1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Routing netlink socket interface: protocol independent part.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
12 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
15 #include <linux/bitops.h>
16 #include <linux/errno.h>
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/kernel.h>
21 #include <linux/timer.h>
22 #include <linux/string.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/fcntl.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/capability.h>
30 #include <linux/skbuff.h>
31 #include <linux/init.h>
32 #include <linux/security.h>
33 #include <linux/mutex.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_vlan.h>
37 #include <linux/pci.h>
38 #include <linux/etherdevice.h>
39 #include <linux/bpf.h>
41 #include <linux/uaccess.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
46 #include <net/protocol.h>
48 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
57 #define RTNL_MAX_TYPE 50
58 #define RTNL_SLAVE_MAX_TYPE 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 modifiy behaviour of doit/dumpit functions
239 * Like rtnl_register, but for use by removable modules.
241 int rtnl_register_module(struct module *owner,
242 int protocol, int msgtype,
243 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
246 return rtnl_register_internal(owner, protocol, msgtype,
247 doit, dumpit, flags);
249 EXPORT_SYMBOL_GPL(rtnl_register_module);
252 * rtnl_register - Register a rtnetlink message type
253 * @protocol: Protocol family or PF_UNSPEC
254 * @msgtype: rtnetlink message type
255 * @doit: Function pointer called for each request message
256 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
257 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
259 * Registers the specified function pointers (at least one of them has
260 * to be non-NULL) to be called whenever a request message for the
261 * specified protocol family and message type is received.
263 * The special protocol family PF_UNSPEC may be used to define fallback
264 * function pointers for the case when no entry for the specific protocol
267 void rtnl_register(int protocol, int msgtype,
268 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
273 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
276 pr_err("Unable to register rtnetlink message handler, "
277 "protocol = %d, message type = %d\n", protocol, msgtype);
281 * rtnl_unregister - Unregister a rtnetlink message type
282 * @protocol: Protocol family or PF_UNSPEC
283 * @msgtype: rtnetlink message type
285 * Returns 0 on success or a negative error code.
287 int rtnl_unregister(int protocol, int msgtype)
289 struct rtnl_link __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 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->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;
546 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
547 if (ops->family == family)
555 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
556 * @ops: struct rtnl_af_ops * to register
558 * Returns 0 on success or a negative error code.
560 void rtnl_af_register(struct rtnl_af_ops *ops)
563 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
566 EXPORT_SYMBOL_GPL(rtnl_af_register);
569 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
570 * @ops: struct rtnl_af_ops * to unregister
572 void rtnl_af_unregister(struct rtnl_af_ops *ops)
575 list_del_rcu(&ops->list);
580 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
582 static size_t rtnl_link_get_af_size(const struct net_device *dev,
585 struct rtnl_af_ops *af_ops;
589 size = nla_total_size(sizeof(struct nlattr));
592 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
593 if (af_ops->get_link_af_size) {
594 /* AF_* + nested data */
595 size += nla_total_size(sizeof(struct nlattr)) +
596 af_ops->get_link_af_size(dev, ext_filter_mask);
604 static bool rtnl_have_link_slave_info(const struct net_device *dev)
606 struct net_device *master_dev;
611 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
612 if (master_dev && master_dev->rtnl_link_ops)
618 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
619 const struct net_device *dev)
621 struct net_device *master_dev;
622 const struct rtnl_link_ops *ops;
623 struct nlattr *slave_data;
626 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
629 ops = master_dev->rtnl_link_ops;
632 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
634 if (ops->fill_slave_info) {
635 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
638 err = ops->fill_slave_info(skb, master_dev, dev);
640 goto err_cancel_slave_data;
641 nla_nest_end(skb, slave_data);
645 err_cancel_slave_data:
646 nla_nest_cancel(skb, slave_data);
650 static int rtnl_link_info_fill(struct sk_buff *skb,
651 const struct net_device *dev)
653 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
659 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
661 if (ops->fill_xstats) {
662 err = ops->fill_xstats(skb, dev);
666 if (ops->fill_info) {
667 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
670 err = ops->fill_info(skb, dev);
672 goto err_cancel_data;
673 nla_nest_end(skb, data);
678 nla_nest_cancel(skb, data);
682 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
684 struct nlattr *linkinfo;
687 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
688 if (linkinfo == NULL)
691 err = rtnl_link_info_fill(skb, dev);
693 goto err_cancel_link;
695 err = rtnl_link_slave_info_fill(skb, dev);
697 goto err_cancel_link;
699 nla_nest_end(skb, linkinfo);
703 nla_nest_cancel(skb, linkinfo);
708 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
710 struct sock *rtnl = net->rtnl;
713 NETLINK_CB(skb).dst_group = group;
715 refcount_inc(&skb->users);
716 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
718 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
722 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
724 struct sock *rtnl = net->rtnl;
726 return nlmsg_unicast(rtnl, skb, pid);
728 EXPORT_SYMBOL(rtnl_unicast);
730 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
731 struct nlmsghdr *nlh, gfp_t flags)
733 struct sock *rtnl = net->rtnl;
737 report = nlmsg_report(nlh);
739 nlmsg_notify(rtnl, skb, pid, group, report, flags);
741 EXPORT_SYMBOL(rtnl_notify);
743 void rtnl_set_sk_err(struct net *net, u32 group, int error)
745 struct sock *rtnl = net->rtnl;
747 netlink_set_err(rtnl, 0, group, error);
749 EXPORT_SYMBOL(rtnl_set_sk_err);
751 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
756 /* nothing is dumped for dst_default_metrics, so just skip the loop */
757 if (metrics == dst_default_metrics.metrics)
760 mx = nla_nest_start_noflag(skb, RTA_METRICS);
764 for (i = 0; i < RTAX_MAX; i++) {
766 if (i == RTAX_CC_ALGO - 1) {
767 char tmp[TCP_CA_NAME_MAX], *name;
769 name = tcp_ca_get_name_by_key(metrics[i], tmp);
772 if (nla_put_string(skb, i + 1, name))
773 goto nla_put_failure;
774 } else if (i == RTAX_FEATURES - 1) {
775 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
779 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
780 if (nla_put_u32(skb, i + 1, user_features))
781 goto nla_put_failure;
783 if (nla_put_u32(skb, i + 1, metrics[i]))
784 goto nla_put_failure;
791 nla_nest_cancel(skb, mx);
795 return nla_nest_end(skb, mx);
798 nla_nest_cancel(skb, mx);
801 EXPORT_SYMBOL(rtnetlink_put_metrics);
803 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
804 long expires, u32 error)
806 struct rta_cacheinfo ci = {
812 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
813 ci.rta_used = dst->__use;
814 ci.rta_clntref = atomic_read(&dst->__refcnt);
819 clock = jiffies_to_clock_t(abs(expires));
820 clock = min_t(unsigned long, clock, INT_MAX);
821 ci.rta_expires = (expires > 0) ? clock : -clock;
823 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
825 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
827 static void set_operstate(struct net_device *dev, unsigned char transition)
829 unsigned char operstate = dev->operstate;
831 switch (transition) {
833 if ((operstate == IF_OPER_DORMANT ||
834 operstate == IF_OPER_TESTING ||
835 operstate == IF_OPER_UNKNOWN) &&
836 !netif_dormant(dev) && !netif_testing(dev))
837 operstate = IF_OPER_UP;
840 case IF_OPER_TESTING:
841 if (operstate == IF_OPER_UP ||
842 operstate == IF_OPER_UNKNOWN)
843 operstate = IF_OPER_TESTING;
846 case IF_OPER_DORMANT:
847 if (operstate == IF_OPER_UP ||
848 operstate == IF_OPER_UNKNOWN)
849 operstate = IF_OPER_DORMANT;
853 if (dev->operstate != operstate) {
854 write_lock_bh(&dev_base_lock);
855 dev->operstate = operstate;
856 write_unlock_bh(&dev_base_lock);
857 netdev_state_change(dev);
861 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
863 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
864 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
867 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
868 const struct ifinfomsg *ifm)
870 unsigned int flags = ifm->ifi_flags;
872 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
874 flags = (flags & ifm->ifi_change) |
875 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
880 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
881 const struct rtnl_link_stats64 *b)
883 a->rx_packets = b->rx_packets;
884 a->tx_packets = b->tx_packets;
885 a->rx_bytes = b->rx_bytes;
886 a->tx_bytes = b->tx_bytes;
887 a->rx_errors = b->rx_errors;
888 a->tx_errors = b->tx_errors;
889 a->rx_dropped = b->rx_dropped;
890 a->tx_dropped = b->tx_dropped;
892 a->multicast = b->multicast;
893 a->collisions = b->collisions;
895 a->rx_length_errors = b->rx_length_errors;
896 a->rx_over_errors = b->rx_over_errors;
897 a->rx_crc_errors = b->rx_crc_errors;
898 a->rx_frame_errors = b->rx_frame_errors;
899 a->rx_fifo_errors = b->rx_fifo_errors;
900 a->rx_missed_errors = b->rx_missed_errors;
902 a->tx_aborted_errors = b->tx_aborted_errors;
903 a->tx_carrier_errors = b->tx_carrier_errors;
904 a->tx_fifo_errors = b->tx_fifo_errors;
905 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
906 a->tx_window_errors = b->tx_window_errors;
908 a->rx_compressed = b->rx_compressed;
909 a->tx_compressed = b->tx_compressed;
911 a->rx_nohandler = b->rx_nohandler;
915 static inline int rtnl_vfinfo_size(const struct net_device *dev,
918 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
919 int num_vfs = dev_num_vf(dev->dev.parent);
920 size_t size = nla_total_size(0);
923 nla_total_size(sizeof(struct ifla_vf_mac)) +
924 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
925 nla_total_size(sizeof(struct ifla_vf_vlan)) +
926 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
927 nla_total_size(MAX_VLAN_LIST_LEN *
928 sizeof(struct ifla_vf_vlan_info)) +
929 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
930 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
931 nla_total_size(sizeof(struct ifla_vf_rate)) +
932 nla_total_size(sizeof(struct ifla_vf_link_state)) +
933 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
934 nla_total_size(0) + /* nest IFLA_VF_STATS */
935 /* IFLA_VF_STATS_RX_PACKETS */
936 nla_total_size_64bit(sizeof(__u64)) +
937 /* IFLA_VF_STATS_TX_PACKETS */
938 nla_total_size_64bit(sizeof(__u64)) +
939 /* IFLA_VF_STATS_RX_BYTES */
940 nla_total_size_64bit(sizeof(__u64)) +
941 /* IFLA_VF_STATS_TX_BYTES */
942 nla_total_size_64bit(sizeof(__u64)) +
943 /* IFLA_VF_STATS_BROADCAST */
944 nla_total_size_64bit(sizeof(__u64)) +
945 /* IFLA_VF_STATS_MULTICAST */
946 nla_total_size_64bit(sizeof(__u64)) +
947 /* IFLA_VF_STATS_RX_DROPPED */
948 nla_total_size_64bit(sizeof(__u64)) +
949 /* IFLA_VF_STATS_TX_DROPPED */
950 nla_total_size_64bit(sizeof(__u64)) +
951 nla_total_size(sizeof(struct ifla_vf_trust)));
957 static size_t rtnl_port_size(const struct net_device *dev,
960 size_t port_size = nla_total_size(4) /* PORT_VF */
961 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
962 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
963 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
964 + nla_total_size(1) /* PROT_VDP_REQUEST */
965 + nla_total_size(2); /* PORT_VDP_RESPONSE */
966 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
967 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
969 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
972 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
973 !(ext_filter_mask & RTEXT_FILTER_VF))
975 if (dev_num_vf(dev->dev.parent))
976 return port_self_size + vf_ports_size +
977 vf_port_size * dev_num_vf(dev->dev.parent);
979 return port_self_size;
982 static size_t rtnl_xdp_size(void)
984 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
985 nla_total_size(1) + /* XDP_ATTACHED */
986 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
987 nla_total_size(4); /* XDP_<mode>_PROG_ID */
992 static size_t rtnl_prop_list_size(const struct net_device *dev)
994 struct netdev_name_node *name_node;
997 if (list_empty(&dev->name_node->list))
999 size = nla_total_size(0);
1000 list_for_each_entry(name_node, &dev->name_node->list, list)
1001 size += nla_total_size(ALTIFNAMSIZ);
1005 static size_t rtnl_proto_down_size(const struct net_device *dev)
1007 size_t size = nla_total_size(1);
1009 if (dev->proto_down_reason)
1010 size += nla_total_size(0) + nla_total_size(4);
1015 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1016 u32 ext_filter_mask)
1018 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1019 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1020 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1021 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1022 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1023 + nla_total_size(sizeof(struct rtnl_link_stats))
1024 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1025 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1026 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1027 + nla_total_size(4) /* IFLA_TXQLEN */
1028 + nla_total_size(4) /* IFLA_WEIGHT */
1029 + nla_total_size(4) /* IFLA_MTU */
1030 + nla_total_size(4) /* IFLA_LINK */
1031 + nla_total_size(4) /* IFLA_MASTER */
1032 + nla_total_size(1) /* IFLA_CARRIER */
1033 + nla_total_size(4) /* IFLA_PROMISCUITY */
1034 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1035 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1036 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1037 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1038 + nla_total_size(1) /* IFLA_OPERSTATE */
1039 + nla_total_size(1) /* IFLA_LINKMODE */
1040 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1041 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1042 + nla_total_size(4) /* IFLA_GROUP */
1043 + nla_total_size(ext_filter_mask
1044 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1045 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1046 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1047 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1048 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1049 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1050 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1051 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1052 + rtnl_xdp_size() /* IFLA_XDP */
1053 + nla_total_size(4) /* IFLA_EVENT */
1054 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1055 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1056 + rtnl_proto_down_size(dev) /* proto down */
1057 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1058 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1059 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1060 + nla_total_size(4) /* IFLA_MIN_MTU */
1061 + nla_total_size(4) /* IFLA_MAX_MTU */
1062 + rtnl_prop_list_size(dev)
1063 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1067 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1069 struct nlattr *vf_ports;
1070 struct nlattr *vf_port;
1074 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1078 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1079 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1081 goto nla_put_failure;
1082 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1083 goto nla_put_failure;
1084 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1085 if (err == -EMSGSIZE)
1086 goto nla_put_failure;
1088 nla_nest_cancel(skb, vf_port);
1091 nla_nest_end(skb, vf_port);
1094 nla_nest_end(skb, vf_ports);
1099 nla_nest_cancel(skb, vf_ports);
1103 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1105 struct nlattr *port_self;
1108 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1112 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1114 nla_nest_cancel(skb, port_self);
1115 return (err == -EMSGSIZE) ? err : 0;
1118 nla_nest_end(skb, port_self);
1123 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1124 u32 ext_filter_mask)
1128 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1129 !(ext_filter_mask & RTEXT_FILTER_VF))
1132 err = rtnl_port_self_fill(skb, dev);
1136 if (dev_num_vf(dev->dev.parent)) {
1137 err = rtnl_vf_ports_fill(skb, dev);
1145 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1148 struct netdev_phys_item_id ppid;
1150 err = dev_get_phys_port_id(dev, &ppid);
1152 if (err == -EOPNOTSUPP)
1157 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1163 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1165 char name[IFNAMSIZ];
1168 err = dev_get_phys_port_name(dev, name, sizeof(name));
1170 if (err == -EOPNOTSUPP)
1175 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1181 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1183 struct netdev_phys_item_id ppid = { };
1186 err = dev_get_port_parent_id(dev, &ppid, false);
1188 if (err == -EOPNOTSUPP)
1193 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1199 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1200 struct net_device *dev)
1202 struct rtnl_link_stats64 *sp;
1203 struct nlattr *attr;
1205 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1206 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1210 sp = nla_data(attr);
1211 dev_get_stats(dev, sp);
1213 attr = nla_reserve(skb, IFLA_STATS,
1214 sizeof(struct rtnl_link_stats));
1218 copy_rtnl_link_stats(nla_data(attr), sp);
1223 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1224 struct net_device *dev,
1226 struct nlattr *vfinfo)
1228 struct ifla_vf_rss_query_en vf_rss_query_en;
1229 struct nlattr *vf, *vfstats, *vfvlanlist;
1230 struct ifla_vf_link_state vf_linkstate;
1231 struct ifla_vf_vlan_info vf_vlan_info;
1232 struct ifla_vf_spoofchk vf_spoofchk;
1233 struct ifla_vf_tx_rate vf_tx_rate;
1234 struct ifla_vf_stats vf_stats;
1235 struct ifla_vf_trust vf_trust;
1236 struct ifla_vf_vlan vf_vlan;
1237 struct ifla_vf_rate vf_rate;
1238 struct ifla_vf_mac vf_mac;
1239 struct ifla_vf_broadcast vf_broadcast;
1240 struct ifla_vf_info ivi;
1241 struct ifla_vf_guid node_guid;
1242 struct ifla_vf_guid port_guid;
1244 memset(&ivi, 0, sizeof(ivi));
1246 /* Not all SR-IOV capable drivers support the
1247 * spoofcheck and "RSS query enable" query. Preset to
1248 * -1 so the user space tool can detect that the driver
1249 * didn't report anything.
1252 ivi.rss_query_en = -1;
1254 /* The default value for VF link state is "auto"
1255 * IFLA_VF_LINK_STATE_AUTO which equals zero
1258 /* VLAN Protocol by default is 802.1Q */
1259 ivi.vlan_proto = htons(ETH_P_8021Q);
1260 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1263 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1264 memset(&node_guid, 0, sizeof(node_guid));
1265 memset(&port_guid, 0, sizeof(port_guid));
1274 vf_rss_query_en.vf =
1277 port_guid.vf = ivi.vf;
1279 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1280 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1281 vf_vlan.vlan = ivi.vlan;
1282 vf_vlan.qos = ivi.qos;
1283 vf_vlan_info.vlan = ivi.vlan;
1284 vf_vlan_info.qos = ivi.qos;
1285 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1286 vf_tx_rate.rate = ivi.max_tx_rate;
1287 vf_rate.min_tx_rate = ivi.min_tx_rate;
1288 vf_rate.max_tx_rate = ivi.max_tx_rate;
1289 vf_spoofchk.setting = ivi.spoofchk;
1290 vf_linkstate.link_state = ivi.linkstate;
1291 vf_rss_query_en.setting = ivi.rss_query_en;
1292 vf_trust.setting = ivi.trusted;
1293 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1295 goto nla_put_vfinfo_failure;
1296 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1297 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1298 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1299 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1301 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1303 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1305 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1307 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1308 sizeof(vf_rss_query_en),
1309 &vf_rss_query_en) ||
1310 nla_put(skb, IFLA_VF_TRUST,
1311 sizeof(vf_trust), &vf_trust))
1312 goto nla_put_vf_failure;
1314 if (dev->netdev_ops->ndo_get_vf_guid &&
1315 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1317 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1319 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1321 goto nla_put_vf_failure;
1323 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1325 goto nla_put_vf_failure;
1326 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1328 nla_nest_cancel(skb, vfvlanlist);
1329 goto nla_put_vf_failure;
1331 nla_nest_end(skb, vfvlanlist);
1332 memset(&vf_stats, 0, sizeof(vf_stats));
1333 if (dev->netdev_ops->ndo_get_vf_stats)
1334 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1336 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1338 goto nla_put_vf_failure;
1339 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1340 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1341 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1342 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1343 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1344 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1345 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1346 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1347 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1348 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1349 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1350 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1351 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1352 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1353 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1354 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1355 nla_nest_cancel(skb, vfstats);
1356 goto nla_put_vf_failure;
1358 nla_nest_end(skb, vfstats);
1359 nla_nest_end(skb, vf);
1363 nla_nest_cancel(skb, vf);
1364 nla_put_vfinfo_failure:
1365 nla_nest_cancel(skb, vfinfo);
1369 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1370 struct net_device *dev,
1371 u32 ext_filter_mask)
1373 struct nlattr *vfinfo;
1376 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1379 num_vfs = dev_num_vf(dev->dev.parent);
1380 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1383 if (!dev->netdev_ops->ndo_get_vf_config)
1386 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1390 for (i = 0; i < num_vfs; i++) {
1391 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1395 nla_nest_end(skb, vfinfo);
1399 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1401 struct rtnl_link_ifmap map;
1403 memset(&map, 0, sizeof(map));
1404 map.mem_start = dev->mem_start;
1405 map.mem_end = dev->mem_end;
1406 map.base_addr = dev->base_addr;
1409 map.port = dev->if_port;
1411 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1417 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1419 const struct bpf_prog *generic_xdp_prog;
1423 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1424 if (!generic_xdp_prog)
1426 return generic_xdp_prog->aux->id;
1429 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1431 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1434 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1436 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1439 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1440 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1441 u32 (*get_prog_id)(struct net_device *dev))
1446 curr_id = get_prog_id(dev);
1451 err = nla_put_u32(skb, attr, curr_id);
1455 if (*mode != XDP_ATTACHED_NONE)
1456 *mode = XDP_ATTACHED_MULTI;
1463 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1470 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1475 mode = XDP_ATTACHED_NONE;
1476 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1477 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1480 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1481 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1484 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1485 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1489 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1493 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1494 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1499 nla_nest_end(skb, xdp);
1503 nla_nest_cancel(skb, xdp);
1507 static u32 rtnl_get_event(unsigned long event)
1509 u32 rtnl_event_type = IFLA_EVENT_NONE;
1513 rtnl_event_type = IFLA_EVENT_REBOOT;
1515 case NETDEV_FEAT_CHANGE:
1516 rtnl_event_type = IFLA_EVENT_FEATURES;
1518 case NETDEV_BONDING_FAILOVER:
1519 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1521 case NETDEV_NOTIFY_PEERS:
1522 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1524 case NETDEV_RESEND_IGMP:
1525 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1527 case NETDEV_CHANGEINFODATA:
1528 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1534 return rtnl_event_type;
1537 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1539 const struct net_device *upper_dev;
1544 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1546 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1552 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1555 int ifindex = dev_get_iflink(dev);
1557 if (force || dev->ifindex != ifindex)
1558 return nla_put_u32(skb, IFLA_LINK, ifindex);
1563 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1564 struct net_device *dev)
1569 ret = dev_get_alias(dev, buf, sizeof(buf));
1570 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1573 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1574 const struct net_device *dev,
1575 struct net *src_net, gfp_t gfp)
1577 bool put_iflink = false;
1579 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1580 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1582 if (!net_eq(dev_net(dev), link_net)) {
1583 int id = peernet2id_alloc(src_net, link_net, gfp);
1585 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1592 return nla_put_iflink(skb, dev, put_iflink);
1595 static int rtnl_fill_link_af(struct sk_buff *skb,
1596 const struct net_device *dev,
1597 u32 ext_filter_mask)
1599 const struct rtnl_af_ops *af_ops;
1600 struct nlattr *af_spec;
1602 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1606 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1610 if (!af_ops->fill_link_af)
1613 af = nla_nest_start_noflag(skb, af_ops->family);
1617 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1619 * Caller may return ENODATA to indicate that there
1620 * was no data to be dumped. This is not an error, it
1621 * means we should trim the attribute header and
1624 if (err == -ENODATA)
1625 nla_nest_cancel(skb, af);
1629 nla_nest_end(skb, af);
1632 nla_nest_end(skb, af_spec);
1636 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1637 const struct net_device *dev)
1639 struct netdev_name_node *name_node;
1642 list_for_each_entry(name_node, &dev->name_node->list, list) {
1643 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1650 static int rtnl_fill_prop_list(struct sk_buff *skb,
1651 const struct net_device *dev)
1653 struct nlattr *prop_list;
1656 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1660 ret = rtnl_fill_alt_ifnames(skb, dev);
1664 nla_nest_end(skb, prop_list);
1668 nla_nest_cancel(skb, prop_list);
1672 static int rtnl_fill_proto_down(struct sk_buff *skb,
1673 const struct net_device *dev)
1678 if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1679 goto nla_put_failure;
1681 preason = dev->proto_down_reason;
1685 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1689 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1690 nla_nest_cancel(skb, pr);
1691 goto nla_put_failure;
1694 nla_nest_end(skb, pr);
1701 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1702 struct net_device *dev, struct net *src_net,
1703 int type, u32 pid, u32 seq, u32 change,
1704 unsigned int flags, u32 ext_filter_mask,
1705 u32 event, int *new_nsid, int new_ifindex,
1706 int tgt_netnsid, gfp_t gfp)
1708 struct ifinfomsg *ifm;
1709 struct nlmsghdr *nlh;
1712 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1716 ifm = nlmsg_data(nlh);
1717 ifm->ifi_family = AF_UNSPEC;
1719 ifm->ifi_type = dev->type;
1720 ifm->ifi_index = dev->ifindex;
1721 ifm->ifi_flags = dev_get_flags(dev);
1722 ifm->ifi_change = change;
1724 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1725 goto nla_put_failure;
1727 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1728 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1729 nla_put_u8(skb, IFLA_OPERSTATE,
1730 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1731 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1732 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1733 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1734 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1735 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1736 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1737 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1738 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1739 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1741 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1743 put_master_ifindex(skb, dev) ||
1744 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1746 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1747 nla_put_ifalias(skb, dev) ||
1748 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1749 atomic_read(&dev->carrier_up_count) +
1750 atomic_read(&dev->carrier_down_count)) ||
1751 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1752 atomic_read(&dev->carrier_up_count)) ||
1753 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1754 atomic_read(&dev->carrier_down_count)))
1755 goto nla_put_failure;
1757 if (rtnl_fill_proto_down(skb, dev))
1758 goto nla_put_failure;
1760 if (event != IFLA_EVENT_NONE) {
1761 if (nla_put_u32(skb, IFLA_EVENT, event))
1762 goto nla_put_failure;
1765 if (rtnl_fill_link_ifmap(skb, dev))
1766 goto nla_put_failure;
1768 if (dev->addr_len) {
1769 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1770 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1771 goto nla_put_failure;
1774 if (rtnl_phys_port_id_fill(skb, dev))
1775 goto nla_put_failure;
1777 if (rtnl_phys_port_name_fill(skb, dev))
1778 goto nla_put_failure;
1780 if (rtnl_phys_switch_id_fill(skb, dev))
1781 goto nla_put_failure;
1783 if (rtnl_fill_stats(skb, dev))
1784 goto nla_put_failure;
1786 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1787 goto nla_put_failure;
1789 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1790 goto nla_put_failure;
1792 if (rtnl_xdp_fill(skb, dev))
1793 goto nla_put_failure;
1795 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1796 if (rtnl_link_fill(skb, dev) < 0)
1797 goto nla_put_failure;
1800 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1801 goto nla_put_failure;
1804 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1805 goto nla_put_failure;
1807 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1808 goto nla_put_failure;
1810 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1811 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1812 goto nla_put_failure;
1815 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1816 goto nla_put_failure_rcu;
1819 if (rtnl_fill_prop_list(skb, dev))
1820 goto nla_put_failure;
1822 nlmsg_end(skb, nlh);
1825 nla_put_failure_rcu:
1828 nlmsg_cancel(skb, nlh);
1832 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1833 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1834 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1835 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1836 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1837 [IFLA_MTU] = { .type = NLA_U32 },
1838 [IFLA_LINK] = { .type = NLA_U32 },
1839 [IFLA_MASTER] = { .type = NLA_U32 },
1840 [IFLA_CARRIER] = { .type = NLA_U8 },
1841 [IFLA_TXQLEN] = { .type = NLA_U32 },
1842 [IFLA_WEIGHT] = { .type = NLA_U32 },
1843 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1844 [IFLA_LINKMODE] = { .type = NLA_U8 },
1845 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1846 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1847 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1848 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1849 * allow 0-length string (needed to remove an alias).
1851 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1852 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1853 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1854 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1855 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1856 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1857 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1858 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1859 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1860 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1861 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1862 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1863 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1864 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1865 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1866 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1867 [IFLA_XDP] = { .type = NLA_NESTED },
1868 [IFLA_EVENT] = { .type = NLA_U32 },
1869 [IFLA_GROUP] = { .type = NLA_U32 },
1870 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
1871 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1872 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1873 [IFLA_MIN_MTU] = { .type = NLA_U32 },
1874 [IFLA_MAX_MTU] = { .type = NLA_U32 },
1875 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
1876 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
1877 .len = ALTIFNAMSIZ - 1 },
1878 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
1879 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1882 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1883 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1884 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1885 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1886 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1889 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1890 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1891 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1892 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1893 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1894 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1895 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1896 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1897 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1898 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1899 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1900 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1901 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1902 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1905 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1906 [IFLA_PORT_VF] = { .type = NLA_U32 },
1907 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1908 .len = PORT_PROFILE_MAX },
1909 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1910 .len = PORT_UUID_MAX },
1911 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1912 .len = PORT_UUID_MAX },
1913 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1914 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1916 /* Unused, but we need to keep it here since user space could
1917 * fill it. It's also broken with regard to NLA_BINARY use in
1918 * combination with structs.
1920 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1921 .len = sizeof(struct ifla_port_vsi) },
1924 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1925 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1926 [IFLA_XDP_FD] = { .type = NLA_S32 },
1927 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
1928 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1929 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1930 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1933 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1935 const struct rtnl_link_ops *ops = NULL;
1936 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1938 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1941 if (linfo[IFLA_INFO_KIND]) {
1942 char kind[MODULE_NAME_LEN];
1944 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1945 ops = rtnl_link_ops_get(kind);
1951 static bool link_master_filtered(struct net_device *dev, int master_idx)
1953 struct net_device *master;
1958 master = netdev_master_upper_dev_get(dev);
1959 if (!master || master->ifindex != master_idx)
1965 static bool link_kind_filtered(const struct net_device *dev,
1966 const struct rtnl_link_ops *kind_ops)
1968 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1974 static bool link_dump_filtered(struct net_device *dev,
1976 const struct rtnl_link_ops *kind_ops)
1978 if (link_master_filtered(dev, master_idx) ||
1979 link_kind_filtered(dev, kind_ops))
1986 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1987 * @sk: netlink socket
1988 * @netnsid: network namespace identifier
1990 * Returns the network namespace identified by netnsid on success or an error
1991 * pointer on failure.
1993 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1997 net = get_net_ns_by_id(sock_net(sk), netnsid);
1999 return ERR_PTR(-EINVAL);
2001 /* For now, the caller is required to have CAP_NET_ADMIN in
2002 * the user namespace owning the target net ns.
2004 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2006 return ERR_PTR(-EACCES);
2010 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2012 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2013 bool strict_check, struct nlattr **tb,
2014 struct netlink_ext_ack *extack)
2019 struct ifinfomsg *ifm;
2021 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2022 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2026 ifm = nlmsg_data(nlh);
2027 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2029 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2032 if (ifm->ifi_index) {
2033 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2037 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2038 IFLA_MAX, ifla_policy,
2042 /* A hack to preserve kernel<->userspace interface.
2043 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2044 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2045 * what iproute2 < v3.9.0 used.
2046 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2047 * attribute, its netlink message is shorter than struct ifinfomsg.
2049 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2050 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2052 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2056 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2058 struct netlink_ext_ack *extack = cb->extack;
2059 const struct nlmsghdr *nlh = cb->nlh;
2060 struct net *net = sock_net(skb->sk);
2061 struct net *tgt_net = net;
2064 struct net_device *dev;
2065 struct hlist_head *head;
2066 struct nlattr *tb[IFLA_MAX+1];
2067 u32 ext_filter_mask = 0;
2068 const struct rtnl_link_ops *kind_ops = NULL;
2069 unsigned int flags = NLM_F_MULTI;
2075 s_idx = cb->args[1];
2077 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2079 if (cb->strict_check)
2085 for (i = 0; i <= IFLA_MAX; ++i) {
2089 /* new attributes should only be added with strict checking */
2091 case IFLA_TARGET_NETNSID:
2092 netnsid = nla_get_s32(tb[i]);
2093 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2094 if (IS_ERR(tgt_net)) {
2095 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2096 return PTR_ERR(tgt_net);
2100 ext_filter_mask = nla_get_u32(tb[i]);
2103 master_idx = nla_get_u32(tb[i]);
2106 kind_ops = linkinfo_to_kind_ops(tb[i]);
2109 if (cb->strict_check) {
2110 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2116 if (master_idx || kind_ops)
2117 flags |= NLM_F_DUMP_FILTERED;
2120 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2122 head = &tgt_net->dev_index_head[h];
2123 hlist_for_each_entry(dev, head, index_hlist) {
2124 if (link_dump_filtered(dev, master_idx, kind_ops))
2128 err = rtnl_fill_ifinfo(skb, dev, net,
2130 NETLINK_CB(cb->skb).portid,
2131 nlh->nlmsg_seq, 0, flags,
2132 ext_filter_mask, 0, NULL, 0,
2133 netnsid, GFP_KERNEL);
2136 if (likely(skb->len))
2150 cb->seq = tgt_net->dev_base_seq;
2151 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2158 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2159 struct netlink_ext_ack *exterr)
2161 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2164 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2166 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2169 /* Examine the link attributes and figure out which
2170 * network namespace we are talking about.
2172 if (tb[IFLA_NET_NS_PID])
2173 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2174 else if (tb[IFLA_NET_NS_FD])
2175 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2177 net = get_net(src_net);
2180 EXPORT_SYMBOL(rtnl_link_get_net);
2182 /* Figure out which network namespace we are talking about by
2183 * examining the link attributes in the following order:
2185 * 1. IFLA_NET_NS_PID
2187 * 3. IFLA_TARGET_NETNSID
2189 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2190 struct nlattr *tb[])
2194 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2195 return rtnl_link_get_net(src_net, tb);
2197 if (!tb[IFLA_TARGET_NETNSID])
2198 return get_net(src_net);
2200 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2202 return ERR_PTR(-EINVAL);
2207 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2208 struct net *src_net,
2209 struct nlattr *tb[], int cap)
2213 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2217 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2219 return ERR_PTR(-EPERM);
2225 /* Verify that rtnetlink requests do not pass additional properties
2226 * potentially referring to different network namespaces.
2228 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2229 struct netlink_ext_ack *extack,
2233 if (netns_id_only) {
2234 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2237 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2241 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2244 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2247 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2253 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2257 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2260 if (tb[IFLA_ADDRESS] &&
2261 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2264 if (tb[IFLA_BROADCAST] &&
2265 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2269 if (tb[IFLA_AF_SPEC]) {
2273 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2274 const struct rtnl_af_ops *af_ops;
2277 af_ops = rtnl_af_lookup(nla_type(af));
2280 return -EAFNOSUPPORT;
2283 if (!af_ops->set_link_af) {
2288 if (af_ops->validate_link_af) {
2289 err = af_ops->validate_link_af(dev, af);
2303 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2306 const struct net_device_ops *ops = dev->netdev_ops;
2308 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2311 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2313 if (dev->type != ARPHRD_INFINIBAND)
2316 return handle_infiniband_guid(dev, ivt, guid_type);
2319 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2321 const struct net_device_ops *ops = dev->netdev_ops;
2324 if (tb[IFLA_VF_MAC]) {
2325 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2327 if (ivm->vf >= INT_MAX)
2330 if (ops->ndo_set_vf_mac)
2331 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2337 if (tb[IFLA_VF_VLAN]) {
2338 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2340 if (ivv->vf >= INT_MAX)
2343 if (ops->ndo_set_vf_vlan)
2344 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2346 htons(ETH_P_8021Q));
2351 if (tb[IFLA_VF_VLAN_LIST]) {
2352 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2353 struct nlattr *attr;
2357 if (!ops->ndo_set_vf_vlan)
2360 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2361 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2362 nla_len(attr) < NLA_HDRLEN) {
2365 if (len >= MAX_VLAN_LIST_LEN)
2367 ivvl[len] = nla_data(attr);
2374 if (ivvl[0]->vf >= INT_MAX)
2376 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2377 ivvl[0]->qos, ivvl[0]->vlan_proto);
2382 if (tb[IFLA_VF_TX_RATE]) {
2383 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2384 struct ifla_vf_info ivf;
2386 if (ivt->vf >= INT_MAX)
2389 if (ops->ndo_get_vf_config)
2390 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2395 if (ops->ndo_set_vf_rate)
2396 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2403 if (tb[IFLA_VF_RATE]) {
2404 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2406 if (ivt->vf >= INT_MAX)
2409 if (ops->ndo_set_vf_rate)
2410 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2417 if (tb[IFLA_VF_SPOOFCHK]) {
2418 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2420 if (ivs->vf >= INT_MAX)
2423 if (ops->ndo_set_vf_spoofchk)
2424 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2430 if (tb[IFLA_VF_LINK_STATE]) {
2431 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2433 if (ivl->vf >= INT_MAX)
2436 if (ops->ndo_set_vf_link_state)
2437 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2443 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2444 struct ifla_vf_rss_query_en *ivrssq_en;
2447 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2448 if (ivrssq_en->vf >= INT_MAX)
2450 if (ops->ndo_set_vf_rss_query_en)
2451 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2452 ivrssq_en->setting);
2457 if (tb[IFLA_VF_TRUST]) {
2458 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2460 if (ivt->vf >= INT_MAX)
2463 if (ops->ndo_set_vf_trust)
2464 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2469 if (tb[IFLA_VF_IB_NODE_GUID]) {
2470 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2472 if (ivt->vf >= INT_MAX)
2474 if (!ops->ndo_set_vf_guid)
2476 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2479 if (tb[IFLA_VF_IB_PORT_GUID]) {
2480 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2482 if (ivt->vf >= INT_MAX)
2484 if (!ops->ndo_set_vf_guid)
2487 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2493 static int do_set_master(struct net_device *dev, int ifindex,
2494 struct netlink_ext_ack *extack)
2496 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2497 const struct net_device_ops *ops;
2501 if (upper_dev->ifindex == ifindex)
2503 ops = upper_dev->netdev_ops;
2504 if (ops->ndo_del_slave) {
2505 err = ops->ndo_del_slave(upper_dev, dev);
2514 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2517 ops = upper_dev->netdev_ops;
2518 if (ops->ndo_add_slave) {
2519 err = ops->ndo_add_slave(upper_dev, dev, extack);
2529 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2530 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2531 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2534 static int do_set_proto_down(struct net_device *dev,
2535 struct nlattr *nl_proto_down,
2536 struct nlattr *nl_proto_down_reason,
2537 struct netlink_ext_ack *extack)
2539 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2540 const struct net_device_ops *ops = dev->netdev_ops;
2541 unsigned long mask = 0;
2546 if (!ops->ndo_change_proto_down) {
2547 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2551 if (nl_proto_down_reason) {
2552 err = nla_parse_nested_deprecated(pdreason,
2553 IFLA_PROTO_DOWN_REASON_MAX,
2554 nl_proto_down_reason,
2555 ifla_proto_down_reason_policy,
2560 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2561 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2565 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2567 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2568 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2570 dev_change_proto_down_reason(dev, mask, value);
2573 if (nl_proto_down) {
2574 proto_down = nla_get_u8(nl_proto_down);
2576 /* Dont turn off protodown if there are active reasons */
2577 if (!proto_down && dev->proto_down_reason) {
2578 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2581 err = dev_change_proto_down(dev,
2590 #define DO_SETLINK_MODIFIED 0x01
2591 /* notify flag means notify + modified. */
2592 #define DO_SETLINK_NOTIFY 0x03
2593 static int do_setlink(const struct sk_buff *skb,
2594 struct net_device *dev, struct ifinfomsg *ifm,
2595 struct netlink_ext_ack *extack,
2596 struct nlattr **tb, char *ifname, int status)
2598 const struct net_device_ops *ops = dev->netdev_ops;
2601 err = validate_linkmsg(dev, tb);
2605 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2606 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2613 err = dev_change_net_namespace(dev, net, ifname);
2617 status |= DO_SETLINK_MODIFIED;
2621 struct rtnl_link_ifmap *u_map;
2624 if (!ops->ndo_set_config) {
2629 if (!netif_device_present(dev)) {
2634 u_map = nla_data(tb[IFLA_MAP]);
2635 k_map.mem_start = (unsigned long) u_map->mem_start;
2636 k_map.mem_end = (unsigned long) u_map->mem_end;
2637 k_map.base_addr = (unsigned short) u_map->base_addr;
2638 k_map.irq = (unsigned char) u_map->irq;
2639 k_map.dma = (unsigned char) u_map->dma;
2640 k_map.port = (unsigned char) u_map->port;
2642 err = ops->ndo_set_config(dev, &k_map);
2646 status |= DO_SETLINK_NOTIFY;
2649 if (tb[IFLA_ADDRESS]) {
2650 struct sockaddr *sa;
2653 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2655 sa = kmalloc(len, GFP_KERNEL);
2660 sa->sa_family = dev->type;
2661 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2663 err = dev_set_mac_address_user(dev, sa, extack);
2667 status |= DO_SETLINK_MODIFIED;
2671 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2674 status |= DO_SETLINK_MODIFIED;
2677 if (tb[IFLA_GROUP]) {
2678 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2679 status |= DO_SETLINK_NOTIFY;
2683 * Interface selected by interface index but interface
2684 * name provided implies that a name change has been
2687 if (ifm->ifi_index > 0 && ifname[0]) {
2688 err = dev_change_name(dev, ifname);
2691 status |= DO_SETLINK_MODIFIED;
2694 if (tb[IFLA_IFALIAS]) {
2695 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2696 nla_len(tb[IFLA_IFALIAS]));
2699 status |= DO_SETLINK_NOTIFY;
2702 if (tb[IFLA_BROADCAST]) {
2703 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2704 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2707 if (ifm->ifi_flags || ifm->ifi_change) {
2708 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2714 if (tb[IFLA_MASTER]) {
2715 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2718 status |= DO_SETLINK_MODIFIED;
2721 if (tb[IFLA_CARRIER]) {
2722 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2725 status |= DO_SETLINK_MODIFIED;
2728 if (tb[IFLA_TXQLEN]) {
2729 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2731 err = dev_change_tx_queue_len(dev, value);
2734 status |= DO_SETLINK_MODIFIED;
2737 if (tb[IFLA_GSO_MAX_SIZE]) {
2738 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2740 if (max_size > GSO_MAX_SIZE) {
2745 if (dev->gso_max_size ^ max_size) {
2746 netif_set_gso_max_size(dev, max_size);
2747 status |= DO_SETLINK_MODIFIED;
2751 if (tb[IFLA_GSO_MAX_SEGS]) {
2752 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2754 if (max_segs > GSO_MAX_SEGS) {
2759 if (dev->gso_max_segs ^ max_segs) {
2760 dev->gso_max_segs = max_segs;
2761 status |= DO_SETLINK_MODIFIED;
2765 if (tb[IFLA_OPERSTATE])
2766 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2768 if (tb[IFLA_LINKMODE]) {
2769 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2771 write_lock_bh(&dev_base_lock);
2772 if (dev->link_mode ^ value)
2773 status |= DO_SETLINK_NOTIFY;
2774 dev->link_mode = value;
2775 write_unlock_bh(&dev_base_lock);
2778 if (tb[IFLA_VFINFO_LIST]) {
2779 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2780 struct nlattr *attr;
2783 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2784 if (nla_type(attr) != IFLA_VF_INFO ||
2785 nla_len(attr) < NLA_HDRLEN) {
2789 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2795 err = do_setvfinfo(dev, vfinfo);
2798 status |= DO_SETLINK_NOTIFY;
2803 if (tb[IFLA_VF_PORTS]) {
2804 struct nlattr *port[IFLA_PORT_MAX+1];
2805 struct nlattr *attr;
2810 if (!ops->ndo_set_vf_port)
2813 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2814 if (nla_type(attr) != IFLA_VF_PORT ||
2815 nla_len(attr) < NLA_HDRLEN) {
2819 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2825 if (!port[IFLA_PORT_VF]) {
2829 vf = nla_get_u32(port[IFLA_PORT_VF]);
2830 err = ops->ndo_set_vf_port(dev, vf, port);
2833 status |= DO_SETLINK_NOTIFY;
2838 if (tb[IFLA_PORT_SELF]) {
2839 struct nlattr *port[IFLA_PORT_MAX+1];
2841 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2843 ifla_port_policy, NULL);
2848 if (ops->ndo_set_vf_port)
2849 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2852 status |= DO_SETLINK_NOTIFY;
2855 if (tb[IFLA_AF_SPEC]) {
2859 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2860 const struct rtnl_af_ops *af_ops;
2864 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2866 err = af_ops->set_link_af(dev, af, extack);
2873 status |= DO_SETLINK_NOTIFY;
2878 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2879 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2880 tb[IFLA_PROTO_DOWN_REASON], extack);
2883 status |= DO_SETLINK_NOTIFY;
2887 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2890 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2892 ifla_xdp_policy, NULL);
2896 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2901 if (xdp[IFLA_XDP_FLAGS]) {
2902 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2903 if (xdp_flags & ~XDP_FLAGS_MASK) {
2907 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2913 if (xdp[IFLA_XDP_FD]) {
2914 int expected_fd = -1;
2916 if (xdp_flags & XDP_FLAGS_REPLACE) {
2917 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2922 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2925 err = dev_change_xdp_fd(dev, extack,
2926 nla_get_s32(xdp[IFLA_XDP_FD]),
2931 status |= DO_SETLINK_NOTIFY;
2936 if (status & DO_SETLINK_MODIFIED) {
2937 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2938 netdev_state_change(dev);
2941 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",
2948 static struct net_device *rtnl_dev_get(struct net *net,
2949 struct nlattr *ifname_attr,
2950 struct nlattr *altifname_attr,
2953 char buffer[ALTIFNAMSIZ];
2958 nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2959 else if (altifname_attr)
2960 nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2965 return __dev_get_by_name(net, ifname);
2968 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2969 struct netlink_ext_ack *extack)
2971 struct net *net = sock_net(skb->sk);
2972 struct ifinfomsg *ifm;
2973 struct net_device *dev;
2975 struct nlattr *tb[IFLA_MAX+1];
2976 char ifname[IFNAMSIZ];
2978 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2979 ifla_policy, extack);
2983 err = rtnl_ensure_unique_netns(tb, extack, false);
2987 if (tb[IFLA_IFNAME])
2988 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2993 ifm = nlmsg_data(nlh);
2994 if (ifm->ifi_index > 0)
2995 dev = __dev_get_by_index(net, ifm->ifi_index);
2996 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2997 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3006 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3011 static int rtnl_group_dellink(const struct net *net, int group)
3013 struct net_device *dev, *aux;
3014 LIST_HEAD(list_kill);
3020 for_each_netdev(net, dev) {
3021 if (dev->group == group) {
3022 const struct rtnl_link_ops *ops;
3025 ops = dev->rtnl_link_ops;
3026 if (!ops || !ops->dellink)
3034 for_each_netdev_safe(net, dev, aux) {
3035 if (dev->group == group) {
3036 const struct rtnl_link_ops *ops;
3038 ops = dev->rtnl_link_ops;
3039 ops->dellink(dev, &list_kill);
3042 unregister_netdevice_many(&list_kill);
3047 int rtnl_delete_link(struct net_device *dev)
3049 const struct rtnl_link_ops *ops;
3050 LIST_HEAD(list_kill);
3052 ops = dev->rtnl_link_ops;
3053 if (!ops || !ops->dellink)
3056 ops->dellink(dev, &list_kill);
3057 unregister_netdevice_many(&list_kill);
3061 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3063 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3064 struct netlink_ext_ack *extack)
3066 struct net *net = sock_net(skb->sk);
3067 struct net *tgt_net = net;
3068 struct net_device *dev = NULL;
3069 struct ifinfomsg *ifm;
3070 struct nlattr *tb[IFLA_MAX+1];
3074 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3075 ifla_policy, extack);
3079 err = rtnl_ensure_unique_netns(tb, extack, true);
3083 if (tb[IFLA_TARGET_NETNSID]) {
3084 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3085 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3086 if (IS_ERR(tgt_net))
3087 return PTR_ERR(tgt_net);
3091 ifm = nlmsg_data(nlh);
3092 if (ifm->ifi_index > 0)
3093 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3094 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3095 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3096 tb[IFLA_ALT_IFNAME], NULL);
3097 else if (tb[IFLA_GROUP])
3098 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3103 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3109 err = rtnl_delete_link(dev);
3118 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3120 unsigned int old_flags;
3123 old_flags = dev->flags;
3124 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3125 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3131 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3132 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3134 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3135 __dev_notify_flags(dev, old_flags, ~0U);
3139 EXPORT_SYMBOL(rtnl_configure_link);
3141 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3142 unsigned char name_assign_type,
3143 const struct rtnl_link_ops *ops,
3144 struct nlattr *tb[],
3145 struct netlink_ext_ack *extack)
3147 struct net_device *dev;
3148 unsigned int num_tx_queues = 1;
3149 unsigned int num_rx_queues = 1;
3151 if (tb[IFLA_NUM_TX_QUEUES])
3152 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3153 else if (ops->get_num_tx_queues)
3154 num_tx_queues = ops->get_num_tx_queues();
3156 if (tb[IFLA_NUM_RX_QUEUES])
3157 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3158 else if (ops->get_num_rx_queues)
3159 num_rx_queues = ops->get_num_rx_queues();
3161 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3162 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3163 return ERR_PTR(-EINVAL);
3166 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3167 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3168 return ERR_PTR(-EINVAL);
3171 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3172 ops->setup, num_tx_queues, num_rx_queues);
3174 return ERR_PTR(-ENOMEM);
3176 dev_net_set(dev, net);
3177 dev->rtnl_link_ops = ops;
3178 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3181 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3184 err = dev_validate_mtu(dev, mtu, extack);
3187 return ERR_PTR(err);
3191 if (tb[IFLA_ADDRESS]) {
3192 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3193 nla_len(tb[IFLA_ADDRESS]));
3194 dev->addr_assign_type = NET_ADDR_SET;
3196 if (tb[IFLA_BROADCAST])
3197 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3198 nla_len(tb[IFLA_BROADCAST]));
3199 if (tb[IFLA_TXQLEN])
3200 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3201 if (tb[IFLA_OPERSTATE])
3202 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3203 if (tb[IFLA_LINKMODE])
3204 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3206 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3207 if (tb[IFLA_GSO_MAX_SIZE])
3208 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3209 if (tb[IFLA_GSO_MAX_SEGS])
3210 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3214 EXPORT_SYMBOL(rtnl_create_link);
3216 static int rtnl_group_changelink(const struct sk_buff *skb,
3217 struct net *net, int group,
3218 struct ifinfomsg *ifm,
3219 struct netlink_ext_ack *extack,
3222 struct net_device *dev, *aux;
3225 for_each_netdev_safe(net, dev, aux) {
3226 if (dev->group == group) {
3227 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3236 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3237 struct nlattr **attr, struct netlink_ext_ack *extack)
3239 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3240 unsigned char name_assign_type = NET_NAME_USER;
3241 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3242 const struct rtnl_link_ops *m_ops = NULL;
3243 struct net_device *master_dev = NULL;
3244 struct net *net = sock_net(skb->sk);
3245 const struct rtnl_link_ops *ops;
3246 struct nlattr *tb[IFLA_MAX + 1];
3247 struct net *dest_net, *link_net;
3248 struct nlattr **slave_data;
3249 char kind[MODULE_NAME_LEN];
3250 struct net_device *dev;
3251 struct ifinfomsg *ifm;
3252 char ifname[IFNAMSIZ];
3253 struct nlattr **data;
3256 #ifdef CONFIG_MODULES
3259 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3260 ifla_policy, extack);
3264 err = rtnl_ensure_unique_netns(tb, extack, false);
3268 if (tb[IFLA_IFNAME])
3269 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3273 ifm = nlmsg_data(nlh);
3274 if (ifm->ifi_index > 0)
3275 dev = __dev_get_by_index(net, ifm->ifi_index);
3276 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3277 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3282 master_dev = netdev_master_upper_dev_get(dev);
3284 m_ops = master_dev->rtnl_link_ops;
3287 err = validate_linkmsg(dev, tb);
3291 if (tb[IFLA_LINKINFO]) {
3292 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3294 ifla_info_policy, NULL);
3298 memset(linkinfo, 0, sizeof(linkinfo));
3300 if (linkinfo[IFLA_INFO_KIND]) {
3301 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3302 ops = rtnl_link_ops_get(kind);
3310 if (ops->maxtype > RTNL_MAX_TYPE)
3313 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3314 err = nla_parse_nested_deprecated(attr, ops->maxtype,
3315 linkinfo[IFLA_INFO_DATA],
3316 ops->policy, extack);
3321 if (ops->validate) {
3322 err = ops->validate(tb, data, extack);
3330 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3333 if (m_ops->slave_maxtype &&
3334 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3335 err = nla_parse_nested_deprecated(slave_attr,
3336 m_ops->slave_maxtype,
3337 linkinfo[IFLA_INFO_SLAVE_DATA],
3338 m_ops->slave_policy,
3342 slave_data = slave_attr;
3349 if (nlh->nlmsg_flags & NLM_F_EXCL)
3351 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3354 if (linkinfo[IFLA_INFO_DATA]) {
3355 if (!ops || ops != dev->rtnl_link_ops ||
3359 err = ops->changelink(dev, tb, data, extack);
3362 status |= DO_SETLINK_NOTIFY;
3365 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3366 if (!m_ops || !m_ops->slave_changelink)
3369 err = m_ops->slave_changelink(master_dev, dev, tb,
3370 slave_data, extack);
3373 status |= DO_SETLINK_NOTIFY;
3376 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3379 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3380 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3381 return rtnl_group_changelink(skb, net,
3382 nla_get_u32(tb[IFLA_GROUP]),
3387 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3391 #ifdef CONFIG_MODULES
3394 request_module("rtnl-link-%s", kind);
3396 ops = rtnl_link_ops_get(kind);
3401 NL_SET_ERR_MSG(extack, "Unknown device type");
3409 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3410 name_assign_type = NET_NAME_ENUM;
3413 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3414 if (IS_ERR(dest_net))
3415 return PTR_ERR(dest_net);
3417 if (tb[IFLA_LINK_NETNSID]) {
3418 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3420 link_net = get_net_ns_by_id(dest_net, id);
3422 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3427 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3433 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3434 name_assign_type, ops, tb, extack);
3440 dev->ifindex = ifm->ifi_index;
3443 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3445 err = register_netdevice(dev);
3451 err = rtnl_configure_link(dev, ifm);
3453 goto out_unregister;
3455 err = dev_change_net_namespace(dev, dest_net, ifname);
3457 goto out_unregister;
3459 if (tb[IFLA_MASTER]) {
3460 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3462 goto out_unregister;
3471 LIST_HEAD(list_kill);
3473 ops->dellink(dev, &list_kill);
3474 unregister_netdevice_many(&list_kill);
3476 unregister_netdevice(dev);
3481 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3482 struct netlink_ext_ack *extack)
3484 struct nlattr **attr;
3487 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3491 ret = __rtnl_newlink(skb, nlh, attr, extack);
3496 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3497 const struct nlmsghdr *nlh,
3499 struct netlink_ext_ack *extack)
3501 struct ifinfomsg *ifm;
3504 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3505 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3509 if (!netlink_strict_get_check(skb))
3510 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3511 ifla_policy, extack);
3513 ifm = nlmsg_data(nlh);
3514 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3516 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3520 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3521 ifla_policy, extack);
3525 for (i = 0; i <= IFLA_MAX; i++) {
3531 case IFLA_ALT_IFNAME:
3533 case IFLA_TARGET_NETNSID:
3536 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3544 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3545 struct netlink_ext_ack *extack)
3547 struct net *net = sock_net(skb->sk);
3548 struct net *tgt_net = net;
3549 struct ifinfomsg *ifm;
3550 struct nlattr *tb[IFLA_MAX+1];
3551 struct net_device *dev = NULL;
3552 struct sk_buff *nskb;
3555 u32 ext_filter_mask = 0;
3557 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3561 err = rtnl_ensure_unique_netns(tb, extack, true);
3565 if (tb[IFLA_TARGET_NETNSID]) {
3566 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3567 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3568 if (IS_ERR(tgt_net))
3569 return PTR_ERR(tgt_net);
3572 if (tb[IFLA_EXT_MASK])
3573 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3576 ifm = nlmsg_data(nlh);
3577 if (ifm->ifi_index > 0)
3578 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3579 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3580 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3581 tb[IFLA_ALT_IFNAME], NULL);
3590 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3594 err = rtnl_fill_ifinfo(nskb, dev, net,
3595 RTM_NEWLINK, NETLINK_CB(skb).portid,
3596 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3597 0, NULL, 0, netnsid, GFP_KERNEL);
3599 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3600 WARN_ON(err == -EMSGSIZE);
3603 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3611 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3612 bool *changed, struct netlink_ext_ack *extack)
3617 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3621 alt_ifname = nla_strdup(attr, GFP_KERNEL);
3625 if (cmd == RTM_NEWLINKPROP) {
3626 err = netdev_name_node_alt_create(dev, alt_ifname);
3629 } else if (cmd == RTM_DELLINKPROP) {
3630 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3642 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3643 struct netlink_ext_ack *extack)
3645 struct net *net = sock_net(skb->sk);
3646 struct nlattr *tb[IFLA_MAX + 1];
3647 struct net_device *dev;
3648 struct ifinfomsg *ifm;
3649 bool changed = false;
3650 struct nlattr *attr;
3653 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3657 err = rtnl_ensure_unique_netns(tb, extack, true);
3661 ifm = nlmsg_data(nlh);
3662 if (ifm->ifi_index > 0)
3663 dev = __dev_get_by_index(net, ifm->ifi_index);
3664 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3665 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3666 tb[IFLA_ALT_IFNAME], NULL);
3673 if (!tb[IFLA_PROP_LIST])
3676 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3677 switch (nla_type(attr)) {
3678 case IFLA_ALT_IFNAME:
3679 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3687 netdev_state_change(dev);
3691 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3692 struct netlink_ext_ack *extack)
3694 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3697 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3698 struct netlink_ext_ack *extack)
3700 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3703 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3705 struct net *net = sock_net(skb->sk);
3706 size_t min_ifinfo_dump_size = 0;
3707 struct nlattr *tb[IFLA_MAX+1];
3708 u32 ext_filter_mask = 0;
3709 struct net_device *dev;
3712 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3713 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3714 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3716 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3717 if (tb[IFLA_EXT_MASK])
3718 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3721 if (!ext_filter_mask)
3722 return NLMSG_GOODSIZE;
3724 * traverse the list of net devices and compute the minimum
3725 * buffer size based upon the filter mask.
3728 for_each_netdev_rcu(net, dev) {
3729 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3730 if_nlmsg_size(dev, ext_filter_mask));
3734 return nlmsg_total_size(min_ifinfo_dump_size);
3737 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3740 int s_idx = cb->family;
3741 int type = cb->nlh->nlmsg_type - RTM_BASE;
3747 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3748 struct rtnl_link __rcu **tab;
3749 struct rtnl_link *link;
3750 rtnl_dumpit_func dumpit;
3752 if (idx < s_idx || idx == PF_PACKET)
3755 if (type < 0 || type >= RTM_NR_MSGTYPES)
3758 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3762 link = rcu_dereference_rtnl(tab[type]);
3766 dumpit = link->dumpit;
3771 memset(&cb->args[0], 0, sizeof(cb->args));
3775 ret = dumpit(skb, cb);
3781 return skb->len ? : ret;
3784 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3785 unsigned int change,
3786 u32 event, gfp_t flags, int *new_nsid,
3789 struct net *net = dev_net(dev);
3790 struct sk_buff *skb;
3792 size_t if_info_size;
3794 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3798 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3799 type, 0, 0, change, 0, 0, event,
3800 new_nsid, new_ifindex, -1, flags);
3802 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3803 WARN_ON(err == -EMSGSIZE);
3810 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3814 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3816 struct net *net = dev_net(dev);
3818 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3821 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3822 unsigned int change, u32 event,
3823 gfp_t flags, int *new_nsid, int new_ifindex)
3825 struct sk_buff *skb;
3827 if (dev->reg_state != NETREG_REGISTERED)
3830 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3833 rtmsg_ifinfo_send(skb, dev, flags);
3836 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3839 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3843 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3844 gfp_t flags, int *new_nsid, int new_ifindex)
3846 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3847 new_nsid, new_ifindex);
3850 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3851 struct net_device *dev,
3852 u8 *addr, u16 vid, u32 pid, u32 seq,
3853 int type, unsigned int flags,
3854 int nlflags, u16 ndm_state)
3856 struct nlmsghdr *nlh;
3859 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3863 ndm = nlmsg_data(nlh);
3864 ndm->ndm_family = AF_BRIDGE;
3867 ndm->ndm_flags = flags;
3869 ndm->ndm_ifindex = dev->ifindex;
3870 ndm->ndm_state = ndm_state;
3872 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3873 goto nla_put_failure;
3875 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3876 goto nla_put_failure;
3878 nlmsg_end(skb, nlh);
3882 nlmsg_cancel(skb, nlh);
3886 static inline size_t rtnl_fdb_nlmsg_size(void)
3888 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3889 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3890 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3894 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3897 struct net *net = dev_net(dev);
3898 struct sk_buff *skb;
3901 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3905 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3906 0, 0, type, NTF_SELF, 0, ndm_state);
3912 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3915 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3919 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3921 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3922 struct nlattr *tb[],
3923 struct net_device *dev,
3924 const unsigned char *addr, u16 vid,
3929 /* If aging addresses are supported device will need to
3930 * implement its own handler for this.
3932 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3933 pr_info("%s: FDB only supports static addresses\n", dev->name);
3938 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3942 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3943 err = dev_uc_add_excl(dev, addr);
3944 else if (is_multicast_ether_addr(addr))
3945 err = dev_mc_add_excl(dev, addr);
3947 /* Only return duplicate errors if NLM_F_EXCL is set */
3948 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3953 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3955 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3956 struct netlink_ext_ack *extack)
3961 if (nla_len(vlan_attr) != sizeof(u16)) {
3962 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3966 vid = nla_get_u16(vlan_attr);
3968 if (!vid || vid >= VLAN_VID_MASK) {
3969 NL_SET_ERR_MSG(extack, "invalid vlan id");
3977 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3978 struct netlink_ext_ack *extack)
3980 struct net *net = sock_net(skb->sk);
3982 struct nlattr *tb[NDA_MAX+1];
3983 struct net_device *dev;
3988 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3993 ndm = nlmsg_data(nlh);
3994 if (ndm->ndm_ifindex == 0) {
3995 NL_SET_ERR_MSG(extack, "invalid ifindex");
3999 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4001 NL_SET_ERR_MSG(extack, "unknown ifindex");
4005 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4006 NL_SET_ERR_MSG(extack, "invalid address");
4010 if (dev->type != ARPHRD_ETHER) {
4011 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4015 addr = nla_data(tb[NDA_LLADDR]);
4017 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4023 /* Support fdb on master device the net/bridge default case */
4024 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4025 netif_is_bridge_port(dev)) {
4026 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4027 const struct net_device_ops *ops = br_dev->netdev_ops;
4029 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4030 nlh->nlmsg_flags, extack);
4034 ndm->ndm_flags &= ~NTF_MASTER;
4037 /* Embedded bridge, macvlan, and any other device support */
4038 if ((ndm->ndm_flags & NTF_SELF)) {
4039 if (dev->netdev_ops->ndo_fdb_add)
4040 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4045 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4049 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4051 ndm->ndm_flags &= ~NTF_SELF;
4059 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4061 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4062 struct nlattr *tb[],
4063 struct net_device *dev,
4064 const unsigned char *addr, u16 vid)
4068 /* If aging addresses are supported device will need to
4069 * implement its own handler for this.
4071 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4072 pr_info("%s: FDB only supports static addresses\n", dev->name);
4076 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4077 err = dev_uc_del(dev, addr);
4078 else if (is_multicast_ether_addr(addr))
4079 err = dev_mc_del(dev, addr);
4083 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4085 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4086 struct netlink_ext_ack *extack)
4088 struct net *net = sock_net(skb->sk);
4090 struct nlattr *tb[NDA_MAX+1];
4091 struct net_device *dev;
4096 if (!netlink_capable(skb, CAP_NET_ADMIN))
4099 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4104 ndm = nlmsg_data(nlh);
4105 if (ndm->ndm_ifindex == 0) {
4106 NL_SET_ERR_MSG(extack, "invalid ifindex");
4110 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4112 NL_SET_ERR_MSG(extack, "unknown ifindex");
4116 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4117 NL_SET_ERR_MSG(extack, "invalid address");
4121 if (dev->type != ARPHRD_ETHER) {
4122 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4126 addr = nla_data(tb[NDA_LLADDR]);
4128 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4134 /* Support fdb on master device the net/bridge default case */
4135 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4136 netif_is_bridge_port(dev)) {
4137 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4138 const struct net_device_ops *ops = br_dev->netdev_ops;
4140 if (ops->ndo_fdb_del)
4141 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4146 ndm->ndm_flags &= ~NTF_MASTER;
4149 /* Embedded bridge, macvlan, and any other device support */
4150 if (ndm->ndm_flags & NTF_SELF) {
4151 if (dev->netdev_ops->ndo_fdb_del)
4152 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4155 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4158 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4160 ndm->ndm_flags &= ~NTF_SELF;
4167 static int nlmsg_populate_fdb(struct sk_buff *skb,
4168 struct netlink_callback *cb,
4169 struct net_device *dev,
4171 struct netdev_hw_addr_list *list)
4173 struct netdev_hw_addr *ha;
4177 portid = NETLINK_CB(cb->skb).portid;
4178 seq = cb->nlh->nlmsg_seq;
4180 list_for_each_entry(ha, &list->list, list) {
4181 if (*idx < cb->args[2])
4184 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4186 RTM_NEWNEIGH, NTF_SELF,
4187 NLM_F_MULTI, NUD_PERMANENT);
4197 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4198 * @skb: socket buffer to store message in
4199 * @cb: netlink callback
4201 * @filter_dev: ignored
4202 * @idx: the number of FDB table entries dumped is added to *@idx
4204 * Default netdevice operation to dump the existing unicast address list.
4205 * Returns number of addresses from list put in skb.
4207 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4208 struct netlink_callback *cb,
4209 struct net_device *dev,
4210 struct net_device *filter_dev,
4215 if (dev->type != ARPHRD_ETHER)
4218 netif_addr_lock_bh(dev);
4219 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4222 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4224 netif_addr_unlock_bh(dev);
4227 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4229 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4230 int *br_idx, int *brport_idx,
4231 struct netlink_ext_ack *extack)
4233 struct nlattr *tb[NDA_MAX + 1];
4237 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4238 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4242 ndm = nlmsg_data(nlh);
4243 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4244 ndm->ndm_flags || ndm->ndm_type) {
4245 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4249 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4250 NDA_MAX, NULL, extack);
4254 *brport_idx = ndm->ndm_ifindex;
4255 for (i = 0; i <= NDA_MAX; ++i) {
4261 if (nla_len(tb[i]) != sizeof(u32)) {
4262 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4265 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4268 if (nla_len(tb[i]) != sizeof(u32)) {
4269 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4272 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4275 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4283 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4284 int *br_idx, int *brport_idx,
4285 struct netlink_ext_ack *extack)
4287 struct nlattr *tb[IFLA_MAX+1];
4290 /* A hack to preserve kernel<->userspace interface.
4291 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4292 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4293 * So, check for ndmsg with an optional u32 attribute (not used here).
4294 * Fortunately these sizes don't conflict with the size of ifinfomsg
4295 * with an optional attribute.
4297 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4298 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4299 nla_attr_size(sizeof(u32)))) {
4300 struct ifinfomsg *ifm;
4302 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4303 tb, IFLA_MAX, ifla_policy,
4307 } else if (err == 0) {
4308 if (tb[IFLA_MASTER])
4309 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4312 ifm = nlmsg_data(nlh);
4313 *brport_idx = ifm->ifi_index;
4318 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4320 struct net_device *dev;
4321 struct net_device *br_dev = NULL;
4322 const struct net_device_ops *ops = NULL;
4323 const struct net_device_ops *cops = NULL;
4324 struct net *net = sock_net(skb->sk);
4325 struct hlist_head *head;
4333 if (cb->strict_check)
4334 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4337 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4343 br_dev = __dev_get_by_index(net, br_idx);
4347 ops = br_dev->netdev_ops;
4351 s_idx = cb->args[1];
4353 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4355 head = &net->dev_index_head[h];
4356 hlist_for_each_entry(dev, head, index_hlist) {
4358 if (brport_idx && (dev->ifindex != brport_idx))
4361 if (!br_idx) { /* user did not specify a specific bridge */
4362 if (netif_is_bridge_port(dev)) {
4363 br_dev = netdev_master_upper_dev_get(dev);
4364 cops = br_dev->netdev_ops;
4367 if (dev != br_dev &&
4368 !netif_is_bridge_port(dev))
4371 if (br_dev != netdev_master_upper_dev_get(dev) &&
4372 !(dev->priv_flags & IFF_EBRIDGE))
4380 if (netif_is_bridge_port(dev)) {
4381 if (cops && cops->ndo_fdb_dump) {
4382 err = cops->ndo_fdb_dump(skb, cb,
4385 if (err == -EMSGSIZE)
4390 if (dev->netdev_ops->ndo_fdb_dump)
4391 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4395 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4397 if (err == -EMSGSIZE)
4402 /* reset fdb offset to 0 for rest of the interfaces */
4418 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4419 struct nlattr **tb, u8 *ndm_flags,
4420 int *br_idx, int *brport_idx, u8 **addr,
4421 u16 *vid, struct netlink_ext_ack *extack)
4426 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4427 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4431 ndm = nlmsg_data(nlh);
4432 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4434 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4438 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4439 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4443 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4444 NDA_MAX, nda_policy, extack);
4448 *ndm_flags = ndm->ndm_flags;
4449 *brport_idx = ndm->ndm_ifindex;
4450 for (i = 0; i <= NDA_MAX; ++i) {
4456 *br_idx = nla_get_u32(tb[i]);
4459 if (nla_len(tb[i]) != ETH_ALEN) {
4460 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4463 *addr = nla_data(tb[i]);
4466 err = fdb_vid_parse(tb[i], vid, extack);
4473 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4481 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4482 struct netlink_ext_ack *extack)
4484 struct net_device *dev = NULL, *br_dev = NULL;
4485 const struct net_device_ops *ops = NULL;
4486 struct net *net = sock_net(in_skb->sk);
4487 struct nlattr *tb[NDA_MAX + 1];
4488 struct sk_buff *skb;
4496 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4497 &brport_idx, &addr, &vid, extack);
4502 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4507 dev = __dev_get_by_index(net, brport_idx);
4509 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4516 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4520 br_dev = __dev_get_by_index(net, br_idx);
4522 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4525 ops = br_dev->netdev_ops;
4529 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4530 if (!netif_is_bridge_port(dev)) {
4531 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4534 br_dev = netdev_master_upper_dev_get(dev);
4536 NL_SET_ERR_MSG(extack, "Master of device not found");
4539 ops = br_dev->netdev_ops;
4541 if (!(ndm_flags & NTF_SELF)) {
4542 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4545 ops = dev->netdev_ops;
4549 if (!br_dev && !dev) {
4550 NL_SET_ERR_MSG(extack, "No device specified");
4554 if (!ops || !ops->ndo_fdb_get) {
4555 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4559 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4565 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4566 NETLINK_CB(in_skb).portid,
4567 nlh->nlmsg_seq, extack);
4571 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4577 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4578 unsigned int attrnum, unsigned int flag)
4581 return nla_put_u8(skb, attrnum, !!(flags & flag));
4585 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4586 struct net_device *dev, u16 mode,
4587 u32 flags, u32 mask, int nlflags,
4589 int (*vlan_fill)(struct sk_buff *skb,
4590 struct net_device *dev,
4593 struct nlmsghdr *nlh;
4594 struct ifinfomsg *ifm;
4595 struct nlattr *br_afspec;
4596 struct nlattr *protinfo;
4597 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4598 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4601 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4605 ifm = nlmsg_data(nlh);
4606 ifm->ifi_family = AF_BRIDGE;
4608 ifm->ifi_type = dev->type;
4609 ifm->ifi_index = dev->ifindex;
4610 ifm->ifi_flags = dev_get_flags(dev);
4611 ifm->ifi_change = 0;
4614 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4615 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4616 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4618 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4620 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4621 (dev->ifindex != dev_get_iflink(dev) &&
4622 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4623 goto nla_put_failure;
4625 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4627 goto nla_put_failure;
4629 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4630 nla_nest_cancel(skb, br_afspec);
4631 goto nla_put_failure;
4634 if (mode != BRIDGE_MODE_UNDEF) {
4635 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4636 nla_nest_cancel(skb, br_afspec);
4637 goto nla_put_failure;
4641 err = vlan_fill(skb, dev, filter_mask);
4643 nla_nest_cancel(skb, br_afspec);
4644 goto nla_put_failure;
4647 nla_nest_end(skb, br_afspec);
4649 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4651 goto nla_put_failure;
4653 if (brport_nla_put_flag(skb, flags, mask,
4654 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4655 brport_nla_put_flag(skb, flags, mask,
4656 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4657 brport_nla_put_flag(skb, flags, mask,
4658 IFLA_BRPORT_FAST_LEAVE,
4659 BR_MULTICAST_FAST_LEAVE) ||
4660 brport_nla_put_flag(skb, flags, mask,
4661 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4662 brport_nla_put_flag(skb, flags, mask,
4663 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4664 brport_nla_put_flag(skb, flags, mask,
4665 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4666 brport_nla_put_flag(skb, flags, mask,
4667 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4668 brport_nla_put_flag(skb, flags, mask,
4669 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4670 brport_nla_put_flag(skb, flags, mask,
4671 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4672 brport_nla_put_flag(skb, flags, mask,
4673 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4674 nla_nest_cancel(skb, protinfo);
4675 goto nla_put_failure;
4678 nla_nest_end(skb, protinfo);
4680 nlmsg_end(skb, nlh);
4683 nlmsg_cancel(skb, nlh);
4684 return err ? err : -EMSGSIZE;
4686 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4688 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4689 bool strict_check, u32 *filter_mask,
4690 struct netlink_ext_ack *extack)
4692 struct nlattr *tb[IFLA_MAX+1];
4696 struct ifinfomsg *ifm;
4698 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4699 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4703 ifm = nlmsg_data(nlh);
4704 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4705 ifm->ifi_change || ifm->ifi_index) {
4706 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4710 err = nlmsg_parse_deprecated_strict(nlh,
4711 sizeof(struct ifinfomsg),
4712 tb, IFLA_MAX, ifla_policy,
4715 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4716 tb, IFLA_MAX, ifla_policy,
4722 /* new attributes should only be added with strict checking */
4723 for (i = 0; i <= IFLA_MAX; ++i) {
4729 *filter_mask = nla_get_u32(tb[i]);
4733 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4742 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4744 const struct nlmsghdr *nlh = cb->nlh;
4745 struct net *net = sock_net(skb->sk);
4746 struct net_device *dev;
4748 u32 portid = NETLINK_CB(cb->skb).portid;
4749 u32 seq = nlh->nlmsg_seq;
4750 u32 filter_mask = 0;
4753 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4755 if (err < 0 && cb->strict_check)
4759 for_each_netdev_rcu(net, dev) {
4760 const struct net_device_ops *ops = dev->netdev_ops;
4761 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4763 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4764 if (idx >= cb->args[0]) {
4765 err = br_dev->netdev_ops->ndo_bridge_getlink(
4766 skb, portid, seq, dev,
4767 filter_mask, NLM_F_MULTI);
4768 if (err < 0 && err != -EOPNOTSUPP) {
4769 if (likely(skb->len))
4778 if (ops->ndo_bridge_getlink) {
4779 if (idx >= cb->args[0]) {
4780 err = ops->ndo_bridge_getlink(skb, portid,
4784 if (err < 0 && err != -EOPNOTSUPP) {
4785 if (likely(skb->len))
4802 static inline size_t bridge_nlmsg_size(void)
4804 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4805 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4806 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4807 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4808 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
4809 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
4810 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
4811 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
4812 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4813 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
4814 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
4817 static int rtnl_bridge_notify(struct net_device *dev)
4819 struct net *net = dev_net(dev);
4820 struct sk_buff *skb;
4821 int err = -EOPNOTSUPP;
4823 if (!dev->netdev_ops->ndo_bridge_getlink)
4826 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4832 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4839 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4842 WARN_ON(err == -EMSGSIZE);
4845 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4849 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4850 struct netlink_ext_ack *extack)
4852 struct net *net = sock_net(skb->sk);
4853 struct ifinfomsg *ifm;
4854 struct net_device *dev;
4855 struct nlattr *br_spec, *attr = NULL;
4856 int rem, err = -EOPNOTSUPP;
4858 bool have_flags = false;
4860 if (nlmsg_len(nlh) < sizeof(*ifm))
4863 ifm = nlmsg_data(nlh);
4864 if (ifm->ifi_family != AF_BRIDGE)
4865 return -EPFNOSUPPORT;
4867 dev = __dev_get_by_index(net, ifm->ifi_index);
4869 NL_SET_ERR_MSG(extack, "unknown ifindex");
4873 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4875 nla_for_each_nested(attr, br_spec, rem) {
4876 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4877 if (nla_len(attr) < sizeof(flags))
4881 flags = nla_get_u16(attr);
4887 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4888 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4890 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4895 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4900 flags &= ~BRIDGE_FLAGS_MASTER;
4903 if ((flags & BRIDGE_FLAGS_SELF)) {
4904 if (!dev->netdev_ops->ndo_bridge_setlink)
4907 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4911 flags &= ~BRIDGE_FLAGS_SELF;
4913 /* Generate event to notify upper layer of bridge
4916 err = rtnl_bridge_notify(dev);
4921 memcpy(nla_data(attr), &flags, sizeof(flags));
4926 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4927 struct netlink_ext_ack *extack)
4929 struct net *net = sock_net(skb->sk);
4930 struct ifinfomsg *ifm;
4931 struct net_device *dev;
4932 struct nlattr *br_spec, *attr = NULL;
4933 int rem, err = -EOPNOTSUPP;
4935 bool have_flags = false;
4937 if (nlmsg_len(nlh) < sizeof(*ifm))
4940 ifm = nlmsg_data(nlh);
4941 if (ifm->ifi_family != AF_BRIDGE)
4942 return -EPFNOSUPPORT;
4944 dev = __dev_get_by_index(net, ifm->ifi_index);
4946 NL_SET_ERR_MSG(extack, "unknown ifindex");
4950 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4952 nla_for_each_nested(attr, br_spec, rem) {
4953 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4954 if (nla_len(attr) < sizeof(flags))
4958 flags = nla_get_u16(attr);
4964 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4965 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4967 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4972 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4976 flags &= ~BRIDGE_FLAGS_MASTER;
4979 if ((flags & BRIDGE_FLAGS_SELF)) {
4980 if (!dev->netdev_ops->ndo_bridge_dellink)
4983 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4987 flags &= ~BRIDGE_FLAGS_SELF;
4989 /* Generate event to notify upper layer of bridge
4992 err = rtnl_bridge_notify(dev);
4997 memcpy(nla_data(attr), &flags, sizeof(flags));
5002 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5004 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5005 (!idxattr || idxattr == attrid);
5008 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5009 static int rtnl_get_offload_stats_attr_size(int attr_id)
5012 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5013 return sizeof(struct rtnl_link_stats64);
5019 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5022 struct nlattr *attr = NULL;
5027 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5028 dev->netdev_ops->ndo_get_offload_stats))
5031 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5032 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5033 if (attr_id < *prividx)
5036 size = rtnl_get_offload_stats_attr_size(attr_id);
5040 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5043 attr = nla_reserve_64bit(skb, attr_id, size,
5044 IFLA_OFFLOAD_XSTATS_UNSPEC);
5046 goto nla_put_failure;
5048 attr_data = nla_data(attr);
5049 memset(attr_data, 0, size);
5050 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5053 goto get_offload_stats_failure;
5064 get_offload_stats_failure:
5069 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5075 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5076 dev->netdev_ops->ndo_get_offload_stats))
5079 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5080 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5081 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5083 size = rtnl_get_offload_stats_attr_size(attr_id);
5084 nla_size += nla_total_size_64bit(size);
5088 nla_size += nla_total_size(0);
5093 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5094 int type, u32 pid, u32 seq, u32 change,
5095 unsigned int flags, unsigned int filter_mask,
5096 int *idxattr, int *prividx)
5098 struct if_stats_msg *ifsm;
5099 struct nlmsghdr *nlh;
5100 struct nlattr *attr;
5101 int s_prividx = *prividx;
5106 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5110 ifsm = nlmsg_data(nlh);
5111 ifsm->family = PF_UNSPEC;
5114 ifsm->ifindex = dev->ifindex;
5115 ifsm->filter_mask = filter_mask;
5117 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5118 struct rtnl_link_stats64 *sp;
5120 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5121 sizeof(struct rtnl_link_stats64),
5124 goto nla_put_failure;
5126 sp = nla_data(attr);
5127 dev_get_stats(dev, sp);
5130 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5131 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5133 if (ops && ops->fill_linkxstats) {
5134 *idxattr = IFLA_STATS_LINK_XSTATS;
5135 attr = nla_nest_start_noflag(skb,
5136 IFLA_STATS_LINK_XSTATS);
5138 goto nla_put_failure;
5140 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5141 nla_nest_end(skb, attr);
5143 goto nla_put_failure;
5148 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5150 const struct rtnl_link_ops *ops = NULL;
5151 const struct net_device *master;
5153 master = netdev_master_upper_dev_get(dev);
5155 ops = master->rtnl_link_ops;
5156 if (ops && ops->fill_linkxstats) {
5157 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5158 attr = nla_nest_start_noflag(skb,
5159 IFLA_STATS_LINK_XSTATS_SLAVE);
5161 goto nla_put_failure;
5163 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5164 nla_nest_end(skb, attr);
5166 goto nla_put_failure;
5171 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5173 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5174 attr = nla_nest_start_noflag(skb,
5175 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5177 goto nla_put_failure;
5179 err = rtnl_get_offload_stats(skb, dev, prividx);
5180 if (err == -ENODATA)
5181 nla_nest_cancel(skb, attr);
5183 nla_nest_end(skb, attr);
5185 if (err && err != -ENODATA)
5186 goto nla_put_failure;
5190 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5191 struct rtnl_af_ops *af_ops;
5193 *idxattr = IFLA_STATS_AF_SPEC;
5194 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5196 goto nla_put_failure;
5199 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5200 if (af_ops->fill_stats_af) {
5204 af = nla_nest_start_noflag(skb,
5208 goto nla_put_failure;
5210 err = af_ops->fill_stats_af(skb, dev);
5212 if (err == -ENODATA) {
5213 nla_nest_cancel(skb, af);
5214 } else if (err < 0) {
5216 goto nla_put_failure;
5219 nla_nest_end(skb, af);
5224 nla_nest_end(skb, attr);
5229 nlmsg_end(skb, nlh);
5234 /* not a multi message or no progress mean a real error */
5235 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5236 nlmsg_cancel(skb, nlh);
5238 nlmsg_end(skb, nlh);
5243 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5248 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5249 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5251 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5252 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5253 int attr = IFLA_STATS_LINK_XSTATS;
5255 if (ops && ops->get_linkxstats_size) {
5256 size += nla_total_size(ops->get_linkxstats_size(dev,
5258 /* for IFLA_STATS_LINK_XSTATS */
5259 size += nla_total_size(0);
5263 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5264 struct net_device *_dev = (struct net_device *)dev;
5265 const struct rtnl_link_ops *ops = NULL;
5266 const struct net_device *master;
5268 /* netdev_master_upper_dev_get can't take const */
5269 master = netdev_master_upper_dev_get(_dev);
5271 ops = master->rtnl_link_ops;
5272 if (ops && ops->get_linkxstats_size) {
5273 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5275 size += nla_total_size(ops->get_linkxstats_size(dev,
5277 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5278 size += nla_total_size(0);
5282 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5283 size += rtnl_get_offload_stats_size(dev);
5285 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5286 struct rtnl_af_ops *af_ops;
5288 /* for IFLA_STATS_AF_SPEC */
5289 size += nla_total_size(0);
5292 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5293 if (af_ops->get_stats_af_size) {
5294 size += nla_total_size(
5295 af_ops->get_stats_af_size(dev));
5298 size += nla_total_size(0);
5307 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5308 bool is_dump, struct netlink_ext_ack *extack)
5310 struct if_stats_msg *ifsm;
5312 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5313 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5320 ifsm = nlmsg_data(nlh);
5322 /* only requests using strict checks can pass data to influence
5323 * the dump. The legacy exception is filter_mask.
5325 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5326 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5329 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5330 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5333 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5334 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5341 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5342 struct netlink_ext_ack *extack)
5344 struct net *net = sock_net(skb->sk);
5345 struct net_device *dev = NULL;
5346 int idxattr = 0, prividx = 0;
5347 struct if_stats_msg *ifsm;
5348 struct sk_buff *nskb;
5352 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5357 ifsm = nlmsg_data(nlh);
5358 if (ifsm->ifindex > 0)
5359 dev = __dev_get_by_index(net, ifsm->ifindex);
5366 filter_mask = ifsm->filter_mask;
5370 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5374 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5375 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5376 0, filter_mask, &idxattr, &prividx);
5378 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5379 WARN_ON(err == -EMSGSIZE);
5382 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5388 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5390 struct netlink_ext_ack *extack = cb->extack;
5391 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5392 struct net *net = sock_net(skb->sk);
5393 unsigned int flags = NLM_F_MULTI;
5394 struct if_stats_msg *ifsm;
5395 struct hlist_head *head;
5396 struct net_device *dev;
5397 u32 filter_mask = 0;
5401 s_idx = cb->args[1];
5402 s_idxattr = cb->args[2];
5403 s_prividx = cb->args[3];
5405 cb->seq = net->dev_base_seq;
5407 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5411 ifsm = nlmsg_data(cb->nlh);
5412 filter_mask = ifsm->filter_mask;
5414 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5418 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5420 head = &net->dev_index_head[h];
5421 hlist_for_each_entry(dev, head, index_hlist) {
5424 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5425 NETLINK_CB(cb->skb).portid,
5426 cb->nlh->nlmsg_seq, 0,
5428 &s_idxattr, &s_prividx);
5429 /* If we ran out of room on the first message,
5432 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5438 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5444 cb->args[3] = s_prividx;
5445 cb->args[2] = s_idxattr;
5452 /* Process one rtnetlink message. */
5454 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5455 struct netlink_ext_ack *extack)
5457 struct net *net = sock_net(skb->sk);
5458 struct rtnl_link *link;
5459 struct module *owner;
5460 int err = -EOPNOTSUPP;
5461 rtnl_doit_func doit;
5467 type = nlh->nlmsg_type;
5473 /* All the messages must have at least 1 byte length */
5474 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5477 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5480 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5484 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5486 rtnl_dumpit_func dumpit;
5487 u32 min_dump_alloc = 0;
5489 link = rtnl_get_link(family, type);
5490 if (!link || !link->dumpit) {
5492 link = rtnl_get_link(family, type);
5493 if (!link || !link->dumpit)
5496 owner = link->owner;
5497 dumpit = link->dumpit;
5499 if (type == RTM_GETLINK - RTM_BASE)
5500 min_dump_alloc = rtnl_calcit(skb, nlh);
5503 /* need to do this before rcu_read_unlock() */
5504 if (!try_module_get(owner))
5505 err = -EPROTONOSUPPORT;
5511 struct netlink_dump_control c = {
5513 .min_dump_alloc = min_dump_alloc,
5516 err = netlink_dump_start(rtnl, skb, nlh, &c);
5517 /* netlink_dump_start() will keep a reference on
5518 * module if dump is still in progress.
5525 link = rtnl_get_link(family, type);
5526 if (!link || !link->doit) {
5528 link = rtnl_get_link(PF_UNSPEC, type);
5529 if (!link || !link->doit)
5533 owner = link->owner;
5534 if (!try_module_get(owner)) {
5535 err = -EPROTONOSUPPORT;
5539 flags = link->flags;
5540 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5544 err = doit(skb, nlh, extack);
5551 link = rtnl_get_link(family, type);
5552 if (link && link->doit)
5553 err = link->doit(skb, nlh, extack);
5569 static void rtnetlink_rcv(struct sk_buff *skb)
5571 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5574 static int rtnetlink_bind(struct net *net, int group)
5577 case RTNLGRP_IPV4_MROUTE_R:
5578 case RTNLGRP_IPV6_MROUTE_R:
5579 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5586 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5588 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5592 case NETDEV_CHANGEMTU:
5593 case NETDEV_CHANGEADDR:
5594 case NETDEV_CHANGENAME:
5595 case NETDEV_FEAT_CHANGE:
5596 case NETDEV_BONDING_FAILOVER:
5597 case NETDEV_POST_TYPE_CHANGE:
5598 case NETDEV_NOTIFY_PEERS:
5599 case NETDEV_CHANGEUPPER:
5600 case NETDEV_RESEND_IGMP:
5601 case NETDEV_CHANGEINFODATA:
5602 case NETDEV_CHANGELOWERSTATE:
5603 case NETDEV_CHANGE_TX_QUEUE_LEN:
5604 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5605 GFP_KERNEL, NULL, 0);
5613 static struct notifier_block rtnetlink_dev_notifier = {
5614 .notifier_call = rtnetlink_event,
5618 static int __net_init rtnetlink_net_init(struct net *net)
5621 struct netlink_kernel_cfg cfg = {
5622 .groups = RTNLGRP_MAX,
5623 .input = rtnetlink_rcv,
5624 .cb_mutex = &rtnl_mutex,
5625 .flags = NL_CFG_F_NONROOT_RECV,
5626 .bind = rtnetlink_bind,
5629 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5636 static void __net_exit rtnetlink_net_exit(struct net *net)
5638 netlink_kernel_release(net->rtnl);
5642 static struct pernet_operations rtnetlink_net_ops = {
5643 .init = rtnetlink_net_init,
5644 .exit = rtnetlink_net_exit,
5647 void __init rtnetlink_init(void)
5649 if (register_pernet_subsys(&rtnetlink_net_ops))
5650 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5652 register_netdevice_notifier(&rtnetlink_dev_notifier);
5654 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5655 rtnl_dump_ifinfo, 0);
5656 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5657 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5658 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5660 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5661 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5662 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5664 rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5665 rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5667 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5668 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5669 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5671 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5672 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5673 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5675 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,