1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Routing netlink socket interface: protocol independent part.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
12 * Vitaly E. Lavrov RTA_OK arithmetic was wrong.
15 #include <linux/bitops.h>
16 #include <linux/errno.h>
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/kernel.h>
21 #include <linux/timer.h>
22 #include <linux/string.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/fcntl.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/capability.h>
30 #include <linux/skbuff.h>
31 #include <linux/init.h>
32 #include <linux/security.h>
33 #include <linux/mutex.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_vlan.h>
37 #include <linux/pci.h>
38 #include <linux/etherdevice.h>
39 #include <linux/bpf.h>
41 #include <linux/uaccess.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
46 #include <net/protocol.h>
48 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
57 #define RTNL_MAX_TYPE 50
58 #define RTNL_SLAVE_MAX_TYPE 40
62 rtnl_dumpit_func dumpit;
68 static DEFINE_MUTEX(rtnl_mutex);
72 mutex_lock(&rtnl_mutex);
74 EXPORT_SYMBOL(rtnl_lock);
76 int rtnl_lock_killable(void)
78 return mutex_lock_killable(&rtnl_mutex);
80 EXPORT_SYMBOL(rtnl_lock_killable);
82 static struct sk_buff *defer_kfree_skb_list;
83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
86 tail->next = defer_kfree_skb_list;
87 defer_kfree_skb_list = head;
90 EXPORT_SYMBOL(rtnl_kfree_skbs);
92 void __rtnl_unlock(void)
94 struct sk_buff *head = defer_kfree_skb_list;
96 defer_kfree_skb_list = NULL;
98 mutex_unlock(&rtnl_mutex);
101 struct sk_buff *next = head->next;
109 void rtnl_unlock(void)
111 /* This fellow will unlock it for us. */
114 EXPORT_SYMBOL(rtnl_unlock);
116 int rtnl_trylock(void)
118 return mutex_trylock(&rtnl_mutex);
120 EXPORT_SYMBOL(rtnl_trylock);
122 int rtnl_is_locked(void)
124 return mutex_is_locked(&rtnl_mutex);
126 EXPORT_SYMBOL(rtnl_is_locked);
128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
130 return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
134 #ifdef CONFIG_PROVE_LOCKING
135 bool lockdep_rtnl_is_held(void)
137 return lockdep_is_held(&rtnl_mutex);
139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
142 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
144 static inline int rtm_msgindex(int msgtype)
146 int msgindex = msgtype - RTM_BASE;
149 * msgindex < 0 implies someone tried to register a netlink
150 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
151 * the message type has not been added to linux/rtnetlink.h
153 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
160 struct rtnl_link __rcu **tab;
162 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
163 protocol = PF_UNSPEC;
165 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
167 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
169 return rcu_dereference_rtnl(tab[msgtype]);
172 static int rtnl_register_internal(struct module *owner,
173 int protocol, int msgtype,
174 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
177 struct rtnl_link *link, *old;
178 struct rtnl_link __rcu **tab;
182 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
183 msgindex = rtm_msgindex(msgtype);
186 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
188 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
192 /* ensures we see the 0 stores */
193 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
196 old = rtnl_dereference(tab[msgindex]);
198 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
202 link = kzalloc(sizeof(*link), GFP_KERNEL);
207 WARN_ON(link->owner && link->owner != owner);
210 WARN_ON(doit && link->doit && link->doit != doit);
213 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
215 link->dumpit = dumpit;
217 link->flags |= flags;
219 /* publish protocol:msgtype */
220 rcu_assign_pointer(tab[msgindex], link);
230 * rtnl_register_module - Register a rtnetlink message type
232 * @owner: module registering the hook (THIS_MODULE)
233 * @protocol: Protocol family or PF_UNSPEC
234 * @msgtype: rtnetlink message type
235 * @doit: Function pointer called for each request message
236 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
237 * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
239 * Like rtnl_register, but for use by removable modules.
241 int rtnl_register_module(struct module *owner,
242 int protocol, int msgtype,
243 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
246 return rtnl_register_internal(owner, protocol, msgtype,
247 doit, dumpit, flags);
249 EXPORT_SYMBOL_GPL(rtnl_register_module);
252 * rtnl_register - Register a rtnetlink message type
253 * @protocol: Protocol family or PF_UNSPEC
254 * @msgtype: rtnetlink message type
255 * @doit: Function pointer called for each request message
256 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
257 * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
259 * Registers the specified function pointers (at least one of them has
260 * to be non-NULL) to be called whenever a request message for the
261 * specified protocol family and message type is received.
263 * The special protocol family PF_UNSPEC may be used to define fallback
264 * function pointers for the case when no entry for the specific protocol
267 void rtnl_register(int protocol, int msgtype,
268 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
273 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
276 pr_err("Unable to register rtnetlink message handler, "
277 "protocol = %d, message type = %d\n", protocol, msgtype);
281 * rtnl_unregister - Unregister a rtnetlink message type
282 * @protocol: Protocol family or PF_UNSPEC
283 * @msgtype: rtnetlink message type
285 * Returns 0 on success or a negative error code.
287 int rtnl_unregister(int protocol, int msgtype)
289 struct rtnl_link __rcu **tab;
290 struct rtnl_link *link;
293 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
294 msgindex = rtm_msgindex(msgtype);
297 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
303 link = rtnl_dereference(tab[msgindex]);
304 rcu_assign_pointer(tab[msgindex], NULL);
307 kfree_rcu(link, rcu);
311 EXPORT_SYMBOL_GPL(rtnl_unregister);
314 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
315 * @protocol : Protocol family or PF_UNSPEC
317 * Identical to calling rtnl_unregster() for all registered message types
318 * of a certain protocol family.
320 void rtnl_unregister_all(int protocol)
322 struct rtnl_link __rcu **tab;
323 struct rtnl_link *link;
326 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
329 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
334 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
335 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
336 link = rtnl_dereference(tab[msgindex]);
340 rcu_assign_pointer(tab[msgindex], NULL);
341 kfree_rcu(link, rcu);
349 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
351 static LIST_HEAD(link_ops);
353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
355 const struct rtnl_link_ops *ops;
357 list_for_each_entry(ops, &link_ops, list) {
358 if (!strcmp(ops->kind, kind))
365 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
366 * @ops: struct rtnl_link_ops * to register
368 * The caller must hold the rtnl_mutex. This function should be used
369 * by drivers that create devices during module initialization. It
370 * must be called before registering the devices.
372 * Returns 0 on success or a negative error code.
374 int __rtnl_link_register(struct rtnl_link_ops *ops)
376 if (rtnl_link_ops_get(ops->kind))
379 /* The check for alloc/setup is here because if ops
380 * does not have that filled up, it is not possible
381 * to use the ops for creating device. So do not
382 * fill up dellink as well. That disables rtnl_dellink.
384 if ((ops->alloc || ops->setup) && !ops->dellink)
385 ops->dellink = unregister_netdevice_queue;
387 list_add_tail(&ops->list, &link_ops);
390 EXPORT_SYMBOL_GPL(__rtnl_link_register);
393 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
394 * @ops: struct rtnl_link_ops * to register
396 * Returns 0 on success or a negative error code.
398 int rtnl_link_register(struct rtnl_link_ops *ops)
402 /* Sanity-check max sizes to avoid stack buffer overflow. */
403 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
404 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
408 err = __rtnl_link_register(ops);
412 EXPORT_SYMBOL_GPL(rtnl_link_register);
414 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
416 struct net_device *dev;
417 LIST_HEAD(list_kill);
419 for_each_netdev(net, dev) {
420 if (dev->rtnl_link_ops == ops)
421 ops->dellink(dev, &list_kill);
423 unregister_netdevice_many(&list_kill);
427 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
428 * @ops: struct rtnl_link_ops * to unregister
430 * The caller must hold the rtnl_mutex and guarantee net_namespace_list
431 * integrity (hold pernet_ops_rwsem for writing to close the race
432 * with setup_net() and cleanup_net()).
434 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
439 __rtnl_kill_links(net, ops);
441 list_del(&ops->list);
443 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
445 /* Return with the rtnl_lock held when there are no network
446 * devices unregistering in any network namespace.
448 static void rtnl_lock_unregistering_all(void)
452 DEFINE_WAIT_FUNC(wait, woken_wake_function);
454 add_wait_queue(&netdev_unregistering_wq, &wait);
456 unregistering = false;
458 /* We held write locked pernet_ops_rwsem, and parallel
459 * setup_net() and cleanup_net() are not possible.
462 if (net->dev_unreg_count > 0) {
463 unregistering = true;
471 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
473 remove_wait_queue(&netdev_unregistering_wq, &wait);
477 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
478 * @ops: struct rtnl_link_ops * to unregister
480 void rtnl_link_unregister(struct rtnl_link_ops *ops)
482 /* Close the race with setup_net() and cleanup_net() */
483 down_write(&pernet_ops_rwsem);
484 rtnl_lock_unregistering_all();
485 __rtnl_link_unregister(ops);
487 up_write(&pernet_ops_rwsem);
489 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
491 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
493 struct net_device *master_dev;
494 const struct rtnl_link_ops *ops;
499 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
503 ops = master_dev->rtnl_link_ops;
504 if (!ops || !ops->get_slave_size)
506 /* IFLA_INFO_SLAVE_DATA + nested data */
507 size = nla_total_size(sizeof(struct nlattr)) +
508 ops->get_slave_size(master_dev, dev);
515 static size_t rtnl_link_get_size(const struct net_device *dev)
517 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
523 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
524 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
527 /* IFLA_INFO_DATA + nested data */
528 size += nla_total_size(sizeof(struct nlattr)) +
531 if (ops->get_xstats_size)
532 /* IFLA_INFO_XSTATS */
533 size += nla_total_size(ops->get_xstats_size(dev));
535 size += rtnl_link_get_slave_info_data_size(dev);
540 static LIST_HEAD(rtnl_af_ops);
542 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
544 const struct rtnl_af_ops *ops;
548 list_for_each_entry(ops, &rtnl_af_ops, list) {
549 if (ops->family == family)
557 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
558 * @ops: struct rtnl_af_ops * to register
560 * Returns 0 on success or a negative error code.
562 void rtnl_af_register(struct rtnl_af_ops *ops)
565 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
568 EXPORT_SYMBOL_GPL(rtnl_af_register);
571 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
572 * @ops: struct rtnl_af_ops * to unregister
574 void rtnl_af_unregister(struct rtnl_af_ops *ops)
577 list_del_rcu(&ops->list);
582 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
584 static size_t rtnl_link_get_af_size(const struct net_device *dev,
587 struct rtnl_af_ops *af_ops;
591 size = nla_total_size(sizeof(struct nlattr));
594 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
595 if (af_ops->get_link_af_size) {
596 /* AF_* + nested data */
597 size += nla_total_size(sizeof(struct nlattr)) +
598 af_ops->get_link_af_size(dev, ext_filter_mask);
606 static bool rtnl_have_link_slave_info(const struct net_device *dev)
608 struct net_device *master_dev;
613 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
614 if (master_dev && master_dev->rtnl_link_ops)
620 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
621 const struct net_device *dev)
623 struct net_device *master_dev;
624 const struct rtnl_link_ops *ops;
625 struct nlattr *slave_data;
628 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
631 ops = master_dev->rtnl_link_ops;
634 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
636 if (ops->fill_slave_info) {
637 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
640 err = ops->fill_slave_info(skb, master_dev, dev);
642 goto err_cancel_slave_data;
643 nla_nest_end(skb, slave_data);
647 err_cancel_slave_data:
648 nla_nest_cancel(skb, slave_data);
652 static int rtnl_link_info_fill(struct sk_buff *skb,
653 const struct net_device *dev)
655 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
661 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
663 if (ops->fill_xstats) {
664 err = ops->fill_xstats(skb, dev);
668 if (ops->fill_info) {
669 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
672 err = ops->fill_info(skb, dev);
674 goto err_cancel_data;
675 nla_nest_end(skb, data);
680 nla_nest_cancel(skb, data);
684 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
686 struct nlattr *linkinfo;
689 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
690 if (linkinfo == NULL)
693 err = rtnl_link_info_fill(skb, dev);
695 goto err_cancel_link;
697 err = rtnl_link_slave_info_fill(skb, dev);
699 goto err_cancel_link;
701 nla_nest_end(skb, linkinfo);
705 nla_nest_cancel(skb, linkinfo);
710 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
712 struct sock *rtnl = net->rtnl;
715 NETLINK_CB(skb).dst_group = group;
717 refcount_inc(&skb->users);
718 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
720 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
724 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
726 struct sock *rtnl = net->rtnl;
728 return nlmsg_unicast(rtnl, skb, pid);
730 EXPORT_SYMBOL(rtnl_unicast);
732 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
733 struct nlmsghdr *nlh, gfp_t flags)
735 struct sock *rtnl = net->rtnl;
739 report = nlmsg_report(nlh);
741 nlmsg_notify(rtnl, skb, pid, group, report, flags);
743 EXPORT_SYMBOL(rtnl_notify);
745 void rtnl_set_sk_err(struct net *net, u32 group, int error)
747 struct sock *rtnl = net->rtnl;
749 netlink_set_err(rtnl, 0, group, error);
751 EXPORT_SYMBOL(rtnl_set_sk_err);
753 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
758 /* nothing is dumped for dst_default_metrics, so just skip the loop */
759 if (metrics == dst_default_metrics.metrics)
762 mx = nla_nest_start_noflag(skb, RTA_METRICS);
766 for (i = 0; i < RTAX_MAX; i++) {
768 if (i == RTAX_CC_ALGO - 1) {
769 char tmp[TCP_CA_NAME_MAX], *name;
771 name = tcp_ca_get_name_by_key(metrics[i], tmp);
774 if (nla_put_string(skb, i + 1, name))
775 goto nla_put_failure;
776 } else if (i == RTAX_FEATURES - 1) {
777 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
781 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
782 if (nla_put_u32(skb, i + 1, user_features))
783 goto nla_put_failure;
785 if (nla_put_u32(skb, i + 1, metrics[i]))
786 goto nla_put_failure;
793 nla_nest_cancel(skb, mx);
797 return nla_nest_end(skb, mx);
800 nla_nest_cancel(skb, mx);
803 EXPORT_SYMBOL(rtnetlink_put_metrics);
805 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
806 long expires, u32 error)
808 struct rta_cacheinfo ci = {
814 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
815 ci.rta_used = dst->__use;
816 ci.rta_clntref = atomic_read(&dst->__refcnt);
821 clock = jiffies_to_clock_t(abs(expires));
822 clock = min_t(unsigned long, clock, INT_MAX);
823 ci.rta_expires = (expires > 0) ? clock : -clock;
825 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
827 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
829 static void set_operstate(struct net_device *dev, unsigned char transition)
831 unsigned char operstate = dev->operstate;
833 switch (transition) {
835 if ((operstate == IF_OPER_DORMANT ||
836 operstate == IF_OPER_TESTING ||
837 operstate == IF_OPER_UNKNOWN) &&
838 !netif_dormant(dev) && !netif_testing(dev))
839 operstate = IF_OPER_UP;
842 case IF_OPER_TESTING:
843 if (operstate == IF_OPER_UP ||
844 operstate == IF_OPER_UNKNOWN)
845 operstate = IF_OPER_TESTING;
848 case IF_OPER_DORMANT:
849 if (operstate == IF_OPER_UP ||
850 operstate == IF_OPER_UNKNOWN)
851 operstate = IF_OPER_DORMANT;
855 if (dev->operstate != operstate) {
856 write_lock_bh(&dev_base_lock);
857 dev->operstate = operstate;
858 write_unlock_bh(&dev_base_lock);
859 netdev_state_change(dev);
863 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
865 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
866 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
869 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
870 const struct ifinfomsg *ifm)
872 unsigned int flags = ifm->ifi_flags;
874 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
876 flags = (flags & ifm->ifi_change) |
877 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
882 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
883 const struct rtnl_link_stats64 *b)
885 a->rx_packets = b->rx_packets;
886 a->tx_packets = b->tx_packets;
887 a->rx_bytes = b->rx_bytes;
888 a->tx_bytes = b->tx_bytes;
889 a->rx_errors = b->rx_errors;
890 a->tx_errors = b->tx_errors;
891 a->rx_dropped = b->rx_dropped;
892 a->tx_dropped = b->tx_dropped;
894 a->multicast = b->multicast;
895 a->collisions = b->collisions;
897 a->rx_length_errors = b->rx_length_errors;
898 a->rx_over_errors = b->rx_over_errors;
899 a->rx_crc_errors = b->rx_crc_errors;
900 a->rx_frame_errors = b->rx_frame_errors;
901 a->rx_fifo_errors = b->rx_fifo_errors;
902 a->rx_missed_errors = b->rx_missed_errors;
904 a->tx_aborted_errors = b->tx_aborted_errors;
905 a->tx_carrier_errors = b->tx_carrier_errors;
906 a->tx_fifo_errors = b->tx_fifo_errors;
907 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
908 a->tx_window_errors = b->tx_window_errors;
910 a->rx_compressed = b->rx_compressed;
911 a->tx_compressed = b->tx_compressed;
913 a->rx_nohandler = b->rx_nohandler;
917 static inline int rtnl_vfinfo_size(const struct net_device *dev,
920 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
921 int num_vfs = dev_num_vf(dev->dev.parent);
922 size_t size = nla_total_size(0);
925 nla_total_size(sizeof(struct ifla_vf_mac)) +
926 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
927 nla_total_size(sizeof(struct ifla_vf_vlan)) +
928 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
929 nla_total_size(MAX_VLAN_LIST_LEN *
930 sizeof(struct ifla_vf_vlan_info)) +
931 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
932 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
933 nla_total_size(sizeof(struct ifla_vf_rate)) +
934 nla_total_size(sizeof(struct ifla_vf_link_state)) +
935 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
936 nla_total_size(0) + /* nest IFLA_VF_STATS */
937 /* IFLA_VF_STATS_RX_PACKETS */
938 nla_total_size_64bit(sizeof(__u64)) +
939 /* IFLA_VF_STATS_TX_PACKETS */
940 nla_total_size_64bit(sizeof(__u64)) +
941 /* IFLA_VF_STATS_RX_BYTES */
942 nla_total_size_64bit(sizeof(__u64)) +
943 /* IFLA_VF_STATS_TX_BYTES */
944 nla_total_size_64bit(sizeof(__u64)) +
945 /* IFLA_VF_STATS_BROADCAST */
946 nla_total_size_64bit(sizeof(__u64)) +
947 /* IFLA_VF_STATS_MULTICAST */
948 nla_total_size_64bit(sizeof(__u64)) +
949 /* IFLA_VF_STATS_RX_DROPPED */
950 nla_total_size_64bit(sizeof(__u64)) +
951 /* IFLA_VF_STATS_TX_DROPPED */
952 nla_total_size_64bit(sizeof(__u64)) +
953 nla_total_size(sizeof(struct ifla_vf_trust)));
959 static size_t rtnl_port_size(const struct net_device *dev,
962 size_t port_size = nla_total_size(4) /* PORT_VF */
963 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
964 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
965 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
966 + nla_total_size(1) /* PROT_VDP_REQUEST */
967 + nla_total_size(2); /* PORT_VDP_RESPONSE */
968 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
969 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
971 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
974 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
975 !(ext_filter_mask & RTEXT_FILTER_VF))
977 if (dev_num_vf(dev->dev.parent))
978 return port_self_size + vf_ports_size +
979 vf_port_size * dev_num_vf(dev->dev.parent);
981 return port_self_size;
984 static size_t rtnl_xdp_size(void)
986 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
987 nla_total_size(1) + /* XDP_ATTACHED */
988 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
989 nla_total_size(4); /* XDP_<mode>_PROG_ID */
994 static size_t rtnl_prop_list_size(const struct net_device *dev)
996 struct netdev_name_node *name_node;
999 if (list_empty(&dev->name_node->list))
1001 size = nla_total_size(0);
1002 list_for_each_entry(name_node, &dev->name_node->list, list)
1003 size += nla_total_size(ALTIFNAMSIZ);
1007 static size_t rtnl_proto_down_size(const struct net_device *dev)
1009 size_t size = nla_total_size(1);
1011 if (dev->proto_down_reason)
1012 size += nla_total_size(0) + nla_total_size(4);
1017 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1018 u32 ext_filter_mask)
1020 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1021 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1022 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1023 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1024 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1025 + nla_total_size(sizeof(struct rtnl_link_stats))
1026 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1027 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1028 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1029 + nla_total_size(4) /* IFLA_TXQLEN */
1030 + nla_total_size(4) /* IFLA_WEIGHT */
1031 + nla_total_size(4) /* IFLA_MTU */
1032 + nla_total_size(4) /* IFLA_LINK */
1033 + nla_total_size(4) /* IFLA_MASTER */
1034 + nla_total_size(1) /* IFLA_CARRIER */
1035 + nla_total_size(4) /* IFLA_PROMISCUITY */
1036 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1037 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1038 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1039 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1040 + nla_total_size(1) /* IFLA_OPERSTATE */
1041 + nla_total_size(1) /* IFLA_LINKMODE */
1042 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1043 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1044 + nla_total_size(4) /* IFLA_GROUP */
1045 + nla_total_size(ext_filter_mask
1046 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1047 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1048 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1049 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1050 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1051 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1052 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1053 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1054 + rtnl_xdp_size() /* IFLA_XDP */
1055 + nla_total_size(4) /* IFLA_EVENT */
1056 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1057 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1058 + rtnl_proto_down_size(dev) /* proto down */
1059 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1060 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1061 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1062 + nla_total_size(4) /* IFLA_MIN_MTU */
1063 + nla_total_size(4) /* IFLA_MAX_MTU */
1064 + rtnl_prop_list_size(dev)
1065 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1069 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1071 struct nlattr *vf_ports;
1072 struct nlattr *vf_port;
1076 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1080 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1081 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1083 goto nla_put_failure;
1084 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1085 goto nla_put_failure;
1086 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1087 if (err == -EMSGSIZE)
1088 goto nla_put_failure;
1090 nla_nest_cancel(skb, vf_port);
1093 nla_nest_end(skb, vf_port);
1096 nla_nest_end(skb, vf_ports);
1101 nla_nest_cancel(skb, vf_ports);
1105 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1107 struct nlattr *port_self;
1110 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1114 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1116 nla_nest_cancel(skb, port_self);
1117 return (err == -EMSGSIZE) ? err : 0;
1120 nla_nest_end(skb, port_self);
1125 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1126 u32 ext_filter_mask)
1130 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1131 !(ext_filter_mask & RTEXT_FILTER_VF))
1134 err = rtnl_port_self_fill(skb, dev);
1138 if (dev_num_vf(dev->dev.parent)) {
1139 err = rtnl_vf_ports_fill(skb, dev);
1147 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1150 struct netdev_phys_item_id ppid;
1152 err = dev_get_phys_port_id(dev, &ppid);
1154 if (err == -EOPNOTSUPP)
1159 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1165 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1167 char name[IFNAMSIZ];
1170 err = dev_get_phys_port_name(dev, name, sizeof(name));
1172 if (err == -EOPNOTSUPP)
1177 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1183 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1185 struct netdev_phys_item_id ppid = { };
1188 err = dev_get_port_parent_id(dev, &ppid, false);
1190 if (err == -EOPNOTSUPP)
1195 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1201 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1202 struct net_device *dev)
1204 struct rtnl_link_stats64 *sp;
1205 struct nlattr *attr;
1207 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1208 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1212 sp = nla_data(attr);
1213 dev_get_stats(dev, sp);
1215 attr = nla_reserve(skb, IFLA_STATS,
1216 sizeof(struct rtnl_link_stats));
1220 copy_rtnl_link_stats(nla_data(attr), sp);
1225 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1226 struct net_device *dev,
1228 struct nlattr *vfinfo)
1230 struct ifla_vf_rss_query_en vf_rss_query_en;
1231 struct nlattr *vf, *vfstats, *vfvlanlist;
1232 struct ifla_vf_link_state vf_linkstate;
1233 struct ifla_vf_vlan_info vf_vlan_info;
1234 struct ifla_vf_spoofchk vf_spoofchk;
1235 struct ifla_vf_tx_rate vf_tx_rate;
1236 struct ifla_vf_stats vf_stats;
1237 struct ifla_vf_trust vf_trust;
1238 struct ifla_vf_vlan vf_vlan;
1239 struct ifla_vf_rate vf_rate;
1240 struct ifla_vf_mac vf_mac;
1241 struct ifla_vf_broadcast vf_broadcast;
1242 struct ifla_vf_info ivi;
1243 struct ifla_vf_guid node_guid;
1244 struct ifla_vf_guid port_guid;
1246 memset(&ivi, 0, sizeof(ivi));
1248 /* Not all SR-IOV capable drivers support the
1249 * spoofcheck and "RSS query enable" query. Preset to
1250 * -1 so the user space tool can detect that the driver
1251 * didn't report anything.
1254 ivi.rss_query_en = -1;
1256 /* The default value for VF link state is "auto"
1257 * IFLA_VF_LINK_STATE_AUTO which equals zero
1260 /* VLAN Protocol by default is 802.1Q */
1261 ivi.vlan_proto = htons(ETH_P_8021Q);
1262 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1265 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1266 memset(&node_guid, 0, sizeof(node_guid));
1267 memset(&port_guid, 0, sizeof(port_guid));
1276 vf_rss_query_en.vf =
1279 port_guid.vf = ivi.vf;
1281 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1282 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1283 vf_vlan.vlan = ivi.vlan;
1284 vf_vlan.qos = ivi.qos;
1285 vf_vlan_info.vlan = ivi.vlan;
1286 vf_vlan_info.qos = ivi.qos;
1287 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1288 vf_tx_rate.rate = ivi.max_tx_rate;
1289 vf_rate.min_tx_rate = ivi.min_tx_rate;
1290 vf_rate.max_tx_rate = ivi.max_tx_rate;
1291 vf_spoofchk.setting = ivi.spoofchk;
1292 vf_linkstate.link_state = ivi.linkstate;
1293 vf_rss_query_en.setting = ivi.rss_query_en;
1294 vf_trust.setting = ivi.trusted;
1295 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1297 goto nla_put_vfinfo_failure;
1298 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1299 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1300 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1301 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1303 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1305 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1307 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1309 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1310 sizeof(vf_rss_query_en),
1311 &vf_rss_query_en) ||
1312 nla_put(skb, IFLA_VF_TRUST,
1313 sizeof(vf_trust), &vf_trust))
1314 goto nla_put_vf_failure;
1316 if (dev->netdev_ops->ndo_get_vf_guid &&
1317 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1319 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1321 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1323 goto nla_put_vf_failure;
1325 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1327 goto nla_put_vf_failure;
1328 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1330 nla_nest_cancel(skb, vfvlanlist);
1331 goto nla_put_vf_failure;
1333 nla_nest_end(skb, vfvlanlist);
1334 memset(&vf_stats, 0, sizeof(vf_stats));
1335 if (dev->netdev_ops->ndo_get_vf_stats)
1336 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1338 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1340 goto nla_put_vf_failure;
1341 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1342 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1343 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1344 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1345 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1346 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1347 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1348 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1349 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1350 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1351 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1352 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1353 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1354 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1355 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1356 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1357 nla_nest_cancel(skb, vfstats);
1358 goto nla_put_vf_failure;
1360 nla_nest_end(skb, vfstats);
1361 nla_nest_end(skb, vf);
1365 nla_nest_cancel(skb, vf);
1366 nla_put_vfinfo_failure:
1367 nla_nest_cancel(skb, vfinfo);
1371 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1372 struct net_device *dev,
1373 u32 ext_filter_mask)
1375 struct nlattr *vfinfo;
1378 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1381 num_vfs = dev_num_vf(dev->dev.parent);
1382 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1385 if (!dev->netdev_ops->ndo_get_vf_config)
1388 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1392 for (i = 0; i < num_vfs; i++) {
1393 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1397 nla_nest_end(skb, vfinfo);
1401 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1403 struct rtnl_link_ifmap map;
1405 memset(&map, 0, sizeof(map));
1406 map.mem_start = dev->mem_start;
1407 map.mem_end = dev->mem_end;
1408 map.base_addr = dev->base_addr;
1411 map.port = dev->if_port;
1413 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1419 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1421 const struct bpf_prog *generic_xdp_prog;
1425 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1426 if (!generic_xdp_prog)
1428 return generic_xdp_prog->aux->id;
1431 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1433 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1436 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1438 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1441 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1442 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1443 u32 (*get_prog_id)(struct net_device *dev))
1448 curr_id = get_prog_id(dev);
1453 err = nla_put_u32(skb, attr, curr_id);
1457 if (*mode != XDP_ATTACHED_NONE)
1458 *mode = XDP_ATTACHED_MULTI;
1465 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1472 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1477 mode = XDP_ATTACHED_NONE;
1478 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1479 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1482 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1483 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1486 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1487 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1491 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1495 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1496 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1501 nla_nest_end(skb, xdp);
1505 nla_nest_cancel(skb, xdp);
1509 static u32 rtnl_get_event(unsigned long event)
1511 u32 rtnl_event_type = IFLA_EVENT_NONE;
1515 rtnl_event_type = IFLA_EVENT_REBOOT;
1517 case NETDEV_FEAT_CHANGE:
1518 rtnl_event_type = IFLA_EVENT_FEATURES;
1520 case NETDEV_BONDING_FAILOVER:
1521 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1523 case NETDEV_NOTIFY_PEERS:
1524 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1526 case NETDEV_RESEND_IGMP:
1527 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1529 case NETDEV_CHANGEINFODATA:
1530 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1536 return rtnl_event_type;
1539 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1541 const struct net_device *upper_dev;
1546 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1548 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1554 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1557 int ifindex = dev_get_iflink(dev);
1559 if (force || dev->ifindex != ifindex)
1560 return nla_put_u32(skb, IFLA_LINK, ifindex);
1565 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1566 struct net_device *dev)
1571 ret = dev_get_alias(dev, buf, sizeof(buf));
1572 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1575 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1576 const struct net_device *dev,
1577 struct net *src_net, gfp_t gfp)
1579 bool put_iflink = false;
1581 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1582 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1584 if (!net_eq(dev_net(dev), link_net)) {
1585 int id = peernet2id_alloc(src_net, link_net, gfp);
1587 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1594 return nla_put_iflink(skb, dev, put_iflink);
1597 static int rtnl_fill_link_af(struct sk_buff *skb,
1598 const struct net_device *dev,
1599 u32 ext_filter_mask)
1601 const struct rtnl_af_ops *af_ops;
1602 struct nlattr *af_spec;
1604 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1608 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1612 if (!af_ops->fill_link_af)
1615 af = nla_nest_start_noflag(skb, af_ops->family);
1619 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1621 * Caller may return ENODATA to indicate that there
1622 * was no data to be dumped. This is not an error, it
1623 * means we should trim the attribute header and
1626 if (err == -ENODATA)
1627 nla_nest_cancel(skb, af);
1631 nla_nest_end(skb, af);
1634 nla_nest_end(skb, af_spec);
1638 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1639 const struct net_device *dev)
1641 struct netdev_name_node *name_node;
1644 list_for_each_entry(name_node, &dev->name_node->list, list) {
1645 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1652 static int rtnl_fill_prop_list(struct sk_buff *skb,
1653 const struct net_device *dev)
1655 struct nlattr *prop_list;
1658 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1662 ret = rtnl_fill_alt_ifnames(skb, dev);
1666 nla_nest_end(skb, prop_list);
1670 nla_nest_cancel(skb, prop_list);
1674 static int rtnl_fill_proto_down(struct sk_buff *skb,
1675 const struct net_device *dev)
1680 if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1681 goto nla_put_failure;
1683 preason = dev->proto_down_reason;
1687 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1691 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1692 nla_nest_cancel(skb, pr);
1693 goto nla_put_failure;
1696 nla_nest_end(skb, pr);
1703 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1704 struct net_device *dev, struct net *src_net,
1705 int type, u32 pid, u32 seq, u32 change,
1706 unsigned int flags, u32 ext_filter_mask,
1707 u32 event, int *new_nsid, int new_ifindex,
1708 int tgt_netnsid, gfp_t gfp)
1710 struct ifinfomsg *ifm;
1711 struct nlmsghdr *nlh;
1714 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1718 ifm = nlmsg_data(nlh);
1719 ifm->ifi_family = AF_UNSPEC;
1721 ifm->ifi_type = dev->type;
1722 ifm->ifi_index = dev->ifindex;
1723 ifm->ifi_flags = dev_get_flags(dev);
1724 ifm->ifi_change = change;
1726 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1727 goto nla_put_failure;
1729 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1730 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1731 nla_put_u8(skb, IFLA_OPERSTATE,
1732 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1733 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1734 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1735 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1736 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1737 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1738 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1739 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1740 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1741 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1743 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1745 put_master_ifindex(skb, dev) ||
1746 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1748 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1749 nla_put_ifalias(skb, dev) ||
1750 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1751 atomic_read(&dev->carrier_up_count) +
1752 atomic_read(&dev->carrier_down_count)) ||
1753 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1754 atomic_read(&dev->carrier_up_count)) ||
1755 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1756 atomic_read(&dev->carrier_down_count)))
1757 goto nla_put_failure;
1759 if (rtnl_fill_proto_down(skb, dev))
1760 goto nla_put_failure;
1762 if (event != IFLA_EVENT_NONE) {
1763 if (nla_put_u32(skb, IFLA_EVENT, event))
1764 goto nla_put_failure;
1767 if (rtnl_fill_link_ifmap(skb, dev))
1768 goto nla_put_failure;
1770 if (dev->addr_len) {
1771 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1772 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1773 goto nla_put_failure;
1776 if (rtnl_phys_port_id_fill(skb, dev))
1777 goto nla_put_failure;
1779 if (rtnl_phys_port_name_fill(skb, dev))
1780 goto nla_put_failure;
1782 if (rtnl_phys_switch_id_fill(skb, dev))
1783 goto nla_put_failure;
1785 if (rtnl_fill_stats(skb, dev))
1786 goto nla_put_failure;
1788 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1789 goto nla_put_failure;
1791 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1792 goto nla_put_failure;
1794 if (rtnl_xdp_fill(skb, dev))
1795 goto nla_put_failure;
1797 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1798 if (rtnl_link_fill(skb, dev) < 0)
1799 goto nla_put_failure;
1802 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1803 goto nla_put_failure;
1806 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1807 goto nla_put_failure;
1809 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1810 goto nla_put_failure;
1812 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1813 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1814 goto nla_put_failure;
1817 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1818 goto nla_put_failure_rcu;
1821 if (rtnl_fill_prop_list(skb, dev))
1822 goto nla_put_failure;
1824 if (dev->dev.parent &&
1825 nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1826 dev_name(dev->dev.parent)))
1827 goto nla_put_failure;
1829 if (dev->dev.parent && dev->dev.parent->bus &&
1830 nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1831 dev->dev.parent->bus->name))
1832 goto nla_put_failure;
1834 nlmsg_end(skb, nlh);
1837 nla_put_failure_rcu:
1840 nlmsg_cancel(skb, nlh);
1844 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1845 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1846 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1847 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1848 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1849 [IFLA_MTU] = { .type = NLA_U32 },
1850 [IFLA_LINK] = { .type = NLA_U32 },
1851 [IFLA_MASTER] = { .type = NLA_U32 },
1852 [IFLA_CARRIER] = { .type = NLA_U8 },
1853 [IFLA_TXQLEN] = { .type = NLA_U32 },
1854 [IFLA_WEIGHT] = { .type = NLA_U32 },
1855 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1856 [IFLA_LINKMODE] = { .type = NLA_U8 },
1857 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1858 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1859 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1860 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1861 * allow 0-length string (needed to remove an alias).
1863 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1864 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1865 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1866 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1867 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1868 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1869 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1870 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1871 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1872 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1873 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1874 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1875 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1876 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1877 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1878 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1879 [IFLA_XDP] = { .type = NLA_NESTED },
1880 [IFLA_EVENT] = { .type = NLA_U32 },
1881 [IFLA_GROUP] = { .type = NLA_U32 },
1882 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
1883 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1884 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1885 [IFLA_MIN_MTU] = { .type = NLA_U32 },
1886 [IFLA_MAX_MTU] = { .type = NLA_U32 },
1887 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
1888 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
1889 .len = ALTIFNAMSIZ - 1 },
1890 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
1891 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1892 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
1893 [IFLA_PARENT_DEV_NAME] = { .type = NLA_NUL_STRING },
1896 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1897 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1898 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1899 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1900 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1903 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1904 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1905 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1906 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1907 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1908 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1909 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1910 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1911 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1912 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1913 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1914 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1915 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1916 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1919 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1920 [IFLA_PORT_VF] = { .type = NLA_U32 },
1921 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1922 .len = PORT_PROFILE_MAX },
1923 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1924 .len = PORT_UUID_MAX },
1925 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1926 .len = PORT_UUID_MAX },
1927 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1928 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1930 /* Unused, but we need to keep it here since user space could
1931 * fill it. It's also broken with regard to NLA_BINARY use in
1932 * combination with structs.
1934 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1935 .len = sizeof(struct ifla_port_vsi) },
1938 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1939 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1940 [IFLA_XDP_FD] = { .type = NLA_S32 },
1941 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
1942 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1943 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1944 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1947 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1949 const struct rtnl_link_ops *ops = NULL;
1950 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1952 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1955 if (linfo[IFLA_INFO_KIND]) {
1956 char kind[MODULE_NAME_LEN];
1958 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1959 ops = rtnl_link_ops_get(kind);
1965 static bool link_master_filtered(struct net_device *dev, int master_idx)
1967 struct net_device *master;
1972 master = netdev_master_upper_dev_get(dev);
1973 if (!master || master->ifindex != master_idx)
1979 static bool link_kind_filtered(const struct net_device *dev,
1980 const struct rtnl_link_ops *kind_ops)
1982 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1988 static bool link_dump_filtered(struct net_device *dev,
1990 const struct rtnl_link_ops *kind_ops)
1992 if (link_master_filtered(dev, master_idx) ||
1993 link_kind_filtered(dev, kind_ops))
2000 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2001 * @sk: netlink socket
2002 * @netnsid: network namespace identifier
2004 * Returns the network namespace identified by netnsid on success or an error
2005 * pointer on failure.
2007 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2011 net = get_net_ns_by_id(sock_net(sk), netnsid);
2013 return ERR_PTR(-EINVAL);
2015 /* For now, the caller is required to have CAP_NET_ADMIN in
2016 * the user namespace owning the target net ns.
2018 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2020 return ERR_PTR(-EACCES);
2024 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2026 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2027 bool strict_check, struct nlattr **tb,
2028 struct netlink_ext_ack *extack)
2033 struct ifinfomsg *ifm;
2035 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2036 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2040 ifm = nlmsg_data(nlh);
2041 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2043 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2046 if (ifm->ifi_index) {
2047 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2051 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2052 IFLA_MAX, ifla_policy,
2056 /* A hack to preserve kernel<->userspace interface.
2057 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2058 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2059 * what iproute2 < v3.9.0 used.
2060 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2061 * attribute, its netlink message is shorter than struct ifinfomsg.
2063 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2064 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2066 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2070 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2072 struct netlink_ext_ack *extack = cb->extack;
2073 const struct nlmsghdr *nlh = cb->nlh;
2074 struct net *net = sock_net(skb->sk);
2075 struct net *tgt_net = net;
2078 struct net_device *dev;
2079 struct hlist_head *head;
2080 struct nlattr *tb[IFLA_MAX+1];
2081 u32 ext_filter_mask = 0;
2082 const struct rtnl_link_ops *kind_ops = NULL;
2083 unsigned int flags = NLM_F_MULTI;
2089 s_idx = cb->args[1];
2091 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2093 if (cb->strict_check)
2099 for (i = 0; i <= IFLA_MAX; ++i) {
2103 /* new attributes should only be added with strict checking */
2105 case IFLA_TARGET_NETNSID:
2106 netnsid = nla_get_s32(tb[i]);
2107 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2108 if (IS_ERR(tgt_net)) {
2109 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2110 return PTR_ERR(tgt_net);
2114 ext_filter_mask = nla_get_u32(tb[i]);
2117 master_idx = nla_get_u32(tb[i]);
2120 kind_ops = linkinfo_to_kind_ops(tb[i]);
2123 if (cb->strict_check) {
2124 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2130 if (master_idx || kind_ops)
2131 flags |= NLM_F_DUMP_FILTERED;
2134 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2136 head = &tgt_net->dev_index_head[h];
2137 hlist_for_each_entry(dev, head, index_hlist) {
2138 if (link_dump_filtered(dev, master_idx, kind_ops))
2142 err = rtnl_fill_ifinfo(skb, dev, net,
2144 NETLINK_CB(cb->skb).portid,
2145 nlh->nlmsg_seq, 0, flags,
2146 ext_filter_mask, 0, NULL, 0,
2147 netnsid, GFP_KERNEL);
2150 if (likely(skb->len))
2164 cb->seq = tgt_net->dev_base_seq;
2165 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2172 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2173 struct netlink_ext_ack *exterr)
2175 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2178 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2180 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2183 /* Examine the link attributes and figure out which
2184 * network namespace we are talking about.
2186 if (tb[IFLA_NET_NS_PID])
2187 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2188 else if (tb[IFLA_NET_NS_FD])
2189 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2191 net = get_net(src_net);
2194 EXPORT_SYMBOL(rtnl_link_get_net);
2196 /* Figure out which network namespace we are talking about by
2197 * examining the link attributes in the following order:
2199 * 1. IFLA_NET_NS_PID
2201 * 3. IFLA_TARGET_NETNSID
2203 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2204 struct nlattr *tb[])
2208 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2209 return rtnl_link_get_net(src_net, tb);
2211 if (!tb[IFLA_TARGET_NETNSID])
2212 return get_net(src_net);
2214 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2216 return ERR_PTR(-EINVAL);
2221 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2222 struct net *src_net,
2223 struct nlattr *tb[], int cap)
2227 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2231 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2233 return ERR_PTR(-EPERM);
2239 /* Verify that rtnetlink requests do not pass additional properties
2240 * potentially referring to different network namespaces.
2242 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2243 struct netlink_ext_ack *extack,
2247 if (netns_id_only) {
2248 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2251 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2255 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2258 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2261 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2267 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2271 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2274 if (tb[IFLA_ADDRESS] &&
2275 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2278 if (tb[IFLA_BROADCAST] &&
2279 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2283 if (tb[IFLA_AF_SPEC]) {
2287 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2288 const struct rtnl_af_ops *af_ops;
2290 af_ops = rtnl_af_lookup(nla_type(af));
2292 return -EAFNOSUPPORT;
2294 if (!af_ops->set_link_af)
2297 if (af_ops->validate_link_af) {
2298 err = af_ops->validate_link_af(dev, af);
2308 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2311 const struct net_device_ops *ops = dev->netdev_ops;
2313 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2316 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2318 if (dev->type != ARPHRD_INFINIBAND)
2321 return handle_infiniband_guid(dev, ivt, guid_type);
2324 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2326 const struct net_device_ops *ops = dev->netdev_ops;
2329 if (tb[IFLA_VF_MAC]) {
2330 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2332 if (ivm->vf >= INT_MAX)
2335 if (ops->ndo_set_vf_mac)
2336 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2342 if (tb[IFLA_VF_VLAN]) {
2343 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2345 if (ivv->vf >= INT_MAX)
2348 if (ops->ndo_set_vf_vlan)
2349 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2351 htons(ETH_P_8021Q));
2356 if (tb[IFLA_VF_VLAN_LIST]) {
2357 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2358 struct nlattr *attr;
2362 if (!ops->ndo_set_vf_vlan)
2365 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2366 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2367 nla_len(attr) < NLA_HDRLEN) {
2370 if (len >= MAX_VLAN_LIST_LEN)
2372 ivvl[len] = nla_data(attr);
2379 if (ivvl[0]->vf >= INT_MAX)
2381 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2382 ivvl[0]->qos, ivvl[0]->vlan_proto);
2387 if (tb[IFLA_VF_TX_RATE]) {
2388 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2389 struct ifla_vf_info ivf;
2391 if (ivt->vf >= INT_MAX)
2394 if (ops->ndo_get_vf_config)
2395 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2400 if (ops->ndo_set_vf_rate)
2401 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2408 if (tb[IFLA_VF_RATE]) {
2409 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2411 if (ivt->vf >= INT_MAX)
2414 if (ops->ndo_set_vf_rate)
2415 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2422 if (tb[IFLA_VF_SPOOFCHK]) {
2423 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2425 if (ivs->vf >= INT_MAX)
2428 if (ops->ndo_set_vf_spoofchk)
2429 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2435 if (tb[IFLA_VF_LINK_STATE]) {
2436 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2438 if (ivl->vf >= INT_MAX)
2441 if (ops->ndo_set_vf_link_state)
2442 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2448 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2449 struct ifla_vf_rss_query_en *ivrssq_en;
2452 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2453 if (ivrssq_en->vf >= INT_MAX)
2455 if (ops->ndo_set_vf_rss_query_en)
2456 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2457 ivrssq_en->setting);
2462 if (tb[IFLA_VF_TRUST]) {
2463 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2465 if (ivt->vf >= INT_MAX)
2468 if (ops->ndo_set_vf_trust)
2469 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2474 if (tb[IFLA_VF_IB_NODE_GUID]) {
2475 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2477 if (ivt->vf >= INT_MAX)
2479 if (!ops->ndo_set_vf_guid)
2481 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2484 if (tb[IFLA_VF_IB_PORT_GUID]) {
2485 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2487 if (ivt->vf >= INT_MAX)
2489 if (!ops->ndo_set_vf_guid)
2492 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2498 static int do_set_master(struct net_device *dev, int ifindex,
2499 struct netlink_ext_ack *extack)
2501 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2502 const struct net_device_ops *ops;
2506 if (upper_dev->ifindex == ifindex)
2508 ops = upper_dev->netdev_ops;
2509 if (ops->ndo_del_slave) {
2510 err = ops->ndo_del_slave(upper_dev, dev);
2519 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2522 ops = upper_dev->netdev_ops;
2523 if (ops->ndo_add_slave) {
2524 err = ops->ndo_add_slave(upper_dev, dev, extack);
2534 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2535 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2536 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2539 static int do_set_proto_down(struct net_device *dev,
2540 struct nlattr *nl_proto_down,
2541 struct nlattr *nl_proto_down_reason,
2542 struct netlink_ext_ack *extack)
2544 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2545 const struct net_device_ops *ops = dev->netdev_ops;
2546 unsigned long mask = 0;
2551 if (!ops->ndo_change_proto_down) {
2552 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2556 if (nl_proto_down_reason) {
2557 err = nla_parse_nested_deprecated(pdreason,
2558 IFLA_PROTO_DOWN_REASON_MAX,
2559 nl_proto_down_reason,
2560 ifla_proto_down_reason_policy,
2565 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2566 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2570 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2572 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2573 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2575 dev_change_proto_down_reason(dev, mask, value);
2578 if (nl_proto_down) {
2579 proto_down = nla_get_u8(nl_proto_down);
2581 /* Don't turn off protodown if there are active reasons */
2582 if (!proto_down && dev->proto_down_reason) {
2583 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2586 err = dev_change_proto_down(dev,
2595 #define DO_SETLINK_MODIFIED 0x01
2596 /* notify flag means notify + modified. */
2597 #define DO_SETLINK_NOTIFY 0x03
2598 static int do_setlink(const struct sk_buff *skb,
2599 struct net_device *dev, struct ifinfomsg *ifm,
2600 struct netlink_ext_ack *extack,
2601 struct nlattr **tb, char *ifname, int status)
2603 const struct net_device_ops *ops = dev->netdev_ops;
2606 err = validate_linkmsg(dev, tb);
2610 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2614 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2621 if (tb[IFLA_NEW_IFINDEX])
2622 new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2626 err = __dev_change_net_namespace(dev, net, ifname, new_ifindex);
2630 status |= DO_SETLINK_MODIFIED;
2634 struct rtnl_link_ifmap *u_map;
2637 if (!ops->ndo_set_config) {
2642 if (!netif_device_present(dev)) {
2647 u_map = nla_data(tb[IFLA_MAP]);
2648 k_map.mem_start = (unsigned long) u_map->mem_start;
2649 k_map.mem_end = (unsigned long) u_map->mem_end;
2650 k_map.base_addr = (unsigned short) u_map->base_addr;
2651 k_map.irq = (unsigned char) u_map->irq;
2652 k_map.dma = (unsigned char) u_map->dma;
2653 k_map.port = (unsigned char) u_map->port;
2655 err = ops->ndo_set_config(dev, &k_map);
2659 status |= DO_SETLINK_NOTIFY;
2662 if (tb[IFLA_ADDRESS]) {
2663 struct sockaddr *sa;
2666 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2668 sa = kmalloc(len, GFP_KERNEL);
2673 sa->sa_family = dev->type;
2674 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2676 err = dev_set_mac_address_user(dev, sa, extack);
2680 status |= DO_SETLINK_MODIFIED;
2684 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2687 status |= DO_SETLINK_MODIFIED;
2690 if (tb[IFLA_GROUP]) {
2691 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2692 status |= DO_SETLINK_NOTIFY;
2696 * Interface selected by interface index but interface
2697 * name provided implies that a name change has been
2700 if (ifm->ifi_index > 0 && ifname[0]) {
2701 err = dev_change_name(dev, ifname);
2704 status |= DO_SETLINK_MODIFIED;
2707 if (tb[IFLA_IFALIAS]) {
2708 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2709 nla_len(tb[IFLA_IFALIAS]));
2712 status |= DO_SETLINK_NOTIFY;
2715 if (tb[IFLA_BROADCAST]) {
2716 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2717 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2720 if (ifm->ifi_flags || ifm->ifi_change) {
2721 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2727 if (tb[IFLA_MASTER]) {
2728 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2731 status |= DO_SETLINK_MODIFIED;
2734 if (tb[IFLA_CARRIER]) {
2735 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2738 status |= DO_SETLINK_MODIFIED;
2741 if (tb[IFLA_TXQLEN]) {
2742 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2744 err = dev_change_tx_queue_len(dev, value);
2747 status |= DO_SETLINK_MODIFIED;
2750 if (tb[IFLA_GSO_MAX_SIZE]) {
2751 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2753 if (max_size > GSO_MAX_SIZE) {
2758 if (dev->gso_max_size ^ max_size) {
2759 netif_set_gso_max_size(dev, max_size);
2760 status |= DO_SETLINK_MODIFIED;
2764 if (tb[IFLA_GSO_MAX_SEGS]) {
2765 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2767 if (max_segs > GSO_MAX_SEGS) {
2772 if (dev->gso_max_segs ^ max_segs) {
2773 dev->gso_max_segs = max_segs;
2774 status |= DO_SETLINK_MODIFIED;
2778 if (tb[IFLA_OPERSTATE])
2779 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2781 if (tb[IFLA_LINKMODE]) {
2782 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2784 write_lock_bh(&dev_base_lock);
2785 if (dev->link_mode ^ value)
2786 status |= DO_SETLINK_NOTIFY;
2787 dev->link_mode = value;
2788 write_unlock_bh(&dev_base_lock);
2791 if (tb[IFLA_VFINFO_LIST]) {
2792 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2793 struct nlattr *attr;
2796 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2797 if (nla_type(attr) != IFLA_VF_INFO ||
2798 nla_len(attr) < NLA_HDRLEN) {
2802 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2808 err = do_setvfinfo(dev, vfinfo);
2811 status |= DO_SETLINK_NOTIFY;
2816 if (tb[IFLA_VF_PORTS]) {
2817 struct nlattr *port[IFLA_PORT_MAX+1];
2818 struct nlattr *attr;
2823 if (!ops->ndo_set_vf_port)
2826 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2827 if (nla_type(attr) != IFLA_VF_PORT ||
2828 nla_len(attr) < NLA_HDRLEN) {
2832 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2838 if (!port[IFLA_PORT_VF]) {
2842 vf = nla_get_u32(port[IFLA_PORT_VF]);
2843 err = ops->ndo_set_vf_port(dev, vf, port);
2846 status |= DO_SETLINK_NOTIFY;
2851 if (tb[IFLA_PORT_SELF]) {
2852 struct nlattr *port[IFLA_PORT_MAX+1];
2854 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2856 ifla_port_policy, NULL);
2861 if (ops->ndo_set_vf_port)
2862 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2865 status |= DO_SETLINK_NOTIFY;
2868 if (tb[IFLA_AF_SPEC]) {
2872 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2873 const struct rtnl_af_ops *af_ops;
2875 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2877 err = af_ops->set_link_af(dev, af, extack);
2881 status |= DO_SETLINK_NOTIFY;
2886 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2887 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2888 tb[IFLA_PROTO_DOWN_REASON], extack);
2891 status |= DO_SETLINK_NOTIFY;
2895 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2898 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2900 ifla_xdp_policy, NULL);
2904 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2909 if (xdp[IFLA_XDP_FLAGS]) {
2910 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2911 if (xdp_flags & ~XDP_FLAGS_MASK) {
2915 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2921 if (xdp[IFLA_XDP_FD]) {
2922 int expected_fd = -1;
2924 if (xdp_flags & XDP_FLAGS_REPLACE) {
2925 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2930 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2933 err = dev_change_xdp_fd(dev, extack,
2934 nla_get_s32(xdp[IFLA_XDP_FD]),
2939 status |= DO_SETLINK_NOTIFY;
2944 if (status & DO_SETLINK_MODIFIED) {
2945 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2946 netdev_state_change(dev);
2949 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",
2956 static struct net_device *rtnl_dev_get(struct net *net,
2957 struct nlattr *ifname_attr,
2958 struct nlattr *altifname_attr,
2961 char buffer[ALTIFNAMSIZ];
2966 nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2967 else if (altifname_attr)
2968 nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2973 return __dev_get_by_name(net, ifname);
2976 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2977 struct netlink_ext_ack *extack)
2979 struct net *net = sock_net(skb->sk);
2980 struct ifinfomsg *ifm;
2981 struct net_device *dev;
2983 struct nlattr *tb[IFLA_MAX+1];
2984 char ifname[IFNAMSIZ];
2986 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2987 ifla_policy, extack);
2991 err = rtnl_ensure_unique_netns(tb, extack, false);
2995 if (tb[IFLA_IFNAME])
2996 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3001 ifm = nlmsg_data(nlh);
3002 if (ifm->ifi_index > 0)
3003 dev = __dev_get_by_index(net, ifm->ifi_index);
3004 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3005 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3014 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3019 static int rtnl_group_dellink(const struct net *net, int group)
3021 struct net_device *dev, *aux;
3022 LIST_HEAD(list_kill);
3028 for_each_netdev(net, dev) {
3029 if (dev->group == group) {
3030 const struct rtnl_link_ops *ops;
3033 ops = dev->rtnl_link_ops;
3034 if (!ops || !ops->dellink)
3042 for_each_netdev_safe(net, dev, aux) {
3043 if (dev->group == group) {
3044 const struct rtnl_link_ops *ops;
3046 ops = dev->rtnl_link_ops;
3047 ops->dellink(dev, &list_kill);
3050 unregister_netdevice_many(&list_kill);
3055 int rtnl_delete_link(struct net_device *dev)
3057 const struct rtnl_link_ops *ops;
3058 LIST_HEAD(list_kill);
3060 ops = dev->rtnl_link_ops;
3061 if (!ops || !ops->dellink)
3064 ops->dellink(dev, &list_kill);
3065 unregister_netdevice_many(&list_kill);
3069 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3071 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3072 struct netlink_ext_ack *extack)
3074 struct net *net = sock_net(skb->sk);
3075 struct net *tgt_net = net;
3076 struct net_device *dev = NULL;
3077 struct ifinfomsg *ifm;
3078 struct nlattr *tb[IFLA_MAX+1];
3082 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3083 ifla_policy, extack);
3087 err = rtnl_ensure_unique_netns(tb, extack, true);
3091 if (tb[IFLA_TARGET_NETNSID]) {
3092 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3093 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3094 if (IS_ERR(tgt_net))
3095 return PTR_ERR(tgt_net);
3099 ifm = nlmsg_data(nlh);
3100 if (ifm->ifi_index > 0)
3101 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3102 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3103 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3104 tb[IFLA_ALT_IFNAME], NULL);
3105 else if (tb[IFLA_GROUP])
3106 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3111 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3117 err = rtnl_delete_link(dev);
3126 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3128 unsigned int old_flags;
3131 old_flags = dev->flags;
3132 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3133 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3139 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3140 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3142 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3143 __dev_notify_flags(dev, old_flags, ~0U);
3147 EXPORT_SYMBOL(rtnl_configure_link);
3149 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3150 unsigned char name_assign_type,
3151 const struct rtnl_link_ops *ops,
3152 struct nlattr *tb[],
3153 struct netlink_ext_ack *extack)
3155 struct net_device *dev;
3156 unsigned int num_tx_queues = 1;
3157 unsigned int num_rx_queues = 1;
3159 if (tb[IFLA_NUM_TX_QUEUES])
3160 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3161 else if (ops->get_num_tx_queues)
3162 num_tx_queues = ops->get_num_tx_queues();
3164 if (tb[IFLA_NUM_RX_QUEUES])
3165 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3166 else if (ops->get_num_rx_queues)
3167 num_rx_queues = ops->get_num_rx_queues();
3169 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3170 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3171 return ERR_PTR(-EINVAL);
3174 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3175 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3176 return ERR_PTR(-EINVAL);
3180 dev = ops->alloc(tb, ifname, name_assign_type,
3181 num_tx_queues, num_rx_queues);
3185 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3186 name_assign_type, ops->setup,
3187 num_tx_queues, num_rx_queues);
3191 return ERR_PTR(-ENOMEM);
3193 dev_net_set(dev, net);
3194 dev->rtnl_link_ops = ops;
3195 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3198 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3201 err = dev_validate_mtu(dev, mtu, extack);
3204 return ERR_PTR(err);
3208 if (tb[IFLA_ADDRESS]) {
3209 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3210 nla_len(tb[IFLA_ADDRESS]));
3211 dev->addr_assign_type = NET_ADDR_SET;
3213 if (tb[IFLA_BROADCAST])
3214 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3215 nla_len(tb[IFLA_BROADCAST]));
3216 if (tb[IFLA_TXQLEN])
3217 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3218 if (tb[IFLA_OPERSTATE])
3219 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3220 if (tb[IFLA_LINKMODE])
3221 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3223 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3224 if (tb[IFLA_GSO_MAX_SIZE])
3225 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3226 if (tb[IFLA_GSO_MAX_SEGS])
3227 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3231 EXPORT_SYMBOL(rtnl_create_link);
3233 static int rtnl_group_changelink(const struct sk_buff *skb,
3234 struct net *net, int group,
3235 struct ifinfomsg *ifm,
3236 struct netlink_ext_ack *extack,
3239 struct net_device *dev, *aux;
3242 for_each_netdev_safe(net, dev, aux) {
3243 if (dev->group == group) {
3244 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3253 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3254 struct nlattr **attr, struct netlink_ext_ack *extack)
3256 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3257 unsigned char name_assign_type = NET_NAME_USER;
3258 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3259 const struct rtnl_link_ops *m_ops = NULL;
3260 struct net_device *master_dev = NULL;
3261 struct net *net = sock_net(skb->sk);
3262 const struct rtnl_link_ops *ops;
3263 struct nlattr *tb[IFLA_MAX + 1];
3264 struct net *dest_net, *link_net;
3265 struct nlattr **slave_data;
3266 char kind[MODULE_NAME_LEN];
3267 struct net_device *dev;
3268 struct ifinfomsg *ifm;
3269 char ifname[IFNAMSIZ];
3270 struct nlattr **data;
3273 #ifdef CONFIG_MODULES
3276 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3277 ifla_policy, extack);
3281 err = rtnl_ensure_unique_netns(tb, extack, false);
3285 if (tb[IFLA_IFNAME])
3286 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3290 ifm = nlmsg_data(nlh);
3291 if (ifm->ifi_index > 0)
3292 dev = __dev_get_by_index(net, ifm->ifi_index);
3293 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3294 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3299 master_dev = netdev_master_upper_dev_get(dev);
3301 m_ops = master_dev->rtnl_link_ops;
3304 err = validate_linkmsg(dev, tb);
3308 if (tb[IFLA_LINKINFO]) {
3309 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3311 ifla_info_policy, NULL);
3315 memset(linkinfo, 0, sizeof(linkinfo));
3317 if (linkinfo[IFLA_INFO_KIND]) {
3318 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3319 ops = rtnl_link_ops_get(kind);
3327 if (ops->maxtype > RTNL_MAX_TYPE)
3330 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3331 err = nla_parse_nested_deprecated(attr, ops->maxtype,
3332 linkinfo[IFLA_INFO_DATA],
3333 ops->policy, extack);
3338 if (ops->validate) {
3339 err = ops->validate(tb, data, extack);
3347 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3350 if (m_ops->slave_maxtype &&
3351 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3352 err = nla_parse_nested_deprecated(slave_attr,
3353 m_ops->slave_maxtype,
3354 linkinfo[IFLA_INFO_SLAVE_DATA],
3355 m_ops->slave_policy,
3359 slave_data = slave_attr;
3366 if (nlh->nlmsg_flags & NLM_F_EXCL)
3368 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3371 if (linkinfo[IFLA_INFO_DATA]) {
3372 if (!ops || ops != dev->rtnl_link_ops ||
3376 err = ops->changelink(dev, tb, data, extack);
3379 status |= DO_SETLINK_NOTIFY;
3382 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3383 if (!m_ops || !m_ops->slave_changelink)
3386 err = m_ops->slave_changelink(master_dev, dev, tb,
3387 slave_data, extack);
3390 status |= DO_SETLINK_NOTIFY;
3393 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3396 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3397 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3398 return rtnl_group_changelink(skb, net,
3399 nla_get_u32(tb[IFLA_GROUP]),
3404 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3408 #ifdef CONFIG_MODULES
3411 request_module("rtnl-link-%s", kind);
3413 ops = rtnl_link_ops_get(kind);
3418 NL_SET_ERR_MSG(extack, "Unknown device type");
3422 if (!ops->alloc && !ops->setup)
3426 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3427 name_assign_type = NET_NAME_ENUM;
3430 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3431 if (IS_ERR(dest_net))
3432 return PTR_ERR(dest_net);
3434 if (tb[IFLA_LINK_NETNSID]) {
3435 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3437 link_net = get_net_ns_by_id(dest_net, id);
3439 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3444 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3450 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3451 name_assign_type, ops, tb, extack);
3457 dev->ifindex = ifm->ifi_index;
3460 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3462 err = register_netdevice(dev);
3468 err = rtnl_configure_link(dev, ifm);
3470 goto out_unregister;
3472 err = dev_change_net_namespace(dev, dest_net, ifname);
3474 goto out_unregister;
3476 if (tb[IFLA_MASTER]) {
3477 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3479 goto out_unregister;
3488 LIST_HEAD(list_kill);
3490 ops->dellink(dev, &list_kill);
3491 unregister_netdevice_many(&list_kill);
3493 unregister_netdevice(dev);
3498 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3499 struct netlink_ext_ack *extack)
3501 struct nlattr **attr;
3504 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3508 ret = __rtnl_newlink(skb, nlh, attr, extack);
3513 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3514 const struct nlmsghdr *nlh,
3516 struct netlink_ext_ack *extack)
3518 struct ifinfomsg *ifm;
3521 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3522 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3526 if (!netlink_strict_get_check(skb))
3527 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3528 ifla_policy, extack);
3530 ifm = nlmsg_data(nlh);
3531 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3533 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3537 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3538 ifla_policy, extack);
3542 for (i = 0; i <= IFLA_MAX; i++) {
3548 case IFLA_ALT_IFNAME:
3550 case IFLA_TARGET_NETNSID:
3553 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3561 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3562 struct netlink_ext_ack *extack)
3564 struct net *net = sock_net(skb->sk);
3565 struct net *tgt_net = net;
3566 struct ifinfomsg *ifm;
3567 struct nlattr *tb[IFLA_MAX+1];
3568 struct net_device *dev = NULL;
3569 struct sk_buff *nskb;
3572 u32 ext_filter_mask = 0;
3574 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3578 err = rtnl_ensure_unique_netns(tb, extack, true);
3582 if (tb[IFLA_TARGET_NETNSID]) {
3583 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3584 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3585 if (IS_ERR(tgt_net))
3586 return PTR_ERR(tgt_net);
3589 if (tb[IFLA_EXT_MASK])
3590 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3593 ifm = nlmsg_data(nlh);
3594 if (ifm->ifi_index > 0)
3595 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3596 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3597 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3598 tb[IFLA_ALT_IFNAME], NULL);
3607 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3611 err = rtnl_fill_ifinfo(nskb, dev, net,
3612 RTM_NEWLINK, NETLINK_CB(skb).portid,
3613 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3614 0, NULL, 0, netnsid, GFP_KERNEL);
3616 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3617 WARN_ON(err == -EMSGSIZE);
3620 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3628 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3629 bool *changed, struct netlink_ext_ack *extack)
3634 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3638 alt_ifname = nla_strdup(attr, GFP_KERNEL);
3642 if (cmd == RTM_NEWLINKPROP) {
3643 err = netdev_name_node_alt_create(dev, alt_ifname);
3646 } else if (cmd == RTM_DELLINKPROP) {
3647 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3659 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3660 struct netlink_ext_ack *extack)
3662 struct net *net = sock_net(skb->sk);
3663 struct nlattr *tb[IFLA_MAX + 1];
3664 struct net_device *dev;
3665 struct ifinfomsg *ifm;
3666 bool changed = false;
3667 struct nlattr *attr;
3670 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3674 err = rtnl_ensure_unique_netns(tb, extack, true);
3678 ifm = nlmsg_data(nlh);
3679 if (ifm->ifi_index > 0)
3680 dev = __dev_get_by_index(net, ifm->ifi_index);
3681 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3682 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3683 tb[IFLA_ALT_IFNAME], NULL);
3690 if (!tb[IFLA_PROP_LIST])
3693 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3694 switch (nla_type(attr)) {
3695 case IFLA_ALT_IFNAME:
3696 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3704 netdev_state_change(dev);
3708 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3709 struct netlink_ext_ack *extack)
3711 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3714 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3715 struct netlink_ext_ack *extack)
3717 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3720 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3722 struct net *net = sock_net(skb->sk);
3723 size_t min_ifinfo_dump_size = 0;
3724 struct nlattr *tb[IFLA_MAX+1];
3725 u32 ext_filter_mask = 0;
3726 struct net_device *dev;
3729 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3730 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3731 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3733 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3734 if (tb[IFLA_EXT_MASK])
3735 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3738 if (!ext_filter_mask)
3739 return NLMSG_GOODSIZE;
3741 * traverse the list of net devices and compute the minimum
3742 * buffer size based upon the filter mask.
3745 for_each_netdev_rcu(net, dev) {
3746 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3747 if_nlmsg_size(dev, ext_filter_mask));
3751 return nlmsg_total_size(min_ifinfo_dump_size);
3754 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3757 int s_idx = cb->family;
3758 int type = cb->nlh->nlmsg_type - RTM_BASE;
3764 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3765 struct rtnl_link __rcu **tab;
3766 struct rtnl_link *link;
3767 rtnl_dumpit_func dumpit;
3769 if (idx < s_idx || idx == PF_PACKET)
3772 if (type < 0 || type >= RTM_NR_MSGTYPES)
3775 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3779 link = rcu_dereference_rtnl(tab[type]);
3783 dumpit = link->dumpit;
3788 memset(&cb->args[0], 0, sizeof(cb->args));
3792 ret = dumpit(skb, cb);
3798 return skb->len ? : ret;
3801 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3802 unsigned int change,
3803 u32 event, gfp_t flags, int *new_nsid,
3806 struct net *net = dev_net(dev);
3807 struct sk_buff *skb;
3809 size_t if_info_size;
3811 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3815 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3816 type, 0, 0, change, 0, 0, event,
3817 new_nsid, new_ifindex, -1, flags);
3819 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3820 WARN_ON(err == -EMSGSIZE);
3827 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3831 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3833 struct net *net = dev_net(dev);
3835 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3838 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3839 unsigned int change, u32 event,
3840 gfp_t flags, int *new_nsid, int new_ifindex)
3842 struct sk_buff *skb;
3844 if (dev->reg_state != NETREG_REGISTERED)
3847 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3850 rtmsg_ifinfo_send(skb, dev, flags);
3853 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3856 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3860 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3861 gfp_t flags, int *new_nsid, int new_ifindex)
3863 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3864 new_nsid, new_ifindex);
3867 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3868 struct net_device *dev,
3869 u8 *addr, u16 vid, u32 pid, u32 seq,
3870 int type, unsigned int flags,
3871 int nlflags, u16 ndm_state)
3873 struct nlmsghdr *nlh;
3876 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3880 ndm = nlmsg_data(nlh);
3881 ndm->ndm_family = AF_BRIDGE;
3884 ndm->ndm_flags = flags;
3886 ndm->ndm_ifindex = dev->ifindex;
3887 ndm->ndm_state = ndm_state;
3889 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3890 goto nla_put_failure;
3892 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3893 goto nla_put_failure;
3895 nlmsg_end(skb, nlh);
3899 nlmsg_cancel(skb, nlh);
3903 static inline size_t rtnl_fdb_nlmsg_size(void)
3905 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3906 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3907 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3911 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3914 struct net *net = dev_net(dev);
3915 struct sk_buff *skb;
3918 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3922 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3923 0, 0, type, NTF_SELF, 0, ndm_state);
3929 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3932 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3936 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3938 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3939 struct nlattr *tb[],
3940 struct net_device *dev,
3941 const unsigned char *addr, u16 vid,
3946 /* If aging addresses are supported device will need to
3947 * implement its own handler for this.
3949 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3950 netdev_info(dev, "default FDB implementation only supports local addresses\n");
3955 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
3959 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3960 err = dev_uc_add_excl(dev, addr);
3961 else if (is_multicast_ether_addr(addr))
3962 err = dev_mc_add_excl(dev, addr);
3964 /* Only return duplicate errors if NLM_F_EXCL is set */
3965 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3970 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3972 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3973 struct netlink_ext_ack *extack)
3978 if (nla_len(vlan_attr) != sizeof(u16)) {
3979 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3983 vid = nla_get_u16(vlan_attr);
3985 if (!vid || vid >= VLAN_VID_MASK) {
3986 NL_SET_ERR_MSG(extack, "invalid vlan id");
3994 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3995 struct netlink_ext_ack *extack)
3997 struct net *net = sock_net(skb->sk);
3999 struct nlattr *tb[NDA_MAX+1];
4000 struct net_device *dev;
4005 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4010 ndm = nlmsg_data(nlh);
4011 if (ndm->ndm_ifindex == 0) {
4012 NL_SET_ERR_MSG(extack, "invalid ifindex");
4016 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4018 NL_SET_ERR_MSG(extack, "unknown ifindex");
4022 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4023 NL_SET_ERR_MSG(extack, "invalid address");
4027 if (dev->type != ARPHRD_ETHER) {
4028 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4032 addr = nla_data(tb[NDA_LLADDR]);
4034 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4040 /* Support fdb on master device the net/bridge default case */
4041 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4042 netif_is_bridge_port(dev)) {
4043 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4044 const struct net_device_ops *ops = br_dev->netdev_ops;
4046 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4047 nlh->nlmsg_flags, extack);
4051 ndm->ndm_flags &= ~NTF_MASTER;
4054 /* Embedded bridge, macvlan, and any other device support */
4055 if ((ndm->ndm_flags & NTF_SELF)) {
4056 if (dev->netdev_ops->ndo_fdb_add)
4057 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4062 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4066 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4068 ndm->ndm_flags &= ~NTF_SELF;
4076 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4078 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4079 struct nlattr *tb[],
4080 struct net_device *dev,
4081 const unsigned char *addr, u16 vid)
4085 /* If aging addresses are supported device will need to
4086 * implement its own handler for this.
4088 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4089 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4093 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4094 err = dev_uc_del(dev, addr);
4095 else if (is_multicast_ether_addr(addr))
4096 err = dev_mc_del(dev, addr);
4100 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4102 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4103 struct netlink_ext_ack *extack)
4105 struct net *net = sock_net(skb->sk);
4107 struct nlattr *tb[NDA_MAX+1];
4108 struct net_device *dev;
4113 if (!netlink_capable(skb, CAP_NET_ADMIN))
4116 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4121 ndm = nlmsg_data(nlh);
4122 if (ndm->ndm_ifindex == 0) {
4123 NL_SET_ERR_MSG(extack, "invalid ifindex");
4127 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4129 NL_SET_ERR_MSG(extack, "unknown ifindex");
4133 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4134 NL_SET_ERR_MSG(extack, "invalid address");
4138 if (dev->type != ARPHRD_ETHER) {
4139 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4143 addr = nla_data(tb[NDA_LLADDR]);
4145 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4151 /* Support fdb on master device the net/bridge default case */
4152 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4153 netif_is_bridge_port(dev)) {
4154 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4155 const struct net_device_ops *ops = br_dev->netdev_ops;
4157 if (ops->ndo_fdb_del)
4158 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4163 ndm->ndm_flags &= ~NTF_MASTER;
4166 /* Embedded bridge, macvlan, and any other device support */
4167 if (ndm->ndm_flags & NTF_SELF) {
4168 if (dev->netdev_ops->ndo_fdb_del)
4169 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4172 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4175 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4177 ndm->ndm_flags &= ~NTF_SELF;
4184 static int nlmsg_populate_fdb(struct sk_buff *skb,
4185 struct netlink_callback *cb,
4186 struct net_device *dev,
4188 struct netdev_hw_addr_list *list)
4190 struct netdev_hw_addr *ha;
4194 portid = NETLINK_CB(cb->skb).portid;
4195 seq = cb->nlh->nlmsg_seq;
4197 list_for_each_entry(ha, &list->list, list) {
4198 if (*idx < cb->args[2])
4201 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4203 RTM_NEWNEIGH, NTF_SELF,
4204 NLM_F_MULTI, NUD_PERMANENT);
4214 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4215 * @skb: socket buffer to store message in
4216 * @cb: netlink callback
4218 * @filter_dev: ignored
4219 * @idx: the number of FDB table entries dumped is added to *@idx
4221 * Default netdevice operation to dump the existing unicast address list.
4222 * Returns number of addresses from list put in skb.
4224 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4225 struct netlink_callback *cb,
4226 struct net_device *dev,
4227 struct net_device *filter_dev,
4232 if (dev->type != ARPHRD_ETHER)
4235 netif_addr_lock_bh(dev);
4236 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4239 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4241 netif_addr_unlock_bh(dev);
4244 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4246 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4247 int *br_idx, int *brport_idx,
4248 struct netlink_ext_ack *extack)
4250 struct nlattr *tb[NDA_MAX + 1];
4254 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4255 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4259 ndm = nlmsg_data(nlh);
4260 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4261 ndm->ndm_flags || ndm->ndm_type) {
4262 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4266 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4267 NDA_MAX, NULL, extack);
4271 *brport_idx = ndm->ndm_ifindex;
4272 for (i = 0; i <= NDA_MAX; ++i) {
4278 if (nla_len(tb[i]) != sizeof(u32)) {
4279 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4282 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4285 if (nla_len(tb[i]) != sizeof(u32)) {
4286 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4289 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4292 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4300 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4301 int *br_idx, int *brport_idx,
4302 struct netlink_ext_ack *extack)
4304 struct nlattr *tb[IFLA_MAX+1];
4307 /* A hack to preserve kernel<->userspace interface.
4308 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4309 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4310 * So, check for ndmsg with an optional u32 attribute (not used here).
4311 * Fortunately these sizes don't conflict with the size of ifinfomsg
4312 * with an optional attribute.
4314 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4315 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4316 nla_attr_size(sizeof(u32)))) {
4317 struct ifinfomsg *ifm;
4319 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4320 tb, IFLA_MAX, ifla_policy,
4324 } else if (err == 0) {
4325 if (tb[IFLA_MASTER])
4326 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4329 ifm = nlmsg_data(nlh);
4330 *brport_idx = ifm->ifi_index;
4335 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4337 struct net_device *dev;
4338 struct net_device *br_dev = NULL;
4339 const struct net_device_ops *ops = NULL;
4340 const struct net_device_ops *cops = NULL;
4341 struct net *net = sock_net(skb->sk);
4342 struct hlist_head *head;
4350 if (cb->strict_check)
4351 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4354 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4360 br_dev = __dev_get_by_index(net, br_idx);
4364 ops = br_dev->netdev_ops;
4368 s_idx = cb->args[1];
4370 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4372 head = &net->dev_index_head[h];
4373 hlist_for_each_entry(dev, head, index_hlist) {
4375 if (brport_idx && (dev->ifindex != brport_idx))
4378 if (!br_idx) { /* user did not specify a specific bridge */
4379 if (netif_is_bridge_port(dev)) {
4380 br_dev = netdev_master_upper_dev_get(dev);
4381 cops = br_dev->netdev_ops;
4384 if (dev != br_dev &&
4385 !netif_is_bridge_port(dev))
4388 if (br_dev != netdev_master_upper_dev_get(dev) &&
4389 !(dev->priv_flags & IFF_EBRIDGE))
4397 if (netif_is_bridge_port(dev)) {
4398 if (cops && cops->ndo_fdb_dump) {
4399 err = cops->ndo_fdb_dump(skb, cb,
4402 if (err == -EMSGSIZE)
4407 if (dev->netdev_ops->ndo_fdb_dump)
4408 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4412 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4414 if (err == -EMSGSIZE)
4419 /* reset fdb offset to 0 for rest of the interfaces */
4435 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4436 struct nlattr **tb, u8 *ndm_flags,
4437 int *br_idx, int *brport_idx, u8 **addr,
4438 u16 *vid, struct netlink_ext_ack *extack)
4443 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4444 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4448 ndm = nlmsg_data(nlh);
4449 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4451 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4455 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4456 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4460 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4461 NDA_MAX, nda_policy, extack);
4465 *ndm_flags = ndm->ndm_flags;
4466 *brport_idx = ndm->ndm_ifindex;
4467 for (i = 0; i <= NDA_MAX; ++i) {
4473 *br_idx = nla_get_u32(tb[i]);
4476 if (nla_len(tb[i]) != ETH_ALEN) {
4477 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4480 *addr = nla_data(tb[i]);
4483 err = fdb_vid_parse(tb[i], vid, extack);
4490 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4498 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4499 struct netlink_ext_ack *extack)
4501 struct net_device *dev = NULL, *br_dev = NULL;
4502 const struct net_device_ops *ops = NULL;
4503 struct net *net = sock_net(in_skb->sk);
4504 struct nlattr *tb[NDA_MAX + 1];
4505 struct sk_buff *skb;
4513 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4514 &brport_idx, &addr, &vid, extack);
4519 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4524 dev = __dev_get_by_index(net, brport_idx);
4526 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4533 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4537 br_dev = __dev_get_by_index(net, br_idx);
4539 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4542 ops = br_dev->netdev_ops;
4546 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4547 if (!netif_is_bridge_port(dev)) {
4548 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4551 br_dev = netdev_master_upper_dev_get(dev);
4553 NL_SET_ERR_MSG(extack, "Master of device not found");
4556 ops = br_dev->netdev_ops;
4558 if (!(ndm_flags & NTF_SELF)) {
4559 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4562 ops = dev->netdev_ops;
4566 if (!br_dev && !dev) {
4567 NL_SET_ERR_MSG(extack, "No device specified");
4571 if (!ops || !ops->ndo_fdb_get) {
4572 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4576 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4582 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4583 NETLINK_CB(in_skb).portid,
4584 nlh->nlmsg_seq, extack);
4588 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4594 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4595 unsigned int attrnum, unsigned int flag)
4598 return nla_put_u8(skb, attrnum, !!(flags & flag));
4602 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4603 struct net_device *dev, u16 mode,
4604 u32 flags, u32 mask, int nlflags,
4606 int (*vlan_fill)(struct sk_buff *skb,
4607 struct net_device *dev,
4610 struct nlmsghdr *nlh;
4611 struct ifinfomsg *ifm;
4612 struct nlattr *br_afspec;
4613 struct nlattr *protinfo;
4614 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4615 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4618 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4622 ifm = nlmsg_data(nlh);
4623 ifm->ifi_family = AF_BRIDGE;
4625 ifm->ifi_type = dev->type;
4626 ifm->ifi_index = dev->ifindex;
4627 ifm->ifi_flags = dev_get_flags(dev);
4628 ifm->ifi_change = 0;
4631 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4632 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4633 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4635 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4637 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4638 (dev->ifindex != dev_get_iflink(dev) &&
4639 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4640 goto nla_put_failure;
4642 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4644 goto nla_put_failure;
4646 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4647 nla_nest_cancel(skb, br_afspec);
4648 goto nla_put_failure;
4651 if (mode != BRIDGE_MODE_UNDEF) {
4652 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4653 nla_nest_cancel(skb, br_afspec);
4654 goto nla_put_failure;
4658 err = vlan_fill(skb, dev, filter_mask);
4660 nla_nest_cancel(skb, br_afspec);
4661 goto nla_put_failure;
4664 nla_nest_end(skb, br_afspec);
4666 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4668 goto nla_put_failure;
4670 if (brport_nla_put_flag(skb, flags, mask,
4671 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4672 brport_nla_put_flag(skb, flags, mask,
4673 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4674 brport_nla_put_flag(skb, flags, mask,
4675 IFLA_BRPORT_FAST_LEAVE,
4676 BR_MULTICAST_FAST_LEAVE) ||
4677 brport_nla_put_flag(skb, flags, mask,
4678 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4679 brport_nla_put_flag(skb, flags, mask,
4680 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4681 brport_nla_put_flag(skb, flags, mask,
4682 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4683 brport_nla_put_flag(skb, flags, mask,
4684 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4685 brport_nla_put_flag(skb, flags, mask,
4686 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4687 brport_nla_put_flag(skb, flags, mask,
4688 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4689 brport_nla_put_flag(skb, flags, mask,
4690 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4691 nla_nest_cancel(skb, protinfo);
4692 goto nla_put_failure;
4695 nla_nest_end(skb, protinfo);
4697 nlmsg_end(skb, nlh);
4700 nlmsg_cancel(skb, nlh);
4701 return err ? err : -EMSGSIZE;
4703 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4705 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4706 bool strict_check, u32 *filter_mask,
4707 struct netlink_ext_ack *extack)
4709 struct nlattr *tb[IFLA_MAX+1];
4713 struct ifinfomsg *ifm;
4715 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4716 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4720 ifm = nlmsg_data(nlh);
4721 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4722 ifm->ifi_change || ifm->ifi_index) {
4723 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4727 err = nlmsg_parse_deprecated_strict(nlh,
4728 sizeof(struct ifinfomsg),
4729 tb, IFLA_MAX, ifla_policy,
4732 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4733 tb, IFLA_MAX, ifla_policy,
4739 /* new attributes should only be added with strict checking */
4740 for (i = 0; i <= IFLA_MAX; ++i) {
4746 *filter_mask = nla_get_u32(tb[i]);
4750 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4759 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4761 const struct nlmsghdr *nlh = cb->nlh;
4762 struct net *net = sock_net(skb->sk);
4763 struct net_device *dev;
4765 u32 portid = NETLINK_CB(cb->skb).portid;
4766 u32 seq = nlh->nlmsg_seq;
4767 u32 filter_mask = 0;
4770 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4772 if (err < 0 && cb->strict_check)
4776 for_each_netdev_rcu(net, dev) {
4777 const struct net_device_ops *ops = dev->netdev_ops;
4778 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4780 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4781 if (idx >= cb->args[0]) {
4782 err = br_dev->netdev_ops->ndo_bridge_getlink(
4783 skb, portid, seq, dev,
4784 filter_mask, NLM_F_MULTI);
4785 if (err < 0 && err != -EOPNOTSUPP) {
4786 if (likely(skb->len))
4795 if (ops->ndo_bridge_getlink) {
4796 if (idx >= cb->args[0]) {
4797 err = ops->ndo_bridge_getlink(skb, portid,
4801 if (err < 0 && err != -EOPNOTSUPP) {
4802 if (likely(skb->len))
4819 static inline size_t bridge_nlmsg_size(void)
4821 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4822 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4823 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4824 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4825 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
4826 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
4827 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
4828 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
4829 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4830 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
4831 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
4834 static int rtnl_bridge_notify(struct net_device *dev)
4836 struct net *net = dev_net(dev);
4837 struct sk_buff *skb;
4838 int err = -EOPNOTSUPP;
4840 if (!dev->netdev_ops->ndo_bridge_getlink)
4843 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4849 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4853 /* Notification info is only filled for bridge ports, not the bridge
4854 * device itself. Therefore, a zero notification length is valid and
4855 * should not result in an error.
4860 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4863 WARN_ON(err == -EMSGSIZE);
4866 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4870 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4871 struct netlink_ext_ack *extack)
4873 struct net *net = sock_net(skb->sk);
4874 struct ifinfomsg *ifm;
4875 struct net_device *dev;
4876 struct nlattr *br_spec, *attr = NULL;
4877 int rem, err = -EOPNOTSUPP;
4879 bool have_flags = false;
4881 if (nlmsg_len(nlh) < sizeof(*ifm))
4884 ifm = nlmsg_data(nlh);
4885 if (ifm->ifi_family != AF_BRIDGE)
4886 return -EPFNOSUPPORT;
4888 dev = __dev_get_by_index(net, ifm->ifi_index);
4890 NL_SET_ERR_MSG(extack, "unknown ifindex");
4894 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4896 nla_for_each_nested(attr, br_spec, rem) {
4897 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4898 if (nla_len(attr) < sizeof(flags))
4902 flags = nla_get_u16(attr);
4908 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4909 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4911 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4916 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4921 flags &= ~BRIDGE_FLAGS_MASTER;
4924 if ((flags & BRIDGE_FLAGS_SELF)) {
4925 if (!dev->netdev_ops->ndo_bridge_setlink)
4928 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4932 flags &= ~BRIDGE_FLAGS_SELF;
4934 /* Generate event to notify upper layer of bridge
4937 err = rtnl_bridge_notify(dev);
4942 memcpy(nla_data(attr), &flags, sizeof(flags));
4947 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4948 struct netlink_ext_ack *extack)
4950 struct net *net = sock_net(skb->sk);
4951 struct ifinfomsg *ifm;
4952 struct net_device *dev;
4953 struct nlattr *br_spec, *attr = NULL;
4954 int rem, err = -EOPNOTSUPP;
4956 bool have_flags = false;
4958 if (nlmsg_len(nlh) < sizeof(*ifm))
4961 ifm = nlmsg_data(nlh);
4962 if (ifm->ifi_family != AF_BRIDGE)
4963 return -EPFNOSUPPORT;
4965 dev = __dev_get_by_index(net, ifm->ifi_index);
4967 NL_SET_ERR_MSG(extack, "unknown ifindex");
4971 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4973 nla_for_each_nested(attr, br_spec, rem) {
4974 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4975 if (nla_len(attr) < sizeof(flags))
4979 flags = nla_get_u16(attr);
4985 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4986 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4988 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4993 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4997 flags &= ~BRIDGE_FLAGS_MASTER;
5000 if ((flags & BRIDGE_FLAGS_SELF)) {
5001 if (!dev->netdev_ops->ndo_bridge_dellink)
5004 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5008 flags &= ~BRIDGE_FLAGS_SELF;
5010 /* Generate event to notify upper layer of bridge
5013 err = rtnl_bridge_notify(dev);
5018 memcpy(nla_data(attr), &flags, sizeof(flags));
5023 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5025 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5026 (!idxattr || idxattr == attrid);
5029 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5030 static int rtnl_get_offload_stats_attr_size(int attr_id)
5033 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5034 return sizeof(struct rtnl_link_stats64);
5040 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5043 struct nlattr *attr = NULL;
5048 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5049 dev->netdev_ops->ndo_get_offload_stats))
5052 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5053 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5054 if (attr_id < *prividx)
5057 size = rtnl_get_offload_stats_attr_size(attr_id);
5061 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5064 attr = nla_reserve_64bit(skb, attr_id, size,
5065 IFLA_OFFLOAD_XSTATS_UNSPEC);
5067 goto nla_put_failure;
5069 attr_data = nla_data(attr);
5070 memset(attr_data, 0, size);
5071 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5074 goto get_offload_stats_failure;
5085 get_offload_stats_failure:
5090 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5096 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5097 dev->netdev_ops->ndo_get_offload_stats))
5100 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5101 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5102 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5104 size = rtnl_get_offload_stats_attr_size(attr_id);
5105 nla_size += nla_total_size_64bit(size);
5109 nla_size += nla_total_size(0);
5114 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5115 int type, u32 pid, u32 seq, u32 change,
5116 unsigned int flags, unsigned int filter_mask,
5117 int *idxattr, int *prividx)
5119 struct if_stats_msg *ifsm;
5120 struct nlmsghdr *nlh;
5121 struct nlattr *attr;
5122 int s_prividx = *prividx;
5127 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5131 ifsm = nlmsg_data(nlh);
5132 ifsm->family = PF_UNSPEC;
5135 ifsm->ifindex = dev->ifindex;
5136 ifsm->filter_mask = filter_mask;
5138 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5139 struct rtnl_link_stats64 *sp;
5141 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5142 sizeof(struct rtnl_link_stats64),
5145 goto nla_put_failure;
5147 sp = nla_data(attr);
5148 dev_get_stats(dev, sp);
5151 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5152 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5154 if (ops && ops->fill_linkxstats) {
5155 *idxattr = IFLA_STATS_LINK_XSTATS;
5156 attr = nla_nest_start_noflag(skb,
5157 IFLA_STATS_LINK_XSTATS);
5159 goto nla_put_failure;
5161 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5162 nla_nest_end(skb, attr);
5164 goto nla_put_failure;
5169 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5171 const struct rtnl_link_ops *ops = NULL;
5172 const struct net_device *master;
5174 master = netdev_master_upper_dev_get(dev);
5176 ops = master->rtnl_link_ops;
5177 if (ops && ops->fill_linkxstats) {
5178 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5179 attr = nla_nest_start_noflag(skb,
5180 IFLA_STATS_LINK_XSTATS_SLAVE);
5182 goto nla_put_failure;
5184 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5185 nla_nest_end(skb, attr);
5187 goto nla_put_failure;
5192 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5194 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5195 attr = nla_nest_start_noflag(skb,
5196 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5198 goto nla_put_failure;
5200 err = rtnl_get_offload_stats(skb, dev, prividx);
5201 if (err == -ENODATA)
5202 nla_nest_cancel(skb, attr);
5204 nla_nest_end(skb, attr);
5206 if (err && err != -ENODATA)
5207 goto nla_put_failure;
5211 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5212 struct rtnl_af_ops *af_ops;
5214 *idxattr = IFLA_STATS_AF_SPEC;
5215 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5217 goto nla_put_failure;
5220 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5221 if (af_ops->fill_stats_af) {
5225 af = nla_nest_start_noflag(skb,
5229 goto nla_put_failure;
5231 err = af_ops->fill_stats_af(skb, dev);
5233 if (err == -ENODATA) {
5234 nla_nest_cancel(skb, af);
5235 } else if (err < 0) {
5237 goto nla_put_failure;
5240 nla_nest_end(skb, af);
5245 nla_nest_end(skb, attr);
5250 nlmsg_end(skb, nlh);
5255 /* not a multi message or no progress mean a real error */
5256 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5257 nlmsg_cancel(skb, nlh);
5259 nlmsg_end(skb, nlh);
5264 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5269 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5270 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5272 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5273 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5274 int attr = IFLA_STATS_LINK_XSTATS;
5276 if (ops && ops->get_linkxstats_size) {
5277 size += nla_total_size(ops->get_linkxstats_size(dev,
5279 /* for IFLA_STATS_LINK_XSTATS */
5280 size += nla_total_size(0);
5284 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5285 struct net_device *_dev = (struct net_device *)dev;
5286 const struct rtnl_link_ops *ops = NULL;
5287 const struct net_device *master;
5289 /* netdev_master_upper_dev_get can't take const */
5290 master = netdev_master_upper_dev_get(_dev);
5292 ops = master->rtnl_link_ops;
5293 if (ops && ops->get_linkxstats_size) {
5294 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5296 size += nla_total_size(ops->get_linkxstats_size(dev,
5298 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5299 size += nla_total_size(0);
5303 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5304 size += rtnl_get_offload_stats_size(dev);
5306 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5307 struct rtnl_af_ops *af_ops;
5309 /* for IFLA_STATS_AF_SPEC */
5310 size += nla_total_size(0);
5313 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5314 if (af_ops->get_stats_af_size) {
5315 size += nla_total_size(
5316 af_ops->get_stats_af_size(dev));
5319 size += nla_total_size(0);
5328 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5329 bool is_dump, struct netlink_ext_ack *extack)
5331 struct if_stats_msg *ifsm;
5333 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5334 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5341 ifsm = nlmsg_data(nlh);
5343 /* only requests using strict checks can pass data to influence
5344 * the dump. The legacy exception is filter_mask.
5346 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5347 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5350 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5351 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5354 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5355 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5362 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5363 struct netlink_ext_ack *extack)
5365 struct net *net = sock_net(skb->sk);
5366 struct net_device *dev = NULL;
5367 int idxattr = 0, prividx = 0;
5368 struct if_stats_msg *ifsm;
5369 struct sk_buff *nskb;
5373 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5378 ifsm = nlmsg_data(nlh);
5379 if (ifsm->ifindex > 0)
5380 dev = __dev_get_by_index(net, ifsm->ifindex);
5387 filter_mask = ifsm->filter_mask;
5391 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5395 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5396 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5397 0, filter_mask, &idxattr, &prividx);
5399 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5400 WARN_ON(err == -EMSGSIZE);
5403 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5409 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5411 struct netlink_ext_ack *extack = cb->extack;
5412 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5413 struct net *net = sock_net(skb->sk);
5414 unsigned int flags = NLM_F_MULTI;
5415 struct if_stats_msg *ifsm;
5416 struct hlist_head *head;
5417 struct net_device *dev;
5418 u32 filter_mask = 0;
5422 s_idx = cb->args[1];
5423 s_idxattr = cb->args[2];
5424 s_prividx = cb->args[3];
5426 cb->seq = net->dev_base_seq;
5428 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5432 ifsm = nlmsg_data(cb->nlh);
5433 filter_mask = ifsm->filter_mask;
5435 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5439 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5441 head = &net->dev_index_head[h];
5442 hlist_for_each_entry(dev, head, index_hlist) {
5445 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5446 NETLINK_CB(cb->skb).portid,
5447 cb->nlh->nlmsg_seq, 0,
5449 &s_idxattr, &s_prividx);
5450 /* If we ran out of room on the first message,
5453 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5459 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5465 cb->args[3] = s_prividx;
5466 cb->args[2] = s_idxattr;
5473 /* Process one rtnetlink message. */
5475 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5476 struct netlink_ext_ack *extack)
5478 struct net *net = sock_net(skb->sk);
5479 struct rtnl_link *link;
5480 struct module *owner;
5481 int err = -EOPNOTSUPP;
5482 rtnl_doit_func doit;
5488 type = nlh->nlmsg_type;
5494 /* All the messages must have at least 1 byte length */
5495 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5498 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5501 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5505 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5507 rtnl_dumpit_func dumpit;
5508 u32 min_dump_alloc = 0;
5510 link = rtnl_get_link(family, type);
5511 if (!link || !link->dumpit) {
5513 link = rtnl_get_link(family, type);
5514 if (!link || !link->dumpit)
5517 owner = link->owner;
5518 dumpit = link->dumpit;
5520 if (type == RTM_GETLINK - RTM_BASE)
5521 min_dump_alloc = rtnl_calcit(skb, nlh);
5524 /* need to do this before rcu_read_unlock() */
5525 if (!try_module_get(owner))
5526 err = -EPROTONOSUPPORT;
5532 struct netlink_dump_control c = {
5534 .min_dump_alloc = min_dump_alloc,
5537 err = netlink_dump_start(rtnl, skb, nlh, &c);
5538 /* netlink_dump_start() will keep a reference on
5539 * module if dump is still in progress.
5546 link = rtnl_get_link(family, type);
5547 if (!link || !link->doit) {
5549 link = rtnl_get_link(PF_UNSPEC, type);
5550 if (!link || !link->doit)
5554 owner = link->owner;
5555 if (!try_module_get(owner)) {
5556 err = -EPROTONOSUPPORT;
5560 flags = link->flags;
5561 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5565 err = doit(skb, nlh, extack);
5572 link = rtnl_get_link(family, type);
5573 if (link && link->doit)
5574 err = link->doit(skb, nlh, extack);
5590 static void rtnetlink_rcv(struct sk_buff *skb)
5592 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5595 static int rtnetlink_bind(struct net *net, int group)
5598 case RTNLGRP_IPV4_MROUTE_R:
5599 case RTNLGRP_IPV6_MROUTE_R:
5600 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5607 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5609 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5613 case NETDEV_CHANGEMTU:
5614 case NETDEV_CHANGEADDR:
5615 case NETDEV_CHANGENAME:
5616 case NETDEV_FEAT_CHANGE:
5617 case NETDEV_BONDING_FAILOVER:
5618 case NETDEV_POST_TYPE_CHANGE:
5619 case NETDEV_NOTIFY_PEERS:
5620 case NETDEV_CHANGEUPPER:
5621 case NETDEV_RESEND_IGMP:
5622 case NETDEV_CHANGEINFODATA:
5623 case NETDEV_CHANGELOWERSTATE:
5624 case NETDEV_CHANGE_TX_QUEUE_LEN:
5625 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5626 GFP_KERNEL, NULL, 0);
5634 static struct notifier_block rtnetlink_dev_notifier = {
5635 .notifier_call = rtnetlink_event,
5639 static int __net_init rtnetlink_net_init(struct net *net)
5642 struct netlink_kernel_cfg cfg = {
5643 .groups = RTNLGRP_MAX,
5644 .input = rtnetlink_rcv,
5645 .cb_mutex = &rtnl_mutex,
5646 .flags = NL_CFG_F_NONROOT_RECV,
5647 .bind = rtnetlink_bind,
5650 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5657 static void __net_exit rtnetlink_net_exit(struct net *net)
5659 netlink_kernel_release(net->rtnl);
5663 static struct pernet_operations rtnetlink_net_ops = {
5664 .init = rtnetlink_net_init,
5665 .exit = rtnetlink_net_exit,
5668 void __init rtnetlink_init(void)
5670 if (register_pernet_subsys(&rtnetlink_net_ops))
5671 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5673 register_netdevice_notifier(&rtnetlink_dev_notifier);
5675 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5676 rtnl_dump_ifinfo, 0);
5677 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5678 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5679 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5681 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5682 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5683 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5685 rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5686 rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5688 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5689 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5690 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5692 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5693 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5694 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5696 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,