2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Routing netlink socket interface: protocol independent part.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
19 #include <linux/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
45 #include <linux/uaccess.h>
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
51 #include <net/protocol.h>
53 #include <net/route.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
64 rtnl_dumpit_func dumpit;
65 rtnl_calcit_func calcit;
68 static DEFINE_MUTEX(rtnl_mutex);
72 mutex_lock(&rtnl_mutex);
74 EXPORT_SYMBOL(rtnl_lock);
76 static struct sk_buff *defer_kfree_skb_list;
77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
80 tail->next = defer_kfree_skb_list;
81 defer_kfree_skb_list = head;
84 EXPORT_SYMBOL(rtnl_kfree_skbs);
86 void __rtnl_unlock(void)
88 struct sk_buff *head = defer_kfree_skb_list;
90 defer_kfree_skb_list = NULL;
92 mutex_unlock(&rtnl_mutex);
95 struct sk_buff *next = head->next;
103 void rtnl_unlock(void)
105 /* This fellow will unlock it for us. */
108 EXPORT_SYMBOL(rtnl_unlock);
110 int rtnl_trylock(void)
112 return mutex_trylock(&rtnl_mutex);
114 EXPORT_SYMBOL(rtnl_trylock);
116 int rtnl_is_locked(void)
118 return mutex_is_locked(&rtnl_mutex);
120 EXPORT_SYMBOL(rtnl_is_locked);
122 #ifdef CONFIG_PROVE_LOCKING
123 bool lockdep_rtnl_is_held(void)
125 return lockdep_is_held(&rtnl_mutex);
127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
130 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
132 static inline int rtm_msgindex(int msgtype)
134 int msgindex = msgtype - RTM_BASE;
137 * msgindex < 0 implies someone tried to register a netlink
138 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
139 * the message type has not been added to linux/rtnetlink.h
141 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
146 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
148 struct rtnl_link *tab;
150 if (protocol <= RTNL_FAMILY_MAX)
151 tab = rtnl_msg_handlers[protocol];
155 if (tab == NULL || tab[msgindex].doit == NULL)
156 tab = rtnl_msg_handlers[PF_UNSPEC];
158 return tab[msgindex].doit;
161 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
163 struct rtnl_link *tab;
165 if (protocol <= RTNL_FAMILY_MAX)
166 tab = rtnl_msg_handlers[protocol];
170 if (tab == NULL || tab[msgindex].dumpit == NULL)
171 tab = rtnl_msg_handlers[PF_UNSPEC];
173 return tab[msgindex].dumpit;
176 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
178 struct rtnl_link *tab;
180 if (protocol <= RTNL_FAMILY_MAX)
181 tab = rtnl_msg_handlers[protocol];
185 if (tab == NULL || tab[msgindex].calcit == NULL)
186 tab = rtnl_msg_handlers[PF_UNSPEC];
188 return tab[msgindex].calcit;
192 * __rtnl_register - Register a rtnetlink message type
193 * @protocol: Protocol family or PF_UNSPEC
194 * @msgtype: rtnetlink message type
195 * @doit: Function pointer called for each request message
196 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
197 * @calcit: Function pointer to calc size of dump message
199 * Registers the specified function pointers (at least one of them has
200 * to be non-NULL) to be called whenever a request message for the
201 * specified protocol family and message type is received.
203 * The special protocol family PF_UNSPEC may be used to define fallback
204 * function pointers for the case when no entry for the specific protocol
207 * Returns 0 on success or a negative error code.
209 int __rtnl_register(int protocol, int msgtype,
210 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
211 rtnl_calcit_func calcit)
213 struct rtnl_link *tab;
216 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
217 msgindex = rtm_msgindex(msgtype);
219 tab = rtnl_msg_handlers[protocol];
221 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
225 rtnl_msg_handlers[protocol] = tab;
229 tab[msgindex].doit = doit;
232 tab[msgindex].dumpit = dumpit;
235 tab[msgindex].calcit = calcit;
239 EXPORT_SYMBOL_GPL(__rtnl_register);
242 * rtnl_register - Register a rtnetlink message type
244 * Identical to __rtnl_register() but panics on failure. This is useful
245 * as failure of this function is very unlikely, it can only happen due
246 * to lack of memory when allocating the chain to store all message
247 * handlers for a protocol. Meant for use in init functions where lack
248 * of memory implies no sense in continuing.
250 void rtnl_register(int protocol, int msgtype,
251 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
252 rtnl_calcit_func calcit)
254 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
255 panic("Unable to register rtnetlink message handler, "
256 "protocol = %d, message type = %d\n",
259 EXPORT_SYMBOL_GPL(rtnl_register);
262 * rtnl_unregister - Unregister a rtnetlink message type
263 * @protocol: Protocol family or PF_UNSPEC
264 * @msgtype: rtnetlink message type
266 * Returns 0 on success or a negative error code.
268 int rtnl_unregister(int protocol, int msgtype)
272 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
273 msgindex = rtm_msgindex(msgtype);
275 if (rtnl_msg_handlers[protocol] == NULL)
278 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
279 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
280 rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
284 EXPORT_SYMBOL_GPL(rtnl_unregister);
287 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
288 * @protocol : Protocol family or PF_UNSPEC
290 * Identical to calling rtnl_unregster() for all registered message types
291 * of a certain protocol family.
293 void rtnl_unregister_all(int protocol)
295 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
297 kfree(rtnl_msg_handlers[protocol]);
298 rtnl_msg_handlers[protocol] = NULL;
300 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
302 static LIST_HEAD(link_ops);
304 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
306 const struct rtnl_link_ops *ops;
308 list_for_each_entry(ops, &link_ops, list) {
309 if (!strcmp(ops->kind, kind))
316 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
317 * @ops: struct rtnl_link_ops * to register
319 * The caller must hold the rtnl_mutex. This function should be used
320 * by drivers that create devices during module initialization. It
321 * must be called before registering the devices.
323 * Returns 0 on success or a negative error code.
325 int __rtnl_link_register(struct rtnl_link_ops *ops)
327 if (rtnl_link_ops_get(ops->kind))
330 /* The check for setup is here because if ops
331 * does not have that filled up, it is not possible
332 * to use the ops for creating device. So do not
333 * fill up dellink as well. That disables rtnl_dellink.
335 if (ops->setup && !ops->dellink)
336 ops->dellink = unregister_netdevice_queue;
338 list_add_tail(&ops->list, &link_ops);
341 EXPORT_SYMBOL_GPL(__rtnl_link_register);
344 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
345 * @ops: struct rtnl_link_ops * to register
347 * Returns 0 on success or a negative error code.
349 int rtnl_link_register(struct rtnl_link_ops *ops)
354 err = __rtnl_link_register(ops);
358 EXPORT_SYMBOL_GPL(rtnl_link_register);
360 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
362 struct net_device *dev;
363 LIST_HEAD(list_kill);
365 for_each_netdev(net, dev) {
366 if (dev->rtnl_link_ops == ops)
367 ops->dellink(dev, &list_kill);
369 unregister_netdevice_many(&list_kill);
373 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
374 * @ops: struct rtnl_link_ops * to unregister
376 * The caller must hold the rtnl_mutex.
378 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
383 __rtnl_kill_links(net, ops);
385 list_del(&ops->list);
387 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
389 /* Return with the rtnl_lock held when there are no network
390 * devices unregistering in any network namespace.
392 static void rtnl_lock_unregistering_all(void)
396 DEFINE_WAIT_FUNC(wait, woken_wake_function);
398 add_wait_queue(&netdev_unregistering_wq, &wait);
400 unregistering = false;
403 if (net->dev_unreg_count > 0) {
404 unregistering = true;
412 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
414 remove_wait_queue(&netdev_unregistering_wq, &wait);
418 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
419 * @ops: struct rtnl_link_ops * to unregister
421 void rtnl_link_unregister(struct rtnl_link_ops *ops)
423 /* Close the race with cleanup_net() */
424 mutex_lock(&net_mutex);
425 rtnl_lock_unregistering_all();
426 __rtnl_link_unregister(ops);
428 mutex_unlock(&net_mutex);
430 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
432 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
434 struct net_device *master_dev;
435 const struct rtnl_link_ops *ops;
437 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
440 ops = master_dev->rtnl_link_ops;
441 if (!ops || !ops->get_slave_size)
443 /* IFLA_INFO_SLAVE_DATA + nested data */
444 return nla_total_size(sizeof(struct nlattr)) +
445 ops->get_slave_size(master_dev, dev);
448 static size_t rtnl_link_get_size(const struct net_device *dev)
450 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
456 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
457 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
460 /* IFLA_INFO_DATA + nested data */
461 size += nla_total_size(sizeof(struct nlattr)) +
464 if (ops->get_xstats_size)
465 /* IFLA_INFO_XSTATS */
466 size += nla_total_size(ops->get_xstats_size(dev));
468 size += rtnl_link_get_slave_info_data_size(dev);
473 static LIST_HEAD(rtnl_af_ops);
475 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
477 const struct rtnl_af_ops *ops;
479 list_for_each_entry(ops, &rtnl_af_ops, list) {
480 if (ops->family == family)
488 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
489 * @ops: struct rtnl_af_ops * to register
491 * Returns 0 on success or a negative error code.
493 void rtnl_af_register(struct rtnl_af_ops *ops)
496 list_add_tail(&ops->list, &rtnl_af_ops);
499 EXPORT_SYMBOL_GPL(rtnl_af_register);
502 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
503 * @ops: struct rtnl_af_ops * to unregister
505 * The caller must hold the rtnl_mutex.
507 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
509 list_del(&ops->list);
511 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
514 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
515 * @ops: struct rtnl_af_ops * to unregister
517 void rtnl_af_unregister(struct rtnl_af_ops *ops)
520 __rtnl_af_unregister(ops);
523 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
525 static size_t rtnl_link_get_af_size(const struct net_device *dev,
528 struct rtnl_af_ops *af_ops;
532 size = nla_total_size(sizeof(struct nlattr));
534 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
535 if (af_ops->get_link_af_size) {
536 /* AF_* + nested data */
537 size += nla_total_size(sizeof(struct nlattr)) +
538 af_ops->get_link_af_size(dev, ext_filter_mask);
545 static bool rtnl_have_link_slave_info(const struct net_device *dev)
547 struct net_device *master_dev;
549 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
550 if (master_dev && master_dev->rtnl_link_ops)
555 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
556 const struct net_device *dev)
558 struct net_device *master_dev;
559 const struct rtnl_link_ops *ops;
560 struct nlattr *slave_data;
563 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
566 ops = master_dev->rtnl_link_ops;
569 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
571 if (ops->fill_slave_info) {
572 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
575 err = ops->fill_slave_info(skb, master_dev, dev);
577 goto err_cancel_slave_data;
578 nla_nest_end(skb, slave_data);
582 err_cancel_slave_data:
583 nla_nest_cancel(skb, slave_data);
587 static int rtnl_link_info_fill(struct sk_buff *skb,
588 const struct net_device *dev)
590 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
596 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
598 if (ops->fill_xstats) {
599 err = ops->fill_xstats(skb, dev);
603 if (ops->fill_info) {
604 data = nla_nest_start(skb, IFLA_INFO_DATA);
607 err = ops->fill_info(skb, dev);
609 goto err_cancel_data;
610 nla_nest_end(skb, data);
615 nla_nest_cancel(skb, data);
619 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
621 struct nlattr *linkinfo;
624 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
625 if (linkinfo == NULL)
628 err = rtnl_link_info_fill(skb, dev);
630 goto err_cancel_link;
632 err = rtnl_link_slave_info_fill(skb, dev);
634 goto err_cancel_link;
636 nla_nest_end(skb, linkinfo);
640 nla_nest_cancel(skb, linkinfo);
645 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
647 struct sock *rtnl = net->rtnl;
650 NETLINK_CB(skb).dst_group = group;
652 refcount_inc(&skb->users);
653 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
655 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
659 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
661 struct sock *rtnl = net->rtnl;
663 return nlmsg_unicast(rtnl, skb, pid);
665 EXPORT_SYMBOL(rtnl_unicast);
667 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
668 struct nlmsghdr *nlh, gfp_t flags)
670 struct sock *rtnl = net->rtnl;
674 report = nlmsg_report(nlh);
676 nlmsg_notify(rtnl, skb, pid, group, report, flags);
678 EXPORT_SYMBOL(rtnl_notify);
680 void rtnl_set_sk_err(struct net *net, u32 group, int error)
682 struct sock *rtnl = net->rtnl;
684 netlink_set_err(rtnl, 0, group, error);
686 EXPORT_SYMBOL(rtnl_set_sk_err);
688 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
693 mx = nla_nest_start(skb, RTA_METRICS);
697 for (i = 0; i < RTAX_MAX; i++) {
699 if (i == RTAX_CC_ALGO - 1) {
700 char tmp[TCP_CA_NAME_MAX], *name;
702 name = tcp_ca_get_name_by_key(metrics[i], tmp);
705 if (nla_put_string(skb, i + 1, name))
706 goto nla_put_failure;
707 } else if (i == RTAX_FEATURES - 1) {
708 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
712 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
713 if (nla_put_u32(skb, i + 1, user_features))
714 goto nla_put_failure;
716 if (nla_put_u32(skb, i + 1, metrics[i]))
717 goto nla_put_failure;
724 nla_nest_cancel(skb, mx);
728 return nla_nest_end(skb, mx);
731 nla_nest_cancel(skb, mx);
734 EXPORT_SYMBOL(rtnetlink_put_metrics);
736 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
737 long expires, u32 error)
739 struct rta_cacheinfo ci = {
740 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
741 .rta_used = dst->__use,
742 .rta_clntref = atomic_read(&(dst->__refcnt)),
750 clock = jiffies_to_clock_t(abs(expires));
751 clock = min_t(unsigned long, clock, INT_MAX);
752 ci.rta_expires = (expires > 0) ? clock : -clock;
754 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
756 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
758 static void set_operstate(struct net_device *dev, unsigned char transition)
760 unsigned char operstate = dev->operstate;
762 switch (transition) {
764 if ((operstate == IF_OPER_DORMANT ||
765 operstate == IF_OPER_UNKNOWN) &&
767 operstate = IF_OPER_UP;
770 case IF_OPER_DORMANT:
771 if (operstate == IF_OPER_UP ||
772 operstate == IF_OPER_UNKNOWN)
773 operstate = IF_OPER_DORMANT;
777 if (dev->operstate != operstate) {
778 write_lock_bh(&dev_base_lock);
779 dev->operstate = operstate;
780 write_unlock_bh(&dev_base_lock);
781 netdev_state_change(dev);
785 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
787 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
788 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
791 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
792 const struct ifinfomsg *ifm)
794 unsigned int flags = ifm->ifi_flags;
796 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
798 flags = (flags & ifm->ifi_change) |
799 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
804 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
805 const struct rtnl_link_stats64 *b)
807 a->rx_packets = b->rx_packets;
808 a->tx_packets = b->tx_packets;
809 a->rx_bytes = b->rx_bytes;
810 a->tx_bytes = b->tx_bytes;
811 a->rx_errors = b->rx_errors;
812 a->tx_errors = b->tx_errors;
813 a->rx_dropped = b->rx_dropped;
814 a->tx_dropped = b->tx_dropped;
816 a->multicast = b->multicast;
817 a->collisions = b->collisions;
819 a->rx_length_errors = b->rx_length_errors;
820 a->rx_over_errors = b->rx_over_errors;
821 a->rx_crc_errors = b->rx_crc_errors;
822 a->rx_frame_errors = b->rx_frame_errors;
823 a->rx_fifo_errors = b->rx_fifo_errors;
824 a->rx_missed_errors = b->rx_missed_errors;
826 a->tx_aborted_errors = b->tx_aborted_errors;
827 a->tx_carrier_errors = b->tx_carrier_errors;
828 a->tx_fifo_errors = b->tx_fifo_errors;
829 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
830 a->tx_window_errors = b->tx_window_errors;
832 a->rx_compressed = b->rx_compressed;
833 a->tx_compressed = b->tx_compressed;
835 a->rx_nohandler = b->rx_nohandler;
839 static inline int rtnl_vfinfo_size(const struct net_device *dev,
842 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
843 int num_vfs = dev_num_vf(dev->dev.parent);
844 size_t size = nla_total_size(0);
847 nla_total_size(sizeof(struct ifla_vf_mac)) +
848 nla_total_size(sizeof(struct ifla_vf_vlan)) +
849 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
850 nla_total_size(MAX_VLAN_LIST_LEN *
851 sizeof(struct ifla_vf_vlan_info)) +
852 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
853 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
854 nla_total_size(sizeof(struct ifla_vf_rate)) +
855 nla_total_size(sizeof(struct ifla_vf_link_state)) +
856 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
857 nla_total_size(0) + /* nest IFLA_VF_STATS */
858 /* IFLA_VF_STATS_RX_PACKETS */
859 nla_total_size_64bit(sizeof(__u64)) +
860 /* IFLA_VF_STATS_TX_PACKETS */
861 nla_total_size_64bit(sizeof(__u64)) +
862 /* IFLA_VF_STATS_RX_BYTES */
863 nla_total_size_64bit(sizeof(__u64)) +
864 /* IFLA_VF_STATS_TX_BYTES */
865 nla_total_size_64bit(sizeof(__u64)) +
866 /* IFLA_VF_STATS_BROADCAST */
867 nla_total_size_64bit(sizeof(__u64)) +
868 /* IFLA_VF_STATS_MULTICAST */
869 nla_total_size_64bit(sizeof(__u64)) +
870 nla_total_size(sizeof(struct ifla_vf_trust)));
876 static size_t rtnl_port_size(const struct net_device *dev,
879 size_t port_size = nla_total_size(4) /* PORT_VF */
880 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
881 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
882 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
883 + nla_total_size(1) /* PROT_VDP_REQUEST */
884 + nla_total_size(2); /* PORT_VDP_RESPONSE */
885 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
886 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
888 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
891 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
892 !(ext_filter_mask & RTEXT_FILTER_VF))
894 if (dev_num_vf(dev->dev.parent))
895 return port_self_size + vf_ports_size +
896 vf_port_size * dev_num_vf(dev->dev.parent);
898 return port_self_size;
901 static size_t rtnl_xdp_size(void)
903 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
904 nla_total_size(1) + /* XDP_ATTACHED */
905 nla_total_size(4); /* XDP_PROG_ID */
910 static noinline size_t if_nlmsg_size(const struct net_device *dev,
913 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
914 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
915 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
916 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
917 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
918 + nla_total_size(sizeof(struct rtnl_link_stats))
919 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
920 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
921 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
922 + nla_total_size(4) /* IFLA_TXQLEN */
923 + nla_total_size(4) /* IFLA_WEIGHT */
924 + nla_total_size(4) /* IFLA_MTU */
925 + nla_total_size(4) /* IFLA_LINK */
926 + nla_total_size(4) /* IFLA_MASTER */
927 + nla_total_size(1) /* IFLA_CARRIER */
928 + nla_total_size(4) /* IFLA_PROMISCUITY */
929 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
930 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
931 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
932 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
933 + nla_total_size(1) /* IFLA_OPERSTATE */
934 + nla_total_size(1) /* IFLA_LINKMODE */
935 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
936 + nla_total_size(4) /* IFLA_LINK_NETNSID */
937 + nla_total_size(4) /* IFLA_GROUP */
938 + nla_total_size(ext_filter_mask
939 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
940 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
941 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
942 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
943 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
944 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
945 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
946 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
947 + rtnl_xdp_size() /* IFLA_XDP */
948 + nla_total_size(4) /* IFLA_EVENT */
949 + nla_total_size(1); /* IFLA_PROTO_DOWN */
953 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
955 struct nlattr *vf_ports;
956 struct nlattr *vf_port;
960 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
964 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
965 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
967 goto nla_put_failure;
968 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
969 goto nla_put_failure;
970 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
971 if (err == -EMSGSIZE)
972 goto nla_put_failure;
974 nla_nest_cancel(skb, vf_port);
977 nla_nest_end(skb, vf_port);
980 nla_nest_end(skb, vf_ports);
985 nla_nest_cancel(skb, vf_ports);
989 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
991 struct nlattr *port_self;
994 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
998 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1000 nla_nest_cancel(skb, port_self);
1001 return (err == -EMSGSIZE) ? err : 0;
1004 nla_nest_end(skb, port_self);
1009 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1010 u32 ext_filter_mask)
1014 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1015 !(ext_filter_mask & RTEXT_FILTER_VF))
1018 err = rtnl_port_self_fill(skb, dev);
1022 if (dev_num_vf(dev->dev.parent)) {
1023 err = rtnl_vf_ports_fill(skb, dev);
1031 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1034 struct netdev_phys_item_id ppid;
1036 err = dev_get_phys_port_id(dev, &ppid);
1038 if (err == -EOPNOTSUPP)
1043 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1049 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1051 char name[IFNAMSIZ];
1054 err = dev_get_phys_port_name(dev, name, sizeof(name));
1056 if (err == -EOPNOTSUPP)
1061 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1067 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1070 struct switchdev_attr attr = {
1072 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1073 .flags = SWITCHDEV_F_NO_RECURSE,
1076 err = switchdev_port_attr_get(dev, &attr);
1078 if (err == -EOPNOTSUPP)
1083 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1090 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1091 struct net_device *dev)
1093 struct rtnl_link_stats64 *sp;
1094 struct nlattr *attr;
1096 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1097 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1101 sp = nla_data(attr);
1102 dev_get_stats(dev, sp);
1104 attr = nla_reserve(skb, IFLA_STATS,
1105 sizeof(struct rtnl_link_stats));
1109 copy_rtnl_link_stats(nla_data(attr), sp);
1114 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1115 struct net_device *dev,
1117 struct nlattr *vfinfo)
1119 struct ifla_vf_rss_query_en vf_rss_query_en;
1120 struct nlattr *vf, *vfstats, *vfvlanlist;
1121 struct ifla_vf_link_state vf_linkstate;
1122 struct ifla_vf_vlan_info vf_vlan_info;
1123 struct ifla_vf_spoofchk vf_spoofchk;
1124 struct ifla_vf_tx_rate vf_tx_rate;
1125 struct ifla_vf_stats vf_stats;
1126 struct ifla_vf_trust vf_trust;
1127 struct ifla_vf_vlan vf_vlan;
1128 struct ifla_vf_rate vf_rate;
1129 struct ifla_vf_mac vf_mac;
1130 struct ifla_vf_info ivi;
1132 memset(&ivi, 0, sizeof(ivi));
1134 /* Not all SR-IOV capable drivers support the
1135 * spoofcheck and "RSS query enable" query. Preset to
1136 * -1 so the user space tool can detect that the driver
1137 * didn't report anything.
1140 ivi.rss_query_en = -1;
1142 /* The default value for VF link state is "auto"
1143 * IFLA_VF_LINK_STATE_AUTO which equals zero
1146 /* VLAN Protocol by default is 802.1Q */
1147 ivi.vlan_proto = htons(ETH_P_8021Q);
1148 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1151 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1160 vf_rss_query_en.vf =
1161 vf_trust.vf = ivi.vf;
1163 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1164 vf_vlan.vlan = ivi.vlan;
1165 vf_vlan.qos = ivi.qos;
1166 vf_vlan_info.vlan = ivi.vlan;
1167 vf_vlan_info.qos = ivi.qos;
1168 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1169 vf_tx_rate.rate = ivi.max_tx_rate;
1170 vf_rate.min_tx_rate = ivi.min_tx_rate;
1171 vf_rate.max_tx_rate = ivi.max_tx_rate;
1172 vf_spoofchk.setting = ivi.spoofchk;
1173 vf_linkstate.link_state = ivi.linkstate;
1174 vf_rss_query_en.setting = ivi.rss_query_en;
1175 vf_trust.setting = ivi.trusted;
1176 vf = nla_nest_start(skb, IFLA_VF_INFO);
1178 goto nla_put_vfinfo_failure;
1179 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1180 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1181 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1183 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1185 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1187 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1189 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1190 sizeof(vf_rss_query_en),
1191 &vf_rss_query_en) ||
1192 nla_put(skb, IFLA_VF_TRUST,
1193 sizeof(vf_trust), &vf_trust))
1194 goto nla_put_vf_failure;
1195 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1197 goto nla_put_vf_failure;
1198 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1200 nla_nest_cancel(skb, vfvlanlist);
1201 goto nla_put_vf_failure;
1203 nla_nest_end(skb, vfvlanlist);
1204 memset(&vf_stats, 0, sizeof(vf_stats));
1205 if (dev->netdev_ops->ndo_get_vf_stats)
1206 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1208 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1210 goto nla_put_vf_failure;
1211 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1212 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1213 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1214 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1215 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1216 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1217 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1218 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1219 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1220 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1221 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1222 vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1223 nla_nest_cancel(skb, vfstats);
1224 goto nla_put_vf_failure;
1226 nla_nest_end(skb, vfstats);
1227 nla_nest_end(skb, vf);
1231 nla_nest_cancel(skb, vf);
1232 nla_put_vfinfo_failure:
1233 nla_nest_cancel(skb, vfinfo);
1237 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1239 struct rtnl_link_ifmap map;
1241 memset(&map, 0, sizeof(map));
1242 map.mem_start = dev->mem_start;
1243 map.mem_end = dev->mem_end;
1244 map.base_addr = dev->base_addr;
1247 map.port = dev->if_port;
1249 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1255 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1257 const struct net_device_ops *ops = dev->netdev_ops;
1258 const struct bpf_prog *generic_xdp_prog;
1263 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1264 if (generic_xdp_prog) {
1265 *prog_id = generic_xdp_prog->aux->id;
1266 return XDP_ATTACHED_SKB;
1269 return XDP_ATTACHED_NONE;
1271 return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1274 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1280 xdp = nla_nest_start(skb, IFLA_XDP);
1284 err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1285 rtnl_xdp_attached_mode(dev, &prog_id));
1290 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1295 nla_nest_end(skb, xdp);
1299 nla_nest_cancel(skb, xdp);
1303 static u32 rtnl_get_event(unsigned long event)
1305 u32 rtnl_event_type = IFLA_EVENT_NONE;
1309 rtnl_event_type = IFLA_EVENT_REBOOT;
1311 case NETDEV_FEAT_CHANGE:
1312 rtnl_event_type = IFLA_EVENT_FEATURES;
1314 case NETDEV_BONDING_FAILOVER:
1315 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1317 case NETDEV_NOTIFY_PEERS:
1318 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1320 case NETDEV_RESEND_IGMP:
1321 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1323 case NETDEV_CHANGEINFODATA:
1324 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1330 return rtnl_event_type;
1333 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1334 int type, u32 pid, u32 seq, u32 change,
1335 unsigned int flags, u32 ext_filter_mask,
1338 struct ifinfomsg *ifm;
1339 struct nlmsghdr *nlh;
1340 struct nlattr *af_spec;
1341 struct rtnl_af_ops *af_ops;
1342 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1345 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1349 ifm = nlmsg_data(nlh);
1350 ifm->ifi_family = AF_UNSPEC;
1352 ifm->ifi_type = dev->type;
1353 ifm->ifi_index = dev->ifindex;
1354 ifm->ifi_flags = dev_get_flags(dev);
1355 ifm->ifi_change = change;
1357 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1358 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1359 nla_put_u8(skb, IFLA_OPERSTATE,
1360 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1361 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1362 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1363 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1364 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1365 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1366 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1367 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1369 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1371 (dev->ifindex != dev_get_iflink(dev) &&
1372 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1374 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1375 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1377 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1379 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1380 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1381 atomic_read(&dev->carrier_changes)) ||
1382 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1383 goto nla_put_failure;
1385 if (event != IFLA_EVENT_NONE) {
1386 if (nla_put_u32(skb, IFLA_EVENT, event))
1387 goto nla_put_failure;
1390 if (rtnl_fill_link_ifmap(skb, dev))
1391 goto nla_put_failure;
1393 if (dev->addr_len) {
1394 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1395 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1396 goto nla_put_failure;
1399 if (rtnl_phys_port_id_fill(skb, dev))
1400 goto nla_put_failure;
1402 if (rtnl_phys_port_name_fill(skb, dev))
1403 goto nla_put_failure;
1405 if (rtnl_phys_switch_id_fill(skb, dev))
1406 goto nla_put_failure;
1408 if (rtnl_fill_stats(skb, dev))
1409 goto nla_put_failure;
1411 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1412 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1413 goto nla_put_failure;
1415 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1416 ext_filter_mask & RTEXT_FILTER_VF) {
1418 struct nlattr *vfinfo;
1419 int num_vfs = dev_num_vf(dev->dev.parent);
1421 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1423 goto nla_put_failure;
1424 for (i = 0; i < num_vfs; i++) {
1425 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1426 goto nla_put_failure;
1429 nla_nest_end(skb, vfinfo);
1432 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1433 goto nla_put_failure;
1435 if (rtnl_xdp_fill(skb, dev))
1436 goto nla_put_failure;
1438 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1439 if (rtnl_link_fill(skb, dev) < 0)
1440 goto nla_put_failure;
1443 if (dev->rtnl_link_ops &&
1444 dev->rtnl_link_ops->get_link_net) {
1445 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1447 if (!net_eq(dev_net(dev), link_net)) {
1448 int id = peernet2id_alloc(dev_net(dev), link_net);
1450 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1451 goto nla_put_failure;
1455 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1456 goto nla_put_failure;
1458 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1459 if (af_ops->fill_link_af) {
1463 if (!(af = nla_nest_start(skb, af_ops->family)))
1464 goto nla_put_failure;
1466 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1469 * Caller may return ENODATA to indicate that there
1470 * was no data to be dumped. This is not an error, it
1471 * means we should trim the attribute header and
1474 if (err == -ENODATA)
1475 nla_nest_cancel(skb, af);
1477 goto nla_put_failure;
1479 nla_nest_end(skb, af);
1483 nla_nest_end(skb, af_spec);
1485 nlmsg_end(skb, nlh);
1489 nlmsg_cancel(skb, nlh);
1493 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1494 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1495 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1496 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1497 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1498 [IFLA_MTU] = { .type = NLA_U32 },
1499 [IFLA_LINK] = { .type = NLA_U32 },
1500 [IFLA_MASTER] = { .type = NLA_U32 },
1501 [IFLA_CARRIER] = { .type = NLA_U8 },
1502 [IFLA_TXQLEN] = { .type = NLA_U32 },
1503 [IFLA_WEIGHT] = { .type = NLA_U32 },
1504 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1505 [IFLA_LINKMODE] = { .type = NLA_U8 },
1506 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1507 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1508 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1509 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1510 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1511 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1512 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1513 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1514 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1515 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1516 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1517 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1518 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1519 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1520 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1521 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1522 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1523 [IFLA_XDP] = { .type = NLA_NESTED },
1524 [IFLA_EVENT] = { .type = NLA_U32 },
1525 [IFLA_GROUP] = { .type = NLA_U32 },
1528 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1529 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1530 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1531 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1532 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1535 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1536 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1537 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1538 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1539 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1540 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1541 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1542 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1543 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1544 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1545 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1546 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1547 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1550 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1551 [IFLA_PORT_VF] = { .type = NLA_U32 },
1552 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1553 .len = PORT_PROFILE_MAX },
1554 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1555 .len = PORT_UUID_MAX },
1556 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1557 .len = PORT_UUID_MAX },
1558 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1559 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1561 /* Unused, but we need to keep it here since user space could
1562 * fill it. It's also broken with regard to NLA_BINARY use in
1563 * combination with structs.
1565 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1566 .len = sizeof(struct ifla_port_vsi) },
1569 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1570 [IFLA_XDP_FD] = { .type = NLA_S32 },
1571 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1572 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1573 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1576 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1578 const struct rtnl_link_ops *ops = NULL;
1579 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1581 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1582 ifla_info_policy, NULL) < 0)
1585 if (linfo[IFLA_INFO_KIND]) {
1586 char kind[MODULE_NAME_LEN];
1588 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1589 ops = rtnl_link_ops_get(kind);
1595 static bool link_master_filtered(struct net_device *dev, int master_idx)
1597 struct net_device *master;
1602 master = netdev_master_upper_dev_get(dev);
1603 if (!master || master->ifindex != master_idx)
1609 static bool link_kind_filtered(const struct net_device *dev,
1610 const struct rtnl_link_ops *kind_ops)
1612 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1618 static bool link_dump_filtered(struct net_device *dev,
1620 const struct rtnl_link_ops *kind_ops)
1622 if (link_master_filtered(dev, master_idx) ||
1623 link_kind_filtered(dev, kind_ops))
1629 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1631 struct net *net = sock_net(skb->sk);
1634 struct net_device *dev;
1635 struct hlist_head *head;
1636 struct nlattr *tb[IFLA_MAX+1];
1637 u32 ext_filter_mask = 0;
1638 const struct rtnl_link_ops *kind_ops = NULL;
1639 unsigned int flags = NLM_F_MULTI;
1645 s_idx = cb->args[1];
1647 cb->seq = net->dev_base_seq;
1649 /* A hack to preserve kernel<->userspace interface.
1650 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1651 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1652 * what iproute2 < v3.9.0 used.
1653 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1654 * attribute, its netlink message is shorter than struct ifinfomsg.
1656 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1657 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1659 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1660 ifla_policy, NULL) >= 0) {
1661 if (tb[IFLA_EXT_MASK])
1662 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1664 if (tb[IFLA_MASTER])
1665 master_idx = nla_get_u32(tb[IFLA_MASTER]);
1667 if (tb[IFLA_LINKINFO])
1668 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1670 if (master_idx || kind_ops)
1671 flags |= NLM_F_DUMP_FILTERED;
1674 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1676 head = &net->dev_index_head[h];
1677 hlist_for_each_entry(dev, head, index_hlist) {
1678 if (link_dump_filtered(dev, master_idx, kind_ops))
1682 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1683 NETLINK_CB(cb->skb).portid,
1684 cb->nlh->nlmsg_seq, 0,
1686 ext_filter_mask, 0);
1689 if (likely(skb->len))
1695 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1709 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1710 struct netlink_ext_ack *exterr)
1712 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1714 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1716 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1719 /* Examine the link attributes and figure out which
1720 * network namespace we are talking about.
1722 if (tb[IFLA_NET_NS_PID])
1723 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1724 else if (tb[IFLA_NET_NS_FD])
1725 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1727 net = get_net(src_net);
1730 EXPORT_SYMBOL(rtnl_link_get_net);
1732 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1735 if (tb[IFLA_ADDRESS] &&
1736 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1739 if (tb[IFLA_BROADCAST] &&
1740 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1744 if (tb[IFLA_AF_SPEC]) {
1748 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1749 const struct rtnl_af_ops *af_ops;
1751 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1752 return -EAFNOSUPPORT;
1754 if (!af_ops->set_link_af)
1757 if (af_ops->validate_link_af) {
1758 err = af_ops->validate_link_af(dev, af);
1768 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1771 const struct net_device_ops *ops = dev->netdev_ops;
1773 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1776 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1778 if (dev->type != ARPHRD_INFINIBAND)
1781 return handle_infiniband_guid(dev, ivt, guid_type);
1784 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1786 const struct net_device_ops *ops = dev->netdev_ops;
1789 if (tb[IFLA_VF_MAC]) {
1790 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1793 if (ops->ndo_set_vf_mac)
1794 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1800 if (tb[IFLA_VF_VLAN]) {
1801 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1804 if (ops->ndo_set_vf_vlan)
1805 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1807 htons(ETH_P_8021Q));
1812 if (tb[IFLA_VF_VLAN_LIST]) {
1813 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1814 struct nlattr *attr;
1818 if (!ops->ndo_set_vf_vlan)
1821 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1822 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1823 nla_len(attr) < NLA_HDRLEN) {
1826 if (len >= MAX_VLAN_LIST_LEN)
1828 ivvl[len] = nla_data(attr);
1835 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1836 ivvl[0]->qos, ivvl[0]->vlan_proto);
1841 if (tb[IFLA_VF_TX_RATE]) {
1842 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1843 struct ifla_vf_info ivf;
1846 if (ops->ndo_get_vf_config)
1847 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1852 if (ops->ndo_set_vf_rate)
1853 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1860 if (tb[IFLA_VF_RATE]) {
1861 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1864 if (ops->ndo_set_vf_rate)
1865 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1872 if (tb[IFLA_VF_SPOOFCHK]) {
1873 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1876 if (ops->ndo_set_vf_spoofchk)
1877 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1883 if (tb[IFLA_VF_LINK_STATE]) {
1884 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1887 if (ops->ndo_set_vf_link_state)
1888 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1894 if (tb[IFLA_VF_RSS_QUERY_EN]) {
1895 struct ifla_vf_rss_query_en *ivrssq_en;
1898 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1899 if (ops->ndo_set_vf_rss_query_en)
1900 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1901 ivrssq_en->setting);
1906 if (tb[IFLA_VF_TRUST]) {
1907 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1910 if (ops->ndo_set_vf_trust)
1911 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1916 if (tb[IFLA_VF_IB_NODE_GUID]) {
1917 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1919 if (!ops->ndo_set_vf_guid)
1922 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1925 if (tb[IFLA_VF_IB_PORT_GUID]) {
1926 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1928 if (!ops->ndo_set_vf_guid)
1931 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1937 static int do_set_master(struct net_device *dev, int ifindex)
1939 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1940 const struct net_device_ops *ops;
1944 if (upper_dev->ifindex == ifindex)
1946 ops = upper_dev->netdev_ops;
1947 if (ops->ndo_del_slave) {
1948 err = ops->ndo_del_slave(upper_dev, dev);
1957 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1960 ops = upper_dev->netdev_ops;
1961 if (ops->ndo_add_slave) {
1962 err = ops->ndo_add_slave(upper_dev, dev);
1972 #define DO_SETLINK_MODIFIED 0x01
1973 /* notify flag means notify + modified. */
1974 #define DO_SETLINK_NOTIFY 0x03
1975 static int do_setlink(const struct sk_buff *skb,
1976 struct net_device *dev, struct ifinfomsg *ifm,
1977 struct netlink_ext_ack *extack,
1978 struct nlattr **tb, char *ifname, int status)
1980 const struct net_device_ops *ops = dev->netdev_ops;
1983 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1984 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1989 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1994 err = dev_change_net_namespace(dev, net, ifname);
1998 status |= DO_SETLINK_MODIFIED;
2002 struct rtnl_link_ifmap *u_map;
2005 if (!ops->ndo_set_config) {
2010 if (!netif_device_present(dev)) {
2015 u_map = nla_data(tb[IFLA_MAP]);
2016 k_map.mem_start = (unsigned long) u_map->mem_start;
2017 k_map.mem_end = (unsigned long) u_map->mem_end;
2018 k_map.base_addr = (unsigned short) u_map->base_addr;
2019 k_map.irq = (unsigned char) u_map->irq;
2020 k_map.dma = (unsigned char) u_map->dma;
2021 k_map.port = (unsigned char) u_map->port;
2023 err = ops->ndo_set_config(dev, &k_map);
2027 status |= DO_SETLINK_NOTIFY;
2030 if (tb[IFLA_ADDRESS]) {
2031 struct sockaddr *sa;
2034 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2036 sa = kmalloc(len, GFP_KERNEL);
2041 sa->sa_family = dev->type;
2042 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2044 err = dev_set_mac_address(dev, sa);
2048 status |= DO_SETLINK_MODIFIED;
2052 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2055 status |= DO_SETLINK_MODIFIED;
2058 if (tb[IFLA_GROUP]) {
2059 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2060 status |= DO_SETLINK_NOTIFY;
2064 * Interface selected by interface index but interface
2065 * name provided implies that a name change has been
2068 if (ifm->ifi_index > 0 && ifname[0]) {
2069 err = dev_change_name(dev, ifname);
2072 status |= DO_SETLINK_MODIFIED;
2075 if (tb[IFLA_IFALIAS]) {
2076 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2077 nla_len(tb[IFLA_IFALIAS]));
2080 status |= DO_SETLINK_NOTIFY;
2083 if (tb[IFLA_BROADCAST]) {
2084 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2085 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2088 if (ifm->ifi_flags || ifm->ifi_change) {
2089 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2094 if (tb[IFLA_MASTER]) {
2095 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2098 status |= DO_SETLINK_MODIFIED;
2101 if (tb[IFLA_CARRIER]) {
2102 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2105 status |= DO_SETLINK_MODIFIED;
2108 if (tb[IFLA_TXQLEN]) {
2109 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2110 unsigned int orig_len = dev->tx_queue_len;
2112 if (dev->tx_queue_len ^ value) {
2113 dev->tx_queue_len = value;
2114 err = call_netdevice_notifiers(
2115 NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2116 err = notifier_to_errno(err);
2118 dev->tx_queue_len = orig_len;
2121 status |= DO_SETLINK_NOTIFY;
2125 if (tb[IFLA_OPERSTATE])
2126 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2128 if (tb[IFLA_LINKMODE]) {
2129 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2131 write_lock_bh(&dev_base_lock);
2132 if (dev->link_mode ^ value)
2133 status |= DO_SETLINK_NOTIFY;
2134 dev->link_mode = value;
2135 write_unlock_bh(&dev_base_lock);
2138 if (tb[IFLA_VFINFO_LIST]) {
2139 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2140 struct nlattr *attr;
2143 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2144 if (nla_type(attr) != IFLA_VF_INFO ||
2145 nla_len(attr) < NLA_HDRLEN) {
2149 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2150 ifla_vf_policy, NULL);
2153 err = do_setvfinfo(dev, vfinfo);
2156 status |= DO_SETLINK_NOTIFY;
2161 if (tb[IFLA_VF_PORTS]) {
2162 struct nlattr *port[IFLA_PORT_MAX+1];
2163 struct nlattr *attr;
2168 if (!ops->ndo_set_vf_port)
2171 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2172 if (nla_type(attr) != IFLA_VF_PORT ||
2173 nla_len(attr) < NLA_HDRLEN) {
2177 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2178 ifla_port_policy, NULL);
2181 if (!port[IFLA_PORT_VF]) {
2185 vf = nla_get_u32(port[IFLA_PORT_VF]);
2186 err = ops->ndo_set_vf_port(dev, vf, port);
2189 status |= DO_SETLINK_NOTIFY;
2194 if (tb[IFLA_PORT_SELF]) {
2195 struct nlattr *port[IFLA_PORT_MAX+1];
2197 err = nla_parse_nested(port, IFLA_PORT_MAX,
2198 tb[IFLA_PORT_SELF], ifla_port_policy,
2204 if (ops->ndo_set_vf_port)
2205 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2208 status |= DO_SETLINK_NOTIFY;
2211 if (tb[IFLA_AF_SPEC]) {
2215 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2216 const struct rtnl_af_ops *af_ops;
2218 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2221 err = af_ops->set_link_af(dev, af);
2225 status |= DO_SETLINK_NOTIFY;
2230 if (tb[IFLA_PROTO_DOWN]) {
2231 err = dev_change_proto_down(dev,
2232 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2235 status |= DO_SETLINK_NOTIFY;
2239 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2242 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2243 ifla_xdp_policy, NULL);
2247 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2252 if (xdp[IFLA_XDP_FLAGS]) {
2253 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2254 if (xdp_flags & ~XDP_FLAGS_MASK) {
2258 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2264 if (xdp[IFLA_XDP_FD]) {
2265 err = dev_change_xdp_fd(dev, extack,
2266 nla_get_s32(xdp[IFLA_XDP_FD]),
2270 status |= DO_SETLINK_NOTIFY;
2275 if (status & DO_SETLINK_MODIFIED) {
2276 if (status & DO_SETLINK_NOTIFY)
2277 netdev_state_change(dev);
2280 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",
2287 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2288 struct netlink_ext_ack *extack)
2290 struct net *net = sock_net(skb->sk);
2291 struct ifinfomsg *ifm;
2292 struct net_device *dev;
2294 struct nlattr *tb[IFLA_MAX+1];
2295 char ifname[IFNAMSIZ];
2297 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2302 if (tb[IFLA_IFNAME])
2303 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2308 ifm = nlmsg_data(nlh);
2309 if (ifm->ifi_index > 0)
2310 dev = __dev_get_by_index(net, ifm->ifi_index);
2311 else if (tb[IFLA_IFNAME])
2312 dev = __dev_get_by_name(net, ifname);
2321 err = validate_linkmsg(dev, tb);
2325 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2330 static int rtnl_group_dellink(const struct net *net, int group)
2332 struct net_device *dev, *aux;
2333 LIST_HEAD(list_kill);
2339 for_each_netdev(net, dev) {
2340 if (dev->group == group) {
2341 const struct rtnl_link_ops *ops;
2344 ops = dev->rtnl_link_ops;
2345 if (!ops || !ops->dellink)
2353 for_each_netdev_safe(net, dev, aux) {
2354 if (dev->group == group) {
2355 const struct rtnl_link_ops *ops;
2357 ops = dev->rtnl_link_ops;
2358 ops->dellink(dev, &list_kill);
2361 unregister_netdevice_many(&list_kill);
2366 int rtnl_delete_link(struct net_device *dev)
2368 const struct rtnl_link_ops *ops;
2369 LIST_HEAD(list_kill);
2371 ops = dev->rtnl_link_ops;
2372 if (!ops || !ops->dellink)
2375 ops->dellink(dev, &list_kill);
2376 unregister_netdevice_many(&list_kill);
2380 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2382 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2383 struct netlink_ext_ack *extack)
2385 struct net *net = sock_net(skb->sk);
2386 struct net_device *dev;
2387 struct ifinfomsg *ifm;
2388 char ifname[IFNAMSIZ];
2389 struct nlattr *tb[IFLA_MAX+1];
2392 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2396 if (tb[IFLA_IFNAME])
2397 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2399 ifm = nlmsg_data(nlh);
2400 if (ifm->ifi_index > 0)
2401 dev = __dev_get_by_index(net, ifm->ifi_index);
2402 else if (tb[IFLA_IFNAME])
2403 dev = __dev_get_by_name(net, ifname);
2404 else if (tb[IFLA_GROUP])
2405 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2412 return rtnl_delete_link(dev);
2415 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2417 unsigned int old_flags;
2420 old_flags = dev->flags;
2421 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2422 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2427 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2429 __dev_notify_flags(dev, old_flags, ~0U);
2432 EXPORT_SYMBOL(rtnl_configure_link);
2434 struct net_device *rtnl_create_link(struct net *net,
2435 const char *ifname, unsigned char name_assign_type,
2436 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2438 struct net_device *dev;
2439 unsigned int num_tx_queues = 1;
2440 unsigned int num_rx_queues = 1;
2442 if (tb[IFLA_NUM_TX_QUEUES])
2443 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2444 else if (ops->get_num_tx_queues)
2445 num_tx_queues = ops->get_num_tx_queues();
2447 if (tb[IFLA_NUM_RX_QUEUES])
2448 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2449 else if (ops->get_num_rx_queues)
2450 num_rx_queues = ops->get_num_rx_queues();
2452 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2453 ops->setup, num_tx_queues, num_rx_queues);
2455 return ERR_PTR(-ENOMEM);
2457 dev_net_set(dev, net);
2458 dev->rtnl_link_ops = ops;
2459 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2462 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2463 if (tb[IFLA_ADDRESS]) {
2464 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2465 nla_len(tb[IFLA_ADDRESS]));
2466 dev->addr_assign_type = NET_ADDR_SET;
2468 if (tb[IFLA_BROADCAST])
2469 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2470 nla_len(tb[IFLA_BROADCAST]));
2471 if (tb[IFLA_TXQLEN])
2472 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2473 if (tb[IFLA_OPERSTATE])
2474 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2475 if (tb[IFLA_LINKMODE])
2476 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2478 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2482 EXPORT_SYMBOL(rtnl_create_link);
2484 static int rtnl_group_changelink(const struct sk_buff *skb,
2485 struct net *net, int group,
2486 struct ifinfomsg *ifm,
2487 struct netlink_ext_ack *extack,
2490 struct net_device *dev, *aux;
2493 for_each_netdev_safe(net, dev, aux) {
2494 if (dev->group == group) {
2495 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2504 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2505 struct netlink_ext_ack *extack)
2507 struct net *net = sock_net(skb->sk);
2508 const struct rtnl_link_ops *ops;
2509 const struct rtnl_link_ops *m_ops = NULL;
2510 struct net_device *dev;
2511 struct net_device *master_dev = NULL;
2512 struct ifinfomsg *ifm;
2513 char kind[MODULE_NAME_LEN];
2514 char ifname[IFNAMSIZ];
2515 struct nlattr *tb[IFLA_MAX+1];
2516 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2517 unsigned char name_assign_type = NET_NAME_USER;
2520 #ifdef CONFIG_MODULES
2523 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2527 if (tb[IFLA_IFNAME])
2528 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2532 ifm = nlmsg_data(nlh);
2533 if (ifm->ifi_index > 0)
2534 dev = __dev_get_by_index(net, ifm->ifi_index);
2537 dev = __dev_get_by_name(net, ifname);
2543 master_dev = netdev_master_upper_dev_get(dev);
2545 m_ops = master_dev->rtnl_link_ops;
2548 err = validate_linkmsg(dev, tb);
2552 if (tb[IFLA_LINKINFO]) {
2553 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2554 tb[IFLA_LINKINFO], ifla_info_policy,
2559 memset(linkinfo, 0, sizeof(linkinfo));
2561 if (linkinfo[IFLA_INFO_KIND]) {
2562 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2563 ops = rtnl_link_ops_get(kind);
2570 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2571 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2572 struct nlattr **data = NULL;
2573 struct nlattr **slave_data = NULL;
2574 struct net *dest_net, *link_net = NULL;
2577 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2578 err = nla_parse_nested(attr, ops->maxtype,
2579 linkinfo[IFLA_INFO_DATA],
2585 if (ops->validate) {
2586 err = ops->validate(tb, data, extack);
2593 if (m_ops->slave_maxtype &&
2594 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2595 err = nla_parse_nested(slave_attr,
2596 m_ops->slave_maxtype,
2597 linkinfo[IFLA_INFO_SLAVE_DATA],
2598 m_ops->slave_policy,
2602 slave_data = slave_attr;
2604 if (m_ops->slave_validate) {
2605 err = m_ops->slave_validate(tb, slave_data,
2615 if (nlh->nlmsg_flags & NLM_F_EXCL)
2617 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2620 if (linkinfo[IFLA_INFO_DATA]) {
2621 if (!ops || ops != dev->rtnl_link_ops ||
2625 err = ops->changelink(dev, tb, data, extack);
2628 status |= DO_SETLINK_NOTIFY;
2631 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2632 if (!m_ops || !m_ops->slave_changelink)
2635 err = m_ops->slave_changelink(master_dev, dev,
2640 status |= DO_SETLINK_NOTIFY;
2643 return do_setlink(skb, dev, ifm, extack, tb, ifname,
2647 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2648 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2649 return rtnl_group_changelink(skb, net,
2650 nla_get_u32(tb[IFLA_GROUP]),
2655 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2659 #ifdef CONFIG_MODULES
2662 request_module("rtnl-link-%s", kind);
2664 ops = rtnl_link_ops_get(kind);
2676 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2677 name_assign_type = NET_NAME_ENUM;
2680 dest_net = rtnl_link_get_net(net, tb);
2681 if (IS_ERR(dest_net))
2682 return PTR_ERR(dest_net);
2685 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2688 if (tb[IFLA_LINK_NETNSID]) {
2689 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2691 link_net = get_net_ns_by_id(dest_net, id);
2697 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2701 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2702 name_assign_type, ops, tb);
2708 dev->ifindex = ifm->ifi_index;
2711 err = ops->newlink(link_net ? : net, dev, tb, data,
2713 /* Drivers should call free_netdev() in ->destructor
2714 * and unregister it on failure after registration
2715 * so that device could be finally freed in rtnl_unlock.
2718 /* If device is not registered at all, free it now */
2719 if (dev->reg_state == NETREG_UNINITIALIZED)
2724 err = register_netdevice(dev);
2730 err = rtnl_configure_link(dev, ifm);
2732 goto out_unregister;
2734 err = dev_change_net_namespace(dev, dest_net, ifname);
2736 goto out_unregister;
2738 if (tb[IFLA_MASTER]) {
2739 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2741 goto out_unregister;
2750 LIST_HEAD(list_kill);
2752 ops->dellink(dev, &list_kill);
2753 unregister_netdevice_many(&list_kill);
2755 unregister_netdevice(dev);
2761 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2762 struct netlink_ext_ack *extack)
2764 struct net *net = sock_net(skb->sk);
2765 struct ifinfomsg *ifm;
2766 char ifname[IFNAMSIZ];
2767 struct nlattr *tb[IFLA_MAX+1];
2768 struct net_device *dev = NULL;
2769 struct sk_buff *nskb;
2771 u32 ext_filter_mask = 0;
2773 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2777 if (tb[IFLA_IFNAME])
2778 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2780 if (tb[IFLA_EXT_MASK])
2781 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2783 ifm = nlmsg_data(nlh);
2784 if (ifm->ifi_index > 0)
2785 dev = __dev_get_by_index(net, ifm->ifi_index);
2786 else if (tb[IFLA_IFNAME])
2787 dev = __dev_get_by_name(net, ifname);
2794 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2798 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2799 nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0);
2801 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2802 WARN_ON(err == -EMSGSIZE);
2805 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2810 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2812 struct net *net = sock_net(skb->sk);
2813 struct net_device *dev;
2814 struct nlattr *tb[IFLA_MAX+1];
2815 u32 ext_filter_mask = 0;
2816 u16 min_ifinfo_dump_size = 0;
2819 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2820 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2821 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2823 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2824 if (tb[IFLA_EXT_MASK])
2825 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2828 if (!ext_filter_mask)
2829 return NLMSG_GOODSIZE;
2831 * traverse the list of net devices and compute the minimum
2832 * buffer size based upon the filter mask.
2834 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2835 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2840 return nlmsg_total_size(min_ifinfo_dump_size);
2843 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2846 int s_idx = cb->family;
2850 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2851 int type = cb->nlh->nlmsg_type-RTM_BASE;
2852 if (idx < s_idx || idx == PF_PACKET)
2854 if (rtnl_msg_handlers[idx] == NULL ||
2855 rtnl_msg_handlers[idx][type].dumpit == NULL)
2858 memset(&cb->args[0], 0, sizeof(cb->args));
2862 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2870 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2871 unsigned int change,
2872 u32 event, gfp_t flags)
2874 struct net *net = dev_net(dev);
2875 struct sk_buff *skb;
2877 size_t if_info_size;
2879 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2883 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event);
2885 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2886 WARN_ON(err == -EMSGSIZE);
2893 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2897 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2899 struct net *net = dev_net(dev);
2901 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2904 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2905 unsigned int change, u32 event,
2908 struct sk_buff *skb;
2910 if (dev->reg_state != NETREG_REGISTERED)
2913 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags);
2915 rtmsg_ifinfo_send(skb, dev, flags);
2918 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2921 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags);
2923 EXPORT_SYMBOL(rtmsg_ifinfo);
2925 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2926 struct net_device *dev,
2927 u8 *addr, u16 vid, u32 pid, u32 seq,
2928 int type, unsigned int flags,
2929 int nlflags, u16 ndm_state)
2931 struct nlmsghdr *nlh;
2934 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2938 ndm = nlmsg_data(nlh);
2939 ndm->ndm_family = AF_BRIDGE;
2942 ndm->ndm_flags = flags;
2944 ndm->ndm_ifindex = dev->ifindex;
2945 ndm->ndm_state = ndm_state;
2947 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2948 goto nla_put_failure;
2950 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2951 goto nla_put_failure;
2953 nlmsg_end(skb, nlh);
2957 nlmsg_cancel(skb, nlh);
2961 static inline size_t rtnl_fdb_nlmsg_size(void)
2963 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2964 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
2965 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
2969 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2972 struct net *net = dev_net(dev);
2973 struct sk_buff *skb;
2976 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2980 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2981 0, 0, type, NTF_SELF, 0, ndm_state);
2987 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2990 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2994 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2996 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2997 struct nlattr *tb[],
2998 struct net_device *dev,
2999 const unsigned char *addr, u16 vid,
3004 /* If aging addresses are supported device will need to
3005 * implement its own handler for this.
3007 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3008 pr_info("%s: FDB only supports static addresses\n", dev->name);
3013 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3017 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3018 err = dev_uc_add_excl(dev, addr);
3019 else if (is_multicast_ether_addr(addr))
3020 err = dev_mc_add_excl(dev, addr);
3022 /* Only return duplicate errors if NLM_F_EXCL is set */
3023 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3028 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3030 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
3035 if (nla_len(vlan_attr) != sizeof(u16)) {
3036 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
3040 vid = nla_get_u16(vlan_attr);
3042 if (!vid || vid >= VLAN_VID_MASK) {
3043 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
3052 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3053 struct netlink_ext_ack *extack)
3055 struct net *net = sock_net(skb->sk);
3057 struct nlattr *tb[NDA_MAX+1];
3058 struct net_device *dev;
3063 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3067 ndm = nlmsg_data(nlh);
3068 if (ndm->ndm_ifindex == 0) {
3069 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3073 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3075 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3079 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3080 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3084 addr = nla_data(tb[NDA_LLADDR]);
3086 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3092 /* Support fdb on master device the net/bridge default case */
3093 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3094 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3095 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3096 const struct net_device_ops *ops = br_dev->netdev_ops;
3098 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3103 ndm->ndm_flags &= ~NTF_MASTER;
3106 /* Embedded bridge, macvlan, and any other device support */
3107 if ((ndm->ndm_flags & NTF_SELF)) {
3108 if (dev->netdev_ops->ndo_fdb_add)
3109 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3113 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3117 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3119 ndm->ndm_flags &= ~NTF_SELF;
3127 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3129 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3130 struct nlattr *tb[],
3131 struct net_device *dev,
3132 const unsigned char *addr, u16 vid)
3136 /* If aging addresses are supported device will need to
3137 * implement its own handler for this.
3139 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3140 pr_info("%s: FDB only supports static addresses\n", dev->name);
3144 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3145 err = dev_uc_del(dev, addr);
3146 else if (is_multicast_ether_addr(addr))
3147 err = dev_mc_del(dev, addr);
3151 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3153 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3154 struct netlink_ext_ack *extack)
3156 struct net *net = sock_net(skb->sk);
3158 struct nlattr *tb[NDA_MAX+1];
3159 struct net_device *dev;
3164 if (!netlink_capable(skb, CAP_NET_ADMIN))
3167 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3171 ndm = nlmsg_data(nlh);
3172 if (ndm->ndm_ifindex == 0) {
3173 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3177 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3179 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3183 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3184 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3188 addr = nla_data(tb[NDA_LLADDR]);
3190 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3196 /* Support fdb on master device the net/bridge default case */
3197 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3198 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3199 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3200 const struct net_device_ops *ops = br_dev->netdev_ops;
3202 if (ops->ndo_fdb_del)
3203 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3208 ndm->ndm_flags &= ~NTF_MASTER;
3211 /* Embedded bridge, macvlan, and any other device support */
3212 if (ndm->ndm_flags & NTF_SELF) {
3213 if (dev->netdev_ops->ndo_fdb_del)
3214 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3217 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3220 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3222 ndm->ndm_flags &= ~NTF_SELF;
3229 static int nlmsg_populate_fdb(struct sk_buff *skb,
3230 struct netlink_callback *cb,
3231 struct net_device *dev,
3233 struct netdev_hw_addr_list *list)
3235 struct netdev_hw_addr *ha;
3239 portid = NETLINK_CB(cb->skb).portid;
3240 seq = cb->nlh->nlmsg_seq;
3242 list_for_each_entry(ha, &list->list, list) {
3243 if (*idx < cb->args[2])
3246 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3248 RTM_NEWNEIGH, NTF_SELF,
3249 NLM_F_MULTI, NUD_PERMANENT);
3259 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3260 * @nlh: netlink message header
3263 * Default netdevice operation to dump the existing unicast address list.
3264 * Returns number of addresses from list put in skb.
3266 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3267 struct netlink_callback *cb,
3268 struct net_device *dev,
3269 struct net_device *filter_dev,
3274 netif_addr_lock_bh(dev);
3275 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3278 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3280 netif_addr_unlock_bh(dev);
3283 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3285 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3287 struct net_device *dev;
3288 struct nlattr *tb[IFLA_MAX+1];
3289 struct net_device *br_dev = NULL;
3290 const struct net_device_ops *ops = NULL;
3291 const struct net_device_ops *cops = NULL;
3292 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3293 struct net *net = sock_net(skb->sk);
3294 struct hlist_head *head;
3302 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3303 IFLA_MAX, ifla_policy, NULL);
3306 } else if (err == 0) {
3307 if (tb[IFLA_MASTER])
3308 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3311 brport_idx = ifm->ifi_index;
3314 br_dev = __dev_get_by_index(net, br_idx);
3318 ops = br_dev->netdev_ops;
3322 s_idx = cb->args[1];
3324 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3326 head = &net->dev_index_head[h];
3327 hlist_for_each_entry(dev, head, index_hlist) {
3329 if (brport_idx && (dev->ifindex != brport_idx))
3332 if (!br_idx) { /* user did not specify a specific bridge */
3333 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3334 br_dev = netdev_master_upper_dev_get(dev);
3335 cops = br_dev->netdev_ops;
3338 if (dev != br_dev &&
3339 !(dev->priv_flags & IFF_BRIDGE_PORT))
3342 if (br_dev != netdev_master_upper_dev_get(dev) &&
3343 !(dev->priv_flags & IFF_EBRIDGE))
3351 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3352 if (cops && cops->ndo_fdb_dump) {
3353 err = cops->ndo_fdb_dump(skb, cb,
3356 if (err == -EMSGSIZE)
3361 if (dev->netdev_ops->ndo_fdb_dump)
3362 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3366 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3368 if (err == -EMSGSIZE)
3373 /* reset fdb offset to 0 for rest of the interfaces */
3389 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3390 unsigned int attrnum, unsigned int flag)
3393 return nla_put_u8(skb, attrnum, !!(flags & flag));
3397 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3398 struct net_device *dev, u16 mode,
3399 u32 flags, u32 mask, int nlflags,
3401 int (*vlan_fill)(struct sk_buff *skb,
3402 struct net_device *dev,
3405 struct nlmsghdr *nlh;
3406 struct ifinfomsg *ifm;
3407 struct nlattr *br_afspec;
3408 struct nlattr *protinfo;
3409 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3410 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3413 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3417 ifm = nlmsg_data(nlh);
3418 ifm->ifi_family = AF_BRIDGE;
3420 ifm->ifi_type = dev->type;
3421 ifm->ifi_index = dev->ifindex;
3422 ifm->ifi_flags = dev_get_flags(dev);
3423 ifm->ifi_change = 0;
3426 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3427 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3428 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3430 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3432 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3433 (dev->ifindex != dev_get_iflink(dev) &&
3434 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3435 goto nla_put_failure;
3437 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3439 goto nla_put_failure;
3441 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3442 nla_nest_cancel(skb, br_afspec);
3443 goto nla_put_failure;
3446 if (mode != BRIDGE_MODE_UNDEF) {
3447 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3448 nla_nest_cancel(skb, br_afspec);
3449 goto nla_put_failure;
3453 err = vlan_fill(skb, dev, filter_mask);
3455 nla_nest_cancel(skb, br_afspec);
3456 goto nla_put_failure;
3459 nla_nest_end(skb, br_afspec);
3461 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3463 goto nla_put_failure;
3465 if (brport_nla_put_flag(skb, flags, mask,
3466 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3467 brport_nla_put_flag(skb, flags, mask,
3468 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3469 brport_nla_put_flag(skb, flags, mask,
3470 IFLA_BRPORT_FAST_LEAVE,
3471 BR_MULTICAST_FAST_LEAVE) ||
3472 brport_nla_put_flag(skb, flags, mask,
3473 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3474 brport_nla_put_flag(skb, flags, mask,
3475 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3476 brport_nla_put_flag(skb, flags, mask,
3477 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3478 brport_nla_put_flag(skb, flags, mask,
3479 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3480 brport_nla_put_flag(skb, flags, mask,
3481 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3482 nla_nest_cancel(skb, protinfo);
3483 goto nla_put_failure;
3486 nla_nest_end(skb, protinfo);
3488 nlmsg_end(skb, nlh);
3491 nlmsg_cancel(skb, nlh);
3492 return err ? err : -EMSGSIZE;
3494 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3496 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3498 struct net *net = sock_net(skb->sk);
3499 struct net_device *dev;
3501 u32 portid = NETLINK_CB(cb->skb).portid;
3502 u32 seq = cb->nlh->nlmsg_seq;
3503 u32 filter_mask = 0;
3506 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3507 struct nlattr *extfilt;
3509 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3512 if (nla_len(extfilt) < sizeof(filter_mask))
3515 filter_mask = nla_get_u32(extfilt);
3520 for_each_netdev_rcu(net, dev) {
3521 const struct net_device_ops *ops = dev->netdev_ops;
3522 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3524 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3525 if (idx >= cb->args[0]) {
3526 err = br_dev->netdev_ops->ndo_bridge_getlink(
3527 skb, portid, seq, dev,
3528 filter_mask, NLM_F_MULTI);
3529 if (err < 0 && err != -EOPNOTSUPP) {
3530 if (likely(skb->len))
3539 if (ops->ndo_bridge_getlink) {
3540 if (idx >= cb->args[0]) {
3541 err = ops->ndo_bridge_getlink(skb, portid,
3545 if (err < 0 && err != -EOPNOTSUPP) {
3546 if (likely(skb->len))
3563 static inline size_t bridge_nlmsg_size(void)
3565 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3566 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3567 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3568 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3569 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3570 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3571 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3572 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3573 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3574 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3575 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3578 static int rtnl_bridge_notify(struct net_device *dev)
3580 struct net *net = dev_net(dev);
3581 struct sk_buff *skb;
3582 int err = -EOPNOTSUPP;
3584 if (!dev->netdev_ops->ndo_bridge_getlink)
3587 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3593 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3600 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3603 WARN_ON(err == -EMSGSIZE);
3606 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3610 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3611 struct netlink_ext_ack *extack)
3613 struct net *net = sock_net(skb->sk);
3614 struct ifinfomsg *ifm;
3615 struct net_device *dev;
3616 struct nlattr *br_spec, *attr = NULL;
3617 int rem, err = -EOPNOTSUPP;
3619 bool have_flags = false;
3621 if (nlmsg_len(nlh) < sizeof(*ifm))
3624 ifm = nlmsg_data(nlh);
3625 if (ifm->ifi_family != AF_BRIDGE)
3626 return -EPFNOSUPPORT;
3628 dev = __dev_get_by_index(net, ifm->ifi_index);
3630 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3634 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3636 nla_for_each_nested(attr, br_spec, rem) {
3637 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3638 if (nla_len(attr) < sizeof(flags))
3642 flags = nla_get_u16(attr);
3648 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3649 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3651 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3656 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3660 flags &= ~BRIDGE_FLAGS_MASTER;
3663 if ((flags & BRIDGE_FLAGS_SELF)) {
3664 if (!dev->netdev_ops->ndo_bridge_setlink)
3667 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3670 flags &= ~BRIDGE_FLAGS_SELF;
3672 /* Generate event to notify upper layer of bridge
3675 err = rtnl_bridge_notify(dev);
3680 memcpy(nla_data(attr), &flags, sizeof(flags));
3685 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3686 struct netlink_ext_ack *extack)
3688 struct net *net = sock_net(skb->sk);
3689 struct ifinfomsg *ifm;
3690 struct net_device *dev;
3691 struct nlattr *br_spec, *attr = NULL;
3692 int rem, err = -EOPNOTSUPP;
3694 bool have_flags = false;
3696 if (nlmsg_len(nlh) < sizeof(*ifm))
3699 ifm = nlmsg_data(nlh);
3700 if (ifm->ifi_family != AF_BRIDGE)
3701 return -EPFNOSUPPORT;
3703 dev = __dev_get_by_index(net, ifm->ifi_index);
3705 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3709 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3711 nla_for_each_nested(attr, br_spec, rem) {
3712 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3713 if (nla_len(attr) < sizeof(flags))
3717 flags = nla_get_u16(attr);
3723 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3724 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3726 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3731 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3735 flags &= ~BRIDGE_FLAGS_MASTER;
3738 if ((flags & BRIDGE_FLAGS_SELF)) {
3739 if (!dev->netdev_ops->ndo_bridge_dellink)
3742 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3746 flags &= ~BRIDGE_FLAGS_SELF;
3748 /* Generate event to notify upper layer of bridge
3751 err = rtnl_bridge_notify(dev);
3756 memcpy(nla_data(attr), &flags, sizeof(flags));
3761 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3763 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3764 (!idxattr || idxattr == attrid);
3767 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3768 static int rtnl_get_offload_stats_attr_size(int attr_id)
3771 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3772 return sizeof(struct rtnl_link_stats64);
3778 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3781 struct nlattr *attr = NULL;
3786 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3787 dev->netdev_ops->ndo_get_offload_stats))
3790 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3791 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3792 if (attr_id < *prividx)
3795 size = rtnl_get_offload_stats_attr_size(attr_id);
3799 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3802 attr = nla_reserve_64bit(skb, attr_id, size,
3803 IFLA_OFFLOAD_XSTATS_UNSPEC);
3805 goto nla_put_failure;
3807 attr_data = nla_data(attr);
3808 memset(attr_data, 0, size);
3809 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3812 goto get_offload_stats_failure;
3823 get_offload_stats_failure:
3828 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3834 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3835 dev->netdev_ops->ndo_get_offload_stats))
3838 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3839 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3840 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3842 size = rtnl_get_offload_stats_attr_size(attr_id);
3843 nla_size += nla_total_size_64bit(size);
3847 nla_size += nla_total_size(0);
3852 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3853 int type, u32 pid, u32 seq, u32 change,
3854 unsigned int flags, unsigned int filter_mask,
3855 int *idxattr, int *prividx)
3857 struct if_stats_msg *ifsm;
3858 struct nlmsghdr *nlh;
3859 struct nlattr *attr;
3860 int s_prividx = *prividx;
3865 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3869 ifsm = nlmsg_data(nlh);
3870 ifsm->ifindex = dev->ifindex;
3871 ifsm->filter_mask = filter_mask;
3873 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3874 struct rtnl_link_stats64 *sp;
3876 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3877 sizeof(struct rtnl_link_stats64),
3880 goto nla_put_failure;
3882 sp = nla_data(attr);
3883 dev_get_stats(dev, sp);
3886 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3887 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3889 if (ops && ops->fill_linkxstats) {
3890 *idxattr = IFLA_STATS_LINK_XSTATS;
3891 attr = nla_nest_start(skb,
3892 IFLA_STATS_LINK_XSTATS);
3894 goto nla_put_failure;
3896 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3897 nla_nest_end(skb, attr);
3899 goto nla_put_failure;
3904 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3906 const struct rtnl_link_ops *ops = NULL;
3907 const struct net_device *master;
3909 master = netdev_master_upper_dev_get(dev);
3911 ops = master->rtnl_link_ops;
3912 if (ops && ops->fill_linkxstats) {
3913 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3914 attr = nla_nest_start(skb,
3915 IFLA_STATS_LINK_XSTATS_SLAVE);
3917 goto nla_put_failure;
3919 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3920 nla_nest_end(skb, attr);
3922 goto nla_put_failure;
3927 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3929 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3930 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3932 goto nla_put_failure;
3934 err = rtnl_get_offload_stats(skb, dev, prividx);
3935 if (err == -ENODATA)
3936 nla_nest_cancel(skb, attr);
3938 nla_nest_end(skb, attr);
3940 if (err && err != -ENODATA)
3941 goto nla_put_failure;
3945 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3946 struct rtnl_af_ops *af_ops;
3948 *idxattr = IFLA_STATS_AF_SPEC;
3949 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3951 goto nla_put_failure;
3953 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3954 if (af_ops->fill_stats_af) {
3958 af = nla_nest_start(skb, af_ops->family);
3960 goto nla_put_failure;
3962 err = af_ops->fill_stats_af(skb, dev);
3964 if (err == -ENODATA)
3965 nla_nest_cancel(skb, af);
3967 goto nla_put_failure;
3969 nla_nest_end(skb, af);
3973 nla_nest_end(skb, attr);
3978 nlmsg_end(skb, nlh);
3983 /* not a multi message or no progress mean a real error */
3984 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3985 nlmsg_cancel(skb, nlh);
3987 nlmsg_end(skb, nlh);
3992 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3997 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3998 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4000 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4001 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4002 int attr = IFLA_STATS_LINK_XSTATS;
4004 if (ops && ops->get_linkxstats_size) {
4005 size += nla_total_size(ops->get_linkxstats_size(dev,
4007 /* for IFLA_STATS_LINK_XSTATS */
4008 size += nla_total_size(0);
4012 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4013 struct net_device *_dev = (struct net_device *)dev;
4014 const struct rtnl_link_ops *ops = NULL;
4015 const struct net_device *master;
4017 /* netdev_master_upper_dev_get can't take const */
4018 master = netdev_master_upper_dev_get(_dev);
4020 ops = master->rtnl_link_ops;
4021 if (ops && ops->get_linkxstats_size) {
4022 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4024 size += nla_total_size(ops->get_linkxstats_size(dev,
4026 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4027 size += nla_total_size(0);
4031 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4032 size += rtnl_get_offload_stats_size(dev);
4034 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4035 struct rtnl_af_ops *af_ops;
4037 /* for IFLA_STATS_AF_SPEC */
4038 size += nla_total_size(0);
4040 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4041 if (af_ops->get_stats_af_size) {
4042 size += nla_total_size(
4043 af_ops->get_stats_af_size(dev));
4046 size += nla_total_size(0);
4054 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4055 struct netlink_ext_ack *extack)
4057 struct net *net = sock_net(skb->sk);
4058 struct net_device *dev = NULL;
4059 int idxattr = 0, prividx = 0;
4060 struct if_stats_msg *ifsm;
4061 struct sk_buff *nskb;
4065 if (nlmsg_len(nlh) < sizeof(*ifsm))
4068 ifsm = nlmsg_data(nlh);
4069 if (ifsm->ifindex > 0)
4070 dev = __dev_get_by_index(net, ifsm->ifindex);
4077 filter_mask = ifsm->filter_mask;
4081 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4085 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4086 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4087 0, filter_mask, &idxattr, &prividx);
4089 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4090 WARN_ON(err == -EMSGSIZE);
4093 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4099 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4101 int h, s_h, err, s_idx, s_idxattr, s_prividx;
4102 struct net *net = sock_net(skb->sk);
4103 unsigned int flags = NLM_F_MULTI;
4104 struct if_stats_msg *ifsm;
4105 struct hlist_head *head;
4106 struct net_device *dev;
4107 u32 filter_mask = 0;
4111 s_idx = cb->args[1];
4112 s_idxattr = cb->args[2];
4113 s_prividx = cb->args[3];
4115 cb->seq = net->dev_base_seq;
4117 if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4120 ifsm = nlmsg_data(cb->nlh);
4121 filter_mask = ifsm->filter_mask;
4125 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4127 head = &net->dev_index_head[h];
4128 hlist_for_each_entry(dev, head, index_hlist) {
4131 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4132 NETLINK_CB(cb->skb).portid,
4133 cb->nlh->nlmsg_seq, 0,
4135 &s_idxattr, &s_prividx);
4136 /* If we ran out of room on the first message,
4139 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4145 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4151 cb->args[3] = s_prividx;
4152 cb->args[2] = s_idxattr;
4159 /* Process one rtnetlink message. */
4161 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4162 struct netlink_ext_ack *extack)
4164 struct net *net = sock_net(skb->sk);
4165 rtnl_doit_func doit;
4171 type = nlh->nlmsg_type;
4177 /* All the messages must have at least 1 byte length */
4178 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4181 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4184 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4187 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4189 rtnl_dumpit_func dumpit;
4190 rtnl_calcit_func calcit;
4191 u16 min_dump_alloc = 0;
4193 dumpit = rtnl_get_dumpit(family, type);
4196 calcit = rtnl_get_calcit(family, type);
4198 min_dump_alloc = calcit(skb, nlh);
4203 struct netlink_dump_control c = {
4205 .min_dump_alloc = min_dump_alloc,
4207 err = netlink_dump_start(rtnl, skb, nlh, &c);
4213 doit = rtnl_get_doit(family, type);
4217 return doit(skb, nlh, extack);
4220 static void rtnetlink_rcv(struct sk_buff *skb)
4223 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4227 static int rtnetlink_bind(struct net *net, int group)
4230 case RTNLGRP_IPV4_MROUTE_R:
4231 case RTNLGRP_IPV6_MROUTE_R:
4232 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4239 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4241 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4245 case NETDEV_CHANGEADDR:
4246 case NETDEV_CHANGENAME:
4247 case NETDEV_FEAT_CHANGE:
4248 case NETDEV_BONDING_FAILOVER:
4249 case NETDEV_NOTIFY_PEERS:
4250 case NETDEV_RESEND_IGMP:
4251 case NETDEV_CHANGEINFODATA:
4252 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4261 static struct notifier_block rtnetlink_dev_notifier = {
4262 .notifier_call = rtnetlink_event,
4266 static int __net_init rtnetlink_net_init(struct net *net)
4269 struct netlink_kernel_cfg cfg = {
4270 .groups = RTNLGRP_MAX,
4271 .input = rtnetlink_rcv,
4272 .cb_mutex = &rtnl_mutex,
4273 .flags = NL_CFG_F_NONROOT_RECV,
4274 .bind = rtnetlink_bind,
4277 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4284 static void __net_exit rtnetlink_net_exit(struct net *net)
4286 netlink_kernel_release(net->rtnl);
4290 static struct pernet_operations rtnetlink_net_ops = {
4291 .init = rtnetlink_net_init,
4292 .exit = rtnetlink_net_exit,
4295 void __init rtnetlink_init(void)
4297 if (register_pernet_subsys(&rtnetlink_net_ops))
4298 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4300 register_netdevice_notifier(&rtnetlink_dev_notifier);
4302 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4303 rtnl_dump_ifinfo, rtnl_calcit);
4304 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4305 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4306 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4308 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4309 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4310 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4312 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4313 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4314 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4316 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4317 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4318 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4320 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,