1 // SPDX-License-Identifier: GPL-2.0-or-later
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>
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>
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>
46 #include <net/protocol.h>
48 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
57 #define RTNL_MAX_TYPE 50
58 #define RTNL_SLAVE_MAX_TYPE 40
62 rtnl_dumpit_func dumpit;
68 static DEFINE_MUTEX(rtnl_mutex);
72 mutex_lock(&rtnl_mutex);
74 EXPORT_SYMBOL(rtnl_lock);
76 int rtnl_lock_killable(void)
78 return mutex_lock_killable(&rtnl_mutex);
80 EXPORT_SYMBOL(rtnl_lock_killable);
82 static struct sk_buff *defer_kfree_skb_list;
83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
86 tail->next = defer_kfree_skb_list;
87 defer_kfree_skb_list = head;
90 EXPORT_SYMBOL(rtnl_kfree_skbs);
92 void __rtnl_unlock(void)
94 struct sk_buff *head = defer_kfree_skb_list;
96 defer_kfree_skb_list = NULL;
98 mutex_unlock(&rtnl_mutex);
101 struct sk_buff *next = head->next;
109 void rtnl_unlock(void)
111 /* This fellow will unlock it for us. */
114 EXPORT_SYMBOL(rtnl_unlock);
116 int rtnl_trylock(void)
118 return mutex_trylock(&rtnl_mutex);
120 EXPORT_SYMBOL(rtnl_trylock);
122 int rtnl_is_locked(void)
124 return mutex_is_locked(&rtnl_mutex);
126 EXPORT_SYMBOL(rtnl_is_locked);
128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
130 return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
134 #ifdef CONFIG_PROVE_LOCKING
135 bool lockdep_rtnl_is_held(void)
137 return lockdep_is_held(&rtnl_mutex);
139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
142 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
144 static inline int rtm_msgindex(int msgtype)
146 int msgindex = msgtype - RTM_BASE;
149 * msgindex < 0 implies someone tried to register a netlink
150 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
151 * the message type has not been added to linux/rtnetlink.h
153 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
160 struct rtnl_link __rcu **tab;
162 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
163 protocol = PF_UNSPEC;
165 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
167 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
169 return rcu_dereference_rtnl(tab[msgtype]);
172 static int rtnl_register_internal(struct module *owner,
173 int protocol, int msgtype,
174 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
177 struct rtnl_link *link, *old;
178 struct rtnl_link __rcu **tab;
182 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
183 msgindex = rtm_msgindex(msgtype);
186 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
188 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
192 /* ensures we see the 0 stores */
193 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
196 old = rtnl_dereference(tab[msgindex]);
198 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
202 link = kzalloc(sizeof(*link), GFP_KERNEL);
207 WARN_ON(link->owner && link->owner != owner);
210 WARN_ON(doit && link->doit && link->doit != doit);
213 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
215 link->dumpit = dumpit;
217 link->flags |= flags;
219 /* publish protocol:msgtype */
220 rcu_assign_pointer(tab[msgindex], link);
230 * rtnl_register_module - Register a rtnetlink message type
232 * @owner: module registering the hook (THIS_MODULE)
233 * @protocol: Protocol family or PF_UNSPEC
234 * @msgtype: rtnetlink message type
235 * @doit: Function pointer called for each request message
236 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
237 * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
239 * Like rtnl_register, but for use by removable modules.
241 int rtnl_register_module(struct module *owner,
242 int protocol, int msgtype,
243 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
246 return rtnl_register_internal(owner, protocol, msgtype,
247 doit, dumpit, flags);
249 EXPORT_SYMBOL_GPL(rtnl_register_module);
252 * rtnl_register - Register a rtnetlink message type
253 * @protocol: Protocol family or PF_UNSPEC
254 * @msgtype: rtnetlink message type
255 * @doit: Function pointer called for each request message
256 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
257 * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
259 * Registers the specified function pointers (at least one of them has
260 * to be non-NULL) to be called whenever a request message for the
261 * specified protocol family and message type is received.
263 * The special protocol family PF_UNSPEC may be used to define fallback
264 * function pointers for the case when no entry for the specific protocol
267 void rtnl_register(int protocol, int msgtype,
268 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
273 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
276 pr_err("Unable to register rtnetlink message handler, "
277 "protocol = %d, message type = %d\n", protocol, msgtype);
281 * rtnl_unregister - Unregister a rtnetlink message type
282 * @protocol: Protocol family or PF_UNSPEC
283 * @msgtype: rtnetlink message type
285 * Returns 0 on success or a negative error code.
287 int rtnl_unregister(int protocol, int msgtype)
289 struct rtnl_link __rcu **tab;
290 struct rtnl_link *link;
293 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
294 msgindex = rtm_msgindex(msgtype);
297 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
303 link = rtnl_dereference(tab[msgindex]);
304 RCU_INIT_POINTER(tab[msgindex], NULL);
307 kfree_rcu(link, rcu);
311 EXPORT_SYMBOL_GPL(rtnl_unregister);
314 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
315 * @protocol : Protocol family or PF_UNSPEC
317 * Identical to calling rtnl_unregster() for all registered message types
318 * of a certain protocol family.
320 void rtnl_unregister_all(int protocol)
322 struct rtnl_link __rcu **tab;
323 struct rtnl_link *link;
326 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
329 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
334 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
335 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
336 link = rtnl_dereference(tab[msgindex]);
340 RCU_INIT_POINTER(tab[msgindex], NULL);
341 kfree_rcu(link, rcu);
349 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
351 static LIST_HEAD(link_ops);
353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
355 const struct rtnl_link_ops *ops;
357 list_for_each_entry(ops, &link_ops, list) {
358 if (!strcmp(ops->kind, kind))
365 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
366 * @ops: struct rtnl_link_ops * to register
368 * The caller must hold the rtnl_mutex. This function should be used
369 * by drivers that create devices during module initialization. It
370 * must be called before registering the devices.
372 * Returns 0 on success or a negative error code.
374 int __rtnl_link_register(struct rtnl_link_ops *ops)
376 if (rtnl_link_ops_get(ops->kind))
379 /* The check for alloc/setup is here because if ops
380 * does not have that filled up, it is not possible
381 * to use the ops for creating device. So do not
382 * fill up dellink as well. That disables rtnl_dellink.
384 if ((ops->alloc || ops->setup) && !ops->dellink)
385 ops->dellink = unregister_netdevice_queue;
387 list_add_tail(&ops->list, &link_ops);
390 EXPORT_SYMBOL_GPL(__rtnl_link_register);
393 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
394 * @ops: struct rtnl_link_ops * to register
396 * Returns 0 on success or a negative error code.
398 int rtnl_link_register(struct rtnl_link_ops *ops)
402 /* Sanity-check max sizes to avoid stack buffer overflow. */
403 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
404 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
408 err = __rtnl_link_register(ops);
412 EXPORT_SYMBOL_GPL(rtnl_link_register);
414 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
416 struct net_device *dev;
417 LIST_HEAD(list_kill);
419 for_each_netdev(net, dev) {
420 if (dev->rtnl_link_ops == ops)
421 ops->dellink(dev, &list_kill);
423 unregister_netdevice_many(&list_kill);
427 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
428 * @ops: struct rtnl_link_ops * to unregister
430 * The caller must hold the rtnl_mutex and guarantee net_namespace_list
431 * integrity (hold pernet_ops_rwsem for writing to close the race
432 * with setup_net() and cleanup_net()).
434 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
439 __rtnl_kill_links(net, ops);
441 list_del(&ops->list);
443 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
445 /* Return with the rtnl_lock held when there are no network
446 * devices unregistering in any network namespace.
448 static void rtnl_lock_unregistering_all(void)
452 DEFINE_WAIT_FUNC(wait, woken_wake_function);
454 add_wait_queue(&netdev_unregistering_wq, &wait);
456 unregistering = false;
458 /* We held write locked pernet_ops_rwsem, and parallel
459 * setup_net() and cleanup_net() are not possible.
462 if (net->dev_unreg_count > 0) {
463 unregistering = true;
471 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
473 remove_wait_queue(&netdev_unregistering_wq, &wait);
477 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
478 * @ops: struct rtnl_link_ops * to unregister
480 void rtnl_link_unregister(struct rtnl_link_ops *ops)
482 /* Close the race with setup_net() and cleanup_net() */
483 down_write(&pernet_ops_rwsem);
484 rtnl_lock_unregistering_all();
485 __rtnl_link_unregister(ops);
487 up_write(&pernet_ops_rwsem);
489 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
491 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
493 struct net_device *master_dev;
494 const struct rtnl_link_ops *ops;
499 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
503 ops = master_dev->rtnl_link_ops;
504 if (!ops || !ops->get_slave_size)
506 /* IFLA_INFO_SLAVE_DATA + nested data */
507 size = nla_total_size(sizeof(struct nlattr)) +
508 ops->get_slave_size(master_dev, dev);
515 static size_t rtnl_link_get_size(const struct net_device *dev)
517 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
523 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
524 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
527 /* IFLA_INFO_DATA + nested data */
528 size += nla_total_size(sizeof(struct nlattr)) +
531 if (ops->get_xstats_size)
532 /* IFLA_INFO_XSTATS */
533 size += nla_total_size(ops->get_xstats_size(dev));
535 size += rtnl_link_get_slave_info_data_size(dev);
540 static LIST_HEAD(rtnl_af_ops);
542 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
544 const struct rtnl_af_ops *ops;
548 list_for_each_entry(ops, &rtnl_af_ops, list) {
549 if (ops->family == family)
557 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
558 * @ops: struct rtnl_af_ops * to register
560 * Returns 0 on success or a negative error code.
562 void rtnl_af_register(struct rtnl_af_ops *ops)
565 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
568 EXPORT_SYMBOL_GPL(rtnl_af_register);
571 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
572 * @ops: struct rtnl_af_ops * to unregister
574 void rtnl_af_unregister(struct rtnl_af_ops *ops)
577 list_del_rcu(&ops->list);
582 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
584 static size_t rtnl_link_get_af_size(const struct net_device *dev,
587 struct rtnl_af_ops *af_ops;
591 size = nla_total_size(sizeof(struct nlattr));
594 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
595 if (af_ops->get_link_af_size) {
596 /* AF_* + nested data */
597 size += nla_total_size(sizeof(struct nlattr)) +
598 af_ops->get_link_af_size(dev, ext_filter_mask);
606 static bool rtnl_have_link_slave_info(const struct net_device *dev)
608 struct net_device *master_dev;
613 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
614 if (master_dev && master_dev->rtnl_link_ops)
620 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
621 const struct net_device *dev)
623 struct net_device *master_dev;
624 const struct rtnl_link_ops *ops;
625 struct nlattr *slave_data;
628 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
631 ops = master_dev->rtnl_link_ops;
634 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
636 if (ops->fill_slave_info) {
637 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
640 err = ops->fill_slave_info(skb, master_dev, dev);
642 goto err_cancel_slave_data;
643 nla_nest_end(skb, slave_data);
647 err_cancel_slave_data:
648 nla_nest_cancel(skb, slave_data);
652 static int rtnl_link_info_fill(struct sk_buff *skb,
653 const struct net_device *dev)
655 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
661 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
663 if (ops->fill_xstats) {
664 err = ops->fill_xstats(skb, dev);
668 if (ops->fill_info) {
669 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
672 err = ops->fill_info(skb, dev);
674 goto err_cancel_data;
675 nla_nest_end(skb, data);
680 nla_nest_cancel(skb, data);
684 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
686 struct nlattr *linkinfo;
689 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
690 if (linkinfo == NULL)
693 err = rtnl_link_info_fill(skb, dev);
695 goto err_cancel_link;
697 err = rtnl_link_slave_info_fill(skb, dev);
699 goto err_cancel_link;
701 nla_nest_end(skb, linkinfo);
705 nla_nest_cancel(skb, linkinfo);
710 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
712 struct sock *rtnl = net->rtnl;
714 return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
717 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
719 struct sock *rtnl = net->rtnl;
721 return nlmsg_unicast(rtnl, skb, pid);
723 EXPORT_SYMBOL(rtnl_unicast);
725 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
726 struct nlmsghdr *nlh, gfp_t flags)
728 struct sock *rtnl = net->rtnl;
730 nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
732 EXPORT_SYMBOL(rtnl_notify);
734 void rtnl_set_sk_err(struct net *net, u32 group, int error)
736 struct sock *rtnl = net->rtnl;
738 netlink_set_err(rtnl, 0, group, error);
740 EXPORT_SYMBOL(rtnl_set_sk_err);
742 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
747 /* nothing is dumped for dst_default_metrics, so just skip the loop */
748 if (metrics == dst_default_metrics.metrics)
751 mx = nla_nest_start_noflag(skb, RTA_METRICS);
755 for (i = 0; i < RTAX_MAX; i++) {
757 if (i == RTAX_CC_ALGO - 1) {
758 char tmp[TCP_CA_NAME_MAX], *name;
760 name = tcp_ca_get_name_by_key(metrics[i], tmp);
763 if (nla_put_string(skb, i + 1, name))
764 goto nla_put_failure;
765 } else if (i == RTAX_FEATURES - 1) {
766 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
770 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
771 if (nla_put_u32(skb, i + 1, user_features))
772 goto nla_put_failure;
774 if (nla_put_u32(skb, i + 1, metrics[i]))
775 goto nla_put_failure;
782 nla_nest_cancel(skb, mx);
786 return nla_nest_end(skb, mx);
789 nla_nest_cancel(skb, mx);
792 EXPORT_SYMBOL(rtnetlink_put_metrics);
794 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
795 long expires, u32 error)
797 struct rta_cacheinfo ci = {
803 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
804 ci.rta_used = dst->__use;
805 ci.rta_clntref = atomic_read(&dst->__refcnt);
810 clock = jiffies_to_clock_t(abs(expires));
811 clock = min_t(unsigned long, clock, INT_MAX);
812 ci.rta_expires = (expires > 0) ? clock : -clock;
814 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
816 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
818 static void set_operstate(struct net_device *dev, unsigned char transition)
820 unsigned char operstate = dev->operstate;
822 switch (transition) {
824 if ((operstate == IF_OPER_DORMANT ||
825 operstate == IF_OPER_TESTING ||
826 operstate == IF_OPER_UNKNOWN) &&
827 !netif_dormant(dev) && !netif_testing(dev))
828 operstate = IF_OPER_UP;
831 case IF_OPER_TESTING:
832 if (operstate == IF_OPER_UP ||
833 operstate == IF_OPER_UNKNOWN)
834 operstate = IF_OPER_TESTING;
837 case IF_OPER_DORMANT:
838 if (operstate == IF_OPER_UP ||
839 operstate == IF_OPER_UNKNOWN)
840 operstate = IF_OPER_DORMANT;
844 if (dev->operstate != operstate) {
845 write_lock(&dev_base_lock);
846 dev->operstate = operstate;
847 write_unlock(&dev_base_lock);
848 netdev_state_change(dev);
852 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
854 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
855 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
858 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
859 const struct ifinfomsg *ifm)
861 unsigned int flags = ifm->ifi_flags;
863 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
865 flags = (flags & ifm->ifi_change) |
866 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
871 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
872 const struct rtnl_link_stats64 *b)
874 a->rx_packets = b->rx_packets;
875 a->tx_packets = b->tx_packets;
876 a->rx_bytes = b->rx_bytes;
877 a->tx_bytes = b->tx_bytes;
878 a->rx_errors = b->rx_errors;
879 a->tx_errors = b->tx_errors;
880 a->rx_dropped = b->rx_dropped;
881 a->tx_dropped = b->tx_dropped;
883 a->multicast = b->multicast;
884 a->collisions = b->collisions;
886 a->rx_length_errors = b->rx_length_errors;
887 a->rx_over_errors = b->rx_over_errors;
888 a->rx_crc_errors = b->rx_crc_errors;
889 a->rx_frame_errors = b->rx_frame_errors;
890 a->rx_fifo_errors = b->rx_fifo_errors;
891 a->rx_missed_errors = b->rx_missed_errors;
893 a->tx_aborted_errors = b->tx_aborted_errors;
894 a->tx_carrier_errors = b->tx_carrier_errors;
895 a->tx_fifo_errors = b->tx_fifo_errors;
896 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
897 a->tx_window_errors = b->tx_window_errors;
899 a->rx_compressed = b->rx_compressed;
900 a->tx_compressed = b->tx_compressed;
902 a->rx_nohandler = b->rx_nohandler;
906 static inline int rtnl_vfinfo_size(const struct net_device *dev,
909 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
910 int num_vfs = dev_num_vf(dev->dev.parent);
911 size_t size = nla_total_size(0);
914 nla_total_size(sizeof(struct ifla_vf_mac)) +
915 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
916 nla_total_size(sizeof(struct ifla_vf_vlan)) +
917 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
918 nla_total_size(MAX_VLAN_LIST_LEN *
919 sizeof(struct ifla_vf_vlan_info)) +
920 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
921 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
922 nla_total_size(sizeof(struct ifla_vf_rate)) +
923 nla_total_size(sizeof(struct ifla_vf_link_state)) +
924 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
925 nla_total_size(0) + /* nest IFLA_VF_STATS */
926 /* IFLA_VF_STATS_RX_PACKETS */
927 nla_total_size_64bit(sizeof(__u64)) +
928 /* IFLA_VF_STATS_TX_PACKETS */
929 nla_total_size_64bit(sizeof(__u64)) +
930 /* IFLA_VF_STATS_RX_BYTES */
931 nla_total_size_64bit(sizeof(__u64)) +
932 /* IFLA_VF_STATS_TX_BYTES */
933 nla_total_size_64bit(sizeof(__u64)) +
934 /* IFLA_VF_STATS_BROADCAST */
935 nla_total_size_64bit(sizeof(__u64)) +
936 /* IFLA_VF_STATS_MULTICAST */
937 nla_total_size_64bit(sizeof(__u64)) +
938 /* IFLA_VF_STATS_RX_DROPPED */
939 nla_total_size_64bit(sizeof(__u64)) +
940 /* IFLA_VF_STATS_TX_DROPPED */
941 nla_total_size_64bit(sizeof(__u64)) +
942 nla_total_size(sizeof(struct ifla_vf_trust)));
948 static size_t rtnl_port_size(const struct net_device *dev,
951 size_t port_size = nla_total_size(4) /* PORT_VF */
952 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
953 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
954 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
955 + nla_total_size(1) /* PROT_VDP_REQUEST */
956 + nla_total_size(2); /* PORT_VDP_RESPONSE */
957 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
958 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
960 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
963 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
964 !(ext_filter_mask & RTEXT_FILTER_VF))
966 if (dev_num_vf(dev->dev.parent))
967 return port_self_size + vf_ports_size +
968 vf_port_size * dev_num_vf(dev->dev.parent);
970 return port_self_size;
973 static size_t rtnl_xdp_size(void)
975 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
976 nla_total_size(1) + /* XDP_ATTACHED */
977 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
978 nla_total_size(4); /* XDP_<mode>_PROG_ID */
983 static size_t rtnl_prop_list_size(const struct net_device *dev)
985 struct netdev_name_node *name_node;
988 if (list_empty(&dev->name_node->list))
990 size = nla_total_size(0);
991 list_for_each_entry(name_node, &dev->name_node->list, list)
992 size += nla_total_size(ALTIFNAMSIZ);
996 static size_t rtnl_proto_down_size(const struct net_device *dev)
998 size_t size = nla_total_size(1);
1000 if (dev->proto_down_reason)
1001 size += nla_total_size(0) + nla_total_size(4);
1006 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1007 u32 ext_filter_mask)
1009 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1010 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1011 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1012 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1013 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1014 + nla_total_size(sizeof(struct rtnl_link_stats))
1015 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1016 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1017 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1018 + nla_total_size(4) /* IFLA_TXQLEN */
1019 + nla_total_size(4) /* IFLA_WEIGHT */
1020 + nla_total_size(4) /* IFLA_MTU */
1021 + nla_total_size(4) /* IFLA_LINK */
1022 + nla_total_size(4) /* IFLA_MASTER */
1023 + nla_total_size(1) /* IFLA_CARRIER */
1024 + nla_total_size(4) /* IFLA_PROMISCUITY */
1025 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1026 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1027 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1028 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1029 + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1030 + nla_total_size(1) /* IFLA_OPERSTATE */
1031 + nla_total_size(1) /* IFLA_LINKMODE */
1032 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1033 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1034 + nla_total_size(4) /* IFLA_GROUP */
1035 + nla_total_size(ext_filter_mask
1036 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1037 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1038 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1039 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1040 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1041 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1042 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1043 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1044 + rtnl_xdp_size() /* IFLA_XDP */
1045 + nla_total_size(4) /* IFLA_EVENT */
1046 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1047 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1048 + rtnl_proto_down_size(dev) /* proto down */
1049 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1050 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1051 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1052 + nla_total_size(4) /* IFLA_MIN_MTU */
1053 + nla_total_size(4) /* IFLA_MAX_MTU */
1054 + rtnl_prop_list_size(dev)
1055 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1059 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1061 struct nlattr *vf_ports;
1062 struct nlattr *vf_port;
1066 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1070 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1071 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1073 goto nla_put_failure;
1074 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1075 goto nla_put_failure;
1076 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1077 if (err == -EMSGSIZE)
1078 goto nla_put_failure;
1080 nla_nest_cancel(skb, vf_port);
1083 nla_nest_end(skb, vf_port);
1086 nla_nest_end(skb, vf_ports);
1091 nla_nest_cancel(skb, vf_ports);
1095 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1097 struct nlattr *port_self;
1100 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1104 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1106 nla_nest_cancel(skb, port_self);
1107 return (err == -EMSGSIZE) ? err : 0;
1110 nla_nest_end(skb, port_self);
1115 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1116 u32 ext_filter_mask)
1120 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1121 !(ext_filter_mask & RTEXT_FILTER_VF))
1124 err = rtnl_port_self_fill(skb, dev);
1128 if (dev_num_vf(dev->dev.parent)) {
1129 err = rtnl_vf_ports_fill(skb, dev);
1137 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1140 struct netdev_phys_item_id ppid;
1142 err = dev_get_phys_port_id(dev, &ppid);
1144 if (err == -EOPNOTSUPP)
1149 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1155 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1157 char name[IFNAMSIZ];
1160 err = dev_get_phys_port_name(dev, name, sizeof(name));
1162 if (err == -EOPNOTSUPP)
1167 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1173 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1175 struct netdev_phys_item_id ppid = { };
1178 err = dev_get_port_parent_id(dev, &ppid, false);
1180 if (err == -EOPNOTSUPP)
1185 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1191 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1192 struct net_device *dev)
1194 struct rtnl_link_stats64 *sp;
1195 struct nlattr *attr;
1197 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1198 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1202 sp = nla_data(attr);
1203 dev_get_stats(dev, sp);
1205 attr = nla_reserve(skb, IFLA_STATS,
1206 sizeof(struct rtnl_link_stats));
1210 copy_rtnl_link_stats(nla_data(attr), sp);
1215 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1216 struct net_device *dev,
1218 struct nlattr *vfinfo)
1220 struct ifla_vf_rss_query_en vf_rss_query_en;
1221 struct nlattr *vf, *vfstats, *vfvlanlist;
1222 struct ifla_vf_link_state vf_linkstate;
1223 struct ifla_vf_vlan_info vf_vlan_info;
1224 struct ifla_vf_spoofchk vf_spoofchk;
1225 struct ifla_vf_tx_rate vf_tx_rate;
1226 struct ifla_vf_stats vf_stats;
1227 struct ifla_vf_trust vf_trust;
1228 struct ifla_vf_vlan vf_vlan;
1229 struct ifla_vf_rate vf_rate;
1230 struct ifla_vf_mac vf_mac;
1231 struct ifla_vf_broadcast vf_broadcast;
1232 struct ifla_vf_info ivi;
1233 struct ifla_vf_guid node_guid;
1234 struct ifla_vf_guid port_guid;
1236 memset(&ivi, 0, sizeof(ivi));
1238 /* Not all SR-IOV capable drivers support the
1239 * spoofcheck and "RSS query enable" query. Preset to
1240 * -1 so the user space tool can detect that the driver
1241 * didn't report anything.
1244 ivi.rss_query_en = -1;
1246 /* The default value for VF link state is "auto"
1247 * IFLA_VF_LINK_STATE_AUTO which equals zero
1250 /* VLAN Protocol by default is 802.1Q */
1251 ivi.vlan_proto = htons(ETH_P_8021Q);
1252 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1255 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1256 memset(&node_guid, 0, sizeof(node_guid));
1257 memset(&port_guid, 0, sizeof(port_guid));
1266 vf_rss_query_en.vf =
1269 port_guid.vf = ivi.vf;
1271 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1272 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1273 vf_vlan.vlan = ivi.vlan;
1274 vf_vlan.qos = ivi.qos;
1275 vf_vlan_info.vlan = ivi.vlan;
1276 vf_vlan_info.qos = ivi.qos;
1277 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1278 vf_tx_rate.rate = ivi.max_tx_rate;
1279 vf_rate.min_tx_rate = ivi.min_tx_rate;
1280 vf_rate.max_tx_rate = ivi.max_tx_rate;
1281 vf_spoofchk.setting = ivi.spoofchk;
1282 vf_linkstate.link_state = ivi.linkstate;
1283 vf_rss_query_en.setting = ivi.rss_query_en;
1284 vf_trust.setting = ivi.trusted;
1285 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1287 goto nla_put_vfinfo_failure;
1288 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1289 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1290 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1291 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1293 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1295 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1297 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1299 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1300 sizeof(vf_rss_query_en),
1301 &vf_rss_query_en) ||
1302 nla_put(skb, IFLA_VF_TRUST,
1303 sizeof(vf_trust), &vf_trust))
1304 goto nla_put_vf_failure;
1306 if (dev->netdev_ops->ndo_get_vf_guid &&
1307 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1309 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1311 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1313 goto nla_put_vf_failure;
1315 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1317 goto nla_put_vf_failure;
1318 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1320 nla_nest_cancel(skb, vfvlanlist);
1321 goto nla_put_vf_failure;
1323 nla_nest_end(skb, vfvlanlist);
1324 memset(&vf_stats, 0, sizeof(vf_stats));
1325 if (dev->netdev_ops->ndo_get_vf_stats)
1326 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1328 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1330 goto nla_put_vf_failure;
1331 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1332 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1333 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1334 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1335 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1336 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1337 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1338 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1339 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1340 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1341 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1342 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1343 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1344 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1345 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1346 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1347 nla_nest_cancel(skb, vfstats);
1348 goto nla_put_vf_failure;
1350 nla_nest_end(skb, vfstats);
1351 nla_nest_end(skb, vf);
1355 nla_nest_cancel(skb, vf);
1356 nla_put_vfinfo_failure:
1357 nla_nest_cancel(skb, vfinfo);
1361 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1362 struct net_device *dev,
1363 u32 ext_filter_mask)
1365 struct nlattr *vfinfo;
1368 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1371 num_vfs = dev_num_vf(dev->dev.parent);
1372 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1375 if (!dev->netdev_ops->ndo_get_vf_config)
1378 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1382 for (i = 0; i < num_vfs; i++) {
1383 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1387 nla_nest_end(skb, vfinfo);
1391 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1393 struct rtnl_link_ifmap map;
1395 memset(&map, 0, sizeof(map));
1396 map.mem_start = dev->mem_start;
1397 map.mem_end = dev->mem_end;
1398 map.base_addr = dev->base_addr;
1401 map.port = dev->if_port;
1403 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1409 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1411 const struct bpf_prog *generic_xdp_prog;
1415 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1416 if (!generic_xdp_prog)
1418 return generic_xdp_prog->aux->id;
1421 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1423 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1426 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1428 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1431 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1432 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1433 u32 (*get_prog_id)(struct net_device *dev))
1438 curr_id = get_prog_id(dev);
1443 err = nla_put_u32(skb, attr, curr_id);
1447 if (*mode != XDP_ATTACHED_NONE)
1448 *mode = XDP_ATTACHED_MULTI;
1455 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1462 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1467 mode = XDP_ATTACHED_NONE;
1468 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1469 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1472 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1473 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1476 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1477 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1481 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1485 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1486 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1491 nla_nest_end(skb, xdp);
1495 nla_nest_cancel(skb, xdp);
1499 static u32 rtnl_get_event(unsigned long event)
1501 u32 rtnl_event_type = IFLA_EVENT_NONE;
1505 rtnl_event_type = IFLA_EVENT_REBOOT;
1507 case NETDEV_FEAT_CHANGE:
1508 rtnl_event_type = IFLA_EVENT_FEATURES;
1510 case NETDEV_BONDING_FAILOVER:
1511 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1513 case NETDEV_NOTIFY_PEERS:
1514 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1516 case NETDEV_RESEND_IGMP:
1517 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1519 case NETDEV_CHANGEINFODATA:
1520 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1526 return rtnl_event_type;
1529 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1531 const struct net_device *upper_dev;
1536 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1538 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1544 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1547 int ifindex = dev_get_iflink(dev);
1549 if (force || dev->ifindex != ifindex)
1550 return nla_put_u32(skb, IFLA_LINK, ifindex);
1555 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1556 struct net_device *dev)
1561 ret = dev_get_alias(dev, buf, sizeof(buf));
1562 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1565 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1566 const struct net_device *dev,
1567 struct net *src_net, gfp_t gfp)
1569 bool put_iflink = false;
1571 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1572 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1574 if (!net_eq(dev_net(dev), link_net)) {
1575 int id = peernet2id_alloc(src_net, link_net, gfp);
1577 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1584 return nla_put_iflink(skb, dev, put_iflink);
1587 static int rtnl_fill_link_af(struct sk_buff *skb,
1588 const struct net_device *dev,
1589 u32 ext_filter_mask)
1591 const struct rtnl_af_ops *af_ops;
1592 struct nlattr *af_spec;
1594 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1598 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1602 if (!af_ops->fill_link_af)
1605 af = nla_nest_start_noflag(skb, af_ops->family);
1609 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1611 * Caller may return ENODATA to indicate that there
1612 * was no data to be dumped. This is not an error, it
1613 * means we should trim the attribute header and
1616 if (err == -ENODATA)
1617 nla_nest_cancel(skb, af);
1621 nla_nest_end(skb, af);
1624 nla_nest_end(skb, af_spec);
1628 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1629 const struct net_device *dev)
1631 struct netdev_name_node *name_node;
1634 list_for_each_entry(name_node, &dev->name_node->list, list) {
1635 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1642 static int rtnl_fill_prop_list(struct sk_buff *skb,
1643 const struct net_device *dev)
1645 struct nlattr *prop_list;
1648 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1652 ret = rtnl_fill_alt_ifnames(skb, dev);
1656 nla_nest_end(skb, prop_list);
1660 nla_nest_cancel(skb, prop_list);
1664 static int rtnl_fill_proto_down(struct sk_buff *skb,
1665 const struct net_device *dev)
1670 if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1671 goto nla_put_failure;
1673 preason = dev->proto_down_reason;
1677 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1681 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1682 nla_nest_cancel(skb, pr);
1683 goto nla_put_failure;
1686 nla_nest_end(skb, pr);
1693 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1694 struct net_device *dev, struct net *src_net,
1695 int type, u32 pid, u32 seq, u32 change,
1696 unsigned int flags, u32 ext_filter_mask,
1697 u32 event, int *new_nsid, int new_ifindex,
1698 int tgt_netnsid, gfp_t gfp)
1700 struct ifinfomsg *ifm;
1701 struct nlmsghdr *nlh;
1702 struct Qdisc *qdisc;
1705 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1709 ifm = nlmsg_data(nlh);
1710 ifm->ifi_family = AF_UNSPEC;
1712 ifm->ifi_type = dev->type;
1713 ifm->ifi_index = dev->ifindex;
1714 ifm->ifi_flags = dev_get_flags(dev);
1715 ifm->ifi_change = change;
1717 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1718 goto nla_put_failure;
1720 qdisc = rtnl_dereference(dev->qdisc);
1721 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1722 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1723 nla_put_u8(skb, IFLA_OPERSTATE,
1724 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1725 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1726 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1727 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1728 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1729 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1730 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1731 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1732 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1733 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1734 nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1736 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1738 put_master_ifindex(skb, dev) ||
1739 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1741 nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1742 nla_put_ifalias(skb, dev) ||
1743 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1744 atomic_read(&dev->carrier_up_count) +
1745 atomic_read(&dev->carrier_down_count)) ||
1746 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1747 atomic_read(&dev->carrier_up_count)) ||
1748 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1749 atomic_read(&dev->carrier_down_count)))
1750 goto nla_put_failure;
1752 if (rtnl_fill_proto_down(skb, dev))
1753 goto nla_put_failure;
1755 if (event != IFLA_EVENT_NONE) {
1756 if (nla_put_u32(skb, IFLA_EVENT, event))
1757 goto nla_put_failure;
1760 if (rtnl_fill_link_ifmap(skb, dev))
1761 goto nla_put_failure;
1763 if (dev->addr_len) {
1764 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1765 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1766 goto nla_put_failure;
1769 if (rtnl_phys_port_id_fill(skb, dev))
1770 goto nla_put_failure;
1772 if (rtnl_phys_port_name_fill(skb, dev))
1773 goto nla_put_failure;
1775 if (rtnl_phys_switch_id_fill(skb, dev))
1776 goto nla_put_failure;
1778 if (rtnl_fill_stats(skb, dev))
1779 goto nla_put_failure;
1781 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1782 goto nla_put_failure;
1784 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1785 goto nla_put_failure;
1787 if (rtnl_xdp_fill(skb, dev))
1788 goto nla_put_failure;
1790 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1791 if (rtnl_link_fill(skb, dev) < 0)
1792 goto nla_put_failure;
1795 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1796 goto nla_put_failure;
1799 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1800 goto nla_put_failure;
1802 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1803 goto nla_put_failure;
1805 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1806 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1807 goto nla_put_failure;
1810 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1811 goto nla_put_failure_rcu;
1814 if (rtnl_fill_prop_list(skb, dev))
1815 goto nla_put_failure;
1817 if (dev->dev.parent &&
1818 nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1819 dev_name(dev->dev.parent)))
1820 goto nla_put_failure;
1822 if (dev->dev.parent && dev->dev.parent->bus &&
1823 nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1824 dev->dev.parent->bus->name))
1825 goto nla_put_failure;
1827 nlmsg_end(skb, nlh);
1830 nla_put_failure_rcu:
1833 nlmsg_cancel(skb, nlh);
1837 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1838 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1839 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1840 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1841 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1842 [IFLA_MTU] = { .type = NLA_U32 },
1843 [IFLA_LINK] = { .type = NLA_U32 },
1844 [IFLA_MASTER] = { .type = NLA_U32 },
1845 [IFLA_CARRIER] = { .type = NLA_U8 },
1846 [IFLA_TXQLEN] = { .type = NLA_U32 },
1847 [IFLA_WEIGHT] = { .type = NLA_U32 },
1848 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1849 [IFLA_LINKMODE] = { .type = NLA_U8 },
1850 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1851 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1852 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1853 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1854 * allow 0-length string (needed to remove an alias).
1856 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1857 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1858 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1859 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1860 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1861 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1862 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1863 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1864 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1865 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1866 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1867 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1868 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1869 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1870 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1871 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1872 [IFLA_XDP] = { .type = NLA_NESTED },
1873 [IFLA_EVENT] = { .type = NLA_U32 },
1874 [IFLA_GROUP] = { .type = NLA_U32 },
1875 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
1876 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1877 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1878 [IFLA_MIN_MTU] = { .type = NLA_U32 },
1879 [IFLA_MAX_MTU] = { .type = NLA_U32 },
1880 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
1881 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
1882 .len = ALTIFNAMSIZ - 1 },
1883 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
1884 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1885 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
1886 [IFLA_PARENT_DEV_NAME] = { .type = NLA_NUL_STRING },
1887 [IFLA_GRO_MAX_SIZE] = { .type = NLA_U32 },
1890 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1891 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1892 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1893 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1894 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1897 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1898 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1899 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1900 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1901 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1902 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1903 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1904 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1905 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1906 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1907 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1908 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1909 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1910 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1913 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1914 [IFLA_PORT_VF] = { .type = NLA_U32 },
1915 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1916 .len = PORT_PROFILE_MAX },
1917 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1918 .len = PORT_UUID_MAX },
1919 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1920 .len = PORT_UUID_MAX },
1921 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1922 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1924 /* Unused, but we need to keep it here since user space could
1925 * fill it. It's also broken with regard to NLA_BINARY use in
1926 * combination with structs.
1928 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1929 .len = sizeof(struct ifla_port_vsi) },
1932 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1933 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1934 [IFLA_XDP_FD] = { .type = NLA_S32 },
1935 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
1936 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1937 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1938 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1941 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1943 const struct rtnl_link_ops *ops = NULL;
1944 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1946 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1949 if (linfo[IFLA_INFO_KIND]) {
1950 char kind[MODULE_NAME_LEN];
1952 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1953 ops = rtnl_link_ops_get(kind);
1959 static bool link_master_filtered(struct net_device *dev, int master_idx)
1961 struct net_device *master;
1966 master = netdev_master_upper_dev_get(dev);
1968 /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
1969 * another invalid value for ifindex to denote "no master".
1971 if (master_idx == -1)
1974 if (!master || master->ifindex != master_idx)
1980 static bool link_kind_filtered(const struct net_device *dev,
1981 const struct rtnl_link_ops *kind_ops)
1983 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1989 static bool link_dump_filtered(struct net_device *dev,
1991 const struct rtnl_link_ops *kind_ops)
1993 if (link_master_filtered(dev, master_idx) ||
1994 link_kind_filtered(dev, kind_ops))
2001 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2002 * @sk: netlink socket
2003 * @netnsid: network namespace identifier
2005 * Returns the network namespace identified by netnsid on success or an error
2006 * pointer on failure.
2008 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2012 net = get_net_ns_by_id(sock_net(sk), netnsid);
2014 return ERR_PTR(-EINVAL);
2016 /* For now, the caller is required to have CAP_NET_ADMIN in
2017 * the user namespace owning the target net ns.
2019 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2021 return ERR_PTR(-EACCES);
2025 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2027 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2028 bool strict_check, struct nlattr **tb,
2029 struct netlink_ext_ack *extack)
2034 struct ifinfomsg *ifm;
2036 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2037 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2041 ifm = nlmsg_data(nlh);
2042 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2044 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2047 if (ifm->ifi_index) {
2048 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2052 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2053 IFLA_MAX, ifla_policy,
2057 /* A hack to preserve kernel<->userspace interface.
2058 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2059 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2060 * what iproute2 < v3.9.0 used.
2061 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2062 * attribute, its netlink message is shorter than struct ifinfomsg.
2064 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2065 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2067 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2071 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2073 struct netlink_ext_ack *extack = cb->extack;
2074 const struct nlmsghdr *nlh = cb->nlh;
2075 struct net *net = sock_net(skb->sk);
2076 struct net *tgt_net = net;
2079 struct net_device *dev;
2080 struct hlist_head *head;
2081 struct nlattr *tb[IFLA_MAX+1];
2082 u32 ext_filter_mask = 0;
2083 const struct rtnl_link_ops *kind_ops = NULL;
2084 unsigned int flags = NLM_F_MULTI;
2090 s_idx = cb->args[1];
2092 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2094 if (cb->strict_check)
2100 for (i = 0; i <= IFLA_MAX; ++i) {
2104 /* new attributes should only be added with strict checking */
2106 case IFLA_TARGET_NETNSID:
2107 netnsid = nla_get_s32(tb[i]);
2108 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2109 if (IS_ERR(tgt_net)) {
2110 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2111 return PTR_ERR(tgt_net);
2115 ext_filter_mask = nla_get_u32(tb[i]);
2118 master_idx = nla_get_u32(tb[i]);
2121 kind_ops = linkinfo_to_kind_ops(tb[i]);
2124 if (cb->strict_check) {
2125 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2131 if (master_idx || kind_ops)
2132 flags |= NLM_F_DUMP_FILTERED;
2135 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2137 head = &tgt_net->dev_index_head[h];
2138 hlist_for_each_entry(dev, head, index_hlist) {
2139 if (link_dump_filtered(dev, master_idx, kind_ops))
2143 err = rtnl_fill_ifinfo(skb, dev, net,
2145 NETLINK_CB(cb->skb).portid,
2146 nlh->nlmsg_seq, 0, flags,
2147 ext_filter_mask, 0, NULL, 0,
2148 netnsid, GFP_KERNEL);
2151 if (likely(skb->len))
2165 cb->seq = tgt_net->dev_base_seq;
2166 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2173 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2174 struct netlink_ext_ack *exterr)
2176 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2179 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2181 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2184 /* Examine the link attributes and figure out which
2185 * network namespace we are talking about.
2187 if (tb[IFLA_NET_NS_PID])
2188 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2189 else if (tb[IFLA_NET_NS_FD])
2190 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2192 net = get_net(src_net);
2195 EXPORT_SYMBOL(rtnl_link_get_net);
2197 /* Figure out which network namespace we are talking about by
2198 * examining the link attributes in the following order:
2200 * 1. IFLA_NET_NS_PID
2202 * 3. IFLA_TARGET_NETNSID
2204 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2205 struct nlattr *tb[])
2209 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2210 return rtnl_link_get_net(src_net, tb);
2212 if (!tb[IFLA_TARGET_NETNSID])
2213 return get_net(src_net);
2215 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2217 return ERR_PTR(-EINVAL);
2222 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2223 struct net *src_net,
2224 struct nlattr *tb[], int cap)
2228 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2232 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2234 return ERR_PTR(-EPERM);
2240 /* Verify that rtnetlink requests do not pass additional properties
2241 * potentially referring to different network namespaces.
2243 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2244 struct netlink_ext_ack *extack,
2248 if (netns_id_only) {
2249 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2252 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2256 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2259 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2262 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2268 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2272 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2273 struct netlink_ext_ack *extack)
2276 if (tb[IFLA_ADDRESS] &&
2277 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2280 if (tb[IFLA_BROADCAST] &&
2281 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2285 if (tb[IFLA_AF_SPEC]) {
2289 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2290 const struct rtnl_af_ops *af_ops;
2292 af_ops = rtnl_af_lookup(nla_type(af));
2294 return -EAFNOSUPPORT;
2296 if (!af_ops->set_link_af)
2299 if (af_ops->validate_link_af) {
2300 err = af_ops->validate_link_af(dev, af, extack);
2307 if (tb[IFLA_GRO_MAX_SIZE]) {
2308 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2310 if (gro_max_size > GRO_MAX_SIZE) {
2311 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2318 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2321 const struct net_device_ops *ops = dev->netdev_ops;
2323 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2326 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2328 if (dev->type != ARPHRD_INFINIBAND)
2331 return handle_infiniband_guid(dev, ivt, guid_type);
2334 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2336 const struct net_device_ops *ops = dev->netdev_ops;
2339 if (tb[IFLA_VF_MAC]) {
2340 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2342 if (ivm->vf >= INT_MAX)
2345 if (ops->ndo_set_vf_mac)
2346 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2352 if (tb[IFLA_VF_VLAN]) {
2353 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2355 if (ivv->vf >= INT_MAX)
2358 if (ops->ndo_set_vf_vlan)
2359 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2361 htons(ETH_P_8021Q));
2366 if (tb[IFLA_VF_VLAN_LIST]) {
2367 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2368 struct nlattr *attr;
2372 if (!ops->ndo_set_vf_vlan)
2375 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2376 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2377 nla_len(attr) < NLA_HDRLEN) {
2380 if (len >= MAX_VLAN_LIST_LEN)
2382 ivvl[len] = nla_data(attr);
2389 if (ivvl[0]->vf >= INT_MAX)
2391 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2392 ivvl[0]->qos, ivvl[0]->vlan_proto);
2397 if (tb[IFLA_VF_TX_RATE]) {
2398 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2399 struct ifla_vf_info ivf;
2401 if (ivt->vf >= INT_MAX)
2404 if (ops->ndo_get_vf_config)
2405 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2410 if (ops->ndo_set_vf_rate)
2411 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2418 if (tb[IFLA_VF_RATE]) {
2419 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2421 if (ivt->vf >= INT_MAX)
2424 if (ops->ndo_set_vf_rate)
2425 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2432 if (tb[IFLA_VF_SPOOFCHK]) {
2433 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2435 if (ivs->vf >= INT_MAX)
2438 if (ops->ndo_set_vf_spoofchk)
2439 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2445 if (tb[IFLA_VF_LINK_STATE]) {
2446 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2448 if (ivl->vf >= INT_MAX)
2451 if (ops->ndo_set_vf_link_state)
2452 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2458 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2459 struct ifla_vf_rss_query_en *ivrssq_en;
2462 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2463 if (ivrssq_en->vf >= INT_MAX)
2465 if (ops->ndo_set_vf_rss_query_en)
2466 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2467 ivrssq_en->setting);
2472 if (tb[IFLA_VF_TRUST]) {
2473 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2475 if (ivt->vf >= INT_MAX)
2478 if (ops->ndo_set_vf_trust)
2479 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2484 if (tb[IFLA_VF_IB_NODE_GUID]) {
2485 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2487 if (ivt->vf >= INT_MAX)
2489 if (!ops->ndo_set_vf_guid)
2491 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2494 if (tb[IFLA_VF_IB_PORT_GUID]) {
2495 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2497 if (ivt->vf >= INT_MAX)
2499 if (!ops->ndo_set_vf_guid)
2502 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2508 static int do_set_master(struct net_device *dev, int ifindex,
2509 struct netlink_ext_ack *extack)
2511 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2512 const struct net_device_ops *ops;
2516 if (upper_dev->ifindex == ifindex)
2518 ops = upper_dev->netdev_ops;
2519 if (ops->ndo_del_slave) {
2520 err = ops->ndo_del_slave(upper_dev, dev);
2529 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2532 ops = upper_dev->netdev_ops;
2533 if (ops->ndo_add_slave) {
2534 err = ops->ndo_add_slave(upper_dev, dev, extack);
2544 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2545 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2546 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2549 static int do_set_proto_down(struct net_device *dev,
2550 struct nlattr *nl_proto_down,
2551 struct nlattr *nl_proto_down_reason,
2552 struct netlink_ext_ack *extack)
2554 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2555 unsigned long mask = 0;
2560 if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2561 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2565 if (nl_proto_down_reason) {
2566 err = nla_parse_nested_deprecated(pdreason,
2567 IFLA_PROTO_DOWN_REASON_MAX,
2568 nl_proto_down_reason,
2569 ifla_proto_down_reason_policy,
2574 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2575 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2579 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2581 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2582 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2584 dev_change_proto_down_reason(dev, mask, value);
2587 if (nl_proto_down) {
2588 proto_down = nla_get_u8(nl_proto_down);
2590 /* Don't turn off protodown if there are active reasons */
2591 if (!proto_down && dev->proto_down_reason) {
2592 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2595 err = dev_change_proto_down(dev,
2604 #define DO_SETLINK_MODIFIED 0x01
2605 /* notify flag means notify + modified. */
2606 #define DO_SETLINK_NOTIFY 0x03
2607 static int do_setlink(const struct sk_buff *skb,
2608 struct net_device *dev, struct ifinfomsg *ifm,
2609 struct netlink_ext_ack *extack,
2610 struct nlattr **tb, char *ifname, int status)
2612 const struct net_device_ops *ops = dev->netdev_ops;
2615 err = validate_linkmsg(dev, tb, extack);
2619 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2620 const char *pat = ifname && ifname[0] ? ifname : NULL;
2624 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2631 if (tb[IFLA_NEW_IFINDEX])
2632 new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2636 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2640 status |= DO_SETLINK_MODIFIED;
2644 struct rtnl_link_ifmap *u_map;
2647 if (!ops->ndo_set_config) {
2652 if (!netif_device_present(dev)) {
2657 u_map = nla_data(tb[IFLA_MAP]);
2658 k_map.mem_start = (unsigned long) u_map->mem_start;
2659 k_map.mem_end = (unsigned long) u_map->mem_end;
2660 k_map.base_addr = (unsigned short) u_map->base_addr;
2661 k_map.irq = (unsigned char) u_map->irq;
2662 k_map.dma = (unsigned char) u_map->dma;
2663 k_map.port = (unsigned char) u_map->port;
2665 err = ops->ndo_set_config(dev, &k_map);
2669 status |= DO_SETLINK_NOTIFY;
2672 if (tb[IFLA_ADDRESS]) {
2673 struct sockaddr *sa;
2676 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2678 sa = kmalloc(len, GFP_KERNEL);
2683 sa->sa_family = dev->type;
2684 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2686 err = dev_set_mac_address_user(dev, sa, extack);
2690 status |= DO_SETLINK_MODIFIED;
2694 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2697 status |= DO_SETLINK_MODIFIED;
2700 if (tb[IFLA_GROUP]) {
2701 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2702 status |= DO_SETLINK_NOTIFY;
2706 * Interface selected by interface index but interface
2707 * name provided implies that a name change has been
2710 if (ifm->ifi_index > 0 && ifname[0]) {
2711 err = dev_change_name(dev, ifname);
2714 status |= DO_SETLINK_MODIFIED;
2717 if (tb[IFLA_IFALIAS]) {
2718 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2719 nla_len(tb[IFLA_IFALIAS]));
2722 status |= DO_SETLINK_NOTIFY;
2725 if (tb[IFLA_BROADCAST]) {
2726 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2727 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2730 if (ifm->ifi_flags || ifm->ifi_change) {
2731 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2737 if (tb[IFLA_MASTER]) {
2738 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2741 status |= DO_SETLINK_MODIFIED;
2744 if (tb[IFLA_CARRIER]) {
2745 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2748 status |= DO_SETLINK_MODIFIED;
2751 if (tb[IFLA_TXQLEN]) {
2752 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2754 err = dev_change_tx_queue_len(dev, value);
2757 status |= DO_SETLINK_MODIFIED;
2760 if (tb[IFLA_GSO_MAX_SIZE]) {
2761 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2763 if (max_size > GSO_MAX_SIZE) {
2768 if (dev->gso_max_size ^ max_size) {
2769 netif_set_gso_max_size(dev, max_size);
2770 status |= DO_SETLINK_MODIFIED;
2774 if (tb[IFLA_GSO_MAX_SEGS]) {
2775 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2777 if (max_segs > GSO_MAX_SEGS) {
2782 if (dev->gso_max_segs ^ max_segs) {
2783 netif_set_gso_max_segs(dev, max_segs);
2784 status |= DO_SETLINK_MODIFIED;
2788 if (tb[IFLA_GRO_MAX_SIZE]) {
2789 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2791 if (dev->gro_max_size ^ gro_max_size) {
2792 netif_set_gro_max_size(dev, gro_max_size);
2793 status |= DO_SETLINK_MODIFIED;
2797 if (tb[IFLA_OPERSTATE])
2798 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2800 if (tb[IFLA_LINKMODE]) {
2801 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2803 write_lock(&dev_base_lock);
2804 if (dev->link_mode ^ value)
2805 status |= DO_SETLINK_NOTIFY;
2806 dev->link_mode = value;
2807 write_unlock(&dev_base_lock);
2810 if (tb[IFLA_VFINFO_LIST]) {
2811 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2812 struct nlattr *attr;
2815 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2816 if (nla_type(attr) != IFLA_VF_INFO ||
2817 nla_len(attr) < NLA_HDRLEN) {
2821 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2827 err = do_setvfinfo(dev, vfinfo);
2830 status |= DO_SETLINK_NOTIFY;
2835 if (tb[IFLA_VF_PORTS]) {
2836 struct nlattr *port[IFLA_PORT_MAX+1];
2837 struct nlattr *attr;
2842 if (!ops->ndo_set_vf_port)
2845 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2846 if (nla_type(attr) != IFLA_VF_PORT ||
2847 nla_len(attr) < NLA_HDRLEN) {
2851 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2857 if (!port[IFLA_PORT_VF]) {
2861 vf = nla_get_u32(port[IFLA_PORT_VF]);
2862 err = ops->ndo_set_vf_port(dev, vf, port);
2865 status |= DO_SETLINK_NOTIFY;
2870 if (tb[IFLA_PORT_SELF]) {
2871 struct nlattr *port[IFLA_PORT_MAX+1];
2873 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2875 ifla_port_policy, NULL);
2880 if (ops->ndo_set_vf_port)
2881 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2884 status |= DO_SETLINK_NOTIFY;
2887 if (tb[IFLA_AF_SPEC]) {
2891 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2892 const struct rtnl_af_ops *af_ops;
2894 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2896 err = af_ops->set_link_af(dev, af, extack);
2900 status |= DO_SETLINK_NOTIFY;
2905 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2906 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2907 tb[IFLA_PROTO_DOWN_REASON], extack);
2910 status |= DO_SETLINK_NOTIFY;
2914 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2917 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2919 ifla_xdp_policy, NULL);
2923 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2928 if (xdp[IFLA_XDP_FLAGS]) {
2929 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2930 if (xdp_flags & ~XDP_FLAGS_MASK) {
2934 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2940 if (xdp[IFLA_XDP_FD]) {
2941 int expected_fd = -1;
2943 if (xdp_flags & XDP_FLAGS_REPLACE) {
2944 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2949 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2952 err = dev_change_xdp_fd(dev, extack,
2953 nla_get_s32(xdp[IFLA_XDP_FD]),
2958 status |= DO_SETLINK_NOTIFY;
2963 if (status & DO_SETLINK_MODIFIED) {
2964 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2965 netdev_state_change(dev);
2968 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",
2975 static struct net_device *rtnl_dev_get(struct net *net,
2976 struct nlattr *ifname_attr,
2977 struct nlattr *altifname_attr,
2980 char buffer[ALTIFNAMSIZ];
2985 nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2986 else if (altifname_attr)
2987 nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2992 return __dev_get_by_name(net, ifname);
2995 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2996 struct netlink_ext_ack *extack)
2998 struct net *net = sock_net(skb->sk);
2999 struct ifinfomsg *ifm;
3000 struct net_device *dev;
3002 struct nlattr *tb[IFLA_MAX+1];
3003 char ifname[IFNAMSIZ];
3005 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3006 ifla_policy, extack);
3010 err = rtnl_ensure_unique_netns(tb, extack, false);
3014 if (tb[IFLA_IFNAME])
3015 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3020 ifm = nlmsg_data(nlh);
3021 if (ifm->ifi_index > 0)
3022 dev = __dev_get_by_index(net, ifm->ifi_index);
3023 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3024 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3033 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3038 static int rtnl_group_dellink(const struct net *net, int group)
3040 struct net_device *dev, *aux;
3041 LIST_HEAD(list_kill);
3047 for_each_netdev(net, dev) {
3048 if (dev->group == group) {
3049 const struct rtnl_link_ops *ops;
3052 ops = dev->rtnl_link_ops;
3053 if (!ops || !ops->dellink)
3061 for_each_netdev_safe(net, dev, aux) {
3062 if (dev->group == group) {
3063 const struct rtnl_link_ops *ops;
3065 ops = dev->rtnl_link_ops;
3066 ops->dellink(dev, &list_kill);
3069 unregister_netdevice_many(&list_kill);
3074 int rtnl_delete_link(struct net_device *dev)
3076 const struct rtnl_link_ops *ops;
3077 LIST_HEAD(list_kill);
3079 ops = dev->rtnl_link_ops;
3080 if (!ops || !ops->dellink)
3083 ops->dellink(dev, &list_kill);
3084 unregister_netdevice_many(&list_kill);
3088 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3090 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3091 struct netlink_ext_ack *extack)
3093 struct net *net = sock_net(skb->sk);
3094 struct net *tgt_net = net;
3095 struct net_device *dev = NULL;
3096 struct ifinfomsg *ifm;
3097 struct nlattr *tb[IFLA_MAX+1];
3101 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3102 ifla_policy, extack);
3106 err = rtnl_ensure_unique_netns(tb, extack, true);
3110 if (tb[IFLA_TARGET_NETNSID]) {
3111 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3112 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3113 if (IS_ERR(tgt_net))
3114 return PTR_ERR(tgt_net);
3118 ifm = nlmsg_data(nlh);
3119 if (ifm->ifi_index > 0)
3120 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3121 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3122 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3123 tb[IFLA_ALT_IFNAME], NULL);
3124 else if (tb[IFLA_GROUP])
3125 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3130 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3136 err = rtnl_delete_link(dev);
3145 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3147 unsigned int old_flags;
3150 old_flags = dev->flags;
3151 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3152 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3158 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3159 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3161 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3162 __dev_notify_flags(dev, old_flags, ~0U);
3166 EXPORT_SYMBOL(rtnl_configure_link);
3168 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3169 unsigned char name_assign_type,
3170 const struct rtnl_link_ops *ops,
3171 struct nlattr *tb[],
3172 struct netlink_ext_ack *extack)
3174 struct net_device *dev;
3175 unsigned int num_tx_queues = 1;
3176 unsigned int num_rx_queues = 1;
3178 if (tb[IFLA_NUM_TX_QUEUES])
3179 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3180 else if (ops->get_num_tx_queues)
3181 num_tx_queues = ops->get_num_tx_queues();
3183 if (tb[IFLA_NUM_RX_QUEUES])
3184 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3185 else if (ops->get_num_rx_queues)
3186 num_rx_queues = ops->get_num_rx_queues();
3188 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3189 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3190 return ERR_PTR(-EINVAL);
3193 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3194 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3195 return ERR_PTR(-EINVAL);
3199 dev = ops->alloc(tb, ifname, name_assign_type,
3200 num_tx_queues, num_rx_queues);
3204 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3205 name_assign_type, ops->setup,
3206 num_tx_queues, num_rx_queues);
3210 return ERR_PTR(-ENOMEM);
3212 dev_net_set(dev, net);
3213 dev->rtnl_link_ops = ops;
3214 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3217 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3220 err = dev_validate_mtu(dev, mtu, extack);
3223 return ERR_PTR(err);
3227 if (tb[IFLA_ADDRESS]) {
3228 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3229 nla_len(tb[IFLA_ADDRESS]));
3230 dev->addr_assign_type = NET_ADDR_SET;
3232 if (tb[IFLA_BROADCAST])
3233 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3234 nla_len(tb[IFLA_BROADCAST]));
3235 if (tb[IFLA_TXQLEN])
3236 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3237 if (tb[IFLA_OPERSTATE])
3238 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3239 if (tb[IFLA_LINKMODE])
3240 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3242 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3243 if (tb[IFLA_GSO_MAX_SIZE])
3244 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3245 if (tb[IFLA_GSO_MAX_SEGS])
3246 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3247 if (tb[IFLA_GRO_MAX_SIZE])
3248 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3252 EXPORT_SYMBOL(rtnl_create_link);
3254 static int rtnl_group_changelink(const struct sk_buff *skb,
3255 struct net *net, int group,
3256 struct ifinfomsg *ifm,
3257 struct netlink_ext_ack *extack,
3260 struct net_device *dev, *aux;
3263 for_each_netdev_safe(net, dev, aux) {
3264 if (dev->group == group) {
3265 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3274 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3275 struct nlattr **attr, struct netlink_ext_ack *extack)
3277 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3278 unsigned char name_assign_type = NET_NAME_USER;
3279 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3280 const struct rtnl_link_ops *m_ops;
3281 struct net_device *master_dev;
3282 struct net *net = sock_net(skb->sk);
3283 const struct rtnl_link_ops *ops;
3284 struct nlattr *tb[IFLA_MAX + 1];
3285 struct net *dest_net, *link_net;
3286 struct nlattr **slave_data;
3287 char kind[MODULE_NAME_LEN];
3288 struct net_device *dev;
3289 struct ifinfomsg *ifm;
3290 char ifname[IFNAMSIZ];
3291 struct nlattr **data;
3294 #ifdef CONFIG_MODULES
3297 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3298 ifla_policy, extack);
3302 err = rtnl_ensure_unique_netns(tb, extack, false);
3306 if (tb[IFLA_IFNAME])
3307 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3311 ifm = nlmsg_data(nlh);
3312 if (ifm->ifi_index > 0)
3313 dev = __dev_get_by_index(net, ifm->ifi_index);
3314 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3315 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3322 master_dev = netdev_master_upper_dev_get(dev);
3324 m_ops = master_dev->rtnl_link_ops;
3327 err = validate_linkmsg(dev, tb, extack);
3331 if (tb[IFLA_LINKINFO]) {
3332 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3334 ifla_info_policy, NULL);
3338 memset(linkinfo, 0, sizeof(linkinfo));
3340 if (linkinfo[IFLA_INFO_KIND]) {
3341 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3342 ops = rtnl_link_ops_get(kind);
3350 if (ops->maxtype > RTNL_MAX_TYPE)
3353 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3354 err = nla_parse_nested_deprecated(attr, ops->maxtype,
3355 linkinfo[IFLA_INFO_DATA],
3356 ops->policy, extack);
3361 if (ops->validate) {
3362 err = ops->validate(tb, data, extack);
3370 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3373 if (m_ops->slave_maxtype &&
3374 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3375 err = nla_parse_nested_deprecated(slave_attr,
3376 m_ops->slave_maxtype,
3377 linkinfo[IFLA_INFO_SLAVE_DATA],
3378 m_ops->slave_policy,
3382 slave_data = slave_attr;
3389 if (nlh->nlmsg_flags & NLM_F_EXCL)
3391 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3394 if (linkinfo[IFLA_INFO_DATA]) {
3395 if (!ops || ops != dev->rtnl_link_ops ||
3399 err = ops->changelink(dev, tb, data, extack);
3402 status |= DO_SETLINK_NOTIFY;
3405 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3406 if (!m_ops || !m_ops->slave_changelink)
3409 err = m_ops->slave_changelink(master_dev, dev, tb,
3410 slave_data, extack);
3413 status |= DO_SETLINK_NOTIFY;
3416 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3419 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3420 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3421 return rtnl_group_changelink(skb, net,
3422 nla_get_u32(tb[IFLA_GROUP]),
3427 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3431 #ifdef CONFIG_MODULES
3434 request_module("rtnl-link-%s", kind);
3436 ops = rtnl_link_ops_get(kind);
3441 NL_SET_ERR_MSG(extack, "Unknown device type");
3445 if (!ops->alloc && !ops->setup)
3449 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3450 name_assign_type = NET_NAME_ENUM;
3453 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3454 if (IS_ERR(dest_net))
3455 return PTR_ERR(dest_net);
3457 if (tb[IFLA_LINK_NETNSID]) {
3458 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3460 link_net = get_net_ns_by_id(dest_net, id);
3462 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3467 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3473 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3474 name_assign_type, ops, tb, extack);
3480 dev->ifindex = ifm->ifi_index;
3483 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3485 err = register_netdevice(dev);
3491 err = rtnl_configure_link(dev, ifm);
3493 goto out_unregister;
3495 err = dev_change_net_namespace(dev, dest_net, ifname);
3497 goto out_unregister;
3499 if (tb[IFLA_MASTER]) {
3500 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3502 goto out_unregister;
3511 LIST_HEAD(list_kill);
3513 ops->dellink(dev, &list_kill);
3514 unregister_netdevice_many(&list_kill);
3516 unregister_netdevice(dev);
3521 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3522 struct netlink_ext_ack *extack)
3524 struct nlattr **attr;
3527 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3531 ret = __rtnl_newlink(skb, nlh, attr, extack);
3536 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3537 const struct nlmsghdr *nlh,
3539 struct netlink_ext_ack *extack)
3541 struct ifinfomsg *ifm;
3544 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3545 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3549 if (!netlink_strict_get_check(skb))
3550 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3551 ifla_policy, extack);
3553 ifm = nlmsg_data(nlh);
3554 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3556 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3560 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3561 ifla_policy, extack);
3565 for (i = 0; i <= IFLA_MAX; i++) {
3571 case IFLA_ALT_IFNAME:
3573 case IFLA_TARGET_NETNSID:
3576 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3584 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3585 struct netlink_ext_ack *extack)
3587 struct net *net = sock_net(skb->sk);
3588 struct net *tgt_net = net;
3589 struct ifinfomsg *ifm;
3590 struct nlattr *tb[IFLA_MAX+1];
3591 struct net_device *dev = NULL;
3592 struct sk_buff *nskb;
3595 u32 ext_filter_mask = 0;
3597 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3601 err = rtnl_ensure_unique_netns(tb, extack, true);
3605 if (tb[IFLA_TARGET_NETNSID]) {
3606 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3607 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3608 if (IS_ERR(tgt_net))
3609 return PTR_ERR(tgt_net);
3612 if (tb[IFLA_EXT_MASK])
3613 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3616 ifm = nlmsg_data(nlh);
3617 if (ifm->ifi_index > 0)
3618 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3619 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3620 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3621 tb[IFLA_ALT_IFNAME], NULL);
3630 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3634 err = rtnl_fill_ifinfo(nskb, dev, net,
3635 RTM_NEWLINK, NETLINK_CB(skb).portid,
3636 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3637 0, NULL, 0, netnsid, GFP_KERNEL);
3639 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3640 WARN_ON(err == -EMSGSIZE);
3643 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3651 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3652 bool *changed, struct netlink_ext_ack *extack)
3657 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3661 alt_ifname = nla_strdup(attr, GFP_KERNEL);
3665 if (cmd == RTM_NEWLINKPROP) {
3666 err = netdev_name_node_alt_create(dev, alt_ifname);
3669 } else if (cmd == RTM_DELLINKPROP) {
3670 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3682 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3683 struct netlink_ext_ack *extack)
3685 struct net *net = sock_net(skb->sk);
3686 struct nlattr *tb[IFLA_MAX + 1];
3687 struct net_device *dev;
3688 struct ifinfomsg *ifm;
3689 bool changed = false;
3690 struct nlattr *attr;
3693 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3697 err = rtnl_ensure_unique_netns(tb, extack, true);
3701 ifm = nlmsg_data(nlh);
3702 if (ifm->ifi_index > 0)
3703 dev = __dev_get_by_index(net, ifm->ifi_index);
3704 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3705 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3706 tb[IFLA_ALT_IFNAME], NULL);
3713 if (!tb[IFLA_PROP_LIST])
3716 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3717 switch (nla_type(attr)) {
3718 case IFLA_ALT_IFNAME:
3719 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3727 netdev_state_change(dev);
3731 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3732 struct netlink_ext_ack *extack)
3734 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3737 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3738 struct netlink_ext_ack *extack)
3740 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3743 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3745 struct net *net = sock_net(skb->sk);
3746 size_t min_ifinfo_dump_size = 0;
3747 struct nlattr *tb[IFLA_MAX+1];
3748 u32 ext_filter_mask = 0;
3749 struct net_device *dev;
3752 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3753 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3754 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3756 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3757 if (tb[IFLA_EXT_MASK])
3758 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3761 if (!ext_filter_mask)
3762 return NLMSG_GOODSIZE;
3764 * traverse the list of net devices and compute the minimum
3765 * buffer size based upon the filter mask.
3768 for_each_netdev_rcu(net, dev) {
3769 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3770 if_nlmsg_size(dev, ext_filter_mask));
3774 return nlmsg_total_size(min_ifinfo_dump_size);
3777 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3780 int s_idx = cb->family;
3781 int type = cb->nlh->nlmsg_type - RTM_BASE;
3787 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3788 struct rtnl_link __rcu **tab;
3789 struct rtnl_link *link;
3790 rtnl_dumpit_func dumpit;
3792 if (idx < s_idx || idx == PF_PACKET)
3795 if (type < 0 || type >= RTM_NR_MSGTYPES)
3798 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3802 link = rcu_dereference_rtnl(tab[type]);
3806 dumpit = link->dumpit;
3811 memset(&cb->args[0], 0, sizeof(cb->args));
3815 ret = dumpit(skb, cb);
3821 return skb->len ? : ret;
3824 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3825 unsigned int change,
3826 u32 event, gfp_t flags, int *new_nsid,
3829 struct net *net = dev_net(dev);
3830 struct sk_buff *skb;
3833 skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
3837 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3838 type, 0, 0, change, 0, 0, event,
3839 new_nsid, new_ifindex, -1, flags);
3841 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3842 WARN_ON(err == -EMSGSIZE);
3849 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3853 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3855 struct net *net = dev_net(dev);
3857 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3860 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3861 unsigned int change, u32 event,
3862 gfp_t flags, int *new_nsid, int new_ifindex)
3864 struct sk_buff *skb;
3866 if (dev->reg_state != NETREG_REGISTERED)
3869 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3872 rtmsg_ifinfo_send(skb, dev, flags);
3875 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3878 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3882 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3883 gfp_t flags, int *new_nsid, int new_ifindex)
3885 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3886 new_nsid, new_ifindex);
3889 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3890 struct net_device *dev,
3891 u8 *addr, u16 vid, u32 pid, u32 seq,
3892 int type, unsigned int flags,
3893 int nlflags, u16 ndm_state)
3895 struct nlmsghdr *nlh;
3898 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3902 ndm = nlmsg_data(nlh);
3903 ndm->ndm_family = AF_BRIDGE;
3906 ndm->ndm_flags = flags;
3908 ndm->ndm_ifindex = dev->ifindex;
3909 ndm->ndm_state = ndm_state;
3911 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3912 goto nla_put_failure;
3914 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3915 goto nla_put_failure;
3917 nlmsg_end(skb, nlh);
3921 nlmsg_cancel(skb, nlh);
3925 static inline size_t rtnl_fdb_nlmsg_size(void)
3927 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3928 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3929 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3933 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3936 struct net *net = dev_net(dev);
3937 struct sk_buff *skb;
3940 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3944 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3945 0, 0, type, NTF_SELF, 0, ndm_state);
3951 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3954 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3958 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3960 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3961 struct nlattr *tb[],
3962 struct net_device *dev,
3963 const unsigned char *addr, u16 vid,
3968 /* If aging addresses are supported device will need to
3969 * implement its own handler for this.
3971 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3972 netdev_info(dev, "default FDB implementation only supports local addresses\n");
3977 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
3981 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3982 err = dev_uc_add_excl(dev, addr);
3983 else if (is_multicast_ether_addr(addr))
3984 err = dev_mc_add_excl(dev, addr);
3986 /* Only return duplicate errors if NLM_F_EXCL is set */
3987 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3992 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3994 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3995 struct netlink_ext_ack *extack)
4000 if (nla_len(vlan_attr) != sizeof(u16)) {
4001 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4005 vid = nla_get_u16(vlan_attr);
4007 if (!vid || vid >= VLAN_VID_MASK) {
4008 NL_SET_ERR_MSG(extack, "invalid vlan id");
4016 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4017 struct netlink_ext_ack *extack)
4019 struct net *net = sock_net(skb->sk);
4021 struct nlattr *tb[NDA_MAX+1];
4022 struct net_device *dev;
4027 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4032 ndm = nlmsg_data(nlh);
4033 if (ndm->ndm_ifindex == 0) {
4034 NL_SET_ERR_MSG(extack, "invalid ifindex");
4038 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4040 NL_SET_ERR_MSG(extack, "unknown ifindex");
4044 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4045 NL_SET_ERR_MSG(extack, "invalid address");
4049 if (dev->type != ARPHRD_ETHER) {
4050 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4054 addr = nla_data(tb[NDA_LLADDR]);
4056 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4062 /* Support fdb on master device the net/bridge default case */
4063 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4064 netif_is_bridge_port(dev)) {
4065 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4066 const struct net_device_ops *ops = br_dev->netdev_ops;
4068 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4069 nlh->nlmsg_flags, extack);
4073 ndm->ndm_flags &= ~NTF_MASTER;
4076 /* Embedded bridge, macvlan, and any other device support */
4077 if ((ndm->ndm_flags & NTF_SELF)) {
4078 if (dev->netdev_ops->ndo_fdb_add)
4079 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4084 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4088 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4090 ndm->ndm_flags &= ~NTF_SELF;
4098 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4100 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4101 struct nlattr *tb[],
4102 struct net_device *dev,
4103 const unsigned char *addr, u16 vid)
4107 /* If aging addresses are supported device will need to
4108 * implement its own handler for this.
4110 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4111 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4115 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4116 err = dev_uc_del(dev, addr);
4117 else if (is_multicast_ether_addr(addr))
4118 err = dev_mc_del(dev, addr);
4122 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4124 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4125 struct netlink_ext_ack *extack)
4127 struct net *net = sock_net(skb->sk);
4129 struct nlattr *tb[NDA_MAX+1];
4130 struct net_device *dev;
4135 if (!netlink_capable(skb, CAP_NET_ADMIN))
4138 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4143 ndm = nlmsg_data(nlh);
4144 if (ndm->ndm_ifindex == 0) {
4145 NL_SET_ERR_MSG(extack, "invalid ifindex");
4149 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4151 NL_SET_ERR_MSG(extack, "unknown ifindex");
4155 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4156 NL_SET_ERR_MSG(extack, "invalid address");
4160 if (dev->type != ARPHRD_ETHER) {
4161 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4165 addr = nla_data(tb[NDA_LLADDR]);
4167 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4173 /* Support fdb on master device the net/bridge default case */
4174 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4175 netif_is_bridge_port(dev)) {
4176 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4177 const struct net_device_ops *ops = br_dev->netdev_ops;
4179 if (ops->ndo_fdb_del)
4180 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4185 ndm->ndm_flags &= ~NTF_MASTER;
4188 /* Embedded bridge, macvlan, and any other device support */
4189 if (ndm->ndm_flags & NTF_SELF) {
4190 if (dev->netdev_ops->ndo_fdb_del)
4191 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4194 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4197 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4199 ndm->ndm_flags &= ~NTF_SELF;
4206 static int nlmsg_populate_fdb(struct sk_buff *skb,
4207 struct netlink_callback *cb,
4208 struct net_device *dev,
4210 struct netdev_hw_addr_list *list)
4212 struct netdev_hw_addr *ha;
4216 portid = NETLINK_CB(cb->skb).portid;
4217 seq = cb->nlh->nlmsg_seq;
4219 list_for_each_entry(ha, &list->list, list) {
4220 if (*idx < cb->args[2])
4223 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4225 RTM_NEWNEIGH, NTF_SELF,
4226 NLM_F_MULTI, NUD_PERMANENT);
4236 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4237 * @skb: socket buffer to store message in
4238 * @cb: netlink callback
4240 * @filter_dev: ignored
4241 * @idx: the number of FDB table entries dumped is added to *@idx
4243 * Default netdevice operation to dump the existing unicast address list.
4244 * Returns number of addresses from list put in skb.
4246 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4247 struct netlink_callback *cb,
4248 struct net_device *dev,
4249 struct net_device *filter_dev,
4254 if (dev->type != ARPHRD_ETHER)
4257 netif_addr_lock_bh(dev);
4258 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4261 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4263 netif_addr_unlock_bh(dev);
4266 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4268 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4269 int *br_idx, int *brport_idx,
4270 struct netlink_ext_ack *extack)
4272 struct nlattr *tb[NDA_MAX + 1];
4276 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4277 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4281 ndm = nlmsg_data(nlh);
4282 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4283 ndm->ndm_flags || ndm->ndm_type) {
4284 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4288 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4289 NDA_MAX, NULL, extack);
4293 *brport_idx = ndm->ndm_ifindex;
4294 for (i = 0; i <= NDA_MAX; ++i) {
4300 if (nla_len(tb[i]) != sizeof(u32)) {
4301 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4304 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4307 if (nla_len(tb[i]) != sizeof(u32)) {
4308 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4311 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4314 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4322 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4323 int *br_idx, int *brport_idx,
4324 struct netlink_ext_ack *extack)
4326 struct nlattr *tb[IFLA_MAX+1];
4329 /* A hack to preserve kernel<->userspace interface.
4330 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4331 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4332 * So, check for ndmsg with an optional u32 attribute (not used here).
4333 * Fortunately these sizes don't conflict with the size of ifinfomsg
4334 * with an optional attribute.
4336 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4337 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4338 nla_attr_size(sizeof(u32)))) {
4339 struct ifinfomsg *ifm;
4341 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4342 tb, IFLA_MAX, ifla_policy,
4346 } else if (err == 0) {
4347 if (tb[IFLA_MASTER])
4348 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4351 ifm = nlmsg_data(nlh);
4352 *brport_idx = ifm->ifi_index;
4357 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4359 struct net_device *dev;
4360 struct net_device *br_dev = NULL;
4361 const struct net_device_ops *ops = NULL;
4362 const struct net_device_ops *cops = NULL;
4363 struct net *net = sock_net(skb->sk);
4364 struct hlist_head *head;
4372 if (cb->strict_check)
4373 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4376 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4382 br_dev = __dev_get_by_index(net, br_idx);
4386 ops = br_dev->netdev_ops;
4390 s_idx = cb->args[1];
4392 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4394 head = &net->dev_index_head[h];
4395 hlist_for_each_entry(dev, head, index_hlist) {
4397 if (brport_idx && (dev->ifindex != brport_idx))
4400 if (!br_idx) { /* user did not specify a specific bridge */
4401 if (netif_is_bridge_port(dev)) {
4402 br_dev = netdev_master_upper_dev_get(dev);
4403 cops = br_dev->netdev_ops;
4406 if (dev != br_dev &&
4407 !netif_is_bridge_port(dev))
4410 if (br_dev != netdev_master_upper_dev_get(dev) &&
4411 !netif_is_bridge_master(dev))
4419 if (netif_is_bridge_port(dev)) {
4420 if (cops && cops->ndo_fdb_dump) {
4421 err = cops->ndo_fdb_dump(skb, cb,
4424 if (err == -EMSGSIZE)
4429 if (dev->netdev_ops->ndo_fdb_dump)
4430 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4434 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4436 if (err == -EMSGSIZE)
4441 /* reset fdb offset to 0 for rest of the interfaces */
4457 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4458 struct nlattr **tb, u8 *ndm_flags,
4459 int *br_idx, int *brport_idx, u8 **addr,
4460 u16 *vid, struct netlink_ext_ack *extack)
4465 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4466 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4470 ndm = nlmsg_data(nlh);
4471 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4473 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4477 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4478 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4482 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4483 NDA_MAX, nda_policy, extack);
4487 *ndm_flags = ndm->ndm_flags;
4488 *brport_idx = ndm->ndm_ifindex;
4489 for (i = 0; i <= NDA_MAX; ++i) {
4495 *br_idx = nla_get_u32(tb[i]);
4498 if (nla_len(tb[i]) != ETH_ALEN) {
4499 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4502 *addr = nla_data(tb[i]);
4505 err = fdb_vid_parse(tb[i], vid, extack);
4512 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4520 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4521 struct netlink_ext_ack *extack)
4523 struct net_device *dev = NULL, *br_dev = NULL;
4524 const struct net_device_ops *ops = NULL;
4525 struct net *net = sock_net(in_skb->sk);
4526 struct nlattr *tb[NDA_MAX + 1];
4527 struct sk_buff *skb;
4535 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4536 &brport_idx, &addr, &vid, extack);
4541 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4546 dev = __dev_get_by_index(net, brport_idx);
4548 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4555 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4559 br_dev = __dev_get_by_index(net, br_idx);
4561 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4564 ops = br_dev->netdev_ops;
4568 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4569 if (!netif_is_bridge_port(dev)) {
4570 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4573 br_dev = netdev_master_upper_dev_get(dev);
4575 NL_SET_ERR_MSG(extack, "Master of device not found");
4578 ops = br_dev->netdev_ops;
4580 if (!(ndm_flags & NTF_SELF)) {
4581 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4584 ops = dev->netdev_ops;
4588 if (!br_dev && !dev) {
4589 NL_SET_ERR_MSG(extack, "No device specified");
4593 if (!ops || !ops->ndo_fdb_get) {
4594 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4598 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4604 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4605 NETLINK_CB(in_skb).portid,
4606 nlh->nlmsg_seq, extack);
4610 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4616 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4617 unsigned int attrnum, unsigned int flag)
4620 return nla_put_u8(skb, attrnum, !!(flags & flag));
4624 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4625 struct net_device *dev, u16 mode,
4626 u32 flags, u32 mask, int nlflags,
4628 int (*vlan_fill)(struct sk_buff *skb,
4629 struct net_device *dev,
4632 struct nlmsghdr *nlh;
4633 struct ifinfomsg *ifm;
4634 struct nlattr *br_afspec;
4635 struct nlattr *protinfo;
4636 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4637 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4640 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4644 ifm = nlmsg_data(nlh);
4645 ifm->ifi_family = AF_BRIDGE;
4647 ifm->ifi_type = dev->type;
4648 ifm->ifi_index = dev->ifindex;
4649 ifm->ifi_flags = dev_get_flags(dev);
4650 ifm->ifi_change = 0;
4653 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4654 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4655 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4657 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4659 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4660 (dev->ifindex != dev_get_iflink(dev) &&
4661 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4662 goto nla_put_failure;
4664 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4666 goto nla_put_failure;
4668 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4669 nla_nest_cancel(skb, br_afspec);
4670 goto nla_put_failure;
4673 if (mode != BRIDGE_MODE_UNDEF) {
4674 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4675 nla_nest_cancel(skb, br_afspec);
4676 goto nla_put_failure;
4680 err = vlan_fill(skb, dev, filter_mask);
4682 nla_nest_cancel(skb, br_afspec);
4683 goto nla_put_failure;
4686 nla_nest_end(skb, br_afspec);
4688 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4690 goto nla_put_failure;
4692 if (brport_nla_put_flag(skb, flags, mask,
4693 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4694 brport_nla_put_flag(skb, flags, mask,
4695 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4696 brport_nla_put_flag(skb, flags, mask,
4697 IFLA_BRPORT_FAST_LEAVE,
4698 BR_MULTICAST_FAST_LEAVE) ||
4699 brport_nla_put_flag(skb, flags, mask,
4700 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4701 brport_nla_put_flag(skb, flags, mask,
4702 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4703 brport_nla_put_flag(skb, flags, mask,
4704 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4705 brport_nla_put_flag(skb, flags, mask,
4706 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4707 brport_nla_put_flag(skb, flags, mask,
4708 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4709 brport_nla_put_flag(skb, flags, mask,
4710 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4711 brport_nla_put_flag(skb, flags, mask,
4712 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4713 nla_nest_cancel(skb, protinfo);
4714 goto nla_put_failure;
4717 nla_nest_end(skb, protinfo);
4719 nlmsg_end(skb, nlh);
4722 nlmsg_cancel(skb, nlh);
4723 return err ? err : -EMSGSIZE;
4725 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4727 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4728 bool strict_check, u32 *filter_mask,
4729 struct netlink_ext_ack *extack)
4731 struct nlattr *tb[IFLA_MAX+1];
4735 struct ifinfomsg *ifm;
4737 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4738 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4742 ifm = nlmsg_data(nlh);
4743 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4744 ifm->ifi_change || ifm->ifi_index) {
4745 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4749 err = nlmsg_parse_deprecated_strict(nlh,
4750 sizeof(struct ifinfomsg),
4751 tb, IFLA_MAX, ifla_policy,
4754 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4755 tb, IFLA_MAX, ifla_policy,
4761 /* new attributes should only be added with strict checking */
4762 for (i = 0; i <= IFLA_MAX; ++i) {
4768 *filter_mask = nla_get_u32(tb[i]);
4772 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4781 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4783 const struct nlmsghdr *nlh = cb->nlh;
4784 struct net *net = sock_net(skb->sk);
4785 struct net_device *dev;
4787 u32 portid = NETLINK_CB(cb->skb).portid;
4788 u32 seq = nlh->nlmsg_seq;
4789 u32 filter_mask = 0;
4792 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4794 if (err < 0 && cb->strict_check)
4798 for_each_netdev_rcu(net, dev) {
4799 const struct net_device_ops *ops = dev->netdev_ops;
4800 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4802 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4803 if (idx >= cb->args[0]) {
4804 err = br_dev->netdev_ops->ndo_bridge_getlink(
4805 skb, portid, seq, dev,
4806 filter_mask, NLM_F_MULTI);
4807 if (err < 0 && err != -EOPNOTSUPP) {
4808 if (likely(skb->len))
4817 if (ops->ndo_bridge_getlink) {
4818 if (idx >= cb->args[0]) {
4819 err = ops->ndo_bridge_getlink(skb, portid,
4823 if (err < 0 && err != -EOPNOTSUPP) {
4824 if (likely(skb->len))
4841 static inline size_t bridge_nlmsg_size(void)
4843 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4844 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4845 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4846 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4847 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
4848 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
4849 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
4850 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
4851 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4852 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
4853 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
4856 static int rtnl_bridge_notify(struct net_device *dev)
4858 struct net *net = dev_net(dev);
4859 struct sk_buff *skb;
4860 int err = -EOPNOTSUPP;
4862 if (!dev->netdev_ops->ndo_bridge_getlink)
4865 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4871 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4875 /* Notification info is only filled for bridge ports, not the bridge
4876 * device itself. Therefore, a zero notification length is valid and
4877 * should not result in an error.
4882 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4885 WARN_ON(err == -EMSGSIZE);
4888 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4892 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4893 struct netlink_ext_ack *extack)
4895 struct net *net = sock_net(skb->sk);
4896 struct ifinfomsg *ifm;
4897 struct net_device *dev;
4898 struct nlattr *br_spec, *attr = NULL;
4899 int rem, err = -EOPNOTSUPP;
4901 bool have_flags = false;
4903 if (nlmsg_len(nlh) < sizeof(*ifm))
4906 ifm = nlmsg_data(nlh);
4907 if (ifm->ifi_family != AF_BRIDGE)
4908 return -EPFNOSUPPORT;
4910 dev = __dev_get_by_index(net, ifm->ifi_index);
4912 NL_SET_ERR_MSG(extack, "unknown ifindex");
4916 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4918 nla_for_each_nested(attr, br_spec, rem) {
4919 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4920 if (nla_len(attr) < sizeof(flags))
4924 flags = nla_get_u16(attr);
4930 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4931 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4933 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4938 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4943 flags &= ~BRIDGE_FLAGS_MASTER;
4946 if ((flags & BRIDGE_FLAGS_SELF)) {
4947 if (!dev->netdev_ops->ndo_bridge_setlink)
4950 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4954 flags &= ~BRIDGE_FLAGS_SELF;
4956 /* Generate event to notify upper layer of bridge
4959 err = rtnl_bridge_notify(dev);
4964 memcpy(nla_data(attr), &flags, sizeof(flags));
4969 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4970 struct netlink_ext_ack *extack)
4972 struct net *net = sock_net(skb->sk);
4973 struct ifinfomsg *ifm;
4974 struct net_device *dev;
4975 struct nlattr *br_spec, *attr = NULL;
4976 int rem, err = -EOPNOTSUPP;
4978 bool have_flags = false;
4980 if (nlmsg_len(nlh) < sizeof(*ifm))
4983 ifm = nlmsg_data(nlh);
4984 if (ifm->ifi_family != AF_BRIDGE)
4985 return -EPFNOSUPPORT;
4987 dev = __dev_get_by_index(net, ifm->ifi_index);
4989 NL_SET_ERR_MSG(extack, "unknown ifindex");
4993 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4995 nla_for_each_nested(attr, br_spec, rem) {
4996 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4997 if (nla_len(attr) < sizeof(flags))
5001 flags = nla_get_u16(attr);
5007 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5008 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5010 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5015 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5019 flags &= ~BRIDGE_FLAGS_MASTER;
5022 if ((flags & BRIDGE_FLAGS_SELF)) {
5023 if (!dev->netdev_ops->ndo_bridge_dellink)
5026 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5030 flags &= ~BRIDGE_FLAGS_SELF;
5032 /* Generate event to notify upper layer of bridge
5035 err = rtnl_bridge_notify(dev);
5040 memcpy(nla_data(attr), &flags, sizeof(flags));
5045 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5047 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5048 (!idxattr || idxattr == attrid);
5051 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5052 static int rtnl_get_offload_stats_attr_size(int attr_id)
5055 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5056 return sizeof(struct rtnl_link_stats64);
5062 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5065 struct nlattr *attr = NULL;
5070 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5071 dev->netdev_ops->ndo_get_offload_stats))
5074 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5075 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5076 if (attr_id < *prividx)
5079 size = rtnl_get_offload_stats_attr_size(attr_id);
5083 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5086 attr = nla_reserve_64bit(skb, attr_id, size,
5087 IFLA_OFFLOAD_XSTATS_UNSPEC);
5089 goto nla_put_failure;
5091 attr_data = nla_data(attr);
5092 memset(attr_data, 0, size);
5093 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5096 goto get_offload_stats_failure;
5107 get_offload_stats_failure:
5112 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5118 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5119 dev->netdev_ops->ndo_get_offload_stats))
5122 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5123 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5124 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5126 size = rtnl_get_offload_stats_attr_size(attr_id);
5127 nla_size += nla_total_size_64bit(size);
5131 nla_size += nla_total_size(0);
5136 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5137 int type, u32 pid, u32 seq, u32 change,
5138 unsigned int flags, unsigned int filter_mask,
5139 int *idxattr, int *prividx)
5141 struct if_stats_msg *ifsm;
5142 struct nlmsghdr *nlh;
5143 struct nlattr *attr;
5144 int s_prividx = *prividx;
5149 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5153 ifsm = nlmsg_data(nlh);
5154 ifsm->family = PF_UNSPEC;
5157 ifsm->ifindex = dev->ifindex;
5158 ifsm->filter_mask = filter_mask;
5160 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5161 struct rtnl_link_stats64 *sp;
5163 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5164 sizeof(struct rtnl_link_stats64),
5167 goto nla_put_failure;
5169 sp = nla_data(attr);
5170 dev_get_stats(dev, sp);
5173 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5174 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5176 if (ops && ops->fill_linkxstats) {
5177 *idxattr = IFLA_STATS_LINK_XSTATS;
5178 attr = nla_nest_start_noflag(skb,
5179 IFLA_STATS_LINK_XSTATS);
5181 goto nla_put_failure;
5183 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5184 nla_nest_end(skb, attr);
5186 goto nla_put_failure;
5191 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5193 const struct rtnl_link_ops *ops = NULL;
5194 const struct net_device *master;
5196 master = netdev_master_upper_dev_get(dev);
5198 ops = master->rtnl_link_ops;
5199 if (ops && ops->fill_linkxstats) {
5200 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5201 attr = nla_nest_start_noflag(skb,
5202 IFLA_STATS_LINK_XSTATS_SLAVE);
5204 goto nla_put_failure;
5206 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5207 nla_nest_end(skb, attr);
5209 goto nla_put_failure;
5214 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5216 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5217 attr = nla_nest_start_noflag(skb,
5218 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5220 goto nla_put_failure;
5222 err = rtnl_get_offload_stats(skb, dev, prividx);
5223 if (err == -ENODATA)
5224 nla_nest_cancel(skb, attr);
5226 nla_nest_end(skb, attr);
5228 if (err && err != -ENODATA)
5229 goto nla_put_failure;
5233 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5234 struct rtnl_af_ops *af_ops;
5236 *idxattr = IFLA_STATS_AF_SPEC;
5237 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5239 goto nla_put_failure;
5242 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5243 if (af_ops->fill_stats_af) {
5247 af = nla_nest_start_noflag(skb,
5251 goto nla_put_failure;
5253 err = af_ops->fill_stats_af(skb, dev);
5255 if (err == -ENODATA) {
5256 nla_nest_cancel(skb, af);
5257 } else if (err < 0) {
5259 goto nla_put_failure;
5262 nla_nest_end(skb, af);
5267 nla_nest_end(skb, attr);
5272 nlmsg_end(skb, nlh);
5277 /* not a multi message or no progress mean a real error */
5278 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5279 nlmsg_cancel(skb, nlh);
5281 nlmsg_end(skb, nlh);
5286 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5289 size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5291 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5292 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5294 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5295 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5296 int attr = IFLA_STATS_LINK_XSTATS;
5298 if (ops && ops->get_linkxstats_size) {
5299 size += nla_total_size(ops->get_linkxstats_size(dev,
5301 /* for IFLA_STATS_LINK_XSTATS */
5302 size += nla_total_size(0);
5306 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5307 struct net_device *_dev = (struct net_device *)dev;
5308 const struct rtnl_link_ops *ops = NULL;
5309 const struct net_device *master;
5311 /* netdev_master_upper_dev_get can't take const */
5312 master = netdev_master_upper_dev_get(_dev);
5314 ops = master->rtnl_link_ops;
5315 if (ops && ops->get_linkxstats_size) {
5316 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5318 size += nla_total_size(ops->get_linkxstats_size(dev,
5320 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5321 size += nla_total_size(0);
5325 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5326 size += rtnl_get_offload_stats_size(dev);
5328 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5329 struct rtnl_af_ops *af_ops;
5331 /* for IFLA_STATS_AF_SPEC */
5332 size += nla_total_size(0);
5335 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5336 if (af_ops->get_stats_af_size) {
5337 size += nla_total_size(
5338 af_ops->get_stats_af_size(dev));
5341 size += nla_total_size(0);
5350 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5351 bool is_dump, struct netlink_ext_ack *extack)
5353 struct if_stats_msg *ifsm;
5355 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5356 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5363 ifsm = nlmsg_data(nlh);
5365 /* only requests using strict checks can pass data to influence
5366 * the dump. The legacy exception is filter_mask.
5368 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5369 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5372 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5373 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5376 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5377 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5384 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5385 struct netlink_ext_ack *extack)
5387 struct net *net = sock_net(skb->sk);
5388 struct net_device *dev = NULL;
5389 int idxattr = 0, prividx = 0;
5390 struct if_stats_msg *ifsm;
5391 struct sk_buff *nskb;
5395 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5400 ifsm = nlmsg_data(nlh);
5401 if (ifsm->ifindex > 0)
5402 dev = __dev_get_by_index(net, ifsm->ifindex);
5409 filter_mask = ifsm->filter_mask;
5413 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5417 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5418 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5419 0, filter_mask, &idxattr, &prividx);
5421 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5422 WARN_ON(err == -EMSGSIZE);
5425 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5431 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5433 struct netlink_ext_ack *extack = cb->extack;
5434 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5435 struct net *net = sock_net(skb->sk);
5436 unsigned int flags = NLM_F_MULTI;
5437 struct if_stats_msg *ifsm;
5438 struct hlist_head *head;
5439 struct net_device *dev;
5440 u32 filter_mask = 0;
5444 s_idx = cb->args[1];
5445 s_idxattr = cb->args[2];
5446 s_prividx = cb->args[3];
5448 cb->seq = net->dev_base_seq;
5450 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5454 ifsm = nlmsg_data(cb->nlh);
5455 filter_mask = ifsm->filter_mask;
5457 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5461 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5463 head = &net->dev_index_head[h];
5464 hlist_for_each_entry(dev, head, index_hlist) {
5467 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5468 NETLINK_CB(cb->skb).portid,
5469 cb->nlh->nlmsg_seq, 0,
5471 &s_idxattr, &s_prividx);
5472 /* If we ran out of room on the first message,
5475 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5481 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5487 cb->args[3] = s_prividx;
5488 cb->args[2] = s_idxattr;
5495 /* Process one rtnetlink message. */
5497 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5498 struct netlink_ext_ack *extack)
5500 struct net *net = sock_net(skb->sk);
5501 struct rtnl_link *link;
5502 struct module *owner;
5503 int err = -EOPNOTSUPP;
5504 rtnl_doit_func doit;
5510 type = nlh->nlmsg_type;
5516 /* All the messages must have at least 1 byte length */
5517 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5520 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5523 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5527 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5529 rtnl_dumpit_func dumpit;
5530 u32 min_dump_alloc = 0;
5532 link = rtnl_get_link(family, type);
5533 if (!link || !link->dumpit) {
5535 link = rtnl_get_link(family, type);
5536 if (!link || !link->dumpit)
5539 owner = link->owner;
5540 dumpit = link->dumpit;
5542 if (type == RTM_GETLINK - RTM_BASE)
5543 min_dump_alloc = rtnl_calcit(skb, nlh);
5546 /* need to do this before rcu_read_unlock() */
5547 if (!try_module_get(owner))
5548 err = -EPROTONOSUPPORT;
5554 struct netlink_dump_control c = {
5556 .min_dump_alloc = min_dump_alloc,
5559 err = netlink_dump_start(rtnl, skb, nlh, &c);
5560 /* netlink_dump_start() will keep a reference on
5561 * module if dump is still in progress.
5568 link = rtnl_get_link(family, type);
5569 if (!link || !link->doit) {
5571 link = rtnl_get_link(PF_UNSPEC, type);
5572 if (!link || !link->doit)
5576 owner = link->owner;
5577 if (!try_module_get(owner)) {
5578 err = -EPROTONOSUPPORT;
5582 flags = link->flags;
5583 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5587 err = doit(skb, nlh, extack);
5594 link = rtnl_get_link(family, type);
5595 if (link && link->doit)
5596 err = link->doit(skb, nlh, extack);
5612 static void rtnetlink_rcv(struct sk_buff *skb)
5614 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5617 static int rtnetlink_bind(struct net *net, int group)
5620 case RTNLGRP_IPV4_MROUTE_R:
5621 case RTNLGRP_IPV6_MROUTE_R:
5622 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5629 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5631 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5635 case NETDEV_CHANGEMTU:
5636 case NETDEV_CHANGEADDR:
5637 case NETDEV_CHANGENAME:
5638 case NETDEV_FEAT_CHANGE:
5639 case NETDEV_BONDING_FAILOVER:
5640 case NETDEV_POST_TYPE_CHANGE:
5641 case NETDEV_NOTIFY_PEERS:
5642 case NETDEV_CHANGEUPPER:
5643 case NETDEV_RESEND_IGMP:
5644 case NETDEV_CHANGEINFODATA:
5645 case NETDEV_CHANGELOWERSTATE:
5646 case NETDEV_CHANGE_TX_QUEUE_LEN:
5647 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5648 GFP_KERNEL, NULL, 0);
5656 static struct notifier_block rtnetlink_dev_notifier = {
5657 .notifier_call = rtnetlink_event,
5661 static int __net_init rtnetlink_net_init(struct net *net)
5664 struct netlink_kernel_cfg cfg = {
5665 .groups = RTNLGRP_MAX,
5666 .input = rtnetlink_rcv,
5667 .cb_mutex = &rtnl_mutex,
5668 .flags = NL_CFG_F_NONROOT_RECV,
5669 .bind = rtnetlink_bind,
5672 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5679 static void __net_exit rtnetlink_net_exit(struct net *net)
5681 netlink_kernel_release(net->rtnl);
5685 static struct pernet_operations rtnetlink_net_ops = {
5686 .init = rtnetlink_net_init,
5687 .exit = rtnetlink_net_exit,
5690 void __init rtnetlink_init(void)
5692 if (register_pernet_subsys(&rtnetlink_net_ops))
5693 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5695 register_netdevice_notifier(&rtnetlink_dev_notifier);
5697 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5698 rtnl_dump_ifinfo, 0);
5699 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5700 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5701 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5703 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5704 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5705 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5707 rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5708 rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5710 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5711 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5712 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5714 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5715 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5716 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5718 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,