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_assign_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_assign_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_bh(&dev_base_lock);
846 dev->operstate = operstate;
847 write_unlock_bh(&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(1) /* IFLA_OPERSTATE */
1030 + nla_total_size(1) /* IFLA_LINKMODE */
1031 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1032 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1033 + nla_total_size(4) /* IFLA_GROUP */
1034 + nla_total_size(ext_filter_mask
1035 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1036 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1037 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1038 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1039 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1040 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1041 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1042 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1043 + rtnl_xdp_size() /* IFLA_XDP */
1044 + nla_total_size(4) /* IFLA_EVENT */
1045 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1046 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1047 + rtnl_proto_down_size(dev) /* proto down */
1048 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1049 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1050 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1051 + nla_total_size(4) /* IFLA_MIN_MTU */
1052 + nla_total_size(4) /* IFLA_MAX_MTU */
1053 + rtnl_prop_list_size(dev)
1054 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1058 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1060 struct nlattr *vf_ports;
1061 struct nlattr *vf_port;
1065 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1069 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1070 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1072 goto nla_put_failure;
1073 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1074 goto nla_put_failure;
1075 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1076 if (err == -EMSGSIZE)
1077 goto nla_put_failure;
1079 nla_nest_cancel(skb, vf_port);
1082 nla_nest_end(skb, vf_port);
1085 nla_nest_end(skb, vf_ports);
1090 nla_nest_cancel(skb, vf_ports);
1094 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1096 struct nlattr *port_self;
1099 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1103 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1105 nla_nest_cancel(skb, port_self);
1106 return (err == -EMSGSIZE) ? err : 0;
1109 nla_nest_end(skb, port_self);
1114 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1115 u32 ext_filter_mask)
1119 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1120 !(ext_filter_mask & RTEXT_FILTER_VF))
1123 err = rtnl_port_self_fill(skb, dev);
1127 if (dev_num_vf(dev->dev.parent)) {
1128 err = rtnl_vf_ports_fill(skb, dev);
1136 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1139 struct netdev_phys_item_id ppid;
1141 err = dev_get_phys_port_id(dev, &ppid);
1143 if (err == -EOPNOTSUPP)
1148 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1154 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1156 char name[IFNAMSIZ];
1159 err = dev_get_phys_port_name(dev, name, sizeof(name));
1161 if (err == -EOPNOTSUPP)
1166 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1172 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1174 struct netdev_phys_item_id ppid = { };
1177 err = dev_get_port_parent_id(dev, &ppid, false);
1179 if (err == -EOPNOTSUPP)
1184 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1190 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1191 struct net_device *dev)
1193 struct rtnl_link_stats64 *sp;
1194 struct nlattr *attr;
1196 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1197 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1201 sp = nla_data(attr);
1202 dev_get_stats(dev, sp);
1204 attr = nla_reserve(skb, IFLA_STATS,
1205 sizeof(struct rtnl_link_stats));
1209 copy_rtnl_link_stats(nla_data(attr), sp);
1214 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1215 struct net_device *dev,
1217 struct nlattr *vfinfo)
1219 struct ifla_vf_rss_query_en vf_rss_query_en;
1220 struct nlattr *vf, *vfstats, *vfvlanlist;
1221 struct ifla_vf_link_state vf_linkstate;
1222 struct ifla_vf_vlan_info vf_vlan_info;
1223 struct ifla_vf_spoofchk vf_spoofchk;
1224 struct ifla_vf_tx_rate vf_tx_rate;
1225 struct ifla_vf_stats vf_stats;
1226 struct ifla_vf_trust vf_trust;
1227 struct ifla_vf_vlan vf_vlan;
1228 struct ifla_vf_rate vf_rate;
1229 struct ifla_vf_mac vf_mac;
1230 struct ifla_vf_broadcast vf_broadcast;
1231 struct ifla_vf_info ivi;
1232 struct ifla_vf_guid node_guid;
1233 struct ifla_vf_guid port_guid;
1235 memset(&ivi, 0, sizeof(ivi));
1237 /* Not all SR-IOV capable drivers support the
1238 * spoofcheck and "RSS query enable" query. Preset to
1239 * -1 so the user space tool can detect that the driver
1240 * didn't report anything.
1243 ivi.rss_query_en = -1;
1245 /* The default value for VF link state is "auto"
1246 * IFLA_VF_LINK_STATE_AUTO which equals zero
1249 /* VLAN Protocol by default is 802.1Q */
1250 ivi.vlan_proto = htons(ETH_P_8021Q);
1251 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1254 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1255 memset(&node_guid, 0, sizeof(node_guid));
1256 memset(&port_guid, 0, sizeof(port_guid));
1265 vf_rss_query_en.vf =
1268 port_guid.vf = ivi.vf;
1270 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1271 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1272 vf_vlan.vlan = ivi.vlan;
1273 vf_vlan.qos = ivi.qos;
1274 vf_vlan_info.vlan = ivi.vlan;
1275 vf_vlan_info.qos = ivi.qos;
1276 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1277 vf_tx_rate.rate = ivi.max_tx_rate;
1278 vf_rate.min_tx_rate = ivi.min_tx_rate;
1279 vf_rate.max_tx_rate = ivi.max_tx_rate;
1280 vf_spoofchk.setting = ivi.spoofchk;
1281 vf_linkstate.link_state = ivi.linkstate;
1282 vf_rss_query_en.setting = ivi.rss_query_en;
1283 vf_trust.setting = ivi.trusted;
1284 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1286 goto nla_put_vfinfo_failure;
1287 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1288 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1289 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1290 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1292 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1294 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1296 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1298 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1299 sizeof(vf_rss_query_en),
1300 &vf_rss_query_en) ||
1301 nla_put(skb, IFLA_VF_TRUST,
1302 sizeof(vf_trust), &vf_trust))
1303 goto nla_put_vf_failure;
1305 if (dev->netdev_ops->ndo_get_vf_guid &&
1306 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1308 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1310 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1312 goto nla_put_vf_failure;
1314 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1316 goto nla_put_vf_failure;
1317 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1319 nla_nest_cancel(skb, vfvlanlist);
1320 goto nla_put_vf_failure;
1322 nla_nest_end(skb, vfvlanlist);
1323 memset(&vf_stats, 0, sizeof(vf_stats));
1324 if (dev->netdev_ops->ndo_get_vf_stats)
1325 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1327 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1329 goto nla_put_vf_failure;
1330 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1331 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1332 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1333 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1334 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1335 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1336 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1337 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1338 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1339 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1340 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1341 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1342 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1343 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1344 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1345 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1346 nla_nest_cancel(skb, vfstats);
1347 goto nla_put_vf_failure;
1349 nla_nest_end(skb, vfstats);
1350 nla_nest_end(skb, vf);
1354 nla_nest_cancel(skb, vf);
1355 nla_put_vfinfo_failure:
1356 nla_nest_cancel(skb, vfinfo);
1360 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1361 struct net_device *dev,
1362 u32 ext_filter_mask)
1364 struct nlattr *vfinfo;
1367 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1370 num_vfs = dev_num_vf(dev->dev.parent);
1371 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1374 if (!dev->netdev_ops->ndo_get_vf_config)
1377 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1381 for (i = 0; i < num_vfs; i++) {
1382 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1386 nla_nest_end(skb, vfinfo);
1390 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1392 struct rtnl_link_ifmap map;
1394 memset(&map, 0, sizeof(map));
1395 map.mem_start = dev->mem_start;
1396 map.mem_end = dev->mem_end;
1397 map.base_addr = dev->base_addr;
1400 map.port = dev->if_port;
1402 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1408 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1410 const struct bpf_prog *generic_xdp_prog;
1414 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1415 if (!generic_xdp_prog)
1417 return generic_xdp_prog->aux->id;
1420 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1422 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1425 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1427 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1430 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1431 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1432 u32 (*get_prog_id)(struct net_device *dev))
1437 curr_id = get_prog_id(dev);
1442 err = nla_put_u32(skb, attr, curr_id);
1446 if (*mode != XDP_ATTACHED_NONE)
1447 *mode = XDP_ATTACHED_MULTI;
1454 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1461 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1466 mode = XDP_ATTACHED_NONE;
1467 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1468 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1471 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1472 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1475 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1476 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1480 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1484 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1485 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1490 nla_nest_end(skb, xdp);
1494 nla_nest_cancel(skb, xdp);
1498 static u32 rtnl_get_event(unsigned long event)
1500 u32 rtnl_event_type = IFLA_EVENT_NONE;
1504 rtnl_event_type = IFLA_EVENT_REBOOT;
1506 case NETDEV_FEAT_CHANGE:
1507 rtnl_event_type = IFLA_EVENT_FEATURES;
1509 case NETDEV_BONDING_FAILOVER:
1510 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1512 case NETDEV_NOTIFY_PEERS:
1513 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1515 case NETDEV_RESEND_IGMP:
1516 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1518 case NETDEV_CHANGEINFODATA:
1519 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1525 return rtnl_event_type;
1528 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1530 const struct net_device *upper_dev;
1535 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1537 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1543 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1546 int ifindex = dev_get_iflink(dev);
1548 if (force || dev->ifindex != ifindex)
1549 return nla_put_u32(skb, IFLA_LINK, ifindex);
1554 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1555 struct net_device *dev)
1560 ret = dev_get_alias(dev, buf, sizeof(buf));
1561 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1564 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1565 const struct net_device *dev,
1566 struct net *src_net, gfp_t gfp)
1568 bool put_iflink = false;
1570 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1571 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1573 if (!net_eq(dev_net(dev), link_net)) {
1574 int id = peernet2id_alloc(src_net, link_net, gfp);
1576 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1583 return nla_put_iflink(skb, dev, put_iflink);
1586 static int rtnl_fill_link_af(struct sk_buff *skb,
1587 const struct net_device *dev,
1588 u32 ext_filter_mask)
1590 const struct rtnl_af_ops *af_ops;
1591 struct nlattr *af_spec;
1593 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1597 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1601 if (!af_ops->fill_link_af)
1604 af = nla_nest_start_noflag(skb, af_ops->family);
1608 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1610 * Caller may return ENODATA to indicate that there
1611 * was no data to be dumped. This is not an error, it
1612 * means we should trim the attribute header and
1615 if (err == -ENODATA)
1616 nla_nest_cancel(skb, af);
1620 nla_nest_end(skb, af);
1623 nla_nest_end(skb, af_spec);
1627 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1628 const struct net_device *dev)
1630 struct netdev_name_node *name_node;
1633 list_for_each_entry(name_node, &dev->name_node->list, list) {
1634 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1641 static int rtnl_fill_prop_list(struct sk_buff *skb,
1642 const struct net_device *dev)
1644 struct nlattr *prop_list;
1647 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1651 ret = rtnl_fill_alt_ifnames(skb, dev);
1655 nla_nest_end(skb, prop_list);
1659 nla_nest_cancel(skb, prop_list);
1663 static int rtnl_fill_proto_down(struct sk_buff *skb,
1664 const struct net_device *dev)
1669 if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1670 goto nla_put_failure;
1672 preason = dev->proto_down_reason;
1676 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1680 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1681 nla_nest_cancel(skb, pr);
1682 goto nla_put_failure;
1685 nla_nest_end(skb, pr);
1692 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1693 struct net_device *dev, struct net *src_net,
1694 int type, u32 pid, u32 seq, u32 change,
1695 unsigned int flags, u32 ext_filter_mask,
1696 u32 event, int *new_nsid, int new_ifindex,
1697 int tgt_netnsid, gfp_t gfp)
1699 struct ifinfomsg *ifm;
1700 struct nlmsghdr *nlh;
1703 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1707 ifm = nlmsg_data(nlh);
1708 ifm->ifi_family = AF_UNSPEC;
1710 ifm->ifi_type = dev->type;
1711 ifm->ifi_index = dev->ifindex;
1712 ifm->ifi_flags = dev_get_flags(dev);
1713 ifm->ifi_change = change;
1715 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1716 goto nla_put_failure;
1718 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1719 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1720 nla_put_u8(skb, IFLA_OPERSTATE,
1721 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1722 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1723 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1724 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1725 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1726 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1727 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1728 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1729 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1730 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1732 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1734 put_master_ifindex(skb, dev) ||
1735 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1737 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1738 nla_put_ifalias(skb, dev) ||
1739 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1740 atomic_read(&dev->carrier_up_count) +
1741 atomic_read(&dev->carrier_down_count)) ||
1742 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1743 atomic_read(&dev->carrier_up_count)) ||
1744 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1745 atomic_read(&dev->carrier_down_count)))
1746 goto nla_put_failure;
1748 if (rtnl_fill_proto_down(skb, dev))
1749 goto nla_put_failure;
1751 if (event != IFLA_EVENT_NONE) {
1752 if (nla_put_u32(skb, IFLA_EVENT, event))
1753 goto nla_put_failure;
1756 if (rtnl_fill_link_ifmap(skb, dev))
1757 goto nla_put_failure;
1759 if (dev->addr_len) {
1760 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1761 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1762 goto nla_put_failure;
1765 if (rtnl_phys_port_id_fill(skb, dev))
1766 goto nla_put_failure;
1768 if (rtnl_phys_port_name_fill(skb, dev))
1769 goto nla_put_failure;
1771 if (rtnl_phys_switch_id_fill(skb, dev))
1772 goto nla_put_failure;
1774 if (rtnl_fill_stats(skb, dev))
1775 goto nla_put_failure;
1777 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1778 goto nla_put_failure;
1780 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1781 goto nla_put_failure;
1783 if (rtnl_xdp_fill(skb, dev))
1784 goto nla_put_failure;
1786 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1787 if (rtnl_link_fill(skb, dev) < 0)
1788 goto nla_put_failure;
1791 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1792 goto nla_put_failure;
1795 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1796 goto nla_put_failure;
1798 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1799 goto nla_put_failure;
1801 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1802 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1803 goto nla_put_failure;
1806 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1807 goto nla_put_failure_rcu;
1810 if (rtnl_fill_prop_list(skb, dev))
1811 goto nla_put_failure;
1813 if (dev->dev.parent &&
1814 nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1815 dev_name(dev->dev.parent)))
1816 goto nla_put_failure;
1818 if (dev->dev.parent && dev->dev.parent->bus &&
1819 nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1820 dev->dev.parent->bus->name))
1821 goto nla_put_failure;
1823 nlmsg_end(skb, nlh);
1826 nla_put_failure_rcu:
1829 nlmsg_cancel(skb, nlh);
1833 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1834 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1835 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1836 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1837 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1838 [IFLA_MTU] = { .type = NLA_U32 },
1839 [IFLA_LINK] = { .type = NLA_U32 },
1840 [IFLA_MASTER] = { .type = NLA_U32 },
1841 [IFLA_CARRIER] = { .type = NLA_U8 },
1842 [IFLA_TXQLEN] = { .type = NLA_U32 },
1843 [IFLA_WEIGHT] = { .type = NLA_U32 },
1844 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1845 [IFLA_LINKMODE] = { .type = NLA_U8 },
1846 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1847 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1848 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1849 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1850 * allow 0-length string (needed to remove an alias).
1852 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1853 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1854 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1855 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1856 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1857 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1858 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1859 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1860 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1861 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1862 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1863 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1864 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1865 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1866 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1867 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1868 [IFLA_XDP] = { .type = NLA_NESTED },
1869 [IFLA_EVENT] = { .type = NLA_U32 },
1870 [IFLA_GROUP] = { .type = NLA_U32 },
1871 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
1872 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1873 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1874 [IFLA_MIN_MTU] = { .type = NLA_U32 },
1875 [IFLA_MAX_MTU] = { .type = NLA_U32 },
1876 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
1877 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
1878 .len = ALTIFNAMSIZ - 1 },
1879 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
1880 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1881 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
1882 [IFLA_PARENT_DEV_NAME] = { .type = NLA_NUL_STRING },
1885 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1886 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1887 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1888 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1889 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1892 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1893 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1894 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1895 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1896 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1897 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1898 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1899 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1900 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1901 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1902 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1903 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1904 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1905 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1908 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1909 [IFLA_PORT_VF] = { .type = NLA_U32 },
1910 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1911 .len = PORT_PROFILE_MAX },
1912 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1913 .len = PORT_UUID_MAX },
1914 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1915 .len = PORT_UUID_MAX },
1916 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1917 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1919 /* Unused, but we need to keep it here since user space could
1920 * fill it. It's also broken with regard to NLA_BINARY use in
1921 * combination with structs.
1923 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1924 .len = sizeof(struct ifla_port_vsi) },
1927 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1928 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1929 [IFLA_XDP_FD] = { .type = NLA_S32 },
1930 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
1931 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1932 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1933 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1936 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1938 const struct rtnl_link_ops *ops = NULL;
1939 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1941 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1944 if (linfo[IFLA_INFO_KIND]) {
1945 char kind[MODULE_NAME_LEN];
1947 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1948 ops = rtnl_link_ops_get(kind);
1954 static bool link_master_filtered(struct net_device *dev, int master_idx)
1956 struct net_device *master;
1961 master = netdev_master_upper_dev_get(dev);
1963 /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
1964 * another invalid value for ifindex to denote "no master".
1966 if (master_idx == -1)
1969 if (!master || master->ifindex != master_idx)
1975 static bool link_kind_filtered(const struct net_device *dev,
1976 const struct rtnl_link_ops *kind_ops)
1978 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1984 static bool link_dump_filtered(struct net_device *dev,
1986 const struct rtnl_link_ops *kind_ops)
1988 if (link_master_filtered(dev, master_idx) ||
1989 link_kind_filtered(dev, kind_ops))
1996 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1997 * @sk: netlink socket
1998 * @netnsid: network namespace identifier
2000 * Returns the network namespace identified by netnsid on success or an error
2001 * pointer on failure.
2003 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2007 net = get_net_ns_by_id(sock_net(sk), netnsid);
2009 return ERR_PTR(-EINVAL);
2011 /* For now, the caller is required to have CAP_NET_ADMIN in
2012 * the user namespace owning the target net ns.
2014 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2016 return ERR_PTR(-EACCES);
2020 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2022 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2023 bool strict_check, struct nlattr **tb,
2024 struct netlink_ext_ack *extack)
2029 struct ifinfomsg *ifm;
2031 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2032 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2036 ifm = nlmsg_data(nlh);
2037 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2039 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2042 if (ifm->ifi_index) {
2043 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2047 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2048 IFLA_MAX, ifla_policy,
2052 /* A hack to preserve kernel<->userspace interface.
2053 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2054 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2055 * what iproute2 < v3.9.0 used.
2056 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2057 * attribute, its netlink message is shorter than struct ifinfomsg.
2059 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2060 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2062 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2066 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2068 struct netlink_ext_ack *extack = cb->extack;
2069 const struct nlmsghdr *nlh = cb->nlh;
2070 struct net *net = sock_net(skb->sk);
2071 struct net *tgt_net = net;
2074 struct net_device *dev;
2075 struct hlist_head *head;
2076 struct nlattr *tb[IFLA_MAX+1];
2077 u32 ext_filter_mask = 0;
2078 const struct rtnl_link_ops *kind_ops = NULL;
2079 unsigned int flags = NLM_F_MULTI;
2085 s_idx = cb->args[1];
2087 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2089 if (cb->strict_check)
2095 for (i = 0; i <= IFLA_MAX; ++i) {
2099 /* new attributes should only be added with strict checking */
2101 case IFLA_TARGET_NETNSID:
2102 netnsid = nla_get_s32(tb[i]);
2103 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2104 if (IS_ERR(tgt_net)) {
2105 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2106 return PTR_ERR(tgt_net);
2110 ext_filter_mask = nla_get_u32(tb[i]);
2113 master_idx = nla_get_u32(tb[i]);
2116 kind_ops = linkinfo_to_kind_ops(tb[i]);
2119 if (cb->strict_check) {
2120 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2126 if (master_idx || kind_ops)
2127 flags |= NLM_F_DUMP_FILTERED;
2130 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2132 head = &tgt_net->dev_index_head[h];
2133 hlist_for_each_entry(dev, head, index_hlist) {
2134 if (link_dump_filtered(dev, master_idx, kind_ops))
2138 err = rtnl_fill_ifinfo(skb, dev, net,
2140 NETLINK_CB(cb->skb).portid,
2141 nlh->nlmsg_seq, 0, flags,
2142 ext_filter_mask, 0, NULL, 0,
2143 netnsid, GFP_KERNEL);
2146 if (likely(skb->len))
2160 cb->seq = tgt_net->dev_base_seq;
2161 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2168 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2169 struct netlink_ext_ack *exterr)
2171 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2174 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2176 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2179 /* Examine the link attributes and figure out which
2180 * network namespace we are talking about.
2182 if (tb[IFLA_NET_NS_PID])
2183 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2184 else if (tb[IFLA_NET_NS_FD])
2185 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2187 net = get_net(src_net);
2190 EXPORT_SYMBOL(rtnl_link_get_net);
2192 /* Figure out which network namespace we are talking about by
2193 * examining the link attributes in the following order:
2195 * 1. IFLA_NET_NS_PID
2197 * 3. IFLA_TARGET_NETNSID
2199 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2200 struct nlattr *tb[])
2204 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2205 return rtnl_link_get_net(src_net, tb);
2207 if (!tb[IFLA_TARGET_NETNSID])
2208 return get_net(src_net);
2210 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2212 return ERR_PTR(-EINVAL);
2217 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2218 struct net *src_net,
2219 struct nlattr *tb[], int cap)
2223 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2227 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2229 return ERR_PTR(-EPERM);
2235 /* Verify that rtnetlink requests do not pass additional properties
2236 * potentially referring to different network namespaces.
2238 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2239 struct netlink_ext_ack *extack,
2243 if (netns_id_only) {
2244 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2247 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2251 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2254 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2257 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2263 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2267 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2268 struct netlink_ext_ack *extack)
2271 if (tb[IFLA_ADDRESS] &&
2272 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2275 if (tb[IFLA_BROADCAST] &&
2276 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2280 if (tb[IFLA_AF_SPEC]) {
2284 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2285 const struct rtnl_af_ops *af_ops;
2287 af_ops = rtnl_af_lookup(nla_type(af));
2289 return -EAFNOSUPPORT;
2291 if (!af_ops->set_link_af)
2294 if (af_ops->validate_link_af) {
2295 err = af_ops->validate_link_af(dev, af, extack);
2305 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2308 const struct net_device_ops *ops = dev->netdev_ops;
2310 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2313 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2315 if (dev->type != ARPHRD_INFINIBAND)
2318 return handle_infiniband_guid(dev, ivt, guid_type);
2321 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2323 const struct net_device_ops *ops = dev->netdev_ops;
2326 if (tb[IFLA_VF_MAC]) {
2327 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2329 if (ivm->vf >= INT_MAX)
2332 if (ops->ndo_set_vf_mac)
2333 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2339 if (tb[IFLA_VF_VLAN]) {
2340 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2342 if (ivv->vf >= INT_MAX)
2345 if (ops->ndo_set_vf_vlan)
2346 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2348 htons(ETH_P_8021Q));
2353 if (tb[IFLA_VF_VLAN_LIST]) {
2354 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2355 struct nlattr *attr;
2359 if (!ops->ndo_set_vf_vlan)
2362 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2363 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2364 nla_len(attr) < NLA_HDRLEN) {
2367 if (len >= MAX_VLAN_LIST_LEN)
2369 ivvl[len] = nla_data(attr);
2376 if (ivvl[0]->vf >= INT_MAX)
2378 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2379 ivvl[0]->qos, ivvl[0]->vlan_proto);
2384 if (tb[IFLA_VF_TX_RATE]) {
2385 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2386 struct ifla_vf_info ivf;
2388 if (ivt->vf >= INT_MAX)
2391 if (ops->ndo_get_vf_config)
2392 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2397 if (ops->ndo_set_vf_rate)
2398 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2405 if (tb[IFLA_VF_RATE]) {
2406 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2408 if (ivt->vf >= INT_MAX)
2411 if (ops->ndo_set_vf_rate)
2412 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2419 if (tb[IFLA_VF_SPOOFCHK]) {
2420 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2422 if (ivs->vf >= INT_MAX)
2425 if (ops->ndo_set_vf_spoofchk)
2426 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2432 if (tb[IFLA_VF_LINK_STATE]) {
2433 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2435 if (ivl->vf >= INT_MAX)
2438 if (ops->ndo_set_vf_link_state)
2439 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2445 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2446 struct ifla_vf_rss_query_en *ivrssq_en;
2449 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2450 if (ivrssq_en->vf >= INT_MAX)
2452 if (ops->ndo_set_vf_rss_query_en)
2453 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2454 ivrssq_en->setting);
2459 if (tb[IFLA_VF_TRUST]) {
2460 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2462 if (ivt->vf >= INT_MAX)
2465 if (ops->ndo_set_vf_trust)
2466 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2471 if (tb[IFLA_VF_IB_NODE_GUID]) {
2472 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2474 if (ivt->vf >= INT_MAX)
2476 if (!ops->ndo_set_vf_guid)
2478 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2481 if (tb[IFLA_VF_IB_PORT_GUID]) {
2482 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2484 if (ivt->vf >= INT_MAX)
2486 if (!ops->ndo_set_vf_guid)
2489 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2495 static int do_set_master(struct net_device *dev, int ifindex,
2496 struct netlink_ext_ack *extack)
2498 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2499 const struct net_device_ops *ops;
2503 if (upper_dev->ifindex == ifindex)
2505 ops = upper_dev->netdev_ops;
2506 if (ops->ndo_del_slave) {
2507 err = ops->ndo_del_slave(upper_dev, dev);
2516 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2519 ops = upper_dev->netdev_ops;
2520 if (ops->ndo_add_slave) {
2521 err = ops->ndo_add_slave(upper_dev, dev, extack);
2531 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2532 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2533 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2536 static int do_set_proto_down(struct net_device *dev,
2537 struct nlattr *nl_proto_down,
2538 struct nlattr *nl_proto_down_reason,
2539 struct netlink_ext_ack *extack)
2541 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2542 const struct net_device_ops *ops = dev->netdev_ops;
2543 unsigned long mask = 0;
2548 if (!ops->ndo_change_proto_down) {
2549 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2553 if (nl_proto_down_reason) {
2554 err = nla_parse_nested_deprecated(pdreason,
2555 IFLA_PROTO_DOWN_REASON_MAX,
2556 nl_proto_down_reason,
2557 ifla_proto_down_reason_policy,
2562 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2563 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2567 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2569 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2570 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2572 dev_change_proto_down_reason(dev, mask, value);
2575 if (nl_proto_down) {
2576 proto_down = nla_get_u8(nl_proto_down);
2578 /* Don't turn off protodown if there are active reasons */
2579 if (!proto_down && dev->proto_down_reason) {
2580 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2583 err = dev_change_proto_down(dev,
2592 #define DO_SETLINK_MODIFIED 0x01
2593 /* notify flag means notify + modified. */
2594 #define DO_SETLINK_NOTIFY 0x03
2595 static int do_setlink(const struct sk_buff *skb,
2596 struct net_device *dev, struct ifinfomsg *ifm,
2597 struct netlink_ext_ack *extack,
2598 struct nlattr **tb, char *ifname, int status)
2600 const struct net_device_ops *ops = dev->netdev_ops;
2603 err = validate_linkmsg(dev, tb, extack);
2607 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2608 const char *pat = ifname && ifname[0] ? ifname : NULL;
2612 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2619 if (tb[IFLA_NEW_IFINDEX])
2620 new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2624 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2628 status |= DO_SETLINK_MODIFIED;
2632 struct rtnl_link_ifmap *u_map;
2635 if (!ops->ndo_set_config) {
2640 if (!netif_device_present(dev)) {
2645 u_map = nla_data(tb[IFLA_MAP]);
2646 k_map.mem_start = (unsigned long) u_map->mem_start;
2647 k_map.mem_end = (unsigned long) u_map->mem_end;
2648 k_map.base_addr = (unsigned short) u_map->base_addr;
2649 k_map.irq = (unsigned char) u_map->irq;
2650 k_map.dma = (unsigned char) u_map->dma;
2651 k_map.port = (unsigned char) u_map->port;
2653 err = ops->ndo_set_config(dev, &k_map);
2657 status |= DO_SETLINK_NOTIFY;
2660 if (tb[IFLA_ADDRESS]) {
2661 struct sockaddr *sa;
2664 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2666 sa = kmalloc(len, GFP_KERNEL);
2671 sa->sa_family = dev->type;
2672 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2674 err = dev_set_mac_address_user(dev, sa, extack);
2678 status |= DO_SETLINK_MODIFIED;
2682 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2685 status |= DO_SETLINK_MODIFIED;
2688 if (tb[IFLA_GROUP]) {
2689 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2690 status |= DO_SETLINK_NOTIFY;
2694 * Interface selected by interface index but interface
2695 * name provided implies that a name change has been
2698 if (ifm->ifi_index > 0 && ifname[0]) {
2699 err = dev_change_name(dev, ifname);
2702 status |= DO_SETLINK_MODIFIED;
2705 if (tb[IFLA_IFALIAS]) {
2706 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2707 nla_len(tb[IFLA_IFALIAS]));
2710 status |= DO_SETLINK_NOTIFY;
2713 if (tb[IFLA_BROADCAST]) {
2714 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2715 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2718 if (ifm->ifi_flags || ifm->ifi_change) {
2719 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2725 if (tb[IFLA_MASTER]) {
2726 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2729 status |= DO_SETLINK_MODIFIED;
2732 if (tb[IFLA_CARRIER]) {
2733 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2736 status |= DO_SETLINK_MODIFIED;
2739 if (tb[IFLA_TXQLEN]) {
2740 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2742 err = dev_change_tx_queue_len(dev, value);
2745 status |= DO_SETLINK_MODIFIED;
2748 if (tb[IFLA_GSO_MAX_SIZE]) {
2749 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2751 if (max_size > GSO_MAX_SIZE) {
2756 if (dev->gso_max_size ^ max_size) {
2757 netif_set_gso_max_size(dev, max_size);
2758 status |= DO_SETLINK_MODIFIED;
2762 if (tb[IFLA_GSO_MAX_SEGS]) {
2763 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2765 if (max_segs > GSO_MAX_SEGS) {
2770 if (dev->gso_max_segs ^ max_segs) {
2771 dev->gso_max_segs = max_segs;
2772 status |= DO_SETLINK_MODIFIED;
2776 if (tb[IFLA_OPERSTATE])
2777 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2779 if (tb[IFLA_LINKMODE]) {
2780 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2782 write_lock_bh(&dev_base_lock);
2783 if (dev->link_mode ^ value)
2784 status |= DO_SETLINK_NOTIFY;
2785 dev->link_mode = value;
2786 write_unlock_bh(&dev_base_lock);
2789 if (tb[IFLA_VFINFO_LIST]) {
2790 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2791 struct nlattr *attr;
2794 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2795 if (nla_type(attr) != IFLA_VF_INFO ||
2796 nla_len(attr) < NLA_HDRLEN) {
2800 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2806 err = do_setvfinfo(dev, vfinfo);
2809 status |= DO_SETLINK_NOTIFY;
2814 if (tb[IFLA_VF_PORTS]) {
2815 struct nlattr *port[IFLA_PORT_MAX+1];
2816 struct nlattr *attr;
2821 if (!ops->ndo_set_vf_port)
2824 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2825 if (nla_type(attr) != IFLA_VF_PORT ||
2826 nla_len(attr) < NLA_HDRLEN) {
2830 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2836 if (!port[IFLA_PORT_VF]) {
2840 vf = nla_get_u32(port[IFLA_PORT_VF]);
2841 err = ops->ndo_set_vf_port(dev, vf, port);
2844 status |= DO_SETLINK_NOTIFY;
2849 if (tb[IFLA_PORT_SELF]) {
2850 struct nlattr *port[IFLA_PORT_MAX+1];
2852 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2854 ifla_port_policy, NULL);
2859 if (ops->ndo_set_vf_port)
2860 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2863 status |= DO_SETLINK_NOTIFY;
2866 if (tb[IFLA_AF_SPEC]) {
2870 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2871 const struct rtnl_af_ops *af_ops;
2873 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2875 err = af_ops->set_link_af(dev, af, extack);
2879 status |= DO_SETLINK_NOTIFY;
2884 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2885 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2886 tb[IFLA_PROTO_DOWN_REASON], extack);
2889 status |= DO_SETLINK_NOTIFY;
2893 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2896 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2898 ifla_xdp_policy, NULL);
2902 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2907 if (xdp[IFLA_XDP_FLAGS]) {
2908 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2909 if (xdp_flags & ~XDP_FLAGS_MASK) {
2913 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2919 if (xdp[IFLA_XDP_FD]) {
2920 int expected_fd = -1;
2922 if (xdp_flags & XDP_FLAGS_REPLACE) {
2923 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2928 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2931 err = dev_change_xdp_fd(dev, extack,
2932 nla_get_s32(xdp[IFLA_XDP_FD]),
2937 status |= DO_SETLINK_NOTIFY;
2942 if (status & DO_SETLINK_MODIFIED) {
2943 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2944 netdev_state_change(dev);
2947 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",
2954 static struct net_device *rtnl_dev_get(struct net *net,
2955 struct nlattr *ifname_attr,
2956 struct nlattr *altifname_attr,
2959 char buffer[ALTIFNAMSIZ];
2964 nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2965 else if (altifname_attr)
2966 nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2971 return __dev_get_by_name(net, ifname);
2974 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2975 struct netlink_ext_ack *extack)
2977 struct net *net = sock_net(skb->sk);
2978 struct ifinfomsg *ifm;
2979 struct net_device *dev;
2981 struct nlattr *tb[IFLA_MAX+1];
2982 char ifname[IFNAMSIZ];
2984 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2985 ifla_policy, extack);
2989 err = rtnl_ensure_unique_netns(tb, extack, false);
2993 if (tb[IFLA_IFNAME])
2994 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2999 ifm = nlmsg_data(nlh);
3000 if (ifm->ifi_index > 0)
3001 dev = __dev_get_by_index(net, ifm->ifi_index);
3002 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3003 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3012 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3017 static int rtnl_group_dellink(const struct net *net, int group)
3019 struct net_device *dev, *aux;
3020 LIST_HEAD(list_kill);
3026 for_each_netdev(net, dev) {
3027 if (dev->group == group) {
3028 const struct rtnl_link_ops *ops;
3031 ops = dev->rtnl_link_ops;
3032 if (!ops || !ops->dellink)
3040 for_each_netdev_safe(net, dev, aux) {
3041 if (dev->group == group) {
3042 const struct rtnl_link_ops *ops;
3044 ops = dev->rtnl_link_ops;
3045 ops->dellink(dev, &list_kill);
3048 unregister_netdevice_many(&list_kill);
3053 int rtnl_delete_link(struct net_device *dev)
3055 const struct rtnl_link_ops *ops;
3056 LIST_HEAD(list_kill);
3058 ops = dev->rtnl_link_ops;
3059 if (!ops || !ops->dellink)
3062 ops->dellink(dev, &list_kill);
3063 unregister_netdevice_many(&list_kill);
3067 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3069 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3070 struct netlink_ext_ack *extack)
3072 struct net *net = sock_net(skb->sk);
3073 struct net *tgt_net = net;
3074 struct net_device *dev = NULL;
3075 struct ifinfomsg *ifm;
3076 struct nlattr *tb[IFLA_MAX+1];
3080 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3081 ifla_policy, extack);
3085 err = rtnl_ensure_unique_netns(tb, extack, true);
3089 if (tb[IFLA_TARGET_NETNSID]) {
3090 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3091 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3092 if (IS_ERR(tgt_net))
3093 return PTR_ERR(tgt_net);
3097 ifm = nlmsg_data(nlh);
3098 if (ifm->ifi_index > 0)
3099 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3100 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3101 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3102 tb[IFLA_ALT_IFNAME], NULL);
3103 else if (tb[IFLA_GROUP])
3104 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3109 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3115 err = rtnl_delete_link(dev);
3124 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3126 unsigned int old_flags;
3129 old_flags = dev->flags;
3130 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3131 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3137 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3138 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3140 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3141 __dev_notify_flags(dev, old_flags, ~0U);
3145 EXPORT_SYMBOL(rtnl_configure_link);
3147 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3148 unsigned char name_assign_type,
3149 const struct rtnl_link_ops *ops,
3150 struct nlattr *tb[],
3151 struct netlink_ext_ack *extack)
3153 struct net_device *dev;
3154 unsigned int num_tx_queues = 1;
3155 unsigned int num_rx_queues = 1;
3157 if (tb[IFLA_NUM_TX_QUEUES])
3158 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3159 else if (ops->get_num_tx_queues)
3160 num_tx_queues = ops->get_num_tx_queues();
3162 if (tb[IFLA_NUM_RX_QUEUES])
3163 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3164 else if (ops->get_num_rx_queues)
3165 num_rx_queues = ops->get_num_rx_queues();
3167 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3168 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3169 return ERR_PTR(-EINVAL);
3172 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3173 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3174 return ERR_PTR(-EINVAL);
3178 dev = ops->alloc(tb, ifname, name_assign_type,
3179 num_tx_queues, num_rx_queues);
3183 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3184 name_assign_type, ops->setup,
3185 num_tx_queues, num_rx_queues);
3189 return ERR_PTR(-ENOMEM);
3191 dev_net_set(dev, net);
3192 dev->rtnl_link_ops = ops;
3193 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3196 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3199 err = dev_validate_mtu(dev, mtu, extack);
3202 return ERR_PTR(err);
3206 if (tb[IFLA_ADDRESS]) {
3207 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3208 nla_len(tb[IFLA_ADDRESS]));
3209 dev->addr_assign_type = NET_ADDR_SET;
3211 if (tb[IFLA_BROADCAST])
3212 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3213 nla_len(tb[IFLA_BROADCAST]));
3214 if (tb[IFLA_TXQLEN])
3215 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3216 if (tb[IFLA_OPERSTATE])
3217 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3218 if (tb[IFLA_LINKMODE])
3219 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3221 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3222 if (tb[IFLA_GSO_MAX_SIZE])
3223 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3224 if (tb[IFLA_GSO_MAX_SEGS])
3225 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3229 EXPORT_SYMBOL(rtnl_create_link);
3231 static int rtnl_group_changelink(const struct sk_buff *skb,
3232 struct net *net, int group,
3233 struct ifinfomsg *ifm,
3234 struct netlink_ext_ack *extack,
3237 struct net_device *dev, *aux;
3240 for_each_netdev_safe(net, dev, aux) {
3241 if (dev->group == group) {
3242 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3251 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3252 struct nlattr **attr, struct netlink_ext_ack *extack)
3254 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3255 unsigned char name_assign_type = NET_NAME_USER;
3256 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3257 const struct rtnl_link_ops *m_ops = NULL;
3258 struct net_device *master_dev = NULL;
3259 struct net *net = sock_net(skb->sk);
3260 const struct rtnl_link_ops *ops;
3261 struct nlattr *tb[IFLA_MAX + 1];
3262 struct net *dest_net, *link_net;
3263 struct nlattr **slave_data;
3264 char kind[MODULE_NAME_LEN];
3265 struct net_device *dev;
3266 struct ifinfomsg *ifm;
3267 char ifname[IFNAMSIZ];
3268 struct nlattr **data;
3271 #ifdef CONFIG_MODULES
3274 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3275 ifla_policy, extack);
3279 err = rtnl_ensure_unique_netns(tb, extack, false);
3283 if (tb[IFLA_IFNAME])
3284 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3288 ifm = nlmsg_data(nlh);
3289 if (ifm->ifi_index > 0)
3290 dev = __dev_get_by_index(net, ifm->ifi_index);
3291 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3292 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3297 master_dev = netdev_master_upper_dev_get(dev);
3299 m_ops = master_dev->rtnl_link_ops;
3302 err = validate_linkmsg(dev, tb, extack);
3306 if (tb[IFLA_LINKINFO]) {
3307 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3309 ifla_info_policy, NULL);
3313 memset(linkinfo, 0, sizeof(linkinfo));
3315 if (linkinfo[IFLA_INFO_KIND]) {
3316 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3317 ops = rtnl_link_ops_get(kind);
3325 if (ops->maxtype > RTNL_MAX_TYPE)
3328 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3329 err = nla_parse_nested_deprecated(attr, ops->maxtype,
3330 linkinfo[IFLA_INFO_DATA],
3331 ops->policy, extack);
3336 if (ops->validate) {
3337 err = ops->validate(tb, data, extack);
3345 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3348 if (m_ops->slave_maxtype &&
3349 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3350 err = nla_parse_nested_deprecated(slave_attr,
3351 m_ops->slave_maxtype,
3352 linkinfo[IFLA_INFO_SLAVE_DATA],
3353 m_ops->slave_policy,
3357 slave_data = slave_attr;
3364 if (nlh->nlmsg_flags & NLM_F_EXCL)
3366 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3369 if (linkinfo[IFLA_INFO_DATA]) {
3370 if (!ops || ops != dev->rtnl_link_ops ||
3374 err = ops->changelink(dev, tb, data, extack);
3377 status |= DO_SETLINK_NOTIFY;
3380 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3381 if (!m_ops || !m_ops->slave_changelink)
3384 err = m_ops->slave_changelink(master_dev, dev, tb,
3385 slave_data, extack);
3388 status |= DO_SETLINK_NOTIFY;
3391 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3394 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3395 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3396 return rtnl_group_changelink(skb, net,
3397 nla_get_u32(tb[IFLA_GROUP]),
3402 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3406 #ifdef CONFIG_MODULES
3409 request_module("rtnl-link-%s", kind);
3411 ops = rtnl_link_ops_get(kind);
3416 NL_SET_ERR_MSG(extack, "Unknown device type");
3420 if (!ops->alloc && !ops->setup)
3424 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3425 name_assign_type = NET_NAME_ENUM;
3428 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3429 if (IS_ERR(dest_net))
3430 return PTR_ERR(dest_net);
3432 if (tb[IFLA_LINK_NETNSID]) {
3433 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3435 link_net = get_net_ns_by_id(dest_net, id);
3437 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3442 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3448 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3449 name_assign_type, ops, tb, extack);
3455 dev->ifindex = ifm->ifi_index;
3458 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3460 err = register_netdevice(dev);
3466 err = rtnl_configure_link(dev, ifm);
3468 goto out_unregister;
3470 err = dev_change_net_namespace(dev, dest_net, ifname);
3472 goto out_unregister;
3474 if (tb[IFLA_MASTER]) {
3475 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3477 goto out_unregister;
3486 LIST_HEAD(list_kill);
3488 ops->dellink(dev, &list_kill);
3489 unregister_netdevice_many(&list_kill);
3491 unregister_netdevice(dev);
3496 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3497 struct netlink_ext_ack *extack)
3499 struct nlattr **attr;
3502 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3506 ret = __rtnl_newlink(skb, nlh, attr, extack);
3511 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3512 const struct nlmsghdr *nlh,
3514 struct netlink_ext_ack *extack)
3516 struct ifinfomsg *ifm;
3519 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3520 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3524 if (!netlink_strict_get_check(skb))
3525 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3526 ifla_policy, extack);
3528 ifm = nlmsg_data(nlh);
3529 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3531 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3535 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3536 ifla_policy, extack);
3540 for (i = 0; i <= IFLA_MAX; i++) {
3546 case IFLA_ALT_IFNAME:
3548 case IFLA_TARGET_NETNSID:
3551 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3559 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3560 struct netlink_ext_ack *extack)
3562 struct net *net = sock_net(skb->sk);
3563 struct net *tgt_net = net;
3564 struct ifinfomsg *ifm;
3565 struct nlattr *tb[IFLA_MAX+1];
3566 struct net_device *dev = NULL;
3567 struct sk_buff *nskb;
3570 u32 ext_filter_mask = 0;
3572 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3576 err = rtnl_ensure_unique_netns(tb, extack, true);
3580 if (tb[IFLA_TARGET_NETNSID]) {
3581 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3582 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3583 if (IS_ERR(tgt_net))
3584 return PTR_ERR(tgt_net);
3587 if (tb[IFLA_EXT_MASK])
3588 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3591 ifm = nlmsg_data(nlh);
3592 if (ifm->ifi_index > 0)
3593 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3594 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3595 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3596 tb[IFLA_ALT_IFNAME], NULL);
3605 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3609 err = rtnl_fill_ifinfo(nskb, dev, net,
3610 RTM_NEWLINK, NETLINK_CB(skb).portid,
3611 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3612 0, NULL, 0, netnsid, GFP_KERNEL);
3614 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3615 WARN_ON(err == -EMSGSIZE);
3618 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3626 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3627 bool *changed, struct netlink_ext_ack *extack)
3632 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3636 alt_ifname = nla_strdup(attr, GFP_KERNEL);
3640 if (cmd == RTM_NEWLINKPROP) {
3641 err = netdev_name_node_alt_create(dev, alt_ifname);
3644 } else if (cmd == RTM_DELLINKPROP) {
3645 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3657 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3658 struct netlink_ext_ack *extack)
3660 struct net *net = sock_net(skb->sk);
3661 struct nlattr *tb[IFLA_MAX + 1];
3662 struct net_device *dev;
3663 struct ifinfomsg *ifm;
3664 bool changed = false;
3665 struct nlattr *attr;
3668 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3672 err = rtnl_ensure_unique_netns(tb, extack, true);
3676 ifm = nlmsg_data(nlh);
3677 if (ifm->ifi_index > 0)
3678 dev = __dev_get_by_index(net, ifm->ifi_index);
3679 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3680 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3681 tb[IFLA_ALT_IFNAME], NULL);
3688 if (!tb[IFLA_PROP_LIST])
3691 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3692 switch (nla_type(attr)) {
3693 case IFLA_ALT_IFNAME:
3694 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3702 netdev_state_change(dev);
3706 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3707 struct netlink_ext_ack *extack)
3709 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3712 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3713 struct netlink_ext_ack *extack)
3715 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3718 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3720 struct net *net = sock_net(skb->sk);
3721 size_t min_ifinfo_dump_size = 0;
3722 struct nlattr *tb[IFLA_MAX+1];
3723 u32 ext_filter_mask = 0;
3724 struct net_device *dev;
3727 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3728 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3729 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3731 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3732 if (tb[IFLA_EXT_MASK])
3733 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3736 if (!ext_filter_mask)
3737 return NLMSG_GOODSIZE;
3739 * traverse the list of net devices and compute the minimum
3740 * buffer size based upon the filter mask.
3743 for_each_netdev_rcu(net, dev) {
3744 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3745 if_nlmsg_size(dev, ext_filter_mask));
3749 return nlmsg_total_size(min_ifinfo_dump_size);
3752 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3755 int s_idx = cb->family;
3756 int type = cb->nlh->nlmsg_type - RTM_BASE;
3762 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3763 struct rtnl_link __rcu **tab;
3764 struct rtnl_link *link;
3765 rtnl_dumpit_func dumpit;
3767 if (idx < s_idx || idx == PF_PACKET)
3770 if (type < 0 || type >= RTM_NR_MSGTYPES)
3773 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3777 link = rcu_dereference_rtnl(tab[type]);
3781 dumpit = link->dumpit;
3786 memset(&cb->args[0], 0, sizeof(cb->args));
3790 ret = dumpit(skb, cb);
3796 return skb->len ? : ret;
3799 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3800 unsigned int change,
3801 u32 event, gfp_t flags, int *new_nsid,
3804 struct net *net = dev_net(dev);
3805 struct sk_buff *skb;
3807 size_t if_info_size;
3809 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3813 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3814 type, 0, 0, change, 0, 0, event,
3815 new_nsid, new_ifindex, -1, flags);
3817 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3818 WARN_ON(err == -EMSGSIZE);
3825 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3829 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3831 struct net *net = dev_net(dev);
3833 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3836 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3837 unsigned int change, u32 event,
3838 gfp_t flags, int *new_nsid, int new_ifindex)
3840 struct sk_buff *skb;
3842 if (dev->reg_state != NETREG_REGISTERED)
3845 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3848 rtmsg_ifinfo_send(skb, dev, flags);
3851 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3854 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3858 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3859 gfp_t flags, int *new_nsid, int new_ifindex)
3861 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3862 new_nsid, new_ifindex);
3865 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3866 struct net_device *dev,
3867 u8 *addr, u16 vid, u32 pid, u32 seq,
3868 int type, unsigned int flags,
3869 int nlflags, u16 ndm_state)
3871 struct nlmsghdr *nlh;
3874 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3878 ndm = nlmsg_data(nlh);
3879 ndm->ndm_family = AF_BRIDGE;
3882 ndm->ndm_flags = flags;
3884 ndm->ndm_ifindex = dev->ifindex;
3885 ndm->ndm_state = ndm_state;
3887 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3888 goto nla_put_failure;
3890 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3891 goto nla_put_failure;
3893 nlmsg_end(skb, nlh);
3897 nlmsg_cancel(skb, nlh);
3901 static inline size_t rtnl_fdb_nlmsg_size(void)
3903 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3904 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3905 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3909 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3912 struct net *net = dev_net(dev);
3913 struct sk_buff *skb;
3916 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3920 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3921 0, 0, type, NTF_SELF, 0, ndm_state);
3927 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3930 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3934 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3936 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3937 struct nlattr *tb[],
3938 struct net_device *dev,
3939 const unsigned char *addr, u16 vid,
3944 /* If aging addresses are supported device will need to
3945 * implement its own handler for this.
3947 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3948 netdev_info(dev, "default FDB implementation only supports local addresses\n");
3953 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
3957 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3958 err = dev_uc_add_excl(dev, addr);
3959 else if (is_multicast_ether_addr(addr))
3960 err = dev_mc_add_excl(dev, addr);
3962 /* Only return duplicate errors if NLM_F_EXCL is set */
3963 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3968 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3970 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3971 struct netlink_ext_ack *extack)
3976 if (nla_len(vlan_attr) != sizeof(u16)) {
3977 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3981 vid = nla_get_u16(vlan_attr);
3983 if (!vid || vid >= VLAN_VID_MASK) {
3984 NL_SET_ERR_MSG(extack, "invalid vlan id");
3992 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3993 struct netlink_ext_ack *extack)
3995 struct net *net = sock_net(skb->sk);
3997 struct nlattr *tb[NDA_MAX+1];
3998 struct net_device *dev;
4003 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4008 ndm = nlmsg_data(nlh);
4009 if (ndm->ndm_ifindex == 0) {
4010 NL_SET_ERR_MSG(extack, "invalid ifindex");
4014 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4016 NL_SET_ERR_MSG(extack, "unknown ifindex");
4020 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4021 NL_SET_ERR_MSG(extack, "invalid address");
4025 if (dev->type != ARPHRD_ETHER) {
4026 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4030 addr = nla_data(tb[NDA_LLADDR]);
4032 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4038 /* Support fdb on master device the net/bridge default case */
4039 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4040 netif_is_bridge_port(dev)) {
4041 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4042 const struct net_device_ops *ops = br_dev->netdev_ops;
4044 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4045 nlh->nlmsg_flags, extack);
4049 ndm->ndm_flags &= ~NTF_MASTER;
4052 /* Embedded bridge, macvlan, and any other device support */
4053 if ((ndm->ndm_flags & NTF_SELF)) {
4054 if (dev->netdev_ops->ndo_fdb_add)
4055 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4060 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4064 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4066 ndm->ndm_flags &= ~NTF_SELF;
4074 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4076 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4077 struct nlattr *tb[],
4078 struct net_device *dev,
4079 const unsigned char *addr, u16 vid)
4083 /* If aging addresses are supported device will need to
4084 * implement its own handler for this.
4086 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4087 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4091 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4092 err = dev_uc_del(dev, addr);
4093 else if (is_multicast_ether_addr(addr))
4094 err = dev_mc_del(dev, addr);
4098 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4100 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4101 struct netlink_ext_ack *extack)
4103 struct net *net = sock_net(skb->sk);
4105 struct nlattr *tb[NDA_MAX+1];
4106 struct net_device *dev;
4111 if (!netlink_capable(skb, CAP_NET_ADMIN))
4114 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4119 ndm = nlmsg_data(nlh);
4120 if (ndm->ndm_ifindex == 0) {
4121 NL_SET_ERR_MSG(extack, "invalid ifindex");
4125 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4127 NL_SET_ERR_MSG(extack, "unknown ifindex");
4131 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4132 NL_SET_ERR_MSG(extack, "invalid address");
4136 if (dev->type != ARPHRD_ETHER) {
4137 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4141 addr = nla_data(tb[NDA_LLADDR]);
4143 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4149 /* Support fdb on master device the net/bridge default case */
4150 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4151 netif_is_bridge_port(dev)) {
4152 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4153 const struct net_device_ops *ops = br_dev->netdev_ops;
4155 if (ops->ndo_fdb_del)
4156 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4161 ndm->ndm_flags &= ~NTF_MASTER;
4164 /* Embedded bridge, macvlan, and any other device support */
4165 if (ndm->ndm_flags & NTF_SELF) {
4166 if (dev->netdev_ops->ndo_fdb_del)
4167 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4170 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4173 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4175 ndm->ndm_flags &= ~NTF_SELF;
4182 static int nlmsg_populate_fdb(struct sk_buff *skb,
4183 struct netlink_callback *cb,
4184 struct net_device *dev,
4186 struct netdev_hw_addr_list *list)
4188 struct netdev_hw_addr *ha;
4192 portid = NETLINK_CB(cb->skb).portid;
4193 seq = cb->nlh->nlmsg_seq;
4195 list_for_each_entry(ha, &list->list, list) {
4196 if (*idx < cb->args[2])
4199 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4201 RTM_NEWNEIGH, NTF_SELF,
4202 NLM_F_MULTI, NUD_PERMANENT);
4212 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4213 * @skb: socket buffer to store message in
4214 * @cb: netlink callback
4216 * @filter_dev: ignored
4217 * @idx: the number of FDB table entries dumped is added to *@idx
4219 * Default netdevice operation to dump the existing unicast address list.
4220 * Returns number of addresses from list put in skb.
4222 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4223 struct netlink_callback *cb,
4224 struct net_device *dev,
4225 struct net_device *filter_dev,
4230 if (dev->type != ARPHRD_ETHER)
4233 netif_addr_lock_bh(dev);
4234 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4237 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4239 netif_addr_unlock_bh(dev);
4242 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4244 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4245 int *br_idx, int *brport_idx,
4246 struct netlink_ext_ack *extack)
4248 struct nlattr *tb[NDA_MAX + 1];
4252 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4253 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4257 ndm = nlmsg_data(nlh);
4258 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4259 ndm->ndm_flags || ndm->ndm_type) {
4260 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4264 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4265 NDA_MAX, NULL, extack);
4269 *brport_idx = ndm->ndm_ifindex;
4270 for (i = 0; i <= NDA_MAX; ++i) {
4276 if (nla_len(tb[i]) != sizeof(u32)) {
4277 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4280 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4283 if (nla_len(tb[i]) != sizeof(u32)) {
4284 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4287 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4290 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4298 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4299 int *br_idx, int *brport_idx,
4300 struct netlink_ext_ack *extack)
4302 struct nlattr *tb[IFLA_MAX+1];
4305 /* A hack to preserve kernel<->userspace interface.
4306 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4307 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4308 * So, check for ndmsg with an optional u32 attribute (not used here).
4309 * Fortunately these sizes don't conflict with the size of ifinfomsg
4310 * with an optional attribute.
4312 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4313 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4314 nla_attr_size(sizeof(u32)))) {
4315 struct ifinfomsg *ifm;
4317 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4318 tb, IFLA_MAX, ifla_policy,
4322 } else if (err == 0) {
4323 if (tb[IFLA_MASTER])
4324 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4327 ifm = nlmsg_data(nlh);
4328 *brport_idx = ifm->ifi_index;
4333 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4335 struct net_device *dev;
4336 struct net_device *br_dev = NULL;
4337 const struct net_device_ops *ops = NULL;
4338 const struct net_device_ops *cops = NULL;
4339 struct net *net = sock_net(skb->sk);
4340 struct hlist_head *head;
4348 if (cb->strict_check)
4349 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4352 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4358 br_dev = __dev_get_by_index(net, br_idx);
4362 ops = br_dev->netdev_ops;
4366 s_idx = cb->args[1];
4368 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4370 head = &net->dev_index_head[h];
4371 hlist_for_each_entry(dev, head, index_hlist) {
4373 if (brport_idx && (dev->ifindex != brport_idx))
4376 if (!br_idx) { /* user did not specify a specific bridge */
4377 if (netif_is_bridge_port(dev)) {
4378 br_dev = netdev_master_upper_dev_get(dev);
4379 cops = br_dev->netdev_ops;
4382 if (dev != br_dev &&
4383 !netif_is_bridge_port(dev))
4386 if (br_dev != netdev_master_upper_dev_get(dev) &&
4387 !(dev->priv_flags & IFF_EBRIDGE))
4395 if (netif_is_bridge_port(dev)) {
4396 if (cops && cops->ndo_fdb_dump) {
4397 err = cops->ndo_fdb_dump(skb, cb,
4400 if (err == -EMSGSIZE)
4405 if (dev->netdev_ops->ndo_fdb_dump)
4406 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4410 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4412 if (err == -EMSGSIZE)
4417 /* reset fdb offset to 0 for rest of the interfaces */
4433 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4434 struct nlattr **tb, u8 *ndm_flags,
4435 int *br_idx, int *brport_idx, u8 **addr,
4436 u16 *vid, struct netlink_ext_ack *extack)
4441 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4442 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4446 ndm = nlmsg_data(nlh);
4447 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4449 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4453 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4454 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4458 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4459 NDA_MAX, nda_policy, extack);
4463 *ndm_flags = ndm->ndm_flags;
4464 *brport_idx = ndm->ndm_ifindex;
4465 for (i = 0; i <= NDA_MAX; ++i) {
4471 *br_idx = nla_get_u32(tb[i]);
4474 if (nla_len(tb[i]) != ETH_ALEN) {
4475 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4478 *addr = nla_data(tb[i]);
4481 err = fdb_vid_parse(tb[i], vid, extack);
4488 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4496 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4497 struct netlink_ext_ack *extack)
4499 struct net_device *dev = NULL, *br_dev = NULL;
4500 const struct net_device_ops *ops = NULL;
4501 struct net *net = sock_net(in_skb->sk);
4502 struct nlattr *tb[NDA_MAX + 1];
4503 struct sk_buff *skb;
4511 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4512 &brport_idx, &addr, &vid, extack);
4517 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4522 dev = __dev_get_by_index(net, brport_idx);
4524 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4531 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4535 br_dev = __dev_get_by_index(net, br_idx);
4537 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4540 ops = br_dev->netdev_ops;
4544 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4545 if (!netif_is_bridge_port(dev)) {
4546 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4549 br_dev = netdev_master_upper_dev_get(dev);
4551 NL_SET_ERR_MSG(extack, "Master of device not found");
4554 ops = br_dev->netdev_ops;
4556 if (!(ndm_flags & NTF_SELF)) {
4557 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4560 ops = dev->netdev_ops;
4564 if (!br_dev && !dev) {
4565 NL_SET_ERR_MSG(extack, "No device specified");
4569 if (!ops || !ops->ndo_fdb_get) {
4570 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4574 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4580 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4581 NETLINK_CB(in_skb).portid,
4582 nlh->nlmsg_seq, extack);
4586 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4592 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4593 unsigned int attrnum, unsigned int flag)
4596 return nla_put_u8(skb, attrnum, !!(flags & flag));
4600 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4601 struct net_device *dev, u16 mode,
4602 u32 flags, u32 mask, int nlflags,
4604 int (*vlan_fill)(struct sk_buff *skb,
4605 struct net_device *dev,
4608 struct nlmsghdr *nlh;
4609 struct ifinfomsg *ifm;
4610 struct nlattr *br_afspec;
4611 struct nlattr *protinfo;
4612 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4613 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4616 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4620 ifm = nlmsg_data(nlh);
4621 ifm->ifi_family = AF_BRIDGE;
4623 ifm->ifi_type = dev->type;
4624 ifm->ifi_index = dev->ifindex;
4625 ifm->ifi_flags = dev_get_flags(dev);
4626 ifm->ifi_change = 0;
4629 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4630 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4631 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4633 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4635 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4636 (dev->ifindex != dev_get_iflink(dev) &&
4637 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4638 goto nla_put_failure;
4640 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4642 goto nla_put_failure;
4644 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4645 nla_nest_cancel(skb, br_afspec);
4646 goto nla_put_failure;
4649 if (mode != BRIDGE_MODE_UNDEF) {
4650 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4651 nla_nest_cancel(skb, br_afspec);
4652 goto nla_put_failure;
4656 err = vlan_fill(skb, dev, filter_mask);
4658 nla_nest_cancel(skb, br_afspec);
4659 goto nla_put_failure;
4662 nla_nest_end(skb, br_afspec);
4664 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4666 goto nla_put_failure;
4668 if (brport_nla_put_flag(skb, flags, mask,
4669 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4670 brport_nla_put_flag(skb, flags, mask,
4671 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4672 brport_nla_put_flag(skb, flags, mask,
4673 IFLA_BRPORT_FAST_LEAVE,
4674 BR_MULTICAST_FAST_LEAVE) ||
4675 brport_nla_put_flag(skb, flags, mask,
4676 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4677 brport_nla_put_flag(skb, flags, mask,
4678 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4679 brport_nla_put_flag(skb, flags, mask,
4680 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4681 brport_nla_put_flag(skb, flags, mask,
4682 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4683 brport_nla_put_flag(skb, flags, mask,
4684 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4685 brport_nla_put_flag(skb, flags, mask,
4686 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4687 brport_nla_put_flag(skb, flags, mask,
4688 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4689 nla_nest_cancel(skb, protinfo);
4690 goto nla_put_failure;
4693 nla_nest_end(skb, protinfo);
4695 nlmsg_end(skb, nlh);
4698 nlmsg_cancel(skb, nlh);
4699 return err ? err : -EMSGSIZE;
4701 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4703 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4704 bool strict_check, u32 *filter_mask,
4705 struct netlink_ext_ack *extack)
4707 struct nlattr *tb[IFLA_MAX+1];
4711 struct ifinfomsg *ifm;
4713 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4714 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4718 ifm = nlmsg_data(nlh);
4719 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4720 ifm->ifi_change || ifm->ifi_index) {
4721 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4725 err = nlmsg_parse_deprecated_strict(nlh,
4726 sizeof(struct ifinfomsg),
4727 tb, IFLA_MAX, ifla_policy,
4730 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4731 tb, IFLA_MAX, ifla_policy,
4737 /* new attributes should only be added with strict checking */
4738 for (i = 0; i <= IFLA_MAX; ++i) {
4744 *filter_mask = nla_get_u32(tb[i]);
4748 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4757 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4759 const struct nlmsghdr *nlh = cb->nlh;
4760 struct net *net = sock_net(skb->sk);
4761 struct net_device *dev;
4763 u32 portid = NETLINK_CB(cb->skb).portid;
4764 u32 seq = nlh->nlmsg_seq;
4765 u32 filter_mask = 0;
4768 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4770 if (err < 0 && cb->strict_check)
4774 for_each_netdev_rcu(net, dev) {
4775 const struct net_device_ops *ops = dev->netdev_ops;
4776 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4778 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4779 if (idx >= cb->args[0]) {
4780 err = br_dev->netdev_ops->ndo_bridge_getlink(
4781 skb, portid, seq, dev,
4782 filter_mask, NLM_F_MULTI);
4783 if (err < 0 && err != -EOPNOTSUPP) {
4784 if (likely(skb->len))
4793 if (ops->ndo_bridge_getlink) {
4794 if (idx >= cb->args[0]) {
4795 err = ops->ndo_bridge_getlink(skb, portid,
4799 if (err < 0 && err != -EOPNOTSUPP) {
4800 if (likely(skb->len))
4817 static inline size_t bridge_nlmsg_size(void)
4819 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4820 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4821 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4822 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4823 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
4824 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
4825 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
4826 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
4827 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4828 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
4829 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
4832 static int rtnl_bridge_notify(struct net_device *dev)
4834 struct net *net = dev_net(dev);
4835 struct sk_buff *skb;
4836 int err = -EOPNOTSUPP;
4838 if (!dev->netdev_ops->ndo_bridge_getlink)
4841 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4847 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4851 /* Notification info is only filled for bridge ports, not the bridge
4852 * device itself. Therefore, a zero notification length is valid and
4853 * should not result in an error.
4858 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4861 WARN_ON(err == -EMSGSIZE);
4864 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4868 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4869 struct netlink_ext_ack *extack)
4871 struct net *net = sock_net(skb->sk);
4872 struct ifinfomsg *ifm;
4873 struct net_device *dev;
4874 struct nlattr *br_spec, *attr = NULL;
4875 int rem, err = -EOPNOTSUPP;
4877 bool have_flags = false;
4879 if (nlmsg_len(nlh) < sizeof(*ifm))
4882 ifm = nlmsg_data(nlh);
4883 if (ifm->ifi_family != AF_BRIDGE)
4884 return -EPFNOSUPPORT;
4886 dev = __dev_get_by_index(net, ifm->ifi_index);
4888 NL_SET_ERR_MSG(extack, "unknown ifindex");
4892 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4894 nla_for_each_nested(attr, br_spec, rem) {
4895 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4896 if (nla_len(attr) < sizeof(flags))
4900 flags = nla_get_u16(attr);
4906 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4907 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4909 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4914 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4919 flags &= ~BRIDGE_FLAGS_MASTER;
4922 if ((flags & BRIDGE_FLAGS_SELF)) {
4923 if (!dev->netdev_ops->ndo_bridge_setlink)
4926 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4930 flags &= ~BRIDGE_FLAGS_SELF;
4932 /* Generate event to notify upper layer of bridge
4935 err = rtnl_bridge_notify(dev);
4940 memcpy(nla_data(attr), &flags, sizeof(flags));
4945 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4946 struct netlink_ext_ack *extack)
4948 struct net *net = sock_net(skb->sk);
4949 struct ifinfomsg *ifm;
4950 struct net_device *dev;
4951 struct nlattr *br_spec, *attr = NULL;
4952 int rem, err = -EOPNOTSUPP;
4954 bool have_flags = false;
4956 if (nlmsg_len(nlh) < sizeof(*ifm))
4959 ifm = nlmsg_data(nlh);
4960 if (ifm->ifi_family != AF_BRIDGE)
4961 return -EPFNOSUPPORT;
4963 dev = __dev_get_by_index(net, ifm->ifi_index);
4965 NL_SET_ERR_MSG(extack, "unknown ifindex");
4969 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4971 nla_for_each_nested(attr, br_spec, rem) {
4972 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4973 if (nla_len(attr) < sizeof(flags))
4977 flags = nla_get_u16(attr);
4983 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4984 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4986 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4991 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4995 flags &= ~BRIDGE_FLAGS_MASTER;
4998 if ((flags & BRIDGE_FLAGS_SELF)) {
4999 if (!dev->netdev_ops->ndo_bridge_dellink)
5002 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5006 flags &= ~BRIDGE_FLAGS_SELF;
5008 /* Generate event to notify upper layer of bridge
5011 err = rtnl_bridge_notify(dev);
5016 memcpy(nla_data(attr), &flags, sizeof(flags));
5021 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5023 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5024 (!idxattr || idxattr == attrid);
5027 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5028 static int rtnl_get_offload_stats_attr_size(int attr_id)
5031 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5032 return sizeof(struct rtnl_link_stats64);
5038 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5041 struct nlattr *attr = NULL;
5046 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5047 dev->netdev_ops->ndo_get_offload_stats))
5050 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5051 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5052 if (attr_id < *prividx)
5055 size = rtnl_get_offload_stats_attr_size(attr_id);
5059 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5062 attr = nla_reserve_64bit(skb, attr_id, size,
5063 IFLA_OFFLOAD_XSTATS_UNSPEC);
5065 goto nla_put_failure;
5067 attr_data = nla_data(attr);
5068 memset(attr_data, 0, size);
5069 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5072 goto get_offload_stats_failure;
5083 get_offload_stats_failure:
5088 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5094 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5095 dev->netdev_ops->ndo_get_offload_stats))
5098 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5099 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5100 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5102 size = rtnl_get_offload_stats_attr_size(attr_id);
5103 nla_size += nla_total_size_64bit(size);
5107 nla_size += nla_total_size(0);
5112 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5113 int type, u32 pid, u32 seq, u32 change,
5114 unsigned int flags, unsigned int filter_mask,
5115 int *idxattr, int *prividx)
5117 struct if_stats_msg *ifsm;
5118 struct nlmsghdr *nlh;
5119 struct nlattr *attr;
5120 int s_prividx = *prividx;
5125 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5129 ifsm = nlmsg_data(nlh);
5130 ifsm->family = PF_UNSPEC;
5133 ifsm->ifindex = dev->ifindex;
5134 ifsm->filter_mask = filter_mask;
5136 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5137 struct rtnl_link_stats64 *sp;
5139 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5140 sizeof(struct rtnl_link_stats64),
5143 goto nla_put_failure;
5145 sp = nla_data(attr);
5146 dev_get_stats(dev, sp);
5149 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5150 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5152 if (ops && ops->fill_linkxstats) {
5153 *idxattr = IFLA_STATS_LINK_XSTATS;
5154 attr = nla_nest_start_noflag(skb,
5155 IFLA_STATS_LINK_XSTATS);
5157 goto nla_put_failure;
5159 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5160 nla_nest_end(skb, attr);
5162 goto nla_put_failure;
5167 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5169 const struct rtnl_link_ops *ops = NULL;
5170 const struct net_device *master;
5172 master = netdev_master_upper_dev_get(dev);
5174 ops = master->rtnl_link_ops;
5175 if (ops && ops->fill_linkxstats) {
5176 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5177 attr = nla_nest_start_noflag(skb,
5178 IFLA_STATS_LINK_XSTATS_SLAVE);
5180 goto nla_put_failure;
5182 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5183 nla_nest_end(skb, attr);
5185 goto nla_put_failure;
5190 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5192 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5193 attr = nla_nest_start_noflag(skb,
5194 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5196 goto nla_put_failure;
5198 err = rtnl_get_offload_stats(skb, dev, prividx);
5199 if (err == -ENODATA)
5200 nla_nest_cancel(skb, attr);
5202 nla_nest_end(skb, attr);
5204 if (err && err != -ENODATA)
5205 goto nla_put_failure;
5209 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5210 struct rtnl_af_ops *af_ops;
5212 *idxattr = IFLA_STATS_AF_SPEC;
5213 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5215 goto nla_put_failure;
5218 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5219 if (af_ops->fill_stats_af) {
5223 af = nla_nest_start_noflag(skb,
5227 goto nla_put_failure;
5229 err = af_ops->fill_stats_af(skb, dev);
5231 if (err == -ENODATA) {
5232 nla_nest_cancel(skb, af);
5233 } else if (err < 0) {
5235 goto nla_put_failure;
5238 nla_nest_end(skb, af);
5243 nla_nest_end(skb, attr);
5248 nlmsg_end(skb, nlh);
5253 /* not a multi message or no progress mean a real error */
5254 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5255 nlmsg_cancel(skb, nlh);
5257 nlmsg_end(skb, nlh);
5262 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5265 size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5267 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5268 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5270 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5271 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5272 int attr = IFLA_STATS_LINK_XSTATS;
5274 if (ops && ops->get_linkxstats_size) {
5275 size += nla_total_size(ops->get_linkxstats_size(dev,
5277 /* for IFLA_STATS_LINK_XSTATS */
5278 size += nla_total_size(0);
5282 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5283 struct net_device *_dev = (struct net_device *)dev;
5284 const struct rtnl_link_ops *ops = NULL;
5285 const struct net_device *master;
5287 /* netdev_master_upper_dev_get can't take const */
5288 master = netdev_master_upper_dev_get(_dev);
5290 ops = master->rtnl_link_ops;
5291 if (ops && ops->get_linkxstats_size) {
5292 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5294 size += nla_total_size(ops->get_linkxstats_size(dev,
5296 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5297 size += nla_total_size(0);
5301 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5302 size += rtnl_get_offload_stats_size(dev);
5304 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5305 struct rtnl_af_ops *af_ops;
5307 /* for IFLA_STATS_AF_SPEC */
5308 size += nla_total_size(0);
5311 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5312 if (af_ops->get_stats_af_size) {
5313 size += nla_total_size(
5314 af_ops->get_stats_af_size(dev));
5317 size += nla_total_size(0);
5326 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5327 bool is_dump, struct netlink_ext_ack *extack)
5329 struct if_stats_msg *ifsm;
5331 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5332 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5339 ifsm = nlmsg_data(nlh);
5341 /* only requests using strict checks can pass data to influence
5342 * the dump. The legacy exception is filter_mask.
5344 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5345 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5348 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5349 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5352 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5353 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5360 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5361 struct netlink_ext_ack *extack)
5363 struct net *net = sock_net(skb->sk);
5364 struct net_device *dev = NULL;
5365 int idxattr = 0, prividx = 0;
5366 struct if_stats_msg *ifsm;
5367 struct sk_buff *nskb;
5371 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5376 ifsm = nlmsg_data(nlh);
5377 if (ifsm->ifindex > 0)
5378 dev = __dev_get_by_index(net, ifsm->ifindex);
5385 filter_mask = ifsm->filter_mask;
5389 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5393 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5394 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5395 0, filter_mask, &idxattr, &prividx);
5397 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5398 WARN_ON(err == -EMSGSIZE);
5401 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5407 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5409 struct netlink_ext_ack *extack = cb->extack;
5410 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5411 struct net *net = sock_net(skb->sk);
5412 unsigned int flags = NLM_F_MULTI;
5413 struct if_stats_msg *ifsm;
5414 struct hlist_head *head;
5415 struct net_device *dev;
5416 u32 filter_mask = 0;
5420 s_idx = cb->args[1];
5421 s_idxattr = cb->args[2];
5422 s_prividx = cb->args[3];
5424 cb->seq = net->dev_base_seq;
5426 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5430 ifsm = nlmsg_data(cb->nlh);
5431 filter_mask = ifsm->filter_mask;
5433 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5437 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5439 head = &net->dev_index_head[h];
5440 hlist_for_each_entry(dev, head, index_hlist) {
5443 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5444 NETLINK_CB(cb->skb).portid,
5445 cb->nlh->nlmsg_seq, 0,
5447 &s_idxattr, &s_prividx);
5448 /* If we ran out of room on the first message,
5451 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5457 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5463 cb->args[3] = s_prividx;
5464 cb->args[2] = s_idxattr;
5471 /* Process one rtnetlink message. */
5473 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5474 struct netlink_ext_ack *extack)
5476 struct net *net = sock_net(skb->sk);
5477 struct rtnl_link *link;
5478 struct module *owner;
5479 int err = -EOPNOTSUPP;
5480 rtnl_doit_func doit;
5486 type = nlh->nlmsg_type;
5492 /* All the messages must have at least 1 byte length */
5493 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5496 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5499 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5503 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5505 rtnl_dumpit_func dumpit;
5506 u32 min_dump_alloc = 0;
5508 link = rtnl_get_link(family, type);
5509 if (!link || !link->dumpit) {
5511 link = rtnl_get_link(family, type);
5512 if (!link || !link->dumpit)
5515 owner = link->owner;
5516 dumpit = link->dumpit;
5518 if (type == RTM_GETLINK - RTM_BASE)
5519 min_dump_alloc = rtnl_calcit(skb, nlh);
5522 /* need to do this before rcu_read_unlock() */
5523 if (!try_module_get(owner))
5524 err = -EPROTONOSUPPORT;
5530 struct netlink_dump_control c = {
5532 .min_dump_alloc = min_dump_alloc,
5535 err = netlink_dump_start(rtnl, skb, nlh, &c);
5536 /* netlink_dump_start() will keep a reference on
5537 * module if dump is still in progress.
5544 link = rtnl_get_link(family, type);
5545 if (!link || !link->doit) {
5547 link = rtnl_get_link(PF_UNSPEC, type);
5548 if (!link || !link->doit)
5552 owner = link->owner;
5553 if (!try_module_get(owner)) {
5554 err = -EPROTONOSUPPORT;
5558 flags = link->flags;
5559 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5563 err = doit(skb, nlh, extack);
5570 link = rtnl_get_link(family, type);
5571 if (link && link->doit)
5572 err = link->doit(skb, nlh, extack);
5588 static void rtnetlink_rcv(struct sk_buff *skb)
5590 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5593 static int rtnetlink_bind(struct net *net, int group)
5596 case RTNLGRP_IPV4_MROUTE_R:
5597 case RTNLGRP_IPV6_MROUTE_R:
5598 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5605 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5607 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5611 case NETDEV_CHANGEMTU:
5612 case NETDEV_CHANGEADDR:
5613 case NETDEV_CHANGENAME:
5614 case NETDEV_FEAT_CHANGE:
5615 case NETDEV_BONDING_FAILOVER:
5616 case NETDEV_POST_TYPE_CHANGE:
5617 case NETDEV_NOTIFY_PEERS:
5618 case NETDEV_CHANGEUPPER:
5619 case NETDEV_RESEND_IGMP:
5620 case NETDEV_CHANGEINFODATA:
5621 case NETDEV_CHANGELOWERSTATE:
5622 case NETDEV_CHANGE_TX_QUEUE_LEN:
5623 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5624 GFP_KERNEL, NULL, 0);
5632 static struct notifier_block rtnetlink_dev_notifier = {
5633 .notifier_call = rtnetlink_event,
5637 static int __net_init rtnetlink_net_init(struct net *net)
5640 struct netlink_kernel_cfg cfg = {
5641 .groups = RTNLGRP_MAX,
5642 .input = rtnetlink_rcv,
5643 .cb_mutex = &rtnl_mutex,
5644 .flags = NL_CFG_F_NONROOT_RECV,
5645 .bind = rtnetlink_bind,
5648 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5655 static void __net_exit rtnetlink_net_exit(struct net *net)
5657 netlink_kernel_release(net->rtnl);
5661 static struct pernet_operations rtnetlink_net_ops = {
5662 .init = rtnetlink_net_init,
5663 .exit = rtnetlink_net_exit,
5666 void __init rtnetlink_init(void)
5668 if (register_pernet_subsys(&rtnetlink_net_ops))
5669 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5671 register_netdevice_notifier(&rtnetlink_dev_notifier);
5673 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5674 rtnl_dump_ifinfo, 0);
5675 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5676 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5677 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5679 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5680 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5681 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5683 rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5684 rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5686 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5687 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5688 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5690 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5691 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5692 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5694 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,