink_ext_ack *extack)
6422 struct br_port_msg *bpm;
6424 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6425 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6426 return -EINVAL;
6429 bpm = nlmsg_data(nlh);
6430 if (bpm->ifindex) {
6431 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6432 return -EINVAL;
6434 if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6435 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6436 return -EINVAL;
6439 return 0;
6442 struct rtnl_mdb_dump_ctx {
6443 long idx;
6446 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6448 struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6449 struct net *net = sock_net(skb->sk);
6450 struct net_device *dev;
6451 int idx, s_idx;
6452 int err;
6454 NL_ASSERT_CTX_FITS(struct rtnl_mdb_dump_ctx);
6456 if (cb->strict_check) {
6457 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6458 if (err)
6459 return err;
6462 s_idx = ctx->idx;
6463 idx = 0;
6465 for_each_netdev(net, dev) {
6466 if (idx < s_idx)
6467 goto skip;
6468 if (!dev->netdev_ops->ndo_mdb_dump)
6469 goto skip;
6471 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6472 if (err == -EMSGSIZE)
6473 goto out;
6474 /* Moving on to next device, reset markers and sequence
6475 * counters since they are all maintained per-device.
6477 memset(cb->ctx, 0, sizeof(cb->ctx));
6478 cb->prev_seq = 0;
6479 cb->seq = 0;
6480 skip:
6481 idx++;
6484 out:
6485 ctx->idx = idx;
6486 return skb->len;
6489 static int rtnl_validate_mdb_entry_get(const struct nlattr *attr,
6490 struct netlink_ext_ack *extack)
6492 struct br_mdb_entry *entry = nla_data(attr);
6494 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6495 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6496 return -EINVAL;
6499 if (entry->ifindex) {
6500 NL_SET_ERR_MSG(extack, "Entry ifindex cannot be specified");
6501 return -EINVAL;
6504 if (entry->state) {
6505 NL_SET_ERR_MSG(extack, "Entry state cannot be specified");
6506 return -EINVAL;
6509 if (entry->flags) {
6510 NL_SET_ERR_MSG(extack, "Entry flags cannot be specified");
6511 return -EINVAL;
6514 if (entry->vid >= VLAN_VID_MASK) {
6515 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6516 return -EINVAL;
6519 if (entry->addr.proto != htons(ETH_P_IP) &&
6520 entry->addr.proto != htons(ETH_P_IPV6) &&
6521 entry->addr.proto != 0) {
6522 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6523 return -EINVAL;
6526 return 0;
6529 static const struct nla_policy mdba_get_policy[MDBA_GET_ENTRY_MAX + 1] = {
6530 [MDBA_GET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6531 rtnl_validate_mdb_entry_get,
6532 sizeof(struct br_mdb_entry)),
6533 [MDBA_GET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6536 static int rtnl_mdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6537 struct netlink_ext_ack *extack)
6539 struct nlattr *tb[MDBA_GET_ENTRY_MAX + 1];
6540 struct net *net = sock_net(in_skb->sk);
6541 struct br_port_msg *bpm;
6542 struct net_device *dev;
6543 int err;
6545 err = nlmsg_parse(nlh, sizeof(struct br_port_msg), tb,
6546 MDBA_GET_ENTRY_MAX, mdba_get_policy, extack);
6547 if (err)
6548 return err;
6550 bpm = nlmsg_data(nlh);
6551 if (!bpm->ifindex) {
6552 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6553 return -EINVAL;
6556 dev = __dev_get_by_index(net, bpm->ifindex);
6557 if (!dev) {
6558 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6559 return -ENODEV;
6562 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_GET_ENTRY)) {
6563 NL_SET_ERR_MSG(extack, "Missing MDBA_GET_ENTRY attribute");
6564 return -EINVAL;
6567 if (!dev->netdev_ops->ndo_mdb_get) {
6568 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6569 return -EOPNOTSUPP;
6572 return dev->netdev_ops->ndo_mdb_get(dev, tb, NETLINK_CB(in_skb).portid,
6573 nlh->nlmsg_seq, extack);
6576 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6577 struct netlink_ext_ack *extack)
6579 struct br_mdb_entry *entry = nla_data(attr);
6581 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6582 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6583 return -EINVAL;
6586 if (entry->ifindex == 0) {
6587 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6588 return -EINVAL;
6591 if (entry->addr.proto == htons(ETH_P_IP)) {
6592 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6593 !ipv4_is_zeronet(entry->addr.u.ip4)) {
6594 NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6595 return -EINVAL;
6597 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6598 NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6599 return -EINVAL;
6601 #if IS_ENABLED(CONFIG_IPV6)
6602 } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6603 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6604 NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6605 return -EINVAL;
6607 #endif
6608 } else if (entry->addr.proto == 0) {
6609 /* L2 mdb */
6610 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6611 NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6612 return -EINVAL;
6614 } else {
6615 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6616 return -EINVAL;
6619 if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6620 NL_SET_ERR_MSG(extack, "Unknown entry state");
6621 return -EINVAL;
6623 if (entry->vid >= VLAN_VID_MASK) {
6624 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6625 return -EINVAL;
6628 return 0;
6631 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6632 [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6633 [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6634 rtnl_validate_mdb_entry,
6635 sizeof(struct br_mdb_entry)),
6636 [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6639 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6640 struct netlink_ext_ack *extack)
6642 struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6643 struct net *net = sock_net(skb->sk);
6644 struct br_port_msg *bpm;
6645 struct net_device *dev;
6646 int err;
6648 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6649 MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6650 if (err)
6651 return err;
6653 bpm = nlmsg_data(nlh);
6654 if (!bpm->ifindex) {
6655 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6656 return -EINVAL;
6659 dev = __dev_get_by_index(net, bpm->ifindex);
6660 if (!dev) {
6661 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6662 return -ENODEV;
6665 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6666 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6667 return -EINVAL;
6670 if (!dev->netdev_ops->ndo_mdb_add) {
6671 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6672 return -EOPNOTSUPP;
6675 return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6678 static int rtnl_validate_mdb_entry_del_bulk(const struct nlattr *attr,
6679 struct netlink_ext_ack *extack)
6681 struct br_mdb_entry *entry = nla_data(attr);
6682 struct br_mdb_entry zero_entry = {};
6684 if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6685 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6686 return -EINVAL;
6689 if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6690 NL_SET_ERR_MSG(extack, "Unknown entry state");
6691 return -EINVAL;
6694 if (entry->flags) {
6695 NL_SET_ERR_MSG(extack, "Entry flags cannot be set");
6696 return -EINVAL;
6699 if (entry->vid >= VLAN_N_VID - 1) {
6700 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6701 return -EINVAL;
6704 if (memcmp(&entry->addr, &zero_entry.addr, sizeof(entry->addr))) {
6705 NL_SET_ERR_MSG(extack, "Entry address cannot be set");
6706 return -EINVAL;
6709 return 0;
6712 static const struct nla_policy mdba_del_bulk_policy[MDBA_SET_ENTRY_MAX + 1] = {
6713 [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6714 rtnl_validate_mdb_entry_del_bulk,
6715 sizeof(struct br_mdb_entry)),
6716 [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6719 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6720 struct netlink_ext_ack *extack)
6722 bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
6723 struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6724 struct net *net = sock_net(skb->sk);
6725 struct br_port_msg *bpm;
6726 struct net_device *dev;
6727 int err;
6729 if (!del_bulk)
6730 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6731 MDBA_SET_ENTRY_MAX, mdba_policy,
6732 extack);
6733 else
6734 err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY_MAX,
6735 mdba_del_bulk_policy, extack);
6736 if (err)
6737 return err;
6739 bpm = nlmsg_data(nlh);
6740 if (!bpm->ifindex) {
6741 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6742 return -EINVAL;
6745 dev = __dev_get_by_index(net, bpm->ifindex);
6746 if (!dev) {
6747 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6748 return -ENODEV;
6751 if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6752 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6753 return -EINVAL;
6756 if (del_bulk) {
6757 if (!dev->netdev_ops->ndo_mdb_del_bulk) {
6758 NL_SET_ERR_MSG(extack, "Device does not support MDB bulk deletion");
6759 return -EOPNOTSUPP;
6761 return dev->netdev_ops->ndo_mdb_del_bulk(dev, tb, extack);
6764 if (!dev->netdev_ops->ndo_mdb_del) {
6765 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6766 return -EOPNOTSUPP;
6769 return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6772 /* Process one rtnetlink message. */
6774 static int rtnl_dumpit(struct sk_buff *skb, struct netlink_callback *cb)
6776 const bool needs_lock = !(cb->flags & RTNL_FLAG_DUMP_UNLOCKED);
6777 rtnl_dumpit_func dumpit = cb->data;
6778 int err;
6780 /* Previous iteration have already finished, avoid calling->dumpit()
6781 * again, it may not expect to be called after it reached the end.
6783 if (!dumpit)
6784 return 0;
6786 if (needs_lock)
6787 rtnl_lock();
6788 err = dumpit(skb, cb);
6789 if (needs_lock)
6790 rtnl_unlock();
6792 /* Old dump handlers used to send NLM_DONE as in a separate recvmsg().
6793 * Some applications which parse netlink manually depend on this.
6795 if (cb->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE) {
6796 if (err < 0 && err != -EMSGSIZE)
6797 return err;
6798 if (!err)
6799 cb->data = NULL;
6801 return skb->len;
6803 return err;
6806 static int rtnetlink_dump_start(struct sock *ssk, struct sk_buff *skb,
6807 const struct nlmsghdr *nlh,
6808 struct netlink_dump_control *control)
6810 if (control->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE ||
6811 !(control->flags & RTNL_FLAG_DUMP_UNLOCKED)) {
6812 WARN_ON(control->data);
6813 control->data = control->dump;
6814 control->dump = rtnl_dumpit;
6817 return netlink_dump_start(ssk, skb, nlh, control);
6820 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6821 struct netlink_ext_ack *extack)
6823 struct net *net = sock_net(skb->sk);
6824 struct rtnl_link *link;
6825 enum rtnl_kinds kind;
6826 struct module *owner;
6827 int err = -EOPNOTSUPP;
6828 rtnl_doit_func doit;
6829 unsigned int flags;
6830 int family;
6831 int type;
6833 type = nlh->nlmsg_type;
6834 if (type > RTM_MAX)
6835 return -EOPNOTSUPP;
6837 type -= RTM_BASE;
6839 /* All the messages must have at least 1 byte length */
6840 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6841 return 0;
6843 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6844 kind = rtnl_msgtype_kind(type);
6846 if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6847 return -EPERM;
6849 rcu_read_lock();
6850 if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6851 struct sock *rtnl;
6852 rtnl_dumpit_func dumpit;
6853 u32 min_dump_alloc = 0;
6855 link = rtnl_get_link(family, type);
6856 if (!link || !link->dumpit) {
6857 family = PF_UNSPEC;
6858 link = rtnl_get_link(family, type);
6859 if (!link || !link->dumpit)
6860 goto err_unlock;
6862 owner = link->owner;
6863 dumpit = link->dumpit;
6864 flags = link->flags;
6866 if (type == RTM_GETLINK - RTM_BASE)
6867 min_dump_alloc = rtnl_calcit(skb, nlh);
6869 err = 0;
6870 /* need to do this before rcu_read_unlock() */
6871 if (!try_module_get(owner))
6872 err = -EPROTONOSUPPORT;
6874 rcu_read_unlock();
6876 rtnl = net->rtnl;
6877 if (err == 0) {
6878 struct netlink_dump_control c = {
6879 .dump = dumpit,
6880 .min_dump_alloc = min_dump_alloc,
6881 .module = owner,
6882 .flags = flags,
6884 err = rtnetlink_dump_start(rtnl, skb, nlh, &c);
6885 /* netlink_dump_start() will keep a reference on
6886 * module if dump is still in progress.
6888 module_put(owner);
6890 return err;
6893 link = rtnl_get_link(family, type);
6894 if (!link || !link->doit) {
6895 family = PF_UNSPEC;
6896 link = rtnl_get_link(PF_UNSPEC, type);
6897 if (!link || !link->doit)
6898 goto out_unlock;
6901 owner = link->owner;
6902 if (!try_module_get(owner)) {
6903 err = -EPROTONOSUPPORT;
6904 goto out_unlock;
6907 flags = link->flags;
6908 if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6909 !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6910 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6911 module_put(owner);
6912 goto err_unlock;
6915 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6916 doit = link->doit;
6917 rcu_read_unlock();
6918 if (doit)
6919 err = doit(skb, nlh, extack);
6920 module_put(owner);
6921 return err;
6923 rcu_read_unlock();
6925 rtnl_lock();
6926 link = rtnl_get_link(family, type);
6927 if (link && link->doit)
6928 err = link->doit(skb, nlh, extack);
6929 rtnl_unlock();
6931 module_put(owner);
6933 return err;
6935 out_unlock:
6936 rcu_read_unlock();
6937 return err;
6939 err_unlock:
6940 rcu_read_unlock();
6941 return -EOPNOTSUPP;
6944 static void rtnetlink_rcv(struct sk_buff *skb)
6946 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6949 static int rtnetlink_bind(struct net *net, int group)
6951 switch (group) {
6952 case RTNLGRP_IPV4_MROUTE_R:
6953 case RTNLGRP_IPV6_MROUTE_R:
6954 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6955 return -EPERM;
6956 break;
6958 return 0;
6961 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6963 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6965 switch (event) {
6966 case NETDEV_REBOOT:
6967 case NETDEV_CHANGEMTU:
6968 case NETDEV_CHANGEADDR:
6969 case NETDEV_CHANGENAME:
6970 case NETDEV_FEAT_CHANGE:
6971 case NETDEV_BONDING_FAILOVER:
6972 case NETDEV_POST_TYPE_CHANGE:
6973 case NETDEV_NOTIFY_PEERS:
6974 case NETDEV_CHANGEUPPER:
6975 case NETDEV_RESEND_IGMP:
6976 case NETDEV_CHANGEINFODATA:
6977 case NETDEV_CHANGELOWERSTATE:
6978 case NETDEV_CHANGE_TX_QUEUE_LEN:
6979 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6980 GFP_KERNEL, NULL, 0, 0, NULL);
6981 break;
6982 default:
6983 break;
6985 return NOTIFY_DONE;
6988 static struct notifier_block rtnetlink_dev_notifier = {
6989 .notifier_call = rtnetlink_event,
6993 static int __net_init rtnetlink_net_init(struct net *net)
6995 struct sock *sk;
6996 struct netlink_kernel_cfg cfg = {
6997 .groups = RTNLGRP_MAX,
6998 .input = rtnetlink_rcv,
6999 .flags = NL_CFG_F_NONROOT_RECV,
7000 .bind = rtnetlink_bind,
7003 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
7004 if (!sk)
7005 return -ENOMEM;
7006 net->rtnl = sk;
7007 return 0;
7010 static void __net_exit rtnetlink_net_exit(struct net *net)
7012 netlink_kernel_release(net->rtnl);
7013 net->rtnl = NULL;
7016 static struct pernet_operations rtnetlink_net_ops = {
7017 .init = rtnetlink_net_init,
7018 .exit = rtnetlink_net_exit,
7021 static const struct rtnl_msg_handler rtnetlink_rtnl_msg_handlers[] __initconst = {
7022 {.msgtype = RTM_NEWLINK, .doit = rtnl_newlink,
7023 .flags = RTNL_FLAG_DOIT_PERNET},
7024 {.msgtype = RTM_DELLINK, .doit = rtnl_dellink,
7025 .flags = RTNL_FLAG_DOIT_PERNET_WIP},
7026 {.msgtype = RTM_GETLINK, .doit = rtnl_getlink,
7027 .dumpit = rtnl_dump_ifinfo, .flags = RTNL_FLAG_DUMP_SPLIT_NLM_DONE},
7028 {.msgtype = RTM_SETLINK, .doit = rtnl_setlink,
7029 .flags = RTNL_FLAG_DOIT_PERNET_WIP},
7030 {.msgtype = RTM_GETADDR, .dumpit = rtnl_dump_all},
7031 {.msgtype = RTM_GETROUTE, .dumpit = rtnl_dump_all},
7032 {.msgtype = RTM_GETNETCONF, .dumpit = rtnl_dump_all},
7033 {.msgtype = RTM_GETSTATS, .doit = rtnl_stats_get,
7034 .dumpit = rtnl_stats_dump},
7035 {.msgtype = RTM_SETSTATS, .doit = rtnl_stats_set},
7036 {.msgtype = RTM_NEWLINKPROP, .doit = rtnl_newlinkprop},
7037 {.msgtype = RTM_DELLINKPROP, .doit = rtnl_dellinkprop},
7038 {.protocol = PF_BRIDGE, .msgtype = RTM_GETLINK,
7039 .dumpit = rtnl_bridge_getlink},
7040 {.protocol = PF_BRIDGE, .msgtype = RTM_DELLINK,
7041 .doit = rtnl_bridge_dellink},
7042 {.protocol = PF_BRIDGE, .msgtype = RTM_SETLINK,
7043 .doit = rtnl_bridge_setlink},
7044 {.protocol = PF_BRIDGE, .msgtype = RTM_NEWNEIGH, .doit = rtnl_fdb_add},
7045 {.protocol = PF_BRIDGE, .msgtype = RTM_DELNEIGH, .doit = rtnl_fdb_del,
7046 .flags = RTNL_FLAG_BULK_DEL_SUPPORTED},
7047 {.protocol = PF_BRIDGE, .msgtype = RTM_GETNEIGH, .doit = rtnl_fdb_get,
7048 .dumpit = rtnl_fdb_dump},
7049 {.protocol = PF_BRIDGE, .msgtype = RTM_NEWMDB, .doit = rtnl_mdb_add},
7050 {.protocol = PF_BRIDGE, .msgtype = RTM_DELMDB, .doit = rtnl_mdb_del,
7051 .flags = RTNL_FLAG_BULK_DEL_SUPPORTED},
7052 {.protocol = PF_BRIDGE, .msgtype = RTM_GETMDB, .doit = rtnl_mdb_get,
7053 .dumpit = rtnl_mdb_dump},
7056 void __init rtnetlink_init(void)
7058 if (register_pernet_subsys(&rtnetlink_net_ops))
7059 panic("rtnetlink_init: cannot initialize rtnetlink\n");
7061 register_netdevice_notifier(&rtnetlink_dev_notifier);
7063 rtnl_register_many(rtnetlink_rtnl_msg_handlers);