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 arithmetics 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 36
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 *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 **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]);
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_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 modifiy 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 modifiy 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 **tab, *link;
292 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
293 msgindex = rtm_msgindex(msgtype);
296 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
302 link = tab[msgindex];
303 rcu_assign_pointer(tab[msgindex], NULL);
306 kfree_rcu(link, rcu);
310 EXPORT_SYMBOL_GPL(rtnl_unregister);
313 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
314 * @protocol : Protocol family or PF_UNSPEC
316 * Identical to calling rtnl_unregster() for all registered message types
317 * of a certain protocol family.
319 void rtnl_unregister_all(int protocol)
321 struct rtnl_link **tab, *link;
324 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
327 tab = rtnl_msg_handlers[protocol];
332 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
333 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
334 link = tab[msgindex];
338 rcu_assign_pointer(tab[msgindex], NULL);
339 kfree_rcu(link, rcu);
347 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
349 static LIST_HEAD(link_ops);
351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
353 const struct rtnl_link_ops *ops;
355 list_for_each_entry(ops, &link_ops, list) {
356 if (!strcmp(ops->kind, kind))
363 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
364 * @ops: struct rtnl_link_ops * to register
366 * The caller must hold the rtnl_mutex. This function should be used
367 * by drivers that create devices during module initialization. It
368 * must be called before registering the devices.
370 * Returns 0 on success or a negative error code.
372 int __rtnl_link_register(struct rtnl_link_ops *ops)
374 if (rtnl_link_ops_get(ops->kind))
377 /* The check for setup is here because if ops
378 * does not have that filled up, it is not possible
379 * to use the ops for creating device. So do not
380 * fill up dellink as well. That disables rtnl_dellink.
382 if (ops->setup && !ops->dellink)
383 ops->dellink = unregister_netdevice_queue;
385 list_add_tail(&ops->list, &link_ops);
388 EXPORT_SYMBOL_GPL(__rtnl_link_register);
391 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
392 * @ops: struct rtnl_link_ops * to register
394 * Returns 0 on success or a negative error code.
396 int rtnl_link_register(struct rtnl_link_ops *ops)
400 /* Sanity-check max sizes to avoid stack buffer overflow. */
401 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
402 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
406 err = __rtnl_link_register(ops);
410 EXPORT_SYMBOL_GPL(rtnl_link_register);
412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
414 struct net_device *dev;
415 LIST_HEAD(list_kill);
417 for_each_netdev(net, dev) {
418 if (dev->rtnl_link_ops == ops)
419 ops->dellink(dev, &list_kill);
421 unregister_netdevice_many(&list_kill);
425 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
426 * @ops: struct rtnl_link_ops * to unregister
428 * The caller must hold the rtnl_mutex and guarantee net_namespace_list
429 * integrity (hold pernet_ops_rwsem for writing to close the race
430 * with setup_net() and cleanup_net()).
432 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
437 __rtnl_kill_links(net, ops);
439 list_del(&ops->list);
441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
443 /* Return with the rtnl_lock held when there are no network
444 * devices unregistering in any network namespace.
446 static void rtnl_lock_unregistering_all(void)
450 DEFINE_WAIT_FUNC(wait, woken_wake_function);
452 add_wait_queue(&netdev_unregistering_wq, &wait);
454 unregistering = false;
456 /* We held write locked pernet_ops_rwsem, and parallel
457 * setup_net() and cleanup_net() are not possible.
460 if (net->dev_unreg_count > 0) {
461 unregistering = true;
469 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
471 remove_wait_queue(&netdev_unregistering_wq, &wait);
475 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
476 * @ops: struct rtnl_link_ops * to unregister
478 void rtnl_link_unregister(struct rtnl_link_ops *ops)
480 /* Close the race with setup_net() and cleanup_net() */
481 down_write(&pernet_ops_rwsem);
482 rtnl_lock_unregistering_all();
483 __rtnl_link_unregister(ops);
485 up_write(&pernet_ops_rwsem);
487 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
491 struct net_device *master_dev;
492 const struct rtnl_link_ops *ops;
497 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
501 ops = master_dev->rtnl_link_ops;
502 if (!ops || !ops->get_slave_size)
504 /* IFLA_INFO_SLAVE_DATA + nested data */
505 size = nla_total_size(sizeof(struct nlattr)) +
506 ops->get_slave_size(master_dev, dev);
513 static size_t rtnl_link_get_size(const struct net_device *dev)
515 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
521 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
522 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
525 /* IFLA_INFO_DATA + nested data */
526 size += nla_total_size(sizeof(struct nlattr)) +
529 if (ops->get_xstats_size)
530 /* IFLA_INFO_XSTATS */
531 size += nla_total_size(ops->get_xstats_size(dev));
533 size += rtnl_link_get_slave_info_data_size(dev);
538 static LIST_HEAD(rtnl_af_ops);
540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
542 const struct rtnl_af_ops *ops;
544 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
545 if (ops->family == family)
553 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
554 * @ops: struct rtnl_af_ops * to register
556 * Returns 0 on success or a negative error code.
558 void rtnl_af_register(struct rtnl_af_ops *ops)
561 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
564 EXPORT_SYMBOL_GPL(rtnl_af_register);
567 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
568 * @ops: struct rtnl_af_ops * to unregister
570 void rtnl_af_unregister(struct rtnl_af_ops *ops)
573 list_del_rcu(&ops->list);
578 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
580 static size_t rtnl_link_get_af_size(const struct net_device *dev,
583 struct rtnl_af_ops *af_ops;
587 size = nla_total_size(sizeof(struct nlattr));
590 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
591 if (af_ops->get_link_af_size) {
592 /* AF_* + nested data */
593 size += nla_total_size(sizeof(struct nlattr)) +
594 af_ops->get_link_af_size(dev, ext_filter_mask);
602 static bool rtnl_have_link_slave_info(const struct net_device *dev)
604 struct net_device *master_dev;
609 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
610 if (master_dev && master_dev->rtnl_link_ops)
616 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
617 const struct net_device *dev)
619 struct net_device *master_dev;
620 const struct rtnl_link_ops *ops;
621 struct nlattr *slave_data;
624 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
627 ops = master_dev->rtnl_link_ops;
630 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
632 if (ops->fill_slave_info) {
633 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
636 err = ops->fill_slave_info(skb, master_dev, dev);
638 goto err_cancel_slave_data;
639 nla_nest_end(skb, slave_data);
643 err_cancel_slave_data:
644 nla_nest_cancel(skb, slave_data);
648 static int rtnl_link_info_fill(struct sk_buff *skb,
649 const struct net_device *dev)
651 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
657 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
659 if (ops->fill_xstats) {
660 err = ops->fill_xstats(skb, dev);
664 if (ops->fill_info) {
665 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
668 err = ops->fill_info(skb, dev);
670 goto err_cancel_data;
671 nla_nest_end(skb, data);
676 nla_nest_cancel(skb, data);
680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
682 struct nlattr *linkinfo;
685 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
686 if (linkinfo == NULL)
689 err = rtnl_link_info_fill(skb, dev);
691 goto err_cancel_link;
693 err = rtnl_link_slave_info_fill(skb, dev);
695 goto err_cancel_link;
697 nla_nest_end(skb, linkinfo);
701 nla_nest_cancel(skb, linkinfo);
706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
708 struct sock *rtnl = net->rtnl;
711 NETLINK_CB(skb).dst_group = group;
713 refcount_inc(&skb->users);
714 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
716 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
722 struct sock *rtnl = net->rtnl;
724 return nlmsg_unicast(rtnl, skb, pid);
726 EXPORT_SYMBOL(rtnl_unicast);
728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
729 struct nlmsghdr *nlh, gfp_t flags)
731 struct sock *rtnl = net->rtnl;
735 report = nlmsg_report(nlh);
737 nlmsg_notify(rtnl, skb, pid, group, report, flags);
739 EXPORT_SYMBOL(rtnl_notify);
741 void rtnl_set_sk_err(struct net *net, u32 group, int error)
743 struct sock *rtnl = net->rtnl;
745 netlink_set_err(rtnl, 0, group, error);
747 EXPORT_SYMBOL(rtnl_set_sk_err);
749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
754 /* nothing is dumped for dst_default_metrics, so just skip the loop */
755 if (metrics == dst_default_metrics.metrics)
758 mx = nla_nest_start_noflag(skb, RTA_METRICS);
762 for (i = 0; i < RTAX_MAX; i++) {
764 if (i == RTAX_CC_ALGO - 1) {
765 char tmp[TCP_CA_NAME_MAX], *name;
767 name = tcp_ca_get_name_by_key(metrics[i], tmp);
770 if (nla_put_string(skb, i + 1, name))
771 goto nla_put_failure;
772 } else if (i == RTAX_FEATURES - 1) {
773 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
777 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
778 if (nla_put_u32(skb, i + 1, user_features))
779 goto nla_put_failure;
781 if (nla_put_u32(skb, i + 1, metrics[i]))
782 goto nla_put_failure;
789 nla_nest_cancel(skb, mx);
793 return nla_nest_end(skb, mx);
796 nla_nest_cancel(skb, mx);
799 EXPORT_SYMBOL(rtnetlink_put_metrics);
801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
802 long expires, u32 error)
804 struct rta_cacheinfo ci = {
810 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
811 ci.rta_used = dst->__use;
812 ci.rta_clntref = atomic_read(&dst->__refcnt);
817 clock = jiffies_to_clock_t(abs(expires));
818 clock = min_t(unsigned long, clock, INT_MAX);
819 ci.rta_expires = (expires > 0) ? clock : -clock;
821 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
825 static void set_operstate(struct net_device *dev, unsigned char transition)
827 unsigned char operstate = dev->operstate;
829 switch (transition) {
831 if ((operstate == IF_OPER_DORMANT ||
832 operstate == IF_OPER_UNKNOWN) &&
834 operstate = IF_OPER_UP;
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 noinline size_t if_nlmsg_size(const struct net_device *dev,
986 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
987 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
988 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
989 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
990 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
991 + nla_total_size(sizeof(struct rtnl_link_stats))
992 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
993 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
994 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
995 + nla_total_size(4) /* IFLA_TXQLEN */
996 + nla_total_size(4) /* IFLA_WEIGHT */
997 + nla_total_size(4) /* IFLA_MTU */
998 + nla_total_size(4) /* IFLA_LINK */
999 + nla_total_size(4) /* IFLA_MASTER */
1000 + nla_total_size(1) /* IFLA_CARRIER */
1001 + nla_total_size(4) /* IFLA_PROMISCUITY */
1002 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1003 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1004 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1005 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1006 + nla_total_size(1) /* IFLA_OPERSTATE */
1007 + nla_total_size(1) /* IFLA_LINKMODE */
1008 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1009 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1010 + nla_total_size(4) /* IFLA_GROUP */
1011 + nla_total_size(ext_filter_mask
1012 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1013 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1014 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1015 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1016 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1017 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1018 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1019 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1020 + rtnl_xdp_size() /* IFLA_XDP */
1021 + nla_total_size(4) /* IFLA_EVENT */
1022 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1023 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1024 + nla_total_size(1) /* IFLA_PROTO_DOWN */
1025 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1026 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1027 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1028 + nla_total_size(4) /* IFLA_MIN_MTU */
1029 + nla_total_size(4) /* IFLA_MAX_MTU */
1033 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1035 struct nlattr *vf_ports;
1036 struct nlattr *vf_port;
1040 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1044 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1045 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1047 goto nla_put_failure;
1048 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1049 goto nla_put_failure;
1050 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1051 if (err == -EMSGSIZE)
1052 goto nla_put_failure;
1054 nla_nest_cancel(skb, vf_port);
1057 nla_nest_end(skb, vf_port);
1060 nla_nest_end(skb, vf_ports);
1065 nla_nest_cancel(skb, vf_ports);
1069 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1071 struct nlattr *port_self;
1074 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1078 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1080 nla_nest_cancel(skb, port_self);
1081 return (err == -EMSGSIZE) ? err : 0;
1084 nla_nest_end(skb, port_self);
1089 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1090 u32 ext_filter_mask)
1094 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1095 !(ext_filter_mask & RTEXT_FILTER_VF))
1098 err = rtnl_port_self_fill(skb, dev);
1102 if (dev_num_vf(dev->dev.parent)) {
1103 err = rtnl_vf_ports_fill(skb, dev);
1111 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1114 struct netdev_phys_item_id ppid;
1116 err = dev_get_phys_port_id(dev, &ppid);
1118 if (err == -EOPNOTSUPP)
1123 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1129 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1131 char name[IFNAMSIZ];
1134 err = dev_get_phys_port_name(dev, name, sizeof(name));
1136 if (err == -EOPNOTSUPP)
1141 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1147 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1149 struct netdev_phys_item_id ppid = { };
1152 err = dev_get_port_parent_id(dev, &ppid, false);
1154 if (err == -EOPNOTSUPP)
1159 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1165 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1166 struct net_device *dev)
1168 struct rtnl_link_stats64 *sp;
1169 struct nlattr *attr;
1171 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1172 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1176 sp = nla_data(attr);
1177 dev_get_stats(dev, sp);
1179 attr = nla_reserve(skb, IFLA_STATS,
1180 sizeof(struct rtnl_link_stats));
1184 copy_rtnl_link_stats(nla_data(attr), sp);
1189 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1190 struct net_device *dev,
1192 struct nlattr *vfinfo)
1194 struct ifla_vf_rss_query_en vf_rss_query_en;
1195 struct nlattr *vf, *vfstats, *vfvlanlist;
1196 struct ifla_vf_link_state vf_linkstate;
1197 struct ifla_vf_vlan_info vf_vlan_info;
1198 struct ifla_vf_spoofchk vf_spoofchk;
1199 struct ifla_vf_tx_rate vf_tx_rate;
1200 struct ifla_vf_stats vf_stats;
1201 struct ifla_vf_trust vf_trust;
1202 struct ifla_vf_vlan vf_vlan;
1203 struct ifla_vf_rate vf_rate;
1204 struct ifla_vf_mac vf_mac;
1205 struct ifla_vf_broadcast vf_broadcast;
1206 struct ifla_vf_info ivi;
1208 memset(&ivi, 0, sizeof(ivi));
1210 /* Not all SR-IOV capable drivers support the
1211 * spoofcheck and "RSS query enable" query. Preset to
1212 * -1 so the user space tool can detect that the driver
1213 * didn't report anything.
1216 ivi.rss_query_en = -1;
1218 /* The default value for VF link state is "auto"
1219 * IFLA_VF_LINK_STATE_AUTO which equals zero
1222 /* VLAN Protocol by default is 802.1Q */
1223 ivi.vlan_proto = htons(ETH_P_8021Q);
1224 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1227 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1236 vf_rss_query_en.vf =
1237 vf_trust.vf = ivi.vf;
1239 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1240 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1241 vf_vlan.vlan = ivi.vlan;
1242 vf_vlan.qos = ivi.qos;
1243 vf_vlan_info.vlan = ivi.vlan;
1244 vf_vlan_info.qos = ivi.qos;
1245 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1246 vf_tx_rate.rate = ivi.max_tx_rate;
1247 vf_rate.min_tx_rate = ivi.min_tx_rate;
1248 vf_rate.max_tx_rate = ivi.max_tx_rate;
1249 vf_spoofchk.setting = ivi.spoofchk;
1250 vf_linkstate.link_state = ivi.linkstate;
1251 vf_rss_query_en.setting = ivi.rss_query_en;
1252 vf_trust.setting = ivi.trusted;
1253 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1255 goto nla_put_vfinfo_failure;
1256 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1257 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1258 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1259 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1261 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1263 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1265 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1267 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1268 sizeof(vf_rss_query_en),
1269 &vf_rss_query_en) ||
1270 nla_put(skb, IFLA_VF_TRUST,
1271 sizeof(vf_trust), &vf_trust))
1272 goto nla_put_vf_failure;
1273 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1275 goto nla_put_vf_failure;
1276 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1278 nla_nest_cancel(skb, vfvlanlist);
1279 goto nla_put_vf_failure;
1281 nla_nest_end(skb, vfvlanlist);
1282 memset(&vf_stats, 0, sizeof(vf_stats));
1283 if (dev->netdev_ops->ndo_get_vf_stats)
1284 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1286 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1288 goto nla_put_vf_failure;
1289 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1290 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1291 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1292 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1293 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1294 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1295 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1296 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1297 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1298 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1299 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1300 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1301 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1302 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1303 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1304 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1305 nla_nest_cancel(skb, vfstats);
1306 goto nla_put_vf_failure;
1308 nla_nest_end(skb, vfstats);
1309 nla_nest_end(skb, vf);
1313 nla_nest_cancel(skb, vf);
1314 nla_put_vfinfo_failure:
1315 nla_nest_cancel(skb, vfinfo);
1319 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1320 struct net_device *dev,
1321 u32 ext_filter_mask)
1323 struct nlattr *vfinfo;
1326 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1329 num_vfs = dev_num_vf(dev->dev.parent);
1330 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1333 if (!dev->netdev_ops->ndo_get_vf_config)
1336 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1340 for (i = 0; i < num_vfs; i++) {
1341 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1345 nla_nest_end(skb, vfinfo);
1349 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1351 struct rtnl_link_ifmap map;
1353 memset(&map, 0, sizeof(map));
1354 map.mem_start = dev->mem_start;
1355 map.mem_end = dev->mem_end;
1356 map.base_addr = dev->base_addr;
1359 map.port = dev->if_port;
1361 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1367 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1369 const struct bpf_prog *generic_xdp_prog;
1373 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1374 if (!generic_xdp_prog)
1376 return generic_xdp_prog->aux->id;
1379 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1381 return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG);
1384 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1386 return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf,
1390 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1391 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1392 u32 (*get_prog_id)(struct net_device *dev))
1397 curr_id = get_prog_id(dev);
1402 err = nla_put_u32(skb, attr, curr_id);
1406 if (*mode != XDP_ATTACHED_NONE)
1407 *mode = XDP_ATTACHED_MULTI;
1414 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1421 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1426 mode = XDP_ATTACHED_NONE;
1427 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1428 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1431 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1432 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1435 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1436 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1440 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1444 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1445 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1450 nla_nest_end(skb, xdp);
1454 nla_nest_cancel(skb, xdp);
1458 static u32 rtnl_get_event(unsigned long event)
1460 u32 rtnl_event_type = IFLA_EVENT_NONE;
1464 rtnl_event_type = IFLA_EVENT_REBOOT;
1466 case NETDEV_FEAT_CHANGE:
1467 rtnl_event_type = IFLA_EVENT_FEATURES;
1469 case NETDEV_BONDING_FAILOVER:
1470 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1472 case NETDEV_NOTIFY_PEERS:
1473 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1475 case NETDEV_RESEND_IGMP:
1476 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1478 case NETDEV_CHANGEINFODATA:
1479 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1485 return rtnl_event_type;
1488 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1490 const struct net_device *upper_dev;
1495 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1497 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1503 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1506 int ifindex = dev_get_iflink(dev);
1508 if (force || dev->ifindex != ifindex)
1509 return nla_put_u32(skb, IFLA_LINK, ifindex);
1514 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1515 struct net_device *dev)
1520 ret = dev_get_alias(dev, buf, sizeof(buf));
1521 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1524 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1525 const struct net_device *dev,
1526 struct net *src_net)
1528 bool put_iflink = false;
1530 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1531 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1533 if (!net_eq(dev_net(dev), link_net)) {
1534 int id = peernet2id_alloc(src_net, link_net);
1536 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1543 return nla_put_iflink(skb, dev, put_iflink);
1546 static int rtnl_fill_link_af(struct sk_buff *skb,
1547 const struct net_device *dev,
1548 u32 ext_filter_mask)
1550 const struct rtnl_af_ops *af_ops;
1551 struct nlattr *af_spec;
1553 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1557 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1561 if (!af_ops->fill_link_af)
1564 af = nla_nest_start_noflag(skb, af_ops->family);
1568 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1570 * Caller may return ENODATA to indicate that there
1571 * was no data to be dumped. This is not an error, it
1572 * means we should trim the attribute header and
1575 if (err == -ENODATA)
1576 nla_nest_cancel(skb, af);
1580 nla_nest_end(skb, af);
1583 nla_nest_end(skb, af_spec);
1587 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1588 struct net_device *dev, struct net *src_net,
1589 int type, u32 pid, u32 seq, u32 change,
1590 unsigned int flags, u32 ext_filter_mask,
1591 u32 event, int *new_nsid, int new_ifindex,
1594 struct ifinfomsg *ifm;
1595 struct nlmsghdr *nlh;
1598 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1602 ifm = nlmsg_data(nlh);
1603 ifm->ifi_family = AF_UNSPEC;
1605 ifm->ifi_type = dev->type;
1606 ifm->ifi_index = dev->ifindex;
1607 ifm->ifi_flags = dev_get_flags(dev);
1608 ifm->ifi_change = change;
1610 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1611 goto nla_put_failure;
1613 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1614 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1615 nla_put_u8(skb, IFLA_OPERSTATE,
1616 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1617 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1618 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1619 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1620 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1621 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1622 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1623 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1624 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1625 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1627 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1629 put_master_ifindex(skb, dev) ||
1630 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1632 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1633 nla_put_ifalias(skb, dev) ||
1634 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1635 atomic_read(&dev->carrier_up_count) +
1636 atomic_read(&dev->carrier_down_count)) ||
1637 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1638 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1639 atomic_read(&dev->carrier_up_count)) ||
1640 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1641 atomic_read(&dev->carrier_down_count)))
1642 goto nla_put_failure;
1644 if (event != IFLA_EVENT_NONE) {
1645 if (nla_put_u32(skb, IFLA_EVENT, event))
1646 goto nla_put_failure;
1649 if (rtnl_fill_link_ifmap(skb, dev))
1650 goto nla_put_failure;
1652 if (dev->addr_len) {
1653 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1654 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1655 goto nla_put_failure;
1658 if (rtnl_phys_port_id_fill(skb, dev))
1659 goto nla_put_failure;
1661 if (rtnl_phys_port_name_fill(skb, dev))
1662 goto nla_put_failure;
1664 if (rtnl_phys_switch_id_fill(skb, dev))
1665 goto nla_put_failure;
1667 if (rtnl_fill_stats(skb, dev))
1668 goto nla_put_failure;
1670 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1671 goto nla_put_failure;
1673 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1674 goto nla_put_failure;
1676 if (rtnl_xdp_fill(skb, dev))
1677 goto nla_put_failure;
1679 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1680 if (rtnl_link_fill(skb, dev) < 0)
1681 goto nla_put_failure;
1684 if (rtnl_fill_link_netnsid(skb, dev, src_net))
1685 goto nla_put_failure;
1688 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1689 goto nla_put_failure;
1691 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1692 goto nla_put_failure;
1696 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1697 goto nla_put_failure_rcu;
1700 nlmsg_end(skb, nlh);
1703 nla_put_failure_rcu:
1706 nlmsg_cancel(skb, nlh);
1710 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1711 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1712 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1713 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1714 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1715 [IFLA_MTU] = { .type = NLA_U32 },
1716 [IFLA_LINK] = { .type = NLA_U32 },
1717 [IFLA_MASTER] = { .type = NLA_U32 },
1718 [IFLA_CARRIER] = { .type = NLA_U8 },
1719 [IFLA_TXQLEN] = { .type = NLA_U32 },
1720 [IFLA_WEIGHT] = { .type = NLA_U32 },
1721 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1722 [IFLA_LINKMODE] = { .type = NLA_U8 },
1723 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1724 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1725 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1726 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1727 * allow 0-length string (needed to remove an alias).
1729 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1730 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1731 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1732 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1733 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1734 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1735 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1736 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1737 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1738 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1739 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1740 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1741 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1742 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1743 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1744 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1745 [IFLA_XDP] = { .type = NLA_NESTED },
1746 [IFLA_EVENT] = { .type = NLA_U32 },
1747 [IFLA_GROUP] = { .type = NLA_U32 },
1748 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
1749 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1750 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1751 [IFLA_MIN_MTU] = { .type = NLA_U32 },
1752 [IFLA_MAX_MTU] = { .type = NLA_U32 },
1755 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1756 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1757 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1758 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1759 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1762 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1763 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1764 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1765 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1766 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1767 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1768 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1769 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1770 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1771 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1772 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1773 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1774 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1775 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1778 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1779 [IFLA_PORT_VF] = { .type = NLA_U32 },
1780 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1781 .len = PORT_PROFILE_MAX },
1782 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1783 .len = PORT_UUID_MAX },
1784 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1785 .len = PORT_UUID_MAX },
1786 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1787 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1789 /* Unused, but we need to keep it here since user space could
1790 * fill it. It's also broken with regard to NLA_BINARY use in
1791 * combination with structs.
1793 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1794 .len = sizeof(struct ifla_port_vsi) },
1797 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1798 [IFLA_XDP_FD] = { .type = NLA_S32 },
1799 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1800 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1801 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1804 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1806 const struct rtnl_link_ops *ops = NULL;
1807 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1809 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1812 if (linfo[IFLA_INFO_KIND]) {
1813 char kind[MODULE_NAME_LEN];
1815 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1816 ops = rtnl_link_ops_get(kind);
1822 static bool link_master_filtered(struct net_device *dev, int master_idx)
1824 struct net_device *master;
1829 master = netdev_master_upper_dev_get(dev);
1830 if (!master || master->ifindex != master_idx)
1836 static bool link_kind_filtered(const struct net_device *dev,
1837 const struct rtnl_link_ops *kind_ops)
1839 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1845 static bool link_dump_filtered(struct net_device *dev,
1847 const struct rtnl_link_ops *kind_ops)
1849 if (link_master_filtered(dev, master_idx) ||
1850 link_kind_filtered(dev, kind_ops))
1857 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1858 * @sk: netlink socket
1859 * @netnsid: network namespace identifier
1861 * Returns the network namespace identified by netnsid on success or an error
1862 * pointer on failure.
1864 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1868 net = get_net_ns_by_id(sock_net(sk), netnsid);
1870 return ERR_PTR(-EINVAL);
1872 /* For now, the caller is required to have CAP_NET_ADMIN in
1873 * the user namespace owning the target net ns.
1875 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1877 return ERR_PTR(-EACCES);
1881 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
1883 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
1884 bool strict_check, struct nlattr **tb,
1885 struct netlink_ext_ack *extack)
1890 struct ifinfomsg *ifm;
1892 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1893 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
1897 ifm = nlmsg_data(nlh);
1898 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
1900 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
1903 if (ifm->ifi_index) {
1904 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
1908 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
1909 IFLA_MAX, ifla_policy,
1913 /* A hack to preserve kernel<->userspace interface.
1914 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1915 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1916 * what iproute2 < v3.9.0 used.
1917 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1918 * attribute, its netlink message is shorter than struct ifinfomsg.
1920 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1921 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1923 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
1927 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1929 struct netlink_ext_ack *extack = cb->extack;
1930 const struct nlmsghdr *nlh = cb->nlh;
1931 struct net *net = sock_net(skb->sk);
1932 struct net *tgt_net = net;
1935 struct net_device *dev;
1936 struct hlist_head *head;
1937 struct nlattr *tb[IFLA_MAX+1];
1938 u32 ext_filter_mask = 0;
1939 const struct rtnl_link_ops *kind_ops = NULL;
1940 unsigned int flags = NLM_F_MULTI;
1946 s_idx = cb->args[1];
1948 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
1950 if (cb->strict_check)
1956 for (i = 0; i <= IFLA_MAX; ++i) {
1960 /* new attributes should only be added with strict checking */
1962 case IFLA_TARGET_NETNSID:
1963 netnsid = nla_get_s32(tb[i]);
1964 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
1965 if (IS_ERR(tgt_net)) {
1966 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
1967 return PTR_ERR(tgt_net);
1971 ext_filter_mask = nla_get_u32(tb[i]);
1974 master_idx = nla_get_u32(tb[i]);
1977 kind_ops = linkinfo_to_kind_ops(tb[i]);
1980 if (cb->strict_check) {
1981 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
1987 if (master_idx || kind_ops)
1988 flags |= NLM_F_DUMP_FILTERED;
1991 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1993 head = &tgt_net->dev_index_head[h];
1994 hlist_for_each_entry(dev, head, index_hlist) {
1995 if (link_dump_filtered(dev, master_idx, kind_ops))
1999 err = rtnl_fill_ifinfo(skb, dev, net,
2001 NETLINK_CB(cb->skb).portid,
2002 nlh->nlmsg_seq, 0, flags,
2003 ext_filter_mask, 0, NULL, 0,
2007 if (likely(skb->len))
2021 cb->seq = net->dev_base_seq;
2022 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2029 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2030 struct netlink_ext_ack *exterr)
2032 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2035 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2037 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2040 /* Examine the link attributes and figure out which
2041 * network namespace we are talking about.
2043 if (tb[IFLA_NET_NS_PID])
2044 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2045 else if (tb[IFLA_NET_NS_FD])
2046 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2048 net = get_net(src_net);
2051 EXPORT_SYMBOL(rtnl_link_get_net);
2053 /* Figure out which network namespace we are talking about by
2054 * examining the link attributes in the following order:
2056 * 1. IFLA_NET_NS_PID
2058 * 3. IFLA_TARGET_NETNSID
2060 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2061 struct nlattr *tb[])
2065 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2066 return rtnl_link_get_net(src_net, tb);
2068 if (!tb[IFLA_TARGET_NETNSID])
2069 return get_net(src_net);
2071 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2073 return ERR_PTR(-EINVAL);
2078 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2079 struct net *src_net,
2080 struct nlattr *tb[], int cap)
2084 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2088 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2090 return ERR_PTR(-EPERM);
2096 /* Verify that rtnetlink requests do not pass additional properties
2097 * potentially referring to different network namespaces.
2099 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2100 struct netlink_ext_ack *extack,
2104 if (netns_id_only) {
2105 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2108 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2112 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2115 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2118 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2124 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2128 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2131 if (tb[IFLA_ADDRESS] &&
2132 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2135 if (tb[IFLA_BROADCAST] &&
2136 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2140 if (tb[IFLA_AF_SPEC]) {
2144 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2145 const struct rtnl_af_ops *af_ops;
2148 af_ops = rtnl_af_lookup(nla_type(af));
2151 return -EAFNOSUPPORT;
2154 if (!af_ops->set_link_af) {
2159 if (af_ops->validate_link_af) {
2160 err = af_ops->validate_link_af(dev, af);
2174 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2177 const struct net_device_ops *ops = dev->netdev_ops;
2179 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2182 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2184 if (dev->type != ARPHRD_INFINIBAND)
2187 return handle_infiniband_guid(dev, ivt, guid_type);
2190 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2192 const struct net_device_ops *ops = dev->netdev_ops;
2195 if (tb[IFLA_VF_MAC]) {
2196 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2199 if (ops->ndo_set_vf_mac)
2200 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2206 if (tb[IFLA_VF_VLAN]) {
2207 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2210 if (ops->ndo_set_vf_vlan)
2211 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2213 htons(ETH_P_8021Q));
2218 if (tb[IFLA_VF_VLAN_LIST]) {
2219 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2220 struct nlattr *attr;
2224 if (!ops->ndo_set_vf_vlan)
2227 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2228 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2229 nla_len(attr) < NLA_HDRLEN) {
2232 if (len >= MAX_VLAN_LIST_LEN)
2234 ivvl[len] = nla_data(attr);
2241 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2242 ivvl[0]->qos, ivvl[0]->vlan_proto);
2247 if (tb[IFLA_VF_TX_RATE]) {
2248 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2249 struct ifla_vf_info ivf;
2252 if (ops->ndo_get_vf_config)
2253 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2258 if (ops->ndo_set_vf_rate)
2259 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2266 if (tb[IFLA_VF_RATE]) {
2267 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2270 if (ops->ndo_set_vf_rate)
2271 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2278 if (tb[IFLA_VF_SPOOFCHK]) {
2279 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2282 if (ops->ndo_set_vf_spoofchk)
2283 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2289 if (tb[IFLA_VF_LINK_STATE]) {
2290 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2293 if (ops->ndo_set_vf_link_state)
2294 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2300 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2301 struct ifla_vf_rss_query_en *ivrssq_en;
2304 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2305 if (ops->ndo_set_vf_rss_query_en)
2306 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2307 ivrssq_en->setting);
2312 if (tb[IFLA_VF_TRUST]) {
2313 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2316 if (ops->ndo_set_vf_trust)
2317 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2322 if (tb[IFLA_VF_IB_NODE_GUID]) {
2323 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2325 if (!ops->ndo_set_vf_guid)
2328 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2331 if (tb[IFLA_VF_IB_PORT_GUID]) {
2332 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2334 if (!ops->ndo_set_vf_guid)
2337 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2343 static int do_set_master(struct net_device *dev, int ifindex,
2344 struct netlink_ext_ack *extack)
2346 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2347 const struct net_device_ops *ops;
2351 if (upper_dev->ifindex == ifindex)
2353 ops = upper_dev->netdev_ops;
2354 if (ops->ndo_del_slave) {
2355 err = ops->ndo_del_slave(upper_dev, dev);
2364 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2367 ops = upper_dev->netdev_ops;
2368 if (ops->ndo_add_slave) {
2369 err = ops->ndo_add_slave(upper_dev, dev, extack);
2379 #define DO_SETLINK_MODIFIED 0x01
2380 /* notify flag means notify + modified. */
2381 #define DO_SETLINK_NOTIFY 0x03
2382 static int do_setlink(const struct sk_buff *skb,
2383 struct net_device *dev, struct ifinfomsg *ifm,
2384 struct netlink_ext_ack *extack,
2385 struct nlattr **tb, char *ifname, int status)
2387 const struct net_device_ops *ops = dev->netdev_ops;
2390 err = validate_linkmsg(dev, tb);
2394 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2395 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2402 err = dev_change_net_namespace(dev, net, ifname);
2406 status |= DO_SETLINK_MODIFIED;
2410 struct rtnl_link_ifmap *u_map;
2413 if (!ops->ndo_set_config) {
2418 if (!netif_device_present(dev)) {
2423 u_map = nla_data(tb[IFLA_MAP]);
2424 k_map.mem_start = (unsigned long) u_map->mem_start;
2425 k_map.mem_end = (unsigned long) u_map->mem_end;
2426 k_map.base_addr = (unsigned short) u_map->base_addr;
2427 k_map.irq = (unsigned char) u_map->irq;
2428 k_map.dma = (unsigned char) u_map->dma;
2429 k_map.port = (unsigned char) u_map->port;
2431 err = ops->ndo_set_config(dev, &k_map);
2435 status |= DO_SETLINK_NOTIFY;
2438 if (tb[IFLA_ADDRESS]) {
2439 struct sockaddr *sa;
2442 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2444 sa = kmalloc(len, GFP_KERNEL);
2449 sa->sa_family = dev->type;
2450 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2452 err = dev_set_mac_address(dev, sa, extack);
2456 status |= DO_SETLINK_MODIFIED;
2460 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2463 status |= DO_SETLINK_MODIFIED;
2466 if (tb[IFLA_GROUP]) {
2467 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2468 status |= DO_SETLINK_NOTIFY;
2472 * Interface selected by interface index but interface
2473 * name provided implies that a name change has been
2476 if (ifm->ifi_index > 0 && ifname[0]) {
2477 err = dev_change_name(dev, ifname);
2480 status |= DO_SETLINK_MODIFIED;
2483 if (tb[IFLA_IFALIAS]) {
2484 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2485 nla_len(tb[IFLA_IFALIAS]));
2488 status |= DO_SETLINK_NOTIFY;
2491 if (tb[IFLA_BROADCAST]) {
2492 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2493 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2496 if (ifm->ifi_flags || ifm->ifi_change) {
2497 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2503 if (tb[IFLA_MASTER]) {
2504 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2507 status |= DO_SETLINK_MODIFIED;
2510 if (tb[IFLA_CARRIER]) {
2511 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2514 status |= DO_SETLINK_MODIFIED;
2517 if (tb[IFLA_TXQLEN]) {
2518 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2520 err = dev_change_tx_queue_len(dev, value);
2523 status |= DO_SETLINK_MODIFIED;
2526 if (tb[IFLA_GSO_MAX_SIZE]) {
2527 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2529 if (max_size > GSO_MAX_SIZE) {
2534 if (dev->gso_max_size ^ max_size) {
2535 netif_set_gso_max_size(dev, max_size);
2536 status |= DO_SETLINK_MODIFIED;
2540 if (tb[IFLA_GSO_MAX_SEGS]) {
2541 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2543 if (max_segs > GSO_MAX_SEGS) {
2548 if (dev->gso_max_segs ^ max_segs) {
2549 dev->gso_max_segs = max_segs;
2550 status |= DO_SETLINK_MODIFIED;
2554 if (tb[IFLA_OPERSTATE])
2555 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2557 if (tb[IFLA_LINKMODE]) {
2558 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2560 write_lock_bh(&dev_base_lock);
2561 if (dev->link_mode ^ value)
2562 status |= DO_SETLINK_NOTIFY;
2563 dev->link_mode = value;
2564 write_unlock_bh(&dev_base_lock);
2567 if (tb[IFLA_VFINFO_LIST]) {
2568 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2569 struct nlattr *attr;
2572 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2573 if (nla_type(attr) != IFLA_VF_INFO ||
2574 nla_len(attr) < NLA_HDRLEN) {
2578 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2584 err = do_setvfinfo(dev, vfinfo);
2587 status |= DO_SETLINK_NOTIFY;
2592 if (tb[IFLA_VF_PORTS]) {
2593 struct nlattr *port[IFLA_PORT_MAX+1];
2594 struct nlattr *attr;
2599 if (!ops->ndo_set_vf_port)
2602 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2603 if (nla_type(attr) != IFLA_VF_PORT ||
2604 nla_len(attr) < NLA_HDRLEN) {
2608 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2614 if (!port[IFLA_PORT_VF]) {
2618 vf = nla_get_u32(port[IFLA_PORT_VF]);
2619 err = ops->ndo_set_vf_port(dev, vf, port);
2622 status |= DO_SETLINK_NOTIFY;
2627 if (tb[IFLA_PORT_SELF]) {
2628 struct nlattr *port[IFLA_PORT_MAX+1];
2630 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2632 ifla_port_policy, NULL);
2637 if (ops->ndo_set_vf_port)
2638 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2641 status |= DO_SETLINK_NOTIFY;
2644 if (tb[IFLA_AF_SPEC]) {
2648 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2649 const struct rtnl_af_ops *af_ops;
2653 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2655 err = af_ops->set_link_af(dev, af);
2662 status |= DO_SETLINK_NOTIFY;
2667 if (tb[IFLA_PROTO_DOWN]) {
2668 err = dev_change_proto_down(dev,
2669 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2672 status |= DO_SETLINK_NOTIFY;
2676 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2679 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2681 ifla_xdp_policy, NULL);
2685 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2690 if (xdp[IFLA_XDP_FLAGS]) {
2691 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2692 if (xdp_flags & ~XDP_FLAGS_MASK) {
2696 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2702 if (xdp[IFLA_XDP_FD]) {
2703 err = dev_change_xdp_fd(dev, extack,
2704 nla_get_s32(xdp[IFLA_XDP_FD]),
2708 status |= DO_SETLINK_NOTIFY;
2713 if (status & DO_SETLINK_MODIFIED) {
2714 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2715 netdev_state_change(dev);
2718 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",
2725 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2726 struct netlink_ext_ack *extack)
2728 struct net *net = sock_net(skb->sk);
2729 struct ifinfomsg *ifm;
2730 struct net_device *dev;
2732 struct nlattr *tb[IFLA_MAX+1];
2733 char ifname[IFNAMSIZ];
2735 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2736 ifla_policy, extack);
2740 err = rtnl_ensure_unique_netns(tb, extack, false);
2744 if (tb[IFLA_IFNAME])
2745 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2750 ifm = nlmsg_data(nlh);
2751 if (ifm->ifi_index > 0)
2752 dev = __dev_get_by_index(net, ifm->ifi_index);
2753 else if (tb[IFLA_IFNAME])
2754 dev = __dev_get_by_name(net, ifname);
2763 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2768 static int rtnl_group_dellink(const struct net *net, int group)
2770 struct net_device *dev, *aux;
2771 LIST_HEAD(list_kill);
2777 for_each_netdev(net, dev) {
2778 if (dev->group == group) {
2779 const struct rtnl_link_ops *ops;
2782 ops = dev->rtnl_link_ops;
2783 if (!ops || !ops->dellink)
2791 for_each_netdev_safe(net, dev, aux) {
2792 if (dev->group == group) {
2793 const struct rtnl_link_ops *ops;
2795 ops = dev->rtnl_link_ops;
2796 ops->dellink(dev, &list_kill);
2799 unregister_netdevice_many(&list_kill);
2804 int rtnl_delete_link(struct net_device *dev)
2806 const struct rtnl_link_ops *ops;
2807 LIST_HEAD(list_kill);
2809 ops = dev->rtnl_link_ops;
2810 if (!ops || !ops->dellink)
2813 ops->dellink(dev, &list_kill);
2814 unregister_netdevice_many(&list_kill);
2818 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2820 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2821 struct netlink_ext_ack *extack)
2823 struct net *net = sock_net(skb->sk);
2824 struct net *tgt_net = net;
2825 struct net_device *dev = NULL;
2826 struct ifinfomsg *ifm;
2827 char ifname[IFNAMSIZ];
2828 struct nlattr *tb[IFLA_MAX+1];
2832 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2833 ifla_policy, extack);
2837 err = rtnl_ensure_unique_netns(tb, extack, true);
2841 if (tb[IFLA_IFNAME])
2842 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2844 if (tb[IFLA_TARGET_NETNSID]) {
2845 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
2846 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
2847 if (IS_ERR(tgt_net))
2848 return PTR_ERR(tgt_net);
2852 ifm = nlmsg_data(nlh);
2853 if (ifm->ifi_index > 0)
2854 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2855 else if (tb[IFLA_IFNAME])
2856 dev = __dev_get_by_name(tgt_net, ifname);
2857 else if (tb[IFLA_GROUP])
2858 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2863 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2869 err = rtnl_delete_link(dev);
2878 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2880 unsigned int old_flags;
2883 old_flags = dev->flags;
2884 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2885 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2891 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2892 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2894 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2895 __dev_notify_flags(dev, old_flags, ~0U);
2899 EXPORT_SYMBOL(rtnl_configure_link);
2901 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
2902 unsigned char name_assign_type,
2903 const struct rtnl_link_ops *ops,
2904 struct nlattr *tb[],
2905 struct netlink_ext_ack *extack)
2907 struct net_device *dev;
2908 unsigned int num_tx_queues = 1;
2909 unsigned int num_rx_queues = 1;
2911 if (tb[IFLA_NUM_TX_QUEUES])
2912 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2913 else if (ops->get_num_tx_queues)
2914 num_tx_queues = ops->get_num_tx_queues();
2916 if (tb[IFLA_NUM_RX_QUEUES])
2917 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2918 else if (ops->get_num_rx_queues)
2919 num_rx_queues = ops->get_num_rx_queues();
2921 if (num_tx_queues < 1 || num_tx_queues > 4096) {
2922 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
2923 return ERR_PTR(-EINVAL);
2926 if (num_rx_queues < 1 || num_rx_queues > 4096) {
2927 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
2928 return ERR_PTR(-EINVAL);
2931 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2932 ops->setup, num_tx_queues, num_rx_queues);
2934 return ERR_PTR(-ENOMEM);
2936 dev_net_set(dev, net);
2937 dev->rtnl_link_ops = ops;
2938 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2941 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2942 if (tb[IFLA_ADDRESS]) {
2943 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2944 nla_len(tb[IFLA_ADDRESS]));
2945 dev->addr_assign_type = NET_ADDR_SET;
2947 if (tb[IFLA_BROADCAST])
2948 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2949 nla_len(tb[IFLA_BROADCAST]));
2950 if (tb[IFLA_TXQLEN])
2951 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2952 if (tb[IFLA_OPERSTATE])
2953 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2954 if (tb[IFLA_LINKMODE])
2955 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2957 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2958 if (tb[IFLA_GSO_MAX_SIZE])
2959 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2960 if (tb[IFLA_GSO_MAX_SEGS])
2961 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2965 EXPORT_SYMBOL(rtnl_create_link);
2967 static int rtnl_group_changelink(const struct sk_buff *skb,
2968 struct net *net, int group,
2969 struct ifinfomsg *ifm,
2970 struct netlink_ext_ack *extack,
2973 struct net_device *dev, *aux;
2976 for_each_netdev_safe(net, dev, aux) {
2977 if (dev->group == group) {
2978 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2987 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2988 struct nlattr **attr, struct netlink_ext_ack *extack)
2990 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
2991 unsigned char name_assign_type = NET_NAME_USER;
2992 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
2993 const struct rtnl_link_ops *m_ops = NULL;
2994 struct net_device *master_dev = NULL;
2995 struct net *net = sock_net(skb->sk);
2996 const struct rtnl_link_ops *ops;
2997 struct nlattr *tb[IFLA_MAX + 1];
2998 struct net *dest_net, *link_net;
2999 struct nlattr **slave_data;
3000 char kind[MODULE_NAME_LEN];
3001 struct net_device *dev;
3002 struct ifinfomsg *ifm;
3003 char ifname[IFNAMSIZ];
3004 struct nlattr **data;
3007 #ifdef CONFIG_MODULES
3010 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3011 ifla_policy, extack);
3015 err = rtnl_ensure_unique_netns(tb, extack, false);
3019 if (tb[IFLA_IFNAME])
3020 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3024 ifm = nlmsg_data(nlh);
3025 if (ifm->ifi_index > 0)
3026 dev = __dev_get_by_index(net, ifm->ifi_index);
3029 dev = __dev_get_by_name(net, ifname);
3035 master_dev = netdev_master_upper_dev_get(dev);
3037 m_ops = master_dev->rtnl_link_ops;
3040 err = validate_linkmsg(dev, tb);
3044 if (tb[IFLA_LINKINFO]) {
3045 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3047 ifla_info_policy, NULL);
3051 memset(linkinfo, 0, sizeof(linkinfo));
3053 if (linkinfo[IFLA_INFO_KIND]) {
3054 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3055 ops = rtnl_link_ops_get(kind);
3063 if (ops->maxtype > RTNL_MAX_TYPE)
3066 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3067 err = nla_parse_nested_deprecated(attr, ops->maxtype,
3068 linkinfo[IFLA_INFO_DATA],
3069 ops->policy, extack);
3074 if (ops->validate) {
3075 err = ops->validate(tb, data, extack);
3083 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3086 if (m_ops->slave_maxtype &&
3087 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3088 err = nla_parse_nested_deprecated(slave_attr,
3089 m_ops->slave_maxtype,
3090 linkinfo[IFLA_INFO_SLAVE_DATA],
3091 m_ops->slave_policy,
3095 slave_data = slave_attr;
3102 if (nlh->nlmsg_flags & NLM_F_EXCL)
3104 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3107 if (linkinfo[IFLA_INFO_DATA]) {
3108 if (!ops || ops != dev->rtnl_link_ops ||
3112 err = ops->changelink(dev, tb, data, extack);
3115 status |= DO_SETLINK_NOTIFY;
3118 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3119 if (!m_ops || !m_ops->slave_changelink)
3122 err = m_ops->slave_changelink(master_dev, dev, tb,
3123 slave_data, extack);
3126 status |= DO_SETLINK_NOTIFY;
3129 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3132 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3133 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3134 return rtnl_group_changelink(skb, net,
3135 nla_get_u32(tb[IFLA_GROUP]),
3140 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3144 #ifdef CONFIG_MODULES
3147 request_module("rtnl-link-%s", kind);
3149 ops = rtnl_link_ops_get(kind);
3154 NL_SET_ERR_MSG(extack, "Unknown device type");
3162 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3163 name_assign_type = NET_NAME_ENUM;
3166 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3167 if (IS_ERR(dest_net))
3168 return PTR_ERR(dest_net);
3170 if (tb[IFLA_LINK_NETNSID]) {
3171 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3173 link_net = get_net_ns_by_id(dest_net, id);
3175 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3180 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3186 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3187 name_assign_type, ops, tb, extack);
3193 dev->ifindex = ifm->ifi_index;
3196 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3197 /* Drivers should call free_netdev() in ->destructor
3198 * and unregister it on failure after registration
3199 * so that device could be finally freed in rtnl_unlock.
3202 /* If device is not registered at all, free it now */
3203 if (dev->reg_state == NETREG_UNINITIALIZED)
3208 err = register_netdevice(dev);
3214 err = rtnl_configure_link(dev, ifm);
3216 goto out_unregister;
3218 err = dev_change_net_namespace(dev, dest_net, ifname);
3220 goto out_unregister;
3222 if (tb[IFLA_MASTER]) {
3223 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3225 goto out_unregister;
3234 LIST_HEAD(list_kill);
3236 ops->dellink(dev, &list_kill);
3237 unregister_netdevice_many(&list_kill);
3239 unregister_netdevice(dev);
3244 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3245 struct netlink_ext_ack *extack)
3247 struct nlattr **attr;
3250 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3254 ret = __rtnl_newlink(skb, nlh, attr, extack);
3259 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3260 const struct nlmsghdr *nlh,
3262 struct netlink_ext_ack *extack)
3264 struct ifinfomsg *ifm;
3267 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3268 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3272 if (!netlink_strict_get_check(skb))
3273 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3274 ifla_policy, extack);
3276 ifm = nlmsg_data(nlh);
3277 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3279 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3283 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3284 ifla_policy, extack);
3288 for (i = 0; i <= IFLA_MAX; i++) {
3295 case IFLA_TARGET_NETNSID:
3298 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3306 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3307 struct netlink_ext_ack *extack)
3309 struct net *net = sock_net(skb->sk);
3310 struct net *tgt_net = net;
3311 struct ifinfomsg *ifm;
3312 char ifname[IFNAMSIZ];
3313 struct nlattr *tb[IFLA_MAX+1];
3314 struct net_device *dev = NULL;
3315 struct sk_buff *nskb;
3318 u32 ext_filter_mask = 0;
3320 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3324 err = rtnl_ensure_unique_netns(tb, extack, true);
3328 if (tb[IFLA_TARGET_NETNSID]) {
3329 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3330 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3331 if (IS_ERR(tgt_net))
3332 return PTR_ERR(tgt_net);
3335 if (tb[IFLA_IFNAME])
3336 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3338 if (tb[IFLA_EXT_MASK])
3339 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3342 ifm = nlmsg_data(nlh);
3343 if (ifm->ifi_index > 0)
3344 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3345 else if (tb[IFLA_IFNAME])
3346 dev = __dev_get_by_name(tgt_net, ifname);
3355 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3359 err = rtnl_fill_ifinfo(nskb, dev, net,
3360 RTM_NEWLINK, NETLINK_CB(skb).portid,
3361 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3362 0, NULL, 0, netnsid);
3364 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3365 WARN_ON(err == -EMSGSIZE);
3368 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3376 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3378 struct net *net = sock_net(skb->sk);
3379 struct net_device *dev;
3380 struct nlattr *tb[IFLA_MAX+1];
3381 u32 ext_filter_mask = 0;
3382 u16 min_ifinfo_dump_size = 0;
3385 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3386 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3387 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3389 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3390 if (tb[IFLA_EXT_MASK])
3391 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3394 if (!ext_filter_mask)
3395 return NLMSG_GOODSIZE;
3397 * traverse the list of net devices and compute the minimum
3398 * buffer size based upon the filter mask.
3401 for_each_netdev_rcu(net, dev) {
3402 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3408 return nlmsg_total_size(min_ifinfo_dump_size);
3411 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3414 int s_idx = cb->family;
3415 int type = cb->nlh->nlmsg_type - RTM_BASE;
3421 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3422 struct rtnl_link **tab;
3423 struct rtnl_link *link;
3424 rtnl_dumpit_func dumpit;
3426 if (idx < s_idx || idx == PF_PACKET)
3429 if (type < 0 || type >= RTM_NR_MSGTYPES)
3432 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3440 dumpit = link->dumpit;
3445 memset(&cb->args[0], 0, sizeof(cb->args));
3449 ret = dumpit(skb, cb);
3455 return skb->len ? : ret;
3458 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3459 unsigned int change,
3460 u32 event, gfp_t flags, int *new_nsid,
3463 struct net *net = dev_net(dev);
3464 struct sk_buff *skb;
3466 size_t if_info_size;
3468 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3472 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3473 type, 0, 0, change, 0, 0, event,
3474 new_nsid, new_ifindex, -1);
3476 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3477 WARN_ON(err == -EMSGSIZE);
3484 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3488 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3490 struct net *net = dev_net(dev);
3492 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3495 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3496 unsigned int change, u32 event,
3497 gfp_t flags, int *new_nsid, int new_ifindex)
3499 struct sk_buff *skb;
3501 if (dev->reg_state != NETREG_REGISTERED)
3504 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3507 rtmsg_ifinfo_send(skb, dev, flags);
3510 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3513 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3517 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3518 gfp_t flags, int *new_nsid, int new_ifindex)
3520 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3521 new_nsid, new_ifindex);
3524 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3525 struct net_device *dev,
3526 u8 *addr, u16 vid, u32 pid, u32 seq,
3527 int type, unsigned int flags,
3528 int nlflags, u16 ndm_state)
3530 struct nlmsghdr *nlh;
3533 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3537 ndm = nlmsg_data(nlh);
3538 ndm->ndm_family = AF_BRIDGE;
3541 ndm->ndm_flags = flags;
3543 ndm->ndm_ifindex = dev->ifindex;
3544 ndm->ndm_state = ndm_state;
3546 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3547 goto nla_put_failure;
3549 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3550 goto nla_put_failure;
3552 nlmsg_end(skb, nlh);
3556 nlmsg_cancel(skb, nlh);
3560 static inline size_t rtnl_fdb_nlmsg_size(void)
3562 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3563 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3564 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3568 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3571 struct net *net = dev_net(dev);
3572 struct sk_buff *skb;
3575 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3579 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3580 0, 0, type, NTF_SELF, 0, ndm_state);
3586 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3589 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3593 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3595 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3596 struct nlattr *tb[],
3597 struct net_device *dev,
3598 const unsigned char *addr, u16 vid,
3603 /* If aging addresses are supported device will need to
3604 * implement its own handler for this.
3606 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3607 pr_info("%s: FDB only supports static addresses\n", dev->name);
3612 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3616 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3617 err = dev_uc_add_excl(dev, addr);
3618 else if (is_multicast_ether_addr(addr))
3619 err = dev_mc_add_excl(dev, addr);
3621 /* Only return duplicate errors if NLM_F_EXCL is set */
3622 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3627 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3629 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3630 struct netlink_ext_ack *extack)
3635 if (nla_len(vlan_attr) != sizeof(u16)) {
3636 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3640 vid = nla_get_u16(vlan_attr);
3642 if (!vid || vid >= VLAN_VID_MASK) {
3643 NL_SET_ERR_MSG(extack, "invalid vlan id");
3651 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3652 struct netlink_ext_ack *extack)
3654 struct net *net = sock_net(skb->sk);
3656 struct nlattr *tb[NDA_MAX+1];
3657 struct net_device *dev;
3662 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3667 ndm = nlmsg_data(nlh);
3668 if (ndm->ndm_ifindex == 0) {
3669 NL_SET_ERR_MSG(extack, "invalid ifindex");
3673 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3675 NL_SET_ERR_MSG(extack, "unknown ifindex");
3679 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3680 NL_SET_ERR_MSG(extack, "invalid address");
3684 if (dev->type != ARPHRD_ETHER) {
3685 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3689 addr = nla_data(tb[NDA_LLADDR]);
3691 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3697 /* Support fdb on master device the net/bridge default case */
3698 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3699 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3700 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3701 const struct net_device_ops *ops = br_dev->netdev_ops;
3703 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3704 nlh->nlmsg_flags, extack);
3708 ndm->ndm_flags &= ~NTF_MASTER;
3711 /* Embedded bridge, macvlan, and any other device support */
3712 if ((ndm->ndm_flags & NTF_SELF)) {
3713 if (dev->netdev_ops->ndo_fdb_add)
3714 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3719 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3723 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3725 ndm->ndm_flags &= ~NTF_SELF;
3733 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3735 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3736 struct nlattr *tb[],
3737 struct net_device *dev,
3738 const unsigned char *addr, u16 vid)
3742 /* If aging addresses are supported device will need to
3743 * implement its own handler for this.
3745 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3746 pr_info("%s: FDB only supports static addresses\n", dev->name);
3750 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3751 err = dev_uc_del(dev, addr);
3752 else if (is_multicast_ether_addr(addr))
3753 err = dev_mc_del(dev, addr);
3757 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3759 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3760 struct netlink_ext_ack *extack)
3762 struct net *net = sock_net(skb->sk);
3764 struct nlattr *tb[NDA_MAX+1];
3765 struct net_device *dev;
3770 if (!netlink_capable(skb, CAP_NET_ADMIN))
3773 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3778 ndm = nlmsg_data(nlh);
3779 if (ndm->ndm_ifindex == 0) {
3780 NL_SET_ERR_MSG(extack, "invalid ifindex");
3784 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3786 NL_SET_ERR_MSG(extack, "unknown ifindex");
3790 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3791 NL_SET_ERR_MSG(extack, "invalid address");
3795 if (dev->type != ARPHRD_ETHER) {
3796 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
3800 addr = nla_data(tb[NDA_LLADDR]);
3802 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3808 /* Support fdb on master device the net/bridge default case */
3809 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3810 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3811 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3812 const struct net_device_ops *ops = br_dev->netdev_ops;
3814 if (ops->ndo_fdb_del)
3815 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3820 ndm->ndm_flags &= ~NTF_MASTER;
3823 /* Embedded bridge, macvlan, and any other device support */
3824 if (ndm->ndm_flags & NTF_SELF) {
3825 if (dev->netdev_ops->ndo_fdb_del)
3826 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3829 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3832 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3834 ndm->ndm_flags &= ~NTF_SELF;
3841 static int nlmsg_populate_fdb(struct sk_buff *skb,
3842 struct netlink_callback *cb,
3843 struct net_device *dev,
3845 struct netdev_hw_addr_list *list)
3847 struct netdev_hw_addr *ha;
3851 portid = NETLINK_CB(cb->skb).portid;
3852 seq = cb->nlh->nlmsg_seq;
3854 list_for_each_entry(ha, &list->list, list) {
3855 if (*idx < cb->args[2])
3858 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3860 RTM_NEWNEIGH, NTF_SELF,
3861 NLM_F_MULTI, NUD_PERMANENT);
3871 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3872 * @skb: socket buffer to store message in
3873 * @cb: netlink callback
3875 * @filter_dev: ignored
3876 * @idx: the number of FDB table entries dumped is added to *@idx
3878 * Default netdevice operation to dump the existing unicast address list.
3879 * Returns number of addresses from list put in skb.
3881 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3882 struct netlink_callback *cb,
3883 struct net_device *dev,
3884 struct net_device *filter_dev,
3889 if (dev->type != ARPHRD_ETHER)
3892 netif_addr_lock_bh(dev);
3893 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3896 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3898 netif_addr_unlock_bh(dev);
3901 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3903 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
3904 int *br_idx, int *brport_idx,
3905 struct netlink_ext_ack *extack)
3907 struct nlattr *tb[NDA_MAX + 1];
3911 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
3912 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
3916 ndm = nlmsg_data(nlh);
3917 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
3918 ndm->ndm_flags || ndm->ndm_type) {
3919 NL_SET_ERR_MSG(extack, "Invalid values in header for fbd dump request");
3923 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
3924 NDA_MAX, NULL, extack);
3928 *brport_idx = ndm->ndm_ifindex;
3929 for (i = 0; i <= NDA_MAX; ++i) {
3935 if (nla_len(tb[i]) != sizeof(u32)) {
3936 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
3939 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
3942 if (nla_len(tb[i]) != sizeof(u32)) {
3943 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
3946 *br_idx = nla_get_u32(tb[NDA_MASTER]);
3949 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
3957 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
3958 int *br_idx, int *brport_idx,
3959 struct netlink_ext_ack *extack)
3961 struct nlattr *tb[IFLA_MAX+1];
3964 /* A hack to preserve kernel<->userspace interface.
3965 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
3966 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
3967 * So, check for ndmsg with an optional u32 attribute (not used here).
3968 * Fortunately these sizes don't conflict with the size of ifinfomsg
3969 * with an optional attribute.
3971 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
3972 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
3973 nla_attr_size(sizeof(u32)))) {
3974 struct ifinfomsg *ifm;
3976 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
3977 tb, IFLA_MAX, ifla_policy,
3981 } else if (err == 0) {
3982 if (tb[IFLA_MASTER])
3983 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
3986 ifm = nlmsg_data(nlh);
3987 *brport_idx = ifm->ifi_index;
3992 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3994 struct net_device *dev;
3995 struct net_device *br_dev = NULL;
3996 const struct net_device_ops *ops = NULL;
3997 const struct net_device_ops *cops = NULL;
3998 struct net *net = sock_net(skb->sk);
3999 struct hlist_head *head;
4007 if (cb->strict_check)
4008 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4011 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4017 br_dev = __dev_get_by_index(net, br_idx);
4021 ops = br_dev->netdev_ops;
4025 s_idx = cb->args[1];
4027 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4029 head = &net->dev_index_head[h];
4030 hlist_for_each_entry(dev, head, index_hlist) {
4032 if (brport_idx && (dev->ifindex != brport_idx))
4035 if (!br_idx) { /* user did not specify a specific bridge */
4036 if (dev->priv_flags & IFF_BRIDGE_PORT) {
4037 br_dev = netdev_master_upper_dev_get(dev);
4038 cops = br_dev->netdev_ops;
4041 if (dev != br_dev &&
4042 !(dev->priv_flags & IFF_BRIDGE_PORT))
4045 if (br_dev != netdev_master_upper_dev_get(dev) &&
4046 !(dev->priv_flags & IFF_EBRIDGE))
4054 if (dev->priv_flags & IFF_BRIDGE_PORT) {
4055 if (cops && cops->ndo_fdb_dump) {
4056 err = cops->ndo_fdb_dump(skb, cb,
4059 if (err == -EMSGSIZE)
4064 if (dev->netdev_ops->ndo_fdb_dump)
4065 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4069 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4071 if (err == -EMSGSIZE)
4076 /* reset fdb offset to 0 for rest of the interfaces */
4092 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4093 struct nlattr **tb, u8 *ndm_flags,
4094 int *br_idx, int *brport_idx, u8 **addr,
4095 u16 *vid, struct netlink_ext_ack *extack)
4100 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4101 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4105 ndm = nlmsg_data(nlh);
4106 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4108 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4112 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4113 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4117 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4118 NDA_MAX, nda_policy, extack);
4122 *ndm_flags = ndm->ndm_flags;
4123 *brport_idx = ndm->ndm_ifindex;
4124 for (i = 0; i <= NDA_MAX; ++i) {
4130 *br_idx = nla_get_u32(tb[i]);
4133 if (nla_len(tb[i]) != ETH_ALEN) {
4134 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4137 *addr = nla_data(tb[i]);
4140 err = fdb_vid_parse(tb[i], vid, extack);
4147 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4155 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4156 struct netlink_ext_ack *extack)
4158 struct net_device *dev = NULL, *br_dev = NULL;
4159 const struct net_device_ops *ops = NULL;
4160 struct net *net = sock_net(in_skb->sk);
4161 struct nlattr *tb[NDA_MAX + 1];
4162 struct sk_buff *skb;
4170 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4171 &brport_idx, &addr, &vid, extack);
4176 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4181 dev = __dev_get_by_index(net, brport_idx);
4183 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4190 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4194 br_dev = __dev_get_by_index(net, br_idx);
4196 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4199 ops = br_dev->netdev_ops;
4203 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4204 if (!(dev->priv_flags & IFF_BRIDGE_PORT)) {
4205 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4208 br_dev = netdev_master_upper_dev_get(dev);
4210 NL_SET_ERR_MSG(extack, "Master of device not found");
4213 ops = br_dev->netdev_ops;
4215 if (!(ndm_flags & NTF_SELF)) {
4216 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4219 ops = dev->netdev_ops;
4223 if (!br_dev && !dev) {
4224 NL_SET_ERR_MSG(extack, "No device specified");
4228 if (!ops || !ops->ndo_fdb_get) {
4229 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4233 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4239 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4240 NETLINK_CB(in_skb).portid,
4241 nlh->nlmsg_seq, extack);
4245 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4251 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4252 unsigned int attrnum, unsigned int flag)
4255 return nla_put_u8(skb, attrnum, !!(flags & flag));
4259 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4260 struct net_device *dev, u16 mode,
4261 u32 flags, u32 mask, int nlflags,
4263 int (*vlan_fill)(struct sk_buff *skb,
4264 struct net_device *dev,
4267 struct nlmsghdr *nlh;
4268 struct ifinfomsg *ifm;
4269 struct nlattr *br_afspec;
4270 struct nlattr *protinfo;
4271 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4272 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4275 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4279 ifm = nlmsg_data(nlh);
4280 ifm->ifi_family = AF_BRIDGE;
4282 ifm->ifi_type = dev->type;
4283 ifm->ifi_index = dev->ifindex;
4284 ifm->ifi_flags = dev_get_flags(dev);
4285 ifm->ifi_change = 0;
4288 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4289 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4290 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4292 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4294 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4295 (dev->ifindex != dev_get_iflink(dev) &&
4296 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4297 goto nla_put_failure;
4299 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4301 goto nla_put_failure;
4303 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4304 nla_nest_cancel(skb, br_afspec);
4305 goto nla_put_failure;
4308 if (mode != BRIDGE_MODE_UNDEF) {
4309 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4310 nla_nest_cancel(skb, br_afspec);
4311 goto nla_put_failure;
4315 err = vlan_fill(skb, dev, filter_mask);
4317 nla_nest_cancel(skb, br_afspec);
4318 goto nla_put_failure;
4321 nla_nest_end(skb, br_afspec);
4323 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4325 goto nla_put_failure;
4327 if (brport_nla_put_flag(skb, flags, mask,
4328 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4329 brport_nla_put_flag(skb, flags, mask,
4330 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4331 brport_nla_put_flag(skb, flags, mask,
4332 IFLA_BRPORT_FAST_LEAVE,
4333 BR_MULTICAST_FAST_LEAVE) ||
4334 brport_nla_put_flag(skb, flags, mask,
4335 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4336 brport_nla_put_flag(skb, flags, mask,
4337 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4338 brport_nla_put_flag(skb, flags, mask,
4339 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4340 brport_nla_put_flag(skb, flags, mask,
4341 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4342 brport_nla_put_flag(skb, flags, mask,
4343 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
4344 nla_nest_cancel(skb, protinfo);
4345 goto nla_put_failure;
4348 nla_nest_end(skb, protinfo);
4350 nlmsg_end(skb, nlh);
4353 nlmsg_cancel(skb, nlh);
4354 return err ? err : -EMSGSIZE;
4356 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4358 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4359 bool strict_check, u32 *filter_mask,
4360 struct netlink_ext_ack *extack)
4362 struct nlattr *tb[IFLA_MAX+1];
4366 struct ifinfomsg *ifm;
4368 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4369 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4373 ifm = nlmsg_data(nlh);
4374 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4375 ifm->ifi_change || ifm->ifi_index) {
4376 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4380 err = nlmsg_parse_deprecated_strict(nlh,
4381 sizeof(struct ifinfomsg),
4382 tb, IFLA_MAX, ifla_policy,
4385 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4386 tb, IFLA_MAX, ifla_policy,
4392 /* new attributes should only be added with strict checking */
4393 for (i = 0; i <= IFLA_MAX; ++i) {
4399 *filter_mask = nla_get_u32(tb[i]);
4403 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4412 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4414 const struct nlmsghdr *nlh = cb->nlh;
4415 struct net *net = sock_net(skb->sk);
4416 struct net_device *dev;
4418 u32 portid = NETLINK_CB(cb->skb).portid;
4419 u32 seq = nlh->nlmsg_seq;
4420 u32 filter_mask = 0;
4423 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4425 if (err < 0 && cb->strict_check)
4429 for_each_netdev_rcu(net, dev) {
4430 const struct net_device_ops *ops = dev->netdev_ops;
4431 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4433 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4434 if (idx >= cb->args[0]) {
4435 err = br_dev->netdev_ops->ndo_bridge_getlink(
4436 skb, portid, seq, dev,
4437 filter_mask, NLM_F_MULTI);
4438 if (err < 0 && err != -EOPNOTSUPP) {
4439 if (likely(skb->len))
4448 if (ops->ndo_bridge_getlink) {
4449 if (idx >= cb->args[0]) {
4450 err = ops->ndo_bridge_getlink(skb, portid,
4454 if (err < 0 && err != -EOPNOTSUPP) {
4455 if (likely(skb->len))
4472 static inline size_t bridge_nlmsg_size(void)
4474 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4475 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4476 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4477 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4478 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
4479 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
4480 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
4481 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
4482 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4483 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
4484 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
4487 static int rtnl_bridge_notify(struct net_device *dev)
4489 struct net *net = dev_net(dev);
4490 struct sk_buff *skb;
4491 int err = -EOPNOTSUPP;
4493 if (!dev->netdev_ops->ndo_bridge_getlink)
4496 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4502 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4509 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4512 WARN_ON(err == -EMSGSIZE);
4515 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4519 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4520 struct netlink_ext_ack *extack)
4522 struct net *net = sock_net(skb->sk);
4523 struct ifinfomsg *ifm;
4524 struct net_device *dev;
4525 struct nlattr *br_spec, *attr = NULL;
4526 int rem, err = -EOPNOTSUPP;
4528 bool have_flags = false;
4530 if (nlmsg_len(nlh) < sizeof(*ifm))
4533 ifm = nlmsg_data(nlh);
4534 if (ifm->ifi_family != AF_BRIDGE)
4535 return -EPFNOSUPPORT;
4537 dev = __dev_get_by_index(net, ifm->ifi_index);
4539 NL_SET_ERR_MSG(extack, "unknown ifindex");
4543 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4545 nla_for_each_nested(attr, br_spec, rem) {
4546 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4547 if (nla_len(attr) < sizeof(flags))
4551 flags = nla_get_u16(attr);
4557 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4558 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4560 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4565 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4570 flags &= ~BRIDGE_FLAGS_MASTER;
4573 if ((flags & BRIDGE_FLAGS_SELF)) {
4574 if (!dev->netdev_ops->ndo_bridge_setlink)
4577 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4581 flags &= ~BRIDGE_FLAGS_SELF;
4583 /* Generate event to notify upper layer of bridge
4586 err = rtnl_bridge_notify(dev);
4591 memcpy(nla_data(attr), &flags, sizeof(flags));
4596 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4597 struct netlink_ext_ack *extack)
4599 struct net *net = sock_net(skb->sk);
4600 struct ifinfomsg *ifm;
4601 struct net_device *dev;
4602 struct nlattr *br_spec, *attr = NULL;
4603 int rem, err = -EOPNOTSUPP;
4605 bool have_flags = false;
4607 if (nlmsg_len(nlh) < sizeof(*ifm))
4610 ifm = nlmsg_data(nlh);
4611 if (ifm->ifi_family != AF_BRIDGE)
4612 return -EPFNOSUPPORT;
4614 dev = __dev_get_by_index(net, ifm->ifi_index);
4616 NL_SET_ERR_MSG(extack, "unknown ifindex");
4620 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4622 nla_for_each_nested(attr, br_spec, rem) {
4623 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4624 if (nla_len(attr) < sizeof(flags))
4628 flags = nla_get_u16(attr);
4634 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4635 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4637 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4642 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4646 flags &= ~BRIDGE_FLAGS_MASTER;
4649 if ((flags & BRIDGE_FLAGS_SELF)) {
4650 if (!dev->netdev_ops->ndo_bridge_dellink)
4653 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4657 flags &= ~BRIDGE_FLAGS_SELF;
4659 /* Generate event to notify upper layer of bridge
4662 err = rtnl_bridge_notify(dev);
4667 memcpy(nla_data(attr), &flags, sizeof(flags));
4672 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4674 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4675 (!idxattr || idxattr == attrid);
4678 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4679 static int rtnl_get_offload_stats_attr_size(int attr_id)
4682 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4683 return sizeof(struct rtnl_link_stats64);
4689 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4692 struct nlattr *attr = NULL;
4697 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4698 dev->netdev_ops->ndo_get_offload_stats))
4701 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4702 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4703 if (attr_id < *prividx)
4706 size = rtnl_get_offload_stats_attr_size(attr_id);
4710 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4713 attr = nla_reserve_64bit(skb, attr_id, size,
4714 IFLA_OFFLOAD_XSTATS_UNSPEC);
4716 goto nla_put_failure;
4718 attr_data = nla_data(attr);
4719 memset(attr_data, 0, size);
4720 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4723 goto get_offload_stats_failure;
4734 get_offload_stats_failure:
4739 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4745 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4746 dev->netdev_ops->ndo_get_offload_stats))
4749 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4750 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4751 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4753 size = rtnl_get_offload_stats_attr_size(attr_id);
4754 nla_size += nla_total_size_64bit(size);
4758 nla_size += nla_total_size(0);
4763 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4764 int type, u32 pid, u32 seq, u32 change,
4765 unsigned int flags, unsigned int filter_mask,
4766 int *idxattr, int *prividx)
4768 struct if_stats_msg *ifsm;
4769 struct nlmsghdr *nlh;
4770 struct nlattr *attr;
4771 int s_prividx = *prividx;
4776 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4780 ifsm = nlmsg_data(nlh);
4781 ifsm->family = PF_UNSPEC;
4784 ifsm->ifindex = dev->ifindex;
4785 ifsm->filter_mask = filter_mask;
4787 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4788 struct rtnl_link_stats64 *sp;
4790 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4791 sizeof(struct rtnl_link_stats64),
4794 goto nla_put_failure;
4796 sp = nla_data(attr);
4797 dev_get_stats(dev, sp);
4800 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4801 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4803 if (ops && ops->fill_linkxstats) {
4804 *idxattr = IFLA_STATS_LINK_XSTATS;
4805 attr = nla_nest_start_noflag(skb,
4806 IFLA_STATS_LINK_XSTATS);
4808 goto nla_put_failure;
4810 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4811 nla_nest_end(skb, attr);
4813 goto nla_put_failure;
4818 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4820 const struct rtnl_link_ops *ops = NULL;
4821 const struct net_device *master;
4823 master = netdev_master_upper_dev_get(dev);
4825 ops = master->rtnl_link_ops;
4826 if (ops && ops->fill_linkxstats) {
4827 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4828 attr = nla_nest_start_noflag(skb,
4829 IFLA_STATS_LINK_XSTATS_SLAVE);
4831 goto nla_put_failure;
4833 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4834 nla_nest_end(skb, attr);
4836 goto nla_put_failure;
4841 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4843 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4844 attr = nla_nest_start_noflag(skb,
4845 IFLA_STATS_LINK_OFFLOAD_XSTATS);
4847 goto nla_put_failure;
4849 err = rtnl_get_offload_stats(skb, dev, prividx);
4850 if (err == -ENODATA)
4851 nla_nest_cancel(skb, attr);
4853 nla_nest_end(skb, attr);
4855 if (err && err != -ENODATA)
4856 goto nla_put_failure;
4860 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4861 struct rtnl_af_ops *af_ops;
4863 *idxattr = IFLA_STATS_AF_SPEC;
4864 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
4866 goto nla_put_failure;
4869 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4870 if (af_ops->fill_stats_af) {
4874 af = nla_nest_start_noflag(skb,
4878 goto nla_put_failure;
4880 err = af_ops->fill_stats_af(skb, dev);
4882 if (err == -ENODATA) {
4883 nla_nest_cancel(skb, af);
4884 } else if (err < 0) {
4886 goto nla_put_failure;
4889 nla_nest_end(skb, af);
4894 nla_nest_end(skb, attr);
4899 nlmsg_end(skb, nlh);
4904 /* not a multi message or no progress mean a real error */
4905 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4906 nlmsg_cancel(skb, nlh);
4908 nlmsg_end(skb, nlh);
4913 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4918 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4919 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4921 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4922 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4923 int attr = IFLA_STATS_LINK_XSTATS;
4925 if (ops && ops->get_linkxstats_size) {
4926 size += nla_total_size(ops->get_linkxstats_size(dev,
4928 /* for IFLA_STATS_LINK_XSTATS */
4929 size += nla_total_size(0);
4933 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4934 struct net_device *_dev = (struct net_device *)dev;
4935 const struct rtnl_link_ops *ops = NULL;
4936 const struct net_device *master;
4938 /* netdev_master_upper_dev_get can't take const */
4939 master = netdev_master_upper_dev_get(_dev);
4941 ops = master->rtnl_link_ops;
4942 if (ops && ops->get_linkxstats_size) {
4943 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4945 size += nla_total_size(ops->get_linkxstats_size(dev,
4947 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4948 size += nla_total_size(0);
4952 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4953 size += rtnl_get_offload_stats_size(dev);
4955 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4956 struct rtnl_af_ops *af_ops;
4958 /* for IFLA_STATS_AF_SPEC */
4959 size += nla_total_size(0);
4962 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4963 if (af_ops->get_stats_af_size) {
4964 size += nla_total_size(
4965 af_ops->get_stats_af_size(dev));
4968 size += nla_total_size(0);
4977 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
4978 bool is_dump, struct netlink_ext_ack *extack)
4980 struct if_stats_msg *ifsm;
4982 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
4983 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
4990 ifsm = nlmsg_data(nlh);
4992 /* only requests using strict checks can pass data to influence
4993 * the dump. The legacy exception is filter_mask.
4995 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
4996 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
4999 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5000 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5003 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5004 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5011 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5012 struct netlink_ext_ack *extack)
5014 struct net *net = sock_net(skb->sk);
5015 struct net_device *dev = NULL;
5016 int idxattr = 0, prividx = 0;
5017 struct if_stats_msg *ifsm;
5018 struct sk_buff *nskb;
5022 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5027 ifsm = nlmsg_data(nlh);
5028 if (ifsm->ifindex > 0)
5029 dev = __dev_get_by_index(net, ifsm->ifindex);
5036 filter_mask = ifsm->filter_mask;
5040 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5044 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5045 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5046 0, filter_mask, &idxattr, &prividx);
5048 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5049 WARN_ON(err == -EMSGSIZE);
5052 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5058 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5060 struct netlink_ext_ack *extack = cb->extack;
5061 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5062 struct net *net = sock_net(skb->sk);
5063 unsigned int flags = NLM_F_MULTI;
5064 struct if_stats_msg *ifsm;
5065 struct hlist_head *head;
5066 struct net_device *dev;
5067 u32 filter_mask = 0;
5071 s_idx = cb->args[1];
5072 s_idxattr = cb->args[2];
5073 s_prividx = cb->args[3];
5075 cb->seq = net->dev_base_seq;
5077 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5081 ifsm = nlmsg_data(cb->nlh);
5082 filter_mask = ifsm->filter_mask;
5084 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5088 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5090 head = &net->dev_index_head[h];
5091 hlist_for_each_entry(dev, head, index_hlist) {
5094 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5095 NETLINK_CB(cb->skb).portid,
5096 cb->nlh->nlmsg_seq, 0,
5098 &s_idxattr, &s_prividx);
5099 /* If we ran out of room on the first message,
5102 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5108 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5114 cb->args[3] = s_prividx;
5115 cb->args[2] = s_idxattr;
5122 /* Process one rtnetlink message. */
5124 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5125 struct netlink_ext_ack *extack)
5127 struct net *net = sock_net(skb->sk);
5128 struct rtnl_link *link;
5129 struct module *owner;
5130 int err = -EOPNOTSUPP;
5131 rtnl_doit_func doit;
5137 type = nlh->nlmsg_type;
5143 /* All the messages must have at least 1 byte length */
5144 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5147 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5150 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5154 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5156 rtnl_dumpit_func dumpit;
5157 u16 min_dump_alloc = 0;
5159 link = rtnl_get_link(family, type);
5160 if (!link || !link->dumpit) {
5162 link = rtnl_get_link(family, type);
5163 if (!link || !link->dumpit)
5166 owner = link->owner;
5167 dumpit = link->dumpit;
5169 if (type == RTM_GETLINK - RTM_BASE)
5170 min_dump_alloc = rtnl_calcit(skb, nlh);
5173 /* need to do this before rcu_read_unlock() */
5174 if (!try_module_get(owner))
5175 err = -EPROTONOSUPPORT;
5181 struct netlink_dump_control c = {
5183 .min_dump_alloc = min_dump_alloc,
5186 err = netlink_dump_start(rtnl, skb, nlh, &c);
5187 /* netlink_dump_start() will keep a reference on
5188 * module if dump is still in progress.
5195 link = rtnl_get_link(family, type);
5196 if (!link || !link->doit) {
5198 link = rtnl_get_link(PF_UNSPEC, type);
5199 if (!link || !link->doit)
5203 owner = link->owner;
5204 if (!try_module_get(owner)) {
5205 err = -EPROTONOSUPPORT;
5209 flags = link->flags;
5210 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5214 err = doit(skb, nlh, extack);
5221 link = rtnl_get_link(family, type);
5222 if (link && link->doit)
5223 err = link->doit(skb, nlh, extack);
5239 static void rtnetlink_rcv(struct sk_buff *skb)
5241 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5244 static int rtnetlink_bind(struct net *net, int group)
5247 case RTNLGRP_IPV4_MROUTE_R:
5248 case RTNLGRP_IPV6_MROUTE_R:
5249 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5256 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5258 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5262 case NETDEV_CHANGEMTU:
5263 case NETDEV_CHANGEADDR:
5264 case NETDEV_CHANGENAME:
5265 case NETDEV_FEAT_CHANGE:
5266 case NETDEV_BONDING_FAILOVER:
5267 case NETDEV_POST_TYPE_CHANGE:
5268 case NETDEV_NOTIFY_PEERS:
5269 case NETDEV_CHANGEUPPER:
5270 case NETDEV_RESEND_IGMP:
5271 case NETDEV_CHANGEINFODATA:
5272 case NETDEV_CHANGELOWERSTATE:
5273 case NETDEV_CHANGE_TX_QUEUE_LEN:
5274 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5275 GFP_KERNEL, NULL, 0);
5283 static struct notifier_block rtnetlink_dev_notifier = {
5284 .notifier_call = rtnetlink_event,
5288 static int __net_init rtnetlink_net_init(struct net *net)
5291 struct netlink_kernel_cfg cfg = {
5292 .groups = RTNLGRP_MAX,
5293 .input = rtnetlink_rcv,
5294 .cb_mutex = &rtnl_mutex,
5295 .flags = NL_CFG_F_NONROOT_RECV,
5296 .bind = rtnetlink_bind,
5299 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5306 static void __net_exit rtnetlink_net_exit(struct net *net)
5308 netlink_kernel_release(net->rtnl);
5312 static struct pernet_operations rtnetlink_net_ops = {
5313 .init = rtnetlink_net_init,
5314 .exit = rtnetlink_net_exit,
5317 void __init rtnetlink_init(void)
5319 if (register_pernet_subsys(&rtnetlink_net_ops))
5320 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5322 register_netdevice_notifier(&rtnetlink_dev_notifier);
5324 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5325 rtnl_dump_ifinfo, 0);
5326 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5327 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5328 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5330 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5331 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5332 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5334 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5335 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5336 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5338 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5339 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5340 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5342 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,