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 },
1880 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
1883 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1884 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1885 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1886 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1887 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1890 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1891 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1892 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1893 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1894 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1895 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1896 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1897 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1898 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1899 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1900 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1901 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1902 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1903 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1906 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1907 [IFLA_PORT_VF] = { .type = NLA_U32 },
1908 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1909 .len = PORT_PROFILE_MAX },
1910 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1911 .len = PORT_UUID_MAX },
1912 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1913 .len = PORT_UUID_MAX },
1914 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1915 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1917 /* Unused, but we need to keep it here since user space could
1918 * fill it. It's also broken with regard to NLA_BINARY use in
1919 * combination with structs.
1921 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1922 .len = sizeof(struct ifla_port_vsi) },
1925 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1926 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1927 [IFLA_XDP_FD] = { .type = NLA_S32 },
1928 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
1929 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1930 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1931 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1934 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1936 const struct rtnl_link_ops *ops = NULL;
1937 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1939 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1942 if (linfo[IFLA_INFO_KIND]) {
1943 char kind[MODULE_NAME_LEN];
1945 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1946 ops = rtnl_link_ops_get(kind);
1952 static bool link_master_filtered(struct net_device *dev, int master_idx)
1954 struct net_device *master;
1959 master = netdev_master_upper_dev_get(dev);
1960 if (!master || master->ifindex != master_idx)
1966 static bool link_kind_filtered(const struct net_device *dev,
1967 const struct rtnl_link_ops *kind_ops)
1969 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1975 static bool link_dump_filtered(struct net_device *dev,
1977 const struct rtnl_link_ops *kind_ops)
1979 if (link_master_filtered(dev, master_idx) ||
1980 link_kind_filtered(dev, kind_ops))
1987 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1988 * @sk: netlink socket
1989 * @netnsid: network namespace identifier
1991 * Returns the network namespace identified by netnsid on success or an error
1992 * pointer on failure.
1994 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1998 net = get_net_ns_by_id(sock_net(sk), netnsid);
2000 return ERR_PTR(-EINVAL);
2002 /* For now, the caller is required to have CAP_NET_ADMIN in
2003 * the user namespace owning the target net ns.
2005 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2007 return ERR_PTR(-EACCES);
2011 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2013 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2014 bool strict_check, struct nlattr **tb,
2015 struct netlink_ext_ack *extack)
2020 struct ifinfomsg *ifm;
2022 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2023 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2027 ifm = nlmsg_data(nlh);
2028 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2030 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2033 if (ifm->ifi_index) {
2034 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2038 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2039 IFLA_MAX, ifla_policy,
2043 /* A hack to preserve kernel<->userspace interface.
2044 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2045 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2046 * what iproute2 < v3.9.0 used.
2047 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2048 * attribute, its netlink message is shorter than struct ifinfomsg.
2050 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2051 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2053 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2057 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2059 struct netlink_ext_ack *extack = cb->extack;
2060 const struct nlmsghdr *nlh = cb->nlh;
2061 struct net *net = sock_net(skb->sk);
2062 struct net *tgt_net = net;
2065 struct net_device *dev;
2066 struct hlist_head *head;
2067 struct nlattr *tb[IFLA_MAX+1];
2068 u32 ext_filter_mask = 0;
2069 const struct rtnl_link_ops *kind_ops = NULL;
2070 unsigned int flags = NLM_F_MULTI;
2076 s_idx = cb->args[1];
2078 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2080 if (cb->strict_check)
2086 for (i = 0; i <= IFLA_MAX; ++i) {
2090 /* new attributes should only be added with strict checking */
2092 case IFLA_TARGET_NETNSID:
2093 netnsid = nla_get_s32(tb[i]);
2094 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2095 if (IS_ERR(tgt_net)) {
2096 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2097 return PTR_ERR(tgt_net);
2101 ext_filter_mask = nla_get_u32(tb[i]);
2104 master_idx = nla_get_u32(tb[i]);
2107 kind_ops = linkinfo_to_kind_ops(tb[i]);
2110 if (cb->strict_check) {
2111 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2117 if (master_idx || kind_ops)
2118 flags |= NLM_F_DUMP_FILTERED;
2121 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2123 head = &tgt_net->dev_index_head[h];
2124 hlist_for_each_entry(dev, head, index_hlist) {
2125 if (link_dump_filtered(dev, master_idx, kind_ops))
2129 err = rtnl_fill_ifinfo(skb, dev, net,
2131 NETLINK_CB(cb->skb).portid,
2132 nlh->nlmsg_seq, 0, flags,
2133 ext_filter_mask, 0, NULL, 0,
2134 netnsid, GFP_KERNEL);
2137 if (likely(skb->len))
2151 cb->seq = tgt_net->dev_base_seq;
2152 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2159 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2160 struct netlink_ext_ack *exterr)
2162 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2165 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2167 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2170 /* Examine the link attributes and figure out which
2171 * network namespace we are talking about.
2173 if (tb[IFLA_NET_NS_PID])
2174 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2175 else if (tb[IFLA_NET_NS_FD])
2176 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2178 net = get_net(src_net);
2181 EXPORT_SYMBOL(rtnl_link_get_net);
2183 /* Figure out which network namespace we are talking about by
2184 * examining the link attributes in the following order:
2186 * 1. IFLA_NET_NS_PID
2188 * 3. IFLA_TARGET_NETNSID
2190 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2191 struct nlattr *tb[])
2195 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2196 return rtnl_link_get_net(src_net, tb);
2198 if (!tb[IFLA_TARGET_NETNSID])
2199 return get_net(src_net);
2201 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2203 return ERR_PTR(-EINVAL);
2208 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2209 struct net *src_net,
2210 struct nlattr *tb[], int cap)
2214 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2218 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2220 return ERR_PTR(-EPERM);
2226 /* Verify that rtnetlink requests do not pass additional properties
2227 * potentially referring to different network namespaces.
2229 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2230 struct netlink_ext_ack *extack,
2234 if (netns_id_only) {
2235 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2238 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2242 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2245 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2248 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2254 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2258 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2261 if (tb[IFLA_ADDRESS] &&
2262 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2265 if (tb[IFLA_BROADCAST] &&
2266 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2270 if (tb[IFLA_AF_SPEC]) {
2274 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2275 const struct rtnl_af_ops *af_ops;
2278 af_ops = rtnl_af_lookup(nla_type(af));
2281 return -EAFNOSUPPORT;
2284 if (!af_ops->set_link_af) {
2289 if (af_ops->validate_link_af) {
2290 err = af_ops->validate_link_af(dev, af);
2304 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2307 const struct net_device_ops *ops = dev->netdev_ops;
2309 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2312 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2314 if (dev->type != ARPHRD_INFINIBAND)
2317 return handle_infiniband_guid(dev, ivt, guid_type);
2320 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2322 const struct net_device_ops *ops = dev->netdev_ops;
2325 if (tb[IFLA_VF_MAC]) {
2326 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2328 if (ivm->vf >= INT_MAX)
2331 if (ops->ndo_set_vf_mac)
2332 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2338 if (tb[IFLA_VF_VLAN]) {
2339 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2341 if (ivv->vf >= INT_MAX)
2344 if (ops->ndo_set_vf_vlan)
2345 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2347 htons(ETH_P_8021Q));
2352 if (tb[IFLA_VF_VLAN_LIST]) {
2353 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2354 struct nlattr *attr;
2358 if (!ops->ndo_set_vf_vlan)
2361 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2362 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2363 nla_len(attr) < NLA_HDRLEN) {
2366 if (len >= MAX_VLAN_LIST_LEN)
2368 ivvl[len] = nla_data(attr);
2375 if (ivvl[0]->vf >= INT_MAX)
2377 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2378 ivvl[0]->qos, ivvl[0]->vlan_proto);
2383 if (tb[IFLA_VF_TX_RATE]) {
2384 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2385 struct ifla_vf_info ivf;
2387 if (ivt->vf >= INT_MAX)
2390 if (ops->ndo_get_vf_config)
2391 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2396 if (ops->ndo_set_vf_rate)
2397 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2404 if (tb[IFLA_VF_RATE]) {
2405 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2407 if (ivt->vf >= INT_MAX)
2410 if (ops->ndo_set_vf_rate)
2411 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2418 if (tb[IFLA_VF_SPOOFCHK]) {
2419 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2421 if (ivs->vf >= INT_MAX)
2424 if (ops->ndo_set_vf_spoofchk)
2425 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2431 if (tb[IFLA_VF_LINK_STATE]) {
2432 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2434 if (ivl->vf >= INT_MAX)
2437 if (ops->ndo_set_vf_link_state)
2438 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2444 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2445 struct ifla_vf_rss_query_en *ivrssq_en;
2448 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2449 if (ivrssq_en->vf >= INT_MAX)
2451 if (ops->ndo_set_vf_rss_query_en)
2452 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2453 ivrssq_en->setting);
2458 if (tb[IFLA_VF_TRUST]) {
2459 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2461 if (ivt->vf >= INT_MAX)
2464 if (ops->ndo_set_vf_trust)
2465 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2470 if (tb[IFLA_VF_IB_NODE_GUID]) {
2471 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2473 if (ivt->vf >= INT_MAX)
2475 if (!ops->ndo_set_vf_guid)
2477 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2480 if (tb[IFLA_VF_IB_PORT_GUID]) {
2481 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2483 if (ivt->vf >= INT_MAX)
2485 if (!ops->ndo_set_vf_guid)
2488 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2494 static int do_set_master(struct net_device *dev, int ifindex,
2495 struct netlink_ext_ack *extack)
2497 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2498 const struct net_device_ops *ops;
2502 if (upper_dev->ifindex == ifindex)
2504 ops = upper_dev->netdev_ops;
2505 if (ops->ndo_del_slave) {
2506 err = ops->ndo_del_slave(upper_dev, dev);
2515 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2518 ops = upper_dev->netdev_ops;
2519 if (ops->ndo_add_slave) {
2520 err = ops->ndo_add_slave(upper_dev, dev, extack);
2530 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2531 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2532 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2535 static int do_set_proto_down(struct net_device *dev,
2536 struct nlattr *nl_proto_down,
2537 struct nlattr *nl_proto_down_reason,
2538 struct netlink_ext_ack *extack)
2540 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2541 const struct net_device_ops *ops = dev->netdev_ops;
2542 unsigned long mask = 0;
2547 if (!ops->ndo_change_proto_down) {
2548 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2552 if (nl_proto_down_reason) {
2553 err = nla_parse_nested_deprecated(pdreason,
2554 IFLA_PROTO_DOWN_REASON_MAX,
2555 nl_proto_down_reason,
2556 ifla_proto_down_reason_policy,
2561 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2562 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2566 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2568 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2569 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2571 dev_change_proto_down_reason(dev, mask, value);
2574 if (nl_proto_down) {
2575 proto_down = nla_get_u8(nl_proto_down);
2577 /* Dont turn off protodown if there are active reasons */
2578 if (!proto_down && dev->proto_down_reason) {
2579 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2582 err = dev_change_proto_down(dev,
2591 #define DO_SETLINK_MODIFIED 0x01
2592 /* notify flag means notify + modified. */
2593 #define DO_SETLINK_NOTIFY 0x03
2594 static int do_setlink(const struct sk_buff *skb,
2595 struct net_device *dev, struct ifinfomsg *ifm,
2596 struct netlink_ext_ack *extack,
2597 struct nlattr **tb, char *ifname, int status)
2599 const struct net_device_ops *ops = dev->netdev_ops;
2602 err = validate_linkmsg(dev, tb);
2606 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2610 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2617 if (tb[IFLA_NEW_IFINDEX])
2618 new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2622 err = __dev_change_net_namespace(dev, net, ifname, new_ifindex);
2626 status |= DO_SETLINK_MODIFIED;
2630 struct rtnl_link_ifmap *u_map;
2633 if (!ops->ndo_set_config) {
2638 if (!netif_device_present(dev)) {
2643 u_map = nla_data(tb[IFLA_MAP]);
2644 k_map.mem_start = (unsigned long) u_map->mem_start;
2645 k_map.mem_end = (unsigned long) u_map->mem_end;
2646 k_map.base_addr = (unsigned short) u_map->base_addr;
2647 k_map.irq = (unsigned char) u_map->irq;
2648 k_map.dma = (unsigned char) u_map->dma;
2649 k_map.port = (unsigned char) u_map->port;
2651 err = ops->ndo_set_config(dev, &k_map);
2655 status |= DO_SETLINK_NOTIFY;
2658 if (tb[IFLA_ADDRESS]) {
2659 struct sockaddr *sa;
2662 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2664 sa = kmalloc(len, GFP_KERNEL);
2669 sa->sa_family = dev->type;
2670 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2672 err = dev_set_mac_address_user(dev, sa, extack);
2676 status |= DO_SETLINK_MODIFIED;
2680 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2683 status |= DO_SETLINK_MODIFIED;
2686 if (tb[IFLA_GROUP]) {
2687 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2688 status |= DO_SETLINK_NOTIFY;
2692 * Interface selected by interface index but interface
2693 * name provided implies that a name change has been
2696 if (ifm->ifi_index > 0 && ifname[0]) {
2697 err = dev_change_name(dev, ifname);
2700 status |= DO_SETLINK_MODIFIED;
2703 if (tb[IFLA_IFALIAS]) {
2704 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2705 nla_len(tb[IFLA_IFALIAS]));
2708 status |= DO_SETLINK_NOTIFY;
2711 if (tb[IFLA_BROADCAST]) {
2712 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2713 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2716 if (ifm->ifi_flags || ifm->ifi_change) {
2717 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2723 if (tb[IFLA_MASTER]) {
2724 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2727 status |= DO_SETLINK_MODIFIED;
2730 if (tb[IFLA_CARRIER]) {
2731 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2734 status |= DO_SETLINK_MODIFIED;
2737 if (tb[IFLA_TXQLEN]) {
2738 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2740 err = dev_change_tx_queue_len(dev, value);
2743 status |= DO_SETLINK_MODIFIED;
2746 if (tb[IFLA_GSO_MAX_SIZE]) {
2747 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2749 if (max_size > GSO_MAX_SIZE) {
2754 if (dev->gso_max_size ^ max_size) {
2755 netif_set_gso_max_size(dev, max_size);
2756 status |= DO_SETLINK_MODIFIED;
2760 if (tb[IFLA_GSO_MAX_SEGS]) {
2761 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2763 if (max_segs > GSO_MAX_SEGS) {
2768 if (dev->gso_max_segs ^ max_segs) {
2769 dev->gso_max_segs = max_segs;
2770 status |= DO_SETLINK_MODIFIED;
2774 if (tb[IFLA_OPERSTATE])
2775 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2777 if (tb[IFLA_LINKMODE]) {
2778 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2780 write_lock_bh(&dev_base_lock);
2781 if (dev->link_mode ^ value)
2782 status |= DO_SETLINK_NOTIFY;
2783 dev->link_mode = value;
2784 write_unlock_bh(&dev_base_lock);
2787 if (tb[IFLA_VFINFO_LIST]) {
2788 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2789 struct nlattr *attr;
2792 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2793 if (nla_type(attr) != IFLA_VF_INFO ||
2794 nla_len(attr) < NLA_HDRLEN) {
2798 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2804 err = do_setvfinfo(dev, vfinfo);
2807 status |= DO_SETLINK_NOTIFY;
2812 if (tb[IFLA_VF_PORTS]) {
2813 struct nlattr *port[IFLA_PORT_MAX+1];
2814 struct nlattr *attr;
2819 if (!ops->ndo_set_vf_port)
2822 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2823 if (nla_type(attr) != IFLA_VF_PORT ||
2824 nla_len(attr) < NLA_HDRLEN) {
2828 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2834 if (!port[IFLA_PORT_VF]) {
2838 vf = nla_get_u32(port[IFLA_PORT_VF]);
2839 err = ops->ndo_set_vf_port(dev, vf, port);
2842 status |= DO_SETLINK_NOTIFY;
2847 if (tb[IFLA_PORT_SELF]) {
2848 struct nlattr *port[IFLA_PORT_MAX+1];
2850 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2852 ifla_port_policy, NULL);
2857 if (ops->ndo_set_vf_port)
2858 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2861 status |= DO_SETLINK_NOTIFY;
2864 if (tb[IFLA_AF_SPEC]) {
2868 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2869 const struct rtnl_af_ops *af_ops;
2873 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2875 err = af_ops->set_link_af(dev, af, extack);
2882 status |= DO_SETLINK_NOTIFY;
2887 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2888 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2889 tb[IFLA_PROTO_DOWN_REASON], extack);
2892 status |= DO_SETLINK_NOTIFY;
2896 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2899 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2901 ifla_xdp_policy, NULL);
2905 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2910 if (xdp[IFLA_XDP_FLAGS]) {
2911 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2912 if (xdp_flags & ~XDP_FLAGS_MASK) {
2916 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2922 if (xdp[IFLA_XDP_FD]) {
2923 int expected_fd = -1;
2925 if (xdp_flags & XDP_FLAGS_REPLACE) {
2926 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2931 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2934 err = dev_change_xdp_fd(dev, extack,
2935 nla_get_s32(xdp[IFLA_XDP_FD]),
2940 status |= DO_SETLINK_NOTIFY;
2945 if (status & DO_SETLINK_MODIFIED) {
2946 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2947 netdev_state_change(dev);
2950 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",
2957 static struct net_device *rtnl_dev_get(struct net *net,
2958 struct nlattr *ifname_attr,
2959 struct nlattr *altifname_attr,
2962 char buffer[ALTIFNAMSIZ];
2967 nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2968 else if (altifname_attr)
2969 nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2974 return __dev_get_by_name(net, ifname);
2977 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2978 struct netlink_ext_ack *extack)
2980 struct net *net = sock_net(skb->sk);
2981 struct ifinfomsg *ifm;
2982 struct net_device *dev;
2984 struct nlattr *tb[IFLA_MAX+1];
2985 char ifname[IFNAMSIZ];
2987 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2988 ifla_policy, extack);
2992 err = rtnl_ensure_unique_netns(tb, extack, false);
2996 if (tb[IFLA_IFNAME])
2997 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3002 ifm = nlmsg_data(nlh);
3003 if (ifm->ifi_index > 0)
3004 dev = __dev_get_by_index(net, ifm->ifi_index);
3005 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3006 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3015 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3020 static int rtnl_group_dellink(const struct net *net, int group)
3022 struct net_device *dev, *aux;
3023 LIST_HEAD(list_kill);
3029 for_each_netdev(net, dev) {
3030 if (dev->group == group) {
3031 const struct rtnl_link_ops *ops;
3034 ops = dev->rtnl_link_ops;
3035 if (!ops || !ops->dellink)
3043 for_each_netdev_safe(net, dev, aux) {
3044 if (dev->group == group) {
3045 const struct rtnl_link_ops *ops;
3047 ops = dev->rtnl_link_ops;
3048 ops->dellink(dev, &list_kill);
3051 unregister_netdevice_many(&list_kill);
3056 int rtnl_delete_link(struct net_device *dev)
3058 const struct rtnl_link_ops *ops;
3059 LIST_HEAD(list_kill);
3061 ops = dev->rtnl_link_ops;
3062 if (!ops || !ops->dellink)
3065 ops->dellink(dev, &list_kill);
3066 unregister_netdevice_many(&list_kill);
3070 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3072 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3073 struct netlink_ext_ack *extack)
3075 struct net *net = sock_net(skb->sk);
3076 struct net *tgt_net = net;
3077 struct net_device *dev = NULL;
3078 struct ifinfomsg *ifm;
3079 struct nlattr *tb[IFLA_MAX+1];
3083 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3084 ifla_policy, extack);
3088 err = rtnl_ensure_unique_netns(tb, extack, true);
3092 if (tb[IFLA_TARGET_NETNSID]) {
3093 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3094 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3095 if (IS_ERR(tgt_net))
3096 return PTR_ERR(tgt_net);
3100 ifm = nlmsg_data(nlh);
3101 if (ifm->ifi_index > 0)
3102 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3103 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3104 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3105 tb[IFLA_ALT_IFNAME], NULL);
3106 else if (tb[IFLA_GROUP])
3107 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3112 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3118 err = rtnl_delete_link(dev);
3127 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3129 unsigned int old_flags;
3132 old_flags = dev->flags;
3133 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3134 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3140 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3141 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3143 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3144 __dev_notify_flags(dev, old_flags, ~0U);
3148 EXPORT_SYMBOL(rtnl_configure_link);
3150 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3151 unsigned char name_assign_type,
3152 const struct rtnl_link_ops *ops,
3153 struct nlattr *tb[],
3154 struct netlink_ext_ack *extack)
3156 struct net_device *dev;
3157 unsigned int num_tx_queues = 1;
3158 unsigned int num_rx_queues = 1;
3160 if (tb[IFLA_NUM_TX_QUEUES])
3161 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3162 else if (ops->get_num_tx_queues)
3163 num_tx_queues = ops->get_num_tx_queues();
3165 if (tb[IFLA_NUM_RX_QUEUES])
3166 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3167 else if (ops->get_num_rx_queues)
3168 num_rx_queues = ops->get_num_rx_queues();
3170 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3171 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3172 return ERR_PTR(-EINVAL);
3175 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3176 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3177 return ERR_PTR(-EINVAL);
3180 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3181 ops->setup, num_tx_queues, num_rx_queues);
3183 return ERR_PTR(-ENOMEM);
3185 dev_net_set(dev, net);
3186 dev->rtnl_link_ops = ops;
3187 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3190 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3193 err = dev_validate_mtu(dev, mtu, extack);
3196 return ERR_PTR(err);
3200 if (tb[IFLA_ADDRESS]) {
3201 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3202 nla_len(tb[IFLA_ADDRESS]));
3203 dev->addr_assign_type = NET_ADDR_SET;
3205 if (tb[IFLA_BROADCAST])
3206 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3207 nla_len(tb[IFLA_BROADCAST]));
3208 if (tb[IFLA_TXQLEN])
3209 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3210 if (tb[IFLA_OPERSTATE])
3211 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3212 if (tb[IFLA_LINKMODE])
3213 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3215 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3216 if (tb[IFLA_GSO_MAX_SIZE])
3217 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3218 if (tb[IFLA_GSO_MAX_SEGS])
3219 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3223 EXPORT_SYMBOL(rtnl_create_link);
3225 static int rtnl_group_changelink(const struct sk_buff *skb,
3226 struct net *net, int group,
3227 struct ifinfomsg *ifm,
3228 struct netlink_ext_ack *extack,
3231 struct net_device *dev, *aux;
3234 for_each_netdev_safe(net, dev, aux) {
3235 if (dev->group == group) {
3236 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3245 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3246 struct nlattr **attr, struct netlink_ext_ack *extack)
3248 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3249 unsigned char name_assign_type = NET_NAME_USER;
3250 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3251 const struct rtnl_link_ops *m_ops = NULL;
3252 struct net_device *master_dev = NULL;
3253 struct net *net = sock_net(skb->sk);
3254 const struct rtnl_link_ops *ops;
3255 struct nlattr *tb[IFLA_MAX + 1];
3256 struct net *dest_net, *link_net;
3257 struct nlattr **slave_data;
3258 char kind[MODULE_NAME_LEN];
3259 struct net_device *dev;
3260 struct ifinfomsg *ifm;
3261 char ifname[IFNAMSIZ];
3262 struct nlattr **data;
3265 #ifdef CONFIG_MODULES
3268 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3269 ifla_policy, extack);
3273 err = rtnl_ensure_unique_netns(tb, extack, false);
3277 if (tb[IFLA_IFNAME])
3278 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3282 ifm = nlmsg_data(nlh);
3283 if (ifm->ifi_index > 0)
3284 dev = __dev_get_by_index(net, ifm->ifi_index);
3285 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3286 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3291 master_dev = netdev_master_upper_dev_get(dev);
3293 m_ops = master_dev->rtnl_link_ops;
3296 err = validate_linkmsg(dev, tb);
3300 if (tb[IFLA_LINKINFO]) {
3301 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3303 ifla_info_policy, NULL);
3307 memset(linkinfo, 0, sizeof(linkinfo));
3309 if (linkinfo[IFLA_INFO_KIND]) {
3310 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3311 ops = rtnl_link_ops_get(kind);
3319 if (ops->maxtype > RTNL_MAX_TYPE)
3322 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3323 err = nla_parse_nested_deprecated(attr, ops->maxtype,
3324 linkinfo[IFLA_INFO_DATA],
3325 ops->policy, extack);
3330 if (ops->validate) {
3331 err = ops->validate(tb, data, extack);
3339 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3342 if (m_ops->slave_maxtype &&
3343 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3344 err = nla_parse_nested_deprecated(slave_attr,
3345 m_ops->slave_maxtype,
3346 linkinfo[IFLA_INFO_SLAVE_DATA],
3347 m_ops->slave_policy,
3351 slave_data = slave_attr;
3358 if (nlh->nlmsg_flags & NLM_F_EXCL)
3360 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3363 if (linkinfo[IFLA_INFO_DATA]) {
3364 if (!ops || ops != dev->rtnl_link_ops ||
3368 err = ops->changelink(dev, tb, data, extack);
3371 status |= DO_SETLINK_NOTIFY;
3374 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3375 if (!m_ops || !m_ops->slave_changelink)
3378 err = m_ops->slave_changelink(master_dev, dev, tb,
3379 slave_data, extack);
3382 status |= DO_SETLINK_NOTIFY;
3385 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3388 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3389 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3390 return rtnl_group_changelink(skb, net,
3391 nla_get_u32(tb[IFLA_GROUP]),
3396 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3400 #ifdef CONFIG_MODULES
3403 request_module("rtnl-link-%s", kind);
3405 ops = rtnl_link_ops_get(kind);
3410 NL_SET_ERR_MSG(extack, "Unknown device type");
3418 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3419 name_assign_type = NET_NAME_ENUM;
3422 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3423 if (IS_ERR(dest_net))
3424 return PTR_ERR(dest_net);
3426 if (tb[IFLA_LINK_NETNSID]) {
3427 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3429 link_net = get_net_ns_by_id(dest_net, id);
3431 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3436 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3442 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3443 name_assign_type, ops, tb, extack);
3449 dev->ifindex = ifm->ifi_index;
3452 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3454 err = register_netdevice(dev);
3460 err = rtnl_configure_link(dev, ifm);
3462 goto out_unregister;
3464 err = dev_change_net_namespace(dev, dest_net, ifname);
3466 goto out_unregister;
3468 if (tb[IFLA_MASTER]) {
3469 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3471 goto out_unregister;
3480 LIST_HEAD(list_kill);
3482 ops->dellink(dev, &list_kill);
3483 unregister_netdevice_many(&list_kill);
3485 unregister_netdevice(dev);
3490 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3491 struct netlink_ext_ack *extack)
3493 struct nlattr **attr;
3496 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3500 ret = __rtnl_newlink(skb, nlh, attr, extack);
3505 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3506 const struct nlmsghdr *nlh,
3508 struct netlink_ext_ack *extack)
3510 struct ifinfomsg *ifm;
3513 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3514 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3518 if (!netlink_strict_get_check(skb))
3519 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3520 ifla_policy, extack);
3522 ifm = nlmsg_data(nlh);
3523 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3525 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3529 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3530 ifla_policy, extack);
3534 for (i = 0; i <= IFLA_MAX; i++) {
3540 case IFLA_ALT_IFNAME:
3542 case IFLA_TARGET_NETNSID:
3545 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3553 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3554 struct netlink_ext_ack *extack)
3556 struct net *net = sock_net(skb->sk);
3557 struct net *tgt_net = net;
3558 struct ifinfomsg *ifm;
3559 struct nlattr *tb[IFLA_MAX+1];
3560 struct net_device *dev = NULL;
3561 struct sk_buff *nskb;
3564 u32 ext_filter_mask = 0;
3566 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3570 err = rtnl_ensure_unique_netns(tb, extack, true);
3574 if (tb[IFLA_TARGET_NETNSID]) {
3575 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3576 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3577 if (IS_ERR(tgt_net))
3578 return PTR_ERR(tgt_net);
3581 if (tb[IFLA_EXT_MASK])
3582 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3585 ifm = nlmsg_data(nlh);
3586 if (ifm->ifi_index > 0)
3587 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3588 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3589 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3590 tb[IFLA_ALT_IFNAME], NULL);
3599 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3603 err = rtnl_fill_ifinfo(nskb, dev, net,
3604 RTM_NEWLINK, NETLINK_CB(skb).portid,
3605 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3606 0, NULL, 0, netnsid, GFP_KERNEL);
3608 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3609 WARN_ON(err == -EMSGSIZE);
3612 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3620 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3621 bool *changed, struct netlink_ext_ack *extack)
3626 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3630 alt_ifname = nla_strdup(attr, GFP_KERNEL);
3634 if (cmd == RTM_NEWLINKPROP) {
3635 err = netdev_name_node_alt_create(dev, alt_ifname);
3638 } else if (cmd == RTM_DELLINKPROP) {
3639 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3651 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3652 struct netlink_ext_ack *extack)
3654 struct net *net = sock_net(skb->sk);
3655 struct nlattr *tb[IFLA_MAX + 1];
3656 struct net_device *dev;
3657 struct ifinfomsg *ifm;
3658 bool changed = false;
3659 struct nlattr *attr;
3662 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3666 err = rtnl_ensure_unique_netns(tb, extack, true);
3670 ifm = nlmsg_data(nlh);
3671 if (ifm->ifi_index > 0)
3672 dev = __dev_get_by_index(net, ifm->ifi_index);
3673 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3674 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3675 tb[IFLA_ALT_IFNAME], NULL);
3682 if (!tb[IFLA_PROP_LIST])
3685 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3686 switch (nla_type(attr)) {
3687 case IFLA_ALT_IFNAME:
3688 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3696 netdev_state_change(dev);
3700 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3701 struct netlink_ext_ack *extack)
3703 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3706 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3707 struct netlink_ext_ack *extack)
3709 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3712 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3714 struct net *net = sock_net(skb->sk);
3715 size_t min_ifinfo_dump_size = 0;
3716 struct nlattr *tb[IFLA_MAX+1];
3717 u32 ext_filter_mask = 0;
3718 struct net_device *dev;
3721 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3722 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3723 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3725 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3726 if (tb[IFLA_EXT_MASK])
3727 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3730 if (!ext_filter_mask)
3731 return NLMSG_GOODSIZE;
3733 * traverse the list of net devices and compute the minimum
3734 * buffer size based upon the filter mask.
3737 for_each_netdev_rcu(net, dev) {
3738 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3739 if_nlmsg_size(dev, ext_filter_mask));
3743 return nlmsg_total_size(min_ifinfo_dump_size);
3746 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3749 int s_idx = cb->family;
3750 int type = cb->nlh->nlmsg_type - RTM_BASE;
3756 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3757 struct rtnl_link __rcu **tab;
3758 struct rtnl_link *link;
3759 rtnl_dumpit_func dumpit;
3761 if (idx < s_idx || idx == PF_PACKET)
3764 if (type < 0 || type >= RTM_NR_MSGTYPES)
3767 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3771 link = rcu_dereference_rtnl(tab[type]);
3775 dumpit = link->dumpit;
3780 memset(&cb->args[0], 0, sizeof(cb->args));
3784 ret = dumpit(skb, cb);
3790 return skb->len ? : ret;
3793 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3794 unsigned int change,
3795 u32 event, gfp_t flags, int *new_nsid,
3798 struct net *net = dev_net(dev);
3799 struct sk_buff *skb;
3801 size_t if_info_size;
3803 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3807 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3808 type, 0, 0, change, 0, 0, event,
3809 new_nsid, new_ifindex, -1, flags);
3811 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3812 WARN_ON(err == -EMSGSIZE);
3819 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3823 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3825 struct net *net = dev_net(dev);
3827 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3830 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3831 unsigned int change, u32 event,
3832 gfp_t flags, int *new_nsid, int new_ifindex)
3834 struct sk_buff *skb;
3836 if (dev->reg_state != NETREG_REGISTERED)
3839 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3842 rtmsg_ifinfo_send(skb, dev, flags);
3845 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3848 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3852 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3853 gfp_t flags, int *new_nsid, int new_ifindex)
3855 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3856 new_nsid, new_ifindex);
3859 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3860 struct net_device *dev,
3861 u8 *addr, u16 vid, u32 pid, u32 seq,
3862 int type, unsigned int flags,
3863 int nlflags, u16 ndm_state)
3865 struct nlmsghdr *nlh;
3868 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3872 ndm = nlmsg_data(nlh);
3873 ndm->ndm_family = AF_BRIDGE;
3876 ndm->ndm_flags = flags;
3878 ndm->ndm_ifindex = dev->ifindex;
3879 ndm->ndm_state = ndm_state;
3881 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3882 goto nla_put_failure;
3884 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3885 goto nla_put_failure;
3887 nlmsg_end(skb, nlh);
3891 nlmsg_cancel(skb, nlh);
3895 static inline size_t rtnl_fdb_nlmsg_size(void)
3897 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3898 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3899 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3903 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3906 struct net *net = dev_net(dev);
3907 struct sk_buff *skb;
3910 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3914 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3915 0, 0, type, NTF_SELF, 0, ndm_state);
3921 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3924 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3928 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3930 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3931 struct nlattr *tb[],
3932 struct net_device *dev,
3933 const unsigned char *addr, u16 vid,
3938 /* If aging addresses are supported device will need to
3939 * implement its own handler for this.
3941 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3942 pr_info("%s: FDB only supports static addresses\n", dev->name);
3947 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3951 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3952 err = dev_uc_add_excl(dev, addr);
3953 else if (is_multicast_ether_addr(addr))
3954 err = dev_mc_add_excl(dev, addr);
3956 /* Only return duplicate errors if NLM_F_EXCL is set */
3957 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3962 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3964 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3965 struct netlink_ext_ack *extack)
3970 if (nla_len(vlan_attr) != sizeof(u16)) {
3971 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3975 vid = nla_get_u16(vlan_attr);
3977 if (!vid || vid >= VLAN_VID_MASK) {
3978 NL_SET_ERR_MSG(extack, "invalid vlan id");
3986 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3987 struct netlink_ext_ack *extack)
3989 struct net *net = sock_net(skb->sk);
3991 struct nlattr *tb[NDA_MAX+1];
3992 struct net_device *dev;
3997 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4002 ndm = nlmsg_data(nlh);
4003 if (ndm->ndm_ifindex == 0) {
4004 NL_SET_ERR_MSG(extack, "invalid ifindex");
4008 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4010 NL_SET_ERR_MSG(extack, "unknown ifindex");
4014 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4015 NL_SET_ERR_MSG(extack, "invalid address");
4019 if (dev->type != ARPHRD_ETHER) {
4020 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4024 addr = nla_data(tb[NDA_LLADDR]);
4026 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4032 /* Support fdb on master device the net/bridge default case */
4033 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4034 netif_is_bridge_port(dev)) {
4035 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4036 const struct net_device_ops *ops = br_dev->netdev_ops;
4038 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4039 nlh->nlmsg_flags, extack);
4043 ndm->ndm_flags &= ~NTF_MASTER;
4046 /* Embedded bridge, macvlan, and any other device support */
4047 if ((ndm->ndm_flags & NTF_SELF)) {
4048 if (dev->netdev_ops->ndo_fdb_add)
4049 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4054 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4058 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4060 ndm->ndm_flags &= ~NTF_SELF;
4068 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4070 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4071 struct nlattr *tb[],
4072 struct net_device *dev,
4073 const unsigned char *addr, u16 vid)
4077 /* If aging addresses are supported device will need to
4078 * implement its own handler for this.
4080 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4081 pr_info("%s: FDB only supports static addresses\n", dev->name);
4085 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4086 err = dev_uc_del(dev, addr);
4087 else if (is_multicast_ether_addr(addr))
4088 err = dev_mc_del(dev, addr);
4092 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4094 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4095 struct netlink_ext_ack *extack)
4097 struct net *net = sock_net(skb->sk);
4099 struct nlattr *tb[NDA_MAX+1];
4100 struct net_device *dev;
4105 if (!netlink_capable(skb, CAP_NET_ADMIN))
4108 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4113 ndm = nlmsg_data(nlh);
4114 if (ndm->ndm_ifindex == 0) {
4115 NL_SET_ERR_MSG(extack, "invalid ifindex");
4119 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4121 NL_SET_ERR_MSG(extack, "unknown ifindex");
4125 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4126 NL_SET_ERR_MSG(extack, "invalid address");
4130 if (dev->type != ARPHRD_ETHER) {
4131 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4135 addr = nla_data(tb[NDA_LLADDR]);
4137 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4143 /* Support fdb on master device the net/bridge default case */
4144 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4145 netif_is_bridge_port(dev)) {
4146 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4147 const struct net_device_ops *ops = br_dev->netdev_ops;
4149 if (ops->ndo_fdb_del)
4150 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4155 ndm->ndm_flags &= ~NTF_MASTER;
4158 /* Embedded bridge, macvlan, and any other device support */
4159 if (ndm->ndm_flags & NTF_SELF) {
4160 if (dev->netdev_ops->ndo_fdb_del)
4161 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4164 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4167 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4169 ndm->ndm_flags &= ~NTF_SELF;
4176 static int nlmsg_populate_fdb(struct sk_buff *skb,
4177 struct netlink_callback *cb,
4178 struct net_device *dev,
4180 struct netdev_hw_addr_list *list)
4182 struct netdev_hw_addr *ha;
4186 portid = NETLINK_CB(cb->skb).portid;
4187 seq = cb->nlh->nlmsg_seq;
4189 list_for_each_entry(ha, &list->list, list) {
4190 if (*idx < cb->args[2])
4193 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4195 RTM_NEWNEIGH, NTF_SELF,
4196 NLM_F_MULTI, NUD_PERMANENT);
4206 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4207 * @skb: socket buffer to store message in
4208 * @cb: netlink callback
4210 * @filter_dev: ignored
4211 * @idx: the number of FDB table entries dumped is added to *@idx
4213 * Default netdevice operation to dump the existing unicast address list.
4214 * Returns number of addresses from list put in skb.
4216 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4217 struct netlink_callback *cb,
4218 struct net_device *dev,
4219 struct net_device *filter_dev,
4224 if (dev->type != ARPHRD_ETHER)
4227 netif_addr_lock_bh(dev);
4228 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4231 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4233 netif_addr_unlock_bh(dev);
4236 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4238 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4239 int *br_idx, int *brport_idx,
4240 struct netlink_ext_ack *extack)
4242 struct nlattr *tb[NDA_MAX + 1];
4246 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4247 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4251 ndm = nlmsg_data(nlh);
4252 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4253 ndm->ndm_flags || ndm->ndm_type) {
4254 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4258 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4259 NDA_MAX, NULL, extack);
4263 *brport_idx = ndm->ndm_ifindex;
4264 for (i = 0; i <= NDA_MAX; ++i) {
4270 if (nla_len(tb[i]) != sizeof(u32)) {
4271 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4274 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4277 if (nla_len(tb[i]) != sizeof(u32)) {
4278 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4281 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4284 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4292 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4293 int *br_idx, int *brport_idx,
4294 struct netlink_ext_ack *extack)
4296 struct nlattr *tb[IFLA_MAX+1];
4299 /* A hack to preserve kernel<->userspace interface.
4300 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4301 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4302 * So, check for ndmsg with an optional u32 attribute (not used here).
4303 * Fortunately these sizes don't conflict with the size of ifinfomsg
4304 * with an optional attribute.
4306 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4307 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4308 nla_attr_size(sizeof(u32)))) {
4309 struct ifinfomsg *ifm;
4311 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4312 tb, IFLA_MAX, ifla_policy,
4316 } else if (err == 0) {
4317 if (tb[IFLA_MASTER])
4318 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4321 ifm = nlmsg_data(nlh);
4322 *brport_idx = ifm->ifi_index;
4327 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4329 struct net_device *dev;
4330 struct net_device *br_dev = NULL;
4331 const struct net_device_ops *ops = NULL;
4332 const struct net_device_ops *cops = NULL;
4333 struct net *net = sock_net(skb->sk);
4334 struct hlist_head *head;
4342 if (cb->strict_check)
4343 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4346 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4352 br_dev = __dev_get_by_index(net, br_idx);
4356 ops = br_dev->netdev_ops;
4360 s_idx = cb->args[1];
4362 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4364 head = &net->dev_index_head[h];
4365 hlist_for_each_entry(dev, head, index_hlist) {
4367 if (brport_idx && (dev->ifindex != brport_idx))
4370 if (!br_idx) { /* user did not specify a specific bridge */
4371 if (netif_is_bridge_port(dev)) {
4372 br_dev = netdev_master_upper_dev_get(dev);
4373 cops = br_dev->netdev_ops;
4376 if (dev != br_dev &&
4377 !netif_is_bridge_port(dev))
4380 if (br_dev != netdev_master_upper_dev_get(dev) &&
4381 !(dev->priv_flags & IFF_EBRIDGE))
4389 if (netif_is_bridge_port(dev)) {
4390 if (cops && cops->ndo_fdb_dump) {
4391 err = cops->ndo_fdb_dump(skb, cb,
4394 if (err == -EMSGSIZE)
4399 if (dev->netdev_ops->ndo_fdb_dump)
4400 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4404 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4406 if (err == -EMSGSIZE)
4411 /* reset fdb offset to 0 for rest of the interfaces */
4427 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4428 struct nlattr **tb, u8 *ndm_flags,
4429 int *br_idx, int *brport_idx, u8 **addr,
4430 u16 *vid, struct netlink_ext_ack *extack)
4435 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4436 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4440 ndm = nlmsg_data(nlh);
4441 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4443 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4447 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4448 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4452 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4453 NDA_MAX, nda_policy, extack);
4457 *ndm_flags = ndm->ndm_flags;
4458 *brport_idx = ndm->ndm_ifindex;
4459 for (i = 0; i <= NDA_MAX; ++i) {
4465 *br_idx = nla_get_u32(tb[i]);
4468 if (nla_len(tb[i]) != ETH_ALEN) {
4469 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4472 *addr = nla_data(tb[i]);
4475 err = fdb_vid_parse(tb[i], vid, extack);
4482 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4490 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4491 struct netlink_ext_ack *extack)
4493 struct net_device *dev = NULL, *br_dev = NULL;
4494 const struct net_device_ops *ops = NULL;
4495 struct net *net = sock_net(in_skb->sk);
4496 struct nlattr *tb[NDA_MAX + 1];
4497 struct sk_buff *skb;
4505 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4506 &brport_idx, &addr, &vid, extack);
4511 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4516 dev = __dev_get_by_index(net, brport_idx);
4518 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4525 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4529 br_dev = __dev_get_by_index(net, br_idx);
4531 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4534 ops = br_dev->netdev_ops;
4538 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4539 if (!netif_is_bridge_port(dev)) {
4540 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4543 br_dev = netdev_master_upper_dev_get(dev);
4545 NL_SET_ERR_MSG(extack, "Master of device not found");
4548 ops = br_dev->netdev_ops;
4550 if (!(ndm_flags & NTF_SELF)) {
4551 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4554 ops = dev->netdev_ops;
4558 if (!br_dev && !dev) {
4559 NL_SET_ERR_MSG(extack, "No device specified");
4563 if (!ops || !ops->ndo_fdb_get) {
4564 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4568 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4574 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4575 NETLINK_CB(in_skb).portid,
4576 nlh->nlmsg_seq, extack);
4580 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4586 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4587 unsigned int attrnum, unsigned int flag)
4590 return nla_put_u8(skb, attrnum, !!(flags & flag));
4594 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4595 struct net_device *dev, u16 mode,
4596 u32 flags, u32 mask, int nlflags,
4598 int (*vlan_fill)(struct sk_buff *skb,
4599 struct net_device *dev,
4602 struct nlmsghdr *nlh;
4603 struct ifinfomsg *ifm;
4604 struct nlattr *br_afspec;
4605 struct nlattr *protinfo;
4606 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4607 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4610 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4614 ifm = nlmsg_data(nlh);
4615 ifm->ifi_family = AF_BRIDGE;
4617 ifm->ifi_type = dev->type;
4618 ifm->ifi_index = dev->ifindex;
4619 ifm->ifi_flags = dev_get_flags(dev);
4620 ifm->ifi_change = 0;
4623 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4624 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4625 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4627 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4629 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4630 (dev->ifindex != dev_get_iflink(dev) &&
4631 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4632 goto nla_put_failure;
4634 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4636 goto nla_put_failure;
4638 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4639 nla_nest_cancel(skb, br_afspec);
4640 goto nla_put_failure;
4643 if (mode != BRIDGE_MODE_UNDEF) {
4644 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4645 nla_nest_cancel(skb, br_afspec);
4646 goto nla_put_failure;
4650 err = vlan_fill(skb, dev, filter_mask);
4652 nla_nest_cancel(skb, br_afspec);
4653 goto nla_put_failure;
4656 nla_nest_end(skb, br_afspec);
4658 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4660 goto nla_put_failure;
4662 if (brport_nla_put_flag(skb, flags, mask,
4663 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4664 brport_nla_put_flag(skb, flags, mask,
4665 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4666 brport_nla_put_flag(skb, flags, mask,
4667 IFLA_BRPORT_FAST_LEAVE,
4668 BR_MULTICAST_FAST_LEAVE) ||
4669 brport_nla_put_flag(skb, flags, mask,
4670 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4671 brport_nla_put_flag(skb, flags, mask,
4672 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4673 brport_nla_put_flag(skb, flags, mask,
4674 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4675 brport_nla_put_flag(skb, flags, mask,
4676 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4677 brport_nla_put_flag(skb, flags, mask,
4678 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4679 brport_nla_put_flag(skb, flags, mask,
4680 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4681 brport_nla_put_flag(skb, flags, mask,
4682 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4683 nla_nest_cancel(skb, protinfo);
4684 goto nla_put_failure;
4687 nla_nest_end(skb, protinfo);
4689 nlmsg_end(skb, nlh);
4692 nlmsg_cancel(skb, nlh);
4693 return err ? err : -EMSGSIZE;
4695 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4697 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4698 bool strict_check, u32 *filter_mask,
4699 struct netlink_ext_ack *extack)
4701 struct nlattr *tb[IFLA_MAX+1];
4705 struct ifinfomsg *ifm;
4707 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4708 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4712 ifm = nlmsg_data(nlh);
4713 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4714 ifm->ifi_change || ifm->ifi_index) {
4715 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4719 err = nlmsg_parse_deprecated_strict(nlh,
4720 sizeof(struct ifinfomsg),
4721 tb, IFLA_MAX, ifla_policy,
4724 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4725 tb, IFLA_MAX, ifla_policy,
4731 /* new attributes should only be added with strict checking */
4732 for (i = 0; i <= IFLA_MAX; ++i) {
4738 *filter_mask = nla_get_u32(tb[i]);
4742 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4751 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4753 const struct nlmsghdr *nlh = cb->nlh;
4754 struct net *net = sock_net(skb->sk);
4755 struct net_device *dev;
4757 u32 portid = NETLINK_CB(cb->skb).portid;
4758 u32 seq = nlh->nlmsg_seq;
4759 u32 filter_mask = 0;
4762 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4764 if (err < 0 && cb->strict_check)
4768 for_each_netdev_rcu(net, dev) {
4769 const struct net_device_ops *ops = dev->netdev_ops;
4770 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4772 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4773 if (idx >= cb->args[0]) {
4774 err = br_dev->netdev_ops->ndo_bridge_getlink(
4775 skb, portid, seq, dev,
4776 filter_mask, NLM_F_MULTI);
4777 if (err < 0 && err != -EOPNOTSUPP) {
4778 if (likely(skb->len))
4787 if (ops->ndo_bridge_getlink) {
4788 if (idx >= cb->args[0]) {
4789 err = ops->ndo_bridge_getlink(skb, portid,
4793 if (err < 0 && err != -EOPNOTSUPP) {
4794 if (likely(skb->len))
4811 static inline size_t bridge_nlmsg_size(void)
4813 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4814 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4815 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4816 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4817 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
4818 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
4819 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
4820 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
4821 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4822 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
4823 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
4826 static int rtnl_bridge_notify(struct net_device *dev)
4828 struct net *net = dev_net(dev);
4829 struct sk_buff *skb;
4830 int err = -EOPNOTSUPP;
4832 if (!dev->netdev_ops->ndo_bridge_getlink)
4835 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4841 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4845 /* Notification info is only filled for bridge ports, not the bridge
4846 * device itself. Therefore, a zero notification length is valid and
4847 * should not result in an error.
4852 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4855 WARN_ON(err == -EMSGSIZE);
4858 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4862 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4863 struct netlink_ext_ack *extack)
4865 struct net *net = sock_net(skb->sk);
4866 struct ifinfomsg *ifm;
4867 struct net_device *dev;
4868 struct nlattr *br_spec, *attr = NULL;
4869 int rem, err = -EOPNOTSUPP;
4871 bool have_flags = false;
4873 if (nlmsg_len(nlh) < sizeof(*ifm))
4876 ifm = nlmsg_data(nlh);
4877 if (ifm->ifi_family != AF_BRIDGE)
4878 return -EPFNOSUPPORT;
4880 dev = __dev_get_by_index(net, ifm->ifi_index);
4882 NL_SET_ERR_MSG(extack, "unknown ifindex");
4886 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4888 nla_for_each_nested(attr, br_spec, rem) {
4889 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4890 if (nla_len(attr) < sizeof(flags))
4894 flags = nla_get_u16(attr);
4900 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4901 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4903 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4908 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4913 flags &= ~BRIDGE_FLAGS_MASTER;
4916 if ((flags & BRIDGE_FLAGS_SELF)) {
4917 if (!dev->netdev_ops->ndo_bridge_setlink)
4920 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4924 flags &= ~BRIDGE_FLAGS_SELF;
4926 /* Generate event to notify upper layer of bridge
4929 err = rtnl_bridge_notify(dev);
4934 memcpy(nla_data(attr), &flags, sizeof(flags));
4939 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4940 struct netlink_ext_ack *extack)
4942 struct net *net = sock_net(skb->sk);
4943 struct ifinfomsg *ifm;
4944 struct net_device *dev;
4945 struct nlattr *br_spec, *attr = NULL;
4946 int rem, err = -EOPNOTSUPP;
4948 bool have_flags = false;
4950 if (nlmsg_len(nlh) < sizeof(*ifm))
4953 ifm = nlmsg_data(nlh);
4954 if (ifm->ifi_family != AF_BRIDGE)
4955 return -EPFNOSUPPORT;
4957 dev = __dev_get_by_index(net, ifm->ifi_index);
4959 NL_SET_ERR_MSG(extack, "unknown ifindex");
4963 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4965 nla_for_each_nested(attr, br_spec, rem) {
4966 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4967 if (nla_len(attr) < sizeof(flags))
4971 flags = nla_get_u16(attr);
4977 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4978 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4980 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4985 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4989 flags &= ~BRIDGE_FLAGS_MASTER;
4992 if ((flags & BRIDGE_FLAGS_SELF)) {
4993 if (!dev->netdev_ops->ndo_bridge_dellink)
4996 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5000 flags &= ~BRIDGE_FLAGS_SELF;
5002 /* Generate event to notify upper layer of bridge
5005 err = rtnl_bridge_notify(dev);
5010 memcpy(nla_data(attr), &flags, sizeof(flags));
5015 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5017 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5018 (!idxattr || idxattr == attrid);
5021 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5022 static int rtnl_get_offload_stats_attr_size(int attr_id)
5025 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5026 return sizeof(struct rtnl_link_stats64);
5032 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5035 struct nlattr *attr = NULL;
5040 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5041 dev->netdev_ops->ndo_get_offload_stats))
5044 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5045 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5046 if (attr_id < *prividx)
5049 size = rtnl_get_offload_stats_attr_size(attr_id);
5053 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5056 attr = nla_reserve_64bit(skb, attr_id, size,
5057 IFLA_OFFLOAD_XSTATS_UNSPEC);
5059 goto nla_put_failure;
5061 attr_data = nla_data(attr);
5062 memset(attr_data, 0, size);
5063 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5066 goto get_offload_stats_failure;
5077 get_offload_stats_failure:
5082 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5088 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5089 dev->netdev_ops->ndo_get_offload_stats))
5092 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5093 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5094 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5096 size = rtnl_get_offload_stats_attr_size(attr_id);
5097 nla_size += nla_total_size_64bit(size);
5101 nla_size += nla_total_size(0);
5106 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5107 int type, u32 pid, u32 seq, u32 change,
5108 unsigned int flags, unsigned int filter_mask,
5109 int *idxattr, int *prividx)
5111 struct if_stats_msg *ifsm;
5112 struct nlmsghdr *nlh;
5113 struct nlattr *attr;
5114 int s_prividx = *prividx;
5119 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5123 ifsm = nlmsg_data(nlh);
5124 ifsm->family = PF_UNSPEC;
5127 ifsm->ifindex = dev->ifindex;
5128 ifsm->filter_mask = filter_mask;
5130 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5131 struct rtnl_link_stats64 *sp;
5133 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5134 sizeof(struct rtnl_link_stats64),
5137 goto nla_put_failure;
5139 sp = nla_data(attr);
5140 dev_get_stats(dev, sp);
5143 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5144 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5146 if (ops && ops->fill_linkxstats) {
5147 *idxattr = IFLA_STATS_LINK_XSTATS;
5148 attr = nla_nest_start_noflag(skb,
5149 IFLA_STATS_LINK_XSTATS);
5151 goto nla_put_failure;
5153 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5154 nla_nest_end(skb, attr);
5156 goto nla_put_failure;
5161 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5163 const struct rtnl_link_ops *ops = NULL;
5164 const struct net_device *master;
5166 master = netdev_master_upper_dev_get(dev);
5168 ops = master->rtnl_link_ops;
5169 if (ops && ops->fill_linkxstats) {
5170 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5171 attr = nla_nest_start_noflag(skb,
5172 IFLA_STATS_LINK_XSTATS_SLAVE);
5174 goto nla_put_failure;
5176 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5177 nla_nest_end(skb, attr);
5179 goto nla_put_failure;
5184 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5186 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5187 attr = nla_nest_start_noflag(skb,
5188 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5190 goto nla_put_failure;
5192 err = rtnl_get_offload_stats(skb, dev, prividx);
5193 if (err == -ENODATA)
5194 nla_nest_cancel(skb, attr);
5196 nla_nest_end(skb, attr);
5198 if (err && err != -ENODATA)
5199 goto nla_put_failure;
5203 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5204 struct rtnl_af_ops *af_ops;
5206 *idxattr = IFLA_STATS_AF_SPEC;
5207 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5209 goto nla_put_failure;
5212 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5213 if (af_ops->fill_stats_af) {
5217 af = nla_nest_start_noflag(skb,
5221 goto nla_put_failure;
5223 err = af_ops->fill_stats_af(skb, dev);
5225 if (err == -ENODATA) {
5226 nla_nest_cancel(skb, af);
5227 } else if (err < 0) {
5229 goto nla_put_failure;
5232 nla_nest_end(skb, af);
5237 nla_nest_end(skb, attr);
5242 nlmsg_end(skb, nlh);
5247 /* not a multi message or no progress mean a real error */
5248 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5249 nlmsg_cancel(skb, nlh);
5251 nlmsg_end(skb, nlh);
5256 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5261 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5262 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5264 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5265 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5266 int attr = IFLA_STATS_LINK_XSTATS;
5268 if (ops && ops->get_linkxstats_size) {
5269 size += nla_total_size(ops->get_linkxstats_size(dev,
5271 /* for IFLA_STATS_LINK_XSTATS */
5272 size += nla_total_size(0);
5276 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5277 struct net_device *_dev = (struct net_device *)dev;
5278 const struct rtnl_link_ops *ops = NULL;
5279 const struct net_device *master;
5281 /* netdev_master_upper_dev_get can't take const */
5282 master = netdev_master_upper_dev_get(_dev);
5284 ops = master->rtnl_link_ops;
5285 if (ops && ops->get_linkxstats_size) {
5286 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5288 size += nla_total_size(ops->get_linkxstats_size(dev,
5290 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5291 size += nla_total_size(0);
5295 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5296 size += rtnl_get_offload_stats_size(dev);
5298 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5299 struct rtnl_af_ops *af_ops;
5301 /* for IFLA_STATS_AF_SPEC */
5302 size += nla_total_size(0);
5305 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5306 if (af_ops->get_stats_af_size) {
5307 size += nla_total_size(
5308 af_ops->get_stats_af_size(dev));
5311 size += nla_total_size(0);
5320 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5321 bool is_dump, struct netlink_ext_ack *extack)
5323 struct if_stats_msg *ifsm;
5325 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5326 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5333 ifsm = nlmsg_data(nlh);
5335 /* only requests using strict checks can pass data to influence
5336 * the dump. The legacy exception is filter_mask.
5338 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5339 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5342 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5343 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5346 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5347 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5354 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5355 struct netlink_ext_ack *extack)
5357 struct net *net = sock_net(skb->sk);
5358 struct net_device *dev = NULL;
5359 int idxattr = 0, prividx = 0;
5360 struct if_stats_msg *ifsm;
5361 struct sk_buff *nskb;
5365 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5370 ifsm = nlmsg_data(nlh);
5371 if (ifsm->ifindex > 0)
5372 dev = __dev_get_by_index(net, ifsm->ifindex);
5379 filter_mask = ifsm->filter_mask;
5383 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5387 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5388 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5389 0, filter_mask, &idxattr, &prividx);
5391 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5392 WARN_ON(err == -EMSGSIZE);
5395 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5401 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5403 struct netlink_ext_ack *extack = cb->extack;
5404 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5405 struct net *net = sock_net(skb->sk);
5406 unsigned int flags = NLM_F_MULTI;
5407 struct if_stats_msg *ifsm;
5408 struct hlist_head *head;
5409 struct net_device *dev;
5410 u32 filter_mask = 0;
5414 s_idx = cb->args[1];
5415 s_idxattr = cb->args[2];
5416 s_prividx = cb->args[3];
5418 cb->seq = net->dev_base_seq;
5420 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5424 ifsm = nlmsg_data(cb->nlh);
5425 filter_mask = ifsm->filter_mask;
5427 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5431 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5433 head = &net->dev_index_head[h];
5434 hlist_for_each_entry(dev, head, index_hlist) {
5437 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5438 NETLINK_CB(cb->skb).portid,
5439 cb->nlh->nlmsg_seq, 0,
5441 &s_idxattr, &s_prividx);
5442 /* If we ran out of room on the first message,
5445 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5451 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5457 cb->args[3] = s_prividx;
5458 cb->args[2] = s_idxattr;
5465 /* Process one rtnetlink message. */
5467 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5468 struct netlink_ext_ack *extack)
5470 struct net *net = sock_net(skb->sk);
5471 struct rtnl_link *link;
5472 struct module *owner;
5473 int err = -EOPNOTSUPP;
5474 rtnl_doit_func doit;
5480 type = nlh->nlmsg_type;
5486 /* All the messages must have at least 1 byte length */
5487 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5490 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5493 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5497 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5499 rtnl_dumpit_func dumpit;
5500 u32 min_dump_alloc = 0;
5502 link = rtnl_get_link(family, type);
5503 if (!link || !link->dumpit) {
5505 link = rtnl_get_link(family, type);
5506 if (!link || !link->dumpit)
5509 owner = link->owner;
5510 dumpit = link->dumpit;
5512 if (type == RTM_GETLINK - RTM_BASE)
5513 min_dump_alloc = rtnl_calcit(skb, nlh);
5516 /* need to do this before rcu_read_unlock() */
5517 if (!try_module_get(owner))
5518 err = -EPROTONOSUPPORT;
5524 struct netlink_dump_control c = {
5526 .min_dump_alloc = min_dump_alloc,
5529 err = netlink_dump_start(rtnl, skb, nlh, &c);
5530 /* netlink_dump_start() will keep a reference on
5531 * module if dump is still in progress.
5538 link = rtnl_get_link(family, type);
5539 if (!link || !link->doit) {
5541 link = rtnl_get_link(PF_UNSPEC, type);
5542 if (!link || !link->doit)
5546 owner = link->owner;
5547 if (!try_module_get(owner)) {
5548 err = -EPROTONOSUPPORT;
5552 flags = link->flags;
5553 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5557 err = doit(skb, nlh, extack);
5564 link = rtnl_get_link(family, type);
5565 if (link && link->doit)
5566 err = link->doit(skb, nlh, extack);
5582 static void rtnetlink_rcv(struct sk_buff *skb)
5584 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5587 static int rtnetlink_bind(struct net *net, int group)
5590 case RTNLGRP_IPV4_MROUTE_R:
5591 case RTNLGRP_IPV6_MROUTE_R:
5592 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5599 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5601 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5605 case NETDEV_CHANGEMTU:
5606 case NETDEV_CHANGEADDR:
5607 case NETDEV_CHANGENAME:
5608 case NETDEV_FEAT_CHANGE:
5609 case NETDEV_BONDING_FAILOVER:
5610 case NETDEV_POST_TYPE_CHANGE:
5611 case NETDEV_NOTIFY_PEERS:
5612 case NETDEV_CHANGEUPPER:
5613 case NETDEV_RESEND_IGMP:
5614 case NETDEV_CHANGEINFODATA:
5615 case NETDEV_CHANGELOWERSTATE:
5616 case NETDEV_CHANGE_TX_QUEUE_LEN:
5617 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5618 GFP_KERNEL, NULL, 0);
5626 static struct notifier_block rtnetlink_dev_notifier = {
5627 .notifier_call = rtnetlink_event,
5631 static int __net_init rtnetlink_net_init(struct net *net)
5634 struct netlink_kernel_cfg cfg = {
5635 .groups = RTNLGRP_MAX,
5636 .input = rtnetlink_rcv,
5637 .cb_mutex = &rtnl_mutex,
5638 .flags = NL_CFG_F_NONROOT_RECV,
5639 .bind = rtnetlink_bind,
5642 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5649 static void __net_exit rtnetlink_net_exit(struct net *net)
5651 netlink_kernel_release(net->rtnl);
5655 static struct pernet_operations rtnetlink_net_ops = {
5656 .init = rtnetlink_net_init,
5657 .exit = rtnetlink_net_exit,
5660 void __init rtnetlink_init(void)
5662 if (register_pernet_subsys(&rtnetlink_net_ops))
5663 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5665 register_netdevice_notifier(&rtnetlink_dev_notifier);
5667 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5668 rtnl_dump_ifinfo, 0);
5669 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5670 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5671 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5673 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5674 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5675 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5677 rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5678 rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5680 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5681 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5682 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5684 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5685 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5686 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5688 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,